2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California.
6 * Copyright (c) 2006-2007 Robert N. M. Watson
7 * Copyright (c) 2010-2011 Juniper Networks, Inc.
10 * Portions of this software were developed by Robert N. M. Watson under
11 * contract to Juniper Networks, Inc.
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
45 #include "opt_inet6.h"
46 #include "opt_ipsec.h"
47 #include "opt_tcpdebug.h"
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/limits.h>
52 #include <sys/malloc.h>
53 #include <sys/refcount.h>
54 #include <sys/kernel.h>
55 #include <sys/sysctl.h>
58 #include <sys/domain.h>
60 #include <sys/socket.h>
61 #include <sys/socketvar.h>
62 #include <sys/protosw.h>
65 #include <sys/syslog.h>
72 #include <net/if_var.h>
73 #include <net/route.h>
76 #include <netinet/in.h>
77 #include <netinet/in_kdtrace.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/in_systm.h>
80 #include <netinet/in_var.h>
81 #include <netinet/ip_var.h>
83 #include <netinet/ip6.h>
84 #include <netinet6/in6_pcb.h>
85 #include <netinet6/ip6_var.h>
86 #include <netinet6/scope6_var.h>
89 #include <netinet/tcp_fastopen.h>
91 #include <netinet/tcp.h>
92 #include <netinet/tcp_fsm.h>
93 #include <netinet/tcp_seq.h>
94 #include <netinet/tcp_timer.h>
95 #include <netinet/tcp_var.h>
96 #include <netinet/tcpip.h>
97 #include <netinet/cc/cc.h>
99 #include <netinet/tcp_pcap.h>
102 #include <netinet/tcp_debug.h>
105 #include <netinet/tcp_offload.h>
107 #include <netipsec/ipsec_support.h>
110 * TCP protocol interface to socket abstraction.
112 static int tcp_attach(struct socket *);
114 static int tcp_connect(struct tcpcb *, struct sockaddr *,
118 static int tcp6_connect(struct tcpcb *, struct sockaddr *,
121 static void tcp_disconnect(struct tcpcb *);
122 static void tcp_usrclosed(struct tcpcb *);
123 static void tcp_fill_info(struct tcpcb *, struct tcp_info *);
126 #define TCPDEBUG0 int ostate = 0
127 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0
128 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
129 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
133 #define TCPDEBUG2(req)
137 * TCP attaches to socket via pru_attach(), reserving space,
138 * and an internet control block.
141 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
144 struct tcpcb *tp = NULL;
149 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
152 error = tcp_attach(so);
156 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
157 so->so_linger = TCP_LINGERTIME;
162 TCPDEBUG2(PRU_ATTACH);
163 TCP_PROBE2(debug__user, tp, PRU_ATTACH);
168 * tcp_detach is called when the socket layer loses its final reference
169 * to the socket, be it a file descriptor reference, a reference from TCP,
170 * etc. At this point, there is only one case in which we will keep around
171 * inpcb state: time wait.
173 * This function can probably be re-absorbed back into tcp_usr_detach() now
174 * that there is a single detach path.
177 tcp_detach(struct socket *so, struct inpcb *inp)
181 INP_INFO_LOCK_ASSERT(&V_tcbinfo);
182 INP_WLOCK_ASSERT(inp);
184 KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
185 KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
189 if (inp->inp_flags & INP_TIMEWAIT) {
191 * There are two cases to handle: one in which the time wait
192 * state is being discarded (INP_DROPPED), and one in which
193 * this connection will remain in timewait. In the former,
194 * it is time to discard all state (except tcptw, which has
195 * already been discarded by the timewait close code, which
196 * should be further up the call stack somewhere). In the
197 * latter case, we detach from the socket, but leave the pcb
198 * present until timewait ends.
200 * XXXRW: Would it be cleaner to free the tcptw here?
202 * Astute question indeed, from twtcp perspective there are
203 * three cases to consider:
205 * #1 tcp_detach is called at tcptw creation time by
206 * tcp_twstart, then do not discard the newly created tcptw
207 * and leave inpcb present until timewait ends
208 * #2 tcp_detach is called at timewait end (or reuse) by
209 * tcp_twclose, then the tcptw has already been discarded
210 * (or reused) and inpcb is freed here
211 * #3 tcp_detach is called() after timewait ends (or reuse)
212 * (e.g. by soclose), then tcptw has already been discarded
213 * (or reused) and inpcb is freed here
215 * In all three cases the tcptw should not be freed here.
217 if (inp->inp_flags & INP_DROPPED) {
219 if (__predict_true(tp == NULL)) {
223 * This case should not happen as in TIMEWAIT
224 * state the inp should not be destroyed before
225 * its tcptw. If INVARIANTS is defined, panic.
228 panic("%s: Panic before an inp double-free: "
229 "INP_TIMEWAIT && INP_DROPPED && tp != NULL"
232 log(LOG_ERR, "%s: Avoid an inp double-free: "
233 "INP_TIMEWAIT && INP_DROPPED && tp != NULL"
244 * If the connection is not in timewait, we consider two
245 * two conditions: one in which no further processing is
246 * necessary (dropped || embryonic), and one in which TCP is
247 * not yet done, but no longer requires the socket, so the
248 * pcb will persist for the time being.
250 * XXXRW: Does the second case still occur?
252 if (inp->inp_flags & INP_DROPPED ||
253 tp->t_state < TCPS_SYN_SENT) {
265 * pru_detach() detaches the TCP protocol from the socket.
266 * If the protocol state is non-embryonic, then can't
267 * do this directly: have to initiate a pru_disconnect(),
268 * which may finish later; embryonic TCB's can just
272 tcp_usr_detach(struct socket *so)
278 KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
279 if (!INP_INFO_WLOCKED(&V_tcbinfo)) {
280 INP_INFO_RLOCK(&V_tcbinfo);
284 KASSERT(inp->inp_socket != NULL,
285 ("tcp_usr_detach: inp_socket == NULL"));
288 INP_INFO_RUNLOCK(&V_tcbinfo);
293 * Give the socket an address.
