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) 2003, 2004 Jeffrey M. Hsu. All rights reserved.
37 * License terms: all terms for the DragonFly license above plus the following:
39 * 4. All advertising materials mentioning features or use of this software
40 * must display the following acknowledgement:
42 * This product includes software developed by Jeffrey M. Hsu
43 * for the DragonFly Project.
45 * This requirement may be waived with permission from Jeffrey Hsu.
46 * This requirement will sunset and may be removed on July 8 2005,
47 * after which the standard DragonFly license (as shown above) will
52 * Copyright (c) 1982, 1986, 1988, 1993
53 * The Regents of the University of California. All rights reserved.
55 * Redistribution and use in source and binary forms, with or without
56 * modification, are permitted provided that the following conditions
58 * 1. Redistributions of source code must retain the above copyright
59 * notice, this list of conditions and the following disclaimer.
60 * 2. Redistributions in binary form must reproduce the above copyright
61 * notice, this list of conditions and the following disclaimer in the
62 * documentation and/or other materials provided with the distribution.
63 * 3. All advertising materials mentioning features or use of this software
64 * must display the following acknowledgement:
65 * This product includes software developed by the University of
66 * California, Berkeley and its contributors.
67 * 4. Neither the name of the University nor the names of its contributors
68 * may be used to endorse or promote products derived from this software
69 * without specific prior written permission.
71 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
72 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
73 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
74 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
75 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
76 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
77 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
78 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
79 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
80 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
83 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
84 * $FreeBSD: src/sys/netinet/tcp_usrreq.c,v 1.51.2.17 2002/10/11 11:46:44 ume Exp $
85 * $DragonFly: src/sys/netinet/tcp_usrreq.c,v 1.27 2004/10/27 03:43:47 dillon Exp $
88 #include "opt_ipsec.h"
89 #include "opt_inet6.h"
90 #include "opt_tcpdebug.h"
92 #include <sys/param.h>
93 #include <sys/systm.h>
94 #include <sys/kernel.h>
95 #include <sys/malloc.h>
96 #include <sys/sysctl.h>
97 #include <sys/globaldata.h>
98 #include <sys/thread.h>
100 #include <sys/mbuf.h>
102 #include <sys/domain.h>
104 #include <sys/socket.h>
105 #include <sys/socketvar.h>
106 #include <sys/protosw.h>
108 #include <sys/msgport2.h>
111 #include <net/netisr.h>
112 #include <net/route.h>
114 #include <netinet/in.h>
115 #include <netinet/in_systm.h>
117 #include <netinet/ip6.h>
119 #include <netinet/in_pcb.h>
121 #include <netinet6/in6_pcb.h>
123 #include <netinet/in_var.h>
124 #include <netinet/ip_var.h>
126 #include <netinet6/ip6_var.h>
128 #include <netinet/tcp.h>
129 #include <netinet/tcp_fsm.h>
130 #include <netinet/tcp_seq.h>
131 #include <netinet/tcp_timer.h>
132 #include <netinet/tcp_var.h>
133 #include <netinet/tcpip.h>
135 #include <netinet/tcp_debug.h>
139 #include <netinet6/ipsec.h>
143 * TCP protocol interface to socket abstraction.
145 extern char *tcpstates[]; /* XXX ??? */
147 static int tcp_attach (struct socket *, struct pru_attach_info *);
148 static int tcp_connect (struct tcpcb *, struct sockaddr *,
151 static int tcp6_connect (struct tcpcb *, struct sockaddr *,
154 static struct tcpcb *
155 tcp_disconnect (struct tcpcb *);
156 static struct tcpcb *
157 tcp_usrclosed (struct tcpcb *);
160 #define TCPDEBUG0 int ostate = 0
161 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0
162 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
163 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
167 #define TCPDEBUG2(req)
171 * TCP attaches to socket via pru_attach(), reserving space,
172 * and an internet control block.
175 tcp_usr_attach(struct socket *so, int proto, struct pru_attach_info *ai)
179 struct inpcb *inp = sotoinpcb(so);
180 struct tcpcb *tp = 0;
189 error = tcp_attach(so, ai);
193 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
194 so->so_linger = TCP_LINGERTIME;
197 TCPDEBUG2(PRU_ATTACH);
203 * pru_detach() detaches the TCP protocol from the socket.
