Index: head/sys/netinet/tcp_input.c =================================================================== --- head/sys/netinet/tcp_input.c (revision 168902) +++ head/sys/netinet/tcp_input.c (revision 168903) @@ -1,3345 +1,3339 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 * $FreeBSD$ */ #include "opt_ipfw.h" /* for ipfw_fwd */ #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 /* for proc0 declaration */ #include #include #include #include #include #include #include #include /* before tcp_seq.h, for tcp_random18() */ #include #include #include #include #include #include #include #include #include /* required for icmp_var.h */ #include /* for ICMP_BANDLIM */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef TCPDEBUG #include #endif /* TCPDEBUG */ #ifdef FAST_IPSEC #include #include #endif /*FAST_IPSEC*/ #ifdef IPSEC #include #include #include #endif /*IPSEC*/ #include #include static const int tcprexmtthresh = 3; struct tcpstat tcpstat; SYSCTL_STRUCT(_net_inet_tcp, TCPCTL_STATS, stats, CTLFLAG_RW, &tcpstat , tcpstat, "TCP statistics (struct tcpstat, netinet/tcp_var.h)"); static int tcp_log_in_vain = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, log_in_vain, CTLFLAG_RW, &tcp_log_in_vain, 0, "Log all incoming TCP segments to closed ports"); static int blackhole = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, blackhole, CTLFLAG_RW, &blackhole, 0, "Do not send RST on segments to closed ports"); int tcp_delack_enabled = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, delayed_ack, CTLFLAG_RW, &tcp_delack_enabled, 0, "Delay ACK to try and piggyback it onto a data packet"); static int drop_synfin = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, drop_synfin, CTLFLAG_RW, &drop_synfin, 0, "Drop TCP packets with SYN+FIN set"); static int tcp_do_rfc3042 = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, rfc3042, CTLFLAG_RW, &tcp_do_rfc3042, 0, "Enable RFC 3042 (Limited Transmit)"); static int tcp_do_rfc3390 = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, rfc3390, CTLFLAG_RW, &tcp_do_rfc3390, 0, "Enable RFC 3390 (Increasing TCP's Initial Congestion Window)"); static int tcp_insecure_rst = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, insecure_rst, CTLFLAG_RW, &tcp_insecure_rst, 0, "Follow the old (insecure) criteria for accepting RST packets"); SYSCTL_NODE(_net_inet_tcp, OID_AUTO, reass, CTLFLAG_RW, 0, "TCP Segment Reassembly Queue"); static int tcp_reass_maxseg = 0; SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, maxsegments, CTLFLAG_RDTUN, &tcp_reass_maxseg, 0, "Global maximum number of TCP Segments in Reassembly Queue"); int tcp_reass_qsize = 0; SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, cursegments, CTLFLAG_RD, &tcp_reass_qsize, 0, "Global number of TCP Segments currently in Reassembly Queue"); static int tcp_reass_maxqlen = 48; SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, maxqlen, CTLFLAG_RW, &tcp_reass_maxqlen, 0, "Maximum number of TCP Segments per individual Reassembly Queue"); static int tcp_reass_overflows = 0; SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, overflows, CTLFLAG_RD, &tcp_reass_overflows, 0, "Global number of TCP Segment Reassembly Queue Overflows"); int tcp_do_autorcvbuf = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, recvbuf_auto, CTLFLAG_RW, &tcp_do_autorcvbuf, 0, "Enable automatic receive buffer sizing"); int tcp_autorcvbuf_inc = 16*1024; SYSCTL_INT(_net_inet_tcp, OID_AUTO, recvbuf_inc, CTLFLAG_RW, &tcp_autorcvbuf_inc, 0, "Incrementor step size of automatic receive buffer"); int tcp_autorcvbuf_max = 256*1024; SYSCTL_INT(_net_inet_tcp, OID_AUTO, recvbuf_max, CTLFLAG_RW, &tcp_autorcvbuf_max, 0, "Max size of automatic receive buffer"); struct inpcbhead tcb; #define tcb6 tcb /* for KAME src sync over BSD*'s */ struct inpcbinfo tcbinfo; static void tcp_dooptions(struct tcpopt *, u_char *, int, int); static int tcp_do_segment(struct mbuf *, struct tcphdr *, struct socket *, struct tcpcb *, int, int); static void tcp_dropwithreset(struct mbuf *, struct tcphdr *, struct tcpcb *, int, int); static void tcp_pulloutofband(struct socket *, struct tcphdr *, struct mbuf *, int); static int tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *); static void tcp_xmit_timer(struct tcpcb *, int); static void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *); static int tcp_timewait(struct inpcb *, struct tcpopt *, struct tcphdr *, struct mbuf *, int); /* Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. */ #ifdef INET6 #define ND6_HINT(tp) \ do { \ if ((tp) && (tp)->t_inpcb && \ ((tp)->t_inpcb->inp_vflag & INP_IPV6) != 0) \ nd6_nud_hint(NULL, NULL, 0); \ } while (0) #else #define ND6_HINT(tp) #endif /* * Indicate whether this ack should be delayed. We can delay the ack if * - there is no delayed ack timer in progress and * - our last ack wasn't a 0-sized window. We never want to delay * the ack that opens up a 0-sized window and * - delayed acks are enabled or * - this is a half-synchronized T/TCP connection. */ #define DELAY_ACK(tp) \ ((!tcp_timer_active(tp, TT_DELACK) && \ (tp->t_flags & TF_RXWIN0SENT) == 0) && \ (tcp_delack_enabled || (tp->t_flags & TF_NEEDSYN))) /* Initialize TCP reassembly queue */ static void tcp_reass_zone_change(void *tag) { tcp_reass_maxseg = nmbclusters / 16; uma_zone_set_max(tcp_reass_zone, tcp_reass_maxseg); } uma_zone_t tcp_reass_zone; void tcp_reass_init() { tcp_reass_maxseg = nmbclusters / 16; TUNABLE_INT_FETCH("net.inet.tcp.reass.maxsegments", &tcp_reass_maxseg); tcp_reass_zone = uma_zcreate("tcpreass", sizeof (struct tseg_qent), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcp_reass_zone, tcp_reass_maxseg); EVENTHANDLER_REGISTER(nmbclusters_change, tcp_reass_zone_change, NULL, EVENTHANDLER_PRI_ANY); } static int tcp_reass(struct tcpcb *tp, struct tcphdr *th, int *tlenp, struct mbuf *m) { struct tseg_qent *q; struct tseg_qent *p = NULL; struct tseg_qent *nq; struct tseg_qent *te = NULL; struct socket *so = tp->t_inpcb->inp_socket; int flags; INP_LOCK_ASSERT(tp->t_inpcb); /* * XXX: tcp_reass() is rather inefficient with its data structures * and should be rewritten (see NetBSD for optimizations). While * doing that it should move to its own file tcp_reass.c. */ /* * Call with th==NULL after become established to * force pre-ESTABLISHED data up to user socket. */ if (th == NULL) goto present; /* * Limit the number of segments in the reassembly queue to prevent * holding on to too many segments (and thus running out of mbufs). * Make sure to let the missing segment through which caused this * queue. Always keep one global queue entry spare to be able to * process the missing segment. */ if (th->th_seq != tp->rcv_nxt && (tcp_reass_qsize + 1 >= tcp_reass_maxseg || tp->t_segqlen >= tcp_reass_maxqlen)) { tcp_reass_overflows++; tcpstat.tcps_rcvmemdrop++; m_freem(m); *tlenp = 0; return (0); } /* * Allocate a new queue entry. If we can't, or hit the zone limit * just drop the pkt. */ te = uma_zalloc(tcp_reass_zone, M_NOWAIT); if (te == NULL) { tcpstat.tcps_rcvmemdrop++; m_freem(m); *tlenp = 0; return (0); } tp->t_segqlen++; tcp_reass_qsize++; /* * Find a segment which begins after this one does. */ LIST_FOREACH(q, &tp->t_segq, tqe_q) { if (SEQ_GT(q->tqe_th->th_seq, th->th_seq)) break; p = q; } /* * If there is a preceding segment, it may provide some of * our data already. If so, drop the data from the incoming * segment. If it provides all of our data, drop us. */ if (p != NULL) { int i; /* conversion to int (in i) handles seq wraparound */ i = p->tqe_th->th_seq + p->tqe_len - th->th_seq; if (i > 0) { if (i >= *tlenp) { tcpstat.tcps_rcvduppack++; tcpstat.tcps_rcvdupbyte += *tlenp; m_freem(m); uma_zfree(tcp_reass_zone, te); tp->t_segqlen--; tcp_reass_qsize--; /* * Try to present any queued data * at the left window edge to the user. * This is needed after the 3-WHS * completes. */ goto present; /* ??? */ } m_adj(m, i); *tlenp -= i; th->th_seq += i; } } tcpstat.tcps_rcvoopack++; tcpstat.tcps_rcvoobyte += *tlenp; /* * While we overlap succeeding segments trim them or, * if they are completely covered, dequeue them. */ while (q) { int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq; if (i <= 0) break; if (i < q->tqe_len) { q->tqe_th->th_seq += i; q->tqe_len -= i; m_adj(q->tqe_m, i); break; } nq = LIST_NEXT(q, tqe_q); LIST_REMOVE(q, tqe_q); m_freem(q->tqe_m); uma_zfree(tcp_reass_zone, q); tp->t_segqlen--; tcp_reass_qsize--; q = nq; } /* Insert the new segment queue entry into place. */ te->tqe_m = m; te->tqe_th = th; te->tqe_len = *tlenp; if (p == NULL) { LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q); } else { LIST_INSERT_AFTER(p, te, tqe_q); } present: /* * Present data to user, advancing rcv_nxt through * completed sequence space. */ if (!TCPS_HAVEESTABLISHED(tp->t_state)) return (0); q = LIST_FIRST(&tp->t_segq); if (!q || q->tqe_th->th_seq != tp->rcv_nxt) return (0); SOCKBUF_LOCK(&so->so_rcv); do { tp->rcv_nxt += q->tqe_len; flags = q->tqe_th->th_flags & TH_FIN; nq = LIST_NEXT(q, tqe_q); LIST_REMOVE(q, tqe_q); if (so->so_rcv.sb_state & SBS_CANTRCVMORE) m_freem(q->tqe_m); else sbappendstream_locked(&so->so_rcv, q->tqe_m); uma_zfree(tcp_reass_zone, q); tp->t_segqlen--; tcp_reass_qsize--; q = nq; } while (q && q->tqe_th->th_seq == tp->rcv_nxt); ND6_HINT(tp); sorwakeup_locked(so); return (flags); } /* * TCP input routine, follows pages 65-76 of the * protocol specification dated September, 1981 very closely. */ #ifdef INET6 int tcp6_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m = *mp; struct in6_ifaddr *ia6; IP6_EXTHDR_CHECK(m, *offp, sizeof(struct tcphdr), IPPROTO_DONE); /* * draft-itojun-ipv6-tcp-to-anycast * better place to put this in? */ ia6 = ip6_getdstifaddr(m); if (ia6 && (ia6->ia6_flags & IN6_IFF_ANYCAST)) { struct ip6_hdr *ip6; ip6 = mtod(m, struct ip6_hdr *); icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, (caddr_t)&ip6->ip6_dst - (caddr_t)ip6); return IPPROTO_DONE; } tcp_input(m, *offp); return IPPROTO_DONE; } #endif void tcp_input(struct mbuf *m, int off0) { struct tcphdr *th; struct ip *ip = NULL; struct ipovly *ipov; struct inpcb *inp = NULL; struct tcpcb *tp = NULL; struct socket *so = NULL; u_char *optp = NULL; int optlen = 0; int len, tlen, off; int drop_hdrlen; int thflags; int rstreason = 0; /* For badport_bandlim accounting purposes */ #ifdef IPFIREWALL_FORWARD struct m_tag *fwd_tag; #endif #ifdef INET6 struct ip6_hdr *ip6 = NULL; int isipv6; char ip6buf[INET6_ADDRSTRLEN]; #else const int isipv6 = 0; #endif struct tcpopt to; /* options in this segment */ #ifdef TCPDEBUG /* * The size of tcp_saveipgen must be the size of the max ip header, * now IPv6. */ u_char tcp_saveipgen[IP6_HDR_LEN]; struct tcphdr tcp_savetcp; short ostate = 0; #endif #ifdef INET6 isipv6 = (mtod(m, struct ip *)->ip_v == 6) ? 1 : 0; #endif to.to_flags = 0; tcpstat.tcps_rcvtotal++; if (isipv6) { #ifdef INET6 /* IP6_EXTHDR_CHECK() is already done at tcp6_input() */ ip6 = mtod(m, struct ip6_hdr *); tlen = sizeof(*ip6) + ntohs(ip6->ip6_plen) - off0; if (in6_cksum(m, IPPROTO_TCP, off0, tlen)) { tcpstat.tcps_rcvbadsum++; goto drop; } th = (struct tcphdr *)((caddr_t)ip6 + off0); /* * Be proactive about unspecified IPv6 address in source. * As we use all-zero to indicate unbounded/unconnected pcb, * unspecified IPv6 address can be used to confuse us. * * Note that packets with unspecified IPv6 destination is * already dropped in ip6_input. */ if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { /* XXX stat */ goto drop; } #else th = NULL; /* XXX: avoid compiler warning */ #endif } else { /* * Get IP and TCP header together in first mbuf. * Note: IP leaves IP header in first mbuf. */ if (off0 > sizeof (struct ip)) { ip_stripoptions(m, (struct mbuf *)0); off0 = sizeof(struct ip); } if (m->m_len < sizeof (struct tcpiphdr)) { if ((m = m_pullup(m, sizeof (struct tcpiphdr))) == NULL) { tcpstat.tcps_rcvshort++; return; } } ip = mtod(m, struct ip *); ipov = (struct ipovly *)ip; th = (struct tcphdr *)((caddr_t)ip + off0); tlen = ip->ip_len; if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) th->th_sum = m->m_pkthdr.csum_data; else th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl(m->m_pkthdr.csum_data + ip->ip_len + IPPROTO_TCP)); th->th_sum ^= 0xffff; #ifdef TCPDEBUG ipov->ih_len = (u_short)tlen; ipov->ih_len = htons(ipov->ih_len); #endif } else { /* * Checksum extended TCP header and data. */ len = sizeof (struct ip) + tlen; bzero(ipov->ih_x1, sizeof(ipov->ih_x1)); ipov->ih_len = (u_short)tlen; ipov->ih_len = htons(ipov->ih_len); th->th_sum = in_cksum(m, len); } if (th->th_sum) { tcpstat.tcps_rcvbadsum++; goto drop; } /* Re-initialization for later version check */ ip->ip_v = IPVERSION; } /* * Check that TCP offset makes sense, * pull out TCP options and adjust length. XXX */ off = th->th_off << 2; if (off < sizeof (struct tcphdr) || off > tlen) { tcpstat.tcps_rcvbadoff++; goto drop; } tlen -= off; /* tlen is used instead of ti->ti_len */ if (off > sizeof (struct tcphdr)) { if (isipv6) { #ifdef INET6 IP6_EXTHDR_CHECK(m, off0, off, ); ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)((caddr_t)ip6 + off0); #endif } else { if (m->m_len < sizeof(struct ip) + off) { if ((m = m_pullup(m, sizeof (struct ip) + off)) == NULL) { tcpstat.tcps_rcvshort++; return; } ip = mtod(m, struct ip *); ipov = (struct ipovly *)ip; th = (struct tcphdr *)((caddr_t)ip + off0); } } optlen = off - sizeof (struct tcphdr); optp = (u_char *)(th + 1); } thflags = th->th_flags; /* * If the drop_synfin option is enabled, drop all packets with * both the SYN and FIN bits set. This prevents e.g. nmap from * identifying the TCP/IP stack. * * This is a violation of the TCP specification. */ if (drop_synfin && (thflags & (TH_SYN|TH_FIN)) == (TH_SYN|TH_FIN)) goto drop; /* * Convert TCP protocol specific fields to host format. */ th->th_seq = ntohl(th->th_seq); th->th_ack = ntohl(th->th_ack); th->th_win = ntohs(th->th_win); th->th_urp = ntohs(th->th_urp); /* * Delay dropping TCP, IP headers, IPv6 ext headers, and TCP options. */ drop_hdrlen = off0 + off; /* * Locate pcb for segment. */ INP_INFO_WLOCK(&tcbinfo); findpcb: INP_INFO_WLOCK_ASSERT(&tcbinfo); #ifdef IPFIREWALL_FORWARD /* Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. */ fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); if (fwd_tag != NULL && isipv6 == 0) { /* IPv6 support is not yet */ struct sockaddr_in *next_hop; next_hop = (struct sockaddr_in *)(fwd_tag+1); /* * Transparently forwarded. Pretend to be the destination. * already got one like this? */ inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, th->th_sport, ip->ip_dst, th->th_dport, 0, m->m_pkthdr.rcvif); if (!inp) { /* It's new. Try to find the ambushing socket. */ inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, th->th_sport, next_hop->sin_addr, next_hop->sin_port ? ntohs(next_hop->sin_port) : th->th_dport, INPLOOKUP_WILDCARD, m->m_pkthdr.rcvif); } /* Remove the tag from the packet. We don't need it anymore. */ m_tag_delete(m, fwd_tag); } else #endif /* IPFIREWALL_FORWARD */ { if (isipv6) { #ifdef INET6 inp = in6_pcblookup_hash(&tcbinfo, &ip6->ip6_src, th->th_sport, &ip6->ip6_dst, th->th_dport, INPLOOKUP_WILDCARD, m->m_pkthdr.rcvif); #endif } else inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, th->th_sport, ip->ip_dst, th->th_dport, INPLOOKUP_WILDCARD, m->m_pkthdr.rcvif); } #if defined(IPSEC) || defined(FAST_IPSEC) #ifdef INET6 if (isipv6 && inp != NULL && ipsec6_in_reject(m, inp)) { #ifdef IPSEC ipsec6stat.in_polvio++; #endif goto dropunlock; } else #endif /* INET6 */ if (inp != NULL && ipsec4_in_reject(m, inp)) { #ifdef IPSEC ipsecstat.in_polvio++; #endif goto dropunlock; } #endif /*IPSEC || FAST_IPSEC*/ /* * If the INPCB does not exist then all data in the incoming * segment is discarded and an appropriate RST is sent back. */ if (inp == NULL) { /* * Log communication attempts to ports that are not * in use. */ if ((tcp_log_in_vain == 1 && (thflags & TH_SYN)) || tcp_log_in_vain == 2) { #ifndef INET6 char dbuf[4*sizeof "123"], sbuf[4*sizeof "123"]; #else char dbuf[INET6_ADDRSTRLEN+2], sbuf[INET6_ADDRSTRLEN+2]; if (isipv6) { strcpy(dbuf, "["); strcat(dbuf, ip6_sprintf(ip6buf, &ip6->ip6_dst)); strcat(dbuf, "]"); strcpy(sbuf, "["); strcat(sbuf, ip6_sprintf(ip6buf, &ip6->ip6_src)); strcat(sbuf, "]"); } else #endif /* INET6 */ { strcpy(dbuf, inet_ntoa(ip->ip_dst)); strcpy(sbuf, inet_ntoa(ip->ip_src)); } log(LOG_INFO, "Connection attempt to TCP %s:%d " "from %s:%d flags:0x%02x\n", dbuf, ntohs(th->th_dport), sbuf, ntohs(th->th_sport), thflags); } /* * When blackholing do not respond with a RST but * completely ignore the segment and drop it. */ if ((blackhole == 1 && (thflags & TH_SYN)) || blackhole == 2) goto dropunlock; rstreason = BANDLIM_RST_CLOSEDPORT; goto dropwithreset; } INP_LOCK(inp); /* Check the minimum TTL for socket. */ if (inp->inp_ip_minttl != 0) { #ifdef INET6 if (isipv6 && inp->inp_ip_minttl > ip6->ip6_hlim) goto dropunlock; else #endif if (inp->inp_ip_minttl > ip->ip_ttl) goto dropunlock; } /* * A previous connection in TIMEWAIT state is supposed to catch * stray or duplicate segments arriving late. If this segment * was a legitimate new connection attempt the old INPCB gets * removed and we can try again to find a listening socket. */ if (inp->inp_vflag & INP_TIMEWAIT) { if (thflags & TH_SYN) tcp_dooptions(&to, optp, optlen, TO_SYN); if (tcp_timewait(inp, &to, th, m, tlen)) goto findpcb; /* tcp_timewait unlocks inp. */ INP_INFO_WUNLOCK(&tcbinfo); return; } /* * The TCPCB may no longer exist if the connection is winding * down or it is in the CLOSED state. Either way we drop the * segment and send an appropriate response. */ tp = intotcpcb(inp); if (tp == NULL) { INP_UNLOCK(inp); rstreason = BANDLIM_RST_CLOSEDPORT; goto dropwithreset; } if (tp->t_state == TCPS_CLOSED) goto dropunlock; /* XXX: dropwithreset??? */ #ifdef MAC INP_LOCK_ASSERT(inp); if (mac_check_inpcb_deliver(inp, m)) goto dropunlock; #endif so = inp->inp_socket; KASSERT(so != NULL, ("%s: so == NULL", __func__)); #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) { ostate = tp->t_state; if (isipv6) bcopy((char *)ip6, (char *)tcp_saveipgen, sizeof(*ip6)); else bcopy((char *)ip, (char *)tcp_saveipgen, sizeof(*ip)); tcp_savetcp = *th; } #endif /* * When the socket is accepting connections (the INPCB is in LISTEN * state) we look into the SYN cache if this is a new connection * attempt or the completion of a previous one. */ if (so->so_options & SO_ACCEPTCONN) { struct in_conninfo inc; bzero(&inc, sizeof(inc)); inc.inc_isipv6 = isipv6; #ifdef INET6 if (isipv6) { inc.inc6_faddr = ip6->ip6_src; inc.inc6_laddr = ip6->ip6_dst; } else #endif { inc.inc_faddr = ip->ip_src; inc.inc_laddr = ip->ip_dst; } inc.inc_fport = th->th_sport; inc.inc_lport = th->th_dport; /* * If the state is LISTEN then ignore segment if it contains * a RST. If the segment contains an ACK then it is bad and * send a RST. If it does not contain a SYN then it is not * interesting; drop it. * * If the state is SYN_RECEIVED (syncache) and seg contains * an ACK, but not for our SYN/ACK, send a RST. If the seg * contains a RST, check the sequence number to see if it * is a valid reset segment. */ if ((thflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { if ((thflags & (TH_RST|TH_ACK|TH_SYN)) == TH_ACK) { /* * Parse the TCP options here because * syncookies need access to the reflected * timestamp. */ tcp_dooptions(&to, optp, optlen, 0); if (!syncache_expand(&inc, &to, th, &so, m)) { /* * No syncache entry or ACK was not * for our SYN/ACK. Send a RST. */ rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } if (so == NULL) { /* * We completed the 3-way handshake * but could not allocate a socket * either due to memory shortage, * listen queue length limits or * global socket limits. */ rstreason = BANDLIM_UNLIMITED; goto dropwithreset; } /* * Socket is created in state SYN_RECEIVED. * Continue processing segment. */ INP_UNLOCK(inp); /* listen socket */ inp = sotoinpcb(so); INP_LOCK(inp); /* new connection */ tp = intotcpcb(inp); /* * Process the segment and the data it * contains. tcp_do_segment() consumes * the mbuf chain and unlocks the inpcb. * XXX: The potential return value of * TIME_WAIT nuked is supposed to be * handled above. */ if (tcp_do_segment(m, th, so, tp, drop_hdrlen, tlen)) goto findpcb; /* TIME_WAIT nuked */ return; } if (thflags & TH_RST) { syncache_chkrst(&inc, th); goto dropunlock; } if (thflags & TH_ACK) { syncache_badack(&inc); tcpstat.tcps_badsyn++; rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } goto dropunlock; } /* * Segment's flags are (SYN) or (SYN|FIN). */ #ifdef INET6 /* * If deprecated address is forbidden, * we do not accept SYN to deprecated interface * address to prevent any new inbound connection from * getting established. * When we do not accept SYN, we send a TCP RST, * with deprecated source address (instead of dropping * it). We compromise it as it is much better for peer * to send a RST, and RST will be the final packet * for the exchange. * * If we do not forbid deprecated addresses, we accept * the SYN packet. RFC2462 does not suggest dropping * SYN in this case. * If we decipher RFC2462 5.5.4, it says like this: * 1. use of deprecated addr with existing * communication is okay - "SHOULD continue to be * used" * 2. use of it with new communication: * (2a) "SHOULD NOT be used if alternate address * with sufficient scope is available" * (2b) nothing mentioned otherwise. * Here we fall into (2b) case as we have no choice in * our source address selection - we must obey the peer. * * The wording in RFC2462 is confusing, and there are * multiple description text for deprecated address * handling - worse, they are not exactly the same. * I believe 5.5.4 is the best one, so we follow 5.5.4. */ if (isipv6 && !ip6_use_deprecated) { struct in6_ifaddr *ia6; if ((ia6 = ip6_getdstifaddr(m)) && (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { INP_UNLOCK(inp); tp = NULL; rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } } #endif /* * Basic sanity checks on incoming SYN requests: * * Don't bother responding if the destination was a * broadcast according to RFC1122 4.2.3.10, p. 104. * * If it is from this socket, drop it, it must be forged. * * Note that it is quite possible to receive unicast * link-layer packets with a broadcast IP address. Use * in_broadcast() to find them. */ if (m->m_flags & (M_BCAST|M_MCAST)) goto dropunlock; if (isipv6) { #ifdef INET6 if (th->th_dport == th->th_sport && IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &ip6->ip6_src)) goto dropunlock; if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto dropunlock; #endif } else { if (th->th_dport == th->th_sport && ip->ip_dst.s_addr == ip->ip_src.s_addr) goto dropunlock; if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || ip->ip_src.s_addr == htonl(INADDR_BROADCAST) || in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) goto dropunlock; } /* * SYN appears to be valid. Create compressed TCP state * for syncache. */ - if (so->so_qlen <= so->so_qlimit) { #ifdef TCPDEBUG - if (so->so_options & SO_DEBUG) - tcp_trace(TA_INPUT, ostate, tp, - (void *)tcp_saveipgen, &tcp_savetcp, 0); + if (so->so_options & SO_DEBUG) + tcp_trace(TA_INPUT, ostate, tp, + (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif - tcp_dooptions(&to, optp, optlen, TO_SYN); - if (!syncache_add(&inc, &to, th, inp, &so, m)) - goto dropunlock; - /* - * Entry added to syncache, mbuf used to - * send SYN-ACK packet. Everything unlocked - * already. - */ - return; - } - /* Catch all. Everthing that makes it down here is junk. */ - goto dropunlock; + tcp_dooptions(&to, optp, optlen, TO_SYN); + syncache_add(&inc, &to, th, inp, &so, m); + /* + * Entry added to syncache and mbuf consumed. + * Everything unlocked already by syncache_add(). + */ + return; } /* * Segment belongs to a connection in SYN_SENT, ESTABLISHED or * later state. tcp_do_segment() always consumes the mbuf chain * and unlocks the inpcb. */ if (tcp_do_segment(m, th, so, tp, drop_hdrlen, tlen)) goto findpcb; /* XXX: TIME_WAIT was nuked. */ return; dropwithreset: INP_INFO_WLOCK_ASSERT(&tcbinfo); tcp_dropwithreset(m, th, tp, tlen, rstreason); m = NULL; /* mbuf chain got consumed. */ dropunlock: INP_INFO_WLOCK_ASSERT(&tcbinfo); if (tp != NULL) INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); drop: INP_INFO_UNLOCK_ASSERT(&tcbinfo); if (m != NULL) m_freem(m); return; } static int tcp_do_segment(struct mbuf *m, struct tcphdr *th, struct socket *so, struct tcpcb *tp, int drop_hdrlen, int tlen) { int thflags, acked, ourfinisacked, needoutput = 0; int headlocked = 1; int rstreason, todrop, win; u_long tiwin; struct tcpopt to; #ifdef TCPDEBUG /* * The size of tcp_saveipgen must be the size of the max ip header, * now IPv6. */ u_char tcp_saveipgen[IP6_HDR_LEN]; struct tcphdr tcp_savetcp; short ostate = 0; #endif thflags = th->th_flags; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(tp->t_inpcb); KASSERT(tp->t_state > TCPS_LISTEN, ("%s: TCPS_LISTEN", __func__)); /* * Segment received on connection. * Reset idle time and keep-alive timer. */ tp->t_rcvtime = ticks; if (TCPS_HAVEESTABLISHED(tp->t_state)) tcp_timer_activate(tp, TT_KEEP, tcp_keepidle); /* * Unscale the window into a 32-bit value. * This value is bogus for the TCPS_SYN_SENT state * and is overwritten later. */ tiwin = th->th_win << tp->snd_scale; /* * Parse options on any incoming segment. */ tcp_dooptions(&to, (u_char *)(th + 1), (th->th_off << 2) - sizeof(struct tcphdr), (thflags & TH_SYN) ? TO_SYN : 0); /* * If echoed timestamp is later than the current time, * fall back to non RFC1323 RTT calculation. Normalize * timestamp if syncookies were used when this connection * was established. */ if ((to.to_flags & TOF_TS) && (to.to_tsecr != 0)) { to.to_tsecr -= tp->ts_offset; if (TSTMP_GT(to.to_tsecr, ticks)) to.to_tsecr = 0; } /* * Process options only when we get SYN/ACK back. The SYN case * for incoming connections is handled in tcp_syncache. * XXX this is traditional behavior, may need to be cleaned up. */ if (tp->t_state == TCPS_SYN_SENT && (thflags & TH_SYN)) { if ((to.to_flags & TOF_SCALE) && (tp->t_flags & TF_REQ_SCALE)) { tp->t_flags |= TF_RCVD_SCALE; tp->snd_scale = to.to_wscale; tp->snd_wnd = th->th_win << tp->snd_scale; tiwin = tp->snd_wnd; } if (to.to_flags & TOF_TS) { tp->t_flags |= TF_RCVD_TSTMP; tp->ts_recent = to.to_tsval; tp->ts_recent_age = ticks; } /* Initial send window, already scaled. */ tp->snd_wnd = th->th_win; if (to.to_flags & TOF_MSS) tcp_mss(tp, to.to_mss); if (tp->sack_enable) { if (!