Index: head/sys/net/if.c =================================================================== --- head/sys/net/if.c (revision 271769) +++ head/sys/net/if.c (revision 271770) @@ -1,4122 +1,4129 @@ /*- * 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 struct ifindex_entry { struct ifnet *ife_ifnet; }; 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_init(void *); static void if_grow(void); 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 if_rtdel(struct radix_node *, void *); 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); static void if_detach_internal(struct ifnet *, int); #ifdef INET6 /* * XXX: declare here to avoid to include many inet6 related files.. * should be more generalized? */ extern void nd6_setmtu(struct ifnet *); #endif 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 ifindex_entry *, 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; struct sx ifnet_sxlock; /* * 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].ife_ifnet == IFNET_HOLD) return (NULL); return (V_ifindex_table[idx].ife_ifnet); } 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 int ifindex_alloc_locked(u_short *idxp) { 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].ife_ifnet == NULL) break; } /* Catch if_index overflow. */ if (idx >= V_if_indexlim) { if_grow(); goto retry; } if (idx > V_if_index) V_if_index = idx; *idxp = idx; return (0); } static void ifindex_free_locked(u_short idx) { IFNET_WLOCK_ASSERT(); V_ifindex_table[idx].ife_ifnet = NULL; while (V_if_index > 0 && V_ifindex_table[V_if_index].ife_ifnet == 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].ife_ifnet = 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 ifaddr *ifa; IFNET_RLOCK_NOSLEEP(); ifa = ifnet_byindex_locked(idx)->if_addr; if (ifa != 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); /* ARGSUSED*/ static void if_init(void *dummy __unused) { IFNET_LOCK_INIT(); if_clone_init(); } SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, 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); #endif static void if_grow(void) { int oldlim; u_int n; struct ifindex_entry *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(); if (ifindex_alloc_locked(&idx) != 0) { IFNET_WUNLOCK(); free(ifp, M_IFNET); return (NULL); } 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. */ 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); 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. * * 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); } static void if_attach_internal(struct ifnet *ifp, int vmove) { 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); 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_get_counter == NULL) ifp->if_get_counter = if_get_counter_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) /* Initialize to max value. */ if (ifp->if_hw_tsomax == 0) ifp->if_hw_tsomax = min(IP_MAXPACKET, 32 * MCLBYTES - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); KASSERT(ifp->if_hw_tsomax <= IP_MAXPACKET && ifp->if_hw_tsomax >= IP_MAXPACKET / 8, ("%s: tsomax outside of range", __func__)); #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; 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 */ TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 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); CURVNET_RESTORE(); } static void if_detach_internal(struct ifnet *ifp, int vmove) { struct ifaddr *ifa; struct radix_node_head *rnh; int i, j; 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) { if (vmove) panic("%s: ifp=%p not on the ifnet tailq %p", __func__, ifp, &V_ifnet); else return; /* XXX this should panic as well? */ } /* * 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); TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); ifa_free(ifa); } } /* * Delete all remaining routes using this interface * Unfortuneatly the only way to do this is to slog through * the entire routing table looking for routes which point * to this interface...oh well... */ for (i = 1; i <= AF_MAX; i++) { for (j = 0; j < rt_numfibs; j++) { rnh = rt_tables_get_rnh(j, i); if (rnh == NULL) continue; RADIX_NODE_HEAD_LOCK(rnh); (void) rnh->rnh_walktree(rnh, if_rtdel, ifp); RADIX_NODE_HEAD_UNLOCK(rnh); } } 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]); } } #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) { u_short idx; /* * Detach from current vnet, but preserve LLADDR info, do not * mark as dead etc. so that the ifnet can be reattached later. */ if_detach_internal(ifp, 1); /* * 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(); if (ifindex_alloc_locked(&idx) != 0) { IFNET_WUNLOCK(); panic("if_index overflow"); } ifp->if_index = idx; ifnet_setbyindex_locked(ifp->if_index, ifp); IFNET_WUNLOCK(); if_attach_internal(ifp, 1); 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); 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); } /* * Delete Routes for a Network Interface * * Called for each routing entry via the rnh->rnh_walktree() call above * to delete all route entries referencing a detaching network interface. * * Arguments: * rn pointer to node in the routing table * arg argument passed to rnh->rnh_walktree() - detaching interface * * Returns: * 0 successful * errno failed - reason indicated * */ static int if_rtdel(struct radix_node *rn, void *arg) { struct rtentry *rt = (struct rtentry *)rn; struct ifnet *ifp = arg; int err; if (rt->rt_ifp == ifp) { /* * Protect (sorta) against walktree recursion problems * with cloned routes */ if ((rt->rt_flags & RTF_UP) == 0) return (0); err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway, rt_mask(rt), rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED, (struct rtentry **) NULL, rt->rt_fibnum); if (err) { log(LOG_WARNING, "if_rtdel: error %d\n", err); } } return (0); } /* * Return counter values from old racy non-pcpu counters. */ uint64_t if_get_counter_default(struct ifnet *ifp, ifnet_counter cnt) { switch (cnt) { case IFCOUNTER_IPACKETS: return (ifp->if_ipackets); case IFCOUNTER_IERRORS: return (ifp->if_ierrors); case IFCOUNTER_OPACKETS: return (ifp->if_opackets); case IFCOUNTER_OERRORS: return (ifp->if_oerrors); case IFCOUNTER_COLLISIONS: return (ifp->if_collisions); case IFCOUNTER_IBYTES: return (ifp->if_ibytes); case IFCOUNTER_OBYTES: return (ifp->if_obytes); case IFCOUNTER_IMCASTS: return (ifp->if_imcasts); case IFCOUNTER_OMCASTS: return (ifp->if_omcasts); case IFCOUNTER_IQDROPS: return (ifp->if_iqdrops); case IFCOUNTER_OQDROPS: return (ifp->if_oqdrops); case IFCOUNTER_NOPROTO: return (ifp->if_noproto); } panic("%s: unknown counter %d", __func__, 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, ifnet_counter cnt, int64_t inc) { switch (cnt) { case IFCOUNTER_IPACKETS: ifp->if_ipackets += inc; break; case IFCOUNTER_IERRORS: ifp->if_ierrors += inc; break; case IFCOUNTER_OPACKETS: ifp->if_opackets += inc; break; case IFCOUNTER_OERRORS: ifp->if_oerrors += inc; break; case IFCOUNTER_COLLISIONS: ifp->if_collisions += inc; break; case IFCOUNTER_IBYTES: ifp->if_ibytes += inc; break; case IFCOUNTER_OBYTES: ifp->if_obytes += inc; break; case IFCOUNTER_IMCASTS: ifp->if_imcasts += inc; break; case IFCOUNTER_OMCASTS: ifp->if_omcasts += inc; break; case IFCOUNTER_IQDROPS: ifp->if_iqdrops += inc; break; case IFCOUNTER_OQDROPS: ifp->if_oqdrops += inc; break; case IFCOUNTER_NOPROTO: ifp->if_noproto += inc; break; default: panic("%s: unknown counter %d", __func__, cnt); } } /* * 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); } } int ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) { int error = 0; struct rtentry *rt = NULL; struct rt_addrinfo info; static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; bzero(&info, sizeof(info)); 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; error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib); if (error == 0 && rt != NULL) { RT_LOCK(rt); ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = ifa->ifa_ifp->if_type; ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = ifa->ifa_ifp->if_index; RT_REMREF(rt); RT_UNLOCK(rt); } else if (error != 0) log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error); return (error); } int ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia) { int error = 0; struct rt_addrinfo info; struct sockaddr_dl null_sdl; bzero(&null_sdl, sizeof(null_sdl)); null_sdl.sdl_len = sizeof(null_sdl); null_sdl.sdl_family = AF_LINK; null_sdl.sdl_type = ifa->ifa_ifp->if_type; null_sdl.sdl_index = ifa->ifa_ifp->if_index; bzero(&info, sizeof(info)); 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; error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib); if (error != 0) log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error); return (error); } int ifa_switch_loopback_route(struct ifaddr *ifa, struct sockaddr *sa, int fib) { struct rtentry *rt; rt = rtalloc1_fib(sa, 0, 0, fib); if (rt == NULL) { log(LOG_DEBUG, "%s: fail", __func__); return (EHOSTUNREACH); } ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = ifa->ifa_ifp->if_type; ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = ifa->ifa_ifp->if_index; RTFREE_LOCKED(rt); return (0); } /* * 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) \ ((((struct sockaddr_dl *)(a1))->sdl_len == \ ((struct sockaddr_dl *)(a2))->sdl_len) && \ (bcmp(LLADDR((struct sockaddr_dl *)(a1)), \ LLADDR((struct sockaddr_dl *)(a2)), \ ((struct sockaddr_dl *)(a1))->sdl_alen) == 0)) /* * Locate an interface based on a complete address. */ /*ARGSUSED*/ static struct ifaddr * ifa_ifwithaddr_internal(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(struct sockaddr *addr) { return (ifa_ifwithaddr_internal(addr, 1)); } int ifa_ifwithaddr_check(struct sockaddr *addr) { return (ifa_ifwithaddr_internal(addr, 0) != NULL); } /* * Locate an interface based on the broadcast address. */ /* ARGSUSED */ struct ifaddr * ifa_ifwithbroadaddr(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(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(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; 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) { struct sockaddr_dl *sdl = (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) { 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(struct sockaddr *addr, struct ifnet *ifp) { struct ifaddr *ifa; 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; RT_LOCK_ASSERT(rt); 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) && IFP2AC(ifp)->ac_netgraph != 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 } 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 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); EVENTHANDLER_INVOKE(iflladdr_event, ifp); 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, 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. */ 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); } #ifdef INET /* * Also send gratuitous ARPs to notify other nodes about * the address change. */ TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { if (ifa->ifa_addr->sa_family == AF_INET) arp_ifinit(ifp, ifa); } #endif } 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); } 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_DROP(ifq); IF_UNLOCK(ifq); m_freem(m); return (0); } if (ifp != NULL) { ifp->if_obytes += m->m_pkthdr.len + adjust; if (m->m_flags & (M_BCAST|M_MCAST)) ifp->if_omcasts++; 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_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); } /* Statistics */ int if_incipackets(if_t ifp, int pkts) { ((struct ifnet *)ifp)->if_ipackets += pkts; return (0); } int if_incopackets(if_t ifp, int pkts) { ((struct ifnet *)ifp)->if_opackets += pkts; return (0); } int if_incierrors(if_t ifp, int ierrors) { ((struct ifnet *)ifp)->if_ierrors += ierrors; return (0); } int if_setierrors(if_t ifp, int ierrors) { ((struct ifnet *)ifp)->if_ierrors = ierrors; return (0); } int if_setoerrors(if_t ifp, int oerrors) { ((struct ifnet *)ifp)->if_oerrors = oerrors; return (0); } int if_incoerrors(if_t ifp, int oerrors) { ((struct ifnet *)ifp)->if_oerrors += oerrors; return (0); } int if_inciqdrops(if_t ifp, int val) { ((struct ifnet *)ifp)->if_iqdrops += val; return (0); } int if_setcollisions(if_t ifp, int collisions) { ((struct ifnet *)ifp)->if_collisions = collisions; return (0); } int if_inccollisions(if_t ifp, int collisions) { ((struct ifnet *)ifp)->if_collisions += collisions; return (0); } int if_setipackets(if_t ifp, int pkts) { ((struct ifnet *)ifp)->if_ipackets = pkts; return (0); } int if_setopackets(if_t ifp, int pkts) { ((struct ifnet *)ifp)->if_opackets = pkts; return (0); } int if_incobytes(if_t ifp, int bytes) { ((struct ifnet *)ifp)->if_obytes += bytes; return (0); } int if_setibytes(if_t ifp, int bytes) { ((struct ifnet *)ifp)->if_ibytes = bytes; return (0); } int if_setobytes(if_t ifp, int bytes) { ((struct