204 * If the protocol state is non-embryonic, then can't
205 * do this directly: have to initiate a pru_disconnect(),
206 * which may finish later; embryonic TCB's can just
210 tcp_usr_detach(struct socket *so)
214 struct inpcb *inp = sotoinpcb(so);
220 return EINVAL; /* XXX */
224 * It's possible for the tcpcb (tp) to disconnect from the inp due
225 * to tcp_drop()->tcp_close() being called. This may occur *after*
226 * the detach message has been queued so we may find a NULL tp here.
228 if ((tp = intotcpcb(inp)) != NULL) {
230 tp = tcp_disconnect(tp);
231 TCPDEBUG2(PRU_DETACH);
237 #define COMMON_START() TCPDEBUG0; \
243 tp = intotcpcb(inp); \
247 #define COMMON_END(req) out: TCPDEBUG2(req); splx(s); return error; goto out
251 * Give the socket an address.
254 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
258 struct inpcb *inp = sotoinpcb(so);
260 struct sockaddr_in *sinp;
265 * Must check for multicast addresses and disallow binding
268 sinp = (struct sockaddr_in *)nam;
269 if (sinp->sin_family == AF_INET &&
270 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
271 error = EAFNOSUPPORT;
274 error = in_pcbbind(inp, nam, td);
277 COMMON_END(PRU_BIND);
283 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
287 struct inpcb *inp = sotoinpcb(so);
289 struct sockaddr_in6 *sin6p;
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 lwkt_msg nm_lmsg;
328 struct inpcb *nm_inp;
329 struct inpcbinfo *nm_pcbinfo;
333 in_pcbinswildcardhash_handler(struct lwkt_msg *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_lmsg, 0);
344 * Prepare to accept connections.
347 tcp_usr_listen(struct socket *so, struct thread *td)
351 struct inpcb *inp = sotoinpcb(so);
358 if (inp->inp_lport == 0) {
359 error = in_pcbbind(inp, NULL, td);
364 tp->t_state = TCPS_LISTEN;
367 * We have to set the flag because we can't have other cpus messing
368 * with our inp's flags.
370 inp->inp_flags |= INP_WILDCARD_MP;
371 for (cpu = 0; cpu < ncpus2; cpu++) {
372 struct netmsg_inswildcard *msg;
374 if (cpu == mycpu->gd_cpuid) {
375 in_pcbinswildcardhash_oncpu(inp, &tcbinfo[cpu]);
379 msg = malloc(sizeof(struct netmsg_inswildcard), M_LWKTMSG,
381 lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0,
382 lwkt_cmd_func(in_pcbinswildcardhash_handler),
385 msg->nm_pcbinfo = &tcbinfo[cpu];
386 lwkt_sendmsg(tcp_cport(cpu), &msg->nm_lmsg);
389 in_pcbinswildcardhash(inp);
391 COMMON_END(PRU_LISTEN);
396 tcp6_usr_listen(struct socket *so, struct thread *td)
400 struct inpcb *inp = sotoinpcb(so);
404 if (inp->inp_lport == 0) {
405 inp->inp_vflag &= ~INP_IPV4;
406 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
407 inp->inp_vflag |= INP_IPV4;
408 error = in6_pcbbind(inp, (struct sockaddr *)0, td);
411 tp->t_state = TCPS_LISTEN;
412 in_pcbinswildcardhash(inp);
413 COMMON_END(PRU_LISTEN);
418 * Initiate connection to peer.
419 * Create a template for use in transmissions on this connection.
420 * Enter SYN_SENT state, and mark socket as connecting.
421 * Start keep-alive timer, and seed output sequence space.
422 * Send initial segment on connection.
425 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
429 struct inpcb *inp = sotoinpcb(so);
431 struct sockaddr_in *sinp;
436 * Must disallow TCP ``connections'' to multicast addresses.
438 sinp = (struct sockaddr_in *)nam;
439 if (sinp->sin_family == AF_INET
440 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
441 error = EAFNOSUPPORT;
445 prison_remote_ip(td, 0, &sinp->sin_addr.s_addr);
447 if ((error = tcp_connect(tp, nam, td)) != 0)
449 error = tcp_output(tp);
450 COMMON_END(PRU_CONNECT);
455 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
459 struct inpcb *inp = sotoinpcb(so);
461 struct sockaddr_in6 *sin6p;
466 * Must disallow TCP ``connections'' to multicast addresses.
