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 1.46 2008/02/09 13:38:03 sephe 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 return EINVAL; /* XXX */
212 * It's possible for the tcpcb (tp) to disconnect from the inp due
213 * to tcp_drop()->tcp_close() being called. This may occur *after*
214 * the detach message has been queued so we may find a NULL tp here.
216 if ((tp = intotcpcb(inp)) != NULL) {
218 tp = tcp_disconnect(tp);
219 TCPDEBUG2(PRU_DETACH);
225 #define COMMON_START(so, inp) \
235 tp = intotcpcb(inp); \
239 #define COMMON_END(req) out: TCPDEBUG2(req); crit_exit(); return error; goto out
243 * Give the socket an address.
246 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
251 struct sockaddr_in *sinp;
253 COMMON_START(so, inp);
256 * Must check for multicast addresses and disallow binding
259 sinp = (struct sockaddr_in *)nam;
260 if (sinp->sin_family == AF_INET &&
261 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
262 error = EAFNOSUPPORT;
265 error = in_pcbbind(inp, nam, td);
268 COMMON_END(PRU_BIND);
274 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
279 struct sockaddr_in6 *sin6p;
281 COMMON_START(so, inp);
284 * Must check for multicast addresses and disallow binding
287 sin6p = (struct sockaddr_in6 *)nam;
288 if (sin6p->sin6_family == AF_INET6 &&
289 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
290 error = EAFNOSUPPORT;
293 inp->inp_vflag &= ~INP_IPV4;
294 inp->inp_vflag |= INP_IPV6;
295 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
296 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
297 inp->inp_vflag |= INP_IPV4;
298 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
299 struct sockaddr_in sin;
301 in6_sin6_2_sin(&sin, sin6p);
302 inp->inp_vflag |= INP_IPV4;
303 inp->inp_vflag &= ~INP_IPV6;
304 error = in_pcbbind(inp, (struct sockaddr *)&sin, td);
308 error = in6_pcbbind(inp, nam, td);
311 COMMON_END(PRU_BIND);
316 struct netmsg_inswildcard {
317 struct netmsg nm_netmsg;
318 struct inpcb *nm_inp;
319 struct inpcbinfo *nm_pcbinfo;
323 in_pcbinswildcardhash_handler(struct netmsg *msg0)
325 struct netmsg_inswildcard *msg = (struct netmsg_inswildcard *)msg0;
327 in_pcbinswildcardhash_oncpu(msg->nm_inp, msg->nm_pcbinfo);
328 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
333 * Prepare to accept connections.
336 tcp_usr_listen(struct socket *so, struct thread *td)
345 COMMON_START(so, inp);
346 if (inp->inp_lport == 0) {
347 error = in_pcbbind(inp, NULL, td);
352 tp->t_state = TCPS_LISTEN;
355 * We have to set the flag because we can't have other cpus
356 * messing with our inp's flags.
358 inp->inp_flags |= INP_WILDCARD_MP;
359 for (cpu = 0; cpu < ncpus2; cpu++) {
360 struct netmsg_inswildcard *msg;
362 if (cpu == mycpu->gd_cpuid) {
363 in_pcbinswildcardhash(inp);
367 msg = kmalloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG,
369 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0,
370 in_pcbinswildcardhash_handler);
372 msg->nm_pcbinfo = &tcbinfo[cpu];
373 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_netmsg.nm_lmsg);
376 in_pcbinswildcardhash(inp);
378 COMMON_END(PRU_LISTEN);
383 tcp6_usr_listen(struct socket *so, struct thread *td)
392 COMMON_START(so, inp);
393 if (inp->inp_lport == 0) {
394 if (!(inp->inp_flags & IN6P_IPV6_V6ONLY))
395 inp->inp_vflag |= INP_IPV4;
397 inp->inp_vflag &= ~INP_IPV4;
398 error = in6_pcbbind(inp, (struct sockaddr *)0, td);
401 tp->t_state = TCPS_LISTEN;
404 * We have to set the flag because we can't have other cpus
405 * messing with our inp's flags.