ifnet *)ifp)->if_obytes = bytes; return (0); } int if_sendq_empty(if_t ifp) { return IFQ_DRV_IS_EMPTY(&((struct ifnet *)ifp)->if_snd); } int if_getiqdrops(if_t ifp) { return ((struct ifnet *)ifp)->if_iqdrops; } int if_incimcasts(if_t ifp, int mcast) { ((struct ifnet *)ifp)->if_imcasts += mcast; return (0); } int if_incomcasts(if_t ifp, int mcast) { ((struct ifnet *)ifp)->if_omcasts += mcast; return (0); } int if_setimcasts(if_t ifp, int mcast) { ((struct ifnet *)ifp)->if_imcasts = mcast; return (0); } 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_drv(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: head/sys/net/if_var.h =================================================================== --- head/sys/net/if_var.h (revision 271769) +++ head/sys/net/if_var.h (revision 271770) @@ -1,620 +1,621 @@ /*- * 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. * * From: @(#)if.h 8.1 (Berkeley) 6/10/93 * $FreeBSD$ */ #ifndef _NET_IF_VAR_H_ #define _NET_IF_VAR_H_ /* * Structures defining a network interface, providing a packet * transport mechanism (ala level 0 of the PUP protocols). * * Each interface accepts output datagrams of a specified maximum * length, and provides higher level routines with input datagrams * received from its medium. * * Output occurs when the routine if_output is called, with three parameters: * (*ifp->if_output)(ifp, m, dst, rt) * Here m is the mbuf chain to be sent and dst is the destination address. * The output routine encapsulates the supplied datagram if necessary, * and then transmits it on its medium. * * On input, each interface unwraps the data received by it, and either * places it on the input queue of an internetwork datagram routine * and posts the associated software interrupt, or passes the datagram to a raw * packet input routine. * * Routines exist for locating interfaces by their addresses * or for locating an interface on a certain network, as well as more general * routing and gateway routines maintaining information used to locate * interfaces. These routines live in the files if.c and route.c */ struct rtentry; /* ifa_rtrequest */ struct rt_addrinfo; /* ifa_rtrequest */ struct socket; struct carp_if; struct carp_softc; struct ifvlantrunk; struct route; /* if_output */ struct vnet; struct ifmedia; struct netmap_adapter; #ifdef _KERNEL #include /* ifqueue only? */ #include #include #endif /* _KERNEL */ #include #include /* XXX */ #include /* struct ifqueue */ #include /* XXX */ #include /* XXX */ #include /* if_link_task */ #define IF_DUNIT_NONE -1 #include TAILQ_HEAD(ifnethead, ifnet); /* we use TAILQs so that the order of */ TAILQ_HEAD(ifaddrhead, ifaddr); /* instantiation is preserved in the list */ TAILQ_HEAD(ifmultihead, ifmultiaddr); TAILQ_HEAD(ifgrouphead, ifg_group); #ifdef _KERNEL VNET_DECLARE(struct pfil_head, link_pfil_hook); /* packet filter hooks */ #define V_link_pfil_hook VNET(link_pfil_hook) #endif /* _KERNEL */ typedef enum { IFCOUNTER_IPACKETS = 1, IFCOUNTER_IERRORS, IFCOUNTER_OPACKETS, IFCOUNTER_OERRORS, IFCOUNTER_COLLISIONS, IFCOUNTER_IBYTES, IFCOUNTER_OBYTES, IFCOUNTER_IMCASTS, IFCOUNTER_OMCASTS, IFCOUNTER_IQDROPS, IFCOUNTER_OQDROPS, IFCOUNTER_NOPROTO, } ifnet_counter; typedef struct ifnet * if_t; typedef void (*if_start_fn_t)(if_t); typedef int (*if_ioctl_fn_t)(if_t, u_long, caddr_t); typedef void (*if_init_fn_t)(void *); typedef void (*if_qflush_fn_t)(if_t); typedef int (*if_transmit_fn_t)(if_t, struct mbuf *); typedef uint64_t (*if_get_counter_t)(if_t, ifnet_counter); /* * Structure defining a network interface. * * Size ILP32: 592 (approx) * LP64: 1048 (approx) */ struct ifnet { /* General book keeping of interface lists. */ TAILQ_ENTRY(ifnet) if_link; /* all struct ifnets are chained */ LIST_ENTRY(ifnet) if_clones; /* interfaces of a cloner */ TAILQ_HEAD(, ifg_list) if_groups; /* linked list of groups per if */ /* protected by if_addr_lock */ u_char if_alloctype; /* if_type at time of allocation */ /* Driver and protocol specific information that remains stable. */ void *if_softc; /* pointer to driver state */ void *if_llsoftc; /* link layer softc */ void *if_l2com; /* pointer to protocol bits */ const char *if_dname; /* driver name */ int if_dunit; /* unit or IF_DUNIT_NONE */ u_short if_index; /* numeric abbreviation for this if */ short if_index_reserved; /* spare space to grow if_index */ char if_xname[IFNAMSIZ]; /* external name (name + unit) */ char *if_description; /* interface description */ /* Variable fields that are touched by the stack and drivers. */ int if_flags; /* up/down, broadcast, etc. */ int if_drv_flags; /* driver-managed status flags */ int if_capabilities; /* interface features & capabilities */ int if_capenable; /* enabled features & capabilities */ void *if_linkmib; /* link-type-specific MIB data */ size_t if_linkmiblen; /* length of above data */ u_int if_refcount; /* reference count */ /* These fields are shared with struct if_data. */ uint8_t if_type; /* ethernet, tokenring, etc */ uint8_t if_addrlen; /* media address length */ uint8_t if_hdrlen; /* media header length */ uint8_t if_link_state; /* current link state */ uint32_t if_mtu; /* maximum transmission unit */ uint32_t if_metric; /* routing metric (external only) */ uint64_t if_baudrate; /* linespeed */ uint64_t if_hwassist; /* HW offload capabilities, see IFCAP */ time_t if_epoch; /* uptime at attach or stat reset */ struct timeval if_lastchange; /* time of last administrative change */ struct ifaltq if_snd; /* output queue (includes altq) */ struct task if_linktask; /* task for link change events */ /* Addresses of different protocol families assigned to this if. */ struct rwlock if_addr_lock; /* lock to protect address lists */ /* * if_addrhead is the list of all addresses associated to * an interface. * Some code in the kernel assumes that first element * of the list has type AF_LINK, and contains sockaddr_dl * addresses which store the link-level address and the name * of the interface. * However, access to the AF_LINK address through this * field is deprecated. Use if_addr or ifaddr_byindex() instead. */ struct ifaddrhead if_addrhead; /* linked list of addresses per if */ struct ifmultihead if_multiaddrs; /* multicast addresses configured */ int if_amcount; /* number of all-multicast requests */ struct ifaddr *if_addr; /* pointer to link-level address */ const u_int8_t *if_broadcastaddr; /* linklevel broadcast bytestring */ struct rwlock if_afdata_lock; void *if_afdata[AF_MAX]; int if_afdata_initialized; /* Additional features hung off the interface. */ u_int if_fib; /* interface FIB */ struct vnet *if_vnet; /* pointer to network stack instance */ struct vnet *if_home_vnet; /* where this ifnet originates from */ struct ifvlantrunk *if_vlantrunk; /* pointer to 802.1q data */ struct bpf_if *if_bpf; /* packet filter structure */ int if_pcount; /* number of promiscuous listeners */ void *if_bridge; /* bridge glue */ void *if_lagg; /* lagg glue */ void *if_pf_kif; /* pf glue */ struct carp_if *if_carp; /* carp interface structure */ struct label *if_label; /* interface MAC label */ struct netmap_adapter *if_netmap; /* netmap(4) softc */ /* Various procedures of the layer2 encapsulation and drivers. */ int (*if_output) /* output routine (enqueue) */ (struct ifnet *, struct mbuf *, const struct sockaddr *, struct route *); void (*if_input) /* input routine (from h/w driver) */ (struct ifnet *, struct mbuf *); if_start_fn_t if_start; /* initiate output routine */ if_ioctl_fn_t if_ioctl; /* ioctl routine */ if_init_fn_t if_init; /* Init routine */ int (*if_resolvemulti) /* validate/resolve multicast */ (struct ifnet *, struct sockaddr **, struct sockaddr *); if_qflush_fn_t if_qflush; /* flush any queue */ if_transmit_fn_t if_transmit; /* initiate output routine */ void (*if_reassign) /* reassign to vnet routine */ (struct ifnet *, struct vnet *, char *); if_get_counter_t if_get_counter; /* get counter values */ /* Stuff that's only temporary and doesn't belong here. */ u_int if_hw_tsomax; /* tso burst length limit, the minimum * is (IP_MAXPACKET / 8). * XXXAO: Have to find a better place * for it eventually. */ /* * Old, racy and expensive statistics, should not be used in * new drivers. */ uint64_t if_ipackets; /* packets received on interface */ uint64_t if_ierrors; /* input errors on interface */ uint64_t if_opackets; /* packets sent on interface */ uint64_t if_oerrors; /* output errors on interface */ uint64_t if_collisions; /* collisions on csma interfaces */ uint64_t if_ibytes; /* total number of octets received */ uint64_t if_obytes; /* total number of octets sent */ uint64_t if_imcasts; /* packets received via multicast */ uint64_t if_omcasts; /* packets sent via multicast */ uint64_t if_iqdrops; /* dropped on input */ uint64_t if_oqdrops; /* dropped on output */ uint64_t if_noproto; /* destined for unsupported protocol */ /* * Spare fields to be added before branching a stable branch, so * that structure can be enhanced without changing the kernel * binary interface. */ }; #include /* XXXAO: temporary unconditional include */ /* for compatibility with other BSDs */ #define if_addrlist if_addrhead #define if_list if_link #define if_name(ifp) ((ifp)->if_xname) /* * Locks for address lists on the network interface. */ #define IF_ADDR_LOCK_INIT(if) rw_init(&(if)->if_addr_lock, "if_addr_lock") #define IF_ADDR_LOCK_DESTROY(if) rw_destroy(&(if)->if_addr_lock) #define IF_ADDR_WLOCK(if) rw_wlock(&(if)->if_addr_lock) #define IF_ADDR_WUNLOCK(if) rw_wunlock(&(if)->if_addr_lock) #define IF_ADDR_RLOCK(if) rw_rlock(&(if)->if_addr_lock) #define IF_ADDR_RUNLOCK(if) rw_runlock(&(if)->if_addr_lock) #define IF_ADDR_LOCK_ASSERT(if) rw_assert(&(if)->if_addr_lock, RA_LOCKED) #define IF_ADDR_WLOCK_ASSERT(if) rw_assert(&(if)->if_addr_lock, RA_WLOCKED) /* * Function variations on locking macros intended to be used by loadable * kernel modules in order to divorce them from the internals of address list * locking. */ void if_addr_rlock(struct ifnet *ifp); /* if_addrhead */ void if_addr_runlock(struct ifnet *ifp); /* if_addrhead */ void if_maddr_rlock(if_t ifp); /* if_multiaddrs */ void if_maddr_runlock(if_t ifp); /* if_multiaddrs */ #ifdef _KERNEL #ifdef _SYS_EVENTHANDLER_H_ /* interface link layer address change event */ typedef void (*iflladdr_event_handler_t)(void *, struct ifnet *); EVENTHANDLER_DECLARE(iflladdr_event, iflladdr_event_handler_t); /* interface address change event */ typedef void (*ifaddr_event_handler_t)(void *, struct ifnet *); EVENTHANDLER_DECLARE(ifaddr_event, ifaddr_event_handler_t); /* new interface arrival event */ typedef void (*ifnet_arrival_event_handler_t)(void *, struct ifnet *); EVENTHANDLER_DECLARE(ifnet_arrival_event, ifnet_arrival_event_handler_t); /* interface departure event */ typedef void (*ifnet_departure_event_handler_t)(void *, struct ifnet *); EVENTHANDLER_DECLARE(ifnet_departure_event, ifnet_departure_event_handler_t); /* Interface link state change event */ typedef void (*ifnet_link_event_handler_t)(void *, struct ifnet *, int); EVENTHANDLER_DECLARE(ifnet_link_event, ifnet_link_event_handler_t); #endif /* _SYS_EVENTHANDLER_H_ */ /* * interface groups */ struct ifg_group { char ifg_group[IFNAMSIZ]; u_int ifg_refcnt; void *ifg_pf_kif; TAILQ_HEAD(, ifg_member) ifg_members; TAILQ_ENTRY(ifg_group) ifg_next; }; struct ifg_member { TAILQ_ENTRY(ifg_member) ifgm_next; struct ifnet *ifgm_ifp; }; struct ifg_list { struct ifg_group *ifgl_group; TAILQ_ENTRY(ifg_list) ifgl_next; }; #ifdef _SYS_EVENTHANDLER_H_ /* group attach event */ typedef void (*group_attach_event_handler_t)(void *, struct ifg_group *); EVENTHANDLER_DECLARE(group_attach_event, group_attach_event_handler_t); /* group detach event */ typedef void (*group_detach_event_handler_t)(void *, struct ifg_group *); EVENTHANDLER_DECLARE(group_detach_event, group_detach_event_handler_t); /* group change event */ typedef void (*group_change_event_handler_t)(void *, const char *); EVENTHANDLER_DECLARE(group_change_event, group_change_event_handler_t); #endif /* _SYS_EVENTHANDLER_H_ */ #define IF_AFDATA_LOCK_INIT(ifp) \ rw_init(&(ifp)->if_afdata_lock, "if_afdata") #define IF_AFDATA_WLOCK(ifp) rw_wlock(&(ifp)->if_afdata_lock) #define IF_AFDATA_RLOCK(ifp) rw_rlock(&(ifp)->if_afdata_lock) #define IF_AFDATA_WUNLOCK(ifp) rw_wunlock(&(ifp)->if_afdata_lock) #define IF_AFDATA_RUNLOCK(ifp) rw_runlock(&(ifp)->if_afdata_lock) #define IF_AFDATA_LOCK(ifp) IF_AFDATA_WLOCK(ifp) #define IF_AFDATA_UNLOCK(ifp) IF_AFDATA_WUNLOCK(ifp) #define IF_AFDATA_TRYLOCK(ifp) rw_try_wlock(&(ifp)->if_afdata_lock) #define IF_AFDATA_DESTROY(ifp) rw_destroy(&(ifp)->if_afdata_lock) #define IF_AFDATA_LOCK_ASSERT(ifp) rw_assert(&(ifp)->if_afdata_lock, RA_LOCKED) #define IF_AFDATA_RLOCK_ASSERT(ifp) rw_assert(&(ifp)->if_afdata_lock, RA_RLOCKED) #define IF_AFDATA_WLOCK_ASSERT(ifp) rw_assert(&(ifp)->if_afdata_lock, RA_WLOCKED) #define