Loading...
1/* SCTP kernel implementation
2 * Copyright (c) 1999-2000 Cisco, Inc.
3 * Copyright (c) 1999-2001 Motorola, Inc.
4 * Copyright (c) 2001-2003 International Business Machines, Corp.
5 * Copyright (c) 2001 Intel Corp.
6 * Copyright (c) 2001 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
8 *
9 * This file is part of the SCTP kernel implementation
10 *
11 * These functions handle all input from the IP layer into SCTP.
12 *
13 * This SCTP implementation is free software;
14 * you can redistribute it and/or modify it under the terms of
15 * the GNU General Public License as published by
16 * the Free Software Foundation; either version 2, or (at your option)
17 * any later version.
18 *
19 * This SCTP implementation is distributed in the hope that it
20 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
21 * ************************
22 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
23 * See the GNU General Public License for more details.
24 *
25 * You should have received a copy of the GNU General Public License
26 * along with GNU CC; see the file COPYING. If not, see
27 * <http://www.gnu.org/licenses/>.
28 *
29 * Please send any bug reports or fixes you make to the
30 * email address(es):
31 * lksctp developers <linux-sctp@vger.kernel.org>
32 *
33 * Written or modified by:
34 * La Monte H.P. Yarroll <piggy@acm.org>
35 * Karl Knutson <karl@athena.chicago.il.us>
36 * Xingang Guo <xingang.guo@intel.com>
37 * Jon Grimm <jgrimm@us.ibm.com>
38 * Hui Huang <hui.huang@nokia.com>
39 * Daisy Chang <daisyc@us.ibm.com>
40 * Sridhar Samudrala <sri@us.ibm.com>
41 * Ardelle Fan <ardelle.fan@intel.com>
42 */
43
44#include <linux/types.h>
45#include <linux/list.h> /* For struct list_head */
46#include <linux/socket.h>
47#include <linux/ip.h>
48#include <linux/time.h> /* For struct timeval */
49#include <linux/slab.h>
50#include <net/ip.h>
51#include <net/icmp.h>
52#include <net/snmp.h>
53#include <net/sock.h>
54#include <net/xfrm.h>
55#include <net/sctp/sctp.h>
56#include <net/sctp/sm.h>
57#include <net/sctp/checksum.h>
58#include <net/net_namespace.h>
59
60/* Forward declarations for internal helpers. */
61static int sctp_rcv_ootb(struct sk_buff *);
62static struct sctp_association *__sctp_rcv_lookup(struct net *net,
63 struct sk_buff *skb,
64 const union sctp_addr *paddr,
65 const union sctp_addr *laddr,
66 struct sctp_transport **transportp);
67static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(struct net *net,
68 const union sctp_addr *laddr);
69static struct sctp_association *__sctp_lookup_association(
70 struct net *net,
71 const union sctp_addr *local,
72 const union sctp_addr *peer,
73 struct sctp_transport **pt);
74
75static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb);
76
77
78/* Calculate the SCTP checksum of an SCTP packet. */
79static inline int sctp_rcv_checksum(struct net *net, struct sk_buff *skb)
80{
81 struct sctphdr *sh = sctp_hdr(skb);
82 __le32 cmp = sh->checksum;
83 __le32 val = sctp_compute_cksum(skb, 0);
84
85 if (val != cmp) {
86 /* CRC failure, dump it. */
87 SCTP_INC_STATS_BH(net, SCTP_MIB_CHECKSUMERRORS);
88 return -1;
89 }
90 return 0;
91}
92
93struct sctp_input_cb {
94 union {
95 struct inet_skb_parm h4;
96#if IS_ENABLED(CONFIG_IPV6)
97 struct inet6_skb_parm h6;
98#endif
99 } header;
100 struct sctp_chunk *chunk;
101};
102#define SCTP_INPUT_CB(__skb) ((struct sctp_input_cb *)&((__skb)->cb[0]))
103
104/*
105 * This is the routine which IP calls when receiving an SCTP packet.
106 */
107int sctp_rcv(struct sk_buff *skb)
108{
109 struct sock *sk;
110 struct sctp_association *asoc;
111 struct sctp_endpoint *ep = NULL;
112 struct sctp_ep_common *rcvr;
113 struct sctp_transport *transport = NULL;
114 struct sctp_chunk *chunk;
115 struct sctphdr *sh;
116 union sctp_addr src;
117 union sctp_addr dest;
118 int family;
119 struct sctp_af *af;
120 struct net *net = dev_net(skb->dev);
121
122 if (skb->pkt_type != PACKET_HOST)
123 goto discard_it;
124
125 SCTP_INC_STATS_BH(net, SCTP_MIB_INSCTPPACKS);
126
127 if (skb_linearize(skb))
128 goto discard_it;
129
130 sh = sctp_hdr(skb);
131
132 /* Pull up the IP and SCTP headers. */
133 __skb_pull(skb, skb_transport_offset(skb));
134 if (skb->len < sizeof(struct sctphdr))
135 goto discard_it;
136
137 skb->csum_valid = 0; /* Previous value not applicable */
138 if (skb_csum_unnecessary(skb))
139 __skb_decr_checksum_unnecessary(skb);
140 else if (!sctp_checksum_disable && sctp_rcv_checksum(net, skb) < 0)
141 goto discard_it;
142 skb->csum_valid = 1;
143
144 skb_pull(skb, sizeof(struct sctphdr));
145
146 /* Make sure we at least have chunk headers worth of data left. */
147 if (skb->len < sizeof(struct sctp_chunkhdr))
148 goto discard_it;
149
150 family = ipver2af(ip_hdr(skb)->version);
151 af = sctp_get_af_specific(family);
152 if (unlikely(!af))
153 goto discard_it;
154
155 /* Initialize local addresses for lookups. */
156 af->from_skb(&src, skb, 1);
157 af->from_skb(&dest, skb, 0);
158
159 /* If the packet is to or from a non-unicast address,
160 * silently discard the packet.
161 *
162 * This is not clearly defined in the RFC except in section
163 * 8.4 - OOTB handling. However, based on the book "Stream Control
164 * Transmission Protocol" 2.1, "It is important to note that the
165 * IP address of an SCTP transport address must be a routable
166 * unicast address. In other words, IP multicast addresses and
167 * IP broadcast addresses cannot be used in an SCTP transport
168 * address."
169 */
170 if (!af->addr_valid(&src, NULL, skb) ||
171 !af->addr_valid(&dest, NULL, skb))
172 goto discard_it;
173
174 asoc = __sctp_rcv_lookup(net, skb, &src, &dest, &transport);
175
176 if (!asoc)
177 ep = __sctp_rcv_lookup_endpoint(net, &dest);
178
179 /* Retrieve the common input handling substructure. */
180 rcvr = asoc ? &asoc->base : &ep->base;
181 sk = rcvr->sk;
182
183 /*
184 * If a frame arrives on an interface and the receiving socket is
185 * bound to another interface, via SO_BINDTODEVICE, treat it as OOTB
186 */
187 if (sk->sk_bound_dev_if && (sk->sk_bound_dev_if != af->skb_iif(skb))) {
188 if (asoc) {
189 sctp_association_put(asoc);
190 asoc = NULL;
191 } else {
192 sctp_endpoint_put(ep);
193 ep = NULL;
194 }
195 sk = net->sctp.ctl_sock;
196 ep = sctp_sk(sk)->ep;
197 sctp_endpoint_hold(ep);
198 rcvr = &ep->base;
199 }
200
201 /*
202 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
203 * An SCTP packet is called an "out of the blue" (OOTB)
204 * packet if it is correctly formed, i.e., passed the
205 * receiver's checksum check, but the receiver is not
206 * able to identify the association to which this
207 * packet belongs.
208 */
209 if (!asoc) {
210 if (sctp_rcv_ootb(skb)) {
211 SCTP_INC_STATS_BH(net, SCTP_MIB_OUTOFBLUES);
212 goto discard_release;
213 }
214 }
215
216 if (!xfrm_policy_check(sk, XFRM_POLICY_IN, skb, family))
217 goto discard_release;
218 nf_reset(skb);
219
220 if (sk_filter(sk, skb))
221 goto discard_release;
222
223 /* Create an SCTP packet structure. */
224 chunk = sctp_chunkify(skb, asoc, sk, GFP_ATOMIC);
225 if (!chunk)
226 goto discard_release;
227 SCTP_INPUT_CB(skb)->chunk = chunk;
228
229 /* Remember what endpoint is to handle this packet. */
230 chunk->rcvr = rcvr;
231
232 /* Remember the SCTP header. */
233 chunk->sctp_hdr = sh;
234
235 /* Set the source and destination addresses of the incoming chunk. */
236 sctp_init_addrs(chunk, &src, &dest);
237
238 /* Remember where we came from. */
239 chunk->transport = transport;
240
241 /* Acquire access to the sock lock. Note: We are safe from other
242 * bottom halves on this lock, but a user may be in the lock too,
243 * so check if it is busy.