296 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
300 struct tcpcb *tp = NULL;
301 struct sockaddr_in *sinp;
303 sinp = (struct sockaddr_in *)nam;
304 if (nam->sa_len != sizeof (*sinp))
307 * Must check for multicast addresses and disallow binding
310 if (sinp->sin_family == AF_INET &&
311 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
312 return (EAFNOSUPPORT);
316 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
318 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
324 INP_HASH_WLOCK(&V_tcbinfo);
325 error = in_pcbbind(inp, nam, td->td_ucred);
326 INP_HASH_WUNLOCK(&V_tcbinfo);
329 TCP_PROBE2(debug__user, tp, PRU_BIND);
338 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
342 struct tcpcb *tp = NULL;
343 struct sockaddr_in6 *sin6p;
345 sin6p = (struct sockaddr_in6 *)nam;
346 if (nam->sa_len != sizeof (*sin6p))
349 * Must check for multicast addresses and disallow binding
352 if (sin6p->sin6_family == AF_INET6 &&
353 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
354 return (EAFNOSUPPORT);
358 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
360 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
366 INP_HASH_WLOCK(&V_tcbinfo);
367 inp->inp_vflag &= ~INP_IPV4;
368 inp->inp_vflag |= INP_IPV6;
370 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
371 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
372 inp->inp_vflag |= INP_IPV4;
373 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
374 struct sockaddr_in sin;
376 in6_sin6_2_sin(&sin, sin6p);
377 inp->inp_vflag |= INP_IPV4;
378 inp->inp_vflag &= ~INP_IPV6;
379 error = in_pcbbind(inp, (struct sockaddr *)&sin,
381 INP_HASH_WUNLOCK(&V_tcbinfo);
386 error = in6_pcbbind(inp, nam, td->td_ucred);
387 INP_HASH_WUNLOCK(&V_tcbinfo);
390 TCP_PROBE2(debug__user, tp, PRU_BIND);
398 * Prepare to accept connections.
401 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
405 struct tcpcb *tp = NULL;
409 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
411 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
418 error = solisten_proto_check(so);
419 INP_HASH_WLOCK(&V_tcbinfo);
420 if (error == 0 && inp->inp_lport == 0)
421 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
422 INP_HASH_WUNLOCK(&V_tcbinfo);
424 tcp_state_change(tp, TCPS_LISTEN);
425 solisten_proto(so, backlog);
427 if ((so->so_options & SO_NO_OFFLOAD) == 0)
428 tcp_offload_listen_start(tp);
434 if (IS_FASTOPEN(tp->t_flags))
435 tp->t_tfo_pending = tcp_fastopen_alloc_counter();
438 TCPDEBUG2(PRU_LISTEN);
439 TCP_PROBE2(debug__user, tp, PRU_LISTEN);
447 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
451 struct tcpcb *tp = NULL;
455 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
457 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
464 error = solisten_proto_check(so);
465 INP_HASH_WLOCK(&V_tcbinfo);
466 if (error == 0 && inp->inp_lport == 0) {
467 inp->inp_vflag &= ~INP_IPV4;
468 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
469 inp->inp_vflag |= INP_IPV4;
470 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
472 INP_HASH_WUNLOCK(&V_tcbinfo);
474 tcp_state_change(tp, TCPS_LISTEN);
475 solisten_proto(so, backlog);
477 if ((so->so_options & SO_NO_OFFLOAD) == 0)
478 tcp_offload_listen_start(tp);
484 if (IS_FASTOPEN(tp->t_flags))
485 tp->t_tfo_pending = tcp_fastopen_alloc_counter();
488 TCPDEBUG2(PRU_LISTEN);
489 TCP_PROBE2(debug__user, tp, PRU_LISTEN);
497 * Initiate connection to peer.
498 * Create a template for use in transmissions on this connection.
499 * Enter SYN_SENT state, and mark socket as connecting.
500 * Start keep-alive timer, and seed output sequence space.
501 * Send initial segment on connection.
504 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
508 struct tcpcb *tp = NULL;
509 struct sockaddr_in *sinp;
511 sinp = (struct sockaddr_in *)nam;
512 if (nam->sa_len != sizeof (*sinp))
515 * Must disallow TCP ``connections'' to multicast addresses.
517 if (sinp->sin_family == AF_INET
518 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
519 return (EAFNOSUPPORT);
520 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
525 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
527 if (inp->inp_flags & INP_TIMEWAIT) {
531 if (inp->inp_flags & INP_DROPPED) {
532 error = ECONNREFUSED;
537 if ((error = tcp_connect(tp, nam, td)) != 0)
540 if (registered_toedevs > 0 &&
541 (so->so_options & SO_NO_OFFLOAD) == 0 &&
542 (error = tcp_offload_connect(so, nam)) == 0)
545 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
546 error = tp->t_fb->tfb_tcp_output(tp);
548 TCPDEBUG2(PRU_CONNECT);
549 TCP_PROBE2(debug__user, tp, PRU_CONNECT);
557 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
561 struct tcpcb *tp = NULL;
562 struct sockaddr_in6 *sin6p;
566 sin6p = (struct sockaddr_in6 *)nam;
567 if (nam->sa_len != sizeof (*sin6p))
570 * Must disallow TCP ``connections'' to multicast addresses.
572 if (sin6p->sin6_family == AF_INET6
573 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
574 return (EAFNOSUPPORT);
577 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
579 if (inp->inp_flags & INP_TIMEWAIT) {
583 if (inp->inp_flags & INP_DROPPED) {
584 error = ECONNREFUSED;
591 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
592 * therefore probably require the hash lock, which isn't held here.
593 * Is this a significant problem?
595 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
596 struct sockaddr_in sin;
598 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
602 if ((inp->inp_vflag & INP_IPV4) == 0) {
603 error = EAFNOSUPPORT;
607 in6_sin6_2_sin(&sin, sin6p);
608 inp->inp_vflag |= INP_IPV4;
609 inp->inp_vflag &= ~INP_IPV6;
610 if ((error = prison_remote_ip4(td->td_ucred,
611 &sin.sin_addr)) != 0)
613 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
616 if (registered_toedevs > 0 &&
617 (so->so_options & SO_NO_OFFLOAD) == 0 &&
618 (error = tcp_offload_connect(so, nam)) == 0)
621 error = tp->t_fb->tfb_tcp_output(tp);
624 if ((inp->inp_vflag & INP_IPV6) == 0) {
625 error = EAFNOSUPPORT;
630 inp->inp_vflag &= ~INP_IPV4;
631 inp->inp_vflag |= INP_IPV6;
632 inp->inp_inc.inc_flags |= INC_ISIPV6;
633 if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
635 if ((error = tcp6_connect(tp, nam, td)) != 0)
638 if (registered_toedevs > 0 &&
639 (so->so_options & SO_NO_OFFLOAD) == 0 &&
640 (error = tcp_offload_connect(so, nam)) == 0)
643 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
644 error = tp->t_fb->tfb_tcp_output(tp);
647 TCPDEBUG2(PRU_CONNECT);
648 TCP_PROBE2(debug__user, tp, PRU_CONNECT);
655 * Initiate disconnect from peer.
