blob: 6d82f400d13c47a36279f861cc2efed80d105254 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/* SCTP kernel reference 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 reference Implementation
10 *
11 * These functions handle all input from the IP layer into SCTP.
12 *
13 * The SCTP reference 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 * The SCTP reference 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, write to
27 * the Free Software Foundation, 59 Temple Place - Suite 330,
28 * Boston, MA 02111-1307, USA.
29 *
30 * Please send any bug reports or fixes you make to the
31 * email address(es):
32 * lksctp developers <lksctp-developers@lists.sourceforge.net>
33 *
34 * Or submit a bug report through the following website:
35 * http://www.sf.net/projects/lksctp
36 *
37 * Written or modified by:
38 * La Monte H.P. Yarroll <piggy@acm.org>
39 * Karl Knutson <karl@athena.chicago.il.us>
40 * Xingang Guo <xingang.guo@intel.com>
41 * Jon Grimm <jgrimm@us.ibm.com>
42 * Hui Huang <hui.huang@nokia.com>
43 * Daisy Chang <daisyc@us.ibm.com>
44 * Sridhar Samudrala <sri@us.ibm.com>
45 * Ardelle Fan <ardelle.fan@intel.com>
46 *
47 * Any bugs reported given to us we will try to fix... any fixes shared will
48 * be incorporated into the next SCTP release.
49 */
50
51#include <linux/types.h>
52#include <linux/list.h> /* For struct list_head */
53#include <linux/socket.h>
54#include <linux/ip.h>
55#include <linux/time.h> /* For struct timeval */
56#include <net/ip.h>
57#include <net/icmp.h>
58#include <net/snmp.h>
59#include <net/sock.h>
60#include <net/xfrm.h>
61#include <net/sctp/sctp.h>
62#include <net/sctp/sm.h>
63
64/* Forward declarations for internal helpers. */
65static int sctp_rcv_ootb(struct sk_buff *);
66static struct sctp_association *__sctp_rcv_lookup(struct sk_buff *skb,
67 const union sctp_addr *laddr,
68 const union sctp_addr *paddr,
69 struct sctp_transport **transportp);
70static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(const union sctp_addr *laddr);
71static struct sctp_association *__sctp_lookup_association(
72 const union sctp_addr *local,
73 const union sctp_addr *peer,
74 struct sctp_transport **pt);
75
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -070076static void sctp_add_backlog(struct sock *sk, struct sk_buff *skb);
77
Linus Torvalds1da177e2005-04-16 15:20:36 -070078
79/* Calculate the SCTP checksum of an SCTP packet. */
80static inline int sctp_rcv_checksum(struct sk_buff *skb)
81{
82 struct sctphdr *sh;
83 __u32 cmp, val;
84 struct sk_buff *list = skb_shinfo(skb)->frag_list;
85
86 sh = (struct sctphdr *) skb->h.raw;
87 cmp = ntohl(sh->checksum);
88
89 val = sctp_start_cksum((__u8 *)sh, skb_headlen(skb));
90
91 for (; list; list = list->next)
92 val = sctp_update_cksum((__u8 *)list->data, skb_headlen(list),
93 val);
94
95 val = sctp_end_cksum(val);
96
97 if (val != cmp) {
98 /* CRC failure, dump it. */
99 SCTP_INC_STATS_BH(SCTP_MIB_CHECKSUMERRORS);
100 return -1;
101 }
102 return 0;
103}
104
David S. Miller79af02c2005-07-08 21:47:49 -0700105struct sctp_input_cb {
106 union {
107 struct inet_skb_parm h4;
108#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
109 struct inet6_skb_parm h6;
110#endif
111 } header;
112 struct sctp_chunk *chunk;
113};
114#define SCTP_INPUT_CB(__skb) ((struct sctp_input_cb *)&((__skb)->cb[0]))
115
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116/*
117 * This is the routine which IP calls when receiving an SCTP packet.
