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,
55 static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(
56 struct net *net, struct sk_buff *skb,
57 const union sctp_addr *laddr,
58 const union sctp_addr *daddr,
60 static struct sctp_association *__sctp_lookup_association(
62 const union sctp_addr *local,
63 const union sctp_addr *peer,
64 struct sctp_transport **pt,
67 static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb);
70 /* Calculate the SCTP checksum of an SCTP packet. */
71 static inline int sctp_rcv_checksum(struct net *net, struct sk_buff *skb)
73 struct sctphdr *sh = sctp_hdr(skb);
74 __le32 cmp = sh->checksum;
75 __le32 val = sctp_compute_cksum(skb, 0);
78 /* CRC failure, dump it. */
79 __SCTP_INC_STATS(net, SCTP_MIB_CHECKSUMERRORS);
86 * This is the routine which IP calls when receiving an SCTP packet.
88 int sctp_rcv(struct sk_buff *skb)
91 struct sctp_association *asoc;
92 struct sctp_endpoint *ep = NULL;
93 struct sctp_ep_common *rcvr;
94 struct sctp_transport *transport = NULL;
95 struct sctp_chunk *chunk;
100 struct net *net = dev_net(skb->dev);
101 bool is_gso = skb_is_gso(skb) && skb_is_gso_sctp(skb);
104 if (skb->pkt_type != PACKET_HOST)
107 __SCTP_INC_STATS(net, SCTP_MIB_INSCTPPACKS);
109 /* If packet is too small to contain a single chunk, let's not
110 * waste time on it anymore.
112 if (skb->len < sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr) +
113 skb_transport_offset(skb))
116 /* If the packet is fragmented and we need to do crc checking,
117 * it's better to just linearize it otherwise crc computing
120 if ((!is_gso && skb_linearize(skb)) ||
121 !pskb_may_pull(skb, sizeof(struct sctphdr)))
124 /* Pull up the IP header. */
125 __skb_pull(skb, skb_transport_offset(skb));
127 skb->csum_valid = 0; /* Previous value not applicable */
128 if (skb_csum_unnecessary(skb))
129 __skb_decr_checksum_unnecessary(skb);
130 else if (!sctp_checksum_disable &&
132 sctp_rcv_checksum(net, skb) < 0)
136 __skb_pull(skb, sizeof(struct sctphdr));
138 family = ipver2af(ip_hdr(skb)->version);
139 af = sctp_get_af_specific(family);
142 SCTP_INPUT_CB(skb)->af = af;
144 /* Initialize local addresses for lookups. */
145 af->from_skb(&src, skb, 1);
146 af->from_skb(&dest, skb, 0);
147 dif = af->skb_iif(skb);
148 sdif = af->skb_sdif(skb);
150 /* If the packet is to or from a non-unicast address,
151 * silently discard the packet.
153 * This is not clearly defined in the RFC except in section
154 * 8.4 - OOTB handling. However, based on the book "Stream Control
155 * Transmission Protocol" 2.1, "It is important to note that the
156 * IP address of an SCTP transport address must be a routable
157 * unicast address. In other words, IP multicast addresses and
158 * IP broadcast addresses cannot be used in an SCTP transport
161 if (!af->addr_valid(&src, NULL, skb) ||
162 !af->addr_valid(&dest, NULL, skb))
165 asoc = __sctp_rcv_lookup(net, skb, &src, &dest, &transport, dif, sdif);
168 ep = __sctp_rcv_lookup_endpoint(net, skb, &dest, &src, dif, sdif);
170 /* Retrieve the common input handling substructure. */
171 rcvr = asoc ? &asoc->base : &ep->base;
175 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
176 * An SCTP packet is called an "out of the blue" (OOTB)
177 * packet if it is correctly formed, i.e., passed the
178 * receiver's checksum check, but the receiver is not
179 * able to identify the association to which this
183 if (sctp_rcv_ootb(skb)) {
184 __SCTP_INC_STATS(net, SCTP_MIB_OUTOFBLUES);
185 goto discard_release;
189 if (!xfrm_policy_check(sk, XFRM_POLICY_IN, skb, family))
190 goto discard_release;
193 if (sk_filter(sk, skb))
194 goto discard_release;
196 /* Create an SCTP packet structure. */
197 chunk = sctp_chunkify(skb, asoc, sk, GFP_ATOMIC);
199 goto discard_release;
200 SCTP_INPUT_CB(skb)->chunk = chunk;
202 /* Remember what endpoint is to handle this packet. */
205 /* Remember the SCTP header. */
206 chunk->sctp_hdr = sctp_hdr(skb);
208 /* Set the source and destination addresses of the incoming chunk. */
209 sctp_init_addrs(chunk, &src, &dest);
211 /* Remember where we came from. */
212 chunk->transport = transport;
214 /* Acquire access to the sock lock. Note: We are safe from other
215 * bottom halves on this lock, but a user may be in the lock too,
216 * so check if it is busy.
