2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California.
4 * Copyright (c) 2006-2007 Robert N. M. Watson
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 4. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
39 #include "opt_inet6.h"
40 #include "opt_tcpdebug.h"
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/malloc.h>
45 #include <sys/kernel.h>
46 #include <sys/sysctl.h>
49 #include <sys/domain.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/protosw.h>
62 #include <net/route.h>
65 #include <netinet/cc.h>
66 #include <netinet/in.h>
67 #include <netinet/in_systm.h>
69 #include <netinet/ip6.h>
71 #include <netinet/in_pcb.h>
73 #include <netinet6/in6_pcb.h>
75 #include <netinet/in_var.h>
76 #include <netinet/ip_var.h>
78 #include <netinet6/ip6_var.h>
79 #include <netinet6/scope6_var.h>
81 #include <netinet/tcp_fsm.h>
82 #include <netinet/tcp_seq.h>
83 #include <netinet/tcp_timer.h>
84 #include <netinet/tcp_var.h>
85 #include <netinet/tcpip.h>
87 #include <netinet/tcp_debug.h>
89 #include <netinet/tcp_offload.h>
92 * TCP protocol interface to socket abstraction.
94 static int tcp_attach(struct socket *);
95 static int tcp_connect(struct tcpcb *, struct sockaddr *,
98 static int tcp6_connect(struct tcpcb *, struct sockaddr *,
101 static void tcp_disconnect(struct tcpcb *);
102 static void tcp_usrclosed(struct tcpcb *);
103 static void tcp_fill_info(struct tcpcb *, struct tcp_info *);
106 #define TCPDEBUG0 int ostate = 0
107 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0
108 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
109 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
113 #define TCPDEBUG2(req)
117 * TCP attaches to socket via pru_attach(), reserving space,
118 * and an internet control block.
121 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
124 struct tcpcb *tp = NULL;
129 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
132 error = tcp_attach(so);
136 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
137 so->so_linger = TCP_LINGERTIME;
142 TCPDEBUG2(PRU_ATTACH);
147 * tcp_detach is called when the socket layer loses its final reference
148 * to the socket, be it a file descriptor reference, a reference from TCP,
149 * etc. At this point, there is only one case in which we will keep around
150 * inpcb state: time wait.
152 * This function can probably be re-absorbed back into tcp_usr_detach() now
153 * that there is a single detach path.
156 tcp_detach(struct socket *so, struct inpcb *inp)
160 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
161 INP_WLOCK_ASSERT(inp);
163 KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
164 KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
168 if (inp->inp_flags & INP_TIMEWAIT) {
170 * There are two cases to handle: one in which the time wait
171 * state is being discarded (INP_DROPPED), and one in which
172 * this connection will remain in timewait. In the former,
173 * it is time to discard all state (except tcptw, which has
174 * already been discarded by the timewait close code, which
175 * should be further up the call stack somewhere). In the
176 * latter case, we detach from the socket, but leave the pcb
177 * present until timewait ends.
179 * XXXRW: Would it be cleaner to free the tcptw here?
181 if (inp->inp_flags & INP_DROPPED) {
182 KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
183 "INP_DROPPED && tp != NULL"));
192 * If the connection is not in timewait, we consider two
193 * two conditions: one in which no further processing is
194 * necessary (dropped || embryonic), and one in which TCP is
195 * not yet done, but no longer requires the socket, so the
196 * pcb will persist for the time being.
198 * XXXRW: Does the second case still occur?
200 if (inp->inp_flags & INP_DROPPED ||
201 tp->t_state < TCPS_SYN_SENT) {
211 * pru_detach() detaches the TCP protocol from the socket.
212 * If the protocol state is non-embryonic, then can't
213 * do this directly: have to initiate a pru_disconnect(),
214 * which may finish later; embryonic TCB's can just
218 tcp_usr_detach(struct socket *so)
223 KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
224 INP_INFO_WLOCK(&V_tcbinfo);
226 KASSERT(inp->inp_socket != NULL,
227 ("tcp_usr_detach: inp_socket == NULL"));
229 INP_INFO_WUNLOCK(&V_tcbinfo);
233 * Give the socket an address.
236 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
240 struct tcpcb *tp = NULL;
241 struct sockaddr_in *sinp;
243 sinp = (struct sockaddr_in *)nam;
244 if (nam->sa_len != sizeof (*sinp))
247 * Must check for multicast addresses and disallow binding
250 if (sinp->sin_family == AF_INET &&
251 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
252 return (EAFNOSUPPORT);
255 INP_INFO_WLOCK(&V_tcbinfo);
257 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
259 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
265 error = in_pcbbind(inp, nam, td->td_ucred);
269 INP_INFO_WUNLOCK(&V_tcbinfo);
276 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
280 struct tcpcb *tp = NULL;
281 struct sockaddr_in6 *sin6p;
283 sin6p = (struct sockaddr_in6 *)nam;
284 if (nam->sa_len != sizeof (*sin6p))
287 * Must check for multicast addresses and disallow binding
290 if (sin6p->sin6_family == AF_INET6 &&
291 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
292 return (EAFNOSUPPORT);
295 INP_INFO_WLOCK(&V_tcbinfo);
297 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
299 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
305 inp->inp_vflag &= ~INP_IPV4;
306 inp->inp_vflag |= INP_IPV6;
307 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
308 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
309 inp->inp_vflag |= INP_IPV4;
310 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
311 struct sockaddr_in sin;
313 in6_sin6_2_sin(&sin, sin6p);
314 inp->inp_vflag |= INP_IPV4;
315 inp->inp_vflag &= ~INP_IPV6;
316 error = in_pcbbind(inp, (struct sockaddr *)&sin,
321 error = in6_pcbbind(inp, nam, td->td_ucred);
325 INP_INFO_WUNLOCK(&V_tcbinfo);
331 * Prepare to accept connections.
