Index: head/sys/netinet/tcp_input.c =================================================================== --- head/sys/netinet/tcp_input.c (revision 167035) +++ head/sys/netinet/tcp_input.c (revision 167036) @@ -1,3401 +1,3405 @@ /*- * 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 "opt_tcp_input.h" #include "opt_tcp_sack.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 connections"); static int blackhole = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, blackhole, CTLFLAG_RW, &blackhole, 0, "Do not send RST when dropping refused connections"); 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"); #ifdef TCP_DROP_SYNFIN 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"); #endif 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; struct mtx *tcbinfo_mtx; static void tcp_dooptions(struct tcpopt *, u_char *, 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) \ ((!callout_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(tp, th, tlenp, m) register struct tcpcb *tp; register 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) { register 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) { register 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(mp, offp, proto) struct mbuf **mp; int *offp, proto; { register 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(m, off0) register struct mbuf *m; int off0; { register struct tcphdr *th; register struct ip *ip = NULL; register struct ipovly *ipov; register struct inpcb *inp = NULL; u_char *optp = NULL; int optlen = 0; int len, tlen, off; int drop_hdrlen; register struct tcpcb *tp = 0; register int thflags; struct socket *so = 0; int todrop, acked, ourfinisacked, needoutput = 0; u_long tiwin; struct tcpopt to; /* options in this segment */ int headlocked = 0; #ifdef IPFIREWALL_FORWARD struct m_tag *fwd_tag; #endif int rstreason; /* For badport_bandlim accounting purposes */ struct ip6_hdr *ip6 = NULL; #ifdef INET6 int isipv6; char ip6buf[INET6_ADDRSTRLEN]; #else const int isipv6 = 0; #endif #ifdef TCPDEBUG /* * The size of tcp_saveipgen must be the size of the max ip header, * now IPv6. */ u_char tcp_saveipgen[40]; struct tcphdr tcp_savetcp; short ostate = 0; #endif #ifdef INET6 isipv6 = (mtod(m, struct ip *)->ip_v == 6) ? 1 : 0; #endif bzero((char *)&to, sizeof(to)); 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))) == 0) { 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)) == 0) { 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; #ifdef TCP_DROP_SYNFIN /* * 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; #endif /* * 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, * until after ip6_savecontrol() is called and before other functions * which don't want those proto headers. * Because ip6_savecontrol() is going to parse the mbuf to * search for data to be passed up to user-land, it wants mbuf * parameters to be unchanged. * XXX: the call of ip6_savecontrol() has been obsoleted based on * latest version of the advanced API (20020110). */ drop_hdrlen = off0 + off; /* * Locate pcb for segment. */ INP_INFO_WLOCK(&tcbinfo); headlocked = 1; findpcb: KASSERT(headlocked, ("tcp_input: findpcb: head not locked")); #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); #ifdef IPFIREWALL_FORWARD } #endif /* IPFIREWALL_FORWARD */ #if defined(IPSEC) || defined(FAST_IPSEC) #ifdef INET6 if (isipv6) { if (inp != NULL && ipsec6_in_reject(m, inp)) { #ifdef IPSEC ipsec6stat.in_polvio++; #endif goto drop; } } else #endif /* INET6 */ if (inp != NULL && ipsec4_in_reject(m, inp)) { #ifdef IPSEC ipsecstat.in_polvio++; #endif goto drop; } #endif /*IPSEC || FAST_IPSEC*/ /* * If the state is CLOSED (i.e., TCB does not exist) then * all data in the incoming segment is discarded. * If the TCB exists but is in CLOSED state, it is embryonic, * but should either do a listen or a connect soon. */ if (inp == NULL) { if (tcp_log_in_vain) { #ifdef INET6 char dbuf[INET6_ADDRSTRLEN+2], sbuf[INET6_ADDRSTRLEN+2]; #else char dbuf[4*sizeof "123"], sbuf[4*sizeof "123"]; #endif if (isipv6) { #ifdef INET6 strcpy(dbuf, "["); strcpy(sbuf, "["); strcat(dbuf, ip6_sprintf(ip6buf, &ip6->ip6_dst)); strcat(sbuf, ip6_sprintf(ip6buf, &ip6->ip6_src)); strcat(dbuf, "]"); strcat(sbuf, "]"); #endif } else { strcpy(dbuf, inet_ntoa(ip->ip_dst)); strcpy(sbuf, inet_ntoa(ip->ip_src)); } switch (tcp_log_in_vain) { case 1: if ((thflags & TH_SYN) == 0) break; /* FALLTHROUGH */ case 2: 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); break; default: break; } } if (blackhole) { switch (blackhole) { case 1: if (thflags & TH_SYN) goto drop; break; case 2: goto drop; default: goto drop; } } 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 drop; else #endif if (inp->inp_ip_minttl > ip->ip_ttl) goto drop; } if (inp->inp_vflag & INP_TIMEWAIT) { /* * The only option of relevance is TOF_CC, and only if * present in a SYN segment. See tcp_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; } tp = intotcpcb(inp); if (tp == 0) { INP_UNLOCK(inp); rstreason = BANDLIM_RST_CLOSEDPORT; goto dropwithreset; } if (tp->t_state == TCPS_CLOSED) goto drop; #ifdef MAC INP_LOCK_ASSERT(inp); if (mac_check_inpcb_deliver(inp, m)) goto drop; #endif so = inp->inp_socket; KASSERT(so != NULL, ("tcp_input: so == NULL")); #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 if (so->so_options & SO_ACCEPTCONN) { struct in_conninfo inc; bzero(&inc, sizeof(inc)); #ifdef INET6 inc.inc_isipv6 = isipv6; #endif if (isipv6) { inc.inc6_faddr = ip6->ip6_src; inc.inc6_laddr = ip6->ip6_dst; } else { 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. */ tcpstat.tcps_badsyn++; rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } if (so == NULL) { /* * Could not complete 3-way handshake, * connection is being closed down, and * syncache has free'd mbuf. */ INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return; } /* * Socket is created in state SYN_RECEIVED. * Continue processing segment. */ INP_UNLOCK(inp); inp = sotoinpcb(so); INP_LOCK(inp); tp = intotcpcb(inp); /* * This is what would have happened in * tcp_output() when the SYN,ACK was sent. */ tp->snd_up = tp->snd_una; tp->snd_max = tp->snd_nxt = tp->iss + 1; tp->last_ack_sent = tp->rcv_nxt; goto after_listen; } if (thflags & TH_RST) { syncache_chkrst(&inc, th); goto drop; } if (thflags & TH_ACK) { syncache_badack(&inc); tcpstat.tcps_badsyn++; rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } goto drop; } /* * 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 /* * If it is from this socket, drop it, it must be forged. * Don't bother responding if the destination was a broadcast. */ if (th->th_dport == th->th_sport) { if (isipv6) { if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &ip6->ip6_src)) goto drop; } else { if (ip->ip_dst.s_addr == ip->ip_src.s_addr) goto drop; } } /* * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN * * 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 drop; if (isipv6) { if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto drop; } else { 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; } /* * SYN appears to be valid; create compressed TCP state * for syncache, or perform t/tcp connection. */ 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); #endif tcp_dooptions(&to, optp, optlen, TO_SYN); if (!syncache_add(&inc, &to, th, inp, &so, m)) goto drop; /* XXX: does not happen */ if (so == NULL) { /* * Entry added to syncache, mbuf used to * send SYN,ACK packet. Everything unlocked * already. */ return; } panic("T/TCP not supported at the moment"); #if 0 /* T/TCP */ /* * Segment passed TAO tests. * XXX: Can't happen at the moment. */ INP_UNLOCK(inp); inp = sotoinpcb(so); INP_LOCK(inp); tp = intotcpcb(inp); tp->t_starttime = ticks; tp->t_state = TCPS_ESTABLISHED; /* * T/TCP logic: * If there is a FIN or if there is data, then * delay SYN,ACK(SYN) in the hope of piggy-backing * it on a response segment. Otherwise must send * ACK now in case the other side is slow starting. */ if (thflags & TH_FIN || tlen != 0) tp->t_flags |= (TF_DELACK | TF_NEEDSYN); else tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN); tiwin = th->th_win << tp->snd_scale; tcpstat.tcps_connects++; soisconnected(so); goto trimthenstep6; #endif /* T/TCP */ } goto drop; } after_listen: KASSERT(headlocked, ("tcp_input: after_listen: head not locked")); INP_LOCK_ASSERT(inp); /* Syncache takes care of sockets in the listen state. */ KASSERT(tp->t_state != TCPS_LISTEN, ("tcp_input: TCPS_LISTEN")); /* * This is the second part of the MSS DoS prevention code (after * minmss on the sending side) and it deals with too many too small * tcp packets in a too short timeframe (1 second). * * For every full second we count the number of received packets * and bytes. If we get a lot of packets per second for this connection * (tcp_minmssoverload) we take a closer look at it and compute the * average packet size for the past second. If that is less than * tcp_minmss we get too many packets with very small payload which * is not good and burdens our system (and every packet generates * a wakeup to the process connected to our socket). We can reasonable * expect this to be small packet DoS attack to exhaust our CPU * cycles. * * Care has to be taken for the minimum packet overload value. This * value defines the minimum number of packets per second before we * start to worry. This must not be too low to avoid killing for * example interactive connections with many small packets like * telnet or SSH. * * Setting either tcp_minmssoverload or tcp_minmss to "0" disables * this check. * * Account for packet if payload packet, skip over ACK, etc. */ if (tcp_minmss && tcp_minmssoverload && tp->t_state == TCPS_ESTABLISHED && tlen > 0) { if ((unsigned int)(tp->rcv_second - ticks) < hz) { tp->rcv_pps++; tp->rcv_byps += tlen + off; if (tp->rcv_pps > tcp_minmssoverload) { if ((tp->rcv_byps / tp->rcv_pps) < tcp_minmss) { printf("too many small tcp packets from " "%s:%u, av. %lubyte/packet, " "dropping connection\n", #ifdef INET6 isipv6 ? ip6_sprintf(ip6buf, &inp->inp_inc.inc6_faddr) : #endif inet_ntoa(inp->inp_inc.inc_faddr), inp->inp_inc.inc_fport, tp->rcv_byps / tp->rcv_pps); KASSERT(headlocked, ("tcp_input: " "after_listen: tcp_drop: head " "not locked")); tp = tcp_drop(tp, ECONNRESET); tcpstat.tcps_minmssdrops++; goto drop; } } } else { tp->rcv_second = ticks + hz; tp->rcv_pps = 1; tp->rcv_byps = tlen + off; } } /* * Segment received on connection. * Reset idle time and keep-alive timer. */ tp->t_rcvtime = ticks; if (TCPS_HAVEESTABLISHED(tp->t_state)) callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp); /* * 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, optp, optlen, (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_requested_s_scale; 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; } if (to.to_flags & TOF_MSS) tcp_mss(tp, to.to_mss); if (tp->sack_enable) { if (!(to.to_flags & TOF_SACK)) 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_nsacks == 0 && TAILQ_EMPTY(&tp->snd_holes)))) { KASSERT(headlocked, ("headlocked")); 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) callout_stop(tp->tt_rexmt); else if (!callout_active(tp->tt_persist)) callout_reset(tp->tt_rexmt, tp->t_rxtcur, tcp_timer_rexmt, tp); 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, ("headlocked")); 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. */ { int win; 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, ("tcp_input: after_listen" ": tcp_drop.2: head not locked")); tp = tcp_drop(tp, ECONNREFUSED); } goto drop; } if ((thflags & TH_SYN) == 0) goto drop; /* Initial send window, already scaled. */ tp->snd_wnd = th->th_win; 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) callout_reset(tp->tt_delack, tcp_delacktime, tcp_timer_delack, tp); 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; callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp); } } 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); callout_stop(tp->tt_rexmt); tp->t_state = TCPS_SYN_RECEIVED; } #if 0 /* T/TCP */ trimthenstep6: #endif KASSERT(headlocked, ("tcp_input: trimthenstep6: head not " "locked")); INP_LOCK_ASSERT(inp); /* * 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, ("timewait")); 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) && SEQ_LT(th->th_seq, tp->last_ack_sent + tp->rcv_wnd)) || (tp->rcv_wnd == 0 && tp->last_ack_sent == th->th_seq)) { switch (tp->t_state) { case TCPS_SYN_RECEIVED: so->so_error = ECONNREFUSED; goto close; case TCPS_ESTABLISHED: if (tp->last_ack_sent != th->th_seq && tcp_insecure_rst == 0) { 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, ("tcp_input: " "trimthenstep6: tcp_close: head not " "locked")); tp = tcp_close(tp); break; case TCPS_CLOSING: case TCPS_LAST_ACK: KASSERT(headlocked, ("trimthenstep6: " "tcp_close.2: head not locked")); tp = tcp_close(tp); break; case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("timewait")); 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, ("trimthenstep6: tcp_close.3: head not " "locked")); 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, ("timewait")); if (thflags & TH_SYN && tp->t_state == TCPS_TIME_WAIT && SEQ_GT(th->th_seq, tp->rcv_nxt)) { KASSERT(headlocked, ("trimthenstep6: " "tcp_close.4: head not locked")); tp = tcp_close(tp); goto findpcb; } /* * 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, ("tcp_input: tcp_drop: trimthenstep6: " "head not locked")); 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; callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp); } /* * 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, ("timewait")); if (SEQ_GT(th->th_ack, tp->snd_max)) { tcpstat.tcps_rcvacktoomuch++; goto dropafterack; } if (tp->sack_enable && (to.to_nsacks > 0 || !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 (!callout_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; callout_stop(tp->tt_rexmt); 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, ("tp->snd_limited too big")); 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, ("dupacks not 1 or 2")); 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), ("sent too much")); 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), ("th_ack <= snd_una")); /* * 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, ("tcp_input: process_ACK: head not " "locked")); INP_LOCK_ASSERT(inp); 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) { callout_stop(tp->tt_rexmt); needoutput = 1; } else if (!callout_active(tp->tt_persist)) callout_reset(tp->tt_rexmt, tp->t_rxtcur, tcp_timer_rexmt, tp); /* * 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)) { register u_int cw = tp->snd_cwnd; register 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); - callout_reset(tp->tt_2msl, tcp_maxidle, + timeout = (tcp_fast_finwait2_recycle) ? + tcp_finwait2_timeout : tcp_maxidle; + callout_reset(tp->tt_2msl, timeout, tcp_timer_2msl, tp); } 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, ("tcp_input: process_ACK: " "head not locked")); tcp_twstart(tp); INP_INFO_WUNLOCK(&tcbinfo); m_freem(m); return; } 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, ("tcp_input: process_ACK:" " tcp_close: head not locked")); 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, ("timewait")); callout_reset(tp->tt_2msl, 2 * tcp_msl, tcp_timer_2msl, tp); goto dropafterack; } } step6: KASSERT(headlocked, ("tcp_input: step6: head not locked")); INP_LOCK_ASSERT(inp); /* * 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, ("tcp_input: dodata: head not locked")); INP_LOCK_ASSERT(inp); /* * 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); /* * Note the amount of data that peer has sent into * our window, in order to estimate the sender's * buffer size. */ len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); } 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, ("tcp_input: dodata: " "TCP_FIN_WAIT_2: head not locked")); tcp_twstart(tp); INP_INFO_WUNLOCK(&tcbinfo); return; /* * In TIME_WAIT state restart the 2 MSL time_wait timer. */ case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("timewait")); callout_reset(tp->tt_2msl, 2 * tcp_msl, tcp_timer_2msl, tp); 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, ("tcp_input: check_delack: head locked")); INP_LOCK_ASSERT(inp); if (tp->t_flags & TF_DELACK) { tp->t_flags &= ~TF_DELACK; callout_reset(tp->tt_delack, tcp_delacktime, tcp_timer_delack, tp); } INP_UNLOCK(inp); return; dropafterack: KASSERT(headlocked, ("tcp_input: dropafterack: head not locked")); /* * 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, ("headlocked should be 1")); INP_INFO_WUNLOCK(&tcbinfo); tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); INP_UNLOCK(inp); m_freem(m); return; dropwithreset: KASSERT(headlocked, ("tcp_input: dropwithreset: head not locked")); /* * Generate a RST, dropping incoming segment. * Make ACK acceptable to originator of segment. * Don't bother to respond if destination was broadcast/multicast. */ if ((thflags & TH_RST) || m->m_flags & (M_BCAST|M_MCAST)) goto drop; if (isipv6) { if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto drop; } else { 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; } /* IPv6 anycast check is done at tcp6_input() */ /* * Perform bandwidth limiting. */ if (badport_bandlim(rstreason) < 0) goto drop; #ifdef TCPDEBUG if (tp == 0 || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif if (thflags & TH_ACK) /* mtod() below is safe as long as hdr dropping is delayed */ tcp_respond(tp, mtod(m, void *), th, m, (tcp_seq)0, th->th_ack, TH_RST); else { if (thflags & TH_SYN) tlen++; /* mtod() below is safe as long as hdr dropping is delayed */ tcp_respond(tp, mtod(m, void *), th, m, th->th_seq+tlen, (tcp_seq)0, TH_RST|TH_ACK); } if (tp != NULL) INP_UNLOCK(inp); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); return; 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(inp); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); m_freem(m); return; } /* * Parse TCP options and place in tcpopt. */ static void tcp_dooptions(to, cp, cnt, flags) 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_requested_s_scale = 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_SACK; break; case TCPOPT_SACK: if (optlen <= 2 || (optlen - 2) % TCPOLEN_SACK != 0) continue; 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(so, th, m, off) struct socket *so; struct tcphdr *th; register struct mbuf *m; int off; /* delayed to be droped hdrlen */ { 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 == 0) break; } panic("tcp_pulloutofband"); } /* * Collect new round-trip time estimate * and update averages and current timeout. */ static void tcp_xmit_timer(tp, rtt) register struct tcpcb *tp; int rtt; { register 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(tp, offer) 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(inc) 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(tp, th) struct tcpcb *tp; struct tcphdr *th; { tcp_seq onxt = tp->snd_nxt; u_long ocwnd = tp->snd_cwnd; callout_stop(tp->tt_rexmt); 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(inp, to, th, m, tlen) 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 167035) +++ head/sys/netinet/tcp_reass.c (revision 167036) @@ -1,3401 +1,3405 @@ /*- * 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 "opt_tcp_input.h" #include "opt_tcp_sack.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 connections"); static int blackhole = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, blackhole, CTLFLAG_RW, &blackhole, 0, "Do not send RST when dropping refused connections"); 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"); #ifdef TCP_DROP_SYNFIN 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"); #endif 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; struct mtx *tcbinfo_mtx; static void tcp_dooptions(struct tcpopt *, u_char *, 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) \ ((!callout_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(tp, th, tlenp, m) register struct tcpcb *tp; register 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) { register 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) { register 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(mp, offp, proto) struct mbuf **mp; int *offp, proto; { register 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(m, off0) register struct mbuf *m; int off0; { register struct tcphdr *th; register struct ip *ip = NULL; register struct ipovly *ipov; register struct inpcb *inp = NULL; u_char *optp = NULL; int optlen = 0; int len, tlen, off; int drop_hdrlen; register struct tcpcb *tp = 0; register int thflags; struct socket *so = 0; int todrop, acked, ourfinisacked, needoutput = 0; u_long tiwin; struct tcpopt to; /* options in this segment */ int headlocked = 0; #ifdef IPFIREWALL_FORWARD struct m_tag *fwd_tag; #endif int rstreason; /* For badport_bandlim accounting purposes */ struct ip6_hdr *ip6 = NULL; #ifdef INET6 int isipv6; char ip6buf[INET6_ADDRSTRLEN]; #else const int isipv6 = 0; #endif #ifdef TCPDEBUG /* * The size of tcp_saveipgen must be the size of the max ip header, * now IPv6. */ u_char tcp_saveipgen[40]; struct tcphdr tcp_savetcp; short ostate = 0; #endif #ifdef INET6 isipv6 = (mtod(m, struct ip *)->ip_v == 6) ? 1 : 0; #endif bzero((char *)&to, sizeof(to)); 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))) == 0) { 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)) == 0) { 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; #ifdef TCP_DROP_SYNFIN /* * 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; #endif /* * 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, * until after ip6_savecontrol() is called and before other functions * which don't want those proto headers. * Because ip6_savecontrol() is going to parse the mbuf to * search for data to be passed up to user-land, it wants mbuf * parameters to be unchanged. * XXX: the call of ip6_savecontrol() has been obsoleted based on * latest version of the advanced API (20020110). */ drop_hdrlen = off0 + off; /* * Locate pcb for segment. */ INP_INFO_WLOCK(&tcbinfo); headlocked = 1; findpcb: KASSERT(headlocked, ("tcp_input: findpcb: head not locked")); #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); #ifdef IPFIREWALL_FORWARD } #endif /* IPFIREWALL_FORWARD */ #if defined(IPSEC) || defined(FAST_IPSEC) #ifdef INET6 if (isipv6) { if (inp != NULL && ipsec6_in_reject(m, inp)) { #ifdef IPSEC ipsec6stat.in_polvio++; #endif goto drop; } } else #endif /* INET6 */ if (inp != NULL && ipsec4_in_reject(m, inp)) { #ifdef IPSEC ipsecstat.in_polvio++; #endif goto drop; } #endif /*IPSEC || FAST_IPSEC*/ /* * If the state is CLOSED (i.e., TCB does not exist) then * all data in the incoming segment is discarded. * If the TCB exists but is in CLOSED state, it is embryonic, * but should either do a listen or a connect soon. */ if (inp == NULL) { if (tcp_log_in_vain) { #ifdef INET6 char dbuf[INET6_ADDRSTRLEN+2], sbuf[INET6_ADDRSTRLEN+2]; #else char dbuf[4*sizeof "123"], sbuf[4*sizeof "123"]; #endif if (isipv6) { #ifdef INET6 strcpy(dbuf, "["); strcpy(sbuf, "["); strcat(dbuf, ip6_sprintf(ip6buf, &ip6->ip6_dst)); strcat(sbuf, ip6_sprintf(ip6buf, &ip6->ip6_src)); strcat(dbuf, "]"); strcat(sbuf, "]"); #endif } else { strcpy(dbuf, inet_ntoa(ip->ip_dst)); strcpy(sbuf, inet_ntoa(ip->ip_src)); } switch (tcp_log_in_vain) { case 1: if ((thflags & TH_SYN) == 0) break; /* FALLTHROUGH */ case 2: 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); break; default: break; } } if (blackhole) { switch (blackhole) { case 1: if (thflags & TH_SYN) goto drop; break; case 2: goto drop; default: goto drop; } } 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 drop; else #endif if (inp->inp_ip_minttl > ip->ip_ttl) goto drop; } if (inp->inp_vflag & INP_TIMEWAIT) { /* * The only option of relevance is TOF_CC, and only if * present in a SYN segment. See tcp_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; } tp = intotcpcb(inp); if (tp == 0) { INP_UNLOCK(inp); rstreason = BANDLIM_RST_CLOSEDPORT; goto dropwithreset; } if (tp->t_state == TCPS_CLOSED) goto drop; #ifdef MAC INP_LOCK_ASSERT(inp); if (mac_check_inpcb_deliver(inp, m)) goto drop; #endif so = inp->inp_socket; KASSERT(so != NULL, ("tcp_input: so == NULL")); #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 if (so->so_options & SO_ACCEPTCONN) { struct in_conninfo inc; bzero(&inc, sizeof(inc)); #ifdef INET6 inc.inc_isipv6 = isipv6; #endif if (isipv6) { inc.inc6_faddr = ip6->ip6_src; inc.inc6_laddr = ip6->ip6_dst; } else { 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. */ tcpstat.tcps_badsyn++; rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } if (so == NULL) { /* * Could not complete 3-way handshake, * connection is being closed down, and * syncache has free'd mbuf. */ INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return; } /* * Socket is created in state SYN_RECEIVED. * Continue processing segment. */ INP_UNLOCK(inp); inp = sotoinpcb(so); INP_LOCK(inp); tp = intotcpcb(inp); /* * This is what would have happened in * tcp_output() when the SYN,ACK was sent. */ tp->snd_up = tp->snd_una; tp->snd_max = tp->snd_nxt = tp->iss + 1; tp->last_ack_sent = tp->rcv_nxt; goto after_listen; } if (thflags & TH_RST) { syncache_chkrst(&inc, th); goto drop; } if (thflags & TH_ACK) { syncache_badack(&inc); tcpstat.tcps_badsyn++; rstreason = BANDLIM_RST_OPENPORT; goto dropwithreset; } goto drop; } /* * 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 /* * If it is from this socket, drop it, it must be forged. * Don't bother responding if the destination was a broadcast. */ if (th->th_dport == th->th_sport) { if (isipv6) { if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &ip6->ip6_src)) goto drop; } else { if (ip->ip_dst.s_addr == ip->ip_src.s_addr) goto drop; } } /* * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN * * 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 drop; if (isipv6) { if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto drop; } else { 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; } /* * SYN appears to be valid; create compressed TCP state * for syncache, or perform t/tcp connection. */ 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); #endif tcp_dooptions(&to, optp, optlen, TO_SYN); if (!syncache_add(&inc, &to, th, inp, &so, m)) goto drop; /* XXX: does not happen */ if (so == NULL) { /* * Entry added to syncache, mbuf used to * send SYN,ACK packet. Everything unlocked * already. */ return; } panic("T/TCP not supported at the moment"); #if 0 /* T/TCP */ /* * Segment passed TAO tests. * XXX: Can't happen at the moment. */ INP_UNLOCK(inp); inp = sotoinpcb(so); INP_LOCK(inp); tp = intotcpcb(inp); tp->t_starttime = ticks; tp->t_state = TCPS_ESTABLISHED; /* * T/TCP logic: * If there is a FIN or if there is data, then * delay SYN,ACK(SYN) in the hope of piggy-backing * it on a response segment. Otherwise must send * ACK now in case the other side is slow starting. */ if (thflags & TH_FIN || tlen != 0) tp->t_flags |= (TF_DELACK | TF_NEEDSYN); else tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN); tiwin = th->th_win << tp->snd_scale; tcpstat.tcps_connects++; soisconnected(so); goto trimthenstep6; #endif /* T/TCP */ } goto drop; } after_listen: KASSERT(headlocked, ("tcp_input: after_listen: head not locked")); INP_LOCK_ASSERT(inp); /* Syncache takes care of sockets in the listen state. */ KASSERT(tp->t_state != TCPS_LISTEN, ("tcp_input: TCPS_LISTEN")); /* * This is the second part of the MSS DoS prevention code (after * minmss on the sending side) and it deals with too many too small * tcp packets in a too short timeframe (1 second). * * For every full second we count the number of received packets * and bytes. If we get a lot of packets per second for this connection * (tcp_minmssoverload) we take a closer look at it and compute the * average packet size for the past second. If that is less than * tcp_minmss we get too many packets with very small payload which * is not good and burdens our system (and every packet generates * a wakeup to the process connected to our socket). We can reasonable * expect this to be small packet DoS attack to exhaust our CPU * cycles. * * Care has to be taken for the minimum packet overload value. This * value defines the minimum number of packets per second before we * start to worry. This must not be too low to avoid killing for * example interactive connections with many small packets like * telnet or SSH. * * Setting either tcp_minmssoverload or tcp_minmss to "0" disables * this check. * * Account for packet if payload packet, skip over ACK, etc. */ if (tcp_minmss && tcp_minmssoverload && tp->t_state == TCPS_ESTABLISHED && tlen > 0) { if ((unsigned int)(tp->rcv_second - ticks) < hz) { tp->rcv_pps++; tp->rcv_byps += tlen + off; if (tp->rcv_pps > tcp_minmssoverload) { if ((tp->rcv_byps / tp->rcv_pps) < tcp_minmss) { printf("too many small tcp packets from " "%s:%u, av. %lubyte/packet, " "dropping connection\n", #ifdef INET6 isipv6 ? ip6_sprintf(ip6buf, &inp->inp_inc.inc6_faddr) : #endif inet_ntoa(inp->inp_inc.inc_faddr), inp->inp_inc.inc_fport, tp->rcv_byps / tp->rcv_pps); KASSERT(headlocked, ("tcp_input: " "after_listen: tcp_drop: head " "not locked")); tp = tcp_drop(tp, ECONNRESET); tcpstat.tcps_minmssdrops++; goto drop; } } } else { tp->rcv_second = ticks + hz; tp->rcv_pps = 1; tp->rcv_byps = tlen + off; } } /* * Segment received on connection. * Reset idle time and keep-alive timer. */ tp->t_rcvtime = ticks; if (TCPS_HAVEESTABLISHED(tp->t_state)) callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp); /* * 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, optp, optlen, (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_requested_s_scale; 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; } if (to.to_flags & TOF_MSS) tcp_mss(tp, to.to_mss); if (tp->sack_enable) { if (!