blob: 1c23060c41a669f38f4e2d244cfb61c85a522be6 [file] [log] [blame]
Vlad Yasevich60c778b2008-01-11 09:57:09 -05001/* SCTP kernel implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001 Intel Corp.
6 * Copyright (c) 2001 La Monte H.P. Yarroll
7 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -05008 * This file is part of the SCTP kernel implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 *
10 * This module provides the abstraction for an SCTP association.
11 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -050012 * This SCTP implementation is free software;
Linus Torvalds1da177e2005-04-16 15:20:36 -070013 * you can redistribute it and/or modify it under the terms of
14 * the GNU General Public License as published by
15 * the Free Software Foundation; either version 2, or (at your option)
16 * any later version.
17 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -050018 * This SCTP implementation is distributed in the hope that it
Linus Torvalds1da177e2005-04-16 15:20:36 -070019 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
20 * ************************
21 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
22 * See the GNU General Public License for more details.
23 *
24 * You should have received a copy of the GNU General Public License
Jeff Kirsher4b2f13a2013-12-06 06:28:48 -080025 * along with GNU CC; see the file COPYING. If not, see
26 * <http://www.gnu.org/licenses/>.
Linus Torvalds1da177e2005-04-16 15:20:36 -070027 *
28 * Please send any bug reports or fixes you make to the
29 * email address(es):
Daniel Borkmann91705c62013-07-23 14:51:47 +020030 * lksctp developers <linux-sctp@vger.kernel.org>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070032 * Written or modified by:
33 * La Monte H.P. Yarroll <piggy@acm.org>
34 * Karl Knutson <karl@athena.chicago.il.us>
35 * Jon Grimm <jgrimm@us.ibm.com>
36 * Xingang Guo <xingang.guo@intel.com>
37 * Hui Huang <hui.huang@nokia.com>
38 * Sridhar Samudrala <sri@us.ibm.com>
39 * Daisy Chang <daisyc@us.ibm.com>
40 * Ryan Layer <rmlayer@us.ibm.com>
41 * Kevin Gao <kevin.gao@intel.com>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042 */
43
Joe Perches145ce502010-08-24 13:21:08 +000044#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
45
Linus Torvalds1da177e2005-04-16 15:20:36 -070046#include <linux/types.h>
47#include <linux/fcntl.h>
48#include <linux/poll.h>
49#include <linux/init.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
51#include <linux/slab.h>
52#include <linux/in.h>
53#include <net/ipv6.h>
54#include <net/sctp/sctp.h>
55#include <net/sctp/sm.h>
56
57/* Forward declarations for internal functions. */
Daniel Borkmannb82e8f32014-06-11 18:19:29 +020058static void sctp_select_active_and_retran_path(struct sctp_association *asoc);
David Howellsc4028952006-11-22 14:57:56 +000059static void sctp_assoc_bh_rcv(struct work_struct *work);
Vlad Yasevicha08de642007-12-20 14:11:47 -080060static void sctp_assoc_free_asconf_acks(struct sctp_association *asoc);
Wei Yongjun8b4472c2011-05-24 21:48:02 +000061static void sctp_assoc_free_asconf_queue(struct sctp_association *asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
Linus Torvalds1da177e2005-04-16 15:20:36 -070063/* 1st Level Abstractions. */
64
65/* Initialize a new association from provided memory. */
66static struct sctp_association *sctp_association_init(struct sctp_association *asoc,
67 const struct sctp_endpoint *ep,
68 const struct sock *sk,
69 sctp_scope_t scope,
Al Virodd0fc662005-10-07 07:46:04 +010070 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070071{
Eric W. Biedermane1fc3b12012-08-07 07:29:57 +000072 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 struct sctp_sock *sp;
74 int i;
Vlad Yasevicha29a5bd2007-09-16 19:31:35 -070075 sctp_paramhdr_t *p;
76 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -070077
78 /* Retrieve the SCTP per socket area. */
79 sp = sctp_sk((struct sock *)sk);
80
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 /* Discarding const is appropriate here. */
82 asoc->ep = (struct sctp_endpoint *)ep;
Linus Torvalds1da177e2005-04-16 15:20:36 -070083 asoc->base.sk = (struct sock *)sk;
Daniel Borkmann2e0c9e72013-06-14 18:24:07 +020084
85 sctp_endpoint_hold(asoc->ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -070086 sock_hold(asoc->base.sk);
87
88 /* Initialize the common base substructure. */
89 asoc->base.type = SCTP_EP_TYPE_ASSOCIATION;
90
91 /* Initialize the object handling fields. */
92 atomic_set(&asoc->base.refcnt, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -070093
94 /* Initialize the bind addr area. */
95 sctp_bind_addr_init(&asoc->base.bind_addr, ep->base.bind_addr.port);
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
97 asoc->state = SCTP_STATE_CLOSED;
Daniel Borkmann52db8822013-06-25 18:17:27 +020098 asoc->cookie_life = ms_to_ktime(sp->assocparams.sasoc_cookie_life);
Vlad Yasevichf68b2e02009-09-04 18:21:00 -040099 asoc->user_frag = sp->user_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
101 /* Set the association max_retrans and RTO values from the
102 * socket values.
103 */
104 asoc->max_retrans = sp->assocparams.sasoc_asocmaxrxt;
Eric W. Biedermane1fc3b12012-08-07 07:29:57 +0000105 asoc->pf_retrans = net->sctp.pf_retrans;
Neil Horman5aa93bc2012-07-21 07:56:07 +0000106
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 asoc->rto_initial = msecs_to_jiffies(sp->rtoinfo.srto_initial);
108 asoc->rto_max = msecs_to_jiffies(sp->rtoinfo.srto_max);
109 asoc->rto_min = msecs_to_jiffies(sp->rtoinfo.srto_min);
110
Frank Filz52ccb8e2005-12-22 11:36:46 -0800111 /* Initialize the association's heartbeat interval based on the
112 * sock configured value.
113 */
114 asoc->hbinterval = msecs_to_jiffies(sp->hbinterval);
115
116 /* Initialize path max retrans value. */
117 asoc->pathmaxrxt = sp->pathmaxrxt;
118
119 /* Initialize default path MTU. */
120 asoc->pathmtu = sp->pathmtu;
121
122 /* Set association default SACK delay */
123 asoc->sackdelay = msecs_to_jiffies(sp->sackdelay);
Wei Yongjund364d922008-05-09 15:13:26 -0700124 asoc->sackfreq = sp->sackfreq;
Frank Filz52ccb8e2005-12-22 11:36:46 -0800125
126 /* Set the association default flags controlling
127 * Heartbeat, SACK delay, and Path MTU Discovery.
