diff --git a/sys/netinet/raw_ip.c b/sys/netinet/raw_ip.c index 1953fd86ca05..12ab675a97c1 100644 --- a/sys/netinet/raw_ip.c +++ b/sys/netinet/raw_ip.c @@ -1,865 +1,867 @@ /* * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95 * $FreeBSD$ */ #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_mac.h" #include "opt_random_ip_id.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef FAST_IPSEC #include #endif /*FAST_IPSEC*/ #ifdef IPSEC #include #endif /*IPSEC*/ struct inpcbhead ripcb; struct inpcbinfo ripcbinfo; /* control hooks for ipfw and dummynet */ ip_fw_ctl_t *ip_fw_ctl_ptr; ip_dn_ctl_t *ip_dn_ctl_ptr; /* * hooks for multicast routing. They all default to NULL, * so leave them not initialized and rely on BSS being set to 0. */ /* The socket used to communicate with the multicast routing daemon. */ struct socket *ip_mrouter; /* The various mrouter and rsvp functions */ int (*ip_mrouter_set)(struct socket *, struct sockopt *); int (*ip_mrouter_get)(struct socket *, struct sockopt *); int (*ip_mrouter_done)(void); int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, struct ip_moptions *); int (*mrt_ioctl)(int, caddr_t); int (*legal_vif_num)(int); u_long (*ip_mcast_src)(int); void (*rsvp_input_p)(struct mbuf *m, int off); int (*ip_rsvp_vif)(struct socket *, struct sockopt *); void (*ip_rsvp_force_done)(struct socket *); /* * Nominal space allocated to a raw ip socket. */ #define RIPSNDQ 8192 #define RIPRCVQ 8192 /* * Raw interface to IP protocol. */ /* * Initialize raw connection block q. */ void rip_init() { INP_INFO_LOCK_INIT(&ripcbinfo, "rip"); LIST_INIT(&ripcb); ripcbinfo.listhead = &ripcb; /* * XXX We don't use the hash list for raw IP, but it's easier * to allocate a one entry hash list than it is to check all * over the place for hashbase == NULL. */ ripcbinfo.hashbase = hashinit(1, M_PCB, &ripcbinfo.hashmask); ripcbinfo.porthashbase = hashinit(1, M_PCB, &ripcbinfo.porthashmask); ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets); } static struct sockaddr_in ripsrc = { sizeof(ripsrc), AF_INET }; static int raw_append(struct inpcb *last, struct ip *ip, struct mbuf *n) { int policyfail = 0; INP_LOCK_ASSERT(last); #if defined(IPSEC) || defined(FAST_IPSEC) /* check AH/ESP integrity. */ if (ipsec4_in_reject(n, last)) { policyfail = 1; #ifdef IPSEC ipsecstat.in_polvio++; #endif /*IPSEC*/ /* do not inject data to pcb */ } #endif /*IPSEC || FAST_IPSEC*/ #ifdef MAC if (!policyfail && mac_check_inpcb_deliver(last, n) != 0) policyfail = 1; #endif if (!policyfail) { struct mbuf *opts = NULL; if ((last->inp_flags & INP_CONTROLOPTS) || (last->inp_socket->so_options & SO_TIMESTAMP)) ip_savecontrol(last, &opts, ip, n); if (sbappendaddr(&last->inp_socket->so_rcv, (struct sockaddr *)&ripsrc, n, opts) == 0) { /* should notify about lost packet */ m_freem(n); if (opts) m_freem(opts); } else sorwakeup(last->inp_socket); } else m_freem(n); return policyfail; } /* * Setup generic address and protocol structures * for raw_input routine, then pass them along with * mbuf chain. */ void rip_input(struct mbuf *m, int off) { struct ip *ip = mtod(m, struct ip *); int proto = ip->ip_p; struct inpcb *inp, *last; INP_INFO_RLOCK(&ripcbinfo); ripsrc.sin_addr = ip->ip_src; last = NULL; LIST_FOREACH(inp, &ripcb, inp_list) { INP_LOCK(inp); if (inp->inp_ip_p && inp->inp_ip_p != proto) { docontinue: INP_UNLOCK(inp); continue; } #ifdef INET6 if ((inp->inp_vflag & INP_IPV4) == 0) goto docontinue; #endif if (inp->inp_laddr.s_addr && inp->inp_laddr.s_addr != ip->ip_dst.s_addr) goto docontinue; if (inp->inp_faddr.s_addr && inp->inp_faddr.s_addr != ip->ip_src.s_addr) goto docontinue; if (jailed(inp->inp_socket->so_cred)) if (htonl(prison_getip(inp->inp_socket->so_cred)) != ip->ip_dst.s_addr) goto docontinue; if (last) { struct mbuf *n; n = m_copy(m, 0, (int)M_COPYALL); if (n != NULL) (void) raw_append(last, ip, n); /* XXX count dropped packet */ INP_UNLOCK(last); } last = inp; } if (last != NULL) { if (raw_append(last, ip, m) != 0) ipstat.ips_delivered--; INP_UNLOCK(last); } else { m_freem(m); ipstat.ips_noproto++; ipstat.ips_delivered--; } INP_INFO_RUNLOCK(&ripcbinfo); } /* * Generate IP header and pass packet to ip_output. * Tack on options user may have setup with control call. */ int rip_output(struct mbuf *m, struct socket *so, u_long dst) { struct ip *ip; struct inpcb *inp = sotoinpcb(so); int flags = (so->so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST; #ifdef MAC - mac_create_mbuf_from_socket(so, m); + INP_LOCK(inp); + mac_create_mbuf_from_inpcb(inp, m); + INP_UNLOCK(inp); #endif /* * If the user handed us a complete IP packet, use it. * Otherwise, allocate an mbuf for a header and fill it in. */ if ((inp->inp_flags & INP_HDRINCL) == 0) { if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { m_freem(m); return(EMSGSIZE); } M_PREPEND(m, sizeof(struct ip), M_TRYWAIT); if (m == NULL) return(ENOBUFS); ip = mtod(m, struct ip *); ip->ip_tos = inp->inp_ip_tos; ip->ip_off = 0; ip->ip_p = inp->inp_ip_p; ip->ip_len = m->m_pkthdr.len; if (jailed(inp->inp_socket->so_cred)) ip->ip_src.s_addr = htonl(prison_getip(inp->inp_socket->so_cred)); else ip->ip_src = inp->inp_laddr; ip->ip_dst.s_addr = dst; ip->ip_ttl = inp->inp_ip_ttl; } else { if (m->m_pkthdr.len > IP_MAXPACKET) { m_freem(m); return(EMSGSIZE); } ip = mtod(m, struct ip *); if (jailed(inp->inp_socket->so_cred)) { if (ip->ip_src.s_addr != htonl(prison_getip(inp->inp_socket->so_cred))) { m_freem(m); return (EPERM); } } /* don't allow both user specified and setsockopt options, and don't allow packet length sizes that will crash */ if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) || (ip->ip_len > m->m_pkthdr.len) || (ip->ip_len < (ip->ip_hl << 2))) { m_freem(m); return EINVAL; } if (ip->ip_id == 0) #ifdef RANDOM_IP_ID ip->ip_id = ip_randomid(); #else ip->ip_id = htons(ip_id++); #endif /* XXX prevent ip_output from overwriting header fields */ flags |= IP_RAWOUTPUT; ipstat.ips_rawout++; } if (inp->inp_flags & INP_ONESBCAST) flags |= IP_SENDONES; return (ip_output(m, inp->inp_options, NULL, flags, inp->inp_moptions, inp)); } /* * Raw IP socket option processing. * * Note that access to all of the IP administrative functions here is * implicitly protected by suser() as gaining access to a raw socket * requires either that the thread pass a suser() check, or that it be * passed a raw socket by another thread that has passed a suser() check. * If FreeBSD moves to a more fine-grained access control mechanism, * additional checks will need to be placed here if the raw IP attachment * check is not equivilent the the check required for these * administrative operations; in some cases, these checks are already * present. */ int rip_ctloutput(struct socket *so, struct sockopt *sopt) { struct inpcb *inp = sotoinpcb(so); int error, optval; if (sopt->sopt_level != IPPROTO_IP) return (EINVAL); error = 0; switch (sopt->sopt_dir) { case SOPT_GET: switch (sopt->sopt_name) { case IP_HDRINCL: optval = inp->inp_flags & INP_HDRINCL; error = sooptcopyout(sopt, &optval, sizeof optval); break; case IP_FW_ADD: /* ADD actually returns the body... */ case IP_FW_GET: if (IPFW_LOADED) error = ip_fw_ctl_ptr(sopt); else error = ENOPROTOOPT; break; case IP_DUMMYNET_GET: if (DUMMYNET_LOADED) error = ip_dn_ctl_ptr(sopt); else error = ENOPROTOOPT; break ; case MRT_INIT: case MRT_DONE: case MRT_ADD_VIF: case MRT_DEL_VIF: case MRT_ADD_MFC: case MRT_DEL_MFC: case MRT_VERSION: case MRT_ASSERT: case MRT_API_SUPPORT: case MRT_API_CONFIG: case MRT_ADD_BW_UPCALL: case MRT_DEL_BW_UPCALL: error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : EOPNOTSUPP; break; default: error = ip_ctloutput(so, sopt); break; } break; case SOPT_SET: switch (sopt->sopt_name) { case IP_HDRINCL: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; if (optval) inp->inp_flags |= INP_HDRINCL; else inp->inp_flags &= ~INP_HDRINCL; break; case IP_FW_ADD: case IP_FW_DEL: case IP_FW_FLUSH: case IP_FW_ZERO: case IP_FW_RESETLOG: if (IPFW_LOADED) error = ip_fw_ctl_ptr(sopt); else error = ENOPROTOOPT; break; case IP_DUMMYNET_CONFIGURE: case IP_DUMMYNET_DEL: case IP_DUMMYNET_FLUSH: if (DUMMYNET_LOADED) error = ip_dn_ctl_ptr(sopt); else error = ENOPROTOOPT ; break ; case IP_RSVP_ON: error = ip_rsvp_init(so); break; case IP_RSVP_OFF: error = ip_rsvp_done(); break; case IP_RSVP_VIF_ON: case IP_RSVP_VIF_OFF: error = ip_rsvp_vif ? ip_rsvp_vif(so, sopt) : EINVAL; break; case MRT_INIT: case MRT_DONE: case MRT_ADD_VIF: case MRT_DEL_VIF: case MRT_ADD_MFC: case MRT_DEL_MFC: case MRT_VERSION: case MRT_ASSERT: case MRT_API_SUPPORT: case MRT_API_CONFIG: case MRT_ADD_BW_UPCALL: case MRT_DEL_BW_UPCALL: error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : EOPNOTSUPP; break; default: error = ip_ctloutput(so, sopt); break; } break; } return (error); } /* * This function exists solely to receive the PRC_IFDOWN messages which * are sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, * and calls in_ifadown() to remove all routes corresponding to that address. * It also receives the PRC_IFUP messages from if_up() and reinstalls the * interface routes. */ void rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct in_ifaddr *ia; struct ifnet *ifp; int err; int flags; switch (cmd) { case PRC_IFDOWN: TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { if (ia->ia_ifa.ifa_addr == sa && (ia->ia_flags & IFA_ROUTE)) { /* * in_ifscrub kills the interface route. */ in_ifscrub(ia->ia_ifp, ia); /* * in_ifadown gets rid of all the rest of * the routes. This is not quite the right * thing to do, but at least if we are running * a routing process they will come back. */ in_ifadown(&ia->ia_ifa, 0); break; } } break; case PRC_IFUP: TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { if (ia->ia_ifa.ifa_addr == sa) break; } if (ia == 0 || (ia->ia_flags & IFA_ROUTE)) return; flags = RTF_UP; ifp = ia->ia_ifa.ifa_ifp; if ((ifp->if_flags & IFF_LOOPBACK) || (ifp->if_flags & IFF_POINTOPOINT)) flags |= RTF_HOST; err = rtinit(&ia->ia_ifa, RTM_ADD, flags); if (err == 0) ia->ia_flags |= IFA_ROUTE; break; } } u_long rip_sendspace = RIPSNDQ; u_long rip_recvspace = RIPRCVQ; SYSCTL_INT(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); SYSCTL_INT(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); static int rip_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; int error; /* XXX why not lower? */ INP_INFO_WLOCK(&ripcbinfo); inp = sotoinpcb(so); if (inp) { /* XXX counter, printf */ INP_INFO_WUNLOCK(&ripcbinfo); return EINVAL; } if (td && jailed(td->td_ucred) && !jail_allow_raw_sockets) { INP_INFO_WUNLOCK(&ripcbinfo); return (EPERM); } if (td && (error = suser_cred(td->td_ucred, PRISON_ROOT)) != 0) { INP_INFO_WUNLOCK(&ripcbinfo); return error; } if (proto >= IPPROTO_MAX || proto < 0) { INP_INFO_WUNLOCK(&ripcbinfo); return EPROTONOSUPPORT; } error = soreserve(so, rip_sendspace, rip_recvspace); if (error) { INP_INFO_WUNLOCK(&ripcbinfo); return error; } error = in_pcballoc(so, &ripcbinfo, "rawinp"); if (error) { INP_INFO_WUNLOCK(&ripcbinfo); return error; } inp = (struct inpcb *)so->so_pcb; INP_LOCK(inp); INP_INFO_WUNLOCK(&ripcbinfo); inp->inp_vflag |= INP_IPV4; inp->inp_ip_p = proto; inp->inp_ip_ttl = ip_defttl; INP_UNLOCK(inp); return 0; } static void rip_pcbdetach(struct socket *so, struct inpcb *inp) { INP_INFO_WLOCK_ASSERT(&ripcbinfo); INP_LOCK_ASSERT(inp); if (so == ip_mrouter && ip_mrouter_done) ip_mrouter_done(); if (ip_rsvp_force_done) ip_rsvp_force_done(so); if (so == ip_rsvpd) ip_rsvp_done(); in_pcbdetach(inp); } static int rip_detach(struct socket *so) { struct inpcb *inp; INP_INFO_WLOCK(&ripcbinfo); inp = sotoinpcb(so); if (inp == 0) { /* XXX counter, printf */ INP_INFO_WUNLOCK(&ripcbinfo); return EINVAL; } INP_LOCK(inp); rip_pcbdetach(so, inp); INP_INFO_WUNLOCK(&ripcbinfo); return 0; } static int rip_abort(struct socket *so) { struct inpcb *inp; INP_INFO_WLOCK(&ripcbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_WUNLOCK(&ripcbinfo); return EINVAL; /* ??? possible? panic instead? */ } INP_LOCK(inp); soisdisconnected(so); if (so->so_state & SS_NOFDREF) rip_pcbdetach(so, inp); else INP_UNLOCK(inp); INP_INFO_WUNLOCK(&ripcbinfo); return 0; } static int rip_disconnect(struct socket *so) { if ((so->so_state & SS_ISCONNECTED) == 0) return ENOTCONN; return rip_abort(so); } static int rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct sockaddr_in *addr = (struct sockaddr_in *)nam; struct inpcb *inp; if (nam->sa_len != sizeof(*addr)) return EINVAL; if (jailed(td->td_ucred)) { if (addr->sin_addr.s_addr == INADDR_ANY) addr->sin_addr.s_addr = htonl(prison_getip(td->td_ucred)); if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr) return (EADDRNOTAVAIL); } if (TAILQ_EMPTY(&ifnet) || (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || (addr->sin_addr.s_addr && ifa_ifwithaddr((struct sockaddr *)addr) == 0)) return EADDRNOTAVAIL; INP_INFO_WLOCK(&ripcbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_WUNLOCK(&ripcbinfo); return EINVAL; } INP_LOCK(inp); inp->inp_laddr = addr->sin_addr; INP_UNLOCK(inp); INP_INFO_WUNLOCK(&ripcbinfo); return 0; } static int rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct sockaddr_in *addr = (struct sockaddr_in *)nam; struct inpcb *inp; if (nam->sa_len != sizeof(*addr)) return EINVAL; if (TAILQ_EMPTY(&ifnet)) return EADDRNOTAVAIL; if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) return EAFNOSUPPORT; INP_INFO_WLOCK(&ripcbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_WUNLOCK(&ripcbinfo); return EINVAL; } INP_LOCK(inp); inp->inp_faddr = addr->sin_addr; soisconnected(so); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&ripcbinfo); return 0; } static int rip_shutdown(struct socket *so) { struct inpcb *inp; INP_INFO_RLOCK(&ripcbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_RUNLOCK(&ripcbinfo); return EINVAL; } INP_LOCK(inp); INP_INFO_RUNLOCK(&ripcbinfo); socantsendmore(so); INP_UNLOCK(inp); return 0; } static int rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { struct inpcb *inp; u_long dst; int ret; INP_INFO_WLOCK(&ripcbinfo); inp = sotoinpcb(so); if (so->so_state & SS_ISCONNECTED) { if (nam) { INP_INFO_WUNLOCK(&ripcbinfo); m_freem(m); return EISCONN; } dst = inp->inp_faddr.s_addr; } else { if (nam == NULL) { INP_INFO_WUNLOCK(&ripcbinfo); m_freem(m); return ENOTCONN; } dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; } INP_LOCK(inp); ret = rip_output(m, so, dst); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&ripcbinfo); return ret; } static int rip_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == 0) { n = ripcbinfo.ipi_count; req->oldidx = 2 * (sizeof xig) + (n + n/8) * sizeof(struct xinpcb); return 0; } if (req->newptr != 0) return EPERM; /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&ripcbinfo); gencnt = ripcbinfo.ipi_gencnt; n = ripcbinfo.ipi_count; INP_INFO_RUNLOCK(&ripcbinfo); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return error; inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == 0) return ENOMEM; INP_INFO_RLOCK(&ripcbinfo); for (inp = LIST_FIRST(ripcbinfo.listhead), i = 0; inp && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_LOCK(inp); if (inp->inp_gencnt <= gencnt && cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) { /* XXX held references? */ inp_list[i++] = inp; } INP_UNLOCK(inp); } INP_INFO_RUNLOCK(&ripcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; if (inp->inp_gencnt <= gencnt) { struct xinpcb xi; xi.xi_len = sizeof xi; /* XXX should avoid extra copy */ bcopy(inp, &xi.xi_inp, sizeof *inp); if (inp->inp_socket) sotoxsocket(inp->inp_socket, &xi.xi_socket); error = SYSCTL_OUT(req, &xi, sizeof xi); } } if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ INP_INFO_RLOCK(&ripcbinfo); xig.xig_gen = ripcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = ripcbinfo.ipi_count; INP_INFO_RUNLOCK(&ripcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return error; } /* * This is the wrapper function for in_setsockaddr. We just pass down * the pcbinfo for in_setpeeraddr to lock. */ static int rip_sockaddr(struct socket *so, struct sockaddr **nam) { return (in_setsockaddr(so, nam, &ripcbinfo)); } /* * This is the wrapper function for in_setpeeraddr. We just pass down * the pcbinfo for in_setpeeraddr to lock. */ static int rip_peeraddr(struct socket *so, struct sockaddr **nam) { return (in_setpeeraddr(so, nam, &ripcbinfo)); } SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0, rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); struct pr_usrreqs rip_usrreqs = { rip_abort, pru_accept_notsupp, rip_attach, rip_bind, rip_connect, pru_connect2_notsupp, in_control, rip_detach, rip_disconnect, pru_listen_notsupp, rip_peeraddr, pru_rcvd_notsupp, pru_rcvoob_notsupp, rip_send, pru_sense_null, rip_shutdown, rip_sockaddr, sosend, soreceive, sopoll, in_pcbsosetlabel }; diff --git a/sys/netinet/tcp_output.c b/sys/netinet/tcp_output.c index 098322ead9c4..46d9337caf03 100644 --- a/sys/netinet/tcp_output.c +++ b/sys/netinet/tcp_output.c @@ -1,1025 +1,1025 @@ /* * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_output.c 8.4 (Berkeley) 5/24/95 * $FreeBSD$ */ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_mac.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #endif #include #define TCPOUTFLAGS #include #include #include #include #include #ifdef TCPDEBUG #include #endif #ifdef IPSEC #include #endif /*IPSEC*/ #ifdef FAST_IPSEC #include #define IPSEC #endif /*FAST_IPSEC*/ #include #ifdef notyet extern struct mbuf *m_copypack(); #endif int path_mtu_discovery = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, path_mtu_discovery, CTLFLAG_RW, &path_mtu_discovery, 1, "Enable Path MTU Discovery"); int ss_fltsz = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, slowstart_flightsize, CTLFLAG_RW, &ss_fltsz, 1, "Slow start flight size"); int ss_fltsz_local = 4; SYSCTL_INT(_net_inet_tcp, OID_AUTO, local_slowstart_flightsize, CTLFLAG_RW, &ss_fltsz_local, 1, "Slow start flight size for local networks"); int tcp_do_newreno = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, newreno, CTLFLAG_RW, &tcp_do_newreno, 0, "Enable NewReno Algorithms"); /* * Tcp output routine: figure out what should be sent and send it. */ int tcp_output(struct tcpcb *tp) { struct socket *so = tp->t_inpcb->inp_socket; long len, recwin, sendwin; int off, flags, error; #ifdef TCP_SIGNATURE int sigoff = 0; #endif struct mbuf *m; struct ip *ip = NULL; struct ipovly *ipov = NULL; struct tcphdr *th; u_char opt[TCP_MAXOLEN]; unsigned ipoptlen, optlen, hdrlen; int idle, sendalot; #if 0 int maxburst = TCP_MAXBURST; #endif struct rmxp_tao tao; #ifdef INET6 struct ip6_hdr *ip6 = NULL; int isipv6; bzero(&tao, sizeof(tao)); isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0; #endif INP_LOCK_ASSERT(tp->t_inpcb); /* * Determine length of data that should be transmitted, * and flags that will be used. * If there is some data or critical controls (SYN, RST) * to send, then transmit; otherwise, investigate further. */ idle = (tp->t_flags & TF_LASTIDLE) || (tp->snd_max == tp->snd_una); if (idle && (ticks - tp->t_rcvtime) >= tp->t_rxtcur) { /* * We have been idle for "a while" and no acks are * expected to clock out any data we send -- * slow start to get ack "clock" running again. * * Set the slow-start flight size depending on whether * this is a local network or not. */ int ss = ss_fltsz; #ifdef INET6 if (isipv6) { if (in6_localaddr(&tp->t_inpcb->in6p_faddr)) ss = ss_fltsz_local; } else #endif /* INET6 */ if (in_localaddr(tp->t_inpcb->inp_faddr)) ss = ss_fltsz_local; tp->snd_cwnd = tp->t_maxseg * ss; } tp->t_flags &= ~TF_LASTIDLE; if (idle) { if (tp->t_flags & TF_MORETOCOME) { tp->t_flags |= TF_LASTIDLE; idle = 0; } } again: sendalot = 0; off = tp->snd_nxt - tp->snd_una; sendwin = min(tp->snd_wnd, tp->snd_cwnd); sendwin = min(sendwin, tp->snd_bwnd); flags = tcp_outflags[tp->t_state]; /* * Get standard flags, and add SYN or FIN if requested by 'hidden' * state flags. */ if (tp->t_flags & TF_NEEDFIN) flags |= TH_FIN; if (tp->t_flags & TF_NEEDSYN) flags |= TH_SYN; /* * If in persist timeout with window of 0, send 1 byte. * Otherwise, if window is small but nonzero * and timer expired, we will send what we can * and go to transmit state. */ if (tp->t_force) { if (sendwin == 0) { /* * If we still have some data to send, then * clear the FIN bit. Usually this would * happen below when it realizes that we * aren't sending all the data. However, * if we have exactly 1 byte of unsent data, * then it won't clear the FIN bit below, * and if we are in persist state, we wind * up sending the packet without recording * that we sent the FIN bit. * * We can't just blindly clear the FIN bit, * because if we don't have any more data * to send then the probe will be the FIN * itself. */ if (off < so->so_snd.sb_cc) flags &= ~TH_FIN; sendwin = 1; } else { callout_stop(tp->tt_persist); tp->t_rxtshift = 0; } } /* * If snd_nxt == snd_max and we have transmitted a FIN, the * offset will be > 0 even if so_snd.sb_cc is 0, resulting in * a negative length. This can also occur when TCP opens up * its congestion window while receiving additional duplicate * acks after fast-retransmit because TCP will reset snd_nxt * to snd_max after the fast-retransmit. * * In the normal retransmit-FIN-only case, however, snd_nxt will * be set to snd_una, the offset will be 0, and the length may * wind up 0. */ len = (long)ulmin(so->so_snd.sb_cc, sendwin) - off; /* * Lop off SYN bit if it has already been sent. However, if this * is SYN-SENT state and if segment contains data and if we don't * know that foreign host supports TAO, suppress sending segment. */ if ((flags & TH_SYN) && SEQ_GT(tp->snd_nxt, tp->snd_una)) { flags &= ~TH_SYN; off--, len++; if (tcp_do_rfc1644) tcp_hc_gettao(&tp->t_inpcb->inp_inc, &tao); if (len > 0 && tp->t_state == TCPS_SYN_SENT && tao.tao_ccsent == 0) return 0; } /* * Be careful not to send data and/or FIN on SYN segments * in cases when no CC option will be sent. * This measure is needed to prevent interoperability problems * with not fully conformant TCP implementations. */ if ((flags & TH_SYN) && ((tp->t_flags & TF_NOOPT) || !(tp->t_flags & TF_REQ_CC) || ((flags & TH_ACK) && !(tp->t_flags & TF_RCVD_CC)))) { len = 0; flags &= ~TH_FIN; } if (len < 0) { /* * If FIN has been sent but not acked, * but we haven't been called to retransmit, * len will be < 0. Otherwise, window shrank * after we sent into it. If window shrank to 0, * cancel pending retransmit, pull snd_nxt back * to (closed) window, and set the persist timer * if it isn't already going. If the window didn't * close completely, just wait for an ACK. */ len = 0; if (sendwin == 0) { callout_stop(tp->tt_rexmt); tp->t_rxtshift = 0; tp->snd_nxt = tp->snd_una; if (!callout_active(tp->tt_persist)) tcp_setpersist(tp); } } /* * len will be >= 0 after this point. Truncate to the maximum * segment length and ensure that FIN is removed if the length * no longer contains the last data byte. */ if (len > tp->t_maxseg) { len = tp->t_maxseg; sendalot = 1; } if (SEQ_LT(tp->snd_nxt + len, tp->snd_una + so->so_snd.sb_cc)) flags &= ~TH_FIN; recwin = sbspace(&so->so_rcv); /* * Sender silly window avoidance. We transmit under the following * conditions when len is non-zero: * * - We have a full segment * - This is the last buffer in a write()/send() and we are * either idle or running NODELAY * - we've timed out (e.g. persist timer) * - we have more then 1/2 the maximum send window's worth of * data (receiver may be limited the window size) * - we need to retransmit */ if (len) { if (len == tp->t_maxseg) goto send; /* * NOTE! on localhost connections an 'ack' from the remote * end may occur synchronously with the output and cause * us to flush a buffer queued with moretocome. XXX * * note: the len + off check is almost certainly unnecessary. */ if (!(tp->t_flags & TF_MORETOCOME) && /* normal case */ (idle || (tp->t_flags & TF_NODELAY)) && len + off >= so->so_snd.sb_cc && (tp->t_flags & TF_NOPUSH) == 0) { goto send; } if (tp->t_force) /* typ. timeout case */ goto send; if (len >= tp->max_sndwnd / 2 && tp->max_sndwnd > 0) goto send; if (SEQ_LT(tp->snd_nxt, tp->snd_max)) /* retransmit case */ goto send; } /* * Compare available window to amount of window * known to peer (as advertised window less * next expected input). If the difference is at least two * max size segments, or at least 50% of the maximum possible * window, then want to send a window update to peer. * Skip this if the connection is in T/TCP half-open state. */ if (recwin > 0 && !