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/route.h>
70 #include <netinet/cc.h>
71 #include <netinet/in.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip_var.h>
77 #include <netinet/ip6.h>
78 #include <netinet6/in6_pcb.h>
79 #include <netinet6/ip6_var.h>
80 #include <netinet6/scope6_var.h>
82 #include <netinet/tcp_fsm.h>
83 #include <netinet/tcp_seq.h>
84 #include <netinet/tcp_timer.h>
85 #include <netinet/tcp_var.h>
86 #include <netinet/tcpip.h>
88 #include <netinet/tcp_debug.h>
90 #include <netinet/tcp_offload.h>
93 * TCP protocol interface to socket abstraction.
95 static int tcp_attach(struct socket *);
97 static int tcp_connect(struct tcpcb *, struct sockaddr *,
101 static int tcp6_connect(struct tcpcb *, struct sockaddr *,
104 static void tcp_disconnect(struct tcpcb *);
105 static void tcp_usrclosed(struct tcpcb *);
106 static void tcp_fill_info(struct tcpcb *, struct tcp_info *);
109 #define TCPDEBUG0 int ostate = 0
110 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0
111 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
112 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
116 #define TCPDEBUG2(req)
120 * TCP attaches to socket via pru_attach(), reserving space,
121 * and an internet control block.
124 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
127 struct tcpcb *tp = NULL;
132 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
135 error = tcp_attach(so);
139 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
140 so->so_linger = TCP_LINGERTIME;
145 TCPDEBUG2(PRU_ATTACH);
150 * tcp_detach is called when the socket layer loses its final reference
151 * to the socket, be it a file descriptor reference, a reference from TCP,
152 * etc. At this point, there is only one case in which we will keep around
153 * inpcb state: time wait.
155 * This function can probably be re-absorbed back into tcp_usr_detach() now
156 * that there is a single detach path.
159 tcp_detach(struct socket *so, struct inpcb *inp)
163 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
164 INP_WLOCK_ASSERT(inp);
166 KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
167 KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
171 if (inp->inp_flags & INP_TIMEWAIT) {
173 * There are two cases to handle: one in which the time wait
174 * state is being discarded (INP_DROPPED), and one in which
175 * this connection will remain in timewait. In the former,
176 * it is time to discard all state (except tcptw, which has
177 * already been discarded by the timewait close code, which
178 * should be further up the call stack somewhere). In the
179 * latter case, we detach from the socket, but leave the pcb
180 * present until timewait ends.
182 * XXXRW: Would it be cleaner to free the tcptw here?
184 if (inp->inp_flags & INP_DROPPED) {
185 KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
186 "INP_DROPPED && tp != NULL"));
195 * If the connection is not in timewait, we consider two
196 * two conditions: one in which no further processing is
197 * necessary (dropped || embryonic), and one in which TCP is
198 * not yet done, but no longer requires the socket, so the
199 * pcb will persist for the time being.
201 * XXXRW: Does the second case still occur?
203 if (inp->inp_flags & INP_DROPPED ||
204 tp->t_state < TCPS_SYN_SENT) {
216 * pru_detach() detaches the TCP protocol from the socket.
217 * If the protocol state is non-embryonic, then can't
218 * do this directly: have to initiate a pru_disconnect(),
219 * which may finish later; embryonic TCB's can just
223 tcp_usr_detach(struct socket *so)
228 KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
229 INP_INFO_WLOCK(&V_tcbinfo);
231 KASSERT(inp->inp_socket != NULL,
232 ("tcp_usr_detach: inp_socket == NULL"));
234 INP_INFO_WUNLOCK(&V_tcbinfo);
239 * Give the socket an address.
242 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
246 struct tcpcb *tp = NULL;
247 struct sockaddr_in *sinp;
249 sinp = (struct sockaddr_in *)nam;
250 if (nam->sa_len != sizeof (*sinp))
253 * Must check for multicast addresses and disallow binding
256 if (sinp->sin_family == AF_INET &&
257 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
258 return (EAFNOSUPPORT);
262 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
264 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
270 INP_HASH_WLOCK(&V_tcbinfo);
271 error = in_pcbbind(inp, nam, td->td_ucred);
272 INP_HASH_WUNLOCK(&V_tcbinfo);
283 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
287 struct tcpcb *tp = NULL;
288 struct sockaddr_in6 *sin6p;
290 sin6p = (struct sockaddr_in6 *)nam;
291 if (nam->sa_len != sizeof (*sin6p))
294 * Must check for multicast addresses and disallow binding
297 if (sin6p->sin6_family == AF_INET6 &&
298 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
299 return (EAFNOSUPPORT);
303 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
305 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
311 INP_HASH_WLOCK(&V_tcbinfo);
312 inp->inp_vflag &= ~INP_IPV4;
313 inp->inp_vflag |= INP_IPV6;
315 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
316 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
317 inp->inp_vflag |= INP_IPV4;
318 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
319 struct sockaddr_in sin;
321 in6_sin6_2_sin(&sin, sin6p);
322 inp->inp_vflag |= INP_IPV4;
323 inp->inp_vflag &= ~INP_IPV6;
324 error = in_pcbbind(inp, (struct sockaddr *)&sin,
326 INP_HASH_WUNLOCK(&V_tcbinfo);
331 error = in6_pcbbind(inp, nam, td->td_ucred);
332 INP_HASH_WUNLOCK(&V_tcbinfo);
342 * Prepare to accept connections.