128 */
129 asoc->param_flags = sp->param_flags;
130
wangweidong9d2c8812013-12-06 09:36:30 +0800131 /* Initialize the maximum number of new data packets that can be sent
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 * in a burst.
133 */
Vlad Yasevich70331572007-03-23 11:34:36 -0700134 asoc->max_burst = sp->max_burst;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135
Vladislav Yasevich1e7d3d92005-11-11 16:06:16 -0800136 /* initialize association timers */
Vladislav Yasevich1e7d3d92005-11-11 16:06:16 -0800137 asoc->timeouts[SCTP_EVENT_TIMEOUT_T1_COOKIE] = asoc->rto_initial;
138 asoc->timeouts[SCTP_EVENT_TIMEOUT_T1_INIT] = asoc->rto_initial;
139 asoc->timeouts[SCTP_EVENT_TIMEOUT_T2_SHUTDOWN] = asoc->rto_initial;
Vladislav Yasevich1e7d3d92005-11-11 16:06:16 -0800140
141 /* sctpimpguide Section 2.12.2
142 * If the 'T5-shutdown-guard' timer is used, it SHOULD be set to the
143 * recommended value of 5 times 'RTO.Max'.
144 */
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900145 asoc->timeouts[SCTP_EVENT_TIMEOUT_T5_SHUTDOWN_GUARD]
Vladislav Yasevich1e7d3d92005-11-11 16:06:16 -0800146 = 5 * asoc->rto_max;
147
Frank Filz52ccb8e2005-12-22 11:36:46 -0800148 asoc->timeouts[SCTP_EVENT_TIMEOUT_SACK] = asoc->sackdelay;
Neil Horman9f70f462013-12-10 06:48:15 -0500149 asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE] = sp->autoclose * HZ;
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900150
Uwe Kleine-König421f91d2010-06-11 12:17:00 +0200151 /* Initializes the timers */
Pavel Emelyanovb24b8a22008-01-23 21:20:07 -0800152 for (i = SCTP_EVENT_TIMEOUT_NONE; i < SCTP_NUM_TIMEOUT_TYPES; ++i)
153 setup_timer(&asoc->timers[i], sctp_timer_events[i],
154 (unsigned long)asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155
156 /* Pull default initialization values from the sock options.
157 * Note: This assumes that the values have already been
158 * validated in the sock.
159 */
160 asoc->c.sinit_max_instreams = sp->initmsg.sinit_max_instreams;
161 asoc->c.sinit_num_ostreams = sp->initmsg.sinit_num_ostreams;
162 asoc->max_init_attempts = sp->initmsg.sinit_max_attempts;
163
164 asoc->max_init_timeo =
165 msecs_to_jiffies(sp->initmsg.sinit_max_init_timeo);
166
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167 /* Set the local window size for receive.
168 * This is also the rcvbuf space per association.
169 * RFC 6 - A SCTP receiver MUST be able to receive a minimum of
170 * 1500 bytes in one SCTP packet.
171 */
Neil Horman049b3ff2005-11-11 16:08:24 -0800172 if ((sk->sk_rcvbuf/2) < SCTP_DEFAULT_MINWINDOW)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 asoc->rwnd = SCTP_DEFAULT_MINWINDOW;
174 else
Neil Horman049b3ff2005-11-11 16:08:24 -0800175 asoc->rwnd = sk->sk_rcvbuf/2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176
177 asoc->a_rwnd = asoc->rwnd;
178
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 /* Use my own max window until I learn something better. */
180 asoc->peer.rwnd = SCTP_DEFAULT_MAXWINDOW;
181
Neil Horman049b3ff2005-11-11 16:08:24 -0800182 /* Initialize the receive memory counter */
183 atomic_set(&asoc->rmem_alloc, 0);
184
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 init_waitqueue_head(&asoc->wait);
186
187 asoc->c.my_vtag = sctp_generate_tag(ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 asoc->c.my_port = ep->base.bind_addr.port;
189
190 asoc->c.initial_tsn = sctp_generate_tsn(ep);
191
192 asoc->next_tsn = asoc->c.initial_tsn;
193
194 asoc->ctsn_ack_point = asoc->next_tsn - 1;
195 asoc->adv_peer_ack_point = asoc->ctsn_ack_point;
196 asoc->highest_sacked = asoc->ctsn_ack_point;
197 asoc->last_cwr_tsn = asoc->ctsn_ack_point;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 /* ADDIP Section 4.1 Asconf Chunk Procedures
200 *
201 * When an endpoint has an ASCONF signaled change to be sent to the
202 * remote endpoint it should do the following:
203 * ...
204 * A2) a serial number should be assigned to the chunk. The serial
205 * number SHOULD be a monotonically increasing number. The serial
206 * numbers SHOULD be initialized at the start of the
207 * association to the same value as the initial TSN.
208 */
209 asoc->addip_serial = asoc->c.initial_tsn;
210
David S. Miller79af02c2005-07-08 21:47:49 -0700211 INIT_LIST_HEAD(&asoc->addip_chunk_list);
Vlad Yasevicha08de642007-12-20 14:11:47 -0800212 INIT_LIST_HEAD(&asoc->asconf_ack_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213
214 /* Make an empty list of remote transport addresses. */
215 INIT_LIST_HEAD(&asoc->peer.transport_addr_list);
216
217 /* RFC 2960 5.1 Normal Establishment of an Association
218 *
219 * After the reception of the first data chunk in an
220 * association the endpoint must immediately respond with a
221 * sack to acknowledge the data chunk. Subsequent
222 * acknowledgements should be done as described in Section
223 * 6.2.
224 *
225 * [We implement this by telling a new association that it
226 * already received one packet.]
227 */
228 asoc->peer.sack_needed = 1;
Neil Horman42448542012-06-30 03:04:26 +0000229 asoc->peer.sack_generation = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230
Vlad Yasevich73d9c4f2007-10-24 17:24:26 -0400231 /* Assume that the peer will tell us if he recognizes ASCONF
232 * as part of INIT exchange.
wangweidong9d2c8812013-12-06 09:36:30 +0800233 * The sctp_addip_noauth option is there for backward compatibility
Vlad Yasevich73d9c4f2007-10-24 17:24:26 -0400234 * and will revert old behavior.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 */
Eric W. Biedermane1fc3b12012-08-07 07:29:57 +0000236 if (net->sctp.addip_noauth)
Vlad Yasevich73d9c4f2007-10-24 17:24:26 -0400237 asoc->peer.asconf_capable = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238
239 /* Create an input queue. */
240 sctp_inq_init(&asoc->base.inqueue);
David Howellsc4028952006-11-22 14:57:56 +0000241 sctp_inq_set_th_handler(&asoc->base.inqueue, sctp_assoc_bh_rcv);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242
243 /* Create an output queue. */
244 sctp_outq_init(asoc, &asoc->outqueue);
245
246 if (!sctp_ulpq_init(&asoc->ulpq, asoc))
247 goto fail_init;
248
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 /* Assume that peer would support both address types unless we are
250 * told otherwise.
