2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California.
6 * Copyright (c) 2006-2007 Robert N. M. Watson
7 * Copyright (c) 2010-2011 Juniper Networks, Inc.
10 * Portions of this software were developed by Robert N. M. Watson under
11 * contract to Juniper Networks, Inc.
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
45 #include "opt_inet6.h"
46 #include "opt_ipsec.h"
47 #include "opt_kern_tls.h"
48 #include "opt_tcpdebug.h"
50 #include <sys/param.h>
51 #include <sys/systm.h>
53 #include <sys/limits.h>
54 #include <sys/malloc.h>
55 #include <sys/refcount.h>
56 #include <sys/kernel.h>
58 #include <sys/qmath.h>
59 #include <sys/sysctl.h>
62 #include <sys/domain.h>
64 #include <sys/socket.h>
65 #include <sys/socketvar.h>
66 #include <sys/protosw.h>
69 #include <sys/syslog.h>
70 #include <sys/stats.h>
77 #include <net/if_var.h>
78 #include <net/route.h>
81 #include <netinet/in.h>
82 #include <netinet/in_kdtrace.h>
83 #include <netinet/in_pcb.h>
84 #include <netinet/in_systm.h>
85 #include <netinet/in_var.h>
86 #include <netinet/ip_var.h>
88 #include <netinet/ip6.h>
89 #include <netinet6/in6_pcb.h>
90 #include <netinet6/ip6_var.h>
91 #include <netinet6/scope6_var.h>
93 #include <netinet/tcp.h>
94 #include <netinet/tcp_fsm.h>
95 #include <netinet/tcp_seq.h>
96 #include <netinet/tcp_timer.h>
97 #include <netinet/tcp_var.h>
98 #include <netinet/tcp_log_buf.h>
99 #include <netinet/tcpip.h>
100 #include <netinet/cc/cc.h>
101 #include <netinet/tcp_fastopen.h>
102 #include <netinet/tcp_hpts.h>
104 #include <netinet/tcp_pcap.h>
107 #include <netinet/tcp_debug.h>
110 #include <netinet/tcp_offload.h>
112 #include <netipsec/ipsec_support.h>
115 #include <vm/vm_param.h>
117 #include <vm/vm_extern.h>
118 #include <vm/vm_map.h>
119 #include <vm/vm_page.h>
122 * TCP protocol interface to socket abstraction.
125 static int tcp_connect(struct tcpcb *, struct sockaddr *,
129 static int tcp6_connect(struct tcpcb *, struct sockaddr *,
132 static void tcp_disconnect(struct tcpcb *);
133 static void tcp_usrclosed(struct tcpcb *);
134 static void tcp_fill_info(struct tcpcb *, struct tcp_info *);
136 static int tcp_pru_options_support(struct tcpcb *tp, int flags);
139 #define TCPDEBUG0 int ostate = 0
140 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0
141 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
142 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
146 #define TCPDEBUG2(req)
150 * tcp_require_unique port requires a globally-unique source port for each
151 * outgoing connection. The default is to require the 4-tuple to be unique.
153 VNET_DEFINE(int, tcp_require_unique_port) = 0;
154 SYSCTL_INT(_net_inet_tcp, OID_AUTO, require_unique_port,
155 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_require_unique_port), 0,
156 "Require globally-unique ephemeral port for outgoing connections");
157 #define V_tcp_require_unique_port VNET(tcp_require_unique_port)
160 * TCP attaches to socket via pru_attach(), reserving space,
161 * and an internet control block.
164 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
167 struct tcpcb *tp = NULL;
172 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
175 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
176 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
181 so->so_rcv.sb_flags |= SB_AUTOSIZE;
182 so->so_snd.sb_flags |= SB_AUTOSIZE;
183 error = in_pcballoc(so, &V_tcbinfo);
188 if (inp->inp_vflag & INP_IPV6PROTO) {
189 inp->inp_vflag |= INP_IPV6;
190 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
191 inp->inp_vflag |= INP_IPV4;
192 inp->in6p_hops = -1; /* use kernel default */
196 inp->inp_vflag |= INP_IPV4;
197 tp = tcp_newtcpcb(inp);
204 tp->t_state = TCPS_CLOSED;
206 TCPSTATES_INC(TCPS_CLOSED);
208 TCPDEBUG2(PRU_ATTACH);
209 TCP_PROBE2(debug__user, tp, PRU_ATTACH);
214 * tcp_usr_detach is called when the socket layer loses its final reference
215 * to the socket, be it a file descriptor reference, a reference from TCP,
216 * etc. At this point, there is only one case in which we will keep around
217 * inpcb state: time wait.
220 tcp_usr_detach(struct socket *so)
226 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
228 KASSERT(so->so_pcb == inp && inp->inp_socket == so,
229 ("%s: socket %p inp %p mismatch", __func__, so, inp));
233 if (inp->inp_flags & INP_TIMEWAIT) {
235 * There are two cases to handle: one in which the time wait
236 * state is being discarded (INP_DROPPED), and one in which
237 * this connection will remain in timewait. In the former,
238 * it is time to discard all state (except tcptw, which has
239 * already been discarded by the timewait close code, which
240 * should be further up the call stack somewhere). In the
241 * latter case, we detach from the socket, but leave the pcb
242 * present until timewait ends.
244 * XXXRW: Would it be cleaner to free the tcptw here?
246 * Astute question indeed, from twtcp perspective there are
247 * four cases to consider:
249 * #1 tcp_usr_detach is called at tcptw creation time by
250 * tcp_twstart, then do not discard the newly created tcptw
251 * and leave inpcb present until timewait ends
252 * #2 tcp_usr_detach is called at tcptw creation time by
253 * tcp_twstart, but connection is local and tw will be
254 * discarded immediately
255 * #3 tcp_usr_detach is called at timewait end (or reuse) by
256 * tcp_twclose, then the tcptw has already been discarded
257 * (or reused) and inpcb is freed here
258 * #4 tcp_usr_detach is called() after timewait ends (or reuse)
259 * (e.g. by soclose), then tcptw has already been discarded
260 * (or reused) and inpcb is freed here
262 * In all three cases the tcptw should not be freed here.
264 if (inp->inp_flags & INP_DROPPED) {
266 if (__predict_true(tp == NULL)) {
270 * This case should not happen as in TIMEWAIT
271 * state the inp should not be destroyed before
272 * its tcptw. If INVARIANTS is defined, panic.
275 panic("%s: Panic before an inp double-free: "
276 "INP_TIMEWAIT && INP_DROPPED && tp != NULL"
279 log(LOG_ERR, "%s: Avoid an inp double-free: "
280 "INP_TIMEWAIT && INP_DROPPED && tp != NULL"
291 * If the connection is not in timewait, we consider two
292 * two conditions: one in which no further processing is
293 * necessary (dropped || embryonic), and one in which TCP is
294 * not yet done, but no longer requires the socket, so the
295 * pcb will persist for the time being.
297 * XXXRW: Does the second case still occur?
299 if (inp->inp_flags & INP_DROPPED ||
300 tp->t_state < TCPS_SYN_SENT) {
313 * Give the socket an address.
316 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
320 struct tcpcb *tp = NULL;
321 struct sockaddr_in *sinp;
323 sinp = (struct sockaddr_in *)nam;
324 if (nam->sa_family != AF_INET) {
326 * Preserve compatibility with old programs.
