projects/linux·blob master · refs

1371 lines · 36.7 KiB

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
9 *
10 * This file is part of the SCTP kernel implementation
11 *
12 * These functions handle all input from the IP layer into SCTP.
13 *
14 * Please send any bug reports or fixes you make to the
15 * email address(es):
16 * lksctp developers <[email protected]>
17 *
18 * Written or modified by:
19 * La Monte H.P. Yarroll <[email protected]>
20 * Karl Knutson <[email protected]>
21 * Xingang Guo <[email protected]>
22 * Jon Grimm <[email protected]>
23 * Hui Huang <[email protected]>
24 * Daisy Chang <[email protected]>
25 * Sridhar Samudrala <[email protected]>
26 * Ardelle Fan <[email protected]>
27 */
28
29#include <linux/types.h>
30#include <linux/list.h> /* For struct list_head */
31#include <linux/socket.h>
32#include <linux/ip.h>
33#include <linux/time.h> /* For struct timeval */
34#include <linux/slab.h>
35#include <net/ip.h>
36#include <net/icmp.h>
37#include <net/snmp.h>
38#include <net/sock.h>
39#include <net/xfrm.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>
46
47/* Forward declarations for internal helpers. */
48static int sctp_rcv_ootb(struct sk_buff *);
49static struct sctp_association *__sctp_rcv_lookup(struct net *net,
50 struct sk_buff *skb,
51 const union sctp_addr *paddr,
52 const union sctp_addr *laddr,
53 struct sctp_transport **transportp,
54 int dif, int sdif);
55static 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,
59 int dif, int sdif);
60static struct sctp_association *__sctp_lookup_association(
61 struct net *net,
62 const union sctp_addr *local,
63 const union sctp_addr *peer,
64 struct sctp_transport **pt,
65 int dif, int sdif);
66
67static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb);
68
69
70/* Calculate the SCTP checksum of an SCTP packet. */
71static inline int sctp_rcv_checksum(struct net *net, struct sk_buff *skb)
72{
73 struct sctphdr *sh = sctp_hdr(skb);
74 __le32 cmp = sh->checksum;
75 __le32 val = sctp_compute_cksum(skb, 0);
76
77 if (val != cmp) {
78 /* CRC failure, dump it. */
79 __SCTP_INC_STATS(net, SCTP_MIB_CHECKSUMERRORS);
80 return -1;
81 }
82 return 0;
83}
84
85/*
86 * This is the routine which IP calls when receiving an SCTP packet.
87 */
88int sctp_rcv(struct sk_buff *skb)
89{
90 struct sock *sk;
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;
96 union sctp_addr src;
97 union sctp_addr dest;
98 int family;
99 struct sctp_af *af;
100 struct net *net = dev_net(skb->dev);
101 bool is_gso = skb_is_gso(skb) && skb_is_gso_sctp(skb);
102 int dif, sdif;
103
104 if (skb->pkt_type != PACKET_HOST)
105 goto discard_it;
106
107 __SCTP_INC_STATS(net, SCTP_MIB_INSCTPPACKS);
108
109 /* If packet is too small to contain a single chunk, let's not
110 * waste time on it anymore.
111 */
112 if (skb->len < sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr) +
113 skb_transport_offset(skb))
114 goto discard_it;
115
116 /* If the packet is fragmented and we need to do crc checking,
117 * it's better to just linearize it otherwise crc computing
118 * takes longer.
119 */
120 if (((!is_gso || skb_cloned(skb)) && skb_linearize(skb)) ||
121 !pskb_may_pull(skb, sizeof(struct sctphdr)))
122 goto discard_it;
123
124 /* Pull up the IP header. */
125 __skb_pull(skb, skb_transport_offset(skb));
126
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 &&
131 !is_gso &&
132 sctp_rcv_checksum(net, skb) < 0)
133 goto discard_it;
134 skb->csum_valid = 1;
135
136 __skb_pull(skb, sizeof(struct sctphdr));
137
138 family = ipver2af(ip_hdr(skb)->version);
139 af = sctp_get_af_specific(family);
140 if (unlikely(!af))
141 goto discard_it;
142 SCTP_INPUT_CB(skb)->af = af;
143
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);
149
150 /* If the packet is to or from a non-unicast address,
151 * silently discard the packet.
152 *
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
159 * address."
160 */
161 if (!af->addr_valid(&src, NULL, skb) ||
162 !af->addr_valid(&dest, NULL, skb))
163 goto discard_it;
164
165 asoc = __sctp_rcv_lookup(net, skb, &src, &dest, &transport, dif, sdif);
166
167 if (!asoc)
168 ep = __sctp_rcv_lookup_endpoint(net, skb, &dest, &src, dif, sdif);
169
170 /* Retrieve the common input handling substructure. */
171 rcvr = asoc ? &asoc->base : &ep->base;
172 sk = rcvr->sk;
173
174 /*
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
180 * packet belongs.
181 */
182 if (!asoc) {
183 if (sctp_rcv_ootb(skb)) {
184 __SCTP_INC_STATS(net, SCTP_MIB_OUTOFBLUES);
185 goto discard_release;
186 }
187 }
188
189 if (!xfrm_policy_check(sk, XFRM_POLICY_IN, skb, family))
190 goto discard_release;
191 nf_reset_ct(skb);
192
193 if (sk_filter(sk, skb) || skb->len < sizeof(struct sctp_chunkhdr))
194 goto discard_release;
195
196 /* Create an SCTP packet structure. */
197 chunk = sctp_chunkify(skb, asoc, sk, GFP_ATOMIC);
198 if (!chunk)
199 goto discard_release;
200 SCTP_INPUT_CB(skb)->chunk = chunk;
201
202 /* Remember what endpoint is to handle this packet. */
203 chunk->rcvr = rcvr;
204
205 /* Remember the SCTP header. */
206 chunk->sctp_hdr = sctp_hdr(skb);
207
208 /* Set the source and destination addresses of the incoming chunk. */
209 sctp_init_addrs(chunk, &src, &dest);
210
211 /* Remember where we came from. */
212 chunk->transport = transport;
213
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.
