/*
* Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
* 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. Neither the name of the project 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 PROJECT 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 PROJECT 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.
*/
/*
* Copyright (c) 1997, 1998, 2000, 2001, 2008 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation
* Facility, NASA Ames Research Center.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. 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 FOUNDATION 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.
*/
/*
* 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.
* 3. 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
*/
/*
* Create template to be used to send tcp packets on a connection.
* Call after host entry created, allocates an mbuf and fills
* in a skeletal tcp/ip header, minimizing the amount of work
* necessary when the connection is used.
*/
struct mbuf *
tcp_template(struct tcpcb *tp)
{
struct inpcb *inp = tp->t_inpcb;
struct tcphdr *n;
struct mbuf *m;
int hlen;
switch (tp->t_family) {
case AF_INET:
hlen = sizeof(struct ip);
if (inp->inp_af == AF_INET)
break;
#ifdef INET6
if (inp->inp_af == AF_INET6) {
/* mapped addr case */
if (IN6_IS_ADDR_V4MAPPED(&in6p_laddr(inp))
&& IN6_IS_ADDR_V4MAPPED(&in6p_faddr(inp)))
break;
}
#endif
return NULL; /*EINVAL*/
#ifdef INET6
case AF_INET6:
hlen = sizeof(struct ip6_hdr);
if (inp != NULL) {
/* more sainty check? */
break;
}
return NULL; /*EINVAL*/
#endif
default:
return NULL; /*EAFNOSUPPORT*/
}
/*
* Compute the pseudo-header portion of the checksum
* now. We incrementally add in the TCP option and
* payload lengths later, and then compute the TCP
* checksum right before the packet is sent off onto
* the wire.
*/
n->th_sum = in_cksum_phdr(ipov->ih_src.s_addr,
ipov->ih_dst.s_addr,
htons(sizeof(struct tcphdr) + IPPROTO_TCP));
break;
}
#ifdef INET6
case AF_INET6:
{
struct ip6_hdr *ip6;
mtod(m, struct ip *)->ip_v = 6;
ip6 = mtod(m, struct ip6_hdr *);
ip6->ip6_nxt = IPPROTO_TCP;
ip6->ip6_plen = htons(sizeof(struct tcphdr));
ip6->ip6_src = in6p_laddr(inp);
ip6->ip6_dst = in6p_faddr(inp);
ip6->ip6_flow = in6p_flowinfo(inp) & IPV6_FLOWINFO_MASK;
if (ip6_auto_flowlabel) {
ip6->ip6_flow &= ~IPV6_FLOWLABEL_MASK;
ip6->ip6_flow |=
(htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
}
ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
ip6->ip6_vfc |= IPV6_VERSION;
/*
* Compute the pseudo-header portion of the checksum
* now. We incrementally add in the TCP option and
* payload lengths later, and then compute the TCP
* checksum right before the packet is sent off onto
* the wire.
*/
n->th_sum = in6_cksum_phdr(&in6p_laddr(inp),
&in6p_faddr(inp), htonl(sizeof(struct tcphdr)),
htonl(IPPROTO_TCP));
break;
}
#endif
}
/*
* Send a single message to the TCP at address specified by
* the given TCP/IP header. If m == 0, 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 tp->t_template. 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.