(to.to_flags & TOF_SACKPERM)) tp->sack_enable = 0; else tp->t_flags |= TF_SACK_PERMIT; } } /* * Header prediction: check for the two common cases * of a uni-directional data xfer. If the packet has * no control flags, is in-sequence, the window didn't * change and we're not retransmitting, it's a * candidate. If the length is zero and the ack moved * forward, we're the sender side of the xfer. Just * free the data acked & wake any higher level process * that was blocked waiting for space. If the length * is non-zero and the ack didn't move, we're the * receiver side. If we're getting packets in-order * (the reassembly queue is empty), add the data to * the socket buffer and note that we need a delayed ack. * Make sure that the hidden state-flags are also off. * Since we check for TCPS_ESTABLISHED above, it can only * be TH_NEEDSYN. */ if (tp->t_state == TCPS_ESTABLISHED && (thflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK && ((tp->t_flags & (TF_NEEDSYN|TF_NEEDFIN)) == 0) && ((to.to_flags & TOF_TS) == 0 || TSTMP_GEQ(to.to_tsval, tp->ts_recent)) && th->th_seq == tp->rcv_nxt && tiwin && tiwin == tp->snd_wnd && tp->snd_nxt == tp->snd_max) { /* * If last ACK falls within this segment's sequence numbers, * record the timestamp. * NOTE that the test is modified according to the latest * proposal of the tcplw@cray.com list (Braden 1993/04/26). */ if ((to.to_flags & TOF_TS) != 0 && SEQ_LEQ(th->th_seq, tp->last_ack_sent)) { tp->ts_recent_age = ticks; tp->ts_recent = to.to_tsval; } if (tlen == 0) { if (SEQ_GT(th->th_ack, tp->snd_una) && SEQ_LEQ(th->th_ack, tp->snd_max) && tp->snd_cwnd >= tp->snd_wnd && ((!tcp_do_newreno && !tp->sack_enable && tp->t_dupacks < tcprexmtthresh) || ((tcp_do_newreno || tp->sack_enable) && !IN_FASTRECOVERY(tp) && (to.to_flags & TOF_SACK) == 0 && TAILQ_EMPTY(&tp->snd_holes)))) { KASSERT(headlocked, ("%s: headlocked", __func__)); INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; /* * this is a pure ack for outstanding data. */ ++tcpstat.tcps_predack; /* * "bad retransmit" recovery */ if (tp->t_rxtshift == 1 && ticks < tp->t_badrxtwin) { ++tcpstat.tcps_sndrexmitbad; tp->snd_cwnd = tp->snd_cwnd_prev; tp->snd_ssthresh = tp->snd_ssthresh_prev; tp->snd_recover = tp->snd_recover_prev; if (tp->t_flags & TF_WASFRECOVERY) ENTER_FASTRECOVERY(tp); tp->snd_nxt = tp->snd_max; tp->t_badrxtwin = 0; } /* * Recalculate the transmit timer / rtt. * * Some boxes send broken timestamp replies * during the SYN+ACK phase, ignore * timestamps of 0 or we could calculate a * huge RTT and blow up the retransmit timer. */ if ((to.to_flags & TOF_TS) != 0 && to.to_tsecr) { if (!tp->t_rttlow || tp->t_rttlow > ticks - to.to_tsecr) tp->t_rttlow = ticks - to.to_tsecr; tcp_xmit_timer(tp, ticks - to.to_tsecr + 1); } else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) { if (!tp->t_rttlow || tp->t_rttlow > ticks - tp->t_rtttime) tp->t_rttlow = ticks - tp->t_rtttime; tcp_xmit_timer(tp, ticks - tp->t_rtttime); } tcp_xmit_bandwidth_limit(tp, th->th_ack); acked = th->th_ack - tp->snd_una; tcpstat.tcps_rcvackpack++; tcpstat.tcps_rcvackbyte += acked; sbdrop(&so->so_snd, acked); if (SEQ_GT(tp->snd_una, tp->snd_recover) && SEQ_LEQ(th->th_ack, tp->snd_recover)) tp->snd_recover = th->th_ack - 1; tp->snd_una = th->th_ack; /* * pull snd_wl2 up to prevent seq wrap relative * to th_ack. */ tp->snd_wl2 = th->th_ack; tp->t_dupacks = 0; m_freem(m); ND6_HINT(tp); /* some progress has been done */ /* * If all outstanding data are acked, stop * retransmit timer, otherwise restart timer * using current (possibly backed-off) value. * If process is waiting for space, * wakeup/selwakeup/signal. If data * are ready to send, let tcp_output * decide between more output or persist. #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif */ if (tp->snd_una == tp->snd_max) tcp_timer_activate(tp, TT_REXMT, 0); else if (!tcp_timer_active(tp, TT_PERSIST)) tcp_timer_activate(tp, TT_REXMT, tp->t_rxtcur); sowwakeup(so); if (so->so_snd.sb_cc) (void) tcp_output(tp); goto check_delack; } } else if (th->th_ack == tp->snd_una && LIST_EMPTY(&tp->t_segq) && tlen <= sbspace(&so->so_rcv)) { int newsize = 0; /* automatic sockbuf scaling */ KASSERT(headlocked, ("%s: headlocked", __func__)); INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; /* * this is a pure, in-sequence data packet * with nothing on the reassembly queue and * we have enough buffer space to take it. */ /* Clean receiver SACK report if present */ if (tp->sack_enable && tp->rcv_numsacks) tcp_clean_sackreport(tp); ++tcpstat.tcps_preddat; tp->rcv_nxt += tlen; /* * Pull snd_wl1 up to prevent seq wrap relative to * th_seq. */ tp->snd_wl1 = th->th_seq; /* * Pull rcv_up up to prevent seq wrap relative to * rcv_nxt. */ tp->rcv_up = tp->rcv_nxt; tcpstat.tcps_rcvpack++; tcpstat.tcps_rcvbyte += tlen; ND6_HINT(tp); /* some progress has been done */ #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif /* * Automatic sizing of receive socket buffer. Often the send * buffer size is not optimally adjusted to the actual network * conditions at hand (delay bandwidth product). Setting the * buffer size too small limits throughput on links with high * bandwidth and high delay (eg. trans-continental/oceanic links). * * On the receive side the socket buffer memory is only rarely * used to any significant extent. This allows us to be much * more aggressive in scaling the receive socket buffer. For * the case that the buffer space is actually used to a large * extent and we run out of kernel memory we can simply drop * the new segments; TCP on the sender will just retransmit it * later. Setting the buffer size too big may only consume too * much kernel memory if the application doesn't read() from * the socket or packet loss or reordering makes use of the * reassembly queue. * * The criteria to step up the receive buffer one notch are: * 1. the number of bytes received during the time it takes * one timestamp to be reflected back to us (the RTT); * 2. received bytes per RTT is within seven eighth of the * current socket buffer size; * 3. receive buffer size has not hit maximal automatic size; * * This algorithm does one step per RTT at most and only if * we receive a bulk stream w/o packet losses or reorderings. * Shrinking the buffer during idle times is not necessary as * it doesn't consume any memory when idle. * * TODO: Only step up if the application is actually serving * the buffer to better manage the socket buffer resources. */ if (tcp_do_autorcvbuf && to.to_tsecr && (so->so_rcv.sb_flags & SB_AUTOSIZE)) { if (to.to_tsecr > tp->rfbuf_ts && to.to_tsecr - tp->rfbuf_ts < hz) { if (tp->rfbuf_cnt > (so->so_rcv.sb_hiwat / 8 * 7) && so->so_rcv.sb_hiwat < tcp_autorcvbuf_max) { newsize = min(so->so_rcv.sb_hiwat + tcp_autorcvbuf_inc, tcp_autorcvbuf_max); } /* Start over with next RTT. */ tp->rfbuf_ts = 0; tp->rfbuf_cnt = 0; } else tp->rfbuf_cnt += tlen; /* add up */ } /* Add data to socket buffer. */ SOCKBUF_LOCK(&so->so_rcv); if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { m_freem(m); } else { /* * Set new socket buffer size. * Give up when limit is reached. */ if (newsize) if (!sbreserve_locked(&so->so_rcv, newsize, so, curthread)) so->so_rcv.sb_flags &= ~SB_AUTOSIZE; m_adj(m, drop_hdrlen); /* delayed header drop */ sbappendstream_locked(&so->so_rcv, m); } sorwakeup_locked(so); if (DELAY_ACK(tp)) { tp->t_flags |= TF_DELACK; } else { tp->t_flags |= TF_ACKNOW; tcp_output(tp); } goto check_delack; } } /* * Calculate amount of space in receive window, * and then do TCP input processing. * Receive window is amount of space in rcv queue, * but not less than advertised window. */ win = sbspace(&so->so_rcv); if (win < 0) win = 0; tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); /* Reset receive buffer auto scaling when not in bulk receive mode. */ tp->rfbuf_ts = 0; tp->rfbuf_cnt = 0; switch (tp->t_state) { /* * If the state is SYN_RECEIVED: * if seg contains an ACK, but not for our SYN/ACK, send a RST. */ case TCPS_SYN_RECEIVED: if ((thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->snd_una) || SEQ_GT(th->th_ack, tp->snd_max))) { rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } break; /* * If the state is SYN_SENT: * if seg contains an ACK, but not for our SYN, drop the input. * if seg contains a RST, then drop the connection. * if seg does not contain SYN, then drop it. * Otherwise this is an acceptable SYN segment * initialize tp->rcv_nxt and tp->irs * if seg contains ack then advance tp->snd_una * if SYN has been acked change to ESTABLISHED else SYN_RCVD state * arrange for segment to be acked (eventually) * continue processing rest of data/controls, beginning with URG */ case TCPS_SYN_SENT: if ((thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->iss) || SEQ_GT(th->th_ack, tp->snd_max))) { rstreason = BANDLIM_UNLIMITED; goto dropwithreset; } if (thflags & TH_RST) { if (thflags & TH_ACK) { KASSERT(headlocked, ("%s: after_listen: " "tcp_drop.2: head not locked", __func__)); tp = tcp_drop(tp, ECONNREFUSED); } goto drop; } if ((thflags & TH_SYN) == 0) goto drop; tp->irs = th->th_seq; tcp_rcvseqinit(tp); if (thflags & TH_ACK) { tcpstat.tcps_connects++; soisconnected(so); #ifdef MAC SOCK_LOCK(so); mac_set_socket_peer_from_mbuf(m, so); SOCK_UNLOCK(so); #endif /* Do window scaling on this connection? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->rcv_scale = tp->request_r_scale; } tp->rcv_adv += tp->rcv_wnd; tp->snd_una++; /* SYN is acked */ /* * If there's data, delay ACK; if there's also a FIN * ACKNOW will be turned on later. */ if (DELAY_ACK(tp) && tlen != 0) tcp_timer_activate(tp, TT_DELACK, tcp_delacktime); else tp->t_flags |= TF_ACKNOW; /* * Received in SYN_SENT[*] state. * Transitions: * SYN_SENT --> ESTABLISHED * SYN_SENT* --> FIN_WAIT_1 */ tp->t_starttime = ticks; if (tp->t_flags & TF_NEEDFIN) { tp->t_state = TCPS_FIN_WAIT_1; tp->t_flags &= ~TF_NEEDFIN; thflags &= ~TH_SYN; } else { tp->t_state = TCPS_ESTABLISHED; tcp_timer_activate(tp, TT_KEEP, tcp_keepidle); } } else { /* * Received initial SYN in SYN-SENT[*] state => * simultaneous open. If segment contains CC option * and there is a cached CC, apply TAO test. * If it succeeds, connection is * half-synchronized. * Otherwise, do 3-way handshake: * SYN-SENT -> SYN-RECEIVED * SYN-SENT* -> SYN-RECEIVED* * If there was no CC option, clear cached CC value. */ tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN); tcp_timer_activate(tp, TT_REXMT, 0); tp->t_state = TCPS_SYN_RECEIVED; } KASSERT(headlocked, ("%s: trimthenstep6: head not locked", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); /* * Advance th->th_seq to correspond to first data byte. * If data, trim to stay within window, * dropping FIN if necessary. */ th->th_seq++; if (tlen > tp->rcv_wnd) { todrop = tlen - tp->rcv_wnd; m_adj(m, -todrop); tlen = tp->rcv_wnd; thflags &= ~TH_FIN; tcpstat.tcps_rcvpackafterwin++; tcpstat.tcps_rcvbyteafterwin += todrop; } tp->snd_wl1 = th->th_seq - 1; tp->rcv_up = th->th_seq; /* * Client side of transaction: already sent SYN and data. * If the remote host used T/TCP to validate the SYN, * our data will be ACK'd; if so, enter normal data segment * processing in the middle of step 5, ack processing. * Otherwise, goto step 6. */ if (thflags & TH_ACK) goto process_ACK; goto step6; /* * If the state is LAST_ACK or CLOSING or TIME_WAIT: * do normal processing. * * NB: Leftover from RFC1644 T/TCP. Cases to be reused later. */ case TCPS_LAST_ACK: case TCPS_CLOSING: case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); break; /* continue normal processing */ } /* * States other than LISTEN or SYN_SENT. * First check the RST flag and sequence number since reset segments * are exempt from the timestamp and connection count tests. This * fixes a bug introduced by the Stevens, vol. 2, p. 960 bugfix * below which allowed reset segments in half the sequence space * to fall though and be processed (which gives forged reset * segments with a random sequence number a 50 percent chance of * killing a connection). * Then check timestamp, if present. * Then check the connection count, if present. * Then check that at least some bytes of segment are within * receive window. If segment begins before rcv_nxt, * drop leading data (and SYN); if nothing left, just ack. * * * 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. * Note: this does not take into account delayed ACKs, so * we should test against last_ack_sent instead of rcv_nxt. * 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. * Note 2: Paul Watson's paper "Slipping in the Window" has shown * that brute force RST attacks are possible. To combat this, * we use a much stricter check while in the ESTABLISHED state, * only accepting RSTs where the sequence number is equal to * last_ack_sent. In all other states (the states in which a * RST is more likely), the more permissive check is used. * If we have multiple segments in flight, the intial reset * segment sequence numbers will be to the left of last_ack_sent, * but they will eventually catch up. * In any case, it never made sense to trim reset segments to * fit the receive window since RFC 1122 says: * 4.2.2.12 RST Segment: RFC-793 Section 3.4 * * A TCP SHOULD allow a received RST segment to include data. * * DISCUSSION * It has been suggested that a RST segment could contain * ASCII text that encoded and explained the cause of the * RST. No standard has yet been established for such * data. * * If the reset segment passes the sequence number test examine * the state: * SYN_RECEIVED STATE: * If passive open, return to LISTEN state. * If active open, inform user that connection was refused. * ESTABLISHED, FIN_WAIT_1, FIN_WAIT_2, CLOSE_WAIT STATES: * Inform user that connection was reset, and close tcb. * CLOSING, LAST_ACK STATES: * Close the tcb. * TIME_WAIT STATE: * Drop the segment - see Stevens, vol. 2, p. 964 and * RFC 1337. */ if (thflags & TH_RST) { if (SEQ_GEQ(th->th_seq, tp->last_ack_sent - 1) && SEQ_LEQ(th->th_seq, tp->last_ack_sent + tp->rcv_wnd)) { switch (tp->t_state) { case TCPS_SYN_RECEIVED: so->so_error = ECONNREFUSED; goto close; case TCPS_ESTABLISHED: if (tcp_insecure_rst == 0 && !(SEQ_GEQ(th->th_seq, tp->rcv_nxt - 1) && SEQ_LEQ(th->th_seq, tp->rcv_nxt + 1)) && !(SEQ_GEQ(th->th_seq, tp->last_ack_sent - 1) && SEQ_LEQ(th->th_seq, tp->last_ack_sent + 1))) { tcpstat.tcps_badrst++; goto drop; } case TCPS_FIN_WAIT_1: case TCPS_FIN_WAIT_2: case TCPS_CLOSE_WAIT: so->so_error = ECONNRESET; close: tp->t_state = TCPS_CLOSED; tcpstat.tcps_drops++; KASSERT(headlocked, ("%s: trimthenstep6: " "tcp_close: head not locked", __func__)); tp = tcp_close(tp); break; case TCPS_CLOSING: case TCPS_LAST_ACK: KASSERT(headlocked, ("%s: trimthenstep6: " "tcp_close.2: head not locked", __func__)); tp = tcp_close(tp); break; case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); break; } } goto drop; } /* * RFC 1323 PAWS: If we have a timestamp reply on this segment * and it's less than ts_recent, drop it. */ if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent && TSTMP_LT(to.to_tsval, tp->ts_recent)) { /* Check to see if ts_recent is over 24 days old. */ if ((int)(ticks - tp->ts_recent_age) > TCP_PAWS_IDLE) { /* * Invalidate ts_recent. If this segment updates * ts_recent, the age will be reset later and ts_recent * will get a valid value. If it does not, setting * ts_recent to zero will at least satisfy the * requirement that zero be placed in the timestamp * echo reply when ts_recent isn't valid. The * age isn't reset until we get a valid ts_recent * because we don't want out-of-order segments to be * dropped when ts_recent is old. */ tp->ts_recent = 0; } else { tcpstat.tcps_rcvduppack++; tcpstat.tcps_rcvdupbyte += tlen; tcpstat.tcps_pawsdrop++; if (tlen) goto dropafterack; goto drop; } } /* * In the SYN-RECEIVED state, validate that the packet belongs to * this connection before trimming the data to fit the receive * window. Check the sequence number versus IRS since we know * the sequence numbers haven't wrapped. This is a partial fix * for the "LAND" DoS attack. */ if (tp->t_state == TCPS_SYN_RECEIVED && SEQ_LT(th->th_seq, tp->irs)) { rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } todrop = tp->rcv_nxt - th->th_seq; if (todrop > 0) { if (thflags & TH_SYN) { thflags &= ~TH_SYN; th->th_seq++; if (th->th_urp > 1) th->th_urp--; else thflags &= ~TH_URG; todrop--; } /* * Following if statement from Stevens, vol. 2, p. 960. */ if (todrop > tlen || (todrop == tlen && (thflags & TH_FIN) == 0)) { /* * Any valid FIN must be to the left of the window. * At this point the FIN must be a duplicate or out * of sequence; drop it. */ thflags &= ~TH_FIN; /* * Send an ACK to resynchronize and drop any data. * But keep on processing for RST or ACK. */ tp->t_flags |= TF_ACKNOW; todrop = tlen; tcpstat.tcps_rcvduppack++; tcpstat.tcps_rcvdupbyte += todrop; } else { tcpstat.tcps_rcvpartduppack++; tcpstat.tcps_rcvpartdupbyte += todrop; } drop_hdrlen += todrop; /* drop from the top afterwards */ th->th_seq += todrop; tlen -= todrop; if (th->th_urp > todrop) th->th_urp -= todrop; else { thflags &= ~TH_URG; th->th_urp = 0; } } /* * If new data are received on a connection after the * user processes are gone, then RST the other end. */ if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT && tlen) { KASSERT(headlocked, ("%s: trimthenstep6: tcp_close.3: head " "not locked", __func__)); tp = tcp_close(tp); tcpstat.tcps_rcvafterclose++; rstreason = BANDLIM_UNLIMITED; goto dropwithreset; } /* * If segment ends after window, drop trailing data * (and PUSH and FIN); if nothing left, just ACK. */ todrop = (th->th_seq+tlen) - (tp->rcv_nxt+tp->rcv_wnd); if (todrop > 0) { tcpstat.tcps_rcvpackafterwin++; if (todrop >= tlen) { tcpstat.tcps_rcvbyteafterwin += tlen; /* * If a new connection request is received * while in TIME_WAIT, drop the old connection * and start over if the sequence numbers * are above the previous ones. */ KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); if (thflags & TH_SYN && tp->t_state == TCPS_TIME_WAIT && SEQ_GT(th->th_seq, tp->rcv_nxt)) { KASSERT(headlocked, ("%s: trimthenstep6: " "tcp_close.4: head not locked", __func__)); tp = tcp_close(tp); /* XXX: Shouldn't be possible. */ return (1); } /* * If window is closed can only take segments at * window edge, and have to drop data and PUSH from * incoming segments. Continue processing, but * remember to ack. Otherwise, drop segment * and ack. */ if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) { tp->t_flags |= TF_ACKNOW; tcpstat.tcps_rcvwinprobe++; } else goto dropafterack; } else tcpstat.tcps_rcvbyteafterwin += todrop; m_adj(m, -todrop); tlen -= todrop; thflags &= ~(TH_PUSH|TH_FIN); } /* * If last ACK falls within this segment's sequence numbers, * record its timestamp. * NOTE: * 1) That the test incorporates suggestions from the latest * proposal of the tcplw@cray.com list (Braden 1993/04/26). * 2) That updating only on newer timestamps interferes with * our earlier PAWS tests, so this check should be solely * predicated on the sequence space of this segment. * 3) That we modify the segment boundary check to be * Last.ACK.Sent <= SEG.SEQ + SEG.Len * instead of RFC1323's * Last.ACK.Sent < SEG.SEQ + SEG.Len, * This modified check allows us to overcome RFC1323's * limitations as described in Stevens TCP/IP Illustrated * Vol. 2 p.869. In such cases, we can still calculate the * RTT correctly when RCV.NXT == Last.ACK.Sent. */ if ((to.to_flags & TOF_TS) != 0 && SEQ_LEQ(th->th_seq, tp->last_ack_sent) && SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen + ((thflags & (TH_SYN|TH_FIN)) != 0))) { tp->ts_recent_age = ticks; tp->ts_recent = to.to_tsval; } /* * If a SYN is in the window, then this is an * error and we send an RST and drop the connection. */ if (thflags & TH_SYN) { KASSERT(headlocked, ("%s: tcp_drop: trimthenstep6: " "head not locked", __func__)); tp = tcp_drop(tp, ECONNRESET); rstreason = BANDLIM_UNLIMITED; goto drop; } /* * If the ACK bit is off: if in SYN-RECEIVED state or SENDSYN * flag is on (half-synchronized state), then queue data for * later processing; else drop segment and return. */ if ((thflags & TH_ACK) == 0) { if (tp->t_state == TCPS_SYN_RECEIVED || (tp->t_flags & TF_NEEDSYN)) goto step6; else if (tp->t_flags & TF_ACKNOW) goto dropafterack; else goto drop; } /* * Ack processing. */ switch (tp->t_state) { /* * In SYN_RECEIVED state, the ack ACKs our SYN, so enter * ESTABLISHED state and continue processing. * The ACK was checked above. */ case TCPS_SYN_RECEIVED: tcpstat.tcps_connects++; soisconnected(so); /* Do window scaling? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->rcv_scale = tp->request_r_scale; tp->snd_wnd = tiwin; } /* * Make transitions: * SYN-RECEIVED -> ESTABLISHED * SYN-RECEIVED* -> FIN-WAIT-1 */ tp->t_starttime = ticks; if (tp->t_flags & TF_NEEDFIN) { tp->t_state = TCPS_FIN_WAIT_1; tp->t_flags &= ~TF_NEEDFIN; } else { tp->t_state = TCPS_ESTABLISHED; tcp_timer_activate(tp, TT_KEEP, tcp_keepidle); } /* * If segment contains data or ACK, will call tcp_reass() * later; if not, do so now to pass queued data to user. */ if (tlen == 0 && (thflags & TH_FIN) == 0) (void) tcp_reass(tp, (struct tcphdr *)0, 0, (struct mbuf *)0); tp->snd_wl1 = th->th_seq - 1; /* FALLTHROUGH */ /* * In ESTABLISHED state: drop duplicate ACKs; ACK out of range * ACKs. If the ack is in the range * tp->snd_una < th->th_ack <= tp->snd_max * then advance tp->snd_una to th->th_ack and drop * data from the retransmission queue. If this ACK reflects * more up to date window information we update our window information. */ case TCPS_ESTABLISHED: case TCPS_FIN_WAIT_1: case TCPS_FIN_WAIT_2: case TCPS_CLOSE_WAIT: case TCPS_CLOSING: case TCPS_LAST_ACK: case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); if (SEQ_GT(th->th_ack, tp->snd_max)) { tcpstat.tcps_rcvacktoomuch++; goto dropafterack; } if (tp->sack_enable && ((to.to_flags & TOF_SACK) || !TAILQ_EMPTY(&tp->snd_holes))) tcp_sack_doack(tp, &to, th->th_ack); if (SEQ_LEQ(th->th_ack, tp->snd_una)) { if (tlen == 0 && tiwin == tp->snd_wnd) { tcpstat.tcps_rcvdupack++; /* * If we have outstanding data (other than * a window probe), this is a completely * duplicate ack (ie, window info didn't * change), the ack is the biggest we've * seen and we've seen exactly our rexmt * threshhold of them, assume a packet * has been dropped and retransmit it. * Kludge snd_nxt & the congestion * window so we send only this one * packet. * * We know we're losing at the current * window size so do congestion avoidance * (set ssthresh to half the current window * and pull our congestion window back to * the new ssthresh). * * Dup acks mean that packets have left the * network (they're now cached at the receiver) * so bump cwnd by the amount in the receiver * to keep a constant cwnd packets in the * network. */ if (!tcp_timer_active(tp, TT_REXMT) || th->th_ack != tp->snd_una) tp->t_dupacks = 0; else if (++tp->t_dupacks > tcprexmtthresh || ((tcp_do_newreno || tp->sack_enable) && IN_FASTRECOVERY(tp))) { if (tp->sack_enable && IN_FASTRECOVERY(tp)) { int awnd; /* * Compute the amount of data in flight first. * We can inject new data into the pipe iff * we have less than 1/2 the original window's * worth of data in flight. */ awnd = (tp->snd_nxt - tp->snd_fack) + tp->sackhint.sack_bytes_rexmit; if (awnd < tp->snd_ssthresh) { tp->snd_cwnd += tp->t_maxseg; if (tp->snd_cwnd > tp->snd_ssthresh) tp->snd_cwnd = tp->snd_ssthresh; } } else tp->snd_cwnd += tp->t_maxseg; (void) tcp_output(tp); goto drop; } else if (tp->t_dupacks == tcprexmtthresh) { tcp_seq onxt = tp->snd_nxt; u_int win; /* * If we're doing sack, check to * see if we're already in sack * recovery. If we're not doing sack, * check to see if we're in newreno * recovery. */ if (tp->sack_enable) { if (IN_FASTRECOVERY(tp)) { tp->t_dupacks = 0; break; } } else if (tcp_do_newreno) { if (SEQ_LEQ(th->th_ack, tp->snd_recover)) { tp->t_dupacks = 0; break; } } win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; if (win < 2) win = 2; tp->snd_ssthresh = win * tp->t_maxseg; ENTER_FASTRECOVERY(tp); tp->snd_recover = tp->snd_max; tcp_timer_activate(tp, TT_REXMT, 0); tp->t_rtttime = 0; if (tp->sack_enable) { tcpstat.tcps_sack_recovery_episode++; tp->sack_newdata = tp->snd_nxt; tp->snd_cwnd = tp->t_maxseg; (void) tcp_output(tp); goto drop; } tp->snd_nxt = th->th_ack; tp->snd_cwnd = tp->t_maxseg; (void) tcp_output(tp); KASSERT(tp->snd_limited <= 2, ("%s: tp->snd_limited too big", __func__)); tp->snd_cwnd = tp->snd_ssthresh + tp->t_maxseg * (tp->t_dupacks - tp->snd_limited); if (SEQ_GT(onxt, tp->snd_nxt)) tp->snd_nxt = onxt; goto drop; } else if (tcp_do_rfc3042) { u_long oldcwnd = tp->snd_cwnd; tcp_seq oldsndmax = tp->snd_max; u_int sent; KASSERT(tp->t_dupacks == 1 || tp->t_dupacks == 2, ("%s: dupacks not 1 or 2", __func__)); if (tp->t_dupacks == 1) tp->snd_limited = 0; tp->snd_cwnd = (tp->snd_nxt - tp->snd_una) + (tp->t_dupacks - tp->snd_limited) * tp->t_maxseg; (void) tcp_output(tp); sent = tp->snd_max - oldsndmax; if (sent > tp->t_maxseg) { KASSERT((tp->t_dupacks == 2 && tp->snd_limited == 0) || (sent == tp->t_maxseg + 1 && tp->t_flags & TF_SENTFIN), ("%s: sent too much", __func__)); tp->snd_limited = 2; } else if (sent > 0) ++tp->snd_limited; tp->snd_cwnd = oldcwnd; goto drop; } } else tp->t_dupacks = 0; break; } KASSERT(SEQ_GT(th->th_ack, tp->snd_una), ("%s: th_ack <= snd_una", __func__)); /* * If the congestion window was inflated to account * for the other side's cached packets, retract it. */ if (tcp_do_newreno || tp->sack_enable) { if (IN_FASTRECOVERY(tp)) { if (SEQ_LT(th->th_ack, tp->snd_recover)) { if (tp->sack_enable) tcp_sack_partialack(tp, th); else tcp_newreno_partial_ack(tp, th); } else { /* * Out of fast recovery. * Window inflation should have left us * with approximately snd_ssthresh * outstanding data. * But in case we would be inclined to * send a burst, better to do it via * the slow start mechanism. */ if (SEQ_GT(th->th_ack + tp->snd_ssthresh, tp->snd_max)) tp->snd_cwnd = tp->snd_max - th->th_ack + tp->t_maxseg; else tp->snd_cwnd = tp->snd_ssthresh; } } } else { if (tp->t_dupacks >= tcprexmtthresh && tp->snd_cwnd > tp->snd_ssthresh) tp->snd_cwnd = tp->snd_ssthresh; } tp->t_dupacks = 0; /* * If we reach this point, ACK is not a duplicate, * i.e., it ACKs something we sent. */ if (tp->t_flags & TF_NEEDSYN) { /* * T/TCP: Connection was half-synchronized, and our * SYN has been ACK'd (so connection is now fully * synchronized). Go to non-starred state, * increment snd_una for ACK of SYN, and check if * we can do window scaling. */ tp->t_flags &= ~TF_NEEDSYN; tp->snd_una++; /* Do window scaling? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->rcv_scale = tp->request_r_scale; /* Send window already scaled. */ } } process_ACK: KASSERT(headlocked, ("%s: process_ACK: head not locked", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); acked = th->th_ack - tp->snd_una; tcpstat.tcps_rcvackpack++; tcpstat.tcps_rcvackbyte += acked; /* * If we just performed our first retransmit, and the ACK * arrives within our recovery window, then it was a mistake * to do the retransmit in the first place. Recover our * original cwnd and ssthresh, and proceed to transmit where * we left off. */ if (tp->t_rxtshift == 1 && ticks < tp->t_badrxtwin) { ++tcpstat.tcps_sndrexmitbad; tp->snd_cwnd = tp->snd_cwnd_prev; tp->snd_ssthresh = tp->snd_ssthresh_prev; tp->snd_recover = tp->snd_recover_prev; if (tp->t_flags & TF_WASFRECOVERY) ENTER_FASTRECOVERY(tp); tp->snd_nxt = tp->snd_max; tp->t_badrxtwin = 0; /* XXX probably not required */ } /* * If we have a timestamp reply, update smoothed * round trip time. If no timestamp is present but * transmit timer is running and timed sequence * number was acked, update smoothed round trip time. * Since we now have an rtt measurement, cancel the * timer backoff (cf., Phil Karn's retransmit alg.). * Recompute the initial retransmit timer. * * Some boxes send broken timestamp replies * during the SYN+ACK phase, ignore * timestamps of 0 or we could calculate a * huge RTT and blow up the retransmit timer. */ if ((to.to_flags & TOF_TS) != 0 && to.to_tsecr) { if (!tp->t_rttlow || tp->t_rttlow > ticks - to.to_tsecr) tp->t_rttlow = ticks - to.to_tsecr; tcp_xmit_timer(tp, ticks - to.to_tsecr + 1); } else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) { if (!tp->t_rttlow || tp->t_rttlow > ticks - tp->t_rtttime) tp->t_rttlow = ticks - tp->t_rtttime; tcp_xmit_timer(tp, ticks - tp->t_rtttime); } tcp_xmit_bandwidth_limit(tp, th->th_ack); /* * If all outstanding data is acked, stop retransmit * timer and remember to restart (more output or persist). * If there is more data to be acked, restart retransmit * timer, using current (possibly backed-off) value. */ if (th->th_ack == tp->snd_max) { tcp_timer_activate(tp, TT_REXMT, 0); needoutput = 1; } else if (!tcp_timer_active(tp, TT_PERSIST)) tcp_timer_activate(tp, TT_REXMT, tp->t_rxtcur); /* * If no data (only SYN) was ACK'd, * skip rest of ACK processing. */ if (acked == 0) goto step6; /* * When new data is acked, open the congestion window. * If the window gives us less than ssthresh packets * in flight, open exponentially (maxseg per packet). * Otherwise open linearly: maxseg per window * (maxseg^2 / cwnd per packet). */ if ((!tcp_do_newreno && !tp->sack_enable) || !IN_FASTRECOVERY(tp)) { u_int cw = tp->snd_cwnd; u_int incr = tp->t_maxseg; if (cw > tp->snd_ssthresh) incr = incr * incr / cw; tp->snd_cwnd = min(cw+incr, TCP_MAXWIN<snd_scale); } SOCKBUF_LOCK(&so->so_snd); if (acked > so->so_snd.sb_cc) { tp->snd_wnd -= so->so_snd.sb_cc; sbdrop_locked(&so->so_snd, (int)so->so_snd.sb_cc); ourfinisacked = 1; } else { sbdrop_locked(&so->so_snd, acked); tp->snd_wnd -= acked; ourfinisacked = 0; } sowwakeup_locked(so); /* detect una wraparound */ if ((tcp_do_newreno || tp->sack_enable) && !IN_FASTRECOVERY(tp) && SEQ_GT(tp->snd_una, tp->snd_recover) && SEQ_LEQ(th->th_ack, tp->snd_recover)) tp->snd_recover = th->th_ack - 1; if ((tcp_do_newreno || tp->sack_enable) && IN_FASTRECOVERY(tp) && SEQ_GEQ(th->th_ack, tp->snd_recover)) EXIT_FASTRECOVERY(tp); tp->snd_una = th->th_ack; if (tp->sack_enable) { if (SEQ_GT(tp->snd_una, tp->snd_recover)) tp->snd_recover = tp->snd_una; } if (SEQ_LT(tp->snd_nxt, tp->snd_una)) tp->snd_nxt = tp->snd_una; switch (tp->t_state) { /* * In FIN_WAIT_1 STATE in addition to the processing * for the ESTABLISHED state if our FIN is now acknowledged * then enter FIN_WAIT_2. */ case TCPS_FIN_WAIT_1: if (ourfinisacked) { /* * If we can't receive any more * data, then closing user can proceed. * Starting the timer is contrary to the * specification, but if we don't get a FIN * we'll hang forever. */ /* XXXjl * we should release the tp also, and use a * compressed state. */ if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { int timeout; soisdisconnected(so); timeout = (tcp_fast_finwait2_recycle) ? tcp_finwait2_timeout : tcp_maxidle; tcp_timer_activate(tp, TT_2MSL, timeout); } tp->t_state = TCPS_FIN_WAIT_2; } break; /* * In CLOSING STATE in addition to the processing for * the ESTABLISHED state if the ACK acknowledges our FIN * then enter the TIME-WAIT state, otherwise ignore * the segment. */ case TCPS_CLOSING: if (ourfinisacked) { KASSERT(headlocked, ("%s: process_ACK: " "head not locked", __func__)); tcp_twstart(tp); INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; m_freem(m); return (0); } break; /* * In LAST_ACK, we may still be waiting for data to drain * and/or to be acked, as well as for the ack of our FIN. * If our FIN is now acknowledged, delete the TCB, * enter the closed state and return. */ case TCPS_LAST_ACK: if (ourfinisacked) { KASSERT(headlocked, ("%s: process_ACK: " "tcp_close: head not locked", __func__)); tp = tcp_close(tp); goto drop; } break; /* * In TIME_WAIT state the only thing that should arrive * is a retransmission of the remote FIN. Acknowledge * it and restart the finack timer. */ case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); tcp_timer_activate(tp, TT_2MSL, 2 * tcp_msl); goto dropafterack; } } step6: KASSERT(headlocked, ("%s: step6: head not locked", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); /* * Update window information. * Don't look at window if no ACK: TAC's send garbage on first SYN. */ if ((thflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, th->th_seq) || (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) || (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) { /* keep track of pure window updates */ if (tlen == 0 && tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) tcpstat.tcps_rcvwinupd++; tp->snd_wnd = tiwin; tp->snd_wl1 = th->th_seq; tp->snd_wl2 = th->th_ack; if (tp->snd_wnd > tp->max_sndwnd) tp->max_sndwnd = tp->snd_wnd; needoutput = 1; } /* * Process segments with URG. */ if ((thflags & TH_URG) && th->th_urp && TCPS_HAVERCVDFIN(tp->t_state) == 0) { /* * This is a kludge, but if we receive and accept * random urgent pointers, we'll crash in * soreceive. It's hard to imagine someone * actually wanting to send this much urgent data. */ SOCKBUF_LOCK(&so->so_rcv); if (th->th_urp + so->so_rcv.sb_cc > sb_max) { th->th_urp = 0; /* XXX */ thflags &= ~TH_URG; /* XXX */ SOCKBUF_UNLOCK(&so->so_rcv); /* XXX */ goto dodata; /* XXX */ } /* * If this segment advances the known urgent pointer, * then mark the data stream. This should not happen * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since * a FIN has been received from the remote side. * In these states we ignore the URG. * * 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 as the original * spec states (in one of two places). */ if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) { tp->rcv_up = th->th_seq + th->th_urp; so->so_oobmark = so->so_rcv.sb_cc + (tp->rcv_up - tp->rcv_nxt) - 1; if (so->so_oobmark == 0) so->so_rcv.sb_state |= SBS_RCVATMARK; sohasoutofband(so); tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); } SOCKBUF_UNLOCK(&so->so_rcv); /* * Remove out of band data so doesn't get presented to user. * This can happen independent of advancing the URG pointer, * but if two URG's are pending at once, some out-of-band * data may creep in... ick. */ if (th->th_urp <= (u_long)tlen && !(so->so_options & SO_OOBINLINE)) { /* hdr drop is delayed */ tcp_pulloutofband(so, th, m, drop_hdrlen); } } else { /* * If no out of band data is expected, * pull receive urgent pointer along * with the receive window. */ if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) tp->rcv_up = tp->rcv_nxt; } dodata: /* XXX */ KASSERT(headlocked, ("%s: dodata: head not locked", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); /* * Process the segment text, merging it into the TCP sequencing queue, * and arranging for acknowledgment of receipt if necessary. * This process logically involves adjusting tp->rcv_wnd as data * is presented to the user (this happens in tcp_usrreq.c, * case PRU_RCVD). If a FIN has already been received on this * connection then we just ignore the text. */ if ((tlen || (thflags & TH_FIN)) && TCPS_HAVERCVDFIN(tp->t_state) == 0) { tcp_seq save_start = th->th_seq; tcp_seq save_end = th->th_seq + tlen; m_adj(m, drop_hdrlen); /* delayed header drop */ /* * Insert segment which includes th into TCP reassembly queue * with control block tp. Set thflags to whether reassembly now * includes a segment with FIN. This handles the common case * inline (segment is the next to be received on an established * connection, and the queue is empty), avoiding linkage into * and removal from the queue and repetition of various * conversions. * Set DELACK for segments received in order, but ack * immediately when segments are out of order (so * fast retransmit can work). */ if (th->th_seq == tp->rcv_nxt && LIST_EMPTY(&tp->t_segq) && TCPS_HAVEESTABLISHED(tp->t_state)) { if (DELAY_ACK(tp)) tp->t_flags |= TF_DELACK; else tp->t_flags |= TF_ACKNOW; tp->rcv_nxt += tlen; thflags = th->th_flags & TH_FIN; tcpstat.tcps_rcvpack++; tcpstat.tcps_rcvbyte += tlen; ND6_HINT(tp); SOCKBUF_LOCK(&so->so_rcv); if (so->so_rcv.sb_state & SBS_CANTRCVMORE) m_freem(m); else sbappendstream_locked(&so->so_rcv, m); sorwakeup_locked(so); } else { thflags = tcp_reass(tp, th, &tlen, m); tp->t_flags |= TF_ACKNOW; } if (tlen > 0 && tp->sack_enable) tcp_update_sack_list(tp, save_start, save_end); #if 0 /* * Note the amount of data that peer has sent into * our window, in order to estimate the sender's * buffer size. * XXX: Unused. */ len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); #endif } else { m_freem(m); thflags &= ~TH_FIN; } /* * If FIN is received ACK the FIN and let the user know * that the connection is closing. */ if (thflags & TH_FIN) { if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { socantrcvmore(so); /* * If connection is half-synchronized * (ie NEEDSYN flag on) then delay ACK, * so it may be piggybacked when SYN is sent. * Otherwise, since we received a FIN then no * more input can be expected, send ACK now. */ if (tp->t_flags & TF_NEEDSYN) tp->t_flags |= TF_DELACK; else tp->t_flags |= TF_ACKNOW; tp->rcv_nxt++; } switch (tp->t_state) { /* * In SYN_RECEIVED and ESTABLISHED STATES * enter the CLOSE_WAIT state. */ case TCPS_SYN_RECEIVED: tp->t_starttime = ticks; /*FALLTHROUGH*/ case TCPS_ESTABLISHED: tp->t_state = TCPS_CLOSE_WAIT; break; /* * If still in FIN_WAIT_1 STATE FIN has not been acked so * enter the CLOSING state. */ case TCPS_FIN_WAIT_1: tp->t_state = TCPS_CLOSING; break; /* * In FIN_WAIT_2 state enter the TIME_WAIT state, * starting the time-wait timer, turning off the other * standard timers. */ case TCPS_FIN_WAIT_2: KASSERT(headlocked == 1, ("%s: dodata: " "TCP_FIN_WAIT_2: head not locked", __func__)); tcp_twstart(tp); INP_INFO_WUNLOCK(&tcbinfo); return (0); /* * In TIME_WAIT state restart the 2 MSL time_wait timer. */ case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); tcp_timer_activate(tp, TT_2MSL, 2 * tcp_msl); break; } } INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif /* * Return any desired output. */ if (needoutput || (tp->t_flags & TF_ACKNOW)) (void) tcp_output(tp); check_delack: KASSERT(headlocked == 0, ("%s: check_delack: head locked", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); if (tp->t_flags & TF_DELACK) { tp->t_flags &= ~TF_DELACK; tcp_timer_activate(tp, TT_DELACK, tcp_delacktime); } INP_UNLOCK(tp->t_inpcb); return (0); dropafterack: KASSERT(headlocked, ("%s: dropafterack: head not locked", __func__)); /* * Generate an ACK dropping incoming segment if it occupies * sequence space, where the ACK reflects our state. * * We can now skip the test for the RST flag since all * paths to this code happen after packets containing * RST have been dropped. * * In the SYN-RECEIVED state, don't send an ACK unless the * segment we received passes the SYN-RECEIVED ACK test. * If it fails send a RST. This breaks the loop in the * "LAND" DoS attack, and also prevents an ACK storm * between two listening ports that have been sent forged * SYN segments, each with the source address of the other. */ if (tp->t_state == TCPS_SYN_RECEIVED && (thflags & TH_ACK) && (SEQ_GT(tp->snd_una, th->th_ack) || SEQ_GT(th->th_ack, tp->snd_max)) ) { rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif KASSERT(headlocked, ("%s: headlocked should be 1", __func__)); INP_INFO_WUNLOCK(&tcbinfo); tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); INP_UNLOCK(tp->t_inpcb); m_freem(m); return (0); dropwithreset: KASSERT(headlocked, ("%s: dropwithreset: head not locked", __func__)); tcp_dropwithreset(m, th, tp, tlen, rstreason); if (tp != NULL) INP_UNLOCK(tp->t_inpcb); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); return (0); drop: /* * Drop space held by incoming segment and return. */ #ifdef TCPDEBUG if (tp == NULL || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif if (tp != NULL) INP_UNLOCK(tp->t_inpcb); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); m_freem(m); return (0); } /* * Issue RST on TCP segment. The mbuf must still include the original * packet header. */ static void tcp_dropwithreset(struct mbuf *m, struct tcphdr *th, struct tcpcb *tp, int tlen, int rstreason) { struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; #endif /* * Generate a RST, dropping incoming segment. * Make ACK acceptable to originator of segment. * Don't bother to respond if destination was broadcast/multicast. * tp may be NULL. */ if ((th->th_flags & TH_RST) || m->m_flags & (M_BCAST|M_MCAST)) goto drop; #ifdef INET6 if (mtod(m, struct ip *)->ip_v == 6) { ip6 = mtod(m, struct ip6_hdr *); if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto drop; /* IPv6 anycast check is done at tcp6_input() */ } else #endif { ip = mtod(m, struct ip *); if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || ip->ip_src.s_addr == htonl(INADDR_BROADCAST) || in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) goto drop; } /* Perform bandwidth limiting. */ if (badport_bandlim(rstreason) < 0) goto drop; /* tcp_respond consumes the mbuf chain. */ if (th->th_flags & TH_ACK) { tcp_respond(tp, mtod(m, void *), th, m, (tcp_seq)0, th->th_ack, TH_RST); } else { if (th->th_flags & TH_SYN) tlen++; tcp_respond(tp, mtod(m, void *), th, m, th->th_seq+tlen, (tcp_seq)0, TH_RST|TH_ACK); } return; drop: m_freem(m); return; } /* * Parse TCP options and place in tcpopt. */ static void tcp_dooptions(struct tcpopt *to, u_char *cp, int cnt, int flags) { int opt, optlen; to->to_flags = 0; for (; cnt > 0; cnt -= optlen, cp += optlen) { opt = cp[0]; if (opt == TCPOPT_EOL) break; if (opt == TCPOPT_NOP) optlen = 1; else { if (cnt < 2) break; optlen = cp[1]; if (optlen < 2 || optlen > cnt) break; } switch (opt) { case TCPOPT_MAXSEG: if (optlen != TCPOLEN_MAXSEG) continue; if (!(flags & TO_SYN)) continue; to->to_flags |= TOF_MSS; bcopy((char *)cp + 2, (char *)&to->to_mss, sizeof(to->to_mss)); to->to_mss = ntohs(to->to_mss); break; case TCPOPT_WINDOW: if (optlen != TCPOLEN_WINDOW) continue; if (!(flags & TO_SYN)) continue; to->to_flags |= TOF_SCALE; to->to_wscale = min(cp[2], TCP_MAX_WINSHIFT); break; case TCPOPT_TIMESTAMP: if (optlen != TCPOLEN_TIMESTAMP) continue; to->to_flags |= TOF_TS; bcopy((char *)cp + 2, (char *)&to->to_tsval, sizeof(to->to_tsval)); to->to_tsval = ntohl(to->to_tsval); bcopy((char *)cp + 6, (char *)&to->to_tsecr, sizeof(to->to_tsecr)); to->to_tsecr = ntohl(to->to_tsecr); break; #ifdef TCP_SIGNATURE /* * XXX In order to reply to a host which has set the * TCP_SIGNATURE option in its initial SYN, we have to * record the fact that the option was observed here * for the syncache code to perform the correct response. */ case TCPOPT_SIGNATURE: if (optlen != TCPOLEN_SIGNATURE) continue; to->to_flags |= (TOF_SIGNATURE | TOF_SIGLEN); break; #endif case TCPOPT_SACK_PERMITTED: if (optlen != TCPOLEN_SACK_PERMITTED) continue; if (!(flags & TO_SYN)) continue; if (!tcp_do_sack) continue; to->to_flags |= TOF_SACKPERM; break; case TCPOPT_SACK: if (optlen <= 2 || (optlen - 2) % TCPOLEN_SACK != 0) continue; if (flags & TO_SYN) continue; to->to_flags |= TOF_SACK; to->to_nsacks = (optlen - 2) / TCPOLEN_SACK; to->to_sacks = cp + 2; tcpstat.tcps_sack_rcv_blocks++; break; default: continue; } } } /* * Pull out of band byte out of a segment so * it doesn't appear in the user's data queue. * It is still reflected in the segment length for * sequencing purposes. */ static void tcp_pulloutofband(struct socket *so, struct tcphdr *th, struct mbuf *m, int off) { int cnt = off + th->th_urp - 1; while (cnt >= 0) { if (m->m_len > cnt) { char *cp = mtod(m, caddr_t) + cnt; struct tcpcb *tp = sototcpcb(so); tp->t_iobc = *cp; tp->t_oobflags |= TCPOOB_HAVEDATA; bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1)); m->m_len--; if (m->m_flags & M_PKTHDR) m->m_pkthdr.len--; return; } cnt -= m->m_len; m = m->m_next; if (m == NULL) break; } panic("tcp_pulloutofband"); } /* * Collect new round-trip time estimate * and update averages and current timeout. */ static void tcp_xmit_timer(struct tcpcb *tp, int rtt) { int delta; INP_LOCK_ASSERT(tp->t_inpcb); tcpstat.tcps_rttupdated++; tp->t_rttupdated++; if (tp->t_srtt != 0) { /* * srtt is stored as fixed point with 5 bits after the * binary point (i.e., scaled by 8). The following magic * is equivalent to the smoothing algorithm in rfc793 with * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed * point). Adjust rtt to origin 0. */ delta = ((rtt - 1) << TCP_DELTA_SHIFT) - (tp->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)); if ((tp->t_srtt += delta) <= 0) tp->t_srtt = 1; /* * We accumulate a smoothed rtt variance (actually, a * smoothed mean difference), then set the retransmit * timer to smoothed rtt + 4 times the smoothed variance. * rttvar is stored as fixed point with 4 bits after the * binary point (scaled by 16). The following is * equivalent to rfc793 smoothing with an alpha of .75 * (rttvar = rttvar*3/4 + |delta| / 4). This replaces * rfc793's wired-in beta. */ if (delta < 0) delta = -delta; delta -= tp->t_rttvar >> (TCP_RTTVAR_SHIFT - TCP_DELTA_SHIFT); if ((tp->t_rttvar += delta) <= 0) tp->t_rttvar = 1; if (tp->t_rttbest > tp->t_srtt + tp->t_rttvar) tp->t_rttbest = tp->t_srtt + tp->t_rttvar; } else { /* * No rtt measurement yet - use the unsmoothed rtt. * Set the variance to half the rtt (so our first * retransmit happens at 3*rtt). */ tp->t_srtt = rtt << TCP_RTT_SHIFT; tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1); tp->t_rttbest = tp->t_srtt + tp->t_rttvar; } tp->t_rtttime = 0; tp->t_rxtshift = 0; /* * the retransmit should happen at rtt + 4 * rttvar. * Because of the way we do the smoothing, srtt and rttvar * will each average +1/2 tick of bias. When we compute * the retransmit timer, we want 1/2 tick of rounding and * 1 extra tick because of +-1/2 tick uncertainty in the * firing of the timer. The bias will give us exactly the * 1.5 tick we need. But, because the bias is * statistical, we have to test that we don't drop below * the minimum feasible timer (which is 2 ticks). */ TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), max(tp->t_rttmin, rtt + 2), TCPTV_REXMTMAX); /* * We received an ack for a packet that wasn't retransmitted; * it is probably safe to discard any error indications we've * received recently. This isn't quite right, but close enough * for now (a route might have failed after we sent a segment, * and the return path might not be symmetrical). */ tp->t_softerror = 0; } /* * Determine a reasonable value for maxseg size. * If the route is known, check route for mtu. * If none, use an mss that can be handled on the outgoing * interface without forcing IP to fragment; if bigger than * an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES * to utilize large mbufs. If no route is found, route has no mtu, * or the destination isn't local, use a default, hopefully conservative * size (usually 512 or the default IP max size, but no more than the mtu * of the interface), as we can't discover anything about intervening * gateways or networks. We also initialize the congestion/slow start * window to be a single segment if the destination isn't local. * While looking at the routing entry, we also initialize other path-dependent * parameters from pre-set or cached values in the routing entry. * * Also take into account the space needed for options that we * send regularly. Make maxseg shorter by that amount to assure * that we can send maxseg amount of data even when the options * are present. Store the upper limit of the length of options plus * data in maxopd. * * * In case of T/TCP, we call this routine during implicit connection * setup as well (offer = -1), to initialize maxseg from the cached * MSS of our peer. * * NOTE that this routine is only called when we process an incoming * segment. Outgoing SYN/ACK MSS settings are handled in tcp_mssopt(). */ void tcp_mss(struct tcpcb *tp, int offer) { int rtt, mss; u_long bufsize; u_long maxmtu; struct inpcb *inp = tp->t_inpcb; struct socket *so; struct hc_metrics_lite metrics; int origoffer = offer; int mtuflags = 0; #ifdef INET6 int isipv6 = ((inp->inp_vflag & INP_IPV6) != 0) ? 