(tp->t_flags & TF_NEEDSYN)) { /* * "adv" is the amount we can increase the window, * taking into account that we are limited by * TCP_MAXWIN << tp->rcv_scale. */ long adv = min(recwin, (long)TCP_MAXWIN << tp->rcv_scale) - (tp->rcv_adv - tp->rcv_nxt); if (adv >= (long) (2 * tp->t_maxseg)) goto send; if (2 * adv >= (long) so->so_rcv.sb_hiwat) goto send; } /* * Send if we owe the peer an ACK, RST, SYN, or urgent data. ACKNOW * is also a catch-all for the retransmit timer timeout case. */ if (tp->t_flags & TF_ACKNOW) goto send; if ((flags & TH_RST) || ((flags & TH_SYN) && (tp->t_flags & TF_NEEDSYN) == 0)) goto send; if (SEQ_GT(tp->snd_up, tp->snd_una)) goto send; /* * If our state indicates that FIN should be sent * and we have not yet done so, then we need to send. */ if (flags & TH_FIN && ((tp->t_flags & TF_SENTFIN) == 0 || tp->snd_nxt == tp->snd_una)) goto send; /* * TCP window updates are not reliable, rather a polling protocol * using ``persist'' packets is used to insure receipt of window * updates. The three ``states'' for the output side are: * idle not doing retransmits or persists * persisting to move a small or zero window * (re)transmitting and thereby not persisting * * callout_active(tp->tt_persist) * is true when we are in persist state. * tp->t_force * is set when we are called to send a persist packet. * callout_active(tp->tt_rexmt) * is set when we are retransmitting * The output side is idle when both timers are zero. * * If send window is too small, there is data to transmit, and no * retransmit or persist is pending, then go to persist state. * If nothing happens soon, send when timer expires: * if window is nonzero, transmit what we can, * otherwise force out a byte. */ if (so->so_snd.sb_cc && !callout_active(tp->tt_rexmt) && !callout_active(tp->tt_persist)) { tp->t_rxtshift = 0; tcp_setpersist(tp); } /* * No reason to send a segment, just return. */ return (0); send: /* * Before ESTABLISHED, force sending of initial options * unless TCP set not to do any options. * NOTE: we assume that the IP/TCP header plus TCP options * always fit in a single mbuf, leaving room for a maximum * link header, i.e. * max_linkhdr + sizeof (struct tcpiphdr) + optlen <= MCLBYTES */ optlen = 0; #ifdef INET6 if (isipv6) hdrlen = sizeof (struct ip6_hdr) + sizeof (struct tcphdr); else #endif hdrlen = sizeof (struct tcpiphdr); if (flags & TH_SYN) { tp->snd_nxt = tp->iss; if ((tp->t_flags & TF_NOOPT) == 0) { u_short mss; opt[0] = TCPOPT_MAXSEG; opt[1] = TCPOLEN_MAXSEG; mss = htons((u_short) tcp_mssopt(&tp->t_inpcb->inp_inc)); (void)memcpy(opt + 2, &mss, sizeof(mss)); optlen = TCPOLEN_MAXSEG; if ((tp->t_flags & TF_REQ_SCALE) && ((flags & TH_ACK) == 0 || (tp->t_flags & TF_RCVD_SCALE))) { *((u_int32_t *)(opt + optlen)) = htonl( TCPOPT_NOP << 24 | TCPOPT_WINDOW << 16 | TCPOLEN_WINDOW << 8 | tp->request_r_scale); optlen += 4; } } } /* * Send a timestamp and echo-reply if this is a SYN and our side * wants to use timestamps (TF_REQ_TSTMP is set) or both our side * and our peer have sent timestamps in our SYN's. */ if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && (flags & TH_RST) == 0 && ((flags & TH_ACK) == 0 || (tp->t_flags & TF_RCVD_TSTMP))) { u_int32_t *lp = (u_int32_t *)(opt + optlen); /* Form timestamp option as shown in appendix A of RFC 1323. */ *lp++ = htonl(TCPOPT_TSTAMP_HDR); *lp++ = htonl(ticks); *lp = htonl(tp->ts_recent); optlen += TCPOLEN_TSTAMP_APPA; } /* * Send `CC-family' options if our side wants to use them (TF_REQ_CC), * options are allowed (!TF_NOOPT) and it's not a RST. */ if ((tp->t_flags & (TF_REQ_CC|TF_NOOPT)) == TF_REQ_CC && (flags & TH_RST) == 0) { switch (flags & (TH_SYN|TH_ACK)) { /* * This is a normal ACK, send CC if we received CC before * from our peer. */ case TH_ACK: if (!(tp->t_flags & TF_RCVD_CC)) break; /*FALLTHROUGH*/ /* * We can only get here in T/TCP's SYN_SENT* state, when * we're a sending a non-SYN segment without waiting for * the ACK of our SYN. A check above assures that we only * do this if our peer understands T/TCP. */ case 0: opt[optlen++] = TCPOPT_NOP; opt[optlen++] = TCPOPT_NOP; opt[optlen++] = TCPOPT_CC; opt[optlen++] = TCPOLEN_CC; *(u_int32_t *)&opt[optlen] = htonl(tp->cc_send); optlen += 4; break; /* * This is our initial SYN, check whether we have to use * CC or CC.new. */ case TH_SYN: opt[optlen++] = TCPOPT_NOP; opt[optlen++] = TCPOPT_NOP; opt[optlen++] = tp->t_flags & TF_SENDCCNEW ? TCPOPT_CCNEW : TCPOPT_CC; opt[optlen++] = TCPOLEN_CC; *(u_int32_t *)&opt[optlen] = htonl(tp->cc_send); optlen += 4; break; /* * This is a SYN,ACK; send CC and CC.echo if we received * CC from our peer. */ case (TH_SYN|TH_ACK): if (tp->t_flags & TF_RCVD_CC) { opt[optlen++] = TCPOPT_NOP; opt[optlen++] = TCPOPT_NOP; opt[optlen++] = TCPOPT_CC; opt[optlen++] = TCPOLEN_CC; *(u_int32_t *)&opt[optlen] = htonl(tp->cc_send); optlen += 4; opt[optlen++] = TCPOPT_NOP; opt[optlen++] = TCPOPT_NOP; opt[optlen++] = TCPOPT_CCECHO; opt[optlen++] = TCPOLEN_CC; *(u_int32_t *)&opt[optlen] = htonl(tp->cc_recv); optlen += 4; } break; } } #ifdef TCP_SIGNATURE #ifdef INET6 if (!isipv6) #endif if (tp->t_flags & TF_SIGNATURE) { int i; u_char *bp; /* Initialize TCP-MD5 option (RFC2385) */ bp = (u_char *)opt + optlen; *bp++ = TCPOPT_SIGNATURE; *bp++ = TCPOLEN_SIGNATURE; sigoff = optlen + 2; for (i = 0; i < TCP_SIGLEN; i++) *bp++ = 0; optlen += TCPOLEN_SIGNATURE; /* Terminate options list and maintain 32-bit alignment. */ *bp++ = TCPOPT_NOP; *bp++ = TCPOPT_EOL; optlen += 2; } #endif /* TCP_SIGNATURE */ hdrlen += optlen; #ifdef INET6 if (isipv6) ipoptlen = ip6_optlen(tp->t_inpcb); else #endif if (tp->t_inpcb->inp_options) ipoptlen = tp->t_inpcb->inp_options->m_len - offsetof(struct ipoption, ipopt_list); else ipoptlen = 0; #ifdef IPSEC ipoptlen += ipsec_hdrsiz_tcp(tp); #endif /* * Adjust data length if insertion of options will * bump the packet length beyond the t_maxopd length. * Clear the FIN bit because we cut off the tail of * the segment. */ if (len + optlen + ipoptlen > tp->t_maxopd) { /* * If there is still more to send, don't close the connection. */ flags &= ~TH_FIN; len = tp->t_maxopd - optlen - ipoptlen; sendalot = 1; } /*#ifdef DIAGNOSTIC*/ #ifdef INET6 if (max_linkhdr + hdrlen > MCLBYTES) #else if (max_linkhdr + hdrlen > MHLEN) #endif panic("tcphdr too big"); /*#endif*/ /* * Grab a header mbuf, attaching a copy of data to * be transmitted, and initialize the header from * the template for sends on this connection. */ if (len) { if (tp->t_force && len == 1) tcpstat.tcps_sndprobe++; else if (SEQ_LT(tp->snd_nxt, tp->snd_max)) { tcpstat.tcps_sndrexmitpack++; tcpstat.tcps_sndrexmitbyte += len; } else { tcpstat.tcps_sndpack++; tcpstat.tcps_sndbyte += len; } #ifdef notyet if ((m = m_copypack(so->so_snd.sb_mb, off, (int)len, max_linkhdr + hdrlen)) == 0) { error = ENOBUFS; goto out; } /* * m_copypack left space for our hdr; use it. */ m->m_len += hdrlen; m->m_data -= hdrlen; #else MGETHDR(m, M_DONTWAIT, MT_HEADER); if (m == NULL) { error = ENOBUFS; goto out; } #ifdef INET6 if (MHLEN < hdrlen + max_linkhdr) { MCLGET(m, M_DONTWAIT); if ((m->m_flags & M_EXT) == 0) { m_freem(m); error = ENOBUFS; goto out; } } #endif m->m_data += max_linkhdr; m->m_len = hdrlen; if (len <= MHLEN - hdrlen - max_linkhdr) { m_copydata(so->so_snd.sb_mb, off, (int) len, mtod(m, caddr_t) + hdrlen); m->m_len += len; } else { m->m_next = m_copy(so->so_snd.sb_mb, off, (int) len); if (m->m_next == 0) { (void) m_free(m); error = ENOBUFS; goto out; } } #endif /* * If we're sending everything we've got, set PUSH. * (This will keep happy those implementations which only * give data to the user when a buffer fills or * a PUSH comes in.) */ if (off + len == so->so_snd.sb_cc) flags |= TH_PUSH; } else { if (tp->t_flags & TF_ACKNOW) tcpstat.tcps_sndacks++; else if (flags & (TH_SYN|TH_FIN|TH_RST)) tcpstat.tcps_sndctrl++; else if (SEQ_GT(tp->snd_up, tp->snd_una)) tcpstat.tcps_sndurg++; else tcpstat.tcps_sndwinup++; MGETHDR(m, M_DONTWAIT, MT_HEADER); if (m == NULL) { error = ENOBUFS; goto out; } #ifdef INET6 if (isipv6 && (MHLEN < hdrlen + max_linkhdr) && MHLEN >= hdrlen) { MH_ALIGN(m, hdrlen); } else #endif m->m_data += max_linkhdr; m->m_len = hdrlen; } m->m_pkthdr.rcvif = (struct ifnet *)0; #ifdef MAC - mac_create_mbuf_from_socket(so, m); + mac_create_mbuf_from_inpcb(tp->t_inpcb, m); #endif #ifdef INET6 if (isipv6) { ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); tcpip_fillheaders(tp->t_inpcb, ip6, th); } else #endif /* INET6 */ { ip = mtod(m, struct ip *); ipov = (struct ipovly *)ip; th = (struct tcphdr *)(ip + 1); tcpip_fillheaders(tp->t_inpcb, ip, th); } /* * Fill in fields, remembering maximum advertised * window for use in delaying messages about window sizes. * If resending a FIN, be sure not to use a new sequence number. */ if (flags & TH_FIN && tp->t_flags & TF_SENTFIN && tp->snd_nxt == tp->snd_max) tp->snd_nxt--; /* * If we are doing retransmissions, then snd_nxt will * not reflect the first unsent octet. For ACK only * packets, we do not want the sequence number of the * retransmitted packet, we want the sequence number * of the next unsent octet. So, if there is no data * (and no SYN or FIN), use snd_max instead of snd_nxt * when filling in ti_seq. But if we are in persist * state, snd_max might reflect one byte beyond the * right edge of the window, so use snd_nxt in that * case, since we know we aren't doing a retransmission. * (retransmit and persist are mutually exclusive...) */ if (len || (flags & (TH_SYN|TH_FIN)) || callout_active(tp->tt_persist)) th->th_seq = htonl(tp->snd_nxt); else th->th_seq = htonl(tp->snd_max); th->th_ack = htonl(tp->rcv_nxt); if (optlen) { bcopy(opt, th + 1, optlen); th->th_off = (sizeof (struct tcphdr) + optlen) >> 2; } th->th_flags = flags; /* * Calculate receive window. Don't shrink window, * but avoid silly window syndrome. */ if (recwin < (long)(so->so_rcv.sb_hiwat / 4) && recwin < (long)tp->t_maxseg) recwin = 0; if (recwin < (long)(tp->rcv_adv - tp->rcv_nxt)) recwin = (long)(tp->rcv_adv - tp->rcv_nxt); if (recwin > (long)TCP_MAXWIN << tp->rcv_scale) recwin = (long)TCP_MAXWIN << tp->rcv_scale; th->th_win = htons((u_short) (recwin >> tp->rcv_scale)); /* * Adjust the RXWIN0SENT flag - indicate that we have advertised * a 0 window. This may cause the remote transmitter to stall. This * flag tells soreceive() to disable delayed acknowledgements when * draining the buffer. This can occur if the receiver is attempting * to read more data then can be buffered prior to transmitting on * the connection. */ if (recwin == 0) tp->t_flags |= TF_RXWIN0SENT; else tp->t_flags &= ~TF_RXWIN0SENT; if (SEQ_GT(tp->snd_up, tp->snd_nxt)) { th->th_urp = htons((u_short)(tp->snd_up - tp->snd_nxt)); th->th_flags |= TH_URG; } else /* * If no urgent pointer to send, then we pull * the urgent pointer to the left edge of the send window * so that it doesn't drift into the send window on sequence * number wraparound. */ tp->snd_up = tp->snd_una; /* drag it along */ #ifdef TCP_SIGNATURE #ifdef INET6 if (!isipv6) #endif if (tp->t_flags & TF_SIGNATURE) tcp_signature_compute(m, sizeof(struct ip), len, optlen, (u_char *)(th + 1) + sigoff, IPSEC_DIR_OUTBOUND); #endif /* * Put TCP length in extended header, and then * checksum extended header and data. */ m->m_pkthdr.len = hdrlen + len; /* in6_cksum() need this */ #ifdef INET6 if (isipv6) /* * ip6_plen is not need to be filled now, and will be filled * in ip6_output. */ th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), sizeof(struct tcphdr) + optlen + len); else #endif /* INET6 */ { m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons(sizeof(struct tcphdr) + IPPROTO_TCP + len + optlen)); /* IP version must be set here for ipv4/ipv6 checking later */ KASSERT(ip->ip_v == IPVERSION, ("%s: IP version incorrect: %d", __func__, ip->ip_v)); } /* * In transmit state, time the transmission and arrange for * the retransmit. In persist state, just set snd_max. */ if (tp->t_force == 0 || !callout_active(tp->tt_persist)) { tcp_seq startseq = tp->snd_nxt; /* * Advance snd_nxt over sequence space of this segment. */ if (flags & (TH_SYN|TH_FIN)) { if (flags & TH_SYN) tp->snd_nxt++; if (flags & TH_FIN) { tp->snd_nxt++; tp->t_flags |= TF_SENTFIN; } } tp->snd_nxt += len; if (SEQ_GT(tp->snd_nxt, tp->snd_max)) { tp->snd_max = tp->snd_nxt; /* * Time this transmission if not a retransmission and * not currently timing anything. */ if (tp->t_rtttime == 0) { tp->t_rtttime = ticks; tp->t_rtseq = startseq; tcpstat.tcps_segstimed++; } } /* * Set retransmit timer if not currently set, * and not doing a pure ack or a keep-alive probe. * Initial value for retransmit timer is smoothed * round-trip time + 2 * round-trip time variance. * Initialize shift counter which is used for backoff * of retransmit time. */ if (!callout_active(tp->tt_rexmt) && tp->snd_nxt != tp->snd_una) { if (callout_active(tp->tt_persist)) { callout_stop(tp->tt_persist); tp->t_rxtshift = 0; } callout_reset(tp->tt_rexmt, tp->t_rxtcur, tcp_timer_rexmt, tp); } } else { /* * Persist case, update snd_max but since we are in * persist mode (no window) we do not update snd_nxt. */ int xlen = len; if (flags & TH_SYN) ++xlen; if (flags & TH_FIN) { ++xlen; tp->t_flags |= TF_SENTFIN; } if (SEQ_GT(tp->snd_nxt + xlen, tp->snd_max)) tp->snd_max = tp->snd_nxt + len; } #ifdef TCPDEBUG /* * Trace. */ if (so->so_options & SO_DEBUG) { u_short save = ipov->ih_len; ipov->ih_len = htons(m->m_pkthdr.len /* - hdrlen + (th->th_off << 2) */); tcp_trace(TA_OUTPUT, tp->t_state, tp, mtod(m, void *), th, 0); ipov->ih_len = save; } #endif /* * Fill in IP length and desired time to live and * send to IP level. There should be a better way * to handle ttl and tos; we could keep them in * the template, but need a way to checksum without them. */ /* * m->m_pkthdr.len should have been set before cksum calcuration, * because in6_cksum() need it. */ #ifdef INET6 if (isipv6) { /* * we separately set hoplimit for every segment, since the * user might want to change the value via setsockopt. * Also, desired default hop limit might be changed via * Neighbor Discovery. */ ip6->ip6_hlim = in6_selecthlim(tp->t_inpcb, NULL); /* TODO: IPv6 IP6TOS_ECT bit on */ error = ip6_output(m, tp->t_inpcb->in6p_outputopts, NULL, (so->so_options & SO_DONTROUTE), NULL, NULL, tp->t_inpcb); } else #endif /* INET6 */ { ip->ip_len = m->m_pkthdr.len; #ifdef INET6 if (INP_CHECK_SOCKAF(so, AF_INET6)) ip->ip_ttl = in6_selecthlim(tp->t_inpcb, NULL); #endif /* INET6 */ /* * If we do path MTU discovery, then we set DF on every packet. * This might not be the best thing to do according to RFC3390 * Section 2. However the tcp hostcache migitates the problem * so it affects only the first tcp connection with a host. */ if (path_mtu_discovery) ip->ip_off |= IP_DF; error = ip_output(m, tp->t_inpcb->inp_options, NULL, (so->so_options & SO_DONTROUTE), 0, tp->t_inpcb); } if (error) { /* * We know that the packet was lost, so back out the * sequence number advance, if any. */ if (tp->t_force == 0 || !callout_active(tp->tt_persist)) { /* * No need to check for TH_FIN here because * the TF_SENTFIN flag handles that case. */ if ((flags & TH_SYN) == 0) tp->snd_nxt -= len; } out: if (error == ENOBUFS) { if (!callout_active(tp->tt_rexmt) && !callout_active(tp->tt_persist)) callout_reset(tp->tt_rexmt, tp->t_rxtcur, tcp_timer_rexmt, tp); tcp_quench(tp->t_inpcb, 0); return (0); } if (error == EMSGSIZE) { /* * ip_output() will have already fixed the route * for us. tcp_mtudisc() will, as its last action, * initiate retransmission, so it is important to * not do so here. */ tcp_mtudisc(tp->t_inpcb, 0); return 0; } if ((error == EHOSTUNREACH || error == ENETDOWN) && TCPS_HAVERCVDSYN(tp->t_state)) { tp->t_softerror = error; return (0); } return (error); } tcpstat.tcps_sndtotal++; /* * Data sent (as far as we can tell). * If this advertises a larger window than any other segment, * then remember the size of the advertised window. * Any pending ACK has now been sent. */ if (recwin > 0 && SEQ_GT(tp->rcv_nxt + recwin, tp->rcv_adv)) tp->rcv_adv = tp->rcv_nxt + recwin; tp->last_ack_sent = tp->rcv_nxt; tp->t_flags &= ~(TF_ACKNOW | TF_DELACK); if (callout_active(tp->tt_delack)) callout_stop(tp->tt_delack); #if 0 /* * This completely breaks TCP if newreno is turned on. What happens * is that if delayed-acks are turned on on the receiver, this code * on the transmitter effectively destroys the TCP window, forcing * it to four packets (1.5Kx4 = 6K window). */ if (sendalot && (!tcp_do_newreno || --maxburst)) goto again; #endif if (sendalot) goto again; return (0); } void tcp_setpersist(tp) register struct tcpcb *tp; { int t = ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1; int tt; if (callout_active(tp->tt_rexmt)) panic("tcp_setpersist: retransmit pending"); /* * Start/restart persistance timer. */ TCPT_RANGESET(tt, t * tcp_backoff[tp->t_rxtshift], TCPTV_PERSMIN, TCPTV_PERSMAX); callout_reset(tp->tt_persist, tt, tcp_timer_persist, tp); if (tp->t_rxtshift < TCP_MAXRXTSHIFT) tp->t_rxtshift++; } diff --git a/sys/netinet/tcp_subr.c b/sys/netinet/tcp_subr.c index 23f08bafc9c3..5fca04c85115 100644 --- a/sys/netinet/tcp_subr.c +++ b/sys/netinet/tcp_subr.c @@ -1,2091 +1,2096 @@ /* * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 * $FreeBSD$ */ #include "opt_compat.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_mac.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef INET6 #include #endif #include #include #ifdef INET6 #include #include #endif #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef TCPDEBUG #include #endif #include #ifdef IPSEC #include #ifdef INET6 #include #endif #endif /*IPSEC*/ #ifdef FAST_IPSEC #include #include #ifdef INET6 #include #endif #include #define IPSEC #endif /*FAST_IPSEC*/ #include #include int tcp_mssdflt = TCP_MSS; SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW, &tcp_mssdflt , 0, "Default TCP Maximum Segment Size"); #ifdef INET6 int tcp_v6mssdflt = TCP6_MSS; SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, CTLFLAG_RW, &tcp_v6mssdflt , 0, "Default TCP Maximum Segment Size for IPv6"); #endif /* * Minimum MSS we accept and use. This prevents DoS attacks where * we are forced to a ridiculous low MSS like 20 and send hundreds * of packets instead of one. The effect scales with the available * bandwidth and quickly saturates the CPU and network interface * with packet generation and sending. Set to zero to disable MINMSS * checking. This setting prevents us from sending too small packets. */ int tcp_minmss = TCP_MINMSS; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_RW, &tcp_minmss , 0, "Minmum TCP Maximum Segment Size"); /* * Number of TCP segments per second we accept from remote host * before we start to calculate average segment size. If average * segment size drops below the minimum TCP MSS we assume a DoS * attack and reset+drop the connection. Care has to be taken not to * set this value too small to not kill interactive type connections * (telnet, SSH) which send many small packets. */ int tcp_minmssoverload = TCP_MINMSSOVERLOAD; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmssoverload, CTLFLAG_RW, &tcp_minmssoverload , 0, "Number of TCP Segments per Second allowed to" "be under the MINMSS Size"); #if 0 static int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; SYSCTL_INT(_net_inet_tcp, TCPCTL_RTTDFLT, rttdflt, CTLFLAG_RW, &tcp_rttdflt , 0, "Default maximum TCP Round Trip Time"); #endif int tcp_do_rfc1323 = 1; SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW, &tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions"); int tcp_do_rfc1644 = 0; SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1644, rfc1644, CTLFLAG_RW, &tcp_do_rfc1644 , 0, "Enable rfc1644 (TTCP) extensions"); static int tcp_tcbhashsize = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RDTUN, &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); static int do_tcpdrain = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, "Enable tcp_drain routine for extra help when low on mbufs"); SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD, &tcbinfo.ipi_count, 0, "Number of active PCBs"); static int icmp_may_rst = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW, &icmp_may_rst, 0, "Certain ICMP unreachable messages may abort connections in SYN_SENT"); static int tcp_isn_reseed_interval = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_RW, &tcp_isn_reseed_interval, 0, "Seconds between reseeding of ISN secret"); /* * TCP bandwidth limiting sysctls. Note that the default lower bound of * 1024 exists only for debugging. A good production default would be * something like 6100. */ static int tcp_inflight_enable = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, inflight_enable, CTLFLAG_RW, &tcp_inflight_enable, 0, "Enable automatic TCP inflight data limiting"); static int tcp_inflight_debug = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, inflight_debug, CTLFLAG_RW, &tcp_inflight_debug, 0, "Debug TCP inflight calculations"); static int tcp_inflight_min = 6144; SYSCTL_INT(_net_inet_tcp, OID_AUTO, inflight_min, CTLFLAG_RW, &tcp_inflight_min, 0, "Lower-bound for TCP inflight window"); static int tcp_inflight_max = TCP_MAXWIN << TCP_MAX_WINSHIFT; SYSCTL_INT(_net_inet_tcp, OID_AUTO, inflight_max, CTLFLAG_RW, &tcp_inflight_max, 0, "Upper-bound for TCP inflight window"); static int tcp_inflight_stab = 20; SYSCTL_INT(_net_inet_tcp, OID_AUTO, inflight_stab, CTLFLAG_RW, &tcp_inflight_stab, 0, "Inflight Algorithm Stabilization 20 = 2 packets"); static struct inpcb *tcp_notify(struct inpcb *, int); static void tcp_discardcb(struct tcpcb *); static void tcp_isn_tick(void *); /* * Target size of TCP PCB hash tables. Must be a power of two. * * Note that this can be overridden by the kernel environment * variable net.inet.tcp.tcbhashsize */ #ifndef TCBHASHSIZE #define TCBHASHSIZE 512 #endif /* * XXX * Callouts should be moved into struct tcp directly. They are currently * separate because the tcpcb structure is exported to userland for sysctl * parsing purposes, which do not know about callouts. */ struct tcpcb_mem { struct tcpcb tcb; struct callout tcpcb_mem_rexmt, tcpcb_mem_persist, tcpcb_mem_keep; struct callout tcpcb_mem_2msl, tcpcb_mem_delack; }; static uma_zone_t tcpcb_zone; static uma_zone_t tcptw_zone; struct callout isn_callout; /* * Tcp initialization */ void tcp_init() { int hashsize = TCBHASHSIZE; tcp_ccgen = 1; tcp_delacktime = TCPTV_DELACK; tcp_keepinit = TCPTV_KEEP_INIT; tcp_keepidle = TCPTV_KEEP_IDLE; tcp_keepintvl = TCPTV_KEEPINTVL; tcp_maxpersistidle = TCPTV_KEEP_IDLE; tcp_msl = TCPTV_MSL; tcp_rexmit_min = TCPTV_MIN; tcp_rexmit_slop = TCPTV_CPU_VAR; INP_INFO_LOCK_INIT(&tcbinfo, "tcp"); LIST_INIT(&tcb); tcbinfo.listhead = &tcb; TUNABLE_INT_FETCH("net.inet.tcp.tcbhashsize", &hashsize); if (!powerof2(hashsize)) { printf("WARNING: TCB hash size not a power of 2\n"); hashsize = 512; /* safe default */ } tcp_tcbhashsize = hashsize; tcbinfo.hashbase = hashinit(hashsize, M_PCB, &tcbinfo.hashmask); tcbinfo.porthashbase = hashinit(hashsize, M_PCB, &tcbinfo.porthashmask); tcbinfo.ipi_zone = uma_zcreate("inpcb", sizeof(struct inpcb), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcbinfo.ipi_zone, maxsockets); #ifdef INET6 #define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) #else /* INET6 */ #define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) #endif /* INET6 */ if (max_protohdr < TCP_MINPROTOHDR) max_protohdr = TCP_MINPROTOHDR; if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) panic("tcp_init"); #undef TCP_MINPROTOHDR /* * These have to be type stable for the benefit of the timers. */ tcpcb_zone = uma_zcreate("tcpcb", sizeof(struct tcpcb_mem), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcpcb_zone, maxsockets); tcptw_zone = uma_zcreate("tcptw", sizeof(struct tcptw), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcptw_zone, maxsockets / 5); tcp_timer_init(); syncache_init(); tcp_hc_init(); tcp_reass_init(); callout_init(&isn_callout, CALLOUT_MPSAFE); tcp_isn_tick(NULL); EVENTHANDLER_REGISTER(shutdown_pre_sync, tcp_fini, NULL, SHUTDOWN_PRI_DEFAULT); } void tcp_fini(xtp) void *xtp; { callout_stop(&isn_callout); } /* * Fill in the IP and TCP headers for an outgoing packet, given the tcpcb. * tcp_template used to store this data in mbufs, but we now recopy it out * of the tcpcb each time to conserve mbufs. */ void tcpip_fillheaders(inp, ip_ptr, tcp_ptr) struct inpcb *inp; void *ip_ptr; void *tcp_ptr; { struct tcphdr *th = (struct tcphdr *)tcp_ptr; #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { struct ip6_hdr *ip6; ip6 = (struct ip6_hdr *)ip_ptr; ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | (inp->in6p_flowinfo & IPV6_FLOWINFO_MASK); ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | (IPV6_VERSION & IPV6_VERSION_MASK); ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = sizeof(struct tcphdr); ip6->ip6_src = inp->in6p_laddr; ip6->ip6_dst = inp->in6p_faddr; } else #endif { struct ip *ip; ip = (struct ip *)ip_ptr; ip->ip_v = IPVERSION; ip->ip_hl = 5; ip->ip_tos = inp->inp_ip_tos; ip->ip_len = 0; ip->ip_id = 0; ip->ip_off = 0; ip->ip_ttl = inp->inp_ip_ttl; ip->ip_sum = 0; ip->ip_p = IPPROTO_TCP; ip->ip_src = inp->inp_laddr; ip->ip_dst = inp->inp_faddr; } th->th_sport = inp->inp_lport; th->th_dport = inp->inp_fport; th->th_seq = 0; th->th_ack = 0; th->th_x2 = 0; th->th_off = 5; th->th_flags = 0; th->th_win = 0; th->th_urp = 0; th->th_sum = 0; /* in_pseudo() is called later for ipv4 */ } /* * Create template to be used to send tcp packets on a connection. * Allocates an mbuf and fills in a skeletal tcp/ip header. The only * use for this function is in keepalives, which use tcp_respond. */ struct tcptemp * tcpip_maketemplate(inp) struct inpcb *inp; { struct mbuf *m; struct tcptemp *n; m = m_get(M_DONTWAIT, MT_HEADER); if (m == NULL) return (0); m->m_len = sizeof(struct tcptemp); n = mtod(m, struct tcptemp *); tcpip_fillheaders(inp, (void *)&n->tt_ipgen, (void *)&n->tt_t); return (n); } /* * Send a single message to the TCP at address specified by * the given TCP/IP header. If m == NULL, then we make a copy * of the tcpiphdr at ti and send directly to the addressed host. * This is used to force keep alive messages out using the TCP * template for a connection. If flags are given then we send * a message back to the TCP which originated the * segment ti, * and discard the mbuf containing it and any other attached mbufs. * * In any case the ack and sequence number of the transmitted * segment are as specified by the parameters. * * NOTE: If m != NULL, then ti must point to *inside* the mbuf. */ void tcp_respond(tp, ipgen, th, m, ack, seq, flags) struct tcpcb *tp; void *ipgen; register struct tcphdr *th; register struct mbuf *m; tcp_seq ack, seq; int flags; { register int tlen; int win = 0; struct ip *ip; struct tcphdr *nth; #ifdef INET6 struct ip6_hdr *ip6; int isipv6; #endif /* INET6 */ int ipflags = 0; - struct inpcb *inp = NULL; + struct inpcb *inp; KASSERT(tp != NULL || m != NULL, ("tcp_respond: tp and m both NULL")); #ifdef INET6 isipv6 = ((struct ip *)ipgen)->ip_v == 6; ip6 = ipgen; #endif /* INET6 */ ip = ipgen; if (tp != NULL) { inp = tp->t_inpcb; KASSERT(inp != NULL, ("tcp control block w/o inpcb")); INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); + } else + inp = NULL; + + if (tp != NULL) { if (!