118 */
119int sctp_rcv(struct sk_buff *skb)
120{
121 struct sock *sk;
122 struct sctp_association *asoc;
123 struct sctp_endpoint *ep = NULL;
124 struct sctp_ep_common *rcvr;
125 struct sctp_transport *transport = NULL;
126 struct sctp_chunk *chunk;
127 struct sctphdr *sh;
128 union sctp_addr src;
129 union sctp_addr dest;
130 int family;
131 struct sctp_af *af;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132
133 if (skb->pkt_type!=PACKET_HOST)
134 goto discard_it;
135
136 SCTP_INC_STATS_BH(SCTP_MIB_INSCTPPACKS);
137
Herbert Xu28cd7752006-10-29 23:46:42 -0800138 if (skb_linearize(skb))
139 goto discard_it;
140
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 sh = (struct sctphdr *) skb->h.raw;
142
143 /* Pull up the IP and SCTP headers. */
144 __skb_pull(skb, skb->h.raw - skb->data);
145 if (skb->len < sizeof(struct sctphdr))
146 goto discard_it;
Sridhar Samudrala503b55f2006-06-17 22:57:28 -0700147 if ((skb->ip_summed != CHECKSUM_UNNECESSARY) &&
148 (sctp_rcv_checksum(skb) < 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 goto discard_it;
150
151 skb_pull(skb, sizeof(struct sctphdr));
152
153 /* Make sure we at least have chunk headers worth of data left. */
154 if (skb->len < sizeof(struct sctp_chunkhdr))
155 goto discard_it;
156
157 family = ipver2af(skb->nh.iph->version);
158 af = sctp_get_af_specific(family);
159 if (unlikely(!af))
160 goto discard_it;
161
162 /* Initialize local addresses for lookups. */
163 af->from_skb(&src, skb, 1);
164 af->from_skb(&dest, skb, 0);
165
166 /* If the packet is to or from a non-unicast address,
167 * silently discard the packet.
168 *
169 * This is not clearly defined in the RFC except in section
170 * 8.4 - OOTB handling. However, based on the book "Stream Control
171 * Transmission Protocol" 2.1, "It is important to note that the
172 * IP address of an SCTP transport address must be a routable
173 * unicast address. In other words, IP multicast addresses and
174 * IP broadcast addresses cannot be used in an SCTP transport
175 * address."
176 */
Vlad Yasevich5636bef2006-06-17 22:55:35 -0700177 if (!af->addr_valid(&src, NULL, skb) ||
178 !af->addr_valid(&dest, NULL, skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 goto discard_it;
180
181 asoc = __sctp_rcv_lookup(skb, &src, &dest, &transport);
182
Neil Horman0fd9a652005-06-13 15:11:24 -0700183 if (!asoc)
184 ep = __sctp_rcv_lookup_endpoint(&dest);
185
186 /* Retrieve the common input handling substructure. */
187 rcvr = asoc ? &asoc->base : &ep->base;
188 sk = rcvr->sk;
189
190 /*
191 * If a frame arrives on an interface and the receiving socket is
192 * bound to another interface, via SO_BINDTODEVICE, treat it as OOTB
193 */
194 if (sk->sk_bound_dev_if && (sk->sk_bound_dev_if != af->skb_iif(skb)))
195 {
Neil Horman0fd9a652005-06-13 15:11:24 -0700196 if (asoc) {
197 sctp_association_put(asoc);
198 asoc = NULL;
199 } else {
200 sctp_endpoint_put(ep);
201 ep = NULL;
202 }
203 sk = sctp_get_ctl_sock();
204 ep = sctp_sk(sk)->ep;
205 sctp_endpoint_hold(ep);
Neil Horman0fd9a652005-06-13 15:11:24 -0700206 rcvr = &ep->base;
207 }
208
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 /*
210 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
211 * An SCTP packet is called an "out of the blue" (OOTB)
212 * packet if it is correctly formed, i.e., passed the
213 * receiver's checksum check, but the receiver is not
214 * able to identify the association to which this
215 * packet belongs.
216 */
217 if (!asoc) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 if (sctp_rcv_ootb(skb)) {
219 SCTP_INC_STATS_BH(SCTP_MIB_OUTOFBLUES);
220 goto discard_release;
221 }
222 }
223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 if (!xfrm_policy_check(sk, XFRM_POLICY_IN, skb, family))
225 goto discard_release;
Patrick McHardyb59c2702006-01-06 23:06:10 -0800226 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700228 if (sk_filter(sk, skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 goto discard_release;
230
231 /* Create an SCTP packet structure. */
232 chunk = sctp_chunkify(skb, asoc, sk);
Herbert Xu2babf9d2006-03-25 01:25:29 -0800233 if (!chunk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 goto discard_release;
David S. Miller79af02c2005-07-08 21:47:49 -0700235 SCTP_INPUT_CB(skb)->chunk = chunk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 /* Remember what endpoint is to handle this packet. */
238 chunk->rcvr = rcvr;
239
240 /* Remember the SCTP header. */
241 chunk->sctp_hdr = sh;
242
243 /* Set the source and destination addresses of the incoming chunk. */
244 sctp_init_addrs(chunk, &src, &dest);
245
246 /* Remember where we came from. */
247 chunk->transport = transport;
248
249 /* Acquire access to the sock lock. Note: We are safe from other
250 * bottom halves on this lock, but a user may be in the lock too,
251 * so check if it is busy.