1 : 0; size_t min_protoh = isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : sizeof (struct tcpiphdr); #else const size_t min_protoh = sizeof(struct tcpiphdr); #endif /* initialize */ #ifdef INET6 if (isipv6) { maxmtu = tcp_maxmtu6(&inp->inp_inc, &mtuflags); tp->t_maxopd = tp->t_maxseg = tcp_v6mssdflt; } else #endif { maxmtu = tcp_maxmtu(&inp->inp_inc, &mtuflags); tp->t_maxopd = tp->t_maxseg = tcp_mssdflt; } so = inp->inp_socket; /* * no route to sender, stay with default mss and return */ if (maxmtu == 0) return; /* what have we got? */ switch (offer) { case 0: /* * Offer == 0 means that there was no MSS on the SYN * segment, in this case we use tcp_mssdflt. */ offer = #ifdef INET6 isipv6 ? tcp_v6mssdflt : #endif tcp_mssdflt; break; case -1: /* * Offer == -1 means that we didn't receive SYN yet. */ /* FALLTHROUGH */ default: /* * Prevent DoS attack with too small MSS. Round up * to at least minmss. */ offer = max(offer, tcp_minmss); /* * Sanity check: make sure that maxopd will be large * enough to allow some data on segments even if the * all the option space is used (40bytes). Otherwise * funny things may happen in tcp_output. */ offer = max(offer, 64); } /* * rmx information is now retrieved from tcp_hostcache */ tcp_hc_get(&inp->inp_inc, &metrics); /* * if there's a discovered mtu int tcp hostcache, use it * else, use the link mtu. */ if (metrics.rmx_mtu) mss = min(metrics.rmx_mtu, maxmtu) - min_protoh; else { #ifdef INET6 if (isipv6) { mss = maxmtu - min_protoh; if (!path_mtu_discovery && !in6_localaddr(&inp->in6p_faddr)) mss = min(mss, tcp_v6mssdflt); } else #endif { mss = maxmtu - min_protoh; if (!path_mtu_discovery && !in_localaddr(inp->inp_faddr)) mss = min(mss, tcp_mssdflt); } } mss = min(mss, offer); /* * maxopd stores the maximum length of data AND options * in a segment; maxseg is the amount of data in a normal * segment. We need to store this value (maxopd) apart * from maxseg, because now every segment carries options * and thus we normally have somewhat less data in segments. */ tp->t_maxopd = mss; /* * origoffer==-1 indicates, that no segments were received yet. * In this case we just guess. */ if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && (origoffer == -1 || (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP)) mss -= TCPOLEN_TSTAMP_APPA; tp->t_maxseg = mss; #if (MCLBYTES & (MCLBYTES - 1)) == 0 if (mss > MCLBYTES) mss &= ~(MCLBYTES-1); #else if (mss > MCLBYTES) mss = mss / MCLBYTES * MCLBYTES; #endif tp->t_maxseg = mss; /* * If there's a pipesize, change the socket buffer to that size, * don't change if sb_hiwat is different than default (then it * has been changed on purpose with setsockopt). * Make the socket buffers an integral number of mss units; * if the mss is larger than the socket buffer, decrease the mss. */ SOCKBUF_LOCK(&so->so_snd); if ((so->so_snd.sb_hiwat == tcp_sendspace) && metrics.rmx_sendpipe) bufsize = metrics.rmx_sendpipe; else bufsize = so->so_snd.sb_hiwat; if (bufsize < mss) mss = bufsize; else { bufsize = roundup(bufsize, mss); if (bufsize > sb_max) bufsize = sb_max; if (bufsize > so->so_snd.sb_hiwat) (void)sbreserve_locked(&so->so_snd, bufsize, so, NULL); } SOCKBUF_UNLOCK(&so->so_snd); tp->t_maxseg = mss; SOCKBUF_LOCK(&so->so_rcv); if ((so->so_rcv.sb_hiwat == tcp_recvspace) && metrics.rmx_recvpipe) bufsize = metrics.rmx_recvpipe; else bufsize = so->so_rcv.sb_hiwat; if (bufsize > mss) { bufsize = roundup(bufsize, mss); if (bufsize > sb_max) bufsize = sb_max; if (bufsize > so->so_rcv.sb_hiwat) (void)sbreserve_locked(&so->so_rcv, bufsize, so, NULL); } SOCKBUF_UNLOCK(&so->so_rcv); /* * While we're here, check the others too */ if (tp->t_srtt == 0 && (rtt = metrics.rmx_rtt)) { tp->t_srtt = rtt; tp->t_rttbest = tp->t_srtt + TCP_RTT_SCALE; tcpstat.tcps_usedrtt++; if (metrics.rmx_rttvar) { tp->t_rttvar = metrics.rmx_rttvar; tcpstat.tcps_usedrttvar++; } else { /* default variation is +- 1 rtt */ tp->t_rttvar = tp->t_srtt * TCP_RTTVAR_SCALE / TCP_RTT_SCALE; } TCPT_RANGESET(tp->t_rxtcur, ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1, tp->t_rttmin, TCPTV_REXMTMAX); } if (metrics.rmx_ssthresh) { /* * There's some sort of gateway or interface * buffer limit on the path. Use this to set * the slow start threshhold, but set the * threshold to no less than 2*mss. */ tp->snd_ssthresh = max(2 * mss, metrics.rmx_ssthresh); tcpstat.tcps_usedssthresh++; } if (metrics.rmx_bandwidth) tp->snd_bandwidth = metrics.rmx_bandwidth; /* * Set the slow-start flight size depending on whether this * is a local network or not. * * Extend this so we cache the cwnd too and retrieve it here. * Make cwnd even bigger than RFC3390 suggests but only if we * have previous experience with the remote host. Be careful * not make cwnd bigger than remote receive window or our own * send socket buffer. Maybe put some additional upper bound * on the retrieved cwnd. Should do incremental updates to * hostcache when cwnd collapses so next connection doesn't * overloads the path again. * * RFC3390 says only do this if SYN or SYN/ACK didn't got lost. * We currently check only in syncache_socket for that. */ #define TCP_METRICS_CWND #ifdef TCP_METRICS_CWND if (metrics.rmx_cwnd) tp->snd_cwnd = max(mss, min(metrics.rmx_cwnd / 2, min(tp->snd_wnd, so->so_snd.sb_hiwat))); else #endif if (tcp_do_rfc3390) tp->snd_cwnd = min(4 * mss, max(2 * mss, 4380)); #ifdef INET6 else if ((isipv6 && in6_localaddr(&inp->in6p_faddr)) || (!isipv6 && in_localaddr(inp->inp_faddr))) #else else if (in_localaddr(inp->inp_faddr)) #endif tp->snd_cwnd = mss * ss_fltsz_local; else tp->snd_cwnd = mss * ss_fltsz; /* Check the interface for TSO capabilities. */ if (mtuflags & CSUM_TSO) tp->t_flags |= TF_TSO; } /* * Determine the MSS option to send on an outgoing SYN. */ int tcp_mssopt(struct in_conninfo *inc) { int mss = 0; u_long maxmtu = 0; u_long thcmtu = 0; size_t min_protoh; #ifdef INET6 int isipv6 = inc->inc_isipv6 ? 1 : 0; #endif KASSERT(inc != NULL, ("tcp_mssopt with NULL in_conninfo pointer")); #ifdef INET6 if (isipv6) { mss = tcp_v6mssdflt; maxmtu = tcp_maxmtu6(inc, NULL); thcmtu = tcp_hc_getmtu(inc); /* IPv4 and IPv6 */ min_protoh = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); } else #endif { mss = tcp_mssdflt; maxmtu = tcp_maxmtu(inc, NULL); thcmtu = tcp_hc_getmtu(inc); /* IPv4 and IPv6 */ min_protoh = sizeof(struct tcpiphdr); } if (maxmtu && thcmtu) mss = min(maxmtu, thcmtu) - min_protoh; else if (maxmtu || thcmtu) mss = max(maxmtu, thcmtu) - min_protoh; return (mss); } /* * On a partial ack arrives, force the retransmission of the * next unacknowledged segment. Do not clear tp->t_dupacks. * By setting snd_nxt to ti_ack, this forces retransmission timer to * be started again. */ static void tcp_newreno_partial_ack(struct tcpcb *tp, struct tcphdr *th) { tcp_seq onxt = tp->snd_nxt; u_long ocwnd = tp->snd_cwnd; tcp_timer_activate(tp, TT_REXMT, 0); tp->t_rtttime = 0; tp->snd_nxt = th->th_ack; /* * Set snd_cwnd to one segment beyond acknowledged offset. * (tp->snd_una has not yet been updated when this function is called.) */ tp->snd_cwnd = tp->t_maxseg + (th->th_ack - tp->snd_una); tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); tp->snd_cwnd = ocwnd; if (SEQ_GT(onxt, tp->snd_nxt)) tp->snd_nxt = onxt; /* * Partial window deflation. Relies on fact that tp->snd_una * not updated yet. */ if (tp->snd_cwnd > th->th_ack - tp->snd_una) tp->snd_cwnd -= th->th_ack - tp->snd_una; else tp->snd_cwnd = 0; tp->snd_cwnd += tp->t_maxseg; } /* * Returns 1 if the TIME_WAIT state was killed and we should start over, * looking for a pcb in the listen state. Returns 0 otherwise. */ static int tcp_timewait(struct inpcb *inp, struct tcpopt *to, struct tcphdr *th, struct mbuf *m, int tlen) { struct tcptw *tw; int thflags; tcp_seq seq; #ifdef INET6 int isipv6 = (mtod(m, struct ip *)->ip_v == 6) ? 1 : 0; #else const int isipv6 = 0; #endif /* tcbinfo lock required for tcp_twclose(), tcp_timer_2msl_reset(). */ INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); /* * XXXRW: Time wait state for inpcb has been recycled, but inpcb is * still present. This is undesirable, but temporarily necessary * until we work out how to handle inpcb's who's timewait state has * been removed. */ tw = intotw(inp); if (tw == NULL) goto drop; thflags = th->th_flags; /* * NOTE: for FIN_WAIT_2 (to be added later), * must validate sequence number before accepting RST */ /* * If the segment contains RST: * Drop the segment - see Stevens, vol. 2, p. 964 and * RFC 1337. */ if (thflags & TH_RST) goto drop; #if 0 /* PAWS not needed at the moment */ /* * RFC 1323 PAWS: If we have a timestamp reply on this segment * and it's less than ts_recent, drop it. */ if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent && TSTMP_LT(to.to_tsval, tp->ts_recent)) { if ((thflags & TH_ACK) == 0) goto drop; goto ack; } /* * ts_recent is never updated because we never accept new segments. */ #endif /* * If a new connection request is received * while in TIME_WAIT, drop the old connection * and start over if the sequence numbers * are above the previous ones. */ if ((thflags & TH_SYN) && SEQ_GT(th->th_seq, tw->rcv_nxt)) { tcp_twclose(tw, 0); return (1); } /* * Drop the the segment if it does not contain an ACK. */ if ((thflags & TH_ACK) == 0) goto drop; /* * Reset the 2MSL timer if this is a duplicate FIN. */ if (thflags & TH_FIN) { seq = th->th_seq + tlen + (thflags & TH_SYN ? 1 : 0); if (seq + 1 == tw->rcv_nxt) tcp_timer_2msl_reset(tw, 1); } /* * Acknowledge the segment if it has data or is not a duplicate ACK. */ if (thflags != TH_ACK || tlen != 0 || th->th_seq != tw->rcv_nxt || th->th_ack != tw->snd_nxt) tcp_twrespond(tw, TH_ACK); goto drop; /* * Generate a RST, dropping incoming segment. * Make ACK acceptable to originator of segment. * Don't bother to respond if destination was broadcast/multicast. */ if (m->m_flags & (M_BCAST|M_MCAST)) goto drop; if (isipv6) { struct ip6_hdr *ip6; /* IPv6 anycast check is done at tcp6_input() */ ip6 = mtod(m, struct ip6_hdr *); if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto drop; } else { struct ip *ip; ip = mtod(m, struct ip *); if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || ip->ip_src.s_addr == htonl(INADDR_BROADCAST) || in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) goto drop; } if (thflags & TH_ACK) { tcp_respond(NULL, mtod(m, void *), th, m, 0, th->th_ack, TH_RST); } else { seq = th->th_seq + (thflags & TH_SYN ? 1 : 0); tcp_respond(NULL, mtod(m, void *), th, m, seq, 0, TH_RST|TH_ACK); } INP_UNLOCK(inp); return (0); drop: INP_UNLOCK(inp); m_freem(m); return (0); } Index: head/sys/netinet/tcp_reass.c =================================================================== --- head/sys/netinet/tcp_reass.c (revision 168902) +++ head/sys/netinet/tcp_reass.c (revision 168903) @@ -1,3345 +1,3339 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 * $FreeBSD$ */ #include "opt_ipfw.h" /* for ipfw_fwd */ #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 /* for proc0 declaration */ #include #include #include #include #include #include #include #include /* before tcp_seq.h, for tcp_random18() */ #include #include #include #include #include #include #include #include #include /* required for icmp_var.h */ #include /* for ICMP_BANDLIM */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef TCPDEBUG #include #endif /* TCPDEBUG */ #ifdef FAST_IPSEC #include #include #endif /*FAST_IPSEC*/ #ifdef IPSEC #include #include #include #endif /*IPSEC*/ #include #include static const int tcprexmtthresh = 3; struct tcpstat tcpstat; SYSCTL_STRUCT(_net_inet_tcp, TCPCTL_STATS, stats, CTLFLAG_RW, &tcpstat , tcpstat, "TCP statistics (struct tcpstat, netinet/tcp_var.h)"); static int tcp_log_in_vain = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, log_in_vain, CTLFLAG_RW, &tcp_log_in_vain, 0, "Log all incoming TCP segments to closed ports"); static int blackhole = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, blackhole, CTLFLAG_RW, &blackhole, 0, "Do not send RST on segments to closed ports"); int tcp_delack_enabled = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, delayed_ack, CTLFLAG_RW, &tcp_delack_enabled, 0, "Delay ACK to try and piggyback it onto a data packet"); static int drop_synfin = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, drop_synfin, CTLFLAG_RW, &drop_synfin, 0, "Drop TCP packets with SYN+FIN set"); static int tcp_do_rfc3042 = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, rfc3042, CTLFLAG_RW, &tcp_do_rfc3042, 0, "Enable RFC 3042 (Limited Transmit)"); static int tcp_do_rfc3390 = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, rfc3390, CTLFLAG_RW, &tcp_do_rfc3390, 0, "Enable RFC 3390 (Increasing TCP's Initial Congestion Window)"); static int tcp_insecure_rst = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, insecure_rst, CTLFLAG_RW, &tcp_insecure_rst, 0, "Follow the old (insecure) criteria for accepting RST packets"); SYSCTL_NODE(_net_inet_tcp, OID_AUTO, reass, CTLFLAG_RW, 0, "TCP Segment Reassembly Queue"); static int tcp_reass_maxseg = 0; SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, maxsegments, CTLFLAG_RDTUN, &tcp_reass_maxseg, 0, "Global maximum number of TCP Segments in Reassembly Queue"); int tcp_reass_qsize = 0; SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, cursegments, CTLFLAG_RD, &tcp_reass_qsize, 0, "Global number of TCP Segments currently in Reassembly Queue"); static int tcp_reass_maxqlen = 48; SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, maxqlen, CTLFLAG_RW, &tcp_reass_maxqlen, 0, "Maximum number of TCP Segments per individual Reassembly Queue"); static int tcp_reass_overflows = 0; SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, overflows, CTLFLAG_RD, &tcp_reass_overflows, 0, "Global number of TCP Segment Reassembly Queue Overflows"); int tcp_do_autorcvbuf = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, recvbuf_auto, CTLFLAG_RW, &tcp_do_autorcvbuf, 0, "Enable automatic receive buffer sizing"); int tcp_autorcvbuf_inc = 16*1024; SYSCTL_INT(_net_inet_tcp, OID_AUTO, recvbuf_inc, CTLFLAG_RW, &tcp_autorcvbuf_inc, 0, "Incrementor step size of automatic receive buffer"); int tcp_autorcvbuf_max = 256*1024; SYSCTL_INT(_net_inet_tcp, OID_AUTO, recvbuf_max, CTLFLAG_RW, &tcp_autorcvbuf_max, 0, "Max size of automatic receive buffer"); struct inpcbhead tcb; #define tcb6 tcb /* for KAME src sync over BSD*'s */ struct inpcbinfo tcbinfo; static void tcp_dooptions(struct tcpopt *, u_char *, int, int); static int tcp_do_segment(struct mbuf *, struct tcphdr *, struct socket *, struct tcpcb *, int, int); static void tcp_dropwithreset(struct mbuf *, struct tcphdr *, struct tcpcb *, int, int); static void tcp_pulloutofband(struct socket *, struct tcphdr *, struct mbuf *, int); static int tcp_reass(struct tcpcb *, struct tcphdr *, int *, struct mbuf *); static void tcp_xmit_timer(struct tcpcb *, int); static void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *); static int tcp_timewait(struct inpcb *, struct tcpopt *, struct tcphdr *, struct mbuf *, int); /* Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. */ #ifdef INET6 #define ND6_HINT(tp) \ do { \ if ((tp) && (tp)->t_inpcb && \ ((tp)->t_inpcb->inp_vflag & INP_IPV6) != 0) \ nd6_nud_hint(NULL, NULL, 0); \ } while (0) #else #define ND6_HINT(tp) #endif /* * Indicate whether this ack should be delayed. We can delay the ack if * - there is no delayed ack timer in progress and * - our last ack wasn't a 0-sized window. We never want to delay * the ack that opens up a 0-sized window and * - delayed acks are enabled or * - this is a half-synchronized T/TCP connection. */ #define DELAY_ACK(tp) \ ((!tcp_timer_active(tp, TT_DELACK) && \ (tp->t_flags & TF_RXWIN0SENT) == 0) && \ (tcp_delack_enabled || (tp->t_flags & TF_NEEDSYN))) /* Initialize TCP reassembly queue */ static void tcp_reass_zone_change(void *tag) { tcp_reass_maxseg = nmbclusters / 16; uma_zone_set_max(tcp_reass_zone, tcp_reass_maxseg); } uma_zone_t tcp_reass_zone; void tcp_reass_init() { tcp_reass_maxseg = nmbclusters / 16; TUNABLE_INT_FETCH("net.inet.tcp.reass.maxsegments", &tcp_reass_maxseg); tcp_reass_zone = uma_zcreate("tcpreass", sizeof (struct tseg_qent), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcp_reass_zone, tcp_reass_maxseg); EVENTHANDLER_REGISTER(nmbclusters_change, tcp_reass_zone_change, NULL, EVENTHANDLER_PRI_ANY); } static int tcp_reass(struct tcpcb *tp, struct tcphdr *th, int *tlenp, struct mbuf *m) { struct tseg_qent *q; struct tseg_qent *p = NULL; struct tseg_qent *nq; struct tseg_qent *te = NULL; struct socket *so = tp->t_inpcb->inp_socket; int flags; INP_LOCK_ASSERT(tp->t_inpcb); /* * XXX: tcp_reass() is rather inefficient with its data structures * and should be rewritten (see NetBSD for optimizations). While * doing that it should move to its own file tcp_reass.c. */ /* * Call with th==NULL after become established to * force pre-ESTABLISHED data up to user socket. */ if (th == NULL) goto present; /* * Limit the number of segments in the reassembly queue to prevent * holding on to too many segments (and thus running out of mbufs). * Make sure to let the missing segment through which caused this * queue. Always keep one global queue entry spare to be able to * process the missing segment. */ if (th->th_seq != tp->rcv_nxt && (tcp_reass_qsize + 1 >= tcp_reass_maxseg || tp->t_segqlen >= tcp_reass_maxqlen)) { tcp_reass_overflows++; tcpstat.tcps_rcvmemdrop++; m_freem(m); *tlenp = 0; return (0); } /* * Allocate a new queue entry. If we can't, or hit the zone limit * just drop the pkt. */ te = uma_zalloc(tcp_reass_zone, M_NOWAIT); if (te == NULL) { tcpstat.tcps_rcvmemdrop++; m_freem(m); *tlenp = 0; return (0); } tp->t_segqlen++; tcp_reass_qsize++; /* * Find a segment which begins after this one does. */ LIST_FOREACH(q, &tp->t_segq, tqe_q) { if (SEQ_GT(q->tqe_th->th_seq, th->th_seq)) break; p = q; } /* * If there is a preceding segment, it may provide some of * our data already. If so, drop the data from the incoming * segment. If it provides all of our data, drop us. */ if (p != NULL) { int i; /* conversion to int (in i) handles seq wraparound */ i = p->tqe_th->th_seq + p->tqe_len - th->th_seq; if (i > 0) { if (i >= *tlenp) { tcpstat.tcps_rcvduppack++; tcpstat.tcps_rcvdupbyte += *tlenp; m_freem(m); uma_zfree(tcp_reass_zone, te); tp->t_segqlen--; tcp_reass_qsize--; /* * Try to present any queued data * at the left window edge to the user. * This is needed after the 3-WHS * completes. */ goto present; /* ??? */ } m_adj(m, i); *tlenp -= i; th->th_seq += i; } } tcpstat.tcps_rcvoopack++; tcpstat.tcps_rcvoobyte += *tlenp; /* * While we overlap succeeding segments trim them or, * if they are completely covered, dequeue them. */ while (q) { int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq; if (i <= 0) break; if (i < q->tqe_len) { q->tqe_th->th_seq += i; q->tqe_len -= i; m_adj(q->tqe_m, i); break; } nq = LIST_NEXT(q, tqe_q); LIST_REMOVE(q, tqe_q); m_freem(q->tqe_m); uma_zfree(tcp_reass_zone, q); tp->t_segqlen--; tcp_reass_qsize--; q = nq; } /* Insert the new segment queue entry into place. */ te->tqe_m = m; te->tqe_th = th; te->tqe_len = *tlenp; if (p == NULL) { LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q); } else { LIST_INSERT_AFTER(p, te, tqe_q); } present: /* * Present data to user, advancing rcv_nxt through * completed sequence space. */ if (!TCPS_HAVEESTABLISHED(tp->t_state)) return (0); q = LIST_FIRST(&tp->t_segq); if (!q || q->tqe_th->th_seq != tp->rcv_nxt) return (0); SOCKBUF_LOCK(&so->so_rcv); do { tp->rcv_nxt += q->tqe_len; flags = q->tqe_th->th_flags & TH_FIN; nq = LIST_NEXT(q, tqe_q); LIST_REMOVE(q, tqe_q); if (so->so_rcv.sb_state & SBS_CANTRCVMORE) m_freem(q->tqe_m); else sbappendstream_locked(&so->so_rcv, q->tqe_m); uma_zfree(tcp_reass_zone, q); tp->t_segqlen--; tcp_reass_qsize--; q = nq; } while (q && q->tqe_th->th_seq == tp->rcv_nxt); ND6_HINT(tp); sorwakeup_locked(so); return (flags); } /* * TCP input routine, follows pages 65-76 of the * protocol specification dated September, 1981 very closely. */ #ifdef INET6 int tcp6_input(struct mbuf **mp, int *offp, int proto) { struct mbuf *m = *mp; struct in6_ifaddr *ia6; IP6_EXTHDR_CHECK(m, *offp, sizeof(struct tcphdr), IPPROTO_DONE); /* * draft-itojun-ipv6-tcp-to-anycast * better place to put this in? */ ia6 = ip6_getdstifaddr(m); if (ia6 && (ia6->ia6_flags & IN6_IFF_ANYCAST)) { struct ip6_hdr *ip6; ip6 = mtod(m, struct ip6_hdr *); icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, (caddr_t)&ip6->ip6_dst - (caddr_t)ip6); return IPPROTO_DONE; } tcp_input(m, *offp); return IPPROTO_DONE; } #endif void tcp_input(struct mbuf *m, int off0) { struct tcphdr *th; struct ip *ip = NULL; struct ipovly *ipov; struct inpcb *inp = NULL; struct tcpcb *tp = NULL; struct socket *so = NULL; u_char *optp = NULL; int optlen = 0; int len, tlen, off; int drop_hdrlen; int thflags; int rstreason = 0; /* For badport_bandlim accounting purposes */ #ifdef IPFIREWALL_FORWARD struct m_tag *fwd_tag; #endif #ifdef INET6 struct ip6_hdr *ip6 = NULL; int isipv6; char ip6buf[INET6_ADDRSTRLEN]; #else const int isipv6 = 0; #endif struct tcpopt to; /* options in this segment */ #ifdef TCPDEBUG /* * The size of tcp_saveipgen must be the size of the max ip header, * now IPv6. */ u_char tcp_saveipgen[IP6_HDR_LEN]; struct tcphdr tcp_savetcp; short ostate = 0; #endif #ifdef INET6 isipv6 = (mtod(m, struct ip *)->ip_v == 6) ? 