328 if (nam->sa_family != AF_UNSPEC ||
329 sinp->sin_addr.s_addr != INADDR_ANY)
330 return (EAFNOSUPPORT);
331 nam->sa_family = AF_INET;
333 if (nam->sa_len != sizeof(*sinp))
337 * Must check for multicast addresses and disallow binding
340 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
341 return (EAFNOSUPPORT);
345 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
347 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
353 INP_HASH_WLOCK(&V_tcbinfo);
354 error = in_pcbbind(inp, nam, td->td_ucred);
355 INP_HASH_WUNLOCK(&V_tcbinfo);
358 TCP_PROBE2(debug__user, tp, PRU_BIND);
367 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
371 struct tcpcb *tp = NULL;
372 struct sockaddr_in6 *sin6;
375 sin6 = (struct sockaddr_in6 *)nam;
376 if (nam->sa_family != AF_INET6)
377 return (EAFNOSUPPORT);
378 if (nam->sa_len != sizeof(*sin6))
382 * Must check for multicast addresses and disallow binding
385 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
386 return (EAFNOSUPPORT);
390 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
392 vflagsav = inp->inp_vflag;
393 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
399 INP_HASH_WLOCK(&V_tcbinfo);
400 inp->inp_vflag &= ~INP_IPV4;
401 inp->inp_vflag |= INP_IPV6;
403 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
404 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
405 inp->inp_vflag |= INP_IPV4;
406 else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
407 struct sockaddr_in sin;
409 in6_sin6_2_sin(&sin, sin6);
410 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
411 error = EAFNOSUPPORT;
412 INP_HASH_WUNLOCK(&V_tcbinfo);
415 inp->inp_vflag |= INP_IPV4;
416 inp->inp_vflag &= ~INP_IPV6;
417 error = in_pcbbind(inp, (struct sockaddr *)&sin,
419 INP_HASH_WUNLOCK(&V_tcbinfo);
424 error = in6_pcbbind(inp, nam, td->td_ucred);
425 INP_HASH_WUNLOCK(&V_tcbinfo);
428 inp->inp_vflag = vflagsav;
430 TCP_PROBE2(debug__user, tp, PRU_BIND);
438 * Prepare to accept connections.
441 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
445 struct tcpcb *tp = NULL;
449 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
451 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
458 error = solisten_proto_check(so);
459 INP_HASH_WLOCK(&V_tcbinfo);
460 if (error == 0 && inp->inp_lport == 0)
461 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
462 INP_HASH_WUNLOCK(&V_tcbinfo);
464 tcp_state_change(tp, TCPS_LISTEN);
465 solisten_proto(so, backlog);
467 if ((so->so_options & SO_NO_OFFLOAD) == 0)
468 tcp_offload_listen_start(tp);
473 if (IS_FASTOPEN(tp->t_flags))
474 tp->t_tfo_pending = tcp_fastopen_alloc_counter();
477 TCPDEBUG2(PRU_LISTEN);
478 TCP_PROBE2(debug__user, tp, PRU_LISTEN);
486 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
490 struct tcpcb *tp = NULL;
495 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
497 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
501 vflagsav = inp->inp_vflag;
505 error = solisten_proto_check(so);
506 INP_HASH_WLOCK(&V_tcbinfo);
507 if (error == 0 && inp->inp_lport == 0) {
508 inp->inp_vflag &= ~INP_IPV4;
509 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
510 inp->inp_vflag |= INP_IPV4;
511 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
513 INP_HASH_WUNLOCK(&V_tcbinfo);
515 tcp_state_change(tp, TCPS_LISTEN);
516 solisten_proto(so, backlog);
518 if ((so->so_options & SO_NO_OFFLOAD) == 0)
519 tcp_offload_listen_start(tp);
524 if (IS_FASTOPEN(tp->t_flags))
525 tp->t_tfo_pending = tcp_fastopen_alloc_counter();
528 inp->inp_vflag = vflagsav;
531 TCPDEBUG2(PRU_LISTEN);
532 TCP_PROBE2(debug__user, tp, PRU_LISTEN);
540 * Initiate connection to peer.
541 * Create a template for use in transmissions on this connection.
542 * Enter SYN_SENT state, and mark socket as connecting.
543 * Start keep-alive timer, and seed output sequence space.
544 * Send initial segment on connection.
547 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
549 struct epoch_tracker et;
552 struct tcpcb *tp = NULL;
553 struct sockaddr_in *sinp;
555 sinp = (struct sockaddr_in *)nam;
556 if (nam->sa_family != AF_INET)
557 return (EAFNOSUPPORT);
558 if (nam->sa_len != sizeof (*sinp))
562 * Must disallow TCP ``connections'' to multicast addresses.
564 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
565 return (EAFNOSUPPORT);
566 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST)
568 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
573 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
575 if (inp->inp_flags & INP_TIMEWAIT) {
579 if (inp->inp_flags & INP_DROPPED) {
580 error = ECONNREFUSED;
586 if ((error = tcp_connect(tp, nam, td)) != 0)
589 if (registered_toedevs > 0 &&
590 (so->so_options & SO_NO_OFFLOAD) == 0 &&
591 (error = tcp_offload_connect(so, nam)) == 0)
594 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
595 error = tp->t_fb->tfb_tcp_output(tp);
599 TCPDEBUG2(PRU_CONNECT);
600 TCP_PROBE2(debug__user, tp, PRU_CONNECT);
608 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
610 struct epoch_tracker et;
613 struct tcpcb *tp = NULL;
614 struct sockaddr_in6 *sin6;
620 sin6 = (struct sockaddr_in6 *)nam;
621 if (nam->sa_family != AF_INET6)
622 return (EAFNOSUPPORT);
623 if (nam->sa_len != sizeof (*sin6))
627 * Must disallow TCP ``connections'' to multicast addresses.
629 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
630 return (EAFNOSUPPORT);
633 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
635 vflagsav = inp->inp_vflag;
636 incflagsav = inp->inp_inc.inc_flags;
637 if (inp->inp_flags & INP_TIMEWAIT) {
641 if (inp->inp_flags & INP_DROPPED) {
642 error = ECONNREFUSED;
649 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
650 * therefore probably require the hash lock, which isn't held here.
651 * Is this a significant problem?
653 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
654 struct sockaddr_in sin;
656 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
660 if ((inp->inp_vflag & INP_IPV4) == 0) {
661 error = EAFNOSUPPORT;
665 in6_sin6_2_sin(&sin, sin6);
666 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
667 error = EAFNOSUPPORT;
670 if (ntohl(sin.sin_addr.s_addr) == INADDR_BROADCAST) {
674 if ((error = prison_remote_ip4(td->td_ucred,
675 &sin.sin_addr)) != 0)
677 inp->inp_vflag |= INP_IPV4;
678 inp->inp_vflag &= ~INP_IPV6;
680 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
683 if (registered_toedevs > 0 &&
684 (so->so_options & SO_NO_OFFLOAD) == 0 &&
685 (error = tcp_offload_connect(so, nam)) == 0)
688 error = tp->t_fb->tfb_tcp_output(tp);
691 if ((inp->inp_vflag & INP_IPV6) == 0) {
692 error = EAFNOSUPPORT;
697 if ((error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr)) != 0)
699 inp->inp_vflag &= ~INP_IPV4;
700 inp->inp_vflag |= INP_IPV6;
701 inp->inp_inc.inc_flags |= INC_ISIPV6;
702 if ((error = tcp6_connect(tp, nam, td)) != 0)
705 if (registered_toedevs > 0 &&
706 (so->so_options & SO_NO_OFFLOAD) == 0 &&
707 (error = tcp_offload_connect(so, nam)) == 0)
710 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
712 error = tp->t_fb->tfb_tcp_output(tp);
719 * If the implicit bind in the connect call fails, restore
720 * the flags we modified.
722 if (error != 0 && inp->inp_lport == 0) {
723 inp->inp_vflag = vflagsav;
724 inp->inp_inc.inc_flags = incflagsav;
727 TCPDEBUG2(PRU_CONNECT);
728 TCP_PROBE2(debug__user, tp, PRU_CONNECT);
735 * Initiate disconnect from peer.
736 * If connection never passed embryonic stage, just drop;
737 * else if don't need to let data drain, then can just drop anyways,
738 * else have to begin TCP shutdown process: mark socket disconnecting,
739 * drain unread data, state switch to reflect user close, and
740 * send segment (e.g. FIN) to peer. Socket will be really disconnected
741 * when peer sends FIN and acks ours.
743 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
746 tcp_usr_disconnect(struct socket *so)
749 struct tcpcb *tp = NULL;
750 struct epoch_tracker et;
756 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
758 if (inp->inp_flags & INP_TIMEWAIT)
760 if (inp->inp_flags & INP_DROPPED) {
768 TCPDEBUG2(PRU_DISCONNECT);
769 TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
777 * Accept a connection. Essentially all the work is done at higher levels;
778 * just return the address of the peer, storing through addr.
