Index: head/sys/netinet/tcp_syncache.c =================================================================== --- head/sys/netinet/tcp_syncache.c (revision 166402) +++ head/sys/netinet/tcp_syncache.c (revision 166403) @@ -1,1527 +1,1533 @@ /*- * Copyright (c) 2001 McAfee, Inc. * Copyright (c) 2006 Andre Oppermann, Internet Business Solutions AG * All rights reserved. * * This software was developed for the FreeBSD Project by Jonathan Lemon * and McAfee Research, the Security Research Division of McAfee, 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. * * 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 #include #include #include #include #include #include #include #include #include /* for proc0 declaration */ #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #include #include #endif #include #include #include #include #include #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 int tcp_syncookiesonly = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, syncookies_only, CTLFLAG_RW, &tcp_syncookiesonly, 0, "Use only TCP SYN cookies"); #define SYNCOOKIE_SECRET_SIZE 8 /* dwords */ #define SYNCOOKIE_LIFETIME 16 /* seconds */ struct syncache { TAILQ_ENTRY(syncache) sc_hash; struct in_conninfo sc_inc; /* addresses */ u_long sc_rxttime; /* retransmit time */ u_int16_t sc_rxmits; /* retransmit counter */ u_int32_t sc_tsreflect; /* timestamp to reflect */ u_int32_t sc_ts; /* our timestamp to send */ u_int32_t sc_tsoff; /* ts offset w/ syncookies */ u_int32_t sc_flowlabel; /* IPv6 flowlabel */ tcp_seq sc_irs; /* seq from peer */ tcp_seq sc_iss; /* our ISS */ struct mbuf *sc_ipopts; /* source route */ u_int16_t sc_peer_mss; /* peer's MSS */ u_int16_t sc_wnd; /* advertised window */ u_int8_t sc_ip_ttl; /* IPv4 TTL */ u_int8_t sc_ip_tos; /* IPv4 TOS */ u_int8_t sc_requested_s_scale:4, sc_requested_r_scale:4; u_int8_t sc_flags; #define SCF_NOOPT 0x01 /* no TCP options */ #define SCF_WINSCALE 0x02 /* negotiated window scaling */ #define SCF_TIMESTAMP 0x04 /* negotiated timestamps */ /* MSS is implicit */ #define SCF_UNREACH 0x10 /* icmp unreachable received */ #define SCF_SIGNATURE 0x20 /* send MD5 digests */ #define SCF_SACK 0x80 /* send SACK option */ #ifdef MAC struct label *sc_label; /* MAC label reference */ #endif }; struct syncache_head { struct mtx sch_mtx; TAILQ_HEAD(sch_head, syncache) sch_bucket; struct callout sch_timer; int sch_nextc; u_int sch_length; u_int sch_oddeven; u_int32_t sch_secbits_odd[SYNCOOKIE_SECRET_SIZE]; u_int32_t sch_secbits_even[SYNCOOKIE_SECRET_SIZE]; u_int sch_reseed; /* time_uptime, seconds */ }; 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 **); static int syncache_respond(struct syncache *, struct mbuf *); static struct socket *syncache_socket(struct syncache *, struct socket *, struct mbuf *m); static void syncache_timer(void *); static void syncookie_generate(struct syncache_head *, struct syncache *, u_int32_t *); static struct syncache *syncookie_lookup(struct in_conninfo *, struct syncache_head *, struct syncache *, struct tcpopt *, 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; /* XXX: unprotected */ u_int cache_limit; u_int rexmt_limit; u_int hash_secret; }; 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, sch, co) do { \ (sc)->sc_rxmits++; \ (sc)->sc_rxttime = ticks + \ TCPTV_RTOBASE * tcp_backoff[(sc)->sc_rxmits - 1]; \ if ((sch)->sch_nextc > (sc)->sc_rxttime) \ (sch)->sch_nextc = (sc)->sc_rxttime; \ if (!TAILQ_EMPTY(&(sch)->sch_bucket) && !(co)) \ callout_reset(&(sch)->sch_timer, \ (sch)->sch_nextc - ticks, \ syncache_timer, (void *)(sch)); \ } while (0) #define SCH_LOCK(sch) mtx_lock(&(sch)->sch_mtx) #define SCH_UNLOCK(sch) mtx_unlock(&(sch)->sch_mtx) #define SCH_LOCK_ASSERT(sch) mtx_assert(&(sch)->sch_mtx, MA_OWNED) /* * Requires the syncache entry to be already removed from the bucket list. */ static void syncache_free(struct syncache *sc) { if (sc->sc_ipopts) (void) m_free(sc->sc_ipopts); #ifdef MAC mac_destroy_syncache(&sc->sc_label); #endif 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.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.bucketlimit", &tcp_syncache.bucket_limit); if (!powerof2(tcp_syncache.hashsize) || tcp_syncache.hashsize == 0) { printf("WARNING: syncache hash size is not a power of 2.\n"); tcp_syncache.hashsize = TCP_SYNCACHE_HASHSIZE; } tcp_syncache.hashmask = tcp_syncache.hashsize - 1; /* Set limits. */ tcp_syncache.cache_limit = tcp_syncache.hashsize * tcp_syncache.bucket_limit; TUNABLE_INT_FETCH("net.inet.tcp.syncache.cachelimit", &tcp_syncache.cache_limit); /* Allocate the hash table. */ MALLOC(tcp_syncache.hashbase, struct syncache_head *, tcp_syncache.hashsize * sizeof(struct syncache_head), M_SYNCACHE, M_WAITOK | M_ZERO); /* Initialize the hash buckets. */ for (i = 0; i < tcp_syncache.hashsize; i++) { TAILQ_INIT(&tcp_syncache.hashbase[i].sch_bucket); mtx_init(&tcp_syncache.hashbase[i].sch_mtx, "tcp_sc_head", NULL, MTX_DEF); callout_init_mtx(&tcp_syncache.hashbase[i].sch_timer, &tcp_syncache.hashbase[i].sch_mtx, 0); tcp_syncache.hashbase[i].sch_length = 0; } /* Create the syncache entry zone. */ tcp_syncache.zone = uma_zcreate("syncache", sizeof(struct syncache), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); uma_zone_set_max(tcp_syncache.zone, tcp_syncache.cache_limit); } /* * Inserts a syncache entry into the specified bucket row. * Locks and unlocks the syncache_head autonomously. */ static void syncache_insert(struct syncache *sc, struct syncache_head *sch) { struct syncache *sc2; SCH_LOCK(sch); /* * Make sure that we don't overflow the per-bucket limit. * If the bucket is full, toss the oldest element. */ if (sch->sch_length >= tcp_syncache.bucket_limit) { KASSERT(!TAILQ_EMPTY(&sch->sch_bucket), ("sch->sch_length incorrect")); sc2 = TAILQ_LAST(&sch->sch_bucket, sch_head); syncache_drop(sc2, sch); tcpstat.tcps_sc_bucketoverflow++; } /* Put it into the bucket. */ TAILQ_INSERT_HEAD(&sch->sch_bucket, sc, sc_hash); sch->sch_length++; /* Reinitialize the bucket row's timer. */ SYNCACHE_TIMEOUT(sc, sch, 1); SCH_UNLOCK(sch); tcp_syncache.cache_count++; tcpstat.tcps_sc_added++; } /* * Remove and free entry from syncache bucket row. * Expects locked syncache head. */ static void syncache_drop(struct syncache *sc, struct syncache_head *sch) { SCH_LOCK_ASSERT(sch); TAILQ_REMOVE(&sch->sch_bucket, sc, sc_hash); sch->sch_length--; syncache_free(sc); tcp_syncache.cache_count--; } /* * 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. * One separate timer for each bucket row. */ static void syncache_timer(void *xsch) { struct syncache_head *sch = (struct syncache_head *)xsch; struct syncache *sc, *nsc; int tick = ticks; /* NB: syncache_head has already been locked by the callout. */ SCH_LOCK_ASSERT(sch); TAILQ_FOREACH_SAFE(sc, &sch->sch_bucket, sc_hash, nsc) { /* * We do not check if the listen socket still exists * and accept the case where the listen socket may be * gone by the time we resend the SYN/ACK. We do * not expect this to happens often. If it does, * then the RST will be sent by the time the remote * host does the SYN/ACK->ACK. */ if (sc->sc_rxttime >= tick) { if (sc->sc_rxttime < sch->sch_nextc) sch->sch_nextc = sc->sc_rxttime; continue; } if (sc->sc_rxmits > tcp_syncache.rexmt_limit) { syncache_drop(sc, sch); tcpstat.tcps_sc_stale++; continue; } (void) syncache_respond(sc, NULL); tcpstat.tcps_sc_retransmitted++; SYNCACHE_TIMEOUT(sc, sch, 0); } if (!TAILQ_EMPTY(&(sch)->sch_bucket)) callout_reset(&(sch)->sch_timer, (sch)->sch_nextc - tick, syncache_timer, (void *)(sch)); } /* * Find an entry in the syncache. * Returns always with locked syncache_head plus a matching entry or NULL. */ struct syncache * syncache_lookup(struct in_conninfo *inc, struct syncache_head **schp) { struct syncache *sc; struct syncache_head *sch; #ifdef INET6 if (inc->inc_isipv6) { sch = &tcp_syncache.hashbase[ SYNCACHE_HASH6(inc, tcp_syncache.hashmask)]; *schp = sch; SCH_LOCK(sch); /* Circle through bucket row to find matching entry. */ 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; SCH_LOCK(sch); /* Circle through bucket row to find matching entry. */ 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); } } SCH_LOCK_ASSERT(*schp); return (NULL); /* always returns with locked sch */ } /* * 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(struct in_conninfo *inc, struct tcphdr *th) { struct syncache *sc; struct syncache_head *sch; sc = syncache_lookup(inc, &sch); /* returns locked sch */ SCH_LOCK_ASSERT(sch); if (sc == NULL) goto done; /* * 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++; } done: SCH_UNLOCK(sch); } void syncache_badack(struct in_conninfo *inc) { struct syncache *sc; struct syncache_head *sch; sc = syncache_lookup(inc, &sch); /* returns locked sch */ SCH_LOCK_ASSERT(sch); if (sc != NULL) { syncache_drop(sc, sch); tcpstat.tcps_sc_badack++; } SCH_UNLOCK(sch); } void syncache_unreach(struct in_conninfo *inc, struct tcphdr *th) { struct syncache *sc; struct syncache_head *sch; sc = syncache_lookup(inc, &sch); /* returns locked sch */ SCH_LOCK_ASSERT(sch); if (sc == NULL) goto done; /* If the sequence number != sc_iss, then it's a bogus ICMP msg */ if (ntohl(th->th_seq) != sc->sc_iss) goto done; /* * 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_rxmits < 3 + 1) { sc->sc_flags |= SCF_UNREACH; goto done; } syncache_drop(sc, sch); tcpstat.tcps_sc_unreach++; done: SCH_UNLOCK(sch); } /* * Build a new TCP socket structure from a syncache entry. */ static struct socket * syncache_socket(struct syncache *sc, struct socket *lso, struct mbuf *m) { struct inpcb *inp = NULL; struct socket *so; struct tcpcb *tp; NET_ASSERT_GIANT(); 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 SOCK_LOCK(so); mac_set_socket_peer_from_mbuf(m, so); SOCK_UNLOCK(so); #endif inp = sotoinpcb(so); INP_LOCK(inp); /* Insert new socket into PCB 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_socket: 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_socket: 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; } /* Override flowlabel from in6_pcbconnect. */ inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK; inp->in6p_flowinfo |= sc->sc_flowlabel; } else #endif { struct in_addr laddr; struct sockaddr_in sin; inp->inp_options = ip_srcroute(m); 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; else { if (sc->sc_flags & SCF_WINSCALE) { tp->t_flags |= TF_REQ_SCALE|TF_RCVD_SCALE; tp->snd_scale = sc->sc_requested_s_scale; tp->request_r_scale = sc->sc_requested_r_scale; } if (sc->sc_flags & SCF_TIMESTAMP) { tp->t_flags |= TF_REQ_TSTMP|TF_RCVD_TSTMP; tp->ts_recent = sc->sc_tsreflect; tp->ts_recent_age = ticks; tp->ts_offset = sc->sc_tsoff; } #ifdef TCP_SIGNATURE if (sc->sc_flags & SCF_SIGNATURE) tp->t_flags |= TF_SIGNATURE; #endif if (sc->sc_flags & SCF_SACK) { tp->sack_enable = 1; tp->t_flags |= TF_SACK_PERMIT; } } /* * 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_rxmits > 1) 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) 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(struct in_conninfo *inc, struct tcpopt *to, struct tcphdr *th, struct socket **lsop, struct mbuf *m) { struct syncache *sc; struct syncache_head *sch; struct socket *so; struct syncache scs; /* * Global TCP locks are held because we manipulate the PCB lists * and create a new socket. */ INP_INFO_WLOCK_ASSERT(&tcbinfo); sc = syncache_lookup(inc, &sch); /* returns locked sch */ SCH_LOCK_ASSERT(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) { SCH_UNLOCK(sch); goto failed; } bzero(&scs, sizeof(scs)); sc = syncookie_lookup(inc, sch, &scs, to, th, *lsop); SCH_UNLOCK(sch); if (sc == NULL) goto failed; tcpstat.tcps_sc_recvcookie++; } else { /* Pull out the entry to unlock the bucket row. */ TAILQ_REMOVE(&sch->sch_bucket, sc, sc_hash); sch->sch_length--; tcp_syncache.cache_count--; SCH_UNLOCK(sch); } /* * If seg contains an ACK, but not for our SYN/ACK, send a RST. */ if (th->th_ack != sc->sc_iss + 1) goto failed; so = syncache_socket(sc, *lsop, 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++; if (sc != &scs) { syncache_insert(sc, sch); /* try again later */ sc = NULL; } goto failed; } else tcpstat.tcps_sc_completed++; *lsop = so; if (sc != &scs) syncache_free(sc); return (1); failed: if (sc != NULL && sc != &scs) syncache_free(sc); return (0); } /* * 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(struct in_conninfo *inc, struct tcpopt *to, struct tcphdr *th, struct inpcb *inp, struct socket **lsop, struct mbuf *m) { struct tcpcb *tp; struct socket *so; struct syncache *sc = NULL; struct syncache_head *sch; struct mbuf *ipopts = NULL; u_int32_t flowtmp; int win, sb_hiwat, ip_ttl, ip_tos, noopt; #ifdef INET6 int autoflowlabel = 0; #endif #ifdef MAC struct label *maclabel; #endif struct syncache scs; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); /* listen socket */ /* * Combine all so/tp operations very early to drop the INP lock as * soon as possible. */ so = *lsop; tp = sototcpcb(so); #ifdef INET6 if (inc->inc_isipv6 && (inp->in6p_flags & IN6P_AUTOFLOWLABEL)) autoflowlabel = 1; #endif ip_ttl = inp->inp_ip_ttl; ip_tos = inp->inp_ip_tos; win = sbspace(&so->so_rcv); sb_hiwat = so->so_rcv.sb_hiwat; noopt = (tp->t_flags & TF_NOOPT); so = NULL; tp = NULL; #ifdef MAC if (mac_init_syncache(&maclabel) != 0) { *lsop = NULL; INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (1); } else mac_init_syncache_from_inpcb(maclabel, inp); #endif INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); /* * Remember the IP options, if any. */ #ifdef INET6 if (!inc->inc_isipv6) #endif ipopts = ip_srcroute(m); /* * 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); /* returns locked entry */ SCH_LOCK_ASSERT(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_tsreflect = to->to_tsval; #ifdef MAC /* * Since we have already unconditionally allocated label * storage, free it up. The syncache entry will already * have an initialized label we can use. */ mac_destroy_syncache(&maclabel); KASSERT(sc->sc_label != NULL, ("%s: label not initialized", __func__)); #endif if (syncache_respond(sc, m) == 0) { SYNCACHE_TIMEOUT(sc, sch, 1); tcpstat.tcps_sndacks++; tcpstat.tcps_sndtotal++; } SCH_UNLOCK(sch); goto done; } sc = uma_zalloc(tcp_syncache.zone, M_NOWAIT | M_ZERO); 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. */ tcpstat.tcps_sc_zonefail++; sc = TAILQ_LAST(&sch->sch_bucket, sch_head); syncache_drop(sc, sch); sc = uma_zalloc(tcp_syncache.zone, M_NOWAIT | M_ZERO); if (sc == NULL) { if (tcp_syncookies) { bzero(&scs, sizeof(scs)); sc = &scs; } else { SCH_UNLOCK(sch); if (ipopts) (void) m_free(ipopts); goto done; } } } /* * Fill in the syncache values. */ #ifdef MAC sc->sc_label = maclabel; #endif sc->sc_ipopts = ipopts; bcopy(inc, &sc->sc_inc, sizeof(struct in_conninfo)); #ifdef INET6 if (!inc->inc_isipv6) #endif { sc->sc_ip_tos = ip_tos; sc->sc_ip_ttl = ip_ttl; } sc->sc_irs = th->th_seq; sc->sc_iss = arc4random(); sc->sc_flags = 0; sc->sc_flowlabel = 0; /* * Initial receive window: clip sbspace to [0 .. TCP_MAXWIN]. * win was derived from socket earlier in the function. */ 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_tsreflect = to->to_tsval; sc->sc_flags |= SCF_TIMESTAMP; } if (to->to_flags & TOF_SCALE) { int wscale = 0; - /* Compute proper scaling value from buffer space */ + /* + * Compute proper scaling value from buffer space. + * Leave enough room for the socket buffer to grow + * with auto sizing. This allows us to scale the + * receive buffer over a wide range while not losing + * any efficiency or fine granularity. + */ while (wscale < TCP_MAX_WINSHIFT && - (TCP_MAXWIN << wscale) < sb_hiwat) + (0x1 << wscale) < tcp_minmss) wscale++; sc->sc_requested_r_scale = wscale; sc->sc_requested_s_scale = to->to_requested_s_scale; sc->sc_flags |= SCF_WINSCALE; } } #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 if (to->to_flags & TOF_SACK) sc->sc_flags |= SCF_SACK; if (to->to_flags & TOF_MSS) sc->sc_peer_mss = to->to_mss; /* peer mss may be zero */ if (noopt) sc->sc_flags |= SCF_NOOPT; if (tcp_syncookies) { syncookie_generate(sch, sc, &flowtmp); #ifdef INET6 if (autoflowlabel) sc->sc_flowlabel = flowtmp; #endif } else { #ifdef INET6 if (autoflowlabel) sc->sc_flowlabel = (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); #endif } SCH_UNLOCK(sch); /* * Do a standard 3-way handshake. */ if (syncache_respond(sc, m) == 0) { if (tcp_syncookies && tcp_syncookiesonly && sc != &scs) syncache_free(sc); else if (sc != &scs) syncache_insert(sc, sch); /* locks and unlocks sch */ #ifdef MAC else mac_destroy_syncache(&sc->sc_label); #endif tcpstat.tcps_sndacks++; tcpstat.tcps_sndtotal++; } else { if (sc != &scs) syncache_free(sc); #ifdef MAC else mac_destroy_syncache(&sc->sc_label); #endif tcpstat.tcps_sc_dropped++; } done: *lsop = NULL; return (1); } static int syncache_respond(struct syncache *sc, struct mbuf *m) { struct ip *ip = NULL; struct tcphdr *th; int optlen, error; u_int16_t tlen, hlen, mssopt; u_int8_t *optp; #ifdef INET6 struct ip6_hdr *ip6 = NULL; #endif hlen = #ifdef INET6 (sc->sc_inc.inc_isipv6) ? sizeof(struct ip6_hdr) : #endif sizeof(struct ip); /* Determine MSS we advertize to other end of connection. */ mssopt = tcp_mssopt(&sc->sc_inc); if (sc->sc_peer_mss) mssopt = max( min(sc->sc_peer_mss, mssopt), tcp_minmss); /* 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); #ifdef TCP_SIGNATURE if (sc->sc_flags & SCF_SIGNATURE) optlen += TCPOLEN_SIGNATURE; #endif if (sc->sc_flags & SCF_SACK) optlen += TCPOLEN_SACK_PERMITTED; optlen = roundup2(optlen, 4); } 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")); /* Create the IP+TCP header from scratch. */ if (m) m_freem(m); m = m_gethdr(M_DONTWAIT, MT_DATA); if (m == NULL) return (ENOBUFS); #ifdef MAC mac_create_mbuf_from_syncache(sc->sc_label, m); #endif m->m_data += max_linkhdr; m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; #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->ip6_flow &= ~IPV6_FLOWLABEL_MASK; ip6->ip6_flow |= sc->sc_flowlabel; 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 = sc->sc_ip_ttl; ip->ip_tos = sc->sc_ip_tos; /* * 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_requested_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); if (sc->sc_ts) *lp++ = htonl(sc->sc_ts); else *lp++ = htonl(ticks); *lp = htonl(sc->sc_tsreflect); optp += TCPOLEN_TSTAMP_APPA; } #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); optp += TCPOLEN_SIGNATURE; } #endif /* TCP_SIGNATURE */ if (sc->sc_flags & SCF_SACK) { *optp++ = TCPOPT_SACK_PERMITTED; *optp++ = TCPOLEN_SACK_PERMITTED; } { /* Pad TCP options to a 4 byte boundary */ int padlen = optlen - (optp - (u_int8_t *)(th + 1)); while (padlen-- > 0) *optp++ = TCPOPT_EOL; } } #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, NULL); } 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); error = ip_output(m, sc->sc_ipopts, NULL, 0, NULL, NULL); } return (error); } /* * The purpose of SYN cookies is to avoid keeping track of all SYN's we * receive and to be able to handle SYN floods from bogus source addresses * (where we will never receive any reply). SYN floods try to exhaust all * our memory and available slots in the SYN cache table to cause a denial * of service to legitimate users of the local host. * * The idea of SYN cookies is to encode and include all necessary information * about the connection setup state within the SYN-ACK we send back and thus * to get along without keeping any local state until the ACK to the SYN-ACK * arrives (if ever). Everything we need to know should be available from * the information we encoded in the SYN-ACK. * * More information about the theory behind SYN cookies and its first * discussion and specification can be found at: * http://cr.yp.to/syncookies.html (overview) * http://cr.yp.to/syncookies/archive (gory details) * * This implementation extends the orginal idea and first implementation * of FreeBSD by using not only the initial sequence number field to store * information but also the timestamp field if present. This way we can * keep track of the entire state we need to know to recreate the session in * its original form. Almost all TCP speakers implement RFC1323 timestamps * these days. For those that do not we still have to live with the known * shortcomings of the ISN only SYN cookies. * * Cookie layers: * * Initial sequence number we send: * 31|................................|0 * DDDDDDDDDDDDDDDDDDDDDDDDDMMMRRRP * D = MD5 Digest (first dword) * M = MSS index * R = Rotation of secret * P = Odd or Even secret * * The MD5 Digest is computed with over following parameters: * a) randomly rotated secret * b) struct in_conninfo containing the remote/local ip/port (IPv4&IPv6) * c) the received initial sequence number from remote host * d) the rotation offset and odd/even bit * * Timestamp we send: * 31|................................