345 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
349 struct tcpcb *tp = NULL;
353 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
355 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
362 error = solisten_proto_check(so);
363 INP_HASH_WLOCK(&V_tcbinfo);
364 if (error == 0 && inp->inp_lport == 0)
365 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
366 INP_HASH_WUNLOCK(&V_tcbinfo);
368 tp->t_state = TCPS_LISTEN;
369 solisten_proto(so, backlog);
370 tcp_offload_listen_open(tp);
375 TCPDEBUG2(PRU_LISTEN);
383 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
387 struct tcpcb *tp = NULL;
391 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
393 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
400 error = solisten_proto_check(so);
401 INP_HASH_WLOCK(&V_tcbinfo);
402 if (error == 0 && inp->inp_lport == 0) {
403 inp->inp_vflag &= ~INP_IPV4;
404 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
405 inp->inp_vflag |= INP_IPV4;
406 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
408 INP_HASH_WUNLOCK(&V_tcbinfo);
410 tp->t_state = TCPS_LISTEN;
411 solisten_proto(so, backlog);
416 TCPDEBUG2(PRU_LISTEN);
424 * Initiate connection to peer.
425 * Create a template for use in transmissions on this connection.
426 * Enter SYN_SENT state, and mark socket as connecting.
427 * Start keep-alive timer, and seed output sequence space.
428 * Send initial segment on connection.
431 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
435 struct tcpcb *tp = NULL;
436 struct sockaddr_in *sinp;
438 sinp = (struct sockaddr_in *)nam;
439 if (nam->sa_len != sizeof (*sinp))
442 * Must disallow TCP ``connections'' to multicast addresses.
444 if (sinp->sin_family == AF_INET
445 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
446 return (EAFNOSUPPORT);
447 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
452 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
454 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
460 if ((error = tcp_connect(tp, nam, td)) != 0)
462 error = tcp_output_connect(so, nam);
464 TCPDEBUG2(PRU_CONNECT);
472 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
476 struct tcpcb *tp = NULL;
477 struct sockaddr_in6 *sin6p;
481 sin6p = (struct sockaddr_in6 *)nam;
482 if (nam->sa_len != sizeof (*sin6p))
485 * Must disallow TCP ``connections'' to multicast addresses.
487 if (sin6p->sin6_family == AF_INET6
488 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
489 return (EAFNOSUPPORT);
492 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
494 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
502 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
503 * therefore probably require the hash lock, which isn't held here.
504 * Is this a significant problem?
506 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
507 struct sockaddr_in sin;
509 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
514 in6_sin6_2_sin(&sin, sin6p);
515 inp->inp_vflag |= INP_IPV4;
516 inp->inp_vflag &= ~INP_IPV6;
517 if ((error = prison_remote_ip4(td->td_ucred,
518 &sin.sin_addr)) != 0)
520 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
522 error = tcp_output_connect(so, nam);
526 inp->inp_vflag &= ~INP_IPV4;
527 inp->inp_vflag |= INP_IPV6;
528 inp->inp_inc.inc_flags |= INC_ISIPV6;
529 if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
531 if ((error = tcp6_connect(tp, nam, td)) != 0)
533 error = tcp_output_connect(so, nam);
536 TCPDEBUG2(PRU_CONNECT);
543 * Initiate disconnect from peer.
544 * If connection never passed embryonic stage, just drop;
545 * else if don't need to let data drain, then can just drop anyways,
546 * else have to begin TCP shutdown process: mark socket disconnecting,
547 * drain unread data, state switch to reflect user close, and
548 * send segment (e.g. FIN) to peer. Socket will be really disconnected
549 * when peer sends FIN and acks ours.
551 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
554 tcp_usr_disconnect(struct socket *so)
557 struct tcpcb *tp = NULL;
561 INP_INFO_WLOCK(&V_tcbinfo);
563 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
565 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
573 TCPDEBUG2(PRU_DISCONNECT);
575 INP_INFO_WUNLOCK(&V_tcbinfo);
581 * Accept a connection. Essentially all the work is done at higher levels;
582 * just return the address of the peer, storing through addr.
584 * The rationale for acquiring the tcbinfo lock here is somewhat complicated,
585 * and is described in detail in the commit log entry for r175612. Acquiring
586 * it delays an accept(2) racing with sonewconn(), which inserts the socket
587 * before the inpcb address/port fields are initialized. A better fix would
588 * prevent the socket from being placed in the listen queue until all fields
589 * are fully initialized.