407 inp->inp_flags |= INP_WILDCARD_MP;
408 for (cpu = 0; cpu < ncpus2; cpu++) {
409 struct netmsg_inswildcard *msg;
411 if (cpu == mycpu->gd_cpuid) {
412 in_pcbinswildcardhash(inp);
416 msg = kmalloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG,
418 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0,
419 in_pcbinswildcardhash_handler);
421 msg->nm_pcbinfo = &tcbinfo[cpu];
422 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_netmsg.nm_lmsg);
425 in_pcbinswildcardhash(inp);
427 COMMON_END(PRU_LISTEN);
432 tcp_output_dispatch(struct netmsg *nmsg)
434 struct lwkt_msg *msg = &nmsg->nm_lmsg;
435 struct tcpcb *tp = msg->u.ms_resultp;
438 error = tcp_output(tp);
439 lwkt_replymsg(msg, error);
443 * Initiate connection to peer.
444 * Create a template for use in transmissions on this connection.
445 * Enter SYN_SENT state, and mark socket as connecting.
446 * Start keep-alive timer, and seed output sequence space.
447 * Send initial segment on connection.
450 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
455 struct sockaddr_in *sinp;
460 COMMON_START(so, inp);
463 * Must disallow TCP ``connections'' to multicast addresses.
465 sinp = (struct sockaddr_in *)nam;
466 if (sinp->sin_family == AF_INET
467 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
468 error = EAFNOSUPPORT;
472 if (!prison_remote_ip(td, (struct sockaddr*)sinp)) {
473 error = EAFNOSUPPORT; /* IPv6 only jail */
477 if ((error = tcp_connect(tp, nam, td)) != 0)
481 port = tcp_addrport(inp->inp_faddr.s_addr, inp->inp_fport,
482 inp->inp_laddr.s_addr, inp->inp_lport);
483 if (port != &curthread->td_msgport) {
485 struct lwkt_msg *msg;
487 netmsg_init(&nmsg, &curthread->td_msgport, 0,
488 tcp_output_dispatch);
490 msg->u.ms_resultp = tp;
492 error = lwkt_domsg(port, msg, 0);
495 error = tcp_output(tp);
496 COMMON_END(PRU_CONNECT);
501 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
506 struct sockaddr_in6 *sin6p;
508 COMMON_START(so, inp);
511 * Must disallow TCP ``connections'' to multicast addresses.
513 sin6p = (struct sockaddr_in6 *)nam;
514 if (sin6p->sin6_family == AF_INET6
515 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
516 error = EAFNOSUPPORT;
520 if (!prison_remote_ip(td, nam)) {
521 error = EAFNOSUPPORT; /* IPv4 only jail */
525 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
526 struct sockaddr_in sin;
528 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
533 in6_sin6_2_sin(&sin, sin6p);
534 inp->inp_vflag |= INP_IPV4;
535 inp->inp_vflag &= ~INP_IPV6;
536 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
538 error = tcp_output(tp);
541 inp->inp_vflag &= ~INP_IPV4;
542 inp->inp_vflag |= INP_IPV6;
543 inp->inp_inc.inc_isipv6 = 1;
544 if ((error = tcp6_connect(tp, nam, td)) != 0)
546 error = tcp_output(tp);
547 COMMON_END(PRU_CONNECT);
552 * Initiate disconnect from peer.
553 * If connection never passed embryonic stage, just drop;
554 * else if don't need to let data drain, then can just drop anyways,
555 * else have to begin TCP shutdown process: mark socket disconnecting,
556 * drain unread data, state switch to reflect user close, and
557 * send segment (e.g. FIN) to peer. Socket will be really disconnected
558 * when peer sends FIN and acks ours.
560 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
563 tcp_usr_disconnect(struct socket *so)
569 COMMON_START(so, inp);
570 tp = tcp_disconnect(tp);
571 COMMON_END(PRU_DISCONNECT);
575 * Accept a connection. Essentially all the work is
576 * done at higher levels; just return the address
577 * of the peer, storing through addr.
580 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
584 struct tcpcb *tp = NULL;
589 if (so->so_state & SS_ISDISCONNECTED) {
590 error = ECONNABORTED;
599 in_setpeeraddr(so, nam);
600 COMMON_END(PRU_ACCEPT);
605 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
609 struct tcpcb *tp = NULL;
615 if (so->so_state & SS_ISDISCONNECTED) {
616 error = ECONNABORTED;
625 in6_mapped_peeraddr(so, nam);
626 COMMON_END(PRU_ACCEPT);
630 * Mark the connection as being incapable of further output.
633 tcp_usr_shutdown(struct socket *so)
639 COMMON_START(so, inp);
641 tp = tcp_usrclosed(tp);
643 error = tcp_output(tp);
644 COMMON_END(PRU_SHUTDOWN);
648 * After a receive, possibly send window update to peer.
