Index: projects/vnet/sys/contrib/ipfilter/netinet/mlfk_ipl.c =================================================================== --- projects/vnet/sys/contrib/ipfilter/netinet/mlfk_ipl.c (revision 294565) +++ projects/vnet/sys/contrib/ipfilter/netinet/mlfk_ipl.c (revision 294566) @@ -1,568 +1,568 @@ /* $FreeBSD$ */ /* * Copyright (C) 2012 by Darren Reed. * * $FreeBSD$ * See the IPFILTER.LICENCE file for details on licencing. */ #include #include #include #include #include #include #include #include #if __FreeBSD_version >= 500000 # include #endif #include #include #include #include "netinet/ipl.h" #include "netinet/ip_compat.h" #include "netinet/ip_fil.h" #include "netinet/ip_state.h" #include "netinet/ip_nat.h" #include "netinet/ip_auth.h" #include "netinet/ip_frag.h" #include "netinet/ip_sync.h" extern ipf_main_softc_t ipfmain; #if __FreeBSD_version >= 502116 static struct cdev *ipf_devs[IPL_LOGSIZE]; #else static dev_t ipf_devs[IPL_LOGSIZE]; #endif static int sysctl_ipf_int ( SYSCTL_HANDLER_ARGS ); static int ipf_modload(void); static int ipf_modunload(void); static int ipf_fbsd_sysctl_create(ipf_main_softc_t*); static int ipf_fbsd_sysctl_destroy(ipf_main_softc_t*); #if (__FreeBSD_version >= 500024) # if (__FreeBSD_version >= 502116) static int ipfopen __P((struct cdev*, int, int, struct thread *)); static int ipfclose __P((struct cdev*, int, int, struct thread *)); # else static int ipfopen __P((dev_t, int, int, struct thread *)); static int ipfclose __P((dev_t, int, int, struct thread *)); # endif /* __FreeBSD_version >= 502116 */ #else static int ipfopen __P((dev_t, int, int, struct proc *)); static int ipfclose __P((dev_t, int, int, struct proc *)); #endif #if (__FreeBSD_version >= 502116) static int ipfread __P((struct cdev*, struct uio *, int)); static int ipfwrite __P((struct cdev*, struct uio *, int)); #else static int ipfread __P((dev_t, struct uio *, int)); static int ipfwrite __P((dev_t, struct uio *, int)); #endif /* __FreeBSD_version >= 502116 */ SYSCTL_DECL(_net_inet); #define SYSCTL_IPF(parent, nbr, name, access, ptr, val, descr) \ SYSCTL_OID(parent, nbr, name, CTLTYPE_INT|access, \ ptr, val, sysctl_ipf_int, "I", descr); #define SYSCTL_DYN_IPF(parent, nbr, name, access,ptr, val, descr) \ SYSCTL_ADD_OID(&ipf_clist, SYSCTL_STATIC_CHILDREN(parent), nbr, name, \ CTLFLAG_DYN|CTLTYPE_INT|access, ptr, val, sysctl_ipf_int, "I", descr) static struct sysctl_ctx_list ipf_clist; #define CTLFLAG_OFF 0x00800000 /* IPFilter must be disabled */ #define CTLFLAG_RWO (CTLFLAG_RW|CTLFLAG_OFF) SYSCTL_NODE(_net_inet, OID_AUTO, ipf, CTLFLAG_RW, 0, "IPF"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_flags, CTLFLAG_RW, &ipfmain.ipf_flags, 0, "IPF flags"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, ipf_pass, CTLFLAG_RW, &ipfmain.ipf_pass, 0, "default pass/block"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_active, CTLFLAG_RD, &ipfmain.ipf_active, 0, "IPF is active"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcpidletimeout, CTLFLAG_RWO, &ipfmain.ipf_tcpidletimeout, 0, "TCP idle timeout in seconds"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcphalfclosed, CTLFLAG_RWO, &ipfmain.ipf_tcphalfclosed, 0, "timeout for half closed TCP sessions"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcpclosewait, CTLFLAG_RWO, &ipfmain.ipf_tcpclosewait, 0, "timeout for TCP sessions in closewait status"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcplastack, CTLFLAG_RWO, &ipfmain.ipf_tcplastack, 0, "timeout for TCP sessions in last ack status"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcptimeout, CTLFLAG_RWO, &ipfmain.ipf_tcptimeout, 0, ""); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_tcpclosed, CTLFLAG_RWO, &ipfmain.ipf_tcpclosed, 0, ""); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_udptimeout, CTLFLAG_RWO, &ipfmain.ipf_udptimeout, 0, "UDP timeout"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_udpacktimeout, CTLFLAG_RWO, &ipfmain.ipf_udpacktimeout, 0, ""); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_icmptimeout, CTLFLAG_RWO, &ipfmain.ipf_icmptimeout, 0, "ICMP timeout"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_running, CTLFLAG_RD, &ipfmain.ipf_running, 0, "IPF is running"); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_chksrc, CTLFLAG_RW, &ipfmain.ipf_chksrc, 0, ""); SYSCTL_IPF(_net_inet_ipf, OID_AUTO, fr_minttl, CTLFLAG_RW, &ipfmain.ipf_minttl, 0, ""); #define CDEV_MAJOR 79 #include #if __FreeBSD_version >= 500043 # include static int ipfpoll(struct cdev *dev, int events, struct thread *td); static struct cdevsw ipf_cdevsw = { #if __FreeBSD_version >= 502103 .d_version = D_VERSION, .d_flags = 0, /* D_NEEDGIANT - Should be SMP safe */ #endif .d_open = ipfopen, .d_close = ipfclose, .d_read = ipfread, .d_write = ipfwrite, .d_ioctl = ipfioctl, .d_poll = ipfpoll, .d_name = "ipf", #if __FreeBSD_version < 600000 .d_maj = CDEV_MAJOR, #endif }; #else static int ipfpoll(dev_t dev, int events, struct proc *td); static struct cdevsw ipf_cdevsw = { /* open */ ipfopen, /* close */ ipfclose, /* read */ ipfread, /* write */ ipfwrite, /* ioctl */ ipfioctl, /* poll */ ipfpoll, /* mmap */ nommap, /* strategy */ nostrategy, /* name */ "ipf", /* maj */ CDEV_MAJOR, /* dump */ nodump, /* psize */ nopsize, /* flags */ 0, # if (__FreeBSD_version < 500043) /* bmaj */ -1, # endif # if (__FreeBSD_version >= 430000) /* kqfilter */ NULL # endif }; #endif static char *ipf_devfiles[] = { IPL_NAME, IPNAT_NAME, IPSTATE_NAME, IPAUTH_NAME, IPSYNC_NAME, IPSCAN_NAME, IPLOOKUP_NAME, NULL }; static int ipfilter_modevent(module_t mod, int type, void *unused) { int error = 0; switch (type) { case MOD_LOAD : error = ipf_modload(); break; case MOD_UNLOAD : error = ipf_modunload(); break; default: error = EINVAL; break; } return error; } static int ipf_modload() { char *defpass, *c, *str; int i, j, error; if (ipf_load_all() != 0) return EIO; if (ipf_create_all(&ipfmain) == NULL) return EIO; if (ipf_fbsd_sysctl_create(&ipfmain) != 0) return EIO; error = ipfattach(&ipfmain); if (error) return error; for (i = 0; i < IPL_LOGSIZE; i++) ipf_devs[i] = NULL; for (i = 0; (str = ipf_devfiles[i]); i++) { c = NULL; for(j = strlen(str); j > 0; j--) if (str[j] == '/') { c = str + j + 1; break; } if (!c) c = str; ipf_devs[i] = make_dev(&ipf_cdevsw, i, 0, 0, 0600, "%s", c); } error = ipf_pfil_hook(); if (error != 0) return error; ipf_event_reg(); if (FR_ISPASS(ipfmain.ipf_pass)) defpass = "pass"; else if (FR_ISBLOCK(ipfmain.ipf_pass)) defpass = "block"; else defpass = "no-match -> block"; printf("%s initialized. Default = %s all, Logging = %s%s\n", ipfilter_version, defpass, #ifdef IPFILTER_LOG "enabled", #else "disabled", #endif #ifdef IPFILTER_COMPILED " (COMPILED)" #else "" #endif ); return 0; } static int ipf_modunload() { int error, i; if (ipfmain.ipf_refcnt) return EBUSY; if (ipf_fbsd_sysctl_destroy(&ipfmain) != 0) return EIO; error = ipf_pfil_unhook(); if (error != 0) return error; if (ipfmain.ipf_running >= 0) { error = ipfdetach(&ipfmain); if (error != 0) return error; ipf_fbsd_sysctl_destroy(&ipfmain); ipf_destroy_all(&ipfmain); ipf_unload_all(); } else error = 0; ipfmain.ipf_running = -2; for (i = 0; ipf_devfiles[i]; i++) { if (ipf_devs[i] != NULL) destroy_dev(ipf_devs[i]); } printf("%s unloaded\n", ipfilter_version); return error; } static moduledata_t ipfiltermod = { "ipfilter", ipfilter_modevent, 0 }; -DECLARE_MODULE(ipfilter, ipfiltermod, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY); +DECLARE_MODULE(ipfilter, ipfiltermod, SI_SUB_FW, SI_ORDER_ANY); #ifdef MODULE_VERSION MODULE_VERSION(ipfilter, 1); #endif #ifdef SYSCTL_IPF int sysctl_ipf_int ( SYSCTL_HANDLER_ARGS ) { int error = 0; if (arg1) error = SYSCTL_OUT(req, arg1, sizeof(int)); else error = SYSCTL_OUT(req, &arg2, sizeof(int)); if (error || !req->newptr) return (error); if (!arg1) error = EPERM; else { if ((oidp->oid_kind & CTLFLAG_OFF) && (ipfmain.ipf_running > 0)) error = EBUSY; else error = SYSCTL_IN(req, arg1, sizeof(int)); } return (error); } #endif static int #if __FreeBSD_version >= 500043 ipfpoll(struct cdev *dev, int events, struct thread *td) #else ipfpoll(dev_t dev, int events, struct proc *td) #endif { int unit = GET_MINOR(dev); int revents; if (unit < 0 || unit > IPL_LOGMAX) return 0; revents = 0; switch (unit) { case IPL_LOGIPF : case IPL_LOGNAT : case IPL_LOGSTATE : #ifdef IPFILTER_LOG if ((events & (POLLIN | POLLRDNORM)) && ipf_log_canread(&ipfmain, unit)) revents |= events & (POLLIN | POLLRDNORM); #endif break; case IPL_LOGAUTH : if ((events & (POLLIN | POLLRDNORM)) && ipf_auth_waiting(&ipfmain)) revents |= events & (POLLIN | POLLRDNORM); break; case IPL_LOGSYNC : if ((events & (POLLIN | POLLRDNORM)) && ipf_sync_canread(&ipfmain)) revents |= events & (POLLIN | POLLRDNORM); if ((events & (POLLOUT | POLLWRNORM)) && ipf_sync_canwrite(&ipfmain)) revents |= events & (POLLOUT | POLLWRNORM); break; case IPL_LOGSCAN : case IPL_LOGLOOKUP : default : break; } if ((revents == 0) && ((events & (POLLIN|POLLRDNORM)) != 0)) selrecord(td, &ipfmain.ipf_selwait[unit]); return revents; } /* * routines below for saving IP headers to buffer */ static int ipfopen(dev, flags #if ((BSD >= 199506) || (__FreeBSD_version >= 220000)) , devtype, p) int devtype; # if (__FreeBSD_version >= 500024) struct thread *p; # else struct proc *p; # endif /* __FreeBSD_version >= 500024 */ #else ) #endif #if (__FreeBSD_version >= 502116) struct cdev *dev; #else dev_t dev; #endif int flags; { int unit = GET_MINOR(dev); int error; if (IPL_LOGMAX < unit) error = ENXIO; else { switch (unit) { case IPL_LOGIPF : case IPL_LOGNAT : case IPL_LOGSTATE : case IPL_LOGAUTH : case IPL_LOGLOOKUP : case IPL_LOGSYNC : #ifdef IPFILTER_SCAN case IPL_LOGSCAN : #endif error = 0; break; default : error = ENXIO; break; } } return error; } static int ipfclose(dev, flags #if ((BSD >= 199506) || (__FreeBSD_version >= 220000)) , devtype, p) int devtype; # if (__FreeBSD_version >= 500024) struct thread *p; # else struct proc *p; # endif /* __FreeBSD_version >= 500024 */ #else ) #endif #if (__FreeBSD_version >= 502116) struct cdev *dev; #else dev_t dev; #endif int flags; { int unit = GET_MINOR(dev); if (IPL_LOGMAX < unit) unit = ENXIO; else unit = 0; return unit; } /* * ipfread/ipflog * both of these must operate with at least splnet() lest they be * called during packet processing and cause an inconsistancy to appear in * the filter lists. */ #if (BSD >= 199306) static int ipfread(dev, uio, ioflag) int ioflag; #else static int ipfread(dev, uio) #endif #if (__FreeBSD_version >= 502116) struct cdev *dev; #else dev_t dev; #endif struct uio *uio; { int unit = GET_MINOR(dev); if (unit < 0) return ENXIO; if (ipfmain.ipf_running < 1) return EIO; if (unit == IPL_LOGSYNC) return ipf_sync_read(&ipfmain, uio); #ifdef IPFILTER_LOG return ipf_log_read(&ipfmain, unit, uio); #else return ENXIO; #endif } /* * ipfwrite * both of these must operate with at least splnet() lest they be * called during packet processing and cause an inconsistancy to appear in * the filter lists. */ #if (BSD >= 199306) static int ipfwrite(dev, uio, ioflag) int ioflag; #else static int ipfwrite(dev, uio) #endif #if (__FreeBSD_version >= 502116) struct cdev *dev; #else dev_t dev; #endif struct uio *uio; { if (ipfmain.ipf_running < 1) return EIO; if (GET_MINOR(dev) == IPL_LOGSYNC) return ipf_sync_write(&ipfmain, uio); return ENXIO; } static int ipf_fbsd_sysctl_create(main_softc) ipf_main_softc_t *main_softc; { ipf_nat_softc_t *nat_softc; ipf_state_softc_t *state_softc; ipf_auth_softc_t *auth_softc; ipf_frag_softc_t *frag_softc; nat_softc = main_softc->ipf_nat_soft; state_softc = main_softc->ipf_state_soft; auth_softc = main_softc->ipf_auth_soft; frag_softc = main_softc->ipf_frag_soft; sysctl_ctx_init(&ipf_clist); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "fr_defnatage", CTLFLAG_RWO, &nat_softc->ipf_nat_defage, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "fr_statesize", CTLFLAG_RWO, &state_softc->ipf_state_size, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "fr_statemax", CTLFLAG_RWO, &state_softc->ipf_state_max, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "ipf_nattable_max", CTLFLAG_RWO, &nat_softc->ipf_nat_table_max, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "ipf_nattable_sz", CTLFLAG_RWO, &nat_softc->ipf_nat_table_sz, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "ipf_natrules_sz", CTLFLAG_RWO, &nat_softc->ipf_nat_maprules_sz, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "ipf_rdrrules_sz", CTLFLAG_RWO, &nat_softc->ipf_nat_rdrrules_sz, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "ipf_hostmap_sz", CTLFLAG_RWO, &nat_softc->ipf_nat_hostmap_sz, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "fr_authsize", CTLFLAG_RWO, &auth_softc->ipf_auth_size, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "fr_authused", CTLFLAG_RD, &auth_softc->ipf_auth_used, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "fr_defaultauthage", CTLFLAG_RW, &auth_softc->ipf_auth_defaultage, 0, ""); SYSCTL_DYN_IPF(_net_inet_ipf, OID_AUTO, "fr_ipfrttl", CTLFLAG_RW, &frag_softc->ipfr_ttl, 0, ""); return 0; } static int ipf_fbsd_sysctl_destroy(main_softc) ipf_main_softc_t *main_softc; { if (sysctl_ctx_free(&ipf_clist)) { printf("sysctl_ctx_free failed"); return(ENOTEMPTY); } return 0; } Index: projects/vnet/sys/dev/usb/net/usb_ethernet.c =================================================================== --- projects/vnet/sys/dev/usb/net/usb_ethernet.c (revision 294565) +++ projects/vnet/sys/dev/usb/net/usb_ethernet.c (revision 294566) @@ -1,645 +1,649 @@ /* $FreeBSD$ */ /*- * Copyright (c) 2009 Andrew Thompson (thompsa@FreeBSD.org) * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static SYSCTL_NODE(_net, OID_AUTO, ue, CTLFLAG_RD, 0, "USB Ethernet parameters"); #define UE_LOCK(_ue) mtx_lock((_ue)->ue_mtx) #define UE_UNLOCK(_ue) mtx_unlock((_ue)->ue_mtx) #define UE_LOCK_ASSERT(_ue, t) mtx_assert((_ue)->ue_mtx, t) MODULE_DEPEND(uether, usb, 1, 1, 1); MODULE_DEPEND(uether, miibus, 1, 1, 1); static struct unrhdr *ueunit; static usb_proc_callback_t ue_attach_post_task; static usb_proc_callback_t ue_promisc_task; static usb_proc_callback_t ue_setmulti_task; static usb_proc_callback_t ue_ifmedia_task; static usb_proc_callback_t ue_tick_task; static usb_proc_callback_t ue_start_task; static usb_proc_callback_t ue_stop_task; static void ue_init(void *); static void ue_start(struct ifnet *); static int ue_ifmedia_upd(struct ifnet *); static void ue_watchdog(void *); /* * Return values: * 0: success * Else: device has been detached */ uint8_t uether_pause(struct usb_ether *ue, unsigned int _ticks) { if (usb_proc_is_gone(&ue->ue_tq)) { /* nothing to do */ return (1); } usb_pause_mtx(ue->ue_mtx, _ticks); return (0); } static void ue_queue_command(struct usb_ether *ue, usb_proc_callback_t *fn, struct usb_proc_msg *t0, struct usb_proc_msg *t1) { struct usb_ether_cfg_task *task; UE_LOCK_ASSERT(ue, MA_OWNED); if (usb_proc_is_gone(&ue->ue_tq)) { return; /* nothing to do */ } /* * NOTE: The task cannot get executed before we drop the * "sc_mtx" mutex. It is safe to update fields in the message * structure after that the message got queued. */ task = (struct usb_ether_cfg_task *) usb_proc_msignal(&ue->ue_tq, t0, t1); /* Setup callback and self pointers */ task->hdr.pm_callback = fn; task->ue = ue; /* * Start and stop must be synchronous! */ if ((fn == ue_start_task) || (fn == ue_stop_task)) usb_proc_mwait(&ue->ue_tq, t0, t1); } struct ifnet * uether_getifp(struct usb_ether *ue) { return (ue->ue_ifp); } struct mii_data * uether_getmii(struct usb_ether *ue) { return (device_get_softc(ue->ue_miibus)); } void * uether_getsc(struct usb_ether *ue) { return (ue->ue_sc); } static int ue_sysctl_parent(SYSCTL_HANDLER_ARGS) { struct usb_ether *ue = arg1; const char *name; name = device_get_nameunit(ue->ue_dev); return SYSCTL_OUT_STR(req, name); } int uether_ifattach(struct usb_ether *ue) { int error; /* check some critical parameters */ if ((ue->ue_dev == NULL) || (ue->ue_udev == NULL) || (ue->ue_mtx == NULL) || (ue->ue_methods == NULL)) return (EINVAL); error = usb_proc_create(&ue->ue_tq, ue->ue_mtx, device_get_nameunit(ue->ue_dev), USB_PRI_MED); if (error) { device_printf(ue->ue_dev, "could not setup taskqueue\n"); goto error; } /* fork rest of the attach code */ UE_LOCK(ue); ue_queue_command(ue, ue_attach_post_task, &ue->ue_sync_task[0].hdr, &ue->ue_sync_task[1].hdr); UE_UNLOCK(ue); error: return (error); } static void ue_attach_post_task(struct usb_proc_msg *_task) { struct usb_ether_cfg_task *task = (struct usb_ether_cfg_task *)_task; struct usb_ether *ue = task->ue; struct ifnet *ifp; int error; char num[14]; /* sufficient for 32 bits */ /* first call driver's post attach routine */ ue->ue_methods->ue_attach_post(ue); UE_UNLOCK(ue); ue->ue_unit = alloc_unr(ueunit); usb_callout_init_mtx(&ue->ue_watchdog, ue->ue_mtx, 0); sysctl_ctx_init(&ue->ue_sysctl_ctx); error = 0; CURVNET_SET_QUIET(vnet0); ifp = if_alloc(IFT_ETHER); if (ifp == NULL) { device_printf(ue->ue_dev, "could not allocate ifnet\n"); goto fail; } ifp->if_softc = ue; if_initname(ifp, "ue", ue->ue_unit); if (ue->ue_methods->ue_attach_post_sub != NULL) { ue->ue_ifp = ifp; error = ue->ue_methods->ue_attach_post_sub(ue); } else { ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; if (ue->ue_methods->ue_ioctl != NULL) ifp->if_ioctl = ue->ue_methods->ue_ioctl; else ifp->if_ioctl = uether_ioctl; ifp->if_start = ue_start; ifp->if_init = ue_init; IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); ifp->if_snd.ifq_drv_maxlen = ifqmaxlen; IFQ_SET_READY(&ifp->if_snd); ue->ue_ifp = ifp; if (ue->ue_methods->ue_mii_upd != NULL && ue->ue_methods->ue_mii_sts != NULL) { /* device_xxx() depends on this */ mtx_lock(&Giant); error = mii_attach(ue->ue_dev, &ue->ue_miibus, ifp, ue_ifmedia_upd, ue->ue_methods->ue_mii_sts, BMSR_DEFCAPMASK, MII_PHY_ANY, MII_OFFSET_ANY, 0); mtx_unlock(&Giant); } } if (error) { device_printf(ue->ue_dev, "attaching PHYs failed\n"); goto fail; } if_printf(ifp, " on %s\n", device_get_nameunit(ue->ue_dev)); ether_ifattach(ifp, ue->ue_eaddr); /* Tell upper layer we support VLAN oversized frames. */ if (ifp->if_capabilities & IFCAP_VLAN_MTU) ifp->if_hdrlen = sizeof(struct ether_vlan_header); CURVNET_RESTORE(); snprintf(num, sizeof(num), "%u", ue->ue_unit); ue->ue_sysctl_oid = SYSCTL_ADD_NODE(&ue->ue_sysctl_ctx, &SYSCTL_NODE_CHILDREN(_net, ue), OID_AUTO, num, CTLFLAG_RD, NULL, ""); SYSCTL_ADD_PROC(&ue->ue_sysctl_ctx, SYSCTL_CHILDREN(ue->ue_sysctl_oid), OID_AUTO, "%parent", CTLTYPE_STRING | CTLFLAG_RD, ue, 0, ue_sysctl_parent, "A", "parent device"); UE_LOCK(ue); return; fail: CURVNET_RESTORE(); free_unr(ueunit, ue->ue_unit); if (ue->ue_ifp != NULL) { if_free(ue->ue_ifp); ue->ue_ifp = NULL; } UE_LOCK(ue); return; } void uether_ifdetach(struct usb_ether *ue) { struct ifnet *ifp; /* wait for any post attach or other command to complete */ usb_proc_drain(&ue->ue_tq); /* read "ifnet" pointer after taskqueue drain */ ifp = ue->ue_ifp; if (ifp != NULL) { /* we are not running any more */ UE_LOCK(ue); ifp->if_drv_flags &= ~IFF_DRV_RUNNING; UE_UNLOCK(ue); /* drain any callouts */ usb_callout_drain(&ue->ue_watchdog); /* detach miibus */ if (ue->ue_miibus != NULL) { mtx_lock(&Giant); /* device_xxx() depends on this */ device_delete_child(ue->ue_dev, ue->ue_miibus); mtx_unlock(&Giant); } /* detach ethernet */ ether_ifdetach(ifp); /* free interface instance */ if_free(ifp); /* free sysctl */ sysctl_ctx_free(&ue->ue_sysctl_ctx); /* free unit */ free_unr(ueunit, ue->ue_unit); } /* free taskqueue, if any */ usb_proc_free(&ue->ue_tq); } uint8_t uether_is_gone(struct usb_ether *ue) { return (usb_proc_is_gone(&ue->ue_tq)); } void uether_init(void *arg) { ue_init(arg); } static void ue_init(void *arg) { struct usb_ether *ue = arg; UE_LOCK(ue); ue_queue_command(ue, ue_start_task, &ue->ue_sync_task[0].hdr, &ue->ue_sync_task[1].hdr); UE_UNLOCK(ue); } static void ue_start_task(struct usb_proc_msg *_task) { struct usb_ether_cfg_task *task = (struct usb_ether_cfg_task *)_task; struct usb_ether *ue = task->ue; struct ifnet *ifp = ue->ue_ifp; UE_LOCK_ASSERT(ue, MA_OWNED); ue->ue_methods->ue_init(ue); if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) return; if (ue->ue_methods->ue_tick != NULL) usb_callout_reset(&ue->ue_watchdog, hz, ue_watchdog, ue); } static void ue_stop_task(struct usb_proc_msg *_task) { struct usb_ether_cfg_task *task = (struct usb_ether_cfg_task *)_task; struct usb_ether *ue = task->ue; UE_LOCK_ASSERT(ue, MA_OWNED); usb_callout_stop(&ue->ue_watchdog); ue->ue_methods->ue_stop(ue); } void uether_start(struct ifnet *ifp) { ue_start(ifp); } static void ue_start(struct ifnet *ifp) { struct usb_ether *ue = ifp->if_softc; if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) return; UE_LOCK(ue); ue->ue_methods->ue_start(ue); UE_UNLOCK(ue); } static void ue_promisc_task(struct usb_proc_msg *_task) { struct usb_ether_cfg_task *task = (struct usb_ether_cfg_task *)_task; struct usb_ether *ue = task->ue; ue->ue_methods->ue_setpromisc(ue); } static void ue_setmulti_task(struct usb_proc_msg *_task) { struct usb_ether_cfg_task *task = (struct usb_ether_cfg_task *)_task; struct usb_ether *ue = task->ue; ue->ue_methods->ue_setmulti(ue); } int uether_ifmedia_upd(struct ifnet *ifp) { return (ue_ifmedia_upd(ifp)); } static int ue_ifmedia_upd(struct ifnet *ifp) { struct usb_ether *ue = ifp->if_softc; /* Defer to process context */ UE_LOCK(ue); ue_queue_command(ue, ue_ifmedia_task, &ue->ue_media_task[0].hdr, &ue->ue_media_task[1].hdr); UE_UNLOCK(ue); return (0); } static void ue_ifmedia_task(struct usb_proc_msg *_task) { struct usb_ether_cfg_task *task = (struct usb_ether_cfg_task *)_task; struct usb_ether *ue = task->ue; struct ifnet *ifp = ue->ue_ifp; ue->ue_methods->ue_mii_upd(ifp); } static void ue_watchdog(void *arg) { struct usb_ether *ue = arg; struct ifnet *ifp = ue->ue_ifp; if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) return; ue_queue_command(ue, ue_tick_task, &ue->ue_tick_task[0].hdr, &ue->ue_tick_task[1].hdr); usb_callout_reset(&ue->ue_watchdog, hz, ue_watchdog, ue); } static void ue_tick_task(struct usb_proc_msg *_task) { struct usb_ether_cfg_task *task = (struct usb_ether_cfg_task *)_task; struct usb_ether *ue = task->ue; struct ifnet *ifp = ue->ue_ifp; if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) return; ue->ue_methods->ue_tick(ue); } int uether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) { struct usb_ether *ue = ifp->if_softc; struct ifreq *ifr = (struct ifreq *)data; struct mii_data *mii; int error = 0; switch (command) { case SIOCSIFFLAGS: UE_LOCK(ue); if (ifp->if_flags & IFF_UP) { if (ifp->if_drv_flags & IFF_DRV_RUNNING) ue_queue_command(ue, ue_promisc_task, &ue->ue_promisc_task[0].hdr, &ue->ue_promisc_task[1].hdr); else ue_queue_command(ue, ue_start_task, &ue->ue_sync_task[0].hdr, &ue->ue_sync_task[1].hdr); } else { ue_queue_command(ue, ue_stop_task, &ue->ue_sync_task[0].hdr, &ue->ue_sync_task[1].hdr); } UE_UNLOCK(ue); break; case SIOCADDMULTI: case SIOCDELMULTI: UE_LOCK(ue); ue_queue_command(ue, ue_setmulti_task, &ue->ue_multi_task[0].hdr, &ue->ue_multi_task[1].hdr); UE_UNLOCK(ue); break; case SIOCGIFMEDIA: case SIOCSIFMEDIA: if (ue->ue_miibus != NULL) { mii = device_get_softc(ue->ue_miibus); error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, command); } else error = ether_ioctl(ifp, command, data); break; default: error = ether_ioctl(ifp, command, data); break; } return (error); } static int uether_modevent(module_t mod, int type, void *data) { switch (type) { case MOD_LOAD: ueunit = new_unrhdr(0, INT_MAX, NULL); break; case MOD_UNLOAD: break; default: return (EOPNOTSUPP); } return (0); } static moduledata_t uether_mod = { "uether", uether_modevent, 0 }; struct mbuf * uether_newbuf(void) { struct mbuf *m_new; m_new = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (m_new == NULL) return (NULL); m_new->m_len = m_new->m_pkthdr.len = MCLBYTES; m_adj(m_new, ETHER_ALIGN); return (m_new); } int uether_rxmbuf(struct usb_ether *ue, struct mbuf *m, unsigned int len) { struct ifnet *ifp = ue->ue_ifp; UE_LOCK_ASSERT(ue, MA_OWNED); /* finalize mbuf */ if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1); m->m_pkthdr.rcvif = ifp; m->m_pkthdr.len = m->m_len = len; /* enqueue for later when the lock can be released */ _IF_ENQUEUE(&ue->ue_rxq, m); return (0); } int uether_rxbuf(struct usb_ether *ue, struct usb_page_cache *pc, unsigned int offset, unsigned int len) { struct ifnet *ifp = ue->ue_ifp; struct mbuf *m; UE_LOCK_ASSERT(ue, MA_OWNED); if (len < ETHER_HDR_LEN || len > MCLBYTES - ETHER_ALIGN) return (1); m = uether_newbuf(); if (m == NULL) { if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1); return (ENOMEM); } usbd_copy_out(pc, offset, mtod(m, uint8_t *), len); /* finalize mbuf */ if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1); m->m_pkthdr.rcvif = ifp; m->m_pkthdr.len = m->m_len = len; /* enqueue for later when the lock can be released */ _IF_ENQUEUE(&ue->ue_rxq, m); return (0); } void uether_rxflush(struct usb_ether *ue) { struct ifnet *ifp = ue->ue_ifp; struct mbuf *m; UE_LOCK_ASSERT(ue, MA_OWNED); for (;;) { _IF_DEQUEUE(&ue->ue_rxq, m); if (m == NULL) break; /* * The USB xfer has been resubmitted so its safe to unlock now. */ UE_UNLOCK(ue); ifp->if_input(ifp, m); UE_LOCK(ue); } } -DECLARE_MODULE(uether, uether_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); +/* + * USB net drivers are run by DRIVER_MODULE() thus SI_SUB_DRIVERS, + * SI_ORDER_MIDDLE. Run uether after that. + */ +DECLARE_MODULE(uether, uether_mod, SI_SUB_DRIVERS, SI_ORDER_ANY); MODULE_VERSION(uether, 1); Index: projects/vnet/sys/kern/kern_shutdown.c =================================================================== --- projects/vnet/sys/kern/kern_shutdown.c (revision 294565) +++ projects/vnet/sys/kern/kern_shutdown.c (revision 294566) @@ -1,901 +1,912 @@ /*- * Copyright (c) 1986, 1988, 1991, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)kern_shutdown.c 8.3 (Berkeley) 1/21/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include "opt_kdb.h" #include "opt_panic.h" #include "opt_sched.h" #include "opt_watchdog.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifndef PANIC_REBOOT_WAIT_TIME #define PANIC_REBOOT_WAIT_TIME 15 /* default to 15 seconds */ #endif static int panic_reboot_wait_time = PANIC_REBOOT_WAIT_TIME; SYSCTL_INT(_kern, OID_AUTO, panic_reboot_wait_time, CTLFLAG_RWTUN, &panic_reboot_wait_time, 0, "Seconds to wait before rebooting after a panic"); /* * Note that stdarg.h and the ANSI style va_start macro is used for both * ANSI and traditional C compilers. */ #include #ifdef KDB #ifdef KDB_UNATTENDED int debugger_on_panic = 0; #else int debugger_on_panic = 1; #endif SYSCTL_INT(_debug, OID_AUTO, debugger_on_panic, CTLFLAG_RWTUN | CTLFLAG_SECURE, &debugger_on_panic, 0, "Run debugger on kernel panic"); #ifdef KDB_TRACE static int trace_on_panic = 1; #else static int trace_on_panic = 0; #endif SYSCTL_INT(_debug, OID_AUTO, trace_on_panic, CTLFLAG_RWTUN | CTLFLAG_SECURE, &trace_on_panic, 0, "Print stack trace on kernel panic"); #endif /* KDB */ static int sync_on_panic = 0; SYSCTL_INT(_kern, OID_AUTO, sync_on_panic, CTLFLAG_RWTUN, &sync_on_panic, 0, "Do a sync before rebooting from a panic"); static SYSCTL_NODE(_kern, OID_AUTO, shutdown, CTLFLAG_RW, 0, "Shutdown environment"); #ifndef DIAGNOSTIC static int show_busybufs; #else static int show_busybufs = 1; #endif SYSCTL_INT(_kern_shutdown, OID_AUTO, show_busybufs, CTLFLAG_RW, &show_busybufs, 0, ""); int suspend_blocked = 0; SYSCTL_INT(_kern, OID_AUTO, suspend_blocked, CTLFLAG_RW, &suspend_blocked, 0, "Block suspend due to a pending shutdown"); /* * Variable panicstr contains argument to first call to panic; used as flag * to indicate that the kernel has already called panic. */ const char *panicstr; int dumping; /* system is dumping */ int rebooting; /* system is rebooting */ static struct dumperinfo dumper; /* our selected dumper */ /* Context information for dump-debuggers. */ static struct pcb dumppcb; /* Registers. */ lwpid_t dumptid; /* Thread ID. */ static struct cdevsw reroot_cdevsw = { .d_version = D_VERSION, .d_name = "reroot", }; static void poweroff_wait(void *, int); static void shutdown_halt(void *junk, int howto); static void shutdown_panic(void *junk, int howto); static void shutdown_reset(void *junk, int howto); static int kern_reroot(void); /* register various local shutdown events */ static void shutdown_conf(void *unused) { EVENTHANDLER_REGISTER(shutdown_final, poweroff_wait, NULL, SHUTDOWN_PRI_FIRST); EVENTHANDLER_REGISTER(shutdown_final, shutdown_halt, NULL, SHUTDOWN_PRI_LAST + 100); EVENTHANDLER_REGISTER(shutdown_final, shutdown_panic, NULL, SHUTDOWN_PRI_LAST + 100); EVENTHANDLER_REGISTER(shutdown_final, shutdown_reset, NULL, SHUTDOWN_PRI_LAST + 200); } SYSINIT(shutdown_conf, SI_SUB_INTRINSIC, SI_ORDER_ANY, shutdown_conf, NULL); /* * The only reason this exists is to create the /dev/reroot/ directory, * used by reroot code in init(8) as a mountpoint for tmpfs. */ static void reroot_conf(void *unused) { int error; struct cdev *cdev; error = make_dev_p(MAKEDEV_CHECKNAME | MAKEDEV_WAITOK, &cdev, &reroot_cdevsw, NULL, UID_ROOT, GID_WHEEL, 0600, "reroot/reroot"); if (error != 0) { printf("%s: failed to create device node, error %d", __func__, error); } } SYSINIT(reroot_conf, SI_SUB_DEVFS, SI_ORDER_ANY, reroot_conf, NULL); /* * The system call that results in a reboot. */ /* ARGSUSED */ int sys_reboot(struct thread *td, struct reboot_args *uap) { int error; error = 0; #ifdef MAC error = mac_system_check_reboot(td->td_ucred, uap->opt); #endif if (error == 0) error = priv_check(td, PRIV_REBOOT); if (error == 0) { if (uap->opt & RB_REROOT) { error = kern_reroot(); } else { mtx_lock(&Giant); kern_reboot(uap->opt); mtx_unlock(&Giant); } } return (error); } /* * Called by events that want to shut down.. e.g on a PC */ void shutdown_nice(int howto) { if (initproc != NULL) { /* Send a signal to init(8) and have it shutdown the world. */ PROC_LOCK(initproc); if (howto & RB_POWEROFF) kern_psignal(initproc, SIGUSR2); else if (howto & RB_HALT) kern_psignal(initproc, SIGUSR1); else kern_psignal(initproc, SIGINT); PROC_UNLOCK(initproc); } else { /* No init(8) running, so simply reboot. */ kern_reboot(howto | RB_NOSYNC); } } static void print_uptime(void) { int f; struct timespec ts; getnanouptime(&ts); printf("Uptime: "); f = 0; if (ts.tv_sec >= 86400) { printf("%ldd", (long)ts.tv_sec / 86400); ts.tv_sec %= 86400; f = 1; } if (f || ts.tv_sec >= 3600) { printf("%ldh", (long)ts.tv_sec / 3600); ts.tv_sec %= 3600; f = 1; } if (f || ts.tv_sec >= 60) { printf("%ldm", (long)ts.tv_sec / 60); ts.tv_sec %= 60; f = 1; } printf("%lds\n", (long)ts.tv_sec); } int doadump(boolean_t textdump) { boolean_t coredump; int error; error = 0; if (dumping) return (EBUSY); if (dumper.dumper == NULL) return (ENXIO); savectx(&dumppcb); dumptid = curthread->td_tid; dumping++; coredump = TRUE; #ifdef DDB if (textdump && textdump_pending) { coredump = FALSE; textdump_dumpsys(&dumper); } #endif if (coredump) error = dumpsys(&dumper); dumping--; return (error); } /* * Shutdown the system cleanly to prepare for reboot, halt, or power off. */ void kern_reboot(int howto) { static int once = 0; #if defined(SMP) /* * Bind us to CPU 0 so that all shutdown code runs there. Some * systems don't shutdown properly (i.e., ACPI power off) if we * run on another processor. */ if (!SCHEDULER_STOPPED()) { thread_lock(curthread); sched_bind(curthread, 0); thread_unlock(curthread); KASSERT(PCPU_GET(cpuid) == 0, ("boot: not running on cpu 0")); } #endif /* We're in the process of rebooting. */ rebooting = 1; /* We are out of the debugger now. */ kdb_active = 0; /* * Do any callouts that should be done BEFORE syncing the filesystems. */ EVENTHANDLER_INVOKE(shutdown_pre_sync, howto); /* * Now sync filesystems */ if (!cold && (howto & RB_NOSYNC) == 0 && once == 0) { once = 1; bufshutdown(show_busybufs); } print_uptime(); cngrab(); /* * Ok, now do things that assume all filesystem activity has * been completed. */ EVENTHANDLER_INVOKE(shutdown_post_sync, howto); if ((howto & (RB_HALT|RB_DUMP)) == RB_DUMP && !cold && !dumping) doadump(TRUE); /* Now that we're going to really halt the system... */ EVENTHANDLER_INVOKE(shutdown_final, howto); for(;;) ; /* safety against shutdown_reset not working */ /* NOTREACHED */ } /* * The system call that results in changing the rootfs. */ static int kern_reroot(void) { struct vnode *oldrootvnode, *vp; struct mount *mp, *devmp; int error; if (curproc != initproc) return (EPERM); /* * Mark the filesystem containing currently-running executable * (the temporary copy of init(8)) busy. */ vp = curproc->p_textvp; error = vn_lock(vp, LK_SHARED); if (error != 0) return (error); mp = vp->v_mount; error = vfs_busy(mp, MBF_NOWAIT); if (error != 0) { vfs_ref(mp); VOP_UNLOCK(vp, 0); error = vfs_busy(mp, 0); vn_lock(vp, LK_SHARED | LK_RETRY); vfs_rel(mp); if (error != 0) { VOP_UNLOCK(vp, 0); return (ENOENT); } if (vp->v_iflag & VI_DOOMED) { VOP_UNLOCK(vp, 0); vfs_unbusy(mp); return (ENOENT); } } VOP_UNLOCK(vp, 0); /* * Remove the filesystem containing currently-running executable * from the mount list, to prevent it from being unmounted * by vfs_unmountall(), and to avoid confusing vfs_mountroot(). * * Also preserve /dev - forcibly unmounting it could cause driver * reinitialization. */ vfs_ref(rootdevmp); devmp = rootdevmp; rootdevmp = NULL; mtx_lock(&mountlist_mtx); TAILQ_REMOVE(&mountlist, mp, mnt_list); TAILQ_REMOVE(&mountlist, devmp, mnt_list); mtx_unlock(&mountlist_mtx); oldrootvnode = rootvnode; /* * Unmount everything except for the two filesystems preserved above. */ vfs_unmountall(); /* * Add /dev back; vfs_mountroot() will move it into its new place. */ mtx_lock(&mountlist_mtx); TAILQ_INSERT_HEAD(&mountlist, devmp, mnt_list); mtx_unlock(&mountlist_mtx); rootdevmp = devmp; vfs_rel(rootdevmp); /* * Mount the new rootfs. */ vfs_mountroot(); /* * Update all references to the old rootvnode. */ mountcheckdirs(oldrootvnode, rootvnode); /* * Add the temporary filesystem back and unbusy it. */ mtx_lock(&mountlist_mtx); TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); mtx_unlock(&mountlist_mtx); vfs_unbusy(mp); return (0); } /* * If the shutdown was a clean halt, behave accordingly. */ static void shutdown_halt(void *junk, int howto) { if (howto & RB_HALT) { printf("\n"); printf("The operating system has halted.\n"); printf("Please press any key to reboot.\n\n"); switch (cngetc()) { case -1: /* No console, just die */ cpu_halt(); /* NOTREACHED */ default: howto &= ~RB_HALT; break; } } } /* * Check to see if the system paniced, pause and then reboot * according to the specified delay. */ static void shutdown_panic(void *junk, int howto) { int loop; if (howto & RB_DUMP) { if (panic_reboot_wait_time != 0) { if (panic_reboot_wait_time != -1) { printf("Automatic reboot in %d seconds - " "press a key on the console to abort\n", panic_reboot_wait_time); for (loop = panic_reboot_wait_time * 10; loop > 0; --loop) { DELAY(1000 * 100); /* 1/10th second */ /* Did user type a key? */ if (cncheckc() != -1) break; } if (!loop) return; } } else { /* zero time specified - reboot NOW */ return; } printf("--> Press a key on the console to reboot,\n"); printf("--> or switch off the system now.\n"); cngetc(); } } /* * Everything done, now reset */ static void shutdown_reset(void *junk, int howto) { printf("Rebooting...\n"); DELAY(1000000); /* wait 1 sec for printf's to complete and be read */ /* * Acquiring smp_ipi_mtx here has a double effect: * - it disables interrupts avoiding CPU0 preemption * by fast handlers (thus deadlocking against other CPUs) * - it avoids deadlocks against smp_rendezvous() or, more * generally, threads busy-waiting, with this spinlock held, * and waiting for responses by threads on other CPUs * (ie. smp_tlb_shootdown()). * * For the !SMP case it just needs to handle the former problem. */ #ifdef SMP mtx_lock_spin(&smp_ipi_mtx); #else spinlock_enter(); #endif /* cpu_boot(howto); */ /* doesn't do anything at the moment */ cpu_reset(); /* NOTREACHED */ /* assuming reset worked */ } #if defined(WITNESS) || defined(INVARIANTS) static int kassert_warn_only = 0; #ifdef KDB static int kassert_do_kdb = 0; #endif #ifdef KTR static int kassert_do_ktr = 0; #endif static int kassert_do_log = 1; static int kassert_log_pps_limit = 4; static int kassert_log_mute_at = 0; static int kassert_log_panic_at = 0; static int kassert_warnings = 0; SYSCTL_NODE(_debug, OID_AUTO, kassert, CTLFLAG_RW, NULL, "kassert options"); SYSCTL_INT(_debug_kassert, OID_AUTO, warn_only, CTLFLAG_RWTUN, &kassert_warn_only, 0, "KASSERT triggers a panic (1) or just a warning (0)"); #ifdef KDB SYSCTL_INT(_debug_kassert, OID_AUTO, do_kdb, CTLFLAG_RWTUN, &kassert_do_kdb, 0, "KASSERT will enter the debugger"); #endif #ifdef KTR SYSCTL_UINT(_debug_kassert, OID_AUTO, do_ktr, CTLFLAG_RWTUN, &kassert_do_ktr, 0, "KASSERT does a KTR, set this to the KTRMASK you want"); #endif SYSCTL_INT(_debug_kassert, OID_AUTO, do_log, CTLFLAG_RWTUN, &kassert_do_log, 0, "KASSERT triggers a panic (1) or just a warning (0)"); SYSCTL_INT(_debug_kassert, OID_AUTO, warnings, CTLFLAG_RWTUN, &kassert_warnings, 0, "number of KASSERTs that have been triggered"); SYSCTL_INT(_debug_kassert, OID_AUTO, log_panic_at, CTLFLAG_RWTUN, &kassert_log_panic_at, 0, "max number of KASSERTS before we will panic"); SYSCTL_INT(_debug_kassert, OID_AUTO, log_pps_limit, CTLFLAG_RWTUN, &kassert_log_pps_limit, 0, "limit number of log messages per second"); SYSCTL_INT(_debug_kassert, OID_AUTO, log_mute_at, CTLFLAG_RWTUN, &kassert_log_mute_at, 0, "max number of KASSERTS to log"); static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS); SYSCTL_PROC(_debug_kassert, OID_AUTO, kassert, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE, NULL, 0, kassert_sysctl_kassert, "I", "set to trigger a test kassert"); static int kassert_sysctl_kassert(SYSCTL_HANDLER_ARGS) { int error, i; error = sysctl_wire_old_buffer(req, sizeof(int)); if (error == 0) { i = 0; error = sysctl_handle_int(oidp, &i, 0, req); } if (error != 0 || req->newptr == NULL) return (error); KASSERT(0, ("kassert_sysctl_kassert triggered kassert %d", i)); return (0); } /* * Called by KASSERT, this decides if we will panic * or if we will log via printf and/or ktr. */ void kassert_panic(const char *fmt, ...) { static char buf[256]; va_list ap; va_start(ap, fmt); (void)vsnprintf(buf, sizeof(buf), fmt, ap); va_end(ap); /* * panic if we're not just warning, or if we've exceeded * kassert_log_panic_at warnings. */ if (!kassert_warn_only || (kassert_log_panic_at > 0 && kassert_warnings >= kassert_log_panic_at)) { va_start(ap, fmt); vpanic(fmt, ap); /* NORETURN */ } #ifdef KTR if (kassert_do_ktr) CTR0(ktr_mask, buf); #endif /* KTR */ /* * log if we've not yet met the mute limit. */ if (kassert_do_log && (kassert_log_mute_at == 0 || kassert_warnings < kassert_log_mute_at)) { static struct timeval lasterr; static int curerr; if (ppsratecheck(&lasterr, &curerr, kassert_log_pps_limit)) { printf("KASSERT failed: %s\n", buf); kdb_backtrace(); } } #ifdef KDB if (kassert_do_kdb) { kdb_enter(KDB_WHY_KASSERT, buf); } #endif atomic_add_int(&kassert_warnings, 1); } #endif /* * Panic is called on unresolvable fatal errors. It prints "panic: mesg", * and then reboots. If we are called twice, then we avoid trying to sync * the disks as this often leads to recursive panics. */ void panic(const char *fmt, ...) { va_list ap; va_start(ap, fmt); vpanic(fmt, ap); } void vpanic(const char *fmt, va_list ap) { #ifdef SMP cpuset_t other_cpus; #endif struct thread *td = curthread; int bootopt, newpanic; static char buf[256]; spinlock_enter(); #ifdef SMP /* * stop_cpus_hard(other_cpus) should prevent multiple CPUs from * concurrently entering panic. Only the winner will proceed * further. */ if (panicstr == NULL && !kdb_active) { other_cpus = all_cpus; CPU_CLR(PCPU_GET(cpuid), &other_cpus); stop_cpus_hard(other_cpus); } /* * Ensure that the scheduler is stopped while panicking, even if panic * has been entered from kdb. */ td->td_stopsched = 1; #endif bootopt = RB_AUTOBOOT; newpanic = 0; if (panicstr) bootopt |= RB_NOSYNC; else { bootopt |= RB_DUMP; panicstr = fmt; newpanic = 1; } if (newpanic) { (void)vsnprintf(buf, sizeof(buf), fmt, ap); panicstr = buf; cngrab(); printf("panic: %s\n", buf); } else { printf("panic: "); vprintf(fmt, ap); printf("\n"); } #ifdef SMP printf("cpuid = %d\n", PCPU_GET(cpuid)); #endif #ifdef KDB if (newpanic && trace_on_panic) kdb_backtrace(); if (debugger_on_panic) kdb_enter(KDB_WHY_PANIC, "panic"); #endif /*thread_lock(td); */ td->td_flags |= TDF_INPANIC; /* thread_unlock(td); */ if (!sync_on_panic) bootopt |= RB_NOSYNC; kern_reboot(bootopt); } /* * Support for poweroff delay. * * Please note that setting this delay too short might power off your machine * before the write cache on your hard disk has been flushed, leading to * soft-updates inconsistencies. */ #ifndef POWEROFF_DELAY # define POWEROFF_DELAY 5000 #endif static int poweroff_delay = POWEROFF_DELAY; SYSCTL_INT(_kern_shutdown, OID_AUTO, poweroff_delay, CTLFLAG_RW, &poweroff_delay, 0, "Delay before poweroff to write disk caches (msec)"); static void poweroff_wait(void *junk, int howto) { if (!(howto & RB_POWEROFF) || poweroff_delay <= 0) return; DELAY(poweroff_delay * 1000); } /* * Some system processes (e.g. syncer) need to be stopped at appropriate * points in their main loops prior to a system shutdown, so that they * won't interfere with the shutdown process (e.g. by holding a disk buf * to cause sync to fail). For each of these system processes, register * shutdown_kproc() as a handler for one of shutdown events. */ static int kproc_shutdown_wait = 60; SYSCTL_INT(_kern_shutdown, OID_AUTO, kproc_shutdown_wait, CTLFLAG_RW, &kproc_shutdown_wait, 0, "Max wait time (sec) to stop for each process"); void kproc_shutdown(void *arg, int howto) { struct proc *p; int error; if (panicstr) return; p = (struct proc *)arg; printf("Waiting (max %d seconds) for system process `%s' to stop...", kproc_shutdown_wait, p->p_comm); error = kproc_suspend(p, kproc_shutdown_wait * hz); if (error == EWOULDBLOCK) printf("timed out\n"); else printf("done\n"); } void kthread_shutdown(void *arg, int howto) { struct thread *td; int error; if (panicstr) return; td = (struct thread *)arg; printf("Waiting (max %d seconds) for system thread `%s' to stop...", kproc_shutdown_wait, td->td_name); error = kthread_suspend(td, kproc_shutdown_wait * hz); if (error == EWOULDBLOCK) printf("timed out\n"); else printf("done\n"); } static char dumpdevname[sizeof(((struct cdev*)NULL)->si_name)]; SYSCTL_STRING(_kern_shutdown, OID_AUTO, dumpdevname, CTLFLAG_RD, dumpdevname, 0, "Device for kernel dumps"); /* Registration of dumpers */ int set_dumper(struct dumperinfo *di, const char *devname, struct thread *td) { size_t wantcopy; int error; error = priv_check(td, PRIV_SETDUMPER); if (error != 0) return (error); if (di == NULL) { bzero(&dumper, sizeof dumper); dumpdevname[0] = '\0'; return (0); } if (dumper.dumper != NULL) return (EBUSY); dumper = *di; wantcopy = strlcpy(dumpdevname, devname, sizeof(dumpdevname)); if (wantcopy >= sizeof(dumpdevname)) { printf("set_dumper: device name truncated from '%s' -> '%s'\n", devname, dumpdevname); } return (0); } /* Call dumper with bounds checking. */ int dump_write(struct dumperinfo *di, void *virtual, vm_offset_t physical, off_t offset, size_t length) { if (length != 0 && (offset < di->mediaoffset || offset - di->mediaoffset + length > di->mediasize)) { printf("Attempt to write outside dump device boundaries.\n" "offset(%jd), mediaoffset(%jd), length(%ju), mediasize(%jd).\n", (intmax_t)offset, (intmax_t)di->mediaoffset, (uintmax_t)length, (intmax_t)di->mediasize); return (ENOSPC); } return (di->dumper(di->priv, virtual, physical, offset, length)); } void mkdumpheader(struct kerneldumpheader *kdh, char *magic, uint32_t archver, uint64_t dumplen, uint32_t blksz) { bzero(kdh, sizeof(*kdh)); strlcpy(kdh->magic, magic, sizeof(kdh->magic)); strlcpy(kdh->architecture, MACHINE_ARCH, sizeof(kdh->architecture)); kdh->version = htod32(KERNELDUMPVERSION); kdh->architectureversion = htod32(archver); kdh->dumplength = htod64(dumplen); kdh->dumptime = htod64(time_second); kdh->blocksize = htod32(blksz); strlcpy(kdh->hostname, prison0.pr_hostname, sizeof(kdh->hostname)); strlcpy(kdh->versionstring, version, sizeof(kdh->versionstring)); if (panicstr != NULL) strlcpy(kdh->panicstring, panicstr, sizeof(kdh->panicstring)); kdh->parity = kerneldump_parity(kdh); } + +#ifdef DDB +DB_SHOW_COMMAND(panic, db_show_panic) +{ + + if (panicstr == NULL) + db_printf("Not paniced\n"); + else + db_printf("panic: %s\n", panicstr); +} +#endif Index: projects/vnet/sys/kern/kern_timeout.c =================================================================== --- projects/vnet/sys/kern/kern_timeout.c (revision 294565) +++ projects/vnet/sys/kern/kern_timeout.c (revision 294566) @@ -1,1617 +1,1655 @@ /*- * Copyright (c) 1982, 1986, 1991, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * From: @(#)kern_clock.c 8.5 (Berkeley) 1/21/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_callout_profiling.h" +#include "opt_ddb.h" #if defined(__arm__) #include "opt_timer.h" #endif #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#ifdef DDB +#include +#include +#endif + #ifdef SMP #include #endif #ifndef NO_EVENTTIMERS DPCPU_DECLARE(sbintime_t, hardclocktime); #endif SDT_PROVIDER_DEFINE(callout_execute); SDT_PROBE_DEFINE1(callout_execute, , , callout__start, "struct callout *"); SDT_PROBE_DEFINE1(callout_execute, , , callout__end, "struct callout *"); #ifdef CALLOUT_PROFILING static int avg_depth; SYSCTL_INT(_debug, OID_AUTO, to_avg_depth, CTLFLAG_RD, &avg_depth, 0, "Average number of items examined per softclock call. Units = 1/1000"); static int avg_gcalls; SYSCTL_INT(_debug, OID_AUTO, to_avg_gcalls, CTLFLAG_RD, &avg_gcalls, 0, "Average number of Giant callouts made per softclock call. Units = 1/1000"); static int avg_lockcalls; SYSCTL_INT(_debug, OID_AUTO, to_avg_lockcalls, CTLFLAG_RD, &avg_lockcalls, 0, "Average number of lock callouts made per softclock call. Units = 1/1000"); static int avg_mpcalls; SYSCTL_INT(_debug, OID_AUTO, to_avg_mpcalls, CTLFLAG_RD, &avg_mpcalls, 0, "Average number of MP callouts made per softclock call. Units = 1/1000"); static int avg_depth_dir; SYSCTL_INT(_debug, OID_AUTO, to_avg_depth_dir, CTLFLAG_RD, &avg_depth_dir, 0, "Average number of direct callouts examined per callout_process call. " "Units = 1/1000"); static int avg_lockcalls_dir; SYSCTL_INT(_debug, OID_AUTO, to_avg_lockcalls_dir, CTLFLAG_RD, &avg_lockcalls_dir, 0, "Average number of lock direct callouts made per " "callout_process call. Units = 1/1000"); static int avg_mpcalls_dir; SYSCTL_INT(_debug, OID_AUTO, to_avg_mpcalls_dir, CTLFLAG_RD, &avg_mpcalls_dir, 0, "Average number of MP direct callouts made per callout_process call. " "Units = 1/1000"); #endif static int ncallout; SYSCTL_INT(_kern, OID_AUTO, ncallout, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &ncallout, 0, "Number of entries in callwheel and size of timeout() preallocation"); #ifdef RSS static int pin_default_swi = 1; static int pin_pcpu_swi = 1; #else static int pin_default_swi = 0; static int pin_pcpu_swi = 0; #endif SYSCTL_INT(_kern, OID_AUTO, pin_default_swi, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &pin_default_swi, 0, "Pin the default (non-per-cpu) swi (shared with PCPU 0 swi)"); SYSCTL_INT(_kern, OID_AUTO, pin_pcpu_swi, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &pin_pcpu_swi, 0, "Pin the per-CPU swis (except PCPU 0, which is also default"); /* * TODO: * allocate more timeout table slots when table overflows. */ u_int callwheelsize, callwheelmask; /* * The callout cpu exec entities represent informations necessary for * describing the state of callouts currently running on the CPU and the ones * necessary for migrating callouts to the new callout cpu. In particular, * the first entry of the array cc_exec_entity holds informations for callout * running in SWI thread context, while the second one holds informations * for callout running directly from hardware interrupt context. * The cached informations are very important for deferring migration when * the migrating callout is already running. */ struct cc_exec { struct callout *cc_curr; void (*cc_drain)(void *); #ifdef SMP void (*ce_migration_func)(void *); void *ce_migration_arg; int ce_migration_cpu; sbintime_t ce_migration_time; sbintime_t ce_migration_prec; #endif bool cc_cancel; bool cc_waiting; }; /* * There is one struct callout_cpu per cpu, holding all relevant * state for the callout processing thread on the individual CPU. */ struct callout_cpu { struct mtx_padalign cc_lock; struct cc_exec cc_exec_entity[2]; struct callout *cc_next; struct callout *cc_callout; struct callout_list *cc_callwheel; struct callout_tailq cc_expireq; struct callout_slist cc_callfree; sbintime_t cc_firstevent; sbintime_t cc_lastscan; void *cc_cookie; u_int cc_bucket; u_int cc_inited; char cc_ktr_event_name[20]; }; #define callout_migrating(c) ((c)->c_iflags & CALLOUT_DFRMIGRATION) #define cc_exec_curr(cc, dir) cc->cc_exec_entity[dir].cc_curr #define cc_exec_drain(cc, dir) cc->cc_exec_entity[dir].cc_drain #define cc_exec_next(cc) cc->cc_next #define cc_exec_cancel(cc, dir) cc->cc_exec_entity[dir].cc_cancel #define cc_exec_waiting(cc, dir) cc->cc_exec_entity[dir].cc_waiting #ifdef SMP #define cc_migration_func(cc, dir) cc->cc_exec_entity[dir].ce_migration_func #define cc_migration_arg(cc, dir) cc->cc_exec_entity[dir].ce_migration_arg #define cc_migration_cpu(cc, dir) cc->cc_exec_entity[dir].ce_migration_cpu #define cc_migration_time(cc, dir) cc->cc_exec_entity[dir].ce_migration_time #define cc_migration_prec(cc, dir) cc->cc_exec_entity[dir].ce_migration_prec struct callout_cpu cc_cpu[MAXCPU]; #define CPUBLOCK MAXCPU #define CC_CPU(cpu) (&cc_cpu[(cpu)]) #define CC_SELF() CC_CPU(PCPU_GET(cpuid)) #else struct callout_cpu cc_cpu; #define CC_CPU(cpu) &cc_cpu #define CC_SELF() &cc_cpu #endif #define CC_LOCK(cc) mtx_lock_spin(&(cc)->cc_lock) #define CC_UNLOCK(cc) mtx_unlock_spin(&(cc)->cc_lock) #define CC_LOCK_ASSERT(cc) mtx_assert(&(cc)->cc_lock, MA_OWNED) static int timeout_cpu; static void callout_cpu_init(struct callout_cpu *cc, int cpu); static void softclock_call_cc(struct callout *c, struct callout_cpu *cc, #ifdef CALLOUT_PROFILING int *mpcalls, int *lockcalls, int *gcalls, #endif int direct); static MALLOC_DEFINE(M_CALLOUT, "callout", "Callout datastructures"); /** * Locked by cc_lock: * cc_curr - If a callout is in progress, it is cc_curr. * If cc_curr is non-NULL, threads waiting in * callout_drain() will be woken up as soon as the * relevant callout completes. * cc_cancel - Changing to 1 with both callout_lock and cc_lock held * guarantees that the current callout will not run. * The softclock() function sets this to 0 before it * drops callout_lock to acquire c_lock, and it calls * the handler only if curr_cancelled is still 0 after * cc_lock is successfully acquired. * cc_waiting - If a thread is waiting in callout_drain(), then * callout_wait is nonzero. Set only when * cc_curr is non-NULL. */ /* * Resets the execution entity tied to a specific callout cpu. */ static void cc_cce_cleanup(struct callout_cpu *cc, int direct) { cc_exec_curr(cc, direct) = NULL; cc_exec_cancel(cc, direct) = false; cc_exec_waiting(cc, direct) = false; #ifdef SMP cc_migration_cpu(cc, direct) = CPUBLOCK; cc_migration_time(cc, direct) = 0; cc_migration_prec(cc, direct) = 0; cc_migration_func(cc, direct) = NULL; cc_migration_arg(cc, direct) = NULL; #endif } /* * Checks if migration is requested by a specific callout cpu. */ static int cc_cce_migrating(struct callout_cpu *cc, int direct) { #ifdef SMP return (cc_migration_cpu(cc, direct) != CPUBLOCK); #else return (0); #endif } /* * Kernel low level callwheel initialization * called on cpu0 during kernel startup. */ static void callout_callwheel_init(void *dummy) { struct callout_cpu *cc; /* * Calculate the size of the callout wheel and the preallocated * timeout() structures. * XXX: Clip callout to result of previous function of maxusers * maximum 384. This is still huge, but acceptable. */ memset(CC_CPU(0), 0, sizeof(cc_cpu)); ncallout = imin(16 + maxproc + maxfiles, 18508); TUNABLE_INT_FETCH("kern.ncallout", &ncallout); /* * Calculate callout wheel size, should be next power of two higher * than 'ncallout'. */ callwheelsize = 1 << fls(ncallout); callwheelmask = callwheelsize - 1; /* * Fetch whether we're pinning the swi's or not. */ TUNABLE_INT_FETCH("kern.pin_default_swi", &pin_default_swi); TUNABLE_INT_FETCH("kern.pin_pcpu_swi", &pin_pcpu_swi); /* * Only cpu0 handles timeout(9) and receives a preallocation. * * XXX: Once all timeout(9) consumers are converted this can * be removed. */ timeout_cpu = PCPU_GET(cpuid); cc = CC_CPU(timeout_cpu); cc->cc_callout = malloc(ncallout * sizeof(struct callout), M_CALLOUT, M_WAITOK); callout_cpu_init(cc, timeout_cpu); } SYSINIT(callwheel_init, SI_SUB_CPU, SI_ORDER_ANY, callout_callwheel_init, NULL); /* * Initialize the per-cpu callout structures. */ static void callout_cpu_init(struct callout_cpu *cc, int cpu) { struct callout *c; int i; mtx_init(&cc->cc_lock, "callout", NULL, MTX_SPIN | MTX_RECURSE); SLIST_INIT(&cc->cc_callfree); cc->cc_inited = 1; cc->cc_callwheel = malloc(sizeof(struct callout_list) * callwheelsize, M_CALLOUT, M_WAITOK); for (i = 0; i < callwheelsize; i++) LIST_INIT(&cc->cc_callwheel[i]); TAILQ_INIT(&cc->cc_expireq); cc->cc_firstevent = SBT_MAX; for (i = 0; i < 2; i++) cc_cce_cleanup(cc, i); snprintf(cc->cc_ktr_event_name, sizeof(cc->cc_ktr_event_name), "callwheel cpu %d", cpu); if (cc->cc_callout == NULL) /* Only cpu0 handles timeout(9) */ return; for (i = 0; i < ncallout; i++) { c = &cc->cc_callout[i]; callout_init(c, 0); c->c_iflags = CALLOUT_LOCAL_ALLOC; SLIST_INSERT_HEAD(&cc->cc_callfree, c, c_links.sle); } } #ifdef SMP /* * Switches the cpu tied to a specific callout. * The function expects a locked incoming callout cpu and returns with * locked outcoming callout cpu. */ static struct callout_cpu * callout_cpu_switch(struct callout *c, struct callout_cpu *cc, int new_cpu) { struct callout_cpu *new_cc; MPASS(c != NULL && cc != NULL); CC_LOCK_ASSERT(cc); /* * Avoid interrupts and preemption firing after the callout cpu * is blocked in order to avoid deadlocks as the new thread * may be willing to acquire the callout cpu lock. */ c->c_cpu = CPUBLOCK; spinlock_enter(); CC_UNLOCK(cc); new_cc = CC_CPU(new_cpu); CC_LOCK(new_cc); spinlock_exit(); c->c_cpu = new_cpu; return (new_cc); } #endif /* * Start standard softclock thread. */ static void start_softclock(void *dummy) { struct callout_cpu *cc; char name[MAXCOMLEN]; #ifdef SMP int cpu; struct intr_event *ie; #endif cc = CC_CPU(timeout_cpu); snprintf(name, sizeof(name), "clock (%d)", timeout_cpu); if (swi_add(&clk_intr_event, name, softclock, cc, SWI_CLOCK, INTR_MPSAFE, &cc->cc_cookie)) panic("died while creating standard software ithreads"); if (pin_default_swi && (intr_event_bind(clk_intr_event, timeout_cpu) != 0)) { printf("%s: timeout clock couldn't be pinned to cpu %d\n", __func__, timeout_cpu); } #ifdef SMP CPU_FOREACH(cpu) { if (cpu == timeout_cpu) continue; cc = CC_CPU(cpu); cc->cc_callout = NULL; /* Only cpu0 handles timeout(9). */ callout_cpu_init(cc, cpu); snprintf(name, sizeof(name), "clock (%d)", cpu); ie = NULL; if (swi_add(&ie, name, softclock, cc, SWI_CLOCK, INTR_MPSAFE, &cc->cc_cookie)) panic("died while creating standard software ithreads"); if (pin_pcpu_swi && (intr_event_bind(ie, cpu) != 0)) { printf("%s: per-cpu clock couldn't be pinned to " "cpu %d\n", __func__, cpu); } } #endif } SYSINIT(start_softclock, SI_SUB_SOFTINTR, SI_ORDER_FIRST, start_softclock, NULL); #define CC_HASH_SHIFT 8 static inline u_int callout_hash(sbintime_t sbt) { return (sbt >> (32 - CC_HASH_SHIFT)); } static inline u_int callout_get_bucket(sbintime_t sbt) { return (callout_hash(sbt) & callwheelmask); } void callout_process(sbintime_t now) { struct callout *tmp, *tmpn; struct callout_cpu *cc; struct callout_list *sc; sbintime_t first, last, max, tmp_max; uint32_t lookahead; u_int firstb, lastb, nowb; #ifdef CALLOUT_PROFILING int depth_dir = 0, mpcalls_dir = 0, lockcalls_dir = 0; #endif cc = CC_SELF(); mtx_lock_spin_flags(&cc->cc_lock, MTX_QUIET); /* Compute the buckets of the last scan and present times. */ firstb = callout_hash(cc->cc_lastscan); cc->cc_lastscan = now; nowb = callout_hash(now); /* Compute the last bucket and minimum time of the bucket after it. */ if (nowb == firstb) lookahead = (SBT_1S / 16); else if (nowb - firstb == 1) lookahead = (SBT_1S / 8); else lookahead = (SBT_1S / 2); first = last = now; first += (lookahead / 2); last += lookahead; last &= (0xffffffffffffffffLLU << (32 - CC_HASH_SHIFT)); lastb = callout_hash(last) - 1; max = last; /* * Check if we wrapped around the entire wheel from the last scan. * In case, we need to scan entirely the wheel for pending callouts. */ if (lastb - firstb >= callwheelsize) { lastb = firstb + callwheelsize - 1; if (nowb - firstb >= callwheelsize) nowb = lastb; } /* Iterate callwheel from firstb to nowb and then up to lastb. */ do { sc = &cc->cc_callwheel[firstb & callwheelmask]; tmp = LIST_FIRST(sc); while (tmp != NULL) { /* Run the callout if present time within allowed. */ if (tmp->c_time <= now) { /* * Consumer told us the callout may be run * directly from hardware interrupt context. */ if (tmp->c_iflags & CALLOUT_DIRECT) { #ifdef CALLOUT_PROFILING ++depth_dir; #endif cc_exec_next(cc) = LIST_NEXT(tmp, c_links.le); cc->cc_bucket = firstb & callwheelmask; LIST_REMOVE(tmp, c_links.le); softclock_call_cc(tmp, cc, #ifdef CALLOUT_PROFILING &mpcalls_dir, &lockcalls_dir, NULL, #endif 1); tmp = cc_exec_next(cc); cc_exec_next(cc) = NULL; } else { tmpn = LIST_NEXT(tmp, c_links.le); LIST_REMOVE(tmp, c_links.le); TAILQ_INSERT_TAIL(&cc->cc_expireq, tmp, c_links.tqe); tmp->c_iflags |= CALLOUT_PROCESSED; tmp = tmpn; } continue; } /* Skip events from distant future. */ if (tmp->c_time >= max) goto next; /* * Event minimal time is bigger than present maximal * time, so it cannot be aggregated. */ if (tmp->c_time > last) { lastb = nowb; goto next; } /* Update first and last time, respecting this event. */ if (tmp->c_time < first) first = tmp->c_time; tmp_max = tmp->c_time + tmp->c_precision; if (tmp_max < last) last = tmp_max; next: tmp = LIST_NEXT(tmp, c_links.le); } /* Proceed with the next bucket. */ firstb++; /* * Stop if we looked after present time and found * some event we can't execute at now. * Stop if we looked far enough into the future. */ } while (((int)(firstb - lastb)) <= 0); cc->cc_firstevent = last; #ifndef NO_EVENTTIMERS cpu_new_callout(curcpu, last, first); #endif #ifdef CALLOUT_PROFILING avg_depth_dir += (depth_dir * 1000 - avg_depth_dir) >> 8; avg_mpcalls_dir += (mpcalls_dir * 1000 - avg_mpcalls_dir) >> 8; avg_lockcalls_dir += (lockcalls_dir * 1000 - avg_lockcalls_dir) >> 8; #endif mtx_unlock_spin_flags(&cc->cc_lock, MTX_QUIET); /* * swi_sched acquires the thread lock, so we don't want to call it * with cc_lock held; incorrect locking order. */ if (!TAILQ_EMPTY(&cc->cc_expireq)) swi_sched(cc->cc_cookie, 0); } static struct callout_cpu * callout_lock(struct callout *c) { struct callout_cpu *cc; int cpu; for (;;) { cpu = c->c_cpu; #ifdef SMP if (cpu == CPUBLOCK) { while (c->c_cpu == CPUBLOCK) cpu_spinwait(); continue; } #endif cc = CC_CPU(cpu); CC_LOCK(cc); if (cpu == c->c_cpu) break; CC_UNLOCK(cc); } return (cc); } static void callout_cc_add(struct callout *c, struct callout_cpu *cc, sbintime_t sbt, sbintime_t precision, void (*func)(void *), void *arg, int cpu, int flags) { int bucket; CC_LOCK_ASSERT(cc); if (sbt < cc->cc_lastscan) sbt = cc->cc_lastscan; c->c_arg = arg; c->c_iflags |= CALLOUT_PENDING; c->c_iflags &= ~CALLOUT_PROCESSED; c->c_flags |= CALLOUT_ACTIVE; if (flags & C_DIRECT_EXEC) c->c_iflags |= CALLOUT_DIRECT; c->c_func = func; c->c_time = sbt; c->c_precision = precision; bucket = callout_get_bucket(c->c_time); CTR3(KTR_CALLOUT, "precision set for %p: %d.%08x", c, (int)(c->c_precision >> 32), (u_int)(c->c_precision & 0xffffffff)); LIST_INSERT_HEAD(&cc->cc_callwheel[bucket], c, c_links.le); if (cc->cc_bucket == bucket) cc_exec_next(cc) = c; #ifndef NO_EVENTTIMERS /* * Inform the eventtimers(4) subsystem there's a new callout * that has been inserted, but only if really required. */ if (SBT_MAX - c->c_time < c->c_precision) c->c_precision = SBT_MAX - c->c_time; sbt = c->c_time + c->c_precision; if (sbt < cc->cc_firstevent) { cc->cc_firstevent = sbt; cpu_new_callout(cpu, sbt, c->c_time); } #endif } static void callout_cc_del(struct callout *c, struct callout_cpu *cc) { if ((c->c_iflags & CALLOUT_LOCAL_ALLOC) == 0) return; c->c_func = NULL; SLIST_INSERT_HEAD(&cc->cc_callfree, c, c_links.sle); } static void softclock_call_cc(struct callout *c, struct callout_cpu *cc, #ifdef CALLOUT_PROFILING int *mpcalls, int *lockcalls, int *gcalls, #endif int direct) { struct rm_priotracker tracker; void (*c_func)(void *); void *c_arg; struct lock_class *class; struct lock_object *c_lock; uintptr_t lock_status; int c_iflags; #ifdef SMP struct callout_cpu *new_cc; void (*new_func)(void *); void *new_arg; int flags, new_cpu; sbintime_t new_prec, new_time; #endif #if defined(DIAGNOSTIC) || defined(CALLOUT_PROFILING) sbintime_t sbt1, sbt2; struct timespec ts2; static sbintime_t maxdt = 2 * SBT_1MS; /* 2 msec */ static timeout_t *lastfunc; #endif KASSERT((c->c_iflags & CALLOUT_PENDING) == CALLOUT_PENDING, ("softclock_call_cc: pend %p %x", c, c->c_iflags)); KASSERT((c->c_flags & CALLOUT_ACTIVE) == CALLOUT_ACTIVE, ("softclock_call_cc: act %p %x", c, c->c_flags)); class = (c->c_lock != NULL) ? LOCK_CLASS(c->c_lock) : NULL; lock_status = 0; if (c->c_flags & CALLOUT_SHAREDLOCK) { if (class == &lock_class_rm) lock_status = (uintptr_t)&tracker; else lock_status = 1; } c_lock = c->c_lock; c_func = c->c_func; c_arg = c->c_arg; c_iflags = c->c_iflags; if (c->c_iflags & CALLOUT_LOCAL_ALLOC) c->c_iflags = CALLOUT_LOCAL_ALLOC; else c->c_iflags &= ~CALLOUT_PENDING; cc_exec_curr(cc, direct) = c; cc_exec_cancel(cc, direct) = false; cc_exec_drain(cc, direct) = NULL; CC_UNLOCK(cc); if (c_lock != NULL) { class->lc_lock(c_lock, lock_status); /* * The callout may have been cancelled * while we switched locks. */ if (cc_exec_cancel(cc, direct)) { class->lc_unlock(c_lock); goto skip; } /* The callout cannot be stopped now. */ cc_exec_cancel(cc, direct) = true; if (c_lock == &Giant.lock_object) { #ifdef CALLOUT_PROFILING (*gcalls)++; #endif CTR3(KTR_CALLOUT, "callout giant %p func %p arg %p", c, c_func, c_arg); } else { #ifdef CALLOUT_PROFILING (*lockcalls)++; #endif CTR3(KTR_CALLOUT, "callout lock %p func %p arg %p", c, c_func, c_arg); } } else { #ifdef CALLOUT_PROFILING (*mpcalls)++; #endif CTR3(KTR_CALLOUT, "callout %p func %p arg %p", c, c_func, c_arg); } KTR_STATE3(KTR_SCHED, "callout", cc->cc_ktr_event_name, "running", "func:%p", c_func, "arg:%p", c_arg, "direct:%d", direct); #if defined(DIAGNOSTIC) || defined(CALLOUT_PROFILING) sbt1 = sbinuptime(); #endif THREAD_NO_SLEEPING(); SDT_PROBE1(callout_execute, , , callout__start, c); c_func(c_arg); SDT_PROBE1(callout_execute, , , callout__end, c); THREAD_SLEEPING_OK(); #if defined(DIAGNOSTIC) || defined(CALLOUT_PROFILING) sbt2 = sbinuptime(); sbt2 -= sbt1; if (sbt2 > maxdt) { if (lastfunc != c_func || sbt2 > maxdt * 2) { ts2 = sbttots(sbt2); printf( "Expensive timeout(9) function: %p(%p) %jd.%09ld s\n", c_func, c_arg, (intmax_t)ts2.tv_sec, ts2.tv_nsec); } maxdt = sbt2; lastfunc = c_func; } #endif KTR_STATE0(KTR_SCHED, "callout", cc->cc_ktr_event_name, "idle"); CTR1(KTR_CALLOUT, "callout %p finished", c); if ((c_iflags & CALLOUT_RETURNUNLOCKED) == 0) class->lc_unlock(c_lock); skip: CC_LOCK(cc); KASSERT(cc_exec_curr(cc, direct) == c, ("mishandled cc_curr")); cc_exec_curr(cc, direct) = NULL; if (cc_exec_drain(cc, direct)) { void (*drain)(void *); drain = cc_exec_drain(cc, direct); cc_exec_drain(cc, direct) = NULL; CC_UNLOCK(cc); drain(c_arg); CC_LOCK(cc); } if (cc_exec_waiting(cc, direct)) { /* * There is someone waiting for the * callout to complete. * If the callout was scheduled for * migration just cancel it. */ if (cc_cce_migrating(cc, direct)) { cc_cce_cleanup(cc, direct); /* * It should be assert here that the callout is not * destroyed but that is not easy. */ c->c_iflags &= ~CALLOUT_DFRMIGRATION; } cc_exec_waiting(cc, direct) = false; CC_UNLOCK(cc); wakeup(&cc_exec_waiting(cc, direct)); CC_LOCK(cc); } else if (cc_cce_migrating(cc, direct)) { KASSERT((c_iflags & CALLOUT_LOCAL_ALLOC) == 0, ("Migrating legacy callout %p", c)); #ifdef SMP /* * If the callout was scheduled for * migration just perform it now. */ new_cpu = cc_migration_cpu(cc, direct); new_time = cc_migration_time(cc, direct); new_prec = cc_migration_prec(cc, direct); new_func = cc_migration_func(cc, direct); new_arg = cc_migration_arg(cc, direct); cc_cce_cleanup(cc, direct); /* * It should be assert here that the callout is not destroyed * but that is not easy. * * As first thing, handle deferred callout stops. */ if (!callout_migrating(c)) { CTR3(KTR_CALLOUT, "deferred cancelled %p func %p arg %p", c, new_func, new_arg); callout_cc_del(c, cc); return; } c->c_iflags &= ~CALLOUT_DFRMIGRATION; new_cc = callout_cpu_switch(c, cc, new_cpu); flags = (direct) ? C_DIRECT_EXEC : 0; callout_cc_add(c, new_cc, new_time, new_prec, new_func, new_arg, new_cpu, flags); CC_UNLOCK(new_cc); CC_LOCK(cc); #else panic("migration should not happen"); #endif } /* * If the current callout is locally allocated (from * timeout(9)) then put it on the freelist. * * Note: we need to check the cached copy of c_iflags because * if it was not local, then it's not safe to deref the * callout pointer. */ KASSERT((c_iflags & CALLOUT_LOCAL_ALLOC) == 0 || c->c_iflags == CALLOUT_LOCAL_ALLOC, ("corrupted callout")); if (c_iflags & CALLOUT_LOCAL_ALLOC) callout_cc_del(c, cc); } /* * The callout mechanism is based on the work of Adam M. Costello and * George Varghese, published in a technical report entitled "Redesigning * the BSD Callout and Timer Facilities" and modified slightly for inclusion * in FreeBSD by Justin T. Gibbs. The original work on the data structures * used in this implementation was published by G. Varghese and T. Lauck in * the paper "Hashed and Hierarchical Timing Wheels: Data Structures for * the Efficient Implementation of a Timer Facility" in the Proceedings of * the 11th ACM Annual Symposium on Operating Systems Principles, * Austin, Texas Nov 1987. */ /* * Software (low priority) clock interrupt. * Run periodic events from timeout queue. */ void softclock(void *arg) { struct callout_cpu *cc; struct callout *c; #ifdef CALLOUT_PROFILING int depth = 0, gcalls = 0, lockcalls = 0, mpcalls = 0; #endif cc = (struct callout_cpu *)arg; CC_LOCK(cc); while ((c = TAILQ_FIRST(&cc->cc_expireq)) != NULL) { TAILQ_REMOVE(&cc->cc_expireq, c, c_links.tqe); softclock_call_cc(c, cc, #ifdef CALLOUT_PROFILING &mpcalls, &lockcalls, &gcalls, #endif 0); #ifdef CALLOUT_PROFILING ++depth; #endif } #ifdef CALLOUT_PROFILING avg_depth += (depth * 1000 - avg_depth) >> 8; avg_mpcalls += (mpcalls * 1000 - avg_mpcalls) >> 8; avg_lockcalls += (lockcalls * 1000 - avg_lockcalls) >> 8; avg_gcalls += (gcalls * 1000 - avg_gcalls) >> 8; #endif CC_UNLOCK(cc); } /* * timeout -- * Execute a function after a specified length of time. * * untimeout -- * Cancel previous timeout function call. * * callout_handle_init -- * Initialize a handle so that using it with untimeout is benign. * * See AT&T BCI Driver Reference Manual for specification. This * implementation differs from that one in that although an * identification value is returned from timeout, the original * arguments to timeout as well as the identifier are used to * identify entries for untimeout. */ struct callout_handle timeout(timeout_t *ftn, void *arg, int to_ticks) { struct callout_cpu *cc; struct callout *new; struct callout_handle handle; cc = CC_CPU(timeout_cpu); CC_LOCK(cc); /* Fill in the next free callout structure. */ new = SLIST_FIRST(&cc->cc_callfree); if (new == NULL) /* XXX Attempt to malloc first */ panic("timeout table full"); SLIST_REMOVE_HEAD(&cc->cc_callfree, c_links.sle); callout_reset(new, to_ticks, ftn, arg); handle.callout = new; CC_UNLOCK(cc); return (handle); } void untimeout(timeout_t *ftn, void *arg, struct callout_handle handle) { struct callout_cpu *cc; /* * Check for a handle that was initialized * by callout_handle_init, but never used * for a real timeout. */ if (handle.callout == NULL) return; cc = callout_lock(handle.callout); if (handle.callout->c_func == ftn && handle.callout->c_arg == arg) callout_stop(handle.callout); CC_UNLOCK(cc); } void callout_handle_init(struct callout_handle *handle) { handle->callout = NULL; } /* * New interface; clients allocate their own callout structures. * * callout_reset() - establish or change a timeout * callout_stop() - disestablish a timeout * callout_init() - initialize a callout structure so that it can * safely be passed to callout_reset() and callout_stop() * * defines three convenience macros: * * callout_active() - returns truth if callout has not been stopped, * drained, or deactivated since the last time the callout was * reset. * callout_pending() - returns truth if callout is still waiting for timeout * callout_deactivate() - marks the callout as having been serviced */ int callout_reset_sbt_on(struct callout *c, sbintime_t sbt, sbintime_t precision, void (*ftn)(void *), void *arg, int cpu, int flags) { sbintime_t to_sbt, pr; struct callout_cpu *cc; int cancelled, direct; int ignore_cpu=0; cancelled = 0; if (cpu == -1) { ignore_cpu = 1; } else if ((cpu >= MAXCPU) || ((CC_CPU(cpu))->cc_inited == 0)) { /* Invalid CPU spec */ panic("Invalid CPU in callout %d", cpu); } if (flags & C_ABSOLUTE) { to_sbt = sbt; } else { if ((flags & C_HARDCLOCK) && (sbt < tick_sbt)) sbt = tick_sbt; if ((flags & C_HARDCLOCK) || #ifdef NO_EVENTTIMERS sbt >= sbt_timethreshold) { to_sbt = getsbinuptime(); /* Add safety belt for the case of hz > 1000. */ to_sbt += tc_tick_sbt - tick_sbt; #else sbt >= sbt_tickthreshold) { /* * Obtain the time of the last hardclock() call on * this CPU directly from the kern_clocksource.c. * This value is per-CPU, but it is equal for all * active ones. */ #ifdef __LP64__ to_sbt = DPCPU_GET(hardclocktime); #else spinlock_enter(); to_sbt = DPCPU_GET(hardclocktime); spinlock_exit(); #endif #endif if ((flags & C_HARDCLOCK) == 0) to_sbt += tick_sbt; } else to_sbt = sbinuptime(); if (SBT_MAX - to_sbt < sbt) to_sbt = SBT_MAX; else to_sbt += sbt; pr = ((C_PRELGET(flags) < 0) ? sbt >> tc_precexp : sbt >> C_PRELGET(flags)); if (pr > precision) precision = pr; } /* * This flag used to be added by callout_cc_add, but the * first time you call this we could end up with the * wrong direct flag if we don't do it before we add. */ if (flags & C_DIRECT_EXEC) { direct = 1; } else { direct = 0; } KASSERT(!direct || c->c_lock == NULL, ("%s: direct callout %p has lock", __func__, c)); cc = callout_lock(c); /* * Don't allow migration of pre-allocated callouts lest they * become unbalanced or handle the case where the user does * not care. */ if ((c->c_iflags & CALLOUT_LOCAL_ALLOC) || ignore_cpu) { cpu = c->c_cpu; } if (cc_exec_curr(cc, direct) == c) { /* * We're being asked to reschedule a callout which is * currently in progress. If there is a lock then we * can cancel the callout if it has not really started. */ if (c->c_lock != NULL && !cc_exec_cancel(cc, direct)) cancelled = cc_exec_cancel(cc, direct) = true; if (cc_exec_waiting(cc, direct)) { /* * Someone has called callout_drain to kill this * callout. Don't reschedule. */ CTR4(KTR_CALLOUT, "%s %p func %p arg %p", cancelled ? "cancelled" : "failed to cancel", c, c->c_func, c->c_arg); CC_UNLOCK(cc); return (cancelled); } #ifdef SMP if (callout_migrating(c)) { /* * This only occurs when a second callout_reset_sbt_on * is made after a previous one moved it into * deferred migration (below). Note we do *not* change * the prev_cpu even though the previous target may * be different. */ cc_migration_cpu(cc, direct) = cpu; cc_migration_time(cc, direct) = to_sbt; cc_migration_prec(cc, direct) = precision; cc_migration_func(cc, direct) = ftn; cc_migration_arg(cc, direct) = arg; cancelled = 1; CC_UNLOCK(cc); return (cancelled); } #endif } if (c->c_iflags & CALLOUT_PENDING) { if ((c->c_iflags & CALLOUT_PROCESSED) == 0) { if (cc_exec_next(cc) == c) cc_exec_next(cc) = LIST_NEXT(c, c_links.le); LIST_REMOVE(c, c_links.le); } else { TAILQ_REMOVE(&cc->cc_expireq, c, c_links.tqe); } cancelled = 1; c->c_iflags &= ~ CALLOUT_PENDING; c->c_flags &= ~ CALLOUT_ACTIVE; } #ifdef SMP /* * If the callout must migrate try to perform it immediately. * If the callout is currently running, just defer the migration * to a more appropriate moment. */ if (c->c_cpu != cpu) { if (cc_exec_curr(cc, direct) == c) { /* * Pending will have been removed since we are * actually executing the callout on another * CPU. That callout should be waiting on the * lock the caller holds. If we set both * active/and/pending after we return and the * lock on the executing callout proceeds, it * will then see pending is true and return. * At the return from the actual callout execution * the migration will occur in softclock_call_cc * and this new callout will be placed on the * new CPU via a call to callout_cpu_switch() which * will get the lock on the right CPU followed * by a call callout_cc_add() which will add it there. * (see above in softclock_call_cc()). */ cc_migration_cpu(cc, direct) = cpu; cc_migration_time(cc, direct) = to_sbt; cc_migration_prec(cc, direct) = precision; cc_migration_func(cc, direct) = ftn; cc_migration_arg(cc, direct) = arg; c->c_iflags |= (CALLOUT_DFRMIGRATION | CALLOUT_PENDING); c->c_flags |= CALLOUT_ACTIVE; CTR6(KTR_CALLOUT, "migration of %p func %p arg %p in %d.%08x to %u deferred", c, c->c_func, c->c_arg, (int)(to_sbt >> 32), (u_int)(to_sbt & 0xffffffff), cpu); CC_UNLOCK(cc); return (cancelled); } cc = callout_cpu_switch(c, cc, cpu); } #endif callout_cc_add(c, cc, to_sbt, precision, ftn, arg, cpu, flags); CTR6(KTR_CALLOUT, "%sscheduled %p func %p arg %p in %d.%08x", cancelled ? "re" : "", c, c->c_func, c->c_arg, (int)(to_sbt >> 32), (u_int)(to_sbt & 0xffffffff)); CC_UNLOCK(cc); return (cancelled); } /* * Common idioms that can be optimized in the future. */ int callout_schedule_on(struct callout *c, int to_ticks, int cpu) { return callout_reset_on(c, to_ticks, c->c_func, c->c_arg, cpu); } int callout_schedule(struct callout *c, int to_ticks) { return callout_reset_on(c, to_ticks, c->c_func, c->c_arg, c->c_cpu); } int _callout_stop_safe(struct callout *c, int safe, void (*drain)(void *)) { struct callout_cpu *cc, *old_cc; struct lock_class *class; int direct, sq_locked, use_lock; int not_on_a_list; if (safe) WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, c->c_lock, "calling %s", __func__); /* * Some old subsystems don't hold Giant while running a callout_stop(), * so just discard this check for the moment. */ if (!safe && c->c_lock != NULL) { if (c->c_lock == &Giant.lock_object) use_lock = mtx_owned(&Giant); else { use_lock = 1; class = LOCK_CLASS(c->c_lock); class->lc_assert(c->c_lock, LA_XLOCKED); } } else use_lock = 0; if (c->c_iflags & CALLOUT_DIRECT) { direct = 1; } else { direct = 0; } sq_locked = 0; old_cc = NULL; again: cc = callout_lock(c); if ((c->c_iflags & (CALLOUT_DFRMIGRATION | CALLOUT_PENDING)) == (CALLOUT_DFRMIGRATION | CALLOUT_PENDING) && ((c->c_flags & CALLOUT_ACTIVE) == CALLOUT_ACTIVE)) { /* * Special case where this slipped in while we * were migrating *as* the callout is about to * execute. The caller probably holds the lock * the callout wants. * * Get rid of the migration first. Then set * the flag that tells this code *not* to * try to remove it from any lists (its not * on one yet). When the callout wheel runs, * it will ignore this callout. */ c->c_iflags &= ~CALLOUT_PENDING; c->c_flags &= ~CALLOUT_ACTIVE; not_on_a_list = 1; } else { not_on_a_list = 0; } /* * If the callout was migrating while the callout cpu lock was * dropped, just drop the sleepqueue lock and check the states * again. */ if (sq_locked != 0 && cc != old_cc) { #ifdef SMP CC_UNLOCK(cc); sleepq_release(&cc_exec_waiting(old_cc, direct)); sq_locked = 0; old_cc = NULL; goto again; #else panic("migration should not happen"); #endif } /* * If the callout isn't pending, it's not on the queue, so * don't attempt to remove it from the queue. We can try to * stop it by other means however. */ if (!(c->c_iflags & CALLOUT_PENDING)) { /* * If it wasn't on the queue and it isn't the current * callout, then we can't stop it, so just bail. * It probably has already been run (if locking * is properly done). You could get here if the caller * calls stop twice in a row for example. The second * call would fall here without CALLOUT_ACTIVE set. */ c->c_flags &= ~CALLOUT_ACTIVE; if (cc_exec_curr(cc, direct) != c) { CTR3(KTR_CALLOUT, "failed to stop %p func %p arg %p", c, c->c_func, c->c_arg); CC_UNLOCK(cc); if (sq_locked) sleepq_release(&cc_exec_waiting(cc, direct)); return (-1); } if (safe) { /* * The current callout is running (or just * about to run) and blocking is allowed, so * just wait for the current invocation to * finish. */ while (cc_exec_curr(cc, direct) == c) { /* * Use direct calls to sleepqueue interface * instead of cv/msleep in order to avoid * a LOR between cc_lock and sleepqueue * chain spinlocks. This piece of code * emulates a msleep_spin() call actually. * * If we already have the sleepqueue chain * locked, then we can safely block. If we * don't already have it locked, however, * we have to drop the cc_lock to lock * it. This opens several races, so we * restart at the beginning once we have * both locks. If nothing has changed, then * we will end up back here with sq_locked * set. */ if (!sq_locked) { CC_UNLOCK(cc); sleepq_lock( &cc_exec_waiting(cc, direct)); sq_locked = 1; old_cc = cc; goto again; } /* * Migration could be cancelled here, but * as long as it is still not sure when it * will be packed up, just let softclock() * take care of it. */ cc_exec_waiting(cc, direct) = true; DROP_GIANT(); CC_UNLOCK(cc); sleepq_add( &cc_exec_waiting(cc, direct), &cc->cc_lock.lock_object, "codrain", SLEEPQ_SLEEP, 0); sleepq_wait( &cc_exec_waiting(cc, direct), 0); sq_locked = 0; old_cc = NULL; /* Reacquire locks previously released. */ PICKUP_GIANT(); CC_LOCK(cc); } } else if (use_lock && !cc_exec_cancel(cc, direct) && (drain == NULL)) { /* * The current callout is waiting for its * lock which we hold. Cancel the callout * and return. After our caller drops the * lock, the callout will be skipped in * softclock(). This *only* works with a * callout_stop() *not* callout_drain() or * callout_async_drain(). */ cc_exec_cancel(cc, direct) = true; CTR3(KTR_CALLOUT, "cancelled %p func %p arg %p", c, c->c_func, c->c_arg); KASSERT(!cc_cce_migrating(cc, direct), ("callout wrongly scheduled for migration")); if (callout_migrating(c)) { c->c_iflags &= ~CALLOUT_DFRMIGRATION; #ifdef SMP cc_migration_cpu(cc, direct) = CPUBLOCK; cc_migration_time(cc, direct) = 0; cc_migration_prec(cc, direct) = 0; cc_migration_func(cc, direct) = NULL; cc_migration_arg(cc, direct) = NULL; #endif } CC_UNLOCK(cc); KASSERT(!sq_locked, ("sleepqueue chain locked")); return (1); } else if (callout_migrating(c)) { /* * The callout is currently being serviced * and the "next" callout is scheduled at * its completion with a migration. We remove * the migration flag so it *won't* get rescheduled, * but we can't stop the one thats running so * we return 0. */ c->c_iflags &= ~CALLOUT_DFRMIGRATION; #ifdef SMP /* * We can't call cc_cce_cleanup here since * if we do it will remove .ce_curr and * its still running. This will prevent a * reschedule of the callout when the * execution completes. */ cc_migration_cpu(cc, direct) = CPUBLOCK; cc_migration_time(cc, direct) = 0; cc_migration_prec(cc, direct) = 0; cc_migration_func(cc, direct) = NULL; cc_migration_arg(cc, direct) = NULL; #endif CTR3(KTR_CALLOUT, "postponing stop %p func %p arg %p", c, c->c_func, c->c_arg); if (drain) { cc_exec_drain(cc, direct) = drain; } CC_UNLOCK(cc); return (0); } CTR3(KTR_CALLOUT, "failed to stop %p func %p arg %p", c, c->c_func, c->c_arg); if (drain) { cc_exec_drain(cc, direct) = drain; } CC_UNLOCK(cc); KASSERT(!sq_locked, ("sleepqueue chain still locked")); return (0); } if (sq_locked) sleepq_release(&cc_exec_waiting(cc, direct)); c->c_iflags &= ~CALLOUT_PENDING; c->c_flags &= ~CALLOUT_ACTIVE; CTR3(KTR_CALLOUT, "cancelled %p func %p arg %p", c, c->c_func, c->c_arg); if (not_on_a_list == 0) { if ((c->c_iflags & CALLOUT_PROCESSED) == 0) { if (cc_exec_next(cc) == c) cc_exec_next(cc) = LIST_NEXT(c, c_links.le); LIST_REMOVE(c, c_links.le); } else { TAILQ_REMOVE(&cc->cc_expireq, c, c_links.tqe); } } callout_cc_del(c, cc); CC_UNLOCK(cc); return (1); } void callout_init(struct callout *c, int mpsafe) { bzero(c, sizeof *c); if (mpsafe) { c->c_lock = NULL; c->c_iflags = CALLOUT_RETURNUNLOCKED; } else { c->c_lock = &Giant.lock_object; c->c_iflags = 0; } c->c_cpu = timeout_cpu; } void _callout_init_lock(struct callout *c, struct lock_object *lock, int flags) { bzero(c, sizeof *c); c->c_lock = lock; KASSERT((flags & ~(CALLOUT_RETURNUNLOCKED | CALLOUT_SHAREDLOCK)) == 0, ("callout_init_lock: bad flags %d", flags)); KASSERT(lock != NULL || (flags & CALLOUT_RETURNUNLOCKED) == 0, ("callout_init_lock: CALLOUT_RETURNUNLOCKED with no lock")); KASSERT(lock == NULL || !(LOCK_CLASS(lock)->lc_flags & (LC_SPINLOCK | LC_SLEEPABLE)), ("%s: invalid lock class", __func__)); c->c_iflags = flags & (CALLOUT_RETURNUNLOCKED | CALLOUT_SHAREDLOCK); c->c_cpu = timeout_cpu; } #ifdef APM_FIXUP_CALLTODO /* * Adjust the kernel calltodo timeout list. This routine is used after * an APM resume to recalculate the calltodo timer list values with the * number of hz's we have been sleeping. The next hardclock() will detect * that there are fired timers and run softclock() to execute them. * * Please note, I have not done an exhaustive analysis of what code this * might break. I am motivated to have my select()'s and alarm()'s that * have expired during suspend firing upon resume so that the applications * which set the timer can do the maintanence the timer was for as close * as possible to the originally intended time. Testing this code for a * week showed that resuming from a suspend resulted in 22 to 25 timers * firing, which seemed independant on whether the suspend was 2 hours or * 2 days. Your milage may vary. - Ken Key */ void adjust_timeout_calltodo(struct timeval *time_change) { register struct callout *p; unsigned long delta_ticks; /* * How many ticks were we asleep? * (stolen from tvtohz()). */ /* Don't do anything */ if (time_change->tv_sec < 0) return; else if (time_change->tv_sec <= LONG_MAX / 1000000) delta_ticks = (time_change->tv_sec * 1000000 + time_change->tv_usec + (tick - 1)) / tick + 1; else if (time_change->tv_sec <= LONG_MAX / hz) delta_ticks = time_change->tv_sec * hz + (time_change->tv_usec + (tick - 1)) / tick + 1; else delta_ticks = LONG_MAX; if (delta_ticks > INT_MAX) delta_ticks = INT_MAX; /* * Now rip through the timer calltodo list looking for timers * to expire. */ /* don't collide with softclock() */ CC_LOCK(cc); for (p = calltodo.c_next; p != NULL; p = p->c_next) { p->c_time -= delta_ticks; /* Break if the timer had more time on it than delta_ticks */ if (p->c_time > 0) break; /* take back the ticks the timer didn't use (p->c_time <= 0) */ delta_ticks = -p->c_time; } CC_UNLOCK(cc); return; } #endif /* APM_FIXUP_CALLTODO */ static int flssbt(sbintime_t sbt) { sbt += (uint64_t)sbt >> 1; if (sizeof(long) >= sizeof(sbintime_t)) return (flsl(sbt)); if (sbt >= SBT_1S) return (flsl(((uint64_t)sbt) >> 32) + 32); return (flsl(sbt)); } /* * Dump immediate statistic snapshot of the scheduled callouts. */ static int sysctl_kern_callout_stat(SYSCTL_HANDLER_ARGS) { struct callout *tmp; struct callout_cpu *cc; struct callout_list *sc; sbintime_t maxpr, maxt, medpr, medt, now, spr, st, t; int ct[64], cpr[64], ccpbk[32]; int error, val, i, count, tcum, pcum, maxc, c, medc; #ifdef SMP int cpu; #endif val = 0; error = sysctl_handle_int(oidp, &val, 0, req); if (error != 0 || req->newptr == NULL) return (error); count = maxc = 0; st = spr = maxt = maxpr = 0; bzero(ccpbk, sizeof(ccpbk)); bzero(ct, sizeof(ct)); bzero(cpr, sizeof(cpr)); now = sbinuptime(); #ifdef SMP CPU_FOREACH(cpu) { cc = CC_CPU(cpu); #else cc = CC_CPU(timeout_cpu); #endif CC_LOCK(cc); for (i = 0; i < callwheelsize; i++) { sc = &cc->cc_callwheel[i]; c = 0; LIST_FOREACH(tmp, sc, c_links.le) { c++; t = tmp->c_time - now; if (t < 0) t = 0; st += t / SBT_1US; spr += tmp->c_precision / SBT_1US; if (t > maxt) maxt = t; if (tmp->c_precision > maxpr) maxpr = tmp->c_precision; ct[flssbt(t)]++; cpr[flssbt(tmp->c_precision)]++; } if (c > maxc) maxc = c; ccpbk[fls(c + c / 2)]++; count += c; } CC_UNLOCK(cc); #ifdef SMP } #endif for (i = 0, tcum = 0; i < 64 && tcum < count / 2; i++) tcum += ct[i]; medt = (i >= 2) ? (((sbintime_t)1) << (i - 2)) : 0; for (i = 0, pcum = 0; i < 64 && pcum < count / 2; i++) pcum += cpr[i]; medpr = (i >= 2) ? (((sbintime_t)1) << (i - 2)) : 0; for (i = 0, c = 0; i < 32 && c < count / 2; i++) c += ccpbk[i]; medc = (i >= 2) ? (1 << (i - 2)) : 0; printf("Scheduled callouts statistic snapshot:\n"); printf(" Callouts: %6d Buckets: %6d*%-3d Bucket size: 0.%06ds\n", count, callwheelsize, mp_ncpus, 1000000 >> CC_HASH_SHIFT); printf(" C/Bk: med %5d avg %6d.%06jd max %6d\n", medc, count / callwheelsize / mp_ncpus, (uint64_t)count * 1000000 / callwheelsize / mp_ncpus % 1000000, maxc); printf(" Time: med %5jd.%06jds avg %6jd.%06jds max %6jd.%06jds\n", medt / SBT_1S, (medt & 0xffffffff) * 1000000 >> 32, (st / count) / 1000000, (st / count) % 1000000, maxt / SBT_1S, (maxt & 0xffffffff) * 1000000 >> 32); printf(" Prec: med %5jd.%06jds avg %6jd.%06jds max %6jd.%06jds\n", medpr / SBT_1S, (medpr & 0xffffffff) * 1000000 >> 32, (spr / count) / 1000000, (spr / count) % 1000000, maxpr / SBT_1S, (maxpr & 0xffffffff) * 1000000 >> 32); printf(" Distribution: \tbuckets\t time\t tcum\t" " prec\t pcum\n"); for (i = 0, tcum = pcum = 0; i < 64; i++) { if (ct[i] == 0 && cpr[i] == 0) continue; t = (i != 0) ? (((sbintime_t)1) << (i - 1)) : 0; tcum += ct[i]; pcum += cpr[i]; printf(" %10jd.%06jds\t 2**%d\t%7d\t%7d\t%7d\t%7d\n", t / SBT_1S, (t & 0xffffffff) * 1000000 >> 32, i - 1 - (32 - CC_HASH_SHIFT), ct[i], tcum, cpr[i], pcum); } return (error); } SYSCTL_PROC(_kern, OID_AUTO, callout_stat, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0, sysctl_kern_callout_stat, "I", "Dump immediate statistic snapshot of the scheduled callouts"); + +#ifdef DDB + +static void +_show_callout(struct callout *c) +{ + + db_printf("callout %p\n", c); +#define C_DB_PRINTF(f, e) db_printf(" %s = " f "\n", #e, c->e); + db_printf(" &c_links = %p\n", &(c->c_links)); + C_DB_PRINTF("%" PRId64, c_time); + C_DB_PRINTF("%" PRId64, c_precision); + C_DB_PRINTF("%p", c_arg); + C_DB_PRINTF("%p", c_func); + C_DB_PRINTF("%p", c_lock); + C_DB_PRINTF("%#x", c_flags); + C_DB_PRINTF("%#x", c_iflags); + C_DB_PRINTF("%d", c_cpu); +#undef C_DB_PRINTF +} + +DB_SHOW_COMMAND(callout, db_show_callout) +{ + + if (!have_addr) { + db_printf("usage: show callout \n"); + return; + } + + _show_callout((struct callout *)addr); +} +#endif /* DDB */ Index: projects/vnet/sys/kern/uipc_domain.c =================================================================== --- projects/vnet/sys/kern/uipc_domain.c (revision 294565) +++ projects/vnet/sys/kern/uipc_domain.c (revision 294566) @@ -1,524 +1,528 @@ /*- * Copyright (c) 1982, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)uipc_domain.c 8.2 (Berkeley) 10/18/93 */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include /* * System initialization * * Note: domain initialization takes place on a per domain basis * as a result of traversing a SYSINIT linker set. Most likely, * each domain would want to call DOMAIN_SET(9) itself, which * would cause the domain to be added just after domaininit() * is called during startup. * * See DOMAIN_SET(9) for details on its use. */ static void domaininit(void *); SYSINIT(domain, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, domaininit, NULL); static void domainfinalize(void *); SYSINIT(domainfin, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST, domainfinalize, NULL); static struct callout pffast_callout; static struct callout pfslow_callout; static void pffasttimo(void *); static void pfslowtimo(void *); struct domain *domains; /* registered protocol domains */ int domain_init_status = 0; static struct mtx dom_mtx; /* domain list lock */ MTX_SYSINIT(domain, &dom_mtx, "domain list", MTX_DEF); /* * Dummy protocol specific user requests function pointer array. * All functions return EOPNOTSUPP. */ struct pr_usrreqs nousrreqs = { .pru_accept = pru_accept_notsupp, .pru_attach = pru_attach_notsupp, .pru_bind = pru_bind_notsupp, .pru_connect = pru_connect_notsupp, .pru_connect2 = pru_connect2_notsupp, .pru_control = pru_control_notsupp, .pru_disconnect = pru_disconnect_notsupp, .pru_listen = pru_listen_notsupp, .pru_peeraddr = pru_peeraddr_notsupp, .pru_rcvd = pru_rcvd_notsupp, .pru_rcvoob = pru_rcvoob_notsupp, .pru_send = pru_send_notsupp, .pru_sense = pru_sense_null, .pru_shutdown = pru_shutdown_notsupp, .pru_sockaddr = pru_sockaddr_notsupp, .pru_sosend = pru_sosend_notsupp, .pru_soreceive = pru_soreceive_notsupp, .pru_sopoll = pru_sopoll_notsupp, }; static void protosw_init(struct protosw *pr) { struct pr_usrreqs *pu; pu = pr->pr_usrreqs; KASSERT(pu != NULL, ("protosw_init: %ssw[%d] has no usrreqs!", pr->pr_domain->dom_name, (int)(pr - pr->pr_domain->dom_protosw))); /* * Protocol switch methods fall into three categories: mandatory, * mandatory but protosw_init() provides a default, and optional. * * For true protocols (i.e., pru_attach != NULL), KASSERT truly * mandatory methods with no defaults, and initialize defaults for * other mandatory methods if the protocol hasn't defined an * implementation (NULL function pointer). */ #if 0 if (pu->pru_attach != NULL) { KASSERT(pu->pru_abort != NULL, ("protosw_init: %ssw[%d] pru_abort NULL", pr->pr_domain->dom_name, (int)(pr - pr->pr_domain->dom_protosw))); KASSERT(pu->pru_send != NULL, ("protosw_init: %ssw[%d] pru_send NULL", pr->pr_domain->dom_name, (int)(pr - pr->pr_domain->dom_protosw))); } #endif #define DEFAULT(foo, bar) if ((foo) == NULL) (foo) = (bar) DEFAULT(pu->pru_accept, pru_accept_notsupp); DEFAULT(pu->pru_bind, pru_bind_notsupp); DEFAULT(pu->pru_bindat, pru_bindat_notsupp); DEFAULT(pu->pru_connect, pru_connect_notsupp); DEFAULT(pu->pru_connect2, pru_connect2_notsupp); DEFAULT(pu->pru_connectat, pru_connectat_notsupp); DEFAULT(pu->pru_control, pru_control_notsupp); DEFAULT(pu->pru_disconnect, pru_disconnect_notsupp); DEFAULT(pu->pru_listen, pru_listen_notsupp); DEFAULT(pu->pru_peeraddr, pru_peeraddr_notsupp); DEFAULT(pu->pru_rcvd, pru_rcvd_notsupp); DEFAULT(pu->pru_rcvoob, pru_rcvoob_notsupp); DEFAULT(pu->pru_sense, pru_sense_null); DEFAULT(pu->pru_shutdown, pru_shutdown_notsupp); DEFAULT(pu->pru_sockaddr, pru_sockaddr_notsupp); DEFAULT(pu->pru_sosend, sosend_generic); DEFAULT(pu->pru_soreceive, soreceive_generic); DEFAULT(pu->pru_sopoll, sopoll_generic); DEFAULT(pu->pru_ready, pru_ready_notsupp); #undef DEFAULT if (pr->pr_init) (*pr->pr_init)(); } /* * Add a new protocol domain to the list of supported domains * Note: you cant unload it again because a socket may be using it. * XXX can't fail at this time. */ void domain_init(void *arg) { struct domain *dp = arg; struct protosw *pr; if (dp->dom_init) (*dp->dom_init)(); for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) protosw_init(pr); /* * update global information about maximums */ max_hdr = max_linkhdr + max_protohdr; max_datalen = MHLEN - max_hdr; if (max_datalen < 1) panic("%s: max_datalen < 1", __func__); } #ifdef VIMAGE void vnet_domain_init(void *arg) { /* Virtualized case is no different -- call init functions. */ domain_init(arg); } void vnet_domain_uninit(void *arg) { struct domain *dp = arg; struct protosw *pr; for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) - if (pr->pr_destroy) + if (pr->pr_destroy) { +#ifdef INVARIANTS + printf("%s: pr %p called pr_destroy\n", __func__, pr); +#endif (*pr->pr_destroy)(); + } if (dp->dom_destroy) (*dp->dom_destroy)(); } #endif /* * Add a new protocol domain to the list of supported domains * Note: you cant unload it again because a socket may be using it. * XXX can't fail at this time. */ void domain_add(void *data) { struct domain *dp; dp = (struct domain *)data; mtx_lock(&dom_mtx); dp->dom_next = domains; domains = dp; KASSERT(domain_init_status >= 1, ("attempt to domain_add(%s) before domaininit()", dp->dom_name)); #ifndef INVARIANTS if (domain_init_status < 1) printf("WARNING: attempt to domain_add(%s) before " "domaininit()\n", dp->dom_name); #endif #ifdef notyet KASSERT(domain_init_status < 2, ("attempt to domain_add(%s) after domainfinalize()", dp->dom_name)); #else if (domain_init_status >= 2) printf("WARNING: attempt to domain_add(%s) after " "domainfinalize()\n", dp->dom_name); #endif mtx_unlock(&dom_mtx); } /* ARGSUSED*/ static void domaininit(void *dummy) { if (max_linkhdr < 16) /* XXX */ max_linkhdr = 16; callout_init(&pffast_callout, 1); callout_init(&pfslow_callout, 1); mtx_lock(&dom_mtx); KASSERT(domain_init_status == 0, ("domaininit called too late!")); domain_init_status = 1; mtx_unlock(&dom_mtx); } /* ARGSUSED*/ static void domainfinalize(void *dummy) { mtx_lock(&dom_mtx); KASSERT(domain_init_status == 1, ("domainfinalize called too late!")); domain_init_status = 2; mtx_unlock(&dom_mtx); callout_reset(&pffast_callout, 1, pffasttimo, NULL); callout_reset(&pfslow_callout, 1, pfslowtimo, NULL); } struct domain * pffinddomain(int family) { struct domain *dp; for (dp = domains; dp != NULL; dp = dp->dom_next) if (dp->dom_family == family) return (dp); return (NULL); } struct protosw * pffindtype(int family, int type) { struct domain *dp; struct protosw *pr; dp = pffinddomain(family); if (dp == NULL) return (NULL); for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) if (pr->pr_type && pr->pr_type == type) return (pr); return (NULL); } struct protosw * pffindproto(int family, int protocol, int type) { struct domain *dp; struct protosw *pr; struct protosw *maybe; maybe = NULL; if (family == 0) return (NULL); dp = pffinddomain(family); if (dp == NULL) return (NULL); for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { if ((pr->pr_protocol == protocol) && (pr->pr_type == type)) return (pr); if (type == SOCK_RAW && pr->pr_type == SOCK_RAW && pr->pr_protocol == 0 && maybe == NULL) maybe = pr; } return (maybe); } /* * The caller must make sure that the new protocol is fully set up and ready to * accept requests before it is registered. */ int pf_proto_register(int family, struct protosw *npr) { VNET_ITERATOR_DECL(vnet_iter); struct domain *dp; struct protosw *pr, *fpr; /* Sanity checks. */ if (family == 0) return (EPFNOSUPPORT); if (npr->pr_type == 0) return (EPROTOTYPE); if (npr->pr_protocol == 0) return (EPROTONOSUPPORT); if (npr->pr_usrreqs == NULL) return (ENXIO); /* Try to find the specified domain based on the family. */ dp = pffinddomain(family); if (dp == NULL) return (EPFNOSUPPORT); /* Initialize backpointer to struct domain. */ npr->pr_domain = dp; fpr = NULL; /* * Protect us against races when two protocol registrations for * the same protocol happen at the same time. */ mtx_lock(&dom_mtx); /* The new protocol must not yet exist. */ for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { if ((pr->pr_type == npr->pr_type) && (pr->pr_protocol == npr->pr_protocol)) { mtx_unlock(&dom_mtx); return (EEXIST); /* XXX: Check only protocol? */ } /* While here, remember the first free spacer. */ if ((fpr == NULL) && (pr->pr_protocol == PROTO_SPACER)) fpr = pr; } /* If no free spacer is found we can't add the new protocol. */ if (fpr == NULL) { mtx_unlock(&dom_mtx); return (ENOMEM); } /* Copy the new struct protosw over the spacer. */ bcopy(npr, fpr, sizeof(*fpr)); /* Job is done, no more protection required. */ mtx_unlock(&dom_mtx); /* Initialize and activate the protocol. */ VNET_LIST_RLOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET_QUIET(vnet_iter); protosw_init(fpr); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK(); return (0); } /* * The caller must make sure the protocol and its functions correctly shut down * all sockets and release all locks and memory references. */ int pf_proto_unregister(int family, int protocol, int type) { struct domain *dp; struct protosw *pr, *dpr; /* Sanity checks. */ if (family == 0) return (EPFNOSUPPORT); if (protocol == 0) return (EPROTONOSUPPORT); if (type == 0) return (EPROTOTYPE); /* Try to find the specified domain based on the family type. */ dp = pffinddomain(family); if (dp == NULL) return (EPFNOSUPPORT); dpr = NULL; /* Lock out everyone else while we are manipulating the protosw. */ mtx_lock(&dom_mtx); /* The protocol must exist and only once. */ for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) { if ((pr->pr_type == type) && (pr->pr_protocol == protocol)) { if (dpr != NULL) { mtx_unlock(&dom_mtx); return (EMLINK); /* Should not happen! */ } else dpr = pr; } } /* Protocol does not exist. */ if (dpr == NULL) { mtx_unlock(&dom_mtx); return (EPROTONOSUPPORT); } /* De-orbit the protocol and make the slot available again. */ dpr->pr_type = 0; dpr->pr_domain = dp; dpr->pr_protocol = PROTO_SPACER; dpr->pr_flags = 0; dpr->pr_input = NULL; dpr->pr_output = NULL; dpr->pr_ctlinput = NULL; dpr->pr_ctloutput = NULL; dpr->pr_init = NULL; dpr->pr_fasttimo = NULL; dpr->pr_slowtimo = NULL; dpr->pr_drain = NULL; dpr->pr_usrreqs = &nousrreqs; /* Job is done, not more protection required. */ mtx_unlock(&dom_mtx); return (0); } void pfctlinput(int cmd, struct sockaddr *sa) { struct domain *dp; struct protosw *pr; for (dp = domains; dp; dp = dp->dom_next) for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) if (pr->pr_ctlinput) (*pr->pr_ctlinput)(cmd, sa, (void *)0); } void pfctlinput2(int cmd, struct sockaddr *sa, void *ctlparam) { struct domain *dp; struct protosw *pr; if (!sa) return; for (dp = domains; dp; dp = dp->dom_next) { /* * the check must be made by xx_ctlinput() anyways, to * make sure we use data item pointed to by ctlparam in * correct way. the following check is made just for safety. */ if (dp->dom_family != sa->sa_family) continue; for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) if (pr->pr_ctlinput) (*pr->pr_ctlinput)(cmd, sa, ctlparam); } } static void pfslowtimo(void *arg) { struct domain *dp; struct protosw *pr; for (dp = domains; dp; dp = dp->dom_next) for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) if (pr->pr_slowtimo) (*pr->pr_slowtimo)(); callout_reset(&pfslow_callout, hz/2, pfslowtimo, NULL); } static void pffasttimo(void *arg) { struct domain *dp; struct protosw *pr; for (dp = domains; dp; dp = dp->dom_next) for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) if (pr->pr_fasttimo) (*pr->pr_fasttimo)(); callout_reset(&pffast_callout, hz/5, pffasttimo, NULL); } Index: projects/vnet/sys/net/if.c =================================================================== --- projects/vnet/sys/net/if.c (revision 294565) +++ projects/vnet/sys/net/if.c (revision 294566) @@ -1,3955 +1,3980 @@ /*- * Copyright (c) 1980, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)if.c 8.5 (Berkeley) 1/9/95 * $FreeBSD$ */ #include "opt_compat.h" #include "opt_inet6.h" #include "opt_inet.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if defined(INET) || defined(INET6) #include #include #include #include #include #ifdef INET #include #endif /* INET */ #ifdef INET6 #include #include #endif /* INET6 */ #endif /* INET || INET6 */ #include #ifdef COMPAT_FREEBSD32 #include #include #endif SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN, &ifqmaxlen, 0, "max send queue size"); /* Log link state change events */ static int log_link_state_change = 1; SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW, &log_link_state_change, 0, "log interface link state change events"); /* Interface description */ static unsigned int ifdescr_maxlen = 1024; SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW, &ifdescr_maxlen, 0, "administrative maximum length for interface description"); static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions"); /* global sx for non-critical path ifdescr */ static struct sx ifdescr_sx; SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr"); void (*bridge_linkstate_p)(struct ifnet *ifp); void (*ng_ether_link_state_p)(struct ifnet *ifp, int state); void (*lagg_linkstate_p)(struct ifnet *ifp, int state); /* These are external hooks for CARP. */ void (*carp_linkstate_p)(struct ifnet *ifp); void (*carp_demote_adj_p)(int, char *); int (*carp_master_p)(struct ifaddr *); #if defined(INET) || defined(INET6) int (*carp_forus_p)(struct ifnet *ifp, u_char *dhost); int (*carp_output_p)(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa); int (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *); int (*carp_attach_p)(struct ifaddr *, int); void (*carp_detach_p)(struct ifaddr *); #endif #ifdef INET int (*carp_iamatch_p)(struct ifaddr *, uint8_t **); #endif #ifdef INET6 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6); caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m, const struct in6_addr *taddr); #endif struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; /* * XXX: Style; these should be sorted alphabetically, and unprototyped * static functions should be prototyped. Currently they are sorted by * declaration order. */ static void if_attachdomain(void *); static void if_attachdomain1(struct ifnet *); static int ifconf(u_long, caddr_t); static void if_freemulti(struct ifmultiaddr *); static void if_grow(void); static void if_input_default(struct ifnet *, struct mbuf *); static void if_route(struct ifnet *, int flag, int fam); static int if_setflag(struct ifnet *, int, int, int *, int); static int if_transmit(struct ifnet *ifp, struct mbuf *m); static void if_unroute(struct ifnet *, int flag, int fam); static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); static int if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int); static void do_link_state_change(void *, int); static int if_getgroup(struct ifgroupreq *, struct ifnet *); static int if_getgroupmembers(struct ifgroupreq *); static void if_delgroups(struct ifnet *); static void if_attach_internal(struct ifnet *, int, struct if_clone *); static int if_detach_internal(struct ifnet *, int, struct if_clone **); #ifdef INET6 /* * XXX: declare here to avoid to include many inet6 related files.. * should be more generalized? */ extern void nd6_setmtu(struct ifnet *); #endif /* ipsec helper hooks */ VNET_DEFINE(struct hhook_head *, ipsec_hhh_in[HHOOK_IPSEC_COUNT]); VNET_DEFINE(struct hhook_head *, ipsec_hhh_out[HHOOK_IPSEC_COUNT]); VNET_DEFINE(int, if_index); int ifqmaxlen = IFQ_MAXLEN; VNET_DEFINE(struct ifnethead, ifnet); /* depend on static init XXX */ VNET_DEFINE(struct ifgrouphead, ifg_head); static VNET_DEFINE(int, if_indexlim) = 8; /* Table of ifnet by index. */ VNET_DEFINE(struct ifnet **, ifindex_table); #define V_if_indexlim VNET(if_indexlim) #define V_ifindex_table VNET(ifindex_table) /* * The global network interface list (V_ifnet) and related state (such as * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and * an rwlock. Either may be acquired shared to stablize the list, but both * must be acquired writable to modify the list. This model allows us to * both stablize the interface list during interrupt thread processing, but * also to stablize it over long-running ioctls, without introducing priority * inversions and deadlocks. */ struct rwlock ifnet_rwlock; RW_SYSINIT_FLAGS(ifnet_rw, &ifnet_rwlock, "ifnet_rw", RW_RECURSE); struct sx ifnet_sxlock; SX_SYSINIT_FLAGS(ifnet_sx, &ifnet_sxlock, "ifnet_sx", SX_RECURSE); /* * The allocation of network interfaces is a rather non-atomic affair; we * need to select an index before we are ready to expose the interface for * use, so will use this pointer value to indicate reservation. */ #define IFNET_HOLD (void *)(uintptr_t)(-1) static if_com_alloc_t *if_com_alloc[256]; static if_com_free_t *if_com_free[256]; static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals"); MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); struct ifnet * ifnet_byindex_locked(u_short idx) { if (idx > V_if_index) return (NULL); if (V_ifindex_table[idx] == IFNET_HOLD) return (NULL); return (V_ifindex_table[idx]); } struct ifnet * ifnet_byindex(u_short idx) { struct ifnet *ifp; IFNET_RLOCK_NOSLEEP(); ifp = ifnet_byindex_locked(idx); IFNET_RUNLOCK_NOSLEEP(); return (ifp); } struct ifnet * ifnet_byindex_ref(u_short idx) { struct ifnet *ifp; IFNET_RLOCK_NOSLEEP(); ifp = ifnet_byindex_locked(idx); if (ifp == NULL || (ifp->if_flags & IFF_DYING)) { IFNET_RUNLOCK_NOSLEEP(); return (NULL); } if_ref(ifp); IFNET_RUNLOCK_NOSLEEP(); return (ifp); } /* * Allocate an ifindex array entry; return 0 on success or an error on * failure. */ static u_short ifindex_alloc(void) { u_short idx; IFNET_WLOCK_ASSERT(); retry: /* * Try to find an empty slot below V_if_index. If we fail, take the * next slot. */ for (idx = 1; idx <= V_if_index; idx++) { if (V_ifindex_table[idx] == NULL) break; } /* Catch if_index overflow. */ if (idx >= V_if_indexlim) { if_grow(); goto retry; } if (idx > V_if_index) V_if_index = idx; return (idx); } static void ifindex_free_locked(u_short idx) { IFNET_WLOCK_ASSERT(); V_ifindex_table[idx] = NULL; while (V_if_index > 0 && V_ifindex_table[V_if_index] == NULL) V_if_index--; } static void ifindex_free(u_short idx) { IFNET_WLOCK(); ifindex_free_locked(idx); IFNET_WUNLOCK(); } static void ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp) { IFNET_WLOCK_ASSERT(); V_ifindex_table[idx] = ifp; } static void ifnet_setbyindex(u_short idx, struct ifnet *ifp) { IFNET_WLOCK(); ifnet_setbyindex_locked(idx, ifp); IFNET_WUNLOCK(); } struct ifaddr * ifaddr_byindex(u_short idx) { struct ifnet *ifp; struct ifaddr *ifa = NULL; IFNET_RLOCK_NOSLEEP(); ifp = ifnet_byindex_locked(idx); if (ifp != NULL && (ifa = ifp->if_addr) != NULL) ifa_ref(ifa); IFNET_RUNLOCK_NOSLEEP(); return (ifa); } /* * Network interface utility routines. * * Routines with ifa_ifwith* names take sockaddr *'s as * parameters. */ static void vnet_if_init(const void *unused __unused) { TAILQ_INIT(&V_ifnet); TAILQ_INIT(&V_ifg_head); IFNET_WLOCK(); if_grow(); /* create initial table */ IFNET_WUNLOCK(); vnet_if_clone_init(); } VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init, NULL); #ifdef VIMAGE static void vnet_if_uninit(const void *unused __unused) { VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p " "not empty", __func__, __LINE__, &V_ifnet)); VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p " "not empty", __func__, __LINE__, &V_ifg_head)); free((caddr_t)V_ifindex_table, M_IFNET); } VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST, vnet_if_uninit, NULL); + +/* + * XXX-BZ VNET; probably along with dom stuff. + * This is very wrong but MC currently implies that interfaces are + * gone before we can free it. This needs to be fied differently + * and this needs to be moved back to SI_SUB_INIT_IF. + */ +static void +vnet_if_return(const void *unused __unused) +{ + struct ifnet *ifp, *nifp; + + /* Return all inherited interfaces to their parent vnets. */ + TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { + if (ifp->if_home_vnet != ifp->if_vnet) + if_vmove(ifp, ifp->if_home_vnet); + } +} +VNET_SYSUNINIT(vnet_if_return, SI_SUB_VNET_DONE, SI_ORDER_ANY, + vnet_if_return, NULL); #endif static void if_grow(void) { int oldlim; u_int n; struct ifnet **e; IFNET_WLOCK_ASSERT(); oldlim = V_if_indexlim; IFNET_WUNLOCK(); n = (oldlim << 1) * sizeof(*e); e = malloc(n, M_IFNET, M_WAITOK | M_ZERO); IFNET_WLOCK(); if (V_if_indexlim != oldlim) { free(e, M_IFNET); return; } if (V_ifindex_table != NULL) { memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2); free((caddr_t)V_ifindex_table, M_IFNET); } V_if_indexlim <<= 1; V_ifindex_table = e; } /* * Allocate a struct ifnet and an index for an interface. A layer 2 * common structure will also be allocated if an allocation routine is * registered for the passed type. */ struct ifnet * if_alloc(u_char type) { struct ifnet *ifp; u_short idx; ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO); IFNET_WLOCK(); idx = ifindex_alloc(); ifnet_setbyindex_locked(idx, IFNET_HOLD); IFNET_WUNLOCK(); ifp->if_index = idx; ifp->if_type = type; ifp->if_alloctype = type; if (if_com_alloc[type] != NULL) { ifp->if_l2com = if_com_alloc[type](type, ifp); if (ifp->if_l2com == NULL) { free(ifp, M_IFNET); ifindex_free(idx); return (NULL); } } IF_ADDR_LOCK_INIT(ifp); TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp); ifp->if_afdata_initialized = 0; IF_AFDATA_LOCK_INIT(ifp); TAILQ_INIT(&ifp->if_addrhead); TAILQ_INIT(&ifp->if_multiaddrs); TAILQ_INIT(&ifp->if_groups); #ifdef MAC mac_ifnet_init(ifp); #endif ifq_init(&ifp->if_snd, ifp); refcount_init(&ifp->if_refcount, 1); /* Index reference. */ for (int i = 0; i < IFCOUNTERS; i++) ifp->if_counters[i] = counter_u64_alloc(M_WAITOK); ifp->if_get_counter = if_get_counter_default; ifnet_setbyindex(ifp->if_index, ifp); return (ifp); } /* * Do the actual work of freeing a struct ifnet, and layer 2 common * structure. This call is made when the last reference to an * interface is released. */ static void if_free_internal(struct ifnet *ifp) { KASSERT((ifp->if_flags & IFF_DYING), ("if_free_internal: interface not dying")); if (if_com_free[ifp->if_alloctype] != NULL) if_com_free[ifp->if_alloctype](ifp->if_l2com, ifp->if_alloctype); #ifdef MAC mac_ifnet_destroy(ifp); #endif /* MAC */ if (ifp->if_description != NULL) free(ifp->if_description, M_IFDESCR); IF_AFDATA_DESTROY(ifp); IF_ADDR_LOCK_DESTROY(ifp); ifq_delete(&ifp->if_snd); for (int i = 0; i < IFCOUNTERS; i++) counter_u64_free(ifp->if_counters[i]); free(ifp, M_IFNET); } /* * Deregister an interface and free the associated storage. */ void if_free(struct ifnet *ifp) { ifp->if_flags |= IFF_DYING; /* XXX: Locking */ CURVNET_SET_QUIET(ifp->if_vnet); IFNET_WLOCK(); KASSERT(ifp == ifnet_byindex_locked(ifp->if_index), ("%s: freeing unallocated ifnet", ifp->if_xname)); ifindex_free_locked(ifp->if_index); IFNET_WUNLOCK(); if (refcount_release(&ifp->if_refcount)) if_free_internal(ifp); CURVNET_RESTORE(); } /* * Interfaces to keep an ifnet type-stable despite the possibility of the * driver calling if_free(). If there are additional references, we defer * freeing the underlying data structure. */ void if_ref(struct ifnet *ifp) { /* We don't assert the ifnet list lock here, but arguably should. */ refcount_acquire(&ifp->if_refcount); } void if_rele(struct ifnet *ifp) { if (!refcount_release(&ifp->if_refcount)) return; if_free_internal(ifp); } void ifq_init(struct ifaltq *ifq, struct ifnet *ifp) { mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); if (ifq->ifq_maxlen == 0) ifq->ifq_maxlen = ifqmaxlen; ifq->altq_type = 0; ifq->altq_disc = NULL; ifq->altq_flags &= ALTQF_CANTCHANGE; ifq->altq_tbr = NULL; ifq->altq_ifp = ifp; } void ifq_delete(struct ifaltq *ifq) { mtx_destroy(&ifq->ifq_mtx); } /* * Perform generic interface initalization tasks and attach the interface * to the list of "active" interfaces. If vmove flag is set on entry * to if_attach_internal(), perform only a limited subset of initialization * tasks, given that we are moving from one vnet to another an ifnet which * has already been fully initialized. * * Note that if_detach_internal() removes group membership unconditionally * even when vmove flag is set, and if_attach_internal() adds only IFG_ALL. * Thus, when if_vmove() is applied to a cloned interface, group membership * is lost while a cloned one always joins a group whose name is * ifc->ifc_name. To recover this after if_detach_internal() and * if_attach_internal(), the cloner should be specified to * if_attach_internal() via ifc. If it is non-NULL, if_attach_internal() * attempts to join a group whose name is ifc->ifc_name. * * XXX: * - The decision to return void and thus require this function to * succeed is questionable. * - We should probably do more sanity checking. For instance we don't * do anything to insure if_xname is unique or non-empty. */ void if_attach(struct ifnet *ifp) { if_attach_internal(ifp, 0, NULL); } /* * Compute the least common TSO limit. */ void if_hw_tsomax_common(if_t ifp, struct ifnet_hw_tsomax *pmax) { /* * 1) If there is no limit currently, take the limit from * the network adapter. * * 2) If the network adapter has a limit below the current * limit, apply it. */ if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 && ifp->if_hw_tsomax < pmax->tsomaxbytes)) { pmax->tsomaxbytes = ifp->if_hw_tsomax; } if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 && ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) { pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; } if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 && ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) { pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; } } /* * Update TSO limit of a network adapter. * * Returns zero if no change. Else non-zero. */ int if_hw_tsomax_update(if_t ifp, struct ifnet_hw_tsomax *pmax) { int retval = 0; if (ifp->if_hw_tsomax != pmax->tsomaxbytes) { ifp->if_hw_tsomax = pmax->tsomaxbytes; retval++; } if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) { ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize; retval++; } if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) { ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount; retval++; } return (retval); } static void if_attach_internal(struct ifnet *ifp, int vmove, struct if_clone *ifc) { unsigned socksize, ifasize; int namelen, masklen; struct sockaddr_dl *sdl; struct ifaddr *ifa; if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index)) panic ("%s: BUG: if_attach called without if_alloc'd input()\n", ifp->if_xname); #ifdef VIMAGE ifp->if_vnet = curvnet; if (ifp->if_home_vnet == NULL) ifp->if_home_vnet = curvnet; #endif if_addgroup(ifp, IFG_ALL); /* Restore group membership for cloned interfaces. */ if (vmove && ifc != NULL) if_clone_addgroup(ifp, ifc); getmicrotime(&ifp->if_lastchange); ifp->if_epoch = time_uptime; KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) || (ifp->if_transmit != NULL && ifp->if_qflush != NULL), ("transmit and qflush must both either be set or both be NULL")); if (ifp->if_transmit == NULL) { ifp->if_transmit = if_transmit; ifp->if_qflush = if_qflush; } if (ifp->if_input == NULL) ifp->if_input = if_input_default; if (!vmove) { #ifdef MAC mac_ifnet_create(ifp); #endif /* * Create a Link Level name for this device. */ namelen = strlen(ifp->if_xname); /* * Always save enough space for any possiable name so we * can do a rename in place later. */ masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; socksize = masklen + ifp->if_addrlen; if (socksize < sizeof(*sdl)) socksize = sizeof(*sdl); socksize = roundup2(socksize, sizeof(long)); ifasize = sizeof(*ifa) + 2 * socksize; ifa = ifa_alloc(ifasize, M_WAITOK); sdl = (struct sockaddr_dl *)(ifa + 1); sdl->sdl_len = socksize; sdl->sdl_family = AF_LINK; bcopy(ifp->if_xname, sdl->sdl_data, namelen); sdl->sdl_nlen = namelen; sdl->sdl_index = ifp->if_index; sdl->sdl_type = ifp->if_type; ifp->if_addr = ifa; ifa->ifa_ifp = ifp; ifa->ifa_rtrequest = link_rtrequest; ifa->ifa_addr = (struct sockaddr *)sdl; sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); ifa->ifa_netmask = (struct sockaddr *)sdl; sdl->sdl_len = masklen; while (namelen != 0) sdl->sdl_data[--namelen] = 0xff; TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); /* Reliably crash if used uninitialized. */ ifp->if_broadcastaddr = NULL; #if defined(INET) || defined(INET6) /* Use defaults for TSO, if nothing is set */ if (ifp->if_hw_tsomax == 0 && ifp->if_hw_tsomaxsegcount == 0 && ifp->if_hw_tsomaxsegsize == 0) { /* * The TSO defaults needs to be such that an * NFS mbuf list of 35 mbufs totalling just * below 64K works and that a chain of mbufs * can be defragged into at most 32 segments: */ ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); ifp->if_hw_tsomaxsegcount = 35; ifp->if_hw_tsomaxsegsize = 2048; /* 2K */ /* XXX some drivers set IFCAP_TSO after ethernet attach */ if (ifp->if_capabilities & IFCAP_TSO) { if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n", ifp->if_hw_tsomax, ifp->if_hw_tsomaxsegcount, ifp->if_hw_tsomaxsegsize); } } #endif } #ifdef VIMAGE else { /* * Update the interface index in the link layer address * of the interface. */ for (ifa = ifp->if_addr; ifa != NULL; ifa = TAILQ_NEXT(ifa, ifa_link)) { if (ifa->ifa_addr->sa_family == AF_LINK) { sdl = (struct sockaddr_dl *)ifa->ifa_addr; sdl->sdl_index = ifp->if_index; } } } #endif IFNET_WLOCK(); TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link); #ifdef VIMAGE curvnet->vnet_ifcnt++; #endif IFNET_WUNLOCK(); if (domain_init_status >= 2) if_attachdomain1(ifp); EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); if (IS_DEFAULT_VNET(curvnet)) devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL); /* Announce the interface. */ rt_ifannouncemsg(ifp, IFAN_ARRIVAL); } static void if_attachdomain(void *dummy) { struct ifnet *ifp; TAILQ_FOREACH(ifp, &V_ifnet, if_link) if_attachdomain1(ifp); } SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND, if_attachdomain, NULL); static void if_attachdomain1(struct ifnet *ifp) { struct domain *dp; /* * Since dp->dom_ifattach calls malloc() with M_WAITOK, we * cannot lock ifp->if_afdata initialization, entirely. */ if (IF_AFDATA_TRYLOCK(ifp) == 0) return; if (ifp->if_afdata_initialized >= domain_init_status) { IF_AFDATA_UNLOCK(ifp); log(LOG_WARNING, "%s called more than once on %s\n", __func__, ifp->if_xname); return; } ifp->if_afdata_initialized = domain_init_status; IF_AFDATA_UNLOCK(ifp); /* address family dependent data region */ bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); for (dp = domains; dp; dp = dp->dom_next) { if (dp->dom_ifattach) ifp->if_afdata[dp->dom_family] = (*dp->dom_ifattach)(ifp); } } /* * Remove any unicast or broadcast network addresses from an interface. */ void if_purgeaddrs(struct ifnet *ifp) { struct ifaddr *ifa, *next; + /* XXX IF_ADDR_R/WLOCK */ TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { if (ifa->ifa_addr->sa_family == AF_LINK) continue; #ifdef INET /* XXX: Ugly!! ad hoc just for INET */ if (ifa->ifa_addr->sa_family == AF_INET) { struct ifaliasreq ifr; bzero(&ifr, sizeof(ifr)); ifr.ifra_addr = *ifa->ifa_addr; if (ifa->ifa_dstaddr) ifr.ifra_broadaddr = *ifa->ifa_dstaddr; if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, NULL) == 0) continue; } #endif /* INET */ #ifdef INET6 if (ifa->ifa_addr->sa_family == AF_INET6) { in6_purgeaddr(ifa); /* ifp_addrhead is already updated */ continue; } #endif /* INET6 */ + IF_ADDR_WLOCK(ifp); TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); + IF_ADDR_WUNLOCK(ifp); ifa_free(ifa); } } /* * Remove any multicast network addresses from an interface when an ifnet * is going away. */ static void if_purgemaddrs(struct ifnet *ifp) { struct ifmultiaddr *ifma; struct ifmultiaddr *next; IF_ADDR_WLOCK(ifp); TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) if_delmulti_locked(ifp, ifma, 1); IF_ADDR_WUNLOCK(ifp); } /* * Detach an interface, removing it from the list of "active" interfaces. * If vmove flag is set on entry to if_detach_internal(), perform only a * limited subset of cleanup tasks, given that we are moving an ifnet from * one vnet to another, where it must be fully operational. * * XXXRW: There are some significant questions about event ordering, and * how to prevent things from starting to use the interface during detach. */ void if_detach(struct ifnet *ifp) { CURVNET_SET_QUIET(ifp->if_vnet); if_detach_internal(ifp, 0, NULL); CURVNET_RESTORE(); } static int if_detach_internal(struct ifnet *ifp, int vmove, struct if_clone **ifcp) { struct ifaddr *ifa; int i; struct domain *dp; struct ifnet *iter; int found = 0; IFNET_WLOCK(); TAILQ_FOREACH(iter, &V_ifnet, if_link) if (iter == ifp) { TAILQ_REMOVE(&V_ifnet, ifp, if_link); found = 1; break; } #ifdef VIMAGE if (found) curvnet->vnet_ifcnt--; #endif IFNET_WUNLOCK(); if (!found) { /* * While we would want to panic here, we cannot * guarantee that the interface is indeed still on * the list given we don't hold locks all the way. */ return (ENOENT); #if 0 if (vmove) panic("%s: ifp=%p not on the ifnet tailq %p", __func__, ifp, &V_ifnet); else return; /* XXX this should panic as well? */ #endif } /* Check if this is a cloned interface or not. */ if (vmove && ifcp != NULL) *ifcp = if_clone_findifc(ifp); /* * Remove/wait for pending events. */ taskqueue_drain(taskqueue_swi, &ifp->if_linktask); /* * Remove routes and flush queues. */ if_down(ifp); #ifdef ALTQ if (ALTQ_IS_ENABLED(&ifp->if_snd)) altq_disable(&ifp->if_snd); if (ALTQ_IS_ATTACHED(&ifp->if_snd)) altq_detach(&ifp->if_snd); #endif if_purgeaddrs(ifp); #ifdef INET in_ifdetach(ifp); #endif #ifdef INET6 /* * Remove all IPv6 kernel structs related to ifp. This should be done * before removing routing entries below, since IPv6 interface direct * routes are expected to be removed by the IPv6-specific kernel API. * Otherwise, the kernel will detect some inconsistency and bark it. */ in6_ifdetach(ifp); #endif if_purgemaddrs(ifp); /* Announce that the interface is gone. */ rt_ifannouncemsg(ifp, IFAN_DEPARTURE); EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); if (IS_DEFAULT_VNET(curvnet)) devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL); if (!vmove) { /* * Prevent further calls into the device driver via ifnet. */ if_dead(ifp); /* * Remove link ifaddr pointer and maybe decrement if_index. * Clean up all addresses. */ ifp->if_addr = NULL; /* We can now free link ifaddr. */ if (!TAILQ_EMPTY(&ifp->if_addrhead)) { ifa = TAILQ_FIRST(&ifp->if_addrhead); + IF_ADDR_WLOCK(ifp); TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); + IF_ADDR_WUNLOCK(ifp); ifa_free(ifa); } } rt_flushifroutes(ifp); if_delgroups(ifp); /* * We cannot hold the lock over dom_ifdetach calls as they might * sleep, for example trying to drain a callout, thus open up the * theoretical race with re-attaching. */ IF_AFDATA_LOCK(ifp); i = ifp->if_afdata_initialized; ifp->if_afdata_initialized = 0; IF_AFDATA_UNLOCK(ifp); for (dp = domains; i > 0 && dp; dp = dp->dom_next) { if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) (*dp->dom_ifdetach)(ifp, ifp->if_afdata[dp->dom_family]); } return (0); } #ifdef VIMAGE /* * if_vmove() performs a limited version of if_detach() in current * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg. * An attempt is made to shrink if_index in current vnet, find an * unused if_index in target vnet and calls if_grow() if necessary, * and finally find an unused if_xname for the target vnet. */ void if_vmove(struct ifnet *ifp, struct vnet *new_vnet) { struct if_clone *ifc; int rc; /* * Detach from current vnet, but preserve LLADDR info, do not * mark as dead etc. so that the ifnet can be reattached later. * If we cannot find it, we lost the race to someone else. */ rc = if_detach_internal(ifp, 1, &ifc); if (rc != 0) return; /* * Unlink the ifnet from ifindex_table[] in current vnet, and shrink * the if_index for that vnet if possible. * * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized, * or we'd lock on one vnet and unlock on another. */ IFNET_WLOCK(); ifindex_free_locked(ifp->if_index); IFNET_WUNLOCK(); /* * Perform interface-specific reassignment tasks, if provided by * the driver. */ if (ifp->if_reassign != NULL) ifp->if_reassign(ifp, new_vnet, NULL); /* * Switch to the context of the target vnet. */ CURVNET_SET_QUIET(new_vnet); IFNET_WLOCK(); ifp->if_index = ifindex_alloc(); ifnet_setbyindex_locked(ifp->if_index, ifp); IFNET_WUNLOCK(); if_attach_internal(ifp, 1, ifc); CURVNET_RESTORE(); } /* * Move an ifnet to or from another child prison/vnet, specified by the jail id. */ static int if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid) { struct prison *pr; struct ifnet *difp; /* Try to find the prison within our visibility. */ sx_slock(&allprison_lock); pr = prison_find_child(td->td_ucred->cr_prison, jid); sx_sunlock(&allprison_lock); if (pr == NULL) return (ENXIO); prison_hold_locked(pr); mtx_unlock(&pr->pr_mtx); /* Do not try to move the iface from and to the same prison. */ if (pr->pr_vnet == ifp->if_vnet) { prison_free(pr); return (EEXIST); } /* Make sure the named iface does not exists in the dst. prison/vnet. */ /* XXX Lock interfaces to avoid races. */ CURVNET_SET_QUIET(pr->pr_vnet); difp = ifunit(ifname); CURVNET_RESTORE(); if (difp != NULL) { prison_free(pr); return (EEXIST); } /* Move the interface into the child jail/vnet. */ if_vmove(ifp, pr->pr_vnet); /* Report the new if_xname back to the userland. */ sprintf(ifname, "%s", ifp->if_xname); prison_free(pr); return (0); } static int if_vmove_reclaim(struct thread *td, char *ifname, int jid) { struct prison *pr; struct vnet *vnet_dst; struct ifnet *ifp; /* Try to find the prison within our visibility. */ sx_slock(&allprison_lock); pr = prison_find_child(td->td_ucred->cr_prison, jid); sx_sunlock(&allprison_lock); if (pr == NULL) return (ENXIO); prison_hold_locked(pr); mtx_unlock(&pr->pr_mtx); /* Make sure the named iface exists in the source prison/vnet. */ CURVNET_SET(pr->pr_vnet); ifp = ifunit(ifname); /* XXX Lock to avoid races. */ if (ifp == NULL) { CURVNET_RESTORE(); prison_free(pr); return (ENXIO); } /* Do not try to move the iface from and to the same prison. */ vnet_dst = TD_TO_VNET(td); if (vnet_dst == ifp->if_vnet) { CURVNET_RESTORE(); prison_free(pr); return (EEXIST); } /* Get interface back from child jail/vnet. */ if_vmove(ifp, vnet_dst); CURVNET_RESTORE(); /* Report the new if_xname back to the userland. */ sprintf(ifname, "%s", ifp->if_xname); prison_free(pr); return (0); } #endif /* VIMAGE */ /* * Add a group to an interface */ int if_addgroup(struct ifnet *ifp, const char *groupname) { struct ifg_list *ifgl; struct ifg_group *ifg = NULL; struct ifg_member *ifgm; int new = 0; if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' && groupname[strlen(groupname) - 1] <= '9') return (EINVAL); IFNET_WLOCK(); TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) { IFNET_WUNLOCK(); return (EEXIST); } if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP, M_NOWAIT)) == NULL) { IFNET_WUNLOCK(); return (ENOMEM); } if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member), M_TEMP, M_NOWAIT)) == NULL) { free(ifgl, M_TEMP); IFNET_WUNLOCK(); return (ENOMEM); } TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) if (!strcmp(ifg->ifg_group, groupname)) break; if (ifg == NULL) { if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group), M_TEMP, M_NOWAIT)) == NULL) { free(ifgl, M_TEMP); free(ifgm, M_TEMP); IFNET_WUNLOCK(); return (ENOMEM); } strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group)); ifg->ifg_refcnt = 0; TAILQ_INIT(&ifg->ifg_members); TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next); new = 1; } ifg->ifg_refcnt++; ifgl->ifgl_group = ifg; ifgm->ifgm_ifp = ifp; IF_ADDR_WLOCK(ifp); TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next); TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next); IF_ADDR_WUNLOCK(ifp); IFNET_WUNLOCK(); if (new) EVENTHANDLER_INVOKE(group_attach_event, ifg); EVENTHANDLER_INVOKE(group_change_event, groupname); return (0); } /* * Remove a group from an interface */ int if_delgroup(struct ifnet *ifp, const char *groupname) { struct ifg_list *ifgl; struct ifg_member *ifgm; IFNET_WLOCK(); TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) break; if (ifgl == NULL) { IFNET_WUNLOCK(); return (ENOENT); } IF_ADDR_WLOCK(ifp); TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); IF_ADDR_WUNLOCK(ifp); TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) if (ifgm->ifgm_ifp == ifp) break; if (ifgm != NULL) { TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next); free(ifgm, M_TEMP); } if (--ifgl->ifgl_group->ifg_refcnt == 0) { TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); IFNET_WUNLOCK(); EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); free(ifgl->ifgl_group, M_TEMP); } else IFNET_WUNLOCK(); free(ifgl, M_TEMP); EVENTHANDLER_INVOKE(group_change_event, groupname); return (0); } /* * Remove an interface from all groups */ static void if_delgroups(struct ifnet *ifp) { struct ifg_list *ifgl; struct ifg_member *ifgm; char groupname[IFNAMSIZ]; IFNET_WLOCK(); while (!TAILQ_EMPTY(&ifp->if_groups)) { ifgl = TAILQ_FIRST(&ifp->if_groups); strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ); IF_ADDR_WLOCK(ifp); - TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); + TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next); // <<<< IF_ADDR_WUNLOCK(ifp); TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next) if (ifgm->ifgm_ifp == ifp) break; if (ifgm != NULL) { TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next); free(ifgm, M_TEMP); } if (--ifgl->ifgl_group->ifg_refcnt == 0) { TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next); IFNET_WUNLOCK(); EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group); free(ifgl->ifgl_group, M_TEMP); } else IFNET_WUNLOCK(); free(ifgl, M_TEMP); EVENTHANDLER_INVOKE(group_change_event, groupname); IFNET_WLOCK(); } IFNET_WUNLOCK(); } /* * Stores all groups from an interface in memory pointed * to by data */ static int if_getgroup(struct ifgroupreq *data, struct ifnet *ifp) { int len, error; struct ifg_list *ifgl; struct ifg_req ifgrq, *ifgp; struct ifgroupreq *ifgr = data; if (ifgr->ifgr_len == 0) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) ifgr->ifgr_len += sizeof(struct ifg_req); IF_ADDR_RUNLOCK(ifp); return (0); } len = ifgr->ifgr_len; ifgp = ifgr->ifgr_groups; /* XXX: wire */ IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) { if (len < sizeof(ifgrq)) { IF_ADDR_RUNLOCK(ifp); return (EINVAL); } bzero(&ifgrq, sizeof ifgrq); strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group, sizeof(ifgrq.ifgrq_group)); if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { IF_ADDR_RUNLOCK(ifp); return (error); } len -= sizeof(ifgrq); ifgp++; } IF_ADDR_RUNLOCK(ifp); return (0); } /* * Stores all members of a group in memory pointed to by data */ static int if_getgroupmembers(struct ifgroupreq *data) { struct ifgroupreq *ifgr = data; struct ifg_group *ifg; struct ifg_member *ifgm; struct ifg_req ifgrq, *ifgp; int len, error; IFNET_RLOCK(); TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next) if (!strcmp(ifg->ifg_group, ifgr->ifgr_name)) break; if (ifg == NULL) { IFNET_RUNLOCK(); return (ENOENT); } if (ifgr->ifgr_len == 0) { TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) ifgr->ifgr_len += sizeof(ifgrq); IFNET_RUNLOCK(); return (0); } len = ifgr->ifgr_len; ifgp = ifgr->ifgr_groups; TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) { if (len < sizeof(ifgrq)) { IFNET_RUNLOCK(); return (EINVAL); } bzero(&ifgrq, sizeof ifgrq); strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname, sizeof(ifgrq.ifgrq_member)); if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) { IFNET_RUNLOCK(); return (error); } len -= sizeof(ifgrq); ifgp++; } IFNET_RUNLOCK(); return (0); } /* * Return counter values from counter(9)s stored in ifnet. */ uint64_t if_get_counter_default(struct ifnet *ifp, ift_counter cnt) { KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); return (counter_u64_fetch(ifp->if_counters[cnt])); } /* * Increase an ifnet counter. Usually used for counters shared * between the stack and a driver, but function supports them all. */ void if_inc_counter(struct ifnet *ifp, ift_counter cnt, int64_t inc) { KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt)); counter_u64_add(ifp->if_counters[cnt], inc); } /* * Copy data from ifnet to userland API structure if_data. */ void if_data_copy(struct ifnet *ifp, struct if_data *ifd) { ifd->ifi_type = ifp->if_type; ifd->ifi_physical = 0; ifd->ifi_addrlen = ifp->if_addrlen; ifd->ifi_hdrlen = ifp->if_hdrlen; ifd->ifi_link_state = ifp->if_link_state; ifd->ifi_vhid = 0; ifd->ifi_datalen = sizeof(struct if_data); ifd->ifi_mtu = ifp->if_mtu; ifd->ifi_metric = ifp->if_metric; ifd->ifi_baudrate = ifp->if_baudrate; ifd->ifi_hwassist = ifp->if_hwassist; ifd->ifi_epoch = ifp->if_epoch; ifd->ifi_lastchange = ifp->if_lastchange; ifd->ifi_ipackets = ifp->if_get_counter(ifp, IFCOUNTER_IPACKETS); ifd->ifi_ierrors = ifp->if_get_counter(ifp, IFCOUNTER_IERRORS); ifd->ifi_opackets = ifp->if_get_counter(ifp, IFCOUNTER_OPACKETS); ifd->ifi_oerrors = ifp->if_get_counter(ifp, IFCOUNTER_OERRORS); ifd->ifi_collisions = ifp->if_get_counter(ifp, IFCOUNTER_COLLISIONS); ifd->ifi_ibytes = ifp->if_get_counter(ifp, IFCOUNTER_IBYTES); ifd->ifi_obytes = ifp->if_get_counter(ifp, IFCOUNTER_OBYTES); ifd->ifi_imcasts = ifp->if_get_counter(ifp, IFCOUNTER_IMCASTS); ifd->ifi_omcasts = ifp->if_get_counter(ifp, IFCOUNTER_OMCASTS); ifd->ifi_iqdrops = ifp->if_get_counter(ifp, IFCOUNTER_IQDROPS); ifd->ifi_oqdrops = ifp->if_get_counter(ifp, IFCOUNTER_OQDROPS); ifd->ifi_noproto = ifp->if_get_counter(ifp, IFCOUNTER_NOPROTO); } /* * Wrapper functions for struct ifnet address list locking macros. These are * used by kernel modules to avoid encoding programming interface or binary * interface assumptions that may be violated when kernel-internal locking * approaches change. */ void if_addr_rlock(struct ifnet *ifp) { IF_ADDR_RLOCK(ifp); } void if_addr_runlock(struct ifnet *ifp) { IF_ADDR_RUNLOCK(ifp); } void if_maddr_rlock(if_t ifp) { IF_ADDR_RLOCK((struct ifnet *)ifp); } void if_maddr_runlock(if_t ifp) { IF_ADDR_RUNLOCK((struct ifnet *)ifp); } /* * Initialization, destruction and refcounting functions for ifaddrs. */ struct ifaddr * ifa_alloc(size_t size, int flags) { struct ifaddr *ifa; KASSERT(size >= sizeof(struct ifaddr), ("%s: invalid size %zu", __func__, size)); ifa = malloc(size, M_IFADDR, M_ZERO | flags); if (ifa == NULL) return (NULL); if ((ifa->ifa_opackets = counter_u64_alloc(flags)) == NULL) goto fail; if ((ifa->ifa_ipackets = counter_u64_alloc(flags)) == NULL) goto fail; if ((ifa->ifa_obytes = counter_u64_alloc(flags)) == NULL) goto fail; if ((ifa->ifa_ibytes = counter_u64_alloc(flags)) == NULL) goto fail; refcount_init(&ifa->ifa_refcnt, 1); return (ifa); fail: /* free(NULL) is okay */ counter_u64_free(ifa->ifa_opackets); counter_u64_free(ifa->ifa_ipackets); counter_u64_free(ifa->ifa_obytes); counter_u64_free(ifa->ifa_ibytes); free(ifa, M_IFADDR); return (NULL); } void ifa_ref(struct ifaddr *ifa) { refcount_acquire(&ifa->ifa_refcnt); } void ifa_free(struct ifaddr *ifa) { if (refcount_release(&ifa->ifa_refcnt)) { counter_u64_free(ifa->ifa_opackets); counter_u64_free(ifa->ifa_ipackets); counter_u64_free(ifa->ifa_obytes); counter_u64_free(ifa->ifa_ibytes); free(ifa, M_IFADDR); } } static int ifa_maintain_loopback_route(int cmd, const char *otype, struct ifaddr *ifa, struct sockaddr *ia) { int error; struct rt_addrinfo info; struct sockaddr_dl null_sdl; struct ifnet *ifp; ifp = ifa->ifa_ifp; bzero(&info, sizeof(info)); if (cmd != RTM_DELETE) info.rti_ifp = V_loif; info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC; info.rti_info[RTAX_DST] = ia; info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl; link_init_sdl(ifp, (struct sockaddr *)&null_sdl, ifp->if_type); error = rtrequest1_fib(cmd, &info, NULL, ifp->if_fib); if (error != 0) log(LOG_DEBUG, "%s: %s failed for interface %s: %u\n", __func__, otype, if_name(ifp), error); return (error); } int ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) { return (ifa_maintain_loopback_route(RTM_ADD, "insertion", ifa, ia)); } int ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) { return (ifa_maintain_loopback_route(RTM_DELETE, "deletion", ifa, ia)); } int ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) { return (ifa_maintain_loopback_route(RTM_CHANGE, "switch", ifa, ia)); } /* * XXX: Because sockaddr_dl has deeper structure than the sockaddr * structs used to represent other address families, it is necessary * to perform a different comparison. */ #define sa_dl_equal(a1, a2) \ ((((const struct sockaddr_dl *)(a1))->sdl_len == \ ((const struct sockaddr_dl *)(a2))->sdl_len) && \ (bcmp(CLLADDR((const struct sockaddr_dl *)(a1)), \ CLLADDR((const struct sockaddr_dl *)(a2)), \ ((const struct sockaddr_dl *)(a1))->sdl_alen) == 0)) /* * Locate an interface based on a complete address. */ /*ARGSUSED*/ static struct ifaddr * ifa_ifwithaddr_internal(const struct sockaddr *addr, int getref) { struct ifnet *ifp; struct ifaddr *ifa; IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != addr->sa_family) continue; if (sa_equal(addr, ifa->ifa_addr)) { if (getref) ifa_ref(ifa); IF_ADDR_RUNLOCK(ifp); goto done; } /* IP6 doesn't have broadcast */ if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr && ifa->ifa_broadaddr->sa_len != 0 && sa_equal(ifa->ifa_broadaddr, addr)) { if (getref) ifa_ref(ifa); IF_ADDR_RUNLOCK(ifp); goto done; } } IF_ADDR_RUNLOCK(ifp); } ifa = NULL; done: IFNET_RUNLOCK_NOSLEEP(); return (ifa); } struct ifaddr * ifa_ifwithaddr(const struct sockaddr *addr) { return (ifa_ifwithaddr_internal(addr, 1)); } int ifa_ifwithaddr_check(const struct sockaddr *addr) { return (ifa_ifwithaddr_internal(addr, 0) != NULL); } /* * Locate an interface based on the broadcast address. */ /* ARGSUSED */ struct ifaddr * ifa_ifwithbroadaddr(const struct sockaddr *addr, int fibnum) { struct ifnet *ifp; struct ifaddr *ifa; IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) continue; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != addr->sa_family) continue; if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr && ifa->ifa_broadaddr->sa_len != 0 && sa_equal(ifa->ifa_broadaddr, addr)) { ifa_ref(ifa); IF_ADDR_RUNLOCK(ifp); goto done; } } IF_ADDR_RUNLOCK(ifp); } ifa = NULL; done: IFNET_RUNLOCK_NOSLEEP(); return (ifa); } /* * Locate the point to point interface with a given destination address. */ /*ARGSUSED*/ struct ifaddr * ifa_ifwithdstaddr(const struct sockaddr *addr, int fibnum) { struct ifnet *ifp; struct ifaddr *ifa; IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { if ((ifp->if_flags & IFF_POINTOPOINT) == 0) continue; if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) continue; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != addr->sa_family) continue; if (ifa->ifa_dstaddr != NULL && sa_equal(addr, ifa->ifa_dstaddr)) { ifa_ref(ifa); IF_ADDR_RUNLOCK(ifp); goto done; } } IF_ADDR_RUNLOCK(ifp); } ifa = NULL; done: IFNET_RUNLOCK_NOSLEEP(); return (ifa); } /* * Find an interface on a specific network. If many, choice * is most specific found. */ struct ifaddr * ifa_ifwithnet(const struct sockaddr *addr, int ignore_ptp, int fibnum) { struct ifnet *ifp; struct ifaddr *ifa; struct ifaddr *ifa_maybe = NULL; u_int af = addr->sa_family; const char *addr_data = addr->sa_data, *cplim; /* * AF_LINK addresses can be looked up directly by their index number, * so do that if we can. */ if (af == AF_LINK) { const struct sockaddr_dl *sdl = (const struct sockaddr_dl *)addr; if (sdl->sdl_index && sdl->sdl_index <= V_if_index) return (ifaddr_byindex(sdl->sdl_index)); } /* * Scan though each interface, looking for ones that have addresses * in this address family and the requested fib. Maintain a reference * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that * kept it stable when we move onto the next interface. */ IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum)) continue; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { const char *cp, *cp2, *cp3; if (ifa->ifa_addr->sa_family != af) next: continue; if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) { /* * This is a bit broken as it doesn't * take into account that the remote end may * be a single node in the network we are * looking for. * The trouble is that we don't know the * netmask for the remote end. */ if (ifa->ifa_dstaddr != NULL && sa_equal(addr, ifa->ifa_dstaddr)) { ifa_ref(ifa); IF_ADDR_RUNLOCK(ifp); goto done; } } else { /* * Scan all the bits in the ifa's address. * If a bit dissagrees with what we are * looking for, mask it with the netmask * to see if it really matters. * (A byte at a time) */ if (ifa->ifa_netmask == 0) continue; cp = addr_data; cp2 = ifa->ifa_addr->sa_data; cp3 = ifa->ifa_netmask->sa_data; cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; while (cp3 < cplim) if ((*cp++ ^ *cp2++) & *cp3++) goto next; /* next address! */ /* * If the netmask of what we just found * is more specific than what we had before * (if we had one), or if the virtual status * of new prefix is better than of the old one, * then remember the new one before continuing * to search for an even better one. */ if (ifa_maybe == NULL || ifa_preferred(ifa_maybe, ifa) || rn_refines((caddr_t)ifa->ifa_netmask, (caddr_t)ifa_maybe->ifa_netmask)) { if (ifa_maybe != NULL) ifa_free(ifa_maybe); ifa_maybe = ifa; ifa_ref(ifa_maybe); } } } IF_ADDR_RUNLOCK(ifp); } ifa = ifa_maybe; ifa_maybe = NULL; done: IFNET_RUNLOCK_NOSLEEP(); if (ifa_maybe != NULL) ifa_free(ifa_maybe); return (ifa); } /* * Find an interface address specific to an interface best matching * a given address. */ struct ifaddr * ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp) { struct ifaddr *ifa; const char *cp, *cp2, *cp3; char *cplim; struct ifaddr *ifa_maybe = NULL; u_int af = addr->sa_family; if (af >= AF_MAX) return (NULL); IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != af) continue; if (ifa_maybe == NULL) ifa_maybe = ifa; if (ifa->ifa_netmask == 0) { if (sa_equal(addr, ifa->ifa_addr) || (ifa->ifa_dstaddr && sa_equal(addr, ifa->ifa_dstaddr))) goto done; continue; } if (ifp->if_flags & IFF_POINTOPOINT) { if (sa_equal(addr, ifa->ifa_dstaddr)) goto done; } else { cp = addr->sa_data; cp2 = ifa->ifa_addr->sa_data; cp3 = ifa->ifa_netmask->sa_data; cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; for (; cp3 < cplim; cp3++) if ((*cp++ ^ *cp2++) & *cp3) break; if (cp3 == cplim) goto done; } } ifa = ifa_maybe; done: if (ifa != NULL) ifa_ref(ifa); IF_ADDR_RUNLOCK(ifp); return (ifa); } /* * See whether new ifa is better than current one: * 1) A non-virtual one is preferred over virtual. * 2) A virtual in master state preferred over any other state. * * Used in several address selecting functions. */ int ifa_preferred(struct ifaddr *cur, struct ifaddr *next) { return (cur->ifa_carp && (!next->ifa_carp || ((*carp_master_p)(next) && !(*carp_master_p)(cur)))); } #include /* * Default action when installing a route with a Link Level gateway. * Lookup an appropriate real ifa to point to. * This should be moved to /sys/net/link.c eventually. */ static void link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) { struct ifaddr *ifa, *oifa; struct sockaddr *dst; struct ifnet *ifp; if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) return; ifa = ifaof_ifpforaddr(dst, ifp); if (ifa) { oifa = rt->rt_ifa; rt->rt_ifa = ifa; ifa_free(oifa); if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) ifa->ifa_rtrequest(cmd, rt, info); } } struct sockaddr_dl * link_alloc_sdl(size_t size, int flags) { return (malloc(size, M_TEMP, flags)); } void link_free_sdl(struct sockaddr *sa) { free(sa, M_TEMP); } /* * Fills in given sdl with interface basic info. * Returns pointer to filled sdl. */ struct sockaddr_dl * link_init_sdl(struct ifnet *ifp, struct sockaddr *paddr, u_char iftype) { struct sockaddr_dl *sdl; sdl = (struct sockaddr_dl *)paddr; memset(sdl, 0, sizeof(struct sockaddr_dl)); sdl->sdl_len = sizeof(struct sockaddr_dl); sdl->sdl_family = AF_LINK; sdl->sdl_index = ifp->if_index; sdl->sdl_type = iftype; return (sdl); } /* * Mark an interface down and notify protocols of * the transition. */ static void if_unroute(struct ifnet *ifp, int flag, int fam) { struct ifaddr *ifa; KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP")); ifp->if_flags &= ~flag; getmicrotime(&ifp->if_lastchange); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) pfctlinput(PRC_IFDOWN, ifa->ifa_addr); ifp->if_qflush(ifp); if (ifp->if_carp) (*carp_linkstate_p)(ifp); rt_ifmsg(ifp); } /* * Mark an interface up and notify protocols of * the transition. */ static void if_route(struct ifnet *ifp, int flag, int fam) { struct ifaddr *ifa; KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP")); ifp->if_flags |= flag; getmicrotime(&ifp->if_lastchange); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) pfctlinput(PRC_IFUP, ifa->ifa_addr); if (ifp->if_carp) (*carp_linkstate_p)(ifp); rt_ifmsg(ifp); #ifdef INET6 in6_if_up(ifp); #endif } void (*vlan_link_state_p)(struct ifnet *); /* XXX: private from if_vlan */ void (*vlan_trunk_cap_p)(struct ifnet *); /* XXX: private from if_vlan */ struct ifnet *(*vlan_trunkdev_p)(struct ifnet *); struct ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t); int (*vlan_tag_p)(struct ifnet *, uint16_t *); int (*vlan_setcookie_p)(struct ifnet *, void *); void *(*vlan_cookie_p)(struct ifnet *); /* * Handle a change in the interface link state. To avoid LORs * between driver lock and upper layer locks, as well as possible * recursions, we post event to taskqueue, and all job * is done in static do_link_state_change(). */ void if_link_state_change(struct ifnet *ifp, int link_state) { /* Return if state hasn't changed. */ if (ifp->if_link_state == link_state) return; ifp->if_link_state = link_state; taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask); } static void do_link_state_change(void *arg, int pending) { struct ifnet *ifp = (struct ifnet *)arg; int link_state = ifp->if_link_state; CURVNET_SET(ifp->if_vnet); /* Notify that the link state has changed. */ rt_ifmsg(ifp); if (ifp->if_vlantrunk != NULL) (*vlan_link_state_p)(ifp); if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) && ifp->if_l2com != NULL) (*ng_ether_link_state_p)(ifp, link_state); if (ifp->if_carp) (*carp_linkstate_p)(ifp); if (ifp->if_bridge) (*bridge_linkstate_p)(ifp); if (ifp->if_lagg) (*lagg_linkstate_p)(ifp, link_state); if (IS_DEFAULT_VNET(curvnet)) devctl_notify("IFNET", ifp->if_xname, (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", NULL); if (pending > 1) if_printf(ifp, "%d link states coalesced\n", pending); if (log_link_state_change) log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname, (link_state == LINK_STATE_UP) ? "UP" : "DOWN" ); EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state); CURVNET_RESTORE(); } /* * Mark an interface down and notify protocols of * the transition. */ void if_down(struct ifnet *ifp) { if_unroute(ifp, IFF_UP, AF_UNSPEC); } /* * Mark an interface up and notify protocols of * the transition. */ void if_up(struct ifnet *ifp) { if_route(ifp, IFF_UP, AF_UNSPEC); } /* * Flush an interface queue. */ void if_qflush(struct ifnet *ifp) { struct mbuf *m, *n; struct ifaltq *ifq; ifq = &ifp->if_snd; IFQ_LOCK(ifq); #ifdef ALTQ if (ALTQ_IS_ENABLED(ifq)) ALTQ_PURGE(ifq); #endif n = ifq->ifq_head; while ((m = n) != 0) { n = m->m_nextpkt; m_freem(m); } ifq->ifq_head = 0; ifq->ifq_tail = 0; ifq->ifq_len = 0; IFQ_UNLOCK(ifq); } /* * Map interface name to interface structure pointer, with or without * returning a reference. */ struct ifnet * ifunit_ref(const char *name) { struct ifnet *ifp; IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 && !(ifp->if_flags & IFF_DYING)) break; } if (ifp != NULL) if_ref(ifp); IFNET_RUNLOCK_NOSLEEP(); return (ifp); } struct ifnet * ifunit(const char *name) { struct ifnet *ifp; IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) break; } IFNET_RUNLOCK_NOSLEEP(); return (ifp); } /* * Hardware specific interface ioctls. */ static int ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) { struct ifreq *ifr; int error = 0; int new_flags, temp_flags; size_t namelen, onamelen; size_t descrlen; char *descrbuf, *odescrbuf; char new_name[IFNAMSIZ]; struct ifaddr *ifa; struct sockaddr_dl *sdl; ifr = (struct ifreq *)data; switch (cmd) { case SIOCGIFINDEX: ifr->ifr_index = ifp->if_index; break; case SIOCGIFFLAGS: temp_flags = ifp->if_flags | ifp->if_drv_flags; ifr->ifr_flags = temp_flags & 0xffff; ifr->ifr_flagshigh = temp_flags >> 16; break; case SIOCGIFCAP: ifr->ifr_reqcap = ifp->if_capabilities; ifr->ifr_curcap = ifp->if_capenable; break; #ifdef MAC case SIOCGIFMAC: error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp); break; #endif case SIOCGIFMETRIC: ifr->ifr_metric = ifp->if_metric; break; case SIOCGIFMTU: ifr->ifr_mtu = ifp->if_mtu; break; case SIOCGIFPHYS: /* XXXGL: did this ever worked? */ ifr->ifr_phys = 0; break; case SIOCGIFDESCR: error = 0; sx_slock(&ifdescr_sx); if (ifp->if_description == NULL) error = ENOMSG; else { /* space for terminating nul */ descrlen = strlen(ifp->if_description) + 1; if (ifr->ifr_buffer.length < descrlen) ifr->ifr_buffer.buffer = NULL; else error = copyout(ifp->if_description, ifr->ifr_buffer.buffer, descrlen); ifr->ifr_buffer.length = descrlen; } sx_sunlock(&ifdescr_sx); break; case SIOCSIFDESCR: error = priv_check(td, PRIV_NET_SETIFDESCR); if (error) return (error); /* * Copy only (length-1) bytes to make sure that * if_description is always nul terminated. The * length parameter is supposed to count the * terminating nul in. */ if (ifr->ifr_buffer.length > ifdescr_maxlen) return (ENAMETOOLONG); else if (ifr->ifr_buffer.length == 0) descrbuf = NULL; else { descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR, M_WAITOK | M_ZERO); error = copyin(ifr->ifr_buffer.buffer, descrbuf, ifr->ifr_buffer.length - 1); if (error) { free(descrbuf, M_IFDESCR); break; } } sx_xlock(&ifdescr_sx); odescrbuf = ifp->if_description; ifp->if_description = descrbuf; sx_xunlock(&ifdescr_sx); getmicrotime(&ifp->if_lastchange); free(odescrbuf, M_IFDESCR); break; case SIOCGIFFIB: ifr->ifr_fib = ifp->if_fib; break; case SIOCSIFFIB: error = priv_check(td, PRIV_NET_SETIFFIB); if (error) return (error); if (ifr->ifr_fib >= rt_numfibs) return (EINVAL); ifp->if_fib = ifr->ifr_fib; break; case SIOCSIFFLAGS: error = priv_check(td, PRIV_NET_SETIFFLAGS); if (error) return (error); /* * Currently, no driver owned flags pass the IFF_CANTCHANGE * check, so we don't need special handling here yet. */ new_flags = (ifr->ifr_flags & 0xffff) | (ifr->ifr_flagshigh << 16); if (ifp->if_flags & IFF_UP && (new_flags & IFF_UP) == 0) { if_down(ifp); } else if (new_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) { if_up(ifp); } /* See if permanently promiscuous mode bit is about to flip */ if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { if (new_flags & IFF_PPROMISC) ifp->if_flags |= IFF_PROMISC; else if (ifp->if_pcount == 0) ifp->if_flags &= ~IFF_PROMISC; log(LOG_INFO, "%s: permanently promiscuous mode %s\n", ifp->if_xname, (new_flags & IFF_PPROMISC) ? "enabled" : "disabled"); } ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | (new_flags &~ IFF_CANTCHANGE); if (ifp->if_ioctl) { (void) (*ifp->if_ioctl)(ifp, cmd, data); } getmicrotime(&ifp->if_lastchange); break; case SIOCSIFCAP: error = priv_check(td, PRIV_NET_SETIFCAP); if (error) return (error); if (ifp->if_ioctl == NULL) return (EOPNOTSUPP); if (ifr->ifr_reqcap & ~ifp->if_capabilities) return (EINVAL); error = (*ifp->if_ioctl)(ifp, cmd, data); if (error == 0) getmicrotime(&ifp->if_lastchange); break; #ifdef MAC case SIOCSIFMAC: error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp); break; #endif case SIOCSIFNAME: error = priv_check(td, PRIV_NET_SETIFNAME); if (error) return (error); error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); if (error != 0) return (error); if (new_name[0] == '\0') return (EINVAL); if (ifunit(new_name) != NULL) return (EEXIST); /* * XXX: Locking. Nothing else seems to lock if_flags, * and there are numerous other races with the * ifunit() checks not being atomic with namespace * changes (renames, vmoves, if_attach, etc). */ ifp->if_flags |= IFF_RENAMING; /* Announce the departure of the interface. */ rt_ifannouncemsg(ifp, IFAN_DEPARTURE); EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); log(LOG_INFO, "%s: changing name to '%s'\n", ifp->if_xname, new_name); IF_ADDR_WLOCK(ifp); strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); ifa = ifp->if_addr; sdl = (struct sockaddr_dl *)ifa->ifa_addr; namelen = strlen(new_name); onamelen = sdl->sdl_nlen; /* * Move the address if needed. This is safe because we * allocate space for a name of length IFNAMSIZ when we * create this in if_attach(). */ if (namelen != onamelen) { bcopy(sdl->sdl_data + onamelen, sdl->sdl_data + namelen, sdl->sdl_alen); } bcopy(new_name, sdl->sdl_data, namelen); sdl->sdl_nlen = namelen; sdl = (struct sockaddr_dl *)ifa->ifa_netmask; bzero(sdl->sdl_data, onamelen); while (namelen != 0) sdl->sdl_data[--namelen] = 0xff; IF_ADDR_WUNLOCK(ifp); EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); /* Announce the return of the interface. */ rt_ifannouncemsg(ifp, IFAN_ARRIVAL); ifp->if_flags &= ~IFF_RENAMING; break; #ifdef VIMAGE case SIOCSIFVNET: error = priv_check(td, PRIV_NET_SETIFVNET); if (error) return (error); error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid); break; #endif case SIOCSIFMETRIC: error = priv_check(td, PRIV_NET_SETIFMETRIC); if (error) return (error); ifp->if_metric = ifr->ifr_metric; getmicrotime(&ifp->if_lastchange); break; case SIOCSIFPHYS: error = priv_check(td, PRIV_NET_SETIFPHYS); if (error) return (error); if (ifp->if_ioctl == NULL) return (EOPNOTSUPP); error = (*ifp->if_ioctl)(ifp, cmd, data); if (error == 0) getmicrotime(&ifp->if_lastchange); break; case SIOCSIFMTU: { u_long oldmtu = ifp->if_mtu; error = priv_check(td, PRIV_NET_SETIFMTU); if (error) return (error); if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) return (EINVAL); if (ifp->if_ioctl == NULL) return (EOPNOTSUPP); error = (*ifp->if_ioctl)(ifp, cmd, data); if (error == 0) { getmicrotime(&ifp->if_lastchange); rt_ifmsg(ifp); } /* * If the link MTU changed, do network layer specific procedure. */ if (ifp->if_mtu != oldmtu) { #ifdef INET6 nd6_setmtu(ifp); #endif rt_updatemtu(ifp); } break; } case SIOCADDMULTI: case SIOCDELMULTI: if (cmd == SIOCADDMULTI) error = priv_check(td, PRIV_NET_ADDMULTI); else error = priv_check(td, PRIV_NET_DELMULTI); if (error) return (error); /* Don't allow group membership on non-multicast interfaces. */ if ((ifp->if_flags & IFF_MULTICAST) == 0) return (EOPNOTSUPP); /* Don't let users screw up protocols' entries. */ if (ifr->ifr_addr.sa_family != AF_LINK) return (EINVAL); if (cmd == SIOCADDMULTI) { struct ifmultiaddr *ifma; /* * Userland is only permitted to join groups once * via the if_addmulti() KPI, because it cannot hold * struct ifmultiaddr * between calls. It may also * lose a race while we check if the membership * already exists. */ IF_ADDR_RLOCK(ifp); ifma = if_findmulti(ifp, &ifr->ifr_addr); IF_ADDR_RUNLOCK(ifp); if (ifma != NULL) error = EADDRINUSE; else error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); } else { error = if_delmulti(ifp, &ifr->ifr_addr); } if (error == 0) getmicrotime(&ifp->if_lastchange); break; case SIOCSIFPHYADDR: case SIOCDIFPHYADDR: #ifdef INET6 case SIOCSIFPHYADDR_IN6: #endif case SIOCSIFMEDIA: case SIOCSIFGENERIC: error = priv_check(td, PRIV_NET_HWIOCTL); if (error) return (error); if (ifp->if_ioctl == NULL) return (EOPNOTSUPP); error = (*ifp->if_ioctl)(ifp, cmd, data); if (error == 0) getmicrotime(&ifp->if_lastchange); break; case SIOCGIFSTATUS: case SIOCGIFPSRCADDR: case SIOCGIFPDSTADDR: case SIOCGIFMEDIA: case SIOCGIFXMEDIA: case SIOCGIFGENERIC: if (ifp->if_ioctl == NULL) return (EOPNOTSUPP); error = (*ifp->if_ioctl)(ifp, cmd, data); break; case SIOCSIFLLADDR: error = priv_check(td, PRIV_NET_SETLLADDR); if (error) return (error); error = if_setlladdr(ifp, ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); break; case SIOCAIFGROUP: { struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; error = priv_check(td, PRIV_NET_ADDIFGROUP); if (error) return (error); if ((error = if_addgroup(ifp, ifgr->ifgr_group))) return (error); break; } case SIOCGIFGROUP: if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp))) return (error); break; case SIOCDIFGROUP: { struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr; error = priv_check(td, PRIV_NET_DELIFGROUP); if (error) return (error); if ((error = if_delgroup(ifp, ifgr->ifgr_group))) return (error); break; } default: error = ENOIOCTL; break; } return (error); } #ifdef COMPAT_FREEBSD32 struct ifconf32 { int32_t ifc_len; union { uint32_t ifcu_buf; uint32_t ifcu_req; } ifc_ifcu; }; #define SIOCGIFCONF32 _IOWR('i', 36, struct ifconf32) #endif /* * Interface ioctls. */ int ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) { struct ifnet *ifp; struct ifreq *ifr; int error; int oif_flags; CURVNET_SET(so->so_vnet); switch (cmd) { case SIOCGIFCONF: error = ifconf(cmd, data); CURVNET_RESTORE(); return (error); #ifdef COMPAT_FREEBSD32 case SIOCGIFCONF32: { struct ifconf32 *ifc32; struct ifconf ifc; ifc32 = (struct ifconf32 *)data; ifc.ifc_len = ifc32->ifc_len; ifc.ifc_buf = PTRIN(ifc32->ifc_buf); error = ifconf(SIOCGIFCONF, (void *)&ifc); CURVNET_RESTORE(); if (error == 0) ifc32->ifc_len = ifc.ifc_len; return (error); } #endif } ifr = (struct ifreq *)data; switch (cmd) { #ifdef VIMAGE case SIOCSIFRVNET: error = priv_check(td, PRIV_NET_SETIFVNET); if (error == 0) error = if_vmove_reclaim(td, ifr->ifr_name, ifr->ifr_jid); CURVNET_RESTORE(); return (error); #endif case SIOCIFCREATE: case SIOCIFCREATE2: error = priv_check(td, PRIV_NET_IFCREATE); if (error == 0) error = if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name), cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL); CURVNET_RESTORE(); return (error); case SIOCIFDESTROY: error = priv_check(td, PRIV_NET_IFDESTROY); if (error == 0) error = if_clone_destroy(ifr->ifr_name); CURVNET_RESTORE(); return (error); case SIOCIFGCLONERS: error = if_clone_list((struct if_clonereq *)data); CURVNET_RESTORE(); return (error); case SIOCGIFGMEMB: error = if_getgroupmembers((struct ifgroupreq *)data); CURVNET_RESTORE(); return (error); #if defined(INET) || defined(INET6) case SIOCSVH: case SIOCGVH: if (carp_ioctl_p == NULL) error = EPROTONOSUPPORT; else error = (*carp_ioctl_p)(ifr, cmd, td); CURVNET_RESTORE(); return (error); #endif } ifp = ifunit_ref(ifr->ifr_name); if (ifp == NULL) { CURVNET_RESTORE(); return (ENXIO); } error = ifhwioctl(cmd, ifp, data, td); if (error != ENOIOCTL) { if_rele(ifp); CURVNET_RESTORE(); return (error); } oif_flags = ifp->if_flags; if (so->so_proto == NULL) { if_rele(ifp); CURVNET_RESTORE(); return (EOPNOTSUPP); } /* * Pass the request on to the socket control method, and if the * latter returns EOPNOTSUPP, directly to the interface. * * Make an exception for the legacy SIOCSIF* requests. Drivers * trust SIOCSIFADDR et al to come from an already privileged * layer, and do not perform any credentials checks or input * validation. */ error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, data, ifp, td)); if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL && cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR && cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK) error = (*ifp->if_ioctl)(ifp, cmd, data); if ((oif_flags ^ ifp->if_flags) & IFF_UP) { #ifdef INET6 if (ifp->if_flags & IFF_UP) in6_if_up(ifp); #endif } if_rele(ifp); CURVNET_RESTORE(); return (error); } /* * The code common to handling reference counted flags, * e.g., in ifpromisc() and if_allmulti(). * The "pflag" argument can specify a permanent mode flag to check, * such as IFF_PPROMISC for promiscuous mode; should be 0 if none. * * Only to be used on stack-owned flags, not driver-owned flags. */ static int if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch) { struct ifreq ifr; int error; int oldflags, oldcount; /* Sanity checks to catch programming errors */ KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0, ("%s: setting driver-owned flag %d", __func__, flag)); if (onswitch) KASSERT(*refcount >= 0, ("%s: increment negative refcount %d for flag %d", __func__, *refcount, flag)); else KASSERT(*refcount > 0, ("%s: decrement non-positive refcount %d for flag %d", __func__, *refcount, flag)); /* In case this mode is permanent, just touch refcount */ if (ifp->if_flags & pflag) { *refcount += onswitch ? 1 : -1; return (0); } /* Save ifnet parameters for if_ioctl() may fail */ oldcount = *refcount; oldflags = ifp->if_flags; /* * See if we aren't the only and touching refcount is enough. * Actually toggle interface flag if we are the first or last. */ if (onswitch) { if ((*refcount)++) return (0); ifp->if_flags |= flag; } else { if (--(*refcount)) return (0); ifp->if_flags &= ~flag; } /* Call down the driver since we've changed interface flags */ if (ifp->if_ioctl == NULL) { error = EOPNOTSUPP; goto recover; } ifr.ifr_flags = ifp->if_flags & 0xffff; ifr.ifr_flagshigh = ifp->if_flags >> 16; error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); if (error) goto recover; /* Notify userland that interface flags have changed */ rt_ifmsg(ifp); return (0); recover: /* Recover after driver error */ *refcount = oldcount; ifp->if_flags = oldflags; return (error); } /* * Set/clear promiscuous mode on interface ifp based on the truth value * of pswitch. The calls are reference counted so that only the first * "on" request actually has an effect, as does the final "off" request. * Results are undefined if the "off" and "on" requests are not matched. */ int ifpromisc(struct ifnet *ifp, int pswitch) { int error; int oldflags = ifp->if_flags; error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC, &ifp->if_pcount, pswitch); /* If promiscuous mode status has changed, log a message */ if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC)) log(LOG_INFO, "%s: promiscuous mode %s\n", ifp->if_xname, (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); return (error); } /* * Return interface configuration * of system. List may be used * in later ioctl's (above) to get * other information. */ /*ARGSUSED*/ static int ifconf(u_long cmd, caddr_t data) { struct ifconf *ifc = (struct ifconf *)data; struct ifnet *ifp; struct ifaddr *ifa; struct ifreq ifr; struct sbuf *sb; int error, full = 0, valid_len, max_len; /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */ max_len = MAXPHYS - 1; /* Prevent hostile input from being able to crash the system */ if (ifc->ifc_len <= 0) return (EINVAL); again: if (ifc->ifc_len <= max_len) { max_len = ifc->ifc_len; full = 1; } sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN); max_len = 0; valid_len = 0; IFNET_RLOCK(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { int addrs; /* * Zero the ifr_name buffer to make sure we don't * disclose the contents of the stack. */ memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name)); if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) >= sizeof(ifr.ifr_name)) { sbuf_delete(sb); IFNET_RUNLOCK(); return (ENAMETOOLONG); } addrs = 0; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { struct sockaddr *sa = ifa->ifa_addr; if (prison_if(curthread->td_ucred, sa) != 0) continue; addrs++; if (sa->sa_len <= sizeof(*sa)) { ifr.ifr_addr = *sa; sbuf_bcat(sb, &ifr, sizeof(ifr)); max_len += sizeof(ifr); } else { sbuf_bcat(sb, &ifr, offsetof(struct ifreq, ifr_addr)); max_len += offsetof(struct ifreq, ifr_addr); sbuf_bcat(sb, sa, sa->sa_len); max_len += sa->sa_len; } if (sbuf_error(sb) == 0) valid_len = sbuf_len(sb); } IF_ADDR_RUNLOCK(ifp); if (addrs == 0) { bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); sbuf_bcat(sb, &ifr, sizeof(ifr)); max_len += sizeof(ifr); if (sbuf_error(sb) == 0) valid_len = sbuf_len(sb); } } IFNET_RUNLOCK(); /* * If we didn't allocate enough space (uncommon), try again. If * we have already allocated as much space as we are allowed, * return what we've got. */ if (valid_len != max_len && !full) { sbuf_delete(sb); goto again; } ifc->ifc_len = valid_len; sbuf_finish(sb); error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len); sbuf_delete(sb); return (error); } /* * Just like ifpromisc(), but for all-multicast-reception mode. */ int if_allmulti(struct ifnet *ifp, int onswitch) { return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch)); } struct ifmultiaddr * if_findmulti(struct ifnet *ifp, const struct sockaddr *sa) { struct ifmultiaddr *ifma; IF_ADDR_LOCK_ASSERT(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (sa->sa_family == AF_LINK) { if (sa_dl_equal(ifma->ifma_addr, sa)) break; } else { if (sa_equal(ifma->ifma_addr, sa)) break; } } return ifma; } /* * Allocate a new ifmultiaddr and initialize based on passed arguments. We * make copies of passed sockaddrs. The ifmultiaddr will not be added to * the ifnet multicast address list here, so the caller must do that and * other setup work (such as notifying the device driver). The reference * count is initialized to 1. */ static struct ifmultiaddr * if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa, int mflags) { struct ifmultiaddr *ifma; struct sockaddr *dupsa; ifma = malloc(sizeof *ifma, M_IFMADDR, mflags | M_ZERO); if (ifma == NULL) return (NULL); dupsa = malloc(sa->sa_len, M_IFMADDR, mflags); if (dupsa == NULL) { free(ifma, M_IFMADDR); return (NULL); } bcopy(sa, dupsa, sa->sa_len); ifma->ifma_addr = dupsa; ifma->ifma_ifp = ifp; ifma->ifma_refcount = 1; ifma->ifma_protospec = NULL; if (llsa == NULL) { ifma->ifma_lladdr = NULL; return (ifma); } dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags); if (dupsa == NULL) { free(ifma->ifma_addr, M_IFMADDR); free(ifma, M_IFMADDR); return (NULL); } bcopy(llsa, dupsa, llsa->sa_len); ifma->ifma_lladdr = dupsa; return (ifma); } /* * if_freemulti: free ifmultiaddr structure and possibly attached related * addresses. The caller is responsible for implementing reference * counting, notifying the driver, handling routing messages, and releasing * any dependent link layer state. */ static void if_freemulti(struct ifmultiaddr *ifma) { KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d", ifma->ifma_refcount)); if (ifma->ifma_lladdr != NULL) free(ifma->ifma_lladdr, M_IFMADDR); free(ifma->ifma_addr, M_IFMADDR); free(ifma, M_IFMADDR); } /* * Register an additional multicast address with a network interface. * * - If the address is already present, bump the reference count on the * address and return. * - If the address is not link-layer, look up a link layer address. * - Allocate address structures for one or both addresses, and attach to the * multicast address list on the interface. If automatically adding a link * layer address, the protocol address will own a reference to the link * layer address, to be freed when it is freed. * - Notify the network device driver of an addition to the multicast address * list. * * 'sa' points to caller-owned memory with the desired multicast address. * * 'retifma' will be used to return a pointer to the resulting multicast * address reference, if desired. */ int if_addmulti(struct ifnet *ifp, struct sockaddr *sa, struct ifmultiaddr **retifma) { struct ifmultiaddr *ifma, *ll_ifma; struct sockaddr *llsa; struct sockaddr_dl sdl; int error; /* * If the address is already present, return a new reference to it; * otherwise, allocate storage and set up a new address. */ IF_ADDR_WLOCK(ifp); ifma = if_findmulti(ifp, sa); if (ifma != NULL) { ifma->ifma_refcount++; if (retifma != NULL) *retifma = ifma; IF_ADDR_WUNLOCK(ifp); return (0); } /* * The address isn't already present; resolve the protocol address * into a link layer address, and then look that up, bump its * refcount or allocate an ifma for that also. * Most link layer resolving functions returns address data which * fits inside default sockaddr_dl structure. However callback * can allocate another sockaddr structure, in that case we need to * free it later. */ llsa = NULL; ll_ifma = NULL; if (ifp->if_resolvemulti != NULL) { /* Provide called function with buffer size information */ sdl.sdl_len = sizeof(sdl); llsa = (struct sockaddr *)&sdl; error = ifp->if_resolvemulti(ifp, &llsa, sa); if (error) goto unlock_out; } /* * Allocate the new address. Don't hook it up yet, as we may also * need to allocate a link layer multicast address. */ ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT); if (ifma == NULL) { error = ENOMEM; goto free_llsa_out; } /* * If a link layer address is found, we'll need to see if it's * already present in the address list, or allocate is as well. * When this block finishes, the link layer address will be on the * list. */ if (llsa != NULL) { ll_ifma = if_findmulti(ifp, llsa); if (ll_ifma == NULL) { ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT); if (ll_ifma == NULL) { --ifma->ifma_refcount; if_freemulti(ifma); error = ENOMEM; goto free_llsa_out; } TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma, ifma_link); } else ll_ifma->ifma_refcount++; ifma->ifma_llifma = ll_ifma; } /* * We now have a new multicast address, ifma, and possibly a new or * referenced link layer address. Add the primary address to the * ifnet address list. */ TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); if (retifma != NULL) *retifma = ifma; /* * Must generate the message while holding the lock so that 'ifma' * pointer is still valid. */ rt_newmaddrmsg(RTM_NEWMADDR, ifma); IF_ADDR_WUNLOCK(ifp); /* * We are certain we have added something, so call down to the * interface to let them know about it. */ if (ifp->if_ioctl != NULL) { (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0); } if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) link_free_sdl(llsa); return (0); free_llsa_out: if ((llsa != NULL) && (llsa != (struct sockaddr *)&sdl)) link_free_sdl(llsa); unlock_out: IF_ADDR_WUNLOCK(ifp); return (error); } /* * Delete a multicast group membership by network-layer group address. * * Returns ENOENT if the entry could not be found. If ifp no longer * exists, results are undefined. This entry point should only be used * from subsystems which do appropriate locking to hold ifp for the * duration of the call. * Network-layer protocol domains must use if_delmulti_ifma(). */ int if_delmulti(struct ifnet *ifp, struct sockaddr *sa) { struct ifmultiaddr *ifma; int lastref; #ifdef INVARIANTS struct ifnet *oifp; IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(oifp, &V_ifnet, if_link) if (ifp == oifp) break; if (ifp != oifp) ifp = NULL; IFNET_RUNLOCK_NOSLEEP(); KASSERT(ifp != NULL, ("%s: ifnet went away", __func__)); #endif if (ifp == NULL) return (ENOENT); IF_ADDR_WLOCK(ifp); lastref = 0; ifma = if_findmulti(ifp, sa); if (ifma != NULL) lastref = if_delmulti_locked(ifp, ifma, 0); IF_ADDR_WUNLOCK(ifp); if (ifma == NULL) return (ENOENT); if (lastref && ifp->if_ioctl != NULL) { (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); } return (0); } /* * Delete all multicast group membership for an interface. * Should be used to quickly flush all multicast filters. */ void if_delallmulti(struct ifnet *ifp) { struct ifmultiaddr *ifma; struct ifmultiaddr *next; IF_ADDR_WLOCK(ifp); TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) if_delmulti_locked(ifp, ifma, 0); IF_ADDR_WUNLOCK(ifp); } /* * Delete a multicast group membership by group membership pointer. * Network-layer protocol domains must use this routine. * * It is safe to call this routine if the ifp disappeared. */ void if_delmulti_ifma(struct ifmultiaddr *ifma) { struct ifnet *ifp; int lastref; ifp = ifma->ifma_ifp; #ifdef DIAGNOSTIC if (ifp == NULL) { printf("%s: ifma_ifp seems to be detached\n", __func__); } else { struct ifnet *oifp; IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(oifp, &V_ifnet, if_link) if (ifp == oifp) break; if (ifp != oifp) { printf("%s: ifnet %p disappeared\n", __func__, ifp); ifp = NULL; } IFNET_RUNLOCK_NOSLEEP(); } #endif /* * If and only if the ifnet instance exists: Acquire the address lock. */ if (ifp != NULL) IF_ADDR_WLOCK(ifp); lastref = if_delmulti_locked(ifp, ifma, 0); if (ifp != NULL) { /* * If and only if the ifnet instance exists: * Release the address lock. * If the group was left: update the hardware hash filter. */ IF_ADDR_WUNLOCK(ifp); if (lastref && ifp->if_ioctl != NULL) { (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0); } } } /* * Perform deletion of network-layer and/or link-layer multicast address. * * Return 0 if the reference count was decremented. * Return 1 if the final reference was released, indicating that the * hardware hash filter should be reprogrammed. */ static int if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching) { struct ifmultiaddr *ll_ifma; if (ifp != NULL && ifma->ifma_ifp != NULL) { KASSERT(ifma->ifma_ifp == ifp, ("%s: inconsistent ifp %p", __func__, ifp)); IF_ADDR_WLOCK_ASSERT(ifp); } ifp = ifma->ifma_ifp; /* * If the ifnet is detaching, null out references to ifnet, * so that upper protocol layers will notice, and not attempt * to obtain locks for an ifnet which no longer exists. The * routing socket announcement must happen before the ifnet * instance is detached from the system. */ if (detaching) { #ifdef DIAGNOSTIC printf("%s: detaching ifnet instance %p\n", __func__, ifp); #endif /* * ifp may already be nulled out if we are being reentered * to delete the ll_ifma. */ if (ifp != NULL) { rt_newmaddrmsg(RTM_DELMADDR, ifma); ifma->ifma_ifp = NULL; } } if (--ifma->ifma_refcount > 0) return 0; /* * If this ifma is a network-layer ifma, a link-layer ifma may * have been associated with it. Release it first if so. */ ll_ifma = ifma->ifma_llifma; if (ll_ifma != NULL) { KASSERT(ifma->ifma_lladdr != NULL, ("%s: llifma w/o lladdr", __func__)); if (detaching) ll_ifma->ifma_ifp = NULL; /* XXX */ if (--ll_ifma->ifma_refcount == 0) { if (ifp != NULL) { TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifma_link); } if_freemulti(ll_ifma); } } if (ifp != NULL) TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); if_freemulti(ifma); /* * The last reference to this instance of struct ifmultiaddr * was released; the hardware should be notified of this change. */ return 1; } /* * Set the link layer address on an interface. * * At this time we only support certain types of interfaces, * and we don't allow the length of the address to change. * * Set noinline to be dtrace-friendly */ __noinline int if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) { struct sockaddr_dl *sdl; struct ifaddr *ifa; struct ifreq ifr; IF_ADDR_RLOCK(ifp); ifa = ifp->if_addr; if (ifa == NULL) { IF_ADDR_RUNLOCK(ifp); return (EINVAL); } ifa_ref(ifa); IF_ADDR_RUNLOCK(ifp); sdl = (struct sockaddr_dl *)ifa->ifa_addr; if (sdl == NULL) { ifa_free(ifa); return (EINVAL); } if (len != sdl->sdl_alen) { /* don't allow length to change */ ifa_free(ifa); return (EINVAL); } switch (ifp->if_type) { case IFT_ETHER: case IFT_FDDI: case IFT_XETHER: case IFT_ISO88025: case IFT_L2VLAN: case IFT_BRIDGE: case IFT_ARCNET: case IFT_IEEE8023ADLAG: case IFT_IEEE80211: bcopy(lladdr, LLADDR(sdl), len); ifa_free(ifa); break; default: ifa_free(ifa); return (ENODEV); } /* * If the interface is already up, we need * to re-init it in order to reprogram its * address filter. */ if ((ifp->if_flags & IFF_UP) != 0) { if (ifp->if_ioctl) { ifp->if_flags &= ~IFF_UP; ifr.ifr_flags = ifp->if_flags & 0xffff; ifr.ifr_flagshigh = ifp->if_flags >> 16; (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); ifp->if_flags |= IFF_UP; ifr.ifr_flags = ifp->if_flags & 0xffff; ifr.ifr_flagshigh = ifp->if_flags >> 16; (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); } } EVENTHANDLER_INVOKE(iflladdr_event, ifp); return (0); } /* * The name argument must be a pointer to storage which will last as * long as the interface does. For physical devices, the result of * device_get_name(dev) is a good choice and for pseudo-devices a * static string works well. */ void if_initname(struct ifnet *ifp, const char *name, int unit) { ifp->if_dname = name; ifp->if_dunit = unit; if (unit != IF_DUNIT_NONE) snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); else strlcpy(ifp->if_xname, name, IFNAMSIZ); } int if_printf(struct ifnet *ifp, const char * fmt, ...) { va_list ap; int retval; retval = printf("%s: ", ifp->if_xname); va_start(ap, fmt); retval += vprintf(fmt, ap); va_end(ap); return (retval); } void if_start(struct ifnet *ifp) { (*(ifp)->if_start)(ifp); } /* * Backwards compatibility interface for drivers * that have not implemented it */ static int if_transmit(struct ifnet *ifp, struct mbuf *m) { int error; IFQ_HANDOFF(ifp, m, error); return (error); } static void if_input_default(struct ifnet *ifp __unused, struct mbuf *m) { m_freem(m); } int if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust) { int active = 0; IF_LOCK(ifq); if (_IF_QFULL(ifq)) { IF_UNLOCK(ifq); if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1); m_freem(m); return (0); } if (ifp != NULL) { if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len + adjust); if (m->m_flags & (M_BCAST|M_MCAST)) if_inc_counter(ifp, IFCOUNTER_OMCASTS, 1); active = ifp->if_drv_flags & IFF_DRV_OACTIVE; } _IF_ENQUEUE(ifq, m); IF_UNLOCK(ifq); if (ifp != NULL && !active) (*(ifp)->if_start)(ifp); return (1); } void if_register_com_alloc(u_char type, if_com_alloc_t *a, if_com_free_t *f) { KASSERT(if_com_alloc[type] == NULL, ("if_register_com_alloc: %d already registered", type)); KASSERT(if_com_free[type] == NULL, ("if_register_com_alloc: %d free already registered", type)); if_com_alloc[type] = a; if_com_free[type] = f; } void if_deregister_com_alloc(u_char type) { KASSERT(if_com_alloc[type] != NULL, ("if_deregister_com_alloc: %d not registered", type)); KASSERT(if_com_free[type] != NULL, ("if_deregister_com_alloc: %d free not registered", type)); if_com_alloc[type] = NULL; if_com_free[type] = NULL; } /* API for driver access to network stack owned ifnet.*/ uint64_t if_setbaudrate(struct ifnet *ifp, uint64_t baudrate) { uint64_t oldbrate; oldbrate = ifp->if_baudrate; ifp->if_baudrate = baudrate; return (oldbrate); } uint64_t if_getbaudrate(if_t ifp) { return (((struct ifnet *)ifp)->if_baudrate); } int if_setcapabilities(if_t ifp, int capabilities) { ((struct ifnet *)ifp)->if_capabilities = capabilities; return (0); } int if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit) { ((struct ifnet *)ifp)->if_capabilities |= setbit; ((struct ifnet *)ifp)->if_capabilities &= ~clearbit; return (0); } int if_getcapabilities(if_t ifp) { return ((struct ifnet *)ifp)->if_capabilities; } int if_setcapenable(if_t ifp, int capabilities) { ((struct ifnet *)ifp)->if_capenable = capabilities; return (0); } int if_setcapenablebit(if_t ifp, int setcap, int clearcap) { if(setcap) ((struct ifnet *)ifp)->if_capenable |= setcap; if(clearcap) ((struct ifnet *)ifp)->if_capenable &= ~clearcap; return (0); } const char * if_getdname(if_t ifp) { return ((struct ifnet *)ifp)->if_dname; } int if_togglecapenable(if_t ifp, int togglecap) { ((struct ifnet *)ifp)->if_capenable ^= togglecap; return (0); } int if_getcapenable(if_t ifp) { return ((struct ifnet *)ifp)->if_capenable; } /* * This is largely undesirable because it ties ifnet to a device, but does * provide flexiblity for an embedded product vendor. Should be used with * the understanding that it violates the interface boundaries, and should be * a last resort only. */ int if_setdev(if_t ifp, void *dev) { return (0); } int if_setdrvflagbits(if_t ifp, int set_flags, int clear_flags) { ((struct ifnet *)ifp)->if_drv_flags |= set_flags; ((struct ifnet *)ifp)->if_drv_flags &= ~clear_flags; return (0); } int if_getdrvflags(if_t ifp) { return ((struct ifnet *)ifp)->if_drv_flags; } int if_setdrvflags(if_t ifp, int flags) { ((struct ifnet *)ifp)->if_drv_flags = flags; return (0); } int if_setflags(if_t ifp, int flags) { ((struct ifnet *)ifp)->if_flags = flags; return (0); } int if_setflagbits(if_t ifp, int set, int clear) { ((struct ifnet *)ifp)->if_flags |= set; ((struct ifnet *)ifp)->if_flags &= ~clear; return (0); } int if_getflags(if_t ifp) { return ((struct ifnet *)ifp)->if_flags; } int if_clearhwassist(if_t ifp) { ((struct ifnet *)ifp)->if_hwassist = 0; return (0); } int if_sethwassistbits(if_t ifp, int toset, int toclear) { ((struct ifnet *)ifp)->if_hwassist |= toset; ((struct ifnet *)ifp)->if_hwassist &= ~toclear; return (0); } int if_sethwassist(if_t ifp, int hwassist_bit) { ((struct ifnet *)ifp)->if_hwassist = hwassist_bit; return (0); } int if_gethwassist(if_t ifp) { return ((struct ifnet *)ifp)->if_hwassist; } int if_setmtu(if_t ifp, int mtu) { ((struct ifnet *)ifp)->if_mtu = mtu; return (0); } int if_getmtu(if_t ifp) { return ((struct ifnet *)ifp)->if_mtu; } int if_getmtu_family(if_t ifp, int family) { struct domain *dp; for (dp = domains; dp; dp = dp->dom_next) { if (dp->dom_family == family && dp->dom_ifmtu != NULL) return (dp->dom_ifmtu((struct ifnet *)ifp)); } return (((struct ifnet *)ifp)->if_mtu); } int if_setsoftc(if_t ifp, void *softc) { ((struct ifnet *)ifp)->if_softc = softc; return (0); } void * if_getsoftc(if_t ifp) { return ((struct ifnet *)ifp)->if_softc; } void if_setrcvif(struct mbuf *m, if_t ifp) { m->m_pkthdr.rcvif = (struct ifnet *)ifp; } void if_setvtag(struct mbuf *m, uint16_t tag) { m->m_pkthdr.ether_vtag = tag; } uint16_t if_getvtag(struct mbuf *m) { return (m->m_pkthdr.ether_vtag); } int if_sendq_empty(if_t ifp) { return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); } struct ifaddr * if_getifaddr(if_t ifp) { return ((struct ifnet *)ifp)->if_addr; } int if_getamcount(if_t ifp) { return ((struct ifnet *)ifp)->if_amcount; } int if_setsendqready(if_t ifp) { IFQ_SET_READY(&((struct ifnet *)ifp)->if_snd); return (0); } int if_setsendqlen(if_t ifp, int tx_desc_count) { IFQ_SET_MAXLEN(&((struct ifnet *)ifp)->if_snd, tx_desc_count); ((struct ifnet *)ifp)->if_snd.ifq_drv_maxlen = tx_desc_count; return (0); } int if_vlantrunkinuse(if_t ifp) { return ((struct ifnet *)ifp)->if_vlantrunk != NULL?1:0; } int if_input(if_t ifp, struct mbuf* sendmp) { (*((struct ifnet *)ifp)->if_input)((struct ifnet *)ifp, sendmp); return (0); } /* XXX */ #ifndef ETH_ADDR_LEN #define ETH_ADDR_LEN 6 #endif int if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max) { struct ifmultiaddr *ifma; uint8_t *lmta = (uint8_t *)mta; int mcnt = 0; TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_LINK) continue; if (mcnt == max) break; bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr), &lmta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN); mcnt++; } *cnt = mcnt; return (0); } int if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max) { int error; if_maddr_rlock(ifp); error = if_setupmultiaddr(ifp, mta, cnt, max); if_maddr_runlock(ifp); return (error); } int if_multiaddr_count(if_t ifp, int max) { struct ifmultiaddr *ifma; int count; count = 0; if_maddr_rlock(ifp); TAILQ_FOREACH(ifma, &((struct ifnet *)ifp)->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_LINK) continue; count++; if (count == max) break; } if_maddr_runlock(ifp); return (count); } struct mbuf * if_dequeue(if_t ifp) { struct mbuf *m; IFQ_DRV_DEQUEUE(&((struct ifnet *)ifp)->if_snd, m); return (m); } int if_sendq_prepend(if_t ifp, struct mbuf *m) { IFQ_DRV_PREPEND(&((struct ifnet *)ifp)->if_snd, m); return (0); } int if_setifheaderlen(if_t ifp, int len) { ((struct ifnet *)ifp)->if_hdrlen = len; return (0); } caddr_t if_getlladdr(if_t ifp) { return (IF_LLADDR((struct ifnet *)ifp)); } void * if_gethandle(u_char type) { return (if_alloc(type)); } void if_bpfmtap(if_t ifh, struct mbuf *m) { struct ifnet *ifp = (struct ifnet *)ifh; BPF_MTAP(ifp, m); } void if_etherbpfmtap(if_t ifh, struct mbuf *m) { struct ifnet *ifp = (struct ifnet *)ifh; ETHER_BPF_MTAP(ifp, m); } void if_vlancap(if_t ifh) { struct ifnet *ifp = (struct ifnet *)ifh; VLAN_CAPABILITIES(ifp); } void if_setinitfn(if_t ifp, void (*init_fn)(void *)) { ((struct ifnet *)ifp)->if_init = init_fn; } void if_setioctlfn(if_t ifp, int (*ioctl_fn)(if_t, u_long, caddr_t)) { ((struct ifnet *)ifp)->if_ioctl = (void *)ioctl_fn; } void if_setstartfn(if_t ifp, void (*start_fn)(if_t)) { ((struct ifnet *)ifp)->if_start = (void *)start_fn; } void if_settransmitfn(if_t ifp, if_transmit_fn_t start_fn) { ((struct ifnet *)ifp)->if_transmit = start_fn; } void if_setqflushfn(if_t ifp, if_qflush_fn_t flush_fn) { ((struct ifnet *)ifp)->if_qflush = flush_fn; } void if_setgetcounterfn(if_t ifp, if_get_counter_t fn) { ifp->if_get_counter = fn; } /* Revisit these - These are inline functions originally. */ int drbr_inuse_drv(if_t ifh, struct buf_ring *br) { return drbr_inuse(ifh, br); } struct mbuf* drbr_dequeue_drv(if_t ifh, struct buf_ring *br) { return drbr_dequeue(ifh, br); } int drbr_needs_enqueue_drv(if_t ifh, struct buf_ring *br) { return drbr_needs_enqueue(ifh, br); } int drbr_enqueue_drv(if_t ifh, struct buf_ring *br, struct mbuf *m) { return drbr_enqueue(ifh, br, m); } Index: projects/vnet/sys/net/if_disc.c =================================================================== --- projects/vnet/sys/net/if_disc.c (revision 294565) +++ projects/vnet/sys/net/if_disc.c (revision 294566) @@ -1,244 +1,244 @@ /*- * Copyright (c) 1982, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * From: @(#)if_loop.c 8.1 (Berkeley) 6/10/93 * $FreeBSD$ */ /* * Discard interface driver for protocol testing and timing. * (Based on the loopback.) */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "opt_inet.h" #include "opt_inet6.h" #ifdef TINY_DSMTU #define DSMTU (1024+512) #else #define DSMTU 65532 #endif struct disc_softc { struct ifnet *sc_ifp; }; static int discoutput(struct ifnet *, struct mbuf *, const struct sockaddr *, struct route *); static int discioctl(struct ifnet *, u_long, caddr_t); static int disc_clone_create(struct if_clone *, int, caddr_t); static void disc_clone_destroy(struct ifnet *); static const char discname[] = "disc"; static MALLOC_DEFINE(M_DISC, discname, "Discard interface"); static VNET_DEFINE(struct if_clone *, disc_cloner); #define V_disc_cloner VNET(disc_cloner) static int disc_clone_create(struct if_clone *ifc, int unit, caddr_t params) { struct ifnet *ifp; struct disc_softc *sc; sc = malloc(sizeof(struct disc_softc), M_DISC, M_WAITOK | M_ZERO); ifp = sc->sc_ifp = if_alloc(IFT_LOOP); if (ifp == NULL) { free(sc, M_DISC); return (ENOSPC); } ifp->if_softc = sc; if_initname(ifp, discname, unit); ifp->if_mtu = DSMTU; /* * IFF_LOOPBACK should not be removed from disc's flags because * it controls what PF-specific routes are magically added when * a network address is assigned to the interface. Things just * won't work as intended w/o such routes because the output * interface selection for a packet is totally route-driven. * A valid alternative to IFF_LOOPBACK can be IFF_BROADCAST or * IFF_POINTOPOINT, but it would result in different properties * of the interface. */ ifp->if_flags = IFF_LOOPBACK | IFF_MULTICAST; ifp->if_drv_flags = IFF_DRV_RUNNING; ifp->if_ioctl = discioctl; ifp->if_output = discoutput; ifp->if_hdrlen = 0; ifp->if_addrlen = 0; ifp->if_snd.ifq_maxlen = 20; if_attach(ifp); bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); return (0); } static void disc_clone_destroy(struct ifnet *ifp) { struct disc_softc *sc; sc = ifp->if_softc; bpfdetach(ifp); if_detach(ifp); if_free(ifp); free(sc, M_DISC); } static void vnet_disc_init(const void *unused __unused) { V_disc_cloner = if_clone_simple(discname, disc_clone_create, disc_clone_destroy, 0); } -VNET_SYSINIT(vnet_disc_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY, +VNET_SYSINIT(vnet_disc_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_disc_init, NULL); static void vnet_disc_uninit(const void *unused __unused) { if_clone_detach(V_disc_cloner); } -VNET_SYSUNINIT(vnet_disc_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY, +VNET_SYSUNINIT(vnet_disc_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_disc_uninit, NULL); static int disc_modevent(module_t mod, int type, void *data) { switch (type) { case MOD_LOAD: case MOD_UNLOAD: break; default: return (EOPNOTSUPP); } return (0); } static moduledata_t disc_mod = { "if_disc", disc_modevent, NULL }; DECLARE_MODULE(if_disc, disc_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); static int discoutput(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, struct route *ro) { u_int32_t af; M_ASSERTPKTHDR(m); /* BPF writes need to be handled specially. */ if (dst->sa_family == AF_UNSPEC) bcopy(dst->sa_data, &af, sizeof(af)); else af = dst->sa_family; if (bpf_peers_present(ifp->if_bpf)) bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); m->m_pkthdr.rcvif = ifp; if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len); m_freem(m); return (0); } /* * Process an ioctl request. */ static int discioctl(struct ifnet *ifp, u_long cmd, caddr_t data) { struct ifreq *ifr = (struct ifreq *)data; int error = 0; switch (cmd) { case SIOCSIFADDR: ifp->if_flags |= IFF_UP; /* * Everything else is done at a higher level. */ break; case SIOCADDMULTI: case SIOCDELMULTI: if (ifr == 0) { error = EAFNOSUPPORT; /* XXX */ break; } switch (ifr->ifr_addr.sa_family) { #ifdef INET case AF_INET: break; #endif #ifdef INET6 case AF_INET6: break; #endif default: error = EAFNOSUPPORT; break; } break; case SIOCSIFMTU: ifp->if_mtu = ifr->ifr_mtu; break; default: error = EINVAL; } return (error); } Index: projects/vnet/sys/net/if_enc.c =================================================================== --- projects/vnet/sys/net/if_enc.c (revision 294565) +++ projects/vnet/sys/net/if_enc.c (revision 294566) @@ -1,404 +1,404 @@ /*- * Copyright (c) 2006 The FreeBSD Project. * Copyright (c) 2015 Andrey V. Elsukov * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #include "opt_inet.h" #include "opt_inet6.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #endif #include #include #define ENCMTU (1024+512) /* XXX this define must have the same value as in OpenBSD */ #define M_CONF 0x0400 /* payload was encrypted (ESP-transport) */ #define M_AUTH 0x0800 /* payload was authenticated (AH or ESP auth) */ #define M_AUTH_AH 0x2000 /* header was authenticated (AH) */ struct enchdr { u_int32_t af; u_int32_t spi; u_int32_t flags; }; struct enc_softc { struct ifnet *sc_ifp; }; static VNET_DEFINE(struct enc_softc *, enc_sc); #define V_enc_sc VNET(enc_sc) static VNET_DEFINE(struct if_clone *, enc_cloner); #define V_enc_cloner VNET(enc_cloner) static int enc_ioctl(struct ifnet *, u_long, caddr_t); static int enc_output(struct ifnet *, struct mbuf *, const struct sockaddr *, struct route *); static int enc_clone_create(struct if_clone *, int, caddr_t); static void enc_clone_destroy(struct ifnet *); static int enc_add_hhooks(struct enc_softc *); static void enc_remove_hhooks(struct enc_softc *); static const char encname[] = "enc"; /* * Before and after are relative to when we are stripping the * outer IP header. */ static VNET_DEFINE(int, filter_mask_in) = IPSEC_ENC_BEFORE; static VNET_DEFINE(int, bpf_mask_in) = IPSEC_ENC_BEFORE; static VNET_DEFINE(int, filter_mask_out) = IPSEC_ENC_BEFORE; static VNET_DEFINE(int, bpf_mask_out) = IPSEC_ENC_BEFORE | IPSEC_ENC_AFTER; #define V_filter_mask_in VNET(filter_mask_in) #define V_bpf_mask_in VNET(bpf_mask_in) #define V_filter_mask_out VNET(filter_mask_out) #define V_bpf_mask_out VNET(bpf_mask_out) static SYSCTL_NODE(_net, OID_AUTO, enc, CTLFLAG_RW, 0, "enc sysctl"); static SYSCTL_NODE(_net_enc, OID_AUTO, in, CTLFLAG_RW, 0, "enc input sysctl"); static SYSCTL_NODE(_net_enc, OID_AUTO, out, CTLFLAG_RW, 0, "enc output sysctl"); SYSCTL_INT(_net_enc_in, OID_AUTO, ipsec_filter_mask, CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(filter_mask_in), 0, "IPsec input firewall filter mask"); SYSCTL_INT(_net_enc_in, OID_AUTO, ipsec_bpf_mask, CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(bpf_mask_in), 0, "IPsec input bpf mask"); SYSCTL_INT(_net_enc_out, OID_AUTO, ipsec_filter_mask, CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(filter_mask_out), 0, "IPsec output firewall filter mask"); SYSCTL_INT(_net_enc_out, OID_AUTO, ipsec_bpf_mask, CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(bpf_mask_out), 0, "IPsec output bpf mask"); static void enc_clone_destroy(struct ifnet *ifp) { struct enc_softc *sc; sc = ifp->if_softc; KASSERT(sc == V_enc_sc, ("sc != ifp->if_softc")); enc_remove_hhooks(sc); bpfdetach(ifp); if_detach(ifp); if_free(ifp); free(sc, M_DEVBUF); V_enc_sc = NULL; } static int enc_clone_create(struct if_clone *ifc, int unit, caddr_t params) { struct ifnet *ifp; struct enc_softc *sc; sc = malloc(sizeof(struct enc_softc), M_DEVBUF, M_WAITOK | M_ZERO); ifp = sc->sc_ifp = if_alloc(IFT_ENC); if (ifp == NULL) { free(sc, M_DEVBUF); return (ENOSPC); } if (V_enc_sc != NULL) { if_free(ifp); free(sc, M_DEVBUF); return (EEXIST); } V_enc_sc = sc; if_initname(ifp, encname, unit); ifp->if_mtu = ENCMTU; ifp->if_ioctl = enc_ioctl; ifp->if_output = enc_output; ifp->if_softc = sc; if_attach(ifp); bpfattach(ifp, DLT_ENC, sizeof(struct enchdr)); if (enc_add_hhooks(sc) != 0) { enc_clone_destroy(ifp); return (ENXIO); } return (0); } static int enc_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, struct route *ro) { m_freem(m); return (0); } static int enc_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) { if (cmd != SIOCSIFFLAGS) return (EINVAL); if (ifp->if_flags & IFF_UP) ifp->if_drv_flags |= IFF_DRV_RUNNING; else ifp->if_drv_flags &= ~IFF_DRV_RUNNING; return (0); } /* * One helper hook function is used by any hook points. * + from hhook_type we can determine the packet direction: * HHOOK_TYPE_IPSEC_IN or HHOOK_TYPE_IPSEC_OUT; * + from hhook_id we can determine address family: AF_INET or AF_INET6; * + udata contains pointer to enc_softc; * + ctx_data contains pointer to struct ipsec_ctx_data. */ static int enc_hhook(int32_t hhook_type, int32_t hhook_id, void *udata, void *ctx_data, void *hdata, struct osd *hosd) { struct enchdr hdr; struct ipsec_ctx_data *ctx; struct enc_softc *sc; struct ifnet *ifp, *rcvif; struct pfil_head *ph; int pdir; sc = (struct enc_softc *)udata; ifp = sc->sc_ifp; if ((ifp->if_flags & IFF_UP) == 0) return (0); ctx = (struct ipsec_ctx_data *)ctx_data; /* XXX: wrong hook point was used by caller? */ if (ctx->af != hhook_id) return (EPFNOSUPPORT); if (((hhook_type == HHOOK_TYPE_IPSEC_IN && (ctx->enc & V_bpf_mask_in) != 0) || (hhook_type == HHOOK_TYPE_IPSEC_OUT && (ctx->enc & V_bpf_mask_out) != 0)) && bpf_peers_present(ifp->if_bpf) != 0) { hdr.af = ctx->af; hdr.spi = ctx->sav->spi; hdr.flags = 0; if (ctx->sav->alg_enc != SADB_EALG_NONE) hdr.flags |= M_CONF; if (ctx->sav->alg_auth != SADB_AALG_NONE) hdr.flags |= M_AUTH; bpf_mtap2(ifp->if_bpf, &hdr, sizeof(hdr), *ctx->mp); } switch (hhook_type) { case HHOOK_TYPE_IPSEC_IN: if (ctx->enc == IPSEC_ENC_BEFORE) { /* Do accounting only once */ if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1); if_inc_counter(ifp, IFCOUNTER_IBYTES, (*ctx->mp)->m_pkthdr.len); } if ((ctx->enc & V_filter_mask_in) == 0) return (0); /* skip pfil processing */ pdir = PFIL_IN; break; case HHOOK_TYPE_IPSEC_OUT: if (ctx->enc == IPSEC_ENC_BEFORE) { /* Do accounting only once */ if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); if_inc_counter(ifp, IFCOUNTER_OBYTES, (*ctx->mp)->m_pkthdr.len); } if ((ctx->enc & V_filter_mask_out) == 0) return (0); /* skip pfil processing */ pdir = PFIL_OUT; break; default: return (EINVAL); } switch (hhook_id) { #ifdef INET case AF_INET: ph = &V_inet_pfil_hook; break; #endif #ifdef INET6 case AF_INET6: ph = &V_inet6_pfil_hook; break; #endif default: ph = NULL; } if (ph == NULL || !PFIL_HOOKED(ph)) return (0); /* Make a packet looks like it was received on enc(4) */ rcvif = (*ctx->mp)->m_pkthdr.rcvif; (*ctx->mp)->m_pkthdr.rcvif = ifp; if (pfil_run_hooks(ph, ctx->mp, ifp, pdir, NULL) != 0 || *ctx->mp == NULL) { *ctx->mp = NULL; /* consumed by filter */ return (EACCES); } (*ctx->mp)->m_pkthdr.rcvif = rcvif; return (0); } static int enc_add_hhooks(struct enc_softc *sc) { struct hookinfo hki; int error; error = EPFNOSUPPORT; hki.hook_func = enc_hhook; hki.hook_helper = NULL; hki.hook_udata = sc; #ifdef INET hki.hook_id = AF_INET; hki.hook_type = HHOOK_TYPE_IPSEC_IN; error = hhook_add_hook(V_ipsec_hhh_in[HHOOK_IPSEC_INET], &hki, HHOOK_WAITOK); if (error != 0) return (error); hki.hook_type = HHOOK_TYPE_IPSEC_OUT; error = hhook_add_hook(V_ipsec_hhh_out[HHOOK_IPSEC_INET], &hki, HHOOK_WAITOK); if (error != 0) return (error); #endif #ifdef INET6 hki.hook_id = AF_INET6; hki.hook_type = HHOOK_TYPE_IPSEC_IN; error = hhook_add_hook(V_ipsec_hhh_in[HHOOK_IPSEC_INET6], &hki, HHOOK_WAITOK); if (error != 0) return (error); hki.hook_type = HHOOK_TYPE_IPSEC_OUT; error = hhook_add_hook(V_ipsec_hhh_out[HHOOK_IPSEC_INET6], &hki, HHOOK_WAITOK); if (error != 0) return (error); #endif return (error); } static void enc_remove_hhooks(struct enc_softc *sc) { struct hookinfo hki; hki.hook_func = enc_hhook; hki.hook_helper = NULL; hki.hook_udata = sc; #ifdef INET hki.hook_id = AF_INET; hki.hook_type = HHOOK_TYPE_IPSEC_IN; hhook_remove_hook(V_ipsec_hhh_in[HHOOK_IPSEC_INET], &hki); hki.hook_type = HHOOK_TYPE_IPSEC_OUT; hhook_remove_hook(V_ipsec_hhh_out[HHOOK_IPSEC_INET], &hki); #endif #ifdef INET6 hki.hook_id = AF_INET6; hki.hook_type = HHOOK_TYPE_IPSEC_IN; hhook_remove_hook(V_ipsec_hhh_in[HHOOK_IPSEC_INET6], &hki); hki.hook_type = HHOOK_TYPE_IPSEC_OUT; hhook_remove_hook(V_ipsec_hhh_out[HHOOK_IPSEC_INET6], &hki); #endif } static void vnet_enc_init(const void *unused __unused) { V_enc_sc = NULL; V_enc_cloner = if_clone_simple(encname, enc_clone_create, enc_clone_destroy, 1); } -VNET_SYSINIT(vnet_enc_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY, +VNET_SYSINIT(vnet_enc_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_enc_init, NULL); static void vnet_enc_uninit(const void *unused __unused) { if_clone_detach(V_enc_cloner); } -VNET_SYSUNINIT(vnet_enc_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY, +VNET_SYSUNINIT(vnet_enc_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_enc_uninit, NULL); static int enc_modevent(module_t mod, int type, void *data) { switch (type) { case MOD_LOAD: case MOD_UNLOAD: break; default: return (EOPNOTSUPP); } return (0); } static moduledata_t enc_mod = { "if_enc", enc_modevent, 0 }; -DECLARE_MODULE(if_enc, enc_mod, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY); +DECLARE_MODULE(if_enc, enc_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); Index: projects/vnet/sys/net/pfil.c =================================================================== --- projects/vnet/sys/net/pfil.c (revision 294565) +++ projects/vnet/sys/net/pfil.c (revision 294566) @@ -1,407 +1,402 @@ /* $FreeBSD$ */ /* $NetBSD: pfil.c,v 1.20 2001/11/12 23:49:46 lukem Exp $ */ /*- * Copyright (c) 1996 Matthew R. Green * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static struct mtx pfil_global_lock; MTX_SYSINIT(pfil_heads_lock, &pfil_global_lock, "pfil_head_list lock", MTX_DEF); static struct packet_filter_hook *pfil_chain_get(int, struct pfil_head *); static int pfil_chain_add(pfil_chain_t *, struct packet_filter_hook *, int); static int pfil_chain_remove(pfil_chain_t *, pfil_func_t, void *); LIST_HEAD(pfilheadhead, pfil_head); VNET_DEFINE(struct pfilheadhead, pfil_head_list); #define V_pfil_head_list VNET(pfil_head_list) VNET_DEFINE(struct rmlock, pfil_lock); #define V_pfil_lock VNET(pfil_lock) /* * pfil_run_hooks() runs the specified packet filter hook chain. */ int pfil_run_hooks(struct pfil_head *ph, struct mbuf **mp, struct ifnet *ifp, int dir, struct inpcb *inp) { struct rm_priotracker rmpt; struct packet_filter_hook *pfh; struct mbuf *m = *mp; int rv = 0; PFIL_RLOCK(ph, &rmpt); KASSERT(ph->ph_nhooks >= 0, ("Pfil hook count dropped < 0")); for (pfh = pfil_chain_get(dir, ph); pfh != NULL; pfh = TAILQ_NEXT(pfh, pfil_chain)) { if (pfh->pfil_func != NULL) { rv = (*pfh->pfil_func)(pfh->pfil_arg, &m, ifp, dir, inp); if (rv != 0 || m == NULL) break; } } PFIL_RUNLOCK(ph, &rmpt); *mp = m; return (rv); } static struct packet_filter_hook * pfil_chain_get(int dir, struct pfil_head *ph) { if (dir == PFIL_IN) return (TAILQ_FIRST(&ph->ph_in)); else if (dir == PFIL_OUT) return (TAILQ_FIRST(&ph->ph_out)); else return (NULL); } /* * pfil_try_rlock() acquires rm reader lock for specified head * if this is immediately possible. */ int pfil_try_rlock(struct pfil_head *ph, struct rm_priotracker *tracker) { return (PFIL_TRY_RLOCK(ph, tracker)); } /* * pfil_rlock() acquires rm reader lock for specified head. */ void pfil_rlock(struct pfil_head *ph, struct rm_priotracker *tracker) { PFIL_RLOCK(ph, tracker); } /* * pfil_runlock() releases reader lock for specified head. */ void pfil_runlock(struct pfil_head *ph, struct rm_priotracker *tracker) { PFIL_RUNLOCK(ph, tracker); } /* * pfil_wlock() acquires writer lock for specified head. */ void pfil_wlock(struct pfil_head *ph) { PFIL_WLOCK(ph); } /* * pfil_wunlock() releases writer lock for specified head. */ void pfil_wunlock(struct pfil_head *ph) { PFIL_WUNLOCK(ph); } /* * pfil_wowned() returns a non-zero value if the current thread owns * an exclusive lock. */ int pfil_wowned(struct pfil_head *ph) { return (PFIL_WOWNED(ph)); } /* * pfil_head_register() registers a pfil_head with the packet filter hook * mechanism. */ int pfil_head_register(struct pfil_head *ph) { struct pfil_head *lph; PFIL_HEADLIST_LOCK(); LIST_FOREACH(lph, &V_pfil_head_list, ph_list) { if (ph->ph_type == lph->ph_type && ph->ph_un.phu_val == lph->ph_un.phu_val) { PFIL_HEADLIST_UNLOCK(); return (EEXIST); } } PFIL_LOCK_INIT(ph); ph->ph_nhooks = 0; TAILQ_INIT(&ph->ph_in); TAILQ_INIT(&ph->ph_out); LIST_INSERT_HEAD(&V_pfil_head_list, ph, ph_list); PFIL_HEADLIST_UNLOCK(); return (0); } /* * pfil_head_unregister() removes a pfil_head from the packet filter hook * mechanism. The producer of the hook promises that all outstanding * invocations of the hook have completed before it unregisters the hook. */ int pfil_head_unregister(struct pfil_head *ph) { struct packet_filter_hook *pfh, *pfnext; PFIL_HEADLIST_LOCK(); LIST_REMOVE(ph, ph_list); PFIL_HEADLIST_UNLOCK(); TAILQ_FOREACH_SAFE(pfh, &ph->ph_in, pfil_chain, pfnext) free(pfh, M_IFADDR); TAILQ_FOREACH_SAFE(pfh, &ph->ph_out, pfil_chain, pfnext) free(pfh, M_IFADDR); PFIL_LOCK_DESTROY(ph); return (0); } /* * pfil_head_get() returns the pfil_head for a given key/dlt. */ struct pfil_head * pfil_head_get(int type, u_long val) { struct pfil_head *ph; PFIL_HEADLIST_LOCK(); LIST_FOREACH(ph, &V_pfil_head_list, ph_list) if (ph->ph_type == type && ph->ph_un.phu_val == val) break; PFIL_HEADLIST_UNLOCK(); return (ph); } /* * pfil_add_hook() adds a function to the packet filter hook. the * flags are: * PFIL_IN call me on incoming packets * PFIL_OUT call me on outgoing packets * PFIL_ALL call me on all of the above * PFIL_WAITOK OK to call malloc with M_WAITOK. */ int pfil_add_hook(pfil_func_t func, void *arg, int flags, struct pfil_head *ph) { struct packet_filter_hook *pfh1 = NULL; struct packet_filter_hook *pfh2 = NULL; int err; if (flags & PFIL_IN) { pfh1 = (struct packet_filter_hook *)malloc(sizeof(*pfh1), M_IFADDR, (flags & PFIL_WAITOK) ? M_WAITOK : M_NOWAIT); if (pfh1 == NULL) { err = ENOMEM; goto error; } } if (flags & PFIL_OUT) { pfh2 = (struct packet_filter_hook *)malloc(sizeof(*pfh1), M_IFADDR, (flags & PFIL_WAITOK) ? M_WAITOK : M_NOWAIT); if (pfh2 == NULL) { err = ENOMEM; goto error; } } PFIL_WLOCK(ph); if (flags & PFIL_IN) { pfh1->pfil_func = func; pfh1->pfil_arg = arg; err = pfil_chain_add(&ph->ph_in, pfh1, flags & ~PFIL_OUT); if (err) goto locked_error; ph->ph_nhooks++; } if (flags & PFIL_OUT) { pfh2->pfil_func = func; pfh2->pfil_arg = arg; err = pfil_chain_add(&ph->ph_out, pfh2, flags & ~PFIL_IN); if (err) { if (flags & PFIL_IN) pfil_chain_remove(&ph->ph_in, func, arg); goto locked_error; } ph->ph_nhooks++; } PFIL_WUNLOCK(ph); return (0); locked_error: PFIL_WUNLOCK(ph); error: if (pfh1 != NULL) free(pfh1, M_IFADDR); if (pfh2 != NULL) free(pfh2, M_IFADDR); return (err); } /* * pfil_remove_hook removes a specific function from the packet filter hook * chain. */ int pfil_remove_hook(pfil_func_t func, void *arg, int flags, struct pfil_head *ph) { int err = 0; PFIL_WLOCK(ph); if (flags & PFIL_IN) { err = pfil_chain_remove(&ph->ph_in, func, arg); if (err == 0) ph->ph_nhooks--; } if ((err == 0) && (flags & PFIL_OUT)) { err = pfil_chain_remove(&ph->ph_out, func, arg); if (err == 0) ph->ph_nhooks--; } PFIL_WUNLOCK(ph); return (err); } /* * Internal: Add a new pfil hook into a hook chain. */ static int pfil_chain_add(pfil_chain_t *chain, struct packet_filter_hook *pfh1, int flags) { struct packet_filter_hook *pfh; /* * First make sure the hook is not already there. */ TAILQ_FOREACH(pfh, chain, pfil_chain) if (pfh->pfil_func == pfh1->pfil_func && pfh->pfil_arg == pfh1->pfil_arg) return (EEXIST); /* * Insert the input list in reverse order of the output list so that * the same path is followed in or out of the kernel. */ if (flags & PFIL_IN) TAILQ_INSERT_HEAD(chain, pfh1, pfil_chain); else TAILQ_INSERT_TAIL(chain, pfh1, pfil_chain); return (0); } /* * Internal: Remove a pfil hook from a hook chain. */ static int pfil_chain_remove(pfil_chain_t *chain, pfil_func_t func, void *arg) { struct packet_filter_hook *pfh; TAILQ_FOREACH(pfh, chain, pfil_chain) if (pfh->pfil_func == func && pfh->pfil_arg == arg) { TAILQ_REMOVE(chain, pfh, pfil_chain); free(pfh, M_IFADDR); return (0); } return (ENOENT); } /* * Stuff that must be initialized for every instance (including the first of * course). */ -static int +static void vnet_pfil_init(const void *unused) { LIST_INIT(&V_pfil_head_list); PFIL_LOCK_INIT_REAL(&V_pfil_lock, "shared"); - return (0); } /* * Called for the removal of each instance. */ -static int +static void vnet_pfil_uninit(const void *unused) { - KASSERT(LIST_EMPTY(&V_pfil_head_list), + VNET_ASSERT(LIST_EMPTY(&V_pfil_head_list), ("%s: pfil_head_list %p not empty", __func__, &V_pfil_head_list)); PFIL_LOCK_DESTROY_REAL(&V_pfil_lock); - return (0); } -/* Define startup order. */ -#define PFIL_SYSINIT_ORDER SI_SUB_PROTO_BEGIN -#define PFIL_MODEVENT_ORDER (SI_ORDER_FIRST) /* On boot slot in here. */ -#define PFIL_VNET_ORDER (PFIL_MODEVENT_ORDER + 2) /* Later still. */ - /* * Starting up. * * VNET_SYSINIT is called for each existing vnet and each new vnet. + * Make sure the pfil bits are first before any possible subsystem which + * might piggyback on the SI_SUB_PROTO_PFIL. */ -VNET_SYSINIT(vnet_pfil_init, PFIL_SYSINIT_ORDER, PFIL_VNET_ORDER, +VNET_SYSINIT(vnet_pfil_init, SI_SUB_PROTO_PFIL, SI_ORDER_FIRST, vnet_pfil_init, NULL); /* * Closing up shop. These are done in REVERSE ORDER. Not called on reboot. * * VNET_SYSUNINIT is called for each exiting vnet as it exits. */ -VNET_SYSUNINIT(vnet_pfil_uninit, PFIL_SYSINIT_ORDER, PFIL_VNET_ORDER, +VNET_SYSUNINIT(vnet_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_FIRST, vnet_pfil_uninit, NULL); Index: projects/vnet/sys/net/vnet.c =================================================================== --- projects/vnet/sys/net/vnet.c (revision 294565) +++ projects/vnet/sys/net/vnet.c (revision 294566) @@ -1,782 +1,774 @@ /*- * Copyright (c) 2004-2009 University of Zagreb * Copyright (c) 2006-2009 FreeBSD Foundation * All rights reserved. * * This software was developed by the University of Zagreb and the * FreeBSD Foundation under sponsorship by the Stichting NLnet and the * FreeBSD Foundation. * * Copyright (c) 2009 Jeffrey Roberson * Copyright (c) 2009 Robert N. M. Watson * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include "opt_kdb.