781 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
784 struct inpcb *inp = NULL;
785 struct tcpcb *tp = NULL;
790 if (so->so_state & SS_ISDISCONNECTED)
791 return (ECONNABORTED);
794 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
796 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
797 error = ECONNABORTED;
804 * We inline in_getpeeraddr and COMMON_END here, so that we can
805 * copy the data of interest and defer the malloc until after we
808 port = inp->inp_fport;
809 addr = inp->inp_faddr;
812 TCPDEBUG2(PRU_ACCEPT);
813 TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
816 *nam = in_sockaddr(port, &addr);
823 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
825 struct inpcb *inp = NULL;
827 struct tcpcb *tp = NULL;
829 struct in6_addr addr6;
830 struct epoch_tracker et;
835 if (so->so_state & SS_ISDISCONNECTED)
836 return (ECONNABORTED);
839 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
842 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
843 error = ECONNABORTED;
850 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
851 * copy the data of interest and defer the malloc until after we
854 if (inp->inp_vflag & INP_IPV4) {
856 port = inp->inp_fport;
857 addr = inp->inp_faddr;
859 port = inp->inp_fport;
860 addr6 = inp->in6p_faddr;
864 TCPDEBUG2(PRU_ACCEPT);
865 TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
870 *nam = in6_v4mapsin6_sockaddr(port, &addr);
872 *nam = in6_sockaddr(port, &addr6);
879 * Mark the connection as being incapable of further output.
882 tcp_usr_shutdown(struct socket *so)
886 struct tcpcb *tp = NULL;
887 struct epoch_tracker et;
892 KASSERT(inp != NULL, ("inp == NULL"));
894 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
902 if (!(inp->inp_flags & INP_DROPPED))
903 error = tp->t_fb->tfb_tcp_output(tp);
906 TCPDEBUG2(PRU_SHUTDOWN);
907 TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
915 * After a receive, possibly send window update to peer.
918 tcp_usr_rcvd(struct socket *so, int flags)
920 struct epoch_tracker et;
922 struct tcpcb *tp = NULL;
927 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
929 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
936 * For passively-created TFO connections, don't attempt a window
937 * update while still in SYN_RECEIVED as this may trigger an early
938 * SYN|ACK. It is preferable to have the SYN|ACK be sent along with
939 * application response data, or failing that, when the DELACK timer
942 if (IS_FASTOPEN(tp->t_flags) &&
943 (tp->t_state == TCPS_SYN_RECEIVED))
947 if (tp->t_flags & TF_TOE)
948 tcp_offload_rcvd(tp);
951 tp->t_fb->tfb_tcp_output(tp);
955 TCP_PROBE2(debug__user, tp, PRU_RCVD);
961 * Do a send by putting data in output queue and updating urgent
962 * marker if URG set. Possibly send more data. Unlike the other
963 * pru_*() routines, the mbuf chains are our responsibility. We
964 * must either enqueue them or free them. The other pru_* routines
965 * generally are caller-frees.
968 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
969 struct sockaddr *nam, struct mbuf *control, struct thread *td)
971 struct epoch_tracker et;
974 struct tcpcb *tp = NULL;
977 struct sockaddr_in sin;
979 struct sockaddr_in *sinp;
990 * We require the pcbinfo "read lock" if we will close the socket
991 * as part of this call.
995 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
997 vflagsav = inp->inp_vflag;
998 incflagsav = inp->inp_inc.inc_flags;
999 restoreflags = false;
1000 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1006 if (control != NULL) {
1007 /* TCP doesn't do control messages (rights, creds, etc) */
1008 if (control->m_len) {
1013 m_freem(control); /* empty control, just free it */
1016 tp = intotcpcb(inp);
1017 if ((flags & PRUS_OOB) != 0 &&
1018 (error = tcp_pru_options_support(tp, PRUS_OOB)) != 0)
1022 if (nam != NULL && tp->t_state < TCPS_SYN_SENT) {
1023 switch (nam->sa_family) {
1026 sinp = (struct sockaddr_in *)nam;
1027 if (sinp->sin_len != sizeof(struct sockaddr_in)) {
1031 if ((inp->inp_vflag & INP_IPV6) != 0) {
1032 error = EAFNOSUPPORT;
1035 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
1036 error = EAFNOSUPPORT;
1039 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) {
1043 if ((error = prison_remote_ip4(td->td_ucred,
1054 struct sockaddr_in6 *sin6;
1056 sin6 = (struct sockaddr_in6 *)nam;
1057 if (sin6->sin6_len != sizeof(*sin6)) {
1061 if ((inp->inp_vflag & INP_IPV6PROTO) == 0) {
1062 error = EAFNOSUPPORT;
1065 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
1066 error = EAFNOSUPPORT;
1069 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1071 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1075 if ((inp->inp_vflag & INP_IPV4) == 0) {
1076 error = EAFNOSUPPORT;
1079 restoreflags = true;
1080 inp->inp_vflag &= ~INP_IPV6;
1082 in6_sin6_2_sin(sinp, sin6);
1084 ntohl(sinp->sin_addr.s_addr))) {
1085 error = EAFNOSUPPORT;
1088 if ((error = prison_remote_ip4(td->td_ucred,
1093 error = EAFNOSUPPORT;
1097 if ((inp->inp_vflag & INP_IPV6) == 0) {
1098 error = EAFNOSUPPORT;
1101 restoreflags = true;
1102 inp->inp_vflag &= ~INP_IPV4;
1103 inp->inp_inc.inc_flags |= INC_ISIPV6;
1104 if ((error = prison_remote_ip6(td->td_ucred,
1113 error = EAFNOSUPPORT;
1117 if (!(flags & PRUS_OOB)) {
1118 sbappendstream(&so->so_snd, m, flags);
1120 if (nam && tp->t_state < TCPS_SYN_SENT) {
1122 * Do implied connect if not yet connected,
1123 * initialize window to default value, and
1124 * initialize maxseg using peer's cached MSS.
1128 error = tcp6_connect(tp, nam, td);
1130 #if defined(INET6) && defined(INET)
1134 error = tcp_connect(tp,
1135 (struct sockaddr *)sinp, td);
1138 * The bind operation in tcp_connect succeeded. We
1139 * no longer want to restore the flags if later
1142 if (error == 0 || inp->inp_lport != 0)
1143 restoreflags = false;
1146 /* m is freed if PRUS_NOTREADY is unset. */
1147 sbflush(&so->so_snd);
1150 if (IS_FASTOPEN(tp->t_flags))
1151 tcp_fastopen_connect(tp);
1153 tp->snd_wnd = TTCP_CLIENT_SND_WND;
1157 if (flags & PRUS_EOF) {
1159 * Close the send side of the connection after
1165 if (TCPS_HAVEESTABLISHED(tp->t_state) &&
1166 ((tp->t_flags2 & TF2_FBYTES_COMPLETE) == 0) &&
1167 (tp->t_fbyte_out == 0) &&
1168 (so->so_snd.sb_ccc > 0)) {
1169 tp->t_fbyte_out = ticks;
1170 if (tp->t_fbyte_out == 0)
1171 tp->t_fbyte_out = 1;
1172 if (tp->t_fbyte_out && tp->t_fbyte_in)
1173 tp->t_flags2 |= TF2_FBYTES_COMPLETE;
1175 if (!(inp->inp_flags & INP_DROPPED) &&
1176 !(flags & PRUS_NOTREADY)) {
1177 if (flags & PRUS_MORETOCOME)
1178 tp->t_flags |= TF_MORETOCOME;
1179 error = tp->t_fb->tfb_tcp_output(tp);
1180 if (flags & PRUS_MORETOCOME)
1181 tp->t_flags &= ~TF_MORETOCOME;
1185 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
1187 SOCKBUF_LOCK(&so->so_snd);
1188 if (sbspace(&so->so_snd) < -512) {
1189 SOCKBUF_UNLOCK(&so->so_snd);
1194 * According to RFC961 (Assigned Protocols),
1195 * the urgent pointer points to the last octet
1196 * of urgent data. We continue, however,
1197 * to consider it to indicate the first octet
1198 * of data past the urgent section.
1199 * Otherwise, snd_up should be one lower.
1201 sbappendstream_locked(&so->so_snd, m, flags);
1202 SOCKBUF_UNLOCK(&so->so_snd);
1204 if (nam && tp->t_state < TCPS_SYN_SENT) {
1206 * Do implied connect if not yet connected,
1207 * initialize window to default value, and
1208 * initialize maxseg using peer's cached MSS.