251 */
252 asoc->peer.ipv4_address = 1;
Wei Yongjuna2c39582009-04-07 16:35:11 +0800253 if (asoc->base.sk->sk_family == PF_INET6)
254 asoc->peer.ipv6_address = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 INIT_LIST_HEAD(&asoc->asocs);
256
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 asoc->default_stream = sp->default_stream;
258 asoc->default_ppid = sp->default_ppid;
259 asoc->default_flags = sp->default_flags;
260 asoc->default_context = sp->default_context;
261 asoc->default_timetolive = sp->default_timetolive;
Ivan Skytte Jorgensen6ab792f2006-12-13 16:34:22 -0800262 asoc->default_rcv_context = sp->default_rcv_context;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263
Vlad Yasevicha29a5bd2007-09-16 19:31:35 -0700264 /* AUTH related initializations */
265 INIT_LIST_HEAD(&asoc->endpoint_shared_keys);
266 err = sctp_auth_asoc_copy_shkeys(ep, asoc, gfp);
267 if (err)
268 goto fail_init;
269
270 asoc->active_key_id = ep->active_key_id;
Xin Long28aa4c22016-07-09 19:47:40 +0800271 asoc->prsctp_enable = ep->prsctp_enable;
Vlad Yasevicha29a5bd2007-09-16 19:31:35 -0700272
Vlad Yasevicha29a5bd2007-09-16 19:31:35 -0700273 /* Save the hmacs and chunks list into this association */
274 if (ep->auth_hmacs_list)
275 memcpy(asoc->c.auth_hmacs, ep->auth_hmacs_list,
276 ntohs(ep->auth_hmacs_list->param_hdr.length));
277 if (ep->auth_chunk_list)
278 memcpy(asoc->c.auth_chunks, ep->auth_chunk_list,
279 ntohs(ep->auth_chunk_list->param_hdr.length));
280
281 /* Get the AUTH random number for this association */
282 p = (sctp_paramhdr_t *)asoc->c.auth_random;
283 p->type = SCTP_PARAM_RANDOM;
284 p->length = htons(sizeof(sctp_paramhdr_t) + SCTP_AUTH_RANDOM_LENGTH);
285 get_random_bytes(p+1, SCTP_AUTH_RANDOM_LENGTH);
286
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 return asoc;
288
289fail_init:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 sock_put(asoc->base.sk);
Daniel Borkmann2e0c9e72013-06-14 18:24:07 +0200291 sctp_endpoint_put(asoc->ep);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 return NULL;
293}
294
295/* Allocate and initialize a new association */
296struct sctp_association *sctp_association_new(const struct sctp_endpoint *ep,
297 const struct sock *sk,
Alexey Dobriyan3182cd82005-07-11 20:57:47 -0700298 sctp_scope_t scope,
Al Virodd0fc662005-10-07 07:46:04 +0100299 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300{
301 struct sctp_association *asoc;
302
Daniel Borkmann939cfa72013-06-17 11:40:04 +0200303 asoc = kzalloc(sizeof(*asoc), gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 if (!asoc)
305 goto fail;
306
307 if (!sctp_association_init(asoc, ep, sk, scope, gfp))
308 goto fail_init;
309
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 SCTP_DBG_OBJCNT_INC(assoc);
Daniel Borkmannbb333812013-06-28 19:49:40 +0200311
312 pr_debug("Created asoc %p\n", asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
314 return asoc;
315
316fail_init:
317 kfree(asoc);
318fail:
319 return NULL;
320}
321
322/* Free this association if possible. There may still be users, so
323 * the actual deallocation may be delayed.
324 */
325void sctp_association_free(struct sctp_association *asoc)
326{
327 struct sock *sk = asoc->base.sk;
328 struct sctp_transport *transport;
329 struct list_head *pos, *temp;
330 int i;
331
Vlad Yasevichde76e692006-10-30 18:55:11 -0800332 /* Only real associations count against the endpoint, so
333 * don't bother for if this is a temporary association.
334 */
Xufeng Zhangd3217b12014-06-12 10:53:36 +0800335 if (!list_empty(&asoc->asocs)) {
Vlad Yasevichde76e692006-10-30 18:55:11 -0800336 list_del(&asoc->asocs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337
Vlad Yasevichde76e692006-10-30 18:55:11 -0800338 /* Decrement the backlog value for a TCP-style listening
339 * socket.
340 */
341 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
342 sk->sk_ack_backlog--;
343 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344
345 /* Mark as dead, so other users can know this structure is
346 * going away.
347 */
Daniel Borkmann0022d2d2013-04-15 03:27:18 +0000348 asoc->base.dead = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349
350 /* Dispose of any data lying around in the outqueue. */
351 sctp_outq_free(&asoc->outqueue);
352
353 /* Dispose of any pending messages for the upper layer. */
354 sctp_ulpq_free(&asoc->ulpq);
355
356 /* Dispose of any pending chunks on the inqueue. */
357 sctp_inq_free(&asoc->base.inqueue);
358
Vlad Yasevich8e1ee182008-10-08 14:18:39 -0700359 sctp_tsnmap_free(&asoc->peer.tsn_map);
360
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 /* Free ssnmap storage. */
362 sctp_ssnmap_free(asoc->ssnmap);
363
364 /* Clean up the bound address list. */
365 sctp_bind_addr_free(&asoc->base.bind_addr);
366
367 /* Do we need to go through all of our timers and
368 * delete them? To be safe we will try to delete all, but we
369 * should be able to go through and make a guess based
370 * on our state.
