2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California.
4 * Copyright (c) 2006-2007 Robert N. M. Watson
5 * Copyright (c) 2010-2011 Juniper Networks, Inc.
8 * Portions of this software were developed by Robert N. M. Watson under
9 * contract to Juniper Networks, Inc.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
43 #include "opt_inet6.h"
44 #include "opt_tcpdebug.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/limits.h>
49 #include <sys/malloc.h>
50 #include <sys/kernel.h>
51 #include <sys/sysctl.h>
54 #include <sys/domain.h>
56 #include <sys/socket.h>
57 #include <sys/socketvar.h>
58 #include <sys/protosw.h>
67 #include <net/if_var.h>
68 #include <net/route.h>
71 #include <netinet/cc.h>
72 #include <netinet/in.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/in_var.h>
76 #include <netinet/ip_var.h>
78 #include <netinet/ip6.h>
79 #include <netinet6/in6_pcb.h>
80 #include <netinet6/ip6_var.h>
81 #include <netinet6/scope6_var.h>
83 #include <netinet/tcp_fsm.h>
84 #include <netinet/tcp_seq.h>
85 #include <netinet/tcp_timer.h>
86 #include <netinet/tcp_var.h>
87 #include <netinet/tcpip.h>
89 #include <netinet/tcp_debug.h>
92 #include <netinet/tcp_offload.h>
96 * TCP protocol interface to socket abstraction.
98 static int tcp_attach(struct socket *);
100 static int tcp_connect(struct tcpcb *, struct sockaddr *,
104 static int tcp6_connect(struct tcpcb *, struct sockaddr *,
107 static void tcp_disconnect(struct tcpcb *);
108 static void tcp_usrclosed(struct tcpcb *);
109 static void tcp_fill_info(struct tcpcb *, struct tcp_info *);
112 #define TCPDEBUG0 int ostate = 0
113 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0
114 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
115 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
119 #define TCPDEBUG2(req)
123 * TCP attaches to socket via pru_attach(), reserving space,
124 * and an internet control block.
127 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
130 struct tcpcb *tp = NULL;
135 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
138 error = tcp_attach(so);
142 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
143 so->so_linger = TCP_LINGERTIME;
148 TCPDEBUG2(PRU_ATTACH);
153 * tcp_detach is called when the socket layer loses its final reference
154 * to the socket, be it a file descriptor reference, a reference from TCP,
155 * etc. At this point, there is only one case in which we will keep around
156 * inpcb state: time wait.
158 * This function can probably be re-absorbed back into tcp_usr_detach() now
159 * that there is a single detach path.
162 tcp_detach(struct socket *so, struct inpcb *inp)
166 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
167 INP_WLOCK_ASSERT(inp);
169 KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
170 KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
174 if (inp->inp_flags & INP_TIMEWAIT) {
176 * There are two cases to handle: one in which the time wait
177 * state is being discarded (INP_DROPPED), and one in which
178 * this connection will remain in timewait. In the former,
179 * it is time to discard all state (except tcptw, which has
180 * already been discarded by the timewait close code, which
181 * should be further up the call stack somewhere). In the
182 * latter case, we detach from the socket, but leave the pcb
183 * present until timewait ends.
185 * XXXRW: Would it be cleaner to free the tcptw here?
187 if (inp->inp_flags & INP_DROPPED) {
188 KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
189 "INP_DROPPED && tp != NULL"));
198 * If the connection is not in timewait, we consider two
199 * two conditions: one in which no further processing is
200 * necessary (dropped || embryonic), and one in which TCP is
201 * not yet done, but no longer requires the socket, so the
202 * pcb will persist for the time being.
204 * XXXRW: Does the second case still occur?
206 if (inp->inp_flags & INP_DROPPED ||
207 tp->t_state < TCPS_SYN_SENT) {
219 * pru_detach() detaches the TCP protocol from the socket.
220 * If the protocol state is non-embryonic, then can't
221 * do this directly: have to initiate a pru_disconnect(),
222 * which may finish later; embryonic TCB's can just
226 tcp_usr_detach(struct socket *so)
231 KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
232 INP_INFO_WLOCK(&V_tcbinfo);
234 KASSERT(inp->inp_socket != NULL,
235 ("tcp_usr_detach: inp_socket == NULL"));
237 INP_INFO_WUNLOCK(&V_tcbinfo);
242 * Give the socket an address.
245 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
249 struct tcpcb *tp = NULL;
250 struct sockaddr_in *sinp;
252 sinp = (struct sockaddr_in *)nam;
253 if (nam->sa_len != sizeof (*sinp))
256 * Must check for multicast addresses and disallow binding
259 if (sinp->sin_family == AF_INET &&
260 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
261 return (EAFNOSUPPORT);
265 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
267 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
273 INP_HASH_WLOCK(&V_tcbinfo);
274 error = in_pcbbind(inp, nam, td->td_ucred);
275 INP_HASH_WUNLOCK(&V_tcbinfo);
286 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
290 struct tcpcb *tp = NULL;
291 struct sockaddr_in6 *sin6p;
293 sin6p = (struct sockaddr_in6 *)nam;
294 if (nam->sa_len != sizeof (*sin6p))
297 * Must check for multicast addresses and disallow binding
300 if (sin6p->sin6_family == AF_INET6 &&
301 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
302 return (EAFNOSUPPORT);
306 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
308 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
314 INP_HASH_WLOCK(&V_tcbinfo);
315 inp->inp_vflag &= ~INP_IPV4;
316 inp->inp_vflag |= INP_IPV6;
318 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
319 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
320 inp->inp_vflag |= INP_IPV4;
321 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
322 struct sockaddr_in sin;
324 in6_sin6_2_sin(&sin, sin6p);
325 inp->inp_vflag |= INP_IPV4;
326 inp->inp_vflag &= ~INP_IPV6;
327 error = in_pcbbind(inp, (struct sockaddr *)&sin,
329 INP_HASH_WUNLOCK(&V_tcbinfo);
334 error = in6_pcbbind(inp, nam, td->td_ucred);
335 INP_HASH_WUNLOCK(&V_tcbinfo);
345 * Prepare to accept connections.
