1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* SCTP kernel implementation
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 International Business Machines, Corp.
6 * Copyright (c) 2001 Intel Corp.
7 * Copyright (c) 2001 Nokia, Inc.
8 * Copyright (c) 2001 La Monte H.P. Yarroll
10 * This file is part of the SCTP kernel implementation
12 * These functions handle all input from the IP layer into SCTP.
14 * Please send any bug reports or fixes you make to the
16 * lksctp developers <linux-sctp@vger.kernel.org>
18 * Written or modified by:
19 * La Monte H.P. Yarroll <piggy@acm.org>
20 * Karl Knutson <karl@athena.chicago.il.us>
21 * Xingang Guo <xingang.guo@intel.com>
22 * Jon Grimm <jgrimm@us.ibm.com>
23 * Hui Huang <hui.huang@nokia.com>
24 * Daisy Chang <daisyc@us.ibm.com>
25 * Sridhar Samudrala <sri@us.ibm.com>
26 * Ardelle Fan <ardelle.fan@intel.com>
29 #include <linux/types.h>
30 #include <linux/list.h> /* For struct list_head */
31 #include <linux/socket.h>
33 #include <linux/time.h> /* For struct timeval */
34 #include <linux/slab.h>
40 #include <net/sctp/sctp.h>
41 #include <net/sctp/sm.h>
42 #include <net/sctp/checksum.h>
43 #include <net/net_namespace.h>
44 #include <linux/rhashtable.h>
45 #include <net/sock_reuseport.h>
47 /* Forward declarations for internal helpers. */
48 static int sctp_rcv_ootb(struct sk_buff *);
49 static struct sctp_association *__sctp_rcv_lookup(struct net *net,
51 const union sctp_addr *paddr,
52 const union sctp_addr *laddr,
53 struct sctp_transport **transportp);
54 static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(
55 struct net *net, struct sk_buff *skb,
56 const union sctp_addr *laddr,
57 const union sctp_addr *daddr);
58 static struct sctp_association *__sctp_lookup_association(
60 const union sctp_addr *local,
61 const union sctp_addr *peer,
62 struct sctp_transport **pt);
64 static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb);
67 /* Calculate the SCTP checksum of an SCTP packet. */
68 static inline int sctp_rcv_checksum(struct net *net, struct sk_buff *skb)
70 struct sctphdr *sh = sctp_hdr(skb);
71 __le32 cmp = sh->checksum;
72 __le32 val = sctp_compute_cksum(skb, 0);
75 /* CRC failure, dump it. */
76 __SCTP_INC_STATS(net, SCTP_MIB_CHECKSUMERRORS);
83 * This is the routine which IP calls when receiving an SCTP packet.
85 int sctp_rcv(struct sk_buff *skb)
88 struct sctp_association *asoc;
89 struct sctp_endpoint *ep = NULL;
90 struct sctp_ep_common *rcvr;
91 struct sctp_transport *transport = NULL;
92 struct sctp_chunk *chunk;
97 struct net *net = dev_net(skb->dev);
98 bool is_gso = skb_is_gso(skb) && skb_is_gso_sctp(skb);
100 if (skb->pkt_type != PACKET_HOST)
103 __SCTP_INC_STATS(net, SCTP_MIB_INSCTPPACKS);
105 /* If packet is too small to contain a single chunk, let's not
106 * waste time on it anymore.
108 if (skb->len < sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr) +
109 skb_transport_offset(skb))
112 /* If the packet is fragmented and we need to do crc checking,
113 * it's better to just linearize it otherwise crc computing
116 if ((!is_gso && skb_linearize(skb)) ||
117 !pskb_may_pull(skb, sizeof(struct sctphdr)))
120 /* Pull up the IP header. */
121 __skb_pull(skb, skb_transport_offset(skb));
123 skb->csum_valid = 0; /* Previous value not applicable */
124 if (skb_csum_unnecessary(skb))
125 __skb_decr_checksum_unnecessary(skb);
126 else if (!sctp_checksum_disable &&
128 sctp_rcv_checksum(net, skb) < 0)
132 __skb_pull(skb, sizeof(struct sctphdr));
134 family = ipver2af(ip_hdr(skb)->version);
135 af = sctp_get_af_specific(family);
138 SCTP_INPUT_CB(skb)->af = af;
140 /* Initialize local addresses for lookups. */
141 af->from_skb(&src, skb, 1);
142 af->from_skb(&dest, skb, 0);
144 /* If the packet is to or from a non-unicast address,
145 * silently discard the packet.