656 * If connection never passed embryonic stage, just drop;
657 * else if don't need to let data drain, then can just drop anyways,
658 * else have to begin TCP shutdown process: mark socket disconnecting,
659 * drain unread data, state switch to reflect user close, and
660 * send segment (e.g. FIN) to peer. Socket will be really disconnected
661 * when peer sends FIN and acks ours.
663 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
666 tcp_usr_disconnect(struct socket *so)
669 struct tcpcb *tp = NULL;
673 INP_INFO_RLOCK(&V_tcbinfo);
675 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
677 if (inp->inp_flags & INP_TIMEWAIT)
679 if (inp->inp_flags & INP_DROPPED) {
687 TCPDEBUG2(PRU_DISCONNECT);
688 TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
690 INP_INFO_RUNLOCK(&V_tcbinfo);
696 * Accept a connection. Essentially all the work is done at higher levels;
697 * just return the address of the peer, storing through addr.
700 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
703 struct inpcb *inp = NULL;
704 struct tcpcb *tp = NULL;
709 if (so->so_state & SS_ISDISCONNECTED)
710 return (ECONNABORTED);
713 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
715 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
716 error = ECONNABORTED;
723 * We inline in_getpeeraddr and COMMON_END here, so that we can
724 * copy the data of interest and defer the malloc until after we
727 port = inp->inp_fport;
728 addr = inp->inp_faddr;
731 TCPDEBUG2(PRU_ACCEPT);
732 TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
735 *nam = in_sockaddr(port, &addr);
742 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
744 struct inpcb *inp = NULL;
746 struct tcpcb *tp = NULL;
748 struct in6_addr addr6;
753 if (so->so_state & SS_ISDISCONNECTED)
754 return (ECONNABORTED);
757 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
758 INP_INFO_RLOCK(&V_tcbinfo);
760 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
761 error = ECONNABORTED;
768 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
769 * copy the data of interest and defer the malloc until after we
772 if (inp->inp_vflag & INP_IPV4) {
774 port = inp->inp_fport;
775 addr = inp->inp_faddr;
777 port = inp->inp_fport;
778 addr6 = inp->in6p_faddr;
782 TCPDEBUG2(PRU_ACCEPT);
783 TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
785 INP_INFO_RUNLOCK(&V_tcbinfo);
788 *nam = in6_v4mapsin6_sockaddr(port, &addr);
790 *nam = in6_sockaddr(port, &addr6);
797 * Mark the connection as being incapable of further output.
800 tcp_usr_shutdown(struct socket *so)
804 struct tcpcb *tp = NULL;
807 INP_INFO_RLOCK(&V_tcbinfo);
809 KASSERT(inp != NULL, ("inp == NULL"));
811 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
819 if (!(inp->inp_flags & INP_DROPPED))
820 error = tp->t_fb->tfb_tcp_output(tp);
823 TCPDEBUG2(PRU_SHUTDOWN);
824 TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
826 INP_INFO_RUNLOCK(&V_tcbinfo);
832 * After a receive, possibly send window update to peer.
835 tcp_usr_rcvd(struct socket *so, int flags)
838 struct tcpcb *tp = NULL;
843 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
845 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
853 * For passively-created TFO connections, don't attempt a window
854 * update while still in SYN_RECEIVED as this may trigger an early
855 * SYN|ACK. It is preferable to have the SYN|ACK be sent along with
856 * application response data, or failing that, when the DELACK timer
859 if (IS_FASTOPEN(tp->t_flags) &&
860 (tp->t_state == TCPS_SYN_RECEIVED))
864 if (tp->t_flags & TF_TOE)
865 tcp_offload_rcvd(tp);
868 tp->t_fb->tfb_tcp_output(tp);
872 TCP_PROBE2(debug__user, tp, PRU_RCVD);
878 * Do a send by putting data in output queue and updating urgent
879 * marker if URG set. Possibly send more data. Unlike the other
880 * pru_*() routines, the mbuf chains are our responsibility. We
881 * must either enqueue them or free them. The other pru_* routines
882 * generally are caller-frees.
885 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
886 struct sockaddr *nam, struct mbuf *control, struct thread *td)
890 struct tcpcb *tp = NULL;
897 * We require the pcbinfo lock if we will close the socket as part of
900 if (flags & PRUS_EOF)
901 INP_INFO_RLOCK(&V_tcbinfo);
903 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
905 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
909 * In case of PRUS_NOTREADY, tcp_usr_ready() is responsible
910 * for freeing memory.
912 if (m && (flags & PRUS_NOTREADY) == 0)
918 isipv6 = nam && nam->sa_family == AF_INET6;
923 /* TCP doesn't do control messages (rights, creds, etc) */
924 if (control->m_len) {
931 m_freem(control); /* empty control, just free it */
933 if (!(flags & PRUS_OOB)) {
934 sbappendstream(&so->so_snd, m, flags);
935 if (nam && tp->t_state < TCPS_SYN_SENT) {
937 * Do implied connect if not yet connected,
938 * initialize window to default value, and
939 * initialize maxseg using peer's cached MSS.
943 error = tcp6_connect(tp, nam, td);
945 #if defined(INET6) && defined(INET)
949 error = tcp_connect(tp, nam, td);
953 tp->snd_wnd = TTCP_CLIENT_SND_WND;
956 if (flags & PRUS_EOF) {
958 * Close the send side of the connection after
961 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
965 if (!(inp->inp_flags & INP_DROPPED) &&
966 !(flags & PRUS_NOTREADY)) {
967 if (flags & PRUS_MORETOCOME)
968 tp->t_flags |= TF_MORETOCOME;
969 error = tp->t_fb->tfb_tcp_output(tp);
970 if (flags & PRUS_MORETOCOME)
971 tp->t_flags &= ~TF_MORETOCOME;
975 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
977 SOCKBUF_LOCK(&so->so_snd);
978 if (sbspace(&so->so_snd) < -512) {
979 SOCKBUF_UNLOCK(&so->so_snd);
985 * According to RFC961 (Assigned Protocols),
986 * the urgent pointer points to the last octet
987 * of urgent data. We continue, however,
988 * to consider it to indicate the first octet
989 * of data past the urgent section.