220 if (sk != rcvr->sk) {
221 /* Our cached sk is different from the rcvr->sk. This is
222 * because migrate()/accept() may have moved the association
223 * to a new socket and released all the sockets. So now we
224 * are holding a lock on the old socket while the user may
225 * be doing something with the new socket. Switch our veiw
233 if (sock_owned_by_user(sk) || !sctp_newsk_ready(sk)) {
234 if (sctp_add_backlog(sk, skb)) {
236 sctp_chunk_free(chunk);
237 skb = NULL; /* sctp_chunk_free already freed the skb */
238 goto discard_release;
240 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_BACKLOG);
242 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_SOFTIRQ);
243 sctp_inq_push(&chunk->rcvr->inqueue, chunk);
248 /* Release the asoc/ep ref we took in the lookup calls. */
250 sctp_transport_put(transport);
252 sctp_endpoint_put(ep);
257 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_DISCARDS);
262 /* Release the asoc/ep ref we took in the lookup calls. */
264 sctp_transport_put(transport);
266 sctp_endpoint_put(ep);
271 /* Process the backlog queue of the socket. Every skb on
272 * the backlog holds a ref on an association or endpoint.
273 * We hold this ref throughout the state machine to make
274 * sure that the structure we need is still around.
276 int sctp_backlog_rcv(struct sock *sk, struct sk_buff *skb)
278 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
279 struct sctp_inq *inqueue = &chunk->rcvr->inqueue;
280 struct sctp_transport *t = chunk->transport;
281 struct sctp_ep_common *rcvr = NULL;
286 /* If the rcvr is dead then the association or endpoint
287 * has been deleted and we can safely drop the chunk
288 * and refs that we are holding.
291 sctp_chunk_free(chunk);
295 if (unlikely(rcvr->sk != sk)) {
296 /* In this case, the association moved from one socket to
297 * another. We are currently sitting on the backlog of the
298 * old socket, so we need to move.
299 * However, since we are here in the process context we
300 * need to take make sure that the user doesn't own
301 * the new socket when we process the packet.
302 * If the new socket is user-owned, queue the chunk to the
303 * backlog of the new socket without dropping any refs.
304 * Otherwise, we can safely push the chunk on the inqueue.
311 if (sock_owned_by_user(sk) || !sctp_newsk_ready(sk)) {
312 if (sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf)))
313 sctp_chunk_free(chunk);
317 sctp_inq_push(inqueue, chunk);
322 /* If the chunk was backloged again, don't drop refs */
326 if (!sctp_newsk_ready(sk)) {
327 if (!sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf)))
329 sctp_chunk_free(chunk);
331 sctp_inq_push(inqueue, chunk);
336 /* Release the refs we took in sctp_add_backlog */
337 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
338 sctp_transport_put(t);
339 else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
340 sctp_endpoint_put(sctp_ep(rcvr));
347 static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb)
349 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
350 struct sctp_transport *t = chunk->transport;
351 struct sctp_ep_common *rcvr = chunk->rcvr;
354 ret = sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf));
356 /* Hold the assoc/ep while hanging on the backlog queue.
357 * This way, we know structures we need will not disappear
360 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
361 sctp_transport_hold(t);
362 else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
363 sctp_endpoint_hold(sctp_ep(rcvr));
371 /* Handle icmp frag needed error. */
372 void sctp_icmp_frag_needed(struct sock *sk, struct sctp_association *asoc,
373 struct sctp_transport *t, __u32 pmtu)
376 (t->pathmtu <= pmtu &&
377 t->pl.probe_size + sctp_transport_pl_hlen(t) <= pmtu))
380 if (sock_owned_by_user(sk)) {
381 atomic_set(&t->mtu_info, pmtu);
382 asoc->pmtu_pending = 1;
387 if (!(t->param_flags & SPP_PMTUD_ENABLE))
388 /* We can't allow retransmitting in such case, as the
389 * retransmission would be sized just as before, and thus we
390 * would get another icmp, and retransmit again.
394 /* Update transports view of the MTU. Return if no update was needed.
395 * If an update wasn't needed/possible, it also doesn't make sense to
396 * try to retransmit now.