592 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
595 struct inpcb *inp = NULL;
596 struct tcpcb *tp = NULL;
601 if (so->so_state & SS_ISDISCONNECTED)
602 return (ECONNABORTED);
605 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
606 INP_INFO_RLOCK(&V_tcbinfo);
608 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
609 error = ECONNABORTED;
616 * We inline in_getpeeraddr and COMMON_END here, so that we can
617 * copy the data of interest and defer the malloc until after we
620 port = inp->inp_fport;
621 addr = inp->inp_faddr;
624 TCPDEBUG2(PRU_ACCEPT);
626 INP_INFO_RUNLOCK(&V_tcbinfo);
628 *nam = in_sockaddr(port, &addr);
635 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
637 struct inpcb *inp = NULL;
639 struct tcpcb *tp = NULL;
641 struct in6_addr addr6;
646 if (so->so_state & SS_ISDISCONNECTED)
647 return (ECONNABORTED);
650 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
651 INP_INFO_RLOCK(&V_tcbinfo);
653 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
654 error = ECONNABORTED;
661 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
662 * copy the data of interest and defer the malloc until after we
665 if (inp->inp_vflag & INP_IPV4) {
667 port = inp->inp_fport;
668 addr = inp->inp_faddr;
670 port = inp->inp_fport;
671 addr6 = inp->in6p_faddr;
675 TCPDEBUG2(PRU_ACCEPT);
677 INP_INFO_RUNLOCK(&V_tcbinfo);
680 *nam = in6_v4mapsin6_sockaddr(port, &addr);
682 *nam = in6_sockaddr(port, &addr6);
689 * Mark the connection as being incapable of further output.
692 tcp_usr_shutdown(struct socket *so)
696 struct tcpcb *tp = NULL;
699 INP_INFO_WLOCK(&V_tcbinfo);
701 KASSERT(inp != NULL, ("inp == NULL"));
703 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
711 if (!(inp->inp_flags & INP_DROPPED))
712 error = tcp_output_disconnect(tp);
715 TCPDEBUG2(PRU_SHUTDOWN);
717 INP_INFO_WUNLOCK(&V_tcbinfo);
723 * After a receive, possibly send window update to peer.
726 tcp_usr_rcvd(struct socket *so, int flags)
729 struct tcpcb *tp = NULL;
734 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
736 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
751 * Do a send by putting data in output queue and updating urgent
752 * marker if URG set. Possibly send more data. Unlike the other
753 * pru_*() routines, the mbuf chains are our responsibility. We
754 * must either enqueue them or free them. The other pru_* routines
755 * generally are caller-frees.
758 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
759 struct sockaddr *nam, struct mbuf *control, struct thread *td)
763 struct tcpcb *tp = NULL;
770 * We require the pcbinfo lock if we will close the socket as part of
773 if (flags & PRUS_EOF)
774 INP_INFO_WLOCK(&V_tcbinfo);
776 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
778 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
787 isipv6 = nam && nam->sa_family == AF_INET6;
792 /* TCP doesn't do control messages (rights, creds, etc) */
793 if (control->m_len) {
800 m_freem(control); /* empty control, just free it */
802 if (!(flags & PRUS_OOB)) {
803 sbappendstream(&so->so_snd, m);
804 if (nam && tp->t_state < TCPS_SYN_SENT) {
806 * Do implied connect if not yet connected,
807 * initialize window to default value, and
808 * initialize maxseg/maxopd using peer's cached
813 error = tcp6_connect(tp, nam, td);
815 #if defined(INET6) && defined(INET)
819 error = tcp_connect(tp, nam, td);
823 tp->snd_wnd = TTCP_CLIENT_SND_WND;
826 if (flags & PRUS_EOF) {
828 * Close the send side of the connection after
831 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
835 if (!(inp->inp_flags & INP_DROPPED)) {
836 if (flags & PRUS_MORETOCOME)
837 tp->t_flags |= TF_MORETOCOME;
838 error = tcp_output_send(tp);
839 if (flags & PRUS_MORETOCOME)
840 tp->t_flags &= ~TF_MORETOCOME;
844 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
846 SOCKBUF_LOCK(&so->so_snd);
847 if (sbspace(&so->so_snd) < -512) {
848 SOCKBUF_UNLOCK(&so->so_snd);
854 * According to RFC961 (Assigned Protocols),
855 * the urgent pointer points to the last octet
856 * of urgent data. We continue, however,
857 * to consider it to indicate the first octet
858 * of data past the urgent section.
859 * Otherwise, snd_up should be one lower.
861 sbappendstream_locked(&so->so_snd, m);
862 SOCKBUF_UNLOCK(&so->so_snd);
863 if (nam && tp->t_state < TCPS_SYN_SENT) {
865 * Do implied connect if not yet connected,
866 * initialize window to default value, and
867 * initialize maxseg/maxopd using peer's cached
872 error = tcp6_connect(tp, nam, td);
874 #if defined(INET6) && defined(INET)
878 error = tcp_connect(tp, nam, td);
882 tp->snd_wnd = TTCP_CLIENT_SND_WND;
885 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
886 tp->t_flags |= TF_FORCEDATA;
887 error = tcp_output_send(tp);
888 tp->t_flags &= ~TF_FORCEDATA;
891 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
892 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
894 if (flags & PRUS_EOF)
895 INP_INFO_WUNLOCK(&V_tcbinfo);
900 * Abort the TCP. Drop the connection abruptly.
903 tcp_usr_abort(struct socket *so)
906 struct tcpcb *tp = NULL;
910 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
912 INP_INFO_WLOCK(&V_tcbinfo);
914 KASSERT(inp->inp_socket != NULL,
915 ("tcp_usr_abort: inp_socket == NULL"));
918 * If we still have full TCP state, and we're not dropped, drop.