(to.to_flags & TOF_SACK)) 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_nsacks == 0 && TAILQ_EMPTY(&tp->snd_holes)))) { KASSERT(headlocked, ("headlocked")); 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) callout_stop(tp->tt_rexmt); else if (!callout_active(tp->tt_persist)) callout_reset(tp->tt_rexmt, tp->t_rxtcur, tcp_timer_rexmt, tp); 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, ("headlocked")); 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. */ { int win; 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, ("tcp_input: after_listen" ": tcp_drop.2: head not locked")); tp = tcp_drop(tp, ECONNREFUSED); } goto drop; } if ((thflags & TH_SYN) == 0) goto drop; /* Initial send window, already scaled. */ tp->snd_wnd = th->th_win; 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) callout_reset(tp->tt_delack, tcp_delacktime, tcp_timer_delack, tp); 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; callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp); } } 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); callout_stop(tp->tt_rexmt); tp->t_state = TCPS_SYN_RECEIVED; } #if 0 /* T/TCP */ trimthenstep6: #endif KASSERT(headlocked, ("tcp_input: trimthenstep6: head not " "locked")); INP_LOCK_ASSERT(inp); /* * 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, ("timewait")); 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) && SEQ_LT(th->th_seq, tp->last_ack_sent + tp->rcv_wnd)) || (tp->rcv_wnd == 0 && tp->last_ack_sent == th->th_seq)) { switch (tp->t_state) { case TCPS_SYN_RECEIVED: so->so_error = ECONNREFUSED; goto close; case TCPS_ESTABLISHED: if (tp->last_ack_sent != th->th_seq && tcp_insecure_rst == 0) { 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, ("tcp_input: " "trimthenstep6: tcp_close: head not " "locked")); tp = tcp_close(tp); break; case TCPS_CLOSING: case TCPS_LAST_ACK: KASSERT(headlocked, ("trimthenstep6: " "tcp_close.2: head not locked")); tp = tcp_close(tp); break; case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("timewait")); 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, ("trimthenstep6: tcp_close.3: head not " "locked")); 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, ("timewait")); if (thflags & TH_SYN && tp->t_state == TCPS_TIME_WAIT && SEQ_GT(th->th_seq, tp->rcv_nxt)) { KASSERT(headlocked, ("trimthenstep6: " "tcp_close.4: head not locked")); tp = tcp_close(tp); goto findpcb; } /* * 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, ("tcp_input: tcp_drop: trimthenstep6: " "head not locked")); 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; callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp); } /* * 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, ("timewait")); if (SEQ_GT(th->th_ack, tp->snd_max)) { tcpstat.tcps_rcvacktoomuch++; goto dropafterack; } if (tp->sack_enable && (to.to_nsacks > 0 || !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 (!callout_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; callout_stop(tp->tt_rexmt); 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, ("tp->snd_limited too big")); 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, ("dupacks not 1 or 2")); 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), ("sent too much")); 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), ("th_ack <= snd_una")); /* * 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, ("tcp_input: process_ACK: head not " "locked")); INP_LOCK_ASSERT(inp); 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) { callout_stop(tp->tt_rexmt); needoutput = 1; } else if (!callout_active(tp->tt_persist)) callout_reset(tp->tt_rexmt, tp->t_rxtcur, tcp_timer_rexmt, tp); /* * 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)) { register u_int cw = tp->snd_cwnd; register 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); - callout_reset(tp->tt_2msl, tcp_maxidle, + timeout = (tcp_fast_finwait2_recycle) ? + tcp_finwait2_timeout : tcp_maxidle; + callout_reset(tp->tt_2msl, timeout, tcp_timer_2msl, tp); } 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, ("tcp_input: process_ACK: " "head not locked")); tcp_twstart(tp); INP_INFO_WUNLOCK(&tcbinfo); m_freem(m); return; } 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, ("tcp_input: process_ACK:" " tcp_close: head not locked")); 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, ("timewait")); callout_reset(tp->tt_2msl, 2 * tcp_msl, tcp_timer_2msl, tp); goto dropafterack; } } step6: KASSERT(headlocked, ("tcp_input: step6: head not locked")); INP_LOCK_ASSERT(inp); /* * 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, ("tcp_input: dodata: head not locked")); INP_LOCK_ASSERT(inp); /* * 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); /* * Note the amount of data that peer has sent into * our window, in order to estimate the sender's * buffer size. */ len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); } 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, ("tcp_input: dodata: " "TCP_FIN_WAIT_2: head not locked")); tcp_twstart(tp); INP_INFO_WUNLOCK(&tcbinfo); return; /* * In TIME_WAIT state restart the 2 MSL time_wait timer. */ case TCPS_TIME_WAIT: KASSERT(tp->t_state != TCPS_TIME_WAIT, ("timewait")); callout_reset(tp->tt_2msl, 2 * tcp_msl, tcp_timer_2msl, tp); 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, ("tcp_input: check_delack: head locked")); INP_LOCK_ASSERT(inp); if (tp->t_flags & TF_DELACK) { tp->t_flags &= ~TF_DELACK; callout_reset(tp->tt_delack, tcp_delacktime, tcp_timer_delack, tp); } INP_UNLOCK(inp); return; dropafterack: KASSERT(headlocked, ("tcp_input: dropafterack: head not locked")); /* * 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, ("headlocked should be 1")); INP_INFO_WUNLOCK(&tcbinfo); tp->t_flags |= TF_ACKNOW; (void) tcp_output(tp); INP_UNLOCK(inp); m_freem(m); return; dropwithreset: KASSERT(headlocked, ("tcp_input: dropwithreset: head not locked")); /* * Generate a RST, dropping incoming segment. * Make ACK acceptable to originator of segment. * Don't bother to respond if destination was broadcast/multicast. */ if ((thflags & TH_RST) || m->m_flags & (M_BCAST|M_MCAST)) goto drop; if (isipv6) { if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) goto drop; } else { 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; } /* IPv6 anycast check is done at tcp6_input() */ /* * Perform bandwidth limiting. */ if (badport_bandlim(rstreason) < 0) goto drop; #ifdef TCPDEBUG if (tp == 0 || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif if (thflags & TH_ACK) /* mtod() below is safe as long as hdr dropping is delayed */ tcp_respond(tp, mtod(m, void *), th, m, (tcp_seq)0, th->th_ack, TH_RST); else { if (thflags & TH_SYN) tlen++; /* mtod() below is safe as long as hdr dropping is delayed */ tcp_respond(tp, mtod(m, void *), th, m, th->th_seq+tlen, (tcp_seq)0, TH_RST|TH_ACK); } if (tp != NULL) INP_UNLOCK(inp); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); return; 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(inp); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); m_freem(m); return; } /* * Parse TCP options and place in tcpopt. */ static void tcp_dooptions(to, cp, cnt, flags) 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_requested_s_scale = 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_SACK; break; case TCPOPT_SACK: if (optlen <= 2 || (optlen - 2) % TCPOLEN_SACK != 0) continue; 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(so, th, m, off) struct socket *so; struct tcphdr *th; register struct mbuf *m; int off; /* delayed to be droped hdrlen */ { 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 == 0) break; } panic("tcp_pulloutofband"); } /* * Collect new round-trip time estimate * and update averages and current timeout. */ static void tcp_xmit_timer(tp, rtt) register struct tcpcb *tp; int rtt; { register 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(tp, offer) 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(inc) 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(tp, th) struct tcpcb *tp; struct tcphdr *th; { tcp_seq onxt = tp->snd_nxt; u_long ocwnd = tp->snd_cwnd; callout_stop(tp->tt_rexmt); 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(inp, to, th, m, tlen) 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_subr.c =================================================================== --- head/sys/netinet/tcp_subr.c (revision 167035) +++ head/sys/netinet/tcp_subr.c (revision 167036) @@ -1,2428 +1,2429 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 * $FreeBSD$ */ #include "opt_compat.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_mac.h" #include "opt_tcpdebug.h" #include "opt_tcp_sack.h" #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef INET6 #include #endif #include #include #ifdef INET6 #include #include #include #endif #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef TCPDEBUG #include #endif #include #ifdef IPSEC #include #ifdef INET6 #include #endif #include #endif /*IPSEC*/ #ifdef FAST_IPSEC #include #include #ifdef INET6 #include #endif #include #define IPSEC #endif /*FAST_IPSEC*/ #include #include #include int tcp_mssdflt = TCP_MSS; SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW, &tcp_mssdflt , 0, "Default TCP Maximum Segment Size"); #ifdef INET6 int tcp_v6mssdflt = TCP6_MSS; SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, CTLFLAG_RW, &tcp_v6mssdflt , 0, "Default TCP Maximum Segment Size for IPv6"); #endif /* * Minimum MSS we accept and use. This prevents DoS attacks where * we are forced to a ridiculous low MSS like 20 and send hundreds * of packets instead of one. The effect scales with the available * bandwidth and quickly saturates the CPU and network interface * with packet generation and sending. Set to zero to disable MINMSS * checking. This setting prevents us from sending too small packets. */ int tcp_minmss = TCP_MINMSS; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_RW, &tcp_minmss , 0, "Minmum TCP Maximum Segment Size"); /* * Number of TCP segments per second we accept from remote host * before we start to calculate average segment size. If average * segment size drops below the minimum TCP MSS we assume a DoS * attack and reset+drop the connection. Care has to be taken not to * set this value too small to not kill interactive type connections * (telnet, SSH) which send many small packets. */ int tcp_minmssoverload = TCP_MINMSSOVERLOAD; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmssoverload, CTLFLAG_RW, &tcp_minmssoverload , 0, "Number of TCP Segments per Second allowed to" "be under the MINMSS Size"); #if 0 static int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; SYSCTL_INT(_net_inet_tcp, TCPCTL_RTTDFLT, rttdflt, CTLFLAG_RW, &tcp_rttdflt , 0, "Default maximum TCP Round Trip Time"); #endif int tcp_do_rfc1323 = 1; SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW, &tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions"); static int tcp_tcbhashsize = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RDTUN, &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); static int do_tcpdrain = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, "Enable tcp_drain routine for extra help when low on mbufs"); SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD, &tcbinfo.ipi_count, 0, "Number of active PCBs"); static int icmp_may_rst = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW, &icmp_may_rst, 0, "Certain ICMP unreachable messages may abort connections in SYN_SENT"); static int tcp_isn_reseed_interval = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_RW, &tcp_isn_reseed_interval, 0, "Seconds between reseeding of ISN secret"); static uma_zone_t tcptw_zone; static int maxtcptw; static int tcptw_auto_size(void) { int halfrange; /* * Max out at half the ephemeral port range so that TIME_WAIT * sockets don't tie up too many ephemeral ports. */ if (ipport_lastauto > ipport_firstauto) halfrange = (ipport_lastauto - ipport_firstauto) / 2; else halfrange = (ipport_firstauto - ipport_lastauto) / 2; /* Protect against goofy port ranges smaller than 32. */ return (imin(imax(halfrange, 32), maxsockets / 5)); } static int sysctl_maxtcptw(SYSCTL_HANDLER_ARGS) { int error, new; if (maxtcptw == 0) new = tcptw_auto_size(); else new = maxtcptw; error = sysctl_handle_int(oidp, &new, sizeof(int), req); if (error == 0 && req->newptr) if (new >= 32) { maxtcptw = new; uma_zone_set_max(tcptw_zone, maxtcptw); } return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, maxtcptw, CTLTYPE_INT|CTLFLAG_RW, &maxtcptw, 0, sysctl_maxtcptw, "IU", "Maximum number of compressed TCP TIME_WAIT entries"); static int nolocaltimewait = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, nolocaltimewait, CTLFLAG_RW, &nolocaltimewait, 0, "Do not create compressed TCP TIME_WAIT entries" "for local connections"); /* * TCP bandwidth limiting sysctls. Note that the default lower bound of * 1024 exists only for debugging. A good production default would be * something like 6100. */ SYSCTL_NODE(_net_inet_tcp, OID_AUTO, inflight, CTLFLAG_RW, 0, "TCP inflight data limiting"); static int tcp_inflight_enable = 1; SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, enable, CTLFLAG_RW, &tcp_inflight_enable, 0, "Enable automatic TCP inflight data limiting"); static int tcp_inflight_debug = 0; SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, debug, CTLFLAG_RW, &tcp_inflight_debug, 0, "Debug TCP inflight calculations"); static int tcp_inflight_rttthresh; SYSCTL_PROC(_net_inet_tcp_inflight, OID_AUTO, rttthresh, CTLTYPE_INT|CTLFLAG_RW, &tcp_inflight_rttthresh, 0, sysctl_msec_to_ticks, "I", "RTT threshold below which inflight will deactivate itself"); static int tcp_inflight_min = 6144; SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, min, CTLFLAG_RW, &tcp_inflight_min, 0, "Lower-bound for TCP inflight window"); static int tcp_inflight_max = TCP_MAXWIN << TCP_MAX_WINSHIFT; SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, max, CTLFLAG_RW, &tcp_inflight_max, 0, "Upper-bound for TCP inflight window"); static int tcp_inflight_stab = 20; SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, stab, CTLFLAG_RW, &tcp_inflight_stab, 0, "Inflight Algorithm Stabilization 20 = 2 packets"); uma_zone_t sack_hole_zone; static struct inpcb *tcp_notify(struct inpcb *, int); static void tcp_isn_tick(void *); /* * Target size of TCP PCB hash tables. Must be a power of two. * * Note that this can be overridden by the kernel environment * variable net.inet.tcp.tcbhashsize */ #ifndef TCBHASHSIZE #define TCBHASHSIZE 512 #endif /* * XXX * Callouts should be moved into struct tcp directly. They are currently * separate because the tcpcb structure is exported to userland for sysctl * parsing purposes, which do not know about callouts. */ struct tcpcb_mem { struct tcpcb tcb; struct callout tcpcb_mem_rexmt, tcpcb_mem_persist, tcpcb_mem_keep; struct callout tcpcb_mem_2msl, tcpcb_mem_delack; }; static uma_zone_t tcpcb_zone; struct callout isn_callout; static struct mtx isn_mtx; #define ISN_LOCK_INIT() mtx_init(&isn_mtx, "isn_mtx", NULL, MTX_DEF) #define ISN_LOCK() mtx_lock(&isn_mtx) #define ISN_UNLOCK() mtx_unlock(&isn_mtx) /* * TCP initialization. */ static void tcp_zone_change(void *tag) { uma_zone_set_max(tcbinfo.ipi_zone, maxsockets); uma_zone_set_max(tcpcb_zone, maxsockets); if (maxtcptw == 0) uma_zone_set_max(tcptw_zone, tcptw_auto_size()); } static int tcp_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "tcpinp"); return (0); } void tcp_init(void) { int hashsize = TCBHASHSIZE; tcp_delacktime = TCPTV_DELACK; tcp_keepinit = TCPTV_KEEP_INIT; tcp_keepidle = TCPTV_KEEP_IDLE; tcp_keepintvl = TCPTV_KEEPINTVL; tcp_maxpersistidle = TCPTV_KEEP_IDLE; tcp_msl = TCPTV_MSL; tcp_rexmit_min = TCPTV_MIN; tcp_rexmit_slop = TCPTV_CPU_VAR; tcp_inflight_rttthresh = TCPTV_INFLIGHT_RTTTHRESH; + tcp_finwait2_timeout = TCPTV_FINWAIT2_TIMEOUT; INP_INFO_LOCK_INIT(&tcbinfo, "tcp"); LIST_INIT(&tcb); tcbinfo.listhead = &tcb; TUNABLE_INT_FETCH("net.inet.tcp.tcbhashsize", &hashsize); if (!powerof2(hashsize)) { printf("WARNING: TCB hash size not a power of 2\n"); hashsize = 512; /* safe default */ } tcp_tcbhashsize = hashsize; tcbinfo.hashbase = hashinit(hashsize, M_PCB, &tcbinfo.hashmask); tcbinfo.porthashbase = hashinit(hashsize, M_PCB, &tcbinfo.porthashmask); tcbinfo.ipi_zone = uma_zcreate("inpcb", sizeof(struct inpcb), NULL, NULL, tcp_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcbinfo.ipi_zone, maxsockets); #ifdef INET6 #define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) #else /* INET6 */ #define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) #endif /* INET6 */ if (max_protohdr < TCP_MINPROTOHDR) max_protohdr = TCP_MINPROTOHDR; if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) panic("tcp_init"); #undef TCP_MINPROTOHDR /* * These have to be type stable for the benefit of the timers. */ tcpcb_zone = uma_zcreate("tcpcb", sizeof(struct tcpcb_mem), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcpcb_zone, maxsockets); tcptw_zone = uma_zcreate("tcptw", sizeof(struct tcptw), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); TUNABLE_INT_FETCH("net.inet.tcp.maxtcptw", &maxtcptw); if (maxtcptw == 0) uma_zone_set_max(tcptw_zone, tcptw_auto_size()); else uma_zone_set_max(tcptw_zone, maxtcptw); tcp_timer_init(); syncache_init(); tcp_hc_init(); tcp_reass_init(); ISN_LOCK_INIT(); callout_init(&isn_callout, CALLOUT_MPSAFE); tcp_isn_tick(NULL); EVENTHANDLER_REGISTER(shutdown_pre_sync, tcp_fini, NULL, SHUTDOWN_PRI_DEFAULT); sack_hole_zone = uma_zcreate("sackhole", sizeof(struct sackhole), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); EVENTHANDLER_REGISTER(maxsockets_change, tcp_zone_change, NULL, EVENTHANDLER_PRI_ANY); } void tcp_fini(void *xtp) { callout_stop(&isn_callout); } /* * Fill in the IP and TCP headers for an outgoing packet, given the tcpcb. * tcp_template used to store this data in mbufs, but we now recopy it out * of the tcpcb each time to conserve mbufs. */ void tcpip_fillheaders(struct inpcb *inp, void *ip_ptr, void *tcp_ptr) { struct tcphdr *th = (struct tcphdr *)tcp_ptr; INP_LOCK_ASSERT(inp); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { struct ip6_hdr *ip6; ip6 = (struct ip6_hdr *)ip_ptr; ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | (inp->in6p_flowinfo & IPV6_FLOWINFO_MASK); ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | (IPV6_VERSION & IPV6_VERSION_MASK); ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = sizeof(struct tcphdr); ip6->ip6_src = inp->in6p_laddr; ip6->ip6_dst = inp->in6p_faddr; } else #endif { struct ip *ip; ip = (struct ip *)ip_ptr; ip->ip_v = IPVERSION; ip->ip_hl = 5; ip->ip_tos = inp->inp_ip_tos; ip->ip_len = 0; ip->ip_id = 0; ip->ip_off = 0; ip->ip_ttl = inp->inp_ip_ttl; ip->ip_sum = 0; ip->ip_p = IPPROTO_TCP; ip->ip_src = inp->inp_laddr; ip->ip_dst = inp->inp_faddr; } th->th_sport = inp->inp_lport; th->th_dport = inp->inp_fport; th->th_seq = 0; th->th_ack = 0; th->th_x2 = 0; th->th_off = 5; th->th_flags = 0; th->th_win = 0; th->th_urp = 0; th->th_sum = 0; /* in_pseudo() is called later for ipv4 */ } /* * Create template to be used to send tcp packets on a connection. * Allocates an mbuf and fills in a skeletal tcp/ip header. The only * use for this function is in keepalives, which use tcp_respond. */ struct tcptemp * tcpip_maketemplate(struct inpcb *inp) { struct mbuf *m; struct tcptemp *n; m = m_get(M_DONTWAIT, MT_DATA); if (m == NULL) return (0); m->m_len = sizeof(struct tcptemp); n = mtod(m, struct tcptemp *); tcpip_fillheaders(inp, (void *)&n->tt_ipgen, (void *)&n->tt_t); return (n); } /* * Send a single message to the TCP at address specified by * the given TCP/IP header. If m == NULL, then we make a copy * of the tcpiphdr at ti and send directly to the addressed host. * This is used to force keep alive messages out using the TCP * template for a connection. If flags are given then we send * a message back to the TCP which originated the * segment ti, * and discard the mbuf containing it and any other attached mbufs. * * In any case the ack and sequence number of the transmitted * segment are as specified by the parameters. * * NOTE: If m != NULL, then ti must point to *inside* the mbuf. */ void tcp_respond(struct tcpcb *tp, void *ipgen, register struct tcphdr *th, register struct mbuf *m, tcp_seq ack, tcp_seq seq, int flags) { register int tlen; int win = 0; struct ip *ip; struct tcphdr *nth; #ifdef INET6 struct ip6_hdr *ip6; int isipv6; #endif /* INET6 */ int ipflags = 0; struct inpcb *inp; KASSERT(tp != NULL || m != NULL, ("tcp_respond: tp and m both NULL")); #ifdef INET6 isipv6 = ((struct ip *)ipgen)->ip_v == 6; ip6 = ipgen; #endif /* INET6 */ ip = ipgen; if (tp != NULL) { inp = tp->t_inpcb; KASSERT(inp != NULL, ("tcp control block w/o inpcb")); INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); } else inp = NULL; if (tp != NULL) { if (!(flags & TH_RST)) { win = sbspace(&inp->inp_socket->so_rcv); if (win > (long)TCP_MAXWIN << tp->rcv_scale) win = (long)TCP_MAXWIN << tp->rcv_scale; } } if (m == NULL) { m = m_gethdr(M_DONTWAIT, MT_DATA); if (m == NULL) return; tlen = 0; m->m_data += max_linkhdr; #ifdef INET6 if (isipv6) { bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(struct ip6_hdr)); ip6 = mtod(m, struct ip6_hdr *); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); ip = mtod(m, struct ip *); nth = (struct tcphdr *)(ip + 1); } bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); flags = TH_ACK; } else { m_freem(m->m_next); m->m_next = NULL; m->m_data = (caddr_t)ipgen; /* m_len is set later */ tlen = 0; #define xchg(a,b,type) { type t; t=a; a=b; b=t; } #ifdef INET6 if (isipv6) { xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long); nth = (struct tcphdr *)(ip + 1); } if (th != nth) { /* * this is usually a case when an extension header * exists between the IPv6 header and the * TCP header. */ nth->th_sport = th->th_sport; nth->th_dport = th->th_dport; } xchg(nth->th_dport, nth->th_sport, n_short); #undef xchg } #ifdef INET6 if (isipv6) { ip6->ip6_flow = 0; ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) + tlen)); tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); } else #endif { tlen += sizeof (struct tcpiphdr); ip->ip_len = tlen; ip->ip_ttl = ip_defttl; if (path_mtu_discovery) ip->ip_off |= IP_DF; } m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; #ifdef MAC if (inp != NULL) { /* * Packet is associated with a socket, so allow the * label of the response to reflect the socket label. */ INP_LOCK_ASSERT(inp); mac_create_mbuf_from_inpcb(inp, m); } else { /* * Packet is not associated with a socket, so possibly * update the label in place. */ mac_reflect_mbuf_tcp(m); } #endif nth->th_seq = htonl(seq); nth->th_ack = htonl(ack); nth->th_x2 = 0; nth->th_off = sizeof (struct tcphdr) >> 2; nth->th_flags = flags; if (tp != NULL) nth->th_win = htons((u_short) (win >> tp->rcv_scale)); else nth->th_win = htons((u_short)win); nth->th_urp = 0; #ifdef INET6 if (isipv6) { nth->th_sum = 0; nth->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), tlen - sizeof(struct ip6_hdr)); ip6->ip6_hlim = in6_selecthlim(tp != NULL ? tp->t_inpcb : NULL, NULL); } else #endif /* INET6 */ { nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p))); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); } #ifdef TCPDEBUG if (tp == NULL || (inp->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); #endif #ifdef INET6 if (isipv6) (void) ip6_output(m, NULL, NULL, ipflags, NULL, NULL, inp); else #endif /* INET6 */ (void) ip_output(m, NULL, NULL, ipflags, NULL, inp); } /* * Create a new TCP control block, making an * empty reassembly queue and hooking it to the argument * protocol control block. The `inp' parameter must have * come from the zone allocator set up in tcp_init(). */ struct tcpcb * tcp_newtcpcb(struct inpcb *inp) { struct tcpcb_mem *tm; struct tcpcb *tp; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ tm = uma_zalloc(tcpcb_zone, M_NOWAIT | M_ZERO); if (tm == NULL) return (NULL); tp = &tm->tcb; /* LIST_INIT(&tp->t_segq); */ /* XXX covered by M_ZERO */ tp->t_maxseg = tp->t_maxopd = #ifdef INET6 isipv6 ? tcp_v6mssdflt : #endif /* INET6 */ tcp_mssdflt; /* Set up our timeouts. */ callout_init(tp->tt_rexmt = &tm->tcpcb_mem_rexmt, NET_CALLOUT_MPSAFE); callout_init(tp->tt_persist = &tm->tcpcb_mem_persist, NET_CALLOUT_MPSAFE); callout_init(tp->tt_keep = &tm->tcpcb_mem_keep, NET_CALLOUT_MPSAFE); callout_init(tp->tt_2msl = &tm->tcpcb_mem_2msl, NET_CALLOUT_MPSAFE); callout_init(tp->tt_delack = &tm->tcpcb_mem_delack, NET_CALLOUT_MPSAFE); if (tcp_do_rfc1323) tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); tp->sack_enable = tcp_do_sack; TAILQ_INIT(&tp->snd_holes); tp->t_inpcb = inp; /* XXX */ /* * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives * reasonable initial retransmit time. */ tp->t_srtt = TCPTV_SRTTBASE; tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; tp->t_rttmin = tcp_rexmit_min; tp->t_rxtcur = TCPTV_RTOBASE; tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->t_rcvtime = ticks; tp->t_bw_rtttime = ticks; /* * IPv4 TTL initialization is necessary for an IPv6 socket as well, * because the socket may be bound to an IPv6 wildcard address, * which may match an IPv4-mapped IPv6 address. */ inp->inp_ip_ttl = ip_defttl; inp->inp_ppcb = tp; return (tp); /* XXX */ } /* * Drop a TCP connection, reporting * the specified error. If connection is synchronized, * then send a RST to peer. */ struct tcpcb * tcp_drop(struct tcpcb *tp, int errno) { struct socket *so = tp->t_inpcb->inp_socket; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(tp->t_inpcb); if (TCPS_HAVERCVDSYN(tp->t_state)) { tp->t_state = TCPS_CLOSED; (void) tcp_output(tp); tcpstat.tcps_drops++; } else tcpstat.tcps_conndrops++; if (errno == ETIMEDOUT && tp->t_softerror) errno = tp->t_softerror; so->so_error = errno; return (tcp_close(tp)); } void tcp_discardcb(struct tcpcb *tp) { struct tseg_qent *q; struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ INP_LOCK_ASSERT(inp); /* * Make sure that all of our timers are stopped before we * delete the PCB. */ callout_stop(tp->tt_rexmt); callout_stop(tp->tt_persist); callout_stop(tp->tt_keep); callout_stop(tp->tt_2msl); callout_stop(tp->tt_delack); /* * If we got enough samples through the srtt filter, * save the rtt and rttvar in the routing entry. * 'Enough' is arbitrarily defined as 4 rtt samples. * 4 samples is enough for the srtt filter to converge * to within enough % of the correct value; fewer samples * and we could save a bogus rtt. The danger is not high * as tcp quickly recovers from everything. * XXX: Works very well but needs some more statistics! */ if (tp->t_rttupdated >= 4) { struct hc_metrics_lite metrics; u_long ssthresh; bzero(&metrics, sizeof(metrics)); /* * Update the ssthresh always when the conditions below * are satisfied. This gives us better new start value * for the congestion avoidance for new connections. * ssthresh is only set if packet loss occured on a session. * * XXXRW: 'so' may be NULL here, and/or socket buffer may be * being torn down. Ideally this code would not use 'so'. */ ssthresh = tp->snd_ssthresh; if (ssthresh != 0 && ssthresh < so->so_snd.sb_hiwat / 2) { /* * convert the limit from user data bytes to * packets then to packet data bytes. */ ssthresh = (ssthresh + tp->t_maxseg / 2) / tp->t_maxseg; if (ssthresh < 2) ssthresh = 2; ssthresh *= (u_long)(tp->t_maxseg + #ifdef INET6 (isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : #endif sizeof (struct tcpiphdr) #ifdef INET6 ) #endif ); } else ssthresh = 0; metrics.rmx_ssthresh = ssthresh; metrics.rmx_rtt = tp->t_srtt; metrics.rmx_rttvar = tp->t_rttvar; /* XXX: This wraps if the pipe is more than 4 Gbit per second */ metrics.rmx_bandwidth = tp->snd_bandwidth; metrics.rmx_cwnd = tp->snd_cwnd; metrics.rmx_sendpipe = 0; metrics.rmx_recvpipe = 0; tcp_hc_update(&inp->inp_inc, &metrics); } /* free the reassembly queue, if any */ while ((q = LIST_FIRST(&tp->t_segq)) != NULL) { LIST_REMOVE(q, tqe_q); m_freem(q->tqe_m); uma_zfree(tcp_reass_zone, q); tp->t_segqlen--; tcp_reass_qsize--; } tcp_free_sackholes(tp); inp->inp_ppcb = NULL; tp->t_inpcb = NULL; uma_zfree(tcpcb_zone, tp); } /* * Attempt to close a TCP control block, marking it as dropped, and freeing * the socket if we hold the only reference. */ struct tcpcb * tcp_close(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); in_pcbdrop(inp); tcpstat.tcps_closed++; KASSERT(inp->inp_socket != NULL, ("tcp_close: inp_socket NULL")); so = inp->inp_socket; soisdisconnected(so); if (inp->inp_vflag & INP_SOCKREF) { KASSERT(so->so_state & SS_PROTOREF, ("tcp_close: !SS_PROTOREF")); inp->inp_vflag &= ~INP_SOCKREF; INP_UNLOCK(inp); ACCEPT_LOCK(); SOCK_LOCK(so); so->so_state &= ~SS_PROTOREF; sofree(so); return (NULL); } return (tp); } void tcp_drain(void) { if (do_tcpdrain) { struct inpcb *inpb; struct tcpcb *tcpb; struct tseg_qent *te; /* * Walk the tcpbs, if existing, and flush the reassembly queue, * if there is one... * XXX: The "Net/3" implementation doesn't imply that the TCP * reassembly queue should be flushed, but in a situation * where we're really low on mbufs, this is potentially * usefull. */ INP_INFO_RLOCK(&tcbinfo); LIST_FOREACH(inpb, tcbinfo.listhead, inp_list) { if (inpb->inp_vflag & INP_TIMEWAIT) continue; INP_LOCK(inpb); if ((tcpb = intotcpcb(inpb)) != NULL) { while ((te = LIST_FIRST(&tcpb->t_segq)) != NULL) { LIST_REMOVE(te, tqe_q); m_freem(te->tqe_m); uma_zfree(tcp_reass_zone, te); tcpb->t_segqlen--; tcp_reass_qsize--; } tcp_clean_sackreport(tcpb); } INP_UNLOCK(inpb); } INP_INFO_RUNLOCK(&tcbinfo); } } /* * Notify a tcp user of an asynchronous error; * store error as soft error, but wake up user * (for now, won't do anything until can select for soft error). * * Do not wake up user since there currently is no mechanism for * reporting soft errors (yet - a kqueue filter may be added). */ static struct inpcb * tcp_notify(struct inpcb *inp, int error) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); if ((inp->inp_vflag & INP_TIMEWAIT) || (inp->inp_vflag & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_notify: tp == NULL")); /* * Ignore some errors if we are hooked up. * If connection hasn't completed, has retransmitted several times, * and receives a second error, give up now. This is better * than waiting a long time to establish a connection that * can never complete. */ if (tp->t_state == TCPS_ESTABLISHED && (error == EHOSTUNREACH || error == ENETUNREACH || error == EHOSTDOWN)) { return (inp); } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && tp->t_softerror) { tp = tcp_drop(tp, error); if (tp != NULL) return (inp); else return (NULL); } else { tp->t_softerror = error; return (inp); } #if 0 wakeup( &so->so_timeo); sorwakeup(so); sowwakeup(so); #endif } static int tcp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == NULL) { n = tcbinfo.ipi_count; req->oldidx = 2 * (sizeof xig) + (n + n/8) * sizeof(struct xtcpcb); return (0); } if (req->newptr != NULL) return (EPERM); /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&tcbinfo); gencnt = tcbinfo.ipi_gencnt; n = tcbinfo.ipi_count; INP_INFO_RUNLOCK(&tcbinfo); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + n * sizeof(struct xtcpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return (error); inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == NULL) return (ENOMEM); INP_INFO_RLOCK(&tcbinfo); for (inp = LIST_FIRST(tcbinfo.listhead), i = 0; inp != NULL && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_LOCK(inp); if (inp->inp_gencnt <= gencnt) { /* * XXX: This use of cr_cansee(), introduced with * TCP state changes, is not quite right, but for * now, better than nothing. */ if (inp->inp_vflag & INP_TIMEWAIT) { if (intotw(inp) != NULL) error = cr_cansee(req->td->td_ucred, intotw(inp)->tw_cred); else error = EINVAL; /* Skip this inp. */ } else error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error == 0) inp_list[i++] = inp; } INP_UNLOCK(inp); } INP_INFO_RUNLOCK(&tcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_LOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xtcpcb xt; void *inp_ppcb; bzero(&xt, sizeof(xt)); xt.xt_len = sizeof xt; /* XXX should avoid extra copy */ bcopy(inp, &xt.xt_inp, sizeof *inp); inp_ppcb = inp->inp_ppcb; if (inp_ppcb == NULL) bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); else if (inp->inp_vflag & INP_TIMEWAIT) { bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); xt.xt_tp.t_state = TCPS_TIME_WAIT; } else bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); if (inp->inp_socket != NULL) sotoxsocket(inp->inp_socket, &xt.xt_socket); else { bzero(&xt.xt_socket, sizeof xt.xt_socket); xt.xt_socket.xso_protocol = IPPROTO_TCP; } xt.xt_inp.inp_gencnt = inp->inp_gencnt; INP_UNLOCK(inp); error = SYSCTL_OUT(req, &xt, sizeof xt); } else INP_UNLOCK(inp); } if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ INP_INFO_RLOCK(&tcbinfo); xig.xig_gen = tcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = tcbinfo.ipi_count; INP_INFO_RUNLOCK(&tcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); static int tcp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error; error = priv_check_cred(req->td->td_ucred, PRIV_NETINET_GETCRED, SUSER_ALLOWJAIL); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); INP_INFO_RLOCK(&tcbinfo); inp = in_pcblookup_hash(&tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, 0, NULL); if (inp == NULL) { error = ENOENT; goto outunlocked; } INP_LOCK(inp); if (inp->inp_socket == NULL) { error = ENOENT; goto out; } error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error) goto out; cru2x(inp->inp_socket->so_cred, &xuc); out: INP_UNLOCK(inp); outunlocked: INP_INFO_RUNLOCK(&tcbinfo); if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp_getcred, "S,xucred", "Get the xucred of a TCP connection"); #ifdef INET6 static int tcp6_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in6 addrs[2]; struct inpcb *inp; int error, mapped = 0; error = priv_check_cred(req->td->td_ucred, PRIV_NETINET_GETCRED, SUSER_ALLOWJAIL); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); if ((error = sa6_embedscope(&addrs[0], ip6_use_defzone)) != 0 || (error = sa6_embedscope(&addrs[1], ip6_use_defzone)) != 0) { return (error); } if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) mapped = 1; else return (EINVAL); } INP_INFO_RLOCK(&tcbinfo); if (mapped == 1) inp = in_pcblookup_hash(&tcbinfo, *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], addrs[1].sin6_port, *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], addrs[0].sin6_port, 0, NULL); else inp = in6_pcblookup_hash(&tcbinfo, &addrs[1].sin6_addr, addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 0, NULL); if (inp == NULL) { error = ENOENT; goto outunlocked; } INP_LOCK(inp); if (inp->inp_socket == NULL) { error = ENOENT; goto out; } error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error) goto out; cru2x(inp->inp_socket->so_cred, &xuc); out: INP_UNLOCK(inp); outunlocked: INP_INFO_RUNLOCK(&tcbinfo); if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection"); #endif void tcp_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct ip *ip = vip; struct tcphdr *th; struct in_addr faddr; struct inpcb *inp; struct tcpcb *tp; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct icmp *icp; struct in_conninfo inc; tcp_seq icmp_tcp_seq; int mtu; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB || cmd == PRC_UNREACH_PORT || cmd == PRC_TIMXCEED_INTRANS) && ip) notify = tcp_drop_syn_sent; /* * Redirects don't need to be handled up here. */ else if (PRC_IS_REDIRECT(cmd)) return; /* * Source quench is depreciated. */ else if (cmd == PRC_QUENCH) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * XXX: We never get this from ICMP, otherwise it makes an * excellent DoS attack on machines with many connections. */ else if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip != NULL) { icp = (struct icmp *)((caddr_t)ip - offsetof(struct icmp, icmp_ip)); th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); INP_INFO_WLOCK(&tcbinfo); inp = in_pcblookup_hash(&tcbinfo, faddr, th->th_dport, ip->ip_src, th->th_sport, 0, NULL); if (inp != NULL) { INP_LOCK(inp); if (!(inp->inp_vflag & INP_TIMEWAIT) && !(inp->inp_vflag & INP_DROPPED) && !(inp->inp_socket == NULL)) { icmp_tcp_seq = htonl(th->th_seq); tp = intotcpcb(inp); if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) && SEQ_LT(icmp_tcp_seq, tp->snd_max)) { if (cmd == PRC_MSGSIZE) { /* * MTU discovery: * If we got a needfrag set the MTU * in the route to the suggested new * value (if given) and then notify. */ bzero(&inc, sizeof(inc)); inc.inc_flags = 0; /* IPv4 */ inc.inc_faddr = faddr; mtu = ntohs(icp->icmp_nextmtu); /* * If no alternative MTU was * proposed, try the next smaller * one. ip->ip_len has already * been swapped in icmp_input(). */ if (!mtu) mtu = ip_next_mtu(ip->ip_len, 1); if (mtu < max(296, (tcp_minmss) + sizeof(struct tcpiphdr))) mtu = 0; if (!mtu) mtu = tcp_mssdflt + sizeof(struct tcpiphdr); /* * Only cache the the MTU if it * is smaller than the interface * or route MTU. tcp_mtudisc() * will do right thing by itself. */ if (mtu <= tcp_maxmtu(&inc, NULL)) tcp_hc_updatemtu(&inc, mtu); } inp = (*notify)(inp, inetctlerrmap[cmd]); } } if (inp != NULL) INP_UNLOCK(inp); } else { inc.inc_fport = th->th_dport; inc.inc_lport = th->th_sport; inc.inc_faddr = faddr; inc.inc_laddr = ip->ip_src; #ifdef INET6 inc.inc_isipv6 = 0; #endif syncache_unreach(&inc, th); } INP_INFO_WUNLOCK(&tcbinfo); } else in_pcbnotifyall(&tcbinfo, faddr, inetctlerrmap[cmd], notify); } #ifdef INET6 void tcp6_ctlinput(int cmd, struct sockaddr *sa, void *d) { struct tcphdr th; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct ip6_hdr *ip6; struct mbuf *m; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; int off; struct tcp_portonly { u_int16_t th_sport; u_int16_t th_dport; } *thp; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; else if (!PRC_IS_REDIRECT(cmd) && ((unsigned)cmd >= PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) return; /* Source quench is depreciated. */ else if (cmd == PRC_QUENCH) return; /* if the parameter is from icmp6, decode it. */ if (d != NULL) { ip6cp = (struct ip6ctlparam *)d; m = ip6cp->ip6c_m; ip6 = ip6cp->ip6c_ip6; off = ip6cp->ip6c_off; sa6_src = ip6cp->ip6c_src; } else { m = NULL; ip6 = NULL; off = 0; /* fool gcc */ sa6_src = &sa6_any; } if (ip6 != NULL) { struct in_conninfo inc; /* * XXX: We assume that when IPV6 is non NULL, * M and OFF are valid. */ /* check if we can safely examine src and dst ports */ if (m->m_pkthdr.len < off + sizeof(*thp)) return; bzero(&th, sizeof(th)); m_copydata(m, off, sizeof(*thp), (caddr_t)&th); in6_pcbnotify(&tcbinfo, sa, th.th_dport, (struct sockaddr *)ip6cp->ip6c_src, th.th_sport, cmd, NULL, notify); inc.inc_fport = th.th_dport; inc.inc_lport = th.th_sport; inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr; inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr; inc.inc_isipv6 = 1; INP_INFO_WLOCK(&tcbinfo); syncache_unreach(&inc, &th); INP_INFO_WUNLOCK(&tcbinfo); } else in6_pcbnotify(&tcbinfo, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify); } #endif /* INET6 */ /* * Following is where TCP initial sequence number generation occurs. * * There are two places where we must use initial sequence numbers: * 1. In SYN-ACK packets. * 2. In SYN packets. * * All ISNs for SYN-ACK packets are generated by the syncache. See * tcp_syncache.c for details. * * The ISNs in SYN packets must be monotonic; TIME_WAIT recycling * depends on this property. In addition, these ISNs should be * unguessable so as to prevent connection hijacking. To satisfy * the requirements of this situation, the algorithm outlined in * RFC 1948 is used, with only small modifications. * * Implementation details: * * Time is based off the system timer, and is corrected so that it * increases by one megabyte per second. This allows for proper * recycling on high speed LANs while still leaving over an hour * before rollover. * * As reading the *exact* system time is too expensive to be done * whenever setting up a TCP connection, we increment the time * offset in two ways. First, a small random positive increment * is added to isn_offset for each connection that is set up. * Second, the function tcp_isn_tick fires once per clock tick * and increments isn_offset as necessary so that sequence numbers * are incremented at approximately ISN_BYTES_PER_SECOND. The * random positive increments serve only to ensure that the same * exact sequence number is never sent out twice (as could otherwise * happen when a port is recycled in less than the system tick * interval.) * * net.inet.tcp.isn_reseed_interval controls the number of seconds * between seeding of isn_secret. This is normally set to zero, * as reseeding should not be necessary. * * Locking of the global variables isn_secret, isn_last_reseed, isn_offset, * isn_offset_old, and isn_ctx is performed using the TCP pcbinfo lock. In * general, this means holding an exclusive (write) lock. */ #define ISN_BYTES_PER_SECOND 1048576 #define ISN_STATIC_INCREMENT 4096 #define ISN_RANDOM_INCREMENT (4096 - 1) static u_char isn_secret[32]; static int isn_last_reseed; static u_int32_t isn_offset, isn_offset_old; static MD5_CTX isn_ctx; tcp_seq tcp_new_isn(struct tcpcb *tp) { u_int32_t md5_buffer[4]; tcp_seq new_isn; INP_LOCK_ASSERT(tp->t_inpcb); ISN_LOCK(); /* Seed if this is the first use, reseed if requested. */ if ((isn_last_reseed == 0) || ((tcp_isn_reseed_interval > 0) && (((u_int)isn_last_reseed + (u_int)tcp_isn_reseed_interval*hz) < (u_int)ticks))) { read_random(&isn_secret, sizeof(isn_secret)); isn_last_reseed = ticks; } /* Compute the md5 hash and return the ISN. */ MD5Init(&isn_ctx); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short)); #ifdef INET6 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr, sizeof(struct in6_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr, sizeof(struct in6_addr)); } else #endif { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr, sizeof(struct in_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr, sizeof(struct in_addr)); } MD5Update(&isn_ctx, (u_char *) &isn_secret, sizeof(isn_secret)); MD5Final((u_char *) &md5_buffer, &isn_ctx); new_isn = (tcp_seq) md5_buffer[0]; isn_offset += ISN_STATIC_INCREMENT + (arc4random() & ISN_RANDOM_INCREMENT); new_isn += isn_offset; ISN_UNLOCK(); return (new_isn); } /* * Increment the offset to the next ISN_BYTES_PER_SECOND / hz boundary * to keep time flowing at a relatively constant rate. If the random * increments have already pushed us past the projected offset, do nothing. */ static void tcp_isn_tick(void *xtp) { u_int32_t projected_offset; ISN_LOCK(); projected_offset = isn_offset_old + ISN_BYTES_PER_SECOND / 100; if (projected_offset > isn_offset) isn_offset = projected_offset; isn_offset_old = isn_offset; callout_reset(&isn_callout, hz/100, tcp_isn_tick, NULL); ISN_UNLOCK(); } /* * When a specific ICMP unreachable message is received and the * connection state is SYN-SENT, drop the connection. This behavior * is controlled by the icmp_may_rst sysctl. */ struct inpcb * tcp_drop_syn_sent(struct inpcb *inp, int errno) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); if ((inp->inp_vflag & INP_TIMEWAIT) || (inp->inp_vflag & INP_DROPPED)) return (inp); tp = intotcpcb(inp); if (tp->t_state != TCPS_SYN_SENT) return (inp); tp = tcp_drop(tp, errno); if (tp != NULL) return (inp); else return (NULL); } /* * When `need fragmentation' ICMP is received, update our idea of the MSS * based on the new value in the route. Also nudge TCP to send something, * since we know the packet we just sent was dropped. * This duplicates some code in the tcp_mss() function in tcp_input.c. */ struct inpcb * tcp_mtudisc(struct inpcb *inp, int errno) { struct tcpcb *tp; struct socket *so = inp->inp_socket; u_int maxmtu; u_int romtu; int mss; #ifdef INET6 int isipv6; #endif /* INET6 */ INP_LOCK_ASSERT(inp); if ((inp->inp_vflag & INP_TIMEWAIT) || (inp->inp_vflag & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_mtudisc: tp == NULL")); #ifdef INET6 isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0; #endif maxmtu = tcp_hc_getmtu(&inp->inp_inc); /* IPv4 and IPv6 */ romtu = #ifdef INET6 isipv6 ? tcp_maxmtu6(&inp->inp_inc, NULL) : #endif /* INET6 */ tcp_maxmtu(&inp->inp_inc, NULL); if (!maxmtu) maxmtu = romtu; else maxmtu = min(maxmtu, romtu); if (!maxmtu) { tp->t_maxopd = tp->t_maxseg = #ifdef INET6 isipv6 ? tcp_v6mssdflt : #endif /* INET6 */ tcp_mssdflt; return (inp); } mss = maxmtu - #ifdef INET6 (isipv6 ? sizeof(struct ip6_hdr) + sizeof(struct tcphdr) : #endif /* INET6 */ sizeof(struct tcpiphdr) #ifdef INET6 ) #endif /* INET6 */ ; /* * XXX - The above conditional probably violates the TCP * spec. The problem is that, since we don't know the * other end's MSS, we are supposed to use a conservative * default. But, if we do that, then MTU discovery will * never actually take place, because the conservative * default is much less than the MTUs typically seen * on the Internet today. For the moment, we'll sweep * this under the carpet. * * The conservative default might not actually be a problem * if the only case this occurs is when sending an initial * SYN with options and data to a host we've never talked * to before. Then, they will reply with an MSS value which * will get recorded and the new parameters should get * recomputed. For Further Study. */ if (tp->t_maxopd <= mss) return (inp); tp->t_maxopd = mss; if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP) mss -= TCPOLEN_TSTAMP_APPA; #if (MCLBYTES & (MCLBYTES - 1)) == 0 if (mss > MCLBYTES) mss &= ~(MCLBYTES-1); #else if (mss > MCLBYTES) mss = mss / MCLBYTES * MCLBYTES; #endif if (so->so_snd.sb_hiwat < mss) mss = so->so_snd.sb_hiwat; tp->t_maxseg = mss; tcpstat.tcps_mturesent++; tp->t_rtttime = 0; tp->snd_nxt = tp->snd_una; tcp_free_sackholes(tp); tp->snd_recover = tp->snd_max; if (tp->sack_enable) EXIT_FASTRECOVERY(tp); tcp_output(tp); return (inp); } /* * Look-up the routing entry to the peer of this inpcb. If no route * is found and it cannot be allocated, then return NULL. This routine * is called by TCP routines that access the rmx structure and by tcp_mss * to get the interface MTU. */ u_long tcp_maxmtu(struct in_conninfo *inc, int *flags) { struct route sro; struct sockaddr_in *dst; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer")); bzero(&sro, sizeof(sro)); if (inc->inc_faddr.s_addr != INADDR_ANY) { dst = (struct sockaddr_in *)&sro.ro_dst; dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = inc->inc_faddr; rtalloc_ign(&sro, RTF_CLONING); } if (sro.ro_rt != NULL) { ifp = sro.ro_rt->rt_ifp; if (sro.ro_rt->rt_rmx.rmx_mtu == 0) maxmtu = ifp->if_mtu; else maxmtu = min(sro.ro_rt->rt_rmx.rmx_mtu, ifp->if_mtu); /* Report additional interface capabilities. */ if (flags != NULL) { if (ifp->if_capenable & IFCAP_TSO4 && ifp->if_hwassist & CSUM_TSO) *flags |= CSUM_TSO; } RTFREE(sro.ro_rt); } return (maxmtu); } #ifdef INET6 u_long tcp_maxmtu6(struct in_conninfo *inc, int *flags) { struct route_in6 sro6; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu6 with NULL in_conninfo pointer")); bzero(&sro6, sizeof(sro6)); if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) { sro6.ro_dst.sin6_family = AF_INET6; sro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); sro6.ro_dst.sin6_addr = inc->inc6_faddr; rtalloc_ign((struct route *)&sro6, RTF_CLONING); } if (sro6.ro_rt != NULL) { ifp = sro6.ro_rt->rt_ifp; if (sro6.ro_rt->rt_rmx.rmx_mtu == 0) maxmtu = IN6_LINKMTU(sro6.ro_rt->rt_ifp); else maxmtu = min(sro6.ro_rt->rt_rmx.rmx_mtu, IN6_LINKMTU(sro6.ro_rt->rt_ifp)); /* Report additional interface capabilities. */ if (flags != NULL) { if (ifp->if_capenable & IFCAP_TSO6 && ifp->if_hwassist & CSUM_TSO) *flags |= CSUM_TSO; } RTFREE(sro6.ro_rt); } return (maxmtu); } #endif /* INET6 */ #ifdef IPSEC /* compute ESP/AH header size for TCP, including outer IP header. */ size_t ipsec_hdrsiz_tcp(struct tcpcb *tp) { struct inpcb *inp; struct mbuf *m; size_t hdrsiz; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; #endif struct tcphdr *th; if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL)) return (0); MGETHDR(m, M_DONTWAIT, MT_DATA); if (!m) return (0); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); tcpip_fillheaders(inp, ip6, th); hdrsiz = ipsec6_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } else #endif /* INET6 */ { ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); tcpip_fillheaders(inp, ip, th); hdrsiz = ipsec4_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } m_free(m); return (hdrsiz); } #endif /*IPSEC*/ /* * Move a TCP connection into TIME_WAIT state. * tcbinfo is locked. * inp is locked, and is unlocked before returning. */ void tcp_twstart(struct tcpcb *tp) { struct tcptw *tw; struct inpcb *inp = tp->t_inpcb; int acknow; struct socket *so; INP_INFO_WLOCK_ASSERT(&tcbinfo); /* tcp_timer_2msl_reset(). */ INP_LOCK_ASSERT(inp); if (nolocaltimewait && in_localip(inp->inp_faddr)) { tp = tcp_close(tp); if (tp != NULL) INP_UNLOCK(inp); return; } tw = uma_zalloc(tcptw_zone, M_NOWAIT); if (tw == NULL) { tw = tcp_timer_2msl_tw(1); if (tw == NULL) { tp = tcp_close(tp); if (tp != NULL) INP_UNLOCK(inp); return; } } tw->tw_inpcb = inp; /* * Recover last window size sent. */ tw->last_win = (tp->rcv_adv - tp->rcv_nxt) >> tp->rcv_scale; /* * Set t_recent if timestamps are used on the connection. */ if ((tp->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP|TF_NOOPT)) == (TF_REQ_TSTMP|TF_RCVD_TSTMP)) tw->t_recent = tp->ts_recent; else tw->t_recent = 0; tw->snd_nxt = tp->snd_nxt; tw->rcv_nxt = tp->rcv_nxt; tw->iss = tp->iss; tw->irs = tp->irs; tw->t_starttime = tp->t_starttime; tw->tw_time = 0; /* XXX * If this code will * be used for fin-wait-2 state also, then we may need * a ts_recent from the last segment. */ acknow = tp->t_flags & TF_ACKNOW; /* * First, discard tcpcb state, which includes stopping its timers and * freeing it. tcp_discardcb() used to also release the inpcb, but * that work is now done in the caller. * * Note: soisdisconnected() call used to be made in tcp_discardcb(), * and might not be needed here any longer. */ tcp_discardcb(tp); so = inp->inp_socket; soisdisconnected(so); SOCK_LOCK(so); tw->tw_cred = crhold(so->so_cred); tw->tw_so_options = so->so_options; SOCK_UNLOCK(so); if (acknow) tcp_twrespond(tw, TH_ACK); inp->inp_ppcb = tw; inp->inp_vflag |= INP_TIMEWAIT; tcp_timer_2msl_reset(tw, 0); /* * If the inpcb owns the sole reference to the socket, then we can * detach and free the socket as it is not needed in time wait. */ if (inp->inp_vflag & INP_SOCKREF) { KASSERT(so->so_state & SS_PROTOREF, ("tcp_twstart: !SS_PROTOREF")); inp->inp_vflag &= ~INP_SOCKREF; INP_UNLOCK(inp); ACCEPT_LOCK(); SOCK_LOCK(so); so->so_state &= ~SS_PROTOREF; sofree(so); } else INP_UNLOCK(inp); } #if 0 /* * The appromixate rate of ISN increase of Microsoft TCP stacks; * the actual rate is slightly higher due to the addition of * random positive increments. * * Most other new OSes use semi-randomized ISN values, so we * do not need to worry about them. */ #define MS_ISN_BYTES_PER_SECOND 250000 /* * Determine if the ISN we will generate has advanced beyond the last * sequence number used by the previous connection. If so, indicate * that it is safe to recycle this tw socket by returning 1. */ int tcp_twrecycleable(struct tcptw *tw) { tcp_seq new_iss = tw->iss; tcp_seq new_irs = tw->irs; INP_INFO_WLOCK_ASSERT(&tcbinfo); new_iss += (ticks - tw->t_starttime) * (ISN_BYTES_PER_SECOND / hz); new_irs += (ticks - tw->t_starttime) * (MS_ISN_BYTES_PER_SECOND / hz); if (SEQ_GT(new_iss, tw->snd_nxt) && SEQ_GT(new_irs, tw->rcv_nxt)) return (1); else return (0); } #endif void tcp_twclose(struct tcptw *tw, int reuse) { struct socket *so; struct inpcb *inp; /* * At this point, we are in one of two situations: * * (1) We have no socket, just an inpcb<->twtcp pair. We can free * all state. * * (2) We have a socket -- if we own a reference, release it and * notify the socket layer. */ inp = tw->tw_inpcb; KASSERT((inp->inp_vflag & INP_TIMEWAIT), ("tcp_twclose: !timewait")); KASSERT(intotw(inp) == tw, ("tcp_twclose: inp_ppcb != tw")); INP_INFO_WLOCK_ASSERT(&tcbinfo); /* tcp_timer_2msl_stop(). */ INP_LOCK_ASSERT(inp); tw->tw_inpcb = NULL; tcp_timer_2msl_stop(tw); inp->inp_ppcb = NULL; in_pcbdrop(inp); so = inp->inp_socket; if (so != NULL) { /* * If there's a socket, handle two cases: first, we own a * strong reference, which we will now release, or we don't * in which case another reference exists (XXXRW: think * about this more), and we don't need to take action. */ if (inp->inp_vflag & INP_SOCKREF) { inp->inp_vflag &= ~INP_SOCKREF; INP_UNLOCK(inp); ACCEPT_LOCK(); SOCK_LOCK(so); KASSERT(so->so_state & SS_PROTOREF, ("tcp_twclose: INP_SOCKREF && !SS_PROTOREF")); so->so_state &= ~SS_PROTOREF; sofree(so); } else { /* * If we don't own the only reference, the socket and * inpcb need to be left around to be handled by * tcp_usr_detach() later. */ INP_UNLOCK(inp); } } else { #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) in6_pcbfree(inp); else #endif in_pcbfree(inp); } tcpstat.tcps_closed++; crfree(tw->tw_cred); tw->tw_cred = NULL; if (reuse) return; uma_zfree(tcptw_zone, tw); } int tcp_twrespond(struct tcptw *tw, int flags) { struct inpcb *inp = tw->tw_inpcb; struct tcphdr *th; struct mbuf *m; struct ip *ip = NULL; u_int8_t *optp; u_int hdrlen, optlen; int error; #ifdef INET6 struct ip6_hdr *ip6 = NULL; int isipv6 = inp->inp_inc.inc_isipv6; #endif INP_LOCK_ASSERT(inp); m = m_gethdr(M_DONTWAIT, MT_DATA); if (m == NULL) return (ENOBUFS); m->m_data += max_linkhdr; #ifdef MAC mac_create_mbuf_from_inpcb(inp, m); #endif #ifdef INET6 if (isipv6) { hdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); tcpip_fillheaders(inp, ip6, th); } else #endif { hdrlen = sizeof(struct tcpiphdr); ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); tcpip_fillheaders(inp, ip, th); } optp = (u_int8_t *)(th + 1); /* * Send a timestamp and echo-reply if both our side and our peer * have sent timestamps in our SYN's and this is not a RST. */ if (tw->t_recent && flags == TH_ACK) { u_int32_t *lp = (u_int32_t *)optp; /* Form timestamp option as shown in appendix A of RFC 1323. */ *lp++ = htonl(TCPOPT_TSTAMP_HDR); *lp++ = htonl(ticks); *lp = htonl(tw->t_recent); optp += TCPOLEN_TSTAMP_APPA; } optlen = optp - (u_int8_t *)(th + 1); m->m_len = hdrlen + optlen; m->m_pkthdr.len = m->m_len; KASSERT(max_linkhdr + m->m_len <= MHLEN, ("tcptw: mbuf too small")); th->th_seq = htonl(tw->snd_nxt); th->th_ack = htonl(tw->rcv_nxt); th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; th->th_flags = flags; th->th_win = htons(tw->last_win); #ifdef INET6 if (isipv6) { th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), sizeof(struct tcphdr) + optlen); ip6->ip6_hlim = in6_selecthlim(inp, NULL); error = ip6_output(m, inp->in6p_outputopts, NULL, (tw->tw_so_options & SO_DONTROUTE), NULL, NULL, inp); } else #endif { th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons(sizeof(struct tcphdr) + optlen + IPPROTO_TCP)); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); ip->ip_len = m->m_pkthdr.len; if (path_mtu_discovery) ip->ip_off |= IP_DF; error = ip_output(m, inp->inp_options, NULL, ((tw->tw_so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0), NULL, inp); } if (flags & TH_ACK) tcpstat.tcps_sndacks++; else tcpstat.tcps_sndctrl++; tcpstat.tcps_sndtotal++; return (error); } /* * TCP BANDWIDTH DELAY PRODUCT WINDOW LIMITING * * This code attempts to calculate the bandwidth-delay product as a * means of determining the optimal window size to maximize bandwidth, * minimize RTT, and avoid the over-allocation of buffers on interfaces and * routers. This code also does a fairly good job keeping RTTs in check * across slow links like modems. We implement an algorithm which is very * similar (but not meant to be) TCP/Vegas. The code operates on the * transmitter side of a TCP connection and so only effects the transmit * side of the connection. * * BACKGROUND: TCP makes no provision for the management of buffer space * at the end points or at the intermediate routers and switches. A TCP * stream, whether using NewReno or not, will eventually buffer as * many packets as it is able and the only reason this typically works is * due to the fairly small default buffers made available for a connection * (typicaly 16K or 32K). As machines use larger windows and/or window * scaling it is now fairly easy for even a single TCP connection to blow-out * all available buffer space not only on the local interface, but on * intermediate routers and switches as well. NewReno makes a misguided * attempt to 'solve' this problem by waiting for an actual failure to occur, * then backing off, then steadily increasing the window again until another * failure occurs, ad-infinitum. This results in terrible oscillation that * is only made worse as network loads increase and the idea of intentionally * blowing out network buffers is, frankly, a terrible way to manage network * resources. * * It is far better to limit the transmit window prior to the failure * condition being achieved. There are two general ways to do this: First * you can 'scan' through different transmit window sizes and locate the * point where the RTT stops increasing, indicating that you have filled the * pipe, then scan backwards until you note that RTT stops decreasing, then * repeat ad-infinitum. This method works in principle but has severe * implementation issues due to RTT variances, timer granularity, and * instability in the algorithm which can lead to many false positives and * create oscillations as well as interact badly with other TCP streams * implementing the same algorithm. * * The second method is to limit the window to the bandwidth delay product * of the link. This is the method we implement. RTT variances and our * own manipulation of the congestion window, bwnd, can potentially * destabilize the algorithm. For this reason we have to stabilize the * elements used to calculate the window. We do this by using the minimum * observed RTT, the long term average of the observed bandwidth, and * by adding two segments worth of slop. It isn't perfect but it is able * to react to changing conditions and gives us a very stable basis on * which to extend the algorithm. */ void tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq) { u_long bw; u_long bwnd; int save_ticks; INP_LOCK_ASSERT(tp->t_inpcb); /* * If inflight_enable is disabled in the middle of a tcp connection, * make sure snd_bwnd is effectively disabled. */ if (tcp_inflight_enable == 0 || tp->t_rttlow < tcp_inflight_rttthresh) { tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_bandwidth = 0; return; } /* * Figure out the bandwidth. Due to the tick granularity this * is a very rough number and it MUST be averaged over a fairly * long period of time. XXX we need to take into account a link * that is not using all available bandwidth, but for now our * slop will ramp us up if this case occurs and the bandwidth later * increases. * * Note: if ticks rollover 'bw' may wind up negative. We must * effectively reset t_bw_rtttime for this case. */ save_ticks = ticks; if ((u_int)(save_ticks - tp->t_bw_rtttime) < 1) return; bw = (int64_t)(ack_seq - tp->t_bw_rtseq) * hz / (save_ticks - tp->t_bw_rtttime); tp->t_bw_rtttime = save_ticks; tp->t_bw_rtseq = ack_seq; if (tp->t_bw_rtttime == 0 || (int)bw < 0) return; bw = ((int64_t)tp->snd_bandwidth * 15 + bw) >> 4; tp->snd_bandwidth = bw; /* * Calculate the semi-static bandwidth delay product, plus two maximal * segments. The additional slop puts us squarely in the sweet * spot and also handles the bandwidth run-up case and stabilization. * Without the slop we could be locking ourselves into a lower * bandwidth. * * Situations Handled: * (1) Prevents over-queueing of packets on LANs, especially on * high speed LANs, allowing larger TCP buffers to be * specified, and also does a good job preventing * over-queueing of packets over choke points like modems * (at least for the transmit side). * * (2) Is able to handle changing network loads (bandwidth * drops so bwnd drops, bandwidth increases so bwnd * increases). * * (3) Theoretically should stabilize in the face of multiple * connections implementing the same algorithm (this may need * a little work). * * (4) Stability value (defaults to 20 = 2 maximal packets) can * be adjusted with a sysctl but typically only needs to be * on very slow connections. A value no smaller then 5 * should be used, but only reduce this default if you have * no other choice. */ #define USERTT ((tp->t_srtt + tp->t_rttbest) / 2) bwnd = (int64_t)bw * USERTT / (hz << TCP_RTT_SHIFT) + tcp_inflight_stab * tp->t_maxseg / 10; #undef USERTT if (tcp_inflight_debug > 0) { static int ltime; if ((u_int)(ticks - ltime) >= hz / tcp_inflight_debug) { ltime = ticks; printf("%p bw %ld rttbest %d srtt %d bwnd %ld\n", tp, bw, tp->t_rttbest, tp->t_srtt, bwnd ); } } if ((long)bwnd < tcp_inflight_min) bwnd = tcp_inflight_min; if (bwnd > tcp_inflight_max) bwnd = tcp_inflight_max; if ((long)bwnd < tp->t_maxseg * 2) bwnd = tp->t_maxseg * 2; tp->snd_bwnd = bwnd; } #ifdef TCP_SIGNATURE /* * Callback function invoked by m_apply() to digest TCP segment data * contained within an mbuf chain. */ static int tcp_signature_apply(void *fstate, void *data, u_int len) { MD5Update(fstate, (u_char *)data, len); return (0); } /* * Compute TCP-MD5 hash of a TCPv4 segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * off0 offset to TCP header within the mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND) * * We do this over ip, tcphdr, segment data, and the key in the SADB. * When called from tcp_input(), we can be sure that th_sum has been * zeroed out and verified already. * * This function is for IPv4 use only. Calling this function with an * IPv6 packet in the mbuf chain will yield undefined results. * * Return 0 if successful, otherwise return -1. * * XXX The key is retrieved from the system's PF_KEY SADB, by keying a * search with the destination IP address, and a 'magic SPI' to be * determined by the application. This is hardcoded elsewhere to 1179 * right now. Another branch of this code exists which uses the SPD to * specify per-application flows but it is unstable. */ int tcp_signature_compute(struct mbuf *m, int off0, int len, int optlen, u_char *buf, u_int direction) { union sockaddr_union dst; struct ippseudo ippseudo; MD5_CTX ctx; int doff; struct ip *ip; struct ipovly *ipovly; struct secasvar *sav; struct tcphdr *th; u_short savecsum; KASSERT(m != NULL, ("NULL mbuf chain")); KASSERT(buf != NULL, ("NULL signature pointer")); /* Extract the destination from the IP header in the mbuf. */ ip = mtod(m, struct ip *); bzero(&dst, sizeof(union sockaddr_union)); dst.sa.sa_len = sizeof(struct sockaddr_in); dst.sa.sa_family = AF_INET; dst.sin.sin_addr = (direction == IPSEC_DIR_INBOUND) ? ip->ip_src : ip->ip_dst; /* Look up an SADB entry which matches the address of the peer. */ sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); if (sav == NULL) { printf("%s: SADB lookup failed for %s\n", __func__, inet_ntoa(dst.sin.sin_addr)); return (EINVAL); } MD5Init(&ctx); ipovly = (struct ipovly *)ip; th = (struct tcphdr *)((u_char *)ip + off0); doff = off0 + sizeof(struct tcphdr) + optlen; /* * Step 1: Update MD5 hash with IP pseudo-header. * * XXX The ippseudo header MUST be digested in network byte order, * or else we'll fail the regression test. Assume all fields we've * been doing arithmetic on have been in host byte order. * XXX One cannot depend on ipovly->ih_len here. When called from * tcp_output(), the underlying ip_len member has not yet been set. */ ippseudo.ippseudo_src = ipovly->ih_src; ippseudo.ippseudo_dst = ipovly->ih_dst; ippseudo.ippseudo_pad = 0; ippseudo.ippseudo_p = IPPROTO_TCP; ippseudo.ippseudo_len = htons(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&ippseudo, sizeof(struct ippseudo)); /* * Step 2: Update MD5 hash with TCP header, excluding options. * The TCP checksum must be set to zero. */ savecsum = th->th_sum; th->th_sum = 0; MD5Update(&ctx, (char *)th, sizeof(struct tcphdr)); th->th_sum = savecsum; /* * Step 3: Update MD5 hash with TCP segment data. * Use m_apply() to avoid an early m_pullup(). */ if (len > 0) m_apply(m, doff, len, tcp_signature_apply, &ctx); /* * Step 4: Update MD5 hash with shared secret. */ MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth)); MD5Final(buf, &ctx); key_sa_recordxfer(sav, m); KEY_FREESAV(&sav); return (0); } #endif /* TCP_SIGNATURE */ static int sysctl_drop(SYSCTL_HANDLER_ARGS) { /* addrs[0] is a foreign socket, addrs[1] is a local one. */ struct sockaddr_storage addrs[2]; struct inpcb *inp; struct tcpcb *tp; struct tcptw *tw; struct sockaddr_in *fin, *lin; #ifdef INET6 struct sockaddr_in6 *fin6, *lin6; struct in6_addr f6, l6; #endif int error; inp = NULL; fin = lin = NULL; #ifdef INET6 fin6 = lin6 = NULL; #endif error = 0; if (req->oldptr != NULL || req->oldlen != 0) return (EINVAL); if (req->newptr == NULL) return (EPERM); if (req->newlen < sizeof(addrs)) return (ENOMEM); error = SYSCTL_IN(req, &addrs, sizeof(addrs)); if (error) return (error); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: fin6 = (struct sockaddr_in6 *)&addrs[0]; lin6 = (struct sockaddr_in6 *)&addrs[1]; if (fin6->sin6_len != sizeof(struct sockaddr_in6) || lin6->sin6_len != sizeof(struct sockaddr_in6)) return (EINVAL); if (IN6_IS_ADDR_V4MAPPED(&fin6->sin6_addr)) { if (!IN6_IS_ADDR_V4MAPPED(&lin6->sin6_addr)) return (EINVAL); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[0]); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[1]); fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; break; } error = sa6_embedscope(fin6, ip6_use_defzone); if (error) return (error); error = sa6_embedscope(lin6, ip6_use_defzone); if (error) return (error); break; #endif case AF_INET: fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; if (fin->sin_len != sizeof(struct sockaddr_in) || lin->sin_len != sizeof(struct sockaddr_in)) return (EINVAL); break; default: return (EINVAL); } INP_INFO_WLOCK(&tcbinfo); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: inp = in6_pcblookup_hash(&tcbinfo, &f6, fin6->sin6_port, &l6, lin6->sin6_port, 0, NULL); break; #endif case AF_INET: inp = in_pcblookup_hash(&tcbinfo, fin->sin_addr, fin->sin_port, lin->sin_addr, lin->sin_port, 0, NULL); break; } if (inp != NULL) { INP_LOCK(inp); if (inp->inp_vflag & INP_TIMEWAIT) { /* * XXXRW: There currently exists a state where an * inpcb is present, but its timewait state has been * discarded. For now, don't allow dropping of this * type of inpcb. */ tw = intotw(inp); if (tw != NULL) tcp_twclose(tw, 0); } else if (!(inp->inp_vflag & INP_DROPPED) && !(inp->inp_socket->so_options & SO_ACCEPTCONN)) { tp = intotcpcb(inp); tcp_drop(tp, ECONNABORTED); } INP_UNLOCK(inp); } else error = ESRCH; INP_INFO_WUNLOCK(&tcbinfo); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_DROP, drop, CTLTYPE_STRUCT|CTLFLAG_WR|CTLFLAG_SKIP, NULL, 0, sysctl_drop, "", "Drop TCP connection"); Index: head/sys/netinet/tcp_timer.c =================================================================== --- head/sys/netinet/tcp_timer.c (revision 167035) +++ head/sys/netinet/tcp_timer.c (revision 167036) @@ -1,600 +1,620 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_timer.c 8.2 (Berkeley) 5/24/95 * $FreeBSD$ */ #include "opt_inet6.h" #include "opt_tcpdebug.h" #include "opt_tcp_sack.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #ifdef TCPDEBUG #include #endif int tcp_keepinit; SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINIT, keepinit, CTLTYPE_INT|CTLFLAG_RW, &tcp_keepinit, 0, sysctl_msec_to_ticks, "I", ""); int tcp_keepidle; SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPIDLE, keepidle, CTLTYPE_INT|CTLFLAG_RW, &tcp_keepidle, 0, sysctl_msec_to_ticks, "I", ""); int tcp_keepintvl; SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINTVL, keepintvl, CTLTYPE_INT|CTLFLAG_RW, &tcp_keepintvl, 0, sysctl_msec_to_ticks, "I", ""); int tcp_delacktime; SYSCTL_PROC(_net_inet_tcp, TCPCTL_DELACKTIME, delacktime, CTLTYPE_INT|CTLFLAG_RW, &tcp_delacktime, 0, sysctl_msec_to_ticks, "I", "Time before a delayed ACK is sent"); int tcp_msl; SYSCTL_PROC(_net_inet_tcp, OID_AUTO, msl, CTLTYPE_INT|CTLFLAG_RW, &tcp_msl, 0, sysctl_msec_to_ticks, "I", "Maximum segment lifetime"); int tcp_rexmit_min; SYSCTL_PROC(_net_inet_tcp, OID_AUTO, rexmit_min, CTLTYPE_INT|CTLFLAG_RW, &tcp_rexmit_min, 0, sysctl_msec_to_ticks, "I", "Minimum Retransmission Timeout"); int tcp_rexmit_slop; SYSCTL_PROC(_net_inet_tcp, OID_AUTO, rexmit_slop, CTLTYPE_INT|CTLFLAG_RW, &tcp_rexmit_slop, 0, sysctl_msec_to_ticks, "I", "Retransmission Timer Slop"); static int always_keepalive = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, always_keepalive, CTLFLAG_RW, &always_keepalive , 0, "Assume SO_KEEPALIVE on all TCP connections"); +int tcp_fast_finwait2_recycle = 0; +SYSCTL_INT(_net_inet_tcp, OID_AUTO, fast_finwait2_recycle, CTLFLAG_RW, + &tcp_fast_finwait2_recycle, 0, "Recycle closed FIN_WAIT_2 connections faster"); + +int tcp_finwait2_timeout; +SYSCTL_PROC(_net_inet_tcp, OID_AUTO, finwait2_timeout, CTLTYPE_INT|CTLFLAG_RW, + &tcp_finwait2_timeout, 0, sysctl_msec_to_ticks, "I", ""); + + static int tcp_keepcnt = TCPTV_KEEPCNT; /* max idle probes */ int tcp_maxpersistidle; /* max idle time in persist */ int tcp_maxidle; /* * Tcp protocol timeout routine called every 500 ms. * Updates timestamps used for TCP * causes finite state machine actions if timers expire. */ void tcp_slowtimo() { tcp_maxidle = tcp_keepcnt * tcp_keepintvl; INP_INFO_WLOCK(&tcbinfo); (void) tcp_timer_2msl_tw(0); INP_INFO_WUNLOCK(&tcbinfo); } int tcp_syn_backoff[TCP_MAXRXTSHIFT + 1] = { 1, 1, 1, 1, 1, 2, 4, 8, 16, 32, 64, 64, 64 }; int tcp_backoff[TCP_MAXRXTSHIFT + 1] = { 1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 512, 512, 512 }; static int tcp_totbackoff = 2559; /* sum of tcp_backoff[] */ static int tcp_timer_race; SYSCTL_INT(_net_inet_tcp, OID_AUTO, timer_race, CTLFLAG_RD, &tcp_timer_race, 0, "Count of t_inpcb races on tcp_discardcb"); /* * TCP timer processing. */ void tcp_timer_delack(xtp) void *xtp; { struct tcpcb *tp = xtp; struct inpcb *inp; INP_INFO_RLOCK(&tcbinfo); inp = tp->t_inpcb; /* * XXXRW: While this assert is in fact correct, bugs in the tcpcb * tear-down mean we need it as a work-around for races between * timers and tcp_discardcb(). * * KASSERT(inp != NULL, ("tcp_timer_delack: inp == NULL")); */ if (inp == NULL) { tcp_timer_race++; INP_INFO_RUNLOCK(&tcbinfo); return; } INP_LOCK(inp); INP_INFO_RUNLOCK(&tcbinfo); if ((inp->inp_vflag & INP_DROPPED) || callout_pending(tp->tt_delack) || !callout_active(tp->tt_delack)) { INP_UNLOCK(inp); return; } callout_deactivate(tp->tt_delack); tp->t_flags |= TF_ACKNOW; tcpstat.tcps_delack++; (void) tcp_output(tp); INP_UNLOCK(inp); } void tcp_timer_2msl(xtp) void *xtp; { struct tcpcb *tp = xtp; struct inpcb *inp; #ifdef TCPDEBUG int ostate; ostate = tp->t_state; #endif /* * XXXRW: Does this actually happen? */ INP_INFO_WLOCK(&tcbinfo); inp = tp->t_inpcb; /* * XXXRW: While this assert is in fact correct, bugs in the tcpcb * tear-down mean we need it as a work-around for races between * timers and tcp_discardcb(). * * KASSERT(inp != NULL, ("tcp_timer_2msl: inp == NULL")); */ if (inp == NULL) { tcp_timer_race++; INP_INFO_WUNLOCK(&tcbinfo); return; } INP_LOCK(inp); tcp_free_sackholes(tp); if ((inp->inp_vflag & INP_DROPPED) || callout_pending(tp->tt_2msl) || !callout_active(tp->tt_2msl)) { INP_UNLOCK(tp->t_inpcb); INP_INFO_WUNLOCK(&tcbinfo); return; } callout_deactivate(tp->tt_2msl); /* * 2 MSL timeout in shutdown went off. If we're closed but * still waiting for peer to close and connection has been idle * too long, or if 2MSL time is up from TIME_WAIT, delete connection * control block. Otherwise, check again in a bit. + * + * If fastrecycle of FIN_WAIT_2, in FIN_WAIT_2 and receiver has closed, + * there's no point in hanging onto FIN_WAIT_2 socket. Just close it. + * Ignore fact that there were recent incoming segments. */ - if (tp->t_state != TCPS_TIME_WAIT && - (ticks - tp->t_rcvtime) <= tcp_maxidle) - callout_reset(tp->tt_2msl, tcp_keepintvl, - tcp_timer_2msl, tp); - else - tp = tcp_close(tp); + if (tcp_fast_finwait2_recycle && tp->t_state == TCPS_FIN_WAIT_2 && + tp->t_inpcb && tp->t_inpcb->inp_socket && + (tp->t_inpcb->inp_socket->so_rcv.sb_state & SBS_CANTRCVMORE)) { + tcpstat.tcps_finwait2_drops++; + tp = tcp_close(tp); + } else { + if (tp->t_state != TCPS_TIME_WAIT && + (ticks - tp->t_rcvtime) <= tcp_maxidle) + callout_reset(tp->tt_2msl, tcp_keepintvl, + tcp_timer_2msl, tp); + else + tp = tcp_close(tp); + } #ifdef TCPDEBUG if (tp != NULL && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0, PRU_SLOWTIMO); #endif if (tp != NULL) INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); } /* * The timed wait queue contains references to each of the TCP sessions * currently in the TIME_WAIT state. The queue pointers, including the * queue pointers in each tcptw structure, are protected using the global * tcbinfo lock, which must be held over queue iteration and modification. */ static TAILQ_HEAD(, tcptw) twq_2msl; void tcp_timer_init(void) { TAILQ_INIT(&twq_2msl); } void tcp_timer_2msl_reset(struct tcptw *tw, int rearm) { INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(tw->tw_inpcb); if (rearm) TAILQ_REMOVE(&twq_2msl, tw, tw_2msl); tw->tw_time = ticks + 2 * tcp_msl; TAILQ_INSERT_TAIL(&twq_2msl, tw, tw_2msl); } void tcp_timer_2msl_stop(struct tcptw *tw) { INP_INFO_WLOCK_ASSERT(&tcbinfo); TAILQ_REMOVE(&twq_2msl, tw, tw_2msl); } struct tcptw * tcp_timer_2msl_tw(int reuse) { struct tcptw *tw; INP_INFO_WLOCK_ASSERT(&tcbinfo); for (;;) { tw = TAILQ_FIRST(&twq_2msl); if (tw == NULL || (!reuse && tw->tw_time > ticks)) break; INP_LOCK(tw->tw_inpcb); tcp_twclose(tw, reuse); if (reuse) return (tw); } return (NULL); } void tcp_timer_keep(xtp) void *xtp; { struct tcpcb *tp = xtp; struct tcptemp *t_template; struct inpcb *inp; #ifdef TCPDEBUG int ostate; ostate = tp->t_state; #endif INP_INFO_WLOCK(&tcbinfo); inp = tp->t_inpcb; /* * XXXRW: While this assert is in fact correct, bugs in the tcpcb * tear-down mean we need it as a work-around for races between * timers and tcp_discardcb(). * * KASSERT(inp != NULL, ("tcp_timer_keep: inp == NULL")); */ if (inp == NULL) { tcp_timer_race++; INP_INFO_WUNLOCK(&tcbinfo); return; } INP_LOCK(inp); if ((inp->inp_vflag & INP_DROPPED) || callout_pending(tp->tt_keep) || !callout_active(tp->tt_keep)) { INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return; } callout_deactivate(tp->tt_keep); /* * Keep-alive timer went off; send something * or drop connection if idle for too long. */ tcpstat.tcps_keeptimeo++; if (tp->t_state < TCPS_ESTABLISHED) goto dropit; if ((always_keepalive || inp->inp_socket->so_options & SO_KEEPALIVE) && tp->t_state <= TCPS_CLOSING) { if ((ticks - tp->t_rcvtime) >= tcp_keepidle + tcp_maxidle) goto dropit; /* * Send a packet designed to force a response * if the peer is up and reachable: * either an ACK if the connection is still alive, * or an RST if the peer has closed the connection * due to timeout or reboot. * Using sequence number tp->snd_una-1 * causes the transmitted zero-length segment * to lie outside the receive window; * by the protocol spec, this requires the * correspondent TCP to respond. */ tcpstat.tcps_keepprobe++; t_template = tcpip_maketemplate(inp); if (t_template) { tcp_respond(tp, t_template->tt_ipgen, &t_template->tt_t, (struct mbuf *)NULL, tp->rcv_nxt, tp->snd_una - 1, 0); (void) m_free(dtom(t_template)); } callout_reset(tp->tt_keep, tcp_keepintvl, tcp_timer_keep, tp); } else callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp); #ifdef TCPDEBUG if (inp->inp_socket->so_options & SO_DEBUG) tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0, PRU_SLOWTIMO); #endif INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return; dropit: tcpstat.tcps_keepdrops++; tp = tcp_drop(tp, ETIMEDOUT); #ifdef TCPDEBUG if (tp != NULL && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0, PRU_SLOWTIMO); #endif if (tp != NULL) INP_UNLOCK(tp->t_inpcb); INP_INFO_WUNLOCK(&tcbinfo); } void tcp_timer_persist(xtp) void *xtp; { struct tcpcb *tp = xtp; struct inpcb *inp; #ifdef TCPDEBUG int ostate; ostate = tp->t_state; #endif INP_INFO_WLOCK(&tcbinfo); inp = tp->t_inpcb; /* * XXXRW: While this assert is in fact correct, bugs in the tcpcb * tear-down mean we need it as a work-around for races between * timers and tcp_discardcb(). * * KASSERT(inp != NULL, ("tcp_timer_persist: inp == NULL")); */ if (inp == NULL) { tcp_timer_race++; INP_INFO_WUNLOCK(&tcbinfo); return; } INP_LOCK(inp); if ((inp->inp_vflag & INP_DROPPED) || callout_pending(tp->tt_persist) || !callout_active(tp->tt_persist)) { INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return; } callout_deactivate(tp->tt_persist); /* * Persistance timer into zero window. * Force a byte to be output, if possible. */ tcpstat.tcps_persisttimeo++; /* * Hack: if the peer is dead/unreachable, we do not * time out if the window is closed. After a full * backoff, drop the connection if the idle time * (no responses to probes) reaches the maximum * backoff that we would use if retransmitting. */ if (tp->t_rxtshift == TCP_MAXRXTSHIFT && ((ticks - tp->t_rcvtime) >= tcp_maxpersistidle || (ticks - tp->t_rcvtime) >= TCP_REXMTVAL(tp) * tcp_totbackoff)) { tcpstat.tcps_persistdrop++; tp = tcp_drop(tp, ETIMEDOUT); goto out; } tcp_setpersist(tp); tp->t_flags |= TF_FORCEDATA; (void) tcp_output(tp); tp->t_flags &= ~TF_FORCEDATA; out: #ifdef TCPDEBUG if (tp != NULL && tp->t_inpcb->inp_socket->so_options & SO_DEBUG) tcp_trace(TA_USER, ostate, tp, NULL, NULL, PRU_SLOWTIMO); #endif if (tp != NULL) INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); } void tcp_timer_rexmt(xtp) void *xtp; { struct tcpcb *tp = xtp; int rexmt; int headlocked; struct inpcb *inp; #ifdef TCPDEBUG int ostate; ostate = tp->t_state; #endif INP_INFO_WLOCK(&tcbinfo); headlocked = 1; inp = tp->t_inpcb; /* * XXXRW: While this assert is in fact correct, bugs in the tcpcb * tear-down mean we need it as a work-around for races between * timers and tcp_discardcb(). * * KASSERT(inp != NULL, ("tcp_timer_rexmt: inp == NULL")); */ if (inp == NULL) { tcp_timer_race++; INP_INFO_WUNLOCK(&tcbinfo); return; } INP_LOCK(inp); if ((inp->inp_vflag & INP_DROPPED) || callout_pending(tp->tt_rexmt) || !callout_active(tp->tt_rexmt)) { INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return; } callout_deactivate(tp->tt_rexmt); tcp_free_sackholes(tp); /* * Retransmission timer went off. Message has not * been acked within retransmit interval. Back off * to a longer retransmit interval and retransmit one segment. */ if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { tp->t_rxtshift = TCP_MAXRXTSHIFT; tcpstat.tcps_timeoutdrop++; tp = tcp_drop(tp, tp->t_softerror ? tp->t_softerror : ETIMEDOUT); goto out; } INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; if (tp->t_rxtshift == 1) { /* * first retransmit; record ssthresh and cwnd so they can * be recovered if this turns out to be a "bad" retransmit. * A retransmit is considered "bad" if an ACK for this * segment is received within RTT/2 interval; the assumption * here is that the ACK was already in flight. See * "On Estimating End-to-End Network Path Properties" by * Allman and Paxson for more details. */ tp->snd_cwnd_prev = tp->snd_cwnd; tp->snd_ssthresh_prev = tp->snd_ssthresh; tp->snd_recover_prev = tp->snd_recover; if (IN_FASTRECOVERY(tp)) tp->t_flags |= TF_WASFRECOVERY; else tp->t_flags &= ~TF_WASFRECOVERY; tp->t_badrxtwin = ticks + (tp->t_srtt >> (TCP_RTT_SHIFT + 1)); } tcpstat.tcps_rexmttimeo++; if (tp->t_state == TCPS_SYN_SENT) rexmt = TCP_REXMTVAL(tp) * tcp_syn_backoff[tp->t_rxtshift]; else rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift]; TCPT_RANGESET(tp->t_rxtcur, rexmt, tp->t_rttmin, TCPTV_REXMTMAX); /* * Disable rfc1323 if we havn't got any response to * our third SYN to work-around some broken terminal servers * (most of which have hopefully been retired) that have bad VJ * header compression code which trashes TCP segments containing * unknown-to-them TCP options. */ if ((tp->t_state == TCPS_SYN_SENT) && (tp->t_rxtshift == 3)) tp->t_flags &= ~(TF_REQ_SCALE|TF_REQ_TSTMP); /* * If we backed off this far, our srtt estimate is probably bogus. * Clobber it so we'll take the next rtt measurement as our srtt; * move the current srtt into rttvar to keep the current * retransmit times until then. */ if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { #ifdef INET6 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) in6_losing(tp->t_inpcb); else #endif tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT); tp->t_srtt = 0; } tp->snd_nxt = tp->snd_una; tp->snd_recover = tp->snd_max; /* * Force a segment to be sent. */ tp->t_flags |= TF_ACKNOW; /* * If timing a segment in this window, stop the timer. */ tp->t_rtttime = 0; /* * Close the congestion window down to one segment * (we'll open it by one segment for each ack we get). * Since we probably have a window's worth of unacked * data accumulated, this "slow start" keeps us from * dumping all that data as back-to-back packets (which * might overwhelm an intermediate gateway). * * There are two phases to the opening: Initially we * open by one mss on each ack. This makes the window * size increase exponentially with time. If the * window is larger than the path can handle, this * exponential growth results in dropped packet(s) * almost immediately. To get more time between * drops but still "push" the network to take advantage * of improving conditions, we switch from exponential * to linear window opening at some threshhold size. * For a threshhold, we use half the current window * size, truncated to a multiple of the mss. * * (the minimum cwnd that will give us exponential * growth is 2 mss. We don't allow the threshhold * to go below this.) */ { u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; if (win < 2) win = 2; tp->snd_cwnd = tp->t_maxseg; tp->snd_ssthresh = win * tp->t_maxseg; tp->t_dupacks = 0; } EXIT_FASTRECOVERY(tp); (void) tcp_output(tp); out: #ifdef TCPDEBUG if (tp != NULL && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0, PRU_SLOWTIMO); #endif if (tp != NULL) INP_UNLOCK(inp); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); } Index: head/sys/netinet/tcp_timer.h =================================================================== --- head/sys/netinet/tcp_timer.h (revision 167035) +++ head/sys/netinet/tcp_timer.h (revision 167036) @@ -1,168 +1,173 @@ /*- * Copyright (c) 1982, 1986, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_timer.h 8.1 (Berkeley) 6/10/93 * $FreeBSD$ */ #ifndef _NETINET_TCP_TIMER_H_ #define _NETINET_TCP_TIMER_H_ /* * The TCPT_REXMT timer is used to force retransmissions. * The TCP has the TCPT_REXMT timer set whenever segments * have been sent for which ACKs are expected but not yet * received. If an ACK is received which advances tp->snd_una, * then the retransmit timer is cleared (if there are no more * outstanding segments) or reset to the base value (if there * are more ACKs expected). Whenever the retransmit timer goes off, * we retransmit one unacknowledged segment, and do a backoff * on the retransmit timer. * * The TCPT_PERSIST timer is used to keep window size information * flowing even if the window goes shut. If all previous transmissions * have been acknowledged (so that there are no retransmissions in progress), * and the window is too small to bother sending anything, then we start * the TCPT_PERSIST timer. When it expires, if the window is nonzero, * we go to transmit state. Otherwise, at intervals send a single byte * into the peer's window to force him to update our window information. * We do this at most as often as TCPT_PERSMIN time intervals, * but no more frequently than the current estimate of round-trip * packet time. The TCPT_PERSIST timer is cleared whenever we receive * a window update from the peer. * * The TCPT_KEEP timer is used to keep connections alive. If an * connection is idle (no segments received) for TCPTV_KEEP_INIT amount of time, * but not yet established, then we drop the connection. Once the connection * is established, if the connection is idle for TCPTV_KEEP_IDLE time * (and keepalives have been enabled on the socket), we begin to probe * the connection. We force the peer to send us a segment by sending: * * This segment is (deliberately) outside the window, and should elicit * an ack segment in response from the peer. If, despite the TCPT_KEEP * initiated segments we cannot elicit a response from a peer in TCPT_MAXIDLE * amount of time probing, then we drop the connection. */ /* * Time constants. */ #define TCPTV_MSL ( 30*hz) /* max seg lifetime (hah!) */ #define TCPTV_SRTTBASE 0 /* base roundtrip time; if 0, no idea yet */ #define TCPTV_RTOBASE ( 3*hz) /* assumed RTO if no info */ #define TCPTV_SRTTDFLT ( 3*hz) /* assumed RTT if no info */ #define TCPTV_PERSMIN ( 5*hz) /* retransmit persistence */ #define TCPTV_PERSMAX ( 60*hz) /* maximum persist interval */ #define TCPTV_KEEP_INIT ( 75*hz) /* initial connect keepalive */ #define TCPTV_KEEP_IDLE (120*60*hz) /* dflt time before probing */ #define TCPTV_KEEPINTVL ( 75*hz) /* default probe interval */ #define TCPTV_KEEPCNT 8 /* max probes before drop */ #define TCPTV_INFLIGHT_RTTTHRESH (10*hz/1000) /* below which inflight disengages, in msec */ +#define TCPTV_FINWAIT2_TIMEOUT (60*hz) /* FIN_WAIT_2 timeout if no receiver */ + /* * Minimum retransmit timer is 3 ticks, for algorithmic stability. * TCPT_RANGESET() will add another TCPTV_CPU_VAR to deal with * the expected worst-case processing variances by the kernels * representing the end points. Such variances do not always show * up in the srtt because the timestamp is often calculated at * the interface rather then at the TCP layer. This value is * typically 50ms. However, it is also possible that delayed * acks (typically 100ms) could create issues so we set the slop * to 200ms to try to cover it. Note that, properly speaking, * delayed-acks should not create a major issue for interactive * environments which 'P'ush the last segment, at least as * long as implementations do the required 'at least one ack * for every two packets' for the non-interactive streaming case. * (maybe the RTO calculation should use 2*RTT instead of RTT * to handle the ack-every-other-packet case). * * The prior minimum of 1*hz (1 second) badly breaks throughput on any * networks faster then a modem that has minor (e.g. 1%) packet loss. */ #define TCPTV_MIN ( 3 ) /* minimum allowable value */ #define TCPTV_CPU_VAR ( hz/5 ) /* cpu variance allowed (200ms) */ #define TCPTV_REXMTMAX ( 64*hz) /* max allowable REXMT value */ #define TCPTV_TWTRUNC 8 /* RTO factor to truncate TW */ #define TCP_LINGERTIME 120 /* linger at most 2 minutes */ #define TCP_MAXRXTSHIFT 12 /* maximum retransmits */ #define TCPTV_DELACK (hz / PR_FASTHZ / 2) /* 100ms timeout */ #ifdef TCPTIMERS static const char *tcptimers[] = { "REXMT", "PERSIST", "KEEP", "2MSL" }; #endif /* * Force a time value to be in a certain range. */ #define TCPT_RANGESET(tv, value, tvmin, tvmax) do { \ (tv) = (value) + tcp_rexmit_slop; \ if ((u_long)(tv) < (u_long)(tvmin)) \ (tv) = (tvmin); \ if ((u_long)(tv) > (u_long)(tvmax)) \ (tv) = (tvmax); \ } while(0) #ifdef _KERNEL extern int tcp_keepinit; /* time to establish connection */ extern int tcp_keepidle; /* time before keepalive probes begin */ extern int tcp_keepintvl; /* time between keepalive probes */ extern int tcp_maxidle; /* time to drop after starting probes */ extern int tcp_delacktime; /* time before sending a delayed ACK */ extern int tcp_maxpersistidle; extern int tcp_rexmit_min; extern int tcp_rexmit_slop; extern int tcp_msl; extern int tcp_ttl; /* time to live for TCP segs */ extern int tcp_backoff[]; struct tcptw; + +extern int tcp_finwait2_timeout; +extern int tcp_fast_finwait2_recycle; void tcp_timer_init(void); void tcp_timer_2msl(void *xtp); struct tcptw * tcp_timer_2msl_tw(int _reuse); /* XXX temporary */ void tcp_timer_2msl_reset(struct tcptw *_tw, int rearm); void tcp_timer_2msl_stop(struct tcptw *_tw); void tcp_timer_keep(void *xtp); void tcp_timer_persist(void *xtp); void tcp_timer_rexmt(void *xtp); void tcp_timer_delack(void *xtp); #endif /* _KERNEL */ #endif /* !_NETINET_TCP_TIMER_H_ */ Index: head/sys/netinet/tcp_timewait.c =================================================================== --- head/sys/netinet/tcp_timewait.c (revision 167035) +++ head/sys/netinet/tcp_timewait.c (revision 167036) @@ -1,2428 +1,2429 @@ /*- * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 * $FreeBSD$ */ #include "opt_compat.