651 tcp_usr_rcvd(struct socket *so, int flags)
657 COMMON_START(so, inp);
659 COMMON_END(PRU_RCVD);
663 * Do a send by putting data in output queue and updating urgent
664 * marker if URG set. Possibly send more data. Unlike the other
665 * pru_*() routines, the mbuf chains are our responsibility. We
666 * must either enqueue them or free them. The other pru_* routines
667 * generally are caller-frees.
670 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
671 struct sockaddr *nam, struct mbuf *control, struct thread *td)
686 * OOPS! we lost a race, the TCP session got reset after
687 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
688 * network interrupt in the non-critical section of sosend().
694 error = ECONNRESET; /* XXX EPIPE? */
700 isipv6 = nam && nam->sa_family == AF_INET6;
705 /* TCP doesn't do control messages (rights, creds, etc) */
706 if (control->m_len) {
713 m_freem(control); /* empty control, just free it */
715 if(!(flags & PRUS_OOB)) {
716 ssb_appendstream(&so->so_snd, m);
717 if (nam && tp->t_state < TCPS_SYN_SENT) {
719 * Do implied connect if not yet connected,
720 * initialize window to default value, and
721 * initialize maxseg/maxopd using peer's cached
726 error = tcp6_connect(tp, nam, td);
729 error = tcp_connect(tp, nam, td);
732 tp->snd_wnd = TTCP_CLIENT_SND_WND;
736 if (flags & PRUS_EOF) {
738 * Close the send side of the connection after
742 tp = tcp_usrclosed(tp);
745 if (flags & PRUS_MORETOCOME)
746 tp->t_flags |= TF_MORETOCOME;
747 error = tcp_output(tp);
748 if (flags & PRUS_MORETOCOME)
749 tp->t_flags &= ~TF_MORETOCOME;
752 if (ssb_space(&so->so_snd) < -512) {
758 * According to RFC961 (Assigned Protocols),
759 * the urgent pointer points to the last octet
760 * of urgent data. We continue, however,
761 * to consider it to indicate the first octet
762 * of data past the urgent section.
763 * Otherwise, snd_up should be one lower.
765 ssb_appendstream(&so->so_snd, m);
766 if (nam && tp->t_state < TCPS_SYN_SENT) {
768 * Do implied connect if not yet connected,
769 * initialize window to default value, and
770 * initialize maxseg/maxopd using peer's cached
775 error = tcp6_connect(tp, nam, td);
778 error = tcp_connect(tp, nam, td);
781 tp->snd_wnd = TTCP_CLIENT_SND_WND;
784 tp->snd_up = tp->snd_una + so->so_snd.ssb_cc;
785 tp->t_flags |= TF_FORCE;
786 error = tcp_output(tp);
787 tp->t_flags &= ~TF_FORCE;
789 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
790 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
797 tcp_usr_abort(struct socket *so)
803 COMMON_START(so, inp);
804 tp = tcp_drop(tp, ECONNABORTED);
805 COMMON_END(PRU_ABORT);
809 * Receive out-of-band data.