1 : 0; #endif to.to_flags = 0; tcpstat.tcps_rcvtotal++; if (isipv6) { #ifdef INET6 /* IP6_EXTHDR_CHECK() is already done at tcp6_input() */ ip6 = mtod(m, struct ip6_hdr *); tlen = sizeof(*ip6) + ntohs(ip6->ip6_plen) - off0; if (in6_cksum(m, IPPROTO_TCP, off0, tlen)) { tcpstat.tcps_rcvbadsum++; goto drop; } th = (struct tcphdr *)((caddr_t)ip6 + off0); /* * Be proactive about unspecified IPv6 address in source. * As we use all-zero to indicate unbounded/unconnected pcb, * unspecified IPv6 address can be used to confuse us. * * Note that packets with unspecified IPv6 destination is * already dropped in ip6_input. */ if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { /* XXX stat */ goto drop; } #else th = NULL; /* XXX: avoid compiler warning */ #endif } else { /* * Get IP and TCP header together in first mbuf. * Note: IP leaves IP header in first mbuf. */ if (off0 > sizeof (struct ip)) { ip_stripoptions(m, (struct mbuf *)0); off0 = sizeof(struct ip); } if (m->m_len < sizeof (struct tcpiphdr)) { if ((m = m_pullup(m, sizeof (struct tcpiphdr))) == NULL) { tcpstat.tcps_rcvshort++; return; } } ip = mtod(m, struct ip *); ipov = (struct ipovly *)ip; th = (struct tcphdr *)((caddr_t)ip + off0); tlen = ip->ip_len; if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) th->th_sum = m->m_pkthdr.csum_data; else th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl(m->m_pkthdr.csum_data + ip->ip_len + IPPROTO_TCP)); th->th_sum ^= 0xffff; #ifdef TCPDEBUG ipov->ih_len = (u_short)tlen; ipov->ih_len = htons(ipov->ih_len); #endif } else { /* * Checksum extended TCP header and data. */ len = sizeof (struct ip) + tlen; bzero(ipov->ih_x1, sizeof(ipov->ih_x1)); ipov->ih_len = (u_short)tlen; ipov->ih_len = htons(ipov->ih_len); th->th_sum = in_cksum(m, len); } if (th->th_sum) { tcpstat.tcps_rcvbadsum++; goto drop; } /* Re-initialization for later version check */ ip->ip_v = IPVERSION; } /* * Check that TCP offset makes sense, * pull out TCP options and adjust length. XXX */ off = th->th_off << 2; if (off < sizeof (struct tcphdr) || off > tlen) { tcpstat.tcps_rcvbadoff++; goto drop; } tlen -= off; /* tlen is used instead of ti->ti_len */ if (off > sizeof (struct tcphdr)) { if (isipv6) { #ifdef INET6 IP6_EXTHDR_CHECK(m, off0, off, ); ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)((caddr_t)ip6 + off0); #endif } else { if (m->m_len < sizeof(struct ip) + off) { if ((m = m_pullup(m, sizeof (struct ip) + off)) == NULL) { tcpstat.tcps_rcvshort++; return; } ip = mtod(m, struct ip *); ipov = (struct ipovly *)ip; th = (struct tcphdr *)((caddr_t)ip + off0); } } optlen = off - sizeof (struct tcphdr); optp = (u_char *)(th + 1); } thflags = th->th_flags; /* * If the drop_synfin option is enabled, drop all packets with * both the SYN and FIN bits set. This prevents e.g. nmap from * identifying the TCP/IP stack. * * This is a violation of the TCP specification. */ if (drop_synfin && (thflags & (TH_SYN|TH_FIN)) == (TH_SYN|TH_FIN)) goto drop; /* * Convert TCP protocol specific fields to host format. */ th->th_seq = ntohl(th->th_seq); th->th_ack = ntohl(th->th_ack); th->th_win = ntohs(th->th_win); th->th_urp = ntohs(th->th_urp); /* * Delay dropping TCP, IP headers, IPv6 ext headers, and TCP options. */ drop_hdrlen = off0 + off; /* * Locate pcb for segment. */ INP_INFO_WLOCK(&tcbinfo); findpcb: INP_INFO_WLOCK_ASSERT(&tcbinfo); #ifdef IPFIREWALL_FORWARD /* Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. */ fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); if (fwd_tag != NULL && isipv6 == 0) { /* IPv6 support is not yet */ struct sockaddr_in *next_hop; next_hop = (struct sockaddr_in *)(fwd_tag+1); /* * Transparently forwarded. Pretend to be the destination. * already got one like this? */ inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, th->th_sport, ip->ip_dst, th->th_dport, 0, m->m_pkthdr.rcvif); if (!inp) { /* It's new. Try to find the ambushing socket. */ inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, th->th_sport, next_hop->sin_addr, next_hop->sin_port ? ntohs(next_hop->sin_port) : th->th_dport, INPLOOKUP_WILDCARD, m->m_pkthdr.rcvif); } /* Remove the tag from the packet. We don't need it anymore. */ m_tag_delete(m, fwd_tag); } else #endif /* IPFIREWALL_FORWARD */ { if (isipv6) { #ifdef INET6 inp = in6_pcblookup_hash(&tcbinfo, &ip6->ip6_src, th->th_sport, &ip6->ip6_dst, th->th_dport, INPLOOKUP_WILDCARD, m->m_pkthdr.rcvif); #endif } else inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, th->th_sport, ip->ip_dst, th->th_dport, INPLOOKUP_WILDCARD, m->m_pkthdr.rcvif); } #if defined(IPSEC) || defined(FAST_IPSEC) #ifdef INET6 if (isipv6 && inp != NULL && ipsec6_in_reject(m, inp)) { #ifdef IPSEC ipsec6stat.in_polvio++; #endif goto dropunlock; } else #endif /* INET6 */ if (inp != NULL && ipsec4_in_reject(m, inp)) { #ifdef IPSEC ipsecstat.in_polvio++; #endif goto dropunlock; } #endif /*IPSEC || FAST_IPSEC*/ /* * If the INPCB does not exist then all data in the incoming * segment is discarded and an appropriate RST is sent back. */ if (inp == NULL) { /* * Log communication attempts to ports that are not * in use. */ if ((tcp_log_in_vain == 1 && (thflags & TH_SYN)) || tcp_log_in_vain == 2) { #ifndef INET6 char dbuf[4*sizeof "123"], sbuf[4*sizeof "123"]; #else char dbuf[INET6_ADDRSTRLEN+2], sbuf[INET6_ADDRSTRLEN+2]; if (isipv6) { strcpy(dbuf, "["); strcat(dbuf, ip6_sprintf(ip6buf, &ip6->ip6_dst)); strcat(dbuf, "]"); strcpy(sbuf, "["); strcat(sbuf, ip6_sprintf(ip6buf, &ip6->ip6_src)); strcat(sbuf, "]"); } else #endif /* INET6 */ { strcpy(dbuf, inet_ntoa(ip->ip_dst)); strcpy(sbuf, inet_ntoa(ip->ip_src)); } log(LOG_INFO, "Connection attempt to TCP %s:%d " "from %s:%d flags:0x%02x\n", dbuf, ntohs(th->th_dport), sbuf, ntohs(th->th_sport), thflags); } /* * When blackholing do not respond with a RST but * completely ignore the segment and drop it. */ if ((blackhole == 1 && (thflags & TH_SYN)) || blackhole == 2) goto dropunlock; rstreason = BANDLIM_RST_CLOSEDPORT; goto dropwithreset; } INP_LOCK(inp); /* Check the minimum TTL for socket. */ if (inp->inp_ip_minttl != 0) { #ifdef INET6 if (isipv6 && inp->inp_ip_minttl > ip6->ip6_hlim) goto dropunlock; else #endif if (inp->inp_ip_minttl > ip->ip_ttl) goto dropunlock; } /* * A previous connection in TIMEWAIT state is supposed to catch * stray or duplicate segments arriving late. If this segment * was a legitimate new connection attempt the old INPCB gets * removed and we can try again to find a listening socket. */ if (inp->inp_vflag & INP_TIMEWAIT) { if (thflags & TH_SYN) tcp_dooptions(&to, optp, optlen, TO_SYN); if (tcp_timewait(inp, &to, th, m, tlen)) goto findpcb; /* tcp_timewait unlocks inp. */ INP_INFO_WUNLOCK(&tcbinfo); return; } /* * The TCPCB may no longer exist if the connection is winding * down or it is in the CLOSED state. Either way we drop the * segment and send an appropriate response. */ tp = intotcpcb(inp); if (tp == NULL) { INP_UNLOCK(inp); rstreason = BANDLIM_RST_CLOSEDPORT; goto dropwithreset; } if (tp->t_state == TCPS_CLOSED) goto dropunlock; /* XXX: dropwithreset??? */ #ifdef MAC INP_LOCK_ASSERT(inp); if (mac_check_inpcb_deliver(inp, m)) goto dropunlock; #endif so = inp->inp_socket; KASSERT(so != NULL, ("%s: so == NULL", __func__)); #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) { ostate = tp->t_state; if (isipv6) bcopy((char *)ip6, (char *)tcp_saveipgen, sizeof(*ip6)); else bcopy((char *)ip, (char *)tcp_saveipgen, sizeof(*ip)); tcp_savetcp = *th; } #endif /* * When the socket is accepting connections (the INPCB is in LISTEN * state) we look into the SYN cache if this is a new connection * attempt or the completion of a previous one. */ if (so->so_options & SO_ACCEPTCONN) { struct in_conninfo inc; bzero(&inc, sizeof(inc)); inc.inc_isipv6 = isipv6; #ifdef INET6 if (isipv6) { inc.inc6_faddr = ip6->ip6_src; inc.inc6_laddr = ip6->ip6_dst; } else #endif { inc.inc_faddr = ip->ip_src; inc.inc_laddr = ip->ip_dst; } inc.inc_fport = th->th_sport; inc.inc_lport = th->th_dport; /* * If the state is LISTEN then ignore segment if it contains * a RST. If the segment contains an ACK then it is bad and * send a RST. If it does not contain a SYN then it is not * interesting; drop it. * * If the state is SYN_RECEIVED (syncache) and seg contains * an ACK, but not for our SYN/ACK, send a RST. If the seg * contains a RST, check the sequence number to see if it * is a valid reset segment. */ if ((thflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { if ((thflags & (TH_RST|TH_ACK|TH_SYN)) == TH_ACK) { /* * Parse the TCP options here because * syncookies need access to the reflected * timestamp. */ tcp_dooptions(&to, optp, optlen, 0); if (!syncache_expand(&inc, &to, th, &so, m)) { /* * No syncache entry or ACK was not * for our SYN/ACK. Send a RST. */ rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } if (so == NULL) { /* * We completed the 3-way handshake * but could not allocate a socket * either due to memory shortage, * listen queue length limits or * global socket limits. */ rstreason = BANDLIM_UNLIMITED; goto dropwithreset; } /* * Socket is created in state SYN_RECEIVED. * Continue processing segment. */ INP_UNLOCK(inp); /* listen socket */ inp = sotoinpcb(so); INP_LOCK(inp); /* new connection */ tp = intotcpcb(inp); /* * Process the segment and the data it * contains. tcp_do_segment() consumes * the mbuf chain and unlocks the inpcb. * XXX: The potential return value of * TIME_WAIT nuked is supposed to be * handled above. */ if (tcp_do_segment(m, th, so, tp, drop_hdrlen, tlen)) goto findpcb; /* TIME_WAIT nuked */ return; } if (thflags & TH_RST) { syncache_chkrst(&inc, th); goto dropunlock; } if (thflags & TH_ACK) { syncache_badack(&inc); tcpstat.tcps_badsyn++; rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } goto dropunlock; } /* * Segment's flags are (SYN) or (SYN|FIN). */ #ifdef INET6 /* * If deprecated address is forbidden, * we do not accept SYN to deprecated interface * address to prevent any new inbound connection from * getting established. * When we do not accept SYN, we send a TCP RST, * with deprecated source address (instead of dropping * it). We compromise it as it is much better for peer * to send a RST, and RST will be the final packet * for the exchange. * * If we do not forbid deprecated addresses, we accept * the SYN packet. RFC2462 does not suggest dropping * SYN in this case. * If we decipher RFC2462 5.5.4, it says like this: * 1. use of deprecated addr with existing * communication is okay - "SHOULD continue to be * used" * 2. use of it with new communication: * (2a) "SHOULD NOT be used if alternate address * with sufficient scope is available" * (2b) nothing mentioned otherwise. * Here we fall into (2b) case as we have no choice in * our source address selection - we must obey the peer. * * The wording in RFC2462 is confusing, and there are * multiple description text for deprecated address * handling - worse, they are not exactly the same. * I believe 5.5.4 is the best one, so we follow 5.5.4. */ if (isipv6 && !ip6_use_deprecated) { struct in6_ifaddr *ia6; if ((ia6 = ip6_getdstifaddr(m)) && (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { INP_UNLOCK(inp); tp = NULL; rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } } #endif /* * Basic sanity checks on incoming SYN requests: * * Don't bother responding if the destination was a * broadcast according to RFC1122 4.2.3.10, p. 104. * * If it is from this socket, drop it, it must be forged. * * Note that it is quite possible to receive unicast * link-layer packets with a broadcast IP address. Use * in_broadcast() to find them. */ if (m->m_flags & (M_BCAST|M_MCAST)) goto dropunlock; if (isipv6) { #ifdef INET6 if (th->th_dport == th->th_sport && IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &ip6->ip6_src)) goto dropunlock; if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto dropunlock; #endif } else { if (th->th_dport == th->th_sport && ip->ip_dst.s_addr == ip->ip_src.s_addr) goto dropunlock; if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || ip->ip_src.s_addr == htonl(INADDR_BROADCAST) || in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) goto dropunlock; } /* * SYN appears to be valid. Create compressed TCP state * for syncache. */ - if (so->so_qlen <= so->so_qlimit) { #ifdef TCPDEBUG - if (so->so_options & SO_DEBUG) - tcp_trace(TA_INPUT, ostate, tp, - (void *)tcp_saveipgen, &tcp_savetcp, 0); + if (so->so_options & SO_DEBUG) + tcp_trace(TA_INPUT, ostate, tp, + (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif - tcp_dooptions(&to, optp, optlen, TO_SYN); - if (!syncache_add(&inc, &to, th, inp, &so, m)) - goto dropunlock; - /* - * Entry added to syncache, mbuf used to - * send SYN-ACK packet. Everything unlocked - * already. - */ - return; - } - /* Catch all. Everthing that makes it down here is junk. */ - goto dropunlock; + tcp_dooptions(&to, optp, optlen, TO_SYN); + syncache_add(&inc, &to, th, inp, &so, m); + /* + * Entry added to syncache and mbuf consumed. + * Everything unlocked already by syncache_add(). + */ + return; } /* * Segment belongs to a connection in SYN_SENT, ESTABLISHED or * later state. tcp_do_segment() always consumes the mbuf chain * and unlocks the inpcb. */ if (tcp_do_segment(m, th, so, tp, drop_hdrlen, tlen)) goto findpcb; /* XXX: TIME_WAIT was nuked. */ return; dropwithreset: INP_INFO_WLOCK_ASSERT(&tcbinfo); tcp_dropwithreset(m, th, tp, tlen, rstreason); m = NULL; /* mbuf chain got consumed. */ dropunlock: INP_INFO_WLOCK_ASSERT(&tcbinfo); if (tp != NULL) INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); drop: INP_INFO_UNLOCK_ASSERT(&tcbinfo); if (m != NULL) m_freem(m); return; } static int tcp_do_segment(struct mbuf *m, struct tcphdr *th, struct socket *so, struct tcpcb *tp, int drop_hdrlen, int tlen) { int thflags, acked, ourfinisacked, needoutput = 0; int headlocked = 1; int rstreason, todrop, win; u_long tiwin; struct tcpopt to; #ifdef TCPDEBUG /* * The size of tcp_saveipgen must be the size of the max ip header, * now IPv6. */ u_char tcp_saveipgen[IP6_HDR_LEN]; struct tcphdr tcp_savetcp; short ostate = 0; #endif thflags = th->th_flags; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(tp->t_inpcb); KASSERT(tp->t_state > TCPS_LISTEN, ("%s: TCPS_LISTEN", __func__)); /* * Segment received on connection. * Reset idle time and keep-alive timer. */ tp->t_rcvtime = ticks; if (TCPS_HAVEESTABLISHED(tp->t_state)) tcp_timer_activate(tp, TT_KEEP, tcp_keepidle); /* * Unscale the window into a 32-bit value. * This value is bogus for the TCPS_SYN_SENT state * and is overwritten later. */ tiwin = th->th_win << tp->snd_scale; /* * Parse options on any incoming segment. */ tcp_dooptions(&to, (u_char *)(th + 1), (th->th_off << 2) - sizeof(struct tcphdr), (thflags & TH_SYN) ? TO_SYN : 0); /* * If echoed timestamp is later than the current time, * fall back to non RFC1323 RTT calculation. Normalize * timestamp if syncookies were used when this connection * was established. */ if ((to.to_flags & TOF_TS) && (to.to_tsecr != 0)) { to.to_tsecr -= tp->ts_offset; if (TSTMP_GT(to.to_tsecr, ticks)) to.to_tsecr = 0; } /* * Process options only when we get SYN/ACK back. The SYN case * for incoming connections is handled in tcp_syncache. * XXX this is traditional behavior, may need to be cleaned up. */ if (tp->t_state == TCPS_SYN_SENT && (thflags & TH_SYN)) { if ((to.to_flags & TOF_SCALE) && (tp->t_flags & TF_REQ_SCALE)) { tp->t_flags |= TF_RCVD_SCALE; tp->snd_scale = to.to_wscale; tp->snd_wnd = th->th_win << tp->snd_scale; tiwin = tp->snd_wnd; } if (to.to_flags & TOF_TS) { tp->t_flags |= TF_RCVD_TSTMP; tp->ts_recent = to.to_tsval; tp->ts_recent_age = ticks; } /* Initial send window, already scaled. */ tp->snd_wnd = th->th_win; if (to.to_flags & TOF_MSS) tcp_mss(tp, to.to_mss); if (tp->sack_enable) { if (!(to.to_flags & TOF_SACKPERM)) tp->sack_enable = 0; else tp->t_flags |= TF_SACK_PERMIT; } } /* * Header prediction: check for the two common cases * of a uni-directional data xfer. If the packet has * no control flags, is in-sequence, the window didn't * change and we're not retransmitting, it's a * candidate. If the length is zero and the ack moved * forward, we're the sender side of the xfer. Just * free the data acked & wake any higher level process * that was blocked waiting for space. If the length * is non-zero and the ack didn't move, we're the * receiver side. If we're getting packets in-order * (the reassembly queue is empty), add the data to * the socket buffer and note that we need a delayed ack. * Make sure that the hidden state-flags are also off. * Since we check for TCPS_ESTABLISHED above, it can only * be TH_NEEDSYN. */ if (tp->t_state == TCPS_ESTABLISHED && (thflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK && ((tp->t_flags & (TF_NEEDSYN|TF_NEEDFIN)) == 0) && ((to.to_flags & TOF_TS) == 0 || TSTMP_GEQ(to.to_tsval, tp->ts_recent)) && th->th_seq == tp->rcv_nxt && tiwin && tiwin == tp->snd_wnd && tp->snd_nxt == tp->snd_max) { /* * If last ACK falls within this segment's sequence numbers, * record the timestamp. * NOTE that the test is modified according to the latest * proposal of the tcplw@cray.com list (Braden 1993/04/26). */ if ((to.to_flags & TOF_TS) != 0 && SEQ_LEQ(th->th_seq, tp->last_ack_sent)) { tp->ts_recent_age = ticks; tp->ts_recent = to.to_tsval; } if (tlen == 0) { if (SEQ_GT(th->th_ack, tp->snd_una) && SEQ_LEQ(th->th_ack, tp->snd_max) && tp->snd_cwnd >= tp->snd_wnd && ((!tcp_do_newreno && !tp->sack_enable && tp->t_dupacks < tcprexmtthresh) || ((tcp_do_newreno || tp->sack_enable) && !IN_FASTRECOVERY(tp) && (to.to_flags & TOF_SACK) == 0 && TAILQ_EMPTY(&tp->snd_holes)))) { KASSERT(headlocked, ("%s: headlocked", __func__)); INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; /* * this is a pure ack for outstanding data. */ ++tcpstat.tcps_predack; /* * "bad retransmit" recovery */ if (tp->t_rxtshift == 1 && ticks < tp->t_badrxtwin) { ++tcpstat.tcps_sndrexmitbad; tp->snd_cwnd = tp->snd_cwnd_prev; tp->snd_ssthresh = tp->snd_ssthresh_prev; tp->snd_recover = tp->snd_recover_prev; if (tp->t_flags & TF_WASFRECOVERY) ENTER_FASTRECOVERY(tp); tp->snd_nxt = tp->snd_max; tp->t_badrxtwin = 0; } /* * Recalculate the transmit timer / rtt. * * Some boxes send broken timestamp replies * during the SYN+ACK phase, ignore * timestamps of 0 or we could calculate a * huge RTT and blow up the retransmit timer. */ if ((to.to_flags & TOF_TS) != 0 && to.to_tsecr) { if (!tp->t_rttlow || tp->t_rttlow > ticks - to.to_tsecr) tp->t_rttlow = ticks - to.to_tsecr; tcp_xmit_timer(tp, ticks - to.to_tsecr + 1); } else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) { if (!tp->t_rttlow || tp->t_rttlow > ticks - tp->t_rtttime) tp->t_rttlow = ticks - tp->t_rtttime; tcp_xmit_timer(tp, ticks - tp->t_rtttime); } tcp_xmit_bandwidth_limit(tp, th->th_ack); acked = th->th_ack - tp->snd_una; tcpstat.tcps_rcvackpack++; tcpstat.tcps_rcvackbyte += acked; sbdrop(&so->so_snd, acked); if (SEQ_GT(tp->snd_una, tp->snd_recover) && SEQ_LEQ(th->th_ack, tp->snd_recover)) tp->snd_recover = th->th_ack - 1; tp->snd_una = th->th_ack; /* * pull snd_wl2 up to prevent seq wrap relative * to th_ack. */ tp->snd_wl2 = th->th_ack; tp->t_dupacks = 0; m_freem(m); ND6_HINT(tp); /* some progress has been done */ /* * If all outstanding data are acked, stop * retransmit timer, otherwise restart timer * using current (possibly backed-off) value. * If process is waiting for space, * wakeup/selwakeup/signal. If data * are ready to send, let tcp_output * decide between more output or persist. #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif */ if (tp->snd_una == tp->snd_max) tcp_timer_activate(tp, TT_REXMT, 0); else if (!tcp_timer_active(tp, TT_PERSIST)) tcp_timer_activate(tp, TT_REXMT, tp->t_rxtcur); sowwakeup(so); if (so->so_snd.sb_cc) (void) tcp_output(tp); goto check_delack; } } else if (th->th_ack == tp->snd_una && LIST_EMPTY(&tp->t_segq) && tlen <= sbspace(&so->so_rcv)) { int newsize = 0; /* automatic sockbuf scaling */ KASSERT(headlocked, ("%s: headlocked", __func__)); INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; /* * this is a pure, in-sequence data packet * with nothing on the reassembly queue and * we have enough buffer space to take it. */ /* Clean receiver SACK report if present */ if (tp->sack_enable && tp->rcv_numsacks) tcp_clean_sackreport(tp); ++tcpstat.tcps_preddat; tp->rcv_nxt += tlen; /* * Pull snd_wl1 up to prevent seq wrap relative to * th_seq. */ tp->snd_wl1 = th->th_seq; /* * Pull rcv_up up to prevent seq wrap relative to * rcv_nxt. */ tp->rcv_up = tp->rcv_nxt; tcpstat.tcps_rcvpack++; tcpstat.tcps_rcvbyte += tlen; ND6_HINT(tp); /* some progress has been done */ #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif /* * Automatic sizing of receive socket buffer. Often the send * buffer size is not optimally adjusted to the actual network * conditions at hand (delay bandwidth product). Setting the * buffer size too small limits throughput on links with high * bandwidth and high delay (eg. trans-continental/oceanic links). * * On the receive side the socket buffer memory is only rarely * used to any significant extent. This allows us to be much * more aggressive in scaling the receive socket buffer. For * the case that the buffer space is actually used to a large * extent and we run out of kernel memory we can simply drop * the new segments; TCP on the sender will just retransmit it * later. Setting the buffer size too big may only consume too * much kernel memory if the application doesn't read() from * the socket or packet loss or reordering makes use of the * reassembly queue. * * The criteria to step up the receive buffer one notch are: * 1. the number of bytes received during the time it takes * one timestamp to be reflected back to us (the RTT); * 2. received bytes per RTT is within seven eighth of the * current socket buffer size; * 3. receive buffer size has not hit maximal automatic size; * * This algorithm does one step per RTT at most and only if * we receive a bulk stream w/o packet losses or reorderings. * Shrinking the buffer during idle times is not necessary as * it doesn't consume any memory when idle. * * TODO: Only step up if the application is actually serving * the buffer to better manage the socket buffer resources. */ if (tcp_do_autorcvbuf && to.to_tsecr && (so->so_rcv.sb_flags & SB_AUTOSIZE)) { if (to.to_tsecr > tp->rfbuf_ts && to.to_tsecr - tp->rfbuf_ts < hz) { if (tp->rfbuf_cnt > (so->so_rcv.sb_hiwat / 8 * 7) && so->so_rcv.sb_hiwat < tcp_autorcvbuf_max) { newsize = min(so->so_rcv.sb_hiwat + tcp_autorcvbuf_inc, tcp_autorcvbuf_max); } /* Start over with next RTT. */ tp->rfbuf_ts = 0; tp->rfbuf_cnt = 0; } else tp->rfbuf_cnt += tlen; /* add up */ } /* Add data to socket buffer. */ SOCKBUF_LOCK(&so->so_rcv); if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { m_freem(m); } else { /* * Set new socket buffer size. * Give up when limit is reached. */ if (newsize) if (!sbreserve_locked(&so->so_rcv, newsize, so, curthread)) so->so_rcv.sb_flags &= ~SB_AUTOSIZE; m_adj(m, drop_hdrlen); /* delayed header drop */ sbappendstream_locked(&so->so_rcv, m); } sorwakeup_locked(so); if (DELAY_ACK(tp)) { tp->t_flags |= TF_DELACK; } else { tp->t_flags |= TF_ACKNOW; tcp_output(tp); } goto check_delack; } } /* * Calculate amount of space in receive window, * and then do TCP input processing. * Receive window is amount of space in rcv queue, * but not less than advertised window. */ win = sbspace(&so->so_rcv); if (win < 0) win = 0; tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); /* Reset receive buffer auto scaling when not in bulk receive mode. */ tp->rfbuf_ts = 0; tp->rfbuf_cnt = 0; switch (tp->t_state) { /* * If the state is SYN_RECEIVED: * if seg contains an ACK, but not for our SYN/ACK, send a RST. */ case TCPS_SYN_RECEIVED: if ((thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->snd_una) || SEQ_GT(th->th_ack, tp->snd_max))) { rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } break; /* * If the state is SYN_SENT: * if seg contains an ACK, but not for our SYN, drop the input. * if seg contains a RST, then drop the connection. * if seg does not contain SYN, then drop it. * Otherwise this is an acceptable SYN segment * initialize tp->rcv_nxt and tp->irs * if seg contains ack then advance tp->snd_una * if SYN has been acked change to ESTABLISHED else SYN_RCVD state * arrange for segment to be acked (eventually) * continue processing rest of data/controls, beginning with URG */ case TCPS_SYN_SENT: if ((thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->iss) || SEQ_GT(th->th_ack, tp->snd_max))) { rstreason = BANDLIM_UNLIMITED; goto dropwithreset; } if (thflags & TH_RST) { if (thflags & TH_ACK) { KASSERT(headlocked, ("%s: after_listen: " "tcp_drop.2: head not locked", __func__)); tp = tcp_drop(tp, ECONNREFUSED); } goto drop; } if ((thflags & TH_SYN) == 0) goto drop; tp->irs = th->th_seq; tcp_rcvseqinit(tp); if (thflags & TH_ACK) { tcpstat.tcps_connects++; soisconnected(so); #ifdef MAC SOCK_LOCK(so); mac_set_socket_peer_from_mbuf(m, so); SOCK_UNLOCK(so); #endif /* Do window scaling on this connection? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->rcv_scale = tp->request_r_scale; } tp->rcv_adv += tp->rcv_wnd; tp->snd_una++; /* SYN is acked */ /* * If there's data, delay ACK; if there's also a FIN * ACKNOW will be turned on later. */ if (DELAY_ACK(tp) && tlen != 0) tcp_timer_activate(tp, TT_DELACK, tcp_delacktime); else tp->t_flags |= TF_ACKNOW; /* * Received in SYN_SENT[*] state. * Transitions: * SYN_SENT --> ESTABLISHED * SYN_SENT* --> FIN_WAIT_1 */ tp->t_starttime = ticks; if (tp->t_flags & TF_NEEDFIN) { tp->t_state = TCPS_FIN_WAIT_1; tp->t_flags &= ~TF_NEEDFIN; thflags &= ~TH_SYN; } else { tp->t_state = TCPS_ESTABLISHED; tcp_timer_activate(tp, TT_KEEP, tcp_keepidle); } } else { /* * Received initial SYN in SYN-SENT[*] state => * simultaneous open. If segment contains CC option * and there is a cached CC, apply TAO test. * If it succeeds, connection is * half-synchronized. * Otherwise, do 3-way handshake: * SYN-SENT -> SYN-RECEIVED * SYN-SENT* -> SYN-RECEIVED* * If there was no CC option, clear cached CC value. */ tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN); tcp_timer_activate(tp, TT_REXMT, 0); tp->t_state = TCPS_SYN_RECEIVED; } KASSERT(headlocked, ("%s: trimthenstep6: head not locked", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); /* * Advance th->th_seq to correspond to first data byte. * If data, trim to stay within window, * dropping FIN if necessary. */ th->th_seq++; if (tlen > tp->rcv_wnd) { todrop = tlen - tp->rcv_wnd; m_adj(m, -todrop); tlen = tp->rcv_wnd; thflags &= ~TH_FIN; tcpstat.tcps_rcvpackafterwin++; tcpstat.tcps_rcvbyteafterwin += todrop; } tp->snd_wl1 = th->th_seq - 1; tp->rcv_up = th->th_seq; /* * Client side of transaction: already sent SYN and data. * If the remote host used T/TCP to validate the SYN, * our data will be ACK'd; if so, enter normal data segment * processing in the middle of step 5, ack processing. * Otherwise, goto step 6. */ if (thflags & TH_ACK) goto process_ACK; goto step6; /* * If the state is LAST_ACK or CLOSING or TIME_WAIT: * do normal processing. * * NB: Leftover from RFC1644 T/TCP. Cases to be reused later. */ case TCPS_LAST_ACK: case TCPS_CLOSING: case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); break; /* continue normal processing */ } /* * States other than LISTEN or SYN_SENT. * First check the RST flag and sequence number since reset segments * are exempt from the timestamp and connection count tests. This * fixes a bug introduced by the Stevens, vol. 2, p. 960 bugfix * below which allowed reset segments in half the sequence space * to fall though and be processed (which gives forged reset * segments with a random sequence number a 50 percent chance of * killing a connection). * Then check timestamp, if present. * Then check the connection count, if present. * Then check that at least some bytes of segment are within * receive window. If segment begins before rcv_nxt, * drop leading data (and SYN); if nothing left, just ack. * * * 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. * Note: this does not take into account delayed ACKs, so * we should test against last_ack_sent instead of rcv_nxt. * 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. * Note 2: Paul Watson's paper "Slipping in the Window" has shown * that brute force RST attacks are possible. To combat this, * we use a much stricter check while in the ESTABLISHED state, * only accepting RSTs where the sequence number is equal to * last_ack_sent. In all other states (the states in which a * RST is more likely), the more permissive check is used. * If we have multiple segments in flight, the intial reset * segment sequence numbers will be to the left of last_ack_sent, * but they will eventually catch up. * In any case, it never made sense to trim reset segments to * fit the receive window since RFC 1122 says: * 4.2.2.12 RST Segment: RFC-793 Section 3.4 * * A TCP SHOULD allow a received RST segment to include data. * * DISCUSSION * It has been suggested that a RST segment could contain * ASCII text that encoded and explained the cause of the * RST. No standard has yet been established for such * data. * * If the reset segment passes the sequence number test examine * the state: * SYN_RECEIVED STATE: * If passive open, return to LISTEN state. * If active open, inform user that connection was refused. * ESTABLISHED, FIN_WAIT_1, FIN_WAIT_2, CLOSE_WAIT STATES: * Inform user that connection was reset, and close tcb. * CLOSING, LAST_ACK STATES: * Close the tcb. * TIME_WAIT STATE: * Drop the segment - see Stevens, vol. 2, p. 964 and * RFC 1337. */ if (thflags & TH_RST) { if (SEQ_GEQ(th->th_seq, tp->last_ack_sent - 1) && SEQ_LEQ(th->th_seq, tp->last_ack_sent + tp->rcv_wnd)) { switch (tp->t_state) { case TCPS_SYN_RECEIVED: so->so_error = ECONNREFUSED; goto close; case TCPS_ESTABLISHED: if (tcp_insecure_rst == 0 && !(SEQ_GEQ(th->th_seq, tp->rcv_nxt - 1) && SEQ_LEQ(th->th_seq, tp->rcv_nxt + 1)) && !(SEQ_GEQ(th->th_seq, tp->last_ack_sent - 1) && SEQ_LEQ(th->th_seq, tp->last_ack_sent + 1))) { tcpstat.tcps_badrst++; goto drop; } case TCPS_FIN_WAIT_1: case TCPS_FIN_WAIT_2: case TCPS_CLOSE_WAIT: so->so_error = ECONNRESET; close: tp->t_state = TCPS_CLOSED; tcpstat.tcps_drops++; KASSERT(headlocked, ("%s: trimthenstep6: " "tcp_close: head not locked", __func__)); tp = tcp_close(tp); break; case TCPS_CLOSING: case TCPS_LAST_ACK: KASSERT(headlocked, ("%s: trimthenstep6: " "tcp_close.2: head not locked", __func__)); tp = tcp_close(tp); break; case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); break; } } goto drop; } /* * RFC 1323 PAWS: If we have a timestamp reply on this segment * and it's less than ts_recent, drop it. */ if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent && TSTMP_LT(to.to_tsval, tp->ts_recent)) { /* Check to see if ts_recent is over 24 days old. */ if ((int)(ticks - tp->ts_recent_age) > TCP_PAWS_IDLE) { /* * Invalidate ts_recent. If this segment updates * ts_recent, the age will be reset later and ts_recent * will get a valid value. If it does not, setting * ts_recent to zero will at least satisfy the * requirement that zero be placed in the timestamp * echo reply when ts_recent isn't valid. The * age isn't reset until we get a valid ts_recent * because we don't want out-of-order segments to be * dropped when ts_recent is old. */ tp->ts_recent = 0; } else { tcpstat.tcps_rcvduppack++; tcpstat.tcps_rcvdupbyte += tlen; tcpstat.tcps_pawsdrop++; if (tlen) goto dropafterack; goto drop; } } /* * In the SYN-RECEIVED state, validate that the packet belongs to * this connection before trimming the data to fit the receive * window. Check the sequence number versus IRS since we know * the sequence numbers haven't wrapped. This is a partial fix * for the "LAND" DoS attack. */ if (tp->t_state == TCPS_SYN_RECEIVED && SEQ_LT(th->th_seq, tp->irs)) { rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } todrop = tp->rcv_nxt - th->th_seq; if (todrop > 0) { if (thflags & TH_SYN) { thflags &= ~TH_SYN; th->th_seq++; if (th->th_urp > 1) th->th_urp--; else thflags &= ~TH_URG; todrop--; } /* * Following if statement from Stevens, vol. 2, p. 960. */ if (todrop > tlen || (todrop == tlen && (thflags & TH_FIN) == 0)) { /* * Any valid FIN must be to the left of the window. * At this point the FIN must be a duplicate or out * of sequence; drop it. */ thflags &= ~TH_FIN; /* * Send an ACK to resynchronize and drop any data. * But keep on processing for RST or ACK. */ tp->t_flags |= TF_ACKNOW; todrop = tlen; tcpstat.tcps_rcvduppack++; tcpstat.tcps_rcvdupbyte += todrop; } else { tcpstat.tcps_rcvpartduppack++; tcpstat.tcps_rcvpartdupbyte += todrop; } drop_hdrlen += todrop; /* drop from the top afterwards */ th->th_seq += todrop; tlen -= todrop; if (th->th_urp > todrop) th->th_urp -= todrop; else { thflags &= ~TH_URG; th->th_urp = 0; } } /* * If new data are received on a connection after the * user processes are gone, then RST the other end. */ if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT && tlen) { KASSERT(headlocked, ("%s: trimthenstep6: tcp_close.3: head " "not locked", __func__)); tp = tcp_close(tp); tcpstat.tcps_rcvafterclose++; rstreason = BANDLIM_UNLIMITED; goto dropwithreset; } /* * If segment ends after window, drop trailing data * (and PUSH and FIN); if nothing left, just ACK. */ todrop = (th->th_seq+tlen) - (tp->rcv_nxt+tp->rcv_wnd); if (todrop > 0) { tcpstat.tcps_rcvpackafterwin++; if (todrop >= tlen) { tcpstat.tcps_rcvbyteafterwin += tlen; /* * If a new connection request is received * while in TIME_WAIT, drop the old connection * and start over if the sequence numbers * are above the previous ones. */ KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); if (thflags & TH_SYN && tp->t_state == TCPS_TIME_WAIT && SEQ_GT(th->th_seq, tp->rcv_nxt)) { KASSERT(headlocked, ("%s: trimthenstep6: " "tcp_close.4: head not locked", __func__)); tp = tcp_close(tp); /* XXX: Shouldn't be possible. */ return (1); } /* * If window is closed can only take segments at * window edge, and have to drop data and PUSH from * incoming segments. Continue processing, but * remember to ack. Otherwise, drop segment * and ack. */ if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) { tp->t_flags |= TF_ACKNOW; tcpstat.tcps_rcvwinprobe++; } else goto dropafterack; } else tcpstat.tcps_rcvbyteafterwin += todrop; m_adj(m, -todrop); tlen -= todrop; thflags &= ~(TH_PUSH|TH_FIN); } /* * If last ACK falls within this segment's sequence numbers, * record its timestamp. * NOTE: * 1) That the test incorporates suggestions from the latest * proposal of the tcplw@cray.com list (Braden 1993/04/26). * 2) That updating only on newer timestamps interferes with * our earlier PAWS tests, so this check should be solely * predicated on the sequence space of this segment. * 3) That we modify the segment boundary check to be * Last.ACK.Sent <= SEG.SEQ + SEG.Len * instead of RFC1323's * Last.ACK.Sent < SEG.SEQ + SEG.Len, * This modified check allows us to overcome RFC1323's * limitations as described in Stevens TCP/IP Illustrated * Vol. 2 p.869. In such cases, we can still calculate the * RTT correctly when RCV.NXT == Last.ACK.Sent. */ if ((to.to_flags & TOF_TS) != 0 && SEQ_LEQ(th->th_seq, tp->last_ack_sent) && SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen + ((thflags & (TH_SYN|TH_FIN)) != 0))) { tp->ts_recent_age = ticks; tp->ts_recent = to.to_tsval; } /* * If a SYN is in the window, then this is an * error and we send an RST and drop the connection. */ if (thflags & TH_SYN) { KASSERT(headlocked, ("%s: tcp_drop: trimthenstep6: " "head not locked", __func__)); tp = tcp_drop(tp, ECONNRESET); rstreason = BANDLIM_UNLIMITED; goto drop; } /* * If the ACK bit is off: if in SYN-RECEIVED state or SENDSYN * flag is on (half-synchronized state), then queue data for * later processing; else drop segment and return. */ if ((thflags & TH_ACK) == 0) { if (tp->t_state == TCPS_SYN_RECEIVED || (tp->t_flags & TF_NEEDSYN)) goto step6; else if (tp->t_flags & TF_ACKNOW) goto dropafterack; else goto drop; } /* * Ack processing. */ switch (tp->t_state) { /* * In SYN_RECEIVED state, the ack ACKs our SYN, so enter * ESTABLISHED state and continue processing. * The ACK was checked above. */ case TCPS_SYN_RECEIVED: tcpstat.tcps_connects++; soisconnected(so); /* Do window scaling? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->rcv_scale = tp->request_r_scale; tp->snd_wnd = tiwin; } /* * Make transitions: * SYN-RECEIVED -> ESTABLISHED * SYN-RECEIVED* -> FIN-WAIT-1 */ tp->t_starttime = ticks; if (tp->t_flags & TF_NEEDFIN) { tp->t_state = TCPS_FIN_WAIT_1; tp->t_flags &= ~TF_NEEDFIN; } else { tp->t_state = TCPS_ESTABLISHED; tcp_timer_activate(tp, TT_KEEP, tcp_keepidle); } /* * If segment contains data or ACK, will call tcp_reass() * later; if not, do so now to pass queued data to user. */ if (tlen == 0 && (thflags & TH_FIN) == 0) (void) tcp_reass(tp, (struct tcphdr *)0, 0, (struct mbuf *)0); tp->snd_wl1 = th->th_seq - 1; /* FALLTHROUGH */ /* * In ESTABLISHED state: drop duplicate ACKs; ACK out of range * ACKs. If the ack is in the range * tp->snd_una < th->th_ack <= tp->snd_max * then advance tp->snd_una to th->th_ack and drop * data from the retransmission queue. If this ACK reflects * more up to date window information we update our window information. */ case TCPS_ESTABLISHED: case TCPS_FIN_WAIT_1: case TCPS_FIN_WAIT_2: case TCPS_CLOSE_WAIT: case TCPS_CLOSING: case TCPS_LAST_ACK: case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); if (SEQ_GT(th->th_ack, tp->snd_max)) { tcpstat.tcps_rcvacktoomuch++; goto dropafterack; } if (tp->sack_enable && ((to.to_flags & TOF_SACK) || !TAILQ_EMPTY(&tp->snd_holes))) tcp_sack_doack(tp, &to, th->th_ack); if (SEQ_LEQ(th->th_ack, tp->snd_una)) { if (tlen == 0 && tiwin == tp->snd_wnd) { tcpstat.tcps_rcvdupack++; /* * If we have outstanding data (other than * a window probe), this is a completely * duplicate ack (ie, window info didn't * change), the ack is the biggest we've * seen and we've seen exactly our rexmt * threshhold of them, assume a packet * has been dropped and retransmit it. * Kludge snd_nxt & the congestion * window so we send only this one * packet. * * We know we're losing at the current * window size so do congestion avoidance * (set ssthresh to half the current window * and pull our congestion window back to * the new ssthresh). * * Dup acks mean that packets have left the * network (they're now cached at the receiver) * so bump cwnd by the amount in the receiver * to keep a constant cwnd packets in the * network. */ if (!tcp_timer_active(tp, TT_REXMT) || th->th_ack != tp->snd_una) tp->t_dupacks = 0; else if (++tp->t_dupacks > tcprexmtthresh || ((tcp_do_newreno || tp->sack_enable) && IN_FASTRECOVERY(tp))) { if (tp->sack_enable && IN_FASTRECOVERY(tp)) { int awnd; /* * Compute the amount of data in flight first. * We can inject new data into the pipe iff * we have less than 1/2 the original window's * worth of data in flight. */ awnd = (tp->snd_nxt - tp->snd_fack) + tp->sackhint.sack_bytes_rexmit; if (awnd < tp->snd_ssthresh) { tp->snd_cwnd += tp->t_maxseg; if (tp->snd_cwnd > tp->snd_ssthresh) tp->snd_cwnd = tp->snd_ssthresh; } } else tp->snd_cwnd += tp->t_maxseg; (void) tcp_output(tp); goto drop; } else if (tp->t_dupacks == tcprexmtthresh) { tcp_seq onxt = tp->snd_nxt; u_int win; /* * If we're doing sack, check to * see if we're already in sack * recovery. If we're not doing sack, * check to see if we're in newreno * recovery. */ if (tp->sack_enable) { if (IN_FASTRECOVERY(tp)) { tp->t_dupacks = 0; break; } } else if (tcp_do_newreno) { if (SEQ_LEQ(th->th_ack, tp->snd_recover)) { tp->t_dupacks = 0; break; } } win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; if (win < 2) win = 2; tp->snd_ssthresh = win * tp->t_maxseg; ENTER_FASTRECOVERY(tp); tp->snd_recover = tp->snd_max; tcp_timer_activate(tp, TT_REXMT, 0); tp->t_rtttime = 0; if (tp->sack_enable) { tcpstat.tcps_sack_recovery_episode++; tp->sack_newdata = tp->snd_nxt; tp->snd_cwnd = tp->t_maxseg; (void) tcp_output(tp); goto drop; } tp->snd_nxt = th->th_ack; tp->snd_cwnd = tp->t_maxseg; (void) tcp_output(tp); KASSERT(tp->snd_limited <= 2, ("%s: tp->snd_limited too big", __func__)); tp->snd_cwnd = tp->snd_ssthresh + tp->t_maxseg * (tp->t_dupacks - tp->snd_limited); if (SEQ_GT(onxt, tp->snd_nxt)) tp->snd_nxt = onxt; goto drop; } else if (tcp_do_rfc3042) { u_long oldcwnd = tp->snd_cwnd; tcp_seq oldsndmax = tp->snd_max; u_int sent; KASSERT(tp->t_dupacks == 1 || tp->t_dupacks == 2, ("%s: dupacks not 1 or 2", __func__)); if (tp->t_dupacks == 1) tp->snd_limited = 0; tp->snd_cwnd = (tp->snd_nxt - tp->snd_una) + (tp->t_dupacks - tp->snd_limited) * tp->t_maxseg; (void) tcp_output(tp); sent = tp->snd_max - oldsndmax; if (sent > tp->t_maxseg) { KASSERT((tp->t_dupacks == 2 && tp->snd_limited == 0) || (sent == tp->t_maxseg + 1 && tp->t_flags & TF_SENTFIN), ("%s: sent too much", __func__)); tp->snd_limited = 2; } else if (sent > 0) ++tp->snd_limited; tp->snd_cwnd = oldcwnd; goto drop; } } else tp->t_dupacks = 0; break; } KASSERT(SEQ_GT(th->th_ack, tp->snd_una), ("%s: th_ack <= snd_una", __func__)); /* * If the congestion window was inflated to account * for the other side's cached packets, retract it. */ if (tcp_do_newreno || tp->sack_enable) { if (IN_FASTRECOVERY(tp)) { if (SEQ_LT(th->th_ack, tp->snd_recover)) { if (tp->sack_enable) tcp_sack_partialack(tp, th); else tcp_newreno_partial_ack(tp, th); } else { /* * Out of fast recovery. * Window inflation should have left us * with approximately snd_ssthresh * outstanding data. * But in case we would be inclined to * send a burst, better to do it via * the slow start mechanism. */ if (SEQ_GT(th->th_ack + tp->snd_ssthresh, tp->snd_max)) tp->snd_cwnd = tp->snd_max - th->th_ack + tp->t_maxseg; else tp->snd_cwnd = tp->snd_ssthresh; } } } else { if (tp->t_dupacks >= tcprexmtthresh && tp->snd_cwnd > tp->snd_ssthresh) tp->snd_cwnd = tp->snd_ssthresh; } tp->t_dupacks = 0; /* * If we reach this point, ACK is not a duplicate, * i.e., it ACKs something we sent. */ if (tp->t_flags & TF_NEEDSYN) { /* * T/TCP: Connection was half-synchronized, and our * SYN has been ACK'd (so connection is now fully * synchronized). Go to non-starred state, * increment snd_una for ACK of SYN, and check if * we can do window scaling. */ tp->t_flags &= ~TF_NEEDSYN; tp->snd_una++; /* Do window scaling? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->rcv_scale = tp->request_r_scale; /* Send window already scaled. */ } } process_ACK: KASSERT(headlocked, ("%s: process_ACK: head not locked", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); acked = th->th_ack - tp->snd_una; tcpstat.tcps_rcvackpack++; tcpstat.tcps_rcvackbyte += acked; /* * If we just performed our first retransmit, and the ACK * arrives within our recovery window, then it was a mistake * to do the retransmit in the first place. Recover our * original cwnd and ssthresh, and proceed to transmit where * we left off. */ if (tp->t_rxtshift == 1 && ticks < tp->t_badrxtwin) { ++tcpstat.tcps_sndrexmitbad; tp->snd_cwnd = tp->snd_cwnd_prev; tp->snd_ssthresh = tp->snd_ssthresh_prev; tp->snd_recover = tp->snd_recover_prev; if (tp->t_flags & TF_WASFRECOVERY) ENTER_FASTRECOVERY(tp); tp->snd_nxt = tp->snd_max; tp->t_badrxtwin = 0; /* XXX probably not required */ } /* * If we have a timestamp reply, update smoothed * round trip time. If no timestamp is present but * transmit timer is running and timed sequence * number was acked, update smoothed round trip time. * Since we now have an rtt measurement, cancel the * timer backoff (cf., Phil Karn's retransmit alg.). * Recompute the initial retransmit timer. * * Some boxes send broken timestamp replies * during the SYN+ACK phase, ignore * timestamps of 0 or we could calculate a * huge RTT and blow up the retransmit timer. */ if ((to.to_flags & TOF_TS) != 0 && to.to_tsecr) { if (!tp->t_rttlow || tp->t_rttlow > ticks - to.to_tsecr) tp->t_rttlow = ticks - to.to_tsecr; tcp_xmit_timer(tp, ticks - to.to_tsecr + 1); } else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) { if (!tp->t_rttlow || tp->t_rttlow > ticks - tp->t_rtttime) tp->t_rttlow = ticks - tp->t_rtttime; tcp_xmit_timer(tp, ticks - tp->t_rtttime); } tcp_xmit_bandwidth_limit(tp, th->th_ack); /* * If all outstanding data is acked, stop retransmit * timer and remember to restart (more output or persist). * If there is more data to be acked, restart retransmit * timer, using current (possibly backed-off) value. */ if (th->th_ack == tp->snd_max) { tcp_timer_activate(tp, TT_REXMT, 0); needoutput = 1; } else if (!tcp_timer_active(tp, TT_PERSIST)) tcp_timer_activate(tp, TT_REXMT, tp->t_rxtcur); /* * If no data (only SYN) was ACK'd, * skip rest of ACK processing. */ if (acked == 0) goto step6; /* * When new data is acked, open the congestion window. * If the window gives us less than ssthresh packets * in flight, open exponentially (maxseg per packet). * Otherwise open linearly: maxseg per window * (maxseg^2 / cwnd per packet). */ if ((!tcp_do_newreno && !tp->sack_enable) || !IN_FASTRECOVERY(tp)) { u_int cw = tp->snd_cwnd; u_int incr = tp->t_maxseg; if (cw > tp->snd_ssthresh) incr = incr * incr / cw; tp->snd_cwnd = min(cw+incr, TCP_MAXWIN<snd_scale); } SOCKBUF_LOCK(&so->so_snd); if (acked > so->so_snd.sb_cc) { tp->snd_wnd -= so->so_snd.sb_cc; sbdrop_locked(&so->so_snd, (int)so->so_snd.sb_cc); ourfinisacked = 1; } else { sbdrop_locked(&so->so_snd, acked); tp->snd_wnd -= acked; ourfinisacked = 0; } sowwakeup_locked(so); /* detect una wraparound */ if ((tcp_do_newreno || tp->sack_enable) && !IN_FASTRECOVERY(tp) && SEQ_GT(tp->snd_una, tp->snd_recover) && SEQ_LEQ(th->th_ack, tp->snd_recover)) tp->snd_recover = th->th_ack - 1; if ((tcp_do_newreno || tp->sack_enable) && IN_FASTRECOVERY(tp) && SEQ_GEQ(th->th_ack, tp->snd_recover)) EXIT_FASTRECOVERY(tp); tp->snd_una = th->th_ack; if (tp->sack_enable) { if (SEQ_GT(tp->snd_una, tp->snd_recover)) tp->snd_recover = tp->snd_una; } if (SEQ_LT(tp->snd_nxt, tp->snd_una)) tp->snd_nxt = tp->snd_una; switch (tp->t_state) { /* * In FIN_WAIT_1 STATE in addition to the processing * for the ESTABLISHED state if our FIN is now acknowledged * then enter FIN_WAIT_2. */ case TCPS_FIN_WAIT_1: if (ourfinisacked) { /* * If we can't receive any more * data, then closing user can proceed. * Starting the timer is contrary to the * specification, but if we don't get a FIN * we'll hang forever. */ /* XXXjl * we should release the tp also, and use a * compressed state. */ if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { int timeout; soisdisconnected(so); timeout = (tcp_fast_finwait2_recycle) ? tcp_finwait2_timeout : tcp_maxidle; tcp_timer_activate(tp, TT_2MSL, timeout); } tp->t_state = TCPS_FIN_WAIT_2; } break; /* * In CLOSING STATE in addition to the processing for * the ESTABLISHED state if the ACK acknowledges our FIN * then enter the TIME-WAIT state, otherwise ignore * the segment. */ case TCPS_CLOSING: if (ourfinisacked) { KASSERT(headlocked, ("%s: process_ACK: " "head not locked", __func__)); tcp_twstart(tp); INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; m_freem(m); return (0); } break; /* * In LAST_ACK, we may still be waiting for data to drain * and/or to be acked, as well as for the ack of our FIN. * If our FIN is now acknowledged, delete the TCB, * enter the closed state and return. */ case TCPS_LAST_ACK: if (ourfinisacked) { KASSERT(headlocked, ("%s: process_ACK: " "tcp_close: head not locked", __func__)); tp = tcp_close(tp); goto drop; } break; /* * In TIME_WAIT state the only thing that should arrive * is a retransmission of the remote FIN. Acknowledge * it and restart the finack timer. */ case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); tcp_timer_activate(tp, TT_2MSL, 2 * tcp_msl); goto dropafterack; } } step6: KASSERT(headlocked, ("%s: step6: head not locked", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); /* * Update window information. * Don't look at window if no ACK: TAC's send garbage on first SYN. */ if ((thflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, th->th_seq) || (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) || (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) { /* keep track of pure window updates */ if (tlen == 0 && tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) tcpstat.tcps_rcvwinupd++; tp->snd_wnd = tiwin; tp->snd_wl1 = th->th_seq; tp->snd_wl2 = th->th_ack; if (tp->snd_wnd > tp->max_sndwnd) tp->max_sndwnd = tp->snd_wnd; needoutput = 1; } /* * Process segments with URG. */ if ((thflags & TH_URG) && th->th_urp && TCPS_HAVERCVDFIN(tp->t_state) == 0) { /* * This is a kludge, but if we receive and accept * random urgent pointers, we'll crash in * soreceive. It's hard to imagine someone * actually wanting to send this much urgent data. */ SOCKBUF_LOCK(&so->so_rcv); if (th->th_urp + so->so_rcv.sb_cc > sb_max) { th->th_urp = 0; /* XXX */ thflags &= ~TH_URG; /* XXX */ SOCKBUF_UNLOCK(&so->so_rcv); /* XXX */ goto dodata; /* XXX */ } /* * If this segment advances the known urgent pointer, * then mark the data stream. This should not happen * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since * a FIN has been received from the remote side. * In these states we ignore the URG. * * 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 as the original * spec states (in one of two places). */ if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) { tp->rcv_up = th->th_seq + th->th_urp; so->so_oobmark = so->so_rcv.sb_cc + (tp->rcv_up - tp->rcv_nxt) - 1; if (so->so_oobmark == 0) so->so_rcv.sb_state |= SBS_RCVATMARK; sohasoutofband(so); tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); } SOCKBUF_UNLOCK(&so->so_rcv); /* * Remove out of band data so doesn't get presented to user. * This can happen independent of advancing the URG pointer, * but if two URG's are pending at once, some out-of-band * data may creep in... ick. */ if (th->th_urp <= (u_long)tlen && !