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_mac.h" #include "opt_tcpdebug.h" #include "opt_tcp_sack.h" #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef INET6 #include #endif #include #include #ifdef INET6 #include #include #include #endif #include #include #include #include #include #include #ifdef INET6 #include #endif #include #ifdef TCPDEBUG #include #endif #include #ifdef IPSEC #include #ifdef INET6 #include #endif #include #endif /*IPSEC*/ #ifdef FAST_IPSEC #include #include #ifdef INET6 #include #endif #include #define IPSEC #endif /*FAST_IPSEC*/ #include #include #include int tcp_mssdflt = TCP_MSS; SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW, &tcp_mssdflt , 0, "Default TCP Maximum Segment Size"); #ifdef INET6 int tcp_v6mssdflt = TCP6_MSS; SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt, CTLFLAG_RW, &tcp_v6mssdflt , 0, "Default TCP Maximum Segment Size for IPv6"); #endif /* * Minimum MSS we accept and use. This prevents DoS attacks where * we are forced to a ridiculous low MSS like 20 and send hundreds * of packets instead of one. The effect scales with the available * bandwidth and quickly saturates the CPU and network interface * with packet generation and sending. Set to zero to disable MINMSS * checking. This setting prevents us from sending too small packets. */ int tcp_minmss = TCP_MINMSS; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_RW, &tcp_minmss , 0, "Minmum TCP Maximum Segment Size"); /* * Number of TCP segments per second we accept from remote host * before we start to calculate average segment size. If average * segment size drops below the minimum TCP MSS we assume a DoS * attack and reset+drop the connection. Care has to be taken not to * set this value too small to not kill interactive type connections * (telnet, SSH) which send many small packets. */ int tcp_minmssoverload = TCP_MINMSSOVERLOAD; SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmssoverload, CTLFLAG_RW, &tcp_minmssoverload , 0, "Number of TCP Segments per Second allowed to" "be under the MINMSS Size"); #if 0 static int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ; SYSCTL_INT(_net_inet_tcp, TCPCTL_RTTDFLT, rttdflt, CTLFLAG_RW, &tcp_rttdflt , 0, "Default maximum TCP Round Trip Time"); #endif int tcp_do_rfc1323 = 1; SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW, &tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions"); static int tcp_tcbhashsize = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RDTUN, &tcp_tcbhashsize, 0, "Size of TCP control-block hashtable"); static int do_tcpdrain = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0, "Enable tcp_drain routine for extra help when low on mbufs"); SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD, &tcbinfo.ipi_count, 0, "Number of active PCBs"); static int icmp_may_rst = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW, &icmp_may_rst, 0, "Certain ICMP unreachable messages may abort connections in SYN_SENT"); static int tcp_isn_reseed_interval = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_RW, &tcp_isn_reseed_interval, 0, "Seconds between reseeding of ISN secret"); static uma_zone_t tcptw_zone; static int maxtcptw; static int tcptw_auto_size(void) { int halfrange; /* * Max out at half the ephemeral port range so that TIME_WAIT * sockets don't tie up too many ephemeral ports. */ if (ipport_lastauto > ipport_firstauto) halfrange = (ipport_lastauto - ipport_firstauto) / 2; else halfrange = (ipport_firstauto - ipport_lastauto) / 2; /* Protect against goofy port ranges smaller than 32. */ return (imin(imax(halfrange, 32), maxsockets / 5)); } static int sysctl_maxtcptw(SYSCTL_HANDLER_ARGS) { int error, new; if (maxtcptw == 0) new = tcptw_auto_size(); else new = maxtcptw; error = sysctl_handle_int(oidp, &new, sizeof(int), req); if (error == 0 && req->newptr) if (new >= 32) { maxtcptw = new; uma_zone_set_max(tcptw_zone, maxtcptw); } return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, maxtcptw, CTLTYPE_INT|CTLFLAG_RW, &maxtcptw, 0, sysctl_maxtcptw, "IU", "Maximum number of compressed TCP TIME_WAIT entries"); static int nolocaltimewait = 0; SYSCTL_INT(_net_inet_tcp, OID_AUTO, nolocaltimewait, CTLFLAG_RW, &nolocaltimewait, 0, "Do not create compressed TCP TIME_WAIT entries" "for local connections"); /* * TCP bandwidth limiting sysctls. Note that the default lower bound of * 1024 exists only for debugging. A good production default would be * something like 6100. */ SYSCTL_NODE(_net_inet_tcp, OID_AUTO, inflight, CTLFLAG_RW, 0, "TCP inflight data limiting"); static int tcp_inflight_enable = 1; SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, enable, CTLFLAG_RW, &tcp_inflight_enable, 0, "Enable automatic TCP inflight data limiting"); static int tcp_inflight_debug = 0; SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, debug, CTLFLAG_RW, &tcp_inflight_debug, 0, "Debug TCP inflight calculations"); static int tcp_inflight_rttthresh; SYSCTL_PROC(_net_inet_tcp_inflight, OID_AUTO, rttthresh, CTLTYPE_INT|CTLFLAG_RW, &tcp_inflight_rttthresh, 0, sysctl_msec_to_ticks, "I", "RTT threshold below which inflight will deactivate itself"); static int tcp_inflight_min = 6144; SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, min, CTLFLAG_RW, &tcp_inflight_min, 0, "Lower-bound for TCP inflight window"); static int tcp_inflight_max = TCP_MAXWIN << TCP_MAX_WINSHIFT; SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, max, CTLFLAG_RW, &tcp_inflight_max, 0, "Upper-bound for TCP inflight window"); static int tcp_inflight_stab = 20; SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, stab, CTLFLAG_RW, &tcp_inflight_stab, 0, "Inflight Algorithm Stabilization 20 = 2 packets"); uma_zone_t sack_hole_zone; static struct inpcb *tcp_notify(struct inpcb *, int); static void tcp_isn_tick(void *); /* * Target size of TCP PCB hash tables. Must be a power of two. * * Note that this can be overridden by the kernel environment * variable net.inet.tcp.tcbhashsize */ #ifndef TCBHASHSIZE #define TCBHASHSIZE 512 #endif /* * XXX * Callouts should be moved into struct tcp directly. They are currently * separate because the tcpcb structure is exported to userland for sysctl * parsing purposes, which do not know about callouts. */ struct tcpcb_mem { struct tcpcb tcb; struct callout tcpcb_mem_rexmt, tcpcb_mem_persist, tcpcb_mem_keep; struct callout tcpcb_mem_2msl, tcpcb_mem_delack; }; static uma_zone_t tcpcb_zone; struct callout isn_callout; static struct mtx isn_mtx; #define ISN_LOCK_INIT() mtx_init(&isn_mtx, "isn_mtx", NULL, MTX_DEF) #define ISN_LOCK() mtx_lock(&isn_mtx) #define ISN_UNLOCK() mtx_unlock(&isn_mtx) /* * TCP initialization. */ static void tcp_zone_change(void *tag) { uma_zone_set_max(tcbinfo.ipi_zone, maxsockets); uma_zone_set_max(tcpcb_zone, maxsockets); if (maxtcptw == 0) uma_zone_set_max(tcptw_zone, tcptw_auto_size()); } static int tcp_inpcb_init(void *mem, int size, int flags) { struct inpcb *inp = mem; INP_LOCK_INIT(inp, "inp", "tcpinp"); return (0); } void tcp_init(void) { int hashsize = TCBHASHSIZE; tcp_delacktime = TCPTV_DELACK; tcp_keepinit = TCPTV_KEEP_INIT; tcp_keepidle = TCPTV_KEEP_IDLE; tcp_keepintvl = TCPTV_KEEPINTVL; tcp_maxpersistidle = TCPTV_KEEP_IDLE; tcp_msl = TCPTV_MSL; tcp_rexmit_min = TCPTV_MIN; tcp_rexmit_slop = TCPTV_CPU_VAR; tcp_inflight_rttthresh = TCPTV_INFLIGHT_RTTTHRESH; + tcp_finwait2_timeout = TCPTV_FINWAIT2_TIMEOUT; INP_INFO_LOCK_INIT(&tcbinfo, "tcp"); LIST_INIT(&tcb); tcbinfo.listhead = &tcb; TUNABLE_INT_FETCH("net.inet.tcp.tcbhashsize", &hashsize); if (!powerof2(hashsize)) { printf("WARNING: TCB hash size not a power of 2\n"); hashsize = 512; /* safe default */ } tcp_tcbhashsize = hashsize; tcbinfo.hashbase = hashinit(hashsize, M_PCB, &tcbinfo.hashmask); tcbinfo.porthashbase = hashinit(hashsize, M_PCB, &tcbinfo.porthashmask); tcbinfo.ipi_zone = uma_zcreate("inpcb", sizeof(struct inpcb), NULL, NULL, tcp_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcbinfo.ipi_zone, maxsockets); #ifdef INET6 #define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) #else /* INET6 */ #define TCP_MINPROTOHDR (sizeof(struct tcpiphdr)) #endif /* INET6 */ if (max_protohdr < TCP_MINPROTOHDR) max_protohdr = TCP_MINPROTOHDR; if (max_linkhdr + TCP_MINPROTOHDR > MHLEN) panic("tcp_init"); #undef TCP_MINPROTOHDR /* * These have to be type stable for the benefit of the timers. */ tcpcb_zone = uma_zcreate("tcpcb", sizeof(struct tcpcb_mem), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); uma_zone_set_max(tcpcb_zone, maxsockets); tcptw_zone = uma_zcreate("tcptw", sizeof(struct tcptw), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); TUNABLE_INT_FETCH("net.inet.tcp.maxtcptw", &maxtcptw); if (maxtcptw == 0) uma_zone_set_max(tcptw_zone, tcptw_auto_size()); else uma_zone_set_max(tcptw_zone, maxtcptw); tcp_timer_init(); syncache_init(); tcp_hc_init(); tcp_reass_init(); ISN_LOCK_INIT(); callout_init(&isn_callout, CALLOUT_MPSAFE); tcp_isn_tick(NULL); EVENTHANDLER_REGISTER(shutdown_pre_sync, tcp_fini, NULL, SHUTDOWN_PRI_DEFAULT); sack_hole_zone = uma_zcreate("sackhole", sizeof(struct sackhole), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); EVENTHANDLER_REGISTER(maxsockets_change, tcp_zone_change, NULL, EVENTHANDLER_PRI_ANY); } void tcp_fini(void *xtp) { callout_stop(&isn_callout); } /* * Fill in the IP and TCP headers for an outgoing packet, given the tcpcb. * tcp_template used to store this data in mbufs, but we now recopy it out * of the tcpcb each time to conserve mbufs. */ void tcpip_fillheaders(struct inpcb *inp, void *ip_ptr, void *tcp_ptr) { struct tcphdr *th = (struct tcphdr *)tcp_ptr; INP_LOCK_ASSERT(inp); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { struct ip6_hdr *ip6; ip6 = (struct ip6_hdr *)ip_ptr; ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) | (inp->in6p_flowinfo & IPV6_FLOWINFO_MASK); ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) | (IPV6_VERSION & IPV6_VERSION_MASK); ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = sizeof(struct tcphdr); ip6->ip6_src = inp->in6p_laddr; ip6->ip6_dst = inp->in6p_faddr; } else #endif { struct ip *ip; ip = (struct ip *)ip_ptr; ip->ip_v = IPVERSION; ip->ip_hl = 5; ip->ip_tos = inp->inp_ip_tos; ip->ip_len = 0; ip->ip_id = 0; ip->ip_off = 0; ip->ip_ttl = inp->inp_ip_ttl; ip->ip_sum = 0; ip->ip_p = IPPROTO_TCP; ip->ip_src = inp->inp_laddr; ip->ip_dst = inp->inp_faddr; } th->th_sport = inp->inp_lport; th->th_dport = inp->inp_fport; th->th_seq = 0; th->th_ack = 0; th->th_x2 = 0; th->th_off = 5; th->th_flags = 0; th->th_win = 0; th->th_urp = 0; th->th_sum = 0; /* in_pseudo() is called later for ipv4 */ } /* * Create template to be used to send tcp packets on a connection. * Allocates an mbuf and fills in a skeletal tcp/ip header. The only * use for this function is in keepalives, which use tcp_respond. */ struct tcptemp * tcpip_maketemplate(struct inpcb *inp) { struct mbuf *m; struct tcptemp *n; m = m_get(M_DONTWAIT, MT_DATA); if (m == NULL) return (0); m->m_len = sizeof(struct tcptemp); n = mtod(m, struct tcptemp *); tcpip_fillheaders(inp, (void *)&n->tt_ipgen, (void *)&n->tt_t); return (n); } /* * Send a single message to the TCP at address specified by * the given TCP/IP header. If m == NULL, then we make a copy * of the tcpiphdr at ti and send directly to the addressed host. * This is used to force keep alive messages out using the TCP * template for a connection. If flags are given then we send * a message back to the TCP which originated the * segment ti, * and discard the mbuf containing it and any other attached mbufs. * * In any case the ack and sequence number of the transmitted * segment are as specified by the parameters. * * NOTE: If m != NULL, then ti must point to *inside* the mbuf. */ void tcp_respond(struct tcpcb *tp, void *ipgen, register struct tcphdr *th, register struct mbuf *m, tcp_seq ack, tcp_seq seq, int flags) { register int tlen; int win = 0; struct ip *ip; struct tcphdr *nth; #ifdef INET6 struct ip6_hdr *ip6; int isipv6; #endif /* INET6 */ int ipflags = 0; struct inpcb *inp; KASSERT(tp != NULL || m != NULL, ("tcp_respond: tp and m both NULL")); #ifdef INET6 isipv6 = ((struct ip *)ipgen)->ip_v == 6; ip6 = ipgen; #endif /* INET6 */ ip = ipgen; if (tp != NULL) { inp = tp->t_inpcb; KASSERT(inp != NULL, ("tcp control block w/o inpcb")); INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); } else inp = NULL; if (tp != NULL) { if (!(flags & TH_RST)) { win = sbspace(&inp->inp_socket->so_rcv); if (win > (long)TCP_MAXWIN << tp->rcv_scale) win = (long)TCP_MAXWIN << tp->rcv_scale; } } if (m == NULL) { m = m_gethdr(M_DONTWAIT, MT_DATA); if (m == NULL) return; tlen = 0; m->m_data += max_linkhdr; #ifdef INET6 if (isipv6) { bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(struct ip6_hdr)); ip6 = mtod(m, struct ip6_hdr *); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); ip = mtod(m, struct ip *); nth = (struct tcphdr *)(ip + 1); } bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr)); flags = TH_ACK; } else { m_freem(m->m_next); m->m_next = NULL; m->m_data = (caddr_t)ipgen; /* m_len is set later */ tlen = 0; #define xchg(a,b,type) { type t; t=a; a=b; b=t; } #ifdef INET6 if (isipv6) { xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr); nth = (struct tcphdr *)(ip6 + 1); } else #endif /* INET6 */ { xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long); nth = (struct tcphdr *)(ip + 1); } if (th != nth) { /* * this is usually a case when an extension header * exists between the IPv6 header and the * TCP header. */ nth->th_sport = th->th_sport; nth->th_dport = th->th_dport; } xchg(nth->th_dport, nth->th_sport, n_short); #undef xchg } #ifdef INET6 if (isipv6) { ip6->ip6_flow = 0; ip6->ip6_vfc = IPV6_VERSION; ip6->ip6_nxt = IPPROTO_TCP; ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) + tlen)); tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr); } else #endif { tlen += sizeof (struct tcpiphdr); ip->ip_len = tlen; ip->ip_ttl = ip_defttl; if (path_mtu_discovery) ip->ip_off |= IP_DF; } m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = NULL; #ifdef MAC if (inp != NULL) { /* * Packet is associated with a socket, so allow the * label of the response to reflect the socket label. */ INP_LOCK_ASSERT(inp); mac_create_mbuf_from_inpcb(inp, m); } else { /* * Packet is not associated with a socket, so possibly * update the label in place. */ mac_reflect_mbuf_tcp(m); } #endif nth->th_seq = htonl(seq); nth->th_ack = htonl(ack); nth->th_x2 = 0; nth->th_off = sizeof (struct tcphdr) >> 2; nth->th_flags = flags; if (tp != NULL) nth->th_win = htons((u_short) (win >> tp->rcv_scale)); else nth->th_win = htons((u_short)win); nth->th_urp = 0; #ifdef INET6 if (isipv6) { nth->th_sum = 0; nth->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), tlen - sizeof(struct ip6_hdr)); ip6->ip6_hlim = in6_selecthlim(tp != NULL ? tp->t_inpcb : NULL, NULL); } else #endif /* INET6 */ { nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p))); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); } #ifdef TCPDEBUG if (tp == NULL || (inp->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0); #endif #ifdef INET6 if (isipv6) (void) ip6_output(m, NULL, NULL, ipflags, NULL, NULL, inp); else #endif /* INET6 */ (void) ip_output(m, NULL, NULL, ipflags, NULL, inp); } /* * Create a new TCP control block, making an * empty reassembly queue and hooking it to the argument * protocol control block. The `inp' parameter must have * come from the zone allocator set up in tcp_init(). */ struct tcpcb * tcp_newtcpcb(struct inpcb *inp) { struct tcpcb_mem *tm; struct tcpcb *tp; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ tm = uma_zalloc(tcpcb_zone, M_NOWAIT | M_ZERO); if (tm == NULL) return (NULL); tp = &tm->tcb; /* LIST_INIT(&tp->t_segq); */ /* XXX covered by M_ZERO */ tp->t_maxseg = tp->t_maxopd = #ifdef INET6 isipv6 ? tcp_v6mssdflt : #endif /* INET6 */ tcp_mssdflt; /* Set up our timeouts. */ callout_init(tp->tt_rexmt = &tm->tcpcb_mem_rexmt, NET_CALLOUT_MPSAFE); callout_init(tp->tt_persist = &tm->tcpcb_mem_persist, NET_CALLOUT_MPSAFE); callout_init(tp->tt_keep = &tm->tcpcb_mem_keep, NET_CALLOUT_MPSAFE); callout_init(tp->tt_2msl = &tm->tcpcb_mem_2msl, NET_CALLOUT_MPSAFE); callout_init(tp->tt_delack = &tm->tcpcb_mem_delack, NET_CALLOUT_MPSAFE); if (tcp_do_rfc1323) tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP); tp->sack_enable = tcp_do_sack; TAILQ_INIT(&tp->snd_holes); tp->t_inpcb = inp; /* XXX */ /* * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives * reasonable initial retransmit time. */ tp->t_srtt = TCPTV_SRTTBASE; tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4; tp->t_rttmin = tcp_rexmit_min; tp->t_rxtcur = TCPTV_RTOBASE; tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->t_rcvtime = ticks; tp->t_bw_rtttime = ticks; /* * IPv4 TTL initialization is necessary for an IPv6 socket as well, * because the socket may be bound to an IPv6 wildcard address, * which may match an IPv4-mapped IPv6 address. */ inp->inp_ip_ttl = ip_defttl; inp->inp_ppcb = tp; return (tp); /* XXX */ } /* * Drop a TCP connection, reporting * the specified error. If connection is synchronized, * then send a RST to peer. */ struct tcpcb * tcp_drop(struct tcpcb *tp, int errno) { struct socket *so = tp->t_inpcb->inp_socket; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(tp->t_inpcb); if (TCPS_HAVERCVDSYN(tp->t_state)) { tp->t_state = TCPS_CLOSED; (void) tcp_output(tp); tcpstat.tcps_drops++; } else tcpstat.tcps_conndrops++; if (errno == ETIMEDOUT && tp->t_softerror) errno = tp->t_softerror; so->so_error = errno; return (tcp_close(tp)); } void tcp_discardcb(struct tcpcb *tp) { struct tseg_qent *q; struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; #ifdef INET6 int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif /* INET6 */ INP_LOCK_ASSERT(inp); /* * Make sure that all of our timers are stopped before we * delete the PCB. */ callout_stop(tp->tt_rexmt); callout_stop(tp->tt_persist); callout_stop(tp->tt_keep); callout_stop(tp->tt_2msl); callout_stop(tp->tt_delack); /* * If we got enough samples through the srtt filter, * save the rtt and rttvar in the routing entry. * 'Enough' is arbitrarily defined as 4 rtt samples. * 4 samples is enough for the srtt filter to converge * to within enough % of the correct value; fewer samples * and we could save a bogus rtt. The danger is not high * as tcp quickly recovers from everything. * XXX: Works very well but needs some more statistics! */ if (tp->t_rttupdated >= 4) { struct hc_metrics_lite metrics; u_long ssthresh; bzero(&metrics, sizeof(metrics)); /* * Update the ssthresh always when the conditions below * are satisfied. This gives us better new start value * for the congestion avoidance for new connections. * ssthresh is only set if packet loss occured on a session. * * XXXRW: 'so' may be NULL here, and/or socket buffer may be * being torn down. Ideally this code would not use 'so'. */ ssthresh = tp->snd_ssthresh; if (ssthresh != 0 && ssthresh < so->so_snd.sb_hiwat / 2) { /* * convert the limit from user data bytes to * packets then to packet data bytes. */ ssthresh = (ssthresh + tp->t_maxseg / 2) / tp->t_maxseg; if (ssthresh < 2) ssthresh = 2; ssthresh *= (u_long)(tp->t_maxseg + #ifdef INET6 (isipv6 ? sizeof (struct ip6_hdr) + sizeof (struct tcphdr) : #endif sizeof (struct tcpiphdr) #ifdef INET6 ) #endif ); } else ssthresh = 0; metrics.rmx_ssthresh = ssthresh; metrics.rmx_rtt = tp->t_srtt; metrics.rmx_rttvar = tp->t_rttvar; /* XXX: This wraps if the pipe is more than 4 Gbit per second */ metrics.rmx_bandwidth = tp->snd_bandwidth; metrics.rmx_cwnd = tp->snd_cwnd; metrics.rmx_sendpipe = 0; metrics.rmx_recvpipe = 0; tcp_hc_update(&inp->inp_inc, &metrics); } /* free the reassembly queue, if any */ while ((q = LIST_FIRST(&tp->t_segq)) != NULL) { LIST_REMOVE(q, tqe_q); m_freem(q->tqe_m); uma_zfree(tcp_reass_zone, q); tp->t_segqlen--; tcp_reass_qsize--; } tcp_free_sackholes(tp); inp->inp_ppcb = NULL; tp->t_inpcb = NULL; uma_zfree(tcpcb_zone, tp); } /* * Attempt to close a TCP control block, marking it as dropped, and freeing * the socket if we hold the only reference. */ struct tcpcb * tcp_close(struct tcpcb *tp) { struct inpcb *inp = tp->t_inpcb; struct socket *so; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); in_pcbdrop(inp); tcpstat.tcps_closed++; KASSERT(inp->inp_socket != NULL, ("tcp_close: inp_socket NULL")); so = inp->inp_socket; soisdisconnected(so); if (inp->inp_vflag & INP_SOCKREF) { KASSERT(so->so_state & SS_PROTOREF, ("tcp_close: !SS_PROTOREF")); inp->inp_vflag &= ~INP_SOCKREF; INP_UNLOCK(inp); ACCEPT_LOCK(); SOCK_LOCK(so); so->so_state &= ~SS_PROTOREF; sofree(so); return (NULL); } return (tp); } void tcp_drain(void) { if (do_tcpdrain) { struct inpcb *inpb; struct tcpcb *tcpb; struct tseg_qent *te; /* * Walk the tcpbs, if existing, and flush the reassembly queue, * if there is one... * XXX: The "Net/3" implementation doesn't imply that the TCP * reassembly queue should be flushed, but in a situation * where we're really low on mbufs, this is potentially * usefull. */ INP_INFO_RLOCK(&tcbinfo); LIST_FOREACH(inpb, tcbinfo.listhead, inp_list) { if (inpb->inp_vflag & INP_TIMEWAIT) continue; INP_LOCK(inpb); if ((tcpb = intotcpcb(inpb)) != NULL) { while ((te = LIST_FIRST(&tcpb->t_segq)) != NULL) { LIST_REMOVE(te, tqe_q); m_freem(te->tqe_m); uma_zfree(tcp_reass_zone, te); tcpb->t_segqlen--; tcp_reass_qsize--; } tcp_clean_sackreport(tcpb); } INP_UNLOCK(inpb); } INP_INFO_RUNLOCK(&tcbinfo); } } /* * Notify a tcp user of an asynchronous error; * store error as soft error, but wake up user * (for now, won't do anything until can select for soft error). * * Do not wake up user since there currently is no mechanism for * reporting soft errors (yet - a kqueue filter may be added). */ static struct inpcb * tcp_notify(struct inpcb *inp, int error) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); if ((inp->inp_vflag & INP_TIMEWAIT) || (inp->inp_vflag & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_notify: tp == NULL")); /* * Ignore some errors if we are hooked up. * If connection hasn't completed, has retransmitted several times, * and receives a second error, give up now. This is better * than waiting a long time to establish a connection that * can never complete. */ if (tp->t_state == TCPS_ESTABLISHED && (error == EHOSTUNREACH || error == ENETUNREACH || error == EHOSTDOWN)) { return (inp); } else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 && tp->t_softerror) { tp = tcp_drop(tp, error); if (tp != NULL) return (inp); else return (NULL); } else { tp->t_softerror = error; return (inp); } #if 0 wakeup( &so->so_timeo); sorwakeup(so); sowwakeup(so); #endif } static int tcp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; struct inpcb *inp, **inp_list; inp_gen_t gencnt; struct xinpgen xig; /* * The process of preparing the TCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == NULL) { n = tcbinfo.ipi_count; req->oldidx = 2 * (sizeof xig) + (n + n/8) * sizeof(struct xtcpcb); return (0); } if (req->newptr != NULL) return (EPERM); /* * OK, now we're committed to doing something. */ INP_INFO_RLOCK(&tcbinfo); gencnt = tcbinfo.ipi_gencnt; n = tcbinfo.ipi_count; INP_INFO_RUNLOCK(&tcbinfo); error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) + n * sizeof(struct xtcpcb)); if (error != 0) return (error); xig.xig_len = sizeof xig; xig.xig_count = n; xig.xig_gen = gencnt; xig.xig_sogen = so_gencnt; error = SYSCTL_OUT(req, &xig, sizeof xig); if (error) return (error); inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); if (inp_list == NULL) return (ENOMEM); INP_INFO_RLOCK(&tcbinfo); for (inp = LIST_FIRST(tcbinfo.listhead), i = 0; inp != NULL && i < n; inp = LIST_NEXT(inp, inp_list)) { INP_LOCK(inp); if (inp->inp_gencnt <= gencnt) { /* * XXX: This use of cr_cansee(), introduced with * TCP state changes, is not quite right, but for * now, better than nothing. */ if (inp->inp_vflag & INP_TIMEWAIT) { if (intotw(inp) != NULL) error = cr_cansee(req->td->td_ucred, intotw(inp)->tw_cred); else error = EINVAL; /* Skip this inp. */ } else error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error == 0) inp_list[i++] = inp; } INP_UNLOCK(inp); } INP_INFO_RUNLOCK(&tcbinfo); n = i; error = 0; for (i = 0; i < n; i++) { inp = inp_list[i]; INP_LOCK(inp); if (inp->inp_gencnt <= gencnt) { struct xtcpcb xt; void *inp_ppcb; bzero(&xt, sizeof(xt)); xt.xt_len = sizeof xt; /* XXX should avoid extra copy */ bcopy(inp, &xt.xt_inp, sizeof *inp); inp_ppcb = inp->inp_ppcb; if (inp_ppcb == NULL) bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); else if (inp->inp_vflag & INP_TIMEWAIT) { bzero((char *) &xt.xt_tp, sizeof xt.xt_tp); xt.xt_tp.t_state = TCPS_TIME_WAIT; } else bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp); if (inp->inp_socket != NULL) sotoxsocket(inp->inp_socket, &xt.xt_socket); else { bzero(&xt.xt_socket, sizeof xt.xt_socket); xt.xt_socket.xso_protocol = IPPROTO_TCP; } xt.xt_inp.inp_gencnt = inp->inp_gencnt; INP_UNLOCK(inp); error = SYSCTL_OUT(req, &xt, sizeof xt); } else INP_UNLOCK(inp); } if (!error) { /* * Give the user an updated idea of our state. * If the generation differs from what we told * her before, she knows that something happened * while we were processing this request, and it * might be necessary to retry. */ INP_INFO_RLOCK(&tcbinfo); xig.xig_gen = tcbinfo.ipi_gencnt; xig.xig_sogen = so_gencnt; xig.xig_count = tcbinfo.ipi_count; INP_INFO_RUNLOCK(&tcbinfo); error = SYSCTL_OUT(req, &xig, sizeof xig); } free(inp_list, M_TEMP); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, tcp_pcblist, "S,xtcpcb", "List of active TCP connections"); static int tcp_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in addrs[2]; struct inpcb *inp; int error; error = priv_check_cred(req->td->td_ucred, PRIV_NETINET_GETCRED, SUSER_ALLOWJAIL); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); INP_INFO_RLOCK(&tcbinfo); inp = in_pcblookup_hash(&tcbinfo, addrs[1].sin_addr, addrs[1].sin_port, addrs[0].sin_addr, addrs[0].sin_port, 0, NULL); if (inp == NULL) { error = ENOENT; goto outunlocked; } INP_LOCK(inp); if (inp->inp_socket == NULL) { error = ENOENT; goto out; } error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error) goto out; cru2x(inp->inp_socket->so_cred, &xuc); out: INP_UNLOCK(inp); outunlocked: INP_INFO_RUNLOCK(&tcbinfo); if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp_getcred, "S,xucred", "Get the xucred of a TCP connection"); #ifdef INET6 static int tcp6_getcred(SYSCTL_HANDLER_ARGS) { struct xucred xuc; struct sockaddr_in6 addrs[2]; struct inpcb *inp; int error, mapped = 0; error = priv_check_cred(req->td->td_ucred, PRIV_NETINET_GETCRED, SUSER_ALLOWJAIL); if (error) return (error); error = SYSCTL_IN(req, addrs, sizeof(addrs)); if (error) return (error); if ((error = sa6_embedscope(&addrs[0], ip6_use_defzone)) != 0 || (error = sa6_embedscope(&addrs[1], ip6_use_defzone)) != 0) { return (error); } if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) { if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr)) mapped = 1; else return (EINVAL); } INP_INFO_RLOCK(&tcbinfo); if (mapped == 1) inp = in_pcblookup_hash(&tcbinfo, *(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12], addrs[1].sin6_port, *(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12], addrs[0].sin6_port, 0, NULL); else inp = in6_pcblookup_hash(&tcbinfo, &addrs[1].sin6_addr, addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 0, NULL); if (inp == NULL) { error = ENOENT; goto outunlocked; } INP_LOCK(inp); if (inp->inp_socket == NULL) { error = ENOENT; goto out; } error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); if (error) goto out; cru2x(inp->inp_socket->so_cred, &xuc); out: INP_UNLOCK(inp); outunlocked: INP_INFO_RUNLOCK(&tcbinfo); if (error == 0) error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); return (error); } SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection"); #endif void tcp_ctlinput(int cmd, struct sockaddr *sa, void *vip) { struct ip *ip = vip; struct tcphdr *th; struct in_addr faddr; struct inpcb *inp; struct tcpcb *tp; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct icmp *icp; struct in_conninfo inc; tcp_seq icmp_tcp_seq; int mtu; faddr = ((struct sockaddr_in *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB || cmd == PRC_UNREACH_PORT || cmd == PRC_TIMXCEED_INTRANS) && ip) notify = tcp_drop_syn_sent; /* * Redirects don't need to be handled up here. */ else if (PRC_IS_REDIRECT(cmd)) return; /* * Source quench is depreciated. */ else if (cmd == PRC_QUENCH) return; /* * Hostdead is ugly because it goes linearly through all PCBs. * XXX: We never get this from ICMP, otherwise it makes an * excellent DoS attack on machines with many connections. */ else if (cmd == PRC_HOSTDEAD) ip = NULL; else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip != NULL) { icp = (struct icmp *)((caddr_t)ip - offsetof(struct icmp, icmp_ip)); th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); INP_INFO_WLOCK(&tcbinfo); inp = in_pcblookup_hash(&tcbinfo, faddr, th->th_dport, ip->ip_src, th->th_sport, 0, NULL); if (inp != NULL) { INP_LOCK(inp); if (!(inp->inp_vflag & INP_TIMEWAIT) && !(inp->inp_vflag & INP_DROPPED) && !