920 if (!(inp->inp_flags & INP_TIMEWAIT) &&
921 !(inp->inp_flags & INP_DROPPED)) {
924 tcp_drop(tp, ECONNABORTED);
925 TCPDEBUG2(PRU_ABORT);
927 if (!(inp->inp_flags & INP_DROPPED)) {
929 so->so_state |= SS_PROTOREF;
931 inp->inp_flags |= INP_SOCKREF;
934 INP_INFO_WUNLOCK(&V_tcbinfo);
938 * TCP socket is closed. Start friendly disconnect.
941 tcp_usr_close(struct socket *so)
944 struct tcpcb *tp = NULL;
948 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
950 INP_INFO_WLOCK(&V_tcbinfo);
952 KASSERT(inp->inp_socket != NULL,
953 ("tcp_usr_close: inp_socket == NULL"));
956 * If we still have full TCP state, and we're not dropped, initiate
959 if (!(inp->inp_flags & INP_TIMEWAIT) &&
960 !(inp->inp_flags & INP_DROPPED)) {
964 TCPDEBUG2(PRU_CLOSE);
966 if (!(inp->inp_flags & INP_DROPPED)) {
968 so->so_state |= SS_PROTOREF;
970 inp->inp_flags |= INP_SOCKREF;
973 INP_INFO_WUNLOCK(&V_tcbinfo);
977 * Receive out-of-band data.
980 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
984 struct tcpcb *tp = NULL;
988 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
990 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
996 if ((so->so_oobmark == 0 &&
997 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
998 so->so_options & SO_OOBINLINE ||
999 tp->t_oobflags & TCPOOB_HADDATA) {
1003 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1004 error = EWOULDBLOCK;
1008 *mtod(m, caddr_t) = tp->t_iobc;
1009 if ((flags & MSG_PEEK) == 0)
1010 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1013 TCPDEBUG2(PRU_RCVOOB);
1019 struct pr_usrreqs tcp_usrreqs = {
1020 .pru_abort = tcp_usr_abort,
1021 .pru_accept = tcp_usr_accept,
1022 .pru_attach = tcp_usr_attach,
1023 .pru_bind = tcp_usr_bind,
1024 .pru_connect = tcp_usr_connect,
1025 .pru_control = in_control,
1026 .pru_detach = tcp_usr_detach,
1027 .pru_disconnect = tcp_usr_disconnect,
1028 .pru_listen = tcp_usr_listen,
1029 .pru_peeraddr = in_getpeeraddr,
1030 .pru_rcvd = tcp_usr_rcvd,
1031 .pru_rcvoob = tcp_usr_rcvoob,
1032 .pru_send = tcp_usr_send,
1033 .pru_shutdown = tcp_usr_shutdown,
1034 .pru_sockaddr = in_getsockaddr,
1035 .pru_sosetlabel = in_pcbsosetlabel,
1036 .pru_close = tcp_usr_close,
1041 struct pr_usrreqs tcp6_usrreqs = {
1042 .pru_abort = tcp_usr_abort,
1043 .pru_accept = tcp6_usr_accept,
1044 .pru_attach = tcp_usr_attach,
1045 .pru_bind = tcp6_usr_bind,
1046 .pru_connect = tcp6_usr_connect,
1047 .pru_control = in6_control,
1048 .pru_detach = tcp_usr_detach,
1049 .pru_disconnect = tcp_usr_disconnect,
1050 .pru_listen = tcp6_usr_listen,
1051 .pru_peeraddr = in6_mapped_peeraddr,
1052 .pru_rcvd = tcp_usr_rcvd,
1053 .pru_rcvoob = tcp_usr_rcvoob,
1054 .pru_send = tcp_usr_send,
1055 .pru_shutdown = tcp_usr_shutdown,
1056 .pru_sockaddr = in6_mapped_sockaddr,
1057 .pru_sosetlabel = in_pcbsosetlabel,
1058 .pru_close = tcp_usr_close,
1064 * Common subroutine to open a TCP connection to remote host specified
1065 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
1066 * port number if needed. Call in_pcbconnect_setup to do the routing and
1067 * to choose a local host address (interface). If there is an existing
1068 * incarnation of the same connection in TIME-WAIT state and if the remote
1069 * host was sending CC options and if the connection duration was < MSL, then
1070 * truncate the previous TIME-WAIT state and proceed.
1071 * Initialize connection parameters and enter SYN-SENT state.
1074 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1076 struct inpcb *inp = tp->t_inpcb, *oinp;
1077 struct socket *so = inp->inp_socket;
1078 struct in_addr laddr;
1082 INP_WLOCK_ASSERT(inp);
1083 INP_HASH_WLOCK(&V_tcbinfo);
1085 if (inp->inp_lport == 0) {
1086 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1092 * Cannot simply call in_pcbconnect, because there might be an
1093 * earlier incarnation of this same connection still in
1094 * TIME_WAIT state, creating an ADDRINUSE error.