334 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
338 struct tcpcb *tp = NULL;
341 INP_INFO_WLOCK(&V_tcbinfo);
343 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
345 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
352 error = solisten_proto_check(so);
353 if (error == 0 && inp->inp_lport == 0)
354 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
356 tp->t_state = TCPS_LISTEN;
357 solisten_proto(so, backlog);
358 tcp_offload_listen_open(tp);
363 TCPDEBUG2(PRU_LISTEN);
365 INP_INFO_WUNLOCK(&V_tcbinfo);
371 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
375 struct tcpcb *tp = NULL;
378 INP_INFO_WLOCK(&V_tcbinfo);
380 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
382 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
389 error = solisten_proto_check(so);
390 if (error == 0 && inp->inp_lport == 0) {
391 inp->inp_vflag &= ~INP_IPV4;
392 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
393 inp->inp_vflag |= INP_IPV4;
394 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
397 tp->t_state = TCPS_LISTEN;
398 solisten_proto(so, backlog);
403 TCPDEBUG2(PRU_LISTEN);
405 INP_INFO_WUNLOCK(&V_tcbinfo);
411 * Initiate connection to peer.
412 * Create a template for use in transmissions on this connection.
413 * Enter SYN_SENT state, and mark socket as connecting.
414 * Start keep-alive timer, and seed output sequence space.
415 * Send initial segment on connection.
418 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
422 struct tcpcb *tp = NULL;
423 struct sockaddr_in *sinp;
425 sinp = (struct sockaddr_in *)nam;
426 if (nam->sa_len != sizeof (*sinp))
429 * Must disallow TCP ``connections'' to multicast addresses.
431 if (sinp->sin_family == AF_INET
432 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
433 return (EAFNOSUPPORT);
434 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
438 INP_INFO_WLOCK(&V_tcbinfo);
440 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
442 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
448 if ((error = tcp_connect(tp, nam, td)) != 0)
450 error = tcp_output_connect(so, nam);
452 TCPDEBUG2(PRU_CONNECT);
454 INP_INFO_WUNLOCK(&V_tcbinfo);
460 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
464 struct tcpcb *tp = NULL;
465 struct sockaddr_in6 *sin6p;
469 sin6p = (struct sockaddr_in6 *)nam;
470 if (nam->sa_len != sizeof (*sin6p))
473 * Must disallow TCP ``connections'' to multicast addresses.
475 if (sin6p->sin6_family == AF_INET6
476 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
477 return (EAFNOSUPPORT);
479 INP_INFO_WLOCK(&V_tcbinfo);
481 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
483 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
489 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
490 struct sockaddr_in sin;
492 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
497 in6_sin6_2_sin(&sin, sin6p);
498 inp->inp_vflag |= INP_IPV4;
499 inp->inp_vflag &= ~INP_IPV6;
500 if ((error = prison_remote_ip4(td->td_ucred,
501 &sin.sin_addr)) != 0)
503 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
505 error = tcp_output_connect(so, nam);
508 inp->inp_vflag &= ~INP_IPV4;
509 inp->inp_vflag |= INP_IPV6;
510 inp->inp_inc.inc_flags |= INC_ISIPV6;
511 if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
513 if ((error = tcp6_connect(tp, nam, td)) != 0)
515 error = tcp_output_connect(so, nam);
518 TCPDEBUG2(PRU_CONNECT);
520 INP_INFO_WUNLOCK(&V_tcbinfo);
526 * Initiate disconnect from peer.
527 * If connection never passed embryonic stage, just drop;
528 * else if don't need to let data drain, then can just drop anyways,
529 * else have to begin TCP shutdown process: mark socket disconnecting,
530 * drain unread data, state switch to reflect user close, and
531 * send segment (e.g. FIN) to peer. Socket will be really disconnected
532 * when peer sends FIN and acks ours.
534 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
537 tcp_usr_disconnect(struct socket *so)
540 struct tcpcb *tp = NULL;
544 INP_INFO_WLOCK(&V_tcbinfo);
546 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
548 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
556 TCPDEBUG2(PRU_DISCONNECT);
558 INP_INFO_WUNLOCK(&V_tcbinfo);
563 * Accept a connection. Essentially all the work is done at higher levels;
564 * just return the address of the peer, storing through addr.