(inp->inp_socket == NULL)) { icmp_tcp_seq = htonl(th->th_seq); tp = intotcpcb(inp); if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) && SEQ_LT(icmp_tcp_seq, tp->snd_max)) { if (cmd == PRC_MSGSIZE) { /* * MTU discovery: * If we got a needfrag set the MTU * in the route to the suggested new * value (if given) and then notify. */ bzero(&inc, sizeof(inc)); inc.inc_flags = 0; /* IPv4 */ inc.inc_faddr = faddr; mtu = ntohs(icp->icmp_nextmtu); /* * If no alternative MTU was * proposed, try the next smaller * one. ip->ip_len has already * been swapped in icmp_input(). */ if (!mtu) mtu = ip_next_mtu(ip->ip_len, 1); if (mtu < max(296, (tcp_minmss) + sizeof(struct tcpiphdr))) mtu = 0; if (!mtu) mtu = tcp_mssdflt + sizeof(struct tcpiphdr); /* * Only cache the the MTU if it * is smaller than the interface * or route MTU. tcp_mtudisc() * will do right thing by itself. */ if (mtu <= tcp_maxmtu(&inc, NULL)) tcp_hc_updatemtu(&inc, mtu); } inp = (*notify)(inp, inetctlerrmap[cmd]); } } if (inp != NULL) INP_UNLOCK(inp); } else { inc.inc_fport = th->th_dport; inc.inc_lport = th->th_sport; inc.inc_faddr = faddr; inc.inc_laddr = ip->ip_src; #ifdef INET6 inc.inc_isipv6 = 0; #endif syncache_unreach(&inc, th); } INP_INFO_WUNLOCK(&tcbinfo); } else in_pcbnotifyall(&tcbinfo, faddr, inetctlerrmap[cmd], notify); } #ifdef INET6 void tcp6_ctlinput(int cmd, struct sockaddr *sa, void *d) { struct tcphdr th; struct inpcb *(*notify)(struct inpcb *, int) = tcp_notify; struct ip6_hdr *ip6; struct mbuf *m; struct ip6ctlparam *ip6cp = NULL; const struct sockaddr_in6 *sa6_src = NULL; int off; struct tcp_portonly { u_int16_t th_sport; u_int16_t th_dport; } *thp; if (sa->sa_family != AF_INET6 || sa->sa_len != sizeof(struct sockaddr_in6)) return; if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; else if (!PRC_IS_REDIRECT(cmd) && ((unsigned)cmd >= PRC_NCMDS || inet6ctlerrmap[cmd] == 0)) return; /* Source quench is depreciated. */ else if (cmd == PRC_QUENCH) return; /* if the parameter is from icmp6, decode it. */ if (d != NULL) { ip6cp = (struct ip6ctlparam *)d; m = ip6cp->ip6c_m; ip6 = ip6cp->ip6c_ip6; off = ip6cp->ip6c_off; sa6_src = ip6cp->ip6c_src; } else { m = NULL; ip6 = NULL; off = 0; /* fool gcc */ sa6_src = &sa6_any; } if (ip6 != NULL) { struct in_conninfo inc; /* * XXX: We assume that when IPV6 is non NULL, * M and OFF are valid. */ /* check if we can safely examine src and dst ports */ if (m->m_pkthdr.len < off + sizeof(*thp)) return; bzero(&th, sizeof(th)); m_copydata(m, off, sizeof(*thp), (caddr_t)&th); in6_pcbnotify(&tcbinfo, sa, th.th_dport, (struct sockaddr *)ip6cp->ip6c_src, th.th_sport, cmd, NULL, notify); inc.inc_fport = th.th_dport; inc.inc_lport = th.th_sport; inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr; inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr; inc.inc_isipv6 = 1; INP_INFO_WLOCK(&tcbinfo); syncache_unreach(&inc, &th); INP_INFO_WUNLOCK(&tcbinfo); } else in6_pcbnotify(&tcbinfo, sa, 0, (const struct sockaddr *)sa6_src, 0, cmd, NULL, notify); } #endif /* INET6 */ /* * Following is where TCP initial sequence number generation occurs. * * There are two places where we must use initial sequence numbers: * 1. In SYN-ACK packets. * 2. In SYN packets. * * All ISNs for SYN-ACK packets are generated by the syncache. See * tcp_syncache.c for details. * * The ISNs in SYN packets must be monotonic; TIME_WAIT recycling * depends on this property. In addition, these ISNs should be * unguessable so as to prevent connection hijacking. To satisfy * the requirements of this situation, the algorithm outlined in * RFC 1948 is used, with only small modifications. * * Implementation details: * * Time is based off the system timer, and is corrected so that it * increases by one megabyte per second. This allows for proper * recycling on high speed LANs while still leaving over an hour * before rollover. * * As reading the *exact* system time is too expensive to be done * whenever setting up a TCP connection, we increment the time * offset in two ways. First, a small random positive increment * is added to isn_offset for each connection that is set up. * Second, the function tcp_isn_tick fires once per clock tick * and increments isn_offset as necessary so that sequence numbers * are incremented at approximately ISN_BYTES_PER_SECOND. The * random positive increments serve only to ensure that the same * exact sequence number is never sent out twice (as could otherwise * happen when a port is recycled in less than the system tick * interval.) * * net.inet.tcp.isn_reseed_interval controls the number of seconds * between seeding of isn_secret. This is normally set to zero, * as reseeding should not be necessary. * * Locking of the global variables isn_secret, isn_last_reseed, isn_offset, * isn_offset_old, and isn_ctx is performed using the TCP pcbinfo lock. In * general, this means holding an exclusive (write) lock. */ #define ISN_BYTES_PER_SECOND 1048576 #define ISN_STATIC_INCREMENT 4096 #define ISN_RANDOM_INCREMENT (4096 - 1) static u_char isn_secret[32]; static int isn_last_reseed; static u_int32_t isn_offset, isn_offset_old; static MD5_CTX isn_ctx; tcp_seq tcp_new_isn(struct tcpcb *tp) { u_int32_t md5_buffer[4]; tcp_seq new_isn; INP_LOCK_ASSERT(tp->t_inpcb); ISN_LOCK(); /* Seed if this is the first use, reseed if requested. */ if ((isn_last_reseed == 0) || ((tcp_isn_reseed_interval > 0) && (((u_int)isn_last_reseed + (u_int)tcp_isn_reseed_interval*hz) < (u_int)ticks))) { read_random(&isn_secret, sizeof(isn_secret)); isn_last_reseed = ticks; } /* Compute the md5 hash and return the ISN. */ MD5Init(&isn_ctx); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short)); #ifdef INET6 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr, sizeof(struct in6_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr, sizeof(struct in6_addr)); } else #endif { MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr, sizeof(struct in_addr)); MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr, sizeof(struct in_addr)); } MD5Update(&isn_ctx, (u_char *) &isn_secret, sizeof(isn_secret)); MD5Final((u_char *) &md5_buffer, &isn_ctx); new_isn = (tcp_seq) md5_buffer[0]; isn_offset += ISN_STATIC_INCREMENT + (arc4random() & ISN_RANDOM_INCREMENT); new_isn += isn_offset; ISN_UNLOCK(); return (new_isn); } /* * Increment the offset to the next ISN_BYTES_PER_SECOND / hz boundary * to keep time flowing at a relatively constant rate. If the random * increments have already pushed us past the projected offset, do nothing. */ static void tcp_isn_tick(void *xtp) { u_int32_t projected_offset; ISN_LOCK(); projected_offset = isn_offset_old + ISN_BYTES_PER_SECOND / 100; if (projected_offset > isn_offset) isn_offset = projected_offset; isn_offset_old = isn_offset; callout_reset(&isn_callout, hz/100, tcp_isn_tick, NULL); ISN_UNLOCK(); } /* * When a specific ICMP unreachable message is received and the * connection state is SYN-SENT, drop the connection. This behavior * is controlled by the icmp_may_rst sysctl. */ struct inpcb * tcp_drop_syn_sent(struct inpcb *inp, int errno) { struct tcpcb *tp; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); if ((inp->inp_vflag & INP_TIMEWAIT) || (inp->inp_vflag & INP_DROPPED)) return (inp); tp = intotcpcb(inp); if (tp->t_state != TCPS_SYN_SENT) return (inp); tp = tcp_drop(tp, errno); if (tp != NULL) return (inp); else return (NULL); } /* * When `need fragmentation' ICMP is received, update our idea of the MSS * based on the new value in the route. Also nudge TCP to send something, * since we know the packet we just sent was dropped. * This duplicates some code in the tcp_mss() function in tcp_input.c. */ struct inpcb * tcp_mtudisc(struct inpcb *inp, int errno) { struct tcpcb *tp; struct socket *so = inp->inp_socket; u_int maxmtu; u_int romtu; int mss; #ifdef INET6 int isipv6; #endif /* INET6 */ INP_LOCK_ASSERT(inp); if ((inp->inp_vflag & INP_TIMEWAIT) || (inp->inp_vflag & INP_DROPPED)) return (inp); tp = intotcpcb(inp); KASSERT(tp != NULL, ("tcp_mtudisc: tp == NULL")); #ifdef INET6 isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0; #endif maxmtu = tcp_hc_getmtu(&inp->inp_inc); /* IPv4 and IPv6 */ romtu = #ifdef INET6 isipv6 ? tcp_maxmtu6(&inp->inp_inc, NULL) : #endif /* INET6 */ tcp_maxmtu(&inp->inp_inc, NULL); if (!maxmtu) maxmtu = romtu; else maxmtu = min(maxmtu, romtu); if (!maxmtu) { tp->t_maxopd = tp->t_maxseg = #ifdef INET6 isipv6 ? tcp_v6mssdflt : #endif /* INET6 */ tcp_mssdflt; return (inp); } mss = maxmtu - #ifdef INET6 (isipv6 ? sizeof(struct ip6_hdr) + sizeof(struct tcphdr) : #endif /* INET6 */ sizeof(struct tcpiphdr) #ifdef INET6 ) #endif /* INET6 */ ; /* * XXX - The above conditional probably violates the TCP * spec. The problem is that, since we don't know the * other end's MSS, we are supposed to use a conservative * default. But, if we do that, then MTU discovery will * never actually take place, because the conservative * default is much less than the MTUs typically seen * on the Internet today. For the moment, we'll sweep * this under the carpet. * * The conservative default might not actually be a problem * if the only case this occurs is when sending an initial * SYN with options and data to a host we've never talked * to before. Then, they will reply with an MSS value which * will get recorded and the new parameters should get * recomputed. For Further Study. */ if (tp->t_maxopd <= mss) return (inp); tp->t_maxopd = mss; if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP && (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP) mss -= TCPOLEN_TSTAMP_APPA; #if (MCLBYTES & (MCLBYTES - 1)) == 0 if (mss > MCLBYTES) mss &= ~(MCLBYTES-1); #else if (mss > MCLBYTES) mss = mss / MCLBYTES * MCLBYTES; #endif if (so->so_snd.sb_hiwat < mss) mss = so->so_snd.sb_hiwat; tp->t_maxseg = mss; tcpstat.tcps_mturesent++; tp->t_rtttime = 0; tp->snd_nxt = tp->snd_una; tcp_free_sackholes(tp); tp->snd_recover = tp->snd_max; if (tp->sack_enable) EXIT_FASTRECOVERY(tp); tcp_output(tp); return (inp); } /* * Look-up the routing entry to the peer of this inpcb. If no route * is found and it cannot be allocated, then return NULL. This routine * is called by TCP routines that access the rmx structure and by tcp_mss * to get the interface MTU. */ u_long tcp_maxmtu(struct in_conninfo *inc, int *flags) { struct route sro; struct sockaddr_in *dst; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu with NULL in_conninfo pointer")); bzero(&sro, sizeof(sro)); if (inc->inc_faddr.s_addr != INADDR_ANY) { dst = (struct sockaddr_in *)&sro.ro_dst; dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = inc->inc_faddr; rtalloc_ign(&sro, RTF_CLONING); } if (sro.ro_rt != NULL) { ifp = sro.ro_rt->rt_ifp; if (sro.ro_rt->rt_rmx.rmx_mtu == 0) maxmtu = ifp->if_mtu; else maxmtu = min(sro.ro_rt->rt_rmx.rmx_mtu, ifp->if_mtu); /* Report additional interface capabilities. */ if (flags != NULL) { if (ifp->if_capenable & IFCAP_TSO4 && ifp->if_hwassist & CSUM_TSO) *flags |= CSUM_TSO; } RTFREE(sro.ro_rt); } return (maxmtu); } #ifdef INET6 u_long tcp_maxmtu6(struct in_conninfo *inc, int *flags) { struct route_in6 sro6; struct ifnet *ifp; u_long maxmtu = 0; KASSERT(inc != NULL, ("tcp_maxmtu6 with NULL in_conninfo pointer")); bzero(&sro6, sizeof(sro6)); if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) { sro6.ro_dst.sin6_family = AF_INET6; sro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); sro6.ro_dst.sin6_addr = inc->inc6_faddr; rtalloc_ign((struct route *)&sro6, RTF_CLONING); } if (sro6.ro_rt != NULL) { ifp = sro6.ro_rt->rt_ifp; if (sro6.ro_rt->rt_rmx.rmx_mtu == 0) maxmtu = IN6_LINKMTU(sro6.ro_rt->rt_ifp); else maxmtu = min(sro6.ro_rt->rt_rmx.rmx_mtu, IN6_LINKMTU(sro6.ro_rt->rt_ifp)); /* Report additional interface capabilities. */ if (flags != NULL) { if (ifp->if_capenable & IFCAP_TSO6 && ifp->if_hwassist & CSUM_TSO) *flags |= CSUM_TSO; } RTFREE(sro6.ro_rt); } return (maxmtu); } #endif /* INET6 */ #ifdef IPSEC /* compute ESP/AH header size for TCP, including outer IP header. */ size_t ipsec_hdrsiz_tcp(struct tcpcb *tp) { struct inpcb *inp; struct mbuf *m; size_t hdrsiz; struct ip *ip; #ifdef INET6 struct ip6_hdr *ip6; #endif struct tcphdr *th; if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL)) return (0); MGETHDR(m, M_DONTWAIT, MT_DATA); if (!m) return (0); #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); tcpip_fillheaders(inp, ip6, th); hdrsiz = ipsec6_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } else #endif /* INET6 */ { ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr); tcpip_fillheaders(inp, ip, th); hdrsiz = ipsec4_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp); } m_free(m); return (hdrsiz); } #endif /*IPSEC*/ /* * Move a TCP connection into TIME_WAIT state. * tcbinfo is locked. * inp is locked, and is unlocked before returning. */ void tcp_twstart(struct tcpcb *tp) { struct tcptw *tw; struct inpcb *inp = tp->t_inpcb; int acknow; struct socket *so; INP_INFO_WLOCK_ASSERT(&tcbinfo); /* tcp_timer_2msl_reset(). */ INP_LOCK_ASSERT(inp); if (nolocaltimewait && in_localip(inp->inp_faddr)) { tp = tcp_close(tp); if (tp != NULL) INP_UNLOCK(inp); return; } tw = uma_zalloc(tcptw_zone, M_NOWAIT); if (tw == NULL) { tw = tcp_timer_2msl_tw(1); if (tw == NULL) { tp = tcp_close(tp); if (tp != NULL) INP_UNLOCK(inp); return; } } tw->tw_inpcb = inp; /* * Recover last window size sent. */ tw->last_win = (tp->rcv_adv - tp->rcv_nxt) >> tp->rcv_scale; /* * Set t_recent if timestamps are used on the connection. */ if ((tp->t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP|TF_NOOPT)) == (TF_REQ_TSTMP|TF_RCVD_TSTMP)) tw->t_recent = tp->ts_recent; else tw->t_recent = 0; tw->snd_nxt = tp->snd_nxt; tw->rcv_nxt = tp->rcv_nxt; tw->iss = tp->iss; tw->irs = tp->irs; tw->t_starttime = tp->t_starttime; tw->tw_time = 0; /* XXX * If this code will * be used for fin-wait-2 state also, then we may need * a ts_recent from the last segment. */ acknow = tp->t_flags & TF_ACKNOW; /* * First, discard tcpcb state, which includes stopping its timers and * freeing it. tcp_discardcb() used to also release the inpcb, but * that work is now done in the caller. * * Note: soisdisconnected() call used to be made in tcp_discardcb(), * and might not be needed here any longer. */ tcp_discardcb(tp); so = inp->inp_socket; soisdisconnected(so); SOCK_LOCK(so); tw->tw_cred = crhold(so->so_cred); tw->tw_so_options = so->so_options; SOCK_UNLOCK(so); if (acknow) tcp_twrespond(tw, TH_ACK); inp->inp_ppcb = tw; inp->inp_vflag |= INP_TIMEWAIT; tcp_timer_2msl_reset(tw, 0); /* * If the inpcb owns the sole reference to the socket, then we can * detach and free the socket as it is not needed in time wait. */ if (inp->inp_vflag & INP_SOCKREF) { KASSERT(so->so_state & SS_PROTOREF, ("tcp_twstart: !SS_PROTOREF")); inp->inp_vflag &= ~INP_SOCKREF; INP_UNLOCK(inp); ACCEPT_LOCK(); SOCK_LOCK(so); so->so_state &= ~SS_PROTOREF; sofree(so); } else INP_UNLOCK(inp); } #if 0 /* * The appromixate rate of ISN increase of Microsoft TCP stacks; * the actual rate is slightly higher due to the addition of * random positive increments. * * Most other new OSes use semi-randomized ISN values, so we * do not need to worry about them. */ #define MS_ISN_BYTES_PER_SECOND 250000 /* * Determine if the ISN we will generate has advanced beyond the last * sequence number used by the previous connection. If so, indicate * that it is safe to recycle this tw socket by returning 1. */ int tcp_twrecycleable(struct tcptw *tw) { tcp_seq new_iss = tw->iss; tcp_seq new_irs = tw->irs; INP_INFO_WLOCK_ASSERT(&tcbinfo); new_iss += (ticks - tw->t_starttime) * (ISN_BYTES_PER_SECOND / hz); new_irs += (ticks - tw->t_starttime) * (MS_ISN_BYTES_PER_SECOND / hz); if (SEQ_GT(new_iss, tw->snd_nxt) && SEQ_GT(new_irs, tw->rcv_nxt)) return (1); else return (0); } #endif void tcp_twclose(struct tcptw *tw, int reuse) { struct socket *so; struct inpcb *inp; /* * At this point, we are in one of two situations: * * (1) We have no socket, just an inpcb<->twtcp pair. We can free * all state. * * (2) We have a socket -- if we own a reference, release it and * notify the socket layer. */ inp = tw->tw_inpcb; KASSERT((inp->inp_vflag & INP_TIMEWAIT), ("tcp_twclose: !timewait")); KASSERT(intotw(inp) == tw, ("tcp_twclose: inp_ppcb != tw")); INP_INFO_WLOCK_ASSERT(&tcbinfo); /* tcp_timer_2msl_stop(). */ INP_LOCK_ASSERT(inp); tw->tw_inpcb = NULL; tcp_timer_2msl_stop(tw); inp->inp_ppcb = NULL; in_pcbdrop(inp); so = inp->inp_socket; if (so != NULL) { /* * If there's a socket, handle two cases: first, we own a * strong reference, which we will now release, or we don't * in which case another reference exists (XXXRW: think * about this more), and we don't need to take action. */ if (inp->inp_vflag & INP_SOCKREF) { inp->inp_vflag &= ~INP_SOCKREF; INP_UNLOCK(inp); ACCEPT_LOCK(); SOCK_LOCK(so); KASSERT(so->so_state & SS_PROTOREF, ("tcp_twclose: INP_SOCKREF && !SS_PROTOREF")); so->so_state &= ~SS_PROTOREF; sofree(so); } else { /* * If we don't own the only reference, the socket and * inpcb need to be left around to be handled by * tcp_usr_detach() later. */ INP_UNLOCK(inp); } } else { #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) in6_pcbfree(inp); else #endif in_pcbfree(inp); } tcpstat.tcps_closed++; crfree(tw->tw_cred); tw->tw_cred = NULL; if (reuse) return; uma_zfree(tcptw_zone, tw); } int tcp_twrespond(struct tcptw *tw, int flags) { struct inpcb *inp = tw->tw_inpcb; struct tcphdr *th; struct mbuf *m; struct ip *ip = NULL; u_int8_t *optp; u_int hdrlen, optlen; int error; #ifdef INET6 struct ip6_hdr *ip6 = NULL; int isipv6 = inp->inp_inc.inc_isipv6; #endif INP_LOCK_ASSERT(inp); m = m_gethdr(M_DONTWAIT, MT_DATA); if (m == NULL) return (ENOBUFS); m->m_data += max_linkhdr; #ifdef MAC mac_create_mbuf_from_inpcb(inp, m); #endif #ifdef INET6 if (isipv6) { hdrlen = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); ip6 = mtod(m, struct ip6_hdr *); th = (struct tcphdr *)(ip6 + 1); tcpip_fillheaders(inp, ip6, th); } else #endif { hdrlen = sizeof(struct tcpiphdr); ip = mtod(m, struct ip *); th = (struct tcphdr *)(ip + 1); tcpip_fillheaders(inp, ip, th); } optp = (u_int8_t *)(th + 1); /* * Send a timestamp and echo-reply if both our side and our peer * have sent timestamps in our SYN's and this is not a RST. */ if (tw->t_recent && flags == TH_ACK) { u_int32_t *lp = (u_int32_t *)optp; /* Form timestamp option as shown in appendix A of RFC 1323. */ *lp++ = htonl(TCPOPT_TSTAMP_HDR); *lp++ = htonl(ticks); *lp = htonl(tw->t_recent); optp += TCPOLEN_TSTAMP_APPA; } optlen = optp - (u_int8_t *)(th + 1); m->m_len = hdrlen + optlen; m->m_pkthdr.len = m->m_len; KASSERT(max_linkhdr + m->m_len <= MHLEN, ("tcptw: mbuf too small")); th->th_seq = htonl(tw->snd_nxt); th->th_ack = htonl(tw->rcv_nxt); th->th_off = (sizeof(struct tcphdr) + optlen) >> 2; th->th_flags = flags; th->th_win = htons(tw->last_win); #ifdef INET6 if (isipv6) { th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), sizeof(struct tcphdr) + optlen); ip6->ip6_hlim = in6_selecthlim(inp, NULL); error = ip6_output(m, inp->in6p_outputopts, NULL, (tw->tw_so_options & SO_DONTROUTE), NULL, NULL, inp); } else #endif { th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htons(sizeof(struct tcphdr) + optlen + IPPROTO_TCP)); m->m_pkthdr.csum_flags = CSUM_TCP; m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); ip->ip_len = m->m_pkthdr.len; if (path_mtu_discovery) ip->ip_off |= IP_DF; error = ip_output(m, inp->inp_options, NULL, ((tw->tw_so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0), NULL, inp); } if (flags & TH_ACK) tcpstat.tcps_sndacks++; else tcpstat.tcps_sndctrl++; tcpstat.tcps_sndtotal++; return (error); } /* * TCP BANDWIDTH DELAY PRODUCT WINDOW LIMITING * * This code attempts to calculate the bandwidth-delay product as a * means of determining the optimal window size to maximize bandwidth, * minimize RTT, and avoid the over-allocation of buffers on interfaces and * routers. This code also does a fairly good job keeping RTTs in check * across slow links like modems. We implement an algorithm which is very * similar (but not meant to be) TCP/Vegas. The code operates on the * transmitter side of a TCP connection and so only effects the transmit * side of the connection. * * BACKGROUND: TCP makes no provision for the management of buffer space * at the end points or at the intermediate routers and switches. A TCP * stream, whether using NewReno or not, will eventually buffer as * many packets as it is able and the only reason this typically works is * due to the fairly small default buffers made available for a connection * (typicaly 16K or 32K). As machines use larger windows and/or window * scaling it is now fairly easy for even a single TCP connection to blow-out * all available buffer space not only on the local interface, but on * intermediate routers and switches as well. NewReno makes a misguided * attempt to 'solve' this problem by waiting for an actual failure to occur, * then backing off, then steadily increasing the window again until another * failure occurs, ad-infinitum. This results in terrible oscillation that * is only made worse as network loads increase and the idea of intentionally * blowing out network buffers is, frankly, a terrible way to manage network * resources. * * It is far better to limit the transmit window prior to the failure * condition being achieved. There are two general ways to do this: First * you can 'scan' through different transmit window sizes and locate the * point where the RTT stops increasing, indicating that you have filled the * pipe, then scan backwards until you note that RTT stops decreasing, then * repeat ad-infinitum. This method works in principle but has severe * implementation issues due to RTT variances, timer granularity, and * instability in the algorithm which can lead to many false positives and * create oscillations as well as interact badly with other TCP streams * implementing the same algorithm. * * The second method is to limit the window to the bandwidth delay product * of the link. This is the method we implement. RTT variances and our * own manipulation of the congestion window, bwnd, can potentially * destabilize the algorithm. For this reason we have to stabilize the * elements used to calculate the window. We do this by using the minimum * observed RTT, the long term average of the observed bandwidth, and * by adding two segments worth of slop. It isn't perfect but it is able * to react to changing conditions and gives us a very stable basis on * which to extend the algorithm. */ void tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq) { u_long bw; u_long bwnd; int save_ticks; INP_LOCK_ASSERT(tp->t_inpcb); /* * If inflight_enable is disabled in the middle of a tcp connection, * make sure snd_bwnd is effectively disabled. */ if (tcp_inflight_enable == 0 || tp->t_rttlow < tcp_inflight_rttthresh) { tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_bandwidth = 0; return; } /* * Figure out the bandwidth. Due to the tick granularity this * is a very rough number and it MUST be averaged over a fairly * long period of time. XXX we need to take into account a link * that is not using all available bandwidth, but for now our * slop will ramp us up if this case occurs and the bandwidth later * increases. * * Note: if ticks rollover 'bw' may wind up negative. We must * effectively reset t_bw_rtttime for this case. */ save_ticks = ticks; if ((u_int)(save_ticks - tp->t_bw_rtttime) < 1) return; bw = (int64_t)(ack_seq - tp->t_bw_rtseq) * hz / (save_ticks - tp->t_bw_rtttime); tp->t_bw_rtttime = save_ticks; tp->t_bw_rtseq = ack_seq; if (tp->t_bw_rtttime == 0 || (int)bw < 0) return; bw = ((int64_t)tp->snd_bandwidth * 15 + bw) >> 4; tp->snd_bandwidth = bw; /* * Calculate the semi-static bandwidth delay product, plus two maximal * segments. The additional slop puts us squarely in the sweet * spot and also handles the bandwidth run-up case and stabilization. * Without the slop we could be locking ourselves into a lower * bandwidth. * * Situations Handled: * (1) Prevents over-queueing of packets on LANs, especially on * high speed LANs, allowing larger TCP buffers to be * specified, and also does a good job preventing * over-queueing of packets over choke points like modems * (at least for the transmit side). * * (2) Is able to handle changing network loads (bandwidth * drops so bwnd drops, bandwidth increases so bwnd * increases). * * (3) Theoretically should stabilize in the face of multiple * connections implementing the same algorithm (this may need * a little work). * * (4) Stability value (defaults to 20 = 2 maximal packets) can * be adjusted with a sysctl but typically only needs to be * on very slow connections. A value no smaller then 5 * should be used, but only reduce this default if you have * no other choice. */ #define USERTT ((tp->t_srtt + tp->t_rttbest) / 2) bwnd = (int64_t)bw * USERTT / (hz << TCP_RTT_SHIFT) + tcp_inflight_stab * tp->t_maxseg / 10; #undef USERTT if (tcp_inflight_debug > 0) { static int ltime; if ((u_int)(ticks - ltime) >= hz / tcp_inflight_debug) { ltime = ticks; printf("%p bw %ld rttbest %d srtt %d bwnd %ld\n", tp, bw, tp->t_rttbest, tp->t_srtt, bwnd ); } } if ((long)bwnd < tcp_inflight_min) bwnd = tcp_inflight_min; if (bwnd > tcp_inflight_max) bwnd = tcp_inflight_max; if ((long)bwnd < tp->t_maxseg * 2) bwnd = tp->t_maxseg * 2; tp->snd_bwnd = bwnd; } #ifdef TCP_SIGNATURE /* * Callback function invoked by m_apply() to digest TCP segment data * contained within an mbuf chain. */ static int tcp_signature_apply(void *fstate, void *data, u_int len) { MD5Update(fstate, (u_char *)data, len); return (0); } /* * Compute TCP-MD5 hash of a TCPv4 segment. (RFC2385) * * Parameters: * m pointer to head of mbuf chain * off0 offset to TCP header within the mbuf chain * len length of TCP segment data, excluding options * optlen length of TCP segment options * buf pointer to storage for computed MD5 digest * direction direction of flow (IPSEC_DIR_INBOUND or OUTBOUND) * * We do this over ip, tcphdr, segment data, and the key in the SADB. * When called from tcp_input(), we can be sure that th_sum has been * zeroed out and verified already. * * This function is for IPv4 use only. Calling this function with an * IPv6 packet in the mbuf chain will yield undefined results. * * Return 0 if successful, otherwise return -1. * * XXX The key is retrieved from the system's PF_KEY SADB, by keying a * search with the destination IP address, and a 'magic SPI' to be * determined by the application. This is hardcoded elsewhere to 1179 * right now. Another branch of this code exists which uses the SPD to * specify per-application flows but it is unstable. */ int tcp_signature_compute(struct mbuf *m, int off0, int len, int optlen, u_char *buf, u_int direction) { union sockaddr_union dst; struct ippseudo ippseudo; MD5_CTX ctx; int doff; struct ip *ip; struct ipovly *ipovly; struct secasvar *sav; struct tcphdr *th; u_short savecsum; KASSERT(m != NULL, ("NULL mbuf chain")); KASSERT(buf != NULL, ("NULL signature pointer")); /* Extract the destination from the IP header in the mbuf. */ ip = mtod(m, struct ip *); bzero(&dst, sizeof(union sockaddr_union)); dst.sa.sa_len = sizeof(struct sockaddr_in); dst.sa.sa_family = AF_INET; dst.sin.sin_addr = (direction == IPSEC_DIR_INBOUND) ? ip->ip_src : ip->ip_dst; /* Look up an SADB entry which matches the address of the peer. */ sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); if (sav == NULL) { printf("%s: SADB lookup failed for %s\n", __func__, inet_ntoa(dst.sin.sin_addr)); return (EINVAL); } MD5Init(&ctx); ipovly = (struct ipovly *)ip; th = (struct tcphdr *)((u_char *)ip + off0); doff = off0 + sizeof(struct tcphdr) + optlen; /* * Step 1: Update MD5 hash with IP pseudo-header. * * XXX The ippseudo header MUST be digested in network byte order, * or else we'll fail the regression test. Assume all fields we've * been doing arithmetic on have been in host byte order. * XXX One cannot depend on ipovly->ih_len here. When called from * tcp_output(), the underlying ip_len member has not yet been set. */ ippseudo.ippseudo_src = ipovly->ih_src; ippseudo.ippseudo_dst = ipovly->ih_dst; ippseudo.ippseudo_pad = 0; ippseudo.ippseudo_p = IPPROTO_TCP; ippseudo.ippseudo_len = htons(len + sizeof(struct tcphdr) + optlen); MD5Update(&ctx, (char *)&ippseudo, sizeof(struct ippseudo)); /* * Step 2: Update MD5 hash with TCP header, excluding options. * The TCP checksum must be set to zero. */ savecsum = th->th_sum; th->th_sum = 0; MD5Update(&ctx, (char *)th, sizeof(struct tcphdr)); th->th_sum = savecsum; /* * Step 3: Update MD5 hash with TCP segment data. * Use m_apply() to avoid an early m_pullup(). */ if (len > 0) m_apply(m, doff, len, tcp_signature_apply, &ctx); /* * Step 4: Update MD5 hash with shared secret. */ MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth)); MD5Final(buf, &ctx); key_sa_recordxfer(sav, m); KEY_FREESAV(&sav); return (0); } #endif /* TCP_SIGNATURE */ static int sysctl_drop(SYSCTL_HANDLER_ARGS) { /* addrs[0] is a foreign socket, addrs[1] is a local one. */ struct sockaddr_storage addrs[2]; struct inpcb *inp; struct tcpcb *tp; struct tcptw *tw; struct sockaddr_in *fin, *lin; #ifdef INET6 struct sockaddr_in6 *fin6, *lin6; struct in6_addr f6, l6; #endif int error; inp = NULL; fin = lin = NULL; #ifdef INET6 fin6 = lin6 = NULL; #endif error = 0; if (req->oldptr != NULL || req->oldlen != 0) return (EINVAL); if (req->newptr == NULL) return (EPERM); if (req->newlen < sizeof(addrs)) return (ENOMEM); error = SYSCTL_IN(req, &addrs, sizeof(addrs)); if (error) return (error); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: fin6 = (struct sockaddr_in6 *)&addrs[0]; lin6 = (struct sockaddr_in6 *)&addrs[1]; if (fin6->sin6_len != sizeof(struct sockaddr_in6) || lin6->sin6_len != sizeof(struct sockaddr_in6)) return (EINVAL); if (IN6_IS_ADDR_V4MAPPED(&fin6->sin6_addr)) { if (!IN6_IS_ADDR_V4MAPPED(&lin6->sin6_addr)) return (EINVAL); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[0]); in6_sin6_2_sin_in_sock((struct sockaddr *)&addrs[1]); fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; break; } error = sa6_embedscope(fin6, ip6_use_defzone); if (error) return (error); error = sa6_embedscope(lin6, ip6_use_defzone); if (error) return (error); break; #endif case AF_INET: fin = (struct sockaddr_in *)&addrs[0]; lin = (struct sockaddr_in *)&addrs[1]; if (fin->sin_len != sizeof(struct sockaddr_in) || lin->sin_len != sizeof(struct sockaddr_in)) return (EINVAL); break; default: return (EINVAL); } INP_INFO_WLOCK(&tcbinfo); switch (addrs[0].ss_family) { #ifdef INET6 case AF_INET6: inp = in6_pcblookup_hash(&tcbinfo, &f6, fin6->sin6_port, &l6, lin6->sin6_port, 0, NULL); break; #endif case AF_INET: inp = in_pcblookup_hash(&tcbinfo, fin->sin_addr, fin->sin_port, lin->sin_addr, lin->sin_port, 0, NULL); break; } if (inp != NULL) { INP_LOCK(inp); if (inp->inp_vflag & INP_TIMEWAIT) { /* * XXXRW: There currently exists a state where an * inpcb is present, but its timewait state has been * discarded. For now, don't allow dropping of this * type of inpcb. */ tw = intotw(inp); if (tw != NULL) tcp_twclose(tw, 0); } else if (!(inp->inp_vflag & INP_DROPPED) && !(inp->inp_socket->so_options & SO_ACCEPTCONN)) { tp = intotcpcb(inp); tcp_drop(tp, ECONNABORTED); } INP_UNLOCK(inp); } else error = ESRCH; INP_INFO_WUNLOCK(&tcbinfo); return (error); } SYSCTL_PROC(_net_inet_tcp, TCPCTL_DROP, drop, CTLTYPE_STRUCT|CTLFLAG_WR|CTLFLAG_SKIP, NULL, 0, sysctl_drop, "", "Drop TCP connection"); Index: head/sys/netinet/tcp_usrreq.c =================================================================== --- head/sys/netinet/tcp_usrreq.c (revision 167035) +++ head/sys/netinet/tcp_usrreq.c (revision 167036) @@ -1,1903 +1,1908 @@ /*- * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. * Copyright (c) 2006-2007 Robert N. M. Watson * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 * $FreeBSD$ */ #include "opt_ddb.