1212 * Not going to contemplate SYN|URG
1214 if (IS_FASTOPEN(tp->t_flags))
1215 tp->t_flags &= ~TF_FASTOPEN;
1218 error = tcp6_connect(tp, nam, td);
1220 #if defined(INET6) && defined(INET)
1224 error = tcp_connect(tp,
1225 (struct sockaddr *)sinp, td);
1228 * The bind operation in tcp_connect succeeded. We
1229 * no longer want to restore the flags if later
1232 if (error == 0 || inp->inp_lport != 0)
1233 restoreflags = false;
1236 /* m is freed if PRUS_NOTREADY is unset. */
1237 sbflush(&so->so_snd);
1240 tp->snd_wnd = TTCP_CLIENT_SND_WND;
1243 tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
1244 if ((flags & PRUS_NOTREADY) == 0) {
1245 tp->t_flags |= TF_FORCEDATA;
1246 error = tp->t_fb->tfb_tcp_output(tp);
1247 tp->t_flags &= ~TF_FORCEDATA;
1250 TCP_LOG_EVENT(tp, NULL,
1251 &inp->inp_socket->so_rcv,
1252 &inp->inp_socket->so_snd,
1253 TCP_LOG_USERSEND, error,
1258 * In case of PRUS_NOTREADY, the caller or tcp_usr_ready() is
1259 * responsible for freeing memory.
1261 if (m != NULL && (flags & PRUS_NOTREADY) == 0)
1265 * If the request was unsuccessful and we changed flags,
1266 * restore the original flags.
1268 if (error != 0 && restoreflags) {
1269 inp->inp_vflag = vflagsav;
1270 inp->inp_inc.inc_flags = incflagsav;
1272 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
1273 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1274 TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
1275 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1282 tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
1284 struct epoch_tracker et;
1289 inp = sotoinpcb(so);
1291 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1293 mb_free_notready(m, count);
1294 return (ECONNRESET);
1296 tp = intotcpcb(inp);
1298 SOCKBUF_LOCK(&so->so_snd);
1299 error = sbready(&so->so_snd, m, count);
1300 SOCKBUF_UNLOCK(&so->so_snd);
1302 NET_EPOCH_ENTER(et);
1303 error = tp->t_fb->tfb_tcp_output(tp);
1312 * Abort the TCP. Drop the connection abruptly.
1315 tcp_usr_abort(struct socket *so)
1318 struct tcpcb *tp = NULL;
1319 struct epoch_tracker et;
1322 inp = sotoinpcb(so);
1323 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1325 NET_EPOCH_ENTER(et);
1327 KASSERT(inp->inp_socket != NULL,
1328 ("tcp_usr_abort: inp_socket == NULL"));
1331 * If we still have full TCP state, and we're not dropped, drop.
1333 if (!(inp->inp_flags & INP_TIMEWAIT) &&
1334 !(inp->inp_flags & INP_DROPPED)) {
1335 tp = intotcpcb(inp);
1337 tp = tcp_drop(tp, ECONNABORTED);
1340 TCPDEBUG2(PRU_ABORT);
1341 TCP_PROBE2(debug__user, tp, PRU_ABORT);
1343 if (!(inp->inp_flags & INP_DROPPED)) {
1345 so->so_state |= SS_PROTOREF;
1347 inp->inp_flags |= INP_SOCKREF;
1355 * TCP socket is closed. Start friendly disconnect.
1358 tcp_usr_close(struct socket *so)
1361 struct tcpcb *tp = NULL;
1362 struct epoch_tracker et;
1365 inp = sotoinpcb(so);
1366 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1368 NET_EPOCH_ENTER(et);
1370 KASSERT(inp->inp_socket != NULL,
1371 ("tcp_usr_close: inp_socket == NULL"));
1374 * If we still have full TCP state, and we're not dropped, initiate
1377 if (!(inp->inp_flags & INP_TIMEWAIT) &&
1378 !(inp->inp_flags & INP_DROPPED)) {
1379 tp = intotcpcb(inp);
1382 TCPDEBUG2(PRU_CLOSE);
1383 TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1385 if (!(inp->inp_flags & INP_DROPPED)) {
1387 so->so_state |= SS_PROTOREF;
1389 inp->inp_flags |= INP_SOCKREF;
1396 tcp_pru_options_support(struct tcpcb *tp, int flags)
1399 * If the specific TCP stack has a pru_options
1400 * specified then it does not always support
1401 * all the PRU_XX options and we must ask it.
1402 * If the function is not specified then all
1403 * of the PRU_XX options are supported.
1407 if (tp->t_fb->tfb_pru_options) {
1408 ret = (*tp->t_fb->tfb_pru_options)(tp, flags);
1414 * Receive out-of-band data.
1417 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1421 struct tcpcb *tp = NULL;
1424 inp = sotoinpcb(so);
1425 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1427 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1431 tp = intotcpcb(inp);
1432 error = tcp_pru_options_support(tp, PRUS_OOB);
1437 if ((so->so_oobmark == 0 &&
1438 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1439 so->so_options & SO_OOBINLINE ||
1440 tp->t_oobflags & TCPOOB_HADDATA) {
1444 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1445 error = EWOULDBLOCK;
1449 *mtod(m, caddr_t) = tp->t_iobc;
1450 if ((flags & MSG_PEEK) == 0)
1451 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1454 TCPDEBUG2(PRU_RCVOOB);
1455 TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1461 struct pr_usrreqs tcp_usrreqs = {
1462 .pru_abort = tcp_usr_abort,
1463 .pru_accept = tcp_usr_accept,
1464 .pru_attach = tcp_usr_attach,
1465 .pru_bind = tcp_usr_bind,
1466 .pru_connect = tcp_usr_connect,
1467 .pru_control = in_control,
1468 .pru_detach = tcp_usr_detach,
1469 .pru_disconnect = tcp_usr_disconnect,
1470 .pru_listen = tcp_usr_listen,
1471 .pru_peeraddr = in_getpeeraddr,
1472 .pru_rcvd = tcp_usr_rcvd,
1473 .pru_rcvoob = tcp_usr_rcvoob,
1474 .pru_send = tcp_usr_send,
1475 .pru_ready = tcp_usr_ready,
1476 .pru_shutdown = tcp_usr_shutdown,
1477 .pru_sockaddr = in_getsockaddr,
1478 .pru_sosetlabel = in_pcbsosetlabel,
1479 .pru_close = tcp_usr_close,
1484 struct pr_usrreqs tcp6_usrreqs = {
1485 .pru_abort = tcp_usr_abort,
1486 .pru_accept = tcp6_usr_accept,
1487 .pru_attach = tcp_usr_attach,
1488 .pru_bind = tcp6_usr_bind,
1489 .pru_connect = tcp6_usr_connect,
1490 .pru_control = in6_control,
1491 .pru_detach = tcp_usr_detach,
1492 .pru_disconnect = tcp_usr_disconnect,
1493 .pru_listen = tcp6_usr_listen,
1494 .pru_peeraddr = in6_mapped_peeraddr,
1495 .pru_rcvd = tcp_usr_rcvd,
1496 .pru_rcvoob = tcp_usr_rcvoob,
1497 .pru_send = tcp_usr_send,
1498 .pru_ready = tcp_usr_ready,
1499 .pru_shutdown = tcp_usr_shutdown,
1500 .pru_sockaddr = in6_mapped_sockaddr,
1501 .pru_sosetlabel = in_pcbsosetlabel,
1502 .pru_close = tcp_usr_close,
1508 * Common subroutine to open a TCP connection to remote host specified
1509 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
1510 * port number if needed. Call in_pcbconnect_setup to do the routing and
1511 * to choose a local host address (interface). If there is an existing
1512 * incarnation of the same connection in TIME-WAIT state and if the remote
1513 * host was sending CC options and if the connection duration was < MSL, then
1514 * truncate the previous TIME-WAIT state and proceed.
1515 * Initialize connection parameters and enter SYN-SENT state.
1518 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1520 struct inpcb *inp = tp->t_inpcb, *oinp;
1521 struct socket *so = inp->inp_socket;
1522 struct in_addr laddr;
1527 INP_WLOCK_ASSERT(inp);
1528 INP_HASH_WLOCK(&V_tcbinfo);
1530 if (V_tcp_require_unique_port && inp->inp_lport == 0) {
1531 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1537 * Cannot simply call in_pcbconnect, because there might be an
1538 * earlier incarnation of this same connection still in
1539 * TIME_WAIT state, creating an ADDRINUSE error.
1541 laddr = inp->inp_laddr;
1542 lport = inp->inp_lport;
1543 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1544 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1545 if (error && oinp == NULL)
1551 /* Handle initial bind if it hadn't been done in advance. */
1552 if (inp->inp_lport == 0) {
1553 inp->inp_lport = lport;
1554 if (in_pcbinshash(inp) != 0) {
1560 inp->inp_laddr = laddr;
1562 INP_HASH_WUNLOCK(&V_tcbinfo);
1565 * Compute window scaling to request:
1566 * Scale to fit into sweet spot. See tcp_syncache.c.