252 */
253 sctp_bh_lock_sock(sk);
254
Sridhar Samudralaac0b0462006-08-22 00:15:33 -0700255 if (sock_owned_by_user(sk)) {
256 SCTP_INC_STATS_BH(SCTP_MIB_IN_PKT_BACKLOG);
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700257 sctp_add_backlog(sk, skb);
Sridhar Samudralaac0b0462006-08-22 00:15:33 -0700258 } else {
259 SCTP_INC_STATS_BH(SCTP_MIB_IN_PKT_SOFTIRQ);
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700260 sctp_inq_push(&chunk->rcvr->inqueue, chunk);
Sridhar Samudralaac0b0462006-08-22 00:15:33 -0700261 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 sctp_bh_unlock_sock(sk);
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700264
265 /* Release the asoc/ep ref we took in the lookup calls. */
266 if (asoc)
267 sctp_association_put(asoc);
268 else
269 sctp_endpoint_put(ep);
Sridhar Samudrala7a48f922006-01-17 11:51:28 -0800270
Herbert Xu2babf9d2006-03-25 01:25:29 -0800271 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
273discard_it:
Sridhar Samudralaac0b0462006-08-22 00:15:33 -0700274 SCTP_INC_STATS_BH(SCTP_MIB_IN_PKT_DISCARDS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 kfree_skb(skb);
Herbert Xu2babf9d2006-03-25 01:25:29 -0800276 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277
278discard_release:
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700279 /* Release the asoc/ep ref we took in the lookup calls. */
Neil Horman0fd9a652005-06-13 15:11:24 -0700280 if (asoc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 sctp_association_put(asoc);
Neil Horman0fd9a652005-06-13 15:11:24 -0700282 else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 sctp_endpoint_put(ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284
285 goto discard_it;
286}
287
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700288/* Process the backlog queue of the socket. Every skb on
289 * the backlog holds a ref on an association or endpoint.
290 * We hold this ref throughout the state machine to make
291 * sure that the structure we need is still around.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 */
293int sctp_backlog_rcv(struct sock *sk, struct sk_buff *skb)
294{
David S. Miller79af02c2005-07-08 21:47:49 -0700295 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700296 struct sctp_inq *inqueue = &chunk->rcvr->inqueue;
Sridhar Samudrala7a48f922006-01-17 11:51:28 -0800297 struct sctp_ep_common *rcvr = NULL;
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700298 int backloged = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299
Sridhar Samudrala7a48f922006-01-17 11:51:28 -0800300 rcvr = chunk->rcvr;
Sridhar Samudralac4d24442006-01-17 11:56:26 -0800301
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700302 /* If the rcvr is dead then the association or endpoint
303 * has been deleted and we can safely drop the chunk
304 * and refs that we are holding.
305 */
306 if (rcvr->dead) {
307 sctp_chunk_free(chunk);
308 goto done;
309 }
Sridhar Samudralac4d24442006-01-17 11:56:26 -0800310
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700311 if (unlikely(rcvr->sk != sk)) {
312 /* In this case, the association moved from one socket to
313 * another. We are currently sitting on the backlog of the
314 * old socket, so we need to move.
315 * However, since we are here in the process context we
316 * need to take make sure that the user doesn't own
317 * the new socket when we process the packet.
318 * If the new socket is user-owned, queue the chunk to the
319 * backlog of the new socket without dropping any refs.
320 * Otherwise, we can safely push the chunk on the inqueue.
321 */
Sridhar Samudrala7a48f922006-01-17 11:51:28 -0800322
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700323 sk = rcvr->sk;
324 sctp_bh_lock_sock(sk);
325
326 if (sock_owned_by_user(sk)) {
327 sk_add_backlog(sk, skb);
328 backloged = 1;
329 } else
330 sctp_inq_push(inqueue, chunk);
331
332 sctp_bh_unlock_sock(sk);
333
334 /* If the chunk was backloged again, don't drop refs */
335 if (backloged)
336 return 0;
337 } else {
338 sctp_inq_push(inqueue, chunk);
339 }
340
341done:
342 /* Release the refs we took in sctp_add_backlog */
343 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
344 sctp_association_put(sctp_assoc(rcvr));
345 else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
346 sctp_endpoint_put(sctp_ep(rcvr));
347 else
348 BUG();
349
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 return 0;
351}
352
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700353static void sctp_add_backlog(struct sock *sk, struct sk_buff *skb)
Sridhar Samudralac4d24442006-01-17 11:56:26 -0800354{
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700355 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
356 struct sctp_ep_common *rcvr = chunk->rcvr;
Sridhar Samudralac4d24442006-01-17 11:56:26 -0800357
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700358 /* Hold the assoc/ep while hanging on the backlog queue.