812 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
818 COMMON_START(so, inp);
819 if ((so->so_oobmark == 0 &&
820 (so->so_state & SS_RCVATMARK) == 0) ||
821 so->so_options & SO_OOBINLINE ||
822 tp->t_oobflags & TCPOOB_HADDATA) {
826 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
831 *mtod(m, caddr_t) = tp->t_iobc;
832 if ((flags & MSG_PEEK) == 0)
833 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
834 COMMON_END(PRU_RCVOOB);
837 /* xxx - should be const */
838 struct pr_usrreqs tcp_usrreqs = {
839 .pru_abort = tcp_usr_abort,
840 .pru_accept = tcp_usr_accept,
841 .pru_attach = tcp_usr_attach,
842 .pru_bind = tcp_usr_bind,
843 .pru_connect = tcp_usr_connect,
844 .pru_connect2 = pru_connect2_notsupp,
845 .pru_control = in_control,
846 .pru_detach = tcp_usr_detach,
847 .pru_disconnect = tcp_usr_disconnect,
848 .pru_listen = tcp_usr_listen,
849 .pru_peeraddr = in_setpeeraddr,
850 .pru_rcvd = tcp_usr_rcvd,
851 .pru_rcvoob = tcp_usr_rcvoob,
852 .pru_send = tcp_usr_send,
853 .pru_sense = pru_sense_null,
854 .pru_shutdown = tcp_usr_shutdown,
855 .pru_sockaddr = in_setsockaddr,
856 .pru_sosend = sosend,
857 .pru_soreceive = soreceive,
862 struct pr_usrreqs tcp6_usrreqs = {
863 .pru_abort = tcp_usr_abort,
864 .pru_accept = tcp6_usr_accept,
865 .pru_attach = tcp_usr_attach,
866 .pru_bind = tcp6_usr_bind,
867 .pru_connect = tcp6_usr_connect,
868 .pru_connect2 = pru_connect2_notsupp,
869 .pru_control = in6_control,
870 .pru_detach = tcp_usr_detach,
871 .pru_disconnect = tcp_usr_disconnect,
872 .pru_listen = tcp6_usr_listen,
873 .pru_peeraddr = in6_mapped_peeraddr,
874 .pru_rcvd = tcp_usr_rcvd,
875 .pru_rcvoob = tcp_usr_rcvoob,
876 .pru_send = tcp_usr_send,
877 .pru_sense = pru_sense_null,
878 .pru_shutdown = tcp_usr_shutdown,
879 .pru_sockaddr = in6_mapped_sockaddr,
880 .pru_sosend = sosend,
881 .pru_soreceive = soreceive,
887 tcp_connect_oncpu(struct tcpcb *tp, struct sockaddr_in *sin,
888 struct sockaddr_in *if_sin)
890 struct inpcb *inp = tp->t_inpcb, *oinp;
891 struct socket *so = inp->inp_socket;
893 struct rmxp_tao *taop;
894 struct rmxp_tao tao_noncached;
896 oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid],
897 sin->sin_addr, sin->sin_port,
898 inp->inp_laddr.s_addr != INADDR_ANY ?
899 inp->inp_laddr : if_sin->sin_addr,
900 inp->inp_lport, 0, NULL);
902 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
903 otp->t_state == TCPS_TIME_WAIT &&
904 (ticks - otp->t_starttime) < tcp_msl &&
905 (otp->t_flags & TF_RCVD_CC))
910 if (inp->inp_laddr.s_addr == INADDR_ANY)
911 inp->inp_laddr = if_sin->sin_addr;
912 inp->inp_faddr = sin->sin_addr;
913 inp->inp_fport = sin->sin_port;
914 inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid];
915 in_pcbinsconnhash(inp);
917 /* Compute window scaling to request. */
918 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
919 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.ssb_hiwat)
920 tp->request_r_scale++;
923 tcpstat.tcps_connattempt++;
924 tp->t_state = TCPS_SYN_SENT;
925 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
926 tp->iss = tcp_new_isn(tp);
930 * Generate a CC value for this connection and
931 * check whether CC or CCnew should be used.
933 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
934 taop = &tao_noncached;
935 bzero(taop, sizeof *taop);
938 tp->cc_send = CC_INC(tcp_ccgen);
939 if (taop->tao_ccsent != 0 &&
940 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
941 taop->tao_ccsent = tp->cc_send;
943 taop->tao_ccsent = 0;
944 tp->t_flags |= TF_SENDCCNEW;
952 struct netmsg_tcp_connect {
953 struct netmsg nm_netmsg;
955 struct sockaddr_in *nm_sin;
956 struct sockaddr_in *nm_ifsin;
960 tcp_connect_handler(netmsg_t netmsg)
962 struct netmsg_tcp_connect *msg = (void *)netmsg;
965 error = tcp_connect_oncpu(msg->nm_tp, msg->nm_sin, msg->nm_ifsin);
966 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, error);
972 * Common subroutine to open a TCP connection to remote host specified
973 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
974 * port number if needed. Call in_pcbladdr to do the routing and to choose
975 * a local host address (interface). If there is an existing incarnation
976 * of the same connection in TIME-WAIT state and if the remote host was
977 * sending CC options and if the connection duration was < MSL, then
978 * truncate the previous TIME-WAIT state and proceed.
979 * Initialize connection parameters and enter SYN-SENT state.
982 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
984 struct inpcb *inp = tp->t_inpcb;
985 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
986 struct sockaddr_in *if_sin;
992 if (inp->inp_lport == 0) {
993 error = in_pcbbind(inp, (struct sockaddr *)NULL, td);
999 * Cannot simply call in_pcbconnect, because there might be an
1000 * earlier incarnation of this same connection still in
1001 * TIME_WAIT state, creating an ADDRINUSE error.
