2 * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2011, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2011, by Michael Tuexen. All rights reserved.
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
9 * a) Redistributions of source code must retain the above copyright notice,
10 * this list of conditions and the following disclaimer.
12 * b) Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in
14 * the documentation and/or other materials provided with the distribution.
16 * c) Neither the name of Cisco Systems, Inc. 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 LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
32 /* $KAME: sctp6_usrreq.c,v 1.38 2005/08/24 08:08:56 suz Exp $ */
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
37 #include <netinet/sctp_os.h>
39 #include <netinet/sctp_pcb.h>
40 #include <netinet/sctp_header.h>
41 #include <netinet/sctp_var.h>
43 #include <netinet6/sctp6_var.h>
45 #include <netinet/sctp_sysctl.h>
46 #include <netinet/sctp_output.h>
47 #include <netinet/sctp_uio.h>
48 #include <netinet/sctp_asconf.h>
49 #include <netinet/sctputil.h>
50 #include <netinet/sctp_indata.h>
51 #include <netinet/sctp_timer.h>
52 #include <netinet/sctp_auth.h>
53 #include <netinet/sctp_input.h>
54 #include <netinet/sctp_output.h>
55 #include <netinet/sctp_bsd_addr.h>
56 #include <netinet/sctp_crc32.h>
57 #include <netinet/udp.h>
60 #include <netipsec/ipsec.h>
62 #include <netipsec/ipsec6.h>
66 extern struct protosw inetsw[];
69 sctp6_input(struct mbuf **i_pak, int *offp, int proto)
74 struct sctp_inpcb *in6p = NULL;
75 struct sctp_nets *net;
80 struct inpcb *in6p_ip;
83 struct sctp_chunkhdr *ch;
84 int length, offset, iphlen;
86 struct sctp_tcb *stcb = NULL;
89 #if !defined(SCTP_WITH_NO_CSUM)
90 uint32_t check, calc_check;
96 /* get the VRF and table id's */
97 if (SCTP_GET_PKT_VRFID(*i_pak, vrf_id)) {
98 SCTP_RELEASE_PKT(*i_pak);
101 m = SCTP_HEADER_TO_CHAIN(*i_pak);
102 pkt_len = SCTP_HEADER_LEN((*i_pak));
104 #ifdef SCTP_PACKET_LOGGING
105 sctp_packet_log(m, pkt_len);
107 ip6 = mtod(m, struct ip6_hdr *);
108 /* Ensure that (sctphdr + sctp_chunkhdr) in a row. */
109 IP6_EXTHDR_GET(sh, struct sctphdr *, m, off,
110 (int)(sizeof(*sh) + sizeof(*ch)));
112 SCTP_STAT_INCR(sctps_hdrops);
115 ch = (struct sctp_chunkhdr *)((caddr_t)sh + sizeof(struct sctphdr));
117 offset = iphlen + sizeof(*sh) + sizeof(*ch);
118 SCTPDBG(SCTP_DEBUG_INPUT1,
119 "sctp6_input() length:%d iphlen:%d\n", pkt_len, iphlen);
122 #if defined(NFAITH) && NFAITH > 0
124 if (faithprefix_p != NULL && (*faithprefix_p) (&ip6->ip6_dst)) {
125 /* XXX send icmp6 host/port unreach? */
128 #endif /* NFAITH defined and > 0 */
129 SCTP_STAT_INCR(sctps_recvpackets);
130 SCTP_STAT_INCR_COUNTER64(sctps_inpackets);
131 SCTPDBG(SCTP_DEBUG_INPUT1, "V6 input gets a packet iphlen:%d pktlen:%d\n",
133 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
134 /* No multi-cast support in SCTP */
137 /* destination port of 0 is illegal, based on RFC2960. */
138 if (sh->dest_port == 0)
141 SCTPDBG(SCTP_DEBUG_CRCOFFLOAD,
142 "sctp_input(): Packet of length %d received on %s with csum_flags 0x%x.\n",
144 if_name(m->m_pkthdr.rcvif),
145 m->m_pkthdr.csum_flags);
146 #if defined(SCTP_WITH_NO_CSUM)
147 SCTP_STAT_INCR(sctps_recvnocrc);
149 if (m->m_pkthdr.csum_flags & CSUM_SCTP_VALID) {
150 SCTP_STAT_INCR(sctps_recvhwcrc);
153 check = sh->checksum; /* save incoming checksum */
154 if ((check == 0) && (SCTP_BASE_SYSCTL(sctp_no_csum_on_loopback)) &&
155 (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &ip6->ip6_dst))) {
156 SCTP_STAT_INCR(sctps_recvnocrc);
159 sh->checksum = 0; /* prepare for calc */
160 calc_check = sctp_calculate_cksum(m, iphlen);
161 SCTP_STAT_INCR(sctps_recvswcrc);