244 */
245 bh_lock_sock(sk);
246
247 if (sk != rcvr->sk) {
248 /* Our cached sk is different from the rcvr->sk. This is
249 * because migrate()/accept() may have moved the association
250 * to a new socket and released all the sockets. So now we
251 * are holding a lock on the old socket while the user may
252 * be doing something with the new socket. Switch our veiw
253 * of the current sk.
254 */
255 bh_unlock_sock(sk);
256 sk = rcvr->sk;
257 bh_lock_sock(sk);
258 }
259
260 if (sock_owned_by_user(sk)) {
261 if (sctp_add_backlog(sk, skb)) {
262 bh_unlock_sock(sk);
263 sctp_chunk_free(chunk);
264 skb = NULL; /* sctp_chunk_free already freed the skb */
265 goto discard_release;
266 }
267 SCTP_INC_STATS_BH(net, SCTP_MIB_IN_PKT_BACKLOG);
268 } else {
269 SCTP_INC_STATS_BH(net, SCTP_MIB_IN_PKT_SOFTIRQ);
270 sctp_inq_push(&chunk->rcvr->inqueue, chunk);
271 }
272
273 bh_unlock_sock(sk);
274
275 /* Release the asoc/ep ref we took in the lookup calls. */
276 if (asoc)
277 sctp_association_put(asoc);
278 else
279 sctp_endpoint_put(ep);
280
281 return 0;
282
283discard_it:
284 SCTP_INC_STATS_BH(net, SCTP_MIB_IN_PKT_DISCARDS);
285 kfree_skb(skb);
286 return 0;
287
288discard_release:
289 /* Release the asoc/ep ref we took in the lookup calls. */
290 if (asoc)
291 sctp_association_put(asoc);
292 else
293 sctp_endpoint_put(ep);
294
295 goto discard_it;
296}
297
298/* Process the backlog queue of the socket. Every skb on
299 * the backlog holds a ref on an association or endpoint.
300 * We hold this ref throughout the state machine to make
301 * sure that the structure we need is still around.
302 */
303int sctp_backlog_rcv(struct sock *sk, struct sk_buff *skb)
304{
305 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
306 struct sctp_inq *inqueue = &chunk->rcvr->inqueue;
307 struct sctp_ep_common *rcvr = NULL;
308 int backloged = 0;
309
310 rcvr = chunk->rcvr;
311
312 /* If the rcvr is dead then the association or endpoint
313 * has been deleted and we can safely drop the chunk
314 * and refs that we are holding.
315 */
316 if (rcvr->dead) {
317 sctp_chunk_free(chunk);
318 goto done;
319 }
320
321 if (unlikely(rcvr->sk != sk)) {
322 /* In this case, the association moved from one socket to
323 * another. We are currently sitting on the backlog of the
324 * old socket, so we need to move.
325 * However, since we are here in the process context we
326 * need to take make sure that the user doesn't own
327 * the new socket when we process the packet.
328 * If the new socket is user-owned, queue the chunk to the
329 * backlog of the new socket without dropping any refs.
330 * Otherwise, we can safely push the chunk on the inqueue.
331 */
332
333 sk = rcvr->sk;
334 bh_lock_sock(sk);
335
336 if (sock_owned_by_user(sk)) {
337 if (sk_add_backlog(sk, skb, sk->sk_rcvbuf))
338 sctp_chunk_free(chunk);
339 else
340 backloged = 1;
341 } else
342 sctp_inq_push(inqueue, chunk);
343
344 bh_unlock_sock(sk);
345
346 /* If the chunk was backloged again, don't drop refs */
347 if (backloged)
348 return 0;
349 } else {
350 sctp_inq_push(inqueue, chunk);
351 }
352
353done:
354 /* Release the refs we took in sctp_add_backlog */
355 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
356 sctp_association_put(sctp_assoc(rcvr));
357 else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
358 sctp_endpoint_put(sctp_ep(rcvr));
359 else
360 BUG();
361
362 return 0;
363}
364
365static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb)
366{
367 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
368 struct sctp_ep_common *rcvr = chunk->rcvr;
369 int ret;
370
371 ret = sk_add_backlog(sk, skb, sk->sk_rcvbuf);
372 if (!ret) {
373 /* Hold the assoc/ep while hanging on the backlog queue.
374 * This way, we know structures we need will not disappear
375 * from us
376 */
377 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
378 sctp_association_hold(sctp_assoc(rcvr));
379 else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
380 sctp_endpoint_hold(sctp_ep(rcvr));
381 else
382 BUG();
383 }
384 return ret;
385
386}
387
388/* Handle icmp frag needed error. */
389void sctp_icmp_frag_needed(struct sock *sk, struct sctp_association *asoc,
390 struct sctp_transport *t, __u32 pmtu)
391{
392 if (!t || (t->pathmtu <= pmtu))
393 return;
394
395 if (sock_owned_by_user(sk)) {
396 asoc->pmtu_pending = 1;
397 t->pmtu_pending = 1;
398 return;
399 }
400
401 if (t->param_flags & SPP_PMTUD_ENABLE) {
402 /* Update transports view of the MTU */
403 sctp_transport_update_pmtu(sk, t, pmtu);
404
405 /* Update association pmtu. */
406 sctp_assoc_sync_pmtu(sk, asoc);
407 }
408
409 /* Retransmit with the new pmtu setting.
410 * Normally, if PMTU discovery is disabled, an ICMP Fragmentation
411 * Needed will never be sent, but if a message was sent before
412 * PMTU discovery was disabled that was larger than the PMTU, it
413 * would not be fragmented, so it must be re-transmitted fragmented.
414 */
415 sctp_retransmit(&asoc->outqueue, t, SCTP_RTXR_PMTUD);
416}
417
418void sctp_icmp_redirect(struct sock *sk, struct sctp_transport *t,
419 struct sk_buff *skb)
420{
421 struct dst_entry *dst;
422
423 if (!t)
424 return;
425 dst = sctp_transport_dst_check(t);
426 if (dst)
427 dst->ops->redirect(dst, sk, skb);
428}
429
430/*
431 * SCTP Implementer's Guide, 2.37 ICMP handling procedures
432 *
433 * ICMP8) If the ICMP code is a "Unrecognized next header type encountered"
434 * or a "Protocol Unreachable" treat this message as an abort
435 * with the T bit set.
436 *
437 * This function sends an event to the state machine, which will abort the
438 * association.
439 *
440 */
441void sctp_icmp_proto_unreachable(struct sock *sk,
442 struct sctp_association *asoc,
443 struct sctp_transport *t)
444{
445 if (sock_owned_by_user(sk)) {
446 if (timer_pending(&t->proto_unreach_timer))
447 return;
448 else {
449 if (!mod_timer(&t->proto_unreach_timer,
450 jiffies + (HZ/20)))
451 sctp_association_hold(asoc);
452 }
453 } else {
454 struct net *net = sock_net(sk);
455
456 pr_debug("%s: unrecognized next header type "
457 "encountered!\n", __func__);
458
459 if (del_timer(&t->proto_unreach_timer))
460 sctp_association_put(asoc);
461
462 sctp_do_sm(net, SCTP_EVENT_T_OTHER,
463 SCTP_ST_OTHER(SCTP_EVENT_ICMP_PROTO_UNREACH),
464 asoc->state, asoc->ep, asoc, t,
465 GFP_ATOMIC);
466 }
467}
468
469/* Common lookup code for icmp/icmpv6 error handler. */
470struct sock *sctp_err_lookup(struct net *net, int family, struct sk_buff *skb,
471 struct sctphdr *sctphdr,
472 struct sctp_association **app,
473 struct sctp_transport **tpp)
474{
475 union sctp_addr saddr;
476 union sctp_addr daddr;
477 struct sctp_af *af;
478 struct sock *sk = NULL;
479 struct sctp_association *asoc;
480 struct sctp_transport *transport = NULL;
481 struct sctp_init_chunk *chunkhdr;
482 __u32 vtag = ntohl(sctphdr->vtag);
483 int len = skb->len - ((void *)sctphdr - (void *)skb->data);
484
485 *app = NULL; *tpp = NULL;
486
487 af = sctp_get_af_specific(family);
488 if (unlikely(!af)) {
489 return NULL;
490 }
491
492 /* Initialize local addresses for lookups. */
493 af->from_skb(&saddr, skb, 1);
494 af->from_skb(&daddr, skb, 0);
495
496 /* Look for an association that matches the incoming ICMP error
497 * packet.
498 */
499 asoc = __sctp_lookup_association(net, &saddr, &daddr, &transport);
500 if (!asoc)
501 return NULL;
502
503 sk = asoc->base.sk;
504
505 /* RFC 4960, Appendix C. ICMP Handling
506 *
507 * ICMP6) An implementation MUST validate that the Verification Tag
508 * contained in the ICMP message matches the Verification Tag of
509 * the peer. If the Verification Tag is not 0 and does NOT
510 * match, discard the ICMP message. If it is 0 and the ICMP
511 * message contains enough bytes to verify that the chunk type is
512 * an INIT chunk and that the Initiate Tag matches the tag of the
513 * peer, continue with ICMP7. If the ICMP message is too short
514 * or the chunk type or the Initiate Tag does not match, silently
515 * discard the packet.