990 * Otherwise, snd_up should be one lower.
992 sbappendstream_locked(&so->so_snd, m, flags);
993 SOCKBUF_UNLOCK(&so->so_snd);
994 if (nam && tp->t_state < TCPS_SYN_SENT) {
996 * Do implied connect if not yet connected,
997 * initialize window to default value, and
998 * initialize maxseg using peer's cached MSS.
1002 error = tcp6_connect(tp, nam, td);
1004 #if defined(INET6) && defined(INET)
1008 error = tcp_connect(tp, nam, td);
1012 tp->snd_wnd = TTCP_CLIENT_SND_WND;
1015 tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
1016 if (!(flags & PRUS_NOTREADY)) {
1017 tp->t_flags |= TF_FORCEDATA;
1018 error = tp->t_fb->tfb_tcp_output(tp);
1019 tp->t_flags &= ~TF_FORCEDATA;
1023 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
1024 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1025 TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
1026 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1028 if (flags & PRUS_EOF)
1029 INP_INFO_RUNLOCK(&V_tcbinfo);
1034 tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
1040 inp = sotoinpcb(so);
1042 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1044 for (int i = 0; i < count; i++)
1046 return (ECONNRESET);
1048 tp = intotcpcb(inp);
1050 SOCKBUF_LOCK(&so->so_snd);
1051 error = sbready(&so->so_snd, m, count);
1052 SOCKBUF_UNLOCK(&so->so_snd);
1054 error = tp->t_fb->tfb_tcp_output(tp);
1061 * Abort the TCP. Drop the connection abruptly.
1064 tcp_usr_abort(struct socket *so)
1067 struct tcpcb *tp = NULL;
1070 inp = sotoinpcb(so);
1071 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1073 INP_INFO_RLOCK(&V_tcbinfo);
1075 KASSERT(inp->inp_socket != NULL,
1076 ("tcp_usr_abort: inp_socket == NULL"));
1079 * If we still have full TCP state, and we're not dropped, drop.
1081 if (!(inp->inp_flags & INP_TIMEWAIT) &&
1082 !(inp->inp_flags & INP_DROPPED)) {
1083 tp = intotcpcb(inp);
1085 tcp_drop(tp, ECONNABORTED);
1086 TCPDEBUG2(PRU_ABORT);
1087 TCP_PROBE2(debug__user, tp, PRU_ABORT);
1089 if (!(inp->inp_flags & INP_DROPPED)) {
1091 so->so_state |= SS_PROTOREF;
1093 inp->inp_flags |= INP_SOCKREF;
1096 INP_INFO_RUNLOCK(&V_tcbinfo);
1100 * TCP socket is closed. Start friendly disconnect.
1103 tcp_usr_close(struct socket *so)
1106 struct tcpcb *tp = NULL;
1109 inp = sotoinpcb(so);
1110 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1112 INP_INFO_RLOCK(&V_tcbinfo);
1114 KASSERT(inp->inp_socket != NULL,
1115 ("tcp_usr_close: inp_socket == NULL"));
1118 * If we still have full TCP state, and we're not dropped, initiate
1121 if (!(inp->inp_flags & INP_TIMEWAIT) &&
1122 !(inp->inp_flags & INP_DROPPED)) {
1123 tp = intotcpcb(inp);
1126 TCPDEBUG2(PRU_CLOSE);
1127 TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1129 if (!(inp->inp_flags & INP_DROPPED)) {
1131 so->so_state |= SS_PROTOREF;
1133 inp->inp_flags |= INP_SOCKREF;
1136 INP_INFO_RUNLOCK(&V_tcbinfo);
1140 * Receive out-of-band data.
1143 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1147 struct tcpcb *tp = NULL;
1150 inp = sotoinpcb(so);
1151 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1153 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1157 tp = intotcpcb(inp);
1159 if ((so->so_oobmark == 0 &&
1160 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1161 so->so_options & SO_OOBINLINE ||
1162 tp->t_oobflags & TCPOOB_HADDATA) {
1166 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1167 error = EWOULDBLOCK;
1171 *mtod(m, caddr_t) = tp->t_iobc;
1172 if ((flags & MSG_PEEK) == 0)
1173 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1176 TCPDEBUG2(PRU_RCVOOB);
1177 TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1183 struct pr_usrreqs tcp_usrreqs = {
1184 .pru_abort = tcp_usr_abort,
1185 .pru_accept = tcp_usr_accept,
1186 .pru_attach = tcp_usr_attach,
1187 .pru_bind = tcp_usr_bind,
1188 .pru_connect = tcp_usr_connect,
1189 .pru_control = in_control,
1190 .pru_detach = tcp_usr_detach,
1191 .pru_disconnect = tcp_usr_disconnect,
1192 .pru_listen = tcp_usr_listen,
1193 .pru_peeraddr = in_getpeeraddr,
1194 .pru_rcvd = tcp_usr_rcvd,
1195 .pru_rcvoob = tcp_usr_rcvoob,
1196 .pru_send = tcp_usr_send,
1197 .pru_ready = tcp_usr_ready,
1198 .pru_shutdown = tcp_usr_shutdown,
1199 .pru_sockaddr = in_getsockaddr,
1200 .pru_sosetlabel = in_pcbsosetlabel,
1201 .pru_close = tcp_usr_close,
1206 struct pr_usrreqs tcp6_usrreqs = {
1207 .pru_abort = tcp_usr_abort,
1208 .pru_accept = tcp6_usr_accept,
1209 .pru_attach = tcp_usr_attach,
1210 .pru_bind = tcp6_usr_bind,
1211 .pru_connect = tcp6_usr_connect,
1212 .pru_control = in6_control,
1213 .pru_detach = tcp_usr_detach,
1214 .pru_disconnect = tcp_usr_disconnect,
1215 .pru_listen = tcp6_usr_listen,
1216 .pru_peeraddr = in6_mapped_peeraddr,
1217 .pru_rcvd = tcp_usr_rcvd,
1218 .pru_rcvoob = tcp_usr_rcvoob,
1219 .pru_send = tcp_usr_send,
1220 .pru_ready = tcp_usr_ready,
1221 .pru_shutdown = tcp_usr_shutdown,
1222 .pru_sockaddr = in6_mapped_sockaddr,
1223 .pru_sosetlabel = in_pcbsosetlabel,
1224 .pru_close = tcp_usr_close,
1230 * Common subroutine to open a TCP connection to remote host specified
1231 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
1232 * port number if needed. Call in_pcbconnect_setup to do the routing and
1233 * to choose a local host address (interface). If there is an existing
1234 * incarnation of the same connection in TIME-WAIT state and if the remote
1235 * host was sending CC options and if the connection duration was < MSL, then
1236 * truncate the previous TIME-WAIT state and proceed.
