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
18 * Written or modified by:
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;
98 struct net *net = dev_net(skb->dev);
99 bool is_gso = skb_is_gso(skb) && skb_is_gso_sctp(skb);
101 if (skb->pkt_type != PACKET_HOST)
104 __SCTP_INC_STATS(net, SCTP_MIB_INSCTPPACKS);
106 /* If packet is too small to contain a single chunk, let's not
107 * waste time on it anymore.
109 if (skb->len < sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr) +
110 skb_transport_offset(skb))
113 /* If the packet is fragmented and we need to do crc checking,
114 * it's better to just linearize it otherwise crc computing
117 if ((!is_gso && skb_linearize(skb)) ||
118 !pskb_may_pull(skb, sizeof(struct sctphdr)))
121 /* Pull up the IP header. */
122 __skb_pull(skb, skb_transport_offset(skb));
124 skb->csum_valid = 0; /* Previous value not applicable */
125 if (skb_csum_unnecessary(skb))
126 __skb_decr_checksum_unnecessary(skb);
127 else if (!sctp_checksum_disable &&
129 sctp_rcv_checksum(net, skb) < 0)
133 __skb_pull(skb, sizeof(struct sctphdr));
135 family = ipver2af(ip_hdr(skb)->version);
136 af = sctp_get_af_specific(family);
139 SCTP_INPUT_CB(skb)->af = af;
141 /* Initialize local addresses for lookups. */
142 af->from_skb(&src, skb, 1);
143 af->from_skb(&dest, skb, 0);
145 /* If the packet is to or from a non-unicast address,
146 * silently discard the packet.
148 * This is not clearly defined in the RFC except in section
149 * 8.4 - OOTB handling. However, based on the book "Stream Control
150 * Transmission Protocol" 2.1, "It is important to note that the
151 * IP address of an SCTP transport address must be a routable
152 * unicast address. In other words, IP multicast addresses and
153 * IP broadcast addresses cannot be used in an SCTP transport
156 if (!af->addr_valid(&src, NULL, skb) ||
157 !af->addr_valid(&dest, NULL, skb))
160 asoc = __sctp_rcv_lookup(net, skb, &src, &dest, &transport);
163 ep = __sctp_rcv_lookup_endpoint(net, skb, &dest, &src);
165 /* Retrieve the common input handling substructure. */
166 rcvr = asoc ? &asoc->base : &ep->base;
170 * If a frame arrives on an interface and the receiving socket is
171 * bound to another interface, via SO_BINDTODEVICE, treat it as OOTB
173 bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
174 if (bound_dev_if && (bound_dev_if != af->skb_iif(skb))) {
176 sctp_transport_put(transport);
180 sctp_endpoint_put(ep);
183 sk = net->sctp.ctl_sock;
184 ep = sctp_sk(sk)->ep;
185 sctp_endpoint_hold(ep);
190 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
191 * An SCTP packet is called an "out of the blue" (OOTB)
192 * packet if it is correctly formed, i.e., passed the
193 * receiver's checksum check, but the receiver is not
194 * able to identify the association to which this
198 if (sctp_rcv_ootb(skb)) {
199 __SCTP_INC_STATS(net, SCTP_MIB_OUTOFBLUES);
200 goto discard_release;
204 if (!xfrm_policy_check(sk, XFRM_POLICY_IN, skb, family))
205 goto discard_release;
208 if (sk_filter(sk, skb))
209 goto discard_release;
211 /* Create an SCTP packet structure. */
212 chunk = sctp_chunkify(skb, asoc, sk, GFP_ATOMIC);
214 goto discard_release;
215 SCTP_INPUT_CB(skb)->chunk = chunk;
217 /* Remember what endpoint is to handle this packet. */
220 /* Remember the SCTP header. */
221 chunk->sctp_hdr = sctp_hdr(skb);
223 /* Set the source and destination addresses of the incoming chunk. */
224 sctp_init_addrs(chunk, &src, &dest);
226 /* Remember where we came from. */
227 chunk->transport = transport;
229 /* Acquire access to the sock lock. Note: We are safe from other
230 * bottom halves on this lock, but a user may be in the lock too,
231 * so check if it is busy.