162 if (calc_check != check) {
163 SCTPDBG(SCTP_DEBUG_INPUT1, "Bad CSUM on SCTP packet calc_check:%x check:%x m:%p phlen:%d\n",
164 calc_check, check, m, iphlen);
165 stcb = sctp_findassociation_addr(m, iphlen, offset - sizeof(*ch),
166 sh, ch, &in6p, &net, vrf_id);
167 if ((net) && (port)) {
168 if (net->port == 0) {
169 sctp_pathmtu_adjustment(in6p, stcb, net, net->mtu - sizeof(struct udphdr));
173 if ((net != NULL) && (m->m_flags & M_FLOWID)) {
174 net->flowid = m->m_pkthdr.flowid;
179 /* in6p's ref-count increased && stcb locked */
180 if ((in6p) && (stcb)) {
181 sctp_send_packet_dropped(stcb, net, m, iphlen, 1);
182 sctp_chunk_output((struct sctp_inpcb *)in6p, stcb, SCTP_OUTPUT_FROM_INPUT_ERROR, SCTP_SO_NOT_LOCKED);
183 } else if ((in6p != NULL) && (stcb == NULL)) {
186 SCTP_STAT_INCR(sctps_badsum);
187 SCTP_STAT_INCR_COUNTER32(sctps_checksumerrors);
190 sh->checksum = calc_check;
196 * Locate pcb and tcb for datagram sctp_findassociation_addr() wants
197 * IP/SCTP/first chunk header...
199 stcb = sctp_findassociation_addr(m, iphlen, offset - sizeof(*ch),
200 sh, ch, &in6p, &net, vrf_id);
201 if ((net) && (port)) {
202 if (net->port == 0) {
203 sctp_pathmtu_adjustment(in6p, stcb, net, net->mtu - sizeof(struct udphdr));
207 if ((net != NULL) && (m->m_flags & M_FLOWID)) {
208 net->flowid = m->m_pkthdr.flowid;
213 /* in6p's ref-count increased */
215 struct sctp_init_chunk *init_chk, chunk_buf;
217 SCTP_STAT_INCR(sctps_noport);
218 if (ch->chunk_type == SCTP_INITIATION) {
220 * we do a trick here to get the INIT tag, dig in
221 * and get the tag from the INIT and put it in the
224 init_chk = (struct sctp_init_chunk *)sctp_m_getptr(m,
225 iphlen + sizeof(*sh), sizeof(*init_chk),
226 (uint8_t *) & chunk_buf);
228 sh->v_tag = init_chk->init.initiate_tag;
232 if (ch->chunk_type == SCTP_SHUTDOWN_ACK) {
233 sctp_send_shutdown_complete2(m, iphlen, sh, vrf_id, port);
236 if (ch->chunk_type == SCTP_SHUTDOWN_COMPLETE) {
239 if (ch->chunk_type != SCTP_ABORT_ASSOCIATION)
240 sctp_send_abort(m, iphlen, sh, 0, NULL, vrf_id, port);
242 } else if (stcb == NULL) {
247 * Check AH/ESP integrity.
249 in6p_ip = (struct inpcb *)in6p;
250 if (in6p_ip && (ipsec6_in_reject(m, in6p_ip))) {
252 MODULE_GLOBAL(ipsec6stat).in_polvio++;
258 * CONTROL chunk processing
260 offset -= sizeof(*ch);
261 ecn_bits = ((ntohl(ip6->ip6_flow) >> 20) & 0x000000ff);
263 /* Length now holds the total packet length payload + iphlen */
264 length = ntohs(ip6->ip6_plen) + iphlen;
266 /* sa_ignore NO_NULL_CHK */
267 sctp_common_input_processing(&m, iphlen, offset, length, sh, ch,
268 in6p, stcb, net, ecn_bits, vrf_id, port);
269 /* inp's ref-count reduced && stcb unlocked */
270 /* XXX this stuff below gets moved to appropriate parts later... */
273 if ((in6p) && refcount_up) {
274 /* reduce ref-count */
275 SCTP_INP_WLOCK(in6p);
276 SCTP_INP_DECR_REF(in6p);
277 SCTP_INP_WUNLOCK(in6p);
283 SCTP_TCB_UNLOCK(stcb);
285 if ((in6p) && refcount_up) {
286 /* reduce ref-count */
287 SCTP_INP_WLOCK(in6p);
288 SCTP_INP_DECR_REF(in6p);
289 SCTP_INP_WUNLOCK(in6p);
298 sctp6_notify_mbuf(struct sctp_inpcb *inp, struct icmp6_hdr *icmp6,
299 struct sctphdr *sh, struct sctp_tcb *stcb, struct sctp_nets *net)
303 if ((inp == NULL) || (stcb == NULL) || (net == NULL) ||
304 (icmp6 == NULL) || (sh == NULL)) {
307 /* First do we even look at it? */
308 if (ntohl(sh->v_tag) != (stcb->asoc.peer_vtag))
311 if (icmp6->icmp6_type != ICMP6_PACKET_TOO_BIG) {
312 /* not PACKET TO BIG */
316 * ok we need to look closely. We could even get smarter and look at
317 * anyone that we sent to in case we get a different ICMP that tells
318 * us there is no way to reach a host, but for this impl, all we
319 * care about is MTU discovery.