398 if (!sctp_transport_update_pmtu(t, pmtu))
401 /* Update association pmtu. */
402 sctp_assoc_sync_pmtu(asoc);
404 /* Retransmit with the new pmtu setting. */
405 sctp_retransmit(&asoc->outqueue, t, SCTP_RTXR_PMTUD);
408 void sctp_icmp_redirect(struct sock *sk, struct sctp_transport *t,
411 struct dst_entry *dst;
413 if (sock_owned_by_user(sk) || !t)
415 dst = sctp_transport_dst_check(t);
417 dst->ops->redirect(dst, sk, skb);
421 * SCTP Implementer's Guide, 2.37 ICMP handling procedures
423 * ICMP8) If the ICMP code is a "Unrecognized next header type encountered"
424 * or a "Protocol Unreachable" treat this message as an abort
425 * with the T bit set.
427 * This function sends an event to the state machine, which will abort the
431 void sctp_icmp_proto_unreachable(struct sock *sk,
432 struct sctp_association *asoc,
433 struct sctp_transport *t)
435 if (sock_owned_by_user(sk)) {
436 if (timer_pending(&t->proto_unreach_timer))
439 if (!mod_timer(&t->proto_unreach_timer,
441 sctp_transport_hold(t);
444 struct net *net = sock_net(sk);
446 pr_debug("%s: unrecognized next header type "
447 "encountered!\n", __func__);
449 if (del_timer(&t->proto_unreach_timer))
450 sctp_transport_put(t);
452 sctp_do_sm(net, SCTP_EVENT_T_OTHER,
453 SCTP_ST_OTHER(SCTP_EVENT_ICMP_PROTO_UNREACH),
454 asoc->state, asoc->ep, asoc, t,
459 /* Common lookup code for icmp/icmpv6 error handler. */
460 struct sock *sctp_err_lookup(struct net *net, int family, struct sk_buff *skb,
461 struct sctphdr *sctphdr,
462 struct sctp_association **app,
463 struct sctp_transport **tpp)
465 struct sctp_init_chunk *chunkhdr, _chunkhdr;
466 union sctp_addr saddr;
467 union sctp_addr daddr;
469 struct sock *sk = NULL;
470 struct sctp_association *asoc;
471 struct sctp_transport *transport = NULL;
472 __u32 vtag = ntohl(sctphdr->vtag);
473 int sdif = inet_sdif(skb);
474 int dif = inet_iif(skb);
476 *app = NULL; *tpp = NULL;
478 af = sctp_get_af_specific(family);
483 /* Initialize local addresses for lookups. */
484 af->from_skb(&saddr, skb, 1);
485 af->from_skb(&daddr, skb, 0);
487 /* Look for an association that matches the incoming ICMP error
490 asoc = __sctp_lookup_association(net, &saddr, &daddr, &transport, dif, sdif);
496 /* RFC 4960, Appendix C. ICMP Handling
498 * ICMP6) An implementation MUST validate that the Verification Tag
499 * contained in the ICMP message matches the Verification Tag of
500 * the peer. If the Verification Tag is not 0 and does NOT
501 * match, discard the ICMP message. If it is 0 and the ICMP
502 * message contains enough bytes to verify that the chunk type is
503 * an INIT chunk and that the Initiate Tag matches the tag of the
504 * peer, continue with ICMP7. If the ICMP message is too short
505 * or the chunk type or the Initiate Tag does not match, silently
506 * discard the packet.
509 /* chunk header + first 4 octects of init header */
510 chunkhdr = skb_header_pointer(skb, skb_transport_offset(skb) +
511 sizeof(struct sctphdr),
512 sizeof(struct sctp_chunkhdr) +
513 sizeof(__be32), &_chunkhdr);
515 chunkhdr->chunk_hdr.type != SCTP_CID_INIT ||
516 ntohl(chunkhdr->init_hdr.init_tag) != asoc->c.my_vtag)
519 } else if (vtag != asoc->c.peer_vtag) {
525 /* If too many ICMPs get dropped on busy
526 * servers this needs to be solved differently.