235 if (sk != rcvr->sk) {
236 /* Our cached sk is different from the rcvr->sk. This is
237 * because migrate()/accept() may have moved the association
238 * to a new socket and released all the sockets. So now we
239 * are holding a lock on the old socket while the user may
240 * be doing something with the new socket. Switch our veiw
248 if (sock_owned_by_user(sk) || !sctp_newsk_ready(sk)) {
249 if (sctp_add_backlog(sk, skb)) {
251 sctp_chunk_free(chunk);
252 skb = NULL; /* sctp_chunk_free already freed the skb */
253 goto discard_release;
255 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_BACKLOG);
257 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_SOFTIRQ);
258 sctp_inq_push(&chunk->rcvr->inqueue, chunk);
263 /* Release the asoc/ep ref we took in the lookup calls. */
265 sctp_transport_put(transport);
267 sctp_endpoint_put(ep);
272 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_DISCARDS);
277 /* Release the asoc/ep ref we took in the lookup calls. */
279 sctp_transport_put(transport);
281 sctp_endpoint_put(ep);
286 /* Process the backlog queue of the socket. Every skb on
287 * the backlog holds a ref on an association or endpoint.
288 * We hold this ref throughout the state machine to make
289 * sure that the structure we need is still around.
291 int sctp_backlog_rcv(struct sock *sk, struct sk_buff *skb)
293 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
294 struct sctp_inq *inqueue = &chunk->rcvr->inqueue;
295 struct sctp_transport *t = chunk->transport;
296 struct sctp_ep_common *rcvr = NULL;
301 /* If the rcvr is dead then the association or endpoint
302 * has been deleted and we can safely drop the chunk
303 * and refs that we are holding.
306 sctp_chunk_free(chunk);
310 if (unlikely(rcvr->sk != sk)) {
311 /* In this case, the association moved from one socket to
312 * another. We are currently sitting on the backlog of the
313 * old socket, so we need to move.
314 * However, since we are here in the process context we
315 * need to take make sure that the user doesn't own
316 * the new socket when we process the packet.
317 * If the new socket is user-owned, queue the chunk to the
318 * backlog of the new socket without dropping any refs.
319 * Otherwise, we can safely push the chunk on the inqueue.
326 if (sock_owned_by_user(sk) || !sctp_newsk_ready(sk)) {
327 if (sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf)))
328 sctp_chunk_free(chunk);
332 sctp_inq_push(inqueue, chunk);
337 /* If the chunk was backloged again, don't drop refs */
341 if (!sctp_newsk_ready(sk)) {
342 if (!sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf)))
344 sctp_chunk_free(chunk);
346 sctp_inq_push(inqueue, chunk);
351 /* Release the refs we took in sctp_add_backlog */
352 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
353 sctp_transport_put(t);
354 else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
355 sctp_endpoint_put(sctp_ep(rcvr));
362 static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb)
364 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
365 struct sctp_transport *t = chunk->transport;
366 struct sctp_ep_common *rcvr = chunk->rcvr;
369 ret = sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf));
371 /* Hold the assoc/ep while hanging on the backlog queue.
372 * This way, we know structures we need will not disappear
375 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
376 sctp_transport_hold(t);
377 else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
378 sctp_endpoint_hold(sctp_ep(rcvr));
386 /* Handle icmp frag needed error. */
387 void sctp_icmp_frag_needed(struct sock *sk, struct sctp_association *asoc,
388 struct sctp_transport *t, __u32 pmtu)
391 (t->pathmtu <= pmtu &&
392 t->pl.probe_size + sctp_transport_pl_hlen(t) <= pmtu))
395 if (sock_owned_by_user(sk)) {
396 atomic_set(&t->mtu_info, pmtu);
397 asoc->pmtu_pending = 1;
402 if (!(t->param_flags & SPP_PMTUD_ENABLE))
403 /* We can't allow retransmitting in such case, as the
404 * retransmission would be sized just as before, and thus we
405 * would get another icmp, and retransmit again.
409 /* Update transports view of the MTU. Return if no update was needed.
410 * If an update wasn't needed/possible, it also doesn't make sense to
411 * try to retransmit now.
413 if (!sctp_transport_update_pmtu(t, pmtu))
416 /* Update association pmtu. */
417 sctp_assoc_sync_pmtu(asoc);
419 /* Retransmit with the new pmtu setting. */
420 sctp_retransmit(&asoc->outqueue, t, SCTP_RTXR_PMTUD);
423 void sctp_icmp_redirect(struct sock *sk, struct sctp_transport *t,
426 struct dst_entry *dst;
428 if (sock_owned_by_user(sk) || !t)
430 dst = sctp_transport_dst_check(t);
432 dst->ops->redirect(dst, sk, skb);
436 * SCTP Implementer's Guide, 2.37 ICMP handling procedures
438 * ICMP8) If the ICMP code is a "Unrecognized next header type encountered"
439 * or a "Protocol Unreachable" treat this message as an abort
440 * with the T bit set.
