2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved.
5 * This code is derived from software contributed to The DragonFly Project
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * Copyright (c) 1982, 1986, 1988, 1993
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
67 * $FreeBSD: src/sys/netinet/tcp_usrreq.c,v 1.51.2.17 2002/10/11 11:46:44 ume Exp $
68 * $DragonFly: src/sys/netinet/tcp_usrreq.c,v 1.51 2008/09/29 20:52:23 dillon Exp $
71 #include "opt_ipsec.h"
72 #include "opt_inet6.h"
73 #include "opt_tcpdebug.h"
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/kernel.h>
78 #include <sys/malloc.h>
79 #include <sys/sysctl.h>
80 #include <sys/globaldata.h>
81 #include <sys/thread.h>
85 #include <sys/domain.h>
87 #include <sys/socket.h>
88 #include <sys/socketvar.h>
89 #include <sys/protosw.h>
91 #include <sys/thread2.h>
92 #include <sys/msgport2.h>
95 #include <net/netisr.h>
96 #include <net/route.h>
98 #include <net/netmsg2.h>
100 #include <netinet/in.h>
101 #include <netinet/in_systm.h>
103 #include <netinet/ip6.h>
105 #include <netinet/in_pcb.h>
107 #include <netinet6/in6_pcb.h>
109 #include <netinet/in_var.h>
110 #include <netinet/ip_var.h>
112 #include <netinet6/ip6_var.h>
114 #include <netinet/tcp.h>
115 #include <netinet/tcp_fsm.h>
116 #include <netinet/tcp_seq.h>
117 #include <netinet/tcp_timer.h>
118 #include <netinet/tcp_var.h>
119 #include <netinet/tcpip.h>
121 #include <netinet/tcp_debug.h>
125 #include <netinet6/ipsec.h>
129 * TCP protocol interface to socket abstraction.
131 extern char *tcpstates[]; /* XXX ??? */
133 static int tcp_attach (struct socket *, struct pru_attach_info *);
134 static int tcp_connect (struct tcpcb *, struct sockaddr *,
137 static int tcp6_connect (struct tcpcb *, struct sockaddr *,
140 static struct tcpcb *
141 tcp_disconnect (struct tcpcb *);
142 static struct tcpcb *
143 tcp_usrclosed (struct tcpcb *);
146 #define TCPDEBUG0 int ostate = 0
147 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0
148 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
149 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
153 #define TCPDEBUG2(req)
157 * TCP attaches to socket via pru_attach(), reserving space,
158 * and an internet control block.
161 tcp_usr_attach(struct socket *so, int proto, struct pru_attach_info *ai)
165 struct tcpcb *tp = 0;
176 error = tcp_attach(so, ai);
180 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
181 so->so_linger = TCP_LINGERTIME;
184 TCPDEBUG2(PRU_ATTACH);
190 * pru_detach() detaches the TCP protocol from the socket.
191 * If the protocol state is non-embryonic, then can't
192 * do this directly: have to initiate a pru_disconnect(),
193 * which may finish later; embryonic TCB's can just
197 tcp_usr_detach(struct socket *so)
208 * If the inp is already detached it may have been due to an async
209 * close. Just return as if no error occured.
217 * It's possible for the tcpcb (tp) to disconnect from the inp due
218 * to tcp_drop()->tcp_close() being called. This may occur *after*
219 * the detach message has been queued so we may find a NULL tp here.
221 if ((tp = intotcpcb(inp)) != NULL) {
223 tp = tcp_disconnect(tp);
224 TCPDEBUG2(PRU_DETACH);
231 * Note: ignore_error is non-zero for certain disconnection races
232 * which we want to silently allow, otherwise close() may return
233 * an unexpected error.
235 #define COMMON_START(so, inp, ignore_error) \
243 return (ignore_error ? 0 : EINVAL); \
245 tp = intotcpcb(inp); \
249 #define COMMON_END(req) out: TCPDEBUG2(req); crit_exit(); return error; goto out
253 * Give the socket an address.