528 if (sock_owned_by_user(sk))
529 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
536 sctp_transport_put(transport);
540 /* Common cleanup code for icmp/icmpv6 error handler. */
541 void sctp_err_finish(struct sock *sk, struct sctp_transport *t)
542 __releases(&((__sk)->sk_lock.slock))
545 sctp_transport_put(t);
548 static void sctp_v4_err_handle(struct sctp_transport *t, struct sk_buff *skb,
549 __u8 type, __u8 code, __u32 info)
551 struct sctp_association *asoc = t->asoc;
552 struct sock *sk = asoc->base.sk;
556 case ICMP_PARAMETERPROB:
559 case ICMP_DEST_UNREACH:
560 if (code > NR_ICMP_UNREACH)
562 if (code == ICMP_FRAG_NEEDED) {
563 sctp_icmp_frag_needed(sk, asoc, t, SCTP_TRUNC4(info));
566 if (code == ICMP_PROT_UNREACH) {
567 sctp_icmp_proto_unreachable(sk, asoc, t);
570 err = icmp_err_convert[code].errno;
572 case ICMP_TIME_EXCEEDED:
573 if (code == ICMP_EXC_FRAGTIME)
579 sctp_icmp_redirect(sk, t, skb);
584 if (!sock_owned_by_user(sk) && inet_test_bit(RECVERR, sk)) {
587 } else { /* Only an error on timeout */
588 WRITE_ONCE(sk->sk_err_soft, err);
593 * This routine is called by the ICMP module when it gets some
594 * sort of error condition. If err < 0 then the socket should
595 * be closed and the error returned to the user. If err > 0
596 * it's just the icmp type << 8 | icmp code. After adjustment
597 * header points to the first 8 bytes of the sctp header. We need
598 * to find the appropriate port.
600 * The locking strategy used here is very "optimistic". When
601 * someone else accesses the socket the ICMP is just dropped
602 * and for some paths there is no check at all.
603 * A more general error queue to queue errors for later handling
604 * is probably better.
607 int sctp_v4_err(struct sk_buff *skb, __u32 info)
609 const struct iphdr *iph = (const struct iphdr *)skb->data;
610 const int type = icmp_hdr(skb)->type;
611 const int code = icmp_hdr(skb)->code;
612 struct net *net = dev_net(skb->dev);
613 struct sctp_transport *transport;
614 struct sctp_association *asoc;
615 __u16 saveip, savesctp;
618 /* Fix up skb to look at the embedded net header. */
619 saveip = skb->network_header;
620 savesctp = skb->transport_header;
621 skb_reset_network_header(skb);
622 skb_set_transport_header(skb, iph->ihl * 4);
623 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &transport);
624 /* Put back, the original values. */
625 skb->network_header = saveip;
626 skb->transport_header = savesctp;
628 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
632 sctp_v4_err_handle(transport, skb, type, code, info);
633 sctp_err_finish(sk, transport);
638 int sctp_udp_v4_err(struct sock *sk, struct sk_buff *skb)
640 struct net *net = dev_net(skb->dev);
641 struct sctp_association *asoc;
642 struct sctp_transport *t;
646 skb->transport_header += sizeof(struct udphdr);
647 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &t);
649 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
653 skb->transport_header -= sizeof(struct udphdr);
654 hdr = (struct icmphdr *)(skb_network_header(skb) - sizeof(struct icmphdr));
655 if (hdr->type == ICMP_REDIRECT) {
656 /* can't be handled without outer iphdr known, leave it to udp_err */
657 sctp_err_finish(sk, t);
660 if (hdr->type == ICMP_DEST_UNREACH && hdr->code == ICMP_FRAG_NEEDED)
661 info = ntohs(hdr->un.frag.mtu);
662 sctp_v4_err_handle(t, skb, hdr->type, hdr->code, info);
664 sctp_err_finish(sk, t);
669 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
671 * This function scans all the chunks in the OOTB packet to determine if
672 * the packet should be discarded right away. If a response might be needed
673 * for this packet, or, if further processing is possible, the packet will
674 * be queued to a proper inqueue for the next phase of handling.
677 * Return 0 - If further processing is needed.
678 * Return 1 - If the packet can be discarded right away.
680 static int sctp_rcv_ootb(struct sk_buff *skb)
682 struct sctp_chunkhdr *ch, _ch;
683 int ch_end, offset = 0;
685 /* Scan through all the chunks in the packet. */
687 /* Make sure we have at least the header there */
688 if (offset + sizeof(_ch) > skb->len)
691 ch = skb_header_pointer(skb, offset, sizeof(*ch), &_ch);
693 /* Break out if chunk length is less then minimal. */
694 if (!ch || ntohs(ch->length) < sizeof(_ch))
697 ch_end = offset + SCTP_PAD4(ntohs(ch->length));
698 if (ch_end > skb->len)
701 /* RFC 8.4, 2) If the OOTB packet contains an ABORT chunk, the
702 * receiver MUST silently discard the OOTB packet and take no
705 if (SCTP_CID_ABORT == ch->type)
708 /* RFC 8.4, 6) If the packet contains a SHUTDOWN COMPLETE
709 * chunk, the receiver should silently discard the packet
710 * and take no further action.