*/
int
tcp_respond(struct tcpcb *tp, struct mbuf *mtemplate, struct mbuf *m,
struct tcphdr *th0, tcp_seq ack, tcp_seq seq, int flags)
{
struct route *ro;
int error, tlen, win = 0;
int hlen;
struct ip *ip;
#ifdef INET6
struct ip6_hdr *ip6;
#endif
int family; /* family on packet, not inpcb! */
struct tcphdr *th;
th = NULL; /* Quell uninitialized warning */
ip = NULL;
#ifdef INET6
ip6 = NULL;
#endif
if (m == NULL) {
if (!mtemplate)
return EINVAL;
/* get family information from template */
switch (mtod(mtemplate, struct ip *)->ip_v) {
case 4:
family = AF_INET;
hlen = sizeof(struct ip);
break;
#ifdef INET6
case 6:
family = AF_INET6;
hlen = sizeof(struct ip6_hdr);
break;
#endif
default:
return EAFNOSUPPORT;
}
MGETHDR(m, M_DONTWAIT, MT_HEADER);
if (m) {
MCLAIM(m, &tcp_tx_mowner);
MCLGET(m, M_DONTWAIT);
if ((m->m_flags & M_EXT) == 0) {
m_free(m);
m = NULL;
}
}
if (m == NULL)
return ENOBUFS;
/*
* Template TCPCB. Rather than zeroing a new TCPCB and initializing
* a bunch of members individually, we maintain this template for the
* static and mostly-static components of the TCPCB, and copy it into
* the new TCPCB instead.
*/
static struct tcpcb tcpcb_template = {
.t_srtt = TCPTV_SRTTBASE,
.t_rttmin = TCPTV_MIN,
/*
* Updates the TCPCB template whenever a parameter that would affect
* the template is changed.
*/
void
tcp_tcpcb_template(void)
{
struct tcpcb *tp = &tcpcb_template;
int flags;
flags = 0;
if (tcp_do_rfc1323 && tcp_do_win_scale)
flags |= TF_REQ_SCALE;
if (tcp_do_rfc1323 && tcp_do_timestamps)
flags |= TF_REQ_TSTMP;
tp->t_flags = flags;
/*
* Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no
* rtt estimate. Set rttvar so that srtt + 2 * rttvar gives
* reasonable initial retransmit time.
*/
tp->t_rttvar = tcp_rttdflt * PR_SLOWHZ << (TCP_RTTVAR_SHIFT + 2 - 1);
TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
TCPTV_MIN, TCPTV_REXMTMAX);
/*
* Create a new TCP control block, making an
* empty reassembly queue and hooking it to the argument
* protocol control block.
*/
struct tcpcb *
tcp_newtcpcb(int family, struct inpcb *inp)
{
struct tcpcb *tp;
int i;
/* XXX Consider using a pool_cache for speed. */
tp = pool_get(&tcpcb_pool, PR_NOWAIT); /* splsoftnet via tcp_usrreq */
if (tp == NULL)
return NULL;
memcpy(tp, &tcpcb_template, sizeof(*tp));
TAILQ_INIT(&tp->segq);
TAILQ_INIT(&tp->timeq);
tp->t_family = family; /* may be overridden later on */
TAILQ_INIT(&tp->snd_holes);
LIST_INIT(&tp->t_sc); /* XXX can template this */
/* Don't sweat this loop; hopefully the compiler will unroll it. */
for (i = 0; i < TCPT_NTIMERS; i++) {
callout_init(&tp->t_timer[i], CALLOUT_MPSAFE);
TCP_TIMER_INIT(tp, i);
}
callout_init(&tp->t_delack_ch, CALLOUT_MPSAFE);
tp->t_inpcb = inp;
/* for IPv6, always try to run path MTU discovery */
tp->t_mtudisc = 1;
break;
#endif /* INET6 */
default:
for (i = 0; i < TCPT_NTIMERS; i++)
callout_destroy(&tp->t_timer[i]);
callout_destroy(&tp->t_delack_ch);
pool_put(&tcpcb_pool, tp); /* splsoftnet via tcp_usrreq */
return NULL;
}
/*
* Initialize our timebase. When we send timestamps, we take
* the delta from tcp_now -- this means each connection always
* gets a timebase of 1, which makes it, among other things,
* more difficult to determine how long a system has been up,
* and thus how many TCP sequence increments have occurred.
*
* We start with 1, because 0 doesn't work with linux, which
* considers timestamp 0 in a SYN packet as a bug and disables
* timestamps.
