xref: /openbmc/linux/net/sctp/input.c (revision e868d61272caa648214046a096e5a6bfc068dc8c)
1 /* 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. */
65 static int sctp_rcv_ootb(struct sk_buff *);
66 static 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);
70 static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(const union sctp_addr *laddr);
71 static struct sctp_association *__sctp_lookup_association(
72 					const union sctp_addr *local,
73 					const union sctp_addr *peer,
74 					struct sctp_transport **pt);
75 
76 static void sctp_add_backlog(struct sock *sk, struct sk_buff *skb);
77 
78 
79 /* Calculate the SCTP checksum of an SCTP packet.  */
80 static inline int sctp_rcv_checksum(struct sk_buff *skb)
81 {
82 	struct sk_buff *list = skb_shinfo(skb)->frag_list;
83 	struct sctphdr *sh = sctp_hdr(skb);
84 	__u32 cmp = ntohl(sh->checksum);
85 	__u32 val = sctp_start_cksum((__u8 *)sh, skb_headlen(skb));
86 
87 	for (; list; list = list->next)
88 		val = sctp_update_cksum((__u8 *)list->data, skb_headlen(list),
89 					val);
90 
91 	val = sctp_end_cksum(val);
92 
93 	if (val != cmp) {
94 		/* CRC failure, dump it. */
95 		SCTP_INC_STATS_BH(SCTP_MIB_CHECKSUMERRORS);
96 		return -1;
97 	}
98 	return 0;
99 }
100 
101 struct sctp_input_cb {
102 	union {
103 		struct inet_skb_parm	h4;
104 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
105 		struct inet6_skb_parm	h6;
106 #endif
107 	} header;
108 	struct sctp_chunk *chunk;
109 };
110 #define SCTP_INPUT_CB(__skb)	((struct sctp_input_cb *)&((__skb)->cb[0]))
111 
112 /*
113  * This is the routine which IP calls when receiving an SCTP packet.
114  */
115 int sctp_rcv(struct sk_buff *skb)
116 {
117 	struct sock *sk;
118 	struct sctp_association *asoc;
119 	struct sctp_endpoint *ep = NULL;
120 	struct sctp_ep_common *rcvr;
121 	struct sctp_transport *transport = NULL;
122 	struct sctp_chunk *chunk;
123 	struct sctphdr *sh;
124 	union sctp_addr src;
125 	union sctp_addr dest;
126 	int family;
127 	struct sctp_af *af;
128 
129 	if (skb->pkt_type!=PACKET_HOST)
130 		goto discard_it;
131 
132 	SCTP_INC_STATS_BH(SCTP_MIB_INSCTPPACKS);
133 
134 	if (skb_linearize(skb))
135 		goto discard_it;
136 
137 	sh = sctp_hdr(skb);
138 
139 	/* Pull up the IP and SCTP headers. */
140 	__skb_pull(skb, skb_transport_offset(skb));
141 	if (skb->len < sizeof(struct sctphdr))
142 		goto discard_it;
143 	if (!skb_csum_unnecessary(skb) && sctp_rcv_checksum(skb) < 0)
144 		goto discard_it;
145 
146 	skb_pull(skb, sizeof(struct sctphdr));
147 
148 	/* Make sure we at least have chunk headers worth of data left. */
149 	if (skb->len < sizeof(struct sctp_chunkhdr))
150 		goto discard_it;
151 
152 	family = ipver2af(ip_hdr(skb)->version);
153 	af = sctp_get_af_specific(family);
154 	if (unlikely(!af))
155 		goto discard_it;
156 
157 	/* Initialize local addresses for lookups. */
158 	af->from_skb(&src, skb, 1);
159 	af->from_skb(&dest, skb, 0);
160 
161 	/* If the packet is to or from a non-unicast address,
162 	 * silently discard the packet.
163 	 *
164 	 * This is not clearly defined in the RFC except in section
165 	 * 8.4 - OOTB handling.  However, based on the book "Stream Control
166 	 * Transmission Protocol" 2.1, "It is important to note that the
167 	 * IP address of an SCTP transport address must be a routable
168 	 * unicast address.  In other words, IP multicast addresses and
169 	 * IP broadcast addresses cannot be used in an SCTP transport
170 	 * address."