359 * This way, we know structures we need will not disappear from us
360 */
361 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
362 sctp_association_hold(sctp_assoc(rcvr));
363 else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
364 sctp_endpoint_hold(sctp_ep(rcvr));
365 else
366 BUG();
Sridhar Samudralac4d24442006-01-17 11:56:26 -0800367
Vladislav Yasevich61c9fed2006-05-19 11:01:18 -0700368 sk_add_backlog(sk, skb);
Sridhar Samudralac4d24442006-01-17 11:56:26 -0800369}
370
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371/* 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{
Frank Filz52ccb8e2005-12-22 11:36:46 -0800375 if (sock_owned_by_user(sk) || !t || (t->pathmtu == pmtu))
376 return;
377
378 if (t->param_flags & SPP_PMTUD_ENABLE) {
379 if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
380 printk(KERN_WARNING "%s: Reported pmtu %d too low, "
381 "using default minimum of %d\n",
382 __FUNCTION__, pmtu,
383 SCTP_DEFAULT_MINSEGMENT);
384 /* Use default minimum segment size and disable
385 * pmtu discovery on this transport.
386 */
387 t->pathmtu = SCTP_DEFAULT_MINSEGMENT;
Vlad Yasevichb56bab42006-09-29 17:09:34 -0700388 t->param_flags = (t->param_flags & ~SPP_PMTUD) |
Frank Filz52ccb8e2005-12-22 11:36:46 -0800389 SPP_PMTUD_DISABLE;
390 } else {
391 t->pathmtu = pmtu;
392 }
393
394 /* Update association pmtu. */
395 sctp_assoc_sync_pmtu(asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 }
397
Frank Filz52ccb8e2005-12-22 11:36:46 -0800398 /* Retransmit with the new pmtu setting.
399 * Normally, if PMTU discovery is disabled, an ICMP Fragmentation
400 * Needed will never be sent, but if a message was sent before
401 * PMTU discovery was disabled that was larger than the PMTU, it
402 * would not be fragmented, so it must be re-transmitted fragmented.
403 */
404 sctp_retransmit(&asoc->outqueue, t, SCTP_RTXR_PMTUD);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405}
406
407/*
408 * SCTP Implementer's Guide, 2.37 ICMP handling procedures
409 *
410 * ICMP8) If the ICMP code is a "Unrecognized next header type encountered"
411 * or a "Protocol Unreachable" treat this message as an abort
412 * with the T bit set.
413 *
414 * This function sends an event to the state machine, which will abort the
415 * association.
416 *
417 */
418void sctp_icmp_proto_unreachable(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 struct sctp_association *asoc,
420 struct sctp_transport *t)
421{
422 SCTP_DEBUG_PRINTK("%s\n", __FUNCTION__);
423
424 sctp_do_sm(SCTP_EVENT_T_OTHER,
425 SCTP_ST_OTHER(SCTP_EVENT_ICMP_PROTO_UNREACH),
Frank Filz3f7a87d2005-06-20 13:14:57 -0700426 asoc->state, asoc->ep, asoc, t,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 GFP_ATOMIC);
428
429}
430
431/* Common lookup code for icmp/icmpv6 error handler. */
432struct sock *sctp_err_lookup(int family, struct sk_buff *skb,
433 struct sctphdr *sctphdr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 struct sctp_association **app,
435 struct sctp_transport **tpp)
436{
437 union sctp_addr saddr;
438 union sctp_addr daddr;
439 struct sctp_af *af;
440 struct sock *sk = NULL;
Sridhar Samudrala8de8c872006-05-19 10:58:12 -0700441 struct sctp_association *asoc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 struct sctp_transport *transport = NULL;
443
Sridhar Samudralad1ad1ff2005-07-18 13:44:10 -0700444 *app = NULL; *tpp = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445
446 af = sctp_get_af_specific(family);
447 if (unlikely(!af)) {
448 return NULL;
449 }
450
451 /* Initialize local addresses for lookups. */
452 af->from_skb(&saddr, skb, 1);
453 af->from_skb(&daddr, skb, 0);
454
455 /* Look for an association that matches the incoming ICMP error
456 * packet.