566 * The rationale for acquiring the tcbinfo lock here is somewhat complicated,
567 * and is described in detail in the commit log entry for r175612. Acquiring
568 * it delays an accept(2) racing with sonewconn(), which inserts the socket
569 * before the inpcb address/port fields are initialized. A better fix would
570 * prevent the socket from being placed in the listen queue until all fields
571 * are fully initialized.
574 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
577 struct inpcb *inp = NULL;
578 struct tcpcb *tp = NULL;
583 if (so->so_state & SS_ISDISCONNECTED)
584 return (ECONNABORTED);
587 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
588 INP_INFO_RLOCK(&V_tcbinfo);
590 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
591 error = ECONNABORTED;
598 * We inline in_getpeeraddr and COMMON_END here, so that we can
599 * copy the data of interest and defer the malloc until after we
602 port = inp->inp_fport;
603 addr = inp->inp_faddr;
606 TCPDEBUG2(PRU_ACCEPT);
608 INP_INFO_RUNLOCK(&V_tcbinfo);
610 *nam = in_sockaddr(port, &addr);
616 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
618 struct inpcb *inp = NULL;
620 struct tcpcb *tp = NULL;
622 struct in6_addr addr6;
627 if (so->so_state & SS_ISDISCONNECTED)
628 return (ECONNABORTED);
631 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
633 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
634 error = ECONNABORTED;
641 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
642 * copy the data of interest and defer the malloc until after we
645 if (inp->inp_vflag & INP_IPV4) {
647 port = inp->inp_fport;
648 addr = inp->inp_faddr;
650 port = inp->inp_fport;
651 addr6 = inp->in6p_faddr;
655 TCPDEBUG2(PRU_ACCEPT);
659 *nam = in6_v4mapsin6_sockaddr(port, &addr);
661 *nam = in6_sockaddr(port, &addr6);
668 * Mark the connection as being incapable of further output.
671 tcp_usr_shutdown(struct socket *so)
675 struct tcpcb *tp = NULL;
678 INP_INFO_WLOCK(&V_tcbinfo);
680 KASSERT(inp != NULL, ("inp == NULL"));
682 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
690 if (!(inp->inp_flags & INP_DROPPED))
691 error = tcp_output_disconnect(tp);
694 TCPDEBUG2(PRU_SHUTDOWN);
696 INP_INFO_WUNLOCK(&V_tcbinfo);
702 * After a receive, possibly send window update to peer.
705 tcp_usr_rcvd(struct socket *so, int flags)
708 struct tcpcb *tp = NULL;
713 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
715 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
730 * Do a send by putting data in output queue and updating urgent
731 * marker if URG set. Possibly send more data. Unlike the other
732 * pru_*() routines, the mbuf chains are our responsibility. We
733 * must either enqueue them or free them. The other pru_* routines
734 * generally are caller-frees.
737 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
738 struct sockaddr *nam, struct mbuf *control, struct thread *td)
742 struct tcpcb *tp = NULL;
750 * We require the pcbinfo lock in two cases:
752 * (1) An implied connect is taking place, which can result in
753 * binding IPs and ports and hence modification of the pcb hash
756 * (2) PRUS_EOF is set, resulting in explicit close on the send.
758 if ((nam != NULL) || (flags & PRUS_EOF)) {
759 INP_INFO_WLOCK(&V_tcbinfo);
763 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
765 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
774 isipv6 = nam && nam->sa_family == AF_INET6;
779 /* TCP doesn't do control messages (rights, creds, etc) */
780 if (control->m_len) {
787 m_freem(control); /* empty control, just free it */
789 if (!(flags & PRUS_OOB)) {
790 sbappendstream(&so->so_snd, m);
791 if (nam && tp->t_state < TCPS_SYN_SENT) {
793 * Do implied connect if not yet connected,
794 * initialize window to default value, and
795 * initialize maxseg/maxopd using peer's cached
798 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
801 error = tcp6_connect(tp, nam, td);
804 error = tcp_connect(tp, nam, td);
807 tp->snd_wnd = TTCP_CLIENT_SND_WND;
810 if (flags & PRUS_EOF) {
812 * Close the send side of the connection after
815 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
820 INP_INFO_WUNLOCK(&V_tcbinfo);
823 if (!(inp->inp_flags & INP_DROPPED)) {
824 if (flags & PRUS_MORETOCOME)
825 tp->t_flags |= TF_MORETOCOME;
826 error = tcp_output_send(tp);
827 if (flags & PRUS_MORETOCOME)
828 tp->t_flags &= ~TF_MORETOCOME;
832 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
834 SOCKBUF_LOCK(&so->so_snd);
835 if (sbspace(&so->so_snd) < -512) {
836 SOCKBUF_UNLOCK(&so->so_snd);
842 * According to RFC961 (Assigned Protocols),
843 * the urgent pointer points to the last octet
844 * of urgent data. We continue, however,
845 * to consider it to indicate the first octet
846 * of data past the urgent section.
847 * Otherwise, snd_up should be one lower.