171 	 */
172 	if (!af->addr_valid(&src, NULL, skb) ||
173 	    !af->addr_valid(&dest, NULL, skb))
174 		goto discard_it;
175 
176 	asoc = __sctp_rcv_lookup(skb, &src, &dest, &transport);
177 
178 	if (!asoc)
179 		ep = __sctp_rcv_lookup_endpoint(&dest);
180 
181 	/* Retrieve the common input handling substructure. */
182 	rcvr = asoc ? &asoc->base : &ep->base;
183 	sk = rcvr->sk;
184 
185 	/*
186 	 * If a frame arrives on an interface and the receiving socket is
187 	 * bound to another interface, via SO_BINDTODEVICE, treat it as OOTB
188 	 */
189 	if (sk->sk_bound_dev_if && (sk->sk_bound_dev_if != af->skb_iif(skb)))
190 	{
191 		if (asoc) {
192 			sctp_association_put(asoc);
193 			asoc = NULL;
194 		} else {
195 			sctp_endpoint_put(ep);
196 			ep = NULL;
197 		}
198 		sk = sctp_get_ctl_sock();
199 		ep = sctp_sk(sk)->ep;
200 		sctp_endpoint_hold(ep);
201 		rcvr = &ep->base;
202 	}
203 
204 	/*
205 	 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
206 	 * An SCTP packet is called an "out of the blue" (OOTB)
207 	 * packet if it is correctly formed, i.e., passed the
208 	 * receiver's checksum check, but the receiver is not
209 	 * able to identify the association to which this
210 	 * packet belongs.
211 	 */
212 	if (!asoc) {
213 		if (sctp_rcv_ootb(skb)) {
214 			SCTP_INC_STATS_BH(SCTP_MIB_OUTOFBLUES);
215 			goto discard_release;
216 		}
217 	}
218 
219 	if (!xfrm_policy_check(sk, XFRM_POLICY_IN, skb, family))
220 		goto discard_release;
221 	nf_reset(skb);
222 
223 	if (sk_filter(sk, skb))
224 		goto discard_release;
225 
226 	/* Create an SCTP packet structure. */
227 	chunk = sctp_chunkify(skb, asoc, sk);
228 	if (!chunk)
229 		goto discard_release;
230 	SCTP_INPUT_CB(skb)->chunk = chunk;
231 
232 	/* Remember what endpoint is to handle this packet. */
233 	chunk->rcvr = rcvr;
234 
235 	/* Remember the SCTP header. */
236 	chunk->sctp_hdr = sh;
237 
238 	/* Set the source and destination addresses of the incoming chunk.  */
239 	sctp_init_addrs(chunk, &src, &dest);
240 
241 	/* Remember where we came from.  */
242 	chunk->transport = transport;
243 
244 	/* Acquire access to the sock lock. Note: We are safe from other
245 	 * bottom halves on this lock, but a user may be in the lock too,
246 	 * so check if it is busy.
247 	 */
248 	sctp_bh_lock_sock(sk);
249 
250 	if (sock_owned_by_user(sk)) {
251 		SCTP_INC_STATS_BH(SCTP_MIB_IN_PKT_BACKLOG);
252 		sctp_add_backlog(sk, skb);
253 	} else {
254 		SCTP_INC_STATS_BH(SCTP_MIB_IN_PKT_SOFTIRQ);
255 		sctp_inq_push(&chunk->rcvr->inqueue, chunk);
256 	}
257 
258 	sctp_bh_unlock_sock(sk);
259 
260 	/* Release the asoc/ep ref we took in the lookup calls. */
261 	if (asoc)
262 		sctp_association_put(asoc);
263 	else
264 		sctp_endpoint_put(ep);
265 
266 	return 0;
267 
268 discard_it:
269 	SCTP_INC_STATS_BH(SCTP_MIB_IN_PKT_DISCARDS);
270 	kfree_skb(skb);
271 	return 0;
272 
273 discard_release:
274 	/* Release the asoc/ep ref we took in the lookup calls. */
275 	if (asoc)
276 		sctp_association_put(asoc);
277 	else
278 		sctp_endpoint_put(ep);
279 
280 	goto discard_it;
281 }
282 
283 /* Process the backlog queue of the socket.  Every skb on
284  * the backlog holds a ref on an association or endpoint.