147 * This is not clearly defined in the RFC except in section
148 * 8.4 - OOTB handling. However, based on the book "Stream Control
149 * Transmission Protocol" 2.1, "It is important to note that the
150 * IP address of an SCTP transport address must be a routable
151 * unicast address. In other words, IP multicast addresses and
152 * IP broadcast addresses cannot be used in an SCTP transport
155 if (!af->addr_valid(&src, NULL, skb) ||
156 !af->addr_valid(&dest, NULL, skb))
159 asoc = __sctp_rcv_lookup(net, skb, &src, &dest, &transport);
162 ep = __sctp_rcv_lookup_endpoint(net, skb, &dest, &src);
164 /* Retrieve the common input handling substructure. */
165 rcvr = asoc ? &asoc->base : &ep->base;
169 * If a frame arrives on an interface and the receiving socket is
170 * bound to another interface, via SO_BINDTODEVICE, treat it as OOTB
172 if (sk->sk_bound_dev_if && (sk->sk_bound_dev_if != af->skb_iif(skb))) {
174 sctp_transport_put(transport);
178 sctp_endpoint_put(ep);
181 sk = net->sctp.ctl_sock;
182 ep = sctp_sk(sk)->ep;
183 sctp_endpoint_hold(ep);
188 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
189 * An SCTP packet is called an "out of the blue" (OOTB)
190 * packet if it is correctly formed, i.e., passed the
191 * receiver's checksum check, but the receiver is not
192 * able to identify the association to which this
196 if (sctp_rcv_ootb(skb)) {
197 __SCTP_INC_STATS(net, SCTP_MIB_OUTOFBLUES);
198 goto discard_release;
202 if (!xfrm_policy_check(sk, XFRM_POLICY_IN, skb, family))
203 goto discard_release;
206 if (sk_filter(sk, skb))
207 goto discard_release;
209 /* Create an SCTP packet structure. */
210 chunk = sctp_chunkify(skb, asoc, sk, GFP_ATOMIC);
212 goto discard_release;
213 SCTP_INPUT_CB(skb)->chunk = chunk;
215 /* Remember what endpoint is to handle this packet. */
218 /* Remember the SCTP header. */
219 chunk->sctp_hdr = sctp_hdr(skb);
221 /* Set the source and destination addresses of the incoming chunk. */
222 sctp_init_addrs(chunk, &src, &dest);
224 /* Remember where we came from. */
225 chunk->transport = transport;
227 /* Acquire access to the sock lock. Note: We are safe from other
228 * bottom halves on this lock, but a user may be in the lock too,
229 * so check if it is busy.
233 if (sk != rcvr->sk) {
234 /* Our cached sk is different from the rcvr->sk. This is
235 * because migrate()/accept() may have moved the association
236 * to a new socket and released all the sockets. So now we
237 * are holding a lock on the old socket while the user may
238 * be doing something with the new socket. Switch our veiw
246 if (sock_owned_by_user(sk) || !sctp_newsk_ready(sk)) {
247 if (sctp_add_backlog(sk, skb)) {
249 sctp_chunk_free(chunk);
250 skb = NULL; /* sctp_chunk_free already freed the skb */
251 goto discard_release;
253 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_BACKLOG);
255 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_SOFTIRQ);
256 sctp_inq_push(&chunk->rcvr->inqueue, chunk);
261 /* Release the asoc/ep ref we took in the lookup calls. */
263 sctp_transport_put(transport);
265 sctp_endpoint_put(ep);
270 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_DISCARDS);
275 /* Release the asoc/ep ref we took in the lookup calls. */
277 sctp_transport_put(transport);
279 sctp_endpoint_put(ep);
284 /* Process the backlog queue of the socket. Every skb on
285 * the backlog holds a ref on an association or endpoint.
286 * We hold this ref throughout the state machine to make
287 * sure that the structure we need is still around.
289 int sctp_backlog_rcv(struct sock *sk, struct sk_buff *skb)
291 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
292 struct sctp_inq *inqueue = &chunk->rcvr->inqueue;
293 struct sctp_transport *t = chunk->transport;
294 struct sctp_ep_common *rcvr = NULL;
299 /* If the rcvr is dead then the association or endpoint
300 * has been deleted and we can safely drop the chunk
301 * and refs that we are holding.
304 sctp_chunk_free(chunk);
308 if (unlikely(rcvr->sk != sk)) {
309 /* In this case, the association moved from one socket to
310 * another. We are currently sitting on the backlog of the
311 * old socket, so we need to move.
312 * However, since we are here in the process context we
313 * need to take make sure that the user doesn't own
314 * the new socket when we process the packet.
315 * If the new socket is user-owned, queue the chunk to the
316 * backlog of the new socket without dropping any refs.
317 * Otherwise, we can safely push the chunk on the inqueue.