256 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
261 struct sockaddr_in *sinp;
263 COMMON_START(so, inp, 0);
266 * Must check for multicast addresses and disallow binding
269 sinp = (struct sockaddr_in *)nam;
270 if (sinp->sin_family == AF_INET &&
271 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
272 error = EAFNOSUPPORT;
275 error = in_pcbbind(inp, nam, td);
278 COMMON_END(PRU_BIND);
284 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
289 struct sockaddr_in6 *sin6p;
291 COMMON_START(so, inp, 0);
294 * Must check for multicast addresses and disallow binding
297 sin6p = (struct sockaddr_in6 *)nam;
298 if (sin6p->sin6_family == AF_INET6 &&
299 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
300 error = EAFNOSUPPORT;
303 inp->inp_vflag &= ~INP_IPV4;
304 inp->inp_vflag |= INP_IPV6;
305 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
306 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
307 inp->inp_vflag |= INP_IPV4;
308 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
309 struct sockaddr_in sin;
311 in6_sin6_2_sin(&sin, sin6p);
312 inp->inp_vflag |= INP_IPV4;
313 inp->inp_vflag &= ~INP_IPV6;
314 error = in_pcbbind(inp, (struct sockaddr *)&sin, td);
318 error = in6_pcbbind(inp, nam, td);
321 COMMON_END(PRU_BIND);
326 struct netmsg_inswildcard {
327 struct netmsg nm_netmsg;
328 struct inpcb *nm_inp;
329 struct inpcbinfo *nm_pcbinfo;
333 in_pcbinswildcardhash_handler(struct netmsg *msg0)
335 struct netmsg_inswildcard *msg = (struct netmsg_inswildcard *)msg0;
337 in_pcbinswildcardhash_oncpu(msg->nm_inp, msg->nm_pcbinfo);
338 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
343 * Prepare to accept connections.
346 tcp_usr_listen(struct socket *so, struct thread *td)
355 COMMON_START(so, inp, 0);
356 if (inp->inp_lport == 0) {
357 error = in_pcbbind(inp, NULL, td);
362 tp->t_state = TCPS_LISTEN;
365 * We have to set the flag because we can't have other cpus
366 * messing with our inp's flags.
368 inp->inp_flags |= INP_WILDCARD_MP;
369 for (cpu = 0; cpu < ncpus2; cpu++) {
370 struct netmsg_inswildcard *msg;
372 if (cpu == mycpu->gd_cpuid) {
373 in_pcbinswildcardhash(inp);
377 msg = kmalloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG,
379 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0,
380 in_pcbinswildcardhash_handler);
382 msg->nm_pcbinfo = &tcbinfo[cpu];
383 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_netmsg.nm_lmsg);
386 in_pcbinswildcardhash(inp);
388 COMMON_END(PRU_LISTEN);
393 tcp6_usr_listen(struct socket *so, struct thread *td)
402 COMMON_START(so, inp, 0);
403 if (inp->inp_lport == 0) {
404 if (!(inp->inp_flags & IN6P_IPV6_V6ONLY))
405 inp->inp_vflag |= INP_IPV4;
407 inp->inp_vflag &= ~INP_IPV4;
408 error = in6_pcbbind(inp, (struct sockaddr *)0, td);
411 tp->t_state = TCPS_LISTEN;
414 * We have to set the flag because we can't have other cpus
415 * messing with our inp's flags.
417 inp->inp_flags |= INP_WILDCARD_MP;
418 for (cpu = 0; cpu < ncpus2; cpu++) {
419 struct netmsg_inswildcard *msg;
421 if (cpu == mycpu->gd_cpuid) {
422 in_pcbinswildcardhash(inp);
426 msg = kmalloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG,
428 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0,
429 in_pcbinswildcardhash_handler);
431 msg->nm_pcbinfo = &tcbinfo[cpu];
432 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_netmsg.nm_lmsg);
435 in_pcbinswildcardhash(inp);
437 COMMON_END(PRU_LISTEN);
443 tcp_output_dispatch(struct netmsg *nmsg)
445 struct lwkt_msg *msg = &nmsg->nm_lmsg;
446 struct tcpcb *tp = msg->u.ms_resultp;
449 error = tcp_output(tp);
450 lwkt_replymsg(msg, error);
455 * Initiate connection to peer.
456 * Create a template for use in transmissions on this connection.
457 * Enter SYN_SENT state, and mark socket as connecting.
458 * Start keep-alive timer, and seed output sequence space.
459 * Send initial segment on connection.