468 sin6p = (struct sockaddr_in6 *)nam;
469 if (sin6p->sin6_family == AF_INET6
470 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
471 error = EAFNOSUPPORT;
475 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
476 struct sockaddr_in sin;
478 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
483 in6_sin6_2_sin(&sin, sin6p);
484 inp->inp_vflag |= INP_IPV4;
485 inp->inp_vflag &= ~INP_IPV6;
486 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
488 error = tcp_output(tp);
491 inp->inp_vflag &= ~INP_IPV4;
492 inp->inp_vflag |= INP_IPV6;
493 inp->inp_inc.inc_isipv6 = 1;
494 if ((error = tcp6_connect(tp, nam, td)) != 0)
496 error = tcp_output(tp);
497 COMMON_END(PRU_CONNECT);
502 * Initiate disconnect from peer.
503 * If connection never passed embryonic stage, just drop;
504 * else if don't need to let data drain, then can just drop anyways,
505 * else have to begin TCP shutdown process: mark socket disconnecting,
506 * drain unread data, state switch to reflect user close, and
507 * send segment (e.g. FIN) to peer. Socket will be really disconnected
508 * when peer sends FIN and acks ours.
510 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
513 tcp_usr_disconnect(struct socket *so)
517 struct inpcb *inp = sotoinpcb(so);
521 tp = tcp_disconnect(tp);
522 COMMON_END(PRU_DISCONNECT);
526 * Accept a connection. Essentially all the work is
527 * done at higher levels; just return the address
528 * of the peer, storing through addr.
531 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
535 struct inpcb *inp = sotoinpcb(so);
536 struct tcpcb *tp = NULL;
539 if (so->so_state & SS_ISDISCONNECTED) {
540 error = ECONNABORTED;
549 in_setpeeraddr(so, nam);
550 COMMON_END(PRU_ACCEPT);
555 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
559 struct inpcb *inp = sotoinpcb(so);
560 struct tcpcb *tp = NULL;
563 if (so->so_state & SS_ISDISCONNECTED) {
564 error = ECONNABORTED;
573 in6_mapped_peeraddr(so, nam);
574 COMMON_END(PRU_ACCEPT);
578 * Mark the connection as being incapable of further output.
581 tcp_usr_shutdown(struct socket *so)
585 struct inpcb *inp = sotoinpcb(so);
590 tp = tcp_usrclosed(tp);
592 error = tcp_output(tp);
593 COMMON_END(PRU_SHUTDOWN);
597 * After a receive, possibly send window update to peer.
600 tcp_usr_rcvd(struct socket *so, int flags)
604 struct inpcb *inp = sotoinpcb(so);
609 COMMON_END(PRU_RCVD);
613 * Do a send by putting data in output queue and updating urgent
614 * marker if URG set. Possibly send more data. Unlike the other
615 * pru_*() routines, the mbuf chains are our responsibility. We
616 * must either enqueue them or free them. The other pru_* routines
617 * generally are caller-frees.
620 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
621 struct sockaddr *nam, struct mbuf *control, struct thread *td)
625 struct inpcb *inp = sotoinpcb(so);
634 * OOPS! we lost a race, the TCP session got reset after
635 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
636 * network interrupt in the non-splnet() section of sosend().
642 error = ECONNRESET; /* XXX EPIPE? */
648 isipv6 = nam && nam->sa_family == AF_INET6;
653 /* TCP doesn't do control messages (rights, creds, etc) */
654 if (control->m_len) {
661 m_freem(control); /* empty control, just free it */
663 if(!(flags & PRUS_OOB)) {
664 sbappend(&so->so_snd, m);
665 if (nam && tp->t_state < TCPS_SYN_SENT) {
667 * Do implied connect if not yet connected,
668 * initialize window to default value, and
669 * initialize maxseg/maxopd using peer's cached
674 error = tcp6_connect(tp, nam, td);
677 error = tcp_connect(tp, nam, td);
680 tp->snd_wnd = TTCP_CLIENT_SND_WND;
684 if (flags & PRUS_EOF) {
686 * Close the send side of the connection after
690 tp = tcp_usrclosed(tp);
693 if (flags & PRUS_MORETOCOME)
694 tp->t_flags |= TF_MORETOCOME;
695 error = tcp_output(tp);
696 if (flags & PRUS_MORETOCOME)
697 tp->t_flags &= ~TF_MORETOCOME;
700 if (sbspace(&so->so_snd) < -512) {
706 * According to RFC961 (Assigned Protocols),
707 * the urgent pointer points to the last octet
708 * of urgent data. We continue, however,
709 * to consider it to indicate the first octet
710 * of data past the urgent section.