324 if (sock_owned_by_user(sk) || !sctp_newsk_ready(sk)) {
325 if (sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf)))
326 sctp_chunk_free(chunk);
330 sctp_inq_push(inqueue, chunk);
335 /* If the chunk was backloged again, don't drop refs */
339 if (!sctp_newsk_ready(sk)) {
340 if (!sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf)))
342 sctp_chunk_free(chunk);
344 sctp_inq_push(inqueue, chunk);
349 /* Release the refs we took in sctp_add_backlog */
350 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
351 sctp_transport_put(t);
352 else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
353 sctp_endpoint_put(sctp_ep(rcvr));
360 static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb)
362 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
363 struct sctp_transport *t = chunk->transport;
364 struct sctp_ep_common *rcvr = chunk->rcvr;
367 ret = sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf));
369 /* Hold the assoc/ep while hanging on the backlog queue.
370 * This way, we know structures we need will not disappear
373 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
374 sctp_transport_hold(t);
375 else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
376 sctp_endpoint_hold(sctp_ep(rcvr));
384 /* Handle icmp frag needed error. */
385 void sctp_icmp_frag_needed(struct sock *sk, struct sctp_association *asoc,
386 struct sctp_transport *t, __u32 pmtu)
389 (t->pathmtu <= pmtu &&
390 t->pl.probe_size + sctp_transport_pl_hlen(t) <= pmtu))
393 if (sock_owned_by_user(sk)) {
394 atomic_set(&t->mtu_info, pmtu);
395 asoc->pmtu_pending = 1;
400 if (!(t->param_flags & SPP_PMTUD_ENABLE))
401 /* We can't allow retransmitting in such case, as the
402 * retransmission would be sized just as before, and thus we
403 * would get another icmp, and retransmit again.
407 /* Update transports view of the MTU. Return if no update was needed.
408 * If an update wasn't needed/possible, it also doesn't make sense to
409 * try to retransmit now.
411 if (!sctp_transport_update_pmtu(t, pmtu))
414 /* Update association pmtu. */
415 sctp_assoc_sync_pmtu(asoc);
417 /* Retransmit with the new pmtu setting. */
418 sctp_retransmit(&asoc->outqueue, t, SCTP_RTXR_PMTUD);
421 void sctp_icmp_redirect(struct sock *sk, struct sctp_transport *t,
424 struct dst_entry *dst;
426 if (sock_owned_by_user(sk) || !t)
428 dst = sctp_transport_dst_check(t);
430 dst->ops->redirect(dst, sk, skb);
434 * SCTP Implementer's Guide, 2.37 ICMP handling procedures
436 * ICMP8) If the ICMP code is a "Unrecognized next header type encountered"
437 * or a "Protocol Unreachable" treat this message as an abort
438 * with the T bit set.
440 * This function sends an event to the state machine, which will abort the
444 void sctp_icmp_proto_unreachable(struct sock *sk,
445 struct sctp_association *asoc,
446 struct sctp_transport *t)
448 if (sock_owned_by_user(sk)) {
449 if (timer_pending(&t->proto_unreach_timer))
452 if (!mod_timer(&t->proto_unreach_timer,
454 sctp_transport_hold(t);
457 struct net *net = sock_net(sk);
459 pr_debug("%s: unrecognized next header type "
460 "encountered!\n", __func__);
462 if (del_timer(&t->proto_unreach_timer))
463 sctp_transport_put(t);
465 sctp_do_sm(net, SCTP_EVENT_T_OTHER,
466 SCTP_ST_OTHER(SCTP_EVENT_ICMP_PROTO_UNREACH),
467 asoc->state, asoc->ep, asoc, t,
472 /* Common lookup code for icmp/icmpv6 error handler. */
473 struct sock *sctp_err_lookup(struct net *net, int family, struct sk_buff *skb,
474 struct sctphdr *sctphdr,
475 struct sctp_association **app,
476 struct sctp_transport **tpp)
478 struct sctp_init_chunk *chunkhdr, _chunkhdr;
479 union sctp_addr saddr;
480 union sctp_addr daddr;
482 struct sock *sk = NULL;
483 struct sctp_association *asoc;
484 struct sctp_transport *transport = NULL;
485 __u32 vtag = ntohl(sctphdr->vtag);
487 *app = NULL; *tpp = NULL;
489 af = sctp_get_af_specific(family);
494 /* Initialize local addresses for lookups. */
495 af->from_skb(&saddr, skb, 1);
496 af->from_skb(&daddr, skb, 0);
498 /* Look for an association that matches the incoming ICMP error
501 asoc = __sctp_lookup_association(net, &saddr, &daddr, &transport);
507 /* RFC 4960, Appendix C. ICMP Handling
509 * ICMP6) An implementation MUST validate that the Verification Tag
510 * contained in the ICMP message matches the Verification Tag of
511 * the peer. If the Verification Tag is not 0 and does NOT
512 * match, discard the ICMP message. If it is 0 and the ICMP
513 * message contains enough bytes to verify that the chunk type is
514 * an INIT chunk and that the Initiate Tag matches the tag of the
515 * peer, continue with ICMP7. If the ICMP message is too short
516 * or the chunk type or the Initiate Tag does not match, silently
517 * discard the packet.