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_tcpdebug.h" #include #include #include #include #include #include #ifdef INET6 #include #endif /* INET6 */ #include #include #include #include #include #ifdef DDB #include #endif #include #include #include #include #ifdef INET6 #include #endif #include #ifdef INET6 #include #endif #include #include #ifdef INET6 #include #include #endif #include #include #include #include #include #include #ifdef TCPDEBUG #include #endif /* * TCP protocol interface to socket abstraction. */ extern char *tcpstates[]; /* XXX ??? */ static int tcp_attach(struct socket *); static int tcp_connect(struct tcpcb *, struct sockaddr *, struct thread *td); #ifdef INET6 static int tcp6_connect(struct tcpcb *, struct sockaddr *, struct thread *td); #endif /* INET6 */ static void tcp_disconnect(struct tcpcb *); static void tcp_usrclosed(struct tcpcb *); static void tcp_fill_info(struct tcpcb *, struct tcp_info *); #ifdef TCPDEBUG #define TCPDEBUG0 int ostate = 0 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \ tcp_trace(TA_USER, ostate, tp, 0, 0, req) #else #define TCPDEBUG0 #define TCPDEBUG1() #define TCPDEBUG2(req) #endif /* * TCP attaches to socket via pru_attach(), reserving space, * and an internet control block. */ static int tcp_usr_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; struct tcpcb *tp = NULL; int error; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL")); TCPDEBUG1(); error = tcp_attach(so); if (error) goto out; if ((so->so_options & SO_LINGER) && so->so_linger == 0) so->so_linger = TCP_LINGERTIME; inp = sotoinpcb(so); tp = intotcpcb(inp); out: TCPDEBUG2(PRU_ATTACH); return error; } /* * tcp_detach is called when the socket layer loses its final reference * to the socket, be it a file descriptor reference, a reference from TCP, * etc. At this point, there is only one case in which we will keep around * inpcb state: time wait. * * This function can probably be re-absorbed back into tcp_usr_detach() now * that there is a single detach path. */ static void tcp_detach(struct socket *so, struct inpcb *inp) { struct tcpcb *tp; #ifdef INET6 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0; #endif INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp")); KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so")); tp = intotcpcb(inp); if (inp->inp_vflag & INP_TIMEWAIT) { /* * There are two cases to handle: one in which the time wait * state is being discarded (INP_DROPPED), and one in which * this connection will remain in timewait. In the former, * it is time to discard all state (except tcptw, which has * already been discarded by the timewait close code, which * should be further up the call stack somewhere). In the * latter case, we detach from the socket, but leave the pcb * present until timewait ends. * * XXXRW: Would it be cleaner to free the tcptw here? */ if (inp->inp_vflag & INP_DROPPED) { KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && " "INP_DROPPED && tp != NULL")); #ifdef INET6 if (isipv6) { in6_pcbdetach(inp); in6_pcbfree(inp); } else { #endif in_pcbdetach(inp); in_pcbfree(inp); #ifdef INET6 } #endif } else { #ifdef INET6 if (isipv6) in6_pcbdetach(inp); else #endif in_pcbdetach(inp); INP_UNLOCK(inp); } } else { /* * If the connection is not in timewait, we consider two * two conditions: one in which no further processing is * necessary (dropped || embryonic), and one in which TCP is * not yet done, but no longer requires the socket, so the * pcb will persist for the time being. * * XXXRW: Does the second case still occur? */ if (inp->inp_vflag & INP_DROPPED || tp->t_state < TCPS_SYN_SENT) { tcp_discardcb(tp); #ifdef INET6 if (isipv6) { in6_pcbdetach(inp); in6_pcbfree(inp); } else { #endif in_pcbdetach(inp); in_pcbfree(inp); #ifdef INET6 } #endif } else { #ifdef INET6 if (isipv6) in6_pcbdetach(inp); else #endif in_pcbdetach(inp); } } } /* * pru_detach() detaches the TCP protocol from the socket. * If the protocol state is non-embryonic, then can't * do this directly: have to initiate a pru_disconnect(), * which may finish later; embryonic TCB's can just * be discarded here. */ static void tcp_usr_detach(struct socket *so) { struct inpcb *inp; struct tcpcb *tp; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL")); INP_INFO_WLOCK(&tcbinfo); INP_LOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_detach: inp_socket == NULL")); TCPDEBUG1(); tcp_detach(so, inp); tp = NULL; TCPDEBUG2(PRU_DETACH); INP_INFO_WUNLOCK(&tcbinfo); } /* * Give the socket an address. */ static int tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in *sinp; sinp = (struct sockaddr_in *)nam; if (nam->sa_len != sizeof (*sinp)) return (EINVAL); /* * Must check for multicast addresses and disallow binding * to them. */ if (sinp->sin_family == AF_INET && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) return (EAFNOSUPPORT); TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); error = in_pcbbind(inp, nam, td->td_ucred); out: TCPDEBUG2(PRU_BIND); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #ifdef INET6 static int tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in6 *sin6p; sin6p = (struct sockaddr_in6 *)nam; if (nam->sa_len != sizeof (*sin6p)) return (EINVAL); /* * Must check for multicast addresses and disallow binding * to them. */ if (sin6p->sin6_family == AF_INET6 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) return (EAFNOSUPPORT); TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) inp->inp_vflag |= INP_IPV4; else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { struct sockaddr_in sin; in6_sin6_2_sin(&sin, sin6p); inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; error = in_pcbbind(inp, (struct sockaddr *)&sin, td->td_ucred); goto out; } } error = in6_pcbbind(inp, nam, td->td_ucred); out: TCPDEBUG2(PRU_BIND); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #endif /* INET6 */ /* * Prepare to accept connections. */ static int tcp_usr_listen(struct socket *so, int backlog, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); SOCK_LOCK(so); error = solisten_proto_check(so); if (error == 0 && inp->inp_lport == 0) error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); if (error == 0) { tp->t_state = TCPS_LISTEN; solisten_proto(so, backlog); } SOCK_UNLOCK(so); out: TCPDEBUG2(PRU_LISTEN); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #ifdef INET6 static int tcp6_usr_listen(struct socket *so, int backlog, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); SOCK_LOCK(so); error = solisten_proto_check(so); if (error == 0 && inp->inp_lport == 0) { inp->inp_vflag &= ~INP_IPV4; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) inp->inp_vflag |= INP_IPV4; error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); } if (error == 0) { tp->t_state = TCPS_LISTEN; solisten_proto(so, backlog); } SOCK_UNLOCK(so); out: TCPDEBUG2(PRU_LISTEN); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #endif /* INET6 */ /* * Initiate connection to peer. * Create a template for use in transmissions on this connection. * Enter SYN_SENT state, and mark socket as connecting. * Start keep-alive timer, and seed output sequence space. * Send initial segment on connection. */ static int tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in *sinp; sinp = (struct sockaddr_in *)nam; if (nam->sa_len != sizeof (*sinp)) return (EINVAL); /* * Must disallow TCP ``connections'' to multicast addresses. */ if (sinp->sin_family == AF_INET && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) return (EAFNOSUPPORT); if (jailed(td->td_ucred)) prison_remote_ip(td->td_ucred, 0, &sinp->sin_addr.s_addr); TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); if ((error = tcp_connect(tp, nam, td)) != 0) goto out; error = tcp_output(tp); out: TCPDEBUG2(PRU_CONNECT); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #ifdef INET6 static int tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; struct sockaddr_in6 *sin6p; TCPDEBUG0; sin6p = (struct sockaddr_in6 *)nam; if (nam->sa_len != sizeof (*sin6p)) return (EINVAL); /* * Must disallow TCP ``connections'' to multicast addresses. */ if (sin6p->sin6_family == AF_INET6 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) return (EAFNOSUPPORT); INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = EINVAL; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) { struct sockaddr_in sin; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { error = EINVAL; goto out; } in6_sin6_2_sin(&sin, sin6p); inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0) goto out; error = tcp_output(tp); goto out; } inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; inp->inp_inc.inc_isipv6 = 1; if ((error = tcp6_connect(tp, nam, td)) != 0) goto out; error = tcp_output(tp); out: TCPDEBUG2(PRU_CONNECT); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } #endif /* INET6 */ /* * Initiate disconnect from peer. * If connection never passed embryonic stage, just drop; * else if don't need to let data drain, then can just drop anyways, * else have to begin TCP shutdown process: mark socket disconnecting, * drain unread data, state switch to reflect user close, and * send segment (e.g. FIN) to peer. Socket will be really disconnected * when peer sends FIN and acks ours. * * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. */ static int tcp_usr_disconnect(struct socket *so) { struct inpcb *inp; struct tcpcb *tp = NULL; int error = 0; TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); tcp_disconnect(tp); out: TCPDEBUG2(PRU_DISCONNECT); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } /* * Accept a connection. Essentially all the work is * done at higher levels; just return the address * of the peer, storing through addr. */ static int tcp_usr_accept(struct socket *so, struct sockaddr **nam) { int error = 0; struct inpcb *inp = NULL; struct tcpcb *tp = NULL; struct in_addr addr; in_port_t port = 0; TCPDEBUG0; if (so->so_state & SS_ISDISCONNECTED) return (ECONNABORTED); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNABORTED; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); /* * We inline in_setpeeraddr and COMMON_END here, so that we can * copy the data of interest and defer the malloc until after we * release the lock. */ port = inp->inp_fport; addr = inp->inp_faddr; out: TCPDEBUG2(PRU_ACCEPT); INP_UNLOCK(inp); if (error == 0) *nam = in_sockaddr(port, &addr); return error; } #ifdef INET6 static int tcp6_usr_accept(struct socket *so, struct sockaddr **nam) { struct inpcb *inp = NULL; int error = 0; struct tcpcb *tp = NULL; struct in_addr addr; struct in6_addr addr6; in_port_t port = 0; int v4 = 0; TCPDEBUG0; if (so->so_state & SS_ISDISCONNECTED) return (ECONNABORTED); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNABORTED; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); /* * We inline in6_mapped_peeraddr and COMMON_END here, so that we can * copy the data of interest and defer the malloc until after we * release the lock. */ if (inp->inp_vflag & INP_IPV4) { v4 = 1; port = inp->inp_fport; addr = inp->inp_faddr; } else { port = inp->inp_fport; addr6 = inp->in6p_faddr; } out: TCPDEBUG2(PRU_ACCEPT); INP_UNLOCK(inp); if (error == 0) { if (v4) *nam = in6_v4mapsin6_sockaddr(port, &addr); else *nam = in6_sockaddr(port, &addr6); } return error; } #endif /* INET6 */ /* * This is the wrapper function for in_setsockaddr. We just pass down * the pcbinfo for in_setsockaddr to lock. We don't want to do the locking * here because in_setsockaddr will call malloc and can block. */ static int tcp_sockaddr(struct socket *so, struct sockaddr **nam) { return (in_setsockaddr(so, nam, &tcbinfo)); } /* * This is the wrapper function for in_setpeeraddr. We just pass down * the pcbinfo for in_setpeeraddr to lock. */ static int tcp_peeraddr(struct socket *so, struct sockaddr **nam) { return (in_setpeeraddr(so, nam, &tcbinfo)); } /* * Mark the connection as being incapable of further output. */ static int tcp_usr_shutdown(struct socket *so) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; INP_INFO_WLOCK(&tcbinfo); inp = sotoinpcb(so); KASSERT(inp != NULL, ("inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); socantsendmore(so); tcp_usrclosed(tp); error = tcp_output(tp); out: TCPDEBUG2(PRU_SHUTDOWN); INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (error); } /* * After a receive, possibly send window update to peer. */ static int tcp_usr_rcvd(struct socket *so, int flags) { struct inpcb *inp; struct tcpcb *tp = NULL; int error = 0; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); tcp_output(tp); out: TCPDEBUG2(PRU_RCVD); INP_UNLOCK(inp); return (error); } /* * Do a send by putting data in output queue and updating urgent * marker if URG set. Possibly send more data. Unlike the other * pru_*() routines, the mbuf chains are our responsibility. We * must either enqueue them or free them. The other pru_* routines * generally are caller-frees. */ static int tcp_usr_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; int headlocked = 0; #ifdef INET6 int isipv6; #endif TCPDEBUG0; /* * We require the pcbinfo lock in two cases: * * (1) An implied connect is taking place, which can result in * binding IPs and ports and hence modification of the pcb hash * chains. * * (2) PRUS_EOF is set, resulting in explicit close on the send. */ if ((nam != NULL) || (flags & PRUS_EOF)) { INP_INFO_WLOCK(&tcbinfo); headlocked = 1; } inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { if (control) m_freem(control); if (m) m_freem(m); error = ECONNRESET; goto out; } #ifdef INET6 isipv6 = nam && nam->sa_family == AF_INET6; #endif /* INET6 */ tp = intotcpcb(inp); TCPDEBUG1(); if (control) { /* TCP doesn't do control messages (rights, creds, etc) */ if (control->m_len) { m_freem(control); if (m) m_freem(m); error = EINVAL; goto out; } m_freem(control); /* empty control, just free it */ } if (!(flags & PRUS_OOB)) { sbappendstream(&so->so_snd, m); if (nam && tp->t_state < TCPS_SYN_SENT) { /* * Do implied connect if not yet connected, * initialize window to default value, and * initialize maxseg/maxopd using peer's cached * MSS. */ INP_INFO_WLOCK_ASSERT(&tcbinfo); #ifdef INET6 if (isipv6) error = tcp6_connect(tp, nam, td); else #endif /* INET6 */ error = tcp_connect(tp, nam, td); if (error) goto out; tp->snd_wnd = TTCP_CLIENT_SND_WND; tcp_mss(tp, -1); } if (flags & PRUS_EOF) { /* * Close the send side of the connection after * the data is sent. */ INP_INFO_WLOCK_ASSERT(&tcbinfo); socantsendmore(so); tcp_usrclosed(tp); } if (headlocked) { INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; } if (tp != NULL) { if (flags & PRUS_MORETOCOME) tp->t_flags |= TF_MORETOCOME; error = tcp_output(tp); if (flags & PRUS_MORETOCOME) tp->t_flags &= ~TF_MORETOCOME; } } else { /* * XXXRW: PRUS_EOF not implemented with PRUS_OOB? */ SOCKBUF_LOCK(&so->so_snd); if (sbspace(&so->so_snd) < -512) { SOCKBUF_UNLOCK(&so->so_snd); m_freem(m); error = ENOBUFS; goto out; } /* * According to RFC961 (Assigned Protocols), * the urgent pointer points to the last octet * of urgent data. We continue, however, * to consider it to indicate the first octet * of data past the urgent section. * Otherwise, snd_up should be one lower. */ sbappendstream_locked(&so->so_snd, m); SOCKBUF_UNLOCK(&so->so_snd); if (nam && tp->t_state < TCPS_SYN_SENT) { /* * Do implied connect if not yet connected, * initialize window to default value, and * initialize maxseg/maxopd using peer's cached * MSS. */ INP_INFO_WLOCK_ASSERT(&tcbinfo); #ifdef INET6 if (isipv6) error = tcp6_connect(tp, nam, td); else #endif /* INET6 */ error = tcp_connect(tp, nam, td); if (error) goto out; tp->snd_wnd = TTCP_CLIENT_SND_WND; tcp_mss(tp, -1); INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; } else if (nam) { INP_INFO_WUNLOCK(&tcbinfo); headlocked = 0; } tp->snd_up = tp->snd_una + so->so_snd.sb_cc; tp->t_flags |= TF_FORCEDATA; error = tcp_output(tp); tp->t_flags &= ~TF_FORCEDATA; } out: TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB : ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); INP_UNLOCK(inp); if (headlocked) INP_INFO_WUNLOCK(&tcbinfo); return (error); } /* * Abort the TCP. Drop the connection abruptly. */ static void tcp_usr_abort(struct socket *so) { struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL")); INP_INFO_WLOCK(&tcbinfo); INP_LOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_abort: inp_socket == NULL")); /* * If we still have full TCP state, and we're not dropped, drop. */ if (!(inp->inp_vflag & INP_TIMEWAIT) && !(inp->inp_vflag & INP_DROPPED)) { tp = intotcpcb(inp); TCPDEBUG1(); tcp_drop(tp, ECONNABORTED); TCPDEBUG2(PRU_ABORT); } if (!(inp->inp_vflag & INP_DROPPED)) { SOCK_LOCK(so); so->so_state |= SS_PROTOREF; SOCK_UNLOCK(so); inp->inp_vflag |= INP_SOCKREF; } INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); } /* * TCP socket is closed. Start friendly disconnect. */ static void tcp_usr_close(struct socket *so) { struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL")); INP_INFO_WLOCK(&tcbinfo); INP_LOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_close: inp_socket == NULL")); /* * If we still have full TCP state, and we're not dropped, initiate * a disconnect. */ if (!(inp->inp_vflag & INP_TIMEWAIT) && !(inp->inp_vflag & INP_DROPPED)) { tp = intotcpcb(inp); TCPDEBUG1(); tcp_disconnect(tp); TCPDEBUG2(PRU_CLOSE); } if (!(inp->inp_vflag & INP_DROPPED)) { SOCK_LOCK(so); so->so_state |= SS_PROTOREF; SOCK_UNLOCK(so); inp->inp_vflag |= INP_SOCKREF; } INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); } /* * Receive out-of-band data. */ static int tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) { int error = 0; struct inpcb *inp; struct tcpcb *tp = NULL; TCPDEBUG0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL")); INP_LOCK(inp); if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); TCPDEBUG1(); if ((so->so_oobmark == 0 && (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) || so->so_options & SO_OOBINLINE || tp->t_oobflags & TCPOOB_HADDATA) { error = EINVAL; goto out; } if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { error = EWOULDBLOCK; goto out; } m->m_len = 1; *mtod(m, caddr_t) = tp->t_iobc; if ((flags & MSG_PEEK) == 0) tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); out: TCPDEBUG2(PRU_RCVOOB); INP_UNLOCK(inp); return (error); } struct pr_usrreqs tcp_usrreqs = { .pru_abort = tcp_usr_abort, .pru_accept = tcp_usr_accept, .pru_attach = tcp_usr_attach, .pru_bind = tcp_usr_bind, .pru_connect = tcp_usr_connect, .pru_control = in_control, .pru_detach = tcp_usr_detach, .pru_disconnect = tcp_usr_disconnect, .pru_listen = tcp_usr_listen, .pru_peeraddr = tcp_peeraddr, .pru_rcvd = tcp_usr_rcvd, .pru_rcvoob = tcp_usr_rcvoob, .pru_send = tcp_usr_send, .pru_shutdown = tcp_usr_shutdown, .pru_sockaddr = tcp_sockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = tcp_usr_close, }; #ifdef INET6 struct pr_usrreqs tcp6_usrreqs = { .pru_abort = tcp_usr_abort, .pru_accept = tcp6_usr_accept, .pru_attach = tcp_usr_attach, .pru_bind = tcp6_usr_bind, .pru_connect = tcp6_usr_connect, .pru_control = in6_control, .pru_detach = tcp_usr_detach, .pru_disconnect = tcp_usr_disconnect, .pru_listen = tcp6_usr_listen, .pru_peeraddr = in6_mapped_peeraddr, .pru_rcvd = tcp_usr_rcvd, .pru_rcvoob = tcp_usr_rcvoob, .pru_send = tcp_usr_send, .pru_shutdown = tcp_usr_shutdown, .pru_sockaddr = in6_mapped_sockaddr, .pru_sosetlabel = in_pcbsosetlabel, .pru_close = tcp_usr_close, }; #endif /* INET6 */ /* * Common subroutine to open a TCP connection to remote host specified * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local * port number if needed. Call in_pcbconnect_setup to do the routing and * to choose a local host address (interface). If there is an existing * incarnation of the same connection in TIME-WAIT state and if the remote * host was sending CC options and if the connection duration was < MSL, then * truncate the previous TIME-WAIT state and proceed. * Initialize connection parameters and enter SYN-SENT state. */ static int tcp_connect(tp, nam, td) register struct tcpcb *tp; struct sockaddr *nam; struct thread *td; { struct inpcb *inp = tp->t_inpcb, *oinp; struct socket *so = inp->inp_socket; struct in_addr laddr; u_short lport; int error; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); if (inp->inp_lport == 0) { error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); if (error) return error; } /* * Cannot simply call in_pcbconnect, because there might be an * earlier incarnation of this same connection still in * TIME_WAIT state, creating an ADDRINUSE error. */ laddr = inp->inp_laddr; lport = inp->inp_lport; error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport, &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred); if (error && oinp == NULL) return error; if (oinp) return EADDRINUSE; inp->inp_laddr = laddr; in_pcbrehash(inp); /* * Compute window scaling to request: * Scale to fit into sweet spot. See tcp_syncache.c. * XXX: This should move to tcp_output(). * XXX: This should be based on the actual MSS. */ while (tp->request_r_scale < TCP_MAX_WINSHIFT && (0x1 << tp->request_r_scale) < tcp_minmss) tp->request_r_scale++; soisconnecting(so); tcpstat.tcps_connattempt++; tp->t_state = TCPS_SYN_SENT; callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); tp->iss = tcp_new_isn(tp); tp->t_bw_rtseq = tp->iss; tcp_sendseqinit(tp); return 0; } #ifdef INET6 static int tcp6_connect(tp, nam, td) register struct tcpcb *tp; struct sockaddr *nam; struct thread *td; { struct inpcb *inp = tp->t_inpcb, *oinp; struct socket *so = inp->inp_socket; struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam; struct in6_addr *addr6; int error; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); if (inp->inp_lport == 0) { error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred); if (error) return error; } /* * Cannot simply call in_pcbconnect, because there might be an * earlier incarnation of this same connection still in * TIME_WAIT state, creating an ADDRINUSE error. * in6_pcbladdr() also handles scope zone IDs. */ error = in6_pcbladdr(inp, nam, &addr6); if (error) return error; oinp = in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr, sin6->sin6_port, IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ? addr6 : &inp->in6p_laddr, inp->inp_lport, 0, NULL); if (oinp) return EADDRINUSE; if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) inp->in6p_laddr = *addr6; inp->in6p_faddr = sin6->sin6_addr; inp->inp_fport = sin6->sin6_port; /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */ inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK; if (inp->in6p_flags & IN6P_AUTOFLOWLABEL) inp->in6p_flowinfo |= (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK); in_pcbrehash(inp); /* Compute window scaling to request. */ while (tp->request_r_scale < TCP_MAX_WINSHIFT && (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat) tp->request_r_scale++; soisconnecting(so); tcpstat.tcps_connattempt++; tp->t_state = TCPS_SYN_SENT; callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp); tp->iss = tcp_new_isn(tp); tp->t_bw_rtseq = tp->iss; tcp_sendseqinit(tp); return 0; } #endif /* INET6 */ /* * Export TCP internal state information via a struct tcp_info, based on the * Linux 2.6 API. Not ABI compatible as our constants are mapped differently * (TCP state machine, etc). We export all information using FreeBSD-native * constants -- for example, the numeric values for tcpi_state will differ * from Linux. */ static void tcp_fill_info(tp, ti) struct tcpcb *tp; struct tcp_info *ti; { INP_LOCK_ASSERT(tp->t_inpcb); bzero(ti, sizeof(*ti)); ti->tcpi_state = tp->t_state; if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP)) ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; if (tp->sack_enable) ti->tcpi_options |= TCPI_OPT_SACK; if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) { ti->tcpi_options |= TCPI_OPT_WSCALE; ti->tcpi_snd_wscale = tp->snd_scale; ti->tcpi_rcv_wscale = tp->rcv_scale; } ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT; ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT; ti->tcpi_snd_ssthresh = tp->snd_ssthresh; ti->tcpi_snd_cwnd = tp->snd_cwnd; /* * FreeBSD-specific extension fields for tcp_info. */ ti->tcpi_rcv_space = tp->rcv_wnd; ti->tcpi_snd_wnd = tp->snd_wnd; ti->tcpi_snd_bwnd = tp->snd_bwnd; } /* * The new sockopt interface makes it possible for us to block in the * copyin/out step (if we take a page fault). Taking a page fault at * splnet() is probably a Bad Thing. (Since sockets and pcbs both now * use TSM, there probably isn't any need for this function to run at * splnet() any more. This needs more examination.) * * XXXRW: The locking here is wrong; we may take a page fault while holding * the inpcb lock. */ int tcp_ctloutput(so, sopt) struct socket *so; struct sockopt *sopt; { int error, opt, optval; struct inpcb *inp; struct tcpcb *tp; struct tcp_info ti; error = 0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL")); INP_LOCK(inp); if (sopt->sopt_level != IPPROTO_TCP) { INP_UNLOCK(inp); #ifdef INET6 if (INP_CHECK_SOCKAF(so, AF_INET6)) error = ip6_ctloutput(so, sopt); else #endif /* INET6 */ error = ip_ctloutput(so, sopt); return (error); } if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) { error = ECONNRESET; goto out; } tp = intotcpcb(inp); switch (sopt->sopt_dir) { case SOPT_SET: switch (sopt->sopt_name) { #ifdef TCP_SIGNATURE case TCP_MD5SIG: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; if (optval > 0) tp->t_flags |= TF_SIGNATURE; else tp->t_flags &= ~TF_SIGNATURE; break; #endif /* TCP_SIGNATURE */ case TCP_NODELAY: case TCP_NOOPT: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; switch (sopt->sopt_name) { case TCP_NODELAY: opt = TF_NODELAY; break; case TCP_NOOPT: opt = TF_NOOPT; break; default: opt = 0; /* dead code to fool gcc */ break; } if (optval) tp->t_flags |= opt; else tp->t_flags &= ~opt; break; case TCP_NOPUSH: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; if (optval) tp->t_flags |= TF_NOPUSH; else { tp->t_flags &= ~TF_NOPUSH; error = tcp_output(tp); } break; case TCP_MAXSEG: error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) break; if (optval > 0 && optval <= tp->t_maxseg && optval + 40 >= tcp_minmss) tp->t_maxseg = optval; else error = EINVAL; break; case TCP_INFO: error = EINVAL; break; default: error = ENOPROTOOPT; break; } break; case SOPT_GET: switch (sopt->sopt_name) { #ifdef TCP_SIGNATURE case TCP_MD5SIG: optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0; error = sooptcopyout(sopt, &optval, sizeof optval); break; #endif case TCP_NODELAY: optval = tp->t_flags & TF_NODELAY; error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_MAXSEG: optval = tp->t_maxseg; error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_NOOPT: optval = tp->t_flags & TF_NOOPT; error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_NOPUSH: optval = tp->t_flags & TF_NOPUSH; error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_INFO: tcp_fill_info(tp, &ti); error = sooptcopyout(sopt, &ti, sizeof ti); break; default: error = ENOPROTOOPT; break; } break; } out: INP_UNLOCK(inp); return (error); } /* * tcp_sendspace and tcp_recvspace are the default send and receive window * sizes, respectively. These are obsolescent (this information should * be set by the route). */ u_long tcp_sendspace = 1024*32; SYSCTL_ULONG(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW, &tcp_sendspace , 0, "Maximum outgoing TCP datagram size"); u_long tcp_recvspace = 1024*64; SYSCTL_ULONG(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW, &tcp_recvspace , 0, "Maximum incoming TCP datagram size"); /* * Attach TCP protocol to socket, allocating * internet protocol control block, tcp control block, * bufer space, and entering LISTEN state if to accept connections. */ static int tcp_attach(so) struct socket *so; { register struct tcpcb *tp; struct inpcb *inp; int error; #ifdef INET6 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0; #endif if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { error = soreserve(so, tcp_sendspace, tcp_recvspace); if (error) return (error); } so->so_rcv.sb_flags |= SB_AUTOSIZE; so->so_snd.sb_flags |= SB_AUTOSIZE; INP_INFO_WLOCK(&tcbinfo); error = in_pcballoc(so, &tcbinfo); if (error) { INP_INFO_WUNLOCK(&tcbinfo); return (error); } inp = sotoinpcb(so); #ifdef INET6 if (isipv6) { inp->inp_vflag |= INP_IPV6; inp->in6p_hops = -1; /* use kernel default */ } else #endif inp->inp_vflag |= INP_IPV4; tp = tcp_newtcpcb(inp); if (tp == NULL) { #ifdef INET6 if (isipv6) { in6_pcbdetach(inp); in6_pcbfree(inp); } else { #endif in_pcbdetach(inp); in_pcbfree(inp); #ifdef INET6 } #endif INP_INFO_WUNLOCK(&tcbinfo); return (ENOBUFS); } tp->t_state = TCPS_CLOSED; INP_UNLOCK(inp); INP_INFO_WUNLOCK(&tcbinfo); return (0); } /* * Initiate (or continue) disconnect. * If embryonic state, just send reset (once). * If in ``let data drain'' option and linger null, just drop. * Otherwise (hard), mark socket disconnecting and drop * current input data; switch states based on user close, and * send segment to peer (with FIN). */ static void tcp_disconnect(tp) register struct tcpcb *tp; { struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(inp); /* * Neither tcp_close() nor tcp_drop() should return NULL, as the * socket is still open. */ if (tp->t_state < TCPS_ESTABLISHED) { tp = tcp_close(tp); KASSERT(tp != NULL, ("tcp_disconnect: tcp_close() returned NULL")); } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) { tp = tcp_drop(tp, 0); KASSERT(tp != NULL, ("tcp_disconnect: tcp_drop() returned NULL")); } else { soisdisconnecting(so); sbflush(&so->so_rcv); tcp_usrclosed(tp); if (!