1237 * Initialize connection parameters and enter SYN-SENT state.
1240 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1242 struct inpcb *inp = tp->t_inpcb, *oinp;
1243 struct socket *so = inp->inp_socket;
1244 struct in_addr laddr;
1248 INP_WLOCK_ASSERT(inp);
1249 INP_HASH_WLOCK(&V_tcbinfo);
1251 if (inp->inp_lport == 0) {
1252 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1258 * Cannot simply call in_pcbconnect, because there might be an
1259 * earlier incarnation of this same connection still in
1260 * TIME_WAIT state, creating an ADDRINUSE error.
1262 laddr = inp->inp_laddr;
1263 lport = inp->inp_lport;
1264 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1265 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1266 if (error && oinp == NULL)
1272 inp->inp_laddr = laddr;
1274 INP_HASH_WUNLOCK(&V_tcbinfo);
1277 * Compute window scaling to request:
1278 * Scale to fit into sweet spot. See tcp_syncache.c.
1279 * XXX: This should move to tcp_output().
1281 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1282 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1283 tp->request_r_scale++;
1286 TCPSTAT_INC(tcps_connattempt);
1287 tcp_state_change(tp, TCPS_SYN_SENT);
1288 tp->iss = tcp_new_isn(tp);
1289 tcp_sendseqinit(tp);
1294 INP_HASH_WUNLOCK(&V_tcbinfo);
1301 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1303 struct inpcb *inp = tp->t_inpcb;
1306 INP_WLOCK_ASSERT(inp);
1307 INP_HASH_WLOCK(&V_tcbinfo);
1309 if (inp->inp_lport == 0) {
1310 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1314 error = in6_pcbconnect(inp, nam, td->td_ucred);
1317 INP_HASH_WUNLOCK(&V_tcbinfo);
1319 /* Compute window scaling to request. */
1320 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1321 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1322 tp->request_r_scale++;
1324 soisconnecting(inp->inp_socket);
1325 TCPSTAT_INC(tcps_connattempt);
1326 tcp_state_change(tp, TCPS_SYN_SENT);
1327 tp->iss = tcp_new_isn(tp);
1328 tcp_sendseqinit(tp);
1333 INP_HASH_WUNLOCK(&V_tcbinfo);
1339 * Export TCP internal state information via a struct tcp_info, based on the
1340 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently
1341 * (TCP state machine, etc). We export all information using FreeBSD-native
1342 * constants -- for example, the numeric values for tcpi_state will differ
1346 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1349 INP_WLOCK_ASSERT(tp->t_inpcb);
1350 bzero(ti, sizeof(*ti));
1352 ti->tcpi_state = tp->t_state;
1353 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1354 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1355 if (tp->t_flags & TF_SACK_PERMIT)
1356 ti->tcpi_options |= TCPI_OPT_SACK;
1357 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1358 ti->tcpi_options |= TCPI_OPT_WSCALE;
1359 ti->tcpi_snd_wscale = tp->snd_scale;
1360 ti->tcpi_rcv_wscale = tp->rcv_scale;
1362 if (tp->t_flags & TF_ECN_PERMIT)
1363 ti->tcpi_options |= TCPI_OPT_ECN;
1365 ti->tcpi_rto = tp->t_rxtcur * tick;
1366 ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick;
1367 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1368 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1370 ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1371 ti->tcpi_snd_cwnd = tp->snd_cwnd;
1374 * FreeBSD-specific extension fields for tcp_info.
1376 ti->tcpi_rcv_space = tp->rcv_wnd;
1377 ti->tcpi_rcv_nxt = tp->rcv_nxt;
1378 ti->tcpi_snd_wnd = tp->snd_wnd;
1379 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */
1380 ti->tcpi_snd_nxt = tp->snd_nxt;
1381 ti->tcpi_snd_mss = tp->t_maxseg;
1382 ti->tcpi_rcv_mss = tp->t_maxseg;
1383 if (tp->t_flags & TF_TOE)
1384 ti->tcpi_options |= TCPI_OPT_TOE;
1385 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1386 ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1387 ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1391 * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1392 * socket option arguments. When it re-acquires the lock after the copy, it
1393 * has to revalidate that the connection is still valid for the socket
1396 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do { \
1398 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
1401 return (ECONNRESET); \
1403 tp = intotcpcb(inp); \
1405 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */)
1408 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1413 struct tcp_function_block *blk;
1414 struct tcp_function_set fsn;
1417 inp = sotoinpcb(so);
1418 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1420 if (sopt->sopt_level != IPPROTO_TCP) {
1422 if (inp->inp_vflag & INP_IPV6PROTO) {
1424 error = ip6_ctloutput(so, sopt);
1427 #if defined(INET6) && defined(INET)
1433 error = ip_ctloutput(so, sopt);
1438 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1440 return (ECONNRESET);
1442 tp = intotcpcb(inp);
1444 * Protect the TCP option TCP_FUNCTION_BLK so
1445 * that a sub-function can *never* overwrite this.
1447 if ((sopt->sopt_dir == SOPT_SET) &&
1448 (sopt->sopt_name == TCP_FUNCTION_BLK)) {
1450 error = sooptcopyin(sopt, &fsn, sizeof fsn,
1454 INP_WLOCK_RECHECK(inp);
1455 blk = find_and_ref_tcp_functions(&fsn);
1460 if (tp->t_fb == blk) {
1461 /* You already have this */
1462 refcount_release(&blk->tfb_refcnt);
1466 if (tp->t_state != TCPS_CLOSED) {
1469 * The user has advanced the state
1470 * past the initial point, we may not
1471 * be able to switch.
1473 if (blk->tfb_tcp_handoff_ok != NULL) {
1475 * Does the stack provide a
1476 * query mechanism, if so it may
1477 * still be possible?
1479 error = (*blk->tfb_tcp_handoff_ok)(tp);
1482 refcount_release(&blk->tfb_refcnt);
1487 if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) {
1488 refcount_release(&blk->tfb_refcnt);
1493 * Release the old refcnt, the
1494 * lookup acquired a ref on the
1497 if (tp->t_fb->tfb_tcp_fb_fini) {
1499 * Tell the stack to cleanup with 0 i.e.