217 */
218 bh_lock_sock(sk);
219
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
226 * of the current sk.
227 */
228 bh_unlock_sock(sk);
229 sk = rcvr->sk;
230 bh_lock_sock(sk);
231 }
232
233 if (sock_owned_by_user(sk) || !sctp_newsk_ready(sk)) {
234 if (sctp_add_backlog(sk, skb)) {
235 bh_unlock_sock(sk);
236 sctp_chunk_free(chunk);
237 skb = NULL; /* sctp_chunk_free already freed the skb */
238 goto discard_release;
239 }
240 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_BACKLOG);
241 } else {
242 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_SOFTIRQ);
243 sctp_inq_push(&chunk->rcvr->inqueue, chunk);
244 }
245
246 bh_unlock_sock(sk);
247
248 /* Release the asoc/ep ref we took in the lookup calls. */
249 if (transport)
250 sctp_transport_put(transport);
251 else
252 sctp_endpoint_put(ep);
253
254 return 0;
255
256discard_it:
257 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_DISCARDS);
258 kfree_skb(skb);
259 return 0;
260
261discard_release:
262 /* Release the asoc/ep ref we took in the lookup calls. */
263 if (transport)
264 sctp_transport_put(transport);
265 else
266 sctp_endpoint_put(ep);
267
268 goto discard_it;
269}
270
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.
275 */
276int sctp_backlog_rcv(struct sock *sk, struct sk_buff *skb)
277{
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;
282 int backloged = 0;
283
284 rcvr = chunk->rcvr;
285
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.
289 */
290 if (rcvr->dead) {
291 sctp_chunk_free(chunk);
292 goto done;
293 }
294
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.
305 */
306
307 sk = rcvr->sk;
308 local_bh_disable();
309 bh_lock_sock(sk);
310
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);
314 else
315 backloged = 1;
316 } else
317 sctp_inq_push(inqueue, chunk);
318
319 bh_unlock_sock(sk);
320 local_bh_enable();
321
322 /* If the chunk was backloged again, don't drop refs */
323 if (backloged)
324 return 0;
325 } else {
326 if (!sctp_newsk_ready(sk)) {
327 if (!sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf)))
328 return 0;
329 sctp_chunk_free(chunk);
330 } else {
331 sctp_inq_push(inqueue, chunk);
332 }
333 }
334
335done:
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));
341 else
342 BUG();
343
344 return 0;
345}
346
347static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb)
348{
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;
352 int ret;
353
354 ret = sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf));
355 if (!ret) {
356 /* Hold the assoc/ep while hanging on the backlog queue.
357 * This way, we know structures we need will not disappear
358 * from us
359 */
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));
364 else
365 BUG();
366 }
367 return ret;
368
369}
370
371/* Handle icmp frag needed error. */
372void sctp_icmp_frag_needed(struct sock *sk, struct sctp_association *asoc,
373 struct sctp_transport *t, __u32 pmtu)
374{
375 if (!t ||
376 (t->pathmtu <= pmtu &&
377 t->pl.probe_size + sctp_transport_pl_hlen(t) <= pmtu))
378 return;
379
380 if (sock_owned_by_user(sk)) {
381 atomic_set(&t->mtu_info, pmtu);
382 asoc->pmtu_pending = 1;
383 t->pmtu_pending = 1;
384 return;
385 }
386
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.
391 */
392 return;
393
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.
397 */
398 if (!sctp_transport_update_pmtu(t, pmtu))
399 return;
400
401 /* Update association pmtu. */
402 sctp_assoc_sync_pmtu(asoc);
403
404 /* Retransmit with the new pmtu setting. */
405 sctp_retransmit(&asoc->outqueue, t, SCTP_RTXR_PMTUD);
406}
407
408void sctp_icmp_redirect(struct sock *sk, struct sctp_transport *t,
409 struct sk_buff *skb)
410{
411 struct dst_entry *dst;
412
413 if (sock_owned_by_user(sk) || !t)
414 return;
415 dst = sctp_transport_dst_check(t);
416 if (dst)
417 dst->ops->redirect(dst, sk, skb);
418}
419
420/*
421 * SCTP Implementer's Guide, 2.37 ICMP handling procedures
422 *
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.
426 *
427 * This function sends an event to the state machine, which will abort the
428 * association.