457 */
458 asoc = __sctp_lookup_association(&saddr, &daddr, &transport);
Sridhar Samudralad1ad1ff2005-07-18 13:44:10 -0700459 if (!asoc)
460 return NULL;
461
462 sk = asoc->base.sk;
463
464 if (ntohl(sctphdr->vtag) != asoc->c.peer_vtag) {
465 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
466 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 }
468
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 sctp_bh_lock_sock(sk);
470
471 /* If too many ICMPs get dropped on busy
472 * servers this needs to be solved differently.
473 */
474 if (sock_owned_by_user(sk))
475 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
476
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 *app = asoc;
478 *tpp = transport;
479 return sk;
480
481out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 if (asoc)
483 sctp_association_put(asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 return NULL;
485}
486
487/* Common cleanup code for icmp/icmpv6 error handler. */
Sridhar Samudralad1ad1ff2005-07-18 13:44:10 -0700488void sctp_err_finish(struct sock *sk, struct sctp_association *asoc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
490 sctp_bh_unlock_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 if (asoc)
492 sctp_association_put(asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493}
494
495/*
496 * This routine is called by the ICMP module when it gets some
497 * sort of error condition. If err < 0 then the socket should
498 * be closed and the error returned to the user. If err > 0
499 * it's just the icmp type << 8 | icmp code. After adjustment
500 * header points to the first 8 bytes of the sctp header. We need
501 * to find the appropriate port.
502 *
503 * The locking strategy used here is very "optimistic". When
504 * someone else accesses the socket the ICMP is just dropped
505 * and for some paths there is no check at all.
506 * A more general error queue to queue errors for later handling
507 * is probably better.
508 *
509 */
510void sctp_v4_err(struct sk_buff *skb, __u32 info)
511{
512 struct iphdr *iph = (struct iphdr *)skb->data;
513 struct sctphdr *sh = (struct sctphdr *)(skb->data + (iph->ihl <<2));
514 int type = skb->h.icmph->type;
515 int code = skb->h.icmph->code;
516 struct sock *sk;
Sridhar Samudrala8de8c872006-05-19 10:58:12 -0700517 struct sctp_association *asoc = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 struct sctp_transport *transport;
519 struct inet_sock *inet;
520 char *saveip, *savesctp;
521 int err;
522
523 if (skb->len < ((iph->ihl << 2) + 8)) {
524 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
525 return;
526 }
527
528 /* Fix up skb to look at the embedded net header. */
529 saveip = skb->nh.raw;
530 savesctp = skb->h.raw;
531 skb->nh.iph = iph;
532 skb->h.raw = (char *)sh;
Sridhar Samudralad1ad1ff2005-07-18 13:44:10 -0700533 sk = sctp_err_lookup(AF_INET, skb, sh, &asoc, &transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 /* Put back, the original pointers. */
535 skb->nh.raw = saveip;
536 skb->h.raw = savesctp;
537 if (!sk) {
538 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
539 return;
540 }
541 /* Warning: The sock lock is held. Remember to call
542 * sctp_err_finish!
543 */
544
545 switch (type) {
546 case ICMP_PARAMETERPROB:
547 err = EPROTO;
548 break;
549 case ICMP_DEST_UNREACH:
550 if (code > NR_ICMP_UNREACH)
551 goto out_unlock;
552
553 /* PMTU discovery (RFC1191) */
554 if (ICMP_FRAG_NEEDED == code) {
555 sctp_icmp_frag_needed(sk, asoc, transport, info);
556 goto out_unlock;
557 }
558 else {
559 if (ICMP_PROT_UNREACH == code) {
Sridhar Samudralad1ad1ff2005-07-18 13:44:10 -0700560 sctp_icmp_proto_unreachable(sk, asoc,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 transport);
562 goto out_unlock;
563 }
564 }
565 err = icmp_err_convert[code].errno;
566 break;
567 case ICMP_TIME_EXCEEDED:
568 /* Ignore any time exceeded errors due to fragment reassembly
569 * timeouts.
570 */
571 if (ICMP_EXC_FRAGTIME == code)
572 goto out_unlock;
573
574 err = EHOSTUNREACH;
575 break;
576 default:
577 goto out_unlock;
578 }
579
580 inet = inet_sk(sk);
581 if (!sock_owned_by_user(sk) && inet->recverr) {
582 sk->sk_err = err;
583 sk->sk_error_report(sk);
584 } else { /* Only an error on timeout */
585 sk->sk_err_soft = err;
586 }
587
588out_unlock:
Sridhar Samudralad1ad1ff2005-07-18 13:44:10 -0700589 sctp_err_finish(sk, asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590}
591
592/*
593 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
594 *
595 * This function scans all the chunks in the OOTB packet to determine if
596 * the packet should be discarded right away. If a response might be needed
597 * for this packet, or, if further processing is possible, the packet will
598 * be queued to a proper inqueue for the next phase of handling.