(flags & TH_RST)) { win = sbspace(&inp->inp_socket->so_rcv); if (win > (long)TCP_MAXWIN << tp->rcv_scale) win = (long)TCP_MAXWIN << tp->rcv_scale; } } if (m == NULL) { m = m_gethdr(M_DONTWAIT, MT_HEADER); if (m == NULL) return; tlen = 0; m->m_data += max_linkhdr; #ifdef INET6 if (isipv6) { bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(struct ip6_hdr)); ip6 = mtod(m, struct ip6_hdr *); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); ip = mtod(m, struct ip *); nth = (struct tcphdr *)(ip + 1); } bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); flags = TH_ACK; } else { m_freem(m->m_next); m->m_next = NULL; m->m_data = (caddr_t)ipgen; /* m_len is set later */ tlen = 0; #define xchg(a,b,type) { type t; t=a; a=b; b=t; } #ifdef INET6 if (isipv6) { xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long); nth = (struct tcphdr *)(ip + 1); } if (th != nth) { /* * this is usually a case when an extension header * exists between the IPv6 header and the * TCP header. */ nth->th_sport = th->th_sport; nth->th_dport = th->th_dport; } xchg(nth->th_dport, nth->th_sport, n_short); #undef xchg } #ifdef INET6 if (isipv6) { ip6->ip6_flow = 0; ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) + tlen)); tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); } else #endif { tlen += sizeof (struct tcpiphdr); ip->ip_len = tlen; ip->ip_ttl = ip_defttl; if (path_mtu_discovery) ip->ip_off |= IP_DF; } m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; #ifdef MAC if (inp != NULL) { /* * Packet is associated with a socket, so allow the * label of the response to reflect the socket label. */ - mac_create_mbuf_from_socket(inp->inp_socket, m); + INP_LOCK_ASSERT(inp); + mac_create_mbuf_from_inpcb(inp, m); } else { /* * Packet is not associated with a socket, so possibly * update the label in place. */ mac_reflect_mbuf_tcp(m); } #endif nth->th_seq = htonl(seq); nth->th_ack = htonl(ack); nth->th_x2 = 0; nth->th_off = sizeof (struct tcphdr) >> 2; nth->th_flags = flags; if (tp != NULL) nth->th_win = htons((u_short) (win >> tp->rcv_scale)); else nth->th_win = htons((u_short)win); nth->th_urp = 0; #ifdef INET6 if (isipv6) { nth->th_sum = 0; nth->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), tlen - sizeof(struct ip6_hdr)); ip6->ip6_hlim = in6_selecthlim(tp != NULL ? tp->t_inpcb : NULL, NULL); } else #endif /* INET6 */ { nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p))); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); } #ifdef TCPDEBUG if (tp == NULL || (inp->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); #endif #ifdef INET6 if (isipv6) (void) ip6_output(m, NULL, NULL, ipflags, NULL, NULL, inp); else #endif /* INET6 */ (void) ip_output(m, NULL, NULL, ipflags, NULL, inp); } /* * Create a new TCP control block, making an * empty reassembly queue and hooking it to the argument * protocol control block. The `inp' parameter must have * come from the zone allocator set up in tcp_init(). */ struct tcpcb * tcp_newtcpcb(inp) struct inpcb *inp; { struct tcpcb_mem *tm; struct tcpcb *tp; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ tm = uma_zalloc(tcpcb_zone, M_NOWAIT | M_ZERO); if (tm == NULL) return (NULL); tp = &tm->tcb; /* LIST_INIT(&tp->t_segq); */ /* XXX covered by M_ZERO */ tp->t_maxseg = tp->t_maxopd = #ifdef INET6 isipv6 ? tcp_v6mssdflt : #endif /* INET6 */ tcp_mssdflt; /* Set up our timeouts. */ callout_init(tp->tt_rexmt = &tm->tcpcb_mem_rexmt, 0); callout_init(tp->tt_persist = &tm->tcpcb_mem_persist, 0); callout_init(tp->tt_keep = &tm->tcpcb_mem_keep, 0); callout_init(tp->tt_2msl = &tm->tcpcb_mem_2msl, 0); callout_init(tp->tt_delack = &tm->tcpcb_mem_delack, 0); if (tcp_do_rfc1323) tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); if (tcp_do_rfc1644) tp->t_flags |= TF_REQ_CC; tp->t_inpcb = inp; /* XXX */ /* * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives * reasonable initial retransmit time. */ tp->t_srtt = TCPTV_SRTTBASE; tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; tp->t_rttmin = tcp_rexmit_min; tp->t_rxtcur = TCPTV_RTOBASE; tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->t_rcvtime = ticks; tp->t_bw_rtttime = ticks; /* * IPv4 TTL initialization is necessary for an IPv6 socket as well, * because the socket may be bound to an IPv6 wildcard address, * which may match an IPv4-mapped IPv6 address. */ inp->inp_ip_ttl = ip_defttl; inp->inp_ppcb = (caddr_t)tp; return (tp); /* XXX */ } /* * Drop a TCP connection, reporting * the specified error. If connection is synchronized, * then send a RST to peer. */ struct tcpcb * tcp_drop(tp, errno) register struct tcpcb *tp; int errno; { struct socket *so = tp->t_inpcb->inp_socket; if (TCPS_HAVERCVDSYN(tp->t_state)) { tp->t_state = TCPS_CLOSED; (void) tcp_output(tp); tcpstat.tcps_drops++; } else tcpstat.tcps_conndrops++; if (errno == ETIMEDOUT && tp->t_softerror) errno = tp->t_softerror; so->so_error = errno; return (tcp_close(tp)); } static void tcp_discardcb(tp) struct tcpcb *tp; { struct tseg_qent *q; struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ /* * Make sure that all of our timers are stopped before we * delete the PCB. */ callout_stop(tp->tt_rexmt); callout_stop(tp->tt_persist); callout_stop(tp->tt_keep); callout_stop(tp->tt_2msl); callout_stop(tp->tt_delack); /* * If we got enough samples through the srtt filter, * save the rtt and rttvar in the routing entry. * 'Enough' is arbitrarily defined as 4 rtt samples. * 4 samples is enough for the srtt filter to converge * to within enough % of the correct value; fewer samples * and we could save a bogus rtt. The danger is not high * as tcp quickly recovers from everything. * XXX: Works very well but needs some more statistics! */ if (tp->t_rttupdated >= 4) { struct hc_metrics_lite metrics; u_long ssthresh; bzero(&metrics, sizeof(metrics)); /* * Update the ssthresh always when the conditions below * are satisfied. This gives us better new start value * for the congestion avoidance for new connections. * ssthresh is only set if packet loss occured on a session. */ ssthresh = tp->snd_ssthresh; if (ssthresh != 0 && ssthresh < so->so_snd.sb_hiwat / 2) { /* * convert the limit from user data bytes to * packets then to packet data bytes. */ ssthresh = (ssthresh + tp->t_maxseg / 2) / tp->t_maxseg; if (ssthresh < 2) ssthresh = 2; ssthresh *= (u_long)(tp->t_maxseg + #ifdef INET6 (isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : #endif sizeof (struct tcpiphdr) #ifdef INET6 ) #endif ); } else ssthresh = 0; metrics.rmx_ssthresh = ssthresh; metrics.rmx_rtt = tp->t_srtt; metrics.rmx_rttvar = tp->t_rttvar; /* XXX: This wraps if the pipe is more than 4 Gbit per second */ metrics.rmx_bandwidth = tp->snd_bandwidth; metrics.rmx_cwnd = tp->snd_cwnd; metrics.rmx_sendpipe = 0; metrics.rmx_recvpipe = 0; tcp_hc_update(&inp->inp_inc, &metrics); } /* free the reassembly queue, if any */ while ((q = LIST_FIRST(&tp->t_segq)) != NULL) { LIST_REMOVE(q, tqe_q); m_freem(q->tqe_m); uma_zfree(tcp_reass_zone, q); tp->t_segqlen--; tcp_reass_qsize--; } inp->inp_ppcb = NULL; tp->t_inpcb = NULL; uma_zfree(tcpcb_zone, tp); soisdisconnected(so); } /* * Close a TCP control block: * discard all space held by the tcp * discard internet protocol block * wake up any sleepers */ struct tcpcb * tcp_close(tp) struct tcpcb *tp; { struct inpcb *inp = tp->t_inpcb; #ifdef INET6 struct socket *so = inp->inp_socket; #endif tcp_discardcb(tp); #ifdef INET6 if (INP_CHECK_SOCKAF(so, AF_INET6)) in6_pcbdetach(inp); else #endif in_pcbdetach(inp); tcpstat.tcps_closed++; return (NULL); } void tcp_drain() { if (do_tcpdrain) { struct inpcb *inpb; struct tcpcb *tcpb; struct tseg_qent *te; /* * Walk the tcpbs, if existing, and flush the reassembly queue, * if there is one... * XXX: The "Net/3" implementation doesn't imply that the TCP * reassembly queue should be flushed, but in a situation * where we're really low on mbufs, this is potentially * usefull. */ INP_INFO_RLOCK(&tcbinfo); LIST_FOREACH(inpb, tcbinfo.listhead, inp_list) { if (inpb->inp_vflag & INP_TIMEWAIT) continue; INP_LOCK(inpb); if ((tcpb = intotcpcb(inpb)) != NULL) { while ((te = LIST_FIRST(&tcpb->t_segq)) != NULL) { LIST_REMOVE(te, tqe_q); m_freem(te->tqe_m); uma_zfree(tcp_reass_zone, te); tcpb->t_segqlen--; tcp_reass_qsize--; } } INP_UNLOCK(inpb); } INP_INFO_RUNLOCK(&tcbinfo); } } /* * Notify a tcp user of an asynchronous error; * store error as soft error, but wake up user * (for now, won't do anything until can select for soft error). * * Do not wake up user since there currently is no mechanism for * reporting soft errors (yet - a kqueue filter may be added). */ static struct inpcb * tcp_notify(inp, error) struct inpcb *inp; int error; { struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb; /* * Ignore some errors if we are hooked up. * If connection hasn't completed, has retransmitted several times, * and receives a second error, give up now. This is better * than waiting a long time to establish a connection that * can never complete. */ if (tp->t_state == TCPS_ESTABLISHED && (error == EHOSTUNREACH || error == ENETUNREACH || error == EHOSTDOWN)) { return inp; } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && tp->t_softerror) { tcp_drop(tp, error); return (struct inpcb *)0; } else { tp->t_softerror = error; return inp; } #if 0 wakeup( &so->so_timeo); sorwakeup(so); sowwakeup(so); #endif } static int tcp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n, s; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == NULL) { n = tcbinfo.ipi_count; req->oldidx = 2 * (sizeof xig) + (n + n/8) * sizeof(struct xtcpcb); return 0; } if (req->newptr != NULL) return EPERM; /* * OK, now we're committed to doing something. */ s = splnet(); INP_INFO_RLOCK(&tcbinfo); gencnt = tcbinfo.ipi_gencnt; n = tcbinfo.ipi_count; INP_INFO_RUNLOCK(&tcbinfo); splx(s); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + n * sizeof(struct xtcpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return error; inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == NULL) return ENOMEM; s = splnet(); INP_INFO_RLOCK(&tcbinfo); for (inp = LIST_FIRST(tcbinfo.listhead), i = 0; inp != NULL && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_LOCK(inp); if (inp->inp_gencnt <= gencnt) { /* * XXX: This use of cr_cansee(), introduced with * TCP state changes, is not quite right, but for * now, better than nothing. */ if (inp->inp_vflag & INP_TIMEWAIT) error = cr_cansee(req->td->td_ucred, intotw(inp)->tw_cred); else error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error == 0) inp_list[i++] = inp; } INP_UNLOCK(inp); } INP_INFO_RUNLOCK(&tcbinfo); splx(s); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; if (inp->inp_gencnt <= gencnt) { struct xtcpcb xt; caddr_t inp_ppcb; xt.xt_len = sizeof xt; /* XXX should avoid extra copy */ bcopy(inp, &xt.xt_inp, sizeof *inp); inp_ppcb = inp->inp_ppcb; if (inp_ppcb == NULL) bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); else if (inp->inp_vflag & INP_TIMEWAIT) { bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); xt.xt_tp.t_state = TCPS_TIME_WAIT; } else bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); if (inp->inp_socket != NULL) sotoxsocket(inp->inp_socket, &xt.xt_socket); else { bzero(&xt.xt_socket, sizeof xt.xt_socket); xt.xt_socket.xso_protocol = IPPROTO_TCP; } xt.xt_inp.inp_gencnt = inp->inp_gencnt; error = SYSCTL_OUT(req, &xt, sizeof xt); } } if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ s = splnet(); INP_INFO_RLOCK(&tcbinfo); xig.xig_gen = tcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = tcbinfo.ipi_count; INP_INFO_RUNLOCK(&tcbinfo); splx(s); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return error; } SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); static int tcp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error, s; error = suser_cred(req->td->td_ucred, PRISON_ROOT); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); s = splnet(); INP_INFO_RLOCK(&tcbinfo); inp = in_pcblookup_hash(&tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, 0, NULL); if (inp == NULL) { error = ENOENT; goto outunlocked; } INP_LOCK(inp); if (inp->inp_socket == NULL) { error = ENOENT; goto out; } error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error) goto out; cru2x(inp->inp_socket->so_cred, &xuc); out: INP_UNLOCK(inp); outunlocked: INP_INFO_RUNLOCK(&tcbinfo); splx(s); if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp_getcred, "S,xucred", "Get the xucred of a TCP connection"); #ifdef INET6 static int tcp6_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in6 addrs[2]; struct inpcb *inp; int error, s, mapped = 0; error = suser_cred(req->td->td_ucred, PRISON_ROOT); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) mapped = 1; else return (EINVAL); } s = splnet(); INP_INFO_RLOCK(&tcbinfo); if (mapped == 1) inp = in_pcblookup_hash(&tcbinfo, *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], addrs[1].sin6_port, *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], addrs[0].sin6_port, 0, NULL); else inp = in6_pcblookup_hash(&tcbinfo, &addrs[1].sin6_addr, addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 0, NULL); if (inp == NULL) { error = ENOENT; goto outunlocked; } INP_LOCK(inp); if (inp->inp_socket == NULL) { error = ENOENT; goto out; } error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error) goto out; cru2x(inp->inp_socket->so_cred, &xuc); out: INP_UNLOCK(inp); outunlocked: INP_INFO_RUNLOCK(&tcbinfo); splx(s); if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection"); #endif void tcp_ctlinput(cmd, sa, vip) int cmd; struct sockaddr *sa; void *vip; { struct ip *ip = vip; struct tcphdr *th; struct in_addr faddr; struct inpcb *inp; struct tcpcb *tp; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; tcp_seq icmp_seq; int s; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (cmd == PRC_QUENCH) notify = tcp_quench; else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB || cmd == PRC_UNREACH_PORT || cmd == PRC_TIMXCEED_INTRANS) && ip) notify = tcp_drop_syn_sent; else if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; /* * Redirects don't need to be handled up here. */ else if (PRC_IS_REDIRECT(cmd)) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * XXX: We never get this from ICMP, otherwise it makes an * excellent DoS attack on machines with many connections. */ else if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip != NULL) { s = splnet(); th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); INP_INFO_WLOCK(&tcbinfo); inp = in_pcblookup_hash(&tcbinfo, faddr, th->th_dport, ip->ip_src, th->th_sport, 0, NULL); if (inp != NULL) { INP_LOCK(inp); if (inp->inp_socket != NULL) { icmp_seq = htonl(th->th_seq); tp = intotcpcb(inp); if (SEQ_GEQ(icmp_seq, tp->snd_una) && SEQ_LT(icmp_seq, tp->snd_max)) inp = (*notify)(inp, inetctlerrmap[cmd]); } if (inp != NULL) INP_UNLOCK(inp); } else { struct in_conninfo inc; inc.inc_fport = th->th_dport; inc.inc_lport = th->th_sport; inc.inc_faddr = faddr; inc.inc_laddr = ip->ip_src; #ifdef INET6 inc.inc_isipv6 = 0; #endif syncache_unreach(&inc, th); } INP_INFO_WUNLOCK(&tcbinfo); splx(s); } else in_pcbnotifyall(&tcbinfo, faddr, inetctlerrmap[cmd], notify); } #ifdef INET6 void tcp6_ctlinput(cmd, sa, d) int cmd; struct sockaddr *sa; void *d; { struct tcphdr th; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct ip6_hdr *ip6; struct mbuf *m; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; int off; struct tcp_portonly { u_int16_t th_sport; u_int16_t th_dport; } *thp; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; if (cmd == PRC_QUENCH) notify = tcp_quench; else if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; else if (!PRC_IS_REDIRECT(cmd) && ((unsigned)cmd >= PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) return; /* if the parameter is from icmp6, decode it. */ if (d != NULL) { ip6cp = (struct ip6ctlparam *)d; m = ip6cp->ip6c_m; ip6 = ip6cp->ip6c_ip6; off = ip6cp->ip6c_off; sa6_src = ip6cp->ip6c_src; } else { m = NULL; ip6 = NULL; off = 0; /* fool gcc */ sa6_src = &sa6_any; } if (ip6 != NULL) { struct in_conninfo inc; /* * XXX: We assume that when IPV6 is non NULL, * M and OFF are valid. */ /* check if we can safely examine src and dst ports */ if (m->m_pkthdr.len < off + sizeof(*thp)) return; bzero(&th, sizeof(th)); m_copydata(m, off, sizeof(*thp), (caddr_t)&th); in6_pcbnotify(&tcb, sa, th.th_dport, (struct sockaddr *)ip6cp->ip6c_src, th.th_sport, cmd, NULL, notify); inc.inc_fport = th.th_dport; inc.inc_lport = th.th_sport; inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr; inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr; inc.inc_isipv6 = 1; syncache_unreach(&inc, &th); } else in6_pcbnotify(&tcb, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify); } #endif /* INET6 */ /* * Following is where TCP initial sequence number generation occurs. * * There are two places where we must use initial sequence numbers: * 1. In SYN-ACK packets. * 2. In SYN packets. * * All ISNs for SYN-ACK packets are generated by the syncache. See * tcp_syncache.c for details. * * The ISNs in SYN packets must be monotonic; TIME_WAIT recycling * depends on this property. In addition, these ISNs should be * unguessable so as to prevent connection hijacking. To satisfy * the requirements of this situation, the algorithm outlined in * RFC 1948 is used, with only small modifications. * * Implementation details: * * Time is based off the system timer, and is corrected so that it * increases by one megabyte per second. This allows for proper * recycling on high speed LANs while still leaving over an hour * before rollover. * * As reading the *exact* system time is too expensive to be done * whenever setting up a TCP connection, we increment the time * offset in two ways. First, a small random positive increment * is added to isn_offset for each connection that is set up. * Second, the function tcp_isn_tick fires once per clock tick * and increments isn_offset as necessary so that sequence numbers * are incremented at approximately ISN_BYTES_PER_SECOND. The * random positive increments serve only to ensure that the same * exact sequence number is never sent out twice (as could otherwise * happen when a port is recycled in less than the system tick * interval.) * * net.inet.tcp.isn_reseed_interval controls the number of seconds * between seeding of isn_secret. This is normally set to zero, * as reseeding should not be necessary. * */ #define ISN_BYTES_PER_SECOND 1048576 #define ISN_STATIC_INCREMENT 4096 #define ISN_RANDOM_INCREMENT (4096 - 1) u_char isn_secret[32]; int isn_last_reseed; u_int32_t isn_offset, isn_offset_old; MD5_CTX isn_ctx; tcp_seq tcp_new_isn(tp) struct tcpcb *tp; { u_int32_t md5_buffer[4]; tcp_seq new_isn; /* Seed if this is the first use, reseed if requested. */ if ((isn_last_reseed == 0) || ((tcp_isn_reseed_interval > 0) && (((u_int)isn_last_reseed + (u_int)tcp_isn_reseed_interval*hz) < (u_int)ticks))) { read_random(&isn_secret, sizeof(isn_secret)); isn_last_reseed = ticks; } /* Compute the md5 hash and return the ISN. */ MD5Init(&isn_ctx); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short)); #ifdef INET6 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr, sizeof(struct in6_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr, sizeof(struct in6_addr)); } else #endif { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr, sizeof(struct in_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr, sizeof(struct in_addr)); } MD5Update(&isn_ctx, (u_char *) &isn_secret, sizeof(isn_secret)); MD5Final((u_char *) &md5_buffer, &isn_ctx); new_isn = (tcp_seq) md5_buffer[0]; isn_offset += ISN_STATIC_INCREMENT + (arc4random() & ISN_RANDOM_INCREMENT); new_isn += isn_offset; return new_isn; } /* * Increment the offset to the next ISN_BYTES_PER_SECOND / hz boundary * to keep time flowing at a relatively constant rate. If the random * increments have already pushed us past the projected offset, do nothing. */ static void tcp_isn_tick(xtp) void *xtp; { u_int32_t projected_offset; projected_offset = isn_offset_old + ISN_BYTES_PER_SECOND / hz; if (projected_offset > isn_offset) isn_offset = projected_offset; isn_offset_old = isn_offset; callout_reset(&isn_callout, 1, tcp_isn_tick, NULL); } /* * When a source quench is received, close congestion window * to one segment. We will gradually open it again as we proceed. */ struct inpcb * tcp_quench(inp, errno) struct inpcb *inp; int errno; { struct tcpcb *tp = intotcpcb(inp); if (tp != NULL) tp->snd_cwnd = tp->t_maxseg; return (inp); } /* * When a specific ICMP unreachable message is received and the * connection state is SYN-SENT, drop the connection. This behavior * is controlled by the icmp_may_rst sysctl. */ struct inpcb * tcp_drop_syn_sent(inp, errno) struct inpcb *inp; int errno; { struct tcpcb *tp = intotcpcb(inp); if (tp != NULL && tp->t_state == TCPS_SYN_SENT) { tcp_drop(tp, errno); return (struct inpcb *)0; } return inp; } /* * When `need fragmentation' ICMP is received, update our idea of the MSS * based on the new value in the route. Also nudge TCP to send something, * since we know the packet we just sent was dropped. * This duplicates some code in the tcp_mss() function in tcp_input.c. */ struct inpcb * tcp_mtudisc(inp, errno) struct inpcb *inp; int errno; { struct tcpcb *tp = intotcpcb(inp); struct rmxp_tao tao; struct socket *so = inp->inp_socket; u_int maxmtu; u_int romtu; int mss; #ifdef INET6 int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ bzero(&tao, sizeof(tao)); if (tp != NULL) { maxmtu = tcp_hc_getmtu(&inp->inp_inc); /* IPv4 and IPv6 */ romtu = #ifdef INET6 isipv6 ? tcp_maxmtu6(&inp->inp_inc) : #endif /* INET6 */ tcp_maxmtu(&inp->inp_inc); if (!maxmtu) maxmtu = romtu; else maxmtu = min(maxmtu, romtu); if (!maxmtu) { tp->t_maxopd = tp->t_maxseg = #ifdef INET6 isipv6 ? tcp_v6mssdflt : #endif /* INET6 */ tcp_mssdflt; return inp; } mss = maxmtu - #ifdef INET6 (isipv6 ? sizeof(struct ip6_hdr) + sizeof(struct tcphdr) : #endif /* INET6 */ sizeof(struct tcpiphdr) #ifdef INET6 ) #endif /* INET6 */ ; if (tcp_do_rfc1644) { tcp_hc_gettao(&inp->inp_inc, &tao); if (tao.tao_mssopt) mss = min(mss, tao.tao_mssopt); } /* * XXX - The above conditional probably violates the TCP * spec. The problem is that, since we don't know the * other end's MSS, we are supposed to use a conservative * default. But, if we do that, then MTU discovery will * never actually take place, because the conservative * default is much less than the MTUs typically seen * on the Internet today. For the moment, we'll sweep * this under the carpet. * * The conservative default might not actually be a problem * if the only case this occurs is when sending an initial * SYN with options and data to a host we've never talked * to before. Then, they will reply with an MSS value which * will get recorded and the new parameters should get * recomputed. For Further Study. */ if (tp->t_maxopd <= mss) return inp; tp->t_maxopd = mss; if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP) mss -= TCPOLEN_TSTAMP_APPA; if ((tp->t_flags & (TF_REQ_CC|TF_NOOPT)) == TF_REQ_CC && (tp->t_flags & TF_RCVD_CC) == TF_RCVD_CC) mss -= TCPOLEN_CC_APPA; #if (MCLBYTES & (MCLBYTES - 1)) == 0 if (mss > MCLBYTES) mss &= ~(MCLBYTES-1); #else if (mss > MCLBYTES) mss = mss / MCLBYTES * MCLBYTES; #endif if (so->so_snd.sb_hiwat < mss) mss = so->so_snd.sb_hiwat; tp->t_maxseg = mss; tcpstat.tcps_mturesent++; tp->t_rtttime = 0; tp->snd_nxt = tp->snd_una; tcp_output(tp); } return inp; } /* * Look-up the routing entry to the peer of this inpcb. If no route * is found and it cannot be allocated, then return NULL. This routine * is called by TCP routines that access the rmx structure and by tcp_mss * to get the interface MTU. */ u_long tcp_maxmtu(inc) struct in_conninfo *inc; { struct route sro; struct sockaddr_in *dst; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer")); bzero(&sro, sizeof(sro)); if (inc->inc_faddr.s_addr != INADDR_ANY) { dst = (struct sockaddr_in *)&sro.ro_dst; dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = inc->inc_faddr; rtalloc_ign(&sro, RTF_CLONING); } if (sro.ro_rt != NULL) { ifp = sro.ro_rt->rt_ifp; if (sro.ro_rt->rt_rmx.rmx_mtu == 0) maxmtu = ifp->if_mtu; else maxmtu = min(sro.ro_rt->rt_rmx.rmx_mtu, ifp->if_mtu); RTFREE(sro.ro_rt); } return (maxmtu); } #ifdef INET6 u_long tcp_maxmtu6(inc) struct in_conninfo *inc; { struct route_in6 sro6; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu6 with NULL in_conninfo pointer")); bzero(&sro6, sizeof(sro6)); if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) { sro6.ro_dst.sin6_family = AF_INET6; sro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); sro6.ro_dst.sin6_addr = inc->inc6_faddr; rtalloc_ign((struct route *)&sro6, RTF_CLONING); } if (sro6.ro_rt != NULL) { ifp = sro6.ro_rt->rt_ifp; if (sro6.ro_rt->rt_rmx.rmx_mtu == 0) maxmtu = IN6_LINKMTU(sro6.ro_rt->rt_ifp); else maxmtu = min(sro6.ro_rt->rt_rmx.rmx_mtu, IN6_LINKMTU(sro6.ro_rt->rt_ifp)); RTFREE(sro6.ro_rt); } return (maxmtu); } #endif /* INET6 */ #ifdef IPSEC /* compute ESP/AH header size for TCP, including outer IP header. */ size_t ipsec_hdrsiz_tcp(tp) struct tcpcb *tp; { struct inpcb *inp; struct mbuf *m; size_t hdrsiz; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; #endif struct tcphdr *th; if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL)) return 0; MGETHDR(m, M_DONTWAIT, MT_DATA); if (!m) return 0; #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); tcpip_fillheaders(inp, ip6, th); hdrsiz = ipsec6_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } else #endif /* INET6 */ { ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); tcpip_fillheaders(inp, ip, th); hdrsiz = ipsec4_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } m_free(m); return hdrsiz; } #endif /*IPSEC*/ /* * Move a TCP connection into TIME_WAIT state. * tcbinfo is unlocked. * inp is locked, and is unlocked before returning. */ void tcp_twstart(tp) struct tcpcb *tp; { struct tcptw *tw; struct inpcb *inp; int tw_time, acknow; struct socket *so; tw = uma_zalloc(tcptw_zone, M_NOWAIT); if (tw == NULL) { tw = tcp_timer_2msl_tw(1); if (tw == NULL) { tcp_close(tp); return; } } inp = tp->t_inpcb; tw->tw_inpcb = inp; /* * Recover last window size sent. */ tw->last_win = (tp->rcv_adv - tp->rcv_nxt) >> tp->rcv_scale; /* * Set t_recent if timestamps are used on the connection. */ if ((tp->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP|TF_NOOPT)) == (TF_REQ_TSTMP|TF_RCVD_TSTMP)) tw->t_recent = tp->ts_recent; else tw->t_recent = 0; tw->snd_nxt = tp->snd_nxt; tw->rcv_nxt = tp->rcv_nxt; tw->iss = tp->iss; tw->irs = tp->irs; tw->cc_recv = tp->cc_recv; tw->cc_send = tp->cc_send; tw->t_starttime = tp->t_starttime; tw->tw_time = 0; /* XXX * If this code will * be used for fin-wait-2 state also, then we may need * a ts_recent from the last segment. */ /* Shorten TIME_WAIT [RFC-1644, p.28] */ if (tp->cc_recv != 0 && (ticks - tp->t_starttime) < tcp_msl) { tw_time = tp->t_rxtcur * TCPTV_TWTRUNC; /* For T/TCP client, force ACK now. */ acknow = 1; } else { tw_time = 2 * tcp_msl; acknow = tp->t_flags & TF_ACKNOW; } tcp_discardcb(tp); so = inp->inp_socket; so->so_pcb = NULL; tw->tw_cred = crhold(so->so_cred); tw->tw_so_options = so->so_options; if (acknow) tcp_twrespond(tw, TH_ACK); sotryfree(so); inp->inp_socket = NULL; inp->inp_ppcb = (caddr_t)tw; inp->inp_vflag |= INP_TIMEWAIT; tcp_timer_2msl_reset(tw, tw_time); INP_UNLOCK(inp); } /* * The appromixate rate of ISN increase of Microsoft TCP stacks; * the actual rate is slightly higher due to the addition of * random positive increments. * * Most other new OSes use semi-randomized ISN values, so we * do not need to worry about them. */ #define MS_ISN_BYTES_PER_SECOND 250000 /* * Determine if the ISN we will generate has advanced beyond the last * sequence number used by the previous connection. If so, indicate * that it is safe to recycle this tw socket by returning 1. */ int tcp_twrecycleable(struct tcptw *tw) { tcp_seq new_iss = tw->iss; tcp_seq new_irs = tw->irs; new_iss += (ticks - tw->t_starttime) * (ISN_BYTES_PER_SECOND / hz); new_irs += (ticks - tw->t_starttime) * (MS_ISN_BYTES_PER_SECOND / hz); if (SEQ_GT(new_iss, tw->snd_nxt) && SEQ_GT(new_irs, tw->rcv_nxt)) return 1; else return 0; } struct tcptw * tcp_twclose(struct tcptw *tw, int reuse) { struct inpcb *inp; inp = tw->tw_inpcb; tw->tw_inpcb = NULL; tcp_timer_2msl_stop(tw); inp->inp_ppcb = NULL; #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) in6_pcbdetach(inp); else #endif in_pcbdetach(inp); tcpstat.tcps_closed++; crfree(tw->tw_cred); tw->tw_cred = NULL; if (reuse) return (tw); uma_zfree(tcptw_zone, tw); return (NULL); } int tcp_twrespond(struct tcptw *tw, int flags) { struct inpcb *inp = tw->tw_inpcb; struct tcphdr *th; struct mbuf *m; struct ip *ip = NULL; u_int8_t *optp; u_int hdrlen, optlen; int error; #ifdef INET6 struct ip6_hdr *ip6 = NULL; int isipv6 = inp->inp_inc.inc_isipv6; #endif m = m_gethdr(M_DONTWAIT, MT_HEADER); if (m == NULL) return (ENOBUFS); m->m_data += max_linkhdr; #ifdef MAC mac_create_mbuf_from_inpcb(inp, m); #endif #ifdef INET6 if (isipv6) { hdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); tcpip_fillheaders(inp, ip6, th); } else #endif { hdrlen = sizeof(struct tcpiphdr); ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); tcpip_fillheaders(inp, ip, th); } optp = (u_int8_t *)(th + 1); /* * Send a timestamp and echo-reply if both our side and our peer * have sent timestamps in our SYN's and this is not a RST. */ if (tw->t_recent && flags == TH_ACK) { u_int32_t *lp = (u_int32_t *)optp; /* Form timestamp option as shown in appendix A of RFC 1323. */ *lp++ = htonl(TCPOPT_TSTAMP_HDR); *lp++ = htonl(ticks); *lp = htonl(tw->t_recent); optp += TCPOLEN_TSTAMP_APPA; } /* * Send `CC-family' options if needed, and it's not a RST. */ if (tw->cc_recv != 0 && flags == TH_ACK) { u_int32_t *lp = (u_int32_t *)optp; *lp++ = htonl(TCPOPT_CC_HDR(TCPOPT_CC)); *lp = htonl(tw->cc_send); optp += TCPOLEN_CC_APPA; } optlen = optp - (u_int8_t *)(th + 1); m->m_len = hdrlen + optlen; m->m_pkthdr.len = m->m_len; KASSERT(max_linkhdr + m->m_len <= MHLEN, ("tcptw: mbuf too small")); th->th_seq = htonl(tw->snd_nxt); th->th_ack = htonl(tw->rcv_nxt); th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; th->th_flags = flags; th->th_win = htons(tw->last_win); #ifdef INET6 if (isipv6) { th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), sizeof(struct tcphdr) + optlen); ip6->ip6_hlim = in6_selecthlim(inp, NULL); error = ip6_output(m, inp->in6p_outputopts, NULL, (tw->tw_so_options & SO_DONTROUTE), NULL, NULL, inp); } else #endif { th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons(sizeof(struct tcphdr) + optlen + IPPROTO_TCP)); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); ip->ip_len = m->m_pkthdr.len; if (path_mtu_discovery) ip->ip_off |= IP_DF; error = ip_output(m, inp->inp_options, NULL, (tw->tw_so_options & SO_DONTROUTE), NULL, inp); } if (flags & TH_ACK) tcpstat.tcps_sndacks++; else tcpstat.tcps_sndctrl++; tcpstat.tcps_sndtotal++; return (error); } /* * TCP BANDWIDTH DELAY PRODUCT WINDOW LIMITING * * This code attempts to calculate the bandwidth-delay product as a * means of determining the optimal window size to maximize bandwidth, * minimize RTT, and avoid the over-allocation of buffers on interfaces and * routers. This code also does a fairly good job keeping RTTs in check * across slow links like modems. We implement an algorithm which is very * similar (but not meant to be) TCP/Vegas. The code operates on the * transmitter side of a TCP connection and so only effects the transmit * side of the connection. * * BACKGROUND: TCP makes no provision for the management of buffer space * at the end points or at the intermediate routers and switches. A TCP * stream, whether using NewReno or not, will eventually buffer as * many packets as it is able and the only reason this typically works is * due to the fairly small default buffers made available for a connection * (typicaly 16K or 32K). As machines use larger windows and/or window * scaling it is now fairly easy for even a single TCP connection to blow-out * all available buffer space not only on the local interface, but on * intermediate routers and switches as well. NewReno makes a misguided * attempt to 'solve' this problem by waiting for an actual failure to occur, * then backing off, then steadily increasing the window again until another * failure occurs, ad-infinitum. This results in terrible oscillation that * is only made worse as network loads increase and the idea of intentionally * blowing out network buffers is, frankly, a terrible way to manage network * resources. * * It is far better to limit the transmit window prior to the failure * condition being achieved. There are two general ways to do this: First * you can 'scan' through different transmit window sizes and locate the * point where the RTT stops increasing, indicating that you have filled the * pipe, then scan backwards until you note that RTT stops decreasing, then * repeat ad-infinitum. This method works in principle but has severe * implementation issues due to RTT variances, timer granularity, and * instability in the algorithm which can lead to many false positives and * create oscillations as well as interact badly with other TCP streams * implementing the same algorithm. * * The second method is to limit the window to the bandwidth delay product * of the link. This is the method we implement. RTT variances and our * own manipulation of the congestion window, bwnd, can potentially * destabilize the algorithm. For this reason we have to stabilize the * elements used to calculate the window. We do this by using the minimum * observed RTT, the long term average of the observed bandwidth, and * by adding two segments worth of slop. It isn't perfect but it is able * to react to changing conditions and gives us a very stable basis on * which to extend the algorithm. */ void tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq) { u_long bw; u_long bwnd; int save_ticks; /* * If inflight_enable is disabled in the middle of a tcp connection, * make sure snd_bwnd is effectively disabled. */ if (tcp_inflight_enable == 0) { tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_bandwidth = 0; return; } /* * Figure out the bandwidth. Due to the tick granularity this * is a very rough number and it MUST be averaged over a fairly * long period of time. XXX we need to take into account a link * that is not using all available bandwidth, but for now our * slop will ramp us up if this case occurs and the bandwidth later * increases. * * Note: if ticks rollover 'bw' may wind up negative. We must * effectively reset t_bw_rtttime for this case. */ save_ticks = ticks; if ((u_int)(save_ticks - tp->t_bw_rtttime) < 1) return; bw = (int64_t)(ack_seq - tp->t_bw_rtseq) * hz / (save_ticks - tp->t_bw_rtttime); tp->t_bw_rtttime = save_ticks; tp->t_bw_rtseq = ack_seq; if (tp->t_bw_rtttime == 0 || (int)bw < 0) return; bw = ((int64_t)tp->snd_bandwidth * 15 + bw) >> 4; tp->snd_bandwidth = bw; /* * Calculate the semi-static bandwidth delay product, plus two maximal * segments. The additional slop puts us squarely in the sweet * spot and also handles the bandwidth run-up case and stabilization. * Without the slop we could be locking ourselves into a lower * bandwidth. * * Situations Handled: * (1) Prevents over-queueing of packets on LANs, especially on * high speed LANs, allowing larger TCP buffers to be * specified, and also does a good job preventing * over-queueing of packets over choke points like modems * (at least for the transmit side). * * (2) Is able to handle changing network loads (bandwidth * drops so bwnd drops, bandwidth increases so bwnd * increases). * * (3) Theoretically should stabilize in the face of multiple * connections implementing the same algorithm (this may need * a little work). * * (4) Stability value (defaults to 20 = 2 maximal packets) can * be adjusted with a sysctl but typically only needs to be * on very slow connections. A value no smaller then 5 * should be used, but only reduce this default if you have * no other choice. */ #define USERTT ((tp->t_srtt + tp->t_rttbest) / 2) bwnd = (int64_t)bw * USERTT / (hz << TCP_RTT_SHIFT) + tcp_inflight_stab * tp->t_maxseg / 10; #undef USERTT if (tcp_inflight_debug > 0) { static int ltime; if ((u_int)(ticks - ltime) >= hz / tcp_inflight_debug) { ltime = ticks; printf("%p bw %ld rttbest %d srtt %d bwnd %ld\n", tp, bw, tp->t_rttbest, tp->t_srtt, bwnd ); } } if ((long)bwnd < tcp_inflight_min) bwnd = tcp_inflight_min; if (bwnd > tcp_inflight_max) bwnd = tcp_inflight_max; if ((long)bwnd < tp->t_maxseg * 2) bwnd = tp->t_maxseg * 2; tp->snd_bwnd = bwnd; } #ifdef TCP_SIGNATURE /* * Callback function invoked by m_apply() to digest TCP segment data * contained within an mbuf chain. */ static int tcp_signature_apply(void *fstate, void *data, u_int len) { MD5Update(fstate, (u_char *)data, len); return (0); } /* * Compute TCP-MD5 hash of a TCPv4 segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * off0 offset to TCP header within the mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND) * * We do this over ip, tcphdr, segment data, and the key in the SADB. * When called from tcp_input(), we can be sure that th_sum has been * zeroed out and verified already. * * This function is for IPv4 use only. Calling this function with an * IPv6 packet in the mbuf chain will yield undefined results. * * Return 0 if successful, otherwise return -1. * * XXX The key is retrieved from the system's PF_KEY SADB, by keying a * search with the destination IP address, and a 'magic SPI' to be * determined by the application. This is hardcoded elsewhere to 1179 * right now. Another branch of this code exists which uses the SPD to * specify per-application flows but it is unstable. */ int tcp_signature_compute(struct mbuf *m, int off0, int len, int optlen, u_char *buf, u_int direction) { union sockaddr_union dst; struct ippseudo ippseudo; MD5_CTX ctx; int doff; struct ip *ip; struct ipovly *ipovly; struct secasvar *sav; struct tcphdr *th; u_short savecsum; KASSERT(m != NULL, ("NULL mbuf chain")); KASSERT(buf != NULL, ("NULL signature pointer")); /* Extract the destination from the IP header in the mbuf. */ ip = mtod(m, struct ip *); bzero(&dst, sizeof(union sockaddr_union)); dst.sa.sa_len = sizeof(struct sockaddr_in); dst.sa.sa_family = AF_INET; dst.sin.sin_addr = (direction == IPSEC_DIR_INBOUND) ? ip->ip_src : ip->ip_dst; /* Look up an SADB entry which matches the address of the peer. */ sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); if (sav == NULL) { printf("%s: SADB lookup failed for %s\n", __func__, inet_ntoa(dst.sin.sin_addr)); return (EINVAL); } MD5Init(&ctx); ipovly = (struct ipovly *)ip; th = (struct tcphdr *)((u_char *)ip + off0); doff = off0 + sizeof(struct tcphdr) + optlen; /* * Step 1: Update MD5 hash with IP pseudo-header. * * XXX The ippseudo header MUST be digested in network byte order, * or else we'll fail the regression test. Assume all fields we've * been doing arithmetic on have been in host byte order. * XXX One cannot depend on ipovly->ih_len here. When called from * tcp_output(), the underlying ip_len member has not yet been set. */ ippseudo.ippseudo_src = ipovly->ih_src; ippseudo.ippseudo_dst = ipovly->ih_dst; ippseudo.ippseudo_pad = 0; ippseudo.ippseudo_p = IPPROTO_TCP; ippseudo.ippseudo_len = htons(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&ippseudo, sizeof(struct ippseudo)); /* * Step 2: Update MD5 hash with TCP header, excluding options. * The TCP checksum must be set to zero. */ savecsum = th->th_sum; th->th_sum = 0; MD5Update(&ctx, (char *)th, sizeof(struct tcphdr)); th->th_sum = savecsum; /* * Step 3: Update MD5 hash with TCP segment data. * Use m_apply() to avoid an early m_pullup(). */ if (len > 0) m_apply(m, doff, len, tcp_signature_apply, &ctx); /* * Step 4: Update MD5 hash with shared secret. */ MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth)); MD5Final(buf, &ctx); key_sa_recordxfer(sav, m); KEY_FREESAV(&sav); return (0); } #endif /* TCP_SIGNATURE */ diff --git a/sys/netinet/tcp_syncache.c b/sys/netinet/tcp_syncache.c index 51575a3779ae..1fb7e99e1b6f 100644 --- a/sys/netinet/tcp_syncache.c +++ b/sys/netinet/tcp_syncache.c @@ -1,1453 +1,1453 @@ /*- * Copyright (c) 2001 Networks Associates Technology, Inc. * All rights reserved. * * This software was developed for the FreeBSD Project by Jonathan Lemon * and NAI Labs, the Security Research Division of Network Associates, Inc. * under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the * DARPA CHATS research program. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote * products derived from this software without specific prior written * permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_mac.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #include #include #include /* for proc0 declaration */ #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #include #include #endif #include #ifdef TCPDEBUG #include #endif #include #include #include #include #ifdef TCPDEBUG #include #endif #ifdef INET6 #include #endif #ifdef IPSEC #include #ifdef INET6 #include #endif #endif /*IPSEC*/ #ifdef FAST_IPSEC #include #ifdef INET6 #include #endif #include #endif /*FAST_IPSEC*/ #include #include static int tcp_syncookies = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, syncookies, CTLFLAG_RW, &tcp_syncookies, 0, "Use TCP SYN cookies if the syncache overflows"); static void syncache_drop(struct syncache *, struct syncache_head *); static void syncache_free(struct syncache *); static void syncache_insert(struct syncache *, struct syncache_head *); struct syncache *syncache_lookup(struct in_conninfo *, struct syncache_head **); #ifdef TCPDEBUG static int syncache_respond(struct syncache *, struct mbuf *, struct socket *); #else static int syncache_respond(struct syncache *, struct mbuf *); #endif static struct socket *syncache_socket(struct syncache *, struct socket *, struct mbuf *m); static void syncache_timer(void *); static u_int32_t syncookie_generate(struct syncache *); static struct syncache *syncookie_lookup(struct in_conninfo *, struct tcphdr *, struct socket *); /* * Transmit the SYN,ACK fewer times than TCP_MAXRXTSHIFT specifies. * 3 retransmits corresponds to a timeout of (1 + 2 + 4 + 8 == 15) seconds, * the odds are that the user has given up attempting to connect by then. */ #define SYNCACHE_MAXREXMTS 3 /* Arbitrary values */ #define TCP_SYNCACHE_HASHSIZE 512 #define TCP_SYNCACHE_BUCKETLIMIT 30 struct tcp_syncache { struct syncache_head *hashbase; uma_zone_t zone; u_int hashsize; u_int hashmask; u_int bucket_limit; u_int cache_count; u_int cache_limit; u_int rexmt_limit; u_int hash_secret; TAILQ_HEAD(, syncache) timerq[SYNCACHE_MAXREXMTS + 1]; struct callout tt_timerq[SYNCACHE_MAXREXMTS + 1]; }; static struct tcp_syncache tcp_syncache; SYSCTL_NODE(_net_inet_tcp, OID_AUTO, syncache, CTLFLAG_RW, 0, "TCP SYN cache"); SYSCTL_INT(_net_inet_tcp_syncache, OID_AUTO, bucketlimit, CTLFLAG_RDTUN, &tcp_syncache.bucket_limit, 0, "Per-bucket hash limit for syncache"); SYSCTL_INT(_net_inet_tcp_syncache, OID_AUTO, cachelimit, CTLFLAG_RDTUN, &tcp_syncache.cache_limit, 0, "Overall entry limit for syncache"); SYSCTL_INT(_net_inet_tcp_syncache, OID_AUTO, count, CTLFLAG_RD, &tcp_syncache.cache_count, 0, "Current number of entries in syncache"); SYSCTL_INT(_net_inet_tcp_syncache, OID_AUTO, hashsize, CTLFLAG_RDTUN, &tcp_syncache.hashsize, 0, "Size of TCP syncache hashtable"); SYSCTL_INT(_net_inet_tcp_syncache, OID_AUTO, rexmtlimit, CTLFLAG_RW, &tcp_syncache.rexmt_limit, 0, "Limit on SYN/ACK retransmissions"); static MALLOC_DEFINE(M_SYNCACHE, "syncache", "TCP syncache"); #define SYNCACHE_HASH(inc, mask) \ ((tcp_syncache.hash_secret ^ \ (inc)->inc_faddr.s_addr ^ \ ((inc)->inc_faddr.s_addr >> 16) ^ \ (inc)->inc_fport ^ (inc)->inc_lport) & mask) #define SYNCACHE_HASH6(inc, mask) \ ((tcp_syncache.hash_secret ^ \ (inc)->inc6_faddr.s6_addr32[0] ^ \ (inc)->inc6_faddr.s6_addr32[3] ^ \ (inc)->inc_fport ^ (inc)->inc_lport) & mask) #define ENDPTS_EQ(a, b) ( \ (a)->ie_fport == (b)->ie_fport && \ (a)->ie_lport == (b)->ie_lport && \ (a)->ie_faddr.s_addr == (b)->ie_faddr.s_addr && \ (a)->ie_laddr.s_addr == (b)->ie_laddr.s_addr \ ) #define ENDPTS6_EQ(a, b) (memcmp(a, b, sizeof(*a)) == 0) #define SYNCACHE_TIMEOUT(sc, slot) do { \ sc->sc_rxtslot = (slot); \ sc->sc_rxttime = ticks + TCPTV_RTOBASE * tcp_backoff[(slot)]; \ TAILQ_INSERT_TAIL(&tcp_syncache.timerq[(slot)], sc, sc_timerq); \ if (!callout_active(&tcp_syncache.tt_timerq[(slot)])) \ callout_reset(&tcp_syncache.tt_timerq[(slot)], \ TCPTV_RTOBASE * tcp_backoff[(slot)], \ syncache_timer, (void *)((intptr_t)(slot))); \ } while (0) static void syncache_free(struct syncache *sc) { if (sc->sc_ipopts) (void) m_free(sc->sc_ipopts); uma_zfree(tcp_syncache.zone, sc); } void syncache_init(void) { int i; tcp_syncache.cache_count = 0; tcp_syncache.hashsize = TCP_SYNCACHE_HASHSIZE; tcp_syncache.bucket_limit = TCP_SYNCACHE_BUCKETLIMIT; tcp_syncache.cache_limit = tcp_syncache.hashsize * tcp_syncache.bucket_limit; tcp_syncache.rexmt_limit = SYNCACHE_MAXREXMTS; tcp_syncache.hash_secret = arc4random(); TUNABLE_INT_FETCH("net.inet.tcp.syncache.hashsize", &tcp_syncache.hashsize); TUNABLE_INT_FETCH("net.inet.tcp.syncache.cachelimit", &tcp_syncache.cache_limit); TUNABLE_INT_FETCH("net.inet.tcp.syncache.bucketlimit", &tcp_syncache.bucket_limit); if (!powerof2(tcp_syncache.hashsize)) { printf("WARNING: syncache hash size is not a power of 2.