285  * We hold this ref throughout the state machine to make
286  * sure that the structure we need is still around.
287  */
288 int sctp_backlog_rcv(struct sock *sk, struct sk_buff *skb)
289 {
290 	struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
291 	struct sctp_inq *inqueue = &chunk->rcvr->inqueue;
292 	struct sctp_ep_common *rcvr = NULL;
293 	int backloged = 0;
294 
295 	rcvr = chunk->rcvr;
296 
297 	/* If the rcvr is dead then the association or endpoint
298 	 * has been deleted and we can safely drop the chunk
299 	 * and refs that we are holding.
300 	 */
301 	if (rcvr->dead) {
302 		sctp_chunk_free(chunk);
303 		goto done;
304 	}
305 
306 	if (unlikely(rcvr->sk != sk)) {
307 		/* In this case, the association moved from one socket to
308 		 * another.  We are currently sitting on the backlog of the
309 		 * old socket, so we need to move.
310 		 * However, since we are here in the process context we
311 		 * need to take make sure that the user doesn't own
312 		 * the new socket when we process the packet.
313 		 * If the new socket is user-owned, queue the chunk to the
314 		 * backlog of the new socket without dropping any refs.
315 		 * Otherwise, we can safely push the chunk on the inqueue.
316 		 */
317 
318 		sk = rcvr->sk;
319 		sctp_bh_lock_sock(sk);
320 
321 		if (sock_owned_by_user(sk)) {
322 			sk_add_backlog(sk, skb);
323 			backloged = 1;
324 		} else
325 			sctp_inq_push(inqueue, chunk);
326 
327 		sctp_bh_unlock_sock(sk);
328 
329 		/* If the chunk was backloged again, don't drop refs */
330 		if (backloged)
331 			return 0;
332 	} else {
333 		sctp_inq_push(inqueue, chunk);
334 	}
335 
336 done:
337 	/* Release the refs we took in sctp_add_backlog */
338 	if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
339 		sctp_association_put(sctp_assoc(rcvr));
340 	else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
341 		sctp_endpoint_put(sctp_ep(rcvr));
342 	else
343 		BUG();
344 
345 	return 0;
346 }
347 
348 static void sctp_add_backlog(struct sock *sk, struct sk_buff *skb)
349 {
350 	struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk;
351 	struct sctp_ep_common *rcvr = chunk->rcvr;
352 
353 	/* Hold the assoc/ep while hanging on the backlog queue.
354 	 * This way, we know structures we need will not disappear from us
355 	 */
356 	if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type)
357 		sctp_association_hold(sctp_assoc(rcvr));
358 	else if (SCTP_EP_TYPE_SOCKET == rcvr->type)
359 		sctp_endpoint_hold(sctp_ep(rcvr));
360 	else
361 		BUG();
362 
363 	sk_add_backlog(sk, skb);
364 }
365 
366 /* Handle icmp frag needed error. */
367 void sctp_icmp_frag_needed(struct sock *sk, struct sctp_association *asoc,
368 			   struct sctp_transport *t, __u32 pmtu)
369 {
370 	if (sock_owned_by_user(sk) || !t || (t->pathmtu == pmtu))
371 		return;
372 
373 	if (t->param_flags & SPP_PMTUD_ENABLE) {
374 		if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
375 			printk(KERN_WARNING "%s: Reported pmtu %d too low, "
376 			       "using default minimum of %d\n",
377 			       __FUNCTION__, pmtu,
378 			       SCTP_DEFAULT_MINSEGMENT);
379 			/* Use default minimum segment size and disable
380 			 * pmtu discovery on this transport.
381 			 */
382 			t->pathmtu = SCTP_DEFAULT_MINSEGMENT;
383 			t->param_flags = (t->param_flags & ~SPP_PMTUD) |
384 				SPP_PMTUD_DISABLE;
385 		} else {
386 			t->pathmtu = pmtu;
387 		}
388 
389 		/* Update association pmtu. */
390 		sctp_assoc_sync_pmtu(asoc);
391 	}
392 
393 	/* Retransmit with the new pmtu setting.