IF_AFDATA_UNLOCK_ASSERT(ifp) rw_assert(&(ifp)->if_afdata_lock, RA_UNLOCKED) /* * 72 was chosen below because it is the size of a TCP/IP * header (40) + the minimum mss (32). */ #define IF_MINMTU 72 #define IF_MAXMTU 65535 #define TOEDEV(ifp) ((ifp)->if_llsoftc) #endif /* _KERNEL */ /* * The ifaddr structure contains information about one address * of an interface. They are maintained by the different address families, * are allocated and attached when an address is set, and are linked * together so all addresses for an interface can be located. * * NOTE: a 'struct ifaddr' is always at the beginning of a larger * chunk of malloc'ed memory, where we store the three addresses * (ifa_addr, ifa_dstaddr and ifa_netmask) referenced here. */ #if defined(_KERNEL) || defined(_WANT_IFADDR) struct ifaddr { struct sockaddr *ifa_addr; /* address of interface */ struct sockaddr *ifa_dstaddr; /* other end of p-to-p link */ #define ifa_broadaddr ifa_dstaddr /* broadcast address interface */ struct sockaddr *ifa_netmask; /* used to determine subnet */ struct ifnet *ifa_ifp; /* back-pointer to interface */ struct carp_softc *ifa_carp; /* pointer to CARP data */ TAILQ_ENTRY(ifaddr) ifa_link; /* queue macro glue */ void (*ifa_rtrequest) /* check or clean routes (+ or -)'d */ (int, struct rtentry *, struct rt_addrinfo *); u_short ifa_flags; /* mostly rt_flags for cloning */ u_int ifa_refcnt; /* references to this structure */ counter_u64_t ifa_ipackets; counter_u64_t ifa_opackets; counter_u64_t ifa_ibytes; counter_u64_t ifa_obytes; }; #endif #ifdef _KERNEL #define IFA_ROUTE RTF_UP /* route installed */ #define IFA_RTSELF RTF_HOST /* loopback route to self installed */ /* For compatibility with other BSDs. SCTP uses it. */ #define ifa_list ifa_link struct ifaddr * ifa_alloc(size_t size, int flags); void ifa_free(struct ifaddr *ifa); void ifa_ref(struct ifaddr *ifa); #endif /* _KERNEL */ /* * Multicast address structure. This is analogous to the ifaddr * structure except that it keeps track of multicast addresses. */ struct ifmultiaddr { TAILQ_ENTRY(ifmultiaddr) ifma_link; /* queue macro glue */ struct sockaddr *ifma_addr; /* address this membership is for */ struct sockaddr *ifma_lladdr; /* link-layer translation, if any */ struct ifnet *ifma_ifp; /* back-pointer to interface */ u_int ifma_refcount; /* reference count */ void *ifma_protospec; /* protocol-specific state, if any */ struct ifmultiaddr *ifma_llifma; /* pointer to ifma for ifma_lladdr */ }; #ifdef _KERNEL extern struct rwlock ifnet_rwlock; extern struct sx ifnet_sxlock; #define IFNET_LOCK_INIT() do { \ rw_init_flags(&ifnet_rwlock, "ifnet_rw", RW_RECURSE); \ sx_init_flags(&ifnet_sxlock, "ifnet_sx", SX_RECURSE); \ } while(0) #define IFNET_WLOCK() do { \ sx_xlock(&ifnet_sxlock); \ rw_wlock(&ifnet_rwlock); \ } while (0) #define IFNET_WUNLOCK() do { \ rw_wunlock(&ifnet_rwlock); \ sx_xunlock(&ifnet_sxlock); \ } while (0) /* * To assert the ifnet lock, you must know not only whether it's for read or * write, but also whether it was acquired with sleep support or not. */ #define IFNET_RLOCK_ASSERT() sx_assert(&ifnet_sxlock, SA_SLOCKED) #define IFNET_RLOCK_NOSLEEP_ASSERT() rw_assert(&ifnet_rwlock, RA_RLOCKED) #define IFNET_WLOCK_ASSERT() do { \ sx_assert(&ifnet_sxlock, SA_XLOCKED); \ rw_assert(&ifnet_rwlock, RA_WLOCKED); \ } while (0) #define IFNET_RLOCK() sx_slock(&ifnet_sxlock) #define IFNET_RLOCK_NOSLEEP() rw_rlock(&ifnet_rwlock) #define IFNET_RUNLOCK() sx_sunlock(&ifnet_sxlock) #define IFNET_RUNLOCK_NOSLEEP() rw_runlock(&ifnet_rwlock) /* * Look up an ifnet given its index; the _ref variant also acquires a * reference that must be freed using if_rele(). It is almost always a bug * to call ifnet_byindex() instead if ifnet_byindex_ref(). */ struct ifnet *ifnet_byindex(u_short idx); struct ifnet *ifnet_byindex_locked(u_short idx); struct ifnet *ifnet_byindex_ref(u_short idx); /* * Given the index, ifaddr_byindex() returns the one and only * link-level ifaddr for the interface. You are not supposed to use * it to traverse the list of addresses associated to the interface. */ struct ifaddr *ifaddr_byindex(u_short idx); VNET_DECLARE(struct ifnethead, ifnet); VNET_DECLARE(struct ifgrouphead, ifg_head); VNET_DECLARE(int, if_index); VNET_DECLARE(struct ifnet *, loif); /* first loopback interface */ #define V_ifnet VNET(ifnet) #define V_ifg_head VNET(ifg_head) #define V_if_index VNET(if_index) #define V_loif VNET(loif) int if_addgroup(struct ifnet *, const char *); int if_delgroup(struct ifnet *, const char *); int if_addmulti(struct ifnet *, struct sockaddr *, struct ifmultiaddr **); int if_allmulti(struct ifnet *, int); struct ifnet* if_alloc(u_char); void if_attach(struct ifnet *); void if_dead(struct ifnet *); int if_delmulti(struct ifnet *, struct sockaddr *); void if_delmulti_ifma(struct ifmultiaddr *); void if_detach(struct ifnet *); void if_vmove(struct ifnet *, struct vnet *); void if_purgeaddrs(struct ifnet *); void if_delallmulti(struct ifnet *); void if_down(struct ifnet *); struct ifmultiaddr * if_findmulti(struct ifnet *, struct sockaddr *); void if_free(struct ifnet *); void if_initname(struct ifnet *, const char *, int); void if_link_state_change(struct ifnet *, int); int if_printf(struct ifnet *, const char *, ...) __printflike(2, 3); void if_ref(struct ifnet *); void if_rele(struct ifnet *); int if_setlladdr(struct ifnet *, const u_char *, int); void if_up(struct ifnet *); int ifioctl(struct socket *, u_long, caddr_t, struct thread *); int ifpromisc(struct ifnet *, int); struct ifnet *ifunit(const char *); struct ifnet *ifunit_ref(const char *); int ifa_add_loopback_route(struct ifaddr *, struct sockaddr *); int ifa_del_loopback_route(struct ifaddr *, struct sockaddr *); int ifa_switch_loopback_route(struct ifaddr *, struct sockaddr *, int fib); struct ifaddr *ifa_ifwithaddr(struct sockaddr *); int ifa_ifwithaddr_check(struct sockaddr *); struct ifaddr *ifa_ifwithbroadaddr(struct sockaddr *, int); struct ifaddr *ifa_ifwithdstaddr(struct sockaddr *, int); struct ifaddr *ifa_ifwithnet(struct sockaddr *, int, int); struct ifaddr *ifa_ifwithroute(int, struct sockaddr *, struct sockaddr *, u_int); struct ifaddr *ifaof_ifpforaddr(struct sockaddr *, struct ifnet *); int ifa_preferred(struct ifaddr *, struct ifaddr *); int if_simloop(struct ifnet *ifp, struct mbuf *m, int af, int hlen); typedef void *if_com_alloc_t(u_char type, struct ifnet *ifp); typedef void if_com_free_t(void *com, u_char type); void if_register_com_alloc(u_char type, if_com_alloc_t *a, if_com_free_t *f); void if_deregister_com_alloc(u_char type); void if_data_copy(struct ifnet *, struct if_data *); uint64_t if_get_counter_default(struct ifnet *, ifnet_counter); void if_inc_counter(struct ifnet *, ifnet_counter, int64_t); #define IF_LLADDR(ifp) \ LLADDR((struct sockaddr_dl *)((ifp)->if_addr->ifa_addr)) uint64_t if_setbaudrate(if_t ifp, uint64_t baudrate); uint64_t if_getbaudrate(if_t ifp); int if_setcapabilities(if_t ifp, int capabilities); int if_setcapabilitiesbit(if_t ifp, int setbit, int clearbit); int if_getcapabilities(if_t ifp); int if_togglecapenable(if_t ifp, int togglecap); int if_setcapenable(if_t ifp, int capenable); int if_setcapenablebit(if_t ifp, int setcap, int clearcap); int if_getcapenable(if_t ifp); const char *if_getdname(if_t ifp); int if_setdev(if_t ifp, void *dev); int if_setdrvflagbits(if_t ifp, int if_setflags, int clear_flags); int if_getdrvflags(if_t ifp); int if_setdrvflags(if_t ifp, int flags); int if_clearhwassist(if_t ifp); int if_sethwassistbits(if_t ifp, int toset, int toclear); int if_sethwassist(if_t ifp, int hwassist_bit); int if_gethwassist(if_t ifp); int if_setsoftc(if_t ifp, void *softc); void *if_getsoftc(if_t ifp); int if_setflags(if_t ifp, int flags); int if_setmtu(if_t ifp, int mtu); int if_getmtu(if_t ifp); int if_setflagbits(if_t ifp, int set, int clear); int if_getflags(if_t ifp); int if_sendq_empty(if_t ifp); int if_setsendqready(if_t ifp); int if_setsendqlen(if_t ifp, int tx_desc_count); int if_input(if_t ifp, struct mbuf* sendmp); int if_sendq_prepend(if_t ifp, struct mbuf *m); struct mbuf *if_dequeue(if_t ifp); int if_setifheaderlen(if_t ifp, int len); void if_setrcvif(struct mbuf *m, if_t ifp); void if_setvtag(struct mbuf *m, u_int16_t tag); u_int16_t if_getvtag(struct mbuf *m); int if_vlantrunkinuse(if_t ifp); caddr_t if_getlladdr(if_t ifp); void *if_gethandle(u_char); void if_bpfmtap(if_t ifp, struct mbuf *m); void if_etherbpfmtap(if_t ifp, struct mbuf *m); void if_vlancap(if_t ifp); int if_setupmultiaddr(if_t ifp, void *mta, int *cnt, int max); int if_multiaddr_array(if_t ifp, void *mta, int *cnt, int max); int if_multiaddr_count(if_t ifp, int max); int if_getamcount(if_t ifp); struct ifaddr * if_getifaddr(if_t ifp); /* Statistics */ int if_incipackets(if_t ifp, int pkt); int if_incopackets(if_t ifp, int pkts); int if_incierrors(if_t ifp, int ierrors); int if_incoerrors(if_t ifp, int oerrors); int if_inciqdrops(if_t ifp, int val); int if_setierrors(if_t ifp, int ierrors); int if_setoerrors(if_t ifp, int oerrors); int if_setcollisions(if_t ifp, int collisions); int if_inccollisions(if_t ifp, int collisions); int if_incobytes(if_t ifp, int bytes); int if_getiqdrops(if_t ifp); int if_incimcasts(if_t ifp, int imcasts); int if_incomcasts(if_t ifp, int imcasts); int if_setipackets(if_t ifp, int pkts); int if_setopackets(if_t ifp, int pkts); int if_setibytes(if_t ifp, int bytes); int if_setobytes(if_t ifp, int bytes); int if_setimcasts(if_t ifp, int pkts); /* Functions */ void if_setinitfn(if_t ifp, void (*)(void *)); void if_setioctlfn(if_t ifp, int (*)(if_t, u_long, caddr_t)); void if_setstartfn(if_t ifp, void (*)(if_t)); void if_settransmitfn(if_t ifp, if_transmit_fn_t); void if_setqflushfn(if_t ifp, if_qflush_fn_t); +void if_setgetcounterfn(if_t ifp, if_get_counter_t); /* Revisit the below. These are inline functions originally */ int drbr_inuse_drv(if_t ifp, struct buf_ring *br); struct mbuf* drbr_dequeue_drv(if_t ifp, struct buf_ring *br); int drbr_needs_enqueue_drv(if_t ifp, struct buf_ring *br); int drbr_enqueue_drv(if_t ifp, struct buf_ring *br, struct mbuf *m); #endif /* _KERNEL */ #endif /* !_NET_IF_VAR_H_ */ Index: head/sys/netgraph/ng_ppp.c =================================================================== --- head/sys/netgraph/ng_ppp.c (revision 271769) +++ head/sys/netgraph/ng_ppp.c (revision 271770) @@ -1,2618 +1,2646 @@ /*- * Copyright (c) 1996-2000 Whistle Communications, Inc. * All rights reserved. * * Subject to the following obligations and disclaimer of warranty, use and * redistribution of this software, in source or object code forms, with or * without modifications are expressly permitted by Whistle Communications; * provided, however, that: * 1. Any and all reproductions of the source or object code must include the * copyright notice above and the following disclaimer of warranties; and * 2. No rights are granted, in any manner or form, to use Whistle * Communications, Inc. trademarks, including the mark "WHISTLE * COMMUNICATIONS" on advertising, endorsements, or otherwise except as * such appears in the above copyright notice or in the software. * * THIS SOFTWARE IS BEING PROVIDED BY WHISTLE COMMUNICATIONS "AS IS", AND * TO THE MAXIMUM EXTENT PERMITTED BY LAW, WHISTLE COMMUNICATIONS MAKES NO * REPRESENTATIONS OR WARRANTIES, EXPRESS OR IMPLIED, REGARDING THIS SOFTWARE, * INCLUDING WITHOUT LIMITATION, ANY AND ALL IMPLIED WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR NON-INFRINGEMENT. * WHISTLE COMMUNICATIONS DOES NOT WARRANT, GUARANTEE, OR MAKE ANY * REPRESENTATIONS REGARDING THE USE OF, OR THE RESULTS OF THE USE OF THIS * SOFTWARE IN TERMS OF ITS CORRECTNESS, ACCURACY, RELIABILITY OR OTHERWISE. * IN NO EVENT SHALL WHISTLE COMMUNICATIONS BE LIABLE FOR ANY DAMAGES * RESULTING FROM OR ARISING OUT OF ANY USE OF THIS SOFTWARE, INCLUDING * WITHOUT LIMITATION, ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, * PUNITIVE, OR CONSEQUENTIAL DAMAGES, PROCUREMENT OF SUBSTITUTE GOODS OR * SERVICES, LOSS OF USE, DATA OR PROFITS, HOWEVER CAUSED AND UNDER 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 WHISTLE COMMUNICATIONS IS ADVISED OF THE POSSIBILITY * OF SUCH DAMAGE. * * Copyright (c) 2007 Alexander Motin * 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 unmodified, 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. * * Authors: Archie Cobbs , Alexander Motin * * $FreeBSD$ * $Whistle: ng_ppp.c,v 1.24 1999/11/01 09:24:52 julian Exp $ */ /* * PPP node type data-flow. * * hook xmit layer recv hook * ------------------------------------ * inet -> -> inet * ipv6 -> -> ipv6 * ipx -> proto -> ipx * atalk -> -> atalk * bypass -> -> bypass * -hcomp_xmit()----------proto_recv()- * vjc_ip <- <- vjc_ip * vjc_comp -> header compression -> vjc_comp * vjc_uncomp -> -> vjc_uncomp * vjc_vjip -> * -comp_xmit()-----------hcomp_recv()- * compress <- compression <- decompress * compress -> -> decompress * -crypt_xmit()-----------comp_recv()- * encrypt <- encryption <- decrypt * encrypt -> -> decrypt * -ml_xmit()-------------crypt_recv()- * multilink * -link_xmit()--------------ml_recv()- * linkX <- link <- linkX * */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef NG_SEPARATE_MALLOC static MALLOC_DEFINE(M_NETGRAPH_PPP, "netgraph_ppp", "netgraph ppp node"); #else #define M_NETGRAPH_PPP M_NETGRAPH #endif #define PROT_VALID(p) (((p) & 0x0101) == 0x0001) #define PROT_COMPRESSABLE(p) (((p) & 0xff00) == 0x0000) /* Some PPP protocol numbers we're interested in */ #define PROT_ATALK 0x0029 #define PROT_COMPD 0x00fd #define PROT_CRYPTD 0x0053 #define PROT_IP 0x0021 #define PROT_IPV6 0x0057 #define PROT_IPX 0x002b #define PROT_LCP 0xc021 #define PROT_MP 0x003d #define PROT_VJCOMP 0x002d #define PROT_VJUNCOMP 0x002f /* Multilink PPP definitions */ #define MP_INITIAL_SEQ 0 /* per RFC 1990 */ #define MP_MIN_LINK_MRU 32 #define MP_SHORT_SEQ_MASK 0x00000fff /* short seq # mask */ #define MP_SHORT_SEQ_HIBIT 0x00000800 /* short seq # high bit */ #define MP_SHORT_FIRST_FLAG 0x00008000 /* first fragment in frame */ #define MP_SHORT_LAST_FLAG 0x00004000 /* last fragment in frame */ #define MP_LONG_SEQ_MASK 0x00ffffff /* long seq # mask */ #define MP_LONG_SEQ_HIBIT 0x00800000 /* long seq # high bit */ #define MP_LONG_FIRST_FLAG 0x80000000 /* first fragment in frame */ #define MP_LONG_LAST_FLAG 0x40000000 /* last fragment in frame */ #define MP_NOSEQ 0x7fffffff /* impossible sequence number */ /* Sign extension of MP sequence numbers */ #define MP_SHORT_EXTEND(s) (((s) & MP_SHORT_SEQ_HIBIT) ? \ ((s) | ~MP_SHORT_SEQ_MASK) \ : ((s) & MP_SHORT_SEQ_MASK)) #define MP_LONG_EXTEND(s) (((s) & MP_LONG_SEQ_HIBIT) ? \ ((s) | ~MP_LONG_SEQ_MASK) \ : ((s) & MP_LONG_SEQ_MASK)) /* Comparision of MP sequence numbers. Note: all sequence numbers except priv->xseq are stored with the sign bit extended. */ #define MP_SHORT_SEQ_DIFF(x,y) MP_SHORT_EXTEND((x) - (y)) #define MP_LONG_SEQ_DIFF(x,y) MP_LONG_EXTEND((x) - (y)) #define MP_RECV_SEQ_DIFF(priv,x,y) \ ((priv)->conf.recvShortSeq ? \ MP_SHORT_SEQ_DIFF((x), (y)) : \ MP_LONG_SEQ_DIFF((x), (y))) /* Increment receive sequence number */ #define MP_NEXT_RECV_SEQ(priv,seq) \ ((priv)->conf.recvShortSeq ? \ MP_SHORT_EXTEND((seq) + 1) : \ MP_LONG_EXTEND((seq) + 1)) /* Don't fragment transmitted packets to parts smaller than this */ #define MP_MIN_FRAG_LEN 32 /* Maximum fragment reasssembly queue length */ #define MP_MAX_QUEUE_LEN 128 /* Fragment queue scanner period */ #define MP_FRAGTIMER_INTERVAL (hz/2) /* Average link overhead. XXX: Should be given by user-level */ #define MP_AVERAGE_LINK_OVERHEAD 16 /* Keep this equal to ng_ppp_hook_names lower! */ #define HOOK_INDEX_MAX 13 /* We store incoming fragments this way */ struct ng_ppp_frag { int seq; /* fragment seq# */ uint8_t first; /* First in packet? */ uint8_t last; /* Last in packet? */ struct timeval timestamp; /* time of reception */ struct mbuf *data; /* Fragment data */ TAILQ_ENTRY(ng_ppp_frag) f_qent; /* Fragment queue */ }; /* Per-link private information */ struct ng_ppp_link { struct ng_ppp_link_conf conf; /* link configuration */ struct ng_ppp_link_stat64 stats; /* link stats */ hook_p hook; /* connection to link data */ int32_t seq; /* highest rec'd seq# - MSEQ */ uint32_t latency; /* calculated link latency */ struct timeval lastWrite; /* time of last write for MP */ int bytesInQueue; /* bytes in the output queue for MP */ }; /* Total per-node private information */ struct ng_ppp_private { struct ng_ppp_bund_conf conf; /* bundle config */ struct ng_ppp_link_stat64 bundleStats; /* bundle stats */ struct ng_ppp_link links[NG_PPP_MAX_LINKS];/* per-link info */ int32_t xseq; /* next out MP seq # */ int32_t mseq; /* min links[i].seq */ uint16_t activeLinks[NG_PPP_MAX_LINKS]; /* indicies */ uint16_t numActiveLinks; /* how many links up */ uint16_t lastLink; /* for round robin */ uint8_t vjCompHooked; /* VJ comp hooked up? */ uint8_t allLinksEqual; /* all xmit the same? */ hook_p hooks[HOOK_INDEX_MAX]; /* non-link hooks */ struct ng_ppp_frag fragsmem[MP_MAX_QUEUE_LEN]; /* fragments storage */ TAILQ_HEAD(ng_ppp_fraglist, ng_ppp_frag) /* fragment queue */ frags; TAILQ_HEAD(ng_ppp_fragfreelist, ng_ppp_frag) /* free fragment queue */ fragsfree; struct callout fragTimer; /* fraq queue check */ struct mtx rmtx; /* recv mutex */ struct mtx xmtx; /* xmit mutex */ }; typedef struct ng_ppp_private *priv_p; /* Netgraph node methods */ static ng_constructor_t ng_ppp_constructor; static ng_rcvmsg_t ng_ppp_rcvmsg; static ng_shutdown_t ng_ppp_shutdown; static ng_newhook_t ng_ppp_newhook; static ng_rcvdata_t ng_ppp_rcvdata; static ng_disconnect_t ng_ppp_disconnect; static ng_rcvdata_t ng_ppp_rcvdata_inet; +static ng_rcvdata_t ng_ppp_rcvdata_inet_fast; static ng_rcvdata_t ng_ppp_rcvdata_ipv6; static ng_rcvdata_t ng_ppp_rcvdata_ipx; static ng_rcvdata_t ng_ppp_rcvdata_atalk; static ng_rcvdata_t ng_ppp_rcvdata_bypass; static ng_rcvdata_t ng_ppp_rcvdata_vjc_ip; static ng_rcvdata_t ng_ppp_rcvdata_vjc_comp; static ng_rcvdata_t ng_ppp_rcvdata_vjc_uncomp; static ng_rcvdata_t ng_ppp_rcvdata_vjc_vjip; static ng_rcvdata_t ng_ppp_rcvdata_compress; static ng_rcvdata_t ng_ppp_rcvdata_decompress; static ng_rcvdata_t ng_ppp_rcvdata_encrypt; static ng_rcvdata_t ng_ppp_rcvdata_decrypt; /* We use integer indicies to refer to the non-link hooks. */ static const struct { char *const name; ng_rcvdata_t *fn; } ng_ppp_hook_names[] = { #define HOOK_INDEX_ATALK 0 { NG_PPP_HOOK_ATALK, ng_ppp_rcvdata_atalk }, #define HOOK_INDEX_BYPASS 1 { NG_PPP_HOOK_BYPASS, ng_ppp_rcvdata_bypass }, #define HOOK_INDEX_COMPRESS 2 { NG_PPP_HOOK_COMPRESS, ng_ppp_rcvdata_compress }, #define HOOK_INDEX_ENCRYPT 3 { NG_PPP_HOOK_ENCRYPT, ng_ppp_rcvdata_encrypt }, #define HOOK_INDEX_DECOMPRESS 4 { NG_PPP_HOOK_DECOMPRESS, ng_ppp_rcvdata_decompress }, #define HOOK_INDEX_DECRYPT 5 { NG_PPP_HOOK_DECRYPT, ng_ppp_rcvdata_decrypt }, #define HOOK_INDEX_INET 6 { NG_PPP_HOOK_INET, ng_ppp_rcvdata_inet }, #define HOOK_INDEX_IPX 7 { NG_PPP_HOOK_IPX, ng_ppp_rcvdata_ipx }, #define HOOK_INDEX_VJC_COMP 8 { NG_PPP_HOOK_VJC_COMP, ng_ppp_rcvdata_vjc_comp }, #define HOOK_INDEX_VJC_IP 9 { NG_PPP_HOOK_VJC_IP, ng_ppp_rcvdata_vjc_ip }, #define HOOK_INDEX_VJC_UNCOMP 10 { NG_PPP_HOOK_VJC_UNCOMP, ng_ppp_rcvdata_vjc_uncomp }, #define HOOK_INDEX_VJC_VJIP 11 { NG_PPP_HOOK_VJC_VJIP, ng_ppp_rcvdata_vjc_vjip }, #define HOOK_INDEX_IPV6 12 { NG_PPP_HOOK_IPV6, ng_ppp_rcvdata_ipv6 }, { NULL, NULL } }; /* Helper functions */ static int ng_ppp_proto_recv(node_p node, item_p item, uint16_t proto, uint16_t linkNum); static int ng_ppp_hcomp_xmit(node_p node, item_p item, uint16_t proto); static int ng_ppp_hcomp_recv(node_p node, item_p item, uint16_t proto, uint16_t linkNum); static int ng_ppp_comp_xmit(node_p node, item_p item, uint16_t proto); static int ng_ppp_comp_recv(node_p node, item_p item, uint16_t proto, uint16_t linkNum); static int ng_ppp_crypt_xmit(node_p node, item_p item, uint16_t proto); static int ng_ppp_crypt_recv(node_p node, item_p item, uint16_t proto, uint16_t linkNum); static int ng_ppp_mp_xmit(node_p node, item_p item, uint16_t proto); static int ng_ppp_mp_recv(node_p node, item_p item, uint16_t proto, uint16_t linkNum); static int ng_ppp_link_xmit(node_p node, item_p item, uint16_t proto, uint16_t linkNum, int plen); static int ng_ppp_bypass(node_p node, item_p item, uint16_t proto, uint16_t linkNum); static void ng_ppp_bump_mseq(node_p node, int32_t new_mseq); static int ng_ppp_frag_drop(node_p node); static int ng_ppp_check_packet(node_p node); static void ng_ppp_get_packet(node_p node, struct mbuf **mp); static int ng_ppp_frag_process(node_p node, item_p oitem); static int ng_ppp_frag_trim(node_p node); static void ng_ppp_frag_timeout(node_p node, hook_p hook, void *arg1, int arg2); static void ng_ppp_frag_checkstale(node_p node); static void ng_ppp_frag_reset(node_p node); static void ng_ppp_mp_strategy(node_p node, int len, int *distrib); static int ng_ppp_intcmp(void *latency, const void *v1, const void *v2); static struct mbuf *ng_ppp_addproto(struct mbuf *m, uint16_t proto, int compOK); static struct mbuf *ng_ppp_cutproto(struct mbuf *m, uint16_t *proto); static struct mbuf *ng_ppp_prepend(struct mbuf *m, const void *buf, int len); static int ng_ppp_config_valid(node_p node, const struct ng_ppp_node_conf *newConf); static void ng_ppp_update(node_p node, int newConf); static void ng_ppp_start_frag_timer(node_p node); static void ng_ppp_stop_frag_timer(node_p node); /* Parse type for struct ng_ppp_mp_state_type */ static const struct ng_parse_fixedarray_info ng_ppp_rseq_array_info = { &ng_parse_hint32_type, NG_PPP_MAX_LINKS }; static const struct ng_parse_type ng_ppp_rseq_array_type = { &ng_parse_fixedarray_type, &ng_ppp_rseq_array_info, }; static const struct ng_parse_struct_field ng_ppp_mp_state_type_fields[] = NG_PPP_MP_STATE_TYPE_INFO(&ng_ppp_rseq_array_type); static const struct ng_parse_type ng_ppp_mp_state_type = { &ng_parse_struct_type, &ng_ppp_mp_state_type_fields }; /* Parse type for struct ng_ppp_link_conf */ static const struct ng_parse_struct_field ng_ppp_link_type_fields[] = NG_PPP_LINK_TYPE_INFO; static const struct ng_parse_type ng_ppp_link_type = { &ng_parse_struct_type, &ng_ppp_link_type_fields }; /* Parse type for struct ng_ppp_bund_conf */ static const struct ng_parse_struct_field ng_ppp_bund_type_fields[] = NG_PPP_BUND_TYPE_INFO; static const struct ng_parse_type ng_ppp_bund_type = { &ng_parse_struct_type, &ng_ppp_bund_type_fields }; /* Parse type for struct ng_ppp_node_conf */ static const struct ng_parse_fixedarray_info ng_ppp_array_info = { &ng_ppp_link_type, NG_PPP_MAX_LINKS }; static const struct ng_parse_type ng_ppp_link_array_type = { &ng_parse_fixedarray_type, &ng_ppp_array_info, }; static const struct ng_parse_struct_field ng_ppp_conf_type_fields[] = NG_PPP_CONFIG_TYPE_INFO(&ng_ppp_bund_type, &ng_ppp_link_array_type); static const struct ng_parse_type ng_ppp_conf_type = { &ng_parse_struct_type, &ng_ppp_conf_type_fields }; /* Parse type for struct ng_ppp_link_stat */ static const struct ng_parse_struct_field ng_ppp_stats_type_fields[] = NG_PPP_STATS_TYPE_INFO; static const struct ng_parse_type ng_ppp_stats_type = { &ng_parse_struct_type, &ng_ppp_stats_type_fields }; /* Parse type for struct ng_ppp_link_stat64 */ static const struct ng_parse_struct_field ng_ppp_stats64_type_fields[] = NG_PPP_STATS64_TYPE_INFO; static const struct ng_parse_type ng_ppp_stats64_type = { &ng_parse_struct_type, &ng_ppp_stats64_type_fields }; /* List of commands and how to convert arguments to/from ASCII */ static const struct ng_cmdlist ng_ppp_cmds[] = { { NGM_PPP_COOKIE, NGM_PPP_SET_CONFIG, "setconfig", &ng_ppp_conf_type, NULL }, { NGM_PPP_COOKIE, NGM_PPP_GET_CONFIG, "getconfig", NULL, &ng_ppp_conf_type }, { NGM_PPP_COOKIE, NGM_PPP_GET_MP_STATE, "getmpstate", NULL, &ng_ppp_mp_state_type }, { NGM_PPP_COOKIE, NGM_PPP_GET_LINK_STATS, "getstats", &ng_parse_int16_type, &ng_ppp_stats_type }, { NGM_PPP_COOKIE, NGM_PPP_CLR_LINK_STATS, "clrstats", &ng_parse_int16_type, NULL }, { NGM_PPP_COOKIE, NGM_PPP_GETCLR_LINK_STATS, "getclrstats", &ng_parse_int16_type, &ng_ppp_stats_type }, { NGM_PPP_COOKIE, NGM_PPP_GET_LINK_STATS64, "getstats64", &ng_parse_int16_type, &ng_ppp_stats64_type }, { NGM_PPP_COOKIE, NGM_PPP_GETCLR_LINK_STATS64, "getclrstats64", &ng_parse_int16_type, &ng_ppp_stats64_type }, { 0 } }; /* Node type descriptor */ static struct ng_type ng_ppp_typestruct = { .version = NG_ABI_VERSION, .name = NG_PPP_NODE_TYPE, .constructor = ng_ppp_constructor, .rcvmsg = ng_ppp_rcvmsg, .shutdown = ng_ppp_shutdown, .newhook = ng_ppp_newhook, .rcvdata = ng_ppp_rcvdata, .disconnect = ng_ppp_disconnect, .cmdlist = ng_ppp_cmds, }; NETGRAPH_INIT(ppp, &ng_ppp_typestruct); /* Address and control field header */ static const uint8_t ng_ppp_acf[2] = { 0xff, 0x03 }; /* Maximum time we'll let a complete incoming packet sit in the queue */ static const struct timeval ng_ppp_max_staleness = { 2, 0 }; /* 2 seconds */ #define ERROUT(x) do { error = (x); goto done; } while (0) /************************************************************************ NETGRAPH NODE STUFF ************************************************************************/ /* * Node type constructor */ static int ng_ppp_constructor(node_p node) { priv_p priv; int i; /* Allocate private structure */ priv = malloc(sizeof(*priv), M_NETGRAPH_PPP, M_WAITOK | M_ZERO); NG_NODE_SET_PRIVATE(node, priv); /* Initialize state */ TAILQ_INIT(&priv->frags); TAILQ_INIT(&priv->fragsfree); for (i = 0; i < MP_MAX_QUEUE_LEN; i++) TAILQ_INSERT_TAIL(&priv->fragsfree, &priv->fragsmem[i], f_qent); for (i = 0; i < NG_PPP_MAX_LINKS; i++) priv->links[i].seq = MP_NOSEQ; ng_callout_init(&priv->fragTimer); mtx_init(&priv->rmtx, "ng_ppp_recv", NULL, MTX_DEF); mtx_init(&priv->xmtx, "ng_ppp_xmit", NULL, MTX_DEF); /* Done */ return (0); } /* * Give our OK for a hook to be added */ static int ng_ppp_newhook(node_p node, hook_p hook, const char *name) { const priv_p priv = NG_NODE_PRIVATE(node); hook_p *hookPtr = NULL; int linkNum = -1; int hookIndex = -1; /* Figure out which hook it is */ if (strncmp(name, NG_PPP_HOOK_LINK_PREFIX, /* a link hook? */ strlen(NG_PPP_HOOK_LINK_PREFIX)) == 0) { const char *cp; char *eptr; cp = name + strlen(NG_PPP_HOOK_LINK_PREFIX); if (!isdigit(*cp) || (cp[0] == '0' && cp[1] != '\0')) return (EINVAL); linkNum = (int)strtoul(cp, &eptr, 10); if (*eptr != '\0' || linkNum < 0 || linkNum >= NG_PPP_MAX_LINKS) return (EINVAL); hookPtr = &priv->links[linkNum].hook; hookIndex = ~linkNum; /* See if hook is already connected. */ if (*hookPtr != NULL) return (EISCONN); /* Disallow more than one link unless multilink is enabled. */ if (priv->links[linkNum].conf.enableLink && !priv->conf.enableMultilink && priv->numActiveLinks >= 1) return (ENODEV); } else { /* must be a non-link hook */ int i; for (i = 0; ng_ppp_hook_names[i].name != NULL; i++) { if (strcmp(name, ng_ppp_hook_names[i].name) == 0) { hookPtr = &priv->hooks[i]; hookIndex = i; break; } } if (ng_ppp_hook_names[i].name == NULL) return (EINVAL); /* no such hook */ /* See if hook is already connected */ if (*hookPtr != NULL) return (EISCONN); /* Every non-linkX hook have it's own function. */ NG_HOOK_SET_RCVDATA(hook, ng_ppp_hook_names[i].fn); } /* OK */ *hookPtr = hook; NG_HOOK_SET_PRIVATE(hook, (void *)(intptr_t)hookIndex); ng_ppp_update(node, 0); return (0); } /* * Receive a control message */ static int ng_ppp_rcvmsg(node_p node, item_p item, hook_p lasthook) { const priv_p priv = NG_NODE_PRIVATE(node); struct ng_mesg *resp = NULL; int error = 0; struct ng_mesg *msg; NGI_GET_MSG(item, msg); switch (msg->header.typecookie) { case NGM_PPP_COOKIE: switch (msg->header.cmd) { case NGM_PPP_SET_CONFIG: { struct ng_ppp_node_conf *const conf = (struct ng_ppp_node_conf *)msg->data; int i; /* Check for invalid or illegal config */ if (msg->header.arglen != sizeof(*conf)) ERROUT(EINVAL); if (!ng_ppp_config_valid(node, conf)) ERROUT(EINVAL); /* Copy config */ priv->conf = conf->bund; for (i = 0; i < NG_PPP_MAX_LINKS; i++) priv->links[i].conf = conf->links[i]; ng_ppp_update(node, 1); break; } case NGM_PPP_GET_CONFIG: { struct ng_ppp_node_conf *conf; int i; NG_MKRESPONSE(resp, msg, sizeof(*conf), M_NOWAIT); if (resp == NULL) ERROUT(ENOMEM); conf = (struct ng_ppp_node_conf *)resp->data; conf->bund = priv->conf; for (i = 0; i < NG_PPP_MAX_LINKS; i++) conf->links[i] = priv->links[i].conf; break; } case NGM_PPP_GET_MP_STATE: { struct ng_ppp_mp_state *info; int i; NG_MKRESPONSE(resp, msg, sizeof(*info), M_NOWAIT); if (resp == NULL) ERROUT(ENOMEM); info = (struct ng_ppp_mp_state *)resp->data; bzero(info, sizeof(*info)); for (i = 0; i < NG_PPP_MAX_LINKS; i++) { if (priv->links[i].seq != MP_NOSEQ) info->rseq[i] = priv->links[i].seq; } info->mseq = priv->mseq; info->xseq = priv->xseq; break; } case NGM_PPP_GET_LINK_STATS: case NGM_PPP_CLR_LINK_STATS: case NGM_PPP_GETCLR_LINK_STATS: case NGM_PPP_GET_LINK_STATS64: case NGM_PPP_GETCLR_LINK_STATS64: { struct ng_ppp_link_stat64 *stats; uint16_t linkNum; /* Process request. */ if (msg->header.arglen != sizeof(uint16_t)) ERROUT(EINVAL); linkNum = *((uint16_t *) msg->data); if (linkNum >= NG_PPP_MAX_LINKS && linkNum != NG_PPP_BUNDLE_LINKNUM) ERROUT(EINVAL); stats = (linkNum == NG_PPP_BUNDLE_LINKNUM) ? &priv->bundleStats : &priv->links[linkNum].stats; /* Make 64bit reply. */ if (msg->header.cmd == NGM_PPP_GET_LINK_STATS64 || msg->header.cmd == NGM_PPP_GETCLR_LINK_STATS64) { NG_MKRESPONSE(resp, msg, sizeof(struct ng_ppp_link_stat64), M_NOWAIT); if (resp == NULL) ERROUT(ENOMEM); bcopy(stats, resp->data, sizeof(*stats)); } else /* Make 32bit reply. */ if (msg->header.cmd == NGM_PPP_GET_LINK_STATS || msg->header.cmd == NGM_PPP_GETCLR_LINK_STATS) { struct ng_ppp_link_stat *rs; NG_MKRESPONSE(resp, msg, sizeof(struct ng_ppp_link_stat), M_NOWAIT); if (resp == NULL) ERROUT(ENOMEM); rs = (struct ng_ppp_link_stat *)resp->data; /* Truncate 64->32 bits. */ rs->xmitFrames = stats->xmitFrames; rs->xmitOctets = stats->xmitOctets; rs->recvFrames = stats->recvFrames; rs->recvOctets = stats->recvOctets; rs->badProtos = stats->badProtos; rs->runts = stats->runts; rs->dupFragments = stats->dupFragments; rs->dropFragments = stats->dropFragments; } /* Clear stats. */ if (msg->header.cmd != NGM_PPP_GET_LINK_STATS && msg->header.cmd != NGM_PPP_GET_LINK_STATS64) bzero(stats, sizeof(*stats)); break; } default: error = EINVAL; break; } break; case NGM_VJC_COOKIE: { /* * Forward it to the vjc node. leave the * old return address alone. * If we have no hook, let NG_RESPOND_MSG * clean up any remaining resources. * Because we have no resp, the item will be freed * along with anything it references. Don't * let msg be freed twice. */ NGI_MSG(item) = msg; /* put it back in the item */ msg = NULL; if ((lasthook = priv->hooks[HOOK_INDEX_VJC_IP])) { NG_FWD_ITEM_HOOK(error, item, lasthook); } return (error); } default: error = EINVAL; break; } done: NG_RESPOND_MSG(error, node, item, resp); NG_FREE_MSG(msg); return (error); } /* * Destroy node */ static int ng_ppp_shutdown(node_p node) { const priv_p priv = NG_NODE_PRIVATE(node); /* Stop fragment queue timer */ ng_ppp_stop_frag_timer(node); /* Take down netgraph node */ ng_ppp_frag_reset(node); mtx_destroy(&priv->rmtx); mtx_destroy(&priv->xmtx); bzero(priv, sizeof(*priv)); free(priv, M_NETGRAPH_PPP); NG_NODE_SET_PRIVATE(node, NULL); NG_NODE_UNREF(node); /* let the node escape */ return (0); } /* * Hook disconnection */ static int ng_ppp_disconnect(hook_p hook) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); const int index = (intptr_t)NG_HOOK_PRIVATE(hook); /* Zero out hook pointer */ if (index < 0) priv->links[~index].hook = NULL; else priv->hooks[index] = NULL; /* Update derived info (or go away if no hooks left). */ if (NG_NODE_NUMHOOKS(node) > 0) ng_ppp_update(node, 0); else if (NG_NODE_IS_VALID(node)) ng_rmnode_self(node); return (0); } /* * Proto layer */ /* * Receive data on a hook inet. */ static int ng_ppp_rcvdata_inet(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); if (!priv->conf.enableIP) { NG_FREE_ITEM(item); return (ENXIO); } return (ng_ppp_hcomp_xmit(NG_HOOK_NODE(hook), item, PROT_IP)); } /* + * Receive data on a hook inet and pass it directly to first link. + */ +static int +ng_ppp_rcvdata_inet_fast(hook_p hook, item_p item) +{ + const node_p node = NG_HOOK_NODE(hook); + const priv_p priv = NG_NODE_PRIVATE(node); + + return (ng_ppp_link_xmit(node, item, PROT_IP, priv->activeLinks[0], + NGI_M(item)->m_pkthdr.len)); +} + +/* * Receive data on a hook ipv6. */ static int ng_ppp_rcvdata_ipv6(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); if (!priv->conf.enableIPv6) { NG_FREE_ITEM(item); return (ENXIO); } return (ng_ppp_hcomp_xmit(NG_HOOK_NODE(hook), item, PROT_IPV6)); } /* * Receive data on a hook atalk. */ static int ng_ppp_rcvdata_atalk(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); if (!priv->conf.enableAtalk) { NG_FREE_ITEM(item); return (ENXIO); } return (ng_ppp_hcomp_xmit(NG_HOOK_NODE(hook), item, PROT_ATALK)); } /* * Receive data on a hook ipx */ static int ng_ppp_rcvdata_ipx(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); if (!priv->conf.enableIPX) { NG_FREE_ITEM(item); return (ENXIO); } return (ng_ppp_hcomp_xmit(NG_HOOK_NODE(hook), item, PROT_IPX)); } /* * Receive data on a hook bypass */ static int ng_ppp_rcvdata_bypass(hook_p hook, item_p item) { uint16_t linkNum; uint16_t proto; struct mbuf *m; NGI_GET_M(item, m); if (m->m_pkthdr.len < 4) { NG_FREE_ITEM(item); return (EINVAL); } if (m->m_len < 4 && (m = m_pullup(m, 4)) == NULL) { NG_FREE_ITEM(item); return (ENOBUFS); } linkNum = be16dec(mtod(m, uint8_t *)); proto = be16dec(mtod(m, uint8_t *) + 2); m_adj(m, 4); NGI_M(item) = m; if (linkNum == NG_PPP_BUNDLE_LINKNUM) return (ng_ppp_hcomp_xmit(NG_HOOK_NODE(hook), item, proto)); else return (ng_ppp_link_xmit(NG_HOOK_NODE(hook), item, proto, linkNum, 0)); } static int ng_ppp_bypass(node_p node, item_p item, uint16_t proto, uint16_t linkNum) { const priv_p priv = NG_NODE_PRIVATE(node); uint16_t hdr[2]; struct mbuf *m; int error; if (priv->hooks[HOOK_INDEX_BYPASS] == NULL) { NG_FREE_ITEM(item); return (ENXIO); } /* Add 4-byte bypass header. */ hdr[0] = htons(linkNum); hdr[1] = htons(proto); NGI_GET_M(item, m); if ((m = ng_ppp_prepend(m, &hdr, 4)) == NULL) { NG_FREE_ITEM(item); return (ENOBUFS); } NGI_M(item) = m; /* Send packet out hook. */ NG_FWD_ITEM_HOOK(error, item, priv->hooks[HOOK_INDEX_BYPASS]); return (error); } static int ng_ppp_proto_recv(node_p node, item_p item, uint16_t proto, uint16_t linkNum) { const priv_p priv = NG_NODE_PRIVATE(node); hook_p outHook = NULL; int error; #ifdef ALIGNED_POINTER struct mbuf *m, *n; NGI_GET_M(item, m); if (!ALIGNED_POINTER(mtod(m, caddr_t), uint32_t)) { n = m_defrag(m, M_NOWAIT); if (n == NULL) { m_freem(m); NG_FREE_ITEM(item); return (ENOBUFS); } m = n; } NGI_M(item) = m; #endif /* ALIGNED_POINTER */ switch (proto) { case PROT_IP: if (priv->conf.enableIP) outHook = priv->hooks[HOOK_INDEX_INET]; break; case PROT_IPV6: if (priv->conf.enableIPv6) outHook = priv->hooks[HOOK_INDEX_IPV6]; break; case PROT_ATALK: if (priv->conf.enableAtalk) outHook = priv->hooks[HOOK_INDEX_ATALK]; break; case PROT_IPX: if (priv->conf.enableIPX) outHook = priv->hooks[HOOK_INDEX_IPX]; break; } if (outHook == NULL) return (ng_ppp_bypass(node, item, proto, linkNum)); /* Send packet out hook. */ NG_FWD_ITEM_HOOK(error, item, outHook); return (error); } /* * Header compression layer */ static int ng_ppp_hcomp_xmit(node_p node, item_p item, uint16_t proto) { const priv_p priv = NG_NODE_PRIVATE(node); if (proto == PROT_IP && priv->conf.enableVJCompression && priv->vjCompHooked) { int error; /* Send packet out hook. */ NG_FWD_ITEM_HOOK(error, item, priv->hooks[HOOK_INDEX_VJC_IP]); return (error); } return (ng_ppp_comp_xmit(node, item, proto)); } /* * Receive data on a hook vjc_comp. */ static int ng_ppp_rcvdata_vjc_comp(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); if (!priv->conf.enableVJCompression) { NG_FREE_ITEM(item); return (ENXIO); } return (ng_ppp_comp_xmit(node, item, PROT_VJCOMP)); } /* * Receive data on a hook vjc_uncomp. */ static int ng_ppp_rcvdata_vjc_uncomp(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); if (!priv->conf.enableVJCompression) { NG_FREE_ITEM(item); return (ENXIO); } return (ng_ppp_comp_xmit(node, item, PROT_VJUNCOMP)); } /* * Receive data on a hook vjc_vjip. */ static int ng_ppp_rcvdata_vjc_vjip(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); if (!priv->conf.enableVJCompression) { NG_FREE_ITEM(item); return (ENXIO); } return (ng_ppp_comp_xmit(node, item, PROT_IP)); } static int ng_ppp_hcomp_recv(node_p node, item_p item, uint16_t proto, uint16_t linkNum) { const priv_p priv = NG_NODE_PRIVATE(node); if (priv->conf.enableVJDecompression && priv->vjCompHooked) { hook_p outHook = NULL; switch (proto) { case PROT_VJCOMP: outHook = priv->hooks[HOOK_INDEX_VJC_COMP]; break; case PROT_VJUNCOMP: outHook = priv->hooks[HOOK_INDEX_VJC_UNCOMP]; break; } if (outHook) { int error; /* Send packet out hook. */ NG_FWD_ITEM_HOOK(error, item, outHook); return (error); } } return (ng_ppp_proto_recv(node, item, proto, linkNum)); } /* * Receive data on a hook vjc_ip. */ static int ng_ppp_rcvdata_vjc_ip(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); if (!priv->conf.enableVJDecompression) { NG_FREE_ITEM(item); return (ENXIO); } return (ng_ppp_proto_recv(node, item, PROT_IP, NG_PPP_BUNDLE_LINKNUM)); } /* * Compression layer */ static int ng_ppp_comp_xmit(node_p node, item_p item, uint16_t proto) { const priv_p priv = NG_NODE_PRIVATE(node); if (priv->conf.enableCompression && proto < 0x4000 && proto != PROT_COMPD && proto != PROT_CRYPTD && priv->hooks[HOOK_INDEX_COMPRESS] != NULL) { struct mbuf *m; int error; NGI_GET_M(item, m); if ((m = ng_ppp_addproto(m, proto, 0)) == NULL) { NG_FREE_ITEM(item); return (ENOBUFS); } NGI_M(item) = m; /* Send packet out hook. */ NG_FWD_ITEM_HOOK(error, item, priv->hooks[HOOK_INDEX_COMPRESS]); return (error); } return (ng_ppp_crypt_xmit(node, item, proto)); } /* * Receive data on a hook compress. */ static int ng_ppp_rcvdata_compress(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); uint16_t proto; switch (priv->conf.enableCompression) { case NG_PPP_COMPRESS_NONE: NG_FREE_ITEM(item); return (ENXIO); case NG_PPP_COMPRESS_FULL: { struct mbuf *m; NGI_GET_M(item, m); if ((m = ng_ppp_cutproto(m, &proto)) == NULL) { NG_FREE_ITEM(item); return (EIO); } NGI_M(item) = m; if (!PROT_VALID(proto)) { NG_FREE_ITEM(item); return (EIO); } } break; default: proto = PROT_COMPD; break; } return (ng_ppp_crypt_xmit(node, item, proto)); } static int ng_ppp_comp_recv(node_p node, item_p item, uint16_t proto, uint16_t linkNum) { const priv_p priv = NG_NODE_PRIVATE(node); if (proto < 0x4000 && ((proto == PROT_COMPD && priv->conf.enableDecompression) || priv->conf.enableDecompression == NG_PPP_DECOMPRESS_FULL) && priv->hooks[HOOK_INDEX_DECOMPRESS] != NULL) { int error; if (priv->conf.enableDecompression == NG_PPP_DECOMPRESS_FULL) { struct mbuf *m; NGI_GET_M(item, m); if ((m = ng_ppp_addproto(m, proto, 0)) == NULL) { NG_FREE_ITEM(item); return (EIO); } NGI_M(item) = m; } /* Send packet out hook. */ NG_FWD_ITEM_HOOK(error, item, priv->hooks[HOOK_INDEX_DECOMPRESS]); return (error); } else if (proto == PROT_COMPD) { /* Disabled protos MUST be silently discarded, but * unsupported MUST not. Let user-level decide this. */ return (ng_ppp_bypass(node, item, proto, linkNum)); } return (ng_ppp_hcomp_recv(node, item, proto, linkNum)); } /* * Receive data on a hook decompress. */ static int ng_ppp_rcvdata_decompress(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); uint16_t proto; struct mbuf *m; if (!priv->conf.enableDecompression) { NG_FREE_ITEM(item); return (ENXIO); } NGI_GET_M(item, m); if ((m = ng_ppp_cutproto(m, &proto)) == NULL) { NG_FREE_ITEM(item); return (EIO); } NGI_M(item) = m; if (!PROT_VALID(proto)) { priv->bundleStats.badProtos++; NG_FREE_ITEM(item); return (EIO); } return (ng_ppp_hcomp_recv(node, item, proto, NG_PPP_BUNDLE_LINKNUM)); } /* * Encryption layer */ static int ng_ppp_crypt_xmit(node_p node, item_p item, uint16_t proto) { const priv_p priv = NG_NODE_PRIVATE(node); if (priv->conf.enableEncryption && proto < 0x4000 && proto != PROT_CRYPTD && priv->hooks[HOOK_INDEX_ENCRYPT] != NULL) { struct mbuf *m; int error; NGI_GET_M(item, m); if ((m = ng_ppp_addproto(m, proto, 0)) == NULL) { NG_FREE_ITEM(item); return (ENOBUFS); } NGI_M(item) = m; /* Send packet out hook. */ NG_FWD_ITEM_HOOK(error, item, priv->hooks[HOOK_INDEX_ENCRYPT]); return (error); } return (ng_ppp_mp_xmit(node, item, proto)); } /* * Receive data on a hook encrypt. */ static int ng_ppp_rcvdata_encrypt(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); if (!priv->conf.enableEncryption) { NG_FREE_ITEM(item); return (ENXIO); } return (ng_ppp_mp_xmit(node, item, PROT_CRYPTD)); } static int ng_ppp_crypt_recv(node_p node, item_p item, uint16_t proto, uint16_t linkNum) { const priv_p priv = NG_NODE_PRIVATE(node); if (proto == PROT_CRYPTD) { if (priv->conf.enableDecryption && priv->hooks[HOOK_INDEX_DECRYPT] != NULL) { int error; /* Send packet out hook. */ NG_FWD_ITEM_HOOK(error, item, priv->hooks[HOOK_INDEX_DECRYPT]); return (error); } else { /* Disabled protos MUST be silently discarded, but * unsupported MUST not. Let user-level decide this. */ return (ng_ppp_bypass(node, item, proto, linkNum)); } } return (ng_ppp_comp_recv(node, item, proto, linkNum)); } /* * Receive data on a hook decrypt. */ static int ng_ppp_rcvdata_decrypt(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); uint16_t proto; struct mbuf *m; if (!priv->conf.enableDecryption) { NG_FREE_ITEM(item); return (ENXIO); } NGI_GET_M(item, m); if ((m = ng_ppp_cutproto(m, &proto)) == NULL) { NG_FREE_ITEM(item); return (EIO); } NGI_M(item) = m; if (!PROT_VALID(proto)) { priv->bundleStats.badProtos++; NG_FREE_ITEM(item); return (EIO); } return (ng_ppp_comp_recv(node, item, proto, NG_PPP_BUNDLE_LINKNUM)); } /* * Link layer */ static int ng_ppp_link_xmit(node_p node, item_p item, uint16_t proto, uint16_t linkNum, int plen) { const priv_p priv = NG_NODE_PRIVATE(node); struct ng_ppp_link *link; int len, error; struct mbuf *m; uint16_t mru; /* Check if link correct. */ if (linkNum >= NG_PPP_MAX_LINKS) { ERROUT(ENETDOWN); } /* Get link pointer (optimization). */ link = &priv->links[linkNum]; /* Check link status (if real). */ if (link->hook == NULL) { ERROUT(ENETDOWN); } /* Extract mbuf. */ NGI_GET_M(item, m); /* Check peer's MRU for this link. */ mru = link->conf.mru; if (mru != 0 && m->m_pkthdr.len > mru) { NG_FREE_M(m); ERROUT(EMSGSIZE); } /* Prepend protocol number, possibly compressed. */ if ((m = ng_ppp_addproto(m, proto, link->conf.enableProtoComp)) == NULL) { ERROUT(ENOBUFS); } /* Prepend address and control field (unless compressed). */ if (proto == PROT_LCP || !link->conf.enableACFComp) { if ((m = ng_ppp_prepend(m, &ng_ppp_acf, 2)) == NULL) ERROUT(ENOBUFS); } /* Deliver frame. */ len = m->m_pkthdr.len; NG_FWD_NEW_DATA(error, item, link->hook, m); mtx_lock(&priv->xmtx); /* Update link stats. */ link->stats.xmitFrames++; link->stats.xmitOctets += len; /* Update bundle stats. */ if (plen > 0) { priv->bundleStats.xmitFrames++; priv->bundleStats.xmitOctets += plen; } /* Update 'bytes in queue' counter. */ if (error == 0) { /* bytesInQueue and lastWrite required only for mp_strategy. */ if (priv->conf.enableMultilink && !priv->allLinksEqual && !priv->conf.enableRoundRobin) { /* If queue was empty, then mark this time. */ if (link->bytesInQueue == 0) getmicrouptime(&link->lastWrite); link->bytesInQueue += len + MP_AVERAGE_LINK_OVERHEAD; /* Limit max queue length to 50 pkts. BW can be defined incorrectly and link may not signal overload. */ if (link->bytesInQueue > 50 * 1600) link->bytesInQueue = 50 * 1600; } } mtx_unlock(&priv->xmtx); return (error); done: NG_FREE_ITEM(item); return (error); } /* * Receive data on a hook linkX. */ static int ng_ppp_rcvdata(hook_p hook, item_p item) { const node_p node = NG_HOOK_NODE(hook); const priv_p priv = NG_NODE_PRIVATE(node); const int index = (intptr_t)NG_HOOK_PRIVATE(hook); const uint16_t linkNum = (uint16_t)~index; struct ng_ppp_link * const link = &priv->links[linkNum]; uint16_t proto; struct mbuf *m; int error = 0; KASSERT(linkNum < NG_PPP_MAX_LINKS, ("%s: bogus index 0x%x", __func__, index)); NGI_GET_M(item, m); mtx_lock(&priv->rmtx); /* Stats */ link->stats.recvFrames++; link->stats.recvOctets += m->m_pkthdr.len; /* Strip address and control fields, if present. */ if (m->m_len < 2 && (m = m_pullup(m, 2)) == NULL) ERROUT(ENOBUFS); if (mtod(m, uint8_t *)[0] == 0xff && mtod(m, uint8_t *)[1] == 0x03) m_adj(m, 2); /* Get protocol number */ if ((m = ng_ppp_cutproto(m, &proto)) == NULL) ERROUT(ENOBUFS); NGI_M(item) = m; /* Put changed m back into item. */ if (!PROT_VALID(proto)) { link->stats.badProtos++; ERROUT(EIO); } /* LCP packets must go directly to bypass. */ if (proto >= 0xB000) { mtx_unlock(&priv->rmtx); return (ng_ppp_bypass(node, item, proto, linkNum)); } /* Other packets are denied on a disabled link. */ if (!link->conf.enableLink) ERROUT(ENXIO); /* Proceed to multilink layer. Mutex will be unlocked inside. */ error = ng_ppp_mp_recv(node, item, proto, linkNum); mtx_assert(&priv->rmtx, MA_NOTOWNED); return (error); done: mtx_unlock(&priv->rmtx); NG_FREE_ITEM(item); return (error); } /* * Multilink layer */ /* * Handle an incoming multi-link fragment * * The fragment reassembly algorithm is somewhat complex. This is mainly * because we are required not to reorder the reconstructed packets, yet * fragments are only guaranteed to arrive in order on a per-link basis. * In other words, when we have a complete packet ready, but the previous * packet is still incomplete, we have to decide between delivering the * complete packet and throwing away the incomplete one, or waiting to * see if the remainder of the incomplete one arrives, at which time we * can deliver both packets, in order. * * This problem is exacerbated by "sequence number slew", which is when * the sequence numbers coming in from different links are far apart from * each other. In particular, certain unnamed equipment (*cough* Ascend) * has been seen to generate sequence number slew of up to 10 on an ISDN * 2B-channel MP link. There is nothing invalid about sequence number slew * but it makes the reasssembly process have to work harder. * * However, the peer is required to transmit fragments in order on each * link. That means if we define MSEQ as the minimum over all links of * the highest sequence number received on that link, then we can always * give up any hope of receiving a fragment with sequence number < MSEQ in * the future (all of this using 'wraparound' sequence number space). * Therefore we can always immediately throw away incomplete packets * missing fragments with sequence numbers < MSEQ. * * Here is an overview of our algorithm: * * o Received fragments are inserted into a queue, for which we * maintain these invariants between calls to this function: * * - Fragments are ordered in the queue by sequence number * - If a complete packet is at the head of the queue, then * the first fragment in the packet has seq# > MSEQ + 1 * (otherwise, we could deliver it immediately) * - If any fragments have seq# < MSEQ, then they are necessarily * part of a packet whose missing seq#'s are all > MSEQ (otherwise, * we can throw them away because they'll never be completed) * - The queue contains at most MP_MAX_QUEUE_LEN fragments * * o We have a periodic timer that checks the queue for the first * complete packet that has been sitting in the queue "too long". * When one is detected, all previous (incomplete) fragments are * discarded, their missing fragments are declared lost and MSEQ * is increased. * * o If we recieve a fragment with seq# < MSEQ, we throw it away * because we've already delcared it lost. * * This assumes linkNum != NG_PPP_BUNDLE_LINKNUM. */ static int ng_ppp_mp_recv(node_p node, item_p item, uint16_t proto, uint16_t linkNum) { const priv_p priv = NG_NODE_PRIVATE(node); struct ng_ppp_link *const link = &priv->links[linkNum]; struct ng_ppp_frag *frag; struct ng_ppp_frag *qent; int i, diff, inserted; struct mbuf *m; int error = 0; if ((!priv->conf.enableMultilink) || proto != PROT_MP) { /* Stats */ priv->bundleStats.recvFrames++; priv->bundleStats.recvOctets += NGI_M(item)->m_pkthdr.len; mtx_unlock(&priv->rmtx); return (ng_ppp_crypt_recv(node, item, proto, linkNum)); } NGI_GET_M(item, m); /* Get a new frag struct from the free queue */ if ((frag = TAILQ_FIRST(&priv->fragsfree)) == NULL) { printf("No free fragments headers in ng_ppp!\n"); NG_FREE_M(m); goto process; } /* Extract fragment information from MP header */ if (priv->conf.recvShortSeq) { uint16_t shdr; if (m->m_pkthdr.len < 2) { link->stats.runts++; NG_FREE_M(m); ERROUT(EINVAL); } if (m->m_len < 2 && (m = m_pullup(m, 2)) == NULL) ERROUT(ENOBUFS); shdr = be16dec(mtod(m, void *)); frag->seq = MP_SHORT_EXTEND(shdr); frag->first = (shdr & MP_SHORT_FIRST_FLAG) != 0; frag->last = (shdr & MP_SHORT_LAST_FLAG) != 0; diff = MP_SHORT_SEQ_DIFF(frag->seq, priv->mseq); m_adj(m, 2); } else { uint32_t lhdr; if (m->m_pkthdr.len < 4) { link->stats.runts++; NG_FREE_M(m); ERROUT(EINVAL); } if (m->m_len < 4 && (m = m_pullup(m, 4)) == NULL) ERROUT(ENOBUFS); lhdr = be32dec(mtod(m, void *)); frag->seq = MP_LONG_EXTEND(lhdr); frag->first = (lhdr & MP_LONG_FIRST_FLAG) != 0; frag->last = (lhdr & MP_LONG_LAST_FLAG) != 0; diff = MP_LONG_SEQ_DIFF(frag->seq, priv->mseq); m_adj(m, 4); } frag->data = m; getmicrouptime(&frag->timestamp); /* If sequence number is < MSEQ, we've already declared this fragment as lost, so we have no choice now but to drop it */ if (diff < 0) { link->stats.dropFragments++; NG_FREE_M(m); ERROUT(0); } /* Update highest received sequence number on this link and MSEQ */ priv->mseq = link->seq = frag->seq; for (i = 0; i < priv->numActiveLinks; i++) { struct ng_ppp_link *const alink = &priv->links[priv->activeLinks[i]]; if (MP_RECV_SEQ_DIFF(priv, alink->seq, priv->mseq) < 0) priv->mseq = alink->seq; } /* Remove frag struct from free queue. */ TAILQ_REMOVE(&priv->fragsfree, frag, f_qent); /* Add fragment to queue, which is sorted by sequence number */ inserted = 0; TAILQ_FOREACH_REVERSE(qent, &priv->frags, ng_ppp_fraglist, f_qent) { diff = MP_RECV_SEQ_DIFF(priv, frag->seq, qent->seq); if (diff > 0) { TAILQ_INSERT_AFTER(&priv->frags, qent, frag, f_qent); inserted = 1; break; } else if (diff == 0) { /* should never happen! */ link->stats.dupFragments++; NG_FREE_M(frag->data); TAILQ_INSERT_HEAD(&priv->fragsfree, frag, f_qent); ERROUT(EINVAL); } } if (!inserted) TAILQ_INSERT_HEAD(&priv->frags, frag, f_qent); process: /* Process the queue */ /* NOTE: rmtx will be unlocked for sending time! */ error = ng_ppp_frag_process(node, item); mtx_unlock(&priv->rmtx); return (error); done: mtx_unlock(&priv->rmtx); NG_FREE_ITEM(item); return (error); } /************************************************************************ HELPER STUFF ************************************************************************/ /* * If new mseq > current then set it and update all active links */ static void ng_ppp_bump_mseq(node_p node, int32_t new_mseq) { const priv_p priv = NG_NODE_PRIVATE(node); int i; if (MP_RECV_SEQ_DIFF(priv, priv->mseq, new_mseq) < 