599 *
600 * Output:
601 * Return 0 - If further processing is needed.
602 * Return 1 - If the packet can be discarded right away.
603 */
604int sctp_rcv_ootb(struct sk_buff *skb)
605{
606 sctp_chunkhdr_t *ch;
607 __u8 *ch_end;
608 sctp_errhdr_t *err;
609
610 ch = (sctp_chunkhdr_t *) skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611
612 /* Scan through all the chunks in the packet. */
Tsutomu Fujiia7d1f1b2006-01-17 11:57:09 -0800613 do {
614 /* Break out if chunk length is less then minimal. */
615 if (ntohs(ch->length) < sizeof(sctp_chunkhdr_t))
616 break;
617
618 ch_end = ((__u8 *)ch) + WORD_ROUND(ntohs(ch->length));
619 if (ch_end > skb->tail)
620 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621
622 /* RFC 8.4, 2) If the OOTB packet contains an ABORT chunk, the
623 * receiver MUST silently discard the OOTB packet and take no
624 * further action.
625 */
626 if (SCTP_CID_ABORT == ch->type)
627 goto discard;
628
629 /* RFC 8.4, 6) If the packet contains a SHUTDOWN COMPLETE
630 * chunk, the receiver should silently discard the packet
631 * and take no further action.
632 */
633 if (SCTP_CID_SHUTDOWN_COMPLETE == ch->type)
634 goto discard;
635
636 /* RFC 8.4, 7) If the packet contains a "Stale cookie" ERROR
637 * or a COOKIE ACK the SCTP Packet should be silently
638 * discarded.
639 */
640 if (SCTP_CID_COOKIE_ACK == ch->type)
641 goto discard;
642
643 if (SCTP_CID_ERROR == ch->type) {
644 sctp_walk_errors(err, ch) {
645 if (SCTP_ERROR_STALE_COOKIE == err->cause)
646 goto discard;
647 }
648 }
649
650 ch = (sctp_chunkhdr_t *) ch_end;
Tsutomu Fujiia7d1f1b2006-01-17 11:57:09 -0800651 } while (ch_end < skb->tail);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
653 return 0;
654
655discard:
656 return 1;
657}
658
659/* Insert endpoint into the hash table. */
660static void __sctp_hash_endpoint(struct sctp_endpoint *ep)
661{
662 struct sctp_ep_common **epp;
663 struct sctp_ep_common *epb;
664 struct sctp_hashbucket *head;
665
666 epb = &ep->base;
667
668 epb->hashent = sctp_ep_hashfn(epb->bind_addr.port);
669 head = &sctp_ep_hashtable[epb->hashent];
670
671 sctp_write_lock(&head->lock);
672 epp = &head->chain;
673 epb->next = *epp;
674 if (epb->next)
675 (*epp)->pprev = &epb->next;
676 *epp = epb;
677 epb->pprev = epp;
678 sctp_write_unlock(&head->lock);
679}
680
681/* Add an endpoint to the hash. Local BH-safe. */
682void sctp_hash_endpoint(struct sctp_endpoint *ep)
683{
684 sctp_local_bh_disable();
685 __sctp_hash_endpoint(ep);
686 sctp_local_bh_enable();
687}
688
689/* Remove endpoint from the hash table. */
690static void __sctp_unhash_endpoint(struct sctp_endpoint *ep)
691{
692 struct sctp_hashbucket *head;
693 struct sctp_ep_common *epb;
694
695 epb = &ep->base;
696
697 epb->hashent = sctp_ep_hashfn(epb->bind_addr.port);
698
699 head = &sctp_ep_hashtable[epb->hashent];
700
701 sctp_write_lock(&head->lock);
702
703 if (epb->pprev) {
704 if (epb->next)
705 epb->next->pprev = epb->pprev;
706 *epb->pprev = epb->next;
707 epb->pprev = NULL;
708 }
709
710 sctp_write_unlock(&head->lock);
711}
712
713/* Remove endpoint from the hash. Local BH-safe. */
714void sctp_unhash_endpoint(struct sctp_endpoint *ep)
715{
716 sctp_local_bh_disable();
717 __sctp_unhash_endpoint(ep);
718 sctp_local_bh_enable();
719}
720
721/* Look up an endpoint. */
722static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(const union sctp_addr *laddr)
723{
724 struct sctp_hashbucket *head;
725 struct sctp_ep_common *epb;
726 struct sctp_endpoint *ep;
727 int hash;
728
729 hash = sctp_ep_hashfn(laddr->v4.sin_port);
730 head = &sctp_ep_hashtable[hash];
731 read_lock(&head->lock);
732 for (epb = head->chain; epb; epb = epb->next) {
733 ep = sctp_ep(epb);
734 if (sctp_endpoint_is_match(ep, laddr))
735 goto hit;
736 }
737
738 ep = sctp_sk((sctp_get_ctl_sock()))->ep;
739 epb = &ep->base;
740
741hit:
742 sctp_endpoint_hold(ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 read_unlock(&head->lock);