394 	 * Normally, if PMTU discovery is disabled, an ICMP Fragmentation
395 	 * Needed will never be sent, but if a message was sent before
396 	 * PMTU discovery was disabled that was larger than the PMTU, it
397 	 * would not be fragmented, so it must be re-transmitted fragmented.
398 	 */
399 	sctp_retransmit(&asoc->outqueue, t, SCTP_RTXR_PMTUD);
400 }
401 
402 /*
403  * SCTP Implementer's Guide, 2.37 ICMP handling procedures
404  *
405  * ICMP8) If the ICMP code is a "Unrecognized next header type encountered"
406  *        or a "Protocol Unreachable" treat this message as an abort
407  *        with the T bit set.
408  *
409  * This function sends an event to the state machine, which will abort the
410  * association.
411  *
412  */
413 void sctp_icmp_proto_unreachable(struct sock *sk,
414 			   struct sctp_association *asoc,
415 			   struct sctp_transport *t)
416 {
417 	SCTP_DEBUG_PRINTK("%s\n",  __FUNCTION__);
418 
419 	sctp_do_sm(SCTP_EVENT_T_OTHER,
420 		   SCTP_ST_OTHER(SCTP_EVENT_ICMP_PROTO_UNREACH),
421 		   asoc->state, asoc->ep, asoc, t,
422 		   GFP_ATOMIC);
423 
424 }
425 
426 /* Common lookup code for icmp/icmpv6 error handler. */
427 struct sock *sctp_err_lookup(int family, struct sk_buff *skb,
428 			     struct sctphdr *sctphdr,
429 			     struct sctp_association **app,
430 			     struct sctp_transport **tpp)
431 {
432 	union sctp_addr saddr;
433 	union sctp_addr daddr;
434 	struct sctp_af *af;
435 	struct sock *sk = NULL;
436 	struct sctp_association *asoc;
437 	struct sctp_transport *transport = NULL;
438 
439 	*app = NULL; *tpp = NULL;
440 
441 	af = sctp_get_af_specific(family);
442 	if (unlikely(!af)) {
443 		return NULL;
444 	}
445 
446 	/* Initialize local addresses for lookups. */
447 	af->from_skb(&saddr, skb, 1);
448 	af->from_skb(&daddr, skb, 0);
449 
450 	/* Look for an association that matches the incoming ICMP error
451 	 * packet.
452 	 */
453 	asoc = __sctp_lookup_association(&saddr, &daddr, &transport);
454 	if (!asoc)
455 		return NULL;
456 
457 	sk = asoc->base.sk;
458 
459 	if (ntohl(sctphdr->vtag) != asoc->c.peer_vtag) {
460 		ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
461 		goto out;
462 	}
463 
464 	sctp_bh_lock_sock(sk);
465 
466 	/* If too many ICMPs get dropped on busy
467 	 * servers this needs to be solved differently.
468 	 */
469 	if (sock_owned_by_user(sk))
470 		NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
471 
472 	*app = asoc;
473 	*tpp = transport;
474 	return sk;
475 
476 out:
477 	if (asoc)
478 		sctp_association_put(asoc);
479 	return NULL;
480 }
481 
482 /* Common cleanup code for icmp/icmpv6 error handler. */
483 void sctp_err_finish(struct sock *sk, struct sctp_association *asoc)
484 {
485 	sctp_bh_unlock_sock(sk);
486 	if (asoc)
487 		sctp_association_put(asoc);
488 }
489 
490 /*
491  * This routine is called by the ICMP module when it gets some
492  * sort of error condition.  If err < 0 then the socket should
493  * be closed and the error returned to the user.  If err > 0
494  * it's just the icmp type << 8 | icmp code.  After adjustment
495  * header points to the first 8 bytes of the sctp header.  We need
496  * to find the appropriate port.
497  *
498  * The locking strategy used here is very "optimistic". When
499  * someone else accesses the socket the ICMP is just dropped
500  * and for some paths there is no check at all.
501  * A more general error queue to queue errors for later handling
502  * is probably better.