348 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
352 struct tcpcb *tp = NULL;
356 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
358 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
365 error = solisten_proto_check(so);
366 INP_HASH_WLOCK(&V_tcbinfo);
367 if (error == 0 && inp->inp_lport == 0)
368 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
369 INP_HASH_WUNLOCK(&V_tcbinfo);
371 tcp_state_change(tp, TCPS_LISTEN);
372 solisten_proto(so, backlog);
374 if ((so->so_options & SO_NO_OFFLOAD) == 0)
375 tcp_offload_listen_start(tp);
381 TCPDEBUG2(PRU_LISTEN);
389 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
393 struct tcpcb *tp = NULL;
397 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
399 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
406 error = solisten_proto_check(so);
407 INP_HASH_WLOCK(&V_tcbinfo);
408 if (error == 0 && inp->inp_lport == 0) {
409 inp->inp_vflag &= ~INP_IPV4;
410 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
411 inp->inp_vflag |= INP_IPV4;
412 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
414 INP_HASH_WUNLOCK(&V_tcbinfo);
416 tcp_state_change(tp, TCPS_LISTEN);
417 solisten_proto(so, backlog);
419 if ((so->so_options & SO_NO_OFFLOAD) == 0)
420 tcp_offload_listen_start(tp);
426 TCPDEBUG2(PRU_LISTEN);
434 * Initiate connection to peer.
435 * Create a template for use in transmissions on this connection.
436 * Enter SYN_SENT state, and mark socket as connecting.
437 * Start keep-alive timer, and seed output sequence space.
438 * Send initial segment on connection.
441 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
445 struct tcpcb *tp = NULL;
446 struct sockaddr_in *sinp;
448 sinp = (struct sockaddr_in *)nam;
449 if (nam->sa_len != sizeof (*sinp))
452 * Must disallow TCP ``connections'' to multicast addresses.
454 if (sinp->sin_family == AF_INET
455 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
456 return (EAFNOSUPPORT);
457 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
462 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
464 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
470 if ((error = tcp_connect(tp, nam, td)) != 0)
473 if (registered_toedevs > 0 &&
474 (so->so_options & SO_NO_OFFLOAD) == 0 &&
475 (error = tcp_offload_connect(so, nam)) == 0)
478 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
479 error = tcp_output(tp);
481 TCPDEBUG2(PRU_CONNECT);
489 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
493 struct tcpcb *tp = NULL;
494 struct sockaddr_in6 *sin6p;
498 sin6p = (struct sockaddr_in6 *)nam;
499 if (nam->sa_len != sizeof (*sin6p))
502 * Must disallow TCP ``connections'' to multicast addresses.
504 if (sin6p->sin6_family == AF_INET6
505 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
506 return (EAFNOSUPPORT);
509 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
511 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
519 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
520 * therefore probably require the hash lock, which isn't held here.
521 * Is this a significant problem?
523 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
524 struct sockaddr_in sin;
526 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
531 in6_sin6_2_sin(&sin, sin6p);
532 inp->inp_vflag |= INP_IPV4;
533 inp->inp_vflag &= ~INP_IPV6;
534 if ((error = prison_remote_ip4(td->td_ucred,
535 &sin.sin_addr)) != 0)
537 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
540 if (registered_toedevs > 0 &&
541 (so->so_options & SO_NO_OFFLOAD) == 0 &&
542 (error = tcp_offload_connect(so, nam)) == 0)
545 error = tcp_output(tp);
549 inp->inp_vflag &= ~INP_IPV4;
550 inp->inp_vflag |= INP_IPV6;
551 inp->inp_inc.inc_flags |= INC_ISIPV6;
552 if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
554 if ((error = tcp6_connect(tp, nam, td)) != 0)
557 if (registered_toedevs > 0 &&
558 (so->so_options & SO_NO_OFFLOAD) == 0 &&
559 (error = tcp_offload_connect(so, nam)) == 0)
562 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
563 error = tcp_output(tp);
566 TCPDEBUG2(PRU_CONNECT);
573 * Initiate disconnect from peer.
574 * If connection never passed embryonic stage, just drop;
575 * else if don't need to let data drain, then can just drop anyways,
576 * else have to begin TCP shutdown process: mark socket disconnecting,
577 * drain unread data, state switch to reflect user close, and
578 * send segment (e.g. FIN) to peer. Socket will be really disconnected
579 * when peer sends FIN and acks ours.
581 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
584 tcp_usr_disconnect(struct socket *so)
587 struct tcpcb *tp = NULL;
591 INP_INFO_WLOCK(&V_tcbinfo);
593 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
595 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
603 TCPDEBUG2(PRU_DISCONNECT);
605 INP_INFO_WUNLOCK(&V_tcbinfo);
611 * Accept a connection. Essentially all the work is done at higher levels;
612 * just return the address of the peer, storing through addr.