442 * This function sends an event to the state machine, which will abort the
446 void sctp_icmp_proto_unreachable(struct sock *sk,
447 struct sctp_association *asoc,
448 struct sctp_transport *t)
450 if (sock_owned_by_user(sk)) {
451 if (timer_pending(&t->proto_unreach_timer))
454 if (!mod_timer(&t->proto_unreach_timer,
456 sctp_transport_hold(t);
459 struct net *net = sock_net(sk);
461 pr_debug("%s: unrecognized next header type "
462 "encountered!\n", __func__);
464 if (del_timer(&t->proto_unreach_timer))
465 sctp_transport_put(t);
467 sctp_do_sm(net, SCTP_EVENT_T_OTHER,
468 SCTP_ST_OTHER(SCTP_EVENT_ICMP_PROTO_UNREACH),
469 asoc->state, asoc->ep, asoc, t,
474 /* Common lookup code for icmp/icmpv6 error handler. */
475 struct sock *sctp_err_lookup(struct net *net, int family, struct sk_buff *skb,
476 struct sctphdr *sctphdr,
477 struct sctp_association **app,
478 struct sctp_transport **tpp)
480 struct sctp_init_chunk *chunkhdr, _chunkhdr;
481 union sctp_addr saddr;
482 union sctp_addr daddr;
484 struct sock *sk = NULL;
485 struct sctp_association *asoc;
486 struct sctp_transport *transport = NULL;
487 __u32 vtag = ntohl(sctphdr->vtag);
489 *app = NULL; *tpp = NULL;
491 af = sctp_get_af_specific(family);
496 /* Initialize local addresses for lookups. */
497 af->from_skb(&saddr, skb, 1);
498 af->from_skb(&daddr, skb, 0);
500 /* Look for an association that matches the incoming ICMP error
503 asoc = __sctp_lookup_association(net, &saddr, &daddr, &transport);
509 /* RFC 4960, Appendix C. ICMP Handling
511 * ICMP6) An implementation MUST validate that the Verification Tag
512 * contained in the ICMP message matches the Verification Tag of
513 * the peer. If the Verification Tag is not 0 and does NOT
514 * match, discard the ICMP message. If it is 0 and the ICMP
515 * message contains enough bytes to verify that the chunk type is
516 * an INIT chunk and that the Initiate Tag matches the tag of the
517 * peer, continue with ICMP7. If the ICMP message is too short
518 * or the chunk type or the Initiate Tag does not match, silently
519 * discard the packet.
522 /* chunk header + first 4 octects of init header */
523 chunkhdr = skb_header_pointer(skb, skb_transport_offset(skb) +
524 sizeof(struct sctphdr),
525 sizeof(struct sctp_chunkhdr) +
526 sizeof(__be32), &_chunkhdr);
528 chunkhdr->chunk_hdr.type != SCTP_CID_INIT ||
529 ntohl(chunkhdr->init_hdr.init_tag) != asoc->c.my_vtag)
532 } else if (vtag != asoc->c.peer_vtag) {
538 /* If too many ICMPs get dropped on busy
539 * servers this needs to be solved differently.
541 if (sock_owned_by_user(sk))
542 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
549 sctp_transport_put(transport);
553 /* Common cleanup code for icmp/icmpv6 error handler. */
554 void sctp_err_finish(struct sock *sk, struct sctp_transport *t)
555 __releases(&((__sk)->sk_lock.slock))
558 sctp_transport_put(t);
561 static void sctp_v4_err_handle(struct sctp_transport *t, struct sk_buff *skb,
562 __u8 type, __u8 code, __u32 info)
564 struct sctp_association *asoc = t->asoc;
565 struct sock *sk = asoc->base.sk;
569 case ICMP_PARAMETERPROB:
572 case ICMP_DEST_UNREACH:
573 if (code > NR_ICMP_UNREACH)
575 if (code == ICMP_FRAG_NEEDED) {
576 sctp_icmp_frag_needed(sk, asoc, t, SCTP_TRUNC4(info));
579 if (code == ICMP_PROT_UNREACH) {
580 sctp_icmp_proto_unreachable(sk, asoc, t);
583 err = icmp_err_convert[code].errno;
585 case ICMP_TIME_EXCEEDED:
586 if (code == ICMP_EXC_FRAGTIME)
592 sctp_icmp_redirect(sk, t, skb);
597 if (!sock_owned_by_user(sk) && inet_sk(sk)->recverr) {
600 } else { /* Only an error on timeout */
601 sk->sk_err_soft = err;
606 * This routine is called by the ICMP module when it gets some
607 * sort of error condition. If err < 0 then the socket should
608 * be closed and the error returned to the user. If err > 0
609 * it's just the icmp type << 8 | icmp code. After adjustment
610 * header points to the first 8 bytes of the sctp header. We need
611 * to find the appropriate port.