1567 * XXX: This should move to tcp_output().
1569 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1570 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1571 tp->request_r_scale++;
1574 TCPSTAT_INC(tcps_connattempt);
1575 tcp_state_change(tp, TCPS_SYN_SENT);
1576 tp->iss = tcp_new_isn(&inp->inp_inc);
1577 if (tp->t_flags & TF_REQ_TSTMP)
1578 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1579 tcp_sendseqinit(tp);
1584 INP_HASH_WUNLOCK(&V_tcbinfo);
1591 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1593 struct inpcb *inp = tp->t_inpcb;
1596 INP_WLOCK_ASSERT(inp);
1597 INP_HASH_WLOCK(&V_tcbinfo);
1599 if (V_tcp_require_unique_port && inp->inp_lport == 0) {
1600 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1604 error = in6_pcbconnect(inp, nam, td->td_ucred);
1607 INP_HASH_WUNLOCK(&V_tcbinfo);
1609 /* Compute window scaling to request. */
1610 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1611 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1612 tp->request_r_scale++;
1614 soisconnecting(inp->inp_socket);
1615 TCPSTAT_INC(tcps_connattempt);
1616 tcp_state_change(tp, TCPS_SYN_SENT);
1617 tp->iss = tcp_new_isn(&inp->inp_inc);
1618 if (tp->t_flags & TF_REQ_TSTMP)
1619 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1620 tcp_sendseqinit(tp);
1625 INP_HASH_WUNLOCK(&V_tcbinfo);
1631 * Export TCP internal state information via a struct tcp_info, based on the
1632 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently
1633 * (TCP state machine, etc). We export all information using FreeBSD-native
1634 * constants -- for example, the numeric values for tcpi_state will differ
1638 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1641 INP_WLOCK_ASSERT(tp->t_inpcb);
1642 bzero(ti, sizeof(*ti));
1644 ti->tcpi_state = tp->t_state;
1645 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1646 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1647 if (tp->t_flags & TF_SACK_PERMIT)
1648 ti->tcpi_options |= TCPI_OPT_SACK;
1649 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1650 ti->tcpi_options |= TCPI_OPT_WSCALE;
1651 ti->tcpi_snd_wscale = tp->snd_scale;
1652 ti->tcpi_rcv_wscale = tp->rcv_scale;
1654 if (tp->t_flags2 & TF2_ECN_PERMIT)
1655 ti->tcpi_options |= TCPI_OPT_ECN;
1657 ti->tcpi_rto = tp->t_rxtcur * tick;
1658 ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick;
1659 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1660 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1662 ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1663 ti->tcpi_snd_cwnd = tp->snd_cwnd;
1666 * FreeBSD-specific extension fields for tcp_info.
1668 ti->tcpi_rcv_space = tp->rcv_wnd;
1669 ti->tcpi_rcv_nxt = tp->rcv_nxt;
1670 ti->tcpi_snd_wnd = tp->snd_wnd;
1671 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */
1672 ti->tcpi_snd_nxt = tp->snd_nxt;
1673 ti->tcpi_snd_mss = tp->t_maxseg;
1674 ti->tcpi_rcv_mss = tp->t_maxseg;
1675 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1676 ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1677 ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1679 if (tp->t_flags & TF_TOE) {
1680 ti->tcpi_options |= TCPI_OPT_TOE;
1681 tcp_offload_tcp_info(tp, ti);
1687 * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1688 * socket option arguments. When it re-acquires the lock after the copy, it
1689 * has to revalidate that the connection is still valid for the socket
1692 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do { \
1694 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
1697 return (ECONNRESET); \
1699 tp = intotcpcb(inp); \
1701 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */)
1704 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1709 struct tcp_function_block *blk;
1710 struct tcp_function_set fsn;
1713 inp = sotoinpcb(so);
1714 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1715 if (sopt->sopt_level != IPPROTO_TCP) {
1717 if (inp->inp_vflag & INP_IPV6PROTO) {
1718 error = ip6_ctloutput(so, sopt);
1720 * In case of the IPV6_USE_MIN_MTU socket option,
1721 * the INC_IPV6MINMTU flag to announce a corresponding
1722 * MSS during the initial handshake.
1723 * If the TCP connection is not in the front states,
1724 * just reduce the MSS being used.
1725 * This avoids the sending of TCP segments which will
1726 * be fragmented at the IPv6 layer.
1729 (sopt->sopt_dir == SOPT_SET) &&
1730 (sopt->sopt_level == IPPROTO_IPV6) &&
1731 (sopt->sopt_name == IPV6_USE_MIN_MTU)) {
1733 if ((inp->inp_flags &
1734 (INP_TIMEWAIT | INP_DROPPED))) {
1736 return (ECONNRESET);
1738 inp->inp_inc.inc_flags |= INC_IPV6MINMTU;
1739 tp = intotcpcb(inp);
1740 if ((tp->t_state >= TCPS_SYN_SENT) &&
1741 (inp->inp_inc.inc_flags & INC_ISIPV6)) {
1742 struct ip6_pktopts *opt;
1744 opt = inp->in6p_outputopts;
1745 if ((opt != NULL) &&
1746 (opt->ip6po_minmtu ==
1747 IP6PO_MINMTU_ALL)) {
1748 if (tp->t_maxseg > TCP6_MSS) {
1749 tp->t_maxseg = TCP6_MSS;
1757 #if defined(INET6) && defined(INET)
1762 error = ip_ctloutput(so, sopt);
1768 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1770 return (ECONNRESET);
1772 tp = intotcpcb(inp);
1774 * Protect the TCP option TCP_FUNCTION_BLK so
1775 * that a sub-function can *never* overwrite this.
1777 if ((sopt->sopt_dir == SOPT_SET) &&
1778 (sopt->sopt_name == TCP_FUNCTION_BLK)) {
1780 error = sooptcopyin(sopt, &fsn, sizeof fsn,
1784 INP_WLOCK_RECHECK(inp);
1785 blk = find_and_ref_tcp_functions(&fsn);
1790 if (tp->t_fb == blk) {
1791 /* You already have this */
1792 refcount_release(&blk->tfb_refcnt);
1796 if (tp->t_state != TCPS_CLOSED) {
1798 * The user has advanced the state
1799 * past the initial point, we may not
1800 * be able to switch.
1802 if (blk->tfb_tcp_handoff_ok != NULL) {
1804 * Does the stack provide a
1805 * query mechanism, if so it may
1806 * still be possible?
1808 error = (*blk->tfb_tcp_handoff_ok)(tp);
1812 refcount_release(&blk->tfb_refcnt);
1817 if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) {
1818 refcount_release(&blk->tfb_refcnt);
1823 * Release the old refcnt, the
1824 * lookup acquired a ref on the
1827 if (tp->t_fb->tfb_tcp_fb_fini) {
1828 struct epoch_tracker et;
1830 * Tell the stack to cleanup with 0 i.e.
1831 * the tcb is not going away.
1833 NET_EPOCH_ENTER(et);
1834 (*tp->t_fb->tfb_tcp_fb_fini)(tp, 0);
1838 /* Assure that we are not on any hpts */
1839 tcp_hpts_remove(tp->t_inpcb, HPTS_REMOVE_ALL);
1841 if (blk->tfb_tcp_fb_init) {
1842 error = (*blk->tfb_tcp_fb_init)(tp);
1844 refcount_release(&blk->tfb_refcnt);
1845 if (tp->t_fb->tfb_tcp_fb_init) {
1846 if((*tp->t_fb->tfb_tcp_fb_init)(tp) != 0) {
1847 /* Fall back failed, drop the connection */
1856 refcount_release(&tp->t_fb->tfb_refcnt);
1859 if (tp->t_flags & TF_TOE) {
1860 tcp_offload_ctloutput(tp, sopt->sopt_dir,
1867 } else if ((sopt->sopt_dir == SOPT_GET) &&
1868 (sopt->sopt_name == TCP_FUNCTION_BLK)) {
1869 strncpy(fsn.function_set_name, tp->t_fb->tfb_tcp_block_name,
1870 TCP_FUNCTION_NAME_LEN_MAX);
1871 fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0';
1872 fsn.pcbcnt = tp->t_fb->tfb_refcnt;
1874 error = sooptcopyout(sopt, &fsn, sizeof fsn);
1877 /* Pass in the INP locked, called must unlock it */
1878 return (tp->t_fb->tfb_tcp_ctloutput(so, sopt, inp, tp));
1882 * If this assert becomes untrue, we need to change the size of the buf
1883 * variable in tcp_default_ctloutput().