503  *
504  */
505 void sctp_v4_err(struct sk_buff *skb, __u32 info)
506 {
507 	struct iphdr *iph = (struct iphdr *)skb->data;
508 	const int ihlen = iph->ihl * 4;
509 	const int type = icmp_hdr(skb)->type;
510 	const int code = icmp_hdr(skb)->code;
511 	struct sock *sk;
512 	struct sctp_association *asoc = NULL;
513 	struct sctp_transport *transport;
514 	struct inet_sock *inet;
515 	sk_buff_data_t saveip, savesctp;
516 	int err;
517 
518 	if (skb->len < ihlen + 8) {
519 		ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
520 		return;
521 	}
522 
523 	/* Fix up skb to look at the embedded net header. */
524 	saveip = skb->network_header;
525 	savesctp = skb->transport_header;
526 	skb_reset_network_header(skb);
527 	skb_set_transport_header(skb, ihlen);
528 	sk = sctp_err_lookup(AF_INET, skb, sctp_hdr(skb), &asoc, &transport);
529 	/* Put back, the original values. */
530 	skb->network_header = saveip;
531 	skb->transport_header = savesctp;
532 	if (!sk) {
533 		ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
534 		return;
535 	}
536 	/* Warning:  The sock lock is held.  Remember to call
537 	 * sctp_err_finish!
538 	 */
539 
540 	switch (type) {
541 	case ICMP_PARAMETERPROB:
542 		err = EPROTO;
543 		break;
544 	case ICMP_DEST_UNREACH:
545 		if (code > NR_ICMP_UNREACH)
546 			goto out_unlock;
547 
548 		/* PMTU discovery (RFC1191) */
549 		if (ICMP_FRAG_NEEDED == code) {
550 			sctp_icmp_frag_needed(sk, asoc, transport, info);
551 			goto out_unlock;
552 		}
553 		else {
554 			if (ICMP_PROT_UNREACH == code) {
555 				sctp_icmp_proto_unreachable(sk, asoc,
556 							    transport);
557 				goto out_unlock;
558 			}
559 		}
560 		err = icmp_err_convert[code].errno;
561 		break;
562 	case ICMP_TIME_EXCEEDED:
563 		/* Ignore any time exceeded errors due to fragment reassembly
564 		 * timeouts.
565 		 */
566 		if (ICMP_EXC_FRAGTIME == code)
567 			goto out_unlock;
568 
569 		err = EHOSTUNREACH;
570 		break;
571 	default:
572 		goto out_unlock;
573 	}
574 
575 	inet = inet_sk(sk);
576 	if (!sock_owned_by_user(sk) && inet->recverr) {
577 		sk->sk_err = err;
578 		sk->sk_error_report(sk);
579 	} else {  /* Only an error on timeout */
580 		sk->sk_err_soft = err;
581 	}
582 
583 out_unlock:
584 	sctp_err_finish(sk, asoc);
585 }
586 
587 /*
588  * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
589  *
590  * This function scans all the chunks in the OOTB packet to determine if
591  * the packet should be discarded right away.  If a response might be needed
592  * for this packet, or, if further processing is possible, the packet will
593  * be queued to a proper inqueue for the next phase of handling.
594  *
595  * Output:
596  * Return 0 - If further processing is needed.
597  * Return 1 - If the packet can be discarded right away.
598  */
599 int sctp_rcv_ootb(struct sk_buff *skb)
600 {
601 	sctp_chunkhdr_t *ch;
602 	__u8 *ch_end;
603 	sctp_errhdr_t *err;
604 
605 	ch = (sctp_chunkhdr_t *) skb->data;
606 
607 	/* Scan through all the chunks in the packet.  */
608 	do {
609 		/* Break out if chunk length is less then minimal. */
610 		if (ntohs(ch->length) < sizeof(sctp_chunkhdr_t))
611 			break;
612 
613 		ch_end = ((__u8 *)ch) + WORD_ROUND(ntohs(ch->length));
614 		if (ch_end > skb_tail_pointer(skb))
615 			break;
616 
617 		/* RFC 8.4, 2) If the OOTB packet contains an ABORT chunk, the
618 		 * receiver MUST silently discard the OOTB packet and take no
619 		 * further action.
620 		 */
621 		if (SCTP_CID_ABORT == ch->type)
622 			goto discard;
623 
624 		/* RFC 8.4, 6) If the packet contains a SHUTDOWN COMPLETE
625 		 * chunk, the receiver should silently discard the packet
626 		 * and take no further action.