614 * The rationale for acquiring the tcbinfo lock here is somewhat complicated,
615 * and is described in detail in the commit log entry for r175612. Acquiring
616 * it delays an accept(2) racing with sonewconn(), which inserts the socket
617 * before the inpcb address/port fields are initialized. A better fix would
618 * prevent the socket from being placed in the listen queue until all fields
619 * are fully initialized.
622 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
625 struct inpcb *inp = NULL;
626 struct tcpcb *tp = NULL;
631 if (so->so_state & SS_ISDISCONNECTED)
632 return (ECONNABORTED);
635 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
636 INP_INFO_RLOCK(&V_tcbinfo);
638 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
639 error = ECONNABORTED;
646 * We inline in_getpeeraddr and COMMON_END here, so that we can
647 * copy the data of interest and defer the malloc until after we
650 port = inp->inp_fport;
651 addr = inp->inp_faddr;
654 TCPDEBUG2(PRU_ACCEPT);
656 INP_INFO_RUNLOCK(&V_tcbinfo);
658 *nam = in_sockaddr(port, &addr);
665 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
667 struct inpcb *inp = NULL;
669 struct tcpcb *tp = NULL;
671 struct in6_addr addr6;
676 if (so->so_state & SS_ISDISCONNECTED)
677 return (ECONNABORTED);
680 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
681 INP_INFO_RLOCK(&V_tcbinfo);
683 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
684 error = ECONNABORTED;
691 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
692 * copy the data of interest and defer the malloc until after we
695 if (inp->inp_vflag & INP_IPV4) {
697 port = inp->inp_fport;
698 addr = inp->inp_faddr;
700 port = inp->inp_fport;
701 addr6 = inp->in6p_faddr;
705 TCPDEBUG2(PRU_ACCEPT);
707 INP_INFO_RUNLOCK(&V_tcbinfo);
710 *nam = in6_v4mapsin6_sockaddr(port, &addr);
712 *nam = in6_sockaddr(port, &addr6);
719 * Mark the connection as being incapable of further output.
722 tcp_usr_shutdown(struct socket *so)
726 struct tcpcb *tp = NULL;
729 INP_INFO_WLOCK(&V_tcbinfo);
731 KASSERT(inp != NULL, ("inp == NULL"));
733 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
741 if (!(inp->inp_flags & INP_DROPPED))
742 error = tcp_output(tp);
745 TCPDEBUG2(PRU_SHUTDOWN);
747 INP_INFO_WUNLOCK(&V_tcbinfo);
753 * After a receive, possibly send window update to peer.
756 tcp_usr_rcvd(struct socket *so, int flags)
759 struct tcpcb *tp = NULL;
764 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
766 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
773 if (tp->t_flags & TF_TOE)
774 tcp_offload_rcvd(tp);
786 * Do a send by putting data in output queue and updating urgent
787 * marker if URG set. Possibly send more data. Unlike the other
788 * pru_*() routines, the mbuf chains are our responsibility. We
789 * must either enqueue them or free them. The other pru_* routines
790 * generally are caller-frees.
793 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
794 struct sockaddr *nam, struct mbuf *control, struct thread *td)
798 struct tcpcb *tp = NULL;
805 * We require the pcbinfo lock if we will close the socket as part of
808 if (flags & PRUS_EOF)
809 INP_INFO_WLOCK(&V_tcbinfo);
811 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
813 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
822 isipv6 = nam && nam->sa_family == AF_INET6;
827 /* TCP doesn't do control messages (rights, creds, etc) */
828 if (control->m_len) {
835 m_freem(control); /* empty control, just free it */
837 if (!(flags & PRUS_OOB)) {
838 sbappendstream(&so->so_snd, m);
839 if (nam && tp->t_state < TCPS_SYN_SENT) {
841 * Do implied connect if not yet connected,
842 * initialize window to default value, and
843 * initialize maxseg/maxopd using peer's cached
848 error = tcp6_connect(tp, nam, td);
850 #if defined(INET6) && defined(INET)
854 error = tcp_connect(tp, nam, td);
858 tp->snd_wnd = TTCP_CLIENT_SND_WND;
861 if (flags & PRUS_EOF) {
863 * Close the send side of the connection after
866 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
870 if (!(inp->inp_flags & INP_DROPPED)) {
871 if (flags & PRUS_MORETOCOME)
872 tp->t_flags |= TF_MORETOCOME;
873 error = tcp_output(tp);
874 if (flags & PRUS_MORETOCOME)
875 tp->t_flags &= ~TF_MORETOCOME;
879 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
881 SOCKBUF_LOCK(&so->so_snd);
882 if (sbspace(&so->so_snd) < -512) {
883 SOCKBUF_UNLOCK(&so->so_snd);
889 * According to RFC961 (Assigned Protocols),
890 * the urgent pointer points to the last octet
891 * of urgent data. We continue, however,
892 * to consider it to indicate the first octet
893 * of data past the urgent section.
894 * Otherwise, snd_up should be one lower.