1096 laddr = inp->inp_laddr;
1097 lport = inp->inp_lport;
1098 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1099 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1100 if (error && oinp == NULL)
1106 inp->inp_laddr = laddr;
1108 INP_HASH_WUNLOCK(&V_tcbinfo);
1111 * Compute window scaling to request:
1112 * Scale to fit into sweet spot. See tcp_syncache.c.
1113 * XXX: This should move to tcp_output().
1115 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1116 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1117 tp->request_r_scale++;
1120 TCPSTAT_INC(tcps_connattempt);
1121 tp->t_state = TCPS_SYN_SENT;
1122 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
1123 tp->iss = tcp_new_isn(tp);
1124 tcp_sendseqinit(tp);
1129 INP_HASH_WUNLOCK(&V_tcbinfo);
1136 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1138 struct inpcb *inp = tp->t_inpcb, *oinp;
1139 struct socket *so = inp->inp_socket;
1140 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
1141 struct in6_addr addr6;
1144 INP_WLOCK_ASSERT(inp);
1145 INP_HASH_WLOCK(&V_tcbinfo);
1147 if (inp->inp_lport == 0) {
1148 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1154 * Cannot simply call in_pcbconnect, because there might be an
1155 * earlier incarnation of this same connection still in
1156 * TIME_WAIT state, creating an ADDRINUSE error.
1157 * in6_pcbladdr() also handles scope zone IDs.
1159 * XXXRW: We wouldn't need to expose in6_pcblookup_hash_locked()
1160 * outside of in6_pcb.c if there were an in6_pcbconnect_setup().
1162 error = in6_pcbladdr(inp, nam, &addr6);
1165 oinp = in6_pcblookup_hash_locked(inp->inp_pcbinfo,
1166 &sin6->sin6_addr, sin6->sin6_port,
1167 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
1170 inp->inp_lport, 0, NULL);
1175 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1176 inp->in6p_laddr = addr6;
1177 inp->in6p_faddr = sin6->sin6_addr;
1178 inp->inp_fport = sin6->sin6_port;
1179 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
1180 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
1181 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
1183 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
1185 INP_HASH_WUNLOCK(&V_tcbinfo);
1187 /* Compute window scaling to request. */
1188 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1189 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1190 tp->request_r_scale++;
1193 TCPSTAT_INC(tcps_connattempt);
1194 tp->t_state = TCPS_SYN_SENT;
1195 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
1196 tp->iss = tcp_new_isn(tp);
1197 tcp_sendseqinit(tp);
1202 INP_HASH_WUNLOCK(&V_tcbinfo);
1208 * Export TCP internal state information via a struct tcp_info, based on the
1209 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently
1210 * (TCP state machine, etc). We export all information using FreeBSD-native
1211 * constants -- for example, the numeric values for tcpi_state will differ
1215 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1218 INP_WLOCK_ASSERT(tp->t_inpcb);
1219 bzero(ti, sizeof(*ti));
1221 ti->tcpi_state = tp->t_state;
1222 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1223 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1224 if (tp->t_flags & TF_SACK_PERMIT)
1225 ti->tcpi_options |= TCPI_OPT_SACK;
1226 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1227 ti->tcpi_options |= TCPI_OPT_WSCALE;
1228 ti->tcpi_snd_wscale = tp->snd_scale;
1229 ti->tcpi_rcv_wscale = tp->rcv_scale;
1232 ti->tcpi_rto = tp->t_rxtcur * tick;
1233 ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick;
1234 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1235 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1237 ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1238 ti->tcpi_snd_cwnd = tp->snd_cwnd;
1241 * FreeBSD-specific extension fields for tcp_info.
1243 ti->tcpi_rcv_space = tp->rcv_wnd;
1244 ti->tcpi_rcv_nxt = tp->rcv_nxt;
1245 ti->tcpi_snd_wnd = tp->snd_wnd;
1246 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */
1247 ti->tcpi_snd_nxt = tp->snd_nxt;
1248 ti->tcpi_snd_mss = tp->t_maxseg;
1249 ti->tcpi_rcv_mss = tp->t_maxseg;
1250 if (tp->t_flags & TF_TOE)
1251 ti->tcpi_options |= TCPI_OPT_TOE;
1252 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1253 ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1254 ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1258 * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1259 * socket option arguments. When it re-acquires the lock after the copy, it
1260 * has to revalidate that the connection is still valid for the socket
1263 #define INP_WLOCK_RECHECK(inp) do { \
1265 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
1267 return (ECONNRESET); \
1269 tp = intotcpcb(inp); \
1273 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1275 int error, opt, optval;
1280 char buf[TCP_CA_NAME_MAX];
1281 struct cc_algo *algo;
1284 inp = sotoinpcb(so);
1285 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1287 if (sopt->sopt_level != IPPROTO_TCP) {
1289 if (inp->inp_vflag & INP_IPV6PROTO) {
1291 error = ip6_ctloutput(so, sopt);
1294 #if defined(INET6) && defined(INET)
1300 error = ip_ctloutput(so, sopt);
1305 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1307 return (ECONNRESET);
1310 switch (sopt->sopt_dir) {
1312 switch (sopt->sopt_name) {
1313 #ifdef TCP_SIGNATURE
1316 error = sooptcopyin(sopt, &optval, sizeof optval,
1321 INP_WLOCK_RECHECK(inp);
1323 tp->t_flags |= TF_SIGNATURE;
1325 tp->t_flags &= ~TF_SIGNATURE;
1328 #endif /* TCP_SIGNATURE */
1332 error = sooptcopyin(sopt, &optval, sizeof optval,
1337 INP_WLOCK_RECHECK(inp);
1338 switch (sopt->sopt_name) {
1346 opt = 0; /* dead code to fool gcc */
1353 tp->t_flags &= ~opt;
1359 error = sooptcopyin(sopt, &optval, sizeof optval,
1364 INP_WLOCK_RECHECK(inp);
1366 tp->t_flags |= TF_NOPUSH;
1367 else if (tp->t_flags & TF_NOPUSH) {
1368 tp->t_flags &= ~TF_NOPUSH;
1369 if (TCPS_HAVEESTABLISHED(tp->t_state))
1370 error = tcp_output(tp);
1377 error = sooptcopyin(sopt, &optval, sizeof optval,
1382 INP_WLOCK_RECHECK(inp);
1383 if (optval > 0 && optval <= tp->t_maxseg &&
1384 optval + 40 >= V_tcp_minmss)
1385 tp->t_maxseg = optval;
1396 case TCP_CONGESTION:
1398 bzero(buf, sizeof(buf));
1399 error = sooptcopyin(sopt, &buf, sizeof(buf), 1);
1402 INP_WLOCK_RECHECK(inp);
1404 * Return EINVAL if we can't find the requested cc algo.