462 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
467 struct sockaddr_in *sinp;
472 COMMON_START(so, inp, 0);
475 * Must disallow TCP ``connections'' to multicast addresses.
477 sinp = (struct sockaddr_in *)nam;
478 if (sinp->sin_family == AF_INET
479 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
480 error = EAFNOSUPPORT;
484 if (!prison_remote_ip(td, (struct sockaddr*)sinp)) {
485 error = EAFNOSUPPORT; /* IPv6 only jail */
489 if ((error = tcp_connect(tp, nam, td)) != 0)
493 port = tcp_addrport(inp->inp_faddr.s_addr, inp->inp_fport,
494 inp->inp_laddr.s_addr, inp->inp_lport);
495 if (port != &curthread->td_msgport) {
497 struct lwkt_msg *msg;
499 netmsg_init(&nmsg, &curthread->td_msgport, 0,
500 tcp_output_dispatch);
502 msg->u.ms_resultp = tp;
504 error = lwkt_domsg(port, msg, 0);
507 error = tcp_output(tp);
508 COMMON_END(PRU_CONNECT);
513 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
518 struct sockaddr_in6 *sin6p;
520 COMMON_START(so, inp, 0);
523 * Must disallow TCP ``connections'' to multicast addresses.
525 sin6p = (struct sockaddr_in6 *)nam;
526 if (sin6p->sin6_family == AF_INET6
527 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
528 error = EAFNOSUPPORT;
532 if (!prison_remote_ip(td, nam)) {
533 error = EAFNOSUPPORT; /* IPv4 only jail */
537 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
538 struct sockaddr_in sin;
540 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
545 in6_sin6_2_sin(&sin, sin6p);
546 inp->inp_vflag |= INP_IPV4;
547 inp->inp_vflag &= ~INP_IPV6;
548 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
550 error = tcp_output(tp);
553 inp->inp_vflag &= ~INP_IPV4;
554 inp->inp_vflag |= INP_IPV6;
555 inp->inp_inc.inc_isipv6 = 1;
556 if ((error = tcp6_connect(tp, nam, td)) != 0)
558 error = tcp_output(tp);
559 COMMON_END(PRU_CONNECT);
564 * Initiate disconnect from peer.
565 * If connection never passed embryonic stage, just drop;
566 * else if don't need to let data drain, then can just drop anyways,
567 * else have to begin TCP shutdown process: mark socket disconnecting,
568 * drain unread data, state switch to reflect user close, and
569 * send segment (e.g. FIN) to peer. Socket will be really disconnected
570 * when peer sends FIN and acks ours.
572 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
575 tcp_usr_disconnect(struct socket *so)
581 COMMON_START(so, inp, 1);
582 tp = tcp_disconnect(tp);
583 COMMON_END(PRU_DISCONNECT);
587 * Accept a connection. Essentially all the work is
588 * done at higher levels; just return the address
589 * of the peer, storing through addr.
592 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
596 struct tcpcb *tp = NULL;
601 if (so->so_state & SS_ISDISCONNECTED) {
602 error = ECONNABORTED;
611 in_setpeeraddr(so, nam);
612 COMMON_END(PRU_ACCEPT);
617 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
621 struct tcpcb *tp = NULL;
627 if (so->so_state & SS_ISDISCONNECTED) {
628 error = ECONNABORTED;
637 in6_mapped_peeraddr(so, nam);
638 COMMON_END(PRU_ACCEPT);
642 * Mark the connection as being incapable of further output.
645 tcp_usr_shutdown(struct socket *so)
651 COMMON_START(so, inp, 0);
653 tp = tcp_usrclosed(tp);
655 error = tcp_output(tp);
656 COMMON_END(PRU_SHUTDOWN);
660 * After a receive, possibly send window update to peer.
663 tcp_usr_rcvd(struct socket *so, int flags)
669 COMMON_START(so, inp, 0);
671 COMMON_END(PRU_RCVD);
675 * Do a send by putting data in output queue and updating urgent
676 * marker if URG set. Possibly send more data. Unlike the other
677 * pru_*() routines, the mbuf chains are our responsibility. We
678 * must either enqueue them or free them. The other pru_* routines
679 * generally are caller-frees.