849 sbappendstream_locked(&so->so_snd, m);
850 SOCKBUF_UNLOCK(&so->so_snd);
851 if (nam && tp->t_state < TCPS_SYN_SENT) {
853 * Do implied connect if not yet connected,
854 * initialize window to default value, and
855 * initialize maxseg/maxopd using peer's cached
858 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
861 error = tcp6_connect(tp, nam, td);
864 error = tcp_connect(tp, nam, td);
867 tp->snd_wnd = TTCP_CLIENT_SND_WND;
869 INP_INFO_WUNLOCK(&V_tcbinfo);
872 INP_INFO_WUNLOCK(&V_tcbinfo);
875 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
876 tp->t_flags |= TF_FORCEDATA;
877 error = tcp_output_send(tp);
878 tp->t_flags &= ~TF_FORCEDATA;
881 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
882 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
885 INP_INFO_WUNLOCK(&V_tcbinfo);
890 * Abort the TCP. Drop the connection abruptly.
893 tcp_usr_abort(struct socket *so)
896 struct tcpcb *tp = NULL;
900 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
902 INP_INFO_WLOCK(&V_tcbinfo);
904 KASSERT(inp->inp_socket != NULL,
905 ("tcp_usr_abort: inp_socket == NULL"));
908 * If we still have full TCP state, and we're not dropped, drop.
910 if (!(inp->inp_flags & INP_TIMEWAIT) &&
911 !(inp->inp_flags & INP_DROPPED)) {
914 tcp_drop(tp, ECONNABORTED);
915 TCPDEBUG2(PRU_ABORT);
917 if (!(inp->inp_flags & INP_DROPPED)) {
919 so->so_state |= SS_PROTOREF;
921 inp->inp_flags |= INP_SOCKREF;
924 INP_INFO_WUNLOCK(&V_tcbinfo);
928 * TCP socket is closed. Start friendly disconnect.
931 tcp_usr_close(struct socket *so)
934 struct tcpcb *tp = NULL;
938 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
940 INP_INFO_WLOCK(&V_tcbinfo);
942 KASSERT(inp->inp_socket != NULL,
943 ("tcp_usr_close: inp_socket == NULL"));
946 * If we still have full TCP state, and we're not dropped, initiate
949 if (!(inp->inp_flags & INP_TIMEWAIT) &&
950 !(inp->inp_flags & INP_DROPPED)) {
954 TCPDEBUG2(PRU_CLOSE);
956 if (!(inp->inp_flags & INP_DROPPED)) {
958 so->so_state |= SS_PROTOREF;
960 inp->inp_flags |= INP_SOCKREF;
963 INP_INFO_WUNLOCK(&V_tcbinfo);
967 * Receive out-of-band data.
970 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
974 struct tcpcb *tp = NULL;
978 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
980 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
986 if ((so->so_oobmark == 0 &&
987 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
988 so->so_options & SO_OOBINLINE ||
989 tp->t_oobflags & TCPOOB_HADDATA) {
993 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
998 *mtod(m, caddr_t) = tp->t_iobc;
999 if ((flags & MSG_PEEK) == 0)
1000 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1003 TCPDEBUG2(PRU_RCVOOB);
1008 struct pr_usrreqs tcp_usrreqs = {
1009 .pru_abort = tcp_usr_abort,
1010 .pru_accept = tcp_usr_accept,
1011 .pru_attach = tcp_usr_attach,
1012 .pru_bind = tcp_usr_bind,
1013 .pru_connect = tcp_usr_connect,
1014 .pru_control = in_control,
1015 .pru_detach = tcp_usr_detach,
1016 .pru_disconnect = tcp_usr_disconnect,
1017 .pru_listen = tcp_usr_listen,
1018 .pru_peeraddr = in_getpeeraddr,
1019 .pru_rcvd = tcp_usr_rcvd,
1020 .pru_rcvoob = tcp_usr_rcvoob,
1021 .pru_send = tcp_usr_send,
1022 .pru_shutdown = tcp_usr_shutdown,
1023 .pru_sockaddr = in_getsockaddr,
1024 .pru_sosetlabel = in_pcbsosetlabel,
1025 .pru_close = tcp_usr_close,
1029 struct pr_usrreqs tcp6_usrreqs = {
1030 .pru_abort = tcp_usr_abort,
1031 .pru_accept = tcp6_usr_accept,
1032 .pru_attach = tcp_usr_attach,
1033 .pru_bind = tcp6_usr_bind,
1034 .pru_connect = tcp6_usr_connect,
1035 .pru_control = in6_control,
1036 .pru_detach = tcp_usr_detach,
1037 .pru_disconnect = tcp_usr_disconnect,
1038 .pru_listen = tcp6_usr_listen,
1039 .pru_peeraddr = in6_mapped_peeraddr,
1040 .pru_rcvd = tcp_usr_rcvd,
1041 .pru_rcvoob = tcp_usr_rcvoob,
1042 .pru_send = tcp_usr_send,
1043 .pru_shutdown = tcp_usr_shutdown,
1044 .pru_sockaddr = in6_mapped_sockaddr,
1045 .pru_sosetlabel = in_pcbsosetlabel,
1046 .pru_close = tcp_usr_close,
1051 * Common subroutine to open a TCP connection to remote host specified
1052 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
1053 * port number if needed. Call in_pcbconnect_setup to do the routing and
1054 * to choose a local host address (interface). If there is an existing
1055 * incarnation of the same connection in TIME-WAIT state and if the remote
1056 * host was sending CC options and if the connection duration was < MSL, then
1057 * truncate the previous TIME-WAIT state and proceed.