896 sbappendstream_locked(&so->so_snd, m);
897 SOCKBUF_UNLOCK(&so->so_snd);
898 if (nam && tp->t_state < TCPS_SYN_SENT) {
900 * Do implied connect if not yet connected,
901 * initialize window to default value, and
902 * initialize maxseg/maxopd using peer's cached
907 error = tcp6_connect(tp, nam, td);
909 #if defined(INET6) && defined(INET)
913 error = tcp_connect(tp, nam, td);
917 tp->snd_wnd = TTCP_CLIENT_SND_WND;
920 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
921 tp->t_flags |= TF_FORCEDATA;
922 error = tcp_output(tp);
923 tp->t_flags &= ~TF_FORCEDATA;
926 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
927 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
929 if (flags & PRUS_EOF)
930 INP_INFO_WUNLOCK(&V_tcbinfo);
935 * Abort the TCP. Drop the connection abruptly.
938 tcp_usr_abort(struct socket *so)
941 struct tcpcb *tp = NULL;
945 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
947 INP_INFO_WLOCK(&V_tcbinfo);
949 KASSERT(inp->inp_socket != NULL,
950 ("tcp_usr_abort: inp_socket == NULL"));
953 * If we still have full TCP state, and we're not dropped, drop.
955 if (!(inp->inp_flags & INP_TIMEWAIT) &&
956 !(inp->inp_flags & INP_DROPPED)) {
959 tcp_drop(tp, ECONNABORTED);
960 TCPDEBUG2(PRU_ABORT);
962 if (!(inp->inp_flags & INP_DROPPED)) {
964 so->so_state |= SS_PROTOREF;
966 inp->inp_flags |= INP_SOCKREF;
969 INP_INFO_WUNLOCK(&V_tcbinfo);
973 * TCP socket is closed. Start friendly disconnect.
976 tcp_usr_close(struct socket *so)
979 struct tcpcb *tp = NULL;
983 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
985 INP_INFO_WLOCK(&V_tcbinfo);
987 KASSERT(inp->inp_socket != NULL,
988 ("tcp_usr_close: inp_socket == NULL"));
991 * If we still have full TCP state, and we're not dropped, initiate
994 if (!(inp->inp_flags & INP_TIMEWAIT) &&
995 !(inp->inp_flags & INP_DROPPED)) {
999 TCPDEBUG2(PRU_CLOSE);
1001 if (!(inp->inp_flags & INP_DROPPED)) {
1003 so->so_state |= SS_PROTOREF;
1005 inp->inp_flags |= INP_SOCKREF;
1008 INP_INFO_WUNLOCK(&V_tcbinfo);
1012 * Receive out-of-band data.
1015 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1019 struct tcpcb *tp = NULL;
1022 inp = sotoinpcb(so);
1023 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1025 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1029 tp = intotcpcb(inp);
1031 if ((so->so_oobmark == 0 &&
1032 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1033 so->so_options & SO_OOBINLINE ||
1034 tp->t_oobflags & TCPOOB_HADDATA) {
1038 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1039 error = EWOULDBLOCK;
1043 *mtod(m, caddr_t) = tp->t_iobc;
1044 if ((flags & MSG_PEEK) == 0)
1045 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1048 TCPDEBUG2(PRU_RCVOOB);
1054 struct pr_usrreqs tcp_usrreqs = {
1055 .pru_abort = tcp_usr_abort,
1056 .pru_accept = tcp_usr_accept,
1057 .pru_attach = tcp_usr_attach,
1058 .pru_bind = tcp_usr_bind,
1059 .pru_connect = tcp_usr_connect,
1060 .pru_control = in_control,
1061 .pru_detach = tcp_usr_detach,
1062 .pru_disconnect = tcp_usr_disconnect,
1063 .pru_listen = tcp_usr_listen,
1064 .pru_peeraddr = in_getpeeraddr,
1065 .pru_rcvd = tcp_usr_rcvd,
1066 .pru_rcvoob = tcp_usr_rcvoob,
1067 .pru_send = tcp_usr_send,
1068 .pru_shutdown = tcp_usr_shutdown,
1069 .pru_sockaddr = in_getsockaddr,
1070 .pru_sosetlabel = in_pcbsosetlabel,
1071 .pru_close = tcp_usr_close,
1076 struct pr_usrreqs tcp6_usrreqs = {
1077 .pru_abort = tcp_usr_abort,
1078 .pru_accept = tcp6_usr_accept,
1079 .pru_attach = tcp_usr_attach,
1080 .pru_bind = tcp6_usr_bind,
1081 .pru_connect = tcp6_usr_connect,
1082 .pru_control = in6_control,
1083 .pru_detach = tcp_usr_detach,
1084 .pru_disconnect = tcp_usr_disconnect,
1085 .pru_listen = tcp6_usr_listen,
1086 .pru_peeraddr = in6_mapped_peeraddr,
1087 .pru_rcvd = tcp_usr_rcvd,
1088 .pru_rcvoob = tcp_usr_rcvoob,
1089 .pru_send = tcp_usr_send,
1090 .pru_shutdown = tcp_usr_shutdown,
1091 .pru_sockaddr = in6_mapped_sockaddr,
1092 .pru_sosetlabel = in_pcbsosetlabel,
1093 .pru_close = tcp_usr_close,
1099 * Common subroutine to open a TCP connection to remote host specified
1100 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
1101 * port number if needed. Call in_pcbconnect_setup to do the routing and
1102 * to choose a local host address (interface). If there is an existing
1103 * incarnation of the same connection in TIME-WAIT state and if the remote
1104 * host was sending CC options and if the connection duration was < MSL, then
1105 * truncate the previous TIME-WAIT state and proceed.