429 *
430 */
431void sctp_icmp_proto_unreachable(struct sock *sk,
432 struct sctp_association *asoc,
433 struct sctp_transport *t)
434{
435 if (sock_owned_by_user(sk)) {
436 if (timer_pending(&t->proto_unreach_timer))
437 return;
438 else {
439 sctp_transport_hold(t);
440 if (mod_timer(&t->proto_unreach_timer,
441 jiffies + (HZ / 20)))
442 sctp_transport_put(t);
443 }
444 } else {
445 struct net *net = sock_net(sk);
446
447 pr_debug("%s: unrecognized next header type "
448 "encountered!\n", __func__);
449
450 if (timer_delete(&t->proto_unreach_timer))
451 sctp_transport_put(t);
452
453 sctp_do_sm(net, SCTP_EVENT_T_OTHER,
454 SCTP_ST_OTHER(SCTP_EVENT_ICMP_PROTO_UNREACH),
455 asoc->state, asoc->ep, asoc, t,
456 GFP_ATOMIC);
457 }
458}
459
460/* Common lookup code for icmp/icmpv6 error handler. */
461struct sock *sctp_err_lookup(struct net *net, int family, struct sk_buff *skb,
462 struct sctphdr *sctphdr,
463 struct sctp_association **app,
464 struct sctp_transport **tpp)
465{
466 struct sctp_init_chunk *chunkhdr, _chunkhdr;
467 union sctp_addr saddr;
468 union sctp_addr daddr;
469 struct sctp_af *af;
470 struct sock *sk = NULL;
471 struct sctp_association *asoc;
472 struct sctp_transport *transport = NULL;
473 __u32 vtag = ntohl(sctphdr->vtag);
474 int sdif = inet_sdif(skb);
475 int dif = inet_iif(skb);
476
477 *app = NULL; *tpp = NULL;
478
479 af = sctp_get_af_specific(family);
480 if (unlikely(!af)) {
481 return NULL;
482 }
483
484 /* Initialize local addresses for lookups. */
485 af->from_skb(&saddr, skb, 1);
486 af->from_skb(&daddr, skb, 0);
487
488 /* Look for an association that matches the incoming ICMP error
489 * packet.
490 */
491 asoc = __sctp_lookup_association(net, &saddr, &daddr, &transport, dif, sdif);
492 if (!asoc)
493 return NULL;
494
495 sk = asoc->base.sk;
496
497 /* RFC 4960, Appendix C. ICMP Handling
498 *
499 * ICMP6) An implementation MUST validate that the Verification Tag
500 * contained in the ICMP message matches the Verification Tag of
501 * the peer. If the Verification Tag is not 0 and does NOT
502 * match, discard the ICMP message. If it is 0 and the ICMP
503 * message contains enough bytes to verify that the chunk type is
504 * an INIT chunk and that the Initiate Tag matches the tag of the
505 * peer, continue with ICMP7. If the ICMP message is too short
506 * or the chunk type or the Initiate Tag does not match, silently
507 * discard the packet.
508 */
509 if (vtag == 0) {
510 /* chunk header + first 4 octects of init header */
511 chunkhdr = skb_header_pointer(skb, skb_transport_offset(skb) +
512 sizeof(struct sctphdr),
513 sizeof(struct sctp_chunkhdr) +
514 sizeof(__be32), &_chunkhdr);
515 if (!chunkhdr ||
516 chunkhdr->chunk_hdr.type != SCTP_CID_INIT ||
517 ntohl(chunkhdr->init_hdr.init_tag) != asoc->c.my_vtag)
518 goto out;
519
520 } else if (vtag != asoc->c.peer_vtag) {
521 goto out;
522 }
523
524 bh_lock_sock(sk);
525
526 /* If too many ICMPs get dropped on busy
527 * servers this needs to be solved differently.
528 */
529 if (sock_owned_by_user(sk))
530 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
531
532 *app = asoc;
533 *tpp = transport;
534 return sk;
535
536out:
537 sctp_transport_put(transport);
538 return NULL;
539}
540
541/* Common cleanup code for icmp/icmpv6 error handler. */
542void sctp_err_finish(struct sock *sk, struct sctp_transport *t)
543 __releases(&((__sk)->sk_lock.slock))
544{
545 bh_unlock_sock(sk);
546 sctp_transport_put(t);
547}
548
549static void sctp_v4_err_handle(struct sctp_transport *t, struct sk_buff *skb,
550 __u8 type, __u8 code, __u32 info)
551{
552 struct sctp_association *asoc = t->asoc;
553 struct sock *sk = asoc->base.sk;
554 int err = 0;
555
556 switch (type) {
557 case ICMP_PARAMETERPROB:
558 err = EPROTO;
559 break;
560 case ICMP_DEST_UNREACH:
561 if (code > NR_ICMP_UNREACH)
562 return;
563 if (code == ICMP_FRAG_NEEDED) {
564 sctp_icmp_frag_needed(sk, asoc, t, SCTP_TRUNC4(info));
565 return;
566 }
567 if (code == ICMP_PROT_UNREACH) {
568 sctp_icmp_proto_unreachable(sk, asoc, t);
569 return;
570 }
571 err = icmp_err_convert[code].errno;
572 break;
573 case ICMP_TIME_EXCEEDED:
574 if (code == ICMP_EXC_FRAGTIME)
575 return;
576
577 err = EHOSTUNREACH;
578 break;
579 case ICMP_REDIRECT:
580 sctp_icmp_redirect(sk, t, skb);
581 return;
582 default:
583 return;
584 }
585 if (!sock_owned_by_user(sk) && inet_test_bit(RECVERR, sk)) {
586 sk->sk_err = err;
587 sk_error_report(sk);
588 } else { /* Only an error on timeout */
589 WRITE_ONCE(sk->sk_err_soft, err);
590 }
591}
592
593/*
594 * This routine is called by the ICMP module when it gets some
595 * sort of error condition. If err < 0 then the socket should
596 * be closed and the error returned to the user. If err > 0
597 * it's just the icmp type << 8 | icmp code. After adjustment
598 * header points to the first 8 bytes of the sctp header. We need
599 * to find the appropriate port.