*/
tp->ts_timebase = tcp_now - 1;
tcp_congctl_select(tp, tcp_congctl_global_name);
return tp;
}
/*
* 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;
KASSERT(tp->t_inpcb != NULL);
so = tp->t_inpcb->inp_socket;
if (so == NULL)
return NULL;
/*
* Close a TCP control block:
* discard all space held by the tcp
* discard internet protocol block
* wake up any sleepers
*/
struct tcpcb *
tcp_close(struct tcpcb *tp)
{
struct inpcb *inp;
struct socket *so;
#ifdef RTV_RTT
struct rtentry *rt = NULL;
#endif
struct route *ro;
int j;
inp = tp->t_inpcb;
so = inp->inp_socket;
ro = &inp->inp_route;
#ifdef RTV_RTT
/*
* If we sent enough data to get some meaningful characteristics,
* save them in the routing entry. 'Enough' is arbitrarily
* defined as the sendpipesize (default 4K) * 16. This would
* give us 16 rtt samples assuming we only get one sample per
* window (the usual case on a long haul net). 16 samples is
* enough for the srtt filter to converge to within 5% of the correct
* value; fewer samples and we could save a very bogus rtt.
*
* Don't update the default route's characteristics and don't
* update anything that the user "locked".
*/
if (SEQ_LT(tp->iss + so->so_snd.sb_hiwat * 16, tp->snd_max) &&
ro && (rt = rtcache_validate(ro)) != NULL &&
!in_nullhost(satocsin(rt_getkey(rt))->sin_addr)) {
u_long i = 0;
if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) {
i = tp->t_srtt *
((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTT_SHIFT + 2));
if (rt->rt_rmx.rmx_rtt && i)
/*
* filter this update to half the old & half
* the new values, converting scale.
* See route.h and tcp_var.h for a
* description of the scaling constants.
*/
rt->rt_rmx.rmx_rtt =
(rt->rt_rmx.rmx_rtt + i) / 2;
else
rt->rt_rmx.rmx_rtt = i;
}
if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) {
i = tp->t_rttvar *
((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTTVAR_SHIFT + 2));
if (rt->rt_rmx.rmx_rttvar && i)
rt->rt_rmx.rmx_rttvar =
(rt->rt_rmx.rmx_rttvar + i) / 2;
else
rt->rt_rmx.rmx_rttvar = i;
}
/*
* update the pipelimit (ssthresh) if it has been updated
* already or if a pipesize was specified & the threshold
* got below half the pipesize. I.e., wait for bad news
* before we start updating, then update on both good
* and bad news.
*/
if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 &&
(i = tp->snd_ssthresh) && rt->rt_rmx.rmx_ssthresh) ||
i < (rt->rt_rmx.rmx_sendpipe / 2)) {
/*
* convert the limit from user data bytes to
* packets then to packet data bytes.
*/
i = (i + tp->t_segsz / 2) / tp->t_segsz;
if (i < 2)
i = 2;
i *= (u_long)(tp->t_segsz + sizeof (struct tcpiphdr));
if (rt->rt_rmx.rmx_ssthresh)
rt->rt_rmx.rmx_ssthresh =
(rt->rt_rmx.rmx_ssthresh + i) / 2;
else
rt->rt_rmx.rmx_ssthresh = i;
}
}
rtcache_unref(rt, ro);
#endif /* RTV_RTT */
/* free the reassembly queue, if any */
TCP_REASS_LOCK(tp);
(void) tcp_freeq(tp);
TCP_REASS_UNLOCK(tp);
/* free the SACK holes list. */
tcp_free_sackholes(tp);
tcp_congctl_release(tp);
syn_cache_cleanup(tp);
if (tp->t_template) {
m_free(tp->t_template);
tp->t_template = NULL;
}
/*
* Detaching the pcb will unlock the socket/tcpcb, and stopping
* the timers can also drop the lock. We need to prevent access
* to the tcpcb as it's half torn down. Flag the pcb as dead
* (prevents access by timers) and only then detach it.