1106 * Initialize connection parameters and enter SYN-SENT state.
1109 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1111 struct inpcb *inp = tp->t_inpcb, *oinp;
1112 struct socket *so = inp->inp_socket;
1113 struct in_addr laddr;
1117 INP_WLOCK_ASSERT(inp);
1118 INP_HASH_WLOCK(&V_tcbinfo);
1120 if (inp->inp_lport == 0) {
1121 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1127 * Cannot simply call in_pcbconnect, because there might be an
1128 * earlier incarnation of this same connection still in
1129 * TIME_WAIT state, creating an ADDRINUSE error.
1131 laddr = inp->inp_laddr;
1132 lport = inp->inp_lport;
1133 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1134 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1135 if (error && oinp == NULL)
1141 inp->inp_laddr = laddr;
1143 INP_HASH_WUNLOCK(&V_tcbinfo);
1146 * Compute window scaling to request:
1147 * Scale to fit into sweet spot. See tcp_syncache.c.
1148 * XXX: This should move to tcp_output().
1150 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1151 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1152 tp->request_r_scale++;
1155 TCPSTAT_INC(tcps_connattempt);
1156 tcp_state_change(tp, TCPS_SYN_SENT);
1157 tp->iss = tcp_new_isn(tp);
1158 tcp_sendseqinit(tp);
1163 INP_HASH_WUNLOCK(&V_tcbinfo);
1170 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1172 struct inpcb *inp = tp->t_inpcb, *oinp;
1173 struct socket *so = inp->inp_socket;
1174 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
1175 struct in6_addr addr6;
1178 INP_WLOCK_ASSERT(inp);
1179 INP_HASH_WLOCK(&V_tcbinfo);
1181 if (inp->inp_lport == 0) {
1182 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1188 * Cannot simply call in_pcbconnect, because there might be an
1189 * earlier incarnation of this same connection still in
1190 * TIME_WAIT state, creating an ADDRINUSE error.
1191 * in6_pcbladdr() also handles scope zone IDs.
1193 * XXXRW: We wouldn't need to expose in6_pcblookup_hash_locked()
1194 * outside of in6_pcb.c if there were an in6_pcbconnect_setup().
1196 error = in6_pcbladdr(inp, nam, &addr6);
1199 oinp = in6_pcblookup_hash_locked(inp->inp_pcbinfo,
1200 &sin6->sin6_addr, sin6->sin6_port,
1201 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
1204 inp->inp_lport, 0, NULL);
1209 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1210 inp->in6p_laddr = addr6;
1211 inp->in6p_faddr = sin6->sin6_addr;
1212 inp->inp_fport = sin6->sin6_port;
1213 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
1214 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
1215 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
1217 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
1219 INP_HASH_WUNLOCK(&V_tcbinfo);
1221 /* Compute window scaling to request. */
1222 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1223 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1224 tp->request_r_scale++;
1227 TCPSTAT_INC(tcps_connattempt);
1228 tcp_state_change(tp, TCPS_SYN_SENT);
1229 tp->iss = tcp_new_isn(tp);
1230 tcp_sendseqinit(tp);
1235 INP_HASH_WUNLOCK(&V_tcbinfo);
1241 * Export TCP internal state information via a struct tcp_info, based on the
1242 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently
1243 * (TCP state machine, etc). We export all information using FreeBSD-native
1244 * constants -- for example, the numeric values for tcpi_state will differ
1248 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1251 INP_WLOCK_ASSERT(tp->t_inpcb);
1252 bzero(ti, sizeof(*ti));
1254 ti->tcpi_state = tp->t_state;
1255 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1256 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1257 if (tp->t_flags & TF_SACK_PERMIT)
1258 ti->tcpi_options |= TCPI_OPT_SACK;
1259 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1260 ti->tcpi_options |= TCPI_OPT_WSCALE;
1261 ti->tcpi_snd_wscale = tp->snd_scale;
1262 ti->tcpi_rcv_wscale = tp->rcv_scale;
1265 ti->tcpi_rto = tp->t_rxtcur * tick;
1266 ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick;
1267 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1268 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1270 ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1271 ti->tcpi_snd_cwnd = tp->snd_cwnd;
1274 * FreeBSD-specific extension fields for tcp_info.