\n"); tcp_syncache.hashsize = 512; /* safe default */ } tcp_syncache.hashmask = tcp_syncache.hashsize - 1; /* Allocate the hash table. */ MALLOC(tcp_syncache.hashbase, struct syncache_head *, tcp_syncache.hashsize * sizeof(struct syncache_head), M_SYNCACHE, M_WAITOK); /* Initialize the hash buckets. */ for (i = 0; i < tcp_syncache.hashsize; i++) { TAILQ_INIT(&tcp_syncache.hashbase[i].sch_bucket); tcp_syncache.hashbase[i].sch_length = 0; } /* Initialize the timer queues. */ for (i = 0; i <= SYNCACHE_MAXREXMTS; i++) { TAILQ_INIT(&tcp_syncache.timerq[i]); callout_init(&tcp_syncache.tt_timerq[i], debug_mpsafenet ? CALLOUT_MPSAFE : 0); } /* * Allocate the syncache entries. Allow the zone to allocate one * more entry than cache limit, so a new entry can bump out an * older one. */ tcp_syncache.zone = uma_zcreate("syncache", sizeof(struct syncache), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcp_syncache.zone, tcp_syncache.cache_limit); tcp_syncache.cache_limit -= 1; } static void syncache_insert(sc, sch) struct syncache *sc; struct syncache_head *sch; { struct syncache *sc2; int i; INP_INFO_WLOCK_ASSERT(&tcbinfo); /* * Make sure that we don't overflow the per-bucket * limit or the total cache size limit. */ if (sch->sch_length >= tcp_syncache.bucket_limit) { /* * The bucket is full, toss the oldest element. */ sc2 = TAILQ_FIRST(&sch->sch_bucket); sc2->sc_tp->ts_recent = ticks; syncache_drop(sc2, sch); tcpstat.tcps_sc_bucketoverflow++; } else if (tcp_syncache.cache_count >= tcp_syncache.cache_limit) { /* * The cache is full. Toss the oldest entry in the * entire cache. This is the front entry in the * first non-empty timer queue with the largest * timeout value. */ for (i = SYNCACHE_MAXREXMTS; i >= 0; i--) { sc2 = TAILQ_FIRST(&tcp_syncache.timerq[i]); if (sc2 != NULL) break; } sc2->sc_tp->ts_recent = ticks; syncache_drop(sc2, NULL); tcpstat.tcps_sc_cacheoverflow++; } /* Initialize the entry's timer. */ SYNCACHE_TIMEOUT(sc, 0); /* Put it into the bucket. */ TAILQ_INSERT_TAIL(&sch->sch_bucket, sc, sc_hash); sch->sch_length++; tcp_syncache.cache_count++; tcpstat.tcps_sc_added++; } static void syncache_drop(sc, sch) struct syncache *sc; struct syncache_head *sch; { INP_INFO_WLOCK_ASSERT(&tcbinfo); if (sch == NULL) { #ifdef INET6 if (sc->sc_inc.inc_isipv6) { sch = &tcp_syncache.hashbase[ SYNCACHE_HASH6(&sc->sc_inc, tcp_syncache.hashmask)]; } else #endif { sch = &tcp_syncache.hashbase[ SYNCACHE_HASH(&sc->sc_inc, tcp_syncache.hashmask)]; } } TAILQ_REMOVE(&sch->sch_bucket, sc, sc_hash); sch->sch_length--; tcp_syncache.cache_count--; TAILQ_REMOVE(&tcp_syncache.timerq[sc->sc_rxtslot], sc, sc_timerq); if (TAILQ_EMPTY(&tcp_syncache.timerq[sc->sc_rxtslot])) callout_stop(&tcp_syncache.tt_timerq[sc->sc_rxtslot]); syncache_free(sc); } /* * Walk the timer queues, looking for SYN,ACKs that need to be retransmitted. * If we have retransmitted an entry the maximum number of times, expire it. */ static void syncache_timer(xslot) void *xslot; { intptr_t slot = (intptr_t)xslot; struct syncache *sc, *nsc; struct inpcb *inp; INP_INFO_WLOCK(&tcbinfo); if (callout_pending(&tcp_syncache.tt_timerq[slot]) || !callout_active(&tcp_syncache.tt_timerq[slot])) { /* XXX can this happen? */ INP_INFO_WUNLOCK(&tcbinfo); return; } callout_deactivate(&tcp_syncache.tt_timerq[slot]); nsc = TAILQ_FIRST(&tcp_syncache.timerq[slot]); while (nsc != NULL) { if (ticks < nsc->sc_rxttime) break; sc = nsc; inp = sc->sc_tp->t_inpcb; if (slot == SYNCACHE_MAXREXMTS || slot >= tcp_syncache.rexmt_limit || inp == NULL || inp->inp_gencnt != sc->sc_inp_gencnt) { nsc = TAILQ_NEXT(sc, sc_timerq); syncache_drop(sc, NULL); tcpstat.tcps_sc_stale++; continue; } /* * syncache_respond() may call back into the syncache to * to modify another entry, so do not obtain the next * entry on the timer chain until it has completed. */ #ifdef TCPDEBUG (void) syncache_respond(sc, NULL, NULL); #else (void) syncache_respond(sc, NULL); #endif nsc = TAILQ_NEXT(sc, sc_timerq); tcpstat.tcps_sc_retransmitted++; TAILQ_REMOVE(&tcp_syncache.timerq[slot], sc, sc_timerq); SYNCACHE_TIMEOUT(sc, slot + 1); } if (nsc != NULL) callout_reset(&tcp_syncache.tt_timerq[slot], nsc->sc_rxttime - ticks, syncache_timer, (void *)(slot)); INP_INFO_WUNLOCK(&tcbinfo); } /* * Find an entry in the syncache. */ struct syncache * syncache_lookup(inc, schp) struct in_conninfo *inc; struct syncache_head **schp; { struct syncache *sc; struct syncache_head *sch; INP_INFO_WLOCK_ASSERT(&tcbinfo); #ifdef INET6 if (inc->inc_isipv6) { sch = &tcp_syncache.hashbase[ SYNCACHE_HASH6(inc, tcp_syncache.hashmask)]; *schp = sch; TAILQ_FOREACH(sc, &sch->sch_bucket, sc_hash) { if (ENDPTS6_EQ(&inc->inc_ie, &sc->sc_inc.inc_ie)) return (sc); } } else #endif { sch = &tcp_syncache.hashbase[ SYNCACHE_HASH(inc, tcp_syncache.hashmask)]; *schp = sch; TAILQ_FOREACH(sc, &sch->sch_bucket, sc_hash) { #ifdef INET6 if (sc->sc_inc.inc_isipv6) continue; #endif if (ENDPTS_EQ(&inc->inc_ie, &sc->sc_inc.inc_ie)) return (sc); } } return (NULL); } /* * This function is called when we get a RST for a * non-existent connection, so that we can see if the * connection is in the syn cache. If it is, zap it. */ void syncache_chkrst(inc, th) struct in_conninfo *inc; struct tcphdr *th; { struct syncache *sc; struct syncache_head *sch; INP_INFO_WLOCK_ASSERT(&tcbinfo); sc = syncache_lookup(inc, &sch); if (sc == NULL) return; /* * If the RST bit is set, check the sequence number to see * if this is a valid reset segment. * RFC 793 page 37: * In all states except SYN-SENT, all reset (RST) segments * are validated by checking their SEQ-fields. A reset is * valid if its sequence number is in the window. * * The sequence number in the reset segment is normally an * echo of our outgoing acknowlegement numbers, but some hosts * send a reset with the sequence number at the rightmost edge * of our receive window, and we have to handle this case. */ if (SEQ_GEQ(th->th_seq, sc->sc_irs) && SEQ_LEQ(th->th_seq, sc->sc_irs + sc->sc_wnd)) { syncache_drop(sc, sch); tcpstat.tcps_sc_reset++; } } void syncache_badack(inc) struct in_conninfo *inc; { struct syncache *sc; struct syncache_head *sch; INP_INFO_WLOCK_ASSERT(&tcbinfo); sc = syncache_lookup(inc, &sch); if (sc != NULL) { syncache_drop(sc, sch); tcpstat.tcps_sc_badack++; } } void syncache_unreach(inc, th) struct in_conninfo *inc; struct tcphdr *th; { struct syncache *sc; struct syncache_head *sch; INP_INFO_WLOCK_ASSERT(&tcbinfo); /* we are called at splnet() here */ sc = syncache_lookup(inc, &sch); if (sc == NULL) return; /* If the sequence number != sc_iss, then it's a bogus ICMP msg */ if (ntohl(th->th_seq) != sc->sc_iss) return; /* * If we've rertransmitted 3 times and this is our second error, * we remove the entry. Otherwise, we allow it to continue on. * This prevents us from incorrectly nuking an entry during a * spurious network outage. * * See tcp_notify(). */ if ((sc->sc_flags & SCF_UNREACH) == 0 || sc->sc_rxtslot < 3) { sc->sc_flags |= SCF_UNREACH; return; } syncache_drop(sc, sch); tcpstat.tcps_sc_unreach++; } /* * Build a new TCP socket structure from a syncache entry. */ static struct socket * syncache_socket(sc, lso, m) struct syncache *sc; struct socket *lso; struct mbuf *m; { struct inpcb *inp = NULL; struct socket *so; struct tcpcb *tp; GIANT_REQUIRED; /* XXX until socket locking */ INP_INFO_WLOCK_ASSERT(&tcbinfo); /* * Ok, create the full blown connection, and set things up * as they would have been set up if we had created the * connection when the SYN arrived. If we can't create * the connection, abort it. */ so = sonewconn(lso, SS_ISCONNECTED); if (so == NULL) { /* * Drop the connection; we will send a RST if the peer * retransmits the ACK, */ tcpstat.tcps_listendrop++; goto abort2; } #ifdef MAC mac_set_socket_peer_from_mbuf(m, so); #endif inp = sotoinpcb(so); INP_LOCK(inp); /* * Insert new socket into hash list. */ inp->inp_inc.inc_isipv6 = sc->sc_inc.inc_isipv6; #ifdef INET6 if (sc->sc_inc.inc_isipv6) { inp->in6p_laddr = sc->sc_inc.inc6_laddr; } else { inp->inp_vflag &= ~INP_IPV6; inp->inp_vflag |= INP_IPV4; #endif inp->inp_laddr = sc->sc_inc.inc_laddr; #ifdef INET6 } #endif inp->inp_lport = sc->sc_inc.inc_lport; if (in_pcbinshash(inp) != 0) { /* * Undo the assignments above if we failed to * put the PCB on the hash lists. */ #ifdef INET6 if (sc->sc_inc.inc_isipv6) inp->in6p_laddr = in6addr_any; else #endif inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_lport = 0; goto abort; } #ifdef IPSEC /* copy old policy into new socket's */ if (ipsec_copy_pcbpolicy(sotoinpcb(lso)->inp_sp, inp->inp_sp)) printf("syncache_expand: could not copy policy\n"); #endif #ifdef FAST_IPSEC /* copy old policy into new socket's */ if (ipsec_copy_policy(sotoinpcb(lso)->inp_sp, inp->inp_sp)) printf("syncache_expand: could not copy policy\n"); #endif #ifdef INET6 if (sc->sc_inc.inc_isipv6) { struct inpcb *oinp = sotoinpcb(lso); struct in6_addr laddr6; struct sockaddr_in6 sin6; /* * Inherit socket options from the listening socket. * Note that in6p_inputopts are not (and should not be) * copied, since it stores previously received options and is * used to detect if each new option is different than the * previous one and hence should be passed to a user. * If we copied in6p_inputopts, a user would not be able to * receive options just after calling the accept system call. */ inp->inp_flags |= oinp->inp_flags & INP_CONTROLOPTS; if (oinp->in6p_outputopts) inp->in6p_outputopts = ip6_copypktopts(oinp->in6p_outputopts, M_NOWAIT); sin6.sin6_family = AF_INET6; sin6.sin6_len = sizeof(sin6); sin6.sin6_addr = sc->sc_inc.inc6_faddr; sin6.sin6_port = sc->sc_inc.inc_fport; sin6.sin6_flowinfo = sin6.sin6_scope_id = 0; laddr6 = inp->in6p_laddr; if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) inp->in6p_laddr = sc->sc_inc.inc6_laddr; if (in6_pcbconnect(inp, (struct sockaddr *)&sin6, thread0.td_ucred)) { inp->in6p_laddr = laddr6; goto abort; } } else #endif { struct in_addr laddr; struct sockaddr_in sin; inp->inp_options = ip_srcroute(); if (inp->inp_options == NULL) { inp->inp_options = sc->sc_ipopts; sc->sc_ipopts = NULL; } sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); sin.sin_addr = sc->sc_inc.inc_faddr; sin.sin_port = sc->sc_inc.inc_fport; bzero((caddr_t)sin.sin_zero, sizeof(sin.sin_zero)); laddr = inp->inp_laddr; if (inp->inp_laddr.s_addr == INADDR_ANY) inp->inp_laddr = sc->sc_inc.inc_laddr; if (in_pcbconnect(inp, (struct sockaddr *)&sin, thread0.td_ucred)) { inp->inp_laddr = laddr; goto abort; } } tp = intotcpcb(inp); tp->t_state = TCPS_SYN_RECEIVED; tp->iss = sc->sc_iss; tp->irs = sc->sc_irs; tcp_rcvseqinit(tp); tcp_sendseqinit(tp); tp->snd_wl1 = sc->sc_irs; tp->rcv_up = sc->sc_irs + 1; tp->rcv_wnd = sc->sc_wnd; tp->rcv_adv += tp->rcv_wnd; tp->t_flags = sototcpcb(lso)->t_flags & (TF_NOPUSH|TF_NODELAY); if (sc->sc_flags & SCF_NOOPT) tp->t_flags |= TF_NOOPT; if (sc->sc_flags & SCF_WINSCALE) { tp->t_flags |= TF_REQ_SCALE|TF_RCVD_SCALE; tp->requested_s_scale = sc->sc_requested_s_scale; tp->request_r_scale = sc->sc_request_r_scale; } if (sc->sc_flags & SCF_TIMESTAMP) { tp->t_flags |= TF_REQ_TSTMP|TF_RCVD_TSTMP; tp->ts_recent = sc->sc_tsrecent; tp->ts_recent_age = ticks; } if (sc->sc_flags & SCF_CC) { /* * Initialization of the tcpcb for transaction; * set SND.WND = SEG.WND, * initialize CCsend and CCrecv. */ tp->t_flags |= TF_REQ_CC|TF_RCVD_CC; tp->cc_send = sc->sc_cc_send; tp->cc_recv = sc->sc_cc_recv; } #ifdef TCP_SIGNATURE if (sc->sc_flags & SCF_SIGNATURE) tp->t_flags |= TF_SIGNATURE; #endif /* * Set up MSS and get cached values from tcp_hostcache. * This might overwrite some of the defaults we just set. */ tcp_mss(tp, sc->sc_peer_mss); /* * If the SYN,ACK was retransmitted, reset cwnd to 1 segment. */ if (sc->sc_rxtslot != 0) tp->snd_cwnd = tp->t_maxseg; callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); INP_UNLOCK(inp); tcpstat.tcps_accepts++; return (so); abort: INP_UNLOCK(inp); abort2: if (so != NULL) (void) soabort(so); return (NULL); } /* * This function gets called when we receive an ACK for a * socket in the LISTEN state. We look up the connection * in the syncache, and if its there, we pull it out of * the cache and turn it into a full-blown connection in * the SYN-RECEIVED state. */ int syncache_expand(inc, th, sop, m) struct in_conninfo *inc; struct tcphdr *th; struct socket **sop; struct mbuf *m; { struct syncache *sc; struct syncache_head *sch; struct socket *so; INP_INFO_WLOCK_ASSERT(&tcbinfo); sc = syncache_lookup(inc, &sch); if (sc == NULL) { /* * There is no syncache entry, so see if this ACK is * a returning syncookie. To do this, first: * A. See if this socket has had a syncache entry dropped in * the past. We don't want to accept a bogus syncookie * if we've never received a SYN. * B. check that the syncookie is valid. If it is, then * cobble up a fake syncache entry, and return. */ if (!tcp_syncookies) return (0); sc = syncookie_lookup(inc, th, *sop); if (sc == NULL) return (0); sch = NULL; tcpstat.tcps_sc_recvcookie++; } /* * If seg contains an ACK, but not for our SYN/ACK, send a RST. */ if (th->th_ack != sc->sc_iss + 1) return (0); so = syncache_socket(sc, *sop, m); if (so == NULL) { #if 0 resetandabort: /* XXXjlemon check this - is this correct? */ (void) tcp_respond(NULL, m, m, th, th->th_seq + tlen, (tcp_seq)0, TH_RST|TH_ACK); #endif m_freem(m); /* XXX only needed for above */ tcpstat.tcps_sc_aborted++; } else tcpstat.tcps_sc_completed++; if (sch == NULL) syncache_free(sc); else syncache_drop(sc, sch); *sop = so; return (1); } /* * Given a LISTEN socket and an inbound SYN request, add * this to the syn cache, and send back a segment: * * to the source. * * IMPORTANT NOTE: We do _NOT_ ACK data that might accompany the SYN. * Doing so would require that we hold onto the data and deliver it * to the application. However, if we are the target of a SYN-flood * DoS attack, an attacker could send data which would eventually * consume all available buffer space if it were ACKed. By not ACKing * the data, we avoid this DoS scenario. */ int syncache_add(inc, to, th, sop, m) struct in_conninfo *inc; struct tcpopt *to; struct tcphdr *th; struct socket **sop; struct mbuf *m; { struct tcpcb *tp; struct socket *so; struct syncache *sc = NULL; struct syncache_head *sch; struct mbuf *ipopts = NULL; struct rmxp_tao tao; int i, win; INP_INFO_WLOCK_ASSERT(&tcbinfo); so = *sop; tp = sototcpcb(so); bzero(&tao, sizeof(tao)); /* * Remember the IP options, if any. */ #ifdef INET6 if (!inc->inc_isipv6) #endif ipopts = ip_srcroute(); /* * See if we already have an entry for this connection. * If we do, resend the SYN,ACK, and reset the retransmit timer. * * XXX * should the syncache be re-initialized with the contents * of the new SYN here (which may have different options?) */ sc = syncache_lookup(inc, &sch); if (sc != NULL) { tcpstat.tcps_sc_dupsyn++; if (ipopts) { /* * If we were remembering a previous source route, * forget it and use the new one we've been given. */ if (sc->sc_ipopts) (void) m_free(sc->sc_ipopts); sc->sc_ipopts = ipopts; } /* * Update timestamp if present. */ if (sc->sc_flags & SCF_TIMESTAMP) sc->sc_tsrecent = to->to_tsval; /* * PCB may have changed, pick up new values. */ sc->sc_tp = tp; sc->sc_inp_gencnt = tp->t_inpcb->inp_gencnt; #ifdef TCPDEBUG if (syncache_respond(sc, m, so) == 0) { #else if (syncache_respond(sc, m) == 0) { #endif /* NB: guarded by INP_INFO_WLOCK(&tcbinfo) */ TAILQ_REMOVE(&tcp_syncache.timerq[sc->sc_rxtslot], sc, sc_timerq); SYNCACHE_TIMEOUT(sc, sc->sc_rxtslot); tcpstat.tcps_sndacks++; tcpstat.tcps_sndtotal++; } *sop = NULL; return (1); } sc = uma_zalloc(tcp_syncache.zone, M_NOWAIT); if (sc == NULL) { /* * The zone allocator couldn't provide more entries. * Treat this as if the cache was full; drop the oldest * entry and insert the new one. */ /* NB: guarded by INP_INFO_WLOCK(&tcbinfo) */ for (i = SYNCACHE_MAXREXMTS; i >= 0; i--) { sc = TAILQ_FIRST(&tcp_syncache.timerq[i]); if (sc != NULL) break; } sc->sc_tp->ts_recent = ticks; syncache_drop(sc, NULL); tcpstat.tcps_sc_zonefail++; sc = uma_zalloc(tcp_syncache.zone, M_NOWAIT); if (sc == NULL) { if (ipopts) (void) m_free(ipopts); return (0); } } /* * Fill in the syncache values. */ bzero(sc, sizeof(*sc)); sc->sc_tp = tp; sc->sc_inp_gencnt = tp->t_inpcb->inp_gencnt; sc->sc_ipopts = ipopts; sc->sc_inc.inc_fport = inc->inc_fport; sc->sc_inc.inc_lport = inc->inc_lport; #ifdef INET6 sc->sc_inc.inc_isipv6 = inc->inc_isipv6; if (inc->inc_isipv6) { sc->sc_inc.inc6_faddr = inc->inc6_faddr; sc->sc_inc.inc6_laddr = inc->inc6_laddr; } else #endif { sc->sc_inc.inc_faddr = inc->inc_faddr; sc->sc_inc.inc_laddr = inc->inc_laddr; } sc->sc_irs = th->th_seq; sc->sc_flags = 0; sc->sc_peer_mss = to->to_flags & TOF_MSS ? to->to_mss : 0; if (tcp_syncookies) sc->sc_iss = syncookie_generate(sc); else sc->sc_iss = arc4random(); /* Initial receive window: clip sbspace to [0 .. TCP_MAXWIN] */ win = sbspace(&so->so_rcv); win = imax(win, 0); win = imin(win, TCP_MAXWIN); sc->sc_wnd = win; if (tcp_do_rfc1323) { /* * A timestamp received in a SYN makes * it ok to send timestamp requests and replies. */ if (to->to_flags & TOF_TS) { sc->sc_tsrecent = to->to_tsval; sc->sc_flags |= SCF_TIMESTAMP; } if (to->to_flags & TOF_SCALE) { int wscale = 0; /* Compute proper scaling value from buffer space */ while (wscale < TCP_MAX_WINSHIFT && (TCP_MAXWIN << wscale) < so->so_rcv.sb_hiwat) wscale++; sc->sc_request_r_scale = wscale; sc->sc_requested_s_scale = to->to_requested_s_scale; sc->sc_flags |= SCF_WINSCALE; } } if (tcp_do_rfc1644) { /* * A CC or CC.new option received in a SYN makes * it ok to send CC in subsequent segments. */ if (to->to_flags & (TOF_CC|TOF_CCNEW)) { sc->sc_cc_recv = to->to_cc; sc->sc_cc_send = CC_INC(tcp_ccgen); sc->sc_flags |= SCF_CC; } } if (tp->t_flags & TF_NOOPT) sc->sc_flags = SCF_NOOPT; #ifdef TCP_SIGNATURE /* * If listening socket requested TCP digests, and received SYN * contains the option, flag this in the syncache so that * syncache_respond() will do the right thing with the SYN+ACK. * XXX Currently we always record the option by default and will * attempt to use it in syncache_respond(). */ if (to->to_flags & TOF_SIGNATURE) sc->sc_flags = SCF_SIGNATURE; #endif /* * XXX * We have the option here of not doing TAO (even if the segment * qualifies) and instead fall back to a normal 3WHS via the syncache. * This allows us to apply synflood protection to TAO-qualifying SYNs * also. However, there should be a hueristic to determine when to * do this, and is not present at the moment. */ /* * Perform TAO test on incoming CC (SEG.CC) option, if any. * - compare SEG.CC against cached CC from the same host, if any. * - if SEG.CC > chached value, SYN must be new and is accepted * immediately: save new CC in the cache, mark the socket * connected, enter ESTABLISHED state, turn on flag to * send a SYN in the next segment. * A virtual advertised window is set in rcv_adv to * initialize SWS prevention. Then enter normal segment * processing: drop SYN, process data and FIN. * - otherwise do a normal 3-way handshake. */ if (tcp_do_rfc1644) tcp_hc_gettao(&sc->sc_inc, &tao); if ((to->to_flags & TOF_CC) != 0) { if (((tp->t_flags & TF_NOPUSH) != 0) && sc->sc_flags & SCF_CC && tao.tao_cc != 0 && CC_GT(to->to_cc, tao.tao_cc)) { sc->sc_rxtslot = 0; so = syncache_socket(sc, *sop, m); if (so != NULL) { tao.tao_cc = to->to_cc; tcp_hc_updatetao(&sc->sc_inc, TCP_HC_TAO_CC, tao.tao_cc, 0); *sop = so; } syncache_free(sc); return (so != NULL); } } else { /* * No CC option, but maybe CC.NEW: invalidate cached value. */ if (tcp_do_rfc1644) { tao.tao_cc = 0; tcp_hc_updatetao(&sc->sc_inc, TCP_HC_TAO_CC, tao.tao_cc, 0); } } /* * TAO test failed or there was no CC option, * do a standard 3-way handshake. */ #ifdef TCPDEBUG if (syncache_respond(sc, m, so) == 0) { #else if (syncache_respond(sc, m) == 0) { #endif syncache_insert(sc, sch); tcpstat.tcps_sndacks++; tcpstat.tcps_sndtotal++; } else { syncache_free(sc); tcpstat.tcps_sc_dropped++; } *sop = NULL; return (1); } #ifdef TCPDEBUG static int syncache_respond(sc, m, so) struct syncache *sc; struct mbuf *m; struct socket *so; #else static int syncache_respond(sc, m) struct syncache *sc; struct mbuf *m; #endif { u_int8_t *optp; int optlen, error; u_int16_t tlen, hlen, mssopt; struct ip *ip = NULL; struct tcphdr *th; struct inpcb *inp; #ifdef INET6 struct ip6_hdr *ip6 = NULL; #endif hlen = #ifdef INET6 (sc->sc_inc.inc_isipv6) ? sizeof(struct ip6_hdr) : #endif sizeof(struct ip); KASSERT((&sc->sc_inc) != NULL, ("syncache_respond with NULL in_conninfo pointer")); /* Determine MSS we advertize to other end of connection */ mssopt = tcp_mssopt(&sc->sc_inc); /* Compute the size of the TCP options. */ if (sc->sc_flags & SCF_NOOPT) { optlen = 0; } else { optlen = TCPOLEN_MAXSEG + ((sc->sc_flags & SCF_WINSCALE) ? 4 : 0) + ((sc->sc_flags & SCF_TIMESTAMP) ? TCPOLEN_TSTAMP_APPA : 0) + ((sc->sc_flags & SCF_CC) ? TCPOLEN_CC_APPA * 2 : 0); #ifdef TCP_SIGNATURE optlen += (sc->sc_flags & SCF_SIGNATURE) ? TCPOLEN_SIGNATURE + 2 : 0; #endif } tlen = hlen + sizeof(struct tcphdr) + optlen; /* * XXX * assume that the entire packet will fit in a header mbuf */ KASSERT(max_linkhdr + tlen <= MHLEN, ("syncache: mbuf too small")); /* * XXX shouldn't this reuse the mbuf if possible ? * Create the IP+TCP header from scratch. */ if (m) m_freem(m); m = m_gethdr(M_DONTWAIT, MT_HEADER); if (m == NULL) return (ENOBUFS); m->m_data += max_linkhdr; m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; inp = sc->sc_tp->t_inpcb; INP_LOCK(inp); #ifdef MAC - mac_create_mbuf_from_socket(inp->inp_socket, m); + mac_create_mbuf_from_inpcb(inp, m); #endif #ifdef INET6 if (sc->sc_inc.inc_isipv6) { ip6 = mtod(m, struct ip6_hdr *); ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_src = sc->sc_inc.inc6_laddr; ip6->ip6_dst = sc->sc_inc.inc6_faddr; ip6->ip6_plen = htons(tlen - hlen); /* ip6_hlim is set after checksum */ /* ip6_flow = ??? */ th = (struct tcphdr *)(ip6 + 1); } else #endif { ip = mtod(m, struct ip *); ip->ip_v = IPVERSION; ip->ip_hl = sizeof(struct ip) >> 2; ip->ip_len = tlen; ip->ip_id = 0; ip->ip_off = 0; ip->ip_sum = 0; ip->ip_p = IPPROTO_TCP; ip->ip_src = sc->sc_inc.inc_laddr; ip->ip_dst = sc->sc_inc.inc_faddr; ip->ip_ttl = inp->inp_ip_ttl; /* XXX */ ip->ip_tos = inp->inp_ip_tos; /* XXX */ /* * See if we should do MTU discovery. Route lookups are * expensive, so we will only unset the DF bit if: * * 1) path_mtu_discovery is disabled * 2) the SCF_UNREACH flag has been set */ if (path_mtu_discovery && ((sc->sc_flags & SCF_UNREACH) == 0)) ip->ip_off |= IP_DF; th = (struct tcphdr *)(ip + 1); } th->th_sport = sc->sc_inc.inc_lport; th->th_dport = sc->sc_inc.inc_fport; th->th_seq = htonl(sc->sc_iss); th->th_ack = htonl(sc->sc_irs + 1); th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; th->th_x2 = 0; th->th_flags = TH_SYN|TH_ACK; th->th_win = htons(sc->sc_wnd); th->th_urp = 0; /* Tack on the TCP options. */ if (optlen != 0) { optp = (u_int8_t *)(th + 1); *optp++ = TCPOPT_MAXSEG; *optp++ = TCPOLEN_MAXSEG; *optp++ = (mssopt >> 8) & 0xff; *optp++ = mssopt & 0xff; if (sc->sc_flags & SCF_WINSCALE) { *((u_int32_t *)optp) = htonl(TCPOPT_NOP << 24 | TCPOPT_WINDOW << 16 | TCPOLEN_WINDOW << 8 | sc->sc_request_r_scale); optp += 4; } if (sc->sc_flags & SCF_TIMESTAMP) { u_int32_t *lp = (u_int32_t *)(optp); /* Form timestamp option per appendix A of RFC 1323. */ *lp++ = htonl(TCPOPT_TSTAMP_HDR); *lp++ = htonl(ticks); *lp = htonl(sc->sc_tsrecent); optp += TCPOLEN_TSTAMP_APPA; } /* * Send CC and CC.echo if we received CC from our peer. */ if (sc->sc_flags & SCF_CC) { u_int32_t *lp = (u_int32_t *)(optp); *lp++ = htonl(TCPOPT_CC_HDR(TCPOPT_CC)); *lp++ = htonl(sc->sc_cc_send); *lp++ = htonl(TCPOPT_CC_HDR(TCPOPT_CCECHO)); *lp = htonl(sc->sc_cc_recv); optp += TCPOLEN_CC_APPA * 2; } #ifdef TCP_SIGNATURE /* * Handle TCP-MD5 passive opener response. */ if (sc->sc_flags & SCF_SIGNATURE) { u_int8_t *bp = optp; int i; *bp++ = TCPOPT_SIGNATURE; *bp++ = TCPOLEN_SIGNATURE; for (i = 0; i < TCP_SIGLEN; i++) *bp++ = 0; tcp_signature_compute(m, sizeof(struct ip), 0, optlen, optp + 2, IPSEC_DIR_OUTBOUND); *bp++ = TCPOPT_NOP; *bp++ = TCPOPT_EOL; optp += TCPOLEN_SIGNATURE + 2; } #endif /* TCP_SIGNATURE */ } #ifdef INET6 if (sc->sc_inc.inc_isipv6) { th->th_sum = 0; th->th_sum = in6_cksum(m, IPPROTO_TCP, hlen, tlen - hlen); ip6->ip6_hlim = in6_selecthlim(NULL, NULL); error = ip6_output(m, NULL, NULL, 0, NULL, NULL, inp); } else #endif { th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons(tlen - hlen + IPPROTO_TCP)); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); #ifdef TCPDEBUG /* * Trace. */ if (so != NULL && so->so_options & SO_DEBUG) { struct tcpcb *tp = sototcpcb(so); tcp_trace(TA_OUTPUT, tp->t_state, tp, mtod(m, void *), th, 0); } #endif error = ip_output(m, sc->sc_ipopts, NULL, 0, NULL, inp); } INP_UNLOCK(inp); return (error); } /* * cookie layers: * * |. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .| * | peer iss | * | MD5(laddr,faddr,secret,lport,fport) |. . . . . . .| * | 0 |(A)| | * (A): peer mss index */ /* * The values below are chosen to minimize the size of the tcp_secret * table, as well as providing roughly a 16 second lifetime for the cookie. */ #define SYNCOOKIE_WNDBITS 5 /* exposed bits for window indexing */ #define SYNCOOKIE_TIMESHIFT 1 /* scale ticks to window time units */ #define SYNCOOKIE_WNDMASK ((1 << SYNCOOKIE_WNDBITS) - 1) #define SYNCOOKIE_NSECRETS (1 << SYNCOOKIE_WNDBITS) #define SYNCOOKIE_TIMEOUT \ (hz * (1 << SYNCOOKIE_WNDBITS) / (1 << SYNCOOKIE_TIMESHIFT)) #define SYNCOOKIE_DATAMASK ((3 << SYNCOOKIE_WNDBITS) | SYNCOOKIE_WNDMASK) static struct { u_int32_t ts_secbits[4]; u_int ts_expire; } tcp_secret[SYNCOOKIE_NSECRETS]; static int tcp_msstab[] = { 0, 536, 1460, 8960 }; static MD5_CTX syn_ctx; #define MD5Add(v) MD5Update(&syn_ctx, (u_char *)&v, sizeof(v)) struct md5_add { u_int32_t laddr, faddr; u_int32_t secbits[4]; u_int16_t lport, fport; }; #ifdef CTASSERT CTASSERT(sizeof(struct md5_add) == 28); #endif /* * Consider the problem of a recreated (and retransmitted) cookie. If the * original SYN was accepted, the connection is established. The second * SYN is inflight, and if it arrives with an ISN that falls within the * receive window, the connection is killed. * * However, since cookies have other problems, this may not be worth * worrying about. */ static u_int32_t syncookie_generate(struct syncache *sc) { u_int32_t md5_buffer[4]; u_int32_t data; int idx, i; struct md5_add add; /* NB: single threaded; could add INP_INFO_WLOCK_ASSERT(&tcbinfo) */ idx = ((ticks << SYNCOOKIE_TIMESHIFT) / hz) & SYNCOOKIE_WNDMASK; if (tcp_secret[idx].ts_expire < ticks) { for (i = 0; i < 4; i++) tcp_secret[idx].ts_secbits[i] = arc4random(); tcp_secret[idx].ts_expire = ticks + SYNCOOKIE_TIMEOUT; } for (data = sizeof(tcp_msstab) / sizeof(int) - 1; data > 0; data--) if (tcp_msstab[data] <= sc->sc_peer_mss) break; data = (data << SYNCOOKIE_WNDBITS) | idx; data ^= sc->sc_irs; /* peer's iss */ MD5Init(&syn_ctx); #ifdef INET6 if (sc->sc_inc.inc_isipv6) { MD5Add(sc->sc_inc.inc6_laddr); MD5Add(sc->sc_inc.inc6_faddr); add.laddr = 0; add.faddr = 0; } else #endif { add.laddr = sc->sc_inc.inc_laddr.s_addr; add.faddr = sc->sc_inc.inc_faddr.s_addr; } add.lport = sc->sc_inc.inc_lport; add.fport = sc->sc_inc.inc_fport; add.secbits[0] = tcp_secret[idx].ts_secbits[0]; add.secbits[1] = tcp_secret[idx].ts_secbits[1]; add.secbits[2] = tcp_secret[idx].ts_secbits[2]; add.secbits[3] = tcp_secret[idx].ts_secbits[3]; MD5Add(add); MD5Final((u_char *)&md5_buffer, &syn_ctx); data ^= (md5_buffer[0] & ~SYNCOOKIE_WNDMASK); return (data); } static struct syncache * syncookie_lookup(inc, th, so) struct in_conninfo *inc; struct tcphdr *th; struct socket *so; { u_int32_t md5_buffer[4]; struct syncache *sc; u_int32_t data; int wnd, idx; struct md5_add add; /* NB: single threaded; could add INP_INFO_WLOCK_ASSERT(&tcbinfo) */ data = (th->th_ack - 1) ^ (th->th_seq - 1); /* remove ISS */ idx = data & SYNCOOKIE_WNDMASK; if (tcp_secret[idx].ts_expire < ticks || sototcpcb(so)->ts_recent + SYNCOOKIE_TIMEOUT < ticks) return (NULL); MD5Init(&syn_ctx); #ifdef INET6 if (inc->inc_isipv6) { MD5Add(inc->inc6_laddr); MD5Add(inc->inc6_faddr); add.laddr = 0; add.faddr = 0; } else #endif { add.laddr = inc->inc_laddr.s_addr; add.faddr = inc->inc_faddr.s_addr; } add.lport = inc->inc_lport; add.fport = inc->inc_fport; add.secbits[0] = tcp_secret[idx].ts_secbits[0]; add.secbits[1] = tcp_secret[idx].ts_secbits[1]; add.secbits[2] = tcp_secret[idx].ts_secbits[2]; add.secbits[3] = tcp_secret[idx].ts_secbits[3]; MD5Add(add); MD5Final((u_char *)&md5_buffer, &syn_ctx); data ^= md5_buffer[0]; if ((data & ~SYNCOOKIE_DATAMASK) != 0) return (NULL); data = data >> SYNCOOKIE_WNDBITS; sc = uma_zalloc(tcp_syncache.zone, M_NOWAIT); if (sc == NULL) return (NULL); /* * Fill in the syncache values. * XXX duplicate code from syncache_add */ sc->sc_ipopts = NULL; sc->sc_inc.inc_fport = inc->inc_fport; sc->sc_inc.inc_lport = inc->inc_lport; #ifdef INET6 sc->sc_inc.inc_isipv6 = inc->inc_isipv6; if (inc->inc_isipv6) { sc->sc_inc.inc6_faddr = inc->inc6_faddr; sc->sc_inc.inc6_laddr = inc->inc6_laddr; } else #endif { sc->sc_inc.inc_faddr = inc->inc_faddr; sc->sc_inc.inc_laddr = inc->inc_laddr; } sc->sc_irs = th->th_seq - 1; sc->sc_iss = th->th_ack - 1; wnd = sbspace(&so->so_rcv); wnd = imax(wnd, 0); wnd = imin(wnd, TCP_MAXWIN); sc->sc_wnd = wnd; sc->sc_flags = 0; sc->sc_rxtslot = 0; sc->sc_peer_mss = tcp_msstab[data]; return (sc); } diff --git a/sys/netinet/tcp_timewait.c b/sys/netinet/tcp_timewait.c index 23f08bafc9c3..5fca04c85115 100644 --- a/sys/netinet/tcp_timewait.c +++ b/sys/netinet/tcp_timewait.c @@ -1,2091 +1,2096 @@ /* * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 * $FreeBSD$ */ #include "opt_compat.