600 *
601 * The locking strategy used here is very "optimistic". When
602 * someone else accesses the socket the ICMP is just dropped
603 * and for some paths there is no check at all.
604 * A more general error queue to queue errors for later handling
605 * is probably better.
606 *
607 */
608int sctp_v4_err(struct sk_buff *skb, __u32 info)
609{
610 const struct iphdr *iph = (const struct iphdr *)skb->data;
611 const int type = icmp_hdr(skb)->type;
612 const int code = icmp_hdr(skb)->code;
613 struct net *net = dev_net(skb->dev);
614 struct sctp_transport *transport;
615 struct sctp_association *asoc;
616 __u16 saveip, savesctp;
617 struct sock *sk;
618
619 /* Fix up skb to look at the embedded net header. */
620 saveip = skb->network_header;
621 savesctp = skb->transport_header;
622 skb_reset_network_header(skb);
623 skb_set_transport_header(skb, iph->ihl * 4);
624 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &transport);
625 /* Put back, the original values. */
626 skb->network_header = saveip;
627 skb->transport_header = savesctp;
628 if (!sk) {
629 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
630 return -ENOENT;
631 }
632
633 sctp_v4_err_handle(transport, skb, type, code, info);
634 sctp_err_finish(sk, transport);
635
636 return 0;
637}
638
639int sctp_udp_v4_err(struct sock *sk, struct sk_buff *skb)
640{
641 struct net *net = dev_net(skb->dev);
642 struct sctp_association *asoc;
643 struct sctp_transport *t;
644 struct icmphdr *hdr;
645 __u32 info = 0;
646
647 skb->transport_header += sizeof(struct udphdr);
648 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &t);
649 if (!sk) {
650 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
651 return -ENOENT;
652 }
653
654 skb->transport_header -= sizeof(struct udphdr);
655 hdr = (struct icmphdr *)(skb_network_header(skb) - sizeof(struct icmphdr));
656 if (hdr->type == ICMP_REDIRECT) {
657 /* can't be handled without outer iphdr known, leave it to udp_err */
658 sctp_err_finish(sk, t);
659 return 0;
660 }
661 if (hdr->type == ICMP_DEST_UNREACH && hdr->code == ICMP_FRAG_NEEDED)
662 info = ntohs(hdr->un.frag.mtu);
663 sctp_v4_err_handle(t, skb, hdr->type, hdr->code, info);
664
665 sctp_err_finish(sk, t);
666 return 1;
667}
668
669/*
670 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
671 *
672 * This function scans all the chunks in the OOTB packet to determine if
673 * the packet should be discarded right away. If a response might be needed
674 * for this packet, or, if further processing is possible, the packet will
675 * be queued to a proper inqueue for the next phase of handling.
676 *
677 * Output:
678 * Return 0 - If further processing is needed.
679 * Return 1 - If the packet can be discarded right away.
680 */
681static int sctp_rcv_ootb(struct sk_buff *skb)
682{
683 struct sctp_chunkhdr *ch, _ch;
684 int ch_end, offset = 0;
685
686 /* Scan through all the chunks in the packet. */
687 do {
688 /* Make sure we have at least the header there */
689 if (offset + sizeof(_ch) > skb->len)
690 break;
691
692 ch = skb_header_pointer(skb, offset, sizeof(*ch), &_ch);
693
694 /* Break out if chunk length is less then minimal. */
695 if (!ch || ntohs(ch->length) < sizeof(_ch))
696 break;
697
698 ch_end = offset + SCTP_PAD4(ntohs(ch->length));
699 if (ch_end > skb->len)
700 break;
701
702 /* RFC 8.4, 2) If the OOTB packet contains an ABORT chunk, the
703 * receiver MUST silently discard the OOTB packet and take no
704 * further action.
705 */
706 if (SCTP_CID_ABORT == ch->type)
707 goto discard;
708
709 /* RFC 8.4, 6) If the packet contains a SHUTDOWN COMPLETE
710 * chunk, the receiver should silently discard the packet
711 * and take no further action.
712 */
713 if (SCTP_CID_SHUTDOWN_COMPLETE == ch->type)
714 goto discard;
715
716 /* RFC 4460, 2.11.2
717 * This will discard packets with INIT chunk bundled as
718 * subsequent chunks in the packet. When INIT is first,
719 * the normal INIT processing will discard the chunk.