1276 ti->tcpi_rcv_space = tp->rcv_wnd;
1277 ti->tcpi_rcv_nxt = tp->rcv_nxt;
1278 ti->tcpi_snd_wnd = tp->snd_wnd;
1279 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */
1280 ti->tcpi_snd_nxt = tp->snd_nxt;
1281 ti->tcpi_snd_mss = tp->t_maxseg;
1282 ti->tcpi_rcv_mss = tp->t_maxseg;
1283 if (tp->t_flags & TF_TOE)
1284 ti->tcpi_options |= TCPI_OPT_TOE;
1285 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1286 ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1287 ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1291 * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1292 * socket option arguments. When it re-acquires the lock after the copy, it
1293 * has to revalidate that the connection is still valid for the socket
1296 #define INP_WLOCK_RECHECK(inp) do { \
1298 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
1300 return (ECONNRESET); \
1302 tp = intotcpcb(inp); \
1306 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1308 int error, opt, optval;
1313 char buf[TCP_CA_NAME_MAX];
1314 struct cc_algo *algo;
1317 inp = sotoinpcb(so);
1318 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1320 if (sopt->sopt_level != IPPROTO_TCP) {
1322 if (inp->inp_vflag & INP_IPV6PROTO) {
1324 error = ip6_ctloutput(so, sopt);
1327 #if defined(INET6) && defined(INET)
1333 error = ip_ctloutput(so, sopt);
1338 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1340 return (ECONNRESET);
1343 switch (sopt->sopt_dir) {
1345 switch (sopt->sopt_name) {
1346 #ifdef TCP_SIGNATURE
1349 error = sooptcopyin(sopt, &optval, sizeof optval,
1354 INP_WLOCK_RECHECK(inp);
1356 tp->t_flags |= TF_SIGNATURE;
1358 tp->t_flags &= ~TF_SIGNATURE;
1359 goto unlock_and_done;
1360 #endif /* TCP_SIGNATURE */
1365 error = sooptcopyin(sopt, &optval, sizeof optval,
1370 INP_WLOCK_RECHECK(inp);
1371 switch (sopt->sopt_name) {
1379 opt = 0; /* dead code to fool gcc */
1386 tp->t_flags &= ~opt;
1389 if (tp->t_flags & TF_TOE) {
1390 tcp_offload_ctloutput(tp, sopt->sopt_dir,
1399 error = sooptcopyin(sopt, &optval, sizeof optval,
1404 INP_WLOCK_RECHECK(inp);
1406 tp->t_flags |= TF_NOPUSH;
1407 else if (tp->t_flags & TF_NOPUSH) {
1408 tp->t_flags &= ~TF_NOPUSH;
1409 if (TCPS_HAVEESTABLISHED(tp->t_state))
1410 error = tcp_output(tp);
1412 goto unlock_and_done;
1416 error = sooptcopyin(sopt, &optval, sizeof optval,
1421 INP_WLOCK_RECHECK(inp);
1422 if (optval > 0 && optval <= tp->t_maxseg &&
1423 optval + 40 >= V_tcp_minmss)
1424 tp->t_maxseg = optval;
1427 goto unlock_and_done;
1434 case TCP_CONGESTION:
1436 bzero(buf, sizeof(buf));
1437 error = sooptcopyin(sopt, &buf, sizeof(buf), 1);
1440 INP_WLOCK_RECHECK(inp);
1442 * Return EINVAL if we can't find the requested cc algo.
1446 STAILQ_FOREACH(algo, &cc_list, entries) {
1447 if (strncmp(buf, algo->name, TCP_CA_NAME_MAX)
1449 /* We've found the requested algo. */
1452 * We hold a write lock over the tcb
1453 * so it's safe to do these things
1454 * without ordering concerns.
1456 if (CC_ALGO(tp)->cb_destroy != NULL)
1457 CC_ALGO(tp)->cb_destroy(tp->ccv);
1460 * If something goes pear shaped
1461 * initialising the new algo,
1462 * fall back to newreno (which
1463 * does not require initialisation).
1465 if (algo->cb_init != NULL)
1466 if (algo->cb_init(tp->ccv) > 0) {
1467 CC_ALGO(tp) = &newreno_cc_algo;
1469 * The only reason init
1471 * because of malloc.
1475 break; /* Break the STAILQ_FOREACH. */
1479 goto unlock_and_done;
1485 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1489 if (ui > (UINT_MAX / hz)) {
1495 INP_WLOCK_RECHECK(inp);
1496 switch (sopt->sopt_name) {
1498 tp->t_keepidle = ui;
1500 * XXX: better check current remaining
1501 * timeout and "merge" it with new value.
1503 if ((tp->t_state > TCPS_LISTEN) &&
1504 (tp->t_state <= TCPS_CLOSING))
1505 tcp_timer_activate(tp, TT_KEEP,
1509 tp->t_keepintvl = ui;
1510 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1511 (TP_MAXIDLE(tp) > 0))
1512 tcp_timer_activate(tp, TT_2MSL,
1516 tp->t_keepinit = ui;
1517 if (tp->t_state == TCPS_SYN_RECEIVED ||
1518 tp->t_state == TCPS_SYN_SENT)
1519 tcp_timer_activate(tp, TT_KEEP,
1523 goto unlock_and_done;
1527 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1531 INP_WLOCK_RECHECK(inp);
1533 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1534 (TP_MAXIDLE(tp) > 0))
1535 tcp_timer_activate(tp, TT_2MSL,
1537 goto unlock_and_done;
1541 error = ENOPROTOOPT;
1547 tp = intotcpcb(inp);
1548 switch (sopt->sopt_name) {
1549 #ifdef TCP_SIGNATURE
1551 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1553 error = sooptcopyout(sopt, &optval, sizeof optval);
1558 optval = tp->t_flags & TF_NODELAY;
1560 error = sooptcopyout(sopt, &optval, sizeof optval);
1563 optval = tp->t_maxseg;
1565 error = sooptcopyout(sopt, &optval, sizeof optval);
1568 optval = tp->t_flags & TF_NOOPT;
1570 error = sooptcopyout(sopt, &optval, sizeof optval);
1573 optval = tp->t_flags & TF_NOPUSH;
1575 error = sooptcopyout(sopt, &optval, sizeof optval);
1578 tcp_fill_info(tp, &ti);
1580 error = sooptcopyout(sopt, &ti, sizeof ti);
1582 case TCP_CONGESTION:
1583 bzero(buf, sizeof(buf));
1584 strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1586 error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX);
1590 error = ENOPROTOOPT;
1597 #undef INP_WLOCK_RECHECK
1600 * Attach TCP protocol to socket, allocating
1601 * internet protocol control block, tcp control block,
1602 * bufer space, and entering LISTEN state if to accept connections.