371 */
372 for (i = SCTP_EVENT_TIMEOUT_NONE; i < SCTP_NUM_TIMEOUT_TYPES; ++i) {
Ying Xue25cc4ae2013-02-03 20:32:57 +0000373 if (del_timer(&asoc->timers[i]))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 sctp_association_put(asoc);
375 }
376
377 /* Free peer's cached cookie. */
Jesper Juhla51482b2005-11-08 09:41:34 -0800378 kfree(asoc->peer.cookie);
Vlad Yasevich730fc3d2007-09-16 19:32:11 -0700379 kfree(asoc->peer.peer_random);
380 kfree(asoc->peer.peer_chunks);
381 kfree(asoc->peer.peer_hmacs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382
383 /* Release the transport structures. */
384 list_for_each_safe(pos, temp, &asoc->peer.transport_addr_list) {
385 transport = list_entry(pos, struct sctp_transport, transports);
Thomas Graf45122ca262012-12-06 09:25:05 +0000386 list_del_rcu(pos);
Xin Long4f008782015-12-30 23:50:47 +0800387 sctp_unhash_transport(transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 sctp_transport_free(transport);
389 }
390
Frank Filz3f7a87d2005-06-20 13:14:57 -0700391 asoc->peer.transport_count = 0;
392
Wei Yongjuna000c012011-05-29 23:23:36 +0000393 sctp_asconf_queue_teardown(asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394
Michio Honda8a07eb02011-04-26 20:19:36 +0900395 /* Free pending address space being deleted */
Markus Elfring7d37d0c2015-01-31 18:10:03 +0100396 kfree(asoc->asconf_addr_del_pending);
Michio Honda8a07eb02011-04-26 20:19:36 +0900397
Vlad Yasevicha29a5bd2007-09-16 19:31:35 -0700398 /* AUTH - Free the endpoint shared keys */
399 sctp_auth_destroy_keys(&asoc->endpoint_shared_keys);
400
401 /* AUTH - Free the association shared key */
402 sctp_auth_key_put(asoc->asoc_shared_key);
403
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 sctp_association_put(asoc);
405}
406
407/* Cleanup and free up an association. */
408static void sctp_association_destroy(struct sctp_association *asoc)
409{
Daniel Borkmannbb333812013-06-28 19:49:40 +0200410 if (unlikely(!asoc->base.dead)) {
411 WARN(1, "Attempt to destroy undead association %p!\n", asoc);
412 return;
413 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414
415 sctp_endpoint_put(asoc->ep);
416 sock_put(asoc->base.sk);
417
418 if (asoc->assoc_id != 0) {
419 spin_lock_bh(&sctp_assocs_id_lock);
420 idr_remove(&sctp_assocs_id, asoc->assoc_id);
421 spin_unlock_bh(&sctp_assocs_id_lock);
422 }
423
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700424 WARN_ON(atomic_read(&asoc->rmem_alloc));
Neil Horman049b3ff2005-11-11 16:08:24 -0800425
Daniel Borkmannff2266c2013-04-15 03:27:17 +0000426 kfree(asoc);
427 SCTP_DBG_OBJCNT_DEC(assoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428}
429
430/* Change the primary destination address for the peer. */
431void sctp_assoc_set_primary(struct sctp_association *asoc,
432 struct sctp_transport *transport)
433{
Vlad Yasevich319fa2a2008-06-16 17:00:29 -0700434 int changeover = 0;
435
436 /* it's a changeover only if we already have a primary path
437 * that we are changing
438 */
439 if (asoc->peer.primary_path != NULL &&
440 asoc->peer.primary_path != transport)
441 changeover = 1 ;
442
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 asoc->peer.primary_path = transport;
444
445 /* Set a default msg_name for events. */
446 memcpy(&asoc->peer.primary_addr, &transport->ipaddr,
447 sizeof(union sctp_addr));
448
449 /* If the primary path is changing, assume that the
450 * user wants to use this new path.
451 */
Sridhar Samudralaad8fec12006-07-21 14:48:50 -0700452 if ((transport->state == SCTP_ACTIVE) ||
453 (transport->state == SCTP_UNKNOWN))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 asoc->peer.active_path = transport;
455
456 /*
457 * SFR-CACC algorithm:
458 * Upon the receipt of a request to change the primary
459 * destination address, on the data structure for the new
460 * primary destination, the sender MUST do the following:
461 *
462 * 1) If CHANGEOVER_ACTIVE is set, then there was a switch
463 * to this destination address earlier. The sender MUST set
464 * CYCLING_CHANGEOVER to indicate that this switch is a
465 * double switch to the same destination address.
Vlad Yaseviche0e9db12009-11-23 15:53:57 -0500466 *
467 * Really, only bother is we have data queued or outstanding on
468 * the association.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 */
Vlad Yaseviche0e9db12009-11-23 15:53:57 -0500470 if (!asoc->outqueue.outstanding_bytes && !asoc->outqueue.out_qlen)
471 return;
472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 if (transport->cacc.changeover_active)
Vlad Yasevich319fa2a2008-06-16 17:00:29 -0700474 transport->cacc.cycling_changeover = changeover;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475
476 /* 2) The sender MUST set CHANGEOVER_ACTIVE to indicate that
477 * a changeover has occurred.
478 */
Vlad Yasevich319fa2a2008-06-16 17:00:29 -0700479 transport->cacc.changeover_active = changeover;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480
481 /* 3) The sender MUST store the next TSN to be sent in
482 * next_tsn_at_change.
483 */
484 transport->cacc.next_tsn_at_change = asoc->next_tsn;
485}
486
Frank Filz3f7a87d2005-06-20 13:14:57 -0700487/* Remove a transport from an association. */
488void sctp_assoc_rm_peer(struct sctp_association *asoc,
489 struct sctp_transport *peer)
490{
491 struct list_head *pos;
492 struct sctp_transport *transport;
493
Daniel Borkmannbb333812013-06-28 19:49:40 +0200494 pr_debug("%s: association:%p addr:%pISpc\n",
495 __func__, asoc, &peer->ipaddr.sa);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700496
497 /* If we are to remove the current retran_path, update it
498 * to the next peer before removing this peer from the list.
499 */
500 if (asoc->peer.retran_path == peer)
501 sctp_assoc_update_retran_path(asoc);
502
503 /* Remove this peer from the list. */
Thomas Graf45122ca262012-12-06 09:25:05 +0000504 list_del_rcu(&peer->transports);
Xin Long4f008782015-12-30 23:50:47 +0800505 /* Remove this peer from the transport hashtable */
506 sctp_unhash_transport(peer);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700507
508 /* Get the first transport of asoc. */
509 pos = asoc->peer.transport_addr_list.next;
510 transport = list_entry(pos, struct sctp_transport, transports);
511
512 /* Update any entries that match the peer to be deleted. */
513 if (asoc->peer.primary_path == peer)
514 sctp_assoc_set_primary(asoc, transport);
515 if (asoc->peer.active_path == peer)
516 asoc->peer.active_path = transport;
Wei Yongjun9494c7c2011-04-12 15:22:22 +0000517 if (asoc->peer.retran_path == peer)
518 asoc->peer.retran_path = transport;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700519 if (asoc->peer.last_data_from == peer)
520 asoc->peer.last_data_from = transport;
521
522 /* If we remove the transport an INIT was last sent to, set it to
523 * NULL. Combined with the update of the retran path above, this
524 * will cause the next INIT to be sent to the next available
525 * transport, maintaining the cycle.
526 */
527 if (asoc->init_last_sent_to == peer)
528 asoc->init_last_sent_to = NULL;
529
Wei Yongjun6345b192009-04-26 23:13:35 +0800530 /* If we remove the transport an SHUTDOWN was last sent to, set it
531 * to NULL. Combined with the update of the retran path above, this
532 * will cause the next SHUTDOWN to be sent to the next available
533 * transport, maintaining the cycle.