613 * The locking strategy used here is very "optimistic". When
614 * someone else accesses the socket the ICMP is just dropped
615 * and for some paths there is no check at all.
616 * A more general error queue to queue errors for later handling
617 * is probably better.
620 int sctp_v4_err(struct sk_buff *skb, __u32 info)
622 const struct iphdr *iph = (const struct iphdr *)skb->data;
623 const int type = icmp_hdr(skb)->type;
624 const int code = icmp_hdr(skb)->code;
625 struct net *net = dev_net(skb->dev);
626 struct sctp_transport *transport;
627 struct sctp_association *asoc;
628 __u16 saveip, savesctp;
631 /* Fix up skb to look at the embedded net header. */
632 saveip = skb->network_header;
633 savesctp = skb->transport_header;
634 skb_reset_network_header(skb);
635 skb_set_transport_header(skb, iph->ihl * 4);
636 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &transport);
637 /* Put back, the original values. */
638 skb->network_header = saveip;
639 skb->transport_header = savesctp;
641 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
645 sctp_v4_err_handle(transport, skb, type, code, info);
646 sctp_err_finish(sk, transport);
651 int sctp_udp_v4_err(struct sock *sk, struct sk_buff *skb)
653 struct net *net = dev_net(skb->dev);
654 struct sctp_association *asoc;
655 struct sctp_transport *t;
659 skb->transport_header += sizeof(struct udphdr);
660 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &t);
662 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
666 skb->transport_header -= sizeof(struct udphdr);
667 hdr = (struct icmphdr *)(skb_network_header(skb) - sizeof(struct icmphdr));
668 if (hdr->type == ICMP_REDIRECT) {
669 /* can't be handled without outer iphdr known, leave it to udp_err */
670 sctp_err_finish(sk, t);
673 if (hdr->type == ICMP_DEST_UNREACH && hdr->code == ICMP_FRAG_NEEDED)
674 info = ntohs(hdr->un.frag.mtu);
675 sctp_v4_err_handle(t, skb, hdr->type, hdr->code, info);
677 sctp_err_finish(sk, t);
682 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
684 * This function scans all the chunks in the OOTB packet to determine if
685 * the packet should be discarded right away. If a response might be needed
686 * for this packet, or, if further processing is possible, the packet will
687 * be queued to a proper inqueue for the next phase of handling.
690 * Return 0 - If further processing is needed.
691 * Return 1 - If the packet can be discarded right away.
693 static int sctp_rcv_ootb(struct sk_buff *skb)
695 struct sctp_chunkhdr *ch, _ch;
696 int ch_end, offset = 0;
698 /* Scan through all the chunks in the packet. */
700 /* Make sure we have at least the header there */
701 if (offset + sizeof(_ch) > skb->len)
704 ch = skb_header_pointer(skb, offset, sizeof(*ch), &_ch);
706 /* Break out if chunk length is less then minimal. */
707 if (!ch || ntohs(ch->length) < sizeof(_ch))
710 ch_end = offset + SCTP_PAD4(ntohs(ch->length));
711 if (ch_end > skb->len)
714 /* RFC 8.4, 2) If the OOTB packet contains an ABORT chunk, the
715 * receiver MUST silently discard the OOTB packet and take no
718 if (SCTP_CID_ABORT == ch->type)
721 /* RFC 8.4, 6) If the packet contains a SHUTDOWN COMPLETE
722 * chunk, the receiver should silently discard the packet
723 * and take no further action.
725 if (SCTP_CID_SHUTDOWN_COMPLETE == ch->type)
729 * This will discard packets with INIT chunk bundled as
730 * subsequent chunks in the packet. When INIT is first,
731 * the normal INIT processing will discard the chunk.