0) { priv->mseq = new_mseq; for (i = 0; i < priv->numActiveLinks; i++) { struct ng_ppp_link *const alink = &priv->links[priv->activeLinks[i]]; if (MP_RECV_SEQ_DIFF(priv, alink->seq, new_mseq) < 0) alink->seq = new_mseq; } } } /* * Examine our list of fragments, and determine if there is a * complete and deliverable packet at the head of the list. * Return 1 if so, zero otherwise. */ static int ng_ppp_check_packet(node_p node) { const priv_p priv = NG_NODE_PRIVATE(node); struct ng_ppp_frag *qent, *qnext; /* Check for empty queue */ if (TAILQ_EMPTY(&priv->frags)) return (0); /* Check first fragment is the start of a deliverable packet */ qent = TAILQ_FIRST(&priv->frags); if (!qent->first || MP_RECV_SEQ_DIFF(priv, qent->seq, priv->mseq) > 1) return (0); /* Check that all the fragments are there */ while (!qent->last) { qnext = TAILQ_NEXT(qent, f_qent); if (qnext == NULL) /* end of queue */ return (0); if (qnext->seq != MP_NEXT_RECV_SEQ(priv, qent->seq)) return (0); qent = qnext; } /* Got one */ return (1); } /* * Pull a completed packet off the head of the incoming fragment queue. * This assumes there is a completed packet there to pull off. */ static void ng_ppp_get_packet(node_p node, struct mbuf **mp) { const priv_p priv = NG_NODE_PRIVATE(node); struct ng_ppp_frag *qent, *qnext; struct mbuf *m = NULL, *tail; qent = TAILQ_FIRST(&priv->frags); KASSERT(!TAILQ_EMPTY(&priv->frags) && qent->first, ("%s: no packet", __func__)); for (tail = NULL; qent != NULL; qent = qnext) { qnext = TAILQ_NEXT(qent, f_qent); KASSERT(!TAILQ_EMPTY(&priv->frags), ("%s: empty q", __func__)); TAILQ_REMOVE(&priv->frags, qent, f_qent); if (tail == NULL) tail = m = qent->data; else { m->m_pkthdr.len += qent->data->m_pkthdr.len; tail->m_next = qent->data; } while (tail->m_next != NULL) tail = tail->m_next; if (qent->last) { qnext = NULL; /* Bump MSEQ if necessary */ ng_ppp_bump_mseq(node, qent->seq); } TAILQ_INSERT_HEAD(&priv->fragsfree, qent, f_qent); } *mp = m; } /* * Trim fragments from the queue whose packets can never be completed. * This assumes a complete packet is NOT at the beginning of the queue. * Returns 1 if fragments were removed, zero otherwise. */ static int ng_ppp_frag_trim(node_p node) { const priv_p priv = NG_NODE_PRIVATE(node); struct ng_ppp_frag *qent, *qnext = NULL; int removed = 0; /* Scan for "dead" fragments and remove them */ while (1) { int dead = 0; /* If queue is empty, we're done */ if (TAILQ_EMPTY(&priv->frags)) break; /* Determine whether first fragment can ever be completed */ TAILQ_FOREACH(qent, &priv->frags, f_qent) { if (MP_RECV_SEQ_DIFF(priv, qent->seq, priv->mseq) >= 0) break; qnext = TAILQ_NEXT(qent, f_qent); KASSERT(qnext != NULL, ("%s: last frag < MSEQ?", __func__)); if (qnext->seq != MP_NEXT_RECV_SEQ(priv, qent->seq) || qent->last || qnext->first) { dead = 1; break; } } if (!dead) break; /* Remove fragment and all others in the same packet */ while ((qent = TAILQ_FIRST(&priv->frags)) != qnext) { KASSERT(!TAILQ_EMPTY(&priv->frags), ("%s: empty q", __func__)); priv->bundleStats.dropFragments++; TAILQ_REMOVE(&priv->frags, qent, f_qent); NG_FREE_M(qent->data); TAILQ_INSERT_HEAD(&priv->fragsfree, qent, f_qent); removed = 1; } } return (removed); } /* * Drop fragments on queue overflow. * Returns 1 if fragments were removed, zero otherwise. */ static int ng_ppp_frag_drop(node_p node) { const priv_p priv = NG_NODE_PRIVATE(node); /* Check queue length */ if (TAILQ_EMPTY(&priv->fragsfree)) { struct ng_ppp_frag *qent; /* Get oldest fragment */ KASSERT(!TAILQ_EMPTY(&priv->frags), ("%s: empty q", __func__)); qent = TAILQ_FIRST(&priv->frags); /* Bump MSEQ if necessary */ ng_ppp_bump_mseq(node, qent->seq); /* Drop it */ priv->bundleStats.dropFragments++; TAILQ_REMOVE(&priv->frags, qent, f_qent); NG_FREE_M(qent->data); TAILQ_INSERT_HEAD(&priv->fragsfree, qent, f_qent); return (1); } return (0); } /* * Run the queue, restoring the queue invariants */ static int ng_ppp_frag_process(node_p node, item_p oitem) { const priv_p priv = NG_NODE_PRIVATE(node); struct mbuf *m; item_p item; uint16_t proto; do { /* Deliver any deliverable packets */ while (ng_ppp_check_packet(node)) { ng_ppp_get_packet(node, &m); if ((m = ng_ppp_cutproto(m, &proto)) == NULL) continue; if (!PROT_VALID(proto)) { priv->bundleStats.badProtos++; NG_FREE_M(m); continue; } if (oitem) { /* If original item present - reuse it. */ item = oitem; oitem = NULL; NGI_M(item) = m; } else { item = ng_package_data(m, NG_NOFLAGS); } if (item != NULL) { /* Stats */ priv->bundleStats.recvFrames++; priv->bundleStats.recvOctets += NGI_M(item)->m_pkthdr.len; /* Drop mutex for the sending time. * Priv may change, but we are ready! */ mtx_unlock(&priv->rmtx); ng_ppp_crypt_recv(node, item, proto, NG_PPP_BUNDLE_LINKNUM); mtx_lock(&priv->rmtx); } } /* Delete dead fragments and try again */ } while (ng_ppp_frag_trim(node) || ng_ppp_frag_drop(node)); /* If we haven't reused original item - free it. */ if (oitem) NG_FREE_ITEM(oitem); /* Done */ return (0); } /* * Check for 'stale' completed packets that need to be delivered * * If a link goes down or has a temporary failure, MSEQ can get * "stuck", because no new incoming fragments appear on that link. * This can cause completed packets to never get delivered if * their sequence numbers are all > MSEQ + 1. * * This routine checks how long all of the completed packets have * been sitting in the queue, and if too long, removes fragments * from the queue and increments MSEQ to allow them to be delivered. */ static void ng_ppp_frag_checkstale(node_p node) { const priv_p priv = NG_NODE_PRIVATE(node); struct ng_ppp_frag *qent, *beg, *end; struct timeval now, age; struct mbuf *m; int seq; item_p item; int endseq; uint16_t proto; now.tv_sec = 0; /* uninitialized state */ while (1) { /* If queue is empty, we're done */ if (TAILQ_EMPTY(&priv->frags)) break; /* Find the first complete packet in the queue */ beg = end = NULL; seq = TAILQ_FIRST(&priv->frags)->seq; TAILQ_FOREACH(qent, &priv->frags, f_qent) { if (qent->first) beg = qent; else if (qent->seq != seq) beg = NULL; if (beg != NULL && qent->last) { end = qent; break; } seq = MP_NEXT_RECV_SEQ(priv, seq); } /* If none found, exit */ if (end == NULL) break; /* Get current time (we assume we've been up for >= 1 second) */ if (now.tv_sec == 0) getmicrouptime(&now); /* Check if packet has been queued too long */ age = now; timevalsub(&age, &beg->timestamp); if (timevalcmp(&age, &ng_ppp_max_staleness, < )) break; /* Throw away junk fragments in front of the completed packet */ while ((qent = TAILQ_FIRST(&priv->frags)) != beg) { KASSERT(!TAILQ_EMPTY(&priv->frags), ("%s: empty q", __func__)); priv->bundleStats.dropFragments++; TAILQ_REMOVE(&priv->frags, qent, f_qent); NG_FREE_M(qent->data); TAILQ_INSERT_HEAD(&priv->fragsfree, qent, f_qent); } /* Extract completed packet */ endseq = end->seq; ng_ppp_get_packet(node, &m); if ((m = ng_ppp_cutproto(m, &proto)) == NULL) continue; if (!PROT_VALID(proto)) { priv->bundleStats.badProtos++; NG_FREE_M(m); continue; } /* Deliver packet */ if ((item = ng_package_data(m, NG_NOFLAGS)) != NULL) { /* Stats */ priv->bundleStats.recvFrames++; priv->bundleStats.recvOctets += NGI_M(item)->m_pkthdr.len; ng_ppp_crypt_recv(node, item, proto, NG_PPP_BUNDLE_LINKNUM); } } } /* * Periodically call ng_ppp_frag_checkstale() */ static void ng_ppp_frag_timeout(node_p node, hook_p hook, void *arg1, int arg2) { /* XXX: is this needed? */ if (NG_NODE_NOT_VALID(node)) return; /* Scan the fragment queue */ ng_ppp_frag_checkstale(node); /* Start timer again */ ng_ppp_start_frag_timer(node); } /* * Deliver a frame out on the bundle, i.e., figure out how to fragment * the frame across the individual PPP links and do so. */ static int ng_ppp_mp_xmit(node_p node, item_p item, uint16_t proto) { const priv_p priv = NG_NODE_PRIVATE(node); const int hdr_len = priv->conf.xmitShortSeq ? 2 : 4; int distrib[NG_PPP_MAX_LINKS]; int firstFragment; int activeLinkNum; struct mbuf *m; int plen; int frags; int32_t seq; /* At least one link must be active */ if (priv->numActiveLinks == 0) { NG_FREE_ITEM(item); return (ENETDOWN); } /* Save length for later stats. */ plen = NGI_M(item)->m_pkthdr.len; if (!priv->conf.enableMultilink) { return (ng_ppp_link_xmit(node, item, proto, priv->activeLinks[0], plen)); } /* Check peer's MRRU for this bundle. */ if (plen > priv->conf.mrru) { NG_FREE_ITEM(item); return (EMSGSIZE); } /* Extract mbuf. */ NGI_GET_M(item, m); /* Prepend protocol number, possibly compressed. */ if ((m = ng_ppp_addproto(m, proto, 1)) == NULL) { NG_FREE_ITEM(item); return (ENOBUFS); } /* Clear distribution plan */ bzero(&distrib, priv->numActiveLinks * sizeof(distrib[0])); mtx_lock(&priv->xmtx); /* Round-robin strategy */ if (priv->conf.enableRoundRobin) { activeLinkNum = priv->lastLink++ % priv->numActiveLinks; distrib[activeLinkNum] = m->m_pkthdr.len; goto deliver; } /* Strategy when all links are equivalent (optimize the common case) */ if (priv->allLinksEqual) { int numFrags, fraction, remain; int i; /* Calculate optimal fragment count */ numFrags = priv->numActiveLinks; if (numFrags > m->m_pkthdr.len / MP_MIN_FRAG_LEN) numFrags = m->m_pkthdr.len / MP_MIN_FRAG_LEN; if (numFrags == 0) numFrags = 1; fraction = m->m_pkthdr.len / numFrags; remain = m->m_pkthdr.len - (fraction * numFrags); /* Assign distribution */ for (i = 0; i < numFrags; i++) { distrib[priv->lastLink++ % priv->numActiveLinks] = fraction + (((remain--) > 0)?1:0); } goto deliver; } /* Strategy when all links are not equivalent */ ng_ppp_mp_strategy(node, m->m_pkthdr.len, distrib); deliver: /* Estimate fragments count */ frags = 0; for (activeLinkNum = priv->numActiveLinks - 1; activeLinkNum >= 0; activeLinkNum--) { const uint16_t linkNum = priv->activeLinks[activeLinkNum]; struct ng_ppp_link *const link = &priv->links[linkNum]; frags += (distrib[activeLinkNum] + link->conf.mru - hdr_len - 1) / (link->conf.mru - hdr_len); } /* Get out initial sequence number */ seq = priv->xseq; /* Update next sequence number */ if (priv->conf.xmitShortSeq) { priv->xseq = (seq + frags) & MP_SHORT_SEQ_MASK; } else { priv->xseq = (seq + frags) & MP_LONG_SEQ_MASK; } mtx_unlock(&priv->xmtx); /* Send alloted portions of frame out on the link(s) */ for (firstFragment = 1, activeLinkNum = priv->numActiveLinks - 1; activeLinkNum >= 0; activeLinkNum--) { const uint16_t linkNum = priv->activeLinks[activeLinkNum]; struct ng_ppp_link *const link = &priv->links[linkNum]; /* Deliver fragment(s) out the next link */ for ( ; distrib[activeLinkNum] > 0; firstFragment = 0) { int len, lastFragment, error; struct mbuf *m2; /* Calculate fragment length; don't exceed link MTU */ len = distrib[activeLinkNum]; if (len > link->conf.mru - hdr_len) len = link->conf.mru - hdr_len; distrib[activeLinkNum] -= len; lastFragment = (len == m->m_pkthdr.len); /* Split off next fragment as "m2" */ m2 = m; if (!lastFragment) { struct mbuf *n = m_split(m, len, M_NOWAIT); if (n == NULL) { NG_FREE_M(m); if (firstFragment) NG_FREE_ITEM(item); return (ENOMEM); } m_tag_copy_chain(n, m, M_NOWAIT); m = n; } /* Prepend MP header */ if (priv->conf.xmitShortSeq) { uint16_t shdr; shdr = seq; seq = (seq + 1) & MP_SHORT_SEQ_MASK; if (firstFragment) shdr |= MP_SHORT_FIRST_FLAG; if (lastFragment) shdr |= MP_SHORT_LAST_FLAG; shdr = htons(shdr); m2 = ng_ppp_prepend(m2, &shdr, 2); } else { uint32_t lhdr; lhdr = seq; seq = (seq + 1) & MP_LONG_SEQ_MASK; if (firstFragment) lhdr |= MP_LONG_FIRST_FLAG; if (lastFragment) lhdr |= MP_LONG_LAST_FLAG; lhdr = htonl(lhdr); m2 = ng_ppp_prepend(m2, &lhdr, 4); } if (m2 == NULL) { if (!lastFragment) m_freem(m); if (firstFragment) NG_FREE_ITEM(item); return (ENOBUFS); } /* Send fragment */ if (firstFragment) { NGI_M(item) = m2; /* Reuse original item. */ } else { item = ng_package_data(m2, NG_NOFLAGS); } if (item != NULL) { error = ng_ppp_link_xmit(node, item, PROT_MP, linkNum, (firstFragment?plen:0)); if (error != 0) { if (!lastFragment) NG_FREE_M(m); return (error); } } } } /* Done */ return (0); } /* * Computing the optimal fragmentation * ----------------------------------- * * This routine tries to compute the optimal fragmentation pattern based * on each link's latency, bandwidth, and calculated additional latency. * The latter quantity is the additional latency caused by previously * written data that has not been transmitted yet. * * This algorithm is only useful when not all of the links have the * same latency and bandwidth values. * * The essential idea is to make the last bit of each fragment of the * frame arrive at the opposite end at the exact same time. This greedy * algorithm is optimal, in that no other scheduling could result in any * packet arriving any sooner unless packets are delivered out of order. * * Suppose link i has bandwidth b_i (in tens of bytes per milisecond) and * latency l_i (in miliseconds). Consider the