(so->so_options & SO_OOBINLINE)) { /* hdr drop is delayed */ tcp_pulloutofband(so, th, m, drop_hdrlen); } } else { /* * If no out of band data is expected, * pull receive urgent pointer along * with the receive window. */ if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) tp->rcv_up = tp->rcv_nxt; } dodata: /* XXX */ KASSERT(headlocked, ("%s: dodata: head not locked", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); /* * Process the segment text, merging it into the TCP sequencing queue, * and arranging for acknowledgment of receipt if necessary. * This process logically involves adjusting tp->rcv_wnd as data * is presented to the user (this happens in tcp_usrreq.c, * case PRU_RCVD). If a FIN has already been received on this * connection then we just ignore the text. */ if ((tlen || (thflags & TH_FIN)) && TCPS_HAVERCVDFIN(tp->t_state) == 0) { tcp_seq save_start = th->th_seq; tcp_seq save_end = th->th_seq + tlen; m_adj(m, drop_hdrlen); /* delayed header drop */ /* * Insert segment which includes th into TCP reassembly queue * with control block tp. Set thflags to whether reassembly now * includes a segment with FIN. This handles the common case * inline (segment is the next to be received on an established * connection, and the queue is empty), avoiding linkage into * and removal from the queue and repetition of various * conversions. * Set DELACK for segments received in order, but ack * immediately when segments are out of order (so * fast retransmit can work). */ if (th->th_seq == tp->rcv_nxt && LIST_EMPTY(&tp->t_segq) && TCPS_HAVEESTABLISHED(tp->t_state)) { if (DELAY_ACK(tp)) tp->t_flags |= TF_DELACK; else tp->t_flags |= TF_ACKNOW; tp->rcv_nxt += tlen; thflags = th->th_flags & TH_FIN; tcpstat.tcps_rcvpack++; tcpstat.tcps_rcvbyte += tlen; ND6_HINT(tp); SOCKBUF_LOCK(&so->so_rcv); if (so->so_rcv.sb_state & SBS_CANTRCVMORE) m_freem(m); else sbappendstream_locked(&so->so_rcv, m); sorwakeup_locked(so); } else { thflags = tcp_reass(tp, th, &tlen, m); tp->t_flags |= TF_ACKNOW; } if (tlen > 0 && tp->sack_enable) tcp_update_sack_list(tp, save_start, save_end); #if 0 /* * Note the amount of data that peer has sent into * our window, in order to estimate the sender's * buffer size. * XXX: Unused. */ len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); #endif } else { m_freem(m); thflags &= ~TH_FIN; } /* * If FIN is received ACK the FIN and let the user know * that the connection is closing. */ if (thflags & TH_FIN) { if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { socantrcvmore(so); /* * If connection is half-synchronized * (ie NEEDSYN flag on) then delay ACK, * so it may be piggybacked when SYN is sent. * Otherwise, since we received a FIN then no * more input can be expected, send ACK now. */ if (tp->t_flags & TF_NEEDSYN) tp->t_flags |= TF_DELACK; else tp->t_flags |= TF_ACKNOW; tp->rcv_nxt++; } switch (tp->t_state) { /* * In SYN_RECEIVED and ESTABLISHED STATES * enter the CLOSE_WAIT state. */ case TCPS_SYN_RECEIVED: tp->t_starttime = ticks; /*FALLTHROUGH*/ case TCPS_ESTABLISHED: tp->t_state = TCPS_CLOSE_WAIT; break; /* * If still in FIN_WAIT_1 STATE FIN has not been acked so * enter the CLOSING state. */ case TCPS_FIN_WAIT_1: tp->t_state = TCPS_CLOSING; break; /* * In FIN_WAIT_2 state enter the TIME_WAIT state, * starting the time-wait timer, turning off the other * standard timers. */ case TCPS_FIN_WAIT_2: KASSERT(headlocked == 1, ("%s: dodata: " "TCP_FIN_WAIT_2: head not locked", __func__)); tcp_twstart(tp); INP_INFO_WUNLOCK(&tcbinfo); return (0); /* * In TIME_WAIT state restart the 2 MSL time_wait timer. */ case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("%s: timewait", __func__)); tcp_timer_activate(tp, TT_2MSL, 2 * tcp_msl); break; } } INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif /* * Return any desired output. */ if (needoutput || (tp->t_flags & TF_ACKNOW)) (void) tcp_output(tp); check_delack: KASSERT(headlocked == 0, ("%s: check_delack: head locked", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); if (tp->t_flags & TF_DELACK) { tp->t_flags &= ~TF_DELACK; tcp_timer_activate(tp, TT_DELACK, tcp_delacktime); } INP_UNLOCK(tp->t_inpcb); return (0); dropafterack: KASSERT(headlocked, ("%s: dropafterack: head not locked", __func__)); /* * Generate an ACK dropping incoming segment if it occupies * sequence space, where the ACK reflects our state. * * We can now skip the test for the RST flag since all * paths to this code happen after packets containing * RST have been dropped. * * In the SYN-RECEIVED state, don't send an ACK unless the * segment we received passes the SYN-RECEIVED ACK test. * If it fails send a RST. This breaks the loop in the * "LAND" DoS attack, and also prevents an ACK storm * between two listening ports that have been sent forged * SYN segments, each with the source address of the other. */ if (tp->t_state == TCPS_SYN_RECEIVED && (thflags & TH_ACK) && (SEQ_GT(tp->snd_una, th->th_ack) || SEQ_GT(th->th_ack, tp->snd_max)) ) { rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } #ifdef TCPDEBUG if (so->so_options & SO_DEBUG) tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif KASSERT(headlocked, ("%s: headlocked should be 1", __func__)); INP_INFO_WUNLOCK(&tcbinfo); tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); INP_UNLOCK(tp->t_inpcb); m_freem(m); return (0); dropwithreset: KASSERT(headlocked, ("%s: dropwithreset: head not locked", __func__)); tcp_dropwithreset(m, th, tp, tlen, rstreason); if (tp != NULL) INP_UNLOCK(tp->t_inpcb); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); return (0); drop: /* * Drop space held by incoming segment and return. */ #ifdef TCPDEBUG if (tp == NULL || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif if (tp != NULL) INP_UNLOCK(tp->t_inpcb); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); m_freem(m); return (0); } /* * Issue RST on TCP segment. The mbuf must still include the original * packet header. */ static void tcp_dropwithreset(struct mbuf *m, struct tcphdr *th, struct tcpcb *tp, int tlen, int rstreason) { struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; #endif /* * Generate a RST, dropping incoming segment. * Make ACK acceptable to originator of segment. * Don't bother to respond if destination was broadcast/multicast. * tp may be NULL. */ if ((th->th_flags & TH_RST) || m->m_flags & (M_BCAST|M_MCAST)) goto drop; #ifdef INET6 if (mtod(m, struct ip *)->ip_v == 6) { ip6 = mtod(m, struct ip6_hdr *); if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto drop; /* IPv6 anycast check is done at tcp6_input() */ } else #endif { ip = mtod(m, struct ip *); if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || ip->ip_src.s_addr == htonl(INADDR_BROADCAST) || in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) goto drop; } /* Perform bandwidth limiting. */ if (badport_bandlim(rstreason) < 0) goto drop; /* tcp_respond consumes the mbuf chain. */ if (th->th_flags & TH_ACK) { tcp_respond(tp, mtod(m, void *), th, m, (tcp_seq)0, th->th_ack, TH_RST); } else { if (th->th_flags & TH_SYN) tlen++; tcp_respond(tp, mtod(m, void *), th, m, th->th_seq+tlen, (tcp_seq)0, TH_RST|TH_ACK); } return; drop: m_freem(m); return; } /* * Parse TCP options and place in tcpopt. */ static void tcp_dooptions(struct tcpopt *to, u_char *cp, int cnt, int flags) { int opt, optlen; to->to_flags = 0; for (; cnt > 0; cnt -= optlen, cp += optlen) { opt = cp[0]; if (opt == TCPOPT_EOL) break; if (opt == TCPOPT_NOP) optlen = 1; else { if (cnt < 2) break; optlen = cp[1]; if (optlen < 2 || optlen > cnt) break; } switch (opt) { case TCPOPT_MAXSEG: if (optlen != TCPOLEN_MAXSEG) continue; if (!(flags & TO_SYN)) continue; to->to_flags |= TOF_MSS; bcopy((char *)cp + 2, (char *)&to->to_mss, sizeof(to->to_mss)); to->to_mss = ntohs(to->to_mss); break; case TCPOPT_WINDOW: if (optlen != TCPOLEN_WINDOW) continue; if (!(flags & TO_SYN)) continue; to->to_flags |= TOF_SCALE; to->to_wscale = min(cp[2], TCP_MAX_WINSHIFT); break; case TCPOPT_TIMESTAMP: if (optlen != TCPOLEN_TIMESTAMP) continue; to->to_flags |= TOF_TS; bcopy((char *)cp + 2, (char *)&to->to_tsval, sizeof(to->to_tsval)); to->to_tsval = ntohl(to->to_tsval); bcopy((char *)cp + 6, (char *)&to->to_tsecr, sizeof(to->to_tsecr)); to->to_tsecr = ntohl(to->to_tsecr); break; #ifdef TCP_SIGNATURE /* * XXX In order to reply to a host which has set the * TCP_SIGNATURE option in its initial SYN, we have to * record the fact that the option was observed here * for the syncache code to perform the correct response. */ case TCPOPT_SIGNATURE: if (optlen != TCPOLEN_SIGNATURE) continue; to->to_flags |= (TOF_SIGNATURE | TOF_SIGLEN); break; #endif case TCPOPT_SACK_PERMITTED: if (optlen != TCPOLEN_SACK_PERMITTED) continue; if (!(flags & TO_SYN)) continue; if (!tcp_do_sack) continue; to->to_flags |= TOF_SACKPERM; break; case TCPOPT_SACK: if (optlen <= 2 || (optlen - 2) % TCPOLEN_SACK != 0) continue; if (flags & TO_SYN) continue; to->to_flags |= TOF_SACK; to->to_nsacks = (optlen - 2) / TCPOLEN_SACK; to->to_sacks = cp + 2; tcpstat.tcps_sack_rcv_blocks++; break; default: continue; } } } /* * Pull out of band byte out of a segment so * it doesn't appear in the user's data queue. * It is still reflected in the segment length for * sequencing purposes. */ static void tcp_pulloutofband(struct socket *so, struct tcphdr *th, struct mbuf *m, int off) { int cnt = off + th->th_urp - 1; while (cnt >= 0) { if (m->m_len > cnt) { char *cp = mtod(m, caddr_t) + cnt; struct tcpcb *tp = sototcpcb(so); tp->t_iobc = *cp; tp->t_oobflags |= TCPOOB_HAVEDATA; bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1)); m->m_len--; if (m->m_flags & M_PKTHDR) m->m_pkthdr.len--; return; } cnt -= m->m_len; m = m->m_next; if (m == NULL) break; } panic("tcp_pulloutofband"); } /* * Collect new round-trip time estimate * and update averages and current timeout. */ static void tcp_xmit_timer(struct tcpcb *tp, int rtt) { int delta; INP_LOCK_ASSERT(tp->t_inpcb); tcpstat.tcps_rttupdated++; tp->t_rttupdated++; if (tp->t_srtt != 0) { /* * srtt is stored as fixed point with 5 bits after the * binary point (i.e., scaled by 8). The following magic * is equivalent to the smoothing algorithm in rfc793 with * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed * point). Adjust rtt to origin 0. */ delta = ((rtt - 1) << TCP_DELTA_SHIFT) - (tp->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)); if ((tp->t_srtt += delta) <= 0) tp->t_srtt = 1; /* * We accumulate a smoothed rtt variance (actually, a * smoothed mean difference), then set the retransmit * timer to smoothed rtt + 4 times the smoothed variance. * rttvar is stored as fixed point with 4 bits after the * binary point (scaled by 16). The following is * equivalent to rfc793 smoothing with an alpha of .75 * (rttvar = rttvar*3/4 + |delta| / 4). This replaces * rfc793's wired-in beta. */ if (delta < 0) delta = -delta; delta -= tp->t_rttvar >> (TCP_RTTVAR_SHIFT - TCP_DELTA_SHIFT); if ((tp->t_rttvar += delta) <= 0) tp->t_rttvar = 1; if (tp->t_rttbest > tp->t_srtt + tp->t_rttvar) tp->t_rttbest = tp->t_srtt + tp->t_rttvar; } else { /* * No rtt measurement yet - use the unsmoothed rtt. * Set the variance to half the rtt (so our first * retransmit happens at 3*rtt). */ tp->t_srtt = rtt << TCP_RTT_SHIFT; tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1); tp->t_rttbest = tp->t_srtt + tp->t_rttvar; } tp->t_rtttime = 0; tp->t_rxtshift = 0; /* * the retransmit should happen at rtt + 4 * rttvar. * Because of the way we do the smoothing, srtt and rttvar * will each average +1/2 tick of bias. When we compute * the retransmit timer, we want 1/2 tick of rounding and * 1 extra tick because of +-1/2 tick uncertainty in the * firing of the timer. The bias will give us exactly the * 1.5 tick we need. But, because the bias is * statistical, we have to test that we don't drop below * the minimum feasible timer (which is 2 ticks). */ TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), max(tp->t_rttmin, rtt + 2), TCPTV_REXMTMAX); /* * We received an ack for a packet that wasn't retransmitted; * it is probably safe to discard any error indications we've * received recently. This isn't quite right, but close enough * for now (a route might have failed after we sent a segment, * and the return path might not be symmetrical). */ tp->t_softerror = 0; } /* * Determine a reasonable value for maxseg size. * If the route is known, check route for mtu. * If none, use an mss that can be handled on the outgoing * interface without forcing IP to fragment; if bigger than * an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES * to utilize large mbufs. If no route is found, route has no mtu, * or the destination isn't local, use a default, hopefully conservative * size (usually 512 or the default IP max size, but no more than the mtu * of the interface), as we can't discover anything about intervening * gateways or networks. We also initialize the congestion/slow start * window to be a single segment if the destination isn't local. * While looking at the routing entry, we also initialize other path-dependent * parameters from pre-set or cached values in the routing entry. * * Also take into account the space needed for options that we * send regularly. Make maxseg shorter by that amount to assure * that we can send maxseg amount of data even when the options * are present. Store the upper limit of the length of options plus * data in maxopd. * * * In case of T/TCP, we call this routine during implicit connection * setup as well (offer = -1), to initialize maxseg from the cached * MSS of our peer. * * NOTE that this routine is only called when we process an incoming * segment. Outgoing SYN/ACK MSS settings are handled in tcp_mssopt(). */ void tcp_mss(struct tcpcb *tp, int offer) { int rtt, mss; u_long bufsize; u_long maxmtu; struct inpcb *inp = tp->t_inpcb; struct socket *so; struct hc_metrics_lite metrics; int origoffer = offer; int mtuflags = 0; #ifdef INET6 int isipv6 = ((inp->inp_vflag & INP_IPV6) != 0) ? 1 : 0; size_t min_protoh = isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : sizeof (struct tcpiphdr); #else const size_t min_protoh = sizeof(struct tcpiphdr); #endif /* initialize */ #ifdef INET6 if (isipv6) { maxmtu = tcp_maxmtu6(&inp->inp_inc, &mtuflags); tp->t_maxopd = tp->t_maxseg = tcp_v6mssdflt; } else #endif { maxmtu = tcp_maxmtu(&inp->inp_inc, &mtuflags); tp->t_maxopd = tp->t_maxseg = tcp_mssdflt; } so = inp->inp_socket; /* * no route to sender, stay with default mss and return */ if (maxmtu == 0) return; /* what have we got? */ switch (offer) { case 0: /* * Offer == 0 means that there was no MSS on the SYN * segment, in this case we use tcp_mssdflt. */ offer = #ifdef INET6 isipv6 ? tcp_v6mssdflt : #endif tcp_mssdflt; break; case -1: /* * Offer == -1 means that we didn't receive SYN yet. */ /* FALLTHROUGH */ default: /* * Prevent DoS attack with too small MSS. Round up * to at least minmss. */ offer = max(offer, tcp_minmss); /* * Sanity check: make sure that maxopd will be large * enough to allow some data on segments even if the * all the option space is used (40bytes). Otherwise * funny things may happen in tcp_output. */ offer = max(offer, 64); } /* * rmx information is now retrieved from tcp_hostcache */ tcp_hc_get(&inp->inp_inc, &metrics); /* * if there's a discovered mtu int tcp hostcache, use it * else, use the link mtu. */ if (metrics.rmx_mtu) mss = min(metrics.rmx_mtu, maxmtu) - min_protoh; else { #ifdef INET6 if (isipv6) { mss = maxmtu - min_protoh; if (!path_mtu_discovery && !in6_localaddr(&inp->in6p_faddr)) mss = min(mss, tcp_v6mssdflt); } else #endif { mss = maxmtu - min_protoh; if (!path_mtu_discovery && !in_localaddr(inp->inp_faddr)) mss = min(mss, tcp_mssdflt); } } mss = min(mss, offer); /* * maxopd stores the maximum length of data AND options * in a segment; maxseg is the amount of data in a normal * segment. We need to store this value (maxopd) apart * from maxseg, because now every segment carries options * and thus we normally have somewhat less data in segments. */ tp->t_maxopd = mss; /* * origoffer==-1 indicates, that no segments were received yet. * In this case we just guess. */ if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && (origoffer == -1 || (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP)) mss -= TCPOLEN_TSTAMP_APPA; tp->t_maxseg = mss; #if (MCLBYTES & (MCLBYTES - 1)) == 0 if (mss > MCLBYTES) mss &= ~(MCLBYTES-1); #else if (mss > MCLBYTES) mss = mss / MCLBYTES * MCLBYTES; #endif tp->t_maxseg = mss; /* * If there's a pipesize, change the socket buffer to that size, * don't change if sb_hiwat is different than default (then it * has been changed on purpose with setsockopt). * Make the socket buffers an integral number of mss units; * if the mss is larger than the socket buffer, decrease the mss. */ SOCKBUF_LOCK(&so->so_snd); if ((so->so_snd.sb_hiwat == tcp_sendspace) && metrics.rmx_sendpipe) bufsize = metrics.rmx_sendpipe; else bufsize = so->so_snd.sb_hiwat; if (bufsize < mss) mss = bufsize; else { bufsize = roundup(bufsize, mss); if (bufsize > sb_max) bufsize = sb_max; if (bufsize > so->so_snd.sb_hiwat) (void)sbreserve_locked(&so->so_snd, bufsize, so, NULL); } SOCKBUF_UNLOCK(&so->so_snd); tp->t_maxseg = mss; SOCKBUF_LOCK(&so->so_rcv); if ((so->so_rcv.sb_hiwat == tcp_recvspace) && metrics.rmx_recvpipe) bufsize = metrics.rmx_recvpipe; else bufsize = so->so_rcv.sb_hiwat; if (bufsize > mss) { bufsize = roundup(bufsize, mss); if (bufsize > sb_max) bufsize = sb_max; if (bufsize > so->so_rcv.sb_hiwat) (void)sbreserve_locked(&so->so_rcv, bufsize, so, NULL); } SOCKBUF_UNLOCK(&so->so_rcv); /* * While we're here, check the others too */ if (tp->t_srtt == 0 && (rtt = metrics.rmx_rtt)) { tp->t_srtt = rtt; tp->t_rttbest = tp->t_srtt + TCP_RTT_SCALE; tcpstat.tcps_usedrtt++; if (metrics.rmx_rttvar) { tp->t_rttvar = metrics.rmx_rttvar; tcpstat.tcps_usedrttvar++; } else { /* default variation is +- 1 rtt */ tp->t_rttvar = tp->t_srtt * TCP_RTTVAR_SCALE / TCP_RTT_SCALE; } TCPT_RANGESET(tp->t_rxtcur, ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1, tp->t_rttmin, TCPTV_REXMTMAX); } if (metrics.rmx_ssthresh) { /* * There's some sort of gateway or interface * buffer limit on the path. Use this to set * the slow start threshhold, but set the * threshold to no less than 2*mss. */ tp->snd_ssthresh = max(2 * mss, metrics.rmx_ssthresh); tcpstat.tcps_usedssthresh++; } if (metrics.rmx_bandwidth) tp->snd_bandwidth = metrics.rmx_bandwidth; /* * Set the slow-start flight size depending on whether this * is a local network or not. * * Extend this so we cache the cwnd too and retrieve it here. * Make cwnd even bigger than RFC3390 suggests but only if we * have previous experience with the remote host. Be careful * not make cwnd bigger than remote receive window or our own * send socket buffer. Maybe put some additional upper bound * on the retrieved cwnd. Should do incremental updates to * hostcache when cwnd collapses so next connection doesn't * overloads the path again. * * RFC3390 says only do this if SYN or SYN/ACK didn't got lost. * We currently check only in syncache_socket for that. */ #define TCP_METRICS_CWND #ifdef TCP_METRICS_CWND if (metrics.rmx_cwnd) tp->snd_cwnd = max(mss, min(metrics.rmx_cwnd / 2, min(tp->snd_wnd, so->so_snd.sb_hiwat))); else #endif if (tcp_do_rfc3390) tp->snd_cwnd = min(4 * mss, max(2 * mss, 4380)); #ifdef INET6 else if ((isipv6 && in6_localaddr(&inp->in6p_faddr)) || (!isipv6 && in_localaddr(inp->inp_faddr))) #else else if (in_localaddr(inp->inp_faddr)) #endif tp->snd_cwnd = mss * ss_fltsz_local; else tp->snd_cwnd = mss * ss_fltsz; /* Check the interface for TSO capabilities. */ if (mtuflags & CSUM_TSO) tp->t_flags |= TF_TSO; } /* * Determine the MSS option to send on an outgoing SYN. */ int tcp_mssopt(struct in_conninfo *inc) { int mss = 0; u_long maxmtu = 0; u_long thcmtu = 0; size_t min_protoh; #ifdef INET6 int isipv6 = inc->inc_isipv6 ? 1 : 0; #endif KASSERT(inc != NULL, ("tcp_mssopt with NULL in_conninfo pointer")); #ifdef INET6 if (isipv6) { mss = tcp_v6mssdflt; maxmtu = tcp_maxmtu6(inc, NULL); thcmtu = tcp_hc_getmtu(inc); /* IPv4 and IPv6 */ min_protoh = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); } else #endif { mss = tcp_mssdflt; maxmtu = tcp_maxmtu(inc, NULL); thcmtu = tcp_hc_getmtu(inc); /* IPv4 and IPv6 */ min_protoh = sizeof(struct tcpiphdr); } if (maxmtu && thcmtu) mss = min(maxmtu, thcmtu) - min_protoh; else if (maxmtu || thcmtu) mss = max(maxmtu, thcmtu) - min_protoh; return (mss); } /* * On a partial ack arrives, force the retransmission of the * next unacknowledged segment. Do not clear tp->t_dupacks. * By setting snd_nxt to ti_ack, this forces retransmission timer to * be started again. */ static void tcp_newreno_partial_ack(struct tcpcb *tp, struct tcphdr *th) { tcp_seq onxt = tp->snd_nxt; u_long ocwnd = tp->snd_cwnd; tcp_timer_activate(tp, TT_REXMT, 0); tp->t_rtttime = 0; tp->snd_nxt = th->th_ack; /* * Set snd_cwnd to one segment beyond acknowledged offset. * (tp->snd_una has not yet been updated when this function is called.) */ tp->snd_cwnd = tp->t_maxseg + (th->th_ack - tp->snd_una); tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); tp->snd_cwnd = ocwnd; if (SEQ_GT(onxt, tp->snd_nxt)) tp->snd_nxt = onxt; /* * Partial window deflation. Relies on fact that tp->snd_una * not updated yet. */ if (tp->snd_cwnd > th->th_ack - tp->snd_una) tp->snd_cwnd -= th->th_ack - tp->snd_una; else tp->snd_cwnd = 0; tp->snd_cwnd += tp->t_maxseg; } /* * Returns 1 if the TIME_WAIT state was killed and we should start over, * looking for a pcb in the listen state. Returns 0 otherwise. */ static int tcp_timewait(struct inpcb *inp, struct tcpopt *to, struct tcphdr *th, struct mbuf *m, int tlen) { struct tcptw *tw; int thflags; tcp_seq seq; #ifdef INET6 int isipv6 = (mtod(m, struct ip *)->ip_v == 6) ? 1 : 0; #else const int isipv6 = 0; #endif /* tcbinfo lock required for tcp_twclose(), tcp_timer_2msl_reset(). */ INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); /* * XXXRW: Time wait state for inpcb has been recycled, but inpcb is * still present. This is undesirable, but temporarily necessary * until we work out how to handle inpcb's who's timewait state has * been removed. */ tw = intotw(inp); if (tw == NULL) goto drop; thflags = th->th_flags; /* * NOTE: for FIN_WAIT_2 (to be added later), * must validate sequence number before accepting RST */ /* * If the segment contains RST: * Drop the segment - see Stevens, vol. 2, p. 964 and * RFC 1337. */ if (thflags & TH_RST) goto drop; #if 0 /* PAWS not needed at the moment */ /* * RFC 1323 PAWS: If we have a timestamp reply on this segment * and it's less than ts_recent, drop it. */ if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent && TSTMP_LT(to.to_tsval, tp->ts_recent)) { if ((thflags & TH_ACK) == 0) goto drop; goto ack; } /* * ts_recent is never updated because we never accept new segments. */ #endif /* * If a new connection request is received * while in TIME_WAIT, drop the old connection * and start over if the sequence numbers * are above the previous ones. */ if ((thflags & TH_SYN) && SEQ_GT(th->th_seq, tw->rcv_nxt)) { tcp_twclose(tw, 0); return (1); } /* * Drop the the segment if it does not contain an ACK. */ if ((thflags & TH_ACK) == 0) goto drop; /* * Reset the 2MSL timer if this is a duplicate FIN. */ if (thflags & TH_FIN) { seq = th->th_seq + tlen + (thflags & TH_SYN ? 1 : 0); if (seq + 1 == tw->rcv_nxt) tcp_timer_2msl_reset(tw, 1); } /* * Acknowledge the segment if it has data or is not a duplicate ACK. */ if (thflags != TH_ACK || tlen != 0 || th->th_seq != tw->rcv_nxt || th->th_ack != tw->snd_nxt) tcp_twrespond(tw, TH_ACK); goto drop; /* * Generate a RST, dropping incoming segment. * Make ACK acceptable to originator of segment. * Don't bother to respond if destination was broadcast/multicast. */ if (m->m_flags & (M_BCAST|M_MCAST)) goto drop; if (isipv6) { struct ip6_hdr *ip6; /* IPv6 anycast check is done at tcp6_input() */ ip6 = mtod(m, struct ip6_hdr *); if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto drop; } else { struct ip *ip; ip = mtod(m, struct ip *); if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || ip->ip_src.s_addr == htonl(INADDR_BROADCAST) || in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) goto drop; } if (thflags & TH_ACK) { tcp_respond(NULL, mtod(m, void *), th, m, 0, th->th_ack, TH_RST); } else { seq = th->th_seq + (thflags & TH_SYN ? 