733 if (SCTP_CID_INIT == ch->type && (void *)ch != skb->data)
737 } while (ch_end < skb->len);
745 /* Insert endpoint into the hash table. */
746 static int __sctp_hash_endpoint(struct sctp_endpoint *ep)
748 struct sock *sk = ep->base.sk;
749 struct net *net = sock_net(sk);
750 struct sctp_hashbucket *head;
752 ep->hashent = sctp_ep_hashfn(net, ep->base.bind_addr.port);
753 head = &sctp_ep_hashtable[ep->hashent];
755 if (sk->sk_reuseport) {
756 bool any = sctp_is_ep_boundall(sk);
757 struct sctp_endpoint *ep2;
758 struct list_head *list;
759 int cnt = 0, err = 1;
761 list_for_each(list, &ep->base.bind_addr.address_list)
764 sctp_for_each_hentry(ep2, &head->chain) {
765 struct sock *sk2 = ep2->base.sk;
767 if (!net_eq(sock_net(sk2), net) || sk2 == sk ||
768 !uid_eq(sock_i_uid(sk2), sock_i_uid(sk)) ||
772 err = sctp_bind_addrs_check(sctp_sk(sk2),
775 err = reuseport_add_sock(sk, sk2, any);
779 } else if (err < 0) {
785 err = reuseport_alloc(sk, any);
791 write_lock(&head->lock);
792 hlist_add_head(&ep->node, &head->chain);
793 write_unlock(&head->lock);
797 /* Add an endpoint to the hash. Local BH-safe. */
798 int sctp_hash_endpoint(struct sctp_endpoint *ep)
803 err = __sctp_hash_endpoint(ep);
809 /* Remove endpoint from the hash table. */
810 static void __sctp_unhash_endpoint(struct sctp_endpoint *ep)
812 struct sock *sk = ep->base.sk;
813 struct sctp_hashbucket *head;
815 ep->hashent = sctp_ep_hashfn(sock_net(sk), ep->base.bind_addr.port);
817 head = &sctp_ep_hashtable[ep->hashent];
819 if (rcu_access_pointer(sk->sk_reuseport_cb))
820 reuseport_detach_sock(sk);
822 write_lock(&head->lock);
823 hlist_del_init(&ep->node);
824 write_unlock(&head->lock);
827 /* Remove endpoint from the hash. Local BH-safe. */
828 void sctp_unhash_endpoint(struct sctp_endpoint *ep)
831 __sctp_unhash_endpoint(ep);
835 static inline __u32 sctp_hashfn(const struct net *net, __be16 lport,
836 const union sctp_addr *paddr, __u32 seed)
840 if (paddr->sa.sa_family == AF_INET6)
841 addr = jhash(&paddr->v6.sin6_addr, 16, seed);
843 addr = (__force __u32)paddr->v4.sin_addr.s_addr;
845 return jhash_3words(addr, ((__force __u32)paddr->v4.sin_port) << 16 |
846 (__force __u32)lport, net_hash_mix(net), seed);
849 /* Look up an endpoint. */
850 static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(
851 struct net *net, struct sk_buff *skb,
852 const union sctp_addr *laddr,
853 const union sctp_addr *paddr)
855 struct sctp_hashbucket *head;
856 struct sctp_endpoint *ep;
861 lport = laddr->v4.sin_port;
862 hash = sctp_ep_hashfn(net, ntohs(lport));
863 head = &sctp_ep_hashtable[hash];
864 read_lock(&head->lock);
865 sctp_for_each_hentry(ep, &head->chain) {
866 if (sctp_endpoint_is_match(ep, net, laddr))
870 ep = sctp_sk(net->sctp.ctl_sock)->ep;
874 if (sk->sk_reuseport) {
875 __u32 phash = sctp_hashfn(net, lport, paddr, 0);
877 sk = reuseport_select_sock(sk, phash, skb,
878 sizeof(struct sctphdr));
880 ep = sctp_sk(sk)->ep;
882 sctp_endpoint_hold(ep);
883 read_unlock(&head->lock);
887 /* rhashtable for transport */
888 struct sctp_hash_cmp_arg {
889 const union sctp_addr *paddr;
890 const struct net *net;
894 static inline int sctp_hash_cmp(struct rhashtable_compare_arg *arg,
897 struct sctp_transport *t = (struct sctp_transport *)ptr;
898 const struct sctp_hash_cmp_arg *x = arg->key;
901 if (!sctp_cmp_addr_exact(&t->ipaddr, x->paddr))
903 if (!sctp_transport_hold(t))
906 if (!net_eq(t->asoc->base.net, x->net))
908 if (x->lport != htons(t->asoc->base.bind_addr.port))
913 sctp_transport_put(t);
917 static inline __u32 sctp_hash_obj(const void *data, u32 len, u32 seed)
919 const struct sctp_transport *t = data;