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #include #endif #include #include #include /*- * This file implements core functions for virtual network stacks: * * - Virtual network stack management functions. * * - Virtual network stack memory allocator, which virtualizes global * variables in the network stack * * - Virtualized SYSINIT's/SYSUNINIT's, which allow network stack subsystems * to register startup/shutdown events to be run for each virtual network * stack instance. */ FEATURE(vimage, "VIMAGE kernel virtualization"); static MALLOC_DEFINE(M_VNET, "vnet", "network stack control block"); /* * The virtual network stack list has two read-write locks, one sleepable and * the other not, so that the list can be stablized and walked in a variety * of network stack contexts. Both must be acquired exclusively to modify * the list, but a read lock of either lock is sufficient to walk the list. */ struct rwlock vnet_rwlock; struct sx vnet_sxlock; #define VNET_LIST_WLOCK() do { \ sx_xlock(&vnet_sxlock); \ rw_wlock(&vnet_rwlock); \ } while (0) #define VNET_LIST_WUNLOCK() do { \ rw_wunlock(&vnet_rwlock); \ sx_xunlock(&vnet_sxlock); \ } while (0) struct vnet_list_head vnet_head; struct vnet *vnet0; /* * The virtual network stack allocator provides storage for virtualized * global variables. These variables are defined/declared using the * VNET_DEFINE()/VNET_DECLARE() macros, which place them in the 'set_vnet' * linker set. The details of the implementation are somewhat subtle, but * allow the majority of most network subsystems to maintain * virtualization-agnostic. * * The virtual network stack allocator handles variables in the base kernel * vs. modules in similar but different ways. In both cases, virtualized * global variables are marked as such by being declared to be part of the * vnet linker set. These "master" copies of global variables serve two * functions: * * (1) They contain static initialization or "default" values for global * variables which will be propagated to each virtual network stack * instance when created. As with normal global variables, they default * to zero-filled. * * (2) They act as unique global names by which the variable can be referred * to, regardless of network stack instance. The single global symbol * will be used to calculate the location of a per-virtual instance * variable at run-time. * * Each virtual network stack instance has a complete copy of each * virtualized global variable, stored in a malloc'd block of memory * referred to by vnet->vnet_data_mem. Critical to the design is that each * per-instance memory block is laid out identically to the master block so * that the offset of each global variable is the same across all blocks. To * optimize run-time access, a precalculated 'base' address, * vnet->vnet_data_base, is stored in each vnet, and is the amount that can * be added to the address of a 'master' instance of a variable to get to the * per-vnet instance. * * Virtualized global variables are handled in a similar manner, but as each * module has its own 'set_vnet' linker set, and we want to keep all * virtualized globals togther, we reserve space in the kernel's linker set * for potential module variables using a per-vnet character array, * 'modspace'. The virtual network stack allocator maintains a free list to * track what space in the array is free (all, initially) and as modules are * linked, allocates portions of the space to specific globals. The kernel * module linker queries the virtual network stack allocator and will * bind references of the global to the location during linking. It also * calls into the virtual network stack allocator, once the memory is * initialized, in order to propagate the new static initializations to all * existing virtual network stack instances so that the soon-to-be executing * module will find every network stack instance with proper default values. */ /* * Number of bytes of data in the 'set_vnet' linker set, and hence the total * size of all kernel virtualized global variables, and the malloc(9) type * that will be used to allocate it. */ #define VNET_BYTES (VNET_STOP - VNET_START) static MALLOC_DEFINE(M_VNET_DATA, "vnet_data", "VNET data"); /* * VNET_MODMIN is the minimum number of bytes we will reserve for the sum of * global variables across all loaded modules. As this actually sizes an * array declared as a virtualized global variable in the kernel itself, and * we want the virtualized global variable space to be page-sized, we may * have more space than that in practice. */ #define VNET_MODMIN 8192 #define VNET_SIZE roundup2(VNET_BYTES, PAGE_SIZE) #define VNET_MODSIZE (VNET_SIZE - (VNET_BYTES - VNET_MODMIN)) /* * Space to store virtualized global variables from loadable kernel modules, * and the free list to manage it. */ static VNET_DEFINE(char, modspace[VNET_MODMIN]); /* * Global lists of subsystem constructor and destructors for vnets. They are * registered via VNET_SYSINIT() and VNET_SYSUNINIT(). Both lists are * protected by the vnet_sysinit_sxlock global lock. */ static TAILQ_HEAD(vnet_sysinit_head, vnet_sysinit) vnet_constructors = TAILQ_HEAD_INITIALIZER(vnet_constructors); static TAILQ_HEAD(vnet_sysuninit_head, vnet_sysinit) vnet_destructors = TAILQ_HEAD_INITIALIZER(vnet_destructors); struct sx vnet_sysinit_sxlock; #define VNET_SYSINIT_WLOCK() sx_xlock(&vnet_sysinit_sxlock); #define VNET_SYSINIT_WUNLOCK() sx_xunlock(&vnet_sysinit_sxlock); #define VNET_SYSINIT_RLOCK() sx_slock(&vnet_sysinit_sxlock); #define VNET_SYSINIT_RUNLOCK() sx_sunlock(&vnet_sysinit_sxlock); struct vnet_data_free { uintptr_t vnd_start; int vnd_len; TAILQ_ENTRY(vnet_data_free) vnd_link; }; static MALLOC_DEFINE(M_VNET_DATA_FREE, "vnet_data_free", "VNET resource accounting"); static TAILQ_HEAD(, vnet_data_free) vnet_data_free_head = TAILQ_HEAD_INITIALIZER(vnet_data_free_head); static struct sx vnet_data_free_lock; SDT_PROVIDER_DEFINE(vnet); SDT_PROBE_DEFINE1(vnet, functions, vnet_alloc, entry, "int"); SDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, alloc, "int", "struct vnet *"); SDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, return, "int", "struct vnet *"); SDT_PROBE_DEFINE2(vnet, functions, vnet_destroy, entry, "int", "struct vnet *"); SDT_PROBE_DEFINE1(vnet, functions, vnet_destroy, return, "int"); #ifdef DDB static void db_show_vnet_print_vs(struct vnet_sysinit *, int); #endif /* * Allocate a virtual network stack. */ struct vnet * vnet_alloc(void) { struct vnet *vnet; SDT_PROBE1(vnet, functions, vnet_alloc, entry, __LINE__); vnet = malloc(sizeof(struct vnet), M_VNET, M_WAITOK | M_ZERO); vnet->vnet_magic_n = VNET_MAGIC_N; SDT_PROBE2(vnet, functions, vnet_alloc, alloc, __LINE__, vnet); /* * Allocate storage for virtualized global variables and copy in * initial values form our 'master' copy. */ vnet->vnet_data_mem = malloc(VNET_SIZE, M_VNET_DATA, M_WAITOK); memcpy(vnet->vnet_data_mem, (void *)VNET_START, VNET_BYTES); /* * All use of vnet-specific data will immediately subtract VNET_START * from the base memory pointer, so pre-calculate that now to avoid * it on each use. */ vnet->vnet_data_base = (uintptr_t)vnet->vnet_data_mem - VNET_START; /* Initialize / attach vnet module instances. */ CURVNET_SET_QUIET(vnet); vnet_sysinit(); CURVNET_RESTORE(); VNET_LIST_WLOCK(); LIST_INSERT_HEAD(&vnet_head, vnet, vnet_le); VNET_LIST_WUNLOCK(); SDT_PROBE2(vnet, functions, vnet_alloc, return, __LINE__, vnet); return (vnet); } /* * Destroy a virtual network stack. */ void vnet_destroy(struct vnet *vnet) { - struct ifnet *ifp, *nifp; SDT_PROBE2(vnet, functions, vnet_destroy, entry, __LINE__, vnet); KASSERT(vnet->vnet_sockcnt == 0, ("%s: vnet still has sockets", __func__)); VNET_LIST_WLOCK(); LIST_REMOVE(vnet, vnet_le); VNET_LIST_WUNLOCK(); CURVNET_SET_QUIET(vnet); - - /* Return all inherited interfaces to their parent vnets. */ - TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { - if (ifp->if_home_vnet != ifp->if_vnet) - if_vmove(ifp, ifp->if_home_vnet); - } - vnet_sysuninit(); CURVNET_RESTORE(); /* * Release storage for the virtual network stack instance. */ free(vnet->vnet_data_mem, M_VNET_DATA); vnet->vnet_data_mem = NULL; vnet->vnet_data_base = 0; vnet->vnet_magic_n = 0xdeadbeef; free(vnet, M_VNET); SDT_PROBE1(vnet, functions, vnet_destroy, return, __LINE__); } /* * Boot time initialization and allocation of virtual network stacks. */ static void vnet_init_prelink(void *arg) { rw_init(&vnet_rwlock, "vnet_rwlock"); sx_init(&vnet_sxlock, "vnet_sxlock"); sx_init(&vnet_sysinit_sxlock, "vnet_sysinit_sxlock"); LIST_INIT(&vnet_head); } SYSINIT(vnet_init_prelink, SI_SUB_VNET_PRELINK, SI_ORDER_FIRST, vnet_init_prelink, NULL); static void vnet0_init(void *arg) { /* Warn people before take off - in case we crash early. */ printf("WARNING: VIMAGE (virtualized network stack) is a highly " "experimental feature.\n"); /* * We MUST clear curvnet in vi_init_done() before going SMP, * otherwise CURVNET_SET() macros would scream about unnecessary * curvnet recursions. */ curvnet = prison0.pr_vnet = vnet0 = vnet_alloc(); } SYSINIT(vnet0_init, SI_SUB_VNET, SI_ORDER_FIRST, vnet0_init, NULL); static void vnet_init_done(void *unused) { curvnet = NULL; } SYSINIT(vnet_init_done, SI_SUB_VNET_DONE, SI_ORDER_FIRST, vnet_init_done, NULL); /* * Once on boot, initialize the modspace freelist to entirely cover modspace. */ static void vnet_data_startup(void *dummy __unused) { struct vnet_data_free *df; df = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO); df->vnd_start = (uintptr_t)&VNET_NAME(modspace); df->vnd_len = VNET_MODMIN; TAILQ_INSERT_HEAD(&vnet_data_free_head, df, vnd_link); sx_init(&vnet_data_free_lock, "vnet_data alloc lock"); } SYSINIT(vnet_data, SI_SUB_KLD, SI_ORDER_FIRST, vnet_data_startup, 0); /* * When a module is loaded and requires storage for a virtualized global * variable, allocate space from the modspace free list. This interface * should be used only by the kernel linker. */ void * vnet_data_alloc(int size) { struct vnet_data_free *df; void *s; s = NULL; size = roundup2(size, sizeof(void *)); sx_xlock(&vnet_data_free_lock); TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) { if (df->vnd_len < size) continue; if (df->vnd_len == size) { s = (void *)df->vnd_start; TAILQ_REMOVE(&vnet_data_free_head, df, vnd_link); free(df, M_VNET_DATA_FREE); break; } s = (void *)df->vnd_start; df->vnd_len -= size; df->vnd_start = df->vnd_start + size; break; } sx_xunlock(&vnet_data_free_lock); return (s); } /* * Free space for a virtualized global variable on module unload. */ void vnet_data_free(void *start_arg, int size) { struct vnet_data_free *df; struct vnet_data_free *dn; uintptr_t start; uintptr_t end; size = roundup2(size, sizeof(void *)); start = (uintptr_t)start_arg; end = start + size; /* * Free a region of space and merge it with as many neighbors as * possible. Keeping the list sorted simplifies this operation. */ sx_xlock(&vnet_data_free_lock); TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) { if (df->vnd_start > end) break; /* * If we expand at the end of an entry we may have to merge * it with the one following it as well. */ if (df->vnd_start + df->vnd_len == start) { df->vnd_len += size; dn = TAILQ_NEXT(df, vnd_link); if (df->vnd_start + df->vnd_len == dn->vnd_start) { df->vnd_len += dn->vnd_len; TAILQ_REMOVE(&vnet_data_free_head, dn, vnd_link); free(dn, M_VNET_DATA_FREE); } sx_xunlock(&vnet_data_free_lock); return; } if (df->vnd_start == end) { df->vnd_start = start; df->vnd_len += size; sx_xunlock(&vnet_data_free_lock); return; } } dn = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO); dn->vnd_start = start; dn->vnd_len = size; if (df) TAILQ_INSERT_BEFORE(df, dn, vnd_link); else TAILQ_INSERT_TAIL(&vnet_data_free_head, dn, vnd_link); sx_xunlock(&vnet_data_free_lock); } /* * When a new virtualized global variable has been allocated, propagate its * initial value to each already-allocated virtual network stack instance. */ void vnet_data_copy(void *start, int size) { struct vnet *vnet; VNET_LIST_RLOCK(); LIST_FOREACH(vnet, &vnet_head, vnet_le) memcpy((void *)((uintptr_t)vnet->vnet_data_base + (uintptr_t)start), start, size); VNET_LIST_RUNLOCK(); } /* * Support for special SYSINIT handlers registered via VNET_SYSINIT() * and VNET_SYSUNINIT(). */ void vnet_register_sysinit(void *arg) { struct vnet_sysinit *vs, *vs2; struct vnet *vnet; vs = arg; KASSERT(vs->subsystem > SI_SUB_VNET, ("vnet sysinit too early")); /* Add the constructor to the global list of vnet constructors. */ VNET_SYSINIT_WLOCK(); TAILQ_FOREACH(vs2, &vnet_constructors, link) { if (vs2->subsystem > vs->subsystem) break; if (vs2->subsystem == vs->subsystem && vs2->order > vs->order) break; } if (vs2 != NULL) TAILQ_INSERT_BEFORE(vs2, vs, link); else TAILQ_INSERT_TAIL(&vnet_constructors, vs, link); /* * Invoke the constructor on all the existing vnets when it is * registered. */ VNET_FOREACH(vnet) { CURVNET_SET_QUIET(vnet); vs->func(vs->arg); CURVNET_RESTORE(); } VNET_SYSINIT_WUNLOCK(); } void vnet_deregister_sysinit(void *arg) { struct vnet_sysinit *vs; vs = arg; /* Remove the constructor from the global list of vnet constructors. */ VNET_SYSINIT_WLOCK(); TAILQ_REMOVE(&vnet_constructors, vs, link); VNET_SYSINIT_WUNLOCK(); } void vnet_register_sysuninit(void *arg) { struct vnet_sysinit *vs, *vs2; vs = arg; /* Add the destructor to the global list of vnet destructors. */ VNET_SYSINIT_WLOCK(); TAILQ_FOREACH(vs2, &vnet_destructors, link) { if (vs2->subsystem > vs->subsystem) break; if (vs2->subsystem == vs->subsystem && vs2->order > vs->order) break; } if (vs2 != NULL) TAILQ_INSERT_BEFORE(vs2, vs, link); else TAILQ_INSERT_TAIL(&vnet_destructors, vs, link); VNET_SYSINIT_WUNLOCK(); } void vnet_deregister_sysuninit(void *arg) { struct vnet_sysinit *vs; struct vnet *vnet; vs = arg; /* * Invoke the destructor on all the existing vnets when it is * deregistered. */ VNET_SYSINIT_WLOCK(); VNET_FOREACH(vnet) { CURVNET_SET_QUIET(vnet); vs->func(vs->arg); CURVNET_RESTORE(); } /* Remove the destructor from the global list of vnet destructors. */ TAILQ_REMOVE(&vnet_destructors, vs, link); VNET_SYSINIT_WUNLOCK(); } /* * Invoke all registered vnet constructors on the current vnet. Used during * vnet construction. The caller is responsible for ensuring the new vnet is * the current vnet and that the vnet_sysinit_sxlock lock is locked. */ void vnet_sysinit(void) { struct vnet_sysinit *vs; VNET_SYSINIT_RLOCK(); TAILQ_FOREACH(vs, &vnet_constructors, link) { vs->func(vs->arg); } VNET_SYSINIT_RUNLOCK(); } /* * Invoke all registered vnet destructors on the current vnet. Used during * vnet destruction. The caller is responsible for ensuring the dying vnet * the current vnet and that the vnet_sysinit_sxlock lock is locked. */ void vnet_sysuninit(void) { struct vnet_sysinit *vs; VNET_SYSINIT_RLOCK(); TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head, link) { vs->func(vs->arg); } VNET_SYSINIT_RUNLOCK(); } /* * EVENTHANDLER(9) extensions. */ /* * Invoke the eventhandler function originally registered with the possibly * registered argument for all virtual network stack instances. * * This iterator can only be used for eventhandlers that do not take any * additional arguments, as we do ignore the variadic arguments from the * EVENTHANDLER_INVOKE() call. */ void vnet_global_eventhandler_iterator_func(void *arg, ...) { VNET_ITERATOR_DECL(vnet_iter); struct eventhandler_entry_vimage *v_ee; /* * There is a bug here in that we should actually cast things to * (struct eventhandler_entry_ ## name *) but that's not easily * possible in here so just re-using the variadic version we * defined for the generic vimage case. */ v_ee = arg; VNET_LIST_RLOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); ((vimage_iterator_func_t)v_ee->func)(v_ee->ee_arg); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK(); } #ifdef VNET_DEBUG struct vnet_recursion { SLIST_ENTRY(vnet_recursion) vnr_le; const char *prev_fn; const char *where_fn; int where_line; struct vnet *old_vnet; struct vnet *new_vnet; }; static SLIST_HEAD(, vnet_recursion) vnet_recursions = SLIST_HEAD_INITIALIZER(vnet_recursions); static void vnet_print_recursion(struct vnet_recursion *vnr, int brief) { if (!brief) printf("CURVNET_SET() recursion in "); printf("%s() line %d, prev in %s()", vnr->where_fn, vnr->where_line, vnr->prev_fn); if (brief) printf(", "); else printf("\n "); printf("%p -> %p\n", vnr->old_vnet, vnr->new_vnet); } void vnet_log_recursion(struct vnet *old_vnet, const char *old_fn, int line) { struct vnet_recursion *vnr; /* Skip already logged recursion events. */ SLIST_FOREACH(vnr, &vnet_recursions, vnr_le) if (vnr->prev_fn == old_fn && vnr->where_fn == curthread->td_vnet_lpush && vnr->where_line == line && (vnr->old_vnet == vnr->new_vnet) == (curvnet == old_vnet)) return; vnr = malloc(sizeof(*vnr), M_VNET, M_NOWAIT | M_ZERO); if (vnr == NULL) panic("%s: malloc failed", __func__); vnr->prev_fn = old_fn; vnr->where_fn = curthread->td_vnet_lpush; vnr->where_line = line; vnr->old_vnet = old_vnet; vnr->new_vnet = curvnet; SLIST_INSERT_HEAD(&vnet_recursions, vnr, vnr_le); vnet_print_recursion(vnr, 0); #ifdef KDB kdb_backtrace(); #endif } #endif /* VNET_DEBUG */ /* * DDB(4). */ #ifdef DDB DB_SHOW_COMMAND(vnets, db_show_vnets) { VNET_ITERATOR_DECL(vnet_iter); VNET_FOREACH(vnet_iter) { db_printf("vnet = %p\n", vnet_iter); db_printf(" vnet_magic_n = 0x%x (%s, orig 0x%x)\n", vnet_iter->vnet_magic_n, (vnet_iter->vnet_magic_n == VNET_MAGIC_N) ? "ok" : "mismatch", VNET_MAGIC_N); db_printf(" vnet_ifcnt = %u\n", vnet_iter->vnet_ifcnt); db_printf(" vnet_sockcnt = %u\n", vnet_iter->vnet_sockcnt); db_printf(" vnet_data_mem = %p\n", vnet_iter->vnet_data_mem); db_printf(" vnet_data_base = 0x%jx\n", (uintmax_t)vnet_iter->vnet_data_base); db_printf("\n"); if (db_pager_quit) break; } } static void db_show_vnet_print_vs(struct vnet_sysinit *vs, int ddb) { const char *vsname, *funcname; c_db_sym_t sym; db_expr_t offset; #define xprint(...) \ if (ddb) \ db_printf(__VA_ARGS__); \ else \ printf(__VA_ARGS__) if (vs == NULL) { xprint("%s: no vnet_sysinit * given\n", __func__); return; } sym = db_search_symbol((vm_offset_t)vs, DB_STGY_ANY, &offset); db_symbol_values(sym, &vsname, NULL); sym = db_search_symbol((vm_offset_t)vs->func, DB_STGY_PROC, &offset); db_symbol_values(sym, &funcname, NULL); xprint("%s(%p)\n", (vsname != NULL) ? vsname : "", vs); xprint(" 0x%08x 0x%08x\n", vs->subsystem, vs->order); xprint(" %p(%s)(%p)\n", vs->func, (funcname != NULL) ? funcname : "", vs->arg); #undef xprint } DB_SHOW_COMMAND(vnet_sysinit, db_show_vnet_sysinit) { struct vnet_sysinit *vs; db_printf("VNET_SYSINIT vs Name(Ptr)\n"); db_printf(" Subsystem Order\n"); db_printf(" Function(Name)(Arg)\n"); TAILQ_FOREACH(vs, &vnet_constructors, link) { db_show_vnet_print_vs(vs, 1); if (db_pager_quit) break; } } DB_SHOW_COMMAND(vnet_sysuninit, db_show_vnet_sysuninit) { struct vnet_sysinit *vs; db_printf("VNET_SYSUNINIT vs Name(Ptr)\n"); db_printf(" Subsystem Order\n"); db_printf(" Function(Name)(Arg)\n"); TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head, link) { db_show_vnet_print_vs(vs, 1); if (db_pager_quit) break; } } #ifdef VNET_DEBUG DB_SHOW_COMMAND(vnetrcrs, db_show_vnetrcrs) { struct vnet_recursion *vnr; SLIST_FOREACH(vnr, &vnet_recursions, vnr_le) vnet_print_recursion(vnr, 1); } #endif #endif /* DDB */ Index: projects/vnet/sys/netinet/igmp.c =================================================================== --- projects/vnet/sys/netinet/igmp.c (revision 294565) +++ projects/vnet/sys/netinet/igmp.c (revision 294566) @@ -1,3667 +1,3674 @@ /*- * Copyright (c) 2007-2009 Bruce Simpson. * Copyright (c) 1988 Stephen Deering. * Copyright (c) 1992, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Stephen Deering of Stanford University. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)igmp.c 8.1 (Berkeley) 7/19/93 */ /* * Internet Group Management Protocol (IGMP) routines. * [RFC1112, RFC2236, RFC3376] * * Written by Steve Deering, Stanford, May 1988. * Modified by Rosen Sharma, Stanford, Aug 1994. * Modified by Bill Fenner, Xerox PARC, Feb 1995. * Modified to fully comply to IGMPv2 by Bill Fenner, Oct 1995. * Significantly rewritten for IGMPv3, VIMAGE, and SMP by Bruce Simpson. * * MULTICAST Revision: 3.5.1.4 */ #include __FBSDID("$FreeBSD$"); +#include "opt_ddb.h" + #include #include #include #include #include #include #include #include #include #include #include #include #include +#ifdef DDB +#include +#endif + #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifndef KTR_IGMPV3 #define KTR_IGMPV3 KTR_INET #endif static struct igmp_ifsoftc * igi_alloc_locked(struct ifnet *); static void igi_delete_locked(const struct ifnet *); static void igmp_dispatch_queue(struct mbufq *, int, const int); static void igmp_fasttimo_vnet(void); static void igmp_final_leave(struct in_multi *, struct igmp_ifsoftc *); static int igmp_handle_state_change(struct in_multi *, struct igmp_ifsoftc *); static int igmp_initial_join(struct in_multi *, struct igmp_ifsoftc *); static int igmp_input_v1_query(struct ifnet *, const struct ip *, const struct igmp *); static int igmp_input_v2_query(struct ifnet *, const struct ip *, const struct igmp *); static int igmp_input_v3_query(struct ifnet *, const struct ip *, /*const*/ struct igmpv3 *); static int igmp_input_v3_group_query(struct in_multi *, struct igmp_ifsoftc *, int, /*const*/ struct igmpv3 *); static int igmp_input_v1_report(struct ifnet *, /*const*/ struct ip *, /*const*/ struct igmp *); static int igmp_input_v2_report(struct ifnet *, /*const*/ struct ip *, /*const*/ struct igmp *); static void igmp_intr(struct mbuf *); static int igmp_isgroupreported(const struct in_addr); static struct mbuf * igmp_ra_alloc(void); #ifdef KTR static char * igmp_rec_type_to_str(const int); #endif static void igmp_set_version(struct igmp_ifsoftc *, const int); static void igmp_slowtimo_vnet(void); static int igmp_v1v2_queue_report(struct in_multi *, const int); static void igmp_v1v2_process_group_timer(struct in_multi *, const int); static void igmp_v1v2_process_querier_timers(struct igmp_ifsoftc *); static void igmp_v2_update_group(struct in_multi *, const int); static void igmp_v3_cancel_link_timers(struct igmp_ifsoftc *); static void igmp_v3_dispatch_general_query(struct igmp_ifsoftc *); static struct mbuf * igmp_v3_encap_report(struct ifnet *, struct mbuf *); static int igmp_v3_enqueue_group_record(struct mbufq *, struct in_multi *, const int, const int, const int); static int igmp_v3_enqueue_filter_change(struct mbufq *, struct in_multi *); static void igmp_v3_process_group_timers(struct igmp_ifsoftc *, struct mbufq *, struct mbufq *, struct in_multi *, const int); static int igmp_v3_merge_state_changes(struct in_multi *, struct mbufq *); static void igmp_v3_suppress_group_record(struct in_multi *); static int sysctl_igmp_default_version(SYSCTL_HANDLER_ARGS); static int sysctl_igmp_gsr(SYSCTL_HANDLER_ARGS); static int sysctl_igmp_ifinfo(SYSCTL_HANDLER_ARGS); static const struct netisr_handler igmp_nh = { .nh_name = "igmp", .nh_handler = igmp_intr, .nh_proto = NETISR_IGMP, .nh_policy = NETISR_POLICY_SOURCE, }; /* * System-wide globals. * * Unlocked access to these is OK, except for the global IGMP output * queue. The IGMP subsystem lock ends up being system-wide for the moment, * because all VIMAGEs have to share a global output queue, as netisrs * themselves are not virtualized. * * Locking: * * The permitted lock order is: IN_MULTI_LOCK, IGMP_LOCK, IF_ADDR_LOCK. * Any may be taken independently; if any are held at the same * time, the above lock order must be followed. * * All output is delegated to the netisr. * Now that Giant has been eliminated, the netisr may be inlined. * * IN_MULTI_LOCK covers in_multi. * * IGMP_LOCK covers igmp_ifsoftc and any global variables in this file, * including the output queue. * * IF_ADDR_LOCK covers if_multiaddrs, which is used for a variety of * per-link state iterators. * * igmp_ifsoftc is valid as long as PF_INET is attached to the interface, * therefore it is not refcounted. * We allow unlocked reads of igmp_ifsoftc when accessed via in_multi. * * Reference counting * * IGMP acquires its own reference every time an in_multi is passed to * it and the group is being joined for the first time. * * IGMP releases its reference(s) on in_multi in a deferred way, * because the operations which process the release run as part of * a loop whose control variables are directly affected by the release * (that, and not recursing on the IF_ADDR_LOCK). * * VIMAGE: Each in_multi corresponds to an ifp, and each ifp corresponds * to a vnet in ifp->if_vnet. * * SMPng: XXX We may potentially race operations on ifma_protospec. * The problem is that we currently lack a clean way of taking the * IF_ADDR_LOCK() between the ifnet and in layers w/o recursing, * as anything which modifies ifma needs to be covered by that lock. * So check for ifma_protospec being NULL before proceeding. */ struct mtx igmp_mtx; struct mbuf *m_raopt; /* Router Alert option */ static MALLOC_DEFINE(M_IGMP, "igmp", "igmp state"); /* * VIMAGE-wide globals. * * The IGMPv3 timers themselves need to run per-image, however, * protosw timers run globally (see tcp). * An ifnet can only be in one vimage at a time, and the loopback * ifnet, loif, is itself virtualized. * It would otherwise be possible to seriously hose IGMP state, * and create inconsistencies in upstream multicast routing, if you have * multiple VIMAGEs running on the same link joining different multicast * groups, UNLESS the "primary IP address" is different. This is because * IGMP for IPv4 does not force link-local addresses to be used for each * node, unlike MLD for IPv6. * Obviously the IGMPv3 per-interface state has per-vimage granularity * also as a result. * * FUTURE: Stop using IFP_TO_IA/INADDR_ANY, and use source address selection * policy to control the address used by IGMP on the link. */ static VNET_DEFINE(int, interface_timers_running); /* IGMPv3 general * query response */ static VNET_DEFINE(int, state_change_timers_running); /* IGMPv3 state-change * retransmit */ static VNET_DEFINE(int, current_state_timers_running); /* IGMPv1/v2 host * report; IGMPv3 g/sg * query response */ #define V_interface_timers_running VNET(interface_timers_running) #define V_state_change_timers_running VNET(state_change_timers_running) #define V_current_state_timers_running VNET(current_state_timers_running) -static VNET_DEFINE(LIST_HEAD(, igmp_ifsoftc), igi_head); +static VNET_DEFINE(LIST_HEAD(, igmp_ifsoftc), igi_head) = + LIST_HEAD_INITIALIZER(igi_head); static VNET_DEFINE(struct igmpstat, igmpstat) = { .igps_version = IGPS_VERSION_3, .igps_len = sizeof(struct igmpstat), }; static VNET_DEFINE(struct timeval, igmp_gsrdelay) = {10, 0}; #define V_igi_head VNET(igi_head) #define V_igmpstat VNET(igmpstat) #define V_igmp_gsrdelay VNET(igmp_gsrdelay) static VNET_DEFINE(int, igmp_recvifkludge) = 1; static VNET_DEFINE(int, igmp_sendra) = 1; static VNET_DEFINE(int, igmp_sendlocal) = 1; static VNET_DEFINE(int, igmp_v1enable) = 1; static VNET_DEFINE(int, igmp_v2enable) = 1; static VNET_DEFINE(int, igmp_legacysupp); static VNET_DEFINE(int, igmp_default_version) = IGMP_VERSION_3; #define V_igmp_recvifkludge VNET(igmp_recvifkludge) #define V_igmp_sendra VNET(igmp_sendra) #define V_igmp_sendlocal VNET(igmp_sendlocal) #define V_igmp_v1enable VNET(igmp_v1enable) #define V_igmp_v2enable VNET(igmp_v2enable) #define V_igmp_legacysupp VNET(igmp_legacysupp) #define V_igmp_default_version VNET(igmp_default_version) /* * Virtualized sysctls. */ SYSCTL_STRUCT(_net_inet_igmp, IGMPCTL_STATS, stats, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(igmpstat), igmpstat, ""); SYSCTL_INT(_net_inet_igmp, OID_AUTO, recvifkludge, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(igmp_recvifkludge), 0, "Rewrite IGMPv1/v2 reports from 0.0.0.0 to contain subnet address"); SYSCTL_INT(_net_inet_igmp, OID_AUTO, sendra, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(igmp_sendra), 0, "Send IP Router Alert option in IGMPv2/v3 messages"); SYSCTL_INT(_net_inet_igmp, OID_AUTO, sendlocal, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(igmp_sendlocal), 0, "Send IGMP membership reports for 224.0.0.0/24 groups"); SYSCTL_INT(_net_inet_igmp, OID_AUTO, v1enable, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(igmp_v1enable), 0, "Enable backwards compatibility with IGMPv1"); SYSCTL_INT(_net_inet_igmp, OID_AUTO, v2enable, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(igmp_v2enable), 0, "Enable backwards compatibility with IGMPv2"); SYSCTL_INT(_net_inet_igmp, OID_AUTO, legacysupp, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(igmp_legacysupp), 0, "Allow v1/v2 reports to suppress v3 group responses"); SYSCTL_PROC(_net_inet_igmp, OID_AUTO, default_version, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, &VNET_NAME(igmp_default_version), 0, sysctl_igmp_default_version, "I", "Default version of IGMP to run on each interface"); SYSCTL_PROC(_net_inet_igmp, OID_AUTO, gsrdelay, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, &VNET_NAME(igmp_gsrdelay.tv_sec), 0, sysctl_igmp_gsr, "I", "Rate limit for IGMPv3 Group-and-Source queries in seconds"); /* * Non-virtualized sysctls. */ static SYSCTL_NODE(_net_inet_igmp, OID_AUTO, ifinfo, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_igmp_ifinfo, "Per-interface IGMPv3 state"); static __inline void igmp_save_context(struct mbuf *m, struct ifnet *ifp) { #ifdef VIMAGE m->m_pkthdr.PH_loc.ptr = ifp->if_vnet; #endif /* VIMAGE */ m->m_pkthdr.flowid = ifp->if_index; } static __inline void igmp_scrub_context(struct mbuf *m) { m->m_pkthdr.PH_loc.ptr = NULL; m->m_pkthdr.flowid = 0; } #ifdef KTR static __inline char * inet_ntoa_haddr(in_addr_t haddr) { struct in_addr ia; ia.s_addr = htonl(haddr); return (inet_ntoa(ia)); } #endif /* * Restore context from a queued IGMP output chain. * Return saved ifindex. * * VIMAGE: The assertion is there to make sure that we * actually called CURVNET_SET() with what's in the mbuf chain. */ static __inline uint32_t igmp_restore_context(struct mbuf *m) { #ifdef notyet #if defined(VIMAGE) && defined(INVARIANTS) KASSERT(curvnet == (m->m_pkthdr.PH_loc.ptr), ("%s: called when curvnet was not restored", __func__)); #endif #endif return (m->m_pkthdr.flowid); } /* * Retrieve or set default IGMP version. * * VIMAGE: Assume curvnet set by caller. * SMPng: NOTE: Serialized by IGMP lock. */ static int sysctl_igmp_default_version(SYSCTL_HANDLER_ARGS) { int error; int new; error = sysctl_wire_old_buffer(req, sizeof(int)); if (error) return (error); IGMP_LOCK(); new = V_igmp_default_version; error = sysctl_handle_int(oidp, &new, 0, req); if (error || !req->newptr) goto out_locked; if (new < IGMP_VERSION_1 || new > IGMP_VERSION_3) { error = EINVAL; goto out_locked; } CTR2(KTR_IGMPV3, "change igmp_default_version from %d to %d", V_igmp_default_version, new); V_igmp_default_version = new; out_locked: IGMP_UNLOCK(); return (error); } /* * Retrieve or set threshold between group-source queries in seconds. * * VIMAGE: Assume curvnet set by caller. * SMPng: NOTE: Serialized by IGMP lock. */ static int sysctl_igmp_gsr(SYSCTL_HANDLER_ARGS) { int error; int i; error = sysctl_wire_old_buffer(req, sizeof(int)); if (error) return (error); IGMP_LOCK(); i = V_igmp_gsrdelay.tv_sec; error = sysctl_handle_int(oidp, &i, 0, req); if (error || !req->newptr) goto out_locked; if (i < -1 || i >= 60) { error = EINVAL; goto out_locked; } CTR2(KTR_IGMPV3, "change igmp_gsrdelay from %d to %d", V_igmp_gsrdelay.tv_sec, i); V_igmp_gsrdelay.tv_sec = i; out_locked: IGMP_UNLOCK(); return (error); } /* * Expose struct igmp_ifsoftc to userland, keyed by ifindex. * For use by ifmcstat(8). * * SMPng: NOTE: Does an unlocked ifindex space read. * VIMAGE: Assume curvnet set by caller. The node handler itself * is not directly virtualized. */ static int sysctl_igmp_ifinfo(SYSCTL_HANDLER_ARGS) { int *name; int error; u_int namelen; struct ifnet *ifp; struct igmp_ifsoftc *igi; name = (int *)arg1; namelen = arg2; if (req->newptr != NULL) return (EPERM); if (namelen != 1) return (EINVAL); error = sysctl_wire_old_buffer(req, sizeof(struct igmp_ifinfo)); if (error) return (error); IN_MULTI_LOCK(); IGMP_LOCK(); if (name[0] <= 0 || name[0] > V_if_index) { error = ENOENT; goto out_locked; } error = ENOENT; ifp = ifnet_byindex(name[0]); if (ifp == NULL) goto out_locked; LIST_FOREACH(igi, &V_igi_head, igi_link) { if (ifp == igi->igi_ifp) { struct igmp_ifinfo info; info.igi_version = igi->igi_version; info.igi_v1_timer = igi->igi_v1_timer; info.igi_v2_timer = igi->igi_v2_timer; info.igi_v3_timer = igi->igi_v3_timer; info.igi_flags = igi->igi_flags; info.igi_rv = igi->igi_rv; info.igi_qi = igi->igi_qi; info.igi_qri = igi->igi_qri; info.igi_uri = igi->igi_uri; error = SYSCTL_OUT(req, &info, sizeof(info)); break; } } out_locked: IGMP_UNLOCK(); IN_MULTI_UNLOCK(); return (error); } /* * Dispatch an entire queue of pending packet chains * using the netisr. * VIMAGE: Assumes the vnet pointer has been set. */ static void igmp_dispatch_queue(struct mbufq *mq, int limit, const int loop) { struct mbuf *m; while ((m = mbufq_dequeue(mq)) != NULL) { CTR3(KTR_IGMPV3, "%s: dispatch %p from %p", __func__, mq, m); if (loop) m->m_flags |= M_IGMP_LOOP; netisr_dispatch(NETISR_IGMP, m); if (--limit == 0) break; } } /* * Filter outgoing IGMP report state by group. * * Reports are ALWAYS suppressed for ALL-HOSTS (224.0.0.1). * If the net.inet.igmp.sendlocal sysctl is 0, then IGMP reports are * disabled for all groups in the 224.0.0.0/24 link-local scope. However, * this may break certain IGMP snooping switches which rely on the old * report behaviour. * * Return zero if the given group is one for which IGMP reports * should be suppressed, or non-zero if reports should be issued. */ static __inline int igmp_isgroupreported(const struct in_addr addr) { if (in_allhosts(addr) || ((!V_igmp_sendlocal && IN_LOCAL_GROUP(ntohl(addr.s_addr))))) return (0); return (1); } /* * Construct a Router Alert option to use in outgoing packets. */ static struct mbuf * igmp_ra_alloc(void) { struct mbuf *m; struct ipoption *p; m = m_get(M_WAITOK, MT_DATA); p = mtod(m, struct ipoption *); p->ipopt_dst.s_addr = INADDR_ANY; p->ipopt_list[0] = IPOPT_RA; /* Router Alert Option */ p->ipopt_list[1] = 0x04; /* 4 bytes long */ p->ipopt_list[2] = IPOPT_EOL; /* End of IP option list */ p->ipopt_list[3] = 0x00; /* pad byte */ m->m_len = sizeof(p->ipopt_dst) + p->ipopt_list[1]; return (m); } /* * Attach IGMP when PF_INET is attached to an interface. */ struct igmp_ifsoftc * igmp_domifattach(struct ifnet *ifp) { struct igmp_ifsoftc *igi; CTR3(KTR_IGMPV3, "%s: called for ifp %p(%s)", __func__, ifp, ifp->if_xname); IGMP_LOCK(); igi = igi_alloc_locked(ifp); if (!(ifp->if_flags & IFF_MULTICAST)) igi->igi_flags |= IGIF_SILENT; IGMP_UNLOCK(); return (igi); } /* * VIMAGE: assume curvnet set by caller. */ static struct igmp_ifsoftc * igi_alloc_locked(/*const*/ struct ifnet *ifp) { struct igmp_ifsoftc *igi; IGMP_LOCK_ASSERT(); igi = malloc(sizeof(struct igmp_ifsoftc), M_IGMP, M_NOWAIT|M_ZERO); if (igi == NULL) goto out; igi->igi_ifp = ifp; igi->igi_version = V_igmp_default_version; igi->igi_flags = 0; igi->igi_rv = IGMP_RV_INIT; igi->igi_qi = IGMP_QI_INIT; igi->igi_qri = IGMP_QRI_INIT; igi->igi_uri = IGMP_URI_INIT; SLIST_INIT(&igi->igi_relinmhead); mbufq_init(&igi->igi_gq, IGMP_MAX_RESPONSE_PACKETS); LIST_INSERT_HEAD(&V_igi_head, igi, igi_link); CTR2(KTR_IGMPV3, "allocate igmp_ifsoftc for ifp %p(%s)", ifp, ifp->if_xname); out: return (igi); } /* * Hook for ifdetach. * * NOTE: Some finalization tasks need to run before the protocol domain * is detached, but also before the link layer does its cleanup. * * SMPNG: igmp_ifdetach() needs to take IF_ADDR_LOCK(). * XXX This is also bitten by unlocked ifma_protospec access. */ void igmp_ifdetach(struct ifnet *ifp) { struct igmp_ifsoftc *igi; struct ifmultiaddr *ifma; struct in_multi *inm, *tinm; CTR3(KTR_IGMPV3, "%s: called for ifp %p(%s)", __func__, ifp, ifp->if_xname); IGMP_LOCK(); igi = ((struct in_ifinfo *)ifp->if_afdata[AF_INET])->ii_igmp; if (igi->igi_version == IGMP_VERSION_3) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET || ifma->ifma_protospec == NULL) continue; #if 0 KASSERT(ifma->ifma_protospec != NULL, ("%s: ifma_protospec is NULL", __func__)); #endif inm = (struct in_multi *)ifma->ifma_protospec; if (inm->inm_state == IGMP_LEAVING_MEMBER) { SLIST_INSERT_HEAD(&igi->igi_relinmhead, inm, inm_nrele); } inm_clear_recorded(inm); } IF_ADDR_RUNLOCK(ifp); /* * Free the in_multi reference(s) for this IGMP lifecycle. */ SLIST_FOREACH_SAFE(inm, &igi->igi_relinmhead, inm_nrele, tinm) { SLIST_REMOVE_HEAD(&igi->igi_relinmhead, inm_nrele); inm_release_locked(inm); } } IGMP_UNLOCK(); } /* * Hook for domifdetach. */ void igmp_domifdetach(struct ifnet *ifp) { - struct igmp_ifsoftc *igi; CTR3(KTR_IGMPV3, "%s: called for ifp %p(%s)", __func__, ifp, ifp->if_xname); IGMP_LOCK(); - - igi = ((struct in_ifinfo *)ifp->if_afdata[AF_INET])->ii_igmp; igi_delete_locked(ifp); - IGMP_UNLOCK(); } static void igi_delete_locked(const struct ifnet *ifp) { struct igmp_ifsoftc *igi, *tigi; CTR3(KTR_IGMPV3, "%s: freeing igmp_ifsoftc for ifp %p(%s)", __func__, ifp, ifp->if_xname); IGMP_LOCK_ASSERT(); LIST_FOREACH_SAFE(igi, &V_igi_head, igi_link, tigi) { if (igi->igi_ifp == ifp) { /* * Free deferred General Query responses. */ mbufq_drain(&igi->igi_gq); LIST_REMOVE(igi, igi_link); KASSERT(SLIST_EMPTY(&igi->igi_relinmhead), ("%s: there are dangling in_multi references", __func__)); free(igi, M_IGMP); return; } } #ifdef INVARIANTS panic("%s: igmp_ifsoftc not found for ifp %p\n", __func__, ifp); #endif } /* * Process a received IGMPv1 query. * Return non-zero if the message should be dropped. * * VIMAGE: The curvnet pointer is derived from the input ifp. */ static int igmp_input_v1_query(struct ifnet *ifp, const struct ip *ip, const struct igmp *igmp) { struct ifmultiaddr *ifma; struct igmp_ifsoftc *igi; struct in_multi *inm; /* * IGMPv1 Host Mmembership Queries SHOULD always be addressed to * 224.0.0.1. They are always treated as General Queries. * igmp_group is always ignored. Do not drop it as a userland * daemon may wish to see it. * XXX SMPng: unlocked increments in igmpstat assumed atomic. */ if (!in_allhosts(ip->ip_dst) || !in_nullhost(igmp->igmp_group)) { IGMPSTAT_INC(igps_rcv_badqueries); return (0); } IGMPSTAT_INC(igps_rcv_gen_queries); IN_MULTI_LOCK(); IGMP_LOCK(); igi = ((struct in_ifinfo *)ifp->if_afdata[AF_INET])->ii_igmp; KASSERT(igi != NULL, ("%s: no igmp_ifsoftc for ifp %p", __func__, ifp)); if (igi->igi_flags & IGIF_LOOPBACK) { CTR2(KTR_IGMPV3, "ignore v1 query on IGIF_LOOPBACK ifp %p(%s)", ifp, ifp->if_xname); goto out_locked; } /* * Switch to IGMPv1 host compatibility mode. */ igmp_set_version(igi, IGMP_VERSION_1); CTR2(KTR_IGMPV3, "process v1 query on ifp %p(%s)", ifp, ifp->if_xname); /* * Start the timers in all of our group records * for the interface on which the query arrived, * except those which are already running. */ IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET || ifma->ifma_protospec == NULL) continue; inm = (struct in_multi *)ifma->ifma_protospec; if (inm->inm_timer != 0) continue; switch (inm->inm_state) { case IGMP_NOT_MEMBER: case IGMP_SILENT_MEMBER: break; case IGMP_G_QUERY_PENDING_MEMBER: case IGMP_SG_QUERY_PENDING_MEMBER: case IGMP_REPORTING_MEMBER: case IGMP_IDLE_MEMBER: case IGMP_LAZY_MEMBER: case IGMP_SLEEPING_MEMBER: case IGMP_AWAKENING_MEMBER: inm->inm_state = IGMP_REPORTING_MEMBER; inm->inm_timer = IGMP_RANDOM_DELAY( IGMP_V1V2_MAX_RI * PR_FASTHZ); V_current_state_timers_running = 1; break; case IGMP_LEAVING_MEMBER: break; } } IF_ADDR_RUNLOCK(ifp); out_locked: IGMP_UNLOCK(); IN_MULTI_UNLOCK(); return (0); } /* * Process a received IGMPv2 general or group-specific query. */ static int igmp_input_v2_query(struct ifnet *ifp, const struct ip *ip, const struct igmp *igmp) { struct ifmultiaddr *ifma; struct igmp_ifsoftc *igi; struct in_multi *inm; int is_general_query; uint16_t timer; is_general_query = 0; /* * Validate address fields upfront. * XXX SMPng: unlocked increments in igmpstat assumed atomic. */ if (in_nullhost(igmp->igmp_group)) { /* * IGMPv2 General Query. * If this was not sent to the all-hosts group, ignore it. */ if (!in_allhosts(ip->ip_dst)) return (0); IGMPSTAT_INC(igps_rcv_gen_queries); is_general_query = 1; } else { /* IGMPv2 Group-Specific Query. */ IGMPSTAT_INC(igps_rcv_group_queries); } IN_MULTI_LOCK(); IGMP_LOCK(); igi = ((struct in_ifinfo *)ifp->if_afdata[AF_INET])->ii_igmp; KASSERT(igi != NULL, ("%s: no igmp_ifsoftc for ifp %p", __func__, ifp)); if (igi->igi_flags & IGIF_LOOPBACK) { CTR2(KTR_IGMPV3, "ignore v2 query on IGIF_LOOPBACK ifp %p(%s)", ifp, ifp->if_xname); goto out_locked; } /* * Ignore v2 query if in v1 Compatibility Mode. */ if (igi->igi_version == IGMP_VERSION_1) goto out_locked; igmp_set_version(igi, IGMP_VERSION_2); timer = igmp->igmp_code * PR_FASTHZ / IGMP_TIMER_SCALE; if (timer == 0) timer = 1; if (is_general_query) { /* * For each reporting group joined on this * interface, kick the report timer. */ CTR2(KTR_IGMPV3, "process v2 general query on ifp %p(%s)", ifp, ifp->if_xname); IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET || ifma->ifma_protospec == NULL) continue; inm = (struct in_multi *)ifma->ifma_protospec; igmp_v2_update_group(inm, timer); } IF_ADDR_RUNLOCK(ifp); } else { /* * Group-specific IGMPv2 query, we need only * look up the single group to process it. */ inm = inm_lookup(ifp, igmp->igmp_group); if (inm != NULL) { CTR3(KTR_IGMPV3, "process v2 query %s on ifp %p(%s)", inet_ntoa(igmp->igmp_group), ifp, ifp->if_xname); igmp_v2_update_group(inm, timer); } } out_locked: IGMP_UNLOCK(); IN_MULTI_UNLOCK(); return (0); } /* * Update the report timer on a group in response to an IGMPv2 query. * * If we are becoming the reporting member for this group, start the timer. * If we already are the reporting member for this group, and timer is * below the threshold, reset it. * * We may be updating the group for the first time since we switched * to IGMPv3. If we are, then we must clear any recorded source lists, * and transition to REPORTING state; the group timer is overloaded * for group and group-source query responses. * * Unlike IGMPv3, the delay per group should be jittered * to avoid bursts of IGMPv2 reports. */ static void igmp_v2_update_group(struct in_multi *inm, const int timer) { CTR4(KTR_IGMPV3, "%s: %s/%s timer=%d", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp->if_xname, timer); IN_MULTI_LOCK_ASSERT(); switch (inm->inm_state) { case IGMP_NOT_MEMBER: case IGMP_SILENT_MEMBER: break; case IGMP_REPORTING_MEMBER: if (inm->inm_timer != 0 && inm->inm_timer <= timer) { CTR1(KTR_IGMPV3, "%s: REPORTING and timer running, " "skipping.", __func__); break; } /* FALLTHROUGH */ case IGMP_SG_QUERY_PENDING_MEMBER: case IGMP_G_QUERY_PENDING_MEMBER: case IGMP_IDLE_MEMBER: case IGMP_LAZY_MEMBER: case IGMP_AWAKENING_MEMBER: CTR1(KTR_IGMPV3, "%s: ->REPORTING", __func__); inm->inm_state = IGMP_REPORTING_MEMBER; inm->inm_timer = IGMP_RANDOM_DELAY(timer); V_current_state_timers_running = 1; break; case IGMP_SLEEPING_MEMBER: CTR1(KTR_IGMPV3, "%s: ->AWAKENING", __func__); inm->inm_state = IGMP_AWAKENING_MEMBER; break; case IGMP_LEAVING_MEMBER: break; } } /* * Process a received IGMPv3 general, group-specific or * group-and-source-specific query. * Assumes m has already been pulled up to the full IGMP message length. * Return 0 if successful, otherwise an appropriate error code is returned. */ static int igmp_input_v3_query(struct ifnet *ifp, const struct ip *ip, /*const*/ struct igmpv3 *igmpv3) { struct igmp_ifsoftc *igi; struct in_multi *inm; int is_general_query; uint32_t maxresp, nsrc, qqi; uint16_t timer; uint8_t qrv; is_general_query = 0; CTR2(KTR_IGMPV3, "process v3 query on ifp %p(%s)", ifp, ifp->if_xname); maxresp = igmpv3->igmp_code; /* in 1/10ths of a second */ if (maxresp >= 128) { maxresp = IGMP_MANT(igmpv3->igmp_code) << (IGMP_EXP(igmpv3->igmp_code) + 3); } /* * Robustness must never be less than 2 for on-wire IGMPv3. * FUTURE: Check if ifp has IGIF_LOOPBACK set, as we will make * an exception for interfaces whose IGMPv3 state changes * are redirected to loopback (e.g. MANET). */ qrv = IGMP_QRV(igmpv3->igmp_misc); if (qrv < 2) { CTR3(KTR_IGMPV3, "%s: clamping qrv %d to %d", __func__, qrv, IGMP_RV_INIT); qrv = IGMP_RV_INIT; } qqi = igmpv3->igmp_qqi; if (qqi >= 128) { qqi = IGMP_MANT(igmpv3->igmp_qqi) << (IGMP_EXP(igmpv3->igmp_qqi) + 3); } timer = maxresp * PR_FASTHZ / IGMP_TIMER_SCALE; if (timer == 0) timer = 1; nsrc = ntohs(igmpv3->igmp_numsrc); /* * Validate address fields and versions upfront before * accepting v3 query. * XXX SMPng: Unlocked access to igmpstat counters here. */ if (in_nullhost(igmpv3->igmp_group)) { /* * IGMPv3 General Query. * * General Queries SHOULD be directed to 224.0.0.1. * A general query with a source list has undefined * behaviour; discard it. */ IGMPSTAT_INC(igps_rcv_gen_queries); if (!in_allhosts(ip->ip_dst) || nsrc > 0) { IGMPSTAT_INC(igps_rcv_badqueries); return (0); } is_general_query = 1; } else { /* Group or group-source specific query. */ if (nsrc == 0) IGMPSTAT_INC(igps_rcv_group_queries); else IGMPSTAT_INC(igps_rcv_gsr_queries); } IN_MULTI_LOCK(); IGMP_LOCK(); igi = ((struct in_ifinfo *)ifp->if_afdata[AF_INET])->ii_igmp; KASSERT(igi != NULL, ("%s: no igmp_ifsoftc for ifp %p", __func__, ifp)); if (igi->igi_flags & IGIF_LOOPBACK) { CTR2(KTR_IGMPV3, "ignore v3 query on IGIF_LOOPBACK ifp %p(%s)", ifp, ifp->if_xname); goto out_locked; } /* * Discard the v3 query if we're in Compatibility Mode. * The RFC is not obviously worded that hosts need to stay in * compatibility mode until the Old Version Querier Present * timer expires. */ if (igi->igi_version != IGMP_VERSION_3) { CTR3(KTR_IGMPV3, "ignore v3 query in v%d mode on ifp %p(%s)", igi->igi_version, ifp, ifp->if_xname); goto out_locked; } igmp_set_version(igi, IGMP_VERSION_3); igi->igi_rv = qrv; igi->igi_qi = qqi; igi->igi_qri = maxresp; CTR4(KTR_IGMPV3, "%s: qrv %d qi %d qri %d", __func__, qrv, qqi, maxresp); if (is_general_query) { /* * Schedule a current-state report on this ifp for * all groups, possibly containing source lists. * If there is a pending General Query response * scheduled earlier than the selected delay, do * not schedule any other reports. * Otherwise, reset the interface timer. */ CTR2(KTR_IGMPV3, "process v3 general query on ifp %p(%s)", ifp, ifp->if_xname); if (igi->igi_v3_timer == 0 || igi->igi_v3_timer >= timer) { igi->igi_v3_timer = IGMP_RANDOM_DELAY(timer); V_interface_timers_running = 1; } } else { /* * Group-source-specific queries are throttled on * a per-group basis to defeat denial-of-service attempts. * Queries for groups we are not a member of on this * link are simply ignored. */ inm = inm_lookup(ifp, igmpv3->igmp_group); if (inm == NULL) goto out_locked; if (nsrc > 0) { if (!ratecheck(&inm->inm_lastgsrtv, &V_igmp_gsrdelay)) { CTR1(KTR_IGMPV3, "%s: GS query throttled.", __func__); IGMPSTAT_INC(igps_drop_gsr_queries); goto out_locked; } } CTR3(KTR_IGMPV3, "process v3 %s query on ifp %p(%s)", inet_ntoa(igmpv3->igmp_group), ifp, ifp->if_xname); /* * If there is a pending General Query response * scheduled sooner than the selected delay, no * further report need be scheduled. * Otherwise, prepare to respond to the * group-specific or group-and-source query. */ if (igi->igi_v3_timer == 0 || igi->igi_v3_timer >= timer) igmp_input_v3_group_query(inm, igi, timer, igmpv3); } out_locked: IGMP_UNLOCK(); IN_MULTI_UNLOCK(); return (0); } /* * Process a recieved IGMPv3 group-specific or group-and-source-specific * query. * Return <0 if any error occured. Currently this is ignored. */ static int igmp_input_v3_group_query(struct in_multi *inm, struct igmp_ifsoftc *igi, int timer, /*const*/ struct igmpv3 *igmpv3) { int retval; uint16_t nsrc; IN_MULTI_LOCK_ASSERT(); IGMP_LOCK_ASSERT(); retval = 0; switch (inm->inm_state) { case IGMP_NOT_MEMBER: case IGMP_SILENT_MEMBER: case IGMP_SLEEPING_MEMBER: case IGMP_LAZY_MEMBER: case IGMP_AWAKENING_MEMBER: case IGMP_IDLE_MEMBER: case IGMP_LEAVING_MEMBER: return (retval); break; case IGMP_REPORTING_MEMBER: case IGMP_G_QUERY_PENDING_MEMBER: case IGMP_SG_QUERY_PENDING_MEMBER: break; } nsrc = ntohs(igmpv3->igmp_numsrc); /* * Deal with group-specific queries upfront. * If any group query is already pending, purge any recorded * source-list state if it exists, and schedule a query response * for this group-specific query. */ if (nsrc == 0) { if (inm->inm_state == IGMP_G_QUERY_PENDING_MEMBER || inm->inm_state == IGMP_SG_QUERY_PENDING_MEMBER) { inm_clear_recorded(inm); timer = min(inm->inm_timer, timer); } inm->inm_state = IGMP_G_QUERY_PENDING_MEMBER; inm->inm_timer = IGMP_RANDOM_DELAY(timer); V_current_state_timers_running = 1; return (retval); } /* * Deal with the case where a group-and-source-specific query has * been received but a group-specific query is already pending. */ if (inm->inm_state == IGMP_G_QUERY_PENDING_MEMBER) { timer = min(inm->inm_timer, timer); inm->inm_timer = IGMP_RANDOM_DELAY(timer); V_current_state_timers_running = 1; return (retval); } /* * Finally, deal with the case where a group-and-source-specific * query has been received, where a response to a previous g-s-r * query exists, or none exists. * In this case, we need to parse the source-list which the Querier * has provided us with and check if we have any source list filter * entries at T1 for these sources. If we do not, there is no need * schedule a report and the query may be dropped. * If we do, we must record them and schedule a current-state * report for those sources. * FIXME: Handling source lists larger than 1 mbuf requires that * we pass the mbuf chain pointer down to this function, and use * m_getptr() to walk the chain. */ if (inm->inm_nsrc > 0) { const struct in_addr *ap; int i, nrecorded; ap = (const struct in_addr *)(igmpv3 + 1); nrecorded = 0; for (i = 0; i < nsrc; i++, ap++) { retval = inm_record_source(inm, ap->s_addr); if (retval < 0) break; nrecorded += retval; } if (nrecorded > 0) { CTR1(KTR_IGMPV3, "%s: schedule response to SG query", __func__); inm->inm_state = IGMP_SG_QUERY_PENDING_MEMBER; inm->inm_timer = IGMP_RANDOM_DELAY(timer); V_current_state_timers_running = 1; } } return (retval); } /* * Process a received IGMPv1 host membership report. * * NOTE: 0.0.0.0 workaround breaks const correctness. */ static int igmp_input_v1_report(struct ifnet *ifp, /*const*/ struct ip *ip, /*const*/ struct igmp *igmp) { struct rm_priotracker in_ifa_tracker; struct in_ifaddr *ia; struct in_multi *inm; IGMPSTAT_INC(igps_rcv_reports); if (ifp->if_flags & IFF_LOOPBACK) return (0); if (!IN_MULTICAST(ntohl(igmp->igmp_group.s_addr)) || !in_hosteq(igmp->igmp_group, ip->ip_dst)) { IGMPSTAT_INC(igps_rcv_badreports); return (EINVAL); } /* * RFC 3376, Section 4.2.13, 9.2, 9.3: * Booting clients may use the source address 0.0.0.0. Some * IGMP daemons may not know how to use IP_RECVIF to determine * the interface upon which this message was received. * Replace 0.0.0.0 with the subnet address if told to do so. */ if (V_igmp_recvifkludge && in_nullhost(ip->ip_src)) { IFP_TO_IA(ifp, ia, &in_ifa_tracker); if (ia != NULL) { ip->ip_src.s_addr = htonl(ia->ia_subnet); ifa_free(&ia->ia_ifa); } } CTR3(KTR_IGMPV3, "process v1 report %s on ifp %p(%s)", inet_ntoa(igmp->igmp_group), ifp, ifp->if_xname); /* * IGMPv1 report suppression. * If we are a member of this group, and our membership should be * reported, stop our group timer and transition to the 'lazy' state. */ IN_MULTI_LOCK(); inm = inm_lookup(ifp, igmp->igmp_group); if (inm != NULL) { struct igmp_ifsoftc *igi; igi = inm->inm_igi; if (igi == NULL) { KASSERT(igi != NULL, ("%s: no igi for ifp %p", __func__, ifp)); goto out_locked; } IGMPSTAT_INC(igps_rcv_ourreports); /* * If we are in IGMPv3 host mode, do not allow the * other host's IGMPv1 report to suppress our reports * unless explicitly configured to do so. */ if (igi->igi_version == IGMP_VERSION_3) { if (V_igmp_legacysupp) igmp_v3_suppress_group_record(inm); goto out_locked; } inm->inm_timer = 0; switch (inm->inm_state) { case IGMP_NOT_MEMBER: case IGMP_SILENT_MEMBER: break; case IGMP_IDLE_MEMBER: case IGMP_LAZY_MEMBER: case IGMP_AWAKENING_MEMBER: CTR3(KTR_IGMPV3, "report suppressed for %s on ifp %p(%s)", inet_ntoa(igmp->igmp_group), ifp, ifp->if_xname); case IGMP_SLEEPING_MEMBER: inm->inm_state = IGMP_SLEEPING_MEMBER; break; case IGMP_REPORTING_MEMBER: CTR3(KTR_IGMPV3, "report suppressed for %s on ifp %p(%s)", inet_ntoa(igmp->igmp_group), ifp, ifp->if_xname); if (igi->igi_version == IGMP_VERSION_1) inm->inm_state = IGMP_LAZY_MEMBER; else if (igi->igi_version == IGMP_VERSION_2) inm->inm_state = IGMP_SLEEPING_MEMBER; break; case IGMP_G_QUERY_PENDING_MEMBER: case IGMP_SG_QUERY_PENDING_MEMBER: case IGMP_LEAVING_MEMBER: break; } } out_locked: IN_MULTI_UNLOCK(); return (0); } /* * Process a received IGMPv2 host membership report. * * NOTE: 0.0.0.0 workaround breaks const correctness. */ static int igmp_input_v2_report(struct ifnet *ifp, /*const*/ struct ip *ip, /*const*/ struct igmp *igmp) { struct rm_priotracker in_ifa_tracker; struct in_ifaddr *ia; struct in_multi *inm; /* * Make sure we don't hear our own membership report. Fast * leave requires knowing that we are the only member of a * group. */ IFP_TO_IA(ifp, ia, &in_ifa_tracker); if (ia != NULL && in_hosteq(ip->ip_src, IA_SIN(ia)->sin_addr)) { ifa_free(&ia->ia_ifa); return (0); } IGMPSTAT_INC(igps_rcv_reports); if (ifp->if_flags & IFF_LOOPBACK) { if (ia != NULL) ifa_free(&ia->ia_ifa); return (0); } if (!IN_MULTICAST(ntohl(igmp->igmp_group.s_addr)) || !in_hosteq(igmp->igmp_group, ip->ip_dst)) { if (ia != NULL) ifa_free(&ia->ia_ifa); IGMPSTAT_INC(igps_rcv_badreports); return (EINVAL); } /* * RFC 3376, Section 4.2.13, 9.2, 9.3: * Booting clients may use the source address 0.0.0.0. Some * IGMP daemons may not know how to use IP_RECVIF to determine * the interface upon which this message was received. * Replace 0.0.0.0 with the subnet address if told to do so. */ if (V_igmp_recvifkludge && in_nullhost(ip->ip_src)) { if (ia != NULL) ip->ip_src.s_addr = htonl(ia->ia_subnet); } if (ia != NULL) ifa_free(&ia->ia_ifa); CTR3(KTR_IGMPV3, "process v2 report %s on ifp %p(%s)", inet_ntoa(igmp->igmp_group), ifp, ifp->if_xname); /* * IGMPv2 report suppression. * If we are a member of this group, and our membership should be * reported, and our group timer is pending or about to be reset, * stop our group timer by transitioning to the 'lazy' state. */ IN_MULTI_LOCK(); inm = inm_lookup(ifp, igmp->igmp_group); if (inm != NULL) { struct igmp_ifsoftc *igi; igi = inm->inm_igi; KASSERT(igi != NULL, ("%s: no igi for ifp %p", __func__, ifp)); IGMPSTAT_INC(igps_rcv_ourreports); /* * If we are in IGMPv3 host mode, do not allow the * other host's IGMPv1 report to suppress our reports * unless explicitly configured to do so. */ if (igi->igi_version == IGMP_VERSION_3) { if (V_igmp_legacysupp) igmp_v3_suppress_group_record(inm); goto out_locked; } inm->inm_timer = 0; switch (inm->inm_state) { case IGMP_NOT_MEMBER: case IGMP_SILENT_MEMBER: case IGMP_SLEEPING_MEMBER: break; case IGMP_REPORTING_MEMBER: case IGMP_IDLE_MEMBER: case IGMP_AWAKENING_MEMBER: CTR3(KTR_IGMPV3, "report suppressed for %s on ifp %p(%s)", inet_ntoa(igmp->igmp_group), ifp, ifp->if_xname); case IGMP_LAZY_MEMBER: inm->inm_state = IGMP_LAZY_MEMBER; break; case IGMP_G_QUERY_PENDING_MEMBER: case IGMP_SG_QUERY_PENDING_MEMBER: case IGMP_LEAVING_MEMBER: break; } } out_locked: IN_MULTI_UNLOCK(); return (0); } int igmp_input(struct mbuf **mp, int *offp, int proto) { int iphlen; struct ifnet *ifp; struct igmp *igmp; struct ip *ip; struct mbuf *m; int igmplen; int minlen; int queryver; CTR3(KTR_IGMPV3, "%s: called w/mbuf (%p,%d)", __func__, *mp, *offp); m = *mp; ifp = m->m_pkthdr.rcvif; *mp = NULL; IGMPSTAT_INC(igps_rcv_total); ip = mtod(m, struct ip *); iphlen = *offp; igmplen = ntohs(ip->ip_len) - iphlen; /* * Validate lengths. */ if (igmplen < IGMP_MINLEN) { IGMPSTAT_INC(igps_rcv_tooshort); m_freem(m); return (IPPROTO_DONE); } /* * Always pullup to the minimum size for v1/v2 or v3 * to amortize calls to m_pullup(). */ minlen = iphlen; if (igmplen >= IGMP_V3_QUERY_MINLEN) minlen += IGMP_V3_QUERY_MINLEN; else minlen += IGMP_MINLEN; if ((!M_WRITABLE(m) || m->m_len < minlen) && (m = m_pullup(m, minlen)) == 0) { IGMPSTAT_INC(igps_rcv_tooshort); return (IPPROTO_DONE); } ip = mtod(m, struct ip *); /* * Validate checksum. */ m->m_data += iphlen; m->m_len -= iphlen; igmp = mtod(m, struct igmp *); if (in_cksum(m, igmplen)) { IGMPSTAT_INC(igps_rcv_badsum); m_freem(m); return (IPPROTO_DONE); } m->m_data -= iphlen; m->m_len += iphlen; /* * IGMP control traffic is link-scope, and must have a TTL of 1. * DVMRP traffic (e.g. mrinfo, mtrace) is an exception; * probe packets may come from beyond the LAN. */ if (igmp->igmp_type != IGMP_DVMRP && ip->ip_ttl != 1) { IGMPSTAT_INC(igps_rcv_badttl); m_freem(m); return (IPPROTO_DONE); } switch (igmp->igmp_type) { case IGMP_HOST_MEMBERSHIP_QUERY: if (igmplen == IGMP_MINLEN) { if (igmp->igmp_code == 0) queryver = IGMP_VERSION_1; else queryver = IGMP_VERSION_2; } else if (igmplen >= IGMP_V3_QUERY_MINLEN) { queryver = IGMP_VERSION_3; } else { IGMPSTAT_INC(igps_rcv_tooshort); m_freem(m); return (IPPROTO_DONE); } switch (queryver) { case IGMP_VERSION_1: IGMPSTAT_INC(igps_rcv_v1v2_queries); if (!V_igmp_v1enable) break; if (igmp_input_v1_query(ifp, ip, igmp) != 0) { m_freem(m); return (IPPROTO_DONE); } break; case IGMP_VERSION_2: IGMPSTAT_INC(igps_rcv_v1v2_queries); if (!V_igmp_v2enable) break; if (igmp_input_v2_query(ifp, ip, igmp) != 0) { m_freem(m); return (IPPROTO_DONE); } break; case IGMP_VERSION_3: { struct igmpv3 *igmpv3; uint16_t igmpv3len; uint16_t nsrc; IGMPSTAT_INC(igps_rcv_v3_queries); igmpv3 = (struct igmpv3 *)igmp; /* * Validate length based on source count. */ nsrc = ntohs(igmpv3->igmp_numsrc); if (nsrc * sizeof(in_addr_t) > UINT16_MAX - iphlen - IGMP_V3_QUERY_MINLEN) { IGMPSTAT_INC(igps_rcv_tooshort); return (IPPROTO_DONE); } /* * m_pullup() may modify m, so pullup in * this scope. */ igmpv3len = iphlen + IGMP_V3_QUERY_MINLEN + sizeof(struct in_addr) * nsrc; if ((!M_WRITABLE(m) || m->m_len < igmpv3len) && (m = m_pullup(m, igmpv3len)) == NULL) { IGMPSTAT_INC(igps_rcv_tooshort); return (IPPROTO_DONE); } igmpv3 = (struct igmpv3 *)(mtod(m, uint8_t *) + iphlen); if (igmp_input_v3_query(ifp, ip, igmpv3) != 0) { m_freem(m); return (IPPROTO_DONE); } } break; } break; case IGMP_v1_HOST_MEMBERSHIP_REPORT: if (!V_igmp_v1enable) break; if (igmp_input_v1_report(ifp, ip, igmp) != 0) { m_freem(m); return (IPPROTO_DONE); } break; case IGMP_v2_HOST_MEMBERSHIP_REPORT: if (!V_igmp_v2enable) break; if (!ip_checkrouteralert(m)) IGMPSTAT_INC(igps_rcv_nora); if (igmp_input_v2_report(ifp, ip, igmp) != 0) { m_freem(m); return (IPPROTO_DONE); } break; case IGMP_v3_HOST_MEMBERSHIP_REPORT: /* * Hosts do not need to process IGMPv3 membership reports, * as report suppression is no longer required. */ if (!ip_checkrouteralert(m)) IGMPSTAT_INC(igps_rcv_nora); break; default: break; } /* * Pass all valid IGMP packets up to any process(es) listening on a * raw IGMP socket. */ *mp = m; return (rip_input(mp, offp, proto)); } /* * Fast timeout handler (global). * VIMAGE: Timeout handlers are expected to service all vimages. */ void igmp_fasttimo(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); igmp_fasttimo_vnet(); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * Fast timeout handler (per-vnet). * Sends are shuffled off to a netisr to deal with Giant. * * VIMAGE: Assume caller has set up our curvnet. */ static void igmp_fasttimo_vnet(void) { struct mbufq scq; /* State-change packets */ struct mbufq qrq; /* Query response packets */ struct ifnet *ifp; struct igmp_ifsoftc *igi; struct ifmultiaddr *ifma; struct in_multi *inm; int loop, uri_fasthz; loop = 0; uri_fasthz = 0; /* * Quick check to see if any work needs to be done, in order to * minimize the overhead of fasttimo processing. * SMPng: XXX Unlocked reads. */ if (!V_current_state_timers_running && !V_interface_timers_running && !V_state_change_timers_running) return; IN_MULTI_LOCK(); IGMP_LOCK(); /* * IGMPv3 General Query response timer processing. */ if (V_interface_timers_running) { CTR1(KTR_IGMPV3, "%s: interface timers running", __func__); V_interface_timers_running = 0; LIST_FOREACH(igi, &V_igi_head, igi_link) { if (igi->igi_v3_timer == 0) { /* Do nothing. */ } else if (--igi->igi_v3_timer == 0) { igmp_v3_dispatch_general_query(igi); } else { V_interface_timers_running = 1; } } } if (!V_current_state_timers_running && !V_state_change_timers_running) goto out_locked; V_current_state_timers_running = 0; V_state_change_timers_running = 0; CTR1(KTR_IGMPV3, "%s: state change timers running", __func__); /* * IGMPv1/v2/v3 host report and state-change timer processing. * Note: Processing a v3 group timer may remove a node. */ LIST_FOREACH(igi, &V_igi_head, igi_link) { ifp = igi->igi_ifp; if (igi->igi_version == IGMP_VERSION_3) { loop = (igi->igi_flags & IGIF_LOOPBACK) ? 1 : 0; uri_fasthz = IGMP_RANDOM_DELAY(igi->igi_uri * PR_FASTHZ); mbufq_init(&qrq, IGMP_MAX_G_GS_PACKETS); mbufq_init(&scq, IGMP_MAX_STATE_CHANGE_PACKETS); } IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET || ifma->ifma_protospec == NULL) continue; inm = (struct in_multi *)ifma->ifma_protospec; switch (igi->igi_version) { case IGMP_VERSION_1: case IGMP_VERSION_2: igmp_v1v2_process_group_timer(inm, igi->igi_version); break; case IGMP_VERSION_3: igmp_v3_process_group_timers(igi, &qrq, &scq, inm, uri_fasthz); break; } } IF_ADDR_RUNLOCK(ifp); if (igi->igi_version == IGMP_VERSION_3) { struct in_multi *tinm; igmp_dispatch_queue(&qrq, 0, loop); igmp_dispatch_queue(&scq, 0, loop); /* * Free the in_multi reference(s) for this * IGMP lifecycle. */ SLIST_FOREACH_SAFE(inm, &igi->igi_relinmhead, inm_nrele, tinm) { SLIST_REMOVE_HEAD(&igi->igi_relinmhead, inm_nrele); inm_release_locked(inm); } } } out_locked: IGMP_UNLOCK(); IN_MULTI_UNLOCK(); } /* * Update host report group timer for IGMPv1/v2. * Will update the global pending timer flags. */ static void igmp_v1v2_process_group_timer(struct in_multi *inm, const int version) { int report_timer_expired; IN_MULTI_LOCK_ASSERT(); IGMP_LOCK_ASSERT(); if (inm->inm_timer == 0) { report_timer_expired = 0; } else if (--inm->inm_timer == 0) { report_timer_expired = 1; } else { V_current_state_timers_running = 1; return; } switch (inm->inm_state) { case IGMP_NOT_MEMBER: case IGMP_SILENT_MEMBER: case IGMP_IDLE_MEMBER: case IGMP_LAZY_MEMBER: case IGMP_SLEEPING_MEMBER: case IGMP_AWAKENING_MEMBER: break; case IGMP_REPORTING_MEMBER: if (report_timer_expired) { inm->inm_state = IGMP_IDLE_MEMBER; (void)igmp_v1v2_queue_report(inm, (version == IGMP_VERSION_2) ? IGMP_v2_HOST_MEMBERSHIP_REPORT : IGMP_v1_HOST_MEMBERSHIP_REPORT); } break; case IGMP_G_QUERY_PENDING_MEMBER: case IGMP_SG_QUERY_PENDING_MEMBER: case IGMP_LEAVING_MEMBER: break; } } /* * Update a group's timers for IGMPv3. * Will update the global pending timer flags. * Note: Unlocked read from igi. */ static void igmp_v3_process_group_timers(struct igmp_ifsoftc *igi, struct mbufq *qrq, struct mbufq *scq, struct in_multi *inm, const int uri_fasthz) { int query_response_timer_expired; int state_change_retransmit_timer_expired; IN_MULTI_LOCK_ASSERT(); IGMP_LOCK_ASSERT(); query_response_timer_expired = 0; state_change_retransmit_timer_expired = 0; /* * During a transition from v1/v2 compatibility mode back to v3, * a group record in REPORTING state may still have its group * timer active. This is a no-op in this function; it is easier * to deal with it here than to complicate the slow-timeout path. */ if (inm->inm_timer == 0) { query_response_timer_expired = 0; } else if (--inm->inm_timer == 0) { query_response_timer_expired = 1; } else { V_current_state_timers_running = 1; } if (inm->inm_sctimer == 0) { state_change_retransmit_timer_expired = 0; } else if (--inm->inm_sctimer == 0) { state_change_retransmit_timer_expired = 1; } else { V_state_change_timers_running = 1; } /* We are in fasttimo, so be quick about it. */ if (!state_change_retransmit_timer_expired && !query_response_timer_expired) return; switch (inm->inm_state) { case IGMP_NOT_MEMBER: case IGMP_SILENT_MEMBER: case IGMP_SLEEPING_MEMBER: case IGMP_LAZY_MEMBER: case IGMP_AWAKENING_MEMBER: case IGMP_IDLE_MEMBER: break; case IGMP_G_QUERY_PENDING_MEMBER: case IGMP_SG_QUERY_PENDING_MEMBER: /* * Respond to a previously pending Group-Specific * or Group-and-Source-Specific query by enqueueing * the appropriate Current-State report for * immediate transmission. */ if (query_response_timer_expired) { int retval; retval = igmp_v3_enqueue_group_record(qrq, inm, 0, 1, (inm->inm_state == IGMP_SG_QUERY_PENDING_MEMBER)); CTR2(KTR_IGMPV3, "%s: enqueue record = %d", __func__, retval); inm->inm_state = IGMP_REPORTING_MEMBER; /* XXX Clear recorded sources for next time. */ inm_clear_recorded(inm); } /* FALLTHROUGH */ case IGMP_REPORTING_MEMBER: case IGMP_LEAVING_MEMBER: if (state_change_retransmit_timer_expired) { /* * State-change retransmission timer fired. * If there are any further pending retransmissions, * set the global pending state-change flag, and * reset the timer. */ if (--inm->inm_scrv > 0) { inm->inm_sctimer = uri_fasthz; V_state_change_timers_running = 1; } /* * Retransmit the previously computed state-change * report. If there are no further pending * retransmissions, the mbuf queue will be consumed. * Update T0 state to T1 as we have now sent * a state-change. */ (void)igmp_v3_merge_state_changes(inm, scq); inm_commit(inm); CTR3(KTR_IGMPV3, "%s: T1 -> T0 for %s/%s", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp->if_xname); /* * If we are leaving the group for good, make sure * we release IGMP's reference to it. * This release must be deferred using a SLIST, * as we are called from a loop which traverses * the in_ifmultiaddr TAILQ. */ if (inm->inm_state == IGMP_LEAVING_MEMBER && inm->inm_scrv == 0) { inm->inm_state = IGMP_NOT_MEMBER; SLIST_INSERT_HEAD(&igi->igi_relinmhead, inm, inm_nrele); } } break; } } /* * Suppress a group's pending response to a group or source/group query. * * Do NOT suppress state changes. This leads to IGMPv3 inconsistency. * Do NOT update ST1/ST0 as this operation merely suppresses * the currently pending group record. * Do NOT suppress the response to a general query. It is possible but * it would require adding another state or flag. */ static void igmp_v3_suppress_group_record(struct in_multi *inm) { IN_MULTI_LOCK_ASSERT(); KASSERT(inm->inm_igi->igi_version == IGMP_VERSION_3, ("%s: not IGMPv3 mode on link", __func__)); if (inm->inm_state != IGMP_G_QUERY_PENDING_MEMBER || inm->inm_state != IGMP_SG_QUERY_PENDING_MEMBER) return; if (inm->inm_state == IGMP_SG_QUERY_PENDING_MEMBER) inm_clear_recorded(inm); inm->inm_timer = 0; inm->inm_state = IGMP_REPORTING_MEMBER; } /* * Switch to a different IGMP version on the given interface, * as per Section 7.2.1. */ static void igmp_set_version(struct igmp_ifsoftc *igi, const int version) { int old_version_timer; IGMP_LOCK_ASSERT(); CTR4(KTR_IGMPV3, "%s: switching to v%d on ifp %p(%s)", __func__, version, igi->igi_ifp, igi->igi_ifp->if_xname); if (version == IGMP_VERSION_1 || version == IGMP_VERSION_2) { /* * Compute the "Older Version Querier Present" timer as per * Section 8.12. */ old_version_timer = igi->igi_rv * igi->igi_qi + igi->igi_qri; old_version_timer *= PR_SLOWHZ; if (version == IGMP_VERSION_1) { igi->igi_v1_timer = old_version_timer; igi->igi_v2_timer = 0; } else if (version == IGMP_VERSION_2) { igi->igi_v1_timer = 0; igi->igi_v2_timer = old_version_timer; } } if (igi->igi_v1_timer == 0 && igi->igi_v2_timer > 0) { if (igi->igi_version != IGMP_VERSION_2) { igi->igi_version = IGMP_VERSION_2; igmp_v3_cancel_link_timers(igi); } } else if (igi->igi_v1_timer > 0) { if (igi->igi_version != IGMP_VERSION_1) { igi->igi_version = IGMP_VERSION_1; igmp_v3_cancel_link_timers(igi); } } } /* * Cancel pending IGMPv3 timers for the given link and all groups * joined on it; state-change, general-query, and group-query timers. * * Only ever called on a transition from v3 to Compatibility mode. Kill * the timers stone dead (this may be expensive for large N groups), they * will be restarted if Compatibility Mode deems that they must be due to * query processing. */ static void igmp_v3_cancel_link_timers(struct igmp_ifsoftc *igi) { struct ifmultiaddr *ifma; struct ifnet *ifp; struct in_multi *inm, *tinm; CTR3(KTR_IGMPV3, "%s: cancel v3 timers on ifp %p(%s)", __func__, igi->igi_ifp, igi->igi_ifp->if_xname); IN_MULTI_LOCK_ASSERT(); IGMP_LOCK_ASSERT(); /* * Stop the v3 General Query Response on this link stone dead. * If fasttimo is woken up due to V_interface_timers_running, * the flag will be cleared if there are no pending link timers. */ igi->igi_v3_timer = 0; /* * Now clear the current-state and state-change report timers * for all memberships scoped to this link. */ ifp = igi->igi_ifp; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET || ifma->ifma_protospec == NULL) continue; inm = (struct in_multi *)ifma->ifma_protospec; switch (inm->inm_state) { case IGMP_NOT_MEMBER: case IGMP_SILENT_MEMBER: case IGMP_IDLE_MEMBER: case IGMP_LAZY_MEMBER: case IGMP_SLEEPING_MEMBER: case IGMP_AWAKENING_MEMBER: /* * These states are either not relevant in v3 mode, * or are unreported. Do nothing. */ break; case IGMP_LEAVING_MEMBER: /* * If we are leaving the group and switching to * compatibility mode, we need to release the final * reference held for issuing the INCLUDE {}, and * transition to REPORTING to ensure the host leave * message is sent upstream to the old querier -- * transition to NOT would lose the leave and race. */ SLIST_INSERT_HEAD(&igi->igi_relinmhead, inm, inm_nrele); /* FALLTHROUGH */ case IGMP_G_QUERY_PENDING_MEMBER: case IGMP_SG_QUERY_PENDING_MEMBER: inm_clear_recorded(inm); /* FALLTHROUGH */ case IGMP_REPORTING_MEMBER: inm->inm_state = IGMP_REPORTING_MEMBER; break; } /* * Always clear state-change and group report timers. * Free any pending IGMPv3 state-change records. */ inm->inm_sctimer = 0; inm->inm_timer = 0; mbufq_drain(&inm->inm_scq); } IF_ADDR_RUNLOCK(ifp); SLIST_FOREACH_SAFE(inm, &igi->igi_relinmhead, inm_nrele, tinm) { SLIST_REMOVE_HEAD(&igi->igi_relinmhead, inm_nrele); inm_release_locked(inm); } } /* * Update the Older Version Querier Present timers for a link. * See Section 7.2.1 of RFC 3376. */ static void igmp_v1v2_process_querier_timers(struct igmp_ifsoftc *igi) { IGMP_LOCK_ASSERT(); if (igi->igi_v1_timer == 0 && igi->igi_v2_timer == 0) { /* * IGMPv1 and IGMPv2 Querier Present timers expired. * * Revert to IGMPv3. */ if (igi->igi_version != IGMP_VERSION_3) { CTR5(KTR_IGMPV3, "%s: transition from v%d -> v%d on %p(%s)", __func__, igi->igi_version, IGMP_VERSION_3, igi->igi_ifp, igi->igi_ifp->if_xname); igi->igi_version = IGMP_VERSION_3; } } else if (igi->igi_v1_timer == 0 && igi->igi_v2_timer > 0) { /* * IGMPv1 Querier Present timer expired, * IGMPv2 Querier Present timer running. * If IGMPv2 was disabled since last timeout, * revert to IGMPv3. * If IGMPv2 is enabled, revert to IGMPv2. */ if (!V_igmp_v2enable) { CTR5(KTR_IGMPV3, "%s: transition from v%d -> v%d on %p(%s)", __func__, igi->igi_version, IGMP_VERSION_3, igi->igi_ifp, igi->igi_ifp->if_xname); igi->igi_v2_timer = 0; igi->igi_version = IGMP_VERSION_3; } else { --igi->igi_v2_timer; if (igi->igi_version != IGMP_VERSION_2) { CTR5(KTR_IGMPV3, "%s: transition from v%d -> v%d on %p(%s)", __func__, igi->igi_version, IGMP_VERSION_2, igi->igi_ifp, igi->igi_ifp->if_xname); igi->igi_version = IGMP_VERSION_2; igmp_v3_cancel_link_timers(igi); } } } else if (igi->igi_v1_timer > 0) { /* * IGMPv1 Querier Present timer running. * Stop IGMPv2 timer if running. * * If IGMPv1 was disabled since last timeout, * revert to IGMPv3. * If IGMPv1 is enabled, reset IGMPv2 timer if running. */ if (!V_igmp_v1enable) { CTR5(KTR_IGMPV3, "%s: transition from v%d -> v%d on %p(%s)", __func__, igi->igi_version, IGMP_VERSION_3, igi->igi_ifp, igi->igi_ifp->if_xname); igi->igi_v1_timer = 0; igi->igi_version = IGMP_VERSION_3; } else { --igi->igi_v1_timer; } if (igi->igi_v2_timer > 0) { CTR3(KTR_IGMPV3, "%s: cancel v2 timer on %p(%s)", __func__, igi->igi_ifp, igi->igi_ifp->if_xname); igi->igi_v2_timer = 0; } } } /* * Global slowtimo handler. * VIMAGE: Timeout handlers are expected to service all vimages. */ void igmp_slowtimo(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); igmp_slowtimo_vnet(); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * Per-vnet slowtimo handler. */ static void igmp_slowtimo_vnet(void) { struct igmp_ifsoftc *igi; IGMP_LOCK(); LIST_FOREACH(igi, &V_igi_head, igi_link) { igmp_v1v2_process_querier_timers(igi); } IGMP_UNLOCK(); } /* * Dispatch an IGMPv1/v2 host report or leave message. * These are always small enough to fit inside a single mbuf. */ static int igmp_v1v2_queue_report(struct in_multi *inm, const int type) { struct ifnet *ifp; struct igmp *igmp; struct ip *ip; struct mbuf *m; IN_MULTI_LOCK_ASSERT(); IGMP_LOCK_ASSERT(); ifp = inm->inm_ifp; m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) return (ENOMEM); M_ALIGN(m, sizeof(struct ip) + sizeof(struct igmp)); m->m_pkthdr.len = sizeof(struct ip) + sizeof(struct igmp); m->m_data += sizeof(struct ip); m->m_len = sizeof(struct igmp); igmp = mtod(m, struct igmp *); igmp->igmp_type = type; igmp->igmp_code = 0; igmp->igmp_group = inm->inm_addr; igmp->igmp_cksum = 0; igmp->igmp_cksum = in_cksum(m, sizeof(struct igmp)); m->m_data -= sizeof(struct ip); m->m_len += sizeof(struct ip); ip = mtod(m, struct ip *); ip->ip_tos = 0; ip->ip_len = htons(sizeof(struct ip) + sizeof(struct igmp)); ip->ip_off = 0; ip->ip_p = IPPROTO_IGMP; ip->ip_src.s_addr = INADDR_ANY; if (type == IGMP_HOST_LEAVE_MESSAGE) ip->ip_dst.s_addr = htonl(INADDR_ALLRTRS_GROUP); else ip->ip_dst = inm->inm_addr; igmp_save_context(m, ifp); m->m_flags |= M_IGMPV2; if (inm->inm_igi->igi_flags & IGIF_LOOPBACK) m->m_flags |= M_IGMP_LOOP; CTR2(KTR_IGMPV3, "%s: netisr_dispatch(NETISR_IGMP, %p)", __func__, m); netisr_dispatch(NETISR_IGMP, m); return (0); } /* * Process a state change from the upper layer for the given IPv4 group. * * Each socket holds a reference on the in_multi in its own ip_moptions. * The socket layer will have made the necessary updates to.the group * state, it is now up to IGMP to issue a state change report if there * has been any change between T0 (when the last state-change was issued) * and T1 (now). * * We use the IGMPv3 state machine at group level. The IGMP module * however makes the decision as to which IGMP protocol version to speak. * A state change *from* INCLUDE {} always means an initial join. * A state change *to* INCLUDE {} always means a final leave. * * FUTURE: If IGIF_V3LITE is enabled for this interface, then we can * save ourselves a bunch of work; any exclusive mode groups need not * compute source filter lists. * * VIMAGE: curvnet should have been set by caller, as this routine * is called from the socket option handlers. */ int igmp_change_state(struct in_multi *inm) { struct igmp_ifsoftc *igi; struct ifnet *ifp; int error; IN_MULTI_LOCK_ASSERT(); error = 0; /* * Try to detect if the upper layer just asked us to change state * for an interface which has now gone away. */ KASSERT(inm->inm_ifma != NULL, ("%s: no ifma", __func__)); ifp = inm->inm_ifma->ifma_ifp; /* * Sanity check that netinet's notion of ifp is the * same as net's. */ KASSERT(inm->inm_ifp == ifp, ("%s: bad ifp", __func__)); IGMP_LOCK(); igi = ((struct in_ifinfo *)ifp->if_afdata[AF_INET])->ii_igmp; KASSERT(igi != NULL, ("%s: no igmp_ifsoftc for ifp %p", __func__, ifp)); /* * If we detect a state transition to or from MCAST_UNDEFINED * for this group, then we are starting or finishing an IGMP * life cycle for this group. */ if (inm->inm_st[1].iss_fmode != inm->inm_st[0].iss_fmode) { CTR3(KTR_IGMPV3, "%s: inm transition %d -> %d", __func__, inm->inm_st[0].iss_fmode, inm->inm_st[1].iss_fmode); if (inm->inm_st[0].iss_fmode == MCAST_UNDEFINED) { CTR1(KTR_IGMPV3, "%s: initial join", __func__); error = igmp_initial_join(inm, igi); goto out_locked; } else if (inm->inm_st[1].iss_fmode == MCAST_UNDEFINED) { CTR1(KTR_IGMPV3, "%s: final leave", __func__); igmp_final_leave(inm, igi); goto out_locked; } } else { CTR1(KTR_IGMPV3, "%s: filter set change", __func__); } error = igmp_handle_state_change(inm, igi); out_locked: IGMP_UNLOCK(); return (error); } /* * Perform the initial join for an IGMP group. * * When joining a group: * If the group should have its IGMP traffic suppressed, do nothing. * IGMPv1 starts sending IGMPv1 host membership reports. * IGMPv2 starts sending IGMPv2 host membership reports. * IGMPv3 will schedule an IGMPv3 state-change report containing the * initial state of the membership. */ static int igmp_initial_join(struct in_multi *inm, struct igmp_ifsoftc *igi) { struct ifnet *ifp; struct mbufq *mq; int error, retval, syncstates; CTR4(KTR_IGMPV3, "%s: initial join %s on ifp %p(%s)", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp, inm->inm_ifp->if_xname); error = 0; syncstates = 1; ifp = inm->inm_ifp; IN_MULTI_LOCK_ASSERT(); IGMP_LOCK_ASSERT(); KASSERT(igi && igi->igi_ifp == ifp, ("%s: inconsistent ifp", __func__)); /* * Groups joined on loopback or marked as 'not reported', * e.g. 224.0.0.1, enter the IGMP_SILENT_MEMBER state and * are never reported in any IGMP protocol exchanges. * All other groups enter the appropriate IGMP state machine * for the version in use on this link. * A link marked as IGIF_SILENT causes IGMP to be completely * disabled for the link. */ if ((ifp->if_flags & IFF_LOOPBACK) || (igi->igi_flags & IGIF_SILENT) || !igmp_isgroupreported(inm->inm_addr)) { CTR1(KTR_IGMPV3, "%s: not kicking state machine for silent group", __func__); inm->inm_state = IGMP_SILENT_MEMBER; inm->inm_timer = 0; } else { /* * Deal with overlapping in_multi lifecycle. * If this group was LEAVING, then make sure * we drop the reference we picked up to keep the * group around for the final INCLUDE {} enqueue. */ if (igi->igi_version == IGMP_VERSION_3 && inm->inm_state == IGMP_LEAVING_MEMBER) inm_release_locked(inm); inm->inm_state = IGMP_REPORTING_MEMBER; switch (igi->igi_version) { case IGMP_VERSION_1: case IGMP_VERSION_2: inm->inm_state = IGMP_IDLE_MEMBER; error = igmp_v1v2_queue_report(inm, (igi->igi_version == IGMP_VERSION_2) ? IGMP_v2_HOST_MEMBERSHIP_REPORT : IGMP_v1_HOST_MEMBERSHIP_REPORT); if (error == 0) { inm->inm_timer = IGMP_RANDOM_DELAY( IGMP_V1V2_MAX_RI * PR_FASTHZ); V_current_state_timers_running = 1; } break; case IGMP_VERSION_3: /* * Defer update of T0 to T1, until the first copy * of the state change has been transmitted. */ syncstates = 0; /* * Immediately enqueue a State-Change Report for * this interface, freeing any previous reports. * Don't kick the timers if there is nothing to do, * or if an error occurred. */ mq = &inm->inm_scq; mbufq_drain(mq); retval = igmp_v3_enqueue_group_record(mq, inm, 1, 0, 0); CTR2(KTR_IGMPV3, "%s: enqueue record = %d", __func__, retval); if (retval <= 0) { error = retval * -1; break; } /* * Schedule transmission of pending state-change * report up to RV times for this link. The timer * will fire at the next igmp_fasttimo (~200ms), * giving us an opportunity to merge the reports. */ if (igi->igi_flags & IGIF_LOOPBACK) { inm->inm_scrv = 1; } else { KASSERT(igi->igi_rv > 1, ("%s: invalid robustness %d", __func__, igi->igi_rv)); inm->inm_scrv = igi->igi_rv; } inm->inm_sctimer = 1; V_state_change_timers_running = 1; error = 0; break; } } /* * Only update the T0 state if state change is atomic, * i.e. we don't need to wait for a timer to fire before we * can consider the state change to have been communicated. */ if (syncstates) { inm_commit(inm); CTR3(KTR_IGMPV3, "%s: T1 -> T0 for %s/%s", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp->if_xname); } return (error); } /* * Issue an intermediate state change during the IGMP life-cycle. */ static int igmp_handle_state_change(struct in_multi *inm, struct igmp_ifsoftc *igi) { struct ifnet *ifp; int retval; CTR4(KTR_IGMPV3, "%s: state change for %s on ifp %p(%s)", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp, inm->inm_ifp->if_xname); ifp = inm->inm_ifp; IN_MULTI_LOCK_ASSERT(); IGMP_LOCK_ASSERT(); KASSERT(igi && igi->igi_ifp == ifp, ("%s: inconsistent ifp", __func__)); if ((ifp->if_flags & IFF_LOOPBACK) || (igi->igi_flags & IGIF_SILENT) || !igmp_isgroupreported(inm->inm_addr) || (igi->igi_version != IGMP_VERSION_3)) { if (!igmp_isgroupreported(inm->inm_addr)) { CTR1(KTR_IGMPV3, "%s: not kicking state machine for silent group", __func__); } CTR1(KTR_IGMPV3, "%s: nothing to do", __func__); inm_commit(inm); CTR3(KTR_IGMPV3, "%s: T1 -> T0 for %s/%s", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp->if_xname); return (0); } mbufq_drain(&inm->inm_scq); retval = igmp_v3_enqueue_group_record(&inm->inm_scq, inm, 1, 0, 0); CTR2(KTR_IGMPV3, "%s: enqueue record = %d", __func__, retval); if (retval <= 0) return (-retval); /* * If record(s) were enqueued, start the state-change * report timer for this group. */ inm->inm_scrv = ((igi->igi_flags & IGIF_LOOPBACK) ? 1 : igi->igi_rv); inm->inm_sctimer = 1; V_state_change_timers_running = 1; return (0); } /* * Perform the final leave for an IGMP group. * * When leaving a group: * IGMPv1 does nothing. * IGMPv2 sends a host leave message, if and only if we are the reporter. * IGMPv3 enqueues a state-change report containing a transition * to INCLUDE {} for immediate transmission. */ static void igmp_final_leave(struct in_multi *inm, struct igmp_ifsoftc *igi) { int syncstates; syncstates = 1; CTR4(KTR_IGMPV3, "%s: final leave %s on ifp %p(%s)", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp, inm->inm_ifp->if_xname); IN_MULTI_LOCK_ASSERT(); IGMP_LOCK_ASSERT(); switch (inm->inm_state) { case IGMP_NOT_MEMBER: case IGMP_SILENT_MEMBER: case IGMP_LEAVING_MEMBER: /* Already leaving or left; do nothing. */ CTR1(KTR_IGMPV3, "%s: not kicking state machine for silent group", __func__); break; case IGMP_REPORTING_MEMBER: case IGMP_IDLE_MEMBER: case IGMP_G_QUERY_PENDING_MEMBER: case IGMP_SG_QUERY_PENDING_MEMBER: if (igi->igi_version == IGMP_VERSION_2) { #ifdef INVARIANTS if (inm->inm_state == IGMP_G_QUERY_PENDING_MEMBER || inm->inm_state == IGMP_SG_QUERY_PENDING_MEMBER) panic("%s: IGMPv3 state reached, not IGMPv3 mode", __func__); #endif igmp_v1v2_queue_report(inm, IGMP_HOST_LEAVE_MESSAGE); inm->inm_state = IGMP_NOT_MEMBER; } else if (igi->igi_version == IGMP_VERSION_3) { /* * Stop group timer and all pending reports. * Immediately enqueue a state-change report * TO_IN {} to be sent on the next fast timeout, * giving us an opportunity to merge reports. */ mbufq_drain(&inm->inm_scq); inm->inm_timer = 0; if (igi->igi_flags & IGIF_LOOPBACK) { inm->inm_scrv = 1; } else { inm->inm_scrv = igi->igi_rv; } CTR4(KTR_IGMPV3, "%s: Leaving %s/%s with %d " "pending retransmissions.", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp->if_xname, inm->inm_scrv); if (inm->inm_scrv == 0) { inm->inm_state = IGMP_NOT_MEMBER; inm->inm_sctimer = 0; } else { int retval; inm_acquire_locked(inm); retval = igmp_v3_enqueue_group_record( &inm->inm_scq, inm, 1, 0, 0); KASSERT(retval != 0, ("%s: enqueue record = %d", __func__, retval)); inm->inm_state = IGMP_LEAVING_MEMBER; inm->inm_sctimer = 1; V_state_change_timers_running = 1; syncstates = 0; } break; } break; case IGMP_LAZY_MEMBER: case IGMP_SLEEPING_MEMBER: case IGMP_AWAKENING_MEMBER: /* Our reports are suppressed; do nothing. */ break; } if (syncstates) { inm_commit(inm); CTR3(KTR_IGMPV3, "%s: T1 -> T0 for %s/%s", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp->if_xname); inm->inm_st[1].iss_fmode = MCAST_UNDEFINED; CTR3(KTR_IGMPV3, "%s: T1 now MCAST_UNDEFINED for %s/%s", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp->if_xname); } } /* * Enqueue an IGMPv3 group record to the given output queue. * * XXX This function could do with having the allocation code * split out, and the multiple-tree-walks coalesced into a single * routine as has been done in igmp_v3_enqueue_filter_change(). * * If is_state_change is zero, a current-state record is appended. * If is_state_change is non-zero, a state-change report is appended. * * If is_group_query is non-zero, an mbuf packet chain is allocated. * If is_group_query is zero, and if there is a packet with free space * at the tail of the queue, it will be appended to providing there * is enough free space. * Otherwise a new mbuf packet chain is allocated. * * If is_source_query is non-zero, each source is checked to see if * it was recorded for a Group-Source query, and will be omitted if * it is not both in-mode and recorded. * * The function will attempt to allocate leading space in the packet * for the IP/IGMP header to be prepended without fragmenting the chain. * * If successful the size of all data appended to the queue is returned, * otherwise an error code less than zero is returned, or zero if * no record(s) were appended. */ static int igmp_v3_enqueue_group_record(struct mbufq *mq, struct in_multi *inm, const int is_state_change, const int is_group_query, const int is_source_query) { struct igmp_grouprec ig; struct igmp_grouprec *pig; struct ifnet *ifp; struct ip_msource *ims, *nims; struct mbuf *m0, *m, *md; int error, is_filter_list_change; int minrec0len, m0srcs, msrcs, nbytes, off; int record_has_sources; int now; int type; in_addr_t naddr; uint8_t mode; IN_MULTI_LOCK_ASSERT(); error = 0; ifp = inm->inm_ifp; is_filter_list_change = 0; m = NULL; m0 = NULL; m0srcs = 0; msrcs = 0; nbytes = 0; nims = NULL; record_has_sources = 1; pig = NULL; type = IGMP_DO_NOTHING; mode = inm->inm_st[1].iss_fmode; /* * If we did not transition out of ASM mode during t0->t1, * and there are no source nodes to process, we can skip * the generation of source records. */ if (inm->inm_st[0].iss_asm > 0 && inm->inm_st[1].iss_asm > 0 && inm->inm_nsrc == 0) record_has_sources = 0; if (is_state_change) { /* * Queue a state change record. * If the mode did not change, and there are non-ASM * listeners or source filters present, * we potentially need to issue two records for the group. * If we are transitioning to MCAST_UNDEFINED, we need * not send any sources. * If there are ASM listeners, and there was no filter * mode transition of any kind, do nothing. */ if (mode != inm->inm_st[0].iss_fmode) { if (mode == MCAST_EXCLUDE) { CTR1(KTR_IGMPV3, "%s: change to EXCLUDE", __func__); type = IGMP_CHANGE_TO_EXCLUDE_MODE; } else { CTR1(KTR_IGMPV3, "%s: change to INCLUDE", __func__); type = IGMP_CHANGE_TO_INCLUDE_MODE; if (mode == MCAST_UNDEFINED) record_has_sources = 0; } } else { if (record_has_sources) { is_filter_list_change = 1; } else { type = IGMP_DO_NOTHING; } } } else { /* * Queue a current state record. */ if (mode == MCAST_EXCLUDE) { type = IGMP_MODE_IS_EXCLUDE; } else if (mode == MCAST_INCLUDE) { type = IGMP_MODE_IS_INCLUDE; KASSERT(inm->inm_st[1].iss_asm == 0, ("%s: inm %p is INCLUDE but ASM count is %d", __func__, inm, inm->inm_st[1].iss_asm)); } } /* * Generate the filter list changes using a separate function. */ if (is_filter_list_change) return (igmp_v3_enqueue_filter_change(mq, inm)); if (type == IGMP_DO_NOTHING) { CTR3(KTR_IGMPV3, "%s: nothing to do for %s/%s", __func__, inet_ntoa(inm->inm_addr), inm->inm_ifp->if_xname); return (0); } /* * If any sources are present, we must be able to fit at least * one in the trailing space of the tail packet's mbuf, * ideally more. */ minrec0len = sizeof(struct igmp_grouprec); if (record_has_sources) minrec0len += sizeof(in_addr_t); CTR4(KTR_IGMPV3, "%s: queueing %s for %s/%s", __func__, igmp_rec_type_to_str(type), inet_ntoa(inm->inm_addr), inm->inm_ifp->if_xname); /* * Check if we have a packet in the tail of the queue for this * group into which the first group record for this group will fit. * Otherwise allocate a new packet. * Always allocate leading space for IP+RA_OPT+IGMP+REPORT. * Note: Group records for G/GSR query responses MUST be sent * in their own packet. */ m0 = mbufq_last(mq); if (!is_group_query && m0 != NULL && (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <= IGMP_V3_REPORT_MAXRECS) && (m0->m_pkthdr.len + minrec0len) < (ifp->if_mtu - IGMP_LEADINGSPACE)) { m0srcs = (ifp->if_mtu - m0->m_pkthdr.len - sizeof(struct igmp_grouprec)) / sizeof(in_addr_t); m = m0; CTR1(KTR_IGMPV3, "%s: use existing packet", __func__); } else { if (mbufq_full(mq)) { CTR1(KTR_IGMPV3, "%s: outbound queue full", __func__); return (-ENOMEM); } m = NULL; m0srcs = (ifp->if_mtu - IGMP_LEADINGSPACE - sizeof(struct igmp_grouprec)) / sizeof(in_addr_t); if (!is_state_change && !is_group_query) { m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (m) m->m_data += IGMP_LEADINGSPACE; } if (m == NULL) { m = m_gethdr(M_NOWAIT, MT_DATA); if (m) M_ALIGN(m, IGMP_LEADINGSPACE); } if (m == NULL) return (-ENOMEM); igmp_save_context(m, ifp); CTR1(KTR_IGMPV3, "%s: allocated first packet", __func__); } /* * Append group record. * If we have sources, we don't know how many yet. */ ig.ig_type = type; ig.ig_datalen = 0; ig.ig_numsrc = 0; ig.ig_group = inm->inm_addr; if (!m_append(m, sizeof(struct igmp_grouprec), (void *)&ig)) { if (m != m0) m_freem(m); CTR1(KTR_IGMPV3, "%s: m_append() failed.", __func__); return (-ENOMEM); } nbytes += sizeof(struct igmp_grouprec); /* * Append as many sources as will fit in the first packet. * If we are appending to a new packet, the chain allocation * may potentially use clusters; use m_getptr() in this case. * If we are appending to an existing packet, we need to obtain * a pointer to the group record after m_append(), in case a new * mbuf was allocated. * Only append sources which are in-mode at t1. If we are * transitioning to MCAST_UNDEFINED state on the group, do not * include source entries. * Only report recorded sources in our filter set when responding * to a group-source query. */ if (record_has_sources) { if (m == m0) { md = m_last(m); pig = (struct igmp_grouprec *)(mtod(md, uint8_t *) + md->m_len - nbytes); } else { md = m_getptr(m, 0, &off); pig = (struct igmp_grouprec *)(mtod(md, uint8_t *) + off); } msrcs = 0; RB_FOREACH_SAFE(ims, ip_msource_tree, &inm->inm_srcs, nims) { CTR2(KTR_IGMPV3, "%s: visit node %s", __func__, inet_ntoa_haddr(ims->ims_haddr)); now = ims_get_mode(inm, ims, 1); CTR2(KTR_IGMPV3, "%s: node is %d", __func__, now); if ((now != mode) || (now == mode && mode == MCAST_UNDEFINED)) { CTR1(KTR_IGMPV3, "%s: skip node", __func__); continue; } if (is_source_query && ims->ims_stp == 0) { CTR1(KTR_IGMPV3, "%s: skip unrecorded node", __func__); continue; } CTR1(KTR_IGMPV3, "%s: append node", __func__); naddr = htonl(ims->ims_haddr); if (!m_append(m, sizeof(in_addr_t), (void *)&naddr)) { if (m != m0) m_freem(m); CTR1(KTR_IGMPV3, "%s: m_append() failed.", __func__); return (-ENOMEM); } nbytes += sizeof(in_addr_t); ++msrcs; if (msrcs == m0srcs) break; } CTR2(KTR_IGMPV3, "%s: msrcs is %d this packet", __func__, msrcs); pig->ig_numsrc = htons(msrcs); nbytes += (msrcs * sizeof(in_addr_t)); } if (is_source_query && msrcs == 0) { CTR1(KTR_IGMPV3, "%s: no recorded sources to report", __func__); if (m != m0) m_freem(m); return (0); } /* * We are good to go with first packet. */ if (m != m0) { CTR1(KTR_IGMPV3, "%s: enqueueing first packet", __func__); m->m_pkthdr.PH_vt.vt_nrecs = 1; mbufq_enqueue(mq, m); } else m->m_pkthdr.PH_vt.vt_nrecs++; /* * No further work needed if no source list in packet(s). */ if (!record_has_sources) return (nbytes); /* * Whilst sources remain to be announced, we need to allocate * a new packet and fill out as many sources as will fit. * Always try for a cluster first. */ while (nims != NULL) { if (mbufq_full(mq)) { CTR1(KTR_IGMPV3, "%s: outbound queue full", __func__); return (-ENOMEM); } m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (m) m->m_data += IGMP_LEADINGSPACE; if (m == NULL) { m = m_gethdr(M_NOWAIT, MT_DATA); if (m) M_ALIGN(m, IGMP_LEADINGSPACE); } if (m == NULL) return (-ENOMEM); igmp_save_context(m, ifp); md = m_getptr(m, 0, &off); pig = (struct igmp_grouprec *)(mtod(md, uint8_t *) + off); CTR1(KTR_IGMPV3, "%s: allocated next packet", __func__); if (!m_append(m, sizeof(struct igmp_grouprec), (void *)&ig)) { if (m != m0) m_freem(m); CTR1(KTR_IGMPV3, "%s: m_append() failed.", __func__); return (-ENOMEM); } m->m_pkthdr.PH_vt.vt_nrecs = 1; nbytes += sizeof(struct igmp_grouprec); m0srcs = (ifp->if_mtu - IGMP_LEADINGSPACE - sizeof(struct igmp_grouprec)) / sizeof(in_addr_t); msrcs = 0; RB_FOREACH_FROM(ims, ip_msource_tree, nims) { CTR2(KTR_IGMPV3, "%s: visit node %s", __func__, inet_ntoa_haddr(ims->ims_haddr)); now = ims_get_mode(inm, ims, 1); if ((now != mode) || (now == mode && mode == MCAST_UNDEFINED)) { CTR1(KTR_IGMPV3, "%s: skip node", __func__); continue; } if (is_source_query && ims->ims_stp == 0) { CTR1(KTR_IGMPV3, "%s: skip unrecorded node", __func__); continue; } CTR1(KTR_IGMPV3, "%s: append node", __func__); naddr = htonl(ims->ims_haddr); if (!m_append(m, sizeof(in_addr_t), (void *)&naddr)) { if (m != m0) m_freem(m); CTR1(KTR_IGMPV3, "%s: m_append() failed.", __func__); return (-ENOMEM); } ++msrcs; if (msrcs == m0srcs) break; } pig->ig_numsrc = htons(msrcs); nbytes += (msrcs * sizeof(in_addr_t)); CTR1(KTR_IGMPV3, "%s: enqueueing next packet", __func__); mbufq_enqueue(mq, m); } return (nbytes); } /* * Type used to mark record pass completion. * We exploit the fact we can cast to this easily from the * current filter modes on each ip_msource node. */ typedef enum { REC_NONE = 0x00, /* MCAST_UNDEFINED */ REC_ALLOW = 0x01, /* MCAST_INCLUDE */ REC_BLOCK = 0x02, /* MCAST_EXCLUDE */ REC_FULL = REC_ALLOW | REC_BLOCK } rectype_t; /* * Enqueue an IGMPv3 filter list change to the given output queue. * * Source list filter state is held in an RB-tree. When the filter list * for a group is changed without changing its mode, we need to compute * the deltas between T0 and T1 for each source in the filter set, * and enqueue the appropriate ALLOW_NEW/BLOCK_OLD records. * * As we may potentially queue two record types, and the entire R-B tree * needs to be walked at once, we break this out into its own function * so we can generate a tightly packed queue of packets. * * XXX This could be written to only use one tree walk, although that makes * serializing into the mbuf chains a bit harder. For now we do two walks * which makes things easier on us, and it may or may not be harder on * the L2 cache. * * If successful the size of all data appended to the queue is returned, * otherwise an error code less than zero is returned, or zero if * no record(s) were appended. */ static int igmp_v3_enqueue_filter_change(struct mbufq *mq, struct in_multi *inm) { static const int MINRECLEN = sizeof(struct igmp_grouprec) + sizeof(in_addr_t); struct ifnet *ifp; struct igmp_grouprec ig; struct igmp_grouprec *pig; struct ip_msource *ims, *nims; struct mbuf *m, *m0, *md; in_addr_t naddr; int m0srcs, nbytes, npbytes, off, rsrcs, schanged; int nallow, nblock; uint8_t mode, now, then; rectype_t crt, drt, nrt; IN_MULTI_LOCK_ASSERT(); if (inm->inm_nsrc == 0 || (inm->inm_st[0].iss_asm > 0 && inm->inm_st[1].iss_asm > 0)) return (0); ifp = inm->inm_ifp; /* interface */ mode = inm->inm_st[1].iss_fmode; /* filter mode at t1 */ crt = REC_NONE; /* current group record type */ drt = REC_NONE; /* mask of completed group record types */ nrt = REC_NONE; /* record type for current node */ m0srcs = 0; /* # source which will fit in current mbuf chain */ nbytes = 0; /* # of bytes appended to group's state-change queue */ npbytes = 0; /* # of bytes appended this packet */ rsrcs = 0; /* # sources encoded in current record */ schanged = 0; /* # nodes encoded in overall filter change */ nallow = 0; /* # of source entries in ALLOW_NEW */ nblock = 0; /* # of source entries in BLOCK_OLD */ nims = NULL; /* next tree node pointer */ /* * For each possible filter record mode. * The first kind of source we encounter tells us which * is the first kind of record we start appending. * If a node transitioned to UNDEFINED at t1, its mode is treated * as the inverse of the group's filter mode. */ while (drt != REC_FULL) { do { m0 = mbufq_last(mq); if (m0 != NULL && (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <= IGMP_V3_REPORT_MAXRECS) && (m0->m_pkthdr.len + MINRECLEN) < (ifp->if_mtu - IGMP_LEADINGSPACE)) { m = m0; m0srcs = (ifp->if_mtu - m0->m_pkthdr.len - sizeof(struct igmp_grouprec)) / sizeof(in_addr_t); CTR1(KTR_IGMPV3, "%s: use previous packet", __func__); } else { m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (m) m->m_data += IGMP_LEADINGSPACE; if (m == NULL) { m = m_gethdr(M_NOWAIT, MT_DATA); if (m) M_ALIGN(m, IGMP_LEADINGSPACE); } if (m == NULL) { CTR1(KTR_IGMPV3, "%s: m_get*() failed", __func__); return (-ENOMEM); } m->m_pkthdr.PH_vt.vt_nrecs = 0; igmp_save_context(m, ifp); m0srcs = (ifp->if_mtu - IGMP_LEADINGSPACE - sizeof(struct igmp_grouprec)) / sizeof(in_addr_t); npbytes = 0; CTR1(KTR_IGMPV3, "%s: allocated new packet", __func__); } /* * Append the IGMP group record header to the * current packet's data area. * Recalculate pointer to free space for next * group record, in case m_append() allocated * a new mbuf or cluster. */ memset(&ig, 0, sizeof(ig)); ig.ig_group = inm->inm_addr; if (!m_append(m, sizeof(ig), (void *)&ig)) { if (m != m0) m_freem(m); CTR1(KTR_IGMPV3, "%s: m_append() failed", __func__); return (-ENOMEM); } npbytes += sizeof(struct igmp_grouprec); if (m != m0) { /* new packet; offset in c hain */ md = m_getptr(m, npbytes - sizeof(struct igmp_grouprec), &off); pig = (struct igmp_grouprec *)(mtod(md, uint8_t *) + off); } else { /* current packet; offset from last append */ md = m_last(m); pig = (struct igmp_grouprec *)(mtod(md, uint8_t *) + md->m_len - sizeof(struct igmp_grouprec)); } /* * Begin walking the tree for this record type * pass, or continue from where we left off * previously if we had to allocate a new packet. * Only report deltas in-mode at t1. * We need not report included sources as allowed * if we are in inclusive mode on the group, * however the converse is not true. */ rsrcs = 0; if (nims == NULL) nims = RB_MIN(ip_msource_tree, &inm->inm_srcs); RB_FOREACH_FROM(ims, ip_msource_tree, nims) { CTR2(KTR_IGMPV3, "%s: visit node %s", __func__, inet_ntoa_haddr(ims->ims_haddr)); now = ims_get_mode(inm, ims, 1); then = ims_get_mode(inm, ims, 0); CTR3(KTR_IGMPV3, "%s: mode: t0 %d, t1 %d", __func__, then, now); if (now == then) { CTR1(KTR_IGMPV3, "%s: skip unchanged", __func__); continue; } if (mode == MCAST_EXCLUDE && now == MCAST_INCLUDE) { CTR1(KTR_IGMPV3, "%s: skip IN src on EX group", __func__); continue; } nrt = (rectype_t)now; if (nrt == REC_NONE) nrt = (rectype_t)(~mode & REC_FULL); if (schanged++ == 0) { crt = nrt; } else if (crt != nrt) continue; naddr = htonl(ims->ims_haddr); if (!m_append(m, sizeof(in_addr_t), (void *)&naddr)) { if (m != m0) m_freem(m); CTR1(KTR_IGMPV3, "%s: m_append() failed", __func__); return (-ENOMEM); } nallow += !!(crt == REC_ALLOW); nblock += !!(crt == REC_BLOCK); if (++rsrcs == m0srcs) break; } /* * If we did not append any tree nodes on this * pass, back out of allocations. */ if (rsrcs == 0) { npbytes -= sizeof(struct igmp_grouprec); if (m != m0) { CTR1(KTR_IGMPV3, "%s: m_free(m)", __func__); m_freem(m); } else { CTR1(KTR_IGMPV3, "%s: m_adj(m, -ig)", __func__); m_adj(m, -((int)sizeof( struct igmp_grouprec))); } continue; } npbytes += (rsrcs * sizeof(in_addr_t)); if (crt == REC_ALLOW) pig->ig_type = IGMP_ALLOW_NEW_SOURCES; else if (crt == REC_BLOCK) pig->ig_type = IGMP_BLOCK_OLD_SOURCES; pig->ig_numsrc = htons(rsrcs); /* * Count the new group record, and enqueue this * packet if it wasn't already queued. */ m->m_pkthdr.PH_vt.vt_nrecs++; if (m != m0) mbufq_enqueue(mq, m); nbytes += npbytes; } while (nims != NULL); drt |= crt; crt = (~crt & REC_FULL); } CTR3(KTR_IGMPV3, "%s: queued %d ALLOW_NEW, %d BLOCK_OLD", __func__, nallow, nblock); return (nbytes); } static int igmp_v3_merge_state_changes(struct in_multi *inm, struct mbufq *scq) { struct mbufq *gq; struct mbuf *m; /* pending state-change */ struct mbuf *m0; /* copy of pending state-change */ struct mbuf *mt; /* last state-change in packet */ int docopy, domerge; u_int recslen; docopy = 0; domerge = 0; recslen = 0; IN_MULTI_LOCK_ASSERT(); IGMP_LOCK_ASSERT(); /* * If there are further pending retransmissions, make a writable * copy of each queued state-change message before merging. */ if (inm->inm_scrv > 0) docopy = 1; gq = &inm->inm_scq; #ifdef KTR if (mbufq_first(gq) == NULL) { CTR2(KTR_IGMPV3, "%s: WARNING: queue for inm %p is empty", __func__, inm); } #endif m = mbufq_first(gq); while (m != NULL) { /* * Only merge the report into the current packet if * there is sufficient space to do so; an IGMPv3 report * packet may only contain 65,535 group records. * Always use a simple mbuf chain concatentation to do this, * as large state changes for single groups may have * allocated clusters. */ domerge = 0; mt = mbufq_last(scq); if (mt != NULL) { recslen = m_length(m, NULL); if ((mt->m_pkthdr.PH_vt.vt_nrecs + m->m_pkthdr.PH_vt.vt_nrecs <= IGMP_V3_REPORT_MAXRECS) && (mt->m_pkthdr.len + recslen <= (inm->inm_ifp->if_mtu - IGMP_LEADINGSPACE))) domerge = 1; } if (!domerge && mbufq_full(gq)) { CTR2(KTR_IGMPV3, "%s: outbound queue full, skipping whole packet %p", __func__, m); mt = m->m_nextpkt; if (!docopy) m_freem(m); m = mt; continue; } if (!docopy) { CTR2(KTR_IGMPV3, "%s: dequeueing %p", __func__, m); m0 = mbufq_dequeue(gq); m = m0->m_nextpkt; } else { CTR2(KTR_IGMPV3, "%s: copying %p", __func__, m); m0 = m_dup(m, M_NOWAIT); if (m0 == NULL) return (ENOMEM); m0->m_nextpkt = NULL; m = m->m_nextpkt; } if (!domerge) { CTR3(KTR_IGMPV3, "%s: queueing %p to scq %p)", __func__, m0, scq); mbufq_enqueue(scq, m0); } else { struct mbuf *mtl; /* last mbuf of packet mt */ CTR3(KTR_IGMPV3, "%s: merging %p with scq tail %p)", __func__, m0, mt); mtl = m_last(mt); m0->m_flags &= ~M_PKTHDR; mt->m_pkthdr.len += recslen; mt->m_pkthdr.PH_vt.vt_nrecs += m0->m_pkthdr.PH_vt.vt_nrecs; mtl->m_next = m0; } } return (0); } /* * Respond to a pending IGMPv3 General Query. */ static void igmp_v3_dispatch_general_query(struct igmp_ifsoftc *igi) { struct ifmultiaddr *ifma; struct ifnet *ifp; struct in_multi *inm; int retval, loop; IN_MULTI_LOCK_ASSERT(); IGMP_LOCK_ASSERT(); KASSERT(igi->igi_version == IGMP_VERSION_3, ("%s: called when version %d", __func__, igi->igi_version)); /* * Check that there are some packets queued. If so, send them first. * For large number of groups the reply to general query can take * many packets, we should finish sending them before starting of * queuing the new reply. */ if (mbufq_len(&igi->igi_gq) != 0) goto send; ifp = igi->igi_ifp; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET || ifma->ifma_protospec == NULL) continue; inm = (struct in_multi *)ifma->ifma_protospec; KASSERT(ifp == inm->inm_ifp, ("%s: inconsistent ifp", __func__)); switch (inm->inm_state) { case IGMP_NOT_MEMBER: case IGMP_SILENT_MEMBER: break; case IGMP_REPORTING_MEMBER: case IGMP_IDLE_MEMBER: case IGMP_LAZY_MEMBER: case IGMP_SLEEPING_MEMBER: case IGMP_AWAKENING_MEMBER: inm->inm_state = IGMP_REPORTING_MEMBER; retval = igmp_v3_enqueue_group_record(&igi->igi_gq, inm, 0, 0, 0); CTR2(KTR_IGMPV3, "%s: enqueue record = %d", __func__, retval); break; case IGMP_G_QUERY_PENDING_MEMBER: case IGMP_SG_QUERY_PENDING_MEMBER: case IGMP_LEAVING_MEMBER: break; } } IF_ADDR_RUNLOCK(ifp); send: loop = (igi->igi_flags & IGIF_LOOPBACK) ? 1 : 0; igmp_dispatch_queue(&igi->igi_gq, IGMP_MAX_RESPONSE_BURST, loop); /* * Slew transmission of bursts over 500ms intervals. */ if (mbufq_first(&igi->igi_gq) != NULL) { igi->igi_v3_timer = 1 + IGMP_RANDOM_DELAY( IGMP_RESPONSE_BURST_INTERVAL); V_interface_timers_running = 1; } } /* * Transmit the next pending IGMP message in the output queue. * * We get called from netisr_processqueue(). A mutex private to igmpoq * will be acquired and released around this routine. * * VIMAGE: Needs to store/restore vnet pointer on a per-mbuf-chain basis. * MRT: Nothing needs to be done, as IGMP traffic is always local to * a link and uses a link-scope multicast address. */ static void igmp_intr(struct mbuf *m) { struct ip_moptions imo; struct ifnet *ifp; struct mbuf *ipopts, *m0; int error; uint32_t ifindex; CTR2(KTR_IGMPV3, "%s: transmit %p", __func__, m); /* * Set VNET image pointer from enqueued mbuf chain * before doing anything else. Whilst we use interface * indexes to guard against interface detach, they are * unique to each VIMAGE and must be retrieved. */ CURVNET_SET((struct vnet *)(m->m_pkthdr.PH_loc.ptr)); ifindex = igmp_restore_context(m); /* * Check if the ifnet still exists. This limits the scope of * any race in the absence of a global ifp lock for low cost * (an array lookup). */ ifp = ifnet_byindex(ifindex); if (ifp == NULL) { CTR3(KTR_IGMPV3, "%s: dropped %p as ifindex %u went away.", __func__, m, ifindex); m_freem(m); IPSTAT_INC(ips_noroute); goto out; } ipopts = V_igmp_sendra ? m_raopt : NULL; imo.imo_multicast_ttl = 1; imo.imo_multicast_vif = -1; imo.imo_multicast_loop = (V_ip_mrouter != NULL); /* * If the user requested that IGMP traffic be explicitly * redirected to the loopback interface (e.g. they are running a * MANET interface and the routing protocol needs to see the * updates), handle this now. */ if (m->m_flags & M_IGMP_LOOP) imo.imo_multicast_ifp = V_loif; else imo.imo_multicast_ifp = ifp; if (m->m_flags & M_IGMPV2) { m0 = m; } else { m0 = igmp_v3_encap_report(ifp, m); if (m0 == NULL) { CTR2(KTR_IGMPV3, "%s: dropped %p", __func__, m); m_freem(m); IPSTAT_INC(ips_odropped); goto out; } } igmp_scrub_context(m0); m_clrprotoflags(m); m0->m_pkthdr.rcvif = V_loif; #ifdef MAC mac_netinet_igmp_send(ifp, m0); #endif error = ip_output(m0, ipopts, NULL, 0, &imo, NULL); if (error) { CTR3(KTR_IGMPV3, "%s: ip_output(%p) = %d", __func__, m0, error); goto out; } IGMPSTAT_INC(igps_snd_reports); out: /* * We must restore the existing vnet pointer before * continuing as we are run from netisr context. */ CURVNET_RESTORE(); } /* * Encapsulate an IGMPv3 report. * * The internal mbuf flag M_IGMPV3_HDR is used to indicate that the mbuf * chain has already had its IP/IGMPv3 header prepended. In this case * the function will not attempt to prepend; the lengths and checksums * will however be re-computed. * * Returns a pointer to the new mbuf chain head, or NULL if the * allocation failed. */ static struct mbuf * igmp_v3_encap_report(struct ifnet *ifp, struct mbuf *m) { struct rm_priotracker in_ifa_tracker; struct igmp_report *igmp; struct ip *ip; int hdrlen, igmpreclen; KASSERT((m->m_flags & M_PKTHDR), ("%s: mbuf chain %p is !M_PKTHDR", __func__, m)); igmpreclen = m_length(m, NULL); hdrlen = sizeof(struct ip) + sizeof(struct igmp_report); if (m->m_flags & M_IGMPV3_HDR) { igmpreclen -= hdrlen; } else { M_PREPEND(m, hdrlen, M_NOWAIT); if (m == NULL) return (NULL); m->m_flags |= M_IGMPV3_HDR; } CTR2(KTR_IGMPV3, "%s: igmpreclen is %d", __func__, igmpreclen); m->m_data += sizeof(struct ip); m->m_len -= sizeof(struct ip); igmp = mtod(m, struct igmp_report *); igmp->ir_type = IGMP_v3_HOST_MEMBERSHIP_REPORT; igmp->ir_rsv1 = 0; igmp->ir_rsv2 = 0; igmp->ir_numgrps = htons(m->m_pkthdr.PH_vt.vt_nrecs); igmp->ir_cksum = 0; igmp->ir_cksum = in_cksum(m, sizeof(struct igmp_report) + igmpreclen); m->m_pkthdr.PH_vt.vt_nrecs = 0; m->m_data -= sizeof(struct ip); m->m_len += sizeof(struct ip); ip = mtod(m, struct ip *); ip->ip_tos = IPTOS_PREC_INTERNETCONTROL; ip->ip_len = htons(hdrlen + igmpreclen); ip->ip_off = htons(IP_DF); ip->ip_p = IPPROTO_IGMP; ip->ip_sum = 0; ip->ip_src.s_addr = INADDR_ANY; if (m->m_flags & M_IGMP_LOOP) { struct in_ifaddr *ia; IFP_TO_IA(ifp, ia, &in_ifa_tracker); if (ia != NULL) { ip->ip_src = ia->ia_addr.sin_addr; ifa_free(&ia->ia_ifa); } } ip->ip_dst.s_addr = htonl(INADDR_ALLRPTS_GROUP); return (m); } #ifdef KTR static char * igmp_rec_type_to_str(const int type) { switch (type) { case IGMP_CHANGE_TO_EXCLUDE_MODE: return "TO_EX"; break; case IGMP_CHANGE_TO_INCLUDE_MODE: return "TO_IN"; break; case IGMP_MODE_IS_EXCLUDE: return "MODE_EX"; break; case IGMP_MODE_IS_INCLUDE: return "MODE_IN"; break; case IGMP_ALLOW_NEW_SOURCES: return "ALLOW_NEW"; break; case IGMP_BLOCK_OLD_SOURCES: return "BLOCK_OLD"; break; default: break; } return "unknown"; } #endif +#ifdef VIMAGE static void -igmp_init(void *unused __unused) +vnet_igmp_uninit(const void *unused __unused) { - CTR1(KTR_IGMPV3, "%s: initializing", __func__); - - IGMP_LOCK_INIT(); - - m_raopt = igmp_ra_alloc(); - - netisr_register(&igmp_nh); -} -SYSINIT(igmp_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, igmp_init, NULL); - -static void -igmp_uninit(void *unused __unused) -{ - CTR1(KTR_IGMPV3, "%s: tearing down", __func__); - netisr_unregister(&igmp_nh); - - m_free(m_raopt); - m_raopt = NULL; - - IGMP_LOCK_DESTROY(); + VNET_ASSERT(LIST_EMPTY(&V_igi_head), + ("%s: igi list %p not empty; ifnets not detached?", __func__, + &V_igi_head)); } -SYSUNINIT(igmp_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, igmp_uninit, NULL); +VNET_SYSUNINIT(vnet_igmp_uninit, SI_SUB_PROTO_MC, SI_ORDER_ANY, + vnet_igmp_uninit, NULL); +#endif -static void -vnet_igmp_init(const void *unused __unused) +#ifdef DDB +DB_SHOW_COMMAND(igi_list, db_show_igi_list) { + struct igmp_ifsoftc *igi, *tigi; + LIST_HEAD(_igi_list, igmp_ifsoftc) *igi_head; - CTR1(KTR_IGMPV3, "%s: initializing", __func__); + if (!have_addr) { + db_printf("usage: show igi_list \n"); + return; + } + igi_head = (struct _igi_list *)addr; - LIST_INIT(&V_igi_head); + LIST_FOREACH_SAFE(igi, igi_head, igi_link, tigi) { + db_printf("igmp_ifsoftc %p:\n", igi); + db_printf(" ifp %p\n", igi->igi_ifp); + db_printf(" version %u\n", igi->igi_version); + db_printf(" v1_timer %u\n", igi->igi_v1_timer); + db_printf(" v2_timer %u\n", igi->igi_v2_timer); + db_printf(" v3_timer %u\n", igi->igi_v3_timer); + db_printf(" flags %#x\n", igi->igi_flags); + db_printf(" rv %u\n", igi->igi_rv); + db_printf(" qi %u\n", igi->igi_qi); + db_printf(" qri %u\n", igi->igi_qri); + db_printf(" uri %u\n", igi->igi_uri); + /* SLIST_HEAD(,in_multi) igi_relinmhead */ + /* struct mbufq igi_gq; */ + db_printf("\n"); + } } -VNET_SYSINIT(vnet_igmp_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_igmp_init, - NULL); +#endif -static void -vnet_igmp_uninit(const void *unused __unused) -{ - - CTR1(KTR_IGMPV3, "%s: tearing down", __func__); - - KASSERT(LIST_EMPTY(&V_igi_head), - ("%s: igi list not empty; ifnets not detached?", __func__)); -} -VNET_SYSUNINIT(vnet_igmp_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, - vnet_igmp_uninit, NULL); - static int igmp_modevent(module_t mod, int type, void *unused __unused) { - switch (type) { - case MOD_LOAD: - case MOD_UNLOAD: - break; - default: - return (EOPNOTSUPP); - } - return (0); + switch (type) { + case MOD_LOAD: + CTR1(KTR_IGMPV3, "%s: initializing", __func__); + IGMP_LOCK_INIT(); + m_raopt = igmp_ra_alloc(); + netisr_register(&igmp_nh); + break; + case MOD_UNLOAD: + CTR1(KTR_IGMPV3, "%s: tearing down", __func__); + netisr_unregister(&igmp_nh); + m_free(m_raopt); + m_raopt = NULL; + IGMP_LOCK_DESTROY(); + break; + default: + return (EOPNOTSUPP); + } + return (0); } static moduledata_t igmp_mod = { "igmp", igmp_modevent, 0 }; -DECLARE_MODULE(igmp, igmp_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); +DECLARE_MODULE(igmp, igmp_mod, SI_SUB_PROTO_MC, SI_ORDER_ANY); Index: projects/vnet/sys/netinet/in_proto.c =================================================================== --- projects/vnet/sys/netinet/in_proto.c (revision 294565) +++ projects/vnet/sys/netinet/in_proto.c (revision 294566) @@ -1,396 +1,378 @@ /*- * Copyright (c) 1982, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)in_proto.c 8.2 (Berkeley) 2/9/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_mrouting.h" #include "opt_ipsec.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_sctp.h" #include "opt_mpath.h" #include #include #include #include #include #include #include #include #include /* * While this file provides the domain and protocol switch tables for IPv4, it * also provides the sysctl node declarations for net.inet.* often shared with * IPv6 for common features or by upper layer protocols. In case of no IPv4 * support compile out everything but these sysctl nodes. */ #ifdef INET #include #include #include #ifdef RADIX_MPATH #include #endif #include #endif /* INET */ #if defined(INET) || defined(INET6) #include #endif #ifdef INET #include #include #include #include #include #include #include #include #include #include #include #include /* * TCP/IP protocol family: IP, ICMP, UDP, TCP. */ static struct pr_usrreqs nousrreqs; #ifdef IPSEC #include #endif /* IPSEC */ #ifdef SCTP #include #include #include #include #endif /* SCTP */ FEATURE(inet, "Internet Protocol version 4"); extern struct domain inetdomain; /* Spacer for loadable protocols. */ #define IPPROTOSPACER \ { \ .pr_domain = &inetdomain, \ .pr_protocol = PROTO_SPACER, \ .pr_usrreqs = &nousrreqs \ } struct protosw inetsw[] = { { .pr_type = 0, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_IP, .pr_init = ip_init, -#ifdef VIMAGE - .pr_destroy = ip_destroy, -#endif .pr_slowtimo = ip_slowtimo, .pr_drain = ip_drain, .pr_usrreqs = &nousrreqs }, { .pr_type = SOCK_DGRAM, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_UDP, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = udp_input, .pr_ctlinput = udp_ctlinput, .pr_ctloutput = udp_ctloutput, .pr_init = udp_init, -#ifdef VIMAGE - .pr_destroy = udp_destroy, -#endif .pr_usrreqs = &udp_usrreqs }, { .pr_type = SOCK_STREAM, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_TCP, .pr_flags = PR_CONNREQUIRED|PR_IMPLOPCL|PR_WANTRCVD, .pr_input = tcp_input, .pr_ctlinput = tcp_ctlinput, .pr_ctloutput = tcp_ctloutput, .pr_init = tcp_init, -#ifdef VIMAGE - .pr_destroy = tcp_destroy, -#endif .pr_slowtimo = tcp_slowtimo, .pr_drain = tcp_drain, .pr_usrreqs = &tcp_usrreqs }, #ifdef SCTP { .pr_type = SOCK_SEQPACKET, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_SCTP, .pr_flags = PR_WANTRCVD, .pr_input = sctp_input, .pr_ctlinput = sctp_ctlinput, .pr_ctloutput = sctp_ctloutput, .pr_init = sctp_init, -#ifdef VIMAGE - .pr_destroy = sctp_finish, -#endif .pr_drain = sctp_drain, .pr_usrreqs = &sctp_usrreqs }, { .pr_type = SOCK_STREAM, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_SCTP, .pr_flags = PR_WANTRCVD, .pr_input = sctp_input, .pr_ctlinput = sctp_ctlinput, .pr_ctloutput = sctp_ctloutput, .pr_drain = sctp_drain, .pr_usrreqs = &sctp_usrreqs }, #endif /* SCTP */ { .pr_type = SOCK_DGRAM, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_UDPLITE, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = udp_input, .pr_ctlinput = udplite_ctlinput, .pr_ctloutput = udp_ctloutput, .pr_init = udplite_init, -#ifdef VIMAGE - .pr_destroy = udplite_destroy, -#endif .pr_usrreqs = &udp_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_RAW, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = rip_input, .pr_ctlinput = rip_ctlinput, .pr_ctloutput = rip_ctloutput, .pr_usrreqs = &rip_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_ICMP, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = icmp_input, .pr_ctloutput = rip_ctloutput, .pr_usrreqs = &rip_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_IGMP, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = igmp_input, .pr_ctloutput = rip_ctloutput, .pr_fasttimo = igmp_fasttimo, .pr_slowtimo = igmp_slowtimo, .pr_usrreqs = &rip_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_RSVP, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = rsvp_input, .pr_ctloutput = rip_ctloutput, .pr_usrreqs = &rip_usrreqs }, #ifdef IPSEC { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_AH, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = ah4_input, .pr_ctlinput = ah4_ctlinput, .pr_usrreqs = &nousrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_ESP, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = esp4_input, .pr_ctlinput = esp4_ctlinput, .pr_usrreqs = &nousrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_IPCOMP, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = ipcomp4_input, .pr_usrreqs = &nousrreqs }, #endif /* IPSEC */ { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_IPV4, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = encap4_input, .pr_ctloutput = rip_ctloutput, .pr_init = encap_init, .pr_usrreqs = &rip_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_MOBILE, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = encap4_input, .pr_ctloutput = rip_ctloutput, .pr_init = encap_init, .pr_usrreqs = &rip_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_ETHERIP, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = encap4_input, .pr_ctloutput = rip_ctloutput, .pr_init = encap_init, .pr_usrreqs = &rip_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_GRE, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = encap4_input, .pr_ctloutput = rip_ctloutput, .pr_init = encap_init, .pr_usrreqs = &rip_usrreqs }, # ifdef INET6 { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_IPV6, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = encap4_input, .pr_ctloutput = rip_ctloutput, .pr_init = encap_init, .pr_usrreqs = &rip_usrreqs }, #endif { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_PIM, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = encap4_input, .pr_ctloutput = rip_ctloutput, .pr_usrreqs = &rip_usrreqs }, /* Spacer n-times for loadable protocols. */ IPPROTOSPACER, IPPROTOSPACER, IPPROTOSPACER, IPPROTOSPACER, IPPROTOSPACER, IPPROTOSPACER, IPPROTOSPACER, IPPROTOSPACER, /* raw wildcard */ { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = rip_input, .pr_ctloutput = rip_ctloutput, .pr_init = rip_init, -#ifdef VIMAGE - .pr_destroy = rip_destroy, -#endif .pr_usrreqs = &rip_usrreqs }, }; extern int in_inithead(void **, int); extern int in_detachhead(void **, int); struct domain inetdomain = { .dom_family = AF_INET, .dom_name = "internet", .dom_protosw = inetsw, .dom_protoswNPROTOSW = &inetsw[sizeof(inetsw)/sizeof(inetsw[0])], #ifdef RADIX_MPATH .dom_rtattach = rn4_mpath_inithead, #else .dom_rtattach = in_inithead, #endif #ifdef VIMAGE .dom_rtdetach = in_detachhead, #endif .dom_ifattach = in_domifattach, .dom_ifdetach = in_domifdetach }; VNET_DOMAIN_SET(inet); #endif /* INET */ SYSCTL_NODE(_net, PF_INET, inet, CTLFLAG_RW, 0, "Internet Family"); SYSCTL_NODE(_net_inet, IPPROTO_IP, ip, CTLFLAG_RW, 0, "IP"); SYSCTL_NODE(_net_inet, IPPROTO_ICMP, icmp, CTLFLAG_RW, 0, "ICMP"); SYSCTL_NODE(_net_inet, IPPROTO_UDP, udp, CTLFLAG_RW, 0, "UDP"); SYSCTL_NODE(_net_inet, IPPROTO_TCP, tcp, CTLFLAG_RW, 0, "TCP"); #ifdef SCTP SYSCTL_NODE(_net_inet, IPPROTO_SCTP, sctp, CTLFLAG_RW, 0, "SCTP"); #endif SYSCTL_NODE(_net_inet, IPPROTO_IGMP, igmp, CTLFLAG_RW, 0, "IGMP"); #ifdef IPSEC /* XXX no protocol # to use, pick something "reserved" */ SYSCTL_NODE(_net_inet, 253, ipsec, CTLFLAG_RW, 0, "IPSEC"); SYSCTL_NODE(_net_inet, IPPROTO_AH, ah, CTLFLAG_RW, 0, "AH"); SYSCTL_NODE(_net_inet, IPPROTO_ESP, esp, CTLFLAG_RW, 0, "ESP"); SYSCTL_NODE(_net_inet, IPPROTO_IPCOMP, ipcomp, CTLFLAG_RW, 0, "IPCOMP"); SYSCTL_NODE(_net_inet, IPPROTO_IPIP, ipip, CTLFLAG_RW, 0, "IPIP"); #endif /* IPSEC */ SYSCTL_NODE(_net_inet, IPPROTO_RAW, raw, CTLFLAG_RW, 0, "RAW"); SYSCTL_NODE(_net_inet, OID_AUTO, accf, CTLFLAG_RW, 0, "Accept filters"); Index: projects/vnet/sys/netinet/ip_divert.c =================================================================== --- projects/vnet/sys/netinet/ip_divert.c (revision 294565) +++ projects/vnet/sys/netinet/ip_divert.c (revision 294566) @@ -1,840 +1,832 @@ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_sctp.h" #ifndef INET #error "IPDIVERT requires INET" #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #endif #ifdef SCTP #include #endif #include /* * Divert sockets */ /* * Allocate enough space to hold a full IP packet */ #define DIVSNDQ (65536 + 100) #define DIVRCVQ (65536 + 100) /* * Divert sockets work in conjunction with ipfw or other packet filters, * see the divert(4) manpage for features. * Packets are selected by the packet filter and tagged with an * MTAG_IPFW_RULE tag carrying the 'divert port' number (as set by * the packet filter) and information on the matching filter rule for * subsequent reinjection. The divert_port is used to put the packet * on the corresponding divert socket, while the rule number is passed * up (at least partially) as the sin_port in the struct sockaddr. * * Packets written to the divert socket carry in sin_addr a * destination address, and in sin_port the number of the filter rule * after which to continue processing. * If the destination address is INADDR_ANY, the packet is treated as * as outgoing and sent to ip_output(); otherwise it is treated as * incoming and sent to ip_input(). * Further, sin_zero carries some information on the interface, * which can be used in the reinject -- see comments in the code. * * On reinjection, processing in ip_input() and ip_output() * will be exactly the same as for the original packet, except that * packet filter processing will start at the rule number after the one * written in the sin_port (ipfw does not allow a rule #0, so sin_port=0 * will apply the entire ruleset to the packet). */ /* Internal variables. */ static VNET_DEFINE(struct inpcbhead, divcb); static VNET_DEFINE(struct inpcbinfo, divcbinfo); #define V_divcb VNET(divcb) #define V_divcbinfo VNET(divcbinfo) static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ static eventhandler_tag ip_divert_event_tag; /* * Initialize divert connection block queue. */ static void div_zone_change(void *tag) { uma_zone_set_max(V_divcbinfo.ipi_zone, maxsockets); } static int div_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "divinp"); return (0); } static void div_inpcb_fini(void *mem, int size) { struct inpcb *inp = mem; INP_LOCK_DESTROY(inp); } static void div_init(void) { /* * XXX We don't use the hash list for divert IP, but it's easier to * allocate one-entry hash lists than it is to check all over the * place for hashbase == NULL. */ in_pcbinfo_init(&V_divcbinfo, "div", &V_divcb, 1, 1, "divcb", div_inpcb_init, div_inpcb_fini, UMA_ZONE_NOFREE, IPI_HASHFIELDS_NONE); } static void div_destroy(void) { in_pcbinfo_destroy(&V_divcbinfo); } +VNET_SYSUNINIT(divert, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, + div_destroy, NULL); /* * IPPROTO_DIVERT is not in the real IP protocol number space; this * function should never be called. Just in case, drop any packets. */ static int div_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m = *mp; KMOD_IPSTAT_INC(ips_noproto); m_freem(m); return (IPPROTO_DONE); } /* * Divert a packet by passing it up to the divert socket at port 'port'. * * Setup generic address and protocol structures for div_input routine, * then pass them along with mbuf chain. */ static void divert_packet(struct mbuf *m, int incoming) { struct ip *ip; struct inpcb *inp; struct socket *sa; u_int16_t nport; struct sockaddr_in divsrc; struct m_tag *mtag; mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL); if (mtag == NULL) { m_freem(m); return; } /* Assure header */ if (m->m_len < sizeof(struct ip) && (m = m_pullup(m, sizeof(struct ip))) == 0) return; ip = mtod(m, struct ip *); /* Delayed checksums are currently not compatible with divert. */ if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { in_delayed_cksum(m); m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; } #ifdef SCTP if (m->m_pkthdr.csum_flags & CSUM_SCTP) { sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); m->m_pkthdr.csum_flags &= ~CSUM_SCTP; } #endif bzero(&divsrc, sizeof(divsrc)); divsrc.sin_len = sizeof(divsrc); divsrc.sin_family = AF_INET; /* record matching rule, in host format */ divsrc.sin_port = ((struct ipfw_rule_ref *)(mtag+1))->rulenum; /* * Record receive interface address, if any. * But only for incoming packets. */ if (incoming) { struct ifaddr *ifa; struct ifnet *ifp; /* Sanity check */ M_ASSERTPKTHDR(m); /* Find IP address for receive interface */ ifp = m->m_pkthdr.rcvif; if_addr_rlock(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET) continue; divsrc.sin_addr = ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr; break; } if_addr_runlock(ifp); } /* * Record the incoming interface name whenever we have one. */ if (m->m_pkthdr.rcvif) { /* * Hide the actual interface name in there in the * sin_zero array. XXX This needs to be moved to a * different sockaddr type for divert, e.g. * sockaddr_div with multiple fields like * sockaddr_dl. Presently we have only 7 bytes * but that will do for now as most interfaces * are 4 or less + 2 or less bytes for unit. * There is probably a faster way of doing this, * possibly taking it from the sockaddr_dl on the iface. * This solves the problem of a P2P link and a LAN interface * having the same address, which can result in the wrong * interface being assigned to the packet when fed back * into the divert socket. Theoretically if the daemon saves * and re-uses the sockaddr_in as suggested in the man pages, * this iface name will come along for the ride. * (see div_output for the other half of this.) */ strlcpy(divsrc.sin_zero, m->m_pkthdr.rcvif->if_xname, sizeof(divsrc.sin_zero)); } /* Put packet on socket queue, if any */ sa = NULL; nport = htons((u_int16_t)(((struct ipfw_rule_ref *)(mtag+1))->info)); INP_INFO_RLOCK(&V_divcbinfo); LIST_FOREACH(inp, &V_divcb, inp_list) { /* XXX why does only one socket match? */ if (inp->inp_lport == nport) { INP_RLOCK(inp); sa = inp->inp_socket; SOCKBUF_LOCK(&sa->so_rcv); if (sbappendaddr_locked(&sa->so_rcv, (struct sockaddr *)&divsrc, m, (struct mbuf *)0) == 0) { SOCKBUF_UNLOCK(&sa->so_rcv); sa = NULL; /* force mbuf reclaim below */ } else sorwakeup_locked(sa); INP_RUNLOCK(inp); break; } } INP_INFO_RUNLOCK(&V_divcbinfo); if (sa == NULL) { m_freem(m); KMOD_IPSTAT_INC(ips_noproto); KMOD_IPSTAT_DEC(ips_delivered); } } /* * Deliver packet back into the IP processing machinery. * * If no address specified, or address is 0.0.0.0, send to ip_output(); * otherwise, send to ip_input() and mark as having been received on * the interface with that address. */ static int div_output(struct socket *so, struct mbuf *m, struct sockaddr_in *sin, struct mbuf *control) { struct ip *const ip = mtod(m, struct ip *); struct m_tag *mtag; struct ipfw_rule_ref *dt; int error = 0; /* * An mbuf may hasn't come from userland, but we pretend * that it has. */ m->m_pkthdr.rcvif = NULL; m->m_nextpkt = NULL; M_SETFIB(m, so->so_fibnum); if (control) m_freem(control); /* XXX */ mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL); if (mtag == NULL) { /* this should be normal */ mtag = m_tag_alloc(MTAG_IPFW_RULE, 0, sizeof(struct ipfw_rule_ref), M_NOWAIT | M_ZERO); if (mtag == NULL) { error = ENOBUFS; goto cantsend; } m_tag_prepend(m, mtag); } dt = (struct ipfw_rule_ref *)(mtag+1); /* Loopback avoidance and state recovery */ if (sin) { int i; /* set the starting point. We provide a non-zero slot, * but a non_matching chain_id to skip that info and use * the rulenum/rule_id. */ dt->slot = 1; /* dummy, chain_id is invalid */ dt->chain_id = 0; dt->rulenum = sin->sin_port+1; /* host format ? */ dt->rule_id = 0; /* * Find receive interface with the given name, stuffed * (if it exists) in the sin_zero[] field. * The name is user supplied data so don't trust its size * or that it is zero terminated. */ for (i = 0; i < sizeof(sin->sin_zero) && sin->sin_zero[i]; i++) ; if ( i > 0 && i < sizeof(sin->sin_zero)) m->m_pkthdr.rcvif = ifunit(sin->sin_zero); } /* Reinject packet into the system as incoming or outgoing */ if (!sin || sin->sin_addr.s_addr == 0) { struct mbuf *options = NULL; struct inpcb *inp; dt->info |= IPFW_IS_DIVERT | IPFW_INFO_OUT; inp = sotoinpcb(so); INP_RLOCK(inp); switch (ip->ip_v) { case IPVERSION: /* * Don't allow both user specified and setsockopt * options, and don't allow packet length sizes that * will crash. */ if ((((ip->ip_hl << 2) != sizeof(struct ip)) && inp->inp_options != NULL) || ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) { error = EINVAL; INP_RUNLOCK(inp); goto cantsend; } break; #ifdef INET6 case IPV6_VERSION >> 4: { struct ip6_hdr *const ip6 = mtod(m, struct ip6_hdr *); /* Don't allow packet length sizes that will crash */ if (((u_short)ntohs(ip6->ip6_plen) > m->m_pkthdr.len)) { error = EINVAL; INP_RUNLOCK(inp); goto cantsend; } break; } #endif default: error = EINVAL; INP_RUNLOCK(inp); goto cantsend; } /* Send packet to output processing */ KMOD_IPSTAT_INC(ips_rawout); /* XXX */ #ifdef MAC mac_inpcb_create_mbuf(inp, m); #endif /* * Get ready to inject the packet into ip_output(). * Just in case socket options were specified on the * divert socket, we duplicate them. This is done * to avoid having to hold the PCB locks over the call * to ip_output(), as doing this results in a number of * lock ordering complexities. * * Note that we set the multicast options argument for * ip_output() to NULL since it should be invariant that * they are not present. */ KASSERT(inp->inp_moptions == NULL, ("multicast options set on a divert socket")); /* * XXXCSJP: It is unclear to me whether or not it makes * sense for divert sockets to have options. However, * for now we will duplicate them with the INP locks * held so we can use them in ip_output() without * requring a reference to the pcb. */ if (inp->inp_options != NULL) { options = m_dup(inp->inp_options, M_NOWAIT); if (options == NULL) { INP_RUNLOCK(inp); error = ENOBUFS; goto cantsend; } } INP_RUNLOCK(inp); switch (ip->ip_v) { case IPVERSION: error = ip_output(m, options, NULL, ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | IP_ALLOWBROADCAST | IP_RAWOUTPUT, NULL, NULL); break; #ifdef INET6 case IPV6_VERSION >> 4: error = ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); break; #endif } if (options != NULL) m_freem(options); } else { dt->info |= IPFW_IS_DIVERT | IPFW_INFO_IN; if (m->m_pkthdr.rcvif == NULL) { /* * No luck with the name, check by IP address. * Clear the port and the ifname to make sure * there are no distractions for ifa_ifwithaddr. */ struct ifaddr *ifa; bzero(sin->sin_zero, sizeof(sin->sin_zero)); sin->sin_port = 0; ifa = ifa_ifwithaddr((struct sockaddr *) sin); if (ifa == NULL) { error = EADDRNOTAVAIL; goto cantsend; } m->m_pkthdr.rcvif = ifa->ifa_ifp; ifa_free(ifa); } #ifdef MAC mac_socket_create_mbuf(so, m); #endif /* Send packet to input processing via netisr */ switch (ip->ip_v) { case IPVERSION: netisr_queue_src(NETISR_IP, (uintptr_t)so, m); break; #ifdef INET6 case IPV6_VERSION >> 4: netisr_queue_src(NETISR_IPV6, (uintptr_t)so, m); break; #endif default: error = EINVAL; goto cantsend; } } return (error); cantsend: m_freem(m); return (error); } static int div_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp == NULL, ("div_attach: inp != NULL")); if (td != NULL) { error = priv_check(td, PRIV_NETINET_DIVERT); if (error) return (error); } error = soreserve(so, div_sendspace, div_recvspace); if (error) return error; INP_INFO_WLOCK(&V_divcbinfo); error = in_pcballoc(so, &V_divcbinfo); if (error) { INP_INFO_WUNLOCK(&V_divcbinfo); return error; } inp = (struct inpcb *)so->so_pcb; INP_INFO_WUNLOCK(&V_divcbinfo); inp->inp_ip_p = proto; inp->inp_vflag |= INP_IPV4; inp->inp_flags |= INP_HDRINCL; INP_WUNLOCK(inp); return 0; } static void div_detach(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("div_detach: inp == NULL")); INP_INFO_WLOCK(&V_divcbinfo); INP_WLOCK(inp); in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(&V_divcbinfo); } static int div_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp != NULL, ("div_bind: inp == NULL")); /* in_pcbbind assumes that nam is a sockaddr_in * and in_pcbbind requires a valid address. Since divert * sockets don't we need to make sure the address is * filled in properly. * XXX -- divert should not be abusing in_pcbind * and should probably have its own family. */ if (nam->sa_family != AF_INET) return EAFNOSUPPORT; ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY; INP_INFO_WLOCK(&V_divcbinfo); INP_WLOCK(inp); INP_HASH_WLOCK(&V_divcbinfo); error = in_pcbbind(inp, nam, td->td_ucred); INP_HASH_WUNLOCK(&V_divcbinfo); INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_divcbinfo); return error; } static int div_shutdown(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("div_shutdown: inp == NULL")); INP_WLOCK(inp); socantsendmore(so); INP_WUNLOCK(inp); return 0; } static int div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { /* Packet must have a header (but that's about it) */ if (m->m_len < sizeof (struct ip) && (m = m_pullup(m, sizeof (struct ip))) == 0) { KMOD_IPSTAT_INC(ips_toosmall); m_freem(m); return EINVAL; } /* Send packet */ return div_output(so, m, (struct sockaddr_in *)nam, control); } static void div_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct in_addr faddr; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (PRC_IS_REDIRECT(cmd)) return; } static int div_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == 0) { n = V_divcbinfo.ipi_count; n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); return 0; } if (req->newptr != 0) return EPERM; /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&V_divcbinfo); gencnt = V_divcbinfo.ipi_gencnt; n = V_divcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_divcbinfo); error = sysctl_wire_old_buffer(req, 2 * sizeof(xig) + n*sizeof(struct xinpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return error; inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == 0) return ENOMEM; INP_INFO_RLOCK(&V_divcbinfo); for (inp = LIST_FIRST(V_divcbinfo.ipi_listhead), i = 0; inp && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt && cr_canseeinpcb(req->td->td_ucred, inp) == 0) { in_pcbref(inp); inp_list[i++] = inp; } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_divcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xinpcb xi; bzero(&xi, sizeof(xi)); xi.xi_len = sizeof xi; /* XXX should avoid extra copy */ bcopy(inp, &xi.xi_inp, sizeof *inp); if (inp->inp_socket) sotoxsocket(inp->inp_socket, &xi.xi_socket); INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xi, sizeof xi); } else INP_RUNLOCK(inp); } INP_INFO_WLOCK(&V_divcbinfo); for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (!in_pcbrele_rlocked(inp)) INP_RUNLOCK(inp); } INP_INFO_WUNLOCK(&V_divcbinfo); if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ INP_INFO_RLOCK(&V_divcbinfo); xig.xig_gen = V_divcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_divcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_divcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return error; } #ifdef SYSCTL_NODE static SYSCTL_NODE(_net_inet, IPPROTO_DIVERT, divert, CTLFLAG_RW, 0, "IPDIVERT"); SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, div_pcblist, "S,xinpcb", "List of active divert sockets"); #endif struct pr_usrreqs div_usrreqs = { .pru_attach = div_attach, .pru_bind = div_bind, .pru_control = in_control, .pru_detach = div_detach, .pru_peeraddr = in_getpeeraddr, .pru_send = div_send, .pru_shutdown = div_shutdown, .pru_sockaddr = in_getsockaddr, .pru_sosetlabel = in_pcbsosetlabel }; struct protosw div_protosw = { .pr_type = SOCK_RAW, .pr_protocol = IPPROTO_DIVERT, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = div_input, .pr_ctlinput = div_ctlinput, .pr_ctloutput = ip_ctloutput, .pr_init = div_init, -#ifdef VIMAGE - .pr_destroy = div_destroy, -#endif .pr_usrreqs = &div_usrreqs }; static int div_modevent(module_t mod, int type, void *unused) { int err = 0; -#ifndef VIMAGE int n; -#endif switch (type) { case MOD_LOAD: /* * Protocol will be initialized by pf_proto_register(). * We don't have to register ip_protox because we are not * a true IP protocol that goes over the wire. */ err = pf_proto_register(PF_INET, &div_protosw); if (err != 0) return (err); ip_divert_ptr = divert_packet; ip_divert_event_tag = EVENTHANDLER_REGISTER(maxsockets_change, div_zone_change, NULL, EVENTHANDLER_PRI_ANY); break; case MOD_QUIESCE: /* * IPDIVERT may normally not be unloaded because of the * potential race conditions. Tell kldunload we can't be * unloaded unless the unload is forced. */ err = EPERM; break; case MOD_UNLOAD: -#ifdef VIMAGE - err = EPERM; - break; -#else /* * Forced unload. * * Module ipdivert can only be unloaded if no sockets are * connected. Maybe this can be changed later to forcefully * disconnect any open sockets. * * XXXRW: Note that there is a slight race here, as a new * socket open request could be spinning on the lock and then * we destroy the lock. */ INP_INFO_WLOCK(&V_divcbinfo); n = V_divcbinfo.ipi_count; if (n != 0) { err = EBUSY; INP_INFO_WUNLOCK(&V_divcbinfo); break; } ip_divert_ptr = NULL; err = pf_proto_unregister(PF_INET, IPPROTO_DIVERT, SOCK_RAW); INP_INFO_WUNLOCK(&V_divcbinfo); div_destroy(); EVENTHANDLER_DEREGISTER(maxsockets_change, ip_divert_event_tag); break; -#endif /* !VIMAGE */ default: err = EOPNOTSUPP; break; } return err; } static moduledata_t ipdivertmod = { "ipdivert", div_modevent, 0 }; -DECLARE_MODULE(ipdivert, ipdivertmod, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY); +DECLARE_MODULE(ipdivert, ipdivertmod, SI_SUB_FW, SI_ORDER_ANY); MODULE_DEPEND(ipdivert, ipfw, 3, 3, 3); MODULE_VERSION(ipdivert, 1); Index: projects/vnet/sys/netinet/ip_mroute.c =================================================================== --- projects/vnet/sys/netinet/ip_mroute.c (revision 294565) +++ projects/vnet/sys/netinet/ip_mroute.c (revision 294566) @@ -1,2948 +1,2948 @@ /*- * Copyright (c) 1989 Stephen Deering * Copyright (c) 1992, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Stephen Deering of Stanford University. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93 */ /* * IP multicast forwarding procedures * * Written by David Waitzman, BBN Labs, August 1988. * Modified by Steve Deering, Stanford, February 1989. * Modified by Mark J. Steiglitz, Stanford, May, 1991 * Modified by Van Jacobson, LBL, January 1993 * Modified by Ajit Thyagarajan, PARC, August 1993 * Modified by Bill Fenner, PARC, April 1995 * Modified by Ahmed Helmy, SGI, June 1996 * Modified by George Edmond Eddy (Rusty), ISI, February 1998 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000 * Modified by Hitoshi Asaeda, WIDE, August 2000 * Modified by Pavlin Radoslavov, ICSI, October 2002 * * MROUTING Revision: 3.5 * and PIM-SMv2 and PIM-DM support, advanced API support, * bandwidth metering and signaling */ /* * TODO: Prefix functions with ipmf_. * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol * domain attachment (if_afdata) so we can track consumers of that service. * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT, * move it to socket options. * TODO: Cleanup LSRR removal further. * TODO: Push RSVP stubs into raw_ip.c. * TODO: Use bitstring.h for vif set. * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded. * TODO: Sync ip6_mroute.c with this file. */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_mrouting.h" #define _PIM_VT 1 #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifndef KTR_IPMF #define KTR_IPMF KTR_INET #endif #define VIFI_INVALID ((vifi_t) -1) static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */ #define V_last_tv_sec VNET(last_tv_sec) static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache"); /* * Locking. We use two locks: one for the virtual interface table and * one for the forwarding table. These locks may be nested in which case * the VIF lock must always be taken first. Note that each lock is used * to cover not only the specific data structure but also related data * structures. */ static struct mtx mrouter_mtx; #define MROUTER_LOCK() mtx_lock(&mrouter_mtx) #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx) #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED) #define MROUTER_LOCK_INIT() \ mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF) #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx) static int ip_mrouter_cnt; /* # of vnets with active mrouters */ static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */ static VNET_PCPUSTAT_DEFINE(struct mrtstat, mrtstat); VNET_PCPUSTAT_SYSINIT(mrtstat); VNET_PCPUSTAT_SYSUNINIT(mrtstat); SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat, mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, " "netinet/ip_mroute.h)"); static VNET_DEFINE(u_long, mfchash); #define V_mfchash VNET(mfchash) #define MFCHASH(a, g) \ ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \ ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash) #define MFCHASHSIZE 256 static u_long mfchashsize; /* Hash size */ static VNET_DEFINE(u_char *, nexpire); /* 0..mfchashsize-1 */ #define V_nexpire VNET(nexpire) static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl); #define V_mfchashtbl VNET(mfchashtbl) static struct mtx mfc_mtx; #define MFC_LOCK() mtx_lock(&mfc_mtx) #define MFC_UNLOCK() mtx_unlock(&mfc_mtx) #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED) #define MFC_LOCK_INIT() \ mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF) #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx) static VNET_DEFINE(vifi_t, numvifs); #define V_numvifs VNET(numvifs) static VNET_DEFINE(struct vif, viftable[MAXVIFS]); #define V_viftable VNET(viftable) SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]", "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); static struct mtx vif_mtx; #define VIF_LOCK() mtx_lock(&vif_mtx) #define VIF_UNLOCK() mtx_unlock(&vif_mtx) #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED) #define VIF_LOCK_INIT() \ mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF) #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx) static eventhandler_tag if_detach_event_tag = NULL; static VNET_DEFINE(struct callout, expire_upcalls_ch); #define V_expire_upcalls_ch VNET(expire_upcalls_ch) #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ #define UPCALL_EXPIRE 6 /* number of timeouts */ /* * Bandwidth meter variables and constants */ static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); /* * Pending timeouts are stored in a hash table, the key being the * expiration time. Periodically, the entries are analysed and processed. */ #define BW_METER_BUCKETS 1024 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]); #define V_bw_meter_timers VNET(bw_meter_timers) static VNET_DEFINE(struct callout, bw_meter_ch); #define V_bw_meter_ch VNET(bw_meter_ch) #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ /* * Pending upcalls are stored in a vector which is flushed when * full, or periodically */ static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]); #define V_bw_upcalls VNET(bw_upcalls) static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */ #define V_bw_upcalls_n VNET(bw_upcalls_n) static VNET_DEFINE(struct callout, bw_upcalls_ch); #define V_bw_upcalls_ch VNET(bw_upcalls_ch) #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ static VNET_PCPUSTAT_DEFINE(struct pimstat, pimstat); VNET_PCPUSTAT_SYSINIT(pimstat); VNET_PCPUSTAT_SYSUNINIT(pimstat); SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM"); SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat, pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)"); static u_long pim_squelch_wholepkt = 0; SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW, &pim_squelch_wholepkt, 0, "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified"); extern struct domain inetdomain; static const struct protosw in_pim_protosw = { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_PIM, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = pim_input, .pr_output = rip_output, .pr_ctloutput = rip_ctloutput, .pr_usrreqs = &rip_usrreqs }; static const struct encaptab *pim_encap_cookie; static int pim_encapcheck(const struct mbuf *, int, int, void *); /* * Note: the PIM Register encapsulation adds the following in front of a * data packet: * * struct pim_encap_hdr { * struct ip ip; * struct pim_encap_pimhdr pim; * } * */ struct pim_encap_pimhdr { struct pim pim; uint32_t flags; }; #define PIM_ENCAP_TTL 64 static struct ip pim_encap_iphdr = { #if BYTE_ORDER == LITTLE_ENDIAN sizeof(struct ip) >> 2, IPVERSION, #else IPVERSION, sizeof(struct ip) >> 2, #endif 0, /* tos */ sizeof(struct ip), /* total length */ 0, /* id */ 0, /* frag offset */ PIM_ENCAP_TTL, IPPROTO_PIM, 0, /* checksum */ }; static struct pim_encap_pimhdr pim_encap_pimhdr = { { PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 0, /* reserved */ 0, /* checksum */ }, 0 /* flags */ }; static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID; #define V_reg_vif_num VNET(reg_vif_num) static VNET_DEFINE(struct ifnet, multicast_register_if); #define V_multicast_register_if VNET(multicast_register_if) /* * Private variables. */ static u_long X_ip_mcast_src(int); static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *, struct ip_moptions *); static int X_ip_mrouter_done(void); static int X_ip_mrouter_get(struct socket *, struct sockopt *); static int X_ip_mrouter_set(struct socket *, struct sockopt *); static int X_legal_vif_num(int); static int X_mrt_ioctl(u_long, caddr_t, int); static int add_bw_upcall(struct bw_upcall *); static int add_mfc(struct mfcctl2 *); static int add_vif(struct vifctl *); static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *); static void bw_meter_process(void); static void bw_meter_receive_packet(struct bw_meter *, int, struct timeval *); static void bw_upcalls_send(void); static int del_bw_upcall(struct bw_upcall *); static int del_mfc(struct mfcctl2 *); static int del_vif(vifi_t); static int del_vif_locked(vifi_t); static void expire_bw_meter_process(void *); static void expire_bw_upcalls_send(void *); static void expire_mfc(struct mfc *); static void expire_upcalls(void *); static void free_bw_list(struct bw_meter *); static int get_sg_cnt(struct sioc_sg_req *); static int get_vif_cnt(struct sioc_vif_req *); static void if_detached_event(void *, struct ifnet *); static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); static int ip_mrouter_init(struct socket *, int); static __inline struct mfc * mfc_find(struct in_addr *, struct in_addr *); static void phyint_send(struct ip *, struct vif *, struct mbuf *); static struct mbuf * pim_register_prepare(struct ip *, struct mbuf *); static int pim_register_send(struct ip *, struct vif *, struct mbuf *, struct mfc *); static int pim_register_send_rp(struct ip *, struct vif *, struct mbuf *, struct mfc *); static int pim_register_send_upcall(struct ip *, struct vif *, struct mbuf *, struct mfc *); static void schedule_bw_meter(struct bw_meter *, struct timeval *); static void send_packet(struct vif *, struct mbuf *); static int set_api_config(uint32_t *); static int set_assert(int); static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); static void unschedule_bw_meter(struct bw_meter *); /* * Kernel multicast forwarding API capabilities and setup. * If more API capabilities are added to the kernel, they should be * recorded in `mrt_api_support'. */ #define MRT_API_VERSION 0x0305 static const int mrt_api_version = MRT_API_VERSION; static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | MRT_MFC_FLAGS_BORDER_VIF | MRT_MFC_RP | MRT_MFC_BW_UPCALL); static VNET_DEFINE(uint32_t, mrt_api_config); #define V_mrt_api_config VNET(mrt_api_config) static VNET_DEFINE(int, pim_assert_enabled); #define V_pim_assert_enabled VNET(pim_assert_enabled) static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */ /* * Find a route for a given origin IP address and multicast group address. * Statistics must be updated by the caller. */ static __inline struct mfc * mfc_find(struct in_addr *o, struct in_addr *g) { struct mfc *rt; MFC_LOCK_ASSERT(); LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) { if (in_hosteq(rt->mfc_origin, *o) && in_hosteq(rt->mfc_mcastgrp, *g) && TAILQ_EMPTY(&rt->mfc_stall)) break; } return (rt); } /* * Handle MRT setsockopt commands to modify the multicast forwarding tables. */ static int X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) { int error, optval; vifi_t vifi; struct vifctl vifc; struct mfcctl2 mfc; struct bw_upcall bw_upcall; uint32_t i; if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT) return EPERM; error = 0; switch (sopt->sopt_name) { case MRT_INIT: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; error = ip_mrouter_init(so, optval); break; case MRT_DONE: error = ip_mrouter_done(); break; case MRT_ADD_VIF: error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); if (error) break; error = add_vif(&vifc); break; case MRT_DEL_VIF: error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); if (error) break; error = del_vif(vifi); break; case MRT_ADD_MFC: case MRT_DEL_MFC: /* * select data size depending on API version. */ if (sopt->sopt_name == MRT_ADD_MFC && V_mrt_api_config & MRT_API_FLAGS_ALL) { error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), sizeof(struct mfcctl2)); } else { error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), sizeof(struct mfcctl)); bzero((caddr_t)&mfc + sizeof(struct mfcctl), sizeof(mfc) - sizeof(struct mfcctl)); } if (error) break; if (sopt->sopt_name == MRT_ADD_MFC) error = add_mfc(&mfc); else error = del_mfc(&mfc); break; case MRT_ASSERT: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; set_assert(optval); break; case MRT_API_CONFIG: error = sooptcopyin(sopt, &i, sizeof i, sizeof i); if (!error) error = set_api_config(&i); if (!error) error = sooptcopyout(sopt, &i, sizeof i); break; case MRT_ADD_BW_UPCALL: case MRT_DEL_BW_UPCALL: error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, sizeof bw_upcall); if (error) break; if (sopt->sopt_name == MRT_ADD_BW_UPCALL) error = add_bw_upcall(&bw_upcall); else error = del_bw_upcall(&bw_upcall); break; default: error = EOPNOTSUPP; break; } return error; } /* * Handle MRT getsockopt commands */ static int X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) { int error; switch (sopt->sopt_name) { case MRT_VERSION: error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version); break; case MRT_ASSERT: error = sooptcopyout(sopt, &V_pim_assert_enabled, sizeof V_pim_assert_enabled); break; case MRT_API_SUPPORT: error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); break; case MRT_API_CONFIG: error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config); break; default: error = EOPNOTSUPP; break; } return error; } /* * Handle ioctl commands to obtain information from the cache */ static int X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused) { int error = 0; /* * Currently the only function calling this ioctl routine is rtioctl(). * Typically, only root can create the raw socket in order to execute * this ioctl method, however the request might be coming from a prison */ error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error) return (error); switch (cmd) { case (SIOCGETVIFCNT): error = get_vif_cnt((struct sioc_vif_req *)data); break; case (SIOCGETSGCNT): error = get_sg_cnt((struct sioc_sg_req *)data); break; default: error = EINVAL; break; } return error; } /* * returns the packet, byte, rpf-failure count for the source group provided */ static int get_sg_cnt(struct sioc_sg_req *req) { struct mfc *rt; MFC_LOCK(); rt = mfc_find(&req->src, &req->grp); if (rt == NULL) { MFC_UNLOCK(); req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; return EADDRNOTAVAIL; } req->pktcnt = rt->mfc_pkt_cnt; req->bytecnt = rt->mfc_byte_cnt; req->wrong_if = rt->mfc_wrong_if; MFC_UNLOCK(); return 0; } /* * returns the input and output packet and byte counts on the vif provided */ static int get_vif_cnt(struct sioc_vif_req *req) { vifi_t vifi = req->vifi; VIF_LOCK(); if (vifi >= V_numvifs) { VIF_UNLOCK(); return EINVAL; } req->icount = V_viftable[vifi].v_pkt_in; req->ocount = V_viftable[vifi].v_pkt_out; req->ibytes = V_viftable[vifi].v_bytes_in; req->obytes = V_viftable[vifi].v_bytes_out; VIF_UNLOCK(); return 0; } static void if_detached_event(void *arg __unused, struct ifnet *ifp) { vifi_t vifi; u_long i; MROUTER_LOCK(); if (V_ip_mrouter == NULL) { MROUTER_UNLOCK(); return; } VIF_LOCK(); MFC_LOCK(); /* * Tear down multicast forwarder state associated with this ifnet. * 1. Walk the vif list, matching vifs against this ifnet. * 2. Walk the multicast forwarding cache (mfc) looking for * inner matches with this vif's index. * 3. Expire any matching multicast forwarding cache entries. * 4. Free vif state. This should disable ALLMULTI on the interface. */ for (vifi = 0; vifi < V_numvifs; vifi++) { if (V_viftable[vifi].v_ifp != ifp) continue; for (i = 0; i < mfchashsize; i++) { struct mfc *rt, *nrt; LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) { if (rt->mfc_parent == vifi) { expire_mfc(rt); } } } del_vif_locked(vifi); } MFC_UNLOCK(); VIF_UNLOCK(); MROUTER_UNLOCK(); } /* * Enable multicast forwarding. */ static int ip_mrouter_init(struct socket *so, int version) { CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__, so->so_type, so->so_proto->pr_protocol); if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; if (version != 1) return ENOPROTOOPT; MROUTER_LOCK(); if (ip_mrouter_unloading) { MROUTER_UNLOCK(); return ENOPROTOOPT; } if (V_ip_mrouter != NULL) { MROUTER_UNLOCK(); return EADDRINUSE; } V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash, HASH_NOWAIT); callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, curvnet); callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send, curvnet); callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, curvnet); V_ip_mrouter = so; ip_mrouter_cnt++; MROUTER_UNLOCK(); CTR1(KTR_IPMF, "%s: done", __func__); return 0; } /* * Disable multicast forwarding. */ static int X_ip_mrouter_done(void) { struct ifnet *ifp; u_long i; vifi_t vifi; MROUTER_LOCK(); if (V_ip_mrouter == NULL) { MROUTER_UNLOCK(); return EINVAL; } /* * Detach/disable hooks to the reset of the system. */ V_ip_mrouter = NULL; ip_mrouter_cnt--; V_mrt_api_config = 0; VIF_LOCK(); /* * For each phyint in use, disable promiscuous reception of all IP * multicasts. */ for (vifi = 0; vifi < V_numvifs; vifi++) { if (!in_nullhost(V_viftable[vifi].v_lcl_addr) && !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { ifp = V_viftable[vifi].v_ifp; if_allmulti(ifp, 0); } } bzero((caddr_t)V_viftable, sizeof(V_viftable)); V_numvifs = 0; V_pim_assert_enabled = 0; VIF_UNLOCK(); callout_stop(&V_expire_upcalls_ch); callout_stop(&V_bw_upcalls_ch); callout_stop(&V_bw_meter_ch); MFC_LOCK(); /* * Free all multicast forwarding cache entries. * Do not use hashdestroy(), as we must perform other cleanup. */ for (i = 0; i < mfchashsize; i++) { struct mfc *rt, *nrt; LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) { expire_mfc(rt); } } free(V_mfchashtbl, M_MRTABLE); V_mfchashtbl = NULL; bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize); V_bw_upcalls_n = 0; bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers)); MFC_UNLOCK(); V_reg_vif_num = VIFI_INVALID; MROUTER_UNLOCK(); CTR1(KTR_IPMF, "%s: done", __func__); return 0; } /* * Set PIM assert processing global */ static int set_assert(int i) { if ((i != 1) && (i != 0)) return EINVAL; V_pim_assert_enabled = i; return 0; } /* * Configure API capabilities */ int set_api_config(uint32_t *apival) { u_long i; /* * We can set the API capabilities only if it is the first operation * after MRT_INIT. I.e.: * - there are no vifs installed * - pim_assert is not enabled * - the MFC table is empty */ if (V_numvifs > 0) { *apival = 0; return EPERM; } if (V_pim_assert_enabled) { *apival = 0; return EPERM; } MFC_LOCK(); for (i = 0; i < mfchashsize; i++) { if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) { MFC_UNLOCK(); *apival = 0; return EPERM; } } MFC_UNLOCK(); V_mrt_api_config = *apival & mrt_api_support; *apival = V_mrt_api_config; return 0; } /* * Add a vif to the vif table */ static int add_vif(struct vifctl *vifcp) { struct vif *vifp = V_viftable + vifcp->vifc_vifi; struct sockaddr_in sin = {sizeof sin, AF_INET}; struct ifaddr *ifa; struct ifnet *ifp; int error; VIF_LOCK(); if (vifcp->vifc_vifi >= MAXVIFS) { VIF_UNLOCK(); return EINVAL; } /* rate limiting is no longer supported by this code */ if (vifcp->vifc_rate_limit != 0) { log(LOG_ERR, "rate limiting is no longer supported\n"); VIF_UNLOCK(); return EINVAL; } if (!in_nullhost(vifp->v_lcl_addr)) { VIF_UNLOCK(); return EADDRINUSE; } if (in_nullhost(vifcp->vifc_lcl_addr)) { VIF_UNLOCK(); return EADDRNOTAVAIL; } /* Find the interface with an address in AF_INET family */ if (vifcp->vifc_flags & VIFF_REGISTER) { /* * XXX: Because VIFF_REGISTER does not really need a valid * local interface (e.g. it could be 127.0.0.2), we don't * check its address. */ ifp = NULL; } else { sin.sin_addr = vifcp->vifc_lcl_addr; ifa = ifa_ifwithaddr((struct sockaddr *)&sin); if (ifa == NULL) { VIF_UNLOCK(); return EADDRNOTAVAIL; } ifp = ifa->ifa_ifp; ifa_free(ifa); } if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) { CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__); VIF_UNLOCK(); return EOPNOTSUPP; } else if (vifcp->vifc_flags & VIFF_REGISTER) { ifp = &V_multicast_register_if; CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp); if (V_reg_vif_num == VIFI_INVALID) { if_initname(&V_multicast_register_if, "register_vif", 0); V_multicast_register_if.if_flags = IFF_LOOPBACK; V_reg_vif_num = vifcp->vifc_vifi; } } else { /* Make sure the interface supports multicast */ if ((ifp->if_flags & IFF_MULTICAST) == 0) { VIF_UNLOCK(); return EOPNOTSUPP; } /* Enable promiscuous reception of all IP multicasts from the if */ error = if_allmulti(ifp, 1); if (error) { VIF_UNLOCK(); return error; } } vifp->v_flags = vifcp->vifc_flags; vifp->v_threshold = vifcp->vifc_threshold; vifp->v_lcl_addr = vifcp->vifc_lcl_addr; vifp->v_rmt_addr = vifcp->vifc_rmt_addr; vifp->v_ifp = ifp; /* initialize per vif pkt counters */ vifp->v_pkt_in = 0; vifp->v_pkt_out = 0; vifp->v_bytes_in = 0; vifp->v_bytes_out = 0; /* Adjust numvifs up if the vifi is higher than numvifs */ if (V_numvifs <= vifcp->vifc_vifi) V_numvifs = vifcp->vifc_vifi + 1; VIF_UNLOCK(); CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__, (int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr), (int)vifcp->vifc_threshold); return 0; } /* * Delete a vif from the vif table */ static int del_vif_locked(vifi_t vifi) { struct vif *vifp; VIF_LOCK_ASSERT(); if (vifi >= V_numvifs) { return EINVAL; } vifp = &V_viftable[vifi]; if (in_nullhost(vifp->v_lcl_addr)) { return EADDRNOTAVAIL; } if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) if_allmulti(vifp->v_ifp, 0); if (vifp->v_flags & VIFF_REGISTER) V_reg_vif_num = VIFI_INVALID; bzero((caddr_t)vifp, sizeof (*vifp)); CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi); /* Adjust numvifs down */ for (vifi = V_numvifs; vifi > 0; vifi--) if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr)) break; V_numvifs = vifi; return 0; } static int del_vif(vifi_t vifi) { int cc; VIF_LOCK(); cc = del_vif_locked(vifi); VIF_UNLOCK(); return cc; } /* * update an mfc entry without resetting counters and S,G addresses. */ static void update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) { int i; rt->mfc_parent = mfccp->mfcc_parent; for (i = 0; i < V_numvifs; i++) { rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config & MRT_MFC_FLAGS_ALL; } /* set the RP address */ if (V_mrt_api_config & MRT_MFC_RP) rt->mfc_rp = mfccp->mfcc_rp; else rt->mfc_rp.s_addr = INADDR_ANY; } /* * fully initialize an mfc entry from the parameter. */ static void init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) { rt->mfc_origin = mfccp->mfcc_origin; rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; update_mfc_params(rt, mfccp); /* initialize pkt counters per src-grp */ rt->mfc_pkt_cnt = 0; rt->mfc_byte_cnt = 0; rt->mfc_wrong_if = 0; timevalclear(&rt->mfc_last_assert); } static void expire_mfc(struct mfc *rt) { struct rtdetq *rte, *nrte; MFC_LOCK_ASSERT(); free_bw_list(rt->mfc_bw_meter); TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) { m_freem(rte->m); TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link); free(rte, M_MRTABLE); } LIST_REMOVE(rt, mfc_hash); free(rt, M_MRTABLE); } /* * Add an mfc entry */ static int add_mfc(struct mfcctl2 *mfccp) { struct mfc *rt; struct rtdetq *rte, *nrte; u_long hash = 0; u_short nstl; VIF_LOCK(); MFC_LOCK(); rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); /* If an entry already exists, just update the fields */ if (rt) { CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x", __func__, inet_ntoa(mfccp->mfcc_origin), (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), mfccp->mfcc_parent); update_mfc_params(rt, mfccp); MFC_UNLOCK(); VIF_UNLOCK(); return (0); } /* * Find the entry for which the upcall was made and update */ nstl = 0; hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp); LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) { if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) && !TAILQ_EMPTY(&rt->mfc_stall)) { CTR5(KTR_IPMF, "%s: add mfc orig %s group %lx parent %x qh %p", __func__, inet_ntoa(mfccp->mfcc_origin), (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), mfccp->mfcc_parent, TAILQ_FIRST(&rt->mfc_stall)); if (nstl++) CTR1(KTR_IPMF, "%s: multiple matches", __func__); init_mfc_params(rt, mfccp); rt->mfc_expire = 0; /* Don't clean this guy up */ V_nexpire[hash]--; /* Free queued packets, but attempt to forward them first. */ TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) { if (rte->ifp != NULL) ip_mdq(rte->m, rte->ifp, rt, -1); m_freem(rte->m); TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link); rt->mfc_nstall--; free(rte, M_MRTABLE); } } } /* * It is possible that an entry is being inserted without an upcall */ if (nstl == 0) { CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__); LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) { if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) { init_mfc_params(rt, mfccp); if (rt->mfc_expire) V_nexpire[hash]--; rt->mfc_expire = 0; break; /* XXX */ } } if (rt == NULL) { /* no upcall, so make a new entry */ rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); if (rt == NULL) { MFC_UNLOCK(); VIF_UNLOCK(); return (ENOBUFS); } init_mfc_params(rt, mfccp); TAILQ_INIT(&rt->mfc_stall); rt->mfc_nstall = 0; rt->mfc_expire = 0; rt->mfc_bw_meter = NULL; /* insert new entry at head of hash chain */ LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash); } } MFC_UNLOCK(); VIF_UNLOCK(); return (0); } /* * Delete an mfc entry */ static int del_mfc(struct mfcctl2 *mfccp) { struct in_addr origin; struct in_addr mcastgrp; struct mfc *rt; origin = mfccp->mfcc_origin; mcastgrp = mfccp->mfcc_mcastgrp; CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__, inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr)); MFC_LOCK(); rt = mfc_find(&origin, &mcastgrp); if (rt == NULL) { MFC_UNLOCK(); return EADDRNOTAVAIL; } /* * free the bw_meter entries */ free_bw_list(rt->mfc_bw_meter); rt->mfc_bw_meter = NULL; LIST_REMOVE(rt, mfc_hash); free(rt, M_MRTABLE); MFC_UNLOCK(); return (0); } /* * Send a message to the routing daemon on the multicast routing socket. */ static int socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) { if (s) { SOCKBUF_LOCK(&s->so_rcv); if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm, NULL) != 0) { sorwakeup_locked(s); return 0; } SOCKBUF_UNLOCK(&s->so_rcv); } m_freem(mm); return -1; } /* * IP multicast forwarding function. This function assumes that the packet * pointed to by "ip" has arrived on (or is about to be sent to) the interface * pointed to by "ifp", and the packet is to be relayed to other networks * that have members of the packet's destination IP multicast group. * * The packet is returned unscathed to the caller, unless it is * erroneous, in which case a non-zero return value tells the caller to * discard it. */ #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo) { struct mfc *rt; int error; vifi_t vifi; CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p", inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp); if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { /* * Packet arrived via a physical interface or * an encapsulated tunnel or a register_vif. */ } else { /* * Packet arrived through a source-route tunnel. * Source-route tunnels are no longer supported. */ return (1); } VIF_LOCK(); MFC_LOCK(); if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) { if (ip->ip_ttl < MAXTTL) ip->ip_ttl++; /* compensate for -1 in *_send routines */ error = ip_mdq(m, ifp, NULL, vifi); MFC_UNLOCK(); VIF_UNLOCK(); return error; } /* * Don't forward a packet with time-to-live of zero or one, * or a packet destined to a local-only group. */ if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) { MFC_UNLOCK(); VIF_UNLOCK(); return 0; } /* * Determine forwarding vifs from the forwarding cache table */ MRTSTAT_INC(mrts_mfc_lookups); rt = mfc_find(&ip->ip_src, &ip->ip_dst); /* Entry exists, so forward if necessary */ if (rt != NULL) { error = ip_mdq(m, ifp, rt, -1); MFC_UNLOCK(); VIF_UNLOCK(); return error; } else { /* * If we don't have a route for packet's origin, * Make a copy of the packet & send message to routing daemon */ struct mbuf *mb0; struct rtdetq *rte; u_long hash; int hlen = ip->ip_hl << 2; MRTSTAT_INC(mrts_mfc_misses); MRTSTAT_INC(mrts_no_route); CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)", inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr)); /* * Allocate mbufs early so that we don't do extra work if we are * just going to fail anyway. Make sure to pullup the header so * that other people can't step on it. */ rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT|M_ZERO); if (rte == NULL) { MFC_UNLOCK(); VIF_UNLOCK(); return ENOBUFS; } mb0 = m_copypacket(m, M_NOWAIT); if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen)) mb0 = m_pullup(mb0, hlen); if (mb0 == NULL) { free(rte, M_MRTABLE); MFC_UNLOCK(); VIF_UNLOCK(); return ENOBUFS; } /* is there an upcall waiting for this flow ? */ hash = MFCHASH(ip->ip_src, ip->ip_dst); LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) { if (in_hosteq(ip->ip_src, rt->mfc_origin) && in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) && !TAILQ_EMPTY(&rt->mfc_stall)) break; } if (rt == NULL) { int i; struct igmpmsg *im; struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; struct mbuf *mm; /* * Locate the vifi for the incoming interface for this packet. * If none found, drop packet. */ for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp; vifi++) ; if (vifi >= V_numvifs) /* vif not found, drop packet */ goto non_fatal; /* no upcall, so make a new entry */ rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); if (rt == NULL) goto fail; /* Make a copy of the header to send to the user level process */ mm = m_copy(mb0, 0, hlen); if (mm == NULL) goto fail1; /* * Send message to routing daemon to install * a route into the kernel table */ im = mtod(mm, struct igmpmsg *); im->im_msgtype = IGMPMSG_NOCACHE; im->im_mbz = 0; im->im_vif = vifi; MRTSTAT_INC(mrts_upcalls); k_igmpsrc.sin_addr = ip->ip_src; if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) { CTR0(KTR_IPMF, "ip_mforward: socket queue full"); MRTSTAT_INC(mrts_upq_sockfull); fail1: free(rt, M_MRTABLE); fail: free(rte, M_MRTABLE); m_freem(mb0); MFC_UNLOCK(); VIF_UNLOCK(); return ENOBUFS; } /* insert new entry at head of hash chain */ rt->mfc_origin.s_addr = ip->ip_src.s_addr; rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; rt->mfc_expire = UPCALL_EXPIRE; V_nexpire[hash]++; for (i = 0; i < V_numvifs; i++) { rt->mfc_ttls[i] = 0; rt->mfc_flags[i] = 0; } rt->mfc_parent = -1; /* clear the RP address */ rt->mfc_rp.s_addr = INADDR_ANY; rt->mfc_bw_meter = NULL; /* initialize pkt counters per src-grp */ rt->mfc_pkt_cnt = 0; rt->mfc_byte_cnt = 0; rt->mfc_wrong_if = 0; timevalclear(&rt->mfc_last_assert); TAILQ_INIT(&rt->mfc_stall); rt->mfc_nstall = 0; /* link into table */ LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash); TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link); rt->mfc_nstall++; } else { /* determine if queue has overflowed */ if (rt->mfc_nstall > MAX_UPQ) { MRTSTAT_INC(mrts_upq_ovflw); non_fatal: free(rte, M_MRTABLE); m_freem(mb0); MFC_UNLOCK(); VIF_UNLOCK(); return (0); } TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link); rt->mfc_nstall++; } rte->m = mb0; rte->ifp = ifp; MFC_UNLOCK(); VIF_UNLOCK(); return 0; } } /* * Clean up the cache entry if upcall is not serviced */ static void expire_upcalls(void *arg) { u_long i; CURVNET_SET((struct vnet *) arg); MFC_LOCK(); for (i = 0; i < mfchashsize; i++) { struct mfc *rt, *nrt; if (V_nexpire[i] == 0) continue; LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) { if (TAILQ_EMPTY(&rt->mfc_stall)) continue; if (rt->mfc_expire == 0 || --rt->mfc_expire > 0) continue; /* * free the bw_meter entries */ while (rt->mfc_bw_meter != NULL) { struct bw_meter *x = rt->mfc_bw_meter; rt->mfc_bw_meter = x->bm_mfc_next; free(x, M_BWMETER); } MRTSTAT_INC(mrts_cache_cleanups); CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__, (u_long)ntohl(rt->mfc_origin.s_addr), (u_long)ntohl(rt->mfc_mcastgrp.s_addr)); expire_mfc(rt); } } MFC_UNLOCK(); callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, curvnet); CURVNET_RESTORE(); } /* * Packet forwarding routine once entry in the cache is made */ static int ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) { struct ip *ip = mtod(m, struct ip *); vifi_t vifi; int plen = ntohs(ip->ip_len); VIF_LOCK_ASSERT(); /* * If xmt_vif is not -1, send on only the requested vif. * * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) */ if (xmt_vif < V_numvifs) { if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER) pim_register_send(ip, V_viftable + xmt_vif, m, rt); else phyint_send(ip, V_viftable + xmt_vif, m); return 1; } /* * Don't forward if it didn't arrive from the parent vif for its origin. */ vifi = rt->mfc_parent; if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) { CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)", __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp); MRTSTAT_INC(mrts_wrong_if); ++rt->mfc_wrong_if; /* * If we are doing PIM assert processing, send a message * to the routing daemon. * * XXX: A PIM-SM router needs the WRONGVIF detection so it * can complete the SPT switch, regardless of the type * of the iif (broadcast media, GRE tunnel, etc). */ if (V_pim_assert_enabled && (vifi < V_numvifs) && V_viftable[vifi].v_ifp) { if (ifp == &V_multicast_register_if) PIMSTAT_INC(pims_rcv_registers_wrongiif); /* Get vifi for the incoming packet */ for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp; vifi++) ; if (vifi >= V_numvifs) return 0; /* The iif is not found: ignore the packet. */ if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) return 0; /* WRONGVIF disabled: ignore the packet */ if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) { struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; struct igmpmsg *im; int hlen = ip->ip_hl << 2; struct mbuf *mm = m_copy(m, 0, hlen); if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen)) mm = m_pullup(mm, hlen); if (mm == NULL) return ENOBUFS; im = mtod(mm, struct igmpmsg *); im->im_msgtype = IGMPMSG_WRONGVIF; im->im_mbz = 0; im->im_vif = vifi; MRTSTAT_INC(mrts_upcalls); k_igmpsrc.sin_addr = im->im_src; if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) { CTR1(KTR_IPMF, "%s: socket queue full", __func__); MRTSTAT_INC(mrts_upq_sockfull); return ENOBUFS; } } } return 0; } /* If I sourced this packet, it counts as output, else it was input. */ if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) { V_viftable[vifi].v_pkt_out++; V_viftable[vifi].v_bytes_out += plen; } else { V_viftable[vifi].v_pkt_in++; V_viftable[vifi].v_bytes_in += plen; } rt->mfc_pkt_cnt++; rt->mfc_byte_cnt += plen; /* * For each vif, decide if a copy of the packet should be forwarded. * Forward if: * - the ttl exceeds the vif's threshold * - there are group members downstream on interface */ for (vifi = 0; vifi < V_numvifs; vifi++) if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { V_viftable[vifi].v_pkt_out++; V_viftable[vifi].v_bytes_out += plen; if (V_viftable[vifi].v_flags & VIFF_REGISTER) pim_register_send(ip, V_viftable + vifi, m, rt); else phyint_send(ip, V_viftable + vifi, m); } /* * Perform upcall-related bw measuring. */ if (rt->mfc_bw_meter != NULL) { struct bw_meter *x; struct timeval now; microtime(&now); MFC_LOCK_ASSERT(); for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) bw_meter_receive_packet(x, plen, &now); } return 0; } /* * Check if a vif number is legal/ok. This is used by in_mcast.c. */ static int X_legal_vif_num(int vif) { int ret; ret = 0; if (vif < 0) return (ret); VIF_LOCK(); if (vif < V_numvifs) ret = 1; VIF_UNLOCK(); return (ret); } /* * Return the local address used by this vif */ static u_long X_ip_mcast_src(int vifi) { in_addr_t addr; addr = INADDR_ANY; if (vifi < 0) return (addr); VIF_LOCK(); if (vifi < V_numvifs) addr = V_viftable[vifi].v_lcl_addr.s_addr; VIF_UNLOCK(); return (addr); } static void phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) { struct mbuf *mb_copy; int hlen = ip->ip_hl << 2; VIF_LOCK_ASSERT(); /* * Make a new reference to the packet; make sure that * the IP header is actually copied, not just referenced, * so that ip_output() only scribbles on the copy. */ mb_copy = m_copypacket(m, M_NOWAIT); if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen)) mb_copy = m_pullup(mb_copy, hlen); if (mb_copy == NULL) return; send_packet(vifp, mb_copy); } static void send_packet(struct vif *vifp, struct mbuf *m) { struct ip_moptions imo; struct in_multi *imm[2]; int error; VIF_LOCK_ASSERT(); imo.imo_multicast_ifp = vifp->v_ifp; imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; imo.imo_multicast_loop = 1; imo.imo_multicast_vif = -1; imo.imo_num_memberships = 0; imo.imo_max_memberships = 2; imo.imo_membership = &imm[0]; /* * Re-entrancy should not be a problem here, because * the packets that we send out and are looped back at us * should get rejected because they appear to come from * the loopback interface, thus preventing looping. */ error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL); CTR3(KTR_IPMF, "%s: vif %td err %d", __func__, (ptrdiff_t)(vifp - V_viftable), error); } /* * Stubs for old RSVP socket shim implementation. */ static int X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused) { return (EOPNOTSUPP); } static void X_ip_rsvp_force_done(struct socket *so __unused) { } static int X_rsvp_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m; m = *mp; *mp = NULL; if (!V_rsvp_on) m_freem(m); return (IPPROTO_DONE); } /* * Code for bandwidth monitors */ /* * Define common interface for timeval-related methods */ #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) static uint32_t compute_bw_meter_flags(struct bw_upcall *req) { uint32_t flags = 0; if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) flags |= BW_METER_UNIT_PACKETS; if (req->bu_flags & BW_UPCALL_UNIT_BYTES) flags |= BW_METER_UNIT_BYTES; if (req->bu_flags & BW_UPCALL_GEQ) flags |= BW_METER_GEQ; if (req->bu_flags & BW_UPCALL_LEQ) flags |= BW_METER_LEQ; return flags; } /* * Add a bw_meter entry */ static int add_bw_upcall(struct bw_upcall *req) { struct mfc *mfc; struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; struct timeval now; struct bw_meter *x; uint32_t flags; if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL)) return EOPNOTSUPP; /* Test if the flags are valid */ if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) return EINVAL; if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) return EINVAL; if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) return EINVAL; /* Test if the threshold time interval is valid */ if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) return EINVAL; flags = compute_bw_meter_flags(req); /* * Find if we have already same bw_meter entry */ MFC_LOCK(); mfc = mfc_find(&req->bu_src, &req->bu_dst); if (mfc == NULL) { MFC_UNLOCK(); return EADDRNOTAVAIL; } for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, &req->bu_threshold.b_time, ==)) && (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && (x->bm_flags & BW_METER_USER_FLAGS) == flags) { MFC_UNLOCK(); return 0; /* XXX Already installed */ } } /* Allocate the new bw_meter entry */ x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); if (x == NULL) { MFC_UNLOCK(); return ENOBUFS; } /* Set the new bw_meter entry */ x->bm_threshold.b_time = req->bu_threshold.b_time; microtime(&now); x->bm_start_time = now; x->bm_threshold.b_packets = req->bu_threshold.b_packets; x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; x->bm_measured.b_packets = 0; x->bm_measured.b_bytes = 0; x->bm_flags = flags; x->bm_time_next = NULL; x->bm_time_hash = BW_METER_BUCKETS; /* Add the new bw_meter entry to the front of entries for this MFC */ x->bm_mfc = mfc; x->bm_mfc_next = mfc->mfc_bw_meter; mfc->mfc_bw_meter = x; schedule_bw_meter(x, &now); MFC_UNLOCK(); return 0; } static void free_bw_list(struct bw_meter *list) { while (list != NULL) { struct bw_meter *x = list; list = list->bm_mfc_next; unschedule_bw_meter(x); free(x, M_BWMETER); } } /* * Delete one or multiple bw_meter entries */ static int del_bw_upcall(struct bw_upcall *req) { struct mfc *mfc; struct bw_meter *x; if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL)) return EOPNOTSUPP; MFC_LOCK(); /* Find the corresponding MFC entry */ mfc = mfc_find(&req->bu_src, &req->bu_dst); if (mfc == NULL) { MFC_UNLOCK(); return EADDRNOTAVAIL; } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { /* * Delete all bw_meter entries for this mfc */ struct bw_meter *list; list = mfc->mfc_bw_meter; mfc->mfc_bw_meter = NULL; free_bw_list(list); MFC_UNLOCK(); return 0; } else { /* Delete a single bw_meter entry */ struct bw_meter *prev; uint32_t flags = 0; flags = compute_bw_meter_flags(req); /* Find the bw_meter entry to delete */ for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; prev = x, x = x->bm_mfc_next) { if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, &req->bu_threshold.b_time, ==)) && (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && (x->bm_flags & BW_METER_USER_FLAGS) == flags) break; } if (x != NULL) { /* Delete entry from the list for this MFC */ if (prev != NULL) prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ else x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ unschedule_bw_meter(x); MFC_UNLOCK(); /* Free the bw_meter entry */ free(x, M_BWMETER); return 0; } else { MFC_UNLOCK(); return EINVAL; } } /* NOTREACHED */ } /* * Perform bandwidth measurement processing that may result in an upcall */ static void bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) { struct timeval delta; MFC_LOCK_ASSERT(); delta = *nowp; BW_TIMEVALDECR(&delta, &x->bm_start_time); if (x->bm_flags & BW_METER_GEQ) { /* * Processing for ">=" type of bw_meter entry */ if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { /* Reset the bw_meter entry */ x->bm_start_time = *nowp; x->bm_measured.b_packets = 0; x->bm_measured.b_bytes = 0; x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; } /* Record that a packet is received */ x->bm_measured.b_packets++; x->bm_measured.b_bytes += plen; /* * Test if we should deliver an upcall */ if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { if (((x->bm_flags & BW_METER_UNIT_PACKETS) && (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || ((x->bm_flags & BW_METER_UNIT_BYTES) && (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { /* Prepare an upcall for delivery */ bw_meter_prepare_upcall(x, nowp); x->bm_flags |= BW_METER_UPCALL_DELIVERED; } } } else if (x->bm_flags & BW_METER_LEQ) { /* * Processing for "<=" type of bw_meter entry */ if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { /* * We are behind time with the multicast forwarding table * scanning for "<=" type of bw_meter entries, so test now * if we should deliver an upcall. */ if (((x->bm_flags & BW_METER_UNIT_PACKETS) && (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || ((x->bm_flags & BW_METER_UNIT_BYTES) && (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { /* Prepare an upcall for delivery */ bw_meter_prepare_upcall(x, nowp); } /* Reschedule the bw_meter entry */ unschedule_bw_meter(x); schedule_bw_meter(x, nowp); } /* Record that a packet is received */ x->bm_measured.b_packets++; x->bm_measured.b_bytes += plen; /* * Test if we should restart the measuring interval */ if ((x->bm_flags & BW_METER_UNIT_PACKETS && x->bm_measured.b_packets <= x->bm_threshold.b_packets) || (x->bm_flags & BW_METER_UNIT_BYTES && x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { /* Don't restart the measuring interval */ } else { /* Do restart the measuring interval */ /* * XXX: note that we don't unschedule and schedule, because this * might be too much overhead per packet. Instead, when we process * all entries for a given timer hash bin, we check whether it is * really a timeout. If not, we reschedule at that time. */ x->bm_start_time = *nowp; x->bm_measured.b_packets = 0; x->bm_measured.b_bytes = 0; x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; } } } /* * Prepare a bandwidth-related upcall */ static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) { struct timeval delta; struct bw_upcall *u; MFC_LOCK_ASSERT(); /* * Compute the measured time interval */ delta = *nowp; BW_TIMEVALDECR(&delta, &x->bm_start_time); /* * If there are too many pending upcalls, deliver them now */ if (V_bw_upcalls_n >= BW_UPCALLS_MAX) bw_upcalls_send(); /* * Set the bw_upcall entry */ u = &V_bw_upcalls[V_bw_upcalls_n++]; u->bu_src = x->bm_mfc->mfc_origin; u->bu_dst = x->bm_mfc->mfc_mcastgrp; u->bu_threshold.b_time = x->bm_threshold.b_time; u->bu_threshold.b_packets = x->bm_threshold.b_packets; u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; u->bu_measured.b_time = delta; u->bu_measured.b_packets = x->bm_measured.b_packets; u->bu_measured.b_bytes = x->bm_measured.b_bytes; u->bu_flags = 0; if (x->bm_flags & BW_METER_UNIT_PACKETS) u->bu_flags |= BW_UPCALL_UNIT_PACKETS; if (x->bm_flags & BW_METER_UNIT_BYTES) u->bu_flags |= BW_UPCALL_UNIT_BYTES; if (x->bm_flags & BW_METER_GEQ) u->bu_flags |= BW_UPCALL_GEQ; if (x->bm_flags & BW_METER_LEQ) u->bu_flags |= BW_UPCALL_LEQ; } /* * Send the pending bandwidth-related upcalls */ static void bw_upcalls_send(void) { struct mbuf *m; int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]); struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; static struct igmpmsg igmpmsg = { 0, /* unused1 */ 0, /* unused2 */ IGMPMSG_BW_UPCALL,/* im_msgtype */ 0, /* im_mbz */ 0, /* im_vif */ 0, /* unused3 */ { 0 }, /* im_src */ { 0 } }; /* im_dst */ MFC_LOCK_ASSERT(); if (V_bw_upcalls_n == 0) return; /* No pending upcalls */ V_bw_upcalls_n = 0; /* * Allocate a new mbuf, initialize it with the header and * the payload for the pending calls. */ m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) { log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); return; } m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]); /* * Send the upcalls * XXX do we need to set the address in k_igmpsrc ? */ MRTSTAT_INC(mrts_upcalls); if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) { log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); MRTSTAT_INC(mrts_upq_sockfull); } } /* * Compute the timeout hash value for the bw_meter entries */ #define BW_METER_TIMEHASH(bw_meter, hash) \ do { \ struct timeval next_timeval = (bw_meter)->bm_start_time; \ \ BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ (hash) = next_timeval.tv_sec; \ if (next_timeval.tv_usec) \ (hash)++; /* XXX: make sure we don't timeout early */ \ (hash) %= BW_METER_BUCKETS; \ } while (0) /* * Schedule a timer to process periodically bw_meter entry of type "<=" * by linking the entry in the proper hash bucket. */ static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) { int time_hash; MFC_LOCK_ASSERT(); if (!(x->bm_flags & BW_METER_LEQ)) return; /* XXX: we schedule timers only for "<=" entries */ /* * Reset the bw_meter entry */ x->bm_start_time = *nowp; x->bm_measured.b_packets = 0; x->bm_measured.b_bytes = 0; x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; /* * Compute the timeout hash value and insert the entry */ BW_METER_TIMEHASH(x, time_hash); x->bm_time_next = V_bw_meter_timers[time_hash]; V_bw_meter_timers[time_hash] = x; x->bm_time_hash = time_hash; } /* * Unschedule the periodic timer that processes bw_meter entry of type "<=" * by removing the entry from the proper hash bucket. */ static void unschedule_bw_meter(struct bw_meter *x) { int time_hash; struct bw_meter *prev, *tmp; MFC_LOCK_ASSERT(); if (!