516 */
517 if (vtag == 0) {
518 chunkhdr = (void *)sctphdr + sizeof(struct sctphdr);
519 if (len < sizeof(struct sctphdr) + sizeof(sctp_chunkhdr_t)
520 + sizeof(__be32) ||
521 chunkhdr->chunk_hdr.type != SCTP_CID_INIT ||
522 ntohl(chunkhdr->init_hdr.init_tag) != asoc->c.my_vtag) {
523 goto out;
524 }
525 } else if (vtag != asoc->c.peer_vtag) {
526 goto out;
527 }
528
529 bh_lock_sock(sk);
530
531 /* If too many ICMPs get dropped on busy
532 * servers this needs to be solved differently.
533 */
534 if (sock_owned_by_user(sk))
535 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
536
537 *app = asoc;
538 *tpp = transport;
539 return sk;
540
541out:
542 sctp_association_put(asoc);
543 return NULL;
544}
545
546/* Common cleanup code for icmp/icmpv6 error handler. */
547void sctp_err_finish(struct sock *sk, struct sctp_association *asoc)
548{
549 bh_unlock_sock(sk);
550 sctp_association_put(asoc);
551}
552
553/*
554 * This routine is called by the ICMP module when it gets some
555 * sort of error condition. If err < 0 then the socket should
556 * be closed and the error returned to the user. If err > 0
557 * it's just the icmp type << 8 | icmp code. After adjustment
558 * header points to the first 8 bytes of the sctp header. We need
559 * to find the appropriate port.
560 *
561 * The locking strategy used here is very "optimistic". When
562 * someone else accesses the socket the ICMP is just dropped
563 * and for some paths there is no check at all.
564 * A more general error queue to queue errors for later handling
565 * is probably better.
566 *
567 */
568void sctp_v4_err(struct sk_buff *skb, __u32 info)
569{
570 const struct iphdr *iph = (const struct iphdr *)skb->data;
571 const int ihlen = iph->ihl * 4;
572 const int type = icmp_hdr(skb)->type;
573 const int code = icmp_hdr(skb)->code;
574 struct sock *sk;
575 struct sctp_association *asoc = NULL;
576 struct sctp_transport *transport;
577 struct inet_sock *inet;
578 __u16 saveip, savesctp;
579 int err;
580 struct net *net = dev_net(skb->dev);
581
582 /* Fix up skb to look at the embedded net header. */
583 saveip = skb->network_header;
584 savesctp = skb->transport_header;
585 skb_reset_network_header(skb);
586 skb_set_transport_header(skb, ihlen);
587 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &transport);
588 /* Put back, the original values. */
589 skb->network_header = saveip;
590 skb->transport_header = savesctp;
591 if (!sk) {
592 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
593 return;
594 }
595 /* Warning: The sock lock is held. Remember to call
596 * sctp_err_finish!
597 */
598
599 switch (type) {
600 case ICMP_PARAMETERPROB:
601 err = EPROTO;
602 break;
603 case ICMP_DEST_UNREACH:
604 if (code > NR_ICMP_UNREACH)
605 goto out_unlock;
606
607 /* PMTU discovery (RFC1191) */
608 if (ICMP_FRAG_NEEDED == code) {
609 sctp_icmp_frag_needed(sk, asoc, transport,
610 WORD_TRUNC(info));
611 goto out_unlock;
612 } else {
613 if (ICMP_PROT_UNREACH == code) {
614 sctp_icmp_proto_unreachable(sk, asoc,
615 transport);
616 goto out_unlock;
617 }
618 }
619 err = icmp_err_convert[code].errno;
620 break;
621 case ICMP_TIME_EXCEEDED:
622 /* Ignore any time exceeded errors due to fragment reassembly
623 * timeouts.
624 */
625 if (ICMP_EXC_FRAGTIME == code)
626 goto out_unlock;
627
628 err = EHOSTUNREACH;
629 break;
630 case ICMP_REDIRECT:
631 sctp_icmp_redirect(sk, transport, skb);
632 /* Fall through to out_unlock. */
633 default:
634 goto out_unlock;
635 }
636
637 inet = inet_sk(sk);
638 if (!sock_owned_by_user(sk) && inet->recverr) {
639 sk->sk_err = err;
640 sk->sk_error_report(sk);
641 } else { /* Only an error on timeout */
642 sk->sk_err_soft = err;
643 }
644
645out_unlock:
646 sctp_err_finish(sk, asoc);
647}
648
649/*
650 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
651 *
652 * This function scans all the chunks in the OOTB packet to determine if
653 * the packet should be discarded right away. If a response might be needed
654 * for this packet, or, if further processing is possible, the packet will
655 * be queued to a proper inqueue for the next phase of handling.
656 *
657 * Output:
658 * Return 0 - If further processing is needed.
659 * Return 1 - If the packet can be discarded right away.
660 */
661static int sctp_rcv_ootb(struct sk_buff *skb)
662{
663 sctp_chunkhdr_t *ch;
664 __u8 *ch_end;
665
666 ch = (sctp_chunkhdr_t *) skb->data;
667
668 /* Scan through all the chunks in the packet. */
669 do {
670 /* Break out if chunk length is less then minimal. */
671 if (ntohs(ch->length) < sizeof(sctp_chunkhdr_t))
672 break;
673
674 ch_end = ((__u8 *)ch) + WORD_ROUND(ntohs(ch->length));
675 if (ch_end > skb_tail_pointer(skb))
676 break;
677
678 /* RFC 8.4, 2) If the OOTB packet contains an ABORT chunk, the
679 * receiver MUST silently discard the OOTB packet and take no
680 * further action.
681 */
682 if (SCTP_CID_ABORT == ch->type)
683 goto discard;
684
685 /* RFC 8.4, 6) If the packet contains a SHUTDOWN COMPLETE
686 * chunk, the receiver should silently discard the packet
687 * and take no further action.
688 */
689 if (SCTP_CID_SHUTDOWN_COMPLETE == ch->type)
690 goto discard;
691
692 /* RFC 4460, 2.11.2
693 * This will discard packets with INIT chunk bundled as
694 * subsequent chunks in the packet. When INIT is first,
695 * the normal INIT processing will discard the chunk.