520 /* chunk header + first 4 octects of init header */
521 chunkhdr = skb_header_pointer(skb, skb_transport_offset(skb) +
522 sizeof(struct sctphdr),
523 sizeof(struct sctp_chunkhdr) +
524 sizeof(__be32), &_chunkhdr);
526 chunkhdr->chunk_hdr.type != SCTP_CID_INIT ||
527 ntohl(chunkhdr->init_hdr.init_tag) != asoc->c.my_vtag)
530 } else if (vtag != asoc->c.peer_vtag) {
536 /* If too many ICMPs get dropped on busy
537 * servers this needs to be solved differently.
539 if (sock_owned_by_user(sk))
540 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
547 sctp_transport_put(transport);
551 /* Common cleanup code for icmp/icmpv6 error handler. */
552 void sctp_err_finish(struct sock *sk, struct sctp_transport *t)
553 __releases(&((__sk)->sk_lock.slock))
556 sctp_transport_put(t);
559 static void sctp_v4_err_handle(struct sctp_transport *t, struct sk_buff *skb,
560 __u8 type, __u8 code, __u32 info)
562 struct sctp_association *asoc = t->asoc;
563 struct sock *sk = asoc->base.sk;
567 case ICMP_PARAMETERPROB:
570 case ICMP_DEST_UNREACH:
571 if (code > NR_ICMP_UNREACH)
573 if (code == ICMP_FRAG_NEEDED) {
574 sctp_icmp_frag_needed(sk, asoc, t, SCTP_TRUNC4(info));
577 if (code == ICMP_PROT_UNREACH) {
578 sctp_icmp_proto_unreachable(sk, asoc, t);
581 err = icmp_err_convert[code].errno;
583 case ICMP_TIME_EXCEEDED:
584 if (code == ICMP_EXC_FRAGTIME)
590 sctp_icmp_redirect(sk, t, skb);
595 if (!sock_owned_by_user(sk) && inet_sk(sk)->recverr) {
598 } else { /* Only an error on timeout */
599 sk->sk_err_soft = err;
604 * This routine is called by the ICMP module when it gets some
605 * sort of error condition. If err < 0 then the socket should
606 * be closed and the error returned to the user. If err > 0
607 * it's just the icmp type << 8 | icmp code. After adjustment
608 * header points to the first 8 bytes of the sctp header. We need
609 * to find the appropriate port.
611 * The locking strategy used here is very "optimistic". When
612 * someone else accesses the socket the ICMP is just dropped
613 * and for some paths there is no check at all.
614 * A more general error queue to queue errors for later handling
615 * is probably better.
618 int sctp_v4_err(struct sk_buff *skb, __u32 info)
620 const struct iphdr *iph = (const struct iphdr *)skb->data;
621 const int type = icmp_hdr(skb)->type;
622 const int code = icmp_hdr(skb)->code;
623 struct net *net = dev_net(skb->dev);
624 struct sctp_transport *transport;
625 struct sctp_association *asoc;
626 __u16 saveip, savesctp;
629 /* Fix up skb to look at the embedded net header. */
630 saveip = skb->network_header;
631 savesctp = skb->transport_header;
632 skb_reset_network_header(skb);
633 skb_set_transport_header(skb, iph->ihl * 4);
634 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &transport);
635 /* Put back, the original values. */
636 skb->network_header = saveip;
637 skb->transport_header = savesctp;
639 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
643 sctp_v4_err_handle(transport, skb, type, code, info);
644 sctp_err_finish(sk, transport);
649 int sctp_udp_v4_err(struct sock *sk, struct sk_buff *skb)
651 struct net *net = dev_net(skb->dev);
652 struct sctp_association *asoc;
653 struct sctp_transport *t;
657 skb->transport_header += sizeof(struct udphdr);
658 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &t);
660 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
664 skb->transport_header -= sizeof(struct udphdr);
665 hdr = (struct icmphdr *)(skb_network_header(skb) - sizeof(struct icmphdr));
666 if (hdr->type == ICMP_REDIRECT) {
667 /* can't be handled without outer iphdr known, leave it to udp_err */
668 sctp_err_finish(sk, t);
671 if (hdr->type == ICMP_DEST_UNREACH && hdr->code == ICMP_FRAG_NEEDED)
672 info = ntohs(hdr->un.frag.mtu);
673 sctp_v4_err_handle(t, skb, hdr->type, hdr->code, info);
675 sctp_err_finish(sk, t);
680 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
682 * This function scans all the chunks in the OOTB packet to determine if
683 * the packet should be discarded right away. If a response might be needed
684 * for this packet, or, if further processing is possible, the packet will
685 * be queued to a proper inqueue for the next phase of handling.