712 if (SCTP_CID_SHUTDOWN_COMPLETE == ch->type)
716 * This will discard packets with INIT chunk bundled as
717 * subsequent chunks in the packet. When INIT is first,
718 * the normal INIT processing will discard the chunk.
720 if (SCTP_CID_INIT == ch->type && (void *)ch != skb->data)
724 } while (ch_end < skb->len);
732 /* Insert endpoint into the hash table. */
733 static int __sctp_hash_endpoint(struct sctp_endpoint *ep)
735 struct sock *sk = ep->base.sk;
736 struct net *net = sock_net(sk);
737 struct sctp_hashbucket *head;
739 ep->hashent = sctp_ep_hashfn(net, ep->base.bind_addr.port);
740 head = &sctp_ep_hashtable[ep->hashent];
742 if (sk->sk_reuseport) {
743 bool any = sctp_is_ep_boundall(sk);
744 struct sctp_endpoint *ep2;
745 struct list_head *list;
746 int cnt = 0, err = 1;
748 list_for_each(list, &ep->base.bind_addr.address_list)
751 sctp_for_each_hentry(ep2, &head->chain) {
752 struct sock *sk2 = ep2->base.sk;
754 if (!net_eq(sock_net(sk2), net) || sk2 == sk ||
755 !uid_eq(sock_i_uid(sk2), sock_i_uid(sk)) ||
759 err = sctp_bind_addrs_check(sctp_sk(sk2),
762 err = reuseport_add_sock(sk, sk2, any);
766 } else if (err < 0) {
772 err = reuseport_alloc(sk, any);
778 write_lock(&head->lock);
779 hlist_add_head(&ep->node, &head->chain);
780 write_unlock(&head->lock);
784 /* Add an endpoint to the hash. Local BH-safe. */
785 int sctp_hash_endpoint(struct sctp_endpoint *ep)
790 err = __sctp_hash_endpoint(ep);
796 /* Remove endpoint from the hash table. */
797 static void __sctp_unhash_endpoint(struct sctp_endpoint *ep)
799 struct sock *sk = ep->base.sk;
800 struct sctp_hashbucket *head;
802 ep->hashent = sctp_ep_hashfn(sock_net(sk), ep->base.bind_addr.port);
804 head = &sctp_ep_hashtable[ep->hashent];
806 if (rcu_access_pointer(sk->sk_reuseport_cb))
807 reuseport_detach_sock(sk);
809 write_lock(&head->lock);
810 hlist_del_init(&ep->node);
811 write_unlock(&head->lock);
814 /* Remove endpoint from the hash. Local BH-safe. */
815 void sctp_unhash_endpoint(struct sctp_endpoint *ep)
818 __sctp_unhash_endpoint(ep);
822 static inline __u32 sctp_hashfn(const struct net *net, __be16 lport,
823 const union sctp_addr *paddr, __u32 seed)
827 if (paddr->sa.sa_family == AF_INET6)
828 addr = jhash(&paddr->v6.sin6_addr, 16, seed);
830 addr = (__force __u32)paddr->v4.sin_addr.s_addr;
832 return jhash_3words(addr, ((__force __u32)paddr->v4.sin_port) << 16 |
833 (__force __u32)lport, net_hash_mix(net), seed);
836 /* Look up an endpoint. */
837 static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(
838 struct net *net, struct sk_buff *skb,
839 const union sctp_addr *laddr,
840 const union sctp_addr *paddr,
843 struct sctp_hashbucket *head;
844 struct sctp_endpoint *ep;
849 lport = laddr->v4.sin_port;
850 hash = sctp_ep_hashfn(net, ntohs(lport));
851 head = &sctp_ep_hashtable[hash];
852 read_lock(&head->lock);
853 sctp_for_each_hentry(ep, &head->chain) {
854 if (sctp_endpoint_is_match(ep, net, laddr, dif, sdif))
858 ep = sctp_sk(net->sctp.ctl_sock)->ep;
862 if (sk->sk_reuseport) {
863 __u32 phash = sctp_hashfn(net, lport, paddr, 0);
865 sk = reuseport_select_sock(sk, phash, skb,
866 sizeof(struct sctphdr));
868 ep = sctp_sk(sk)->ep;
870 sctp_endpoint_hold(ep);
871 read_unlock(&head->lock);
875 /* rhashtable for transport */
876 struct sctp_hash_cmp_arg {
877 const union sctp_addr *paddr;
878 const struct net *net;
882 static inline int sctp_hash_cmp(struct rhashtable_compare_arg *arg,
885 struct sctp_transport *t = (struct sctp_transport *)ptr;
886 const struct sctp_hash_cmp_arg *x = arg->key;
889 if (!sctp_cmp_addr_exact(&t->ipaddr, x->paddr))
891 if (!sctp_transport_hold(t))
894 if (!net_eq(t->asoc->base.net, x->net))
896 if (x->lport != htons(t->asoc->base.bind_addr.port))
901 sctp_transport_put(t);
905 static inline __u32 sctp_hash_obj(const void *data, u32 len, u32 seed)