*/
tp->t_flags |= TF_DEAD;
inp->inp_ppcb = NULL;
soisdisconnected(so);
inpcb_destroy(inp);
/*
* pcb is no longer visble elsewhere, so we can safely release
* the lock in callout_halt() if needed.
*/
TCP_STATINC(TCP_STAT_CLOSED);
for (j = 0; j < TCPT_NTIMERS; j++) {
callout_halt(&tp->t_timer[j], softnet_lock);
callout_destroy(&tp->t_timer[j]);
}
callout_halt(&tp->t_delack_ch, softnet_lock);
callout_destroy(&tp->t_delack_ch);
pool_put(&tcpcb_pool, tp);
return NULL;
}
int
tcp_freeq(struct tcpcb *tp)
{
struct ipqent *qe;
int rv = 0;
/*
* Protocol drain routine. Called when memory is in short supply.
* Called from pr_fasttimo thus a callout context.
*/
void
tcp_drain(void)
{
struct inpcb *inp;
struct tcpcb *tp;
mutex_enter(softnet_lock);
KERNEL_LOCK(1, NULL);
/*
* Free the sequence queue of all TCP connections.
*/
TAILQ_FOREACH(inp, &tcbtable.inpt_queue, inp_queue) {
tp = intotcpcb(inp);
if (tp != NULL) {
/*
* If the tcpcb is already busy,
* just bail out now.
*/
if (tcp_reass_lock_try(tp) == 0)
continue;
if (tcp_freeq(tp))
TCP_STATINC(TCP_STAT_CONNSDRAINED);
TCP_REASS_UNLOCK(tp);
}
}
/*
* 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).
*/
void
tcp_notify(struct inpcb *inp, int error)
{
struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb;
struct socket *so = inp->inp_socket;
/*
* 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;
} else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 &&
tp->t_rxtshift > 3 && tp->t_softerror)
so->so_error = error;
else
tp->t_softerror = error;
cv_broadcast(&so->so_cv);
sorwakeup(so);
sowwakeup(so);
}
if (sa->sa_family != AF_INET6 ||
sa->sa_len != sizeof(struct sockaddr_in6))
return NULL;
if ((unsigned)cmd >= PRC_NCMDS)
return NULL;
else if (cmd == PRC_QUENCH) {
/*
* Don't honor ICMP Source Quench messages meant for
* TCP connections.
*/
return NULL;
} else if (PRC_IS_REDIRECT(cmd))
notify = in6pcb_rtchange, d = NULL;
else if (cmd == PRC_MSGSIZE)
; /* special code is present, see below */
else if (cmd == PRC_HOSTDEAD)
d = NULL;
else if (inet6ctlerrmap[cmd] == 0)
return NULL;
/* if the parameter is from icmp6, decode it. */
if (d != NULL) {
struct ip6ctlparam *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;
sa6_src = &sa6_any;
off = 0;
}
if (ip6) {
/* check if we can safely examine src and dst ports */
if (m->m_pkthdr.len < off + sizeof(th)) {
if (cmd == PRC_MSGSIZE)
icmp6_mtudisc_update((struct ip6ctlparam *)d, 0);
return NULL;
}
/*
* Check to see if we have a valid TCP connection
* corresponding to the address in the ICMPv6 message
* payload.
*/
if (in6pcb_lookup(&tcbtable, &sa6->sin6_addr,
th.th_dport,
(const struct in6_addr *)&sa6_src->sin6_addr,
th.th_sport, 0, 0))
valid++;
/*
* Depending on the value of "valid" and routing table
* size (mtudisc_{hi,lo}wat), we will:
* - recalculate the new MTU and create the
* corresponding routing entry, or
* - ignore the MTU change notification.