711 * Otherwise, snd_up should be one lower.
713 sbappend(&so->so_snd, m);
714 if (nam && tp->t_state < TCPS_SYN_SENT) {
716 * Do implied connect if not yet connected,
717 * initialize window to default value, and
718 * initialize maxseg/maxopd using peer's cached
723 error = tcp6_connect(tp, nam, td);
726 error = tcp_connect(tp, nam, td);
729 tp->snd_wnd = TTCP_CLIENT_SND_WND;
732 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
733 tp->t_flags |= TF_FORCE;
734 error = tcp_output(tp);
735 tp->t_flags &= ~TF_FORCE;
737 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
738 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
745 tcp_usr_abort(struct socket *so)
749 struct inpcb *inp = sotoinpcb(so);
753 tp = tcp_drop(tp, ECONNABORTED);
754 COMMON_END(PRU_ABORT);
758 * Receive out-of-band data.
761 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
765 struct inpcb *inp = sotoinpcb(so);
769 if ((so->so_oobmark == 0 &&
770 (so->so_state & SS_RCVATMARK) == 0) ||
771 so->so_options & SO_OOBINLINE ||
772 tp->t_oobflags & TCPOOB_HADDATA) {
776 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
781 *mtod(m, caddr_t) = tp->t_iobc;
782 if ((flags & MSG_PEEK) == 0)
783 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
784 COMMON_END(PRU_RCVOOB);
787 /* xxx - should be const */
788 struct pr_usrreqs tcp_usrreqs = {
789 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
790 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
791 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
792 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
793 in_setsockaddr, sosend, soreceive, sopoll
797 struct pr_usrreqs tcp6_usrreqs = {
798 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind,
799 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach,
800 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd,
801 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
802 in6_mapped_sockaddr, sosend, soreceive, sopoll
807 tcp_connect_oncpu(struct tcpcb *tp, struct sockaddr_in *sin,
808 struct sockaddr_in *if_sin)
810 struct inpcb *inp = tp->t_inpcb, *oinp;
811 struct socket *so = inp->inp_socket;
813 struct rmxp_tao *taop;
814 struct rmxp_tao tao_noncached;
816 oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid],
817 sin->sin_addr, sin->sin_port,
818 inp->inp_laddr.s_addr != INADDR_ANY ?
819 inp->inp_laddr : if_sin->sin_addr,
820 inp->inp_lport, 0, NULL);
822 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
823 otp->t_state == TCPS_TIME_WAIT &&
824 (ticks - otp->t_starttime) < tcp_msl &&
825 (otp->t_flags & TF_RCVD_CC))
826 (void) tcp_close(otp);
830 if (inp->inp_laddr.s_addr == INADDR_ANY)
831 inp->inp_laddr = if_sin->sin_addr;
832 inp->inp_faddr = sin->sin_addr;
833 inp->inp_fport = sin->sin_port;
834 inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid];
835 in_pcbinsconnhash(inp);
837 /* Compute window scaling to request. */
838 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
839 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
840 tp->request_r_scale++;
843 tcpstat.tcps_connattempt++;
844 tp->t_state = TCPS_SYN_SENT;
845 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
846 tp->iss = tcp_new_isn(tp);
847 tp->t_bw_rtseq = tp->iss;
851 * Generate a CC value for this connection and
852 * check whether CC or CCnew should be used.
854 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
855 taop = &tao_noncached;
856 bzero(taop, sizeof(*taop));
859 tp->cc_send = CC_INC(tcp_ccgen);
860 if (taop->tao_ccsent != 0 &&
861 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
862 taop->tao_ccsent = tp->cc_send;
864 taop->tao_ccsent = 0;
865 tp->t_flags |= TF_SENDCCNEW;
873 struct netmsg_tcp_connect {
874 struct lwkt_msg nm_lmsg;
876 struct sockaddr_in *nm_sin;
877 struct sockaddr_in *nm_ifsin;
881 tcp_connect_handler(lwkt_msg_t lmsg)
883 struct netmsg_tcp_connect *msg = (void *)lmsg;
886 error = tcp_connect_oncpu(msg->nm_tp, msg->nm_sin, msg->nm_ifsin);
887 lwkt_replymsg(lmsg, error);
894 * Common subroutine to open a TCP connection to remote host specified
895 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
896 * port number if needed. Call in_pcbladdr to do the routing and to choose
897 * a local host address (interface). If there is an existing incarnation
898 * of the same connection in TIME-WAIT state and if the remote host was
899 * sending CC options and if the connection duration was < MSL, then
900 * truncate the previous TIME-WAIT state and proceed.