1886 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN);
1887 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN);
1892 copyin_tls_enable(struct sockopt *sopt, struct tls_enable *tls)
1894 struct tls_enable_v0 tls_v0;
1897 if (sopt->sopt_valsize == sizeof(tls_v0)) {
1898 error = sooptcopyin(sopt, &tls_v0, sizeof(tls_v0),
1902 memset(tls, 0, sizeof(*tls));
1903 tls->cipher_key = tls_v0.cipher_key;
1904 tls->iv = tls_v0.iv;
1905 tls->auth_key = tls_v0.auth_key;
1906 tls->cipher_algorithm = tls_v0.cipher_algorithm;
1907 tls->cipher_key_len = tls_v0.cipher_key_len;
1908 tls->iv_len = tls_v0.iv_len;
1909 tls->auth_algorithm = tls_v0.auth_algorithm;
1910 tls->auth_key_len = tls_v0.auth_key_len;
1911 tls->flags = tls_v0.flags;
1912 tls->tls_vmajor = tls_v0.tls_vmajor;
1913 tls->tls_vminor = tls_v0.tls_vminor;
1917 return (sooptcopyin(sopt, tls, sizeof(*tls), sizeof(*tls)));
1922 tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp)
1924 int error, opt, optval;
1928 struct tls_enable tls;
1930 struct cc_algo *algo;
1931 char *pbuf, buf[TCP_LOG_ID_LEN];
1933 struct statsblob *sbp;
1938 * For TCP_CCALGOOPT forward the control to CC module, for both
1939 * SOPT_SET and SOPT_GET.
1941 switch (sopt->sopt_name) {
1944 if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT)
1946 pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO);
1947 error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize,
1948 sopt->sopt_valsize);
1953 INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP));
1954 if (CC_ALGO(tp)->ctl_output != NULL)
1955 error = CC_ALGO(tp)->ctl_output(tp->ccv, sopt, pbuf);
1959 if (error == 0 && sopt->sopt_dir == SOPT_GET)
1960 error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize);
1965 switch (sopt->sopt_dir) {
1967 switch (sopt->sopt_name) {
1968 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
1970 if (!TCPMD5_ENABLED()) {
1972 return (ENOPROTOOPT);
1974 error = TCPMD5_PCBCTL(inp, sopt);
1977 goto unlock_and_done;
1984 error = sooptcopyin(sopt, &optval, sizeof optval,
1989 INP_WLOCK_RECHECK(inp);
1990 switch (sopt->sopt_name) {
2001 opt = 0; /* dead code to fool gcc */
2008 tp->t_flags &= ~opt;
2011 if (tp->t_flags & TF_TOE) {
2012 tcp_offload_ctloutput(tp, sopt->sopt_dir,
2021 error = sooptcopyin(sopt, &optval, sizeof optval,
2026 INP_WLOCK_RECHECK(inp);
2028 tp->t_flags |= TF_NOPUSH;
2029 else if (tp->t_flags & TF_NOPUSH) {
2030 tp->t_flags &= ~TF_NOPUSH;
2031 if (TCPS_HAVEESTABLISHED(tp->t_state)) {
2032 struct epoch_tracker et;
2034 NET_EPOCH_ENTER(et);
2035 error = tp->t_fb->tfb_tcp_output(tp);
2039 goto unlock_and_done;
2041 case TCP_REMOTE_UDP_ENCAPS_PORT:
2043 error = sooptcopyin(sopt, &optval, sizeof optval,
2047 if ((optval < TCP_TUNNELING_PORT_MIN) ||
2048 (optval > TCP_TUNNELING_PORT_MAX)) {
2049 /* Its got to be in range */
2052 if ((V_tcp_udp_tunneling_port == 0) && (optval != 0)) {
2053 /* You have to have enabled a UDP tunneling port first */
2056 INP_WLOCK_RECHECK(inp);
2057 if (tp->t_state != TCPS_CLOSED) {
2058 /* You can't change after you are connected */
2061 /* Ok we are all good set the port */
2062 tp->t_port = htons(optval);
2064 goto unlock_and_done;
2068 error = sooptcopyin(sopt, &optval, sizeof optval,
2073 INP_WLOCK_RECHECK(inp);
2074 if (optval > 0 && optval <= tp->t_maxseg &&
2075 optval + 40 >= V_tcp_minmss)
2076 tp->t_maxseg = optval;
2079 goto unlock_and_done;
2089 error = sooptcopyin(sopt, &optval, sizeof optval,
2095 sbp = stats_blob_alloc(
2096 V_tcp_perconn_stats_dflt_tpl, 0);
2100 INP_WLOCK_RECHECK(inp);
2101 if ((tp->t_stats != NULL && sbp == NULL) ||
2102 (tp->t_stats == NULL && sbp != NULL)) {
2103 struct statsblob *t = tp->t_stats;
2109 stats_blob_destroy(sbp);
2111 return (EOPNOTSUPP);
2115 case TCP_CONGESTION:
2117 error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1);
2120 buf[sopt->sopt_valsize] = '\0';
2121 INP_WLOCK_RECHECK(inp);
2123 STAILQ_FOREACH(algo, &cc_list, entries)
2124 if (strncmp(buf, algo->name,
2125 TCP_CA_NAME_MAX) == 0)
2134 * We hold a write lock over the tcb so it's safe to
2135 * do these things without ordering concerns.
2137 if (CC_ALGO(tp)->cb_destroy != NULL)
2138 CC_ALGO(tp)->cb_destroy(tp->ccv);
2142 * If something goes pear shaped initialising the new
2143 * algo, fall back to newreno (which does not
2144 * require initialisation).
2146 if (algo->cb_init != NULL &&
2147 algo->cb_init(tp->ccv) != 0) {
2148 CC_ALGO(tp) = &newreno_cc_algo;
2150 * The only reason init should fail is
2151 * because of malloc.
2158 case TCP_REUSPORT_LB_NUMA:
2160 error = sooptcopyin(sopt, &optval, sizeof(optval),
2162 INP_WLOCK_RECHECK(inp);
2164 error = in_pcblbgroup_numa(inp, optval);
2169 case TCP_TXTLS_ENABLE:
2171 error = copyin_tls_enable(sopt, &tls);
2174 error = ktls_enable_tx(so, &tls);
2176 case TCP_TXTLS_MODE:
2178 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2182 INP_WLOCK_RECHECK(inp);
2183 error = ktls_set_tx_mode(so, ui);
2186 case TCP_RXTLS_ENABLE:
2188 error = sooptcopyin(sopt, &tls, sizeof(tls),
2192 error = ktls_enable_rx(so, &tls);
2200 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2204 if (ui > (UINT_MAX / hz)) {
2210 INP_WLOCK_RECHECK(inp);
2211 switch (sopt->sopt_name) {
2213 tp->t_keepidle = ui;
2215 * XXX: better check current remaining
2216 * timeout and "merge" it with new value.
2218 if ((tp->t_state > TCPS_LISTEN) &&
2219 (tp->t_state <= TCPS_CLOSING))
2220 tcp_timer_activate(tp, TT_KEEP,
2224 tp->t_keepintvl = ui;
2225 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
2226 (TP_MAXIDLE(tp) > 0))
2227 tcp_timer_activate(tp, TT_2MSL,
2231 tp->t_keepinit = ui;
2232 if (tp->t_state == TCPS_SYN_RECEIVED ||
2233 tp->t_state == TCPS_SYN_SENT)
2234 tcp_timer_activate(tp, TT_KEEP,
2238 goto unlock_and_done;
2242 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2246 INP_WLOCK_RECHECK(inp);
2248 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
2249 (TP_MAXIDLE(tp) > 0))
2250 tcp_timer_activate(tp, TT_2MSL,
2252 goto unlock_and_done;
2258 error = sooptcopyin(sopt, &optval, sizeof optval,
2263 INP_WLOCK_RECHECK(inp);
2265 tcp_pcap_set_sock_max(TCP_PCAP_OUT ?