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_mac.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef INET6 #include #endif #include #include #ifdef INET6 #include #include #endif #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef TCPDEBUG #include #endif #include #ifdef IPSEC #include #ifdef INET6 #include #endif #endif /*IPSEC*/ #ifdef FAST_IPSEC #include #include #ifdef INET6 #include #endif #include #define IPSEC #endif /*FAST_IPSEC*/ #include #include int tcp_mssdflt = TCP_MSS; SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW, &tcp_mssdflt , 0, "Default TCP Maximum Segment Size"); #ifdef INET6 int tcp_v6mssdflt = TCP6_MSS; SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, CTLFLAG_RW, &tcp_v6mssdflt , 0, "Default TCP Maximum Segment Size for IPv6"); #endif /* * Minimum MSS we accept and use. This prevents DoS attacks where * we are forced to a ridiculous low MSS like 20 and send hundreds * of packets instead of one. The effect scales with the available * bandwidth and quickly saturates the CPU and network interface * with packet generation and sending. Set to zero to disable MINMSS * checking. This setting prevents us from sending too small packets. */ int tcp_minmss = TCP_MINMSS; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_RW, &tcp_minmss , 0, "Minmum TCP Maximum Segment Size"); /* * Number of TCP segments per second we accept from remote host * before we start to calculate average segment size. If average * segment size drops below the minimum TCP MSS we assume a DoS * attack and reset+drop the connection. Care has to be taken not to * set this value too small to not kill interactive type connections * (telnet, SSH) which send many small packets. */ int tcp_minmssoverload = TCP_MINMSSOVERLOAD; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmssoverload, CTLFLAG_RW, &tcp_minmssoverload , 0, "Number of TCP Segments per Second allowed to" "be under the MINMSS Size"); #if 0 static int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; SYSCTL_INT(_net_inet_tcp, TCPCTL_RTTDFLT, rttdflt, CTLFLAG_RW, &tcp_rttdflt , 0, "Default maximum TCP Round Trip Time"); #endif int tcp_do_rfc1323 = 1; SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW, &tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions"); int tcp_do_rfc1644 = 0; SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1644, rfc1644, CTLFLAG_RW, &tcp_do_rfc1644 , 0, "Enable rfc1644 (TTCP) extensions"); static int tcp_tcbhashsize = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RDTUN, &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); static int do_tcpdrain = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, "Enable tcp_drain routine for extra help when low on mbufs"); SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD, &tcbinfo.ipi_count, 0, "Number of active PCBs"); static int icmp_may_rst = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW, &icmp_may_rst, 0, "Certain ICMP unreachable messages may abort connections in SYN_SENT"); static int tcp_isn_reseed_interval = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_RW, &tcp_isn_reseed_interval, 0, "Seconds between reseeding of ISN secret"); /* * TCP bandwidth limiting sysctls. Note that the default lower bound of * 1024 exists only for debugging. A good production default would be * something like 6100. */ static int tcp_inflight_enable = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, inflight_enable, CTLFLAG_RW, &tcp_inflight_enable, 0, "Enable automatic TCP inflight data limiting"); static int tcp_inflight_debug = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, inflight_debug, CTLFLAG_RW, &tcp_inflight_debug, 0, "Debug TCP inflight calculations"); static int tcp_inflight_min = 6144; SYSCTL_INT(_net_inet_tcp, OID_AUTO, inflight_min, CTLFLAG_RW, &tcp_inflight_min, 0, "Lower-bound for TCP inflight window"); static int tcp_inflight_max = TCP_MAXWIN << TCP_MAX_WINSHIFT; SYSCTL_INT(_net_inet_tcp, OID_AUTO, inflight_max, CTLFLAG_RW, &tcp_inflight_max, 0, "Upper-bound for TCP inflight window"); static int tcp_inflight_stab = 20; SYSCTL_INT(_net_inet_tcp, OID_AUTO, inflight_stab, CTLFLAG_RW, &tcp_inflight_stab, 0, "Inflight Algorithm Stabilization 20 = 2 packets"); static struct inpcb *tcp_notify(struct inpcb *, int); static void tcp_discardcb(struct tcpcb *); static void tcp_isn_tick(void *); /* * Target size of TCP PCB hash tables. Must be a power of two. * * Note that this can be overridden by the kernel environment * variable net.inet.tcp.tcbhashsize */ #ifndef TCBHASHSIZE #define TCBHASHSIZE 512 #endif /* * XXX * Callouts should be moved into struct tcp directly. They are currently * separate because the tcpcb structure is exported to userland for sysctl * parsing purposes, which do not know about callouts. */ struct tcpcb_mem { struct tcpcb tcb; struct callout tcpcb_mem_rexmt, tcpcb_mem_persist, tcpcb_mem_keep; struct callout tcpcb_mem_2msl, tcpcb_mem_delack; }; static uma_zone_t tcpcb_zone; static uma_zone_t tcptw_zone; struct callout isn_callout; /* * Tcp initialization */ void tcp_init() { int hashsize = TCBHASHSIZE; tcp_ccgen = 1; tcp_delacktime = TCPTV_DELACK; tcp_keepinit = TCPTV_KEEP_INIT; tcp_keepidle = TCPTV_KEEP_IDLE; tcp_keepintvl = TCPTV_KEEPINTVL; tcp_maxpersistidle = TCPTV_KEEP_IDLE; tcp_msl = TCPTV_MSL; tcp_rexmit_min = TCPTV_MIN; tcp_rexmit_slop = TCPTV_CPU_VAR; INP_INFO_LOCK_INIT(&tcbinfo, "tcp"); LIST_INIT(&tcb); tcbinfo.listhead = &tcb; TUNABLE_INT_FETCH("net.inet.tcp.tcbhashsize", &hashsize); if (!powerof2(hashsize)) { printf("WARNING: TCB hash size not a power of 2\n"); hashsize = 512; /* safe default */ } tcp_tcbhashsize = hashsize; tcbinfo.hashbase = hashinit(hashsize, M_PCB, &tcbinfo.hashmask); tcbinfo.porthashbase = hashinit(hashsize, M_PCB, &tcbinfo.porthashmask); tcbinfo.ipi_zone = uma_zcreate("inpcb", sizeof(struct inpcb), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcbinfo.ipi_zone, maxsockets); #ifdef INET6 #define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) #else /* INET6 */ #define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) #endif /* INET6 */ if (max_protohdr < TCP_MINPROTOHDR) max_protohdr = TCP_MINPROTOHDR; if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) panic("tcp_init"); #undef TCP_MINPROTOHDR /* * These have to be type stable for the benefit of the timers. */ tcpcb_zone = uma_zcreate("tcpcb", sizeof(struct tcpcb_mem), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcpcb_zone, maxsockets); tcptw_zone = uma_zcreate("tcptw", sizeof(struct tcptw), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcptw_zone, maxsockets / 5); tcp_timer_init(); syncache_init(); tcp_hc_init(); tcp_reass_init(); callout_init(&isn_callout, CALLOUT_MPSAFE); tcp_isn_tick(NULL); EVENTHANDLER_REGISTER(shutdown_pre_sync, tcp_fini, NULL, SHUTDOWN_PRI_DEFAULT); } void tcp_fini(xtp) void *xtp; { callout_stop(&isn_callout); } /* * Fill in the IP and TCP headers for an outgoing packet, given the tcpcb. * tcp_template used to store this data in mbufs, but we now recopy it out * of the tcpcb each time to conserve mbufs. */ void tcpip_fillheaders(inp, ip_ptr, tcp_ptr) struct inpcb *inp; void *ip_ptr; void *tcp_ptr; { struct tcphdr *th = (struct tcphdr *)tcp_ptr; #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { struct ip6_hdr *ip6; ip6 = (struct ip6_hdr *)ip_ptr; ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | (inp->in6p_flowinfo & IPV6_FLOWINFO_MASK); ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | (IPV6_VERSION & IPV6_VERSION_MASK); ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = sizeof(struct tcphdr); ip6->ip6_src = inp->in6p_laddr; ip6->ip6_dst = inp->in6p_faddr; } else #endif { struct ip *ip; ip = (struct ip *)ip_ptr; ip->ip_v = IPVERSION; ip->ip_hl = 5; ip->ip_tos = inp->inp_ip_tos; ip->ip_len = 0; ip->ip_id = 0; ip->ip_off = 0; ip->ip_ttl = inp->inp_ip_ttl; ip->ip_sum = 0; ip->ip_p = IPPROTO_TCP; ip->ip_src = inp->inp_laddr; ip->ip_dst = inp->inp_faddr; } th->th_sport = inp->inp_lport; th->th_dport = inp->inp_fport; th->th_seq = 0; th->th_ack = 0; th->th_x2 = 0; th->th_off = 5; th->th_flags = 0; th->th_win = 0; th->th_urp = 0; th->th_sum = 0; /* in_pseudo() is called later for ipv4 */ } /* * Create template to be used to send tcp packets on a connection. * Allocates an mbuf and fills in a skeletal tcp/ip header. The only * use for this function is in keepalives, which use tcp_respond. */ struct tcptemp * tcpip_maketemplate(inp) struct inpcb *inp; { struct mbuf *m; struct tcptemp *n; m = m_get(M_DONTWAIT, MT_HEADER); if (m == NULL) return (0); m->m_len = sizeof(struct tcptemp); n = mtod(m, struct tcptemp *); tcpip_fillheaders(inp, (void *)&n->tt_ipgen, (void *)&n->tt_t); return (n); } /* * Send a single message to the TCP at address specified by * the given TCP/IP header. If m == NULL, then we make a copy * of the tcpiphdr at ti and send directly to the addressed host. * This is used to force keep alive messages out using the TCP * template for a connection. If flags are given then we send * a message back to the TCP which originated the * segment ti, * and discard the mbuf containing it and any other attached mbufs. * * In any case the ack and sequence number of the transmitted * segment are as specified by the parameters. * * NOTE: If m != NULL, then ti must point to *inside* the mbuf. */ void tcp_respond(tp, ipgen, th, m, ack, seq, flags) struct tcpcb *tp; void *ipgen; register struct tcphdr *th; register struct mbuf *m; tcp_seq ack, seq; int flags; { register int tlen; int win = 0; struct ip *ip; struct tcphdr *nth; #ifdef INET6 struct ip6_hdr *ip6; int isipv6; #endif /* INET6 */ int ipflags = 0; - struct inpcb *inp = NULL; + struct inpcb *inp; KASSERT(tp != NULL || m != NULL, ("tcp_respond: tp and m both NULL")); #ifdef INET6 isipv6 = ((struct ip *)ipgen)->ip_v == 6; ip6 = ipgen; #endif /* INET6 */ ip = ipgen; if (tp != NULL) { inp = tp->t_inpcb; KASSERT(inp != NULL, ("tcp control block w/o inpcb")); INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); + } else + inp = NULL; + + if (tp != NULL) { if (!(flags & TH_RST)) { win = sbspace(&inp->inp_socket->so_rcv); if (win > (long)TCP_MAXWIN << tp->rcv_scale) win = (long)TCP_MAXWIN << tp->rcv_scale; } } if (m == NULL) { m = m_gethdr(M_DONTWAIT, MT_HEADER); if (m == NULL) return; tlen = 0; m->m_data += max_linkhdr; #ifdef INET6 if (isipv6) { bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(struct ip6_hdr)); ip6 = mtod(m, struct ip6_hdr *); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); ip = mtod(m, struct ip *); nth = (struct tcphdr *)(ip + 1); } bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); flags = TH_ACK; } else { m_freem(m->m_next); m->m_next = NULL; m->m_data = (caddr_t)ipgen; /* m_len is set later */ tlen = 0; #define xchg(a,b,type) { type t; t=a; a=b; b=t; } #ifdef INET6 if (isipv6) { xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long); nth = (struct tcphdr *)(ip + 1); } if (th != nth) { /* * this is usually a case when an extension header * exists between the IPv6 header and the * TCP header. */ nth->th_sport = th->th_sport; nth->th_dport = th->th_dport; } xchg(nth->th_dport, nth->th_sport, n_short); #undef xchg } #ifdef INET6 if (isipv6) { ip6->ip6_flow = 0; ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) + tlen)); tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); } else #endif { tlen += sizeof (struct tcpiphdr); ip->ip_len = tlen; ip->ip_ttl = ip_defttl; if (path_mtu_discovery) ip->ip_off |= IP_DF; } m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; #ifdef MAC if (inp != NULL) { /* * Packet is associated with a socket, so allow the * label of the response to reflect the socket label. */ - mac_create_mbuf_from_socket(inp->inp_socket, m); + INP_LOCK_ASSERT(inp); + mac_create_mbuf_from_inpcb(inp, m); } else { /* * Packet is not associated with a socket, so possibly * update the label in place. */ mac_reflect_mbuf_tcp(m); } #endif nth->th_seq = htonl(seq); nth->th_ack = htonl(ack); nth->th_x2 = 0; nth->th_off = sizeof (struct tcphdr) >> 2; nth->th_flags = flags; if (tp != NULL) nth->th_win = htons((u_short) (win >> tp->rcv_scale)); else nth->th_win = htons((u_short)win); nth->th_urp = 0; #ifdef INET6 if (isipv6) { nth->th_sum = 0; nth->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), tlen - sizeof(struct ip6_hdr)); ip6->ip6_hlim = in6_selecthlim(tp != NULL ? tp->t_inpcb : NULL, NULL); } else #endif /* INET6 */ { nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p))); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); } #ifdef TCPDEBUG if (tp == NULL || (inp->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); #endif #ifdef INET6 if (isipv6) (void) ip6_output(m, NULL, NULL, ipflags, NULL, NULL, inp); else #endif /* INET6 */ (void) ip_output(m, NULL, NULL, ipflags, NULL, inp); } /* * Create a new TCP control block, making an * empty reassembly queue and hooking it to the argument * protocol control block. The `inp' parameter must have * come from the zone allocator set up in tcp_init(). */ struct tcpcb * tcp_newtcpcb(inp) struct inpcb *inp; { struct tcpcb_mem *tm; struct tcpcb *tp; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ tm = uma_zalloc(tcpcb_zone, M_NOWAIT | M_ZERO); if (tm == NULL) return (NULL); tp = &tm->tcb; /* LIST_INIT(&tp->t_segq); */ /* XXX covered by M_ZERO */ tp->t_maxseg = tp->t_maxopd = #ifdef INET6 isipv6 ? tcp_v6mssdflt : #endif /* INET6 */ tcp_mssdflt; /* Set up our timeouts. */ callout_init(tp->tt_rexmt = &tm->tcpcb_mem_rexmt, 0); callout_init(tp->tt_persist = &tm->tcpcb_mem_persist, 0); callout_init(tp->tt_keep = &tm->tcpcb_mem_keep, 0); callout_init(tp->tt_2msl = &tm->tcpcb_mem_2msl, 0); callout_init(tp->tt_delack = &tm->tcpcb_mem_delack, 0); if (tcp_do_rfc1323) tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); if (tcp_do_rfc1644) tp->t_flags |= TF_REQ_CC; tp->t_inpcb = inp; /* XXX */ /* * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives * reasonable initial retransmit time. */ tp->t_srtt = TCPTV_SRTTBASE; tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; tp->t_rttmin = tcp_rexmit_min; tp->t_rxtcur = TCPTV_RTOBASE; tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->t_rcvtime = ticks; tp->t_bw_rtttime = ticks; /* * IPv4 TTL initialization is necessary for an IPv6 socket as well, * because the socket may be bound to an IPv6 wildcard address, * which may match an IPv4-mapped IPv6 address. */ inp->inp_ip_ttl = ip_defttl; inp->inp_ppcb = (caddr_t)tp; return (tp); /* XXX */ } /* * Drop a TCP connection, reporting * the specified error. If connection is synchronized, * then send a RST to peer. */ struct tcpcb * tcp_drop(tp, errno) register struct tcpcb *tp; int errno; { struct socket *so = tp->t_inpcb->inp_socket; if (TCPS_HAVERCVDSYN(tp->t_state)) { tp->t_state = TCPS_CLOSED; (void) tcp_output(tp); tcpstat.tcps_drops++; } else tcpstat.tcps_conndrops++; if (errno == ETIMEDOUT && tp->t_softerror) errno = tp->t_softerror; so->so_error = errno; return (tcp_close(tp)); } static void tcp_discardcb(tp) struct tcpcb *tp; { struct tseg_qent *q; struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ /* * Make sure that all of our timers are stopped before we * delete the PCB. */ callout_stop(tp->tt_rexmt); callout_stop(tp->tt_persist); callout_stop(tp->tt_keep); callout_stop(tp->tt_2msl); callout_stop(tp->tt_delack); /* * If we got enough samples through the srtt filter, * save the rtt and rttvar in the routing entry. * 'Enough' is arbitrarily defined as 4 rtt samples. * 4 samples is enough for the srtt filter to converge * to within enough % of the correct value; fewer samples * and we could save a bogus rtt. The danger is not high * as tcp quickly recovers from everything. * XXX: Works very well but needs some more statistics! */ if (tp->t_rttupdated >= 4) { struct hc_metrics_lite metrics; u_long ssthresh; bzero(&metrics, sizeof(metrics)); /* * Update the ssthresh always when the conditions below * are satisfied. This gives us better new start value * for the congestion avoidance for new connections. * ssthresh is only set if packet loss occured on a session. */ ssthresh = tp->snd_ssthresh; if (ssthresh != 0 && ssthresh < so->so_snd.sb_hiwat / 2) { /* * convert the limit from user data bytes to * packets then to packet data bytes. */ ssthresh = (ssthresh + tp->t_maxseg / 2) / tp->t_maxseg; if (ssthresh < 2) ssthresh = 2; ssthresh *= (u_long)(tp->t_maxseg + #ifdef INET6 (isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : #endif sizeof (struct tcpiphdr) #ifdef INET6 ) #endif ); } else ssthresh = 0; metrics.rmx_ssthresh = ssthresh; metrics.rmx_rtt = tp->t_srtt; metrics.rmx_rttvar = tp->t_rttvar; /* XXX: This wraps if the pipe is more than 4 Gbit per second */ metrics.rmx_bandwidth = tp->snd_bandwidth; metrics.rmx_cwnd = tp->snd_cwnd; metrics.rmx_sendpipe = 0; metrics.rmx_recvpipe = 0; tcp_hc_update(&inp->inp_inc, &metrics); } /* free the reassembly queue, if any */ while ((q = LIST_FIRST(&tp->t_segq)) != NULL) { LIST_REMOVE(q, tqe_q); m_freem(q->tqe_m); uma_zfree(tcp_reass_zone, q); tp->t_segqlen--; tcp_reass_qsize--; } inp->inp_ppcb = NULL; tp->t_inpcb = NULL; uma_zfree(tcpcb_zone, tp); soisdisconnected(so); } /* * Close a TCP control block: * discard all space held by the tcp * discard internet protocol block * wake up any sleepers */ struct tcpcb * tcp_close(tp) struct tcpcb *tp; { struct inpcb *inp = tp->t_inpcb; #ifdef INET6 struct socket *so = inp->inp_socket; #endif tcp_discardcb(tp); #ifdef INET6 if (INP_CHECK_SOCKAF(so, AF_INET6)) in6_pcbdetach(inp); else #endif in_pcbdetach(inp); tcpstat.tcps_closed++; return (NULL); } void tcp_drain() { if (do_tcpdrain) { struct inpcb *inpb; struct tcpcb *tcpb; struct tseg_qent *te; /* * Walk the tcpbs, if existing, and flush the reassembly queue, * if there is one... * XXX: The "Net/3" implementation doesn't imply that the TCP * reassembly queue should be flushed, but in a situation * where we're really low on mbufs, this is potentially * usefull. */ INP_INFO_RLOCK(&tcbinfo); LIST_FOREACH(inpb, tcbinfo.listhead, inp_list) { if (inpb->inp_vflag & INP_TIMEWAIT) continue; INP_LOCK(inpb); if ((tcpb = intotcpcb(inpb)) != NULL) { while ((te = LIST_FIRST(&tcpb->t_segq)) != NULL) { LIST_REMOVE(te, tqe_q); m_freem(te->tqe_m); uma_zfree(tcp_reass_zone, te); tcpb->t_segqlen--; tcp_reass_qsize--; } } INP_UNLOCK(inpb); } INP_INFO_RUNLOCK(&tcbinfo); } } /* * Notify a tcp user of an asynchronous error; * store error as soft error, but wake up user * (for now, won't do anything until can select for soft error). * * Do not wake up user since there currently is no mechanism for * reporting soft errors (yet - a kqueue filter may be added). */ static struct inpcb * tcp_notify(inp, error) struct inpcb *inp; int error; { struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb; /* * Ignore some errors if we are hooked up. * If connection hasn't completed, has retransmitted several times, * and receives a second error, give up now. This is better * than waiting a long time to establish a connection that * can never complete. */ if (tp->t_state == TCPS_ESTABLISHED && (error == EHOSTUNREACH || error == ENETUNREACH || error == EHOSTDOWN)) { return inp; } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && tp->t_softerror) { tcp_drop(tp, error); return (struct inpcb *)0; } else { tp->t_softerror = error; return inp; } #if 0 wakeup( &so->so_timeo); sorwakeup(so); sowwakeup(so); #endif } static int tcp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n, s; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == NULL) { n = tcbinfo.ipi_count; req->oldidx = 2 * (sizeof xig) + (n + n/8) * sizeof(struct xtcpcb); return 0; } if (req->newptr != NULL) return EPERM; /* * OK, now we're committed to doing something. */ s = splnet(); INP_INFO_RLOCK(&tcbinfo); gencnt = tcbinfo.ipi_gencnt; n = tcbinfo.ipi_count; INP_INFO_RUNLOCK(&tcbinfo); splx(s); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + n * sizeof(struct xtcpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return error; inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == NULL) return ENOMEM; s = splnet(); INP_INFO_RLOCK(&tcbinfo); for (inp = LIST_FIRST(tcbinfo.listhead), i = 0; inp != NULL && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_LOCK(inp); if (inp->inp_gencnt <= gencnt) { /* * XXX: This use of cr_cansee(), introduced with * TCP state changes, is not quite right, but for * now, better than nothing. */ if (inp->inp_vflag & INP_TIMEWAIT) error = cr_cansee(req->td->td_ucred, intotw(inp)->tw_cred); else error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error == 0) inp_list[i++] = inp; } INP_UNLOCK(inp); } INP_INFO_RUNLOCK(&tcbinfo); splx(s); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; if (inp->inp_gencnt <= gencnt) { struct xtcpcb xt; caddr_t inp_ppcb; xt.xt_len = sizeof xt; /* XXX should avoid extra copy */ bcopy(inp, &xt.xt_inp, sizeof *inp); inp_ppcb = inp->inp_ppcb; if (inp_ppcb == NULL) bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); else if (inp->inp_vflag & INP_TIMEWAIT) { bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); xt.xt_tp.t_state = TCPS_TIME_WAIT; } else bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); if (inp->inp_socket != NULL) sotoxsocket(inp->inp_socket, &xt.xt_socket); else { bzero(&xt.xt_socket, sizeof xt.xt_socket); xt.xt_socket.xso_protocol = IPPROTO_TCP; } xt.xt_inp.inp_gencnt = inp->inp_gencnt; error = SYSCTL_OUT(req, &xt, sizeof xt); } } if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ s = splnet(); INP_INFO_RLOCK(&tcbinfo); xig.xig_gen = tcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = tcbinfo.ipi_count; INP_INFO_RUNLOCK(&tcbinfo); splx(s); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return error; } SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); static int tcp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error, s; error = suser_cred(req->td->td_ucred, PRISON_ROOT); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); s = splnet(); INP_INFO_RLOCK(&tcbinfo); inp = in_pcblookup_hash(&tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, 0, NULL); if (inp == NULL) { error = ENOENT; goto outunlocked; } INP_LOCK(inp); if (inp->inp_socket == NULL) { error = ENOENT; goto out; } error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error) goto out; cru2x(inp->inp_socket->so_cred, &xuc); out: INP_UNLOCK(inp); outunlocked: INP_INFO_RUNLOCK(&tcbinfo); splx(s); if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp_getcred, "S,xucred", "Get the xucred of a TCP connection"); #ifdef INET6 static int tcp6_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in6 addrs[2]; struct inpcb *inp; int error, s, mapped = 0; error = suser_cred(req->td->td_ucred, PRISON_ROOT); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) mapped = 1; else return (EINVAL); } s = splnet(); INP_INFO_RLOCK(&tcbinfo); if (mapped == 1) inp = in_pcblookup_hash(&tcbinfo, *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], addrs[1].sin6_port, *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], addrs[0].sin6_port, 