321 nxtsz = ntohl(icmp6->icmp6_mtu);
322 /* Stop any PMTU timer */
323 sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, NULL, SCTP_FROM_SCTP6_USRREQ + SCTP_LOC_1);
325 /* Adjust destination size limit */
326 if (net->mtu > nxtsz) {
329 net->mtu -= sizeof(struct udphdr);
332 /* now what about the ep? */
333 if (stcb->asoc.smallest_mtu > nxtsz) {
334 struct sctp_tmit_chunk *chk;
336 /* Adjust that too */
337 stcb->asoc.smallest_mtu = nxtsz;
338 /* now off to subtract IP_DF flag if needed */
340 TAILQ_FOREACH(chk, &stcb->asoc.send_queue, sctp_next) {
341 if ((uint32_t) (chk->send_size + IP_HDR_SIZE) > nxtsz) {
342 chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
345 TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
346 if ((uint32_t) (chk->send_size + IP_HDR_SIZE) > nxtsz) {
348 * For this guy we also mark for immediate
349 * resend since we sent to big of chunk
351 chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
352 if (chk->sent != SCTP_DATAGRAM_RESEND)
353 stcb->asoc.sent_queue_retran_cnt++;
354 chk->sent = SCTP_DATAGRAM_RESEND;
355 chk->rec.data.doing_fast_retransmit = 0;
357 chk->sent = SCTP_DATAGRAM_RESEND;
358 /* Clear any time so NO RTT is being done */
359 chk->sent_rcv_time.tv_sec = 0;
360 chk->sent_rcv_time.tv_usec = 0;
361 stcb->asoc.total_flight -= chk->send_size;
362 net->flight_size -= chk->send_size;
366 sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, NULL);
369 SCTP_TCB_UNLOCK(stcb);
375 sctp6_notify(struct sctp_inpcb *inp,
376 struct icmp6_hdr *icmph,
379 struct sctp_tcb *stcb,
380 struct sctp_nets *net)
382 #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
390 if ((inp == NULL) || (stcb == NULL) || (net == NULL) ||
391 (sh == NULL) || (to == NULL)) {
393 SCTP_TCB_UNLOCK(stcb);
396 /* First job is to verify the vtag matches what I would send */
397 if (ntohl(sh->v_tag) != (stcb->asoc.peer_vtag)) {
398 SCTP_TCB_UNLOCK(stcb);
401 if (icmph->icmp6_type != ICMP_UNREACH) {
402 /* We only care about unreachable */
403 SCTP_TCB_UNLOCK(stcb);
406 if ((icmph->icmp6_code == ICMP_UNREACH_NET) ||
407 (icmph->icmp6_code == ICMP_UNREACH_HOST) ||
408 (icmph->icmp6_code == ICMP_UNREACH_NET_UNKNOWN) ||
409 (icmph->icmp6_code == ICMP_UNREACH_HOST_UNKNOWN) ||
410 (icmph->icmp6_code == ICMP_UNREACH_ISOLATED) ||
411 (icmph->icmp6_code == ICMP_UNREACH_NET_PROHIB) ||
412 (icmph->icmp6_code == ICMP_UNREACH_HOST_PROHIB) ||
413 (icmph->icmp6_code == ICMP_UNREACH_FILTER_PROHIB)) {
416 * Hmm reachablity problems we must examine closely. If its
417 * not reachable, we may have lost a network. Or if there is
418 * NO protocol at the other end named SCTP. well we consider
421 if (net->dest_state & SCTP_ADDR_REACHABLE) {
422 /* Ok that destination is NOT reachable */
423 SCTP_PRINTF("ICMP (thresh %d/%d) takes interface %p down\n",
425 net->failure_threshold,
428 net->dest_state &= ~SCTP_ADDR_REACHABLE;
429 net->dest_state |= SCTP_ADDR_NOT_REACHABLE;
431 * JRS 5/14/07 - If a destination is unreachable,
432 * the PF bit is turned off. This allows an
433 * unambiguous use of the PF bit for destinations
434 * that are reachable but potentially failed. If the
435 * destination is set to the unreachable state, also
436 * set the destination to the PF state.