1003 error = in_pcbladdr(inp, nam, &if_sin, td);
1008 port = tcp_addrport(sin->sin_addr.s_addr, sin->sin_port,
1009 inp->inp_laddr.s_addr ?
1010 inp->inp_laddr.s_addr : if_sin->sin_addr.s_addr,
1013 if (port != &curthread->td_msgport) {
1014 struct netmsg_tcp_connect msg;
1015 struct route *ro = &inp->inp_route;
1018 * in_pcbladdr() may have allocated a route entry for us
1019 * on the current CPU, but we need a route entry on the
1020 * target CPU, so free it here.
1022 if (ro->ro_rt != NULL)
1024 bzero(ro, sizeof(*ro));
1026 netmsg_init(&msg.nm_netmsg, &curthread->td_msgport, 0,
1027 tcp_connect_handler);
1030 msg.nm_ifsin = if_sin;
1031 error = lwkt_domsg(port, &msg.nm_netmsg.nm_lmsg, 0);
1034 error = tcp_connect_oncpu(tp, sin, if_sin);
1041 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1043 struct inpcb *inp = tp->t_inpcb, *oinp;
1044 struct socket *so = inp->inp_socket;
1046 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
1047 struct in6_addr *addr6;
1048 struct rmxp_tao *taop;
1049 struct rmxp_tao tao_noncached;
1052 if (inp->inp_lport == 0) {
1053 error = in6_pcbbind(inp, (struct sockaddr *)0, td);
1059 * Cannot simply call in_pcbconnect, because there might be an
1060 * earlier incarnation of this same connection still in
1061 * TIME_WAIT state, creating an ADDRINUSE error.
1063 error = in6_pcbladdr(inp, nam, &addr6, td);
1066 oinp = in6_pcblookup_hash(inp->inp_cpcbinfo,
1067 &sin6->sin6_addr, sin6->sin6_port,
1068 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
1069 addr6 : &inp->in6p_laddr,
1070 inp->inp_lport, 0, NULL);
1072 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
1073 otp->t_state == TCPS_TIME_WAIT &&
1074 (ticks - otp->t_starttime) < tcp_msl &&
1075 (otp->t_flags & TF_RCVD_CC))
1076 otp = tcp_close(otp);
1078 return (EADDRINUSE);
1080 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1081 inp->in6p_laddr = *addr6;
1082 inp->in6p_faddr = sin6->sin6_addr;
1083 inp->inp_fport = sin6->sin6_port;
1084 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL)
1085 inp->in6p_flowinfo = sin6->sin6_flowinfo;
1086 in_pcbinsconnhash(inp);
1088 /* Compute window scaling to request. */
1089 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1090 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.ssb_hiwat)
1091 tp->request_r_scale++;
1094 tcpstat.tcps_connattempt++;
1095 tp->t_state = TCPS_SYN_SENT;
1096 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
1097 tp->iss = tcp_new_isn(tp);
1098 tcp_sendseqinit(tp);
1101 * Generate a CC value for this connection and
1102 * check whether CC or CCnew should be used.
1104 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
1105 taop = &tao_noncached;
1106 bzero(taop, sizeof *taop);
1109 tp->cc_send = CC_INC(tcp_ccgen);
1110 if (taop->tao_ccsent != 0 &&
1111 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
1112 taop->tao_ccsent = tp->cc_send;
1114 taop->tao_ccsent = 0;
1115 tp->t_flags |= TF_SENDCCNEW;
1123 * The new sockopt interface makes it possible for us to block in the
1124 * copyin/out step (if we take a page fault). Taking a page fault while
1125 * in a critical section is probably a Bad Thing. (Since sockets and pcbs
1126 * both now use TSM, there probably isn't any need for this function to
1127 * run in a critical section any more. This needs more examination.)