688 * Return 0 - If further processing is needed.
689 * Return 1 - If the packet can be discarded right away.
691 static int sctp_rcv_ootb(struct sk_buff *skb)
693 struct sctp_chunkhdr *ch, _ch;
694 int ch_end, offset = 0;
696 /* Scan through all the chunks in the packet. */
698 /* Make sure we have at least the header there */
699 if (offset + sizeof(_ch) > skb->len)
702 ch = skb_header_pointer(skb, offset, sizeof(*ch), &_ch);
704 /* Break out if chunk length is less then minimal. */
705 if (!ch || ntohs(ch->length) < sizeof(_ch))
708 ch_end = offset + SCTP_PAD4(ntohs(ch->length));
709 if (ch_end > skb->len)
712 /* RFC 8.4, 2) If the OOTB packet contains an ABORT chunk, the
713 * receiver MUST silently discard the OOTB packet and take no
716 if (SCTP_CID_ABORT == ch->type)
719 /* RFC 8.4, 6) If the packet contains a SHUTDOWN COMPLETE
720 * chunk, the receiver should silently discard the packet
721 * and take no further action.
723 if (SCTP_CID_SHUTDOWN_COMPLETE == ch->type)
727 * This will discard packets with INIT chunk bundled as
728 * subsequent chunks in the packet. When INIT is first,
729 * the normal INIT processing will discard the chunk.
731 if (SCTP_CID_INIT == ch->type && (void *)ch != skb->data)
735 } while (ch_end < skb->len);
743 /* Insert endpoint into the hash table. */
744 static int __sctp_hash_endpoint(struct sctp_endpoint *ep)
746 struct sock *sk = ep->base.sk;
747 struct net *net = sock_net(sk);
748 struct sctp_hashbucket *head;
750 ep->hashent = sctp_ep_hashfn(net, ep->base.bind_addr.port);
751 head = &sctp_ep_hashtable[ep->hashent];
753 if (sk->sk_reuseport) {
754 bool any = sctp_is_ep_boundall(sk);
755 struct sctp_endpoint *ep2;
756 struct list_head *list;
757 int cnt = 0, err = 1;
759 list_for_each(list, &ep->base.bind_addr.address_list)
762 sctp_for_each_hentry(ep2, &head->chain) {
763 struct sock *sk2 = ep2->base.sk;
765 if (!net_eq(sock_net(sk2), net) || sk2 == sk ||
766 !uid_eq(sock_i_uid(sk2), sock_i_uid(sk)) ||
770 err = sctp_bind_addrs_check(sctp_sk(sk2),
773 err = reuseport_add_sock(sk, sk2, any);
777 } else if (err < 0) {
783 err = reuseport_alloc(sk, any);
789 write_lock(&head->lock);
790 hlist_add_head(&ep->node, &head->chain);
791 write_unlock(&head->lock);
795 /* Add an endpoint to the hash. Local BH-safe. */
796 int sctp_hash_endpoint(struct sctp_endpoint *ep)
801 err = __sctp_hash_endpoint(ep);
807 /* Remove endpoint from the hash table. */
808 static void __sctp_unhash_endpoint(struct sctp_endpoint *ep)
810 struct sock *sk = ep->base.sk;
811 struct sctp_hashbucket *head;
813 ep->hashent = sctp_ep_hashfn(sock_net(sk), ep->base.bind_addr.port);
815 head = &sctp_ep_hashtable[ep->hashent];
817 if (rcu_access_pointer(sk->sk_reuseport_cb))
818 reuseport_detach_sock(sk);
820 write_lock(&head->lock);
821 hlist_del_init(&ep->node);
822 write_unlock(&head->lock);
825 /* Remove endpoint from the hash. Local BH-safe. */
826 void sctp_unhash_endpoint(struct sctp_endpoint *ep)
829 __sctp_unhash_endpoint(ep);
833 static inline __u32 sctp_hashfn(const struct net *net, __be16 lport,
834 const union sctp_addr *paddr, __u32 seed)
838 if (paddr->sa.sa_family == AF_INET6)
839 addr = jhash(&paddr->v6.sin6_addr, 16, seed);
841 addr = (__force __u32)paddr->v4.sin_addr.s_addr;
843 return jhash_3words(addr, ((__force __u32)paddr->v4.sin_port) << 16 |
844 (__force __u32)lport, net_hash_mix(net), seed);
847 /* Look up an endpoint. */
848 static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(
849 struct net *net, struct sk_buff *skb,
850 const union sctp_addr *laddr,
851 const union sctp_addr *paddr)
853 struct sctp_hashbucket *head;
854 struct sctp_endpoint *ep;
859 lport = laddr->v4.sin_port;
860 hash = sctp_ep_hashfn(net, ntohs(lport));
861 head = &sctp_ep_hashtable[hash];
862 read_lock(&head->lock);
863 sctp_for_each_hentry(ep, &head->chain) {
864 if (sctp_endpoint_is_match(ep, net, laddr))
868 ep = sctp_sk(net->sctp.ctl_sock)->ep;
872 if (sk->sk_reuseport) {
873 __u32 phash = sctp_hashfn(net, lport, paddr, 0);
875 sk = reuseport_select_sock(sk, phash, skb,
876 sizeof(struct sctphdr));
878 ep = sctp_sk(sk)->ep;
880 sctp_endpoint_hold(ep);
881 read_unlock(&head->lock);
885 /* rhashtable for transport */
886 struct sctp_hash_cmp_arg {