921 return sctp_hashfn(t->asoc->base.net,
922 htons(t->asoc->base.bind_addr.port),
926 static inline __u32 sctp_hash_key(const void *data, u32 len, u32 seed)
928 const struct sctp_hash_cmp_arg *x = data;
930 return sctp_hashfn(x->net, x->lport, x->paddr, seed);
933 static const struct rhashtable_params sctp_hash_params = {
934 .head_offset = offsetof(struct sctp_transport, node),
935 .hashfn = sctp_hash_key,
936 .obj_hashfn = sctp_hash_obj,
937 .obj_cmpfn = sctp_hash_cmp,
938 .automatic_shrinking = true,
941 int sctp_transport_hashtable_init(void)
943 return rhltable_init(&sctp_transport_hashtable, &sctp_hash_params);
946 void sctp_transport_hashtable_destroy(void)
948 rhltable_destroy(&sctp_transport_hashtable);
951 int sctp_hash_transport(struct sctp_transport *t)
953 struct sctp_transport *transport;
954 struct rhlist_head *tmp, *list;
955 struct sctp_hash_cmp_arg arg;
961 arg.net = t->asoc->base.net;
962 arg.paddr = &t->ipaddr;
963 arg.lport = htons(t->asoc->base.bind_addr.port);
966 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
969 rhl_for_each_entry_rcu(transport, tmp, list, node)
970 if (transport->asoc->ep == t->asoc->ep) {
976 err = rhltable_insert_key(&sctp_transport_hashtable, &arg,
977 &t->node, sctp_hash_params);
979 pr_err_once("insert transport fail, errno %d\n", err);
984 void sctp_unhash_transport(struct sctp_transport *t)
989 rhltable_remove(&sctp_transport_hashtable, &t->node,
993 /* return a transport with holding it */
994 struct sctp_transport *sctp_addrs_lookup_transport(
996 const union sctp_addr *laddr,
997 const union sctp_addr *paddr)
999 struct rhlist_head *tmp, *list;
1000 struct sctp_transport *t;
1001 struct sctp_hash_cmp_arg arg = {
1004 .lport = laddr->v4.sin_port,
1007 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
1010 rhl_for_each_entry_rcu(t, tmp, list, node) {
1011 if (!sctp_transport_hold(t))
1014 if (sctp_bind_addr_match(&t->asoc->base.bind_addr,
1015 laddr, sctp_sk(t->asoc->base.sk)))
1017 sctp_transport_put(t);
1023 /* return a transport without holding it, as it's only used under sock lock */
1024 struct sctp_transport *sctp_epaddr_lookup_transport(
1025 const struct sctp_endpoint *ep,
1026 const union sctp_addr *paddr)
1028 struct rhlist_head *tmp, *list;
1029 struct sctp_transport *t;
1030 struct sctp_hash_cmp_arg arg = {
1032 .net = ep->base.net,
1033 .lport = htons(ep->base.bind_addr.port),
1036 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
1039 rhl_for_each_entry_rcu(t, tmp, list, node)
1040 if (ep == t->asoc->ep)
1046 /* Look up an association. */
1047 static struct sctp_association *__sctp_lookup_association(
1049 const union sctp_addr *local,
1050 const union sctp_addr *peer,
1051 struct sctp_transport **pt)
1053 struct sctp_transport *t;
1054 struct sctp_association *asoc = NULL;
1056 t = sctp_addrs_lookup_transport(net, local, peer);
1067 /* Look up an association. protected by RCU read lock */
1069 struct sctp_association *sctp_lookup_association(struct net *net,
1070 const union sctp_addr *laddr,
1071 const union sctp_addr *paddr,
1072 struct sctp_transport **transportp)
1074 struct sctp_association *asoc;
1077 asoc = __sctp_lookup_association(net, laddr, paddr, transportp);
1083 /* Is there an association matching the given local and peer addresses? */
1084 bool sctp_has_association(struct net *net,
1085 const union sctp_addr *laddr,
1086 const union sctp_addr *paddr)
1088 struct sctp_transport *transport;
1090 if (sctp_lookup_association(net, laddr, paddr, &transport)) {
1091 sctp_transport_put(transport);
1099 * SCTP Implementors Guide, 2.18 Handling of address
1100 * parameters within the INIT or INIT-ACK.