1408 STAILQ_FOREACH(algo, &cc_list, entries) {
1409 if (strncmp(buf, algo->name, TCP_CA_NAME_MAX)
1411 /* We've found the requested algo. */
1414 * We hold a write lock over the tcb
1415 * so it's safe to do these things
1416 * without ordering concerns.
1418 if (CC_ALGO(tp)->cb_destroy != NULL)
1419 CC_ALGO(tp)->cb_destroy(tp->ccv);
1422 * If something goes pear shaped
1423 * initialising the new algo,
1424 * fall back to newreno (which
1425 * does not require initialisation).
1427 if (algo->cb_init != NULL)
1428 if (algo->cb_init(tp->ccv) > 0) {
1429 CC_ALGO(tp) = &newreno_cc_algo;
1431 * The only reason init
1433 * because of malloc.
1437 break; /* Break the STAILQ_FOREACH. */
1448 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1452 if (ui > (UINT_MAX / hz)) {
1458 INP_WLOCK_RECHECK(inp);
1459 switch (sopt->sopt_name) {
1461 tp->t_keepidle = ui;
1463 * XXX: better check current remaining
1464 * timeout and "merge" it with new value.
1466 if ((tp->t_state > TCPS_LISTEN) &&
1467 (tp->t_state <= TCPS_CLOSING))
1468 tcp_timer_activate(tp, TT_KEEP,
1472 tp->t_keepintvl = ui;
1473 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1474 (TP_MAXIDLE(tp) > 0))
1475 tcp_timer_activate(tp, TT_2MSL,
1479 tp->t_keepinit = ui;
1480 if (tp->t_state == TCPS_SYN_RECEIVED ||
1481 tp->t_state == TCPS_SYN_SENT)
1482 tcp_timer_activate(tp, TT_KEEP,
1491 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1495 INP_WLOCK_RECHECK(inp);
1497 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1498 (TP_MAXIDLE(tp) > 0))
1499 tcp_timer_activate(tp, TT_2MSL,
1506 error = ENOPROTOOPT;
1512 tp = intotcpcb(inp);
1513 switch (sopt->sopt_name) {
1514 #ifdef TCP_SIGNATURE
1516 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1518 error = sooptcopyout(sopt, &optval, sizeof optval);
1523 optval = tp->t_flags & TF_NODELAY;
1525 error = sooptcopyout(sopt, &optval, sizeof optval);
1528 optval = tp->t_maxseg;
1530 error = sooptcopyout(sopt, &optval, sizeof optval);
1533 optval = tp->t_flags & TF_NOOPT;
1535 error = sooptcopyout(sopt, &optval, sizeof optval);
1538 optval = tp->t_flags & TF_NOPUSH;
1540 error = sooptcopyout(sopt, &optval, sizeof optval);
1543 tcp_fill_info(tp, &ti);
1545 error = sooptcopyout(sopt, &ti, sizeof ti);
1547 case TCP_CONGESTION:
1548 bzero(buf, sizeof(buf));
1549 strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1551 error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX);
1555 error = ENOPROTOOPT;
1562 #undef INP_WLOCK_RECHECK
1565 * tcp_sendspace and tcp_recvspace are the default send and receive window
1566 * sizes, respectively. These are obsolescent (this information should
1567 * be set by the route).
1569 u_long tcp_sendspace = 1024*32;
1570 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1571 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1572 u_long tcp_recvspace = 1024*64;
1573 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1574 &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1577 * Attach TCP protocol to socket, allocating
1578 * internet protocol control block, tcp control block,
1579 * bufer space, and entering LISTEN state if to accept connections.