534 */
535 if (asoc->shutdown_last_sent_to == peer)
536 asoc->shutdown_last_sent_to = NULL;
537
Wei Yongjun10a43ce2009-04-26 23:14:42 +0800538 /* If we remove the transport an ASCONF was last sent to, set it to
539 * NULL.
540 */
541 if (asoc->addip_last_asconf &&
542 asoc->addip_last_asconf->transport == peer)
543 asoc->addip_last_asconf->transport = NULL;
544
Vlad Yasevich31b02e12009-09-04 18:21:00 -0400545 /* If we have something on the transmitted list, we have to
546 * save it off. The best place is the active path.
547 */
548 if (!list_empty(&peer->transmitted)) {
549 struct sctp_transport *active = asoc->peer.active_path;
550 struct sctp_chunk *ch;
551
552 /* Reset the transport of each chunk on this list */
553 list_for_each_entry(ch, &peer->transmitted,
554 transmitted_list) {
555 ch->transport = NULL;
556 ch->rtt_in_progress = 0;
557 }
558
559 list_splice_tail_init(&peer->transmitted,
560 &active->transmitted);
561
562 /* Start a T3 timer here in case it wasn't running so
563 * that these migrated packets have a chance to get
wangweidong2bccbad2013-10-26 16:06:30 +0800564 * retransmitted.
Vlad Yasevich31b02e12009-09-04 18:21:00 -0400565 */
566 if (!timer_pending(&active->T3_rtx_timer))
567 if (!mod_timer(&active->T3_rtx_timer,
568 jiffies + active->rto))
569 sctp_transport_hold(active);
570 }
571
Frank Filz3f7a87d2005-06-20 13:14:57 -0700572 asoc->peer.transport_count--;
573
574 sctp_transport_free(peer);
575}
576
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577/* Add a transport address to an association. */
578struct sctp_transport *sctp_assoc_add_peer(struct sctp_association *asoc,
579 const union sctp_addr *addr,
Al Virodd0fc662005-10-07 07:46:04 +0100580 const gfp_t gfp,
Frank Filz3f7a87d2005-06-20 13:14:57 -0700581 const int peer_state)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582{
Eric W. Biederman89bf3452012-08-07 07:26:14 +0000583 struct net *net = sock_net(asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 struct sctp_transport *peer;
585 struct sctp_sock *sp;
586 unsigned short port;
587
588 sp = sctp_sk(asoc->base.sk);
589
590 /* AF_INET and AF_INET6 share common port field. */
Al Viro4bdf4b52006-11-20 17:10:20 -0800591 port = ntohs(addr->v4.sin_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
Daniel Borkmannbb333812013-06-28 19:49:40 +0200593 pr_debug("%s: association:%p addr:%pISpc state:%d\n", __func__,
594 asoc, &addr->sa, peer_state);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700595
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 /* Set the port if it has not been set yet. */
597 if (0 == asoc->peer.port)
598 asoc->peer.port = port;
599
600 /* Check to see if this is a duplicate. */
601 peer = sctp_assoc_lookup_paddr(asoc, addr);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700602 if (peer) {
Vlad Yasevichadd52372008-09-18 16:28:27 -0700603 /* An UNKNOWN state is only set on transports added by
604 * user in sctp_connectx() call. Such transports should be
605 * considered CONFIRMED per RFC 4960, Section 5.4.
606 */
Sridhar Samudralaad8fec12006-07-21 14:48:50 -0700607 if (peer->state == SCTP_UNKNOWN) {
Vlad Yasevichadd52372008-09-18 16:28:27 -0700608 peer->state = SCTP_ACTIVE;
Sridhar Samudralaad8fec12006-07-21 14:48:50 -0700609 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610 return peer;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700611 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
Eric W. Biederman89bf3452012-08-07 07:26:14 +0000613 peer = sctp_transport_new(net, addr, gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 if (!peer)
615 return NULL;
616
617 sctp_transport_set_owner(peer, asoc);
618
Frank Filz52ccb8e2005-12-22 11:36:46 -0800619 /* Initialize the peer's heartbeat interval based on the
620 * association configured value.
621 */
622 peer->hbinterval = asoc->hbinterval;
623
624 /* Set the path max_retrans. */
625 peer->pathmaxrxt = asoc->pathmaxrxt;
626
wangweidong2bccbad2013-10-26 16:06:30 +0800627 /* And the partial failure retrans threshold */
Neil Horman5aa93bc2012-07-21 07:56:07 +0000628 peer->pf_retrans = asoc->pf_retrans;
629
Frank Filz52ccb8e2005-12-22 11:36:46 -0800630 /* Initialize the peer's SACK delay timeout based on the
631 * association configured value.
632 */
633 peer->sackdelay = asoc->sackdelay;
Wei Yongjund364d922008-05-09 15:13:26 -0700634 peer->sackfreq = asoc->sackfreq;
Frank Filz52ccb8e2005-12-22 11:36:46 -0800635
636 /* Enable/disable heartbeat, SACK delay, and path MTU discovery
637 * based on association setting.
638 */
639 peer->param_flags = asoc->param_flags;
640
Vlad Yasevich8da645e2009-09-04 18:21:01 -0400641 sctp_transport_route(peer, NULL, sp);
642
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 /* Initialize the pmtu of the transport. */
Vlad Yasevich8da645e2009-09-04 18:21:01 -0400644 if (peer->param_flags & SPP_PMTUD_DISABLE) {
645 if (asoc->pathmtu)
646 peer->pathmtu = asoc->pathmtu;
647 else
648 peer->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
649 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
651 /* If this is the first transport addr on this association,
652 * initialize the association PMTU to the peer's PMTU.
653 * If not and the current association PMTU is higher than the new
654 * peer's PMTU, reset the association PMTU to the new peer's PMTU.
655 */
Frank Filz52ccb8e2005-12-22 11:36:46 -0800656 if (asoc->pathmtu)
657 asoc->pathmtu = min_t(int, peer->pathmtu, asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 else
Frank Filz52ccb8e2005-12-22 11:36:46 -0800659 asoc->pathmtu = peer->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
Daniel Borkmannbb333812013-06-28 19:49:40 +0200661 pr_debug("%s: association:%p PMTU set to %d\n", __func__, asoc,
662 asoc->pathmtu);
663
Florian Westphal6d0ccba2008-07-18 23:04:39 -0700664 peer->pmtu_pending = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
Vlad Yasevichf68b2e02009-09-04 18:21:00 -0400666 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
668 /* The asoc->peer.port might not be meaningful yet, but
669 * initialize the packet structure anyway.