720 */
721 if (SCTP_CID_INIT == ch->type && (void *)ch != skb->data)
722 goto discard;
723
724 offset = ch_end;
725 } while (ch_end < skb->len);
726
727 return 0;
728
729discard:
730 return 1;
731}
732
733/* Insert endpoint into the hash table. */
734static int __sctp_hash_endpoint(struct sctp_endpoint *ep)
735{
736 struct sock *sk = ep->base.sk;
737 struct net *net = sock_net(sk);
738 struct sctp_hashbucket *head;
739 int err = 0;
740
741 ep->hashent = sctp_ep_hashfn(net, ep->base.bind_addr.port);
742 head = &sctp_ep_hashtable[ep->hashent];
743
744 write_lock(&head->lock);
745 if (sk->sk_reuseport) {
746 bool any = sctp_is_ep_boundall(sk);
747 struct sctp_endpoint *ep2;
748 struct list_head *list;
749 int cnt = 0;
750
751 err = 1;
752
753 list_for_each(list, &ep->base.bind_addr.address_list)
754 cnt++;
755
756 sctp_for_each_hentry(ep2, &head->chain) {
757 struct sock *sk2 = ep2->base.sk;
758
759 if (!net_eq(sock_net(sk2), net) || sk2 == sk ||
760 !uid_eq(sk_uid(sk2), sk_uid(sk)) ||
761 !sk2->sk_reuseport)
762 continue;
763
764 err = sctp_bind_addrs_check(sctp_sk(sk2),
765 sctp_sk(sk), cnt);
766 if (!err) {
767 err = reuseport_add_sock(sk, sk2, any);
768 if (err)
769 goto out;
770 break;
771 } else if (err < 0) {
772 goto out;
773 }
774 }
775
776 if (err) {
777 err = reuseport_alloc(sk, any);
778 if (err)
779 goto out;
780 }
781 }
782
783 hlist_add_head(&ep->node, &head->chain);
784out:
785 write_unlock(&head->lock);
786 return err;
787}
788
789/* Add an endpoint to the hash. Local BH-safe. */
790int sctp_hash_endpoint(struct sctp_endpoint *ep)
791{
792 int err;
793
794 local_bh_disable();
795 err = __sctp_hash_endpoint(ep);
796 local_bh_enable();
797
798 return err;
799}
800
801/* Remove endpoint from the hash table. */
802static void __sctp_unhash_endpoint(struct sctp_endpoint *ep)
803{
804 struct sock *sk = ep->base.sk;
805 struct sctp_hashbucket *head;
806
807 ep->hashent = sctp_ep_hashfn(sock_net(sk), ep->base.bind_addr.port);
808
809 head = &sctp_ep_hashtable[ep->hashent];
810
811 write_lock(&head->lock);
812 if (rcu_access_pointer(sk->sk_reuseport_cb))
813 reuseport_detach_sock(sk);
814 hlist_del_init(&ep->node);
815 write_unlock(&head->lock);
816}
817
818/* Remove endpoint from the hash. Local BH-safe. */
819void sctp_unhash_endpoint(struct sctp_endpoint *ep)
820{
821 local_bh_disable();
822 __sctp_unhash_endpoint(ep);
823 local_bh_enable();
824}
825
826static inline __u32 sctp_hashfn(const struct net *net, __be16 lport,
827 const union sctp_addr *paddr, __u32 seed)
828{
829 __u32 addr;
830
831 if (paddr->sa.sa_family == AF_INET6)
832 addr = jhash(&paddr->v6.sin6_addr, 16, seed);
833 else
834 addr = (__force __u32)paddr->v4.sin_addr.s_addr;
835
836 return jhash_3words(addr, ((__force __u32)paddr->v4.sin_port) << 16 |
837 (__force __u32)lport, net_hash_mix(net), seed);
838}
839
840/* Look up an endpoint. */
841static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(
842 struct net *net, struct sk_buff *skb,
843 const union sctp_addr *laddr,
844 const union sctp_addr *paddr,
845 int dif, int sdif)
846{
847 struct sctp_hashbucket *head;
848 struct sctp_endpoint *ep;
849 struct sock *sk;
850 __be16 lport;
851 int hash;
852
853 lport = laddr->v4.sin_port;
854 hash = sctp_ep_hashfn(net, ntohs(lport));
855 head = &sctp_ep_hashtable[hash];
856 read_lock(&head->lock);
857 sctp_for_each_hentry(ep, &head->chain) {
858 if (sctp_endpoint_is_match(ep, net, laddr, dif, sdif))
859 goto hit;
860 }
861
862 ep = sctp_sk(net->sctp.ctl_sock)->ep;
863
864hit:
865 sk = ep->base.sk;
866 if (sk->sk_reuseport) {
867 __u32 phash = sctp_hashfn(net, lport, paddr, 0);
868
869 sk = reuseport_select_sock(sk, phash, skb,
870 sizeof(struct sctphdr));
871 if (sk)
872 ep = sctp_sk(sk)->ep;
873 }
874 sctp_endpoint_hold(ep);
875 read_unlock(&head->lock);
876 return ep;
877}
878
879/* rhashtable for transport */
880struct sctp_hash_cmp_arg {
881 const union sctp_addr *paddr;
882 const struct net *net;
883 __be16 lport;
884};
885
886static inline int sctp_hash_cmp(struct rhashtable_compare_arg *arg,
887 const void *ptr)
888{
889 struct sctp_transport *t = (struct sctp_transport *)ptr;
890 const struct sctp_hash_cmp_arg *x = arg->key;
891 int err = 1;
892
893 if (!sctp_cmp_addr_exact(&t->ipaddr, x->paddr))
894 return err;
895 if (!sctp_transport_hold(t))
896 return err;
897
898 if (!net_eq(t->asoc->base.net, x->net))
899 goto out;
900 if (x->lport != htons(t->asoc->base.bind_addr.port))
901 goto out;
902
903 err = 0;
904out:
905 sctp_transport_put(t);
906 return err;
907}
908
909static inline __u32 sctp_hash_obj(const void *data, u32 len, u32 seed)
910{
911 const struct sctp_transport *t = data;
912
913 return sctp_hashfn(t->asoc->base.net,
914 htons(t->asoc->base.bind_addr.port),
915 &t->ipaddr, seed);
916}
917
918static inline __u32 sctp_hash_key(const void *data, u32 len, u32 seed)
919{
920 const struct sctp_hash_cmp_arg *x = data;
921
922 return sctp_hashfn(x->net, x->lport, x->paddr, seed);