|0 * DDDDDDDDDDDDDDDDDDDDDDSSSSRRRRA5 * D = MD5 Digest (third dword) (only as filler) * S = Requested send window scale * R = Requested receive window scale * A = SACK allowed * 5 = TCP-MD5 enabled (not implemented yet) * XORed with MD5 Digest (forth dword) * * The timestamp isn't cryptographically secure and doesn't need to be. * The double use of the MD5 digest dwords ties it to a specific remote/ * local host/port, remote initial sequence number and our local time * limited secret. A received timestamp is reverted (XORed) and then * the contained MD5 dword is compared to the computed one to ensure the * timestamp belongs to the SYN-ACK we sent. The other parameters may * have been tampered with but this isn't different from supplying bogus * values in the SYN in the first place. * * Some problems with SYN cookies remain however: * 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. * * Notes: * A heuristic to determine when to accept syn cookies is not necessary. * An ACK flood would cause the syncookie verification to be attempted, * but a SYN flood causes syncookies to be generated. Both are of equal * cost, so there's no point in trying to optimize the ACK flood case. * Also, if you don't process certain ACKs for some reason, then all someone * would have to do is launch a SYN and ACK flood at the same time, which * would stop cookie verification and defeat the entire purpose of syncookies. */ static int tcp_sc_msstab[] = { 0, 256, 468, 536, 996, 1452, 1460, 8960 }; static void syncookie_generate(struct syncache_head *sch, struct syncache *sc, u_int32_t *flowlabel) { MD5_CTX ctx; u_int32_t md5_buffer[MD5_DIGEST_LENGTH / sizeof(u_int32_t)]; u_int32_t data; u_int32_t *secbits; u_int off, pmss, mss; int i; SCH_LOCK_ASSERT(sch); /* Which of the two secrets to use. */ secbits = sch->sch_oddeven ? sch->sch_secbits_odd : sch->sch_secbits_even; /* Reseed secret if too old. */ if (sch->sch_reseed < time_uptime) { sch->sch_oddeven = sch->sch_oddeven ? 0 : 1; /* toggle */ secbits = sch->sch_oddeven ? sch->sch_secbits_odd : sch->sch_secbits_even; for (i = 0; i < SYNCOOKIE_SECRET_SIZE; i++) secbits[i] = arc4random(); sch->sch_reseed = time_uptime + SYNCOOKIE_LIFETIME; } /* Secret rotation offset. */ off = sc->sc_iss & 0x7; /* iss was randomized before */ /* Maximum segment size calculation. */ pmss = max( min(sc->sc_peer_mss, tcp_mssopt(&sc->sc_inc)), tcp_minmss); for (mss = sizeof(tcp_sc_msstab) / sizeof(int) - 1; mss > 0; mss--) if (tcp_sc_msstab[mss] <= pmss) break; /* Fold parameters and MD5 digest into the ISN we will send. */ data = sch->sch_oddeven;/* odd or even secret, 1 bit */ data |= off << 1; /* secret offset, derived from iss, 3 bits */ data |= mss << 4; /* mss, 3 bits */ MD5Init(&ctx); MD5Update(&ctx, ((u_int8_t *)secbits) + off, SYNCOOKIE_SECRET_SIZE * sizeof(*secbits) - off); MD5Update(&ctx, secbits, off); MD5Update(&ctx, &sc->sc_inc, sizeof(sc->sc_inc)); MD5Update(&ctx, &sc->sc_irs, sizeof(sc->sc_irs)); MD5Update(&ctx, &data, sizeof(data)); MD5Final((u_int8_t *)&md5_buffer, &ctx); data |= (md5_buffer[0] << 7); sc->sc_iss = data; #ifdef INET6 *flowlabel = md5_buffer[1] & IPV6_FLOWLABEL_MASK; #endif /* Additional parameters are stored in the timestamp if present. */ if (sc->sc_flags & SCF_TIMESTAMP) { data = ((sc->sc_flags & SCF_SIGNATURE) ? 1 : 0); /* TCP-MD5, 1 bit */ data |= ((sc->sc_flags & SCF_SACK) ? 1 : 0) << 1; /* SACK, 1 bit */ data |= sc->sc_requested_s_scale << 2; /* SWIN scale, 4 bits */ data |= sc->sc_requested_r_scale << 6; /* RWIN scale, 4 bits */ data |= md5_buffer[2] << 10; /* more digest bits */ data ^= md5_buffer[3]; sc->sc_ts = data; sc->sc_tsoff = data - ticks; /* after XOR */ } else sc->sc_ts = 0; return; } static struct syncache * syncookie_lookup(struct in_conninfo *inc, struct syncache_head *sch, struct syncache *sc, struct tcpopt *to, struct tcphdr *th, struct socket *so) { MD5_CTX ctx; u_int32_t md5_buffer[MD5_DIGEST_LENGTH / sizeof(u_int32_t)]; u_int32_t data = 0; u_int32_t *secbits; tcp_seq ack, seq; int off, mss, wnd, flags; SCH_LOCK_ASSERT(sch); /* * Pull information out of SYN-ACK/ACK and * revert sequence number advances. */ ack = th->th_ack - 1; seq = th->th_seq - 1; off = (ack >> 1) & 0x7; mss = (ack >> 4) & 0x7; flags = ack & 0x7f; /* Which of the two secrets to use. */ secbits = (flags & 0x1) ? sch->sch_secbits_odd : sch->sch_secbits_even; /* * The secret wasn't updated for the lifetime of a syncookie, * so this SYN-ACK/ACK is either too old (replay) or totally bogus. */ if (sch->sch_reseed < time_uptime) { return (NULL); } /* Recompute the digest so we can compare it. */ MD5Init(&ctx); MD5Update(&ctx, ((u_int8_t *)secbits) + off, SYNCOOKIE_SECRET_SIZE * sizeof(*secbits) - off); MD5Update(&ctx, secbits, off); MD5Update(&ctx, inc, sizeof(*inc)); MD5Update(&ctx, &seq, sizeof(seq)); MD5Update(&ctx, &flags, sizeof(flags)); MD5Final((u_int8_t *)&md5_buffer, &ctx); /* Does the digest part of or ACK'ed ISS match? */ if ((ack & (~0x7f)) != (md5_buffer[0] << 7)) return (NULL); /* Does the digest part of our reflected timestamp match? */ if (to->to_flags & TOF_TS) { data = md5_buffer[3] ^ to->to_tsecr; if ((data & (~0x3ff)) != (md5_buffer[2] << 10)) return (NULL); } /* Fill in the syncache values. */ bcopy(inc, &sc->sc_inc, sizeof(struct in_conninfo)); sc->sc_ipopts = NULL; sc->sc_irs = seq; sc->sc_iss = ack; #ifdef INET6 if (inc->inc_isipv6) { if (sotoinpcb(so)->in6p_flags & IN6P_AUTOFLOWLABEL) sc->sc_flowlabel = md5_buffer[1] & IPV6_FLOWLABEL_MASK; } else #endif { sc->sc_ip_ttl = sotoinpcb(so)->inp_ip_ttl; sc->sc_ip_tos = sotoinpcb(so)->inp_ip_tos; } /* Additional parameters that were encoded in the timestamp. */ if (data) { sc->sc_flags |= SCF_TIMESTAMP; sc->sc_tsreflect = to->to_tsval; sc->sc_tsoff = to->to_tsecr - ticks; sc->sc_flags |= (data & 0x1) ? SCF_SIGNATURE : 0; sc->sc_flags |= ((data >> 1) & 0x1) ? SCF_SACK : 0; sc->sc_requested_s_scale = min((data >> 2) & 0xf, TCP_MAX_WINSHIFT); sc->sc_requested_r_scale = min((data >> 6) & 0xf, TCP_MAX_WINSHIFT); if (sc->sc_requested_s_scale || sc->sc_requested_r_scale) sc->sc_flags |= SCF_WINSCALE; } else sc->sc_flags |= SCF_NOOPT; wnd = sbspace(&so->so_rcv); wnd = imax(wnd, 0); wnd = imin(wnd, TCP_MAXWIN); sc->sc_wnd = wnd; sc->sc_rxmits = 0; sc->sc_peer_mss = tcp_sc_msstab[mss]; return (sc); } Index: head/sys/netinet/tcp_usrreq.c =================================================================== --- head/sys/netinet/tcp_usrreq.c (revision 166402) +++ head/sys/netinet/tcp_usrreq.c (revision 166403) @@ -1,1572 +1,1577 @@ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. * Copyright (c) 2006 Robert N. M. Watson * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 * $FreeBSD$ */ #include "opt_inet.h" #include "opt_inet6.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #ifdef INET6 #include #endif /* INET6 */ #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 #include #ifdef TCPDEBUG #include #endif /* * TCP protocol interface to socket abstraction. */ extern char *tcpstates[]; /* XXX ??? */ static int tcp_attach(struct socket *); static int tcp_connect(struct tcpcb *, struct sockaddr *, struct thread *td); #ifdef INET6 static int tcp6_connect(struct tcpcb *, struct sockaddr *, struct thread *td); #endif /* INET6 */ static void tcp_disconnect(struct tcpcb *); static void tcp_usrclosed(struct tcpcb *); static void tcp_fill_info(struct tcpcb *, struct tcp_info *); #ifdef TCPDEBUG #define TCPDEBUG0 int ostate = 0 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ tcp_trace(TA_USER, ostate, tp, 0, 0, req) #else #define TCPDEBUG0 #define TCPDEBUG1() #define TCPDEBUG2(req) #endif /* * TCP attaches to socket via pru_attach(), reserving space, * and an internet control block. */ static int tcp_usr_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; struct tcpcb *tp = NULL; int error; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL")); TCPDEBUG1(); error = tcp_attach(so); if (error) goto out; if ((so->so_options & SO_LINGER) && so->so_linger == 0) so->so_linger = TCP_LINGERTIME; inp = sotoinpcb(so); tp = intotcpcb(inp); out: TCPDEBUG2(PRU_ATTACH); return error; } /* * tcp_detach is called when the socket layer loses its final reference * to the socket, be it a file descriptor reference, a reference from TCP, * etc. At this point, there is only one case in which we will keep around * inpcb state: time wait. * * This function can probably be re-absorbed back into tcp_usr_detach() now * that there is a single detach path. */ static void tcp_detach(struct socket *so, struct inpcb *inp) { struct tcpcb *tp; #ifdef INET6 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0; #endif INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp")); KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so")); tp = intotcpcb(inp); if (inp->inp_vflag & INP_TIMEWAIT) { /* * There are two cases to handle: one in which the time wait * state is being discarded (INP_DROPPED), and one in which * this connection will remain in timewait. In the former, * it is time to discard all state (except tcptw, which has * already been discarded by the timewait close code, which * should be further up the call stack somewhere). In the * latter case, we detach from the socket, but leave the pcb * present until timewait ends. * * XXXRW: Would it be cleaner to free the tcptw here? */ if (inp->inp_vflag & INP_DROPPED) { KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && " "INP_DROPPED && tp != NULL")); #ifdef INET6 if (isipv6) { in6_pcbdetach(inp); in6_pcbfree(inp); } else { #endif in_pcbdetach(inp); in_pcbfree(inp); #ifdef INET6 } #endif } else { #ifdef INET6 if (isipv6) in6_pcbdetach(inp); else #endif in_pcbdetach(inp); INP_UNLOCK(inp); } } else { /* * If the connection is not in timewait, we consider two * two conditions: one in which no further processing is * necessary (dropped || embryonic), and one in which TCP is * not yet done, but no longer requires the socket, so the * pcb will persist for the time being. * * XXXRW: Does the second case still occur? */ if (inp->inp_vflag & INP_DROPPED || tp->t_state < TCPS_SYN_SENT) { tcp_discardcb(tp); #ifdef INET6 if (isipv6) { in6_pcbdetach(inp); in6_pcbfree(inp); } else { #endif in_pcbdetach(inp); in_pcbfree(inp); #ifdef INET6 } #endif } else { #ifdef INET6 if (isipv6) in6_pcbdetach(inp); else #endif in_pcbdetach(inp); } } } /* * pru_detach() detaches the TCP protocol from the socket. * If the protocol state is non-embryonic, then can't * do this directly: have to initiate a pru_disconnect(), * which may finish later; embryonic TCB's can just * be discarded here. */ static void tcp_usr_detach(struct socket *so) { struct inpcb *inp; struct tcpcb *tp; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL")); INP_INFO_WLOCK(&tcbinfo); INP_LOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_detach: inp_socket == NULL")); TCPDEBUG1(); tcp_detach(so, inp); tp = NULL; TCPDEBUG2(PRU_DETACH); INP_INFO_WUNLOCK(&tcbinfo); } /* * Give the socket an address. */ static int tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in *sinp; sinp = (struct sockaddr_in *)nam; if (nam->sa_len != sizeof (*sinp)) return (EINVAL); /* * Must check for multicast addresses and disallow binding * to them. */ if (sinp->sin_family == AF_INET && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) return (EAFNOSUPPORT); TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); error = in_pcbbind(inp, nam, td->td_ucred); out: TCPDEBUG2(PRU_BIND); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #ifdef INET6 static int tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in6 *sin6p; sin6p = (struct sockaddr_in6 *)nam; if (nam->sa_len != sizeof (*sin6p)) return (EINVAL); /* * Must check for multicast addresses and disallow binding * to them. */ if (sin6p->sin6_family == AF_INET6 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) return (EAFNOSUPPORT); TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) inp->inp_vflag |= INP_IPV4; else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { struct sockaddr_in sin; in6_sin6_2_sin(&sin, sin6p); inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; error = in_pcbbind(inp, (struct sockaddr *)&sin, td->td_ucred); goto out; } } error = in6_pcbbind(inp, nam, td->td_ucred); out: TCPDEBUG2(PRU_BIND); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #endif /* INET6 */ /* * Prepare to accept connections. */ static int tcp_usr_listen(struct socket *so, int backlog, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); SOCK_LOCK(so); error = solisten_proto_check(so); if (error == 0 && inp->inp_lport == 0) error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); if (error == 0) { tp->t_state = TCPS_LISTEN; solisten_proto(so, backlog); } SOCK_UNLOCK(so); out: TCPDEBUG2(PRU_LISTEN); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #ifdef INET6 static int tcp6_usr_listen(struct socket *so, int backlog, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); SOCK_LOCK(so); error = solisten_proto_check(so); if (error == 0 && inp->inp_lport == 0) { inp->inp_vflag &= ~INP_IPV4; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) inp->inp_vflag |= INP_IPV4; error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); } if (error == 0) { tp->t_state = TCPS_LISTEN; solisten_proto(so, backlog); } SOCK_UNLOCK(so); out: TCPDEBUG2(PRU_LISTEN); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #endif /* INET6 */ /* * Initiate connection to peer. * Create a template for use in transmissions on this connection. * Enter SYN_SENT state, and mark socket as connecting. * Start keep-alive timer, and seed output sequence space. * Send initial segment on connection. */ static int tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in *sinp; sinp = (struct sockaddr_in *)nam; if (nam->sa_len != sizeof (*sinp)) return (EINVAL); /* * Must disallow TCP ``connections'' to multicast addresses. */ if (sinp->sin_family == AF_INET && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) return (EAFNOSUPPORT); if (jailed(td->td_ucred)) prison_remote_ip(td->td_ucred, 0, &sinp->sin_addr.s_addr); TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); if ((error = tcp_connect(tp, nam, td)) != 0) goto out; error = tcp_output(tp); out: TCPDEBUG2(PRU_CONNECT); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #ifdef INET6 static int tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in6 *sin6p; TCPDEBUG0; sin6p = (struct sockaddr_in6 *)nam; if (nam->sa_len != sizeof (*sin6p)) return (EINVAL); /* * Must disallow TCP ``connections'' to multicast addresses. */ if (sin6p->sin6_family == AF_INET6 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) return (EAFNOSUPPORT); INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { struct sockaddr_in sin; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { error = EINVAL; goto out; } in6_sin6_2_sin(&sin, sin6p); inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) goto out; error = tcp_output(tp); goto out; } inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; inp->inp_inc.inc_isipv6 = 1; if ((error = tcp6_connect(tp, nam, td)) != 0) goto out; error = tcp_output(tp); out: TCPDEBUG2(PRU_CONNECT); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #endif /* INET6 */ /* * Initiate disconnect from peer. * If connection never passed embryonic stage, just drop; * else if don't need to let data drain, then can just drop anyways, * else have to begin TCP shutdown process: mark socket disconnecting, * drain unread data, state switch to reflect user close, and * send segment (e.g. FIN) to peer. Socket will be really disconnected * when peer sends FIN and acks ours. * * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. */ static int tcp_usr_disconnect(struct socket *so) { struct inpcb *inp; struct tcpcb *tp = NULL; int error = 0; TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); tcp_disconnect(tp); out: TCPDEBUG2(PRU_DISCONNECT); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } /* * Accept a connection. Essentially all the work is * done at higher levels; just return the address * of the peer, storing through addr. */ static int tcp_usr_accept(struct socket *so, struct sockaddr **nam) { int error = 0; struct inpcb *inp = NULL; struct tcpcb *tp = NULL; struct in_addr addr; in_port_t port = 0; TCPDEBUG0; if (so->so_state & SS_ISDISCONNECTED) return (ECONNABORTED); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNABORTED; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); /* * We inline in_setpeeraddr and COMMON_END here, so that we can * copy the data of interest and defer the malloc until after we * release the lock. */ port = inp->inp_fport; addr = inp->inp_faddr; out: TCPDEBUG2(PRU_ACCEPT); INP_UNLOCK(inp); if (error == 0) *nam = in_sockaddr(port, &addr); return error; } #ifdef INET6 static int tcp6_usr_accept(struct socket *so, struct sockaddr **nam) { struct inpcb *inp = NULL; int error = 0; struct tcpcb *tp = NULL; struct in_addr addr; struct in6_addr addr6; in_port_t port = 0; int v4 = 0; TCPDEBUG0; if (so->so_state & SS_ISDISCONNECTED) return (ECONNABORTED); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNABORTED; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); /* * We inline in6_mapped_peeraddr and COMMON_END here, so that we can * copy the data of interest and defer the malloc until after we * release the lock. */ if (inp->inp_vflag & INP_IPV4) { v4 = 1; port = inp->inp_fport; addr = inp->inp_faddr; } else { port = inp->inp_fport; addr6 = inp->in6p_faddr; } out: TCPDEBUG2(PRU_ACCEPT); INP_UNLOCK(inp); if (error == 0) { if (v4) *nam = in6_v4mapsin6_sockaddr(port, &addr); else *nam = in6_sockaddr(port, &addr6); } return error; } #endif /* INET6 */ /* * 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 can block. */ static int tcp_sockaddr(struct socket *so, struct sockaddr **nam) { return (in_setsockaddr(so, nam, &tcbinfo)); } /* * This is the wrapper function for in_setpeeraddr. We just pass down * the pcbinfo for in_setpeeraddr to lock. */ static int tcp_peeraddr(struct socket *so, struct sockaddr **nam) { return (in_setpeeraddr(so, nam, &tcbinfo)); } /* * Mark the connection as being incapable of further output. */ static int tcp_usr_shutdown(struct socket *so) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); socantsendmore(so); tcp_usrclosed(tp); error = tcp_output(tp); out: TCPDEBUG2(PRU_SHUTDOWN); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } /* * After a receive, possibly send window update to peer. */ static int tcp_usr_rcvd(struct socket *so, int flags) { struct inpcb *inp; struct tcpcb *tp = NULL; int error = 0; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); tcp_output(tp); out: TCPDEBUG2(PRU_RCVD); INP_UNLOCK(inp); return (error); } /* * Do a send by putting data in output queue and updating urgent * marker if URG set. Possibly send more data. Unlike the other * pru_*() routines, the mbuf chains are our responsibility. We * must either enqueue them or free them. The other pru_* routines * generally are caller-frees. */ static int tcp_usr_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; int headlocked = 0; #ifdef INET6 int isipv6; #endif TCPDEBUG0; /* * We require the pcbinfo lock in two cases: * * (1) An implied connect is taking place, which can result in * binding IPs and ports and hence modification of the pcb hash * chains. * * (2) PRUS_EOF is set, resulting in explicit close on the send. */ if ((nam != NULL) || (flags & PRUS_EOF)) { INP_INFO_WLOCK(&tcbinfo); headlocked = 1; } inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { if (control) m_freem(control); if (m) m_freem(m); error = ECONNRESET; goto out; } #ifdef INET6 isipv6 = nam && nam->sa_family == AF_INET6; #endif /* INET6 */ tp = intotcpcb(inp); TCPDEBUG1(); if (control) { /* TCP doesn't do control messages (rights, creds, etc) */ if (control->m_len) { m_freem(control); if (m) m_freem(m); error = EINVAL; goto out; } m_freem(control); /* empty control, just free it */ } if (!(flags & PRUS_OOB)) { sbappendstream(&so->so_snd, m); if (nam && tp->t_state < TCPS_SYN_SENT) { /* * Do implied connect if not yet connected, * initialize window to default value, and * initialize maxseg/maxopd using peer's cached * MSS. */ INP_INFO_WLOCK_ASSERT(&tcbinfo); #ifdef INET6 if (isipv6) error = tcp6_connect(tp, nam, td); else #endif /* INET6 */ error = tcp_connect(tp, nam, td); if (error) goto out; tp->snd_wnd = TTCP_CLIENT_SND_WND; tcp_mss(tp, -1); } if (flags & PRUS_EOF) { /* * Close the send side of the connection after * the data is sent. */ INP_INFO_WLOCK_ASSERT(&tcbinfo); socantsendmore(so); tcp_usrclosed(tp); } if (headlocked) { INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; } if (tp != NULL) { if (flags & PRUS_MORETOCOME) tp->t_flags |= TF_MORETOCOME; error = tcp_output(tp); if (flags & PRUS_MORETOCOME) tp->t_flags &= ~TF_MORETOCOME; } } else { /* * XXXRW: PRUS_EOF not implemented with PRUS_OOB? */ SOCKBUF_LOCK(&so->so_snd); if (sbspace(&so->so_snd) < -512) { SOCKBUF_UNLOCK(&so->so_snd); m_freem(m); error = ENOBUFS; goto out; } /* * According to RFC961 (Assigned Protocols), * the urgent pointer points to the last octet * of urgent data. We continue, however, * to consider it to indicate the first octet * of data past the urgent section. * Otherwise, snd_up should be one lower. */ sbappendstream_locked(&so->so_snd, m); SOCKBUF_UNLOCK(&so->so_snd); if (nam && tp->t_state < TCPS_SYN_SENT) { /* * Do implied connect if not yet connected, * initialize window to default value, and * initialize maxseg/maxopd using peer's cached * MSS. */ INP_INFO_WLOCK_ASSERT(&tcbinfo); #ifdef INET6 if (isipv6) error = tcp6_connect(tp, nam, td); else #endif /* INET6 */ error = tcp_connect(tp, nam, td); if (error) goto out; tp->snd_wnd = TTCP_CLIENT_SND_WND; tcp_mss(tp, -1); INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; } else if (nam) { INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; } tp->snd_up = tp->snd_una + so->so_snd.sb_cc; tp->t_flags |= TF_FORCEDATA; error = tcp_output(tp); tp->t_flags &= ~TF_FORCEDATA; } out: TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB : ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); INP_UNLOCK(inp); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); return (error); } /* * Abort the TCP. Drop the connection abruptly. */ static void tcp_usr_abort(struct socket *so) { struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL")); INP_INFO_WLOCK(&tcbinfo); INP_LOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_abort: inp_socket == NULL")); /* * If we still have full TCP state, and we're not dropped, drop. */ if (!