627 		 */
628 		if (SCTP_CID_SHUTDOWN_COMPLETE == ch->type)
629 			goto discard;
630 
631 		/* RFC 8.4, 7) If the packet contains a "Stale cookie" ERROR
632 		 * or a COOKIE ACK the SCTP Packet should be silently
633 		 * discarded.
634 		 */
635 		if (SCTP_CID_COOKIE_ACK == ch->type)
636 			goto discard;
637 
638 		if (SCTP_CID_ERROR == ch->type) {
639 			sctp_walk_errors(err, ch) {
640 				if (SCTP_ERROR_STALE_COOKIE == err->cause)
641 					goto discard;
642 			}
643 		}
644 
645 		ch = (sctp_chunkhdr_t *) ch_end;
646 	} while (ch_end < skb_tail_pointer(skb));
647 
648 	return 0;
649 
650 discard:
651 	return 1;
652 }
653 
654 /* Insert endpoint into the hash table.  */
655 static void __sctp_hash_endpoint(struct sctp_endpoint *ep)
656 {
657 	struct sctp_ep_common **epp;
658 	struct sctp_ep_common *epb;
659 	struct sctp_hashbucket *head;
660 
661 	epb = &ep->base;
662 
663 	epb->hashent = sctp_ep_hashfn(epb->bind_addr.port);
664 	head = &sctp_ep_hashtable[epb->hashent];
665 
666 	sctp_write_lock(&head->lock);
667 	epp = &head->chain;
668 	epb->next = *epp;
669 	if (epb->next)
670 		(*epp)->pprev = &epb->next;
671 	*epp = epb;
672 	epb->pprev = epp;
673 	sctp_write_unlock(&head->lock);
674 }
675 
676 /* Add an endpoint to the hash. Local BH-safe. */
677 void sctp_hash_endpoint(struct sctp_endpoint *ep)
678 {
679 	sctp_local_bh_disable();
680 	__sctp_hash_endpoint(ep);
681 	sctp_local_bh_enable();
682 }
683 
684 /* Remove endpoint from the hash table.  */
685 static void __sctp_unhash_endpoint(struct sctp_endpoint *ep)
686 {
687 	struct sctp_hashbucket *head;
688 	struct sctp_ep_common *epb;
689 
690 	epb = &ep->base;
691 
692 	epb->hashent = sctp_ep_hashfn(epb->bind_addr.port);
693 
694 	head = &sctp_ep_hashtable[epb->hashent];
695 
696 	sctp_write_lock(&head->lock);
697 
698 	if (epb->pprev) {
699 		if (epb->next)
700 			epb->next->pprev = epb->pprev;
701 		*epb->pprev = epb->next;
702 		epb->pprev = NULL;
703 	}
704 
705 	sctp_write_unlock(&head->lock);
706 }
707 
708 /* Remove endpoint from the hash.  Local BH-safe. */
709 void sctp_unhash_endpoint(struct sctp_endpoint *ep)
710 {
711 	sctp_local_bh_disable();
712 	__sctp_unhash_endpoint(ep);
713 	sctp_local_bh_enable();
714 }
715 
716 /* Look up an endpoint. */
717 static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(const union sctp_addr *laddr)
718 {
719 	struct sctp_hashbucket *head;
720 	struct sctp_ep_common *epb;
721 	struct sctp_endpoint *ep;
722 	int hash;
723 
724 	hash = sctp_ep_hashfn(ntohs(laddr->v4.sin_port));
725 	head = &sctp_ep_hashtable[hash];
726 	read_lock(&head->lock);
727 	for (epb = head->chain; epb; epb = epb->next) {
728 		ep = sctp_ep(epb);
729 		if (sctp_endpoint_is_match(ep, laddr))
730 			goto hit;
731 	}
732 
733 	ep = sctp_sk((sctp_get_ctl_sock()))->ep;
734 	epb = &ep->base;
735 
736 hit:
737 	sctp_endpoint_hold(ep);
738 	read_unlock(&head->lock);
739 	return ep;
740 }
741 
742 /* Insert association into the hash table.  */
743 static void __sctp_hash_established(struct sctp_association *asoc)
744 {
745 	struct sctp_ep_common **epp;
746 	struct sctp_ep_common *epb;
747 	struct sctp_hashbucket *head;
748 
749 	epb = &asoc->base;
750 
751 	/* Calculate which chain this entry will belong to. */
752 	epb->hashent = sctp_assoc_hashfn(epb->bind_addr.port, asoc->peer.port);
753 
754 	head = &sctp_assoc_hashtable[epb->hashent];
755 