901 * Initialize connection parameters and enter SYN-SENT state.
904 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
906 struct inpcb *inp = tp->t_inpcb;
907 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
908 struct sockaddr_in *if_sin;
914 if (inp->inp_lport == 0) {
915 error = in_pcbbind(inp, (struct sockaddr *)NULL, td);
921 * Cannot simply call in_pcbconnect, because there might be an
922 * earlier incarnation of this same connection still in
923 * TIME_WAIT state, creating an ADDRINUSE error.
925 error = in_pcbladdr(inp, nam, &if_sin);
930 port = tcp_addrport(sin->sin_addr.s_addr, sin->sin_port,
931 inp->inp_laddr.s_addr ?
932 inp->inp_laddr.s_addr : if_sin->sin_addr.s_addr,
935 if (port->mp_td != curthread) {
936 struct netmsg_tcp_connect msg;
938 lwkt_initmsg(&msg.nm_lmsg, &curthread->td_msgport, 0,
939 lwkt_cmd_func(tcp_connect_handler), lwkt_cmd_op_none);
942 msg.nm_ifsin = if_sin;
943 error = lwkt_domsg(port, &msg.nm_lmsg);
946 error = tcp_connect_oncpu(tp, sin, if_sin);
953 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
955 struct inpcb *inp = tp->t_inpcb, *oinp;
956 struct socket *so = inp->inp_socket;
958 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
959 struct in6_addr *addr6;
960 struct rmxp_tao *taop;
961 struct rmxp_tao tao_noncached;
964 if (inp->inp_lport == 0) {
965 error = in6_pcbbind(inp, (struct sockaddr *)0, td);
971 * Cannot simply call in_pcbconnect, because there might be an
972 * earlier incarnation of this same connection still in
973 * TIME_WAIT state, creating an ADDRINUSE error.
975 error = in6_pcbladdr(inp, nam, &addr6);
978 oinp = in6_pcblookup_hash(inp->inp_cpcbinfo,
979 &sin6->sin6_addr, sin6->sin6_port,
980 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
981 addr6 : &inp->in6p_laddr,
982 inp->inp_lport, 0, NULL);
984 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
985 otp->t_state == TCPS_TIME_WAIT &&
986 (ticks - otp->t_starttime) < tcp_msl &&
987 (otp->t_flags & TF_RCVD_CC))
988 otp = tcp_close(otp);
992 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
993 inp->in6p_laddr = *addr6;
994 inp->in6p_faddr = sin6->sin6_addr;
995 inp->inp_fport = sin6->sin6_port;
996 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL)
997 inp->in6p_flowinfo = sin6->sin6_flowinfo;
998 in_pcbinsconnhash(inp);
1000 /* Compute window scaling to request. */
1001 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1002 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
1003 tp->request_r_scale++;
1006 tcpstat.tcps_connattempt++;
1007 tp->t_state = TCPS_SYN_SENT;
1008 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
1009 tp->iss = tcp_new_isn(tp);
1010 tp->t_bw_rtseq = tp->iss;
1011 tcp_sendseqinit(tp);
1014 * Generate a CC value for this connection and
1015 * check whether CC or CCnew should be used.
1017 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
1018 taop = &tao_noncached;
1019 bzero(taop, sizeof(*taop));
1022 tp->cc_send = CC_INC(tcp_ccgen);
1023 if (taop->tao_ccsent != 0 &&
1024 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
1025 taop->tao_ccsent = tp->cc_send;
1027 taop->tao_ccsent = 0;
1028 tp->t_flags |= TF_SENDCCNEW;
1036 * The new sockopt interface makes it possible for us to block in the
1037 * copyin/out step (if we take a page fault). Taking a page fault at
1038 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now
1039 * use TSM, there probably isn't any need for this function to run at
1040 * splnet() any more. This needs more examination.)