*/
icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
/*
* no need to call in6pcb_notify, it should have been
* called via callback if necessary
*/
return NULL;
}
/* assumes that ip header and tcp header are contiguous on mbuf */
void *
tcp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
{
struct ip *ip = v;
struct tcphdr *th;
struct icmp *icp;
extern const int inetctlerrmap[];
void (*notify)(struct inpcb *, int) = tcp_notify;
int errno;
int nmatch;
struct tcpcb *tp;
u_int mtu;
tcp_seq seq;
struct inpcb *inp;
#ifdef INET6
struct in6_addr src6, dst6;
#endif
if (sa->sa_family != AF_INET ||
sa->sa_len != sizeof(struct sockaddr_in))
return NULL;
if ((unsigned)cmd >= PRC_NCMDS)
return NULL;
errno = inetctlerrmap[cmd];
if (cmd == PRC_QUENCH)
/*
* Don't honor ICMP Source Quench messages meant for
* TCP connections.
*/
return NULL;
else if (PRC_IS_REDIRECT(cmd))
notify = inpcb_rtchange, ip = 0;
else if (cmd == PRC_MSGSIZE && ip && ip->ip_v == 4) {
/*
* Check to see if we have a valid TCP connection
* corresponding to the address in the ICMP message
* payload.
*
* Boundary check is made in icmp_input(), with ICMP_ADVLENMIN.
*/
th = (struct tcphdr *)((char *)ip + (ip->ip_hl << 2));
#ifdef INET6
in6_in_2_v4mapin6(&ip->ip_src, &src6);
in6_in_2_v4mapin6(&ip->ip_dst, &dst6);
#endif
if ((inp = inpcb_lookup(&tcbtable, ip->ip_dst,
th->th_dport, ip->ip_src, th->th_sport, 0)) != NULL)
;
#ifdef INET6
else if ((inp = in6pcb_lookup(&tcbtable, &dst6,
th->th_dport, &src6, th->th_sport, 0, 0)) != NULL)
;
#endif
else
return NULL;
/*
* Now that we've validated that we are actually communicating
* with the host indicated in the ICMP message, locate the
* ICMP header, recalculate the new MTU, and create the
* corresponding routing entry.
*/
icp = (struct icmp *)((char *)ip -
offsetof(struct icmp, icmp_ip));
tp = intotcpcb(inp);
if (tp == NULL)
return NULL;
seq = ntohl(th->th_seq);
if (SEQ_LT(seq, tp->snd_una) || SEQ_GT(seq, tp->snd_max))
return NULL;
/*
* If the ICMP message advertises a Next-Hop MTU
* equal or larger than the maximum packet size we have
* ever sent, drop the message.
*/
mtu = (u_int)ntohs(icp->icmp_nextmtu);
if (mtu >= tp->t_pmtud_mtu_sent)
return NULL;
if (mtu >= tcp_hdrsz(tp) + tp->t_pmtud_mss_acked) {
/*
* Calculate new MTU, and create corresponding
* route (traditional PMTUD).
*/
tp->t_flags &= ~TF_PMTUD_PEND;
icmp_mtudisc(icp, ip->ip_dst);
} else {
/*
* Record the information got in the ICMP
* message; act on it later.
* If we had already recorded an ICMP message,
* replace the old one only if the new message
* refers to an older TCP segment
*/
if (tp->t_flags & TF_PMTUD_PEND) {
if (SEQ_LT(tp->t_pmtud_th_seq, seq))
return NULL;
} else
tp->t_flags |= TF_PMTUD_PEND;
tp->t_pmtud_th_seq = seq;
tp->t_pmtud_nextmtu = icp->icmp_nextmtu;
tp->t_pmtud_ip_len = icp->icmp_ip.ip_len;
tp->t_pmtud_ip_hl = icp->icmp_ip.ip_hl;
}
return NULL;
} else if (cmd == PRC_HOSTDEAD)
ip = 0;
else if (errno == 0)
return NULL;
if (ip && ip->ip_v == 4 && sa->sa_family == AF_INET) {
th = (struct tcphdr *)((char *)ip + (ip->ip_hl << 2));
nmatch = inpcb_notify(&tcbtable, satocsin(sa)->sin_addr,
th->th_dport, ip->ip_src, th->th_sport, errno, notify);
if (nmatch == 0 && syn_cache_count &&
(inetctlerrmap[cmd] == EHOSTUNREACH ||
inetctlerrmap[cmd] == ENETUNREACH ||
inetctlerrmap[cmd] == EHOSTDOWN)) {
struct sockaddr_in sin;
memset(&sin, 0, sizeof(sin));
sin.sin_len = sizeof(sin);
sin.sin_family = AF_INET;
sin.sin_port = th->th_sport;
sin.sin_addr = ip->ip_src;
syn_cache_unreach((struct sockaddr *)&sin, sa, th);
}
/*
* When a source quench is received, we are being notified of congestion.