(x->bm_flags & BW_METER_LEQ)) return; /* XXX: we schedule timers only for "<=" entries */ /* * Compute the timeout hash value and delete the entry */ time_hash = x->bm_time_hash; if (time_hash >= BW_METER_BUCKETS) return; /* Entry was not scheduled */ for (prev = NULL, tmp = V_bw_meter_timers[time_hash]; tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) if (tmp == x) break; if (tmp == NULL) panic("unschedule_bw_meter: bw_meter entry not found"); if (prev != NULL) prev->bm_time_next = x->bm_time_next; else V_bw_meter_timers[time_hash] = x->bm_time_next; x->bm_time_next = NULL; x->bm_time_hash = BW_METER_BUCKETS; } /* * Process all "<=" type of bw_meter that should be processed now, * and for each entry prepare an upcall if necessary. Each processed * entry is rescheduled again for the (periodic) processing. * * This is run periodically (once per second normally). On each round, * all the potentially matching entries are in the hash slot that we are * looking at. */ static void bw_meter_process() { uint32_t loops; int i; struct timeval now, process_endtime; microtime(&now); if (V_last_tv_sec == now.tv_sec) return; /* nothing to do */ loops = now.tv_sec - V_last_tv_sec; V_last_tv_sec = now.tv_sec; if (loops > BW_METER_BUCKETS) loops = BW_METER_BUCKETS; MFC_LOCK(); /* * Process all bins of bw_meter entries from the one after the last * processed to the current one. On entry, i points to the last bucket * visited, so we need to increment i at the beginning of the loop. */ for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { struct bw_meter *x, *tmp_list; if (++i >= BW_METER_BUCKETS) i = 0; /* Disconnect the list of bw_meter entries from the bin */ tmp_list = V_bw_meter_timers[i]; V_bw_meter_timers[i] = NULL; /* Process the list of bw_meter entries */ while (tmp_list != NULL) { x = tmp_list; tmp_list = tmp_list->bm_time_next; /* Test if the time interval is over */ process_endtime = x->bm_start_time; BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); if (BW_TIMEVALCMP(&process_endtime, &now, >)) { /* Not yet: reschedule, but don't reset */ int time_hash; BW_METER_TIMEHASH(x, time_hash); if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { /* * XXX: somehow the bin processing is a bit ahead of time. * Put the entry in the next bin. */ if (++time_hash >= BW_METER_BUCKETS) time_hash = 0; } x->bm_time_next = V_bw_meter_timers[time_hash]; V_bw_meter_timers[time_hash] = x; x->bm_time_hash = time_hash; continue; } /* * Test if we should deliver an upcall */ if (((x->bm_flags & BW_METER_UNIT_PACKETS) && (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || ((x->bm_flags & BW_METER_UNIT_BYTES) && (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { /* Prepare an upcall for delivery */ bw_meter_prepare_upcall(x, &now); } /* * Reschedule for next processing */ schedule_bw_meter(x, &now); } } /* Send all upcalls that are pending delivery */ bw_upcalls_send(); MFC_UNLOCK(); } /* * A periodic function for sending all upcalls that are pending delivery */ static void expire_bw_upcalls_send(void *arg) { CURVNET_SET((struct vnet *) arg); MFC_LOCK(); bw_upcalls_send(); MFC_UNLOCK(); callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send, curvnet); CURVNET_RESTORE(); } /* * A periodic function for periodic scanning of the multicast forwarding * table for processing all "<=" bw_meter entries. */ static void expire_bw_meter_process(void *arg) { CURVNET_SET((struct vnet *) arg); if (V_mrt_api_config & MRT_MFC_BW_UPCALL) bw_meter_process(); callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, curvnet); CURVNET_RESTORE(); } /* * End of bandwidth monitoring code */ /* * Send the packet up to the user daemon, or eventually do kernel encapsulation * */ static int pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m, struct mfc *rt) { struct mbuf *mb_copy, *mm; /* * Do not send IGMP_WHOLEPKT notifications to userland, if the * rendezvous point was unspecified, and we were told not to. */ if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) && in_nullhost(rt->mfc_rp)) return 0; mb_copy = pim_register_prepare(ip, m); if (mb_copy == NULL) return ENOBUFS; /* * Send all the fragments. Note that the mbuf for each fragment * is freed by the sending machinery. */ for (mm = mb_copy; mm; mm = mb_copy) { mb_copy = mm->m_nextpkt; mm->m_nextpkt = 0; mm = m_pullup(mm, sizeof(struct ip)); if (mm != NULL) { ip = mtod(mm, struct ip *); if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) { pim_register_send_rp(ip, vifp, mm, rt); } else { pim_register_send_upcall(ip, vifp, mm, rt); } } } return 0; } /* * Return a copy of the data packet that is ready for PIM Register * encapsulation. * XXX: Note that in the returned copy the IP header is a valid one. */ static struct mbuf * pim_register_prepare(struct ip *ip, struct mbuf *m) { struct mbuf *mb_copy = NULL; int mtu; /* Take care of delayed checksums */ if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { in_delayed_cksum(m); m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; } /* * Copy the old packet & pullup its IP header into the * new mbuf so we can modify it. */ mb_copy = m_copypacket(m, M_NOWAIT); if (mb_copy == NULL) return NULL; mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); if (mb_copy == NULL) return NULL; /* take care of the TTL */ ip = mtod(mb_copy, struct ip *); --ip->ip_ttl; /* Compute the MTU after the PIM Register encapsulation */ mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); if (ntohs(ip->ip_len) <= mtu) { /* Turn the IP header into a valid one */ ip->ip_sum = 0; ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); } else { /* Fragment the packet */ mb_copy->m_pkthdr.csum_flags |= CSUM_IP; if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) { m_freem(mb_copy); return NULL; } } return mb_copy; } /* * Send an upcall with the data packet to the user-level process. */ static int pim_register_send_upcall(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy, struct mfc *rt) { struct mbuf *mb_first; int len = ntohs(ip->ip_len); struct igmpmsg *im; struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; VIF_LOCK_ASSERT(); /* * Add a new mbuf with an upcall header */ mb_first = m_gethdr(M_NOWAIT, MT_DATA); if (mb_first == NULL) { m_freem(mb_copy); return ENOBUFS; } mb_first->m_data += max_linkhdr; mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); mb_first->m_len = sizeof(struct igmpmsg); mb_first->m_next = mb_copy; /* Send message to routing daemon */ im = mtod(mb_first, struct igmpmsg *); im->im_msgtype = IGMPMSG_WHOLEPKT; im->im_mbz = 0; im->im_vif = vifp - V_viftable; im->im_src = ip->ip_src; im->im_dst = ip->ip_dst; k_igmpsrc.sin_addr = ip->ip_src; MRTSTAT_INC(mrts_upcalls); if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) { CTR1(KTR_IPMF, "%s: socket queue full", __func__); MRTSTAT_INC(mrts_upq_sockfull); return ENOBUFS; } /* Keep statistics */ PIMSTAT_INC(pims_snd_registers_msgs); PIMSTAT_ADD(pims_snd_registers_bytes, len); return 0; } /* * Encapsulate the data packet in PIM Register message and send it to the RP. */ static int pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy, struct mfc *rt) { struct mbuf *mb_first; struct ip *ip_outer; struct pim_encap_pimhdr *pimhdr; int len = ntohs(ip->ip_len); vifi_t vifi = rt->mfc_parent; VIF_LOCK_ASSERT(); if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) { m_freem(mb_copy); return EADDRNOTAVAIL; /* The iif vif is invalid */ } /* * Add a new mbuf with the encapsulating header */ mb_first = m_gethdr(M_NOWAIT, MT_DATA); if (mb_first == NULL) { m_freem(mb_copy); return ENOBUFS; } mb_first->m_data += max_linkhdr; mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); mb_first->m_next = mb_copy; mb_first->m_pkthdr.len = len + mb_first->m_len; /* * Fill in the encapsulating IP and PIM header */ ip_outer = mtod(mb_first, struct ip *); *ip_outer = pim_encap_iphdr; ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr)); ip_outer->ip_src = V_viftable[vifi].v_lcl_addr; ip_outer->ip_dst = rt->mfc_rp; /* * Copy the inner header TOS to the outer header, and take care of the * IP_DF bit. */ ip_outer->ip_tos = ip->ip_tos; if (ip->ip_off & htons(IP_DF)) ip_outer->ip_off |= htons(IP_DF); ip_fillid(ip_outer); pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer + sizeof(pim_encap_iphdr)); *pimhdr = pim_encap_pimhdr; /* If the iif crosses a border, set the Border-bit */ if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config) pimhdr->flags |= htonl(PIM_BORDER_REGISTER); mb_first->m_data += sizeof(pim_encap_iphdr); pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); mb_first->m_data -= sizeof(pim_encap_iphdr); send_packet(vifp, mb_first); /* Keep statistics */ PIMSTAT_INC(pims_snd_registers_msgs); PIMSTAT_ADD(pims_snd_registers_bytes, len); return 0; } /* * pim_encapcheck() is called by the encap4_input() path at runtime to * determine if a packet is for PIM; allowing PIM to be dynamically loaded * into the kernel. */ static int pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg) { #ifdef DIAGNOSTIC KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM")); #endif if (proto != IPPROTO_PIM) return 0; /* not for us; reject the datagram. */ return 64; /* claim the datagram. */ } /* * PIM-SMv2 and PIM-DM messages processing. * Receives and verifies the PIM control messages, and passes them * up to the listening socket, using rip_input(). * The only message with special processing is the PIM_REGISTER message * (used by PIM-SM): the PIM header is stripped off, and the inner packet * is passed to if_simloop(). */ int pim_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m = *mp; struct ip *ip = mtod(m, struct ip *); struct pim *pim; int iphlen = *offp; int minlen; int datalen = ntohs(ip->ip_len) - iphlen; int ip_tos; *mp = NULL; /* Keep statistics */ PIMSTAT_INC(pims_rcv_total_msgs); PIMSTAT_ADD(pims_rcv_total_bytes, datalen); /* * Validate lengths */ if (datalen < PIM_MINLEN) { PIMSTAT_INC(pims_rcv_tooshort); CTR3(KTR_IPMF, "%s: short packet (%d) from %s", __func__, datalen, inet_ntoa(ip->ip_src)); m_freem(m); return (IPPROTO_DONE); } /* * If the packet is at least as big as a REGISTER, go agead * and grab the PIM REGISTER header size, to avoid another * possible m_pullup() later. * * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 */ minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); /* * Get the IP and PIM headers in contiguous memory, and * possibly the PIM REGISTER header. */ if (m->m_len < minlen && (m = m_pullup(m, minlen)) == 0) { CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__); return (IPPROTO_DONE); } /* m_pullup() may have given us a new mbuf so reset ip. */ ip = mtod(m, struct ip *); ip_tos = ip->ip_tos; /* adjust mbuf to point to the PIM header */ m->m_data += iphlen; m->m_len -= iphlen; pim = mtod(m, struct pim *); /* * Validate checksum. If PIM REGISTER, exclude the data packet. * * XXX: some older PIMv2 implementations don't make this distinction, * so for compatibility reason perform the checksum over part of the * message, and if error, then over the whole message. */ if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { /* do nothing, checksum okay */ } else if (in_cksum(m, datalen)) { PIMSTAT_INC(pims_rcv_badsum); CTR1(KTR_IPMF, "%s: invalid checksum", __func__); m_freem(m); return (IPPROTO_DONE); } /* PIM version check */ if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { PIMSTAT_INC(pims_rcv_badversion); CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__, (int)PIM_VT_V(pim->pim_vt), PIM_VERSION); m_freem(m); return (IPPROTO_DONE); } /* restore mbuf back to the outer IP */ m->m_data -= iphlen; m->m_len += iphlen; if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { /* * Since this is a REGISTER, we'll make a copy of the register * headers ip + pim + u_int32 + encap_ip, to be passed up to the * routing daemon. */ struct sockaddr_in dst = { sizeof(dst), AF_INET }; struct mbuf *mcp; struct ip *encap_ip; u_int32_t *reghdr; struct ifnet *vifp; VIF_LOCK(); if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) { VIF_UNLOCK(); CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__, (int)V_reg_vif_num); m_freem(m); return (IPPROTO_DONE); } /* XXX need refcnt? */ vifp = V_viftable[V_reg_vif_num].v_ifp; VIF_UNLOCK(); /* * Validate length */ if (datalen < PIM_REG_MINLEN) { PIMSTAT_INC(pims_rcv_tooshort); PIMSTAT_INC(pims_rcv_badregisters); CTR1(KTR_IPMF, "%s: register packet size too small", __func__); m_freem(m); return (IPPROTO_DONE); } reghdr = (u_int32_t *)(pim + 1); encap_ip = (struct ip *)(reghdr + 1); CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d", __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len)); /* verify the version number of the inner packet */ if (encap_ip->ip_v != IPVERSION) { PIMSTAT_INC(pims_rcv_badregisters); CTR1(KTR_IPMF, "%s: bad encap ip version", __func__); m_freem(m); return (IPPROTO_DONE); } /* verify the inner packet is destined to a mcast group */ if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { PIMSTAT_INC(pims_rcv_badregisters); CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__, inet_ntoa(encap_ip->ip_dst)); m_freem(m); return (IPPROTO_DONE); } /* If a NULL_REGISTER, pass it to the daemon */ if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) goto pim_input_to_daemon; /* * Copy the TOS from the outer IP header to the inner IP header. */ if (encap_ip->ip_tos != ip_tos) { /* Outer TOS -> inner TOS */ encap_ip->ip_tos = ip_tos; /* Recompute the inner header checksum. Sigh... */ /* adjust mbuf to point to the inner IP header */ m->m_data += (iphlen + PIM_MINLEN); m->m_len -= (iphlen + PIM_MINLEN); encap_ip->ip_sum = 0; encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); /* restore mbuf to point back to the outer IP header */ m->m_data -= (iphlen + PIM_MINLEN); m->m_len += (iphlen + PIM_MINLEN); } /* * Decapsulate the inner IP packet and loopback to forward it * as a normal multicast packet. Also, make a copy of the * outer_iphdr + pimhdr + reghdr + encap_iphdr * to pass to the daemon later, so it can take the appropriate * actions (e.g., send back PIM_REGISTER_STOP). * XXX: here m->m_data points to the outer IP header. */ mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); if (mcp == NULL) { CTR1(KTR_IPMF, "%s: m_copy() failed", __func__); m_freem(m); return (IPPROTO_DONE); } /* Keep statistics */ /* XXX: registers_bytes include only the encap. mcast pkt */ PIMSTAT_INC(pims_rcv_registers_msgs); PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len)); /* * forward the inner ip packet; point m_data at the inner ip. */ m_adj(m, iphlen + PIM_MINLEN); CTR4(KTR_IPMF, "%s: forward decap'd REGISTER: src %lx dst %lx vif %d", __func__, (u_long)ntohl(encap_ip->ip_src.s_addr), (u_long)ntohl(encap_ip->ip_dst.s_addr), (int)V_reg_vif_num); /* NB: vifp was collected above; can it change on us? */ if_simloop(vifp, m, dst.sin_family, 0); /* prepare the register head to send to the mrouting daemon */ m = mcp; } pim_input_to_daemon: /* * Pass the PIM message up to the daemon; if it is a Register message, * pass the 'head' only up to the daemon. This includes the * outer IP header, PIM header, PIM-Register header and the * inner IP header. * XXX: the outer IP header pkt size of a Register is not adjust to * reflect the fact that the inner multicast data is truncated. */ *mp = m; rip_input(mp, offp, proto); return (IPPROTO_DONE); } static int sysctl_mfctable(SYSCTL_HANDLER_ARGS) { struct mfc *rt; int error, i; if (req->newptr) return (EPERM); if (V_mfchashtbl == NULL) /* XXX unlocked */ return (0); error = sysctl_wire_old_buffer(req, 0); if (error) return (error); MFC_LOCK(); for (i = 0; i < mfchashsize; i++) { LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) { error = SYSCTL_OUT(req, rt, sizeof(struct mfc)); if (error) goto out_locked; } } out_locked: MFC_UNLOCK(); return (error); } static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, sysctl_mfctable, "IPv4 Multicast Forwarding Table " "(struct *mfc[mfchashsize], netinet/ip_mroute.h)"); static void vnet_mroute_init(const void *unused __unused) { MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO); bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers)); callout_init(&V_expire_upcalls_ch, 1); callout_init(&V_bw_upcalls_ch, 1); callout_init(&V_bw_meter_ch, 1); } -VNET_SYSINIT(vnet_mroute_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mroute_init, +VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init, NULL); static void vnet_mroute_uninit(const void *unused __unused) { FREE(V_nexpire, M_MRTABLE); V_nexpire = NULL; } -VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, +VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE, vnet_mroute_uninit, NULL); static int ip_mroute_modevent(module_t mod, int type, void *unused) { switch (type) { case MOD_LOAD: MROUTER_LOCK_INIT(); if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, if_detached_event, NULL, EVENTHANDLER_PRI_ANY); if (if_detach_event_tag == NULL) { printf("ip_mroute: unable to register " "ifnet_departure_event handler\n"); MROUTER_LOCK_DESTROY(); return (EINVAL); } MFC_LOCK_INIT(); VIF_LOCK_INIT(); mfchashsize = MFCHASHSIZE; if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) && !powerof2(mfchashsize)) { printf("WARNING: %s not a power of 2; using default\n", "net.inet.ip.mfchashsize"); mfchashsize = MFCHASHSIZE; } pim_squelch_wholepkt = 0; TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt", &pim_squelch_wholepkt); pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM, pim_encapcheck, &in_pim_protosw, NULL); if (pim_encap_cookie == NULL) { printf("ip_mroute: unable to attach pim encap\n"); VIF_LOCK_DESTROY(); MFC_LOCK_DESTROY(); MROUTER_LOCK_DESTROY(); return (EINVAL); } ip_mcast_src = X_ip_mcast_src; ip_mforward = X_ip_mforward; ip_mrouter_done = X_ip_mrouter_done; ip_mrouter_get = X_ip_mrouter_get; ip_mrouter_set = X_ip_mrouter_set; ip_rsvp_force_done = X_ip_rsvp_force_done; ip_rsvp_vif = X_ip_rsvp_vif; legal_vif_num = X_legal_vif_num; mrt_ioctl = X_mrt_ioctl; rsvp_input_p = X_rsvp_input; break; case MOD_UNLOAD: /* * Typically module unload happens after the user-level * process has shutdown the kernel services (the check * below insures someone can't just yank the module out * from under a running process). But if the module is * just loaded and then unloaded w/o starting up a user * process we still need to cleanup. */ MROUTER_LOCK(); if (ip_mrouter_cnt != 0) { MROUTER_UNLOCK(); return (EINVAL); } ip_mrouter_unloading = 1; MROUTER_UNLOCK(); EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag); if (pim_encap_cookie) { encap_detach(pim_encap_cookie); pim_encap_cookie = NULL; } ip_mcast_src = NULL; ip_mforward = NULL; ip_mrouter_done = NULL; ip_mrouter_get = NULL; ip_mrouter_set = NULL; ip_rsvp_force_done = NULL; ip_rsvp_vif = NULL; legal_vif_num = NULL; mrt_ioctl = NULL; rsvp_input_p = NULL; VIF_LOCK_DESTROY(); MFC_LOCK_DESTROY(); MROUTER_LOCK_DESTROY(); break; default: return EOPNOTSUPP; } return 0; } static moduledata_t ip_mroutemod = { "ip_mroute", ip_mroute_modevent, 0 }; -DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_MIDDLE); +DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE); Index: projects/vnet/sys/netinet/ip_var.h =================================================================== --- projects/vnet/sys/netinet/ip_var.h (revision 294565) +++ projects/vnet/sys/netinet/ip_var.h (revision 294566) @@ -1,304 +1,301 @@ /*- * Copyright (c) 1982, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ip_var.h 8.2 (Berkeley) 1/9/95 * $FreeBSD$ */ #ifndef _NETINET_IP_VAR_H_ #define _NETINET_IP_VAR_H_ #include /* * Overlay for ip header used by other protocols (tcp, udp). */ struct ipovly { u_char ih_x1[9]; /* (unused) */ u_char ih_pr; /* protocol */ u_short ih_len; /* protocol length */ struct in_addr ih_src; /* source internet address */ struct in_addr ih_dst; /* destination internet address */ }; #ifdef _KERNEL /* * Ip reassembly queue structure. Each fragment * being reassembled is attached to one of these structures. * They are timed out after ipq_ttl drops to 0, and may also * be reclaimed if memory becomes tight. */ struct ipq { TAILQ_ENTRY(ipq) ipq_list; /* to other reass headers */ u_char ipq_ttl; /* time for reass q to live */ u_char ipq_p; /* protocol of this fragment */ u_short ipq_id; /* sequence id for reassembly */ struct mbuf *ipq_frags; /* to ip headers of fragments */ struct in_addr ipq_src,ipq_dst; u_char ipq_nfrags; /* # frags in this packet */ struct label *ipq_label; /* MAC label */ }; #endif /* _KERNEL */ /* * Structure stored in mbuf in inpcb.ip_options * and passed to ip_output when ip options are in use. * The actual length of the options (including ipopt_dst) * is in m_len. */ #define MAX_IPOPTLEN 40 struct ipoption { struct in_addr ipopt_dst; /* first-hop dst if source routed */ char ipopt_list[MAX_IPOPTLEN]; /* options proper */ }; /* * Structure attached to inpcb.ip_moptions and * passed to ip_output when IP multicast options are in use. * This structure is lazy-allocated. */ struct ip_moptions { struct ifnet *imo_multicast_ifp; /* ifp for outgoing multicasts */ struct in_addr imo_multicast_addr; /* ifindex/addr on MULTICAST_IF */ u_long imo_multicast_vif; /* vif num outgoing multicasts */ u_char imo_multicast_ttl; /* TTL for outgoing multicasts */ u_char imo_multicast_loop; /* 1 => hear sends if a member */ u_short imo_num_memberships; /* no. memberships this socket */ u_short imo_max_memberships; /* max memberships this socket */ struct in_multi **imo_membership; /* group memberships */ struct in_mfilter *imo_mfilters; /* source filters */ STAILQ_ENTRY(ip_moptions) imo_link; }; struct ipstat { uint64_t ips_total; /* total packets received */ uint64_t ips_badsum; /* checksum bad */ uint64_t ips_tooshort; /* packet too short */ uint64_t ips_toosmall; /* not enough data */ uint64_t ips_badhlen; /* ip header length < data size */ uint64_t ips_badlen; /* ip length < ip header length */ uint64_t ips_fragments; /* fragments received */ uint64_t ips_fragdropped; /* frags dropped (dups, out of space) */ uint64_t ips_fragtimeout; /* fragments timed out */ uint64_t ips_forward; /* packets forwarded */ uint64_t ips_fastforward; /* packets fast forwarded */ uint64_t ips_cantforward; /* packets rcvd for unreachable dest */ uint64_t ips_redirectsent; /* packets forwarded on same net */ uint64_t ips_noproto; /* unknown or unsupported protocol */ uint64_t ips_delivered; /* datagrams delivered to upper level*/ uint64_t ips_localout; /* total ip packets generated here */ uint64_t ips_odropped; /* lost packets due to nobufs, etc. */ uint64_t ips_reassembled; /* total packets reassembled ok */ uint64_t ips_fragmented; /* datagrams successfully fragmented */ uint64_t ips_ofragments; /* output fragments created */ uint64_t ips_cantfrag; /* don't fragment flag was set, etc. */ uint64_t ips_badoptions; /* error in option processing */ uint64_t ips_noroute; /* packets discarded due to no route */ uint64_t ips_badvers; /* ip version != 4 */ uint64_t ips_rawout; /* total raw ip packets generated */ uint64_t ips_toolong; /* ip length > max ip packet size */ uint64_t ips_notmember; /* multicasts for unregistered grps */ uint64_t ips_nogif; /* no match gif found */ uint64_t ips_badaddr; /* invalid address on header */ }; #ifdef _KERNEL #include #include VNET_PCPUSTAT_DECLARE(struct ipstat, ipstat); /* * In-kernel consumers can use these accessor macros directly to update * stats. */ #define IPSTAT_ADD(name, val) \ VNET_PCPUSTAT_ADD(struct ipstat, ipstat, name, (val)) #define IPSTAT_SUB(name, val) IPSTAT_ADD(name, -(val)) #define IPSTAT_INC(name) IPSTAT_ADD(name, 1) #define IPSTAT_DEC(name) IPSTAT_SUB(name, 1) /* * Kernel module consumers must use this accessor macro. */ void kmod_ipstat_inc(int statnum); #define KMOD_IPSTAT_INC(name) \ kmod_ipstat_inc(offsetof(struct ipstat, name) / sizeof(uint64_t)) void kmod_ipstat_dec(int statnum); #define KMOD_IPSTAT_DEC(name) \ kmod_ipstat_dec(offsetof(struct ipstat, name) / sizeof(uint64_t)) /* flags passed to ip_output as last parameter */ #define IP_FORWARDING 0x1 /* most of ip header exists */ #define IP_RAWOUTPUT 0x2 /* raw ip header exists */ #define IP_SENDONES 0x4 /* send all-ones broadcast */ #define IP_SENDTOIF 0x8 /* send on specific ifnet */ #define IP_ROUTETOIF SO_DONTROUTE /* 0x10 bypass routing tables */ #define IP_ALLOWBROADCAST SO_BROADCAST /* 0x20 can send broadcast packets */ #define IP_NODEFAULTFLOWID 0x40 /* Don't set the flowid from inp */ #ifdef __NO_STRICT_ALIGNMENT #define IP_HDR_ALIGNED_P(ip) 1 #else #define IP_HDR_ALIGNED_P(ip) ((((intptr_t) (ip)) & 3) == 0) #endif struct ip; struct inpcb; struct route; struct sockopt; VNET_DECLARE(int, ip_defttl); /* default IP ttl */ VNET_DECLARE(int, ipforwarding); /* ip forwarding */ #ifdef IPSTEALTH VNET_DECLARE(int, ipstealth); /* stealth forwarding */ #endif extern u_char ip_protox[]; VNET_DECLARE(struct socket *, ip_rsvpd); /* reservation protocol daemon*/ VNET_DECLARE(struct socket *, ip_mrouter); /* multicast routing daemon */ extern int (*legal_vif_num)(int); extern u_long (*ip_mcast_src)(int); VNET_DECLARE(int, rsvp_on); VNET_DECLARE(int, drop_redirect); extern struct pr_usrreqs rip_usrreqs; #define V_ip_id VNET(ip_id) #define V_ip_defttl VNET(ip_defttl) #define V_ipforwarding VNET(ipforwarding) #ifdef IPSTEALTH #define V_ipstealth VNET(ipstealth) #endif #define V_ip_rsvpd VNET(ip_rsvpd) #define V_ip_mrouter VNET(ip_mrouter) #define V_rsvp_on VNET(rsvp_on) #define V_drop_redirect VNET(drop_redirect) void inp_freemoptions(struct ip_moptions *); int inp_getmoptions(struct inpcb *, struct sockopt *); int inp_setmoptions(struct inpcb *, struct sockopt *); int ip_ctloutput(struct socket *, struct sockopt *sopt); void ip_drain(void); int ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, u_long if_hwassist_flags); void ip_forward(struct mbuf *m, int srcrt); void ip_init(void); #ifdef VIMAGE void ip_destroy(void); #endif extern int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, struct ip_moptions *); int ip_output(struct mbuf *, struct mbuf *, struct route *, int, struct ip_moptions *, struct inpcb *); int ipproto_register(short); int ipproto_unregister(short); struct mbuf * ip_reass(struct mbuf *); void ip_savecontrol(struct inpcb *, struct mbuf **, struct ip *, struct mbuf *); void ip_slowtimo(void); void ip_fillid(struct ip *); int rip_ctloutput(struct socket *, struct sockopt *); void rip_ctlinput(int, struct sockaddr *, void *); void rip_init(void); -#ifdef VIMAGE -void rip_destroy(void); -#endif int rip_input(struct mbuf **, int *, int); int rip_output(struct mbuf *, struct socket *, ...); int ipip_input(struct mbuf **, int *, int); int rsvp_input(struct mbuf **, int *, int); int ip_rsvp_init(struct socket *); int ip_rsvp_done(void); extern int (*ip_rsvp_vif)(struct socket *, struct sockopt *); extern void (*ip_rsvp_force_done)(struct socket *); extern int (*rsvp_input_p)(struct mbuf **, int *, int); VNET_DECLARE(struct pfil_head, inet_pfil_hook); /* packet filter hooks */ #define V_inet_pfil_hook VNET(inet_pfil_hook) void in_delayed_cksum(struct mbuf *m); /* Hooks for ipfw, dummynet, divert etc. Most are declared in raw_ip.c */ /* * Reference to an ipfw or packet filter rule that can be carried * outside critical sections. * A rule is identified by rulenum:rule_id which is ordered. * In version chain_id the rule can be found in slot 'slot', so * we don't need a lookup if chain_id == chain->id. * * On exit from the firewall this structure refers to the rule after * the matching one (slot points to the new rule; rulenum:rule_id-1 * is the matching rule), and additional info (e.g. info often contains * the insn argument or tablearg in the low 16 bits, in host format). * On entry, the structure is valid if slot>0, and refers to the starting * rules. 'info' contains the reason for reinject, e.g. divert port, * divert direction, and so on. */ struct ipfw_rule_ref { uint32_t slot; /* slot for matching rule */ uint32_t rulenum; /* matching rule number */ uint32_t rule_id; /* matching rule id */ uint32_t chain_id; /* ruleset id */ uint32_t info; /* see below */ }; enum { IPFW_INFO_MASK = 0x0000ffff, IPFW_INFO_OUT = 0x00000000, /* outgoing, just for convenience */ IPFW_INFO_IN = 0x80000000, /* incoming, overloads dir */ IPFW_ONEPASS = 0x40000000, /* One-pass, do not reinject */ IPFW_IS_MASK = 0x30000000, /* which source ? */ IPFW_IS_DIVERT = 0x20000000, IPFW_IS_DUMMYNET =0x10000000, IPFW_IS_PIPE = 0x08000000, /* pipe=1, queue = 0 */ }; #define MTAG_IPFW 1148380143 /* IPFW-tagged cookie */ #define MTAG_IPFW_RULE 1262273568 /* rule reference */ #define MTAG_IPFW_CALL 1308397630 /* call stack */ struct ip_fw_args; typedef int (*ip_fw_chk_ptr_t)(struct ip_fw_args *args); typedef int (*ip_fw_ctl_ptr_t)(struct sockopt *); VNET_DECLARE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr); #define V_ip_fw_ctl_ptr VNET(ip_fw_ctl_ptr) /* Divert hooks. */ extern void (*ip_divert_ptr)(struct mbuf *m, int incoming); /* ng_ipfw hooks -- XXX make it the same as divert and dummynet */ extern int (*ng_ipfw_input_p)(struct mbuf **, int, struct ip_fw_args *, int); extern int (*ip_dn_ctl_ptr)(struct sockopt *); extern int (*ip_dn_io_ptr)(struct mbuf **, int, struct ip_fw_args *); #endif /* _KERNEL */ #endif /* !_NETINET_IP_VAR_H_ */ Index: projects/vnet/sys/netinet/raw_ip.c =================================================================== --- projects/vnet/sys/netinet/raw_ip.c (revision 294565) +++ projects/vnet/sys/netinet/raw_ip.c (revision 294566) @@ -1,1131 +1,1132 @@ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef IPSEC #include #endif /*IPSEC*/ #include #include VNET_DEFINE(int, ip_defttl) = IPDEFTTL; SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_defttl), 0, "Maximum TTL on IP packets"); VNET_DEFINE(struct inpcbhead, ripcb); VNET_DEFINE(struct inpcbinfo, ripcbinfo); #define V_ripcb VNET(ripcb) #define V_ripcbinfo VNET(ripcbinfo) /* * Control and data hooks for ipfw, dummynet, divert and so on. * The data hooks are not used here but it is convenient * to keep them all in one place. */ VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL; VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL; int (*ip_dn_ctl_ptr)(struct sockopt *); int (*ip_dn_io_ptr)(struct mbuf **, int, struct ip_fw_args *); void (*ip_divert_ptr)(struct mbuf *, int); int (*ng_ipfw_input_p)(struct mbuf **, int, struct ip_fw_args *, int); #ifdef INET /* * Hooks for multicast routing. They all default to NULL, so leave them not * initialized and rely on BSS being set to 0. */ /* * The socket used to communicate with the multicast routing daemon. */ VNET_DEFINE(struct socket *, ip_mrouter); /* * The various mrouter and rsvp functions. */ int (*ip_mrouter_set)(struct socket *, struct sockopt *); int (*ip_mrouter_get)(struct socket *, struct sockopt *); int (*ip_mrouter_done)(void); int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, struct ip_moptions *); int (*mrt_ioctl)(u_long, caddr_t, int); int (*legal_vif_num)(int); u_long (*ip_mcast_src)(int); int (*rsvp_input_p)(struct mbuf **, int *, int); int (*ip_rsvp_vif)(struct socket *, struct sockopt *); void (*ip_rsvp_force_done)(struct socket *); #endif /* INET */ u_long rip_sendspace = 9216; SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); u_long rip_recvspace = 9216; SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); /* * Hash functions */ #define INP_PCBHASH_RAW_SIZE 256 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \ (((proto) + (laddr) + (faddr)) % (mask) + 1) #ifdef INET static void rip_inshash(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct inpcbhead *pcbhash; int hash; INP_INFO_WLOCK_ASSERT(pcbinfo); INP_WLOCK_ASSERT(inp); if (inp->inp_ip_p != 0 && inp->inp_laddr.s_addr != INADDR_ANY && inp->inp_faddr.s_addr != INADDR_ANY) { hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr, inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask); } else hash = 0; pcbhash = &pcbinfo->ipi_hashbase[hash]; LIST_INSERT_HEAD(pcbhash, inp, inp_hash); } static void rip_delhash(struct inpcb *inp) { INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); INP_WLOCK_ASSERT(inp); LIST_REMOVE(inp, inp_hash); } #endif /* INET */ /* * Raw interface to IP protocol. */ /* * Initialize raw connection block q. */ static void rip_zone_change(void *tag) { uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); } static int rip_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "rawinp"); return (0); } void rip_init(void) { in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE, 1, "ripcb", rip_inpcb_init, NULL, UMA_ZONE_NOFREE, IPI_HASHFIELDS_NONE); EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL, EVENTHANDLER_PRI_ANY); } #ifdef VIMAGE -void +static void rip_destroy(void) { in_pcbinfo_destroy(&V_ripcbinfo); } +VNET_SYSUNINIT(raw_ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, rip_destroy, NULL); #endif #ifdef INET static int rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, struct sockaddr_in *ripsrc) { int policyfail = 0; INP_LOCK_ASSERT(last); #ifdef IPSEC /* check AH/ESP integrity. */ if (ipsec4_in_reject(n, last)) { policyfail = 1; } #endif /* IPSEC */ #ifdef MAC if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) policyfail = 1; #endif /* Check the minimum TTL for socket. */ if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) policyfail = 1; if (!policyfail) { struct mbuf *opts = NULL; struct socket *so; so = last->inp_socket; if ((last->inp_flags & INP_CONTROLOPTS) || (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) ip_savecontrol(last, &opts, ip, n); SOCKBUF_LOCK(&so->so_rcv); if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)ripsrc, n, opts) == 0) { /* should notify about lost packet */ m_freem(n); if (opts) m_freem(opts); SOCKBUF_UNLOCK(&so->so_rcv); } else sorwakeup_locked(so); } else m_freem(n); return (policyfail); } /* * Setup generic address and protocol structures for raw_input routine, then * pass them along with mbuf chain. */ int rip_input(struct mbuf **mp, int *offp, int proto) { struct ifnet *ifp; struct mbuf *m = *mp; struct ip *ip = mtod(m, struct ip *); struct inpcb *inp, *last; struct sockaddr_in ripsrc; int hash; *mp = NULL; bzero(&ripsrc, sizeof(ripsrc)); ripsrc.sin_len = sizeof(ripsrc); ripsrc.sin_family = AF_INET; ripsrc.sin_addr = ip->ip_src; last = NULL; ifp = m->m_pkthdr.rcvif; hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); INP_INFO_RLOCK(&V_ripcbinfo); LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { if (inp->inp_ip_p != proto) continue; #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) continue; if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) continue; if (jailed_without_vnet(inp->inp_cred)) { /* * XXX: If faddr was bound to multicast group, * jailed raw socket will drop datagram. */ if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) continue; } if (last != NULL) { struct mbuf *n; n = m_copy(m, 0, (int)M_COPYALL); if (n != NULL) (void) rip_append(last, ip, n, &ripsrc); /* XXX count dropped packet */ INP_RUNLOCK(last); } INP_RLOCK(inp); last = inp; } LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { if (inp->inp_ip_p && inp->inp_ip_p != proto) continue; #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (!in_nullhost(inp->inp_laddr) && !in_hosteq(inp->inp_laddr, ip->ip_dst)) continue; if (!in_nullhost(inp->inp_faddr) && !in_hosteq(inp->inp_faddr, ip->ip_src)) continue; if (jailed_without_vnet(inp->inp_cred)) { /* * Allow raw socket in jail to receive multicast; * assume process had PRIV_NETINET_RAW at attach, * and fall through into normal filter path if so. */ if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) continue; } /* * If this raw socket has multicast state, and we * have received a multicast, check if this socket * should receive it, as multicast filtering is now * the responsibility of the transport layer. */ if (inp->inp_moptions != NULL && IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { /* * If the incoming datagram is for IGMP, allow it * through unconditionally to the raw socket. * * In the case of IGMPv2, we may not have explicitly * joined the group, and may have set IFF_ALLMULTI * on the interface. imo_multi_filter() may discard * control traffic we actually need to see. * * Userland multicast routing daemons should continue * filter the control traffic appropriately. */ int blocked; blocked = MCAST_PASS; if (proto != IPPROTO_IGMP) { struct sockaddr_in group; bzero(&group, sizeof(struct sockaddr_in)); group.sin_len = sizeof(struct sockaddr_in); group.sin_family = AF_INET; group.sin_addr = ip->ip_dst; blocked = imo_multi_filter(inp->inp_moptions, ifp, (struct sockaddr *)&group, (struct sockaddr *)&ripsrc); } if (blocked != MCAST_PASS) { IPSTAT_INC(ips_notmember); continue; } } if (last != NULL) { struct mbuf *n; n = m_copy(m, 0, (int)M_COPYALL); if (n != NULL) (void) rip_append(last, ip, n, &ripsrc); /* XXX count dropped packet */ INP_RUNLOCK(last); } INP_RLOCK(inp); last = inp; } INP_INFO_RUNLOCK(&V_ripcbinfo); if (last != NULL) { if (rip_append(last, ip, m, &ripsrc) != 0) IPSTAT_INC(ips_delivered); INP_RUNLOCK(last); } else { m_freem(m); IPSTAT_INC(ips_noproto); IPSTAT_DEC(ips_delivered); } return (IPPROTO_DONE); } /* * Generate IP header and pass packet to ip_output. Tack on options user may * have setup with control call. */ int rip_output(struct mbuf *m, struct socket *so, ...) { struct ip *ip; int error; struct inpcb *inp = sotoinpcb(so); va_list ap; u_long dst; int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | IP_ALLOWBROADCAST; va_start(ap, so); dst = va_arg(ap, u_long); va_end(ap); /* * If the user handed us a complete IP packet, use it. Otherwise, * allocate an mbuf for a header and fill it in. */ if ((inp->inp_flags & INP_HDRINCL) == 0) { if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { m_freem(m); return(EMSGSIZE); } M_PREPEND(m, sizeof(struct ip), M_NOWAIT); if (m == NULL) return(ENOBUFS); INP_RLOCK(inp); ip = mtod(m, struct ip *); ip->ip_tos = inp->inp_ip_tos; if (inp->inp_flags & INP_DONTFRAG) ip->ip_off = htons(IP_DF); else ip->ip_off = htons(0); ip->ip_p = inp->inp_ip_p; ip->ip_len = htons(m->m_pkthdr.len); ip->ip_src = inp->inp_laddr; ip->ip_dst.s_addr = dst; if (jailed(inp->inp_cred)) { /* * prison_local_ip4() would be good enough but would * let a source of INADDR_ANY pass, which we do not * want to see from jails. */ if (ip->ip_src.s_addr == INADDR_ANY) { error = in_pcbladdr(inp, &ip->ip_dst, &ip->ip_src, inp->inp_cred); } else { error = prison_local_ip4(inp->inp_cred, &ip->ip_src); } if (error != 0) { INP_RUNLOCK(inp); m_freem(m); return (error); } } ip->ip_ttl = inp->inp_ip_ttl; } else { if (m->m_pkthdr.len > IP_MAXPACKET) { m_freem(m); return(EMSGSIZE); } INP_RLOCK(inp); ip = mtod(m, struct ip *); error = prison_check_ip4(inp->inp_cred, &ip->ip_src); if (error != 0) { INP_RUNLOCK(inp); m_freem(m); return (error); } /* * Don't allow both user specified and setsockopt options, * and don't allow packet length sizes that will crash. */ if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) || (ntohs(ip->ip_len) > m->m_pkthdr.len) || (ntohs(ip->ip_len) < (ip->ip_hl << 2))) { INP_RUNLOCK(inp); m_freem(m); return (EINVAL); } /* * This doesn't allow application to specify ID of zero, * but we got this limitation from the beginning of history. */ if (ip->ip_id == 0) ip_fillid(ip); /* * XXX prevent ip_output from overwriting header fields. */ flags |= IP_RAWOUTPUT; IPSTAT_INC(ips_rawout); } if (inp->inp_flags & INP_ONESBCAST) flags |= IP_SENDONES; #ifdef MAC mac_inpcb_create_mbuf(inp, m); #endif error = ip_output(m, inp->inp_options, NULL, flags, inp->inp_moptions, inp); INP_RUNLOCK(inp); return (error); } /* * Raw IP socket option processing. * * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could * only be created by a privileged process, and as such, socket option * operations to manage system properties on any raw socket were allowed to * take place without explicit additional access control checks. However, * raw sockets can now also be created in jail(), and therefore explicit * checks are now required. Likewise, raw sockets can be used by a process * after it gives up privilege, so some caution is required. For options * passed down to the IP layer via ip_ctloutput(), checks are assumed to be * performed in ip_ctloutput() and therefore no check occurs here. * Unilaterally checking priv_check() here breaks normal IP socket option * operations on raw sockets. * * When adding new socket options here, make sure to add access control * checks here as necessary. * * XXX-BZ inp locking? */ int rip_ctloutput(struct socket *so, struct sockopt *sopt) { struct inpcb *inp = sotoinpcb(so); int error, optval; if (sopt->sopt_level != IPPROTO_IP) { if ((sopt->sopt_level == SOL_SOCKET) && (sopt->sopt_name == SO_SETFIB)) { inp->inp_inc.inc_fibnum = so->so_fibnum; return (0); } return (EINVAL); } error = 0; switch (sopt->sopt_dir) { case SOPT_GET: switch (sopt->sopt_name) { case IP_HDRINCL: optval = inp->inp_flags & INP_HDRINCL; error = sooptcopyout(sopt, &optval, sizeof optval); break; case IP_FW3: /* generic ipfw v.3 functions */ case IP_FW_ADD: /* ADD actually returns the body... */ case IP_FW_GET: case IP_FW_TABLE_GETSIZE: case IP_FW_TABLE_LIST: case IP_FW_NAT_GET_CONFIG: case IP_FW_NAT_GET_LOG: if (V_ip_fw_ctl_ptr != NULL) error = V_ip_fw_ctl_ptr(sopt); else error = ENOPROTOOPT; break; case IP_DUMMYNET3: /* generic dummynet v.3 functions */ case IP_DUMMYNET_GET: if (ip_dn_ctl_ptr != NULL) error = ip_dn_ctl_ptr(sopt); else error = ENOPROTOOPT; break ; case MRT_INIT: case MRT_DONE: case MRT_ADD_VIF: case MRT_DEL_VIF: case MRT_ADD_MFC: case MRT_DEL_MFC: case MRT_VERSION: case MRT_ASSERT: case MRT_API_SUPPORT: case MRT_API_CONFIG: case MRT_ADD_BW_UPCALL: case MRT_DEL_BW_UPCALL: error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error != 0) return (error); error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : EOPNOTSUPP; break; default: error = ip_ctloutput(so, sopt); break; } break; case SOPT_SET: switch (sopt->sopt_name) { case IP_HDRINCL: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; if (optval) inp->inp_flags |= INP_HDRINCL; else inp->inp_flags &= ~INP_HDRINCL; break; case IP_FW3: /* generic ipfw v.3 functions */ case IP_FW_ADD: case IP_FW_DEL: case IP_FW_FLUSH: case IP_FW_ZERO: case IP_FW_RESETLOG: case IP_FW_TABLE_ADD: case IP_FW_TABLE_DEL: case IP_FW_TABLE_FLUSH: case IP_FW_NAT_CFG: case IP_FW_NAT_DEL: if (V_ip_fw_ctl_ptr != NULL) error = V_ip_fw_ctl_ptr(sopt); else error = ENOPROTOOPT; break; case IP_DUMMYNET3: /* generic dummynet v.3 functions */ case IP_DUMMYNET_CONFIGURE: case IP_DUMMYNET_DEL: case IP_DUMMYNET_FLUSH: if (ip_dn_ctl_ptr != NULL) error = ip_dn_ctl_ptr(sopt); else error = ENOPROTOOPT ; break ; case IP_RSVP_ON: error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error != 0) return (error); error = ip_rsvp_init(so); break; case IP_RSVP_OFF: error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error != 0) return (error); error = ip_rsvp_done(); break; case IP_RSVP_VIF_ON: case IP_RSVP_VIF_OFF: error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error != 0) return (error); error = ip_rsvp_vif ? ip_rsvp_vif(so, sopt) : EINVAL; break; case MRT_INIT: case MRT_DONE: case MRT_ADD_VIF: case MRT_DEL_VIF: case MRT_ADD_MFC: case MRT_DEL_MFC: case MRT_VERSION: case MRT_ASSERT: case MRT_API_SUPPORT: case MRT_API_CONFIG: case MRT_ADD_BW_UPCALL: case MRT_DEL_BW_UPCALL: error = priv_check(curthread, PRIV_NETINET_MROUTE); if (error != 0) return (error); error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : EOPNOTSUPP; break; default: error = ip_ctloutput(so, sopt); break; } break; } return (error); } /* * This function exists solely to receive the PRC_IFDOWN messages which are * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls * in_ifadown() to remove all routes corresponding to that address. It also * receives the PRC_IFUP messages from if_up() and reinstalls the interface * routes. */ void rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct rm_priotracker in_ifa_tracker; struct in_ifaddr *ia; struct ifnet *ifp; int err; int flags; switch (cmd) { case PRC_IFDOWN: IN_IFADDR_RLOCK(&in_ifa_tracker); TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { if (ia->ia_ifa.ifa_addr == sa && (ia->ia_flags & IFA_ROUTE)) { ifa_ref(&ia->ia_ifa); IN_IFADDR_RUNLOCK(&in_ifa_tracker); /* * in_scrubprefix() kills the interface route. */ in_scrubprefix(ia, 0); /* * in_ifadown gets rid of all the rest of the * routes. This is not quite the right thing * to do, but at least if we are running a * routing process they will come back. */ in_ifadown(&ia->ia_ifa, 0); ifa_free(&ia->ia_ifa); break; } } if (ia == NULL) /* If ia matched, already unlocked. */ IN_IFADDR_RUNLOCK(&in_ifa_tracker); break; case PRC_IFUP: IN_IFADDR_RLOCK(&in_ifa_tracker); TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { if (ia->ia_ifa.ifa_addr == sa) break; } if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) { IN_IFADDR_RUNLOCK(&in_ifa_tracker); return; } ifa_ref(&ia->ia_ifa); IN_IFADDR_RUNLOCK(&in_ifa_tracker); flags = RTF_UP; ifp = ia->ia_ifa.ifa_ifp; if ((ifp->if_flags & IFF_LOOPBACK) || (ifp->if_flags & IFF_POINTOPOINT)) flags |= RTF_HOST; err = ifa_del_loopback_route((struct ifaddr *)ia, sa); err = rtinit(&ia->ia_ifa, RTM_ADD, flags); if (err == 0) ia->ia_flags |= IFA_ROUTE; err = ifa_add_loopback_route((struct ifaddr *)ia, sa); ifa_free(&ia->ia_ifa); break; } } static int rip_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; int error; inp = sotoinpcb(so); KASSERT(inp == NULL, ("rip_attach: inp != NULL")); error = priv_check(td, PRIV_NETINET_RAW); if (error) return (error); if (proto >= IPPROTO_MAX || proto < 0) return EPROTONOSUPPORT; error = soreserve(so, rip_sendspace, rip_recvspace); if (error) return (error); INP_INFO_WLOCK(&V_ripcbinfo); error = in_pcballoc(so, &V_ripcbinfo); if (error) { INP_INFO_WUNLOCK(&V_ripcbinfo); return (error); } inp = (struct inpcb *)so->so_pcb; inp->inp_vflag |= INP_IPV4; inp->inp_ip_p = proto; inp->inp_ip_ttl = V_ip_defttl; rip_inshash(inp); INP_INFO_WUNLOCK(&V_ripcbinfo); INP_WUNLOCK(inp); return (0); } static void rip_detach(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_detach: inp == NULL")); KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, ("rip_detach: not closed")); INP_INFO_WLOCK(&V_ripcbinfo); INP_WLOCK(inp); rip_delhash(inp); if (so == V_ip_mrouter && ip_mrouter_done) ip_mrouter_done(); if (ip_rsvp_force_done) ip_rsvp_force_done(so); if (so == V_ip_rsvpd) ip_rsvp_done(); in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(&V_ripcbinfo); } static void rip_dodisconnect(struct socket *so, struct inpcb *inp) { struct inpcbinfo *pcbinfo; pcbinfo = inp->inp_pcbinfo; INP_INFO_WLOCK(pcbinfo); INP_WLOCK(inp); rip_delhash(inp); inp->inp_faddr.s_addr = INADDR_ANY; rip_inshash(inp); SOCK_LOCK(so); so->so_state &= ~SS_ISCONNECTED; SOCK_UNLOCK(so); INP_WUNLOCK(inp); INP_INFO_WUNLOCK(pcbinfo); } static void rip_abort(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_abort: inp == NULL")); rip_dodisconnect(so, inp); } static void rip_close(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_close: inp == NULL")); rip_dodisconnect(so, inp); } static int rip_disconnect(struct socket *so) { struct inpcb *inp; if ((so->so_state & SS_ISCONNECTED) == 0) return (ENOTCONN); inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); rip_dodisconnect(so, inp); return (0); } static int rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct sockaddr_in *addr = (struct sockaddr_in *)nam; struct inpcb *inp; int error; if (nam->sa_len != sizeof(*addr)) return (EINVAL); error = prison_check_ip4(td->td_ucred, &addr->sin_addr); if (error != 0) return (error); inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_bind: inp == NULL")); if (TAILQ_EMPTY(&V_ifnet) || (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || (addr->sin_addr.s_addr && (inp->inp_flags & INP_BINDANY) == 0 && ifa_ifwithaddr_check((struct sockaddr *)addr) == 0)) return (EADDRNOTAVAIL); INP_INFO_WLOCK(&V_ripcbinfo); INP_WLOCK(inp); rip_delhash(inp); inp->inp_laddr = addr->sin_addr; rip_inshash(inp); INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_ripcbinfo); return (0); } static int rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct sockaddr_in *addr = (struct sockaddr_in *)nam; struct inpcb *inp; if (nam->sa_len != sizeof(*addr)) return (EINVAL); if (TAILQ_EMPTY(&V_ifnet)) return (EADDRNOTAVAIL); if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) return (EAFNOSUPPORT); inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_connect: inp == NULL")); INP_INFO_WLOCK(&V_ripcbinfo); INP_WLOCK(inp); rip_delhash(inp); inp->inp_faddr = addr->sin_addr; rip_inshash(inp); soisconnected(so); INP_WUNLOCK(inp); INP_INFO_WUNLOCK(&V_ripcbinfo); return (0); } static int rip_shutdown(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); INP_WLOCK(inp); socantsendmore(so); INP_WUNLOCK(inp); return (0); } static int rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { struct inpcb *inp; u_long dst; inp = sotoinpcb(so); KASSERT(inp != NULL, ("rip_send: inp == NULL")); /* * Note: 'dst' reads below are unlocked. */ if (so->so_state & SS_ISCONNECTED) { if (nam) { m_freem(m); return (EISCONN); } dst = inp->inp_faddr.s_addr; /* Unlocked read. */ } else { if (nam == NULL) { m_freem(m); return (ENOTCONN); } dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; } return (rip_output(m, so, dst)); } #endif /* INET */ static int rip_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == 0) { n = V_ripcbinfo.ipi_count; n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); return (0); } if (req->newptr != 0) return (EPERM); /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&V_ripcbinfo); gencnt = V_ripcbinfo.ipi_gencnt; n = V_ripcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_ripcbinfo); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return (error); inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == 0) return (ENOMEM); INP_INFO_RLOCK(&V_ripcbinfo); for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt && cr_canseeinpcb(req->td->td_ucred, inp) == 0) { in_pcbref(inp); inp_list[i++] = inp; } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_ripcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xinpcb xi; bzero(&xi, sizeof(xi)); xi.xi_len = sizeof xi; /* XXX should avoid extra copy */ bcopy(inp, &xi.xi_inp, sizeof *inp); if (inp->inp_socket) sotoxsocket(inp->inp_socket, &xi.xi_socket); INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xi, sizeof xi); } else INP_RUNLOCK(inp); } INP_INFO_WLOCK(&V_ripcbinfo); for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (!in_pcbrele_rlocked(inp)) INP_RUNLOCK(inp); } INP_INFO_WUNLOCK(&V_ripcbinfo); if (!error) { /* * Give the user an updated idea of our state. If the * generation differs from what we told her before, she knows * that something happened while we were processing this * request, and it might be necessary to retry. */ INP_INFO_RLOCK(&V_ripcbinfo); xig.xig_gen = V_ripcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_ripcbinfo.ipi_count; INP_INFO_RUNLOCK(&V_ripcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); #ifdef INET struct pr_usrreqs rip_usrreqs = { .pru_abort = rip_abort, .pru_attach = rip_attach, .pru_bind = rip_bind, .pru_connect = rip_connect, .pru_control = in_control, .pru_detach = rip_detach, .pru_disconnect = rip_disconnect, .pru_peeraddr = in_getpeeraddr, .pru_send = rip_send, .pru_shutdown = rip_shutdown, .pru_sockaddr = in_getsockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = rip_close, }; #endif /* INET */ Index: projects/vnet/sys/netinet/sctp_bsd_addr.c =================================================================== --- projects/vnet/sys/netinet/sctp_bsd_addr.c (revision 294565) +++ projects/vnet/sys/netinet/sctp_bsd_addr.c (revision 294566) @@ -1,547 +1,550 @@ /*- * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* Declare all of our malloc named types */ MALLOC_DEFINE(SCTP_M_MAP, "sctp_map", "sctp asoc map descriptor"); MALLOC_DEFINE(SCTP_M_STRMI, "sctp_stri", "sctp stream in array"); MALLOC_DEFINE(SCTP_M_STRMO, "sctp_stro", "sctp stream out array"); MALLOC_DEFINE(SCTP_M_ASC_ADDR, "sctp_aadr", "sctp asconf address"); MALLOC_DEFINE(SCTP_M_ASC_IT, "sctp_a_it", "sctp asconf iterator"); MALLOC_DEFINE(SCTP_M_AUTH_CL, "sctp_atcl", "sctp auth chunklist"); MALLOC_DEFINE(SCTP_M_AUTH_KY, "sctp_atky", "sctp auth key"); MALLOC_DEFINE(SCTP_M_AUTH_HL, "sctp_athm", "sctp auth hmac list"); MALLOC_DEFINE(SCTP_M_AUTH_IF, "sctp_athi", "sctp auth info"); MALLOC_DEFINE(SCTP_M_STRESET, "sctp_stre", "sctp stream reset"); MALLOC_DEFINE(SCTP_M_CMSG, "sctp_cmsg", "sctp CMSG buffer"); MALLOC_DEFINE(SCTP_M_COPYAL, "sctp_cpal", "sctp copy all"); MALLOC_DEFINE(SCTP_M_VRF, "sctp_vrf", "sctp vrf struct"); MALLOC_DEFINE(SCTP_M_IFA, "sctp_ifa", "sctp ifa struct"); MALLOC_DEFINE(SCTP_M_IFN, "sctp_ifn", "sctp ifn struct"); MALLOC_DEFINE(SCTP_M_TIMW, "sctp_timw", "sctp time block"); MALLOC_DEFINE(SCTP_M_MVRF, "sctp_mvrf", "sctp mvrf pcb list"); MALLOC_DEFINE(SCTP_M_ITER, "sctp_iter", "sctp iterator control"); MALLOC_DEFINE(SCTP_M_SOCKOPT, "sctp_socko", "sctp socket option"); MALLOC_DEFINE(SCTP_M_MCORE, "sctp_mcore", "sctp mcore queue"); /* Global NON-VNET structure that controls the iterator */ struct iterator_control sctp_it_ctl; void sctp_wakeup_iterator(void) { wakeup(&sctp_it_ctl.iterator_running); } static void sctp_iterator_thread(void *v SCTP_UNUSED) { SCTP_IPI_ITERATOR_WQ_LOCK(); /* In FreeBSD this thread never terminates. */ for (;;) { msleep(&sctp_it_ctl.iterator_running, &sctp_it_ctl.ipi_iterator_wq_mtx, 0, "waiting_for_work", 0); sctp_iterator_worker(); } } void sctp_startup_iterator(void) { if (sctp_it_ctl.thread_proc) { /* You only get one */ return; } /* Initialize global locks here, thus only once. */ SCTP_ITERATOR_LOCK_INIT(); SCTP_IPI_ITERATOR_WQ_INIT(); TAILQ_INIT(&sctp_it_ctl.iteratorhead); kproc_create(sctp_iterator_thread, (void *)NULL, &sctp_it_ctl.thread_proc, RFPROC, SCTP_KTHREAD_PAGES, SCTP_KTRHEAD_NAME); } #ifdef INET6 void sctp_gather_internal_ifa_flags(struct sctp_ifa *ifa) { struct in6_ifaddr *ifa6; ifa6 = (struct in6_ifaddr *)ifa->ifa; ifa->flags = ifa6->ia6_flags; if (!MODULE_GLOBAL(ip6_use_deprecated)) { if (ifa->flags & IN6_IFF_DEPRECATED) { ifa->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE; } else { ifa->localifa_flags &= ~SCTP_ADDR_IFA_UNUSEABLE; } } else { ifa->localifa_flags &= ~SCTP_ADDR_IFA_UNUSEABLE; } if (ifa->flags & (IN6_IFF_DETACHED | IN6_IFF_ANYCAST | IN6_IFF_NOTREADY)) { ifa->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE; } else { ifa->localifa_flags &= ~SCTP_ADDR_IFA_UNUSEABLE; } } #endif /* INET6 */ static uint32_t sctp_is_desired_interface_type(struct ifnet *ifn) { int result; /* check the interface type to see if it's one we care about */ switch (ifn->if_type) { case IFT_ETHER: case IFT_ISO88023: case IFT_ISO88024: case IFT_ISO88025: case IFT_ISO88026: case IFT_STARLAN: case IFT_P10: case IFT_P80: case IFT_HY: case IFT_FDDI: case IFT_XETHER: case IFT_ISDNBASIC: case IFT_ISDNPRIMARY: case IFT_PTPSERIAL: case IFT_OTHER: case IFT_PPP: case IFT_LOOP: case IFT_SLIP: case IFT_GIF: case IFT_L2VLAN: case IFT_STF: case IFT_IP: case IFT_IPOVERCDLC: case IFT_IPOVERCLAW: case IFT_PROPVIRTUAL: /* NetGraph Virtual too */ case IFT_VIRTUALIPADDRESS: result = 1; break; default: result = 0; } return (result); } static void sctp_init_ifns_for_vrf(int vrfid) { /* * Here we must apply ANY locks needed by the IFN we access and also * make sure we lock any IFA that exists as we float through the * list of IFA's */ struct ifnet *ifn; struct ifaddr *ifa; struct sctp_ifa *sctp_ifa; uint32_t ifa_flags; #ifdef INET6 struct in6_ifaddr *ifa6; #endif IFNET_RLOCK(); TAILQ_FOREACH(ifn, &MODULE_GLOBAL(ifnet), if_list) { if (sctp_is_desired_interface_type(ifn) == 0) { /* non desired type */ continue; } IF_ADDR_RLOCK(ifn); TAILQ_FOREACH(ifa, &ifn->if_addrlist, ifa_list) { if (ifa->ifa_addr == NULL) { continue; } switch (ifa->ifa_addr->sa_family) { #ifdef INET case AF_INET: if (((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr == 0) { continue; } break; #endif #ifdef INET6 case AF_INET6: if (IN6_IS_ADDR_UNSPECIFIED(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr)) { /* skip unspecifed addresses */ continue; } break; #endif default: continue; } switch (ifa->ifa_addr->sa_family) { #ifdef INET case AF_INET: ifa_flags = 0; break; #endif #ifdef INET6 case AF_INET6: ifa6 = (struct in6_ifaddr *)ifa; ifa_flags = ifa6->ia6_flags; break; #endif default: ifa_flags = 0; break; } sctp_ifa = sctp_add_addr_to_vrf(vrfid, (void *)ifn, ifn->if_index, ifn->if_type, ifn->if_xname, (void *)ifa, ifa->ifa_addr, ifa_flags, 0); if (sctp_ifa) { sctp_ifa->localifa_flags &= ~SCTP_ADDR_DEFER_USE; } } IF_ADDR_RUNLOCK(ifn); } IFNET_RUNLOCK(); } void sctp_init_vrf_list(int vrfid) { if (vrfid > SCTP_MAX_VRF_ID) /* can't do that */ return; /* Don't care about return here */ (void)sctp_allocate_vrf(vrfid); /* * Now we need to build all the ifn's for this vrf and there * addresses */ sctp_init_ifns_for_vrf(vrfid); } void sctp_addr_change(struct ifaddr *ifa, int cmd) { uint32_t ifa_flags = 0; + if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) + return; + /* * BSD only has one VRF, if this changes we will need to hook in the * right things here to get the id to pass to the address managment * routine. */ if (SCTP_BASE_VAR(first_time) == 0) { /* Special test to see if my ::1 will showup with this */ SCTP_BASE_VAR(first_time) = 1; sctp_init_ifns_for_vrf(SCTP_DEFAULT_VRFID); } if ((cmd != RTM_ADD) && (cmd != RTM_DELETE)) { /* don't know what to do with this */ return; } if (ifa->ifa_addr == NULL) { return; } if (sctp_is_desired_interface_type(ifa->ifa_ifp) == 0) { /* non desired type */ return; } switch (ifa->ifa_addr->sa_family) { #ifdef INET case AF_INET: if (((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr == 0) { return; } break; #endif #ifdef INET6 case AF_INET6: ifa_flags = ((struct in6_ifaddr *)ifa)->ia6_flags; if (IN6_IS_ADDR_UNSPECIFIED(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr)) { /* skip unspecifed addresses */ return; } break; #endif default: /* non inet/inet6 skip */ return; } if (cmd == RTM_ADD) { (void)sctp_add_addr_to_vrf(SCTP_DEFAULT_VRFID, (void *)ifa->ifa_ifp, ifa->ifa_ifp->if_index, ifa->ifa_ifp->if_type, ifa->ifa_ifp->if_xname, (void *)ifa, ifa->ifa_addr, ifa_flags, 1); } else { sctp_del_addr_from_vrf(SCTP_DEFAULT_VRFID, ifa->ifa_addr, ifa->ifa_ifp->if_index, ifa->ifa_ifp->if_xname); /* * We don't bump refcount here so when it completes the * final delete will happen. */ } } void sctp_add_or_del_interfaces(int (*pred) (struct ifnet *), int add){ struct ifnet *ifn; struct ifaddr *ifa; IFNET_RLOCK(); TAILQ_FOREACH(ifn, &MODULE_GLOBAL(ifnet), if_list) { if (!(*pred) (ifn)) { continue; } TAILQ_FOREACH(ifa, &ifn->if_addrlist, ifa_list) { sctp_addr_change(ifa, add ? RTM_ADD : RTM_DELETE); } } IFNET_RUNLOCK(); } struct mbuf * sctp_get_mbuf_for_msg(unsigned int space_needed, int want_header, int how, int allonebuf, int type) { struct mbuf *m = NULL; m = m_getm2(NULL, space_needed, how, type, want_header ? M_PKTHDR : 0); if (m == NULL) { /* bad, no memory */ return (m); } if (allonebuf) { if (SCTP_BUF_SIZE(m) < space_needed) { m_freem(m); return (NULL); } } if (SCTP_BUF_NEXT(m)) { sctp_m_freem(SCTP_BUF_NEXT(m)); SCTP_BUF_NEXT(m) = NULL; } #ifdef SCTP_MBUF_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MBUF_LOGGING_ENABLE) { sctp_log_mb(m, SCTP_MBUF_IALLOC); } #endif return (m); } #ifdef SCTP_PACKET_LOGGING void sctp_packet_log(struct mbuf *m) { int *lenat, thisone; void *copyto; uint32_t *tick_tock; int length; int total_len; int grabbed_lock = 0; int value, newval, thisend, thisbegin; /* * Buffer layout. -sizeof this entry (total_len) -previous end * (value) -ticks of log (ticks) o -ip packet o -as logged - * where this started (thisbegin) x <--end points here */ length = SCTP_HEADER_LEN(m); total_len = SCTP_SIZE32((length + (4 * sizeof(int)))); /* Log a packet to the buffer. */ if (total_len > SCTP_PACKET_LOG_SIZE) { /* Can't log this packet I have not a buffer big enough */ return; } if (length < (int)(SCTP_MIN_V4_OVERHEAD + sizeof(struct sctp_cookie_ack_chunk))) { return; } atomic_add_int(&SCTP_BASE_VAR(packet_log_writers), 1); try_again: if (SCTP_BASE_VAR(packet_log_writers) > SCTP_PKTLOG_WRITERS_NEED_LOCK) { SCTP_IP_PKTLOG_LOCK(); grabbed_lock = 1; again_locked: value = SCTP_BASE_VAR(packet_log_end); newval = SCTP_BASE_VAR(packet_log_end) + total_len; if (newval >= SCTP_PACKET_LOG_SIZE) { /* we wrapped */ thisbegin = 0; thisend = total_len; } else { thisbegin = SCTP_BASE_VAR(packet_log_end); thisend = newval; } if (!(atomic_cmpset_int(&SCTP_BASE_VAR(packet_log_end), value, thisend))) { goto again_locked; } } else { value = SCTP_BASE_VAR(packet_log_end); newval = SCTP_BASE_VAR(packet_log_end) + total_len; if (newval >= SCTP_PACKET_LOG_SIZE) { /* we wrapped */ thisbegin = 0; thisend = total_len; } else { thisbegin = SCTP_BASE_VAR(packet_log_end); thisend = newval; } if (!(atomic_cmpset_int(&SCTP_BASE_VAR(packet_log_end), value, thisend))) { goto try_again; } } /* Sanity check */ if (thisend >= SCTP_PACKET_LOG_SIZE) { SCTP_PRINTF("Insanity stops a log thisbegin:%d thisend:%d writers:%d lock:%d end:%d\n", thisbegin, thisend, SCTP_BASE_VAR(packet_log_writers), grabbed_lock, SCTP_BASE_VAR(packet_log_end)); SCTP_BASE_VAR(packet_log_end) = 0; goto no_log; } lenat = (int *)&SCTP_BASE_VAR(packet_log_buffer)[thisbegin]; *lenat = total_len; lenat++; *lenat = value; lenat++; tick_tock = (uint32_t *) lenat; lenat++; *tick_tock = sctp_get_tick_count(); copyto = (void *)lenat; thisone = thisend - sizeof(int); lenat = (int *)&SCTP_BASE_VAR(packet_log_buffer)[thisone]; *lenat = thisbegin; if (grabbed_lock) { SCTP_IP_PKTLOG_UNLOCK(); grabbed_lock = 0; } m_copydata(m, 0, length, (caddr_t)copyto); no_log: if (grabbed_lock) { SCTP_IP_PKTLOG_UNLOCK(); } atomic_subtract_int(&SCTP_BASE_VAR(packet_log_writers), 1); } int sctp_copy_out_packet_log(uint8_t * target, int length) { /* * We wind through the packet log starting at start copying up to * length bytes out. We return the number of bytes copied. */ int tocopy, this_copy; int *lenat; int did_delay = 0; tocopy = length; if (length < (int)(2 * sizeof(int))) { /* not enough room */ return (0); } if (SCTP_PKTLOG_WRITERS_NEED_LOCK) { atomic_add_int(&SCTP_BASE_VAR(packet_log_writers), SCTP_PKTLOG_WRITERS_NEED_LOCK); again: if ((did_delay == 0) && (SCTP_BASE_VAR(packet_log_writers) != SCTP_PKTLOG_WRITERS_NEED_LOCK)) { /* * we delay here for just a moment hoping the * writer(s) that were present when we entered will * have left and we only have locking ones that will * contend with us for the lock. This does not * assure 100% access, but its good enough for a * logging facility like this. */ did_delay = 1; DELAY(10); goto again; } } SCTP_IP_PKTLOG_LOCK(); lenat = (int *)target; *lenat = SCTP_BASE_VAR(packet_log_end); lenat++; this_copy = min((length - sizeof(int)), SCTP_PACKET_LOG_SIZE); memcpy((void *)lenat, (void *)SCTP_BASE_VAR(packet_log_buffer), this_copy); if (SCTP_PKTLOG_WRITERS_NEED_LOCK) { atomic_subtract_int(&SCTP_BASE_VAR(packet_log_writers), SCTP_PKTLOG_WRITERS_NEED_LOCK); } SCTP_IP_PKTLOG_UNLOCK(); return (this_copy + sizeof(int)); } #endif Index: projects/vnet/sys/netinet/sctp_pcb.c =================================================================== --- projects/vnet/sys/netinet/sctp_pcb.c (revision 294565) +++ projects/vnet/sys/netinet/sctp_pcb.c (revision 294566) @@ -1,7045 +1,7067 @@ /*- * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #if defined(INET) || defined(INET6) #include #endif #ifdef INET6 #include #endif #include #include #include VNET_DEFINE(struct sctp_base_info, system_base_info); /* FIX: we don't handle multiple link local scopes */ /* "scopeless" replacement IN6_ARE_ADDR_EQUAL */ #ifdef INET6 int SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b) { struct sockaddr_in6 tmp_a, tmp_b; memcpy(&tmp_a, a, sizeof(struct sockaddr_in6)); if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) { return (0); } memcpy(&tmp_b, b, sizeof(struct sockaddr_in6)); if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) { return (0); } return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr)); } #endif void sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb) { /* * We really don't need to lock this, but I will just because it * does not hurt. */ SCTP_INP_INFO_RLOCK(); spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep); spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc); spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr); spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr); spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk); spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq); spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq); spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks); SCTP_INP_INFO_RUNLOCK(); } /*- * Addresses are added to VRF's (Virtual Router's). For BSD we * have only the default VRF 0. We maintain a hash list of * VRF's. Each VRF has its own list of sctp_ifn's. Each of * these has a list of addresses. When we add a new address * to a VRF we lookup the ifn/ifn_index, if the ifn does * not exist we create it and add it to the list of IFN's * within the VRF. Once we have the sctp_ifn, we add the * address to the list. So we look something like: * * hash-vrf-table * vrf-> ifn-> ifn -> ifn * vrf | * ... +--ifa-> ifa -> ifa * vrf * * We keep these separate lists since the SCTP subsystem will * point to these from its source address selection nets structure. * When an address is deleted it does not happen right away on * the SCTP side, it gets scheduled. What we do when a * delete happens is immediately remove the address from * the master list and decrement the refcount. As our * addip iterator works through and frees the src address * selection pointing to the sctp_ifa, eventually the refcount * will reach 0 and we will delete it. Note that it is assumed * that any locking on system level ifn/ifa is done at the * caller of these functions and these routines will only * lock the SCTP structures as they add or delete things. * * Other notes on VRF concepts. * - An endpoint can be in multiple VRF's * - An association lives within a VRF and only one VRF. * - Any incoming packet we can deduce the VRF for by * looking at the mbuf/pak inbound (for BSD its VRF=0 :D) * - Any downward send call or connect call must supply the * VRF via ancillary data or via some sort of set default * VRF socket option call (again for BSD no brainer since * the VRF is always 0). * - An endpoint may add multiple VRF's to it. * - Listening sockets can accept associations in any * of the VRF's they are in but the assoc will end up * in only one VRF (gotten from the packet or connect/send). * */ struct sctp_vrf * sctp_allocate_vrf(int vrf_id) { struct sctp_vrf *vrf = NULL; struct sctp_vrflist *bucket; /* First allocate the VRF structure */ vrf = sctp_find_vrf(vrf_id); if (vrf) { /* Already allocated */ return (vrf); } SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf), SCTP_M_VRF); if (vrf == NULL) { /* No memory */ #ifdef INVARIANTS panic("No memory for VRF:%d", vrf_id); #endif return (NULL); } /* setup the VRF */ memset(vrf, 0, sizeof(struct sctp_vrf)); vrf->vrf_id = vrf_id; LIST_INIT(&vrf->ifnlist); vrf->total_ifa_count = 0; vrf->refcount = 0; /* now also setup table ids */ SCTP_INIT_VRF_TABLEID(vrf); /* Init the HASH of addresses */ vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE, &vrf->vrf_addr_hashmark); if (vrf->vrf_addr_hash == NULL) { /* No memory */ #ifdef INVARIANTS panic("No memory for VRF:%d", vrf_id); #endif SCTP_FREE(vrf, SCTP_M_VRF); return (NULL); } /* Add it to the hash table */ bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))]; LIST_INSERT_HEAD(bucket, vrf, next_vrf); atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1); return (vrf); } struct sctp_ifn * sctp_find_ifn(void *ifn, uint32_t ifn_index) { struct sctp_ifn *sctp_ifnp; struct sctp_ifnlist *hash_ifn_head; /* * We assume the lock is held for the addresses if that's wrong * problems could occur :-) */ hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))]; LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) { if (sctp_ifnp->ifn_index == ifn_index) { return (sctp_ifnp); } if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) { return (sctp_ifnp); } } return (NULL); } struct sctp_vrf * sctp_find_vrf(uint32_t vrf_id) { struct sctp_vrflist *bucket; struct sctp_vrf *liste; bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))]; LIST_FOREACH(liste, bucket, next_vrf) { if (vrf_id == liste->vrf_id) { return (liste); } } return (NULL); } void sctp_free_vrf(struct sctp_vrf *vrf) { if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) { if (vrf->vrf_addr_hash) { SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark); vrf->vrf_addr_hash = NULL; } /* We zero'd the count */ LIST_REMOVE(vrf, next_vrf); SCTP_FREE(vrf, SCTP_M_VRF); atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1); } } void sctp_free_ifn(struct sctp_ifn *sctp_ifnp) { if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) { /* We zero'd the count */ if (sctp_ifnp->vrf) { sctp_free_vrf(sctp_ifnp->vrf); } SCTP_FREE(sctp_ifnp, SCTP_M_IFN); atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1); } } void sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu) { struct sctp_ifn *sctp_ifnp; sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index); if (sctp_ifnp != NULL) { sctp_ifnp->ifn_mtu = mtu; } } void sctp_free_ifa(struct sctp_ifa *sctp_ifap) { if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) { /* We zero'd the count */ if (sctp_ifap->ifn_p) { sctp_free_ifn(sctp_ifap->ifn_p); } SCTP_FREE(sctp_ifap, SCTP_M_IFA); atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1); } } static void sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock) { struct sctp_ifn *found; found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index); if (found == NULL) { /* Not in the list.. sorry */ return; } if (hold_addr_lock == 0) SCTP_IPI_ADDR_WLOCK(); LIST_REMOVE(sctp_ifnp, next_bucket); LIST_REMOVE(sctp_ifnp, next_ifn); SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index, sctp_ifnp->registered_af); if (hold_addr_lock == 0) SCTP_IPI_ADDR_WUNLOCK(); /* Take away the reference, and possibly free it */ sctp_free_ifn(sctp_ifnp); } void sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr, const char *if_name, uint32_t ifn_index) { struct sctp_vrf *vrf; struct sctp_ifa *sctp_ifap; SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); goto out; } sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n"); goto out; } if (sctp_ifap->ifn_p == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n"); goto out; } if (if_name) { if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) { SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n", sctp_ifap->ifn_p->ifn_name, if_name); goto out; } } else { if (sctp_ifap->ifn_p->ifn_index != ifn_index) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n", sctp_ifap->ifn_p->ifn_index, ifn_index); goto out; } } sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID); sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE; out: SCTP_IPI_ADDR_RUNLOCK(); } void sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr, const char *if_name, uint32_t ifn_index) { struct sctp_vrf *vrf; struct sctp_ifa *sctp_ifap; SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); goto out; } sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n"); goto out; } if (sctp_ifap->ifn_p == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n"); goto out; } if (if_name) { if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) { SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n", sctp_ifap->ifn_p->ifn_name, if_name); goto out; } } else { if (sctp_ifap->ifn_p->ifn_index != ifn_index) { SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n", sctp_ifap->ifn_p->ifn_index, ifn_index); goto out; } } sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE); sctp_ifap->localifa_flags |= SCTP_ADDR_VALID; out: SCTP_IPI_ADDR_RUNLOCK(); } /*- * Add an ifa to an ifn. * Register the interface as necessary. * NOTE: ADDR write lock MUST be held. */ static void sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap) { int ifa_af; LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa); sctp_ifap->ifn_p = sctp_ifnp; atomic_add_int(&sctp_ifap->ifn_p->refcount, 1); /* update address counts */ sctp_ifnp->ifa_count++; ifa_af = sctp_ifap->address.sa.sa_family; switch (ifa_af) { #ifdef INET case AF_INET: sctp_ifnp->num_v4++; break; #endif #ifdef INET6 case AF_INET6: sctp_ifnp->num_v6++; break; #endif default: break; } if (sctp_ifnp->ifa_count == 1) { /* register the new interface */ SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af); sctp_ifnp->registered_af = ifa_af; } } /*- * Remove an ifa from its ifn. * If no more addresses exist, remove the ifn too. Otherwise, re-register * the interface based on the remaining address families left. * NOTE: ADDR write lock MUST be held. */ static void sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap) { LIST_REMOVE(sctp_ifap, next_ifa); if (sctp_ifap->ifn_p) { /* update address counts */ sctp_ifap->ifn_p->ifa_count--; switch (sctp_ifap->address.sa.sa_family) { #ifdef INET case AF_INET: sctp_ifap->ifn_p->num_v4--; break; #endif #ifdef INET6 case AF_INET6: sctp_ifap->ifn_p->num_v6--; break; #endif default: break; } if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) { /* remove the ifn, possibly freeing it */ sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED); } else { /* re-register address family type, if needed */ if ((sctp_ifap->ifn_p->num_v6 == 0) && (sctp_ifap->ifn_p->registered_af == AF_INET6)) { SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6); SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET); sctp_ifap->ifn_p->registered_af = AF_INET; } else if ((sctp_ifap->ifn_p->num_v4 == 0) && (sctp_ifap->ifn_p->registered_af == AF_INET)) { SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET); SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6); sctp_ifap->ifn_p->registered_af = AF_INET6; } /* free the ifn refcount */ sctp_free_ifn(sctp_ifap->ifn_p); } sctp_ifap->ifn_p = NULL; } } struct sctp_ifa * sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index, uint32_t ifn_type, const char *if_name, void *ifa, struct sockaddr *addr, uint32_t ifa_flags, int dynamic_add) { struct sctp_vrf *vrf; struct sctp_ifn *sctp_ifnp = NULL; struct sctp_ifa *sctp_ifap = NULL; struct sctp_ifalist *hash_addr_head; struct sctp_ifnlist *hash_ifn_head; uint32_t hash_of_addr; int new_ifn_af = 0; #ifdef SCTP_DEBUG SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id); SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr); #endif SCTP_IPI_ADDR_WLOCK(); sctp_ifnp = sctp_find_ifn(ifn, ifn_index); if (sctp_ifnp) { vrf = sctp_ifnp->vrf; } else { vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { vrf = sctp_allocate_vrf(vrf_id); if (vrf == NULL) { SCTP_IPI_ADDR_WUNLOCK(); return (NULL); } } } if (sctp_ifnp == NULL) { /* * build one and add it, can't hold lock until after malloc * done though. */ SCTP_IPI_ADDR_WUNLOCK(); SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *, sizeof(struct sctp_ifn), SCTP_M_IFN); if (sctp_ifnp == NULL) { #ifdef INVARIANTS panic("No memory for IFN"); #endif return (NULL); } memset(sctp_ifnp, 0, sizeof(struct sctp_ifn)); sctp_ifnp->ifn_index = ifn_index; sctp_ifnp->ifn_p = ifn; sctp_ifnp->ifn_type = ifn_type; sctp_ifnp->refcount = 0; sctp_ifnp->vrf = vrf; atomic_add_int(&vrf->refcount, 1); sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family); if (if_name != NULL) { snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name); } else { snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown"); } hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))]; LIST_INIT(&sctp_ifnp->ifalist); SCTP_IPI_ADDR_WLOCK(); LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket); LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn); atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1); new_ifn_af = 1; } sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap) { /* Hmm, it already exists? */ if ((sctp_ifap->ifn_p) && (sctp_ifap->ifn_p->ifn_index == ifn_index)) { SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n", sctp_ifap->ifn_p->ifn_name, ifn_index, (void *)sctp_ifap); if (new_ifn_af) { /* Remove the created one that we don't want */ sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED); } if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) { /* easy to solve, just switch back to active */ SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n"); sctp_ifap->localifa_flags = SCTP_ADDR_VALID; sctp_ifap->ifn_p = sctp_ifnp; atomic_add_int(&sctp_ifap->ifn_p->refcount, 1); } exit_stage_left: SCTP_IPI_ADDR_WUNLOCK(); return (sctp_ifap); } else { if (sctp_ifap->ifn_p) { /* * The last IFN gets the address, remove the * old one */ SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n", (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name, sctp_ifap->ifn_p->ifn_index, if_name, ifn_index); /* remove the address from the old ifn */ sctp_remove_ifa_from_ifn(sctp_ifap); /* move the address over to the new ifn */ sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap); goto exit_stage_left; } else { /* repair ifnp which was NULL ? */ sctp_ifap->localifa_flags = SCTP_ADDR_VALID; SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n", (void *)sctp_ifnp, (void *)sctp_ifap); sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap); } goto exit_stage_left; } } SCTP_IPI_ADDR_WUNLOCK(); SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA); if (sctp_ifap == NULL) { #ifdef INVARIANTS panic("No memory for IFA"); #endif return (NULL); } memset(sctp_ifap, 0, sizeof(struct sctp_ifa)); sctp_ifap->ifn_p = sctp_ifnp; atomic_add_int(&sctp_ifnp->refcount, 1); sctp_ifap->vrf_id = vrf_id; sctp_ifap->ifa = ifa; memcpy(&sctp_ifap->address, addr, addr->sa_len); sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE; sctp_ifap->flags = ifa_flags; /* Set scope */ switch (sctp_ifap->address.sa.sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin; sin = &sctp_ifap->address.sin; if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) || (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) { sctp_ifap->src_is_loop = 1; } if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { sctp_ifap->src_is_priv = 1; } sctp_ifnp->num_v4++; if (new_ifn_af) new_ifn_af = AF_INET; break; } #endif #ifdef INET6 case AF_INET6: { /* ok to use deprecated addresses? */ struct sockaddr_in6 *sin6; sin6 = &sctp_ifap->address.sin6; if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) || (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) { sctp_ifap->src_is_loop = 1; } if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { sctp_ifap->src_is_priv = 1; } sctp_ifnp->num_v6++; if (new_ifn_af) new_ifn_af = AF_INET6; break; } #endif default: new_ifn_af = 0; break; } hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa); if ((sctp_ifap->src_is_priv == 0) && (sctp_ifap->src_is_loop == 0)) { sctp_ifap->src_is_glob = 1; } SCTP_IPI_ADDR_WLOCK(); hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)]; LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket); sctp_ifap->refcount = 1; LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa); sctp_ifnp->ifa_count++; vrf->total_ifa_count++; atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1); if (new_ifn_af) { SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af); sctp_ifnp->registered_af = new_ifn_af; } SCTP_IPI_ADDR_WUNLOCK(); if (dynamic_add) { /* * Bump up the refcount so that when the timer completes it * will drop back down. */ struct sctp_laddr *wi; atomic_add_int(&sctp_ifap->refcount, 1); wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (wi == NULL) { /* * Gak, what can we do? We have lost an address * change can you say HOSED? */ SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n"); /* Opps, must decrement the count */ sctp_del_addr_from_vrf(vrf_id, addr, ifn_index, if_name); return (NULL); } SCTP_INCR_LADDR_COUNT(); bzero(wi, sizeof(*wi)); (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); wi->ifa = sctp_ifap; wi->action = SCTP_ADD_IP_ADDRESS; SCTP_WQ_ADDR_LOCK(); LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); SCTP_WQ_ADDR_UNLOCK(); sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, (struct sctp_inpcb *)NULL, (struct sctp_tcb *)NULL, (struct sctp_nets *)NULL); } else { /* it's ready for use */ sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE; } return (sctp_ifap); } void sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr, uint32_t ifn_index, const char *if_name) { struct sctp_vrf *vrf; struct sctp_ifa *sctp_ifap = NULL; SCTP_IPI_ADDR_WLOCK(); vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id); goto out_now; } #ifdef SCTP_DEBUG SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id); SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr); #endif sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED); if (sctp_ifap) { /* Validate the delete */ if (sctp_ifap->ifn_p) { int valid = 0; /*- * The name has priority over the ifn_index * if its given. We do this especially for * panda who might recycle indexes fast. */ if (if_name) { if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) { /* They match its a correct delete */ valid = 1; } } if (!valid) { /* last ditch check ifn_index */ if (ifn_index == sctp_ifap->ifn_p->ifn_index) { valid = 1; } } if (!valid) { SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n", ifn_index, ((if_name == NULL) ? "NULL" : if_name)); SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n", sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name); SCTP_IPI_ADDR_WUNLOCK(); return; } } SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap); sctp_ifap->localifa_flags &= SCTP_ADDR_VALID; /* * We don't set the flag. This means that the structure will * hang around in EP's that have bound specific to it until * they close. This gives us TCP like behavior if someone * removes an address (or for that matter adds it right * back). */ /* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */ vrf->total_ifa_count--; LIST_REMOVE(sctp_ifap, next_bucket); sctp_remove_ifa_from_ifn(sctp_ifap); } #ifdef SCTP_DEBUG else { SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:", ifn_index); SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr); } #endif out_now: SCTP_IPI_ADDR_WUNLOCK(); if (sctp_ifap) { struct sctp_laddr *wi; wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (wi == NULL) { /* * Gak, what can we do? We have lost an address * change can you say HOSED? */ SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n"); /* Oops, must decrement the count */ sctp_free_ifa(sctp_ifap); return; } SCTP_INCR_LADDR_COUNT(); bzero(wi, sizeof(*wi)); (void)SCTP_GETTIME_TIMEVAL(&wi->start_time); wi->ifa = sctp_ifap; wi->action = SCTP_DEL_IP_ADDRESS; SCTP_WQ_ADDR_LOCK(); /* * Should this really be a tailq? As it is we will process * the newest first :-0 */ LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr); SCTP_WQ_ADDR_UNLOCK(); sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ, (struct sctp_inpcb *)NULL, (struct sctp_tcb *)NULL, (struct sctp_nets *)NULL); } return; } static int sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to) { int loopback_scope; #if defined(INET) int ipv4_local_scope, ipv4_addr_legal; #endif #if defined(INET6) int local_scope, site_scope, ipv6_addr_legal; #endif struct sctp_vrf *vrf; struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa; loopback_scope = stcb->asoc.scope.loopback_scope; #if defined(INET) ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope; ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal; #endif #if defined(INET6) local_scope = stcb->asoc.scope.local_scope; site_scope = stcb->asoc.scope.site_scope; ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal; #endif SCTP_IPI_ADDR_RLOCK(); vrf = sctp_find_vrf(stcb->asoc.vrf_id); if (vrf == NULL) { /* no vrf, no addresses */ SCTP_IPI_ADDR_RUNLOCK(); return (0); } if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { if ((loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { continue; } LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if (sctp_is_addr_restricted(stcb, sctp_ifa) && (!sctp_is_addr_pending(stcb, sctp_ifa))) { /* * We allow pending addresses, where * we have sent an asconf-add to be * considered valid. */ continue; } if (sctp_ifa->address.sa.sa_family != to->sa_family) { continue; } switch (sctp_ifa->address.sa.sa_family) { #ifdef INET case AF_INET: if (ipv4_addr_legal) { struct sockaddr_in *sin, *rsin; sin = &sctp_ifa->address.sin; rsin = (struct sockaddr_in *)to; if ((ipv4_local_scope == 0) && IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { continue; } if (prison_check_ip4(stcb->sctp_ep->ip_inp.inp.inp_cred, &sin->sin_addr) != 0) { continue; } if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { SCTP_IPI_ADDR_RUNLOCK(); return (1); } } break; #endif #ifdef INET6 case AF_INET6: if (ipv6_addr_legal) { struct sockaddr_in6 *sin6, *rsin6; sin6 = &sctp_ifa->address.sin6; rsin6 = (struct sockaddr_in6 *)to; if (prison_check_ip6(stcb->sctp_ep->ip_inp.inp.inp_cred, &sin6->sin6_addr) != 0) { continue; } if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { if (local_scope == 0) continue; if (sin6->sin6_scope_id == 0) { if (sa6_recoverscope(sin6) != 0) continue; } } if ((site_scope == 0) && (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { continue; } if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { SCTP_IPI_ADDR_RUNLOCK(); return (1); } } break; #endif default: /* TSNH */ break; } } } } else { struct sctp_laddr *laddr; LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n"); continue; } if (sctp_is_addr_restricted(stcb, laddr->ifa) && (!sctp_is_addr_pending(stcb, laddr->ifa))) { /* * We allow pending addresses, where we have * sent an asconf-add to be considered * valid. */ continue; } if (laddr->ifa->address.sa.sa_family != to->sa_family) { continue; } switch (to->sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin, *rsin; sin = &laddr->ifa->address.sin; rsin = (struct sockaddr_in *)to; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { SCTP_IPI_ADDR_RUNLOCK(); return (1); } break; } #endif #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = &laddr->ifa->address.sin6; rsin6 = (struct sockaddr_in6 *)to; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { SCTP_IPI_ADDR_RUNLOCK(); return (1); } break; } #endif default: /* TSNH */ break; } } } SCTP_IPI_ADDR_RUNLOCK(); return (0); } static struct sctp_tcb * sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from, struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id) { /**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */ /* * If we support the TCP model, then we must now dig through to see * if we can find our endpoint in the list of tcp ep's. */ uint16_t lport, rport; struct sctppcbhead *ephead; struct sctp_inpcb *inp; struct sctp_laddr *laddr; struct sctp_tcb *stcb; struct sctp_nets *net; if ((to == NULL) || (from == NULL)) { return (NULL); } switch (to->sa_family) { #ifdef INET case AF_INET: if (from->sa_family == AF_INET) { lport = ((struct sockaddr_in *)to)->sin_port; rport = ((struct sockaddr_in *)from)->sin_port; } else { return (NULL); } break; #endif #ifdef INET6 case AF_INET6: if (from->sa_family == AF_INET6) { lport = ((struct sockaddr_in6 *)to)->sin6_port; rport = ((struct sockaddr_in6 *)from)->sin6_port; } else { return (NULL); } break; #endif default: return (NULL); } ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))]; /* * Ok now for each of the guys in this bucket we must look and see: * - Does the remote port match. - Does there single association's * addresses match this address (to). If so we update p_ep to point * to this ep and return the tcb from it. */ LIST_FOREACH(inp, ephead, sctp_hash) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(inp); continue; } if (lport != inp->sctp_lport) { SCTP_INP_RUNLOCK(inp); continue; } switch (to->sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin; sin = (struct sockaddr_in *)to; if (prison_check_ip4(inp->ip_inp.inp.inp_cred, &sin->sin_addr) != 0) { SCTP_INP_RUNLOCK(inp); continue; } break; } #endif #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)to; if (prison_check_ip6(inp->ip_inp.inp.inp_cred, &sin6->sin6_addr) != 0) { SCTP_INP_RUNLOCK(inp); continue; } break; } #endif default: SCTP_INP_RUNLOCK(inp); continue; } if (inp->def_vrf_id != vrf_id) { SCTP_INP_RUNLOCK(inp); continue; } /* check to see if the ep has one of the addresses */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { /* We are NOT bound all, so look further */ int match = 0; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __func__); continue; } if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n"); continue; } if (laddr->ifa->address.sa.sa_family == to->sa_family) { /* see if it matches */ #ifdef INET if (from->sa_family == AF_INET) { struct sockaddr_in *intf_addr, *sin; intf_addr = &laddr->ifa->address.sin; sin = (struct sockaddr_in *)to; if (sin->sin_addr.s_addr == intf_addr->sin_addr.s_addr) { match = 1; break; } } #endif #ifdef INET6 if (from->sa_family == AF_INET6) { struct sockaddr_in6 *intf_addr6; struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *) to; intf_addr6 = &laddr->ifa->address.sin6; if (SCTP6_ARE_ADDR_EQUAL(sin6, intf_addr6)) { match = 1; break; } } #endif } } if (match == 0) { /* This endpoint does not have this address */ SCTP_INP_RUNLOCK(inp); continue; } } /* * Ok if we hit here the ep has the address, does it hold * the tcb? */ /* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */ stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { SCTP_INP_RUNLOCK(inp); continue; } SCTP_TCB_LOCK(stcb); if (!sctp_does_stcb_own_this_addr(stcb, to)) { SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); continue; } if (stcb->rport != rport) { /* remote port does not match. */ SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); continue; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); continue; } if (!sctp_does_stcb_own_this_addr(stcb, to)) { SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); continue; } /* Does this TCB have a matching address? */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (net->ro._l_addr.sa.sa_family != from->sa_family) { /* not the same family, can't be a match */ continue; } switch (from->sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin, *rsin; sin = (struct sockaddr_in *)&net->ro._l_addr; rsin = (struct sockaddr_in *)from; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { /* found it */ if (netp != NULL) { *netp = net; } /* * Update the endpoint * pointer */ *inp_p = inp; SCTP_INP_RUNLOCK(inp); return (stcb); } break; } #endif #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; rsin6 = (struct sockaddr_in6 *)from; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { /* found it */ if (netp != NULL) { *netp = net; } /* * Update the endpoint * pointer */ *inp_p = inp; SCTP_INP_RUNLOCK(inp); return (stcb); } break; } #endif default: /* TSNH */ break; } } SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); } return (NULL); } /* * rules for use * * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an * stcb, both will be locked (locked_tcb and stcb) but decrement will be done * (if locked == NULL). 3) Decrement happens on return ONLY if locked == * NULL. */ struct sctp_tcb * sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote, struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb) { struct sctpasochead *head; struct sctp_inpcb *inp; struct sctp_tcb *stcb = NULL; struct sctp_nets *net; uint16_t rport; inp = *inp_p; switch (remote->sa_family) { #ifdef INET case AF_INET: rport = (((struct sockaddr_in *)remote)->sin_port); break; #endif #ifdef INET6 case AF_INET6: rport = (((struct sockaddr_in6 *)remote)->sin6_port); break; #endif default: return (NULL); } if (locked_tcb) { /* * UN-lock so we can do proper locking here this occurs when * called from load_addresses_from_init. */ atomic_add_int(&locked_tcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(locked_tcb); } SCTP_INP_INFO_RLOCK(); if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /*- * Now either this guy is our listener or it's the * connector. If it is the one that issued the connect, then * it's only chance is to be the first TCB in the list. If * it is the acceptor, then do the special_lookup to hash * and find the real inp. */ if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) { /* to is peer addr, from is my addr */ stcb = sctp_tcb_special_locate(inp_p, remote, local, netp, inp->def_vrf_id); if ((stcb != NULL) && (locked_tcb == NULL)) { /* we have a locked tcb, lower refcount */ SCTP_INP_DECR_REF(inp); } if ((locked_tcb != NULL) && (locked_tcb != stcb)) { SCTP_INP_RLOCK(locked_tcb->sctp_ep); SCTP_TCB_LOCK(locked_tcb); atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); SCTP_INP_RUNLOCK(locked_tcb->sctp_ep); } SCTP_INP_INFO_RUNLOCK(); return (stcb); } else { SCTP_INP_WLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { goto null_return; } stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { goto null_return; } SCTP_TCB_LOCK(stcb); if (stcb->rport != rport) { /* remote port does not match. */ SCTP_TCB_UNLOCK(stcb); goto null_return; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); goto null_return; } if (local && !sctp_does_stcb_own_this_addr(stcb, local)) { SCTP_TCB_UNLOCK(stcb); goto null_return; } /* now look at the list of remote addresses */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { #ifdef INVARIANTS if (net == (TAILQ_NEXT(net, sctp_next))) { panic("Corrupt net list"); } #endif if (net->ro._l_addr.sa.sa_family != remote->sa_family) { /* not the same family */ continue; } switch (remote->sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin, *rsin; sin = (struct sockaddr_in *) &net->ro._l_addr; rsin = (struct sockaddr_in *)remote; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #endif #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; rsin6 = (struct sockaddr_in6 *)remote; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #endif default: /* TSNH */ break; } } SCTP_TCB_UNLOCK(stcb); } } else { SCTP_INP_WLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { goto null_return; } head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport, inp->sctp_hashmark)]; LIST_FOREACH(stcb, head, sctp_tcbhash) { if (stcb->rport != rport) { /* remote port does not match */ continue; } SCTP_TCB_LOCK(stcb); if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); continue; } if (local && !sctp_does_stcb_own_this_addr(stcb, local)) { SCTP_TCB_UNLOCK(stcb); continue; } /* now look at the list of remote addresses */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { #ifdef INVARIANTS if (net == (TAILQ_NEXT(net, sctp_next))) { panic("Corrupt net list"); } #endif if (net->ro._l_addr.sa.sa_family != remote->sa_family) { /* not the same family */ continue; } switch (remote->sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin, *rsin; sin = (struct sockaddr_in *) &net->ro._l_addr; rsin = (struct sockaddr_in *)remote; if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #endif #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6, *rsin6; sin6 = (struct sockaddr_in6 *) &net->ro._l_addr; rsin6 = (struct sockaddr_in6 *)remote; if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) { /* found it */ if (netp != NULL) { *netp = net; } if (locked_tcb == NULL) { SCTP_INP_DECR_REF(inp); } else if (locked_tcb != stcb) { SCTP_TCB_LOCK(locked_tcb); } if (locked_tcb) { atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); return (stcb); } break; } #endif default: /* TSNH */ break; } } SCTP_TCB_UNLOCK(stcb); } } null_return: /* clean up for returning null */ if (locked_tcb) { SCTP_TCB_LOCK(locked_tcb); atomic_subtract_int(&locked_tcb->asoc.refcnt, 1); } SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_RUNLOCK(); /* not found */ return (NULL); } /* * Find an association for a specific endpoint using the association id given * out in the COMM_UP notification */ struct sctp_tcb * sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock) { /* * Use my the assoc_id to find a endpoint */ struct sctpasochead *head; struct sctp_tcb *stcb; uint32_t id; if (inp == NULL) { SCTP_PRINTF("TSNH ep_associd\n"); return (NULL); } if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_PRINTF("TSNH ep_associd0\n"); return (NULL); } id = (uint32_t) asoc_id; head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)]; if (head == NULL) { /* invalid id TSNH */ SCTP_PRINTF("TSNH ep_associd1\n"); return (NULL); } LIST_FOREACH(stcb, head, sctp_tcbasocidhash) { if (stcb->asoc.assoc_id == id) { if (inp != stcb->sctp_ep) { /* * some other guy has the same id active (id * collision ??). */ SCTP_PRINTF("TSNH ep_associd2\n"); continue; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { continue; } if (want_lock) { SCTP_TCB_LOCK(stcb); } return (stcb); } } return (NULL); } struct sctp_tcb * sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock) { struct sctp_tcb *stcb; SCTP_INP_RLOCK(inp); stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock); SCTP_INP_RUNLOCK(inp); return (stcb); } /* * Endpoint probe expects that the INP_INFO is locked. */ static struct sctp_inpcb * sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head, uint16_t lport, uint32_t vrf_id) { struct sctp_inpcb *inp; struct sctp_laddr *laddr; #ifdef INET struct sockaddr_in *sin; #endif #ifdef INET6 struct sockaddr_in6 *sin6; struct sockaddr_in6 *intf_addr6; #endif int fnd; #ifdef INET sin = NULL; #endif #ifdef INET6 sin6 = NULL; #endif switch (nam->sa_family) { #ifdef INET case AF_INET: sin = (struct sockaddr_in *)nam; break; #endif #ifdef INET6 case AF_INET6: sin6 = (struct sockaddr_in6 *)nam; break; #endif default: /* unsupported family */ return (NULL); } if (head == NULL) return (NULL); LIST_FOREACH(inp, head, sctp_hash) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(inp); continue; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) && (inp->sctp_lport == lport)) { /* got it */ switch (nam->sa_family) { #ifdef INET case AF_INET: if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* * IPv4 on a IPv6 socket with ONLY * IPv6 set */ SCTP_INP_RUNLOCK(inp); continue; } if (prison_check_ip4(inp->ip_inp.inp.inp_cred, &sin->sin_addr) != 0) { SCTP_INP_RUNLOCK(inp); continue; } break; #endif #ifdef INET6 case AF_INET6: /* * A V6 address and the endpoint is NOT * bound V6 */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { SCTP_INP_RUNLOCK(inp); continue; } if (prison_check_ip6(inp->ip_inp.inp.inp_cred, &sin6->sin6_addr) != 0) { SCTP_INP_RUNLOCK(inp); continue; } break; #endif default: break; } /* does a VRF id match? */ fnd = 0; if (inp->def_vrf_id == vrf_id) fnd = 1; SCTP_INP_RUNLOCK(inp); if (!fnd) continue; return (inp); } SCTP_INP_RUNLOCK(inp); } switch (nam->sa_family) { #ifdef INET case AF_INET: if (sin->sin_addr.s_addr == INADDR_ANY) { /* Can't hunt for one that has no address specified */ return (NULL); } break; #endif #ifdef INET6 case AF_INET6: if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { /* Can't hunt for one that has no address specified */ return (NULL); } break; #endif default: break; } /* * ok, not bound to all so see if we can find a EP bound to this * address. */ LIST_FOREACH(inp, head, sctp_hash) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(inp); continue; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) { SCTP_INP_RUNLOCK(inp); continue; } /* * Ok this could be a likely candidate, look at all of its * addresses */ if (inp->sctp_lport != lport) { SCTP_INP_RUNLOCK(inp); continue; } /* does a VRF id match? */ fnd = 0; if (inp->def_vrf_id == vrf_id) fnd = 1; if (!fnd) { SCTP_INP_RUNLOCK(inp); continue; } LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __func__); continue; } SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ", (void *)laddr->ifa); if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n"); continue; } if (laddr->ifa->address.sa.sa_family == nam->sa_family) { /* possible, see if it matches */ switch (nam->sa_family) { #ifdef INET case AF_INET: if (sin->sin_addr.s_addr == laddr->ifa->address.sin.sin_addr.s_addr) { SCTP_INP_RUNLOCK(inp); return (inp); } break; #endif #ifdef INET6 case AF_INET6: intf_addr6 = &laddr->ifa->address.sin6; if (SCTP6_ARE_ADDR_EQUAL(sin6, intf_addr6)) { SCTP_INP_RUNLOCK(inp); return (inp); } break; #endif } } } SCTP_INP_RUNLOCK(inp); } return (NULL); } static struct sctp_inpcb * sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id) { struct sctppcbhead *head; struct sctp_inpcb *t_inp; int fnd; head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))]; LIST_FOREACH(t_inp, head, sctp_hash) { if (t_inp->sctp_lport != lport) { continue; } /* is it in the VRF in question */ fnd = 0; if (t_inp->def_vrf_id == vrf_id) fnd = 1; if (!fnd) continue; /* This one is in use. */ /* check the v6/v4 binding issue */ if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(t_inp)) { if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { /* collision in V6 space */ return (t_inp); } else { /* inp is BOUND_V4 no conflict */ continue; } } else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { /* t_inp is bound v4 and v6, conflict always */ return (t_inp); } else { /* t_inp is bound only V4 */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && SCTP_IPV6_V6ONLY(inp)) { /* no conflict */ continue; } /* else fall through to conflict */ } return (t_inp); } return (NULL); } int sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp) { /* For 1-2-1 with port reuse */ struct sctppcbhead *head; struct sctp_inpcb *tinp, *ninp; if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) { /* only works with port reuse on */ return (-1); } if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) { return (0); } SCTP_INP_RUNLOCK(inp); SCTP_INP_INFO_WLOCK(); head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))]; /* Kick out all non-listeners to the TCP hash */ LIST_FOREACH_SAFE(tinp, head, sctp_hash, ninp) { if (tinp->sctp_lport != inp->sctp_lport) { continue; } if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { continue; } if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { continue; } if (tinp->sctp_socket->so_qlimit) { continue; } SCTP_INP_WLOCK(tinp); LIST_REMOVE(tinp, sctp_hash); head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))]; tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL; LIST_INSERT_HEAD(head, tinp, sctp_hash); SCTP_INP_WUNLOCK(tinp); } SCTP_INP_WLOCK(inp); /* Pull from where he was */ LIST_REMOVE(inp, sctp_hash); inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL; head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))]; LIST_INSERT_HEAD(head, inp, sctp_hash); SCTP_INP_WUNLOCK(inp); SCTP_INP_RLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return (0); } struct sctp_inpcb * sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock, uint32_t vrf_id) { /* * First we check the hash table to see if someone has this port * bound with just the port. */ struct sctp_inpcb *inp; struct sctppcbhead *head; int lport; unsigned int i; #ifdef INET struct sockaddr_in *sin; #endif #ifdef INET6 struct sockaddr_in6 *sin6; #endif switch (nam->sa_family) { #ifdef INET case AF_INET: sin = (struct sockaddr_in *)nam; lport = sin->sin_port; break; #endif #ifdef INET6 case AF_INET6: sin6 = (struct sockaddr_in6 *)nam; lport = sin6->sin6_port; break; #endif default: return (NULL); } /* * I could cheat here and just cast to one of the types but we will * do it right. It also provides the check against an Unsupported * type too. */ /* Find the head of the ALLADDR chain */ if (have_lock == 0) { SCTP_INP_INFO_RLOCK(); } head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))]; inp = sctp_endpoint_probe(nam, head, lport, vrf_id); /* * If the TCP model exists it could be that the main listening * endpoint is gone but there still exists a connected socket for * this guy. If so we can return the first one that we find. This * may NOT be the correct one so the caller should be wary on the * returned INP. Currently the only caller that sets find_tcp_pool * is in bindx where we are verifying that a user CAN bind the * address. He either has bound it already, or someone else has, or * its open to bind, so this is good enough. */ if (inp == NULL && find_tcp_pool) { for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) { head = &SCTP_BASE_INFO(sctp_tcpephash)[i]; inp = sctp_endpoint_probe(nam, head, lport, vrf_id); if (inp) { break; } } } if (inp) { SCTP_INP_INCR_REF(inp); } if (have_lock == 0) { SCTP_INP_INFO_RUNLOCK(); } return (inp); } /* * Find an association for an endpoint with the pointer to whom you want to * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may * need to change the *to to some other struct like a mbuf... */ struct sctp_tcb * sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to, struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool, uint32_t vrf_id) { struct sctp_inpcb *inp = NULL; struct sctp_tcb *stcb; SCTP_INP_INFO_RLOCK(); if (find_tcp_pool) { if (inp_p != NULL) { stcb = sctp_tcb_special_locate(inp_p, from, to, netp, vrf_id); } else { stcb = sctp_tcb_special_locate(&inp, from, to, netp, vrf_id); } if (stcb != NULL) { SCTP_INP_INFO_RUNLOCK(); return (stcb); } } inp = sctp_pcb_findep(to, 0, 1, vrf_id); if (inp_p != NULL) { *inp_p = inp; } SCTP_INP_INFO_RUNLOCK(); if (inp == NULL) { return (NULL); } /* * ok, we have an endpoint, now lets find the assoc for it (if any) * we now place the source address or from in the to of the find * endpoint call. Since in reality this chain is used from the * inbound packet side. */ if (inp_p != NULL) { stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to, NULL); } else { stcb = sctp_findassociation_ep_addr(&inp, from, netp, to, NULL); } return (stcb); } /* * This routine will grub through the mbuf that is a INIT or INIT-ACK and * find all addresses that the sender has specified in any address list. Each * address will be used to lookup the TCB and see if one exits. */ static struct sctp_tcb * sctp_findassociation_special_addr(struct mbuf *m, int offset, struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, struct sockaddr *dst) { struct sctp_paramhdr *phdr, parm_buf; #if defined(INET) || defined(INET6) struct sctp_tcb *stcb; uint16_t ptype; #endif uint16_t plen; #ifdef INET struct sockaddr_in sin4; #endif #ifdef INET6 struct sockaddr_in6 sin6; #endif #ifdef INET memset(&sin4, 0, sizeof(sin4)); sin4.sin_len = sizeof(sin4); sin4.sin_family = AF_INET; sin4.sin_port = sh->src_port; #endif #ifdef INET6 memset(&sin6, 0, sizeof(sin6)); sin6.sin6_len = sizeof(sin6); sin6.sin6_family = AF_INET6; sin6.sin6_port = sh->src_port; #endif offset += sizeof(struct sctp_init_chunk); phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); while (phdr != NULL) { /* now we must see if we want the parameter */ #if defined(INET) || defined(INET6) ptype = ntohs(phdr->param_type); #endif plen = ntohs(phdr->param_length); if (plen == 0) { break; } #ifdef INET if (ptype == SCTP_IPV4_ADDRESS && plen == sizeof(struct sctp_ipv4addr_param)) { /* Get the rest of the address */ struct sctp_ipv4addr_param ip4_parm, *p4; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm))); if (phdr == NULL) { return (NULL); } p4 = (struct sctp_ipv4addr_param *)phdr; memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr)); /* look it up */ stcb = sctp_findassociation_ep_addr(inp_p, (struct sockaddr *)&sin4, netp, dst, NULL); if (stcb != NULL) { return (stcb); } } #endif #ifdef INET6 if (ptype == SCTP_IPV6_ADDRESS && plen == sizeof(struct sctp_ipv6addr_param)) { /* Get the rest of the address */ struct sctp_ipv6addr_param ip6_parm, *p6; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm))); if (phdr == NULL) { return (NULL); } p6 = (struct sctp_ipv6addr_param *)phdr; memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr)); /* look it up */ stcb = sctp_findassociation_ep_addr(inp_p, (struct sockaddr *)&sin6, netp, dst, NULL); if (stcb != NULL) { return (stcb); } } #endif offset += SCTP_SIZE32(plen); phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); } return (NULL); } static struct sctp_tcb * sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport, uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag) { /* * Use my vtag to hash. If we find it we then verify the source addr * is in the assoc. If all goes well we save a bit on rec of a * packet. */ struct sctpasochead *head; struct sctp_nets *net; struct sctp_tcb *stcb; SCTP_INP_INFO_RLOCK(); head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag, SCTP_BASE_INFO(hashasocmark))]; LIST_FOREACH(stcb, head, sctp_asocs) { SCTP_INP_RLOCK(stcb->sctp_ep); if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { SCTP_INP_RUNLOCK(stcb->sctp_ep); continue; } if (stcb->sctp_ep->def_vrf_id != vrf_id) { SCTP_INP_RUNLOCK(stcb->sctp_ep); continue; } SCTP_TCB_LOCK(stcb); SCTP_INP_RUNLOCK(stcb->sctp_ep); if (stcb->asoc.my_vtag == vtag) { /* candidate */ if (stcb->rport != rport) { SCTP_TCB_UNLOCK(stcb); continue; } if (stcb->sctp_ep->sctp_lport != lport) { SCTP_TCB_UNLOCK(stcb); continue; } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); continue; } /* RRS:Need toaddr check here */ if (sctp_does_stcb_own_this_addr(stcb, to) == 0) { /* Endpoint does not own this address */ SCTP_TCB_UNLOCK(stcb); continue; } if (remote_tag) { /* * If we have both vtags that's all we match * on */ if (stcb->asoc.peer_vtag == remote_tag) { /* * If both tags match we consider it * conclusive and check NO * source/destination addresses */ goto conclusive; } } if (skip_src_check) { conclusive: if (from) { *netp = sctp_findnet(stcb, from); } else { *netp = NULL; /* unknown */ } if (inp_p) *inp_p = stcb->sctp_ep; SCTP_INP_INFO_RUNLOCK(); return (stcb); } net = sctp_findnet(stcb, from); if (net) { /* yep its him. */ *netp = net; SCTP_STAT_INCR(sctps_vtagexpress); *inp_p = stcb->sctp_ep; SCTP_INP_INFO_RUNLOCK(); return (stcb); } else { /* * not him, this should only happen in rare * cases so I peg it. */ SCTP_STAT_INCR(sctps_vtagbogus); } } SCTP_TCB_UNLOCK(stcb); } SCTP_INP_INFO_RUNLOCK(); return (NULL); } /* * Find an association with the pointer to the inbound IP packet. This can be * a IPv4 or IPv6 packet. */ struct sctp_tcb * sctp_findassociation_addr(struct mbuf *m, int offset, struct sockaddr *src, struct sockaddr *dst, struct sctphdr *sh, struct sctp_chunkhdr *ch, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id) { int find_tcp_pool; struct sctp_tcb *stcb; struct sctp_inpcb *inp; if (sh->v_tag) { /* we only go down this path if vtag is non-zero */ stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag), inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0); if (stcb) { return (stcb); } } find_tcp_pool = 0; if ((ch->chunk_type != SCTP_INITIATION) && (ch->chunk_type != SCTP_INITIATION_ACK) && (ch->chunk_type != SCTP_COOKIE_ACK) && (ch->chunk_type != SCTP_COOKIE_ECHO)) { /* Other chunk types go to the tcp pool. */ find_tcp_pool = 1; } if (inp_p) { stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp, find_tcp_pool, vrf_id); inp = *inp_p; } else { stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp, find_tcp_pool, vrf_id); } SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp); if (stcb == NULL && inp) { /* Found a EP but not this address */ if ((ch->chunk_type == SCTP_INITIATION) || (ch->chunk_type == SCTP_INITIATION_ACK)) { /*- * special hook, we do NOT return linp or an * association that is linked to an existing * association that is under the TCP pool (i.e. no * listener exists). The endpoint finding routine * will always find a listener before examining the * TCP pool. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) { if (inp_p) { *inp_p = NULL; } return (NULL); } stcb = sctp_findassociation_special_addr(m, offset, sh, &inp, netp, dst); if (inp_p != NULL) { *inp_p = inp; } } } SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb); return (stcb); } /* * lookup an association by an ASCONF lookup address. * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup */ struct sctp_tcb * sctp_findassociation_ep_asconf(struct mbuf *m, int offset, struct sockaddr *dst, struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id) { struct sctp_tcb *stcb; union sctp_sockstore remote_store; struct sctp_paramhdr parm_buf, *phdr; int ptype; int zero_address = 0; #ifdef INET struct sockaddr_in *sin; #endif #ifdef INET6 struct sockaddr_in6 *sin6; #endif memset(&remote_store, 0, sizeof(remote_store)); phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), &parm_buf, sizeof(struct sctp_paramhdr)); if (phdr == NULL) { SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n", __func__); return NULL; } ptype = (int)((uint32_t) ntohs(phdr->param_type)); /* get the correlation address */ switch (ptype) { #ifdef INET6 case SCTP_IPV6_ADDRESS: { /* ipv6 address param */ struct sctp_ipv6addr_param *p6, p6_buf; if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) { return NULL; } p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), &p6_buf.ph, sizeof(*p6)); if (p6 == NULL) { SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n", __func__); return (NULL); } sin6 = &remote_store.sin6; sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); sin6->sin6_port = sh->src_port; memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr)); if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) zero_address = 1; break; } #endif #ifdef INET case SCTP_IPV4_ADDRESS: { /* ipv4 address param */ struct sctp_ipv4addr_param *p4, p4_buf; if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) { return NULL; } p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk), &p4_buf.ph, sizeof(*p4)); if (p4 == NULL) { SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n", __func__); return (NULL); } sin = &remote_store.sin; sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_port = sh->src_port; memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr)); if (sin->sin_addr.s_addr == INADDR_ANY) zero_address = 1; break; } #endif default: /* invalid address param type */ return NULL; } if (zero_address) { stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p, netp, sh->src_port, sh->dest_port, 1, vrf_id, 0); if (stcb != NULL) { SCTP_INP_DECR_REF(*inp_p); } } else { stcb = sctp_findassociation_ep_addr(inp_p, &remote_store.sa, netp, dst, NULL); } return (stcb); } /* * allocate a sctp_inpcb and setup a temporary binding to a port/all * addresses. This way if we don't get a bind we by default pick a ephemeral * port with all addresses bound. */ int sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id) { /* * we get called when a new endpoint starts up. We need to allocate * the sctp_inpcb structure from the zone and init it. Mark it as * unbound and find a port that we can use as an ephemeral with * INADDR_ANY. If the user binds later no problem we can then add in * the specific addresses. And setup the default parameters for the * EP. */ int i, error; struct sctp_inpcb *inp; struct sctp_pcb *m; struct timeval time; sctp_sharedkey_t *null_key; error = 0; SCTP_INP_INFO_WLOCK(); inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb); if (inp == NULL) { SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n"); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); return (ENOBUFS); } /* zap it */ bzero(inp, sizeof(*inp)); /* bump generations */ /* setup socket pointers */ inp->sctp_socket = so; inp->ip_inp.inp.inp_socket = so; inp->ip_inp.inp.inp_cred = crhold(so->so_cred); #ifdef INET6 if (INP_SOCKAF(so) == AF_INET6) { if (MODULE_GLOBAL(ip6_auto_flowlabel)) { inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL; } if (MODULE_GLOBAL(ip6_v6only)) { inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY; } } #endif inp->sctp_associd_counter = 1; inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT; inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT; inp->max_cwnd = 0; inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off); inp->ecn_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_ecn_enable); inp->prsctp_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pr_enable); inp->auth_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_auth_enable); inp->asconf_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_asconf_enable); inp->reconfig_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_reconfig_enable); inp->nrsack_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_nrsack_enable); inp->pktdrop_supported = (uint8_t) SCTP_BASE_SYSCTL(sctp_pktdrop_enable); inp->fibnum = so->so_fibnum; /* init the small hash table we use to track asocid <-> tcb */ inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark); if (inp->sctp_asocidhash == NULL) { crfree(inp->ip_inp.inp.inp_cred); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); SCTP_INP_INFO_WUNLOCK(); return (ENOBUFS); } #ifdef IPSEC error = ipsec_init_policy(so, &inp->ip_inp.inp.inp_sp); if (error != 0) { crfree(inp->ip_inp.inp.inp_cred); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); SCTP_INP_INFO_WUNLOCK(); return error; } #endif /* IPSEC */ SCTP_INCR_EP_COUNT(); inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl); SCTP_INP_INFO_WUNLOCK(); so->so_pcb = (caddr_t)inp; if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) { /* UDP style socket */ inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE | SCTP_PCB_FLAGS_UNBOUND); /* Be sure it is NON-BLOCKING IO for UDP */ /* SCTP_SET_SO_NBIO(so); */ } else if (SCTP_SO_TYPE(so) == SOCK_STREAM) { /* TCP style socket */ inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE | SCTP_PCB_FLAGS_UNBOUND); /* Be sure we have blocking IO by default */ SCTP_CLEAR_SO_NBIO(so); } else { /* * unsupported socket type (RAW, etc)- in case we missed it * in protosw */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP); so->so_pcb = NULL; crfree(inp->ip_inp.inp.inp_cred); #ifdef IPSEC ipsec_delete_pcbpolicy(&inp->ip_inp.inp); #endif SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); return (EOPNOTSUPP); } if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) { sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) { sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize), &inp->sctp_hashmark); if (inp->sctp_tcbhash == NULL) { SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n"); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); so->so_pcb = NULL; crfree(inp->ip_inp.inp.inp_cred); #ifdef IPSEC ipsec_delete_pcbpolicy(&inp->ip_inp.inp); #endif SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); return (ENOBUFS); } inp->def_vrf_id = vrf_id; SCTP_INP_INFO_WLOCK(); SCTP_INP_LOCK_INIT(inp); INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp"); SCTP_INP_READ_INIT(inp); SCTP_ASOC_CREATE_LOCK_INIT(inp); /* lock the new ep */ SCTP_INP_WLOCK(inp); /* add it to the info area */ LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list); SCTP_INP_INFO_WUNLOCK(); TAILQ_INIT(&inp->read_queue); LIST_INIT(&inp->sctp_addr_list); LIST_INIT(&inp->sctp_asoc_list); #ifdef SCTP_TRACK_FREED_ASOCS /* TEMP CODE */ LIST_INIT(&inp->sctp_asoc_free_list); #endif /* Init the timer structure for signature change */ SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer); inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE; /* now init the actual endpoint default data */ m = &inp->sctp_ep; /* setup the base timeout information */ m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC); /* needed ? */ m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC); /* needed ? */ m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default)); m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default)); m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default)); m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default)); m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default)); /* all max/min max are in ms */ m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default); m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default); m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default); m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default); m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default); m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default); m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default); m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default); m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold); m->sctp_sws_sender = SCTP_SWS_SENDER_DEF; m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF; m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default); m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default); m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module); m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module); m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default); /* number of streams to pre-open on a association */ m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default); /* Add adaptation cookie */ m->adaptation_layer_indicator = 0; m->adaptation_layer_indicator_provided = 0; /* seed random number generator */ m->random_counter = 1; m->store_at = SCTP_SIGNATURE_SIZE; SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers)); sctp_fill_random_store(m); /* Minimum cookie size */ m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) + sizeof(struct sctp_state_cookie); m->size_of_a_cookie += SCTP_SIGNATURE_SIZE; /* Setup the initial secret */ (void)SCTP_GETTIME_TIMEVAL(&time); m->time_of_secret_change = time.tv_sec; for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) { m->secret_key[0][i] = sctp_select_initial_TSN(m); } sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL); /* How long is a cookie good for ? */ m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default)); /* * Initialize authentication parameters */ m->local_hmacs = sctp_default_supported_hmaclist(); m->local_auth_chunks = sctp_alloc_chunklist(); if (inp->asconf_supported) { sctp_auth_add_chunk(SCTP_ASCONF, m->local_auth_chunks); sctp_auth_add_chunk(SCTP_ASCONF_ACK, m->local_auth_chunks); } m->default_dscp = 0; #ifdef INET6 m->default_flowlabel = 0; #endif m->port = 0; /* encapsulation disabled by default */ LIST_INIT(&m->shared_keys); /* add default NULL key as key id 0 */ null_key = sctp_alloc_sharedkey(); sctp_insert_sharedkey(&m->shared_keys, null_key); SCTP_INP_WUNLOCK(inp); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 12); #endif return (error); } void sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp, struct sctp_tcb *stcb) { struct sctp_nets *net; uint16_t lport, rport; struct sctppcbhead *head; struct sctp_laddr *laddr, *oladdr; atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(old_inp); SCTP_INP_WLOCK(new_inp); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); new_inp->sctp_ep.time_of_secret_change = old_inp->sctp_ep.time_of_secret_change; memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key, sizeof(old_inp->sctp_ep.secret_key)); new_inp->sctp_ep.current_secret_number = old_inp->sctp_ep.current_secret_number; new_inp->sctp_ep.last_secret_number = old_inp->sctp_ep.last_secret_number; new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie; /* make it so new data pours into the new socket */ stcb->sctp_socket = new_inp->sctp_socket; stcb->sctp_ep = new_inp; /* Copy the port across */ lport = new_inp->sctp_lport = old_inp->sctp_lport; rport = stcb->rport; /* Pull the tcb from the old association */ LIST_REMOVE(stcb, sctp_tcbhash); LIST_REMOVE(stcb, sctp_tcblist); if (stcb->asoc.in_asocid_hash) { LIST_REMOVE(stcb, sctp_tcbasocidhash); } /* Now insert the new_inp into the TCP connected hash */ head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))]; LIST_INSERT_HEAD(head, new_inp, sctp_hash); /* Its safe to access */ new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND; /* Now move the tcb into the endpoint list */ LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist); /* * Question, do we even need to worry about the ep-hash since we * only have one connection? Probably not :> so lets get rid of it * and not suck up any kernel memory in that. */ if (stcb->asoc.in_asocid_hash) { struct sctpasochead *lhd; lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id, new_inp->hashasocidmark)]; LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash); } /* Ok. Let's restart timer. */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp, stcb, net); } SCTP_INP_INFO_WUNLOCK(); if (new_inp->sctp_tcbhash != NULL) { SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark); new_inp->sctp_tcbhash = NULL; } if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { /* Subset bound, so copy in the laddr list from the old_inp */ LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) { laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (laddr == NULL) { /* * Gak, what can we do? This assoc is really * HOSED. We probably should send an abort * here. */ SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n"); continue; } SCTP_INCR_LADDR_COUNT(); bzero(laddr, sizeof(*laddr)); (void)SCTP_GETTIME_TIMEVAL(&laddr->start_time); laddr->ifa = oladdr->ifa; atomic_add_int(&laddr->ifa->refcount, 1); LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr, sctp_nxt_addr); new_inp->laddr_count++; if (oladdr == stcb->asoc.last_used_address) { stcb->asoc.last_used_address = laddr; } } } /* * Now any running timers need to be adjusted since we really don't * care if they are running or not just blast in the new_inp into * all of them. */ stcb->asoc.dack_timer.ep = (void *)new_inp; stcb->asoc.asconf_timer.ep = (void *)new_inp; stcb->asoc.strreset_timer.ep = (void *)new_inp; stcb->asoc.shut_guard_timer.ep = (void *)new_inp; stcb->asoc.autoclose_timer.ep = (void *)new_inp; stcb->asoc.delayed_event_timer.ep = (void *)new_inp; stcb->asoc.delete_prim_timer.ep = (void *)new_inp; /* now what about the nets? */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { net->pmtu_timer.ep = (void *)new_inp; net->hb_timer.ep = (void *)new_inp; net->rxt_timer.ep = (void *)new_inp; } SCTP_INP_WUNLOCK(new_inp); SCTP_INP_WUNLOCK(old_inp); } /* sctp_ifap is used to bypass normal local address validation checks */ int sctp_inpcb_bind(struct socket *so, struct sockaddr *addr, struct sctp_ifa *sctp_ifap, struct thread *p) { /* bind a ep to a socket address */ struct sctppcbhead *head; struct sctp_inpcb *inp, *inp_tmp; struct inpcb *ip_inp; int port_reuse_active = 0; int bindall; uint16_t lport; int error; uint32_t vrf_id; lport = 0; bindall = 1; inp = (struct sctp_inpcb *)so->so_pcb; ip_inp = (struct inpcb *)so->so_pcb; #ifdef SCTP_DEBUG if (addr) { SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n", ntohs(((struct sockaddr_in *)addr)->sin_port)); SCTPDBG(SCTP_DEBUG_PCB1, "Addr: "); SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr); } #endif if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { /* already did a bind, subsequent binds NOT allowed ! */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } #ifdef INVARIANTS if (p == NULL) panic("null proc/thread"); #endif if (addr != NULL) { switch (addr->sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin; /* IPV6_V6ONLY socket? */ if (SCTP_IPV6_V6ONLY(ip_inp)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } if (addr->sa_len != sizeof(*sin)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } sin = (struct sockaddr_in *)addr; lport = sin->sin_port; /* * For LOOPBACK the prison_local_ip4() call * will transmute the ip address to the * proper value. */ if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); return (error); } if (sin->sin_addr.s_addr != INADDR_ANY) { bindall = 0; } break; } #endif #ifdef INET6 case AF_INET6: { /* * Only for pure IPv6 Address. (No IPv4 * Mapped!) */ struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (addr->sa_len != sizeof(*sin6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } lport = sin6->sin6_port; /* * For LOOPBACK the prison_local_ip6() call * will transmute the ipv6 address to the * proper value. */ if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr, (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); return (error); } if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { bindall = 0; /* KAME hack: embed scopeid */ if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } } /* this must be cleared for ifa_ifwithaddr() */ sin6->sin6_scope_id = 0; break; } #endif default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT); return (EAFNOSUPPORT); } } SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); /* Setup a vrf_id to be the default for the non-bind-all case. */ vrf_id = inp->def_vrf_id; /* increase our count due to the unlock we do */ SCTP_INP_INCR_REF(inp); if (lport) { /* * Did the caller specify a port? if so we must see if an ep * already has this one bound. */ /* got to be root to get at low ports */ if (ntohs(lport) < IPPORT_RESERVED) { if (p && (error = priv_check(p, PRIV_NETINET_RESERVEDPORT) )) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return (error); } } SCTP_INP_WUNLOCK(inp); if (bindall) { vrf_id = inp->def_vrf_id; inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id); if (inp_tmp != NULL) { /* * lock guy returned and lower count note * that we are not bound so inp_tmp should * NEVER be inp. And it is this inp * (inp_tmp) that gets the reference bump, * so we must lower it. */ SCTP_INP_DECR_REF(inp_tmp); /* unlock info */ if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { /* * Ok, must be one-2-one and * allowing port re-use */ port_reuse_active = 1; goto continue_anyway; } SCTP_INP_DECR_REF(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } } else { inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id); if (inp_tmp != NULL) { /* * lock guy returned and lower count note * that we are not bound so inp_tmp should * NEVER be inp. And it is this inp * (inp_tmp) that gets the reference bump, * so we must lower it. */ SCTP_INP_DECR_REF(inp_tmp); /* unlock info */ if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { /* * Ok, must be one-2-one and * allowing port re-use */ port_reuse_active = 1; goto continue_anyway; } SCTP_INP_DECR_REF(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } } continue_anyway: SCTP_INP_WLOCK(inp); if (bindall) { /* verify that no lport is not used by a singleton */ if ((port_reuse_active == 0) && (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) { /* Sorry someone already has this one bound */ if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) { port_reuse_active = 1; } else { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } } } } else { uint16_t first, last, candidate; uint16_t count; int done; if (ip_inp->inp_flags & INP_HIGHPORT) { first = MODULE_GLOBAL(ipport_hifirstauto); last = MODULE_GLOBAL(ipport_hilastauto); } else if (ip_inp->inp_flags & INP_LOWPORT) { if (p && (error = priv_check(p, PRIV_NETINET_RESERVEDPORT) )) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error); return (error); } first = MODULE_GLOBAL(ipport_lowfirstauto); last = MODULE_GLOBAL(ipport_lowlastauto); } else { first = MODULE_GLOBAL(ipport_firstauto); last = MODULE_GLOBAL(ipport_lastauto); } if (first > last) { uint16_t temp; temp = first; first = last; last = temp; } count = last - first + 1; /* number of candidates */ candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count); done = 0; while (!done) { if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) { done = 1; } if (!done) { if (--count == 0) { SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE); return (EADDRINUSE); } if (candidate == last) candidate = first; else candidate = candidate + 1; } } lport = htons(candidate); } SCTP_INP_DECR_REF(inp); if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { /* * this really should not happen. The guy did a non-blocking * bind and then did a close at the same time. */ SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } /* ok we look clear to give out this port, so lets setup the binding */ if (bindall) { /* binding to all addresses, so just set in the proper flags */ inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL; /* set the automatic addr changes from kernel flag */ if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF); sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); } else { sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF); sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); } if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS); } else { sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS); } /* * set the automatic mobility_base from kernel flag (by * micchie) */ if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) { sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } else { sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } /* * set the automatic mobility_fasthandoff from kernel flag * (by micchie) */ if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) { sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } else { sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF); sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED); } } else { /* * bind specific, make sure flags is off and add a new * address structure to the sctp_addr_list inside the ep * structure. * * We will need to allocate one and insert it at the head. The * socketopt call can just insert new addresses in there as * well. It will also have to do the embed scope kame hack * too (before adding). */ struct sctp_ifa *ifa; union sctp_sockstore store; memset(&store, 0, sizeof(store)); switch (addr->sa_family) { #ifdef INET case AF_INET: memcpy(&store.sin, addr, sizeof(struct sockaddr_in)); store.sin.sin_port = 0; break; #endif #ifdef INET6 case AF_INET6: memcpy(&store.sin6, addr, sizeof(struct sockaddr_in6)); store.sin6.sin6_port = 0; break; #endif default: break; } /* * first find the interface with the bound address need to * zero out the port to find the address! yuck! can't do * this earlier since need port for sctp_pcb_findep() */ if (sctp_ifap != NULL) { ifa = sctp_ifap; } else { /* * Note for BSD we hit here always other O/S's will * pass things in via the sctp_ifap argument * (Panda). */ ifa = sctp_find_ifa_by_addr(&store.sa, vrf_id, SCTP_ADDR_NOT_LOCKED); } if (ifa == NULL) { /* Can't find an interface with that address */ SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL); return (EADDRNOTAVAIL); } #ifdef INET6 if (addr->sa_family == AF_INET6) { /* GAK, more FIXME IFA lock? */ if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { /* Can't bind a non-existent addr. */ SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } } #endif /* we're not bound all */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL; /* allow bindx() to send ASCONF's for binding changes */ sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF); /* clear automatic addr changes from kernel flag */ sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); /* add this address to the endpoint list */ error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0); if (error != 0) { SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return (error); } inp->laddr_count++; } /* find the bucket */ if (port_reuse_active) { /* Put it into tcp 1-2-1 hash */ head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))]; inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL; } else { head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))]; } /* put it in the bucket */ LIST_INSERT_HEAD(head, inp, sctp_hash); SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n", (void *)head, ntohs(lport), port_reuse_active); /* set in the port */ inp->sctp_lport = lport; /* turn off just the unbound flag */ inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND; SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return (0); } static void sctp_iterator_inp_being_freed(struct sctp_inpcb *inp) { struct sctp_iterator *it, *nit; /* * We enter with the only the ITERATOR_LOCK in place and a write * lock on the inp_info stuff. */ it = sctp_it_ctl.cur_it; if (it && (it->vn != curvnet)) { /* Its not looking at our VNET */ return; } if (it && (it->inp == inp)) { /* * This is tricky and we hold the iterator lock, but when it * returns and gets the lock (when we release it) the * iterator will try to operate on inp. We need to stop that * from happening. But of course the iterator has a * reference on the stcb and inp. We can mark it and it will * stop. * * If its a single iterator situation, we set the end iterator * flag. Otherwise we set the iterator to go to the next * inp. * */ if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT; } else { sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP; } } /* * Now go through and remove any single reference to our inp that * may be still pending on the list */ SCTP_IPI_ITERATOR_WQ_LOCK(); TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) { if (it->vn != curvnet) { continue; } if (it->inp == inp) { /* This one points to me is it inp specific? */ if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) { /* Remove and free this one */ TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); if (it->function_atend != NULL) { (*it->function_atend) (it->pointer, it->val); } SCTP_FREE(it, SCTP_M_ITER); } else { it->inp = LIST_NEXT(it->inp, sctp_list); if (it->inp) { SCTP_INP_INCR_REF(it->inp); } } /* * When its put in the refcnt is incremented so decr * it */ SCTP_INP_DECR_REF(inp); } } SCTP_IPI_ITERATOR_WQ_UNLOCK(); } /* release sctp_inpcb unbind the port */ void sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from) { /* * Here we free a endpoint. We must find it (if it is in the Hash * table) and remove it from there. Then we must also find it in the * overall list and remove it from there. After all removals are * complete then any timer has to be stopped. Then start the actual * freeing. a) Any local lists. b) Any associations. c) The hash of * all associations. d) finally the ep itself. */ struct sctp_tcb *asoc, *nasoc; struct sctp_laddr *laddr, *nladdr; struct inpcb *ip_pcb; struct socket *so; int being_refed = 0; struct sctp_queued_to_read *sq, *nsq; int cnt; sctp_sharedkey_t *shared_key, *nshared_key; #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 0); #endif SCTP_ITERATOR_LOCK(); /* mark any iterators on the list or being processed */ sctp_iterator_inp_being_freed(inp); SCTP_ITERATOR_UNLOCK(); so = inp->sctp_socket; if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { /* been here before.. eeks.. get out of here */ SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 1); #endif return; } SCTP_ASOC_CREATE_LOCK(inp); SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) { inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP; /* socket is gone, so no more wakeups allowed */ inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE; inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT; inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT; } /* First time through we have the socket lock, after that no more. */ sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL, SCTP_FROM_SCTP_PCB + SCTP_LOC_1); if (inp->control) { sctp_m_freem(inp->control); inp->control = NULL; } if (inp->pkt) { sctp_m_freem(inp->pkt); inp->pkt = NULL; } ip_pcb = &inp->ip_inp.inp; /* we could just cast the main pointer * here but I will be nice :> (i.e. * ip_pcb = ep;) */ if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) { int cnt_in_sd; cnt_in_sd = 0; LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) { SCTP_TCB_LOCK(asoc); if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* Skip guys being freed */ cnt_in_sd++; if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) { /* * Special case - we did not start a * kill timer on the asoc due to it * was not closed. So go ahead and * start it now. */ asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE; sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL); } SCTP_TCB_UNLOCK(asoc); continue; } if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) && (asoc->asoc.total_output_queue_size == 0)) { /* * If we have data in queue, we don't want * to just free since the app may have done, * send()/close or connect/send/close. And * it wants the data to get across first. */ /* Just abandon things in the front states */ if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) { cnt_in_sd++; } continue; } /* Disconnect the socket please */ asoc->sctp_socket = NULL; asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET; if ((asoc->asoc.size_on_reasm_queue > 0) || (asoc->asoc.control_pdapi) || (asoc->asoc.size_on_all_streams > 0) || (so && (so->so_rcv.sb_cc > 0))) { /* Left with Data unread */ struct mbuf *op_err; op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3; sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) { cnt_in_sd++; } continue; } else if (TAILQ_EMPTY(&asoc->asoc.send_queue) && TAILQ_EMPTY(&asoc->asoc.sent_queue) && (asoc->asoc.stream_queue_cnt == 0)) { if (asoc->asoc.locked_on_sending) { goto abort_anyway; } if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { struct sctp_nets *netp; /* * there is nothing queued to send, * so I send shutdown */ if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_stop_timers_for_shutdown(asoc); if (asoc->asoc.alternate) { netp = asoc->asoc.alternate; } else { netp = asoc->asoc.primary_destination; } sctp_send_shutdown(asoc, netp); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc, netp); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc, asoc->asoc.primary_destination); sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED); } } else { /* mark into shutdown pending */ struct sctp_stream_queue_pending *sp; asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING; sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc, asoc->asoc.primary_destination); if (asoc->asoc.locked_on_sending) { sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue), sctp_streamhead); if (sp == NULL) { SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n", (void *)asoc->asoc.locked_on_sending, asoc->asoc.locked_on_sending->stream_no); } else { if ((sp->length == 0) && (sp->msg_is_complete == 0)) asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT; } } if (TAILQ_EMPTY(&asoc->asoc.send_queue) && TAILQ_EMPTY(&asoc->asoc.sent_queue) && (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) { struct mbuf *op_err; abort_anyway: op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5; sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) { cnt_in_sd++; } continue; } else { sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); } } cnt_in_sd++; SCTP_TCB_UNLOCK(asoc); } /* now is there some left in our SHUTDOWN state? */ if (cnt_in_sd) { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 2); #endif inp->sctp_socket = NULL; SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return; } } inp->sctp_socket = NULL; if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) != SCTP_PCB_FLAGS_UNBOUND) { /* * ok, this guy has been bound. It's port is somewhere in * the SCTP_BASE_INFO(hash table). Remove it! */ LIST_REMOVE(inp, sctp_hash); inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND; } /* * If there is a timer running to kill us, forget it, since it may * have a contest on the INP lock.. which would cause us to die ... */ cnt = 0; LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) { SCTP_TCB_LOCK(asoc); if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) { asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE; sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL); } cnt++; SCTP_TCB_UNLOCK(asoc); continue; } /* Free associations that are NOT killing us */ if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) && ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) { struct mbuf *op_err; op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7; sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); } else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { cnt++; SCTP_TCB_UNLOCK(asoc); continue; } if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) { cnt++; } } if (cnt) { /* Ok we have someone out there that will kill us */ (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 3); #endif SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return; } if (SCTP_INP_LOCK_CONTENDED(inp)) being_refed++; if (SCTP_INP_READ_CONTENDED(inp)) being_refed++; if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp)) being_refed++; if ((inp->refcount) || (being_refed) || (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) { (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 4); #endif sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL); SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); return; } inp->sctp_ep.signature_change.type = 0; inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE; /* * Remove it from the list .. last thing we need a lock for. */ LIST_REMOVE(inp, sctp_list); SCTP_INP_WUNLOCK(inp); SCTP_ASOC_CREATE_UNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); /* * Now we release all locks. Since this INP cannot be found anymore * except possibly by the kill timer that might be running. We call * the drain function here. It should hit the case were it sees the * ACTIVE flag cleared and exit out freeing us to proceed and * destroy everything. */ if (from != SCTP_CALLED_FROM_INPKILL_TIMER) { (void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer); } else { /* Probably un-needed */ (void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer); } #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 5); #endif if ((inp->sctp_asocidhash) != NULL) { SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark); inp->sctp_asocidhash = NULL; } /* sa_ignore FREED_MEMORY */ TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) { /* Its only abandoned if it had data left */ if (sq->length) SCTP_STAT_INCR(sctps_left_abandon); TAILQ_REMOVE(&inp->read_queue, sq, next); sctp_free_remote_addr(sq->whoFrom); if (so) so->so_rcv.sb_cc -= sq->length; if (sq->data) { sctp_m_freem(sq->data); sq->data = NULL; } /* * no need to free the net count, since at this point all * assoc's are gone. */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq); SCTP_DECR_READQ_COUNT(); } /* Now the sctp_pcb things */ /* * free each asoc if it is not already closed/free. we can't use the * macro here since le_next will get freed as part of the * sctp_free_assoc() call. */ #ifdef IPSEC ipsec_delete_pcbpolicy(ip_pcb); #endif if (ip_pcb->inp_options) { (void)sctp_m_free(ip_pcb->inp_options); ip_pcb->inp_options = 0; } #ifdef INET6 if (ip_pcb->inp_vflag & INP_IPV6) { struct in6pcb *in6p; in6p = (struct in6pcb *)inp; ip6_freepcbopts(in6p->in6p_outputopts); } #endif /* INET6 */ ip_pcb->inp_vflag = 0; /* free up authentication fields */ if (inp->sctp_ep.local_auth_chunks != NULL) sctp_free_chunklist(inp->sctp_ep.local_auth_chunks); if (inp->sctp_ep.local_hmacs != NULL) sctp_free_hmaclist(inp->sctp_ep.local_hmacs); LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) { LIST_REMOVE(shared_key, next); sctp_free_sharedkey(shared_key); /* sa_ignore FREED_MEMORY */ } /* * if we have an address list the following will free the list of * ifaddr's that are set into this ep. Again macro limitations here, * since the LIST_FOREACH could be a bad idea. */ LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) { sctp_remove_laddr(laddr); } #ifdef SCTP_TRACK_FREED_ASOCS /* TEMP CODE */ LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) { LIST_REMOVE(asoc, sctp_tcblist); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc); SCTP_DECR_ASOC_COUNT(); } /* *** END TEMP CODE *** */ #endif /* Now lets see about freeing the EP hash table. */ if (inp->sctp_tcbhash != NULL) { SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark); inp->sctp_tcbhash = NULL; } /* Now we must put the ep memory back into the zone pool */ crfree(inp->ip_inp.inp.inp_cred); INP_LOCK_DESTROY(&inp->ip_inp.inp); SCTP_INP_LOCK_DESTROY(inp); SCTP_INP_READ_DESTROY(inp); SCTP_ASOC_CREATE_LOCK_DESTROY(inp); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp); SCTP_DECR_EP_COUNT(); } struct sctp_nets * sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr) { struct sctp_nets *net; /* locate the address */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr)) return (net); } return (NULL); } int sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id) { struct sctp_ifa *sctp_ifa; sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED); if (sctp_ifa) { return (1); } else { return (0); } } /* * add's a remote endpoint address, done with the INIT/INIT-ACK as well as * when a ASCONF arrives that adds it. It will also initialize all the cwnd * stats of stuff. */ int sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr, struct sctp_nets **netp, int set_scope, int from) { /* * The following is redundant to the same lines in the * sctp_aloc_assoc() but is needed since others call the add address * function */ struct sctp_nets *net, *netfirst; int addr_inscope; SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ", from); SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr); netfirst = sctp_findnet(stcb, newaddr); if (netfirst) { /* * Lie and return ok, we don't want to make the association * go away for this behavior. It will happen in the TCP * model in a connected socket. It does not reach the hash * table until after the association is built so it can't be * found. Mark as reachable, since the initial creation will * have been cleared and the NOT_IN_ASSOC flag will have * been added... and we don't want to end up removing it * back out. */ if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) { netfirst->dest_state = (SCTP_ADDR_REACHABLE | SCTP_ADDR_UNCONFIRMED); } else { netfirst->dest_state = SCTP_ADDR_REACHABLE; } return (0); } addr_inscope = 1; switch (newaddr->sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin; sin = (struct sockaddr_in *)newaddr; if (sin->sin_addr.s_addr == 0) { /* Invalid address */ return (-1); } /* zero out the bzero area */ memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); /* assure len is set */ sin->sin_len = sizeof(struct sockaddr_in); if (set_scope) { if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) { stcb->asoc.scope.ipv4_local_scope = 1; } } else { /* Validate the address is in scope */ if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) && (stcb->asoc.scope.ipv4_local_scope == 0)) { addr_inscope = 0; } } break; } #endif #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)newaddr; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { /* Invalid address */ return (-1); } /* assure len is set */ sin6->sin6_len = sizeof(struct sockaddr_in6); if (set_scope) { if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) { stcb->asoc.scope.loopback_scope = 1; stcb->asoc.scope.local_scope = 0; stcb->asoc.scope.ipv4_local_scope = 1; stcb->asoc.scope.site_scope = 1; } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { /* * If the new destination is a * LINK_LOCAL we must have common * site scope. Don't set the local * scope since we may not share all * links, only loopback can do this. * Links on the local network would * also be on our private network * for v4 too. */ stcb->asoc.scope.ipv4_local_scope = 1; stcb->asoc.scope.site_scope = 1; } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) { /* * If the new destination is * SITE_LOCAL then we must have site * scope in common. */ stcb->asoc.scope.site_scope = 1; } } else { /* Validate the address is in scope */ if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) && (stcb->asoc.scope.loopback_scope == 0)) { addr_inscope = 0; } else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) && (stcb->asoc.scope.local_scope == 0)) { addr_inscope = 0; } else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) && (stcb->asoc.scope.site_scope == 0)) { addr_inscope = 0; } } break; } #endif default: /* not supported family type */ return (-1); } net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets); if (net == NULL) { return (-1); } SCTP_INCR_RADDR_COUNT(); bzero(net, sizeof(struct sctp_nets)); (void)SCTP_GETTIME_TIMEVAL(&net->start_time); memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len); switch (newaddr->sa_family) { #ifdef INET case AF_INET: ((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport; break; #endif #ifdef INET6 case AF_INET6: ((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport; break; #endif default: break; } net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id); if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) { stcb->asoc.scope.loopback_scope = 1; stcb->asoc.scope.ipv4_local_scope = 1; stcb->asoc.scope.local_scope = 0; stcb->asoc.scope.site_scope = 1; addr_inscope = 1; } net->failure_threshold = stcb->asoc.def_net_failure; net->pf_threshold = stcb->asoc.def_net_pf_threshold; if (addr_inscope == 0) { net->dest_state = (SCTP_ADDR_REACHABLE | SCTP_ADDR_OUT_OF_SCOPE); } else { if (from == SCTP_ADDR_IS_CONFIRMED) /* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */ net->dest_state = SCTP_ADDR_REACHABLE; else net->dest_state = SCTP_ADDR_REACHABLE | SCTP_ADDR_UNCONFIRMED; } /* * We set this to 0, the timer code knows that this means its an * initial value */ net->rto_needed = 1; net->RTO = 0; net->RTO_measured = 0; stcb->asoc.numnets++; net->ref_count = 1; net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1; net->port = stcb->asoc.port; net->dscp = stcb->asoc.default_dscp; #ifdef INET6 net->flowlabel = stcb->asoc.default_flowlabel; #endif if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) { net->dest_state |= SCTP_ADDR_NOHB; } else { net->dest_state &= ~SCTP_ADDR_NOHB; } if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) { net->dest_state |= SCTP_ADDR_NO_PMTUD; } else { net->dest_state &= ~SCTP_ADDR_NO_PMTUD; } net->heart_beat_delay = stcb->asoc.heart_beat_delay; /* Init the timer structure */ SCTP_OS_TIMER_INIT(&net->rxt_timer.timer); SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer); SCTP_OS_TIMER_INIT(&net->hb_timer.timer); /* Now generate a route for this guy */ #ifdef INET6 /* KAME hack: embed scopeid */ if (newaddr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; (void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)); sin6->sin6_scope_id = 0; } #endif SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id, stcb->sctp_ep->fibnum); if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) { /* Get source address */ net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep, stcb, (sctp_route_t *) & net->ro, net, 0, stcb->asoc.vrf_id); if (net->ro._s_addr != NULL) { net->src_addr_selected = 1; /* Now get the interface MTU */ if (net->ro._s_addr->ifn_p != NULL) { net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p); } } else { net->src_addr_selected = 0; } if (net->mtu > 0) { uint32_t rmtu; rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt); if (rmtu == 0) { /* * Start things off to match mtu of * interface please. */ SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa, net->ro.ro_rt, net->mtu); } else { /* * we take the route mtu over the interface, * since the route may be leading out the * loopback, or a different interface. */ net->mtu = rmtu; } } } else { net->src_addr_selected = 0; } if (net->mtu == 0) { switch (newaddr->sa_family) { #ifdef INET case AF_INET: net->mtu = SCTP_DEFAULT_MTU; break; #endif #ifdef INET6 case AF_INET6: net->mtu = 1280; break; #endif default: break; } } #if defined(INET) || defined(INET6) if (net->port) { net->mtu -= (uint32_t) sizeof(struct udphdr); } #endif if (from == SCTP_ALLOC_ASOC) { stcb->asoc.smallest_mtu = net->mtu; } if (stcb->asoc.smallest_mtu > net->mtu) { stcb->asoc.smallest_mtu = net->mtu; } #ifdef INET6 if (newaddr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&net->ro._l_addr; (void)sa6_recoverscope(sin6); } #endif /* JRS - Use the congestion control given in the CC module */ if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL) (*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net); /* * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning * of assoc (2005/06/27, iyengar@cis.udel.edu) */ net->find_pseudo_cumack = 1; net->find_rtx_pseudo_cumack = 1; /* Choose an initial flowid. */ net->flowid = stcb->asoc.my_vtag ^ ntohs(stcb->rport) ^ ntohs(stcb->sctp_ep->sctp_lport); net->flowtype = M_HASHTYPE_OPAQUE; if (netp) { *netp = net; } netfirst = TAILQ_FIRST(&stcb->asoc.nets); if (net->ro.ro_rt == NULL) { /* Since we have no route put it at the back */ TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next); } else if (netfirst == NULL) { /* We are the first one in the pool. */ TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); } else if (netfirst->ro.ro_rt == NULL) { /* * First one has NO route. Place this one ahead of the first * one. */ TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); } else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) { /* * This one has a different interface than the one at the * top of the list. Place it ahead. */ TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next); } else { /* * Ok we have the same interface as the first one. Move * forward until we find either a) one with a NULL route... * insert ahead of that b) one with a different ifp.. insert * after that. c) end of the list.. insert at the tail. */ struct sctp_nets *netlook; do { netlook = TAILQ_NEXT(netfirst, sctp_next); if (netlook == NULL) { /* End of the list */ TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next); break; } else if (netlook->ro.ro_rt == NULL) { /* next one has NO route */ TAILQ_INSERT_BEFORE(netfirst, net, sctp_next); break; } else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) { TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook, net, sctp_next); break; } /* Shift forward */ netfirst = netlook; } while (netlook != NULL); } /* got to have a primary set */ if (stcb->asoc.primary_destination == 0) { stcb->asoc.primary_destination = net; } else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) && (net->ro.ro_rt) && ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) { /* No route to current primary adopt new primary */ stcb->asoc.primary_destination = net; } /* Validate primary is first */ net = TAILQ_FIRST(&stcb->asoc.nets); if ((net != stcb->asoc.primary_destination) && (stcb->asoc.primary_destination)) { /* * first one on the list is NOT the primary sctp_cmpaddr() * is much more efficient if the primary is the first on the * list, make it so. */ TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); } return (0); } static uint32_t sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb) { uint32_t id; struct sctpasochead *head; struct sctp_tcb *lstcb; SCTP_INP_WLOCK(inp); try_again: if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { /* TSNH */ SCTP_INP_WUNLOCK(inp); return (0); } /* * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC, * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC. */ if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) { inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1; } id = inp->sctp_associd_counter; inp->sctp_associd_counter++; lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0); if (lstcb) { goto try_again; } head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)]; LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash); stcb->asoc.in_asocid_hash = 1; SCTP_INP_WUNLOCK(inp); return id; } /* * allocate an association and add it to the endpoint. The caller must be * careful to add all additional addresses once they are know right away or * else the assoc will be may experience a blackout scenario. */ struct sctp_tcb * sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr, int *error, uint32_t override_tag, uint32_t vrf_id, uint16_t o_streams, struct thread *p ) { /* note the p argument is only valid in unbound sockets */ struct sctp_tcb *stcb; struct sctp_association *asoc; struct sctpasochead *head; uint16_t rport; int err; /* * Assumption made here: Caller has done a * sctp_findassociation_ep_addr(ep, addr's); to make sure the * address does not exist already. */ if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) { /* Hit max assoc, sorry no more */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); *error = ENOBUFS; return (NULL); } if (firstaddr == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTP_INP_RLOCK(inp); if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) || (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) { /* * If its in the TCP pool, its NOT allowed to create an * association. The parent listener needs to call * sctp_aloc_assoc.. or the one-2-many socket. If a peeled * off, or connected one does this.. its an error. */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) || (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } } SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:"); #ifdef SCTP_DEBUG if (firstaddr) { SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr); switch (firstaddr->sa_family) { #ifdef INET case AF_INET: SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n", ntohs(((struct sockaddr_in *)firstaddr)->sin_port)); break; #endif #ifdef INET6 case AF_INET6: SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n", ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port)); break; #endif default: break; } } else { SCTPDBG(SCTP_DEBUG_PCB3, "None\n"); } #endif /* SCTP_DEBUG */ switch (firstaddr->sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin; sin = (struct sockaddr_in *)firstaddr; if ((ntohs(sin->sin_port) == 0) || (sin->sin_addr.s_addr == INADDR_ANY) || (sin->sin_addr.s_addr == INADDR_BROADCAST) || IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { /* Invalid address */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } rport = sin->sin_port; break; } #endif #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)firstaddr; if ((ntohs(sin6->sin6_port) == 0) || IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) || IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { /* Invalid address */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } rport = sin6->sin6_port; break; } #endif default: /* not supported family type */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTP_INP_RUNLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { /* * If you have not performed a bind, then we need to do the * ephemeral bind for you. */ if ((err = sctp_inpcb_bind(inp->sctp_socket, (struct sockaddr *)NULL, (struct sctp_ifa *)NULL, p ))) { /* bind error, probably perm */ *error = err; return (NULL); } } stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb); if (stcb == NULL) { /* out of memory? */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); *error = ENOMEM; return (NULL); } SCTP_INCR_ASOC_COUNT(); bzero(stcb, sizeof(*stcb)); asoc = &stcb->asoc; asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb); SCTP_TCB_LOCK_INIT(stcb); SCTP_TCB_SEND_LOCK_INIT(stcb); stcb->rport = rport; /* setup back pointer's */ stcb->sctp_ep = inp; stcb->sctp_socket = inp->sctp_socket; if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id, o_streams))) { /* failed */ SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); LIST_REMOVE(stcb, sctp_tcbasocidhash); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_DECR_ASOC_COUNT(); *error = err; return (NULL); } /* and the port */ SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) { /* inpcb freed while alloc going on */ SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); LIST_REMOVE(stcb, sctp_tcbasocidhash); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_INP_WUNLOCK(inp); SCTP_INP_INFO_WUNLOCK(); SCTP_DECR_ASOC_COUNT(); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); *error = EINVAL; return (NULL); } SCTP_TCB_LOCK(stcb); /* now that my_vtag is set, add it to the hash */ head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))]; /* put it in the bucket in the vtag hash of assoc's for the system */ LIST_INSERT_HEAD(head, stcb, sctp_asocs); SCTP_INP_INFO_WUNLOCK(); if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) { /* failure.. memory error? */ if (asoc->strmout) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); asoc->strmout = NULL; } if (asoc->mapping_array) { SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); asoc->mapping_array = NULL; } if (asoc->nr_mapping_array) { SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP); asoc->nr_mapping_array = NULL; } SCTP_DECR_ASOC_COUNT(); SCTP_TCB_UNLOCK(stcb); SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); LIST_REMOVE(stcb, sctp_tcbasocidhash); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_INP_WUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS); *error = ENOBUFS; return (NULL); } /* Init all the timers */ SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer); SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer); SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer); SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer); SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer); SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer); SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer); LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist); /* now file the port under the hash as well */ if (inp->sctp_tcbhash != NULL) { head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport, inp->sctp_hashmark)]; LIST_INSERT_HEAD(head, stcb, sctp_tcbhash); } SCTP_INP_WUNLOCK(inp); SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb); return (stcb); } void sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net) { struct sctp_association *asoc; asoc = &stcb->asoc; asoc->numnets--; TAILQ_REMOVE(&asoc->nets, net, sctp_next); if (net == asoc->primary_destination) { /* Reset primary */ struct sctp_nets *lnet; lnet = TAILQ_FIRST(&asoc->nets); /* * Mobility adaptation Ideally, if deleted destination is * the primary, it becomes a fast retransmission trigger by * the subsequent SET PRIMARY. (by micchie) */ if (sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_BASE) || sctp_is_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_FASTHANDOFF)) { SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n"); if (asoc->deleted_primary != NULL) { SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n"); goto out; } asoc->deleted_primary = net; atomic_add_int(&net->ref_count, 1); memset(&net->lastsa, 0, sizeof(net->lastsa)); memset(&net->lastsv, 0, sizeof(net->lastsv)); sctp_mobility_feature_on(stcb->sctp_ep, SCTP_MOBILITY_PRIM_DELETED); sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED, stcb->sctp_ep, stcb, NULL); } out: /* Try to find a confirmed primary */ asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0); } if (net == asoc->last_data_chunk_from) { /* Reset primary */ asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets); } if (net == asoc->last_control_chunk_from) { /* Clear net */ asoc->last_control_chunk_from = NULL; } if (net == stcb->asoc.alternate) { sctp_free_remote_addr(stcb->asoc.alternate); stcb->asoc.alternate = NULL; } sctp_free_remote_addr(net); } /* * remove a remote endpoint address from an association, it will fail if the * address does not exist. */ int sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr) { /* * Here we need to remove a remote address. This is quite simple, we * first find it in the list of address for the association * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE * on that item. Note we do not allow it to be removed if there are * no other addresses. */ struct sctp_association *asoc; struct sctp_nets *net, *nnet; asoc = &stcb->asoc; /* locate the address */ TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) { if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) { continue; } if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr, remaddr)) { /* we found the guy */ if (asoc->numnets < 2) { /* Must have at LEAST two remote addresses */ return (-1); } else { sctp_remove_net(stcb, net); return (0); } } } /* not found. */ return (-2); } void sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport) { struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; int found = 0; int i; chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if ((twait_block->vtag_block[i].v_tag == tag) && (twait_block->vtag_block[i].lport == lport) && (twait_block->vtag_block[i].rport == rport)) { twait_block->vtag_block[i].tv_sec_at_expire = 0; twait_block->vtag_block[i].v_tag = 0; twait_block->vtag_block[i].lport = 0; twait_block->vtag_block[i].rport = 0; found = 1; break; } } if (found) break; } } int sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport) { struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; int found = 0; int i; SCTP_INP_INFO_WLOCK(); chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if ((twait_block->vtag_block[i].v_tag == tag) && (twait_block->vtag_block[i].lport == lport) && (twait_block->vtag_block[i].rport == rport)) { found = 1; break; } } if (found) break; } SCTP_INP_INFO_WUNLOCK(); return (found); } void sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport) { struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; struct timeval now; int set, i; if (time == 0) { /* Its disabled */ return; } (void)SCTP_GETTIME_TIMEVAL(&now); chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; set = 0; LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { /* Block(s) present, lets find space, and expire on the fly */ for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if ((twait_block->vtag_block[i].v_tag == 0) && !set) { twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time; twait_block->vtag_block[i].v_tag = tag; twait_block->vtag_block[i].lport = lport; twait_block->vtag_block[i].rport = rport; set = 1; } else if ((twait_block->vtag_block[i].v_tag) && ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) { /* Audit expires this guy */ twait_block->vtag_block[i].tv_sec_at_expire = 0; twait_block->vtag_block[i].v_tag = 0; twait_block->vtag_block[i].lport = 0; twait_block->vtag_block[i].rport = 0; if (set == 0) { /* Reuse it for my new tag */ twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time; twait_block->vtag_block[i].v_tag = tag; twait_block->vtag_block[i].lport = lport; twait_block->vtag_block[i].rport = rport; set = 1; } } } if (set) { /* * We only do up to the block where we can place our * tag for audits */ break; } } /* Need to add a new block to chain */ if (!set) { SCTP_MALLOC(twait_block, struct sctp_tagblock *, sizeof(struct sctp_tagblock), SCTP_M_TIMW); if (twait_block == NULL) { #ifdef INVARIANTS panic("Can not alloc tagblock"); #endif return; } memset(twait_block, 0, sizeof(struct sctp_tagblock)); LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock); twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time; twait_block->vtag_block[0].v_tag = tag; twait_block->vtag_block[0].lport = lport; twait_block->vtag_block[0].rport = rport; } } /*- * Free the association after un-hashing the remote port. This * function ALWAYS returns holding NO LOCK on the stcb. It DOES * expect that the input to this function IS a locked TCB. * It will return 0, if it did NOT destroy the association (instead * it unlocks it. It will return NON-zero if it either destroyed the * association OR the association is already destroyed. */ int sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location) { int i; struct sctp_association *asoc; struct sctp_nets *net, *nnet; struct sctp_laddr *laddr, *naddr; struct sctp_tmit_chunk *chk, *nchk; struct sctp_asconf_addr *aparam, *naparam; struct sctp_asconf_ack *aack, *naack; struct sctp_stream_reset_list *strrst, *nstrrst; struct sctp_queued_to_read *sq, *nsq; struct sctp_stream_queue_pending *sp, *nsp; sctp_sharedkey_t *shared_key, *nshared_key; struct socket *so; /* first, lets purge the entry from the hash table. */ #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 6); #endif if (stcb->asoc.state == 0) { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 7); #endif /* there is no asoc, really TSNH :-0 */ return (1); } if (stcb->asoc.alternate) { sctp_free_remote_addr(stcb->asoc.alternate); stcb->asoc.alternate = NULL; } /* TEMP CODE */ if (stcb->freed_from_where == 0) { /* Only record the first place free happened from */ stcb->freed_from_where = from_location; } /* TEMP CODE */ asoc = &stcb->asoc; if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) /* nothing around */ so = NULL; else so = inp->sctp_socket; /* * We used timer based freeing if a reader or writer is in the way. * So we first check if we are actually being called from a timer, * if so we abort early if a reader or writer is still in the way. */ if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) && (from_inpcbfree == SCTP_NORMAL_PROC)) { /* * is it the timer driving us? if so are the reader/writers * gone? */ if (stcb->asoc.refcnt) { /* nope, reader or writer in the way */ sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); /* no asoc destroyed */ SCTP_TCB_UNLOCK(stcb); #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 8); #endif return (0); } } /* now clean up any other timers */ (void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer); asoc->dack_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); /*- * For stream reset we don't blast this unless * it is a str-reset timer, it might be the * free-asoc timer which we DON'T want to * disturb. */ if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET) asoc->strreset_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer); asoc->asconf_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer); asoc->autoclose_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer); asoc->shut_guard_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer); asoc->delayed_event_timer.self = NULL; /* Mobility adaptation */ (void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer); asoc->delete_prim_timer.self = NULL; TAILQ_FOREACH(net, &asoc->nets, sctp_next) { (void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer); net->rxt_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer); net->pmtu_timer.self = NULL; (void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer); net->hb_timer.self = NULL; } /* Now the read queue needs to be cleaned up (only once) */ if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) { stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED; SCTP_INP_READ_LOCK(inp); TAILQ_FOREACH(sq, &inp->read_queue, next) { if (sq->stcb == stcb) { sq->do_not_ref_stcb = 1; sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn; /* * If there is no end, there never will be * now. */ if (sq->end_added == 0) { /* Held for PD-API clear that. */ sq->pdapi_aborted = 1; sq->held_length = 0; if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) { /* * Need to add a PD-API * aborted indication. * Setting the control_pdapi * assures that it will be * added right after this * msg. */ uint32_t strseq; stcb->asoc.control_pdapi = sq; strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn; sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION, stcb, SCTP_PARTIAL_DELIVERY_ABORTED, (void *)&strseq, SCTP_SO_LOCKED); stcb->asoc.control_pdapi = NULL; } } /* Add an end to wake them */ sq->end_added = 1; } } SCTP_INP_READ_UNLOCK(inp); if (stcb->block_entry) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET); stcb->block_entry->error = ECONNRESET; stcb->block_entry = NULL; } } if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) { /* * Someone holds a reference OR the socket is unaccepted * yet. */ if ((stcb->asoc.refcnt) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE; sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); } SCTP_TCB_UNLOCK(stcb); if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) /* nothing around */ so = NULL; if (so) { /* Wake any reader/writers */ sctp_sorwakeup(inp, so); sctp_sowwakeup(inp, so); } #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 9); #endif /* no asoc destroyed */ return (0); } #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, stcb, 10); #endif /* * When I reach here, no others want to kill the assoc yet.. and I * own the lock. Now its possible an abort comes in when I do the * lock exchange below to grab all the locks to do the final take * out. to prevent this we increment the count, which will start a * timer and blow out above thus assuring us that we hold exclusive * killing of the asoc. Note that after getting back the TCB lock we * will go ahead and increment the counter back up and stop any * timer a passing stranger may have started :-S */ if (from_inpcbfree == SCTP_NORMAL_PROC) { atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_INP_INFO_WLOCK(); SCTP_INP_WLOCK(inp); SCTP_TCB_LOCK(stcb); } /* Double check the GONE flag */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) /* nothing around */ so = NULL; if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* * For TCP type we need special handling when we are * connected. We also include the peel'ed off ones to. */ if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED; inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED; if (so) { SOCK_LOCK(so); if (so->so_rcv.sb_cc == 0) { so->so_state &= ~(SS_ISCONNECTING | SS_ISDISCONNECTING | SS_ISCONFIRMING | SS_ISCONNECTED); } socantrcvmore_locked(so); sctp_sowwakeup(inp, so); sctp_sorwakeup(inp, so); SCTP_SOWAKEUP(so); } } } /* * Make it invalid too, that way if its about to run it will abort * and return. */ /* re-increment the lock */ if (from_inpcbfree == SCTP_NORMAL_PROC) { atomic_add_int(&stcb->asoc.refcnt, -1); } if (stcb->asoc.refcnt) { stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE; sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL); if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_INFO_WUNLOCK(); SCTP_INP_WUNLOCK(inp); } SCTP_TCB_UNLOCK(stcb); return (0); } asoc->state = 0; if (inp->sctp_tcbhash) { LIST_REMOVE(stcb, sctp_tcbhash); } if (stcb->asoc.in_asocid_hash) { LIST_REMOVE(stcb, sctp_tcbasocidhash); } /* Now lets remove it from the list of ALL associations in the EP */ LIST_REMOVE(stcb, sctp_tcblist); if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_INCR_REF(inp); SCTP_INP_WUNLOCK(inp); } /* pull from vtag hash */ LIST_REMOVE(stcb, sctp_asocs); sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait), inp->sctp_lport, stcb->rport); /* * Now restop the timers to be sure this is paranoia at is finest! */ (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer); (void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer); TAILQ_FOREACH(net, &asoc->nets, sctp_next) { (void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer); (void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer); (void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer); } asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE; /* * The chunk lists and such SHOULD be empty but we check them just * in case. */ /* anything on the wheel needs to be removed */ for (i = 0; i < asoc->streamoutcnt; i++) { struct sctp_stream_out *outs; outs = &asoc->strmout[i]; /* now clean up any chunks here */ TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) { TAILQ_REMOVE(&outs->outqueue, sp, next); sctp_free_spbufspace(stcb, asoc, sp); if (sp->data) { if (so) { /* Still an open socket - report */ sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb, 0, (void *)sp, SCTP_SO_LOCKED); } if (sp->data) { sctp_m_freem(sp->data); sp->data = NULL; sp->tail_mbuf = NULL; sp->length = 0; } } if (sp->net) { sctp_free_remote_addr(sp->net); sp->net = NULL; } sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED); } } /* sa_ignore FREED_MEMORY */ TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) { TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp); SCTP_FREE(strrst, SCTP_M_STRESET); } TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) { TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next); if (sq->data) { sctp_m_freem(sq->data); sq->data = NULL; } sctp_free_remote_addr(sq->whoFrom); sq->whoFrom = NULL; sq->stcb = NULL; /* Free the ctl entry */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq); SCTP_DECR_READQ_COUNT(); /* sa_ignore FREED_MEMORY */ } TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) { TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); asoc->free_chunk_cnt--; /* sa_ignore FREED_MEMORY */ } /* pending send queue SHOULD be empty */ TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) { if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) { asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--; #ifdef INVARIANTS } else { panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number); #endif } TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next); if (chk->data) { if (so) { /* Still a socket? */ sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb, 0, chk, SCTP_SO_LOCKED); } if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); if (chk->whoTo) { sctp_free_remote_addr(chk->whoTo); chk->whoTo = NULL; } SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ } /* sent queue SHOULD be empty */ TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) { if (chk->sent != SCTP_DATAGRAM_NR_ACKED) { if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) { asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--; #ifdef INVARIANTS } else { panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number); #endif } } TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next); if (chk->data) { if (so) { /* Still a socket? */ sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb, 0, chk, SCTP_SO_LOCKED); } if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ } #ifdef INVARIANTS for (i = 0; i < stcb->asoc.streamoutcnt; i++) { if (stcb->asoc.strmout[i].chunks_on_queues > 0) { panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i); } } #endif /* control queue MAY not be empty */ TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ } /* ASCONF queue MAY not be empty */ TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ } TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) { TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } if (chk->holds_key_ref) sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED); sctp_free_remote_addr(chk->whoTo); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk); SCTP_DECR_CHK_COUNT(); /* sa_ignore FREED_MEMORY */ } if (asoc->mapping_array) { SCTP_FREE(asoc->mapping_array, SCTP_M_MAP); asoc->mapping_array = NULL; } if (asoc->nr_mapping_array) { SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP); asoc->nr_mapping_array = NULL; } /* the stream outs */ if (asoc->strmout) { SCTP_FREE(asoc->strmout, SCTP_M_STRMO); asoc->strmout = NULL; } asoc->strm_realoutsize = asoc->streamoutcnt = 0; if (asoc->strmin) { struct sctp_queued_to_read *ctl, *nctl; for (i = 0; i < asoc->streamincnt; i++) { TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) { TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next); sctp_free_remote_addr(ctl->whoFrom); if (ctl->data) { sctp_m_freem(ctl->data); ctl->data = NULL; } /* * We don't free the address here since all * the net's were freed above. */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl); SCTP_DECR_READQ_COUNT(); } } SCTP_FREE(asoc->strmin, SCTP_M_STRMI); asoc->strmin = NULL; } asoc->streamincnt = 0; TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) { #ifdef INVARIANTS if (SCTP_BASE_INFO(ipi_count_raddr) == 0) { panic("no net's left alloc'ed, or list points to itself"); } #endif TAILQ_REMOVE(&asoc->nets, net, sctp_next); sctp_free_remote_addr(net); } LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) { /* sa_ignore FREED_MEMORY */ sctp_remove_laddr(laddr); } /* pending asconf (address) parameters */ TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) { /* sa_ignore FREED_MEMORY */ TAILQ_REMOVE(&asoc->asconf_queue, aparam, next); SCTP_FREE(aparam, SCTP_M_ASC_ADDR); } TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) { /* sa_ignore FREED_MEMORY */ TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next); if (aack->data != NULL) { sctp_m_freem(aack->data); } SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack); } /* clean up auth stuff */ if (asoc->local_hmacs) sctp_free_hmaclist(asoc->local_hmacs); if (asoc->peer_hmacs) sctp_free_hmaclist(asoc->peer_hmacs); if (asoc->local_auth_chunks) sctp_free_chunklist(asoc->local_auth_chunks); if (asoc->peer_auth_chunks) sctp_free_chunklist(asoc->peer_auth_chunks); sctp_free_authinfo(&asoc->authinfo); LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) { LIST_REMOVE(shared_key, next); sctp_free_sharedkey(shared_key); /* sa_ignore FREED_MEMORY */ } /* Insert new items here :> */ /* Get rid of LOCK */ SCTP_TCB_UNLOCK(stcb); SCTP_TCB_LOCK_DESTROY(stcb); SCTP_TCB_SEND_LOCK_DESTROY(stcb); if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_INFO_WUNLOCK(); SCTP_INP_RLOCK(inp); } #ifdef SCTP_TRACK_FREED_ASOCS if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* now clean up the tasoc itself */ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_DECR_ASOC_COUNT(); } else { LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist); } #else SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb); SCTP_DECR_ASOC_COUNT(); #endif if (from_inpcbfree == SCTP_NORMAL_PROC) { if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) { /* * If its NOT the inp_free calling us AND sctp_close * as been called, we call back... */ SCTP_INP_RUNLOCK(inp); /* * This will start the kill timer (if we are the * last one) since we hold an increment yet. But * this is the only safe way to do this since * otherwise if the socket closes at the same time * we are here we might collide in the cleanup. */ sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE, SCTP_CALLED_DIRECTLY_NOCMPSET); SCTP_INP_DECR_REF(inp); goto out_of; } else { /* The socket is still open. */ SCTP_INP_DECR_REF(inp); } } if (from_inpcbfree == SCTP_NORMAL_PROC) { SCTP_INP_RUNLOCK(inp); } out_of: /* destroyed the asoc */ #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 11); #endif return (1); } /* * determine if a destination is "reachable" based upon the addresses bound * to the current endpoint (e.g. only v4 or v6 currently bound) */ /* * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use * assoc level v4/v6 flags, as the assoc *may* not have the same address * types bound as its endpoint */ int sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr) { struct sctp_inpcb *inp; int answer; /* * No locks here, the TCB, in all cases is already locked and an * assoc is up. There is either a INP lock by the caller applied (in * asconf case when deleting an address) or NOT in the HB case, * however if HB then the INP increment is up and the INP will not * be removed (on top of the fact that we have a TCB lock). So we * only want to read the sctp_flags, which is either bound-all or * not.. no protection needed since once an assoc is up you can't be * changing your binding. */ inp = stcb->sctp_ep; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* if bound all, destination is not restricted */ /* * RRS: Question during lock work: Is this correct? If you * are bound-all you still might need to obey the V4--V6 * flags??? IMO this bound-all stuff needs to be removed! */ return (1); } /* NOTE: all "scope" checks are done when local addresses are added */ switch (destaddr->sa_family) { #ifdef INET6 case AF_INET6: answer = inp->ip_inp.inp.inp_vflag & INP_IPV6; break; #endif #ifdef INET case AF_INET: answer = inp->ip_inp.inp.inp_vflag & INP_IPV4; break; #endif default: /* invalid family, so it's unreachable */ answer = 0; break; } return (answer); } /* * update the inp_vflags on an endpoint */ static void sctp_update_ep_vflag(struct sctp_inpcb *inp) { struct sctp_laddr *laddr; /* first clear the flag */ inp->ip_inp.inp.inp_vflag = 0; /* set the flag based on addresses on the ep list */ LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __func__); continue; } if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) { continue; } switch (laddr->ifa->address.sa.sa_family) { #ifdef INET6 case AF_INET6: inp->ip_inp.inp.inp_vflag |= INP_IPV6; break; #endif #ifdef INET case AF_INET: inp->ip_inp.inp.inp_vflag |= INP_IPV4; break; #endif default: break; } } } /* * Add the address to the endpoint local address list There is nothing to be * done if we are bound to all addresses */ void sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action) { struct sctp_laddr *laddr; int fnd, error = 0; fnd = 0; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* You are already bound to all. You have it already */ return; } #ifdef INET6 if (ifa->address.sa.sa_family == AF_INET6) { if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { /* Can't bind a non-useable addr. */ return; } } #endif /* first, is it already present? */ LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == ifa) { fnd = 1; break; } } if (fnd == 0) { /* Not in the ep list */ error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action); if (error != 0) return; inp->laddr_count++; /* update inp_vflag flags */ switch (ifa->address.sa.sa_family) { #ifdef INET6 case AF_INET6: inp->ip_inp.inp.inp_vflag |= INP_IPV6; break; #endif #ifdef INET case AF_INET: inp->ip_inp.inp.inp_vflag |= INP_IPV4; break; #endif default: break; } } return; } /* * select a new (hopefully reachable) destination net (should only be used * when we deleted an ep addr that is the only usable source address to reach * the destination net) */ static void sctp_select_primary_destination(struct sctp_tcb *stcb) { struct sctp_nets *net; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { /* for now, we'll just pick the first reachable one we find */ if (net->dest_state & SCTP_ADDR_UNCONFIRMED) continue; if (sctp_destination_is_reachable(stcb, (struct sockaddr *)&net->ro._l_addr)) { /* found a reachable destination */ stcb->asoc.primary_destination = net; } } /* I can't there from here! ...we're gonna die shortly... */ } /* * Delete the address from the endpoint local address list There is nothing * to be done if we are bound to all addresses */ void sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa) { struct sctp_laddr *laddr; int fnd; fnd = 0; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* You are already bound to all. You have it already */ return; } LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == ifa) { fnd = 1; break; } } if (fnd && (inp->laddr_count < 2)) { /* can't delete unless there are at LEAST 2 addresses */ return; } if (fnd) { /* * clean up any use of this address go through our * associations and clear any last_used_address that match * this one for each assoc, see if a new primary_destination * is needed */ struct sctp_tcb *stcb; /* clean up "next_addr_touse" */ if (inp->next_addr_touse == laddr) /* delete this address */ inp->next_addr_touse = NULL; /* clean up "last_used_address" */ LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { struct sctp_nets *net; SCTP_TCB_LOCK(stcb); if (stcb->asoc.last_used_address == laddr) /* delete this address */ stcb->asoc.last_used_address = NULL; /* * Now spin through all the nets and purge any ref * to laddr */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (net->ro._s_addr && (net->ro._s_addr->ifa == laddr->ifa)) { /* Yep, purge src address selected */ sctp_rtentry_t *rt; /* delete this address if cached */ rt = net->ro.ro_rt; if (rt != NULL) { RTFREE(rt); net->ro.ro_rt = NULL; } sctp_free_ifa(net->ro._s_addr); net->ro._s_addr = NULL; net->src_addr_selected = 0; } } SCTP_TCB_UNLOCK(stcb); } /* for each tcb */ /* remove it from the ep list */ sctp_remove_laddr(laddr); inp->laddr_count--; /* update inp_vflag flags */ sctp_update_ep_vflag(inp); } return; } /* * Add the address to the TCB local address restricted list. * This is a "pending" address list (eg. addresses waiting for an * ASCONF-ACK response) and cannot be used as a valid source address. */ void sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) { struct sctp_laddr *laddr; struct sctpladdr *list; /* * Assumes TCB is locked.. and possibly the INP. May need to * confirm/fix that if we need it and is not the case. */ list = &stcb->asoc.sctp_restricted_addrs; #ifdef INET6 if (ifa->address.sa.sa_family == AF_INET6) { if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) { /* Can't bind a non-existent addr. */ return; } } #endif /* does the address already exist? */ LIST_FOREACH(laddr, list, sctp_nxt_addr) { if (laddr->ifa == ifa) { return; } } /* add to the list */ (void)sctp_insert_laddr(list, ifa, 0); return; } /* * insert an laddr entry with the given ifa for the desired list */ int sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act) { struct sctp_laddr *laddr; laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr); if (laddr == NULL) { /* out of memory? */ SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL); return (EINVAL); } SCTP_INCR_LADDR_COUNT(); bzero(laddr, sizeof(*laddr)); (void)SCTP_GETTIME_TIMEVAL(&laddr->start_time); laddr->ifa = ifa; laddr->action = act; atomic_add_int(&ifa->refcount, 1); /* insert it */ LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr); return (0); } /* * Remove an laddr entry from the local address list (on an assoc) */ void sctp_remove_laddr(struct sctp_laddr *laddr) { /* remove from the list */ LIST_REMOVE(laddr, sctp_nxt_addr); sctp_free_ifa(laddr->ifa); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr); SCTP_DECR_LADDR_COUNT(); } /* * Remove a local address from the TCB local address restricted list */ void sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa) { struct sctp_inpcb *inp; struct sctp_laddr *laddr; /* * This is called by asconf work. It is assumed that a) The TCB is * locked and b) The INP is locked. This is true in as much as I can * trace through the entry asconf code where I did these locks. * Again, the ASCONF code is a bit different in that it does lock * the INP during its work often times. This must be since we don't * want other proc's looking up things while what they are looking * up is changing :-D */ inp = stcb->sctp_ep; /* if subset bound and don't allow ASCONF's, can't delete last */ if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) && sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) { if (stcb->sctp_ep->laddr_count < 2) { /* can't delete last address */ return; } } LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) { /* remove the address if it exists */ if (laddr->ifa == NULL) continue; if (laddr->ifa == ifa) { sctp_remove_laddr(laddr); return; } } /* address not found! */ return; } /* * Temporarily remove for __APPLE__ until we use the Tiger equivalents */ /* sysctl */ static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC; static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR; #if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP) struct sctp_mcore_ctrl *sctp_mcore_workers = NULL; int *sctp_cpuarry = NULL; void sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use) { /* Queue a packet to a processor for the specified core */ struct sctp_mcore_queue *qent; struct sctp_mcore_ctrl *wkq; int need_wake = 0; if (sctp_mcore_workers == NULL) { /* Something went way bad during setup */ sctp_input_with_port(m, off, 0); return; } SCTP_MALLOC(qent, struct sctp_mcore_queue *, (sizeof(struct sctp_mcore_queue)), SCTP_M_MCORE); if (qent == NULL) { /* This is trouble */ sctp_input_with_port(m, off, 0); return; } qent->vn = curvnet; qent->m = m; qent->off = off; qent->v6 = 0; wkq = &sctp_mcore_workers[cpu_to_use]; SCTP_MCORE_QLOCK(wkq); TAILQ_INSERT_TAIL(&wkq->que, qent, next); if (wkq->running == 0) { need_wake = 1; } SCTP_MCORE_QUNLOCK(wkq); if (need_wake) { wakeup(&wkq->running); } } static void sctp_mcore_thread(void *arg) { struct sctp_mcore_ctrl *wkq; struct sctp_mcore_queue *qent; wkq = (struct sctp_mcore_ctrl *)arg; struct mbuf *m; int off, v6; /* Wait for first tickle */ SCTP_MCORE_LOCK(wkq); wkq->running = 0; msleep(&wkq->running, &wkq->core_mtx, 0, "wait for pkt", 0); SCTP_MCORE_UNLOCK(wkq); /* Bind to our cpu */ thread_lock(curthread); sched_bind(curthread, wkq->cpuid); thread_unlock(curthread); /* Now lets start working */ SCTP_MCORE_LOCK(wkq); /* Now grab lock and go */ for (;;) { SCTP_MCORE_QLOCK(wkq); skip_sleep: wkq->running = 1; qent = TAILQ_FIRST(&wkq->que); if (qent) { TAILQ_REMOVE(&wkq->que, qent, next); SCTP_MCORE_QUNLOCK(wkq); CURVNET_SET(qent->vn); m = qent->m; off = qent->off; v6 = qent->v6; SCTP_FREE(qent, SCTP_M_MCORE); if (v6 == 0) { sctp_input_with_port(m, off, 0); } else { SCTP_PRINTF("V6 not yet supported\n"); sctp_m_freem(m); } CURVNET_RESTORE(); SCTP_MCORE_QLOCK(wkq); } wkq->running = 0; if (!TAILQ_EMPTY(&wkq->que)) { goto skip_sleep; } SCTP_MCORE_QUNLOCK(wkq); msleep(&wkq->running, &wkq->core_mtx, 0, "wait for pkt", 0); } } static void sctp_startup_mcore_threads(void) { int i, cpu; if (mp_ncpus == 1) return; if (sctp_mcore_workers != NULL) { /* * Already been here in some previous vnet? */ return; } SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *, ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)), SCTP_M_MCORE); if (sctp_mcore_workers == NULL) { /* TSNH I hope */ return; } memset(sctp_mcore_workers, 0, ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl))); /* Init the structures */ for (i = 0; i <= mp_maxid; i++) { TAILQ_INIT(&sctp_mcore_workers[i].que); SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]); SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]); sctp_mcore_workers[i].cpuid = i; } if (sctp_cpuarry == NULL) { SCTP_MALLOC(sctp_cpuarry, int *, (mp_ncpus * sizeof(int)), SCTP_M_MCORE); i = 0; CPU_FOREACH(cpu) { sctp_cpuarry[i] = cpu; i++; } } /* Now start them all */ CPU_FOREACH(cpu) { (void)kproc_create(sctp_mcore_thread, (void *)&sctp_mcore_workers[cpu], &sctp_mcore_workers[cpu].thread_proc, RFPROC, SCTP_KTHREAD_PAGES, SCTP_MCORE_NAME); } } #endif void sctp_pcb_init() { /* * SCTP initialization for the PCB structures should be called by * the sctp_init() funciton. */ int i; struct timeval tv; if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) { /* error I was called twice */ return; } SCTP_BASE_VAR(sctp_pcb_initialized) = 1; #if defined(SCTP_LOCAL_TRACE_BUF) bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log)); #endif #if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT) SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *, ((mp_maxid + 1) * sizeof(struct sctpstat)), SCTP_M_MCORE); #endif (void)SCTP_GETTIME_TIMEVAL(&tv); #if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT) bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1))); SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec; SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec; #else bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat)); SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec; SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec; #endif /* init the empty list of (All) Endpoints */ LIST_INIT(&SCTP_BASE_INFO(listhead)); /* init the hash table of endpoints */ TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize)); TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize)); TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale)); SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31), &SCTP_BASE_INFO(hashasocmark)); SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize), &SCTP_BASE_INFO(hashmark)); SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize), &SCTP_BASE_INFO(hashtcpmark)); SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize); SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH, &SCTP_BASE_INFO(hashvrfmark)); SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE, &SCTP_BASE_INFO(vrf_ifn_hashmark)); /* init the zones */ /* * FIX ME: Should check for NULL returns, but if it does fail we are * doomed to panic anyways... add later maybe. */ SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep", sizeof(struct sctp_inpcb), maxsockets); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc", sizeof(struct sctp_tcb), sctp_max_number_of_assoc); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr", sizeof(struct sctp_laddr), (sctp_max_number_of_assoc * sctp_scale_up_for_address)); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr", sizeof(struct sctp_nets), (sctp_max_number_of_assoc * sctp_scale_up_for_address)); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk", sizeof(struct sctp_tmit_chunk), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq", sizeof(struct sctp_queued_to_read), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out", sizeof(struct sctp_stream_queue_pending), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf", sizeof(struct sctp_asconf), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack", sizeof(struct sctp_asconf_ack), (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale))); /* Master Lock INIT for info structure */ SCTP_INP_INFO_LOCK_INIT(); SCTP_STATLOG_INIT_LOCK(); SCTP_IPI_COUNT_INIT(); SCTP_IPI_ADDR_INIT(); #ifdef SCTP_PACKET_LOGGING SCTP_IP_PKTLOG_INIT(); #endif LIST_INIT(&SCTP_BASE_INFO(addr_wq)); SCTP_WQ_ADDR_INIT(); /* not sure if we need all the counts */ SCTP_BASE_INFO(ipi_count_ep) = 0; /* assoc/tcb zone info */ SCTP_BASE_INFO(ipi_count_asoc) = 0; /* local addrlist zone info */ SCTP_BASE_INFO(ipi_count_laddr) = 0; /* remote addrlist zone info */ SCTP_BASE_INFO(ipi_count_raddr) = 0; /* chunk info */ SCTP_BASE_INFO(ipi_count_chunk) = 0; /* socket queue zone info */ SCTP_BASE_INFO(ipi_count_readq) = 0; /* stream out queue cont */ SCTP_BASE_INFO(ipi_count_strmoq) = 0; SCTP_BASE_INFO(ipi_free_strmoq) = 0; SCTP_BASE_INFO(ipi_free_chunks) = 0; SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer)); /* Init the TIMEWAIT list */ for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) { LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]); } sctp_startup_iterator(); #if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP) sctp_startup_mcore_threads(); #endif /* * INIT the default VRF which for BSD is the only one, other O/S's * may have more. But initially they must start with one and then * add the VRF's as addresses are added. */ sctp_init_vrf_list(SCTP_DEFAULT_VRF); } /* * Assumes that the SCTP_BASE_INFO() lock is NOT held. */ void sctp_pcb_finish(void) { struct sctp_vrflist *vrf_bucket; struct sctp_vrf *vrf, *nvrf; struct sctp_ifn *ifn, *nifn; struct sctp_ifa *ifa, *nifa; struct sctpvtaghead *chain; struct sctp_tagblock *twait_block, *prev_twait_block; struct sctp_laddr *wi, *nwi; int i; struct sctp_iterator *it, *nit; + if (SCTP_BASE_VAR(sctp_pcb_initialized) == 0) { + printf("%s: race condition on teardown.\n", __func__); + return; + } + SCTP_BASE_VAR(sctp_pcb_initialized) = 0; + /* * In FreeBSD the iterator thread never exits but we do clean up. * The only way FreeBSD reaches here is if we have VRF's but we * still add the ifdef to make it compile on old versions. */ +retry: + while (sctp_it_ctl.iterator_running != 0) + DELAY(1); SCTP_IPI_ITERATOR_WQ_LOCK(); + /* + * sctp_iterator_worker() might be working on an it entry without + * holding the lock. We won't find it on the list either and + * continue and free/destroy it. While holding the lock, spin, to + * avoid the race condition as sctp_iterator_worker() will have to + * wait to re-aquire the lock. + */ + if (sctp_it_ctl.cur_it != NULL || sctp_it_ctl.iterator_running != 0) { + SCTP_IPI_ITERATOR_WQ_UNLOCK(); + printf("%s: Iterator running while we held the lock. Retry.