1043 tcp_ctloutput(so, sopt)
1045 struct sockopt *sopt;
1047 int error, opt, optval, s;
1052 s = splnet(); /* XXX */
1053 inp = sotoinpcb(so);
1056 return (ECONNRESET);
1058 if (sopt->sopt_level != IPPROTO_TCP) {
1060 if (INP_CHECK_SOCKAF(so, AF_INET6))
1061 error = ip6_ctloutput(so, sopt);
1064 error = ip_ctloutput(so, sopt);
1068 tp = intotcpcb(inp);
1070 switch (sopt->sopt_dir) {
1072 switch (sopt->sopt_name) {
1075 error = sooptcopyin(sopt, &optval, sizeof optval,
1080 switch (sopt->sopt_name) {
1088 opt = 0; /* dead code to fool gcc */
1095 tp->t_flags &= ~opt;
1099 error = sooptcopyin(sopt, &optval, sizeof optval,
1105 tp->t_flags |= TF_NOPUSH;
1107 tp->t_flags &= ~TF_NOPUSH;
1108 error = tcp_output(tp);
1113 error = sooptcopyin(sopt, &optval, sizeof optval,
1118 if (optval > 0 && optval <= tp->t_maxseg)
1119 tp->t_maxseg = optval;
1125 error = ENOPROTOOPT;
1131 switch (sopt->sopt_name) {
1133 optval = tp->t_flags & TF_NODELAY;
1136 optval = tp->t_maxseg;
1139 optval = tp->t_flags & TF_NOOPT;
1142 optval = tp->t_flags & TF_NOPUSH;
1145 error = ENOPROTOOPT;
1149 error = sooptcopyout(sopt, &optval, sizeof optval);
1157 * tcp_sendspace and tcp_recvspace are the default send and receive window
1158 * sizes, respectively. These are obsolescent (this information should
1159 * be set by the route).
1161 u_long tcp_sendspace = 1024*32;
1162 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1163 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1164 u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */
1165 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1166 &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1169 * Attach TCP protocol to socket, allocating
1170 * internet protocol control block, tcp control block,
1171 * bufer space, and entering LISTEN state if to accept connections.
1174 tcp_attach(struct socket *so, struct pru_attach_info *ai)
1181 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL;
1184 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1185 error = soreserve(so, tcp_sendspace, tcp_recvspace,
1190 cpu = mycpu->gd_cpuid;
1191 error = in_pcballoc(so, &tcbinfo[cpu]);
1194 inp = sotoinpcb(so);
1197 inp->inp_vflag |= INP_IPV6;
1198 inp->in6p_hops = -1; /* use kernel default */
1202 inp->inp_vflag |= INP_IPV4;
1203 tp = tcp_newtcpcb(inp);
1205 int nofd = so->so_state & SS_NOFDREF; /* XXX */
1207 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */
1214 so->so_state |= nofd;
1217 tp->t_state = TCPS_CLOSED;
1222 * Initiate (or continue) disconnect.
1223 * If embryonic state, just send reset (once).
1224 * If in ``let data drain'' option and linger null, just drop.
1225 * Otherwise (hard), mark socket disconnecting and drop
1226 * current input data; switch states based on user close, and
1227 * send segment to peer (with FIN).
1229 static struct tcpcb *
1233 struct socket *so = tp->t_inpcb->inp_socket;
1235 if (tp->t_state < TCPS_ESTABLISHED)
1237 else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1238 tp = tcp_drop(tp, 0);
1240 soisdisconnecting(so);
1241 sbflush(&so->so_rcv);
1242 tp = tcp_usrclosed(tp);
1244 (void) tcp_output(tp);
1250 * User issued close, and wish to trail through shutdown states:
1251 * if never received SYN, just forget it. If got a SYN from peer,
1252 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1253 * If already got a FIN from peer, then almost done; go to LAST_ACK
1254 * state. In all other cases, have already sent FIN to peer (e.g.
1255 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1256 * for peer to send FIN or not respond to keep-alives, etc.
1257 * We can let the user exit from the close as soon as the FIN is acked.
1259 static struct tcpcb *
1264 switch (tp->t_state) {
1268 tp->t_state = TCPS_CLOSED;
1273 case TCPS_SYN_RECEIVED:
1274 tp->t_flags |= TF_NEEDFIN;
1277 case TCPS_ESTABLISHED:
1278 tp->t_state = TCPS_FIN_WAIT_1;
1281 case TCPS_CLOSE_WAIT:
1282 tp->t_state = TCPS_LAST_ACK;
1285 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1286 soisdisconnected(tp->t_inpcb->inp_socket);
1287 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
1288 if (tp->t_state == TCPS_FIN_WAIT_2)
1289 callout_reset(tp->tt_2msl, tcp_maxidle,
1290 tcp_timer_2msl, tp);