682 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
683 struct sockaddr *nam, struct mbuf *control, struct thread *td)
698 * OOPS! we lost a race, the TCP session got reset after
699 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
700 * network interrupt in the non-critical section of sosend().
706 error = ECONNRESET; /* XXX EPIPE? */
712 isipv6 = nam && nam->sa_family == AF_INET6;
717 /* TCP doesn't do control messages (rights, creds, etc) */
718 if (control->m_len) {
725 m_freem(control); /* empty control, just free it */
727 if(!(flags & PRUS_OOB)) {
728 ssb_appendstream(&so->so_snd, m);
729 if (nam && tp->t_state < TCPS_SYN_SENT) {
731 * Do implied connect if not yet connected,
732 * initialize window to default value, and
733 * initialize maxseg/maxopd using peer's cached
738 error = tcp6_connect(tp, nam, td);
741 error = tcp_connect(tp, nam, td);
744 tp->snd_wnd = TTCP_CLIENT_SND_WND;
748 if (flags & PRUS_EOF) {
750 * Close the send side of the connection after
754 tp = tcp_usrclosed(tp);
757 if (flags & PRUS_MORETOCOME)
758 tp->t_flags |= TF_MORETOCOME;
759 error = tcp_output(tp);
760 if (flags & PRUS_MORETOCOME)
761 tp->t_flags &= ~TF_MORETOCOME;
764 if (ssb_space(&so->so_snd) < -512) {
770 * According to RFC961 (Assigned Protocols),
771 * the urgent pointer points to the last octet
772 * of urgent data. We continue, however,
773 * to consider it to indicate the first octet
774 * of data past the urgent section.
775 * Otherwise, snd_up should be one lower.
777 ssb_appendstream(&so->so_snd, m);
778 if (nam && tp->t_state < TCPS_SYN_SENT) {
780 * Do implied connect if not yet connected,
781 * initialize window to default value, and
782 * initialize maxseg/maxopd using peer's cached
787 error = tcp6_connect(tp, nam, td);
790 error = tcp_connect(tp, nam, td);
793 tp->snd_wnd = TTCP_CLIENT_SND_WND;
796 tp->snd_up = tp->snd_una + so->so_snd.ssb_cc;
797 tp->t_flags |= TF_FORCE;
798 error = tcp_output(tp);
799 tp->t_flags &= ~TF_FORCE;
801 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
802 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
809 tcp_usr_abort(struct socket *so)
815 COMMON_START(so, inp, 1);
816 tp = tcp_drop(tp, ECONNABORTED);
817 COMMON_END(PRU_ABORT);
821 * Receive out-of-band data.
824 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
830 COMMON_START(so, inp, 0);
831 if ((so->so_oobmark == 0 &&
832 (so->so_state & SS_RCVATMARK) == 0) ||
833 so->so_options & SO_OOBINLINE ||
834 tp->t_oobflags & TCPOOB_HADDATA) {
838 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
843 *mtod(m, caddr_t) = tp->t_iobc;
844 if ((flags & MSG_PEEK) == 0)
845 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
846 COMMON_END(PRU_RCVOOB);
849 /* xxx - should be const */
850 struct pr_usrreqs tcp_usrreqs = {
851 .pru_abort = tcp_usr_abort,
852 .pru_accept = tcp_usr_accept,
853 .pru_attach = tcp_usr_attach,
854 .pru_bind = tcp_usr_bind,
855 .pru_connect = tcp_usr_connect,
856 .pru_connect2 = pru_connect2_notsupp,
857 .pru_control = in_control,
858 .pru_detach = tcp_usr_detach,
859 .pru_disconnect = tcp_usr_disconnect,
860 .pru_listen = tcp_usr_listen,
861 .pru_peeraddr = in_setpeeraddr,
862 .pru_rcvd = tcp_usr_rcvd,
863 .pru_rcvoob = tcp_usr_rcvoob,
864 .pru_send = tcp_usr_send,
865 .pru_sense = pru_sense_null,
866 .pru_shutdown = tcp_usr_shutdown,
867 .pru_sockaddr = in_setsockaddr,
868 .pru_sosend = sosend,
869 .pru_soreceive = soreceive,
874 struct pr_usrreqs tcp6_usrreqs = {
875 .pru_abort = tcp_usr_abort,
876 .pru_accept = tcp6_usr_accept,
877 .pru_attach = tcp_usr_attach,
878 .pru_bind = tcp6_usr_bind,
879 .pru_connect = tcp6_usr_connect,