439 * Add debug message here if destination is not in
442 /* Stop any running T3 timers here? */
443 if ((stcb->asoc.sctp_cmt_on_off > 0) &&
444 (stcb->asoc.sctp_cmt_pf > 0)) {
445 net->dest_state &= ~SCTP_ADDR_PF;
446 SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from PF to unreachable.\n",
449 net->error_count = net->failure_threshold + 1;
450 sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
451 stcb, SCTP_FAILED_THRESHOLD,
452 (void *)net, SCTP_SO_NOT_LOCKED);
454 SCTP_TCB_UNLOCK(stcb);
455 } else if ((icmph->icmp6_code == ICMP_UNREACH_PROTOCOL) ||
456 (icmph->icmp6_code == ICMP_UNREACH_PORT)) {
458 * Here the peer is either playing tricks on us, including
459 * an address that belongs to someone who does not support
460 * SCTP OR was a userland implementation that shutdown and
461 * now is dead. In either case treat it like a OOTB abort
464 reason = SCTP_PEER_FAULTY;
465 sctp_abort_notification(stcb, reason, SCTP_SO_NOT_LOCKED);
466 #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
467 so = SCTP_INP_SO(inp);
468 atomic_add_int(&stcb->asoc.refcnt, 1);
469 SCTP_TCB_UNLOCK(stcb);
470 SCTP_SOCKET_LOCK(so, 1);
472 atomic_subtract_int(&stcb->asoc.refcnt, 1);
474 (void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_2);
475 #if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
476 SCTP_SOCKET_UNLOCK(so, 1);
477 /* SCTP_TCB_UNLOCK(stcb); MT: I think this is not needed. */
479 /* no need to unlock here, since the TCB is gone */
481 SCTP_TCB_UNLOCK(stcb);
488 sctp6_ctlinput(int cmd, struct sockaddr *pktdst, void *d)
491 struct ip6ctlparam *ip6cp = NULL;
494 vrf_id = SCTP_DEFAULT_VRFID;
496 if (pktdst->sa_family != AF_INET6 ||
497 pktdst->sa_len != sizeof(struct sockaddr_in6))
500 if ((unsigned)cmd >= PRC_NCMDS)
502 if (PRC_IS_REDIRECT(cmd)) {
504 } else if (inet6ctlerrmap[cmd] == 0) {
507 /* if the parameter is from icmp6, decode it. */
509 ip6cp = (struct ip6ctlparam *)d;
511 ip6cp = (struct ip6ctlparam *)NULL;
516 * XXX: We assume that when IPV6 is non NULL, M and OFF are
519 /* check if we can safely examine src and dst ports */
520 struct sctp_inpcb *inp = NULL;
521 struct sctp_tcb *stcb = NULL;
522 struct sctp_nets *net = NULL;
523 struct sockaddr_in6 final;
525 if (ip6cp->ip6c_m == NULL)
528 bzero(&sh, sizeof(sh));
529 bzero(&final, sizeof(final));
532 m_copydata(ip6cp->ip6c_m, ip6cp->ip6c_off, sizeof(sh),
534 ip6cp->ip6c_src->sin6_port = sh.src_port;
535 final.sin6_len = sizeof(final);
536 final.sin6_family = AF_INET6;
537 final.sin6_addr = ((struct sockaddr_in6 *)pktdst)->sin6_addr;
538 final.sin6_port = sh.dest_port;
539 stcb = sctp_findassociation_addr_sa((struct sockaddr *)ip6cp->ip6c_src,
540 (struct sockaddr *)&final,
541 &inp, &net, 1, vrf_id);
542 /* inp's ref-count increased && stcb locked */
543 if (stcb != NULL && inp && (inp->sctp_socket != NULL)) {
544 if (cmd == PRC_MSGSIZE) {
545 sctp6_notify_mbuf(inp,
550 /* inp's ref-count reduced && stcb unlocked */
552 sctp6_notify(inp, ip6cp->ip6c_icmp6, &sh,
553 (struct sockaddr *)&final,
555 /* inp's ref-count reduced && stcb unlocked */
558 if (PRC_IS_REDIRECT(cmd) && inp) {
559 in6_rtchange((struct in6pcb *)inp,
560 inet6ctlerrmap[cmd]);
563 /* reduce inp's ref-count */
565 SCTP_INP_DECR_REF(inp);
566 SCTP_INP_WUNLOCK(inp);
569 SCTP_TCB_UNLOCK(stcb);
575 * this routine can probably be collasped into the one in sctp_userreq.c
576 * since they do the same thing and now we lookup with a sockaddr
579 sctp6_getcred(SYSCTL_HANDLER_ARGS)
582 struct sockaddr_in6 addrs[2];
583 struct sctp_inpcb *inp;
584 struct sctp_nets *net;
585 struct sctp_tcb *stcb;
589 vrf_id = SCTP_DEFAULT_VRFID;
591 error = priv_check(req->td, PRIV_NETINET_GETCRED);
595 if (req->newlen != sizeof(addrs)) {
596 SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
599 if (req->oldlen != sizeof(struct ucred)) {
600 SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
603 error = SYSCTL_IN(req, addrs, sizeof(addrs));
607 stcb = sctp_findassociation_addr_sa(sin6tosa(&addrs[0]),
609 &inp, &net, 1, vrf_id);
610 if (stcb == NULL || inp == NULL || inp->sctp_socket == NULL) {
611 if ((inp != NULL) && (stcb == NULL)) {
612 /* reduce ref-count */
614 SCTP_INP_DECR_REF(inp);
617 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ENOENT);
621 SCTP_TCB_UNLOCK(stcb);
623 * We use the write lock here, only since in the error leg we need
624 * it. If we used RLOCK, then we would have to
625 * wlock/decr/unlock/rlock. Which in theory could create a hole.