756 	sctp_write_lock(&head->lock);
757 	epp = &head->chain;
758 	epb->next = *epp;
759 	if (epb->next)
760 		(*epp)->pprev = &epb->next;
761 	*epp = epb;
762 	epb->pprev = epp;
763 	sctp_write_unlock(&head->lock);
764 }
765 
766 /* Add an association to the hash. Local BH-safe. */
767 void sctp_hash_established(struct sctp_association *asoc)
768 {
769 	if (asoc->temp)
770 		return;
771 
772 	sctp_local_bh_disable();
773 	__sctp_hash_established(asoc);
774 	sctp_local_bh_enable();
775 }
776 
777 /* Remove association from the hash table.  */
778 static void __sctp_unhash_established(struct sctp_association *asoc)
779 {
780 	struct sctp_hashbucket *head;
781 	struct sctp_ep_common *epb;
782 
783 	epb = &asoc->base;
784 
785 	epb->hashent = sctp_assoc_hashfn(epb->bind_addr.port,
786 					 asoc->peer.port);
787 
788 	head = &sctp_assoc_hashtable[epb->hashent];
789 
790 	sctp_write_lock(&head->lock);
791 
792 	if (epb->pprev) {
793 		if (epb->next)
794 			epb->next->pprev = epb->pprev;
795 		*epb->pprev = epb->next;
796 		epb->pprev = NULL;
797 	}
798 
799 	sctp_write_unlock(&head->lock);
800 }
801 
802 /* Remove association from the hash table.  Local BH-safe. */
803 void sctp_unhash_established(struct sctp_association *asoc)
804 {
805 	if (asoc->temp)
806 		return;
807 
808 	sctp_local_bh_disable();
809 	__sctp_unhash_established(asoc);
810 	sctp_local_bh_enable();
811 }
812 
813 /* Look up an association. */
814 static struct sctp_association *__sctp_lookup_association(
815 					const union sctp_addr *local,
816 					const union sctp_addr *peer,
817 					struct sctp_transport **pt)
818 {
819 	struct sctp_hashbucket *head;
820 	struct sctp_ep_common *epb;
821 	struct sctp_association *asoc;
822 	struct sctp_transport *transport;
823 	int hash;
824 
825 	/* Optimize here for direct hit, only listening connections can
826 	 * have wildcards anyways.
827 	 */
828 	hash = sctp_assoc_hashfn(ntohs(local->v4.sin_port), ntohs(peer->v4.sin_port));
829 	head = &sctp_assoc_hashtable[hash];
830 	read_lock(&head->lock);
831 	for (epb = head->chain; epb; epb = epb->next) {
832 		asoc = sctp_assoc(epb);
833 		transport = sctp_assoc_is_match(asoc, local, peer);
834 		if (transport)
835 			goto hit;
836 	}
837 
838 	read_unlock(&head->lock);
839 
840 	return NULL;
841 
842 hit:
843 	*pt = transport;
844 	sctp_association_hold(asoc);
845 	read_unlock(&head->lock);
846 	return asoc;
847 }
848 
849 /* Look up an association. BH-safe. */
850 SCTP_STATIC
851 struct sctp_association *sctp_lookup_association(const union sctp_addr *laddr,
852 						 const union sctp_addr *paddr,
853 					    struct sctp_transport **transportp)
854 {
855 	struct sctp_association *asoc;
856 
857 	sctp_local_bh_disable();
858 	asoc = __sctp_lookup_association(laddr, paddr, transportp);
859 	sctp_local_bh_enable();
860 
861 	return asoc;
862 }
863 
864 /* Is there an association matching the given local and peer addresses? */
865 int sctp_has_association(const union sctp_addr *laddr,
866 			 const union sctp_addr *paddr)
867 {
868 	struct sctp_association *asoc;
869 	struct sctp_transport *transport;
870 
871 	if ((asoc = sctp_lookup_association(laddr, paddr, &transport))) {
872 		sctp_association_put(asoc);
873 		return 1;
874 	}
875 
876 	return 0;
877 }
878 
879 /*
880  * SCTP Implementors Guide, 2.18 Handling of address
881  * parameters within the INIT or INIT-ACK.