880 .pru_connect2 = pru_connect2_notsupp,
881 .pru_control = in6_control,
882 .pru_detach = tcp_usr_detach,
883 .pru_disconnect = tcp_usr_disconnect,
884 .pru_listen = tcp6_usr_listen,
885 .pru_peeraddr = in6_mapped_peeraddr,
886 .pru_rcvd = tcp_usr_rcvd,
887 .pru_rcvoob = tcp_usr_rcvoob,
888 .pru_send = tcp_usr_send,
889 .pru_sense = pru_sense_null,
890 .pru_shutdown = tcp_usr_shutdown,
891 .pru_sockaddr = in6_mapped_sockaddr,
892 .pru_sosend = sosend,
893 .pru_soreceive = soreceive,
899 tcp_connect_oncpu(struct tcpcb *tp, struct sockaddr_in *sin,
900 struct sockaddr_in *if_sin)
902 struct inpcb *inp = tp->t_inpcb, *oinp;
903 struct socket *so = inp->inp_socket;
905 struct rmxp_tao *taop;
906 struct rmxp_tao tao_noncached;
908 oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid],
909 sin->sin_addr, sin->sin_port,
910 inp->inp_laddr.s_addr != INADDR_ANY ?
911 inp->inp_laddr : if_sin->sin_addr,
912 inp->inp_lport, 0, NULL);
914 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
915 otp->t_state == TCPS_TIME_WAIT &&
916 (ticks - otp->t_starttime) < tcp_msl &&
917 (otp->t_flags & TF_RCVD_CC))
922 if (inp->inp_laddr.s_addr == INADDR_ANY)
923 inp->inp_laddr = if_sin->sin_addr;
924 inp->inp_faddr = sin->sin_addr;
925 inp->inp_fport = sin->sin_port;
926 inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid];
927 in_pcbinsconnhash(inp);
929 /* Compute window scaling to request. */
930 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
931 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.ssb_hiwat)
932 tp->request_r_scale++;
935 tcpstat.tcps_connattempt++;
936 tp->t_state = TCPS_SYN_SENT;
937 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
938 tp->iss = tcp_new_isn(tp);
942 * Generate a CC value for this connection and
943 * check whether CC or CCnew should be used.
945 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
946 taop = &tao_noncached;
947 bzero(taop, sizeof *taop);
950 tp->cc_send = CC_INC(tcp_ccgen);
951 if (taop->tao_ccsent != 0 &&
952 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
953 taop->tao_ccsent = tp->cc_send;
955 taop->tao_ccsent = 0;
956 tp->t_flags |= TF_SENDCCNEW;
964 struct netmsg_tcp_connect {
965 struct netmsg nm_netmsg;
967 struct sockaddr_in *nm_sin;
968 struct sockaddr_in *nm_ifsin;
972 tcp_connect_handler(netmsg_t netmsg)
974 struct netmsg_tcp_connect *msg = (void *)netmsg;
977 error = tcp_connect_oncpu(msg->nm_tp, msg->nm_sin, msg->nm_ifsin);
978 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, error);
984 * Common subroutine to open a TCP connection to remote host specified
985 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
986 * port number if needed. Call in_pcbladdr to do the routing and to choose
987 * a local host address (interface). If there is an existing incarnation
988 * of the same connection in TIME-WAIT state and if the remote host was
989 * sending CC options and if the connection duration was < MSL, then
990 * truncate the previous TIME-WAIT state and proceed.
991 * Initialize connection parameters and enter SYN-SENT state.
994 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
996 struct inpcb *inp = tp->t_inpcb;
997 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
998 struct sockaddr_in *if_sin;
1004 if (inp->inp_lport == 0) {
1005 error = in_pcbbind(inp, (struct sockaddr *)NULL, td);
1011 * Cannot simply call in_pcbconnect, because there might be an
1012 * earlier incarnation of this same connection still in
1013 * TIME_WAIT state, creating an ADDRINUSE error.
1015 error = in_pcbladdr(inp, nam, &if_sin, td);
1020 port = tcp_addrport(sin->sin_addr.s_addr, sin->sin_port,
1021 inp->inp_laddr.s_addr ?