696 */
697 if (SCTP_CID_INIT == ch->type && (void *)ch != skb->data)
698 goto discard;
699
700 ch = (sctp_chunkhdr_t *) ch_end;
701 } while (ch_end < skb_tail_pointer(skb));
702
703 return 0;
704
705discard:
706 return 1;
707}
708
709/* Insert endpoint into the hash table. */
710static void __sctp_hash_endpoint(struct sctp_endpoint *ep)
711{
712 struct net *net = sock_net(ep->base.sk);
713 struct sctp_ep_common *epb;
714 struct sctp_hashbucket *head;
715
716 epb = &ep->base;
717
718 epb->hashent = sctp_ep_hashfn(net, epb->bind_addr.port);
719 head = &sctp_ep_hashtable[epb->hashent];
720
721 write_lock(&head->lock);
722 hlist_add_head(&epb->node, &head->chain);
723 write_unlock(&head->lock);
724}
725
726/* Add an endpoint to the hash. Local BH-safe. */
727void sctp_hash_endpoint(struct sctp_endpoint *ep)
728{
729 local_bh_disable();
730 __sctp_hash_endpoint(ep);
731 local_bh_enable();
732}
733
734/* Remove endpoint from the hash table. */
735static void __sctp_unhash_endpoint(struct sctp_endpoint *ep)
736{
737 struct net *net = sock_net(ep->base.sk);
738 struct sctp_hashbucket *head;
739 struct sctp_ep_common *epb;
740
741 epb = &ep->base;
742
743 epb->hashent = sctp_ep_hashfn(net, epb->bind_addr.port);
744
745 head = &sctp_ep_hashtable[epb->hashent];
746
747 write_lock(&head->lock);
748 hlist_del_init(&epb->node);
749 write_unlock(&head->lock);
750}
751
752/* Remove endpoint from the hash. Local BH-safe. */
753void sctp_unhash_endpoint(struct sctp_endpoint *ep)
754{
755 local_bh_disable();
756 __sctp_unhash_endpoint(ep);
757 local_bh_enable();
758}
759
760/* Look up an endpoint. */
761static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(struct net *net,
762 const union sctp_addr *laddr)
763{
764 struct sctp_hashbucket *head;
765 struct sctp_ep_common *epb;
766 struct sctp_endpoint *ep;
767 int hash;
768
769 hash = sctp_ep_hashfn(net, ntohs(laddr->v4.sin_port));
770 head = &sctp_ep_hashtable[hash];
771 read_lock(&head->lock);
772 sctp_for_each_hentry(epb, &head->chain) {
773 ep = sctp_ep(epb);
774 if (sctp_endpoint_is_match(ep, net, laddr))
775 goto hit;
776 }
777
778 ep = sctp_sk(net->sctp.ctl_sock)->ep;
779
780hit:
781 sctp_endpoint_hold(ep);
782 read_unlock(&head->lock);
783 return ep;
784}
785
786/* rhashtable for transport */
787struct sctp_hash_cmp_arg {
788 const struct sctp_endpoint *ep;
789 const union sctp_addr *laddr;
790 const union sctp_addr *paddr;
791 const struct net *net;
792};
793
794static inline int sctp_hash_cmp(struct rhashtable_compare_arg *arg,
795 const void *ptr)
796{
797 const struct sctp_hash_cmp_arg *x = arg->key;
798 const struct sctp_transport *t = ptr;
799 struct sctp_association *asoc = t->asoc;
800 const struct net *net = x->net;
801
802 if (!sctp_cmp_addr_exact(&t->ipaddr, x->paddr))
803 return 1;
804 if (!net_eq(sock_net(asoc->base.sk), net))
805 return 1;
806 if (x->ep) {
807 if (x->ep != asoc->ep)
808 return 1;
809 } else {
810 if (x->laddr->v4.sin_port != htons(asoc->base.bind_addr.port))
811 return 1;
812 if (!sctp_bind_addr_match(&asoc->base.bind_addr,
813 x->laddr, sctp_sk(asoc->base.sk)))
814 return 1;
815 }
816
817 return 0;
818}
819
820static inline u32 sctp_hash_obj(const void *data, u32 len, u32 seed)
821{
822 const struct sctp_transport *t = data;
823 const union sctp_addr *paddr = &t->ipaddr;
824 const struct net *net = sock_net(t->asoc->base.sk);
825 u16 lport = htons(t->asoc->base.bind_addr.port);
826 u32 addr;
827
828 if (paddr->sa.sa_family == AF_INET6)
829 addr = jhash(&paddr->v6.sin6_addr, 16, seed);
830 else
831 addr = paddr->v4.sin_addr.s_addr;
832
833 return jhash_3words(addr, ((__u32)paddr->v4.sin_port) << 16 |
834 (__force __u32)lport, net_hash_mix(net), seed);
835}
836
837static inline u32 sctp_hash_key(const void *data, u32 len, u32 seed)
838{
839 const struct sctp_hash_cmp_arg *x = data;
840 const union sctp_addr *paddr = x->paddr;
841 const struct net *net = x->net;
842 u16 lport;
843 u32 addr;
844
845 lport = x->ep ? htons(x->ep->base.bind_addr.port) :
846 x->laddr->v4.sin_port;
847 if (paddr->sa.sa_family == AF_INET6)
848 addr = jhash(&paddr->v6.sin6_addr, 16, seed);
849 else
850 addr = paddr->v4.sin_addr.s_addr;
851
852 return jhash_3words(addr, ((__u32)paddr->v4.sin_port) << 16 |
853 (__force __u32)lport, net_hash_mix(net), seed);
854}
855
856static const struct rhashtable_params sctp_hash_params = {
857 .head_offset = offsetof(struct sctp_transport, node),
858 .hashfn = sctp_hash_key,
859 .obj_hashfn = sctp_hash_obj,
860 .obj_cmpfn = sctp_hash_cmp,
861 .automatic_shrinking = true,
862};
863
864int sctp_transport_hashtable_init(void)
865{
866 return rhashtable_init(&sctp_transport_hashtable, &sctp_hash_params);
867}
868
869void sctp_transport_hashtable_destroy(void)
870{
871 rhashtable_destroy(&sctp_transport_hashtable);
872}
873
874void sctp_hash_transport(struct sctp_transport *t)
875{
876 struct sctp_hash_cmp_arg arg;
877
878 if (t->asoc->temp)
879 return;
880
881 arg.ep = t->asoc->ep;
882 arg.paddr = &t->ipaddr;
883 arg.net = sock_net(t->asoc->base.sk);
884
885reinsert:
886 if (rhashtable_lookup_insert_key(&sctp_transport_hashtable, &arg,
887 &t->node, sctp_hash_params) == -EBUSY)
888 goto reinsert;
889}
890
891void sctp_unhash_transport(struct sctp_transport *t)
892{
893 if (t->asoc->temp)
894 return;
895
896 rhashtable_remove_fast(&sctp_transport_hashtable, &t->node,
897 sctp_hash_params);
898}
899
900struct sctp_transport *sctp_addrs_lookup_transport(
901 struct net *net,
902 const union sctp_addr *laddr,
903 const union sctp_addr *paddr)
904{
905 struct sctp_hash_cmp_arg arg = {
906 .ep = NULL,
907 .laddr = laddr,
908 .paddr = paddr,
909 .net = net,
910 };
911
912 return rhashtable_lookup_fast(&sctp_transport_hashtable, &arg,
913 sctp_hash_params);
914}
915
916struct sctp_transport *sctp_epaddr_lookup_transport(
917 const struct sctp_endpoint *ep,
918 const union sctp_addr *paddr)
919{
920 struct net *net = sock_net(ep->base.sk);
921 struct sctp_hash_cmp_arg arg = {
922 .ep = ep,
923 .paddr = paddr,
924 .net = net,
925 };
926
927 return rhashtable_lookup_fast(&sctp_transport_hashtable, &arg,
928 sctp_hash_params);
929}
930
931/* Look up an association. */
932static struct sctp_association *__sctp_lookup_association(
933 struct net *net,
934 const union sctp_addr *local,
935 const union sctp_addr *peer,
936 struct sctp_transport **pt)
937{
938 struct sctp_transport *t;
939 struct sctp_association *asoc = NULL;
940
941 t = sctp_addrs_lookup_transport(net, local, peer);
942 if (!t || !sctp_transport_hold(t))
943 goto out;
944
945 asoc = t->asoc;
946 sctp_association_hold(asoc);
947 *pt = t;
948
949 sctp_transport_put(t);
950
951out:
952 return asoc;
953}
954
955/* Look up an association. protected by RCU read lock */
956static
957struct sctp_association *sctp_lookup_association(struct net *net,
958 const union sctp_addr *laddr,
959 const union sctp_addr *paddr,
960 struct sctp_transport **transportp)
961{
962 struct sctp_association *asoc;
963
964 rcu_read_lock();
965 asoc = __sctp_lookup_association(net, laddr, paddr, transportp);
966 rcu_read_unlock();
967
968 return asoc;
969}
970
971/* Is there an association matching the given local and peer addresses? */
972int sctp_has_association(struct net *net,
973 const union sctp_addr *laddr,
974 const union sctp_addr *paddr)
975{
976 struct sctp_association *asoc;
977 struct sctp_transport *transport;
978
979 if ((asoc = sctp_lookup_association(net, laddr, paddr, &transport))) {
980 sctp_association_put(asoc);
981 return 1;
982 }
983
984 return 0;
985}
986
987/*
988 * SCTP Implementors Guide, 2.18 Handling of address
989 * parameters within the INIT or INIT-ACK.
990 *
991 * D) When searching for a matching TCB upon reception of an INIT
992 * or INIT-ACK chunk the receiver SHOULD use not only the
993 * source address of the packet (containing the INIT or
994 * INIT-ACK) but the receiver SHOULD also use all valid
995 * address parameters contained within the chunk.
996 *
997 * 2.18.3 Solution description
998 *
999 * This new text clearly specifies to an implementor the need
1000 * to look within the INIT or INIT-ACK. Any implementation that
1001 * does not do this, may not be able to establish associations
1002 * in certain circumstances.
1003 *
1004 */
1005static struct sctp_association *__sctp_rcv_init_lookup(struct net *net,
1006 struct sk_buff *skb,
1007 const union sctp_addr *laddr, struct sctp_transport **transportp)
1008{
1009 struct sctp_association *asoc;
1010 union sctp_addr addr;
1011 union sctp_addr *paddr = &addr;
1012 struct sctphdr *sh = sctp_hdr(skb);
1013 union sctp_params params;
1014 sctp_init_chunk_t *init;
1015 struct sctp_transport *transport;
1016 struct sctp_af *af;
1017
1018 /*
1019 * This code will NOT touch anything inside the chunk--it is
1020 * strictly READ-ONLY.
1021 *
1022 * RFC 2960 3 SCTP packet Format
1023 *
1024 * Multiple chunks can be bundled into one SCTP packet up to
1025 * the MTU size, except for the INIT, INIT ACK, and SHUTDOWN
1026 * COMPLETE chunks. These chunks MUST NOT be bundled with any
1027 * other chunk in a packet. See Section 6.10 for more details
1028 * on chunk bundling.
1029 */
1030
1031 /* Find the start of the TLVs and the end of the chunk. This is
1032 * the region we search for address parameters.