1500 * the tcb is not going away.
1502 (*tp->t_fb->tfb_tcp_fb_fini)(tp, 0);
1504 refcount_release(&tp->t_fb->tfb_refcnt);
1506 if (tp->t_fb->tfb_tcp_fb_init) {
1507 (*tp->t_fb->tfb_tcp_fb_init)(tp);
1510 if (tp->t_flags & TF_TOE) {
1511 tcp_offload_ctloutput(tp, sopt->sopt_dir,
1517 } else if ((sopt->sopt_dir == SOPT_GET) &&
1518 (sopt->sopt_name == TCP_FUNCTION_BLK)) {
1519 strcpy(fsn.function_set_name, tp->t_fb->tfb_tcp_block_name);
1520 fsn.pcbcnt = tp->t_fb->tfb_refcnt;
1522 error = sooptcopyout(sopt, &fsn, sizeof fsn);
1525 /* Pass in the INP locked, called must unlock it */
1526 return (tp->t_fb->tfb_tcp_ctloutput(so, sopt, inp, tp));
1530 tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp)
1532 int error, opt, optval;
1535 struct cc_algo *algo;
1536 char *pbuf, buf[TCP_CA_NAME_MAX];
1540 * For TCP_CCALGOOPT forward the control to CC module, for both
1541 * SOPT_SET and SOPT_GET.
1543 switch (sopt->sopt_name) {
1546 pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO);
1547 error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize,
1548 sopt->sopt_valsize);
1553 INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP));
1554 if (CC_ALGO(tp)->ctl_output != NULL)
1555 error = CC_ALGO(tp)->ctl_output(tp->ccv, sopt, pbuf);
1559 if (error == 0 && sopt->sopt_dir == SOPT_GET)
1560 error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize);
1565 switch (sopt->sopt_dir) {
1567 switch (sopt->sopt_name) {
1568 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
1570 if (!TCPMD5_ENABLED()) {
1572 return (ENOPROTOOPT);
1574 error = TCPMD5_PCBCTL(inp, sopt);
1577 goto unlock_and_done;
1583 error = sooptcopyin(sopt, &optval, sizeof optval,
1588 INP_WLOCK_RECHECK(inp);
1589 switch (sopt->sopt_name) {
1597 opt = 0; /* dead code to fool gcc */
1604 tp->t_flags &= ~opt;
1607 if (tp->t_flags & TF_TOE) {
1608 tcp_offload_ctloutput(tp, sopt->sopt_dir,
1617 error = sooptcopyin(sopt, &optval, sizeof optval,
1622 INP_WLOCK_RECHECK(inp);
1624 tp->t_flags |= TF_NOPUSH;
1625 else if (tp->t_flags & TF_NOPUSH) {
1626 tp->t_flags &= ~TF_NOPUSH;
1627 if (TCPS_HAVEESTABLISHED(tp->t_state))
1628 error = tp->t_fb->tfb_tcp_output(tp);
1630 goto unlock_and_done;
1634 error = sooptcopyin(sopt, &optval, sizeof optval,
1639 INP_WLOCK_RECHECK(inp);
1640 if (optval > 0 && optval <= tp->t_maxseg &&
1641 optval + 40 >= V_tcp_minmss)
1642 tp->t_maxseg = optval;
1645 goto unlock_and_done;
1652 case TCP_CONGESTION:
1654 error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1);
1657 buf[sopt->sopt_valsize] = '\0';
1658 INP_WLOCK_RECHECK(inp);
1660 STAILQ_FOREACH(algo, &cc_list, entries)
1661 if (strncmp(buf, algo->name,
1662 TCP_CA_NAME_MAX) == 0)
1671 * We hold a write lock over the tcb so it's safe to
1672 * do these things without ordering concerns.
1674 if (CC_ALGO(tp)->cb_destroy != NULL)
1675 CC_ALGO(tp)->cb_destroy(tp->ccv);
1678 * If something goes pear shaped initialising the new
1679 * algo, fall back to newreno (which does not
1680 * require initialisation).
1682 if (algo->cb_init != NULL &&
1683 algo->cb_init(tp->ccv) != 0) {
1684 CC_ALGO(tp) = &newreno_cc_algo;
1686 * The only reason init should fail is
1687 * because of malloc.
1698 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1702 if (ui > (UINT_MAX / hz)) {
1708 INP_WLOCK_RECHECK(inp);
1709 switch (sopt->sopt_name) {
1711 tp->t_keepidle = ui;
1713 * XXX: better check current remaining
1714 * timeout and "merge" it with new value.
1716 if ((tp->t_state > TCPS_LISTEN) &&
1717 (tp->t_state <= TCPS_CLOSING))
1718 tcp_timer_activate(tp, TT_KEEP,
1722 tp->t_keepintvl = ui;
1723 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1724 (TP_MAXIDLE(tp) > 0))
1725 tcp_timer_activate(tp, TT_2MSL,
1729 tp->t_keepinit = ui;
1730 if (tp->t_state == TCPS_SYN_RECEIVED ||
1731 tp->t_state == TCPS_SYN_SENT)
1732 tcp_timer_activate(tp, TT_KEEP,
1736 goto unlock_and_done;
1740 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1744 INP_WLOCK_RECHECK(inp);
1746 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1747 (TP_MAXIDLE(tp) > 0))
1748 tcp_timer_activate(tp, TT_2MSL,
1750 goto unlock_and_done;
1756 error = sooptcopyin(sopt, &optval, sizeof optval,
1761 INP_WLOCK_RECHECK(inp);
1763 tcp_pcap_set_sock_max(TCP_PCAP_OUT ?