\n", + __func__); + goto retry; + } TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) { if (it->vn != curvnet) { continue; } TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); if (it->function_atend != NULL) { (*it->function_atend) (it->pointer, it->val); } SCTP_FREE(it, SCTP_M_ITER); } SCTP_IPI_ITERATOR_WQ_UNLOCK(); SCTP_ITERATOR_LOCK(); if ((sctp_it_ctl.cur_it) && (sctp_it_ctl.cur_it->vn == curvnet)) { sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT; } SCTP_ITERATOR_UNLOCK(); - SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer)); + SCTP_OS_TIMER_STOP_DRAIN(&SCTP_BASE_INFO(addr_wq_timer.timer)); SCTP_WQ_ADDR_LOCK(); LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) { LIST_REMOVE(wi, sctp_nxt_addr); SCTP_DECR_LADDR_COUNT(); if (wi->action == SCTP_DEL_IP_ADDRESS) { SCTP_FREE(wi->ifa, SCTP_M_IFA); } SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi); } SCTP_WQ_ADDR_UNLOCK(); /* * free the vrf/ifn/ifa lists and hashes (be sure address monitor is * destroyed first). */ vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))]; LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) { LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) { LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) { /* free the ifa */ LIST_REMOVE(ifa, next_bucket); LIST_REMOVE(ifa, next_ifa); SCTP_FREE(ifa, SCTP_M_IFA); } /* free the ifn */ LIST_REMOVE(ifn, next_bucket); LIST_REMOVE(ifn, next_ifn); SCTP_FREE(ifn, SCTP_M_IFN); } SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark); /* free the vrf */ LIST_REMOVE(vrf, next_vrf); SCTP_FREE(vrf, SCTP_M_VRF); } /* free the vrf hashes */ SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark)); SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark)); /* * free the TIMEWAIT list elements malloc'd in the function * sctp_add_vtag_to_timewait()... */ for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) { chain = &SCTP_BASE_INFO(vtag_timewait)[i]; if (!LIST_EMPTY(chain)) { prev_twait_block = NULL; LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { if (prev_twait_block) { SCTP_FREE(prev_twait_block, SCTP_M_TIMW); } prev_twait_block = twait_block; } SCTP_FREE(prev_twait_block, SCTP_M_TIMW); } } /* free the locks and mutexes */ #ifdef SCTP_PACKET_LOGGING SCTP_IP_PKTLOG_DESTROY(); #endif SCTP_IPI_ADDR_DESTROY(); SCTP_STATLOG_DESTROY(); SCTP_INP_INFO_LOCK_DESTROY(); SCTP_WQ_ADDR_DESTROY(); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf)); SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack)); /* Get rid of other stuff to */ if (SCTP_BASE_INFO(sctp_asochash) != NULL) SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark)); if (SCTP_BASE_INFO(sctp_ephash) != NULL) SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark)); if (SCTP_BASE_INFO(sctp_tcpephash) != NULL) SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark)); #if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT) SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE); #endif } int sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m, int offset, int limit, struct sockaddr *src, struct sockaddr *dst, struct sockaddr *altsa) { /* * grub through the INIT pulling addresses and loading them to the * nets structure in the asoc. The from address in the mbuf should * also be loaded (if it is not already). This routine can be called * with either INIT or INIT-ACK's as long as the m points to the IP * packet and the offset points to the beginning of the parameters. */ struct sctp_inpcb *inp; struct sctp_nets *net, *nnet, *net_tmp; struct sctp_paramhdr *phdr, parm_buf; struct sctp_tcb *stcb_tmp; uint16_t ptype, plen; struct sockaddr *sa; uint8_t random_store[SCTP_PARAM_BUFFER_SIZE]; struct sctp_auth_random *p_random = NULL; uint16_t random_len = 0; uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE]; struct sctp_auth_hmac_algo *hmacs = NULL; uint16_t hmacs_len = 0; uint8_t saw_asconf = 0; uint8_t saw_asconf_ack = 0; uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE]; struct sctp_auth_chunk_list *chunks = NULL; uint16_t num_chunks = 0; sctp_key_t *new_key; uint32_t keylen; int got_random = 0, got_hmacs = 0, got_chklist = 0; uint8_t peer_supports_ecn; uint8_t peer_supports_prsctp; uint8_t peer_supports_auth; uint8_t peer_supports_asconf; uint8_t peer_supports_asconf_ack; uint8_t peer_supports_reconfig; uint8_t peer_supports_nrsack; uint8_t peer_supports_pktdrop; #ifdef INET struct sockaddr_in sin; #endif #ifdef INET6 struct sockaddr_in6 sin6; #endif /* First get the destination address setup too. */ #ifdef INET memset(&sin, 0, sizeof(sin)); sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); sin.sin_port = stcb->rport; #endif #ifdef INET6 memset(&sin6, 0, sizeof(sin6)); sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_port = stcb->rport; #endif if (altsa) { sa = altsa; } else { sa = src; } peer_supports_ecn = 0; peer_supports_prsctp = 0; peer_supports_auth = 0; peer_supports_asconf = 0; peer_supports_reconfig = 0; peer_supports_nrsack = 0; peer_supports_pktdrop = 0; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { /* mark all addresses that we have currently on the list */ net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC; } /* does the source address already exist? if so skip it */ inp = stcb->sctp_ep; atomic_add_int(&stcb->asoc.refcnt, 1); stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) { /* we must add the source address */ /* no scope set here since we have a tcb already. */ switch (sa->sa_family) { #ifdef INET case AF_INET: if (stcb->asoc.scope.ipv4_addr_legal) { if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) { return (-1); } } break; #endif #ifdef INET6 case AF_INET6: if (stcb->asoc.scope.ipv6_addr_legal) { if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) { return (-2); } } break; #endif default: break; } } else { if (net_tmp != NULL && stcb_tmp == stcb) { net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; } else if (stcb_tmp != stcb) { /* It belongs to another association? */ if (stcb_tmp) SCTP_TCB_UNLOCK(stcb_tmp); return (-3); } } if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-4); } /* now we must go through each of the params. */ phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); while (phdr) { ptype = ntohs(phdr->param_type); plen = ntohs(phdr->param_length); /* * SCTP_PRINTF("ptype => %0x, plen => %d\n", * (uint32_t)ptype, (int)plen); */ if (offset + plen > limit) { break; } if (plen == 0) { break; } #ifdef INET if (ptype == SCTP_IPV4_ADDRESS) { if (stcb->asoc.scope.ipv4_addr_legal) { struct sctp_ipv4addr_param *p4, p4_buf; /* ok get the v4 address and check/add */ phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&p4_buf, sizeof(p4_buf)); if (plen != sizeof(struct sctp_ipv4addr_param) || phdr == NULL) { return (-5); } p4 = (struct sctp_ipv4addr_param *)phdr; sin.sin_addr.s_addr = p4->addr; if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { /* Skip multi-cast addresses */ goto next_param; } if ((sin.sin_addr.s_addr == INADDR_BROADCAST) || (sin.sin_addr.s_addr == INADDR_ANY)) { goto next_param; } sa = (struct sockaddr *)&sin; inp = stcb->sctp_ep; atomic_add_int(&stcb->asoc.refcnt, 1); stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net, dst, stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) { /* we must add the source address */ /* * no scope set since we have a tcb * already */ /* * we must validate the state again * here */ add_it_now: if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-7); } if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) { return (-8); } } else if (stcb_tmp == stcb) { if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-10); } if (net != NULL) { /* clear flag */ net->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; } } else { /* * strange, address is in another * assoc? straighten out locks. */ if (stcb_tmp) { if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) { struct mbuf *op_err; char msg[SCTP_DIAG_INFO_LEN]; /* * in setup state we * abort this guy */ snprintf(msg, sizeof(msg), "%s:%d at %s", __FILE__, __LINE__, __func__); op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), msg); sctp_abort_an_association(stcb_tmp->sctp_ep, stcb_tmp, op_err, SCTP_SO_NOT_LOCKED); goto add_it_now; } SCTP_TCB_UNLOCK(stcb_tmp); } if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-12); } return (-13); } } } else #endif #ifdef INET6 if (ptype == SCTP_IPV6_ADDRESS) { if (stcb->asoc.scope.ipv6_addr_legal) { /* ok get the v6 address and check/add */ struct sctp_ipv6addr_param *p6, p6_buf; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&p6_buf, sizeof(p6_buf)); if (plen != sizeof(struct sctp_ipv6addr_param) || phdr == NULL) { return (-14); } p6 = (struct sctp_ipv6addr_param *)phdr; memcpy((caddr_t)&sin6.sin6_addr, p6->addr, sizeof(p6->addr)); if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) { /* Skip multi-cast addresses */ goto next_param; } if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) { /* * Link local make no sense without * scope */ goto next_param; } sa = (struct sockaddr *)&sin6; inp = stcb->sctp_ep; atomic_add_int(&stcb->asoc.refcnt, 1); stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net, dst, stcb); atomic_add_int(&stcb->asoc.refcnt, -1); if (stcb_tmp == NULL && (inp == stcb->sctp_ep || inp == NULL)) { /* * we must validate the state again * here */ add_it_now6: if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-16); } /* * we must add the address, no scope * set */ if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) { return (-17); } } else if (stcb_tmp == stcb) { /* * we must validate the state again * here */ if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-19); } if (net != NULL) { /* clear flag */ net->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC; } } else { /* * strange, address is in another * assoc? straighten out locks. */ if (stcb_tmp) { if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) { struct mbuf *op_err; char msg[SCTP_DIAG_INFO_LEN]; /* * in setup state we * abort this guy */ snprintf(msg, sizeof(msg), "%s:%d at %s", __FILE__, __LINE__, __func__); op_err = sctp_generate_cause(SCTP_BASE_SYSCTL(sctp_diag_info_code), msg); sctp_abort_an_association(stcb_tmp->sctp_ep, stcb_tmp, op_err, SCTP_SO_NOT_LOCKED); goto add_it_now6; } SCTP_TCB_UNLOCK(stcb_tmp); } if (stcb->asoc.state == 0) { /* the assoc was freed? */ return (-21); } return (-22); } } } else #endif if (ptype == SCTP_ECN_CAPABLE) { peer_supports_ecn = 1; } else if (ptype == SCTP_ULP_ADAPTATION) { if (stcb->asoc.state != SCTP_STATE_OPEN) { struct sctp_adaptation_layer_indication ai, *aip; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&ai, sizeof(ai)); aip = (struct sctp_adaptation_layer_indication *)phdr; if (aip) { stcb->asoc.peers_adaptation = ntohl(aip->indication); stcb->asoc.adaptation_needed = 1; } } } else if (ptype == SCTP_SET_PRIM_ADDR) { struct sctp_asconf_addr_param lstore, *fee; int lptype; struct sockaddr *lsa = NULL; #ifdef INET struct sctp_asconf_addrv4_param *fii; #endif if (stcb->asoc.asconf_supported == 0) { return (-100); } if (plen > sizeof(lstore)) { return (-23); } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&lstore, min(plen, sizeof(lstore))); if (phdr == NULL) { return (-24); } fee = (struct sctp_asconf_addr_param *)phdr; lptype = ntohs(fee->addrp.ph.param_type); switch (lptype) { #ifdef INET case SCTP_IPV4_ADDRESS: if (plen != sizeof(struct sctp_asconf_addrv4_param)) { SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n", (int)sizeof(struct sctp_asconf_addrv4_param), plen); } else { fii = (struct sctp_asconf_addrv4_param *)fee; sin.sin_addr.s_addr = fii->addrp.addr; lsa = (struct sockaddr *)&sin; } break; #endif #ifdef INET6 case SCTP_IPV6_ADDRESS: if (plen != sizeof(struct sctp_asconf_addr_param)) { SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n", (int)sizeof(struct sctp_asconf_addr_param), plen); } else { memcpy(sin6.sin6_addr.s6_addr, fee->addrp.addr, sizeof(fee->addrp.addr)); lsa = (struct sockaddr *)&sin6; } break; #endif default: break; } if (lsa) { (void)sctp_set_primary_addr(stcb, sa, NULL); } } else if (ptype == SCTP_HAS_NAT_SUPPORT) { stcb->asoc.peer_supports_nat = 1; } else if (ptype == SCTP_PRSCTP_SUPPORTED) { /* Peer supports pr-sctp */ peer_supports_prsctp = 1; } else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) { /* A supported extension chunk */ struct sctp_supported_chunk_types_param *pr_supported; uint8_t local_store[SCTP_PARAM_BUFFER_SIZE]; int num_ent, i; phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen)); if (phdr == NULL) { return (-25); } pr_supported = (struct sctp_supported_chunk_types_param *)phdr; num_ent = plen - sizeof(struct sctp_paramhdr); for (i = 0; i < num_ent; i++) { switch (pr_supported->chunk_types[i]) { case SCTP_ASCONF: peer_supports_asconf = 1; break; case SCTP_ASCONF_ACK: peer_supports_asconf_ack = 1; break; case SCTP_FORWARD_CUM_TSN: peer_supports_prsctp = 1; break; case SCTP_PACKET_DROPPED: peer_supports_pktdrop = 1; break; case SCTP_NR_SELECTIVE_ACK: peer_supports_nrsack = 1; break; case SCTP_STREAM_RESET: peer_supports_reconfig = 1; break; case SCTP_AUTHENTICATION: peer_supports_auth = 1; break; default: /* one I have not learned yet */ break; } } } else if (ptype == SCTP_RANDOM) { if (plen > sizeof(random_store)) break; if (got_random) { /* already processed a RANDOM */ goto next_param; } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)random_store, min(sizeof(random_store), plen)); if (phdr == NULL) return (-26); p_random = (struct sctp_auth_random *)phdr; random_len = plen - sizeof(*p_random); /* enforce the random length */ if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) { SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n"); return (-27); } got_random = 1; } else if (ptype == SCTP_HMAC_LIST) { uint16_t num_hmacs; uint16_t i; if (plen > sizeof(hmacs_store)) break; if (got_hmacs) { /* already processed a HMAC list */ goto next_param; } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)hmacs_store, min(plen, sizeof(hmacs_store))); if (phdr == NULL) return (-28); hmacs = (struct sctp_auth_hmac_algo *)phdr; hmacs_len = plen - sizeof(*hmacs); num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]); /* validate the hmac list */ if (sctp_verify_hmac_param(hmacs, num_hmacs)) { return (-29); } if (stcb->asoc.peer_hmacs != NULL) sctp_free_hmaclist(stcb->asoc.peer_hmacs); stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs); if (stcb->asoc.peer_hmacs != NULL) { for (i = 0; i < num_hmacs; i++) { (void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs, ntohs(hmacs->hmac_ids[i])); } } got_hmacs = 1; } else if (ptype == SCTP_CHUNK_LIST) { int i; if (plen > sizeof(chunks_store)) break; if (got_chklist) { /* already processed a Chunks list */ goto next_param; } phdr = sctp_get_next_param(m, offset, (struct sctp_paramhdr *)chunks_store, min(plen, sizeof(chunks_store))); if (phdr == NULL) return (-30); chunks = (struct sctp_auth_chunk_list *)phdr; num_chunks = plen - sizeof(*chunks); if (stcb->asoc.peer_auth_chunks != NULL) sctp_clear_chunklist(stcb->asoc.peer_auth_chunks); else stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist(); for (i = 0; i < num_chunks; i++) { (void)sctp_auth_add_chunk(chunks->chunk_types[i], stcb->asoc.peer_auth_chunks); /* record asconf/asconf-ack if listed */ if (chunks->chunk_types[i] == SCTP_ASCONF) saw_asconf = 1; if (chunks->chunk_types[i] == SCTP_ASCONF_ACK) saw_asconf_ack = 1; } got_chklist = 1; } else if ((ptype == SCTP_HEARTBEAT_INFO) || (ptype == SCTP_STATE_COOKIE) || (ptype == SCTP_UNRECOG_PARAM) || (ptype == SCTP_COOKIE_PRESERVE) || (ptype == SCTP_SUPPORTED_ADDRTYPE) || (ptype == SCTP_ADD_IP_ADDRESS) || (ptype == SCTP_DEL_IP_ADDRESS) || (ptype == SCTP_ERROR_CAUSE_IND) || (ptype == SCTP_SUCCESS_REPORT)) { /* don't care */ ; } else { if ((ptype & 0x8000) == 0x0000) { /* * must stop processing the rest of the * param's. Any report bits were handled * with the call to * sctp_arethere_unrecognized_parameters() * when the INIT or INIT-ACK was first seen. */ break; } } next_param: offset += SCTP_SIZE32(plen); if (offset >= limit) { break; } phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf)); } /* Now check to see if we need to purge any addresses */ TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) { if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) == SCTP_ADDR_NOT_IN_ASSOC) { /* This address has been removed from the asoc */ /* remove and free it */ stcb->asoc.numnets--; TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next); sctp_free_remote_addr(net); if (net == stcb->asoc.primary_destination) { stcb->asoc.primary_destination = NULL; sctp_select_primary_destination(stcb); } } } if ((stcb->asoc.ecn_supported == 1) && (peer_supports_ecn == 0)) { stcb->asoc.ecn_supported = 0; } if ((stcb->asoc.prsctp_supported == 1) && (peer_supports_prsctp == 0)) { stcb->asoc.prsctp_supported = 0; } if ((stcb->asoc.auth_supported == 1) && ((peer_supports_auth == 0) || (got_random == 0) || (got_hmacs == 0))) { stcb->asoc.auth_supported = 0; } if ((stcb->asoc.asconf_supported == 1) && ((peer_supports_asconf == 0) || (peer_supports_asconf_ack == 0) || (stcb->asoc.auth_supported == 0) || (saw_asconf == 0) || (saw_asconf_ack == 0))) { stcb->asoc.asconf_supported = 0; } if ((stcb->asoc.reconfig_supported == 1) && (peer_supports_reconfig == 0)) { stcb->asoc.reconfig_supported = 0; } if ((stcb->asoc.nrsack_supported == 1) && (peer_supports_nrsack == 0)) { stcb->asoc.nrsack_supported = 0; } if ((stcb->asoc.pktdrop_supported == 1) && (peer_supports_pktdrop == 0)) { stcb->asoc.pktdrop_supported = 0; } /* validate authentication required parameters */ if ((peer_supports_auth == 0) && (got_chklist == 1)) { /* peer does not support auth but sent a chunks list? */ return (-31); } if ((peer_supports_asconf == 1) && (peer_supports_auth == 0)) { /* peer supports asconf but not auth? */ return (-32); } else if ((peer_supports_asconf == 1) && (peer_supports_auth == 1) && ((saw_asconf == 0) || (saw_asconf_ack == 0))) { return (-33); } /* concatenate the full random key */ keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len; if (chunks != NULL) { keylen += sizeof(*chunks) + num_chunks; } new_key = sctp_alloc_key(keylen); if (new_key != NULL) { /* copy in the RANDOM */ if (p_random != NULL) { keylen = sizeof(*p_random) + random_len; bcopy(p_random, new_key->key, keylen); } /* append in the AUTH chunks */ if (chunks != NULL) { bcopy(chunks, new_key->key + keylen, sizeof(*chunks) + num_chunks); keylen += sizeof(*chunks) + num_chunks; } /* append in the HMACs */ if (hmacs != NULL) { bcopy(hmacs, new_key->key + keylen, sizeof(*hmacs) + hmacs_len); } } else { /* failed to get memory for the key */ return (-34); } if (stcb->asoc.authinfo.peer_random != NULL) sctp_free_key(stcb->asoc.authinfo.peer_random); stcb->asoc.authinfo.peer_random = new_key; sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid); sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid); return (0); } int sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa, struct sctp_nets *net) { /* make sure the requested primary address exists in the assoc */ if (net == NULL && sa) net = sctp_findnet(stcb, sa); if (net == NULL) { /* didn't find the requested primary address! */ return (-1); } else { /* set the primary address */ if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { /* Must be confirmed, so queue to set */ net->dest_state |= SCTP_ADDR_REQ_PRIMARY; return (0); } stcb->asoc.primary_destination = net; if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) { sctp_free_remote_addr(stcb->asoc.alternate); stcb->asoc.alternate = NULL; } net = TAILQ_FIRST(&stcb->asoc.nets); if (net != stcb->asoc.primary_destination) { /* * first one on the list is NOT the primary * sctp_cmpaddr() is much more efficient if the * primary is the first on the list, make it so. */ TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next); } return (0); } } int sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now) { /* * This function serves two purposes. It will see if a TAG can be * re-used and return 1 for yes it is ok and 0 for don't use that * tag. A secondary function it will do is purge out old tags that * can be removed. */ struct sctpvtaghead *chain; struct sctp_tagblock *twait_block; struct sctpasochead *head; struct sctp_tcb *stcb; int i; SCTP_INP_INFO_RLOCK(); head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag, SCTP_BASE_INFO(hashasocmark))]; LIST_FOREACH(stcb, head, sctp_asocs) { /* * We choose not to lock anything here. TCB's can't be * removed since we have the read lock, so they can't be * freed on us, same thing for the INP. I may be wrong with * this assumption, but we will go with it for now :-) */ if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) { continue; } if (stcb->asoc.my_vtag == tag) { /* candidate */ if (stcb->rport != rport) { continue; } if (stcb->sctp_ep->sctp_lport != lport) { continue; } /* Its a used tag set */ SCTP_INP_INFO_RUNLOCK(); return (0); } } chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)]; /* Now what about timed wait ? */ LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) { /* * Block(s) are present, lets see if we have this tag in the * list */ for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) { if (twait_block->vtag_block[i].v_tag == 0) { /* not used */ continue; } else if ((long)twait_block->vtag_block[i].tv_sec_at_expire < now->tv_sec) { /* Audit expires this guy */ twait_block->vtag_block[i].tv_sec_at_expire = 0; twait_block->vtag_block[i].v_tag = 0; twait_block->vtag_block[i].lport = 0; twait_block->vtag_block[i].rport = 0; } else if ((twait_block->vtag_block[i].v_tag == tag) && (twait_block->vtag_block[i].lport == lport) && (twait_block->vtag_block[i].rport == rport)) { /* Bad tag, sorry :< */ SCTP_INP_INFO_RUNLOCK(); return (0); } } } SCTP_INP_INFO_RUNLOCK(); return (1); } static void sctp_drain_mbufs(struct sctp_tcb *stcb) { /* * We must hunt this association for MBUF's past the cumack (i.e. * out of order data that we can renege on). */ struct sctp_association *asoc; struct sctp_tmit_chunk *chk, *nchk; uint32_t cumulative_tsn_p1; struct sctp_queued_to_read *ctl, *nctl; int cnt, strmat; uint32_t gap, i; int fnd = 0; /* We look for anything larger than the cum-ack + 1 */ asoc = &stcb->asoc; if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) { /* none we can reneg on. */ return; } SCTP_STAT_INCR(sctps_protocol_drains_done); cumulative_tsn_p1 = asoc->cumulative_tsn + 1; cnt = 0; /* First look in the re-assembly queue */ TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) { if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) { /* Yep it is above cum-ack */ cnt++; SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn); asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size); sctp_ucount_decr(asoc->cnt_on_reasm_queue); SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next); if (chk->data) { sctp_m_freem(chk->data); chk->data = NULL; } sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED); } } /* Ok that was fun, now we will drain all the inbound streams? */ for (strmat = 0; strmat < asoc->streamincnt; strmat++) { TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) { if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) { /* Yep it is above cum-ack */ cnt++; SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn); asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length); sctp_ucount_decr(asoc->cnt_on_all_streams); SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap); TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next); if (ctl->data) { sctp_m_freem(ctl->data); ctl->data = NULL; } sctp_free_remote_addr(ctl->whoFrom); SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl); SCTP_DECR_READQ_COUNT(); } } } if (cnt) { /* We must back down to see what the new highest is */ for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) { SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn); if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) { asoc->highest_tsn_inside_map = i; fnd = 1; break; } } if (!fnd) { asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1; } /* * Question, should we go through the delivery queue? The * only reason things are on here is the app not reading OR * a p-d-api up. An attacker COULD send enough in to * initiate the PD-API and then send a bunch of stuff to * other streams... these would wind up on the delivery * queue.. and then we would not get to them. But in order * to do this I then have to back-track and un-deliver * sequence numbers in streams.. el-yucko. I think for now * we will NOT look at the delivery queue and leave it to be * something to consider later. An alternative would be to * abort the P-D-API with a notification and then deliver * the data.... Or another method might be to keep track of * how many times the situation occurs and if we see a * possible attack underway just abort the association. */ #ifdef SCTP_DEBUG SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt); #endif /* * Now do we need to find a new * asoc->highest_tsn_inside_map? */ asoc->last_revoke_count = cnt; (void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer); /* sa_ignore NO_NULL_CHK */ sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED); sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED); } /* * Another issue, in un-setting the TSN's in the mapping array we * DID NOT adjust the highest_tsn marker. This will cause one of * two things to occur. It may cause us to do extra work in checking * for our mapping array movement. More importantly it may cause us * to SACK every datagram. This may not be a bad thing though since * we will recover once we get our cum-ack above and all this stuff * we dumped recovered. */ } void sctp_drain() { /* * We must walk the PCB lists for ALL associations here. The system * is LOW on MBUF's and needs help. This is where reneging will * occur. We really hope this does NOT happen! */ VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); struct sctp_inpcb *inp; struct sctp_tcb *stcb; SCTP_STAT_INCR(sctps_protocol_drain_calls); if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) { #ifdef VIMAGE continue; #else return; #endif } SCTP_INP_INFO_RLOCK(); LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) { /* For each endpoint */ SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { /* For each association */ SCTP_TCB_LOCK(stcb); sctp_drain_mbufs(stcb); SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } SCTP_INP_INFO_RUNLOCK(); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * start a new iterator * iterates through all endpoints and associations based on the pcb_state * flags and asoc_state. "af" (mandatory) is executed for all matching * assocs and "ef" (optional) is executed when the iterator completes. * "inpf" (optional) is executed for each new endpoint as it is being * iterated through. inpe (optional) is called when the inp completes * its way through all the stcbs. */ int sctp_initiate_iterator(inp_func inpf, asoc_func af, inp_func inpe, uint32_t pcb_state, uint32_t pcb_features, uint32_t asoc_state, void *argp, uint32_t argi, end_func ef, struct sctp_inpcb *s_inp, uint8_t chunk_output_off) { struct sctp_iterator *it = NULL; if (af == NULL) { return (-1); } SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator), SCTP_M_ITER); if (it == NULL) { SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM); return (ENOMEM); } memset(it, 0, sizeof(*it)); it->function_assoc = af; it->function_inp = inpf; if (inpf) it->done_current_ep = 0; else it->done_current_ep = 1; it->function_atend = ef; it->pointer = argp; it->val = argi; it->pcb_flags = pcb_state; it->pcb_features = pcb_features; it->asoc_state = asoc_state; it->function_inp_end = inpe; it->no_chunk_output = chunk_output_off; it->vn = curvnet; if (s_inp) { /* Assume lock is held here */ it->inp = s_inp; SCTP_INP_INCR_REF(it->inp); it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP; } else { SCTP_INP_INFO_RLOCK(); it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead)); if (it->inp) { SCTP_INP_INCR_REF(it->inp); } SCTP_INP_INFO_RUNLOCK(); it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP; } SCTP_IPI_ITERATOR_WQ_LOCK(); TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr); if (sctp_it_ctl.iterator_running == 0) { sctp_wakeup_iterator(); } SCTP_IPI_ITERATOR_WQ_UNLOCK(); /* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */ return (0); } Index: projects/vnet/sys/netinet/sctp_usrreq.c =================================================================== --- projects/vnet/sys/netinet/sctp_usrreq.c (revision 294565) +++ projects/vnet/sys/netinet/sctp_usrreq.c (revision 294566) @@ -1,7419 +1,7419 @@ /*- * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #include #include #include #include extern struct sctp_cc_functions sctp_cc_functions[]; extern struct sctp_ss_functions sctp_ss_functions[]; void sctp_init(void) { u_long sb_max_adj; /* Initialize and modify the sysctled variables */ sctp_init_sysctls(); if ((nmbclusters / 8) > SCTP_ASOC_MAX_CHUNKS_ON_QUEUE) SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue) = (nmbclusters / 8); /* * Allow a user to take no more than 1/2 the number of clusters or * the SB_MAX whichever is smaller for the send window. */ sb_max_adj = (u_long)((u_quad_t) (SB_MAX) * MCLBYTES / (MSIZE + MCLBYTES)); SCTP_BASE_SYSCTL(sctp_sendspace) = min(sb_max_adj, (((uint32_t) nmbclusters / 2) * SCTP_DEFAULT_MAXSEGMENT)); /* * Now for the recv window, should we take the same amount? or * should I do 1/2 the SB_MAX instead in the SB_MAX min above. For * now I will just copy. */ SCTP_BASE_SYSCTL(sctp_recvspace) = SCTP_BASE_SYSCTL(sctp_sendspace); SCTP_BASE_VAR(first_time) = 0; SCTP_BASE_VAR(sctp_pcb_initialized) = 0; sctp_pcb_init(); #if defined(SCTP_PACKET_LOGGING) SCTP_BASE_VAR(packet_log_writers) = 0; SCTP_BASE_VAR(packet_log_end) = 0; bzero(&SCTP_BASE_VAR(packet_log_buffer), SCTP_PACKET_LOG_SIZE); #endif } -void +static void sctp_finish(void) { sctp_pcb_finish(); } - +VNET_SYSUNINIT(sctp, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, sctp_finish, NULL); void sctp_pathmtu_adjustment(struct sctp_tcb *stcb, uint16_t nxtsz) { struct sctp_tmit_chunk *chk; uint16_t overhead; /* Adjust that too */ stcb->asoc.smallest_mtu = nxtsz; /* now off to subtract IP_DF flag if needed */ overhead = IP_HDR_SIZE; if (sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks)) { overhead += sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id); } TAILQ_FOREACH(chk, &stcb->asoc.send_queue, sctp_next) { if ((chk->send_size + overhead) > nxtsz) { chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; } } TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) { if ((chk->send_size + overhead) > nxtsz) { /* * For this guy we also mark for immediate resend * since we sent to big of chunk */ chk->flags |= CHUNK_FLAGS_FRAGMENT_OK; if (chk->sent < SCTP_DATAGRAM_RESEND) { sctp_flight_size_decrease(chk); sctp_total_flight_decrease(stcb, chk); chk->sent = SCTP_DATAGRAM_RESEND; sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt); chk->rec.data.doing_fast_retransmit = 0; if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) { sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_PMTU, chk->whoTo->flight_size, chk->book_size, (uintptr_t) chk->whoTo, chk->rec.data.TSN_seq); } /* Clear any time so NO RTT is being done */ chk->do_rtt = 0; } } } } #ifdef INET static void sctp_notify_mbuf(struct sctp_inpcb *inp, struct sctp_tcb *stcb, struct sctp_nets *net, struct ip *ip, struct sctphdr *sh) { struct icmp *icmph; int totsz, tmr_stopped = 0; uint16_t nxtsz; /* protection */ if ((inp == NULL) || (stcb == NULL) || (net == NULL) || (ip == NULL) || (sh == NULL)) { if (stcb != NULL) { SCTP_TCB_UNLOCK(stcb); } return; } /* First job is to verify the vtag matches what I would send */ if (ntohl(sh->v_tag) != (stcb->asoc.peer_vtag)) { SCTP_TCB_UNLOCK(stcb); return; } icmph = (struct icmp *)((caddr_t)ip - (sizeof(struct icmp) - sizeof(struct ip))); if (icmph->icmp_type != ICMP_UNREACH) { /* We only care about unreachable */ SCTP_TCB_UNLOCK(stcb); return; } if (icmph->icmp_code != ICMP_UNREACH_NEEDFRAG) { /* not a unreachable message due to frag. */ SCTP_TCB_UNLOCK(stcb); return; } totsz = ntohs(ip->ip_len); nxtsz = ntohs(icmph->icmp_nextmtu); if (nxtsz == 0) { /* * old type router that does not tell us what the next size * mtu is. Rats we will have to guess (in a educated fashion * of course) */ nxtsz = sctp_get_prev_mtu(totsz); } /* Stop any PMTU timer */ if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { tmr_stopped = 1; sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_1); } /* Adjust destination size limit */ if (net->mtu > nxtsz) { net->mtu = nxtsz; if (net->port) { net->mtu -= sizeof(struct udphdr); } } /* now what about the ep? */ if (stcb->asoc.smallest_mtu > nxtsz) { sctp_pathmtu_adjustment(stcb, nxtsz); } if (tmr_stopped) sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); SCTP_TCB_UNLOCK(stcb); } void sctp_notify(struct sctp_inpcb *inp, struct ip *ip, struct sctphdr *sh, struct sockaddr *to, struct sctp_tcb *stcb, struct sctp_nets *net) { #if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING) struct socket *so; #endif struct icmp *icmph; /* protection */ if ((inp == NULL) || (stcb == NULL) || (net == NULL) || (sh == NULL) || (to == NULL)) { if (stcb) SCTP_TCB_UNLOCK(stcb); return; } /* First job is to verify the vtag matches what I would send */ if (ntohl(sh->v_tag) != (stcb->asoc.peer_vtag)) { SCTP_TCB_UNLOCK(stcb); return; } icmph = (struct icmp *)((caddr_t)ip - (sizeof(struct icmp) - sizeof(struct ip))); if (icmph->icmp_type != ICMP_UNREACH) { /* We only care about unreachable */ SCTP_TCB_UNLOCK(stcb); return; } if ((icmph->icmp_code == ICMP_UNREACH_NET) || (icmph->icmp_code == ICMP_UNREACH_HOST) || (icmph->icmp_code == ICMP_UNREACH_NET_UNKNOWN) || (icmph->icmp_code == ICMP_UNREACH_HOST_UNKNOWN) || (icmph->icmp_code == ICMP_UNREACH_ISOLATED) || (icmph->icmp_code == ICMP_UNREACH_NET_PROHIB) || (icmph->icmp_code == ICMP_UNREACH_HOST_PROHIB) || (icmph->icmp_code == ICMP_UNREACH_FILTER_PROHIB)) { /* * Hmm reachablity problems we must examine closely. If its * not reachable, we may have lost a network. Or if there is * NO protocol at the other end named SCTP. well we consider * it a OOTB abort. */ if (net->dest_state & SCTP_ADDR_REACHABLE) { /* Ok that destination is NOT reachable */ net->dest_state &= ~SCTP_ADDR_REACHABLE; net->dest_state &= ~SCTP_ADDR_PF; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, 0, (void *)net, SCTP_SO_NOT_LOCKED); } SCTP_TCB_UNLOCK(stcb); } else if ((icmph->icmp_code == ICMP_UNREACH_PROTOCOL) || (icmph->icmp_code == ICMP_UNREACH_PORT)) { /* * Here the peer is either playing tricks on us, including * an address that belongs to someone who does not support * SCTP OR was a userland implementation that shutdown and * now is dead. In either case treat it like a OOTB abort * with no TCB */ sctp_abort_notification(stcb, 1, 0, NULL, SCTP_SO_NOT_LOCKED); #if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING) so = SCTP_INP_SO(inp); atomic_add_int(&stcb->asoc.refcnt, 1); SCTP_TCB_UNLOCK(stcb); SCTP_SOCKET_LOCK(so, 1); SCTP_TCB_LOCK(stcb); atomic_subtract_int(&stcb->asoc.refcnt, 1); #endif (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_2); #if defined(__APPLE__) || defined(SCTP_SO_LOCK_TESTING) SCTP_SOCKET_UNLOCK(so, 1); /* SCTP_TCB_UNLOCK(stcb); MT: I think this is not needed. */ #endif /* no need to unlock here, since the TCB is gone */ } else { SCTP_TCB_UNLOCK(stcb); } } #endif #ifdef INET void sctp_ctlinput(cmd, sa, vip) int cmd; struct sockaddr *sa; void *vip; { struct ip *ip = vip; struct sctphdr *sh; uint32_t vrf_id; /* FIX, for non-bsd is this right? */ vrf_id = SCTP_DEFAULT_VRFID; if (sa->sa_family != AF_INET || ((struct sockaddr_in *)sa)->sin_addr.s_addr == INADDR_ANY) { return; } if (PRC_IS_REDIRECT(cmd)) { ip = 0; } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) { return; } if (ip) { struct sctp_inpcb *inp = NULL; struct sctp_tcb *stcb = NULL; struct sctp_nets *net = NULL; struct sockaddr_in to, from; sh = (struct sctphdr *)((caddr_t)ip + (ip->ip_hl << 2)); bzero(&to, sizeof(to)); bzero(&from, sizeof(from)); from.sin_family = to.sin_family = AF_INET; from.sin_len = to.sin_len = sizeof(to); from.sin_port = sh->src_port; from.sin_addr = ip->ip_src; to.sin_port = sh->dest_port; to.sin_addr = ip->ip_dst; /* * 'to' holds the dest of the packet that failed to be sent. * 'from' holds our local endpoint address. Thus we reverse * the to and the from in the lookup. */ stcb = sctp_findassociation_addr_sa((struct sockaddr *)&to, (struct sockaddr *)&from, &inp, &net, 1, vrf_id); if (stcb != NULL && inp && (inp->sctp_socket != NULL)) { if (cmd != PRC_MSGSIZE) { sctp_notify(inp, ip, sh, (struct sockaddr *)&to, stcb, net); } else { /* handle possible ICMP size messages */ sctp_notify_mbuf(inp, stcb, net, ip, sh); } } else { if ((stcb == NULL) && (inp != NULL)) { /* reduce ref-count */ SCTP_INP_WLOCK(inp); SCTP_INP_DECR_REF(inp); SCTP_INP_WUNLOCK(inp); } if (stcb) { SCTP_TCB_UNLOCK(stcb); } } } return; } #endif static int sctp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct sctp_inpcb *inp; struct sctp_nets *net; struct sctp_tcb *stcb; int error; uint32_t vrf_id; /* FIX, for non-bsd is this right? */ vrf_id = SCTP_DEFAULT_VRFID; error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); stcb = sctp_findassociation_addr_sa(sintosa(&addrs[1]), sintosa(&addrs[0]), &inp, &net, 1, vrf_id); if (stcb == NULL || inp == NULL || inp->sctp_socket == NULL) { if ((inp != NULL) && (stcb == NULL)) { /* reduce ref-count */ SCTP_INP_WLOCK(inp); SCTP_INP_DECR_REF(inp); goto cred_can_cont; } SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; goto out; } SCTP_TCB_UNLOCK(stcb); /* * We use the write lock here, only since in the error leg we need * it. If we used RLOCK, then we would have to * wlock/decr/unlock/rlock. Which in theory could create a hole. * Better to use higher wlock. */ SCTP_INP_WLOCK(inp); cred_can_cont: error = cr_canseesocket(req->td->td_ucred, inp->sctp_socket); if (error) { SCTP_INP_WUNLOCK(inp); goto out; } cru2x(inp->sctp_socket->so_cred, &xuc); SCTP_INP_WUNLOCK(inp); error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); out: return (error); } SYSCTL_PROC(_net_inet_sctp, OID_AUTO, getcred, CTLTYPE_OPAQUE | CTLFLAG_RW, 0, 0, sctp_getcred, "S,ucred", "Get the ucred of a SCTP connection"); #ifdef INET static void sctp_abort(struct socket *so) { struct sctp_inpcb *inp; uint32_t flags; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { return; } sctp_must_try_again: flags = inp->sctp_flags; #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 17); #endif if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && (atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 16); #endif sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_AFTER_CMPSET_OFCLOSE); SOCK_LOCK(so); SCTP_SB_CLEAR(so->so_snd); /* * same for the rcv ones, they are only here for the * accounting/select. */ SCTP_SB_CLEAR(so->so_rcv); /* Now null out the reference, we are completely detached. */ so->so_pcb = NULL; SOCK_UNLOCK(so); } else { flags = inp->sctp_flags; if ((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) { goto sctp_must_try_again; } } return; } static int sctp_attach(struct socket *so, int proto SCTP_UNUSED, struct thread *p SCTP_UNUSED) { struct sctp_inpcb *inp; struct inpcb *ip_inp; int error; uint32_t vrf_id = SCTP_DEFAULT_VRFID; inp = (struct sctp_inpcb *)so->so_pcb; if (inp != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { error = SCTP_SORESERVE(so, SCTP_BASE_SYSCTL(sctp_sendspace), SCTP_BASE_SYSCTL(sctp_recvspace)); if (error) { return (error); } } error = sctp_inpcb_alloc(so, vrf_id); if (error) { return (error); } inp = (struct sctp_inpcb *)so->so_pcb; SCTP_INP_WLOCK(inp); inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUND_V6; /* I'm not v6! */ ip_inp = &inp->ip_inp.inp; ip_inp->inp_vflag |= INP_IPV4; ip_inp->inp_ip_ttl = MODULE_GLOBAL(ip_defttl); SCTP_INP_WUNLOCK(inp); return (0); } static int sctp_bind(struct socket *so, struct sockaddr *addr, struct thread *p) { struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } if (addr != NULL) { if ((addr->sa_family != AF_INET) || (addr->sa_len != sizeof(struct sockaddr_in))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } } return (sctp_inpcb_bind(so, addr, NULL, p)); } #endif void sctp_close(struct socket *so) { struct sctp_inpcb *inp; uint32_t flags; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) return; /* * Inform all the lower layer assoc that we are done. */ sctp_must_try_again: flags = inp->sctp_flags; #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 17); #endif if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && (atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) { if (((so->so_options & SO_LINGER) && (so->so_linger == 0)) || (so->so_rcv.sb_cc > 0)) { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 13); #endif sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT, SCTP_CALLED_AFTER_CMPSET_OFCLOSE); } else { #ifdef SCTP_LOG_CLOSING sctp_log_closing(inp, NULL, 14); #endif sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE, SCTP_CALLED_AFTER_CMPSET_OFCLOSE); } /* * The socket is now detached, no matter what the state of * the SCTP association. */ SOCK_LOCK(so); SCTP_SB_CLEAR(so->so_snd); /* * same for the rcv ones, they are only here for the * accounting/select. */ SCTP_SB_CLEAR(so->so_rcv); /* Now null out the reference, we are completely detached. */ so->so_pcb = NULL; SOCK_UNLOCK(so); } else { flags = inp->sctp_flags; if ((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) { goto sctp_must_try_again; } } return; } int sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *p); int sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *p) { struct sctp_inpcb *inp; int error; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { if (control) { sctp_m_freem(control); control = NULL; } SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); sctp_m_freem(m); return (EINVAL); } /* Got to have an to address if we are NOT a connected socket */ if ((addr == NULL) && ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) || (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE))) { goto connected_type; } else if (addr == NULL) { SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EDESTADDRREQ); error = EDESTADDRREQ; sctp_m_freem(m); if (control) { sctp_m_freem(control); control = NULL; } return (error); } #ifdef INET6 if (addr->sa_family != AF_INET) { /* must be a v4 address! */ SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EDESTADDRREQ); sctp_m_freem(m); if (control) { sctp_m_freem(control); control = NULL; } error = EDESTADDRREQ; return (error); } #endif /* INET6 */ connected_type: /* now what about control */ if (control) { if (inp->control) { SCTP_PRINTF("huh? control set?\n"); sctp_m_freem(inp->control); inp->control = NULL; } inp->control = control; } /* Place the data */ if (inp->pkt) { SCTP_BUF_NEXT(inp->pkt_last) = m; inp->pkt_last = m; } else { inp->pkt_last = inp->pkt = m; } if ( /* FreeBSD uses a flag passed */ ((flags & PRUS_MORETOCOME) == 0) ) { /* * note with the current version this code will only be used * by OpenBSD-- NetBSD, FreeBSD, and MacOS have methods for * re-defining sosend to use the sctp_sosend. One can * optionally switch back to this code (by changing back the * definitions) but this is not advisable. This code is used * by FreeBSD when sending a file with sendfile() though. */ int ret; ret = sctp_output(inp, inp->pkt, addr, inp->control, p, flags); inp->pkt = NULL; inp->control = NULL; return (ret); } else { return (0); } } int sctp_disconnect(struct socket *so) { struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); return (ENOTCONN); } SCTP_INP_RLOCK(inp); if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { if (LIST_EMPTY(&inp->sctp_asoc_list)) { /* No connection */ SCTP_INP_RUNLOCK(inp); return (0); } else { struct sctp_association *asoc; struct sctp_tcb *stcb; stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } SCTP_TCB_LOCK(stcb); asoc = &stcb->asoc; if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { /* We are about to be freed, out of here */ SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); return (0); } if (((so->so_options & SO_LINGER) && (so->so_linger == 0)) || (so->so_rcv.sb_cc > 0)) { if (SCTP_GET_STATE(asoc) != SCTP_STATE_COOKIE_WAIT) { /* Left with Data unread */ struct mbuf *err; err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_NOWAIT, 1, MT_DATA); if (err) { /* * Fill in the user * initiated abort */ struct sctp_paramhdr *ph; ph = mtod(err, struct sctp_paramhdr *); SCTP_BUF_LEN(err) = sizeof(struct sctp_paramhdr); ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT); ph->param_length = htons(SCTP_BUF_LEN(err)); } sctp_send_abort_tcb(stcb, err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); } SCTP_INP_RUNLOCK(inp); if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_3); /* No unlock tcb assoc is gone */ return (0); } if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->stream_queue_cnt == 0)) { /* there is nothing queued to send, so done */ if (asoc->locked_on_sending) { goto abort_anyway; } if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) { /* only send SHUTDOWN 1st time thru */ struct sctp_nets *netp; if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_stop_timers_for_shutdown(stcb); if (stcb->asoc.alternate) { netp = stcb->asoc.alternate; } else { netp = stcb->asoc.primary_destination; } sctp_send_shutdown(stcb, netp); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, netp); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, netp); sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_LOCKED); } } else { /* * we still got (or just got) data to send, * so set SHUTDOWN_PENDING */ /* * XXX sockets draft says that SCTP_EOF * should be sent with no data. currently, * we will allow user data to be sent first * and move to SHUTDOWN-PENDING */ struct sctp_nets *netp; if (stcb->asoc.alternate) { netp = stcb->asoc.alternate; } else { netp = stcb->asoc.primary_destination; } asoc->state |= SCTP_STATE_SHUTDOWN_PENDING; sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, netp); if (asoc->locked_on_sending) { /* Locked to send out the data */ struct sctp_stream_queue_pending *sp; sp = TAILQ_LAST(&asoc->locked_on_sending->outqueue, sctp_streamhead); if (sp == NULL) { SCTP_PRINTF("Error, sp is NULL, locked on sending is non-null strm:%d\n", asoc->locked_on_sending->stream_no); } else { if ((sp->length == 0) && (sp->msg_is_complete == 0)) asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT; } } if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { struct mbuf *op_err; abort_anyway: op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_USRREQ + SCTP_LOC_4; sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED); SCTP_STAT_INCR_COUNTER32(sctps_aborted); if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) || (SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) { SCTP_STAT_DECR_GAUGE32(sctps_currestab); } SCTP_INP_RUNLOCK(inp); (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_5); return (0); } else { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); } } soisdisconnecting(so); SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); return (0); } /* not reached */ } else { /* UDP model does not support this */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); return (EOPNOTSUPP); } } int sctp_flush(struct socket *so, int how) { /* * We will just clear out the values and let subsequent close clear * out the data, if any. Note if the user did a shutdown(SHUT_RD) * they will not be able to read the data, the socket will block * that from happening. */ struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } SCTP_INP_RLOCK(inp); /* For the 1 to many model this does nothing */ if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { SCTP_INP_RUNLOCK(inp); return (0); } SCTP_INP_RUNLOCK(inp); if ((how == PRU_FLUSH_RD) || (how == PRU_FLUSH_RDWR)) { /* * First make sure the sb will be happy, we don't use these * except maybe the count */ SCTP_INP_WLOCK(inp); SCTP_INP_READ_LOCK(inp); inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_CANT_READ; SCTP_INP_READ_UNLOCK(inp); SCTP_INP_WUNLOCK(inp); so->so_rcv.sb_cc = 0; so->so_rcv.sb_mbcnt = 0; so->so_rcv.sb_mb = NULL; } if ((how == PRU_FLUSH_WR) || (how == PRU_FLUSH_RDWR)) { /* * First make sure the sb will be happy, we don't use these * except maybe the count */ so->so_snd.sb_cc = 0; so->so_snd.sb_mbcnt = 0; so->so_snd.sb_mb = NULL; } return (0); } int sctp_shutdown(struct socket *so) { struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } SCTP_INP_RLOCK(inp); /* For UDP model this is a invalid call */ if (!((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL))) { /* Restore the flags that the soshutdown took away. */ SOCKBUF_LOCK(&so->so_rcv); so->so_rcv.sb_state &= ~SBS_CANTRCVMORE; SOCKBUF_UNLOCK(&so->so_rcv); /* This proc will wakeup for read and do nothing (I hope) */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); return (EOPNOTSUPP); } else { /* * Ok, if we reach here its the TCP model and it is either a * SHUT_WR or SHUT_RDWR. This means we put the shutdown flag * against it. */ struct sctp_tcb *stcb; struct sctp_association *asoc; struct sctp_nets *netp; if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | SS_ISDISCONNECTING)) == 0) { SCTP_INP_RUNLOCK(inp); return (ENOTCONN); } socantsendmore(so); stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { /* * Ok, we hit the case that the shutdown call was * made after an abort or something. Nothing to do * now. */ SCTP_INP_RUNLOCK(inp); return (0); } SCTP_TCB_LOCK(stcb); asoc = &stcb->asoc; if (asoc->state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); return (0); } if ((SCTP_GET_STATE(asoc) != SCTP_STATE_COOKIE_WAIT) && (SCTP_GET_STATE(asoc) != SCTP_STATE_COOKIE_ECHOED) && (SCTP_GET_STATE(asoc) != SCTP_STATE_OPEN)) { /* * If we are not in or before ESTABLISHED, there is * no protocol action required. */ SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); return (0); } if (stcb->asoc.alternate) { netp = stcb->asoc.alternate; } else { netp = stcb->asoc.primary_destination; } if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) && TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->stream_queue_cnt == 0)) { if (asoc->locked_on_sending) { goto abort_anyway; } /* there is nothing queued to send, so I'm done... */ SCTP_STAT_DECR_GAUGE32(sctps_currestab); SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT); SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); sctp_stop_timers_for_shutdown(stcb); sctp_send_shutdown(stcb, netp); sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, stcb->sctp_ep, stcb, netp); } else { /* * We still got (or just got) data to send, so set * SHUTDOWN_PENDING. */ SCTP_ADD_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING); if (asoc->locked_on_sending) { /* Locked to send out the data */ struct sctp_stream_queue_pending *sp; sp = TAILQ_LAST(&asoc->locked_on_sending->outqueue, sctp_streamhead); if (sp == NULL) { SCTP_PRINTF("Error, sp is NULL, locked on sending is non-null strm:%d\n", asoc->locked_on_sending->stream_no); } else { if ((sp->length == 0) && (sp->msg_is_complete == 0)) { SCTP_ADD_SUBSTATE(asoc, SCTP_STATE_PARTIAL_MSG_LEFT); } } } if (TAILQ_EMPTY(&asoc->send_queue) && TAILQ_EMPTY(&asoc->sent_queue) && (asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) { struct mbuf *op_err; abort_anyway: op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, ""); stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_USRREQ + SCTP_LOC_6; sctp_abort_an_association(stcb->sctp_ep, stcb, op_err, SCTP_SO_LOCKED); SCTP_INP_RUNLOCK(inp); return (0); } } sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb, netp); /* * XXX: Why do this in the case where we have still data * queued? */ sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED); SCTP_TCB_UNLOCK(stcb); SCTP_INP_RUNLOCK(inp); return (0); } } /* * copies a "user" presentable address and removes embedded scope, etc. * returns 0 on success, 1 on error */ static uint32_t sctp_fill_user_address(struct sockaddr_storage *ss, struct sockaddr *sa) { #ifdef INET6 struct sockaddr_in6 lsa6; sa = (struct sockaddr *)sctp_recover_scope((struct sockaddr_in6 *)sa, &lsa6); #endif memcpy(ss, sa, sa->sa_len); return (0); } /* * NOTE: assumes addr lock is held */ static size_t sctp_fill_up_addresses_vrf(struct sctp_inpcb *inp, struct sctp_tcb *stcb, size_t limit, struct sockaddr_storage *sas, uint32_t vrf_id) { struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa; size_t actual; int loopback_scope; #if defined(INET) int ipv4_local_scope, ipv4_addr_legal; #endif #if defined(INET6) int local_scope, site_scope, ipv6_addr_legal; #endif struct sctp_vrf *vrf; actual = 0; if (limit <= 0) return (actual); if (stcb) { /* Turn on all the appropriate scope */ loopback_scope = stcb->asoc.scope.loopback_scope; #if defined(INET) ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope; ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal; #endif #if defined(INET6) local_scope = stcb->asoc.scope.local_scope; site_scope = stcb->asoc.scope.site_scope; ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal; #endif } else { /* Use generic values for endpoints. */ loopback_scope = 1; #if defined(INET) ipv4_local_scope = 1; #endif #if defined(INET6) local_scope = 1; site_scope = 1; #endif if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { #if defined(INET6) ipv6_addr_legal = 1; #endif #if defined(INET) if (SCTP_IPV6_V6ONLY(inp)) { ipv4_addr_legal = 0; } else { ipv4_addr_legal = 1; } #endif } else { #if defined(INET6) ipv6_addr_legal = 0; #endif #if defined(INET) ipv4_addr_legal = 1; #endif } } vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { return (0); } if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { if ((loopback_scope == 0) && SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) { /* Skip loopback if loopback_scope not set */ continue; } LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { if (stcb) { /* * For the BOUND-ALL case, the list * associated with a TCB is Always * considered a reverse list.. i.e. * it lists addresses that are NOT * part of the association. If this * is one of those we must skip it. */ if (sctp_is_addr_restricted(stcb, sctp_ifa)) { continue; } } switch (sctp_ifa->address.sa.sa_family) { #ifdef INET case AF_INET: if (ipv4_addr_legal) { struct sockaddr_in *sin; sin = &sctp_ifa->address.sin; if (sin->sin_addr.s_addr == 0) { /* * we skip * unspecifed * addresses */ continue; } if (prison_check_ip4(inp->ip_inp.inp.inp_cred, &sin->sin_addr) != 0) { continue; } if ((ipv4_local_scope == 0) && (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) { continue; } #ifdef INET6 if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { in6_sin_2_v4mapsin6(sin, (struct sockaddr_in6 *)sas); ((struct sockaddr_in6 *)sas)->sin6_port = inp->sctp_lport; sas = (struct sockaddr_storage *)((caddr_t)sas + sizeof(struct sockaddr_in6)); actual += sizeof(struct sockaddr_in6); } else { #endif memcpy(sas, sin, sizeof(*sin)); ((struct sockaddr_in *)sas)->sin_port = inp->sctp_lport; sas = (struct sockaddr_storage *)((caddr_t)sas + sizeof(*sin)); actual += sizeof(*sin); #ifdef INET6 } #endif if (actual >= limit) { return (actual); } } else { continue; } break; #endif #ifdef INET6 case AF_INET6: if (ipv6_addr_legal) { struct sockaddr_in6 *sin6; sin6 = &sctp_ifa->address.sin6; if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { /* * we skip * unspecifed * addresses */ continue; } if (prison_check_ip6(inp->ip_inp.inp.inp_cred, &sin6->sin6_addr) != 0) { continue; } if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { if (local_scope == 0) continue; if (sin6->sin6_scope_id == 0) { if (sa6_recoverscope(sin6) != 0) /* * * bad * * li * nk * * loc * al * * add * re * ss * */ continue; } } if ((site_scope == 0) && (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) { continue; } memcpy(sas, sin6, sizeof(*sin6)); ((struct sockaddr_in6 *)sas)->sin6_port = inp->sctp_lport; sas = (struct sockaddr_storage *)((caddr_t)sas + sizeof(*sin6)); actual += sizeof(*sin6); if (actual >= limit) { return (actual); } } else { continue; } break; #endif default: /* TSNH */ break; } } } } else { struct sctp_laddr *laddr; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (stcb) { if (sctp_is_addr_restricted(stcb, laddr->ifa)) { continue; } } if (sctp_fill_user_address(sas, &laddr->ifa->address.sa)) continue; switch (laddr->ifa->address.sa.sa_family) { #ifdef INET case AF_INET: ((struct sockaddr_in *)sas)->sin_port = inp->sctp_lport; break; #endif #ifdef INET6 case AF_INET6: ((struct sockaddr_in6 *)sas)->sin6_port = inp->sctp_lport; break; #endif default: /* TSNH */ break; } sas = (struct sockaddr_storage *)((caddr_t)sas + laddr->ifa->address.sa.sa_len); actual += laddr->ifa->address.sa.sa_len; if (actual >= limit) { return (actual); } } } return (actual); } static size_t sctp_fill_up_addresses(struct sctp_inpcb *inp, struct sctp_tcb *stcb, size_t limit, struct sockaddr_storage *sas) { size_t size = 0; SCTP_IPI_ADDR_RLOCK(); /* fill up addresses for the endpoint's default vrf */ size = sctp_fill_up_addresses_vrf(inp, stcb, limit, sas, inp->def_vrf_id); SCTP_IPI_ADDR_RUNLOCK(); return (size); } /* * NOTE: assumes addr lock is held */ static int sctp_count_max_addresses_vrf(struct sctp_inpcb *inp, uint32_t vrf_id) { int cnt = 0; struct sctp_vrf *vrf = NULL; /* * In both sub-set bound an bound_all cases we return the MAXIMUM * number of addresses that you COULD get. In reality the sub-set * bound may have an exclusion list for a given TCB OR in the * bound-all case a TCB may NOT include the loopback or other * addresses as well. */ vrf = sctp_find_vrf(vrf_id); if (vrf == NULL) { return (0); } if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { struct sctp_ifn *sctp_ifn; struct sctp_ifa *sctp_ifa; LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) { LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) { /* Count them if they are the right type */ switch (sctp_ifa->address.sa.sa_family) { #ifdef INET case AF_INET: #ifdef INET6 if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) cnt += sizeof(struct sockaddr_in6); else cnt += sizeof(struct sockaddr_in); #else cnt += sizeof(struct sockaddr_in); #endif break; #endif #ifdef INET6 case AF_INET6: cnt += sizeof(struct sockaddr_in6); break; #endif default: break; } } } } else { struct sctp_laddr *laddr; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { switch (laddr->ifa->address.sa.sa_family) { #ifdef INET case AF_INET: #ifdef INET6 if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) cnt += sizeof(struct sockaddr_in6); else cnt += sizeof(struct sockaddr_in); #else cnt += sizeof(struct sockaddr_in); #endif break; #endif #ifdef INET6 case AF_INET6: cnt += sizeof(struct sockaddr_in6); break; #endif default: break; } } } return (cnt); } static int sctp_count_max_addresses(struct sctp_inpcb *inp) { int cnt = 0; SCTP_IPI_ADDR_RLOCK(); /* count addresses for the endpoint's default VRF */ cnt = sctp_count_max_addresses_vrf(inp, inp->def_vrf_id); SCTP_IPI_ADDR_RUNLOCK(); return (cnt); } static int sctp_do_connect_x(struct socket *so, struct sctp_inpcb *inp, void *optval, size_t optsize, void *p, int delay) { int error = 0; int creat_lock_on = 0; struct sctp_tcb *stcb = NULL; struct sockaddr *sa; int num_v6 = 0, num_v4 = 0, *totaddrp, totaddr; uint32_t vrf_id; int bad_addresses = 0; sctp_assoc_t *a_id; SCTPDBG(SCTP_DEBUG_PCB1, "Connectx called\n"); if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { /* We are already connected AND the TCP model */ SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE); return (EADDRINUSE); } if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE))) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { SCTP_INP_RLOCK(inp); stcb = LIST_FIRST(&inp->sctp_asoc_list); SCTP_INP_RUNLOCK(inp); } if (stcb) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); return (EALREADY); } SCTP_INP_INCR_REF(inp); SCTP_ASOC_CREATE_LOCK(inp); creat_lock_on = 1; if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EFAULT); error = EFAULT; goto out_now; } totaddrp = (int *)optval; totaddr = *totaddrp; sa = (struct sockaddr *)(totaddrp + 1); stcb = sctp_connectx_helper_find(inp, sa, &totaddr, &num_v4, &num_v6, &error, (optsize - sizeof(int)), &bad_addresses); if ((stcb != NULL) || bad_addresses) { /* Already have or am bring up an association */ SCTP_ASOC_CREATE_UNLOCK(inp); creat_lock_on = 0; if (stcb) SCTP_TCB_UNLOCK(stcb); if (bad_addresses == 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); error = EALREADY; } goto out_now; } #ifdef INET6 if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && (num_v6 > 0)) { error = EINVAL; goto out_now; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && (num_v4 > 0)) { struct in6pcb *inp6; inp6 = (struct in6pcb *)inp; if (SCTP_IPV6_V6ONLY(inp6)) { /* * if IPV6_V6ONLY flag, ignore connections destined * to a v4 addr or v4-mapped addr */ SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_now; } } #endif /* INET6 */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == SCTP_PCB_FLAGS_UNBOUND) { /* Bind a ephemeral port */ error = sctp_inpcb_bind(so, NULL, NULL, p); if (error) { goto out_now; } } /* FIX ME: do we want to pass in a vrf on the connect call? */ vrf_id = inp->def_vrf_id; /* We are GOOD to go */ stcb = sctp_aloc_assoc(inp, sa, &error, 0, vrf_id, inp->sctp_ep.pre_open_stream_count, (struct thread *)p ); if (stcb == NULL) { /* Gak! no memory */ goto out_now; } if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) { stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED; /* Set the connected flag so we can queue data */ soisconnecting(so); } SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_COOKIE_WAIT); /* move to second address */ switch (sa->sa_family) { #ifdef INET case AF_INET: sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in)); break; #endif #ifdef INET6 case AF_INET6: sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in6)); break; #endif default: break; } error = 0; sctp_connectx_helper_add(stcb, sa, (totaddr - 1), &error); /* Fill in the return id */ if (error) { (void)sctp_free_assoc(inp, stcb, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_7); goto out_now; } a_id = (sctp_assoc_t *) optval; *a_id = sctp_get_associd(stcb); /* initialize authentication parameters for the assoc */ sctp_initialize_auth_params(inp, stcb); if (delay) { /* doing delayed connection */ stcb->asoc.delayed_connection = 1; sctp_timer_start(SCTP_TIMER_TYPE_INIT, inp, stcb, stcb->asoc.primary_destination); } else { (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); } SCTP_TCB_UNLOCK(stcb); if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) { stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED; /* Set the connected flag so we can queue data */ soisconnecting(so); } out_now: if (creat_lock_on) { SCTP_ASOC_CREATE_UNLOCK(inp); } SCTP_INP_DECR_REF(inp); return (error); } #define SCTP_FIND_STCB(inp, stcb, assoc_id) { \ if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||\ (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { \ SCTP_INP_RLOCK(inp); \ stcb = LIST_FIRST(&inp->sctp_asoc_list); \ if (stcb) { \ SCTP_TCB_LOCK(stcb); \ } \ SCTP_INP_RUNLOCK(inp); \ } else if (assoc_id > SCTP_ALL_ASSOC) { \ stcb = sctp_findassociation_ep_asocid(inp, assoc_id, 1); \ if (stcb == NULL) { \ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); \ error = ENOENT; \ break; \ } \ } else { \ stcb = NULL; \ } \ } #define SCTP_CHECK_AND_CAST(destp, srcp, type, size) {\ if (size < sizeof(type)) { \ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); \ error = EINVAL; \ break; \ } else { \ destp = (type *)srcp; \ } \ } static int sctp_getopt(struct socket *so, int optname, void *optval, size_t *optsize, void *p) { struct sctp_inpcb *inp = NULL; int error, val = 0; struct sctp_tcb *stcb = NULL; if (optval == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return EINVAL; } error = 0; switch (optname) { case SCTP_NODELAY: case SCTP_AUTOCLOSE: case SCTP_EXPLICIT_EOR: case SCTP_AUTO_ASCONF: case SCTP_DISABLE_FRAGMENTS: case SCTP_I_WANT_MAPPED_V4_ADDR: case SCTP_USE_EXT_RCVINFO: SCTP_INP_RLOCK(inp); switch (optname) { case SCTP_DISABLE_FRAGMENTS: val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NO_FRAGMENT); break; case SCTP_I_WANT_MAPPED_V4_ADDR: val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4); break; case SCTP_AUTO_ASCONF: if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* only valid for bound all sockets */ val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto flags_out; } break; case SCTP_EXPLICIT_EOR: val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR); break; case SCTP_NODELAY: val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NODELAY); break; case SCTP_USE_EXT_RCVINFO: val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO); break; case SCTP_AUTOCLOSE: if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE)) val = TICKS_TO_SEC(inp->sctp_ep.auto_close_time); else val = 0; break; default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); error = ENOPROTOOPT; } /* end switch (sopt->sopt_name) */ if (*optsize < sizeof(val)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } flags_out: SCTP_INP_RUNLOCK(inp); if (error == 0) { /* return the option value */ *(int *)optval = val; *optsize = sizeof(val); } break; case SCTP_GET_PACKET_LOG: { #ifdef SCTP_PACKET_LOGGING uint8_t *target; int ret; SCTP_CHECK_AND_CAST(target, optval, uint8_t, *optsize); ret = sctp_copy_out_packet_log(target, (int)*optsize); *optsize = ret; #else SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; #endif break; } case SCTP_REUSE_PORT: { uint32_t *value; if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE)) { /* Can't do this for a 1-m socket */ error = EINVAL; break; } SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); *value = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE); *optsize = sizeof(uint32_t); break; } case SCTP_PARTIAL_DELIVERY_POINT: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); *value = inp->partial_delivery_point; *optsize = sizeof(uint32_t); break; } case SCTP_FRAGMENT_INTERLEAVE: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE)) { if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS)) { *value = SCTP_FRAG_LEVEL_2; } else { *value = SCTP_FRAG_LEVEL_1; } } else { *value = SCTP_FRAG_LEVEL_0; } *optsize = sizeof(uint32_t); break; } case SCTP_CMT_ON_OFF: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.sctp_cmt_on_off; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->sctp_cmt_on_off; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_PLUGGABLE_CC: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.congestion_control_module; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->sctp_ep.sctp_default_cc_module; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_CC_OPTION: { struct sctp_cc_option *cc_opt; SCTP_CHECK_AND_CAST(cc_opt, optval, struct sctp_cc_option, *optsize); SCTP_FIND_STCB(inp, stcb, cc_opt->aid_value.assoc_id); if (stcb == NULL) { error = EINVAL; } else { if (stcb->asoc.cc_functions.sctp_cwnd_socket_option == NULL) { error = ENOTSUP; } else { error = (*stcb->asoc.cc_functions.sctp_cwnd_socket_option) (stcb, 0, cc_opt); *optsize = sizeof(struct sctp_cc_option); } SCTP_TCB_UNLOCK(stcb); } break; } case SCTP_PLUGGABLE_SS: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.stream_scheduling_module; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->sctp_ep.sctp_default_ss_module; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_SS_VALUE: { struct sctp_stream_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_stream_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { if ((av->stream_id >= stcb->asoc.streamoutcnt) || (stcb->asoc.ss_functions.sctp_ss_get_value(stcb, &stcb->asoc, &stcb->asoc.strmout[av->stream_id], &av->stream_value) < 0)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } else { *optsize = sizeof(struct sctp_stream_value); } SCTP_TCB_UNLOCK(stcb); } else { /* * Can't get stream value without * association */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } break; } case SCTP_GET_ADDR_LEN: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); error = EINVAL; #ifdef INET if (av->assoc_value == AF_INET) { av->assoc_value = sizeof(struct sockaddr_in); error = 0; } #endif #ifdef INET6 if (av->assoc_value == AF_INET6) { av->assoc_value = sizeof(struct sockaddr_in6); error = 0; } #endif if (error) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); } else { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_GET_ASSOC_NUMBER: { uint32_t *value, cnt; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); cnt = 0; SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { cnt++; } SCTP_INP_RUNLOCK(inp); *value = cnt; *optsize = sizeof(uint32_t); break; } case SCTP_GET_ASSOC_ID_LIST: { struct sctp_assoc_ids *ids; unsigned int at, limit; SCTP_CHECK_AND_CAST(ids, optval, struct sctp_assoc_ids, *optsize); at = 0; limit = (*optsize - sizeof(uint32_t)) / sizeof(sctp_assoc_t); SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { if (at < limit) { ids->gaids_assoc_id[at++] = sctp_get_associd(stcb); } else { error = EINVAL; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } SCTP_INP_RUNLOCK(inp); if (error == 0) { ids->gaids_number_of_ids = at; *optsize = ((at * sizeof(sctp_assoc_t)) + sizeof(uint32_t)); } break; } case SCTP_CONTEXT: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.context; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->sctp_context; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_VRF_ID: { uint32_t *default_vrfid; SCTP_CHECK_AND_CAST(default_vrfid, optval, uint32_t, *optsize); *default_vrfid = inp->def_vrf_id; *optsize = sizeof(uint32_t); break; } case SCTP_GET_ASOC_VRF: { struct sctp_assoc_value *id; SCTP_CHECK_AND_CAST(id, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, id->assoc_id); if (stcb == NULL) { error = EINVAL; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); } else { id->assoc_value = stcb->asoc.vrf_id; *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_GET_VRF_IDS: { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; break; } case SCTP_GET_NONCE_VALUES: { struct sctp_get_nonce_values *gnv; SCTP_CHECK_AND_CAST(gnv, optval, struct sctp_get_nonce_values, *optsize); SCTP_FIND_STCB(inp, stcb, gnv->gn_assoc_id); if (stcb) { gnv->gn_peers_tag = stcb->asoc.peer_vtag; gnv->gn_local_tag = stcb->asoc.my_vtag; SCTP_TCB_UNLOCK(stcb); *optsize = sizeof(struct sctp_get_nonce_values); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); error = ENOTCONN; } break; } case SCTP_DELAYED_SACK: { struct sctp_sack_info *sack; SCTP_CHECK_AND_CAST(sack, optval, struct sctp_sack_info, *optsize); SCTP_FIND_STCB(inp, stcb, sack->sack_assoc_id); if (stcb) { sack->sack_delay = stcb->asoc.delayed_ack; sack->sack_freq = stcb->asoc.sack_freq; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (sack->sack_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); sack->sack_delay = TICKS_TO_MSEC(inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV]); sack->sack_freq = inp->sctp_ep.sctp_sack_freq; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_sack_info); } break; } case SCTP_GET_SNDBUF_USE: { struct sctp_sockstat *ss; SCTP_CHECK_AND_CAST(ss, optval, struct sctp_sockstat, *optsize); SCTP_FIND_STCB(inp, stcb, ss->ss_assoc_id); if (stcb) { ss->ss_total_sndbuf = stcb->asoc.total_output_queue_size; ss->ss_total_recv_buf = (stcb->asoc.size_on_reasm_queue + stcb->asoc.size_on_all_streams); SCTP_TCB_UNLOCK(stcb); *optsize = sizeof(struct sctp_sockstat); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); error = ENOTCONN; } break; } case SCTP_MAX_BURST: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.max_burst; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->sctp_ep.max_burst; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_MAXSEG: { struct sctp_assoc_value *av; int ovh; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = sctp_get_frag_point(stcb, &stcb->asoc); SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MED_OVERHEAD; } else { ovh = SCTP_MED_V4_OVERHEAD; } if (inp->sctp_frag_point >= SCTP_DEFAULT_MAXSEGMENT) av->assoc_value = 0; else av->assoc_value = inp->sctp_frag_point - ovh; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_GET_STAT_LOG: error = sctp_fill_stat_log(optval, optsize); break; case SCTP_EVENTS: { struct sctp_event_subscribe *events; SCTP_CHECK_AND_CAST(events, optval, struct sctp_event_subscribe, *optsize); memset(events, 0, sizeof(struct sctp_event_subscribe)); SCTP_INP_RLOCK(inp); if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT)) events->sctp_data_io_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT)) events->sctp_association_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVPADDREVNT)) events->sctp_address_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT)) events->sctp_send_failure_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVPEERERR)) events->sctp_peer_error_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT)) events->sctp_shutdown_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT)) events->sctp_partial_delivery_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT)) events->sctp_adaptation_layer_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTHEVNT)) events->sctp_authentication_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_DRYEVNT)) events->sctp_sender_dry_event = 1; if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT)) events->sctp_stream_reset_event = 1; SCTP_INP_RUNLOCK(inp); *optsize = sizeof(struct sctp_event_subscribe); break; } case SCTP_ADAPTATION_LAYER: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); SCTP_INP_RLOCK(inp); *value = inp->sctp_ep.adaptation_layer_indicator; SCTP_INP_RUNLOCK(inp); *optsize = sizeof(uint32_t); break; } case SCTP_SET_INITIAL_DBG_SEQ: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); SCTP_INP_RLOCK(inp); *value = inp->sctp_ep.initial_sequence_debug; SCTP_INP_RUNLOCK(inp); *optsize = sizeof(uint32_t); break; } case SCTP_GET_LOCAL_ADDR_SIZE: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); SCTP_INP_RLOCK(inp); *value = sctp_count_max_addresses(inp); SCTP_INP_RUNLOCK(inp); *optsize = sizeof(uint32_t); break; } case SCTP_GET_REMOTE_ADDR_SIZE: { uint32_t *value; size_t size; struct sctp_nets *net; SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize); /* FIXME MT: change to sctp_assoc_value? */ SCTP_FIND_STCB(inp, stcb, (sctp_assoc_t) * value); if (stcb) { size = 0; /* Count the sizes */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { switch (net->ro._l_addr.sa.sa_family) { #ifdef INET case AF_INET: #ifdef INET6 if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { size += sizeof(struct sockaddr_in6); } else { size += sizeof(struct sockaddr_in); } #else size += sizeof(struct sockaddr_in); #endif break; #endif #ifdef INET6 case AF_INET6: size += sizeof(struct sockaddr_in6); break; #endif default: break; } } SCTP_TCB_UNLOCK(stcb); *value = (uint32_t) size; *optsize = sizeof(uint32_t); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); error = ENOTCONN; } break; } case SCTP_GET_PEER_ADDRESSES: /* * Get the address information, an array is passed in to * fill up we pack it. */ { size_t cpsz, left; struct sockaddr_storage *sas; struct sctp_nets *net; struct sctp_getaddresses *saddr; SCTP_CHECK_AND_CAST(saddr, optval, struct sctp_getaddresses, *optsize); SCTP_FIND_STCB(inp, stcb, saddr->sget_assoc_id); if (stcb) { left = (*optsize) - sizeof(struct sctp_getaddresses); *optsize = sizeof(struct sctp_getaddresses); sas = (struct sockaddr_storage *)&saddr->addr[0]; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { switch (net->ro._l_addr.sa.sa_family) { #ifdef INET case AF_INET: #ifdef INET6 if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { cpsz = sizeof(struct sockaddr_in6); } else { cpsz = sizeof(struct sockaddr_in); } #else cpsz = sizeof(struct sockaddr_in); #endif break; #endif #ifdef INET6 case AF_INET6: cpsz = sizeof(struct sockaddr_in6); break; #endif default: cpsz = 0; break; } if (cpsz == 0) { break; } if (left < cpsz) { /* not enough room. */ break; } #if defined(INET) && defined(INET6) if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) && (net->ro._l_addr.sa.sa_family == AF_INET)) { /* Must map the address */ in6_sin_2_v4mapsin6(&net->ro._l_addr.sin, (struct sockaddr_in6 *)sas); } else { memcpy(sas, &net->ro._l_addr, cpsz); } #else memcpy(sas, &net->ro._l_addr, cpsz); #endif ((struct sockaddr_in *)sas)->sin_port = stcb->rport; sas = (struct sockaddr_storage *)((caddr_t)sas + cpsz); left -= cpsz; *optsize += cpsz; } SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; } break; } case SCTP_GET_LOCAL_ADDRESSES: { size_t limit, actual; struct sockaddr_storage *sas; struct sctp_getaddresses *saddr; SCTP_CHECK_AND_CAST(saddr, optval, struct sctp_getaddresses, *optsize); SCTP_FIND_STCB(inp, stcb, saddr->sget_assoc_id); sas = (struct sockaddr_storage *)&saddr->addr[0]; limit = *optsize - sizeof(sctp_assoc_t); actual = sctp_fill_up_addresses(inp, stcb, limit, sas); if (stcb) { SCTP_TCB_UNLOCK(stcb); } *optsize = sizeof(struct sockaddr_storage) + actual; break; } case SCTP_PEER_ADDR_PARAMS: { struct sctp_paddrparams *paddrp; struct sctp_nets *net; struct sockaddr *addr; #if defined(INET) && defined(INET6) struct sockaddr_in sin_store; #endif SCTP_CHECK_AND_CAST(paddrp, optval, struct sctp_paddrparams, *optsize); SCTP_FIND_STCB(inp, stcb, paddrp->spp_assoc_id); #if defined(INET) && defined(INET6) if (paddrp->spp_address.ss_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&paddrp->spp_address; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { in6_sin6_2_sin(&sin_store, sin6); addr = (struct sockaddr *)&sin_store; } else { addr = (struct sockaddr *)&paddrp->spp_address; } } else { addr = (struct sockaddr *)&paddrp->spp_address; } #else addr = (struct sockaddr *)&paddrp->spp_address; #endif if (stcb != NULL) { net = sctp_findnet(stcb, addr); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ net = NULL; SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if ((stcb != NULL) && (net == NULL)) { #ifdef INET if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; if (sin->sin_addr.s_addr != INADDR_ANY) { error = EINVAL; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else #endif #ifdef INET6 if (addr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { error = EINVAL; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else #endif { error = EAFNOSUPPORT; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } if (stcb != NULL) { /* Applies to the specific association */ paddrp->spp_flags = 0; if (net != NULL) { paddrp->spp_hbinterval = net->heart_beat_delay; paddrp->spp_pathmaxrxt = net->failure_threshold; paddrp->spp_pathmtu = net->mtu; switch (net->ro._l_addr.sa.sa_family) { #ifdef INET case AF_INET: paddrp->spp_pathmtu -= SCTP_MIN_V4_OVERHEAD; break; #endif #ifdef INET6 case AF_INET6: paddrp->spp_pathmtu -= SCTP_MIN_V4_OVERHEAD; break; #endif default: break; } /* get flags for HB */ if (net->dest_state & SCTP_ADDR_NOHB) { paddrp->spp_flags |= SPP_HB_DISABLE; } else { paddrp->spp_flags |= SPP_HB_ENABLE; } /* get flags for PMTU */ if (net->dest_state & SCTP_ADDR_NO_PMTUD) { paddrp->spp_flags |= SPP_PMTUD_DISABLE; } else { paddrp->spp_flags |= SPP_PMTUD_ENABLE; } if (net->dscp & 0x01) { paddrp->spp_dscp = net->dscp & 0xfc; paddrp->spp_flags |= SPP_DSCP; } #ifdef INET6 if ((net->ro._l_addr.sa.sa_family == AF_INET6) && (net->flowlabel & 0x80000000)) { paddrp->spp_ipv6_flowlabel = net->flowlabel & 0x000fffff; paddrp->spp_flags |= SPP_IPV6_FLOWLABEL; } #endif } else { /* * No destination so return default * value */ paddrp->spp_pathmaxrxt = stcb->asoc.def_net_failure; paddrp->spp_pathmtu = 0; if (stcb->asoc.default_dscp & 0x01) { paddrp->spp_dscp = stcb->asoc.default_dscp & 0xfc; paddrp->spp_flags |= SPP_DSCP; } #ifdef INET6 if (stcb->asoc.default_flowlabel & 0x80000000) { paddrp->spp_ipv6_flowlabel = stcb->asoc.default_flowlabel & 0x000fffff; paddrp->spp_flags |= SPP_IPV6_FLOWLABEL; } #endif /* default settings should be these */ if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) { paddrp->spp_flags |= SPP_HB_DISABLE; } else { paddrp->spp_flags |= SPP_HB_ENABLE; } if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) { paddrp->spp_flags |= SPP_PMTUD_DISABLE; } else { paddrp->spp_flags |= SPP_PMTUD_ENABLE; } paddrp->spp_hbinterval = stcb->asoc.heart_beat_delay; } paddrp->spp_assoc_id = sctp_get_associd(stcb); SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (paddrp->spp_assoc_id == SCTP_FUTURE_ASSOC)) { /* Use endpoint defaults */ SCTP_INP_RLOCK(inp); paddrp->spp_pathmaxrxt = inp->sctp_ep.def_net_failure; paddrp->spp_hbinterval = TICKS_TO_MSEC(inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT]); paddrp->spp_assoc_id = SCTP_FUTURE_ASSOC; /* get inp's default */ if (inp->sctp_ep.default_dscp & 0x01) { paddrp->spp_dscp = inp->sctp_ep.default_dscp & 0xfc; paddrp->spp_flags |= SPP_DSCP; } #ifdef INET6 if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) && (inp->sctp_ep.default_flowlabel & 0x80000000)) { paddrp->spp_ipv6_flowlabel = inp->sctp_ep.default_flowlabel & 0x000fffff; paddrp->spp_flags |= SPP_IPV6_FLOWLABEL; } #endif /* can't return this */ paddrp->spp_pathmtu = 0; if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) { paddrp->spp_flags |= SPP_HB_ENABLE; } else { paddrp->spp_flags |= SPP_HB_DISABLE; } if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) { paddrp->spp_flags |= SPP_PMTUD_ENABLE; } else { paddrp->spp_flags |= SPP_PMTUD_DISABLE; } SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_paddrparams); } break; } case SCTP_GET_PEER_ADDR_INFO: { struct sctp_paddrinfo *paddri; struct sctp_nets *net; struct sockaddr *addr; #if defined(INET) && defined(INET6) struct sockaddr_in sin_store; #endif SCTP_CHECK_AND_CAST(paddri, optval, struct sctp_paddrinfo, *optsize); SCTP_FIND_STCB(inp, stcb, paddri->spinfo_assoc_id); #if defined(INET) && defined(INET6) if (paddri->spinfo_address.ss_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&paddri->spinfo_address; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { in6_sin6_2_sin(&sin_store, sin6); addr = (struct sockaddr *)&sin_store; } else { addr = (struct sockaddr *)&paddri->spinfo_address; } } else { addr = (struct sockaddr *)&paddri->spinfo_address; } #else addr = (struct sockaddr *)&paddri->spinfo_address; #endif if (stcb != NULL) { net = sctp_findnet(stcb, addr); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ net = NULL; SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if ((stcb != NULL) && (net != NULL)) { if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { /* It's unconfirmed */ paddri->spinfo_state = SCTP_UNCONFIRMED; } else if (net->dest_state & SCTP_ADDR_REACHABLE) { /* It's active */ paddri->spinfo_state = SCTP_ACTIVE; } else { /* It's inactive */ paddri->spinfo_state = SCTP_INACTIVE; } paddri->spinfo_cwnd = net->cwnd; paddri->spinfo_srtt = net->lastsa >> SCTP_RTT_SHIFT; paddri->spinfo_rto = net->RTO; paddri->spinfo_assoc_id = sctp_get_associd(stcb); paddri->spinfo_mtu = net->mtu; switch (addr->sa_family) { #if defined(INET) case AF_INET: paddri->spinfo_mtu -= SCTP_MIN_V4_OVERHEAD; break; #endif #if defined(INET6) case AF_INET6: paddri->spinfo_mtu -= SCTP_MIN_OVERHEAD; break; #endif default: break; } SCTP_TCB_UNLOCK(stcb); *optsize = sizeof(struct sctp_paddrinfo); } else { if (stcb != NULL) { SCTP_TCB_UNLOCK(stcb); } SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; } break; } case SCTP_PCB_STATUS: { struct sctp_pcbinfo *spcb; SCTP_CHECK_AND_CAST(spcb, optval, struct sctp_pcbinfo, *optsize); sctp_fill_pcbinfo(spcb); *optsize = sizeof(struct sctp_pcbinfo); break; } case SCTP_STATUS: { struct sctp_nets *net; struct sctp_status *sstat; SCTP_CHECK_AND_CAST(sstat, optval, struct sctp_status, *optsize); SCTP_FIND_STCB(inp, stcb, sstat->sstat_assoc_id); if (stcb == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } sstat->sstat_state = sctp_map_assoc_state(stcb->asoc.state); sstat->sstat_assoc_id = sctp_get_associd(stcb); sstat->sstat_rwnd = stcb->asoc.peers_rwnd; sstat->sstat_unackdata = stcb->asoc.sent_queue_cnt; /* * We can't include chunks that have been passed to * the socket layer. Only things in queue. */ sstat->sstat_penddata = (stcb->asoc.cnt_on_reasm_queue + stcb->asoc.cnt_on_all_streams); sstat->sstat_instrms = stcb->asoc.streamincnt; sstat->sstat_outstrms = stcb->asoc.streamoutcnt; sstat->sstat_fragmentation_point = sctp_get_frag_point(stcb, &stcb->asoc); memcpy(&sstat->sstat_primary.spinfo_address, &stcb->asoc.primary_destination->ro._l_addr, ((struct sockaddr *)(&stcb->asoc.primary_destination->ro._l_addr))->sa_len); net = stcb->asoc.primary_destination; ((struct sockaddr_in *)&sstat->sstat_primary.spinfo_address)->sin_port = stcb->rport; /* * Again the user can get info from sctp_constants.h * for what the state of the network is. */ if (net->dest_state & SCTP_ADDR_UNCONFIRMED) { /* It's unconfirmed */ sstat->sstat_primary.spinfo_state = SCTP_UNCONFIRMED; } else if (net->dest_state & SCTP_ADDR_REACHABLE) { /* It's active */ sstat->sstat_primary.spinfo_state = SCTP_ACTIVE; } else { /* It's inactive */ sstat->sstat_primary.spinfo_state = SCTP_INACTIVE; } sstat->sstat_primary.spinfo_cwnd = net->cwnd; sstat->sstat_primary.spinfo_srtt = net->lastsa >> SCTP_RTT_SHIFT; sstat->sstat_primary.spinfo_rto = net->RTO; sstat->sstat_primary.spinfo_mtu = net->mtu; switch (stcb->asoc.primary_destination->ro._l_addr.sa.sa_family) { #if defined(INET) case AF_INET: sstat->sstat_primary.spinfo_mtu -= SCTP_MIN_V4_OVERHEAD; break; #endif #if defined(INET6) case AF_INET6: sstat->sstat_primary.spinfo_mtu -= SCTP_MIN_OVERHEAD; break; #endif default: break; } sstat->sstat_primary.spinfo_assoc_id = sctp_get_associd(stcb); SCTP_TCB_UNLOCK(stcb); *optsize = sizeof(struct sctp_status); break; } case SCTP_RTOINFO: { struct sctp_rtoinfo *srto; SCTP_CHECK_AND_CAST(srto, optval, struct sctp_rtoinfo, *optsize); SCTP_FIND_STCB(inp, stcb, srto->srto_assoc_id); if (stcb) { srto->srto_initial = stcb->asoc.initial_rto; srto->srto_max = stcb->asoc.maxrto; srto->srto_min = stcb->asoc.minrto; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (srto->srto_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); srto->srto_initial = inp->sctp_ep.initial_rto; srto->srto_max = inp->sctp_ep.sctp_maxrto; srto->srto_min = inp->sctp_ep.sctp_minrto; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_rtoinfo); } break; } case SCTP_TIMEOUTS: { struct sctp_timeouts *stimo; SCTP_CHECK_AND_CAST(stimo, optval, struct sctp_timeouts, *optsize); SCTP_FIND_STCB(inp, stcb, stimo->stimo_assoc_id); if (stcb) { stimo->stimo_init = stcb->asoc.timoinit; stimo->stimo_data = stcb->asoc.timodata; stimo->stimo_sack = stcb->asoc.timosack; stimo->stimo_shutdown = stcb->asoc.timoshutdown; stimo->stimo_heartbeat = stcb->asoc.timoheartbeat; stimo->stimo_cookie = stcb->asoc.timocookie; stimo->stimo_shutdownack = stcb->asoc.timoshutdownack; SCTP_TCB_UNLOCK(stcb); *optsize = sizeof(struct sctp_timeouts); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } break; } case SCTP_ASSOCINFO: { struct sctp_assocparams *sasoc; SCTP_CHECK_AND_CAST(sasoc, optval, struct sctp_assocparams, *optsize); SCTP_FIND_STCB(inp, stcb, sasoc->sasoc_assoc_id); if (stcb) { sasoc->sasoc_cookie_life = TICKS_TO_MSEC(stcb->asoc.cookie_life); sasoc->sasoc_asocmaxrxt = stcb->asoc.max_send_times; sasoc->sasoc_number_peer_destinations = stcb->asoc.numnets; sasoc->sasoc_peer_rwnd = stcb->asoc.peers_rwnd; sasoc->sasoc_local_rwnd = stcb->asoc.my_rwnd; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (sasoc->sasoc_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); sasoc->sasoc_cookie_life = TICKS_TO_MSEC(inp->sctp_ep.def_cookie_life); sasoc->sasoc_asocmaxrxt = inp->sctp_ep.max_send_times; sasoc->sasoc_number_peer_destinations = 0; sasoc->sasoc_peer_rwnd = 0; sasoc->sasoc_local_rwnd = sbspace(&inp->sctp_socket->so_rcv); SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assocparams); } break; } case SCTP_DEFAULT_SEND_PARAM: { struct sctp_sndrcvinfo *s_info; SCTP_CHECK_AND_CAST(s_info, optval, struct sctp_sndrcvinfo, *optsize); SCTP_FIND_STCB(inp, stcb, s_info->sinfo_assoc_id); if (stcb) { memcpy(s_info, &stcb->asoc.def_send, sizeof(stcb->asoc.def_send)); SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (s_info->sinfo_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); memcpy(s_info, &inp->def_send, sizeof(inp->def_send)); SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_sndrcvinfo); } break; } case SCTP_INITMSG: { struct sctp_initmsg *sinit; SCTP_CHECK_AND_CAST(sinit, optval, struct sctp_initmsg, *optsize); SCTP_INP_RLOCK(inp); sinit->sinit_num_ostreams = inp->sctp_ep.pre_open_stream_count; sinit->sinit_max_instreams = inp->sctp_ep.max_open_streams_intome; sinit->sinit_max_attempts = inp->sctp_ep.max_init_times; sinit->sinit_max_init_timeo = inp->sctp_ep.initial_init_rto_max; SCTP_INP_RUNLOCK(inp); *optsize = sizeof(struct sctp_initmsg); break; } case SCTP_PRIMARY_ADDR: /* we allow a "get" operation on this */ { struct sctp_setprim *ssp; SCTP_CHECK_AND_CAST(ssp, optval, struct sctp_setprim, *optsize); SCTP_FIND_STCB(inp, stcb, ssp->ssp_assoc_id); if (stcb) { union sctp_sockstore *addr; addr = &stcb->asoc.primary_destination->ro._l_addr; switch (addr->sa.sa_family) { #ifdef INET case AF_INET: #ifdef INET6 if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) { in6_sin_2_v4mapsin6(&addr->sin, (struct sockaddr_in6 *)&ssp->ssp_addr); } else { memcpy(&ssp->ssp_addr, &addr->sin, sizeof(struct sockaddr_in)); } #else memcpy(&ssp->ssp_addr, &addr->sin, sizeof(struct sockaddr_in)); #endif break; #endif #ifdef INET6 case AF_INET6: memcpy(&ssp->ssp_addr, &addr->sin6, sizeof(struct sockaddr_in6)); break; #endif default: break; } SCTP_TCB_UNLOCK(stcb); *optsize = sizeof(struct sctp_setprim); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } break; } case SCTP_HMAC_IDENT: { struct sctp_hmacalgo *shmac; sctp_hmaclist_t *hmaclist; uint32_t size; int i; SCTP_CHECK_AND_CAST(shmac, optval, struct sctp_hmacalgo, *optsize); SCTP_INP_RLOCK(inp); hmaclist = inp->sctp_ep.local_hmacs; if (hmaclist == NULL) { /* no HMACs to return */ *optsize = sizeof(*shmac); SCTP_INP_RUNLOCK(inp); break; } /* is there room for all of the hmac ids? */ size = sizeof(*shmac) + (hmaclist->num_algo * sizeof(shmac->shmac_idents[0])); if ((size_t)(*optsize) < size) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_INP_RUNLOCK(inp); break; } /* copy in the list */ shmac->shmac_number_of_idents = hmaclist->num_algo; for (i = 0; i < hmaclist->num_algo; i++) { shmac->shmac_idents[i] = hmaclist->hmac[i]; } SCTP_INP_RUNLOCK(inp); *optsize = size; break; } case SCTP_AUTH_ACTIVE_KEY: { struct sctp_authkeyid *scact; SCTP_CHECK_AND_CAST(scact, optval, struct sctp_authkeyid, *optsize); SCTP_FIND_STCB(inp, stcb, scact->scact_assoc_id); if (stcb) { /* get the active key on the assoc */ scact->scact_keynumber = stcb->asoc.authinfo.active_keyid; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (scact->scact_assoc_id == SCTP_FUTURE_ASSOC)) { /* get the endpoint active key */ SCTP_INP_RLOCK(inp); scact->scact_keynumber = inp->sctp_ep.default_keyid; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_authkeyid); } break; } case SCTP_LOCAL_AUTH_CHUNKS: { struct sctp_authchunks *sac; sctp_auth_chklist_t *chklist = NULL; size_t size = 0; SCTP_CHECK_AND_CAST(sac, optval, struct sctp_authchunks, *optsize); SCTP_FIND_STCB(inp, stcb, sac->gauth_assoc_id); if (stcb) { /* get off the assoc */ chklist = stcb->asoc.local_auth_chunks; /* is there enough space? */ size = sctp_auth_get_chklist_size(chklist); if (*optsize < (sizeof(struct sctp_authchunks) + size)) { error = EINVAL; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); } else { /* copy in the chunks */ (void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks); sac->gauth_number_of_chunks = (uint32_t) size; *optsize = sizeof(struct sctp_authchunks) + size; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (sac->gauth_assoc_id == SCTP_FUTURE_ASSOC)) { /* get off the endpoint */ SCTP_INP_RLOCK(inp); chklist = inp->sctp_ep.local_auth_chunks; /* is there enough space? */ size = sctp_auth_get_chklist_size(chklist); if (*optsize < (sizeof(struct sctp_authchunks) + size)) { error = EINVAL; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); } else { /* copy in the chunks */ (void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks); sac->gauth_number_of_chunks = (uint32_t) size; *optsize = sizeof(struct sctp_authchunks) + size; } SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_PEER_AUTH_CHUNKS: { struct sctp_authchunks *sac; sctp_auth_chklist_t *chklist = NULL; size_t size = 0; SCTP_CHECK_AND_CAST(sac, optval, struct sctp_authchunks, *optsize); SCTP_FIND_STCB(inp, stcb, sac->gauth_assoc_id); if (stcb) { /* get off the assoc */ chklist = stcb->asoc.peer_auth_chunks; /* is there enough space? */ size = sctp_auth_get_chklist_size(chklist); if (*optsize < (sizeof(struct sctp_authchunks) + size)) { error = EINVAL; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); } else { /* copy in the chunks */ (void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks); sac->gauth_number_of_chunks = (uint32_t) size; *optsize = sizeof(struct sctp_authchunks) + size; } SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; } break; } case SCTP_EVENT: { struct sctp_event *event; uint32_t event_type; SCTP_CHECK_AND_CAST(event, optval, struct sctp_event, *optsize); SCTP_FIND_STCB(inp, stcb, event->se_assoc_id); switch (event->se_type) { case SCTP_ASSOC_CHANGE: event_type = SCTP_PCB_FLAGS_RECVASSOCEVNT; break; case SCTP_PEER_ADDR_CHANGE: event_type = SCTP_PCB_FLAGS_RECVPADDREVNT; break; case SCTP_REMOTE_ERROR: event_type = SCTP_PCB_FLAGS_RECVPEERERR; break; case SCTP_SEND_FAILED: event_type = SCTP_PCB_FLAGS_RECVSENDFAILEVNT; break; case SCTP_SHUTDOWN_EVENT: event_type = SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT; break; case SCTP_ADAPTATION_INDICATION: event_type = SCTP_PCB_FLAGS_ADAPTATIONEVNT; break; case SCTP_PARTIAL_DELIVERY_EVENT: event_type = SCTP_PCB_FLAGS_PDAPIEVNT; break; case SCTP_AUTHENTICATION_EVENT: event_type = SCTP_PCB_FLAGS_AUTHEVNT; break; case SCTP_STREAM_RESET_EVENT: event_type = SCTP_PCB_FLAGS_STREAM_RESETEVNT; break; case SCTP_SENDER_DRY_EVENT: event_type = SCTP_PCB_FLAGS_DRYEVNT; break; case SCTP_NOTIFICATIONS_STOPPED_EVENT: event_type = 0; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTSUP); error = ENOTSUP; break; case SCTP_ASSOC_RESET_EVENT: event_type = SCTP_PCB_FLAGS_ASSOC_RESETEVNT; break; case SCTP_STREAM_CHANGE_EVENT: event_type = SCTP_PCB_FLAGS_STREAM_CHANGEEVNT; break; case SCTP_SEND_FAILED_EVENT: event_type = SCTP_PCB_FLAGS_RECVNSENDFAILEVNT; break; default: event_type = 0; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } if (event_type > 0) { if (stcb) { event->se_on = sctp_stcb_is_feature_on(inp, stcb, event_type); SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (event->se_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); event->se_on = sctp_is_feature_on(inp, event_type); SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } } if (error == 0) { *optsize = sizeof(struct sctp_event); } break; } case SCTP_RECVRCVINFO: { int onoff; if (*optsize < sizeof(int)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } else { SCTP_INP_RLOCK(inp); onoff = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVRCVINFO); SCTP_INP_RUNLOCK(inp); } if (error == 0) { /* return the option value */ *(int *)optval = onoff; *optsize = sizeof(int); } break; } case SCTP_RECVNXTINFO: { int onoff; if (*optsize < sizeof(int)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } else { SCTP_INP_RLOCK(inp); onoff = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVNXTINFO); SCTP_INP_RUNLOCK(inp); } if (error == 0) { /* return the option value */ *(int *)optval = onoff; *optsize = sizeof(int); } break; } case SCTP_DEFAULT_SNDINFO: { struct sctp_sndinfo *info; SCTP_CHECK_AND_CAST(info, optval, struct sctp_sndinfo, *optsize); SCTP_FIND_STCB(inp, stcb, info->snd_assoc_id); if (stcb) { info->snd_sid = stcb->asoc.def_send.sinfo_stream; info->snd_flags = stcb->asoc.def_send.sinfo_flags; info->snd_flags &= 0xfff0; info->snd_ppid = stcb->asoc.def_send.sinfo_ppid; info->snd_context = stcb->asoc.def_send.sinfo_context; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (info->snd_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); info->snd_sid = inp->def_send.sinfo_stream; info->snd_flags = inp->def_send.sinfo_flags; info->snd_flags &= 0xfff0; info->snd_ppid = inp->def_send.sinfo_ppid; info->snd_context = inp->def_send.sinfo_context; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_sndinfo); } break; } case SCTP_DEFAULT_PRINFO: { struct sctp_default_prinfo *info; SCTP_CHECK_AND_CAST(info, optval, struct sctp_default_prinfo, *optsize); SCTP_FIND_STCB(inp, stcb, info->pr_assoc_id); if (stcb) { info->pr_policy = PR_SCTP_POLICY(stcb->asoc.def_send.sinfo_flags); info->pr_value = stcb->asoc.def_send.sinfo_timetolive; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (info->pr_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); info->pr_policy = PR_SCTP_POLICY(inp->def_send.sinfo_flags); info->pr_value = inp->def_send.sinfo_timetolive; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_default_prinfo); } break; } case SCTP_PEER_ADDR_THLDS: { struct sctp_paddrthlds *thlds; struct sctp_nets *net; struct sockaddr *addr; #if defined(INET) && defined(INET6) struct sockaddr_in sin_store; #endif SCTP_CHECK_AND_CAST(thlds, optval, struct sctp_paddrthlds, *optsize); SCTP_FIND_STCB(inp, stcb, thlds->spt_assoc_id); #if defined(INET) && defined(INET6) if (thlds->spt_address.ss_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&thlds->spt_address; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { in6_sin6_2_sin(&sin_store, sin6); addr = (struct sockaddr *)&sin_store; } else { addr = (struct sockaddr *)&thlds->spt_address; } } else { addr = (struct sockaddr *)&thlds->spt_address; } #else addr = (struct sockaddr *)&thlds->spt_address; #endif if (stcb != NULL) { net = sctp_findnet(stcb, addr); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ net = NULL; SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if ((stcb != NULL) && (net == NULL)) { #ifdef INET if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; if (sin->sin_addr.s_addr != INADDR_ANY) { error = EINVAL; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else #endif #ifdef INET6 if (addr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { error = EINVAL; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else #endif { error = EAFNOSUPPORT; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } if (stcb != NULL) { if (net != NULL) { thlds->spt_pathmaxrxt = net->failure_threshold; thlds->spt_pathpfthld = net->pf_threshold; } else { thlds->spt_pathmaxrxt = stcb->asoc.def_net_failure; thlds->spt_pathpfthld = stcb->asoc.def_net_pf_threshold; } thlds->spt_assoc_id = sctp_get_associd(stcb); SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (thlds->spt_assoc_id == SCTP_FUTURE_ASSOC)) { /* Use endpoint defaults */ SCTP_INP_RLOCK(inp); thlds->spt_pathmaxrxt = inp->sctp_ep.def_net_failure; thlds->spt_pathpfthld = inp->sctp_ep.def_net_pf_threshold; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_paddrthlds); } break; } case SCTP_REMOTE_UDP_ENCAPS_PORT: { struct sctp_udpencaps *encaps; struct sctp_nets *net; struct sockaddr *addr; #if defined(INET) && defined(INET6) struct sockaddr_in sin_store; #endif SCTP_CHECK_AND_CAST(encaps, optval, struct sctp_udpencaps, *optsize); SCTP_FIND_STCB(inp, stcb, encaps->sue_assoc_id); #if defined(INET) && defined(INET6) if (encaps->sue_address.ss_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&encaps->sue_address; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { in6_sin6_2_sin(&sin_store, sin6); addr = (struct sockaddr *)&sin_store; } else { addr = (struct sockaddr *)&encaps->sue_address; } } else { addr = (struct sockaddr *)&encaps->sue_address; } #else addr = (struct sockaddr *)&encaps->sue_address; #endif if (stcb) { net = sctp_findnet(stcb, addr); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ net = NULL; SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if ((stcb != NULL) && (net == NULL)) { #ifdef INET if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; if (sin->sin_addr.s_addr != INADDR_ANY) { error = EINVAL; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else #endif #ifdef INET6 if (addr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { error = EINVAL; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else #endif { error = EAFNOSUPPORT; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } if (stcb != NULL) { if (net) { encaps->sue_port = net->port; } else { encaps->sue_port = stcb->asoc.port; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (encaps->sue_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); encaps->sue_port = inp->sctp_ep.port; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_udpencaps); } break; } case SCTP_ECN_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.ecn_supported; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->ecn_supported; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_PR_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.prsctp_supported; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->prsctp_supported; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_AUTH_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.auth_supported; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->auth_supported; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_ASCONF_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.asconf_supported; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->asconf_supported; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_RECONFIG_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.reconfig_supported; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->reconfig_supported; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_NRSACK_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.nrsack_supported; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->nrsack_supported; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_PKTDROP_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.pktdrop_supported; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->pktdrop_supported; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_ENABLE_STREAM_RESET: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = (uint32_t) stcb->asoc.local_strreset_support; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = (uint32_t) inp->local_strreset_support; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } case SCTP_PR_STREAM_STATUS: { struct sctp_prstatus *sprstat; uint16_t sid; uint16_t policy; SCTP_CHECK_AND_CAST(sprstat, optval, struct sctp_prstatus, *optsize); SCTP_FIND_STCB(inp, stcb, sprstat->sprstat_assoc_id); sid = sprstat->sprstat_sid; policy = sprstat->sprstat_policy; #if defined(SCTP_DETAILED_STR_STATS) if ((stcb != NULL) && (sid < stcb->asoc.streamoutcnt) && (policy != SCTP_PR_SCTP_NONE) && ((policy <= SCTP_PR_SCTP_MAX) || (policy == SCTP_PR_SCTP_ALL))) { if (policy == SCTP_PR_SCTP_ALL) { sprstat->sprstat_abandoned_unsent = stcb->asoc.strmout[sid].abandoned_unsent[0]; sprstat->sprstat_abandoned_sent = stcb->asoc.strmout[sid].abandoned_sent[0]; } else { sprstat->sprstat_abandoned_unsent = stcb->asoc.strmout[sid].abandoned_unsent[policy]; sprstat->sprstat_abandoned_sent = stcb->asoc.strmout[sid].abandoned_sent[policy]; } #else if ((stcb != NULL) && (sid < stcb->asoc.streamoutcnt) && (policy == SCTP_PR_SCTP_ALL)) { sprstat->sprstat_abandoned_unsent = stcb->asoc.strmout[sid].abandoned_unsent[0]; sprstat->sprstat_abandoned_sent = stcb->asoc.strmout[sid].abandoned_sent[0]; #endif SCTP_TCB_UNLOCK(stcb); *optsize = sizeof(struct sctp_prstatus); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } break; } case SCTP_PR_ASSOC_STATUS: { struct sctp_prstatus *sprstat; uint16_t policy; SCTP_CHECK_AND_CAST(sprstat, optval, struct sctp_prstatus, *optsize); SCTP_FIND_STCB(inp, stcb, sprstat->sprstat_assoc_id); policy = sprstat->sprstat_policy; if ((stcb != NULL) && (policy != SCTP_PR_SCTP_NONE) && ((policy <= SCTP_PR_SCTP_MAX) || (policy == SCTP_PR_SCTP_ALL))) { if (policy == SCTP_PR_SCTP_ALL) { sprstat->sprstat_abandoned_unsent = stcb->asoc.abandoned_unsent[0]; sprstat->sprstat_abandoned_sent = stcb->asoc.abandoned_sent[0]; } else { sprstat->sprstat_abandoned_unsent = stcb->asoc.abandoned_unsent[policy]; sprstat->sprstat_abandoned_sent = stcb->asoc.abandoned_sent[policy]; } SCTP_TCB_UNLOCK(stcb); *optsize = sizeof(struct sctp_prstatus); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } break; } case SCTP_MAX_CWND: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { av->assoc_value = stcb->asoc.max_cwnd; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_RLOCK(inp); av->assoc_value = inp->max_cwnd; SCTP_INP_RUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } if (error == 0) { *optsize = sizeof(struct sctp_assoc_value); } break; } default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); error = ENOPROTOOPT; break; } /* end switch (sopt->sopt_name) */ if (error) { *optsize = 0; } return (error); } static int sctp_setopt(struct socket *so, int optname, void *optval, size_t optsize, void *p) { int error, set_opt; uint32_t *mopt; struct sctp_tcb *stcb = NULL; struct sctp_inpcb *inp = NULL; uint32_t vrf_id; if (optval == NULL) { SCTP_PRINTF("optval is NULL\n"); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_PRINTF("inp is NULL?\n"); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } vrf_id = inp->def_vrf_id; error = 0; switch (optname) { case SCTP_NODELAY: case SCTP_AUTOCLOSE: case SCTP_AUTO_ASCONF: case SCTP_EXPLICIT_EOR: case SCTP_DISABLE_FRAGMENTS: case SCTP_USE_EXT_RCVINFO: case SCTP_I_WANT_MAPPED_V4_ADDR: /* copy in the option value */ SCTP_CHECK_AND_CAST(mopt, optval, uint32_t, optsize); set_opt = 0; if (error) break; switch (optname) { case SCTP_DISABLE_FRAGMENTS: set_opt = SCTP_PCB_FLAGS_NO_FRAGMENT; break; case SCTP_AUTO_ASCONF: /* * NOTE: we don't really support this flag */ if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { /* only valid for bound all sockets */ if ((SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) && (*mopt != 0)) { /* forbidden by admin */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EPERM); return (EPERM); } set_opt = SCTP_PCB_FLAGS_AUTO_ASCONF; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } break; case SCTP_EXPLICIT_EOR: set_opt = SCTP_PCB_FLAGS_EXPLICIT_EOR; break; case SCTP_USE_EXT_RCVINFO: set_opt = SCTP_PCB_FLAGS_EXT_RCVINFO; break; case SCTP_I_WANT_MAPPED_V4_ADDR: if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { set_opt = SCTP_PCB_FLAGS_NEEDS_MAPPED_V4; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } break; case SCTP_NODELAY: set_opt = SCTP_PCB_FLAGS_NODELAY; break; case SCTP_AUTOCLOSE: if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } set_opt = SCTP_PCB_FLAGS_AUTOCLOSE; /* * The value is in ticks. Note this does not effect * old associations, only new ones. */ inp->sctp_ep.auto_close_time = SEC_TO_TICKS(*mopt); break; } SCTP_INP_WLOCK(inp); if (*mopt != 0) { sctp_feature_on(inp, set_opt); } else { sctp_feature_off(inp, set_opt); } SCTP_INP_WUNLOCK(inp); break; case SCTP_REUSE_PORT: { SCTP_CHECK_AND_CAST(mopt, optval, uint32_t, optsize); if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) { /* Can't set it after we are bound */ error = EINVAL; break; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE)) { /* Can't do this for a 1-m socket */ error = EINVAL; break; } if (optval) sctp_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE); else sctp_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE); break; } case SCTP_PARTIAL_DELIVERY_POINT: { uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, optsize); if (*value > SCTP_SB_LIMIT_RCV(so)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } inp->partial_delivery_point = *value; break; } case SCTP_FRAGMENT_INTERLEAVE: /* not yet until we re-write sctp_recvmsg() */ { uint32_t *level; SCTP_CHECK_AND_CAST(level, optval, uint32_t, optsize); if (*level == SCTP_FRAG_LEVEL_2) { sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else if (*level == SCTP_FRAG_LEVEL_1) { sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else if (*level == SCTP_FRAG_LEVEL_0) { sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE); sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } break; } case SCTP_CMT_ON_OFF: if (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); if (av->assoc_value > SCTP_CMT_MAX) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { stcb->asoc.sctp_cmt_on_off = av->assoc_value; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); inp->sctp_cmt_on_off = av->assoc_value; SCTP_INP_WUNLOCK(inp); } if ((av->assoc_id == SCTP_CURRENT_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); stcb->asoc.sctp_cmt_on_off = av->assoc_value; SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); error = ENOPROTOOPT; } break; case SCTP_PLUGGABLE_CC: { struct sctp_assoc_value *av; struct sctp_nets *net; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); if ((av->assoc_value != SCTP_CC_RFC2581) && (av->assoc_value != SCTP_CC_HSTCP) && (av->assoc_value != SCTP_CC_HTCP) && (av->assoc_value != SCTP_CC_RTCC)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { stcb->asoc.cc_functions = sctp_cc_functions[av->assoc_value]; stcb->asoc.congestion_control_module = av->assoc_value; if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net); } } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); inp->sctp_ep.sctp_default_cc_module = av->assoc_value; SCTP_INP_WUNLOCK(inp); } if ((av->assoc_id == SCTP_CURRENT_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); stcb->asoc.cc_functions = sctp_cc_functions[av->assoc_value]; stcb->asoc.congestion_control_module = av->assoc_value; if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net); } } SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_CC_OPTION: { struct sctp_cc_option *cc_opt; SCTP_CHECK_AND_CAST(cc_opt, optval, struct sctp_cc_option, optsize); SCTP_FIND_STCB(inp, stcb, cc_opt->aid_value.assoc_id); if (stcb == NULL) { if (cc_opt->aid_value.assoc_id == SCTP_CURRENT_ASSOC) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); if (stcb->asoc.cc_functions.sctp_cwnd_socket_option) { (*stcb->asoc.cc_functions.sctp_cwnd_socket_option) (stcb, 1, cc_opt); } SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } else { error = EINVAL; } } else { if (stcb->asoc.cc_functions.sctp_cwnd_socket_option == NULL) { error = ENOTSUP; } else { error = (*stcb->asoc.cc_functions.sctp_cwnd_socket_option) (stcb, 1, cc_opt); } SCTP_TCB_UNLOCK(stcb); } break; } case SCTP_PLUGGABLE_SS: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); if ((av->assoc_value != SCTP_SS_DEFAULT) && (av->assoc_value != SCTP_SS_ROUND_ROBIN) && (av->assoc_value != SCTP_SS_ROUND_ROBIN_PACKET) && (av->assoc_value != SCTP_SS_PRIORITY) && (av->assoc_value != SCTP_SS_FAIR_BANDWITH) && (av->assoc_value != SCTP_SS_FIRST_COME)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { stcb->asoc.ss_functions.sctp_ss_clear(stcb, &stcb->asoc, 1, 1); stcb->asoc.ss_functions = sctp_ss_functions[av->assoc_value]; stcb->asoc.stream_scheduling_module = av->assoc_value; stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 1); SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); inp->sctp_ep.sctp_default_ss_module = av->assoc_value; SCTP_INP_WUNLOCK(inp); } if ((av->assoc_id == SCTP_CURRENT_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); stcb->asoc.ss_functions.sctp_ss_clear(stcb, &stcb->asoc, 1, 1); stcb->asoc.ss_functions = sctp_ss_functions[av->assoc_value]; stcb->asoc.stream_scheduling_module = av->assoc_value; stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 1); SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_SS_VALUE: { struct sctp_stream_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_stream_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { if ((av->stream_id >= stcb->asoc.streamoutcnt) || (stcb->asoc.ss_functions.sctp_ss_set_value(stcb, &stcb->asoc, &stcb->asoc.strmout[av->stream_id], av->stream_value) < 0)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { if (av->assoc_id == SCTP_CURRENT_ASSOC) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); if (av->stream_id < stcb->asoc.streamoutcnt) { stcb->asoc.ss_functions.sctp_ss_set_value(stcb, &stcb->asoc, &stcb->asoc.strmout[av->stream_id], av->stream_value); } SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } else { /* * Can't set stream value without * association */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_CLR_STAT_LOG: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; break; case SCTP_CONTEXT: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { stcb->asoc.context = av->assoc_value; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); inp->sctp_context = av->assoc_value; SCTP_INP_WUNLOCK(inp); } if ((av->assoc_id == SCTP_CURRENT_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); stcb->asoc.context = av->assoc_value; SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_VRF_ID: { uint32_t *default_vrfid; SCTP_CHECK_AND_CAST(default_vrfid, optval, uint32_t, optsize); if (*default_vrfid > SCTP_MAX_VRF_ID) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } inp->def_vrf_id = *default_vrfid; break; } case SCTP_DEL_VRF_ID: { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; break; } case SCTP_ADD_VRF_ID: { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; break; } case SCTP_DELAYED_SACK: { struct sctp_sack_info *sack; SCTP_CHECK_AND_CAST(sack, optval, struct sctp_sack_info, optsize); SCTP_FIND_STCB(inp, stcb, sack->sack_assoc_id); if (sack->sack_delay) { if (sack->sack_delay > SCTP_MAX_SACK_DELAY) sack->sack_delay = SCTP_MAX_SACK_DELAY; if (MSEC_TO_TICKS(sack->sack_delay) < 1) { sack->sack_delay = TICKS_TO_MSEC(1); } } if (stcb) { if (sack->sack_delay) { stcb->asoc.delayed_ack = sack->sack_delay; } if (sack->sack_freq) { stcb->asoc.sack_freq = sack->sack_freq; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (sack->sack_assoc_id == SCTP_FUTURE_ASSOC) || (sack->sack_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); if (sack->sack_delay) { inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(sack->sack_delay); } if (sack->sack_freq) { inp->sctp_ep.sctp_sack_freq = sack->sack_freq; } SCTP_INP_WUNLOCK(inp); } if ((sack->sack_assoc_id == SCTP_CURRENT_ASSOC) || (sack->sack_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); if (sack->sack_delay) { stcb->asoc.delayed_ack = sack->sack_delay; } if (sack->sack_freq) { stcb->asoc.sack_freq = sack->sack_freq; } SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_AUTH_CHUNK: { struct sctp_authchunk *sauth; SCTP_CHECK_AND_CAST(sauth, optval, struct sctp_authchunk, optsize); SCTP_INP_WLOCK(inp); if (sctp_auth_add_chunk(sauth->sauth_chunk, inp->sctp_ep.local_auth_chunks)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_INP_WUNLOCK(inp); break; } case SCTP_AUTH_KEY: { struct sctp_authkey *sca; struct sctp_keyhead *shared_keys; sctp_sharedkey_t *shared_key; sctp_key_t *key = NULL; size_t size; SCTP_CHECK_AND_CAST(sca, optval, struct sctp_authkey, optsize); if (sca->sca_keylength == 0) { size = optsize - sizeof(struct sctp_authkey); } else { if (sca->sca_keylength + sizeof(struct sctp_authkey) <= optsize) { size = sca->sca_keylength; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } } SCTP_FIND_STCB(inp, stcb, sca->sca_assoc_id); if (stcb) { shared_keys = &stcb->asoc.shared_keys; /* clear the cached keys for this key id */ sctp_clear_cachedkeys(stcb, sca->sca_keynumber); /* * create the new shared key and * insert/replace it */ if (size > 0) { key = sctp_set_key(sca->sca_key, (uint32_t) size); if (key == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; SCTP_TCB_UNLOCK(stcb); break; } } shared_key = sctp_alloc_sharedkey(); if (shared_key == NULL) { sctp_free_key(key); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; SCTP_TCB_UNLOCK(stcb); break; } shared_key->key = key; shared_key->keyid = sca->sca_keynumber; error = sctp_insert_sharedkey(shared_keys, shared_key); SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (sca->sca_assoc_id == SCTP_FUTURE_ASSOC) || (sca->sca_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); shared_keys = &inp->sctp_ep.shared_keys; /* * clear the cached keys on all * assocs for this key id */ sctp_clear_cachedkeys_ep(inp, sca->sca_keynumber); /* * create the new shared key and * insert/replace it */ if (size > 0) { key = sctp_set_key(sca->sca_key, (uint32_t) size); if (key == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; SCTP_INP_WUNLOCK(inp); break; } } shared_key = sctp_alloc_sharedkey(); if (shared_key == NULL) { sctp_free_key(key); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; SCTP_INP_WUNLOCK(inp); break; } shared_key->key = key; shared_key->keyid = sca->sca_keynumber; error = sctp_insert_sharedkey(shared_keys, shared_key); SCTP_INP_WUNLOCK(inp); } if ((sca->sca_assoc_id == SCTP_CURRENT_ASSOC) || (sca->sca_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); shared_keys = &stcb->asoc.shared_keys; /* * clear the cached keys for * this key id */ sctp_clear_cachedkeys(stcb, sca->sca_keynumber); /* * create the new shared key * and insert/replace it */ if (size > 0) { key = sctp_set_key(sca->sca_key, (uint32_t) size); if (key == NULL) { SCTP_TCB_UNLOCK(stcb); continue; } } shared_key = sctp_alloc_sharedkey(); if (shared_key == NULL) { sctp_free_key(key); SCTP_TCB_UNLOCK(stcb); continue; } shared_key->key = key; shared_key->keyid = sca->sca_keynumber; error = sctp_insert_sharedkey(shared_keys, shared_key); SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_HMAC_IDENT: { struct sctp_hmacalgo *shmac; sctp_hmaclist_t *hmaclist; uint16_t hmacid; uint32_t i; SCTP_CHECK_AND_CAST(shmac, optval, struct sctp_hmacalgo, optsize); if ((optsize < sizeof(struct sctp_hmacalgo) + shmac->shmac_number_of_idents * sizeof(uint16_t)) || (shmac->shmac_number_of_idents > 0xffff)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } hmaclist = sctp_alloc_hmaclist((uint16_t) shmac->shmac_number_of_idents); if (hmaclist == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; break; } for (i = 0; i < shmac->shmac_number_of_idents; i++) { hmacid = shmac->shmac_idents[i]; if (sctp_auth_add_hmacid(hmaclist, hmacid)) { /* invalid HMACs were found */ ; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; sctp_free_hmaclist(hmaclist); goto sctp_set_hmac_done; } } for (i = 0; i < hmaclist->num_algo; i++) { if (hmaclist->hmac[i] == SCTP_AUTH_HMAC_ID_SHA1) { /* already in list */ break; } } if (i == hmaclist->num_algo) { /* not found in list */ sctp_free_hmaclist(hmaclist); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } /* set it on the endpoint */ SCTP_INP_WLOCK(inp); if (inp->sctp_ep.local_hmacs) sctp_free_hmaclist(inp->sctp_ep.local_hmacs); inp->sctp_ep.local_hmacs = hmaclist; SCTP_INP_WUNLOCK(inp); sctp_set_hmac_done: break; } case SCTP_AUTH_ACTIVE_KEY: { struct sctp_authkeyid *scact; SCTP_CHECK_AND_CAST(scact, optval, struct sctp_authkeyid, optsize); SCTP_FIND_STCB(inp, stcb, scact->scact_assoc_id); /* set the active key on the right place */ if (stcb) { /* set the active key on the assoc */ if (sctp_auth_setactivekey(stcb, scact->scact_keynumber)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (scact->scact_assoc_id == SCTP_FUTURE_ASSOC) || (scact->scact_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); if (sctp_auth_setactivekey_ep(inp, scact->scact_keynumber)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_INP_WUNLOCK(inp); } if ((scact->scact_assoc_id == SCTP_CURRENT_ASSOC) || (scact->scact_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); sctp_auth_setactivekey(stcb, scact->scact_keynumber); SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_AUTH_DELETE_KEY: { struct sctp_authkeyid *scdel; SCTP_CHECK_AND_CAST(scdel, optval, struct sctp_authkeyid, optsize); SCTP_FIND_STCB(inp, stcb, scdel->scact_assoc_id); /* delete the key from the right place */ if (stcb) { if (sctp_delete_sharedkey(stcb, scdel->scact_keynumber)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (scdel->scact_assoc_id == SCTP_FUTURE_ASSOC) || (scdel->scact_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); if (sctp_delete_sharedkey_ep(inp, scdel->scact_keynumber)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_INP_WUNLOCK(inp); } if ((scdel->scact_assoc_id == SCTP_CURRENT_ASSOC) || (scdel->scact_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); sctp_delete_sharedkey(stcb, scdel->scact_keynumber); SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_AUTH_DEACTIVATE_KEY: { struct sctp_authkeyid *keyid; SCTP_CHECK_AND_CAST(keyid, optval, struct sctp_authkeyid, optsize); SCTP_FIND_STCB(inp, stcb, keyid->scact_assoc_id); /* deactivate the key from the right place */ if (stcb) { if (sctp_deact_sharedkey(stcb, keyid->scact_keynumber)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (keyid->scact_assoc_id == SCTP_FUTURE_ASSOC) || (keyid->scact_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); if (sctp_deact_sharedkey_ep(inp, keyid->scact_keynumber)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_INP_WUNLOCK(inp); } if ((keyid->scact_assoc_id == SCTP_CURRENT_ASSOC) || (keyid->scact_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); sctp_deact_sharedkey(stcb, keyid->scact_keynumber); SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_ENABLE_STREAM_RESET: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); if (av->assoc_value & (~SCTP_ENABLE_VALUE_MASK)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { stcb->asoc.local_strreset_support = (uint8_t) av->assoc_value; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); inp->local_strreset_support = (uint8_t) av->assoc_value; SCTP_INP_WUNLOCK(inp); } if ((av->assoc_id == SCTP_CURRENT_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); stcb->asoc.local_strreset_support = (uint8_t) av->assoc_value; SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_RESET_STREAMS: { struct sctp_reset_streams *strrst; int i, send_out = 0; int send_in = 0; SCTP_CHECK_AND_CAST(strrst, optval, struct sctp_reset_streams, optsize); SCTP_FIND_STCB(inp, stcb, strrst->srs_assoc_id); if (stcb == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; break; } if (stcb->asoc.reconfig_supported == 0) { /* * Peer does not support the chunk type. */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; SCTP_TCB_UNLOCK(stcb); break; } if (sizeof(struct sctp_reset_streams) + strrst->srs_number_streams * sizeof(uint16_t) > optsize) { error = EINVAL; SCTP_TCB_UNLOCK(stcb); break; } if (strrst->srs_flags & SCTP_STREAM_RESET_INCOMING) { send_in = 1; if (stcb->asoc.stream_reset_outstanding) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); error = EALREADY; SCTP_TCB_UNLOCK(stcb); break; } } if (strrst->srs_flags & SCTP_STREAM_RESET_OUTGOING) { send_out = 1; } if ((strrst->srs_number_streams > SCTP_MAX_STREAMS_AT_ONCE_RESET) && send_in) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM); error = ENOMEM; SCTP_TCB_UNLOCK(stcb); break; } if ((send_in == 0) && (send_out == 0)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_TCB_UNLOCK(stcb); break; } for (i = 0; i < strrst->srs_number_streams; i++) { if ((send_in) && (strrst->srs_stream_list[i] > stcb->asoc.streamincnt)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } if ((send_out) && (strrst->srs_stream_list[i] > stcb->asoc.streamoutcnt)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } } if (error) { SCTP_TCB_UNLOCK(stcb); break; } if (send_out) { int cnt; uint16_t strm; if (strrst->srs_number_streams) { for (i = 0, cnt = 0; i < strrst->srs_number_streams; i++) { strm = strrst->srs_stream_list[i]; if (stcb->asoc.strmout[strm].state == SCTP_STREAM_OPEN) { stcb->asoc.strmout[strm].state = SCTP_STREAM_RESET_PENDING; cnt++; } } } else { /* Its all */ for (i = 0, cnt = 0; i < stcb->asoc.streamoutcnt; i++) { if (stcb->asoc.strmout[i].state == SCTP_STREAM_OPEN) { stcb->asoc.strmout[i].state = SCTP_STREAM_RESET_PENDING; cnt++; } } } } if (send_in) { error = sctp_send_str_reset_req(stcb, strrst->srs_number_streams, strrst->srs_stream_list, send_in, 0, 0, 0, 0, 0); } else { error = sctp_send_stream_reset_out_if_possible(stcb, SCTP_SO_LOCKED); } if (error == 0) { sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_REQ, SCTP_SO_LOCKED); } else { /* * For outgoing streams don't report any * problems in sending the request to the * application. XXX: Double check resetting * incoming streams. */ error = 0; } SCTP_TCB_UNLOCK(stcb); break; } case SCTP_ADD_STREAMS: { struct sctp_add_streams *stradd; uint8_t addstream = 0; uint16_t add_o_strmcnt = 0; uint16_t add_i_strmcnt = 0; SCTP_CHECK_AND_CAST(stradd, optval, struct sctp_add_streams, optsize); SCTP_FIND_STCB(inp, stcb, stradd->sas_assoc_id); if (stcb == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; break; } if (stcb->asoc.reconfig_supported == 0) { /* * Peer does not support the chunk type. */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; SCTP_TCB_UNLOCK(stcb); break; } if (stcb->asoc.stream_reset_outstanding) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); error = EALREADY; SCTP_TCB_UNLOCK(stcb); break; } if ((stradd->sas_outstrms == 0) && (stradd->sas_instrms == 0)) { error = EINVAL; goto skip_stuff; } if (stradd->sas_outstrms) { addstream = 1; /* We allocate here */ add_o_strmcnt = stradd->sas_outstrms; if ((((int)add_o_strmcnt) + ((int)stcb->asoc.streamoutcnt)) > 0x0000ffff) { /* You can't have more than 64k */ error = EINVAL; goto skip_stuff; } } if (stradd->sas_instrms) { int cnt; addstream |= 2; /* * We allocate inside * sctp_send_str_reset_req() */ add_i_strmcnt = stradd->sas_instrms; cnt = add_i_strmcnt; cnt += stcb->asoc.streamincnt; if (cnt > 0x0000ffff) { /* You can't have more than 64k */ error = EINVAL; goto skip_stuff; } if (cnt > (int)stcb->asoc.max_inbound_streams) { /* More than you are allowed */ error = EINVAL; goto skip_stuff; } } error = sctp_send_str_reset_req(stcb, 0, NULL, 0, 0, addstream, add_o_strmcnt, add_i_strmcnt, 0); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_REQ, SCTP_SO_LOCKED); skip_stuff: SCTP_TCB_UNLOCK(stcb); break; } case SCTP_RESET_ASSOC: { int i; uint32_t *value; SCTP_CHECK_AND_CAST(value, optval, uint32_t, optsize); SCTP_FIND_STCB(inp, stcb, (sctp_assoc_t) * value); if (stcb == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; break; } if (stcb->asoc.reconfig_supported == 0) { /* * Peer does not support the chunk type. */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); error = EOPNOTSUPP; SCTP_TCB_UNLOCK(stcb); break; } if (stcb->asoc.stream_reset_outstanding) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); error = EALREADY; SCTP_TCB_UNLOCK(stcb); break; } /* * Is there any data pending in the send or sent * queues? */ if (!TAILQ_EMPTY(&stcb->asoc.send_queue) || !TAILQ_EMPTY(&stcb->asoc.sent_queue)) { busy_out: error = EBUSY; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); SCTP_TCB_UNLOCK(stcb); break; } /* Do any streams have data queued? */ for (i = 0; i < stcb->asoc.streamoutcnt; i++) { if (!TAILQ_EMPTY(&stcb->asoc.strmout[i].outqueue)) { goto busy_out; } } error = sctp_send_str_reset_req(stcb, 0, NULL, 0, 1, 0, 0, 0, 0); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_REQ, SCTP_SO_LOCKED); SCTP_TCB_UNLOCK(stcb); break; } case SCTP_CONNECT_X: if (optsize < (sizeof(int) + sizeof(struct sockaddr_in))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } error = sctp_do_connect_x(so, inp, optval, optsize, p, 0); break; case SCTP_CONNECT_X_DELAYED: if (optsize < (sizeof(int) + sizeof(struct sockaddr_in))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } error = sctp_do_connect_x(so, inp, optval, optsize, p, 1); break; case SCTP_CONNECT_X_COMPLETE: { struct sockaddr *sa; /* FIXME MT: check correct? */ SCTP_CHECK_AND_CAST(sa, optval, struct sockaddr, optsize); /* find tcb */ if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { SCTP_INP_RLOCK(inp); stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb) { SCTP_TCB_LOCK(stcb); } SCTP_INP_RUNLOCK(inp); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, sa, NULL, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if (stcb == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); error = ENOENT; break; } if (stcb->asoc.delayed_connection == 1) { stcb->asoc.delayed_connection = 0; (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); sctp_timer_stop(SCTP_TIMER_TYPE_INIT, inp, stcb, stcb->asoc.primary_destination, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_8); sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); } else { /* * already expired or did not use delayed * connectx */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); error = EALREADY; } SCTP_TCB_UNLOCK(stcb); break; } case SCTP_MAX_BURST: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { stcb->asoc.max_burst = av->assoc_value; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); inp->sctp_ep.max_burst = av->assoc_value; SCTP_INP_WUNLOCK(inp); } if ((av->assoc_id == SCTP_CURRENT_ASSOC) || (av->assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); stcb->asoc.max_burst = av->assoc_value; SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_MAXSEG: { struct sctp_assoc_value *av; int ovh; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { ovh = SCTP_MED_OVERHEAD; } else { ovh = SCTP_MED_V4_OVERHEAD; } if (stcb) { if (av->assoc_value) { stcb->asoc.sctp_frag_point = (av->assoc_value + ovh); } else { stcb->asoc.sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); /* * FIXME MT: I think this is not in * tune with the API ID */ if (av->assoc_value) { inp->sctp_frag_point = (av->assoc_value + ovh); } else { inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT; } SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_EVENTS: { struct sctp_event_subscribe *events; SCTP_CHECK_AND_CAST(events, optval, struct sctp_event_subscribe, optsize); SCTP_INP_WLOCK(inp); if (events->sctp_data_io_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT); } if (events->sctp_association_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT); } if (events->sctp_address_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVPADDREVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVPADDREVNT); } if (events->sctp_send_failure_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT); } if (events->sctp_peer_error_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVPEERERR); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVPEERERR); } if (events->sctp_shutdown_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT); } if (events->sctp_partial_delivery_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_PDAPIEVNT); } if (events->sctp_adaptation_layer_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT); } if (events->sctp_authentication_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTHEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTHEVNT); } if (events->sctp_sender_dry_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_DRYEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_DRYEVNT); } if (events->sctp_stream_reset_event) { sctp_feature_on(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT); } SCTP_INP_WUNLOCK(inp); SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); if (events->sctp_association_event) { sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_RECVASSOCEVNT); } else { sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_RECVASSOCEVNT); } if (events->sctp_address_event) { sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_RECVPADDREVNT); } else { sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_RECVPADDREVNT); } if (events->sctp_send_failure_event) { sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_RECVSENDFAILEVNT); } else { sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_RECVSENDFAILEVNT); } if (events->sctp_peer_error_event) { sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_RECVPEERERR); } else { sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_RECVPEERERR); } if (events->sctp_shutdown_event) { sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT); } else { sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT); } if (events->sctp_partial_delivery_event) { sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT); } else { sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT); } if (events->sctp_adaptation_layer_event) { sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_ADAPTATIONEVNT); } else { sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_ADAPTATIONEVNT); } if (events->sctp_authentication_event) { sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_AUTHEVNT); } else { sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_AUTHEVNT); } if (events->sctp_sender_dry_event) { sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_DRYEVNT); } else { sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_DRYEVNT); } if (events->sctp_stream_reset_event) { sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_STREAM_RESETEVNT); } else { sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_STREAM_RESETEVNT); } SCTP_TCB_UNLOCK(stcb); } /* * Send up the sender dry event only for 1-to-1 * style sockets. */ if (events->sctp_sender_dry_event) { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb) { SCTP_TCB_LOCK(stcb); if (TAILQ_EMPTY(&stcb->asoc.send_queue) && TAILQ_EMPTY(&stcb->asoc.sent_queue) && (stcb->asoc.stream_queue_cnt == 0)) { sctp_ulp_notify(SCTP_NOTIFY_SENDER_DRY, stcb, 0, NULL, SCTP_SO_LOCKED); } SCTP_TCB_UNLOCK(stcb); } } } SCTP_INP_RUNLOCK(inp); break; } case SCTP_ADAPTATION_LAYER: { struct sctp_setadaptation *adap_bits; SCTP_CHECK_AND_CAST(adap_bits, optval, struct sctp_setadaptation, optsize); SCTP_INP_WLOCK(inp); inp->sctp_ep.adaptation_layer_indicator = adap_bits->ssb_adaptation_ind; inp->sctp_ep.adaptation_layer_indicator_provided = 1; SCTP_INP_WUNLOCK(inp); break; } #ifdef SCTP_DEBUG case SCTP_SET_INITIAL_DBG_SEQ: { uint32_t *vvv; SCTP_CHECK_AND_CAST(vvv, optval, uint32_t, optsize); SCTP_INP_WLOCK(inp); inp->sctp_ep.initial_sequence_debug = *vvv; SCTP_INP_WUNLOCK(inp); break; } #endif case SCTP_DEFAULT_SEND_PARAM: { struct sctp_sndrcvinfo *s_info; SCTP_CHECK_AND_CAST(s_info, optval, struct sctp_sndrcvinfo, optsize); SCTP_FIND_STCB(inp, stcb, s_info->sinfo_assoc_id); if (stcb) { if (s_info->sinfo_stream < stcb->asoc.streamoutcnt) { memcpy(&stcb->asoc.def_send, s_info, min(optsize, sizeof(stcb->asoc.def_send))); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (s_info->sinfo_assoc_id == SCTP_FUTURE_ASSOC) || (s_info->sinfo_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); memcpy(&inp->def_send, s_info, min(optsize, sizeof(inp->def_send))); SCTP_INP_WUNLOCK(inp); } if ((s_info->sinfo_assoc_id == SCTP_CURRENT_ASSOC) || (s_info->sinfo_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); if (s_info->sinfo_stream < stcb->asoc.streamoutcnt) { memcpy(&stcb->asoc.def_send, s_info, min(optsize, sizeof(stcb->asoc.def_send))); } SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_PEER_ADDR_PARAMS: { struct sctp_paddrparams *paddrp; struct sctp_nets *net; struct sockaddr *addr; #if defined(INET) && defined(INET6) struct sockaddr_in sin_store; #endif SCTP_CHECK_AND_CAST(paddrp, optval, struct sctp_paddrparams, optsize); SCTP_FIND_STCB(inp, stcb, paddrp->spp_assoc_id); #if defined(INET) && defined(INET6) if (paddrp->spp_address.ss_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&paddrp->spp_address; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { in6_sin6_2_sin(&sin_store, sin6); addr = (struct sockaddr *)&sin_store; } else { addr = (struct sockaddr *)&paddrp->spp_address; } } else { addr = (struct sockaddr *)&paddrp->spp_address; } #else addr = (struct sockaddr *)&paddrp->spp_address; #endif if (stcb != NULL) { net = sctp_findnet(stcb, addr); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ net = NULL; SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if ((stcb != NULL) && (net == NULL)) { #ifdef INET if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; if (sin->sin_addr.s_addr != INADDR_ANY) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); SCTP_TCB_UNLOCK(stcb); error = EINVAL; break; } } else #endif #ifdef INET6 if (addr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); SCTP_TCB_UNLOCK(stcb); error = EINVAL; break; } } else #endif { error = EAFNOSUPPORT; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } /* sanity checks */ if ((paddrp->spp_flags & SPP_HB_ENABLE) && (paddrp->spp_flags & SPP_HB_DISABLE)) { if (stcb) SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } if ((paddrp->spp_flags & SPP_PMTUD_ENABLE) && (paddrp->spp_flags & SPP_PMTUD_DISABLE)) { if (stcb) SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } if (stcb != NULL) { /************************TCB SPECIFIC SET ******************/ if (net != NULL) { /************************NET SPECIFIC SET ******************/ if (paddrp->spp_flags & SPP_HB_DISABLE) { if (!(net->dest_state & SCTP_ADDR_UNCONFIRMED) && !(net->dest_state & SCTP_ADDR_NOHB)) { sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_9); } net->dest_state |= SCTP_ADDR_NOHB; } if (paddrp->spp_flags & SPP_HB_ENABLE) { if (paddrp->spp_hbinterval) { net->heart_beat_delay = paddrp->spp_hbinterval; } else if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) { net->heart_beat_delay = 0; } sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_10); sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net); net->dest_state &= ~SCTP_ADDR_NOHB; } if (paddrp->spp_flags & SPP_HB_DEMAND) { /* on demand HB */ sctp_send_hb(stcb, net, SCTP_SO_LOCKED); sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_SOCKOPT, SCTP_SO_LOCKED); sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net); } if ((paddrp->spp_flags & SPP_PMTUD_DISABLE) && (paddrp->spp_pathmtu >= SCTP_SMALLEST_PMTU)) { if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_11); } net->dest_state |= SCTP_ADDR_NO_PMTUD; net->mtu = paddrp->spp_pathmtu; switch (net->ro._l_addr.sa.sa_family) { #ifdef INET case AF_INET: net->mtu += SCTP_MIN_V4_OVERHEAD; break; #endif #ifdef INET6 case AF_INET6: net->mtu += SCTP_MIN_OVERHEAD; break; #endif default: break; } if (net->mtu < stcb->asoc.smallest_mtu) { sctp_pathmtu_adjustment(stcb, net->mtu); } } if (paddrp->spp_flags & SPP_PMTUD_ENABLE) { if (!SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); } net->dest_state &= ~SCTP_ADDR_NO_PMTUD; } if (paddrp->spp_pathmaxrxt) { if (net->dest_state & SCTP_ADDR_PF) { if (net->error_count > paddrp->spp_pathmaxrxt) { net->dest_state &= ~SCTP_ADDR_PF; } } else { if ((net->error_count <= paddrp->spp_pathmaxrxt) && (net->error_count > net->pf_threshold)) { net->dest_state |= SCTP_ADDR_PF; sctp_send_hb(stcb, net, SCTP_SO_LOCKED); sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_12); sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); } } if (net->dest_state & SCTP_ADDR_REACHABLE) { if (net->error_count > paddrp->spp_pathmaxrxt) { net->dest_state &= ~SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, 0, net, SCTP_SO_LOCKED); } } else { if (net->error_count <= paddrp->spp_pathmaxrxt) { net->dest_state |= SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, 0, net, SCTP_SO_LOCKED); } } net->failure_threshold = paddrp->spp_pathmaxrxt; } if (paddrp->spp_flags & SPP_DSCP) { net->dscp = paddrp->spp_dscp & 0xfc; net->dscp |= 0x01; } #ifdef INET6 if (paddrp->spp_flags & SPP_IPV6_FLOWLABEL) { if (net->ro._l_addr.sa.sa_family == AF_INET6) { net->flowlabel = paddrp->spp_ipv6_flowlabel & 0x000fffff; net->flowlabel |= 0x80000000; } } #endif } else { /************************ASSOC ONLY -- NO NET SPECIFIC SET ******************/ if (paddrp->spp_pathmaxrxt != 0) { stcb->asoc.def_net_failure = paddrp->spp_pathmaxrxt; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (net->dest_state & SCTP_ADDR_PF) { if (net->error_count > paddrp->spp_pathmaxrxt) { net->dest_state &= ~SCTP_ADDR_PF; } } else { if ((net->error_count <= paddrp->spp_pathmaxrxt) && (net->error_count > net->pf_threshold)) { net->dest_state |= SCTP_ADDR_PF; sctp_send_hb(stcb, net, SCTP_SO_LOCKED); sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_13); sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); } } if (net->dest_state & SCTP_ADDR_REACHABLE) { if (net->error_count > paddrp->spp_pathmaxrxt) { net->dest_state &= ~SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, 0, net, SCTP_SO_LOCKED); } } else { if (net->error_count <= paddrp->spp_pathmaxrxt) { net->dest_state |= SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, 0, net, SCTP_SO_LOCKED); } } net->failure_threshold = paddrp->spp_pathmaxrxt; } } if (paddrp->spp_flags & SPP_HB_ENABLE) { if (paddrp->spp_hbinterval != 0) { stcb->asoc.heart_beat_delay = paddrp->spp_hbinterval; } else if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) { stcb->asoc.heart_beat_delay = 0; } /* Turn back on the timer */ TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (paddrp->spp_hbinterval != 0) { net->heart_beat_delay = paddrp->spp_hbinterval; } else if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) { net->heart_beat_delay = 0; } if (net->dest_state & SCTP_ADDR_NOHB) { net->dest_state &= ~SCTP_ADDR_NOHB; } sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_14); sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net); } sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT); } if (paddrp->spp_flags & SPP_HB_DISABLE) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (!(net->dest_state & SCTP_ADDR_NOHB)) { net->dest_state |= SCTP_ADDR_NOHB; if (!(net->dest_state & SCTP_ADDR_UNCONFIRMED)) { sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_15); } } } sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT); } if ((paddrp->spp_flags & SPP_PMTUD_DISABLE) && (paddrp->spp_pathmtu >= SCTP_SMALLEST_PMTU)) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_16); } net->dest_state |= SCTP_ADDR_NO_PMTUD; net->mtu = paddrp->spp_pathmtu; switch (net->ro._l_addr.sa.sa_family) { #ifdef INET case AF_INET: net->mtu += SCTP_MIN_V4_OVERHEAD; break; #endif #ifdef INET6 case AF_INET6: net->mtu += SCTP_MIN_OVERHEAD; break; #endif default: break; } if (net->mtu < stcb->asoc.smallest_mtu) { sctp_pathmtu_adjustment(stcb, net->mtu); } } sctp_stcb_feature_on(inp, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD); } if (paddrp->spp_flags & SPP_PMTUD_ENABLE) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (!SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) { sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net); } net->dest_state &= ~SCTP_ADDR_NO_PMTUD; } sctp_stcb_feature_off(inp, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD); } if (paddrp->spp_flags & SPP_DSCP) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { net->dscp = paddrp->spp_dscp & 0xfc; net->dscp |= 0x01; } stcb->asoc.default_dscp = paddrp->spp_dscp & 0xfc; stcb->asoc.default_dscp |= 0x01; } #ifdef INET6 if (paddrp->spp_flags & SPP_IPV6_FLOWLABEL) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if (net->ro._l_addr.sa.sa_family == AF_INET6) { net->flowlabel = paddrp->spp_ipv6_flowlabel & 0x000fffff; net->flowlabel |= 0x80000000; } } stcb->asoc.default_flowlabel = paddrp->spp_ipv6_flowlabel & 0x000fffff; stcb->asoc.default_flowlabel |= 0x80000000; } #endif } SCTP_TCB_UNLOCK(stcb); } else { /************************NO TCB, SET TO default stuff ******************/ if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (paddrp->spp_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); /* * For the TOS/FLOWLABEL stuff you * set it with the options on the * socket */ if (paddrp->spp_pathmaxrxt != 0) { inp->sctp_ep.def_net_failure = paddrp->spp_pathmaxrxt; } if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = 0; else if (paddrp->spp_hbinterval != 0) { if (paddrp->spp_hbinterval > SCTP_MAX_HB_INTERVAL) paddrp->spp_hbinterval = SCTP_MAX_HB_INTERVAL; inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(paddrp->spp_hbinterval); } if (paddrp->spp_flags & SPP_HB_ENABLE) { if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO) { inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = 0; } else if (paddrp->spp_hbinterval) { inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(paddrp->spp_hbinterval); } sctp_feature_off(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT); } else if (paddrp->spp_flags & SPP_HB_DISABLE) { sctp_feature_on(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT); } if (paddrp->spp_flags & SPP_PMTUD_ENABLE) { sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_NOT_PMTUD); } else if (paddrp->spp_flags & SPP_PMTUD_DISABLE) { sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_NOT_PMTUD); } if (paddrp->spp_flags & SPP_DSCP) { inp->sctp_ep.default_dscp = paddrp->spp_dscp & 0xfc; inp->sctp_ep.default_dscp |= 0x01; } #ifdef INET6 if (paddrp->spp_flags & SPP_IPV6_FLOWLABEL) { if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { inp->sctp_ep.default_flowlabel = paddrp->spp_ipv6_flowlabel & 0x000fffff; inp->sctp_ep.default_flowlabel |= 0x80000000; } } #endif SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_RTOINFO: { struct sctp_rtoinfo *srto; uint32_t new_init, new_min, new_max; SCTP_CHECK_AND_CAST(srto, optval, struct sctp_rtoinfo, optsize); SCTP_FIND_STCB(inp, stcb, srto->srto_assoc_id); if (stcb) { if (srto->srto_initial) new_init = srto->srto_initial; else new_init = stcb->asoc.initial_rto; if (srto->srto_max) new_max = srto->srto_max; else new_max = stcb->asoc.maxrto; if (srto->srto_min) new_min = srto->srto_min; else new_min = stcb->asoc.minrto; if ((new_min <= new_init) && (new_init <= new_max)) { stcb->asoc.initial_rto = new_init; stcb->asoc.maxrto = new_max; stcb->asoc.minrto = new_min; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (srto->srto_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); if (srto->srto_initial) new_init = srto->srto_initial; else new_init = inp->sctp_ep.initial_rto; if (srto->srto_max) new_max = srto->srto_max; else new_max = inp->sctp_ep.sctp_maxrto; if (srto->srto_min) new_min = srto->srto_min; else new_min = inp->sctp_ep.sctp_minrto; if ((new_min <= new_init) && (new_init <= new_max)) { inp->sctp_ep.initial_rto = new_init; inp->sctp_ep.sctp_maxrto = new_max; inp->sctp_ep.sctp_minrto = new_min; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_ASSOCINFO: { struct sctp_assocparams *sasoc; SCTP_CHECK_AND_CAST(sasoc, optval, struct sctp_assocparams, optsize); SCTP_FIND_STCB(inp, stcb, sasoc->sasoc_assoc_id); if (sasoc->sasoc_cookie_life) { /* boundary check the cookie life */ if (sasoc->sasoc_cookie_life < 1000) sasoc->sasoc_cookie_life = 1000; if (sasoc->sasoc_cookie_life > SCTP_MAX_COOKIE_LIFE) { sasoc->sasoc_cookie_life = SCTP_MAX_COOKIE_LIFE; } } if (stcb) { if (sasoc->sasoc_asocmaxrxt) stcb->asoc.max_send_times = sasoc->sasoc_asocmaxrxt; if (sasoc->sasoc_cookie_life) { stcb->asoc.cookie_life = MSEC_TO_TICKS(sasoc->sasoc_cookie_life); } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (sasoc->sasoc_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); if (sasoc->sasoc_asocmaxrxt) inp->sctp_ep.max_send_times = sasoc->sasoc_asocmaxrxt; if (sasoc->sasoc_cookie_life) { inp->sctp_ep.def_cookie_life = MSEC_TO_TICKS(sasoc->sasoc_cookie_life); } SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_INITMSG: { struct sctp_initmsg *sinit; SCTP_CHECK_AND_CAST(sinit, optval, struct sctp_initmsg, optsize); SCTP_INP_WLOCK(inp); if (sinit->sinit_num_ostreams) inp->sctp_ep.pre_open_stream_count = sinit->sinit_num_ostreams; if (sinit->sinit_max_instreams) inp->sctp_ep.max_open_streams_intome = sinit->sinit_max_instreams; if (sinit->sinit_max_attempts) inp->sctp_ep.max_init_times = sinit->sinit_max_attempts; if (sinit->sinit_max_init_timeo) inp->sctp_ep.initial_init_rto_max = sinit->sinit_max_init_timeo; SCTP_INP_WUNLOCK(inp); break; } case SCTP_PRIMARY_ADDR: { struct sctp_setprim *spa; struct sctp_nets *net; struct sockaddr *addr; #if defined(INET) && defined(INET6) struct sockaddr_in sin_store; #endif SCTP_CHECK_AND_CAST(spa, optval, struct sctp_setprim, optsize); SCTP_FIND_STCB(inp, stcb, spa->ssp_assoc_id); #if defined(INET) && defined(INET6) if (spa->ssp_addr.ss_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&spa->ssp_addr; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { in6_sin6_2_sin(&sin_store, sin6); addr = (struct sockaddr *)&sin_store; } else { addr = (struct sockaddr *)&spa->ssp_addr; } } else { addr = (struct sockaddr *)&spa->ssp_addr; } #else addr = (struct sockaddr *)&spa->ssp_addr; #endif if (stcb != NULL) { net = sctp_findnet(stcb, addr); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ net = NULL; SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if ((stcb != NULL) && (net != NULL)) { if (net != stcb->asoc.primary_destination) { if (!(net->dest_state & SCTP_ADDR_UNCONFIRMED)) { /* Ok we need to set it */ if (sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, net) == 0) { if ((stcb->asoc.alternate) && (!(net->dest_state & SCTP_ADDR_PF)) && (net->dest_state & SCTP_ADDR_REACHABLE)) { sctp_free_remote_addr(stcb->asoc.alternate); stcb->asoc.alternate = NULL; } } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } if (stcb != NULL) { SCTP_TCB_UNLOCK(stcb); } break; } case SCTP_SET_DYNAMIC_PRIMARY: { union sctp_sockstore *ss; error = priv_check(curthread, PRIV_NETINET_RESERVEDPORT); if (error) break; SCTP_CHECK_AND_CAST(ss, optval, union sctp_sockstore, optsize); /* SUPER USER CHECK? */ error = sctp_dynamic_set_primary(&ss->sa, vrf_id); break; } case SCTP_SET_PEER_PRIMARY_ADDR: { struct sctp_setpeerprim *sspp; struct sockaddr *addr; #if defined(INET) && defined(INET6) struct sockaddr_in sin_store; #endif SCTP_CHECK_AND_CAST(sspp, optval, struct sctp_setpeerprim, optsize); SCTP_FIND_STCB(inp, stcb, sspp->sspp_assoc_id); if (stcb != NULL) { struct sctp_ifa *ifa; #if defined(INET) && defined(INET6) if (sspp->sspp_addr.ss_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&sspp->sspp_addr; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { in6_sin6_2_sin(&sin_store, sin6); addr = (struct sockaddr *)&sin_store; } else { addr = (struct sockaddr *)&sspp->sspp_addr; } } else { addr = (struct sockaddr *)&sspp->sspp_addr; } #else addr = (struct sockaddr *)&sspp->sspp_addr; #endif ifa = sctp_find_ifa_by_addr(addr, stcb->asoc.vrf_id, SCTP_ADDR_NOT_LOCKED); if (ifa == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_of_it; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { /* * Must validate the ifa found is in * our ep */ struct sctp_laddr *laddr; int found = 0; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa == NULL) { SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: NULL ifa\n", __func__); continue; } if (laddr->ifa == ifa) { found = 1; break; } } if (!found) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_of_it; } } else { switch (addr->sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; if (prison_check_ip4(inp->ip_inp.inp.inp_cred, &sin->sin_addr) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_of_it; } break; } #endif #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (prison_check_ip6(inp->ip_inp.inp.inp_cred, &sin6->sin6_addr) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_of_it; } break; } #endif default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_of_it; } } if (sctp_set_primary_ip_address_sa(stcb, addr) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } out_of_it: SCTP_TCB_UNLOCK(stcb); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } break; } case SCTP_BINDX_ADD_ADDR: { struct sctp_getaddresses *addrs; struct thread *td; td = (struct thread *)p; SCTP_CHECK_AND_CAST(addrs, optval, struct sctp_getaddresses, optsize); #ifdef INET if (addrs->addr->sa_family == AF_INET) { if (optsize < sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } if (td != NULL && (error = prison_local_ip4(td->td_ucred, &(((struct sockaddr_in *)(addrs->addr))->sin_addr)))) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else #endif #ifdef INET6 if (addrs->addr->sa_family == AF_INET6) { if (optsize < sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } if (td != NULL && (error = prison_local_ip6(td->td_ucred, &(((struct sockaddr_in6 *)(addrs->addr))->sin6_addr), (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else #endif { error = EAFNOSUPPORT; break; } sctp_bindx_add_address(so, inp, addrs->addr, addrs->sget_assoc_id, vrf_id, &error, p); break; } case SCTP_BINDX_REM_ADDR: { struct sctp_getaddresses *addrs; struct thread *td; td = (struct thread *)p; SCTP_CHECK_AND_CAST(addrs, optval, struct sctp_getaddresses, optsize); #ifdef INET if (addrs->addr->sa_family == AF_INET) { if (optsize < sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } if (td != NULL && (error = prison_local_ip4(td->td_ucred, &(((struct sockaddr_in *)(addrs->addr))->sin_addr)))) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else #endif #ifdef INET6 if (addrs->addr->sa_family == AF_INET6) { if (optsize < sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } if (td != NULL && (error = prison_local_ip6(td->td_ucred, &(((struct sockaddr_in6 *)(addrs->addr))->sin6_addr), (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) { SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } else #endif { error = EAFNOSUPPORT; break; } sctp_bindx_delete_address(inp, addrs->addr, addrs->sget_assoc_id, vrf_id, &error); break; } case SCTP_EVENT: { struct sctp_event *event; uint32_t event_type; SCTP_CHECK_AND_CAST(event, optval, struct sctp_event, optsize); SCTP_FIND_STCB(inp, stcb, event->se_assoc_id); switch (event->se_type) { case SCTP_ASSOC_CHANGE: event_type = SCTP_PCB_FLAGS_RECVASSOCEVNT; break; case SCTP_PEER_ADDR_CHANGE: event_type = SCTP_PCB_FLAGS_RECVPADDREVNT; break; case SCTP_REMOTE_ERROR: event_type = SCTP_PCB_FLAGS_RECVPEERERR; break; case SCTP_SEND_FAILED: event_type = SCTP_PCB_FLAGS_RECVSENDFAILEVNT; break; case SCTP_SHUTDOWN_EVENT: event_type = SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT; break; case SCTP_ADAPTATION_INDICATION: event_type = SCTP_PCB_FLAGS_ADAPTATIONEVNT; break; case SCTP_PARTIAL_DELIVERY_EVENT: event_type = SCTP_PCB_FLAGS_PDAPIEVNT; break; case SCTP_AUTHENTICATION_EVENT: event_type = SCTP_PCB_FLAGS_AUTHEVNT; break; case SCTP_STREAM_RESET_EVENT: event_type = SCTP_PCB_FLAGS_STREAM_RESETEVNT; break; case SCTP_SENDER_DRY_EVENT: event_type = SCTP_PCB_FLAGS_DRYEVNT; break; case SCTP_NOTIFICATIONS_STOPPED_EVENT: event_type = 0; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTSUP); error = ENOTSUP; break; case SCTP_ASSOC_RESET_EVENT: event_type = SCTP_PCB_FLAGS_ASSOC_RESETEVNT; break; case SCTP_STREAM_CHANGE_EVENT: event_type = SCTP_PCB_FLAGS_STREAM_CHANGEEVNT; break; case SCTP_SEND_FAILED_EVENT: event_type = SCTP_PCB_FLAGS_RECVNSENDFAILEVNT; break; default: event_type = 0; SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } if (event_type > 0) { if (stcb) { if (event->se_on) { sctp_stcb_feature_on(inp, stcb, event_type); if (event_type == SCTP_PCB_FLAGS_DRYEVNT) { if (TAILQ_EMPTY(&stcb->asoc.send_queue) && TAILQ_EMPTY(&stcb->asoc.sent_queue) && (stcb->asoc.stream_queue_cnt == 0)) { sctp_ulp_notify(SCTP_NOTIFY_SENDER_DRY, stcb, 0, NULL, SCTP_SO_LOCKED); } } } else { sctp_stcb_feature_off(inp, stcb, event_type); } SCTP_TCB_UNLOCK(stcb); } else { /* * We don't want to send up a storm * of events, so return an error for * sender dry events */ if ((event_type == SCTP_PCB_FLAGS_DRYEVNT) && ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) == 0) && ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) && ((event->se_assoc_id == SCTP_ALL_ASSOC) || (event->se_assoc_id == SCTP_CURRENT_ASSOC))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTSUP); error = ENOTSUP; break; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (event->se_assoc_id == SCTP_FUTURE_ASSOC) || (event->se_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); if (event->se_on) { sctp_feature_on(inp, event_type); } else { sctp_feature_off(inp, event_type); } SCTP_INP_WUNLOCK(inp); } if ((event->se_assoc_id == SCTP_CURRENT_ASSOC) || (event->se_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); if (event->se_on) { sctp_stcb_feature_on(inp, stcb, event_type); } else { sctp_stcb_feature_off(inp, stcb, event_type); } SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } } break; } case SCTP_RECVRCVINFO: { int *onoff; SCTP_CHECK_AND_CAST(onoff, optval, int, optsize); SCTP_INP_WLOCK(inp); if (*onoff != 0) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVRCVINFO); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVRCVINFO); } SCTP_INP_WUNLOCK(inp); break; } case SCTP_RECVNXTINFO: { int *onoff; SCTP_CHECK_AND_CAST(onoff, optval, int, optsize); SCTP_INP_WLOCK(inp); if (*onoff != 0) { sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVNXTINFO); } else { sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVNXTINFO); } SCTP_INP_WUNLOCK(inp); break; } case SCTP_DEFAULT_SNDINFO: { struct sctp_sndinfo *info; uint16_t policy; SCTP_CHECK_AND_CAST(info, optval, struct sctp_sndinfo, optsize); SCTP_FIND_STCB(inp, stcb, info->snd_assoc_id); if (stcb) { if (info->snd_sid < stcb->asoc.streamoutcnt) { stcb->asoc.def_send.sinfo_stream = info->snd_sid; policy = PR_SCTP_POLICY(stcb->asoc.def_send.sinfo_flags); stcb->asoc.def_send.sinfo_flags = info->snd_flags; stcb->asoc.def_send.sinfo_flags |= policy; stcb->asoc.def_send.sinfo_ppid = info->snd_ppid; stcb->asoc.def_send.sinfo_context = info->snd_context; } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (info->snd_assoc_id == SCTP_FUTURE_ASSOC) || (info->snd_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); inp->def_send.sinfo_stream = info->snd_sid; policy = PR_SCTP_POLICY(inp->def_send.sinfo_flags); inp->def_send.sinfo_flags = info->snd_flags; inp->def_send.sinfo_flags |= policy; inp->def_send.sinfo_ppid = info->snd_ppid; inp->def_send.sinfo_context = info->snd_context; SCTP_INP_WUNLOCK(inp); } if ((info->snd_assoc_id == SCTP_CURRENT_ASSOC) || (info->snd_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); if (info->snd_sid < stcb->asoc.streamoutcnt) { stcb->asoc.def_send.sinfo_stream = info->snd_sid; policy = PR_SCTP_POLICY(stcb->asoc.def_send.sinfo_flags); stcb->asoc.def_send.sinfo_flags = info->snd_flags; stcb->asoc.def_send.sinfo_flags |= policy; stcb->asoc.def_send.sinfo_ppid = info->snd_ppid; stcb->asoc.def_send.sinfo_context = info->snd_context; } SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_DEFAULT_PRINFO: { struct sctp_default_prinfo *info; SCTP_CHECK_AND_CAST(info, optval, struct sctp_default_prinfo, optsize); SCTP_FIND_STCB(inp, stcb, info->pr_assoc_id); if (info->pr_policy > SCTP_PR_SCTP_MAX) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; break; } if (stcb) { stcb->asoc.def_send.sinfo_flags &= 0xfff0; stcb->asoc.def_send.sinfo_flags |= info->pr_policy; stcb->asoc.def_send.sinfo_timetolive = info->pr_value; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (info->pr_assoc_id == SCTP_FUTURE_ASSOC) || (info->pr_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_WLOCK(inp); inp->def_send.sinfo_flags &= 0xfff0; inp->def_send.sinfo_flags |= info->pr_policy; inp->def_send.sinfo_timetolive = info->pr_value; SCTP_INP_WUNLOCK(inp); } if ((info->pr_assoc_id == SCTP_CURRENT_ASSOC) || (info->pr_assoc_id == SCTP_ALL_ASSOC)) { SCTP_INP_RLOCK(inp); LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) { SCTP_TCB_LOCK(stcb); stcb->asoc.def_send.sinfo_flags &= 0xfff0; stcb->asoc.def_send.sinfo_flags |= info->pr_policy; stcb->asoc.def_send.sinfo_timetolive = info->pr_value; SCTP_TCB_UNLOCK(stcb); } SCTP_INP_RUNLOCK(inp); } } break; } case SCTP_PEER_ADDR_THLDS: /* Applies to the specific association */ { struct sctp_paddrthlds *thlds; struct sctp_nets *net; struct sockaddr *addr; #if defined(INET) && defined(INET6) struct sockaddr_in sin_store; #endif SCTP_CHECK_AND_CAST(thlds, optval, struct sctp_paddrthlds, optsize); SCTP_FIND_STCB(inp, stcb, thlds->spt_assoc_id); #if defined(INET) && defined(INET6) if (thlds->spt_address.ss_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&thlds->spt_address; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { in6_sin6_2_sin(&sin_store, sin6); addr = (struct sockaddr *)&sin_store; } else { addr = (struct sockaddr *)&thlds->spt_address; } } else { addr = (struct sockaddr *)&thlds->spt_address; } #else addr = (struct sockaddr *)&thlds->spt_address; #endif if (stcb != NULL) { net = sctp_findnet(stcb, addr); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ net = NULL; SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if ((stcb != NULL) && (net == NULL)) { #ifdef INET if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; if (sin->sin_addr.s_addr != INADDR_ANY) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); SCTP_TCB_UNLOCK(stcb); error = EINVAL; break; } } else #endif #ifdef INET6 if (addr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); SCTP_TCB_UNLOCK(stcb); error = EINVAL; break; } } else #endif { error = EAFNOSUPPORT; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } if (stcb != NULL) { if (net != NULL) { net->failure_threshold = thlds->spt_pathmaxrxt; net->pf_threshold = thlds->spt_pathpfthld; if (net->dest_state & SCTP_ADDR_PF) { if ((net->error_count > net->failure_threshold) || (net->error_count <= net->pf_threshold)) { net->dest_state &= ~SCTP_ADDR_PF; } } else { if ((net->error_count > net->pf_threshold) && (net->error_count <= net->failure_threshold)) { net->dest_state |= SCTP_ADDR_PF; sctp_send_hb(stcb, net, SCTP_SO_LOCKED); sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_17); sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); } } if (net->dest_state & SCTP_ADDR_REACHABLE) { if (net->error_count > net->failure_threshold) { net->dest_state &= ~SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, 0, net, SCTP_SO_LOCKED); } } else { if (net->error_count <= net->failure_threshold) { net->dest_state |= SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, 0, net, SCTP_SO_LOCKED); } } } else { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { net->failure_threshold = thlds->spt_pathmaxrxt; net->pf_threshold = thlds->spt_pathpfthld; if (net->dest_state & SCTP_ADDR_PF) { if ((net->error_count > net->failure_threshold) || (net->error_count <= net->pf_threshold)) { net->dest_state &= ~SCTP_ADDR_PF; } } else { if ((net->error_count > net->pf_threshold) && (net->error_count <= net->failure_threshold)) { net->dest_state |= SCTP_ADDR_PF; sctp_send_hb(stcb, net, SCTP_SO_LOCKED); sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_18); sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, stcb->sctp_ep, stcb, net); } } if (net->dest_state & SCTP_ADDR_REACHABLE) { if (net->error_count > net->failure_threshold) { net->dest_state &= ~SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN, stcb, 0, net, SCTP_SO_LOCKED); } } else { if (net->error_count <= net->failure_threshold) { net->dest_state |= SCTP_ADDR_REACHABLE; sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb, 0, net, SCTP_SO_LOCKED); } } } stcb->asoc.def_net_failure = thlds->spt_pathmaxrxt; stcb->asoc.def_net_pf_threshold = thlds->spt_pathpfthld; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (thlds->spt_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); inp->sctp_ep.def_net_failure = thlds->spt_pathmaxrxt; inp->sctp_ep.def_net_pf_threshold = thlds->spt_pathpfthld; SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_REMOTE_UDP_ENCAPS_PORT: { struct sctp_udpencaps *encaps; struct sctp_nets *net; struct sockaddr *addr; #if defined(INET) && defined(INET6) struct sockaddr_in sin_store; #endif SCTP_CHECK_AND_CAST(encaps, optval, struct sctp_udpencaps, optsize); SCTP_FIND_STCB(inp, stcb, encaps->sue_assoc_id); #if defined(INET) && defined(INET6) if (encaps->sue_address.ss_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)&encaps->sue_address; if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { in6_sin6_2_sin(&sin_store, sin6); addr = (struct sockaddr *)&sin_store; } else { addr = (struct sockaddr *)&encaps->sue_address; } } else { addr = (struct sockaddr *)&encaps->sue_address; } #else addr = (struct sockaddr *)&encaps->sue_address; #endif if (stcb != NULL) { net = sctp_findnet(stcb, addr); } else { /* * We increment here since * sctp_findassociation_ep_addr() wil do a * decrement if it finds the stcb as long as * the locked tcb (last argument) is NOT a * TCB.. aka NULL. */ net = NULL; SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, addr, &net, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } } if ((stcb != NULL) && (net == NULL)) { #ifdef INET if (addr->sa_family == AF_INET) { struct sockaddr_in *sin; sin = (struct sockaddr_in *)addr; if (sin->sin_addr.s_addr != INADDR_ANY) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); SCTP_TCB_UNLOCK(stcb); error = EINVAL; break; } } else #endif #ifdef INET6 if (addr->sa_family == AF_INET6) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)addr; if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); SCTP_TCB_UNLOCK(stcb); error = EINVAL; break; } } else #endif { error = EAFNOSUPPORT; SCTP_TCB_UNLOCK(stcb); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); break; } } if (stcb != NULL) { if (net != NULL) { net->port = encaps->sue_port; } else { stcb->asoc.port = encaps->sue_port; } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (encaps->sue_assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); inp->sctp_ep.port = encaps->sue_port; SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_ECN_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); if (av->assoc_value == 0) { inp->ecn_supported = 0; } else { inp->ecn_supported = 1; } SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_PR_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); if (av->assoc_value == 0) { inp->prsctp_supported = 0; } else { inp->prsctp_supported = 1; } SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_AUTH_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { if ((av->assoc_value == 0) && (inp->asconf_supported == 1)) { /* * AUTH is required for * ASCONF */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } else { SCTP_INP_WLOCK(inp); if (av->assoc_value == 0) { inp->auth_supported = 0; } else { inp->auth_supported = 1; } SCTP_INP_WUNLOCK(inp); } } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_ASCONF_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { if ((av->assoc_value != 0) && (inp->auth_supported == 0)) { /* * AUTH is required for * ASCONF */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } else { SCTP_INP_WLOCK(inp); if (av->assoc_value == 0) { inp->asconf_supported = 0; sctp_auth_delete_chunk(SCTP_ASCONF, inp->sctp_ep.local_auth_chunks); sctp_auth_delete_chunk(SCTP_ASCONF_ACK, inp->sctp_ep.local_auth_chunks); } else { inp->asconf_supported = 1; sctp_auth_add_chunk(SCTP_ASCONF, inp->sctp_ep.local_auth_chunks); sctp_auth_add_chunk(SCTP_ASCONF_ACK, inp->sctp_ep.local_auth_chunks); } SCTP_INP_WUNLOCK(inp); } } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_RECONFIG_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); if (av->assoc_value == 0) { inp->reconfig_supported = 0; } else { inp->reconfig_supported = 1; } SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_NRSACK_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); if (av->assoc_value == 0) { inp->nrsack_supported = 0; } else { inp->nrsack_supported = 1; } SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_PKTDROP_SUPPORTED: { struct sctp_assoc_value *av; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); if (av->assoc_value == 0) { inp->pktdrop_supported = 0; } else { inp->pktdrop_supported = 1; } SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } case SCTP_MAX_CWND: { struct sctp_assoc_value *av; struct sctp_nets *net; SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize); SCTP_FIND_STCB(inp, stcb, av->assoc_id); if (stcb) { stcb->asoc.max_cwnd = av->assoc_value; if (stcb->asoc.max_cwnd > 0) { TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { if ((net->cwnd > stcb->asoc.max_cwnd) && (net->cwnd > (net->mtu - sizeof(struct sctphdr)))) { net->cwnd = stcb->asoc.max_cwnd; if (net->cwnd < (net->mtu - sizeof(struct sctphdr))) { net->cwnd = net->mtu - sizeof(struct sctphdr); } } } } SCTP_TCB_UNLOCK(stcb); } else { if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) || (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) || (av->assoc_id == SCTP_FUTURE_ASSOC)) { SCTP_INP_WLOCK(inp); inp->max_cwnd = av->assoc_value; SCTP_INP_WUNLOCK(inp); } else { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } } break; } default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT); error = ENOPROTOOPT; break; } /* end switch (opt) */ return (error); } int sctp_ctloutput(struct socket *so, struct sockopt *sopt) { void *optval = NULL; size_t optsize = 0; void *p; int error = 0; struct sctp_inpcb *inp; if ((sopt->sopt_level == SOL_SOCKET) && (sopt->sopt_name == SO_SETFIB)) { inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(so->so_pcb, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOBUFS); return (EINVAL); } SCTP_INP_WLOCK(inp); inp->fibnum = so->so_fibnum; SCTP_INP_WUNLOCK(inp); return (0); } if (sopt->sopt_level != IPPROTO_SCTP) { /* wrong proto level... send back up to IP */ #ifdef INET6 if (INP_CHECK_SOCKAF(so, AF_INET6)) error = ip6_ctloutput(so, sopt); #endif /* INET6 */ #if defined(INET) && defined(INET6) else #endif #ifdef INET error = ip_ctloutput(so, sopt); #endif return (error); } optsize = sopt->sopt_valsize; if (optsize) { SCTP_MALLOC(optval, void *, optsize, SCTP_M_SOCKOPT); if (optval == NULL) { SCTP_LTRACE_ERR_RET(so->so_pcb, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOBUFS); return (ENOBUFS); } error = sooptcopyin(sopt, optval, optsize, optsize); if (error) { SCTP_FREE(optval, SCTP_M_SOCKOPT); goto out; } } p = (void *)sopt->sopt_td; if (sopt->sopt_dir == SOPT_SET) { error = sctp_setopt(so, sopt->sopt_name, optval, optsize, p); } else if (sopt->sopt_dir == SOPT_GET) { error = sctp_getopt(so, sopt->sopt_name, optval, &optsize, p); } else { SCTP_LTRACE_ERR_RET(so->so_pcb, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; } if ((error == 0) && (optval != NULL)) { error = sooptcopyout(sopt, optval, optsize); SCTP_FREE(optval, SCTP_M_SOCKOPT); } else if (optval != NULL) { SCTP_FREE(optval, SCTP_M_SOCKOPT); } out: return (error); } #ifdef INET static int sctp_connect(struct socket *so, struct sockaddr *addr, struct thread *p) { int error = 0; int create_lock_on = 0; uint32_t vrf_id; struct sctp_inpcb *inp; struct sctp_tcb *stcb = NULL; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { /* I made the same as TCP since we are not setup? */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } if (addr == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return EINVAL; } switch (addr->sa_family) { #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6p; if (addr->sa_len != sizeof(struct sockaddr_in6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } sin6p = (struct sockaddr_in6 *)addr; if (p != NULL && (error = prison_remote_ip6(p->td_ucred, &sin6p->sin6_addr)) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); return (error); } break; } #endif #ifdef INET case AF_INET: { struct sockaddr_in *sinp; if (addr->sa_len != sizeof(struct sockaddr_in)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (EINVAL); } sinp = (struct sockaddr_in *)addr; if (p != NULL && (error = prison_remote_ip4(p->td_ucred, &sinp->sin_addr)) != 0) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error); return (error); } break; } #endif default: SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EAFNOSUPPORT); return (EAFNOSUPPORT); } SCTP_INP_INCR_REF(inp); SCTP_ASOC_CREATE_LOCK(inp); create_lock_on = 1; if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) || (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) { /* Should I really unlock ? */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EFAULT); error = EFAULT; goto out_now; } #ifdef INET6 if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) && (addr->sa_family == AF_INET6)) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_now; } #endif if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == SCTP_PCB_FLAGS_UNBOUND) { /* Bind a ephemeral port */ error = sctp_inpcb_bind(so, NULL, NULL, p); if (error) { goto out_now; } } /* Now do we connect? */ if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) && (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE))) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); error = EINVAL; goto out_now; } if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { /* We are already connected AND the TCP model */ SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE); error = EADDRINUSE; goto out_now; } if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { SCTP_INP_RLOCK(inp); stcb = LIST_FIRST(&inp->sctp_asoc_list); SCTP_INP_RUNLOCK(inp); } else { /* * We increment here since sctp_findassociation_ep_addr() * will do a decrement if it finds the stcb as long as the * locked tcb (last argument) is NOT a TCB.. aka NULL. */ SCTP_INP_INCR_REF(inp); stcb = sctp_findassociation_ep_addr(&inp, addr, NULL, NULL, NULL); if (stcb == NULL) { SCTP_INP_DECR_REF(inp); } else { SCTP_TCB_UNLOCK(stcb); } } if (stcb != NULL) { /* Already have or am bring up an association */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY); error = EALREADY; goto out_now; } vrf_id = inp->def_vrf_id; /* We are GOOD to go */ stcb = sctp_aloc_assoc(inp, addr, &error, 0, vrf_id, inp->sctp_ep.pre_open_stream_count, p); if (stcb == NULL) { /* Gak! no memory */ goto out_now; } if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) { stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED; /* Set the connected flag so we can queue data */ soisconnecting(so); } SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_COOKIE_WAIT); (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered); /* initialize authentication parameters for the assoc */ sctp_initialize_auth_params(inp, stcb); sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED); SCTP_TCB_UNLOCK(stcb); out_now: if (create_lock_on) { SCTP_ASOC_CREATE_UNLOCK(inp); } SCTP_INP_DECR_REF(inp); return (error); } #endif int sctp_listen(struct socket *so, int backlog, struct thread *p) { /* * Note this module depends on the protocol processing being called * AFTER any socket level flags and backlog are applied to the * socket. The traditional way that the socket flags are applied is * AFTER protocol processing. We have made a change to the * sys/kern/uipc_socket.c module to reverse this but this MUST be in * place if the socket API for SCTP is to work properly. */ int error = 0; struct sctp_inpcb *inp; inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { /* I made the same as TCP since we are not setup? */ SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) { /* See if we have a listener */ struct sctp_inpcb *tinp; union sctp_sockstore store; if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) { /* not bound all */ struct sctp_laddr *laddr; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { memcpy(&store, &laddr->ifa->address, sizeof(store)); switch (store.sa.sa_family) { #ifdef INET case AF_INET: store.sin.sin_port = inp->sctp_lport; break; #endif #ifdef INET6 case AF_INET6: store.sin6.sin6_port = inp->sctp_lport; break; #endif default: break; } tinp = sctp_pcb_findep(&store.sa, 0, 0, inp->def_vrf_id); if (tinp && (tinp != inp) && ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) == 0) && ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && (tinp->sctp_socket->so_qlimit)) { /* * we have a listener already and * its not this inp. */ SCTP_INP_DECR_REF(tinp); return (EADDRINUSE); } else if (tinp) { SCTP_INP_DECR_REF(tinp); } } } else { /* Setup a local addr bound all */ memset(&store, 0, sizeof(store)); #ifdef INET6 if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) { store.sa.sa_family = AF_INET6; store.sa.sa_len = sizeof(struct sockaddr_in6); } #endif #ifdef INET if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) { store.sa.sa_family = AF_INET; store.sa.sa_len = sizeof(struct sockaddr_in); } #endif switch (store.sa.sa_family) { #ifdef INET case AF_INET: store.sin.sin_port = inp->sctp_lport; break; #endif #ifdef INET6 case AF_INET6: store.sin6.sin6_port = inp->sctp_lport; break; #endif default: break; } tinp = sctp_pcb_findep(&store.sa, 0, 0, inp->def_vrf_id); if (tinp && (tinp != inp) && ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) == 0) && ((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) && (tinp->sctp_socket->so_qlimit)) { /* * we have a listener already and its not * this inp. */ SCTP_INP_DECR_REF(tinp); return (EADDRINUSE); } else if (tinp) { SCTP_INP_DECR_REF(tinp); } } } SCTP_INP_RLOCK(inp); #ifdef SCTP_LOCK_LOGGING if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) { sctp_log_lock(inp, (struct sctp_tcb *)NULL, SCTP_LOG_LOCK_SOCK); } #endif SOCK_LOCK(so); error = solisten_proto_check(so); SOCK_UNLOCK(so); if (error) { SCTP_INP_RUNLOCK(inp); return (error); } if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) && (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { /* * The unlucky case - We are in the tcp pool with this guy. * - Someone else is in the main inp slot. - We must move * this guy (the listener) to the main slot - We must then * move the guy that was listener to the TCP Pool. */ if (sctp_swap_inpcb_for_listen(inp)) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE); return (EADDRINUSE); } } if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) && (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) { /* We are already connected AND the TCP model */ SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE); return (EADDRINUSE); } SCTP_INP_RUNLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) { /* We must do a bind. */ if ((error = sctp_inpcb_bind(so, NULL, NULL, p))) { /* bind error, probably perm */ return (error); } } SOCK_LOCK(so); /* It appears for 7.0 and on, we must always call this. */ solisten_proto(so, backlog); if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { /* remove the ACCEPTCONN flag for one-to-many sockets */ so->so_options &= ~SO_ACCEPTCONN; } if (backlog == 0) { /* turning off listen */ so->so_options &= ~SO_ACCEPTCONN; } SOCK_UNLOCK(so); return (error); } static int sctp_defered_wakeup_cnt = 0; int sctp_accept(struct socket *so, struct sockaddr **addr) { struct sctp_tcb *stcb; struct sctp_inpcb *inp; union sctp_sockstore store; #ifdef INET6 int error; #endif inp = (struct sctp_inpcb *)so->so_pcb; if (inp == NULL) { SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } SCTP_INP_RLOCK(inp); if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP); return (EOPNOTSUPP); } if (so->so_state & SS_ISDISCONNECTED) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ECONNABORTED); return (ECONNABORTED); } stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } SCTP_TCB_LOCK(stcb); SCTP_INP_RUNLOCK(inp); store = stcb->asoc.primary_destination->ro._l_addr; stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE; SCTP_TCB_UNLOCK(stcb); switch (store.sa.sa_family) { #ifdef INET case AF_INET: { struct sockaddr_in *sin; SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin); if (sin == NULL) return (ENOMEM); sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_port = store.sin.sin_port; sin->sin_addr = store.sin.sin_addr; *addr = (struct sockaddr *)sin; break; } #endif #ifdef INET6 case AF_INET6: { struct sockaddr_in6 *sin6; SCTP_MALLOC_SONAME(sin6, struct sockaddr_in6 *, sizeof *sin6); if (sin6 == NULL) return (ENOMEM); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); sin6->sin6_port = store.sin6.sin6_port; sin6->sin6_addr = store.sin6.sin6_addr; if ((error = sa6_recoverscope(sin6)) != 0) { SCTP_FREE_SONAME(sin6); return (error); } *addr = (struct sockaddr *)sin6; break; } #endif default: /* TSNH */ break; } /* Wake any delayed sleep action */ if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) { SCTP_INP_WLOCK(inp); inp->sctp_flags &= ~SCTP_PCB_FLAGS_DONT_WAKE; if (inp->sctp_flags & SCTP_PCB_FLAGS_WAKEOUTPUT) { inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT; SCTP_INP_WUNLOCK(inp); SOCKBUF_LOCK(&inp->sctp_socket->so_snd); if (sowriteable(inp->sctp_socket)) { sowwakeup_locked(inp->sctp_socket); } else { SOCKBUF_UNLOCK(&inp->sctp_socket->so_snd); } SCTP_INP_WLOCK(inp); } if (inp->sctp_flags & SCTP_PCB_FLAGS_WAKEINPUT) { inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT; SCTP_INP_WUNLOCK(inp); SOCKBUF_LOCK(&inp->sctp_socket->so_rcv); if (soreadable(inp->sctp_socket)) { sctp_defered_wakeup_cnt++; sorwakeup_locked(inp->sctp_socket); } else { SOCKBUF_UNLOCK(&inp->sctp_socket->so_rcv); } SCTP_INP_WLOCK(inp); } SCTP_INP_WUNLOCK(inp); } if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) { SCTP_TCB_LOCK(stcb); sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_19); } return (0); } #ifdef INET int sctp_ingetaddr(struct socket *so, struct sockaddr **addr) { struct sockaddr_in *sin; uint32_t vrf_id; struct sctp_inpcb *inp; struct sctp_ifa *sctp_ifa; /* * Do the malloc first in case it blocks. */ SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin); if (sin == NULL) return (ENOMEM); sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); inp = (struct sctp_inpcb *)so->so_pcb; if (!inp) { SCTP_FREE_SONAME(sin); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } SCTP_INP_RLOCK(inp); sin->sin_port = inp->sctp_lport; if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) { if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) { struct sctp_tcb *stcb; struct sockaddr_in *sin_a; struct sctp_nets *net; int fnd; stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb == NULL) { goto notConn; } fnd = 0; sin_a = NULL; SCTP_TCB_LOCK(stcb); TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { sin_a = (struct sockaddr_in *)&net->ro._l_addr; if (sin_a == NULL) /* this will make coverity happy */ continue; if (sin_a->sin_family == AF_INET) { fnd = 1; break; } } if ((!fnd) || (sin_a == NULL)) { /* punt */ SCTP_TCB_UNLOCK(stcb); goto notConn; } vrf_id = inp->def_vrf_id; sctp_ifa = sctp_source_address_selection(inp, stcb, (sctp_route_t *) & net->ro, net, 0, vrf_id); if (sctp_ifa) { sin->sin_addr = sctp_ifa->address.sin.sin_addr; sctp_free_ifa(sctp_ifa); } SCTP_TCB_UNLOCK(stcb); } else { /* For the bound all case you get back 0 */ notConn: sin->sin_addr.s_addr = 0; } } else { /* Take the first IPv4 address in the list */ struct sctp_laddr *laddr; int fnd = 0; LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) { if (laddr->ifa->address.sa.sa_family == AF_INET) { struct sockaddr_in *sin_a; sin_a = &laddr->ifa->address.sin; sin->sin_addr = sin_a->sin_addr; fnd = 1; break; } } if (!fnd) { SCTP_FREE_SONAME(sin); SCTP_INP_RUNLOCK(inp); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); return (ENOENT); } } SCTP_INP_RUNLOCK(inp); (*addr) = (struct sockaddr *)sin; return (0); } int sctp_peeraddr(struct socket *so, struct sockaddr **addr) { struct sockaddr_in *sin; int fnd; struct sockaddr_in *sin_a; struct sctp_inpcb *inp; struct sctp_tcb *stcb; struct sctp_nets *net; /* Do the malloc first in case it blocks. */ SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin); if (sin == NULL) return (ENOMEM); sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); inp = (struct sctp_inpcb *)so->so_pcb; if ((inp == NULL) || ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0)) { /* UDP type and listeners will drop out here */ SCTP_FREE_SONAME(sin); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN); return (ENOTCONN); } SCTP_INP_RLOCK(inp); stcb = LIST_FIRST(&inp->sctp_asoc_list); if (stcb) { SCTP_TCB_LOCK(stcb); } SCTP_INP_RUNLOCK(inp); if (stcb == NULL) { SCTP_FREE_SONAME(sin); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); return (ECONNRESET); } fnd = 0; TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) { sin_a = (struct sockaddr_in *)&net->ro._l_addr; if (sin_a->sin_family == AF_INET) { fnd = 1; sin->sin_port = stcb->rport; sin->sin_addr = sin_a->sin_addr; break; } } SCTP_TCB_UNLOCK(stcb); if (!fnd) { /* No IPv4 address */ SCTP_FREE_SONAME(sin); SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); return (ENOENT); } (*addr) = (struct sockaddr *)sin; return (0); } struct pr_usrreqs sctp_usrreqs = { .pru_abort = sctp_abort, .pru_accept = sctp_accept, .pru_attach = sctp_attach, .pru_bind = sctp_bind, .pru_connect = sctp_connect, .pru_control = in_control, .pru_close = sctp_close, .pru_detach = sctp_close, .pru_sopoll = sopoll_generic, .pru_flush = sctp_flush, .pru_disconnect = sctp_disconnect, .pru_listen = sctp_listen, .pru_peeraddr = sctp_peeraddr, .pru_send = sctp_sendm, .pru_shutdown = sctp_shutdown, .pru_sockaddr = sctp_ingetaddr, .pru_sosend = sctp_sosend, .pru_soreceive = sctp_soreceive }; #endif Index: projects/vnet/sys/netinet/sctp_var.h =================================================================== --- projects/vnet/sys/netinet/sctp_var.h (revision 294565) +++ projects/vnet/sys/netinet/sctp_var.h (revision 294566) @@ -1,356 +1,355 @@ /*- * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved. * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved. * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: * * a) Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. * * b) Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the distribution. * * c) Neither the name of Cisco Systems, Inc. nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #ifndef _NETINET_SCTP_VAR_H_ #define _NETINET_SCTP_VAR_H_ #include #if defined(_KERNEL) || defined(__Userspace__) extern struct pr_usrreqs sctp_usrreqs; #define sctp_feature_on(inp, feature) (inp->sctp_features |= feature) #define sctp_feature_off(inp, feature) (inp->sctp_features &= ~feature) #define sctp_is_feature_on(inp, feature) ((inp->sctp_features & feature) == feature) #define sctp_is_feature_off(inp, feature) ((inp->sctp_features & feature) == 0) #define sctp_stcb_feature_on(inp, stcb, feature) {\ if (stcb) { \ stcb->asoc.sctp_features |= feature; \ } else if (inp) { \ inp->sctp_features |= feature; \ } \ } #define sctp_stcb_feature_off(inp, stcb, feature) {\ if (stcb) { \ stcb->asoc.sctp_features &= ~feature; \ } else if (inp) { \ inp->sctp_features &= ~feature; \ } \ } #define sctp_stcb_is_feature_on(inp, stcb, feature) \ (((stcb != NULL) && \ ((stcb->asoc.sctp_features & feature) == feature)) || \ ((stcb == NULL) && (inp != NULL) && \ ((inp->sctp_features & feature) == feature))) #define sctp_stcb_is_feature_off(inp, stcb, feature) \ (((stcb != NULL) && \ ((stcb->asoc.sctp_features & feature) == 0)) || \ ((stcb == NULL) && (inp != NULL) && \ ((inp->sctp_features & feature) == 0)) || \ ((stcb == NULL) && (inp == NULL))) /* managing mobility_feature in inpcb (by micchie) */ #define sctp_mobility_feature_on(inp, feature) (inp->sctp_mobility_features |= feature) #define sctp_mobility_feature_off(inp, feature) (inp->sctp_mobility_features &= ~feature) #define sctp_is_mobility_feature_on(inp, feature) (inp->sctp_mobility_features & feature) #define sctp_is_mobility_feature_off(inp, feature) ((inp->sctp_mobility_features & feature) == 0) #define sctp_maxspace(sb) (max((sb)->sb_hiwat,SCTP_MINIMAL_RWND)) #define sctp_sbspace(asoc, sb) ((long) ((sctp_maxspace(sb) > (asoc)->sb_cc) ? (sctp_maxspace(sb) - (asoc)->sb_cc) : 0)) #define sctp_sbspace_failedmsgs(sb) ((long) ((sctp_maxspace(sb) > (sb)->sb_cc) ? (sctp_maxspace(sb) - (sb)->sb_cc) : 0)) #define sctp_sbspace_sub(a,b) ((a > b) ? (a - b) : 0) /* * I tried to cache the readq entries at one point. But the reality * is that it did not add any performance since this meant we had to * lock the STCB on read. And at that point once you have to do an * extra lock, it really does not matter if the lock is in the ZONE * stuff or in our code. Note that this same problem would occur with * an mbuf cache as well so it is not really worth doing, at least * right now :-D */ #define sctp_free_a_readq(_stcb, _readq) { \ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), (_readq)); \ SCTP_DECR_READQ_COUNT(); \ } #define sctp_alloc_a_readq(_stcb, _readq) { \ (_readq) = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_readq), struct sctp_queued_to_read); \ if ((_readq)) { \ SCTP_INCR_READQ_COUNT(); \ } \ } #define sctp_free_a_strmoq(_stcb, _strmoq, _so_locked) { \ if ((_strmoq)->holds_key_ref) { \ sctp_auth_key_release(stcb, sp->auth_keyid, _so_locked); \ (_strmoq)->holds_key_ref = 0; \ } \ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), (_strmoq)); \ SCTP_DECR_STRMOQ_COUNT(); \ } #define sctp_alloc_a_strmoq(_stcb, _strmoq) { \ (_strmoq) = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_strmoq), struct sctp_stream_queue_pending); \ if ((_strmoq)) { \ memset(_strmoq, 0, sizeof(struct sctp_stream_queue_pending)); \ SCTP_INCR_STRMOQ_COUNT(); \ (_strmoq)->holds_key_ref = 0; \ } \ } #define sctp_free_a_chunk(_stcb, _chk, _so_locked) { \ if ((_chk)->holds_key_ref) {\ sctp_auth_key_release((_stcb), (_chk)->auth_keyid, _so_locked); \ (_chk)->holds_key_ref = 0; \ } \ if (_stcb) { \ SCTP_TCB_LOCK_ASSERT((_stcb)); \ if ((_chk)->whoTo) { \ sctp_free_remote_addr((_chk)->whoTo); \ (_chk)->whoTo = NULL; \ } \ if (((_stcb)->asoc.free_chunk_cnt > SCTP_BASE_SYSCTL(sctp_asoc_free_resc_limit)) || \ (SCTP_BASE_INFO(ipi_free_chunks) > SCTP_BASE_SYSCTL(sctp_system_free_resc_limit))) { \ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), (_chk)); \ SCTP_DECR_CHK_COUNT(); \ } else { \ TAILQ_INSERT_TAIL(&(_stcb)->asoc.free_chunks, (_chk), sctp_next); \ (_stcb)->asoc.free_chunk_cnt++; \ atomic_add_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); \ } \ } else { \ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), (_chk)); \ SCTP_DECR_CHK_COUNT(); \ } \ } #define sctp_alloc_a_chunk(_stcb, _chk) { \ if (TAILQ_EMPTY(&(_stcb)->asoc.free_chunks)) { \ (_chk) = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_chunk), struct sctp_tmit_chunk); \ if ((_chk)) { \ SCTP_INCR_CHK_COUNT(); \ (_chk)->whoTo = NULL; \ (_chk)->holds_key_ref = 0; \ } \ } else { \ (_chk) = TAILQ_FIRST(&(_stcb)->asoc.free_chunks); \ TAILQ_REMOVE(&(_stcb)->asoc.free_chunks, (_chk), sctp_next); \ atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1); \ (_chk)->holds_key_ref = 0; \ SCTP_STAT_INCR(sctps_cached_chk); \ (_stcb)->asoc.free_chunk_cnt--; \ } \ } #define sctp_free_remote_addr(__net) { \ if ((__net)) { \ if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&(__net)->ref_count)) { \ (void)SCTP_OS_TIMER_STOP(&(__net)->rxt_timer.timer); \ (void)SCTP_OS_TIMER_STOP(&(__net)->pmtu_timer.timer); \ (void)SCTP_OS_TIMER_STOP(&(__net)->hb_timer.timer); \ if ((__net)->ro.ro_rt) { \ RTFREE((__net)->ro.ro_rt); \ (__net)->ro.ro_rt = NULL; \ } \ if ((__net)->src_addr_selected) { \ sctp_free_ifa((__net)->ro._s_addr); \ (__net)->ro._s_addr = NULL; \ } \ (__net)->src_addr_selected = 0; \ (__net)->dest_state &= ~SCTP_ADDR_REACHABLE; \ SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_net), (__net)); \ SCTP_DECR_RADDR_COUNT(); \ } \ } \ } #define sctp_sbfree(ctl, stcb, sb, m) { \ SCTP_SAVE_ATOMIC_DECREMENT(&(sb)->sb_cc, SCTP_BUF_LEN((m))); \ SCTP_SAVE_ATOMIC_DECREMENT(&(sb)->sb_mbcnt, MSIZE); \ if (((ctl)->do_not_ref_stcb == 0) && stcb) {\ SCTP_SAVE_ATOMIC_DECREMENT(&(stcb)->asoc.sb_cc, SCTP_BUF_LEN((m))); \ SCTP_SAVE_ATOMIC_DECREMENT(&(stcb)->asoc.my_rwnd_control_len, MSIZE); \ } \ if (SCTP_BUF_TYPE(m) != MT_DATA && SCTP_BUF_TYPE(m) != MT_HEADER && \ SCTP_BUF_TYPE(m) != MT_OOBDATA) \ atomic_subtract_int(&(sb)->sb_ctl,SCTP_BUF_LEN((m))); \ } #define sctp_sballoc(stcb, sb, m) { \ atomic_add_int(&(sb)->sb_cc,SCTP_BUF_LEN((m))); \ atomic_add_int(&(sb)->sb_mbcnt, MSIZE); \ if (stcb) { \ atomic_add_int(&(stcb)->asoc.sb_cc,SCTP_BUF_LEN((m))); \ atomic_add_int(&(stcb)->asoc.my_rwnd_control_len, MSIZE); \ } \ if (SCTP_BUF_TYPE(m) != MT_DATA && SCTP_BUF_TYPE(m) != MT_HEADER && \ SCTP_BUF_TYPE(m) != MT_OOBDATA) \ atomic_add_int(&(sb)->sb_ctl,SCTP_BUF_LEN((m))); \ } #define sctp_ucount_incr(val) { \ val++; \ } #define sctp_ucount_decr(val) { \ if (val > 0) { \ val--; \ } else { \ val = 0; \ } \ } #define sctp_mbuf_crush(data) do { \ struct mbuf *_m; \ _m = (data); \ while (_m && (SCTP_BUF_LEN(_m) == 0)) { \ (data) = SCTP_BUF_NEXT(_m); \ SCTP_BUF_NEXT(_m) = NULL; \ sctp_m_free(_m); \ _m = (data); \ } \ } while (0) #define sctp_flight_size_decrease(tp1) do { \ if (tp1->whoTo->flight_size >= tp1->book_size) \ tp1->whoTo->flight_size -= tp1->book_size; \ else \ tp1->whoTo->flight_size = 0; \ } while (0) #define sctp_flight_size_increase(tp1) do { \ (tp1)->whoTo->flight_size += (tp1)->book_size; \ } while (0) #ifdef SCTP_FS_SPEC_LOG #define sctp_total_flight_decrease(stcb, tp1) do { \ if (stcb->asoc.fs_index > SCTP_FS_SPEC_LOG_SIZE) \ stcb->asoc.fs_index = 0;\ stcb->asoc.fslog[stcb->asoc.fs_index].total_flight = stcb->asoc.total_flight; \ stcb->asoc.fslog[stcb->asoc.fs_index].tsn = tp1->rec.data.TSN_seq; \ stcb->asoc.fslog[stcb->asoc.fs_index].book = tp1->book_size; \ stcb->asoc.fslog[stcb->asoc.fs_index].sent = tp1->sent; \ stcb->asoc.fslog[stcb->asoc.fs_index].incr = 0; \ stcb->asoc.fslog[stcb->asoc.fs_index].decr = 1; \ stcb->asoc.fs_index++; \ tp1->window_probe = 0; \ if (stcb->asoc.total_flight >= tp1->book_size) { \ stcb->asoc.total_flight -= tp1->book_size; \ if (stcb->asoc.total_flight_count > 0) \ stcb->asoc.total_flight_count--; \ } else { \ stcb->asoc.total_flight = 0; \ stcb->asoc.total_flight_count = 0; \ } \ } while (0) #define sctp_total_flight_increase(stcb, tp1) do { \ if (stcb->asoc.fs_index > SCTP_FS_SPEC_LOG_SIZE) \ stcb->asoc.fs_index = 0;\ stcb->asoc.fslog[stcb->asoc.fs_index].total_flight = stcb->asoc.total_flight; \ stcb->asoc.fslog[stcb->asoc.fs_index].tsn = tp1->rec.data.TSN_seq; \ stcb->asoc.fslog[stcb->asoc.fs_index].book = tp1->book_size; \ stcb->asoc.fslog[stcb->asoc.fs_index].sent = tp1->sent; \ stcb->asoc.fslog[stcb->asoc.fs_index].incr = 1; \ stcb->asoc.fslog[stcb->asoc.fs_index].decr = 0; \ stcb->asoc.fs_index++; \ (stcb)->asoc.total_flight_count++; \ (stcb)->asoc.total_flight += (tp1)->book_size; \ } while (0) #else #define sctp_total_flight_decrease(stcb, tp1) do { \ tp1->window_probe = 0; \ if (stcb->asoc.total_flight >= tp1->book_size) { \ stcb->asoc.total_flight -= tp1->book_size; \ if (stcb->asoc.total_flight_count > 0) \ stcb->asoc.total_flight_count--; \ } else { \ stcb->asoc.total_flight = 0; \ stcb->asoc.total_flight_count = 0; \ } \ } while (0) #define sctp_total_flight_increase(stcb, tp1) do { \ (stcb)->asoc.total_flight_count++; \ (stcb)->asoc.total_flight += (tp1)->book_size; \ } while (0) #endif #define SCTP_PF_ENABLED(_net) (_net->pf_threshold < _net->failure_threshold) #define SCTP_NET_IS_PF(_net) (_net->pf_threshold < _net->error_count) struct sctp_nets; struct sctp_inpcb; struct sctp_tcb; struct sctphdr; void sctp_close(struct socket *so); int sctp_disconnect(struct socket *so); void sctp_ctlinput(int, struct sockaddr *, void *); int sctp_ctloutput(struct socket *, struct sockopt *); #ifdef INET void sctp_input_with_port(struct mbuf *, int, uint16_t); int sctp_input(struct mbuf **, int *, int); #endif void sctp_pathmtu_adjustment(struct sctp_tcb *, uint16_t); void sctp_drain(void); void sctp_init(void); -void sctp_finish(void); int sctp_flush(struct socket *, int); int sctp_shutdown(struct socket *); void sctp_notify(struct sctp_inpcb *, struct ip *ip, struct sctphdr *, struct sockaddr *, struct sctp_tcb *, struct sctp_nets *); int sctp_bindx(struct socket *, int, struct sockaddr_storage *, int, int, struct proc *); /* can't use sctp_assoc_t here */ int sctp_peeloff(struct socket *, struct socket *, int, caddr_t, int *); int sctp_ingetaddr(struct socket *, struct sockaddr **); int sctp_peeraddr(struct socket *, struct sockaddr **); int sctp_listen(struct socket *, int, struct thread *); int sctp_accept(struct socket *, struct sockaddr **); #endif /* _KERNEL */ #endif /* !_NETINET_SCTP_VAR_H_ */ Index: projects/vnet/sys/netinet/tcp_subr.c =================================================================== --- projects/vnet/sys/netinet/tcp_subr.c (revision 294565) +++ projects/vnet/sys/netinet/tcp_subr.c (revision 294566) @@ -1,2867 +1,2868 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #include #include #endif #ifdef TCP_RFC7413 #include #endif #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef TCPPCAP #include #endif #ifdef TCPDEBUG #include #endif #ifdef INET6 #include #endif #ifdef TCP_OFFLOAD #include #endif #ifdef IPSEC #include #include #ifdef INET6 #include #endif #include #include #endif /*IPSEC*/ #include #include #include VNET_DEFINE(int, tcp_mssdflt) = TCP_MSS; #ifdef INET6 VNET_DEFINE(int, tcp_v6mssdflt) = TCP6_MSS; #endif struct rwlock tcp_function_lock; static int sysctl_net_inet_tcp_mss_check(SYSCTL_HANDLER_ARGS) { int error, new; new = V_tcp_mssdflt; error = sysctl_handle_int(oidp, &new, 0, req); if (error == 0 && req->newptr) { if (new < TCP_MINMSS) error = EINVAL; else V_tcp_mssdflt = new; } return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(tcp_mssdflt), 0, &sysctl_net_inet_tcp_mss_check, "I", "Default TCP Maximum Segment Size"); #ifdef INET6 static int sysctl_net_inet_tcp_mss_v6_check(SYSCTL_HANDLER_ARGS) { int error, new; new = V_tcp_v6mssdflt; error = sysctl_handle_int(oidp, &new, 0, req); if (error == 0 && req->newptr) { if (new < TCP_MINMSS) error = EINVAL; else V_tcp_v6mssdflt = new; } return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(tcp_v6mssdflt), 0, &sysctl_net_inet_tcp_mss_v6_check, "I", "Default TCP Maximum Segment Size for IPv6"); #endif /* INET6 */ /* * Minimum MSS we accept and use. This prevents DoS attacks where * we are forced to a ridiculous low MSS like 20 and send hundreds * of packets instead of one. The effect scales with the available * bandwidth and quickly saturates the CPU and network interface * with packet generation and sending. Set to zero to disable MINMSS * checking. This setting prevents us from sending too small packets. */ VNET_DEFINE(int, tcp_minmss) = TCP_MINMSS; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_minmss), 0, "Minimum TCP Maximum Segment Size"); VNET_DEFINE(int, tcp_do_rfc1323) = 1; SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_do_rfc1323), 0, "Enable rfc1323 (high performance TCP) extensions"); static int tcp_log_debug = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, log_debug, CTLFLAG_RW, &tcp_log_debug, 0, "Log errors caused by incoming TCP segments"); static int tcp_tcbhashsize; SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); static int do_tcpdrain = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, "Enable tcp_drain routine for extra help when low on mbufs"); SYSCTL_UINT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(tcbinfo.ipi_count), 0, "Number of active PCBs"); static VNET_DEFINE(int, icmp_may_rst) = 1; #define V_icmp_may_rst VNET(icmp_may_rst) SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp_may_rst), 0, "Certain ICMP unreachable messages may abort connections in SYN_SENT"); static VNET_DEFINE(int, tcp_isn_reseed_interval) = 0; #define V_tcp_isn_reseed_interval VNET(tcp_isn_reseed_interval) SYSCTL_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_isn_reseed_interval), 0, "Seconds between reseeding of ISN secret"); static int tcp_soreceive_stream; SYSCTL_INT(_net_inet_tcp, OID_AUTO, soreceive_stream, CTLFLAG_RDTUN, &tcp_soreceive_stream, 0, "Using soreceive_stream for TCP sockets"); #ifdef TCP_SIGNATURE static int tcp_sig_checksigs = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, signature_verify_input, CTLFLAG_RW, &tcp_sig_checksigs, 0, "Verify RFC2385 digests on inbound traffic"); #endif VNET_DEFINE(uma_zone_t, sack_hole_zone); #define V_sack_hole_zone VNET(sack_hole_zone) VNET_DEFINE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST+1]); static struct inpcb *tcp_notify(struct inpcb *, int); static struct inpcb *tcp_mtudisc_notify(struct inpcb *, int); static void tcp_mtudisc(struct inpcb *, int); static char * tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr); static void tcp_timer_discard(struct tcpcb *, uint32_t); static struct tcp_function_block tcp_def_funcblk = { "default", tcp_output, tcp_do_segment, tcp_default_ctloutput, NULL, NULL, NULL, NULL, NULL, NULL, NULL, 0, 0 }; struct tcp_funchead t_functions; static struct tcp_function_block *tcp_func_set_ptr = &tcp_def_funcblk; static struct tcp_function_block * find_tcp_functions_locked(struct tcp_function_set *fs) { struct tcp_function *f; struct tcp_function_block *blk=NULL; TAILQ_FOREACH(f, &t_functions, tf_next) { if (strcmp(f->tf_fb->tfb_tcp_block_name, fs->function_set_name) == 0) { blk = f->tf_fb; break; } } return(blk); } static struct tcp_function_block * find_tcp_fb_locked(struct tcp_function_block *blk, struct tcp_function **s) { struct tcp_function_block *rblk=NULL; struct tcp_function *f; TAILQ_FOREACH(f, &t_functions, tf_next) { if (f->tf_fb == blk) { rblk = blk; if (s) { *s = f; } break; } } return (rblk); } struct tcp_function_block * find_and_ref_tcp_functions(struct tcp_function_set *fs) { struct tcp_function_block *blk; rw_rlock(&tcp_function_lock); blk = find_tcp_functions_locked(fs); if (blk) refcount_acquire(&blk->tfb_refcnt); rw_runlock(&tcp_function_lock); return(blk); } struct tcp_function_block * find_and_ref_tcp_fb(struct tcp_function_block *blk) { struct tcp_function_block *rblk; rw_rlock(&tcp_function_lock); rblk = find_tcp_fb_locked(blk, NULL); if (rblk) refcount_acquire(&rblk->tfb_refcnt); rw_runlock(&tcp_function_lock); return(rblk); } static int sysctl_net_inet_default_tcp_functions(SYSCTL_HANDLER_ARGS) { int error=ENOENT; struct tcp_function_set fs; struct tcp_function_block *blk; memset(&fs, 0, sizeof(fs)); rw_rlock(&tcp_function_lock); blk = find_tcp_fb_locked(tcp_func_set_ptr, NULL); if (blk) { /* Found him */ strcpy(fs.function_set_name, blk->tfb_tcp_block_name); fs.pcbcnt = blk->tfb_refcnt; } rw_runlock(&tcp_function_lock); error = sysctl_handle_string(oidp, fs.function_set_name, sizeof(fs.function_set_name), req); /* Check for error or no change */ if (error != 0 || req->newptr == NULL) return(error); rw_wlock(&tcp_function_lock); blk = find_tcp_functions_locked(&fs); if ((blk == NULL) || (blk->tfb_flags & TCP_FUNC_BEING_REMOVED)) { error = ENOENT; goto done; } tcp_func_set_ptr = blk; done: rw_wunlock(&tcp_function_lock); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, functions_default, CTLTYPE_STRING | CTLFLAG_RW, NULL, 0, sysctl_net_inet_default_tcp_functions, "A", "Set/get the default TCP functions"); static int sysctl_net_inet_list_available(SYSCTL_HANDLER_ARGS) { int error, cnt, linesz; struct tcp_function *f; char *buffer, *cp; size_t bufsz, outsz; cnt = 0; rw_rlock(&tcp_function_lock); TAILQ_FOREACH(f, &t_functions, tf_next) { cnt++; } rw_runlock(&tcp_function_lock); bufsz = (cnt+2) * (TCP_FUNCTION_NAME_LEN_MAX + 12) + 1; buffer = malloc(bufsz, M_TEMP, M_WAITOK); error = 0; cp = buffer; linesz = snprintf(cp, bufsz, "\n%-32s%c %s\n", "Stack", 'D', "PCB count"); cp += linesz; bufsz -= linesz; outsz = linesz; rw_rlock(&tcp_function_lock); TAILQ_FOREACH(f, &t_functions, tf_next) { linesz = snprintf(cp, bufsz, "%-32s%c %u\n", f->tf_fb->tfb_tcp_block_name, (f->tf_fb == tcp_func_set_ptr) ? '*' : ' ', f->tf_fb->tfb_refcnt); if (linesz >= bufsz) { error = EOVERFLOW; break; } cp += linesz; bufsz -= linesz; outsz += linesz; } rw_runlock(&tcp_function_lock); if (error == 0) error = sysctl_handle_string(oidp, buffer, outsz + 1, req); free(buffer, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, functions_available, CTLTYPE_STRING|CTLFLAG_RD, NULL, 0, sysctl_net_inet_list_available, "A", "list available TCP Function sets"); /* * Target size of TCP PCB hash tables. Must be a power of two. * * Note that this can be overridden by the kernel environment * variable net.inet.tcp.tcbhashsize */ #ifndef TCBHASHSIZE #define TCBHASHSIZE 0 #endif /* * XXX * Callouts should be moved into struct tcp directly. They are currently * separate because the tcpcb structure is exported to userland for sysctl * parsing purposes, which do not know about callouts. */ struct tcpcb_mem { struct tcpcb tcb; struct tcp_timer tt; struct cc_var ccv; struct osd osd; }; static VNET_DEFINE(uma_zone_t, tcpcb_zone); #define V_tcpcb_zone VNET(tcpcb_zone) MALLOC_DEFINE(M_TCPLOG, "tcplog", "TCP address and flags print buffers"); MALLOC_DEFINE(M_TCPFUNCTIONS, "tcpfunc", "TCP function set memory"); static struct mtx isn_mtx; #define ISN_LOCK_INIT() mtx_init(&isn_mtx, "isn_mtx", NULL, MTX_DEF) #define ISN_LOCK() mtx_lock(&isn_mtx) #define ISN_UNLOCK() mtx_unlock(&isn_mtx) /* * TCP initialization. */ static void tcp_zone_change(void *tag) { uma_zone_set_max(V_tcbinfo.ipi_zone, maxsockets); uma_zone_set_max(V_tcpcb_zone, maxsockets); tcp_tw_zone_change(); } static int tcp_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "tcpinp"); return (0); } /* * Take a value and get the next power of 2 that doesn't overflow. * Used to size the tcp_inpcb hash buckets. */ static int maketcp_hashsize(int size) { int hashsize; /* * auto tune. * get the next power of 2 higher than maxsockets. */ hashsize = 1 << fls(size); /* catch overflow, and just go one power of 2 smaller */ if (hashsize < size) { hashsize = 1 << (fls(size) - 1); } return (hashsize); } int register_tcp_functions(struct tcp_function_block *blk, int wait) { struct tcp_function_block *lblk; struct tcp_function *n; struct tcp_function_set fs; if ((blk->tfb_tcp_output == NULL) || (blk->tfb_tcp_do_segment == NULL) || (blk->tfb_tcp_ctloutput == NULL) || (strlen(blk->tfb_tcp_block_name) == 0)) { /* * These functions are required and you * need a name. */ return (EINVAL); } if (blk->tfb_tcp_timer_stop_all || blk->tfb_tcp_timers_left || blk->tfb_tcp_timer_activate || blk->tfb_tcp_timer_active || blk->tfb_tcp_timer_stop) { /* * If you define one timer function you * must have them all. */ if ((blk->tfb_tcp_timer_stop_all == NULL) || (blk->tfb_tcp_timers_left == NULL) || (blk->tfb_tcp_timer_activate == NULL) || (blk->tfb_tcp_timer_active == NULL) || (blk->tfb_tcp_timer_stop == NULL)) { return (EINVAL); } } n = malloc(sizeof(struct tcp_function), M_TCPFUNCTIONS, wait); if (n == NULL) { return (ENOMEM); } n->tf_fb = blk; strcpy(fs.function_set_name, blk->tfb_tcp_block_name); rw_wlock(&tcp_function_lock); lblk = find_tcp_functions_locked(&fs); if (lblk) { /* Duplicate name space not allowed */ rw_wunlock(&tcp_function_lock); free(n, M_TCPFUNCTIONS); return (EALREADY); } refcount_init(&blk->tfb_refcnt, 0); blk->tfb_flags = 0; TAILQ_INSERT_TAIL(&t_functions, n, tf_next); rw_wunlock(&tcp_function_lock); return(0); } int deregister_tcp_functions(struct tcp_function_block *blk) { struct tcp_function_block *lblk; struct tcp_function *f; int error=ENOENT; if (strcmp(blk->tfb_tcp_block_name, "default") == 0) { /* You can't un-register the default */ return (EPERM); } rw_wlock(&tcp_function_lock); if (blk == tcp_func_set_ptr) { /* You can't free the current default */ rw_wunlock(&tcp_function_lock); return (EBUSY); } if (blk->tfb_refcnt) { /* Still tcb attached, mark it. */ blk->tfb_flags |= TCP_FUNC_BEING_REMOVED; rw_wunlock(&tcp_function_lock); return (EBUSY); } lblk = find_tcp_fb_locked(blk, &f); if (lblk) { /* Found */ TAILQ_REMOVE(&t_functions, f, tf_next); f->tf_fb = NULL; free(f, M_TCPFUNCTIONS); error = 0; } rw_wunlock(&tcp_function_lock); return (error); } void tcp_init(void) { const char *tcbhash_tuneable; int hashsize; tcbhash_tuneable = "net.inet.tcp.tcbhashsize"; if (hhook_head_register(HHOOK_TYPE_TCP, HHOOK_TCP_EST_IN, &V_tcp_hhh[HHOOK_TCP_EST_IN], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register helper hook\n", __func__); if (hhook_head_register(HHOOK_TYPE_TCP, HHOOK_TCP_EST_OUT, &V_tcp_hhh[HHOOK_TCP_EST_OUT], HHOOK_NOWAIT|HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register helper hook\n", __func__); hashsize = TCBHASHSIZE; TUNABLE_INT_FETCH(tcbhash_tuneable, &hashsize); if (hashsize == 0) { /* * Auto tune the hash size based on maxsockets. * A perfect hash would have a 1:1 mapping * (hashsize = maxsockets) however it's been * suggested that O(2) average is better. */ hashsize = maketcp_hashsize(maxsockets / 4); /* * Our historical default is 512, * do not autotune lower than this. */ if (hashsize < 512) hashsize = 512; if (bootverbose && IS_DEFAULT_VNET(curvnet)) printf("%s: %s auto tuned to %d\n", __func__, tcbhash_tuneable, hashsize); } /* * We require a hashsize to be a power of two. * Previously if it was not a power of two we would just reset it * back to 512, which could be a nasty surprise if you did not notice * the error message. * Instead what we do is clip it to the closest power of two lower * than the specified hash value. */ if (!powerof2(hashsize)) { int oldhashsize = hashsize; hashsize = maketcp_hashsize(hashsize); /* prevent absurdly low value */ if (hashsize < 16) hashsize = 16; printf("%s: WARNING: TCB hash size not a power of 2, " "clipped from %d to %d.