626 * Better to use higher wlock.
630 error = cr_canseesocket(req->td->td_ucred, inp->sctp_socket);
632 SCTP_INP_WUNLOCK(inp);
635 cru2x(inp->sctp_socket->so_cred, &xuc);
636 SCTP_INP_WUNLOCK(inp);
637 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
642 SYSCTL_PROC(_net_inet6_sctp6, OID_AUTO, getcred, CTLTYPE_OPAQUE | CTLFLAG_RW,
644 sctp6_getcred, "S,ucred", "Get the ucred of a SCTP6 connection");
647 /* This is the same as the sctp_abort() could be made common */
649 sctp6_abort(struct socket *so)
651 struct sctp_inpcb *inp;
654 inp = (struct sctp_inpcb *)so->so_pcb;
656 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
660 flags = inp->sctp_flags;
661 #ifdef SCTP_LOG_CLOSING
662 sctp_log_closing(inp, NULL, 17);
664 if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) &&
665 (atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) {
666 #ifdef SCTP_LOG_CLOSING
667 sctp_log_closing(inp, NULL, 16);
669 sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT,
670 SCTP_CALLED_AFTER_CMPSET_OFCLOSE);
672 SCTP_SB_CLEAR(so->so_snd);
674 * same for the rcv ones, they are only here for the
677 SCTP_SB_CLEAR(so->so_rcv);
678 /* Now null out the reference, we are completely detached. */
682 flags = inp->sctp_flags;
683 if ((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) {
684 goto sctp_must_try_again;
691 sctp6_attach(struct socket *so, int proto, struct thread *p)
695 struct sctp_inpcb *inp;
696 uint32_t vrf_id = SCTP_DEFAULT_VRFID;
698 inp = (struct sctp_inpcb *)so->so_pcb;
700 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
703 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
704 error = SCTP_SORESERVE(so, SCTP_BASE_SYSCTL(sctp_sendspace), SCTP_BASE_SYSCTL(sctp_recvspace));
708 error = sctp_inpcb_alloc(so, vrf_id);
711 inp = (struct sctp_inpcb *)so->so_pcb;
713 inp->sctp_flags |= SCTP_PCB_FLAGS_BOUND_V6; /* I'm v6! */
714 inp6 = (struct in6pcb *)inp;
716 inp6->inp_vflag |= INP_IPV6;
717 inp6->in6p_hops = -1; /* use kernel default */
718 inp6->in6p_cksum = -1; /* just to be sure */
721 * XXX: ugly!! IPv4 TTL initialization is necessary for an IPv6
722 * socket as well, because the socket may be bound to an IPv6
723 * wildcard address, which may match an IPv4-mapped IPv6 address.