1033 */
1034 init = (sctp_init_chunk_t *)skb->data;
1035
1036 /* Walk the parameters looking for embedded addresses. */
1037 sctp_walk_params(params, init, init_hdr.params) {
1038
1039 /* Note: Ignoring hostname addresses. */
1040 af = sctp_get_af_specific(param_type2af(params.p->type));
1041 if (!af)
1042 continue;
1043
1044 af->from_addr_param(paddr, params.addr, sh->source, 0);
1045
1046 asoc = __sctp_lookup_association(net, laddr, paddr, &transport);
1047 if (asoc)
1048 return asoc;
1049 }
1050
1051 return NULL;
1052}
1053
1054/* ADD-IP, Section 5.2
1055 * When an endpoint receives an ASCONF Chunk from the remote peer
1056 * special procedures may be needed to identify the association the
1057 * ASCONF Chunk is associated with. To properly find the association
1058 * the following procedures SHOULD be followed:
1059 *
1060 * D2) If the association is not found, use the address found in the
1061 * Address Parameter TLV combined with the port number found in the
1062 * SCTP common header. If found proceed to rule D4.
1063 *
1064 * D2-ext) If more than one ASCONF Chunks are packed together, use the
1065 * address found in the ASCONF Address Parameter TLV of each of the
1066 * subsequent ASCONF Chunks. If found, proceed to rule D4.
1067 */
1068static struct sctp_association *__sctp_rcv_asconf_lookup(
1069 struct net *net,
1070 sctp_chunkhdr_t *ch,
1071 const union sctp_addr *laddr,
1072 __be16 peer_port,
1073 struct sctp_transport **transportp)
1074{
1075 sctp_addip_chunk_t *asconf = (struct sctp_addip_chunk *)ch;
1076 struct sctp_af *af;
1077 union sctp_addr_param *param;
1078 union sctp_addr paddr;
1079
1080 /* Skip over the ADDIP header and find the Address parameter */
1081 param = (union sctp_addr_param *)(asconf + 1);
1082
1083 af = sctp_get_af_specific(param_type2af(param->p.type));
1084 if (unlikely(!af))
1085 return NULL;
1086
1087 af->from_addr_param(&paddr, param, peer_port, 0);
1088
1089 return __sctp_lookup_association(net, laddr, &paddr, transportp);
1090}
1091
1092
1093/* SCTP-AUTH, Section 6.3:
1094* If the receiver does not find a STCB for a packet containing an AUTH
1095* chunk as the first chunk and not a COOKIE-ECHO chunk as the second
1096* chunk, it MUST use the chunks after the AUTH chunk to look up an existing
1097* association.
1098*
1099* This means that any chunks that can help us identify the association need
1100* to be looked at to find this association.
1101*/
1102static struct sctp_association *__sctp_rcv_walk_lookup(struct net *net,
1103 struct sk_buff *skb,
1104 const union sctp_addr *laddr,
1105 struct sctp_transport **transportp)
1106{
1107 struct sctp_association *asoc = NULL;
1108 sctp_chunkhdr_t *ch;
1109 int have_auth = 0;
1110 unsigned int chunk_num = 1;
1111 __u8 *ch_end;
1112
1113 /* Walk through the chunks looking for AUTH or ASCONF chunks
1114 * to help us find the association.
1115 */
1116 ch = (sctp_chunkhdr_t *) skb->data;
1117 do {
1118 /* Break out if chunk length is less then minimal. */
1119 if (ntohs(ch->length) < sizeof(sctp_chunkhdr_t))
1120 break;
1121
1122 ch_end = ((__u8 *)ch) + WORD_ROUND(ntohs(ch->length));
1123 if (ch_end > skb_tail_pointer(skb))
1124 break;
1125
1126 switch (ch->type) {
1127 case SCTP_CID_AUTH:
1128 have_auth = chunk_num;
1129 break;
1130
1131 case SCTP_CID_COOKIE_ECHO:
1132 /* If a packet arrives containing an AUTH chunk as
1133 * a first chunk, a COOKIE-ECHO chunk as the second
1134 * chunk, and possibly more chunks after them, and
1135 * the receiver does not have an STCB for that
1136 * packet, then authentication is based on
1137 * the contents of the COOKIE- ECHO chunk.
1138 */
1139 if (have_auth == 1 && chunk_num == 2)
1140 return NULL;
1141 break;
1142
1143 case SCTP_CID_ASCONF:
1144 if (have_auth || net->sctp.addip_noauth)
1145 asoc = __sctp_rcv_asconf_lookup(
1146 net, ch, laddr,
1147 sctp_hdr(skb)->source,
1148 transportp);
1149 default:
1150 break;
1151 }
1152
1153 if (asoc)
1154 break;
1155
1156 ch = (sctp_chunkhdr_t *) ch_end;
1157 chunk_num++;
1158 } while (ch_end < skb_tail_pointer(skb));
1159
1160 return asoc;
1161}
1162
1163/*
1164 * There are circumstances when we need to look inside the SCTP packet
1165 * for information to help us find the association. Examples
1166 * include looking inside of INIT/INIT-ACK chunks or after the AUTH
1167 * chunks.
1168 */
1169static struct sctp_association *__sctp_rcv_lookup_harder(struct net *net,
1170 struct sk_buff *skb,
1171 const union sctp_addr *laddr,
1172 struct sctp_transport **transportp)
1173{
1174 sctp_chunkhdr_t *ch;
1175
1176 ch = (sctp_chunkhdr_t *) skb->data;
1177
1178 /* The code below will attempt to walk the chunk and extract
1179 * parameter information. Before we do that, we need to verify
1180 * that the chunk length doesn't cause overflow. Otherwise, we'll
1181 * walk off the end.
1182 */
1183 if (WORD_ROUND(ntohs(ch->length)) > skb->len)
1184 return NULL;
1185
1186 /* If this is INIT/INIT-ACK look inside the chunk too. */
1187 if (ch->type == SCTP_CID_INIT || ch->type == SCTP_CID_INIT_ACK)
1188 return __sctp_rcv_init_lookup(net, skb, laddr, transportp);
1189
1190 return __sctp_rcv_walk_lookup(net, skb, laddr, transportp);
1191}
1192
1193/* Lookup an association for an inbound skb. */
1194static struct sctp_association *__sctp_rcv_lookup(struct net *net,
1195 struct sk_buff *skb,
1196 const union sctp_addr *paddr,
1197 const union sctp_addr *laddr,
1198 struct sctp_transport **transportp)
1199{
1200 struct sctp_association *asoc;
1201
1202 asoc = __sctp_lookup_association(net, laddr, paddr, transportp);
1203
1204 /* Further lookup for INIT/INIT-ACK packets.
1205 * SCTP Implementors Guide, 2.18 Handling of address
1206 * parameters within the INIT or INIT-ACK.
1207 */
1208 if (!asoc)
1209 asoc = __sctp_rcv_lookup_harder(net, skb, laddr, transportp);
1210
1211 return asoc;
1212}
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 <linux-sctp@vger.kernel.org>
17 *
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>
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_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))
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 if (!mod_timer(&t->proto_unreach_timer,
440 jiffies + (HZ/20)))
441 sctp_transport_hold(t);
442 }
443 } else {
444 struct net *net = sock_net(sk);
445
446 pr_debug("%s: unrecognized next header type "
447 "encountered!\n", __func__);
448
449 if (del_timer(&t->proto_unreach_timer))
450 sctp_transport_put(t);
451
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,
455 GFP_ATOMIC);
456 }
457}
458
459/* Common lookup code for icmp/icmpv6 error handler. */
460struct 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)
464{
465 struct sctp_init_chunk *chunkhdr, _chunkhdr;
466 union sctp_addr saddr;
467 union sctp_addr daddr;
468 struct sctp_af *af;
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);
475
476 *app = NULL; *tpp = NULL;
477
478 af = sctp_get_af_specific(family);
479 if (unlikely(!af)) {
480 return NULL;
481 }
482
483 /* Initialize local addresses for lookups. */
484 af->from_skb(&saddr, skb, 1);
485 af->from_skb(&daddr, skb, 0);
486
487 /* Look for an association that matches the incoming ICMP error
488 * packet.
489 */
490 asoc = __sctp_lookup_association(net, &saddr, &daddr, &transport, dif, sdif);
491 if (!asoc)
492 return NULL;
493
494 sk = asoc->base.sk;
495
496 /* RFC 4960, Appendix C. ICMP Handling
497 *
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.
507 */
508 if (vtag == 0) {
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);
514 if (!chunkhdr ||
515 chunkhdr->chunk_hdr.type != SCTP_CID_INIT ||
516 ntohl(chunkhdr->init_hdr.init_tag) != asoc->c.my_vtag)
517 goto out;
518
519 } else if (vtag != asoc->c.peer_vtag) {
520 goto out;
521 }
522
523 bh_lock_sock(sk);
524
525 /* If too many ICMPs get dropped on busy
526 * servers this needs to be solved differently.