\n", __func__, oldhashsize, hashsize); } in_pcbinfo_init(&V_tcbinfo, "tcp", &V_tcb, hashsize, hashsize, "tcp_inpcb", tcp_inpcb_init, NULL, UMA_ZONE_NOFREE, IPI_HASHFIELDS_4TUPLE); /* * These have to be type stable for the benefit of the timers. */ V_tcpcb_zone = uma_zcreate("tcpcb", sizeof(struct tcpcb_mem), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(V_tcpcb_zone, maxsockets); uma_zone_set_warning(V_tcpcb_zone, "kern.ipc.maxsockets limit reached"); tcp_tw_init(); syncache_init(); tcp_hc_init(); TUNABLE_INT_FETCH("net.inet.tcp.sack.enable", &V_tcp_do_sack); V_sack_hole_zone = uma_zcreate("sackhole", sizeof(struct sackhole), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); /* Skip initialization of globals for non-default instances. */ if (!IS_DEFAULT_VNET(curvnet)) return; tcp_reass_global_init(); /* XXX virtualize those bellow? */ tcp_delacktime = TCPTV_DELACK; tcp_keepinit = TCPTV_KEEP_INIT; tcp_keepidle = TCPTV_KEEP_IDLE; tcp_keepintvl = TCPTV_KEEPINTVL; tcp_maxpersistidle = TCPTV_KEEP_IDLE; tcp_msl = TCPTV_MSL; tcp_rexmit_min = TCPTV_MIN; if (tcp_rexmit_min < 1) tcp_rexmit_min = 1; tcp_rexmit_slop = TCPTV_CPU_VAR; tcp_finwait2_timeout = TCPTV_FINWAIT2_TIMEOUT; tcp_tcbhashsize = hashsize; /* Setup the tcp function block list */ TAILQ_INIT(&t_functions); rw_init_flags(&tcp_function_lock, "tcp_func_lock" , 0); register_tcp_functions(&tcp_def_funcblk, M_WAITOK); if (tcp_soreceive_stream) { #ifdef INET tcp_usrreqs.pru_soreceive = soreceive_stream; #endif #ifdef INET6 tcp6_usrreqs.pru_soreceive = soreceive_stream; #endif /* INET6 */ } #ifdef INET6 #define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) #else /* INET6 */ #define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) #endif /* INET6 */ if (max_protohdr < TCP_MINPROTOHDR) max_protohdr = TCP_MINPROTOHDR; if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) panic("tcp_init"); #undef TCP_MINPROTOHDR ISN_LOCK_INIT(); EVENTHANDLER_REGISTER(shutdown_pre_sync, tcp_fini, NULL, SHUTDOWN_PRI_DEFAULT); EVENTHANDLER_REGISTER(maxsockets_change, tcp_zone_change, NULL, EVENTHANDLER_PRI_ANY); #ifdef TCPPCAP tcp_pcap_init(); #endif #ifdef TCP_RFC7413 tcp_fastopen_init(); #endif } #ifdef VIMAGE -void +static void tcp_destroy(void) { int error; #ifdef TCP_RFC7413 tcp_fastopen_destroy(); #endif tcp_hc_destroy(); syncache_destroy(); tcp_tw_destroy(); in_pcbinfo_destroy(&V_tcbinfo); uma_zdestroy(V_sack_hole_zone); uma_zdestroy(V_tcpcb_zone); error = hhook_head_deregister(V_tcp_hhh[HHOOK_TCP_EST_IN]); if (error != 0) { printf("%s: WARNING: unable to deregister helper hook " "type=%d, id=%d: error %d returned\n", __func__, HHOOK_TYPE_TCP, HHOOK_TCP_EST_IN, error); } error = hhook_head_deregister(V_tcp_hhh[HHOOK_TCP_EST_OUT]); if (error != 0) { printf("%s: WARNING: unable to deregister helper hook " "type=%d, id=%d: error %d returned\n", __func__, HHOOK_TYPE_TCP, HHOOK_TCP_EST_OUT, error); } } +VNET_SYSUNINIT(tcp, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, tcp_destroy, NULL); #endif void tcp_fini(void *xtp) { } /* * Fill in the IP and TCP headers for an outgoing packet, given the tcpcb. * tcp_template used to store this data in mbufs, but we now recopy it out * of the tcpcb each time to conserve mbufs. */ void tcpip_fillheaders(struct inpcb *inp, void *ip_ptr, void *tcp_ptr) { struct tcphdr *th = (struct tcphdr *)tcp_ptr; INP_WLOCK_ASSERT(inp); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { struct ip6_hdr *ip6; ip6 = (struct ip6_hdr *)ip_ptr; ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | (inp->inp_flow & IPV6_FLOWINFO_MASK); ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | (IPV6_VERSION & IPV6_VERSION_MASK); ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = htons(sizeof(struct tcphdr)); ip6->ip6_src = inp->in6p_laddr; ip6->ip6_dst = inp->in6p_faddr; } #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET { struct ip *ip; ip = (struct ip *)ip_ptr; ip->ip_v = IPVERSION; ip->ip_hl = 5; ip->ip_tos = inp->inp_ip_tos; ip->ip_len = 0; ip->ip_id = 0; ip->ip_off = 0; ip->ip_ttl = inp->inp_ip_ttl; ip->ip_sum = 0; ip->ip_p = IPPROTO_TCP; ip->ip_src = inp->inp_laddr; ip->ip_dst = inp->inp_faddr; } #endif /* INET */ th->th_sport = inp->inp_lport; th->th_dport = inp->inp_fport; th->th_seq = 0; th->th_ack = 0; th->th_x2 = 0; th->th_off = 5; th->th_flags = 0; th->th_win = 0; th->th_urp = 0; th->th_sum = 0; /* in_pseudo() is called later for ipv4 */ } /* * Create template to be used to send tcp packets on a connection. * Allocates an mbuf and fills in a skeletal tcp/ip header. The only * use for this function is in keepalives, which use tcp_respond. */ struct tcptemp * tcpip_maketemplate(struct inpcb *inp) { struct tcptemp *t; t = malloc(sizeof(*t), M_TEMP, M_NOWAIT); if (t == NULL) return (NULL); tcpip_fillheaders(inp, (void *)&t->tt_ipgen, (void *)&t->tt_t); return (t); } /* * Send a single message to the TCP at address specified by * the given TCP/IP header. If m == NULL, then we make a copy * of the tcpiphdr at th and send directly to the addressed host. * This is used to force keep alive messages out using the TCP * template for a connection. If flags are given then we send * a message back to the TCP which originated the segment th, * and discard the mbuf containing it and any other attached mbufs. * * In any case the ack and sequence number of the transmitted * segment are as specified by the parameters. * * NOTE: If m != NULL, then th must point to *inside* the mbuf. */ void tcp_respond(struct tcpcb *tp, void *ipgen, struct tcphdr *th, struct mbuf *m, tcp_seq ack, tcp_seq seq, int flags) { int tlen; int win = 0; struct ip *ip; struct tcphdr *nth; #ifdef INET6 struct ip6_hdr *ip6; int isipv6; #endif /* INET6 */ int ipflags = 0; struct inpcb *inp; KASSERT(tp != NULL || m != NULL, ("tcp_respond: tp and m both NULL")); #ifdef INET6 isipv6 = ((struct ip *)ipgen)->ip_v == (IPV6_VERSION >> 4); ip6 = ipgen; #endif /* INET6 */ ip = ipgen; if (tp != NULL) { inp = tp->t_inpcb; KASSERT(inp != NULL, ("tcp control block w/o inpcb")); INP_WLOCK_ASSERT(inp); } else inp = NULL; if (tp != NULL) { if (!(flags & TH_RST)) { win = sbspace(&inp->inp_socket->so_rcv); if (win > (long)TCP_MAXWIN << tp->rcv_scale) win = (long)TCP_MAXWIN << tp->rcv_scale; } } if (m == NULL) { m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) return; tlen = 0; m->m_data += max_linkhdr; #ifdef INET6 if (isipv6) { bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(struct ip6_hdr)); ip6 = mtod(m, struct ip6_hdr *); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); ip = mtod(m, struct ip *); nth = (struct tcphdr *)(ip + 1); } bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); flags = TH_ACK; } else { /* * reuse the mbuf. * XXX MRT We inherrit the FIB, which is lucky. */ m_freem(m->m_next); m->m_next = NULL; m->m_data = (caddr_t)ipgen; /* m_len is set later */ tlen = 0; #define xchg(a,b,type) { type t; t=a; a=b; b=t; } #ifdef INET6 if (isipv6) { xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, uint32_t); nth = (struct tcphdr *)(ip + 1); } if (th != nth) { /* * this is usually a case when an extension header * exists between the IPv6 header and the * TCP header. */ nth->th_sport = th->th_sport; nth->th_dport = th->th_dport; } xchg(nth->th_dport, nth->th_sport, uint16_t); #undef xchg } #ifdef INET6 if (isipv6) { ip6->ip6_flow = 0; ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); ip6->ip6_plen = htons(tlen - sizeof(*ip6)); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { tlen += sizeof (struct tcpiphdr); ip->ip_len = htons(tlen); ip->ip_ttl = V_ip_defttl; if (V_path_mtu_discovery) ip->ip_off |= htons(IP_DF); } #endif m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; #ifdef MAC if (inp != NULL) { /* * Packet is associated with a socket, so allow the * label of the response to reflect the socket label. */ INP_WLOCK_ASSERT(inp); mac_inpcb_create_mbuf(inp, m); } else { /* * Packet is not associated with a socket, so possibly * update the label in place. */ mac_netinet_tcp_reply(m); } #endif nth->th_seq = htonl(seq); nth->th_ack = htonl(ack); nth->th_x2 = 0; nth->th_off = sizeof (struct tcphdr) >> 2; nth->th_flags = flags; if (tp != NULL) nth->th_win = htons((u_short) (win >> tp->rcv_scale)); else nth->th_win = htons((u_short)win); nth->th_urp = 0; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); #ifdef INET6 if (isipv6) { m->m_pkthdr.csum_flags = CSUM_TCP_IPV6; nth->th_sum = in6_cksum_pseudo(ip6, tlen - sizeof(struct ip6_hdr), IPPROTO_TCP, 0); ip6->ip6_hlim = in6_selecthlim(tp != NULL ? tp->t_inpcb : NULL, NULL); } #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET { m->m_pkthdr.csum_flags = CSUM_TCP; nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p))); } #endif /* INET */ #ifdef TCPDEBUG if (tp == NULL || (inp->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); #endif TCP_PROBE3(debug__input, tp, th, mtod(m, const char *)); if (flags & TH_RST) TCP_PROBE5(accept__refused, NULL, NULL, mtod(m, const char *), tp, nth); TCP_PROBE5(send, NULL, tp, mtod(m, const char *), tp, nth); #ifdef INET6 if (isipv6) (void) ip6_output(m, NULL, NULL, ipflags, NULL, NULL, inp); #endif /* INET6 */ #if defined(INET) && defined(INET6) else #endif #ifdef INET (void) ip_output(m, NULL, NULL, ipflags, NULL, inp); #endif } /* * Create a new TCP control block, making an * empty reassembly queue and hooking it to the argument * protocol control block. The `inp' parameter must have * come from the zone allocator set up in tcp_init(). */ struct tcpcb * tcp_newtcpcb(struct inpcb *inp) { struct tcpcb_mem *tm; struct tcpcb *tp; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ tm = uma_zalloc(V_tcpcb_zone, M_NOWAIT | M_ZERO); if (tm == NULL) return (NULL); tp = &tm->tcb; /* Initialise cc_var struct for this tcpcb. */ tp->ccv = &tm->ccv; tp->ccv->type = IPPROTO_TCP; tp->ccv->ccvc.tcp = tp; rw_rlock(&tcp_function_lock); tp->t_fb = tcp_func_set_ptr; refcount_acquire(&tp->t_fb->tfb_refcnt); rw_runlock(&tcp_function_lock); if (tp->t_fb->tfb_tcp_fb_init) { (*tp->t_fb->tfb_tcp_fb_init)(tp); } /* * Use the current system default CC algorithm. */ CC_LIST_RLOCK(); KASSERT(!STAILQ_EMPTY(&cc_list), ("cc_list is empty!")); CC_ALGO(tp) = CC_DEFAULT(); CC_LIST_RUNLOCK(); if (CC_ALGO(tp)->cb_init != NULL) if (CC_ALGO(tp)->cb_init(tp->ccv) > 0) { if (tp->t_fb->tfb_tcp_fb_fini) (*tp->t_fb->tfb_tcp_fb_fini)(tp); refcount_release(&tp->t_fb->tfb_refcnt); uma_zfree(V_tcpcb_zone, tm); return (NULL); } tp->osd = &tm->osd; if (khelp_init_osd(HELPER_CLASS_TCP, tp->osd)) { if (tp->t_fb->tfb_tcp_fb_fini) (*tp->t_fb->tfb_tcp_fb_fini)(tp); refcount_release(&tp->t_fb->tfb_refcnt); uma_zfree(V_tcpcb_zone, tm); return (NULL); } #ifdef VIMAGE tp->t_vnet = inp->inp_vnet; #endif tp->t_timers = &tm->tt; /* LIST_INIT(&tp->t_segq); */ /* XXX covered by M_ZERO */ tp->t_maxseg = tp->t_maxopd = #ifdef INET6 isipv6 ? V_tcp_v6mssdflt : #endif /* INET6 */ V_tcp_mssdflt; /* Set up our timeouts. */ callout_init(&tp->t_timers->tt_rexmt, 1); callout_init(&tp->t_timers->tt_persist, 1); callout_init(&tp->t_timers->tt_keep, 1); callout_init(&tp->t_timers->tt_2msl, 1); callout_init(&tp->t_timers->tt_delack, 1); if (V_tcp_do_rfc1323) tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); if (V_tcp_do_sack) tp->t_flags |= TF_SACK_PERMIT; TAILQ_INIT(&tp->snd_holes); /* * The tcpcb will hold a reference on its inpcb until tcp_discardcb() * is called. */ in_pcbref(inp); /* Reference for tcpcb */ tp->t_inpcb = inp; /* * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives * reasonable initial retransmit time. */ tp->t_srtt = TCPTV_SRTTBASE; tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; tp->t_rttmin = tcp_rexmit_min; tp->t_rxtcur = TCPTV_RTOBASE; tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->t_rcvtime = ticks; /* * IPv4 TTL initialization is necessary for an IPv6 socket as well, * because the socket may be bound to an IPv6 wildcard address, * which may match an IPv4-mapped IPv6 address. */ inp->inp_ip_ttl = V_ip_defttl; inp->inp_ppcb = tp; #ifdef TCPPCAP /* * Init the TCP PCAP queues. */ tcp_pcap_tcpcb_init(tp); #endif return (tp); /* XXX */ } /* * Switch the congestion control algorithm back to NewReno for any active * control blocks using an algorithm which is about to go away. * This ensures the CC framework can allow the unload to proceed without leaving * any dangling pointers which would trigger a panic. * Returning non-zero would inform the CC framework that something went wrong * and it would be unsafe to allow the unload to proceed. However, there is no * way for this to occur with this implementation so we always return zero. */ int tcp_ccalgounload(struct cc_algo *unload_algo) { struct cc_algo *tmpalgo; struct inpcb *inp; struct tcpcb *tp; VNET_ITERATOR_DECL(vnet_iter); /* * Check all active control blocks across all network stacks and change * any that are using "unload_algo" back to NewReno. If "unload_algo" * requires cleanup code to be run, call it. */ VNET_LIST_RLOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); INP_INFO_WLOCK(&V_tcbinfo); /* * New connections already part way through being initialised * with the CC algo we're removing will not race with this code * because the INP_INFO_WLOCK is held during initialisation. We * therefore don't enter the loop below until the connection * list has stabilised. */ LIST_FOREACH(inp, &V_tcb, inp_list) { INP_WLOCK(inp); /* Important to skip tcptw structs. */ if (!(inp->inp_flags & INP_TIMEWAIT) && (tp = intotcpcb(inp)) != NULL) { /* * By holding INP_WLOCK here, we are assured * that the connection is not currently * executing inside the CC module's functions * i.e. it is safe to make the switch back to * NewReno. */ if (CC_ALGO(tp) == unload_algo) { tmpalgo = CC_ALGO(tp); /* NewReno does not require any init. */ CC_ALGO(tp) = &newreno_cc_algo; if (tmpalgo->cb_destroy != NULL) tmpalgo->cb_destroy(tp->ccv); } } INP_WUNLOCK(inp); } INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK(); return (0); } /* * Drop a TCP connection, reporting * the specified error. If connection is synchronized, * then send a RST to peer. */ struct tcpcb * tcp_drop(struct tcpcb *tp, int errno) { struct socket *so = tp->t_inpcb->inp_socket; INP_INFO_LOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(tp->t_inpcb); if (TCPS_HAVERCVDSYN(tp->t_state)) { tcp_state_change(tp, TCPS_CLOSED); (void) tp->t_fb->tfb_tcp_output(tp); TCPSTAT_INC(tcps_drops); } else TCPSTAT_INC(tcps_conndrops); if (errno == ETIMEDOUT && tp->t_softerror) errno = tp->t_softerror; so->so_error = errno; return (tcp_close(tp)); } void tcp_discardcb(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ int released; INP_WLOCK_ASSERT(inp); /* * Make sure that all of our timers are stopped before we delete the * PCB. * * If stopping a timer fails, we schedule a discard function in same * callout, and the last discard function called will take care of * deleting the tcpcb. */ tcp_timer_stop(tp, TT_REXMT); tcp_timer_stop(tp, TT_PERSIST); tcp_timer_stop(tp, TT_KEEP); tcp_timer_stop(tp, TT_2MSL); tcp_timer_stop(tp, TT_DELACK); if (tp->t_fb->tfb_tcp_timer_stop_all) { /* Call the stop-all function of the methods */ tp->t_fb->tfb_tcp_timer_stop_all(tp); } /* * If we got enough samples through the srtt filter, * save the rtt and rttvar in the routing entry. * 'Enough' is arbitrarily defined as 4 rtt samples. * 4 samples is enough for the srtt filter to converge * to within enough % of the correct value; fewer samples * and we could save a bogus rtt. The danger is not high * as tcp quickly recovers from everything. * XXX: Works very well but needs some more statistics! */ if (tp->t_rttupdated >= 4) { struct hc_metrics_lite metrics; u_long ssthresh; bzero(&metrics, sizeof(metrics)); /* * Update the ssthresh always when the conditions below * are satisfied. This gives us better new start value * for the congestion avoidance for new connections. * ssthresh is only set if packet loss occured on a session. * * XXXRW: 'so' may be NULL here, and/or socket buffer may be * being torn down. Ideally this code would not use 'so'. */ ssthresh = tp->snd_ssthresh; if (ssthresh != 0 && ssthresh < so->so_snd.sb_hiwat / 2) { /* * convert the limit from user data bytes to * packets then to packet data bytes. */ ssthresh = (ssthresh + tp->t_maxseg / 2) / tp->t_maxseg; if (ssthresh < 2) ssthresh = 2; ssthresh *= (u_long)(tp->t_maxseg + #ifdef INET6 (isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : #endif sizeof (struct tcpiphdr) #ifdef INET6 ) #endif ); } else ssthresh = 0; metrics.rmx_ssthresh = ssthresh; metrics.rmx_rtt = tp->t_srtt; metrics.rmx_rttvar = tp->t_rttvar; metrics.rmx_cwnd = tp->snd_cwnd; metrics.rmx_sendpipe = 0; metrics.rmx_recvpipe = 0; tcp_hc_update(&inp->inp_inc, &metrics); } /* free the reassembly queue, if any */ tcp_reass_flush(tp); #ifdef TCP_OFFLOAD /* Disconnect offload device, if any. */ if (tp->t_flags & TF_TOE) tcp_offload_detach(tp); #endif tcp_free_sackholes(tp); #ifdef TCPPCAP /* Free the TCP PCAP queues. */ tcp_pcap_drain(&(tp->t_inpkts)); tcp_pcap_drain(&(tp->t_outpkts)); #endif /* Allow the CC algorithm to clean up after itself. */ if (CC_ALGO(tp)->cb_destroy != NULL) CC_ALGO(tp)->cb_destroy(tp->ccv); khelp_destroy_osd(tp->osd); CC_ALGO(tp) = NULL; inp->inp_ppcb = NULL; if ((tp->t_timers->tt_flags & TT_MASK) == 0) { /* We own the last reference on tcpcb, let's free it. */ if ((tp->t_fb->tfb_tcp_timers_left) && (tp->t_fb->tfb_tcp_timers_left(tp))) { /* Some fb timers left running! */ return; } if (tp->t_fb->tfb_tcp_fb_fini) (*tp->t_fb->tfb_tcp_fb_fini)(tp); refcount_release(&tp->t_fb->tfb_refcnt); tp->t_inpcb = NULL; uma_zfree(V_tcpcb_zone, tp); released = in_pcbrele_wlocked(inp); KASSERT(!released, ("%s: inp %p should not have been released " "here", __func__, inp)); } } void tcp_timer_2msl_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_2MSL); } void tcp_timer_keep_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_KEEP); } void tcp_timer_persist_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_PERSIST); } void tcp_timer_rexmt_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_REXMT); } void tcp_timer_delack_discard(void *xtp) { tcp_timer_discard((struct tcpcb *)xtp, TT_DELACK); } void tcp_timer_discard(struct tcpcb *tp, uint32_t timer_type) { struct inpcb *inp; CURVNET_SET(tp->t_vnet); INP_INFO_RLOCK(&V_tcbinfo); inp = tp->t_inpcb; KASSERT(inp != NULL, ("%s: tp %p tp->t_inpcb == NULL", __func__, tp)); INP_WLOCK(inp); KASSERT((tp->t_timers->tt_flags & TT_STOPPED) != 0, ("%s: tcpcb has to be stopped here", __func__)); KASSERT((tp->t_timers->tt_flags & timer_type) != 0, ("%s: discard callout should be running", __func__)); tp->t_timers->tt_flags &= ~timer_type; if ((tp->t_timers->tt_flags & TT_MASK) == 0) { /* We own the last reference on this tcpcb, let's free it. */ if ((tp->t_fb->tfb_tcp_timers_left) && (tp->t_fb->tfb_tcp_timers_left(tp))) { /* Some fb timers left running! */ goto leave; } if (tp->t_fb->tfb_tcp_fb_fini) (*tp->t_fb->tfb_tcp_fb_fini)(tp); refcount_release(&tp->t_fb->tfb_refcnt); tp->t_inpcb = NULL; uma_zfree(V_tcpcb_zone, tp); if (in_pcbrele_wlocked(inp)) { INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); return; } } leave: INP_WUNLOCK(inp); INP_INFO_RUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } /* * Attempt to close a TCP control block, marking it as dropped, and freeing * the socket if we hold the only reference. */ struct tcpcb * tcp_close(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so; INP_INFO_LOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); #ifdef TCP_OFFLOAD if (tp->t_state == TCPS_LISTEN) tcp_offload_listen_stop(tp); #endif #ifdef TCP_RFC7413 /* * This releases the TFO pending counter resource for TFO listen * sockets as well as passively-created TFO sockets that transition * from SYN_RECEIVED to CLOSED. */ if (tp->t_tfo_pending) { tcp_fastopen_decrement_counter(tp->t_tfo_pending); tp->t_tfo_pending = NULL; } #endif in_pcbdrop(inp); TCPSTAT_INC(tcps_closed); KASSERT(inp->inp_socket != NULL, ("tcp_close: inp_socket NULL")); so = inp->inp_socket; soisdisconnected(so); if (inp->inp_flags & INP_SOCKREF) { KASSERT(so->so_state & SS_PROTOREF, ("tcp_close: !SS_PROTOREF")); inp->inp_flags &= ~INP_SOCKREF; INP_WUNLOCK(inp); ACCEPT_LOCK(); SOCK_LOCK(so); so->so_state &= ~SS_PROTOREF; sofree(so); return (NULL); } return (tp); } void tcp_drain(void) { VNET_ITERATOR_DECL(vnet_iter); if (!do_tcpdrain) return; VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); struct inpcb *inpb; struct tcpcb *tcpb; /* * Walk the tcpbs, if existing, and flush the reassembly queue, * if there is one... * XXX: The "Net/3" implementation doesn't imply that the TCP * reassembly queue should be flushed, but in a situation * where we're really low on mbufs, this is potentially * useful. */ INP_INFO_WLOCK(&V_tcbinfo); LIST_FOREACH(inpb, V_tcbinfo.ipi_listhead, inp_list) { if (inpb->inp_flags & INP_TIMEWAIT) continue; INP_WLOCK(inpb); if ((tcpb = intotcpcb(inpb)) != NULL) { tcp_reass_flush(tcpb); tcp_clean_sackreport(tcpb); } INP_WUNLOCK(inpb); } INP_INFO_WUNLOCK(&V_tcbinfo); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * Notify a tcp user of an asynchronous error; * store error as soft error, but wake up user * (for now, won't do anything until can select for soft error). * * Do not wake up user since there currently is no mechanism for * reporting soft errors (yet - a kqueue filter may be added). */ static struct inpcb * tcp_notify(struct inpcb *inp, int error) { struct tcpcb *tp; INP_INFO_LOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_notify: tp == NULL")); /* * Ignore some errors if we are hooked up. * If connection hasn't completed, has retransmitted several times, * and receives a second error, give up now. This is better * than waiting a long time to establish a connection that * can never complete. */ if (tp->t_state == TCPS_ESTABLISHED && (error == EHOSTUNREACH || error == ENETUNREACH || error == EHOSTDOWN)) { return (inp); } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && tp->t_softerror) { tp = tcp_drop(tp, error); if (tp != NULL) return (inp); else return (NULL); } else { tp->t_softerror = error; return (inp); } #if 0 wakeup( &so->so_timeo); sorwakeup(so); sowwakeup(so); #endif } static int tcp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, m, n, pcb_count; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == NULL) { n = V_tcbinfo.ipi_count + syncache_pcbcount(); n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xtcpcb); return (0); } if (req->newptr != NULL) return (EPERM); /* * OK, now we're committed to doing something. */ INP_LIST_RLOCK(&V_tcbinfo); gencnt = V_tcbinfo.ipi_gencnt; n = V_tcbinfo.ipi_count; INP_LIST_RUNLOCK(&V_tcbinfo); m = syncache_pcbcount(); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + (n + m) * sizeof(struct xtcpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n + m; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return (error); error = syncache_pcblist(req, m, &pcb_count); if (error) return (error); inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == NULL) return (ENOMEM); INP_INFO_WLOCK(&V_tcbinfo); for (inp = LIST_FIRST(V_tcbinfo.ipi_listhead), i = 0; inp != NULL && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt) { /* * XXX: This use of cr_cansee(), introduced with * TCP state changes, is not quite right, but for * now, better than nothing. */ if (inp->inp_flags & INP_TIMEWAIT) { if (intotw(inp) != NULL) error = cr_cansee(req->td->td_ucred, intotw(inp)->tw_cred); else error = EINVAL; /* Skip this inp. */ } else error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) { in_pcbref(inp); inp_list[i++] = inp; } } INP_WUNLOCK(inp); } INP_INFO_WUNLOCK(&V_tcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xtcpcb xt; void *inp_ppcb; bzero(&xt, sizeof(xt)); xt.xt_len = sizeof xt; /* XXX should avoid extra copy */ bcopy(inp, &xt.xt_inp, sizeof *inp); inp_ppcb = inp->inp_ppcb; if (inp_ppcb == NULL) bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); else if (inp->inp_flags & INP_TIMEWAIT) { bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); xt.xt_tp.t_state = TCPS_TIME_WAIT; } else { bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); if (xt.xt_tp.t_timers) tcp_timer_to_xtimer(&xt.xt_tp, xt.xt_tp.t_timers, &xt.xt_timer); } if (inp->inp_socket != NULL) sotoxsocket(inp->inp_socket, &xt.xt_socket); else { bzero(&xt.xt_socket, sizeof xt.xt_socket); xt.xt_socket.xso_protocol = IPPROTO_TCP; } xt.xt_inp.inp_gencnt = inp->inp_gencnt; INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xt, sizeof xt); } else INP_RUNLOCK(inp); } INP_INFO_RLOCK(&V_tcbinfo); for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (!in_pcbrele_rlocked(inp)) INP_RUNLOCK(inp); } INP_INFO_RUNLOCK(&V_tcbinfo); if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ INP_LIST_RLOCK(&V_tcbinfo); xig.xig_gen = V_tcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_tcbinfo.ipi_count + pcb_count; INP_LIST_RUNLOCK(&V_tcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); #ifdef INET static int tcp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error; error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); inp = in_pcblookup(&V_tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); } else error = ENOENT; if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp_getcred, "S,xucred", "Get the xucred of a TCP connection"); #endif /* INET */ #ifdef INET6 static int tcp6_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in6 addrs[2]; struct inpcb *inp; int error; #ifdef INET int mapped = 0; #endif error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 || (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) { return (error); } if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { #ifdef INET if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) mapped = 1; else #endif return (EINVAL); } #ifdef INET if (mapped == 1) inp = in_pcblookup(&V_tcbinfo, *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], addrs[1].sin6_port, *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], addrs[0].sin6_port, INPLOOKUP_RLOCKPCB, NULL); else #endif inp = in6_pcblookup(&V_tcbinfo, &addrs[1].sin6_addr, addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); } else error = ENOENT; if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection"); #endif /* INET6 */ #ifdef INET void tcp_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct ip *ip = vip; struct tcphdr *th; struct in_addr faddr; struct inpcb *inp; struct tcpcb *tp; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct icmp *icp; struct in_conninfo inc; tcp_seq icmp_tcp_seq; int mtu; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc_notify; else if (V_icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB || cmd == PRC_UNREACH_PORT || cmd == PRC_TIMXCEED_INTRANS) && ip) notify = tcp_drop_syn_sent; /* * Redirects don't need to be handled up here. */ else if (PRC_IS_REDIRECT(cmd)) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * XXX: We never get this from ICMP, otherwise it makes an * excellent DoS attack on machines with many connections. */ else if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip == NULL) { in_pcbnotifyall(&V_tcbinfo, faddr, inetctlerrmap[cmd], notify); return; } icp = (struct icmp *)((caddr_t)ip - offsetof(struct icmp, icmp_ip)); th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); INP_INFO_RLOCK(&V_tcbinfo); inp = in_pcblookup(&V_tcbinfo, faddr, th->th_dport, ip->ip_src, th->th_sport, INPLOOKUP_WLOCKPCB, NULL); if (inp != NULL) { if (!(inp->inp_flags & INP_TIMEWAIT) && !(inp->inp_flags & INP_DROPPED) && !(inp->inp_socket == NULL)) { icmp_tcp_seq = ntohl(th->th_seq); tp = intotcpcb(inp); if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) && SEQ_LT(icmp_tcp_seq, tp->snd_max)) { if (cmd == PRC_MSGSIZE) { /* * MTU discovery: * If we got a needfrag set the MTU * in the route to the suggested new * value (if given) and then notify. */ mtu = ntohs(icp->icmp_nextmtu); /* * If no alternative MTU was * proposed, try the next smaller * one. */ if (!mtu) mtu = ip_next_mtu( ntohs(ip->ip_len), 1); if (mtu < V_tcp_minmss + sizeof(struct tcpiphdr)) mtu = V_tcp_minmss + sizeof(struct tcpiphdr); /* * Only process the offered MTU if it * is smaller than the current one. */ if (mtu < tp->t_maxopd + sizeof(struct tcpiphdr)) { bzero(&inc, sizeof(inc)); inc.inc_faddr = faddr; inc.inc_fibnum = inp->inp_inc.inc_fibnum; tcp_hc_updatemtu(&inc, mtu); tcp_mtudisc(inp, mtu); } } else inp = (*notify)(inp, inetctlerrmap[cmd]); } } if (inp != NULL) INP_WUNLOCK(inp); } else { bzero(&inc, sizeof(inc)); inc.inc_fport = th->th_dport; inc.inc_lport = th->th_sport; inc.inc_faddr = faddr; inc.inc_laddr = ip->ip_src; syncache_unreach(&inc, th); } INP_INFO_RUNLOCK(&V_tcbinfo); } #endif /* INET */ #ifdef INET6 void tcp6_ctlinput(int cmd, struct sockaddr *sa, void *d) { struct tcphdr th; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct ip6_hdr *ip6; struct mbuf *m; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; int off; struct tcp_portonly { u_int16_t th_sport; u_int16_t th_dport; } *thp; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc_notify; else if (!PRC_IS_REDIRECT(cmd) && ((unsigned)cmd >= PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) return; /* if the parameter is from icmp6, decode it. */ if (d != NULL) { ip6cp = (struct ip6ctlparam *)d; m = ip6cp->ip6c_m; ip6 = ip6cp->ip6c_ip6; off = ip6cp->ip6c_off; sa6_src = ip6cp->ip6c_src; } else { m = NULL; ip6 = NULL; off = 0; /* fool gcc */ sa6_src = &sa6_any; } if (ip6 != NULL) { struct in_conninfo inc; /* * XXX: We assume that when IPV6 is non NULL, * M and OFF are valid. */ /* check if we can safely examine src and dst ports */ if (m->m_pkthdr.len < off + sizeof(*thp)) return; bzero(&th, sizeof(th)); m_copydata(m, off, sizeof(*thp), (caddr_t)&th); in6_pcbnotify(&V_tcbinfo, sa, th.th_dport, (struct sockaddr *)ip6cp->ip6c_src, th.th_sport, cmd, NULL, notify); bzero(&inc, sizeof(inc)); inc.inc_fport = th.th_dport; inc.inc_lport = th.th_sport; inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr; inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr; inc.inc_flags |= INC_ISIPV6; INP_INFO_RLOCK(&V_tcbinfo); syncache_unreach(&inc, &th); INP_INFO_RUNLOCK(&V_tcbinfo); } else in6_pcbnotify(&V_tcbinfo, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify); } #endif /* INET6 */ /* * Following is where TCP initial sequence number generation occurs. * * There are two places where we must use initial sequence numbers: * 1. In SYN-ACK packets. * 2. In SYN packets. * * All ISNs for SYN-ACK packets are generated by the syncache. See * tcp_syncache.c for details. * * The ISNs in SYN packets must be monotonic; TIME_WAIT recycling * depends on this property. In addition, these ISNs should be * unguessable so as to prevent connection hijacking. To satisfy * the requirements of this situation, the algorithm outlined in * RFC 1948 is used, with only small modifications. * * Implementation details: * * Time is based off the system timer, and is corrected so that it * increases by one megabyte per second. This allows for proper * recycling on high speed LANs while still leaving over an hour * before rollover. * * As reading the *exact* system time is too expensive to be done * whenever setting up a TCP connection, we increment the time * offset in two ways. First, a small random positive increment * is added to isn_offset for each connection that is set up. * Second, the function tcp_isn_tick fires once per clock tick * and increments isn_offset as necessary so that sequence numbers * are incremented at approximately ISN_BYTES_PER_SECOND. The * random positive increments serve only to ensure that the same * exact sequence number is never sent out twice (as could otherwise * happen when a port is recycled in less than the system tick * interval.) * * net.inet.tcp.isn_reseed_interval controls the number of seconds * between seeding of isn_secret. This is normally set to zero, * as reseeding should not be necessary. * * Locking of the global variables isn_secret, isn_last_reseed, isn_offset, * isn_offset_old, and isn_ctx is performed using the TCP pcbinfo lock. In * general, this means holding an exclusive (write) lock. */ #define ISN_BYTES_PER_SECOND 1048576 #define ISN_STATIC_INCREMENT 4096 #define ISN_RANDOM_INCREMENT (4096 - 1) static VNET_DEFINE(u_char, isn_secret[32]); static VNET_DEFINE(int, isn_last); static VNET_DEFINE(int, isn_last_reseed); static VNET_DEFINE(u_int32_t, isn_offset); static VNET_DEFINE(u_int32_t, isn_offset_old); #define V_isn_secret VNET(isn_secret) #define V_isn_last VNET(isn_last) #define V_isn_last_reseed VNET(isn_last_reseed) #define V_isn_offset VNET(isn_offset) #define V_isn_offset_old VNET(isn_offset_old) tcp_seq tcp_new_isn(struct tcpcb *tp) { MD5_CTX isn_ctx; u_int32_t md5_buffer[4]; tcp_seq new_isn; u_int32_t projected_offset; INP_WLOCK_ASSERT(tp->t_inpcb); ISN_LOCK(); /* Seed if this is the first use, reseed if requested. */ if ((V_isn_last_reseed == 0) || ((V_tcp_isn_reseed_interval > 0) && (((u_int)V_isn_last_reseed + (u_int)V_tcp_isn_reseed_interval*hz) < (u_int)ticks))) { read_random(&V_isn_secret, sizeof(V_isn_secret)); V_isn_last_reseed = ticks; } /* Compute the md5 hash and return the ISN. */ MD5Init(&isn_ctx); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short)); #ifdef INET6 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr, sizeof(struct in6_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr, sizeof(struct in6_addr)); } else #endif { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr, sizeof(struct in_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr, sizeof(struct in_addr)); } MD5Update(&isn_ctx, (u_char *) &V_isn_secret, sizeof(V_isn_secret)); MD5Final((u_char *) &md5_buffer, &isn_ctx); new_isn = (tcp_seq) md5_buffer[0]; V_isn_offset += ISN_STATIC_INCREMENT + (arc4random() & ISN_RANDOM_INCREMENT); if (ticks != V_isn_last) { projected_offset = V_isn_offset_old + ISN_BYTES_PER_SECOND / hz * (ticks - V_isn_last); if (SEQ_GT(projected_offset, V_isn_offset)) V_isn_offset = projected_offset; V_isn_offset_old = V_isn_offset; V_isn_last = ticks; } new_isn += V_isn_offset; ISN_UNLOCK(); return (new_isn); } /* * When a specific ICMP unreachable message is received and the * connection state is SYN-SENT, drop the connection. This behavior * is controlled by the icmp_may_rst sysctl. */ struct inpcb * tcp_drop_syn_sent(struct inpcb *inp, int errno) { struct tcpcb *tp; INP_INFO_RLOCK_ASSERT(&V_tcbinfo); INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return (inp); tp = intotcpcb(inp); if (tp->t_state != TCPS_SYN_SENT) return (inp); tp = tcp_drop(tp, errno); if (tp != NULL) return (inp); else return (NULL); } /* * When `need fragmentation' ICMP is received, update our idea of the MSS * based on the new value. Also nudge TCP to send something, since we * know the packet we just sent was dropped. * This duplicates some code in the tcp_mss() function in tcp_input.c. */ static struct inpcb * tcp_mtudisc_notify(struct inpcb *inp, int error) { tcp_mtudisc(inp, -1); return (inp); } static void tcp_mtudisc(struct inpcb *inp, int mtuoffer) { struct tcpcb *tp; struct socket *so; INP_WLOCK_ASSERT(inp); if ((inp->inp_flags & INP_TIMEWAIT) || (inp->inp_flags & INP_DROPPED)) return; tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_mtudisc: tp == NULL")); tcp_mss_update(tp, -1, mtuoffer, NULL, NULL); so = inp->inp_socket; SOCKBUF_LOCK(&so->so_snd); /* If the mss is larger than the socket buffer, decrease the mss. */ if (so->so_snd.sb_hiwat < tp->t_maxseg) tp->t_maxseg = so->so_snd.sb_hiwat; SOCKBUF_UNLOCK(&so->so_snd); TCPSTAT_INC(tcps_mturesent); tp->t_rtttime = 0; tp->snd_nxt = tp->snd_una; tcp_free_sackholes(tp); tp->snd_recover = tp->snd_max; if (tp->t_flags & TF_SACK_PERMIT) EXIT_FASTRECOVERY(tp->t_flags); tp->t_fb->tfb_tcp_output(tp); } #ifdef INET /* * Look-up the routing entry to the peer of this inpcb. If no route * is found and it cannot be allocated, then return 0. This routine * is called by TCP routines that access the rmx structure and by * tcp_mss_update to get the peer/interface MTU. */ u_long tcp_maxmtu(struct in_conninfo *inc, struct tcp_ifcap *cap) { struct route sro; struct sockaddr_in *dst; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer")); bzero(&sro, sizeof(sro)); if (inc->inc_faddr.s_addr != INADDR_ANY) { dst = (struct sockaddr_in *)&sro.ro_dst; dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = inc->inc_faddr; in_rtalloc_ign(&sro, 0, inc->inc_fibnum); } if (sro.ro_rt != NULL) { ifp = sro.ro_rt->rt_ifp; if (sro.ro_rt->rt_mtu == 0) maxmtu = ifp->if_mtu; else maxmtu = min(sro.ro_rt->rt_mtu, ifp->if_mtu); /* Report additional interface capabilities. */ if (cap != NULL) { if (ifp->if_capenable & IFCAP_TSO4 && ifp->if_hwassist & CSUM_TSO) { cap->ifcap |= CSUM_TSO; cap->tsomax = ifp->if_hw_tsomax; cap->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; cap->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; } } RTFREE(sro.ro_rt); } return (maxmtu); } #endif /* INET */ #ifdef INET6 u_long tcp_maxmtu6(struct in_conninfo *inc, struct tcp_ifcap *cap) { struct route_in6 sro6; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu6 with NULL in_conninfo pointer")); bzero(&sro6, sizeof(sro6)); if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) { sro6.ro_dst.sin6_family = AF_INET6; sro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); sro6.ro_dst.sin6_addr = inc->inc6_faddr; in6_rtalloc_ign(&sro6, 0, inc->inc_fibnum); } if (sro6.ro_rt != NULL) { ifp = sro6.ro_rt->rt_ifp; if (sro6.ro_rt->rt_mtu == 0) maxmtu = IN6_LINKMTU(sro6.ro_rt->rt_ifp); else maxmtu = min(sro6.ro_rt->rt_mtu, IN6_LINKMTU(sro6.ro_rt->rt_ifp)); /* Report additional interface capabilities. */ if (cap != NULL) { if (ifp->if_capenable & IFCAP_TSO6 && ifp->if_hwassist & CSUM_TSO) { cap->ifcap |= CSUM_TSO; cap->tsomax = ifp->if_hw_tsomax; cap->tsomaxsegcount = ifp->if_hw_tsomaxsegcount; cap->tsomaxsegsize = ifp->if_hw_tsomaxsegsize; } } RTFREE(sro6.ro_rt); } return (maxmtu); } #endif /* INET6 */ #ifdef IPSEC /* compute ESP/AH header size for TCP, including outer IP header. */ size_t ipsec_hdrsiz_tcp(struct tcpcb *tp) { struct inpcb *inp; struct mbuf *m; size_t hdrsiz; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; #endif struct tcphdr *th; if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL) || (!key_havesp(IPSEC_DIR_OUTBOUND))) return (0); m = m_gethdr(M_NOWAIT, MT_DATA); if (!m) return (0); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); tcpip_fillheaders(inp, ip6, th); hdrsiz = ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } else #endif /* INET6 */ { ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); tcpip_fillheaders(inp, ip, th); hdrsiz = ipsec_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } m_free(m); return (hdrsiz); } #endif /* IPSEC */ #ifdef TCP_SIGNATURE /* * Callback function invoked by m_apply() to digest TCP segment data * contained within an mbuf chain. */ static int tcp_signature_apply(void *fstate, void *data, u_int len) { MD5Update(fstate, (u_char *)data, len); return (0); } /* * XXX The key is retrieved from the system's PF_KEY SADB, by keying a * search with the destination IP address, and a 'magic SPI' to be * determined by the application. This is hardcoded elsewhere to 1179 */ struct secasvar * tcp_get_sav(struct mbuf *m, u_int direction) { union sockaddr_union dst; struct secasvar *sav; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; char ip6buf[INET6_ADDRSTRLEN]; #endif /* Extract the destination from the IP header in the mbuf. */ bzero(&dst, sizeof(union sockaddr_union)); ip = mtod(m, struct ip *); #ifdef INET6 ip6 = NULL; /* Make the compiler happy. */ #endif switch (ip->ip_v) { #ifdef INET case IPVERSION: dst.sa.sa_len = sizeof(struct sockaddr_in); dst.sa.sa_family = AF_INET; dst.sin.sin_addr = (direction == IPSEC_DIR_INBOUND) ? ip->ip_src : ip->ip_dst; break; #endif #ifdef INET6 case (IPV6_VERSION >> 4): ip6 = mtod(m, struct ip6_hdr *); dst.sa.sa_len = sizeof(struct sockaddr_in6); dst.sa.sa_family = AF_INET6; dst.sin6.sin6_addr = (direction == IPSEC_DIR_INBOUND) ? ip6->ip6_src : ip6->ip6_dst; break; #endif default: return (NULL); /* NOTREACHED */ break; } /* Look up an SADB entry which matches the address of the peer. */ sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); if (sav == NULL) { ipseclog((LOG_ERR, "%s: SADB lookup failed for %s\n", __func__, (ip->ip_v == IPVERSION) ? inet_ntoa(dst.sin.sin_addr) : #ifdef INET6 (ip->ip_v == (IPV6_VERSION >> 4)) ? ip6_sprintf(ip6buf, &dst.sin6.sin6_addr) : #endif "(unsupported)")); } return (sav); } /* * Compute TCP-MD5 hash of a TCP segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * sav pointer to security assosiation * * We do this over ip, tcphdr, segment data, and the key in the SADB. * When called from tcp_input(), we can be sure that th_sum has been * zeroed out and verified already. * * Releases reference to SADB key before return. * * Return 0 if successful, otherwise return -1. * */ int tcp_signature_do_compute(struct mbuf *m, int len, int optlen, u_char *buf, struct secasvar *sav) { #ifdef INET struct ippseudo ippseudo; #endif MD5_CTX ctx; int doff; struct ip *ip; #ifdef INET struct ipovly *ipovly; #endif struct tcphdr *th; #ifdef INET6 struct ip6_hdr *ip6; struct in6_addr in6; uint32_t plen; uint16_t nhdr; #endif u_short savecsum; KASSERT(m != NULL, ("NULL mbuf chain")); KASSERT(buf != NULL, ("NULL signature pointer")); /* Extract the destination from the IP header in the mbuf. */ ip = mtod(m, struct ip *); #ifdef INET6 ip6 = NULL; /* Make the compiler happy. */ #endif MD5Init(&ctx); /* * Step 1: Update MD5 hash with IP(v6) pseudo-header. * * XXX The ippseudo header MUST be digested in network byte order, * or else we'll fail the regression test. Assume all fields we've * been doing arithmetic on have been in host byte order. * XXX One cannot depend on ipovly->ih_len here. When called from * tcp_output(), the underlying ip_len member has not yet been set. */ switch (ip->ip_v) { #ifdef INET case IPVERSION: ipovly = (struct ipovly *)ip; ippseudo.ippseudo_src = ipovly->ih_src; ippseudo.ippseudo_dst = ipovly->ih_dst; ippseudo.ippseudo_pad = 0; ippseudo.ippseudo_p = IPPROTO_TCP; ippseudo.ippseudo_len = htons(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&ippseudo, sizeof(struct ippseudo)); th = (struct tcphdr *)((u_char *)ip + sizeof(struct ip)); doff = sizeof(struct ip) + sizeof(struct tcphdr) + optlen; break; #endif #ifdef INET6 /* * RFC 2385, 2.0 Proposal * For IPv6, the pseudo-header is as described in RFC 2460, namely the * 128-bit source IPv6 address, 128-bit destination IPv6 address, zero- * extended next header value (to form 32 bits), and 32-bit segment * length. * Note: Upper-Layer Packet Length comes before Next Header. */ case (IPV6_VERSION >> 4): in6 = ip6->ip6_src; in6_clearscope(&in6); MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr)); in6 = ip6->ip6_dst; in6_clearscope(&in6); MD5Update(&ctx, (char *)&in6, sizeof(struct in6_addr)); plen = htonl(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&plen, sizeof(uint32_t)); nhdr = 0; MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); nhdr = IPPROTO_TCP; MD5Update(&ctx, (char *)&nhdr, sizeof(uint8_t)); th = (struct tcphdr *)((u_char *)ip6 + sizeof(struct ip6_hdr)); doff = sizeof(struct ip6_hdr) + sizeof(struct tcphdr) + optlen; break; #endif default: KEY_FREESAV(&sav); return (-1); /* NOTREACHED */ break; } /* * Step 2: Update MD5 hash with TCP header, excluding options. * The TCP checksum must be set to zero. */ savecsum = th->th_sum; th->th_sum = 0; MD5Update(&ctx, (char *)th, sizeof(struct tcphdr)); th->th_sum = savecsum; /* * Step 3: Update MD5 hash with TCP segment data. * Use m_apply() to avoid an early m_pullup(). */ if (len > 0) m_apply(m, doff, len, tcp_signature_apply, &ctx); /* * Step 4: Update MD5 hash with shared secret. */ MD5Update(&ctx, sav->key_auth->key_data, _KEYLEN(sav->key_auth)); MD5Final(buf, &ctx); key_sa_recordxfer(sav, m); KEY_FREESAV(&sav); return (0); } /* * Compute TCP-MD5 hash of a TCP segment. (RFC2385) * * Return 0 if successful, otherwise return -1. */ int tcp_signature_compute(struct mbuf *m, int _unused, int len, int optlen, u_char *buf, u_int direction) { struct secasvar *sav; if ((sav = tcp_get_sav(m, direction)) == NULL) return (-1); return (tcp_signature_do_compute(m, len, optlen, buf, sav)); } /* * Verify the TCP-MD5 hash of a TCP segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND) * * Return 1 if successful, otherwise return 0. */ int tcp_signature_verify(struct mbuf *m, int off0, int tlen, int optlen, struct tcpopt *to, struct tcphdr *th, u_int tcpbflag) { char tmpdigest[TCP_SIGLEN]; if (tcp_sig_checksigs == 0) return (1); if ((tcpbflag & TF_SIGNATURE) == 0) { if ((to->to_flags & TOF_SIGNATURE) != 0) { /* * If this socket is not expecting signature but * the segment contains signature just fail. */ TCPSTAT_INC(tcps_sig_err_sigopt); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } /* Signature is not expected, and not present in segment. */ return (1); } /* * If this socket is expecting signature but the segment does not * contain any just fail. */ if ((to->to_flags & TOF_SIGNATURE) == 0) { TCPSTAT_INC(tcps_sig_err_nosigopt); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } if (tcp_signature_compute(m, off0, tlen, optlen, &tmpdigest[0], IPSEC_DIR_INBOUND) == -1) { TCPSTAT_INC(tcps_sig_err_buildsig); TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } if (bcmp(to->to_signature, &tmpdigest[0], TCP_SIGLEN) != 0) { TCPSTAT_INC(tcps_sig_rcvbadsig); return (0); } TCPSTAT_INC(tcps_sig_rcvgoodsig); return (1); } #endif /* TCP_SIGNATURE */ static int sysctl_drop(SYSCTL_HANDLER_ARGS) { /* addrs[0] is a foreign socket, addrs[1] is a local one. */ struct sockaddr_storage addrs[2]; struct inpcb *inp; struct tcpcb *tp; struct tcptw *tw; struct sockaddr_in *fin, *lin; #ifdef INET6 struct sockaddr_in6 *fin6, *lin6; #endif int error; inp = NULL; fin = lin = NULL; #ifdef INET6 fin6 = lin6 = NULL; #endif error = 0; if (req->oldptr != NULL || req->oldlen != 0) return (EINVAL); if (req->newptr == NULL) return (EPERM); if (req->newlen < sizeof(addrs)) return (ENOMEM); error = SYSCTL_IN(req, &addrs, sizeof(addrs)); if (error) return (error); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: fin6 = (struct sockaddr_in6 *)&addrs[0]; lin6 = (struct sockaddr_in6 *)&addrs[1]; if (fin6->sin6_len != sizeof(struct sockaddr_in6) || lin6->sin6_len != sizeof(struct sockaddr_in6)) return (EINVAL); if (IN6_IS_ADDR_V4MAPPED(&fin6->sin6_addr)) { if (!IN6_IS_ADDR_V4MAPPED(&lin6->sin6_addr)) return (EINVAL); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[0]); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[1]); fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; break; } error = sa6_embedscope(fin6, V_ip6_use_defzone); if (error) return (error); error = sa6_embedscope(lin6, V_ip6_use_defzone); if (error) return (error); break; #endif #ifdef INET case AF_INET: fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; if (fin->sin_len != sizeof(struct sockaddr_in) || lin->sin_len != sizeof(struct sockaddr_in)) return (EINVAL); break; #endif default: return (EINVAL); } INP_INFO_RLOCK(&V_tcbinfo); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: inp = in6_pcblookup(&V_tcbinfo, &fin6->sin6_addr, fin6->sin6_port, &lin6->sin6_addr, lin6->sin6_port, INPLOOKUP_WLOCKPCB, NULL); break; #endif #ifdef INET case AF_INET: inp = in_pcblookup(&V_tcbinfo, fin->sin_addr, fin->sin_port, lin->sin_addr, lin->sin_port, INPLOOKUP_WLOCKPCB, NULL); break; #endif } if (inp != NULL) { if (inp->inp_flags & INP_TIMEWAIT) { /* * XXXRW: There currently exists a state where an * inpcb is present, but its timewait state has been * discarded. For now, don't allow dropping of this * type of inpcb. */ tw = intotw(inp); if (tw != NULL) tcp_twclose(tw, 0); else INP_WUNLOCK(inp); } else if (!(inp->inp_flags & INP_DROPPED) && !(inp->inp_socket->so_options & SO_ACCEPTCONN)) { tp = intotcpcb(inp); tp = tcp_drop(tp, ECONNABORTED); if (tp != NULL) INP_WUNLOCK(inp); } else INP_WUNLOCK(inp); } else error = ESRCH; INP_INFO_RUNLOCK(&V_tcbinfo); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_DROP, drop, CTLFLAG_VNET | CTLTYPE_STRUCT | CTLFLAG_WR | CTLFLAG_SKIP, NULL, 0, sysctl_drop, "", "Drop TCP connection"); /* * Generate a standardized TCP log line for use throughout the * tcp subsystem. Memory allocation is done with M_NOWAIT to * allow use in the interrupt context. * * NB: The caller MUST free(s, M_TCPLOG) the returned string. * NB: The function may return NULL if memory allocation failed. * * Due to header inclusion and ordering limitations the struct ip * and ip6_hdr pointers have to be passed as void pointers. */ char * tcp_log_vain(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { /* Is logging enabled? */ if (tcp_log_in_vain == 0) return (NULL); return (tcp_log_addr(inc, th, ip4hdr, ip6hdr)); } char * tcp_log_addrs(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { /* Is logging enabled? */ if (tcp_log_debug == 0) return (NULL); return (tcp_log_addr(inc, th, ip4hdr, ip6hdr)); } static char * tcp_log_addr(struct in_conninfo *inc, struct tcphdr *th, void *ip4hdr, const void *ip6hdr) { char *s, *sp; size_t size; struct ip *ip; #ifdef INET6 const struct ip6_hdr *ip6; ip6 = (const struct ip6_hdr *)ip6hdr; #endif /* INET6 */ ip = (struct ip *)ip4hdr; /* * The log line looks like this: * "TCP: [1.2.3.4]:50332 to [1.2.3.4]:80 tcpflags 0x2" */ size = sizeof("TCP: []:12345 to []:12345 tcpflags 0x2<>") + sizeof(PRINT_TH_FLAGS) + 1 + #ifdef INET6 2 * INET6_ADDRSTRLEN; #else 2 * INET_ADDRSTRLEN; #endif /* INET6 */ s = malloc(size, M_TCPLOG, M_ZERO|M_NOWAIT); if (s == NULL) return (NULL); strcat(s, "TCP: ["); sp = s + strlen(s); if (inc && ((inc->inc_flags & INC_ISIPV6) == 0)) { inet_ntoa_r(inc->inc_faddr, sp); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(inc->inc_fport)); sp = s + strlen(s); inet_ntoa_r(inc->inc_laddr, sp); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(inc->inc_lport)); #ifdef INET6 } else if (inc) { ip6_sprintf(sp, &inc->inc6_faddr); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(inc->inc_fport)); sp = s + strlen(s); ip6_sprintf(sp, &inc->inc6_laddr); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(inc->inc_lport)); } else if (ip6 && th) { ip6_sprintf(sp, &ip6->ip6_src); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(th->th_sport)); sp = s + strlen(s); ip6_sprintf(sp, &ip6->ip6_dst); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(th->th_dport)); #endif /* INET6 */ #ifdef INET } else if (ip && th) { inet_ntoa_r(ip->ip_src, sp); sp = s + strlen(s); sprintf(sp, "]:%i to [", ntohs(th->th_sport)); sp = s + strlen(s); inet_ntoa_r(ip->ip_dst, sp); sp = s + strlen(s); sprintf(sp, "]:%i", ntohs(th->th_dport)); #endif /* INET */ } else { free(s, M_TCPLOG); return (NULL); } sp = s + strlen(s); if (th) sprintf(sp, " tcpflags 0x%b", th->th_flags, PRINT_TH_FLAGS); if (*(s + size - 1) != '\0') panic("%s: string too long", __func__); return (s); } /* * A subroutine which makes it easy to track TCP state changes with DTrace. * This function shouldn't be called for t_state initializations that don't * correspond to actual TCP state transitions. */ void tcp_state_change(struct tcpcb *tp, int newstate) { #if defined(KDTRACE_HOOKS) int pstate = tp->t_state; #endif tp->t_state = newstate; TCP_PROBE6(state__change, NULL, tp, NULL, tp, NULL, pstate); } Index: projects/vnet/sys/netinet/tcp_var.h =================================================================== --- projects/vnet/sys/netinet/tcp_var.h (revision 294565) +++ projects/vnet/sys/netinet/tcp_var.h (revision 294566) @@ -1,866 +1,863 @@ /*- * Copyright (c) 1982, 1986, 1993, 1994, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_var.h 8.4 (Berkeley) 5/24/95 * $FreeBSD$ */ #ifndef _NETINET_TCP_VAR_H_ #define _NETINET_TCP_VAR_H_ #include #ifdef _KERNEL #include #include /* * Kernel variables for tcp. */ VNET_DECLARE(int, tcp_do_rfc1323); #define V_tcp_do_rfc1323 VNET(tcp_do_rfc1323) #endif /* _KERNEL */ /* TCP segment queue entry */ struct tseg_qent { LIST_ENTRY(tseg_qent) tqe_q; int tqe_len; /* TCP segment data length */ struct tcphdr *tqe_th; /* a pointer to tcp header */ struct mbuf *tqe_m; /* mbuf contains packet */ }; LIST_HEAD(tsegqe_head, tseg_qent); struct sackblk { tcp_seq start; /* start seq no. of sack block */ tcp_seq end; /* end seq no. */ }; struct sackhole { tcp_seq start; /* start seq no. of hole */ tcp_seq end; /* end seq no. */ tcp_seq rxmit; /* next seq. no in hole to be retransmitted */ TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */ }; struct sackhint { struct sackhole *nexthole; int sack_bytes_rexmit; tcp_seq last_sack_ack; /* Most recent/largest sacked ack */ int ispare; /* explicit pad for 64bit alignment */ int sacked_bytes; /* * Total sacked bytes reported by the * receiver via sack option */ uint32_t _pad1[1]; /* TBD */ uint64_t _pad[1]; /* TBD */ }; struct tcptemp { u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */ struct tcphdr tt_t; }; #define tcp6cb tcpcb /* for KAME src sync over BSD*'s */ /* * TODO: We yet need to brave plowing in * to tcp_input() and the pru_usrreq() block. * Right now these go to the old standards which * are somewhat ok, but in the long term may * need to be changed. If we do tackle tcp_input() * then we need to get rid of the tcp_do_segment() * function below. */ /* Flags for tcp functions */ #define TCP_FUNC_BEING_REMOVED 0x01 /* Can no longer be referenced */ struct tcpcb; struct inpcb; struct sockopt; struct socket; struct tcp_function_block { char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX]; int (*tfb_tcp_output)(struct tcpcb *); void (*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *, struct socket *, struct tcpcb *, int, int, uint8_t, int); int (*tfb_tcp_ctloutput)(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp); /* Optional memory allocation/free routine */ void (*tfb_tcp_fb_init)(struct tcpcb *); void (*tfb_tcp_fb_fini)(struct tcpcb *); /* Optional timers, must define all if you define one */ int (*tfb_tcp_timer_stop_all)(struct tcpcb *); int (*tfb_tcp_timers_left)(struct tcpcb *); void (*tfb_tcp_timer_activate)(struct tcpcb *, uint32_t, u_int); int (*tfb_tcp_timer_active)(struct tcpcb *, uint32_t); void (*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t); volatile uint32_t tfb_refcnt; uint32_t tfb_flags; }; struct tcp_function { TAILQ_ENTRY(tcp_function) tf_next; struct tcp_function_block *tf_fb; }; TAILQ_HEAD(tcp_funchead, tcp_function); /* * Tcp control block, one per tcp; fields: * Organized for 16 byte cacheline efficiency. */ struct tcpcb { struct tsegqe_head t_segq; /* segment reassembly queue */ void *t_pspare[2]; /* new reassembly queue */ int t_segqlen; /* segment reassembly queue length */ int t_dupacks; /* consecutive dup acks recd */ struct tcp_timer *t_timers; /* All the TCP timers in one struct */ struct inpcb *t_inpcb; /* back pointer to internet pcb */ int t_state; /* state of this connection */ u_int t_flags; struct vnet *t_vnet; /* back pointer to parent vnet */ tcp_seq snd_una; /* sent but unacknowledged */ tcp_seq snd_max; /* highest sequence number sent; * used to recognize retransmits */ tcp_seq snd_nxt; /* send next */ tcp_seq snd_up; /* send urgent pointer */ tcp_seq snd_wl1; /* window update seg seq number */ tcp_seq snd_wl2; /* window update seg ack number */ tcp_seq iss; /* initial send sequence number */ tcp_seq irs; /* initial receive sequence number */ tcp_seq rcv_nxt; /* receive next */ tcp_seq rcv_adv; /* advertised window */ u_long rcv_wnd; /* receive window */ tcp_seq rcv_up; /* receive urgent pointer */ u_long snd_wnd; /* send window */ u_long snd_cwnd; /* congestion-controlled window */ u_long snd_spare1; /* unused */ u_long snd_ssthresh; /* snd_cwnd size threshold for * for slow start exponential to * linear switch */ u_long snd_spare2; /* unused */ tcp_seq snd_recover; /* for use in NewReno Fast Recovery */ u_int t_maxopd; /* mss plus options */ u_int t_rcvtime; /* inactivity time */ u_int t_starttime; /* time connection was established */ u_int t_rtttime; /* RTT measurement start time */ tcp_seq t_rtseq; /* sequence number being timed */ u_int t_bw_spare1; /* unused */ tcp_seq t_bw_spare2; /* unused */ int t_rxtcur; /* current retransmit value (ticks) */ u_int t_maxseg; /* maximum segment size */ int t_srtt; /* smoothed round-trip time */ int t_rttvar; /* variance in round-trip time */ int t_rxtshift; /* log(2) of rexmt exp. backoff */ u_int t_rttmin; /* minimum rtt allowed */ u_int t_rttbest; /* best rtt we've seen */ u_long t_rttupdated; /* number of times rtt sampled */ u_long max_sndwnd; /* largest window peer has offered */ int t_softerror; /* possible error not yet reported */ /* out-of-band data */ char t_oobflags; /* have some */ char t_iobc; /* input character */ /* RFC 1323 variables */ u_char snd_scale; /* window scaling for send window */ u_char rcv_scale; /* window scaling for recv window */ u_char request_r_scale; /* pending window scaling */ u_int32_t ts_recent; /* timestamp echo data */ u_int ts_recent_age; /* when last updated */ u_int32_t ts_offset; /* our timestamp offset */ tcp_seq last_ack_sent; /* experimental */ u_long snd_cwnd_prev; /* cwnd prior to retransmit */ u_long snd_ssthresh_prev; /* ssthresh prior to retransmit */ tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */ int t_sndzerowin; /* zero-window updates sent */ u_int t_badrxtwin; /* window for retransmit recovery */ u_char snd_limited; /* segments limited transmitted */ /* SACK related state */ int snd_numholes; /* number of holes seen by sender */ TAILQ_HEAD(sackhole_head, sackhole) snd_holes; /* SACK scoreboard (sorted) */ tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/ int rcv_numsacks; /* # distinct sack blks present */ struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */ tcp_seq sack_newdata; /* New data xmitted in this recovery episode starts at this seq number */ struct sackhint sackhint; /* SACK scoreboard hint */ int t_rttlow; /* smallest observerved RTT */ u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */ int rfbuf_cnt; /* recv buffer autoscaling byte count */ struct toedev *tod; /* toedev handling this connection */ int t_sndrexmitpack; /* retransmit packets sent */ int t_rcvoopack; /* out-of-order packets received */ void *t_toe; /* TOE pcb pointer */ int t_bytes_acked; /* # bytes acked during current RTT */ struct cc_algo *cc_algo; /* congestion control algorithm */ struct cc_var *ccv; /* congestion control specific vars */ struct osd *osd; /* storage for Khelp module data */ u_int t_keepinit; /* time to establish connection */ u_int t_keepidle; /* time before keepalive probes begin */ u_int t_keepintvl; /* interval between keepalives */ u_int t_keepcnt; /* number of keepalives before close */ u_int t_tsomax; /* TSO total burst length limit in bytes */ u_int t_tsomaxsegcount; /* TSO maximum segment count */ u_int t_tsomaxsegsize; /* TSO maximum segment size in bytes */ u_int t_pmtud_saved_maxopd; /* pre-blackhole MSS */ u_int t_flags2; /* More tcpcb flags storage */ #if defined(_KERNEL) && defined(TCP_RFC7413) uint32_t t_ispare[6]; /* 5 UTO, 1 TBD */ uint64_t t_tfo_cookie; /* TCP Fast Open cookie */ #else uint32_t t_ispare[8]; /* 5 UTO, 3 TBD */ #endif struct tcp_function_block *t_fb;/* TCP function call block */ void *t_fb_ptr; /* Pointer to t_fb specific data */ #if defined(_KERNEL) && defined(TCP_RFC7413) unsigned int *t_tfo_pending; /* TCP Fast Open pending counter */ void *t_pspare2[1]; /* 1 TCP_SIGNATURE */ #else void *t_pspare2[2]; /* 1 TCP_SIGNATURE, 1 TBD */ #endif #if defined(_KERNEL) && defined(TCPPCAP) struct mbufq t_inpkts; /* List of saved input packets. */ struct mbufq t_outpkts; /* List of saved output packets. */ #ifdef _LP64 uint64_t _pad[0]; /* all used! */ #else uint64_t _pad[2]; /* 2 are available */ #endif /* _LP64 */ #else uint64_t _pad[6]; #endif /* defined(_KERNEL) && defined(TCPPCAP) */ }; /* * Flags and utility macros for the t_flags field. */ #define TF_ACKNOW 0x000001 /* ack peer immediately */ #define TF_DELACK 0x000002 /* ack, but try to delay it */ #define TF_NODELAY 0x000004 /* don't delay packets to coalesce */ #define TF_NOOPT 0x000008 /* don't use tcp options */ #define TF_SENTFIN 0x000010 /* have sent FIN */ #define TF_REQ_SCALE 0x000020 /* have/will request window scaling */ #define TF_RCVD_SCALE 0x000040 /* other side has requested scaling */ #define TF_REQ_TSTMP 0x000080 /* have/will request timestamps */ #define TF_RCVD_TSTMP 0x000100 /* a timestamp was received in SYN */ #define TF_SACK_PERMIT 0x000200 /* other side said I could SACK */ #define TF_NEEDSYN 0x000400 /* send SYN (implicit state) */ #define TF_NEEDFIN 0x000800 /* send FIN (implicit state) */ #define TF_NOPUSH 0x001000 /* don't push */ #define TF_PREVVALID 0x002000 /* saved values for bad rxmit valid */ #define TF_MORETOCOME 0x010000 /* More data to be appended to sock */ #define TF_LQ_OVERFLOW 0x020000 /* listen queue overflow */ #define TF_LASTIDLE 0x040000 /* connection was previously idle */ #define TF_RXWIN0SENT 0x080000 /* sent a receiver win 0 in response */ #define TF_FASTRECOVERY 0x100000 /* in NewReno Fast Recovery */ #define TF_WASFRECOVERY 0x200000 /* was in NewReno Fast Recovery */ #define TF_SIGNATURE 0x400000 /* require MD5 digests (RFC2385) */ #define TF_FORCEDATA 0x800000 /* force out a byte */ #define TF_TSO 0x1000000 /* TSO enabled on this connection */ #define TF_TOE 0x2000000 /* this connection is offloaded */ #define TF_ECN_PERMIT 0x4000000 /* connection ECN-ready */ #define TF_ECN_SND_CWR 0x8000000 /* ECN CWR in queue */ #define TF_ECN_SND_ECE 0x10000000 /* ECN ECE in queue */ #define TF_CONGRECOVERY 0x20000000 /* congestion recovery mode */ #define TF_WASCRECOVERY 0x40000000 /* was in congestion recovery */ #define TF_FASTOPEN 0x80000000 /* TCP Fast Open indication */ #define IN_FASTRECOVERY(t_flags) (t_flags & TF_FASTRECOVERY) #define ENTER_FASTRECOVERY(t_flags) t_flags |= TF_FASTRECOVERY #define EXIT_FASTRECOVERY(t_flags) t_flags &= ~TF_FASTRECOVERY #define IN_CONGRECOVERY(t_flags) (t_flags & TF_CONGRECOVERY) #define ENTER_CONGRECOVERY(t_flags) t_flags |= TF_CONGRECOVERY #define EXIT_CONGRECOVERY(t_flags) t_flags &= ~TF_CONGRECOVERY #define IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY)) #define ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY) #define EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY) #define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una) /* * Flags for the t_oobflags field. */ #define TCPOOB_HAVEDATA 0x01 #define TCPOOB_HADDATA 0x02 #ifdef TCP_SIGNATURE /* * Defines which are needed by the xform_tcp module and tcp_[in|out]put * for SADB verification and lookup. */ #define TCP_SIGLEN 16 /* length of computed digest in bytes */ #define TCP_KEYLEN_MIN 1 /* minimum length of TCP-MD5 key */ #define TCP_KEYLEN_MAX 80 /* maximum length of TCP-MD5 key */ /* * Only a single SA per host may be specified at this time. An SPI is * needed in order for the KEY_ALLOCSA() lookup to work. */ #define TCP_SIG_SPI 0x1000 #endif /* TCP_SIGNATURE */ /* * Flags for PLPMTU handling, t_flags2 */ #define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */ #define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */ #define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */ /* * Structure to hold TCP options that are only used during segment * processing (in tcp_input), but not held in the tcpcb. * It's basically used to reduce the number of parameters * to tcp_dooptions and tcp_addoptions. * The binary order of the to_flags is relevant for packing of the * options in tcp_addoptions. */ struct tcpopt { u_int64_t to_flags; /* which options are present */ #define TOF_MSS 0x0001 /* maximum segment size */ #define TOF_SCALE 0x0002 /* window scaling */ #define TOF_SACKPERM 0x0004 /* SACK permitted */ #define TOF_TS 0x0010 /* timestamp */ #define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */ #define TOF_SACK 0x0080 /* Peer sent SACK option */ #define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */ #define TOF_MAXOPT 0x0200 u_int32_t to_tsval; /* new timestamp */ u_int32_t to_tsecr; /* reflected timestamp */ u_char *to_sacks; /* pointer to the first SACK blocks */ u_char *to_signature; /* pointer to the TCP-MD5 signature */ u_char *to_tfo_cookie; /* pointer to the TFO cookie */ u_int16_t to_mss; /* maximum segment size */ u_int8_t to_wscale; /* window scaling */ u_int8_t to_nsacks; /* number of SACK blocks */ u_int8_t to_tfo_len; /* TFO cookie length */ u_int32_t to_spare; /* UTO */ }; /* * Flags for tcp_dooptions. */ #define TO_SYN 0x01 /* parse SYN-only options */ struct hc_metrics_lite { /* must stay in sync with hc_metrics */ u_long rmx_mtu; /* MTU for this path */ u_long rmx_ssthresh; /* outbound gateway buffer limit */ u_long rmx_rtt; /* estimated round trip time */ u_long rmx_rttvar; /* estimated rtt variance */ u_long rmx_cwnd; /* congestion window */ u_long rmx_sendpipe; /* outbound delay-bandwidth product */ u_long rmx_recvpipe; /* inbound delay-bandwidth product */ }; /* * Used by tcp_maxmtu() to communicate interface specific features * and limits at the time of connection setup. */ struct tcp_ifcap { int ifcap; u_int tsomax; u_int tsomaxsegcount; u_int tsomaxsegsize; }; #ifndef _NETINET_IN_PCB_H_ struct in_conninfo; #endif /* _NETINET_IN_PCB_H_ */ struct tcptw { struct inpcb *tw_inpcb; /* XXX back pointer to internet pcb */ tcp_seq snd_nxt; tcp_seq rcv_nxt; tcp_seq iss; tcp_seq irs; u_short last_win; /* cached window value */ u_short tw_so_options; /* copy of so_options */ struct ucred *tw_cred; /* user credentials */ u_int32_t t_recent; u_int32_t ts_offset; /* our timestamp offset */ u_int t_starttime; int tw_time; TAILQ_ENTRY(tcptw) tw_2msl; void *tw_pspare; /* TCP_SIGNATURE */ u_int *tw_spare; /* TCP_SIGNATURE */ }; #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb) #define intotw(ip) ((struct tcptw *)(ip)->inp_ppcb) #define sototcpcb(so) (intotcpcb(sotoinpcb(so))) /* * The smoothed round-trip time and estimated variance * are stored as fixed point numbers scaled by the values below. * For convenience, these scales are also used in smoothing the average * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed). * With these scales, srtt has 3 bits to the right of the binary point, * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the * binary point, and is smoothed with an ALPHA of 0.75. */ #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */ #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */ #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */ #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */ #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */ /* * The initial retransmission should happen at rtt + 4 * rttvar. * Because of the way we do the smoothing, srtt and rttvar * will each average +1/2 tick of bias. When we compute * the retransmit timer, we want 1/2 tick of rounding and * 1 extra tick because of +-1/2 tick uncertainty in the * firing of the timer. The bias will give us exactly the * 1.5 tick we need. But, because the bias is * statistical, we have to test that we don't drop below * the minimum feasible timer (which is 2 ticks). * This version of the macro adapted from a paper by Lawrence * Brakmo and Larry Peterson which outlines a problem caused * by insufficient precision in the original implementation, * which results in inappropriately large RTO values for very * fast networks. */ #define TCP_REXMTVAL(tp) \ max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \ + (tp)->t_rttvar) >> TCP_DELTA_SHIFT) /* * TCP statistics. * Many of these should be kept per connection, * but that's inconvenient at the moment. */ struct tcpstat { uint64_t tcps_connattempt; /* connections initiated */ uint64_t tcps_accepts; /* connections accepted */ uint64_t tcps_connects; /* connections established */ uint64_t tcps_drops; /* connections dropped */ uint64_t tcps_conndrops; /* embryonic connections dropped */ uint64_t tcps_minmssdrops; /* average minmss too low drops */ uint64_t tcps_closed; /* conn. closed (includes drops) */ uint64_t tcps_segstimed; /* segs where we tried to get rtt */ uint64_t tcps_rttupdated; /* times we succeeded */ uint64_t tcps_delack; /* delayed acks sent */ uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */ uint64_t tcps_rexmttimeo; /* retransmit timeouts */ uint64_t tcps_persisttimeo; /* persist timeouts */ uint64_t tcps_keeptimeo; /* keepalive timeouts */ uint64_t tcps_keepprobe; /* keepalive probes sent */ uint64_t tcps_keepdrops; /* connections dropped in keepalive */ uint64_t tcps_sndtotal; /* total packets sent */ uint64_t tcps_sndpack; /* data packets sent */ uint64_t tcps_sndbyte; /* data bytes sent */ uint64_t tcps_sndrexmitpack; /* data packets retransmitted */ uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */ uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */ uint64_t tcps_sndacks; /* ack-only packets sent */ uint64_t tcps_sndprobe; /* window probes sent */ uint64_t tcps_sndurg; /* packets sent with URG only */ uint64_t tcps_sndwinup; /* window update-only packets sent */ uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */ uint64_t tcps_rcvtotal; /* total packets received */ uint64_t tcps_rcvpack; /* packets received in sequence */ uint64_t tcps_rcvbyte; /* bytes received in sequence */ uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */ uint64_t tcps_rcvbadoff; /* packets received with bad offset */ uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */ uint64_t tcps_rcvshort; /* packets received too short */ uint64_t tcps_rcvduppack; /* duplicate-only packets received */ uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */ uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */ uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */ uint64_t tcps_rcvoopack; /* out-of-order packets received */ uint64_t tcps_rcvoobyte; /* out-of-order bytes received */ uint64_t tcps_rcvpackafterwin; /* packets with data after window */ uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */ uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */ uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */ uint64_t tcps_rcvdupack; /* rcvd duplicate acks */ uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */ uint64_t tcps_rcvackpack; /* rcvd ack packets */ uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */ uint64_t tcps_rcvwinupd; /* rcvd window update packets */ uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */ uint64_t tcps_predack; /* times hdr predict ok for acks */ uint64_t tcps_preddat; /* times hdr predict ok for data pkts */ uint64_t tcps_pcbcachemiss; uint64_t tcps_cachedrtt; /* times cached RTT in route updated */ uint64_t tcps_cachedrttvar; /* times cached rttvar updated */ uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */ uint64_t tcps_usedrtt; /* times RTT initialized from route */ uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */ uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/ uint64_t tcps_persistdrop; /* timeout in persist state */ uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */ uint64_t tcps_mturesent; /* resends due to MTU discovery */ uint64_t tcps_listendrop; /* listen queue overflows */ uint64_t tcps_badrst; /* ignored RSTs in the window */ uint64_t tcps_sc_added; /* entry added to syncache */ uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */ uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */ uint64_t tcps_sc_dropped; /* could not reply to packet */ uint64_t tcps_sc_completed; /* successful extraction of entry */ uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */ uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */ uint64_t tcps_sc_reset; /* RST removed entry from syncache */ uint64_t tcps_sc_stale; /* timed out or listen socket gone */ uint64_t tcps_sc_aborted; /* syncache entry aborted */ uint64_t tcps_sc_badack; /* removed due to bad ACK */ uint64_t tcps_sc_unreach; /* ICMP unreachable received */ uint64_t tcps_sc_zonefail; /* zalloc() failed */ uint64_t tcps_sc_sendcookie; /* SYN cookie sent */ uint64_t tcps_sc_recvcookie; /* SYN cookie received */ uint64_t tcps_hc_added; /* entry added to hostcache */ uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */ uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */ /* SACK related stats */ uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */ uint64_t tcps_sack_rexmits; /* SACK rexmit segments */ uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */ uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */ uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */ uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */ /* ECN related stats */ uint64_t tcps_ecn_ce; /* ECN Congestion Experienced */ uint64_t tcps_ecn_ect0; /* ECN Capable Transport */ uint64_t tcps_ecn_ect1; /* ECN Capable Transport */ uint64_t tcps_ecn_shs; /* ECN successful handshakes */ uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */ /* TCP_SIGNATURE related stats */ uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */ uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */ uint64_t tcps_sig_err_buildsig; /* Mismatching signature received */ uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */ uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */ uint64_t _pad[12]; /* 6 UTO, 6 TBD */ }; #define tcps_rcvmemdrop tcps_rcvreassfull /* compat */ #ifdef _KERNEL #define TI_UNLOCKED 1 #define TI_RLOCKED 2 #include VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */ /* * In-kernel consumers can use these accessor macros directly to update * stats. */ #define TCPSTAT_ADD(name, val) \ VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val)) #define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1) /* * Kernel module consumers must use this accessor macro. */ void kmod_tcpstat_inc(int statnum); #define KMOD_TCPSTAT_INC(name) \ kmod_tcpstat_inc(offsetof(struct tcpstat, name) / sizeof(uint64_t)) /* * TCP specific helper hook point identifiers. */ #define HHOOK_TCP_EST_IN 0 #define HHOOK_TCP_EST_OUT 1 #define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT struct tcp_hhook_data { struct tcpcb *tp; struct tcphdr *th; struct tcpopt *to; long len; int tso; tcp_seq curack; }; #endif /* * TCB structure exported to user-land via sysctl(3). * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been * included. Not all of our clients do. */ #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_) struct xtcp_timer { int tt_rexmt; /* retransmit timer */ int tt_persist; /* retransmit persistence */ int tt_keep; /* keepalive */ int tt_2msl; /* 2*msl TIME_WAIT timer */ int tt_delack; /* delayed ACK timer */ int t_rcvtime; /* Time since last packet received */ }; struct xtcpcb { size_t xt_len; struct inpcb xt_inp; struct tcpcb xt_tp; struct xsocket xt_socket; struct xtcp_timer xt_timer; u_quad_t xt_alignment_hack; }; #endif /* * Identifiers for TCP sysctl nodes */ #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */ #define TCPCTL_MSSDFLT 3 /* MSS default */ #define TCPCTL_STATS 4 /* statistics (read-only) */ #define TCPCTL_RTTDFLT 5 /* default RTT estimate */ #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */ #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */ #define TCPCTL_SENDSPACE 8 /* send buffer space */ #define TCPCTL_RECVSPACE 9 /* receive buffer space */ #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */ #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */ #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */ #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */ #define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */ #define TCPCTL_DROP 15 /* drop tcp connection */ #ifdef _KERNEL #ifdef SYSCTL_DECL SYSCTL_DECL(_net_inet_tcp); SYSCTL_DECL(_net_inet_tcp_sack); MALLOC_DECLARE(M_TCPLOG); #endif VNET_DECLARE(struct inpcbhead, tcb); /* queue of active tcpcb's */ VNET_DECLARE(struct inpcbinfo, tcbinfo); extern int tcp_log_in_vain; VNET_DECLARE(int, tcp_mssdflt); /* XXX */ VNET_DECLARE(int, tcp_minmss); VNET_DECLARE(int, tcp_delack_enabled); VNET_DECLARE(int, tcp_do_rfc3390); VNET_DECLARE(int, tcp_initcwnd_segments); VNET_DECLARE(int, tcp_sendspace); VNET_DECLARE(int, tcp_recvspace); VNET_DECLARE(int, path_mtu_discovery); VNET_DECLARE(int, tcp_do_rfc3465); VNET_DECLARE(int, tcp_abc_l_var); #define V_tcb VNET(tcb) #define V_tcbinfo VNET(tcbinfo) #define V_tcp_mssdflt VNET(tcp_mssdflt) #define V_tcp_minmss VNET(tcp_minmss) #define V_tcp_delack_enabled VNET(tcp_delack_enabled) #define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390) #define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments) #define V_tcp_sendspace VNET(tcp_sendspace) #define V_tcp_recvspace VNET(tcp_recvspace) #define V_path_mtu_discovery VNET(path_mtu_discovery) #define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465) #define V_tcp_abc_l_var VNET(tcp_abc_l_var) VNET_DECLARE(int, tcp_do_sack); /* SACK enabled/disabled */ VNET_DECLARE(int, tcp_sc_rst_sock_fail); /* RST on sock alloc failure */ #define V_tcp_do_sack VNET(tcp_do_sack) #define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail) VNET_DECLARE(int, tcp_do_ecn); /* TCP ECN enabled/disabled */ VNET_DECLARE(int, tcp_ecn_maxretries); #define V_tcp_do_ecn VNET(tcp_do_ecn) #define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries) VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]); #define V_tcp_hhh VNET(tcp_hhh) VNET_DECLARE(int, tcp_do_rfc6675_pipe); #define V_tcp_do_rfc6675_pipe VNET(tcp_do_rfc6675_pipe) int tcp_addoptions(struct tcpopt *, u_char *); int tcp_ccalgounload(struct cc_algo *unload_algo); struct tcpcb * tcp_close(struct tcpcb *); void tcp_discardcb(struct tcpcb *); void tcp_twstart(struct tcpcb *); void tcp_twclose(struct tcptw *, int); void tcp_ctlinput(int, struct sockaddr *, void *); int tcp_ctloutput(struct socket *, struct sockopt *); struct tcpcb * tcp_drop(struct tcpcb *, int); void tcp_drain(void); void tcp_init(void); -#ifdef VIMAGE -void tcp_destroy(void); -#endif void tcp_fini(void *); char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, void *, const void *); char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, void *, const void *); int tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *); void tcp_reass_global_init(void); void tcp_reass_flush(struct tcpcb *); void tcp_dooptions(struct tcpopt *, u_char *, int, int); void tcp_dropwithreset(struct mbuf *, struct tcphdr *, struct tcpcb *, int, int); void tcp_pulloutofband(struct socket *, struct tcphdr *, struct mbuf *, int); void tcp_xmit_timer(struct tcpcb *, int); void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *); void cc_ack_received(struct tcpcb *tp, struct tcphdr *th, uint16_t type); void cc_conn_init(struct tcpcb *tp); void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th); void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type); void hhook_run_tcp_est_in(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to); int tcp_input(struct mbuf **, int *, int); void tcp_do_segment(struct mbuf *, struct tcphdr *, struct socket *, struct tcpcb *, int, int, uint8_t, int); int register_tcp_functions(struct tcp_function_block *blk, int wait); int deregister_tcp_functions(struct tcp_function_block *blk); struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs); struct tcp_function_block *find_and_ref_tcp_fb(struct tcp_function_block *blk); int tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp); u_long tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *); u_long tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *); void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *, struct tcp_ifcap *); void tcp_mss(struct tcpcb *, int); int tcp_mssopt(struct in_conninfo *); struct inpcb * tcp_drop_syn_sent(struct inpcb *, int); struct tcpcb * tcp_newtcpcb(struct inpcb *); int tcp_output(struct tcpcb *); void tcp_state_change(struct tcpcb *, int); void tcp_respond(struct tcpcb *, void *, struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int); void tcp_tw_init(void); #ifdef VIMAGE void tcp_tw_destroy(void); #endif void tcp_tw_zone_change(void); int tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *, struct mbuf *, int); void tcp_setpersist(struct tcpcb *); #ifdef TCP_SIGNATURE struct secasvar; struct secasvar *tcp_get_sav(struct mbuf *, u_int); int tcp_signature_do_compute(struct mbuf *, int, int, u_char *, struct secasvar *); int tcp_signature_compute(struct mbuf *, int, int, int, u_char *, u_int); int tcp_signature_verify(struct mbuf *, int, int, int, struct tcpopt *, struct tcphdr *, u_int); int tcp_signature_check(struct mbuf *m, int off0, int tlen, int optlen, struct tcpopt *to, struct tcphdr *th, u_int tcpbflag); #endif void tcp_slowtimo(void); struct tcptemp * tcpip_maketemplate(struct inpcb *); void tcpip_fillheaders(struct inpcb *, void *, void *); void tcp_timer_activate(struct tcpcb *, uint32_t, u_int); int tcp_timer_active(struct tcpcb *, uint32_t); void tcp_timer_stop(struct tcpcb *, uint32_t); void tcp_trace(short, short, struct tcpcb *, void *, struct tcphdr *, int); /* * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo) */ void tcp_hc_init(void); #ifdef VIMAGE void tcp_hc_destroy(void); #endif void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *); u_long tcp_hc_getmtu(struct in_conninfo *); void tcp_hc_updatemtu(struct in_conninfo *, u_long); void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *); extern struct pr_usrreqs tcp_usrreqs; tcp_seq tcp_new_isn(struct tcpcb *); int tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq); void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend); void tcp_clean_sackreport(struct tcpcb *tp); void tcp_sack_adjust(struct tcpcb *tp); struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt); void tcp_sack_partialack(struct tcpcb *, struct tcphdr *); void tcp_free_sackholes(struct tcpcb *tp); int tcp_newreno(struct tcpcb *, struct tcphdr *); u_long tcp_seq_subtract(u_long, u_long ); int tcp_compute_pipe(struct tcpcb *); static inline void tcp_fields_to_host(struct tcphdr *th) { th->th_seq = ntohl(th->th_seq); th->th_ack = ntohl(th->th_ack); th->th_win = ntohs(th->th_win); th->th_urp = ntohs(th->th_urp); } #ifdef TCP_SIGNATURE static inline void tcp_fields_to_net(struct tcphdr *th) { th->th_seq = htonl(th->th_seq); th->th_ack = htonl(th->th_ack); th->th_win = htons(th->th_win); th->th_urp = htons(th->th_urp); } #endif #endif /* _KERNEL */ #endif /* _NETINET_TCP_VAR_H_ */ Index: projects/vnet/sys/netinet/udp_usrreq.c =================================================================== --- projects/vnet/sys/netinet/udp_usrreq.c (revision 294565) +++ projects/vnet/sys/netinet/udp_usrreq.c (revision 294566) @@ -1,1915 +1,1918 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. * Copyright (c) 2008 Robert N. M. Watson * Copyright (c) 2010-2011 Juniper Networks, Inc. * Copyright (c) 2014 Kevin Lo * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 */ #include __FBSDID("$FreeBSD$"); #include "opt_ipfw.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #ifdef IPSEC #include #include #endif #include #include /* * UDP and UDP-Lite protocols implementation. * Per RFC 768, August, 1980. * Per RFC 3828, July, 2004. */ /* * BSD 4.2 defaulted the udp checksum to be off. Turning off udp checksums * removes the only data integrity mechanism for packets and malformed * packets that would otherwise be discarded due to bad checksums, and may * cause problems (especially for NFS data blocks). */ VNET_DEFINE(int, udp_cksum) = 1; SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(udp_cksum), 0, "compute udp checksum"); int udp_log_in_vain = 0; SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW, &udp_log_in_vain, 0, "Log all incoming UDP packets"); VNET_DEFINE(int, udp_blackhole) = 0; SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(udp_blackhole), 0, "Do not send port unreachables for refused connects"); u_long udp_sendspace = 9216; /* really max datagram size */ SYSCTL_ULONG(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW, &udp_sendspace, 0, "Maximum outgoing UDP datagram size"); u_long udp_recvspace = 40 * (1024 + #ifdef INET6 sizeof(struct sockaddr_in6) #else sizeof(struct sockaddr_in) #endif ); /* 40 1K datagrams */ SYSCTL_ULONG(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW, &udp_recvspace, 0, "Maximum space for incoming UDP datagrams"); VNET_DEFINE(struct inpcbhead, udb); /* from udp_var.h */ VNET_DEFINE(struct inpcbinfo, udbinfo); VNET_DEFINE(struct inpcbhead, ulitecb); VNET_DEFINE(struct inpcbinfo, ulitecbinfo); static VNET_DEFINE(uma_zone_t, udpcb_zone); #define V_udpcb_zone VNET(udpcb_zone) #ifndef UDBHASHSIZE #define UDBHASHSIZE 128 #endif VNET_PCPUSTAT_DEFINE(struct udpstat, udpstat); /* from udp_var.h */ VNET_PCPUSTAT_SYSINIT(udpstat); SYSCTL_VNET_PCPUSTAT(_net_inet_udp, UDPCTL_STATS, stats, struct udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)"); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(udpstat); #endif /* VIMAGE */ #ifdef INET static void udp_detach(struct socket *so); static int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *, struct mbuf *, struct thread *); #endif #ifdef IPSEC #ifdef IPSEC_NAT_T #define UF_ESPINUDP_ALL (UF_ESPINUDP_NON_IKE|UF_ESPINUDP) #ifdef INET static struct mbuf *udp4_espdecap(struct inpcb *, struct mbuf *, int); #endif #endif /* IPSEC_NAT_T */ #endif /* IPSEC */ static void udp_zone_change(void *tag) { uma_zone_set_max(V_udbinfo.ipi_zone, maxsockets); uma_zone_set_max(V_udpcb_zone, maxsockets); } static int udp_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp; inp = mem; INP_LOCK_INIT(inp, "inp", "udpinp"); return (0); } static int udplite_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp; inp = mem; INP_LOCK_INIT(inp, "inp", "udpliteinp"); return (0); } void udp_init(void) { /* * For now default to 2-tuple UDP hashing - until the fragment * reassembly code can also update the flowid. * * Once we can calculate the flowid that way and re-establish * a 4-tuple, flip this to 4-tuple. */ in_pcbinfo_init(&V_udbinfo, "udp", &V_udb, UDBHASHSIZE, UDBHASHSIZE, "udp_inpcb", udp_inpcb_init, NULL, 0, IPI_HASHFIELDS_2TUPLE); V_udpcb_zone = uma_zcreate("udpcb", sizeof(struct udpcb), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); uma_zone_set_max(V_udpcb_zone, maxsockets); uma_zone_set_warning(V_udpcb_zone, "kern.ipc.maxsockets limit reached"); EVENTHANDLER_REGISTER(maxsockets_change, udp_zone_change, NULL, EVENTHANDLER_PRI_ANY); } void udplite_init(void) { in_pcbinfo_init(&V_ulitecbinfo, "udplite", &V_ulitecb, UDBHASHSIZE, UDBHASHSIZE, "udplite_inpcb", udplite_inpcb_init, NULL, 0, IPI_HASHFIELDS_2TUPLE); } /* * Kernel module interface for updating udpstat. The argument is an index * into udpstat treated as an array of u_long. While this encodes the * general layout of udpstat into the caller, it doesn't encode its location, * so that future changes to add, for example, per-CPU stats support won't * cause binary compatibility problems for kernel modules. */ void kmod_udpstat_inc(int statnum) { counter_u64_add(VNET(udpstat)[statnum], 1); } int udp_newudpcb(struct inpcb *inp) { struct udpcb *up; up = uma_zalloc(V_udpcb_zone, M_NOWAIT | M_ZERO); if (up == NULL) return (ENOBUFS); inp->inp_ppcb = up; return (0); } void udp_discardcb(struct udpcb *up) { uma_zfree(V_udpcb_zone, up); } #ifdef VIMAGE -void +static void udp_destroy(void) { in_pcbinfo_destroy(&V_udbinfo); uma_zdestroy(V_udpcb_zone); } +VNET_SYSUNINIT(udp, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, udp_destroy, NULL); -void +static void udplite_destroy(void) { in_pcbinfo_destroy(&V_ulitecbinfo); } +VNET_SYSUNINIT(udplite, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, udplite_destroy, + NULL); #endif #ifdef INET /* * Subroutine of udp_input(), which appends the provided mbuf chain to the * passed pcb/socket. The caller must provide a sockaddr_in via udp_in that * contains the source address. If the socket ends up being an IPv6 socket, * udp_append() will convert to a sockaddr_in6 before passing the address * into the socket code. * * In the normal case udp_append() will return 0, indicating that you * must unlock the inp. However if a tunneling protocol is in place we increment * the inpcb refcnt and unlock the inp, on return from the tunneling protocol we * then decrement the reference count. If the inp_rele returns 1, indicating the * inp is gone, we return that to the caller to tell them *not* to unlock * the inp. In the case of multi-cast this will cause the distribution * to stop (though most tunneling protocols known currently do *not* use * multicast). */ static int udp_append(struct inpcb *inp, struct ip *ip, struct mbuf *n, int off, struct sockaddr_in *udp_in) { struct sockaddr *append_sa; struct socket *so; struct mbuf *opts = 0; #ifdef INET6 struct sockaddr_in6 udp_in6; #endif struct udpcb *up; INP_LOCK_ASSERT(inp); /* * Engage the tunneling protocol. */ up = intoudpcb(inp); if (up->u_tun_func != NULL) { in_pcbref(inp); INP_RUNLOCK(inp); (*up->u_tun_func)(n, off, inp, (struct sockaddr *)udp_in, up->u_tun_ctx); INP_RLOCK(inp); return (in_pcbrele_rlocked(inp)); } off += sizeof(struct udphdr); #ifdef IPSEC /* Check AH/ESP integrity. */ if (ipsec4_in_reject(n, inp)) { m_freem(n); return (0); } #ifdef IPSEC_NAT_T up = intoudpcb(inp); KASSERT(up != NULL, ("%s: udpcb NULL", __func__)); if (up->u_flags & UF_ESPINUDP_ALL) { /* IPSec UDP encaps. */ n = udp4_espdecap(inp, n, off); if (n == NULL) /* Consumed. */ return (0); } #endif /* IPSEC_NAT_T */ #endif /* IPSEC */ #ifdef MAC if (mac_inpcb_check_deliver(inp, n) != 0) { m_freem(n); return (0); } #endif /* MAC */ if (inp->inp_flags & INP_CONTROLOPTS || inp->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) { #ifdef INET6 if (inp->inp_vflag & INP_IPV6) (void)ip6_savecontrol_v4(inp, n, &opts, NULL); else #endif /* INET6 */ ip_savecontrol(inp, &opts, ip, n); } #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { bzero(&udp_in6, sizeof(udp_in6)); udp_in6.sin6_len = sizeof(udp_in6); udp_in6.sin6_family = AF_INET6; in6_sin_2_v4mapsin6(udp_in, &udp_in6); append_sa = (struct sockaddr *)&udp_in6; } else #endif /* INET6 */ append_sa = (struct sockaddr *)udp_in; m_adj(n, off); so = inp->inp_socket; SOCKBUF_LOCK(&so->so_rcv); if (sbappendaddr_locked(&so->so_rcv, append_sa, n, opts) == 0) { SOCKBUF_UNLOCK(&so->so_rcv); m_freem(n); if (opts) m_freem(opts); UDPSTAT_INC(udps_fullsock); } else sorwakeup_locked(so); return (0); } int udp_input(struct mbuf **mp, int *offp, int proto) { struct ip *ip; struct udphdr *uh; struct ifnet *ifp; struct inpcb *inp; uint16_t len, ip_len; struct inpcbinfo *pcbinfo; struct ip save_ip; struct sockaddr_in udp_in; struct mbuf *m; struct m_tag *fwd_tag; int cscov_partial, iphlen; m = *mp; iphlen = *offp; ifp = m->m_pkthdr.rcvif; *mp = NULL; UDPSTAT_INC(udps_ipackets); /* * Strip IP options, if any; should skip this, make available to * user, and use on returned packets, but we don't yet have a way to * check the checksum with options still present. */ if (iphlen > sizeof (struct ip)) { ip_stripoptions(m); iphlen = sizeof(struct ip); } /* * Get IP and UDP header together in first mbuf. */ ip = mtod(m, struct ip *); if (m->m_len < iphlen + sizeof(struct udphdr)) { if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == NULL) { UDPSTAT_INC(udps_hdrops); return (IPPROTO_DONE); } ip = mtod(m, struct ip *); } uh = (struct udphdr *)((caddr_t)ip + iphlen); cscov_partial = (proto == IPPROTO_UDPLITE) ? 1 : 0; /* * Destination port of 0 is illegal, based on RFC768. */ if (uh->uh_dport == 0) goto badunlocked; /* * Construct sockaddr format source address. Stuff source address * and datagram in user buffer. */ bzero(&udp_in, sizeof(udp_in)); udp_in.sin_len = sizeof(udp_in); udp_in.sin_family = AF_INET; udp_in.sin_port = uh->uh_sport; udp_in.sin_addr = ip->ip_src; /* * Make mbuf data length reflect UDP length. If not enough data to * reflect UDP length, drop. */ len = ntohs((u_short)uh->uh_ulen); ip_len = ntohs(ip->ip_len) - iphlen; if (proto == IPPROTO_UDPLITE && (len == 0 || len == ip_len)) { /* Zero means checksum over the complete packet. */ if (len == 0) len = ip_len; cscov_partial = 0; } if (ip_len != len) { if (len > ip_len || len < sizeof(struct udphdr)) { UDPSTAT_INC(udps_badlen); goto badunlocked; } if (proto == IPPROTO_UDP) m_adj(m, len - ip_len); } /* * Save a copy of the IP header in case we want restore it for * sending an ICMP error message in response. */ if (!V_udp_blackhole) save_ip = *ip; else memset(&save_ip, 0, sizeof(save_ip)); /* * Checksum extended UDP header and data. */ if (uh->uh_sum) { u_short uh_sum; if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID) && !cscov_partial) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) uh_sum = m->m_pkthdr.csum_data; else uh_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl((u_short)len + m->m_pkthdr.csum_data + proto)); uh_sum ^= 0xffff; } else { char b[9]; bcopy(((struct ipovly *)ip)->ih_x1, b, 9); bzero(((struct ipovly *)ip)->ih_x1, 9); ((struct ipovly *)ip)->ih_len = (proto == IPPROTO_UDP) ? uh->uh_ulen : htons(ip_len); uh_sum = in_cksum(m, len + sizeof (struct ip)); bcopy(b, ((struct ipovly *)ip)->ih_x1, 9); } if (uh_sum) { UDPSTAT_INC(udps_badsum); m_freem(m); return (IPPROTO_DONE); } } else { if (proto == IPPROTO_UDP) { UDPSTAT_INC(udps_nosum); } else { /* UDPLite requires a checksum */ /* XXX: What is the right UDPLite MIB counter here? */ m_freem(m); return (IPPROTO_DONE); } } pcbinfo = udp_get_inpcbinfo(proto); if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || in_broadcast(ip->ip_dst, ifp)) { struct inpcb *last; struct inpcbhead *pcblist; struct ip_moptions *imo; INP_INFO_RLOCK(pcbinfo); pcblist = udp_get_pcblist(proto); last = NULL; LIST_FOREACH(inp, pcblist, inp_list) { if (inp->inp_lport != uh->uh_dport) continue; #ifdef INET6 if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_laddr.s_addr != INADDR_ANY && inp->inp_laddr.s_addr != ip->ip_dst.s_addr) continue; if (inp->inp_faddr.s_addr != INADDR_ANY && inp->inp_faddr.s_addr != ip->ip_src.s_addr) continue; if (inp->inp_fport != 0 && inp->inp_fport != uh->uh_sport) continue; INP_RLOCK(inp); /* * XXXRW: Because we weren't holding either the inpcb * or the hash lock when we checked for a match * before, we should probably recheck now that the * inpcb lock is held. */ /* * Handle socket delivery policy for any-source * and source-specific multicast. [RFC3678] */ imo = inp->inp_moptions; if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { struct sockaddr_in group; int blocked; if (imo == NULL) { INP_RUNLOCK(inp); continue; } bzero(&group, sizeof(struct sockaddr_in)); group.sin_len = sizeof(struct sockaddr_in); group.sin_family = AF_INET; group.sin_addr = ip->ip_dst; blocked = imo_multi_filter(imo, ifp, (struct sockaddr *)&group, (struct sockaddr *)&udp_in); if (blocked != MCAST_PASS) { if (blocked == MCAST_NOTGMEMBER) IPSTAT_INC(ips_notmember); if (blocked == MCAST_NOTSMEMBER || blocked == MCAST_MUTED) UDPSTAT_INC(udps_filtermcast); INP_RUNLOCK(inp); continue; } } if (last != NULL) { struct mbuf *n; if ((n = m_copy(m, 0, M_COPYALL)) != NULL) { UDP_PROBE(receive, NULL, last, ip, last, uh); if (udp_append(last, ip, n, iphlen, &udp_in)) { goto inp_lost; } } INP_RUNLOCK(last); } last = inp; /* * Don't look for additional matches if this one does * not have either the SO_REUSEPORT or SO_REUSEADDR * socket options set. This heuristic avoids * searching through all pcbs in the common case of a * non-shared port. It assumes that an application * will never clear these options after setting them. */ if ((last->inp_socket->so_options & (SO_REUSEPORT|SO_REUSEADDR)) == 0) break; } if (last == NULL) { /* * No matching pcb found; discard datagram. (No need * to send an ICMP Port Unreachable for a broadcast * or multicast datgram.) */ UDPSTAT_INC(udps_noportbcast); if (inp) INP_RUNLOCK(inp); INP_INFO_RUNLOCK(pcbinfo); goto badunlocked; } UDP_PROBE(receive, NULL, last, ip, last, uh); if (udp_append(last, ip, m, iphlen, &udp_in) == 0) INP_RUNLOCK(last); inp_lost: INP_INFO_RUNLOCK(pcbinfo); return (IPPROTO_DONE); } /* * Locate pcb for datagram. */ /* * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. */ if ((m->m_flags & M_IP_NEXTHOP) && (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { struct sockaddr_in *next_hop; next_hop = (struct sockaddr_in *)(fwd_tag + 1); /* * Transparently forwarded. Pretend to be the destination. * Already got one like this? */ inp = in_pcblookup_mbuf(pcbinfo, ip->ip_src, uh->uh_sport, ip->ip_dst, uh->uh_dport, INPLOOKUP_RLOCKPCB, ifp, m); if (!inp) { /* * It's new. Try to find the ambushing socket. * Because we've rewritten the destination address, * any hardware-generated hash is ignored. */ inp = in_pcblookup(pcbinfo, ip->ip_src, uh->uh_sport, next_hop->sin_addr, next_hop->sin_port ? htons(next_hop->sin_port) : uh->uh_dport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, ifp); } /* Remove the tag from the packet. We don't need it anymore. */ m_tag_delete(m, fwd_tag); m->m_flags &= ~M_IP_NEXTHOP; } else inp = in_pcblookup_mbuf(pcbinfo, ip->ip_src, uh->uh_sport, ip->ip_dst, uh->uh_dport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, ifp, m); if (inp == NULL) { if (udp_log_in_vain) { char buf[4*sizeof "123"]; strcpy(buf, inet_ntoa(ip->ip_dst)); log(LOG_INFO, "Connection attempt to UDP %s:%d from %s:%d\n", buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src), ntohs(uh->uh_sport)); } UDPSTAT_INC(udps_noport); if (m->m_flags & (M_BCAST | M_MCAST)) { UDPSTAT_INC(udps_noportbcast); goto badunlocked; } if (V_udp_blackhole) goto badunlocked; if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0) goto badunlocked; *ip = save_ip; icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); return (IPPROTO_DONE); } /* * Check the minimum TTL for socket. */ INP_RLOCK_ASSERT(inp); if (inp->inp_ip_minttl && inp->inp_ip_minttl > ip->ip_ttl) { INP_RUNLOCK(inp); m_freem(m); return (IPPROTO_DONE); } if (cscov_partial) { struct udpcb *up; up = intoudpcb(inp); if (up->u_rxcslen == 0 || up->u_rxcslen > len) { INP_RUNLOCK(inp); m_freem(m); return (IPPROTO_DONE); } } UDP_PROBE(receive, NULL, inp, ip, inp, uh); if (udp_append(inp, ip, m, iphlen, &udp_in) == 0) INP_RUNLOCK(inp); return (IPPROTO_DONE); badunlocked: m_freem(m); return (IPPROTO_DONE); } #endif /* INET */ /* * Notify a udp user of an asynchronous error; just wake up so that they can * collect error status. */ struct inpcb * udp_notify(struct inpcb *inp, int errno) { /* * While udp_ctlinput() always calls udp_notify() with a read lock * when invoking it directly, in_pcbnotifyall() currently uses write * locks due to sharing code with TCP. For now, accept either a read * or a write lock, but a read lock is sufficient. */ INP_LOCK_ASSERT(inp); inp->inp_socket->so_error = errno; sorwakeup(inp->inp_socket); sowwakeup(inp->inp_socket); return (inp); } #ifdef INET static void udp_common_ctlinput(int cmd, struct sockaddr *sa, void *vip, struct inpcbinfo *pcbinfo) { struct ip *ip = vip; struct udphdr *uh; struct in_addr faddr; struct inpcb *inp; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; /* * Redirects don't need to be handled up here. */ if (PRC_IS_REDIRECT(cmd)) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * * XXX: We never get this from ICMP, otherwise it makes an excellent * DoS attack on machines with many connections. */ if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip != NULL) { uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); inp = in_pcblookup(pcbinfo, faddr, uh->uh_dport, ip->ip_src, uh->uh_sport, INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { INP_RLOCK_ASSERT(inp); if (inp->inp_socket != NULL) { udp_notify(inp, inetctlerrmap[cmd]); } INP_RUNLOCK(inp); } } else in_pcbnotifyall(pcbinfo, faddr, inetctlerrmap[cmd], udp_notify); } void udp_ctlinput(int cmd, struct sockaddr *sa, void *vip) { return (udp_common_ctlinput(cmd, sa, vip, &V_udbinfo)); } void udplite_ctlinput(int cmd, struct sockaddr *sa, void *vip) { return (udp_common_ctlinput(cmd, sa, vip, &V_ulitecbinfo)); } #endif /* INET */ static int udp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the PCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == 0) { n = V_udbinfo.ipi_count; n += imax(n / 8, 10); req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); return (0); } if (req->newptr != 0) return (EPERM); /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&V_udbinfo); gencnt = V_udbinfo.ipi_gencnt; n = V_udbinfo.ipi_count; INP_INFO_RUNLOCK(&V_udbinfo); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + n * sizeof(struct xinpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return (error); inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == 0) return (ENOMEM); INP_INFO_RLOCK(&V_udbinfo); for (inp = LIST_FIRST(V_udbinfo.ipi_listhead), i = 0; inp && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_WLOCK(inp); if (inp->inp_gencnt <= gencnt && cr_canseeinpcb(req->td->td_ucred, inp) == 0) { in_pcbref(inp); inp_list[i++] = inp; } INP_WUNLOCK(inp); } INP_INFO_RUNLOCK(&V_udbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xinpcb xi; bzero(&xi, sizeof(xi)); xi.xi_len = sizeof xi; /* XXX should avoid extra copy */ bcopy(inp, &xi.xi_inp, sizeof *inp); if (inp->inp_socket) sotoxsocket(inp->inp_socket, &xi.xi_socket); xi.xi_inp.inp_gencnt = inp->inp_gencnt; INP_RUNLOCK(inp); error = SYSCTL_OUT(req, &xi, sizeof xi); } else INP_RUNLOCK(inp); } INP_INFO_WLOCK(&V_udbinfo); for (i = 0; i < n; i++) { inp = inp_list[i]; INP_RLOCK(inp); if (!in_pcbrele_rlocked(inp)) INP_RUNLOCK(inp); } INP_INFO_WUNLOCK(&V_udbinfo); if (!error) { /* * Give the user an updated idea of our state. If the * generation differs from what we told her before, she knows * that something happened while we were processing this * request, and it might be necessary to retry. */ INP_INFO_RLOCK(&V_udbinfo); xig.xig_gen = V_udbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = V_udbinfo.ipi_count; INP_INFO_RUNLOCK(&V_udbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, udp_pcblist, "S,xinpcb", "List of active UDP sockets"); #ifdef INET static int udp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error; error = priv_check(req->td, PRIV_NETINET_GETCRED); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); inp = in_pcblookup(&V_udbinfo, addrs[1].sin_addr, addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { INP_RLOCK_ASSERT(inp); if (inp->inp_socket == NULL) error = ENOENT; if (error == 0) error = cr_canseeinpcb(req->td->td_ucred, inp); if (error == 0) cru2x(inp->inp_cred, &xuc); INP_RUNLOCK(inp); } else error = ENOENT; if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, udp_getcred, "S,xucred", "Get the xucred of a UDP connection"); #endif /* INET */ int udp_ctloutput(struct socket *so, struct sockopt *sopt) { struct inpcb *inp; struct udpcb *up; int isudplite, error, optval; error = 0; isudplite = (so->so_proto->pr_protocol == IPPROTO_UDPLITE) ? 1 : 0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); INP_WLOCK(inp); if (sopt->sopt_level != so->so_proto->pr_protocol) { #ifdef INET6 if (INP_CHECK_SOCKAF(so, AF_INET6)) { INP_WUNLOCK(inp); error = ip6_ctloutput(so, sopt); } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET { INP_WUNLOCK(inp); error = ip_ctloutput(so, sopt); } #endif return (error); } switch (sopt->sopt_dir) { case SOPT_SET: switch (sopt->sopt_name) { case UDP_ENCAP: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; inp = sotoinpcb(so); KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); INP_WLOCK(inp); #ifdef IPSEC_NAT_T up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); #endif switch (optval) { case 0: /* Clear all UDP encap. */ #ifdef IPSEC_NAT_T up->u_flags &= ~UF_ESPINUDP_ALL; #endif break; #ifdef IPSEC_NAT_T case UDP_ENCAP_ESPINUDP: case UDP_ENCAP_ESPINUDP_NON_IKE: up->u_flags &= ~UF_ESPINUDP_ALL; if (optval == UDP_ENCAP_ESPINUDP) up->u_flags |= UF_ESPINUDP; else if (optval == UDP_ENCAP_ESPINUDP_NON_IKE) up->u_flags |= UF_ESPINUDP_NON_IKE; break; #endif default: error = EINVAL; break; } INP_WUNLOCK(inp); break; case UDPLITE_SEND_CSCOV: case UDPLITE_RECV_CSCOV: if (!isudplite) { INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval)); if (error != 0) break; inp = sotoinpcb(so); KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); INP_WLOCK(inp); up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); if ((optval != 0 && optval < 8) || (optval > 65535)) { INP_WUNLOCK(inp); error = EINVAL; break; } if (sopt->sopt_name == UDPLITE_SEND_CSCOV) up->u_txcslen = optval; else up->u_rxcslen = optval; INP_WUNLOCK(inp); break; default: INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } break; case SOPT_GET: switch (sopt->sopt_name) { #ifdef IPSEC_NAT_T case UDP_ENCAP: up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); optval = up->u_flags & UF_ESPINUDP_ALL; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; #endif case UDPLITE_SEND_CSCOV: case UDPLITE_RECV_CSCOV: if (!isudplite) { INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); if (sopt->sopt_name == UDPLITE_SEND_CSCOV) optval = up->u_txcslen; else optval = up->u_rxcslen; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof(optval)); break; default: INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } break; } return (error); } #ifdef INET #define UH_WLOCKED 2 #define UH_RLOCKED 1 #define UH_UNLOCKED 0 static int udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td) { struct udpiphdr *ui; int len = m->m_pkthdr.len; struct in_addr faddr, laddr; struct cmsghdr *cm; struct inpcbinfo *pcbinfo; struct sockaddr_in *sin, src; int cscov_partial = 0; int error = 0; int ipflags; u_short fport, lport; int unlock_udbinfo; u_char tos; uint8_t pr; uint16_t cscov = 0; uint32_t flowid = 0; uint8_t flowtype = M_HASHTYPE_NONE; /* * udp_output() may need to temporarily bind or connect the current * inpcb. As such, we don't know up front whether we will need the * pcbinfo lock or not. Do any work to decide what is needed up * front before acquiring any locks. */ if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { if (control) m_freem(control); m_freem(m); return (EMSGSIZE); } src.sin_family = 0; INP_RLOCK(inp); tos = inp->inp_ip_tos; if (control != NULL) { /* * XXX: Currently, we assume all the optional information is * stored in a single mbuf. */ if (control->m_next) { INP_RUNLOCK(inp); m_freem(control); m_freem(m); return (EINVAL); } for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { cm = mtod(control, struct cmsghdr *); if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) { error = EINVAL; break; } if (cm->cmsg_level != IPPROTO_IP) continue; switch (cm->cmsg_type) { case IP_SENDSRCADDR: if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_addr))) { error = EINVAL; break; } bzero(&src, sizeof(src)); src.sin_family = AF_INET; src.sin_len = sizeof(src); src.sin_port = inp->inp_lport; src.sin_addr = *(struct in_addr *)CMSG_DATA(cm); break; case IP_TOS: if (cm->cmsg_len != CMSG_LEN(sizeof(u_char))) { error = EINVAL; break; } tos = *(u_char *)CMSG_DATA(cm); break; case IP_FLOWID: if (cm->cmsg_len != CMSG_LEN(sizeof(uint32_t))) { error = EINVAL; break; } flowid = *(uint32_t *) CMSG_DATA(cm); break; case IP_FLOWTYPE: if (cm->cmsg_len != CMSG_LEN(sizeof(uint32_t))) { error = EINVAL; break; } flowtype = *(uint32_t *) CMSG_DATA(cm); break; #ifdef RSS case IP_RSSBUCKETID: if (cm->cmsg_len != CMSG_LEN(sizeof(uint32_t))) { error = EINVAL; break; } /* This is just a placeholder for now */ break; #endif /* RSS */ default: error = ENOPROTOOPT; break; } if (error) break; } m_freem(control); } if (error) { INP_RUNLOCK(inp); m_freem(m); return (error); } /* * Depending on whether or not the application has bound or connected * the socket, we may have to do varying levels of work. The optimal * case is for a connected UDP socket, as a global lock isn't * required at all. * * In order to decide which we need, we require stability of the * inpcb binding, which we ensure by acquiring a read lock on the * inpcb. This doesn't strictly follow the lock order, so we play * the trylock and retry game; note that we may end up with more * conservative locks than required the second time around, so later * assertions have to accept that. Further analysis of the number of * misses under contention is required. * * XXXRW: Check that hash locking update here is correct. */ pr = inp->inp_socket->so_proto->pr_protocol; pcbinfo = udp_get_inpcbinfo(pr); sin = (struct sockaddr_in *)addr; if (sin != NULL && (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0)) { INP_RUNLOCK(inp); INP_WLOCK(inp); INP_HASH_WLOCK(pcbinfo); unlock_udbinfo = UH_WLOCKED; } else if ((sin != NULL && ( (sin->sin_addr.s_addr == INADDR_ANY) || (sin->sin_addr.s_addr == INADDR_BROADCAST) || (inp->inp_laddr.s_addr == INADDR_ANY) || (inp->inp_lport == 0))) || (src.sin_family == AF_INET)) { INP_HASH_RLOCK(pcbinfo); unlock_udbinfo = UH_RLOCKED; } else unlock_udbinfo = UH_UNLOCKED; /* * If the IP_SENDSRCADDR control message was specified, override the * source address for this datagram. Its use is invalidated if the * address thus specified is incomplete or clobbers other inpcbs. */ laddr = inp->inp_laddr; lport = inp->inp_lport; if (src.sin_family == AF_INET) { INP_HASH_LOCK_ASSERT(pcbinfo); if ((lport == 0) || (laddr.s_addr == INADDR_ANY && src.sin_addr.s_addr == INADDR_ANY)) { error = EINVAL; goto release; } error = in_pcbbind_setup(inp, (struct sockaddr *)&src, &laddr.s_addr, &lport, td->td_ucred); if (error) goto release; } /* * If a UDP socket has been connected, then a local address/port will * have been selected and bound. * * If a UDP socket has not been connected to, then an explicit * destination address must be used, in which case a local * address/port may not have been selected and bound. */ if (sin != NULL) { INP_LOCK_ASSERT(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { error = EISCONN; goto release; } /* * Jail may rewrite the destination address, so let it do * that before we use it. */ error = prison_remote_ip4(td->td_ucred, &sin->sin_addr); if (error) goto release; /* * If a local address or port hasn't yet been selected, or if * the destination address needs to be rewritten due to using * a special INADDR_ constant, invoke in_pcbconnect_setup() * to do the heavy lifting. Once a port is selected, we * commit the binding back to the socket; we also commit the * binding of the address if in jail. * * If we already have a valid binding and we're not * requesting a destination address rewrite, use a fast path. */ if (inp->inp_laddr.s_addr == INADDR_ANY || inp->inp_lport == 0 || sin->sin_addr.s_addr == INADDR_ANY || sin->sin_addr.s_addr == INADDR_BROADCAST) { INP_HASH_LOCK_ASSERT(pcbinfo); error = in_pcbconnect_setup(inp, addr, &laddr.s_addr, &lport, &faddr.s_addr, &fport, NULL, td->td_ucred); if (error) goto release; /* * XXXRW: Why not commit the port if the address is * !INADDR_ANY? */ /* Commit the local port if newly assigned. */ if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); /* * Remember addr if jailed, to prevent * rebinding. */ if (prison_flag(td->td_ucred, PR_IP4)) inp->inp_laddr = laddr; inp->inp_lport = lport; if (in_pcbinshash(inp) != 0) { inp->inp_lport = 0; error = EAGAIN; goto release; } inp->inp_flags |= INP_ANONPORT; } } else { faddr = sin->sin_addr; fport = sin->sin_port; } } else { INP_LOCK_ASSERT(inp); faddr = inp->inp_faddr; fport = inp->inp_fport; if (faddr.s_addr == INADDR_ANY) { error = ENOTCONN; goto release; } } /* * Calculate data length and get a mbuf for UDP, IP, and possible * link-layer headers. Immediate slide the data pointer back forward * since we won't use that space at this layer. */ M_PREPEND(m, sizeof(struct udpiphdr) + max_linkhdr, M_NOWAIT); if (m == NULL) { error = ENOBUFS; goto release; } m->m_data += max_linkhdr; m->m_len -= max_linkhdr; m->m_pkthdr.len -= max_linkhdr; /* * Fill in mbuf with extended UDP header and addresses and length put * into network format. */ ui = mtod(m, struct udpiphdr *); bzero(ui->ui_x1, sizeof(ui->ui_x1)); /* XXX still needed? */ ui->ui_pr = pr; ui->ui_src = laddr; ui->ui_dst = faddr; ui->ui_sport = lport; ui->ui_dport = fport; ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr)); if (pr == IPPROTO_UDPLITE) { struct udpcb *up; uint16_t plen; up = intoudpcb(inp); cscov = up->u_txcslen; plen = (u_short)len + sizeof(struct udphdr); if (cscov >= plen) cscov = 0; ui->ui_len = htons(plen); ui->ui_ulen = htons(cscov); /* * For UDP-Lite, checksum coverage length of zero means * the entire UDPLite packet is covered by the checksum. */ cscov_partial = (cscov == 0) ? 0 : 1; } else ui->ui_v = IPVERSION << 4; /* * Set the Don't Fragment bit in the IP header. */ if (inp->inp_flags & INP_DONTFRAG) { struct ip *ip; ip = (struct ip *)&ui->ui_i; ip->ip_off |= htons(IP_DF); } ipflags = 0; if (inp->inp_socket->so_options & SO_DONTROUTE) ipflags |= IP_ROUTETOIF; if (inp->inp_socket->so_options & SO_BROADCAST) ipflags |= IP_ALLOWBROADCAST; if (inp->inp_flags & INP_ONESBCAST) ipflags |= IP_SENDONES; #ifdef MAC mac_inpcb_create_mbuf(inp, m); #endif /* * Set up checksum and output datagram. */ ui->ui_sum = 0; if (pr == IPPROTO_UDPLITE) { if (inp->inp_flags & INP_ONESBCAST) faddr.s_addr = INADDR_BROADCAST; if (cscov_partial) { if ((ui->ui_sum = in_cksum(m, sizeof(struct ip) + cscov)) == 0) ui->ui_sum = 0xffff; } else { if ((ui->ui_sum = in_cksum(m, sizeof(struct udpiphdr) + len)) == 0) ui->ui_sum = 0xffff; } } else if (V_udp_cksum) { if (inp->inp_flags & INP_ONESBCAST) faddr.s_addr = INADDR_BROADCAST; ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr, htons((u_short)len + sizeof(struct udphdr) + pr)); m->m_pkthdr.csum_flags = CSUM_UDP; m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); } ((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len); ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */ ((struct ip *)ui)->ip_tos = tos; /* XXX */ UDPSTAT_INC(udps_opackets); /* * Setup flowid / RSS information for outbound socket. * * Once the UDP code decides to set a flowid some other way, * this allows the flowid to be overridden by userland. */ if (flowtype != M_HASHTYPE_NONE) { m->m_pkthdr.flowid = flowid; M_HASHTYPE_SET(m, flowtype); #ifdef RSS } else { uint32_t hash_val, hash_type; /* * Calculate an appropriate RSS hash for UDP and * UDP Lite. * * The called function will take care of figuring out * whether a 2-tuple or 4-tuple hash is required based * on the currently configured scheme. * * Later later on connected socket values should be * cached in the inpcb and reused, rather than constantly * re-calculating it. * * UDP Lite is a different protocol number and will * likely end up being hashed as a 2-tuple until * RSS / NICs grow UDP Lite protocol awareness. */ if (rss_proto_software_hash_v4(faddr, laddr, fport, lport, pr, &hash_val, &hash_type) == 0) { m->m_pkthdr.flowid = hash_val; M_HASHTYPE_SET(m, hash_type); } #endif } #ifdef RSS /* * Don't override with the inp cached flowid value. * * Depending upon the kind of send being done, the inp * flowid/flowtype values may actually not be appropriate * for this particular socket send. * * We should either leave the flowid at zero (which is what is * currently done) or set it to some software generated * hash value based on the packet contents. */ ipflags |= IP_NODEFAULTFLOWID; #endif /* RSS */ if (unlock_udbinfo == UH_WLOCKED) INP_HASH_WUNLOCK(pcbinfo); else if (unlock_udbinfo == UH_RLOCKED) INP_HASH_RUNLOCK(pcbinfo); UDP_PROBE(send, NULL, inp, &ui->ui_i, inp, &ui->ui_u); error = ip_output(m, inp->inp_options, NULL, ipflags, inp->inp_moptions, inp); if (unlock_udbinfo == UH_WLOCKED) INP_WUNLOCK(inp); else INP_RUNLOCK(inp); return (error); release: if (unlock_udbinfo == UH_WLOCKED) { INP_HASH_WUNLOCK(pcbinfo); INP_WUNLOCK(inp); } else if (unlock_udbinfo == UH_RLOCKED) { INP_HASH_RUNLOCK(pcbinfo); INP_RUNLOCK(inp); } else INP_RUNLOCK(inp); m_freem(m); return (error); } #if defined(IPSEC) && defined(IPSEC_NAT_T) /* * Potentially decap ESP in UDP frame. Check for an ESP header * and optional marker; if present, strip the UDP header and * push the result through IPSec. * * Returns mbuf to be processed (potentially re-allocated) or * NULL if consumed and/or processed. */ static struct mbuf * udp4_espdecap(struct inpcb *inp, struct mbuf *m, int off) { size_t minlen, payload, skip, iphlen; caddr_t data; struct udpcb *up; struct m_tag *tag; struct udphdr *udphdr; struct ip *ip; INP_RLOCK_ASSERT(inp); /* * Pull up data so the longest case is contiguous: * IP/UDP hdr + non ESP marker + ESP hdr. */ minlen = off + sizeof(uint64_t) + sizeof(struct esp); if (minlen > m->m_pkthdr.len) minlen = m->m_pkthdr.len; if ((m = m_pullup(m, minlen)) == NULL) { IPSECSTAT_INC(ips_in_inval); return (NULL); /* Bypass caller processing. */ } data = mtod(m, caddr_t); /* Points to ip header. */ payload = m->m_len - off; /* Size of payload. */ if (payload == 1 && data[off] == '\xff') return (m); /* NB: keepalive packet, no decap. */ up = intoudpcb(inp); KASSERT(up != NULL, ("%s: udpcb NULL", __func__)); KASSERT((up->u_flags & UF_ESPINUDP_ALL) != 0, ("u_flags 0x%x", up->u_flags)); /* * Check that the payload is large enough to hold an * ESP header and compute the amount of data to remove. * * NB: the caller has already done a pullup for us. * XXX can we assume alignment and eliminate bcopys? */ if (up->u_flags & UF_ESPINUDP_NON_IKE) { /* * draft-ietf-ipsec-nat-t-ike-0[01].txt and * draft-ietf-ipsec-udp-encaps-(00/)01.txt, ignoring * possible AH mode non-IKE marker+non-ESP marker * from draft-ietf-ipsec-udp-encaps-00.txt. */ uint64_t marker; if (payload <= sizeof(uint64_t) + sizeof(struct esp)) return (m); /* NB: no decap. */ bcopy(data + off, &marker, sizeof(uint64_t)); if (marker != 0) /* Non-IKE marker. */ return (m); /* NB: no decap. */ skip = sizeof(uint64_t) + sizeof(struct udphdr); } else { uint32_t spi; if (payload <= sizeof(struct esp)) { IPSECSTAT_INC(ips_in_inval); m_freem(m); return (NULL); /* Discard. */ } bcopy(data + off, &spi, sizeof(uint32_t)); if (spi == 0) /* Non-ESP marker. */ return (m); /* NB: no decap. */ skip = sizeof(struct udphdr); } /* * Setup a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember * the UDP ports. This is required if we want to select * the right SPD for multiple hosts behind same NAT. * * NB: ports are maintained in network byte order everywhere * in the NAT-T code. */ tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS, 2 * sizeof(uint16_t), M_NOWAIT); if (tag == NULL) { IPSECSTAT_INC(ips_in_nomem); m_freem(m); return (NULL); /* Discard. */ } iphlen = off - sizeof(struct udphdr); udphdr = (struct udphdr *)(data + iphlen); ((uint16_t *)(tag + 1))[0] = udphdr->uh_sport; ((uint16_t *)(tag + 1))[1] = udphdr->uh_dport; m_tag_prepend(m, tag); /* * Remove the UDP header (and possibly the non ESP marker) * IP header length is iphlen * Before: * <--- off ---> * +----+------+-----+ * | IP | UDP | ESP | * +----+------+-----+ * <-skip-> * After: * +----+-----+ * | IP | ESP | * +----+-----+ * <-skip-> */ ovbcopy(data, data + skip, iphlen); m_adj(m, skip); ip = mtod(m, struct ip *); ip->ip_len = htons(ntohs(ip->ip_len) - skip); ip->ip_p = IPPROTO_ESP; /* * We cannot yet update the cksums so clear any * h/w cksum flags as they are no longer valid. */ if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID|CSUM_PSEUDO_HDR); (void) ipsec_common_input(m, iphlen, offsetof(struct ip, ip_p), AF_INET, ip->ip_p); return (NULL); /* NB: consumed, bypass processing. */ } #endif /* defined(IPSEC) && defined(IPSEC_NAT_T) */ static void udp_abort(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_abort: inp == NULL")); INP_WLOCK(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { INP_HASH_WLOCK(pcbinfo); in_pcbdisconnect(inp); inp->inp_laddr.s_addr = INADDR_ANY; INP_HASH_WUNLOCK(pcbinfo); soisdisconnected(so); } INP_WUNLOCK(inp); } static int udp_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; struct inpcbinfo *pcbinfo; int error; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp == NULL, ("udp_attach: inp != NULL")); error = soreserve(so, udp_sendspace, udp_recvspace); if (error) return (error); INP_INFO_WLOCK(pcbinfo); error = in_pcballoc(so, pcbinfo); if (error) { INP_INFO_WUNLOCK(pcbinfo); return (error); } inp = sotoinpcb(so); inp->inp_vflag |= INP_IPV4; inp->inp_ip_ttl = V_ip_defttl; error = udp_newudpcb(inp); if (error) { in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(pcbinfo); return (error); } INP_WUNLOCK(inp); INP_INFO_WUNLOCK(pcbinfo); return (0); } #endif /* INET */ int udp_set_kernel_tunneling(struct socket *so, udp_tun_func_t f, void *ctx) { struct inpcb *inp; struct udpcb *up; KASSERT(so->so_type == SOCK_DGRAM, ("udp_set_kernel_tunneling: !dgram")); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_set_kernel_tunneling: inp == NULL")); INP_WLOCK(inp); up = intoudpcb(inp); if (up->u_tun_func != NULL) { INP_WUNLOCK(inp); return (EBUSY); } up->u_tun_func = f; up->u_tun_ctx = ctx; INP_WUNLOCK(inp); return (0); } #ifdef INET static int udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; struct inpcbinfo *pcbinfo; int error; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_bind: inp == NULL")); INP_WLOCK(inp); INP_HASH_WLOCK(pcbinfo); error = in_pcbbind(inp, nam, td->td_ucred); INP_HASH_WUNLOCK(pcbinfo); INP_WUNLOCK(inp); return (error); } static void udp_close(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_close: inp == NULL")); INP_WLOCK(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { INP_HASH_WLOCK(pcbinfo); in_pcbdisconnect(inp); inp->inp_laddr.s_addr = INADDR_ANY; INP_HASH_WUNLOCK(pcbinfo); soisdisconnected(so); } INP_WUNLOCK(inp); } static int udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; struct inpcbinfo *pcbinfo; struct sockaddr_in *sin; int error; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_connect: inp == NULL")); INP_WLOCK(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { INP_WUNLOCK(inp); return (EISCONN); } sin = (struct sockaddr_in *)nam; error = prison_remote_ip4(td->td_ucred, &sin->sin_addr); if (error != 0) { INP_WUNLOCK(inp); return (error); } INP_HASH_WLOCK(pcbinfo); error = in_pcbconnect(inp, nam, td->td_ucred); INP_HASH_WUNLOCK(pcbinfo); if (error == 0) soisconnected(so); INP_WUNLOCK(inp); return (error); } static void udp_detach(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; struct udpcb *up; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_detach: inp == NULL")); KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, ("udp_detach: not disconnected")); INP_INFO_WLOCK(pcbinfo); INP_WLOCK(inp); up = intoudpcb(inp); KASSERT(up != NULL, ("%s: up == NULL", __func__)); inp->inp_ppcb = NULL; in_pcbdetach(inp); in_pcbfree(inp); INP_INFO_WUNLOCK(pcbinfo); udp_discardcb(up); } static int udp_disconnect(struct socket *so) { struct inpcb *inp; struct inpcbinfo *pcbinfo; pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol); inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_disconnect: inp == NULL")); INP_WLOCK(inp); if (inp->inp_faddr.s_addr == INADDR_ANY) { INP_WUNLOCK(inp); return (ENOTCONN); } INP_HASH_WLOCK(pcbinfo); in_pcbdisconnect(inp); inp->inp_laddr.s_addr = INADDR_ANY; INP_HASH_WUNLOCK(pcbinfo); SOCK_LOCK(so); so->so_state &= ~SS_ISCONNECTED; /* XXX */ SOCK_UNLOCK(so); INP_WUNLOCK(inp); return (0); } static int udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_send: inp == NULL")); return (udp_output(inp, m, addr, control, td)); } #endif /* INET */ int udp_shutdown(struct socket *so) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("udp_shutdown: inp == NULL")); INP_WLOCK(inp); socantsendmore(so); INP_WUNLOCK(inp); return (0); } #ifdef INET struct pr_usrreqs udp_usrreqs = { .pru_abort = udp_abort, .pru_attach = udp_attach, .pru_bind = udp_bind, .pru_connect = udp_connect, .pru_control = in_control, .pru_detach = udp_detach, .pru_disconnect = udp_disconnect, .pru_peeraddr = in_getpeeraddr, .pru_send = udp_send, .pru_soreceive = soreceive_dgram, .pru_sosend = sosend_dgram, .pru_shutdown = udp_shutdown, .pru_sockaddr = in_getsockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = udp_close, }; #endif /* INET */ Index: projects/vnet/sys/netinet/udp_var.h =================================================================== --- projects/vnet/sys/netinet/udp_var.h (revision 294565) +++ projects/vnet/sys/netinet/udp_var.h (revision 294566) @@ -1,186 +1,182 @@ /*- * Copyright (c) 1982, 1986, 1989, 1993 * The Regents of the University of California. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)udp_var.h 8.1 (Berkeley) 6/10/93 * $FreeBSD$ */ #ifndef _NETINET_UDP_VAR_H_ #define _NETINET_UDP_VAR_H_ /* * UDP kernel structures and variables. */ struct udpiphdr { struct ipovly ui_i; /* overlaid ip structure */ struct udphdr ui_u; /* udp header */ }; #define ui_x1 ui_i.ih_x1 #define ui_v ui_i.ih_x1[0] #define ui_pr ui_i.ih_pr #define ui_len ui_i.ih_len #define ui_src ui_i.ih_src #define ui_dst ui_i.ih_dst #define ui_sport ui_u.uh_sport #define ui_dport ui_u.uh_dport #define ui_ulen ui_u.uh_ulen #define ui_sum ui_u.uh_sum struct inpcb; struct mbuf; typedef void(*udp_tun_func_t)(struct mbuf *, int off, struct inpcb *, const struct sockaddr *, void *); /* * UDP control block; one per udp. */ struct udpcb { udp_tun_func_t u_tun_func; /* UDP kernel tunneling callback. */ u_int u_flags; /* Generic UDP flags. */ uint16_t u_rxcslen; /* Coverage for incoming datagrams. */ uint16_t u_txcslen; /* Coverage for outgoing datagrams. */ void *u_tun_ctx; /* Tunneling callback context. */ }; #define intoudpcb(ip) ((struct udpcb *)(ip)->inp_ppcb) #define sotoudpcb(so) (intoudpcb(sotoinpcb(so))) /* IPsec: ESP in UDP tunneling: */ #define UF_ESPINUDP_NON_IKE 0x00000001 /* w/ non-IKE marker .. */ /* .. per draft-ietf-ipsec-nat-t-ike-0[01], * and draft-ietf-ipsec-udp-encaps-(00/)01.txt */ #define UF_ESPINUDP 0x00000002 /* w/ non-ESP marker. */ struct udpstat { /* input statistics: */ uint64_t udps_ipackets; /* total input packets */ uint64_t udps_hdrops; /* packet shorter than header */ uint64_t udps_badsum; /* checksum error */ uint64_t udps_nosum; /* no checksum */ uint64_t udps_badlen; /* data length larger than packet */ uint64_t udps_noport; /* no socket on port */ uint64_t udps_noportbcast; /* of above, arrived as broadcast */ uint64_t udps_fullsock; /* not delivered, input socket full */ uint64_t udpps_pcbcachemiss; /* input packets missing pcb cache */ uint64_t udpps_pcbhashmiss; /* input packets not for hashed pcb */ /* output statistics: */ uint64_t udps_opackets; /* total output packets */ uint64_t udps_fastout; /* output packets on fast path */ /* of no socket on port, arrived as multicast */ uint64_t udps_noportmcast; uint64_t udps_filtermcast; /* blocked by multicast filter */ }; #ifdef _KERNEL #include VNET_PCPUSTAT_DECLARE(struct udpstat, udpstat); /* * In-kernel consumers can use these accessor macros directly to update * stats. */ #define UDPSTAT_ADD(name, val) \ VNET_PCPUSTAT_ADD(struct udpstat, udpstat, name, (val)) #define UDPSTAT_INC(name) UDPSTAT_ADD(name, 1) /* * Kernel module consumers must use this accessor macro. */ void kmod_udpstat_inc(int statnum); #define KMOD_UDPSTAT_INC(name) \ kmod_udpstat_inc(offsetof(struct udpstat, name) / sizeof(uint64_t)) #endif /* * Identifiers for UDP sysctl nodes. */ #define UDPCTL_CHECKSUM 1 /* checksum UDP packets */ #define UDPCTL_STATS 2 /* statistics (read-only) */ #define UDPCTL_MAXDGRAM 3 /* max datagram size */ #define UDPCTL_RECVSPACE 4 /* default receive buffer space */ #define UDPCTL_PCBLIST 5 /* list of PCBs for UDP sockets */ #ifdef _KERNEL #include SYSCTL_DECL(_net_inet_udp); extern struct pr_usrreqs udp_usrreqs; VNET_DECLARE(struct inpcbhead, udb); VNET_DECLARE(struct inpcbinfo, udbinfo); VNET_DECLARE(struct inpcbhead, ulitecb); VNET_DECLARE(struct inpcbinfo, ulitecbinfo); #define V_udb VNET(udb) #define V_udbinfo VNET(udbinfo) #define V_ulitecb VNET(ulitecb) #define V_ulitecbinfo VNET(ulitecbinfo) extern u_long udp_sendspace; extern u_long udp_recvspace; VNET_DECLARE(int, udp_cksum); VNET_DECLARE(int, udp_blackhole); #define V_udp_cksum VNET(udp_cksum) #define V_udp_blackhole VNET(udp_blackhole) extern int udp_log_in_vain; static __inline struct inpcbinfo * udp_get_inpcbinfo(int protocol) { return (protocol == IPPROTO_UDP) ? &V_udbinfo : &V_ulitecbinfo; } static __inline struct inpcbhead * udp_get_pcblist(int protocol) { return (protocol == IPPROTO_UDP) ? &V_udb : &V_ulitecb; } int udp_newudpcb(struct inpcb *); void udp_discardcb(struct udpcb *); void udp_ctlinput(int, struct sockaddr *, void *); void udplite_ctlinput(int, struct sockaddr *, void *); int udp_ctloutput(struct socket *, struct sockopt *); void udp_init(void); void udplite_init(void); -#ifdef VIMAGE -void udp_destroy(void); -void udplite_destroy(void); -#endif int udp_input(struct mbuf **, int *, int); void udplite_input(struct mbuf *, int); struct inpcb *udp_notify(struct inpcb *inp, int errno); int udp_shutdown(struct socket *so); int udp_set_kernel_tunneling(struct socket *so, udp_tun_func_t f, void *ctx); #endif /* _KERNEL */ #endif /* _NETINET_UDP_VAR_H_ */ Index: projects/vnet/sys/netinet6/in6.c =================================================================== --- projects/vnet/sys/netinet6/in6.c (revision 294565) +++ projects/vnet/sys/netinet6/in6.c (revision 294566) @@ -1,2524 +1,2524 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $ */ /*- * Copyright (c) 1982, 1986, 1991, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)in.c 8.2 (Berkeley) 11/15/93 */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include "opt_inet.h" #include "opt_inet6.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix); #define V_icmp6_nodeinfo_oldmcprefix VNET(icmp6_nodeinfo_oldmcprefix) /* * Definitions of some costant IP6 addresses. */ const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT; const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT; const struct in6_addr in6addr_nodelocal_allnodes = IN6ADDR_NODELOCAL_ALLNODES_INIT; const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT; const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT; const struct in6_addr in6addr_linklocal_allv2routers = IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT; const struct in6_addr in6mask0 = IN6MASK0; const struct in6_addr in6mask32 = IN6MASK32; const struct in6_addr in6mask64 = IN6MASK64; const struct in6_addr in6mask96 = IN6MASK96; const struct in6_addr in6mask128 = IN6MASK128; const struct sockaddr_in6 sa6_any = { sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 }; static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *, struct in6_aliasreq *, int); static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *); static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *, struct in6_ifaddr *, int); static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *, struct in6_aliasreq *, int flags); static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *, struct in6_ifaddr *, int, int); static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *, struct in6_ifaddr *, int); #define ifa2ia6(ifa) ((struct in6_ifaddr *)(ifa)) #define ia62ifa(ia6) (&((ia6)->ia_ifa)) void in6_newaddrmsg(struct in6_ifaddr *ia, int cmd) { struct sockaddr_dl gateway; struct sockaddr_in6 mask, addr; struct rtentry rt; /* * initialize for rtmsg generation */ bzero(&gateway, sizeof(gateway)); gateway.sdl_len = sizeof(gateway); gateway.sdl_family = AF_LINK; bzero(&rt, sizeof(rt)); rt.rt_gateway = (struct sockaddr *)&gateway; memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask)); memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr)); rt_mask(&rt) = (struct sockaddr *)&mask; rt_key(&rt) = (struct sockaddr *)&addr; rt.rt_flags = RTF_HOST | RTF_STATIC; if (cmd == RTM_ADD) rt.rt_flags |= RTF_UP; /* Announce arrival of local address to all FIBs. */ rt_newaddrmsg(cmd, &ia->ia_ifa, 0, &rt); } int in6_mask2len(struct in6_addr *mask, u_char *lim0) { int x = 0, y; u_char *lim = lim0, *p; /* ignore the scope_id part */ if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask)) lim = (u_char *)mask + sizeof(*mask); for (p = (u_char *)mask; p < lim; x++, p++) { if (*p != 0xff) break; } y = 0; if (p < lim) { for (y = 0; y < 8; y++) { if ((*p & (0x80 >> y)) == 0) break; } } /* * when the limit pointer is given, do a stricter check on the * remaining bits. */ if (p < lim) { if (y != 0 && (*p & (0x00ff >> y)) != 0) return (-1); for (p = p + 1; p < lim; p++) if (*p != 0) return (-1); } return x * 8 + y; } #ifdef COMPAT_FREEBSD32 struct in6_ndifreq32 { char ifname[IFNAMSIZ]; uint32_t ifindex; }; #define SIOCGDEFIFACE32_IN6 _IOWR('i', 86, struct in6_ndifreq32) #endif int in6_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td) { struct in6_ifreq *ifr = (struct in6_ifreq *)data; struct in6_ifaddr *ia = NULL; struct in6_aliasreq *ifra = (struct in6_aliasreq *)data; struct sockaddr_in6 *sa6; int carp_attached = 0; int error; u_long ocmd = cmd; /* * Compat to make pre-10.x ifconfig(8) operable. */ if (cmd == OSIOCAIFADDR_IN6) cmd = SIOCAIFADDR_IN6; switch (cmd) { case SIOCGETSGCNT_IN6: case SIOCGETMIFCNT_IN6: /* * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c. * We cannot see how that would be needed, so do not adjust the * KPI blindly; more likely should clean up the IPv4 variant. */ return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP); } switch (cmd) { case SIOCAADDRCTL_POLICY: case SIOCDADDRCTL_POLICY: if (td != NULL) { error = priv_check(td, PRIV_NETINET_ADDRCTRL6); if (error) return (error); } return (in6_src_ioctl(cmd, data)); } if (ifp == NULL) return (EOPNOTSUPP); switch (cmd) { case SIOCSNDFLUSH_IN6: case SIOCSPFXFLUSH_IN6: case SIOCSRTRFLUSH_IN6: case SIOCSDEFIFACE_IN6: case SIOCSIFINFO_FLAGS: case SIOCSIFINFO_IN6: if (td != NULL) { error = priv_check(td, PRIV_NETINET_ND6); if (error) return (error); } /* FALLTHROUGH */ case OSIOCGIFINFO_IN6: case SIOCGIFINFO_IN6: case SIOCGNBRINFO_IN6: case SIOCGDEFIFACE_IN6: return (nd6_ioctl(cmd, data, ifp)); #ifdef COMPAT_FREEBSD32 case SIOCGDEFIFACE32_IN6: { struct in6_ndifreq ndif; struct in6_ndifreq32 *ndif32; error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif, ifp); if (error) return (error); ndif32 = (struct in6_ndifreq32 *)data; ndif32->ifindex = ndif.ifindex; return (0); } #endif } switch (cmd) { case SIOCSIFPREFIX_IN6: case SIOCDIFPREFIX_IN6: case SIOCAIFPREFIX_IN6: case SIOCCIFPREFIX_IN6: case SIOCSGIFPREFIX_IN6: case SIOCGIFPREFIX_IN6: log(LOG_NOTICE, "prefix ioctls are now invalidated. " "please use ifconfig.\n"); return (EOPNOTSUPP); } switch (cmd) { case SIOCSSCOPE6: if (td != NULL) { error = priv_check(td, PRIV_NETINET_SCOPE6); if (error) return (error); } /* FALLTHROUGH */ case SIOCGSCOPE6: case SIOCGSCOPE6DEF: return (scope6_ioctl(cmd, data, ifp)); } /* * Find address for this interface, if it exists. * * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation * only, and used the first interface address as the target of other * operations (without checking ifra_addr). This was because netinet * code/API assumed at most 1 interface address per interface. * Since IPv6 allows a node to assign multiple addresses * on a single interface, we almost always look and check the * presence of ifra_addr, and reject invalid ones here. * It also decreases duplicated code among SIOC*_IN6 operations. */ switch (cmd) { case SIOCAIFADDR_IN6: case SIOCSIFPHYADDR_IN6: sa6 = &ifra->ifra_addr; break; case SIOCSIFADDR_IN6: case SIOCGIFADDR_IN6: case SIOCSIFDSTADDR_IN6: case SIOCSIFNETMASK_IN6: case SIOCGIFDSTADDR_IN6: case SIOCGIFNETMASK_IN6: case SIOCDIFADDR_IN6: case SIOCGIFPSRCADDR_IN6: case SIOCGIFPDSTADDR_IN6: case SIOCGIFAFLAG_IN6: case SIOCSNDFLUSH_IN6: case SIOCSPFXFLUSH_IN6: case SIOCSRTRFLUSH_IN6: case SIOCGIFALIFETIME_IN6: case SIOCSIFALIFETIME_IN6: case SIOCGIFSTAT_IN6: case SIOCGIFSTAT_ICMP6: sa6 = &ifr->ifr_addr; break; case SIOCSIFADDR: case SIOCSIFBRDADDR: case SIOCSIFDSTADDR: case SIOCSIFNETMASK: /* * Although we should pass any non-INET6 ioctl requests * down to driver, we filter some legacy INET requests. * Drivers trust SIOCSIFADDR et al to come from an already * privileged layer, and do not perform any credentials * checks or input validation. */ return (EINVAL); default: sa6 = NULL; break; } if (sa6 && sa6->sin6_family == AF_INET6) { if (sa6->sin6_scope_id != 0) error = sa6_embedscope(sa6, 0); else error = in6_setscope(&sa6->sin6_addr, ifp, NULL); if (error != 0) return (error); if (td != NULL && (error = prison_check_ip6(td->td_ucred, &sa6->sin6_addr)) != 0) return (error); ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr); } else ia = NULL; switch (cmd) { case SIOCSIFADDR_IN6: case SIOCSIFDSTADDR_IN6: case SIOCSIFNETMASK_IN6: /* * Since IPv6 allows a node to assign multiple addresses * on a single interface, SIOCSIFxxx ioctls are deprecated. */ /* we decided to obsolete this command (20000704) */ error = EINVAL; goto out; case SIOCDIFADDR_IN6: /* * for IPv4, we look for existing in_ifaddr here to allow * "ifconfig if0 delete" to remove the first IPv4 address on * the interface. For IPv6, as the spec allows multiple * interface address from the day one, we consider "remove the * first one" semantics to be not preferable. */ if (ia == NULL) { error = EADDRNOTAVAIL; goto out; } /* FALLTHROUGH */ case SIOCAIFADDR_IN6: /* * We always require users to specify a valid IPv6 address for * the corresponding operation. */ if (ifra->ifra_addr.sin6_family != AF_INET6 || ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) { error = EAFNOSUPPORT; goto out; } if (td != NULL) { error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ? PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR); if (error) goto out; } /* FALLTHROUGH */ case SIOCGIFSTAT_IN6: case SIOCGIFSTAT_ICMP6: if (ifp->if_afdata[AF_INET6] == NULL) { error = EPFNOSUPPORT; goto out; } break; case SIOCGIFADDR_IN6: /* This interface is basically deprecated. use SIOCGIFCONF. */ /* FALLTHROUGH */ case SIOCGIFAFLAG_IN6: case SIOCGIFNETMASK_IN6: case SIOCGIFDSTADDR_IN6: case SIOCGIFALIFETIME_IN6: /* must think again about its semantics */ if (ia == NULL) { error = EADDRNOTAVAIL; goto out; } break; case SIOCSIFALIFETIME_IN6: { struct in6_addrlifetime *lt; if (td != NULL) { error = priv_check(td, PRIV_NETINET_ALIFETIME6); if (error) goto out; } if (ia == NULL) { error = EADDRNOTAVAIL; goto out; } /* sanity for overflow - beware unsigned */ lt = &ifr->ifr_ifru.ifru_lifetime; if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME && lt->ia6t_vltime + time_uptime < time_uptime) { error = EINVAL; goto out; } if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME && lt->ia6t_pltime + time_uptime < time_uptime) { error = EINVAL; goto out; } break; } } switch (cmd) { case SIOCGIFADDR_IN6: ifr->ifr_addr = ia->ia_addr; if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0) goto out; break; case SIOCGIFDSTADDR_IN6: if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { error = EINVAL; goto out; } /* * XXX: should we check if ifa_dstaddr is NULL and return * an error? */ ifr->ifr_dstaddr = ia->ia_dstaddr; if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0) goto out; break; case SIOCGIFNETMASK_IN6: ifr->ifr_addr = ia->ia_prefixmask; break; case SIOCGIFAFLAG_IN6: ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags; break; case SIOCGIFSTAT_IN6: COUNTER_ARRAY_COPY(((struct in6_ifextra *) ifp->if_afdata[AF_INET6])->in6_ifstat, &ifr->ifr_ifru.ifru_stat, sizeof(struct in6_ifstat) / sizeof(uint64_t)); break; case SIOCGIFSTAT_ICMP6: COUNTER_ARRAY_COPY(((struct in6_ifextra *) ifp->if_afdata[AF_INET6])->icmp6_ifstat, &ifr->ifr_ifru.ifru_icmp6stat, sizeof(struct icmp6_ifstat) / sizeof(uint64_t)); break; case SIOCGIFALIFETIME_IN6: ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime; if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) { time_t maxexpire; struct in6_addrlifetime *retlt = &ifr->ifr_ifru.ifru_lifetime; /* * XXX: adjust expiration time assuming time_t is * signed. */ maxexpire = (-1) & ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); if (ia->ia6_lifetime.ia6t_vltime < maxexpire - ia->ia6_updatetime) { retlt->ia6t_expire = ia->ia6_updatetime + ia->ia6_lifetime.ia6t_vltime; } else retlt->ia6t_expire = maxexpire; } if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) { time_t maxexpire; struct in6_addrlifetime *retlt = &ifr->ifr_ifru.ifru_lifetime; /* * XXX: adjust expiration time assuming time_t is * signed. */ maxexpire = (-1) & ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); if (ia->ia6_lifetime.ia6t_pltime < maxexpire - ia->ia6_updatetime) { retlt->ia6t_preferred = ia->ia6_updatetime + ia->ia6_lifetime.ia6t_pltime; } else retlt->ia6t_preferred = maxexpire; } break; case SIOCSIFALIFETIME_IN6: ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime; /* for sanity */ if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) { ia->ia6_lifetime.ia6t_expire = time_uptime + ia->ia6_lifetime.ia6t_vltime; } else ia->ia6_lifetime.ia6t_expire = 0; if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) { ia->ia6_lifetime.ia6t_preferred = time_uptime + ia->ia6_lifetime.ia6t_pltime; } else ia->ia6_lifetime.ia6t_preferred = 0; break; case SIOCAIFADDR_IN6: { struct nd_prefixctl pr0; struct nd_prefix *pr; /* * first, make or update the interface address structure, * and link it to the list. */ if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0) goto out; if (ia != NULL) ifa_free(&ia->ia_ifa); if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr)) == NULL) { /* * this can happen when the user specify the 0 valid * lifetime. */ break; } if (cmd == ocmd && ifra->ifra_vhid > 0) { if (carp_attach_p != NULL) error = (*carp_attach_p)(&ia->ia_ifa, ifra->ifra_vhid); else error = EPROTONOSUPPORT; if (error) goto out; else carp_attached = 1; } /* * then, make the prefix on-link on the interface. * XXX: we'd rather create the prefix before the address, but * we need at least one address to install the corresponding * interface route, so we configure the address first. */ /* * convert mask to prefix length (prefixmask has already * been validated in in6_update_ifa(). */ bzero(&pr0, sizeof(pr0)); pr0.ndpr_ifp = ifp; pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr, NULL); if (pr0.ndpr_plen == 128) { /* we don't need to install a host route. */ goto aifaddr_out; } pr0.ndpr_prefix = ifra->ifra_addr; /* apply the mask for safety. */ IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr, &ifra->ifra_prefixmask.sin6_addr); /* * XXX: since we don't have an API to set prefix (not address) * lifetimes, we just use the same lifetimes as addresses. * The (temporarily) installed lifetimes can be overridden by * later advertised RAs (when accept_rtadv is non 0), which is * an intended behavior. */ pr0.ndpr_raf_onlink = 1; /* should be configurable? */ pr0.ndpr_raf_auto = ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0); pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime; pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime; /* add the prefix if not yet. */ if ((pr = nd6_prefix_lookup(&pr0)) == NULL) { /* * nd6_prelist_add will install the corresponding * interface route. */ if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) { if (carp_attached) (*carp_detach_p)(&ia->ia_ifa); goto out; } if (pr == NULL) { if (carp_attached) (*carp_detach_p)(&ia->ia_ifa); log(LOG_ERR, "nd6_prelist_add succeeded but " "no prefix\n"); error = EINVAL; goto out; } } /* relate the address to the prefix */ if (ia->ia6_ndpr == NULL) { ia->ia6_ndpr = pr; pr->ndpr_refcnt++; /* * If this is the first autoconf address from the * prefix, create a temporary address as well * (when required). */ if ((ia->ia6_flags & IN6_IFF_AUTOCONF) && V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) { int e; if ((e = in6_tmpifadd(ia, 1, 0)) != 0) { log(LOG_NOTICE, "in6_control: failed " "to create a temporary address, " "errno=%d\n", e); } } } /* * this might affect the status of autoconfigured addresses, * that is, this address might make other addresses detached. */ pfxlist_onlink_check(); aifaddr_out: if (error != 0 || ia == NULL) break; /* * Try to clear the flag when a new IPv6 address is added * onto an IFDISABLED interface and it succeeds. */ if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) { struct in6_ndireq nd; memset(&nd, 0, sizeof(nd)); nd.ndi.flags = ND_IFINFO(ifp)->flags; nd.ndi.flags &= ~ND6_IFF_IFDISABLED; if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0) log(LOG_NOTICE, "SIOCAIFADDR_IN6: " "SIOCSIFINFO_FLAGS for -ifdisabled " "failed."); /* * Ignore failure of clearing the flag intentionally. * The failure means address duplication was detected. */ } EVENTHANDLER_INVOKE(ifaddr_event, ifp); break; } case SIOCDIFADDR_IN6: { struct nd_prefix *pr; /* * If the address being deleted is the only one that owns * the corresponding prefix, expire the prefix as well. * XXX: theoretically, we don't have to worry about such * relationship, since we separate the address management * and the prefix management. We do this, however, to provide * as much backward compatibility as possible in terms of * the ioctl operation. * Note that in6_purgeaddr() will decrement ndpr_refcnt. */ pr = ia->ia6_ndpr; in6_purgeaddr(&ia->ia_ifa); if (pr && pr->ndpr_refcnt == 0) prelist_remove(pr); EVENTHANDLER_INVOKE(ifaddr_event, ifp); break; } default: if (ifp->if_ioctl == NULL) { error = EOPNOTSUPP; goto out; } error = (*ifp->if_ioctl)(ifp, cmd, data); goto out; } error = 0; out: if (ia != NULL) ifa_free(&ia->ia_ifa); return (error); } /* * Join necessary multicast groups. Factored out from in6_update_ifa(). * This entire work should only be done once, for the default FIB. */ static int in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra, struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol) { char ip6buf[INET6_ADDRSTRLEN]; struct in6_addr mltaddr; struct in6_multi_mship *imm; int delay, error; KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__)); /* Join solicited multicast addr for new host id. */ bzero(&mltaddr, sizeof(struct in6_addr)); mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL; mltaddr.s6_addr32[2] = htonl(1); mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3]; mltaddr.s6_addr8[12] = 0xff; if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) { /* XXX: should not happen */ log(LOG_ERR, "%s: in6_setscope failed\n", __func__); goto cleanup; } delay = error = 0; if ((flags & IN6_IFAUPDATE_DADDELAY)) { /* * We need a random delay for DAD on the address being * configured. It also means delaying transmission of the * corresponding MLD report to avoid report collision. * [RFC 4861, Section 6.3.7] */ delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz); } imm = in6_joingroup(ifp, &mltaddr, &error, delay); if (imm == NULL) { nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s " "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp), error)); goto cleanup; } LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain); *in6m_sol = imm->i6mm_maddr; /* * Join link-local all-nodes address. */ mltaddr = in6addr_linklocal_allnodes; if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) goto cleanup; /* XXX: should not fail */ imm = in6_joingroup(ifp, &mltaddr, &error, 0); if (imm == NULL) { nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s " "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp), error)); goto cleanup; } LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain); /* * Join node information group address. */ delay = 0; if ((flags & IN6_IFAUPDATE_DADDELAY)) { /* * The spec does not say anything about delay for this group, * but the same logic should apply. */ delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz); } if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) { /* XXX jinmei */ imm = in6_joingroup(ifp, &mltaddr, &error, delay); if (imm == NULL) nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s " "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp), error)); /* XXX not very fatal, go on... */ else LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain); } if (V_icmp6_nodeinfo_oldmcprefix && in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) { imm = in6_joingroup(ifp, &mltaddr, &error, delay); if (imm == NULL) nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s " "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp), error)); /* XXX not very fatal, go on... */ else LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain); } /* * Join interface-local all-nodes address. * (ff01::1%ifN, and ff01::%ifN/32) */ mltaddr = in6addr_nodelocal_allnodes; if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) goto cleanup; /* XXX: should not fail */ imm = in6_joingroup(ifp, &mltaddr, &error, 0); if (imm == NULL) { nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s " "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp), error)); goto cleanup; } LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain); cleanup: return (error); } /* * Update parameters of an IPv6 interface address. * If necessary, a new entry is created and linked into address chains. * This function is separated from in6_control(). */ int in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, struct in6_ifaddr *ia, int flags) { int error, hostIsNew = 0; if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0) return (error); if (ia == NULL) { hostIsNew = 1; if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL) return (ENOBUFS); } error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags); if (error != 0) { if (hostIsNew != 0) { in6_unlink_ifa(ia, ifp); ifa_free(&ia->ia_ifa); } return (error); } if (hostIsNew) error = in6_broadcast_ifa(ifp, ifra, ia, flags); return (error); } /* * Fill in basic IPv6 address request info. */ void in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr, const struct in6_addr *mask) { memset(ifra, 0, sizeof(struct in6_aliasreq)); ifra->ifra_addr.sin6_family = AF_INET6; ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6); if (addr != NULL) ifra->ifra_addr.sin6_addr = *addr; ifra->ifra_prefixmask.sin6_family = AF_INET6; ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6); if (mask != NULL) ifra->ifra_prefixmask.sin6_addr = *mask; } static int in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra, struct in6_ifaddr *ia, int flags) { int plen = -1; struct sockaddr_in6 dst6; struct in6_addrlifetime *lt; char ip6buf[INET6_ADDRSTRLEN]; /* Validate parameters */ if (ifp == NULL || ifra == NULL) /* this maybe redundant */ return (EINVAL); /* * The destination address for a p2p link must have a family * of AF_UNSPEC or AF_INET6. */ if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && ifra->ifra_dstaddr.sin6_family != AF_INET6 && ifra->ifra_dstaddr.sin6_family != AF_UNSPEC) return (EAFNOSUPPORT); /* * Validate address */ if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) || ifra->ifra_addr.sin6_family != AF_INET6) return (EINVAL); /* * validate ifra_prefixmask. don't check sin6_family, netmask * does not carry fields other than sin6_len. */ if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6)) return (EINVAL); /* * Because the IPv6 address architecture is classless, we require * users to specify a (non 0) prefix length (mask) for a new address. * We also require the prefix (when specified) mask is valid, and thus * reject a non-consecutive mask. */ if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0) return (EINVAL); if (ifra->ifra_prefixmask.sin6_len != 0) { plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr, (u_char *)&ifra->ifra_prefixmask + ifra->ifra_prefixmask.sin6_len); if (plen <= 0) return (EINVAL); } else { /* * In this case, ia must not be NULL. We just use its prefix * length. */ plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); } /* * If the destination address on a p2p interface is specified, * and the address is a scoped one, validate/set the scope * zone identifier. */ dst6 = ifra->ifra_dstaddr; if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 && (dst6.sin6_family == AF_INET6)) { struct in6_addr in6_tmp; u_int32_t zoneid; in6_tmp = dst6.sin6_addr; if (in6_setscope(&in6_tmp, ifp, &zoneid)) return (EINVAL); /* XXX: should be impossible */ if (dst6.sin6_scope_id != 0) { if (dst6.sin6_scope_id != zoneid) return (EINVAL); } else /* user omit to specify the ID. */ dst6.sin6_scope_id = zoneid; /* convert into the internal form */ if (sa6_embedscope(&dst6, 0)) return (EINVAL); /* XXX: should be impossible */ } /* Modify original ifra_dstaddr to reflect changes */ ifra->ifra_dstaddr = dst6; /* * The destination address can be specified only for a p2p or a * loopback interface. If specified, the corresponding prefix length * must be 128. */ if (ifra->ifra_dstaddr.sin6_family == AF_INET6) { if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) { /* XXX: noisy message */ nd6log((LOG_INFO, "in6_update_ifa: a destination can " "be specified for a p2p or a loopback IF only\n")); return (EINVAL); } if (plen != 128) { nd6log((LOG_INFO, "in6_update_ifa: prefixlen should " "be 128 when dstaddr is specified\n")); return (EINVAL); } } /* lifetime consistency check */ lt = &ifra->ifra_lifetime; if (lt->ia6t_pltime > lt->ia6t_vltime) return (EINVAL); if (lt->ia6t_vltime == 0) { /* * the following log might be noisy, but this is a typical * configuration mistake or a tool's bug. */ nd6log((LOG_INFO, "in6_update_ifa: valid lifetime is 0 for %s\n", ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr))); if (ia == NULL) return (0); /* there's nothing to do */ } /* Check prefix mask */ if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) { /* * We prohibit changing the prefix length of an existing * address, because * + such an operation should be rare in IPv6, and * + the operation would confuse prefix management. */ if (ia->ia_prefixmask.sin6_len != 0 && in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) { nd6log((LOG_INFO, "in6_validate_ifa: the prefix length " "of an existing %s address should not be changed\n", ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); return (EINVAL); } } return (0); } /* * Allocate a new ifaddr and link it into chains. */ static struct in6_ifaddr * in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags) { struct in6_ifaddr *ia; /* * When in6_alloc_ifa() is called in a process of a received * RA, it is called under an interrupt context. So, we should * call malloc with M_NOWAIT. */ ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT); if (ia == NULL) return (NULL); LIST_INIT(&ia->ia6_memberships); /* Initialize the address and masks, and put time stamp */ ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr; ia->ia_addr.sin6_family = AF_INET6; ia->ia_addr.sin6_len = sizeof(ia->ia_addr); /* XXX: Can we assign ,sin6_addr and skip the rest? */ ia->ia_addr = ifra->ifra_addr; ia->ia6_createtime = time_uptime; if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) { /* * Some functions expect that ifa_dstaddr is not * NULL for p2p interfaces. */ ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr; } else { ia->ia_ifa.ifa_dstaddr = NULL; } /* set prefix mask if any */ ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask; if (ifra->ifra_prefixmask.sin6_len != 0) { ia->ia_prefixmask.sin6_family = AF_INET6; ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len; ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr; } ia->ia_ifp = ifp; ifa_ref(&ia->ia_ifa); /* if_addrhead */ IF_ADDR_WLOCK(ifp); TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); IF_ADDR_WUNLOCK(ifp); ifa_ref(&ia->ia_ifa); /* in6_ifaddrhead */ IN6_IFADDR_WLOCK(); TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link); LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash); IN6_IFADDR_WUNLOCK(); return (ia); } /* * Update/configure interface address parameters: * * 1) Update lifetime * 2) Update interface metric ad flags * 3) Notify other subsystems */ static int in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra, struct in6_ifaddr *ia, int hostIsNew, int flags) { int error; /* update timestamp */ ia->ia6_updatetime = time_uptime; /* * Set lifetimes. We do not refer to ia6t_expire and ia6t_preferred * to see if the address is deprecated or invalidated, but initialize * these members for applications. */ ia->ia6_lifetime = ifra->ifra_lifetime; if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) { ia->ia6_lifetime.ia6t_expire = time_uptime + ia->ia6_lifetime.ia6t_vltime; } else ia->ia6_lifetime.ia6t_expire = 0; if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) { ia->ia6_lifetime.ia6t_preferred = time_uptime + ia->ia6_lifetime.ia6t_pltime; } else ia->ia6_lifetime.ia6t_preferred = 0; /* * backward compatibility - if IN6_IFF_DEPRECATED is set from the * userland, make it deprecated. */ if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) { ia->ia6_lifetime.ia6t_pltime = 0; ia->ia6_lifetime.ia6t_preferred = time_uptime; } /* * configure address flags. */ ia->ia6_flags = ifra->ifra_flags; /* * Make the address tentative before joining multicast addresses, * so that corresponding MLD responses would not have a tentative * source address. */ ia->ia6_flags &= ~IN6_IFF_DUPLICATED; /* safety */ /* * DAD should be performed for an new address or addresses on * an interface with ND6_IFF_IFDISABLED. */ if (in6if_do_dad(ifp) && (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED))) ia->ia6_flags |= IN6_IFF_TENTATIVE; /* notify other subsystems */ error = in6_notify_ifa(ifp, ia, ifra, hostIsNew); return (error); } /* * Do link-level ifa job: * 1) Add lle entry for added address * 2) Notifies routing socket users about new address * 3) join appropriate multicast group * 4) start DAD if enabled */ static int in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, struct in6_ifaddr *ia, int flags) { struct in6_multi *in6m_sol; int error = 0; /* Add local address to lltable, if necessary (ex. on p2p link). */ if ((error = nd6_add_ifa_lle(ia)) != 0) { in6_purgeaddr(&ia->ia_ifa); ifa_free(&ia->ia_ifa); return (error); } /* Join necessary multicast groups. */ in6m_sol = NULL; if ((ifp->if_flags & IFF_MULTICAST) != 0) { error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol); if (error != 0) { in6_purgeaddr(&ia->ia_ifa); ifa_free(&ia->ia_ifa); return (error); } } /* Perform DAD, if the address is TENTATIVE. */ if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) { int delay, mindelay, maxdelay; delay = 0; if ((flags & IN6_IFAUPDATE_DADDELAY)) { /* * We need to impose a delay before sending an NS * for DAD. Check if we also needed a delay for the * corresponding MLD message. If we did, the delay * should be larger than the MLD delay (this could be * relaxed a bit, but this simple logic is at least * safe). * XXX: Break data hiding guidelines and look at * state for the solicited multicast group. */ mindelay = 0; if (in6m_sol != NULL && in6m_sol->in6m_state == MLD_REPORTING_MEMBER) { mindelay = in6m_sol->in6m_timer; } maxdelay = MAX_RTR_SOLICITATION_DELAY * hz; if (maxdelay - mindelay == 0) delay = 0; else { delay = (arc4random() % (maxdelay - mindelay)) + mindelay; } } nd6_dad_start((struct ifaddr *)ia, delay); } in6_newaddrmsg(ia, RTM_ADD); ifa_free(&ia->ia_ifa); return (error); } void in6_purgeaddr(struct ifaddr *ifa) { struct ifnet *ifp = ifa->ifa_ifp; struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa; struct in6_multi_mship *imm; int plen, error; if (ifa->ifa_carp) (*carp_detach_p)(ifa); /* * Remove the loopback route to the interface address. * The check for the current setting of "nd6_useloopback" * is not needed. */ if (ia->ia_flags & IFA_RTSELF) { error = ifa_del_loopback_route((struct ifaddr *)ia, (struct sockaddr *)&ia->ia_addr); if (error == 0) ia->ia_flags &= ~IFA_RTSELF; } /* stop DAD processing */ nd6_dad_stop(ifa); /* Leave multicast groups. */ while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) { LIST_REMOVE(imm, i6mm_chain); in6_leavegroup(imm); } plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */ if ((ia->ia_flags & IFA_ROUTE) && plen == 128) { error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags | (ia->ia_dstaddr.sin6_family == AF_INET6) ? RTF_HOST : 0); if (error != 0) log(LOG_INFO, "%s: err=%d, destination address delete " "failed\n", __func__, error); ia->ia_flags &= ~IFA_ROUTE; } in6_newaddrmsg(ia, RTM_DELETE); in6_unlink_ifa(ia, ifp); } static void in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp) { char ip6buf[INET6_ADDRSTRLEN]; int remove_lle; IF_ADDR_WLOCK(ifp); TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link); IF_ADDR_WUNLOCK(ifp); ifa_free(&ia->ia_ifa); /* if_addrhead */ /* * Defer the release of what might be the last reference to the * in6_ifaddr so that it can't be freed before the remainder of the * cleanup. */ IN6_IFADDR_WLOCK(); TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link); LIST_REMOVE(ia, ia6_hash); IN6_IFADDR_WUNLOCK(); /* * Release the reference to the base prefix. There should be a * positive reference. */ remove_lle = 0; if (ia->ia6_ndpr == NULL) { nd6log((LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address " "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia)))); } else { ia->ia6_ndpr->ndpr_refcnt--; /* Do not delete lles within prefix if refcont != 0 */ if (ia->ia6_ndpr->ndpr_refcnt == 0) remove_lle = 1; ia->ia6_ndpr = NULL; } nd6_rem_ifa_lle(ia, remove_lle); /* * Also, if the address being removed is autoconf'ed, call * pfxlist_onlink_check() since the release might affect the status of * other (detached) addresses. */ if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) { pfxlist_onlink_check(); } ifa_free(&ia->ia_ifa); /* in6_ifaddrhead */ } /* * Notifies other subsystems about address change/arrival: * 1) Notifies device handler on the first IPv6 address assignment * 2) Handle routing table changes for P2P links and route * 3) Handle routing table changes for address host route */ static int in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia, struct in6_aliasreq *ifra, int hostIsNew) { int error = 0, plen, ifacount = 0; struct ifaddr *ifa; struct sockaddr_in6 *pdst; char ip6buf[INET6_ADDRSTRLEN]; /* * Give the interface a chance to initialize * if this is its first address, */ if (hostIsNew != 0) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; ifacount++; } IF_ADDR_RUNLOCK(ifp); } if (ifacount <= 1 && ifp->if_ioctl) { error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia); if (error) return (error); } /* * If a new destination address is specified, scrub the old one and * install the new destination. Note that the interface must be * p2p or loopback. */ pdst = &ifra->ifra_dstaddr; if (pdst->sin6_family == AF_INET6 && !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) { if ((ia->ia_flags & IFA_ROUTE) != 0 && (rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0)) { nd6log((LOG_ERR, "in6_update_ifa_internal: failed to " "remove a route to the old destination: %s\n", ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); /* proceed anyway... */ } else ia->ia_flags &= ~IFA_ROUTE; ia->ia_dstaddr = *pdst; } /* * If a new destination address is specified for a point-to-point * interface, install a route to the destination as an interface * direct route. * XXX: the logic below rejects assigning multiple addresses on a p2p * interface that share the same destination. */ plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */ if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) { int rtflags = RTF_UP | RTF_HOST; /* * Handle the case for ::1 . */ if (ifp->if_flags & IFF_LOOPBACK) ia->ia_flags |= IFA_RTSELF; error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags); if (error) return (error); ia->ia_flags |= IFA_ROUTE; } /* * add a loopback route to self if not exists */ if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) { error = ifa_add_loopback_route((struct ifaddr *)ia, (struct sockaddr *)&ia->ia_addr); if (error == 0) ia->ia_flags |= IFA_RTSELF; } return (error); } /* * Find an IPv6 interface link-local address specific to an interface. * ifaddr is returned referenced. */ struct in6_ifaddr * in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags) { struct ifaddr *ifa; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) { if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0) continue; ifa_ref(ifa); break; } } IF_ADDR_RUNLOCK(ifp); return ((struct in6_ifaddr *)ifa); } /* * find the internet address corresponding to a given address. * ifaddr is returned referenced. */ struct in6_ifaddr * in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid) { struct rm_priotracker in6_ifa_tracker; struct in6_ifaddr *ia; IN6_IFADDR_RLOCK(&in6_ifa_tracker); LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) { if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) { if (zoneid != 0 && zoneid != ia->ia_addr.sin6_scope_id) continue; ifa_ref(&ia->ia_ifa); break; } } IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (ia); } /* * find the internet address corresponding to a given interface and address. * ifaddr is returned referenced. */ struct in6_ifaddr * in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr) { struct ifaddr *ifa; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) { ifa_ref(ifa); break; } } IF_ADDR_RUNLOCK(ifp); return ((struct in6_ifaddr *)ifa); } /* * Find a link-local scoped address on ifp and return it if any. */ struct in6_ifaddr * in6ifa_llaonifp(struct ifnet *ifp) { struct sockaddr_in6 *sin6; struct ifaddr *ifa; if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) return (NULL); - if_addr_rlock(ifp); + IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) || IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) || IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr)) break; } - if_addr_runlock(ifp); + IF_ADDR_RUNLOCK(ifp); return ((struct in6_ifaddr *)ifa); } /* * Convert IP6 address to printable (loggable) representation. Caller * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long. */ static char digits[] = "0123456789abcdef"; char * ip6_sprintf(char *ip6buf, const struct in6_addr *addr) { int i, cnt = 0, maxcnt = 0, idx = 0, index = 0; char *cp; const u_int16_t *a = (const u_int16_t *)addr; const u_int8_t *d; int dcolon = 0, zero = 0; cp = ip6buf; for (i = 0; i < 8; i++) { if (*(a + i) == 0) { cnt++; if (cnt == 1) idx = i; } else if (maxcnt < cnt) { maxcnt = cnt; index = idx; cnt = 0; } } if (maxcnt < cnt) { maxcnt = cnt; index = idx; } for (i = 0; i < 8; i++) { if (dcolon == 1) { if (*a == 0) { if (i == 7) *cp++ = ':'; a++; continue; } else dcolon = 2; } if (*a == 0) { if (dcolon == 0 && *(a + 1) == 0 && i == index) { if (i == 0) *cp++ = ':'; *cp++ = ':'; dcolon = 1; } else { *cp++ = '0'; *cp++ = ':'; } a++; continue; } d = (const u_char *)a; /* Try to eliminate leading zeros in printout like in :0001. */ zero = 1; *cp = digits[*d >> 4]; if (*cp != '0') { zero = 0; cp++; } *cp = digits[*d++ & 0xf]; if (zero == 0 || (*cp != '0')) { zero = 0; cp++; } *cp = digits[*d >> 4]; if (zero == 0 || (*cp != '0')) { zero = 0; cp++; } *cp++ = digits[*d & 0xf]; *cp++ = ':'; a++; } *--cp = '\0'; return (ip6buf); } int in6_localaddr(struct in6_addr *in6) { struct rm_priotracker in6_ifa_tracker; struct in6_ifaddr *ia; if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6)) return 1; IN6_IFADDR_RLOCK(&in6_ifa_tracker); TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr, &ia->ia_prefixmask.sin6_addr)) { IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return 1; } } IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (0); } /* * Return 1 if an internet address is for the local host and configured * on one of its interfaces. */ int in6_localip(struct in6_addr *in6) { struct rm_priotracker in6_ifa_tracker; struct in6_ifaddr *ia; IN6_IFADDR_RLOCK(&in6_ifa_tracker); LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) { if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) { IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (1); } } IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (0); } /* * Return 1 if an internet address is configured on an interface. */ int in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr) { struct in6_addr in6; struct ifaddr *ifa; struct in6_ifaddr *ia6; in6 = *addr; if (in6_clearscope(&in6)) return (0); in6_setscope(&in6, ifp, NULL); IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; ia6 = (struct in6_ifaddr *)ifa; if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6)) { IF_ADDR_RUNLOCK(ifp); return (1); } } IF_ADDR_RUNLOCK(ifp); return (0); } int in6_is_addr_deprecated(struct sockaddr_in6 *sa6) { struct rm_priotracker in6_ifa_tracker; struct in6_ifaddr *ia; IN6_IFADDR_RLOCK(&in6_ifa_tracker); LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) { if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) { if (ia->ia6_flags & IN6_IFF_DEPRECATED) { IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (1); /* true */ } break; } } IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (0); /* false */ } /* * return length of part which dst and src are equal * hard coding... */ int in6_matchlen(struct in6_addr *src, struct in6_addr *dst) { int match = 0; u_char *s = (u_char *)src, *d = (u_char *)dst; u_char *lim = s + 16, r; while (s < lim) if ((r = (*d++ ^ *s++)) != 0) { while (r < 128) { match++; r <<= 1; } break; } else match += 8; return match; } /* XXX: to be scope conscious */ int in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len) { int bytelen, bitlen; /* sanity check */ if (0 > len || len > 128) { log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n", len); return (0); } bytelen = len / 8; bitlen = len % 8; if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen)) return (0); if (bitlen != 0 && p1->s6_addr[bytelen] >> (8 - bitlen) != p2->s6_addr[bytelen] >> (8 - bitlen)) return (0); return (1); } void in6_prefixlen2mask(struct in6_addr *maskp, int len) { u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff}; int bytelen, bitlen, i; /* sanity check */ if (0 > len || len > 128) { log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n", len); return; } bzero(maskp, sizeof(*maskp)); bytelen = len / 8; bitlen = len % 8; for (i = 0; i < bytelen; i++) maskp->s6_addr[i] = 0xff; if (bitlen) maskp->s6_addr[bytelen] = maskarray[bitlen - 1]; } /* * return the best address out of the same scope. if no address was * found, return the first valid address from designated IF. */ struct in6_ifaddr * in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst) { int dst_scope = in6_addrscope(dst), blen = -1, tlen; struct ifaddr *ifa; struct in6_ifaddr *besta = 0; struct in6_ifaddr *dep[2]; /* last-resort: deprecated */ dep[0] = dep[1] = NULL; /* * We first look for addresses in the same scope. * If there is one, return it. * If two or more, return one which matches the dst longest. * If none, return one of global addresses assigned other ifs. */ IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST) continue; /* XXX: is there any case to allow anycast? */ if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) continue; /* don't use this interface */ if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED) continue; if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) { if (V_ip6_use_deprecated) dep[0] = (struct in6_ifaddr *)ifa; continue; } if (dst_scope == in6_addrscope(IFA_IN6(ifa))) { /* * call in6_matchlen() as few as possible */ if (besta) { if (blen == -1) blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst); tlen = in6_matchlen(IFA_IN6(ifa), dst); if (tlen > blen) { blen = tlen; besta = (struct in6_ifaddr *)ifa; } } else besta = (struct in6_ifaddr *)ifa; } } if (besta) { ifa_ref(&besta->ia_ifa); IF_ADDR_RUNLOCK(ifp); return (besta); } TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST) continue; /* XXX: is there any case to allow anycast? */ if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) continue; /* don't use this interface */ if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED) continue; if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) { if (V_ip6_use_deprecated) dep[1] = (struct in6_ifaddr *)ifa; continue; } if (ifa != NULL) ifa_ref(ifa); IF_ADDR_RUNLOCK(ifp); return (struct in6_ifaddr *)ifa; } /* use the last-resort values, that are, deprecated addresses */ if (dep[0]) { ifa_ref((struct ifaddr *)dep[0]); IF_ADDR_RUNLOCK(ifp); return dep[0]; } if (dep[1]) { ifa_ref((struct ifaddr *)dep[1]); IF_ADDR_RUNLOCK(ifp); return dep[1]; } IF_ADDR_RUNLOCK(ifp); return NULL; } /* * perform DAD when interface becomes IFF_UP. */ void in6_if_up(struct ifnet *ifp) { struct ifaddr *ifa; struct in6_ifaddr *ia; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; ia = (struct in6_ifaddr *)ifa; if (ia->ia6_flags & IN6_IFF_TENTATIVE) { /* * The TENTATIVE flag was likely set by hand * beforehand, implicitly indicating the need for DAD. * We may be able to skip the random delay in this * case, but we impose delays just in case. */ nd6_dad_start(ifa, arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz)); } } IF_ADDR_RUNLOCK(ifp); /* * special cases, like 6to4, are handled in in6_ifattach */ in6_ifattach(ifp, NULL); } int in6if_do_dad(struct ifnet *ifp) { if ((ifp->if_flags & IFF_LOOPBACK) != 0) return (0); if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) || (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD)) return (0); /* * Our DAD routine requires the interface up and running. * However, some interfaces can be up before the RUNNING * status. Additionaly, users may try to assign addresses * before the interface becomes up (or running). * This function returns EAGAIN in that case. * The caller should mark "tentative" on the address instead of * performing DAD immediately. */ if (!((ifp->if_flags & IFF_UP) && (ifp->if_drv_flags & IFF_DRV_RUNNING))) return (EAGAIN); return (1); } /* * Calculate max IPv6 MTU through all the interfaces and store it * to in6_maxmtu. */ void in6_setmaxmtu(void) { unsigned long maxmtu = 0; struct ifnet *ifp; IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { /* this function can be called during ifnet initialization */ if (!ifp->if_afdata[AF_INET6]) continue; if ((ifp->if_flags & IFF_LOOPBACK) == 0 && IN6_LINKMTU(ifp) > maxmtu) maxmtu = IN6_LINKMTU(ifp); } IFNET_RUNLOCK_NOSLEEP(); if (maxmtu) /* update only when maxmtu is positive */ V_in6_maxmtu = maxmtu; } /* * Provide the length of interface identifiers to be used for the link attached * to the given interface. The length should be defined in "IPv6 over * xxx-link" document. Note that address architecture might also define * the length for a particular set of address prefixes, regardless of the * link type. As clarified in rfc2462bis, those two definitions should be * consistent, and those really are as of August 2004. */ int in6_if2idlen(struct ifnet *ifp) { switch (ifp->if_type) { case IFT_ETHER: /* RFC2464 */ case IFT_PROPVIRTUAL: /* XXX: no RFC. treat it as ether */ case IFT_L2VLAN: /* ditto */ case IFT_IEEE80211: /* ditto */ case IFT_BRIDGE: /* bridge(4) only does Ethernet-like links */ case IFT_INFINIBAND: return (64); case IFT_FDDI: /* RFC2467 */ return (64); case IFT_ISO88025: /* RFC2470 (IPv6 over Token Ring) */ return (64); case IFT_PPP: /* RFC2472 */ return (64); case IFT_ARCNET: /* RFC2497 */ return (64); case IFT_FRELAY: /* RFC2590 */ return (64); case IFT_IEEE1394: /* RFC3146 */ return (64); case IFT_GIF: return (64); /* draft-ietf-v6ops-mech-v2-07 */ case IFT_LOOP: return (64); /* XXX: is this really correct? */ default: /* * Unknown link type: * It might be controversial to use the today's common constant * of 64 for these cases unconditionally. For full compliance, * we should return an error in this case. On the other hand, * if we simply miss the standard for the link type or a new * standard is defined for a new link type, the IFID length * is very likely to be the common constant. As a compromise, * we always use the constant, but make an explicit notice * indicating the "unknown" case. */ printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type); return (64); } } #include struct in6_llentry { struct llentry base; }; #define IN6_LLTBL_DEFAULT_HSIZE 32 #define IN6_LLTBL_HASH(k, h) \ (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1)) /* * Do actual deallocation of @lle. * Called by LLE_FREE_LOCKED when number of references * drops to zero. */ static void in6_lltable_destroy_lle(struct llentry *lle) { LLE_WUNLOCK(lle); LLE_LOCK_DESTROY(lle); LLE_REQ_DESTROY(lle); free(lle, M_LLTABLE); } static struct llentry * in6_lltable_new(const struct in6_addr *addr6, u_int flags) { struct in6_llentry *lle; lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); if (lle == NULL) /* NB: caller generates msg */ return NULL; lle->base.r_l3addr.addr6 = *addr6; lle->base.lle_refcnt = 1; lle->base.lle_free = in6_lltable_destroy_lle; LLE_LOCK_INIT(&lle->base); LLE_REQ_INIT(&lle->base); callout_init(&lle->base.lle_timer, 1); return (&lle->base); } static int in6_lltable_match_prefix(const struct sockaddr *saddr, const struct sockaddr *smask, u_int flags, struct llentry *lle) { const struct in6_addr *addr, *mask, *lle_addr; addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr; mask = &((const struct sockaddr_in6 *)smask)->sin6_addr; lle_addr = &lle->r_l3addr.addr6; if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0) return (0); if (lle->la_flags & LLE_IFADDR) { /* * Delete LLE_IFADDR records IFF address & flag matches. * Note that addr is the interface address within prefix * being matched. */ if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) && (flags & LLE_STATIC) != 0) return (1); return (0); } /* flags & LLE_STATIC means deleting both dynamic and static entries */ if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)) return (1); return (0); } static void in6_lltable_free_entry(struct lltable *llt, struct llentry *lle) { struct ifnet *ifp; LLE_WLOCK_ASSERT(lle); KASSERT(llt != NULL, ("lltable is NULL")); /* Unlink entry from table */ if ((lle->la_flags & LLE_LINKED) != 0) { ifp = llt->llt_ifp; IF_AFDATA_WLOCK_ASSERT(ifp); lltable_unlink_entry(llt, lle); } if (callout_stop(&lle->lle_timer) > 0) LLE_REMREF(lle); llentry_free(lle); } static int in6_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) { const struct sockaddr_in6 *sin6; struct nhop6_basic nh6; struct in6_addr dst; uint32_t scopeid; int error; char ip6buf[INET6_ADDRSTRLEN]; KASSERT(l3addr->sa_family == AF_INET6, ("sin_family %d", l3addr->sa_family)); /* Our local addresses are always only installed on the default FIB. */ sin6 = (const struct sockaddr_in6 *)l3addr; in6_splitscope(&sin6->sin6_addr, &dst, &scopeid); error = fib6_lookup_nh_basic(RT_DEFAULT_FIB, &dst, scopeid, 0, 0, &nh6); if (error != 0 || (nh6.nh_flags & NHF_GATEWAY) || nh6.nh_ifp != ifp) { struct ifaddr *ifa; /* * Create an ND6 cache for an IPv6 neighbor * that is not covered by our own prefix. */ ifa = ifaof_ifpforaddr(l3addr, ifp); if (ifa != NULL) { ifa_free(ifa); return 0; } log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n", ip6_sprintf(ip6buf, &sin6->sin6_addr)); return EINVAL; } return 0; } static inline uint32_t in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize) { return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize)); } static uint32_t in6_lltable_hash(const struct llentry *lle, uint32_t hsize) { return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize)); } static void in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) { struct sockaddr_in6 *sin6; sin6 = (struct sockaddr_in6 *)sa; bzero(sin6, sizeof(*sin6)); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(*sin6); sin6->sin6_addr = lle->r_l3addr.addr6; } static inline struct llentry * in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst) { struct llentry *lle; struct llentries *lleh; u_int hashidx; hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize); lleh = &llt->lle_head[hashidx]; LIST_FOREACH(lle, lleh, lle_next) { if (lle->la_flags & LLE_DELETED) continue; if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst)) break; } return (lle); } static void in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle) { lle->la_flags |= LLE_DELETED; EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); #ifdef DIAGNOSTIC log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); #endif llentry_free(lle); } static struct llentry * in6_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) { const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr; struct ifnet *ifp = llt->llt_ifp; struct llentry *lle; KASSERT(l3addr->sa_family == AF_INET6, ("sin_family %d", l3addr->sa_family)); /* * A route that covers the given address must have * been installed 1st because we are doing a resolution, * verify this. */ if (!(flags & LLE_IFADDR) && in6_lltable_rtcheck(ifp, flags, l3addr) != 0) return (NULL); lle = in6_lltable_new(&sin6->sin6_addr, flags); if (lle == NULL) { log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); return (NULL); } lle->la_flags = flags; if ((flags & LLE_IFADDR) == LLE_IFADDR) { lltable_set_entry_addr(ifp, lle, IF_LLADDR(ifp)); lle->la_flags |= LLE_STATIC; } if ((lle->la_flags & LLE_STATIC) != 0) lle->ln_state = ND6_LLINFO_REACHABLE; return (lle); } static struct llentry * in6_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) { const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr; struct llentry *lle; IF_AFDATA_LOCK_ASSERT(llt->llt_ifp); KASSERT(l3addr->sa_family == AF_INET6, ("sin_family %d", l3addr->sa_family)); lle = in6_lltable_find_dst(llt, &sin6->sin6_addr); if (lle == NULL) return (NULL); KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) != (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X", flags)); if (flags & LLE_UNLOCKED) return (lle); if (flags & LLE_EXCLUSIVE) LLE_WLOCK(lle); else LLE_RLOCK(lle); return (lle); } static int in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle, struct sysctl_req *wr) { struct ifnet *ifp = llt->llt_ifp; /* XXX stack use */ struct { struct rt_msghdr rtm; struct sockaddr_in6 sin6; /* * ndp.c assumes that sdl is word aligned */ #ifdef __LP64__ uint32_t pad; #endif struct sockaddr_dl sdl; } ndpc; struct sockaddr_dl *sdl; int error; bzero(&ndpc, sizeof(ndpc)); /* skip deleted entries */ if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) return (0); /* Skip if jailed and not a valid IP of the prison. */ lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6); if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0) return (0); /* * produce a msg made of: * struct rt_msghdr; * struct sockaddr_in6 (IPv6) * struct sockaddr_dl; */ ndpc.rtm.rtm_msglen = sizeof(ndpc); ndpc.rtm.rtm_version = RTM_VERSION; ndpc.rtm.rtm_type = RTM_GET; ndpc.rtm.rtm_flags = RTF_UP; ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; if (V_deembed_scopeid) sa6_recoverscope(&ndpc.sin6); /* publish */ if (lle->la_flags & LLE_PUB) ndpc.rtm.rtm_flags |= RTF_ANNOUNCE; sdl = &ndpc.sdl; sdl->sdl_family = AF_LINK; sdl->sdl_len = sizeof(*sdl); sdl->sdl_alen = ifp->if_addrlen; sdl->sdl_index = ifp->if_index; sdl->sdl_type = ifp->if_type; bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); if (lle->la_expire != 0) ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire + lle->lle_remtime / hz + time_second - time_uptime; ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); if (lle->la_flags & LLE_STATIC) ndpc.rtm.rtm_flags |= RTF_STATIC; if (lle->la_flags & LLE_IFADDR) ndpc.rtm.rtm_flags |= RTF_PINNED; if (lle->ln_router != 0) ndpc.rtm.rtm_flags |= RTF_GATEWAY; ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked; /* Store state in rmx_weight value */ ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state; ndpc.rtm.rtm_index = ifp->if_index; error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc)); return (error); } static struct lltable * in6_lltattach(struct ifnet *ifp) { struct lltable *llt; llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE); llt->llt_af = AF_INET6; llt->llt_ifp = ifp; llt->llt_lookup = in6_lltable_lookup; llt->llt_alloc_entry = in6_lltable_alloc; llt->llt_delete_entry = in6_lltable_delete_entry; llt->llt_dump_entry = in6_lltable_dump_entry; llt->llt_hash = in6_lltable_hash; llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry; llt->llt_free_entry = in6_lltable_free_entry; llt->llt_match_prefix = in6_lltable_match_prefix; lltable_link(llt); return (llt); } void * in6_domifattach(struct ifnet *ifp) { struct in6_ifextra *ext; /* There are not IPv6-capable interfaces. */ switch (ifp->if_type) { case IFT_PFLOG: case IFT_PFSYNC: case IFT_USB: return (NULL); } ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK); bzero(ext, sizeof(*ext)); ext->in6_ifstat = malloc(sizeof(counter_u64_t) * sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK); COUNTER_ARRAY_ALLOC(ext->in6_ifstat, sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK); ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) * sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK); COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat, sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK); ext->nd_ifinfo = nd6_ifattach(ifp); ext->scope6_id = scope6_ifattach(ifp); ext->lltable = in6_lltattach(ifp); ext->mld_ifinfo = mld_domifattach(ifp); return ext; } int in6_domifmtu(struct ifnet *ifp) { if (ifp->if_afdata[AF_INET6] == NULL) return ifp->if_mtu; return (IN6_LINKMTU(ifp)); } void in6_domifdetach(struct ifnet *ifp, void *aux) { struct in6_ifextra *ext = (struct in6_ifextra *)aux; mld_domifdetach(ifp); scope6_ifdetach(ext->scope6_id); nd6_ifdetach(ext->nd_ifinfo); lltable_free(ext->lltable); COUNTER_ARRAY_FREE(ext->in6_ifstat, sizeof(struct in6_ifstat) / sizeof(uint64_t)); free(ext->in6_ifstat, M_IFADDR); COUNTER_ARRAY_FREE(ext->icmp6_ifstat, sizeof(struct icmp6_ifstat) / sizeof(uint64_t)); free(ext->icmp6_ifstat, M_IFADDR); free(ext, M_IFADDR); } /* * Convert sockaddr_in6 to sockaddr_in. Original sockaddr_in6 must be * v4 mapped addr or v4 compat addr */ void in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6) { bzero(sin, sizeof(*sin)); sin->sin_len = sizeof(struct sockaddr_in); sin->sin_family = AF_INET; sin->sin_port = sin6->sin6_port; sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3]; } /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */ void in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6) { bzero(sin6, sizeof(*sin6)); sin6->sin6_len = sizeof(struct sockaddr_in6); sin6->sin6_family = AF_INET6; sin6->sin6_port = sin->sin_port; sin6->sin6_addr.s6_addr32[0] = 0; sin6->sin6_addr.s6_addr32[1] = 0; sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr; } /* Convert sockaddr_in6 into sockaddr_in. */ void in6_sin6_2_sin_in_sock(struct sockaddr *nam) { struct sockaddr_in *sin_p; struct sockaddr_in6 sin6; /* * Save original sockaddr_in6 addr and convert it * to sockaddr_in. */ sin6 = *(struct sockaddr_in6 *)nam; sin_p = (struct sockaddr_in *)nam; in6_sin6_2_sin(sin_p, &sin6); } /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */ void in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam) { struct sockaddr_in *sin_p; struct sockaddr_in6 *sin6_p; sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK); sin_p = (struct sockaddr_in *)*nam; in6_sin_2_v4mapsin6(sin_p, sin6_p); free(*nam, M_SONAME); *nam = (struct sockaddr *)sin6_p; } Index: projects/vnet/sys/netinet6/in6_ifattach.c =================================================================== --- projects/vnet/sys/netinet6/in6_ifattach.c (revision 294565) +++ projects/vnet/sys/netinet6/in6_ifattach.c (revision 294566) @@ -1,892 +1,918 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: in6_ifattach.c,v 1.118 2001/05/24 07:44:00 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include VNET_DEFINE(unsigned long, in6_maxmtu) = 0; #ifdef IP6_AUTO_LINKLOCAL VNET_DEFINE(int, ip6_auto_linklocal) = IP6_AUTO_LINKLOCAL; #else VNET_DEFINE(int, ip6_auto_linklocal) = 1; /* enabled by default */ #endif VNET_DEFINE(struct callout, in6_tmpaddrtimer_ch); #define V_in6_tmpaddrtimer_ch VNET(in6_tmpaddrtimer_ch) VNET_DECLARE(struct inpcbinfo, ripcbinfo); #define V_ripcbinfo VNET(ripcbinfo) static int get_rand_ifid(struct ifnet *, struct in6_addr *); static int generate_tmp_ifid(u_int8_t *, const u_int8_t *, u_int8_t *); static int get_ifid(struct ifnet *, struct ifnet *, struct in6_addr *); static int in6_ifattach_linklocal(struct ifnet *, struct ifnet *); static int in6_ifattach_loopback(struct ifnet *); static void in6_purgemaddrs(struct ifnet *); #define EUI64_GBIT 0x01 #define EUI64_UBIT 0x02 #define EUI64_TO_IFID(in6) do {(in6)->s6_addr[8] ^= EUI64_UBIT; } while (0) #define EUI64_GROUP(in6) ((in6)->s6_addr[8] & EUI64_GBIT) #define EUI64_INDIVIDUAL(in6) (!EUI64_GROUP(in6)) #define EUI64_LOCAL(in6) ((in6)->s6_addr[8] & EUI64_UBIT) #define EUI64_UNIVERSAL(in6) (!EUI64_LOCAL(in6)) #define IFID_LOCAL(in6) (!EUI64_LOCAL(in6)) #define IFID_UNIVERSAL(in6) (!EUI64_UNIVERSAL(in6)) /* * Generate a last-resort interface identifier, when the machine has no * IEEE802/EUI64 address sources. * The goal here is to get an interface identifier that is * (1) random enough and (2) does not change across reboot. * We currently use MD5(hostname) for it. * * in6 - upper 64bits are preserved */ static int get_rand_ifid(struct ifnet *ifp, struct in6_addr *in6) { MD5_CTX ctxt; struct prison *pr; u_int8_t digest[16]; int hostnamelen; pr = curthread->td_ucred->cr_prison; mtx_lock(&pr->pr_mtx); hostnamelen = strlen(pr->pr_hostname); #if 0 /* we need at least several letters as seed for ifid */ if (hostnamelen < 3) { mtx_unlock(&pr->pr_mtx); return -1; } #endif /* generate 8 bytes of pseudo-random value. */ bzero(&ctxt, sizeof(ctxt)); MD5Init(&ctxt); MD5Update(&ctxt, pr->pr_hostname, hostnamelen); mtx_unlock(&pr->pr_mtx); MD5Final(digest, &ctxt); /* assumes sizeof(digest) > sizeof(ifid) */ bcopy(digest, &in6->s6_addr[8], 8); /* make sure to set "u" bit to local, and "g" bit to individual. */ in6->s6_addr[8] &= ~EUI64_GBIT; /* g bit to "individual" */ in6->s6_addr[8] |= EUI64_UBIT; /* u bit to "local" */ /* convert EUI64 into IPv6 interface identifier */ EUI64_TO_IFID(in6); return 0; } static int generate_tmp_ifid(u_int8_t *seed0, const u_int8_t *seed1, u_int8_t *ret) { MD5_CTX ctxt; u_int8_t seed[16], digest[16], nullbuf[8]; u_int32_t val32; /* If there's no history, start with a random seed. */ bzero(nullbuf, sizeof(nullbuf)); if (bcmp(nullbuf, seed0, sizeof(nullbuf)) == 0) { int i; for (i = 0; i < 2; i++) { val32 = arc4random(); bcopy(&val32, seed + sizeof(val32) * i, sizeof(val32)); } } else bcopy(seed0, seed, 8); /* copy the right-most 64-bits of the given address */ /* XXX assumption on the size of IFID */ bcopy(seed1, &seed[8], 8); if (0) { /* for debugging purposes only */ int i; printf("generate_tmp_ifid: new randomized ID from: "); for (i = 0; i < 16; i++) printf("%02x", seed[i]); printf(" "); } /* generate 16 bytes of pseudo-random value. */ bzero(&ctxt, sizeof(ctxt)); MD5Init(&ctxt); MD5Update(&ctxt, seed, sizeof(seed)); MD5Final(digest, &ctxt); /* * RFC 3041 3.2.1. (3) * Take the left-most 64-bits of the MD5 digest and set bit 6 (the * left-most bit is numbered 0) to zero. */ bcopy(digest, ret, 8); ret[0] &= ~EUI64_UBIT; /* * XXX: we'd like to ensure that the generated value is not zero * for simplicity. If the caclculated digest happens to be zero, * use a random non-zero value as the last resort. */ if (bcmp(nullbuf, ret, sizeof(nullbuf)) == 0) { nd6log((LOG_INFO, "generate_tmp_ifid: computed MD5 value is zero.