1058 * Initialize connection parameters and enter SYN-SENT state.
1061 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1063 struct inpcb *inp = tp->t_inpcb, *oinp;
1064 struct socket *so = inp->inp_socket;
1065 struct in_addr laddr;
1069 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1070 INP_WLOCK_ASSERT(inp);
1072 if (inp->inp_lport == 0) {
1073 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1079 * Cannot simply call in_pcbconnect, because there might be an
1080 * earlier incarnation of this same connection still in
1081 * TIME_WAIT state, creating an ADDRINUSE error.
1083 laddr = inp->inp_laddr;
1084 lport = inp->inp_lport;
1085 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1086 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1087 if (error && oinp == NULL)
1091 inp->inp_laddr = laddr;
1095 * Compute window scaling to request:
1096 * Scale to fit into sweet spot. See tcp_syncache.c.
1097 * XXX: This should move to tcp_output().
1099 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1100 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1101 tp->request_r_scale++;
1104 TCPSTAT_INC(tcps_connattempt);
1105 tp->t_state = TCPS_SYN_SENT;
1106 tcp_timer_activate(tp, TT_KEEP, tcp_keepinit);
1107 tp->iss = tcp_new_isn(tp);
1108 tcp_sendseqinit(tp);
1115 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1117 struct inpcb *inp = tp->t_inpcb, *oinp;
1118 struct socket *so = inp->inp_socket;
1119 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
1120 struct in6_addr addr6;
1123 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1124 INP_WLOCK_ASSERT(inp);
1126 if (inp->inp_lport == 0) {
1127 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1133 * Cannot simply call in_pcbconnect, because there might be an
1134 * earlier incarnation of this same connection still in
1135 * TIME_WAIT state, creating an ADDRINUSE error.
1136 * in6_pcbladdr() also handles scope zone IDs.
1138 error = in6_pcbladdr(inp, nam, &addr6);
1141 oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
1142 &sin6->sin6_addr, sin6->sin6_port,
1143 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
1146 inp->inp_lport, 0, NULL);
1149 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1150 inp->in6p_laddr = addr6;
1151 inp->in6p_faddr = sin6->sin6_addr;
1152 inp->inp_fport = sin6->sin6_port;
1153 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
1154 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
1155 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
1157 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
1160 /* Compute window scaling to request. */
1161 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1162 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1163 tp->request_r_scale++;
1166 TCPSTAT_INC(tcps_connattempt);
1167 tp->t_state = TCPS_SYN_SENT;
1168 tcp_timer_activate(tp, TT_KEEP, tcp_keepinit);
1169 tp->iss = tcp_new_isn(tp);
1170 tcp_sendseqinit(tp);
1177 * Export TCP internal state information via a struct tcp_info, based on the
1178 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently
1179 * (TCP state machine, etc). We export all information using FreeBSD-native
1180 * constants -- for example, the numeric values for tcpi_state will differ
1184 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1187 INP_WLOCK_ASSERT(tp->t_inpcb);
1188 bzero(ti, sizeof(*ti));
1190 ti->tcpi_state = tp->t_state;
1191 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1192 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1193 if (tp->t_flags & TF_SACK_PERMIT)
1194 ti->tcpi_options |= TCPI_OPT_SACK;
1195 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1196 ti->tcpi_options |= TCPI_OPT_WSCALE;
1197 ti->tcpi_snd_wscale = tp->snd_scale;
1198 ti->tcpi_rcv_wscale = tp->rcv_scale;
1201 ti->tcpi_rto = tp->t_rxtcur * tick;
1202 ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick;
1203 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1204 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1206 ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1207 ti->tcpi_snd_cwnd = tp->snd_cwnd;
1210 * FreeBSD-specific extension fields for tcp_info.