527 */
528 if (sock_owned_by_user(sk))
529 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
530
531 *app = asoc;
532 *tpp = transport;
533 return sk;
534
535out:
536 sctp_transport_put(transport);
537 return NULL;
538}
539
540/* Common cleanup code for icmp/icmpv6 error handler. */
541void sctp_err_finish(struct sock *sk, struct sctp_transport *t)
542 __releases(&((__sk)->sk_lock.slock))
543{
544 bh_unlock_sock(sk);
545 sctp_transport_put(t);
546}
547
548static void sctp_v4_err_handle(struct sctp_transport *t, struct sk_buff *skb,
549 __u8 type, __u8 code, __u32 info)
550{
551 struct sctp_association *asoc = t->asoc;
552 struct sock *sk = asoc->base.sk;
553 int err = 0;
554
555 switch (type) {
556 case ICMP_PARAMETERPROB:
557 err = EPROTO;
558 break;
559 case ICMP_DEST_UNREACH:
560 if (code > NR_ICMP_UNREACH)
561 return;
562 if (code == ICMP_FRAG_NEEDED) {
563 sctp_icmp_frag_needed(sk, asoc, t, SCTP_TRUNC4(info));
564 return;
565 }
566 if (code == ICMP_PROT_UNREACH) {
567 sctp_icmp_proto_unreachable(sk, asoc, t);
568 return;
569 }
570 err = icmp_err_convert[code].errno;
571 break;
572 case ICMP_TIME_EXCEEDED:
573 if (code == ICMP_EXC_FRAGTIME)
574 return;
575
576 err = EHOSTUNREACH;
577 break;
578 case ICMP_REDIRECT:
579 sctp_icmp_redirect(sk, t, skb);
580 return;
581 default:
582 return;
583 }
584 if (!sock_owned_by_user(sk) && inet_test_bit(RECVERR, sk)) {
585 sk->sk_err = err;
586 sk_error_report(sk);
587 } else { /* Only an error on timeout */
588 WRITE_ONCE(sk->sk_err_soft, err);
589 }
590}
591
592/*
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.
599 *
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.
605 *
606 */
607int sctp_v4_err(struct sk_buff *skb, __u32 info)
608{
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;
616 struct sock *sk;
617
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;
627 if (!sk) {
628 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
629 return -ENOENT;
630 }
631
632 sctp_v4_err_handle(transport, skb, type, code, info);
633 sctp_err_finish(sk, transport);
634
635 return 0;
636}
637
638int sctp_udp_v4_err(struct sock *sk, struct sk_buff *skb)
639{
640 struct net *net = dev_net(skb->dev);
641 struct sctp_association *asoc;
642 struct sctp_transport *t;
643 struct icmphdr *hdr;
644 __u32 info = 0;
645
646 skb->transport_header += sizeof(struct udphdr);
647 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &t);
648 if (!sk) {
649 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
650 return -ENOENT;
651 }
652
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);
658 return 0;
659 }
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);
663
664 sctp_err_finish(sk, t);
665 return 1;
666}
667
668/*
669 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
670 *
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.
675 *
676 * Output:
677 * Return 0 - If further processing is needed.
678 * Return 1 - If the packet can be discarded right away.
679 */
680static int sctp_rcv_ootb(struct sk_buff *skb)
681{
682 struct sctp_chunkhdr *ch, _ch;
683 int ch_end, offset = 0;
684
685 /* Scan through all the chunks in the packet. */
686 do {
687 /* Make sure we have at least the header there */
688 if (offset + sizeof(_ch) > skb->len)
689 break;
690
691 ch = skb_header_pointer(skb, offset, sizeof(*ch), &_ch);
692
693 /* Break out if chunk length is less then minimal. */
694 if (!ch || ntohs(ch->length) < sizeof(_ch))
695 break;
696
697 ch_end = offset + SCTP_PAD4(ntohs(ch->length));
698 if (ch_end > skb->len)
699 break;
700
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
703 * further action.
704 */
705 if (SCTP_CID_ABORT == ch->type)
706 goto discard;
707
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.
711 */
712 if (SCTP_CID_SHUTDOWN_COMPLETE == ch->type)
713 goto discard;
714
715 /* RFC 4460, 2.11.2
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.
719 */
720 if (SCTP_CID_INIT == ch->type && (void *)ch != skb->data)
721 goto discard;
722
723 offset = ch_end;
724 } while (ch_end < skb->len);
725
726 return 0;
727
728discard:
729 return 1;
730}
731
732/* Insert endpoint into the hash table. */
733static int __sctp_hash_endpoint(struct sctp_endpoint *ep)
734{
735 struct sock *sk = ep->base.sk;
736 struct net *net = sock_net(sk);
737 struct sctp_hashbucket *head;
738 int err = 0;
739
740 ep->hashent = sctp_ep_hashfn(net, ep->base.bind_addr.port);
741 head = &sctp_ep_hashtable[ep->hashent];
742
743 write_lock(&head->lock);
744 if (sk->sk_reuseport) {
745 bool any = sctp_is_ep_boundall(sk);
746 struct sctp_endpoint *ep2;
747 struct list_head *list;
748 int cnt = 0;
749
750 err = 1;
751
752 list_for_each(list, &ep->base.bind_addr.address_list)
753 cnt++;
754
755 sctp_for_each_hentry(ep2, &head->chain) {
756 struct sock *sk2 = ep2->base.sk;
757
758 if (!net_eq(sock_net(sk2), net) || sk2 == sk ||
759 !uid_eq(sock_i_uid(sk2), sock_i_uid(sk)) ||
760 !sk2->sk_reuseport)
761 continue;
762
763 err = sctp_bind_addrs_check(sctp_sk(sk2),
764 sctp_sk(sk), cnt);
765 if (!err) {
766 err = reuseport_add_sock(sk, sk2, any);
767 if (err)
768 goto out;
769 break;
770 } else if (err < 0) {
771 goto out;
772 }
773 }
774
775 if (err) {
776 err = reuseport_alloc(sk, any);
777 if (err)
778 goto out;
779 }
780 }
781
782 hlist_add_head(&ep->node, &head->chain);
783out:
784 write_unlock(&head->lock);
785 return err;
786}
787
788/* Add an endpoint to the hash. Local BH-safe. */
789int sctp_hash_endpoint(struct sctp_endpoint *ep)
790{
791 int err;
792
793 local_bh_disable();
794 err = __sctp_hash_endpoint(ep);
795 local_bh_enable();
796
797 return err;
798}
799
800/* Remove endpoint from the hash table. */
801static void __sctp_unhash_endpoint(struct sctp_endpoint *ep)
802{
803 struct sock *sk = ep->base.sk;
804 struct sctp_hashbucket *head;
805
806 ep->hashent = sctp_ep_hashfn(sock_net(sk), ep->base.bind_addr.port);
807
808 head = &sctp_ep_hashtable[ep->hashent];
809
810 write_lock(&head->lock);
811 if (rcu_access_pointer(sk->sk_reuseport_cb))
812 reuseport_detach_sock(sk);
813 hlist_del_init(&ep->node);
814 write_unlock(&head->lock);
815}
816
817/* Remove endpoint from the hash. Local BH-safe. */
818void sctp_unhash_endpoint(struct sctp_endpoint *ep)
819{
820 local_bh_disable();
821 __sctp_unhash_endpoint(ep);
822 local_bh_enable();
823}
824
825static inline __u32 sctp_hashfn(const struct net *net, __be16 lport,
826 const union sctp_addr *paddr, __u32 seed)
827{
828 __u32 addr;
829
830 if (paddr->sa.sa_family == AF_INET6)
831 addr = jhash(&paddr->v6.sin6_addr, 16, seed);
832 else
833 addr = (__force __u32)paddr->v4.sin_addr.s_addr;
834
835 return jhash_3words(addr, ((__force __u32)paddr->v4.sin_port) << 16 |
836 (__force __u32)lport, net_hash_mix(net), seed);
837}
838
839/* Look up an endpoint. */
840static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(
841 struct net *net, struct sk_buff *skb,
842 const union sctp_addr *laddr,
843 const union sctp_addr *paddr,
844 int dif, int sdif)
845{
846 struct sctp_hashbucket *head;
847 struct sctp_endpoint *ep;
848 struct sock *sk;
849 __be16 lport;
850 int hash;
851
852 lport = laddr->v4.sin_port;
853 hash = sctp_ep_hashfn(net, ntohs(lport));
854 head = &sctp_ep_hashtable[hash];
855 read_lock(&head->lock);
856 sctp_for_each_hentry(ep, &head->chain) {
857 if (sctp_endpoint_is_match(ep, net, laddr, dif, sdif))
858 goto hit;
859 }
860
861 ep = sctp_sk(net->sctp.ctl_sock)->ep;
862
863hit:
864 sk = ep->base.sk;
865 if (sk->sk_reuseport) {