1764 &(tp->t_outpkts) : &(tp->t_inpkts),
1768 goto unlock_and_done;
1774 if (!V_tcp_fastopen_enabled)
1777 error = sooptcopyin(sopt, &optval, sizeof optval,
1782 INP_WLOCK_RECHECK(inp);
1784 tp->t_flags |= TF_FASTOPEN;
1785 if ((tp->t_state == TCPS_LISTEN) &&
1786 (tp->t_tfo_pending == NULL))
1788 tcp_fastopen_alloc_counter();
1790 tp->t_flags &= ~TF_FASTOPEN;
1791 goto unlock_and_done;
1796 error = ENOPROTOOPT;
1802 tp = intotcpcb(inp);
1803 switch (sopt->sopt_name) {
1804 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
1806 if (!TCPMD5_ENABLED()) {
1808 return (ENOPROTOOPT);
1810 error = TCPMD5_PCBCTL(inp, sopt);
1815 optval = tp->t_flags & TF_NODELAY;
1817 error = sooptcopyout(sopt, &optval, sizeof optval);
1820 optval = tp->t_maxseg;
1822 error = sooptcopyout(sopt, &optval, sizeof optval);
1825 optval = tp->t_flags & TF_NOOPT;
1827 error = sooptcopyout(sopt, &optval, sizeof optval);
1830 optval = tp->t_flags & TF_NOPUSH;
1832 error = sooptcopyout(sopt, &optval, sizeof optval);
1835 tcp_fill_info(tp, &ti);
1837 error = sooptcopyout(sopt, &ti, sizeof ti);
1839 case TCP_CONGESTION:
1840 len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1842 error = sooptcopyout(sopt, buf, len + 1);
1848 switch (sopt->sopt_name) {
1850 ui = TP_KEEPIDLE(tp) / hz;
1853 ui = TP_KEEPINTVL(tp) / hz;
1856 ui = TP_KEEPINIT(tp) / hz;
1859 ui = TP_KEEPCNT(tp);
1863 error = sooptcopyout(sopt, &ui, sizeof(ui));
1868 optval = tcp_pcap_get_sock_max(TCP_PCAP_OUT ?
1869 &(tp->t_outpkts) : &(tp->t_inpkts));
1871 error = sooptcopyout(sopt, &optval, sizeof optval);
1877 optval = tp->t_flags & TF_FASTOPEN;
1879 error = sooptcopyout(sopt, &optval, sizeof optval);
1884 error = ENOPROTOOPT;
1891 #undef INP_WLOCK_RECHECK
1892 #undef INP_WLOCK_RECHECK_CLEANUP
1895 * Attach TCP protocol to socket, allocating
1896 * internet protocol control block, tcp control block,
1897 * bufer space, and entering LISTEN state if to accept connections.
1900 tcp_attach(struct socket *so)
1906 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1907 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
1911 so->so_rcv.sb_flags |= SB_AUTOSIZE;
1912 so->so_snd.sb_flags |= SB_AUTOSIZE;
1913 INP_INFO_RLOCK(&V_tcbinfo);
1914 error = in_pcballoc(so, &V_tcbinfo);
1916 INP_INFO_RUNLOCK(&V_tcbinfo);
1919 inp = sotoinpcb(so);
1921 if (inp->inp_vflag & INP_IPV6PROTO) {
1922 inp->inp_vflag |= INP_IPV6;
1923 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
1924 inp->inp_vflag |= INP_IPV4;
1925 inp->in6p_hops = -1; /* use kernel default */
1929 inp->inp_vflag |= INP_IPV4;
1930 tp = tcp_newtcpcb(inp);
1934 INP_INFO_RUNLOCK(&V_tcbinfo);
1937 tp->t_state = TCPS_CLOSED;
1939 INP_INFO_RUNLOCK(&V_tcbinfo);
1940 TCPSTATES_INC(TCPS_CLOSED);
1945 * Initiate (or continue) disconnect.
1946 * If embryonic state, just send reset (once).
1947 * If in ``let data drain'' option and linger null, just drop.
1948 * Otherwise (hard), mark socket disconnecting and drop
1949 * current input data; switch states based on user close, and
1950 * send segment to peer (with FIN).
1953 tcp_disconnect(struct tcpcb *tp)
1955 struct inpcb *inp = tp->t_inpcb;
1956 struct socket *so = inp->inp_socket;
1958 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
1959 INP_WLOCK_ASSERT(inp);
1962 * Neither tcp_close() nor tcp_drop() should return NULL, as the
1963 * socket is still open.
1965 if (tp->t_state < TCPS_ESTABLISHED) {
1968 ("tcp_disconnect: tcp_close() returned NULL"));
1969 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1970 tp = tcp_drop(tp, 0);
1972 ("tcp_disconnect: tcp_drop() returned NULL"));
1974 soisdisconnecting(so);
1975 sbflush(&so->so_rcv);
1977 if (!(inp->inp_flags & INP_DROPPED))
1978 tp->t_fb->tfb_tcp_output(tp);
1983 * User issued close, and wish to trail through shutdown states:
1984 * if never received SYN, just forget it. If got a SYN from peer,
1985 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1986 * If already got a FIN from peer, then almost done; go to LAST_ACK
1987 * state. In all other cases, have already sent FIN to peer (e.g.
1988 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1989 * for peer to send FIN or not respond to keep-alives, etc.
1990 * We can let the user exit from the close as soon as the FIN is acked.
1993 tcp_usrclosed(struct tcpcb *tp)
1996 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
1997 INP_WLOCK_ASSERT(tp->t_inpcb);
1999 switch (tp->t_state) {
2002 tcp_offload_listen_stop(tp);
2004 tcp_state_change(tp, TCPS_CLOSED);
2009 * tcp_close() should never return NULL here as the socket is
2013 ("tcp_usrclosed: tcp_close() returned NULL"));
2017 case TCPS_SYN_RECEIVED:
2018 tp->t_flags |= TF_NEEDFIN;
2021 case TCPS_ESTABLISHED:
2022 tcp_state_change(tp, TCPS_FIN_WAIT_1);
2025 case TCPS_CLOSE_WAIT:
2026 tcp_state_change(tp, TCPS_LAST_ACK);
2029 if (tp->t_state >= TCPS_FIN_WAIT_2) {
2030 soisdisconnected(tp->t_inpcb->inp_socket);
2031 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
2032 if (tp->t_state == TCPS_FIN_WAIT_2) {
2035 timeout = (tcp_fast_finwait2_recycle) ?