725 inp6->inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
728 * Hmm what about the IPSEC stuff that is missing here but in
731 SCTP_INP_WUNLOCK(inp);
736 sctp6_bind(struct socket *so, struct sockaddr *addr, struct thread *p)
738 struct sctp_inpcb *inp;
742 inp = (struct sctp_inpcb *)so->so_pcb;
744 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
748 if ((addr->sa_family == AF_INET6) &&
749 (addr->sa_len != sizeof(struct sockaddr_in6))) {
750 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
753 if ((addr->sa_family == AF_INET) &&
754 (addr->sa_len != sizeof(struct sockaddr_in))) {
755 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
759 inp6 = (struct in6pcb *)inp;
760 inp6->inp_vflag &= ~INP_IPV4;
761 inp6->inp_vflag |= INP_IPV6;
762 if ((addr != NULL) && (SCTP_IPV6_V6ONLY(inp6) == 0)) {
763 if (addr->sa_family == AF_INET) {
764 /* binding v4 addr to v6 socket, so reset flags */
765 inp6->inp_vflag |= INP_IPV4;
766 inp6->inp_vflag &= ~INP_IPV6;
768 struct sockaddr_in6 *sin6_p;
770 sin6_p = (struct sockaddr_in6 *)addr;
772 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr)) {
773 inp6->inp_vflag |= INP_IPV4;
774 } else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
775 struct sockaddr_in sin;
777 in6_sin6_2_sin(&sin, sin6_p);
778 inp6->inp_vflag |= INP_IPV4;
779 inp6->inp_vflag &= ~INP_IPV6;
780 error = sctp_inpcb_bind(so, (struct sockaddr *)&sin, NULL, p);
784 } else if (addr != NULL) {
785 /* IPV6_V6ONLY socket */
786 if (addr->sa_family == AF_INET) {
787 /* can't bind v4 addr to v6 only socket! */
788 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
791 struct sockaddr_in6 *sin6_p;
793 sin6_p = (struct sockaddr_in6 *)addr;
795 if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
796 /* can't bind v4-mapped addrs either! */
797 /* NOTE: we don't support SIIT */
798 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
803 error = sctp_inpcb_bind(so, addr, NULL, p);
809 sctp6_close(struct socket *so)
814 /* This could be made common with sctp_detach() since they are identical */
818 sctp6_disconnect(struct socket *so)
820 return (sctp_disconnect(so));
825 sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
826 struct mbuf *control, struct thread *p);
830 sctp6_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
831 struct mbuf *control, struct thread *p)
833 struct sctp_inpcb *inp;
837 struct sockaddr_in6 *sin6;
840 /* No SPL needed since sctp_output does this */
842 inp = (struct sctp_inpcb *)so->so_pcb;
845 SCTP_RELEASE_PKT(control);
849 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
852 inp6 = (struct in6pcb *)inp;
854 * For the TCP model we may get a NULL addr, if we are a connected
857 if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) &&
864 SCTP_RELEASE_PKT(control);
867 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EDESTADDRREQ);
868 return (EDESTADDRREQ);
871 sin6 = (struct sockaddr_in6 *)addr;
872 if (SCTP_IPV6_V6ONLY(inp6)) {
874 * if IPV6_V6ONLY flag, we discard datagrams destined to a
875 * v4 addr or v4-mapped addr
877 if (addr->sa_family == AF_INET) {
878 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
881 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
882 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
886 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
887 if (!MODULE_GLOBAL(ip6_v6only)) {
888 struct sockaddr_in sin;
890 /* convert v4-mapped into v4 addr and send */
891 in6_sin6_2_sin(&sin, sin6);
892 return sctp_sendm(so, flags, m, (struct sockaddr *)&sin,
895 /* mapped addresses aren't enabled */
896 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
902 /* now what about control */
905 SCTP_PRINTF("huh? control set?\n");
906 SCTP_RELEASE_PKT(inp->control);
909 inp->control = control;
913 SCTP_BUF_NEXT(inp->pkt_last) = m;
916 inp->pkt_last = inp->pkt = m;
919 /* FreeBSD and MacOSX uses a flag passed */
920 ((flags & PRUS_MORETOCOME) == 0)
923 * note with the current version this code will only be used
924 * by OpenBSD, NetBSD and FreeBSD have methods for
925 * re-defining sosend() to use sctp_sosend(). One can
926 * optionaly switch back to this code (by changing back the
927 * defininitions but this is not advisable.
931 ret = sctp_output(inp, inp->pkt, addr, inp->control, p, flags);
941 sctp6_connect(struct socket *so, struct sockaddr *addr, struct thread *p)
945 struct sctp_inpcb *inp;
947 struct sctp_tcb *stcb;
950 struct sockaddr_in6 *sin6;
951 struct sockaddr_storage ss;
955 inp6 = (struct in6pcb *)so->so_pcb;
956 inp = (struct sctp_inpcb *)so->so_pcb;
958 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ECONNRESET);
959 return (ECONNRESET); /* I made the same as TCP since we are
963 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
966 if ((addr->sa_family == AF_INET6) && (addr->sa_len != sizeof(struct sockaddr_in6))) {
967 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
970 if ((addr->sa_family == AF_INET) && (addr->sa_len != sizeof(struct sockaddr_in))) {
971 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
974 vrf_id = inp->def_vrf_id;
975 SCTP_ASOC_CREATE_LOCK(inp);