907 const struct sctp_transport *t = data;
909 return sctp_hashfn(t->asoc->base.net,
910 htons(t->asoc->base.bind_addr.port),
914 static inline __u32 sctp_hash_key(const void *data, u32 len, u32 seed)
916 const struct sctp_hash_cmp_arg *x = data;
918 return sctp_hashfn(x->net, x->lport, x->paddr, seed);
921 static const struct rhashtable_params sctp_hash_params = {
922 .head_offset = offsetof(struct sctp_transport, node),
923 .hashfn = sctp_hash_key,
924 .obj_hashfn = sctp_hash_obj,
925 .obj_cmpfn = sctp_hash_cmp,
926 .automatic_shrinking = true,
929 int sctp_transport_hashtable_init(void)
931 return rhltable_init(&sctp_transport_hashtable, &sctp_hash_params);
934 void sctp_transport_hashtable_destroy(void)
936 rhltable_destroy(&sctp_transport_hashtable);
939 int sctp_hash_transport(struct sctp_transport *t)
941 struct sctp_transport *transport;
942 struct rhlist_head *tmp, *list;
943 struct sctp_hash_cmp_arg arg;
949 arg.net = t->asoc->base.net;
950 arg.paddr = &t->ipaddr;
951 arg.lport = htons(t->asoc->base.bind_addr.port);
954 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
957 rhl_for_each_entry_rcu(transport, tmp, list, node)
958 if (transport->asoc->ep == t->asoc->ep) {
964 err = rhltable_insert_key(&sctp_transport_hashtable, &arg,
965 &t->node, sctp_hash_params);
967 pr_err_once("insert transport fail, errno %d\n", err);
972 void sctp_unhash_transport(struct sctp_transport *t)
977 rhltable_remove(&sctp_transport_hashtable, &t->node,
981 bool sctp_sk_bound_dev_eq(struct net *net, int bound_dev_if, int dif, int sdif)
983 bool l3mdev_accept = true;
985 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
986 l3mdev_accept = !!READ_ONCE(net->sctp.l3mdev_accept);
988 return inet_bound_dev_eq(l3mdev_accept, bound_dev_if, dif, sdif);
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,
998 struct rhlist_head *tmp, *list;
999 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 bound_dev_if = READ_ONCE(t->asoc->base.sk->sk_bound_dev_if);
1015 if (sctp_sk_bound_dev_eq(net, bound_dev_if, dif, sdif) &&
1016 sctp_bind_addr_match(&t->asoc->base.bind_addr,
1017 laddr, sctp_sk(t->asoc->base.sk)))
1019 sctp_transport_put(t);
1025 /* return a transport without holding it, as it's only used under sock lock */
1026 struct sctp_transport *sctp_epaddr_lookup_transport(
1027 const struct sctp_endpoint *ep,
1028 const union sctp_addr *paddr)
1030 struct rhlist_head *tmp, *list;
1031 struct sctp_transport *t;
1032 struct sctp_hash_cmp_arg arg = {
1034 .net = ep->base.net,
1035 .lport = htons(ep->base.bind_addr.port),
1038 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
1041 rhl_for_each_entry_rcu(t, tmp, list, node)
1042 if (ep == t->asoc->ep)
1048 /* Look up an association. */
1049 static struct sctp_association *__sctp_lookup_association(
1051 const union sctp_addr *local,
1052 const union sctp_addr *peer,
1053 struct sctp_transport **pt,
1056 struct sctp_transport *t;
1057 struct sctp_association *asoc = NULL;
1059 t = sctp_addrs_lookup_transport(net, local, peer, dif, sdif);
1070 /* Look up an association. protected by RCU read lock */
1072 struct sctp_association *sctp_lookup_association(struct net *net,
1073 const union sctp_addr *laddr,
1074 const union sctp_addr *paddr,
1075 struct sctp_transport **transportp,
1078 struct sctp_association *asoc;
1081 asoc = __sctp_lookup_association(net, laddr, paddr, transportp, dif, sdif);
1087 /* Is there an association matching the given local and peer addresses? */
1088 bool sctp_has_association(struct net *net,
1089 const union sctp_addr *laddr,
1090 const union sctp_addr *paddr,
1093 struct sctp_transport *transport;
1095 if (sctp_lookup_association(net, laddr, paddr, &transport, dif, sdif)) {
1096 sctp_transport_put(transport);
1104 * SCTP Implementors Guide, 2.18 Handling of address
1105 * parameters within the INIT or INIT-ACK.