923}
924
925static const struct rhashtable_params sctp_hash_params = {
926 .head_offset = offsetof(struct sctp_transport, node),
927 .hashfn = sctp_hash_key,
928 .obj_hashfn = sctp_hash_obj,
929 .obj_cmpfn = sctp_hash_cmp,
930 .automatic_shrinking = true,
931};
932
933int sctp_transport_hashtable_init(void)
934{
935 return rhltable_init(&sctp_transport_hashtable, &sctp_hash_params);
936}
937
938void sctp_transport_hashtable_destroy(void)
939{
940 rhltable_destroy(&sctp_transport_hashtable);
941}
942
943int sctp_hash_transport(struct sctp_transport *t)
944{
945 struct sctp_transport *transport;
946 struct rhlist_head *tmp, *list;
947 struct sctp_hash_cmp_arg arg;
948 int err;
949
950 if (t->asoc->temp)
951 return 0;
952
953 arg.net = t->asoc->base.net;
954 arg.paddr = &t->ipaddr;
955 arg.lport = htons(t->asoc->base.bind_addr.port);
956
957 rcu_read_lock();
958 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
959 sctp_hash_params);
960
961 rhl_for_each_entry_rcu(transport, tmp, list, node)
962 if (transport->asoc->ep == t->asoc->ep) {
963 rcu_read_unlock();
964 return -EEXIST;
965 }
966 rcu_read_unlock();
967
968 err = rhltable_insert_key(&sctp_transport_hashtable, &arg,
969 &t->node, sctp_hash_params);
970 if (err)
971 pr_err_once("insert transport fail, errno %d\n", err);
972
973 return err;
974}
975
976void sctp_unhash_transport(struct sctp_transport *t)
977{
978 if (t->asoc->temp)
979 return;
980
981 rhltable_remove(&sctp_transport_hashtable, &t->node,
982 sctp_hash_params);
983}
984
985bool sctp_sk_bound_dev_eq(struct net *net, int bound_dev_if, int dif, int sdif)
986{
987 bool l3mdev_accept = true;
988
989#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
990 l3mdev_accept = !!READ_ONCE(net->sctp.l3mdev_accept);
991#endif
992 return inet_bound_dev_eq(l3mdev_accept, bound_dev_if, dif, sdif);
993}
994
995/* return a transport with holding it */
996struct sctp_transport *sctp_addrs_lookup_transport(
997 struct net *net,
998 const union sctp_addr *laddr,
999 const union sctp_addr *paddr,
1000 int dif, int sdif)
1001{
1002 struct rhlist_head *tmp, *list;
1003 struct sctp_transport *t;
1004 int bound_dev_if;
1005 struct sctp_hash_cmp_arg arg = {
1006 .paddr = paddr,
1007 .net = net,
1008 .lport = laddr->v4.sin_port,
1009 };
1010
1011 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
1012 sctp_hash_params);
1013
1014 rhl_for_each_entry_rcu(t, tmp, list, node) {
1015 if (!sctp_transport_hold(t))
1016 continue;
1017
1018 bound_dev_if = READ_ONCE(t->asoc->base.sk->sk_bound_dev_if);
1019 if (sctp_sk_bound_dev_eq(net, bound_dev_if, dif, sdif) &&
1020 sctp_bind_addr_match(&t->asoc->base.bind_addr,
1021 laddr, sctp_sk(t->asoc->base.sk)))
1022 return t;
1023 sctp_transport_put(t);
1024 }
1025
1026 return NULL;
1027}
1028
1029/* return a transport without holding it, as it's only used under sock lock */
1030struct sctp_transport *sctp_epaddr_lookup_transport(
1031 const struct sctp_endpoint *ep,
1032 const union sctp_addr *paddr)
1033{
1034 struct rhlist_head *tmp, *list;
1035 struct sctp_transport *t;
1036 struct sctp_hash_cmp_arg arg = {
1037 .paddr = paddr,
1038 .net = ep->base.net,
1039 .lport = htons(ep->base.bind_addr.port),
1040 };
1041
1042 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
1043 sctp_hash_params);
1044
1045 rhl_for_each_entry_rcu(t, tmp, list, node)
1046 if (ep == t->asoc->ep)
1047 return t;
1048
1049 return NULL;
1050}
1051
1052/* Look up an association. */
1053static struct sctp_association *__sctp_lookup_association(
1054 struct net *net,
1055 const union sctp_addr *local,
1056 const union sctp_addr *peer,
1057 struct sctp_transport **pt,
1058 int dif, int sdif)
1059{
1060 struct sctp_transport *t;
1061 struct sctp_association *asoc = NULL;
1062
1063 t = sctp_addrs_lookup_transport(net, local, peer, dif, sdif);
1064 if (!t)
1065 goto out;
1066
1067 asoc = t->asoc;
1068 *pt = t;
1069
1070out:
1071 return asoc;
1072}
1073
1074/* Look up an association. protected by RCU read lock */
1075static
1076struct sctp_association *sctp_lookup_association(struct net *net,
1077 const union sctp_addr *laddr,
1078 const union sctp_addr *paddr,
1079 struct sctp_transport **transportp,
1080 int dif, int sdif)
1081{
1082 struct sctp_association *asoc;
1083
1084 rcu_read_lock();
1085 asoc = __sctp_lookup_association(net, laddr, paddr, transportp, dif, sdif);
1086 rcu_read_unlock();
1087
1088 return asoc;
1089}
1090
1091/* Is there an association matching the given local and peer addresses? */
1092bool sctp_has_association(struct net *net,
1093 const union sctp_addr *laddr,
1094 const union sctp_addr *paddr,
1095 int dif, int sdif)
1096{
1097 struct sctp_transport *transport;
1098
1099 if (sctp_lookup_association(net, laddr, paddr, &transport, dif, sdif)) {
1100 sctp_transport_put(transport);
1101 return true;
1102 }
1103
1104 return false;
1105}
1106
1107/*
1108 * SCTP Implementors Guide, 2.18 Handling of address
1109 * parameters within the INIT or INIT-ACK.