670 */
671 sctp_packet_init(&peer->packet, peer, asoc->base.bind_addr.port,
672 asoc->peer.port);
673
674 /* 7.2.1 Slow-Start
675 *
676 * o The initial cwnd before DATA transmission or after a sufficiently
677 * long idle period MUST be set to
678 * min(4*MTU, max(2*MTU, 4380 bytes))
679 *
680 * o The initial value of ssthresh MAY be arbitrarily high
681 * (for example, implementations MAY use the size of the
682 * receiver advertised window).
683 */
Frank Filz52ccb8e2005-12-22 11:36:46 -0800684 peer->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685
686 /* At this point, we may not have the receiver's advertised window,
687 * so initialize ssthresh to the default value and it will be set
688 * later when we process the INIT.
689 */
690 peer->ssthresh = SCTP_DEFAULT_MAXWINDOW;
691
692 peer->partial_bytes_acked = 0;
693 peer->flight_size = 0;
Vlad Yasevich46d5a802009-11-23 15:54:00 -0500694 peer->burst_limited = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 /* Set the transport's RTO.initial value */
697 peer->rto = asoc->rto_initial;
Michele Baldessari196d6752012-12-01 04:49:42 +0000698 sctp_max_rto(asoc, peer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699
Frank Filz3f7a87d2005-06-20 13:14:57 -0700700 /* Set the peer's active state. */
701 peer->state = peer_state;
702
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 /* Attach the remote transport to our asoc. */
Thomas Graf45122ca262012-12-06 09:25:05 +0000704 list_add_tail_rcu(&peer->transports, &asoc->peer.transport_addr_list);
Frank Filz3f7a87d2005-06-20 13:14:57 -0700705 asoc->peer.transport_count++;
Xin Long4f008782015-12-30 23:50:47 +0800706 /* Add this peer into the transport hashtable */
707 sctp_hash_transport(peer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708
709 /* If we do not yet have a primary path, set one. */
710 if (!asoc->peer.primary_path) {
711 sctp_assoc_set_primary(asoc, peer);
712 asoc->peer.retran_path = peer;
713 }
714
Vlad Yasevichfbdf5012010-04-30 22:39:26 -0400715 if (asoc->peer.active_path == asoc->peer.retran_path &&
716 peer->state != SCTP_UNCONFIRMED) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 asoc->peer.retran_path = peer;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700718 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719
720 return peer;
721}
722
723/* Delete a transport address from an association. */
724void sctp_assoc_del_peer(struct sctp_association *asoc,
725 const union sctp_addr *addr)
726{
727 struct list_head *pos;
728 struct list_head *temp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 struct sctp_transport *transport;
730
731 list_for_each_safe(pos, temp, &asoc->peer.transport_addr_list) {
732 transport = list_entry(pos, struct sctp_transport, transports);
733 if (sctp_cmp_addr_exact(addr, &transport->ipaddr)) {
Frank Filz3f7a87d2005-06-20 13:14:57 -0700734 /* Do book keeping for removing the peer and free it. */
735 sctp_assoc_rm_peer(asoc, transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 break;
737 }
738 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739}
740
741/* Lookup a transport by address. */
742struct sctp_transport *sctp_assoc_lookup_paddr(
743 const struct sctp_association *asoc,
744 const union sctp_addr *address)
745{
746 struct sctp_transport *t;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
748 /* Cycle through all transports searching for a peer address. */
749
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700750 list_for_each_entry(t, &asoc->peer.transport_addr_list,
751 transports) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 if (sctp_cmp_addr_exact(address, &t->ipaddr))
753 return t;
754 }
755
756 return NULL;
757}
758
Vlad Yasevich42e30bf2007-12-20 14:08:56 -0800759/* Remove all transports except a give one */
760void sctp_assoc_del_nonprimary_peers(struct sctp_association *asoc,
761 struct sctp_transport *primary)
762{
763 struct sctp_transport *temp;
764 struct sctp_transport *t;
765
766 list_for_each_entry_safe(t, temp, &asoc->peer.transport_addr_list,
767 transports) {
768 /* if the current transport is not the primary one, delete it */
769 if (t != primary)
770 sctp_assoc_rm_peer(asoc, t);
771 }
Vlad Yasevich42e30bf2007-12-20 14:08:56 -0800772}
773
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774/* Engage in transport control operations.
775 * Mark the transport up or down and send a notification to the user.
776 * Select and update the new active and retran paths.
777 */
778void sctp_assoc_control_transport(struct sctp_association *asoc,
779 struct sctp_transport *transport,
780 sctp_transport_cmd_t command,
781 sctp_sn_error_t error)
782{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 struct sctp_ulpevent *event;
Al Viro0906e202006-11-20 17:03:01 -0800784 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 int spc_state = 0;
Neil Horman5aa93bc2012-07-21 07:56:07 +0000786 bool ulp_notify = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
788 /* Record the transition on the transport. */
789 switch (command) {
790 case SCTP_TRANSPORT_UP:
Vlad Yasevich1ae4114d2007-03-23 11:32:26 -0700791 /* If we are moving from UNCONFIRMED state due
792 * to heartbeat success, report the SCTP_ADDR_CONFIRMED
793 * state to the user, otherwise report SCTP_ADDR_AVAILABLE.
794 */
795 if (SCTP_UNCONFIRMED == transport->state &&
796 SCTP_HEARTBEAT_SUCCESS == error)
797 spc_state = SCTP_ADDR_CONFIRMED;
798 else
799 spc_state = SCTP_ADDR_AVAILABLE;
Neil Horman5aa93bc2012-07-21 07:56:07 +0000800 /* Don't inform ULP about transition from PF to
Daniel Borkmannac4f9592013-08-09 15:09:08 +0200801 * active state and set cwnd to 1 MTU, see SCTP
Neil Horman5aa93bc2012-07-21 07:56:07 +0000802 * Quick failover draft section 5.1, point 5
803 */
804 if (transport->state == SCTP_PF) {
805 ulp_notify = false;
Daniel Borkmannac4f9592013-08-09 15:09:08 +0200806 transport->cwnd = asoc->pathmtu;
Neil Horman5aa93bc2012-07-21 07:56:07 +0000807 }
Frank Filz3f7a87d2005-06-20 13:14:57 -0700808 transport->state = SCTP_ACTIVE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 break;
810
811 case SCTP_TRANSPORT_DOWN:
Vlad Yasevich40187882009-06-23 11:28:05 -0400812 /* If the transport was never confirmed, do not transition it
813 * to inactive state. Also, release the cached route since
814 * there may be a better route next time.