0, NULL); else inp = in6_pcblookup_hash(&tcbinfo, &addrs[1].sin6_addr, addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 0, NULL); if (inp == NULL) { error = ENOENT; goto outunlocked; } INP_LOCK(inp); if (inp->inp_socket == NULL) { error = ENOENT; goto out; } error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error) goto out; cru2x(inp->inp_socket->so_cred, &xuc); out: INP_UNLOCK(inp); outunlocked: INP_INFO_RUNLOCK(&tcbinfo); splx(s); if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection"); #endif void tcp_ctlinput(cmd, sa, vip) int cmd; struct sockaddr *sa; void *vip; { struct ip *ip = vip; struct tcphdr *th; struct in_addr faddr; struct inpcb *inp; struct tcpcb *tp; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; tcp_seq icmp_seq; int s; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (cmd == PRC_QUENCH) notify = tcp_quench; else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB || cmd == PRC_UNREACH_PORT || cmd == PRC_TIMXCEED_INTRANS) && ip) notify = tcp_drop_syn_sent; else if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; /* * Redirects don't need to be handled up here. */ else if (PRC_IS_REDIRECT(cmd)) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * XXX: We never get this from ICMP, otherwise it makes an * excellent DoS attack on machines with many connections. */ else if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip != NULL) { s = splnet(); th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); INP_INFO_WLOCK(&tcbinfo); inp = in_pcblookup_hash(&tcbinfo, faddr, th->th_dport, ip->ip_src, th->th_sport, 0, NULL); if (inp != NULL) { INP_LOCK(inp); if (inp->inp_socket != NULL) { icmp_seq = htonl(th->th_seq); tp = intotcpcb(inp); if (SEQ_GEQ(icmp_seq, tp->snd_una) && SEQ_LT(icmp_seq, tp->snd_max)) inp = (*notify)(inp, inetctlerrmap[cmd]); } if (inp != NULL) INP_UNLOCK(inp); } else { struct in_conninfo inc; inc.inc_fport = th->th_dport; inc.inc_lport = th->th_sport; inc.inc_faddr = faddr; inc.inc_laddr = ip->ip_src; #ifdef INET6 inc.inc_isipv6 = 0; #endif syncache_unreach(&inc, th); } INP_INFO_WUNLOCK(&tcbinfo); splx(s); } else in_pcbnotifyall(&tcbinfo, faddr, inetctlerrmap[cmd], notify); } #ifdef INET6 void tcp6_ctlinput(cmd, sa, d) int cmd; struct sockaddr *sa; void *d; { struct tcphdr th; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct ip6_hdr *ip6; struct mbuf *m; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; int off; struct tcp_portonly { u_int16_t th_sport; u_int16_t th_dport; } *thp; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; if (cmd == PRC_QUENCH) notify = tcp_quench; else if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; else if (!PRC_IS_REDIRECT(cmd) && ((unsigned)cmd >= PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) return; /* if the parameter is from icmp6, decode it. */ if (d != NULL) { ip6cp = (struct ip6ctlparam *)d; m = ip6cp->ip6c_m; ip6 = ip6cp->ip6c_ip6; off = ip6cp->ip6c_off; sa6_src = ip6cp->ip6c_src; } else { m = NULL; ip6 = NULL; off = 0; /* fool gcc */ sa6_src = &sa6_any; } if (ip6 != NULL) { struct in_conninfo inc; /* * XXX: We assume that when IPV6 is non NULL, * M and OFF are valid. */ /* check if we can safely examine src and dst ports */ if (m->m_pkthdr.len < off + sizeof(*thp)) return; bzero(&th, sizeof(th)); m_copydata(m, off, sizeof(*thp), (caddr_t)&th); in6_pcbnotify(&tcb, sa, th.th_dport, (struct sockaddr *)ip6cp->ip6c_src, th.th_sport, cmd, NULL, notify); inc.inc_fport = th.th_dport; inc.inc_lport = th.th_sport; inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr; inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr; inc.inc_isipv6 = 1; syncache_unreach(&inc, &th); } else in6_pcbnotify(&tcb, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify); } #endif /* INET6 */ /* * Following is where TCP initial sequence number generation occurs. * * There are two places where we must use initial sequence numbers: * 1. In SYN-ACK packets. * 2. In SYN packets. * * All ISNs for SYN-ACK packets are generated by the syncache. See * tcp_syncache.c for details. * * The ISNs in SYN packets must be monotonic; TIME_WAIT recycling * depends on this property. In addition, these ISNs should be * unguessable so as to prevent connection hijacking. To satisfy * the requirements of this situation, the algorithm outlined in * RFC 1948 is used, with only small modifications. * * Implementation details: * * Time is based off the system timer, and is corrected so that it * increases by one megabyte per second. This allows for proper * recycling on high speed LANs while still leaving over an hour * before rollover. * * As reading the *exact* system time is too expensive to be done * whenever setting up a TCP connection, we increment the time * offset in two ways. First, a small random positive increment * is added to isn_offset for each connection that is set up. * Second, the function tcp_isn_tick fires once per clock tick * and increments isn_offset as necessary so that sequence numbers * are incremented at approximately ISN_BYTES_PER_SECOND. The * random positive increments serve only to ensure that the same * exact sequence number is never sent out twice (as could otherwise * happen when a port is recycled in less than the system tick * interval.) * * net.inet.tcp.isn_reseed_interval controls the number of seconds * between seeding of isn_secret. This is normally set to zero, * as reseeding should not be necessary. * */ #define ISN_BYTES_PER_SECOND 1048576 #define ISN_STATIC_INCREMENT 4096 #define ISN_RANDOM_INCREMENT (4096 - 1) u_char isn_secret[32]; int isn_last_reseed; u_int32_t isn_offset, isn_offset_old; MD5_CTX isn_ctx; tcp_seq tcp_new_isn(tp) struct tcpcb *tp; { u_int32_t md5_buffer[4]; tcp_seq new_isn; /* Seed if this is the first use, reseed if requested. */ if ((isn_last_reseed == 0) || ((tcp_isn_reseed_interval > 0) && (((u_int)isn_last_reseed + (u_int)tcp_isn_reseed_interval*hz) < (u_int)ticks))) { read_random(&isn_secret, sizeof(isn_secret)); isn_last_reseed = ticks; } /* Compute the md5 hash and return the ISN. */ MD5Init(&isn_ctx); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short)); #ifdef INET6 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr, sizeof(struct in6_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr, sizeof(struct in6_addr)); } else #endif { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr, sizeof(struct in_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr, sizeof(struct in_addr)); } MD5Update(&isn_ctx, (u_char *) &isn_secret, sizeof(isn_secret)); MD5Final((u_char *) &md5_buffer, &isn_ctx); new_isn = (tcp_seq) md5_buffer[0]; isn_offset += ISN_STATIC_INCREMENT + (arc4random() & ISN_RANDOM_INCREMENT); new_isn += isn_offset; return new_isn; } /* * Increment the offset to the next ISN_BYTES_PER_SECOND / hz boundary * to keep time flowing at a relatively constant rate. If the random * increments have already pushed us past the projected offset, do nothing. */ static void tcp_isn_tick(xtp) void *xtp; { u_int32_t projected_offset; projected_offset = isn_offset_old + ISN_BYTES_PER_SECOND / hz; if (projected_offset > isn_offset) isn_offset = projected_offset; isn_offset_old = isn_offset; callout_reset(&isn_callout, 1, tcp_isn_tick, NULL); } /* * When a source quench is received, close congestion window * to one segment. We will gradually open it again as we proceed. */ struct inpcb * tcp_quench(inp, errno) struct inpcb *inp; int errno; { struct tcpcb *tp = intotcpcb(inp); if (tp != NULL) tp->snd_cwnd = tp->t_maxseg; return (inp); } /* * When a specific ICMP unreachable message is received and the * connection state is SYN-SENT, drop the connection. This behavior * is controlled by the icmp_may_rst sysctl. */ struct inpcb * tcp_drop_syn_sent(inp, errno) struct inpcb *inp; int errno; { struct tcpcb *tp = intotcpcb(inp); if (tp != NULL && tp->t_state == TCPS_SYN_SENT) { tcp_drop(tp, errno); return (struct inpcb *)0; } return inp; } /* * When `need fragmentation' ICMP is received, update our idea of the MSS * based on the new value in the route. Also nudge TCP to send something, * since we know the packet we just sent was dropped. * This duplicates some code in the tcp_mss() function in tcp_input.c. */ struct inpcb * tcp_mtudisc(inp, errno) struct inpcb *inp; int errno; { struct tcpcb *tp = intotcpcb(inp); struct rmxp_tao tao; struct socket *so = inp->inp_socket; u_int maxmtu; u_int romtu; int mss; #ifdef INET6 int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ bzero(&tao, sizeof(tao)); if (tp != NULL) { maxmtu = tcp_hc_getmtu(&inp->inp_inc); /* IPv4 and IPv6 */ romtu = #ifdef INET6 isipv6 ? tcp_maxmtu6(&inp->inp_inc) : #endif /* INET6 */ tcp_maxmtu(&inp->inp_inc); if (!maxmtu) maxmtu = romtu; else maxmtu = min(maxmtu, romtu); if (!maxmtu) { tp->t_maxopd = tp->t_maxseg = #ifdef INET6 isipv6 ? tcp_v6mssdflt : #endif /* INET6 */ tcp_mssdflt; return inp; } mss = maxmtu - #ifdef INET6 (isipv6 ? sizeof(struct ip6_hdr) + sizeof(struct tcphdr) : #endif /* INET6 */ sizeof(struct tcpiphdr) #ifdef INET6 ) #endif /* INET6 */ ; if (tcp_do_rfc1644) { tcp_hc_gettao(&inp->inp_inc, &tao); if (tao.tao_mssopt) mss = min(mss, tao.tao_mssopt); } /* * XXX - The above conditional probably violates the TCP * spec. The problem is that, since we don't know the * other end's MSS, we are supposed to use a conservative * default. But, if we do that, then MTU discovery will * never actually take place, because the conservative * default is much less than the MTUs typically seen * on the Internet today. For the moment, we'll sweep * this under the carpet. * * The conservative default might not actually be a problem * if the only case this occurs is when sending an initial * SYN with options and data to a host we've never talked * to before. Then, they will reply with an MSS value which * will get recorded and the new parameters should get * recomputed. For Further Study. */ if (tp->t_maxopd <= mss) return inp; tp->t_maxopd = mss; if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP) mss -= TCPOLEN_TSTAMP_APPA; if ((tp->t_flags & (TF_REQ_CC|TF_NOOPT)) == TF_REQ_CC && (tp->t_flags & TF_RCVD_CC) == TF_RCVD_CC) mss -= TCPOLEN_CC_APPA; #if (MCLBYTES & (MCLBYTES - 1)) == 0 if (mss > MCLBYTES) mss &= ~(MCLBYTES-1); #else if (mss > MCLBYTES) mss = mss / MCLBYTES * MCLBYTES; #endif if (so->so_snd.sb_hiwat < mss) mss = so->so_snd.sb_hiwat; tp->t_maxseg = mss; tcpstat.tcps_mturesent++; tp->t_rtttime = 0; tp->snd_nxt = tp->snd_una; tcp_output(tp); } return inp; } /* * Look-up the routing entry to the peer of this inpcb. If no route * is found and it cannot be allocated, then return NULL. This routine * is called by TCP routines that access the rmx structure and by tcp_mss * to get the interface MTU. */ u_long tcp_maxmtu(inc) struct in_conninfo *inc; { struct route sro; struct sockaddr_in *dst; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer")); bzero(&sro, sizeof(sro)); if (inc->inc_faddr.s_addr != INADDR_ANY) { dst = (struct sockaddr_in *)&sro.ro_dst; dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = inc->inc_faddr; rtalloc_ign(&sro, RTF_CLONING); } if (sro.ro_rt != NULL) { ifp = sro.ro_rt->rt_ifp; if (sro.ro_rt->rt_rmx.rmx_mtu == 0) maxmtu = ifp->if_mtu; else maxmtu = min(sro.ro_rt->rt_rmx.rmx_mtu, ifp->if_mtu); RTFREE(sro.ro_rt); } return (maxmtu); } #ifdef INET6 u_long tcp_maxmtu6(inc) struct in_conninfo *inc; { struct route_in6 sro6; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu6 with NULL in_conninfo pointer")); bzero(&sro6, sizeof(sro6)); if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) { sro6.ro_dst.sin6_family = AF_INET6; sro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); sro6.ro_dst.sin6_addr = inc->inc6_faddr; rtalloc_ign((struct route *)&sro6, RTF_CLONING); } if (sro6.ro_rt != NULL) { ifp = sro6.ro_rt->rt_ifp; if (sro6.ro_rt->rt_rmx.rmx_mtu == 0) maxmtu = IN6_LINKMTU(sro6.ro_rt->rt_ifp); else maxmtu = min(sro6.ro_rt->rt_rmx.rmx_mtu, IN6_LINKMTU(sro6.ro_rt->rt_ifp)); RTFREE(sro6.ro_rt); } return (maxmtu); } #endif /* INET6 */ #ifdef IPSEC /* compute ESP/AH header size for TCP, including outer IP header. */ size_t ipsec_hdrsiz_tcp(tp) struct tcpcb *tp; { struct inpcb *inp; struct mbuf *m; size_t hdrsiz; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; #endif struct tcphdr *th; if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL)) return 0; MGETHDR(m, M_DONTWAIT, MT_DATA); if (!m) return 0; #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); tcpip_fillheaders(inp, ip6, th); hdrsiz = ipsec6_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } else #endif /* INET6 */ { ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); tcpip_fillheaders(inp, ip, th); hdrsiz = ipsec4_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } m_free(m); return hdrsiz; } #endif /*IPSEC*/ /* * Move a TCP connection into TIME_WAIT state. * tcbinfo is unlocked. * inp is locked, and is unlocked before returning. */ void tcp_twstart(tp) struct tcpcb *tp; { struct tcptw *tw; struct inpcb *inp; int tw_time, acknow; struct socket *so; tw = uma_zalloc(tcptw_zone, M_NOWAIT); if (tw == NULL) { tw = tcp_timer_2msl_tw(1); if (tw == NULL) { tcp_close(tp); return; } } inp = tp->t_inpcb; tw->tw_inpcb = inp; /* * Recover last window size sent. */ tw->last_win = (tp->rcv_adv - tp->rcv_nxt) >> tp->rcv_scale; /* * Set t_recent if timestamps are used on the connection. */ if ((tp->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP|TF_NOOPT)) == (TF_REQ_TSTMP|TF_RCVD_TSTMP)) tw->t_recent = tp->ts_recent; else tw->t_recent = 0; tw->snd_nxt = tp->snd_nxt; tw->rcv_nxt = tp->rcv_nxt; tw->iss = tp->iss; tw->irs = tp->irs; tw->cc_recv = tp->cc_recv; tw->cc_send = tp->cc_send; tw->t_starttime = tp->t_starttime; tw->tw_time = 0; /* XXX * If this code will * be used for fin-wait-2 state also, then we may need * a ts_recent from the last segment. */ /* Shorten TIME_WAIT [RFC-1644, p.28] */ if (tp->cc_recv != 0 && (ticks - tp->t_starttime) < tcp_msl) { tw_time = tp->t_rxtcur * TCPTV_TWTRUNC; /* For T/TCP client, force ACK now. */ acknow = 1; } else { tw_time = 2 * tcp_msl; acknow = tp->t_flags & TF_ACKNOW; } tcp_discardcb(tp); so = inp->inp_socket; so->so_pcb = NULL; tw->tw_cred = crhold(so->so_cred); tw->tw_so_options = so->so_options; if (acknow) tcp_twrespond(tw, TH_ACK); sotryfree(so); inp->inp_socket = NULL; inp->inp_ppcb = (caddr_t)tw; inp->inp_vflag |= INP_TIMEWAIT; tcp_timer_2msl_reset(tw, tw_time); INP_UNLOCK(inp); } /* * The appromixate rate of ISN increase of Microsoft TCP stacks; * the actual rate is slightly higher due to the addition of * random positive increments. * * Most other new OSes use semi-randomized ISN values, so we * do not need to worry about them. */ #define MS_ISN_BYTES_PER_SECOND 250000 /* * Determine if the ISN we will generate has advanced beyond the last * sequence number used by the previous connection. If so, indicate * that it is safe to recycle this tw socket by returning 1. */ int tcp_twrecycleable(struct tcptw *tw) { tcp_seq new_iss = tw->iss; tcp_seq new_irs = tw->irs; new_iss += (ticks - tw->t_starttime) * (ISN_BYTES_PER_SECOND / hz); new_irs += (ticks - tw->t_starttime) * (MS_ISN_BYTES_PER_SECOND / hz); if (SEQ_GT(new_iss, tw->snd_nxt) && SEQ_GT(new_irs, tw->rcv_nxt)) return 1; else return 0; } struct tcptw * tcp_twclose(struct tcptw *tw, int reuse) { struct inpcb *inp; inp = tw->tw_inpcb; tw->tw_inpcb = NULL; tcp_timer_2msl_stop(tw); inp->inp_ppcb = NULL; #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) in6_pcbdetach(inp); else #endif in_pcbdetach(inp); tcpstat.tcps_closed++; crfree(tw->tw_cred); tw->tw_cred = NULL; if (reuse) return (tw); uma_zfree(tcptw_zone, tw); return (NULL); } int tcp_twrespond(struct tcptw *tw, int flags) { struct inpcb *inp = tw->tw_inpcb; struct tcphdr *th; struct mbuf *m; struct ip *ip = NULL; u_int8_t *optp; u_int hdrlen, optlen; int error; #ifdef INET6 struct ip6_hdr *ip6 = NULL; int isipv6 = inp->inp_inc.inc_isipv6; #endif m = m_gethdr(M_DONTWAIT, MT_HEADER); if (m == NULL) return (ENOBUFS); m->m_data += max_linkhdr; #ifdef MAC mac_create_mbuf_from_inpcb(inp, m); #endif #ifdef INET6 if (isipv6) { hdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); tcpip_fillheaders(inp, ip6, th); } else #endif { hdrlen = sizeof(struct tcpiphdr); ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); tcpip_fillheaders(inp, ip, th); } optp = (u_int8_t *)(th + 1); /* * Send a timestamp and echo-reply if both our side and our peer * have sent timestamps in our SYN's and this is not a RST. */ if (tw->t_recent && flags == TH_ACK) { u_int32_t *lp = (u_int32_t *)optp; /* Form timestamp option as shown in appendix A of RFC 1323. */ *lp++ = htonl(TCPOPT_TSTAMP_HDR); *lp++ = htonl(ticks); *lp = htonl(tw->t_recent); optp += TCPOLEN_TSTAMP_APPA; } /* * Send `CC-family' options if needed, and it's not a RST. */ if (tw->cc_recv != 0 && flags == TH_ACK) { u_int32_t *lp = (u_int32_t *)optp; *lp++ = htonl(TCPOPT_CC_HDR(TCPOPT_CC)); *lp = htonl(tw->cc_send); optp += TCPOLEN_CC_APPA; } optlen = optp - (u_int8_t *)(th + 1); m->m_len = hdrlen + optlen; m->m_pkthdr.len = m->m_len; KASSERT(max_linkhdr + m->m_len <= MHLEN, ("tcptw: mbuf too small")); th->th_seq = htonl(tw->snd_nxt); th->th_ack = htonl(tw->rcv_nxt); th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; th->th_flags = flags; th->th_win = htons(tw->last_win); #ifdef INET6 if (isipv6) { th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), sizeof(struct tcphdr) + optlen); ip6->ip6_hlim = in6_selecthlim(inp, NULL); error = ip6_output(m, inp->in6p_outputopts, NULL, (tw->tw_so_options & SO_DONTROUTE), NULL, NULL, inp); } else #endif { th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons(sizeof(struct tcphdr) + optlen + IPPROTO_TCP)); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); ip->ip_len = m->m_pkthdr.len; if (path_mtu_discovery) ip->ip_off |= IP_DF; error = ip_output(m, inp->inp_options, NULL, (tw->tw_so_options & SO_DONTROUTE), NULL, inp); } if (flags & TH_ACK) tcpstat.tcps_sndacks++; else tcpstat.tcps_sndctrl++; tcpstat.tcps_sndtotal++; return (error); } /* * TCP BANDWIDTH DELAY PRODUCT WINDOW LIMITING * * This code attempts to calculate the bandwidth-delay product as a * means of determining the optimal window size to maximize bandwidth, * minimize RTT, and avoid the over-allocation of buffers on interfaces and * routers. This code also does a fairly good job keeping RTTs in check * across slow links like modems. We implement an algorithm which is very * similar (but not meant to be) TCP/Vegas. The code operates on the * transmitter side of a TCP connection and so only effects the transmit * side of the connection. * * BACKGROUND: TCP makes no provision for the management of buffer space * at the end points or at the intermediate routers and switches. A TCP * stream, whether using NewReno or not, will eventually buffer as * many packets as it is able and the only reason this typically works is * due to the fairly small default buffers made available for a connection * (typicaly 16K or 32K). As machines use larger windows and/or window * scaling it is now fairly easy for even a single TCP connection to blow-out * all available buffer space not only on the local interface, but on * intermediate routers and switches as well. NewReno makes a misguided * attempt to 'solve' this problem by waiting for an actual failure to occur, * then backing off, then steadily increasing the window again until another * failure occurs, ad-infinitum. This results in terrible oscillation that * is only made worse as network loads increase and the idea of intentionally * blowing out network buffers is, frankly, a terrible way to manage network * resources. * * It is far better to limit the transmit window prior to the failure * condition being achieved. There are two general ways to do this: First * you can 'scan' through different transmit window sizes and locate the * point where the RTT stops increasing, indicating that you have filled the * pipe, then scan backwards until you note that RTT stops decreasing, then * repeat ad-infinitum. This method works in principle but has severe * implementation issues due to RTT variances, timer granularity, and * instability in the algorithm which can lead to many false positives and * create oscillations as well as interact badly with other TCP streams * implementing the same algorithm. * * The second method is to limit the window to the bandwidth delay product * of the link. This is the method we implement. RTT variances and our * own manipulation of the congestion window, bwnd, can potentially * destabilize the algorithm. For this reason we have to stabilize the * elements used to calculate the window. We do this by using the minimum * observed RTT, the long term average of the observed bandwidth, and * by adding two segments worth of slop. It isn't perfect but it is able * to react to changing conditions and gives us a very stable basis on * which to extend the algorithm. */ void tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq) { u_long bw; u_long bwnd; int save_ticks; /* * If inflight_enable is disabled in the middle of a tcp connection, * make sure snd_bwnd is effectively disabled. */ if (tcp_inflight_enable == 0) { tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_bandwidth = 0; return; } /* * Figure out the bandwidth. Due to the tick granularity this * is a very rough number and it MUST be averaged over a fairly * long period of time. XXX we need to take into account a link * that is not using all available bandwidth, but for now our * slop will ramp us up if this case occurs and the bandwidth later * increases. * * Note: if ticks rollover 'bw' may wind up negative. We must * effectively reset t_bw_rtttime for this case. */ save_ticks = ticks; if ((u_int)(save_ticks - tp->t_bw_rtttime) < 1) return; bw = (int64_t)(ack_seq - tp->t_bw_rtseq) * hz / (save_ticks - tp->t_bw_rtttime); tp->t_bw_rtttime = save_ticks; tp->t_bw_rtseq = ack_seq; if (tp->t_bw_rtttime == 0 || (int)bw < 0) return; bw = ((int64_t)tp->snd_bandwidth * 15 + bw) >> 4; tp->snd_bandwidth = bw; /* * Calculate the semi-static bandwidth delay product, plus two maximal * segments. The additional slop puts us squarely in the sweet * spot and also handles the bandwidth run-up case and stabilization. * Without the slop we could be locking ourselves into a lower * bandwidth. * * Situations Handled: * (1) Prevents over-queueing of packets on LANs, especially on * high speed LANs, allowing larger TCP buffers to be * specified, and also does a good job preventing * over-queueing of packets over choke points like modems * (at least for the transmit side). * * (2) Is able to handle changing network loads (bandwidth * drops so bwnd drops, bandwidth increases so bwnd * increases). * * (3) Theoretically should stabilize in the face of multiple * connections implementing the same algorithm (this may need * a little work). * * (4) Stability value (defaults to 20 = 2 maximal packets) can * be adjusted with a sysctl but typically only needs to be * on very slow connections. A value no smaller then 5 * should be used, but only reduce this default if you have * no other choice. */ #define USERTT ((tp->t_srtt + tp->t_rttbest) / 2) bwnd = (int64_t)bw * USERTT / (hz << TCP_RTT_SHIFT) + tcp_inflight_stab * tp->t_maxseg / 10; #undef USERTT if (tcp_inflight_debug > 0) { static int ltime; if ((u_int)(ticks - ltime) >= hz / tcp_inflight_debug) { ltime = ticks; printf("%p bw %ld rttbest %d srtt %d bwnd %ld\n", tp, bw, tp->t_rttbest, tp->t_srtt, bwnd ); } } if ((long)bwnd < tcp_inflight_min) bwnd = tcp_inflight_min; if (bwnd > tcp_inflight_max) bwnd = tcp_inflight_max; if ((long)bwnd < tp->t_maxseg * 2) bwnd = tp->t_maxseg * 2; tp->snd_bwnd = bwnd; } #ifdef TCP_SIGNATURE /* * Callback function invoked by m_apply() to digest TCP segment data * contained within an mbuf chain. */ static int tcp_signature_apply(void *fstate, void *data, u_int len) { MD5Update(fstate, (u_char *)data, len); return (0); } /* * Compute TCP-MD5 hash of a TCPv4 segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * off0 offset to TCP header within the mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND) * * We do this over ip, tcphdr, segment data, and the key in the SADB. * When called from tcp_input(), we can be sure that th_sum has been * zeroed out and verified already. * * This function is for IPv4 use only. Calling this function with an * IPv6 packet in the mbuf chain will yield undefined results. * * Return 0 if successful, otherwise return -1. * * XXX The key is retrieved from the system's PF_KEY SADB, by keying a * search with the destination IP address, and a 'magic SPI' to be * determined by the application. This is hardcoded elsewhere to 1179 * right now. Another branch of this code exists which uses the SPD to * specify per-application flows but it is unstable. */ int tcp_signature_compute(struct mbuf *m, int off0, int len, int optlen, u_char *buf, u_int direction) { union sockaddr_union dst; struct ippseudo ippseudo; MD5_CTX ctx; int doff; struct ip *ip; struct ipovly *ipovly; struct secasvar *sav; struct tcphdr *th; u_short savecsum; KASSERT(m != NULL, ("NULL mbuf chain")); KASSERT(buf != NULL, ("NULL signature pointer")); /* Extract the destination from the IP header in the mbuf. */ ip = mtod(m, struct ip *); bzero(&dst, sizeof(union sockaddr_union)); dst.sa.sa_len = sizeof(struct sockaddr_in); dst.sa.sa_family = AF_INET; dst.sin.sin_addr = (direction == IPSEC_DIR_INBOUND) ? ip->ip_src : ip->ip_dst; /* Look up an SADB entry which matches the address of the peer. */ sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); if (sav == NULL) { printf("%s: SADB lookup failed for %s\n", __func__, inet_ntoa(dst.sin.sin_addr)); return (EINVAL); } MD5Init(&ctx); ipovly = (struct ipovly *)ip; th = (struct tcphdr *)((u_char *)ip + off0); doff = off0 + sizeof(struct tcphdr) + optlen; /* * Step 1: Update MD5 hash with IP pseudo-header. * * XXX The ippseudo header MUST be digested in network byte order, * or else we'll fail the regression test. Assume all fields we've * been doing arithmetic on have been in host byte order. * XXX One cannot depend on ipovly->ih_len here. When called from * tcp_output(), the underlying ip_len member has not yet been set. */ ippseudo.ippseudo_src = ipovly->ih_src; ippseudo.ippseudo_dst = ipovly->ih_dst; ippseudo.ippseudo_pad = 0; ippseudo.ippseudo_p = IPPROTO_TCP; ippseudo.ippseudo_len = htons(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&ippseudo, sizeof(struct ippseudo)); /* * Step 2: Update MD5 hash with TCP header, excluding options. * The TCP checksum must be set to zero. */ savecsum = th->th_sum; th->th_sum = 0; MD5Update(&ctx, (char *)th, sizeof(struct tcphdr)); th->th_sum = savecsum; /* * Step 3: Update MD5 hash with TCP segment data. * Use m_apply() to avoid an early m_pullup(). */ if (len > 0) m_apply(m, doff, len, tcp_signature_apply, &ctx); /* * Step 4: Update MD5 hash with shared secret. */ MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth)); MD5Final(buf, &ctx); key_sa_recordxfer(sav, m); KEY_FREESAV(&sav); return (0); } #endif /* TCP_SIGNATURE */ diff --git a/sys/netinet/udp_usrreq.c b/sys/netinet/udp_usrreq.c index 868049134d48..9c541dba8524 100644 --- a/sys/netinet/udp_usrreq.c +++ b/sys/netinet/udp_usrreq.c @@ -1,1125 +1,1125 @@ /* * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 * $FreeBSD$ */ #include "opt_ipsec.h" #include "opt_inet6.h" #include "opt_mac.