1110 *
1111 * D) When searching for a matching TCB upon reception of an INIT
1112 * or INIT-ACK chunk the receiver SHOULD use not only the
1113 * source address of the packet (containing the INIT or
1114 * INIT-ACK) but the receiver SHOULD also use all valid
1115 * address parameters contained within the chunk.
1116 *
1117 * 2.18.3 Solution description
1118 *
1119 * This new text clearly specifies to an implementor the need
1120 * to look within the INIT or INIT-ACK. Any implementation that
1121 * does not do this, may not be able to establish associations
1122 * in certain circumstances.
1123 *
1124 */
1125static struct sctp_association *__sctp_rcv_init_lookup(struct net *net,
1126 struct sk_buff *skb,
1127 const union sctp_addr *laddr, struct sctp_transport **transportp,
1128 int dif, int sdif)
1129{
1130 struct sctp_association *asoc;
1131 union sctp_addr addr;
1132 union sctp_addr *paddr = &addr;
1133 struct sctphdr *sh = sctp_hdr(skb);
1134 union sctp_params params;
1135 struct sctp_init_chunk *init;
1136 struct sctp_af *af;
1137
1138 /*
1139 * This code will NOT touch anything inside the chunk--it is
1140 * strictly READ-ONLY.
1141 *
1142 * RFC 2960 3 SCTP packet Format
1143 *
1144 * Multiple chunks can be bundled into one SCTP packet up to
1145 * the MTU size, except for the INIT, INIT ACK, and SHUTDOWN
1146 * COMPLETE chunks. These chunks MUST NOT be bundled with any
1147 * other chunk in a packet. See Section 6.10 for more details
1148 * on chunk bundling.
1149 */
1150
1151 /* Find the start of the TLVs and the end of the chunk. This is
1152 * the region we search for address parameters.
1153 */
1154 init = (struct sctp_init_chunk *)skb->data;
1155
1156 /* Walk the parameters looking for embedded addresses. */
1157 sctp_walk_params(params, init) {
1158
1159 /* Note: Ignoring hostname addresses. */
1160 af = sctp_get_af_specific(param_type2af(params.p->type));
1161 if (!af)
1162 continue;
1163
1164 if (!af->from_addr_param(paddr, params.addr, sh->source, 0))
1165 continue;
1166
1167 asoc = __sctp_lookup_association(net, laddr, paddr, transportp, dif, sdif);
1168 if (asoc)
1169 return asoc;
1170 }
1171
1172 return NULL;
1173}
1174
1175/* ADD-IP, Section 5.2
1176 * When an endpoint receives an ASCONF Chunk from the remote peer
1177 * special procedures may be needed to identify the association the
1178 * ASCONF Chunk is associated with. To properly find the association
1179 * the following procedures SHOULD be followed:
1180 *
1181 * D2) If the association is not found, use the address found in the
1182 * Address Parameter TLV combined with the port number found in the
1183 * SCTP common header. If found proceed to rule D4.
1184 *
1185 * D2-ext) If more than one ASCONF Chunks are packed together, use the
1186 * address found in the ASCONF Address Parameter TLV of each of the
1187 * subsequent ASCONF Chunks. If found, proceed to rule D4.
1188 */
1189static struct sctp_association *__sctp_rcv_asconf_lookup(
1190 struct net *net,
1191 struct sctp_chunkhdr *ch,
1192 const union sctp_addr *laddr,
1193 __be16 peer_port,
1194 struct sctp_transport **transportp,
1195 int dif, int sdif)
1196{
1197 struct sctp_addip_chunk *asconf = (struct sctp_addip_chunk *)ch;
1198 struct sctp_af *af;
1199 union sctp_addr_param *param;
1200 union sctp_addr paddr;
1201
1202 if (ntohs(ch->length) < sizeof(*asconf) + sizeof(struct sctp_paramhdr))
1203 return NULL;
1204
1205 /* Skip over the ADDIP header and find the Address parameter */
1206 param = (union sctp_addr_param *)(asconf + 1);
1207
1208 /* The whole address parameter must lie within the chunk before
1209 * af->from_addr_param() reads the variable-length address; otherwise a
1210 * truncated trailing ASCONF chunk lets it read uninitialized bytes past
1211 * the parameter.
1212 */
1213 if (sizeof(*asconf) + ntohs(param->p.length) > ntohs(ch->length))
1214 return NULL;
1215
1216 af = sctp_get_af_specific(param_type2af(param->p.type));
1217 if (unlikely(!af))
1218 return NULL;
1219
1220 if (!af->from_addr_param(&paddr, param, peer_port, 0))
1221 return NULL;
1222
1223 return __sctp_lookup_association(net, laddr, &paddr, transportp, dif, sdif);
1224}
1225
1226
1227/* SCTP-AUTH, Section 6.3:
1228* If the receiver does not find a STCB for a packet containing an AUTH
1229* chunk as the first chunk and not a COOKIE-ECHO chunk as the second
1230* chunk, it MUST use the chunks after the AUTH chunk to look up an existing
1231* association.
1232*
1233* This means that any chunks that can help us identify the association need
1234* to be looked at to find this association.