866 __u32 phash = sctp_hashfn(net, lport, paddr, 0);
867
868 sk = reuseport_select_sock(sk, phash, skb,
869 sizeof(struct sctphdr));
870 if (sk)
871 ep = sctp_sk(sk)->ep;
872 }
873 sctp_endpoint_hold(ep);
874 read_unlock(&head->lock);
875 return ep;
876}
877
878/* rhashtable for transport */
879struct sctp_hash_cmp_arg {
880 const union sctp_addr *paddr;
881 const struct net *net;
882 __be16 lport;
883};
884
885static inline int sctp_hash_cmp(struct rhashtable_compare_arg *arg,
886 const void *ptr)
887{
888 struct sctp_transport *t = (struct sctp_transport *)ptr;
889 const struct sctp_hash_cmp_arg *x = arg->key;
890 int err = 1;
891
892 if (!sctp_cmp_addr_exact(&t->ipaddr, x->paddr))
893 return err;
894 if (!sctp_transport_hold(t))
895 return err;
896
897 if (!net_eq(t->asoc->base.net, x->net))
898 goto out;
899 if (x->lport != htons(t->asoc->base.bind_addr.port))
900 goto out;
901
902 err = 0;
903out:
904 sctp_transport_put(t);
905 return err;
906}
907
908static inline __u32 sctp_hash_obj(const void *data, u32 len, u32 seed)
909{
910 const struct sctp_transport *t = data;
911
912 return sctp_hashfn(t->asoc->base.net,
913 htons(t->asoc->base.bind_addr.port),
914 &t->ipaddr, seed);
915}
916
917static inline __u32 sctp_hash_key(const void *data, u32 len, u32 seed)
918{
919 const struct sctp_hash_cmp_arg *x = data;
920
921 return sctp_hashfn(x->net, x->lport, x->paddr, seed);
922}
923
924static const struct rhashtable_params sctp_hash_params = {
925 .head_offset = offsetof(struct sctp_transport, node),
926 .hashfn = sctp_hash_key,
927 .obj_hashfn = sctp_hash_obj,
928 .obj_cmpfn = sctp_hash_cmp,
929 .automatic_shrinking = true,
930};
931
932int sctp_transport_hashtable_init(void)
933{
934 return rhltable_init(&sctp_transport_hashtable, &sctp_hash_params);
935}
936
937void sctp_transport_hashtable_destroy(void)
938{
939 rhltable_destroy(&sctp_transport_hashtable);
940}
941
942int sctp_hash_transport(struct sctp_transport *t)
943{
944 struct sctp_transport *transport;
945 struct rhlist_head *tmp, *list;
946 struct sctp_hash_cmp_arg arg;
947 int err;
948
949 if (t->asoc->temp)
950 return 0;
951
952 arg.net = t->asoc->base.net;
953 arg.paddr = &t->ipaddr;
954 arg.lport = htons(t->asoc->base.bind_addr.port);
955
956 rcu_read_lock();
957 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
958 sctp_hash_params);
959
960 rhl_for_each_entry_rcu(transport, tmp, list, node)
961 if (transport->asoc->ep == t->asoc->ep) {
962 rcu_read_unlock();
963 return -EEXIST;
964 }
965 rcu_read_unlock();
966
967 err = rhltable_insert_key(&sctp_transport_hashtable, &arg,
968 &t->node, sctp_hash_params);
969 if (err)
970 pr_err_once("insert transport fail, errno %d\n", err);
971
972 return err;
973}
974
975void sctp_unhash_transport(struct sctp_transport *t)
976{
977 if (t->asoc->temp)
978 return;
979
980 rhltable_remove(&sctp_transport_hashtable, &t->node,
981 sctp_hash_params);
982}
983
984bool sctp_sk_bound_dev_eq(struct net *net, int bound_dev_if, int dif, int sdif)
985{
986 bool l3mdev_accept = true;
987
988#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
989 l3mdev_accept = !!READ_ONCE(net->sctp.l3mdev_accept);
990#endif
991 return inet_bound_dev_eq(l3mdev_accept, bound_dev_if, dif, sdif);
992}
993
994/* return a transport with holding it */
995struct sctp_transport *sctp_addrs_lookup_transport(
996 struct net *net,
997 const union sctp_addr *laddr,
998 const union sctp_addr *paddr,
999 int dif, int sdif)
1000{
1001 struct rhlist_head *tmp, *list;
1002 struct sctp_transport *t;
1003 int bound_dev_if;
1004 struct sctp_hash_cmp_arg arg = {
1005 .paddr = paddr,
1006 .net = net,
1007 .lport = laddr->v4.sin_port,
1008 };
1009
1010 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
1011 sctp_hash_params);
1012
1013 rhl_for_each_entry_rcu(t, tmp, list, node) {
1014 if (!sctp_transport_hold(t))
1015 continue;
1016
1017 bound_dev_if = READ_ONCE(t->asoc->base.sk->sk_bound_dev_if);
1018 if (sctp_sk_bound_dev_eq(net, bound_dev_if, dif, sdif) &&
1019 sctp_bind_addr_match(&t->asoc->base.bind_addr,
1020 laddr, sctp_sk(t->asoc->base.sk)))
1021 return t;
1022 sctp_transport_put(t);
1023 }
1024
1025 return NULL;
1026}
1027
1028/* return a transport without holding it, as it's only used under sock lock */
1029struct sctp_transport *sctp_epaddr_lookup_transport(
1030 const struct sctp_endpoint *ep,
1031 const union sctp_addr *paddr)
1032{
1033 struct rhlist_head *tmp, *list;
1034 struct sctp_transport *t;
1035 struct sctp_hash_cmp_arg arg = {
1036 .paddr = paddr,
1037 .net = ep->base.net,
1038 .lport = htons(ep->base.bind_addr.port),
1039 };
1040
1041 list = rhltable_lookup(&sctp_transport_hashtable, &arg,
1042 sctp_hash_params);
1043
1044 rhl_for_each_entry_rcu(t, tmp, list, node)
1045 if (ep == t->asoc->ep)
1046 return t;
1047
1048 return NULL;
1049}
1050
1051/* Look up an association. */
1052static struct sctp_association *__sctp_lookup_association(
1053 struct net *net,
1054 const union sctp_addr *local,
1055 const union sctp_addr *peer,
1056 struct sctp_transport **pt,
1057 int dif, int sdif)
1058{
1059 struct sctp_transport *t;
1060 struct sctp_association *asoc = NULL;
1061
1062 t = sctp_addrs_lookup_transport(net, local, peer, dif, sdif);
1063 if (!t)
1064 goto out;
1065
1066 asoc = t->asoc;
1067 *pt = t;
1068
1069out:
1070 return asoc;
1071}
1072
1073/* Look up an association. protected by RCU read lock */
1074static
1075struct sctp_association *sctp_lookup_association(struct net *net,
1076 const union sctp_addr *laddr,
1077 const union sctp_addr *paddr,
1078 struct sctp_transport **transportp,
1079 int dif, int sdif)
1080{
1081 struct sctp_association *asoc;
1082
1083 rcu_read_lock();
1084 asoc = __sctp_lookup_association(net, laddr, paddr, transportp, dif, sdif);
1085 rcu_read_unlock();
1086
1087 return asoc;
1088}
1089
1090/* Is there an association matching the given local and peer addresses? */
1091bool sctp_has_association(struct net *net,
1092 const union sctp_addr *laddr,
1093 const union sctp_addr *paddr,
1094 int dif, int sdif)
1095{
1096 struct sctp_transport *transport;
1097
1098 if (sctp_lookup_association(net, laddr, paddr, &transport, dif, sdif)) {
1099 sctp_transport_put(transport);
1100 return true;
1101 }
1102
1103 return false;
1104}
1105
1106/*
1107 * SCTP Implementors Guide, 2.18 Handling of address
1108 * parameters within the INIT or INIT-ACK.
1109 *
1110 * D) When searching for a matching TCB upon reception of an INIT
1111 * or INIT-ACK chunk the receiver SHOULD use not only the
1112 * source address of the packet (containing the INIT or
1113 * INIT-ACK) but the receiver SHOULD also use all valid
1114 * address parameters contained within the chunk.
1115 *
1116 * 2.18.3 Solution description
1117 *
1118 * This new text clearly specifies to an implementor the need
1119 * to look within the INIT or INIT-ACK. Any implementation that
1120 * does not do this, may not be able to establish associations
1121 * in certain circumstances.
1122 *
1123 */
1124static struct sctp_association *__sctp_rcv_init_lookup(struct net *net,
1125 struct sk_buff *skb,
1126 const union sctp_addr *laddr, struct sctp_transport **transportp,
1127 int dif, int sdif)
1128{
1129 struct sctp_association *asoc;
1130 union sctp_addr addr;
1131 union sctp_addr *paddr = &addr;
1132 struct sctphdr *sh = sctp_hdr(skb);
1133 union sctp_params params;
1134 struct sctp_init_chunk *init;
1135 struct sctp_af *af;
1136
1137 /*
1138 * This code will NOT touch anything inside the chunk--it is
1139 * strictly READ-ONLY.