2036 tcp_finwait2_timeout : TP_MAXIDLE(tp);
2037 tcp_timer_activate(tp, TT_2MSL, timeout);
2044 db_print_indent(int indent)
2048 for (i = 0; i < indent; i++)
2053 db_print_tstate(int t_state)
2058 db_printf("TCPS_CLOSED");
2062 db_printf("TCPS_LISTEN");
2066 db_printf("TCPS_SYN_SENT");
2069 case TCPS_SYN_RECEIVED:
2070 db_printf("TCPS_SYN_RECEIVED");
2073 case TCPS_ESTABLISHED:
2074 db_printf("TCPS_ESTABLISHED");
2077 case TCPS_CLOSE_WAIT:
2078 db_printf("TCPS_CLOSE_WAIT");
2081 case TCPS_FIN_WAIT_1:
2082 db_printf("TCPS_FIN_WAIT_1");
2086 db_printf("TCPS_CLOSING");
2090 db_printf("TCPS_LAST_ACK");
2093 case TCPS_FIN_WAIT_2:
2094 db_printf("TCPS_FIN_WAIT_2");
2097 case TCPS_TIME_WAIT:
2098 db_printf("TCPS_TIME_WAIT");
2102 db_printf("unknown");
2108 db_print_tflags(u_int t_flags)
2113 if (t_flags & TF_ACKNOW) {
2114 db_printf("%sTF_ACKNOW", comma ? ", " : "");
2117 if (t_flags & TF_DELACK) {
2118 db_printf("%sTF_DELACK", comma ? ", " : "");
2121 if (t_flags & TF_NODELAY) {
2122 db_printf("%sTF_NODELAY", comma ? ", " : "");
2125 if (t_flags & TF_NOOPT) {
2126 db_printf("%sTF_NOOPT", comma ? ", " : "");
2129 if (t_flags & TF_SENTFIN) {
2130 db_printf("%sTF_SENTFIN", comma ? ", " : "");
2133 if (t_flags & TF_REQ_SCALE) {
2134 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
2137 if (t_flags & TF_RCVD_SCALE) {
2138 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
2141 if (t_flags & TF_REQ_TSTMP) {
2142 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
2145 if (t_flags & TF_RCVD_TSTMP) {
2146 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
2149 if (t_flags & TF_SACK_PERMIT) {
2150 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
2153 if (t_flags & TF_NEEDSYN) {
2154 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
2157 if (t_flags & TF_NEEDFIN) {
2158 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
2161 if (t_flags & TF_NOPUSH) {
2162 db_printf("%sTF_NOPUSH", comma ? ", " : "");
2165 if (t_flags & TF_MORETOCOME) {
2166 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
2169 if (t_flags & TF_LQ_OVERFLOW) {
2170 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
2173 if (t_flags & TF_LASTIDLE) {
2174 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
2177 if (t_flags & TF_RXWIN0SENT) {
2178 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
2181 if (t_flags & TF_FASTRECOVERY) {
2182 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
2185 if (t_flags & TF_CONGRECOVERY) {
2186 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
2189 if (t_flags & TF_WASFRECOVERY) {
2190 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
2193 if (t_flags & TF_SIGNATURE) {
2194 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
2197 if (t_flags & TF_FORCEDATA) {
2198 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
2201 if (t_flags & TF_TSO) {
2202 db_printf("%sTF_TSO", comma ? ", " : "");
2205 if (t_flags & TF_ECN_PERMIT) {
2206 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
2209 if (t_flags & TF_FASTOPEN) {
2210 db_printf("%sTF_FASTOPEN", comma ? ", " : "");
2216 db_print_toobflags(char t_oobflags)
2221 if (t_oobflags & TCPOOB_HAVEDATA) {
2222 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
2225 if (t_oobflags & TCPOOB_HADDATA) {
2226 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
2232 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
2235 db_print_indent(indent);
2236 db_printf("%s at %p\n", name, tp);
2240 db_print_indent(indent);
2241 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n",
2242 LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
2244 db_print_indent(indent);
2245 db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n",
2246 &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
2248 db_print_indent(indent);
2249 db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl,
2250 &tp->t_timers->tt_delack, tp->t_inpcb);
2252 db_print_indent(indent);
2253 db_printf("t_state: %d (", tp->t_state);
2254 db_print_tstate(tp->t_state);
2257 db_print_indent(indent);
2258 db_printf("t_flags: 0x%x (", tp->t_flags);
2259 db_print_tflags(tp->t_flags);
2262 db_print_indent(indent);
2263 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n",
2264 tp->snd_una, tp->snd_max, tp->snd_nxt);
2266 db_print_indent(indent);
2267 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n",
2268 tp->snd_up, tp->snd_wl1, tp->snd_wl2);
2270 db_print_indent(indent);
2271 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n",
2272 tp->iss, tp->irs, tp->rcv_nxt);
2274 db_print_indent(indent);
2275 db_printf("rcv_adv: 0x%08x rcv_wnd: %u rcv_up: 0x%08x\n",
2276 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
2278 db_print_indent(indent);
2279 db_printf("snd_wnd: %u snd_cwnd: %u\n",
2280 tp->snd_wnd, tp->snd_cwnd);
2282 db_print_indent(indent);
2283 db_printf("snd_ssthresh: %u snd_recover: "
2284 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
2286 db_print_indent(indent);
2287 db_printf("t_rcvtime: %u t_startime: %u\n",
2288 tp->t_rcvtime, tp->t_starttime);
2290 db_print_indent(indent);
2291 db_printf("t_rttime: %u t_rtsq: 0x%08x\n",
2292 tp->t_rtttime, tp->t_rtseq);
2294 db_print_indent(indent);
2295 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n",
2296 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
2298 db_print_indent(indent);
2299 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u "
2300 "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
2303 db_print_indent(indent);
2304 db_printf("t_rttupdated: %lu max_sndwnd: %u t_softerror: %d\n",
2305 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
2307 db_print_indent(indent);
2308 db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
2309 db_print_toobflags(tp->t_oobflags);
2310 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc);
2312 db_print_indent(indent);
2313 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n",
2314 tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
2316 db_print_indent(indent);
2317 db_printf("ts_recent: %u ts_recent_age: %u\n",
2318 tp->ts_recent, tp->ts_recent_age);
2320 db_print_indent(indent);
2321 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: "
2322 "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
2324 db_print_indent(indent);
2325 db_printf("snd_ssthresh_prev: %u snd_recover_prev: 0x%08x "
2326 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2327 tp->snd_recover_prev, tp->t_badrxtwin);
2329 db_print_indent(indent);
2330 db_printf("snd_numholes: %d snd_holes first: %p\n",
2331 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2333 db_print_indent(indent);
2334 db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: "
2335 "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
2337 /* Skip sackblks, sackhint. */
2339 db_print_indent(indent);
2340 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n",
2341 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2344 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2349 db_printf("usage: show tcpcb <addr>\n");
2352 tp = (struct tcpcb *)addr;
2354 db_print_tcpcb(tp, "tcpcb", 0);