1107 * D) When searching for a matching TCB upon reception of an INIT
1108 * or INIT-ACK chunk the receiver SHOULD use not only the
1109 * source address of the packet (containing the INIT or
1110 * INIT-ACK) but the receiver SHOULD also use all valid
1111 * address parameters contained within the chunk.
1113 * 2.18.3 Solution description
1115 * This new text clearly specifies to an implementor the need
1116 * to look within the INIT or INIT-ACK. Any implementation that
1117 * does not do this, may not be able to establish associations
1118 * in certain circumstances.
1121 static struct sctp_association *__sctp_rcv_init_lookup(struct net *net,
1122 struct sk_buff *skb,
1123 const union sctp_addr *laddr, struct sctp_transport **transportp,
1126 struct sctp_association *asoc;
1127 union sctp_addr addr;
1128 union sctp_addr *paddr = &addr;
1129 struct sctphdr *sh = sctp_hdr(skb);
1130 union sctp_params params;
1131 struct sctp_init_chunk *init;
1135 * This code will NOT touch anything inside the chunk--it is
1136 * strictly READ-ONLY.
1138 * RFC 2960 3 SCTP packet Format
1140 * Multiple chunks can be bundled into one SCTP packet up to
1141 * the MTU size, except for the INIT, INIT ACK, and SHUTDOWN
1142 * COMPLETE chunks. These chunks MUST NOT be bundled with any
1143 * other chunk in a packet. See Section 6.10 for more details
1144 * on chunk bundling.
1147 /* Find the start of the TLVs and the end of the chunk. This is
1148 * the region we search for address parameters.
1150 init = (struct sctp_init_chunk *)skb->data;
1152 /* Walk the parameters looking for embedded addresses. */
1153 sctp_walk_params(params, init) {
1155 /* Note: Ignoring hostname addresses. */
1156 af = sctp_get_af_specific(param_type2af(params.p->type));
1160 if (!af->from_addr_param(paddr, params.addr, sh->source, 0))
1163 asoc = __sctp_lookup_association(net, laddr, paddr, transportp, dif, sdif);
1171 /* ADD-IP, Section 5.2
1172 * When an endpoint receives an ASCONF Chunk from the remote peer
1173 * special procedures may be needed to identify the association the
1174 * ASCONF Chunk is associated with. To properly find the association
1175 * the following procedures SHOULD be followed:
1177 * D2) If the association is not found, use the address found in the
1178 * Address Parameter TLV combined with the port number found in the
1179 * SCTP common header. If found proceed to rule D4.
1181 * D2-ext) If more than one ASCONF Chunks are packed together, use the
1182 * address found in the ASCONF Address Parameter TLV of each of the
1183 * subsequent ASCONF Chunks. If found, proceed to rule D4.
1185 static struct sctp_association *__sctp_rcv_asconf_lookup(
1187 struct sctp_chunkhdr *ch,
1188 const union sctp_addr *laddr,
1190 struct sctp_transport **transportp,
1193 struct sctp_addip_chunk *asconf = (struct sctp_addip_chunk *)ch;
1195 union sctp_addr_param *param;
1196 union sctp_addr paddr;
1198 if (ntohs(ch->length) < sizeof(*asconf) + sizeof(struct sctp_paramhdr))
1201 /* Skip over the ADDIP header and find the Address parameter */
1202 param = (union sctp_addr_param *)(asconf + 1);
1204 af = sctp_get_af_specific(param_type2af(param->p.type));
1208 if (!af->from_addr_param(&paddr, param, peer_port, 0))
1211 return __sctp_lookup_association(net, laddr, &paddr, transportp, dif, sdif);
1215 /* SCTP-AUTH, Section 6.3:
1216 * If the receiver does not find a STCB for a packet containing an AUTH
1217 * chunk as the first chunk and not a COOKIE-ECHO chunk as the second
1218 * chunk, it MUST use the chunks after the AUTH chunk to look up an existing
1221 * This means that any chunks that can help us identify the association need
1222 * to be looked at to find this association.