1582 tcp_attach(struct socket *so)
1588 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1589 error = soreserve(so, tcp_sendspace, tcp_recvspace);
1593 so->so_rcv.sb_flags |= SB_AUTOSIZE;
1594 so->so_snd.sb_flags |= SB_AUTOSIZE;
1595 INP_INFO_WLOCK(&V_tcbinfo);
1596 error = in_pcballoc(so, &V_tcbinfo);
1598 INP_INFO_WUNLOCK(&V_tcbinfo);
1601 inp = sotoinpcb(so);
1603 if (inp->inp_vflag & INP_IPV6PROTO) {
1604 inp->inp_vflag |= INP_IPV6;
1605 inp->in6p_hops = -1; /* use kernel default */
1609 inp->inp_vflag |= INP_IPV4;
1610 tp = tcp_newtcpcb(inp);
1614 INP_INFO_WUNLOCK(&V_tcbinfo);
1617 tp->t_state = TCPS_CLOSED;
1619 INP_INFO_WUNLOCK(&V_tcbinfo);
1624 * Initiate (or continue) disconnect.
1625 * If embryonic state, just send reset (once).
1626 * If in ``let data drain'' option and linger null, just drop.
1627 * Otherwise (hard), mark socket disconnecting and drop
1628 * current input data; switch states based on user close, and
1629 * send segment to peer (with FIN).
1632 tcp_disconnect(struct tcpcb *tp)
1634 struct inpcb *inp = tp->t_inpcb;
1635 struct socket *so = inp->inp_socket;
1637 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1638 INP_WLOCK_ASSERT(inp);
1641 * Neither tcp_close() nor tcp_drop() should return NULL, as the
1642 * socket is still open.
1644 if (tp->t_state < TCPS_ESTABLISHED) {
1647 ("tcp_disconnect: tcp_close() returned NULL"));
1648 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1649 tp = tcp_drop(tp, 0);
1651 ("tcp_disconnect: tcp_drop() returned NULL"));
1653 soisdisconnecting(so);
1654 sbflush(&so->so_rcv);
1656 if (!(inp->inp_flags & INP_DROPPED))
1657 tcp_output_disconnect(tp);
1662 * User issued close, and wish to trail through shutdown states:
1663 * if never received SYN, just forget it. If got a SYN from peer,
1664 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1665 * If already got a FIN from peer, then almost done; go to LAST_ACK
1666 * state. In all other cases, have already sent FIN to peer (e.g.
1667 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1668 * for peer to send FIN or not respond to keep-alives, etc.
1669 * We can let the user exit from the close as soon as the FIN is acked.
1672 tcp_usrclosed(struct tcpcb *tp)
1675 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1676 INP_WLOCK_ASSERT(tp->t_inpcb);
1678 switch (tp->t_state) {
1680 tcp_offload_listen_close(tp);
1683 tp->t_state = TCPS_CLOSED;
1686 * tcp_close() should never return NULL here as the socket is
1690 ("tcp_usrclosed: tcp_close() returned NULL"));
1694 case TCPS_SYN_RECEIVED:
1695 tp->t_flags |= TF_NEEDFIN;
1698 case TCPS_ESTABLISHED:
1699 tp->t_state = TCPS_FIN_WAIT_1;
1702 case TCPS_CLOSE_WAIT:
1703 tp->t_state = TCPS_LAST_ACK;
1706 if (tp->t_state >= TCPS_FIN_WAIT_2) {
1707 soisdisconnected(tp->t_inpcb->inp_socket);
1708 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
1709 if (tp->t_state == TCPS_FIN_WAIT_2) {
1712 timeout = (tcp_fast_finwait2_recycle) ?
1713 tcp_finwait2_timeout : TP_MAXIDLE(tp);
1714 tcp_timer_activate(tp, TT_2MSL, timeout);
1721 db_print_indent(int indent)
1725 for (i = 0; i < indent; i++)
1730 db_print_tstate(int t_state)
1735 db_printf("TCPS_CLOSED");
1739 db_printf("TCPS_LISTEN");
1743 db_printf("TCPS_SYN_SENT");
1746 case TCPS_SYN_RECEIVED:
1747 db_printf("TCPS_SYN_RECEIVED");
1750 case TCPS_ESTABLISHED:
1751 db_printf("TCPS_ESTABLISHED");
1754 case TCPS_CLOSE_WAIT:
1755 db_printf("TCPS_CLOSE_WAIT");
1758 case TCPS_FIN_WAIT_1:
1759 db_printf("TCPS_FIN_WAIT_1");
1763 db_printf("TCPS_CLOSING");
1767 db_printf("TCPS_LAST_ACK");
1770 case TCPS_FIN_WAIT_2:
1771 db_printf("TCPS_FIN_WAIT_2");
1774 case TCPS_TIME_WAIT:
1775 db_printf("TCPS_TIME_WAIT");
1779 db_printf("unknown");
1785 db_print_tflags(u_int t_flags)
1790 if (t_flags & TF_ACKNOW) {
1791 db_printf("%sTF_ACKNOW", comma ? ", " : "");
1794 if (t_flags & TF_DELACK) {
1795 db_printf("%sTF_DELACK", comma ? ", " : "");
1798 if (t_flags & TF_NODELAY) {
1799 db_printf("%sTF_NODELAY", comma ? ", " : "");
1802 if (t_flags & TF_NOOPT) {
1803 db_printf("%sTF_NOOPT", comma ? ", " : "");
1806 if (t_flags & TF_SENTFIN) {
1807 db_printf("%sTF_SENTFIN", comma ? ", " : "");
1810 if (t_flags & TF_REQ_SCALE) {
1811 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
1814 if (t_flags & TF_RCVD_SCALE) {
1815 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
1818 if (t_flags & TF_REQ_TSTMP) {
1819 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
1822 if (t_flags & TF_RCVD_TSTMP) {
1823 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
1826 if (t_flags & TF_SACK_PERMIT) {
1827 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
1830 if (t_flags & TF_NEEDSYN) {
1831 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
1834 if (t_flags & TF_NEEDFIN) {
1835 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
1838 if (t_flags & TF_NOPUSH) {
1839 db_printf("%sTF_NOPUSH", comma ? ", " : "");
1842 if (t_flags & TF_MORETOCOME) {
1843 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