887 const union sctp_addr *paddr;
888 const struct net *net;
892 static inline int sctp_hash_cmp(struct rhashtable_compare_arg *arg,
895 struct sctp_transport *t = (struct sctp_transport *)ptr;
896 const struct sctp_hash_cmp_arg *x = arg->key;
899 if (!sctp_cmp_addr_exact(&t->ipaddr, x->paddr))
901 if (!sctp_transport_hold(t))
904 if (!net_eq(t->asoc->base.net, x->net))
906 if (x->lport != htons(t->asoc->base.bind_addr.port))
911 sctp_transport_put(t);
915 static inline __u32 sctp_hash_obj(const void *data, u32 len, u32 seed)
917 const struct sctp_transport *t = data;
919 return sctp_hashfn(t->asoc->base.net,
920 htons(t->asoc->base.bind_addr.port),
924 static inline __u32 sctp_hash_key(const void *data, u32 len, u32 seed)
926 const struct sctp_hash_cmp_arg *x = data;
928 return sctp_hashfn(x->net, x->lport, x->paddr, seed);
931 static const struct rhashtable_params sctp_hash_params = {
932 .head_offset = offsetof(struct sctp_transport, node),
933 .hashfn = sctp_hash_key,
934 .obj_hashfn = sctp_hash_obj,
935 .obj_cmpfn = sctp_hash_cmp,
936 .automatic_shrinking = true,
939 int sctp_transport_hashtable_init(void)
941 return rhltable_init(&sctp_transport_hashtable, &sctp_hash_params);
944 void sctp_transport_hashtable_destroy(void)
946 rhltable_destroy(&sctp_transport_hashtable);
949 int sctp_hash_transport(struct sctp_transport *t)
951 struct sctp_transport *transport;
952 struct rhlist_head *tmp, *list;
953 struct sctp_hash_cmp_arg arg;
959 arg.net = t->asoc->base.net;
960 arg.paddr = &t->ipaddr;
961 arg.lport = htons(t->asoc->base.bind_addr.port);
964 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
967 rhl_for_each_entry_rcu(transport, tmp, list, node)
968 if (transport->asoc->ep == t->asoc->ep) {
974 err = rhltable_insert_key(&sctp_transport_hashtable, &arg,
975 &t->node, sctp_hash_params);
977 pr_err_once("insert transport fail, errno %d\n", err);
982 void sctp_unhash_transport(struct sctp_transport *t)
987 rhltable_remove(&sctp_transport_hashtable, &t->node,
991 /* return a transport with holding it */
992 struct sctp_transport *sctp_addrs_lookup_transport(
994 const union sctp_addr *laddr,
995 const union sctp_addr *paddr)
997 struct rhlist_head *tmp, *list;
998 struct sctp_transport *t;
999 struct sctp_hash_cmp_arg arg = {
1002 .lport = laddr->v4.sin_port,
1005 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
1008 rhl_for_each_entry_rcu(t, tmp, list, node) {
1009 if (!sctp_transport_hold(t))
1012 if (sctp_bind_addr_match(&t->asoc->base.bind_addr,
1013 laddr, sctp_sk(t->asoc->base.sk)))
1015 sctp_transport_put(t);
1021 /* return a transport without holding it, as it's only used under sock lock */
1022 struct sctp_transport *sctp_epaddr_lookup_transport(
1023 const struct sctp_endpoint *ep,
1024 const union sctp_addr *paddr)
1026 struct rhlist_head *tmp, *list;
1027 struct sctp_transport *t;
1028 struct sctp_hash_cmp_arg arg = {
1030 .net = ep->base.net,
1031 .lport = htons(ep->base.bind_addr.port),
1034 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
1037 rhl_for_each_entry_rcu(t, tmp, list, node)
1038 if (ep == t->asoc->ep)
1044 /* Look up an association. */
1045 static struct sctp_association *__sctp_lookup_association(
1047 const union sctp_addr *local,
1048 const union sctp_addr *peer,
1049 struct sctp_transport **pt)
1051 struct sctp_transport *t;
1052 struct sctp_association *asoc = NULL;
1054 t = sctp_addrs_lookup_transport(net, local, peer);
1065 /* Look up an association. protected by RCU read lock */
1067 struct sctp_association *sctp_lookup_association(struct net *net,
1068 const union sctp_addr *laddr,
1069 const union sctp_addr *paddr,
1070 struct sctp_transport **transportp)
1072 struct sctp_association *asoc;
1075 asoc = __sctp_lookup_association(net, laddr, paddr, transportp);
1081 /* Is there an association matching the given local and peer addresses? */
1082 bool sctp_has_association(struct net *net,
1083 const union sctp_addr *laddr,
1084 const union sctp_addr *paddr)
1086 struct sctp_transport *transport;
1088 if (sctp_lookup_association(net, laddr, paddr, &transport)) {
1089 sctp_transport_put(transport);
1097 * SCTP Implementors Guide, 2.18 Handling of address
1098 * parameters within the INIT or INIT-ACK.