Vlad Yasevichcc756892007-08-24 19:30:25 +0900815 */
816 if (transport->state != SCTP_UNCONFIRMED)
817 transport->state = SCTP_INACTIVE;
Vlad Yasevich40187882009-06-23 11:28:05 -0400818 else {
819 dst_release(transport->dst);
820 transport->dst = NULL;
zhuyj061079a2014-08-20 17:31:43 +0800821 ulp_notify = false;
Vlad Yasevich40187882009-06-23 11:28:05 -0400822 }
Vlad Yasevichcc756892007-08-24 19:30:25 +0900823
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 spc_state = SCTP_ADDR_UNREACHABLE;
825 break;
826
Neil Horman5aa93bc2012-07-21 07:56:07 +0000827 case SCTP_TRANSPORT_PF:
828 transport->state = SCTP_PF;
829 ulp_notify = false;
830 break;
831
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 default:
833 return;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -0700834 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835
Daniel Borkmannb82e8f32014-06-11 18:19:29 +0200836 /* Generate and send a SCTP_PEER_ADDR_CHANGE notification
837 * to the user.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 */
Neil Horman5aa93bc2012-07-21 07:56:07 +0000839 if (ulp_notify) {
840 memset(&addr, 0, sizeof(struct sockaddr_storage));
841 memcpy(&addr, &transport->ipaddr,
842 transport->af_specific->sockaddr_len);
Daniel Borkmannb82e8f32014-06-11 18:19:29 +0200843
Neil Horman5aa93bc2012-07-21 07:56:07 +0000844 event = sctp_ulpevent_make_peer_addr_change(asoc, &addr,
845 0, spc_state, error, GFP_ATOMIC);
846 if (event)
847 sctp_ulpq_tail_event(&asoc->ulpq, event);
848 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
850 /* Select new active and retran paths. */
Daniel Borkmannb82e8f32014-06-11 18:19:29 +0200851 sctp_select_active_and_retran_path(asoc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852}
853
854/* Hold a reference to an association. */
855void sctp_association_hold(struct sctp_association *asoc)
856{
857 atomic_inc(&asoc->base.refcnt);
858}
859
860/* Release a reference to an association and cleanup
861 * if there are no more references.
862 */
863void sctp_association_put(struct sctp_association *asoc)
864{
865 if (atomic_dec_and_test(&asoc->base.refcnt))
866 sctp_association_destroy(asoc);
867}
868
869/* Allocate the next TSN, Transmission Sequence Number, for the given
870 * association.
871 */
872__u32 sctp_association_get_next_tsn(struct sctp_association *asoc)
873{
874 /* From Section 1.6 Serial Number Arithmetic:
875 * Transmission Sequence Numbers wrap around when they reach
876 * 2**32 - 1. That is, the next TSN a DATA chunk MUST use
877 * after transmitting TSN = 2*32 - 1 is TSN = 0.
878 */
879 __u32 retval = asoc->next_tsn;
880 asoc->next_tsn++;
881 asoc->unack_data++;
882
883 return retval;
884}
885
886/* Compare two addresses to see if they match. Wildcard addresses
887 * only match themselves.
888 */
889int sctp_cmp_addr_exact(const union sctp_addr *ss1,
890 const union sctp_addr *ss2)
891{
892 struct sctp_af *af;
893
894 af = sctp_get_af_specific(ss1->sa.sa_family);
895 if (unlikely(!af))
896 return 0;
897
898 return af->cmp_addr(ss1, ss2);
899}
900
901/* Return an ecne chunk to get prepended to a packet.
902 * Note: We are sly and return a shared, prealloced chunk. FIXME:
903 * No we don't, but we could/should.
904 */
905struct sctp_chunk *sctp_get_ecne_prepend(struct sctp_association *asoc)
906{
wangweidong8b7318d2013-12-06 09:36:28 +0800907 if (!asoc->need_ecne)
908 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909
910 /* Send ECNE if needed.
911 * Not being able to allocate a chunk here is not deadly.
912 */
wangweidong8b7318d2013-12-06 09:36:28 +0800913 return sctp_make_ecne(asoc, asoc->last_ecne_tsn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914}
915
916/*
917 * Find which transport this TSN was sent on.
918 */
919struct sctp_transport *sctp_assoc_lookup_tsn(struct sctp_association *asoc,
920 __u32 tsn)
921{
922 struct sctp_transport *active;
923 struct sctp_transport *match;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 struct sctp_transport *transport;
925 struct sctp_chunk *chunk;
Al Virodbc16db2006-11-20 17:01:42 -0800926 __be32 key = htonl(tsn);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927
928 match = NULL;
929
930 /*
931 * FIXME: In general, find a more efficient data structure for
932 * searching.
933 */
934
935 /*
936 * The general strategy is to search each transport's transmitted
937 * list. Return which transport this TSN lives on.
938 *
939 * Let's be hopeful and check the active_path first.
940 * Another optimization would be to know if there is only one
941 * outbound path and not have to look for the TSN at all.
942 *
943 */
944
945 active = asoc->peer.active_path;
946
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700947 list_for_each_entry(chunk, &active->transmitted,
948 transmitted_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949
950 if (key == chunk->subh.data_hdr->tsn) {
951 match = active;
952 goto out;
953 }
954 }
955
956 /* If not found, go search all the other transports. */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700957 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
958 transports) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959
960 if (transport == active)
Xufeng Zhang2317f442013-03-07 21:39:37 +0000961 continue;
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700962 list_for_each_entry(chunk, &transport->transmitted,
963 transmitted_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 if (key == chunk->subh.data_hdr->tsn) {
965 match = transport;
966 goto out;
967 }
968 }
969 }
970out:
971 return match;
972}
973
974/* Is this the association we are looking for? */
975struct sctp_transport *sctp_assoc_is_match(struct sctp_association *asoc,
Eric W. Biederman4110cc22012-08-06 08:41:13 +0000976 struct net *net,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 const union sctp_addr *laddr,
978 const union sctp_addr *paddr)
979{
980 struct sctp_transport *transport;
981
Al Viroe2fcced2006-11-20 17:08:41 -0800982 if ((htons(asoc->base.bind_addr.port) == laddr->v4.sin_port) &&
Eric W. Biederman4110cc22012-08-06 08:41:13 +0000983 (htons(asoc->peer.port) == paddr->v4.sin_port) &&
984 net_eq(sock_net(asoc->base.sk), net)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 transport = sctp_assoc_lookup_paddr(asoc, paddr);
986 if (!transport)
987 goto out;
988
989 if (sctp_bind_addr_match(&asoc->base.bind_addr, laddr,
990 sctp_sk(asoc->base.sk)))
991 goto out;
992 }
993 transport = NULL;
994
995out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 return transport;
997}
998
999/* Do delayed input processing. This is scheduled by sctp_rcv(). */
David Howellsc4028952006-11-22 14:57:56 +00001000static void sctp_assoc_bh_rcv(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001{
David Howellsc4028952006-11-22 14:57:56 +00001002 struct sctp_association *asoc =
1003 container_of(work, struct sctp_association,
1004 base.inqueue.immediate);
Eric W. Biederman55e26eb2012-08-07 07:25:24 +00001005 struct net *net = sock_net(asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 struct sctp_endpoint *ep;
1007 struct sctp_chunk *chunk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 struct sctp_inq *inqueue;
1009 int state;
1010 sctp_subtype_t subtype;
1011 int error = 0;
1012
1013 /* The association should be held so we should be safe. */
1014 ep = asoc->ep;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015
1016 inqueue = &asoc->base.inqueue;
1017 sctp_association_hold(asoc);
1018 while (NULL != (chunk = sctp_inq_pop(inqueue))) {
1019 state = asoc->state;
1020 subtype = SCTP_ST_CHUNK(chunk->chunk_hdr->type);
1021
Vlad Yasevichbbd0d592007-10-03 17:51:34 -07001022 /* SCTP-AUTH, Section 6.3:
1023 * The receiver has a list of chunk types which it expects
1024 * to be received only after an AUTH-chunk. This list has
1025 * been sent to the peer during the association setup. It
1026 * MUST silently discard these chunks if they are not placed
1027 * after an AUTH chunk in the packet.