(inp->inp_vflag & INP_DROPPED)) tcp_output(tp); } } /* * User issued close, and wish to trail through shutdown states: * if never received SYN, just forget it. If got a SYN from peer, * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. * If already got a FIN from peer, then almost done; go to LAST_ACK * state. In all other cases, have already sent FIN to peer (e.g. * after PRU_SHUTDOWN), and just have to play tedious game waiting * for peer to send FIN or not respond to keep-alives, etc. * We can let the user exit from the close as soon as the FIN is acked. */ static void tcp_usrclosed(tp) register struct tcpcb *tp; { INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(tp->t_inpcb); switch (tp->t_state) { case TCPS_CLOSED: case TCPS_LISTEN: tp->t_state = TCPS_CLOSED; tp = tcp_close(tp); /* * tcp_close() should never return NULL here as the socket is * still open. */ KASSERT(tp != NULL, ("tcp_usrclosed: tcp_close() returned NULL")); break; case TCPS_SYN_SENT: case TCPS_SYN_RECEIVED: tp->t_flags |= TF_NEEDFIN; break; case TCPS_ESTABLISHED: tp->t_state = TCPS_FIN_WAIT_1; break; case TCPS_CLOSE_WAIT: tp->t_state = TCPS_LAST_ACK; break; } if (tp && tp->t_state >= TCPS_FIN_WAIT_2) { soisdisconnected(tp->t_inpcb->inp_socket); /* To prevent the connection hanging in FIN_WAIT_2 forever. */ - if (tp->t_state == TCPS_FIN_WAIT_2) - callout_reset(tp->tt_2msl, tcp_maxidle, + if (tp->t_state == TCPS_FIN_WAIT_2) { + int timeout; + + timeout = (tcp_fast_finwait2_recycle) ? + tcp_finwait2_timeout : tcp_maxidle; + callout_reset(tp->tt_2msl, timeout, tcp_timer_2msl, tp); + } } } #ifdef DDB static void db_print_indent(int indent) { int i; for (i = 0; i < indent; i++) db_printf(" "); } static void db_print_tstate(int t_state) { switch (t_state) { case TCPS_CLOSED: db_printf("TCPS_CLOSED"); return; case TCPS_LISTEN: db_printf("TCPS_LISTEN"); return; case TCPS_SYN_SENT: db_printf("TCPS_SYN_SENT"); return; case TCPS_SYN_RECEIVED: db_printf("TCPS_SYN_RECEIVED"); return; case TCPS_ESTABLISHED: db_printf("TCPS_ESTABLISHED"); return; case TCPS_CLOSE_WAIT: db_printf("TCPS_CLOSE_WAIT"); return; case TCPS_FIN_WAIT_1: db_printf("TCPS_FIN_WAIT_1"); return; case TCPS_CLOSING: db_printf("TCPS_CLOSING"); return; case TCPS_LAST_ACK: db_printf("TCPS_LAST_ACK"); return; case TCPS_FIN_WAIT_2: db_printf("TCPS_FIN_WAIT_2"); return; case TCPS_TIME_WAIT: db_printf("TCPS_TIME_WAIT"); return; default: db_printf("unknown"); return; } } static void db_print_tflags(u_int t_flags) { int comma; comma = 0; if (t_flags & TF_ACKNOW) { db_printf("%sTF_ACKNOW", comma ? ", " : ""); comma = 1; } if (t_flags & TF_DELACK) { db_printf("%sTF_DELACK", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NODELAY) { db_printf("%sTF_NODELAY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NOOPT) { db_printf("%sTF_NOOPT", comma ? ", " : ""); comma = 1; } if (t_flags & TF_SENTFIN) { db_printf("%sTF_SENTFIN", comma ? ", " : ""); comma = 1; } if (t_flags & TF_REQ_SCALE) { db_printf("%sTF_REQ_SCALE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_RCVD_SCALE) { db_printf("%sTF_RECVD_SCALE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_REQ_TSTMP) { db_printf("%sTF_REQ_TSTMP", comma ? ", " : ""); comma = 1; } if (t_flags & TF_RCVD_TSTMP) { db_printf("%sTF_RCVD_TSTMP", comma ? ", " : ""); comma = 1; } if (t_flags & TF_SACK_PERMIT) { db_printf("%sTF_SACK_PERMIT", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NEEDSYN) { db_printf("%sTF_NEEDSYN", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NEEDFIN) { db_printf("%sTF_NEEDFIN", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NOPUSH) { db_printf("%sTF_NOPUSH", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NOPUSH) { db_printf("%sTF_NOPUSH", comma ? ", " : ""); comma = 1; } if (t_flags & TF_MORETOCOME) { db_printf("%sTF_MORETOCOME", comma ? ", " : ""); comma = 1; } if (t_flags & TF_LQ_OVERFLOW) { db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : ""); comma = 1; } if (t_flags & TF_LASTIDLE) { db_printf("%sTF_LASTIDLE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_RXWIN0SENT) { db_printf("%sTF_RXWIN0SENT", comma ? ", " : ""); comma = 1; } if (t_flags & TF_FASTRECOVERY) { db_printf("%sTF_FASTRECOVERY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_WASFRECOVERY) { db_printf("%sTF_WASFRECOVERY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_SIGNATURE) { db_printf("%sTF_SIGNATURE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_FORCEDATA) { db_printf("%sTF_FORCEDATA", comma ? ", " : ""); comma = 1; } if (t_flags & TF_TSO) { db_printf("%sTF_TSO", comma ? ", " : ""); comma = 1; } } static void db_print_toobflags(char t_oobflags) { int comma; comma = 0; if (t_oobflags & TCPOOB_HAVEDATA) { db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : ""); comma = 1; } if (t_oobflags & TCPOOB_HADDATA) { db_printf("%sTCPOOB_HADDATA", comma ? ", " : ""); comma = 1; } } static void db_print_tcpcb(struct tcpcb *tp, const char *name, int indent) { db_print_indent(indent); db_printf("%s at %p\n", name, tp); indent += 2; db_print_indent(indent); db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n", LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks); db_print_indent(indent); db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n", tp->tt_rexmt, tp->tt_persist, tp->tt_keep); db_print_indent(indent); db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", tp->tt_2msl, tp->tt_delack, tp->t_inpcb); db_print_indent(indent); db_printf("t_state: %d (", tp->t_state); db_print_tstate(tp->t_state); db_printf(")\n"); db_print_indent(indent); db_printf("t_flags: 0x%x (", tp->t_flags); db_print_tflags(tp->t_flags); db_printf(")\n"); db_print_indent(indent); db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n", tp->snd_una, tp->snd_max, tp->snd_nxt); db_print_indent(indent); db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n", tp->snd_up, tp->snd_wl1, tp->snd_wl2); db_print_indent(indent); db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n", tp->iss, tp->irs, tp->rcv_nxt); db_print_indent(indent); db_printf("rcv_adv: 0x%08x rcv_wnd: %lu rcv_up: 0x%08x\n", tp->rcv_adv, tp->rcv_wnd, tp->rcv_up); db_print_indent(indent); db_printf("snd_wnd: %lu snd_cwnd: %lu snd_bwnd: %lu\n", tp->snd_wnd, tp->snd_cwnd, tp->snd_bwnd); db_print_indent(indent); db_printf("snd_ssthresh: %lu snd_bandwidth: %lu snd_recover: " "0x%08x\n", tp->snd_ssthresh, tp->snd_bandwidth, tp->snd_recover); db_print_indent(indent); db_printf("t_maxopd: %u t_rcvtime: %lu t_startime: %lu\n", tp->t_maxopd, tp->t_rcvtime, tp->t_starttime); db_print_indent(indent); db_printf("t_rttime: %d t_rtsq: 0x%08x t_bw_rtttime: %d\n", tp->t_rtttime, tp->t_rtseq, tp->t_bw_rtttime); db_print_indent(indent); db_printf("t_bw_rtseq: 0x%08x t_rxtcur: %d t_maxseg: %u " "t_srtt: %d\n", tp->t_bw_rtseq, tp->t_rxtcur, tp->t_maxseg, tp->t_srtt); db_print_indent(indent); db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u " "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin, tp->t_rttbest); db_print_indent(indent); db_printf("t_rttupdated: %lu max_sndwnd: %lu t_softerror: %d\n", tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror); db_print_indent(indent); db_printf("t_oobflags: 0x%x (", tp->t_oobflags); db_print_toobflags(tp->t_oobflags); db_printf(") t_iobc: 0x%02x\n", tp->t_iobc); db_print_indent(indent); db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n", tp->snd_scale, tp->rcv_scale, tp->request_r_scale); db_print_indent(indent); db_printf("requested_s_scale: %u ts_recent: %u ts_recent_age: " "%lu\n", tp->requested_s_scale, tp->ts_recent, tp->ts_recent_age); db_print_indent(indent); db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: " "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev); db_print_indent(indent); db_printf("snd_ssthresh_prev: %lu snd_recover_prev: 0x%08x " "t_badrxtwin: %lu\n", tp->snd_ssthresh_prev, tp->snd_recover_prev, tp->t_badrxtwin); db_print_indent(indent); db_printf("snd_limited: %u rcv_second: %lu rcv_pps: %lu " "tcv_byps: %lu\n", tp->snd_limited, tp->rcv_second, tp->rcv_pps, tp->rcv_byps); db_print_indent(indent); db_printf("sack_enable: %d snd_numholes: %d snd_holes first: %p\n", tp->sack_enable, tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes)); db_print_indent(indent); db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: " "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata); /* Skip sackblks, sackhint. */ db_print_indent(indent); db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n", tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt); } DB_SHOW_COMMAND(tcpcb, db_show_tcpcb) { struct tcpcb *tp; if (!have_addr) { db_printf("usage: show tcpcb \n"); return; } tp = (struct tcpcb *)addr; db_print_tcpcb(tp, "tcpcb", 0); } #endif Index: head/sys/netinet/tcp_var.h =================================================================== --- head/sys/netinet/tcp_var.h (revision 167035) +++ head/sys/netinet/tcp_var.h (revision 167036) @@ -1,575 +1,578 @@ /*- * 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 callout *tt_rexmt; /* retransmit timer */ struct callout *tt_persist; /* retransmit persistence */ struct callout *tt_keep; /* keepalive */ struct callout *tt_2msl; /* 2*msl TIME_WAIT timer */ struct callout *tt_delack; /* delayed ACK 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 */ /* anti DoS counters */ u_long rcv_second; /* start of interval second */ u_long rcv_pps; /* received packets per second */ u_long rcv_byps; /* received bytes per second */ /* 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. */ struct tcpopt { u_long to_flags; /* which options are present */ #define TOF_TS 0x0001 /* timestamp */ #define TOF_MSS 0x0010 /* maximum segment size */ #define TOF_SCALE 0x0020 /* window scaling */ #define TOF_SIGNATURE 0x0040 /* signature option present */ #define TOF_SIGLEN 0x0080 /* signature length valid (RFC2385) */ #define TOF_SACK 0x0100 /* Peer sent SACK option */ u_int32_t to_tsval; u_int32_t to_tsecr; u_int16_t to_mss; u_int8_t to_requested_s_scale; u_int8_t to_nsacks; /* number of SACK blocks */ u_char *to_sacks; /* pointer to the first SACK blocks */ }; /* * 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_minmssoverload; 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 */ 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 *); struct tcpcb * tcp_timers(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 *, 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_ */ Index: head/usr.bin/netstat/inet.c =================================================================== --- head/usr.bin/netstat/inet.c (revision 167035) +++ head/usr.bin/netstat/inet.c (revision 167036) @@ -1,881 +1,882 @@ /* * Copyright (c) 1983, 1988, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #if 0 #ifndef lint static char sccsid[] = "@(#)inet.c 8.5 (Berkeley) 5/24/95"; #endif /* not lint */ #endif #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif /* INET6 */ #include #include #include #include #include #include #include #include #include #define TCPSTATES #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "netstat.h" char *inetname (struct in_addr *); void inetprint (struct in_addr *, int, const char *, int); #ifdef INET6 static int udp_done, tcp_done; #endif /* INET6 */ /* * Print a summary of connections related to an Internet * protocol. For TCP, also give state of connection. * Listening processes (aflag) are suppressed unless the * -a (all) flag is specified. */ void protopr(u_long proto, /* for sysctl version we pass proto # */ const char *name, int af1) { int istcp; static int first = 1; char *buf; const char *mibvar, *vchar; struct tcpcb *tp = NULL; struct inpcb *inp; struct xinpgen *xig, *oxig; struct xsocket *so; size_t len; istcp = 0; switch (proto) { case IPPROTO_TCP: #ifdef INET6 if (tcp_done != 0) return; else tcp_done = 1; #endif istcp = 1; mibvar = "net.inet.tcp.pcblist"; break; case IPPROTO_UDP: #ifdef INET6 if (udp_done != 0) return; else udp_done = 1; #endif mibvar = "net.inet.udp.pcblist"; break; case IPPROTO_DIVERT: mibvar = "net.inet.divert.pcblist"; break; default: mibvar = "net.inet.raw.pcblist"; break; } len = 0; if (sysctlbyname(mibvar, 0, &len, 0, 0) < 0) { if (errno != ENOENT) warn("sysctl: %s", mibvar); return; } if ((buf = malloc(len)) == 0) { warnx("malloc %lu bytes", (u_long)len); return; } if (sysctlbyname(mibvar, buf, &len, 0, 0) < 0) { warn("sysctl: %s", mibvar); free(buf); return; } oxig = xig = (struct xinpgen *)buf; for (xig = (struct xinpgen *)((char *)xig + xig->xig_len); xig->xig_len > sizeof(struct xinpgen); xig = (struct xinpgen *)((char *)xig + xig->xig_len)) { if (istcp) { tp = &((struct xtcpcb *)xig)->xt_tp; inp = &((struct xtcpcb *)xig)->xt_inp; so = &((struct xtcpcb *)xig)->xt_socket; } else { inp = &((struct xinpcb *)xig)->xi_inp; so = &((struct xinpcb *)xig)->xi_socket; } /* Ignore sockets for protocols other than the desired one. */ if (so->xso_protocol != (int)proto) continue; /* Ignore PCBs which were freed during copyout. */ if (inp->inp_gencnt > oxig->xig_gen) continue; if ((af1 == AF_INET && (inp->inp_vflag & INP_IPV4) == 0) #ifdef INET6 || (af1 == AF_INET6 && (inp->inp_vflag & INP_IPV6) == 0) #endif /* INET6 */ || (af1 == AF_UNSPEC && ((inp->inp_vflag & INP_IPV4) == 0 #ifdef INET6 && (inp->inp_vflag & INP_IPV6) == 0 #endif /* INET6 */ )) ) continue; if (!aflag && ( (istcp && tp->t_state == TCPS_LISTEN) || (af1 == AF_INET && inet_lnaof(inp->inp_laddr) == INADDR_ANY) #ifdef INET6 || (af1 == AF_INET6 && IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) #endif /* INET6 */ || (af1 == AF_UNSPEC && (((inp->inp_vflag & INP_IPV4) != 0 && inet_lnaof(inp->inp_laddr) == INADDR_ANY) #ifdef INET6 || ((inp->inp_vflag & INP_IPV6) != 0 && IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) #endif )) )) continue; if (first) { if (!Lflag) { printf("Active Internet connections"); if (aflag) printf(" (including servers)"); } else printf( "Current listen queue sizes (qlen/incqlen/maxqlen)"); putchar('\n'); if (Aflag) printf("%-8.8s ", "Socket"); if (Lflag) printf("%-5.5s %-14.14s %-22.22s\n", "Proto", "Listen", "Local Address"); else printf((Aflag && !Wflag) ? "%-5.5s %-6.6s %-6.6s %-18.18s %-18.18s %s\n" : "%-5.5s %-6.6s %-6.6s %-22.22s %-22.22s %s\n", "Proto", "Recv-Q", "Send-Q", "Local Address", "Foreign Address", "(state)"); first = 0; } if (Lflag && so->so_qlimit == 0) continue; if (Aflag) { if (istcp) printf("%8lx ", (u_long)inp->inp_ppcb); else printf("%8lx ", (u_long)so->so_pcb); } #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) vchar = ((inp->inp_vflag & INP_IPV4) != 0) ? "46" : "6 "; else #endif vchar = ((inp->inp_vflag & INP_IPV4) != 0) ? "4 " : " "; printf("%-3.3s%-2.2s ", name, vchar); if (Lflag) { char buf1[15]; snprintf(buf1, 15, "%d/%d/%d", so->so_qlen, so->so_incqlen, so->so_qlimit); printf("%-14.14s ", buf1); } else { printf("%6u %6u ", so->so_rcv.sb_cc, so->so_snd.sb_cc); } if (numeric_port) { if (inp->inp_vflag & INP_IPV4) { inetprint(&inp->inp_laddr, (int)inp->inp_lport, name, 1); if (!Lflag) inetprint(&inp->inp_faddr, (int)inp->inp_fport, name, 1); } #ifdef INET6 else if (inp->inp_vflag & INP_IPV6) { inet6print(&inp->in6p_laddr, (int)inp->inp_lport, name, 1); if (!Lflag) inet6print(&inp->in6p_faddr, (int)inp->inp_fport, name, 1); } /* else nothing printed now */ #endif /* INET6 */ } else if (inp->inp_flags & INP_ANONPORT) { if (inp->inp_vflag & INP_IPV4) { inetprint(&inp->inp_laddr, (int)inp->inp_lport, name, 1); if (!Lflag) inetprint(&inp->inp_faddr, (int)inp->inp_fport, name, 0); } #ifdef INET6 else if (inp->inp_vflag & INP_IPV6) { inet6print(&inp->in6p_laddr, (int)inp->inp_lport, name, 1); if (!Lflag) inet6print(&inp->in6p_faddr, (int)inp->inp_fport, name, 0); } /* else nothing printed now */ #endif /* INET6 */ } else { if (inp->inp_vflag & INP_IPV4) { inetprint(&inp->inp_laddr, (int)inp->inp_lport, name, 0); if (!Lflag) inetprint(&inp->inp_faddr, (int)inp->inp_fport, name, inp->inp_lport != inp->inp_fport); } #ifdef INET6 else if (inp->inp_vflag & INP_IPV6) { inet6print(&inp->in6p_laddr, (int)inp->inp_lport, name, 0); if (!Lflag) inet6print(&inp->in6p_faddr, (int)inp->inp_fport, name, inp->inp_lport != inp->inp_fport); } /* else nothing printed now */ #endif /* INET6 */ } if (istcp && !Lflag) { if (tp->t_state < 0 || tp->t_state >= TCP_NSTATES) printf("%d", tp->t_state); else { printf("%s", tcpstates[tp->t_state]); #if defined(TF_NEEDSYN) && defined(TF_NEEDFIN) /* Show T/TCP `hidden state' */ if (tp->t_flags & (TF_NEEDSYN|TF_NEEDFIN)) putchar('*'); #endif /* defined(TF_NEEDSYN) && defined(TF_NEEDFIN) */ } } putchar('\n'); } if (xig != oxig && xig->xig_gen != oxig->xig_gen) { if (oxig->xig_count > xig->xig_count) { printf("Some %s sockets may have been deleted.\n", name); } else if (oxig->xig_count < xig->xig_count) { printf("Some %s sockets may have been created.\n", name); } else { printf("Some %s sockets may have been created or deleted.\n", name); } } free(buf); } /* * Dump TCP statistics structure. */ void tcp_stats(u_long off __unused, const char *name, int af1 __unused) { struct tcpstat tcpstat, zerostat; size_t len = sizeof tcpstat; if (zflag) memset(&zerostat, 0, len); if (sysctlbyname("net.inet.tcp.stats", &tcpstat, &len, zflag ? &zerostat : NULL, zflag ? len : 0) < 0) { warn("sysctl: net.inet.tcp.stats"); return; } #ifdef INET6 if (tcp_done != 0) return; else tcp_done = 1; #endif printf ("%s:\n", name); #define p(f, m) if (tcpstat.f || sflag <= 1) \ printf(m, tcpstat.f, plural(tcpstat.f)) #define p1a(f, m) if (tcpstat.f || sflag <= 1) \ printf(m, tcpstat.f) #define p2(f1, f2, m) if (tcpstat.f1 || tcpstat.f2 || sflag <= 1) \ printf(m, tcpstat.f1, plural(tcpstat.f1), tcpstat.f2, plural(tcpstat.f2)) #define p2a(f1, f2, m) if (tcpstat.f1 || tcpstat.f2 || sflag <= 1) \ printf(m, tcpstat.f1, plural(tcpstat.f1), tcpstat.f2) #define p3(f, m) if (tcpstat.f || sflag <= 1) \ printf(m, tcpstat.f, plurales(tcpstat.f)) p(tcps_sndtotal, "\t%lu packet%s sent\n"); p2(tcps_sndpack,tcps_sndbyte, "\t\t%lu data packet%s (%lu byte%s)\n"); p2(tcps_sndrexmitpack, tcps_sndrexmitbyte, "\t\t%lu data packet%s (%lu byte%s) retransmitted\n"); p(tcps_sndrexmitbad, "\t\t%lu data packet%s unnecessarily retransmitted\n"); p(tcps_mturesent, "\t\t%lu resend%s initiated by MTU discovery\n"); p2a(tcps_sndacks, tcps_delack, "\t\t%lu ack-only packet%s (%lu delayed)\n"); p(tcps_sndurg, "\t\t%lu URG only packet%s\n"); p(tcps_sndprobe, "\t\t%lu window probe packet%s\n"); p(tcps_sndwinup, "\t\t%lu window update packet%s\n"); p(tcps_sndctrl, "\t\t%lu control packet%s\n"); p(tcps_rcvtotal, "\t%lu packet%s received\n"); p2(tcps_rcvackpack, tcps_rcvackbyte, "\t\t%lu ack%s (for %lu byte%s)\n"); p(tcps_rcvdupack, "\t\t%lu duplicate ack%s\n"); p(tcps_rcvacktoomuch, "\t\t%lu ack%s for unsent data\n"); p2(tcps_rcvpack, tcps_rcvbyte, "\t\t%lu packet%s (%lu byte%s) received in-sequence\n"); p2(tcps_rcvduppack, tcps_rcvdupbyte, "\t\t%lu completely duplicate packet%s (%lu byte%s)\n"); p(tcps_pawsdrop, "\t\t%lu old duplicate packet%s\n"); p2(tcps_rcvpartduppack, tcps_rcvpartdupbyte, "\t\t%lu packet%s with some dup. data (%lu byte%s duped)\n"); p2(tcps_rcvoopack, tcps_rcvoobyte, "\t\t%lu out-of-order packet%s (%lu byte%s)\n"); p2(tcps_rcvpackafterwin, tcps_rcvbyteafterwin, "\t\t%lu packet%s (%lu byte%s) of data after window\n"); p(tcps_rcvwinprobe, "\t\t%lu window probe%s\n"); p(tcps_rcvwinupd, "\t\t%lu window update packet%s\n"); p(tcps_rcvafterclose, "\t\t%lu packet%s received after close\n"); p(tcps_rcvbadsum, "\t\t%lu discarded for bad checksum%s\n"); p(tcps_rcvbadoff, "\t\t%lu discarded for bad header offset field%s\n"); p1a(tcps_rcvshort, "\t\t%lu discarded because packet too short\n"); p1a(tcps_rcvmemdrop, "\t\t%lu discarded due to memory problems\n"); p(tcps_connattempt, "\t%lu connection request%s\n"); p(tcps_accepts, "\t%lu connection accept%s\n"); p(tcps_badsyn, "\t%lu bad connection attempt%s\n"); p(tcps_listendrop, "\t%lu listen queue overflow%s\n"); p(tcps_badrst, "\t%lu ignored RSTs in the window%s\n"); p(tcps_connects, "\t%lu connection%s established (including accepts)\n"); p2(tcps_closed, tcps_drops, "\t%lu connection%s closed (including %lu drop%s)\n"); p(tcps_cachedrtt, "\t\t%lu connection%s updated cached RTT on close\n"); p(tcps_cachedrttvar, "\t\t%lu connection%s updated cached RTT variance on close\n"); p(tcps_cachedssthresh, "\t\t%lu connection%s updated cached ssthresh on close\n"); p(tcps_conndrops, "\t%lu embryonic connection%s dropped\n"); p2(tcps_rttupdated, tcps_segstimed, "\t%lu segment%s updated rtt (of %lu attempt%s)\n"); p(tcps_rexmttimeo, "\t%lu retransmit timeout%s\n"); p(tcps_timeoutdrop, "\t\t%lu connection%s dropped by rexmit timeout\n"); p(tcps_persisttimeo, "\t%lu persist timeout%s\n"); p(tcps_persistdrop, "\t\t%lu connection%s dropped by persist timeout\n"); + p(tcps_finwait2_drops, "\t%lu Connection%s (fin_wait_2) dropped because of timeout\n"); p(tcps_keeptimeo, "\t%lu keepalive timeout%s\n"); p(tcps_keepprobe, "\t\t%lu keepalive probe%s sent\n"); p(tcps_keepdrops, "\t\t%lu connection%s dropped by keepalive\n"); p(tcps_predack, "\t%lu correct ACK header prediction%s\n"); p(tcps_preddat, "\t%lu correct data packet header prediction%s\n"); p(tcps_sc_added, "\t%lu syncache entrie%s added\n"); p1a(tcps_sc_retransmitted, "\t\t%lu retransmitted\n"); p1a(tcps_sc_dupsyn, "\t\t%lu dupsyn\n"); p1a(tcps_sc_dropped, "\t\t%lu dropped\n"); p1a(tcps_sc_completed, "\t\t%lu completed\n"); p1a(tcps_sc_bucketoverflow, "\t\t%lu bucket overflow\n"); p1a(tcps_sc_cacheoverflow, "\t\t%lu cache overflow\n"); p1a(tcps_sc_reset, "\t\t%lu reset\n"); p1a(tcps_sc_stale, "\t\t%lu stale\n"); p1a(tcps_sc_aborted, "\t\t%lu aborted\n"); p1a(tcps_sc_badack, "\t\t%lu badack\n"); p1a(tcps_sc_unreach, "\t\t%lu unreach\n"); p(tcps_sc_zonefail, "\t\t%lu zone failure%s\n"); p(tcps_sc_sendcookie, "\t%lu cookie%s sent\n"); p(tcps_sc_recvcookie, "\t%lu cookie%s received\n"); p(tcps_sack_recovery_episode, "\t%lu SACK recovery episode%s\n"); p(tcps_sack_rexmits, "\t%lu segment rexmit%s in SACK recovery episodes\n"); p(tcps_sack_rexmit_bytes, "\t%lu byte rexmit%s in SACK recovery episodes\n"); p(tcps_sack_rcv_blocks, "\t%lu SACK option%s (SACK blocks) received\n"); p(tcps_sack_send_blocks, "\t%lu SACK option%s (SACK blocks) sent\n"); p1a(tcps_sack_sboverflow, "\t%lu SACK scoreboard overflow\n"); #undef p #undef p1a #undef p2 #undef p2a #undef p3 } /* * Dump UDP statistics structure. */ void udp_stats(u_long off __unused, const char *name, int af1 __unused) { struct udpstat udpstat, zerostat; size_t len = sizeof udpstat; u_long delivered; if (zflag) memset(&zerostat, 0, len); if (sysctlbyname("net.inet.udp.stats", &udpstat, &len, zflag ? &zerostat : NULL, zflag ? len : 0) < 0) { warn("sysctl: net.inet.udp.stats"); return; } #ifdef INET6 if (udp_done != 0) return; else udp_done = 1; #endif printf("%s:\n", name); #define p(f, m) if (udpstat.f || sflag <= 1) \ printf(m, udpstat.f, plural(udpstat.f)) #define p1a(f, m) if (udpstat.f || sflag <= 1) \ printf(m, udpstat.f) p(udps_ipackets, "\t%lu datagram%s received\n"); p1a(udps_hdrops, "\t%lu with incomplete header\n"); p1a(udps_badlen, "\t%lu with bad data length field\n"); p1a(udps_badsum, "\t%lu with bad checksum\n"); p1a(udps_nosum, "\t%lu with no checksum\n"); p1a(udps_noport, "\t%lu dropped due to no socket\n"); p(udps_noportbcast, "\t%lu broadcast/multicast datagram%s dropped due to no socket\n"); p1a(udps_fullsock, "\t%lu dropped due to full socket buffers\n"); p1a(udpps_pcbhashmiss, "\t%lu not for hashed pcb\n"); delivered = udpstat.udps_ipackets - udpstat.udps_hdrops - udpstat.udps_badlen - udpstat.udps_badsum - udpstat.udps_noport - udpstat.udps_noportbcast - udpstat.udps_fullsock; if (delivered || sflag <= 1) printf("\t%lu delivered\n", delivered); p(udps_opackets, "\t%lu datagram%s output\n"); #undef p #undef p1a } /* * Dump CARP statistics structure. */ void carp_stats(u_long off __unused, const char *name, int af1 __unused) { struct carpstats carpstat, zerostat; size_t len = sizeof(struct carpstats); if (zflag) memset(&zerostat, 0, len); if (sysctlbyname("net.inet.carp.stats", &carpstat, &len, zflag ? &zerostat : NULL, zflag ? len : 0) < 0) { if (errno != ENOENT) warn("sysctl: net.inet.carp.stats"); return; } printf("%s:\n", name); #define p(f, m) if (carpstat.f || sflag <= 1) \ printf(m, (uintmax_t)carpstat.f, plural(carpstat.f)) #define p2(f, m) if (carpstat.f || sflag <= 1) \ printf(m, (uintmax_t)carpstat.f) p(carps_ipackets, "\t%ju packet%s received (IPv4)\n"); p(carps_ipackets6, "\t%ju packet%s received (IPv6)\n"); p(carps_badttl, "\t\t%ju packet%s discarded for wrong TTL\n"); p(carps_hdrops, "\t\t%ju packet%s shorter than header\n"); p(carps_badsum, "\t\t%ju discarded for bad checksum%s\n"); p(carps_badver, "\t\t%ju discarded packet%s with a bad version\n"); p2(carps_badlen, "\t\t%ju discarded because packet too short\n"); p2(carps_badauth, "\t\t%ju discarded for bad authentication\n"); p2(carps_badvhid, "\t\t%ju discarded for bad vhid\n"); p2(carps_badaddrs, "\t\t%ju discarded because of a bad address list\n"); p(carps_opackets, "\t%ju packet%s sent (IPv4)\n"); p(carps_opackets6, "\t%ju packet%s sent (IPv6)\n"); p2(carps_onomem, "\t\t%ju send failed due to mbuf memory error\n"); #if notyet p(carps_ostates, "\t\t%s state update%s sent\n"); #endif #undef p #undef p2 } /* * Dump IP statistics structure. */ void ip_stats(u_long off __unused, const char *name, int af1 __unused) { struct ipstat ipstat, zerostat; size_t len = sizeof ipstat; if (zflag) memset(&zerostat, 0, len); if (sysctlbyname("net.inet.ip.stats", &ipstat, &len, zflag ? &zerostat : NULL, zflag ? len : 0) < 0) { warn("sysctl: net.inet.ip.stats"); return; } printf("%s:\n", name); #define p(f, m) if (ipstat.f || sflag <= 1) \ printf(m, ipstat.f, plural(ipstat.f)) #define p1a(f, m) if (ipstat.f || sflag <= 1) \ printf(m, ipstat.f) p(ips_total, "\t%lu total packet%s received\n"); p(ips_badsum, "\t%lu bad header checksum%s\n"); p1a(ips_toosmall, "\t%lu with size smaller than minimum\n"); p1a(ips_tooshort, "\t%lu with data size < data length\n"); p1a(ips_toolong, "\t%lu with ip length > max ip packet size\n"); p1a(ips_badhlen, "\t%lu with header length < data size\n"); p1a(ips_badlen, "\t%lu with data length < header length\n"); p1a(ips_badoptions, "\t%lu with bad options\n"); p1a(ips_badvers, "\t%lu with incorrect version number\n"); p(ips_fragments, "\t%lu fragment%s received\n"); p(ips_fragdropped, "\t%lu fragment%s dropped (dup or out of space)\n"); p(ips_fragtimeout, "\t%lu fragment%s dropped after timeout\n"); p(ips_reassembled, "\t%lu packet%s reassembled ok\n"); p(ips_delivered, "\t%lu packet%s for this host\n"); p(ips_noproto, "\t%lu packet%s for unknown/unsupported protocol\n"); p(ips_forward, "\t%lu packet%s forwarded"); p(ips_fastforward, " (%lu packet%s fast forwarded)"); if (ipstat.ips_forward || sflag <= 1) putchar('\n'); p(ips_cantforward, "\t%lu packet%s not forwardable\n"); p(ips_notmember, "\t%lu packet%s received for unknown multicast group\n"); p(ips_redirectsent, "\t%lu redirect%s sent\n"); p(ips_localout, "\t%lu packet%s sent from this host\n"); p(ips_rawout, "\t%lu packet%s sent with fabricated ip header\n"); p(ips_odropped, "\t%lu output packet%s dropped due to no bufs, etc.\n"); p(ips_noroute, "\t%lu output packet%s discarded due to no route\n"); p(ips_fragmented, "\t%lu output datagram%s fragmented\n"); p(ips_ofragments, "\t%lu fragment%s created\n"); p(ips_cantfrag, "\t%lu datagram%s that can't be fragmented\n"); p(ips_nogif, "\t%lu tunneling packet%s that can't find gif\n"); p(ips_badaddr, "\t%lu datagram%s with bad address in header\n"); #undef p #undef p1a } static const char *icmpnames[] = { "echo reply", "#1", "#2", "destination unreachable", "source quench", "routing redirect", "#6", "#7", "echo", "router advertisement", "router solicitation", "time exceeded", "parameter problem", "time stamp", "time stamp reply", "information request", "information request reply", "address mask request", "address mask reply", }; /* * Dump ICMP statistics. */ void icmp_stats(u_long off __unused, const char *name, int af1 __unused) { struct icmpstat icmpstat, zerostat; int i, first; int mib[4]; /* CTL_NET + PF_INET + IPPROTO_ICMP + req */ size_t len; mib[0] = CTL_NET; mib[1] = PF_INET; mib[2] = IPPROTO_ICMP; mib[3] = ICMPCTL_STATS; len = sizeof icmpstat; if (zflag) memset(&zerostat, 0, len); if (sysctl(mib, 4, &icmpstat, &len, zflag ? &zerostat : NULL, zflag ? len : 0) < 0) { warn("sysctl: net.inet.icmp.stats"); return; } printf("%s:\n", name); #define p(f, m) if (icmpstat.f || sflag <= 1) \ printf(m, icmpstat.f, plural(icmpstat.f)) #define p1a(f, m) if (icmpstat.f || sflag <= 1) \ printf(m, icmpstat.f) #define p2(f, m) if (icmpstat.f || sflag <= 1) \ printf(m, icmpstat.f, plurales(icmpstat.f)) p(icps_error, "\t%lu call%s to icmp_error\n"); p(icps_oldicmp, "\t%lu error%s not generated in response to an icmp message\n"); for (first = 1, i = 0; i < ICMP_MAXTYPE + 1; i++) if (icmpstat.icps_outhist[i] != 0) { if (first) { printf("\tOutput histogram:\n"); first = 0; } printf("\t\t%s: %lu\n", icmpnames[i], icmpstat.icps_outhist[i]); } p(icps_badcode, "\t%lu message%s with bad code fields\n"); p(icps_tooshort, "\t%lu message%s < minimum length\n"); p(icps_checksum, "\t%lu bad checksum%s\n"); p(icps_badlen, "\t%lu message%s with bad length\n"); p1a(icps_bmcastecho, "\t%lu multicast echo requests ignored\n"); p1a(icps_bmcasttstamp, "\t%lu multicast timestamp requests ignored\n"); for (first = 1, i = 0; i < ICMP_MAXTYPE + 1; i++) if (icmpstat.icps_inhist[i] != 0) { if (first) { printf("\tInput histogram:\n"); first = 0; } printf("\t\t%s: %lu\n", icmpnames[i], icmpstat.icps_inhist[i]); } p(icps_reflect, "\t%lu message response%s generated\n"); p2(icps_badaddr, "\t%lu invalid return address%s\n"); p(icps_noroute, "\t%lu no return route%s\n"); #undef p #undef p1a #undef p2 mib[3] = ICMPCTL_MASKREPL; len = sizeof i; if (sysctl(mib, 4, &i, &len, (void *)0, 0) < 0) return; printf("\tICMP address mask responses are %sabled\n", i ? "en" : "dis"); } /* * Dump IGMP statistics structure. */ void igmp_stats(u_long off __unused, const char *name, int af1 __unused) { struct igmpstat igmpstat, zerostat; size_t len = sizeof igmpstat; if (zflag) memset(&zerostat, 0, len); if (sysctlbyname("net.inet.igmp.stats", &igmpstat, &len, zflag ? &zerostat : NULL, zflag ? len : 0) < 0) { warn("sysctl: net.inet.igmp.stats"); return; } printf("%s:\n", name); #define p(f, m) if (igmpstat.f || sflag <= 1) \ printf(m, igmpstat.f, plural(igmpstat.f)) #define py(f, m) if (igmpstat.f || sflag <= 1) \ printf(m, igmpstat.f, igmpstat.f != 1 ? "ies" : "y") p(igps_rcv_total, "\t%u message%s received\n"); p(igps_rcv_tooshort, "\t%u message%s received with too few bytes\n"); p(igps_rcv_badsum, "\t%u message%s received with bad checksum\n"); py(igps_rcv_queries, "\t%u membership quer%s received\n"); py(igps_rcv_badqueries, "\t%u membership quer%s received with invalid field(s)\n"); p(igps_rcv_reports, "\t%u membership report%s received\n"); p(igps_rcv_badreports, "\t%u membership report%s received with invalid field(s)\n"); p(igps_rcv_ourreports, "\t%u membership report%s received for groups to which we belong\n"); p(igps_snd_reports, "\t%u membership report%s sent\n"); #undef p #undef py } /* * Dump PIM statistics structure. */ void pim_stats(u_long off __unused, const char *name, int af1 __unused) { struct pimstat pimstat, zerostat; size_t len = sizeof pimstat; if (zflag) memset(&zerostat, 0, len); if (sysctlbyname("net.inet.pim.stats", &pimstat, &len, zflag ? &zerostat : NULL, zflag ? len : 0) < 0) { if (errno != ENOENT) warn("sysctl: net.inet.pim.stats"); return; } printf("%s:\n", name); #define p(f, m) if (pimstat.f || sflag <= 1) \ printf(m, (uintmax_t)pimstat.f, plural(pimstat.f)) #define py(f, m) if (pimstat.f || sflag <= 1) \ printf(m, (uintmax_t)pimstat.f, pimstat.f != 1 ? "ies" : "y") p(pims_rcv_total_msgs, "\t%ju message%s received\n"); p(pims_rcv_total_bytes, "\t%ju byte%s received\n"); p(pims_rcv_tooshort, "\t%ju message%s received with too few bytes\n"); p(pims_rcv_badsum, "\t%ju message%s received with bad checksum\n"); p(pims_rcv_badversion, "\t%ju message%s received with bad version\n"); p(pims_rcv_registers_msgs, "\t%ju data register message%s received\n"); p(pims_rcv_registers_bytes, "\t%ju data register byte%s received\n"); p(pims_rcv_registers_wrongiif, "\t%ju data register message%s received on wrong iif\n"); p(pims_rcv_badregisters, "\t%ju bad register%s received\n"); p(pims_snd_registers_msgs, "\t%ju data register message%s sent\n"); p(pims_snd_registers_bytes, "\t%ju data register byte%s sent\n"); #undef p #undef py } /* * Pretty print an Internet address (net address + port). */ void inetprint(struct in_addr *in, int port, const char *proto, int num_port) { struct servent *sp = 0; char line[80], *cp; int width; if (Wflag) sprintf(line, "%s.", inetname(in)); else sprintf(line, "%.*s.", (Aflag && !num_port) ? 12 : 16, inetname(in)); cp = index(line, '\0'); if (!num_port && port) sp = getservbyport((int)port, proto); if (sp || port == 0) sprintf(cp, "%.15s ", sp ? sp->s_name : "*"); else sprintf(cp, "%d ", ntohs((u_short)port)); width = (Aflag && !Wflag) ? 18 : 22; if (Wflag) printf("%-*s ", width, line); else printf("%-*.*s ", width, width, line); } /* * Construct an Internet address representation. * If numeric_addr has been supplied, give * numeric value, otherwise try for symbolic name. */ char * inetname(struct in_addr *inp) { char *cp; static char line[MAXHOSTNAMELEN]; struct hostent *hp; struct netent *np; cp = 0; if (!numeric_addr && inp->s_addr != INADDR_ANY) { int net = inet_netof(*inp); int lna = inet_lnaof(*inp); if (lna == INADDR_ANY) { np = getnetbyaddr(net, AF_INET); if (np) cp = np->n_name; } if (cp == 0) { hp = gethostbyaddr((char *)inp, sizeof (*inp), AF_INET); if (hp) { cp = hp->h_name; trimdomain(cp, strlen(cp)); } } } if (inp->s_addr == INADDR_ANY) strcpy(line, "*"); else if (cp) { strncpy(line, cp, sizeof(line) - 1); line[sizeof(line) - 1] = '\0'; } else { inp->s_addr = ntohl(inp->s_addr); #define C(x) ((u_int)((x) & 0xff)) sprintf(line, "%u.%u.%u.%u", C(inp->s_addr >> 24), C(inp->s_addr >> 16), C(inp->s_addr >> 8), C(inp->s_addr)); } return (line); }