1102 * D) When searching for a matching TCB upon reception of an INIT
1103 * or INIT-ACK chunk the receiver SHOULD use not only the
1104 * source address of the packet (containing the INIT or
1105 * INIT-ACK) but the receiver SHOULD also use all valid
1106 * address parameters contained within the chunk.
1108 * 2.18.3 Solution description
1110 * This new text clearly specifies to an implementor the need
1111 * to look within the INIT or INIT-ACK. Any implementation that
1112 * does not do this, may not be able to establish associations
1113 * in certain circumstances.
1116 static struct sctp_association *__sctp_rcv_init_lookup(struct net *net,
1117 struct sk_buff *skb,
1118 const union sctp_addr *laddr, struct sctp_transport **transportp)
1120 struct sctp_association *asoc;
1121 union sctp_addr addr;
1122 union sctp_addr *paddr = &addr;
1123 struct sctphdr *sh = sctp_hdr(skb);
1124 union sctp_params params;
1125 struct sctp_init_chunk *init;
1129 * This code will NOT touch anything inside the chunk--it is
1130 * strictly READ-ONLY.
1132 * RFC 2960 3 SCTP packet Format
1134 * Multiple chunks can be bundled into one SCTP packet up to
1135 * the MTU size, except for the INIT, INIT ACK, and SHUTDOWN
1136 * COMPLETE chunks. These chunks MUST NOT be bundled with any
1137 * other chunk in a packet. See Section 6.10 for more details
1138 * on chunk bundling.
1141 /* Find the start of the TLVs and the end of the chunk. This is
1142 * the region we search for address parameters.
1144 init = (struct sctp_init_chunk *)skb->data;
1146 /* Walk the parameters looking for embedded addresses. */
1147 sctp_walk_params(params, init, init_hdr.params) {
1149 /* Note: Ignoring hostname addresses. */
1150 af = sctp_get_af_specific(param_type2af(params.p->type));
1154 if (!af->from_addr_param(paddr, params.addr, sh->source, 0))
1157 asoc = __sctp_lookup_association(net, laddr, paddr, transportp);
1165 /* ADD-IP, Section 5.2
1166 * When an endpoint receives an ASCONF Chunk from the remote peer
1167 * special procedures may be needed to identify the association the
1168 * ASCONF Chunk is associated with. To properly find the association
1169 * the following procedures SHOULD be followed:
1171 * D2) If the association is not found, use the address found in the
1172 * Address Parameter TLV combined with the port number found in the
1173 * SCTP common header. If found proceed to rule D4.
1175 * D2-ext) If more than one ASCONF Chunks are packed together, use the
1176 * address found in the ASCONF Address Parameter TLV of each of the
1177 * subsequent ASCONF Chunks. If found, proceed to rule D4.
1179 static struct sctp_association *__sctp_rcv_asconf_lookup(
1181 struct sctp_chunkhdr *ch,
1182 const union sctp_addr *laddr,
1184 struct sctp_transport **transportp)
1186 struct sctp_addip_chunk *asconf = (struct sctp_addip_chunk *)ch;
1188 union sctp_addr_param *param;
1189 union sctp_addr paddr;
1191 if (ntohs(ch->length) < sizeof(*asconf) + sizeof(struct sctp_paramhdr))
1194 /* Skip over the ADDIP header and find the Address parameter */
1195 param = (union sctp_addr_param *)(asconf + 1);
1197 af = sctp_get_af_specific(param_type2af(param->p.type));
1201 if (!af->from_addr_param(&paddr, param, peer_port, 0))
1204 return __sctp_lookup_association(net, laddr, &paddr, transportp);
1208 /* SCTP-AUTH, Section 6.3:
1209 * If the receiver does not find a STCB for a packet containing an AUTH
1210 * chunk as the first chunk and not a COOKIE-ECHO chunk as the second
1211 * chunk, it MUST use the chunks after the AUTH chunk to look up an existing
1214 * This means that any chunks that can help us identify the association need
1215 * to be looked at to find this association.