977 if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) ==
978 SCTP_PCB_FLAGS_UNBOUND) {
979 /* Bind a ephemeral port */
980 SCTP_INP_RUNLOCK(inp);
981 error = sctp6_bind(so, NULL, p);
983 SCTP_ASOC_CREATE_UNLOCK(inp);
989 if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) &&
990 (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) {
991 /* We are already connected AND the TCP model */
992 SCTP_INP_RUNLOCK(inp);
993 SCTP_ASOC_CREATE_UNLOCK(inp);
994 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EADDRINUSE);
998 sin6 = (struct sockaddr_in6 *)addr;
999 if (SCTP_IPV6_V6ONLY(inp6)) {
1001 * if IPV6_V6ONLY flag, ignore connections destined to a v4
1002 * addr or v4-mapped addr
1004 if (addr->sa_family == AF_INET) {
1005 SCTP_INP_RUNLOCK(inp);
1006 SCTP_ASOC_CREATE_UNLOCK(inp);
1007 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
1010 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1011 SCTP_INP_RUNLOCK(inp);
1012 SCTP_ASOC_CREATE_UNLOCK(inp);
1013 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
1017 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1018 if (!MODULE_GLOBAL(ip6_v6only)) {
1019 /* convert v4-mapped into v4 addr */
1020 in6_sin6_2_sin((struct sockaddr_in *)&ss, sin6);
1021 addr = (struct sockaddr *)&ss;
1023 /* mapped addresses aren't enabled */
1024 SCTP_INP_RUNLOCK(inp);
1025 SCTP_ASOC_CREATE_UNLOCK(inp);
1026 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
1031 addr = addr; /* for true v6 address case */
1033 /* Now do we connect? */
1034 if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
1035 stcb = LIST_FIRST(&inp->sctp_asoc_list);
1037 SCTP_TCB_UNLOCK(stcb);
1039 SCTP_INP_RUNLOCK(inp);
1041 SCTP_INP_RUNLOCK(inp);
1042 SCTP_INP_WLOCK(inp);
1043 SCTP_INP_INCR_REF(inp);
1044 SCTP_INP_WUNLOCK(inp);
1045 stcb = sctp_findassociation_ep_addr(&inp, addr, NULL, NULL, NULL);
1047 SCTP_INP_WLOCK(inp);
1048 SCTP_INP_DECR_REF(inp);
1049 SCTP_INP_WUNLOCK(inp);
1054 /* Already have or am bring up an association */
1055 SCTP_ASOC_CREATE_UNLOCK(inp);
1056 SCTP_TCB_UNLOCK(stcb);
1057 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EALREADY);
1060 /* We are GOOD to go */
1061 stcb = sctp_aloc_assoc(inp, addr, &error, 0, vrf_id, p);
1062 SCTP_ASOC_CREATE_UNLOCK(inp);
1064 /* Gak! no memory */
1067 if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
1068 stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED;
1069 /* Set the connected flag so we can queue data */
1072 stcb->asoc.state = SCTP_STATE_COOKIE_WAIT;
1073 (void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered);
1075 /* initialize authentication parameters for the assoc */
1076 sctp_initialize_auth_params(inp, stcb);
1078 sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED);
1079 SCTP_TCB_UNLOCK(stcb);
1084 sctp6_getaddr(struct socket *so, struct sockaddr **addr)
1086 struct sockaddr_in6 *sin6;
1087 struct sctp_inpcb *inp;
1089 struct sctp_ifa *sctp_ifa;
1094 * Do the malloc first in case it blocks.
1096 SCTP_MALLOC_SONAME(sin6, struct sockaddr_in6 *, sizeof *sin6);
1099 sin6->sin6_family = AF_INET6;
1100 sin6->sin6_len = sizeof(*sin6);
1102 inp = (struct sctp_inpcb *)so->so_pcb;
1104 SCTP_FREE_SONAME(sin6);
1105 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ECONNRESET);
1108 SCTP_INP_RLOCK(inp);
1109 sin6->sin6_port = inp->sctp_lport;
1110 if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
1111 /* For the bound all case you get back 0 */
1112 if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
1113 struct sctp_tcb *stcb;
1114 struct sockaddr_in6 *sin_a6;
1115 struct sctp_nets *net;
1118 stcb = LIST_FIRST(&inp->sctp_asoc_list);
1124 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1125 sin_a6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1127 /* this will make coverity happy */
1130 if (sin_a6->sin6_family == AF_INET6) {
1135 if ((!fnd) || (sin_a6 == NULL)) {
1139 vrf_id = inp->def_vrf_id;
1140 sctp_ifa = sctp_source_address_selection(inp, stcb, (sctp_route_t *) & net->ro, net, 0, vrf_id);
1142 sin6->sin6_addr = sctp_ifa->address.sin6.sin6_addr;
1145 /* For the bound all case you get back 0 */
1147 memset(&sin6->sin6_addr, 0, sizeof(sin6->sin6_addr));
1150 /* Take the first IPv6 address in the list */
1151 struct sctp_laddr *laddr;
1154 LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1155 if (laddr->ifa->address.sa.sa_family == AF_INET6) {
1156 struct sockaddr_in6 *sin_a;
1158 sin_a = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
1159 sin6->sin6_addr = sin_a->sin6_addr;
1165 SCTP_FREE_SONAME(sin6);
1166 SCTP_INP_RUNLOCK(inp);
1167 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ENOENT);
1171 SCTP_INP_RUNLOCK(inp);
1172 /* Scoping things for v6 */
1173 if ((error = sa6_recoverscope(sin6)) != 0) {
1174 SCTP_FREE_SONAME(sin6);
1177 (*addr) = (struct sockaddr *)sin6;
1182 sctp6_peeraddr(struct socket *so, struct sockaddr **addr)
1184 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)*addr;
1186 struct sockaddr_in6 *sin_a6;
1187 struct sctp_inpcb *inp;
1188 struct sctp_tcb *stcb;
1189 struct sctp_nets *net;
1194 * Do the malloc first in case it blocks.