2266 &(tp->t_outpkts) : &(tp->t_inpkts),
2270 goto unlock_and_done;
2273 case TCP_FASTOPEN: {
2274 struct tcp_fastopen tfo_optval;
2277 if (!V_tcp_fastopen_client_enable &&
2278 !V_tcp_fastopen_server_enable)
2281 error = sooptcopyin(sopt, &tfo_optval,
2282 sizeof(tfo_optval), sizeof(int));
2286 INP_WLOCK_RECHECK(inp);
2287 if ((tp->t_state != TCPS_CLOSED) &&
2288 (tp->t_state != TCPS_LISTEN)) {
2290 goto unlock_and_done;
2292 if (tfo_optval.enable) {
2293 if (tp->t_state == TCPS_LISTEN) {
2294 if (!V_tcp_fastopen_server_enable) {
2296 goto unlock_and_done;
2299 if (tp->t_tfo_pending == NULL)
2301 tcp_fastopen_alloc_counter();
2304 * If a pre-shared key was provided,
2305 * stash it in the client cookie
2306 * field of the tcpcb for use during
2309 if (sopt->sopt_valsize ==
2310 sizeof(tfo_optval)) {
2311 memcpy(tp->t_tfo_cookie.client,
2313 TCP_FASTOPEN_PSK_LEN);
2314 tp->t_tfo_client_cookie_len =
2315 TCP_FASTOPEN_PSK_LEN;
2318 tp->t_flags |= TF_FASTOPEN;
2320 tp->t_flags &= ~TF_FASTOPEN;
2321 goto unlock_and_done;
2327 error = sooptcopyin(sopt, &optval, sizeof optval,
2332 INP_WLOCK_RECHECK(inp);
2333 error = tcp_log_state_change(tp, optval);
2334 goto unlock_and_done;
2343 error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0);
2346 buf[sopt->sopt_valsize] = '\0';
2347 INP_WLOCK_RECHECK(inp);
2348 error = tcp_log_set_id(tp, buf);
2349 /* tcp_log_set_id() unlocks the INP. */
2356 sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0);
2359 buf[sopt->sopt_valsize] = '\0';
2360 INP_WLOCK_RECHECK(inp);
2361 if (sopt->sopt_name == TCP_LOGDUMP) {
2362 error = tcp_log_dump_tp_logbuf(tp, buf,
2366 tcp_log_dump_tp_bucket_logbufs(tp, buf);
2368 * tcp_log_dump_tp_bucket_logbufs() drops the
2377 error = ENOPROTOOPT;
2383 tp = intotcpcb(inp);
2384 switch (sopt->sopt_name) {
2385 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
2387 if (!TCPMD5_ENABLED()) {
2389 return (ENOPROTOOPT);
2391 error = TCPMD5_PCBCTL(inp, sopt);
2396 optval = tp->t_flags & TF_NODELAY;
2398 error = sooptcopyout(sopt, &optval, sizeof optval);
2401 optval = tp->t_maxseg;
2403 error = sooptcopyout(sopt, &optval, sizeof optval);
2405 case TCP_REMOTE_UDP_ENCAPS_PORT:
2406 optval = ntohs(tp->t_port);
2408 error = sooptcopyout(sopt, &optval, sizeof optval);
2411 optval = tp->t_flags & TF_NOOPT;
2413 error = sooptcopyout(sopt, &optval, sizeof optval);
2416 optval = tp->t_flags & TF_NOPUSH;
2418 error = sooptcopyout(sopt, &optval, sizeof optval);
2421 tcp_fill_info(tp, &ti);
2423 error = sooptcopyout(sopt, &ti, sizeof ti);
2429 TYPEOF_MEMBER(struct statsblob, flags) sbflags = 0;
2432 socklen_t outsbsz = sopt->sopt_valsize;
2433 if (tp->t_stats == NULL)
2435 else if (outsbsz >= tp->t_stats->cursz)
2436 outsbsz = tp->t_stats->cursz;
2437 else if (outsbsz >= sizeof(struct statsblob))
2438 outsbsz = sizeof(struct statsblob);
2445 sbp = sopt->sopt_val;
2446 nheld = atop(round_page(((vm_offset_t)sbp) +
2447 (vm_size_t)outsbsz) - trunc_page((vm_offset_t)sbp));
2448 vm_page_t ma[nheld];
2449 if (vm_fault_quick_hold_pages(
2450 &curproc->p_vmspace->vm_map, (vm_offset_t)sbp,
2451 outsbsz, VM_PROT_READ | VM_PROT_WRITE, ma,
2457 if ((error = copyin_nofault(&(sbp->flags), &sbflags,
2458 SIZEOF_MEMBER(struct statsblob, flags))))
2461 INP_WLOCK_RECHECK(inp);
2462 error = stats_blob_snapshot(&sbp, outsbsz, tp->t_stats,
2463 sbflags | SB_CLONE_USRDSTNOFAULT);
2465 sopt->sopt_valsize = outsbsz;
2467 vm_page_unhold_pages(ma, nheld);
2474 case TCP_CONGESTION:
2475 len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
2477 error = sooptcopyout(sopt, buf, len + 1);
2483 switch (sopt->sopt_name) {
2485 ui = TP_KEEPIDLE(tp) / hz;
2488 ui = TP_KEEPINTVL(tp) / hz;
2491 ui = TP_KEEPINIT(tp) / hz;
2494 ui = TP_KEEPCNT(tp);
2498 error = sooptcopyout(sopt, &ui, sizeof(ui));
2503 optval = tcp_pcap_get_sock_max(TCP_PCAP_OUT ?
2504 &(tp->t_outpkts) : &(tp->t_inpkts));
2506 error = sooptcopyout(sopt, &optval, sizeof optval);
2510 optval = tp->t_flags & TF_FASTOPEN;
2512 error = sooptcopyout(sopt, &optval, sizeof optval);
2516 optval = tp->t_logstate;
2518 error = sooptcopyout(sopt, &optval, sizeof(optval));
2521 /* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */
2522 error = tcp_log_getlogbuf(sopt, tp);
2525 len = tcp_log_get_id(tp, buf);
2527 error = sooptcopyout(sopt, buf, len + 1);
2536 case TCP_TXTLS_MODE:
2537 optval = ktls_get_tx_mode(so);
2539 error = sooptcopyout(sopt, &optval, sizeof(optval));
2541 case TCP_RXTLS_MODE:
2542 optval = ktls_get_rx_mode(so);
2544 error = sooptcopyout(sopt, &optval, sizeof(optval));
2548 optval = tp->t_flags & TF_LRD;
2550 error = sooptcopyout(sopt, &optval, sizeof optval);
2554 error = ENOPROTOOPT;
2561 #undef INP_WLOCK_RECHECK
2562 #undef INP_WLOCK_RECHECK_CLEANUP
2565 * Initiate (or continue) disconnect.
2566 * If embryonic state, just send reset (once).
2567 * If in ``let data drain'' option and linger null, just drop.
2568 * Otherwise (hard), mark socket disconnecting and drop
2569 * current input data; switch states based on user close, and
2570 * send segment to peer (with FIN).
2573 tcp_disconnect(struct tcpcb *tp)
2575 struct inpcb *inp = tp->t_inpcb;
2576 struct socket *so = inp->inp_socket;
2579 INP_WLOCK_ASSERT(inp);
2582 * Neither tcp_close() nor tcp_drop() should return NULL, as the
2583 * socket is still open.
2585 if (tp->t_state < TCPS_ESTABLISHED &&
2586 !(tp->t_state > TCPS_LISTEN && IS_FASTOPEN(tp->t_flags))) {
2589 ("tcp_disconnect: tcp_close() returned NULL"));
2590 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
2591 tp = tcp_drop(tp, 0);
2593 ("tcp_disconnect: tcp_drop() returned NULL"));
2595 soisdisconnecting(so);
2596 sbflush(&so->so_rcv);
2598 if (!(inp->inp_flags & INP_DROPPED))
2599 tp->t_fb->tfb_tcp_output(tp);
2604 * User issued close, and wish to trail through shutdown states:
2605 * if never received SYN, just forget it. If got a SYN from peer,
2606 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
2607 * If already got a FIN from peer, then almost done; go to LAST_ACK
2608 * state. In all other cases, have already sent FIN to peer (e.g.
2609 * after PRU_SHUTDOWN), and just have to play tedious game waiting
2610 * for peer to send FIN or not respond to keep-alives, etc.
2611 * We can let the user exit from the close as soon as the FIN is acked.