744 return ep;
745}
746
747/* Insert association into the hash table. */
748static void __sctp_hash_established(struct sctp_association *asoc)
749{
750 struct sctp_ep_common **epp;
751 struct sctp_ep_common *epb;
752 struct sctp_hashbucket *head;
753
754 epb = &asoc->base;
755
756 /* Calculate which chain this entry will belong to. */
757 epb->hashent = sctp_assoc_hashfn(epb->bind_addr.port, asoc->peer.port);
758
759 head = &sctp_assoc_hashtable[epb->hashent];
760
761 sctp_write_lock(&head->lock);
762 epp = &head->chain;
763 epb->next = *epp;
764 if (epb->next)
765 (*epp)->pprev = &epb->next;
766 *epp = epb;
767 epb->pprev = epp;
768 sctp_write_unlock(&head->lock);
769}
770
771/* Add an association to the hash. Local BH-safe. */
772void sctp_hash_established(struct sctp_association *asoc)
773{
Vlad Yasevichde76e692006-10-30 18:55:11 -0800774 if (asoc->temp)
775 return;
776
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 sctp_local_bh_disable();
778 __sctp_hash_established(asoc);
779 sctp_local_bh_enable();
780}
781
782/* Remove association from the hash table. */
783static void __sctp_unhash_established(struct sctp_association *asoc)
784{
785 struct sctp_hashbucket *head;
786 struct sctp_ep_common *epb;
787
788 epb = &asoc->base;
789
790 epb->hashent = sctp_assoc_hashfn(epb->bind_addr.port,
791 asoc->peer.port);
792
793 head = &sctp_assoc_hashtable[epb->hashent];
794
795 sctp_write_lock(&head->lock);
796
797 if (epb->pprev) {
798 if (epb->next)
799 epb->next->pprev = epb->pprev;
800 *epb->pprev = epb->next;
801 epb->pprev = NULL;
802 }
803
804 sctp_write_unlock(&head->lock);
805}
806
807/* Remove association from the hash table. Local BH-safe. */
808void sctp_unhash_established(struct sctp_association *asoc)
809{
Vlad Yasevichde76e692006-10-30 18:55:11 -0800810 if (asoc->temp)
811 return;
812
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 sctp_local_bh_disable();
814 __sctp_unhash_established(asoc);
815 sctp_local_bh_enable();
816}
817
818/* Look up an association. */
819static struct sctp_association *__sctp_lookup_association(
820 const union sctp_addr *local,
821 const union sctp_addr *peer,
822 struct sctp_transport **pt)
823{
824 struct sctp_hashbucket *head;
825 struct sctp_ep_common *epb;
826 struct sctp_association *asoc;
827 struct sctp_transport *transport;
828 int hash;
829
830 /* Optimize here for direct hit, only listening connections can
831 * have wildcards anyways.
832 */
833 hash = sctp_assoc_hashfn(local->v4.sin_port, peer->v4.sin_port);
834 head = &sctp_assoc_hashtable[hash];
835 read_lock(&head->lock);
836 for (epb = head->chain; epb; epb = epb->next) {
837 asoc = sctp_assoc(epb);
838 transport = sctp_assoc_is_match(asoc, local, peer);
839 if (transport)
840 goto hit;
841 }
842
843 read_unlock(&head->lock);
844
845 return NULL;
846
847hit:
848 *pt = transport;
849 sctp_association_hold(asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 read_unlock(&head->lock);
851 return asoc;
852}
853
854/* Look up an association. BH-safe. */
855SCTP_STATIC
856struct sctp_association *sctp_lookup_association(const union sctp_addr *laddr,
857 const union sctp_addr *paddr,
858 struct sctp_transport **transportp)
859{
860 struct sctp_association *asoc;
861
862 sctp_local_bh_disable();
863 asoc = __sctp_lookup_association(laddr, paddr, transportp);
864 sctp_local_bh_enable();
865
866 return asoc;
867}
868
869/* Is there an association matching the given local and peer addresses? */
870int sctp_has_association(const union sctp_addr *laddr,
871 const union sctp_addr *paddr)
872{
873 struct sctp_association *asoc;
874 struct sctp_transport *transport;
875
876 if ((asoc = sctp_lookup_association(laddr, paddr, &transport))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 sctp_association_put(asoc);
878 return 1;
879 }
880
881 return 0;
882}
883
884/*
885 * SCTP Implementors Guide, 2.18 Handling of address
886 * parameters within the INIT or INIT-ACK.