1196 inp = (struct sctp_inpcb *)so->so_pcb;
1197 if ((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0) {
1198 /* UDP type and listeners will drop out here */
1199 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ENOTCONN);
1202 SCTP_MALLOC_SONAME(sin6, struct sockaddr_in6 *, sizeof *sin6);
1205 sin6->sin6_family = AF_INET6;
1206 sin6->sin6_len = sizeof(*sin6);
1208 /* We must recapture incase we blocked */
1209 inp = (struct sctp_inpcb *)so->so_pcb;
1211 SCTP_FREE_SONAME(sin6);
1212 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ECONNRESET);
1215 SCTP_INP_RLOCK(inp);
1216 stcb = LIST_FIRST(&inp->sctp_asoc_list);
1218 SCTP_TCB_LOCK(stcb);
1220 SCTP_INP_RUNLOCK(inp);
1222 SCTP_FREE_SONAME(sin6);
1223 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ECONNRESET);
1227 TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1228 sin_a6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1229 if (sin_a6->sin6_family == AF_INET6) {
1231 sin6->sin6_port = stcb->rport;
1232 sin6->sin6_addr = sin_a6->sin6_addr;
1236 SCTP_TCB_UNLOCK(stcb);
1238 /* No IPv4 address */
1239 SCTP_FREE_SONAME(sin6);
1240 SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, ENOENT);
1243 if ((error = sa6_recoverscope(sin6)) != 0)
1245 *addr = (struct sockaddr *)sin6;
1250 sctp6_in6getaddr(struct socket *so, struct sockaddr **nam)
1252 struct sockaddr *addr;
1253 struct in6pcb *inp6 = sotoin6pcb(so);
1257 SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
1260 /* allow v6 addresses precedence */
1261 error = sctp6_getaddr(so, nam);
1263 /* try v4 next if v6 failed */
1264 error = sctp_ingetaddr(so, nam);
1269 /* if I'm V6ONLY, convert it to v4-mapped */
1270 if (SCTP_IPV6_V6ONLY(inp6)) {
1271 struct sockaddr_in6 sin6;
1273 in6_sin_2_v4mapsin6((struct sockaddr_in *)addr, &sin6);
1274 memcpy(addr, &sin6, sizeof(struct sockaddr_in6));
1283 sctp6_getpeeraddr(struct socket *so, struct sockaddr **nam)
1285 struct sockaddr *addr = *nam;
1286 struct in6pcb *inp6 = sotoin6pcb(so);
1290 SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP6_USRREQ, EINVAL);
1293 /* allow v6 addresses precedence */
1294 error = sctp6_peeraddr(so, nam);
1296 /* try v4 next if v6 failed */
1297 error = sctp_peeraddr(so, nam);
1301 /* if I'm V6ONLY, convert it to v4-mapped */
1302 if (SCTP_IPV6_V6ONLY(inp6)) {
1303 struct sockaddr_in6 sin6;
1305 in6_sin_2_v4mapsin6((struct sockaddr_in *)addr, &sin6);
1306 memcpy(addr, &sin6, sizeof(struct sockaddr_in6));
1312 struct pr_usrreqs sctp6_usrreqs = {
1313 .pru_abort = sctp6_abort,
1314 .pru_accept = sctp_accept,
1315 .pru_attach = sctp6_attach,
1316 .pru_bind = sctp6_bind,
1317 .pru_connect = sctp6_connect,
1318 .pru_control = in6_control,
1319 .pru_close = sctp6_close,
1320 .pru_detach = sctp6_close,
1321 .pru_sopoll = sopoll_generic,
1322 .pru_flush = sctp_flush,
1323 .pru_disconnect = sctp6_disconnect,
1324 .pru_listen = sctp_listen,
1325 .pru_peeraddr = sctp6_getpeeraddr,
1326 .pru_send = sctp6_send,
1327 .pru_shutdown = sctp_shutdown,
1328 .pru_sockaddr = sctp6_in6getaddr,
1329 .pru_sosend = sctp_sosend,
1330 .pru_soreceive = sctp_soreceive