1140 *
1141 * RFC 2960 3 SCTP packet Format
1142 *
1143 * Multiple chunks can be bundled into one SCTP packet up to
1144 * the MTU size, except for the INIT, INIT ACK, and SHUTDOWN
1145 * COMPLETE chunks. These chunks MUST NOT be bundled with any
1146 * other chunk in a packet. See Section 6.10 for more details
1147 * on chunk bundling.
1148 */
1149
1150 /* Find the start of the TLVs and the end of the chunk. This is
1151 * the region we search for address parameters.
1152 */
1153 init = (struct sctp_init_chunk *)skb->data;
1154
1155 /* Walk the parameters looking for embedded addresses. */
1156 sctp_walk_params(params, init) {
1157
1158 /* Note: Ignoring hostname addresses. */
1159 af = sctp_get_af_specific(param_type2af(params.p->type));
1160 if (!af)
1161 continue;
1162
1163 if (!af->from_addr_param(paddr, params.addr, sh->source, 0))
1164 continue;
1165
1166 asoc = __sctp_lookup_association(net, laddr, paddr, transportp, dif, sdif);
1167 if (asoc)
1168 return asoc;
1169 }
1170
1171 return NULL;
1172}
1173
1174/* ADD-IP, Section 5.2
1175 * When an endpoint receives an ASCONF Chunk from the remote peer
1176 * special procedures may be needed to identify the association the
1177 * ASCONF Chunk is associated with. To properly find the association
1178 * the following procedures SHOULD be followed:
1179 *
1180 * D2) If the association is not found, use the address found in the
1181 * Address Parameter TLV combined with the port number found in the
1182 * SCTP common header. If found proceed to rule D4.
1183 *
1184 * D2-ext) If more than one ASCONF Chunks are packed together, use the
1185 * address found in the ASCONF Address Parameter TLV of each of the
1186 * subsequent ASCONF Chunks. If found, proceed to rule D4.
1187 */
1188static struct sctp_association *__sctp_rcv_asconf_lookup(
1189 struct net *net,
1190 struct sctp_chunkhdr *ch,
1191 const union sctp_addr *laddr,
1192 __be16 peer_port,
1193 struct sctp_transport **transportp,
1194 int dif, int sdif)
1195{
1196 struct sctp_addip_chunk *asconf = (struct sctp_addip_chunk *)ch;
1197 struct sctp_af *af;
1198 union sctp_addr_param *param;
1199 union sctp_addr paddr;
1200
1201 if (ntohs(ch->length) < sizeof(*asconf) + sizeof(struct sctp_paramhdr))
1202 return NULL;
1203
1204 /* Skip over the ADDIP header and find the Address parameter */
1205 param = (union sctp_addr_param *)(asconf + 1);
1206
1207 af = sctp_get_af_specific(param_type2af(param->p.type));
1208 if (unlikely(!af))
1209 return NULL;
1210
1211 if (!af->from_addr_param(&paddr, param, peer_port, 0))
1212 return NULL;
1213
1214 return __sctp_lookup_association(net, laddr, &paddr, transportp, dif, sdif);
1215}
1216
1217
1218/* SCTP-AUTH, Section 6.3:
1219* If the receiver does not find a STCB for a packet containing an AUTH
1220* chunk as the first chunk and not a COOKIE-ECHO chunk as the second
1221* chunk, it MUST use the chunks after the AUTH chunk to look up an existing
1222* association.
1223*
1224* This means that any chunks that can help us identify the association need
1225* to be looked at to find this association.
1226*/
1227static struct sctp_association *__sctp_rcv_walk_lookup(struct net *net,
1228 struct sk_buff *skb,
1229 const union sctp_addr *laddr,
1230 struct sctp_transport **transportp,
1231 int dif, int sdif)
1232{
1233 struct sctp_association *asoc = NULL;
1234 struct sctp_chunkhdr *ch;
1235 int have_auth = 0;
1236 unsigned int chunk_num = 1;
1237 __u8 *ch_end;
1238
1239 /* Walk through the chunks looking for AUTH or ASCONF chunks
1240 * to help us find the association.
1241 */
1242 ch = (struct sctp_chunkhdr *)skb->data;
1243 do {
1244 /* Break out if chunk length is less then minimal. */
1245 if (ntohs(ch->length) < sizeof(*ch))
1246 break;
1247
1248 ch_end = ((__u8 *)ch) + SCTP_PAD4(ntohs(ch->length));
1249 if (ch_end > skb_tail_pointer(skb))
1250 break;
1251
1252 switch (ch->type) {
1253 case SCTP_CID_AUTH:
1254 have_auth = chunk_num;
1255 break;
1256
1257 case SCTP_CID_COOKIE_ECHO:
1258 /* If a packet arrives containing an AUTH chunk as
1259 * a first chunk, a COOKIE-ECHO chunk as the second
1260 * chunk, and possibly more chunks after them, and
1261 * the receiver does not have an STCB for that
1262 * packet, then authentication is based on
1263 * the contents of the COOKIE- ECHO chunk.
1264 */
1265 if (have_auth == 1 && chunk_num == 2)
1266 return NULL;
1267 break;
1268
1269 case SCTP_CID_ASCONF:
1270 if (have_auth || net->sctp.addip_noauth)
1271 asoc = __sctp_rcv_asconf_lookup(
1272 net, ch, laddr,
1273 sctp_hdr(skb)->source,
1274 transportp, dif, sdif);
1275 break;
1276 default:
1277 break;
1278 }
1279
1280 if (asoc)
1281 break;
1282
1283 ch = (struct sctp_chunkhdr *)ch_end;
1284 chunk_num++;
1285 } while (ch_end + sizeof(*ch) < skb_tail_pointer(skb));
1286
1287 return asoc;
1288}
1289
1290/*
1291 * There are circumstances when we need to look inside the SCTP packet
1292 * for information to help us find the association. Examples
1293 * include looking inside of INIT/INIT-ACK chunks or after the AUTH
1294 * chunks.
1295 */
1296static struct sctp_association *__sctp_rcv_lookup_harder(struct net *net,
1297 struct sk_buff *skb,
1298 const union sctp_addr *laddr,
1299 struct sctp_transport **transportp,
1300 int dif, int sdif)
1301{
1302 struct sctp_chunkhdr *ch;
1303
1304 /* We do not allow GSO frames here as we need to linearize and
1305 * then cannot guarantee frame boundaries. This shouldn't be an
1306 * issue as packets hitting this are mostly INIT or INIT-ACK and
1307 * those cannot be on GSO-style anyway.
1308 */
1309 if (skb_is_gso(skb) && skb_is_gso_sctp(skb))
1310 return NULL;
1311
1312 ch = (struct sctp_chunkhdr *)skb->data;
1313
1314 /* The code below will attempt to walk the chunk and extract
1315 * parameter information. Before we do that, we need to verify
1316 * that the chunk length doesn't cause overflow. Otherwise, we'll
1317 * walk off the end.
1318 */
1319 if (SCTP_PAD4(ntohs(ch->length)) > skb->len)
1320 return NULL;
1321
1322 /* If this is INIT/INIT-ACK look inside the chunk too. */
1323 if (ch->type == SCTP_CID_INIT || ch->type == SCTP_CID_INIT_ACK)
1324 return __sctp_rcv_init_lookup(net, skb, laddr, transportp, dif, sdif);
1325
1326 return __sctp_rcv_walk_lookup(net, skb, laddr, transportp, dif, sdif);
1327}
1328
1329/* Lookup an association for an inbound skb. */
1330static struct sctp_association *__sctp_rcv_lookup(struct net *net,
1331 struct sk_buff *skb,
1332 const union sctp_addr *paddr,
1333 const union sctp_addr *laddr,
1334 struct sctp_transport **transportp,
1335 int dif, int sdif)
1336{
1337 struct sctp_association *asoc;
1338
1339 asoc = __sctp_lookup_association(net, laddr, paddr, transportp, dif, sdif);
1340 if (asoc)
1341 goto out;
1342
1343 /* Further lookup for INIT/INIT-ACK packets.
1344 * SCTP Implementors Guide, 2.18 Handling of address
1345 * parameters within the INIT or INIT-ACK.
1346 */
1347 asoc = __sctp_rcv_lookup_harder(net, skb, laddr, transportp, dif, sdif);
1348 if (asoc)
1349 goto out;
1350
1351 if (paddr->sa.sa_family == AF_INET)
1352 pr_debug("sctp: asoc not found for src:%pI4:%d dst:%pI4:%d\n",
1353 &laddr->v4.sin_addr, ntohs(laddr->v4.sin_port),
1354 &paddr->v4.sin_addr, ntohs(paddr->v4.sin_port));
1355 else
1356 pr_debug("sctp: asoc not found for src:%pI6:%d dst:%pI6:%d\n",
1357 &laddr->v6.sin6_addr, ntohs(laddr->v6.sin6_port),
1358 &paddr->v6.sin6_addr, ntohs(paddr->v6.sin6_port));
1359
1360out:
1361 return asoc;
1362}