1224 static struct sctp_association *__sctp_rcv_walk_lookup(struct net *net,
1225 struct sk_buff *skb,
1226 const union sctp_addr *laddr,
1227 struct sctp_transport **transportp,
1230 struct sctp_association *asoc = NULL;
1231 struct sctp_chunkhdr *ch;
1233 unsigned int chunk_num = 1;
1236 /* Walk through the chunks looking for AUTH or ASCONF chunks
1237 * to help us find the association.
1239 ch = (struct sctp_chunkhdr *)skb->data;
1241 /* Break out if chunk length is less then minimal. */
1242 if (ntohs(ch->length) < sizeof(*ch))
1245 ch_end = ((__u8 *)ch) + SCTP_PAD4(ntohs(ch->length));
1246 if (ch_end > skb_tail_pointer(skb))
1251 have_auth = chunk_num;
1254 case SCTP_CID_COOKIE_ECHO:
1255 /* If a packet arrives containing an AUTH chunk as
1256 * a first chunk, a COOKIE-ECHO chunk as the second
1257 * chunk, and possibly more chunks after them, and
1258 * the receiver does not have an STCB for that
1259 * packet, then authentication is based on
1260 * the contents of the COOKIE- ECHO chunk.
1262 if (have_auth == 1 && chunk_num == 2)
1266 case SCTP_CID_ASCONF:
1267 if (have_auth || net->sctp.addip_noauth)
1268 asoc = __sctp_rcv_asconf_lookup(
1270 sctp_hdr(skb)->source,
1271 transportp, dif, sdif);
1280 ch = (struct sctp_chunkhdr *)ch_end;
1282 } while (ch_end + sizeof(*ch) < skb_tail_pointer(skb));
1288 * There are circumstances when we need to look inside the SCTP packet
1289 * for information to help us find the association. Examples
1290 * include looking inside of INIT/INIT-ACK chunks or after the AUTH
1293 static struct sctp_association *__sctp_rcv_lookup_harder(struct net *net,
1294 struct sk_buff *skb,
1295 const union sctp_addr *laddr,
1296 struct sctp_transport **transportp,
1299 struct sctp_chunkhdr *ch;
1301 /* We do not allow GSO frames here as we need to linearize and
1302 * then cannot guarantee frame boundaries. This shouldn't be an
1303 * issue as packets hitting this are mostly INIT or INIT-ACK and
1304 * those cannot be on GSO-style anyway.
1306 if (skb_is_gso(skb) && skb_is_gso_sctp(skb))
1309 ch = (struct sctp_chunkhdr *)skb->data;
1311 /* The code below will attempt to walk the chunk and extract
1312 * parameter information. Before we do that, we need to verify
1313 * that the chunk length doesn't cause overflow. Otherwise, we'll
1316 if (SCTP_PAD4(ntohs(ch->length)) > skb->len)
1319 /* If this is INIT/INIT-ACK look inside the chunk too. */
1320 if (ch->type == SCTP_CID_INIT || ch->type == SCTP_CID_INIT_ACK)
1321 return __sctp_rcv_init_lookup(net, skb, laddr, transportp, dif, sdif);
1323 return __sctp_rcv_walk_lookup(net, skb, laddr, transportp, dif, sdif);
1326 /* Lookup an association for an inbound skb. */
1327 static struct sctp_association *__sctp_rcv_lookup(struct net *net,
1328 struct sk_buff *skb,
1329 const union sctp_addr *paddr,
1330 const union sctp_addr *laddr,
1331 struct sctp_transport **transportp,
1334 struct sctp_association *asoc;
1336 asoc = __sctp_lookup_association(net, laddr, paddr, transportp, dif, sdif);
1340 /* Further lookup for INIT/INIT-ACK packets.
1341 * SCTP Implementors Guide, 2.18 Handling of address
1342 * parameters within the INIT or INIT-ACK.
1344 asoc = __sctp_rcv_lookup_harder(net, skb, laddr, transportp, dif, sdif);
1348 if (paddr->sa.sa_family == AF_INET)
1349 pr_debug("sctp: asoc not found for src:%pI4:%d dst:%pI4:%d\n",
1350 &laddr->v4.sin_addr, ntohs(laddr->v4.sin_port),
1351 &paddr->v4.sin_addr, ntohs(paddr->v4.sin_port));
1353 pr_debug("sctp: asoc not found for src:%pI6:%d dst:%pI6:%d\n",
1354 &laddr->v6.sin6_addr, ntohs(laddr->v6.sin6_port),
1355 &paddr->v6.sin6_addr, ntohs(paddr->v6.sin6_port));