1605 tcp_attach(struct socket *so)
1611 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1612 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
1616 so->so_rcv.sb_flags |= SB_AUTOSIZE;
1617 so->so_snd.sb_flags |= SB_AUTOSIZE;
1618 INP_INFO_WLOCK(&V_tcbinfo);
1619 error = in_pcballoc(so, &V_tcbinfo);
1621 INP_INFO_WUNLOCK(&V_tcbinfo);
1624 inp = sotoinpcb(so);
1626 if (inp->inp_vflag & INP_IPV6PROTO) {
1627 inp->inp_vflag |= INP_IPV6;
1628 inp->in6p_hops = -1; /* use kernel default */
1632 inp->inp_vflag |= INP_IPV4;
1633 tp = tcp_newtcpcb(inp);
1637 INP_INFO_WUNLOCK(&V_tcbinfo);
1640 tp->t_state = TCPS_CLOSED;
1642 INP_INFO_WUNLOCK(&V_tcbinfo);
1647 * Initiate (or continue) disconnect.
1648 * If embryonic state, just send reset (once).
1649 * If in ``let data drain'' option and linger null, just drop.
1650 * Otherwise (hard), mark socket disconnecting and drop
1651 * current input data; switch states based on user close, and
1652 * send segment to peer (with FIN).
1655 tcp_disconnect(struct tcpcb *tp)
1657 struct inpcb *inp = tp->t_inpcb;
1658 struct socket *so = inp->inp_socket;
1660 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1661 INP_WLOCK_ASSERT(inp);
1664 * Neither tcp_close() nor tcp_drop() should return NULL, as the
1665 * socket is still open.
1667 if (tp->t_state < TCPS_ESTABLISHED) {
1670 ("tcp_disconnect: tcp_close() returned NULL"));
1671 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1672 tp = tcp_drop(tp, 0);
1674 ("tcp_disconnect: tcp_drop() returned NULL"));
1676 soisdisconnecting(so);
1677 sbflush(&so->so_rcv);
1679 if (!(inp->inp_flags & INP_DROPPED))
1685 * User issued close, and wish to trail through shutdown states:
1686 * if never received SYN, just forget it. If got a SYN from peer,
1687 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1688 * If already got a FIN from peer, then almost done; go to LAST_ACK
1689 * state. In all other cases, have already sent FIN to peer (e.g.
1690 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1691 * for peer to send FIN or not respond to keep-alives, etc.
1692 * We can let the user exit from the close as soon as the FIN is acked.
1695 tcp_usrclosed(struct tcpcb *tp)
1698 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1699 INP_WLOCK_ASSERT(tp->t_inpcb);
1701 switch (tp->t_state) {
1704 tcp_offload_listen_stop(tp);
1708 tcp_state_change(tp, TCPS_CLOSED);
1711 * tcp_close() should never return NULL here as the socket is
1715 ("tcp_usrclosed: tcp_close() returned NULL"));
1719 case TCPS_SYN_RECEIVED:
1720 tp->t_flags |= TF_NEEDFIN;
1723 case TCPS_ESTABLISHED:
1724 tcp_state_change(tp, TCPS_FIN_WAIT_1);
1727 case TCPS_CLOSE_WAIT:
1728 tcp_state_change(tp, TCPS_LAST_ACK);
1731 if (tp->t_state >= TCPS_FIN_WAIT_2) {
1732 soisdisconnected(tp->t_inpcb->inp_socket);
1733 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
1734 if (tp->t_state == TCPS_FIN_WAIT_2) {
1737 timeout = (tcp_fast_finwait2_recycle) ?