1022 inp->inp_laddr.s_addr : if_sin->sin_addr.s_addr,
1025 if (port != &curthread->td_msgport) {
1026 struct netmsg_tcp_connect msg;
1027 struct route *ro = &inp->inp_route;
1030 * in_pcbladdr() may have allocated a route entry for us
1031 * on the current CPU, but we need a route entry on the
1032 * target CPU, so free it here.
1034 if (ro->ro_rt != NULL)
1036 bzero(ro, sizeof(*ro));
1038 netmsg_init(&msg.nm_netmsg, &curthread->td_msgport, 0,
1039 tcp_connect_handler);
1042 msg.nm_ifsin = if_sin;
1043 error = lwkt_domsg(port, &msg.nm_netmsg.nm_lmsg, 0);
1046 error = tcp_connect_oncpu(tp, sin, if_sin);
1053 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1055 struct inpcb *inp = tp->t_inpcb, *oinp;
1056 struct socket *so = inp->inp_socket;
1058 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
1059 struct in6_addr *addr6;
1060 struct rmxp_tao *taop;
1061 struct rmxp_tao tao_noncached;
1064 if (inp->inp_lport == 0) {
1065 error = in6_pcbbind(inp, (struct sockaddr *)0, td);
1071 * Cannot simply call in_pcbconnect, because there might be an
1072 * earlier incarnation of this same connection still in
1073 * TIME_WAIT state, creating an ADDRINUSE error.
1075 error = in6_pcbladdr(inp, nam, &addr6, td);
1078 oinp = in6_pcblookup_hash(inp->inp_cpcbinfo,
1079 &sin6->sin6_addr, sin6->sin6_port,
1080 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
1081 addr6 : &inp->in6p_laddr,
1082 inp->inp_lport, 0, NULL);
1084 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
1085 otp->t_state == TCPS_TIME_WAIT &&
1086 (ticks - otp->t_starttime) < tcp_msl &&
1087 (otp->t_flags & TF_RCVD_CC))
1088 otp = tcp_close(otp);
1090 return (EADDRINUSE);
1092 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1093 inp->in6p_laddr = *addr6;
1094 inp->in6p_faddr = sin6->sin6_addr;
1095 inp->inp_fport = sin6->sin6_port;
1096 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != 0)
1097 inp->in6p_flowinfo = sin6->sin6_flowinfo;
1098 in_pcbinsconnhash(inp);
1100 /* Compute window scaling to request. */
1101 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1102 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.ssb_hiwat)
1103 tp->request_r_scale++;
1106 tcpstat.tcps_connattempt++;
1107 tp->t_state = TCPS_SYN_SENT;
1108 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
1109 tp->iss = tcp_new_isn(tp);
1110 tcp_sendseqinit(tp);
1113 * Generate a CC value for this connection and
1114 * check whether CC or CCnew should be used.
1116 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
1117 taop = &tao_noncached;
1118 bzero(taop, sizeof *taop);
1121 tp->cc_send = CC_INC(tcp_ccgen);
1122 if (taop->tao_ccsent != 0 &&
1123 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
1124 taop->tao_ccsent = tp->cc_send;
1126 taop->tao_ccsent = 0;
1127 tp->t_flags |= TF_SENDCCNEW;
1135 * The new sockopt interface makes it possible for us to block in the
1136 * copyin/out step (if we take a page fault). Taking a page fault while
1137 * in a critical section is probably a Bad Thing. (Since sockets and pcbs
1138 * both now use TSM, there probably isn't any need for this function to
1139 * run in a critical section any more. This needs more examination.)