882  *
883  * D) When searching for a matching TCB upon reception of an INIT
884  *    or INIT-ACK chunk the receiver SHOULD use not only the
885  *    source address of the packet (containing the INIT or
886  *    INIT-ACK) but the receiver SHOULD also use all valid
887  *    address parameters contained within the chunk.
888  *
889  * 2.18.3 Solution description
890  *
891  * This new text clearly specifies to an implementor the need
892  * to look within the INIT or INIT-ACK. Any implementation that
893  * does not do this, may not be able to establish associations
894  * in certain circumstances.
895  *
896  */
897 static struct sctp_association *__sctp_rcv_init_lookup(struct sk_buff *skb,
898 	const union sctp_addr *laddr, struct sctp_transport **transportp)
899 {
900 	struct sctp_association *asoc;
901 	union sctp_addr addr;
902 	union sctp_addr *paddr = &addr;
903 	struct sctphdr *sh = sctp_hdr(skb);
904 	sctp_chunkhdr_t *ch;
905 	union sctp_params params;
906 	sctp_init_chunk_t *init;
907 	struct sctp_transport *transport;
908 	struct sctp_af *af;
909 
910 	ch = (sctp_chunkhdr_t *) skb->data;
911 
912 	/* If this is INIT/INIT-ACK look inside the chunk too. */
913 	switch (ch->type) {
914 	case SCTP_CID_INIT:
915 	case SCTP_CID_INIT_ACK:
916 		break;
917 	default:
918 		return NULL;
919 	}
920 
921 	/* The code below will attempt to walk the chunk and extract
922 	 * parameter information.  Before we do that, we need to verify
923 	 * that the chunk length doesn't cause overflow.  Otherwise, we'll
924 	 * walk off the end.
925 	 */
926 	if (WORD_ROUND(ntohs(ch->length)) > skb->len)
927 		return NULL;
928 
929 	/*
930 	 * This code will NOT touch anything inside the chunk--it is
931 	 * strictly READ-ONLY.
932 	 *
933 	 * RFC 2960 3  SCTP packet Format
934 	 *
935 	 * Multiple chunks can be bundled into one SCTP packet up to
936 	 * the MTU size, except for the INIT, INIT ACK, and SHUTDOWN
937 	 * COMPLETE chunks.  These chunks MUST NOT be bundled with any
938 	 * other chunk in a packet.  See Section 6.10 for more details
939 	 * on chunk bundling.
940 	 */
941 
942 	/* Find the start of the TLVs and the end of the chunk.  This is
943 	 * the region we search for address parameters.
944 	 */
945 	init = (sctp_init_chunk_t *)skb->data;
946 
947 	/* Walk the parameters looking for embedded addresses. */
948 	sctp_walk_params(params, init, init_hdr.params) {
949 
950 		/* Note: Ignoring hostname addresses. */
951 		af = sctp_get_af_specific(param_type2af(params.p->type));
952 		if (!af)
953 			continue;
954 
955 		af->from_addr_param(paddr, params.addr, sh->source, 0);
956 
957 		asoc = __sctp_lookup_association(laddr, paddr, &transport);
958 		if (asoc)
959 			return asoc;
960 	}
961 
962 	return NULL;
963 }
964 
965 /* Lookup an association for an inbound skb. */
966 static struct sctp_association *__sctp_rcv_lookup(struct sk_buff *skb,
967 				      const union sctp_addr *paddr,
968 				      const union sctp_addr *laddr,
969 				      struct sctp_transport **transportp)
970 {
971 	struct sctp_association *asoc;
972 
973 	asoc = __sctp_lookup_association(laddr, paddr, transportp);
974 
975 	/* Further lookup for INIT/INIT-ACK packets.
976 	 * SCTP Implementors Guide, 2.18 Handling of address
977 	 * parameters within the INIT or INIT-ACK.
978 	 */
979 	if (!asoc)
980 		asoc = __sctp_rcv_init_lookup(skb, laddr, transportp);
981 
982 	return asoc;
983 }
984