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 #ifdef INET6 #include #endif #include #include #include #ifdef INET6 #include #endif #include #include #ifdef FAST_IPSEC #include #endif /*FAST_IPSEC*/ #ifdef IPSEC #include #endif /*IPSEC*/ #include /* * UDP protocol implementation. * Per RFC 768, August, 1980. */ #ifndef COMPAT_42 static int udpcksum = 1; #else static int udpcksum = 0; /* XXX */ #endif SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW, &udpcksum, 0, ""); int log_in_vain = 0; SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW, &log_in_vain, 0, "Log all incoming UDP packets"); static int blackhole = 0; SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW, &blackhole, 0, "Do not send port unreachables for refused connects"); static int strict_mcast_mship = 0; SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW, &strict_mcast_mship, 0, "Only send multicast to member sockets"); struct inpcbhead udb; /* from udp_var.h */ #define udb6 udb /* for KAME src sync over BSD*'s */ struct inpcbinfo udbinfo; #ifndef UDBHASHSIZE #define UDBHASHSIZE 16 #endif struct udpstat udpstat; /* from udp_var.h */ SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW, &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)"); static struct sockaddr_in udp_in = { sizeof(udp_in), AF_INET }; #ifdef INET6 struct udp_in6 { struct sockaddr_in6 uin6_sin; u_char uin6_init_done : 1; } udp_in6 = { { sizeof(udp_in6.uin6_sin), AF_INET6 }, 0 }; struct udp_ip6 { struct ip6_hdr uip6_ip6; u_char uip6_init_done : 1; } udp_ip6; #endif /* INET6 */ static void udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off); #ifdef INET6 static void ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip); #endif static int udp_detach(struct socket *so); static int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *, struct mbuf *, struct thread *); void udp_init() { INP_INFO_LOCK_INIT(&udbinfo, "udp"); LIST_INIT(&udb); udbinfo.listhead = &udb; udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask); udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.porthashmask); udbinfo.ipi_zone = uma_zcreate("udpcb", sizeof(struct inpcb), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(udbinfo.ipi_zone, maxsockets); } void udp_input(m, off) register struct mbuf *m; int off; { int iphlen = off; register struct ip *ip; register struct udphdr *uh; register struct inpcb *inp; struct mbuf *opts = 0; int len; struct ip save_ip; udpstat.udps_ipackets++; /* * Strip IP options, if any; should skip this, * make available to user, and use on returned packets, * but we don't yet have a way to check the checksum * with options still present. */ if (iphlen > sizeof (struct ip)) { ip_stripoptions(m, (struct mbuf *)0); iphlen = sizeof(struct ip); } /* * Get IP and UDP header together in first mbuf. */ ip = mtod(m, struct ip *); if (m->m_len < iphlen + sizeof(struct udphdr)) { if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) { udpstat.udps_hdrops++; return; } ip = mtod(m, struct ip *); } uh = (struct udphdr *)((caddr_t)ip + iphlen); /* destination port of 0 is illegal, based on RFC768. */ if (uh->uh_dport == 0) goto badunlocked; /* * Construct sockaddr format source address. * Stuff source address and datagram in user buffer. */ udp_in.sin_port = uh->uh_sport; udp_in.sin_addr = ip->ip_src; #ifdef INET6 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0; #endif /* * Make mbuf data length reflect UDP length. * If not enough data to reflect UDP length, drop. */ len = ntohs((u_short)uh->uh_ulen); if (ip->ip_len != len) { if (len > ip->ip_len || len < sizeof(struct udphdr)) { udpstat.udps_badlen++; goto badunlocked; } m_adj(m, len - ip->ip_len); /* ip->ip_len = len; */ } /* * Save a copy of the IP header in case we want restore it * for sending an ICMP error message in response. */ if (!blackhole) save_ip = *ip; /* * Checksum extended UDP header and data. */ if (uh->uh_sum) { if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) uh->uh_sum = m->m_pkthdr.csum_data; else uh->uh_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl((u_short)len + m->m_pkthdr.csum_data + IPPROTO_UDP)); uh->uh_sum ^= 0xffff; } else { char b[9]; bcopy(((struct ipovly *)ip)->ih_x1, b, 9); bzero(((struct ipovly *)ip)->ih_x1, 9); ((struct ipovly *)ip)->ih_len = uh->uh_ulen; uh->uh_sum = in_cksum(m, len + sizeof (struct ip)); bcopy(b, ((struct ipovly *)ip)->ih_x1, 9); } if (uh->uh_sum) { udpstat.udps_badsum++; m_freem(m); return; } } else udpstat.udps_nosum++; INP_INFO_RLOCK(&udbinfo); if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) { struct inpcb *last; /* * Deliver a multicast or broadcast datagram to *all* sockets * for which the local and remote addresses and ports match * those of the incoming datagram. This allows more than * one process to receive multi/broadcasts on the same port. * (This really ought to be done for unicast datagrams as * well, but that would cause problems with existing * applications that open both address-specific sockets and * a wildcard socket listening to the same port -- they would * end up receiving duplicates of every unicast datagram. * Those applications open the multiple sockets to overcome an * inadequacy of the UDP socket interface, but for backwards * compatibility we avoid the problem here rather than * fixing the interface. Maybe 4.5BSD will remedy this?) */ /* * Locate pcb(s) for datagram. * (Algorithm copied from raw_intr().) */ last = NULL; LIST_FOREACH(inp, &udb, inp_list) { INP_LOCK(inp); if (inp->inp_lport != uh->uh_dport) { docontinue: INP_UNLOCK(inp); continue; } #ifdef INET6 if ((inp->inp_vflag & INP_IPV4) == 0) goto docontinue; #endif if (inp->inp_laddr.s_addr != INADDR_ANY) { if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) goto docontinue; } if (inp->inp_faddr.s_addr != INADDR_ANY) { if (inp->inp_faddr.s_addr != ip->ip_src.s_addr || inp->inp_fport != uh->uh_sport) goto docontinue; } /* * Check multicast packets to make sure they are only * sent to sockets with multicast memberships for the * packet's destination address and arrival interface */ #define MSHIP(_inp, n) ((_inp)->inp_moptions->imo_membership[(n)]) #define NMSHIPS(_inp) ((_inp)->inp_moptions->imo_num_memberships) if (strict_mcast_mship && inp->inp_moptions != NULL) { int mship, foundmship = 0; for (mship = 0; mship < NMSHIPS(inp); mship++) { if (MSHIP(inp, mship)->inm_addr.s_addr == ip->ip_dst.s_addr && MSHIP(inp, mship)->inm_ifp == m->m_pkthdr.rcvif) { foundmship = 1; break; } } if (foundmship == 0) goto docontinue; } #undef NMSHIPS #undef MSHIP if (last != NULL) { struct mbuf *n; n = m_copy(m, 0, M_COPYALL); if (n != NULL) udp_append(last, ip, n, iphlen + sizeof(struct udphdr)); INP_UNLOCK(last); } last = inp; /* * Don't look for additional matches if this one does * not have either the SO_REUSEPORT or SO_REUSEADDR * socket options set. This heuristic avoids searching * through all pcbs in the common case of a non-shared * port. It * assumes that an application will never * clear these options after setting them. */ if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0) break; } if (last == NULL) { /* * No matching pcb found; discard datagram. * (No need to send an ICMP Port Unreachable * for a broadcast or multicast datgram.) */ udpstat.udps_noportbcast++; goto badheadlocked; } INP_INFO_RUNLOCK(&udbinfo); udp_append(last, ip, m, iphlen + sizeof(struct udphdr)); INP_UNLOCK(last); return; } /* * Locate pcb for datagram. */ inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport, ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif); if (inp == NULL) { if (log_in_vain) { char buf[4*sizeof "123"]; strcpy(buf, inet_ntoa(ip->ip_dst)); log(LOG_INFO, "Connection attempt to UDP %s:%d from %s:%d\n", buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src), ntohs(uh->uh_sport)); } udpstat.udps_noport++; if (m->m_flags & (M_BCAST | M_MCAST)) { udpstat.udps_noportbcast++; goto badheadlocked; } if (blackhole) goto badheadlocked; if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0) goto badheadlocked; *ip = save_ip; ip->ip_len += iphlen; icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); INP_INFO_RUNLOCK(&udbinfo); return; } INP_LOCK(inp); INP_INFO_RUNLOCK(&udbinfo); udp_append(inp, ip, m, iphlen + sizeof(struct udphdr)); INP_UNLOCK(inp); return; badheadlocked: INP_INFO_RUNLOCK(&udbinfo); if (inp) INP_UNLOCK(inp); badunlocked: m_freem(m); if (opts) m_freem(opts); return; } #ifdef INET6 static void ip_2_ip6_hdr(ip6, ip) struct ip6_hdr *ip6; struct ip *ip; { bzero(ip6, sizeof(*ip6)); ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_plen = ip->ip_len; ip6->ip6_nxt = ip->ip_p; ip6->ip6_hlim = ip->ip_ttl; ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_SMP; ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr; ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr; } #endif /* * subroutine of udp_input(), mainly for source code readability. * caller must properly init udp_ip6 and udp_in6 beforehand. */ static void udp_append(last, ip, n, off) struct inpcb *last; struct ip *ip; struct mbuf *n; int off; { struct sockaddr *append_sa; struct mbuf *opts = 0; INP_LOCK_ASSERT(last); #if defined(IPSEC) || defined(FAST_IPSEC) /* check AH/ESP integrity. */ if (ipsec4_in_reject(n, last)) { #ifdef IPSEC ipsecstat.in_polvio++; #endif /*IPSEC*/ m_freem(n); return; } #endif /*IPSEC || FAST_IPSEC*/ #ifdef MAC if (mac_check_inpcb_deliver(last, n) != 0) { m_freem(n); return; } #endif if (last->inp_flags & INP_CONTROLOPTS || last->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) { #ifdef INET6 if (last->inp_vflag & INP_IPV6) { int savedflags; if (udp_ip6.uip6_init_done == 0) { ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip); udp_ip6.uip6_init_done = 1; } savedflags = last->inp_flags; last->inp_flags &= ~INP_UNMAPPABLEOPTS; ip6_savecontrol(last, n, &opts); last->inp_flags = savedflags; } else #endif ip_savecontrol(last, &opts, ip, n); } #ifdef INET6 if (last->inp_vflag & INP_IPV6) { if (udp_in6.uin6_init_done == 0) { in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin); udp_in6.uin6_init_done = 1; } append_sa = (struct sockaddr *)&udp_in6.uin6_sin; } else #endif append_sa = (struct sockaddr *)&udp_in; m_adj(n, off); if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) { m_freem(n); if (opts) m_freem(opts); udpstat.udps_fullsock++; } else sorwakeup(last->inp_socket); } /* * Notify a udp user of an asynchronous error; * just wake up so that he can collect error status. */ struct inpcb * udp_notify(inp, errno) register struct inpcb *inp; int errno; { inp->inp_socket->so_error = errno; sorwakeup(inp->inp_socket); sowwakeup(inp->inp_socket); return inp; } void udp_ctlinput(cmd, sa, vip) int cmd; struct sockaddr *sa; void *vip; { struct ip *ip = vip; struct udphdr *uh; struct inpcb *(*notify)(struct inpcb *, int) = udp_notify; struct in_addr faddr; struct inpcb *inp; int s; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; /* * Redirects don't need to be handled up here. */ if (PRC_IS_REDIRECT(cmd)) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * XXX: We never get this from ICMP, otherwise it makes an * excellent DoS attack on machines with many connections. */ if (cmd == PRC_HOSTDEAD) ip = 0; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip) { s = splnet(); uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); INP_INFO_RLOCK(&udbinfo); inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport, ip->ip_src, uh->uh_sport, 0, NULL); if (inp != NULL) { INP_LOCK(inp); if (inp->inp_socket != NULL) { (*notify)(inp, inetctlerrmap[cmd]); } INP_UNLOCK(inp); } INP_INFO_RUNLOCK(&udbinfo); splx(s); } else in_pcbnotifyall(&udbinfo, faddr, inetctlerrmap[cmd], notify); } static int udp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n, s; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == 0) { n = udbinfo.ipi_count; req->oldidx = 2 * (sizeof xig) + (n + n/8) * sizeof(struct xinpcb); return 0; } if (req->newptr != 0) return EPERM; /* * OK, now we're committed to doing something. */ s = splnet(); INP_INFO_RLOCK(&udbinfo); gencnt = udbinfo.ipi_gencnt; n = udbinfo.ipi_count; INP_INFO_RUNLOCK(&udbinfo); splx(s); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + n * sizeof(struct xinpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return error; inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == 0) return ENOMEM; s = splnet(); INP_INFO_RLOCK(&udbinfo); for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_LOCK(inp); if (inp->inp_gencnt <= gencnt && cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) inp_list[i++] = inp; INP_UNLOCK(inp); } INP_INFO_RUNLOCK(&udbinfo); splx(s); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; if (inp->inp_gencnt <= gencnt) { struct xinpcb xi; xi.xi_len = sizeof xi; /* XXX should avoid extra copy */ bcopy(inp, &xi.xi_inp, sizeof *inp); if (inp->inp_socket) sotoxsocket(inp->inp_socket, &xi.xi_socket); xi.xi_inp.inp_gencnt = inp->inp_gencnt; error = SYSCTL_OUT(req, &xi, sizeof xi); } } if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ s = splnet(); INP_INFO_RLOCK(&udbinfo); xig.xig_gen = udbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = udbinfo.ipi_count; INP_INFO_RUNLOCK(&udbinfo); splx(s); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return error; } SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, udp_pcblist, "S,xinpcb", "List of active UDP sockets"); static int udp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error, s; error = suser_cred(req->td->td_ucred, PRISON_ROOT); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); s = splnet(); INP_INFO_RLOCK(&udbinfo); inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, 1, NULL); if (inp == NULL || inp->inp_socket == NULL) { error = ENOENT; goto out; } error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error) goto out; cru2x(inp->inp_socket->so_cred, &xuc); out: INP_INFO_RUNLOCK(&udbinfo); splx(s); if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, udp_getcred, "S,xucred", "Get the xucred of a UDP connection"); static int udp_output(inp, m, addr, control, td) register struct inpcb *inp; struct mbuf *m; struct sockaddr *addr; struct mbuf *control; struct thread *td; { register struct udpiphdr *ui; register int len = m->m_pkthdr.len; struct in_addr faddr, laddr; struct cmsghdr *cm; struct sockaddr_in *sin, src; int error = 0; int ipflags; u_short fport, lport; INP_LOCK_ASSERT(inp); #ifdef MAC - mac_create_mbuf_from_socket(inp->inp_socket, m); + mac_create_mbuf_from_inpcb(inp, m); #endif if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { error = EMSGSIZE; if (control) m_freem(control); goto release; } src.sin_addr.s_addr = INADDR_ANY; if (control != NULL) { /* * XXX: Currently, we assume all the optional information * is stored in a single mbuf. */ if (control->m_next) { error = EINVAL; m_freem(control); goto release; } for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len), control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { cm = mtod(control, struct cmsghdr *); if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0 || cm->cmsg_len > control->m_len) { error = EINVAL; break; } if (cm->cmsg_level != IPPROTO_IP) continue; switch (cm->cmsg_type) { case IP_SENDSRCADDR: if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_addr))) { error = EINVAL; break; } bzero(&src, sizeof(src)); src.sin_family = AF_INET; src.sin_len = sizeof(src); src.sin_port = inp->inp_lport; src.sin_addr = *(struct in_addr *)CMSG_DATA(cm); break; default: error = ENOPROTOOPT; break; } if (error) break; } m_freem(control); } if (error) goto release; laddr = inp->inp_laddr; lport = inp->inp_lport; if (src.sin_addr.s_addr != INADDR_ANY) { if (lport == 0) { error = EINVAL; goto release; } error = in_pcbbind_setup(inp, (struct sockaddr *)&src, &laddr.s_addr, &lport, td->td_ucred); if (error) goto release; } if (addr) { sin = (struct sockaddr_in *)addr; if (td && jailed(td->td_ucred)) prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr); if (inp->inp_faddr.s_addr != INADDR_ANY) { error = EISCONN; goto release; } error = in_pcbconnect_setup(inp, addr, &laddr.s_addr, &lport, &faddr.s_addr, &fport, NULL, td->td_ucred); if (error) goto release; /* Commit the local port if newly assigned. */ if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { inp->inp_lport = lport; if (in_pcbinshash(inp) != 0) { inp->inp_lport = 0; error = EAGAIN; goto release; } inp->inp_flags |= INP_ANONPORT; } } else { faddr = inp->inp_faddr; fport = inp->inp_fport; if (faddr.s_addr == INADDR_ANY) { error = ENOTCONN; goto release; } } /* * Calculate data length and get a mbuf * for UDP and IP headers. */ M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); if (m == 0) { error = ENOBUFS; goto release; } /* * Fill in mbuf with extended UDP header * and addresses and length put into network format. */ ui = mtod(m, struct udpiphdr *); bzero(ui->ui_x1, sizeof(ui->ui_x1)); /* XXX still needed? */ ui->ui_pr = IPPROTO_UDP; ui->ui_src = laddr; ui->ui_dst = faddr; ui->ui_sport = lport; ui->ui_dport = fport; ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr)); ipflags = inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST); if (inp->inp_flags & INP_ONESBCAST) ipflags |= IP_SENDONES; /* * Set up checksum and output datagram. */ if (udpcksum) { if (inp->inp_flags & INP_ONESBCAST) faddr.s_addr = INADDR_BROADCAST; ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr, htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP)); m->m_pkthdr.csum_flags = CSUM_UDP; m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); } else { ui->ui_sum = 0; } ((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len; ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */ ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */ udpstat.udps_opackets++; error = ip_output(m, inp->inp_options, NULL, ipflags, inp->inp_moptions, inp); return (error); release: m_freem(m); return (error); } u_long udp_sendspace = 9216; /* really max datagram size */ /* 40 1K datagrams */ SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW, &udp_sendspace, 0, "Maximum outgoing UDP datagram size"); u_long udp_recvspace = 40 * (1024 + #ifdef INET6 sizeof(struct sockaddr_in6) #else sizeof(struct sockaddr_in) #endif ); SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW, &udp_recvspace, 0, "Maximum space for incoming UDP datagrams"); static int udp_abort(struct socket *so) { struct inpcb *inp; int s; INP_INFO_WLOCK(&udbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_WUNLOCK(&udbinfo); return EINVAL; /* ??? possible? panic instead? */ } INP_LOCK(inp); soisdisconnected(so); s = splnet(); in_pcbdetach(inp); INP_INFO_WUNLOCK(&udbinfo); splx(s); return 0; } static int udp_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; int s, error; INP_INFO_WLOCK(&udbinfo); inp = sotoinpcb(so); if (inp != 0) { INP_INFO_WUNLOCK(&udbinfo); return EINVAL; } error = soreserve(so, udp_sendspace, udp_recvspace); if (error) { INP_INFO_WUNLOCK(&udbinfo); return error; } s = splnet(); error = in_pcballoc(so, &udbinfo, "udpinp"); splx(s); if (error) { INP_INFO_WUNLOCK(&udbinfo); return error; } inp = (struct inpcb *)so->so_pcb; INP_LOCK(inp); INP_INFO_WUNLOCK(&udbinfo); inp->inp_vflag |= INP_IPV4; inp->inp_ip_ttl = ip_defttl; INP_UNLOCK(inp); return 0; } static int udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; int s, error; INP_INFO_WLOCK(&udbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_WUNLOCK(&udbinfo); return EINVAL; } INP_LOCK(inp); s = splnet(); error = in_pcbbind(inp, nam, td->td_ucred); splx(s); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&udbinfo); return error; } static int udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct inpcb *inp; int s, error; struct sockaddr_in *sin; INP_INFO_WLOCK(&udbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_WUNLOCK(&udbinfo); return EINVAL; } INP_LOCK(inp); if (inp->inp_faddr.s_addr != INADDR_ANY) { INP_UNLOCK(inp); INP_INFO_WUNLOCK(&udbinfo); return EISCONN; } s = splnet(); sin = (struct sockaddr_in *)nam; if (td && jailed(td->td_ucred)) prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr); error = in_pcbconnect(inp, nam, td->td_ucred); splx(s); if (error == 0) soisconnected(so); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&udbinfo); return error; } static int udp_detach(struct socket *so) { struct inpcb *inp; int s; INP_INFO_WLOCK(&udbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_WUNLOCK(&udbinfo); return EINVAL; } INP_LOCK(inp); s = splnet(); in_pcbdetach(inp); INP_INFO_WUNLOCK(&udbinfo); splx(s); return 0; } static int udp_disconnect(struct socket *so) { struct inpcb *inp; int s; INP_INFO_WLOCK(&udbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_WUNLOCK(&udbinfo); return EINVAL; } INP_LOCK(inp); if (inp->inp_faddr.s_addr == INADDR_ANY) { INP_INFO_WUNLOCK(&udbinfo); INP_UNLOCK(inp); return ENOTCONN; } s = splnet(); in_pcbdisconnect(inp); inp->inp_laddr.s_addr = INADDR_ANY; INP_UNLOCK(inp); INP_INFO_WUNLOCK(&udbinfo); splx(s); so->so_state &= ~SS_ISCONNECTED; /* XXX */ return 0; } static int udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, struct mbuf *control, struct thread *td) { struct inpcb *inp; int ret; INP_INFO_WLOCK(&udbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_WUNLOCK(&udbinfo); m_freem(m); return EINVAL; } INP_LOCK(inp); ret = udp_output(inp, m, addr, control, td); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&udbinfo); return ret; } int udp_shutdown(struct socket *so) { struct inpcb *inp; INP_INFO_RLOCK(&udbinfo); inp = sotoinpcb(so); if (inp == 0) { INP_INFO_RUNLOCK(&udbinfo); return EINVAL; } INP_LOCK(inp); INP_INFO_RUNLOCK(&udbinfo); socantsendmore(so); INP_UNLOCK(inp); return 0; } /* * This is the wrapper function for in_setsockaddr. We just pass down * the pcbinfo for in_setsockaddr to lock. We don't want to do the locking * here because in_setsockaddr will call malloc and might block. */ static int udp_sockaddr(struct socket *so, struct sockaddr **nam) { return (in_setsockaddr(so, nam, &udbinfo)); } /* * This is the wrapper function for in_setpeeraddr. We just pass down * the pcbinfo for in_setpeeraddr to lock. */ static int udp_peeraddr(struct socket *so, struct sockaddr **nam) { return (in_setpeeraddr(so, nam, &udbinfo)); } struct pr_usrreqs udp_usrreqs = { udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect, pru_connect2_notsupp, in_control, udp_detach, udp_disconnect, pru_listen_notsupp, udp_peeraddr, pru_rcvd_notsupp, pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown, udp_sockaddr, sosend, soreceive, sopoll, in_pcbsosetlabel };