\n")); val32 = arc4random(); val32 = 1 + (val32 % (0xffffffff - 1)); } /* * RFC 3041 3.2.1. (4) * Take the rightmost 64-bits of the MD5 digest and save them in * stable storage as the history value to be used in the next * iteration of the algorithm. */ bcopy(&digest[8], seed0, 8); if (0) { /* for debugging purposes only */ int i; printf("to: "); for (i = 0; i < 16; i++) printf("%02x", digest[i]); printf("\n"); } return 0; } /* * Get interface identifier for the specified interface. * XXX assumes single sockaddr_dl (AF_LINK address) per an interface * * in6 - upper 64bits are preserved */ int in6_get_hw_ifid(struct ifnet *ifp, struct in6_addr *in6) { struct ifaddr *ifa; struct sockaddr_dl *sdl; u_int8_t *addr; size_t addrlen; static u_int8_t allzero[8] = { 0, 0, 0, 0, 0, 0, 0, 0 }; static u_int8_t allone[8] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_LINK) continue; sdl = (struct sockaddr_dl *)ifa->ifa_addr; if (sdl == NULL) continue; if (sdl->sdl_alen == 0) continue; goto found; } IF_ADDR_RUNLOCK(ifp); return -1; found: IF_ADDR_LOCK_ASSERT(ifp); addr = LLADDR(sdl); addrlen = sdl->sdl_alen; /* get EUI64 */ switch (ifp->if_type) { case IFT_BRIDGE: case IFT_ETHER: case IFT_L2VLAN: case IFT_FDDI: case IFT_ISO88025: case IFT_ATM: case IFT_IEEE1394: case IFT_IEEE80211: /* IEEE802/EUI64 cases - what others? */ /* IEEE1394 uses 16byte length address starting with EUI64 */ if (addrlen > 8) addrlen = 8; /* look at IEEE802/EUI64 only */ if (addrlen != 8 && addrlen != 6) { IF_ADDR_RUNLOCK(ifp); return -1; } /* * check for invalid MAC address - on bsdi, we see it a lot * since wildboar configures all-zero MAC on pccard before * card insertion. */ if (bcmp(addr, allzero, addrlen) == 0) { IF_ADDR_RUNLOCK(ifp); return -1; } if (bcmp(addr, allone, addrlen) == 0) { IF_ADDR_RUNLOCK(ifp); return -1; } /* make EUI64 address */ if (addrlen == 8) bcopy(addr, &in6->s6_addr[8], 8); else if (addrlen == 6) { in6->s6_addr[8] = addr[0]; in6->s6_addr[9] = addr[1]; in6->s6_addr[10] = addr[2]; in6->s6_addr[11] = 0xff; in6->s6_addr[12] = 0xfe; in6->s6_addr[13] = addr[3]; in6->s6_addr[14] = addr[4]; in6->s6_addr[15] = addr[5]; } break; case IFT_ARCNET: if (addrlen != 1) { IF_ADDR_RUNLOCK(ifp); return -1; } if (!addr[0]) { IF_ADDR_RUNLOCK(ifp); return -1; } bzero(&in6->s6_addr[8], 8); in6->s6_addr[15] = addr[0]; /* * due to insufficient bitwidth, we mark it local. */ in6->s6_addr[8] &= ~EUI64_GBIT; /* g bit to "individual" */ in6->s6_addr[8] |= EUI64_UBIT; /* u bit to "local" */ break; case IFT_GIF: case IFT_STF: /* * RFC2893 says: "SHOULD use IPv4 address as ifid source". * however, IPv4 address is not very suitable as unique * identifier source (can be renumbered). * we don't do this. */ IF_ADDR_RUNLOCK(ifp); return -1; default: IF_ADDR_RUNLOCK(ifp); return -1; } /* sanity check: g bit must not indicate "group" */ if (EUI64_GROUP(in6)) { IF_ADDR_RUNLOCK(ifp); return -1; } /* convert EUI64 into IPv6 interface identifier */ EUI64_TO_IFID(in6); /* * sanity check: ifid must not be all zero, avoid conflict with * subnet router anycast */ if ((in6->s6_addr[8] & ~(EUI64_GBIT | EUI64_UBIT)) == 0x00 && bcmp(&in6->s6_addr[9], allzero, 7) == 0) { IF_ADDR_RUNLOCK(ifp); return -1; } IF_ADDR_RUNLOCK(ifp); return 0; } /* * Get interface identifier for the specified interface. If it is not * available on ifp0, borrow interface identifier from other information * sources. * * altifp - secondary EUI64 source */ static int get_ifid(struct ifnet *ifp0, struct ifnet *altifp, struct in6_addr *in6) { struct ifnet *ifp; /* first, try to get it from the interface itself */ if (in6_get_hw_ifid(ifp0, in6) == 0) { nd6log((LOG_DEBUG, "%s: got interface identifier from itself\n", if_name(ifp0))); goto success; } /* try secondary EUI64 source. this basically is for ATM PVC */ if (altifp && in6_get_hw_ifid(altifp, in6) == 0) { nd6log((LOG_DEBUG, "%s: got interface identifier from %s\n", if_name(ifp0), if_name(altifp))); goto success; } /* next, try to get it from some other hardware interface */ IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { if (ifp == ifp0) continue; if (in6_get_hw_ifid(ifp, in6) != 0) continue; /* * to borrow ifid from other interface, ifid needs to be * globally unique */ if (IFID_UNIVERSAL(in6)) { nd6log((LOG_DEBUG, "%s: borrow interface identifier from %s\n", if_name(ifp0), if_name(ifp))); IFNET_RUNLOCK_NOSLEEP(); goto success; } } IFNET_RUNLOCK_NOSLEEP(); /* last resort: get from random number source */ if (get_rand_ifid(ifp, in6) == 0) { nd6log((LOG_DEBUG, "%s: interface identifier generated by random number\n", if_name(ifp0))); goto success; } printf("%s: failed to get interface identifier\n", if_name(ifp0)); return -1; success: nd6log((LOG_INFO, "%s: ifid: %02x:%02x:%02x:%02x:%02x:%02x:%02x:%02x\n", if_name(ifp0), in6->s6_addr[8], in6->s6_addr[9], in6->s6_addr[10], in6->s6_addr[11], in6->s6_addr[12], in6->s6_addr[13], in6->s6_addr[14], in6->s6_addr[15])); return 0; } /* * altifp - secondary EUI64 source */ static int in6_ifattach_linklocal(struct ifnet *ifp, struct ifnet *altifp) { struct in6_ifaddr *ia; struct in6_aliasreq ifra; struct nd_prefixctl pr0; int error; /* * configure link-local address. */ in6_prepare_ifra(&ifra, NULL, &in6mask64); ifra.ifra_addr.sin6_addr.s6_addr32[0] = htonl(0xfe800000); ifra.ifra_addr.sin6_addr.s6_addr32[1] = 0; if ((ifp->if_flags & IFF_LOOPBACK) != 0) { ifra.ifra_addr.sin6_addr.s6_addr32[2] = 0; ifra.ifra_addr.sin6_addr.s6_addr32[3] = htonl(1); } else { if (get_ifid(ifp, altifp, &ifra.ifra_addr.sin6_addr) != 0) { nd6log((LOG_ERR, "%s: no ifid available\n", if_name(ifp))); return (-1); } } if (in6_setscope(&ifra.ifra_addr.sin6_addr, ifp, NULL)) return (-1); /* link-local addresses should NEVER expire. */ ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME; ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME; /* * Now call in6_update_ifa() to do a bunch of procedures to configure * a link-local address. We can set the 3rd argument to NULL, because * we know there's no other link-local address on the interface * and therefore we are adding one (instead of updating one). */ if ((error = in6_update_ifa(ifp, &ifra, NULL, IN6_IFAUPDATE_DADDELAY)) != 0) { /* * XXX: When the interface does not support IPv6, this call * would fail in the SIOCSIFADDR ioctl. I believe the * notification is rather confusing in this case, so just * suppress it. (jinmei@kame.net 20010130) */ if (error != EAFNOSUPPORT) nd6log((LOG_NOTICE, "in6_ifattach_linklocal: failed to " "configure a link-local address on %s " "(errno=%d)\n", if_name(ifp), error)); return (-1); } ia = in6ifa_ifpforlinklocal(ifp, 0); /* ia must not be NULL */ KASSERT(ia != NULL, ("%s: ia == NULL, ifp=%p", __func__, ifp)); ifa_free(&ia->ia_ifa); /* * Make the link-local prefix (fe80::%link/64) as on-link. * Since we'd like to manage prefixes separately from addresses, * we make an ND6 prefix structure for the link-local prefix, * and add it to the prefix list as a never-expire prefix. * XXX: this change might affect some existing code base... */ bzero(&pr0, sizeof(pr0)); pr0.ndpr_ifp = ifp; /* this should be 64 at this moment. */ pr0.ndpr_plen = in6_mask2len(&ifra.ifra_prefixmask.sin6_addr, NULL); pr0.ndpr_prefix = ifra.ifra_addr; /* apply the mask for safety. (nd6_prelist_add will apply it again) */ IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr, &in6mask64); /* * Initialize parameters. The link-local prefix must always be * on-link, and its lifetimes never expire. */ pr0.ndpr_raf_onlink = 1; pr0.ndpr_raf_auto = 1; /* probably meaningless */ pr0.ndpr_vltime = ND6_INFINITE_LIFETIME; pr0.ndpr_pltime = ND6_INFINITE_LIFETIME; /* * Since there is no other link-local addresses, nd6_prefix_lookup() * probably returns NULL. However, we cannot always expect the result. * For example, if we first remove the (only) existing link-local * address, and then reconfigure another one, the prefix is still * valid with referring to the old link-local address. */ if (nd6_prefix_lookup(&pr0) == NULL) { if ((error = nd6_prelist_add(&pr0, NULL, NULL)) != 0) return (error); } return 0; } /* * ifp - must be IFT_LOOP */ static int in6_ifattach_loopback(struct ifnet *ifp) { struct in6_aliasreq ifra; int error; in6_prepare_ifra(&ifra, &in6addr_loopback, &in6mask128); /* * Always initialize ia_dstaddr (= broadcast address) to loopback * address. Follows IPv4 practice - see in_ifinit(). */ ifra.ifra_dstaddr.sin6_len = sizeof(struct sockaddr_in6); ifra.ifra_dstaddr.sin6_family = AF_INET6; ifra.ifra_dstaddr.sin6_addr = in6addr_loopback; /* the loopback address should NEVER expire. */ ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME; ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME; /* * We are sure that this is a newly assigned address, so we can set * NULL to the 3rd arg. */ if ((error = in6_update_ifa(ifp, &ifra, NULL, 0)) != 0) { nd6log((LOG_ERR, "in6_ifattach_loopback: failed to configure " "the loopback address on %s (errno=%d)\n", if_name(ifp), error)); return (-1); } return 0; } /* * compute NI group address, based on the current hostname setting. * see RFC 4620. * * when ifp == NULL, the caller is responsible for filling scopeid. * * If oldmcprefix == 1, FF02:0:0:0:0:2::/96 is used for NI group address * while it is FF02:0:0:0:0:2:FF00::/104 in RFC 4620. */ static int in6_nigroup0(struct ifnet *ifp, const char *name, int namelen, struct in6_addr *in6, int oldmcprefix) { struct prison *pr; const char *p; u_char *q; MD5_CTX ctxt; u_int8_t digest[16]; char l; char n[64]; /* a single label must not exceed 63 chars */ /* * If no name is given and namelen is -1, * we try to do the hostname lookup ourselves. */ if (!name && namelen == -1) { pr = curthread->td_ucred->cr_prison; mtx_lock(&pr->pr_mtx); name = pr->pr_hostname; namelen = strlen(name); } else pr = NULL; if (!name || !namelen) { if (pr != NULL) mtx_unlock(&pr->pr_mtx); return -1; } p = name; while (p && *p && *p != '.' && p - name < namelen) p++; if (p == name || p - name > sizeof(n) - 1) { if (pr != NULL) mtx_unlock(&pr->pr_mtx); return -1; /* label too long */ } l = p - name; strncpy(n, name, l); if (pr != NULL) mtx_unlock(&pr->pr_mtx); n[(int)l] = '\0'; for (q = n; *q; q++) { if ('A' <= *q && *q <= 'Z') *q = *q - 'A' + 'a'; } /* generate 16 bytes of pseudo-random value. */ bzero(&ctxt, sizeof(ctxt)); MD5Init(&ctxt); MD5Update(&ctxt, &l, sizeof(l)); MD5Update(&ctxt, n, l); MD5Final(digest, &ctxt); bzero(in6, sizeof(*in6)); in6->s6_addr16[0] = IPV6_ADDR_INT16_MLL; in6->s6_addr8[11] = 2; if (oldmcprefix == 0) { in6->s6_addr8[12] = 0xff; /* Copy the first 24 bits of 128-bit hash into the address. */ bcopy(digest, &in6->s6_addr8[13], 3); } else { /* Copy the first 32 bits of 128-bit hash into the address. */ bcopy(digest, &in6->s6_addr32[3], sizeof(in6->s6_addr32[3])); } if (in6_setscope(in6, ifp, NULL)) return (-1); /* XXX: should not fail */ return 0; } int in6_nigroup(struct ifnet *ifp, const char *name, int namelen, struct in6_addr *in6) { return (in6_nigroup0(ifp, name, namelen, in6, 0)); } int in6_nigroup_oldmcprefix(struct ifnet *ifp, const char *name, int namelen, struct in6_addr *in6) { return (in6_nigroup0(ifp, name, namelen, in6, 1)); } /* * XXX multiple loopback interface needs more care. for instance, * nodelocal address needs to be configured onto only one of them. * XXX multiple link-local address case * * altifp - secondary EUI64 source */ void in6_ifattach(struct ifnet *ifp, struct ifnet *altifp) { struct in6_ifaddr *ia; struct in6_addr in6; if (ifp->if_afdata[AF_INET6] == NULL) return; /* * quirks based on interface type */ switch (ifp->if_type) { case IFT_STF: /* * 6to4 interface is a very special kind of beast. * no multicast, no linklocal. RFC2529 specifies how to make * linklocals for 6to4 interface, but there's no use and * it is rather harmful to have one. */ ND_IFINFO(ifp)->flags &= ~ND6_IFF_AUTO_LINKLOCAL; break; default: break; } /* * usually, we require multicast capability to the interface */ if ((ifp->if_flags & IFF_MULTICAST) == 0) { nd6log((LOG_INFO, "in6_ifattach: " "%s is not multicast capable, IPv6 not enabled\n", if_name(ifp))); return; } /* * assign loopback address for loopback interface. * XXX multiple loopback interface case. */ if ((ifp->if_flags & IFF_LOOPBACK) != 0) { struct ifaddr *ifa; in6 = in6addr_loopback; ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &in6); if (ifa == NULL) { if (in6_ifattach_loopback(ifp) != 0) return; } else ifa_free(ifa); } /* * assign a link-local address, if there's none. */ if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL) { int error; ia = in6ifa_ifpforlinklocal(ifp, 0); if (ia == NULL) { error = in6_ifattach_linklocal(ifp, altifp); #if 0 if (error) log(LOG_NOTICE, "in6_ifattach_linklocal: " "failed to add a link-local addr to %s\n", if_name(ifp)); #endif } else ifa_free(&ia->ia_ifa); } /* update dynamically. */ if (V_in6_maxmtu < ifp->if_mtu) V_in6_maxmtu = ifp->if_mtu; } /* * NOTE: in6_ifdetach() does not support loopback if at this moment. * We don't need this function in bsdi, because interfaces are never removed * from the ifnet list in bsdi. */ void in6_ifdetach(struct ifnet *ifp) { struct ifaddr *ifa, *next; if (ifp->if_afdata[AF_INET6] == NULL) return; /* remove neighbor management table */ nd6_purge(ifp); /* * nuke any of IPv6 addresses we have * XXX: all addresses should be already removed */ TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; in6_purgeaddr(ifa); } in6_pcbpurgeif0(&V_udbinfo, ifp); in6_pcbpurgeif0(&V_ulitecbinfo, ifp); in6_pcbpurgeif0(&V_ripcbinfo, ifp); /* leave from all multicast groups joined */ in6_purgemaddrs(ifp); /* * remove neighbor management table. we call it twice just to make * sure we nuke everything. maybe we need just one call. * XXX: since the first call did not release addresses, some prefixes * might remain. We should call nd6_purge() again to release the * prefixes after removing all addresses above. * (Or can we just delay calling nd6_purge until at this point?) */ nd6_purge(ifp); } int in6_get_tmpifid(struct ifnet *ifp, u_int8_t *retbuf, const u_int8_t *baseid, int generate) { u_int8_t nullbuf[8]; struct nd_ifinfo *ndi = ND_IFINFO(ifp); bzero(nullbuf, sizeof(nullbuf)); if (bcmp(ndi->randomid, nullbuf, sizeof(nullbuf)) == 0) { /* we've never created a random ID. Create a new one. */ generate = 1; } if (generate) { bcopy(baseid, ndi->randomseed1, sizeof(ndi->randomseed1)); /* generate_tmp_ifid will update seedn and buf */ (void)generate_tmp_ifid(ndi->randomseed0, ndi->randomseed1, ndi->randomid); } bcopy(ndi->randomid, retbuf, 8); return (0); } void in6_tmpaddrtimer(void *arg) { CURVNET_SET((struct vnet *) arg); struct nd_ifinfo *ndi; u_int8_t nullbuf[8]; struct ifnet *ifp; callout_reset(&V_in6_tmpaddrtimer_ch, (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor - V_ip6_temp_regen_advance) * hz, in6_tmpaddrtimer, curvnet); bzero(nullbuf, sizeof(nullbuf)); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { if (ifp->if_afdata[AF_INET6] == NULL) continue; ndi = ND_IFINFO(ifp); if (bcmp(ndi->randomid, nullbuf, sizeof(nullbuf)) != 0) { /* * We've been generating a random ID on this interface. * Create a new one. */ (void)generate_tmp_ifid(ndi->randomseed0, ndi->randomseed1, ndi->randomid); } } CURVNET_RESTORE(); } static void in6_purgemaddrs(struct ifnet *ifp) { LIST_HEAD(,in6_multi) purgeinms; struct in6_multi *inm, *tinm; struct ifmultiaddr *ifma; LIST_INIT(&purgeinms); IN6_MULTI_LOCK(); /* * Extract list of in6_multi associated with the detaching ifp * which the PF_INET6 layer is about to release. * We need to do this as IF_ADDR_LOCK() may be re-acquired * by code further down. */ IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET6 || ifma->ifma_protospec == NULL) continue; inm = (struct in6_multi *)ifma->ifma_protospec; LIST_INSERT_HEAD(&purgeinms, inm, in6m_entry); } IF_ADDR_RUNLOCK(ifp); LIST_FOREACH_SAFE(inm, &purgeinms, in6m_entry, tinm) { LIST_REMOVE(inm, in6m_entry); in6m_release_locked(inm); } mld_ifdetach(ifp); IN6_MULTI_UNLOCK(); } + +void +in6_ifattach_destroy(void) +{ + + callout_drain(&V_in6_tmpaddrtimer_ch); +} + +static void +in6_ifattach_init(void *dummy) +{ + + /* Timer for regeneranation of temporary addresses randomize ID. */ + callout_init(&V_in6_tmpaddrtimer_ch, 0); + callout_reset(&V_in6_tmpaddrtimer_ch, + (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor - + V_ip6_temp_regen_advance) * hz, + in6_tmpaddrtimer, curvnet); +} + +/* + * Cheat. + * This must be after route_init(), which is now SI_ORDER_THIRD. + */ +SYSINIT(in6_ifattach_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, + in6_ifattach_init, NULL); Index: projects/vnet/sys/netinet6/in6_ifattach.h =================================================================== --- projects/vnet/sys/netinet6/in6_ifattach.h (revision 294565) +++ projects/vnet/sys/netinet6/in6_ifattach.h (revision 294566) @@ -1,46 +1,47 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: in6_ifattach.h,v 1.14 2001/02/08 12:48:39 jinmei Exp $ * $FreeBSD$ */ #ifndef _NETINET6_IN6_IFATTACH_H_ #define _NETINET6_IN6_IFATTACH_H_ #ifdef _KERNEL void in6_ifattach(struct ifnet *, struct ifnet *); +void in6_ifattach_destroy(void); void in6_ifdetach(struct ifnet *); int in6_get_tmpifid(struct ifnet *, u_int8_t *, const u_int8_t *, int); void in6_tmpaddrtimer(void *); int in6_get_hw_ifid(struct ifnet *, struct in6_addr *); int in6_nigroup(struct ifnet *, const char *, int, struct in6_addr *); int in6_nigroup_oldmcprefix(struct ifnet *, const char *, int, struct in6_addr *); #endif /* _KERNEL */ #endif /* _NETINET6_IN6_IFATTACH_H_ */ Index: projects/vnet/sys/netinet6/in6_proto.c =================================================================== --- projects/vnet/sys/netinet6/in6_proto.c (revision 294565) +++ projects/vnet/sys/netinet6/in6_proto.c (revision 294566) @@ -1,627 +1,624 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: in6_proto.c,v 1.91 2001/05/27 13:28:35 itojun Exp $ */ /*- * Copyright (c) 1982, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)in_proto.c 8.1 (Berkeley) 6/10/93 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_ipstealth.h" #include "opt_sctp.h" #include "opt_mpath.h" #include "opt_route.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef RADIX_MPATH #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef SCTP #include #include #include #include #include #endif /* SCTP */ #ifdef IPSEC #include #include #endif /* IPSEC */ #include /* * TCP/IP protocol family: IP6, ICMP6, UDP, TCP. */ FEATURE(inet6, "Internet Protocol version 6"); extern struct domain inet6domain; static struct pr_usrreqs nousrreqs; #define PR_LISTEN 0 #define PR_ABRTACPTDIS 0 /* Spacer for loadable protocols. */ #define IP6PROTOSPACER \ { \ .pr_domain = &inet6domain, \ .pr_protocol = PROTO_SPACER, \ .pr_usrreqs = &nousrreqs \ } struct protosw inet6sw[] = { { .pr_type = 0, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_IPV6, .pr_init = ip6_init, -#ifdef VIMAGE - .pr_destroy = ip6_destroy, -#endif .pr_slowtimo = frag6_slowtimo, .pr_drain = frag6_drain, .pr_usrreqs = &nousrreqs, }, { .pr_type = SOCK_DGRAM, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_UDP, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = udp6_input, .pr_ctlinput = udp6_ctlinput, .pr_ctloutput = ip6_ctloutput, #ifndef INET /* Do not call initialization twice. */ .pr_init = udp_init, #endif .pr_usrreqs = &udp6_usrreqs, }, { .pr_type = SOCK_STREAM, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_TCP, .pr_flags = PR_CONNREQUIRED|PR_WANTRCVD|PR_LISTEN, .pr_input = tcp6_input, .pr_ctlinput = tcp6_ctlinput, .pr_ctloutput = tcp_ctloutput, #ifndef INET /* don't call initialization and timeout routines twice */ .pr_init = tcp_init, .pr_slowtimo = tcp_slowtimo, #endif .pr_drain = tcp_drain, .pr_usrreqs = &tcp6_usrreqs, }, #ifdef SCTP { .pr_type = SOCK_SEQPACKET, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_SCTP, .pr_flags = PR_WANTRCVD, .pr_input = sctp6_input, .pr_ctlinput = sctp6_ctlinput, .pr_ctloutput = sctp_ctloutput, .pr_drain = sctp_drain, #ifndef INET /* Do not call initialization twice. */ .pr_init = sctp_init, #endif .pr_usrreqs = &sctp6_usrreqs }, { .pr_type = SOCK_STREAM, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_SCTP, .pr_flags = PR_WANTRCVD, .pr_input = sctp6_input, .pr_ctlinput = sctp6_ctlinput, .pr_ctloutput = sctp_ctloutput, .pr_drain = sctp_drain, .pr_usrreqs = &sctp6_usrreqs }, #endif /* SCTP */ { .pr_type = SOCK_DGRAM, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_UDPLITE, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = udp6_input, .pr_ctlinput = udplite6_ctlinput, .pr_ctloutput = udp_ctloutput, #ifndef INET /* Do not call initialization twice. */ .pr_init = udplite_init, #endif .pr_usrreqs = &udp6_usrreqs, }, { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_RAW, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = rip6_input, .pr_output = rip6_output, .pr_ctlinput = rip6_ctlinput, .pr_ctloutput = rip6_ctloutput, #ifndef INET /* Do not call initialization twice. */ .pr_init = rip_init, #endif .pr_usrreqs = &rip6_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_ICMPV6, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = icmp6_input, .pr_output = rip6_output, .pr_ctlinput = rip6_ctlinput, .pr_ctloutput = rip6_ctloutput, .pr_fasttimo = icmp6_fasttimo, .pr_slowtimo = icmp6_slowtimo, .pr_usrreqs = &rip6_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_DSTOPTS, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = dest6_input, .pr_usrreqs = &nousrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_ROUTING, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = route6_input, .pr_usrreqs = &nousrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_FRAGMENT, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = frag6_input, .pr_usrreqs = &nousrreqs }, #ifdef IPSEC { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_AH, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = ipsec6_common_input, .pr_usrreqs = &nousrreqs, }, { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_ESP, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = ipsec6_common_input, .pr_ctlinput = esp6_ctlinput, .pr_usrreqs = &nousrreqs, }, { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_IPCOMP, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = ipsec6_common_input, .pr_usrreqs = &nousrreqs, }, #endif /* IPSEC */ #ifdef INET { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_IPV4, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = encap6_input, .pr_output = rip6_output, .pr_ctloutput = rip6_ctloutput, .pr_init = encap_init, .pr_usrreqs = &rip6_usrreqs }, #endif /* INET */ { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_IPV6, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = encap6_input, .pr_output = rip6_output, .pr_ctloutput = rip6_ctloutput, .pr_init = encap_init, .pr_usrreqs = &rip6_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_GRE, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = encap6_input, .pr_output = rip6_output, .pr_ctloutput = rip6_ctloutput, .pr_init = encap_init, .pr_usrreqs = &rip6_usrreqs }, { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_PIM, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = encap6_input, .pr_output = rip6_output, .pr_ctloutput = rip6_ctloutput, .pr_usrreqs = &rip6_usrreqs }, /* Spacer n-times for loadable protocols. */ IP6PROTOSPACER, IP6PROTOSPACER, IP6PROTOSPACER, IP6PROTOSPACER, IP6PROTOSPACER, IP6PROTOSPACER, IP6PROTOSPACER, IP6PROTOSPACER, /* raw wildcard */ { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = rip6_input, .pr_output = rip6_output, .pr_ctloutput = rip6_ctloutput, .pr_usrreqs = &rip6_usrreqs }, }; extern int in6_inithead(void **, int); #ifdef VIMAGE extern int in6_detachhead(void **, int); #endif struct domain inet6domain = { .dom_family = AF_INET6, .dom_name = "internet6", .dom_protosw = (struct protosw *)inet6sw, .dom_protoswNPROTOSW = (struct protosw *) &inet6sw[sizeof(inet6sw)/sizeof(inet6sw[0])], #ifdef RADIX_MPATH .dom_rtattach = rn6_mpath_inithead, #else .dom_rtattach = in6_inithead, #endif #ifdef VIMAGE .dom_rtdetach = in6_detachhead, #endif .dom_ifattach = in6_domifattach, .dom_ifdetach = in6_domifdetach, .dom_ifmtu = in6_domifmtu }; VNET_DOMAIN_SET(inet6); /* * Internet configuration info */ #ifndef IPV6FORWARDING #ifdef GATEWAY6 #define IPV6FORWARDING 1 /* forward IP6 packets not for us */ #else #define IPV6FORWARDING 0 /* don't forward IP6 packets not for us */ #endif /* GATEWAY6 */ #endif /* !IPV6FORWARDING */ #ifndef IPV6_SENDREDIRECTS #define IPV6_SENDREDIRECTS 1 #endif VNET_DEFINE(int, ip6_forwarding) = IPV6FORWARDING; /* act as router? */ VNET_DEFINE(int, ip6_sendredirects) = IPV6_SENDREDIRECTS; VNET_DEFINE(int, ip6_defhlim) = IPV6_DEFHLIM; VNET_DEFINE(int, ip6_defmcasthlim) = IPV6_DEFAULT_MULTICAST_HOPS; VNET_DEFINE(int, ip6_accept_rtadv) = 0; VNET_DEFINE(int, ip6_no_radr) = 0; VNET_DEFINE(int, ip6_norbit_raif) = 0; VNET_DEFINE(int, ip6_rfc6204w3) = 0; VNET_DEFINE(int, ip6_maxfragpackets); /* initialized in frag6.c:frag6_init() */ VNET_DEFINE(int, ip6_maxfrags); /* initialized in frag6.c:frag6_init() */ VNET_DEFINE(int, ip6_log_interval) = 5; VNET_DEFINE(int, ip6_hdrnestlimit) = 15;/* How many header options will we * process? */ VNET_DEFINE(int, ip6_dad_count) = 1; /* DupAddrDetectionTransmits */ VNET_DEFINE(int, ip6_auto_flowlabel) = 1; VNET_DEFINE(int, ip6_use_deprecated) = 1;/* allow deprecated addr * (RFC2462 5.5.4) */ VNET_DEFINE(int, ip6_rr_prune) = 5; /* router renumbering prefix * walk list every 5 sec. */ VNET_DEFINE(int, ip6_mcast_pmtu) = 0; /* enable pMTU discovery for multicast? */ VNET_DEFINE(int, ip6_v6only) = 1; VNET_DEFINE(time_t, ip6_log_time) = (time_t)0L; #ifdef IPSTEALTH VNET_DEFINE(int, ip6stealth) = 0; #endif VNET_DEFINE(int, nd6_onlink_ns_rfc4861) = 0;/* allow 'on-link' nd6 NS * (RFC 4861) */ /* icmp6 */ /* * BSDI4 defines these variables in in_proto.c... * XXX: what if we don't define INET? Should we define pmtu6_expire * or so? (jinmei@kame.net 19990310) */ VNET_DEFINE(int, pmtu_expire) = 60*10; VNET_DEFINE(int, pmtu_probe) = 60*2; /* ICMPV6 parameters */ VNET_DEFINE(int, icmp6_rediraccept) = 1;/* accept and process redirects */ VNET_DEFINE(int, icmp6_redirtimeout) = 10 * 60; /* 10 minutes */ VNET_DEFINE(int, icmp6errppslim) = 100; /* 100pps */ /* control how to respond to NI queries */ VNET_DEFINE(int, icmp6_nodeinfo) = (ICMP6_NODEINFO_FQDNOK|ICMP6_NODEINFO_NODEADDROK); VNET_DEFINE(int, icmp6_nodeinfo_oldmcprefix) = 1; /* * sysctl related items. */ SYSCTL_NODE(_net, PF_INET6, inet6, CTLFLAG_RW, 0, "Internet6 Family"); /* net.inet6 */ SYSCTL_NODE(_net_inet6, IPPROTO_IPV6, ip6, CTLFLAG_RW, 0, "IP6"); SYSCTL_NODE(_net_inet6, IPPROTO_ICMPV6, icmp6, CTLFLAG_RW, 0, "ICMP6"); SYSCTL_NODE(_net_inet6, IPPROTO_UDP, udp6, CTLFLAG_RW, 0, "UDP6"); SYSCTL_NODE(_net_inet6, IPPROTO_TCP, tcp6, CTLFLAG_RW, 0, "TCP6"); #ifdef SCTP SYSCTL_NODE(_net_inet6, IPPROTO_SCTP, sctp6, CTLFLAG_RW, 0, "SCTP6"); #endif #ifdef IPSEC SYSCTL_NODE(_net_inet6, IPPROTO_ESP, ipsec6, CTLFLAG_RW, 0, "IPSEC6"); #endif /* IPSEC */ /* net.inet6.ip6 */ static int sysctl_ip6_temppltime(SYSCTL_HANDLER_ARGS) { int error = 0; int old; error = SYSCTL_OUT(req, arg1, sizeof(int)); if (error || !req->newptr) return (error); old = V_ip6_temp_preferred_lifetime; error = SYSCTL_IN(req, arg1, sizeof(int)); if (V_ip6_temp_preferred_lifetime < V_ip6_desync_factor + V_ip6_temp_regen_advance) { V_ip6_temp_preferred_lifetime = old; return (EINVAL); } return (error); } static int sysctl_ip6_tempvltime(SYSCTL_HANDLER_ARGS) { int error = 0; int old; error = SYSCTL_OUT(req, arg1, sizeof(int)); if (error || !req->newptr) return (error); old = V_ip6_temp_valid_lifetime; error = SYSCTL_IN(req, arg1, sizeof(int)); if (V_ip6_temp_valid_lifetime < V_ip6_temp_preferred_lifetime) { V_ip6_temp_preferred_lifetime = old; return (EINVAL); } return (error); } SYSCTL_INT(_net_inet6_ip6, IPV6CTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_forwarding), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_sendredirects), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_DEFHLIM, hlim, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_defhlim), 0, ""); SYSCTL_VNET_PCPUSTAT(_net_inet6_ip6, IPV6CTL_STATS, stats, struct ip6stat, ip6stat, "IP6 statistics (struct ip6stat, netinet6/ip6_var.h)"); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_MAXFRAGPACKETS, maxfragpackets, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_maxfragpackets), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_ACCEPT_RTADV, accept_rtadv, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_accept_rtadv), 0, "Default value of per-interface flag for accepting ICMPv6 Router" "Advertisement messages"); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_NO_RADR, no_radr, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_no_radr), 0, "Default value of per-interface flag to control whether routers " "sending ICMPv6 RA messages on that interface are added into the " "default router list."); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_NORBIT_RAIF, norbit_raif, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_norbit_raif), 0, "Always set 0 to R flag in ICMPv6 NA messages when accepting RA" " on the interface."); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RFC6204W3, rfc6204w3, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_rfc6204w3), 0, "Accept the default router list from ICMPv6 RA messages even " "when packet forwarding enabled."); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_LOG_INTERVAL, log_interval, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_log_interval), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_HDRNESTLIMIT, hdrnestlimit, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_hdrnestlimit), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_DAD_COUNT, dad_count, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_dad_count), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_AUTO_FLOWLABEL, auto_flowlabel, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_auto_flowlabel), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_DEFMCASTHLIM, defmcasthlim, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_defmcasthlim), 0, ""); SYSCTL_STRING(_net_inet6_ip6, IPV6CTL_KAME_VERSION, kame_version, CTLFLAG_RD, __KAME_VERSION, 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_USE_DEPRECATED, use_deprecated, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_use_deprecated), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RR_PRUNE, rr_prune, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_rr_prune), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_USETEMPADDR, use_tempaddr, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_use_tempaddr), 0, ""); SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_TEMPPLTIME, temppltime, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ip6_temp_preferred_lifetime), 0, sysctl_ip6_temppltime, "I", ""); SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_TEMPVLTIME, tempvltime, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, &VNET_NAME(ip6_temp_valid_lifetime), 0, sysctl_ip6_tempvltime, "I", ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_V6ONLY, v6only, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_v6only), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_AUTO_LINKLOCAL, auto_linklocal, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_auto_linklocal), 0, "Default value of per-interface flag for automatically adding an IPv6" " link-local address to interfaces when attached"); SYSCTL_VNET_PCPUSTAT(_net_inet6_ip6, IPV6CTL_RIP6STATS, rip6stats, struct rip6stat, rip6stat, "Raw IP6 statistics (struct rip6stat, netinet6/raw_ip6.h)"); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_PREFER_TEMPADDR, prefer_tempaddr, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_prefer_tempaddr), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_USE_DEFAULTZONE, use_defaultzone, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_use_defzone), 0,""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_MAXFRAGS, maxfrags, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_maxfrags), 0, ""); SYSCTL_INT(_net_inet6_ip6, IPV6CTL_MCAST_PMTU, mcast_pmtu, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_mcast_pmtu), 0, ""); #ifdef IPSTEALTH SYSCTL_INT(_net_inet6_ip6, IPV6CTL_STEALTH, stealth, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6stealth), 0, ""); #endif /* net.inet6.icmp6 */ SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_REDIRACCEPT, rediraccept, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_rediraccept), 0, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_REDIRTIMEOUT, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_redirtimeout), 0, ""); SYSCTL_VNET_PCPUSTAT(_net_inet6_icmp6, ICMPV6CTL_STATS, stats, struct icmp6stat, icmp6stat, "ICMPv6 statistics (struct icmp6stat, netinet/icmp6.h)"); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_PRUNE, nd6_prune, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_prune), 0, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_DELAY, nd6_delay, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_delay), 0, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_UMAXTRIES, nd6_umaxtries, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_umaxtries), 0, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MMAXTRIES, nd6_mmaxtries, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_mmaxtries), 0, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_USELOOPBACK, nd6_useloopback, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_useloopback), 0, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_NODEINFO, nodeinfo, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_nodeinfo), 0, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_NODEINFO_OLDMCPREFIX, nodeinfo_oldmcprefix, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_nodeinfo_oldmcprefix), 0, "Join old IPv6 NI group address in draft-ietf-ipngwg-icmp-name-lookup" " for compatibility with KAME implememtation."); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ERRPPSLIMIT, errppslimit, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6errppslim), 0, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXNUDHINT, nd6_maxnudhint, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxnudhint), 0, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_DEBUG, nd6_debug, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_debug), 0, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_ONLINKNSRFC4861, nd6_onlink_ns_rfc4861, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_onlink_ns_rfc4861), 0, "Accept 'on-link' nd6 NS in compliance with RFC 4861."); Index: projects/vnet/sys/netinet6/ip6_input.c =================================================================== --- projects/vnet/sys/netinet6/ip6_input.c (revision 294565) +++ projects/vnet/sys/netinet6/ip6_input.c (revision 294566) @@ -1,1736 +1,1703 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $ */ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipfw.h" #include "opt_ipsec.h" #include "opt_route.h" #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET #include #include #endif /* INET */ #include #include #include #include #include #include #include #include #include #ifdef IPSEC #include #include #include #endif /* IPSEC */ #include extern struct domain inet6domain; u_char ip6_protox[IPPROTO_MAX]; VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); VNET_DEFINE(u_long, in6_ifaddrhmask); static struct netisr_handler ip6_nh = { .nh_name = "ip6", .nh_handler = ip6_input, .nh_proto = NETISR_IPV6, #ifdef RSS .nh_m2cpuid = rss_soft_m2cpuid_v6, .nh_policy = NETISR_POLICY_CPU, .nh_dispatch = NETISR_DISPATCH_HYBRID, #else .nh_policy = NETISR_POLICY_FLOW, #endif }; #ifdef RSS static struct netisr_handler ip6_direct_nh = { .nh_name = "ip6_direct", .nh_handler = ip6_direct_input, .nh_proto = NETISR_IPV6_DIRECT, .nh_m2cpuid = rss_soft_m2cpuid_v6, .nh_policy = NETISR_POLICY_CPU, .nh_dispatch = NETISR_DISPATCH_HYBRID, }; #endif -VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch); -#define V_in6_tmpaddrtimer_ch VNET(in6_tmpaddrtimer_ch) - VNET_DEFINE(struct pfil_head, inet6_pfil_hook); VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); VNET_PCPUSTAT_SYSINIT(ip6stat); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(ip6stat); #endif /* VIMAGE */ struct rmlock in6_ifaddr_lock; RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); -static void ip6_init2(void *); static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); #ifdef PULLDOWN_TEST static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int); #endif /* * IP6 initialization: fill in IP6 protocol switch table. * All protocols not implemented in kernel go to raw IP6 protocol handler. */ void ip6_init(void) { struct protosw *pr; int i; TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", &V_ip6_auto_linklocal); TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); TAILQ_INIT(&V_in6_ifaddrhead); V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, &V_in6_ifaddrhmask); /* Initialize packet filter hooks. */ V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF; V_inet6_pfil_hook.ph_af = AF_INET6; if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0) printf("%s: WARNING: unable to register pfil hook, " "error %d\n", __func__, i); if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register input helper hook\n", __func__); if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) printf("%s: WARNING: unable to register output helper hook\n", __func__); scope6_init(); addrsel_policy_init(); nd6_init(); frag6_init(); V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; /* Skip global initialization stuff for non-default instances. */ if (!IS_DEFAULT_VNET(curvnet)) return; pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) panic("ip6_init"); /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ for (i = 0; i < IPPROTO_MAX; i++) ip6_protox[i] = pr - inet6sw; /* * Cycle through IP protocols and put them into the appropriate place * in ip6_protox[]. */ for (pr = inet6domain.dom_protosw; pr < inet6domain.dom_protoswNPROTOSW; pr++) if (pr->pr_domain->dom_family == PF_INET6 && pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { /* Be careful to only index valid IP protocols. */ if (pr->pr_protocol < IPPROTO_MAX) ip6_protox[pr->pr_protocol] = pr - inet6sw; } netisr_register(&ip6_nh); #ifdef RSS netisr_register(&ip6_direct_nh); #endif } /* * The protocol to be inserted into ip6_protox[] must be already registered * in inet6sw[], either statically or through pf_proto_register(). */ int ip6proto_register(short ip6proto) { struct protosw *pr; /* Sanity checks. */ if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) return (EPROTONOSUPPORT); /* * The protocol slot must not be occupied by another protocol * already. An index pointing to IPPROTO_RAW is unused. */ pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) return (EPFNOSUPPORT); if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ return (EEXIST); /* * Find the protocol position in inet6sw[] and set the index. */ for (pr = inet6domain.dom_protosw; pr < inet6domain.dom_protoswNPROTOSW; pr++) { if (pr->pr_domain->dom_family == PF_INET6 && pr->pr_protocol && pr->pr_protocol == ip6proto) { ip6_protox[pr->pr_protocol] = pr - inet6sw; return (0); } } return (EPROTONOSUPPORT); } int ip6proto_unregister(short ip6proto) { struct protosw *pr; /* Sanity checks. */ if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) return (EPROTONOSUPPORT); /* Check if the protocol was indeed registered. */ pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); if (pr == NULL) return (EPFNOSUPPORT); if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ return (ENOENT); /* Reset the protocol slot to IPPROTO_RAW. */ ip6_protox[ip6proto] = pr - inet6sw; return (0); } #ifdef VIMAGE void ip6_destroy() { int error; if ((error = pfil_head_unregister(&V_inet6_pfil_hook)) != 0) printf("%s: WARNING: unable to unregister pfil hook, " "error %d\n", __func__, error); error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); if (error != 0) { printf("%s: WARNING: unable to deregister input helper hook " "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " "error %d returned\n", __func__, error); } error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); if (error != 0) { printf("%s: WARNING: unable to deregister output helper hook " "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " "error %d returned\n", __func__, error); } hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); nd6_destroy(); - callout_drain(&V_in6_tmpaddrtimer_ch); + in6_ifattach_destroy(); } #endif - -static int -ip6_init2_vnet(const void *unused __unused) -{ - - /* nd6_timer_init */ - callout_init(&V_nd6_timer_ch, 0); - callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); - - /* timer for regeneranation of temporary addresses randomize ID */ - callout_init(&V_in6_tmpaddrtimer_ch, 0); - callout_reset(&V_in6_tmpaddrtimer_ch, - (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor - - V_ip6_temp_regen_advance) * hz, - in6_tmpaddrtimer, curvnet); - - return (0); -} - -static void -ip6_init2(void *dummy) -{ - - ip6_init2_vnet(NULL); -} - -/* cheat */ -/* This must be after route_init(), which is now SI_ORDER_THIRD */ -SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL); static int ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off, int *nxt, int *ours) { struct ip6_hdr *ip6; struct ip6_hbh *hbh; if (ip6_hopopts_input(plen, rtalert, &m, off)) { #if 0 /*touches NULL pointer*/ in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); #endif goto out; /* m have already been freed */ } /* adjust pointer */ ip6 = mtod(m, struct ip6_hdr *); /* * if the payload length field is 0 and the next header field * indicates Hop-by-Hop Options header, then a Jumbo Payload * option MUST be included. */ if (ip6->ip6_plen == 0 && *plen == 0) { /* * Note that if a valid jumbo payload option is * contained, ip6_hopopts_input() must set a valid * (non-zero) payload length to the variable plen. */ IP6STAT_INC(ip6s_badoptions); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); goto out; } #ifndef PULLDOWN_TEST /* ip6_hopopts_input() ensures that mbuf is contiguous */ hbh = (struct ip6_hbh *)(ip6 + 1); #else IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); if (hbh == NULL) { IP6STAT_INC(ip6s_tooshort); goto out; } #endif *nxt = hbh->ip6h_nxt; /* * If we are acting as a router and the packet contains a * router alert option, see if we know the option value. * Currently, we only support the option value for MLD, in which * case we should pass the packet to the multicast routing * daemon. */ if (*rtalert != ~0) { switch (*rtalert) { case IP6OPT_RTALERT_MLD: if (V_ip6_forwarding) *ours = 1; break; default: /* * RFC2711 requires unrecognized values must be * silently ignored. */ break; } } return (0); out: return (1); } #ifdef RSS /* * IPv6 direct input routine. * * This is called when reinjecting completed fragments where * all of the previous checking and book-keeping has been done. */ void ip6_direct_input(struct mbuf *m) { int off, nxt; int nest; struct m_tag *mtag; struct ip6_direct_ctx *ip6dc; mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); ip6dc = (struct ip6_direct_ctx *)(mtag + 1); nxt = ip6dc->ip6dc_nxt; off = ip6dc->ip6dc_off; nest = 0; m_tag_delete(m, mtag); while (nxt != IPPROTO_DONE) { if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { IP6STAT_INC(ip6s_toomanyhdr); goto bad; } /* * protection against faulty packet - there should be * more sanity checks in header chain processing. */ if (m->m_pkthdr.len < off) { IP6STAT_INC(ip6s_tooshort); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); goto bad; } #ifdef IPSEC /* * enforce IPsec policy checking if we are seeing last header. * note that we do not visit this with protocols with pcb layer * code - like udp/tcp/raw ip. */ if (ip6_ipsec_input(m, nxt)) goto bad; #endif /* IPSEC */ nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); } return; bad: m_freem(m); } #endif void ip6_input(struct mbuf *m) { struct in6_addr odst; struct ip6_hdr *ip6; struct in6_ifaddr *ia; u_int32_t plen; u_int32_t rtalert = ~0; int off = sizeof(struct ip6_hdr), nest; int nxt, ours = 0; int srcrt = 0; #ifdef IPSEC /* * should the inner packet be considered authentic? * see comment in ah4_input(). * NB: m cannot be NULL when passed to the input routine */ m->m_flags &= ~M_AUTHIPHDR; m->m_flags &= ~M_AUTHIPDGM; #endif /* IPSEC */ if (m->m_flags & M_FASTFWD_OURS) { /* * Firewall changed destination to local. */ m->m_flags &= ~M_FASTFWD_OURS; ours = 1; ip6 = mtod(m, struct ip6_hdr *); goto hbhcheck; } /* * mbuf statistics */ if (m->m_flags & M_EXT) { if (m->m_next) IP6STAT_INC(ip6s_mext2m); else IP6STAT_INC(ip6s_mext1); } else { if (m->m_next) { if (m->m_flags & M_LOOP) { IP6STAT_INC(ip6s_m2m[V_loif->if_index]); } else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX) IP6STAT_INC( ip6s_m2m[m->m_pkthdr.rcvif->if_index]); else IP6STAT_INC(ip6s_m2m[0]); } else IP6STAT_INC(ip6s_m1); } /* drop the packet if IPv6 operation is disabled on the IF */ if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) goto bad; in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive); IP6STAT_INC(ip6s_total); #ifndef PULLDOWN_TEST /* * L2 bridge code and some other code can return mbuf chain * that does not conform to KAME requirement. too bad. * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? */ if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { struct mbuf *n; if (m->m_pkthdr.len > MHLEN) n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); else n = m_gethdr(M_NOWAIT, MT_DATA); if (n == NULL) { m_freem(m); return; /* ENOBUFS */ } m_move_pkthdr(n, m); m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); n->m_len = n->m_pkthdr.len; m_freem(m); m = n; } IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */); #endif if (m->m_len < sizeof(struct ip6_hdr)) { struct ifnet *inifp; inifp = m->m_pkthdr.rcvif; if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { IP6STAT_INC(ip6s_toosmall); in6_ifstat_inc(inifp, ifs6_in_hdrerr); return; } } ip6 = mtod(m, struct ip6_hdr *); if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { IP6STAT_INC(ip6s_badvers); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); goto bad; } IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6); /* * Check against address spoofing/corruption. */ if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { /* * XXX: "badscope" is not very suitable for a multicast source. */ IP6STAT_INC(ip6s_badscope); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); goto bad; } if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && !(m->m_flags & M_LOOP)) { /* * In this case, the packet should come from the loopback * interface. However, we cannot just check the if_flags, * because ip6_mloopback() passes the "actual" interface * as the outgoing/incoming interface. */ IP6STAT_INC(ip6s_badscope); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); goto bad; } if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { /* * RFC4291 2.7: * Nodes must not originate a packet to a multicast address * whose scop field contains the reserved value 0; if such * a packet is received, it must be silently dropped. */ IP6STAT_INC(ip6s_badscope); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); goto bad; } #ifdef ALTQ if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { /* packet is dropped by traffic conditioner */ return; } #endif /* * The following check is not documented in specs. A malicious * party may be able to use IPv4 mapped addr to confuse tcp/udp stack * and bypass security checks (act as if it was from 127.0.0.1 by using * IPv6 src ::ffff:127.0.0.1). Be cautious. * * This check chokes if we are in an SIIT cloud. As none of BSDs * support IPv4-less kernel compilation, we cannot support SIIT * environment at all. So, it makes more sense for us to reject any * malicious packets for non-SIIT environment, than try to do a * partial support for SIIT environment. */ if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { IP6STAT_INC(ip6s_badscope); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); goto bad; } #if 0 /* * Reject packets with IPv4 compatible addresses (auto tunnel). * * The code forbids auto tunnel relay case in RFC1933 (the check is * stronger than RFC1933). We may want to re-enable it if mech-xx * is revised to forbid relaying case. */ if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { IP6STAT_INC(ip6s_badscope); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); goto bad; } #endif #ifdef IPSEC /* * Bypass packet filtering for packets previously handled by IPsec. */ if (ip6_ipsec_filtertunnel(m)) goto passin; #endif /* IPSEC */ /* * Run through list of hooks for input packets. * * NB: Beware of the destination address changing * (e.g. by NAT rewriting). When this happens, * tell ip6_forward to do the right thing. */ odst = ip6->ip6_dst; /* Jump over all PFIL processing if hooks are not active. */ if (!PFIL_HOOKED(&V_inet6_pfil_hook)) goto passin; if (pfil_run_hooks(&V_inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL)) return; if (m == NULL) /* consumed by filter */ return; ip6 = mtod(m, struct ip6_hdr *); srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); if (m->m_flags & M_FASTFWD_OURS) { m->m_flags &= ~M_FASTFWD_OURS; ours = 1; goto hbhcheck; } if ((m->m_flags & M_IP6_NEXTHOP) && m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { /* * Directly ship the packet on. This allows forwarding * packets originally destined to us to some other directly * connected host. */ ip6_forward(m, 1); return; } passin: /* * Disambiguate address scope zones (if there is ambiguity). * We first make sure that the original source or destination address * is not in our internal form for scoped addresses. Such addresses * are not necessarily invalid spec-wise, but we cannot accept them due * to the usage conflict. * in6_setscope() then also checks and rejects the cases where src or * dst are the loopback address and the receiving interface * is not loopback. */ if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { IP6STAT_INC(ip6s_badscope); /* XXX */ goto bad; } if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) || in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) { IP6STAT_INC(ip6s_badscope); goto bad; } /* * Multicast check. Assume packet is for us to avoid * prematurely taking locks. */ if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { ours = 1; in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast); goto hbhcheck; } /* * Unicast check * XXX: For now we keep link-local IPv6 addresses with embedded * scope zone id, therefore we use zero zoneid here. */ ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */); if (ia != NULL) { if (ia->ia6_flags & IN6_IFF_NOTREADY) { char ip6bufs[INET6_ADDRSTRLEN]; char ip6bufd[INET6_ADDRSTRLEN]; /* address is not ready, so discard the packet. */ nd6log((LOG_INFO, "ip6_input: packet to an unready address %s->%s\n", ip6_sprintf(ip6bufs, &ip6->ip6_src), ip6_sprintf(ip6bufd, &ip6->ip6_dst))); ifa_free(&ia->ia_ifa); goto bad; } /* Count the packet in the ip address stats */ counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); ifa_free(&ia->ia_ifa); ours = 1; goto hbhcheck; } /* * Now there is no reason to process the packet if it's not our own * and we're not a router. */ if (!V_ip6_forwarding) { IP6STAT_INC(ip6s_cantforward); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); goto bad; } hbhcheck: /* * Process Hop-by-Hop options header if it's contained. * m may be modified in ip6_hopopts_input(). * If a JumboPayload option is included, plen will also be modified. */ plen = (u_int32_t)ntohs(ip6->ip6_plen); if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { if (ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours) != 0) return; } else nxt = ip6->ip6_nxt; /* * Use mbuf flags to propagate Router Alert option to * ICMPv6 layer, as hop-by-hop options have been stripped. */ if (rtalert != ~0) m->m_flags |= M_RTALERT_MLD; /* * Check that the amount of data in the buffers * is as at least much as the IPv6 header would have us expect. * Trim mbufs if longer than we expect. * Drop packet if shorter than we expect. */ if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { IP6STAT_INC(ip6s_tooshort); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); goto bad; } if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { if (m->m_len == m->m_pkthdr.len) { m->m_len = sizeof(struct ip6_hdr) + plen; m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; } else m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); } /* * Forward if desirable. */ if (V_ip6_mrouter && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { /* * If we are acting as a multicast router, all * incoming multicast packets are passed to the * kernel-level multicast forwarding function. * The packet is returned (relatively) intact; if * ip6_mforward() returns a non-zero value, the packet * must be discarded, else it may be accepted below. * * XXX TODO: Check hlim and multicast scope here to avoid * unnecessarily calling into ip6_mforward(). */ if (ip6_mforward && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) { IP6STAT_INC(ip6s_cantforward); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); goto bad; } } else if (!ours) { ip6_forward(m, srcrt); return; } ip6 = mtod(m, struct ip6_hdr *); /* * Malicious party may be able to use IPv4 mapped addr to confuse * tcp/udp stack and bypass security checks (act as if it was from * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. * * For SIIT end node behavior, you may want to disable the check. * However, you will become vulnerable to attacks using IPv4 mapped * source. */ if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { IP6STAT_INC(ip6s_badscope); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); goto bad; } /* * Tell launch routine the next header */ IP6STAT_INC(ip6s_delivered); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_deliver); nest = 0; while (nxt != IPPROTO_DONE) { if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { IP6STAT_INC(ip6s_toomanyhdr); goto bad; } /* * protection against faulty packet - there should be * more sanity checks in header chain processing. */ if (m->m_pkthdr.len < off) { IP6STAT_INC(ip6s_tooshort); in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); goto bad; } #ifdef IPSEC /* * enforce IPsec policy checking if we are seeing last header. * note that we do not visit this with protocols with pcb layer * code - like udp/tcp/raw ip. */ if (ip6_ipsec_input(m, nxt)) goto bad; #endif /* IPSEC */ nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); } return; bad: m_freem(m); } /* * Hop-by-Hop options header processing. If a valid jumbo payload option is * included, the real payload length will be stored in plenp. * * rtalertp - XXX: should be stored more smart way */ static int ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, struct mbuf **mp, int *offp) { struct mbuf *m = *mp; int off = *offp, hbhlen; struct ip6_hbh *hbh; /* validation of the length of the header */ #ifndef PULLDOWN_TEST IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); hbhlen = (hbh->ip6h_len + 1) << 3; IP6_EXTHDR_CHECK(m, off, hbhlen, -1); hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); #else IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); if (hbh == NULL) { IP6STAT_INC(ip6s_tooshort); return -1; } hbhlen = (hbh->ip6h_len + 1) << 3; IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), hbhlen); if (hbh == NULL) { IP6STAT_INC(ip6s_tooshort); return -1; } #endif off += hbhlen; hbhlen -= sizeof(struct ip6_hbh); if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), hbhlen, rtalertp, plenp) < 0) return (-1); *offp = off; *mp = m; return (0); } /* * Search header for all Hop-by-hop options and process each option. * This function is separate from ip6_hopopts_input() in order to * handle a case where the sending node itself process its hop-by-hop * options header. In such a case, the function is called from ip6_output(). * * The function assumes that hbh header is located right after the IPv6 header * (RFC2460 p7), opthead is pointer into data content in m, and opthead to * opthead + hbhlen is located in contiguous memory region. */ int ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, u_int32_t *rtalertp, u_int32_t *plenp) { struct ip6_hdr *ip6; int optlen = 0; u_int8_t *opt = opthead; u_int16_t rtalert_val; u_int32_t jumboplen; const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { switch (*opt) { case IP6OPT_PAD1: optlen = 1; break; case IP6OPT_PADN: if (hbhlen < IP6OPT_MINLEN) { IP6STAT_INC(ip6s_toosmall); goto bad; } optlen = *(opt + 1) + 2; break; case IP6OPT_ROUTER_ALERT: /* XXX may need check for alignment */ if (hbhlen < IP6OPT_RTALERT_LEN) { IP6STAT_INC(ip6s_toosmall); goto bad; } if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { /* XXX stat */ icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, erroff + opt + 1 - opthead); return (-1); } optlen = IP6OPT_RTALERT_LEN; bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); *rtalertp = ntohs(rtalert_val); break; case IP6OPT_JUMBO: /* XXX may need check for alignment */ if (hbhlen < IP6OPT_JUMBO_LEN) { IP6STAT_INC(ip6s_toosmall); goto bad; } if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { /* XXX stat */ icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, erroff + opt + 1 - opthead); return (-1); } optlen = IP6OPT_JUMBO_LEN; /* * IPv6 packets that have non 0 payload length * must not contain a jumbo payload option. */ ip6 = mtod(m, struct ip6_hdr *); if (ip6->ip6_plen) { IP6STAT_INC(ip6s_badoptions); icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, erroff + opt - opthead); return (-1); } /* * We may see jumbolen in unaligned location, so * we'd need to perform bcopy(). */ bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); jumboplen = (u_int32_t)htonl(jumboplen); #if 1 /* * if there are multiple jumbo payload options, * *plenp will be non-zero and the packet will be * rejected. * the behavior may need some debate in ipngwg - * multiple options does not make sense, however, * there's no explicit mention in specification. */ if (*plenp != 0) { IP6STAT_INC(ip6s_badoptions); icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, erroff + opt + 2 - opthead); return (-1); } #endif /* * jumbo payload length must be larger than 65535. */ if (jumboplen <= IPV6_MAXPACKET) { IP6STAT_INC(ip6s_badoptions); icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, erroff + opt + 2 - opthead); return (-1); } *plenp = jumboplen; break; default: /* unknown option */ if (hbhlen < IP6OPT_MINLEN) { IP6STAT_INC(ip6s_toosmall); goto bad; } optlen = ip6_unknown_opt(opt, m, erroff + opt - opthead); if (optlen == -1) return (-1); optlen += 2; break; } } return (0); bad: m_freem(m); return (-1); } /* * Unknown option processing. * The third argument `off' is the offset from the IPv6 header to the option, * which is necessary if the IPv6 header the and option header and IPv6 header * is not contiguous in order to return an ICMPv6 error. */ int ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) { struct ip6_hdr *ip6; switch (IP6OPT_TYPE(*optp)) { case IP6OPT_TYPE_SKIP: /* ignore the option */ return ((int)*(optp + 1)); case IP6OPT_TYPE_DISCARD: /* silently discard */ m_freem(m); return (-1); case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ IP6STAT_INC(ip6s_badoptions); icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); return (-1); case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ IP6STAT_INC(ip6s_badoptions); ip6 = mtod(m, struct ip6_hdr *); if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || (m->m_flags & (M_BCAST|M_MCAST))) m_freem(m); else icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); return (-1); } m_freem(m); /* XXX: NOTREACHED */ return (-1); } /* * Create the "control" list for this pcb. * These functions will not modify mbuf chain at all. * * With KAME mbuf chain restriction: * The routine will be called from upper layer handlers like tcp6_input(). * Thus the routine assumes that the caller (tcp6_input) have already * called IP6_EXTHDR_CHECK() and all the extension headers are located in the * very first mbuf on the mbuf chain. * * ip6_savecontrol_v4 will handle those options that are possible to be * set on a v4-mapped socket. * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those * options and handle the v6-only ones itself. */ struct mbuf ** ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, int *v4only) { struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); #ifdef SO_TIMESTAMP if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { struct timeval tv; microtime(&tv); *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), SCM_TIMESTAMP, SOL_SOCKET); if (*mp) mp = &(*mp)->m_next; } #endif #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) /* RFC 2292 sec. 5 */ if ((inp->inp_flags & IN6P_PKTINFO) != 0) { struct in6_pktinfo pi6; if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { #ifdef INET struct ip *ip; ip = mtod(m, struct ip *); pi6.ipi6_addr.s6_addr32[0] = 0; pi6.ipi6_addr.s6_addr32[1] = 0; pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; #else /* We won't hit this code */ bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); #endif } else { bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); in6_clearscope(&pi6.ipi6_addr); /* XXX */ } pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; *mp = sbcreatecontrol((caddr_t) &pi6, sizeof(struct in6_pktinfo), IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); if (*mp) mp = &(*mp)->m_next; } if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { int hlim; if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { #ifdef INET struct ip *ip; ip = mtod(m, struct ip *); hlim = ip->ip_ttl; #else /* We won't hit this code */ hlim = 0; #endif } else { hlim = ip6->ip6_hlim & 0xff; } *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6); if (*mp) mp = &(*mp)->m_next; } if ((inp->inp_flags & IN6P_TCLASS) != 0) { int tclass; if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { #ifdef INET struct ip *ip; ip = mtod(m, struct ip *); tclass = ip->ip_tos; #else /* We won't hit this code */ tclass = 0; #endif } else { u_int32_t flowinfo; flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); flowinfo >>= 20; tclass = flowinfo & 0xff; } *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int), IPV6_TCLASS, IPPROTO_IPV6); if (*mp) mp = &(*mp)->m_next; } if (v4only != NULL) { if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { *v4only = 1; } else { *v4only = 0; } } return (mp); } void ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) { struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); int v4only = 0; mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); if (v4only) return; /* * IPV6_HOPOPTS socket option. Recall that we required super-user * privilege for the option (see ip6_ctloutput), but it might be too * strict, since there might be some hop-by-hop options which can be * returned to normal user. * See also RFC 2292 section 6 (or RFC 3542 section 8). */ if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { /* * Check if a hop-by-hop options header is contatined in the * received packet, and if so, store the options as ancillary * data. Note that a hop-by-hop options header must be * just after the IPv6 header, which is assured through the * IPv6 input processing. */ if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { struct ip6_hbh *hbh; int hbhlen = 0; #ifdef PULLDOWN_TEST struct mbuf *ext; #endif #ifndef PULLDOWN_TEST hbh = (struct ip6_hbh *)(ip6 + 1); hbhlen = (hbh->ip6h_len + 1) << 3; #else ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), ip6->ip6_nxt); if (ext == NULL) { IP6STAT_INC(ip6s_tooshort); return; } hbh = mtod(ext, struct ip6_hbh *); hbhlen = (hbh->ip6h_len + 1) << 3; if (hbhlen != ext->m_len) { m_freem(ext); IP6STAT_INC(ip6s_tooshort); return; } #endif /* * XXX: We copy the whole header even if a * jumbo payload option is included, the option which * is to be removed before returning according to * RFC2292. * Note: this constraint is removed in RFC3542 */ *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), IPPROTO_IPV6); if (*mp) mp = &(*mp)->m_next; #ifdef PULLDOWN_TEST m_freem(ext); #endif } } if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); /* * Search for destination options headers or routing * header(s) through the header chain, and stores each * header as ancillary data. * Note that the order of the headers remains in * the chain of ancillary data. */ while (1) { /* is explicit loop prevention necessary? */ struct ip6_ext *ip6e = NULL; int elen; #ifdef PULLDOWN_TEST struct mbuf *ext = NULL; #endif /* * if it is not an extension header, don't try to * pull it from the chain. */ switch (nxt) { case IPPROTO_DSTOPTS: case IPPROTO_ROUTING: case IPPROTO_HOPOPTS: case IPPROTO_AH: /* is it possible? */ break; default: goto loopend; } #ifndef PULLDOWN_TEST if (off + sizeof(*ip6e) > m->m_len) goto loopend; ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); if (nxt == IPPROTO_AH) elen = (ip6e->ip6e_len + 2) << 2; else elen = (ip6e->ip6e_len + 1) << 3; if (off + elen > m->m_len) goto loopend; #else ext = ip6_pullexthdr(m, off, nxt); if (ext == NULL) { IP6STAT_INC(ip6s_tooshort); return; } ip6e = mtod(ext, struct ip6_ext *); if (nxt == IPPROTO_AH) elen = (ip6e->ip6e_len + 2) << 2; else elen = (ip6e->ip6e_len + 1) << 3; if (elen != ext->m_len) { m_freem(ext); IP6STAT_INC(ip6s_tooshort); return; } #endif switch (nxt) { case IPPROTO_DSTOPTS: if (!(in6p->inp_flags & IN6P_DSTOPTS)) break; *mp = sbcreatecontrol((caddr_t)ip6e, elen, IS2292(in6p, IPV6_2292DSTOPTS, IPV6_DSTOPTS), IPPROTO_IPV6); if (*mp) mp = &(*mp)->m_next; break; case IPPROTO_ROUTING: if (!(in6p->inp_flags & IN6P_RTHDR)) break; *mp = sbcreatecontrol((caddr_t)ip6e, elen, IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), IPPROTO_IPV6); if (*mp) mp = &(*mp)->m_next; break; case IPPROTO_HOPOPTS: case IPPROTO_AH: /* is it possible? */ break; default: /* * other cases have been filtered in the above. * none will visit this case. here we supply * the code just in case (nxt overwritten or * other cases). */ #ifdef PULLDOWN_TEST m_freem(ext); #endif goto loopend; } /* proceed with the next header. */ off += elen; nxt = ip6e->ip6e_nxt; ip6e = NULL; #ifdef PULLDOWN_TEST m_freem(ext); ext = NULL; #endif } loopend: ; } if (in6p->inp_flags2 & INP_RECVFLOWID) { uint32_t flowid, flow_type; flowid = m->m_pkthdr.flowid; flow_type = M_HASHTYPE_GET(m); /* * XXX should handle the failure of one or the * other - don't populate both? */ *mp = sbcreatecontrol((caddr_t) &flowid, sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6); if (*mp) mp = &(*mp)->m_next; *mp = sbcreatecontrol((caddr_t) &flow_type, sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6); if (*mp) mp = &(*mp)->m_next; } #ifdef RSS if (in6p->inp_flags2 & INP_RECVRSSBUCKETID) { uint32_t flowid, flow_type; uint32_t rss_bucketid; flowid = m->m_pkthdr.flowid; flow_type = M_HASHTYPE_GET(m); if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { *mp = sbcreatecontrol((caddr_t) &rss_bucketid, sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6); if (*mp) mp = &(*mp)->m_next; } } #endif } #undef IS2292 void ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) { struct socket *so; struct mbuf *m_mtu; struct ip6_mtuinfo mtuctl; KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); /* * Notify the error by sending IPV6_PATHMTU ancillary data if * application wanted to know the MTU value. * NOTE: we notify disconnected sockets, because some udp * applications keep sending sockets disconnected. * NOTE: our implementation doesn't notify connected sockets that has * foreign address that is different than given destination addresses * (this is permitted by RFC 3542). */ if ((inp->inp_flags & IN6P_MTU) == 0 || ( !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) return; mtuctl.ip6m_mtu = mtu; mtuctl.ip6m_addr = *dst; if (sa6_recoverscope(&mtuctl.ip6m_addr)) return; if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) return; so = inp->inp_socket; if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) == 0) { m_freem(m_mtu); /* XXX: should count statistics */ } else sorwakeup(so); } #ifdef PULLDOWN_TEST /* * pull single extension header from mbuf chain. returns single mbuf that * contains the result, or NULL on error. */ static struct mbuf * ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) { struct ip6_ext ip6e; size_t elen; struct mbuf *n; #ifdef DIAGNOSTIC switch (nxt) { case IPPROTO_DSTOPTS: case IPPROTO_ROUTING: case IPPROTO_HOPOPTS: case IPPROTO_AH: /* is it possible? */ break; default: printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); } #endif m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); if (nxt == IPPROTO_AH) elen = (ip6e.ip6e_len + 2) << 2; else elen = (ip6e.ip6e_len + 1) << 3; if (elen > MLEN) n = m_getcl(M_NOWAIT, MT_DATA, 0); else n = m_get(M_NOWAIT, MT_DATA); if (n == NULL) return NULL; m_copydata(m, off, elen, mtod(n, caddr_t)); n->m_len = elen; return n; } #endif /* * Get pointer to the previous header followed by the header * currently processed. * XXX: This function supposes that * M includes all headers, * the next header field and the header length field of each header * are valid, and * the sum of each header length equals to OFF. * Because of these assumptions, this function must be called very * carefully. Moreover, it will not be used in the near future when * we develop `neater' mechanism to process extension headers. */ char * ip6_get_prevhdr(const struct mbuf *m, int off) { struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); if (off == sizeof(struct ip6_hdr)) return (&ip6->ip6_nxt); else { int len, nxt; struct ip6_ext *ip6e = NULL; nxt = ip6->ip6_nxt; len = sizeof(struct ip6_hdr); while (len < off) { ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len); switch (nxt) { case IPPROTO_FRAGMENT: len += sizeof(struct ip6_frag); break; case IPPROTO_AH: len += (ip6e->ip6e_len + 2) << 2; break; default: len += (ip6e->ip6e_len + 1) << 3; break; } nxt = ip6e->ip6e_nxt; } if (ip6e) return (&ip6e->ip6e_nxt); else return NULL; } } /* * get next header offset. m will be retained. */ int ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) { struct ip6_hdr ip6; struct ip6_ext ip6e; struct ip6_frag fh; /* just in case */ if (m == NULL) panic("ip6_nexthdr: m == NULL"); if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) return -1; switch (proto) { case IPPROTO_IPV6: if (m->m_pkthdr.len < off + sizeof(ip6)) return -1; m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); if (nxtp) *nxtp = ip6.ip6_nxt; off += sizeof(ip6); return off; case IPPROTO_FRAGMENT: /* * terminate parsing if it is not the first fragment, * it does not make sense to parse through it. */ if (m->m_pkthdr.len < off + sizeof(fh)) return -1; m_copydata(m, off, sizeof(fh), (caddr_t)&fh); /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ if (fh.ip6f_offlg & IP6F_OFF_MASK) return -1; if (nxtp) *nxtp = fh.ip6f_nxt; off += sizeof(struct ip6_frag); return off; case IPPROTO_AH: if (m->m_pkthdr.len < off + sizeof(ip6e)) return -1; m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); if (nxtp) *nxtp = ip6e.ip6e_nxt; off += (ip6e.ip6e_len + 2) << 2; return off; case IPPROTO_HOPOPTS: case IPPROTO_ROUTING: case IPPROTO_DSTOPTS: if (m->m_pkthdr.len < off + sizeof(ip6e)) return -1; m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); if (nxtp) *nxtp = ip6e.ip6e_nxt; off += (ip6e.ip6e_len + 1) << 3; return off; case IPPROTO_NONE: case IPPROTO_ESP: case IPPROTO_IPCOMP: /* give up */ return -1; default: return -1; } /* NOTREACHED */ } /* * get offset for the last header in the chain. m will be kept untainted. */ int ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) { int newoff; int nxt; if (!nxtp) { nxt = -1; nxtp = &nxt; } while (1) { newoff = ip6_nexthdr(m, off, proto, nxtp); if (newoff < 0) return off; else if (newoff < off) return -1; /* invalid */ else if (newoff == off) return newoff; off = newoff; proto = *nxtp; } } /* * System control for IP6 */ u_char inet6ctlerrmap[PRC_NCMDS] = { 0, 0, 0, 0, 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, EMSGSIZE, EHOSTUNREACH, 0, 0, 0, 0, 0, 0, ENOPROTOOPT }; Index: projects/vnet/sys/netinet6/ip6_mroute.c =================================================================== --- projects/vnet/sys/netinet6/ip6_mroute.c (revision 294565) +++ projects/vnet/sys/netinet6/ip6_mroute.c (revision 294566) @@ -1,1969 +1,1969 @@ /*- * Copyright (C) 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $ */ /*- * Copyright (c) 1989 Stephen Deering * Copyright (c) 1992, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Stephen Deering of Stanford University. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93 * BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp */ /* * IP multicast forwarding procedures * * Written by David Waitzman, BBN Labs, August 1988. * Modified by Steve Deering, Stanford, February 1989. * Modified by Mark J. Steiglitz, Stanford, May, 1991 * Modified by Van Jacobson, LBL, January 1993 * Modified by Ajit Thyagarajan, PARC, August 1993 * Modified by Bill Fenner, PARC, April 1994 * * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support */ #include __FBSDID("$FreeBSD$"); #include "opt_inet6.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry"); static int ip6_mdq(struct mbuf *, struct ifnet *, struct mf6c *); static void phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *); static int register_send(struct ip6_hdr *, struct mif6 *, struct mbuf *); static int set_pim6(int *); static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in6 *); extern int in6_mcast_loop; extern struct domain inet6domain; static const struct encaptab *pim6_encap_cookie; static const struct protosw in6_pim_protosw = { .pr_type = SOCK_RAW, .pr_domain = &inet6domain, .pr_protocol = IPPROTO_PIM, .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, .pr_input = pim6_input, .pr_output = rip6_output, .pr_ctloutput = rip6_ctloutput, .pr_usrreqs = &rip6_usrreqs }; static int pim6_encapcheck(const struct mbuf *, int, int, void *); static VNET_DEFINE(int, ip6_mrouter_ver) = 0; #define V_ip6_mrouter_ver VNET(ip6_mrouter_ver) SYSCTL_DECL(_net_inet6); SYSCTL_DECL(_net_inet6_ip6); static SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM"); static struct mrt6stat mrt6stat; SYSCTL_STRUCT(_net_inet6_ip6, OID_AUTO, mrt6stat, CTLFLAG_RW, &mrt6stat, mrt6stat, "Multicast Routing Statistics (struct mrt6stat, netinet6/ip6_mroute.h)"); #define MRT6STAT_INC(name) mrt6stat.name += 1 #define NO_RTE_FOUND 0x1 #define RTE_FOUND 0x2 static struct mtx mrouter6_mtx; #define MROUTER6_LOCK() mtx_lock(&mrouter6_mtx) #define MROUTER6_UNLOCK() mtx_unlock(&mrouter6_mtx) #define MROUTER6_LOCK_ASSERT() do { \ mtx_assert(&mrouter6_mtx, MA_OWNED); \ NET_ASSERT_GIANT(); \ } while (0) #define MROUTER6_LOCK_INIT() \ mtx_init(&mrouter6_mtx, "IPv6 multicast forwarding", NULL, MTX_DEF) #define MROUTER6_LOCK_DESTROY() mtx_destroy(&mrouter6_mtx) static struct mf6c *mf6ctable[MF6CTBLSIZ]; SYSCTL_OPAQUE(_net_inet6_ip6, OID_AUTO, mf6ctable, CTLFLAG_RD, &mf6ctable, sizeof(mf6ctable), "S,*mf6ctable[MF6CTBLSIZ]", "IPv6 Multicast Forwarding Table (struct *mf6ctable[MF6CTBLSIZ], " "netinet6/ip6_mroute.h)"); static struct mtx mfc6_mtx; #define MFC6_LOCK() mtx_lock(&mfc6_mtx) #define MFC6_UNLOCK() mtx_unlock(&mfc6_mtx) #define MFC6_LOCK_ASSERT() do { \ mtx_assert(&mfc6_mtx, MA_OWNED); \ NET_ASSERT_GIANT(); \ } while (0) #define MFC6_LOCK_INIT() \ mtx_init(&mfc6_mtx, "IPv6 multicast forwarding cache", NULL, MTX_DEF) #define MFC6_LOCK_DESTROY() mtx_destroy(&mfc6_mtx) static u_char n6expire[MF6CTBLSIZ]; static struct mif6 mif6table[MAXMIFS]; static int sysctl_mif6table(SYSCTL_HANDLER_ARGS) { struct mif6_sctl *out; int error; out = malloc(sizeof(struct mif6_sctl) * MAXMIFS, M_TEMP, M_WAITOK); for (int i = 0; i < MAXMIFS; i++) { out[i].m6_flags = mif6table[i].m6_flags; out[i].m6_rate_limit = mif6table[i].m6_rate_limit; out[i].m6_lcl_addr = mif6table[i].m6_lcl_addr; if (mif6table[i].m6_ifp != NULL) out[i].m6_ifp = mif6table[i].m6_ifp->if_index; else out[i].m6_ifp = 0; out[i].m6_pkt_in = mif6table[i].m6_pkt_in; out[i].m6_pkt_out = mif6table[i].m6_pkt_out; out[i].m6_bytes_in = mif6table[i].m6_bytes_in; out[i].m6_bytes_out = mif6table[i].m6_bytes_out; } error = SYSCTL_OUT(req, out, sizeof(struct mif6_sctl) * MAXMIFS); free(out, M_TEMP); return (error); } SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, mif6table, CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, sysctl_mif6table, "S,mif6_sctl[MAXMIFS]", "IPv6 Multicast Interfaces (struct mif6_sctl[MAXMIFS], " "netinet6/ip6_mroute.h)"); static struct mtx mif6_mtx; #define MIF6_LOCK() mtx_lock(&mif6_mtx) #define MIF6_UNLOCK() mtx_unlock(&mif6_mtx) #define MIF6_LOCK_ASSERT() mtx_assert(&mif6_mtx, MA_OWNED) #define MIF6_LOCK_INIT() \ mtx_init(&mif6_mtx, "IPv6 multicast interfaces", NULL, MTX_DEF) #define MIF6_LOCK_DESTROY() mtx_destroy(&mif6_mtx) #ifdef MRT6DEBUG static VNET_DEFINE(u_int, mrt6debug) = 0; /* debug level */ #define V_mrt6debug VNET(mrt6debug) #define DEBUG_MFC 0x02 #define DEBUG_FORWARD 0x04 #define DEBUG_EXPIRE 0x08 #define DEBUG_XMIT 0x10 #define DEBUG_REG 0x20 #define DEBUG_PIM 0x40 #define DEBUG_ERR 0x80 #define DEBUG_ANY 0x7f #define MRT6_DLOG(m, fmt, ...) \ if (V_mrt6debug & (m)) \ log(((m) & DEBUG_ERR) ? LOG_ERR: LOG_DEBUG, \ "%s: " fmt "\n", __func__, ##__VA_ARGS__) #else #define MRT6_DLOG(m, fmt, ...) #endif static void expire_upcalls(void *); #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ #define UPCALL_EXPIRE 6 /* number of timeouts */ /* * XXX TODO: maintain a count to if_allmulti() calls in struct ifnet. */ /* * 'Interfaces' associated with decapsulator (so we can tell * packets that went through it from ones that get reflected * by a broken gateway). Different from IPv4 register_if, * these interfaces are linked into the system ifnet list, * because per-interface IPv6 statistics are maintained in * ifp->if_afdata. But it does not have any routes point * to them. I.e., packets can't be sent this way. They * only exist as a placeholder for multicast source * verification. */ static struct ifnet *multicast_register_if6; #define ENCAP_HOPS 64 /* * Private variables. */ static mifi_t nummifs = 0; static mifi_t reg_mif_num = (mifi_t)-1; static struct pim6stat pim6stat; SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_RW, &pim6stat, pim6stat, "PIM Statistics (struct pim6stat, netinet6/pim6_var.h)"); #define PIM6STAT_INC(name) pim6stat.name += 1 static VNET_DEFINE(int, pim6); #define V_pim6 VNET(pim6) /* * Hash function for a source, group entry */ #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \ (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \ (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \ (g).s6_addr32[2] ^ (g).s6_addr32[3]) /* * Find a route for a given origin IPv6 address and Multicast group address. */ #define MF6CFIND(o, g, rt) do { \ struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \ rt = NULL; \ while (_rt) { \ if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \ IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \ (_rt->mf6c_stall == NULL)) { \ rt = _rt; \ break; \ } \ _rt = _rt->mf6c_next; \ } \ if (rt == NULL) { \ MRT6STAT_INC(mrt6s_mfc_misses); \ } \ } while (/*CONSTCOND*/ 0) /* * Macros to compute elapsed time efficiently * Borrowed from Van Jacobson's scheduling code * XXX: replace with timersub() ? */ #define TV_DELTA(a, b, delta) do { \ int xxs; \ \ delta = (a).tv_usec - (b).tv_usec; \ if ((xxs = (a).tv_sec - (b).tv_sec)) { \ switch (xxs) { \ case 2: \ delta += 1000000; \ /* FALLTHROUGH */ \ case 1: \ delta += 1000000; \ break; \ default: \ delta += (1000000 * xxs); \ } \ } \ } while (/*CONSTCOND*/ 0) /* XXX: replace with timercmp(a, b, <) ? */ #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) #ifdef UPCALL_TIMING #define UPCALL_MAX 50 static u_long upcall_data[UPCALL_MAX + 1]; static void collate(); #endif /* UPCALL_TIMING */ static int ip6_mrouter_init(struct socket *, int, int); static int add_m6fc(struct mf6cctl *); static int add_m6if(struct mif6ctl *); static int del_m6fc(struct mf6cctl *); static int del_m6if(mifi_t *); static int del_m6if_locked(mifi_t *); static int get_mif6_cnt(struct sioc_mif_req6 *); static int get_sg_cnt(struct sioc_sg_req6 *); static struct callout expire_upcalls_ch; int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *); int X_ip6_mrouter_done(void); int X_ip6_mrouter_set(struct socket *, struct sockopt *); int X_ip6_mrouter_get(struct socket *, struct sockopt *); int X_mrt6_ioctl(u_long, caddr_t); /* * Handle MRT setsockopt commands to modify the multicast routing tables. */ int X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt) { int error = 0; int optval; struct mif6ctl mifc; struct mf6cctl mfcc; mifi_t mifi; if (so != V_ip6_mrouter && sopt->sopt_name != MRT6_INIT) return (EPERM); switch (sopt->sopt_name) { case MRT6_INIT: #ifdef MRT6_OINIT case MRT6_OINIT: #endif error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval)); if (error) break; error = ip6_mrouter_init(so, optval, sopt->sopt_name); break; case MRT6_DONE: error = X_ip6_mrouter_done(); break; case MRT6_ADD_MIF: error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc)); if (error) break; error = add_m6if(&mifc); break; case MRT6_ADD_MFC: error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc)); if (error) break; error = add_m6fc(&mfcc); break; case MRT6_DEL_MFC: error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc)); if (error) break; error = del_m6fc(&mfcc); break; case MRT6_DEL_MIF: error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi)); if (error) break; error = del_m6if(&mifi); break; case MRT6_PIM: error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval)); if (error) break; error = set_pim6(&optval); break; default: error = EOPNOTSUPP; break; } return (error); } /* * Handle MRT getsockopt commands */ int X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt) { int error = 0; if (so != V_ip6_mrouter) return (EACCES); switch (sopt->sopt_name) { case MRT6_PIM: error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6)); break; } return (error); } /* * Handle ioctl commands to obtain information from the cache */ int X_mrt6_ioctl(u_long cmd, caddr_t data) { int ret; ret = EINVAL; switch (cmd) { case SIOCGETSGCNT_IN6: ret = get_sg_cnt((struct sioc_sg_req6 *)data); break; case SIOCGETMIFCNT_IN6: ret = get_mif6_cnt((struct sioc_mif_req6 *)data); break; default: break; } return (ret); } /* * returns the packet, byte, rpf-failure count for the source group provided */ static int get_sg_cnt(struct sioc_sg_req6 *req) { struct mf6c *rt; int ret; ret = 0; MFC6_LOCK(); MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt); if (rt == NULL) { ret = ESRCH; } else { req->pktcnt = rt->mf6c_pkt_cnt; req->bytecnt = rt->mf6c_byte_cnt; req->wrong_if = rt->mf6c_wrong_if; } MFC6_UNLOCK(); return (ret); } /* * returns the input and output packet and byte counts on the mif provided */ static int get_mif6_cnt(struct sioc_mif_req6 *req) { mifi_t mifi; int ret; ret = 0; mifi = req->mifi; MIF6_LOCK(); if (mifi >= nummifs) { ret = EINVAL; } else { req->icount = mif6table[mifi].m6_pkt_in; req->ocount = mif6table[mifi].m6_pkt_out; req->ibytes = mif6table[mifi].m6_bytes_in; req->obytes = mif6table[mifi].m6_bytes_out; } MIF6_UNLOCK(); return (ret); } static int set_pim6(int *i) { if ((*i != 1) && (*i != 0)) return (EINVAL); V_pim6 = *i; return (0); } /* * Enable multicast routing */ static int ip6_mrouter_init(struct socket *so, int v, int cmd) { MRT6_DLOG(DEBUG_ANY, "so_type = %d, pr_protocol = %d", so->so_type, so->so_proto->pr_protocol); if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_ICMPV6) return (EOPNOTSUPP); if (v != 1) return (ENOPROTOOPT); MROUTER6_LOCK(); if (V_ip6_mrouter != NULL) { MROUTER6_UNLOCK(); return (EADDRINUSE); } V_ip6_mrouter = so; V_ip6_mrouter_ver = cmd; bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); bzero((caddr_t)n6expire, sizeof(n6expire)); V_pim6 = 0;/* used for stubbing out/in pim stuff */ callout_init(&expire_upcalls_ch, 0); callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); MROUTER6_UNLOCK(); MRT6_DLOG(DEBUG_ANY, "finished"); return (0); } /* * Disable IPv6 multicast forwarding. */ int X_ip6_mrouter_done(void) { mifi_t mifi; u_long i; struct mf6c *rt; struct rtdetq *rte; MROUTER6_LOCK(); if (V_ip6_mrouter == NULL) { MROUTER6_UNLOCK(); return (EINVAL); } /* * For each phyint in use, disable promiscuous reception of all IPv6 * multicasts. */ for (mifi = 0; mifi < nummifs; mifi++) { if (mif6table[mifi].m6_ifp && !(mif6table[mifi].m6_flags & MIFF_REGISTER)) { if_allmulti(mif6table[mifi].m6_ifp, 0); } } bzero((caddr_t)mif6table, sizeof(mif6table)); nummifs = 0; V_pim6 = 0; /* used to stub out/in pim specific code */ callout_stop(&expire_upcalls_ch); /* * Free all multicast forwarding cache entries. */ MFC6_LOCK(); for (i = 0; i < MF6CTBLSIZ; i++) { rt = mf6ctable[i]; while (rt) { struct mf6c *frt; for (rte = rt->mf6c_stall; rte != NULL; ) { struct rtdetq *n = rte->next; m_freem(rte->m); free(rte, M_MRTABLE6); rte = n; } frt = rt; rt = rt->mf6c_next; free(frt, M_MRTABLE6); } } bzero((caddr_t)mf6ctable, sizeof(mf6ctable)); MFC6_UNLOCK(); /* * Reset register interface */ if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) { if_detach(multicast_register_if6); if_free(multicast_register_if6); reg_mif_num = (mifi_t)-1; multicast_register_if6 = NULL; } V_ip6_mrouter = NULL; V_ip6_mrouter_ver = 0; MROUTER6_UNLOCK(); MRT6_DLOG(DEBUG_ANY, "finished"); return (0); } static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; /* * Add a mif to the mif table */ static int add_m6if(struct mif6ctl *mifcp) { struct mif6 *mifp; struct ifnet *ifp; int error; MIF6_LOCK(); if (mifcp->mif6c_mifi >= MAXMIFS) { MIF6_UNLOCK(); return (EINVAL); } mifp = mif6table + mifcp->mif6c_mifi; if (mifp->m6_ifp != NULL) { MIF6_UNLOCK(); return (EADDRINUSE); /* XXX: is it appropriate? */ } if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > V_if_index) { MIF6_UNLOCK(); return (ENXIO); } ifp = ifnet_byindex(mifcp->mif6c_pifi); if (mifcp->mif6c_flags & MIFF_REGISTER) { if (reg_mif_num == (mifi_t)-1) { ifp = if_alloc(IFT_OTHER); if_initname(ifp, "register_mif", 0); ifp->if_flags |= IFF_LOOPBACK; if_attach(ifp); multicast_register_if6 = ifp; reg_mif_num = mifcp->mif6c_mifi; /* * it is impossible to guess the ifindex of the * register interface. So mif6c_pifi is automatically * calculated. */ mifcp->mif6c_pifi = ifp->if_index; } else { ifp = multicast_register_if6; } } else { /* Make sure the interface supports multicast */ if ((ifp->if_flags & IFF_MULTICAST) == 0) { MIF6_UNLOCK(); return (EOPNOTSUPP); } error = if_allmulti(ifp, 1); if (error) { MIF6_UNLOCK(); return (error); } } mifp->m6_flags = mifcp->mif6c_flags; mifp->m6_ifp = ifp; /* initialize per mif pkt counters */ mifp->m6_pkt_in = 0; mifp->m6_pkt_out = 0; mifp->m6_bytes_in = 0; mifp->m6_bytes_out = 0; /* Adjust nummifs up if the mifi is higher than nummifs */ if (nummifs <= mifcp->mif6c_mifi) nummifs = mifcp->mif6c_mifi + 1; MIF6_UNLOCK(); MRT6_DLOG(DEBUG_ANY, "mif #%d, phyint %s", mifcp->mif6c_mifi, if_name(ifp)); return (0); } /* * Delete a mif from the mif table */ static int del_m6if_locked(mifi_t *mifip) { struct mif6 *mifp = mif6table + *mifip; mifi_t mifi; struct ifnet *ifp; MIF6_LOCK_ASSERT(); if (*mifip >= nummifs) return (EINVAL); if (mifp->m6_ifp == NULL) return (EINVAL); if (!(mifp->m6_flags & MIFF_REGISTER)) { /* XXX: TODO: Maintain an ALLMULTI refcount in struct ifnet. */ ifp = mifp->m6_ifp; if_allmulti(ifp, 0); } else { if (reg_mif_num != (mifi_t)-1 && multicast_register_if6 != NULL) { if_detach(multicast_register_if6); if_free(multicast_register_if6); reg_mif_num = (mifi_t)-1; multicast_register_if6 = NULL; } } bzero((caddr_t)mifp, sizeof(*mifp)); /* Adjust nummifs down */ for (mifi = nummifs; mifi > 0; mifi--) if (mif6table[mifi - 1].m6_ifp) break; nummifs = mifi; MRT6_DLOG(DEBUG_ANY, "mif %d, nummifs %d", *mifip, nummifs); return (0); } static int del_m6if(mifi_t *mifip) { int cc; MIF6_LOCK(); cc = del_m6if_locked(mifip); MIF6_UNLOCK(); return (cc); } /* * Add an mfc entry */ static int add_m6fc(struct mf6cctl *mfccp) { struct mf6c *rt; u_long hash; struct rtdetq *rte; u_short nstl; char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN]; MFC6_LOCK(); MF6CFIND(mfccp->mf6cc_origin.sin6_addr, mfccp->mf6cc_mcastgrp.sin6_addr, rt); /* If an entry already exists, just update the fields */ if (rt) { MRT6_DLOG(DEBUG_MFC, "no upcall o %s g %s p %x", ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr), ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr), mfccp->mf6cc_parent); rt->mf6c_parent = mfccp->mf6cc_parent; rt->mf6c_ifset = mfccp->mf6cc_ifset; MFC6_UNLOCK(); return (0); } /* * Find the entry for which the upcall was made and update */ hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr, mfccp->mf6cc_mcastgrp.sin6_addr); for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) { if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, &mfccp->mf6cc_origin.sin6_addr) && IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, &mfccp->mf6cc_mcastgrp.sin6_addr) && (rt->mf6c_stall != NULL)) { if (nstl++) log(LOG_ERR, "add_m6fc: %s o %s g %s p %x dbx %p\n", "multiple kernel entries", ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr), ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr), mfccp->mf6cc_parent, rt->mf6c_stall); MRT6_DLOG(DEBUG_MFC, "o %s g %s p %x dbg %p", ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr), ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr), mfccp->mf6cc_parent, rt->mf6c_stall); rt->mf6c_origin = mfccp->mf6cc_origin; rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; rt->mf6c_parent = mfccp->mf6cc_parent; rt->mf6c_ifset = mfccp->mf6cc_ifset; /* initialize pkt counters per src-grp */ rt->mf6c_pkt_cnt = 0; rt->mf6c_byte_cnt = 0; rt->mf6c_wrong_if = 0; rt->mf6c_expire = 0; /* Don't clean this guy up */ n6expire[hash]--; /* free packets Qed at the end of this entry */ for (rte = rt->mf6c_stall; rte != NULL; ) { struct rtdetq *n = rte->next; ip6_mdq(rte->m, rte->ifp, rt); m_freem(rte->m); #ifdef UPCALL_TIMING collate(&(rte->t)); #endif /* UPCALL_TIMING */ free(rte, M_MRTABLE6); rte = n; } rt->mf6c_stall = NULL; } } /* * It is possible that an entry is being inserted without an upcall */ if (nstl == 0) { MRT6_DLOG(DEBUG_MFC, "no upcall h %lu o %s g %s p %x", hash, ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr), ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr), mfccp->mf6cc_parent); for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, &mfccp->mf6cc_origin.sin6_addr)&& IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, &mfccp->mf6cc_mcastgrp.sin6_addr)) { rt->mf6c_origin = mfccp->mf6cc_origin; rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; rt->mf6c_parent = mfccp->mf6cc_parent; rt->mf6c_ifset = mfccp->mf6cc_ifset; /* initialize pkt counters per src-grp */ rt->mf6c_pkt_cnt = 0; rt->mf6c_byte_cnt = 0; rt->mf6c_wrong_if = 0; if (rt->mf6c_expire) n6expire[hash]--; rt->mf6c_expire = 0; } } if (rt == NULL) { /* no upcall, so make a new entry */ rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT); if (rt == NULL) { MFC6_UNLOCK(); return (ENOBUFS); } /* insert new entry at head of hash chain */ rt->mf6c_origin = mfccp->mf6cc_origin; rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp; rt->mf6c_parent = mfccp->mf6cc_parent; rt->mf6c_ifset = mfccp->mf6cc_ifset; /* initialize pkt counters per src-grp */ rt->mf6c_pkt_cnt = 0; rt->mf6c_byte_cnt = 0; rt->mf6c_wrong_if = 0; rt->mf6c_expire = 0; rt->mf6c_stall = NULL; /* link into table */ rt->mf6c_next = mf6ctable[hash]; mf6ctable[hash] = rt; } } MFC6_UNLOCK(); return (0); } #ifdef UPCALL_TIMING /* * collect delay statistics on the upcalls */ static void collate(struct timeval *t) { u_long d; struct timeval tp; u_long delta; GET_TIME(tp); if (TV_LT(*t, tp)) { TV_DELTA(tp, *t, delta); d = delta >> 10; if (d > UPCALL_MAX) d = UPCALL_MAX; ++upcall_data[d]; } } #endif /* UPCALL_TIMING */ /* * Delete an mfc entry */ static int del_m6fc(struct mf6cctl *mfccp) { #ifdef MRT6DEBUG char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN]; #endif struct sockaddr_in6 origin; struct sockaddr_in6 mcastgrp; struct mf6c *rt; struct mf6c **nptr; u_long hash; origin = mfccp->mf6cc_origin; mcastgrp = mfccp->mf6cc_mcastgrp; hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr); MRT6_DLOG(DEBUG_MFC, "orig %s mcastgrp %s", ip6_sprintf(ip6bufo, &origin.sin6_addr), ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr)); MFC6_LOCK(); nptr = &mf6ctable[hash]; while ((rt = *nptr) != NULL) { if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr, &rt->mf6c_origin.sin6_addr) && IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr, &rt->mf6c_mcastgrp.sin6_addr) && rt->mf6c_stall == NULL) break; nptr = &rt->mf6c_next; } if (rt == NULL) { MFC6_UNLOCK(); return (EADDRNOTAVAIL); } *nptr = rt->mf6c_next; free(rt, M_MRTABLE6); MFC6_UNLOCK(); return (0); } static int socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src) { if (s) { if (sbappendaddr(&s->so_rcv, (struct sockaddr *)src, mm, (struct mbuf *)0) != 0) { sorwakeup(s); return (0); } } m_freem(mm); return (-1); } /* * IPv6 multicast forwarding function. This function assumes that the packet * pointed to by "ip6" has arrived on (or is about to be sent to) the interface * pointed to by "ifp", and the packet is to be relayed to other networks * that have members of the packet's destination IPv6 multicast group. * * The packet is returned unscathed to the caller, unless it is * erroneous, in which case a non-zero return value tells the caller to * discard it. * * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff * this function is called in the originating context (i.e., not when * forwarding a packet from other node). ip6_output(), which is currently the * only function that calls this function is called in the originating context, * explicitly ensures this condition. It is caller's responsibility to ensure * that if this function is called from somewhere else in the originating * context in the future. */ int X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m) { struct rtdetq *rte; struct mbuf *mb0; struct mf6c *rt; struct mif6 *mifp; struct mbuf *mm; u_long hash; mifi_t mifi; char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; #ifdef UPCALL_TIMING struct timeval tp; GET_TIME(tp); #endif /* UPCALL_TIMING */ MRT6_DLOG(DEBUG_FORWARD, "src %s, dst %s, ifindex %d", ip6_sprintf(ip6bufs, &ip6->ip6_src), ip6_sprintf(ip6bufd, &ip6->ip6_dst), ifp->if_index); /* * Don't forward a packet with Hop limit of zero or one, * or a packet destined to a local-only group. */ if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) || IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst)) return (0); ip6->ip6_hlim--; /* * Source address check: do not forward packets with unspecified * source. It was discussed in July 2000, on ipngwg mailing list. * This is rather more serious than unicast cases, because some * MLD packets can be sent with the unspecified source address * (although such packets must normally set 1 to the hop limit field). */ if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { IP6STAT_INC(ip6s_cantforward); if (V_ip6_log_time + V_ip6_log_interval < time_uptime) { V_ip6_log_time = time_uptime; log(LOG_DEBUG, "cannot forward " "from %s to %s nxt %d received on %s\n", ip6_sprintf(ip6bufs, &ip6->ip6_src), ip6_sprintf(ip6bufd, &ip6->ip6_dst), ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif)); } return (0); } MFC6_LOCK(); /* * Determine forwarding mifs from the forwarding cache table */ MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt); MRT6STAT_INC(mrt6s_mfc_lookups); /* Entry exists, so forward if necessary */ if (rt) { MFC6_UNLOCK(); return (ip6_mdq(m, ifp, rt)); } /* * If we don't have a route for packet's origin, * Make a copy of the packet & send message to routing daemon. */ MRT6STAT_INC(mrt6s_no_route); MRT6_DLOG(DEBUG_FORWARD | DEBUG_MFC, "no rte s %s g %s", ip6_sprintf(ip6bufs, &ip6->ip6_src), ip6_sprintf(ip6bufd, &ip6->ip6_dst)); /* * Allocate mbufs early so that we don't do extra work if we * are just going to fail anyway. */ rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE6, M_NOWAIT); if (rte == NULL) { MFC6_UNLOCK(); return (ENOBUFS); } mb0 = m_copy(m, 0, M_COPYALL); /* * Pullup packet header if needed before storing it, * as other references may modify it in the meantime. */ if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < sizeof(struct ip6_hdr))) mb0 = m_pullup(mb0, sizeof(struct ip6_hdr)); if (mb0 == NULL) { free(rte, M_MRTABLE6); MFC6_UNLOCK(); return (ENOBUFS); } /* is there an upcall waiting for this packet? */ hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst); for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) { if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &rt->mf6c_origin.sin6_addr) && IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &rt->mf6c_mcastgrp.sin6_addr) && (rt->mf6c_stall != NULL)) break; } if (rt == NULL) { struct mrt6msg *im; #ifdef MRT6_OINIT struct omrt6msg *oim; #endif /* no upcall, so make a new entry */ rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT); if (rt == NULL) { free(rte, M_MRTABLE6); m_freem(mb0); MFC6_UNLOCK(); return (ENOBUFS); } /* * Make a copy of the header to send to the user * level process */ mm = m_copy(mb0, 0, sizeof(struct ip6_hdr)); if (mm == NULL) { free(rte, M_MRTABLE6); m_freem(mb0); free(rt, M_MRTABLE6); MFC6_UNLOCK(); return (ENOBUFS); } /* * Send message to routing daemon */ sin6.sin6_addr = ip6->ip6_src; im = NULL; #ifdef MRT6_OINIT oim = NULL; #endif switch (V_ip6_mrouter_ver) { #ifdef MRT6_OINIT case MRT6_OINIT: oim = mtod(mm, struct omrt6msg *); oim->im6_msgtype = MRT6MSG_NOCACHE; oim->im6_mbz = 0; break; #endif case MRT6_INIT: im = mtod(mm, struct mrt6msg *); im->im6_msgtype = MRT6MSG_NOCACHE; im->im6_mbz = 0; break; default: free(rte, M_MRTABLE6); m_freem(mb0); free(rt, M_MRTABLE6); MFC6_UNLOCK(); return (EINVAL); } MRT6_DLOG(DEBUG_FORWARD, "getting the iif info in the kernel"); for (mifp = mif6table, mifi = 0; mifi < nummifs && mifp->m6_ifp != ifp; mifp++, mifi++) ; switch (V_ip6_mrouter_ver) { #ifdef MRT6_OINIT case MRT6_OINIT: oim->im6_mif = mifi; break; #endif case MRT6_INIT: im->im6_mif = mifi; break; } if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) { log(LOG_WARNING, "ip6_mforward: ip6_mrouter " "socket queue full\n"); MRT6STAT_INC(mrt6s_upq_sockfull); free(rte, M_MRTABLE6); m_freem(mb0); free(rt, M_MRTABLE6); MFC6_UNLOCK(); return (ENOBUFS); } MRT6STAT_INC(mrt6s_upcalls); /* insert new entry at head of hash chain */ bzero(rt, sizeof(*rt)); rt->mf6c_origin.sin6_family = AF_INET6; rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6); rt->mf6c_origin.sin6_addr = ip6->ip6_src; rt->mf6c_mcastgrp.sin6_family = AF_INET6; rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6); rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst; rt->mf6c_expire = UPCALL_EXPIRE; n6expire[hash]++; rt->mf6c_parent = MF6C_INCOMPLETE_PARENT; /* link into table */ rt->mf6c_next = mf6ctable[hash]; mf6ctable[hash] = rt; /* Add this entry to the end of the queue */ rt->mf6c_stall = rte; } else { /* determine if q has overflowed */ struct rtdetq **p; int npkts = 0; for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next) if (++npkts > MAX_UPQ6) { MRT6STAT_INC(mrt6s_upq_ovflw); free(rte, M_MRTABLE6); m_freem(mb0); MFC6_UNLOCK(); return (0); } /* Add this entry to the end of the queue */ *p = rte; } rte->next = NULL; rte->m = mb0; rte->ifp = ifp; #ifdef UPCALL_TIMING rte->t = tp; #endif /* UPCALL_TIMING */ MFC6_UNLOCK(); return (0); } /* * Clean up cache entries if upcalls are not serviced * Call from the Slow Timeout mechanism, every half second. */ static void expire_upcalls(void *unused) { #ifdef MRT6DEBUG char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN]; #endif struct rtdetq *rte; struct mf6c *mfc, **nptr; u_long i; MFC6_LOCK(); for (i = 0; i < MF6CTBLSIZ; i++) { if (n6expire[i] == 0) continue; nptr = &mf6ctable[i]; while ((mfc = *nptr) != NULL) { rte = mfc->mf6c_stall; /* * Skip real cache entries * Make sure it wasn't marked to not expire (shouldn't happen) * If it expires now */ if (rte != NULL && mfc->mf6c_expire != 0 && --mfc->mf6c_expire == 0) { MRT6_DLOG(DEBUG_EXPIRE, "expiring (%s %s)", ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr), ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr)); /* * drop all the packets * free the mbuf with the pkt, if, timing info */ do { struct rtdetq *n = rte->next; m_freem(rte->m); free(rte, M_MRTABLE6); rte = n; } while (rte != NULL); MRT6STAT_INC(mrt6s_cache_cleanups); n6expire[i]--; *nptr = mfc->mf6c_next; free(mfc, M_MRTABLE6); } else { nptr = &mfc->mf6c_next; } } } MFC6_UNLOCK(); callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); } /* * Packet forwarding routine once entry in the cache is made */ static int ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt) { struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); mifi_t mifi, iif; struct mif6 *mifp; int plen = m->m_pkthdr.len; struct in6_addr src0, dst0; /* copies for local work */ u_int32_t iszone, idzone, oszone, odzone; int error = 0; /* * Macro to send packet on mif. Since RSVP packets don't get counted on * input, they shouldn't get counted on output, so statistics keeping is * separate. */ #define MC6_SEND(ip6, mifp, m) do { \ if ((mifp)->m6_flags & MIFF_REGISTER) \ register_send((ip6), (mifp), (m)); \ else \ phyint_send((ip6), (mifp), (m)); \ } while (/*CONSTCOND*/ 0) /* * Don't forward if it didn't arrive from the parent mif * for its origin. */ mifi = rt->mf6c_parent; if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) { /* came in the wrong interface */ MRT6_DLOG(DEBUG_FORWARD, "wrong if: ifid %d mifi %d mififid %x", ifp->if_index, mifi, mif6table[mifi].m6_ifp->if_index); MRT6STAT_INC(mrt6s_wrong_if); rt->mf6c_wrong_if++; /* * If we are doing PIM processing, and we are forwarding * packets on this interface, send a message to the * routing daemon. */ /* have to make sure this is a valid mif */ if (mifi < nummifs && mif6table[mifi].m6_ifp) if (V_pim6 && (m->m_flags & M_LOOP) == 0) { /* * Check the M_LOOP flag to avoid an * unnecessary PIM assert. * XXX: M_LOOP is an ad-hoc hack... */ static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; struct mbuf *mm; struct mrt6msg *im; #ifdef MRT6_OINIT struct omrt6msg *oim; #endif mm = m_copy(m, 0, sizeof(struct ip6_hdr)); if (mm && (!M_WRITABLE(mm) || mm->m_len < sizeof(struct ip6_hdr))) mm = m_pullup(mm, sizeof(struct ip6_hdr)); if (mm == NULL) return (ENOBUFS); #ifdef MRT6_OINIT oim = NULL; #endif im = NULL; switch (V_ip6_mrouter_ver) { #ifdef MRT6_OINIT case MRT6_OINIT: oim = mtod(mm, struct omrt6msg *); oim->im6_msgtype = MRT6MSG_WRONGMIF; oim->im6_mbz = 0; break; #endif case MRT6_INIT: im = mtod(mm, struct mrt6msg *); im->im6_msgtype = MRT6MSG_WRONGMIF; im->im6_mbz = 0; break; default: m_freem(mm); return (EINVAL); } for (mifp = mif6table, iif = 0; iif < nummifs && mifp && mifp->m6_ifp != ifp; mifp++, iif++) ; switch (V_ip6_mrouter_ver) { #ifdef MRT6_OINIT case MRT6_OINIT: oim->im6_mif = iif; sin6.sin6_addr = oim->im6_src; break; #endif case MRT6_INIT: im->im6_mif = iif; sin6.sin6_addr = im->im6_src; break; } MRT6STAT_INC(mrt6s_upcalls); if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) { MRT6_DLOG(DEBUG_ANY, "ip6_mrouter socket queue full"); MRT6STAT_INC(mrt6s_upq_sockfull); return (ENOBUFS); } /* if socket Q full */ } /* if PIM */ return (0); } /* if wrong iif */ /* If I sourced this packet, it counts as output, else it was input. */ if (m->m_pkthdr.rcvif == NULL) { /* XXX: is rcvif really NULL when output?? */ mif6table[mifi].m6_pkt_out++; mif6table[mifi].m6_bytes_out += plen; } else { mif6table[mifi].m6_pkt_in++; mif6table[mifi].m6_bytes_in += plen; } rt->mf6c_pkt_cnt++; rt->mf6c_byte_cnt += plen; /* * For each mif, forward a copy of the packet if there are group * members downstream on the interface. */ src0 = ip6->ip6_src; dst0 = ip6->ip6_dst; if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 || (error = in6_setscope(&dst0, ifp, &idzone)) != 0) { IP6STAT_INC(ip6s_badscope); return (error); } for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++) { if (IF_ISSET(mifi, &rt->mf6c_ifset)) { /* * check if the outgoing packet is going to break * a scope boundary. * XXX For packets through PIM register tunnel * interface, we believe a routing daemon. */ if (!(mif6table[rt->mf6c_parent].m6_flags & MIFF_REGISTER) && !(mif6table[mifi].m6_flags & MIFF_REGISTER)) { if (in6_setscope(&src0, mif6table[mifi].m6_ifp, &oszone) || in6_setscope(&dst0, mif6table[mifi].m6_ifp, &odzone) || iszone != oszone || idzone != odzone) { IP6STAT_INC(ip6s_badscope); continue; } } mifp->m6_pkt_out++; mifp->m6_bytes_out += plen; MC6_SEND(ip6, mifp, m); } } return (0); } static void phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m) { #ifdef MRT6DEBUG char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; #endif struct mbuf *mb_copy; struct ifnet *ifp = mifp->m6_ifp; int error = 0; u_long linkmtu; /* * Make a new reference to the packet; make sure that * the IPv6 header is actually copied, not just referenced, * so that ip6_output() only scribbles on the copy. */ mb_copy = m_copy(m, 0, M_COPYALL); if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr))) mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr)); if (mb_copy == NULL) { return; } /* set MCAST flag to the outgoing packet */ mb_copy->m_flags |= M_MCAST; /* * If we sourced the packet, call ip6_output since we may devide * the packet into fragments when the packet is too big for the * outgoing interface. * Otherwise, we can simply send the packet to the interface * sending queue. */ if (m->m_pkthdr.rcvif == NULL) { struct ip6_moptions im6o; im6o.im6o_multicast_ifp = ifp; /* XXX: ip6_output will override ip6->ip6_hlim */ im6o.im6o_multicast_hlim = ip6->ip6_hlim; im6o.im6o_multicast_loop = 1; error = ip6_output(mb_copy, NULL, NULL, IPV6_FORWARDING, &im6o, NULL, NULL); MRT6_DLOG(DEBUG_XMIT, "mif %u err %d", (uint16_t)(mifp - mif6table), error); return; } /* * If configured to loop back multicasts by default, * loop back a copy now. */ if (in6_mcast_loop) ip6_mloopback(ifp, m); /* * Put the packet into the sending queue of the outgoing interface * if it would fit in the MTU of the interface. */ linkmtu = IN6_LINKMTU(ifp); if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) { struct sockaddr_in6 dst6; bzero(&dst6, sizeof(dst6)); dst6.sin6_len = sizeof(struct sockaddr_in6); dst6.sin6_family = AF_INET6; dst6.sin6_addr = ip6->ip6_dst; IP_PROBE(send, NULL, NULL, ip6, ifp, NULL, ip6); /* * We just call if_output instead of nd6_output here, since * we need no ND for a multicast forwarded packet...right? */ m_clrprotoflags(m); /* Avoid confusing lower layers. */ error = (*ifp->if_output)(ifp, mb_copy, (struct sockaddr *)&dst6, NULL); MRT6_DLOG(DEBUG_XMIT, "mif %u err %d", (uint16_t)(mifp - mif6table), error); } else { /* * pMTU discovery is intentionally disabled by default, since * various router may notify pMTU in multicast, which can be * a DDoS to a router */ if (V_ip6_mcast_pmtu) icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu); else { MRT6_DLOG(DEBUG_XMIT, " packet too big on %s o %s " "g %s size %d (discarded)", if_name(ifp), ip6_sprintf(ip6bufs, &ip6->ip6_src), ip6_sprintf(ip6bufd, &ip6->ip6_dst), mb_copy->m_pkthdr.len); m_freem(mb_copy); /* simply discard the packet */ } } } static int register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m) { #ifdef MRT6DEBUG char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; #endif struct mbuf *mm; int i, len = m->m_pkthdr.len; static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 }; struct mrt6msg *im6; MRT6_DLOG(DEBUG_ANY, "src %s dst %s", ip6_sprintf(ip6bufs, &ip6->ip6_src), ip6_sprintf(ip6bufd, &ip6->ip6_dst)); PIM6STAT_INC(pim6s_snd_registers); /* Make a copy of the packet to send to the user level process. */ mm = m_gethdr(M_NOWAIT, MT_DATA); if (mm == NULL) return (ENOBUFS); mm->m_data += max_linkhdr; mm->m_len = sizeof(struct ip6_hdr); if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { m_freem(mm); return (ENOBUFS); } i = MHLEN - M_LEADINGSPACE(mm); if (i > len) i = len; mm = m_pullup(mm, i); if (mm == NULL) return (ENOBUFS); /* TODO: check it! */ mm->m_pkthdr.len = len + sizeof(struct ip6_hdr); /* * Send message to routing daemon */ sin6.sin6_addr = ip6->ip6_src; im6 = mtod(mm, struct mrt6msg *); im6->im6_msgtype = MRT6MSG_WHOLEPKT; im6->im6_mbz = 0; im6->im6_mif = mif - mif6table; /* iif info is not given for reg. encap.n */ MRT6STAT_INC(mrt6s_upcalls); if (socket_send(V_ip6_mrouter, mm, &sin6) < 0) { MRT6_DLOG(DEBUG_ANY, "ip6_mrouter socket queue full"); MRT6STAT_INC(mrt6s_upq_sockfull); return (ENOBUFS); } return (0); } /* * pim6_encapcheck() is called by the encap6_input() path at runtime to * determine if a packet is for PIM; allowing PIM to be dynamically loaded * into the kernel. */ static int pim6_encapcheck(const struct mbuf *m, int off, int proto, void *arg) { #ifdef DIAGNOSTIC KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM")); #endif if (proto != IPPROTO_PIM) return 0; /* not for us; reject the datagram. */ return 64; /* claim the datagram. */ } /* * PIM sparse mode hook * Receives the pim control messages, and passes them up to the listening * socket, using rip6_input. * The only message processed is the REGISTER pim message; the pim header * is stripped off, and the inner packet is passed to register_mforward. */ int pim6_input(struct mbuf **mp, int *offp, int proto) { struct pim *pim; /* pointer to a pim struct */ struct ip6_hdr *ip6; int pimlen; struct mbuf *m = *mp; int minlen; int off = *offp; PIM6STAT_INC(pim6s_rcv_total); ip6 = mtod(m, struct ip6_hdr *); pimlen = m->m_pkthdr.len - *offp; /* * Validate lengths */ if (pimlen < PIM_MINLEN) { PIM6STAT_INC(pim6s_rcv_tooshort); MRT6_DLOG(DEBUG_PIM, "PIM packet too short"); m_freem(m); return (IPPROTO_DONE); } /* * if the packet is at least as big as a REGISTER, go ahead * and grab the PIM REGISTER header size, to avoid another * possible m_pullup() later. * * PIM_MINLEN == pimhdr + u_int32 == 8 * PIM6_REG_MINLEN == pimhdr + reghdr + eip6hdr == 4 + 4 + 40 */ minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN; /* * Make sure that the IP6 and PIM headers in contiguous memory, and * possibly the PIM REGISTER header */ #ifndef PULLDOWN_TEST IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE); /* adjust pointer */ ip6 = mtod(m, struct ip6_hdr *); /* adjust mbuf to point to the PIM header */ pim = (struct pim *)((caddr_t)ip6 + off); #else IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen); if (pim == NULL) { PIM6STAT_INC(pim6s_rcv_tooshort); return (IPPROTO_DONE); } #endif #define PIM6_CHECKSUM #ifdef PIM6_CHECKSUM { int cksumlen; /* * Validate checksum. * If PIM REGISTER, exclude the data packet */ if (pim->pim_type == PIM_REGISTER) cksumlen = PIM_MINLEN; else cksumlen = pimlen; if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) { PIM6STAT_INC(pim6s_rcv_badsum); MRT6_DLOG(DEBUG_PIM, "invalid checksum"); m_freem(m); return (IPPROTO_DONE); } } #endif /* PIM_CHECKSUM */ /* PIM version check */ if (pim->pim_ver != PIM_VERSION) { PIM6STAT_INC(pim6s_rcv_badversion); MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "incorrect version %d, expecting %d", pim->pim_ver, PIM_VERSION); m_freem(m); return (IPPROTO_DONE); } if (pim->pim_type == PIM_REGISTER) { /* * since this is a REGISTER, we'll make a copy of the register * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the * routing daemon. */ static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 }; struct mbuf *mcp; struct ip6_hdr *eip6; u_int32_t *reghdr; int rc; #ifdef MRT6DEBUG char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; #endif PIM6STAT_INC(pim6s_rcv_registers); if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) { MRT6_DLOG(DEBUG_PIM, "register mif not set: %d", reg_mif_num); m_freem(m); return (IPPROTO_DONE); } reghdr = (u_int32_t *)(pim + 1); if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) goto pim6_input_to_daemon; /* * Validate length */ if (pimlen < PIM6_REG_MINLEN) { PIM6STAT_INC(pim6s_rcv_tooshort); PIM6STAT_INC(pim6s_rcv_badregisters); MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "register packet " "size too small %d from %s", pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src)); m_freem(m); return (IPPROTO_DONE); } eip6 = (struct ip6_hdr *) (reghdr + 1); MRT6_DLOG(DEBUG_PIM, "eip6: %s -> %s, eip6 plen %d", ip6_sprintf(ip6bufs, &eip6->ip6_src), ip6_sprintf(ip6bufd, &eip6->ip6_dst), ntohs(eip6->ip6_plen)); /* verify the version number of the inner packet */ if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { PIM6STAT_INC(pim6s_rcv_badregisters); MRT6_DLOG(DEBUG_ANY, "invalid IP version (%d) " "of the inner packet", (eip6->ip6_vfc & IPV6_VERSION)); m_freem(m); return (IPPROTO_NONE); } /* verify the inner packet is destined to a mcast group */ if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) { PIM6STAT_INC(pim6s_rcv_badregisters); MRT6_DLOG(DEBUG_PIM, "inner packet of register " "is not multicast %s", ip6_sprintf(ip6bufd, &eip6->ip6_dst)); m_freem(m); return (IPPROTO_DONE); } /* * make a copy of the whole header to pass to the daemon later. */ mcp = m_copy(m, 0, off + PIM6_REG_MINLEN); if (mcp == NULL) { MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "pim register: " "could not copy register head"); m_freem(m); return (IPPROTO_DONE); } /* * forward the inner ip6 packet; point m_data at the inner ip6. */ m_adj(m, off + PIM_MINLEN); MRT6_DLOG(DEBUG_PIM, "forwarding decapsulated register: " "src %s, dst %s, mif %d", ip6_sprintf(ip6bufs, &eip6->ip6_src), ip6_sprintf(ip6bufd, &eip6->ip6_dst), reg_mif_num); rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m, dst.sin6_family, 0); /* prepare the register head to send to the mrouting daemon */ m = mcp; } /* * Pass the PIM message up to the daemon; if it is a register message * pass the 'head' only up to the daemon. This includes the * encapsulator ip6 header, pim header, register header and the * encapsulated ip6 header. */ pim6_input_to_daemon: rip6_input(&m, offp, proto); return (IPPROTO_DONE); } static int ip6_mroute_modevent(module_t mod, int type, void *unused) { switch (type) { case MOD_LOAD: MROUTER6_LOCK_INIT(); MFC6_LOCK_INIT(); MIF6_LOCK_INIT(); pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM, pim6_encapcheck, (const struct protosw *)&in6_pim_protosw, NULL); if (pim6_encap_cookie == NULL) { printf("ip6_mroute: unable to attach pim6 encap\n"); MIF6_LOCK_DESTROY(); MFC6_LOCK_DESTROY(); MROUTER6_LOCK_DESTROY(); return (EINVAL); } ip6_mforward = X_ip6_mforward; ip6_mrouter_done = X_ip6_mrouter_done; ip6_mrouter_get = X_ip6_mrouter_get; ip6_mrouter_set = X_ip6_mrouter_set; mrt6_ioctl = X_mrt6_ioctl; break; case MOD_UNLOAD: if (V_ip6_mrouter != NULL) return EINVAL; if (pim6_encap_cookie) { encap_detach(pim6_encap_cookie); pim6_encap_cookie = NULL; } X_ip6_mrouter_done(); ip6_mforward = NULL; ip6_mrouter_done = NULL; ip6_mrouter_get = NULL; ip6_mrouter_set = NULL; mrt6_ioctl = NULL; MIF6_LOCK_DESTROY(); MFC6_LOCK_DESTROY(); MROUTER6_LOCK_DESTROY(); break; default: return (EOPNOTSUPP); } return (0); } static moduledata_t ip6_mroutemod = { "ip6_mroute", ip6_mroute_modevent, 0 }; -DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); +DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_ANY); Index: projects/vnet/sys/netinet6/mld6.c =================================================================== --- projects/vnet/sys/netinet6/mld6.c (revision 294565) +++ projects/vnet/sys/netinet6/mld6.c (revision 294566) @@ -1,3320 +1,3320 @@ /*- * Copyright (c) 2009 Bruce Simpson. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote * products derived from this software without specific prior written * permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: mld6.c,v 1.27 2001/04/04 05:17:30 itojun Exp $ */ /*- * Copyright (c) 1988 Stephen Deering. * Copyright (c) 1992, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Stephen Deering of Stanford University. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)igmp.c 8.1 (Berkeley) 7/19/93 */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifndef KTR_MLD #define KTR_MLD KTR_INET6 #endif static struct mld_ifsoftc * mli_alloc_locked(struct ifnet *); static void mli_delete_locked(const struct ifnet *); static void mld_dispatch_packet(struct mbuf *); static void mld_dispatch_queue(struct mbufq *, int); static void mld_final_leave(struct in6_multi *, struct mld_ifsoftc *); static void mld_fasttimo_vnet(void); static int mld_handle_state_change(struct in6_multi *, struct mld_ifsoftc *); static int mld_initial_join(struct in6_multi *, struct mld_ifsoftc *, const int); #ifdef KTR static char * mld_rec_type_to_str(const int); #endif static void mld_set_version(struct mld_ifsoftc *, const int); static void mld_slowtimo_vnet(void); static int mld_v1_input_query(struct ifnet *, const struct ip6_hdr *, /*const*/ struct mld_hdr *); static int mld_v1_input_report(struct ifnet *, const struct ip6_hdr *, /*const*/ struct mld_hdr *); static void mld_v1_process_group_timer(struct mld_ifsoftc *, struct in6_multi *); static void mld_v1_process_querier_timers(struct mld_ifsoftc *); static int mld_v1_transmit_report(struct in6_multi *, const int); static void mld_v1_update_group(struct in6_multi *, const int); static void mld_v2_cancel_link_timers(struct mld_ifsoftc *); static void mld_v2_dispatch_general_query(struct mld_ifsoftc *); static struct mbuf * mld_v2_encap_report(struct ifnet *, struct mbuf *); static int mld_v2_enqueue_filter_change(struct mbufq *, struct in6_multi *); static int mld_v2_enqueue_group_record(struct mbufq *, struct in6_multi *, const int, const int, const int, const int); static int mld_v2_input_query(struct ifnet *, const struct ip6_hdr *, struct mbuf *, const int, const int); static int mld_v2_merge_state_changes(struct in6_multi *, struct mbufq *); static void mld_v2_process_group_timers(struct mld_ifsoftc *, struct mbufq *, struct mbufq *, struct in6_multi *, const int); static int mld_v2_process_group_query(struct in6_multi *, struct mld_ifsoftc *mli, int, struct mbuf *, const int); static int sysctl_mld_gsr(SYSCTL_HANDLER_ARGS); static int sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS); /* * Normative references: RFC 2710, RFC 3590, RFC 3810. * * Locking: * * The MLD subsystem lock ends up being system-wide for the moment, * but could be per-VIMAGE later on. * * The permitted lock order is: IN6_MULTI_LOCK, MLD_LOCK, IF_ADDR_LOCK. * Any may be taken independently; if any are held at the same * time, the above lock order must be followed. * * IN6_MULTI_LOCK covers in_multi. * * MLD_LOCK covers per-link state and any global variables in this file. * * IF_ADDR_LOCK covers if_multiaddrs, which is used for a variety of * per-link state iterators. * * XXX LOR PREVENTION * A special case for IPv6 is the in6_setscope() routine. ip6_output() * will not accept an ifp; it wants an embedded scope ID, unlike * ip_output(), which happily takes the ifp given to it. The embedded * scope ID is only used by MLD to select the outgoing interface. * * During interface attach and detach, MLD will take MLD_LOCK *after* * the IF_AFDATA_LOCK. * As in6_setscope() takes IF_AFDATA_LOCK then SCOPE_LOCK, we can't call * it with MLD_LOCK held without triggering an LOR. A netisr with indirect * dispatch could work around this, but we'd rather not do that, as it * can introduce other races. * * As such, we exploit the fact that the scope ID is just the interface * index, and embed it in the IPv6 destination address accordingly. * This is potentially NOT VALID for MLDv1 reports, as they * are always sent to the multicast group itself; as MLDv2 * reports are always sent to ff02::16, this is not an issue * when MLDv2 is in use. * * This does not however eliminate the LOR when ip6_output() itself * calls in6_setscope() internally whilst MLD_LOCK is held. This will * trigger a LOR warning in WITNESS when the ifnet is detached. * * The right answer is probably to make IF_AFDATA_LOCK an rwlock, given * how it's used across the network stack. Here we're simply exploiting * the fact that MLD runs at a similar layer in the stack to scope6.c. * * VIMAGE: * * Each in6_multi corresponds to an ifp, and each ifp corresponds * to a vnet in ifp->if_vnet. */ static struct mtx mld_mtx; static MALLOC_DEFINE(M_MLD, "mld", "mld state"); #define MLD_EMBEDSCOPE(pin6, zoneid) \ if (IN6_IS_SCOPE_LINKLOCAL(pin6) || \ IN6_IS_ADDR_MC_INTFACELOCAL(pin6)) \ (pin6)->s6_addr16[1] = htons((zoneid) & 0xFFFF) \ /* * VIMAGE-wide globals. */ static VNET_DEFINE(struct timeval, mld_gsrdelay) = {10, 0}; static VNET_DEFINE(LIST_HEAD(, mld_ifsoftc), mli_head); static VNET_DEFINE(int, interface_timers_running6); static VNET_DEFINE(int, state_change_timers_running6); static VNET_DEFINE(int, current_state_timers_running6); #define V_mld_gsrdelay VNET(mld_gsrdelay) #define V_mli_head VNET(mli_head) #define V_interface_timers_running6 VNET(interface_timers_running6) #define V_state_change_timers_running6 VNET(state_change_timers_running6) #define V_current_state_timers_running6 VNET(current_state_timers_running6) SYSCTL_DECL(_net_inet6); /* Note: Not in any common header. */ SYSCTL_NODE(_net_inet6, OID_AUTO, mld, CTLFLAG_RW, 0, "IPv6 Multicast Listener Discovery"); /* * Virtualized sysctls. */ SYSCTL_PROC(_net_inet6_mld, OID_AUTO, gsrdelay, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, &VNET_NAME(mld_gsrdelay.tv_sec), 0, sysctl_mld_gsr, "I", "Rate limit for MLDv2 Group-and-Source queries in seconds"); /* * Non-virtualized sysctls. */ static SYSCTL_NODE(_net_inet6_mld, OID_AUTO, ifinfo, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mld_ifinfo, "Per-interface MLDv2 state"); static int mld_v1enable = 1; SYSCTL_INT(_net_inet6_mld, OID_AUTO, v1enable, CTLFLAG_RWTUN, &mld_v1enable, 0, "Enable fallback to MLDv1"); static int mld_use_allow = 1; SYSCTL_INT(_net_inet6_mld, OID_AUTO, use_allow, CTLFLAG_RWTUN, &mld_use_allow, 0, "Use ALLOW/BLOCK for RFC 4604 SSM joins/leaves"); /* * Packed Router Alert option structure declaration. */ struct mld_raopt { struct ip6_hbh hbh; struct ip6_opt pad; struct ip6_opt_router ra; } __packed; /* * Router Alert hop-by-hop option header. */ static struct mld_raopt mld_ra = { .hbh = { 0, 0 }, .pad = { .ip6o_type = IP6OPT_PADN, 0 }, .ra = { .ip6or_type = IP6OPT_ROUTER_ALERT, .ip6or_len = IP6OPT_RTALERT_LEN - 2, .ip6or_value[0] = ((IP6OPT_RTALERT_MLD >> 8) & 0xFF), .ip6or_value[1] = (IP6OPT_RTALERT_MLD & 0xFF) } }; static struct ip6_pktopts mld_po; static __inline void mld_save_context(struct mbuf *m, struct ifnet *ifp) { #ifdef VIMAGE m->m_pkthdr.PH_loc.ptr = ifp->if_vnet; #endif /* VIMAGE */ m->m_pkthdr.flowid = ifp->if_index; } static __inline void mld_scrub_context(struct mbuf *m) { m->m_pkthdr.PH_loc.ptr = NULL; m->m_pkthdr.flowid = 0; } /* * Restore context from a queued output chain. * Return saved ifindex. * * VIMAGE: The assertion is there to make sure that we * actually called CURVNET_SET() with what's in the mbuf chain. */ static __inline uint32_t mld_restore_context(struct mbuf *m) { #if defined(VIMAGE) && defined(INVARIANTS) KASSERT(curvnet == m->m_pkthdr.PH_loc.ptr, ("%s: called when curvnet was not restored", __func__)); #endif return (m->m_pkthdr.flowid); } /* * Retrieve or set threshold between group-source queries in seconds. * * VIMAGE: Assume curvnet set by caller. * SMPng: NOTE: Serialized by MLD lock. */ static int sysctl_mld_gsr(SYSCTL_HANDLER_ARGS) { int error; int i; error = sysctl_wire_old_buffer(req, sizeof(int)); if (error) return (error); MLD_LOCK(); i = V_mld_gsrdelay.tv_sec; error = sysctl_handle_int(oidp, &i, 0, req); if (error || !req->newptr) goto out_locked; if (i < -1 || i >= 60) { error = EINVAL; goto out_locked; } CTR2(KTR_MLD, "change mld_gsrdelay from %d to %d", V_mld_gsrdelay.tv_sec, i); V_mld_gsrdelay.tv_sec = i; out_locked: MLD_UNLOCK(); return (error); } /* * Expose struct mld_ifsoftc to userland, keyed by ifindex. * For use by ifmcstat(8). * * SMPng: NOTE: Does an unlocked ifindex space read. * VIMAGE: Assume curvnet set by caller. The node handler itself * is not directly virtualized. */ static int sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS) { int *name; int error; u_int namelen; struct ifnet *ifp; struct mld_ifsoftc *mli; name = (int *)arg1; namelen = arg2; if (req->newptr != NULL) return (EPERM); if (namelen != 1) return (EINVAL); error = sysctl_wire_old_buffer(req, sizeof(struct mld_ifinfo)); if (error) return (error); IN6_MULTI_LOCK(); MLD_LOCK(); if (name[0] <= 0 || name[0] > V_if_index) { error = ENOENT; goto out_locked; } error = ENOENT; ifp = ifnet_byindex(name[0]); if (ifp == NULL) goto out_locked; LIST_FOREACH(mli, &V_mli_head, mli_link) { if (ifp == mli->mli_ifp) { struct mld_ifinfo info; info.mli_version = mli->mli_version; info.mli_v1_timer = mli->mli_v1_timer; info.mli_v2_timer = mli->mli_v2_timer; info.mli_flags = mli->mli_flags; info.mli_rv = mli->mli_rv; info.mli_qi = mli->mli_qi; info.mli_qri = mli->mli_qri; info.mli_uri = mli->mli_uri; error = SYSCTL_OUT(req, &info, sizeof(info)); break; } } out_locked: MLD_UNLOCK(); IN6_MULTI_UNLOCK(); return (error); } /* * Dispatch an entire queue of pending packet chains. * VIMAGE: Assumes the vnet pointer has been set. */ static void mld_dispatch_queue(struct mbufq *mq, int limit) { struct mbuf *m; while ((m = mbufq_dequeue(mq)) != NULL) { CTR3(KTR_MLD, "%s: dispatch %p from %p", __func__, mq, m); mld_dispatch_packet(m); if (--limit == 0) break; } } /* * Filter outgoing MLD report state by group. * * Reports are ALWAYS suppressed for ALL-HOSTS (ff02::1) * and node-local addresses. However, kernel and socket consumers * always embed the KAME scope ID in the address provided, so strip it * when performing comparison. * Note: This is not the same as the *multicast* scope. * * Return zero if the given group is one for which MLD reports * should be suppressed, or non-zero if reports should be issued. */ static __inline int mld_is_addr_reported(const struct in6_addr *addr) { KASSERT(IN6_IS_ADDR_MULTICAST(addr), ("%s: not multicast", __func__)); if (IPV6_ADDR_MC_SCOPE(addr) == IPV6_ADDR_SCOPE_NODELOCAL) return (0); if (IPV6_ADDR_MC_SCOPE(addr) == IPV6_ADDR_SCOPE_LINKLOCAL) { struct in6_addr tmp = *addr; in6_clearscope(&tmp); if (IN6_ARE_ADDR_EQUAL(&tmp, &in6addr_linklocal_allnodes)) return (0); } return (1); } /* * Attach MLD when PF_INET6 is attached to an interface. * * SMPng: Normally called with IF_AFDATA_LOCK held. */ struct mld_ifsoftc * mld_domifattach(struct ifnet *ifp) { struct mld_ifsoftc *mli; CTR3(KTR_MLD, "%s: called for ifp %p(%s)", __func__, ifp, if_name(ifp)); MLD_LOCK(); mli = mli_alloc_locked(ifp); if (!(ifp->if_flags & IFF_MULTICAST)) mli->mli_flags |= MLIF_SILENT; if (mld_use_allow) mli->mli_flags |= MLIF_USEALLOW; MLD_UNLOCK(); return (mli); } /* * VIMAGE: assume curvnet set by caller. */ static struct mld_ifsoftc * mli_alloc_locked(/*const*/ struct ifnet *ifp) { struct mld_ifsoftc *mli; MLD_LOCK_ASSERT(); mli = malloc(sizeof(struct mld_ifsoftc), M_MLD, M_NOWAIT|M_ZERO); if (mli == NULL) goto out; mli->mli_ifp = ifp; mli->mli_version = MLD_VERSION_2; mli->mli_flags = 0; mli->mli_rv = MLD_RV_INIT; mli->mli_qi = MLD_QI_INIT; mli->mli_qri = MLD_QRI_INIT; mli->mli_uri = MLD_URI_INIT; SLIST_INIT(&mli->mli_relinmhead); mbufq_init(&mli->mli_gq, MLD_MAX_RESPONSE_PACKETS); LIST_INSERT_HEAD(&V_mli_head, mli, mli_link); CTR2(KTR_MLD, "allocate mld_ifsoftc for ifp %p(%s)", ifp, if_name(ifp)); out: return (mli); } /* * Hook for ifdetach. * * NOTE: Some finalization tasks need to run before the protocol domain * is detached, but also before the link layer does its cleanup. * Run before link-layer cleanup; cleanup groups, but do not free MLD state. * * SMPng: Caller must hold IN6_MULTI_LOCK(). * Must take IF_ADDR_LOCK() to cover if_multiaddrs iterator. * XXX This routine is also bitten by unlocked ifma_protospec access. */ void mld_ifdetach(struct ifnet *ifp) { struct mld_ifsoftc *mli; struct ifmultiaddr *ifma; struct in6_multi *inm, *tinm; CTR3(KTR_MLD, "%s: called for ifp %p(%s)", __func__, ifp, if_name(ifp)); IN6_MULTI_LOCK_ASSERT(); MLD_LOCK(); mli = MLD_IFINFO(ifp); if (mli->mli_version == MLD_VERSION_2) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET6 || ifma->ifma_protospec == NULL) continue; inm = (struct in6_multi *)ifma->ifma_protospec; if (inm->in6m_state == MLD_LEAVING_MEMBER) { SLIST_INSERT_HEAD(&mli->mli_relinmhead, inm, in6m_nrele); } in6m_clear_recorded(inm); } IF_ADDR_RUNLOCK(ifp); SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead, in6m_nrele, tinm) { SLIST_REMOVE_HEAD(&mli->mli_relinmhead, in6m_nrele); in6m_release_locked(inm); } } MLD_UNLOCK(); } /* * Hook for domifdetach. * Runs after link-layer cleanup; free MLD state. * * SMPng: Normally called with IF_AFDATA_LOCK held. */ void mld_domifdetach(struct ifnet *ifp) { CTR3(KTR_MLD, "%s: called for ifp %p(%s)", __func__, ifp, if_name(ifp)); MLD_LOCK(); mli_delete_locked(ifp); MLD_UNLOCK(); } static void mli_delete_locked(const struct ifnet *ifp) { struct mld_ifsoftc *mli, *tmli; CTR3(KTR_MLD, "%s: freeing mld_ifsoftc for ifp %p(%s)", __func__, ifp, if_name(ifp)); MLD_LOCK_ASSERT(); LIST_FOREACH_SAFE(mli, &V_mli_head, mli_link, tmli) { if (mli->mli_ifp == ifp) { /* * Free deferred General Query responses. */ mbufq_drain(&mli->mli_gq); LIST_REMOVE(mli, mli_link); KASSERT(SLIST_EMPTY(&mli->mli_relinmhead), ("%s: there are dangling in_multi references", __func__)); free(mli, M_MLD); return; } } #ifdef INVARIANTS panic("%s: mld_ifsoftc not found for ifp %p\n", __func__, ifp); #endif } /* * Process a received MLDv1 general or address-specific query. * Assumes that the query header has been pulled up to sizeof(mld_hdr). * * NOTE: Can't be fully const correct as we temporarily embed scope ID in * mld_addr. This is OK as we own the mbuf chain. */ static int mld_v1_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6, /*const*/ struct mld_hdr *mld) { struct ifmultiaddr *ifma; struct mld_ifsoftc *mli; struct in6_multi *inm; int is_general_query; uint16_t timer; #ifdef KTR char ip6tbuf[INET6_ADDRSTRLEN]; #endif is_general_query = 0; if (!mld_v1enable) { CTR3(KTR_MLD, "ignore v1 query %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, if_name(ifp)); return (0); } /* * RFC3810 Section 6.2: MLD queries must originate from * a router's link-local address. */ if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) { CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &ip6->ip6_src), ifp, if_name(ifp)); return (0); } /* * Do address field validation upfront before we accept * the query. */ if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) { /* * MLDv1 General Query. * If this was not sent to the all-nodes group, ignore it. */ struct in6_addr dst; dst = ip6->ip6_dst; in6_clearscope(&dst); if (!IN6_ARE_ADDR_EQUAL(&dst, &in6addr_linklocal_allnodes)) return (EINVAL); is_general_query = 1; } else { /* * Embed scope ID of receiving interface in MLD query for * lookup whilst we don't hold other locks. */ in6_setscope(&mld->mld_addr, ifp, NULL); } IN6_MULTI_LOCK(); MLD_LOCK(); /* * Switch to MLDv1 host compatibility mode. */ mli = MLD_IFINFO(ifp); KASSERT(mli != NULL, ("%s: no mld_ifsoftc for ifp %p", __func__, ifp)); mld_set_version(mli, MLD_VERSION_1); timer = (ntohs(mld->mld_maxdelay) * PR_FASTHZ) / MLD_TIMER_SCALE; if (timer == 0) timer = 1; IF_ADDR_RLOCK(ifp); if (is_general_query) { /* * For each reporting group joined on this * interface, kick the report timer. */ CTR2(KTR_MLD, "process v1 general query on ifp %p(%s)", ifp, if_name(ifp)); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET6 || ifma->ifma_protospec == NULL) continue; inm = (struct in6_multi *)ifma->ifma_protospec; mld_v1_update_group(inm, timer); } } else { /* * MLDv1 Group-Specific Query. * If this is a group-specific MLDv1 query, we need only * look up the single group to process it. */ inm = in6m_lookup_locked(ifp, &mld->mld_addr); if (inm != NULL) { CTR3(KTR_MLD, "process v1 query %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, if_name(ifp)); mld_v1_update_group(inm, timer); } /* XXX Clear embedded scope ID as userland won't expect it. */ in6_clearscope(&mld->mld_addr); } IF_ADDR_RUNLOCK(ifp); MLD_UNLOCK(); IN6_MULTI_UNLOCK(); return (0); } /* * Update the report timer on a group in response to an MLDv1 query. * * If we are becoming the reporting member for this group, start the timer. * If we already are the reporting member for this group, and timer is * below the threshold, reset it. * * We may be updating the group for the first time since we switched * to MLDv2. If we are, then we must clear any recorded source lists, * and transition to REPORTING state; the group timer is overloaded * for group and group-source query responses. * * Unlike MLDv2, the delay per group should be jittered * to avoid bursts of MLDv1 reports. */ static void mld_v1_update_group(struct in6_multi *inm, const int timer) { #ifdef KTR char ip6tbuf[INET6_ADDRSTRLEN]; #endif CTR4(KTR_MLD, "%s: %s/%s timer=%d", __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr), if_name(inm->in6m_ifp), timer); IN6_MULTI_LOCK_ASSERT(); switch (inm->in6m_state) { case MLD_NOT_MEMBER: case MLD_SILENT_MEMBER: break; case MLD_REPORTING_MEMBER: if (inm->in6m_timer != 0 && inm->in6m_timer <= timer) { CTR1(KTR_MLD, "%s: REPORTING and timer running, " "skipping.", __func__); break; } /* FALLTHROUGH */ case MLD_SG_QUERY_PENDING_MEMBER: case MLD_G_QUERY_PENDING_MEMBER: case MLD_IDLE_MEMBER: case MLD_LAZY_MEMBER: case MLD_AWAKENING_MEMBER: CTR1(KTR_MLD, "%s: ->REPORTING", __func__); inm->in6m_state = MLD_REPORTING_MEMBER; inm->in6m_timer = MLD_RANDOM_DELAY(timer); V_current_state_timers_running6 = 1; break; case MLD_SLEEPING_MEMBER: CTR1(KTR_MLD, "%s: ->AWAKENING", __func__); inm->in6m_state = MLD_AWAKENING_MEMBER; break; case MLD_LEAVING_MEMBER: break; } } /* * Process a received MLDv2 general, group-specific or * group-and-source-specific query. * * Assumes that the query header has been pulled up to sizeof(mldv2_query). * * Return 0 if successful, otherwise an appropriate error code is returned. */ static int mld_v2_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6, struct mbuf *m, const int off, const int icmp6len) { struct mld_ifsoftc *mli; struct mldv2_query *mld; struct in6_multi *inm; uint32_t maxdelay, nsrc, qqi; int is_general_query; uint16_t timer; uint8_t qrv; #ifdef KTR char ip6tbuf[INET6_ADDRSTRLEN]; #endif is_general_query = 0; /* * RFC3810 Section 6.2: MLD queries must originate from * a router's link-local address. */ if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) { CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &ip6->ip6_src), ifp, if_name(ifp)); return (0); } CTR2(KTR_MLD, "input v2 query on ifp %p(%s)", ifp, if_name(ifp)); mld = (struct mldv2_query *)(mtod(m, uint8_t *) + off); maxdelay = ntohs(mld->mld_maxdelay); /* in 1/10ths of a second */ if (maxdelay >= 32768) { maxdelay = (MLD_MRC_MANT(maxdelay) | 0x1000) << (MLD_MRC_EXP(maxdelay) + 3); } timer = (maxdelay * PR_FASTHZ) / MLD_TIMER_SCALE; if (timer == 0) timer = 1; qrv = MLD_QRV(mld->mld_misc); if (qrv < 2) { CTR3(KTR_MLD, "%s: clamping qrv %d to %d", __func__, qrv, MLD_RV_INIT); qrv = MLD_RV_INIT; } qqi = mld->mld_qqi; if (qqi >= 128) { qqi = MLD_QQIC_MANT(mld->mld_qqi) << (MLD_QQIC_EXP(mld->mld_qqi) + 3); } nsrc = ntohs(mld->mld_numsrc); if (nsrc > MLD_MAX_GS_SOURCES) return (EMSGSIZE); if (icmp6len < sizeof(struct mldv2_query) + (nsrc * sizeof(struct in6_addr))) return (EMSGSIZE); /* * Do further input validation upfront to avoid resetting timers * should we need to discard this query. */ if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) { /* * A general query with a source list has undefined * behaviour; discard it. */ if (nsrc > 0) return (EINVAL); is_general_query = 1; } else { /* * Embed scope ID of receiving interface in MLD query for * lookup whilst we don't hold other locks (due to KAME * locking lameness). We own this mbuf chain just now. */ in6_setscope(&mld->mld_addr, ifp, NULL); } IN6_MULTI_LOCK(); MLD_LOCK(); mli = MLD_IFINFO(ifp); KASSERT(mli != NULL, ("%s: no mld_ifsoftc for ifp %p", __func__, ifp)); /* * Discard the v2 query if we're in Compatibility Mode. * The RFC is pretty clear that hosts need to stay in MLDv1 mode * until the Old Version Querier Present timer expires. */ if (mli->mli_version != MLD_VERSION_2) goto out_locked; mld_set_version(mli, MLD_VERSION_2); mli->mli_rv = qrv; mli->mli_qi = qqi; mli->mli_qri = maxdelay; CTR4(KTR_MLD, "%s: qrv %d qi %d maxdelay %d", __func__, qrv, qqi, maxdelay); if (is_general_query) { /* * MLDv2 General Query. * * Schedule a current-state report on this ifp for * all groups, possibly containing source lists. * * If there is a pending General Query response * scheduled earlier than the selected delay, do * not schedule any other reports. * Otherwise, reset the interface timer. */ CTR2(KTR_MLD, "process v2 general query on ifp %p(%s)", ifp, if_name(ifp)); if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer) { mli->mli_v2_timer = MLD_RANDOM_DELAY(timer); V_interface_timers_running6 = 1; } } else { /* * MLDv2 Group-specific or Group-and-source-specific Query. * * Group-source-specific queries are throttled on * a per-group basis to defeat denial-of-service attempts. * Queries for groups we are not a member of on this * link are simply ignored. */ IF_ADDR_RLOCK(ifp); inm = in6m_lookup_locked(ifp, &mld->mld_addr); if (inm == NULL) { IF_ADDR_RUNLOCK(ifp); goto out_locked; } if (nsrc > 0) { if (!ratecheck(&inm->in6m_lastgsrtv, &V_mld_gsrdelay)) { CTR1(KTR_MLD, "%s: GS query throttled.", __func__); IF_ADDR_RUNLOCK(ifp); goto out_locked; } } CTR2(KTR_MLD, "process v2 group query on ifp %p(%s)", ifp, if_name(ifp)); /* * If there is a pending General Query response * scheduled sooner than the selected delay, no * further report need be scheduled. * Otherwise, prepare to respond to the * group-specific or group-and-source query. */ if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer) mld_v2_process_group_query(inm, mli, timer, m, off); /* XXX Clear embedded scope ID as userland won't expect it. */ in6_clearscope(&mld->mld_addr); IF_ADDR_RUNLOCK(ifp); } out_locked: MLD_UNLOCK(); IN6_MULTI_UNLOCK(); return (0); } /* * Process a recieved MLDv2 group-specific or group-and-source-specific * query. * Return <0 if any error occured. Currently this is ignored. */ static int mld_v2_process_group_query(struct in6_multi *inm, struct mld_ifsoftc *mli, int timer, struct mbuf *m0, const int off) { struct mldv2_query *mld; int retval; uint16_t nsrc; IN6_MULTI_LOCK_ASSERT(); MLD_LOCK_ASSERT(); retval = 0; mld = (struct mldv2_query *)(mtod(m0, uint8_t *) + off); switch (inm->in6m_state) { case MLD_NOT_MEMBER: case MLD_SILENT_MEMBER: case MLD_SLEEPING_MEMBER: case MLD_LAZY_MEMBER: case MLD_AWAKENING_MEMBER: case MLD_IDLE_MEMBER: case MLD_LEAVING_MEMBER: return (retval); break; case MLD_REPORTING_MEMBER: case MLD_G_QUERY_PENDING_MEMBER: case MLD_SG_QUERY_PENDING_MEMBER: break; } nsrc = ntohs(mld->mld_numsrc); /* * Deal with group-specific queries upfront. * If any group query is already pending, purge any recorded * source-list state if it exists, and schedule a query response * for this group-specific query. */ if (nsrc == 0) { if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER || inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER) { in6m_clear_recorded(inm); timer = min(inm->in6m_timer, timer); } inm->in6m_state = MLD_G_QUERY_PENDING_MEMBER; inm->in6m_timer = MLD_RANDOM_DELAY(timer); V_current_state_timers_running6 = 1; return (retval); } /* * Deal with the case where a group-and-source-specific query has * been received but a group-specific query is already pending. */ if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER) { timer = min(inm->in6m_timer, timer); inm->in6m_timer = MLD_RANDOM_DELAY(timer); V_current_state_timers_running6 = 1; return (retval); } /* * Finally, deal with the case where a group-and-source-specific * query has been received, where a response to a previous g-s-r * query exists, or none exists. * In this case, we need to parse the source-list which the Querier * has provided us with and check if we have any source list filter * entries at T1 for these sources. If we do not, there is no need * schedule a report and the query may be dropped. * If we do, we must record them and schedule a current-state * report for those sources. */ if (inm->in6m_nsrc > 0) { struct mbuf *m; uint8_t *sp; int i, nrecorded; int soff; m = m0; soff = off + sizeof(struct mldv2_query); nrecorded = 0; for (i = 0; i < nsrc; i++) { sp = mtod(m, uint8_t *) + soff; retval = in6m_record_source(inm, (const struct in6_addr *)sp); if (retval < 0) break; nrecorded += retval; soff += sizeof(struct in6_addr); if (soff >= m->m_len) { soff = soff - m->m_len; m = m->m_next; if (m == NULL) break; } } if (nrecorded > 0) { CTR1(KTR_MLD, "%s: schedule response to SG query", __func__); inm->in6m_state = MLD_SG_QUERY_PENDING_MEMBER; inm->in6m_timer = MLD_RANDOM_DELAY(timer); V_current_state_timers_running6 = 1; } } return (retval); } /* * Process a received MLDv1 host membership report. * Assumes mld points to mld_hdr in pulled up mbuf chain. * * NOTE: Can't be fully const correct as we temporarily embed scope ID in * mld_addr. This is OK as we own the mbuf chain. */ static int mld_v1_input_report(struct ifnet *ifp, const struct ip6_hdr *ip6, /*const*/ struct mld_hdr *mld) { struct in6_addr src, dst; struct in6_ifaddr *ia; struct in6_multi *inm; #ifdef KTR char ip6tbuf[INET6_ADDRSTRLEN]; #endif if (!mld_v1enable) { CTR3(KTR_MLD, "ignore v1 report %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, if_name(ifp)); return (0); } if (ifp->if_flags & IFF_LOOPBACK) return (0); /* * MLDv1 reports must originate from a host's link-local address, * or the unspecified address (when booting). */ src = ip6->ip6_src; in6_clearscope(&src); if (!IN6_IS_SCOPE_LINKLOCAL(&src) && !IN6_IS_ADDR_UNSPECIFIED(&src)) { CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &ip6->ip6_src), ifp, if_name(ifp)); return (EINVAL); } /* * RFC2710 Section 4: MLDv1 reports must pertain to a multicast * group, and must be directed to the group itself. */ dst = ip6->ip6_dst; in6_clearscope(&dst); if (!IN6_IS_ADDR_MULTICAST(&mld->mld_addr) || !IN6_ARE_ADDR_EQUAL(&mld->mld_addr, &dst)) { CTR3(KTR_MLD, "ignore v1 query dst %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &ip6->ip6_dst), ifp, if_name(ifp)); return (EINVAL); } /* * Make sure we don't hear our own membership report, as fast * leave requires knowing that we are the only member of a * group. Assume we used the link-local address if available, * otherwise look for ::. * * XXX Note that scope ID comparison is needed for the address * returned by in6ifa_ifpforlinklocal(), but SHOULD NOT be * performed for the on-wire address. */ ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST); if ((ia && IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, IA6_IN6(ia))) || (ia == NULL && IN6_IS_ADDR_UNSPECIFIED(&src))) { if (ia != NULL) ifa_free(&ia->ia_ifa); return (0); } if (ia != NULL) ifa_free(&ia->ia_ifa); CTR3(KTR_MLD, "process v1 report %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, if_name(ifp)); /* * Embed scope ID of receiving interface in MLD query for lookup * whilst we don't hold other locks (due to KAME locking lameness). */ if (!IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) in6_setscope(&mld->mld_addr, ifp, NULL); IN6_MULTI_LOCK(); MLD_LOCK(); IF_ADDR_RLOCK(ifp); /* * MLDv1 report suppression. * If we are a member of this group, and our membership should be * reported, and our group timer is pending or about to be reset, * stop our group timer by transitioning to the 'lazy' state. */ inm = in6m_lookup_locked(ifp, &mld->mld_addr); if (inm != NULL) { struct mld_ifsoftc *mli; mli = inm->in6m_mli; KASSERT(mli != NULL, ("%s: no mli for ifp %p", __func__, ifp)); /* * If we are in MLDv2 host mode, do not allow the * other host's MLDv1 report to suppress our reports. */ if (mli->mli_version == MLD_VERSION_2) goto out_locked; inm->in6m_timer = 0; switch (inm->in6m_state) { case MLD_NOT_MEMBER: case MLD_SILENT_MEMBER: case MLD_SLEEPING_MEMBER: break; case MLD_REPORTING_MEMBER: case MLD_IDLE_MEMBER: case MLD_AWAKENING_MEMBER: CTR3(KTR_MLD, "report suppressed for %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, if_name(ifp)); case MLD_LAZY_MEMBER: inm->in6m_state = MLD_LAZY_MEMBER; break; case MLD_G_QUERY_PENDING_MEMBER: case MLD_SG_QUERY_PENDING_MEMBER: case MLD_LEAVING_MEMBER: break; } } out_locked: IF_ADDR_RUNLOCK(ifp); MLD_UNLOCK(); IN6_MULTI_UNLOCK(); /* XXX Clear embedded scope ID as userland won't expect it. */ in6_clearscope(&mld->mld_addr); return (0); } /* * MLD input path. * * Assume query messages which fit in a single ICMPv6 message header * have been pulled up. * Assume that userland will want to see the message, even if it * otherwise fails kernel input validation; do not free it. * Pullup may however free the mbuf chain m if it fails. * * Return IPPROTO_DONE if we freed m. Otherwise, return 0. */ int mld_input(struct mbuf *m, int off, int icmp6len) { struct ifnet *ifp; struct ip6_hdr *ip6; struct mld_hdr *mld; int mldlen; CTR3(KTR_MLD, "%s: called w/mbuf (%p,%d)", __func__, m, off); ifp = m->m_pkthdr.rcvif; ip6 = mtod(m, struct ip6_hdr *); /* Pullup to appropriate size. */ mld = (struct mld_hdr *)(mtod(m, uint8_t *) + off); if (mld->mld_type == MLD_LISTENER_QUERY && icmp6len >= sizeof(struct mldv2_query)) { mldlen = sizeof(struct mldv2_query); } else { mldlen = sizeof(struct mld_hdr); } IP6_EXTHDR_GET(mld, struct mld_hdr *, m, off, mldlen); if (mld == NULL) { ICMP6STAT_INC(icp6s_badlen); return (IPPROTO_DONE); } /* * Userland needs to see all of this traffic for implementing * the endpoint discovery portion of multicast routing. */ switch (mld->mld_type) { case MLD_LISTENER_QUERY: icmp6_ifstat_inc(ifp, ifs6_in_mldquery); if (icmp6len == sizeof(struct mld_hdr)) { if (mld_v1_input_query(ifp, ip6, mld) != 0) return (0); } else if (icmp6len >= sizeof(struct mldv2_query)) { if (mld_v2_input_query(ifp, ip6, m, off, icmp6len) != 0) return (0); } break; case MLD_LISTENER_REPORT: icmp6_ifstat_inc(ifp, ifs6_in_mldreport); if (mld_v1_input_report(ifp, ip6, mld) != 0) return (0); break; case MLDV2_LISTENER_REPORT: icmp6_ifstat_inc(ifp, ifs6_in_mldreport); break; case MLD_LISTENER_DONE: icmp6_ifstat_inc(ifp, ifs6_in_mlddone); break; default: break; } return (0); } /* * Fast timeout handler (global). * VIMAGE: Timeout handlers are expected to service all vimages. */ void mld_fasttimo(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); mld_fasttimo_vnet(); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * Fast timeout handler (per-vnet). * * VIMAGE: Assume caller has set up our curvnet. */ static void mld_fasttimo_vnet(void) { struct mbufq scq; /* State-change packets */ struct mbufq qrq; /* Query response packets */ struct ifnet *ifp; struct mld_ifsoftc *mli; struct ifmultiaddr *ifma; struct in6_multi *inm, *tinm; int uri_fasthz; uri_fasthz = 0; /* * Quick check to see if any work needs to be done, in order to * minimize the overhead of fasttimo processing. * SMPng: XXX Unlocked reads. */ if (!V_current_state_timers_running6 && !V_interface_timers_running6 && !V_state_change_timers_running6) return; IN6_MULTI_LOCK(); MLD_LOCK(); /* * MLDv2 General Query response timer processing. */ if (V_interface_timers_running6) { CTR1(KTR_MLD, "%s: interface timers running", __func__); V_interface_timers_running6 = 0; LIST_FOREACH(mli, &V_mli_head, mli_link) { if (mli->mli_v2_timer == 0) { /* Do nothing. */ } else if (--mli->mli_v2_timer == 0) { mld_v2_dispatch_general_query(mli); } else { V_interface_timers_running6 = 1; } } } if (!V_current_state_timers_running6 && !V_state_change_timers_running6) goto out_locked; V_current_state_timers_running6 = 0; V_state_change_timers_running6 = 0; CTR1(KTR_MLD, "%s: state change timers running", __func__); /* * MLD host report and state-change timer processing. * Note: Processing a v2 group timer may remove a node. */ LIST_FOREACH(mli, &V_mli_head, mli_link) { ifp = mli->mli_ifp; if (mli->mli_version == MLD_VERSION_2) { uri_fasthz = MLD_RANDOM_DELAY(mli->mli_uri * PR_FASTHZ); mbufq_init(&qrq, MLD_MAX_G_GS_PACKETS); mbufq_init(&scq, MLD_MAX_STATE_CHANGE_PACKETS); } IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET6 || ifma->ifma_protospec == NULL) continue; inm = (struct in6_multi *)ifma->ifma_protospec; switch (mli->mli_version) { case MLD_VERSION_1: mld_v1_process_group_timer(mli, inm); break; case MLD_VERSION_2: mld_v2_process_group_timers(mli, &qrq, &scq, inm, uri_fasthz); break; } } IF_ADDR_RUNLOCK(ifp); switch (mli->mli_version) { case MLD_VERSION_1: /* * Transmit reports for this lifecycle. This * is done while not holding IF_ADDR_LOCK * since this can call * in6ifa_ifpforlinklocal() which locks * IF_ADDR_LOCK internally as well as * ip6_output() to transmit a packet. */ SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead, in6m_nrele, tinm) { SLIST_REMOVE_HEAD(&mli->mli_relinmhead, in6m_nrele); (void)mld_v1_transmit_report(inm, MLD_LISTENER_REPORT); } break; case MLD_VERSION_2: mld_dispatch_queue(&qrq, 0); mld_dispatch_queue(&scq, 0); /* * Free the in_multi reference(s) for * this lifecycle. */ SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead, in6m_nrele, tinm) { SLIST_REMOVE_HEAD(&mli->mli_relinmhead, in6m_nrele); in6m_release_locked(inm); } break; } } out_locked: MLD_UNLOCK(); IN6_MULTI_UNLOCK(); } /* * Update host report group timer. * Will update the global pending timer flags. */ static void mld_v1_process_group_timer(struct mld_ifsoftc *mli, struct in6_multi *inm) { int report_timer_expired; IN6_MULTI_LOCK_ASSERT(); MLD_LOCK_ASSERT(); if (inm->in6m_timer == 0) { report_timer_expired = 0; } else if (--inm->in6m_timer == 0) { report_timer_expired = 1; } else { V_current_state_timers_running6 = 1; return; } switch (inm->in6m_state) { case MLD_NOT_MEMBER: case MLD_SILENT_MEMBER: case MLD_IDLE_MEMBER: case MLD_LAZY_MEMBER: case MLD_SLEEPING_MEMBER: case MLD_AWAKENING_MEMBER: break; case MLD_REPORTING_MEMBER: if (report_timer_expired) { inm->in6m_state = MLD_IDLE_MEMBER; SLIST_INSERT_HEAD(&mli->mli_relinmhead, inm, in6m_nrele); } break; case MLD_G_QUERY_PENDING_MEMBER: case MLD_SG_QUERY_PENDING_MEMBER: case MLD_LEAVING_MEMBER: break; } } /* * Update a group's timers for MLDv2. * Will update the global pending timer flags. * Note: Unlocked read from mli. */ static void mld_v2_process_group_timers(struct mld_ifsoftc *mli, struct mbufq *qrq, struct mbufq *scq, struct in6_multi *inm, const int uri_fasthz) { int query_response_timer_expired; int state_change_retransmit_timer_expired; #ifdef KTR char ip6tbuf[INET6_ADDRSTRLEN]; #endif IN6_MULTI_LOCK_ASSERT(); MLD_LOCK_ASSERT(); query_response_timer_expired = 0; state_change_retransmit_timer_expired = 0; /* * During a transition from compatibility mode back to MLDv2, * a group record in REPORTING state may still have its group * timer active. This is a no-op in this function; it is easier * to deal with it here than to complicate the slow-timeout path. */ if (inm->in6m_timer == 0) { query_response_timer_expired = 0; } else if (--inm->in6m_timer == 0) { query_response_timer_expired = 1; } else { V_current_state_timers_running6 = 1; } if (inm->in6m_sctimer == 0) { state_change_retransmit_timer_expired = 0; } else if (--inm->in6m_sctimer == 0) { state_change_retransmit_timer_expired = 1; } else { V_state_change_timers_running6 = 1; } /* We are in fasttimo, so be quick about it. */ if (!state_change_retransmit_timer_expired && !query_response_timer_expired) return; switch (inm->in6m_state) { case MLD_NOT_MEMBER: case MLD_SILENT_MEMBER: case MLD_SLEEPING_MEMBER: case MLD_LAZY_MEMBER: case MLD_AWAKENING_MEMBER: case MLD_IDLE_MEMBER: break; case MLD_G_QUERY_PENDING_MEMBER: case MLD_SG_QUERY_PENDING_MEMBER: /* * Respond to a previously pending Group-Specific * or Group-and-Source-Specific query by enqueueing * the appropriate Current-State report for * immediate transmission. */ if (query_response_timer_expired) { int retval; retval = mld_v2_enqueue_group_record(qrq, inm, 0, 1, (inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER), 0); CTR2(KTR_MLD, "%s: enqueue record = %d", __func__, retval); inm->in6m_state = MLD_REPORTING_MEMBER; in6m_clear_recorded(inm); } /* FALLTHROUGH */ case MLD_REPORTING_MEMBER: case MLD_LEAVING_MEMBER: if (state_change_retransmit_timer_expired) { /* * State-change retransmission timer fired. * If there are any further pending retransmissions, * set the global pending state-change flag, and * reset the timer. */ if (--inm->in6m_scrv > 0) { inm->in6m_sctimer = uri_fasthz; V_state_change_timers_running6 = 1; } /* * Retransmit the previously computed state-change * report. If there are no further pending * retransmissions, the mbuf queue will be consumed. * Update T0 state to T1 as we have now sent * a state-change. */ (void)mld_v2_merge_state_changes(inm, scq); in6m_commit(inm); CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr), if_name(inm->in6m_ifp)); /* * If we are leaving the group for good, make sure * we release MLD's reference to it. * This release must be deferred using a SLIST, * as we are called from a loop which traverses * the in_ifmultiaddr TAILQ. */ if (inm->in6m_state == MLD_LEAVING_MEMBER && inm->in6m_scrv == 0) { inm->in6m_state = MLD_NOT_MEMBER; SLIST_INSERT_HEAD(&mli->mli_relinmhead, inm, in6m_nrele); } } break; } } /* * Switch to a different version on the given interface, * as per Section 9.12. */ static void mld_set_version(struct mld_ifsoftc *mli, const int version) { int old_version_timer; MLD_LOCK_ASSERT(); CTR4(KTR_MLD, "%s: switching to v%d on ifp %p(%s)", __func__, version, mli->mli_ifp, if_name(mli->mli_ifp)); if (version == MLD_VERSION_1) { /* * Compute the "Older Version Querier Present" timer as per * Section 9.12. */ old_version_timer = (mli->mli_rv * mli->mli_qi) + mli->mli_qri; old_version_timer *= PR_SLOWHZ; mli->mli_v1_timer = old_version_timer; } if (mli->mli_v1_timer > 0 && mli->mli_version != MLD_VERSION_1) { mli->mli_version = MLD_VERSION_1; mld_v2_cancel_link_timers(mli); } } /* * Cancel pending MLDv2 timers for the given link and all groups * joined on it; state-change, general-query, and group-query timers. */ static void mld_v2_cancel_link_timers(struct mld_ifsoftc *mli) { struct ifmultiaddr *ifma; struct ifnet *ifp; struct in6_multi *inm, *tinm; CTR3(KTR_MLD, "%s: cancel v2 timers on ifp %p(%s)", __func__, mli->mli_ifp, if_name(mli->mli_ifp)); IN6_MULTI_LOCK_ASSERT(); MLD_LOCK_ASSERT(); /* * Fast-track this potentially expensive operation * by checking all the global 'timer pending' flags. */ if (!V_interface_timers_running6 && !V_state_change_timers_running6 && !V_current_state_timers_running6) return; mli->mli_v2_timer = 0; ifp = mli->mli_ifp; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET6) continue; inm = (struct in6_multi *)ifma->ifma_protospec; switch (inm->in6m_state) { case MLD_NOT_MEMBER: case MLD_SILENT_MEMBER: case MLD_IDLE_MEMBER: case MLD_LAZY_MEMBER: case MLD_SLEEPING_MEMBER: case MLD_AWAKENING_MEMBER: break; case MLD_LEAVING_MEMBER: /* * If we are leaving the group and switching * version, we need to release the final * reference held for issuing the INCLUDE {}. */ SLIST_INSERT_HEAD(&mli->mli_relinmhead, inm, in6m_nrele); /* FALLTHROUGH */ case MLD_G_QUERY_PENDING_MEMBER: case MLD_SG_QUERY_PENDING_MEMBER: in6m_clear_recorded(inm); /* FALLTHROUGH */ case MLD_REPORTING_MEMBER: inm->in6m_sctimer = 0; inm->in6m_timer = 0; inm->in6m_state = MLD_REPORTING_MEMBER; /* * Free any pending MLDv2 state-change records. */ mbufq_drain(&inm->in6m_scq); break; } } IF_ADDR_RUNLOCK(ifp); SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead, in6m_nrele, tinm) { SLIST_REMOVE_HEAD(&mli->mli_relinmhead, in6m_nrele); in6m_release_locked(inm); } } /* * Global slowtimo handler. * VIMAGE: Timeout handlers are expected to service all vimages. */ void mld_slowtimo(void) { VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK_NOSLEEP(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); mld_slowtimo_vnet(); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK_NOSLEEP(); } /* * Per-vnet slowtimo handler. */ static void mld_slowtimo_vnet(void) { struct mld_ifsoftc *mli; MLD_LOCK(); LIST_FOREACH(mli, &V_mli_head, mli_link) { mld_v1_process_querier_timers(mli); } MLD_UNLOCK(); } /* * Update the Older Version Querier Present timers for a link. * See Section 9.12 of RFC 3810. */ static void mld_v1_process_querier_timers(struct mld_ifsoftc *mli) { MLD_LOCK_ASSERT(); if (mli->mli_version != MLD_VERSION_2 && --mli->mli_v1_timer == 0) { /* * MLDv1 Querier Present timer expired; revert to MLDv2. */ CTR5(KTR_MLD, "%s: transition from v%d -> v%d on %p(%s)", __func__, mli->mli_version, MLD_VERSION_2, mli->mli_ifp, if_name(mli->mli_ifp)); mli->mli_version = MLD_VERSION_2; } } /* * Transmit an MLDv1 report immediately. */ static int mld_v1_transmit_report(struct in6_multi *in6m, const int type) { struct ifnet *ifp; struct in6_ifaddr *ia; struct ip6_hdr *ip6; struct mbuf *mh, *md; struct mld_hdr *mld; IN6_MULTI_LOCK_ASSERT(); MLD_LOCK_ASSERT(); ifp = in6m->in6m_ifp; ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST); /* ia may be NULL if link-local address is tentative. */ mh = m_gethdr(M_NOWAIT, MT_DATA); if (mh == NULL) { if (ia != NULL) ifa_free(&ia->ia_ifa); return (ENOMEM); } md = m_get(M_NOWAIT, MT_DATA); if (md == NULL) { m_free(mh); if (ia != NULL) ifa_free(&ia->ia_ifa); return (ENOMEM); } mh->m_next = md; /* * FUTURE: Consider increasing alignment by ETHER_HDR_LEN, so * that ether_output() does not need to allocate another mbuf * for the header in the most common case. */ M_ALIGN(mh, sizeof(struct ip6_hdr)); mh->m_pkthdr.len = sizeof(struct ip6_hdr) + sizeof(struct mld_hdr); mh->m_len = sizeof(struct ip6_hdr); ip6 = mtod(mh, struct ip6_hdr *); ip6->ip6_flow = 0; ip6->ip6_vfc &= ~IPV6_VERSION_MASK; ip6->ip6_vfc |= IPV6_VERSION; ip6->ip6_nxt = IPPROTO_ICMPV6; ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any; ip6->ip6_dst = in6m->in6m_addr; md->m_len = sizeof(struct mld_hdr); mld = mtod(md, struct mld_hdr *); mld->mld_type = type; mld->mld_code = 0; mld->mld_cksum = 0; mld->mld_maxdelay = 0; mld->mld_reserved = 0; mld->mld_addr = in6m->in6m_addr; in6_clearscope(&mld->mld_addr); mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), sizeof(struct mld_hdr)); mld_save_context(mh, ifp); mh->m_flags |= M_MLDV1; mld_dispatch_packet(mh); if (ia != NULL) ifa_free(&ia->ia_ifa); return (0); } /* * Process a state change from the upper layer for the given IPv6 group. * * Each socket holds a reference on the in_multi in its own ip_moptions. * The socket layer will have made the necessary updates to.the group * state, it is now up to MLD to issue a state change report if there * has been any change between T0 (when the last state-change was issued) * and T1 (now). * * We use the MLDv2 state machine at group level. The MLd module * however makes the decision as to which MLD protocol version to speak. * A state change *from* INCLUDE {} always means an initial join. * A state change *to* INCLUDE {} always means a final leave. * * If delay is non-zero, and the state change is an initial multicast * join, the state change report will be delayed by 'delay' ticks * in units of PR_FASTHZ if MLDv1 is active on the link; otherwise * the initial MLDv2 state change report will be delayed by whichever * is sooner, a pending state-change timer or delay itself. * * VIMAGE: curvnet should have been set by caller, as this routine * is called from the socket option handlers. */ int mld_change_state(struct in6_multi *inm, const int delay) { struct mld_ifsoftc *mli; struct ifnet *ifp; int error; IN6_MULTI_LOCK_ASSERT(); error = 0; /* * Try to detect if the upper layer just asked us to change state * for an interface which has now gone away. */ KASSERT(inm->in6m_ifma != NULL, ("%s: no ifma", __func__)); ifp = inm->in6m_ifma->ifma_ifp; if (ifp != NULL) { /* * Sanity check that netinet6's notion of ifp is the * same as net's. */ KASSERT(inm->in6m_ifp == ifp, ("%s: bad ifp", __func__)); } MLD_LOCK(); mli = MLD_IFINFO(ifp); KASSERT(mli != NULL, ("%s: no mld_ifsoftc for ifp %p", __func__, ifp)); /* * If we detect a state transition to or from MCAST_UNDEFINED * for this group, then we are starting or finishing an MLD * life cycle for this group. */ if (inm->in6m_st[1].iss_fmode != inm->in6m_st[0].iss_fmode) { CTR3(KTR_MLD, "%s: inm transition %d -> %d", __func__, inm->in6m_st[0].iss_fmode, inm->in6m_st[1].iss_fmode); if (inm->in6m_st[0].iss_fmode == MCAST_UNDEFINED) { CTR1(KTR_MLD, "%s: initial join", __func__); error = mld_initial_join(inm, mli, delay); goto out_locked; } else if (inm->in6m_st[1].iss_fmode == MCAST_UNDEFINED) { CTR1(KTR_MLD, "%s: final leave", __func__); mld_final_leave(inm, mli); goto out_locked; } } else { CTR1(KTR_MLD, "%s: filter set change", __func__); } error = mld_handle_state_change(inm, mli); out_locked: MLD_UNLOCK(); return (error); } /* * Perform the initial join for an MLD group. * * When joining a group: * If the group should have its MLD traffic suppressed, do nothing. * MLDv1 starts sending MLDv1 host membership reports. * MLDv2 will schedule an MLDv2 state-change report containing the * initial state of the membership. * * If the delay argument is non-zero, then we must delay sending the * initial state change for delay ticks (in units of PR_FASTHZ). */ static int mld_initial_join(struct in6_multi *inm, struct mld_ifsoftc *mli, const int delay) { struct ifnet *ifp; struct mbufq *mq; int error, retval, syncstates; int odelay; #ifdef KTR char ip6tbuf[INET6_ADDRSTRLEN]; #endif CTR4(KTR_MLD, "%s: initial join %s on ifp %p(%s)", __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr), inm->in6m_ifp, if_name(inm->in6m_ifp)); error = 0; syncstates = 1; ifp = inm->in6m_ifp; IN6_MULTI_LOCK_ASSERT(); MLD_LOCK_ASSERT(); KASSERT(mli && mli->mli_ifp == ifp, ("%s: inconsistent ifp", __func__)); /* * Groups joined on loopback or marked as 'not reported', * enter the MLD_SILENT_MEMBER state and * are never reported in any protocol exchanges. * All other groups enter the appropriate state machine * for the version in use on this link. * A link marked as MLIF_SILENT causes MLD to be completely * disabled for the link. */ if ((ifp->if_flags & IFF_LOOPBACK) || (mli->mli_flags & MLIF_SILENT) || !mld_is_addr_reported(&inm->in6m_addr)) { CTR1(KTR_MLD, "%s: not kicking state machine for silent group", __func__); inm->in6m_state = MLD_SILENT_MEMBER; inm->in6m_timer = 0; } else { /* * Deal with overlapping in_multi lifecycle. * If this group was LEAVING, then make sure * we drop the reference we picked up to keep the * group around for the final INCLUDE {} enqueue. */ if (mli->mli_version == MLD_VERSION_2 && inm->in6m_state == MLD_LEAVING_MEMBER) in6m_release_locked(inm); inm->in6m_state = MLD_REPORTING_MEMBER; switch (mli->mli_version) { case MLD_VERSION_1: /* * If a delay was provided, only use it if * it is greater than the delay normally * used for an MLDv1 state change report, * and delay sending the initial MLDv1 report * by not transitioning to the IDLE state. */ odelay = MLD_RANDOM_DELAY(MLD_V1_MAX_RI * PR_FASTHZ); if (delay) { inm->in6m_timer = max(delay, odelay); V_current_state_timers_running6 = 1; } else { inm->in6m_state = MLD_IDLE_MEMBER; error = mld_v1_transmit_report(inm, MLD_LISTENER_REPORT); if (error == 0) { inm->in6m_timer = odelay; V_current_state_timers_running6 = 1; } } break; case MLD_VERSION_2: /* * Defer update of T0 to T1, until the first copy * of the state change has been transmitted. */ syncstates = 0; /* * Immediately enqueue a State-Change Report for * this interface, freeing any previous reports. * Don't kick the timers if there is nothing to do, * or if an error occurred. */ mq = &inm->in6m_scq; mbufq_drain(mq); retval = mld_v2_enqueue_group_record(mq, inm, 1, 0, 0, (mli->mli_flags & MLIF_USEALLOW)); CTR2(KTR_MLD, "%s: enqueue record = %d", __func__, retval); if (retval <= 0) { error = retval * -1; break; } /* * Schedule transmission of pending state-change * report up to RV times for this link. The timer * will fire at the next mld_fasttimo (~200ms), * giving us an opportunity to merge the reports. * * If a delay was provided to this function, only * use this delay if sooner than the existing one. */ KASSERT(mli->mli_rv > 1, ("%s: invalid robustness %d", __func__, mli->mli_rv)); inm->in6m_scrv = mli->mli_rv; if (delay) { if (inm->in6m_sctimer > 1) { inm->in6m_sctimer = min(inm->in6m_sctimer, delay); } else inm->in6m_sctimer = delay; } else inm->in6m_sctimer = 1; V_state_change_timers_running6 = 1; error = 0; break; } } /* * Only update the T0 state if state change is atomic, * i.e. we don't need to wait for a timer to fire before we * can consider the state change to have been communicated. */ if (syncstates) { in6m_commit(inm); CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr), if_name(inm->in6m_ifp)); } return (error); } /* * Issue an intermediate state change during the life-cycle. */ static int mld_handle_state_change(struct in6_multi *inm, struct mld_ifsoftc *mli) { struct ifnet *ifp; int retval; #ifdef KTR char ip6tbuf[INET6_ADDRSTRLEN]; #endif CTR4(KTR_MLD, "%s: state change for %s on ifp %p(%s)", __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr), inm->in6m_ifp, if_name(inm->in6m_ifp)); ifp = inm->in6m_ifp; IN6_MULTI_LOCK_ASSERT(); MLD_LOCK_ASSERT(); KASSERT(mli && mli->mli_ifp == ifp, ("%s: inconsistent ifp", __func__)); if ((ifp->if_flags & IFF_LOOPBACK) || (mli->mli_flags & MLIF_SILENT) || !mld_is_addr_reported(&inm->in6m_addr) || (mli->mli_version != MLD_VERSION_2)) { if (!mld_is_addr_reported(&inm->in6m_addr)) { CTR1(KTR_MLD, "%s: not kicking state machine for silent group", __func__); } CTR1(KTR_MLD, "%s: nothing to do", __func__); in6m_commit(inm); CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr), if_name(inm->in6m_ifp)); return (0); } mbufq_drain(&inm->in6m_scq); retval = mld_v2_enqueue_group_record(&inm->in6m_scq, inm, 1, 0, 0, (mli->mli_flags & MLIF_USEALLOW)); CTR2(KTR_MLD, "%s: enqueue record = %d", __func__, retval); if (retval <= 0) return (-retval); /* * If record(s) were enqueued, start the state-change * report timer for this group. */ inm->in6m_scrv = mli->mli_rv; inm->in6m_sctimer = 1; V_state_change_timers_running6 = 1; return (0); } /* * Perform the final leave for a multicast address. * * When leaving a group: * MLDv1 sends a DONE message, if and only if we are the reporter. * MLDv2 enqueues a state-change report containing a transition * to INCLUDE {} for immediate transmission. */ static void mld_final_leave(struct in6_multi *inm, struct mld_ifsoftc *mli) { int syncstates; #ifdef KTR char ip6tbuf[INET6_ADDRSTRLEN]; #endif syncstates = 1; CTR4(KTR_MLD, "%s: final leave %s on ifp %p(%s)", __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr), inm->in6m_ifp, if_name(inm->in6m_ifp)); IN6_MULTI_LOCK_ASSERT(); MLD_LOCK_ASSERT(); switch (inm->in6m_state) { case MLD_NOT_MEMBER: case MLD_SILENT_MEMBER: case MLD_LEAVING_MEMBER: /* Already leaving or left; do nothing. */ CTR1(KTR_MLD, "%s: not kicking state machine for silent group", __func__); break; case MLD_REPORTING_MEMBER: case MLD_IDLE_MEMBER: case MLD_G_QUERY_PENDING_MEMBER: case MLD_SG_QUERY_PENDING_MEMBER: if (mli->mli_version == MLD_VERSION_1) { #ifdef INVARIANTS if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER || inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER) panic("%s: MLDv2 state reached, not MLDv2 mode", __func__); #endif mld_v1_transmit_report(inm, MLD_LISTENER_DONE); inm->in6m_state = MLD_NOT_MEMBER; V_current_state_timers_running6 = 1; } else if (mli->mli_version == MLD_VERSION_2) { /* * Stop group timer and all pending reports. * Immediately enqueue a state-change report * TO_IN {} to be sent on the next fast timeout, * giving us an opportunity to merge reports. */ mbufq_drain(&inm->in6m_scq); inm->in6m_timer = 0; inm->in6m_scrv = mli->mli_rv; CTR4(KTR_MLD, "%s: Leaving %s/%s with %d " "pending retransmissions.", __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr), if_name(inm->in6m_ifp), inm->in6m_scrv); if (inm->in6m_scrv == 0) { inm->in6m_state = MLD_NOT_MEMBER; inm->in6m_sctimer = 0; } else { int retval; in6m_acquire_locked(inm); retval = mld_v2_enqueue_group_record( &inm->in6m_scq, inm, 1, 0, 0, (mli->mli_flags & MLIF_USEALLOW)); KASSERT(retval != 0, ("%s: enqueue record = %d", __func__, retval)); inm->in6m_state = MLD_LEAVING_MEMBER; inm->in6m_sctimer = 1; V_state_change_timers_running6 = 1; syncstates = 0; } break; } break; case MLD_LAZY_MEMBER: case MLD_SLEEPING_MEMBER: case MLD_AWAKENING_MEMBER: /* Our reports are suppressed; do nothing. */ break; } if (syncstates) { in6m_commit(inm); CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr), if_name(inm->in6m_ifp)); inm->in6m_st[1].iss_fmode = MCAST_UNDEFINED; CTR3(KTR_MLD, "%s: T1 now MCAST_UNDEFINED for %p/%s", __func__, &inm->in6m_addr, if_name(inm->in6m_ifp)); } } /* * Enqueue an MLDv2 group record to the given output queue. * * If is_state_change is zero, a current-state record is appended. * If is_state_change is non-zero, a state-change report is appended. * * If is_group_query is non-zero, an mbuf packet chain is allocated. * If is_group_query is zero, and if there is a packet with free space * at the tail of the queue, it will be appended to providing there * is enough free space. * Otherwise a new mbuf packet chain is allocated. * * If is_source_query is non-zero, each source is checked to see if * it was recorded for a Group-Source query, and will be omitted if * it is not both in-mode and recorded. * * If use_block_allow is non-zero, state change reports for initial join * and final leave, on an inclusive mode group with a source list, will be * rewritten to use the ALLOW_NEW and BLOCK_OLD record types, respectively. * * The function will attempt to allocate leading space in the packet * for the IPv6+ICMP headers to be prepended without fragmenting the chain. * * If successful the size of all data appended to the queue is returned, * otherwise an error code less than zero is returned, or zero if * no record(s) were appended. */ static int mld_v2_enqueue_group_record(struct mbufq *mq, struct in6_multi *inm, const int is_state_change, const int is_group_query, const int is_source_query, const int use_block_allow) { struct mldv2_record mr; struct mldv2_record *pmr; struct ifnet *ifp; struct ip6_msource *ims, *nims; struct mbuf *m0, *m, *md; int error, is_filter_list_change; int minrec0len, m0srcs, msrcs, nbytes, off; int record_has_sources; int now; int type; uint8_t mode; #ifdef KTR char ip6tbuf[INET6_ADDRSTRLEN]; #endif IN6_MULTI_LOCK_ASSERT(); error = 0; ifp = inm->in6m_ifp; is_filter_list_change = 0; m = NULL; m0 = NULL; m0srcs = 0; msrcs = 0; nbytes = 0; nims = NULL; record_has_sources = 1; pmr = NULL; type = MLD_DO_NOTHING; mode = inm->in6m_st[1].iss_fmode; /* * If we did not transition out of ASM mode during t0->t1, * and there are no source nodes to process, we can skip * the generation of source records. */ if (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0 && inm->in6m_nsrc == 0) record_has_sources = 0; if (is_state_change) { /* * Queue a state change record. * If the mode did not change, and there are non-ASM * listeners or source filters present, * we potentially need to issue two records for the group. * If there are ASM listeners, and there was no filter * mode transition of any kind, do nothing. * * If we are transitioning to MCAST_UNDEFINED, we need * not send any sources. A transition to/from this state is * considered inclusive with some special treatment. * * If we are rewriting initial joins/leaves to use * ALLOW/BLOCK, and the group's membership is inclusive, * we need to send sources in all cases. */ if (mode != inm->in6m_st[0].iss_fmode) { if (mode == MCAST_EXCLUDE) { CTR1(KTR_MLD, "%s: change to EXCLUDE", __func__); type = MLD_CHANGE_TO_EXCLUDE_MODE; } else { CTR1(KTR_MLD, "%s: change to INCLUDE", __func__); if (use_block_allow) { /* * XXX * Here we're interested in state * edges either direction between * MCAST_UNDEFINED and MCAST_INCLUDE. * Perhaps we should just check * the group state, rather than * the filter mode. */ if (mode == MCAST_UNDEFINED) { type = MLD_BLOCK_OLD_SOURCES; } else { type = MLD_ALLOW_NEW_SOURCES; } } else { type = MLD_CHANGE_TO_INCLUDE_MODE; if (mode == MCAST_UNDEFINED) record_has_sources = 0; } } } else { if (record_has_sources) { is_filter_list_change = 1; } else { type = MLD_DO_NOTHING; } } } else { /* * Queue a current state record. */ if (mode == MCAST_EXCLUDE) { type = MLD_MODE_IS_EXCLUDE; } else if (mode == MCAST_INCLUDE) { type = MLD_MODE_IS_INCLUDE; KASSERT(inm->in6m_st[1].iss_asm == 0, ("%s: inm %p is INCLUDE but ASM count is %d", __func__, inm, inm->in6m_st[1].iss_asm)); } } /* * Generate the filter list changes using a separate function. */ if (is_filter_list_change) return (mld_v2_enqueue_filter_change(mq, inm)); if (type == MLD_DO_NOTHING) { CTR3(KTR_MLD, "%s: nothing to do for %s/%s", __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr), if_name(inm->in6m_ifp)); return (0); } /* * If any sources are present, we must be able to fit at least * one in the trailing space of the tail packet's mbuf, * ideally more. */ minrec0len = sizeof(struct mldv2_record); if (record_has_sources) minrec0len += sizeof(struct in6_addr); CTR4(KTR_MLD, "%s: queueing %s for %s/%s", __func__, mld_rec_type_to_str(type), ip6_sprintf(ip6tbuf, &inm->in6m_addr), if_name(inm->in6m_ifp)); /* * Check if we have a packet in the tail of the queue for this * group into which the first group record for this group will fit. * Otherwise allocate a new packet. * Always allocate leading space for IP6+RA+ICMPV6+REPORT. * Note: Group records for G/GSR query responses MUST be sent * in their own packet. */ m0 = mbufq_last(mq); if (!is_group_query && m0 != NULL && (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <= MLD_V2_REPORT_MAXRECS) && (m0->m_pkthdr.len + minrec0len) < (ifp->if_mtu - MLD_MTUSPACE)) { m0srcs = (ifp->if_mtu - m0->m_pkthdr.len - sizeof(struct mldv2_record)) / sizeof(struct in6_addr); m = m0; CTR1(KTR_MLD, "%s: use existing packet", __func__); } else { if (mbufq_full(mq)) { CTR1(KTR_MLD, "%s: outbound queue full", __func__); return (-ENOMEM); } m = NULL; m0srcs = (ifp->if_mtu - MLD_MTUSPACE - sizeof(struct mldv2_record)) / sizeof(struct in6_addr); if (!is_state_change && !is_group_query) m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (m == NULL) m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) return (-ENOMEM); mld_save_context(m, ifp); CTR1(KTR_MLD, "%s: allocated first packet", __func__); } /* * Append group record. * If we have sources, we don't know how many yet. */ mr.mr_type = type; mr.mr_datalen = 0; mr.mr_numsrc = 0; mr.mr_addr = inm->in6m_addr; in6_clearscope(&mr.mr_addr); if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) { if (m != m0) m_freem(m); CTR1(KTR_MLD, "%s: m_append() failed.", __func__); return (-ENOMEM); } nbytes += sizeof(struct mldv2_record); /* * Append as many sources as will fit in the first packet. * If we are appending to a new packet, the chain allocation * may potentially use clusters; use m_getptr() in this case. * If we are appending to an existing packet, we need to obtain * a pointer to the group record after m_append(), in case a new * mbuf was allocated. * * Only append sources which are in-mode at t1. If we are * transitioning to MCAST_UNDEFINED state on the group, and * use_block_allow is zero, do not include source entries. * Otherwise, we need to include this source in the report. * * Only report recorded sources in our filter set when responding * to a group-source query. */ if (record_has_sources) { if (m == m0) { md = m_last(m); pmr = (struct mldv2_record *)(mtod(md, uint8_t *) + md->m_len - nbytes); } else { md = m_getptr(m, 0, &off); pmr = (struct mldv2_record *)(mtod(md, uint8_t *) + off); } msrcs = 0; RB_FOREACH_SAFE(ims, ip6_msource_tree, &inm->in6m_srcs, nims) { CTR2(KTR_MLD, "%s: visit node %s", __func__, ip6_sprintf(ip6tbuf, &ims->im6s_addr)); now = im6s_get_mode(inm, ims, 1); CTR2(KTR_MLD, "%s: node is %d", __func__, now); if ((now != mode) || (now == mode && (!use_block_allow && mode == MCAST_UNDEFINED))) { CTR1(KTR_MLD, "%s: skip node", __func__); continue; } if (is_source_query && ims->im6s_stp == 0) { CTR1(KTR_MLD, "%s: skip unrecorded node", __func__); continue; } CTR1(KTR_MLD, "%s: append node", __func__); if (!m_append(m, sizeof(struct in6_addr), (void *)&ims->im6s_addr)) { if (m != m0) m_freem(m); CTR1(KTR_MLD, "%s: m_append() failed.", __func__); return (-ENOMEM); } nbytes += sizeof(struct in6_addr); ++msrcs; if (msrcs == m0srcs) break; } CTR2(KTR_MLD, "%s: msrcs is %d this packet", __func__, msrcs); pmr->mr_numsrc = htons(msrcs); nbytes += (msrcs * sizeof(struct in6_addr)); } if (is_source_query && msrcs == 0) { CTR1(KTR_MLD, "%s: no recorded sources to report", __func__); if (m != m0) m_freem(m); return (0); } /* * We are good to go with first packet. */ if (m != m0) { CTR1(KTR_MLD, "%s: enqueueing first packet", __func__); m->m_pkthdr.PH_vt.vt_nrecs = 1; mbufq_enqueue(mq, m); } else m->m_pkthdr.PH_vt.vt_nrecs++; /* * No further work needed if no source list in packet(s). */ if (!record_has_sources) return (nbytes); /* * Whilst sources remain to be announced, we need to allocate * a new packet and fill out as many sources as will fit. * Always try for a cluster first. */ while (nims != NULL) { if (mbufq_full(mq)) { CTR1(KTR_MLD, "%s: outbound queue full", __func__); return (-ENOMEM); } m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (m == NULL) m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) return (-ENOMEM); mld_save_context(m, ifp); md = m_getptr(m, 0, &off); pmr = (struct mldv2_record *)(mtod(md, uint8_t *) + off); CTR1(KTR_MLD, "%s: allocated next packet", __func__); if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) { if (m != m0) m_freem(m); CTR1(KTR_MLD, "%s: m_append() failed.", __func__); return (-ENOMEM); } m->m_pkthdr.PH_vt.vt_nrecs = 1; nbytes += sizeof(struct mldv2_record); m0srcs = (ifp->if_mtu - MLD_MTUSPACE - sizeof(struct mldv2_record)) / sizeof(struct in6_addr); msrcs = 0; RB_FOREACH_FROM(ims, ip6_msource_tree, nims) { CTR2(KTR_MLD, "%s: visit node %s", __func__, ip6_sprintf(ip6tbuf, &ims->im6s_addr)); now = im6s_get_mode(inm, ims, 1); if ((now != mode) || (now == mode && (!use_block_allow && mode == MCAST_UNDEFINED))) { CTR1(KTR_MLD, "%s: skip node", __func__); continue; } if (is_source_query && ims->im6s_stp == 0) { CTR1(KTR_MLD, "%s: skip unrecorded node", __func__); continue; } CTR1(KTR_MLD, "%s: append node", __func__); if (!m_append(m, sizeof(struct in6_addr), (void *)&ims->im6s_addr)) { if (m != m0) m_freem(m); CTR1(KTR_MLD, "%s: m_append() failed.", __func__); return (-ENOMEM); } ++msrcs; if (msrcs == m0srcs) break; } pmr->mr_numsrc = htons(msrcs); nbytes += (msrcs * sizeof(struct in6_addr)); CTR1(KTR_MLD, "%s: enqueueing next packet", __func__); mbufq_enqueue(mq, m); } return (nbytes); } /* * Type used to mark record pass completion. * We exploit the fact we can cast to this easily from the * current filter modes on each ip_msource node. */ typedef enum { REC_NONE = 0x00, /* MCAST_UNDEFINED */ REC_ALLOW = 0x01, /* MCAST_INCLUDE */ REC_BLOCK = 0x02, /* MCAST_EXCLUDE */ REC_FULL = REC_ALLOW | REC_BLOCK } rectype_t; /* * Enqueue an MLDv2 filter list change to the given output queue. * * Source list filter state is held in an RB-tree. When the filter list * for a group is changed without changing its mode, we need to compute * the deltas between T0 and T1 for each source in the filter set, * and enqueue the appropriate ALLOW_NEW/BLOCK_OLD records. * * As we may potentially queue two record types, and the entire R-B tree * needs to be walked at once, we break this out into its own function * so we can generate a tightly packed queue of packets. * * XXX This could be written to only use one tree walk, although that makes * serializing into the mbuf chains a bit harder. For now we do two walks * which makes things easier on us, and it may or may not be harder on * the L2 cache. * * If successful the size of all data appended to the queue is returned, * otherwise an error code less than zero is returned, or zero if * no record(s) were appended. */ static int mld_v2_enqueue_filter_change(struct mbufq *mq, struct in6_multi *inm) { static const int MINRECLEN = sizeof(struct mldv2_record) + sizeof(struct in6_addr); struct ifnet *ifp; struct mldv2_record mr; struct mldv2_record *pmr; struct ip6_msource *ims, *nims; struct mbuf *m, *m0, *md; int m0srcs, nbytes, npbytes, off, rsrcs, schanged; int nallow, nblock; uint8_t mode, now, then; rectype_t crt, drt, nrt; #ifdef KTR char ip6tbuf[INET6_ADDRSTRLEN]; #endif IN6_MULTI_LOCK_ASSERT(); if (inm->in6m_nsrc == 0 || (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0)) return (0); ifp = inm->in6m_ifp; /* interface */ mode = inm->in6m_st[1].iss_fmode; /* filter mode at t1 */ crt = REC_NONE; /* current group record type */ drt = REC_NONE; /* mask of completed group record types */ nrt = REC_NONE; /* record type for current node */ m0srcs = 0; /* # source which will fit in current mbuf chain */ npbytes = 0; /* # of bytes appended this packet */ nbytes = 0; /* # of bytes appended to group's state-change queue */ rsrcs = 0; /* # sources encoded in current record */ schanged = 0; /* # nodes encoded in overall filter change */ nallow = 0; /* # of source entries in ALLOW_NEW */ nblock = 0; /* # of source entries in BLOCK_OLD */ nims = NULL; /* next tree node pointer */ /* * For each possible filter record mode. * The first kind of source we encounter tells us which * is the first kind of record we start appending. * If a node transitioned to UNDEFINED at t1, its mode is treated * as the inverse of the group's filter mode. */ while (drt != REC_FULL) { do { m0 = mbufq_last(mq); if (m0 != NULL && (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <= MLD_V2_REPORT_MAXRECS) && (m0->m_pkthdr.len + MINRECLEN) < (ifp->if_mtu - MLD_MTUSPACE)) { m = m0; m0srcs = (ifp->if_mtu - m0->m_pkthdr.len - sizeof(struct mldv2_record)) / sizeof(struct in6_addr); CTR1(KTR_MLD, "%s: use previous packet", __func__); } else { m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); if (m == NULL) m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) { CTR1(KTR_MLD, "%s: m_get*() failed", __func__); return (-ENOMEM); } m->m_pkthdr.PH_vt.vt_nrecs = 0; mld_save_context(m, ifp); m0srcs = (ifp->if_mtu - MLD_MTUSPACE - sizeof(struct mldv2_record)) / sizeof(struct in6_addr); npbytes = 0; CTR1(KTR_MLD, "%s: allocated new packet", __func__); } /* * Append the MLD group record header to the * current packet's data area. * Recalculate pointer to free space for next * group record, in case m_append() allocated * a new mbuf or cluster. */ memset(&mr, 0, sizeof(mr)); mr.mr_addr = inm->in6m_addr; in6_clearscope(&mr.mr_addr); if (!m_append(m, sizeof(mr), (void *)&mr)) { if (m != m0) m_freem(m); CTR1(KTR_MLD, "%s: m_append() failed", __func__); return (-ENOMEM); } npbytes += sizeof(struct mldv2_record); if (m != m0) { /* new packet; offset in chain */ md = m_getptr(m, npbytes - sizeof(struct mldv2_record), &off); pmr = (struct mldv2_record *)(mtod(md, uint8_t *) + off); } else { /* current packet; offset from last append */ md = m_last(m); pmr = (struct mldv2_record *)(mtod(md, uint8_t *) + md->m_len - sizeof(struct mldv2_record)); } /* * Begin walking the tree for this record type * pass, or continue from where we left off * previously if we had to allocate a new packet. * Only report deltas in-mode at t1. * We need not report included sources as allowed * if we are in inclusive mode on the group, * however the converse is not true. */ rsrcs = 0; if (nims == NULL) { nims = RB_MIN(ip6_msource_tree, &inm->in6m_srcs); } RB_FOREACH_FROM(ims, ip6_msource_tree, nims) { CTR2(KTR_MLD, "%s: visit node %s", __func__, ip6_sprintf(ip6tbuf, &ims->im6s_addr)); now = im6s_get_mode(inm, ims, 1); then = im6s_get_mode(inm, ims, 0); CTR3(KTR_MLD, "%s: mode: t0 %d, t1 %d", __func__, then, now); if (now == then) { CTR1(KTR_MLD, "%s: skip unchanged", __func__); continue; } if (mode == MCAST_EXCLUDE && now == MCAST_INCLUDE) { CTR1(KTR_MLD, "%s: skip IN src on EX group", __func__); continue; } nrt = (rectype_t)now; if (nrt == REC_NONE) nrt = (rectype_t)(~mode & REC_FULL); if (schanged++ == 0) { crt = nrt; } else if (crt != nrt) continue; if (!m_append(m, sizeof(struct in6_addr), (void *)&ims->im6s_addr)) { if (m != m0) m_freem(m); CTR1(KTR_MLD, "%s: m_append() failed", __func__); return (-ENOMEM); } nallow += !!(crt == REC_ALLOW); nblock += !!(crt == REC_BLOCK); if (++rsrcs == m0srcs) break; } /* * If we did not append any tree nodes on this * pass, back out of allocations. */ if (rsrcs == 0) { npbytes -= sizeof(struct mldv2_record); if (m != m0) { CTR1(KTR_MLD, "%s: m_free(m)", __func__); m_freem(m); } else { CTR1(KTR_MLD, "%s: m_adj(m, -mr)", __func__); m_adj(m, -((int)sizeof( struct mldv2_record))); } continue; } npbytes += (rsrcs * sizeof(struct in6_addr)); if (crt == REC_ALLOW) pmr->mr_type = MLD_ALLOW_NEW_SOURCES; else if (crt == REC_BLOCK) pmr->mr_type = MLD_BLOCK_OLD_SOURCES; pmr->mr_numsrc = htons(rsrcs); /* * Count the new group record, and enqueue this * packet if it wasn't already queued. */ m->m_pkthdr.PH_vt.vt_nrecs++; if (m != m0) mbufq_enqueue(mq, m); nbytes += npbytes; } while (nims != NULL); drt |= crt; crt = (~crt & REC_FULL); } CTR3(KTR_MLD, "%s: queued %d ALLOW_NEW, %d BLOCK_OLD", __func__, nallow, nblock); return (nbytes); } static int mld_v2_merge_state_changes(struct in6_multi *inm, struct mbufq *scq) { struct mbufq *gq; struct mbuf *m; /* pending state-change */ struct mbuf *m0; /* copy of pending state-change */ struct mbuf *mt; /* last state-change in packet */ int docopy, domerge; u_int recslen; docopy = 0; domerge = 0; recslen = 0; IN6_MULTI_LOCK_ASSERT(); MLD_LOCK_ASSERT(); /* * If there are further pending retransmissions, make a writable * copy of each queued state-change message before merging. */ if (inm->in6m_scrv > 0) docopy = 1; gq = &inm->in6m_scq; #ifdef KTR if (mbufq_first(gq) == NULL) { CTR2(KTR_MLD, "%s: WARNING: queue for inm %p is empty", __func__, inm); } #endif m = mbufq_first(gq); while (m != NULL) { /* * Only merge the report into the current packet if * there is sufficient space to do so; an MLDv2 report * packet may only contain 65,535 group records. * Always use a simple mbuf chain concatentation to do this, * as large state changes for single groups may have * allocated clusters. */ domerge = 0; mt = mbufq_last(scq); if (mt != NULL) { recslen = m_length(m, NULL); if ((mt->m_pkthdr.PH_vt.vt_nrecs + m->m_pkthdr.PH_vt.vt_nrecs <= MLD_V2_REPORT_MAXRECS) && (mt->m_pkthdr.len + recslen <= (inm->in6m_ifp->if_mtu - MLD_MTUSPACE))) domerge = 1; } if (!domerge && mbufq_full(gq)) { CTR2(KTR_MLD, "%s: outbound queue full, skipping whole packet %p", __func__, m); mt = m->m_nextpkt; if (!docopy) m_freem(m); m = mt; continue; } if (!docopy) { CTR2(KTR_MLD, "%s: dequeueing %p", __func__, m); m0 = mbufq_dequeue(gq); m = m0->m_nextpkt; } else { CTR2(KTR_MLD, "%s: copying %p", __func__, m); m0 = m_dup(m, M_NOWAIT); if (m0 == NULL) return (ENOMEM); m0->m_nextpkt = NULL; m = m->m_nextpkt; } if (!domerge) { CTR3(KTR_MLD, "%s: queueing %p to scq %p)", __func__, m0, scq); mbufq_enqueue(scq, m0); } else { struct mbuf *mtl; /* last mbuf of packet mt */ CTR3(KTR_MLD, "%s: merging %p with ifscq tail %p)", __func__, m0, mt); mtl = m_last(mt); m0->m_flags &= ~M_PKTHDR; mt->m_pkthdr.len += recslen; mt->m_pkthdr.PH_vt.vt_nrecs += m0->m_pkthdr.PH_vt.vt_nrecs; mtl->m_next = m0; } } return (0); } /* * Respond to a pending MLDv2 General Query. */ static void mld_v2_dispatch_general_query(struct mld_ifsoftc *mli) { struct ifmultiaddr *ifma; struct ifnet *ifp; struct in6_multi *inm; int retval; IN6_MULTI_LOCK_ASSERT(); MLD_LOCK_ASSERT(); KASSERT(mli->mli_version == MLD_VERSION_2, ("%s: called when version %d", __func__, mli->mli_version)); /* * Check that there are some packets queued. If so, send them first. * For large number of groups the reply to general query can take * many packets, we should finish sending them before starting of * queuing the new reply. */ if (mbufq_len(&mli->mli_gq) != 0) goto send; ifp = mli->mli_ifp; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_INET6 || ifma->ifma_protospec == NULL) continue; inm = (struct in6_multi *)ifma->ifma_protospec; KASSERT(ifp == inm->in6m_ifp, ("%s: inconsistent ifp", __func__)); switch (inm->in6m_state) { case MLD_NOT_MEMBER: case MLD_SILENT_MEMBER: break; case MLD_REPORTING_MEMBER: case MLD_IDLE_MEMBER: case MLD_LAZY_MEMBER: case MLD_SLEEPING_MEMBER: case MLD_AWAKENING_MEMBER: inm->in6m_state = MLD_REPORTING_MEMBER; retval = mld_v2_enqueue_group_record(&mli->mli_gq, inm, 0, 0, 0, 0); CTR2(KTR_MLD, "%s: enqueue record = %d", __func__, retval); break; case MLD_G_QUERY_PENDING_MEMBER: case MLD_SG_QUERY_PENDING_MEMBER: case MLD_LEAVING_MEMBER: break; } } IF_ADDR_RUNLOCK(ifp); send: mld_dispatch_queue(&mli->mli_gq, MLD_MAX_RESPONSE_BURST); /* * Slew transmission of bursts over 500ms intervals. */ if (mbufq_first(&mli->mli_gq) != NULL) { mli->mli_v2_timer = 1 + MLD_RANDOM_DELAY( MLD_RESPONSE_BURST_INTERVAL); V_interface_timers_running6 = 1; } } /* * Transmit the next pending message in the output queue. * * VIMAGE: Needs to store/restore vnet pointer on a per-mbuf-chain basis. * MRT: Nothing needs to be done, as MLD traffic is always local to * a link and uses a link-scope multicast address. */ static void mld_dispatch_packet(struct mbuf *m) { struct ip6_moptions im6o; struct ifnet *ifp; struct ifnet *oifp; struct mbuf *m0; struct mbuf *md; struct ip6_hdr *ip6; struct mld_hdr *mld; int error; int off; int type; uint32_t ifindex; CTR2(KTR_MLD, "%s: transmit %p", __func__, m); /* * Set VNET image pointer from enqueued mbuf chain * before doing anything else. Whilst we use interface * indexes to guard against interface detach, they are * unique to each VIMAGE and must be retrieved. */ ifindex = mld_restore_context(m); /* * Check if the ifnet still exists. This limits the scope of * any race in the absence of a global ifp lock for low cost * (an array lookup). */ ifp = ifnet_byindex(ifindex); if (ifp == NULL) { CTR3(KTR_MLD, "%s: dropped %p as ifindex %u went away.", __func__, m, ifindex); m_freem(m); IP6STAT_INC(ip6s_noroute); goto out; } im6o.im6o_multicast_hlim = 1; im6o.im6o_multicast_loop = (V_ip6_mrouter != NULL); im6o.im6o_multicast_ifp = ifp; if (m->m_flags & M_MLDV1) { m0 = m; } else { m0 = mld_v2_encap_report(ifp, m); if (m0 == NULL) { CTR2(KTR_MLD, "%s: dropped %p", __func__, m); IP6STAT_INC(ip6s_odropped); goto out; } } mld_scrub_context(m0); m_clrprotoflags(m); m0->m_pkthdr.rcvif = V_loif; ip6 = mtod(m0, struct ip6_hdr *); #if 0 (void)in6_setscope(&ip6->ip6_dst, ifp, NULL); /* XXX LOR */ #else /* * XXX XXX Break some KPI rules to prevent an LOR which would * occur if we called in6_setscope() at transmission. * See comments at top of file. */ MLD_EMBEDSCOPE(&ip6->ip6_dst, ifp->if_index); #endif /* * Retrieve the ICMPv6 type before handoff to ip6_output(), * so we can bump the stats. */ md = m_getptr(m0, sizeof(struct ip6_hdr), &off); mld = (struct mld_hdr *)(mtod(md, uint8_t *) + off); type = mld->mld_type; error = ip6_output(m0, &mld_po, NULL, IPV6_UNSPECSRC, &im6o, &oifp, NULL); if (error) { CTR3(KTR_MLD, "%s: ip6_output(%p) = %d", __func__, m0, error); goto out; } ICMP6STAT_INC(icp6s_outhist[type]); if (oifp != NULL) { icmp6_ifstat_inc(oifp, ifs6_out_msg); switch (type) { case MLD_LISTENER_REPORT: case MLDV2_LISTENER_REPORT: icmp6_ifstat_inc(oifp, ifs6_out_mldreport); break; case MLD_LISTENER_DONE: icmp6_ifstat_inc(oifp, ifs6_out_mlddone); break; } } out: return; } /* * Encapsulate an MLDv2 report. * * KAME IPv6 requires that hop-by-hop options be passed separately, * and that the IPv6 header be prepended in a separate mbuf. * * Returns a pointer to the new mbuf chain head, or NULL if the * allocation failed. */ static struct mbuf * mld_v2_encap_report(struct ifnet *ifp, struct mbuf *m) { struct mbuf *mh; struct mldv2_report *mld; struct ip6_hdr *ip6; struct in6_ifaddr *ia; int mldreclen; KASSERT(ifp != NULL, ("%s: null ifp", __func__)); KASSERT((m->m_flags & M_PKTHDR), ("%s: mbuf chain %p is !M_PKTHDR", __func__, m)); /* * RFC3590: OK to send as :: or tentative during DAD. */ ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST); if (ia == NULL) CTR1(KTR_MLD, "%s: warning: ia is NULL", __func__); mh = m_gethdr(M_NOWAIT, MT_DATA); if (mh == NULL) { if (ia != NULL) ifa_free(&ia->ia_ifa); m_freem(m); return (NULL); } M_ALIGN(mh, sizeof(struct ip6_hdr) + sizeof(struct mldv2_report)); mldreclen = m_length(m, NULL); CTR2(KTR_MLD, "%s: mldreclen is %d", __func__, mldreclen); mh->m_len = sizeof(struct ip6_hdr) + sizeof(struct mldv2_report); mh->m_pkthdr.len = sizeof(struct ip6_hdr) + sizeof(struct mldv2_report) + mldreclen; ip6 = mtod(mh, struct ip6_hdr *); ip6->ip6_flow = 0; ip6->ip6_vfc &= ~IPV6_VERSION_MASK; ip6->ip6_vfc |= IPV6_VERSION; ip6->ip6_nxt = IPPROTO_ICMPV6; ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any; if (ia != NULL) ifa_free(&ia->ia_ifa); ip6->ip6_dst = in6addr_linklocal_allv2routers; /* scope ID will be set in netisr */ mld = (struct mldv2_report *)(ip6 + 1); mld->mld_type = MLDV2_LISTENER_REPORT; mld->mld_code = 0; mld->mld_cksum = 0; mld->mld_v2_reserved = 0; mld->mld_v2_numrecs = htons(m->m_pkthdr.PH_vt.vt_nrecs); m->m_pkthdr.PH_vt.vt_nrecs = 0; mh->m_next = m; mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), sizeof(struct mldv2_report) + mldreclen); return (mh); } #ifdef KTR static char * mld_rec_type_to_str(const int type) { switch (type) { case MLD_CHANGE_TO_EXCLUDE_MODE: return "TO_EX"; break; case MLD_CHANGE_TO_INCLUDE_MODE: return "TO_IN"; break; case MLD_MODE_IS_EXCLUDE: return "MODE_EX"; break; case MLD_MODE_IS_INCLUDE: return "MODE_IN"; break; case MLD_ALLOW_NEW_SOURCES: return "ALLOW_NEW"; break; case MLD_BLOCK_OLD_SOURCES: return "BLOCK_OLD"; break; default: break; } return "unknown"; } #endif static void mld_init(void *unused __unused) { CTR1(KTR_MLD, "%s: initializing", __func__); MLD_LOCK_INIT(); ip6_initpktopts(&mld_po); mld_po.ip6po_hlim = 1; mld_po.ip6po_hbh = &mld_ra.hbh; mld_po.ip6po_prefer_tempaddr = IP6PO_TEMPADDR_NOTPREFER; mld_po.ip6po_flags = IP6PO_DONTFRAG; } -SYSINIT(mld_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, mld_init, NULL); +SYSINIT(mld_init, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE, mld_init, NULL); static void mld_uninit(void *unused __unused) { CTR1(KTR_MLD, "%s: tearing down", __func__); MLD_LOCK_DESTROY(); } -SYSUNINIT(mld_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, mld_uninit, NULL); +SYSUNINIT(mld_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE, mld_uninit, NULL); static void vnet_mld_init(const void *unused __unused) { CTR1(KTR_MLD, "%s: initializing", __func__); LIST_INIT(&V_mli_head); } -VNET_SYSINIT(vnet_mld_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mld_init, +VNET_SYSINIT(vnet_mld_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mld_init, NULL); static void vnet_mld_uninit(const void *unused __unused) { CTR1(KTR_MLD, "%s: tearing down", __func__); KASSERT(LIST_EMPTY(&V_mli_head), ("%s: mli list not empty; ifnets not detached?", __func__)); } -VNET_SYSUNINIT(vnet_mld_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mld_uninit, +VNET_SYSUNINIT(vnet_mld_uninit, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mld_uninit, NULL); static int mld_modevent(module_t mod, int type, void *unused __unused) { switch (type) { case MOD_LOAD: case MOD_UNLOAD: break; default: return (EOPNOTSUPP); } return (0); } static moduledata_t mld_mod = { "mld", mld_modevent, 0 }; -DECLARE_MODULE(mld, mld_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); +DECLARE_MODULE(mld, mld_mod, SI_SUB_PROTO_MC, SI_ORDER_ANY); Index: projects/vnet/sys/netinet6/nd6.c =================================================================== --- projects/vnet/sys/netinet6/nd6.c (revision 294565) +++ projects/vnet/sys/netinet6/nd6.c (revision 294566) @@ -1,2569 +1,2573 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ #define SIN6(s) ((const struct sockaddr_in6 *)(s)) /* timer values */ VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for * local traffic */ VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage * collection timer */ /* preventing too many loops in ND option parsing */ static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper * layer hints */ static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved * ND entries */ #define V_nd6_maxndopt VNET(nd6_maxndopt) #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) #ifdef ND6_DEBUG VNET_DEFINE(int, nd6_debug) = 1; #else VNET_DEFINE(int, nd6_debug) = 0; #endif static eventhandler_tag lle_event_eh; /* for debugging? */ #if 0 static int nd6_inuse, nd6_allocated; #endif VNET_DEFINE(struct nd_drhead, nd_defrouter); VNET_DEFINE(struct nd_prhead, nd_prefix); VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int); static int nd6_is_new_addr_neighbor(const struct sockaddr_in6 *, struct ifnet *); static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); static void nd6_slowtimo(void *); static int regen_tmpaddr(struct in6_ifaddr *); static void nd6_free(struct llentry *, int); static void nd6_free_redirect(const struct llentry *); static void nd6_llinfo_timer(void *); static void nd6_llinfo_settimer_locked(struct llentry *, long); static void clear_llinfo_pqueue(struct llentry *); static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *); static int nd6_resolve_slow(struct ifnet *, struct mbuf *, const struct sockaddr_in6 *, u_char *, uint32_t *); static int nd6_need_cache(struct ifnet *); static VNET_DEFINE(struct callout, nd6_slowtimo_ch); #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) VNET_DEFINE(struct callout, nd6_timer_ch); +#define V_nd6_timer_ch VNET(nd6_timer_ch) static void nd6_lle_event(void *arg __unused, struct llentry *lle, int evt) { struct rt_addrinfo rtinfo; struct sockaddr_in6 dst; struct sockaddr_dl gw; struct ifnet *ifp; int type; LLE_WLOCK_ASSERT(lle); if (lltable_get_af(lle->lle_tbl) != AF_INET6) return; switch (evt) { case LLENTRY_RESOLVED: type = RTM_ADD; KASSERT(lle->la_flags & LLE_VALID, ("%s: %p resolved but not valid?", __func__, lle)); break; case LLENTRY_EXPIRED: type = RTM_DELETE; break; default: return; } ifp = lltable_get_ifp(lle->lle_tbl); bzero(&dst, sizeof(dst)); bzero(&gw, sizeof(gw)); bzero(&rtinfo, sizeof(rtinfo)); lltable_fill_sa_entry(lle, (struct sockaddr *)&dst); dst.sin6_scope_id = in6_getscopezone(ifp, in6_addrscope(&dst.sin6_addr)); gw.sdl_len = sizeof(struct sockaddr_dl); gw.sdl_family = AF_LINK; gw.sdl_alen = ifp->if_addrlen; gw.sdl_index = ifp->if_index; gw.sdl_type = ifp->if_type; if (evt == LLENTRY_RESOLVED) bcopy(&lle->ll_addr, gw.sdl_data, ifp->if_addrlen); rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst; rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw; rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY; rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | ( type == RTM_ADD ? RTF_UP: 0), 0, RT_DEFAULT_FIB); } void nd6_init(void) { LIST_INIT(&V_nd_prefix); /* initialization of the default router list */ TAILQ_INIT(&V_nd_defrouter); - /* start timer */ + /* Start timers. */ callout_init(&V_nd6_slowtimo_ch, 0); callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, nd6_slowtimo, curvnet); + + callout_init(&V_nd6_timer_ch, 0); + callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); nd6_dad_init(); if (IS_DEFAULT_VNET(curvnet)) lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event, NULL, EVENTHANDLER_PRI_ANY); } #ifdef VIMAGE void nd6_destroy() { callout_drain(&V_nd6_slowtimo_ch); callout_drain(&V_nd6_timer_ch); if (IS_DEFAULT_VNET(curvnet)) EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh); } #endif struct nd_ifinfo * nd6_ifattach(struct ifnet *ifp) { struct nd_ifinfo *nd; nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO); nd->initialized = 1; nd->chlim = IPV6_DEFHLIM; nd->basereachable = REACHABLE_TIME; nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); nd->retrans = RETRANS_TIMER; nd->flags = ND6_IFF_PERFORMNUD; /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by * default regardless of the V_ip6_auto_linklocal configuration to * give a reasonable default behavior. */ if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) || (ifp->if_flags & IFF_LOOPBACK)) nd->flags |= ND6_IFF_AUTO_LINKLOCAL; /* * A loopback interface does not need to accept RTADV. * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by * default regardless of the V_ip6_accept_rtadv configuration to * prevent the interface from accepting RA messages arrived * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV. */ if (V_ip6_accept_rtadv && !