1212 ti->tcpi_rcv_space = tp->rcv_wnd;
1213 ti->tcpi_rcv_nxt = tp->rcv_nxt;
1214 ti->tcpi_snd_wnd = tp->snd_wnd;
1215 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */
1216 ti->tcpi_snd_nxt = tp->snd_nxt;
1217 ti->tcpi_snd_mss = tp->t_maxseg;
1218 ti->tcpi_rcv_mss = tp->t_maxseg;
1219 if (tp->t_flags & TF_TOE)
1220 ti->tcpi_options |= TCPI_OPT_TOE;
1221 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1222 ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1223 ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1227 * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1228 * socket option arguments. When it re-acquires the lock after the copy, it
1229 * has to revalidate that the connection is still valid for the socket
1232 #define INP_WLOCK_RECHECK(inp) do { \
1234 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
1236 return (ECONNRESET); \
1238 tp = intotcpcb(inp); \
1242 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1244 int error, opt, optval;
1248 char buf[TCP_CA_NAME_MAX];
1249 struct cc_algo *algo;
1252 inp = sotoinpcb(so);
1253 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1255 if (sopt->sopt_level != IPPROTO_TCP) {
1257 if (inp->inp_vflag & INP_IPV6PROTO) {
1259 error = ip6_ctloutput(so, sopt);
1263 error = ip_ctloutput(so, sopt);
1269 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1271 return (ECONNRESET);
1274 switch (sopt->sopt_dir) {
1276 switch (sopt->sopt_name) {
1277 #ifdef TCP_SIGNATURE
1280 error = sooptcopyin(sopt, &optval, sizeof optval,
1285 INP_WLOCK_RECHECK(inp);
1287 tp->t_flags |= TF_SIGNATURE;
1289 tp->t_flags &= ~TF_SIGNATURE;
1292 #endif /* TCP_SIGNATURE */
1296 error = sooptcopyin(sopt, &optval, sizeof optval,
1301 INP_WLOCK_RECHECK(inp);
1302 switch (sopt->sopt_name) {
1310 opt = 0; /* dead code to fool gcc */
1317 tp->t_flags &= ~opt;
1323 error = sooptcopyin(sopt, &optval, sizeof optval,
1328 INP_WLOCK_RECHECK(inp);
1330 tp->t_flags |= TF_NOPUSH;
1332 tp->t_flags &= ~TF_NOPUSH;
1333 error = tcp_output(tp);
1340 error = sooptcopyin(sopt, &optval, sizeof optval,
1345 INP_WLOCK_RECHECK(inp);
1346 if (optval > 0 && optval <= tp->t_maxseg &&
1347 optval + 40 >= V_tcp_minmss)
1348 tp->t_maxseg = optval;
1359 case TCP_CONGESTION:
1361 bzero(buf, sizeof(buf));
1362 error = sooptcopyin(sopt, &buf, sizeof(buf), 1);
1365 INP_WLOCK_RECHECK(inp);
1367 * Return EINVAL if we can't find the requested cc algo.
1371 STAILQ_FOREACH(algo, &cc_list, entries) {
1372 if (strncmp(buf, algo->name, TCP_CA_NAME_MAX)
1374 /* We've found the requested algo. */
1377 * We hold a write lock over the tcb
1378 * so it's safe to do these things
1379 * without ordering concerns.
1381 if (CC_ALGO(tp)->cb_destroy != NULL)
1382 CC_ALGO(tp)->cb_destroy(tp->ccv);
1385 * If something goes pear shaped
1386 * initialising the new algo,
1387 * fall back to newreno (which
1388 * does not require initialisation).
1390 if (algo->cb_init != NULL)
1391 if (algo->cb_init(tp->ccv) > 0) {
1392 CC_ALGO(tp) = &newreno_cc_algo;
1394 * The only reason init
1396 * because of malloc.
1400 break; /* Break the STAILQ_FOREACH. */
1409 error = ENOPROTOOPT;
1415 tp = intotcpcb(inp);
1416 switch (sopt->sopt_name) {
1417 #ifdef TCP_SIGNATURE
1419 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1421 error = sooptcopyout(sopt, &optval, sizeof optval);
1426 optval = tp->t_flags & TF_NODELAY;
1428 error = sooptcopyout(sopt, &optval, sizeof optval);
1431 optval = tp->t_maxseg;
1433 error = sooptcopyout(sopt, &optval, sizeof optval);
1436 optval = tp->t_flags & TF_NOOPT;
1438 error = sooptcopyout(sopt, &optval, sizeof optval);
1441 optval = tp->t_flags & TF_NOPUSH;
1443 error = sooptcopyout(sopt, &optval, sizeof optval);
1446 tcp_fill_info(tp, &ti);
1448 error = sooptcopyout(sopt, &ti, sizeof ti);
1450 case TCP_CONGESTION:
1451 bzero(buf, sizeof(buf));
1452 strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1454 error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX);
1458 error = ENOPROTOOPT;
1465 #undef INP_WLOCK_RECHECK
1468 * tcp_sendspace and tcp_recvspace are the default send and receive window
1469 * sizes, respectively. These are obsolescent (this information should
1470 * be set by the route).
1472 u_long tcp_sendspace = 1024*32;
1473 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1474 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1475 u_long tcp_recvspace = 1024*64;
1476 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1477 &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1480 * Attach TCP protocol to socket, allocating
1481 * internet protocol control block, tcp control block,
1482 * bufer space, and entering LISTEN state if to accept connections.
1485 tcp_attach(struct socket *so)
1491 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1492 error = soreserve(so, tcp_sendspace, tcp_recvspace);
1496 so->so_rcv.sb_flags |= SB_AUTOSIZE;
1497 so->so_snd.sb_flags |= SB_AUTOSIZE;
1498 INP_INFO_WLOCK(&V_tcbinfo);
1499 error = in_pcballoc(so, &V_tcbinfo);
1501 INP_INFO_WUNLOCK(&V_tcbinfo);
1504 inp = sotoinpcb(so);
1506 if (inp->inp_vflag & INP_IPV6PROTO) {
1507 inp->inp_vflag |= INP_IPV6;
1508 inp->in6p_hops = -1; /* use kernel default */
1512 inp->inp_vflag |= INP_IPV4;
1513 tp = tcp_newtcpcb(inp);
1517 INP_INFO_WUNLOCK(&V_tcbinfo);
1520 tp->t_state = TCPS_CLOSED;
1522 INP_INFO_WUNLOCK(&V_tcbinfo);
1527 * Initiate (or continue) disconnect.