1846 if (t_flags & TF_LQ_OVERFLOW) {
1847 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
1850 if (t_flags & TF_LASTIDLE) {
1851 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
1854 if (t_flags & TF_RXWIN0SENT) {
1855 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
1858 if (t_flags & TF_FASTRECOVERY) {
1859 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
1862 if (t_flags & TF_CONGRECOVERY) {
1863 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
1866 if (t_flags & TF_WASFRECOVERY) {
1867 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
1870 if (t_flags & TF_SIGNATURE) {
1871 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
1874 if (t_flags & TF_FORCEDATA) {
1875 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
1878 if (t_flags & TF_TSO) {
1879 db_printf("%sTF_TSO", comma ? ", " : "");
1882 if (t_flags & TF_ECN_PERMIT) {
1883 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
1889 db_print_toobflags(char t_oobflags)
1894 if (t_oobflags & TCPOOB_HAVEDATA) {
1895 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
1898 if (t_oobflags & TCPOOB_HADDATA) {
1899 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
1905 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
1908 db_print_indent(indent);
1909 db_printf("%s at %p\n", name, tp);
1913 db_print_indent(indent);
1914 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n",
1915 LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
1917 db_print_indent(indent);
1918 db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n",
1919 &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
1921 db_print_indent(indent);
1922 db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl,
1923 &tp->t_timers->tt_delack, tp->t_inpcb);
1925 db_print_indent(indent);
1926 db_printf("t_state: %d (", tp->t_state);
1927 db_print_tstate(tp->t_state);
1930 db_print_indent(indent);
1931 db_printf("t_flags: 0x%x (", tp->t_flags);
1932 db_print_tflags(tp->t_flags);
1935 db_print_indent(indent);
1936 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n",
1937 tp->snd_una, tp->snd_max, tp->snd_nxt);
1939 db_print_indent(indent);
1940 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n",
1941 tp->snd_up, tp->snd_wl1, tp->snd_wl2);
1943 db_print_indent(indent);
1944 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n",
1945 tp->iss, tp->irs, tp->rcv_nxt);
1947 db_print_indent(indent);
1948 db_printf("rcv_adv: 0x%08x rcv_wnd: %lu rcv_up: 0x%08x\n",
1949 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
1951 db_print_indent(indent);
1952 db_printf("snd_wnd: %lu snd_cwnd: %lu\n",
1953 tp->snd_wnd, tp->snd_cwnd);
1955 db_print_indent(indent);
1956 db_printf("snd_ssthresh: %lu snd_recover: "
1957 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
1959 db_print_indent(indent);
1960 db_printf("t_maxopd: %u t_rcvtime: %u t_startime: %u\n",
1961 tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
1963 db_print_indent(indent);
1964 db_printf("t_rttime: %u t_rtsq: 0x%08x\n",
1965 tp->t_rtttime, tp->t_rtseq);
1967 db_print_indent(indent);
1968 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n",
1969 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
1971 db_print_indent(indent);
1972 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u "
1973 "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
1976 db_print_indent(indent);
1977 db_printf("t_rttupdated: %lu max_sndwnd: %lu t_softerror: %d\n",
1978 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
1980 db_print_indent(indent);
1981 db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
1982 db_print_toobflags(tp->t_oobflags);
1983 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc);
1985 db_print_indent(indent);
1986 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n",
1987 tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
1989 db_print_indent(indent);
1990 db_printf("ts_recent: %u ts_recent_age: %u\n",
1991 tp->ts_recent, tp->ts_recent_age);
1993 db_print_indent(indent);
1994 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: "
1995 "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
1997 db_print_indent(indent);
1998 db_printf("snd_ssthresh_prev: %lu snd_recover_prev: 0x%08x "
1999 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2000 tp->snd_recover_prev, tp->t_badrxtwin);
2002 db_print_indent(indent);
2003 db_printf("snd_numholes: %d snd_holes first: %p\n",
2004 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2006 db_print_indent(indent);
2007 db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: "
2008 "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
2010 /* Skip sackblks, sackhint. */
2012 db_print_indent(indent);
2013 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n",
2014 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2017 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2022 db_printf("usage: show tcpcb <addr>\n");
2025 tp = (struct tcpcb *)addr;
2027 db_print_tcpcb(tp, "tcpcb", 0);