1142 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1144 int error, opt, optval;
1149 crit_enter(); /* XXX */
1153 return (ECONNRESET);
1155 if (sopt->sopt_level != IPPROTO_TCP) {
1157 if (INP_CHECK_SOCKAF(so, AF_INET6))
1158 error = ip6_ctloutput(so, sopt);
1161 error = ip_ctloutput(so, sopt);
1165 tp = intotcpcb(inp);
1167 switch (sopt->sopt_dir) {
1169 error = soopt_to_kbuf(sopt, &optval, sizeof optval,
1173 switch (sopt->sopt_name) {
1176 switch (sopt->sopt_name) {
1184 opt = 0; /* dead code to fool gcc */
1191 tp->t_flags &= ~opt;
1196 tp->t_flags |= TF_NOPUSH;
1198 tp->t_flags &= ~TF_NOPUSH;
1199 error = tcp_output(tp);
1204 if (optval > 0 && optval <= tp->t_maxseg)
1205 tp->t_maxseg = optval;
1211 error = ENOPROTOOPT;
1217 switch (sopt->sopt_name) {
1219 optval = tp->t_flags & TF_NODELAY;
1222 optval = tp->t_maxseg;
1225 optval = tp->t_flags & TF_NOOPT;
1228 optval = tp->t_flags & TF_NOPUSH;
1231 error = ENOPROTOOPT;
1235 soopt_from_kbuf(sopt, &optval, sizeof optval);
1243 * tcp_sendspace and tcp_recvspace are the default send and receive window
1244 * sizes, respectively. These are obsolescent (this information should
1245 * be set by the route).
1247 * Use a default that does not require tcp window scaling to be turned
1248 * on. Individual programs or the administrator can increase the default.
1250 u_long tcp_sendspace = 57344; /* largest multiple of PAGE_SIZE < 64k */
1251 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1252 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1253 u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */
1254 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1255 &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1258 * Attach TCP protocol to socket, allocating
1259 * internet protocol control block, tcp control block,
1260 * bufer space, and entering LISTEN state if to accept connections.
1263 tcp_attach(struct socket *so, struct pru_attach_info *ai)
1270 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0;
1273 if (so->so_snd.ssb_hiwat == 0 || so->so_rcv.ssb_hiwat == 0) {
1274 error = soreserve(so, tcp_sendspace, tcp_recvspace,
1279 cpu = mycpu->gd_cpuid;
1280 error = in_pcballoc(so, &tcbinfo[cpu]);
1286 inp->inp_vflag |= INP_IPV6;
1287 inp->in6p_hops = -1; /* use kernel default */
1291 inp->inp_vflag |= INP_IPV4;
1292 tp = tcp_newtcpcb(inp);
1294 int nofd = so->so_state & SS_NOFDREF; /* XXX */
1296 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */
1303 so->so_state |= nofd;
1306 tp->t_state = TCPS_CLOSED;
1311 * Initiate (or continue) disconnect.
1312 * If embryonic state, just send reset (once).
1313 * If in ``let data drain'' option and linger null, just drop.
1314 * Otherwise (hard), mark socket disconnecting and drop
1315 * current input data; switch states based on user close, and
1316 * send segment to peer (with FIN).
1318 static struct tcpcb *
1319 tcp_disconnect(struct tcpcb *tp)
1321 struct socket *so = tp->t_inpcb->inp_socket;
1323 if (tp->t_state < TCPS_ESTABLISHED)
1325 else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1326 tp = tcp_drop(tp, 0);
1328 soisdisconnecting(so);
1329 sbflush(&so->so_rcv.sb);
1330 tp = tcp_usrclosed(tp);
1338 * User issued close, and wish to trail through shutdown states:
1339 * if never received SYN, just forget it. If got a SYN from peer,
1340 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1341 * If already got a FIN from peer, then almost done; go to LAST_ACK
1342 * state. In all other cases, have already sent FIN to peer (e.g.
1343 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1344 * for peer to send FIN or not respond to keep-alives, etc.
1345 * We can let the user exit from the close as soon as the FIN is acked.
1347 static struct tcpcb *
1348 tcp_usrclosed(struct tcpcb *tp)
1351 switch (tp->t_state) {
1355 tp->t_state = TCPS_CLOSED;
1360 case TCPS_SYN_RECEIVED:
1361 tp->t_flags |= TF_NEEDFIN;
1364 case TCPS_ESTABLISHED:
1365 tp->t_state = TCPS_FIN_WAIT_1;
1368 case TCPS_CLOSE_WAIT:
1369 tp->t_state = TCPS_LAST_ACK;
1372 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1373 soisdisconnected(tp->t_inpcb->inp_socket);
1374 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
1375 if (tp->t_state == TCPS_FIN_WAIT_2)
1376 callout_reset(tp->tt_2msl, tcp_maxidle,
1377 tcp_timer_2msl, tp);