(ifp->if_flags & IFF_LOOPBACK) && (ifp->if_type != IFT_BRIDGE)) nd->flags |= ND6_IFF_ACCEPT_RTADV; if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) nd->flags |= ND6_IFF_NO_RADR; /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ nd6_setmtu0(ifp, nd); return nd; } void nd6_ifdetach(struct nd_ifinfo *nd) { free(nd, M_IP6NDP); } /* * Reset ND level link MTU. This function is called when the physical MTU * changes, which means we might have to adjust the ND level MTU. */ void nd6_setmtu(struct ifnet *ifp) { if (ifp->if_afdata[AF_INET6] == NULL) return; nd6_setmtu0(ifp, ND_IFINFO(ifp)); } /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ void nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) { u_int32_t omaxmtu; omaxmtu = ndi->maxmtu; switch (ifp->if_type) { case IFT_ARCNET: ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ break; case IFT_FDDI: ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ break; case IFT_ISO88025: ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); break; default: ndi->maxmtu = ifp->if_mtu; break; } /* * Decreasing the interface MTU under IPV6 minimum MTU may cause * undesirable situation. We thus notify the operator of the change * explicitly. The check for omaxmtu is necessary to restrict the * log to the case of changing the MTU, not initializing it. */ if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { log(LOG_NOTICE, "nd6_setmtu0: " "new link MTU on %s (%lu) is too small for IPv6\n", if_name(ifp), (unsigned long)ndi->maxmtu); } if (ndi->maxmtu > V_in6_maxmtu) in6_setmaxmtu(); /* check all interfaces just in case */ } void nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) { bzero(ndopts, sizeof(*ndopts)); ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; ndopts->nd_opts_last = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); if (icmp6len == 0) { ndopts->nd_opts_done = 1; ndopts->nd_opts_search = NULL; } } /* * Take one ND option. */ struct nd_opt_hdr * nd6_option(union nd_opts *ndopts) { struct nd_opt_hdr *nd_opt; int olen; KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", __func__)); if (ndopts->nd_opts_search == NULL) return NULL; if (ndopts->nd_opts_done) return NULL; nd_opt = ndopts->nd_opts_search; /* make sure nd_opt_len is inside the buffer */ if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { bzero(ndopts, sizeof(*ndopts)); return NULL; } olen = nd_opt->nd_opt_len << 3; if (olen == 0) { /* * Message validation requires that all included * options have a length that is greater than zero. */ bzero(ndopts, sizeof(*ndopts)); return NULL; } ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); if (ndopts->nd_opts_search > ndopts->nd_opts_last) { /* option overruns the end of buffer, invalid */ bzero(ndopts, sizeof(*ndopts)); return NULL; } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { /* reached the end of options chain */ ndopts->nd_opts_done = 1; ndopts->nd_opts_search = NULL; } return nd_opt; } /* * Parse multiple ND options. * This function is much easier to use, for ND routines that do not need * multiple options of the same type. */ int nd6_options(union nd_opts *ndopts) { struct nd_opt_hdr *nd_opt; int i = 0; KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", __func__)); if (ndopts->nd_opts_search == NULL) return 0; while (1) { nd_opt = nd6_option(ndopts); if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { /* * Message validation requires that all included * options have a length that is greater than zero. */ ICMP6STAT_INC(icp6s_nd_badopt); bzero(ndopts, sizeof(*ndopts)); return -1; } if (nd_opt == NULL) goto skip1; switch (nd_opt->nd_opt_type) { case ND_OPT_SOURCE_LINKADDR: case ND_OPT_TARGET_LINKADDR: case ND_OPT_MTU: case ND_OPT_REDIRECTED_HEADER: case ND_OPT_NONCE: if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { nd6log((LOG_INFO, "duplicated ND6 option found (type=%d)\n", nd_opt->nd_opt_type)); /* XXX bark? */ } else { ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt; } break; case ND_OPT_PREFIX_INFORMATION: if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt; } ndopts->nd_opts_pi_end = (struct nd_opt_prefix_info *)nd_opt; break; /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ case ND_OPT_RDNSS: /* RFC 6106 */ case ND_OPT_DNSSL: /* RFC 6106 */ /* * Silently ignore options we know and do not care about * in the kernel. */ break; default: /* * Unknown options must be silently ignored, * to accomodate future extension to the protocol. */ nd6log((LOG_DEBUG, "nd6_options: unsupported option %d - " "option ignored\n", nd_opt->nd_opt_type)); } skip1: i++; if (i > V_nd6_maxndopt) { ICMP6STAT_INC(icp6s_nd_toomanyopt); nd6log((LOG_INFO, "too many loop in nd opt\n")); break; } if (ndopts->nd_opts_done) break; } return 0; } /* * ND6 timer routine to handle ND6 entries */ static void nd6_llinfo_settimer_locked(struct llentry *ln, long tick) { int canceled; LLE_WLOCK_ASSERT(ln); if (tick < 0) { ln->la_expire = 0; ln->ln_ntick = 0; canceled = callout_stop(&ln->lle_timer); } else { ln->la_expire = time_uptime + tick / hz; LLE_ADDREF(ln); if (tick > INT_MAX) { ln->ln_ntick = tick - INT_MAX; canceled = callout_reset(&ln->lle_timer, INT_MAX, nd6_llinfo_timer, ln); } else { ln->ln_ntick = 0; canceled = callout_reset(&ln->lle_timer, tick, nd6_llinfo_timer, ln); } } if (canceled > 0) LLE_REMREF(ln); } /* * Gets source address of the first packet in hold queue * and stores it in @src. * Returns pointer to @src (if hold queue is not empty) or NULL. * * Set noinline to be dtrace-friendly */ static __noinline struct in6_addr * nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src) { struct ip6_hdr hdr; struct mbuf *m; if (ln->la_hold == NULL) return (NULL); /* * assume every packet in la_hold has the same IP header */ m = ln->la_hold; if (sizeof(hdr) > m->m_len) return (NULL); m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr); *src = hdr.ip6_src; return (src); } /* * Checks if we need to switch from STALE state. * * RFC 4861 requires switching from STALE to DELAY state * on first packet matching entry, waiting V_nd6_delay and * transition to PROBE state (if upper layer confirmation was * not received). * * This code performs a bit differently: * On packet hit we don't change state (but desired state * can be guessed by control plane). However, after V_nd6_delay * seconds code will transition to PROBE state (so DELAY state * is kinda skipped in most situations). * * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so * we perform the following upon entering STALE state: * * 1) Arm timer to run each V_nd6_delay seconds to make sure that * if packet was transmitted at the start of given interval, we * would be able to switch to PROBE state in V_nd6_delay seconds * as user expects. * * 2) Reschedule timer until original V_nd6_gctimer expires keeping * lle in STALE state (remaining timer value stored in lle_remtime). * * 3) Reschedule timer if packet was transmitted less that V_nd6_delay * seconds ago. * * Returns non-zero value if the entry is still STALE (storing * the next timer interval in @pdelay). * * Returns zero value if original timer expired or we need to switch to * PROBE (store that in @do_switch variable). */ static int nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch) { int nd_delay, nd_gctimer, r_skip_req; time_t lle_hittime; long delay; *do_switch = 0; nd_gctimer = V_nd6_gctimer; nd_delay = V_nd6_delay; LLE_REQ_LOCK(lle); r_skip_req = lle->r_skip_req; lle_hittime = lle->lle_hittime; LLE_REQ_UNLOCK(lle); if (r_skip_req > 0) { /* * Nonzero r_skip_req value was set upon entering * STALE state. Since value was not changed, no * packets were passed using this lle. Ask for * timer reschedule and keep STALE state. */ delay = (long)(MIN(nd_gctimer, nd_delay)); delay *= hz; if (lle->lle_remtime > delay) lle->lle_remtime -= delay; else { delay = lle->lle_remtime; lle->lle_remtime = 0; } if (delay == 0) { /* * The original ng6_gctime timeout ended, * no more rescheduling. */ return (0); } *pdelay = delay; return (1); } /* * Packet received. Verify timestamp */ delay = (long)(time_uptime - lle_hittime); if (delay < nd_delay) { /* * V_nd6_delay still not passed since the first * hit in STALE state. * Reshedule timer and return. */ *pdelay = (long)(nd_delay - delay) * hz; return (1); } /* Request switching to probe */ *do_switch = 1; return (0); } /* * Switch @lle state to new state optionally arming timers. * * Set noinline to be dtrace-friendly */ __noinline void nd6_llinfo_setstate(struct llentry *lle, int newstate) { struct ifnet *ifp; int nd_gctimer, nd_delay; long delay, remtime; delay = 0; remtime = 0; switch (newstate) { case ND6_LLINFO_INCOMPLETE: ifp = lle->lle_tbl->llt_ifp; delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000; break; case ND6_LLINFO_REACHABLE: if (!ND6_LLINFO_PERMANENT(lle)) { ifp = lle->lle_tbl->llt_ifp; delay = (long)ND_IFINFO(ifp)->reachable * hz; } break; case ND6_LLINFO_STALE: /* * Notify fast path that we want to know if any packet * is transmitted by setting r_skip_req. */ LLE_REQ_LOCK(lle); lle->r_skip_req = 1; LLE_REQ_UNLOCK(lle); nd_delay = V_nd6_delay; nd_gctimer = V_nd6_gctimer; delay = (long)(MIN(nd_gctimer, nd_delay)) * hz; remtime = (long)nd_gctimer * hz - delay; break; case ND6_LLINFO_DELAY: lle->la_asked = 0; delay = (long)V_nd6_delay * hz; break; } if (delay > 0) nd6_llinfo_settimer_locked(lle, delay); lle->lle_remtime = remtime; lle->ln_state = newstate; } /* * Timer-dependent part of nd state machine. * * Set noinline to be dtrace-friendly */ static __noinline void nd6_llinfo_timer(void *arg) { struct llentry *ln; struct in6_addr *dst, *pdst, *psrc, src; struct ifnet *ifp; struct nd_ifinfo *ndi = NULL; int do_switch, send_ns; long delay; KASSERT(arg != NULL, ("%s: arg NULL", __func__)); ln = (struct llentry *)arg; LLE_WLOCK(ln); if (callout_pending(&ln->lle_timer)) { /* * Here we are a bit odd here in the treatment of * active/pending. If the pending bit is set, it got * rescheduled before I ran. The active * bit we ignore, since if it was stopped * in ll_tablefree() and was currently running * it would have return 0 so the code would * not have deleted it since the callout could * not be stopped so we want to go through * with the delete here now. If the callout * was restarted, the pending bit will be back on and * we just want to bail since the callout_reset would * return 1 and our reference would have been removed * by nd6_llinfo_settimer_locked above since canceled * would have been 1. */ LLE_WUNLOCK(ln); return; } ifp = ln->lle_tbl->llt_ifp; CURVNET_SET(ifp->if_vnet); ndi = ND_IFINFO(ifp); send_ns = 0; dst = &ln->r_l3addr.addr6; pdst = dst; if (ln->ln_ntick > 0) { if (ln->ln_ntick > INT_MAX) { ln->ln_ntick -= INT_MAX; nd6_llinfo_settimer_locked(ln, INT_MAX); } else { ln->ln_ntick = 0; nd6_llinfo_settimer_locked(ln, ln->ln_ntick); } goto done; } if (ln->la_flags & LLE_STATIC) { goto done; } if (ln->la_flags & LLE_DELETED) { nd6_free(ln, 0); ln = NULL; goto done; } switch (ln->ln_state) { case ND6_LLINFO_INCOMPLETE: if (ln->la_asked < V_nd6_mmaxtries) { ln->la_asked++; send_ns = 1; /* Send NS to multicast address */ pdst = NULL; } else { struct mbuf *m = ln->la_hold; if (m) { struct mbuf *m0; /* * assuming every packet in la_hold has the * same IP header. Send error after unlock. */ m0 = m->m_nextpkt; m->m_nextpkt = NULL; ln->la_hold = m0; clear_llinfo_pqueue(ln); } EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_TIMEDOUT); nd6_free(ln, 0); ln = NULL; if (m != NULL) icmp6_error2(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, 0, ifp); } break; case ND6_LLINFO_REACHABLE: if (!ND6_LLINFO_PERMANENT(ln)) nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); break; case ND6_LLINFO_STALE: if (nd6_is_stale(ln, &delay, &do_switch) != 0) { /* * No packet has used this entry and GC timeout * has not been passed. Reshedule timer and * return. */ nd6_llinfo_settimer_locked(ln, delay); break; } if (do_switch == 0) { /* * GC timer has ended and entry hasn't been used. * Run Garbage collector (RFC 4861, 5.3) */ if (!ND6_LLINFO_PERMANENT(ln)) { EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); nd6_free(ln, 1); ln = NULL; } break; } /* Entry has been used AND delay timer has ended. */ /* FALLTHROUGH */ case ND6_LLINFO_DELAY: if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { /* We need NUD */ ln->la_asked = 1; nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE); send_ns = 1; } else nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */ break; case ND6_LLINFO_PROBE: if (ln->la_asked < V_nd6_umaxtries) { ln->la_asked++; send_ns = 1; } else { EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); nd6_free(ln, 0); ln = NULL; } break; default: panic("%s: paths in a dark night can be confusing: %d", __func__, ln->ln_state); } done: if (send_ns != 0) { nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); psrc = nd6_llinfo_get_holdsrc(ln, &src); LLE_FREE_LOCKED(ln); ln = NULL; nd6_ns_output(ifp, psrc, pdst, dst, NULL); } if (ln != NULL) LLE_FREE_LOCKED(ln); CURVNET_RESTORE(); } /* * ND6 timer routine to expire default route list and prefix list */ void nd6_timer(void *arg) { CURVNET_SET((struct vnet *) arg); struct nd_defrouter *dr, *ndr; struct nd_prefix *pr, *npr; struct in6_ifaddr *ia6, *nia6; callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, nd6_timer, curvnet); /* expire default router list */ TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { if (dr->expire && dr->expire < time_uptime) defrtrlist_del(dr); } /* * expire interface addresses. * in the past the loop was inside prefix expiry processing. * However, from a stricter speci-confrmance standpoint, we should * rather separate address lifetimes and prefix lifetimes. * * XXXRW: in6_ifaddrhead locking. */ addrloop: TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { /* check address lifetime */ if (IFA6_IS_INVALID(ia6)) { int regen = 0; /* * If the expiring address is temporary, try * regenerating a new one. This would be useful when * we suspended a laptop PC, then turned it on after a * period that could invalidate all temporary * addresses. Although we may have to restart the * loop (see below), it must be after purging the * address. Otherwise, we'd see an infinite loop of * regeneration. */ if (V_ip6_use_tempaddr && (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { if (regen_tmpaddr(ia6) == 0) regen = 1; } in6_purgeaddr(&ia6->ia_ifa); if (regen) goto addrloop; /* XXX: see below */ } else if (IFA6_IS_DEPRECATED(ia6)) { int oldflags = ia6->ia6_flags; ia6->ia6_flags |= IN6_IFF_DEPRECATED; /* * If a temporary address has just become deprecated, * regenerate a new one if possible. */ if (V_ip6_use_tempaddr && (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && (oldflags & IN6_IFF_DEPRECATED) == 0) { if (regen_tmpaddr(ia6) == 0) { /* * A new temporary address is * generated. * XXX: this means the address chain * has changed while we are still in * the loop. Although the change * would not cause disaster (because * it's not a deletion, but an * addition,) we'd rather restart the * loop just for safety. Or does this * significantly reduce performance?? */ goto addrloop; } } } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) { /* * Schedule DAD for a tentative address. This happens * if the interface was down or not running * when the address was configured. */ int delay; delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz); nd6_dad_start((struct ifaddr *)ia6, delay); } else { /* * Check status of the interface. If it is down, * mark the address as tentative for future DAD. */ if ((ia6->ia_ifp->if_flags & IFF_UP) == 0 || (ia6->ia_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || (ND_IFINFO(ia6->ia_ifp)->flags & ND6_IFF_IFDISABLED) != 0) { ia6->ia6_flags &= ~IN6_IFF_DUPLICATED; ia6->ia6_flags |= IN6_IFF_TENTATIVE; } /* * A new RA might have made a deprecated address * preferred. */ ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; } } /* expire prefix list */ LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { /* * check prefix lifetime. * since pltime is just for autoconf, pltime processing for * prefix is not necessary. */ if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) { /* * address expiration and prefix expiration are * separate. NEVER perform in6_purgeaddr here. */ prelist_remove(pr); } } CURVNET_RESTORE(); } /* * ia6 - deprecated/invalidated temporary address */ static int regen_tmpaddr(struct in6_ifaddr *ia6) { struct ifaddr *ifa; struct ifnet *ifp; struct in6_ifaddr *public_ifa6 = NULL; ifp = ia6->ia_ifa.ifa_ifp; IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { struct in6_ifaddr *it6; if (ifa->ifa_addr->sa_family != AF_INET6) continue; it6 = (struct in6_ifaddr *)ifa; /* ignore no autoconf addresses. */ if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) continue; /* ignore autoconf addresses with different prefixes. */ if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) continue; /* * Now we are looking at an autoconf address with the same * prefix as ours. If the address is temporary and is still * preferred, do not create another one. It would be rare, but * could happen, for example, when we resume a laptop PC after * a long period. */ if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && !IFA6_IS_DEPRECATED(it6)) { public_ifa6 = NULL; break; } /* * This is a public autoconf address that has the same prefix * as ours. If it is preferred, keep it. We can't break the * loop here, because there may be a still-preferred temporary * address with the prefix. */ if (!IFA6_IS_DEPRECATED(it6)) public_ifa6 = it6; } if (public_ifa6 != NULL) ifa_ref(&public_ifa6->ia_ifa); IF_ADDR_RUNLOCK(ifp); if (public_ifa6 != NULL) { int e; if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { ifa_free(&public_ifa6->ia_ifa); log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" " tmp addr,errno=%d\n", e); return (-1); } ifa_free(&public_ifa6->ia_ifa); return (0); } return (-1); } /* * Nuke neighbor cache/prefix/default router management table, right before * ifp goes away. */ void nd6_purge(struct ifnet *ifp) { struct nd_defrouter *dr, *ndr; struct nd_prefix *pr, *npr; /* * Nuke default router list entries toward ifp. * We defer removal of default router list entries that is installed * in the routing table, in order to keep additional side effects as * small as possible. */ TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { if (dr->installed) continue; if (dr->ifp == ifp) defrtrlist_del(dr); } TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { if (!dr->installed) continue; if (dr->ifp == ifp) defrtrlist_del(dr); } /* Nuke prefix list entries toward ifp */ LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { if (pr->ndpr_ifp == ifp) { /* * Because if_detach() does *not* release prefixes * while purging addresses the reference count will * still be above zero. We therefore reset it to * make sure that the prefix really gets purged. */ pr->ndpr_refcnt = 0; /* * Previously, pr->ndpr_addr is removed as well, * but I strongly believe we don't have to do it. * nd6_purge() is only called from in6_ifdetach(), * which removes all the associated interface addresses * by itself. * (jinmei@kame.net 20010129) */ prelist_remove(pr); } } /* cancel default outgoing interface setting */ if (V_nd6_defifindex == ifp->if_index) nd6_setdefaultiface(0); if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { /* Refresh default router list. */ defrouter_select(); } /* XXXXX * We do not nuke the neighbor cache entries here any more * because the neighbor cache is kept in if_afdata[AF_INET6]. * nd6_purge() is invoked by in6_ifdetach() which is called * from if_detach() where everything gets purged. So let * in6_domifdetach() do the actual L2 table purging work. */ } /* * the caller acquires and releases the lock on the lltbls * Returns the llentry locked */ struct llentry * nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp) { struct sockaddr_in6 sin6; struct llentry *ln; bzero(&sin6, sizeof(sin6)); sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_family = AF_INET6; sin6.sin6_addr = *addr6; IF_AFDATA_LOCK_ASSERT(ifp); ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)&sin6); return (ln); } struct llentry * nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp) { struct sockaddr_in6 sin6; struct llentry *ln; bzero(&sin6, sizeof(sin6)); sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_family = AF_INET6; sin6.sin6_addr = *addr6; ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6); if (ln != NULL) ln->ln_state = ND6_LLINFO_NOSTATE; return (ln); } /* * Test whether a given IPv6 address is a neighbor or not, ignoring * the actual neighbor cache. The neighbor cache is ignored in order * to not reenter the routing code from within itself. */ static int nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) { struct nd_prefix *pr; struct ifaddr *dstaddr; /* * A link-local address is always a neighbor. * XXX: a link does not necessarily specify a single interface. */ if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { struct sockaddr_in6 sin6_copy; u_int32_t zone; /* * We need sin6_copy since sa6_recoverscope() may modify the * content (XXX). */ sin6_copy = *addr; if (sa6_recoverscope(&sin6_copy)) return (0); /* XXX: should be impossible */ if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) return (0); if (sin6_copy.sin6_scope_id == zone) return (1); else return (0); } /* * If the address matches one of our addresses, * it should be a neighbor. * If the address matches one of our on-link prefixes, it should be a * neighbor. */ LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { if (pr->ndpr_ifp != ifp) continue; if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { struct rtentry *rt; /* Always use the default FIB here. */ rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 0, 0, RT_DEFAULT_FIB); if (rt == NULL) continue; /* * This is the case where multiple interfaces * have the same prefix, but only one is installed * into the routing table and that prefix entry * is not the one being examined here. In the case * where RADIX_MPATH is enabled, multiple route * entries (of the same rt_key value) will be * installed because the interface addresses all * differ. */ if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { RTFREE_LOCKED(rt); continue; } RTFREE_LOCKED(rt); } if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, &addr->sin6_addr, &pr->ndpr_mask)) return (1); } /* * If the address is assigned on the node of the other side of * a p2p interface, the address should be a neighbor. */ dstaddr = ifa_ifwithdstaddr((const struct sockaddr *)addr, RT_ALL_FIBS); if (dstaddr != NULL) { if (dstaddr->ifa_ifp == ifp) { ifa_free(dstaddr); return (1); } ifa_free(dstaddr); } /* * If the default router list is empty, all addresses are regarded * as on-link, and thus, as a neighbor. */ if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && TAILQ_EMPTY(&V_nd_defrouter) && V_nd6_defifindex == ifp->if_index) { return (1); } return (0); } /* * Detect if a given IPv6 address identifies a neighbor on a given link. * XXX: should take care of the destination of a p2p link? */ int nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) { struct llentry *lle; int rc = 0; IF_AFDATA_UNLOCK_ASSERT(ifp); if (nd6_is_new_addr_neighbor(addr, ifp)) return (1); /* * Even if the address matches none of our addresses, it might be * in the neighbor cache. */ IF_AFDATA_RLOCK(ifp); if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { LLE_RUNLOCK(lle); rc = 1; } IF_AFDATA_RUNLOCK(ifp); return (rc); } /* * Free an nd6 llinfo entry. * Since the function would cause significant changes in the kernel, DO NOT * make it global, unless you have a strong reason for the change, and are sure * that the change is safe. * * Set noinline to be dtrace-friendly */ static __noinline void nd6_free(struct llentry *ln, int gc) { struct nd_defrouter *dr; struct ifnet *ifp; LLE_WLOCK_ASSERT(ln); /* * we used to have pfctlinput(PRC_HOSTDEAD) here. * even though it is not harmful, it was not really necessary. */ /* cancel timer */ nd6_llinfo_settimer_locked(ln, -1); ifp = ln->lle_tbl->llt_ifp; if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { dr = defrouter_lookup(&ln->r_l3addr.addr6, ifp); if (dr != NULL && dr->expire && ln->ln_state == ND6_LLINFO_STALE && gc) { /* * If the reason for the deletion is just garbage * collection, and the neighbor is an active default * router, do not delete it. Instead, reset the GC * timer using the router's lifetime. * Simply deleting the entry would affect default * router selection, which is not necessarily a good * thing, especially when we're using router preference * values. * XXX: the check for ln_state would be redundant, * but we intentionally keep it just in case. */ if (dr->expire > time_uptime) nd6_llinfo_settimer_locked(ln, (dr->expire - time_uptime) * hz); else nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); LLE_REMREF(ln); LLE_WUNLOCK(ln); return; } if (dr) { /* * Unreachablity of a router might affect the default * router selection and on-link detection of advertised * prefixes. */ /* * Temporarily fake the state to choose a new default * router and to perform on-link determination of * prefixes correctly. * Below the state will be set correctly, * or the entry itself will be deleted. */ ln->ln_state = ND6_LLINFO_INCOMPLETE; } if (ln->ln_router || dr) { /* * We need to unlock to avoid a LOR with rt6_flush() with the * rnh and for the calls to pfxlist_onlink_check() and * defrouter_select() in the block further down for calls * into nd6_lookup(). We still hold a ref. */ LLE_WUNLOCK(ln); /* * rt6_flush must be called whether or not the neighbor * is in the Default Router List. * See a corresponding comment in nd6_na_input(). */ rt6_flush(&ln->r_l3addr.addr6, ifp); } if (dr) { /* * Since defrouter_select() does not affect the * on-link determination and MIP6 needs the check * before the default router selection, we perform * the check now. */ pfxlist_onlink_check(); /* * Refresh default router list. */ defrouter_select(); } /* * If this entry was added by an on-link redirect, remove the * corresponding host route. */ if (ln->la_flags & LLE_REDIRECT) nd6_free_redirect(ln); if (ln->ln_router || dr) LLE_WLOCK(ln); } /* * Save to unlock. We still hold an extra reference and will not * free(9) in llentry_free() if someone else holds one as well. */ LLE_WUNLOCK(ln); IF_AFDATA_LOCK(ifp); LLE_WLOCK(ln); /* Guard against race with other llentry_free(). */ if (ln->la_flags & LLE_LINKED) { /* Remove callout reference */ LLE_REMREF(ln); lltable_unlink_entry(ln->lle_tbl, ln); } IF_AFDATA_UNLOCK(ifp); llentry_free(ln); } static int nd6_isdynrte(const struct rtentry *rt, void *xap) { if (rt->rt_flags == (RTF_UP | RTF_HOST | RTF_DYNAMIC)) return (1); return (0); } /* * Remove the rtentry for the given llentry, * both of which were installed by a redirect. */ static void nd6_free_redirect(const struct llentry *ln) { int fibnum; struct sockaddr_in6 sin6; struct rt_addrinfo info; lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6); memset(&info, 0, sizeof(info)); info.rti_info[RTAX_DST] = (struct sockaddr *)&sin6; info.rti_filter = nd6_isdynrte; for (fibnum = 0; fibnum < rt_numfibs; fibnum++) rtrequest1_fib(RTM_DELETE, &info, NULL, fibnum); } /* * Rejuvenate this function for routing operations related * processing. */ void nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) { struct sockaddr_in6 *gateway; struct nd_defrouter *dr; struct ifnet *ifp; gateway = (struct sockaddr_in6 *)rt->rt_gateway; ifp = rt->rt_ifp; switch (req) { case RTM_ADD: break; case RTM_DELETE: if (!ifp) return; /* * Only indirect routes are interesting. */ if ((rt->rt_flags & RTF_GATEWAY) == 0) return; /* * check for default route */ if (IN6_ARE_ADDR_EQUAL(&in6addr_any, &SIN6(rt_key(rt))->sin6_addr)) { dr = defrouter_lookup(&gateway->sin6_addr, ifp); if (dr != NULL) dr->installed = 0; } break; } } int nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) { struct in6_ndireq *ndi = (struct in6_ndireq *)data; struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; int error = 0; if (ifp->if_afdata[AF_INET6] == NULL) return (EPFNOSUPPORT); switch (cmd) { case OSIOCGIFINFO_IN6: #define ND ndi->ndi /* XXX: old ndp(8) assumes a positive value for linkmtu. */ bzero(&ND, sizeof(ND)); ND.linkmtu = IN6_LINKMTU(ifp); ND.maxmtu = ND_IFINFO(ifp)->maxmtu; ND.basereachable = ND_IFINFO(ifp)->basereachable; ND.reachable = ND_IFINFO(ifp)->reachable; ND.retrans = ND_IFINFO(ifp)->retrans; ND.flags = ND_IFINFO(ifp)->flags; ND.recalctm = ND_IFINFO(ifp)->recalctm; ND.chlim = ND_IFINFO(ifp)->chlim; break; case SIOCGIFINFO_IN6: ND = *ND_IFINFO(ifp); break; case SIOCSIFINFO_IN6: /* * used to change host variables from userland. * intented for a use on router to reflect RA configurations. */ /* 0 means 'unspecified' */ if (ND.linkmtu != 0) { if (ND.linkmtu < IPV6_MMTU || ND.linkmtu > IN6_LINKMTU(ifp)) { error = EINVAL; break; } ND_IFINFO(ifp)->linkmtu = ND.linkmtu; } if (ND.basereachable != 0) { int obasereachable = ND_IFINFO(ifp)->basereachable; ND_IFINFO(ifp)->basereachable = ND.basereachable; if (ND.basereachable != obasereachable) ND_IFINFO(ifp)->reachable = ND_COMPUTE_RTIME(ND.basereachable); } if (ND.retrans != 0) ND_IFINFO(ifp)->retrans = ND.retrans; if (ND.chlim != 0) ND_IFINFO(ifp)->chlim = ND.chlim; /* FALLTHROUGH */ case SIOCSIFINFO_FLAGS: { struct ifaddr *ifa; struct in6_ifaddr *ia; if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && !(ND.flags & ND6_IFF_IFDISABLED)) { /* ifdisabled 1->0 transision */ /* * If the interface is marked as ND6_IFF_IFDISABLED and * has an link-local address with IN6_IFF_DUPLICATED, * do not clear ND6_IFF_IFDISABLED. * See RFC 4862, Section 5.4.5. */ IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; ia = (struct in6_ifaddr *)ifa; if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) break; } IF_ADDR_RUNLOCK(ifp); if (ifa != NULL) { /* LLA is duplicated. */ ND.flags |= ND6_IFF_IFDISABLED; log(LOG_ERR, "Cannot enable an interface" " with a link-local address marked" " duplicate.\n"); } else { ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; if (ifp->if_flags & IFF_UP) in6_if_up(ifp); } } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && (ND.flags & ND6_IFF_IFDISABLED)) { /* ifdisabled 0->1 transision */ /* Mark all IPv6 address as tentative. */ ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; if (V_ip6_dad_count > 0 && (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) { IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; ia = (struct in6_ifaddr *)ifa; ia->ia6_flags |= IN6_IFF_TENTATIVE; } IF_ADDR_RUNLOCK(ifp); } } if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { /* auto_linklocal 0->1 transision */ /* If no link-local address on ifp, configure */ ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; in6_ifattach(ifp, NULL); } else if (!(ND.flags & ND6_IFF_IFDISABLED) && ifp->if_flags & IFF_UP) { /* * When the IF already has * ND6_IFF_AUTO_LINKLOCAL, no link-local * address is assigned, and IFF_UP, try to * assign one. */ IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family != AF_INET6) continue; ia = (struct in6_ifaddr *)ifa; if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) break; } IF_ADDR_RUNLOCK(ifp); if (ifa != NULL) /* No LLA is configured. */ in6_ifattach(ifp, NULL); } } } ND_IFINFO(ifp)->flags = ND.flags; break; #undef ND case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ /* sync kernel routing table with the default router list */ defrouter_reset(); defrouter_select(); break; case SIOCSPFXFLUSH_IN6: { /* flush all the prefix advertised by routers */ struct nd_prefix *pr, *next; LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { struct in6_ifaddr *ia, *ia_next; if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) continue; /* XXX */ /* do we really have to remove addresses as well? */ /* XXXRW: in6_ifaddrhead locking. */ TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, ia_next) { if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) continue; if (ia->ia6_ndpr == pr) in6_purgeaddr(&ia->ia_ifa); } prelist_remove(pr); } break; } case SIOCSRTRFLUSH_IN6: { /* flush all the default routers */ struct nd_defrouter *dr, *next; defrouter_reset(); TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) { defrtrlist_del(dr); } defrouter_select(); break; } case SIOCGNBRINFO_IN6: { struct llentry *ln; struct in6_addr nb_addr = nbi->addr; /* make local for safety */ if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) return (error); IF_AFDATA_RLOCK(ifp); ln = nd6_lookup(&nb_addr, 0, ifp); IF_AFDATA_RUNLOCK(ifp); if (ln == NULL) { error = EINVAL; break; } nbi->state = ln->ln_state; nbi->asked = ln->la_asked; nbi->isrouter = ln->ln_router; if (ln->la_expire == 0) nbi->expire = 0; else nbi->expire = ln->la_expire + ln->lle_remtime / hz + (time_second - time_uptime); LLE_RUNLOCK(ln); break; } case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ ndif->ifindex = V_nd6_defifindex; break; case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ return (nd6_setdefaultiface(ndif->ifindex)); } return (error); } /* * Calculates new isRouter value based on provided parameters and * returns it. */ static int nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr, int ln_router) { /* * ICMP6 type dependent behavior. * * NS: clear IsRouter if new entry * RS: clear IsRouter * RA: set IsRouter if there's lladdr * redir: clear IsRouter if new entry * * RA case, (1): * The spec says that we must set IsRouter in the following cases: * - If lladdr exist, set IsRouter. This means (1-5). * - If it is old entry (!newentry), set IsRouter. This means (7). * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. * A quetion arises for (1) case. (1) case has no lladdr in the * neighbor cache, this is similar to (6). * This case is rare but we figured that we MUST NOT set IsRouter. * * is_new old_addr new_addr NS RS RA redir * D R * 0 n n (1) c ? s * 0 y n (2) c s s * 0 n y (3) c s s * 0 y y (4) c s s * 0 y y (5) c s s * 1 -- n (6) c c c s * 1 -- y (7) c c s c s * * (c=clear s=set) */ switch (type & 0xff) { case ND_NEIGHBOR_SOLICIT: /* * New entry must have is_router flag cleared. */ if (is_new) /* (6-7) */ ln_router = 0; break; case ND_REDIRECT: /* * If the icmp is a redirect to a better router, always set the * is_router flag. Otherwise, if the entry is newly created, * clear the flag. [RFC 2461, sec 8.3] */ if (code == ND_REDIRECT_ROUTER) ln_router = 1; else { if (is_new) /* (6-7) */ ln_router = 0; } break; case ND_ROUTER_SOLICIT: /* * is_router flag must always be cleared. */ ln_router = 0; break; case ND_ROUTER_ADVERT: /* * Mark an entry with lladdr as a router. */ if ((!is_new && (old_addr || new_addr)) || /* (2-5) */ (is_new && new_addr)) { /* (7) */ ln_router = 1; } break; } return (ln_router); } /* * Create neighbor cache entry and cache link-layer address, * on reception of inbound ND6 packets. (RS/RA/NS/redirect) * * type - ICMP6 type * code - type dependent information * */ void nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, int lladdrlen, int type, int code) { struct llentry *ln = NULL, *ln_tmp; int is_newentry; int do_update; int olladdr; int llchange; int flags; uint16_t router = 0; struct sockaddr_in6 sin6; struct mbuf *chain = NULL; IF_AFDATA_UNLOCK_ASSERT(ifp); KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__)); KASSERT(from != NULL, ("%s: from == NULL", __func__)); /* nothing must be updated for unspecified address */ if (IN6_IS_ADDR_UNSPECIFIED(from)) return; /* * Validation about ifp->if_addrlen and lladdrlen must be done in * the caller. * * XXX If the link does not have link-layer adderss, what should * we do? (ifp->if_addrlen == 0) * Spec says nothing in sections for RA, RS and NA. There's small * description on it in NS section (RFC 2461 7.2.3). */ flags = lladdr ? LLE_EXCLUSIVE : 0; IF_AFDATA_RLOCK(ifp); ln = nd6_lookup(from, flags, ifp); IF_AFDATA_RUNLOCK(ifp); is_newentry = 0; if (ln == NULL) { flags |= LLE_EXCLUSIVE; ln = nd6_alloc(from, 0, ifp); if (ln == NULL) return; /* * Since we already know all the data for the new entry, * fill it before insertion. */ if (lladdr != NULL) lltable_set_entry_addr(ifp, ln, lladdr); IF_AFDATA_WLOCK(ifp); LLE_WLOCK(ln); /* Prefer any existing lle over newly-created one */ ln_tmp = nd6_lookup(from, LLE_EXCLUSIVE, ifp); if (ln_tmp == NULL) lltable_link_entry(LLTABLE6(ifp), ln); IF_AFDATA_WUNLOCK(ifp); if (ln_tmp == NULL) { /* No existing lle, mark as new entry (6,7) */ is_newentry = 1; nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); if (lladdr != NULL) /* (7) */ EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); } else { lltable_free_entry(LLTABLE6(ifp), ln); ln = ln_tmp; ln_tmp = NULL; } } /* do nothing if static ndp is set */ if ((ln->la_flags & LLE_STATIC)) { if (flags & LLE_EXCLUSIVE) LLE_WUNLOCK(ln); else LLE_RUNLOCK(ln); return; } olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; if (olladdr && lladdr) { llchange = bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen); } else if (!olladdr && lladdr) llchange = 1; else llchange = 0; /* * newentry olladdr lladdr llchange (*=record) * 0 n n -- (1) * 0 y n -- (2) * 0 n y y (3) * STALE * 0 y y n (4) * * 0 y y y (5) * STALE * 1 -- n -- (6) NOSTATE(= PASSIVE) * 1 -- y -- (7) * STALE */ do_update = 0; if (is_newentry == 0 && llchange != 0) { do_update = 1; /* (3,5) */ /* * Record source link-layer address * XXX is it dependent to ifp->if_type? */ if (lltable_try_set_entry_addr(ifp, ln, lladdr) == 0) { /* Entry was deleted */ return; } nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); if (ln->la_hold != NULL) nd6_grab_holdchain(ln, &chain, &sin6); } /* Calculates new router status */ router = nd6_is_router(type, code, is_newentry, olladdr, lladdr != NULL ? 1 : 0, ln->ln_router); ln->ln_router = router; /* Mark non-router redirects with special flag */ if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER) ln->la_flags |= LLE_REDIRECT; if (flags & LLE_EXCLUSIVE) LLE_WUNLOCK(ln); else LLE_RUNLOCK(ln); if (chain != NULL) nd6_flush_holdchain(ifp, ifp, chain, &sin6); /* * When the link-layer address of a router changes, select the * best router again. In particular, when the neighbor entry is newly * created, it might affect the selection policy. * Question: can we restrict the first condition to the "is_newentry" * case? * XXX: when we hear an RA from a new router with the link-layer * address option, defrouter_select() is called twice, since * defrtrlist_update called the function as well. However, I believe * we can compromise the overhead, since it only happens the first * time. * XXX: although defrouter_select() should not have a bad effect * for those are not autoconfigured hosts, we explicitly avoid such * cases for safety. */ if ((do_update || is_newentry) && router && ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { /* * guaranteed recursion */ defrouter_select(); } } static void nd6_slowtimo(void *arg) { CURVNET_SET((struct vnet *) arg); struct nd_ifinfo *nd6if; struct ifnet *ifp; callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, nd6_slowtimo, curvnet); IFNET_RLOCK_NOSLEEP(); TAILQ_FOREACH(ifp, &V_ifnet, if_link) { if (ifp->if_afdata[AF_INET6] == NULL) continue; nd6if = ND_IFINFO(ifp); if (nd6if->basereachable && /* already initialized */ (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { /* * Since reachable time rarely changes by router * advertisements, we SHOULD insure that a new random * value gets recomputed at least once every few hours. * (RFC 2461, 6.3.4) */ nd6if->recalctm = V_nd6_recalc_reachtm_interval; nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); } } IFNET_RUNLOCK_NOSLEEP(); CURVNET_RESTORE(); } void nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain, struct sockaddr_in6 *sin6) { LLE_WLOCK_ASSERT(ln); *chain = ln->la_hold; ln->la_hold = NULL; lltable_fill_sa_entry(ln, (struct sockaddr *)sin6); if (ln->ln_state == ND6_LLINFO_STALE) { /* * The first time we send a packet to a * neighbor whose entry is STALE, we have * to change the state to DELAY and a sets * a timer to expire in DELAY_FIRST_PROBE_TIME * seconds to ensure do neighbor unreachability * detection on expiration. * (RFC 2461 7.3.3) */ nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY); } } int nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, struct sockaddr_in6 *dst, struct route *ro) { int error; int ip6len; struct ip6_hdr *ip6; struct m_tag *mtag; #ifdef MAC mac_netinet6_nd6_send(ifp, m); #endif /* * If called from nd6_ns_output() (NS), nd6_na_output() (NA), * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA * as handled by rtsol and rtadvd), mbufs will be tagged for SeND * to be diverted to user space. When re-injected into the kernel, * send_output() will directly dispatch them to the outgoing interface. */ if (send_sendso_input_hook != NULL) { mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); if (mtag != NULL) { ip6 = mtod(m, struct ip6_hdr *); ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); /* Use the SEND socket */ error = send_sendso_input_hook(m, ifp, SND_OUT, ip6len); /* -1 == no app on SEND socket */ if (error == 0 || error != -1) return (error); } } m_clrprotoflags(m); /* Avoid confusing lower layers. */ IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL, mtod(m, struct ip6_hdr *)); if ((ifp->if_flags & IFF_LOOPBACK) == 0) origifp = ifp; error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro); return (error); } /* * Do L2 address resolution for @sa_dst address. Stores found * address in @desten buffer. Copy of lle ln_flags can be also * saved in @pflags if @pflags is non-NULL. * * If destination LLE does not exists or lle state modification * is required, call "slow" version. * * Return values: * - 0 on success (address copied to buffer). * - EWOULDBLOCK (no local error, but address is still unresolved) * - other errors (alloc failure, etc) */ int nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m, const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags) { struct llentry *ln = NULL; const struct sockaddr_in6 *dst6; if (pflags != NULL) *pflags = 0; dst6 = (const struct sockaddr_in6 *)sa_dst; /* discard the packet if IPv6 operation is disabled on the interface */ if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { m_freem(m); return (ENETDOWN); /* better error? */ } if (m != NULL && m->m_flags & M_MCAST) { switch (ifp->if_type) { case IFT_ETHER: case IFT_FDDI: case IFT_L2VLAN: case IFT_IEEE80211: case IFT_BRIDGE: case IFT_ISO88025: ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr, desten); return (0); default: m_freem(m); return (EAFNOSUPPORT); } } IF_AFDATA_RLOCK(ifp); ln = nd6_lookup(&dst6->sin6_addr, LLE_UNLOCKED, ifp); if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) { /* Entry found, let's copy lle info */ bcopy(&ln->ll_addr, desten, ifp->if_addrlen); if (pflags != NULL) *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR); /* Check if we have feedback request from nd6 timer */ if (ln->r_skip_req != 0) { LLE_REQ_LOCK(ln); ln->r_skip_req = 0; /* Notify that entry was used */ ln->lle_hittime = time_uptime; LLE_REQ_UNLOCK(ln); } IF_AFDATA_RUNLOCK(ifp); return (0); } IF_AFDATA_RUNLOCK(ifp); return (nd6_resolve_slow(ifp, m, dst6, desten, pflags)); } /* * Do L2 address resolution for @sa_dst address. Stores found * address in @desten buffer. Copy of lle ln_flags can be also * saved in @pflags if @pflags is non-NULL. * * Heavy version. * Function assume that destination LLE does not exist, * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired. * * Set noinline to be dtrace-friendly */ static __noinline int nd6_resolve_slow(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags) { struct llentry *lle = NULL, *lle_tmp; struct in6_addr *psrc, src; int send_ns; /* * Address resolution or Neighbor Unreachability Detection * for the next hop. * At this point, the destination of the packet must be a unicast * or an anycast address(i.e. not a multicast). */ if (lle == NULL) { IF_AFDATA_RLOCK(ifp); lle = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); IF_AFDATA_RUNLOCK(ifp); if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) { /* * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), * the condition below is not very efficient. But we believe * it is tolerable, because this should be a rare case. */ lle = nd6_alloc(&dst->sin6_addr, 0, ifp); if (lle == NULL) { char ip6buf[INET6_ADDRSTRLEN]; log(LOG_DEBUG, "nd6_output: can't allocate llinfo for %s " "(ln=%p)\n", ip6_sprintf(ip6buf, &dst->sin6_addr), lle); m_freem(m); return (ENOBUFS); } IF_AFDATA_WLOCK(ifp); LLE_WLOCK(lle); /* Prefer any existing entry over newly-created one */ lle_tmp = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); if (lle_tmp == NULL) lltable_link_entry(LLTABLE6(ifp), lle); IF_AFDATA_WUNLOCK(ifp); if (lle_tmp != NULL) { lltable_free_entry(LLTABLE6(ifp), lle); lle = lle_tmp; lle_tmp = NULL; } } } if (lle == NULL) { if (!(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { m_freem(m); return (ENOBUFS); } if (m != NULL) m_freem(m); return (ENOBUFS); } LLE_WLOCK_ASSERT(lle); /* * The first time we send a packet to a neighbor whose entry is * STALE, we have to change the state to DELAY and a sets a timer to * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do * neighbor unreachability detection on expiration. * (RFC 2461 7.3.3) */ if (lle->ln_state == ND6_LLINFO_STALE) nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY); /* * If the neighbor cache entry has a state other than INCOMPLETE * (i.e. its link-layer address is already resolved), just * send the packet. */ if (lle->ln_state > ND6_LLINFO_INCOMPLETE) { bcopy(&lle->ll_addr, desten, ifp->if_addrlen); if (pflags != NULL) *pflags = lle->la_flags; LLE_WUNLOCK(lle); return (0); } /* * There is a neighbor cache entry, but no ethernet address * response yet. Append this latest packet to the end of the * packet queue in the mbuf, unless the number of the packet * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, * the oldest packet in the queue will be removed. */ if (lle->la_hold != NULL) { struct mbuf *m_hold; int i; i = 0; for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){ i++; if (m_hold->m_nextpkt == NULL) { m_hold->m_nextpkt = m; break; } } while (i >= V_nd6_maxqueuelen) { m_hold = lle->la_hold; lle->la_hold = lle->la_hold->m_nextpkt; m_freem(m_hold); i--; } } else { lle->la_hold = m; } /* * If there has been no NS for the neighbor after entering the * INCOMPLETE state, send the first solicitation. * Note that for newly-created lle la_asked will be 0, * so we will transition from ND6_LLINFO_NOSTATE to * ND6_LLINFO_INCOMPLETE state here. */ psrc = NULL; send_ns = 0; if (lle->la_asked == 0) { lle->la_asked++; send_ns = 1; psrc = nd6_llinfo_get_holdsrc(lle, &src); nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE); } LLE_WUNLOCK(lle); if (send_ns != 0) nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL); return (EWOULDBLOCK); } int nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, struct sockaddr_in6 *dst) { struct mbuf *m, *m_head; struct ifnet *outifp; int error = 0; m_head = chain; if ((ifp->if_flags & IFF_LOOPBACK) != 0) outifp = origifp; else outifp = ifp; while (m_head) { m = m_head; m_head = m_head->m_nextpkt; error = nd6_output_ifp(ifp, origifp, m, dst, NULL); } /* * XXX * note that intermediate errors are blindly ignored */ return (error); } static int nd6_need_cache(struct ifnet *ifp) { /* * XXX: we currently do not make neighbor cache on any interface * other than ARCnet, Ethernet, FDDI and GIF. * * RFC2893 says: * - unidirectional tunnels needs no ND */ switch (ifp->if_type) { case IFT_ARCNET: case IFT_ETHER: case IFT_FDDI: case IFT_IEEE1394: case IFT_L2VLAN: case IFT_IEEE80211: case IFT_INFINIBAND: case IFT_BRIDGE: case IFT_PROPVIRTUAL: return (1); default: return (0); } } /* * Add pernament ND6 link-layer record for given * interface address. * * Very similar to IPv4 arp_ifinit(), but: * 1) IPv6 DAD is performed in different place * 2) It is called by IPv6 protocol stack in contrast to * arp_ifinit() which is typically called in SIOCSIFADDR * driver ioctl handler. * */ int nd6_add_ifa_lle(struct in6_ifaddr *ia) { struct ifnet *ifp; struct llentry *ln, *ln_tmp; struct sockaddr *dst; ifp = ia->ia_ifa.ifa_ifp; if (nd6_need_cache(ifp) == 0) return (0); ia->ia_ifa.ifa_rtrequest = nd6_rtrequest; dst = (struct sockaddr *)&ia->ia_addr; ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst); if (ln == NULL) return (ENOBUFS); IF_AFDATA_WLOCK(ifp); LLE_WLOCK(ln); /* Unlink any entry if exists */ ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst); if (ln_tmp != NULL) lltable_unlink_entry(LLTABLE6(ifp), ln_tmp); lltable_link_entry(LLTABLE6(ifp), ln); IF_AFDATA_WUNLOCK(ifp); if (ln_tmp != NULL) EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED); EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); LLE_WUNLOCK(ln); if (ln_tmp != NULL) llentry_free(ln_tmp); return (0); } /* * Removes either all lle entries for given @ia, or lle * corresponding to @ia address. */ void nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all) { struct sockaddr_in6 mask, addr; struct sockaddr *saddr, *smask; struct ifnet *ifp; ifp = ia->ia_ifa.ifa_ifp; memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr)); memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask)); saddr = (struct sockaddr *)&addr; smask = (struct sockaddr *)&mask; if (all != 0) lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC); else lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr); } static void clear_llinfo_pqueue(struct llentry *ln) { struct mbuf *m_hold, *m_hold_next; for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { m_hold_next = m_hold->m_nextpkt; m_freem(m_hold); } ln->la_hold = NULL; return; } static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); #ifdef SYSCTL_DECL SYSCTL_DECL(_net_inet6_icmp6); #endif SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, CTLFLAG_RD, nd6_sysctl_drlist, ""); SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, CTLFLAG_RD, nd6_sysctl_prlist, ""); SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), ""); static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) { struct in6_defrouter d; struct nd_defrouter *dr; int error; if (req->newptr) return (EPERM); bzero(&d, sizeof(d)); d.rtaddr.sin6_family = AF_INET6; d.rtaddr.sin6_len = sizeof(d.rtaddr); /* * XXX locking */ TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { d.rtaddr.sin6_addr = dr->rtaddr; error = sa6_recoverscope(&d.rtaddr); if (error != 0) return (error); d.flags = dr->flags; d.rtlifetime = dr->rtlifetime; d.expire = dr->expire + (time_second - time_uptime); d.if_index = dr->ifp->if_index; error = SYSCTL_OUT(req, &d, sizeof(d)); if (error != 0) return (error); } return (0); } static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) { struct in6_prefix p; struct sockaddr_in6 s6; struct nd_prefix *pr; struct nd_pfxrouter *pfr; time_t maxexpire; int error; char ip6buf[INET6_ADDRSTRLEN]; if (req->newptr) return (EPERM); bzero(&p, sizeof(p)); p.origin = PR_ORIG_RA; bzero(&s6, sizeof(s6)); s6.sin6_family = AF_INET6; s6.sin6_len = sizeof(s6); /* * XXX locking */ LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { p.prefix = pr->ndpr_prefix; if (sa6_recoverscope(&p.prefix)) { log(LOG_ERR, "scope error in prefix list (%s)\n", ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); /* XXX: press on... */ } p.raflags = pr->ndpr_raf; p.prefixlen = pr->ndpr_plen; p.vltime = pr->ndpr_vltime; p.pltime = pr->ndpr_pltime; p.if_index = pr->ndpr_ifp->if_index; if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) p.expire = 0; else { /* XXX: we assume time_t is signed. */ maxexpire = (-1) & ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) p.expire = pr->ndpr_lastupdate + pr->ndpr_vltime + (time_second - time_uptime); else p.expire = maxexpire; } p.refcnt = pr->ndpr_refcnt; p.flags = pr->ndpr_stateflags; p.advrtrs = 0; LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) p.advrtrs++; error = SYSCTL_OUT(req, &p, sizeof(p)); if (error != 0) return (error); LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { s6.sin6_addr = pfr->router->rtaddr; if (sa6_recoverscope(&s6)) log(LOG_ERR, "scope error in prefix list (%s)\n", ip6_sprintf(ip6buf, &pfr->router->rtaddr)); error = SYSCTL_OUT(req, &s6, sizeof(s6)); if (error != 0) return (error); } } return (0); } Index: projects/vnet/sys/netinet6/nd6.h =================================================================== --- projects/vnet/sys/netinet6/nd6.h (revision 294565) +++ projects/vnet/sys/netinet6/nd6.h (revision 294566) @@ -1,458 +1,455 @@ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $KAME: nd6.h,v 1.76 2001/12/18 02:10:31 itojun Exp $ * $FreeBSD$ */ #ifndef _NETINET6_ND6_H_ #define _NETINET6_ND6_H_ /* see net/route.h, or net/if_inarp.h */ #ifndef RTF_ANNOUNCE #define RTF_ANNOUNCE RTF_PROTO2 #endif #include #include struct llentry; #define ND6_LLINFO_NOSTATE -2 /* * We don't need the WAITDELETE state any more, but we keep the definition * in a comment line instead of removing it. This is necessary to avoid * unintentionally reusing the value for another purpose, which might * affect backward compatibility with old applications. * (20000711 jinmei@kame.net) */ /* #define ND6_LLINFO_WAITDELETE -1 */ #define ND6_LLINFO_INCOMPLETE 0 #define ND6_LLINFO_REACHABLE 1 #define ND6_LLINFO_STALE 2 #define ND6_LLINFO_DELAY 3 #define ND6_LLINFO_PROBE 4 #define ND6_IS_LLINFO_PROBREACH(n) ((n)->ln_state > ND6_LLINFO_INCOMPLETE) #define ND6_LLINFO_PERMANENT(n) (((n)->la_expire == 0) && ((n)->ln_state > ND6_LLINFO_INCOMPLETE)) struct nd_ifinfo { u_int32_t linkmtu; /* LinkMTU */ u_int32_t maxmtu; /* Upper bound of LinkMTU */ u_int32_t basereachable; /* BaseReachableTime */ u_int32_t reachable; /* Reachable Time */ u_int32_t retrans; /* Retrans Timer */ u_int32_t flags; /* Flags */ int recalctm; /* BaseReacable re-calculation timer */ u_int8_t chlim; /* CurHopLimit */ u_int8_t initialized; /* Flag to see the entry is initialized */ /* the following 3 members are for privacy extension for addrconf */ u_int8_t randomseed0[8]; /* upper 64 bits of MD5 digest */ u_int8_t randomseed1[8]; /* lower 64 bits (usually the EUI64 IFID) */ u_int8_t randomid[8]; /* current random ID */ }; #define ND6_IFF_PERFORMNUD 0x1 #define ND6_IFF_ACCEPT_RTADV 0x2 #define ND6_IFF_PREFER_SOURCE 0x4 /* Not used in FreeBSD. */ #define ND6_IFF_IFDISABLED 0x8 /* IPv6 operation is disabled due to * DAD failure. (XXX: not ND-specific) */ #define ND6_IFF_DONT_SET_IFROUTE 0x10 #define ND6_IFF_AUTO_LINKLOCAL 0x20 #define ND6_IFF_NO_RADR 0x40 #define ND6_IFF_NO_PREFER_IFACE 0x80 /* XXX: not related to ND. */ #define ND6_IFF_NO_DAD 0x100 #ifdef _KERNEL #define ND_IFINFO(ifp) \ (((struct in6_ifextra *)(ifp)->if_afdata[AF_INET6])->nd_ifinfo) #define IN6_LINKMTU(ifp) \ ((ND_IFINFO(ifp)->linkmtu && ND_IFINFO(ifp)->linkmtu < (ifp)->if_mtu) \ ? ND_IFINFO(ifp)->linkmtu \ : ((ND_IFINFO(ifp)->maxmtu && ND_IFINFO(ifp)->maxmtu < (ifp)->if_mtu) \ ? ND_IFINFO(ifp)->maxmtu : (ifp)->if_mtu)) #endif struct in6_nbrinfo { char ifname[IFNAMSIZ]; /* if name, e.g. "en0" */ struct in6_addr addr; /* IPv6 address of the neighbor */ long asked; /* number of queries already sent for this addr */ int isrouter; /* if it acts as a router */ int state; /* reachability state */ int expire; /* lifetime for NDP state transition */ }; #define DRLSTSIZ 10 #define PRLSTSIZ 10 struct in6_drlist { char ifname[IFNAMSIZ]; struct { struct in6_addr rtaddr; u_char flags; u_short rtlifetime; u_long expire; u_short if_index; } defrouter[DRLSTSIZ]; }; struct in6_defrouter { struct sockaddr_in6 rtaddr; u_char flags; u_short rtlifetime; u_long expire; u_short if_index; }; #ifdef _KERNEL struct in6_oprlist { char ifname[IFNAMSIZ]; struct { struct in6_addr prefix; struct prf_ra raflags; u_char prefixlen; u_char origin; u_long vltime; u_long pltime; u_long expire; u_short if_index; u_short advrtrs; /* number of advertisement routers */ struct in6_addr advrtr[DRLSTSIZ]; /* XXX: explicit limit */ } prefix[PRLSTSIZ]; }; #endif struct in6_prlist { char ifname[IFNAMSIZ]; struct { struct in6_addr prefix; struct prf_ra raflags; u_char prefixlen; u_char origin; u_int32_t vltime; u_int32_t pltime; time_t expire; u_short if_index; u_short advrtrs; /* number of advertisement routers */ struct in6_addr advrtr[DRLSTSIZ]; /* XXX: explicit limit */ } prefix[PRLSTSIZ]; }; struct in6_prefix { struct sockaddr_in6 prefix; struct prf_ra raflags; u_char prefixlen; u_char origin; u_int32_t vltime; u_int32_t pltime; time_t expire; u_int32_t flags; int refcnt; u_short if_index; u_short advrtrs; /* number of advertisement routers */ /* struct sockaddr_in6 advrtr[] */ }; #ifdef _KERNEL struct in6_ondireq { char ifname[IFNAMSIZ]; struct { u_int32_t linkmtu; /* LinkMTU */ u_int32_t maxmtu; /* Upper bound of LinkMTU */ u_int32_t basereachable; /* BaseReachableTime */ u_int32_t reachable; /* Reachable Time */ u_int32_t retrans; /* Retrans Timer */ u_int32_t flags; /* Flags */ int recalctm; /* BaseReacable re-calculation timer */ u_int8_t chlim; /* CurHopLimit */ u_int8_t receivedra; } ndi; }; #endif struct in6_ndireq { char ifname[IFNAMSIZ]; struct nd_ifinfo ndi; }; struct in6_ndifreq { char ifname[IFNAMSIZ]; u_long ifindex; }; /* Prefix status */ #define NDPRF_ONLINK 0x1 #define NDPRF_DETACHED 0x2 /* protocol constants */ #define MAX_RTR_SOLICITATION_DELAY 1 /* 1sec */ #define RTR_SOLICITATION_INTERVAL 4 /* 4sec */ #define MAX_RTR_SOLICITATIONS 3 #define ND6_INFINITE_LIFETIME 0xffffffff #ifdef _KERNEL /* node constants */ #define MAX_REACHABLE_TIME 3600000 /* msec */ #define REACHABLE_TIME 30000 /* msec */ #define RETRANS_TIMER 1000 /* msec */ #define MIN_RANDOM_FACTOR 512 /* 1024 * 0.5 */ #define MAX_RANDOM_FACTOR 1536 /* 1024 * 1.5 */ #define DEF_TEMP_VALID_LIFETIME 604800 /* 1 week */ #define DEF_TEMP_PREFERRED_LIFETIME 86400 /* 1 day */ #define TEMPADDR_REGEN_ADVANCE 5 /* sec */ #define MAX_TEMP_DESYNC_FACTOR 600 /* 10 min */ #define ND_COMPUTE_RTIME(x) \ (((MIN_RANDOM_FACTOR * (x >> 10)) + (arc4random() & \ ((MAX_RANDOM_FACTOR - MIN_RANDOM_FACTOR) * (x >> 10)))) /1000) TAILQ_HEAD(nd_drhead, nd_defrouter); struct nd_defrouter { TAILQ_ENTRY(nd_defrouter) dr_entry; struct in6_addr rtaddr; u_char flags; /* flags on RA message */ u_short rtlifetime; u_long expire; struct ifnet *ifp; int installed; /* is installed into kernel routing table */ }; struct nd_prefixctl { struct ifnet *ndpr_ifp; /* prefix */ struct sockaddr_in6 ndpr_prefix; u_char ndpr_plen; u_int32_t ndpr_vltime; /* advertised valid lifetime */ u_int32_t ndpr_pltime; /* advertised preferred lifetime */ struct prf_ra ndpr_flags; }; struct nd_prefix { struct ifnet *ndpr_ifp; LIST_ENTRY(nd_prefix) ndpr_entry; struct sockaddr_in6 ndpr_prefix; /* prefix */ struct in6_addr ndpr_mask; /* netmask derived from the prefix */ u_int32_t ndpr_vltime; /* advertised valid lifetime */ u_int32_t ndpr_pltime; /* advertised preferred lifetime */ time_t ndpr_expire; /* expiration time of the prefix */ time_t ndpr_preferred; /* preferred time of the prefix */ time_t ndpr_lastupdate; /* reception time of last advertisement */ struct prf_ra ndpr_flags; u_int32_t ndpr_stateflags; /* actual state flags */ /* list of routers that advertise the prefix: */ LIST_HEAD(pr_rtrhead, nd_pfxrouter) ndpr_advrtrs; u_char ndpr_plen; int ndpr_refcnt; /* reference couter from addresses */ }; #define ndpr_raf ndpr_flags #define ndpr_raf_onlink ndpr_flags.onlink #define ndpr_raf_auto ndpr_flags.autonomous #define ndpr_raf_router ndpr_flags.router /* * Message format for use in obtaining information about prefixes * from inet6 sysctl function */ struct inet6_ndpr_msghdr { u_short inpm_msglen; /* to skip over non-understood messages */ u_char inpm_version; /* future binary compatibility */ u_char inpm_type; /* message type */ struct in6_addr inpm_prefix; u_long prm_vltim; u_long prm_pltime; u_long prm_expire; u_long prm_preferred; struct in6_prflags prm_flags; u_short prm_index; /* index for associated ifp */ u_char prm_plen; /* length of prefix in bits */ }; #define prm_raf_onlink prm_flags.prf_ra.onlink #define prm_raf_auto prm_flags.prf_ra.autonomous #define prm_statef_onlink prm_flags.prf_state.onlink #define prm_rrf_decrvalid prm_flags.prf_rr.decrvalid #define prm_rrf_decrprefd prm_flags.prf_rr.decrprefd struct nd_pfxrouter { LIST_ENTRY(nd_pfxrouter) pfr_entry; struct nd_defrouter *router; }; LIST_HEAD(nd_prhead, nd_prefix); /* nd6.c */ VNET_DECLARE(int, nd6_prune); VNET_DECLARE(int, nd6_delay); VNET_DECLARE(int, nd6_umaxtries); VNET_DECLARE(int, nd6_mmaxtries); VNET_DECLARE(int, nd6_useloopback); VNET_DECLARE(int, nd6_maxnudhint); VNET_DECLARE(int, nd6_gctimer); VNET_DECLARE(struct nd_drhead, nd_defrouter); VNET_DECLARE(struct nd_prhead, nd_prefix); VNET_DECLARE(int, nd6_debug); VNET_DECLARE(int, nd6_onlink_ns_rfc4861); #define V_nd6_prune VNET(nd6_prune) #define V_nd6_delay VNET(nd6_delay) #define V_nd6_umaxtries VNET(nd6_umaxtries) #define V_nd6_mmaxtries VNET(nd6_mmaxtries) #define V_nd6_useloopback VNET(nd6_useloopback) #define V_nd6_maxnudhint VNET(nd6_maxnudhint) #define V_nd6_gctimer VNET(nd6_gctimer) #define V_nd_defrouter VNET(nd_defrouter) #define V_nd_prefix VNET(nd_prefix) #define V_nd6_debug VNET(nd6_debug) #define V_nd6_onlink_ns_rfc4861 VNET(nd6_onlink_ns_rfc4861) #define nd6log(x) do { if (V_nd6_debug) log x; } while (/*CONSTCOND*/ 0) -VNET_DECLARE(struct callout, nd6_timer_ch); -#define V_nd6_timer_ch VNET(nd6_timer_ch) - /* nd6_rtr.c */ VNET_DECLARE(int, nd6_defifindex); VNET_DECLARE(int, ip6_desync_factor); /* seconds */ VNET_DECLARE(u_int32_t, ip6_temp_preferred_lifetime); /* seconds */ VNET_DECLARE(u_int32_t, ip6_temp_valid_lifetime); /* seconds */ VNET_DECLARE(int, ip6_temp_regen_advance); /* seconds */ #define V_nd6_defifindex VNET(nd6_defifindex) #define V_ip6_desync_factor VNET(ip6_desync_factor) #define V_ip6_temp_preferred_lifetime VNET(ip6_temp_preferred_lifetime) #define V_ip6_temp_valid_lifetime VNET(ip6_temp_valid_lifetime) #define V_ip6_temp_regen_advance VNET(ip6_temp_regen_advance) union nd_opts { struct nd_opt_hdr *nd_opt_array[16]; /* max = ND_OPT_NONCE */ struct { struct nd_opt_hdr *zero; struct nd_opt_hdr *src_lladdr; struct nd_opt_hdr *tgt_lladdr; struct nd_opt_prefix_info *pi_beg; /* multiple opts, start */ struct nd_opt_rd_hdr *rh; struct nd_opt_mtu *mtu; struct nd_opt_hdr *__res6; struct nd_opt_hdr *__res7; struct nd_opt_hdr *__res8; struct nd_opt_hdr *__res9; struct nd_opt_hdr *__res10; struct nd_opt_hdr *__res11; struct nd_opt_hdr *__res12; struct nd_opt_hdr *__res13; struct nd_opt_nonce *nonce; struct nd_opt_hdr *__res15; struct nd_opt_hdr *search; /* multiple opts */ struct nd_opt_hdr *last; /* multiple opts */ int done; struct nd_opt_prefix_info *pi_end;/* multiple opts, end */ } nd_opt_each; }; #define nd_opts_src_lladdr nd_opt_each.src_lladdr #define nd_opts_tgt_lladdr nd_opt_each.tgt_lladdr #define nd_opts_pi nd_opt_each.pi_beg #define nd_opts_pi_end nd_opt_each.pi_end #define nd_opts_rh nd_opt_each.rh #define nd_opts_mtu nd_opt_each.mtu #define nd_opts_nonce nd_opt_each.nonce #define nd_opts_search nd_opt_each.search #define nd_opts_last nd_opt_each.last #define nd_opts_done nd_opt_each.done /* XXX: need nd6_var.h?? */ /* nd6.c */ void nd6_init(void); #ifdef VIMAGE void nd6_destroy(void); #endif struct nd_ifinfo *nd6_ifattach(struct ifnet *); void nd6_ifdetach(struct nd_ifinfo *); int nd6_is_addr_neighbor(const struct sockaddr_in6 *, struct ifnet *); void nd6_option_init(void *, int, union nd_opts *); struct nd_opt_hdr *nd6_option(union nd_opts *); int nd6_options(union nd_opts *); struct llentry *nd6_lookup(const struct in6_addr *, int, struct ifnet *); struct llentry *nd6_alloc(const struct in6_addr *, int, struct ifnet *); void nd6_setmtu(struct ifnet *); void nd6_llinfo_setstate(struct llentry *lle, int newstate); void nd6_timer(void *); void nd6_purge(struct ifnet *); int nd6_resolve(struct ifnet *, int, struct mbuf *, const struct sockaddr *, u_char *, uint32_t *); int nd6_ioctl(u_long, caddr_t, struct ifnet *); void nd6_cache_lladdr(struct ifnet *, struct in6_addr *, char *, int, int, int); void nd6_grab_holdchain(struct llentry *, struct mbuf **, struct sockaddr_in6 *); int nd6_flush_holdchain(struct ifnet *, struct ifnet *, struct mbuf *, struct sockaddr_in6 *); int nd6_add_ifa_lle(struct in6_ifaddr *); void nd6_rem_ifa_lle(struct in6_ifaddr *, int); int nd6_output_ifp(struct ifnet *, struct ifnet *, struct mbuf *, struct sockaddr_in6 *, struct route *); /* nd6_nbr.c */ void nd6_na_input(struct mbuf *, int, int); void nd6_na_output(struct ifnet *, const struct in6_addr *, const struct in6_addr *, u_long, int, struct sockaddr *); void nd6_ns_input(struct mbuf *, int, int); void nd6_ns_output(struct ifnet *, const struct in6_addr *, const struct in6_addr *, const struct in6_addr *, uint8_t *); caddr_t nd6_ifptomac(struct ifnet *); void nd6_dad_init(void); void nd6_dad_start(struct ifaddr *, int); void nd6_dad_stop(struct ifaddr *); /* nd6_rtr.c */ void nd6_rs_input(struct mbuf *, int, int); void nd6_ra_input(struct mbuf *, int, int); void prelist_del(struct nd_prefix *); void defrouter_reset(void); void defrouter_select(void); void defrtrlist_del(struct nd_defrouter *); void prelist_remove(struct nd_prefix *); int nd6_prelist_add(struct nd_prefixctl *, struct nd_defrouter *, struct nd_prefix **); void pfxlist_onlink_check(void); struct nd_defrouter *defrouter_lookup(struct in6_addr *, struct ifnet *); struct nd_prefix *nd6_prefix_lookup(struct nd_prefixctl *); void rt6_flush(struct in6_addr *, struct ifnet *); int nd6_setdefaultiface(int); int in6_tmpifadd(const struct in6_ifaddr *, int, int); #endif /* _KERNEL */ #endif /* _NETINET6_ND6_H_ */ Index: projects/vnet/sys/netipsec/ipsec.c =================================================================== --- projects/vnet/sys/netipsec/ipsec.c (revision 294565) +++ projects/vnet/sys/netipsec/ipsec.c (revision 294566) @@ -1,1745 +1,1745 @@ /* $FreeBSD$ */ /* $KAME: ipsec.c,v 1.103 2001/05/24 07:14:18 sakane Exp $ */ /*- * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the project nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * IPsec controller part. */ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef INET6 #include #endif #include #include #ifdef INET6 #include #endif #include #include #include /*XXX*/ #include #include #include #include #include #include #include #ifdef IPSEC_DEBUG VNET_DEFINE(int, ipsec_debug) = 1; #else VNET_DEFINE(int, ipsec_debug) = 0; #endif /* NB: name changed so netstat doesn't use it. */ VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec4stat); VNET_PCPUSTAT_SYSINIT(ipsec4stat); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(ipsec4stat); #endif /* VIMAGE */ VNET_DEFINE(int, ip4_ah_offsetmask) = 0; /* maybe IP_DF? */ /* DF bit on encap. 0: clear 1: set 2: copy */ VNET_DEFINE(int, ip4_ipsec_dfbit) = 0; VNET_DEFINE(int, ip4_esp_trans_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip4_esp_net_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip4_ah_trans_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip4_ah_net_deflev) = IPSEC_LEVEL_USE; /* ECN ignore(-1)/forbidden(0)/allowed(1) */ VNET_DEFINE(int, ip4_ipsec_ecn) = 0; VNET_DEFINE(int, ip4_esp_randpad) = -1; static VNET_DEFINE(struct secpolicy, def_policy); #define V_def_policy VNET(def_policy) /* * Crypto support requirements: * * 1 require hardware support * -1 require software support * 0 take anything */ VNET_DEFINE(int, crypto_support) = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE; FEATURE(ipsec, "Internet Protocol Security (IPsec)"); #ifdef IPSEC_NAT_T FEATURE(ipsec_natt, "UDP Encapsulation of IPsec ESP Packets ('NAT-T')"); #endif SYSCTL_DECL(_net_inet_ipsec); /* net.inet.ipsec */ SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_POLICY, def_policy, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(def_policy).policy, 0, "IPsec default policy."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_trans_deflev), 0, "Default ESP transport mode level"); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_net_deflev), 0, "Default ESP tunnel mode level."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_trans_deflev), 0, "AH transfer mode default level."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_net_deflev), 0, "AH tunnel mode default level."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS, ah_cleartos, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0, "If set clear type-of-service field when doing AH computation."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK, ah_offsetmask, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_offsetmask), 0, "If not set clear offset field mask when doing AH computation."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DFBIT, dfbit, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_dfbit), 0, "Do not fragment bit on encap."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_ECN, ecn, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_ecn), 0, "Explicit Congestion Notification handling."); SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0, "Enable IPsec debugging output when set."); SYSCTL_INT(_net_inet_ipsec, OID_AUTO, crypto_support, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(crypto_support), 0, "Crypto driver selection."); SYSCTL_VNET_PCPUSTAT(_net_inet_ipsec, OID_AUTO, ipsecstats, struct ipsecstat, ipsec4stat, "IPsec IPv4 statistics."); #ifdef REGRESSION /* * When set to 1, IPsec will send packets with the same sequence number. * This allows to verify if the other side has proper replay attacks detection. */ VNET_DEFINE(int, ipsec_replay) = 0; SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_replay, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_replay), 0, "Emulate replay attack"); /* * When set 1, IPsec will send packets with corrupted HMAC. * This allows to verify if the other side properly detects modified packets. */ VNET_DEFINE(int, ipsec_integrity) = 0; SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_integrity, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_integrity), 0, "Emulate man-in-the-middle attack"); #endif #ifdef INET6 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec6stat); VNET_PCPUSTAT_SYSINIT(ipsec6stat); #ifdef VIMAGE VNET_PCPUSTAT_SYSUNINIT(ipsec6stat); #endif /* VIMAGE */ VNET_DEFINE(int, ip6_esp_trans_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip6_esp_net_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip6_ah_trans_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip6_ah_net_deflev) = IPSEC_LEVEL_USE; VNET_DEFINE(int, ip6_ipsec_ecn) = 0; /* ECN ignore(-1)/forbidden(0)/allowed(1) */ SYSCTL_DECL(_net_inet6_ipsec6); /* net.inet6.ipsec6 */ SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY, def_policy, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(def_policy).policy, 0, "IPsec default policy."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_trans_deflev), 0, "Default ESP transport mode level."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_net_deflev), 0, "Default ESP tunnel mode level."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_trans_deflev), 0, "AH transfer mode default level."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_net_deflev), 0, "AH tunnel mode default level."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_ECN, ecn, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ipsec_ecn), 0, "Explicit Congestion Notification handling."); SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0, "Enable IPsec debugging output when set."); SYSCTL_VNET_PCPUSTAT(_net_inet6_ipsec6, IPSECCTL_STATS, ipsecstats, struct ipsecstat, ipsec6stat, "IPsec IPv6 statistics."); #endif /* INET6 */ static int ipsec_in_reject(struct secpolicy *, struct mbuf *); static int ipsec_setspidx_inpcb(struct mbuf *, struct inpcb *); static int ipsec_setspidx(struct mbuf *, struct secpolicyindex *, int); static void ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *, int); static int ipsec4_setspidx_ipaddr(struct mbuf *, struct secpolicyindex *); #ifdef INET6 static void ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *, int); static int ipsec6_setspidx_ipaddr(struct mbuf *, struct secpolicyindex *); #endif static void ipsec_delpcbpolicy(struct inpcbpolicy *); static struct secpolicy *ipsec_deepcopy_policy(struct secpolicy *src); static void vshiftl(unsigned char *, int, int); MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy"); /* * Return a held reference to the default SP. */ static struct secpolicy * key_allocsp_default(const char* where, int tag) { struct secpolicy *sp; KEYDEBUG(KEYDEBUG_IPSEC_STAMP, printf("DP key_allocsp_default from %s:%u\n", where, tag)); sp = &V_def_policy; if (sp->policy != IPSEC_POLICY_DISCARD && sp->policy != IPSEC_POLICY_NONE) { ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n", sp->policy, IPSEC_POLICY_NONE)); sp->policy = IPSEC_POLICY_NONE; } key_addref(sp); KEYDEBUG(KEYDEBUG_IPSEC_STAMP, printf("DP key_allocsp_default returns SP:%p (%u)\n", sp, sp->refcnt)); return (sp); } #define KEY_ALLOCSP_DEFAULT() \ key_allocsp_default(__FILE__, __LINE__) /* * For OUTBOUND packet having a socket. Searching SPD for packet, * and return a pointer to SP. * OUT: NULL: no apropreate SP found, the following value is set to error. * 0 : bypass * EACCES : discard packet. * ENOENT : ipsec_acquire() in progress, maybe. * others : error occured. * others: a pointer to SP * * NOTE: IPv6 mapped adddress concern is implemented here. */ struct secpolicy * ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir) { struct secpolicy *sp; IPSEC_ASSERT(tdbi != NULL, ("null tdbi")); IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND, ("invalid direction %u", dir)); sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir); if (sp == NULL) /*XXX????*/ sp = KEY_ALLOCSP_DEFAULT(); IPSEC_ASSERT(sp != NULL, ("null SP")); return (sp); } /* * For OUTBOUND packet having a socket. Searching SPD for packet, * and return a pointer to SP. * OUT: NULL: no apropreate SP found, the following value is set to error. * 0 : bypass * EACCES : discard packet. * ENOENT : ipsec_acquire() in progress, maybe. * others : error occured. * others: a pointer to SP * * NOTE: IPv6 mapped adddress concern is implemented here. */ static struct secpolicy * ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb *inp, int *error) { struct inpcbpolicy *pcbsp; struct secpolicy *currsp = NULL; /* Policy on socket. */ struct secpolicy *sp; IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(inp != NULL, ("null inpcb")); IPSEC_ASSERT(error != NULL, ("null error")); IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND, ("invalid direction %u", dir)); if (!key_havesp(dir)) { /* No SP found, use system default. */ sp = KEY_ALLOCSP_DEFAULT(); return (sp); } /* Set spidx in pcb. */ *error = ipsec_setspidx_inpcb(m, inp); if (*error) return (NULL); pcbsp = inp->inp_sp; IPSEC_ASSERT(pcbsp != NULL, ("null pcbsp")); switch (dir) { case IPSEC_DIR_INBOUND: currsp = pcbsp->sp_in; break; case IPSEC_DIR_OUTBOUND: currsp = pcbsp->sp_out; break; } IPSEC_ASSERT(currsp != NULL, ("null currsp")); if (pcbsp->priv) { /* When privilieged socket. */ switch (currsp->policy) { case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_IPSEC: key_addref(currsp); sp = currsp; break; case IPSEC_POLICY_ENTRUST: /* Look for a policy in SPD. */ sp = KEY_ALLOCSP(&currsp->spidx, dir); if (sp == NULL) /* No SP found. */ sp = KEY_ALLOCSP_DEFAULT(); break; default: ipseclog((LOG_ERR, "%s: Invalid policy for PCB %d\n", __func__, currsp->policy)); *error = EINVAL; return (NULL); } } else { /* Unpriv, SPD has policy. */ sp = KEY_ALLOCSP(&currsp->spidx, dir); if (sp == NULL) { /* No SP found. */ switch (currsp->policy) { case IPSEC_POLICY_BYPASS: ipseclog((LOG_ERR, "%s: Illegal policy for " "non-priviliged defined %d\n", __func__, currsp->policy)); *error = EINVAL; return (NULL); case IPSEC_POLICY_ENTRUST: sp = KEY_ALLOCSP_DEFAULT(); break; case IPSEC_POLICY_IPSEC: key_addref(currsp); sp = currsp; break; default: ipseclog((LOG_ERR, "%s: Invalid policy for " "PCB %d\n", __func__, currsp->policy)); *error = EINVAL; return (NULL); } } } IPSEC_ASSERT(sp != NULL, ("null SP (priv %u policy %u", pcbsp->priv, currsp->policy)); KEYDEBUG(KEYDEBUG_IPSEC_STAMP, printf("DP %s (priv %u policy %u) allocate SP:%p (refcnt %u)\n", __func__, pcbsp->priv, currsp->policy, sp, sp->refcnt)); return (sp); } /* * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet, * and return a pointer to SP. * OUT: positive: a pointer to the entry for security policy leaf matched. * NULL: no apropreate SP found, the following value is set to error. * 0 : bypass * EACCES : discard packet. * ENOENT : ipsec_acquire() in progress, maybe. * others : error occured. */ struct secpolicy * ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int *error) { struct secpolicyindex spidx; struct secpolicy *sp; IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(error != NULL, ("null error")); IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND, ("invalid direction %u", dir)); sp = NULL; *error = 0; if (key_havesp(dir)) { /* Make an index to look for a policy. */ *error = ipsec_setspidx(m, &spidx, 0); if (*error != 0) { DPRINTF(("%s: setpidx failed, dir %u\n", __func__, dir)); return (NULL); } spidx.dir = dir; sp = KEY_ALLOCSP(&spidx, dir); } if (sp == NULL) /* No SP found, use system default. */ sp = KEY_ALLOCSP_DEFAULT(); IPSEC_ASSERT(sp != NULL, ("null SP")); return (sp); } struct secpolicy * ipsec4_checkpolicy(struct mbuf *m, u_int dir, int *error, struct inpcb *inp) { struct secpolicy *sp; *error = 0; if (inp == NULL) sp = ipsec_getpolicybyaddr(m, dir, error); else sp = ipsec_getpolicybysock(m, dir, inp, error); if (sp == NULL) { IPSEC_ASSERT(*error != 0, ("getpolicy failed w/o error")); IPSECSTAT_INC(ips_out_inval); return (NULL); } IPSEC_ASSERT(*error == 0, ("sp w/ error set to %u", *error)); switch (sp->policy) { case IPSEC_POLICY_ENTRUST: default: printf("%s: invalid policy %u\n", __func__, sp->policy); /* FALLTHROUGH */ case IPSEC_POLICY_DISCARD: IPSECSTAT_INC(ips_out_polvio); *error = -EINVAL; /* Packet is discarded by caller. */ break; case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: KEY_FREESP(&sp); sp = NULL; /* NB: force NULL result. */ break; case IPSEC_POLICY_IPSEC: if (sp->req == NULL) /* Acquire a SA. */ *error = key_spdacquire(sp); break; } if (*error != 0) { KEY_FREESP(&sp); sp = NULL; } return (sp); } static int ipsec_setspidx_inpcb(struct mbuf *m, struct inpcb *inp) { int error; IPSEC_ASSERT(inp != NULL, ("null inp")); IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp")); IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL, ("null sp_in || sp_out")); error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1); if (error == 0) { inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND; inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx; inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND; } else { bzero(&inp->inp_sp->sp_in->spidx, sizeof (inp->inp_sp->sp_in->spidx)); bzero(&inp->inp_sp->sp_out->spidx, sizeof (inp->inp_sp->sp_in->spidx)); } return (error); } /* * Configure security policy index (src/dst/proto/sport/dport) * by looking at the content of mbuf. * The caller is responsible for error recovery (like clearing up spidx). */ static int ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport) { struct ip *ip = NULL; struct ip ipbuf; u_int v; struct mbuf *n; int len; int error; IPSEC_ASSERT(m != NULL, ("null mbuf")); /* * Validate m->m_pkthdr.len. We see incorrect length if we * mistakenly call this function with inconsistent mbuf chain * (like 4.4BSD tcp/udp processing). XXX Should we panic here? */ len = 0; for (n = m; n; n = n->m_next) len += n->m_len; if (m->m_pkthdr.len != len) { KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s: pkthdr len(%d) mismatch (%d), ignored.\n", __func__, len, m->m_pkthdr.len)); return (EINVAL); } if (m->m_pkthdr.len < sizeof(struct ip)) { KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s: pkthdr len(%d) too small (v4), ignored.\n", __func__, m->m_pkthdr.len)); return (EINVAL); } if (m->m_len >= sizeof(*ip)) ip = mtod(m, struct ip *); else { m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf); ip = &ipbuf; } v = ip->ip_v; switch (v) { case 4: error = ipsec4_setspidx_ipaddr(m, spidx); if (error) return (error); ipsec4_get_ulp(m, spidx, needport); return (0); #ifdef INET6 case 6: if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) { KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s: pkthdr len(%d) too small (v6), " "ignored\n", __func__, m->m_pkthdr.len)); return (EINVAL); } error = ipsec6_setspidx_ipaddr(m, spidx); if (error) return (error); ipsec6_get_ulp(m, spidx, needport); return (0); #endif default: KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s: " "unknown IP version %u, ignored.\n", __func__, v)); return (EINVAL); } } static void ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport) { u_int8_t nxt; int off; /* Sanity check. */ IPSEC_ASSERT(m != NULL, ("null mbuf")); IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),("packet too short")); if (m->m_len >= sizeof (struct ip)) { struct ip *ip = mtod(m, struct ip *); if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) goto done; off = ip->ip_hl << 2; nxt = ip->ip_p; } else { struct ip ih; m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih); if (ih.ip_off & htons(IP_MF | IP_OFFMASK)) goto done; off = ih.ip_hl << 2; nxt = ih.ip_p; } while (off < m->m_pkthdr.len) { struct ip6_ext ip6e; struct tcphdr th; struct udphdr uh; switch (nxt) { case IPPROTO_TCP: spidx->ul_proto = nxt; if (!needport) goto done_proto; if (off + sizeof(struct tcphdr) > m->m_pkthdr.len) goto done; m_copydata(m, off, sizeof (th), (caddr_t) &th); spidx->src.sin.sin_port = th.th_sport; spidx->dst.sin.sin_port = th.th_dport; return; case IPPROTO_UDP: spidx->ul_proto = nxt; if (!needport) goto done_proto; if (off + sizeof(struct udphdr) > m->m_pkthdr.len) goto done; m_copydata(m, off, sizeof (uh), (caddr_t) &uh); spidx->src.sin.sin_port = uh.uh_sport; spidx->dst.sin.sin_port = uh.uh_dport; return; case IPPROTO_AH: if (off + sizeof(ip6e) > m->m_pkthdr.len) goto done; /* XXX Sigh, this works but is totally bogus. */ m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e); off += (ip6e.ip6e_len + 2) << 2; nxt = ip6e.ip6e_nxt; break; case IPPROTO_ICMP: default: /* XXX Intermediate headers??? */ spidx->ul_proto = nxt; goto done_proto; } } done: spidx->ul_proto = IPSEC_ULPROTO_ANY; done_proto: spidx->src.sin.sin_port = IPSEC_PORT_ANY; spidx->dst.sin.sin_port = IPSEC_PORT_ANY; } /* Assumes that m is sane. */ static int ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx) { static const struct sockaddr_in template = { sizeof (struct sockaddr_in), AF_INET, 0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 } }; spidx->src.sin = template; spidx->dst.sin = template; if (m->m_len < sizeof (struct ip)) { m_copydata(m, offsetof(struct ip, ip_src), sizeof (struct in_addr), (caddr_t) &spidx->src.sin.sin_addr); m_copydata(m, offsetof(struct ip, ip_dst), sizeof (struct in_addr), (caddr_t) &spidx->dst.sin.sin_addr); } else { struct ip *ip = mtod(m, struct ip *); spidx->src.sin.sin_addr = ip->ip_src; spidx->dst.sin.sin_addr = ip->ip_dst; } spidx->prefs = sizeof(struct in_addr) << 3; spidx->prefd = sizeof(struct in_addr) << 3; return (0); } #ifdef INET6 static void ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport) { int off, nxt; struct tcphdr th; struct udphdr uh; struct icmp6_hdr ih; /* Sanity check. */ if (m == NULL) panic("%s: NULL pointer was passed.\n", __func__); KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s:\n", __func__); kdebug_mbuf(m)); /* Set default. */ spidx->ul_proto = IPSEC_ULPROTO_ANY; ((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY; ((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY; nxt = -1; off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); if (off < 0 || m->m_pkthdr.len < off) return; switch (nxt) { case IPPROTO_TCP: spidx->ul_proto = nxt; if (!needport) break; if (off + sizeof(struct tcphdr) > m->m_pkthdr.len) break; m_copydata(m, off, sizeof(th), (caddr_t)&th); ((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport; ((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport; break; case IPPROTO_UDP: spidx->ul_proto = nxt; if (!needport) break; if (off + sizeof(struct udphdr) > m->m_pkthdr.len) break; m_copydata(m, off, sizeof(uh), (caddr_t)&uh); ((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport; ((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport; break; case IPPROTO_ICMPV6: spidx->ul_proto = nxt; if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len) break; m_copydata(m, off, sizeof(ih), (caddr_t)&ih); ((struct sockaddr_in6 *)&spidx->src)->sin6_port = htons((uint16_t)ih.icmp6_type); ((struct sockaddr_in6 *)&spidx->dst)->sin6_port = htons((uint16_t)ih.icmp6_code); break; default: /* XXX Intermediate headers??? */ spidx->ul_proto = nxt; break; } } /* Assumes that m is sane. */ static int ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx) { struct ip6_hdr *ip6 = NULL; struct ip6_hdr ip6buf; struct sockaddr_in6 *sin6; if (m->m_len >= sizeof(*ip6)) ip6 = mtod(m, struct ip6_hdr *); else { m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf); ip6 = &ip6buf; } sin6 = (struct sockaddr_in6 *)&spidx->src; bzero(sin6, sizeof(*sin6)); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(struct sockaddr_in6); bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src)); if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) { sin6->sin6_addr.s6_addr16[1] = 0; sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]); } spidx->prefs = sizeof(struct in6_addr) << 3; sin6 = (struct sockaddr_in6 *)&spidx->dst; bzero(sin6, sizeof(*sin6)); sin6->sin6_family = AF_INET6; sin6->sin6_len = sizeof(struct sockaddr_in6); bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst)); if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) { sin6->sin6_addr.s6_addr16[1] = 0; sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]); } spidx->prefd = sizeof(struct in6_addr) << 3; return (0); } #endif int ipsec_run_hhooks(struct ipsec_ctx_data *ctx, int type) { int idx; switch (ctx->af) { #ifdef INET case AF_INET: idx = HHOOK_IPSEC_INET; break; #endif #ifdef INET6 case AF_INET6: idx = HHOOK_IPSEC_INET6; break; #endif default: return (EPFNOSUPPORT); } if (type == HHOOK_TYPE_IPSEC_IN) HHOOKS_RUN_IF(V_ipsec_hhh_in[idx], ctx, NULL); else HHOOKS_RUN_IF(V_ipsec_hhh_out[idx], ctx, NULL); if (*ctx->mp == NULL) return (EACCES); return (0); } static void ipsec_delpcbpolicy(struct inpcbpolicy *p) { free(p, M_IPSEC_INPCB); } /* Initialize policy in PCB. */ int ipsec_init_policy(struct socket *so, struct inpcbpolicy **pcb_sp) { struct inpcbpolicy *new; /* Sanity check. */ if (so == NULL || pcb_sp == NULL) panic("%s: NULL pointer was passed.\n", __func__); new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy), M_IPSEC_INPCB, M_NOWAIT|M_ZERO); if (new == NULL) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return (ENOBUFS); } new->priv = IPSEC_IS_PRIVILEGED_SO(so); if ((new->sp_in = KEY_NEWSP()) == NULL) { ipsec_delpcbpolicy(new); return (ENOBUFS); } new->sp_in->policy = IPSEC_POLICY_ENTRUST; if ((new->sp_out = KEY_NEWSP()) == NULL) { KEY_FREESP(&new->sp_in); ipsec_delpcbpolicy(new); return (ENOBUFS); } new->sp_out->policy = IPSEC_POLICY_ENTRUST; *pcb_sp = new; return (0); } /* Copy old IPsec policy into new. */ int ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new) { struct secpolicy *sp; sp = ipsec_deepcopy_policy(old->sp_in); if (sp) { KEY_FREESP(&new->sp_in); new->sp_in = sp; } else return (ENOBUFS); sp = ipsec_deepcopy_policy(old->sp_out); if (sp) { KEY_FREESP(&new->sp_out); new->sp_out = sp; } else return (ENOBUFS); new->priv = old->priv; return (0); } struct ipsecrequest * ipsec_newisr(void) { struct ipsecrequest *p; p = malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT|M_ZERO); if (p != NULL) IPSECREQUEST_LOCK_INIT(p); return (p); } void ipsec_delisr(struct ipsecrequest *p) { IPSECREQUEST_LOCK_DESTROY(p); free(p, M_IPSEC_SR); } /* Deep-copy a policy in PCB. */ static struct secpolicy * ipsec_deepcopy_policy(struct secpolicy *src) { struct ipsecrequest *newchain = NULL; struct ipsecrequest *p; struct ipsecrequest **q; struct ipsecrequest *r; struct secpolicy *dst; if (src == NULL) return (NULL); dst = KEY_NEWSP(); if (dst == NULL) return (NULL); /* * Deep-copy IPsec request chain. This is required since struct * ipsecrequest is not reference counted. */ q = &newchain; for (p = src->req; p; p = p->next) { *q = ipsec_newisr(); if (*q == NULL) goto fail; (*q)->saidx.proto = p->saidx.proto; (*q)->saidx.mode = p->saidx.mode; (*q)->level = p->level; (*q)->saidx.reqid = p->saidx.reqid; bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src)); bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst)); (*q)->sp = dst; q = &((*q)->next); } dst->req = newchain; dst->policy = src->policy; /* Do not touch the refcnt fields. */ return (dst); fail: for (p = newchain; p; p = r) { r = p->next; ipsec_delisr(p); p = NULL; } return (NULL); } /* Set policy and IPsec request if present. */ static int ipsec_set_policy_internal(struct secpolicy **pcb_sp, int optname, caddr_t request, size_t len, struct ucred *cred) { struct sadb_x_policy *xpl; struct secpolicy *newsp = NULL; int error; /* Sanity check. */ if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL) return (EINVAL); if (len < sizeof(*xpl)) return (EINVAL); xpl = (struct sadb_x_policy *)request; KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s: passed policy\n", __func__); kdebug_sadb_x_policy((struct sadb_ext *)xpl)); /* Check policy type. */ /* ipsec_set_policy_internal() accepts IPSEC, ENTRUST and BYPASS. */ if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE) return (EINVAL); /* Check privileged socket. */ if (cred != NULL && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) { error = priv_check_cred(cred, PRIV_NETINET_IPSEC, 0); if (error) return (EACCES); } /* Allocating new SP entry. */ if ((newsp = key_msg2sp(xpl, len, &error)) == NULL) return (error); /* Clear old SP and set new SP. */ KEY_FREESP(pcb_sp); *pcb_sp = newsp; KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s: new policy\n", __func__); kdebug_secpolicy(newsp)); return (0); } int ipsec_set_policy(struct inpcb *inp, int optname, caddr_t request, size_t len, struct ucred *cred) { struct sadb_x_policy *xpl; struct secpolicy **pcb_sp; /* Sanity check. */ if (inp == NULL || request == NULL) return (EINVAL); if (len < sizeof(*xpl)) return (EINVAL); xpl = (struct sadb_x_policy *)request; /* Select direction. */ switch (xpl->sadb_x_policy_dir) { case IPSEC_DIR_INBOUND: pcb_sp = &inp->inp_sp->sp_in; break; case IPSEC_DIR_OUTBOUND: pcb_sp = &inp->inp_sp->sp_out; break; default: ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__, xpl->sadb_x_policy_dir)); return (EINVAL); } return (ipsec_set_policy_internal(pcb_sp, optname, request, len, cred)); } int ipsec_get_policy(struct inpcb *inp, caddr_t request, size_t len, struct mbuf **mp) { struct sadb_x_policy *xpl; struct secpolicy *pcb_sp; /* Sanity check. */ if (inp == NULL || request == NULL || mp == NULL) return (EINVAL); IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp")); if (len < sizeof(*xpl)) return (EINVAL); xpl = (struct sadb_x_policy *)request; /* Select direction. */ switch (xpl->sadb_x_policy_dir) { case IPSEC_DIR_INBOUND: pcb_sp = inp->inp_sp->sp_in; break; case IPSEC_DIR_OUTBOUND: pcb_sp = inp->inp_sp->sp_out; break; default: ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__, xpl->sadb_x_policy_dir)); return (EINVAL); } /* Sanity check. Should be an IPSEC_ASSERT. */ if (pcb_sp == NULL) return (EINVAL); *mp = key_sp2msg(pcb_sp); if (!*mp) { ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__)); return (ENOBUFS); } (*mp)->m_type = MT_DATA; KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s:\n", __func__); kdebug_mbuf(*mp)); return (0); } /* Delete policy in PCB. */ int ipsec_delete_pcbpolicy(struct inpcb *inp) { IPSEC_ASSERT(inp != NULL, ("null inp")); if (inp->inp_sp == NULL) return (0); if (inp->inp_sp->sp_in != NULL) KEY_FREESP(&inp->inp_sp->sp_in); if (inp->inp_sp->sp_out != NULL) KEY_FREESP(&inp->inp_sp->sp_out); ipsec_delpcbpolicy(inp->inp_sp); inp->inp_sp = NULL; return (0); } /* * Return current level. * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned. */ u_int ipsec_get_reqlevel(struct ipsecrequest *isr) { u_int level = 0; u_int esp_trans_deflev, esp_net_deflev; u_int ah_trans_deflev, ah_net_deflev; IPSEC_ASSERT(isr != NULL && isr->sp != NULL, ("null argument")); IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family, ("af family mismatch, src %u, dst %u", isr->sp->spidx.src.sa.sa_family, isr->sp->spidx.dst.sa.sa_family)); /* XXX Note that we have ipseclog() expanded here - code sync issue. */ #define IPSEC_CHECK_DEFAULT(lev) \ (((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE \ && (lev) != IPSEC_LEVEL_UNIQUE) \ ? (V_ipsec_debug \ ? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\ (lev), IPSEC_LEVEL_REQUIRE) \ : 0), \ (lev) = IPSEC_LEVEL_REQUIRE, \ (lev) \ : (lev)) /* Set default level. */ switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) { #ifdef INET case AF_INET: esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev); esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev); ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev); ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev); break; #endif #ifdef INET6 case AF_INET6: esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev); esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev); ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev); ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev); break; #endif /* INET6 */ default: panic("%s: unknown af %u", __func__, isr->sp->spidx.src.sa.sa_family); } #undef IPSEC_CHECK_DEFAULT /* Set level. */ switch (isr->level) { case IPSEC_LEVEL_DEFAULT: switch (isr->saidx.proto) { case IPPROTO_ESP: if (isr->saidx.mode == IPSEC_MODE_TUNNEL) level = esp_net_deflev; else level = esp_trans_deflev; break; case IPPROTO_AH: if (isr->saidx.mode == IPSEC_MODE_TUNNEL) level = ah_net_deflev; else level = ah_trans_deflev; break; case IPPROTO_IPCOMP: /* * We don't really care, as IPcomp document says that * we shouldn't compress small packets. */ level = IPSEC_LEVEL_USE; break; default: panic("%s: Illegal protocol defined %u\n", __func__, isr->saidx.proto); } break; case IPSEC_LEVEL_USE: case IPSEC_LEVEL_REQUIRE: level = isr->level; break; case IPSEC_LEVEL_UNIQUE: level = IPSEC_LEVEL_REQUIRE; break; default: panic("%s: Illegal IPsec level %u\n", __func__, isr->level); } return (level); } /* * Check security policy requirements against the actual * packet contents. Return one if the packet should be * reject as "invalid"; otherwiser return zero to have the * packet treated as "valid". * * OUT: * 0: valid * 1: invalid */ static int ipsec_in_reject(struct secpolicy *sp, struct mbuf *m) { struct ipsecrequest *isr; int need_auth; KEYDEBUG(KEYDEBUG_IPSEC_DATA, printf("%s: using SP\n", __func__); kdebug_secpolicy(sp)); /* Check policy. */ switch (sp->policy) { case IPSEC_POLICY_DISCARD: return (1); case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: return (0); } IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC, ("invalid policy %u", sp->policy)); /* XXX Should compare policy against IPsec header history. */ need_auth = 0; for (isr = sp->req; isr != NULL; isr = isr->next) { if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE) continue; switch (isr->saidx.proto) { case IPPROTO_ESP: if ((m->m_flags & M_DECRYPTED) == 0) { KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s: ESP m_flags:%x\n", __func__, m->m_flags)); return (1); } if (!need_auth && isr->sav != NULL && isr->sav->tdb_authalgxform != NULL && (m->m_flags & M_AUTHIPDGM) == 0) { KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s: ESP/AH m_flags:%x\n", __func__, m->m_flags)); return (1); } break; case IPPROTO_AH: need_auth = 1; if ((m->m_flags & M_AUTHIPHDR) == 0) { KEYDEBUG(KEYDEBUG_IPSEC_DUMP, printf("%s: AH m_flags:%x\n", __func__, m->m_flags)); return (1); } break; case IPPROTO_IPCOMP: /* * We don't really care, as IPcomp document * says that we shouldn't compress small * packets. IPComp policy should always be * treated as being in "use" level. */ break; } } return (0); /* Valid. */ } /* * Non zero return value means security policy DISCARD or policy violation. */ static int ipsec46_in_reject(struct mbuf *m, struct inpcb *inp) { struct secpolicy *sp; int error; int result; if (!key_havesp(IPSEC_DIR_INBOUND)) return 0; IPSEC_ASSERT(m != NULL, ("null mbuf")); /* Get SP for this packet. */ if (inp == NULL) sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, &error); else sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error); if (sp != NULL) { result = ipsec_in_reject(sp, m); KEY_FREESP(&sp); } else { result = 1; /* treat errors as policy violation */ } return (result); } /* * Check AH/ESP integrity. * This function is called from tcp_input(), udp_input(), * and {ah,esp}4_input for tunnel mode. */ int ipsec4_in_reject(struct mbuf *m, struct inpcb *inp) { int result; result = ipsec46_in_reject(m, inp); if (result) IPSECSTAT_INC(ips_in_polvio); return (result); } #ifdef INET6 /* * Check AH/ESP integrity. * This function is called from tcp6_input(), udp6_input(), * and {ah,esp}6_input for tunnel mode. */ int ipsec6_in_reject(struct mbuf *m, struct inpcb *inp) { int result; result = ipsec46_in_reject(m, inp); if (result) IPSEC6STAT_INC(ips_in_polvio); return (result); } #endif /* * Compute the byte size to be occupied by IPsec header. * In case it is tunnelled, it includes the size of outer IP header. * NOTE: SP passed is freed in this function. */ static size_t ipsec_hdrsiz_internal(struct secpolicy *sp) { struct ipsecrequest *isr; size_t size; KEYDEBUG(KEYDEBUG_IPSEC_DATA, printf("%s: using SP\n", __func__); kdebug_secpolicy(sp)); switch (sp->policy) { case IPSEC_POLICY_DISCARD: case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: return (0); } IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC, ("invalid policy %u", sp->policy)); size = 0; for (isr = sp->req; isr != NULL; isr = isr->next) { size_t clen = 0; switch (isr->saidx.proto) { case IPPROTO_ESP: clen = esp_hdrsiz(isr->sav); break; case IPPROTO_AH: clen = ah_hdrsiz(isr->sav); break; case IPPROTO_IPCOMP: clen = sizeof(struct ipcomp); break; } if (isr->saidx.mode == IPSEC_MODE_TUNNEL) { switch (isr->saidx.dst.sa.sa_family) { case AF_INET: clen += sizeof(struct ip); break; #ifdef INET6 case AF_INET6: clen += sizeof(struct ip6_hdr); break; #endif default: ipseclog((LOG_ERR, "%s: unknown AF %d in " "IPsec tunnel SA\n", __func__, ((struct sockaddr *)&isr->saidx.dst)->sa_family)); break; } } size += clen; } return (size); } /* * This function is called from ipsec_hdrsiz_tcp(), ip_ipsec_mtu(), * disabled ip6_ipsec_mtu() and ip6_forward(). */ size_t ipsec_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp) { struct secpolicy *sp; int error; size_t size; if (!key_havesp(dir)) return 0; IPSEC_ASSERT(m != NULL, ("null mbuf")); /* Get SP for this packet. */ if (inp == NULL) sp = ipsec_getpolicybyaddr(m, dir, &error); else sp = ipsec_getpolicybysock(m, dir, inp, &error); if (sp != NULL) { size = ipsec_hdrsiz_internal(sp); KEYDEBUG(KEYDEBUG_IPSEC_DATA, printf("%s: size:%lu.\n", __func__, (unsigned long)size)); KEY_FREESP(&sp); } else { size = 0; /* XXX Should be panic? * -> No, we are called w/o knowing if * IPsec processing is needed. */ } return (size); } /* * Check the variable replay window. * ipsec_chkreplay() performs replay check before ICV verification. * ipsec_updatereplay() updates replay bitmap. This must be called after * ICV verification (it also performs replay check, which is usually done * beforehand). * 0 (zero) is returned if packet disallowed, 1 if packet permitted. * * Based on RFC 2401. */ int ipsec_chkreplay(u_int32_t seq, struct secasvar *sav) { const struct secreplay *replay; u_int32_t diff; int fr; u_int32_t wsizeb; /* Constant: bits of window size. */ int frlast; /* Constant: last frame. */ IPSEC_ASSERT(sav != NULL, ("Null SA")); IPSEC_ASSERT(sav->replay != NULL, ("Null replay state")); replay = sav->replay; if (replay->wsize == 0) return (1); /* No need to check replay. */ /* Constant. */ frlast = replay->wsize - 1; wsizeb = replay->wsize << 3; /* Sequence number of 0 is invalid. */ if (seq == 0) return (0); /* First time is always okay. */ if (replay->count == 0) return (1); if (seq > replay->lastseq) { /* Larger sequences are okay. */ return (1); } else { /* seq is equal or less than lastseq. */ diff = replay->lastseq - seq; /* Over range to check, i.e. too old or wrapped. */ if (diff >= wsizeb) return (0); fr = frlast - diff / 8; /* This packet already seen? */ if ((replay->bitmap)[fr] & (1 << (diff % 8))) return (0); /* Out of order but good. */ return (1); } } /* * Check replay counter whether to update or not. * OUT: 0: OK * 1: NG */ int ipsec_updatereplay(u_int32_t seq, struct secasvar *sav) { char buf[128]; struct secreplay *replay; u_int32_t diff; int fr; u_int32_t wsizeb; /* Constant: bits of window size. */ int frlast; /* Constant: last frame. */ IPSEC_ASSERT(sav != NULL, ("Null SA")); IPSEC_ASSERT(sav->replay != NULL, ("Null replay state")); replay = sav->replay; if (replay->wsize == 0) goto ok; /* No need to check replay. */ /* Constant. */ frlast = replay->wsize - 1; wsizeb = replay->wsize << 3; /* Sequence number of 0 is invalid. */ if (seq == 0) return (1); /* First time. */ if (replay->count == 0) { replay->lastseq = seq; bzero(replay->bitmap, replay->wsize); (replay->bitmap)[frlast] = 1; goto ok; } if (seq > replay->lastseq) { /* seq is larger than lastseq. */ diff = seq - replay->lastseq; /* New larger sequence number. */ if (diff < wsizeb) { /* In window. */ /* Set bit for this packet. */ vshiftl(replay->bitmap, diff, replay->wsize); (replay->bitmap)[frlast] |= 1; } else { /* This packet has a "way larger". */ bzero(replay->bitmap, replay->wsize); (replay->bitmap)[frlast] = 1; } replay->lastseq = seq; /* Larger is good. */ } else { /* seq is equal or less than lastseq. */ diff = replay->lastseq - seq; /* Over range to check, i.e. too old or wrapped. */ if (diff >= wsizeb) return (1); fr = frlast - diff / 8; /* This packet already seen? */ if ((replay->bitmap)[fr] & (1 << (diff % 8))) return (1); /* Mark as seen. */ (replay->bitmap)[fr] |= (1 << (diff % 8)); /* Out of order but good. */ } ok: if (replay->count == ~0) { /* Set overflow flag. */ replay->overflow++; /* Don't increment, no more packets accepted. */ if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0) return (1); ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n", __func__, replay->overflow, ipsec_logsastr(sav, buf, sizeof(buf)))); } replay->count++; return (0); } /* * Shift variable length buffer to left. * IN: bitmap: pointer to the buffer * nbit: the number of to shift. * wsize: buffer size (bytes). */ static void vshiftl(unsigned char *bitmap, int nbit, int wsize) { int s, j, i; unsigned char over; for (j = 0; j < nbit; j += 8) { s = (nbit - j < 8) ? (nbit - j): 8; bitmap[0] <<= s; for (i = 1; i < wsize; i++) { over = (bitmap[i] >> (8 - s)); bitmap[i] <<= s; bitmap[i-1] |= over; } } } /* Return a printable string for the address. */ char* ipsec_address(union sockaddr_union* sa, char *buf, socklen_t size) { switch (sa->sa.sa_family) { #ifdef INET case AF_INET: return (inet_ntop(AF_INET, &sa->sin.sin_addr, buf, size)); #endif /* INET */ #ifdef INET6 case AF_INET6: return (inet_ntop(AF_INET6, &sa->sin6.sin6_addr, buf, size)); #endif /* INET6 */ default: return ("(unknown address family)"); } } char * ipsec_logsastr(struct secasvar *sav, char *buf, size_t size) { char sbuf[INET6_ADDRSTRLEN], dbuf[INET6_ADDRSTRLEN]; IPSEC_ASSERT(sav->sah->saidx.src.sa.sa_family == sav->sah->saidx.dst.sa.sa_family, ("address family mismatch")); snprintf(buf, size, "SA(SPI=%08lx src=%s dst=%s)", (u_long)ntohl(sav->spi), ipsec_address(&sav->sah->saidx.src, sbuf, sizeof(sbuf)), ipsec_address(&sav->sah->saidx.dst, dbuf, sizeof(dbuf))); return (buf); } void ipsec_dumpmbuf(struct mbuf *m) { int totlen; int i; u_char *p; totlen = 0; printf("---\n"); while (m) { p = mtod(m, u_char *); for (i = 0; i < m->m_len; i++) { printf("%02x ", p[i]); totlen++; if (totlen % 16 == 0) printf("\n"); } m = m->m_next; } if (totlen % 16 != 0) printf("\n"); printf("---\n"); } static void def_policy_init(const void *unused __unused) { bzero(&V_def_policy, sizeof(struct secpolicy)); V_def_policy.policy = IPSEC_POLICY_NONE; V_def_policy.refcnt = 1; } -VNET_SYSINIT(def_policy_init, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, +VNET_SYSINIT(def_policy_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, def_policy_init, NULL); /* XXX This stuff doesn't belong here... */ static struct xformsw* xforms = NULL; /* * Register a transform; typically at system startup. */ void xform_register(struct xformsw* xsp) { xsp->xf_next = xforms; xforms = xsp; } /* * Initialize transform support in an sav. */ int xform_init(struct secasvar *sav, int xftype) { struct xformsw *xsp; if (sav->tdb_xform != NULL) /* Previously initialized. */ return (0); for (xsp = xforms; xsp; xsp = xsp->xf_next) if (xsp->xf_type == xftype) return ((*xsp->xf_init)(sav, xsp)); return (EINVAL); } Index: projects/vnet/sys/netipsec/xform_tcp.c =================================================================== --- projects/vnet/sys/netipsec/xform_tcp.c (revision 294565) +++ projects/vnet/sys/netipsec/xform_tcp.c (revision 294566) @@ -1,170 +1,170 @@ /* $FreeBSD$ */ /*- * Copyright (c) 2003 Bruce M. Simpson * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ /* TCP MD5 Signature Option (RFC2385) */ #include "opt_inet.h" #include "opt_inet6.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #endif #include #include /* * Initialize a TCP-MD5 SA. Called when the SA is being set up. * * We don't need to set up the tdb prefixed fields, as we don't use the * opencrypto code; we just perform a key length check. * * XXX: Currently we only allow a single 'magic' SPI to be used. * * This allows per-host granularity without affecting the userland * interface, which is a simple socket option toggle switch, * TCP_SIGNATURE_ENABLE. * * To allow per-service granularity requires that we have a means * of mapping port to SPI. The mandated way of doing this is to * use SPD entries to specify packet flows which get the TCP-MD5 * treatment, however the code to do this is currently unstable * and unsuitable for production use. * * Therefore we use this compromise in the meantime. */ static int tcpsignature_init(struct secasvar *sav, struct xformsw *xsp) { int keylen; if (sav->spi != htonl(TCP_SIG_SPI)) { DPRINTF(("%s: SPI must be TCP_SIG_SPI (0x1000)\n", __func__)); return (EINVAL); } if (sav->alg_auth != SADB_X_AALG_TCP_MD5) { DPRINTF(("%s: unsupported authentication algorithm %u\n", __func__, sav->alg_auth)); return (EINVAL); } if (sav->key_auth == NULL) { DPRINTF(("%s: no authentication key present\n", __func__)); return (EINVAL); } keylen = _KEYLEN(sav->key_auth); if ((keylen < TCP_KEYLEN_MIN) || (keylen > TCP_KEYLEN_MAX)) { DPRINTF(("%s: invalid key length %u\n", __func__, keylen)); return (EINVAL); } return (0); } /* * Paranoia. * * Called when the SA is deleted. */ static int tcpsignature_zeroize(struct secasvar *sav) { if (sav->key_auth) bzero(sav->key_auth->key_data, _KEYLEN(sav->key_auth)); sav->tdb_cryptoid = 0; sav->tdb_authalgxform = NULL; sav->tdb_xform = NULL; return (0); } /* * Verify that an input packet passes authentication. * Called from the ipsec layer. * We do this from within tcp itself, so this routine is just a stub. */ static int tcpsignature_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff) { return (0); } /* * Prepend the authentication header. * Called from the ipsec layer. * We do this from within tcp itself, so this routine is just a stub. */ static int tcpsignature_output(struct mbuf *m, struct ipsecrequest *isr, struct mbuf **mp, int skip, int protoff) { return (EINVAL); } static struct xformsw tcpsignature_xformsw = { XF_TCPSIGNATURE, XFT_AUTH, "TCPMD5", tcpsignature_init, tcpsignature_zeroize, tcpsignature_input, tcpsignature_output }; static void tcpsignature_attach(void) { xform_register(&tcpsignature_xformsw); } -SYSINIT(tcpsignature_xform_init, SI_SUB_DRIVERS, SI_ORDER_FIRST, +SYSINIT(tcpsignature_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, tcpsignature_attach, NULL); Index: projects/vnet/sys/netpfil/ipfw/dn_sched.h =================================================================== --- projects/vnet/sys/netpfil/ipfw/dn_sched.h (revision 294565) +++ projects/vnet/sys/netpfil/ipfw/dn_sched.h (revision 294566) @@ -1,192 +1,192 @@ /* * Copyright (c) 2010 Riccardo Panicucci, Luigi Rizzo, Universita` di Pisa * All rights reserved * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * The API to write a packet scheduling algorithm for dummynet. * * $FreeBSD$ */ #ifndef _DN_SCHED_H #define _DN_SCHED_H #define DN_MULTIQUEUE 0x01 /* * Descriptor for a scheduling algorithm. * Contains all function pointers for a given scheduler * This is typically created when a module is loaded, and stored * in a global list of schedulers. */ struct dn_alg { uint32_t type; /* the scheduler type */ const char *name; /* scheduler name */ uint32_t flags; /* DN_MULTIQUEUE if supports multiple queues */ /* * The following define the size of 3 optional data structures * that may need to be allocated at runtime, and are appended * to each of the base data structures: scheduler, sched.inst, * and queue. We don't have a per-flowset structure. */ /* + parameters attached to the template, e.g. * default queue sizes, weights, quantum size, and so on; */ size_t schk_datalen; /* + per-instance parameters, such as timestamps, * containers for queues, etc; */ size_t si_datalen; size_t q_datalen; /* per-queue parameters (e.g. S,F) */ /* * Methods implemented by the scheduler: * enqueue enqueue packet 'm' on scheduler 's', queue 'q'. * q is NULL for !MULTIQUEUE. * Return 0 on success, 1 on drop (packet consumed anyways). * Note that q should be interpreted only as a hint * on the flow that the mbuf belongs to: while a * scheduler will normally enqueue m into q, it is ok * to leave q alone and put the mbuf elsewhere. * This function is called in two cases: * - when a new packet arrives to the scheduler; * - when a scheduler is reconfigured. In this case the * call is issued by the new_queue callback, with a * non empty queue (q) and m pointing to the first * mbuf in the queue. For this reason, the function * should internally check for (m != q->mq.head) * before calling dn_enqueue(). * * dequeue Called when scheduler instance 's' can * dequeue a packet. Return NULL if none are available. * XXX what about non work-conserving ? * * config called on 'sched X config ...', normally writes * in the area of size sch_arg * * destroy called on 'sched delete', frees everything * in sch_arg (other parts are handled by more specific * functions) * * new_sched called when a new instance is created, e.g. * to create the local queue for !MULTIQUEUE, set V or * copy parameters for WFQ, and so on. * * free_sched called when deleting an instance, cleans * extra data in the per-instance area. * * new_fsk called when a flowset is linked to a scheduler, * e.g. to validate parameters such as weights etc. * free_fsk when a flowset is unlinked from a scheduler. * (probably unnecessary) * * new_queue called to set the per-queue parameters, * e.g. S and F, adjust sum of weights in the parent, etc. * * The new_queue callback is normally called from when * creating a new queue. In some cases (such as a * scheduler change or reconfiguration) it can be called * with a non empty queue. In this case, the queue * In case of non empty queue, the new_queue callback could * need to call the enqueue function. In this case, * the callback should eventually call enqueue() passing * as m the first element in the queue. * * free_queue actions related to a queue removal, e.g. undo * all the above. If the queue has data in it, also remove * from the scheduler. This can e.g. happen during a reconfigure. */ int (*enqueue)(struct dn_sch_inst *, struct dn_queue *, struct mbuf *); struct mbuf * (*dequeue)(struct dn_sch_inst *); int (*config)(struct dn_schk *); int (*destroy)(struct dn_schk*); int (*new_sched)(struct dn_sch_inst *); int (*free_sched)(struct dn_sch_inst *); int (*new_fsk)(struct dn_fsk *f); int (*free_fsk)(struct dn_fsk *f); int (*new_queue)(struct dn_queue *q); int (*free_queue)(struct dn_queue *q); /* run-time fields */ int ref_count; /* XXX number of instances in the system */ SLIST_ENTRY(dn_alg) next; /* Next scheduler in the list */ }; /* MSVC does not support initializers so we need this ugly macro */ #ifdef _WIN32 #define _SI(fld) #else #define _SI(fld) fld #endif /* * Additionally, dummynet exports some functions and macros * to be used by schedulers: */ void dn_free_pkts(struct mbuf *mnext); int dn_enqueue(struct dn_queue *q, struct mbuf* m, int drop); /* bound a variable between min and max */ int ipdn_bound_var(int *v, int dflt, int lo, int hi, const char *msg); /* * Extract the head of a queue, update stats. Must be the very last * thing done on a dequeue as the queue itself may go away. */ static __inline struct mbuf* dn_dequeue(struct dn_queue *q) { struct mbuf *m = q->mq.head; if (m == NULL) return NULL; q->mq.head = m->m_nextpkt; q->mq.count--; /* Update stats for the queue */ q->ni.length--; q->ni.len_bytes -= m->m_pkthdr.len; if (q->_si) { q->_si->ni.length--; q->_si->ni.len_bytes -= m->m_pkthdr.len; } if (q->ni.length == 0) /* queue is now idle */ q->q_time = dn_cfg.curr_time; return m; } int dn_sched_modevent(module_t mod, int cmd, void *arg); #define DECLARE_DNSCHED_MODULE(name, dnsched) \ static moduledata_t name##_mod = { \ #name, dn_sched_modevent, dnsched \ }; \ DECLARE_MODULE(name, name##_mod, \ - SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY); \ + SI_SUB_FW, SI_ORDER_ANY); \ MODULE_DEPEND(name, dummynet, 3, 3, 3) #endif /* _DN_SCHED_H */ Index: projects/vnet/sys/netpfil/ipfw/ip_dummynet.c =================================================================== --- projects/vnet/sys/netpfil/ipfw/ip_dummynet.c (revision 294565) +++ projects/vnet/sys/netpfil/ipfw/ip_dummynet.c (revision 294566) @@ -1,2321 +1,2321 @@ /*- * Copyright (c) 1998-2002,2010 Luigi Rizzo, Universita` di Pisa * Portions Copyright (c) 2000 Akamba Corp. * All rights reserved * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); /* * Configuration and internal object management for dummynet. */ #include "opt_inet6.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* IFNAMSIZ, struct ifaddr, ifq head, lock.h mutex.h */ #include #include /* ip_output(), IP_FORWARDING */ #include #include #include #include #include #include /* which objects to copy */ #define DN_C_LINK 0x01 #define DN_C_SCH 0x02 #define DN_C_FLOW 0x04 #define DN_C_FS 0x08 #define DN_C_QUEUE 0x10 /* we use this argument in case of a schk_new */ struct schk_new_arg { struct dn_alg *fp; struct dn_sch *sch; }; /*---- callout hooks. ----*/ static struct callout dn_timeout; static int dn_gone; static struct task dn_task; static struct taskqueue *dn_tq = NULL; static void dummynet(void *arg) { (void)arg; /* UNUSED */ taskqueue_enqueue_fast(dn_tq, &dn_task); } void dn_reschedule(void) { if (dn_gone != 0) return; callout_reset_sbt(&dn_timeout, tick_sbt, 0, dummynet, NULL, C_HARDCLOCK | C_DIRECT_EXEC); } /*----- end of callout hooks -----*/ /* Return a scheduler descriptor given the type or name. */ static struct dn_alg * find_sched_type(int type, char *name) { struct dn_alg *d; SLIST_FOREACH(d, &dn_cfg.schedlist, next) { if (d->type == type || (name && !strcasecmp(d->name, name))) return d; } return NULL; /* not found */ } int ipdn_bound_var(int *v, int dflt, int lo, int hi, const char *msg) { int oldv = *v; const char *op = NULL; if (dflt < lo) dflt = lo; if (dflt > hi) dflt = hi; if (oldv < lo) { *v = dflt; op = "Bump"; } else if (oldv > hi) { *v = hi; op = "Clamp"; } else return *v; if (op && msg) printf("%s %s to %d (was %d)\n", op, msg, *v, oldv); return *v; } /*---- flow_id mask, hash and compare functions ---*/ /* * The flow_id includes the 5-tuple, the queue/pipe number * which we store in the extra area in host order, * and for ipv6 also the flow_id6. * XXX see if we want the tos byte (can store in 'flags') */ static struct ipfw_flow_id * flow_id_mask(struct ipfw_flow_id *mask, struct ipfw_flow_id *id) { int is_v6 = IS_IP6_FLOW_ID(id); id->dst_port &= mask->dst_port; id->src_port &= mask->src_port; id->proto &= mask->proto; id->extra &= mask->extra; if (is_v6) { APPLY_MASK(&id->dst_ip6, &mask->dst_ip6); APPLY_MASK(&id->src_ip6, &mask->src_ip6); id->flow_id6 &= mask->flow_id6; } else { id->dst_ip &= mask->dst_ip; id->src_ip &= mask->src_ip; } return id; } /* computes an OR of two masks, result in dst and also returned */ static struct ipfw_flow_id * flow_id_or(struct ipfw_flow_id *src, struct ipfw_flow_id *dst) { int is_v6 = IS_IP6_FLOW_ID(dst); dst->dst_port |= src->dst_port; dst->src_port |= src->src_port; dst->proto |= src->proto; dst->extra |= src->extra; if (is_v6) { #define OR_MASK(_d, _s) \ (_d)->__u6_addr.__u6_addr32[0] |= (_s)->__u6_addr.__u6_addr32[0]; \ (_d)->__u6_addr.__u6_addr32[1] |= (_s)->__u6_addr.__u6_addr32[1]; \ (_d)->__u6_addr.__u6_addr32[2] |= (_s)->__u6_addr.__u6_addr32[2]; \ (_d)->__u6_addr.__u6_addr32[3] |= (_s)->__u6_addr.__u6_addr32[3]; OR_MASK(&dst->dst_ip6, &src->dst_ip6); OR_MASK(&dst->src_ip6, &src->src_ip6); #undef OR_MASK dst->flow_id6 |= src->flow_id6; } else { dst->dst_ip |= src->dst_ip; dst->src_ip |= src->src_ip; } return dst; } static int nonzero_mask(struct ipfw_flow_id *m) { if (m->dst_port || m->src_port || m->proto || m->extra) return 1; if (IS_IP6_FLOW_ID(m)) { return m->dst_ip6.__u6_addr.__u6_addr32[0] || m->dst_ip6.__u6_addr.__u6_addr32[1] || m->dst_ip6.__u6_addr.__u6_addr32[2] || m->dst_ip6.__u6_addr.__u6_addr32[3] || m->src_ip6.__u6_addr.__u6_addr32[0] || m->src_ip6.__u6_addr.__u6_addr32[1] || m->src_ip6.__u6_addr.__u6_addr32[2] || m->src_ip6.__u6_addr.__u6_addr32[3] || m->flow_id6; } else { return m->dst_ip || m->src_ip; } } /* XXX we may want a better hash function */ static uint32_t flow_id_hash(struct ipfw_flow_id *id) { uint32_t i; if (IS_IP6_FLOW_ID(id)) { uint32_t *d = (uint32_t *)&id->dst_ip6; uint32_t *s = (uint32_t *)&id->src_ip6; i = (d[0] ) ^ (d[1]) ^ (d[2] ) ^ (d[3]) ^ (d[0] >> 15) ^ (d[1] >> 15) ^ (d[2] >> 15) ^ (d[3] >> 15) ^ (s[0] << 1) ^ (s[1] << 1) ^ (s[2] << 1) ^ (s[3] << 1) ^ (s[0] << 16) ^ (s[1] << 16) ^ (s[2] << 16) ^ (s[3] << 16) ^ (id->dst_port << 1) ^ (id->src_port) ^ (id->extra) ^ (id->proto ) ^ (id->flow_id6); } else { i = (id->dst_ip) ^ (id->dst_ip >> 15) ^ (id->src_ip << 1) ^ (id->src_ip >> 16) ^ (id->extra) ^ (id->dst_port << 1) ^ (id->src_port) ^ (id->proto); } return i; } /* Like bcmp, returns 0 if ids match, 1 otherwise. */ static int flow_id_cmp(struct ipfw_flow_id *id1, struct ipfw_flow_id *id2) { int is_v6 = IS_IP6_FLOW_ID(id1); if (!is_v6) { if (IS_IP6_FLOW_ID(id2)) return 1; /* different address families */ return (id1->dst_ip == id2->dst_ip && id1->src_ip == id2->src_ip && id1->dst_port == id2->dst_port && id1->src_port == id2->src_port && id1->proto == id2->proto && id1->extra == id2->extra) ? 0 : 1; } /* the ipv6 case */ return ( !bcmp(&id1->dst_ip6,&id2->dst_ip6, sizeof(id1->dst_ip6)) && !bcmp(&id1->src_ip6,&id2->src_ip6, sizeof(id1->src_ip6)) && id1->dst_port == id2->dst_port && id1->src_port == id2->src_port && id1->proto == id2->proto && id1->extra == id2->extra && id1->flow_id6 == id2->flow_id6) ? 0 : 1; } /*--------- end of flow-id mask, hash and compare ---------*/ /*--- support functions for the qht hashtable ---- * Entries are hashed by flow-id */ static uint32_t q_hash(uintptr_t key, int flags, void *arg) { /* compute the hash slot from the flow id */ struct ipfw_flow_id *id = (flags & DNHT_KEY_IS_OBJ) ? &((struct dn_queue *)key)->ni.fid : (struct ipfw_flow_id *)key; return flow_id_hash(id); } static int q_match(void *obj, uintptr_t key, int flags, void *arg) { struct dn_queue *o = (struct dn_queue *)obj; struct ipfw_flow_id *id2; if (flags & DNHT_KEY_IS_OBJ) { /* compare pointers */ id2 = &((struct dn_queue *)key)->ni.fid; } else { id2 = (struct ipfw_flow_id *)key; } return (0 == flow_id_cmp(&o->ni.fid, id2)); } /* * create a new queue instance for the given 'key'. */ static void * q_new(uintptr_t key, int flags, void *arg) { struct dn_queue *q, *template = arg; struct dn_fsk *fs = template->fs; int size = sizeof(*q) + fs->sched->fp->q_datalen; q = malloc(size, M_DUMMYNET, M_NOWAIT | M_ZERO); if (q == NULL) { D("no memory for new queue"); return NULL; } set_oid(&q->ni.oid, DN_QUEUE, size); if (fs->fs.flags & DN_QHT_HASH) q->ni.fid = *(struct ipfw_flow_id *)key; q->fs = fs; q->_si = template->_si; q->_si->q_count++; if (fs->sched->fp->new_queue) fs->sched->fp->new_queue(q); dn_cfg.queue_count++; return q; } /* * Notify schedulers that a queue is going away. * If (flags & DN_DESTROY), also free the packets. * The version for callbacks is called q_delete_cb(). */ static void dn_delete_queue(struct dn_queue *q, int flags) { struct dn_fsk *fs = q->fs; // D("fs %p si %p\n", fs, q->_si); /* notify the parent scheduler that the queue is going away */ if (fs && fs->sched->fp->free_queue) fs->sched->fp->free_queue(q); q->_si->q_count--; q->_si = NULL; if (flags & DN_DESTROY) { if (q->mq.head) dn_free_pkts(q->mq.head); bzero(q, sizeof(*q)); // safety free(q, M_DUMMYNET); dn_cfg.queue_count--; } } static int q_delete_cb(void *q, void *arg) { int flags = (int)(uintptr_t)arg; dn_delete_queue(q, flags); return (flags & DN_DESTROY) ? DNHT_SCAN_DEL : 0; } /* * calls dn_delete_queue/q_delete_cb on all queues, * which notifies the parent scheduler and possibly drains packets. * flags & DN_DESTROY: drains queues and destroy qht; */ static void qht_delete(struct dn_fsk *fs, int flags) { ND("fs %d start flags %d qht %p", fs->fs.fs_nr, flags, fs->qht); if (!fs->qht) return; if (fs->fs.flags & DN_QHT_HASH) { dn_ht_scan(fs->qht, q_delete_cb, (void *)(uintptr_t)flags); if (flags & DN_DESTROY) { dn_ht_free(fs->qht, 0); fs->qht = NULL; } } else { dn_delete_queue((struct dn_queue *)(fs->qht), flags); if (flags & DN_DESTROY) fs->qht = NULL; } } /* * Find and possibly create the queue for a MULTIQUEUE scheduler. * We never call it for !MULTIQUEUE (the queue is in the sch_inst). */ struct dn_queue * ipdn_q_find(struct dn_fsk *fs, struct dn_sch_inst *si, struct ipfw_flow_id *id) { struct dn_queue template; template._si = si; template.fs = fs; if (fs->fs.flags & DN_QHT_HASH) { struct ipfw_flow_id masked_id; if (fs->qht == NULL) { fs->qht = dn_ht_init(NULL, fs->fs.buckets, offsetof(struct dn_queue, q_next), q_hash, q_match, q_new); if (fs->qht == NULL) return NULL; } masked_id = *id; flow_id_mask(&fs->fsk_mask, &masked_id); return dn_ht_find(fs->qht, (uintptr_t)&masked_id, DNHT_INSERT, &template); } else { if (fs->qht == NULL) fs->qht = q_new(0, 0, &template); return (struct dn_queue *)fs->qht; } } /*--- end of queue hash table ---*/ /*--- support functions for the sch_inst hashtable ---- * * These are hashed by flow-id */ static uint32_t si_hash(uintptr_t key, int flags, void *arg) { /* compute the hash slot from the flow id */ struct ipfw_flow_id *id = (flags & DNHT_KEY_IS_OBJ) ? &((struct dn_sch_inst *)key)->ni.fid : (struct ipfw_flow_id *)key; return flow_id_hash(id); } static int si_match(void *obj, uintptr_t key, int flags, void *arg) { struct dn_sch_inst *o = obj; struct ipfw_flow_id *id2; id2 = (flags & DNHT_KEY_IS_OBJ) ? &((struct dn_sch_inst *)key)->ni.fid : (struct ipfw_flow_id *)key; return flow_id_cmp(&o->ni.fid, id2) == 0; } /* * create a new instance for the given 'key' * Allocate memory for instance, delay line and scheduler private data. */ static void * si_new(uintptr_t key, int flags, void *arg) { struct dn_schk *s = arg; struct dn_sch_inst *si; int l = sizeof(*si) + s->fp->si_datalen; si = malloc(l, M_DUMMYNET, M_NOWAIT | M_ZERO); if (si == NULL) goto error; /* Set length only for the part passed up to userland. */ set_oid(&si->ni.oid, DN_SCH_I, sizeof(struct dn_flow)); set_oid(&(si->dline.oid), DN_DELAY_LINE, sizeof(struct delay_line)); /* mark si and dline as outside the event queue */ si->ni.oid.id = si->dline.oid.id = -1; si->sched = s; si->dline.si = si; if (s->fp->new_sched && s->fp->new_sched(si)) { D("new_sched error"); goto error; } if (s->sch.flags & DN_HAVE_MASK) si->ni.fid = *(struct ipfw_flow_id *)key; dn_cfg.si_count++; return si; error: if (si) { bzero(si, sizeof(*si)); // safety free(si, M_DUMMYNET); } return NULL; } /* * Callback from siht to delete all scheduler instances. Remove * si and delay line from the system heap, destroy all queues. * We assume that all flowset have been notified and do not * point to us anymore. */ static int si_destroy(void *_si, void *arg) { struct dn_sch_inst *si = _si; struct dn_schk *s = si->sched; struct delay_line *dl = &si->dline; if (dl->oid.subtype) /* remove delay line from event heap */ heap_extract(&dn_cfg.evheap, dl); dn_free_pkts(dl->mq.head); /* drain delay line */ if (si->kflags & DN_ACTIVE) /* remove si from event heap */ heap_extract(&dn_cfg.evheap, si); if (s->fp->free_sched) s->fp->free_sched(si); bzero(si, sizeof(*si)); /* safety */ free(si, M_DUMMYNET); dn_cfg.si_count--; return DNHT_SCAN_DEL; } /* * Find the scheduler instance for this packet. If we need to apply * a mask, do on a local copy of the flow_id to preserve the original. * Assume siht is always initialized if we have a mask. */ struct dn_sch_inst * ipdn_si_find(struct dn_schk *s, struct ipfw_flow_id *id) { if (s->sch.flags & DN_HAVE_MASK) { struct ipfw_flow_id id_t = *id; flow_id_mask(&s->sch.sched_mask, &id_t); return dn_ht_find(s->siht, (uintptr_t)&id_t, DNHT_INSERT, s); } if (!s->siht) s->siht = si_new(0, 0, s); return (struct dn_sch_inst *)s->siht; } /* callback to flush credit for the scheduler instance */ static int si_reset_credit(void *_si, void *arg) { struct dn_sch_inst *si = _si; struct dn_link *p = &si->sched->link; si->credit = p->burst + (dn_cfg.io_fast ? p->bandwidth : 0); return 0; } static void schk_reset_credit(struct dn_schk *s) { if (s->sch.flags & DN_HAVE_MASK) dn_ht_scan(s->siht, si_reset_credit, NULL); else if (s->siht) si_reset_credit(s->siht, NULL); } /*---- end of sch_inst hashtable ---------------------*/ /*------------------------------------------------------- * flowset hash (fshash) support. Entries are hashed by fs_nr. * New allocations are put in the fsunlinked list, from which * they are removed when they point to a specific scheduler. */ static uint32_t fsk_hash(uintptr_t key, int flags, void *arg) { uint32_t i = !(flags & DNHT_KEY_IS_OBJ) ? key : ((struct dn_fsk *)key)->fs.fs_nr; return ( (i>>8)^(i>>4)^i ); } static int fsk_match(void *obj, uintptr_t key, int flags, void *arg) { struct dn_fsk *fs = obj; int i = !(flags & DNHT_KEY_IS_OBJ) ? key : ((struct dn_fsk *)key)->fs.fs_nr; return (fs->fs.fs_nr == i); } static void * fsk_new(uintptr_t key, int flags, void *arg) { struct dn_fsk *fs; fs = malloc(sizeof(*fs), M_DUMMYNET, M_NOWAIT | M_ZERO); if (fs) { set_oid(&fs->fs.oid, DN_FS, sizeof(fs->fs)); dn_cfg.fsk_count++; fs->drain_bucket = 0; SLIST_INSERT_HEAD(&dn_cfg.fsu, fs, sch_chain); } return fs; } /* * detach flowset from its current scheduler. Flags as follows: * DN_DETACH removes from the fsk_list * DN_DESTROY deletes individual queues * DN_DELETE_FS destroys the flowset (otherwise goes in unlinked). */ static void fsk_detach(struct dn_fsk *fs, int flags) { if (flags & DN_DELETE_FS) flags |= DN_DESTROY; ND("fs %d from sched %d flags %s %s %s", fs->fs.fs_nr, fs->fs.sched_nr, (flags & DN_DELETE_FS) ? "DEL_FS":"", (flags & DN_DESTROY) ? "DEL":"", (flags & DN_DETACH) ? "DET":""); if (flags & DN_DETACH) { /* detach from the list */ struct dn_fsk_head *h; h = fs->sched ? &fs->sched->fsk_list : &dn_cfg.fsu; SLIST_REMOVE(h, fs, dn_fsk, sch_chain); } /* Free the RED parameters, they will be recomputed on * subsequent attach if needed. */ if (fs->w_q_lookup) free(fs->w_q_lookup, M_DUMMYNET); fs->w_q_lookup = NULL; qht_delete(fs, flags); if (fs->sched && fs->sched->fp->free_fsk) fs->sched->fp->free_fsk(fs); fs->sched = NULL; if (flags & DN_DELETE_FS) { bzero(fs, sizeof(*fs)); /* safety */ free(fs, M_DUMMYNET); dn_cfg.fsk_count--; } else { SLIST_INSERT_HEAD(&dn_cfg.fsu, fs, sch_chain); } } /* * Detach or destroy all flowsets in a list. * flags specifies what to do: * DN_DESTROY: flush all queues * DN_DELETE_FS: DN_DESTROY + destroy flowset * DN_DELETE_FS implies DN_DESTROY */ static void fsk_detach_list(struct dn_fsk_head *h, int flags) { struct dn_fsk *fs; int n = 0; /* only for stats */ ND("head %p flags %x", h, flags); while ((fs = SLIST_FIRST(h))) { SLIST_REMOVE_HEAD(h, sch_chain); n++; fsk_detach(fs, flags); } ND("done %d flowsets", n); } /* * called on 'queue X delete' -- removes the flowset from fshash, * deletes all queues for the flowset, and removes the flowset. */ static int delete_fs(int i, int locked) { struct dn_fsk *fs; int err = 0; if (!locked) DN_BH_WLOCK(); fs = dn_ht_find(dn_cfg.fshash, i, DNHT_REMOVE, NULL); ND("fs %d found %p", i, fs); if (fs) { fsk_detach(fs, DN_DETACH | DN_DELETE_FS); err = 0; } else err = EINVAL; if (!locked) DN_BH_WUNLOCK(); return err; } /*----- end of flowset hashtable support -------------*/ /*------------------------------------------------------------ * Scheduler hash. When searching by index we pass sched_nr, * otherwise we pass struct dn_sch * which is the first field in * struct dn_schk so we can cast between the two. We use this trick * because in the create phase (but it should be fixed). */ static uint32_t schk_hash(uintptr_t key, int flags, void *_arg) { uint32_t i = !(flags & DNHT_KEY_IS_OBJ) ? key : ((struct dn_schk *)key)->sch.sched_nr; return ( (i>>8)^(i>>4)^i ); } static int schk_match(void *obj, uintptr_t key, int flags, void *_arg) { struct dn_schk *s = (struct dn_schk *)obj; int i = !(flags & DNHT_KEY_IS_OBJ) ? key : ((struct dn_schk *)key)->sch.sched_nr; return (s->sch.sched_nr == i); } /* * Create the entry and intialize with the sched hash if needed. * Leave s->fp unset so we can tell whether a dn_ht_find() returns * a new object or a previously existing one. */ static void * schk_new(uintptr_t key, int flags, void *arg) { struct schk_new_arg *a = arg; struct dn_schk *s; int l = sizeof(*s) +a->fp->schk_datalen; s = malloc(l, M_DUMMYNET, M_NOWAIT | M_ZERO); if (s == NULL) return NULL; set_oid(&s->link.oid, DN_LINK, sizeof(s->link)); s->sch = *a->sch; // copy initial values s->link.link_nr = s->sch.sched_nr; SLIST_INIT(&s->fsk_list); /* initialize the hash table or create the single instance */ s->fp = a->fp; /* si_new needs this */ s->drain_bucket = 0; if (s->sch.flags & DN_HAVE_MASK) { s->siht = dn_ht_init(NULL, s->sch.buckets, offsetof(struct dn_sch_inst, si_next), si_hash, si_match, si_new); if (s->siht == NULL) { free(s, M_DUMMYNET); return NULL; } } s->fp = NULL; /* mark as a new scheduler */ dn_cfg.schk_count++; return s; } /* * Callback for sched delete. Notify all attached flowsets to * detach from the scheduler, destroy the internal flowset, and * all instances. The scheduler goes away too. * arg is 0 (only detach flowsets and destroy instances) * DN_DESTROY (detach & delete queues, delete schk) * or DN_DELETE_FS (delete queues and flowsets, delete schk) */ static int schk_delete_cb(void *obj, void *arg) { struct dn_schk *s = obj; #if 0 int a = (int)arg; ND("sched %d arg %s%s", s->sch.sched_nr, a&DN_DESTROY ? "DEL ":"", a&DN_DELETE_FS ? "DEL_FS":""); #endif fsk_detach_list(&s->fsk_list, arg ? DN_DESTROY : 0); /* no more flowset pointing to us now */ if (s->sch.flags & DN_HAVE_MASK) { dn_ht_scan(s->siht, si_destroy, NULL); dn_ht_free(s->siht, 0); } else if (s->siht) si_destroy(s->siht, NULL); if (s->profile) { free(s->profile, M_DUMMYNET); s->profile = NULL; } s->siht = NULL; if (s->fp->destroy) s->fp->destroy(s); bzero(s, sizeof(*s)); // safety free(obj, M_DUMMYNET); dn_cfg.schk_count--; return DNHT_SCAN_DEL; } /* * called on a 'sched X delete' command. Deletes a single scheduler. * This is done by removing from the schedhash, unlinking all * flowsets and deleting their traffic. */ static int delete_schk(int i) { struct dn_schk *s; s = dn_ht_find(dn_cfg.schedhash, i, DNHT_REMOVE, NULL); ND("%d %p", i, s); if (!s) return EINVAL; delete_fs(i + DN_MAX_ID, 1); /* first delete internal fs */ /* then detach flowsets, delete traffic */ schk_delete_cb(s, (void*)(uintptr_t)DN_DESTROY); return 0; } /*--- end of schk hashtable support ---*/ static int copy_obj(char **start, char *end, void *_o, const char *msg, int i) { struct dn_id *o = _o; int have = end - *start; if (have < o->len || o->len == 0 || o->type == 0) { D("(WARN) type %d %s %d have %d need %d", o->type, msg, i, have, o->len); return 1; } ND("type %d %s %d len %d", o->type, msg, i, o->len); bcopy(_o, *start, o->len); if (o->type == DN_LINK) { /* Adjust burst parameter for link */ struct dn_link *l = (struct dn_link *)*start; l->burst = div64(l->burst, 8 * hz); l->delay = l->delay * 1000 / hz; } else if (o->type == DN_SCH) { /* Set id->id to the number of instances */ struct dn_schk *s = _o; struct dn_id *id = (struct dn_id *)(*start); id->id = (s->sch.flags & DN_HAVE_MASK) ? dn_ht_entries(s->siht) : (s->siht ? 1 : 0); } *start += o->len; return 0; } /* Specific function to copy a queue. * Copies only the user-visible part of a queue (which is in * a struct dn_flow), and sets len accordingly. */ static int copy_obj_q(char **start, char *end, void *_o, const char *msg, int i) { struct dn_id *o = _o; int have = end - *start; int len = sizeof(struct dn_flow); /* see above comment */ if (have < len || o->len == 0 || o->type != DN_QUEUE) { D("ERROR type %d %s %d have %d need %d", o->type, msg, i, have, len); return 1; } ND("type %d %s %d len %d", o->type, msg, i, len); bcopy(_o, *start, len); ((struct dn_id*)(*start))->len = len; *start += len; return 0; } static int copy_q_cb(void *obj, void *arg) { struct dn_queue *q = obj; struct copy_args *a = arg; struct dn_flow *ni = (struct dn_flow *)(*a->start); if (copy_obj_q(a->start, a->end, &q->ni, "queue", -1)) return DNHT_SCAN_END; ni->oid.type = DN_FLOW; /* override the DN_QUEUE */ ni->oid.id = si_hash((uintptr_t)&ni->fid, 0, NULL); return 0; } static int copy_q(struct copy_args *a, struct dn_fsk *fs, int flags) { if (!fs->qht) return 0; if (fs->fs.flags & DN_QHT_HASH) dn_ht_scan(fs->qht, copy_q_cb, a); else copy_q_cb(fs->qht, a); return 0; } /* * This routine only copies the initial part of a profile ? XXX */ static int copy_profile(struct copy_args *a, struct dn_profile *p) { int have = a->end - *a->start; /* XXX here we check for max length */ int profile_len = sizeof(struct dn_profile) - ED_MAX_SAMPLES_NO*sizeof(int); if (p == NULL) return 0; if (have < profile_len) { D("error have %d need %d", have, profile_len); return 1; } bcopy(p, *a->start, profile_len); ((struct dn_id *)(*a->start))->len = profile_len; *a->start += profile_len; return 0; } static int copy_flowset(struct copy_args *a, struct dn_fsk *fs, int flags) { struct dn_fs *ufs = (struct dn_fs *)(*a->start); if (!fs) return 0; ND("flowset %d", fs->fs.fs_nr); if (copy_obj(a->start, a->end, &fs->fs, "flowset", fs->fs.fs_nr)) return DNHT_SCAN_END; ufs->oid.id = (fs->fs.flags & DN_QHT_HASH) ? dn_ht_entries(fs->qht) : (fs->qht ? 1 : 0); if (flags) { /* copy queues */ copy_q(a, fs, 0); } return 0; } static int copy_si_cb(void *obj, void *arg) { struct dn_sch_inst *si = obj; struct copy_args *a = arg; struct dn_flow *ni = (struct dn_flow *)(*a->start); if (copy_obj(a->start, a->end, &si->ni, "inst", si->sched->sch.sched_nr)) return DNHT_SCAN_END; ni->oid.type = DN_FLOW; /* override the DN_SCH_I */ ni->oid.id = si_hash((uintptr_t)si, DNHT_KEY_IS_OBJ, NULL); return 0; } static int copy_si(struct copy_args *a, struct dn_schk *s, int flags) { if (s->sch.flags & DN_HAVE_MASK) dn_ht_scan(s->siht, copy_si_cb, a); else if (s->siht) copy_si_cb(s->siht, a); return 0; } /* * compute a list of children of a scheduler and copy up */ static int copy_fsk_list(struct copy_args *a, struct dn_schk *s, int flags) { struct dn_fsk *fs; struct dn_id *o; uint32_t *p; int n = 0, space = sizeof(*o); SLIST_FOREACH(fs, &s->fsk_list, sch_chain) { if (fs->fs.fs_nr < DN_MAX_ID) n++; } space += n * sizeof(uint32_t); DX(3, "sched %d has %d flowsets", s->sch.sched_nr, n); if (a->end - *(a->start) < space) return DNHT_SCAN_END; o = (struct dn_id *)(*(a->start)); o->len = space; *a->start += o->len; o->type = DN_TEXT; p = (uint32_t *)(o+1); SLIST_FOREACH(fs, &s->fsk_list, sch_chain) if (fs->fs.fs_nr < DN_MAX_ID) *p++ = fs->fs.fs_nr; return 0; } static int copy_data_helper(void *_o, void *_arg) { struct copy_args *a = _arg; uint32_t *r = a->extra->r; /* start of first range */ uint32_t *lim; /* first invalid pointer */ int n; lim = (uint32_t *)((char *)(a->extra) + a->extra->o.len); if (a->type == DN_LINK || a->type == DN_SCH) { /* pipe|sched show, we receive a dn_schk */ struct dn_schk *s = _o; n = s->sch.sched_nr; if (a->type == DN_SCH && n >= DN_MAX_ID) return 0; /* not a scheduler */ if (a->type == DN_LINK && n <= DN_MAX_ID) return 0; /* not a pipe */ /* see if the object is within one of our ranges */ for (;r < lim; r += 2) { if (n < r[0] || n > r[1]) continue; /* Found a valid entry, copy and we are done */ if (a->flags & DN_C_LINK) { if (copy_obj(a->start, a->end, &s->link, "link", n)) return DNHT_SCAN_END; if (copy_profile(a, s->profile)) return DNHT_SCAN_END; if (copy_flowset(a, s->fs, 0)) return DNHT_SCAN_END; } if (a->flags & DN_C_SCH) { if (copy_obj(a->start, a->end, &s->sch, "sched", n)) return DNHT_SCAN_END; /* list all attached flowsets */ if (copy_fsk_list(a, s, 0)) return DNHT_SCAN_END; } if (a->flags & DN_C_FLOW) copy_si(a, s, 0); break; } } else if (a->type == DN_FS) { /* queue show, skip internal flowsets */ struct dn_fsk *fs = _o; n = fs->fs.fs_nr; if (n >= DN_MAX_ID) return 0; /* see if the object is within one of our ranges */ for (;r < lim; r += 2) { if (n < r[0] || n > r[1]) continue; if (copy_flowset(a, fs, 0)) return DNHT_SCAN_END; copy_q(a, fs, 0); break; /* we are done */ } } return 0; } static inline struct dn_schk * locate_scheduler(int i) { return dn_ht_find(dn_cfg.schedhash, i, 0, NULL); } /* * red parameters are in fixed point arithmetic. */ static int config_red(struct dn_fsk *fs) { int64_t s, idle, weight, w0; int t, i; fs->w_q = fs->fs.w_q; fs->max_p = fs->fs.max_p; ND("called"); /* Doing stuff that was in userland */ i = fs->sched->link.bandwidth; s = (i <= 0) ? 0 : hz * dn_cfg.red_avg_pkt_size * 8 * SCALE(1) / i; idle = div64((s * 3) , fs->w_q); /* s, fs->w_q scaled; idle not scaled */ fs->lookup_step = div64(idle , dn_cfg.red_lookup_depth); /* fs->lookup_step not scaled, */ if (!fs->lookup_step) fs->lookup_step = 1; w0 = weight = SCALE(1) - fs->w_q; //fs->w_q scaled for (t = fs->lookup_step; t > 1; --t) weight = SCALE_MUL(weight, w0); fs->lookup_weight = (int)(weight); // scaled /* Now doing stuff that was in kerneland */ fs->min_th = SCALE(fs->fs.min_th); fs->max_th = SCALE(fs->fs.max_th); if (fs->fs.max_th == fs->fs.min_th) fs->c_1 = fs->max_p; else fs->c_1 = SCALE((int64_t)(fs->max_p)) / (fs->fs.max_th - fs->fs.min_th); fs->c_2 = SCALE_MUL(fs->c_1, SCALE(fs->fs.min_th)); if (fs->fs.flags & DN_IS_GENTLE_RED) { fs->c_3 = (SCALE(1) - fs->max_p) / fs->fs.max_th; fs->c_4 = SCALE(1) - 2 * fs->max_p; } /* If the lookup table already exist, free and create it again. */ if (fs->w_q_lookup) { free(fs->w_q_lookup, M_DUMMYNET); fs->w_q_lookup = NULL; } if (dn_cfg.red_lookup_depth == 0) { printf("\ndummynet: net.inet.ip.dummynet.red_lookup_depth" "must be > 0\n"); fs->fs.flags &= ~DN_IS_RED; fs->fs.flags &= ~DN_IS_GENTLE_RED; return (EINVAL); } fs->lookup_depth = dn_cfg.red_lookup_depth; fs->w_q_lookup = (u_int *)malloc(fs->lookup_depth * sizeof(int), M_DUMMYNET, M_NOWAIT); if (fs->w_q_lookup == NULL) { printf("dummynet: sorry, cannot allocate red lookup table\n"); fs->fs.flags &= ~DN_IS_RED; fs->fs.flags &= ~DN_IS_GENTLE_RED; return(ENOSPC); } /* Fill the lookup table with (1 - w_q)^x */ fs->w_q_lookup[0] = SCALE(1) - fs->w_q; for (i = 1; i < fs->lookup_depth; i++) fs->w_q_lookup[i] = SCALE_MUL(fs->w_q_lookup[i - 1], fs->lookup_weight); if (dn_cfg.red_avg_pkt_size < 1) dn_cfg.red_avg_pkt_size = 512; fs->avg_pkt_size = dn_cfg.red_avg_pkt_size; if (dn_cfg.red_max_pkt_size < 1) dn_cfg.red_max_pkt_size = 1500; fs->max_pkt_size = dn_cfg.red_max_pkt_size; ND("exit"); return 0; } /* Scan all flowset attached to this scheduler and update red */ static void update_red(struct dn_schk *s) { struct dn_fsk *fs; SLIST_FOREACH(fs, &s->fsk_list, sch_chain) { if (fs && (fs->fs.flags & DN_IS_RED)) config_red(fs); } } /* attach flowset to scheduler s, possibly requeue */ static void fsk_attach(struct dn_fsk *fs, struct dn_schk *s) { ND("remove fs %d from fsunlinked, link to sched %d", fs->fs.fs_nr, s->sch.sched_nr); SLIST_REMOVE(&dn_cfg.fsu, fs, dn_fsk, sch_chain); fs->sched = s; SLIST_INSERT_HEAD(&s->fsk_list, fs, sch_chain); if (s->fp->new_fsk) s->fp->new_fsk(fs); /* XXX compute fsk_mask */ fs->fsk_mask = fs->fs.flow_mask; if (fs->sched->sch.flags & DN_HAVE_MASK) flow_id_or(&fs->sched->sch.sched_mask, &fs->fsk_mask); if (fs->qht) { /* * we must drain qht according to the old * type, and reinsert according to the new one. * The requeue is complex -- in general we need to * reclassify every single packet. * For the time being, let's hope qht is never set * when we reach this point. */ D("XXX TODO requeue from fs %d to sch %d", fs->fs.fs_nr, s->sch.sched_nr); fs->qht = NULL; } /* set the new type for qht */ if (nonzero_mask(&fs->fsk_mask)) fs->fs.flags |= DN_QHT_HASH; else fs->fs.flags &= ~DN_QHT_HASH; /* XXX config_red() can fail... */ if (fs->fs.flags & DN_IS_RED) config_red(fs); } /* update all flowsets which may refer to this scheduler */ static void update_fs(struct dn_schk *s) { struct dn_fsk *fs, *tmp; SLIST_FOREACH_SAFE(fs, &dn_cfg.fsu, sch_chain, tmp) { if (s->sch.sched_nr != fs->fs.sched_nr) { D("fs %d for sch %d not %d still unlinked", fs->fs.fs_nr, fs->fs.sched_nr, s->sch.sched_nr); continue; } fsk_attach(fs, s); } } /* * Configuration -- to preserve backward compatibility we use * the following scheme (N is 65536) * NUMBER SCHED LINK FLOWSET * 1 .. N-1 (1)WFQ (2)WFQ (3)queue * N+1 .. 2N-1 (4)FIFO (5)FIFO (6)FIFO for sched 1..N-1 * 2N+1 .. 3N-1 -- -- (7)FIFO for sched N+1..2N-1 * * "pipe i config" configures #1, #2 and #3 * "sched i config" configures #1 and possibly #6 * "queue i config" configures #3 * #1 is configured with 'pipe i config' or 'sched i config' * #2 is configured with 'pipe i config', and created if not * existing with 'sched i config' * #3 is configured with 'queue i config' * #4 is automatically configured after #1, can only be FIFO * #5 is automatically configured after #2 * #6 is automatically created when #1 is !MULTIQUEUE, * and can be updated. * #7 is automatically configured after #2 */ /* * configure a link (and its FIFO instance) */ static int config_link(struct dn_link *p, struct dn_id *arg) { int i; if (p->oid.len != sizeof(*p)) { D("invalid pipe len %d", p->oid.len); return EINVAL; } i = p->link_nr; if (i <= 0 || i >= DN_MAX_ID) return EINVAL; /* * The config program passes parameters as follows: * bw = bits/second (0 means no limits), * delay = ms, must be translated into ticks. * qsize = slots/bytes * burst ??? */ p->delay = (p->delay * hz) / 1000; /* Scale burst size: bytes -> bits * hz */ p->burst *= 8 * hz; DN_BH_WLOCK(); /* do it twice, base link and FIFO link */ for (; i < 2*DN_MAX_ID; i += DN_MAX_ID) { struct dn_schk *s = locate_scheduler(i); if (s == NULL) { DN_BH_WUNLOCK(); D("sched %d not found", i); return EINVAL; } /* remove profile if exists */ if (s->profile) { free(s->profile, M_DUMMYNET); s->profile = NULL; } /* copy all parameters */ s->link.oid = p->oid; s->link.link_nr = i; s->link.delay = p->delay; if (s->link.bandwidth != p->bandwidth) { /* XXX bandwidth changes, need to update red params */ s->link.bandwidth = p->bandwidth; update_red(s); } s->link.burst = p->burst; schk_reset_credit(s); } dn_cfg.id++; DN_BH_WUNLOCK(); return 0; } /* * configure a flowset. Can be called from inside with locked=1, */ static struct dn_fsk * config_fs(struct dn_fs *nfs, struct dn_id *arg, int locked) { int i; struct dn_fsk *fs; if (nfs->oid.len != sizeof(*nfs)) { D("invalid flowset len %d", nfs->oid.len); return NULL; } i = nfs->fs_nr; if (i <= 0 || i >= 3*DN_MAX_ID) return NULL; ND("flowset %d", i); /* XXX other sanity checks */ if (nfs->flags & DN_QSIZE_BYTES) { ipdn_bound_var(&nfs->qsize, 16384, 1500, dn_cfg.byte_limit, NULL); // "queue byte size"); } else { ipdn_bound_var(&nfs->qsize, 50, 1, dn_cfg.slot_limit, NULL); // "queue slot size"); } if (nfs->flags & DN_HAVE_MASK) { /* make sure we have some buckets */ ipdn_bound_var((int *)&nfs->buckets, dn_cfg.hash_size, 1, dn_cfg.max_hash_size, "flowset buckets"); } else { nfs->buckets = 1; /* we only need 1 */ } if (!locked) DN_BH_WLOCK(); do { /* exit with break when done */ struct dn_schk *s; int flags = nfs->sched_nr ? DNHT_INSERT : 0; int j; int oldc = dn_cfg.fsk_count; fs = dn_ht_find(dn_cfg.fshash, i, flags, NULL); if (fs == NULL) { D("missing sched for flowset %d", i); break; } /* grab some defaults from the existing one */ if (nfs->sched_nr == 0) /* reuse */ nfs->sched_nr = fs->fs.sched_nr; for (j = 0; j < sizeof(nfs->par)/sizeof(nfs->par[0]); j++) { if (nfs->par[j] == -1) /* reuse */ nfs->par[j] = fs->fs.par[j]; } if (bcmp(&fs->fs, nfs, sizeof(*nfs)) == 0) { ND("flowset %d unchanged", i); break; /* no change, nothing to do */ } if (oldc != dn_cfg.fsk_count) /* new item */ dn_cfg.id++; s = locate_scheduler(nfs->sched_nr); /* detach from old scheduler if needed, preserving * queues if we need to reattach. Then update the * configuration, and possibly attach to the new sched. */ DX(2, "fs %d changed sched %d@%p to %d@%p", fs->fs.fs_nr, fs->fs.sched_nr, fs->sched, nfs->sched_nr, s); if (fs->sched) { int flags = s ? DN_DETACH : (DN_DETACH | DN_DESTROY); flags |= DN_DESTROY; /* XXX temporary */ fsk_detach(fs, flags); } fs->fs = *nfs; /* copy configuration */ if (s != NULL) fsk_attach(fs, s); } while (0); if (!locked) DN_BH_WUNLOCK(); return fs; } /* * config/reconfig a scheduler and its FIFO variant. * For !MULTIQUEUE schedulers, also set up the flowset. * * On reconfigurations (detected because s->fp is set), * detach existing flowsets preserving traffic, preserve link, * and delete the old scheduler creating a new one. */ static int config_sched(struct dn_sch *_nsch, struct dn_id *arg) { struct dn_schk *s; struct schk_new_arg a; /* argument for schk_new */ int i; struct dn_link p; /* copy of oldlink */ struct dn_profile *pf = NULL; /* copy of old link profile */ /* Used to preserv mask parameter */ struct ipfw_flow_id new_mask; int new_buckets = 0; int new_flags = 0; int pipe_cmd; int err = ENOMEM; a.sch = _nsch; if (a.sch->oid.len != sizeof(*a.sch)) { D("bad sched len %d", a.sch->oid.len); return EINVAL; } i = a.sch->sched_nr; if (i <= 0 || i >= DN_MAX_ID) return EINVAL; /* make sure we have some buckets */ if (a.sch->flags & DN_HAVE_MASK) ipdn_bound_var((int *)&a.sch->buckets, dn_cfg.hash_size, 1, dn_cfg.max_hash_size, "sched buckets"); /* XXX other sanity checks */ bzero(&p, sizeof(p)); pipe_cmd = a.sch->flags & DN_PIPE_CMD; a.sch->flags &= ~DN_PIPE_CMD; //XXX do it even if is not set? if (pipe_cmd) { /* Copy mask parameter */ new_mask = a.sch->sched_mask; new_buckets = a.sch->buckets; new_flags = a.sch->flags; } DN_BH_WLOCK(); again: /* run twice, for wfq and fifo */ /* * lookup the type. If not supplied, use the previous one * or default to WF2Q+. Otherwise, return an error. */ dn_cfg.id++; a.fp = find_sched_type(a.sch->oid.subtype, a.sch->name); if (a.fp != NULL) { /* found. Lookup or create entry */ s = dn_ht_find(dn_cfg.schedhash, i, DNHT_INSERT, &a); } else if (a.sch->oid.subtype == 0 && !a.sch->name[0]) { /* No type. search existing s* or retry with WF2Q+ */ s = dn_ht_find(dn_cfg.schedhash, i, 0, &a); if (s != NULL) { a.fp = s->fp; /* Scheduler exists, skip to FIFO scheduler * if command was pipe config... */ if (pipe_cmd) goto next; } else { /* New scheduler, create a wf2q+ with no mask * if command was pipe config... */ if (pipe_cmd) { /* clear mask parameter */ bzero(&a.sch->sched_mask, sizeof(new_mask)); a.sch->buckets = 0; a.sch->flags &= ~DN_HAVE_MASK; } a.sch->oid.subtype = DN_SCHED_WF2QP; goto again; } } else { D("invalid scheduler type %d %s", a.sch->oid.subtype, a.sch->name); err = EINVAL; goto error; } /* normalize name and subtype */ a.sch->oid.subtype = a.fp->type; bzero(a.sch->name, sizeof(a.sch->name)); strlcpy(a.sch->name, a.fp->name, sizeof(a.sch->name)); if (s == NULL) { D("cannot allocate scheduler %d", i); goto error; } /* restore existing link if any */ if (p.link_nr) { s->link = p; if (!pf || pf->link_nr != p.link_nr) { /* no saved value */ s->profile = NULL; /* XXX maybe not needed */ } else { s->profile = malloc(sizeof(struct dn_profile), M_DUMMYNET, M_NOWAIT | M_ZERO); if (s->profile == NULL) { D("cannot allocate profile"); goto error; //XXX } bcopy(pf, s->profile, sizeof(*pf)); } } p.link_nr = 0; if (s->fp == NULL) { DX(2, "sched %d new type %s", i, a.fp->name); } else if (s->fp != a.fp || bcmp(a.sch, &s->sch, sizeof(*a.sch)) ) { /* already existing. */ DX(2, "sched %d type changed from %s to %s", i, s->fp->name, a.fp->name); DX(4, " type/sub %d/%d -> %d/%d", s->sch.oid.type, s->sch.oid.subtype, a.sch->oid.type, a.sch->oid.subtype); if (s->link.link_nr == 0) D("XXX WARNING link 0 for sched %d", i); p = s->link; /* preserve link */ if (s->profile) {/* preserve profile */ if (!pf) pf = malloc(sizeof(*pf), M_DUMMYNET, M_NOWAIT | M_ZERO); if (pf) /* XXX should issue a warning otherwise */ bcopy(s->profile, pf, sizeof(*pf)); } /* remove from the hash */ dn_ht_find(dn_cfg.schedhash, i, DNHT_REMOVE, NULL); /* Detach flowsets, preserve queues. */ // schk_delete_cb(s, NULL); // XXX temporarily, kill queues schk_delete_cb(s, (void *)DN_DESTROY); goto again; } else { DX(4, "sched %d unchanged type %s", i, a.fp->name); } /* complete initialization */ s->sch = *a.sch; s->fp = a.fp; s->cfg = arg; // XXX schk_reset_credit(s); /* create the internal flowset if needed, * trying to reuse existing ones if available */ if (!(s->fp->flags & DN_MULTIQUEUE) && !s->fs) { s->fs = dn_ht_find(dn_cfg.fshash, i, 0, NULL); if (!s->fs) { struct dn_fs fs; bzero(&fs, sizeof(fs)); set_oid(&fs.oid, DN_FS, sizeof(fs)); fs.fs_nr = i + DN_MAX_ID; fs.sched_nr = i; s->fs = config_fs(&fs, NULL, 1 /* locked */); } if (!s->fs) { schk_delete_cb(s, (void *)DN_DESTROY); D("error creating internal fs for %d", i); goto error; } } /* call init function after the flowset is created */ if (s->fp->config) s->fp->config(s); update_fs(s); next: if (i < DN_MAX_ID) { /* now configure the FIFO instance */ i += DN_MAX_ID; if (pipe_cmd) { /* Restore mask parameter for FIFO */ a.sch->sched_mask = new_mask; a.sch->buckets = new_buckets; a.sch->flags = new_flags; } else { /* sched config shouldn't modify the FIFO scheduler */ if (dn_ht_find(dn_cfg.schedhash, i, 0, &a) != NULL) { /* FIFO already exist, don't touch it */ err = 0; /* and this is not an error */ goto error; } } a.sch->sched_nr = i; a.sch->oid.subtype = DN_SCHED_FIFO; bzero(a.sch->name, sizeof(a.sch->name)); goto again; } err = 0; error: DN_BH_WUNLOCK(); if (pf) free(pf, M_DUMMYNET); return err; } /* * attach a profile to a link */ static int config_profile(struct dn_profile *pf, struct dn_id *arg) { struct dn_schk *s; int i, olen, err = 0; if (pf->oid.len < sizeof(*pf)) { D("short profile len %d", pf->oid.len); return EINVAL; } i = pf->link_nr; if (i <= 0 || i >= DN_MAX_ID) return EINVAL; /* XXX other sanity checks */ DN_BH_WLOCK(); for (; i < 2*DN_MAX_ID; i += DN_MAX_ID) { s = locate_scheduler(i); if (s == NULL) { err = EINVAL; break; } dn_cfg.id++; /* * If we had a profile and the new one does not fit, * or it is deleted, then we need to free memory. */ if (s->profile && (pf->samples_no == 0 || s->profile->oid.len < pf->oid.len)) { free(s->profile, M_DUMMYNET); s->profile = NULL; } if (pf->samples_no == 0) continue; /* * new profile, possibly allocate memory * and copy data. */ if (s->profile == NULL) s->profile = malloc(pf->oid.len, M_DUMMYNET, M_NOWAIT | M_ZERO); if (s->profile == NULL) { D("no memory for profile %d", i); err = ENOMEM; break; } /* preserve larger length XXX double check */ olen = s->profile->oid.len; if (olen < pf->oid.len) olen = pf->oid.len; bcopy(pf, s->profile, pf->oid.len); s->profile->oid.len = olen; } DN_BH_WUNLOCK(); return err; } /* * Delete all objects: */ static void dummynet_flush(void) { /* delete all schedulers and related links/queues/flowsets */ dn_ht_scan(dn_cfg.schedhash, schk_delete_cb, (void *)(uintptr_t)DN_DELETE_FS); /* delete all remaining (unlinked) flowsets */ DX(4, "still %d unlinked fs", dn_cfg.fsk_count); dn_ht_free(dn_cfg.fshash, DNHT_REMOVE); fsk_detach_list(&dn_cfg.fsu, DN_DELETE_FS); /* Reinitialize system heap... */ heap_init(&dn_cfg.evheap, 16, offsetof(struct dn_id, id)); } /* * Main handler for configuration. We are guaranteed to be called * with an oid which is at least a dn_id. * - the first object is the command (config, delete, flush, ...) * - config_link must be issued after the corresponding config_sched * - parameters (DN_TXT) for an object must preceed the object * processed on a config_sched. */ int do_config(void *p, int l) { struct dn_id *next, *o; int err = 0, err2 = 0; struct dn_id *arg = NULL; uintptr_t *a; o = p; if (o->id != DN_API_VERSION) { D("invalid api version got %d need %d", o->id, DN_API_VERSION); return EINVAL; } for (; l >= sizeof(*o); o = next) { struct dn_id *prev = arg; if (o->len < sizeof(*o) || l < o->len) { D("bad len o->len %d len %d", o->len, l); err = EINVAL; break; } l -= o->len; next = (struct dn_id *)((char *)o + o->len); err = 0; switch (o->type) { default: D("cmd %d not implemented", o->type); break; #ifdef EMULATE_SYSCTL /* sysctl emulation. * if we recognize the command, jump to the correct * handler and return */ case DN_SYSCTL_SET: err = kesysctl_emu_set(p, l); return err; #endif case DN_CMD_CONFIG: /* simply a header */ break; case DN_CMD_DELETE: /* the argument is in the first uintptr_t after o */ a = (uintptr_t *)(o+1); if (o->len < sizeof(*o) + sizeof(*a)) { err = EINVAL; break; } switch (o->subtype) { case DN_LINK: /* delete base and derived schedulers */ DN_BH_WLOCK(); err = delete_schk(*a); err2 = delete_schk(*a + DN_MAX_ID); DN_BH_WUNLOCK(); if (!err) err = err2; break; default: D("invalid delete type %d", o->subtype); err = EINVAL; break; case DN_FS: err = (*a <1 || *a >= DN_MAX_ID) ? EINVAL : delete_fs(*a, 0) ; break; } break; case DN_CMD_FLUSH: DN_BH_WLOCK(); dummynet_flush(); DN_BH_WUNLOCK(); break; case DN_TEXT: /* store argument the next block */ prev = NULL; arg = o; break; case DN_LINK: err = config_link((struct dn_link *)o, arg); break; case DN_PROFILE: err = config_profile((struct dn_profile *)o, arg); break; case DN_SCH: err = config_sched((struct dn_sch *)o, arg); break; case DN_FS: err = (NULL==config_fs((struct dn_fs *)o, arg, 0)); break; } if (prev) arg = NULL; if (err != 0) break; } return err; } static int compute_space(struct dn_id *cmd, struct copy_args *a) { int x = 0, need = 0; int profile_size = sizeof(struct dn_profile) - ED_MAX_SAMPLES_NO*sizeof(int); /* NOTE about compute space: * NP = dn_cfg.schk_count * NSI = dn_cfg.si_count * NF = dn_cfg.fsk_count * NQ = dn_cfg.queue_count * - ipfw pipe show * (NP/2)*(dn_link + dn_sch + dn_id + dn_fs) only half scheduler * link, scheduler template, flowset * integrated in scheduler and header * for flowset list * (NSI)*(dn_flow) all scheduler instance (includes * the queue instance) * - ipfw sched show * (NP/2)*(dn_link + dn_sch + dn_id + dn_fs) only half scheduler * link, scheduler template, flowset * integrated in scheduler and header * for flowset list * (NSI * dn_flow) all scheduler instances * (NF * sizeof(uint_32)) space for flowset list linked to scheduler * (NQ * dn_queue) all queue [XXXfor now not listed] * - ipfw queue show * (NF * dn_fs) all flowset * (NQ * dn_queue) all queues */ switch (cmd->subtype) { default: return -1; /* XXX where do LINK and SCH differ ? */ /* 'ipfw sched show' could list all queues associated to * a scheduler. This feature for now is disabled */ case DN_LINK: /* pipe show */ x = DN_C_LINK | DN_C_SCH | DN_C_FLOW; need += dn_cfg.schk_count * (sizeof(struct dn_fs) + profile_size) / 2; need += dn_cfg.fsk_count * sizeof(uint32_t); break; case DN_SCH: /* sched show */ need += dn_cfg.schk_count * (sizeof(struct dn_fs) + profile_size) / 2; need += dn_cfg.fsk_count * sizeof(uint32_t); x = DN_C_SCH | DN_C_LINK | DN_C_FLOW; break; case DN_FS: /* queue show */ x = DN_C_FS | DN_C_QUEUE; break; case DN_GET_COMPAT: /* compatibility mode */ need = dn_compat_calc_size(); break; } a->flags = x; if (x & DN_C_SCH) { need += dn_cfg.schk_count * sizeof(struct dn_sch) / 2; /* NOT also, each fs might be attached to a sched */ need += dn_cfg.schk_count * sizeof(struct dn_id) / 2; } if (x & DN_C_FS) need += dn_cfg.fsk_count * sizeof(struct dn_fs); if (x & DN_C_LINK) { need += dn_cfg.schk_count * sizeof(struct dn_link) / 2; } /* * When exporting a queue to userland, only pass up the * struct dn_flow, which is the only visible part. */ if (x & DN_C_QUEUE) need += dn_cfg.queue_count * sizeof(struct dn_flow); if (x & DN_C_FLOW) need += dn_cfg.si_count * (sizeof(struct dn_flow)); return need; } /* * If compat != NULL dummynet_get is called in compatibility mode. * *compat will be the pointer to the buffer to pass to ipfw */ int dummynet_get(struct sockopt *sopt, void **compat) { int have, i, need, error; char *start = NULL, *buf; size_t sopt_valsize; struct dn_id *cmd; struct copy_args a; struct copy_range r; int l = sizeof(struct dn_id); bzero(&a, sizeof(a)); bzero(&r, sizeof(r)); /* save and restore original sopt_valsize around copyin */ sopt_valsize = sopt->sopt_valsize; cmd = &r.o; if (!compat) { /* copy at least an oid, and possibly a full object */ error = sooptcopyin(sopt, cmd, sizeof(r), sizeof(*cmd)); sopt->sopt_valsize = sopt_valsize; if (error) goto done; l = cmd->len; #ifdef EMULATE_SYSCTL /* sysctl emulation. */ if (cmd->type == DN_SYSCTL_GET) return kesysctl_emu_get(sopt); #endif if (l > sizeof(r)) { /* request larger than default, allocate buffer */ cmd = malloc(l, M_DUMMYNET, M_WAITOK); error = sooptcopyin(sopt, cmd, l, l); sopt->sopt_valsize = sopt_valsize; if (error) goto done; } } else { /* compatibility */ error = 0; cmd->type = DN_CMD_GET; cmd->len = sizeof(struct dn_id); cmd->subtype = DN_GET_COMPAT; // cmd->id = sopt_valsize; D("compatibility mode"); } a.extra = (struct copy_range *)cmd; if (cmd->len == sizeof(*cmd)) { /* no range, create a default */ uint32_t *rp = (uint32_t *)(cmd + 1); cmd->len += 2* sizeof(uint32_t); rp[0] = 1; rp[1] = DN_MAX_ID - 1; if (cmd->subtype == DN_LINK) { rp[0] += DN_MAX_ID; rp[1] += DN_MAX_ID; } } /* Count space (under lock) and allocate (outside lock). * Exit with lock held if we manage to get enough buffer. * Try a few times then give up. */ for (have = 0, i = 0; i < 10; i++) { DN_BH_WLOCK(); need = compute_space(cmd, &a); /* if there is a range, ignore value from compute_space() */ if (l > sizeof(*cmd)) need = sopt_valsize - sizeof(*cmd); if (need < 0) { DN_BH_WUNLOCK(); error = EINVAL; goto done; } need += sizeof(*cmd); cmd->id = need; if (have >= need) break; DN_BH_WUNLOCK(); if (start) free(start, M_DUMMYNET); start = NULL; if (need > sopt_valsize) break; have = need; start = malloc(have, M_DUMMYNET, M_WAITOK | M_ZERO); } if (start == NULL) { if (compat) { *compat = NULL; error = 1; // XXX } else { error = sooptcopyout(sopt, cmd, sizeof(*cmd)); } goto done; } ND("have %d:%d sched %d, %d:%d links %d, %d:%d flowsets %d, " "%d:%d si %d, %d:%d queues %d", dn_cfg.schk_count, sizeof(struct dn_sch), DN_SCH, dn_cfg.schk_count, sizeof(struct dn_link), DN_LINK, dn_cfg.fsk_count, sizeof(struct dn_fs), DN_FS, dn_cfg.si_count, sizeof(struct dn_flow), DN_SCH_I, dn_cfg.queue_count, sizeof(struct dn_queue), DN_QUEUE); sopt->sopt_valsize = sopt_valsize; a.type = cmd->subtype; if (compat == NULL) { bcopy(cmd, start, sizeof(*cmd)); ((struct dn_id*)(start))->len = sizeof(struct dn_id); buf = start + sizeof(*cmd); } else buf = start; a.start = &buf; a.end = start + have; /* start copying other objects */ if (compat) { a.type = DN_COMPAT_PIPE; dn_ht_scan(dn_cfg.schedhash, copy_data_helper_compat, &a); a.type = DN_COMPAT_QUEUE; dn_ht_scan(dn_cfg.fshash, copy_data_helper_compat, &a); } else if (a.type == DN_FS) { dn_ht_scan(dn_cfg.fshash, copy_data_helper, &a); } else { dn_ht_scan(dn_cfg.schedhash, copy_data_helper, &a); } DN_BH_WUNLOCK(); if (compat) { *compat = start; sopt->sopt_valsize = buf - start; /* free() is done by ip_dummynet_compat() */ start = NULL; //XXX hack } else { error = sooptcopyout(sopt, start, buf - start); } done: if (cmd && cmd != &r.o) free(cmd, M_DUMMYNET); if (start) free(start, M_DUMMYNET); return error; } /* Callback called on scheduler instance to delete it if idle */ static int drain_scheduler_cb(void *_si, void *arg) { struct dn_sch_inst *si = _si; if ((si->kflags & DN_ACTIVE) || si->dline.mq.head != NULL) return 0; if (si->sched->fp->flags & DN_MULTIQUEUE) { if (si->q_count == 0) return si_destroy(si, NULL); else return 0; } else { /* !DN_MULTIQUEUE */ if ((si+1)->ni.length == 0) return si_destroy(si, NULL); else return 0; } return 0; /* unreachable */ } /* Callback called on scheduler to check if it has instances */ static int drain_scheduler_sch_cb(void *_s, void *arg) { struct dn_schk *s = _s; if (s->sch.flags & DN_HAVE_MASK) { dn_ht_scan_bucket(s->siht, &s->drain_bucket, drain_scheduler_cb, NULL); s->drain_bucket++; } else { if (s->siht) { if (drain_scheduler_cb(s->siht, NULL) == DNHT_SCAN_DEL) s->siht = NULL; } } return 0; } /* Called every tick, try to delete a 'bucket' of scheduler */ void dn_drain_scheduler(void) { dn_ht_scan_bucket(dn_cfg.schedhash, &dn_cfg.drain_sch, drain_scheduler_sch_cb, NULL); dn_cfg.drain_sch++; } /* Callback called on queue to delete if it is idle */ static int drain_queue_cb(void *_q, void *arg) { struct dn_queue *q = _q; if (q->ni.length == 0) { dn_delete_queue(q, DN_DESTROY); return DNHT_SCAN_DEL; /* queue is deleted */ } return 0; /* queue isn't deleted */ } /* Callback called on flowset used to check if it has queues */ static int drain_queue_fs_cb(void *_fs, void *arg) { struct dn_fsk *fs = _fs; if (fs->fs.flags & DN_QHT_HASH) { /* Flowset has a hash table for queues */ dn_ht_scan_bucket(fs->qht, &fs->drain_bucket, drain_queue_cb, NULL); fs->drain_bucket++; } else { /* No hash table for this flowset, null the pointer * if the queue is deleted */ if (fs->qht) { if (drain_queue_cb(fs->qht, NULL) == DNHT_SCAN_DEL) fs->qht = NULL; } } return 0; } /* Called every tick, try to delete a 'bucket' of queue */ void dn_drain_queue(void) { /* scan a bucket of flowset */ dn_ht_scan_bucket(dn_cfg.fshash, &dn_cfg.drain_fs, drain_queue_fs_cb, NULL); dn_cfg.drain_fs++; } /* * Handler for the various dummynet socket options */ static int ip_dn_ctl(struct sockopt *sopt) { void *p = NULL; int error, l; error = priv_check(sopt->sopt_td, PRIV_NETINET_DUMMYNET); if (error) return (error); /* Disallow sets in really-really secure mode. */ if (sopt->sopt_dir == SOPT_SET) { error = securelevel_ge(sopt->sopt_td->td_ucred, 3); if (error) return (error); } switch (sopt->sopt_name) { default : D("dummynet: unknown option %d", sopt->sopt_name); error = EINVAL; break; case IP_DUMMYNET_FLUSH: case IP_DUMMYNET_CONFIGURE: case IP_DUMMYNET_DEL: /* remove a pipe or queue */ case IP_DUMMYNET_GET: D("dummynet: compat option %d", sopt->sopt_name); error = ip_dummynet_compat(sopt); break; case IP_DUMMYNET3 : if (sopt->sopt_dir == SOPT_GET) { error = dummynet_get(sopt, NULL); break; } l = sopt->sopt_valsize; if (l < sizeof(struct dn_id) || l > 12000) { D("argument len %d invalid", l); break; } p = malloc(l, M_TEMP, M_WAITOK); // XXX can it fail ? error = sooptcopyin(sopt, p, l, l); if (error) break ; error = do_config(p, l); break; } if (p != NULL) free(p, M_TEMP); return error ; } static void ip_dn_init(void) { if (dn_cfg.init_done) return; printf("DUMMYNET %p with IPv6 initialized (100409)\n", curvnet); dn_cfg.init_done = 1; /* Set defaults here. MSVC does not accept initializers, * and this is also useful for vimages */ /* queue limits */ dn_cfg.slot_limit = 100; /* Foot shooting limit for queues. */ dn_cfg.byte_limit = 1024 * 1024; dn_cfg.expire = 1; /* RED parameters */ dn_cfg.red_lookup_depth = 256; /* default lookup table depth */ dn_cfg.red_avg_pkt_size = 512; /* default medium packet size */ dn_cfg.red_max_pkt_size = 1500; /* default max packet size */ /* hash tables */ dn_cfg.max_hash_size = 65536; /* max in the hash tables */ dn_cfg.hash_size = 64; /* default hash size */ /* create hash tables for schedulers and flowsets. * In both we search by key and by pointer. */ dn_cfg.schedhash = dn_ht_init(NULL, dn_cfg.hash_size, offsetof(struct dn_schk, schk_next), schk_hash, schk_match, schk_new); dn_cfg.fshash = dn_ht_init(NULL, dn_cfg.hash_size, offsetof(struct dn_fsk, fsk_next), fsk_hash, fsk_match, fsk_new); /* bucket index to drain object */ dn_cfg.drain_fs = 0; dn_cfg.drain_sch = 0; heap_init(&dn_cfg.evheap, 16, offsetof(struct dn_id, id)); SLIST_INIT(&dn_cfg.fsu); SLIST_INIT(&dn_cfg.schedlist); DN_LOCK_INIT(); TASK_INIT(&dn_task, 0, dummynet_task, curvnet); dn_tq = taskqueue_create_fast("dummynet", M_WAITOK, taskqueue_thread_enqueue, &dn_tq); taskqueue_start_threads(&dn_tq, 1, PI_NET, "dummynet"); callout_init(&dn_timeout, 1); dn_reschedule(); /* Initialize curr_time adjustment mechanics. */ getmicrouptime(&dn_cfg.prev_t); } static void ip_dn_destroy(int last) { DN_BH_WLOCK(); /* ensure no more callouts are started */ dn_gone = 1; /* check for last */ if (last) { ND("removing last instance\n"); ip_dn_ctl_ptr = NULL; ip_dn_io_ptr = NULL; } dummynet_flush(); DN_BH_WUNLOCK(); callout_drain(&dn_timeout); taskqueue_drain(dn_tq, &dn_task); taskqueue_free(dn_tq); dn_ht_free(dn_cfg.schedhash, 0); dn_ht_free(dn_cfg.fshash, 0); heap_free(&dn_cfg.evheap); DN_LOCK_DESTROY(); } static int dummynet_modevent(module_t mod, int type, void *data) { if (type == MOD_LOAD) { if (ip_dn_io_ptr) { printf("DUMMYNET already loaded\n"); return EEXIST ; } ip_dn_init(); ip_dn_ctl_ptr = ip_dn_ctl; ip_dn_io_ptr = dummynet_io; return 0; } else if (type == MOD_UNLOAD) { ip_dn_destroy(1 /* last */); return 0; } else return EOPNOTSUPP; } /* modevent helpers for the modules */ static int load_dn_sched(struct dn_alg *d) { struct dn_alg *s; if (d == NULL) return 1; /* error */ ip_dn_init(); /* just in case, we need the lock */ /* Check that mandatory funcs exists */ if (d->enqueue == NULL || d->dequeue == NULL) { D("missing enqueue or dequeue for %s", d->name); return 1; } /* Search if scheduler already exists */ DN_BH_WLOCK(); SLIST_FOREACH(s, &dn_cfg.schedlist, next) { if (strcmp(s->name, d->name) == 0) { D("%s already loaded", d->name); break; /* scheduler already exists */ } } if (s == NULL) SLIST_INSERT_HEAD(&dn_cfg.schedlist, d, next); DN_BH_WUNLOCK(); D("dn_sched %s %sloaded", d->name, s ? "not ":""); return s ? 1 : 0; } static int unload_dn_sched(struct dn_alg *s) { struct dn_alg *tmp, *r; int err = EINVAL; ND("called for %s", s->name); DN_BH_WLOCK(); SLIST_FOREACH_SAFE(r, &dn_cfg.schedlist, next, tmp) { if (strcmp(s->name, r->name) != 0) continue; ND("ref_count = %d", r->ref_count); err = (r->ref_count != 0) ? EBUSY : 0; if (err == 0) SLIST_REMOVE(&dn_cfg.schedlist, r, dn_alg, next); break; } DN_BH_WUNLOCK(); D("dn_sched %s %sunloaded", s->name, err ? "not ":""); return err; } int dn_sched_modevent(module_t mod, int cmd, void *arg) { struct dn_alg *sch = arg; if (cmd == MOD_LOAD) return load_dn_sched(sch); else if (cmd == MOD_UNLOAD) return unload_dn_sched(sch); else return EINVAL; } static moduledata_t dummynet_mod = { "dummynet", dummynet_modevent, NULL }; -#define DN_SI_SUB SI_SUB_PROTO_IFATTACHDOMAIN +#define DN_SI_SUB SI_SUB_FW #define DN_MODEV_ORD (SI_ORDER_ANY - 128) /* after ipfw */ DECLARE_MODULE(dummynet, dummynet_mod, DN_SI_SUB, DN_MODEV_ORD); MODULE_DEPEND(dummynet, ipfw, 3, 3, 3); MODULE_VERSION(dummynet, 3); /* * Starting up. Done in order after dummynet_modevent() has been called. * VNET_SYSINIT is also called for each existing vnet and each new vnet. */ //VNET_SYSINIT(vnet_dn_init, DN_SI_SUB, DN_MODEV_ORD+2, ip_dn_init, NULL); /* * Shutdown handlers up shop. These are done in REVERSE ORDER, but still * after dummynet_modevent() has been called. Not called on reboot. * VNET_SYSUNINIT is also called for each exiting vnet as it exits. * or when the module is unloaded. */ //VNET_SYSUNINIT(vnet_dn_uninit, DN_SI_SUB, DN_MODEV_ORD+2, ip_dn_destroy, NULL); /* end of file */ Index: projects/vnet/sys/netpfil/ipfw/ip_fw2.c =================================================================== --- projects/vnet/sys/netpfil/ipfw/ip_fw2.c (revision 294565) +++ projects/vnet/sys/netpfil/ipfw/ip_fw2.c (revision 294566) @@ -1,2915 +1,2915 @@ /*- * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); /* * The FreeBSD IP packet firewall, main file */ #include "opt_ipfw.h" #include "opt_ipdivert.h" #include "opt_inet.h" #ifndef INET #error "IPFIREWALL requires INET" #endif /* INET */ #include "opt_inet6.h" #include "opt_ipsec.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* for ETHERTYPE_IP */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #include #endif #include #include /* XXX for in_cksum */ #ifdef MAC #include #endif /* * static variables followed by global ones. * All ipfw global variables are here. */ static VNET_DEFINE(int, fw_deny_unknown_exthdrs); #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) static VNET_DEFINE(int, fw_permit_single_frag6) = 1; #define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6) #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT static int default_to_accept = 1; #else static int default_to_accept; #endif VNET_DEFINE(int, autoinc_step); VNET_DEFINE(int, fw_one_pass) = 1; VNET_DEFINE(unsigned int, fw_tables_max); VNET_DEFINE(unsigned int, fw_tables_sets) = 0; /* Don't use set-aware tables */ /* Use 128 tables by default */ static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT; #ifndef LINEAR_SKIPTO static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, int tablearg, int jump_backwards); #define JUMP(ch, f, num, targ, back) jump_fast(ch, f, num, targ, back) #else static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, int tablearg, int jump_backwards); #define JUMP(ch, f, num, targ, back) jump_linear(ch, f, num, targ, back) #endif /* * Each rule belongs to one of 32 different sets (0..31). * The variable set_disable contains one bit per set. * If the bit is set, all rules in the corresponding set * are disabled. Set RESVD_SET(31) is reserved for the default rule * and rules that are not deleted by the flush command, * and CANNOT be disabled. * Rules in set RESVD_SET can only be deleted individually. */ VNET_DEFINE(u_int32_t, set_disable); #define V_set_disable VNET(set_disable) VNET_DEFINE(int, fw_verbose); /* counter for ipfw_log(NULL...) */ VNET_DEFINE(u_int64_t, norule_counter); VNET_DEFINE(int, verbose_limit); /* layer3_chain contains the list of rules for layer 3 */ VNET_DEFINE(struct ip_fw_chain, layer3_chain); /* ipfw_vnet_ready controls when we are open for business */ VNET_DEFINE(int, ipfw_vnet_ready) = 0; VNET_DEFINE(int, ipfw_nat_ready) = 0; ipfw_nat_t *ipfw_nat_ptr = NULL; struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; ipfw_nat_cfg_t *ipfw_nat_del_ptr; ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; #ifdef SYSCTL_NODE uint32_t dummy_def = IPFW_DEFAULT_RULE; static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS); static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS); SYSBEGIN(f3) SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, "Only do a single pass through ipfw when using dummynet(4)"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, "Rule number auto-increment step"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, "Log matches to ipfw rules"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, "Set upper limit of matches of ipfw rules logged"); SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, &dummy_def, 0, "The default/max possible rule number."); SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU", "Maximum number of concurrently used tables"); SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets, CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_tables_sets, "IU", "Use per-set namespace for tables"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, &default_to_accept, 0, "Make the default rule accept all packets."); TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, "Number of static rules"); #ifdef INET6 SYSCTL_DECL(_net_inet6_ip6); SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0, "Deny packets with unknown IPv6 Extension Headers"); SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0, "Permit single packet IPv6 fragments"); #endif /* INET6 */ SYSEND #endif /* SYSCTL_NODE */ /* * Some macros used in the various matching options. * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T * Other macros just cast void * into the appropriate type */ #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) #define TCP(p) ((struct tcphdr *)(p)) #define SCTP(p) ((struct sctphdr *)(p)) #define UDP(p) ((struct udphdr *)(p)) #define ICMP(p) ((struct icmphdr *)(p)) #define ICMP6(p) ((struct icmp6_hdr *)(p)) static __inline int icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) { int type = icmp->icmp_type; return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<icmp_type; return (type <= ICMP_MAXTYPE && (TT & (1<arg1 or cmd->d[0]. * * We scan options and store the bits we find set. We succeed if * * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear * * The code is sometimes optimized not to store additional variables. */ static int flags_match(ipfw_insn *cmd, u_int8_t bits) { u_char want_clear; bits = ~bits; if ( ((cmd->arg1 & 0xff) & bits) != 0) return 0; /* some bits we want set were clear */ want_clear = (cmd->arg1 >> 8) & 0xff; if ( (want_clear & bits) != want_clear) return 0; /* some bits we want clear were set */ return 1; } static int ipopts_match(struct ip *ip, ipfw_insn *cmd) { int optlen, bits = 0; u_char *cp = (u_char *)(ip + 1); int x = (ip->ip_hl << 2) - sizeof (struct ip); for (; x > 0; x -= optlen, cp += optlen) { int opt = cp[IPOPT_OPTVAL]; if (opt == IPOPT_EOL) break; if (opt == IPOPT_NOP) optlen = 1; else { optlen = cp[IPOPT_OLEN]; if (optlen <= 0 || optlen > x) return 0; /* invalid or truncated */ } switch (opt) { default: break; case IPOPT_LSRR: bits |= IP_FW_IPOPT_LSRR; break; case IPOPT_SSRR: bits |= IP_FW_IPOPT_SSRR; break; case IPOPT_RR: bits |= IP_FW_IPOPT_RR; break; case IPOPT_TS: bits |= IP_FW_IPOPT_TS; break; } } return (flags_match(cmd, bits)); } static int tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) { int optlen, bits = 0; u_char *cp = (u_char *)(tcp + 1); int x = (tcp->th_off << 2) - sizeof(struct tcphdr); for (; x > 0; x -= optlen, cp += optlen) { int opt = cp[0]; if (opt == TCPOPT_EOL) break; if (opt == TCPOPT_NOP) optlen = 1; else { optlen = cp[1]; if (optlen <= 0) break; } switch (opt) { default: break; case TCPOPT_MAXSEG: bits |= IP_FW_TCPOPT_MSS; break; case TCPOPT_WINDOW: bits |= IP_FW_TCPOPT_WINDOW; break; case TCPOPT_SACK_PERMITTED: case TCPOPT_SACK: bits |= IP_FW_TCPOPT_SACK; break; case TCPOPT_TIMESTAMP: bits |= IP_FW_TCPOPT_TS; break; } } return (flags_match(cmd, bits)); } static int iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, uint32_t *tablearg) { if (ifp == NULL) /* no iface with this packet, match fails */ return (0); /* Check by name or by IP address */ if (cmd->name[0] != '\0') { /* match by name */ if (cmd->name[0] == '\1') /* use tablearg to match */ return ipfw_lookup_table_extended(chain, cmd->p.kidx, 0, &ifp->if_index, tablearg); /* Check name */ if (cmd->p.glob) { if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) return(1); } else { if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) return(1); } } else { #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */ struct ifaddr *ia; if_addr_rlock(ifp); TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { if (ia->ifa_addr->sa_family != AF_INET) continue; if (cmd->p.ip.s_addr == ((struct sockaddr_in *) (ia->ifa_addr))->sin_addr.s_addr) { if_addr_runlock(ifp); return(1); /* match */ } } if_addr_runlock(ifp); #endif /* __FreeBSD__ */ } return(0); /* no match, fail ... */ } /* * The verify_path function checks if a route to the src exists and * if it is reachable via ifp (when provided). * * The 'verrevpath' option checks that the interface that an IP packet * arrives on is the same interface that traffic destined for the * packet's source address would be routed out of. * The 'versrcreach' option just checks that the source address is * reachable via any route (except default) in the routing table. * These two are a measure to block forged packets. This is also * commonly known as "anti-spoofing" or Unicast Reverse Path * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs * is purposely reminiscent of the Cisco IOS command, * * ip verify unicast reverse-path * ip verify unicast source reachable-via any * * which implements the same functionality. But note that the syntax * is misleading, and the check may be performed on all IP packets * whether unicast, multicast, or broadcast. */ static int verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) { #if defined(USERSPACE) || !defined(__FreeBSD__) return 0; #else struct nhop4_basic nh4; if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0) return (0); /* * If ifp is provided, check for equality with rtentry. * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, * in order to pass packets injected back by if_simloop(): * routing entry (via lo0) for our own address * may exist, so we need to handle routing assymetry. */ if (ifp != NULL && ifp != nh4.nh_ifp) return (0); /* if no ifp provided, check if rtentry is not default route */ if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0) return (0); /* or if this is a blackhole/reject route */ if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) return (0); /* found valid route */ return 1; #endif /* __FreeBSD__ */ } #ifdef INET6 /* * ipv6 specific rules here... */ static __inline int icmp6type_match (int type, ipfw_insn_u32 *cmd) { return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); } static int flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) { int i; for (i=0; i <= cmd->o.arg1; ++i ) if (curr_flow == cmd->d[i] ) return 1; return 0; } /* support for IP6_*_ME opcodes */ static const struct in6_addr lla_mask = {{{ 0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }}}; static int ipfw_localip6(struct in6_addr *in6) { struct rm_priotracker in6_ifa_tracker; struct in6_ifaddr *ia; if (IN6_IS_ADDR_MULTICAST(in6)) return (0); if (!IN6_IS_ADDR_LINKLOCAL(in6)) return (in6_localip(in6)); IN6_IFADDR_RLOCK(&in6_ifa_tracker); TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) continue; if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr, in6, &lla_mask)) { IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (1); } } IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); return (0); } static int verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib) { struct nhop6_basic nh6; if (IN6_IS_SCOPE_LINKLOCAL(src)) return (1); if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0) return (0); /* If ifp is provided, check for equality with route table. */ if (ifp != NULL && ifp != nh6.nh_ifp) return (0); /* if no ifp provided, check if rtentry is not default route */ if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0) return (0); /* or if this is a blackhole/reject route */ if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) return (0); /* found valid route */ return 1; } static int is_icmp6_query(int icmp6_type) { if ((icmp6_type <= ICMP6_MAXTYPE) && (icmp6_type == ICMP6_ECHO_REQUEST || icmp6_type == ICMP6_MEMBERSHIP_QUERY || icmp6_type == ICMP6_WRUREQUEST || icmp6_type == ICMP6_FQDN_QUERY || icmp6_type == ICMP6_NI_QUERY)) return (1); return (0); } static void send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) { struct mbuf *m; m = args->m; if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { struct tcphdr *tcp; tcp = (struct tcphdr *)((char *)ip6 + hlen); if ((tcp->th_flags & TH_RST) == 0) { struct mbuf *m0; m0 = ipfw_send_pkt(args->m, &(args->f_id), ntohl(tcp->th_seq), ntohl(tcp->th_ack), tcp->th_flags | TH_RST); if (m0 != NULL) ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); } FREE_PKT(m); } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ #if 0 /* * Unlike above, the mbufs need to line up with the ip6 hdr, * as the contents are read. We need to m_adj() the * needed amount. * The mbuf will however be thrown away so we can adjust it. * Remember we did an m_pullup on it already so we * can make some assumptions about contiguousness. */ if (args->L3offset) m_adj(m, args->L3offset); #endif icmp6_error(m, ICMP6_DST_UNREACH, code, 0); } else FREE_PKT(m); args->m = NULL; } #endif /* INET6 */ /* * sends a reject message, consuming the mbuf passed as an argument. */ static void send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip) { #if 0 /* XXX When ip is not guaranteed to be at mtod() we will * need to account for this */ * The mbuf will however be thrown away so we can adjust it. * Remember we did an m_pullup on it already so we * can make some assumptions about contiguousness. */ if (args->L3offset) m_adj(m, args->L3offset); #endif if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); } else if (args->f_id.proto == IPPROTO_TCP) { struct tcphdr *const tcp = L3HDR(struct tcphdr, mtod(args->m, struct ip *)); if ( (tcp->th_flags & TH_RST) == 0) { struct mbuf *m; m = ipfw_send_pkt(args->m, &(args->f_id), ntohl(tcp->th_seq), ntohl(tcp->th_ack), tcp->th_flags | TH_RST); if (m != NULL) ip_output(m, NULL, NULL, 0, NULL, NULL); } FREE_PKT(args->m); } else FREE_PKT(args->m); args->m = NULL; } /* * Support for uid/gid/jail lookup. These tests are expensive * (because we may need to look into the list of active sockets) * so we cache the results. ugid_lookupp is 0 if we have not * yet done a lookup, 1 if we succeeded, and -1 if we tried * and failed. The function always returns the match value. * We could actually spare the variable and use *uc, setting * it to '(void *)check_uidgid if we have no info, NULL if * we tried and failed, or any other value if successful. */ static int check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp, struct ucred **uc) { #if defined(USERSPACE) return 0; // not supported in userspace #else #ifndef __FreeBSD__ /* XXX */ return cred_check(insn, proto, oif, dst_ip, dst_port, src_ip, src_port, (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); #else /* FreeBSD */ struct in_addr src_ip, dst_ip; struct inpcbinfo *pi; struct ipfw_flow_id *id; struct inpcb *pcb, *inp; struct ifnet *oif; int lookupflags; int match; id = &args->f_id; inp = args->inp; oif = args->oif; /* * Check to see if the UDP or TCP stack supplied us with * the PCB. If so, rather then holding a lock and looking * up the PCB, we can use the one that was supplied. */ if (inp && *ugid_lookupp == 0) { INP_LOCK_ASSERT(inp); if (inp->inp_socket != NULL) { *uc = crhold(inp->inp_cred); *ugid_lookupp = 1; } else *ugid_lookupp = -1; } /* * If we have already been here and the packet has no * PCB entry associated with it, then we can safely * assume that this is a no match. */ if (*ugid_lookupp == -1) return (0); if (id->proto == IPPROTO_TCP) { lookupflags = 0; pi = &V_tcbinfo; } else if (id->proto == IPPROTO_UDP) { lookupflags = INPLOOKUP_WILDCARD; pi = &V_udbinfo; } else return 0; lookupflags |= INPLOOKUP_RLOCKPCB; match = 0; if (*ugid_lookupp == 0) { if (id->addr_type == 6) { #ifdef INET6 if (oif == NULL) pcb = in6_pcblookup_mbuf(pi, &id->src_ip6, htons(id->src_port), &id->dst_ip6, htons(id->dst_port), lookupflags, oif, args->m); else pcb = in6_pcblookup_mbuf(pi, &id->dst_ip6, htons(id->dst_port), &id->src_ip6, htons(id->src_port), lookupflags, oif, args->m); #else *ugid_lookupp = -1; return (0); #endif } else { src_ip.s_addr = htonl(id->src_ip); dst_ip.s_addr = htonl(id->dst_ip); if (oif == NULL) pcb = in_pcblookup_mbuf(pi, src_ip, htons(id->src_port), dst_ip, htons(id->dst_port), lookupflags, oif, args->m); else pcb = in_pcblookup_mbuf(pi, dst_ip, htons(id->dst_port), src_ip, htons(id->src_port), lookupflags, oif, args->m); } if (pcb != NULL) { INP_RLOCK_ASSERT(pcb); *uc = crhold(pcb->inp_cred); *ugid_lookupp = 1; INP_RUNLOCK(pcb); } if (*ugid_lookupp == 0) { /* * We tried and failed, set the variable to -1 * so we will not try again on this packet. */ *ugid_lookupp = -1; return (0); } } if (insn->o.opcode == O_UID) match = ((*uc)->cr_uid == (uid_t)insn->d[0]); else if (insn->o.opcode == O_GID) match = groupmember((gid_t)insn->d[0], *uc); else if (insn->o.opcode == O_JAIL) match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); return (match); #endif /* __FreeBSD__ */ #endif /* not supported in userspace */ } /* * Helper function to set args with info on the rule after the matching * one. slot is precise, whereas we guess rule_id as they are * assigned sequentially. */ static inline void set_match(struct ip_fw_args *args, int slot, struct ip_fw_chain *chain) { args->rule.chain_id = chain->id; args->rule.slot = slot + 1; /* we use 0 as a marker */ args->rule.rule_id = 1 + chain->map[slot]->id; args->rule.rulenum = chain->map[slot]->rulenum; } #ifndef LINEAR_SKIPTO /* * Helper function to enable cached rule lookups using * cached_id and cached_pos fields in ipfw rule. */ static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, int tablearg, int jump_backwards) { int f_pos; /* If possible use cached f_pos (in f->cached_pos), * whose version is written in f->cached_id * (horrible hacks to avoid changing the ABI). */ if (num != IP_FW_TARG && f->cached_id == chain->id) f_pos = f->cached_pos; else { int i = IP_FW_ARG_TABLEARG(chain, num, skipto); /* make sure we do not jump backward */ if (jump_backwards == 0 && i <= f->rulenum) i = f->rulenum + 1; if (chain->idxmap != NULL) f_pos = chain->idxmap[i]; else f_pos = ipfw_find_rule(chain, i, 0); /* update the cache */ if (num != IP_FW_TARG) { f->cached_id = chain->id; f->cached_pos = f_pos; } } return (f_pos); } #else /* * Helper function to enable real fast rule lookups. */ static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, int tablearg, int jump_backwards) { int f_pos; num = IP_FW_ARG_TABLEARG(chain, num, skipto); /* make sure we do not jump backward */ if (jump_backwards == 0 && num <= f->rulenum) num = f->rulenum + 1; f_pos = chain->idxmap[num]; return (f_pos); } #endif #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f) /* * The main check routine for the firewall. * * All arguments are in args so we can modify them and return them * back to the caller. * * Parameters: * * args->m (in/out) The packet; we set to NULL when/if we nuke it. * Starts with the IP header. * args->eh (in) Mac header if present, NULL for layer3 packet. * args->L3offset Number of bytes bypassed if we came from L2. * e.g. often sizeof(eh) ** NOTYET ** * args->oif Outgoing interface, NULL if packet is incoming. * The incoming interface is in the mbuf. (in) * args->divert_rule (in/out) * Skip up to the first rule past this rule number; * upon return, non-zero port number for divert or tee. * * args->rule Pointer to the last matching rule (in/out) * args->next_hop Socket we are forwarding to (out). * args->next_hop6 IPv6 next hop we are forwarding to (out). * args->f_id Addresses grabbed from the packet (out) * args->rule.info a cookie depending on rule action * * Return value: * * IP_FW_PASS the packet must be accepted * IP_FW_DENY the packet must be dropped * IP_FW_DIVERT divert packet, port in m_tag * IP_FW_TEE tee packet, port in m_tag * IP_FW_DUMMYNET to dummynet, pipe in args->cookie * IP_FW_NETGRAPH into netgraph, cookie args->cookie * args->rule contains the matching rule, * args->rule.info has additional information. * */ int ipfw_chk(struct ip_fw_args *args) { /* * Local variables holding state while processing a packet: * * IMPORTANT NOTE: to speed up the processing of rules, there * are some assumption on the values of the variables, which * are documented here. Should you change them, please check * the implementation of the various instructions to make sure * that they still work. * * args->eh The MAC header. It is non-null for a layer2 * packet, it is NULL for a layer-3 packet. * **notyet** * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. * * m | args->m Pointer to the mbuf, as received from the caller. * It may change if ipfw_chk() does an m_pullup, or if it * consumes the packet because it calls send_reject(). * XXX This has to change, so that ipfw_chk() never modifies * or consumes the buffer. * ip is the beginning of the ip(4 or 6) header. * Calculated by adding the L3offset to the start of data. * (Until we start using L3offset, the packet is * supposed to start with the ip header). */ struct mbuf *m = args->m; struct ip *ip = mtod(m, struct ip *); /* * For rules which contain uid/gid or jail constraints, cache * a copy of the users credentials after the pcb lookup has been * executed. This will speed up the processing of rules with * these types of constraints, as well as decrease contention * on pcb related locks. */ #ifndef __FreeBSD__ struct bsd_ucred ucred_cache; #else struct ucred *ucred_cache = NULL; #endif int ucred_lookup = 0; /* * oif | args->oif If NULL, ipfw_chk has been called on the * inbound path (ether_input, ip_input). * If non-NULL, ipfw_chk has been called on the outbound path * (ether_output, ip_output). */ struct ifnet *oif = args->oif; int f_pos = 0; /* index of current rule in the array */ int retval = 0; /* * hlen The length of the IP header. */ u_int hlen = 0; /* hlen >0 means we have an IP pkt */ /* * offset The offset of a fragment. offset != 0 means that * we have a fragment at this offset of an IPv4 packet. * offset == 0 means that (if this is an IPv4 packet) * this is the first or only fragment. * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header * or there is a single packet fragement (fragement header added * without needed). We will treat a single packet fragment as if * there was no fragment header (or log/block depending on the * V_fw_permit_single_frag6 sysctl setting). */ u_short offset = 0; u_short ip6f_mf = 0; /* * Local copies of addresses. They are only valid if we have * an IP packet. * * proto The protocol. Set to 0 for non-ip packets, * or to the protocol read from the packet otherwise. * proto != 0 means that we have an IPv4 packet. * * src_port, dst_port port numbers, in HOST format. Only * valid for TCP and UDP packets. * * src_ip, dst_ip ip addresses, in NETWORK format. * Only valid for IPv4 packets. */ uint8_t proto; uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */ struct in_addr src_ip, dst_ip; /* NOTE: network format */ uint16_t iplen=0; int pktlen; uint16_t etype = 0; /* Host order stored ether type */ /* * dyn_dir = MATCH_UNKNOWN when rules unchecked, * MATCH_NONE when checked and not matched (q = NULL), * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) */ int dyn_dir = MATCH_UNKNOWN; ipfw_dyn_rule *q = NULL; struct ip_fw_chain *chain = &V_layer3_chain; /* * We store in ulp a pointer to the upper layer protocol header. * In the ipv4 case this is easy to determine from the header, * but for ipv6 we might have some additional headers in the middle. * ulp is NULL if not found. */ void *ulp = NULL; /* upper layer protocol pointer. */ /* XXX ipv6 variables */ int is_ipv6 = 0; uint8_t icmp6_type = 0; uint16_t ext_hd = 0; /* bits vector for extension header filtering */ /* end of ipv6 variables */ int is_ipv4 = 0; int done = 0; /* flag to exit the outer loop */ IPFW_RLOCK_TRACKER; if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) return (IP_FW_PASS); /* accept */ dst_ip.s_addr = 0; /* make sure it is initialized */ src_ip.s_addr = 0; /* make sure it is initialized */ pktlen = m->m_pkthdr.len; args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ proto = args->f_id.proto = 0; /* mark f_id invalid */ /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ /* * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, * then it sets p to point at the offset "len" in the mbuf. WARNING: the * pointer might become stale after other pullups (but we never use it * this way). */ #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) #define PULLUP_LEN(_len, p, T) \ do { \ int x = (_len) + T; \ if ((m)->m_len < x) { \ args->m = m = m_pullup(m, x); \ if (m == NULL) \ goto pullup_failed; \ } \ p = (mtod(m, char *) + (_len)); \ } while (0) /* * if we have an ether header, */ if (args->eh) etype = ntohs(args->eh->ether_type); /* Identify IP packets and fill up variables. */ if (pktlen >= sizeof(struct ip6_hdr) && (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; is_ipv6 = 1; args->f_id.addr_type = 6; hlen = sizeof(struct ip6_hdr); proto = ip6->ip6_nxt; /* Search extension headers to find upper layer protocols */ while (ulp == NULL && offset == 0) { switch (proto) { case IPPROTO_ICMPV6: PULLUP_TO(hlen, ulp, struct icmp6_hdr); icmp6_type = ICMP6(ulp)->icmp6_type; break; case IPPROTO_TCP: PULLUP_TO(hlen, ulp, struct tcphdr); dst_port = TCP(ulp)->th_dport; src_port = TCP(ulp)->th_sport; /* save flags for dynamic rules */ args->f_id._flags = TCP(ulp)->th_flags; break; case IPPROTO_SCTP: PULLUP_TO(hlen, ulp, struct sctphdr); src_port = SCTP(ulp)->src_port; dst_port = SCTP(ulp)->dest_port; break; case IPPROTO_UDP: PULLUP_TO(hlen, ulp, struct udphdr); dst_port = UDP(ulp)->uh_dport; src_port = UDP(ulp)->uh_sport; break; case IPPROTO_HOPOPTS: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_hbh); ext_hd |= EXT_HOPOPTS; hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; ulp = NULL; break; case IPPROTO_ROUTING: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_rthdr); switch (((struct ip6_rthdr *)ulp)->ip6r_type) { case 0: ext_hd |= EXT_RTHDR0; break; case 2: ext_hd |= EXT_RTHDR2; break; default: if (V_fw_verbose) printf("IPFW2: IPV6 - Unknown " "Routing Header type(%d)\n", ((struct ip6_rthdr *) ulp)->ip6r_type); if (V_fw_deny_unknown_exthdrs) return (IP_FW_DENY); break; } ext_hd |= EXT_ROUTING; hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; ulp = NULL; break; case IPPROTO_FRAGMENT: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_frag); ext_hd |= EXT_FRAGMENT; hlen += sizeof (struct ip6_frag); proto = ((struct ip6_frag *)ulp)->ip6f_nxt; offset = ((struct ip6_frag *)ulp)->ip6f_offlg & IP6F_OFF_MASK; ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg & IP6F_MORE_FRAG; if (V_fw_permit_single_frag6 == 0 && offset == 0 && ip6f_mf == 0) { if (V_fw_verbose) printf("IPFW2: IPV6 - Invalid " "Fragment Header\n"); if (V_fw_deny_unknown_exthdrs) return (IP_FW_DENY); break; } args->f_id.extra = ntohl(((struct ip6_frag *)ulp)->ip6f_ident); ulp = NULL; break; case IPPROTO_DSTOPTS: /* RFC 2460 */ PULLUP_TO(hlen, ulp, struct ip6_hbh); ext_hd |= EXT_DSTOPTS; hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; ulp = NULL; break; case IPPROTO_AH: /* RFC 2402 */ PULLUP_TO(hlen, ulp, struct ip6_ext); ext_hd |= EXT_AH; hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; proto = ((struct ip6_ext *)ulp)->ip6e_nxt; ulp = NULL; break; case IPPROTO_ESP: /* RFC 2406 */ PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ /* Anything past Seq# is variable length and * data past this ext. header is encrypted. */ ext_hd |= EXT_ESP; break; case IPPROTO_NONE: /* RFC 2460 */ /* * Packet ends here, and IPv6 header has * already been pulled up. If ip6e_len!=0 * then octets must be ignored. */ ulp = ip; /* non-NULL to get out of loop. */ break; case IPPROTO_OSPFIGP: /* XXX OSPF header check? */ PULLUP_TO(hlen, ulp, struct ip6_ext); break; case IPPROTO_PIM: /* XXX PIM header check? */ PULLUP_TO(hlen, ulp, struct pim); break; case IPPROTO_CARP: PULLUP_TO(hlen, ulp, struct carp_header); if (((struct carp_header *)ulp)->carp_version != CARP_VERSION) return (IP_FW_DENY); if (((struct carp_header *)ulp)->carp_type != CARP_ADVERTISEMENT) return (IP_FW_DENY); break; case IPPROTO_IPV6: /* RFC 2893 */ PULLUP_TO(hlen, ulp, struct ip6_hdr); break; case IPPROTO_IPV4: /* RFC 2893 */ PULLUP_TO(hlen, ulp, struct ip); break; default: if (V_fw_verbose) printf("IPFW2: IPV6 - Unknown " "Extension Header(%d), ext_hd=%x\n", proto, ext_hd); if (V_fw_deny_unknown_exthdrs) return (IP_FW_DENY); PULLUP_TO(hlen, ulp, struct ip6_ext); break; } /*switch */ } ip = mtod(m, struct ip *); ip6 = (struct ip6_hdr *)ip; args->f_id.src_ip6 = ip6->ip6_src; args->f_id.dst_ip6 = ip6->ip6_dst; args->f_id.src_ip = 0; args->f_id.dst_ip = 0; args->f_id.flow_id6 = ntohl(ip6->ip6_flow); } else if (pktlen >= sizeof(struct ip) && (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { is_ipv4 = 1; hlen = ip->ip_hl << 2; args->f_id.addr_type = 4; /* * Collect parameters into local variables for faster matching. */ proto = ip->ip_p; src_ip = ip->ip_src; dst_ip = ip->ip_dst; offset = ntohs(ip->ip_off) & IP_OFFMASK; iplen = ntohs(ip->ip_len); pktlen = iplen < pktlen ? iplen : pktlen; if (offset == 0) { switch (proto) { case IPPROTO_TCP: PULLUP_TO(hlen, ulp, struct tcphdr); dst_port = TCP(ulp)->th_dport; src_port = TCP(ulp)->th_sport; /* save flags for dynamic rules */ args->f_id._flags = TCP(ulp)->th_flags; break; case IPPROTO_SCTP: PULLUP_TO(hlen, ulp, struct sctphdr); src_port = SCTP(ulp)->src_port; dst_port = SCTP(ulp)->dest_port; break; case IPPROTO_UDP: PULLUP_TO(hlen, ulp, struct udphdr); dst_port = UDP(ulp)->uh_dport; src_port = UDP(ulp)->uh_sport; break; case IPPROTO_ICMP: PULLUP_TO(hlen, ulp, struct icmphdr); //args->f_id.flags = ICMP(ulp)->icmp_type; break; default: break; } } ip = mtod(m, struct ip *); args->f_id.src_ip = ntohl(src_ip.s_addr); args->f_id.dst_ip = ntohl(dst_ip.s_addr); } #undef PULLUP_TO if (proto) { /* we may have port numbers, store them */ args->f_id.proto = proto; args->f_id.src_port = src_port = ntohs(src_port); args->f_id.dst_port = dst_port = ntohs(dst_port); } IPFW_PF_RLOCK(chain); if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ IPFW_PF_RUNLOCK(chain); return (IP_FW_PASS); /* accept */ } if (args->rule.slot) { /* * Packet has already been tagged as a result of a previous * match on rule args->rule aka args->rule_id (PIPE, QUEUE, * REASS, NETGRAPH, DIVERT/TEE...) * Validate the slot and continue from the next one * if still present, otherwise do a lookup. */ f_pos = (args->rule.chain_id == chain->id) ? args->rule.slot : ipfw_find_rule(chain, args->rule.rulenum, args->rule.rule_id); } else { f_pos = 0; } /* * Now scan the rules, and parse microinstructions for each rule. * We have two nested loops and an inner switch. Sometimes we * need to break out of one or both loops, or re-enter one of * the loops with updated variables. Loop variables are: * * f_pos (outer loop) points to the current rule. * On output it points to the matching rule. * done (outer loop) is used as a flag to break the loop. * l (inner loop) residual length of current rule. * cmd points to the current microinstruction. * * We break the inner loop by setting l=0 and possibly * cmdlen=0 if we don't want to advance cmd. * We break the outer loop by setting done=1 * We can restart the inner loop by setting l>0 and f_pos, f, cmd * as needed. */ for (; f_pos < chain->n_rules; f_pos++) { ipfw_insn *cmd; uint32_t tablearg = 0; int l, cmdlen, skip_or; /* skip rest of OR block */ struct ip_fw *f; f = chain->map[f_pos]; if (V_set_disable & (1 << f->set) ) continue; skip_or = 0; for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; l -= cmdlen, cmd += cmdlen) { int match; /* * check_body is a jump target used when we find a * CHECK_STATE, and need to jump to the body of * the target rule. */ /* check_body: */ cmdlen = F_LEN(cmd); /* * An OR block (insn_1 || .. || insn_n) has the * F_OR bit set in all but the last instruction. * The first match will set "skip_or", and cause * the following instructions to be skipped until * past the one with the F_OR bit clear. */ if (skip_or) { /* skip this instruction */ if ((cmd->len & F_OR) == 0) skip_or = 0; /* next one is good */ continue; } match = 0; /* set to 1 if we succeed */ switch (cmd->opcode) { /* * The first set of opcodes compares the packet's * fields with some pattern, setting 'match' if a * match is found. At the end of the loop there is * logic to deal with F_NOT and F_OR flags associated * with the opcode. */ case O_NOP: match = 1; break; case O_FORWARD_MAC: printf("ipfw: opcode %d unimplemented\n", cmd->opcode); break; case O_GID: case O_UID: case O_JAIL: /* * We only check offset == 0 && proto != 0, * as this ensures that we have a * packet with the ports info. */ if (offset != 0) break; if (proto == IPPROTO_TCP || proto == IPPROTO_UDP) match = check_uidgid( (ipfw_insn_u32 *)cmd, args, &ucred_lookup, #ifdef __FreeBSD__ &ucred_cache); #else (void *)&ucred_cache); #endif break; case O_RECV: match = iface_match(m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd, chain, &tablearg); break; case O_XMIT: match = iface_match(oif, (ipfw_insn_if *)cmd, chain, &tablearg); break; case O_VIA: match = iface_match(oif ? oif : m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd, chain, &tablearg); break; case O_MACADDR2: if (args->eh != NULL) { /* have MAC header */ u_int32_t *want = (u_int32_t *) ((ipfw_insn_mac *)cmd)->addr; u_int32_t *mask = (u_int32_t *) ((ipfw_insn_mac *)cmd)->mask; u_int32_t *hdr = (u_int32_t *)args->eh; match = ( want[0] == (hdr[0] & mask[0]) && want[1] == (hdr[1] & mask[1]) && want[2] == (hdr[2] & mask[2]) ); } break; case O_MAC_TYPE: if (args->eh != NULL) { u_int16_t *p = ((ipfw_insn_u16 *)cmd)->ports; int i; for (i = cmdlen - 1; !match && i>0; i--, p += 2) match = (etype >= p[0] && etype <= p[1]); } break; case O_FRAG: match = (offset != 0); break; case O_IN: /* "out" is "not in" */ match = (oif == NULL); break; case O_LAYER2: match = (args->eh != NULL); break; case O_DIVERTED: { /* For diverted packets, args->rule.info * contains the divert port (in host format) * reason and direction. */ uint32_t i = args->rule.info; match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT && cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2); } break; case O_PROTO: /* * We do not allow an arg of 0 so the * check of "proto" only suffices. */ match = (proto == cmd->arg1); break; case O_IP_SRC: match = is_ipv4 && (((ipfw_insn_ip *)cmd)->addr.s_addr == src_ip.s_addr); break; case O_IP_SRC_LOOKUP: case O_IP_DST_LOOKUP: if (is_ipv4) { uint32_t key = (cmd->opcode == O_IP_DST_LOOKUP) ? dst_ip.s_addr : src_ip.s_addr; uint32_t v = 0; if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { /* generic lookup. The key must be * in 32bit big-endian format. */ v = ((ipfw_insn_u32 *)cmd)->d[1]; if (v == 0) key = dst_ip.s_addr; else if (v == 1) key = src_ip.s_addr; else if (v == 6) /* dscp */ key = (ip->ip_tos >> 2) & 0x3f; else if (offset != 0) break; else if (proto != IPPROTO_TCP && proto != IPPROTO_UDP) break; else if (v == 2) key = dst_port; else if (v == 3) key = src_port; #ifndef USERSPACE else if (v == 4 || v == 5) { check_uidgid( (ipfw_insn_u32 *)cmd, args, &ucred_lookup, #ifdef __FreeBSD__ &ucred_cache); if (v == 4 /* O_UID */) key = ucred_cache->cr_uid; else if (v == 5 /* O_JAIL */) key = ucred_cache->cr_prison->pr_id; #else /* !__FreeBSD__ */ (void *)&ucred_cache); if (v ==4 /* O_UID */) key = ucred_cache.uid; else if (v == 5 /* O_JAIL */) key = ucred_cache.xid; #endif /* !