* Close the congestion window down to the Loss Window (one segment).
* We will gradually open it again as we proceed.
*/
void
tcp_quench(struct inpcb *inp)
{
struct tcpcb *tp = intotcpcb(inp);
/*
* On receipt of path MTU corrections, flush old route and replace it
* with the new one. Retransmit all unacknowledged packets, to ensure
* that all packets will be received.
*/
void
tcp_mtudisc(struct inpcb *inp, int errno)
{
struct tcpcb *tp = intotcpcb(inp);
struct rtentry *rt;
if (tp == NULL)
return;
rt = inpcb_rtentry(inp);
if (rt != NULL) {
/*
* If this was not a host route, remove and realloc.
*/
if ((rt->rt_flags & RTF_HOST) == 0) {
inpcb_rtentry_unref(rt, inp);
inpcb_rtchange(inp, errno);
if ((rt = inpcb_rtentry(inp)) == NULL)
return;
}
/*
* Slow start out of the error condition. We
* use the MTU because we know it's smaller
* than the previously transmitted segment.
*
* Note: This is more conservative than the
* suggestion in draft-floyd-incr-init-win-03.
*/
if (rt->rt_rmx.rmx_mtu != 0)
tp->snd_cwnd =
TCP_INITIAL_WINDOW(tcp_init_win,
rt->rt_rmx.rmx_mtu);
inpcb_rtentry_unref(rt, inp);
}
rt = in6pcb_rtentry(inp);
if (rt != NULL) {
/*
* If this was not a host route, remove and realloc.
*/
if ((rt->rt_flags & RTF_HOST) == 0) {
in6pcb_rtentry_unref(rt, inp);
in6pcb_rtchange(inp, errno);
rt = in6pcb_rtentry(inp);
if (rt == NULL)
return;
}
/*
* Slow start out of the error condition. We
* use the MTU because we know it's smaller
* than the previously transmitted segment.
*
* Note: This is more conservative than the
* suggestion in draft-floyd-incr-init-win-03.
*/
if (rt->rt_rmx.rmx_mtu != 0) {
tp->snd_cwnd = TCP_INITIAL_WINDOW(tcp_init_win,
rt->rt_rmx.rmx_mtu);
}
in6pcb_rtentry_unref(rt, inp);
}
/*
* Compute the MSS to advertise to the peer. Called only during
* the 3-way handshake. If we are the server (peer initiated
* connection), we are called with a pointer to the interface
* on which the SYN packet arrived. If we are the client (we
* initiated connection), we are called with a pointer to the
* interface out which this connection should go.
*
* NOTE: Do not subtract IP option/extension header size nor IPsec
* header size from MSS advertisement. MSS option must hold the maximum
* segment size we can accept, so it must always be:
* max(if mtu) - ip header - tcp header
*/
u_long
tcp_mss_to_advertise(const struct ifnet *ifp, int af)
{
extern u_long in_maxmtu;
u_long mss = 0;
u_long hdrsiz;
/*
* In order to avoid defeating path MTU discovery on the peer,
* we advertise the max MTU of all attached networks as our MSS,
* per RFC 1191, section 3.1.
*
* We provide the option to advertise just the MTU of
* the interface on which we hope this connection will
* be receiving. If we are responding to a SYN, we
* will have a pretty good idea about this, but when
* initiating a connection there is a bit more doubt.