1217 static struct sctp_association *__sctp_rcv_walk_lookup(struct net *net,
1218 struct sk_buff *skb,
1219 const union sctp_addr *laddr,
1220 struct sctp_transport **transportp)
1222 struct sctp_association *asoc = NULL;
1223 struct sctp_chunkhdr *ch;
1225 unsigned int chunk_num = 1;
1228 /* Walk through the chunks looking for AUTH or ASCONF chunks
1229 * to help us find the association.
1231 ch = (struct sctp_chunkhdr *)skb->data;
1233 /* Break out if chunk length is less then minimal. */
1234 if (ntohs(ch->length) < sizeof(*ch))
1237 ch_end = ((__u8 *)ch) + SCTP_PAD4(ntohs(ch->length));
1238 if (ch_end > skb_tail_pointer(skb))
1243 have_auth = chunk_num;
1246 case SCTP_CID_COOKIE_ECHO:
1247 /* If a packet arrives containing an AUTH chunk as
1248 * a first chunk, a COOKIE-ECHO chunk as the second
1249 * chunk, and possibly more chunks after them, and
1250 * the receiver does not have an STCB for that
1251 * packet, then authentication is based on
1252 * the contents of the COOKIE- ECHO chunk.
1254 if (have_auth == 1 && chunk_num == 2)
1258 case SCTP_CID_ASCONF:
1259 if (have_auth || net->sctp.addip_noauth)
1260 asoc = __sctp_rcv_asconf_lookup(
1262 sctp_hdr(skb)->source,
1272 ch = (struct sctp_chunkhdr *)ch_end;
1274 } while (ch_end + sizeof(*ch) < skb_tail_pointer(skb));
1280 * There are circumstances when we need to look inside the SCTP packet
1281 * for information to help us find the association. Examples
1282 * include looking inside of INIT/INIT-ACK chunks or after the AUTH
1285 static struct sctp_association *__sctp_rcv_lookup_harder(struct net *net,
1286 struct sk_buff *skb,
1287 const union sctp_addr *laddr,
1288 struct sctp_transport **transportp)
1290 struct sctp_chunkhdr *ch;
1292 /* We do not allow GSO frames here as we need to linearize and
1293 * then cannot guarantee frame boundaries. This shouldn't be an
1294 * issue as packets hitting this are mostly INIT or INIT-ACK and
1295 * those cannot be on GSO-style anyway.
1297 if (skb_is_gso(skb) && skb_is_gso_sctp(skb))
1300 ch = (struct sctp_chunkhdr *)skb->data;
1302 /* The code below will attempt to walk the chunk and extract
1303 * parameter information. Before we do that, we need to verify
1304 * that the chunk length doesn't cause overflow. Otherwise, we'll
1307 if (SCTP_PAD4(ntohs(ch->length)) > skb->len)
1310 /* If this is INIT/INIT-ACK look inside the chunk too. */
1311 if (ch->type == SCTP_CID_INIT || ch->type == SCTP_CID_INIT_ACK)
1312 return __sctp_rcv_init_lookup(net, skb, laddr, transportp);
1314 return __sctp_rcv_walk_lookup(net, skb, laddr, transportp);
1317 /* Lookup an association for an inbound skb. */
1318 static struct sctp_association *__sctp_rcv_lookup(struct net *net,
1319 struct sk_buff *skb,
1320 const union sctp_addr *paddr,
1321 const union sctp_addr *laddr,
1322 struct sctp_transport **transportp)
1324 struct sctp_association *asoc;
1326 asoc = __sctp_lookup_association(net, laddr, paddr, transportp);
1330 /* Further lookup for INIT/INIT-ACK packets.
1331 * SCTP Implementors Guide, 2.18 Handling of address
1332 * parameters within the INIT or INIT-ACK.
1334 asoc = __sctp_rcv_lookup_harder(net, skb, laddr, transportp);
1338 if (paddr->sa.sa_family == AF_INET)
1339 pr_debug("sctp: asoc not found for src:%pI4:%d dst:%pI4:%d\n",
1340 &laddr->v4.sin_addr, ntohs(laddr->v4.sin_port),
1341 &paddr->v4.sin_addr, ntohs(paddr->v4.sin_port));
1343 pr_debug("sctp: asoc not found for src:%pI6:%d dst:%pI6:%d\n",
1344 &laddr->v6.sin6_addr, ntohs(laddr->v6.sin6_port),
1345 &paddr->v6.sin6_addr, ntohs(paddr->v6.sin6_port));