1528 * If embryonic state, just send reset (once).
1529 * If in ``let data drain'' option and linger null, just drop.
1530 * Otherwise (hard), mark socket disconnecting and drop
1531 * current input data; switch states based on user close, and
1532 * send segment to peer (with FIN).
1535 tcp_disconnect(struct tcpcb *tp)
1537 struct inpcb *inp = tp->t_inpcb;
1538 struct socket *so = inp->inp_socket;
1540 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1541 INP_WLOCK_ASSERT(inp);
1544 * Neither tcp_close() nor tcp_drop() should return NULL, as the
1545 * socket is still open.
1547 if (tp->t_state < TCPS_ESTABLISHED) {
1550 ("tcp_disconnect: tcp_close() returned NULL"));
1551 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1552 tp = tcp_drop(tp, 0);
1554 ("tcp_disconnect: tcp_drop() returned NULL"));
1556 soisdisconnecting(so);
1557 sbflush(&so->so_rcv);
1559 if (!(inp->inp_flags & INP_DROPPED))
1560 tcp_output_disconnect(tp);
1565 * User issued close, and wish to trail through shutdown states:
1566 * if never received SYN, just forget it. If got a SYN from peer,
1567 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1568 * If already got a FIN from peer, then almost done; go to LAST_ACK
1569 * state. In all other cases, have already sent FIN to peer (e.g.
1570 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1571 * for peer to send FIN or not respond to keep-alives, etc.
1572 * We can let the user exit from the close as soon as the FIN is acked.
1575 tcp_usrclosed(struct tcpcb *tp)
1578 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1579 INP_WLOCK_ASSERT(tp->t_inpcb);
1581 switch (tp->t_state) {
1583 tcp_offload_listen_close(tp);
1586 tp->t_state = TCPS_CLOSED;
1589 * tcp_close() should never return NULL here as the socket is
1593 ("tcp_usrclosed: tcp_close() returned NULL"));
1597 case TCPS_SYN_RECEIVED:
1598 tp->t_flags |= TF_NEEDFIN;
1601 case TCPS_ESTABLISHED:
1602 tp->t_state = TCPS_FIN_WAIT_1;
1605 case TCPS_CLOSE_WAIT:
1606 tp->t_state = TCPS_LAST_ACK;
1609 if (tp->t_state >= TCPS_FIN_WAIT_2) {
1610 soisdisconnected(tp->t_inpcb->inp_socket);
1611 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
1612 if (tp->t_state == TCPS_FIN_WAIT_2) {
1615 timeout = (tcp_fast_finwait2_recycle) ?
1616 tcp_finwait2_timeout : tcp_maxidle;
1617 tcp_timer_activate(tp, TT_2MSL, timeout);
1624 db_print_indent(int indent)
1628 for (i = 0; i < indent; i++)
1633 db_print_tstate(int t_state)
1638 db_printf("TCPS_CLOSED");
1642 db_printf("TCPS_LISTEN");
1646 db_printf("TCPS_SYN_SENT");
1649 case TCPS_SYN_RECEIVED:
1650 db_printf("TCPS_SYN_RECEIVED");
1653 case TCPS_ESTABLISHED:
1654 db_printf("TCPS_ESTABLISHED");
1657 case TCPS_CLOSE_WAIT:
1658 db_printf("TCPS_CLOSE_WAIT");
1661 case TCPS_FIN_WAIT_1:
1662 db_printf("TCPS_FIN_WAIT_1");
1666 db_printf("TCPS_CLOSING");
1670 db_printf("TCPS_LAST_ACK");
1673 case TCPS_FIN_WAIT_2:
1674 db_printf("TCPS_FIN_WAIT_2");
1677 case TCPS_TIME_WAIT:
1678 db_printf("TCPS_TIME_WAIT");
1682 db_printf("unknown");
1688 db_print_tflags(u_int t_flags)
1693 if (t_flags & TF_ACKNOW) {
1694 db_printf("%sTF_ACKNOW", comma ? ", " : "");
1697 if (t_flags & TF_DELACK) {
1698 db_printf("%sTF_DELACK", comma ? ", " : "");
1701 if (t_flags & TF_NODELAY) {
1702 db_printf("%sTF_NODELAY", comma ? ", " : "");
1705 if (t_flags & TF_NOOPT) {
1706 db_printf("%sTF_NOOPT", comma ? ", " : "");
1709 if (t_flags & TF_SENTFIN) {
1710 db_printf("%sTF_SENTFIN", comma ? ", " : "");
1713 if (t_flags & TF_REQ_SCALE) {
1714 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
1717 if (t_flags & TF_RCVD_SCALE) {
1718 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
1721 if (t_flags & TF_REQ_TSTMP) {
1722 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
1725 if (t_flags & TF_RCVD_TSTMP) {
1726 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
1729 if (t_flags & TF_SACK_PERMIT) {
1730 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
1733 if (t_flags & TF_NEEDSYN) {
1734 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
1737 if (t_flags & TF_NEEDFIN) {
1738 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
1741 if (t_flags & TF_NOPUSH) {
1742 db_printf("%sTF_NOPUSH", comma ? ", " : "");
1745 if (t_flags & TF_NOPUSH) {
1746 db_printf("%sTF_NOPUSH", comma ? ", " : "");
1749 if (t_flags & TF_MORETOCOME) {
1750 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
1753 if (t_flags & TF_LQ_OVERFLOW) {
1754 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
1757 if (t_flags & TF_LASTIDLE) {
1758 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
1761 if (t_flags & TF_RXWIN0SENT) {
1762 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
1765 if (t_flags & TF_FASTRECOVERY) {