2614 tcp_usrclosed(struct tcpcb *tp)
2618 INP_WLOCK_ASSERT(tp->t_inpcb);
2620 switch (tp->t_state) {
2623 tcp_offload_listen_stop(tp);
2625 tcp_state_change(tp, TCPS_CLOSED);
2630 * tcp_close() should never return NULL here as the socket is
2634 ("tcp_usrclosed: tcp_close() returned NULL"));
2638 case TCPS_SYN_RECEIVED:
2639 tp->t_flags |= TF_NEEDFIN;
2642 case TCPS_ESTABLISHED:
2643 tcp_state_change(tp, TCPS_FIN_WAIT_1);
2646 case TCPS_CLOSE_WAIT:
2647 tcp_state_change(tp, TCPS_LAST_ACK);
2650 if (tp->t_state >= TCPS_FIN_WAIT_2) {
2651 soisdisconnected(tp->t_inpcb->inp_socket);
2652 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
2653 if (tp->t_state == TCPS_FIN_WAIT_2) {
2656 timeout = (tcp_fast_finwait2_recycle) ?
2657 tcp_finwait2_timeout : TP_MAXIDLE(tp);
2658 tcp_timer_activate(tp, TT_2MSL, timeout);
2665 db_print_indent(int indent)
2669 for (i = 0; i < indent; i++)
2674 db_print_tstate(int t_state)
2679 db_printf("TCPS_CLOSED");
2683 db_printf("TCPS_LISTEN");
2687 db_printf("TCPS_SYN_SENT");
2690 case TCPS_SYN_RECEIVED:
2691 db_printf("TCPS_SYN_RECEIVED");
2694 case TCPS_ESTABLISHED:
2695 db_printf("TCPS_ESTABLISHED");
2698 case TCPS_CLOSE_WAIT:
2699 db_printf("TCPS_CLOSE_WAIT");
2702 case TCPS_FIN_WAIT_1:
2703 db_printf("TCPS_FIN_WAIT_1");
2707 db_printf("TCPS_CLOSING");
2711 db_printf("TCPS_LAST_ACK");
2714 case TCPS_FIN_WAIT_2:
2715 db_printf("TCPS_FIN_WAIT_2");
2718 case TCPS_TIME_WAIT:
2719 db_printf("TCPS_TIME_WAIT");
2723 db_printf("unknown");
2729 db_print_tflags(u_int t_flags)
2734 if (t_flags & TF_ACKNOW) {
2735 db_printf("%sTF_ACKNOW", comma ? ", " : "");
2738 if (t_flags & TF_DELACK) {
2739 db_printf("%sTF_DELACK", comma ? ", " : "");
2742 if (t_flags & TF_NODELAY) {
2743 db_printf("%sTF_NODELAY", comma ? ", " : "");
2746 if (t_flags & TF_NOOPT) {
2747 db_printf("%sTF_NOOPT", comma ? ", " : "");
2750 if (t_flags & TF_SENTFIN) {
2751 db_printf("%sTF_SENTFIN", comma ? ", " : "");
2754 if (t_flags & TF_REQ_SCALE) {
2755 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
2758 if (t_flags & TF_RCVD_SCALE) {
2759 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
2762 if (t_flags & TF_REQ_TSTMP) {
2763 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
2766 if (t_flags & TF_RCVD_TSTMP) {
2767 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
2770 if (t_flags & TF_SACK_PERMIT) {
2771 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
2774 if (t_flags & TF_NEEDSYN) {
2775 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
2778 if (t_flags & TF_NEEDFIN) {
2779 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
2782 if (t_flags & TF_NOPUSH) {
2783 db_printf("%sTF_NOPUSH", comma ? ", " : "");
2786 if (t_flags & TF_MORETOCOME) {
2787 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
2790 if (t_flags & TF_LQ_OVERFLOW) {
2791 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
2794 if (t_flags & TF_LASTIDLE) {
2795 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
2798 if (t_flags & TF_RXWIN0SENT) {
2799 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
2802 if (t_flags & TF_FASTRECOVERY) {
2803 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
2806 if (t_flags & TF_CONGRECOVERY) {
2807 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
2810 if (t_flags & TF_WASFRECOVERY) {
2811 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
2814 if (t_flags & TF_SIGNATURE) {
2815 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
2818 if (t_flags & TF_FORCEDATA) {
2819 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
2822 if (t_flags & TF_TSO) {
2823 db_printf("%sTF_TSO", comma ? ", " : "");
2826 if (t_flags & TF_FASTOPEN) {
2827 db_printf("%sTF_FASTOPEN", comma ? ", " : "");
2833 db_print_tflags2(u_int t_flags2)
2838 if (t_flags2 & TF2_ECN_PERMIT) {
2839 db_printf("%sTF2_ECN_PERMIT", comma ? ", " : "");
2845 db_print_toobflags(char t_oobflags)
2850 if (t_oobflags & TCPOOB_HAVEDATA) {
2851 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
2854 if (t_oobflags & TCPOOB_HADDATA) {
2855 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
2861 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
2864 db_print_indent(indent);
2865 db_printf("%s at %p\n", name, tp);
2869 db_print_indent(indent);
2870 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n",
2871 TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
2873 db_print_indent(indent);
2874 db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n",
2875 &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
2877 db_print_indent(indent);
2878 db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl,
2879 &tp->t_timers->tt_delack, tp->t_inpcb);
2881 db_print_indent(indent);
2882 db_printf("t_state: %d (", tp->t_state);
2883 db_print_tstate(tp->t_state);
2886 db_print_indent(indent);
2887 db_printf("t_flags: 0x%x (", tp->t_flags);
2888 db_print_tflags(tp->t_flags);
2891 db_print_indent(indent);
2892 db_printf("t_flags2: 0x%x (", tp->t_flags2);
2893 db_print_tflags2(tp->t_flags2);
2896 db_print_indent(indent);
2897 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n",
2898 tp->snd_una, tp->snd_max, tp->snd_nxt);
2900 db_print_indent(indent);
2901 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n",
2902 tp->snd_up, tp->snd_wl1, tp->snd_wl2);
2904 db_print_indent(indent);
2905 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n",
2906 tp->iss, tp->irs, tp->rcv_nxt);
2908 db_print_indent(indent);
2909 db_printf("rcv_adv: 0x%08x rcv_wnd: %u rcv_up: 0x%08x\n",
2910 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
2912 db_print_indent(indent);
2913 db_printf("snd_wnd: %u snd_cwnd: %u\n",
2914 tp->snd_wnd, tp->snd_cwnd);
2916 db_print_indent(indent);
2917 db_printf("snd_ssthresh: %u snd_recover: "
2918 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
2920 db_print_indent(indent);
2921 db_printf("t_rcvtime: %u t_startime: %u\n",
2922 tp->t_rcvtime, tp->t_starttime);
2924 db_print_indent(indent);
2925 db_printf("t_rttime: %u t_rtsq: 0x%08x\n",
2926 tp->t_rtttime, tp->t_rtseq);
2928 db_print_indent(indent);
2929 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n",
2930 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
2932 db_print_indent(indent);
2933 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u "
2934 "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
2937 db_print_indent(indent);
2938 db_printf("t_rttupdated: %lu max_sndwnd: %u t_softerror: %d\n",
2939 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
2941 db_print_indent(indent);
2942 db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
2943 db_print_toobflags(tp->t_oobflags);
2944 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc);
2946 db_print_indent(indent);
2947 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n",
2948 tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
2950 db_print_indent(indent);
2951 db_printf("ts_recent: %u ts_recent_age: %u\n",
2952 tp->ts_recent, tp->ts_recent_age);
2954 db_print_indent(indent);
2955 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: "
2956 "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
2958 db_print_indent(indent);
2959 db_printf("snd_ssthresh_prev: %u snd_recover_prev: 0x%08x "
2960 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2961 tp->snd_recover_prev, tp->t_badrxtwin);
2963 db_print_indent(indent);
2964 db_printf("snd_numholes: %d snd_holes first: %p\n",
2965 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2967 db_print_indent(indent);
2968 db_printf("snd_fack: 0x%08x rcv_numsacks: %d\n",
2969 tp->snd_fack, tp->rcv_numsacks);
2971 /* Skip sackblks, sackhint. */
2973 db_print_indent(indent);
2974 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n",
2975 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2978 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2983 db_printf("usage: show tcpcb <addr>\n");
2986 tp = (struct tcpcb *)addr;
2988 db_print_tcpcb(tp, "tcpcb", 0);