__FreeBSD__ */ } #endif /* !USERSPACE */ else break; } match = ipfw_lookup_table(chain, cmd->arg1, key, &v); if (!match) break; if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) match = ((ipfw_insn_u32 *)cmd)->d[0] == v; else tablearg = v; } else if (is_ipv6) { uint32_t v = 0; void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ? &args->f_id.dst_ip6: &args->f_id.src_ip6; match = ipfw_lookup_table_extended(chain, cmd->arg1, sizeof(struct in6_addr), pkey, &v); if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) match = ((ipfw_insn_u32 *)cmd)->d[0] == v; if (match) tablearg = v; } break; case O_IP_FLOW_LOOKUP: { uint32_t v = 0; match = ipfw_lookup_table_extended(chain, cmd->arg1, 0, &args->f_id, &v); if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) match = ((ipfw_insn_u32 *)cmd)->d[0] == v; if (match) tablearg = v; } break; case O_IP_SRC_MASK: case O_IP_DST_MASK: if (is_ipv4) { uint32_t a = (cmd->opcode == O_IP_DST_MASK) ? dst_ip.s_addr : src_ip.s_addr; uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; int i = cmdlen-1; for (; !match && i>0; i-= 2, p+= 2) match = (p[0] == (a & p[1])); } break; case O_IP_SRC_ME: if (is_ipv4) { struct ifnet *tif; INADDR_TO_IFP(src_ip, tif); match = (tif != NULL); break; } #ifdef INET6 /* FALLTHROUGH */ case O_IP6_SRC_ME: match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6); #endif break; case O_IP_DST_SET: case O_IP_SRC_SET: if (is_ipv4) { u_int32_t *d = (u_int32_t *)(cmd+1); u_int32_t addr = cmd->opcode == O_IP_DST_SET ? args->f_id.dst_ip : args->f_id.src_ip; if (addr < d[0]) break; addr -= d[0]; /* subtract base */ match = (addr < cmd->arg1) && ( d[ 1 + (addr>>5)] & (1<<(addr & 0x1f)) ); } break; case O_IP_DST: match = is_ipv4 && (((ipfw_insn_ip *)cmd)->addr.s_addr == dst_ip.s_addr); break; case O_IP_DST_ME: if (is_ipv4) { struct ifnet *tif; INADDR_TO_IFP(dst_ip, tif); match = (tif != NULL); break; } #ifdef INET6 /* FALLTHROUGH */ case O_IP6_DST_ME: match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6); #endif break; case O_IP_SRCPORT: case O_IP_DSTPORT: /* * offset == 0 && proto != 0 is enough * to guarantee that we have a * packet with port info. */ if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) && offset == 0) { u_int16_t x = (cmd->opcode == O_IP_SRCPORT) ? src_port : dst_port ; u_int16_t *p = ((ipfw_insn_u16 *)cmd)->ports; int i; for (i = cmdlen - 1; !match && i>0; i--, p += 2) match = (x>=p[0] && x<=p[1]); } break; case O_ICMPTYPE: match = (offset == 0 && proto==IPPROTO_ICMP && icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); break; #ifdef INET6 case O_ICMP6TYPE: match = is_ipv6 && offset == 0 && proto==IPPROTO_ICMPV6 && icmp6type_match( ICMP6(ulp)->icmp6_type, (ipfw_insn_u32 *)cmd); break; #endif /* INET6 */ case O_IPOPT: match = (is_ipv4 && ipopts_match(ip, cmd) ); break; case O_IPVER: match = (is_ipv4 && cmd->arg1 == ip->ip_v); break; case O_IPID: case O_IPLEN: case O_IPTTL: if (is_ipv4) { /* only for IP packets */ uint16_t x; uint16_t *p; int i; if (cmd->opcode == O_IPLEN) x = iplen; else if (cmd->opcode == O_IPTTL) x = ip->ip_ttl; else /* must be IPID */ x = ntohs(ip->ip_id); if (cmdlen == 1) { match = (cmd->arg1 == x); break; } /* otherwise we have ranges */ p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for (; !match && i>0; i--, p += 2) match = (x >= p[0] && x <= p[1]); } break; case O_IPPRECEDENCE: match = (is_ipv4 && (cmd->arg1 == (ip->ip_tos & 0xe0)) ); break; case O_IPTOS: match = (is_ipv4 && flags_match(cmd, ip->ip_tos)); break; case O_DSCP: { uint32_t *p; uint16_t x; p = ((ipfw_insn_u32 *)cmd)->d; if (is_ipv4) x = ip->ip_tos >> 2; else if (is_ipv6) { uint8_t *v; v = &((struct ip6_hdr *)ip)->ip6_vfc; x = (*v & 0x0F) << 2; v++; x |= *v >> 6; } else break; /* DSCP bitmask is stored as low_u32 high_u32 */ if (x > 32) match = *(p + 1) & (1 << (x - 32)); else match = *p & (1 << x); } break; case O_TCPDATALEN: if (proto == IPPROTO_TCP && offset == 0) { struct tcphdr *tcp; uint16_t x; uint16_t *p; int i; tcp = TCP(ulp); x = iplen - ((ip->ip_hl + tcp->th_off) << 2); if (cmdlen == 1) { match = (cmd->arg1 == x); break; } /* otherwise we have ranges */ p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for (; !match && i>0; i--, p += 2) match = (x >= p[0] && x <= p[1]); } break; case O_TCPFLAGS: match = (proto == IPPROTO_TCP && offset == 0 && flags_match(cmd, TCP(ulp)->th_flags)); break; case O_TCPOPTS: if (proto == IPPROTO_TCP && offset == 0 && ulp){ PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2)); match = tcpopts_match(TCP(ulp), cmd); } break; case O_TCPSEQ: match = (proto == IPPROTO_TCP && offset == 0 && ((ipfw_insn_u32 *)cmd)->d[0] == TCP(ulp)->th_seq); break; case O_TCPACK: match = (proto == IPPROTO_TCP && offset == 0 && ((ipfw_insn_u32 *)cmd)->d[0] == TCP(ulp)->th_ack); break; case O_TCPWIN: if (proto == IPPROTO_TCP && offset == 0) { uint16_t x; uint16_t *p; int i; x = ntohs(TCP(ulp)->th_win); if (cmdlen == 1) { match = (cmd->arg1 == x); break; } /* Otherwise we have ranges. */ p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for (; !match && i > 0; i--, p += 2) match = (x >= p[0] && x <= p[1]); } break; case O_ESTAB: /* reject packets which have SYN only */ /* XXX should i also check for TH_ACK ? */ match = (proto == IPPROTO_TCP && offset == 0 && (TCP(ulp)->th_flags & (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); break; case O_ALTQ: { struct pf_mtag *at; struct m_tag *mtag; ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; /* * ALTQ uses mbuf tags from another * packet filtering system - pf(4). * We allocate a tag in its format * and fill it in, pretending to be pf(4). */ match = 1; at = pf_find_mtag(m); if (at != NULL && at->qid != 0) break; mtag = m_tag_get(PACKET_TAG_PF, sizeof(struct pf_mtag), M_NOWAIT | M_ZERO); if (mtag == NULL) { /* * Let the packet fall back to the * default ALTQ. */ break; } m_tag_prepend(m, mtag); at = (struct pf_mtag *)(mtag + 1); at->qid = altq->qid; at->hdr = ip; break; } case O_LOG: ipfw_log(chain, f, hlen, args, m, oif, offset | ip6f_mf, tablearg, ip); match = 1; break; case O_PROB: match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); break; case O_VERREVPATH: /* Outgoing packets automatically pass/match */ match = ((oif != NULL) || (m->m_pkthdr.rcvif == NULL) || ( #ifdef INET6 is_ipv6 ? verify_path6(&(args->f_id.src_ip6), m->m_pkthdr.rcvif, args->f_id.fib) : #endif verify_path(src_ip, m->m_pkthdr.rcvif, args->f_id.fib))); break; case O_VERSRCREACH: /* Outgoing packets automatically pass/match */ match = (hlen > 0 && ((oif != NULL) || #ifdef INET6 is_ipv6 ? verify_path6(&(args->f_id.src_ip6), NULL, args->f_id.fib) : #endif verify_path(src_ip, NULL, args->f_id.fib))); break; case O_ANTISPOOF: /* Outgoing packets automatically pass/match */ if (oif == NULL && hlen > 0 && ( (is_ipv4 && in_localaddr(src_ip)) #ifdef INET6 || (is_ipv6 && in6_localaddr(&(args->f_id.src_ip6))) #endif )) match = #ifdef INET6 is_ipv6 ? verify_path6( &(args->f_id.src_ip6), m->m_pkthdr.rcvif, args->f_id.fib) : #endif verify_path(src_ip, m->m_pkthdr.rcvif, args->f_id.fib); else match = 1; break; case O_IPSEC: #ifdef IPSEC match = (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); #endif /* otherwise no match */ break; #ifdef INET6 case O_IP6_SRC: match = is_ipv6 && IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, &((ipfw_insn_ip6 *)cmd)->addr6); break; case O_IP6_DST: match = is_ipv6 && IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, &((ipfw_insn_ip6 *)cmd)->addr6); break; case O_IP6_SRC_MASK: case O_IP6_DST_MASK: if (is_ipv6) { int i = cmdlen - 1; struct in6_addr p; struct in6_addr *d = &((ipfw_insn_ip6 *)cmd)->addr6; for (; !match && i > 0; d += 2, i -= F_INSN_SIZE(struct in6_addr) * 2) { p = (cmd->opcode == O_IP6_SRC_MASK) ? args->f_id.src_ip6: args->f_id.dst_ip6; APPLY_MASK(&p, &d[1]); match = IN6_ARE_ADDR_EQUAL(&d[0], &p); } } break; case O_FLOW6ID: match = is_ipv6 && flow6id_match(args->f_id.flow_id6, (ipfw_insn_u32 *) cmd); break; case O_EXT_HDR: match = is_ipv6 && (ext_hd & ((ipfw_insn *) cmd)->arg1); break; case O_IP6: match = is_ipv6; break; #endif case O_IP4: match = is_ipv4; break; case O_TAG: { struct m_tag *mtag; uint32_t tag = TARG(cmd->arg1, tag); /* Packet is already tagged with this tag? */ mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); /* We have `untag' action when F_NOT flag is * present. And we must remove this mtag from * mbuf and reset `match' to zero (`match' will * be inversed later). * Otherwise we should allocate new mtag and * push it into mbuf. */ if (cmd->len & F_NOT) { /* `untag' action */ if (mtag != NULL) m_tag_delete(m, mtag); match = 0; } else { if (mtag == NULL) { mtag = m_tag_alloc( MTAG_IPFW, tag, 0, M_NOWAIT); if (mtag != NULL) m_tag_prepend(m, mtag); } match = 1; } break; } case O_FIB: /* try match the specified fib */ if (args->f_id.fib == cmd->arg1) match = 1; break; case O_SOCKARG: { #ifndef USERSPACE /* not supported in userspace */ struct inpcb *inp = args->inp; struct inpcbinfo *pi; if (is_ipv6) /* XXX can we remove this ? */ break; if (proto == IPPROTO_TCP) pi = &V_tcbinfo; else if (proto == IPPROTO_UDP) pi = &V_udbinfo; else break; /* * XXXRW: so_user_cookie should almost * certainly be inp_user_cookie? */ /* For incomming packet, lookup up the inpcb using the src/dest ip/port tuple */ if (inp == NULL) { inp = in_pcblookup(pi, src_ip, htons(src_port), dst_ip, htons(dst_port), INPLOOKUP_RLOCKPCB, NULL); if (inp != NULL) { tablearg = inp->inp_socket->so_user_cookie; if (tablearg) match = 1; INP_RUNLOCK(inp); } } else { if (inp->inp_socket) { tablearg = inp->inp_socket->so_user_cookie; if (tablearg) match = 1; } } #endif /* !USERSPACE */ break; } case O_TAGGED: { struct m_tag *mtag; uint32_t tag = TARG(cmd->arg1, tag); if (cmdlen == 1) { match = m_tag_locate(m, MTAG_IPFW, tag, NULL) != NULL; break; } /* we have ranges */ for (mtag = m_tag_first(m); mtag != NULL && !match; mtag = m_tag_next(m, mtag)) { uint16_t *p; int i; if (mtag->m_tag_cookie != MTAG_IPFW) continue; p = ((ipfw_insn_u16 *)cmd)->ports; i = cmdlen - 1; for(; !match && i > 0; i--, p += 2) match = mtag->m_tag_id >= p[0] && mtag->m_tag_id <= p[1]; } break; } /* * The second set of opcodes represents 'actions', * i.e. the terminal part of a rule once the packet * matches all previous patterns. * Typically there is only one action for each rule, * and the opcode is stored at the end of the rule * (but there are exceptions -- see below). * * In general, here we set retval and terminate the * outer loop (would be a 'break 3' in some language, * but we need to set l=0, done=1) * * Exceptions: * O_COUNT and O_SKIPTO actions: * instead of terminating, we jump to the next rule * (setting l=0), or to the SKIPTO target (setting * f/f_len, cmd and l as needed), respectively. * * O_TAG, O_LOG and O_ALTQ action parameters: * perform some action and set match = 1; * * O_LIMIT and O_KEEP_STATE: these opcodes are * not real 'actions', and are stored right * before the 'action' part of the rule. * These opcodes try to install an entry in the * state tables; if successful, we continue with * the next opcode (match=1; break;), otherwise * the packet must be dropped (set retval, * break loops with l=0, done=1) * * O_PROBE_STATE and O_CHECK_STATE: these opcodes * cause a lookup of the state table, and a jump * to the 'action' part of the parent rule * if an entry is found, or * (CHECK_STATE only) a jump to the next rule if * the entry is not found. * The result of the lookup is cached so that * further instances of these opcodes become NOPs. * The jump to the next rule is done by setting * l=0, cmdlen=0. */ case O_LIMIT: case O_KEEP_STATE: if (ipfw_install_state(chain, f, (ipfw_insn_limit *)cmd, args, tablearg)) { /* error or limit violation */ retval = IP_FW_DENY; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ } match = 1; break; case O_PROBE_STATE: case O_CHECK_STATE: /* * dynamic rules are checked at the first * keep-state or check-state occurrence, * with the result being stored in dyn_dir. * The compiler introduces a PROBE_STATE * instruction for us when we have a * KEEP_STATE (because PROBE_STATE needs * to be run first). */ if (dyn_dir == MATCH_UNKNOWN && (q = ipfw_lookup_dyn_rule(&args->f_id, &dyn_dir, proto == IPPROTO_TCP ? TCP(ulp) : NULL)) != NULL) { /* * Found dynamic entry, update stats * and jump to the 'action' part of * the parent rule by setting * f, cmd, l and clearing cmdlen. */ IPFW_INC_DYN_COUNTER(q, pktlen); /* XXX we would like to have f_pos * readily accessible in the dynamic * rule, instead of having to * lookup q->rule. */ f = q->rule; f_pos = ipfw_find_rule(chain, f->rulenum, f->id); cmd = ACTION_PTR(f); l = f->cmd_len - f->act_ofs; ipfw_dyn_unlock(q); cmdlen = 0; match = 1; break; } /* * Dynamic entry not found. If CHECK_STATE, * skip to next rule, if PROBE_STATE just * ignore and continue with next opcode. */ if (cmd->opcode == O_CHECK_STATE) l = 0; /* exit inner loop */ match = 1; break; case O_ACCEPT: retval = 0; /* accept */ l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_PIPE: case O_QUEUE: set_match(args, f_pos, chain); args->rule.info = TARG(cmd->arg1, pipe); if (cmd->opcode == O_PIPE) args->rule.info |= IPFW_IS_PIPE; if (V_fw_one_pass) args->rule.info |= IPFW_ONEPASS; retval = IP_FW_DUMMYNET; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_DIVERT: case O_TEE: if (args->eh) /* not on layer 2 */ break; /* otherwise this is terminal */ l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ retval = (cmd->opcode == O_DIVERT) ? IP_FW_DIVERT : IP_FW_TEE; set_match(args, f_pos, chain); args->rule.info = TARG(cmd->arg1, divert); break; case O_COUNT: IPFW_INC_RULE_COUNTER(f, pktlen); l = 0; /* exit inner loop */ break; case O_SKIPTO: IPFW_INC_RULE_COUNTER(f, pktlen); f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0); /* * Skip disabled rules, and re-enter * the inner loop with the correct * f_pos, f, l and cmd. * Also clear cmdlen and skip_or */ for (; f_pos < chain->n_rules - 1 && (V_set_disable & (1 << chain->map[f_pos]->set)); f_pos++) ; /* Re-enter the inner loop at the skipto rule. */ f = chain->map[f_pos]; l = f->cmd_len; cmd = f->cmd; match = 1; cmdlen = 0; skip_or = 0; continue; break; /* not reached */ case O_CALLRETURN: { /* * Implementation of `subroutine' call/return, * in the stack carried in an mbuf tag. This * is different from `skipto' in that any call * address is possible (`skipto' must prevent * backward jumps to avoid endless loops). * We have `return' action when F_NOT flag is * present. The `m_tag_id' field is used as * stack pointer. */ struct m_tag *mtag; uint16_t jmpto, *stack; #define IS_CALL ((cmd->len & F_NOT) == 0) #define IS_RETURN ((cmd->len & F_NOT) != 0) /* * Hand-rolled version of m_tag_locate() with * wildcard `type'. * If not already tagged, allocate new tag. */ mtag = m_tag_first(m); while (mtag != NULL) { if (mtag->m_tag_cookie == MTAG_IPFW_CALL) break; mtag = m_tag_next(m, mtag); } if (mtag == NULL && IS_CALL) { mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, IPFW_CALLSTACK_SIZE * sizeof(uint16_t), M_NOWAIT); if (mtag != NULL) m_tag_prepend(m, mtag); } /* * On error both `call' and `return' just * continue with next rule. */ if (IS_RETURN && (mtag == NULL || mtag->m_tag_id == 0)) { l = 0; /* exit inner loop */ break; } if (IS_CALL && (mtag == NULL || mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) { printf("ipfw: call stack error, " "go to next rule\n"); l = 0; /* exit inner loop */ break; } IPFW_INC_RULE_COUNTER(f, pktlen); stack = (uint16_t *)(mtag + 1); /* * The `call' action may use cached f_pos * (in f->next_rule), whose version is written * in f->next_rule. * The `return' action, however, doesn't have * fixed jump address in cmd->arg1 and can't use * cache. */ if (IS_CALL) { stack[mtag->m_tag_id] = f->rulenum; mtag->m_tag_id++; f_pos = JUMP(chain, f, cmd->arg1, tablearg, 1); } else { /* `return' action */ mtag->m_tag_id--; jmpto = stack[mtag->m_tag_id] + 1; f_pos = ipfw_find_rule(chain, jmpto, 0); } /* * Skip disabled rules, and re-enter * the inner loop with the correct * f_pos, f, l and cmd. * Also clear cmdlen and skip_or */ for (; f_pos < chain->n_rules - 1 && (V_set_disable & (1 << chain->map[f_pos]->set)); f_pos++) ; /* Re-enter the inner loop at the dest rule. */ f = chain->map[f_pos]; l = f->cmd_len; cmd = f->cmd; cmdlen = 0; skip_or = 0; continue; break; /* NOTREACHED */ } #undef IS_CALL #undef IS_RETURN case O_REJECT: /* * Drop the packet and send a reject notice * if the packet is not ICMP (or is an ICMP * query), and it is not multicast/broadcast. */ if (hlen > 0 && is_ipv4 && offset == 0 && (proto != IPPROTO_ICMP || is_icmp_query(ICMP(ulp))) && !(m->m_flags & (M_BCAST|M_MCAST)) && !IN_MULTICAST(ntohl(dst_ip.s_addr))) { send_reject(args, cmd->arg1, iplen, ip); m = args->m; } /* FALLTHROUGH */ #ifdef INET6 case O_UNREACH6: if (hlen > 0 && is_ipv6 && ((offset & IP6F_OFF_MASK) == 0) && (proto != IPPROTO_ICMPV6 || (is_icmp6_query(icmp6_type) == 1)) && !(m->m_flags & (M_BCAST|M_MCAST)) && !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { send_reject6( args, cmd->arg1, hlen, (struct ip6_hdr *)ip); m = args->m; } /* FALLTHROUGH */ #endif case O_DENY: retval = IP_FW_DENY; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_FORWARD_IP: if (args->eh) /* not valid on layer2 pkts */ break; if (q == NULL || q->rule != f || dyn_dir == MATCH_FORWARD) { struct sockaddr_in *sa; sa = &(((ipfw_insn_sa *)cmd)->sa); if (sa->sin_addr.s_addr == INADDR_ANY) { #ifdef INET6 /* * We use O_FORWARD_IP opcode for * fwd rule with tablearg, but tables * now support IPv6 addresses. And * when we are inspecting IPv6 packet, * we can use nh6 field from * table_value as next_hop6 address. */ if (is_ipv6) { struct sockaddr_in6 *sa6; sa6 = args->next_hop6 = &args->hopstore6; sa6->sin6_family = AF_INET6; sa6->sin6_len = sizeof(*sa6); sa6->sin6_addr = TARG_VAL( chain, tablearg, nh6); /* * Set sin6_scope_id only for * link-local unicast addresses. */ if (IN6_IS_ADDR_LINKLOCAL( &sa6->sin6_addr)) sa6->sin6_scope_id = TARG_VAL(chain, tablearg, zoneid); } else #endif { sa = args->next_hop = &args->hopstore; sa->sin_family = AF_INET; sa->sin_len = sizeof(*sa); sa->sin_addr.s_addr = htonl( TARG_VAL(chain, tablearg, nh4)); } } else { args->next_hop = sa; } } retval = IP_FW_PASS; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; #ifdef INET6 case O_FORWARD_IP6: if (args->eh) /* not valid on layer2 pkts */ break; if (q == NULL || q->rule != f || dyn_dir == MATCH_FORWARD) { struct sockaddr_in6 *sin6; sin6 = &(((ipfw_insn_sa6 *)cmd)->sa); args->next_hop6 = sin6; } retval = IP_FW_PASS; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; #endif case O_NETGRAPH: case O_NGTEE: set_match(args, f_pos, chain); args->rule.info = TARG(cmd->arg1, netgraph); if (V_fw_one_pass) args->rule.info |= IPFW_ONEPASS; retval = (cmd->opcode == O_NETGRAPH) ? IP_FW_NETGRAPH : IP_FW_NGTEE; l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ break; case O_SETFIB: { uint32_t fib; IPFW_INC_RULE_COUNTER(f, pktlen); fib = TARG(cmd->arg1, fib) & 0x7FFF; if (fib >= rt_numfibs) fib = 0; M_SETFIB(m, fib); args->f_id.fib = fib; l = 0; /* exit inner loop */ break; } case O_SETDSCP: { uint16_t code; code = TARG(cmd->arg1, dscp) & 0x3F; l = 0; /* exit inner loop */ if (is_ipv4) { uint16_t old; old = *(uint16_t *)ip; ip->ip_tos = (code << 2) | (ip->ip_tos & 0x03); ip->ip_sum = cksum_adjust(ip->ip_sum, old, *(uint16_t *)ip); } else if (is_ipv6) { uint8_t *v; v = &((struct ip6_hdr *)ip)->ip6_vfc; *v = (*v & 0xF0) | (code >> 2); v++; *v = (*v & 0x3F) | ((code & 0x03) << 6); } else break; IPFW_INC_RULE_COUNTER(f, pktlen); break; } case O_NAT: l = 0; /* exit inner loop */ done = 1; /* exit outer loop */ if (!IPFW_NAT_LOADED) { retval = IP_FW_DENY; break; } struct cfg_nat *t; int nat_id; set_match(args, f_pos, chain); /* Check if this is 'global' nat rule */ if (cmd->arg1 == 0) { retval = ipfw_nat_ptr(args, NULL, m); break; } t = ((ipfw_insn_nat *)cmd)->nat; if (t == NULL) { nat_id = TARG(cmd->arg1, nat); t = (*lookup_nat_ptr)(&chain->nat, nat_id); if (t == NULL) { retval = IP_FW_DENY; break; } if (cmd->arg1 != IP_FW_TARG) ((ipfw_insn_nat *)cmd)->nat = t; } retval = ipfw_nat_ptr(args, t, m); break; case O_REASS: { int ip_off; IPFW_INC_RULE_COUNTER(f, pktlen); l = 0; /* in any case exit inner loop */ ip_off = ntohs(ip->ip_off); /* if not fragmented, go to next rule */ if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) break; args->m = m = ip_reass(m); /* * do IP header checksum fixup. */ if (m == NULL) { /* fragment got swallowed */ retval = IP_FW_DENY; } else { /* good, packet complete */ int hlen; ip = mtod(m, struct ip *); hlen = ip->ip_hl << 2; ip->ip_sum = 0; if (hlen == sizeof(struct ip)) ip->ip_sum = in_cksum_hdr(ip); else ip->ip_sum = in_cksum(m, hlen); retval = IP_FW_REASS; set_match(args, f_pos, chain); } done = 1; /* exit outer loop */ break; } default: panic("-- unknown opcode %d\n", cmd->opcode); } /* end of switch() on opcodes */ /* * if we get here with l=0, then match is irrelevant. */ if (cmd->len & F_NOT) match = !match; if (match) { if (cmd->len & F_OR) skip_or = 1; } else { if (!(cmd->len & F_OR)) /* not an OR block, */ break; /* try next rule */ } } /* end of inner loop, scan opcodes */ #undef PULLUP_LEN if (done) break; /* next_rule:; */ /* try next rule */ } /* end of outer for, scan rules */ if (done) { struct ip_fw *rule = chain->map[f_pos]; /* Update statistics */ IPFW_INC_RULE_COUNTER(rule, pktlen); } else { retval = IP_FW_DENY; printf("ipfw: ouch!, skip past end of rules, denying packet\n"); } IPFW_PF_RUNLOCK(chain); #ifdef __FreeBSD__ if (ucred_cache != NULL) crfree(ucred_cache); #endif return (retval); pullup_failed: if (V_fw_verbose) printf("ipfw: pullup failed\n"); return (IP_FW_DENY); } /* * Set maximum number of tables that can be used in given VNET ipfw instance. */ #ifdef SYSCTL_NODE static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS) { int error; unsigned int ntables; ntables = V_fw_tables_max; error = sysctl_handle_int(oidp, &ntables, 0, req); /* Read operation or some error */ if ((error != 0) || (req->newptr == NULL)) return (error); return (ipfw_resize_tables(&V_layer3_chain, ntables)); } /* * Switches table namespace between global and per-set. */ static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS) { int error; unsigned int sets; sets = V_fw_tables_sets; error = sysctl_handle_int(oidp, &sets, 0, req); /* Read operation or some error */ if ((error != 0) || (req->newptr == NULL)) return (error); return (ipfw_switch_tables_namespace(&V_layer3_chain, sets)); } #endif /* * Module and VNET glue */ /* * Stuff that must be initialised only on boot or module load */ static int ipfw_init(void) { int error = 0; /* * Only print out this stuff the first time around, * when called from the sysinit code. */ printf("ipfw2 " #ifdef INET6 "(+ipv6) " #endif "initialized, divert %s, nat %s, " "default to %s, logging ", #ifdef IPDIVERT "enabled", #else "loadable", #endif #ifdef IPFIREWALL_NAT "enabled", #else "loadable", #endif default_to_accept ? "accept" : "deny"); /* * Note: V_xxx variables can be accessed here but the vnet specific * initializer may not have been called yet for the VIMAGE case. * Tuneables will have been processed. We will print out values for * the default vnet. * XXX This should all be rationalized AFTER 8.0 */ if (V_fw_verbose == 0) printf("disabled\n"); else if (V_verbose_limit == 0) printf("unlimited\n"); else printf("limited to %d packets/entry by default\n", V_verbose_limit); /* Check user-supplied table count for validness */ if (default_fw_tables > IPFW_TABLES_MAX) default_fw_tables = IPFW_TABLES_MAX; ipfw_init_sopt_handler(); ipfw_log_bpf(1); /* init */ ipfw_iface_init(); return (error); } /* * Called for the removal of the last instance only on module unload. */ static void ipfw_destroy(void) { ipfw_iface_destroy(); ipfw_log_bpf(0); /* uninit */ ipfw_destroy_sopt_handler(); printf("IP firewall unloaded\n"); } /* * Stuff that must be initialized for every instance * (including the first of course). */ static int vnet_ipfw_init(const void *unused) { int error, first; struct ip_fw *rule = NULL; struct ip_fw_chain *chain; chain = &V_layer3_chain; first = IS_DEFAULT_VNET(curvnet) ? 1 : 0; /* First set up some values that are compile time options */ V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ V_fw_deny_unknown_exthdrs = 1; #ifdef IPFIREWALL_VERBOSE V_fw_verbose = 1; #endif #ifdef IPFIREWALL_VERBOSE_LIMIT V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; #endif #ifdef IPFIREWALL_NAT LIST_INIT(&chain->nat); #endif /* Init shared services hash table */ ipfw_init_srv(chain); ipfw_init_obj_rewriter(); ipfw_init_counters(); /* insert the default rule and create the initial map */ chain->n_rules = 1; chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO); rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw)); /* Set initial number of tables */ V_fw_tables_max = default_fw_tables; error = ipfw_init_tables(chain, first); if (error) { printf("ipfw2: setting up tables failed\n"); free(chain->map, M_IPFW); free(rule, M_IPFW); return (ENOSPC); } /* fill and insert the default rule */ rule->act_ofs = 0; rule->rulenum = IPFW_DEFAULT_RULE; rule->cmd_len = 1; rule->set = RESVD_SET; rule->cmd[0].len = 1; rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; chain->default_rule = chain->map[0] = rule; chain->id = rule->id = 1; /* Pre-calculate rules length for legacy dump format */ chain->static_len = sizeof(struct ip_fw_rule0); IPFW_LOCK_INIT(chain); ipfw_dyn_init(chain); #ifdef LINEAR_SKIPTO ipfw_init_skipto_cache(chain); #endif /* First set up some values that are compile time options */ V_ipfw_vnet_ready = 1; /* Open for business */ /* * Hook the sockopt handler and pfil hooks for ipv4 and ipv6. * Even if the latter two fail we still keep the module alive * because the sockopt and layer2 paths are still useful. * ipfw[6]_hook return 0 on success, ENOENT on failure, * so we can ignore the exact return value and just set a flag. * * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so * changes in the underlying (per-vnet) variables trigger * immediate hook()/unhook() calls. * In layer2 we have the same behaviour, except that V_ether_ipfw * is checked on each packet because there are no pfil hooks. */ V_ip_fw_ctl_ptr = ipfw_ctl3; error = ipfw_attach_hooks(1); return (error); } /* * Called for the removal of each instance. */ static int vnet_ipfw_uninit(const void *unused) { struct ip_fw *reap; struct ip_fw_chain *chain = &V_layer3_chain; int i, last; V_ipfw_vnet_ready = 0; /* tell new callers to go away */ /* * disconnect from ipv4, ipv6, layer2 and sockopt. * Then grab, release and grab again the WLOCK so we make * sure the update is propagated and nobody will be in. */ (void)ipfw_attach_hooks(0 /* detach */); V_ip_fw_ctl_ptr = NULL; last = IS_DEFAULT_VNET(curvnet) ? 1 : 0; IPFW_UH_WLOCK(chain); IPFW_UH_WUNLOCK(chain); ipfw_dyn_uninit(0); /* run the callout_drain */ IPFW_UH_WLOCK(chain); reap = NULL; IPFW_WLOCK(chain); for (i = 0; i < chain->n_rules; i++) ipfw_reap_add(chain, &reap, chain->map[i]); free(chain->map, M_IPFW); #ifdef LINEAR_SKIPTO ipfw_destroy_skipto_cache(chain); #endif IPFW_WUNLOCK(chain); IPFW_UH_WUNLOCK(chain); ipfw_destroy_tables(chain, last); if (reap != NULL) ipfw_reap_rules(reap); vnet_ipfw_iface_destroy(chain); ipfw_destroy_srv(chain); IPFW_LOCK_DESTROY(chain); ipfw_dyn_uninit(1); /* free the remaining parts */ ipfw_destroy_counters(); ipfw_destroy_obj_rewriter(); return (0); } /* * Module event handler. * In general we have the choice of handling most of these events by the * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to * use the SYSINIT handlers as they are more capable of expressing the * flow of control during module and vnet operations, so this is just * a skeleton. Note there is no SYSINIT equivalent of the module * SHUTDOWN handler, but we don't have anything to do in that case anyhow. */ static int ipfw_modevent(module_t mod, int type, void *unused) { int err = 0; switch (type) { case MOD_LOAD: /* Called once at module load or * system boot if compiled in. */ break; case MOD_QUIESCE: /* Called before unload. May veto unloading. */ break; case MOD_UNLOAD: /* Called during unload. */ break; case MOD_SHUTDOWN: /* Called during system shutdown. */ break; default: err = EOPNOTSUPP; break; } return err; } static moduledata_t ipfwmod = { "ipfw", ipfw_modevent, 0 }; /* Define startup order. */ -#define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN +#define IPFW_SI_SUB_FIREWALL SI_SUB_FW #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); FEATURE(ipfw_ctl3, "ipfw new sockopt calls"); MODULE_VERSION(ipfw, 3); /* should declare some dependencies here */ /* * Starting up. Done in order after ipfwmod() has been called. * VNET_SYSINIT is also called for each existing vnet and each new vnet. */ SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, ipfw_init, NULL); VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, vnet_ipfw_init, NULL); /* * Closing up shop. These are done in REVERSE ORDER, but still * after ipfwmod() has been called. Not called on reboot. * VNET_SYSUNINIT is also called for each exiting vnet as it exits. * or when the module is unloaded. */ SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, ipfw_destroy, NULL); VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, vnet_ipfw_uninit, NULL); /* end of file */ Index: projects/vnet/sys/netpfil/ipfw/ip_fw_nat.c =================================================================== --- projects/vnet/sys/netpfil/ipfw/ip_fw_nat.c (revision 294565) +++ projects/vnet/sys/netpfil/ipfw/ip_fw_nat.c (revision 294566) @@ -1,1236 +1,1236 @@ /*- * Copyright (c) 2008 Paolo Pisati * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* XXX for in_cksum */ struct cfg_spool { LIST_ENTRY(cfg_spool) _next; /* chain of spool instances */ struct in_addr addr; uint16_t port; }; /* Nat redirect configuration. */ struct cfg_redir { LIST_ENTRY(cfg_redir) _next; /* chain of redir instances */ uint16_t mode; /* type of redirect mode */ uint16_t proto; /* protocol: tcp/udp */ struct in_addr laddr; /* local ip address */ struct in_addr paddr; /* public ip address */ struct in_addr raddr; /* remote ip address */ uint16_t lport; /* local port */ uint16_t pport; /* public port */ uint16_t rport; /* remote port */ uint16_t pport_cnt; /* number of public ports */ uint16_t rport_cnt; /* number of remote ports */ struct alias_link **alink; u_int16_t spool_cnt; /* num of entry in spool chain */ /* chain of spool instances */ LIST_HEAD(spool_chain, cfg_spool) spool_chain; }; /* Nat configuration data struct. */ struct cfg_nat { /* chain of nat instances */ LIST_ENTRY(cfg_nat) _next; int id; /* nat id */ struct in_addr ip; /* nat ip address */ struct libalias *lib; /* libalias instance */ int mode; /* aliasing mode */ int redir_cnt; /* number of entry in spool chain */ /* chain of redir instances */ LIST_HEAD(redir_chain, cfg_redir) redir_chain; char if_name[IF_NAMESIZE]; /* interface name */ }; static eventhandler_tag ifaddr_event_tag; static void ifaddr_change(void *arg __unused, struct ifnet *ifp) { struct cfg_nat *ptr; struct ifaddr *ifa; struct ip_fw_chain *chain; KASSERT(curvnet == ifp->if_vnet, ("curvnet(%p) differs from iface vnet(%p)", curvnet, ifp->if_vnet)); chain = &V_layer3_chain; IPFW_UH_WLOCK(chain); /* Check every nat entry... */ LIST_FOREACH(ptr, &chain->nat, _next) { /* ...using nic 'ifp->if_xname' as dynamic alias address. */ if (strncmp(ptr->if_name, ifp->if_xname, IF_NAMESIZE) != 0) continue; if_addr_rlock(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr == NULL) continue; if (ifa->ifa_addr->sa_family != AF_INET) continue; IPFW_WLOCK(chain); ptr->ip = ((struct sockaddr_in *) (ifa->ifa_addr))->sin_addr; LibAliasSetAddress(ptr->lib, ptr->ip); IPFW_WUNLOCK(chain); } if_addr_runlock(ifp); } IPFW_UH_WUNLOCK(chain); } /* * delete the pointers for nat entry ix, or all of them if ix < 0 */ static void flush_nat_ptrs(struct ip_fw_chain *chain, const int ix) { int i; ipfw_insn_nat *cmd; IPFW_WLOCK_ASSERT(chain); for (i = 0; i < chain->n_rules; i++) { cmd = (ipfw_insn_nat *)ACTION_PTR(chain->map[i]); /* XXX skip log and the like ? */ if (cmd->o.opcode == O_NAT && cmd->nat != NULL && (ix < 0 || cmd->nat->id == ix)) cmd->nat = NULL; } } static void del_redir_spool_cfg(struct cfg_nat *n, struct redir_chain *head) { struct cfg_redir *r, *tmp_r; struct cfg_spool *s, *tmp_s; int i, num; LIST_FOREACH_SAFE(r, head, _next, tmp_r) { num = 1; /* Number of alias_link to delete. */ switch (r->mode) { case NAT44_REDIR_PORT: num = r->pport_cnt; /* FALLTHROUGH */ case NAT44_REDIR_ADDR: case NAT44_REDIR_PROTO: /* Delete all libalias redirect entry. */ for (i = 0; i < num; i++) LibAliasRedirectDelete(n->lib, r->alink[i]); /* Del spool cfg if any. */ LIST_FOREACH_SAFE(s, &r->spool_chain, _next, tmp_s) { LIST_REMOVE(s, _next); free(s, M_IPFW); } free(r->alink, M_IPFW); LIST_REMOVE(r, _next); free(r, M_IPFW); break; default: printf("unknown redirect mode: %u\n", r->mode); /* XXX - panic?!?!? */ break; } } } static int add_redir_spool_cfg(char *buf, struct cfg_nat *ptr) { struct cfg_redir *r; struct cfg_spool *s; struct nat44_cfg_redir *ser_r; struct nat44_cfg_spool *ser_s; int cnt, off, i; for (cnt = 0, off = 0; cnt < ptr->redir_cnt; cnt++) { ser_r = (struct nat44_cfg_redir *)&buf[off]; r = malloc(sizeof(*r), M_IPFW, M_WAITOK | M_ZERO); r->mode = ser_r->mode; r->laddr = ser_r->laddr; r->paddr = ser_r->paddr; r->raddr = ser_r->raddr; r->lport = ser_r->lport; r->pport = ser_r->pport; r->rport = ser_r->rport; r->pport_cnt = ser_r->pport_cnt; r->rport_cnt = ser_r->rport_cnt; r->proto = ser_r->proto; r->spool_cnt = ser_r->spool_cnt; //memcpy(r, ser_r, SOF_REDIR); LIST_INIT(&r->spool_chain); off += sizeof(struct nat44_cfg_redir); r->alink = malloc(sizeof(struct alias_link *) * r->pport_cnt, M_IPFW, M_WAITOK | M_ZERO); switch (r->mode) { case NAT44_REDIR_ADDR: r->alink[0] = LibAliasRedirectAddr(ptr->lib, r->laddr, r->paddr); break; case NAT44_REDIR_PORT: for (i = 0 ; i < r->pport_cnt; i++) { /* If remotePort is all ports, set it to 0. */ u_short remotePortCopy = r->rport + i; if (r->rport_cnt == 1 && r->rport == 0) remotePortCopy = 0; r->alink[i] = LibAliasRedirectPort(ptr->lib, r->laddr, htons(r->lport + i), r->raddr, htons(remotePortCopy), r->paddr, htons(r->pport + i), r->proto); if (r->alink[i] == NULL) { r->alink[0] = NULL; break; } } break; case NAT44_REDIR_PROTO: r->alink[0] = LibAliasRedirectProto(ptr->lib ,r->laddr, r->raddr, r->paddr, r->proto); break; default: printf("unknown redirect mode: %u\n", r->mode); break; } if (r->alink[0] == NULL) { printf("LibAliasRedirect* returned NULL\n"); free(r->alink, M_IPFW); free(r, M_IPFW); return (EINVAL); } /* LSNAT handling. */ for (i = 0; i < r->spool_cnt; i++) { ser_s = (struct nat44_cfg_spool *)&buf[off]; s = malloc(sizeof(*s), M_IPFW, M_WAITOK | M_ZERO); s->addr = ser_s->addr; s->port = ser_s->port; LibAliasAddServer(ptr->lib, r->alink[0], s->addr, htons(s->port)); off += sizeof(struct nat44_cfg_spool); /* Hook spool entry. */ LIST_INSERT_HEAD(&r->spool_chain, s, _next); } /* And finally hook this redir entry. */ LIST_INSERT_HEAD(&ptr->redir_chain, r, _next); } return (0); } static void free_nat_instance(struct cfg_nat *ptr) { del_redir_spool_cfg(ptr, &ptr->redir_chain); LibAliasUninit(ptr->lib); free(ptr, M_IPFW); } /* * ipfw_nat - perform mbuf header translation. * * Note V_layer3_chain has to be locked while calling ipfw_nat() in * 'global' operation mode (t == NULL). * */ static int ipfw_nat(struct ip_fw_args *args, struct cfg_nat *t, struct mbuf *m) { struct mbuf *mcl; struct ip *ip; /* XXX - libalias duct tape */ int ldt, retval, found; struct ip_fw_chain *chain; char *c; ldt = 0; retval = 0; mcl = m_megapullup(m, m->m_pkthdr.len); if (mcl == NULL) { args->m = NULL; return (IP_FW_DENY); } ip = mtod(mcl, struct ip *); /* * XXX - Libalias checksum offload 'duct tape': * * locally generated packets have only pseudo-header checksum * calculated and libalias will break it[1], so mark them for * later fix. Moreover there are cases when libalias modifies * tcp packet data[2], mark them for later fix too. * * [1] libalias was never meant to run in kernel, so it does * not have any knowledge about checksum offloading, and * expects a packet with a full internet checksum. * Unfortunately, packets generated locally will have just the * pseudo header calculated, and when libalias tries to adjust * the checksum it will actually compute a wrong value. * * [2] when libalias modifies tcp's data content, full TCP * checksum has to be recomputed: the problem is that * libalias does not have any idea about checksum offloading. * To work around this, we do not do checksumming in LibAlias, * but only mark the packets in th_x2 field. If we receive a * marked packet, we calculate correct checksum for it * aware of offloading. Why such a terrible hack instead of * recalculating checksum for each packet? * Because the previous checksum was not checked! * Recalculating checksums for EVERY packet will hide ALL * transmission errors. Yes, marked packets still suffer from * this problem. But, sigh, natd(8) has this problem, too. * * TODO: -make libalias mbuf aware (so * it can handle delayed checksum and tso) */ if (mcl->m_pkthdr.rcvif == NULL && mcl->m_pkthdr.csum_flags & CSUM_DELAY_DATA) ldt = 1; c = mtod(mcl, char *); /* Check if this is 'global' instance */ if (t == NULL) { if (args->oif == NULL) { /* Wrong direction, skip processing */ args->m = mcl; return (IP_FW_NAT); } found = 0; chain = &V_layer3_chain; IPFW_RLOCK_ASSERT(chain); /* Check every nat entry... */ LIST_FOREACH(t, &chain->nat, _next) { if ((t->mode & PKT_ALIAS_SKIP_GLOBAL) != 0) continue; retval = LibAliasOutTry(t->lib, c, mcl->m_len + M_TRAILINGSPACE(mcl), 0); if (retval == PKT_ALIAS_OK) { /* Nat instance recognises state */ found = 1; break; } } if (found != 1) { /* No instance found, return ignore */ args->m = mcl; return (IP_FW_NAT); } } else { if (args->oif == NULL) retval = LibAliasIn(t->lib, c, mcl->m_len + M_TRAILINGSPACE(mcl)); else retval = LibAliasOut(t->lib, c, mcl->m_len + M_TRAILINGSPACE(mcl)); } /* * We drop packet when: * 1. libalias returns PKT_ALIAS_ERROR; * 2. For incoming packets: * a) for unresolved fragments; * b) libalias returns PKT_ALIAS_IGNORED and * PKT_ALIAS_DENY_INCOMING flag is set. */ if (retval == PKT_ALIAS_ERROR || (args->oif == NULL && (retval == PKT_ALIAS_UNRESOLVED_FRAGMENT || (retval == PKT_ALIAS_IGNORED && (t->mode & PKT_ALIAS_DENY_INCOMING) != 0)))) { /* XXX - should i add some logging? */ m_free(mcl); args->m = NULL; return (IP_FW_DENY); } if (retval == PKT_ALIAS_RESPOND) mcl->m_flags |= M_SKIP_FIREWALL; mcl->m_pkthdr.len = mcl->m_len = ntohs(ip->ip_len); /* * XXX - libalias checksum offload * 'duct tape' (see above) */ if ((ip->ip_off & htons(IP_OFFMASK)) == 0 && ip->ip_p == IPPROTO_TCP) { struct tcphdr *th; th = (struct tcphdr *)(ip + 1); if (th->th_x2) ldt = 1; } if (ldt) { struct tcphdr *th; struct udphdr *uh; uint16_t ip_len, cksum; ip_len = ntohs(ip->ip_len); cksum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons(ip->ip_p + ip_len - (ip->ip_hl << 2))); switch (ip->ip_p) { case IPPROTO_TCP: th = (struct tcphdr *)(ip + 1); /* * Maybe it was set in * libalias... */ th->th_x2 = 0; th->th_sum = cksum; mcl->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); break; case IPPROTO_UDP: uh = (struct udphdr *)(ip + 1); uh->uh_sum = cksum; mcl->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); break; } /* No hw checksum offloading: do it ourselves */ if ((mcl->m_pkthdr.csum_flags & CSUM_DELAY_DATA) == 0) { in_delayed_cksum(mcl); mcl->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; } } args->m = mcl; return (IP_FW_NAT); } static struct cfg_nat * lookup_nat(struct nat_list *l, int nat_id) { struct cfg_nat *res; LIST_FOREACH(res, l, _next) { if (res->id == nat_id) break; } return res; } static struct cfg_nat * lookup_nat_name(struct nat_list *l, char *name) { struct cfg_nat *res; int id; char *errptr; id = strtol(name, &errptr, 10); if (id == 0 || *errptr != '\0') return (NULL); LIST_FOREACH(res, l, _next) { if (res->id == id) break; } return (res); } /* IP_FW3 configuration routines */ static void nat44_config(struct ip_fw_chain *chain, struct nat44_cfg_nat *ucfg) { struct cfg_nat *ptr, *tcfg; int gencnt; /* * Find/create nat rule. */ IPFW_UH_WLOCK(chain); gencnt = chain->gencnt; ptr = lookup_nat_name(&chain->nat, ucfg->name); if (ptr == NULL) { IPFW_UH_WUNLOCK(chain); /* New rule: allocate and init new instance. */ ptr = malloc(sizeof(struct cfg_nat), M_IPFW, M_WAITOK | M_ZERO); ptr->lib = LibAliasInit(NULL); LIST_INIT(&ptr->redir_chain); } else { /* Entry already present: temporarily unhook it. */ IPFW_WLOCK(chain); LIST_REMOVE(ptr, _next); flush_nat_ptrs(chain, ptr->id); IPFW_WUNLOCK(chain); IPFW_UH_WUNLOCK(chain); } /* * Basic nat (re)configuration. */ ptr->id = strtol(ucfg->name, NULL, 10); /* * XXX - what if this rule doesn't nat any ip and just * redirect? * do we set aliasaddress to 0.0.0.0? */ ptr->ip = ucfg->ip; ptr->redir_cnt = ucfg->redir_cnt; ptr->mode = ucfg->mode; strlcpy(ptr->if_name, ucfg->if_name, sizeof(ptr->if_name)); LibAliasSetMode(ptr->lib, ptr->mode, ~0); LibAliasSetAddress(ptr->lib, ptr->ip); /* * Redir and LSNAT configuration. */ /* Delete old cfgs. */ del_redir_spool_cfg(ptr, &ptr->redir_chain); /* Add new entries. */ add_redir_spool_cfg((char *)(ucfg + 1), ptr); IPFW_UH_WLOCK(chain); /* Extra check to avoid race with another ipfw_nat_cfg() */ tcfg = NULL; if (gencnt != chain->gencnt) tcfg = lookup_nat_name(&chain->nat, ucfg->name); IPFW_WLOCK(chain); if (tcfg != NULL) LIST_REMOVE(tcfg, _next); LIST_INSERT_HEAD(&chain->nat, ptr, _next); IPFW_WUNLOCK(chain); chain->gencnt++; IPFW_UH_WUNLOCK(chain); if (tcfg != NULL) free_nat_instance(ptr); } /* * Creates/configure nat44 instance * Data layout (v0)(current): * Request: [ ipfw_obj_header nat44_cfg_nat .. ] * * Returns 0 on success */ static int nat44_cfg(struct ip_fw_chain *chain, ip_fw3_opheader *op3, struct sockopt_data *sd) { ipfw_obj_header *oh; struct nat44_cfg_nat *ucfg; int id; size_t read; char *errptr; /* Check minimum header size */ if (sd->valsize < (sizeof(*oh) + sizeof(*ucfg))) return (EINVAL); oh = (ipfw_obj_header *)sd->kbuf; /* Basic length checks for TLVs */ if (oh->ntlv.head.length != sizeof(oh->ntlv)) return (EINVAL); ucfg = (struct nat44_cfg_nat *)(oh + 1); /* Check if name is properly terminated and looks like number */ if (strnlen(ucfg->name, sizeof(ucfg->name)) == sizeof(ucfg->name)) return (EINVAL); id = strtol(ucfg->name, &errptr, 10); if (id == 0 || *errptr != '\0') return (EINVAL); read = sizeof(*oh) + sizeof(*ucfg); /* Check number of redirs */ if (sd->valsize < read + ucfg->redir_cnt*sizeof(struct nat44_cfg_redir)) return (EINVAL); nat44_config(chain, ucfg); return (0); } /* * Destroys given nat instances. * Data layout (v0)(current): * Request: [ ipfw_obj_header ] * * Returns 0 on success */ static int nat44_destroy(struct ip_fw_chain *chain, ip_fw3_opheader *op3, struct sockopt_data *sd) { ipfw_obj_header *oh; struct cfg_nat *ptr; ipfw_obj_ntlv *ntlv; /* Check minimum header size */ if (sd->valsize < sizeof(*oh)) return (EINVAL); oh = (ipfw_obj_header *)sd->kbuf; /* Basic length checks for TLVs */ if (oh->ntlv.head.length != sizeof(oh->ntlv)) return (EINVAL); ntlv = &oh->ntlv; /* Check if name is properly terminated */ if (strnlen(ntlv->name, sizeof(ntlv->name)) == sizeof(ntlv->name)) return (EINVAL); IPFW_UH_WLOCK(chain); ptr = lookup_nat_name(&chain->nat, ntlv->name); if (ptr == NULL) { IPFW_UH_WUNLOCK(chain); return (ESRCH); } IPFW_WLOCK(chain); LIST_REMOVE(ptr, _next); flush_nat_ptrs(chain, ptr->id); IPFW_WUNLOCK(chain); IPFW_UH_WUNLOCK(chain); free_nat_instance(ptr); return (0); } static void export_nat_cfg(struct cfg_nat *ptr, struct nat44_cfg_nat *ucfg) { snprintf(ucfg->name, sizeof(ucfg->name), "%d", ptr->id); ucfg->ip = ptr->ip; ucfg->redir_cnt = ptr->redir_cnt; ucfg->mode = ptr->mode; strlcpy(ucfg->if_name, ptr->if_name, sizeof(ucfg->if_name)); } /* * Gets config for given nat instance * Data layout (v0)(current): * Request: [ ipfw_obj_header nat44_cfg_nat .. ] * * Returns 0 on success */ static int nat44_get_cfg(struct ip_fw_chain *chain, ip_fw3_opheader *op3, struct sockopt_data *sd) { ipfw_obj_header *oh; struct nat44_cfg_nat *ucfg; struct cfg_nat *ptr; struct cfg_redir *r; struct cfg_spool *s; struct nat44_cfg_redir *ser_r; struct nat44_cfg_spool *ser_s; size_t sz; sz = sizeof(*oh) + sizeof(*ucfg); /* Check minimum header size */ if (sd->valsize < sz) return (EINVAL); oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz); /* Basic length checks for TLVs */ if (oh->ntlv.head.length != sizeof(oh->ntlv)) return (EINVAL); ucfg = (struct nat44_cfg_nat *)(oh + 1); /* Check if name is properly terminated */ if (strnlen(ucfg->name, sizeof(ucfg->name)) == sizeof(ucfg->name)) return (EINVAL); IPFW_UH_RLOCK(chain); ptr = lookup_nat_name(&chain->nat, ucfg->name); if (ptr == NULL) { IPFW_UH_RUNLOCK(chain); return (ESRCH); } export_nat_cfg(ptr, ucfg); /* Estimate memory amount */ sz = sizeof(ipfw_obj_header) + sizeof(struct nat44_cfg_nat); LIST_FOREACH(r, &ptr->redir_chain, _next) { sz += sizeof(struct nat44_cfg_redir); LIST_FOREACH(s, &r->spool_chain, _next) sz += sizeof(struct nat44_cfg_spool); } ucfg->size = sz; if (sd->valsize < sz) { /* * Submitted buffer size is not enough. * WE've already filled in @ucfg structure with * relevant info including size, so we * can return. Buffer will be flushed automatically. */ IPFW_UH_RUNLOCK(chain); return (ENOMEM); } /* Size OK, let's copy data */ LIST_FOREACH(r, &ptr->redir_chain, _next) { ser_r = (struct nat44_cfg_redir *)ipfw_get_sopt_space(sd, sizeof(*ser_r)); ser_r->mode = r->mode; ser_r->laddr = r->laddr; ser_r->paddr = r->paddr; ser_r->raddr = r->raddr; ser_r->lport = r->lport; ser_r->pport = r->pport; ser_r->rport = r->rport; ser_r->pport_cnt = r->pport_cnt; ser_r->rport_cnt = r->rport_cnt; ser_r->proto = r->proto; ser_r->spool_cnt = r->spool_cnt; LIST_FOREACH(s, &r->spool_chain, _next) { ser_s = (struct nat44_cfg_spool *)ipfw_get_sopt_space( sd, sizeof(*ser_s)); ser_s->addr = s->addr; ser_s->port = s->port; } } IPFW_UH_RUNLOCK(chain); return (0); } /* * Lists all nat44 instances currently available in kernel. * Data layout (v0)(current): * Request: [ ipfw_obj_lheader ] * Reply: [ ipfw_obj_lheader nat44_cfg_nat x N ] * * Returns 0 on success */ static int nat44_list_nat(struct ip_fw_chain *chain, ip_fw3_opheader *op3, struct sockopt_data *sd) { ipfw_obj_lheader *olh; struct nat44_cfg_nat *ucfg; struct cfg_nat *ptr; int nat_count; /* Check minimum header size */ if (sd->valsize < sizeof(ipfw_obj_lheader)) return (EINVAL); olh = (ipfw_obj_lheader *)ipfw_get_sopt_header(sd, sizeof(*olh)); IPFW_UH_RLOCK(chain); nat_count = 0; LIST_FOREACH(ptr, &chain->nat, _next) nat_count++; olh->count = nat_count; olh->objsize = sizeof(struct nat44_cfg_nat); olh->size = sizeof(*olh) + olh->count * olh->objsize; if (sd->valsize < olh->size) { IPFW_UH_RUNLOCK(chain); return (ENOMEM); } LIST_FOREACH(ptr, &chain->nat, _next) { ucfg = (struct nat44_cfg_nat *)ipfw_get_sopt_space(sd, sizeof(*ucfg)); export_nat_cfg(ptr, ucfg); } IPFW_UH_RUNLOCK(chain); return (0); } /* * Gets log for given nat instance * Data layout (v0)(current): * Request: [ ipfw_obj_header nat44_cfg_nat ] * Reply: [ ipfw_obj_header nat44_cfg_nat LOGBUFFER ] * * Returns 0 on success */ static int nat44_get_log(struct ip_fw_chain *chain, ip_fw3_opheader *op3, struct sockopt_data *sd) { ipfw_obj_header *oh; struct nat44_cfg_nat *ucfg; struct cfg_nat *ptr; void *pbuf; size_t sz; sz = sizeof(*oh) + sizeof(*ucfg); /* Check minimum header size */ if (sd->valsize < sz) return (EINVAL); oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz); /* Basic length checks for TLVs */ if (oh->ntlv.head.length != sizeof(oh->ntlv)) return (EINVAL); ucfg = (struct nat44_cfg_nat *)(oh + 1); /* Check if name is properly terminated */ if (strnlen(ucfg->name, sizeof(ucfg->name)) == sizeof(ucfg->name)) return (EINVAL); IPFW_UH_RLOCK(chain); ptr = lookup_nat_name(&chain->nat, ucfg->name); if (ptr == NULL) { IPFW_UH_RUNLOCK(chain); return (ESRCH); } if (ptr->lib->logDesc == NULL) { IPFW_UH_RUNLOCK(chain); return (ENOENT); } export_nat_cfg(ptr, ucfg); /* Estimate memory amount */ ucfg->size = sizeof(struct nat44_cfg_nat) + LIBALIAS_BUF_SIZE; if (sd->valsize < sz + sizeof(*oh)) { /* * Submitted buffer size is not enough. * WE've already filled in @ucfg structure with * relevant info including size, so we * can return. Buffer will be flushed automatically. */ IPFW_UH_RUNLOCK(chain); return (ENOMEM); } pbuf = (void *)ipfw_get_sopt_space(sd, LIBALIAS_BUF_SIZE); memcpy(pbuf, ptr->lib->logDesc, LIBALIAS_BUF_SIZE); IPFW_UH_RUNLOCK(chain); return (0); } static struct ipfw_sopt_handler scodes[] = { { IP_FW_NAT44_XCONFIG, 0, HDIR_SET, nat44_cfg }, { IP_FW_NAT44_DESTROY, 0, HDIR_SET, nat44_destroy }, { IP_FW_NAT44_XGETCONFIG, 0, HDIR_GET, nat44_get_cfg }, { IP_FW_NAT44_LIST_NAT, 0, HDIR_GET, nat44_list_nat }, { IP_FW_NAT44_XGETLOG, 0, HDIR_GET, nat44_get_log }, }; /* * Legacy configuration routines */ struct cfg_spool_legacy { LIST_ENTRY(cfg_spool_legacy) _next; struct in_addr addr; u_short port; }; struct cfg_redir_legacy { LIST_ENTRY(cfg_redir) _next; u_int16_t mode; struct in_addr laddr; struct in_addr paddr; struct in_addr raddr; u_short lport; u_short pport; u_short rport; u_short pport_cnt; u_short rport_cnt; int proto; struct alias_link **alink; u_int16_t spool_cnt; LIST_HEAD(, cfg_spool_legacy) spool_chain; }; struct cfg_nat_legacy { LIST_ENTRY(cfg_nat_legacy) _next; int id; struct in_addr ip; char if_name[IF_NAMESIZE]; int mode; struct libalias *lib; int redir_cnt; LIST_HEAD(, cfg_redir_legacy) redir_chain; }; static int ipfw_nat_cfg(struct sockopt *sopt) { struct cfg_nat_legacy *cfg; struct nat44_cfg_nat *ucfg; struct cfg_redir_legacy *rdir; struct nat44_cfg_redir *urdir; char *buf; size_t len, len2; int error, i; len = sopt->sopt_valsize; len2 = len + 128; /* * Allocate 2x buffer to store converted structures. * new redir_cfg has shrinked, so we're sure that * new buffer size is enough. */ buf = malloc(roundup2(len, 8) + len2, M_TEMP, M_WAITOK | M_ZERO); error = sooptcopyin(sopt, buf, len, sizeof(struct cfg_nat_legacy)); if (error != 0) goto out; cfg = (struct cfg_nat_legacy *)buf; if (cfg->id < 0) { error = EINVAL; goto out; } ucfg = (struct nat44_cfg_nat *)&buf[roundup2(len, 8)]; snprintf(ucfg->name, sizeof(ucfg->name), "%d", cfg->id); strlcpy(ucfg->if_name, cfg->if_name, sizeof(ucfg->if_name)); ucfg->ip = cfg->ip; ucfg->mode = cfg->mode; ucfg->redir_cnt = cfg->redir_cnt; if (len < sizeof(*cfg) + cfg->redir_cnt * sizeof(*rdir)) { error = EINVAL; goto out; } urdir = (struct nat44_cfg_redir *)(ucfg + 1); rdir = (struct cfg_redir_legacy *)(cfg + 1); for (i = 0; i < cfg->redir_cnt; i++) { urdir->mode = rdir->mode; urdir->laddr = rdir->laddr; urdir->paddr = rdir->paddr; urdir->raddr = rdir->raddr; urdir->lport = rdir->lport; urdir->pport = rdir->pport; urdir->rport = rdir->rport; urdir->pport_cnt = rdir->pport_cnt; urdir->rport_cnt = rdir->rport_cnt; urdir->proto = rdir->proto; urdir->spool_cnt = rdir->spool_cnt; urdir++; rdir++; } nat44_config(&V_layer3_chain, ucfg); out: free(buf, M_TEMP); return (error); } static int ipfw_nat_del(struct sockopt *sopt) { struct cfg_nat *ptr; struct ip_fw_chain *chain = &V_layer3_chain; int i; sooptcopyin(sopt, &i, sizeof i, sizeof i); /* XXX validate i */ IPFW_UH_WLOCK(chain); ptr = lookup_nat(&chain->nat, i); if (ptr == NULL) { IPFW_UH_WUNLOCK(chain); return (EINVAL); } IPFW_WLOCK(chain); LIST_REMOVE(ptr, _next); flush_nat_ptrs(chain, i); IPFW_WUNLOCK(chain); IPFW_UH_WUNLOCK(chain); free_nat_instance(ptr); return (0); } static int ipfw_nat_get_cfg(struct sockopt *sopt) { struct ip_fw_chain *chain = &V_layer3_chain; struct cfg_nat *n; struct cfg_nat_legacy *ucfg; struct cfg_redir *r; struct cfg_spool *s; struct cfg_redir_legacy *ser_r; struct cfg_spool_legacy *ser_s; char *data; int gencnt, nat_cnt, len, error; nat_cnt = 0; len = sizeof(nat_cnt); IPFW_UH_RLOCK(chain); retry: gencnt = chain->gencnt; /* Estimate memory amount */ LIST_FOREACH(n, &chain->nat, _next) { nat_cnt++; len += sizeof(struct cfg_nat_legacy); LIST_FOREACH(r, &n->redir_chain, _next) { len += sizeof(struct cfg_redir_legacy); LIST_FOREACH(s, &r->spool_chain, _next) len += sizeof(struct cfg_spool_legacy); } } IPFW_UH_RUNLOCK(chain); data = malloc(len, M_TEMP, M_WAITOK | M_ZERO); bcopy(&nat_cnt, data, sizeof(nat_cnt)); nat_cnt = 0; len = sizeof(nat_cnt); IPFW_UH_RLOCK(chain); if (gencnt != chain->gencnt) { free(data, M_TEMP); goto retry; } /* Serialize all the data. */ LIST_FOREACH(n, &chain->nat, _next) { ucfg = (struct cfg_nat_legacy *)&data[len]; ucfg->id = n->id; ucfg->ip = n->ip; ucfg->redir_cnt = n->redir_cnt; ucfg->mode = n->mode; strlcpy(ucfg->if_name, n->if_name, sizeof(ucfg->if_name)); len += sizeof(struct cfg_nat_legacy); LIST_FOREACH(r, &n->redir_chain, _next) { ser_r = (struct cfg_redir_legacy *)&data[len]; ser_r->mode = r->mode; ser_r->laddr = r->laddr; ser_r->paddr = r->paddr; ser_r->raddr = r->raddr; ser_r->lport = r->lport; ser_r->pport = r->pport; ser_r->rport = r->rport; ser_r->pport_cnt = r->pport_cnt; ser_r->rport_cnt = r->rport_cnt; ser_r->proto = r->proto; ser_r->spool_cnt = r->spool_cnt; len += sizeof(struct cfg_redir_legacy); LIST_FOREACH(s, &r->spool_chain, _next) { ser_s = (struct cfg_spool_legacy *)&data[len]; ser_s->addr = s->addr; ser_s->port = s->port; len += sizeof(struct cfg_spool_legacy); } } } IPFW_UH_RUNLOCK(chain); error = sooptcopyout(sopt, data, len); free(data, M_TEMP); return (error); } static int ipfw_nat_get_log(struct sockopt *sopt) { uint8_t *data; struct cfg_nat *ptr; int i, size; struct ip_fw_chain *chain; IPFW_RLOCK_TRACKER; chain = &V_layer3_chain; IPFW_RLOCK(chain); /* one pass to count, one to copy the data */ i = 0; LIST_FOREACH(ptr, &chain->nat, _next) { if (ptr->lib->logDesc == NULL) continue; i++; } size = i * (LIBALIAS_BUF_SIZE + sizeof(int)); data = malloc(size, M_IPFW, M_NOWAIT | M_ZERO); if (data == NULL) { IPFW_RUNLOCK(chain); return (ENOSPC); } i = 0; LIST_FOREACH(ptr, &chain->nat, _next) { if (ptr->lib->logDesc == NULL) continue; bcopy(&ptr->id, &data[i], sizeof(int)); i += sizeof(int); bcopy(ptr->lib->logDesc, &data[i], LIBALIAS_BUF_SIZE); i += LIBALIAS_BUF_SIZE; } IPFW_RUNLOCK(chain); sooptcopyout(sopt, data, size); free(data, M_IPFW); return(0); } static int vnet_ipfw_nat_init(const void *arg __unused) { V_ipfw_nat_ready = 1; return (0); } static int vnet_ipfw_nat_uninit(const void *arg __unused) { struct cfg_nat *ptr, *ptr_temp; struct ip_fw_chain *chain; chain = &V_layer3_chain; IPFW_WLOCK(chain); LIST_FOREACH_SAFE(ptr, &chain->nat, _next, ptr_temp) { LIST_REMOVE(ptr, _next); free_nat_instance(ptr); } flush_nat_ptrs(chain, -1 /* flush all */); V_ipfw_nat_ready = 0; IPFW_WUNLOCK(chain); return (0); } static void ipfw_nat_init(void) { /* init ipfw hooks */ ipfw_nat_ptr = ipfw_nat; lookup_nat_ptr = lookup_nat; ipfw_nat_cfg_ptr = ipfw_nat_cfg; ipfw_nat_del_ptr = ipfw_nat_del; ipfw_nat_get_cfg_ptr = ipfw_nat_get_cfg; ipfw_nat_get_log_ptr = ipfw_nat_get_log; IPFW_ADD_SOPT_HANDLER(1, scodes); ifaddr_event_tag = EVENTHANDLER_REGISTER(ifaddr_event, ifaddr_change, NULL, EVENTHANDLER_PRI_ANY); } static void ipfw_nat_destroy(void) { EVENTHANDLER_DEREGISTER(ifaddr_event, ifaddr_event_tag); /* deregister ipfw_nat */ IPFW_DEL_SOPT_HANDLER(1, scodes); ipfw_nat_ptr = NULL; lookup_nat_ptr = NULL; ipfw_nat_cfg_ptr = NULL; ipfw_nat_del_ptr = NULL; ipfw_nat_get_cfg_ptr = NULL; ipfw_nat_get_log_ptr = NULL; } static int ipfw_nat_modevent(module_t mod, int type, void *unused) { int err = 0; switch (type) { case MOD_LOAD: break; case MOD_UNLOAD: break; default: return EOPNOTSUPP; break; } return err; } static moduledata_t ipfw_nat_mod = { "ipfw_nat", ipfw_nat_modevent, 0 }; /* Define startup order. */ -#define IPFW_NAT_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN +#define IPFW_NAT_SI_SUB_FIREWALL SI_SUB_FW #define IPFW_NAT_MODEVENT_ORDER (SI_ORDER_ANY - 128) /* after ipfw */ #define IPFW_NAT_MODULE_ORDER (IPFW_NAT_MODEVENT_ORDER + 1) #define IPFW_NAT_VNET_ORDER (IPFW_NAT_MODEVENT_ORDER + 2) DECLARE_MODULE(ipfw_nat, ipfw_nat_mod, IPFW_NAT_SI_SUB_FIREWALL, SI_ORDER_ANY); MODULE_DEPEND(ipfw_nat, libalias, 1, 1, 1); MODULE_DEPEND(ipfw_nat, ipfw, 3, 3, 3); MODULE_VERSION(ipfw_nat, 1); SYSINIT(ipfw_nat_init, IPFW_NAT_SI_SUB_FIREWALL, IPFW_NAT_MODULE_ORDER, ipfw_nat_init, NULL); VNET_SYSINIT(vnet_ipfw_nat_init, IPFW_NAT_SI_SUB_FIREWALL, IPFW_NAT_VNET_ORDER, vnet_ipfw_nat_init, NULL); SYSUNINIT(ipfw_nat_destroy, IPFW_NAT_SI_SUB_FIREWALL, IPFW_NAT_MODULE_ORDER, ipfw_nat_destroy, NULL); VNET_SYSUNINIT(vnet_ipfw_nat_uninit, IPFW_NAT_SI_SUB_FIREWALL, IPFW_NAT_VNET_ORDER, vnet_ipfw_nat_uninit, NULL); /* end of file */ Index: projects/vnet/sys/netpfil/pf/if_pfsync.c =================================================================== --- projects/vnet/sys/netpfil/pf/if_pfsync.c (revision 294565) +++ projects/vnet/sys/netpfil/pf/if_pfsync.c (revision 294566) @@ -1,2421 +1,2422 @@ /*- * Copyright (c) 2002 Michael Shalayeff * Copyright (c) 2012 Gleb Smirnoff * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF * THE POSSIBILITY OF SUCH DAMAGE. */ /*- * Copyright (c) 2009 David Gwynne * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above * copyright notice and this permission notice appear in all copies. * * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ /* * $OpenBSD: if_pfsync.c,v 1.110 2009/02/24 05:39:19 dlg Exp $ * * Revisions picked from OpenBSD after revision 1.110 import: * 1.119 - don't m_copydata() beyond the len of mbuf in pfsync_input() * 1.118, 1.124, 1.148, 1.149, 1.151, 1.171 - fixes to bulk updates * 1.120, 1.175 - use monotonic time_uptime * 1.122 - reduce number of updates for non-TCP sessions * 1.125, 1.127 - rewrite merge or stale processing * 1.128 - cleanups * 1.146 - bzero() mbuf before sparsely filling it with data * 1.170 - SIOCSIFMTU checks * 1.126, 1.142 - deferred packets processing * 1.173 - correct expire time processing */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_pf.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define PFSYNC_MINPKT ( \ sizeof(struct ip) + \ sizeof(struct pfsync_header) + \ sizeof(struct pfsync_subheader) ) struct pfsync_pkt { struct ip *ip; struct in_addr src; u_int8_t flags; }; static int pfsync_upd_tcp(struct pf_state *, struct pfsync_state_peer *, struct pfsync_state_peer *); static int pfsync_in_clr(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_ins(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_iack(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_upd(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_upd_c(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_ureq(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_del(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_del_c(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_bus(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_tdb(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_eof(struct pfsync_pkt *, struct mbuf *, int, int); static int pfsync_in_error(struct pfsync_pkt *, struct mbuf *, int, int); static int (*pfsync_acts[])(struct pfsync_pkt *, struct mbuf *, int, int) = { pfsync_in_clr, /* PFSYNC_ACT_CLR */ pfsync_in_ins, /* PFSYNC_ACT_INS */ pfsync_in_iack, /* PFSYNC_ACT_INS_ACK */ pfsync_in_upd, /* PFSYNC_ACT_UPD */ pfsync_in_upd_c, /* PFSYNC_ACT_UPD_C */ pfsync_in_ureq, /* PFSYNC_ACT_UPD_REQ */ pfsync_in_del, /* PFSYNC_ACT_DEL */ pfsync_in_del_c, /* PFSYNC_ACT_DEL_C */ pfsync_in_error, /* PFSYNC_ACT_INS_F */ pfsync_in_error, /* PFSYNC_ACT_DEL_F */ pfsync_in_bus, /* PFSYNC_ACT_BUS */ pfsync_in_tdb, /* PFSYNC_ACT_TDB */ pfsync_in_eof /* PFSYNC_ACT_EOF */ }; struct pfsync_q { void (*write)(struct pf_state *, void *); size_t len; u_int8_t action; }; /* we have one of these for every PFSYNC_S_ */ static void pfsync_out_state(struct pf_state *, void *); static void pfsync_out_iack(struct pf_state *, void *); static void pfsync_out_upd_c(struct pf_state *, void *); static void pfsync_out_del(struct pf_state *, void *); static struct pfsync_q pfsync_qs[] = { { pfsync_out_state, sizeof(struct pfsync_state), PFSYNC_ACT_INS }, { pfsync_out_iack, sizeof(struct pfsync_ins_ack), PFSYNC_ACT_INS_ACK }, { pfsync_out_state, sizeof(struct pfsync_state), PFSYNC_ACT_UPD }, { pfsync_out_upd_c, sizeof(struct pfsync_upd_c), PFSYNC_ACT_UPD_C }, { pfsync_out_del, sizeof(struct pfsync_del_c), PFSYNC_ACT_DEL_C } }; static void pfsync_q_ins(struct pf_state *, int); static void pfsync_q_del(struct pf_state *); static void pfsync_update_state(struct pf_state *); struct pfsync_upd_req_item { TAILQ_ENTRY(pfsync_upd_req_item) ur_entry; struct pfsync_upd_req ur_msg; }; struct pfsync_deferral { struct pfsync_softc *pd_sc; TAILQ_ENTRY(pfsync_deferral) pd_entry; u_int pd_refs; struct callout pd_tmo; struct pf_state *pd_st; struct mbuf *pd_m; }; struct pfsync_softc { /* Configuration */ struct ifnet *sc_ifp; struct ifnet *sc_sync_if; struct ip_moptions sc_imo; struct in_addr sc_sync_peer; uint32_t sc_flags; #define PFSYNCF_OK 0x00000001 #define PFSYNCF_DEFER 0x00000002 #define PFSYNCF_PUSH 0x00000004 uint8_t sc_maxupdates; struct ip sc_template; struct callout sc_tmo; struct mtx sc_mtx; /* Queued data */ size_t sc_len; TAILQ_HEAD(, pf_state) sc_qs[PFSYNC_S_COUNT]; TAILQ_HEAD(, pfsync_upd_req_item) sc_upd_req_list; TAILQ_HEAD(, pfsync_deferral) sc_deferrals; u_int sc_deferred; void *sc_plus; size_t sc_pluslen; /* Bulk update info */ struct mtx sc_bulk_mtx; uint32_t sc_ureq_sent; int sc_bulk_tries; uint32_t sc_ureq_received; int sc_bulk_hashid; uint64_t sc_bulk_stateid; uint32_t sc_bulk_creatorid; struct callout sc_bulk_tmo; struct callout sc_bulkfail_tmo; }; #define PFSYNC_LOCK(sc) mtx_lock(&(sc)->sc_mtx) #define PFSYNC_UNLOCK(sc) mtx_unlock(&(sc)->sc_mtx) #define PFSYNC_LOCK_ASSERT(sc) mtx_assert(&(sc)->sc_mtx, MA_OWNED) #define PFSYNC_BLOCK(sc) mtx_lock(&(sc)->sc_bulk_mtx) #define PFSYNC_BUNLOCK(sc) mtx_unlock(&(sc)->sc_bulk_mtx) #define PFSYNC_BLOCK_ASSERT(sc) mtx_assert(&(sc)->sc_bulk_mtx, MA_OWNED) static const char pfsyncname[] = "pfsync"; static MALLOC_DEFINE(M_PFSYNC, pfsyncname, "pfsync(4) data"); static VNET_DEFINE(struct pfsync_softc *, pfsyncif) = NULL; #define V_pfsyncif VNET(pfsyncif) static VNET_DEFINE(void *, pfsync_swi_cookie) = NULL; #define V_pfsync_swi_cookie VNET(pfsync_swi_cookie) static VNET_DEFINE(struct pfsyncstats, pfsyncstats); #define V_pfsyncstats VNET(pfsyncstats) static VNET_DEFINE(int, pfsync_carp_adj) = CARP_MAXSKEW; #define V_pfsync_carp_adj VNET(pfsync_carp_adj) static void pfsync_timeout(void *); static void pfsync_push(struct pfsync_softc *); static void pfsyncintr(void *); static int pfsync_multicast_setup(struct pfsync_softc *, struct ifnet *, void *); static void pfsync_multicast_cleanup(struct pfsync_softc *); static void pfsync_pointers_init(void); static void pfsync_pointers_uninit(void); static int pfsync_init(void); static void pfsync_uninit(void); SYSCTL_NODE(_net, OID_AUTO, pfsync, CTLFLAG_RW, 0, "PFSYNC"); SYSCTL_STRUCT(_net_pfsync, OID_AUTO, stats, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(pfsyncstats), pfsyncstats, "PFSYNC statistics (struct pfsyncstats, net/if_pfsync.h)"); SYSCTL_INT(_net_pfsync, OID_AUTO, carp_demotion_factor, CTLFLAG_RW, &VNET_NAME(pfsync_carp_adj), 0, "pfsync's CARP demotion factor adjustment"); static int pfsync_clone_create(struct if_clone *, int, caddr_t); static void pfsync_clone_destroy(struct ifnet *); static int pfsync_alloc_scrub_memory(struct pfsync_state_peer *, struct pf_state_peer *); static int pfsyncoutput(struct ifnet *, struct mbuf *, const struct sockaddr *, struct route *); static int pfsyncioctl(struct ifnet *, u_long, caddr_t); static int pfsync_defer(struct pf_state *, struct mbuf *); static void pfsync_undefer(struct pfsync_deferral *, int); static void pfsync_undefer_state(struct pf_state *, int); static void pfsync_defer_tmo(void *); static void pfsync_request_update(u_int32_t, u_int64_t); static void pfsync_update_state_req(struct pf_state *); static void pfsync_drop(struct pfsync_softc *); static void pfsync_sendout(int); static void pfsync_send_plus(void *, size_t); static void pfsync_bulk_start(void); static void pfsync_bulk_status(u_int8_t); static void pfsync_bulk_update(void *); static void pfsync_bulk_fail(void *); #ifdef IPSEC static void pfsync_update_net_tdb(struct pfsync_tdb *); #endif #define PFSYNC_MAX_BULKTRIES 12 VNET_DEFINE(struct if_clone *, pfsync_cloner); #define V_pfsync_cloner VNET(pfsync_cloner) static int pfsync_clone_create(struct if_clone *ifc, int unit, caddr_t param) { struct pfsync_softc *sc; struct ifnet *ifp; int q; if (unit != 0) return (EINVAL); sc = malloc(sizeof(struct pfsync_softc), M_PFSYNC, M_WAITOK | M_ZERO); sc->sc_flags |= PFSYNCF_OK; for (q = 0; q < PFSYNC_S_COUNT; q++) TAILQ_INIT(&sc->sc_qs[q]); TAILQ_INIT(&sc->sc_upd_req_list); TAILQ_INIT(&sc->sc_deferrals); sc->sc_len = PFSYNC_MINPKT; sc->sc_maxupdates = 128; ifp = sc->sc_ifp = if_alloc(IFT_PFSYNC); if (ifp == NULL) { free(sc, M_PFSYNC); return (ENOSPC); } if_initname(ifp, pfsyncname, unit); ifp->if_softc = sc; ifp->if_ioctl = pfsyncioctl; ifp->if_output = pfsyncoutput; ifp->if_type = IFT_PFSYNC; ifp->if_snd.ifq_maxlen = ifqmaxlen; ifp->if_hdrlen = sizeof(struct pfsync_header); ifp->if_mtu = ETHERMTU; mtx_init(&sc->sc_mtx, pfsyncname, NULL, MTX_DEF); mtx_init(&sc->sc_bulk_mtx, "pfsync bulk", NULL, MTX_DEF); callout_init(&sc->sc_tmo, 1); callout_init_mtx(&sc->sc_bulk_tmo, &sc->sc_bulk_mtx, 0); callout_init_mtx(&sc->sc_bulkfail_tmo, &sc->sc_bulk_mtx, 0); if_attach(ifp); bpfattach(ifp, DLT_PFSYNC, PFSYNC_HDRLEN); V_pfsyncif = sc; return (0); } static void pfsync_clone_destroy(struct ifnet *ifp) { struct pfsync_softc *sc = ifp->if_softc; /* * At this stage, everything should have already been * cleared by pfsync_uninit(), and we have only to * drain callouts. */ while (sc->sc_deferred > 0) { struct pfsync_deferral *pd = TAILQ_FIRST(&sc->sc_deferrals); TAILQ_REMOVE(&sc->sc_deferrals, pd, pd_entry); sc->sc_deferred--; if (callout_stop(&pd->pd_tmo) > 0) { pf_release_state(pd->pd_st); m_freem(pd->pd_m); free(pd, M_PFSYNC); } else { pd->pd_refs++; callout_drain(&pd->pd_tmo); free(pd, M_PFSYNC); } } callout_drain(&sc->sc_tmo); callout_drain(&sc->sc_bulkfail_tmo); callout_drain(&sc->sc_bulk_tmo); if (!(sc->sc_flags & PFSYNCF_OK) && carp_demote_adj_p) (*carp_demote_adj_p)(-V_pfsync_carp_adj, "pfsync destroy"); bpfdetach(ifp); if_detach(ifp); pfsync_drop(sc); if_free(ifp); if (sc->sc_imo.imo_membership) pfsync_multicast_cleanup(sc); mtx_destroy(&sc->sc_mtx); mtx_destroy(&sc->sc_bulk_mtx); free(sc, M_PFSYNC); V_pfsyncif = NULL; } static int pfsync_alloc_scrub_memory(struct pfsync_state_peer *s, struct pf_state_peer *d) { if (s->scrub.scrub_flag && d->scrub == NULL) { d->scrub = uma_zalloc(V_pf_state_scrub_z, M_NOWAIT | M_ZERO); if (d->scrub == NULL) return (ENOMEM); } return (0); } static int pfsync_state_import(struct pfsync_state *sp, u_int8_t flags) { struct pfsync_softc *sc = V_pfsyncif; #ifndef __NO_STRICT_ALIGNMENT struct pfsync_state_key key[2]; #endif struct pfsync_state_key *kw, *ks; struct pf_state *st = NULL; struct pf_state_key *skw = NULL, *sks = NULL; struct pf_rule *r = NULL; struct pfi_kif *kif; int error; PF_RULES_RASSERT(); if (sp->creatorid == 0) { if (V_pf_status.debug >= PF_DEBUG_MISC) printf("%s: invalid creator id: %08x\n", __func__, ntohl(sp->creatorid)); return (EINVAL); } if ((kif = pfi_kif_find(sp->ifname)) == NULL) { if (V_pf_status.debug >= PF_DEBUG_MISC) printf("%s: unknown interface: %s\n", __func__, sp->ifname); if (flags & PFSYNC_SI_IOCTL) return (EINVAL); return (0); /* skip this state */ } /* * If the ruleset checksums match or the state is coming from the ioctl, * it's safe to associate the state with the rule of that number. */ if (sp->rule != htonl(-1) && sp->anchor == htonl(-1) && (flags & (PFSYNC_SI_IOCTL | PFSYNC_SI_CKSUM)) && ntohl(sp->rule) < pf_main_ruleset.rules[PF_RULESET_FILTER].active.rcount) r = pf_main_ruleset.rules[ PF_RULESET_FILTER].active.ptr_array[ntohl(sp->rule)]; else r = &V_pf_default_rule; if ((r->max_states && counter_u64_fetch(r->states_cur) >= r->max_states)) goto cleanup; /* * XXXGL: consider M_WAITOK in ioctl path after. */ if ((st = uma_zalloc(V_pf_state_z, M_NOWAIT | M_ZERO)) == NULL) goto cleanup; if ((skw = uma_zalloc(V_pf_state_key_z, M_NOWAIT)) == NULL) goto cleanup; #ifndef __NO_STRICT_ALIGNMENT bcopy(&sp->key, key, sizeof(struct pfsync_state_key) * 2); kw = &key[PF_SK_WIRE]; ks = &key[PF_SK_STACK]; #else kw = &sp->key[PF_SK_WIRE]; ks = &sp->key[PF_SK_STACK]; #endif if (PF_ANEQ(&kw->addr[0], &ks->addr[0], sp->af) || PF_ANEQ(&kw->addr[1], &ks->addr[1], sp->af) || kw->port[0] != ks->port[0] || kw->port[1] != ks->port[1]) { sks = uma_zalloc(V_pf_state_key_z, M_NOWAIT); if (sks == NULL) goto cleanup; } else sks = skw; /* allocate memory for scrub info */ if (pfsync_alloc_scrub_memory(&sp->src, &st->src) || pfsync_alloc_scrub_memory(&sp->dst, &st->dst)) goto cleanup; /* Copy to state key(s). */ skw->addr[0] = kw->addr[0]; skw->addr[1] = kw->addr[1]; skw->port[0] = kw->port[0]; skw->port[1] = kw->port[1]; skw->proto = sp->proto; skw->af = sp->af; if (sks != skw) { sks->addr[0] = ks->addr[0]; sks->addr[1] = ks->addr[1]; sks->port[0] = ks->port[0]; sks->port[1] = ks->port[1]; sks->proto = sp->proto; sks->af = sp->af; } /* copy to state */ bcopy(&sp->rt_addr, &st->rt_addr, sizeof(st->rt_addr)); st->creation = time_uptime - ntohl(sp->creation); st->expire = time_uptime; if (sp->expire) { uint32_t timeout; timeout = r->timeout[sp->timeout]; if (!timeout) timeout = V_pf_default_rule.timeout[sp->timeout]; /* sp->expire may have been adaptively scaled by export. */ st->expire -= timeout - ntohl(sp->expire); } st->direction = sp->direction; st->log = sp->log; st->timeout = sp->timeout; st->state_flags = sp->state_flags; st->id = sp->id; st->creatorid = sp->creatorid; pf_state_peer_ntoh(&sp->src, &st->src); pf_state_peer_ntoh(&sp->dst, &st->dst); st->rule.ptr = r; st->nat_rule.ptr = NULL; st->anchor.ptr = NULL; st->rt_kif = NULL; st->pfsync_time = time_uptime; st->sync_state = PFSYNC_S_NONE; if (!(flags & PFSYNC_SI_IOCTL)) st->state_flags |= PFSTATE_NOSYNC; if ((error = pf_state_insert(kif, skw, sks, st)) != 0) goto cleanup_state; /* XXX when we have nat_rule/anchors, use STATE_INC_COUNTERS */ counter_u64_add(r->states_cur, 1); counter_u64_add(r->states_tot, 1); if (!(flags & PFSYNC_SI_IOCTL)) { st->state_flags &= ~PFSTATE_NOSYNC; if (st->state_flags & PFSTATE_ACK) { pfsync_q_ins(st, PFSYNC_S_IACK); pfsync_push(sc); } } st->state_flags &= ~PFSTATE_ACK; PF_STATE_UNLOCK(st); return (0); cleanup: error = ENOMEM; if (skw == sks) sks = NULL; if (skw != NULL) uma_zfree(V_pf_state_key_z, skw); if (sks != NULL) uma_zfree(V_pf_state_key_z, sks); cleanup_state: /* pf_state_insert() frees the state keys. */ if (st) { if (st->dst.scrub) uma_zfree(V_pf_state_scrub_z, st->dst.scrub); if (st->src.scrub) uma_zfree(V_pf_state_scrub_z, st->src.scrub); uma_zfree(V_pf_state_z, st); } return (error); } static int pfsync_input(struct mbuf **mp, int *offp __unused, int proto __unused) { struct pfsync_softc *sc = V_pfsyncif; struct pfsync_pkt pkt; struct mbuf *m = *mp; struct ip *ip = mtod(m, struct ip *); struct pfsync_header *ph; struct pfsync_subheader subh; int offset, len; int rv; uint16_t count; *mp = NULL; V_pfsyncstats.pfsyncs_ipackets++; /* Verify that we have a sync interface configured. */ if (!sc || !sc->sc_sync_if || !V_pf_status.running || (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) goto done; /* verify that the packet came in on the right interface */ if (sc->sc_sync_if != m->m_pkthdr.rcvif) { V_pfsyncstats.pfsyncs_badif++; goto done; } if_inc_counter(sc->sc_ifp, IFCOUNTER_IPACKETS, 1); if_inc_counter(sc->sc_ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); /* verify that the IP TTL is 255. */ if (ip->ip_ttl != PFSYNC_DFLTTL) { V_pfsyncstats.pfsyncs_badttl++; goto done; } offset = ip->ip_hl << 2; if (m->m_pkthdr.len < offset + sizeof(*ph)) { V_pfsyncstats.pfsyncs_hdrops++; goto done; } if (offset + sizeof(*ph) > m->m_len) { if (m_pullup(m, offset + sizeof(*ph)) == NULL) { V_pfsyncstats.pfsyncs_hdrops++; return (IPPROTO_DONE); } ip = mtod(m, struct ip *); } ph = (struct pfsync_header *)((char *)ip + offset); /* verify the version */ if (ph->version != PFSYNC_VERSION) { V_pfsyncstats.pfsyncs_badver++; goto done; } len = ntohs(ph->len) + offset; if (m->m_pkthdr.len < len) { V_pfsyncstats.pfsyncs_badlen++; goto done; } /* Cheaper to grab this now than having to mess with mbufs later */ pkt.ip = ip; pkt.src = ip->ip_src; pkt.flags = 0; /* * Trusting pf_chksum during packet processing, as well as seeking * in interface name tree, require holding PF_RULES_RLOCK(). */ PF_RULES_RLOCK(); if (!bcmp(&ph->pfcksum, &V_pf_status.pf_chksum, PF_MD5_DIGEST_LENGTH)) pkt.flags |= PFSYNC_SI_CKSUM; offset += sizeof(*ph); while (offset <= len - sizeof(subh)) { m_copydata(m, offset, sizeof(subh), (caddr_t)&subh); offset += sizeof(subh); if (subh.action >= PFSYNC_ACT_MAX) { V_pfsyncstats.pfsyncs_badact++; PF_RULES_RUNLOCK(); goto done; } count = ntohs(subh.count); V_pfsyncstats.pfsyncs_iacts[subh.action] += count; rv = (*pfsync_acts[subh.action])(&pkt, m, offset, count); if (rv == -1) { PF_RULES_RUNLOCK(); return (IPPROTO_DONE); } offset += rv; } PF_RULES_RUNLOCK(); done: m_freem(m); return (IPPROTO_DONE); } static int pfsync_in_clr(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { struct pfsync_clr *clr; struct mbuf *mp; int len = sizeof(*clr) * count; int i, offp; u_int32_t creatorid; mp = m_pulldown(m, offset, len, &offp); if (mp == NULL) { V_pfsyncstats.pfsyncs_badlen++; return (-1); } clr = (struct pfsync_clr *)(mp->m_data + offp); for (i = 0; i < count; i++) { creatorid = clr[i].creatorid; if (clr[i].ifname[0] != '\0' && pfi_kif_find(clr[i].ifname) == NULL) continue; for (int i = 0; i <= pf_hashmask; i++) { struct pf_idhash *ih = &V_pf_idhash[i]; struct pf_state *s; relock: PF_HASHROW_LOCK(ih); LIST_FOREACH(s, &ih->states, entry) { if (s->creatorid == creatorid) { s->state_flags |= PFSTATE_NOSYNC; pf_unlink_state(s, PF_ENTER_LOCKED); goto relock; } } PF_HASHROW_UNLOCK(ih); } } return (len); } static int pfsync_in_ins(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { struct mbuf *mp; struct pfsync_state *sa, *sp; int len = sizeof(*sp) * count; int i, offp; mp = m_pulldown(m, offset, len, &offp); if (mp == NULL) { V_pfsyncstats.pfsyncs_badlen++; return (-1); } sa = (struct pfsync_state *)(mp->m_data + offp); for (i = 0; i < count; i++) { sp = &sa[i]; /* Check for invalid values. */ if (sp->timeout >= PFTM_MAX || sp->src.state > PF_TCPS_PROXY_DST || sp->dst.state > PF_TCPS_PROXY_DST || sp->direction > PF_OUT || (sp->af != AF_INET && sp->af != AF_INET6)) { if (V_pf_status.debug >= PF_DEBUG_MISC) printf("%s: invalid value\n", __func__); V_pfsyncstats.pfsyncs_badval++; continue; } if (pfsync_state_import(sp, pkt->flags) == ENOMEM) /* Drop out, but process the rest of the actions. */ break; } return (len); } static int pfsync_in_iack(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { struct pfsync_ins_ack *ia, *iaa; struct pf_state *st; struct mbuf *mp; int len = count * sizeof(*ia); int offp, i; mp = m_pulldown(m, offset, len, &offp); if (mp == NULL) { V_pfsyncstats.pfsyncs_badlen++; return (-1); } iaa = (struct pfsync_ins_ack *)(mp->m_data + offp); for (i = 0; i < count; i++) { ia = &iaa[i]; st = pf_find_state_byid(ia->id, ia->creatorid); if (st == NULL) continue; if (st->state_flags & PFSTATE_ACK) { PFSYNC_LOCK(V_pfsyncif); pfsync_undefer_state(st, 0); PFSYNC_UNLOCK(V_pfsyncif); } PF_STATE_UNLOCK(st); } /* * XXX this is not yet implemented, but we know the size of the * message so we can skip it. */ return (count * sizeof(struct pfsync_ins_ack)); } static int pfsync_upd_tcp(struct pf_state *st, struct pfsync_state_peer *src, struct pfsync_state_peer *dst) { int sync = 0; PF_STATE_LOCK_ASSERT(st); /* * The state should never go backwards except * for syn-proxy states. Neither should the * sequence window slide backwards. */ if ((st->src.state > src->state && (st->src.state < PF_TCPS_PROXY_SRC || src->state >= PF_TCPS_PROXY_SRC)) || (st->src.state == src->state && SEQ_GT(st->src.seqlo, ntohl(src->seqlo)))) sync++; else pf_state_peer_ntoh(src, &st->src); if ((st->dst.state > dst->state) || (st->dst.state >= TCPS_SYN_SENT && SEQ_GT(st->dst.seqlo, ntohl(dst->seqlo)))) sync++; else pf_state_peer_ntoh(dst, &st->dst); return (sync); } static int pfsync_in_upd(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { struct pfsync_softc *sc = V_pfsyncif; struct pfsync_state *sa, *sp; struct pf_state *st; int sync; struct mbuf *mp; int len = count * sizeof(*sp); int offp, i; mp = m_pulldown(m, offset, len, &offp); if (mp == NULL) { V_pfsyncstats.pfsyncs_badlen++; return (-1); } sa = (struct pfsync_state *)(mp->m_data + offp); for (i = 0; i < count; i++) { sp = &sa[i]; /* check for invalid values */ if (sp->timeout >= PFTM_MAX || sp->src.state > PF_TCPS_PROXY_DST || sp->dst.state > PF_TCPS_PROXY_DST) { if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pfsync_input: PFSYNC_ACT_UPD: " "invalid value\n"); } V_pfsyncstats.pfsyncs_badval++; continue; } st = pf_find_state_byid(sp->id, sp->creatorid); if (st == NULL) { /* insert the update */ if (pfsync_state_import(sp, 0)) V_pfsyncstats.pfsyncs_badstate++; continue; } if (st->state_flags & PFSTATE_ACK) { PFSYNC_LOCK(sc); pfsync_undefer_state(st, 1); PFSYNC_UNLOCK(sc); } if (st->key[PF_SK_WIRE]->proto == IPPROTO_TCP) sync = pfsync_upd_tcp(st, &sp->src, &sp->dst); else { sync = 0; /* * Non-TCP protocol state machine always go * forwards */ if (st->src.state > sp->src.state) sync++; else pf_state_peer_ntoh(&sp->src, &st->src); if (st->dst.state > sp->dst.state) sync++; else pf_state_peer_ntoh(&sp->dst, &st->dst); } if (sync < 2) { pfsync_alloc_scrub_memory(&sp->dst, &st->dst); pf_state_peer_ntoh(&sp->dst, &st->dst); st->expire = time_uptime; st->timeout = sp->timeout; } st->pfsync_time = time_uptime; if (sync) { V_pfsyncstats.pfsyncs_stale++; pfsync_update_state(st); PF_STATE_UNLOCK(st); PFSYNC_LOCK(sc); pfsync_push(sc); PFSYNC_UNLOCK(sc); continue; } PF_STATE_UNLOCK(st); } return (len); } static int pfsync_in_upd_c(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { struct pfsync_softc *sc = V_pfsyncif; struct pfsync_upd_c *ua, *up; struct pf_state *st; int len = count * sizeof(*up); int sync; struct mbuf *mp; int offp, i; mp = m_pulldown(m, offset, len, &offp); if (mp == NULL) { V_pfsyncstats.pfsyncs_badlen++; return (-1); } ua = (struct pfsync_upd_c *)(mp->m_data + offp); for (i = 0; i < count; i++) { up = &ua[i]; /* check for invalid values */ if (up->timeout >= PFTM_MAX || up->src.state > PF_TCPS_PROXY_DST || up->dst.state > PF_TCPS_PROXY_DST) { if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pfsync_input: " "PFSYNC_ACT_UPD_C: " "invalid value\n"); } V_pfsyncstats.pfsyncs_badval++; continue; } st = pf_find_state_byid(up->id, up->creatorid); if (st == NULL) { /* We don't have this state. Ask for it. */ PFSYNC_LOCK(sc); pfsync_request_update(up->creatorid, up->id); PFSYNC_UNLOCK(sc); continue; } if (st->state_flags & PFSTATE_ACK) { PFSYNC_LOCK(sc); pfsync_undefer_state(st, 1); PFSYNC_UNLOCK(sc); } if (st->key[PF_SK_WIRE]->proto == IPPROTO_TCP) sync = pfsync_upd_tcp(st, &up->src, &up->dst); else { sync = 0; /* * Non-TCP protocol state machine always go * forwards */ if (st->src.state > up->src.state) sync++; else pf_state_peer_ntoh(&up->src, &st->src); if (st->dst.state > up->dst.state) sync++; else pf_state_peer_ntoh(&up->dst, &st->dst); } if (sync < 2) { pfsync_alloc_scrub_memory(&up->dst, &st->dst); pf_state_peer_ntoh(&up->dst, &st->dst); st->expire = time_uptime; st->timeout = up->timeout; } st->pfsync_time = time_uptime; if (sync) { V_pfsyncstats.pfsyncs_stale++; pfsync_update_state(st); PF_STATE_UNLOCK(st); PFSYNC_LOCK(sc); pfsync_push(sc); PFSYNC_UNLOCK(sc); continue; } PF_STATE_UNLOCK(st); } return (len); } static int pfsync_in_ureq(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { struct pfsync_upd_req *ur, *ura; struct mbuf *mp; int len = count * sizeof(*ur); int i, offp; struct pf_state *st; mp = m_pulldown(m, offset, len, &offp); if (mp == NULL) { V_pfsyncstats.pfsyncs_badlen++; return (-1); } ura = (struct pfsync_upd_req *)(mp->m_data + offp); for (i = 0; i < count; i++) { ur = &ura[i]; if (ur->id == 0 && ur->creatorid == 0) pfsync_bulk_start(); else { st = pf_find_state_byid(ur->id, ur->creatorid); if (st == NULL) { V_pfsyncstats.pfsyncs_badstate++; continue; } if (st->state_flags & PFSTATE_NOSYNC) { PF_STATE_UNLOCK(st); continue; } pfsync_update_state_req(st); PF_STATE_UNLOCK(st); } } return (len); } static int pfsync_in_del(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { struct mbuf *mp; struct pfsync_state *sa, *sp; struct pf_state *st; int len = count * sizeof(*sp); int offp, i; mp = m_pulldown(m, offset, len, &offp); if (mp == NULL) { V_pfsyncstats.pfsyncs_badlen++; return (-1); } sa = (struct pfsync_state *)(mp->m_data + offp); for (i = 0; i < count; i++) { sp = &sa[i]; st = pf_find_state_byid(sp->id, sp->creatorid); if (st == NULL) { V_pfsyncstats.pfsyncs_badstate++; continue; } st->state_flags |= PFSTATE_NOSYNC; pf_unlink_state(st, PF_ENTER_LOCKED); } return (len); } static int pfsync_in_del_c(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { struct mbuf *mp; struct pfsync_del_c *sa, *sp; struct pf_state *st; int len = count * sizeof(*sp); int offp, i; mp = m_pulldown(m, offset, len, &offp); if (mp == NULL) { V_pfsyncstats.pfsyncs_badlen++; return (-1); } sa = (struct pfsync_del_c *)(mp->m_data + offp); for (i = 0; i < count; i++) { sp = &sa[i]; st = pf_find_state_byid(sp->id, sp->creatorid); if (st == NULL) { V_pfsyncstats.pfsyncs_badstate++; continue; } st->state_flags |= PFSTATE_NOSYNC; pf_unlink_state(st, PF_ENTER_LOCKED); } return (len); } static int pfsync_in_bus(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { struct pfsync_softc *sc = V_pfsyncif; struct pfsync_bus *bus; struct mbuf *mp; int len = count * sizeof(*bus); int offp; PFSYNC_BLOCK(sc); /* If we're not waiting for a bulk update, who cares. */ if (sc->sc_ureq_sent == 0) { PFSYNC_BUNLOCK(sc); return (len); } mp = m_pulldown(m, offset, len, &offp); if (mp == NULL) { PFSYNC_BUNLOCK(sc); V_pfsyncstats.pfsyncs_badlen++; return (-1); } bus = (struct pfsync_bus *)(mp->m_data + offp); switch (bus->status) { case PFSYNC_BUS_START: callout_reset(&sc->sc_bulkfail_tmo, 4 * hz + V_pf_limits[PF_LIMIT_STATES].limit / ((sc->sc_ifp->if_mtu - PFSYNC_MINPKT) / sizeof(struct pfsync_state)), pfsync_bulk_fail, sc); if (V_pf_status.debug >= PF_DEBUG_MISC) printf("pfsync: received bulk update start\n"); break; case PFSYNC_BUS_END: if (time_uptime - ntohl(bus->endtime) >= sc->sc_ureq_sent) { /* that's it, we're happy */ sc->sc_ureq_sent = 0; sc->sc_bulk_tries = 0; callout_stop(&sc->sc_bulkfail_tmo); if (!(sc->sc_flags & PFSYNCF_OK) && carp_demote_adj_p) (*carp_demote_adj_p)(-V_pfsync_carp_adj, "pfsync bulk done"); sc->sc_flags |= PFSYNCF_OK; if (V_pf_status.debug >= PF_DEBUG_MISC) printf("pfsync: received valid " "bulk update end\n"); } else { if (V_pf_status.debug >= PF_DEBUG_MISC) printf("pfsync: received invalid " "bulk update end: bad timestamp\n"); } break; } PFSYNC_BUNLOCK(sc); return (len); } static int pfsync_in_tdb(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { int len = count * sizeof(struct pfsync_tdb); #if defined(IPSEC) struct pfsync_tdb *tp; struct mbuf *mp; int offp; int i; int s; mp = m_pulldown(m, offset, len, &offp); if (mp == NULL) { V_pfsyncstats.pfsyncs_badlen++; return (-1); } tp = (struct pfsync_tdb *)(mp->m_data + offp); for (i = 0; i < count; i++) pfsync_update_net_tdb(&tp[i]); #endif return (len); } #if defined(IPSEC) /* Update an in-kernel tdb. Silently fail if no tdb is found. */ static void pfsync_update_net_tdb(struct pfsync_tdb *pt) { struct tdb *tdb; int s; /* check for invalid values */ if (ntohl(pt->spi) <= SPI_RESERVED_MAX || (pt->dst.sa.sa_family != AF_INET && pt->dst.sa.sa_family != AF_INET6)) goto bad; tdb = gettdb(pt->spi, &pt->dst, pt->sproto); if (tdb) { pt->rpl = ntohl(pt->rpl); pt->cur_bytes = (unsigned long long)be64toh(pt->cur_bytes); /* Neither replay nor byte counter should ever decrease. */ if (pt->rpl < tdb->tdb_rpl || pt->cur_bytes < tdb->tdb_cur_bytes) { goto bad; } tdb->tdb_rpl = pt->rpl; tdb->tdb_cur_bytes = pt->cur_bytes; } return; bad: if (V_pf_status.debug >= PF_DEBUG_MISC) printf("pfsync_insert: PFSYNC_ACT_TDB_UPD: " "invalid value\n"); V_pfsyncstats.pfsyncs_badstate++; return; } #endif static int pfsync_in_eof(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { /* check if we are at the right place in the packet */ if (offset != m->m_pkthdr.len) V_pfsyncstats.pfsyncs_badlen++; /* we're done. free and let the caller return */ m_freem(m); return (-1); } static int pfsync_in_error(struct pfsync_pkt *pkt, struct mbuf *m, int offset, int count) { V_pfsyncstats.pfsyncs_badact++; m_freem(m); return (-1); } static int pfsyncoutput(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, struct route *rt) { m_freem(m); return (0); } /* ARGSUSED */ static int pfsyncioctl(struct ifnet *ifp, u_long cmd, caddr_t data) { struct pfsync_softc *sc = ifp->if_softc; struct ifreq *ifr = (struct ifreq *)data; struct pfsyncreq pfsyncr; int error; switch (cmd) { case SIOCSIFFLAGS: PFSYNC_LOCK(sc); if (ifp->if_flags & IFF_UP) { ifp->if_drv_flags |= IFF_DRV_RUNNING; PFSYNC_UNLOCK(sc); pfsync_pointers_init(); } else { ifp->if_drv_flags &= ~IFF_DRV_RUNNING; PFSYNC_UNLOCK(sc); pfsync_pointers_uninit(); } break; case SIOCSIFMTU: if (!sc->sc_sync_if || ifr->ifr_mtu <= PFSYNC_MINPKT || ifr->ifr_mtu > sc->sc_sync_if->if_mtu) return (EINVAL); if (ifr->ifr_mtu < ifp->if_mtu) { PFSYNC_LOCK(sc); if (sc->sc_len > PFSYNC_MINPKT) pfsync_sendout(1); PFSYNC_UNLOCK(sc); } ifp->if_mtu = ifr->ifr_mtu; break; case SIOCGETPFSYNC: bzero(&pfsyncr, sizeof(pfsyncr)); PFSYNC_LOCK(sc); if (sc->sc_sync_if) { strlcpy(pfsyncr.pfsyncr_syncdev, sc->sc_sync_if->if_xname, IFNAMSIZ); } pfsyncr.pfsyncr_syncpeer = sc->sc_sync_peer; pfsyncr.pfsyncr_maxupdates = sc->sc_maxupdates; pfsyncr.pfsyncr_defer = (PFSYNCF_DEFER == (sc->sc_flags & PFSYNCF_DEFER)); PFSYNC_UNLOCK(sc); return (copyout(&pfsyncr, ifr->ifr_data, sizeof(pfsyncr))); case SIOCSETPFSYNC: { struct ip_moptions *imo = &sc->sc_imo; struct ifnet *sifp; struct ip *ip; void *mship = NULL; if ((error = priv_check(curthread, PRIV_NETINET_PF)) != 0) return (error); if ((error = copyin(ifr->ifr_data, &pfsyncr, sizeof(pfsyncr)))) return (error); if (pfsyncr.pfsyncr_maxupdates > 255) return (EINVAL); if (pfsyncr.pfsyncr_syncdev[0] == 0) sifp = NULL; else if ((sifp = ifunit_ref(pfsyncr.pfsyncr_syncdev)) == NULL) return (EINVAL); if (sifp != NULL && ( pfsyncr.pfsyncr_syncpeer.s_addr == 0 || pfsyncr.pfsyncr_syncpeer.s_addr == htonl(INADDR_PFSYNC_GROUP))) mship = malloc((sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_PFSYNC, M_WAITOK | M_ZERO); PFSYNC_LOCK(sc); if (pfsyncr.pfsyncr_syncpeer.s_addr == 0) sc->sc_sync_peer.s_addr = htonl(INADDR_PFSYNC_GROUP); else sc->sc_sync_peer.s_addr = pfsyncr.pfsyncr_syncpeer.s_addr; sc->sc_maxupdates = pfsyncr.pfsyncr_maxupdates; if (pfsyncr.pfsyncr_defer) { sc->sc_flags |= PFSYNCF_DEFER; pfsync_defer_ptr = pfsync_defer; } else { sc->sc_flags &= ~PFSYNCF_DEFER; pfsync_defer_ptr = NULL; } if (sifp == NULL) { if (sc->sc_sync_if) if_rele(sc->sc_sync_if); sc->sc_sync_if = NULL; if (imo->imo_membership) pfsync_multicast_cleanup(sc); PFSYNC_UNLOCK(sc); break; } if (sc->sc_len > PFSYNC_MINPKT && (sifp->if_mtu < sc->sc_ifp->if_mtu || (sc->sc_sync_if != NULL && sifp->if_mtu < sc->sc_sync_if->if_mtu) || sifp->if_mtu < MCLBYTES - sizeof(struct ip))) pfsync_sendout(1); if (imo->imo_membership) pfsync_multicast_cleanup(sc); if (sc->sc_sync_peer.s_addr == htonl(INADDR_PFSYNC_GROUP)) { error = pfsync_multicast_setup(sc, sifp, mship); if (error) { if_rele(sifp); free(mship, M_PFSYNC); return (error); } } if (sc->sc_sync_if) if_rele(sc->sc_sync_if); sc->sc_sync_if = sifp; ip = &sc->sc_template; bzero(ip, sizeof(*ip)); ip->ip_v = IPVERSION; ip->ip_hl = sizeof(sc->sc_template) >> 2; ip->ip_tos = IPTOS_LOWDELAY; /* len and id are set later. */ ip->ip_off = htons(IP_DF); ip->ip_ttl = PFSYNC_DFLTTL; ip->ip_p = IPPROTO_PFSYNC; ip->ip_src.s_addr = INADDR_ANY; ip->ip_dst.s_addr = sc->sc_sync_peer.s_addr; /* Request a full state table update. */ if ((sc->sc_flags & PFSYNCF_OK) && carp_demote_adj_p) (*carp_demote_adj_p)(V_pfsync_carp_adj, "pfsync bulk start"); sc->sc_flags &= ~PFSYNCF_OK; if (V_pf_status.debug >= PF_DEBUG_MISC) printf("pfsync: requesting bulk update\n"); pfsync_request_update(0, 0); PFSYNC_UNLOCK(sc); PFSYNC_BLOCK(sc); sc->sc_ureq_sent = time_uptime; callout_reset(&sc->sc_bulkfail_tmo, 5 * hz, pfsync_bulk_fail, sc); PFSYNC_BUNLOCK(sc); break; } default: return (ENOTTY); } return (0); } static void pfsync_out_state(struct pf_state *st, void *buf) { struct pfsync_state *sp = buf; pfsync_state_export(sp, st); } static void pfsync_out_iack(struct pf_state *st, void *buf) { struct pfsync_ins_ack *iack = buf; iack->id = st->id; iack->creatorid = st->creatorid; } static void pfsync_out_upd_c(struct pf_state *st, void *buf) { struct pfsync_upd_c *up = buf; bzero(up, sizeof(*up)); up->id = st->id; pf_state_peer_hton(&st->src, &up->src); pf_state_peer_hton(&st->dst, &up->dst); up->creatorid = st->creatorid; up->timeout = st->timeout; } static void pfsync_out_del(struct pf_state *st, void *buf) { struct pfsync_del_c *dp = buf; dp->id = st->id; dp->creatorid = st->creatorid; st->state_flags |= PFSTATE_NOSYNC; } static void pfsync_drop(struct pfsync_softc *sc) { struct pf_state *st, *next; struct pfsync_upd_req_item *ur; int q; for (q = 0; q < PFSYNC_S_COUNT; q++) { if (TAILQ_EMPTY(&sc->sc_qs[q])) continue; TAILQ_FOREACH_SAFE(st, &sc->sc_qs[q], sync_list, next) { KASSERT(st->sync_state == q, ("%s: st->sync_state == q", __func__)); st->sync_state = PFSYNC_S_NONE; pf_release_state(st); } TAILQ_INIT(&sc->sc_qs[q]); } while ((ur = TAILQ_FIRST(&sc->sc_upd_req_list)) != NULL) { TAILQ_REMOVE(&sc->sc_upd_req_list, ur, ur_entry); free(ur, M_PFSYNC); } sc->sc_plus = NULL; sc->sc_len = PFSYNC_MINPKT; } static void pfsync_sendout(int schedswi) { struct pfsync_softc *sc = V_pfsyncif; struct ifnet *ifp = sc->sc_ifp; struct mbuf *m; struct ip *ip; struct pfsync_header *ph; struct pfsync_subheader *subh; struct pf_state *st; struct pfsync_upd_req_item *ur; int offset; int q, count = 0; KASSERT(sc != NULL, ("%s: null sc", __func__)); KASSERT(sc->sc_len > PFSYNC_MINPKT, ("%s: sc_len %zu", __func__, sc->sc_len)); PFSYNC_LOCK_ASSERT(sc); if (ifp->if_bpf == NULL && sc->sc_sync_if == NULL) { pfsync_drop(sc); return; } m = m_get2(max_linkhdr + sc->sc_len, M_NOWAIT, MT_DATA, M_PKTHDR); if (m == NULL) { if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1); V_pfsyncstats.pfsyncs_onomem++; return; } m->m_data += max_linkhdr; m->m_len = m->m_pkthdr.len = sc->sc_len; /* build the ip header */ ip = (struct ip *)m->m_data; bcopy(&sc->sc_template, ip, sizeof(*ip)); offset = sizeof(*ip); ip->ip_len = htons(m->m_pkthdr.len); ip_fillid(ip); /* build the pfsync header */ ph = (struct pfsync_header *)(m->m_data + offset); bzero(ph, sizeof(*ph)); offset += sizeof(*ph); ph->version = PFSYNC_VERSION; ph->len = htons(sc->sc_len - sizeof(*ip)); bcopy(V_pf_status.pf_chksum, ph->pfcksum, PF_MD5_DIGEST_LENGTH); /* walk the queues */ for (q = 0; q < PFSYNC_S_COUNT; q++) { if (TAILQ_EMPTY(&sc->sc_qs[q])) continue; subh = (struct pfsync_subheader *)(m->m_data + offset); offset += sizeof(*subh); count = 0; TAILQ_FOREACH(st, &sc->sc_qs[q], sync_list) { KASSERT(st->sync_state == q, ("%s: st->sync_state == q", __func__)); /* * XXXGL: some of write methods do unlocked reads * of state data :( */ pfsync_qs[q].write(st, m->m_data + offset); offset += pfsync_qs[q].len; st->sync_state = PFSYNC_S_NONE; pf_release_state(st); count++; } TAILQ_INIT(&sc->sc_qs[q]); bzero(subh, sizeof(*subh)); subh->action = pfsync_qs[q].action; subh->count = htons(count); V_pfsyncstats.pfsyncs_oacts[pfsync_qs[q].action] += count; } if (!TAILQ_EMPTY(&sc->sc_upd_req_list)) { subh = (struct pfsync_subheader *)(m->m_data + offset); offset += sizeof(*subh); count = 0; while ((ur = TAILQ_FIRST(&sc->sc_upd_req_list)) != NULL) { TAILQ_REMOVE(&sc->sc_upd_req_list, ur, ur_entry); bcopy(&ur->ur_msg, m->m_data + offset, sizeof(ur->ur_msg)); offset += sizeof(ur->ur_msg); free(ur, M_PFSYNC); count++; } bzero(subh, sizeof(*subh)); subh->action = PFSYNC_ACT_UPD_REQ; subh->count = htons(count); V_pfsyncstats.pfsyncs_oacts[PFSYNC_ACT_UPD_REQ] += count; } /* has someone built a custom region for us to add? */ if (sc->sc_plus != NULL) { bcopy(sc->sc_plus, m->m_data + offset, sc->sc_pluslen); offset += sc->sc_pluslen; sc->sc_plus = NULL; } subh = (struct pfsync_subheader *)(m->m_data + offset); offset += sizeof(*subh); bzero(subh, sizeof(*subh)); subh->action = PFSYNC_ACT_EOF; subh->count = htons(1); V_pfsyncstats.pfsyncs_oacts[PFSYNC_ACT_EOF]++; /* we're done, let's put it on the wire */ if (ifp->if_bpf) { m->m_data += sizeof(*ip); m->m_len = m->m_pkthdr.len = sc->sc_len - sizeof(*ip); BPF_MTAP(ifp, m); m->m_data -= sizeof(*ip); m->m_len = m->m_pkthdr.len = sc->sc_len; } if (sc->sc_sync_if == NULL) { sc->sc_len = PFSYNC_MINPKT; m_freem(m); return; } if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1); if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len); sc->sc_len = PFSYNC_MINPKT; if (!_IF_QFULL(&sc->sc_ifp->if_snd)) _IF_ENQUEUE(&sc->sc_ifp->if_snd, m); else { m_freem(m); if_inc_counter(sc->sc_ifp, IFCOUNTER_OQDROPS, 1); } if (schedswi) swi_sched(V_pfsync_swi_cookie, 0); } static void pfsync_insert_state(struct pf_state *st) { struct pfsync_softc *sc = V_pfsyncif; if (st->state_flags & PFSTATE_NOSYNC) return; if ((st->rule.ptr->rule_flag & PFRULE_NOSYNC) || st->key[PF_SK_WIRE]->proto == IPPROTO_PFSYNC) { st->state_flags |= PFSTATE_NOSYNC; return; } KASSERT(st->sync_state == PFSYNC_S_NONE, ("%s: st->sync_state %u", __func__, st->sync_state)); PFSYNC_LOCK(sc); if (sc->sc_len == PFSYNC_MINPKT) callout_reset(&sc->sc_tmo, 1 * hz, pfsync_timeout, V_pfsyncif); pfsync_q_ins(st, PFSYNC_S_INS); PFSYNC_UNLOCK(sc); st->sync_updates = 0; } static int pfsync_defer(struct pf_state *st, struct mbuf *m) { struct pfsync_softc *sc = V_pfsyncif; struct pfsync_deferral *pd; if (m->m_flags & (M_BCAST|M_MCAST)) return (0); PFSYNC_LOCK(sc); if (sc == NULL || !(sc->sc_ifp->if_flags & IFF_DRV_RUNNING) || !(sc->sc_flags & PFSYNCF_DEFER)) { PFSYNC_UNLOCK(sc); return (0); } if (sc->sc_deferred >= 128) pfsync_undefer(TAILQ_FIRST(&sc->sc_deferrals), 0); pd = malloc(sizeof(*pd), M_PFSYNC, M_NOWAIT); if (pd == NULL) return (0); sc->sc_deferred++; m->m_flags |= M_SKIP_FIREWALL; st->state_flags |= PFSTATE_ACK; pd->pd_sc = sc; pd->pd_refs = 0; pd->pd_st = st; pf_ref_state(st); pd->pd_m = m; TAILQ_INSERT_TAIL(&sc->sc_deferrals, pd, pd_entry); callout_init_mtx(&pd->pd_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED); callout_reset(&pd->pd_tmo, 10, pfsync_defer_tmo, pd); pfsync_push(sc); return (1); } static void pfsync_undefer(struct pfsync_deferral *pd, int drop) { struct pfsync_softc *sc = pd->pd_sc; struct mbuf *m = pd->pd_m; struct pf_state *st = pd->pd_st; PFSYNC_LOCK_ASSERT(sc); TAILQ_REMOVE(&sc->sc_deferrals, pd, pd_entry); sc->sc_deferred--; pd->pd_st->state_flags &= ~PFSTATE_ACK; /* XXX: locking! */ free(pd, M_PFSYNC); pf_release_state(st); if (drop) m_freem(m); else { _IF_ENQUEUE(&sc->sc_ifp->if_snd, m); pfsync_push(sc); } } static void pfsync_defer_tmo(void *arg) { struct pfsync_deferral *pd = arg; struct pfsync_softc *sc = pd->pd_sc; struct mbuf *m = pd->pd_m; struct pf_state *st = pd->pd_st; PFSYNC_LOCK_ASSERT(sc); CURVNET_SET(m->m_pkthdr.rcvif->if_vnet); TAILQ_REMOVE(&sc->sc_deferrals, pd, pd_entry); sc->sc_deferred--; pd->pd_st->state_flags &= ~PFSTATE_ACK; /* XXX: locking! */ if (pd->pd_refs == 0) free(pd, M_PFSYNC); PFSYNC_UNLOCK(sc); ip_output(m, NULL, NULL, 0, NULL, NULL); pf_release_state(st); CURVNET_RESTORE(); } static void pfsync_undefer_state(struct pf_state *st, int drop) { struct pfsync_softc *sc = V_pfsyncif; struct pfsync_deferral *pd; PFSYNC_LOCK_ASSERT(sc); TAILQ_FOREACH(pd, &sc->sc_deferrals, pd_entry) { if (pd->pd_st == st) { if (callout_stop(&pd->pd_tmo) > 0) pfsync_undefer(pd, drop); return; } } panic("%s: unable to find deferred state", __func__); } static void pfsync_update_state(struct pf_state *st) { struct pfsync_softc *sc = V_pfsyncif; int sync = 0; PF_STATE_LOCK_ASSERT(st); PFSYNC_LOCK(sc); if (st->state_flags & PFSTATE_ACK) pfsync_undefer_state(st, 0); if (st->state_flags & PFSTATE_NOSYNC) { if (st->sync_state != PFSYNC_S_NONE) pfsync_q_del(st); PFSYNC_UNLOCK(sc); return; } if (sc->sc_len == PFSYNC_MINPKT) callout_reset(&sc->sc_tmo, 1 * hz, pfsync_timeout, V_pfsyncif); switch (st->sync_state) { case PFSYNC_S_UPD_C: case PFSYNC_S_UPD: case PFSYNC_S_INS: /* we're already handling it */ if (st->key[PF_SK_WIRE]->proto == IPPROTO_TCP) { st->sync_updates++; if (st->sync_updates >= sc->sc_maxupdates) sync = 1; } break; case PFSYNC_S_IACK: pfsync_q_del(st); case PFSYNC_S_NONE: pfsync_q_ins(st, PFSYNC_S_UPD_C); st->sync_updates = 0; break; default: panic("%s: unexpected sync state %d", __func__, st->sync_state); } if (sync || (time_uptime - st->pfsync_time) < 2) pfsync_push(sc); PFSYNC_UNLOCK(sc); } static void pfsync_request_update(u_int32_t creatorid, u_int64_t id) { struct pfsync_softc *sc = V_pfsyncif; struct pfsync_upd_req_item *item; size_t nlen = sizeof(struct pfsync_upd_req); PFSYNC_LOCK_ASSERT(sc); /* * This code does a bit to prevent multiple update requests for the * same state being generated. It searches current subheader queue, * but it doesn't lookup into queue of already packed datagrams. */ TAILQ_FOREACH(item, &sc->sc_upd_req_list, ur_entry) if (item->ur_msg.id == id && item->ur_msg.creatorid == creatorid) return; item = malloc(sizeof(*item), M_PFSYNC, M_NOWAIT); if (item == NULL) return; /* XXX stats */ item->ur_msg.id = id; item->ur_msg.creatorid = creatorid; if (TAILQ_EMPTY(&sc->sc_upd_req_list)) nlen += sizeof(struct pfsync_subheader); if (sc->sc_len + nlen > sc->sc_ifp->if_mtu) { pfsync_sendout(1); nlen = sizeof(struct pfsync_subheader) + sizeof(struct pfsync_upd_req); } TAILQ_INSERT_TAIL(&sc->sc_upd_req_list, item, ur_entry); sc->sc_len += nlen; } static void pfsync_update_state_req(struct pf_state *st) { struct pfsync_softc *sc = V_pfsyncif; PF_STATE_LOCK_ASSERT(st); PFSYNC_LOCK(sc); if (st->state_flags & PFSTATE_NOSYNC) { if (st->sync_state != PFSYNC_S_NONE) pfsync_q_del(st); PFSYNC_UNLOCK(sc); return; } switch (st->sync_state) { case PFSYNC_S_UPD_C: case PFSYNC_S_IACK: pfsync_q_del(st); case PFSYNC_S_NONE: pfsync_q_ins(st, PFSYNC_S_UPD); pfsync_push(sc); break; case PFSYNC_S_INS: case PFSYNC_S_UPD: case PFSYNC_S_DEL: /* we're already handling it */ break; default: panic("%s: unexpected sync state %d", __func__, st->sync_state); } PFSYNC_UNLOCK(sc); } static void pfsync_delete_state(struct pf_state *st) { struct pfsync_softc *sc = V_pfsyncif; PFSYNC_LOCK(sc); if (st->state_flags & PFSTATE_ACK) pfsync_undefer_state(st, 1); if (st->state_flags & PFSTATE_NOSYNC) { if (st->sync_state != PFSYNC_S_NONE) pfsync_q_del(st); PFSYNC_UNLOCK(sc); return; } if (sc->sc_len == PFSYNC_MINPKT) callout_reset(&sc->sc_tmo, 1 * hz, pfsync_timeout, V_pfsyncif); switch (st->sync_state) { case PFSYNC_S_INS: /* We never got to tell the world so just forget about it. */ pfsync_q_del(st); break; case PFSYNC_S_UPD_C: case PFSYNC_S_UPD: case PFSYNC_S_IACK: pfsync_q_del(st); /* FALLTHROUGH to putting it on the del list */ case PFSYNC_S_NONE: pfsync_q_ins(st, PFSYNC_S_DEL); break; default: panic("%s: unexpected sync state %d", __func__, st->sync_state); } PFSYNC_UNLOCK(sc); } static void pfsync_clear_states(u_int32_t creatorid, const char *ifname) { struct pfsync_softc *sc = V_pfsyncif; struct { struct pfsync_subheader subh; struct pfsync_clr clr; } __packed r; bzero(&r, sizeof(r)); r.subh.action = PFSYNC_ACT_CLR; r.subh.count = htons(1); V_pfsyncstats.pfsyncs_oacts[PFSYNC_ACT_CLR]++; strlcpy(r.clr.ifname, ifname, sizeof(r.clr.ifname)); r.clr.creatorid = creatorid; PFSYNC_LOCK(sc); pfsync_send_plus(&r, sizeof(r)); PFSYNC_UNLOCK(sc); } static void pfsync_q_ins(struct pf_state *st, int q) { struct pfsync_softc *sc = V_pfsyncif; size_t nlen = pfsync_qs[q].len; PFSYNC_LOCK_ASSERT(sc); KASSERT(st->sync_state == PFSYNC_S_NONE, ("%s: st->sync_state %u", __func__, st->sync_state)); KASSERT(sc->sc_len >= PFSYNC_MINPKT, ("pfsync pkt len is too low %zu", sc->sc_len)); if (TAILQ_EMPTY(&sc->sc_qs[q])) nlen += sizeof(struct pfsync_subheader); if (sc->sc_len + nlen > sc->sc_ifp->if_mtu) { pfsync_sendout(1); nlen = sizeof(struct pfsync_subheader) + pfsync_qs[q].len; } sc->sc_len += nlen; TAILQ_INSERT_TAIL(&sc->sc_qs[q], st, sync_list); st->sync_state = q; pf_ref_state(st); } static void pfsync_q_del(struct pf_state *st) { struct pfsync_softc *sc = V_pfsyncif; int q = st->sync_state; PFSYNC_LOCK_ASSERT(sc); KASSERT(st->sync_state != PFSYNC_S_NONE, ("%s: st->sync_state != PFSYNC_S_NONE", __func__)); sc->sc_len -= pfsync_qs[q].len; TAILQ_REMOVE(&sc->sc_qs[q], st, sync_list); st->sync_state = PFSYNC_S_NONE; pf_release_state(st); if (TAILQ_EMPTY(&sc->sc_qs[q])) sc->sc_len -= sizeof(struct pfsync_subheader); } static void pfsync_bulk_start(void) { struct pfsync_softc *sc = V_pfsyncif; if (V_pf_status.debug >= PF_DEBUG_MISC) printf("pfsync: received bulk update request\n"); PFSYNC_BLOCK(sc); sc->sc_ureq_received = time_uptime; sc->sc_bulk_hashid = 0; sc->sc_bulk_stateid = 0; pfsync_bulk_status(PFSYNC_BUS_START); callout_reset(&sc->sc_bulk_tmo, 1, pfsync_bulk_update, sc); PFSYNC_BUNLOCK(sc); } static void pfsync_bulk_update(void *arg) { struct pfsync_softc *sc = arg; struct pf_state *s; int i, sent = 0; PFSYNC_BLOCK_ASSERT(sc); CURVNET_SET(sc->sc_ifp->if_vnet); /* * Start with last state from previous invocation. * It may had gone, in this case start from the * hash slot. */ s = pf_find_state_byid(sc->sc_bulk_stateid, sc->sc_bulk_creatorid); if (s != NULL) i = PF_IDHASH(s); else i = sc->sc_bulk_hashid; for (; i <= pf_hashmask; i++) { struct pf_idhash *ih = &V_pf_idhash[i]; if (s != NULL) PF_HASHROW_ASSERT(ih); else { PF_HASHROW_LOCK(ih); s = LIST_FIRST(&ih->states); } for (; s; s = LIST_NEXT(s, entry)) { if (sent > 1 && (sc->sc_ifp->if_mtu - sc->sc_len) < sizeof(struct pfsync_state)) { /* We've filled a packet. */ sc->sc_bulk_hashid = i; sc->sc_bulk_stateid = s->id; sc->sc_bulk_creatorid = s->creatorid; PF_HASHROW_UNLOCK(ih); callout_reset(&sc->sc_bulk_tmo, 1, pfsync_bulk_update, sc); goto full; } if (s->sync_state == PFSYNC_S_NONE && s->timeout < PFTM_MAX && s->pfsync_time <= sc->sc_ureq_received) { pfsync_update_state_req(s); sent++; } } PF_HASHROW_UNLOCK(ih); } /* We're done. */ pfsync_bulk_status(PFSYNC_BUS_END); full: CURVNET_RESTORE(); } static void pfsync_bulk_status(u_int8_t status) { struct { struct pfsync_subheader subh; struct pfsync_bus bus; } __packed r; struct pfsync_softc *sc = V_pfsyncif; bzero(&r, sizeof(r)); r.subh.action = PFSYNC_ACT_BUS; r.subh.count = htons(1); V_pfsyncstats.pfsyncs_oacts[PFSYNC_ACT_BUS]++; r.bus.creatorid = V_pf_status.hostid; r.bus.endtime = htonl(time_uptime - sc->sc_ureq_received); r.bus.status = status; PFSYNC_LOCK(sc); pfsync_send_plus(&r, sizeof(r)); PFSYNC_UNLOCK(sc); } static void pfsync_bulk_fail(void *arg) { struct pfsync_softc *sc = arg; CURVNET_SET(sc->sc_ifp->if_vnet); PFSYNC_BLOCK_ASSERT(sc); if (sc->sc_bulk_tries++ < PFSYNC_MAX_BULKTRIES) { /* Try again */ callout_reset(&sc->sc_bulkfail_tmo, 5 * hz, pfsync_bulk_fail, V_pfsyncif); PFSYNC_LOCK(sc); pfsync_request_update(0, 0); PFSYNC_UNLOCK(sc); } else { /* Pretend like the transfer was ok. */ sc->sc_ureq_sent = 0; sc->sc_bulk_tries = 0; PFSYNC_LOCK(sc); if (!(sc->sc_flags & PFSYNCF_OK) && carp_demote_adj_p) (*carp_demote_adj_p)(-V_pfsync_carp_adj, "pfsync bulk fail"); sc->sc_flags |= PFSYNCF_OK; PFSYNC_UNLOCK(sc); if (V_pf_status.debug >= PF_DEBUG_MISC) printf("pfsync: failed to receive bulk update\n"); } CURVNET_RESTORE(); } static void pfsync_send_plus(void *plus, size_t pluslen) { struct pfsync_softc *sc = V_pfsyncif; PFSYNC_LOCK_ASSERT(sc); if (sc->sc_len + pluslen > sc->sc_ifp->if_mtu) pfsync_sendout(1); sc->sc_plus = plus; sc->sc_len += (sc->sc_pluslen = pluslen); pfsync_sendout(1); } static void pfsync_timeout(void *arg) { struct pfsync_softc *sc = arg; CURVNET_SET(sc->sc_ifp->if_vnet); PFSYNC_LOCK(sc); pfsync_push(sc); PFSYNC_UNLOCK(sc); CURVNET_RESTORE(); } static void pfsync_push(struct pfsync_softc *sc) { PFSYNC_LOCK_ASSERT(sc); sc->sc_flags |= PFSYNCF_PUSH; swi_sched(V_pfsync_swi_cookie, 0); } static void pfsyncintr(void *arg) { struct pfsync_softc *sc = arg; struct mbuf *m, *n; CURVNET_SET(sc->sc_ifp->if_vnet); PFSYNC_LOCK(sc); if ((sc->sc_flags & PFSYNCF_PUSH) && sc->sc_len > PFSYNC_MINPKT) { pfsync_sendout(0); sc->sc_flags &= ~PFSYNCF_PUSH; } _IF_DEQUEUE_ALL(&sc->sc_ifp->if_snd, m); PFSYNC_UNLOCK(sc); for (; m != NULL; m = n) { n = m->m_nextpkt; m->m_nextpkt = NULL; /* * We distinguish between a deferral packet and our * own pfsync packet based on M_SKIP_FIREWALL * flag. This is XXX. */ if (m->m_flags & M_SKIP_FIREWALL) ip_output(m, NULL, NULL, 0, NULL, NULL); else if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL) == 0) V_pfsyncstats.pfsyncs_opackets++; else V_pfsyncstats.pfsyncs_oerrors++; } CURVNET_RESTORE(); } static int pfsync_multicast_setup(struct pfsync_softc *sc, struct ifnet *ifp, void *mship) { struct ip_moptions *imo = &sc->sc_imo; int error; if (!(ifp->if_flags & IFF_MULTICAST)) return (EADDRNOTAVAIL); imo->imo_membership = (struct in_multi **)mship; imo->imo_max_memberships = IP_MIN_MEMBERSHIPS; imo->imo_multicast_vif = -1; if ((error = in_joingroup(ifp, &sc->sc_sync_peer, NULL, &imo->imo_membership[0])) != 0) { imo->imo_membership = NULL; return (error); } imo->imo_num_memberships++; imo->imo_multicast_ifp = ifp; imo->imo_multicast_ttl = PFSYNC_DFLTTL; imo->imo_multicast_loop = 0; return (0); } static void pfsync_multicast_cleanup(struct pfsync_softc *sc) { struct ip_moptions *imo = &sc->sc_imo; in_leavegroup(imo->imo_membership[0], NULL); free(imo->imo_membership, M_PFSYNC); imo->imo_membership = NULL; imo->imo_multicast_ifp = NULL; } #ifdef INET extern struct domain inetdomain; static struct protosw in_pfsync_protosw = { .pr_type = SOCK_RAW, .pr_domain = &inetdomain, .pr_protocol = IPPROTO_PFSYNC, .pr_flags = PR_ATOMIC|PR_ADDR, .pr_input = pfsync_input, .pr_output = rip_output, .pr_ctloutput = rip_ctloutput, .pr_usrreqs = &rip_usrreqs }; #endif static void pfsync_pointers_init() { PF_RULES_WLOCK(); pfsync_state_import_ptr = pfsync_state_import; pfsync_insert_state_ptr = pfsync_insert_state; pfsync_update_state_ptr = pfsync_update_state; pfsync_delete_state_ptr = pfsync_delete_state; pfsync_clear_states_ptr = pfsync_clear_states; pfsync_defer_ptr = pfsync_defer; PF_RULES_WUNLOCK(); } static void pfsync_pointers_uninit() { PF_RULES_WLOCK(); pfsync_state_import_ptr = NULL; pfsync_insert_state_ptr = NULL; pfsync_update_state_ptr = NULL; pfsync_delete_state_ptr = NULL; pfsync_clear_states_ptr = NULL; pfsync_defer_ptr = NULL; PF_RULES_WUNLOCK(); } static int pfsync_init() { VNET_ITERATOR_DECL(vnet_iter); int error = 0; VNET_LIST_RLOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); V_pfsync_cloner = if_clone_simple(pfsyncname, pfsync_clone_create, pfsync_clone_destroy, 1); error = swi_add(NULL, pfsyncname, pfsyncintr, V_pfsyncif, SWI_NET, INTR_MPSAFE, &V_pfsync_swi_cookie); CURVNET_RESTORE(); if (error) goto fail_locked; } VNET_LIST_RUNLOCK(); #ifdef INET error = pf_proto_register(PF_INET, &in_pfsync_protosw); if (error) goto fail; error = ipproto_register(IPPROTO_PFSYNC); if (error) { pf_proto_unregister(PF_INET, IPPROTO_PFSYNC, SOCK_RAW); goto fail; } #endif pfsync_pointers_init(); return (0); fail: VNET_LIST_RLOCK(); fail_locked: VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); if (V_pfsync_swi_cookie) { swi_remove(V_pfsync_swi_cookie); if_clone_detach(V_pfsync_cloner); } CURVNET_RESTORE(); } VNET_LIST_RUNLOCK(); return (error); } static void pfsync_uninit() { VNET_ITERATOR_DECL(vnet_iter); pfsync_pointers_uninit(); ipproto_unregister(IPPROTO_PFSYNC); pf_proto_unregister(PF_INET, IPPROTO_PFSYNC, SOCK_RAW); VNET_LIST_RLOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); if_clone_detach(V_pfsync_cloner); swi_remove(V_pfsync_swi_cookie); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK(); } static int pfsync_modevent(module_t mod, int type, void *data) { int error = 0; switch (type) { case MOD_LOAD: error = pfsync_init(); break; case MOD_QUIESCE: /* * Module should not be unloaded due to race conditions. */ error = EBUSY; break; case MOD_UNLOAD: pfsync_uninit(); break; default: error = EINVAL; break; } return (error); } static moduledata_t pfsync_mod = { pfsyncname, pfsync_modevent, 0 }; #define PFSYNC_MODVER 1 +/* XXX-BZ recheck the r229853 comment once the shuffling is done. */ DECLARE_MODULE(pfsync, pfsync_mod, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY); MODULE_VERSION(pfsync, PFSYNC_MODVER); MODULE_DEPEND(pfsync, pf, PF_MODVER, PF_MODVER, PF_MODVER); Index: projects/vnet/sys/netpfil/pf/pf_ioctl.c =================================================================== --- projects/vnet/sys/netpfil/pf/pf_ioctl.c (revision 294565) +++ projects/vnet/sys/netpfil/pf/pf_ioctl.c (revision 294566) @@ -1,3793 +1,3793 @@ /*- * Copyright (c) 2001 Daniel Hartmeier * Copyright (c) 2002,2003 Henning Brauer * Copyright (c) 2012 Gleb Smirnoff * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * - Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * - Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials provided * with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. * * Effort sponsored in part by the Defense Advanced Research Projects * Agency (DARPA) and Air Force Research Laboratory, Air Force * Materiel Command, USAF, under agreement number F30602-01-2-0537. * * $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $ */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include "opt_bpf.h" #include "opt_pf.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif /* INET6 */ #ifdef ALTQ #include #endif static int pfattach(void); static struct pf_pool *pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t, u_int8_t, u_int8_t, u_int8_t); static void pf_mv_pool(struct pf_palist *, struct pf_palist *); static void pf_empty_pool(struct pf_palist *); static int pfioctl(struct cdev *, u_long, caddr_t, int, struct thread *); #ifdef ALTQ static int pf_begin_altq(u_int32_t *); static int pf_rollback_altq(u_int32_t); static int pf_commit_altq(u_int32_t); static int pf_enable_altq(struct pf_altq *); static int pf_disable_altq(struct pf_altq *); static u_int32_t pf_qname2qid(char *); static void pf_qid_unref(u_int32_t); #endif /* ALTQ */ static int pf_begin_rules(u_int32_t *, int, const char *); static int pf_rollback_rules(u_int32_t, int, char *); static int pf_setup_pfsync_matching(struct pf_ruleset *); static void pf_hash_rule(MD5_CTX *, struct pf_rule *); static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *); static int pf_commit_rules(u_int32_t, int, char *); static int pf_addr_setup(struct pf_ruleset *, struct pf_addr_wrap *, sa_family_t); static void pf_addr_copyout(struct pf_addr_wrap *); VNET_DEFINE(struct pf_rule, pf_default_rule); #ifdef ALTQ static VNET_DEFINE(int, pf_altq_running); #define V_pf_altq_running VNET(pf_altq_running) #endif #define TAGID_MAX 50000 struct pf_tagname { TAILQ_ENTRY(pf_tagname) entries; char name[PF_TAG_NAME_SIZE]; uint16_t tag; int ref; }; TAILQ_HEAD(pf_tags, pf_tagname); #define V_pf_tags VNET(pf_tags) VNET_DEFINE(struct pf_tags, pf_tags); #define V_pf_qids VNET(pf_qids) VNET_DEFINE(struct pf_tags, pf_qids); static MALLOC_DEFINE(M_PFTAG, "pf_tag", "pf(4) tag names"); static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db"); static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules"); #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE) #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE #endif static u_int16_t tagname2tag(struct pf_tags *, char *); static u_int16_t pf_tagname2tag(char *); static void tag_unref(struct pf_tags *, u_int16_t); #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x struct cdev *pf_dev; /* * XXX - These are new and need to be checked when moveing to a new version */ static void pf_clear_states(void); static int pf_clear_tables(void); static void pf_clear_srcnodes(struct pf_src_node *); static void pf_kill_srcnodes(struct pfioc_src_node_kill *); static void pf_tbladdr_copyout(struct pf_addr_wrap *); /* * Wrapper functions for pfil(9) hooks */ #ifdef INET static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, struct inpcb *inp); static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, struct inpcb *inp); #endif #ifdef INET6 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, struct inpcb *inp); static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, struct inpcb *inp); #endif static int hook_pf(void); static int dehook_pf(void); static int shutdown_pf(void); static int pf_load(void); static int pf_unload(void); static struct cdevsw pf_cdevsw = { .d_ioctl = pfioctl, .d_name = PF_NAME, .d_version = D_VERSION, }; static volatile VNET_DEFINE(int, pf_pfil_hooked); #define V_pf_pfil_hooked VNET(pf_pfil_hooked) VNET_DEFINE(int, pf_end_threads); struct rwlock pf_rules_lock; struct sx pf_ioctl_lock; /* pfsync */ pfsync_state_import_t *pfsync_state_import_ptr = NULL; pfsync_insert_state_t *pfsync_insert_state_ptr = NULL; pfsync_update_state_t *pfsync_update_state_ptr = NULL; pfsync_delete_state_t *pfsync_delete_state_ptr = NULL; pfsync_clear_states_t *pfsync_clear_states_ptr = NULL; pfsync_defer_t *pfsync_defer_ptr = NULL; /* pflog */ pflog_packet_t *pflog_packet_ptr = NULL; static int pfattach(void) { u_int32_t *my_timeout = V_pf_default_rule.timeout; int error; if (IS_DEFAULT_VNET(curvnet)) pf_mtag_initialize(); pf_initialize(); pfr_initialize(); pfi_initialize(); pf_normalize_init(); V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT; V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT; RB_INIT(&V_pf_anchors); pf_init_ruleset(&pf_main_ruleset); /* default rule should never be garbage collected */ V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next; #ifdef PF_DEFAULT_TO_DROP V_pf_default_rule.action = PF_DROP; #else V_pf_default_rule.action = PF_PASS; #endif V_pf_default_rule.nr = -1; V_pf_default_rule.rtableid = -1; V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK); V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK); V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK); /* initialize default timeouts */ my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL; my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL; my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL; my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL; my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL; my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL; my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL; my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL; my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL; my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL; my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL; my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL; my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL; my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL; my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL; my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL; my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL; my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL; my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START; my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END; bzero(&V_pf_status, sizeof(V_pf_status)); V_pf_status.debug = PF_DEBUG_URGENT; V_pf_pfil_hooked = 0; /* XXX do our best to avoid a conflict */ V_pf_status.hostid = arc4random(); for (int i = 0; i < PFRES_MAX; i++) V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK); for (int i = 0; i < LCNT_MAX; i++) V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK); for (int i = 0; i < FCNT_MAX; i++) V_pf_status.fcounters[i] = counter_u64_alloc(M_WAITOK); for (int i = 0; i < SCNT_MAX; i++) V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK); if ((error = kproc_create(pf_purge_thread, curvnet, NULL, 0, 0, "pf purge")) != 0) /* XXXGL: leaked all above. */ return (error); if ((error = swi_add(NULL, "pf send", pf_intr, curvnet, SWI_NET, INTR_MPSAFE, &V_pf_swi_cookie)) != 0) /* XXXGL: leaked all above. */ return (error); return (0); } static struct pf_pool * pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action, u_int32_t rule_number, u_int8_t r_last, u_int8_t active, u_int8_t check_ticket) { struct pf_ruleset *ruleset; struct pf_rule *rule; int rs_num; ruleset = pf_find_ruleset(anchor); if (ruleset == NULL) return (NULL); rs_num = pf_get_ruleset_number(rule_action); if (rs_num >= PF_RULESET_MAX) return (NULL); if (active) { if (check_ticket && ticket != ruleset->rules[rs_num].active.ticket) return (NULL); if (r_last) rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, pf_rulequeue); else rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); } else { if (check_ticket && ticket != ruleset->rules[rs_num].inactive.ticket) return (NULL); if (r_last) rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, pf_rulequeue); else rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr); } if (!r_last) { while ((rule != NULL) && (rule->nr != rule_number)) rule = TAILQ_NEXT(rule, entries); } if (rule == NULL) return (NULL); return (&rule->rpool); } static void pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb) { struct pf_pooladdr *mv_pool_pa; while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) { TAILQ_REMOVE(poola, mv_pool_pa, entries); TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries); } } static void pf_empty_pool(struct pf_palist *poola) { struct pf_pooladdr *pa; while ((pa = TAILQ_FIRST(poola)) != NULL) { switch (pa->addr.type) { case PF_ADDR_DYNIFTL: pfi_dynaddr_remove(pa->addr.p.dyn); break; case PF_ADDR_TABLE: /* XXX: this could be unfinished pooladdr on pabuf */ if (pa->addr.p.tbl != NULL) pfr_detach_table(pa->addr.p.tbl); break; } if (pa->kif) pfi_kif_unref(pa->kif); TAILQ_REMOVE(poola, pa, entries); free(pa, M_PFRULE); } } static void pf_unlink_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule) { PF_RULES_WASSERT(); TAILQ_REMOVE(rulequeue, rule, entries); PF_UNLNKDRULES_LOCK(); rule->rule_flag |= PFRULE_REFS; TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries); PF_UNLNKDRULES_UNLOCK(); } void pf_free_rule(struct pf_rule *rule) { PF_RULES_WASSERT(); if (rule->tag) tag_unref(&V_pf_tags, rule->tag); if (rule->match_tag) tag_unref(&V_pf_tags, rule->match_tag); #ifdef ALTQ if (rule->pqid != rule->qid) pf_qid_unref(rule->pqid); pf_qid_unref(rule->qid); #endif switch (rule->src.addr.type) { case PF_ADDR_DYNIFTL: pfi_dynaddr_remove(rule->src.addr.p.dyn); break; case PF_ADDR_TABLE: pfr_detach_table(rule->src.addr.p.tbl); break; } switch (rule->dst.addr.type) { case PF_ADDR_DYNIFTL: pfi_dynaddr_remove(rule->dst.addr.p.dyn); break; case PF_ADDR_TABLE: pfr_detach_table(rule->dst.addr.p.tbl); break; } if (rule->overload_tbl) pfr_detach_table(rule->overload_tbl); if (rule->kif) pfi_kif_unref(rule->kif); pf_anchor_remove(rule); pf_empty_pool(&rule->rpool.list); counter_u64_free(rule->states_cur); counter_u64_free(rule->states_tot); counter_u64_free(rule->src_nodes); free(rule, M_PFRULE); } static u_int16_t tagname2tag(struct pf_tags *head, char *tagname) { struct pf_tagname *tag, *p = NULL; u_int16_t new_tagid = 1; PF_RULES_WASSERT(); TAILQ_FOREACH(tag, head, entries) if (strcmp(tagname, tag->name) == 0) { tag->ref++; return (tag->tag); } /* * to avoid fragmentation, we do a linear search from the beginning * and take the first free slot we find. if there is none or the list * is empty, append a new entry at the end. */ /* new entry */ if (!TAILQ_EMPTY(head)) for (p = TAILQ_FIRST(head); p != NULL && p->tag == new_tagid; p = TAILQ_NEXT(p, entries)) new_tagid = p->tag + 1; if (new_tagid > TAGID_MAX) return (0); /* allocate and fill new struct pf_tagname */ tag = malloc(sizeof(*tag), M_PFTAG, M_NOWAIT|M_ZERO); if (tag == NULL) return (0); strlcpy(tag->name, tagname, sizeof(tag->name)); tag->tag = new_tagid; tag->ref++; if (p != NULL) /* insert new entry before p */ TAILQ_INSERT_BEFORE(p, tag, entries); else /* either list empty or no free slot in between */ TAILQ_INSERT_TAIL(head, tag, entries); return (tag->tag); } static void tag_unref(struct pf_tags *head, u_int16_t tag) { struct pf_tagname *p, *next; PF_RULES_WASSERT(); for (p = TAILQ_FIRST(head); p != NULL; p = next) { next = TAILQ_NEXT(p, entries); if (tag == p->tag) { if (--p->ref == 0) { TAILQ_REMOVE(head, p, entries); free(p, M_PFTAG); } break; } } } static u_int16_t pf_tagname2tag(char *tagname) { return (tagname2tag(&V_pf_tags, tagname)); } #ifdef ALTQ static u_int32_t pf_qname2qid(char *qname) { return ((u_int32_t)tagname2tag(&V_pf_qids, qname)); } static void pf_qid_unref(u_int32_t qid) { tag_unref(&V_pf_qids, (u_int16_t)qid); } static int pf_begin_altq(u_int32_t *ticket) { struct pf_altq *altq; int error = 0; PF_RULES_WASSERT(); /* Purge the old altq list */ while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) { TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries); if (altq->qname[0] == 0 && (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { /* detach and destroy the discipline */ error = altq_remove(altq); } else pf_qid_unref(altq->qid); free(altq, M_PFALTQ); } if (error) return (error); *ticket = ++V_ticket_altqs_inactive; V_altqs_inactive_open = 1; return (0); } static int pf_rollback_altq(u_int32_t ticket) { struct pf_altq *altq; int error = 0; PF_RULES_WASSERT(); if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) return (0); /* Purge the old altq list */ while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) { TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries); if (altq->qname[0] == 0 && (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { /* detach and destroy the discipline */ error = altq_remove(altq); } else pf_qid_unref(altq->qid); free(altq, M_PFALTQ); } V_altqs_inactive_open = 0; return (error); } static int pf_commit_altq(u_int32_t ticket) { struct pf_altqqueue *old_altqs; struct pf_altq *altq; int err, error = 0; PF_RULES_WASSERT(); if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive) return (EBUSY); /* swap altqs, keep the old. */ old_altqs = V_pf_altqs_active; V_pf_altqs_active = V_pf_altqs_inactive; V_pf_altqs_inactive = old_altqs; V_ticket_altqs_active = V_ticket_altqs_inactive; /* Attach new disciplines */ TAILQ_FOREACH(altq, V_pf_altqs_active, entries) { if (altq->qname[0] == 0 && (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { /* attach the discipline */ error = altq_pfattach(altq); if (error == 0 && V_pf_altq_running) error = pf_enable_altq(altq); if (error != 0) return (error); } } /* Purge the old altq list */ while ((altq = TAILQ_FIRST(V_pf_altqs_inactive)) != NULL) { TAILQ_REMOVE(V_pf_altqs_inactive, altq, entries); if (altq->qname[0] == 0 && (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { /* detach and destroy the discipline */ if (V_pf_altq_running) error = pf_disable_altq(altq); err = altq_pfdetach(altq); if (err != 0 && error == 0) error = err; err = altq_remove(altq); if (err != 0 && error == 0) error = err; } else pf_qid_unref(altq->qid); free(altq, M_PFALTQ); } V_altqs_inactive_open = 0; return (error); } static int pf_enable_altq(struct pf_altq *altq) { struct ifnet *ifp; struct tb_profile tb; int error = 0; if ((ifp = ifunit(altq->ifname)) == NULL) return (EINVAL); if (ifp->if_snd.altq_type != ALTQT_NONE) error = altq_enable(&ifp->if_snd); /* set tokenbucket regulator */ if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) { tb.rate = altq->ifbandwidth; tb.depth = altq->tbrsize; error = tbr_set(&ifp->if_snd, &tb); } return (error); } static int pf_disable_altq(struct pf_altq *altq) { struct ifnet *ifp; struct tb_profile tb; int error; if ((ifp = ifunit(altq->ifname)) == NULL) return (EINVAL); /* * when the discipline is no longer referenced, it was overridden * by a new one. if so, just return. */ if (altq->altq_disc != ifp->if_snd.altq_disc) return (0); error = altq_disable(&ifp->if_snd); if (error == 0) { /* clear tokenbucket regulator */ tb.rate = 0; error = tbr_set(&ifp->if_snd, &tb); } return (error); } void pf_altq_ifnet_event(struct ifnet *ifp, int remove) { struct ifnet *ifp1; struct pf_altq *a1, *a2, *a3; u_int32_t ticket; int error = 0; /* Interrupt userland queue modifications */ if (V_altqs_inactive_open) pf_rollback_altq(V_ticket_altqs_inactive); /* Start new altq ruleset */ if (pf_begin_altq(&ticket)) return; /* Copy the current active set */ TAILQ_FOREACH(a1, V_pf_altqs_active, entries) { a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT); if (a2 == NULL) { error = ENOMEM; break; } bcopy(a1, a2, sizeof(struct pf_altq)); if (a2->qname[0] != 0) { if ((a2->qid = pf_qname2qid(a2->qname)) == 0) { error = EBUSY; free(a2, M_PFALTQ); break; } a2->altq_disc = NULL; TAILQ_FOREACH(a3, V_pf_altqs_inactive, entries) { if (strncmp(a3->ifname, a2->ifname, IFNAMSIZ) == 0 && a3->qname[0] == 0) { a2->altq_disc = a3->altq_disc; break; } } } /* Deactivate the interface in question */ a2->local_flags &= ~PFALTQ_FLAG_IF_REMOVED; if ((ifp1 = ifunit(a2->ifname)) == NULL || (remove && ifp1 == ifp)) { a2->local_flags |= PFALTQ_FLAG_IF_REMOVED; } else { error = altq_add(a2); if (ticket != V_ticket_altqs_inactive) error = EBUSY; if (error) { free(a2, M_PFALTQ); break; } } TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries); } if (error != 0) pf_rollback_altq(ticket); else pf_commit_altq(ticket); } #endif /* ALTQ */ static int pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor) { struct pf_ruleset *rs; struct pf_rule *rule; PF_RULES_WASSERT(); if (rs_num < 0 || rs_num >= PF_RULESET_MAX) return (EINVAL); rs = pf_find_or_create_ruleset(anchor); if (rs == NULL) return (EINVAL); while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); rs->rules[rs_num].inactive.rcount--; } *ticket = ++rs->rules[rs_num].inactive.ticket; rs->rules[rs_num].inactive.open = 1; return (0); } static int pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor) { struct pf_ruleset *rs; struct pf_rule *rule; PF_RULES_WASSERT(); if (rs_num < 0 || rs_num >= PF_RULESET_MAX) return (EINVAL); rs = pf_find_ruleset(anchor); if (rs == NULL || !rs->rules[rs_num].inactive.open || rs->rules[rs_num].inactive.ticket != ticket) return (0); while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) { pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule); rs->rules[rs_num].inactive.rcount--; } rs->rules[rs_num].inactive.open = 0; return (0); } #define PF_MD5_UPD(st, elm) \ MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm)) #define PF_MD5_UPD_STR(st, elm) \ MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm)) #define PF_MD5_UPD_HTONL(st, elm, stor) do { \ (stor) = htonl((st)->elm); \ MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\ } while (0) #define PF_MD5_UPD_HTONS(st, elm, stor) do { \ (stor) = htons((st)->elm); \ MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\ } while (0) static void pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr) { PF_MD5_UPD(pfr, addr.type); switch (pfr->addr.type) { case PF_ADDR_DYNIFTL: PF_MD5_UPD(pfr, addr.v.ifname); PF_MD5_UPD(pfr, addr.iflags); break; case PF_ADDR_TABLE: PF_MD5_UPD(pfr, addr.v.tblname); break; case PF_ADDR_ADDRMASK: /* XXX ignore af? */ PF_MD5_UPD(pfr, addr.v.a.addr.addr32); PF_MD5_UPD(pfr, addr.v.a.mask.addr32); break; } PF_MD5_UPD(pfr, port[0]); PF_MD5_UPD(pfr, port[1]); PF_MD5_UPD(pfr, neg); PF_MD5_UPD(pfr, port_op); } static void pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule) { u_int16_t x; u_int32_t y; pf_hash_rule_addr(ctx, &rule->src); pf_hash_rule_addr(ctx, &rule->dst); PF_MD5_UPD_STR(rule, label); PF_MD5_UPD_STR(rule, ifname); PF_MD5_UPD_STR(rule, match_tagname); PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */ PF_MD5_UPD_HTONL(rule, os_fingerprint, y); PF_MD5_UPD_HTONL(rule, prob, y); PF_MD5_UPD_HTONL(rule, uid.uid[0], y); PF_MD5_UPD_HTONL(rule, uid.uid[1], y); PF_MD5_UPD(rule, uid.op); PF_MD5_UPD_HTONL(rule, gid.gid[0], y); PF_MD5_UPD_HTONL(rule, gid.gid[1], y); PF_MD5_UPD(rule, gid.op); PF_MD5_UPD_HTONL(rule, rule_flag, y); PF_MD5_UPD(rule, action); PF_MD5_UPD(rule, direction); PF_MD5_UPD(rule, af); PF_MD5_UPD(rule, quick); PF_MD5_UPD(rule, ifnot); PF_MD5_UPD(rule, match_tag_not); PF_MD5_UPD(rule, natpass); PF_MD5_UPD(rule, keep_state); PF_MD5_UPD(rule, proto); PF_MD5_UPD(rule, type); PF_MD5_UPD(rule, code); PF_MD5_UPD(rule, flags); PF_MD5_UPD(rule, flagset); PF_MD5_UPD(rule, allow_opts); PF_MD5_UPD(rule, rt); PF_MD5_UPD(rule, tos); } static int pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) { struct pf_ruleset *rs; struct pf_rule *rule, **old_array; struct pf_rulequeue *old_rules; int error; u_int32_t old_rcount; PF_RULES_WASSERT(); if (rs_num < 0 || rs_num >= PF_RULESET_MAX) return (EINVAL); rs = pf_find_ruleset(anchor); if (rs == NULL || !rs->rules[rs_num].inactive.open || ticket != rs->rules[rs_num].inactive.ticket) return (EBUSY); /* Calculate checksum for the main ruleset */ if (rs == &pf_main_ruleset) { error = pf_setup_pfsync_matching(rs); if (error != 0) return (error); } /* Swap rules, keep the old. */ old_rules = rs->rules[rs_num].active.ptr; old_rcount = rs->rules[rs_num].active.rcount; old_array = rs->rules[rs_num].active.ptr_array; rs->rules[rs_num].active.ptr = rs->rules[rs_num].inactive.ptr; rs->rules[rs_num].active.ptr_array = rs->rules[rs_num].inactive.ptr_array; rs->rules[rs_num].active.rcount = rs->rules[rs_num].inactive.rcount; rs->rules[rs_num].inactive.ptr = old_rules; rs->rules[rs_num].inactive.ptr_array = old_array; rs->rules[rs_num].inactive.rcount = old_rcount; rs->rules[rs_num].active.ticket = rs->rules[rs_num].inactive.ticket; pf_calc_skip_steps(rs->rules[rs_num].active.ptr); /* Purge the old rule list. */ while ((rule = TAILQ_FIRST(old_rules)) != NULL) pf_unlink_rule(old_rules, rule); if (rs->rules[rs_num].inactive.ptr_array) free(rs->rules[rs_num].inactive.ptr_array, M_TEMP); rs->rules[rs_num].inactive.ptr_array = NULL; rs->rules[rs_num].inactive.rcount = 0; rs->rules[rs_num].inactive.open = 0; pf_remove_if_empty_ruleset(rs); return (0); } static int pf_setup_pfsync_matching(struct pf_ruleset *rs) { MD5_CTX ctx; struct pf_rule *rule; int rs_cnt; u_int8_t digest[PF_MD5_DIGEST_LENGTH]; MD5Init(&ctx); for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) { /* XXX PF_RULESET_SCRUB as well? */ if (rs_cnt == PF_RULESET_SCRUB) continue; if (rs->rules[rs_cnt].inactive.ptr_array) free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP); rs->rules[rs_cnt].inactive.ptr_array = NULL; if (rs->rules[rs_cnt].inactive.rcount) { rs->rules[rs_cnt].inactive.ptr_array = malloc(sizeof(caddr_t) * rs->rules[rs_cnt].inactive.rcount, M_TEMP, M_NOWAIT); if (!rs->rules[rs_cnt].inactive.ptr_array) return (ENOMEM); } TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr, entries) { pf_hash_rule(&ctx, rule); (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule; } } MD5Final(digest, &ctx); memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum)); return (0); } static int pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr, sa_family_t af) { int error = 0; switch (addr->type) { case PF_ADDR_TABLE: addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname); if (addr->p.tbl == NULL) error = ENOMEM; break; case PF_ADDR_DYNIFTL: error = pfi_dynaddr_setup(addr, af); break; } return (error); } static void pf_addr_copyout(struct pf_addr_wrap *addr) { switch (addr->type) { case PF_ADDR_DYNIFTL: pfi_dynaddr_copyout(addr); break; case PF_ADDR_TABLE: pf_tbladdr_copyout(addr); break; } } static int pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td) { int error = 0; /* XXX keep in sync with switch() below */ if (securelevel_gt(td->td_ucred, 2)) switch (cmd) { case DIOCGETRULES: case DIOCGETRULE: case DIOCGETADDRS: case DIOCGETADDR: case DIOCGETSTATE: case DIOCSETSTATUSIF: case DIOCGETSTATUS: case DIOCCLRSTATUS: case DIOCNATLOOK: case DIOCSETDEBUG: case DIOCGETSTATES: case DIOCGETTIMEOUT: case DIOCCLRRULECTRS: case DIOCGETLIMIT: case DIOCGETALTQS: case DIOCGETALTQ: case DIOCGETQSTATS: case DIOCGETRULESETS: case DIOCGETRULESET: case DIOCRGETTABLES: case DIOCRGETTSTATS: case DIOCRCLRTSTATS: case DIOCRCLRADDRS: case DIOCRADDADDRS: case DIOCRDELADDRS: case DIOCRSETADDRS: case DIOCRGETADDRS: case DIOCRGETASTATS: case DIOCRCLRASTATS: case DIOCRTSTADDRS: case DIOCOSFPGET: case DIOCGETSRCNODES: case DIOCCLRSRCNODES: case DIOCIGETIFACES: case DIOCGIFSPEED: case DIOCSETIFFLAG: case DIOCCLRIFFLAG: break; case DIOCRCLRTABLES: case DIOCRADDTABLES: case DIOCRDELTABLES: case DIOCRSETTFLAGS: if (((struct pfioc_table *)addr)->pfrio_flags & PFR_FLAG_DUMMY) break; /* dummy operation ok */ return (EPERM); default: return (EPERM); } if (!(flags & FWRITE)) switch (cmd) { case DIOCGETRULES: case DIOCGETADDRS: case DIOCGETADDR: case DIOCGETSTATE: case DIOCGETSTATUS: case DIOCGETSTATES: case DIOCGETTIMEOUT: case DIOCGETLIMIT: case DIOCGETALTQS: case DIOCGETALTQ: case DIOCGETQSTATS: case DIOCGETRULESETS: case DIOCGETRULESET: case DIOCNATLOOK: case DIOCRGETTABLES: case DIOCRGETTSTATS: case DIOCRGETADDRS: case DIOCRGETASTATS: case DIOCRTSTADDRS: case DIOCOSFPGET: case DIOCGETSRCNODES: case DIOCIGETIFACES: case DIOCGIFSPEED: break; case DIOCRCLRTABLES: case DIOCRADDTABLES: case DIOCRDELTABLES: case DIOCRCLRTSTATS: case DIOCRCLRADDRS: case DIOCRADDADDRS: case DIOCRDELADDRS: case DIOCRSETADDRS: case DIOCRSETTFLAGS: if (((struct pfioc_table *)addr)->pfrio_flags & PFR_FLAG_DUMMY) { flags |= FWRITE; /* need write lock for dummy */ break; /* dummy operation ok */ } return (EACCES); case DIOCGETRULE: if (((struct pfioc_rule *)addr)->action == PF_GET_CLR_CNTR) return (EACCES); break; default: return (EACCES); } CURVNET_SET(TD_TO_VNET(td)); switch (cmd) { case DIOCSTART: sx_xlock(&pf_ioctl_lock); if (V_pf_status.running) error = EEXIST; else { int cpu; error = hook_pf(); if (error) { DPFPRINTF(PF_DEBUG_MISC, ("pf: pfil registration failed\n")); break; } V_pf_status.running = 1; V_pf_status.since = time_second; CPU_FOREACH(cpu) V_pf_stateid[cpu] = time_second; DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n")); } break; case DIOCSTOP: sx_xlock(&pf_ioctl_lock); if (!V_pf_status.running) error = ENOENT; else { V_pf_status.running = 0; error = dehook_pf(); if (error) { V_pf_status.running = 1; DPFPRINTF(PF_DEBUG_MISC, ("pf: pfil unregistration failed\n")); } V_pf_status.since = time_second; DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n")); } break; case DIOCADDRULE: { struct pfioc_rule *pr = (struct pfioc_rule *)addr; struct pf_ruleset *ruleset; struct pf_rule *rule, *tail; struct pf_pooladdr *pa; struct pfi_kif *kif = NULL; int rs_num; if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) { error = EINVAL; break; } #ifndef INET if (pr->rule.af == AF_INET) { error = EAFNOSUPPORT; break; } #endif /* INET */ #ifndef INET6 if (pr->rule.af == AF_INET6) { error = EAFNOSUPPORT; break; } #endif /* INET6 */ rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK); bcopy(&pr->rule, rule, sizeof(struct pf_rule)); if (rule->ifname[0]) kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK); rule->states_cur = counter_u64_alloc(M_WAITOK); rule->states_tot = counter_u64_alloc(M_WAITOK); rule->src_nodes = counter_u64_alloc(M_WAITOK); rule->cuid = td->td_ucred->cr_ruid; rule->cpid = td->td_proc ? td->td_proc->p_pid : 0; TAILQ_INIT(&rule->rpool.list); #define ERROUT(x) { error = (x); goto DIOCADDRULE_error; } PF_RULES_WLOCK(); pr->anchor[sizeof(pr->anchor) - 1] = 0; ruleset = pf_find_ruleset(pr->anchor); if (ruleset == NULL) ERROUT(EINVAL); rs_num = pf_get_ruleset_number(pr->rule.action); if (rs_num >= PF_RULESET_MAX) ERROUT(EINVAL); if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) { DPFPRINTF(PF_DEBUG_MISC, ("ticket: %d != [%d]%d\n", pr->ticket, rs_num, ruleset->rules[rs_num].inactive.ticket)); ERROUT(EBUSY); } if (pr->pool_ticket != V_ticket_pabuf) { DPFPRINTF(PF_DEBUG_MISC, ("pool_ticket: %d != %d\n", pr->pool_ticket, V_ticket_pabuf)); ERROUT(EBUSY); } tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr, pf_rulequeue); if (tail) rule->nr = tail->nr + 1; else rule->nr = 0; if (rule->ifname[0]) { rule->kif = pfi_kif_attach(kif, rule->ifname); pfi_kif_ref(rule->kif); } else rule->kif = NULL; if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs) error = EBUSY; #ifdef ALTQ /* set queue IDs */ if (rule->qname[0] != 0) { if ((rule->qid = pf_qname2qid(rule->qname)) == 0) error = EBUSY; else if (rule->pqname[0] != 0) { if ((rule->pqid = pf_qname2qid(rule->pqname)) == 0) error = EBUSY; } else rule->pqid = rule->qid; } #endif if (rule->tagname[0]) if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) error = EBUSY; if (rule->match_tagname[0]) if ((rule->match_tag = pf_tagname2tag(rule->match_tagname)) == 0) error = EBUSY; if (rule->rt && !rule->direction) error = EINVAL; if (!rule->log) rule->logif = 0; if (rule->logif >= PFLOGIFS_MAX) error = EINVAL; if (pf_addr_setup(ruleset, &rule->src.addr, rule->af)) error = ENOMEM; if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af)) error = ENOMEM; if (pf_anchor_setup(rule, ruleset, pr->anchor_call)) error = EINVAL; TAILQ_FOREACH(pa, &V_pf_pabuf, entries) if (pa->addr.type == PF_ADDR_TABLE) { pa->addr.p.tbl = pfr_attach_table(ruleset, pa->addr.v.tblname); if (pa->addr.p.tbl == NULL) error = ENOMEM; } if (rule->overload_tblname[0]) { if ((rule->overload_tbl = pfr_attach_table(ruleset, rule->overload_tblname)) == NULL) error = EINVAL; else rule->overload_tbl->pfrkt_flags |= PFR_TFLAG_ACTIVE; } pf_mv_pool(&V_pf_pabuf, &rule->rpool.list); if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) || (rule->action == PF_BINAT)) && rule->anchor == NULL) || (rule->rt > PF_FASTROUTE)) && (TAILQ_FIRST(&rule->rpool.list) == NULL)) error = EINVAL; if (error) { pf_free_rule(rule); PF_RULES_WUNLOCK(); break; } rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list); rule->evaluations = rule->packets[0] = rule->packets[1] = rule->bytes[0] = rule->bytes[1] = 0; TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr, rule, entries); ruleset->rules[rs_num].inactive.rcount++; PF_RULES_WUNLOCK(); break; #undef ERROUT DIOCADDRULE_error: PF_RULES_WUNLOCK(); counter_u64_free(rule->states_cur); counter_u64_free(rule->states_tot); counter_u64_free(rule->src_nodes); free(rule, M_PFRULE); if (kif) free(kif, PFI_MTYPE); break; } case DIOCGETRULES: { struct pfioc_rule *pr = (struct pfioc_rule *)addr; struct pf_ruleset *ruleset; struct pf_rule *tail; int rs_num; PF_RULES_WLOCK(); pr->anchor[sizeof(pr->anchor) - 1] = 0; ruleset = pf_find_ruleset(pr->anchor); if (ruleset == NULL) { PF_RULES_WUNLOCK(); error = EINVAL; break; } rs_num = pf_get_ruleset_number(pr->rule.action); if (rs_num >= PF_RULESET_MAX) { PF_RULES_WUNLOCK(); error = EINVAL; break; } tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr, pf_rulequeue); if (tail) pr->nr = tail->nr + 1; else pr->nr = 0; pr->ticket = ruleset->rules[rs_num].active.ticket; PF_RULES_WUNLOCK(); break; } case DIOCGETRULE: { struct pfioc_rule *pr = (struct pfioc_rule *)addr; struct pf_ruleset *ruleset; struct pf_rule *rule; int rs_num, i; PF_RULES_WLOCK(); pr->anchor[sizeof(pr->anchor) - 1] = 0; ruleset = pf_find_ruleset(pr->anchor); if (ruleset == NULL) { PF_RULES_WUNLOCK(); error = EINVAL; break; } rs_num = pf_get_ruleset_number(pr->rule.action); if (rs_num >= PF_RULESET_MAX) { PF_RULES_WUNLOCK(); error = EINVAL; break; } if (pr->ticket != ruleset->rules[rs_num].active.ticket) { PF_RULES_WUNLOCK(); error = EBUSY; break; } rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr); while ((rule != NULL) && (rule->nr != pr->nr)) rule = TAILQ_NEXT(rule, entries); if (rule == NULL) { PF_RULES_WUNLOCK(); error = EBUSY; break; } bcopy(rule, &pr->rule, sizeof(struct pf_rule)); pr->rule.u_states_cur = counter_u64_fetch(rule->states_cur); pr->rule.u_states_tot = counter_u64_fetch(rule->states_tot); pr->rule.u_src_nodes = counter_u64_fetch(rule->src_nodes); if (pf_anchor_copyout(ruleset, rule, pr)) { PF_RULES_WUNLOCK(); error = EBUSY; break; } pf_addr_copyout(&pr->rule.src.addr); pf_addr_copyout(&pr->rule.dst.addr); for (i = 0; i < PF_SKIP_COUNT; ++i) if (rule->skip[i].ptr == NULL) pr->rule.skip[i].nr = -1; else pr->rule.skip[i].nr = rule->skip[i].ptr->nr; if (pr->action == PF_GET_CLR_CNTR) { rule->evaluations = 0; rule->packets[0] = rule->packets[1] = 0; rule->bytes[0] = rule->bytes[1] = 0; counter_u64_zero(rule->states_tot); } PF_RULES_WUNLOCK(); break; } case DIOCCHANGERULE: { struct pfioc_rule *pcr = (struct pfioc_rule *)addr; struct pf_ruleset *ruleset; struct pf_rule *oldrule = NULL, *newrule = NULL; struct pfi_kif *kif = NULL; struct pf_pooladdr *pa; u_int32_t nr = 0; int rs_num; if (pcr->action < PF_CHANGE_ADD_HEAD || pcr->action > PF_CHANGE_GET_TICKET) { error = EINVAL; break; } if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) { error = EINVAL; break; } if (pcr->action != PF_CHANGE_REMOVE) { #ifndef INET if (pcr->rule.af == AF_INET) { error = EAFNOSUPPORT; break; } #endif /* INET */ #ifndef INET6 if (pcr->rule.af == AF_INET6) { error = EAFNOSUPPORT; break; } #endif /* INET6 */ newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK); bcopy(&pcr->rule, newrule, sizeof(struct pf_rule)); if (newrule->ifname[0]) kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK); newrule->states_cur = counter_u64_alloc(M_WAITOK); newrule->states_tot = counter_u64_alloc(M_WAITOK); newrule->src_nodes = counter_u64_alloc(M_WAITOK); newrule->cuid = td->td_ucred->cr_ruid; newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0; TAILQ_INIT(&newrule->rpool.list); } #define ERROUT(x) { error = (x); goto DIOCCHANGERULE_error; } PF_RULES_WLOCK(); if (!(pcr->action == PF_CHANGE_REMOVE || pcr->action == PF_CHANGE_GET_TICKET) && pcr->pool_ticket != V_ticket_pabuf) ERROUT(EBUSY); ruleset = pf_find_ruleset(pcr->anchor); if (ruleset == NULL) ERROUT(EINVAL); rs_num = pf_get_ruleset_number(pcr->rule.action); if (rs_num >= PF_RULESET_MAX) ERROUT(EINVAL); if (pcr->action == PF_CHANGE_GET_TICKET) { pcr->ticket = ++ruleset->rules[rs_num].active.ticket; ERROUT(0); } else if (pcr->ticket != ruleset->rules[rs_num].active.ticket) ERROUT(EINVAL); if (pcr->action != PF_CHANGE_REMOVE) { if (newrule->ifname[0]) { newrule->kif = pfi_kif_attach(kif, newrule->ifname); pfi_kif_ref(newrule->kif); } else newrule->kif = NULL; if (newrule->rtableid > 0 && newrule->rtableid >= rt_numfibs) error = EBUSY; #ifdef ALTQ /* set queue IDs */ if (newrule->qname[0] != 0) { if ((newrule->qid = pf_qname2qid(newrule->qname)) == 0) error = EBUSY; else if (newrule->pqname[0] != 0) { if ((newrule->pqid = pf_qname2qid(newrule->pqname)) == 0) error = EBUSY; } else newrule->pqid = newrule->qid; } #endif /* ALTQ */ if (newrule->tagname[0]) if ((newrule->tag = pf_tagname2tag(newrule->tagname)) == 0) error = EBUSY; if (newrule->match_tagname[0]) if ((newrule->match_tag = pf_tagname2tag( newrule->match_tagname)) == 0) error = EBUSY; if (newrule->rt && !newrule->direction) error = EINVAL; if (!newrule->log) newrule->logif = 0; if (newrule->logif >= PFLOGIFS_MAX) error = EINVAL; if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af)) error = ENOMEM; if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af)) error = ENOMEM; if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call)) error = EINVAL; TAILQ_FOREACH(pa, &V_pf_pabuf, entries) if (pa->addr.type == PF_ADDR_TABLE) { pa->addr.p.tbl = pfr_attach_table(ruleset, pa->addr.v.tblname); if (pa->addr.p.tbl == NULL) error = ENOMEM; } if (newrule->overload_tblname[0]) { if ((newrule->overload_tbl = pfr_attach_table( ruleset, newrule->overload_tblname)) == NULL) error = EINVAL; else newrule->overload_tbl->pfrkt_flags |= PFR_TFLAG_ACTIVE; } pf_mv_pool(&V_pf_pabuf, &newrule->rpool.list); if (((((newrule->action == PF_NAT) || (newrule->action == PF_RDR) || (newrule->action == PF_BINAT) || (newrule->rt > PF_FASTROUTE)) && !newrule->anchor)) && (TAILQ_FIRST(&newrule->rpool.list) == NULL)) error = EINVAL; if (error) { pf_free_rule(newrule); PF_RULES_WUNLOCK(); break; } newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list); newrule->evaluations = 0; newrule->packets[0] = newrule->packets[1] = 0; newrule->bytes[0] = newrule->bytes[1] = 0; } pf_empty_pool(&V_pf_pabuf); if (pcr->action == PF_CHANGE_ADD_HEAD) oldrule = TAILQ_FIRST( ruleset->rules[rs_num].active.ptr); else if (pcr->action == PF_CHANGE_ADD_TAIL) oldrule = TAILQ_LAST( ruleset->rules[rs_num].active.ptr, pf_rulequeue); else { oldrule = TAILQ_FIRST( ruleset->rules[rs_num].active.ptr); while ((oldrule != NULL) && (oldrule->nr != pcr->nr)) oldrule = TAILQ_NEXT(oldrule, entries); if (oldrule == NULL) { if (newrule != NULL) pf_free_rule(newrule); PF_RULES_WUNLOCK(); error = EINVAL; break; } } if (pcr->action == PF_CHANGE_REMOVE) { pf_unlink_rule(ruleset->rules[rs_num].active.ptr, oldrule); ruleset->rules[rs_num].active.rcount--; } else { if (oldrule == NULL) TAILQ_INSERT_TAIL( ruleset->rules[rs_num].active.ptr, newrule, entries); else if (pcr->action == PF_CHANGE_ADD_HEAD || pcr->action == PF_CHANGE_ADD_BEFORE) TAILQ_INSERT_BEFORE(oldrule, newrule, entries); else TAILQ_INSERT_AFTER( ruleset->rules[rs_num].active.ptr, oldrule, newrule, entries); ruleset->rules[rs_num].active.rcount++; } nr = 0; TAILQ_FOREACH(oldrule, ruleset->rules[rs_num].active.ptr, entries) oldrule->nr = nr++; ruleset->rules[rs_num].active.ticket++; pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr); pf_remove_if_empty_ruleset(ruleset); PF_RULES_WUNLOCK(); break; #undef ERROUT DIOCCHANGERULE_error: PF_RULES_WUNLOCK(); if (newrule != NULL) { counter_u64_free(newrule->states_cur); counter_u64_free(newrule->states_tot); counter_u64_free(newrule->src_nodes); free(newrule, M_PFRULE); } if (kif != NULL) free(kif, PFI_MTYPE); break; } case DIOCCLRSTATES: { struct pf_state *s; struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr; u_int i, killed = 0; for (i = 0; i <= pf_hashmask; i++) { struct pf_idhash *ih = &V_pf_idhash[i]; relock_DIOCCLRSTATES: PF_HASHROW_LOCK(ih); LIST_FOREACH(s, &ih->states, entry) if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname, s->kif->pfik_name)) { /* * Don't send out individual * delete messages. */ s->state_flags |= PFSTATE_NOSYNC; pf_unlink_state(s, PF_ENTER_LOCKED); killed++; goto relock_DIOCCLRSTATES; } PF_HASHROW_UNLOCK(ih); } psk->psk_killed = killed; if (pfsync_clear_states_ptr != NULL) pfsync_clear_states_ptr(V_pf_status.hostid, psk->psk_ifname); break; } case DIOCKILLSTATES: { struct pf_state *s; struct pf_state_key *sk; struct pf_addr *srcaddr, *dstaddr; u_int16_t srcport, dstport; struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr; u_int i, killed = 0; if (psk->psk_pfcmp.id) { if (psk->psk_pfcmp.creatorid == 0) psk->psk_pfcmp.creatorid = V_pf_status.hostid; if ((s = pf_find_state_byid(psk->psk_pfcmp.id, psk->psk_pfcmp.creatorid))) { pf_unlink_state(s, PF_ENTER_LOCKED); psk->psk_killed = 1; } break; } for (i = 0; i <= pf_hashmask; i++) { struct pf_idhash *ih = &V_pf_idhash[i]; relock_DIOCKILLSTATES: PF_HASHROW_LOCK(ih); LIST_FOREACH(s, &ih->states, entry) { sk = s->key[PF_SK_WIRE]; if (s->direction == PF_OUT) { srcaddr = &sk->addr[1]; dstaddr = &sk->addr[0]; srcport = sk->port[1]; dstport = sk->port[0]; } else { srcaddr = &sk->addr[0]; dstaddr = &sk->addr[1]; srcport = sk->port[0]; dstport = sk->port[1]; } if ((!psk->psk_af || sk->af == psk->psk_af) && (!psk->psk_proto || psk->psk_proto == sk->proto) && PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr, &psk->psk_src.addr.v.a.mask, srcaddr, sk->af) && PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr, &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af) && (psk->psk_src.port_op == 0 || pf_match_port(psk->psk_src.port_op, psk->psk_src.port[0], psk->psk_src.port[1], srcport)) && (psk->psk_dst.port_op == 0 || pf_match_port(psk->psk_dst.port_op, psk->psk_dst.port[0], psk->psk_dst.port[1], dstport)) && (!psk->psk_label[0] || (s->rule.ptr->label[0] && !strcmp(psk->psk_label, s->rule.ptr->label))) && (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname, s->kif->pfik_name))) { pf_unlink_state(s, PF_ENTER_LOCKED); killed++; goto relock_DIOCKILLSTATES; } } PF_HASHROW_UNLOCK(ih); } psk->psk_killed = killed; break; } case DIOCADDSTATE: { struct pfioc_state *ps = (struct pfioc_state *)addr; struct pfsync_state *sp = &ps->state; if (sp->timeout >= PFTM_MAX) { error = EINVAL; break; } if (pfsync_state_import_ptr != NULL) { PF_RULES_RLOCK(); error = pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL); PF_RULES_RUNLOCK(); } else error = EOPNOTSUPP; break; } case DIOCGETSTATE: { struct pfioc_state *ps = (struct pfioc_state *)addr; struct pf_state *s; s = pf_find_state_byid(ps->state.id, ps->state.creatorid); if (s == NULL) { error = ENOENT; break; } pfsync_state_export(&ps->state, s); PF_STATE_UNLOCK(s); break; } case DIOCGETSTATES: { struct pfioc_states *ps = (struct pfioc_states *)addr; struct pf_state *s; struct pfsync_state *pstore, *p; int i, nr; if (ps->ps_len == 0) { nr = uma_zone_get_cur(V_pf_state_z); ps->ps_len = sizeof(struct pfsync_state) * nr; break; } p = pstore = malloc(ps->ps_len, M_TEMP, M_WAITOK); nr = 0; for (i = 0; i <= pf_hashmask; i++) { struct pf_idhash *ih = &V_pf_idhash[i]; PF_HASHROW_LOCK(ih); LIST_FOREACH(s, &ih->states, entry) { if (s->timeout == PFTM_UNLINKED) continue; if ((nr+1) * sizeof(*p) > ps->ps_len) { PF_HASHROW_UNLOCK(ih); goto DIOCGETSTATES_full; } pfsync_state_export(p, s); p++; nr++; } PF_HASHROW_UNLOCK(ih); } DIOCGETSTATES_full: error = copyout(pstore, ps->ps_states, sizeof(struct pfsync_state) * nr); if (error) { free(pstore, M_TEMP); break; } ps->ps_len = sizeof(struct pfsync_state) * nr; free(pstore, M_TEMP); break; } case DIOCGETSTATUS: { struct pf_status *s = (struct pf_status *)addr; PF_RULES_RLOCK(); s->running = V_pf_status.running; s->since = V_pf_status.since; s->debug = V_pf_status.debug; s->hostid = V_pf_status.hostid; s->states = V_pf_status.states; s->src_nodes = V_pf_status.src_nodes; for (int i = 0; i < PFRES_MAX; i++) s->counters[i] = counter_u64_fetch(V_pf_status.counters[i]); for (int i = 0; i < LCNT_MAX; i++) s->lcounters[i] = counter_u64_fetch(V_pf_status.lcounters[i]); for (int i = 0; i < FCNT_MAX; i++) s->fcounters[i] = counter_u64_fetch(V_pf_status.fcounters[i]); for (int i = 0; i < SCNT_MAX; i++) s->scounters[i] = counter_u64_fetch(V_pf_status.scounters[i]); bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ); bcopy(V_pf_status.pf_chksum, s->pf_chksum, PF_MD5_DIGEST_LENGTH); pfi_update_status(s->ifname, s); PF_RULES_RUNLOCK(); break; } case DIOCSETSTATUSIF: { struct pfioc_if *pi = (struct pfioc_if *)addr; if (pi->ifname[0] == 0) { bzero(V_pf_status.ifname, IFNAMSIZ); break; } PF_RULES_WLOCK(); strlcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ); PF_RULES_WUNLOCK(); break; } case DIOCCLRSTATUS: { PF_RULES_WLOCK(); for (int i = 0; i < PFRES_MAX; i++) counter_u64_zero(V_pf_status.counters[i]); for (int i = 0; i < FCNT_MAX; i++) counter_u64_zero(V_pf_status.fcounters[i]); for (int i = 0; i < SCNT_MAX; i++) counter_u64_zero(V_pf_status.scounters[i]); V_pf_status.since = time_second; if (*V_pf_status.ifname) pfi_update_status(V_pf_status.ifname, NULL); PF_RULES_WUNLOCK(); break; } case DIOCNATLOOK: { struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr; struct pf_state_key *sk; struct pf_state *state; struct pf_state_key_cmp key; int m = 0, direction = pnl->direction; int sidx, didx; /* NATLOOK src and dst are reversed, so reverse sidx/didx */ sidx = (direction == PF_IN) ? 1 : 0; didx = (direction == PF_IN) ? 0 : 1; if (!pnl->proto || PF_AZERO(&pnl->saddr, pnl->af) || PF_AZERO(&pnl->daddr, pnl->af) || ((pnl->proto == IPPROTO_TCP || pnl->proto == IPPROTO_UDP) && (!pnl->dport || !pnl->sport))) error = EINVAL; else { bzero(&key, sizeof(key)); key.af = pnl->af; key.proto = pnl->proto; PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af); key.port[sidx] = pnl->sport; PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af); key.port[didx] = pnl->dport; state = pf_find_state_all(&key, direction, &m); if (m > 1) error = E2BIG; /* more than one state */ else if (state != NULL) { /* XXXGL: not locked read */ sk = state->key[sidx]; PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af); pnl->rsport = sk->port[sidx]; PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af); pnl->rdport = sk->port[didx]; } else error = ENOENT; } break; } case DIOCSETTIMEOUT: { struct pfioc_tm *pt = (struct pfioc_tm *)addr; int old; if (pt->timeout < 0 || pt->timeout >= PFTM_MAX || pt->seconds < 0) { error = EINVAL; break; } PF_RULES_WLOCK(); old = V_pf_default_rule.timeout[pt->timeout]; if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0) pt->seconds = 1; V_pf_default_rule.timeout[pt->timeout] = pt->seconds; if (pt->timeout == PFTM_INTERVAL && pt->seconds < old) wakeup(pf_purge_thread); pt->seconds = old; PF_RULES_WUNLOCK(); break; } case DIOCGETTIMEOUT: { struct pfioc_tm *pt = (struct pfioc_tm *)addr; if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) { error = EINVAL; break; } PF_RULES_RLOCK(); pt->seconds = V_pf_default_rule.timeout[pt->timeout]; PF_RULES_RUNLOCK(); break; } case DIOCGETLIMIT: { struct pfioc_limit *pl = (struct pfioc_limit *)addr; if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) { error = EINVAL; break; } PF_RULES_RLOCK(); pl->limit = V_pf_limits[pl->index].limit; PF_RULES_RUNLOCK(); break; } case DIOCSETLIMIT: { struct pfioc_limit *pl = (struct pfioc_limit *)addr; int old_limit; PF_RULES_WLOCK(); if (pl->index < 0 || pl->index >= PF_LIMIT_MAX || V_pf_limits[pl->index].zone == NULL) { PF_RULES_WUNLOCK(); error = EINVAL; break; } uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit); old_limit = V_pf_limits[pl->index].limit; V_pf_limits[pl->index].limit = pl->limit; pl->limit = old_limit; PF_RULES_WUNLOCK(); break; } case DIOCSETDEBUG: { u_int32_t *level = (u_int32_t *)addr; PF_RULES_WLOCK(); V_pf_status.debug = *level; PF_RULES_WUNLOCK(); break; } case DIOCCLRRULECTRS: { /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */ struct pf_ruleset *ruleset = &pf_main_ruleset; struct pf_rule *rule; PF_RULES_WLOCK(); TAILQ_FOREACH(rule, ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) { rule->evaluations = 0; rule->packets[0] = rule->packets[1] = 0; rule->bytes[0] = rule->bytes[1] = 0; } PF_RULES_WUNLOCK(); break; } case DIOCGIFSPEED: { struct pf_ifspeed *psp = (struct pf_ifspeed *)addr; struct pf_ifspeed ps; struct ifnet *ifp; if (psp->ifname[0] != 0) { /* Can we completely trust user-land? */ strlcpy(ps.ifname, psp->ifname, IFNAMSIZ); ifp = ifunit(ps.ifname); if (ifp != NULL) psp->baudrate = ifp->if_baudrate; else error = EINVAL; } else error = EINVAL; break; } #ifdef ALTQ case DIOCSTARTALTQ: { struct pf_altq *altq; PF_RULES_WLOCK(); /* enable all altq interfaces on active list */ TAILQ_FOREACH(altq, V_pf_altqs_active, entries) { if (altq->qname[0] == 0 && (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { error = pf_enable_altq(altq); if (error != 0) break; } } if (error == 0) V_pf_altq_running = 1; PF_RULES_WUNLOCK(); DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n")); break; } case DIOCSTOPALTQ: { struct pf_altq *altq; PF_RULES_WLOCK(); /* disable all altq interfaces on active list */ TAILQ_FOREACH(altq, V_pf_altqs_active, entries) { if (altq->qname[0] == 0 && (altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) { error = pf_disable_altq(altq); if (error != 0) break; } } if (error == 0) V_pf_altq_running = 0; PF_RULES_WUNLOCK(); DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n")); break; } case DIOCADDALTQ: { struct pfioc_altq *pa = (struct pfioc_altq *)addr; struct pf_altq *altq, *a; struct ifnet *ifp; altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK); bcopy(&pa->altq, altq, sizeof(struct pf_altq)); altq->local_flags = 0; PF_RULES_WLOCK(); if (pa->ticket != V_ticket_altqs_inactive) { PF_RULES_WUNLOCK(); free(altq, M_PFALTQ); error = EBUSY; break; } /* * if this is for a queue, find the discipline and * copy the necessary fields */ if (altq->qname[0] != 0) { if ((altq->qid = pf_qname2qid(altq->qname)) == 0) { PF_RULES_WUNLOCK(); error = EBUSY; free(altq, M_PFALTQ); break; } altq->altq_disc = NULL; TAILQ_FOREACH(a, V_pf_altqs_inactive, entries) { if (strncmp(a->ifname, altq->ifname, IFNAMSIZ) == 0 && a->qname[0] == 0) { altq->altq_disc = a->altq_disc; break; } } } if ((ifp = ifunit(altq->ifname)) == NULL) altq->local_flags |= PFALTQ_FLAG_IF_REMOVED; else error = altq_add(altq); if (error) { PF_RULES_WUNLOCK(); free(altq, M_PFALTQ); break; } TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries); bcopy(altq, &pa->altq, sizeof(struct pf_altq)); PF_RULES_WUNLOCK(); break; } case DIOCGETALTQS: { struct pfioc_altq *pa = (struct pfioc_altq *)addr; struct pf_altq *altq; PF_RULES_RLOCK(); pa->nr = 0; TAILQ_FOREACH(altq, V_pf_altqs_active, entries) pa->nr++; pa->ticket = V_ticket_altqs_active; PF_RULES_RUNLOCK(); break; } case DIOCGETALTQ: { struct pfioc_altq *pa = (struct pfioc_altq *)addr; struct pf_altq *altq; u_int32_t nr; PF_RULES_RLOCK(); if (pa->ticket != V_ticket_altqs_active) { PF_RULES_RUNLOCK(); error = EBUSY; break; } nr = 0; altq = TAILQ_FIRST(V_pf_altqs_active); while ((altq != NULL) && (nr < pa->nr)) { altq = TAILQ_NEXT(altq, entries); nr++; } if (altq == NULL) { PF_RULES_RUNLOCK(); error = EBUSY; break; } bcopy(altq, &pa->altq, sizeof(struct pf_altq)); PF_RULES_RUNLOCK(); break; } case DIOCCHANGEALTQ: /* CHANGEALTQ not supported yet! */ error = ENODEV; break; case DIOCGETQSTATS: { struct pfioc_qstats *pq = (struct pfioc_qstats *)addr; struct pf_altq *altq; u_int32_t nr; int nbytes; PF_RULES_RLOCK(); if (pq->ticket != V_ticket_altqs_active) { PF_RULES_RUNLOCK(); error = EBUSY; break; } nbytes = pq->nbytes; nr = 0; altq = TAILQ_FIRST(V_pf_altqs_active); while ((altq != NULL) && (nr < pq->nr)) { altq = TAILQ_NEXT(altq, entries); nr++; } if (altq == NULL) { PF_RULES_RUNLOCK(); error = EBUSY; break; } if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) { PF_RULES_RUNLOCK(); error = ENXIO; break; } PF_RULES_RUNLOCK(); error = altq_getqstats(altq, pq->buf, &nbytes); if (error == 0) { pq->scheduler = altq->scheduler; pq->nbytes = nbytes; } break; } #endif /* ALTQ */ case DIOCBEGINADDRS: { struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; PF_RULES_WLOCK(); pf_empty_pool(&V_pf_pabuf); pp->ticket = ++V_ticket_pabuf; PF_RULES_WUNLOCK(); break; } case DIOCADDADDR: { struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; struct pf_pooladdr *pa; struct pfi_kif *kif = NULL; #ifndef INET if (pp->af == AF_INET) { error = EAFNOSUPPORT; break; } #endif /* INET */ #ifndef INET6 if (pp->af == AF_INET6) { error = EAFNOSUPPORT; break; } #endif /* INET6 */ if (pp->addr.addr.type != PF_ADDR_ADDRMASK && pp->addr.addr.type != PF_ADDR_DYNIFTL && pp->addr.addr.type != PF_ADDR_TABLE) { error = EINVAL; break; } pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK); bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr)); if (pa->ifname[0]) kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK); PF_RULES_WLOCK(); if (pp->ticket != V_ticket_pabuf) { PF_RULES_WUNLOCK(); if (pa->ifname[0]) free(kif, PFI_MTYPE); free(pa, M_PFRULE); error = EBUSY; break; } if (pa->ifname[0]) { pa->kif = pfi_kif_attach(kif, pa->ifname); pfi_kif_ref(pa->kif); } else pa->kif = NULL; if (pa->addr.type == PF_ADDR_DYNIFTL && ((error = pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) { if (pa->ifname[0]) pfi_kif_unref(pa->kif); PF_RULES_WUNLOCK(); free(pa, M_PFRULE); break; } TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries); PF_RULES_WUNLOCK(); break; } case DIOCGETADDRS: { struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; struct pf_pool *pool; struct pf_pooladdr *pa; PF_RULES_RLOCK(); pp->nr = 0; pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action, pp->r_num, 0, 1, 0); if (pool == NULL) { PF_RULES_RUNLOCK(); error = EBUSY; break; } TAILQ_FOREACH(pa, &pool->list, entries) pp->nr++; PF_RULES_RUNLOCK(); break; } case DIOCGETADDR: { struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr; struct pf_pool *pool; struct pf_pooladdr *pa; u_int32_t nr = 0; PF_RULES_RLOCK(); pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action, pp->r_num, 0, 1, 1); if (pool == NULL) { PF_RULES_RUNLOCK(); error = EBUSY; break; } pa = TAILQ_FIRST(&pool->list); while ((pa != NULL) && (nr < pp->nr)) { pa = TAILQ_NEXT(pa, entries); nr++; } if (pa == NULL) { PF_RULES_RUNLOCK(); error = EBUSY; break; } bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr)); pf_addr_copyout(&pp->addr.addr); PF_RULES_RUNLOCK(); break; } case DIOCCHANGEADDR: { struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr; struct pf_pool *pool; struct pf_pooladdr *oldpa = NULL, *newpa = NULL; struct pf_ruleset *ruleset; struct pfi_kif *kif = NULL; if (pca->action < PF_CHANGE_ADD_HEAD || pca->action > PF_CHANGE_REMOVE) { error = EINVAL; break; } if (pca->addr.addr.type != PF_ADDR_ADDRMASK && pca->addr.addr.type != PF_ADDR_DYNIFTL && pca->addr.addr.type != PF_ADDR_TABLE) { error = EINVAL; break; } if (pca->action != PF_CHANGE_REMOVE) { #ifndef INET if (pca->af == AF_INET) { error = EAFNOSUPPORT; break; } #endif /* INET */ #ifndef INET6 if (pca->af == AF_INET6) { error = EAFNOSUPPORT; break; } #endif /* INET6 */ newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK); bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr)); if (newpa->ifname[0]) kif = malloc(sizeof(*kif), PFI_MTYPE, M_WAITOK); newpa->kif = NULL; } #define ERROUT(x) { error = (x); goto DIOCCHANGEADDR_error; } PF_RULES_WLOCK(); ruleset = pf_find_ruleset(pca->anchor); if (ruleset == NULL) ERROUT(EBUSY); pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action, pca->r_num, pca->r_last, 1, 1); if (pool == NULL) ERROUT(EBUSY); if (pca->action != PF_CHANGE_REMOVE) { if (newpa->ifname[0]) { newpa->kif = pfi_kif_attach(kif, newpa->ifname); pfi_kif_ref(newpa->kif); kif = NULL; } switch (newpa->addr.type) { case PF_ADDR_DYNIFTL: error = pfi_dynaddr_setup(&newpa->addr, pca->af); break; case PF_ADDR_TABLE: newpa->addr.p.tbl = pfr_attach_table(ruleset, newpa->addr.v.tblname); if (newpa->addr.p.tbl == NULL) error = ENOMEM; break; } if (error) goto DIOCCHANGEADDR_error; } switch (pca->action) { case PF_CHANGE_ADD_HEAD: oldpa = TAILQ_FIRST(&pool->list); break; case PF_CHANGE_ADD_TAIL: oldpa = TAILQ_LAST(&pool->list, pf_palist); break; default: oldpa = TAILQ_FIRST(&pool->list); for (int i = 0; oldpa && i < pca->nr; i++) oldpa = TAILQ_NEXT(oldpa, entries); if (oldpa == NULL) ERROUT(EINVAL); } if (pca->action == PF_CHANGE_REMOVE) { TAILQ_REMOVE(&pool->list, oldpa, entries); switch (oldpa->addr.type) { case PF_ADDR_DYNIFTL: pfi_dynaddr_remove(oldpa->addr.p.dyn); break; case PF_ADDR_TABLE: pfr_detach_table(oldpa->addr.p.tbl); break; } if (oldpa->kif) pfi_kif_unref(oldpa->kif); free(oldpa, M_PFRULE); } else { if (oldpa == NULL) TAILQ_INSERT_TAIL(&pool->list, newpa, entries); else if (pca->action == PF_CHANGE_ADD_HEAD || pca->action == PF_CHANGE_ADD_BEFORE) TAILQ_INSERT_BEFORE(oldpa, newpa, entries); else TAILQ_INSERT_AFTER(&pool->list, oldpa, newpa, entries); } pool->cur = TAILQ_FIRST(&pool->list); PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af); PF_RULES_WUNLOCK(); break; #undef ERROUT DIOCCHANGEADDR_error: if (newpa->kif) pfi_kif_unref(newpa->kif); PF_RULES_WUNLOCK(); if (newpa != NULL) free(newpa, M_PFRULE); if (kif != NULL) free(kif, PFI_MTYPE); break; } case DIOCGETRULESETS: { struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; struct pf_ruleset *ruleset; struct pf_anchor *anchor; PF_RULES_RLOCK(); pr->path[sizeof(pr->path) - 1] = 0; if ((ruleset = pf_find_ruleset(pr->path)) == NULL) { PF_RULES_RUNLOCK(); error = ENOENT; break; } pr->nr = 0; if (ruleset->anchor == NULL) { /* XXX kludge for pf_main_ruleset */ RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors) if (anchor->parent == NULL) pr->nr++; } else { RB_FOREACH(anchor, pf_anchor_node, &ruleset->anchor->children) pr->nr++; } PF_RULES_RUNLOCK(); break; } case DIOCGETRULESET: { struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; struct pf_ruleset *ruleset; struct pf_anchor *anchor; u_int32_t nr = 0; PF_RULES_RLOCK(); pr->path[sizeof(pr->path) - 1] = 0; if ((ruleset = pf_find_ruleset(pr->path)) == NULL) { PF_RULES_RUNLOCK(); error = ENOENT; break; } pr->name[0] = 0; if (ruleset->anchor == NULL) { /* XXX kludge for pf_main_ruleset */ RB_FOREACH(anchor, pf_anchor_global, &V_pf_anchors) if (anchor->parent == NULL && nr++ == pr->nr) { strlcpy(pr->name, anchor->name, sizeof(pr->name)); break; } } else { RB_FOREACH(anchor, pf_anchor_node, &ruleset->anchor->children) if (nr++ == pr->nr) { strlcpy(pr->name, anchor->name, sizeof(pr->name)); break; } } if (!pr->name[0]) error = EBUSY; PF_RULES_RUNLOCK(); break; } case DIOCRCLRTABLES: { struct pfioc_table *io = (struct pfioc_table *)addr; if (io->pfrio_esize != 0) { error = ENODEV; break; } PF_RULES_WLOCK(); error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); break; } case DIOCRADDTABLES: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_table *pfrts; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_table)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_table); pfrts = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->pfrio_buffer, pfrts, totlen); if (error) { free(pfrts, M_TEMP); break; } PF_RULES_WLOCK(); error = pfr_add_tables(pfrts, io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); free(pfrts, M_TEMP); break; } case DIOCRDELTABLES: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_table *pfrts; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_table)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_table); pfrts = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->pfrio_buffer, pfrts, totlen); if (error) { free(pfrts, M_TEMP); break; } PF_RULES_WLOCK(); error = pfr_del_tables(pfrts, io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); free(pfrts, M_TEMP); break; } case DIOCRGETTABLES: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_table *pfrts; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_table)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_table); pfrts = malloc(totlen, M_TEMP, M_WAITOK); PF_RULES_RLOCK(); error = pfr_get_tables(&io->pfrio_table, pfrts, &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_RUNLOCK(); if (error == 0) error = copyout(pfrts, io->pfrio_buffer, totlen); free(pfrts, M_TEMP); break; } case DIOCRGETTSTATS: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_tstats *pfrtstats; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_tstats)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_tstats); pfrtstats = malloc(totlen, M_TEMP, M_WAITOK); PF_RULES_WLOCK(); error = pfr_get_tstats(&io->pfrio_table, pfrtstats, &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); if (error == 0) error = copyout(pfrtstats, io->pfrio_buffer, totlen); free(pfrtstats, M_TEMP); break; } case DIOCRCLRTSTATS: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_table *pfrts; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_table)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_table); pfrts = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->pfrio_buffer, pfrts, totlen); if (error) { free(pfrts, M_TEMP); break; } PF_RULES_WLOCK(); error = pfr_clr_tstats(pfrts, io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); free(pfrts, M_TEMP); break; } case DIOCRSETTFLAGS: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_table *pfrts; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_table)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_table); pfrts = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->pfrio_buffer, pfrts, totlen); if (error) { free(pfrts, M_TEMP); break; } PF_RULES_WLOCK(); error = pfr_set_tflags(pfrts, io->pfrio_size, io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange, &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); free(pfrts, M_TEMP); break; } case DIOCRCLRADDRS: { struct pfioc_table *io = (struct pfioc_table *)addr; if (io->pfrio_esize != 0) { error = ENODEV; break; } PF_RULES_WLOCK(); error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); break; } case DIOCRADDADDRS: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_addr *pfras; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); break; } PF_RULES_WLOCK(); error = pfr_add_addrs(&io->pfrio_table, pfras, io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) error = copyout(pfras, io->pfrio_buffer, totlen); free(pfras, M_TEMP); break; } case DIOCRDELADDRS: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_addr *pfras; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); break; } PF_RULES_WLOCK(); error = pfr_del_addrs(&io->pfrio_table, pfras, io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) error = copyout(pfras, io->pfrio_buffer, totlen); free(pfras, M_TEMP); break; } case DIOCRSETADDRS: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_addr *pfras; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); break; } PF_RULES_WLOCK(); error = pfr_set_addrs(&io->pfrio_table, pfras, io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd, &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags | PFR_FLAG_USERIOCTL, 0); PF_RULES_WUNLOCK(); if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) error = copyout(pfras, io->pfrio_buffer, totlen); free(pfras, M_TEMP); break; } case DIOCRGETADDRS: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_addr *pfras; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = malloc(totlen, M_TEMP, M_WAITOK); PF_RULES_RLOCK(); error = pfr_get_addrs(&io->pfrio_table, pfras, &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_RUNLOCK(); if (error == 0) error = copyout(pfras, io->pfrio_buffer, totlen); free(pfras, M_TEMP); break; } case DIOCRGETASTATS: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_astats *pfrastats; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_astats)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_astats); pfrastats = malloc(totlen, M_TEMP, M_WAITOK); PF_RULES_RLOCK(); error = pfr_get_astats(&io->pfrio_table, pfrastats, &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_RUNLOCK(); if (error == 0) error = copyout(pfrastats, io->pfrio_buffer, totlen); free(pfrastats, M_TEMP); break; } case DIOCRCLRASTATS: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_addr *pfras; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); break; } PF_RULES_WLOCK(); error = pfr_clr_astats(&io->pfrio_table, pfras, io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK) error = copyout(pfras, io->pfrio_buffer, totlen); free(pfras, M_TEMP); break; } case DIOCRTSTADDRS: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_addr *pfras; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); break; } PF_RULES_RLOCK(); error = pfr_tst_addrs(&io->pfrio_table, pfras, io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_RUNLOCK(); if (error == 0) error = copyout(pfras, io->pfrio_buffer, totlen); free(pfras, M_TEMP); break; } case DIOCRINADEFINE: { struct pfioc_table *io = (struct pfioc_table *)addr; struct pfr_addr *pfras; size_t totlen; if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; break; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); break; } PF_RULES_WLOCK(); error = pfr_ina_define(&io->pfrio_table, pfras, io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr, io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL); PF_RULES_WUNLOCK(); free(pfras, M_TEMP); break; } case DIOCOSFPADD: { struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; PF_RULES_WLOCK(); error = pf_osfp_add(io); PF_RULES_WUNLOCK(); break; } case DIOCOSFPGET: { struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr; PF_RULES_RLOCK(); error = pf_osfp_get(io); PF_RULES_RUNLOCK(); break; } case DIOCXBEGIN: { struct pfioc_trans *io = (struct pfioc_trans *)addr; struct pfioc_trans_e *ioes, *ioe; size_t totlen; int i; if (io->esize != sizeof(*ioe)) { error = ENODEV; break; } totlen = sizeof(struct pfioc_trans_e) * io->size; ioes = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->array, ioes, totlen); if (error) { free(ioes, M_TEMP); break; } PF_RULES_WLOCK(); for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { switch (ioe->rs_num) { #ifdef ALTQ case PF_RULESET_ALTQ: if (ioe->anchor[0]) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); error = EINVAL; goto fail; } if ((error = pf_begin_altq(&ioe->ticket))) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); goto fail; } break; #endif /* ALTQ */ case PF_RULESET_TABLE: { struct pfr_table table; bzero(&table, sizeof(table)); strlcpy(table.pfrt_anchor, ioe->anchor, sizeof(table.pfrt_anchor)); if ((error = pfr_ina_begin(&table, &ioe->ticket, NULL, 0))) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); goto fail; } break; } default: if ((error = pf_begin_rules(&ioe->ticket, ioe->rs_num, ioe->anchor))) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); goto fail; } break; } } PF_RULES_WUNLOCK(); error = copyout(ioes, io->array, totlen); free(ioes, M_TEMP); break; } case DIOCXROLLBACK: { struct pfioc_trans *io = (struct pfioc_trans *)addr; struct pfioc_trans_e *ioe, *ioes; size_t totlen; int i; if (io->esize != sizeof(*ioe)) { error = ENODEV; break; } totlen = sizeof(struct pfioc_trans_e) * io->size; ioes = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->array, ioes, totlen); if (error) { free(ioes, M_TEMP); break; } PF_RULES_WLOCK(); for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { switch (ioe->rs_num) { #ifdef ALTQ case PF_RULESET_ALTQ: if (ioe->anchor[0]) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); error = EINVAL; goto fail; } if ((error = pf_rollback_altq(ioe->ticket))) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); goto fail; /* really bad */ } break; #endif /* ALTQ */ case PF_RULESET_TABLE: { struct pfr_table table; bzero(&table, sizeof(table)); strlcpy(table.pfrt_anchor, ioe->anchor, sizeof(table.pfrt_anchor)); if ((error = pfr_ina_rollback(&table, ioe->ticket, NULL, 0))) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); goto fail; /* really bad */ } break; } default: if ((error = pf_rollback_rules(ioe->ticket, ioe->rs_num, ioe->anchor))) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); goto fail; /* really bad */ } break; } } PF_RULES_WUNLOCK(); free(ioes, M_TEMP); break; } case DIOCXCOMMIT: { struct pfioc_trans *io = (struct pfioc_trans *)addr; struct pfioc_trans_e *ioe, *ioes; struct pf_ruleset *rs; size_t totlen; int i; if (io->esize != sizeof(*ioe)) { error = ENODEV; break; } totlen = sizeof(struct pfioc_trans_e) * io->size; ioes = malloc(totlen, M_TEMP, M_WAITOK); error = copyin(io->array, ioes, totlen); if (error) { free(ioes, M_TEMP); break; } PF_RULES_WLOCK(); /* First makes sure everything will succeed. */ for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { switch (ioe->rs_num) { #ifdef ALTQ case PF_RULESET_ALTQ: if (ioe->anchor[0]) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); error = EINVAL; goto fail; } if (!V_altqs_inactive_open || ioe->ticket != V_ticket_altqs_inactive) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); error = EBUSY; goto fail; } break; #endif /* ALTQ */ case PF_RULESET_TABLE: rs = pf_find_ruleset(ioe->anchor); if (rs == NULL || !rs->topen || ioe->ticket != rs->tticket) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); error = EBUSY; goto fail; } break; default: if (ioe->rs_num < 0 || ioe->rs_num >= PF_RULESET_MAX) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); error = EINVAL; goto fail; } rs = pf_find_ruleset(ioe->anchor); if (rs == NULL || !rs->rules[ioe->rs_num].inactive.open || rs->rules[ioe->rs_num].inactive.ticket != ioe->ticket) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); error = EBUSY; goto fail; } break; } } /* Now do the commit - no errors should happen here. */ for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { switch (ioe->rs_num) { #ifdef ALTQ case PF_RULESET_ALTQ: if ((error = pf_commit_altq(ioe->ticket))) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); goto fail; /* really bad */ } break; #endif /* ALTQ */ case PF_RULESET_TABLE: { struct pfr_table table; bzero(&table, sizeof(table)); strlcpy(table.pfrt_anchor, ioe->anchor, sizeof(table.pfrt_anchor)); if ((error = pfr_ina_commit(&table, ioe->ticket, NULL, NULL, 0))) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); goto fail; /* really bad */ } break; } default: if ((error = pf_commit_rules(ioe->ticket, ioe->rs_num, ioe->anchor))) { PF_RULES_WUNLOCK(); free(ioes, M_TEMP); goto fail; /* really bad */ } break; } } PF_RULES_WUNLOCK(); free(ioes, M_TEMP); break; } case DIOCGETSRCNODES: { struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr; struct pf_srchash *sh; struct pf_src_node *n, *p, *pstore; uint32_t i, nr = 0; if (psn->psn_len == 0) { for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask; i++, sh++) { PF_HASHROW_LOCK(sh); LIST_FOREACH(n, &sh->nodes, entry) nr++; PF_HASHROW_UNLOCK(sh); } psn->psn_len = sizeof(struct pf_src_node) * nr; break; } p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK); for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask; i++, sh++) { PF_HASHROW_LOCK(sh); LIST_FOREACH(n, &sh->nodes, entry) { int secs = time_uptime, diff; if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len) break; bcopy(n, p, sizeof(struct pf_src_node)); if (n->rule.ptr != NULL) p->rule.nr = n->rule.ptr->nr; p->creation = secs - p->creation; if (p->expire > secs) p->expire -= secs; else p->expire = 0; /* Adjust the connection rate estimate. */ diff = secs - n->conn_rate.last; if (diff >= n->conn_rate.seconds) p->conn_rate.count = 0; else p->conn_rate.count -= n->conn_rate.count * diff / n->conn_rate.seconds; p++; nr++; } PF_HASHROW_UNLOCK(sh); } error = copyout(pstore, psn->psn_src_nodes, sizeof(struct pf_src_node) * nr); if (error) { free(pstore, M_TEMP); break; } psn->psn_len = sizeof(struct pf_src_node) * nr; free(pstore, M_TEMP); break; } case DIOCCLRSRCNODES: { pf_clear_srcnodes(NULL); pf_purge_expired_src_nodes(); break; } case DIOCKILLSRCNODES: pf_kill_srcnodes((struct pfioc_src_node_kill *)addr); break; case DIOCSETHOSTID: { u_int32_t *hostid = (u_int32_t *)addr; PF_RULES_WLOCK(); if (*hostid == 0) V_pf_status.hostid = arc4random(); else V_pf_status.hostid = *hostid; PF_RULES_WUNLOCK(); break; } case DIOCOSFPFLUSH: PF_RULES_WLOCK(); pf_osfp_flush(); PF_RULES_WUNLOCK(); break; case DIOCIGETIFACES: { struct pfioc_iface *io = (struct pfioc_iface *)addr; struct pfi_kif *ifstore; size_t bufsiz; if (io->pfiio_esize != sizeof(struct pfi_kif)) { error = ENODEV; break; } bufsiz = io->pfiio_size * sizeof(struct pfi_kif); ifstore = malloc(bufsiz, M_TEMP, M_WAITOK); PF_RULES_RLOCK(); pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size); PF_RULES_RUNLOCK(); error = copyout(ifstore, io->pfiio_buffer, bufsiz); free(ifstore, M_TEMP); break; } case DIOCSETIFFLAG: { struct pfioc_iface *io = (struct pfioc_iface *)addr; PF_RULES_WLOCK(); error = pfi_set_flags(io->pfiio_name, io->pfiio_flags); PF_RULES_WUNLOCK(); break; } case DIOCCLRIFFLAG: { struct pfioc_iface *io = (struct pfioc_iface *)addr; PF_RULES_WLOCK(); error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags); PF_RULES_WUNLOCK(); break; } default: error = ENODEV; break; } fail: if (sx_xlocked(&pf_ioctl_lock)) sx_xunlock(&pf_ioctl_lock); CURVNET_RESTORE(); return (error); } void pfsync_state_export(struct pfsync_state *sp, struct pf_state *st) { bzero(sp, sizeof(struct pfsync_state)); /* copy from state key */ sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0]; sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1]; sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0]; sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1]; sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0]; sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1]; sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0]; sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1]; sp->proto = st->key[PF_SK_WIRE]->proto; sp->af = st->key[PF_SK_WIRE]->af; /* copy from state */ strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname)); bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr)); sp->creation = htonl(time_uptime - st->creation); sp->expire = pf_state_expires(st); if (sp->expire <= time_uptime) sp->expire = htonl(0); else sp->expire = htonl(sp->expire - time_uptime); sp->direction = st->direction; sp->log = st->log; sp->timeout = st->timeout; sp->state_flags = st->state_flags; if (st->src_node) sp->sync_flags |= PFSYNC_FLAG_SRCNODE; if (st->nat_src_node) sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE; sp->id = st->id; sp->creatorid = st->creatorid; pf_state_peer_hton(&st->src, &sp->src); pf_state_peer_hton(&st->dst, &sp->dst); if (st->rule.ptr == NULL) sp->rule = htonl(-1); else sp->rule = htonl(st->rule.ptr->nr); if (st->anchor.ptr == NULL) sp->anchor = htonl(-1); else sp->anchor = htonl(st->anchor.ptr->nr); if (st->nat_rule.ptr == NULL) sp->nat_rule = htonl(-1); else sp->nat_rule = htonl(st->nat_rule.ptr->nr); pf_state_counter_hton(st->packets[0], sp->packets[0]); pf_state_counter_hton(st->packets[1], sp->packets[1]); pf_state_counter_hton(st->bytes[0], sp->bytes[0]); pf_state_counter_hton(st->bytes[1], sp->bytes[1]); } static void pf_tbladdr_copyout(struct pf_addr_wrap *aw) { struct pfr_ktable *kt; KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type)); kt = aw->p.tbl; if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) kt = kt->pfrkt_root; aw->p.tbl = NULL; aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ? kt->pfrkt_cnt : -1; } /* * XXX - Check for version missmatch!!! */ static void pf_clear_states(void) { struct pf_state *s; u_int i; for (i = 0; i <= pf_hashmask; i++) { struct pf_idhash *ih = &V_pf_idhash[i]; relock: PF_HASHROW_LOCK(ih); LIST_FOREACH(s, &ih->states, entry) { s->timeout = PFTM_PURGE; /* Don't send out individual delete messages. */ s->sync_state = PFSTATE_NOSYNC; pf_unlink_state(s, PF_ENTER_LOCKED); goto relock; } PF_HASHROW_UNLOCK(ih); } } static int pf_clear_tables(void) { struct pfioc_table io; int error; bzero(&io, sizeof(io)); error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel, io.pfrio_flags); return (error); } static void pf_clear_srcnodes(struct pf_src_node *n) { struct pf_state *s; int i; for (i = 0; i <= pf_hashmask; i++) { struct pf_idhash *ih = &V_pf_idhash[i]; PF_HASHROW_LOCK(ih); LIST_FOREACH(s, &ih->states, entry) { if (n == NULL || n == s->src_node) s->src_node = NULL; if (n == NULL || n == s->nat_src_node) s->nat_src_node = NULL; } PF_HASHROW_UNLOCK(ih); } if (n == NULL) { struct pf_srchash *sh; for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask; i++, sh++) { PF_HASHROW_LOCK(sh); LIST_FOREACH(n, &sh->nodes, entry) { n->expire = 1; n->states = 0; } PF_HASHROW_UNLOCK(sh); } } else { /* XXX: hash slot should already be locked here. */ n->expire = 1; n->states = 0; } } static void pf_kill_srcnodes(struct pfioc_src_node_kill *psnk) { struct pf_src_node_list kill; LIST_INIT(&kill); for (int i = 0; i <= pf_srchashmask; i++) { struct pf_srchash *sh = &V_pf_srchash[i]; struct pf_src_node *sn, *tmp; PF_HASHROW_LOCK(sh); LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp) if (PF_MATCHA(psnk->psnk_src.neg, &psnk->psnk_src.addr.v.a.addr, &psnk->psnk_src.addr.v.a.mask, &sn->addr, sn->af) && PF_MATCHA(psnk->psnk_dst.neg, &psnk->psnk_dst.addr.v.a.addr, &psnk->psnk_dst.addr.v.a.mask, &sn->raddr, sn->af)) { pf_unlink_src_node(sn); LIST_INSERT_HEAD(&kill, sn, entry); sn->expire = 1; } PF_HASHROW_UNLOCK(sh); } for (int i = 0; i <= pf_hashmask; i++) { struct pf_idhash *ih = &V_pf_idhash[i]; struct pf_state *s; PF_HASHROW_LOCK(ih); LIST_FOREACH(s, &ih->states, entry) { if (s->src_node && s->src_node->expire == 1) s->src_node = NULL; if (s->nat_src_node && s->nat_src_node->expire == 1) s->nat_src_node = NULL; } PF_HASHROW_UNLOCK(ih); } psnk->psnk_killed = pf_free_src_nodes(&kill); } /* * XXX - Check for version missmatch!!! */ /* * Duplicate pfctl -Fa operation to get rid of as much as we can. */ static int shutdown_pf(void) { int error = 0; u_int32_t t[5]; char nn = '\0'; V_pf_status.running = 0; counter_u64_free(V_pf_default_rule.states_cur); counter_u64_free(V_pf_default_rule.states_tot); counter_u64_free(V_pf_default_rule.src_nodes); for (int i = 0; i < PFRES_MAX; i++) counter_u64_free(V_pf_status.counters[i]); for (int i = 0; i < LCNT_MAX; i++) counter_u64_free(V_pf_status.lcounters[i]); for (int i = 0; i < FCNT_MAX; i++) counter_u64_free(V_pf_status.fcounters[i]); for (int i = 0; i < SCNT_MAX; i++) counter_u64_free(V_pf_status.scounters[i]); do { if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) { DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n")); break; } if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) { DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n")); break; /* XXX: rollback? */ } if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn)) != 0) { DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n")); break; /* XXX: rollback? */ } if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn)) != 0) { DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n")); break; /* XXX: rollback? */ } if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn)) != 0) { DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n")); break; /* XXX: rollback? */ } /* XXX: these should always succeed here */ pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn); pf_commit_rules(t[1], PF_RULESET_FILTER, &nn); pf_commit_rules(t[2], PF_RULESET_NAT, &nn); pf_commit_rules(t[3], PF_RULESET_BINAT, &nn); pf_commit_rules(t[4], PF_RULESET_RDR, &nn); if ((error = pf_clear_tables()) != 0) break; #ifdef ALTQ if ((error = pf_begin_altq(&t[0])) != 0) { DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n")); break; } pf_commit_altq(t[0]); #endif pf_clear_states(); pf_clear_srcnodes(NULL); /* status does not use malloced mem so no need to cleanup */ /* fingerprints and interfaces have thier own cleanup code */ } while(0); return (error); } #ifdef INET static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, struct inpcb *inp) { int chk; chk = pf_test(PF_IN, ifp, m, inp); if (chk && *m) { m_freem(*m); *m = NULL; } return (chk); } static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, struct inpcb *inp) { int chk; chk = pf_test(PF_OUT, ifp, m, inp); if (chk && *m) { m_freem(*m); *m = NULL; } return (chk); } #endif #ifdef INET6 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, struct inpcb *inp) { int chk; /* * In case of loopback traffic IPv6 uses the real interface in * order to support scoped addresses. In order to support stateful * filtering we have change this to lo0 as it is the case in IPv4. */ CURVNET_SET(ifp->if_vnet); chk = pf_test6(PF_IN, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp); CURVNET_RESTORE(); if (chk && *m) { m_freem(*m); *m = NULL; } return chk; } static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir, struct inpcb *inp) { int chk; CURVNET_SET(ifp->if_vnet); chk = pf_test6(PF_OUT, ifp, m, inp); CURVNET_RESTORE(); if (chk && *m) { m_freem(*m); *m = NULL; } return chk; } #endif /* INET6 */ static int hook_pf(void) { #ifdef INET struct pfil_head *pfh_inet; #endif #ifdef INET6 struct pfil_head *pfh_inet6; #endif if (V_pf_pfil_hooked) return (0); #ifdef INET pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET); if (pfh_inet == NULL) return (ESRCH); /* XXX */ pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet); pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet); #endif #ifdef INET6 pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6); if (pfh_inet6 == NULL) { #ifdef INET pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet); pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet); #endif return (ESRCH); /* XXX */ } pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6); pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6); #endif V_pf_pfil_hooked = 1; return (0); } static int dehook_pf(void) { #ifdef INET struct pfil_head *pfh_inet; #endif #ifdef INET6 struct pfil_head *pfh_inet6; #endif if (V_pf_pfil_hooked == 0) return (0); #ifdef INET pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET); if (pfh_inet == NULL) return (ESRCH); /* XXX */ pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet); pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet); #endif #ifdef INET6 pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6); if (pfh_inet6 == NULL) return (ESRCH); /* XXX */ pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6); pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6); #endif V_pf_pfil_hooked = 0; return (0); } static int pf_load(void) { int error; VNET_ITERATOR_DECL(vnet_iter); VNET_LIST_RLOCK(); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); V_pf_pfil_hooked = 0; V_pf_end_threads = 0; TAILQ_INIT(&V_pf_tags); TAILQ_INIT(&V_pf_qids); CURVNET_RESTORE(); } VNET_LIST_RUNLOCK(); rw_init(&pf_rules_lock, "pf rulesets"); sx_init(&pf_ioctl_lock, "pf ioctl"); pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME); if ((error = pfattach()) != 0) return (error); return (0); } static int pf_unload(void) { int error = 0; V_pf_status.running = 0; swi_remove(V_pf_swi_cookie); error = dehook_pf(); if (error) { /* * Should not happen! * XXX Due to error code ESRCH, kldunload will show * a message like 'No such process'. */ printf("%s : pfil unregisteration fail\n", __FUNCTION__); return error; } PF_RULES_WLOCK(); shutdown_pf(); V_pf_end_threads = 1; while (V_pf_end_threads < 2) { wakeup_one(pf_purge_thread); rw_sleep(pf_purge_thread, &pf_rules_lock, 0, "pftmo", 0); } PF_RULES_WUNLOCK(); pf_normalize_cleanup(); pfi_cleanup(); pfr_cleanup(); pf_osfp_flush(); pf_cleanup(); if (IS_DEFAULT_VNET(curvnet)) pf_mtag_cleanup(); destroy_dev(pf_dev); rw_destroy(&pf_rules_lock); sx_destroy(&pf_ioctl_lock); return (error); } static int pf_modevent(module_t mod, int type, void *data) { int error = 0; switch(type) { case MOD_LOAD: error = pf_load(); break; case MOD_QUIESCE: /* * Module should not be unloaded due to race conditions. */ error = EBUSY; break; case MOD_UNLOAD: error = pf_unload(); break; default: error = EINVAL; break; } return (error); } static moduledata_t pf_mod = { "pf", pf_modevent, 0 }; -DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST); +DECLARE_MODULE(pf, pf_mod, SI_SUB_FW, SI_ORDER_FIRST); MODULE_VERSION(pf, PF_MODVER); Index: projects/vnet/sys/sys/kernel.h =================================================================== --- projects/vnet/sys/sys/kernel.h (revision 294565) +++ projects/vnet/sys/sys/kernel.h (revision 294566) @@ -1,366 +1,369 @@ /*- * Copyright (c) 1995 Terrence R. Lambert * All rights reserved. * * Copyright (c) 1990, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)kernel.h 8.3 (Berkeley) 1/21/94 * $FreeBSD$ */ #ifndef _SYS_KERNEL_H_ #define _SYS_KERNEL_H_ #include #ifdef _KERNEL /* for intrhook below */ #include /* Global variables for the kernel. */ /* 1.1 */ extern char kernelname[MAXPATHLEN]; extern int tick; /* usec per tick (1000000 / hz) */ extern int hz; /* system clock's frequency */ extern int psratio; /* ratio: prof / stat */ extern int stathz; /* statistics clock's frequency */ extern int profhz; /* profiling clock's frequency */ extern int profprocs; /* number of process's profiling */ extern volatile int ticks; #endif /* _KERNEL */ /* * Enumerated types for known system startup interfaces. * * Startup occurs in ascending numeric order; the list entries are * sorted prior to attempting startup to guarantee order. Items * of the same level are arbitrated for order based on the 'order' * element. * * These numbers are arbitrary and are chosen ONLY for ordering; the * enumeration values are explicit rather than implicit to provide * for binary compatibility with inserted elements. * * The SI_SUB_LAST value must have the highest lexical value. */ enum sysinit_sub_id { SI_SUB_DUMMY = 0x0000000, /* not executed; for linker*/ SI_SUB_DONE = 0x0000001, /* processed*/ SI_SUB_TUNABLES = 0x0700000, /* establish tunable values */ SI_SUB_COPYRIGHT = 0x0800001, /* first use of console*/ SI_SUB_VM = 0x1000000, /* virtual memory system init*/ SI_SUB_KMEM = 0x1800000, /* kernel memory*/ SI_SUB_HYPERVISOR = 0x1A40000, /* * Hypervisor detection and * virtualization support * setup. */ SI_SUB_WITNESS = 0x1A80000, /* witness initialization */ SI_SUB_MTX_POOL_DYNAMIC = 0x1AC0000, /* dynamic mutex pool */ SI_SUB_LOCK = 0x1B00000, /* various locks */ SI_SUB_EVENTHANDLER = 0x1C00000, /* eventhandler init */ SI_SUB_VNET_PRELINK = 0x1E00000, /* vnet init before modules */ SI_SUB_KLD = 0x2000000, /* KLD and module setup */ SI_SUB_CPU = 0x2100000, /* CPU resource(s)*/ SI_SUB_RACCT = 0x2110000, /* resource accounting */ SI_SUB_KDTRACE = 0x2140000, /* Kernel dtrace hooks */ SI_SUB_RANDOM = 0x2160000, /* random number generator */ SI_SUB_MAC = 0x2180000, /* TrustedBSD MAC subsystem */ SI_SUB_MAC_POLICY = 0x21C0000, /* TrustedBSD MAC policies */ SI_SUB_MAC_LATE = 0x21D0000, /* TrustedBSD MAC subsystem */ SI_SUB_VNET = 0x21E0000, /* vnet 0 */ SI_SUB_INTRINSIC = 0x2200000, /* proc 0*/ SI_SUB_VM_CONF = 0x2300000, /* config VM, set limits*/ SI_SUB_DDB_SERVICES = 0x2380000, /* capture, scripting, etc. */ SI_SUB_RUN_QUEUE = 0x2400000, /* set up run queue*/ SI_SUB_KTRACE = 0x2480000, /* ktrace */ SI_SUB_OPENSOLARIS = 0x2490000, /* OpenSolaris compatibility */ SI_SUB_AUDIT = 0x24C0000, /* audit */ SI_SUB_CREATE_INIT = 0x2500000, /* create init process*/ SI_SUB_SCHED_IDLE = 0x2600000, /* required idle procs */ SI_SUB_MBUF = 0x2700000, /* mbuf subsystem */ SI_SUB_INTR = 0x2800000, /* interrupt threads */ SI_SUB_SOFTINTR = 0x2800001, /* start soft interrupt thread */ SI_SUB_DEVFS = 0x2F00000, /* devfs ready for devices */ SI_SUB_INIT_IF = 0x3000000, /* prep for net interfaces */ SI_SUB_NETGRAPH = 0x3010000, /* Let Netgraph initialize */ SI_SUB_DTRACE = 0x3020000, /* DTrace subsystem */ SI_SUB_DTRACE_PROVIDER = 0x3048000, /* DTrace providers */ SI_SUB_DTRACE_ANON = 0x308C000, /* DTrace anon enabling */ SI_SUB_DRIVERS = 0x3100000, /* Let Drivers initialize */ SI_SUB_CONFIGURE = 0x3800000, /* Configure devices */ SI_SUB_VFS = 0x4000000, /* virtual filesystem*/ SI_SUB_CLOCKS = 0x4800000, /* real time and stat clocks*/ SI_SUB_SYSV_SHM = 0x6400000, /* System V shared memory*/ SI_SUB_SYSV_SEM = 0x6800000, /* System V semaphores*/ SI_SUB_SYSV_MSG = 0x6C00000, /* System V message queues*/ SI_SUB_P1003_1B = 0x6E00000, /* P1003.1B realtime */ SI_SUB_PSEUDO = 0x7000000, /* pseudo devices*/ SI_SUB_EXEC = 0x7400000, /* execve() handlers */ SI_SUB_PROTO_BEGIN = 0x8000000, /* VNET initialization */ + SI_SUB_PROTO_PFIL = 0x8100000, /* Intialize pfil before FWs */ SI_SUB_PROTO_IF = 0x8400000, /* interfaces*/ SI_SUB_PROTO_DOMAININIT = 0x8600000, /* domain registration system */ + SI_SUB_PROTO_MC = 0x8700000, /* Multicast */ SI_SUB_PROTO_DOMAIN = 0x8800000, /* domains (address families?)*/ - SI_SUB_PROTO_IFATTACHDOMAIN = 0x8800001, /* domain dependent data init*/ + SI_SUB_FW = 0x8806000, /* Firewalls */ + SI_SUB_PROTO_IFATTACHDOMAIN = 0x8808000,/* domain dependent data init */ SI_SUB_PROTO_END = 0x8ffffff, /* VNET helper functions */ SI_SUB_KPROF = 0x9000000, /* kernel profiling*/ SI_SUB_KICK_SCHEDULER = 0xa000000, /* start the timeout events*/ SI_SUB_INT_CONFIG_HOOKS = 0xa800000, /* Interrupts enabled config */ SI_SUB_ROOT_CONF = 0xb000000, /* Find root devices */ SI_SUB_INTRINSIC_POST = 0xd000000, /* proc 0 cleanup*/ SI_SUB_SYSCALLS = 0xd800000, /* register system calls */ SI_SUB_VNET_DONE = 0xdc00000, /* vnet registration complete */ SI_SUB_KTHREAD_INIT = 0xe000000, /* init process*/ SI_SUB_KTHREAD_PAGE = 0xe400000, /* pageout daemon*/ SI_SUB_KTHREAD_VM = 0xe800000, /* vm daemon*/ SI_SUB_KTHREAD_BUF = 0xea00000, /* buffer daemon*/ SI_SUB_KTHREAD_UPDATE = 0xec00000, /* update daemon*/ SI_SUB_KTHREAD_IDLE = 0xee00000, /* idle procs*/ SI_SUB_SMP = 0xf000000, /* start the APs*/ SI_SUB_RACCTD = 0xf100000, /* start racctd*/ SI_SUB_LAST = 0xfffffff /* final initialization */ }; /* * Some enumerated orders; "ANY" sorts last. */ enum sysinit_elem_order { SI_ORDER_FIRST = 0x0000000, /* first*/ SI_ORDER_SECOND = 0x0000001, /* second*/ SI_ORDER_THIRD = 0x0000002, /* third*/ SI_ORDER_FOURTH = 0x0000003, /* fourth*/ SI_ORDER_MIDDLE = 0x1000000, /* somewhere in the middle */ SI_ORDER_ANY = 0xfffffff /* last*/ }; /* * A system initialization call instance * * At the moment there is one instance of sysinit. We probably do not * want two which is why this code is if'd out, but we definitely want * to discern SYSINIT's which take non-constant data pointers and * SYSINIT's which take constant data pointers, * * The C_* macros take functions expecting const void * arguments * while the non-C_* macros take functions expecting just void * arguments. * * With -Wcast-qual on, the compiler issues warnings: * - if we pass non-const data or functions taking non-const data * to a C_* macro. * * - if we pass const data to the normal macros * * However, no warning is issued if we pass a function taking const data * through a normal non-const macro. This is ok because the function is * saying it won't modify the data so we don't care whether the data is * modifiable or not. */ typedef void (*sysinit_nfunc_t)(void *); typedef void (*sysinit_cfunc_t)(const void *); struct sysinit { enum sysinit_sub_id subsystem; /* subsystem identifier*/ enum sysinit_elem_order order; /* init order within subsystem*/ sysinit_cfunc_t func; /* function */ const void *udata; /* multiplexer/argument */ }; /* * Default: no special processing * * The C_ version of SYSINIT is for data pointers to const * data ( and functions taking data pointers to const data ). * At the moment it is no different from SYSINIT and thus * still results in warnings. * * The casts are necessary to have the compiler produce the * correct warnings when -Wcast-qual is used. * */ #define C_SYSINIT(uniquifier, subsystem, order, func, ident) \ static struct sysinit uniquifier ## _sys_init = { \ subsystem, \ order, \ func, \ (ident) \ }; \ DATA_SET(sysinit_set,uniquifier ## _sys_init) #define SYSINIT(uniquifier, subsystem, order, func, ident) \ C_SYSINIT(uniquifier, subsystem, order, \ (sysinit_cfunc_t)(sysinit_nfunc_t)func, (void *)(ident)) /* * Called on module unload: no special processing */ #define C_SYSUNINIT(uniquifier, subsystem, order, func, ident) \ static struct sysinit uniquifier ## _sys_uninit = { \ subsystem, \ order, \ func, \ (ident) \ }; \ DATA_SET(sysuninit_set,uniquifier ## _sys_uninit) #define SYSUNINIT(uniquifier, subsystem, order, func, ident) \ C_SYSUNINIT(uniquifier, subsystem, order, \ (sysinit_cfunc_t)(sysinit_nfunc_t)func, (void *)(ident)) void sysinit_add(struct sysinit **set, struct sysinit **set_end); /* * Infrastructure for tunable 'constants'. Value may be specified at compile * time or kernel load time. Rules relating tunables together can be placed * in a SYSINIT function at SI_SUB_TUNABLES with SI_ORDER_ANY. * * WARNING: developers should never use the reserved suffixes specified in * loader.conf(5) for any tunables or conflicts will result. */ /* * int * please avoid using for new tunables! */ extern void tunable_int_init(void *); struct tunable_int { const char *path; int *var; }; #define TUNABLE_INT(path, var) \ static struct tunable_int __CONCAT(__tunable_int_, __LINE__) = { \ (path), \ (var), \ }; \ SYSINIT(__CONCAT(__Tunable_init_, __LINE__), \ SI_SUB_TUNABLES, SI_ORDER_MIDDLE, tunable_int_init, \ &__CONCAT(__tunable_int_, __LINE__)) #define TUNABLE_INT_FETCH(path, var) getenv_int((path), (var)) /* * long */ extern void tunable_long_init(void *); struct tunable_long { const char *path; long *var; }; #define TUNABLE_LONG(path, var) \ static struct tunable_long __CONCAT(__tunable_long_, __LINE__) = { \ (path), \ (var), \ }; \ SYSINIT(__CONCAT(__Tunable_init_, __LINE__), \ SI_SUB_TUNABLES, SI_ORDER_MIDDLE, tunable_long_init,\ &__CONCAT(__tunable_long_, __LINE__)) #define TUNABLE_LONG_FETCH(path, var) getenv_long((path), (var)) /* * unsigned long */ extern void tunable_ulong_init(void *); struct tunable_ulong { const char *path; unsigned long *var; }; #define TUNABLE_ULONG(path, var) \ static struct tunable_ulong __CONCAT(__tunable_ulong_, __LINE__) = { \ (path), \ (var), \ }; \ SYSINIT(__CONCAT(__Tunable_init_, __LINE__), \ SI_SUB_TUNABLES, SI_ORDER_MIDDLE, tunable_ulong_init, \ &__CONCAT(__tunable_ulong_, __LINE__)) #define TUNABLE_ULONG_FETCH(path, var) getenv_ulong((path), (var)) /* * quad */ extern void tunable_quad_init(void *); struct tunable_quad { const char *path; quad_t *var; }; #define TUNABLE_QUAD(path, var) \ static struct tunable_quad __CONCAT(__tunable_quad_, __LINE__) = { \ (path), \ (var), \ }; \ SYSINIT(__CONCAT(__Tunable_init_, __LINE__), \ SI_SUB_TUNABLES, SI_ORDER_MIDDLE, tunable_quad_init, \ &__CONCAT(__tunable_quad_, __LINE__)) #define TUNABLE_QUAD_FETCH(path, var) getenv_quad((path), (var)) extern void tunable_str_init(void *); struct tunable_str { const char *path; char *var; int size; }; #define TUNABLE_STR(path, var, size) \ static struct tunable_str __CONCAT(__tunable_str_, __LINE__) = { \ (path), \ (var), \ (size), \ }; \ SYSINIT(__CONCAT(__Tunable_init_, __LINE__), \ SI_SUB_TUNABLES, SI_ORDER_MIDDLE, tunable_str_init, \ &__CONCAT(__tunable_str_, __LINE__)) #define TUNABLE_STR_FETCH(path, var, size) \ getenv_string((path), (var), (size)) struct intr_config_hook { TAILQ_ENTRY(intr_config_hook) ich_links; void (*ich_func)(void *arg); void *ich_arg; }; int config_intrhook_establish(struct intr_config_hook *hook); void config_intrhook_disestablish(struct intr_config_hook *hook); #endif /* !_SYS_KERNEL_H_*/