1130 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1132 int error, opt, optval;
1137 crit_enter(); /* XXX */
1141 return (ECONNRESET);
1143 if (sopt->sopt_level != IPPROTO_TCP) {
1145 if (INP_CHECK_SOCKAF(so, AF_INET6))
1146 error = ip6_ctloutput(so, sopt);
1149 error = ip_ctloutput(so, sopt);
1153 tp = intotcpcb(inp);
1155 switch (sopt->sopt_dir) {
1157 switch (sopt->sopt_name) {
1160 error = sooptcopyin(sopt, &optval, sizeof optval,
1165 switch (sopt->sopt_name) {
1173 opt = 0; /* dead code to fool gcc */
1180 tp->t_flags &= ~opt;
1184 error = sooptcopyin(sopt, &optval, sizeof optval,
1190 tp->t_flags |= TF_NOPUSH;
1192 tp->t_flags &= ~TF_NOPUSH;
1193 error = tcp_output(tp);
1198 error = sooptcopyin(sopt, &optval, sizeof optval,
1203 if (optval > 0 && optval <= tp->t_maxseg)
1204 tp->t_maxseg = optval;
1210 error = ENOPROTOOPT;
1216 switch (sopt->sopt_name) {
1218 optval = tp->t_flags & TF_NODELAY;
1221 optval = tp->t_maxseg;
1224 optval = tp->t_flags & TF_NOOPT;
1227 optval = tp->t_flags & TF_NOPUSH;
1230 error = ENOPROTOOPT;
1234 error = sooptcopyout(sopt, &optval, sizeof optval);
1242 * tcp_sendspace and tcp_recvspace are the default send and receive window
1243 * sizes, respectively. These are obsolescent (this information should
1244 * be set by the route).
1246 u_long tcp_sendspace = 1024*32;
1247 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1248 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1249 u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */
1250 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1251 &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1254 * Attach TCP protocol to socket, allocating
1255 * internet protocol control block, tcp control block,
1256 * bufer space, and entering LISTEN state if to accept connections.
1259 tcp_attach(struct socket *so, struct pru_attach_info *ai)
1266 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL;
1269 if (so->so_snd.ssb_hiwat == 0 || so->so_rcv.ssb_hiwat == 0) {
1270 error = soreserve(so, tcp_sendspace, tcp_recvspace,
1275 cpu = mycpu->gd_cpuid;
1276 error = in_pcballoc(so, &tcbinfo[cpu]);
1282 inp->inp_vflag |= INP_IPV6;
1283 inp->in6p_hops = -1; /* use kernel default */
1287 inp->inp_vflag |= INP_IPV4;
1288 tp = tcp_newtcpcb(inp);
1290 int nofd = so->so_state & SS_NOFDREF; /* XXX */
1292 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */
1299 so->so_state |= nofd;
1302 tp->t_state = TCPS_CLOSED;
1307 * Initiate (or continue) disconnect.
1308 * If embryonic state, just send reset (once).
1309 * If in ``let data drain'' option and linger null, just drop.
1310 * Otherwise (hard), mark socket disconnecting and drop
1311 * current input data; switch states based on user close, and
1312 * send segment to peer (with FIN).
1314 static struct tcpcb *
1315 tcp_disconnect(struct tcpcb *tp)
1317 struct socket *so = tp->t_inpcb->inp_socket;
1319 if (tp->t_state < TCPS_ESTABLISHED)
1321 else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1322 tp = tcp_drop(tp, 0);
1324 soisdisconnecting(so);
1325 sbflush(&so->so_rcv.sb);
1326 tp = tcp_usrclosed(tp);
1334 * User issued close, and wish to trail through shutdown states:
1335 * if never received SYN, just forget it. If got a SYN from peer,
1336 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1337 * If already got a FIN from peer, then almost done; go to LAST_ACK
1338 * state. In all other cases, have already sent FIN to peer (e.g.
1339 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1340 * for peer to send FIN or not respond to keep-alives, etc.
1341 * We can let the user exit from the close as soon as the FIN is acked.
1343 static struct tcpcb *
1344 tcp_usrclosed(struct tcpcb *tp)
1347 switch (tp->t_state) {
1351 tp->t_state = TCPS_CLOSED;
1356 case TCPS_SYN_RECEIVED:
1357 tp->t_flags |= TF_NEEDFIN;
1360 case TCPS_ESTABLISHED:
1361 tp->t_state = TCPS_FIN_WAIT_1;
1364 case TCPS_CLOSE_WAIT:
1365 tp->t_state = TCPS_LAST_ACK;
1368 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1369 soisdisconnected(tp->t_inpcb->inp_socket);
1370 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
1371 if (tp->t_state == TCPS_FIN_WAIT_2)
1372 callout_reset(tp->tt_2msl, tcp_maxidle,
1373 tcp_timer_2msl, tp);