1100 * D) When searching for a matching TCB upon reception of an INIT
1101 * or INIT-ACK chunk the receiver SHOULD use not only the
1102 * source address of the packet (containing the INIT or
1103 * INIT-ACK) but the receiver SHOULD also use all valid
1104 * address parameters contained within the chunk.
1106 * 2.18.3 Solution description
1108 * This new text clearly specifies to an implementor the need
1109 * to look within the INIT or INIT-ACK. Any implementation that
1110 * does not do this, may not be able to establish associations
1111 * in certain circumstances.
1114 static struct sctp_association *__sctp_rcv_init_lookup(struct net *net,
1115 struct sk_buff *skb,
1116 const union sctp_addr *laddr, struct sctp_transport **transportp)
1118 struct sctp_association *asoc;
1119 union sctp_addr addr;
1120 union sctp_addr *paddr = &addr;
1121 struct sctphdr *sh = sctp_hdr(skb);
1122 union sctp_params params;
1123 struct sctp_init_chunk *init;
1127 * This code will NOT touch anything inside the chunk--it is
1128 * strictly READ-ONLY.
1130 * RFC 2960 3 SCTP packet Format
1132 * Multiple chunks can be bundled into one SCTP packet up to
1133 * the MTU size, except for the INIT, INIT ACK, and SHUTDOWN
1134 * COMPLETE chunks. These chunks MUST NOT be bundled with any
1135 * other chunk in a packet. See Section 6.10 for more details
1136 * on chunk bundling.
1139 /* Find the start of the TLVs and the end of the chunk. This is
1140 * the region we search for address parameters.
1142 init = (struct sctp_init_chunk *)skb->data;
1144 /* Walk the parameters looking for embedded addresses. */
1145 sctp_walk_params(params, init, init_hdr.params) {
1147 /* Note: Ignoring hostname addresses. */
1148 af = sctp_get_af_specific(param_type2af(params.p->type));
1152 if (!af->from_addr_param(paddr, params.addr, sh->source, 0))
1155 asoc = __sctp_lookup_association(net, laddr, paddr, transportp);
1163 /* ADD-IP, Section 5.2
1164 * When an endpoint receives an ASCONF Chunk from the remote peer
1165 * special procedures may be needed to identify the association the
1166 * ASCONF Chunk is associated with. To properly find the association
1167 * the following procedures SHOULD be followed:
1169 * D2) If the association is not found, use the address found in the
1170 * Address Parameter TLV combined with the port number found in the
1171 * SCTP common header. If found proceed to rule D4.
1173 * D2-ext) If more than one ASCONF Chunks are packed together, use the
1174 * address found in the ASCONF Address Parameter TLV of each of the
1175 * subsequent ASCONF Chunks. If found, proceed to rule D4.
1177 static struct sctp_association *__sctp_rcv_asconf_lookup(
1179 struct sctp_chunkhdr *ch,
1180 const union sctp_addr *laddr,
1182 struct sctp_transport **transportp)
1184 struct sctp_addip_chunk *asconf = (struct sctp_addip_chunk *)ch;
1186 union sctp_addr_param *param;
1187 union sctp_addr paddr;
1189 if (ntohs(ch->length) < sizeof(*asconf) + sizeof(struct sctp_paramhdr))
1192 /* Skip over the ADDIP header and find the Address parameter */
1193 param = (union sctp_addr_param *)(asconf + 1);
1195 af = sctp_get_af_specific(param_type2af(param->p.type));
1199 if (!af->from_addr_param(&paddr, param, peer_port, 0))
1202 return __sctp_lookup_association(net, laddr, &paddr, transportp);
1206 /* SCTP-AUTH, Section 6.3:
1207 * If the receiver does not find a STCB for a packet containing an AUTH
1208 * chunk as the first chunk and not a COOKIE-ECHO chunk as the second
1209 * chunk, it MUST use the chunks after the AUTH chunk to look up an existing
1212 * This means that any chunks that can help us identify the association need
1213 * to be looked at to find this association.