1235*/
1236static struct sctp_association *__sctp_rcv_walk_lookup(struct net *net,
1237 struct sk_buff *skb,
1238 const union sctp_addr *laddr,
1239 struct sctp_transport **transportp,
1240 int dif, int sdif)
1241{
1242 struct sctp_association *asoc = NULL;
1243 struct sctp_chunkhdr *ch;
1244 int have_auth = 0;
1245 unsigned int chunk_num = 1;
1246 __u8 *ch_end;
1247
1248 /* Walk through the chunks looking for AUTH or ASCONF chunks
1249 * to help us find the association.
1250 */
1251 ch = (struct sctp_chunkhdr *)skb->data;
1252 do {
1253 /* Break out if chunk length is less then minimal. */
1254 if (ntohs(ch->length) < sizeof(*ch))
1255 break;
1256
1257 ch_end = ((__u8 *)ch) + SCTP_PAD4(ntohs(ch->length));
1258 if (ch_end > skb_tail_pointer(skb))
1259 break;
1260
1261 switch (ch->type) {
1262 case SCTP_CID_AUTH:
1263 have_auth = chunk_num;
1264 break;
1265
1266 case SCTP_CID_COOKIE_ECHO:
1267 /* If a packet arrives containing an AUTH chunk as
1268 * a first chunk, a COOKIE-ECHO chunk as the second
1269 * chunk, and possibly more chunks after them, and
1270 * the receiver does not have an STCB for that
1271 * packet, then authentication is based on
1272 * the contents of the COOKIE- ECHO chunk.
1273 */
1274 if (have_auth == 1 && chunk_num == 2)
1275 return NULL;
1276 break;
1277
1278 case SCTP_CID_ASCONF:
1279 if (have_auth || net->sctp.addip_noauth)
1280 asoc = __sctp_rcv_asconf_lookup(
1281 net, ch, laddr,
1282 sctp_hdr(skb)->source,
1283 transportp, dif, sdif);
1284 break;
1285 default:
1286 break;
1287 }
1288
1289 if (asoc)
1290 break;
1291
1292 ch = (struct sctp_chunkhdr *)ch_end;
1293 chunk_num++;
1294 } while (ch_end + sizeof(*ch) < skb_tail_pointer(skb));
1295
1296 return asoc;
1297}
1298
1299/*
1300 * There are circumstances when we need to look inside the SCTP packet
1301 * for information to help us find the association. Examples
1302 * include looking inside of INIT/INIT-ACK chunks or after the AUTH
1303 * chunks.
1304 */
1305static struct sctp_association *__sctp_rcv_lookup_harder(struct net *net,
1306 struct sk_buff *skb,
1307 const union sctp_addr *laddr,
1308 struct sctp_transport **transportp,
1309 int dif, int sdif)
1310{
1311 struct sctp_chunkhdr *ch;
1312
1313 /* We do not allow GSO frames here as we need to linearize and
1314 * then cannot guarantee frame boundaries. This shouldn't be an
1315 * issue as packets hitting this are mostly INIT or INIT-ACK and
1316 * those cannot be on GSO-style anyway.
1317 */
1318 if (skb_is_gso(skb) && skb_is_gso_sctp(skb))
1319 return NULL;
1320
1321 ch = (struct sctp_chunkhdr *)skb->data;
1322
1323 /* The code below will attempt to walk the chunk and extract
1324 * parameter information. Before we do that, we need to verify
1325 * that the chunk length doesn't cause overflow. Otherwise, we'll
1326 * walk off the end.
1327 */
1328 if (SCTP_PAD4(ntohs(ch->length)) > skb->len)
1329 return NULL;
1330
1331 /* If this is INIT/INIT-ACK look inside the chunk too. */
1332 if (ch->type == SCTP_CID_INIT || ch->type == SCTP_CID_INIT_ACK)
1333 return __sctp_rcv_init_lookup(net, skb, laddr, transportp, dif, sdif);
1334
1335 return __sctp_rcv_walk_lookup(net, skb, laddr, transportp, dif, sdif);
1336}
1337
1338/* Lookup an association for an inbound skb. */
1339static struct sctp_association *__sctp_rcv_lookup(struct net *net,
1340 struct sk_buff *skb,
1341 const union sctp_addr *paddr,
1342 const union sctp_addr *laddr,
1343 struct sctp_transport **transportp,
1344 int dif, int sdif)
1345{
1346 struct sctp_association *asoc;
1347
1348 asoc = __sctp_lookup_association(net, laddr, paddr, transportp, dif, sdif);
1349 if (asoc)
1350 goto out;
1351
1352 /* Further lookup for INIT/INIT-ACK packets.
1353 * SCTP Implementors Guide, 2.18 Handling of address
1354 * parameters within the INIT or INIT-ACK.
1355 */
1356 asoc = __sctp_rcv_lookup_harder(net, skb, laddr, transportp, dif, sdif);
1357 if (asoc)
1358 goto out;
1359
1360 if (paddr->sa.sa_family == AF_INET)
1361 pr_debug("sctp: asoc not found for src:%pI4:%d dst:%pI4:%d\n",
1362 &laddr->v4.sin_addr, ntohs(laddr->v4.sin_port),
1363 &paddr->v4.sin_addr, ntohs(paddr->v4.sin_port));
1364 else
1365 pr_debug("sctp: asoc not found for src:%pI6:%d dst:%pI6:%d\n",
1366 &laddr->v6.sin6_addr, ntohs(laddr->v6.sin6_port),
1367 &paddr->v6.sin6_addr, ntohs(paddr->v6.sin6_port));
1368
1369out:
1370 return asoc;
1371}