1766 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
1769 if (t_flags & TF_CONGRECOVERY) {
1770 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
1773 if (t_flags & TF_WASFRECOVERY) {
1774 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
1777 if (t_flags & TF_SIGNATURE) {
1778 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
1781 if (t_flags & TF_FORCEDATA) {
1782 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
1785 if (t_flags & TF_TSO) {
1786 db_printf("%sTF_TSO", comma ? ", " : "");
1789 if (t_flags & TF_ECN_PERMIT) {
1790 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
1796 db_print_toobflags(char t_oobflags)
1801 if (t_oobflags & TCPOOB_HAVEDATA) {
1802 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
1805 if (t_oobflags & TCPOOB_HADDATA) {
1806 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
1812 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
1815 db_print_indent(indent);
1816 db_printf("%s at %p\n", name, tp);
1820 db_print_indent(indent);
1821 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n",
1822 LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
1824 db_print_indent(indent);
1825 db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n",
1826 &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
1828 db_print_indent(indent);
1829 db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl,
1830 &tp->t_timers->tt_delack, tp->t_inpcb);
1832 db_print_indent(indent);
1833 db_printf("t_state: %d (", tp->t_state);
1834 db_print_tstate(tp->t_state);
1837 db_print_indent(indent);
1838 db_printf("t_flags: 0x%x (", tp->t_flags);
1839 db_print_tflags(tp->t_flags);
1842 db_print_indent(indent);
1843 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n",
1844 tp->snd_una, tp->snd_max, tp->snd_nxt);
1846 db_print_indent(indent);
1847 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n",
1848 tp->snd_up, tp->snd_wl1, tp->snd_wl2);
1850 db_print_indent(indent);
1851 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n",
1852 tp->iss, tp->irs, tp->rcv_nxt);
1854 db_print_indent(indent);
1855 db_printf("rcv_adv: 0x%08x rcv_wnd: %lu rcv_up: 0x%08x\n",
1856 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
1858 db_print_indent(indent);
1859 db_printf("snd_wnd: %lu snd_cwnd: %lu\n",
1860 tp->snd_wnd, tp->snd_cwnd);
1862 db_print_indent(indent);
1863 db_printf("snd_ssthresh: %lu snd_recover: "
1864 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
1866 db_print_indent(indent);
1867 db_printf("t_maxopd: %u t_rcvtime: %u t_startime: %u\n",
1868 tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
1870 db_print_indent(indent);
1871 db_printf("t_rttime: %u t_rtsq: 0x%08x\n",
1872 tp->t_rtttime, tp->t_rtseq);
1874 db_print_indent(indent);
1875 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n",
1876 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
1878 db_print_indent(indent);
1879 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u "
1880 "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
1883 db_print_indent(indent);
1884 db_printf("t_rttupdated: %lu max_sndwnd: %lu t_softerror: %d\n",
1885 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
1887 db_print_indent(indent);
1888 db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
1889 db_print_toobflags(tp->t_oobflags);
1890 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc);
1892 db_print_indent(indent);
1893 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n",
1894 tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
1896 db_print_indent(indent);
1897 db_printf("ts_recent: %u ts_recent_age: %u\n",
1898 tp->ts_recent, tp->ts_recent_age);
1900 db_print_indent(indent);
1901 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: "
1902 "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
1904 db_print_indent(indent);
1905 db_printf("snd_ssthresh_prev: %lu snd_recover_prev: 0x%08x "
1906 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
1907 tp->snd_recover_prev, tp->t_badrxtwin);
1909 db_print_indent(indent);
1910 db_printf("snd_numholes: %d snd_holes first: %p\n",
1911 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
1913 db_print_indent(indent);
1914 db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: "
1915 "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
1917 /* Skip sackblks, sackhint. */
1919 db_print_indent(indent);
1920 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n",
1921 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
1924 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
1929 db_printf("usage: show tcpcb <addr>\n");
1932 tp = (struct tcpcb *)addr;
1934 db_print_tcpcb(tp, "tcpcb", 0);