1215 static struct sctp_association *__sctp_rcv_walk_lookup(struct net *net,
1216 struct sk_buff *skb,
1217 const union sctp_addr *laddr,
1218 struct sctp_transport **transportp)
1220 struct sctp_association *asoc = NULL;
1221 struct sctp_chunkhdr *ch;
1223 unsigned int chunk_num = 1;
1226 /* Walk through the chunks looking for AUTH or ASCONF chunks
1227 * to help us find the association.
1229 ch = (struct sctp_chunkhdr *)skb->data;
1231 /* Break out if chunk length is less then minimal. */
1232 if (ntohs(ch->length) < sizeof(*ch))
1235 ch_end = ((__u8 *)ch) + SCTP_PAD4(ntohs(ch->length));
1236 if (ch_end > skb_tail_pointer(skb))
1241 have_auth = chunk_num;
1244 case SCTP_CID_COOKIE_ECHO:
1245 /* If a packet arrives containing an AUTH chunk as
1246 * a first chunk, a COOKIE-ECHO chunk as the second
1247 * chunk, and possibly more chunks after them, and
1248 * the receiver does not have an STCB for that
1249 * packet, then authentication is based on
1250 * the contents of the COOKIE- ECHO chunk.
1252 if (have_auth == 1 && chunk_num == 2)
1256 case SCTP_CID_ASCONF:
1257 if (have_auth || net->sctp.addip_noauth)
1258 asoc = __sctp_rcv_asconf_lookup(
1260 sctp_hdr(skb)->source,
1270 ch = (struct sctp_chunkhdr *)ch_end;
1272 } while (ch_end + sizeof(*ch) < skb_tail_pointer(skb));
1278 * There are circumstances when we need to look inside the SCTP packet
1279 * for information to help us find the association. Examples
1280 * include looking inside of INIT/INIT-ACK chunks or after the AUTH
1283 static struct sctp_association *__sctp_rcv_lookup_harder(struct net *net,
1284 struct sk_buff *skb,
1285 const union sctp_addr *laddr,
1286 struct sctp_transport **transportp)
1288 struct sctp_chunkhdr *ch;
1290 /* We do not allow GSO frames here as we need to linearize and
1291 * then cannot guarantee frame boundaries. This shouldn't be an
1292 * issue as packets hitting this are mostly INIT or INIT-ACK and
1293 * those cannot be on GSO-style anyway.
1295 if (skb_is_gso(skb) && skb_is_gso_sctp(skb))
1298 ch = (struct sctp_chunkhdr *)skb->data;
1300 /* The code below will attempt to walk the chunk and extract
1301 * parameter information. Before we do that, we need to verify
1302 * that the chunk length doesn't cause overflow. Otherwise, we'll
1305 if (SCTP_PAD4(ntohs(ch->length)) > skb->len)
1308 /* If this is INIT/INIT-ACK look inside the chunk too. */
1309 if (ch->type == SCTP_CID_INIT || ch->type == SCTP_CID_INIT_ACK)
1310 return __sctp_rcv_init_lookup(net, skb, laddr, transportp);
1312 return __sctp_rcv_walk_lookup(net, skb, laddr, transportp);
1315 /* Lookup an association for an inbound skb. */
1316 static struct sctp_association *__sctp_rcv_lookup(struct net *net,
1317 struct sk_buff *skb,
1318 const union sctp_addr *paddr,
1319 const union sctp_addr *laddr,
1320 struct sctp_transport **transportp)
1322 struct sctp_association *asoc;
1324 asoc = __sctp_lookup_association(net, laddr, paddr, transportp);
1328 /* Further lookup for INIT/INIT-ACK packets.
1329 * SCTP Implementors Guide, 2.18 Handling of address
1330 * parameters within the INIT or INIT-ACK.
1332 asoc = __sctp_rcv_lookup_harder(net, skb, laddr, transportp);
1336 if (paddr->sa.sa_family == AF_INET)
1337 pr_debug("sctp: asoc not found for src:%pI4:%d dst:%pI4:%d\n",
1338 &laddr->v4.sin_addr, ntohs(laddr->v4.sin_port),
1339 &paddr->v4.sin_addr, ntohs(paddr->v4.sin_port));
1341 pr_debug("sctp: asoc not found for src:%pI6:%d dst:%pI6:%d\n",
1342 &laddr->v6.sin6_addr, ntohs(laddr->v6.sin6_port),
1343 &paddr->v6.sin6_addr, ntohs(paddr->v6.sin6_port));