*
* We also need to ensure that loopback has a large enough
* MSS, as the loopback MTU is never included in in_maxmtu.
*/
if (ifp != NULL)
switch (af) {
#ifdef INET6
case AF_INET6: /* FALLTHROUGH */
#endif
case AF_INET:
mss = ifp->if_mtu;
break;
}
if (tcp_mss_ifmtu == 0)
switch (af) {
#ifdef INET6
case AF_INET6: /* FALLTHROUGH */
#endif
case AF_INET:
mss = uimax(in_maxmtu, mss);
break;
}
/*
* Set connection variables based on the peer's advertised MSS.
* We are passed the TCPCB for the actual connection. If we
* are the server, we are called by the compressed state engine
* when the 3-way handshake is complete. If we are the client,
* we are called when we receive the SYN,ACK from the server.
*
* NOTE: Our advertised MSS value must be initialized in the TCPCB
* before this routine is called!
*/
void
tcp_mss_from_peer(struct tcpcb *tp, int offer)
{
struct socket *so;
#if defined(RTV_SPIPE) || defined(RTV_SSTHRESH)
struct rtentry *rt;
#endif
u_long bufsize;
int mss;
/*
* As per RFC1122, use the default MSS value, unless they
* sent us an offer. Do not accept offers less than 256 bytes.
*/
mss = tcp_mssdflt;
if (offer)
mss = offer;
mss = uimax(mss, 256); /* sanity */
tp->t_peermss = mss;
mss -= tcp_optlen(tp);
if (tp->t_inpcb->inp_af == AF_INET)
mss -= ip_optlen(tp->t_inpcb);
#ifdef INET6
if (tp->t_inpcb->inp_af == AF_INET6)
mss -= ip6_optlen(tp->t_inpcb);
#endif
/*
* XXX XXX What if mss goes negative or zero? This can happen if a
* socket has large IPv6 options. We crash below.
*/
/*
* If there's a pipesize, change the socket buffer to that size.
* Make the socket buffer an integral number of MSS units. If
* the MSS is larger than the socket buffer, artificially decrease
* the MSS.
*/
#ifdef RTV_SPIPE
if (rt != NULL && rt->rt_rmx.rmx_sendpipe != 0)
bufsize = rt->rt_rmx.rmx_sendpipe;
else
#endif
{
KASSERT(so != NULL);
bufsize = so->so_snd.sb_hiwat;
}
if (bufsize < mss)
mss = bufsize;
else {
bufsize = roundup(bufsize, mss);
if (bufsize > sb_max)
bufsize = sb_max;
(void) sbreserve(&so->so_snd, bufsize, so);
}
tp->t_segsz = mss;
#ifdef RTV_SSTHRESH
if (rt != NULL && rt->rt_rmx.rmx_ssthresh) {
/*
* There's some sort of gateway or interface buffer
* limit on the path. Use this to set the slow
* start threshold, but set the threshold to no less
* than 2 * MSS.
*/
tp->snd_ssthresh = uimax(2 * mss, rt->rt_rmx.rmx_ssthresh);
}
#endif
#if defined(RTV_SPIPE) || defined(RTV_SSTHRESH)
inpcb_rtentry_unref(rt, tp->t_inpcb);
#endif
}
/*
* Processing necessary when a TCP connection is established.