887 *
888 * D) When searching for a matching TCB upon reception of an INIT
889 * or INIT-ACK chunk the receiver SHOULD use not only the
890 * source address of the packet (containing the INIT or
891 * INIT-ACK) but the receiver SHOULD also use all valid
892 * address parameters contained within the chunk.
893 *
894 * 2.18.3 Solution description
895 *
896 * This new text clearly specifies to an implementor the need
897 * to look within the INIT or INIT-ACK. Any implementation that
898 * does not do this, may not be able to establish associations
899 * in certain circumstances.
900 *
901 */
902static struct sctp_association *__sctp_rcv_init_lookup(struct sk_buff *skb,
903 const union sctp_addr *laddr, struct sctp_transport **transportp)
904{
905 struct sctp_association *asoc;
906 union sctp_addr addr;
907 union sctp_addr *paddr = &addr;
908 struct sctphdr *sh = (struct sctphdr *) skb->h.raw;
909 sctp_chunkhdr_t *ch;
910 union sctp_params params;
911 sctp_init_chunk_t *init;
912 struct sctp_transport *transport;
913 struct sctp_af *af;
914
915 ch = (sctp_chunkhdr_t *) skb->data;
916
917 /* If this is INIT/INIT-ACK look inside the chunk too. */
918 switch (ch->type) {
919 case SCTP_CID_INIT:
920 case SCTP_CID_INIT_ACK:
921 break;
922 default:
923 return NULL;
924 }
925
926 /* The code below will attempt to walk the chunk and extract
927 * parameter information. Before we do that, we need to verify
928 * that the chunk length doesn't cause overflow. Otherwise, we'll
929 * walk off the end.
930 */
931 if (WORD_ROUND(ntohs(ch->length)) > skb->len)
932 return NULL;
933
934 /*
935 * This code will NOT touch anything inside the chunk--it is
936 * strictly READ-ONLY.
937 *
938 * RFC 2960 3 SCTP packet Format
939 *
940 * Multiple chunks can be bundled into one SCTP packet up to
941 * the MTU size, except for the INIT, INIT ACK, and SHUTDOWN
942 * COMPLETE chunks. These chunks MUST NOT be bundled with any
943 * other chunk in a packet. See Section 6.10 for more details
944 * on chunk bundling.
945 */
946
947 /* Find the start of the TLVs and the end of the chunk. This is
948 * the region we search for address parameters.
949 */
950 init = (sctp_init_chunk_t *)skb->data;
951
952 /* Walk the parameters looking for embedded addresses. */
953 sctp_walk_params(params, init, init_hdr.params) {
954
955 /* Note: Ignoring hostname addresses. */
956 af = sctp_get_af_specific(param_type2af(params.p->type));
957 if (!af)
958 continue;
959
960 af->from_addr_param(paddr, params.addr, ntohs(sh->source), 0);
961
962 asoc = __sctp_lookup_association(laddr, paddr, &transport);
963 if (asoc)
964 return asoc;
965 }
966
967 return NULL;
968}
969
970/* Lookup an association for an inbound skb. */
971static struct sctp_association *__sctp_rcv_lookup(struct sk_buff *skb,
972 const union sctp_addr *paddr,
973 const union sctp_addr *laddr,
974 struct sctp_transport **transportp)
975{
976 struct sctp_association *asoc;
977
978 asoc = __sctp_lookup_association(laddr, paddr, transportp);
979
980 /* Further lookup for INIT/INIT-ACK packets.
981 * SCTP Implementors Guide, 2.18 Handling of address
982 * parameters within the INIT or INIT-ACK.
983 */
984 if (!asoc)
985 asoc = __sctp_rcv_init_lookup(skb, laddr, transportp);
986
987 return asoc;
988}