function function f_i(t) * which is equal to the number of bytes that will have arrived at * the peer after t miliseconds if we start writing continuously at * time t = 0. Then f_i(t) = b_i * (t - l_i) = ((b_i * t) - (l_i * b_i). * That is, f_i(t) is a line with slope b_i and y-intersect -(l_i * b_i). * Note that the y-intersect is always <= zero because latency can't be * negative. Note also that really the function is f_i(t) except when * f_i(t) is negative, in which case the function is zero. To take * care of this, let Q_i(t) = { if (f_i(t) > 0) return 1; else return 0; }. * So the actual number of bytes that will have arrived at the peer after * t miliseconds is f_i(t) * Q_i(t). * * At any given time, each link has some additional latency a_i >= 0 * due to previously written fragment(s) which are still in the queue. * This value is easily computed from the time since last transmission, * the previous latency value, the number of bytes written, and the * link's bandwidth. * * Assume that l_i includes any a_i already, and that the links are * sorted by latency, so that l_i <= l_{i+1}. * * Let N be the total number of bytes in the current frame we are sending. * * Suppose we were to start writing bytes at time t = 0 on all links * simultaneously, which is the most we can possibly do. Then let * F(t) be equal to the total number of bytes received by the peer * after t miliseconds. Then F(t) = Sum_i (f_i(t) * Q_i(t)). * * Our goal is simply this: fragment the frame across the links such * that the peer is able to reconstruct the completed frame as soon as * possible, i.e., at the least possible value of t. Call this value t_0. * * Then it follows that F(t_0) = N. Our strategy is first to find the value * of t_0, and then deduce how many bytes to write to each link. * * Rewriting F(t_0): * * t_0 = ( N + Sum_i ( l_i * b_i * Q_i(t_0) ) ) / Sum_i ( b_i * Q_i(t_0) ) * * Now, we note that Q_i(t) is constant for l_i <= t <= l_{i+1}. t_0 will * lie in one of these ranges. To find it, we just need to find the i such * that F(l_i) <= N <= F(l_{i+1}). Then we compute all the constant values * for Q_i() in this range, plug in the remaining values, solving for t_0. * * Once t_0 is known, then the number of bytes to send on link i is * just f_i(t_0) * Q_i(t_0). * * In other words, we start allocating bytes to the links one at a time. * We keep adding links until the frame is completely sent. Some links * may not get any bytes because their latency is too high. * * Is all this work really worth the trouble? Depends on the situation. * The bigger the ratio of computer speed to link speed, and the more * important total bundle latency is (e.g., for interactive response time), * the more it's worth it. There is however the cost of calling this * function for every frame. The running time is O(n^2) where n is the * number of links that receive a non-zero number of bytes. * * Since latency is measured in miliseconds, the "resolution" of this * algorithm is one milisecond. * * To avoid this algorithm altogether, configure all links to have the * same latency and bandwidth. */ static void ng_ppp_mp_strategy(node_p node, int len, int *distrib) { const priv_p priv = NG_NODE_PRIVATE(node); int latency[NG_PPP_MAX_LINKS]; int sortByLatency[NG_PPP_MAX_LINKS]; int activeLinkNum; int t0, total, topSum, botSum; struct timeval now; int i, numFragments; /* If only one link, this gets real easy */ if (priv->numActiveLinks == 1) { distrib[0] = len; return; } /* Get current time */ getmicrouptime(&now); /* Compute latencies for each link at this point in time */ for (activeLinkNum = 0; activeLinkNum < priv->numActiveLinks; activeLinkNum++) { struct ng_ppp_link *alink; struct timeval diff; int xmitBytes; /* Start with base latency value */ alink = &priv->links[priv->activeLinks[activeLinkNum]]; latency[activeLinkNum] = alink->latency; sortByLatency[activeLinkNum] = activeLinkNum; /* see below */ /* Any additional latency? */ if (alink->bytesInQueue == 0) continue; /* Compute time delta since last write */ diff = now; timevalsub(&diff, &alink->lastWrite); /* alink->bytesInQueue will be changed, mark change time. */ alink->lastWrite = now; if (now.tv_sec < 0 || diff.tv_sec >= 10) { /* sanity */ alink->bytesInQueue = 0; continue; } /* How many bytes could have transmitted since last write? */ xmitBytes = (alink->conf.bandwidth * 10 * diff.tv_sec) + (alink->conf.bandwidth * (diff.tv_usec / 1000)) / 100; alink->bytesInQueue -= xmitBytes; if (alink->bytesInQueue < 0) alink->bytesInQueue = 0; else latency[activeLinkNum] += (100 * alink->bytesInQueue) / alink->conf.bandwidth; } /* Sort active links by latency */ qsort_r(sortByLatency, priv->numActiveLinks, sizeof(*sortByLatency), latency, ng_ppp_intcmp); /* Find the interval we need (add links in sortByLatency[] order) */ for (numFragments = 1; numFragments < priv->numActiveLinks; numFragments++) { for (total = i = 0; i < numFragments; i++) { int flowTime; flowTime = latency[sortByLatency[numFragments]] - latency[sortByLatency[i]]; total += ((flowTime * priv->links[ priv->activeLinks[sortByLatency[i]]].conf.bandwidth) + 99) / 100; } if (total >= len) break; } /* Solve for t_0 in that interval */ for (topSum = botSum = i = 0; i < numFragments; i++) { int bw = priv->links[ priv->activeLinks[sortByLatency[i]]].conf.bandwidth; topSum += latency[sortByLatency[i]] * bw; /* / 100 */ botSum += bw; /* / 100 */ } t0 = ((len * 100) + topSum + botSum / 2) / botSum; /* Compute f_i(t_0) all i */ for (total = i = 0; i < numFragments; i++) { int bw = priv->links[ priv->activeLinks[sortByLatency[i]]].conf.bandwidth; distrib[sortByLatency[i]] = (bw * (t0 - latency[sortByLatency[i]]) + 50) / 100; total += distrib[sortByLatency[i]]; } /* Deal with any rounding error */ if (total < len) { struct ng_ppp_link *fastLink = &priv->links[priv->activeLinks[sortByLatency[0]]]; int fast = 0; /* Find the fastest link */ for (i = 1; i < numFragments; i++) { struct ng_ppp_link *const link = &priv->links[priv->activeLinks[sortByLatency[i]]]; if (link->conf.bandwidth > fastLink->conf.bandwidth) { fast = i; fastLink = link; } } distrib[sortByLatency[fast]] += len - total; } else while (total > len) { struct ng_ppp_link *slowLink = &priv->links[priv->activeLinks[sortByLatency[0]]]; int delta, slow = 0; /* Find the slowest link that still has bytes to remove */ for (i = 1; i < numFragments; i++) { struct ng_ppp_link *const link = &priv->links[priv->activeLinks[sortByLatency[i]]]; if (distrib[sortByLatency[slow]] == 0 || (distrib[sortByLatency[i]] > 0 && link->conf.bandwidth < slowLink->conf.bandwidth)) { slow = i; slowLink = link; } } delta = total - len; if (delta > distrib[sortByLatency[slow]]) delta = distrib[sortByLatency[slow]]; distrib[sortByLatency[slow]] -= delta; total -= delta; } } /* * Compare two integers */ static int ng_ppp_intcmp(void *latency, const void *v1, const void *v2) { const int index1 = *((const int *) v1); const int index2 = *((const int *) v2); return ((int *)latency)[index1] - ((int *)latency)[index2]; } /* * Prepend a possibly compressed PPP protocol number in front of a frame */ static struct mbuf * ng_ppp_addproto(struct mbuf *m, uint16_t proto, int compOK) { if (compOK && PROT_COMPRESSABLE(proto)) { uint8_t pbyte = (uint8_t)proto; return ng_ppp_prepend(m, &pbyte, 1); } else { uint16_t pword = htons((uint16_t)proto); return ng_ppp_prepend(m, &pword, 2); } } /* * Cut a possibly compressed PPP protocol number from the front of a frame. */ static struct mbuf * ng_ppp_cutproto(struct mbuf *m, uint16_t *proto) { *proto = 0; if (m->m_len < 1 && (m = m_pullup(m, 1)) == NULL) return (NULL); *proto = *mtod(m, uint8_t *); m_adj(m, 1); if (!PROT_VALID(*proto)) { if (m->m_len < 1 && (m = m_pullup(m, 1)) == NULL) return (NULL); *proto = (*proto << 8) + *mtod(m, uint8_t *); m_adj(m, 1); } return (m); } /* * Prepend some bytes to an mbuf. */ static struct mbuf * ng_ppp_prepend(struct mbuf *m, const void *buf, int len) { M_PREPEND(m, len, M_NOWAIT); if (m == NULL || (m->m_len < len && (m = m_pullup(m, len)) == NULL)) return (NULL); bcopy(buf, mtod(m, uint8_t *), len); return (m); } /* * Update private information that is derived from other private information */ static void ng_ppp_update(node_p node, int newConf) { const priv_p priv = NG_NODE_PRIVATE(node); int i; /* Update active status for VJ Compression */ priv->vjCompHooked = priv->hooks[HOOK_INDEX_VJC_IP] != NULL && priv->hooks[HOOK_INDEX_VJC_COMP] != NULL && priv->hooks[HOOK_INDEX_VJC_UNCOMP] != NULL && priv->hooks[HOOK_INDEX_VJC_VJIP] != NULL; /* Increase latency for each link an amount equal to one MP header */ if (newConf) { for (i = 0; i < NG_PPP_MAX_LINKS; i++) { int hdrBytes; if (priv->links[i].conf.bandwidth == 0) continue; hdrBytes = MP_AVERAGE_LINK_OVERHEAD + (priv->links[i].conf.enableACFComp ? 0 : 2) + (priv->links[i].conf.enableProtoComp ? 1 : 2) + (priv->conf.xmitShortSeq ? 2 : 4); priv->links[i].latency = priv->links[i].conf.latency + (hdrBytes / priv->links[i].conf.bandwidth + 50) / 100; } } /* Update list of active links */ bzero(&priv->activeLinks, sizeof(priv->activeLinks)); priv->numActiveLinks = 0; priv->allLinksEqual = 1; for (i = 0; i < NG_PPP_MAX_LINKS; i++) { struct ng_ppp_link *const link = &priv->links[i]; /* Is link active? */ if (link->conf.enableLink && link->hook != NULL) { struct ng_ppp_link *link0; /* Add link to list of active links */ priv->activeLinks[priv->numActiveLinks++] = i; link0 = &priv->links[priv->activeLinks[0]]; /* Determine if all links are still equal */ if (link->latency != link0->latency || link->conf.bandwidth != link0->conf.bandwidth) priv->allLinksEqual = 0; /* Initialize rec'd sequence number */ if (link->seq == MP_NOSEQ) { link->seq = (link == link0) ? MP_INITIAL_SEQ : link0->seq; } } else link->seq = MP_NOSEQ; } /* Update MP state as multi-link is active or not */ if (priv->conf.enableMultilink && priv->numActiveLinks > 0) ng_ppp_start_frag_timer(node); else { ng_ppp_stop_frag_timer(node); ng_ppp_frag_reset(node); priv->xseq = MP_INITIAL_SEQ; priv->mseq = MP_INITIAL_SEQ; for (i = 0; i < NG_PPP_MAX_LINKS; i++) { struct ng_ppp_link *const link = &priv->links[i]; bzero(&link->lastWrite, sizeof(link->lastWrite)); link->bytesInQueue = 0; link->seq = MP_NOSEQ; } + } + + if (priv->hooks[HOOK_INDEX_INET] != NULL) { + if (priv->conf.enableIP == 1 && + priv->numActiveLinks == 1 && + priv->conf.enableMultilink == 0 && + priv->conf.enableCompression == 0 && + priv->conf.enableEncryption == 0 && + priv->conf.enableVJCompression == 0) + NG_HOOK_SET_RCVDATA(priv->hooks[HOOK_INDEX_INET], + ng_ppp_rcvdata_inet_fast); + else + NG_HOOK_SET_RCVDATA(priv->hooks[HOOK_INDEX_INET], + ng_ppp_rcvdata_inet); } } /* * Determine if a new configuration would represent a valid change * from the current configuration and link activity status. */ static int ng_ppp_config_valid(node_p node, const struct ng_ppp_node_conf *newConf) { const priv_p priv = NG_NODE_PRIVATE(node); int i, newNumLinksActive; /* Check per-link config and count how many links would be active */ for (newNumLinksActive = i = 0; i < NG_PPP_MAX_LINKS; i++) { if (newConf->links[i].enableLink && priv->links[i].hook != NULL) newNumLinksActive++; if (!newConf->links[i].enableLink) continue; if (newConf->links[i].mru < MP_MIN_LINK_MRU) return (0); if (newConf->links[i].bandwidth == 0) return (0); if (newConf->links[i].bandwidth > NG_PPP_MAX_BANDWIDTH) return (0); if (newConf->links[i].latency > NG_PPP_MAX_LATENCY) return (0); } /* Disallow changes to multi-link configuration while MP is active */ if (priv->numActiveLinks > 0 && newNumLinksActive > 0) { if (!priv->conf.enableMultilink != !newConf->bund.enableMultilink || !priv->conf.xmitShortSeq != !newConf->bund.xmitShortSeq || !priv->conf.recvShortSeq != !newConf->bund.recvShortSeq) return (0); } /* At most one link can be active unless multi-link is enabled */ if (!newConf->bund.enableMultilink && newNumLinksActive > 1) return (0); /* Configuration change would be valid */ return (1); } /* * Free all entries in the fragment queue */ static void ng_ppp_frag_reset(node_p node) { const priv_p priv = NG_NODE_PRIVATE(node); struct ng_ppp_frag *qent, *qnext; for (qent = TAILQ_FIRST(&priv->frags); qent; qent = qnext) { qnext = TAILQ_NEXT(qent, f_qent); NG_FREE_M(qent->data); TAILQ_INSERT_HEAD(&priv->fragsfree, qent, f_qent); } TAILQ_INIT(&priv->frags); } /* * Start fragment queue timer */ static void ng_ppp_start_frag_timer(node_p node) { const priv_p priv = NG_NODE_PRIVATE(node); if (!(callout_pending(&priv->fragTimer))) ng_callout(&priv->fragTimer, node, NULL, MP_FRAGTIMER_INTERVAL, ng_ppp_frag_timeout, NULL, 0); } /* * Stop fragment queue timer */ static void ng_ppp_stop_frag_timer(node_p node) { const priv_p priv = NG_NODE_PRIVATE(node); if (callout_pending(&priv->fragTimer)) ng_uncallout(&priv->fragTimer, node); }