(inp->inp_vflag & INP_TIMEWAIT) && !(inp->inp_vflag & INP_DROPPED)) { tp = intotcpcb(inp); TCPDEBUG1(); tcp_drop(tp, ECONNABORTED); TCPDEBUG2(PRU_ABORT); } if (!(inp->inp_vflag & INP_DROPPED)) { SOCK_LOCK(so); so->so_state |= SS_PROTOREF; SOCK_UNLOCK(so); inp->inp_vflag |= INP_SOCKREF; } INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); } /* * TCP socket is closed. Start friendly disconnect. */ static void tcp_usr_close(struct socket *so) { struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL")); INP_INFO_WLOCK(&tcbinfo); INP_LOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_close: inp_socket == NULL")); /* * If we still have full TCP state, and we're not dropped, initiate * a disconnect. */ if (!(inp->inp_vflag & INP_TIMEWAIT) && !(inp->inp_vflag & INP_DROPPED)) { tp = intotcpcb(inp); TCPDEBUG1(); tcp_disconnect(tp); TCPDEBUG2(PRU_CLOSE); } if (!(inp->inp_vflag & INP_DROPPED)) { SOCK_LOCK(so); so->so_state |= SS_PROTOREF; SOCK_UNLOCK(so); inp->inp_vflag |= INP_SOCKREF; } INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); } /* * Receive out-of-band data. */ static int tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); if ((so->so_oobmark == 0 && (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) || so->so_options & SO_OOBINLINE || tp->t_oobflags & TCPOOB_HADDATA) { error = EINVAL; goto out; } if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { error = EWOULDBLOCK; goto out; } m->m_len = 1; *mtod(m, caddr_t) = tp->t_iobc; if ((flags & MSG_PEEK) == 0) tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); out: TCPDEBUG2(PRU_RCVOOB); INP_UNLOCK(inp); return (error); } struct pr_usrreqs tcp_usrreqs = { .pru_abort = tcp_usr_abort, .pru_accept = tcp_usr_accept, .pru_attach = tcp_usr_attach, .pru_bind = tcp_usr_bind, .pru_connect = tcp_usr_connect, .pru_control = in_control, .pru_detach = tcp_usr_detach, .pru_disconnect = tcp_usr_disconnect, .pru_listen = tcp_usr_listen, .pru_peeraddr = tcp_peeraddr, .pru_rcvd = tcp_usr_rcvd, .pru_rcvoob = tcp_usr_rcvoob, .pru_send = tcp_usr_send, .pru_shutdown = tcp_usr_shutdown, .pru_sockaddr = tcp_sockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = tcp_usr_close, }; #ifdef INET6 struct pr_usrreqs tcp6_usrreqs = { .pru_abort = tcp_usr_abort, .pru_accept = tcp6_usr_accept, .pru_attach = tcp_usr_attach, .pru_bind = tcp6_usr_bind, .pru_connect = tcp6_usr_connect, .pru_control = in6_control, .pru_detach = tcp_usr_detach, .pru_disconnect = tcp_usr_disconnect, .pru_listen = tcp6_usr_listen, .pru_peeraddr = in6_mapped_peeraddr, .pru_rcvd = tcp_usr_rcvd, .pru_rcvoob = tcp_usr_rcvoob, .pru_send = tcp_usr_send, .pru_shutdown = tcp_usr_shutdown, .pru_sockaddr = in6_mapped_sockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = tcp_usr_close, }; #endif /* INET6 */ /* * Common subroutine to open a TCP connection to remote host specified * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local * port number if needed. Call in_pcbconnect_setup to do the routing and * to choose a local host address (interface). If there is an existing * incarnation of the same connection in TIME-WAIT state and if the remote * host was sending CC options and if the connection duration was < MSL, then * truncate the previous TIME-WAIT state and proceed. * Initialize connection parameters and enter SYN-SENT state. */ static int tcp_connect(tp, nam, td) register struct tcpcb *tp; struct sockaddr *nam; struct thread *td; { struct inpcb *inp = tp->t_inpcb, *oinp; struct socket *so = inp->inp_socket; struct in_addr laddr; u_short lport; int error; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); if (inp->inp_lport == 0) { error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); if (error) return error; } /* * Cannot simply call in_pcbconnect, because there might be an * earlier incarnation of this same connection still in * TIME_WAIT state, creating an ADDRINUSE error. */ laddr = inp->inp_laddr; lport = inp->inp_lport; error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport, &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred); if (error && oinp == NULL) return error; if (oinp) return EADDRINUSE; inp->inp_laddr = laddr; in_pcbrehash(inp); - /* Compute window scaling to request. */ + /* + * Compute window scaling to request: + * Scale to fit into sweet spot. See tcp_syncache.c. + * XXX: This should move to tcp_output(). + * XXX: This should be based on the actual MSS. + */ while (tp->request_r_scale < TCP_MAX_WINSHIFT && - (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) + (0x1 << tp->request_r_scale) < tcp_minmss) tp->request_r_scale++; soisconnecting(so); tcpstat.tcps_connattempt++; tp->t_state = TCPS_SYN_SENT; callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); tp->iss = tcp_new_isn(tp); tp->t_bw_rtseq = tp->iss; tcp_sendseqinit(tp); return 0; } #ifdef INET6 static int tcp6_connect(tp, nam, td) register struct tcpcb *tp; struct sockaddr *nam; struct thread *td; { struct inpcb *inp = tp->t_inpcb, *oinp; struct socket *so = inp->inp_socket; struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; struct in6_addr *addr6; int error; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); if (inp->inp_lport == 0) { error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); if (error) return error; } /* * Cannot simply call in_pcbconnect, because there might be an * earlier incarnation of this same connection still in * TIME_WAIT state, creating an ADDRINUSE error. * in6_pcbladdr() also handles scope zone IDs. */ error = in6_pcbladdr(inp, nam, &addr6); if (error) return error; oinp = in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr, sin6->sin6_port, IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? addr6 : &inp->in6p_laddr, inp->inp_lport, 0, NULL); if (oinp) return EADDRINUSE; if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) inp->in6p_laddr = *addr6; inp->in6p_faddr = sin6->sin6_addr; inp->inp_fport = sin6->sin6_port; /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK; if (inp->in6p_flags & IN6P_AUTOFLOWLABEL) inp->in6p_flowinfo |= (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); in_pcbrehash(inp); /* Compute window scaling to request. */ while (tp->request_r_scale < TCP_MAX_WINSHIFT && (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) tp->request_r_scale++; soisconnecting(so); tcpstat.tcps_connattempt++; tp->t_state = TCPS_SYN_SENT; callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); tp->iss = tcp_new_isn(tp); tp->t_bw_rtseq = tp->iss; tcp_sendseqinit(tp); return 0; } #endif /* INET6 */ /* * Export TCP internal state information via a struct tcp_info, based on the * Linux 2.6 API. Not ABI compatible as our constants are mapped differently * (TCP state machine, etc). We export all information using FreeBSD-native * constants -- for example, the numeric values for tcpi_state will differ * from Linux. */ static void tcp_fill_info(tp, ti) struct tcpcb *tp; struct tcp_info *ti; { INP_LOCK_ASSERT(tp->t_inpcb); bzero(ti, sizeof(*ti)); ti->tcpi_state = tp->t_state; if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP)) ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; if (tp->sack_enable) ti->tcpi_options |= TCPI_OPT_SACK; if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) { ti->tcpi_options |= TCPI_OPT_WSCALE; ti->tcpi_snd_wscale = tp->snd_scale; ti->tcpi_rcv_wscale = tp->rcv_scale; } ti->tcpi_snd_ssthresh = tp->snd_ssthresh; ti->tcpi_snd_cwnd = tp->snd_cwnd; /* * FreeBSD-specific extension fields for tcp_info. */ ti->tcpi_rcv_space = tp->rcv_wnd; ti->tcpi_snd_wnd = tp->snd_wnd; ti->tcpi_snd_bwnd = tp->snd_bwnd; } /* * The new sockopt interface makes it possible for us to block in the * copyin/out step (if we take a page fault). Taking a page fault at * splnet() is probably a Bad Thing. (Since sockets and pcbs both now * use TSM, there probably isn't any need for this function to run at * splnet() any more. This needs more examination.) * * XXXRW: The locking here is wrong; we may take a page fault while holding * the inpcb lock. */ int tcp_ctloutput(so, sopt) struct socket *so; struct sockopt *sopt; { int error, opt, optval; struct inpcb *inp; struct tcpcb *tp; struct tcp_info ti; error = 0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL")); INP_LOCK(inp); if (sopt->sopt_level != IPPROTO_TCP) { INP_UNLOCK(inp); #ifdef INET6 if (INP_CHECK_SOCKAF(so, AF_INET6)) error = ip6_ctloutput(so, sopt); else #endif /* INET6 */ error = ip_ctloutput(so, sopt); return (error); } if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); switch (sopt->sopt_dir) { case SOPT_SET: switch (sopt->sopt_name) { #ifdef TCP_SIGNATURE case TCP_MD5SIG: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; if (optval > 0) tp->t_flags |= TF_SIGNATURE; else tp->t_flags &= ~TF_SIGNATURE; break; #endif /* TCP_SIGNATURE */ case TCP_NODELAY: case TCP_NOOPT: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; switch (sopt->sopt_name) { case TCP_NODELAY: opt = TF_NODELAY; break; case TCP_NOOPT: opt = TF_NOOPT; break; default: opt = 0; /* dead code to fool gcc */ break; } if (optval) tp->t_flags |= opt; else tp->t_flags &= ~opt; break; case TCP_NOPUSH: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; if (optval) tp->t_flags |= TF_NOPUSH; else { tp->t_flags &= ~TF_NOPUSH; error = tcp_output(tp); } break; case TCP_MAXSEG: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; if (optval > 0 && optval <= tp->t_maxseg && optval + 40 >= tcp_minmss) tp->t_maxseg = optval; else error = EINVAL; break; case TCP_INFO: error = EINVAL; break; default: error = ENOPROTOOPT; break; } break; case SOPT_GET: switch (sopt->sopt_name) { #ifdef TCP_SIGNATURE case TCP_MD5SIG: optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; error = sooptcopyout(sopt, &optval, sizeof optval); break; #endif case TCP_NODELAY: optval = tp->t_flags & TF_NODELAY; error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_MAXSEG: optval = tp->t_maxseg; error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_NOOPT: optval = tp->t_flags & TF_NOOPT; error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_NOPUSH: optval = tp->t_flags & TF_NOPUSH; error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_INFO: tcp_fill_info(tp, &ti); error = sooptcopyout(sopt, &ti, sizeof ti); break; default: error = ENOPROTOOPT; break; } break; } out: INP_UNLOCK(inp); return (error); } /* * tcp_sendspace and tcp_recvspace are the default send and receive window * sizes, respectively. These are obsolescent (this information should * be set by the route). */ u_long tcp_sendspace = 1024*32; SYSCTL_ULONG(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); u_long tcp_recvspace = 1024*64; SYSCTL_ULONG(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); /* * Attach TCP protocol to socket, allocating * internet protocol control block, tcp control block, * bufer space, and entering LISTEN state if to accept connections. */ static int tcp_attach(so) struct socket *so; { register struct tcpcb *tp; struct inpcb *inp; int error; #ifdef INET6 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0; #endif if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { error = soreserve(so, tcp_sendspace, tcp_recvspace); if (error) return (error); } INP_INFO_WLOCK(&tcbinfo); error = in_pcballoc(so, &tcbinfo); if (error) { INP_INFO_WUNLOCK(&tcbinfo); return (error); } inp = sotoinpcb(so); #ifdef INET6 if (isipv6) { inp->inp_vflag |= INP_IPV6; inp->in6p_hops = -1; /* use kernel default */ } else #endif inp->inp_vflag |= INP_IPV4; tp = tcp_newtcpcb(inp); if (tp == NULL) { #ifdef INET6 if (isipv6) { in6_pcbdetach(inp); in6_pcbfree(inp); } else { #endif in_pcbdetach(inp); in_pcbfree(inp); #ifdef INET6 } #endif INP_INFO_WUNLOCK(&tcbinfo); return (ENOBUFS); } tp->t_state = TCPS_CLOSED; INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (0); } /* * Initiate (or continue) disconnect. * If embryonic state, just send reset (once). * If in ``let data drain'' option and linger null, just drop. * Otherwise (hard), mark socket disconnecting and drop * current input data; switch states based on user close, and * send segment to peer (with FIN). */ static void tcp_disconnect(tp) register struct tcpcb *tp; { struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); /* * Neither tcp_close() nor tcp_drop() should return NULL, as the * socket is still open. */ if (tp->t_state < TCPS_ESTABLISHED) { tp = tcp_close(tp); KASSERT(tp != NULL, ("tcp_disconnect: tcp_close() returned NULL")); } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) { tp = tcp_drop(tp, 0); KASSERT(tp != NULL, ("tcp_disconnect: tcp_drop() returned NULL")); } else { soisdisconnecting(so); sbflush(&so->so_rcv); tcp_usrclosed(tp); if (!(inp->inp_vflag & INP_DROPPED)) tcp_output(tp); } } /* * User issued close, and wish to trail through shutdown states: * if never received SYN, just forget it. If got a SYN from peer, * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. * If already got a FIN from peer, then almost done; go to LAST_ACK * state. In all other cases, have already sent FIN to peer (e.g. * after PRU_SHUTDOWN), and just have to play tedious game waiting * for peer to send FIN or not respond to keep-alives, etc. * We can let the user exit from the close as soon as the FIN is acked. */ static void tcp_usrclosed(tp) register struct tcpcb *tp; { INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(tp->t_inpcb); switch (tp->t_state) { case TCPS_CLOSED: case TCPS_LISTEN: tp->t_state = TCPS_CLOSED; tp = tcp_close(tp); /* * tcp_close() should never return NULL here as the socket is * still open. */ KASSERT(tp != NULL, ("tcp_usrclosed: tcp_close() returned NULL")); break; case TCPS_SYN_SENT: case TCPS_SYN_RECEIVED: tp->t_flags |= TF_NEEDFIN; break; case TCPS_ESTABLISHED: tp->t_state = TCPS_FIN_WAIT_1; break; case TCPS_CLOSE_WAIT: tp->t_state = TCPS_LAST_ACK; break; } if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { soisdisconnected(tp->t_inpcb->inp_socket); /* To prevent the connection hanging in FIN_WAIT_2 forever. */ if (tp->t_state == TCPS_FIN_WAIT_2) callout_reset(tp->tt_2msl, tcp_maxidle, tcp_timer_2msl, tp); } }