1 : 0); tcp_respond(NULL, mtod(m, void *), th, m, seq, 0, TH_RST|TH_ACK); } INP_UNLOCK(inp); return (0); drop: INP_UNLOCK(inp); m_freem(m); return (0); } Index: head/sys/netinet/tcp_syncache.c =================================================================== --- head/sys/netinet/tcp_syncache.c (revision 168902) +++ head/sys/netinet/tcp_syncache.c (revision 168903) @@ -1,1489 +1,1489 @@ /*- * 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 *); 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); 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->snd_max = tp->iss + 1; tp->snd_nxt = tp->iss + 1; tp->rcv_up = sc->sc_irs + 1; tp->rcv_wnd = sc->sc_wnd; tp->rcv_adv += tp->rcv_wnd; tp->last_ack_sent = tp->rcv_nxt; 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; tcp_timer_activate(tp, TT_KEEP, tcp_keepinit); 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 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; *lsop = syncache_socket(sc, *lsop, m); if (*lsop == NULL) tcpstat.tcps_sc_aborted++; else tcpstat.tcps_sc_completed++; if (sc != &scs) syncache_free(sc); return (1); failed: if (sc != NULL && sc != &scs) syncache_free(sc); *lsop = NULL; 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 +void 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) { INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); goto done; } 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) && (to->to_flags & TOF_TS)) sc->sc_tsreflect = to->to_tsval; else sc->sc_flags &= ~SCF_TIMESTAMP; #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) == 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++; if ((sc = TAILQ_LAST(&sch->sch_bucket, sch_head)) != NULL) 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. * 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. * * RFC1323: The Window field in a SYN (i.e., a * or ) segment itself is never scaled. */ while (wscale < TCP_MAX_WINSHIFT && (0x1 << wscale) < tcp_minmss) wscale++; sc->sc_requested_r_scale = wscale; sc->sc_requested_s_scale = to->to_wscale; 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) == 0) { if (tcp_syncookies && tcp_syncookiesonly && sc != &scs) syncache_free(sc); else if (sc != &scs) syncache_insert(sc, sch); /* locks and unlocks sch */ tcpstat.tcps_sndacks++; tcpstat.tcps_sndtotal++; } else { if (sc != &scs) syncache_free(sc); tcpstat.tcps_sc_dropped++; } done: #ifdef MAC if (sc == &scs) mac_destroy_syncache(&maclabel); #endif *lsop = NULL; m_freem(m); - return (1); + return; } static int syncache_respond(struct syncache *sc) { struct ip *ip = NULL; struct mbuf *m; struct tcphdr *th; int optlen, error; u_int16_t hlen, tlen, mssopt; struct tcpopt to; #ifdef INET6 struct ip6_hdr *ip6 = NULL; #endif hlen = #ifdef INET6 (sc->sc_inc.inc_isipv6) ? sizeof(struct ip6_hdr) : #endif sizeof(struct ip); tlen = hlen + sizeof(struct tcphdr); /* 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); /* XXX: Assume that the entire packet will fit in a header mbuf. */ KASSERT(max_linkhdr + tlen + MAX_TCPOPTLEN <= MHLEN, ("syncache: mbuf too small")); /* Create the IP+TCP header from scratch. */ 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) >> 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 ((sc->sc_flags & SCF_NOOPT) == 0) { to.to_flags = 0; to.to_mss = mssopt; to.to_flags = TOF_MSS; if (sc->sc_flags & SCF_WINSCALE) { to.to_wscale = sc->sc_requested_r_scale; to.to_flags |= TOF_SCALE; } if (sc->sc_flags & SCF_TIMESTAMP) { /* Virgin timestamp or TCP cookie enhanced one. */ to.to_tsval = sc->sc_ts ? sc->sc_ts : ticks; to.to_tsecr = sc->sc_tsreflect; to.to_flags |= TOF_TS; } if (sc->sc_flags & SCF_SACK) to.to_flags |= TOF_SACKPERM; #ifdef TCP_SIGNATURE if (sc->sc_flags & SCF_SIGNATURE) to.to_flags |= TOF_SIGNATURE; #endif optlen = tcp_addoptions(&to, (u_char *)(th + 1)); #ifdef TCP_SIGNATURE tcp_signature_compute(m, sizeof(struct ip), 0, optlen, to.to_signature, IPSEC_DIR_OUTBOUND); #endif /* Adjust headers by option size. */ th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; m->m_len += optlen; m->m_pkthdr.len += optlen; #ifdef INET6 if (sc->sc_inc.inc_isipv6) ip6->ip6_plen = htons(ntohs(ip6->ip6_plen) + optlen); else #endif ip->ip_len += optlen; } else optlen = 0; #ifdef INET6 if (sc->sc_inc.inc_isipv6) { th->th_sum = 0; th->th_sum = in6_cksum(m, IPPROTO_TCP, hlen, tlen + optlen - 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 + optlen - 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_var.h =================================================================== --- head/sys/netinet/tcp_var.h (revision 168902) +++ head/sys/netinet/tcp_var.h (revision 168903) @@ -1,575 +1,575 @@ /*- * Copyright (c) 1982, 1986, 1993, 1994, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_var.h 8.4 (Berkeley) 5/24/95 * $FreeBSD$ */ #ifndef _NETINET_TCP_VAR_H_ #define _NETINET_TCP_VAR_H_ #include /* * Kernel variables for tcp. */ extern int tcp_do_rfc1323; /* TCP segment queue entry */ struct tseg_qent { LIST_ENTRY(tseg_qent) tqe_q; int tqe_len; /* TCP segment data length */ struct tcphdr *tqe_th; /* a pointer to tcp header */ struct mbuf *tqe_m; /* mbuf contains packet */ }; LIST_HEAD(tsegqe_head, tseg_qent); extern int tcp_reass_qsize; extern struct uma_zone *tcp_reass_zone; struct sackblk { tcp_seq start; /* start seq no. of sack block */ tcp_seq end; /* end seq no. */ }; struct sackhole { tcp_seq start; /* start seq no. of hole */ tcp_seq end; /* end seq no. */ tcp_seq rxmit; /* next seq. no in hole to be retransmitted */ TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */ }; struct sackhint { struct sackhole *nexthole; int sack_bytes_rexmit; }; struct tcptemp { u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */ struct tcphdr tt_t; }; #define tcp6cb tcpcb /* for KAME src sync over BSD*'s */ /* * Tcp control block, one per tcp; fields: * Organized for 16 byte cacheline efficiency. */ struct tcpcb { struct tsegqe_head t_segq; /* segment reassembly queue */ int t_segqlen; /* segment reassembly queue length */ int t_dupacks; /* consecutive dup acks recd */ struct tcp_timer *t_timers; /* retransmit timer */ struct inpcb *t_inpcb; /* back pointer to internet pcb */ int t_state; /* state of this connection */ u_int t_flags; #define TF_ACKNOW 0x000001 /* ack peer immediately */ #define TF_DELACK 0x000002 /* ack, but try to delay it */ #define TF_NODELAY 0x000004 /* don't delay packets to coalesce */ #define TF_NOOPT 0x000008 /* don't use tcp options */ #define TF_SENTFIN 0x000010 /* have sent FIN */ #define TF_REQ_SCALE 0x000020 /* have/will request window scaling */ #define TF_RCVD_SCALE 0x000040 /* other side has requested scaling */ #define TF_REQ_TSTMP 0x000080 /* have/will request timestamps */ #define TF_RCVD_TSTMP 0x000100 /* a timestamp was received in SYN */ #define TF_SACK_PERMIT 0x000200 /* other side said I could SACK */ #define TF_NEEDSYN 0x000400 /* send SYN (implicit state) */ #define TF_NEEDFIN 0x000800 /* send FIN (implicit state) */ #define TF_NOPUSH 0x001000 /* don't push */ #define TF_MORETOCOME 0x010000 /* More data to be appended to sock */ #define TF_LQ_OVERFLOW 0x020000 /* listen queue overflow */ #define TF_LASTIDLE 0x040000 /* connection was previously idle */ #define TF_RXWIN0SENT 0x080000 /* sent a receiver win 0 in response */ #define TF_FASTRECOVERY 0x100000 /* in NewReno Fast Recovery */ #define TF_WASFRECOVERY 0x200000 /* was in NewReno Fast Recovery */ #define TF_SIGNATURE 0x400000 /* require MD5 digests (RFC2385) */ #define TF_FORCEDATA 0x800000 /* force out a byte */ #define TF_TSO 0x1000000 /* TSO enabled on this connection */ tcp_seq snd_una; /* send unacknowledged */ tcp_seq snd_max; /* highest sequence number sent; * used to recognize retransmits */ tcp_seq snd_nxt; /* send next */ tcp_seq snd_up; /* send urgent pointer */ tcp_seq snd_wl1; /* window update seg seq number */ tcp_seq snd_wl2; /* window update seg ack number */ tcp_seq iss; /* initial send sequence number */ tcp_seq irs; /* initial receive sequence number */ tcp_seq rcv_nxt; /* receive next */ tcp_seq rcv_adv; /* advertised window */ u_long rcv_wnd; /* receive window */ tcp_seq rcv_up; /* receive urgent pointer */ u_long snd_wnd; /* send window */ u_long snd_cwnd; /* congestion-controlled window */ u_long snd_bwnd; /* bandwidth-controlled window */ u_long snd_ssthresh; /* snd_cwnd size threshold for * for slow start exponential to * linear switch */ u_long snd_bandwidth; /* calculated bandwidth or 0 */ tcp_seq snd_recover; /* for use in NewReno Fast Recovery */ u_int t_maxopd; /* mss plus options */ u_long t_rcvtime; /* inactivity time */ u_long t_starttime; /* time connection was established */ int t_rtttime; /* round trip time */ tcp_seq t_rtseq; /* sequence number being timed */ int t_bw_rtttime; /* used for bandwidth calculation */ tcp_seq t_bw_rtseq; /* used for bandwidth calculation */ int t_rxtcur; /* current retransmit value (ticks) */ u_int t_maxseg; /* maximum segment size */ int t_srtt; /* smoothed round-trip time */ int t_rttvar; /* variance in round-trip time */ int t_rxtshift; /* log(2) of rexmt exp. backoff */ u_int t_rttmin; /* minimum rtt allowed */ u_int t_rttbest; /* best rtt we've seen */ u_long t_rttupdated; /* number of times rtt sampled */ u_long max_sndwnd; /* largest window peer has offered */ int t_softerror; /* possible error not yet reported */ /* out-of-band data */ char t_oobflags; /* have some */ char t_iobc; /* input character */ #define TCPOOB_HAVEDATA 0x01 #define TCPOOB_HADDATA 0x02 /* RFC 1323 variables */ u_char snd_scale; /* window scaling for send window */ u_char rcv_scale; /* window scaling for recv window */ u_char request_r_scale; /* pending window scaling */ u_char requested_s_scale; /* unused, to be reused later */ u_int32_t ts_recent; /* timestamp echo data */ u_long ts_recent_age; /* when last updated */ u_int32_t ts_offset; /* our timestamp offset */ tcp_seq last_ack_sent; /* experimental */ u_long snd_cwnd_prev; /* cwnd prior to retransmit */ u_long snd_ssthresh_prev; /* ssthresh prior to retransmit */ tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */ u_long t_badrxtwin; /* window for retransmit recovery */ u_char snd_limited; /* segments limited transmitted */ /* SACK related state */ int sack_enable; /* enable SACK for this connection */ int snd_numholes; /* number of holes seen by sender */ TAILQ_HEAD(sackhole_head, sackhole) snd_holes; /* SACK scoreboard (sorted) */ tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/ int rcv_numsacks; /* # distinct sack blks present */ struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */ tcp_seq sack_newdata; /* New data xmitted in this recovery episode starts at this seq number */ struct sackhint sackhint; /* SACK scoreboard hint */ int t_rttlow; /* smallest observerved RTT */ u_int32_t rfbuf_ts; /* recv buffer autoscaling timestamp */ int rfbuf_cnt; /* recv buffer autoscaling byte count */ }; #define IN_FASTRECOVERY(tp) (tp->t_flags & TF_FASTRECOVERY) #define ENTER_FASTRECOVERY(tp) tp->t_flags |= TF_FASTRECOVERY #define EXIT_FASTRECOVERY(tp) tp->t_flags &= ~TF_FASTRECOVERY #ifdef TCP_SIGNATURE /* * Defines which are needed by the xform_tcp module and tcp_[in|out]put * for SADB verification and lookup. */ #define TCP_SIGLEN 16 /* length of computed digest in bytes */ #define TCP_KEYLEN_MIN 1 /* minimum length of TCP-MD5 key */ #define TCP_KEYLEN_MAX 80 /* maximum length of TCP-MD5 key */ /* * Only a single SA per host may be specified at this time. An SPI is * needed in order for the KEY_ALLOCSA() lookup to work. */ #define TCP_SIG_SPI 0x1000 #endif /* TCP_SIGNATURE */ /* * Structure to hold TCP options that are only used during segment * processing (in tcp_input), but not held in the tcpcb. * It's basically used to reduce the number of parameters * to tcp_dooptions and tcp_addoptions. * The binary order of the to_flags is relevant for packing of the * options in tcp_addoptions. */ struct tcpopt { u_long to_flags; /* which options are present */ #define TOF_MSS 0x0001 /* maximum segment size */ #define TOF_SCALE 0x0002 /* window scaling */ #define TOF_SACKPERM 0x0004 /* SACK permitted */ #define TOF_TS 0x0010 /* timestamp */ #define TOF_SIGNATURE 0x0040 /* signature option present */ #define TOF_SIGLEN 0x0080 /* signature length valid (RFC2385) */ #define TOF_SACK 0x0100 /* Peer sent SACK option */ #define TOF_MAXOPT 0x0200 u_int32_t to_tsval; /* our new timestamp */ u_int32_t to_tsecr; /* reflected timestamp */ u_int16_t to_mss; /* maximum segment size */ u_int8_t to_wscale; /* window scaling */ u_int8_t to_nsacks; /* number of SACK blocks */ u_char *to_sacks; /* pointer to the first SACK blocks */ u_char *to_signature; /* pointer to the MD5 signature */ }; /* * Flags for tcp_dooptions. */ #define TO_SYN 0x01 /* parse SYN-only options */ struct hc_metrics_lite { /* must stay in sync with hc_metrics */ u_long rmx_mtu; /* MTU for this path */ u_long rmx_ssthresh; /* outbound gateway buffer limit */ u_long rmx_rtt; /* estimated round trip time */ u_long rmx_rttvar; /* estimated rtt variance */ u_long rmx_bandwidth; /* estimated bandwidth */ u_long rmx_cwnd; /* congestion window */ u_long rmx_sendpipe; /* outbound delay-bandwidth product */ u_long rmx_recvpipe; /* inbound delay-bandwidth product */ }; #ifndef _NETINET_IN_PCB_H_ struct in_conninfo; #endif /* _NETINET_IN_PCB_H_ */ struct tcptw { struct inpcb *tw_inpcb; /* XXX back pointer to internet pcb */ tcp_seq snd_nxt; tcp_seq rcv_nxt; tcp_seq iss; tcp_seq irs; u_short last_win; /* cached window value */ u_short tw_so_options; /* copy of so_options */ struct ucred *tw_cred; /* user credentials */ u_long t_recent; u_long t_starttime; int tw_time; TAILQ_ENTRY(tcptw) tw_2msl; }; #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb) #define intotw(ip) ((struct tcptw *)(ip)->inp_ppcb) #define sototcpcb(so) (intotcpcb(sotoinpcb(so))) /* * The smoothed round-trip time and estimated variance * are stored as fixed point numbers scaled by the values below. * For convenience, these scales are also used in smoothing the average * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed). * With these scales, srtt has 3 bits to the right of the binary point, * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the * binary point, and is smoothed with an ALPHA of 0.75. */ #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */ #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */ #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */ #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */ #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */ /* * The initial retransmission should happen at rtt + 4 * rttvar. * Because of the way we do the smoothing, srtt and rttvar * will each average +1/2 tick of bias. When we compute * the retransmit timer, we want 1/2 tick of rounding and * 1 extra tick because of +-1/2 tick uncertainty in the * firing of the timer. The bias will give us exactly the * 1.5 tick we need. But, because the bias is * statistical, we have to test that we don't drop below * the minimum feasible timer (which is 2 ticks). * This version of the macro adapted from a paper by Lawrence * Brakmo and Larry Peterson which outlines a problem caused * by insufficient precision in the original implementation, * which results in inappropriately large RTO values for very * fast networks. */ #define TCP_REXMTVAL(tp) \ max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \ + (tp)->t_rttvar) >> TCP_DELTA_SHIFT) /* * TCP statistics. * Many of these should be kept per connection, * but that's inconvenient at the moment. */ struct tcpstat { u_long tcps_connattempt; /* connections initiated */ u_long tcps_accepts; /* connections accepted */ u_long tcps_connects; /* connections established */ u_long tcps_drops; /* connections dropped */ u_long tcps_conndrops; /* embryonic connections dropped */ u_long tcps_minmssdrops; /* average minmss too low drops */ u_long tcps_closed; /* conn. closed (includes drops) */ u_long tcps_segstimed; /* segs where we tried to get rtt */ u_long tcps_rttupdated; /* times we succeeded */ u_long tcps_delack; /* delayed acks sent */ u_long tcps_timeoutdrop; /* conn. dropped in rxmt timeout */ u_long tcps_rexmttimeo; /* retransmit timeouts */ u_long tcps_persisttimeo; /* persist timeouts */ u_long tcps_keeptimeo; /* keepalive timeouts */ u_long tcps_keepprobe; /* keepalive probes sent */ u_long tcps_keepdrops; /* connections dropped in keepalive */ u_long tcps_sndtotal; /* total packets sent */ u_long tcps_sndpack; /* data packets sent */ u_long tcps_sndbyte; /* data bytes sent */ u_long tcps_sndrexmitpack; /* data packets retransmitted */ u_long tcps_sndrexmitbyte; /* data bytes retransmitted */ u_long tcps_sndrexmitbad; /* unnecessary packet retransmissions */ u_long tcps_sndacks; /* ack-only packets sent */ u_long tcps_sndprobe; /* window probes sent */ u_long tcps_sndurg; /* packets sent with URG only */ u_long tcps_sndwinup; /* window update-only packets sent */ u_long tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */ u_long tcps_rcvtotal; /* total packets received */ u_long tcps_rcvpack; /* packets received in sequence */ u_long tcps_rcvbyte; /* bytes received in sequence */ u_long tcps_rcvbadsum; /* packets received with ccksum errs */ u_long tcps_rcvbadoff; /* packets received with bad offset */ u_long tcps_rcvmemdrop; /* packets dropped for lack of memory */ u_long tcps_rcvshort; /* packets received too short */ u_long tcps_rcvduppack; /* duplicate-only packets received */ u_long tcps_rcvdupbyte; /* duplicate-only bytes received */ u_long tcps_rcvpartduppack; /* packets with some duplicate data */ u_long tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */ u_long tcps_rcvoopack; /* out-of-order packets received */ u_long tcps_rcvoobyte; /* out-of-order bytes received */ u_long tcps_rcvpackafterwin; /* packets with data after window */ u_long tcps_rcvbyteafterwin; /* bytes rcvd after window */ u_long tcps_rcvafterclose; /* packets rcvd after "close" */ u_long tcps_rcvwinprobe; /* rcvd window probe packets */ u_long tcps_rcvdupack; /* rcvd duplicate acks */ u_long tcps_rcvacktoomuch; /* rcvd acks for unsent data */ u_long tcps_rcvackpack; /* rcvd ack packets */ u_long tcps_rcvackbyte; /* bytes acked by rcvd acks */ u_long tcps_rcvwinupd; /* rcvd window update packets */ u_long tcps_pawsdrop; /* segments dropped due to PAWS */ u_long tcps_predack; /* times hdr predict ok for acks */ u_long tcps_preddat; /* times hdr predict ok for data pkts */ u_long tcps_pcbcachemiss; u_long tcps_cachedrtt; /* times cached RTT in route updated */ u_long tcps_cachedrttvar; /* times cached rttvar updated */ u_long tcps_cachedssthresh; /* times cached ssthresh updated */ u_long tcps_usedrtt; /* times RTT initialized from route */ u_long tcps_usedrttvar; /* times RTTVAR initialized from rt */ u_long tcps_usedssthresh; /* times ssthresh initialized from rt*/ u_long tcps_persistdrop; /* timeout in persist state */ u_long tcps_badsyn; /* bogus SYN, e.g. premature ACK */ u_long tcps_mturesent; /* resends due to MTU discovery */ u_long tcps_listendrop; /* listen queue overflows */ u_long tcps_badrst; /* ignored RSTs in the window */ u_long tcps_sc_added; /* entry added to syncache */ u_long tcps_sc_retransmitted; /* syncache entry was retransmitted */ u_long tcps_sc_dupsyn; /* duplicate SYN packet */ u_long tcps_sc_dropped; /* could not reply to packet */ u_long tcps_sc_completed; /* successful extraction of entry */ u_long tcps_sc_bucketoverflow; /* syncache per-bucket limit hit */ u_long tcps_sc_cacheoverflow; /* syncache cache limit hit */ u_long tcps_sc_reset; /* RST removed entry from syncache */ u_long tcps_sc_stale; /* timed out or listen socket gone */ u_long tcps_sc_aborted; /* syncache entry aborted */ u_long tcps_sc_badack; /* removed due to bad ACK */ u_long tcps_sc_unreach; /* ICMP unreachable received */ u_long tcps_sc_zonefail; /* zalloc() failed */ u_long tcps_sc_sendcookie; /* SYN cookie sent */ u_long tcps_sc_recvcookie; /* SYN cookie received */ u_long tcps_hc_added; /* entry added to hostcache */ u_long tcps_hc_bucketoverflow; /* hostcache per bucket limit hit */ u_long tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */ /* SACK related stats */ u_long tcps_sack_recovery_episode; /* SACK recovery episodes */ u_long tcps_sack_rexmits; /* SACK rexmit segments */ u_long tcps_sack_rexmit_bytes; /* SACK rexmit bytes */ u_long tcps_sack_rcv_blocks; /* SACK blocks (options) received */ u_long tcps_sack_send_blocks; /* SACK blocks (options) sent */ u_long tcps_sack_sboverflow; /* times scoreboard overflowed */ }; /* * TCB structure exported to user-land via sysctl(3). * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been * included. Not all of our clients do. */ #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_) struct xtcpcb { size_t xt_len; struct inpcb xt_inp; struct tcpcb xt_tp; struct xsocket xt_socket; u_quad_t xt_alignment_hack; }; #endif /* * Names for TCP sysctl objects */ #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */ #define TCPCTL_MSSDFLT 3 /* MSS default */ #define TCPCTL_STATS 4 /* statistics (read-only) */ #define TCPCTL_RTTDFLT 5 /* default RTT estimate */ #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */ #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */ #define TCPCTL_SENDSPACE 8 /* send buffer space */ #define TCPCTL_RECVSPACE 9 /* receive buffer space */ #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */ #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */ #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */ #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */ #define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */ #define TCPCTL_DROP 15 /* drop tcp connection */ #define TCPCTL_MAXID 16 #define TCPCTL_FINWAIT2_TIMEOUT 17 #define TCPCTL_NAMES { \ { 0, 0 }, \ { "rfc1323", CTLTYPE_INT }, \ { "mssdflt", CTLTYPE_INT }, \ { "stats", CTLTYPE_STRUCT }, \ { "rttdflt", CTLTYPE_INT }, \ { "keepidle", CTLTYPE_INT }, \ { "keepintvl", CTLTYPE_INT }, \ { "sendspace", CTLTYPE_INT }, \ { "recvspace", CTLTYPE_INT }, \ { "keepinit", CTLTYPE_INT }, \ { "pcblist", CTLTYPE_STRUCT }, \ { "delacktime", CTLTYPE_INT }, \ { "v6mssdflt", CTLTYPE_INT }, \ { "maxid", CTLTYPE_INT }, \ } #ifdef _KERNEL #ifdef SYSCTL_DECL SYSCTL_DECL(_net_inet_tcp); SYSCTL_DECL(_net_inet_tcp_sack); #endif extern struct inpcbhead tcb; /* head of queue of active tcpcb's */ extern struct inpcbinfo tcbinfo; extern struct tcpstat tcpstat; /* tcp statistics */ extern int tcp_mssdflt; /* XXX */ extern int tcp_minmss; extern int tcp_delack_enabled; extern int tcp_do_newreno; extern int path_mtu_discovery; extern int ss_fltsz; extern int ss_fltsz_local; extern int tcp_do_sack; /* SACK enabled/disabled */ int tcp_addoptions(struct tcpopt *, u_char *); struct tcpcb * tcp_close(struct tcpcb *); void tcp_discardcb(struct tcpcb *); void tcp_twstart(struct tcpcb *); #if 0 int tcp_twrecycleable(struct tcptw *tw); #endif void tcp_twclose(struct tcptw *_tw, int _reuse); void tcp_ctlinput(int, struct sockaddr *, void *); int tcp_ctloutput(struct socket *, struct sockopt *); struct tcpcb * tcp_drop(struct tcpcb *, int); void tcp_drain(void); void tcp_fasttimo(void); void tcp_init(void); void tcp_fini(void *); void tcp_reass_init(void); void tcp_input(struct mbuf *, int); u_long tcp_maxmtu(struct in_conninfo *, int *); u_long tcp_maxmtu6(struct in_conninfo *, int *); void tcp_mss(struct tcpcb *, int); int tcp_mssopt(struct in_conninfo *); struct inpcb * tcp_drop_syn_sent(struct inpcb *, int); struct inpcb * tcp_mtudisc(struct inpcb *, int); struct tcpcb * tcp_newtcpcb(struct inpcb *); int tcp_output(struct tcpcb *); void tcp_respond(struct tcpcb *, void *, struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int); int tcp_twrespond(struct tcptw *, int); void tcp_setpersist(struct tcpcb *); #ifdef TCP_SIGNATURE int tcp_signature_compute(struct mbuf *, int, int, int, u_char *, u_int); #endif void tcp_slowtimo(void); struct tcptemp * tcpip_maketemplate(struct inpcb *); void tcpip_fillheaders(struct inpcb *, void *, void *); void tcp_timer_activate(struct tcpcb *, int, u_int); int tcp_timer_active(struct tcpcb *, int); void tcp_trace(int, int, struct tcpcb *, void *, struct tcphdr *, int); void tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq); void syncache_init(void); void syncache_unreach(struct in_conninfo *, struct tcphdr *); int syncache_expand(struct in_conninfo *, struct tcpopt *, struct tcphdr *, struct socket **, struct mbuf *); -int syncache_add(struct in_conninfo *, struct tcpopt *, +void syncache_add(struct in_conninfo *, struct tcpopt *, struct tcphdr *, struct inpcb *, struct socket **, struct mbuf *); void syncache_chkrst(struct in_conninfo *, struct tcphdr *); void syncache_badack(struct in_conninfo *); /* * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo) */ void tcp_hc_init(void); void tcp_hc_get(struct in_conninfo *, struct hc_metrics_lite *); u_long tcp_hc_getmtu(struct in_conninfo *); void tcp_hc_updatemtu(struct in_conninfo *, u_long); void tcp_hc_update(struct in_conninfo *, struct hc_metrics_lite *); extern struct pr_usrreqs tcp_usrreqs; extern u_long tcp_sendspace; extern u_long tcp_recvspace; tcp_seq tcp_new_isn(struct tcpcb *); void tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq); void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart, tcp_seq rcv_lastend); void tcp_clean_sackreport(struct tcpcb *tp); void tcp_sack_adjust(struct tcpcb *tp); struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt); void tcp_sack_partialack(struct tcpcb *, struct tcphdr *); void tcp_free_sackholes(struct tcpcb *tp); int tcp_newreno(struct tcpcb *, struct tcphdr *); u_long tcp_seq_subtract(u_long, u_long ); #endif /* _KERNEL */ #endif /* _NETINET_TCP_VAR_H_ */