1738 tcp_finwait2_timeout : TP_MAXIDLE(tp);
1739 tcp_timer_activate(tp, TT_2MSL, timeout);
1746 db_print_indent(int indent)
1750 for (i = 0; i < indent; i++)
1755 db_print_tstate(int t_state)
1760 db_printf("TCPS_CLOSED");
1764 db_printf("TCPS_LISTEN");
1768 db_printf("TCPS_SYN_SENT");
1771 case TCPS_SYN_RECEIVED:
1772 db_printf("TCPS_SYN_RECEIVED");
1775 case TCPS_ESTABLISHED:
1776 db_printf("TCPS_ESTABLISHED");
1779 case TCPS_CLOSE_WAIT:
1780 db_printf("TCPS_CLOSE_WAIT");
1783 case TCPS_FIN_WAIT_1:
1784 db_printf("TCPS_FIN_WAIT_1");
1788 db_printf("TCPS_CLOSING");
1792 db_printf("TCPS_LAST_ACK");
1795 case TCPS_FIN_WAIT_2:
1796 db_printf("TCPS_FIN_WAIT_2");
1799 case TCPS_TIME_WAIT:
1800 db_printf("TCPS_TIME_WAIT");
1804 db_printf("unknown");
1810 db_print_tflags(u_int t_flags)
1815 if (t_flags & TF_ACKNOW) {
1816 db_printf("%sTF_ACKNOW", comma ? ", " : "");
1819 if (t_flags & TF_DELACK) {
1820 db_printf("%sTF_DELACK", comma ? ", " : "");
1823 if (t_flags & TF_NODELAY) {
1824 db_printf("%sTF_NODELAY", comma ? ", " : "");
1827 if (t_flags & TF_NOOPT) {
1828 db_printf("%sTF_NOOPT", comma ? ", " : "");
1831 if (t_flags & TF_SENTFIN) {
1832 db_printf("%sTF_SENTFIN", comma ? ", " : "");
1835 if (t_flags & TF_REQ_SCALE) {
1836 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
1839 if (t_flags & TF_RCVD_SCALE) {
1840 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
1843 if (t_flags & TF_REQ_TSTMP) {
1844 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
1847 if (t_flags & TF_RCVD_TSTMP) {
1848 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
1851 if (t_flags & TF_SACK_PERMIT) {
1852 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
1855 if (t_flags & TF_NEEDSYN) {
1856 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
1859 if (t_flags & TF_NEEDFIN) {
1860 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
1863 if (t_flags & TF_NOPUSH) {
1864 db_printf("%sTF_NOPUSH", comma ? ", " : "");
1867 if (t_flags & TF_MORETOCOME) {
1868 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
1871 if (t_flags & TF_LQ_OVERFLOW) {
1872 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
1875 if (t_flags & TF_LASTIDLE) {
1876 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
1879 if (t_flags & TF_RXWIN0SENT) {
1880 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
1883 if (t_flags & TF_FASTRECOVERY) {
1884 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
1887 if (t_flags & TF_CONGRECOVERY) {
1888 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
1891 if (t_flags & TF_WASFRECOVERY) {
1892 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
1895 if (t_flags & TF_SIGNATURE) {
1896 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
1899 if (t_flags & TF_FORCEDATA) {
1900 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
1903 if (t_flags & TF_TSO) {
1904 db_printf("%sTF_TSO", comma ? ", " : "");
1907 if (t_flags & TF_ECN_PERMIT) {
1908 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
1914 db_print_toobflags(char t_oobflags)
1919 if (t_oobflags & TCPOOB_HAVEDATA) {
1920 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
1923 if (t_oobflags & TCPOOB_HADDATA) {
1924 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
1930 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
1933 db_print_indent(indent);
1934 db_printf("%s at %p\n", name, tp);
1938 db_print_indent(indent);
1939 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n",
1940 LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
1942 db_print_indent(indent);
1943 db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n",
1944 &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
1946 db_print_indent(indent);
1947 db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl,
1948 &tp->t_timers->tt_delack, tp->t_inpcb);
1950 db_print_indent(indent);
1951 db_printf("t_state: %d (", tp->t_state);
1952 db_print_tstate(tp->t_state);
1955 db_print_indent(indent);
1956 db_printf("t_flags: 0x%x (", tp->t_flags);
1957 db_print_tflags(tp->t_flags);
1960 db_print_indent(indent);
1961 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n",
1962 tp->snd_una, tp->snd_max, tp->snd_nxt);
1964 db_print_indent(indent);
1965 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n",
1966 tp->snd_up, tp->snd_wl1, tp->snd_wl2);
1968 db_print_indent(indent);
1969 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n",
1970 tp->iss, tp->irs, tp->rcv_nxt);
1972 db_print_indent(indent);
1973 db_printf("rcv_adv: 0x%08x rcv_wnd: %lu rcv_up: 0x%08x\n",
1974 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
1976 db_print_indent(indent);
1977 db_printf("snd_wnd: %lu snd_cwnd: %lu\n",
1978 tp->snd_wnd, tp->snd_cwnd);
1980 db_print_indent(indent);
1981 db_printf("snd_ssthresh: %lu snd_recover: "
1982 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
1984 db_print_indent(indent);
1985 db_printf("t_maxopd: %u t_rcvtime: %u t_startime: %u\n",
1986 tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
1988 db_print_indent(indent);
1989 db_printf("t_rttime: %u t_rtsq: 0x%08x\n",
1990 tp->t_rtttime, tp->t_rtseq);
1992 db_print_indent(indent);
1993 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n",
1994 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
1996 db_print_indent(indent);
1997 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u "
1998 "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
2001 db_print_indent(indent);
2002 db_printf("t_rttupdated: %lu max_sndwnd: %lu t_softerror: %d\n",
2003 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
2005 db_print_indent(indent);
2006 db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
2007 db_print_toobflags(tp->t_oobflags);
2008 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc);
2010 db_print_indent(indent);
2011 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n",
2012 tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
2014 db_print_indent(indent);
2015 db_printf("ts_recent: %u ts_recent_age: %u\n",
2016 tp->ts_recent, tp->ts_recent_age);
2018 db_print_indent(indent);
2019 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: "
2020 "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
2022 db_print_indent(indent);
2023 db_printf("snd_ssthresh_prev: %lu snd_recover_prev: 0x%08x "
2024 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2025 tp->snd_recover_prev, tp->t_badrxtwin);
2027 db_print_indent(indent);
2028 db_printf("snd_numholes: %d snd_holes first: %p\n",
2029 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2031 db_print_indent(indent);
2032 db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: "
2033 "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
2035 /* Skip sackblks, sackhint. */
2037 db_print_indent(indent);
2038 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n",
2039 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2042 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2047 db_printf("usage: show tcpcb <addr>\n");
2050 tp = (struct tcpcb *)addr;
2052 db_print_tcpcb(tp, "tcpcb", 0);