*/
void
tcp_established(struct tcpcb *tp)
{
struct socket *so;
#ifdef RTV_RPIPE
struct rtentry *rt;
#endif
u_long bufsize;
KASSERT(tp->t_inpcb != NULL);
so = NULL;
rt = NULL;
/* This is a while() to reduce the dreadful stairstepping below */
while (tp->t_inpcb->inp_af == AF_INET) {
so = tp->t_inpcb->inp_socket;
#if defined(RTV_RPIPE)
rt = inpcb_rtentry(tp->t_inpcb);
#endif
if (__predict_true(tcp_msl_enable)) {
if (in4p_laddr(tp->t_inpcb).s_addr == INADDR_LOOPBACK) {
tp->t_msl = tcp_msl_loop ? tcp_msl_loop : (TCPTV_MSL >> 2);
break;
}
if (__predict_false(tcp_rttlocal)) {
/* This may be adjusted by tcp_input */
tp->t_msl = tcp_msl_local ? tcp_msl_local : (TCPTV_MSL >> 1);
break;
}
if (in_localaddr(in4p_faddr(tp->t_inpcb))) {
tp->t_msl = tcp_msl_local ? tcp_msl_local : (TCPTV_MSL >> 1);
break;
}
}
tp->t_msl = tcp_msl_remote ? tcp_msl_remote : TCPTV_MSL;
break;
}
/* Clamp to a reasonable range. */
tp->t_msl = MIN(tp->t_msl, TCP_MAXMSL);
#ifdef INET6
while (tp->t_inpcb->inp_af == AF_INET6) {
so = tp->t_inpcb->inp_socket;
#if defined(RTV_RPIPE)
rt = in6pcb_rtentry(tp->t_inpcb);
#endif
if (__predict_true(tcp_msl_enable)) {
extern const struct in6_addr in6addr_loopback;
/*
* Check if there's an initial rtt or rttvar. Convert from the
* route-table units to scaled multiples of the slow timeout timer.
* Called only during the 3-way handshake.
*/
void
tcp_rmx_rtt(struct tcpcb *tp)
{
#ifdef RTV_RTT
struct rtentry *rt = NULL;
int rtt;
KASSERT(tp->t_inpcb != NULL);
rt = inpcb_rtentry(tp->t_inpcb);
if (rt == NULL)
return;
if (tp->t_srtt == 0 && (rtt = rt->rt_rmx.rmx_rtt)) {
/*
* XXX The lock bit for MTU indicates that the value
* is also a minimum value; this is subject to time.
*/
if (rt->rt_rmx.rmx_locks & RTV_RTT)
TCPT_RANGESET(tp->t_rttmin,
rtt / (RTM_RTTUNIT / PR_SLOWHZ),
TCPTV_MIN, TCPTV_REXMTMAX);
tp->t_srtt = rtt /
((RTM_RTTUNIT / PR_SLOWHZ) >> (TCP_RTT_SHIFT + 2));
if (rt->rt_rmx.rmx_rttvar) {
tp->t_rttvar = rt->rt_rmx.rmx_rttvar /
((RTM_RTTUNIT / PR_SLOWHZ) >>
(TCP_RTTVAR_SHIFT + 2));
} else {
/* Default variation is +- 1 rtt */
tp->t_rttvar =
tp->t_srtt >> (TCP_RTT_SHIFT - TCP_RTTVAR_SHIFT);
}
TCPT_RANGESET(tp->t_rxtcur,
((tp->t_srtt >> 2) + tp->t_rttvar) >> (1 + 2),
tp->t_rttmin, TCPTV_REXMTMAX);
}
inpcb_rtentry_unref(rt, tp->t_inpcb);
#endif
}
tcp_seq tcp_iss_seq = 0; /* tcp initial seq # */
/*
* Get a new sequence value given a tcp control block
*/
tcp_seq
tcp_new_iss(struct tcpcb *tp)
{
if (!tp || !tp->t_template || !(inp = tp->t_inpcb))
return 0;
switch (tp->t_family) {
case AF_INET6:
/* XXX: should use correct direction. */
hdrsiz = ipsec_hdrsiz(tp->t_template, IPSEC_DIR_OUTBOUND, inp);
break;
case AF_INET:
/* mapped address case - tricky */
default:
hdrsiz = 0;
break;
}
return hdrsiz;
}
#endif
#endif /*IPSEC*/
/*
* Determine the length of the TCP options for this connection.
*
* XXX: What do we do for SACK, when we add that? Just reserve
* all of the space? Otherwise we can't exactly be incrementing
* cwnd by an amount that varies depending on the amount we last
* had to SACK!
*/