1028 */
1029 if (sctp_auth_recv_cid(subtype.chunk, asoc) && !chunk->auth)
1030 continue;
1031
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 /* Remember where the last DATA chunk came from so we
1033 * know where to send the SACK.
1034 */
1035 if (sctp_chunk_is_data(chunk))
1036 asoc->peer.last_data_from = chunk->transport;
Michele Baldessari196d6752012-12-01 04:49:42 +00001037 else {
Eric W. Biederman55e26eb2012-08-07 07:25:24 +00001038 SCTP_INC_STATS(net, SCTP_MIB_INCTRLCHUNKS);
Michele Baldessari196d6752012-12-01 04:49:42 +00001039 asoc->stats.ictrlchunks++;
1040 if (chunk->chunk_hdr->type == SCTP_CID_SACK)
1041 asoc->stats.isacks++;
1042 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043
1044 if (chunk->transport)
Daniel Borkmanne575235f2014-06-11 18:19:30 +02001045 chunk->transport->last_time_heard = ktime_get();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046
1047 /* Run through the state machine. */
Eric W. Biederman55e26eb2012-08-07 07:25:24 +00001048 error = sctp_do_sm(net, SCTP_EVENT_T_CHUNK, subtype,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 state, ep, asoc, chunk, GFP_ATOMIC);
1050
1051 /* Check to see if the association is freed in response to
1052 * the incoming chunk. If so, get out of the while loop.
1053 */
1054 if (asoc->base.dead)
1055 break;
1056
1057 /* If there is an error on chunk, discard this packet. */
1058 if (error && chunk)
1059 chunk->pdiscard = 1;
1060 }
1061 sctp_association_put(asoc);
1062}
1063
1064/* This routine moves an association from its old sk to a new sk. */
1065void sctp_assoc_migrate(struct sctp_association *assoc, struct sock *newsk)
1066{
1067 struct sctp_sock *newsp = sctp_sk(newsk);
1068 struct sock *oldsk = assoc->base.sk;
1069
1070 /* Delete the association from the old endpoint's list of
1071 * associations.
1072 */
1073 list_del_init(&assoc->asocs);
1074
1075 /* Decrement the backlog value for a TCP-style socket. */
1076 if (sctp_style(oldsk, TCP))
1077 oldsk->sk_ack_backlog--;
1078
1079 /* Release references to the old endpoint and the sock. */
1080 sctp_endpoint_put(assoc->ep);
1081 sock_put(assoc->base.sk);
1082
1083 /* Get a reference to the new endpoint. */
1084 assoc->ep = newsp->ep;
1085 sctp_endpoint_hold(assoc->ep);
1086
1087 /* Get a reference to the new sock. */
1088 assoc->base.sk = newsk;
1089 sock_hold(assoc->base.sk);
1090
1091 /* Add the association to the new endpoint's list of associations. */
1092 sctp_endpoint_add_asoc(newsp->ep, assoc);
1093}
1094
1095/* Update an association (possibly from unexpected COOKIE-ECHO processing). */
1096void sctp_assoc_update(struct sctp_association *asoc,
1097 struct sctp_association *new)
1098{
1099 struct sctp_transport *trans;
1100 struct list_head *pos, *temp;
1101
1102 /* Copy in new parameters of peer. */
1103 asoc->c = new->c;
1104 asoc->peer.rwnd = new->peer.rwnd;
1105 asoc->peer.sack_needed = new->peer.sack_needed;
Daniel Borkmann1be9a952014-07-22 15:22:45 +02001106 asoc->peer.auth_capable = new->peer.auth_capable;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 asoc->peer.i = new->peer.i;
Vlad Yasevich8e1ee182008-10-08 14:18:39 -07001108 sctp_tsnmap_init(&asoc->peer.tsn_map, SCTP_TSN_MAP_INITIAL,
1109 asoc->peer.i.initial_tsn, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110
1111 /* Remove any peer addresses not present in the new association. */
1112 list_for_each_safe(pos, temp, &asoc->peer.transport_addr_list) {
1113 trans = list_entry(pos, struct sctp_transport, transports);
Vlad Yasevich0c427492010-04-28 08:47:19 +00001114 if (!sctp_assoc_lookup_paddr(new, &trans->ipaddr)) {
1115 sctp_assoc_rm_peer(asoc, trans);
1116 continue;
1117 }
Vlad Yasevich749bf922007-03-19 17:02:30 -07001118
1119 if (asoc->state >= SCTP_STATE_ESTABLISHED)
1120 sctp_transport_reset(trans);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 }
1122
1123 /* If the case is A (association restart), use
1124 * initial_tsn as next_tsn. If the case is B, use
1125 * current next_tsn in case data sent to peer
1126 * has been discarded and needs retransmission.
1127 */
1128 if (asoc->state >= SCTP_STATE_ESTABLISHED) {
1129 asoc->next_tsn = new->next_tsn;
1130 asoc->ctsn_ack_point = new->ctsn_ack_point;
1131 asoc->adv_peer_ack_point = new->adv_peer_ack_point;
1132
1133 /* Reinitialize SSN for both local streams
1134 * and peer's streams.
1135 */
1136 sctp_ssnmap_clear(asoc->ssnmap);
1137
Vlad Yasevich0b58a812007-03-19 17:01:17 -07001138 /* Flush the ULP reassembly and ordered queue.
1139 * Any data there will now be stale and will
1140 * cause problems.
1141 */
1142 sctp_ulpq_flush(&asoc->ulpq);
1143
Vlad Yasevich749bf922007-03-19 17:02:30 -07001144 /* reset the overall association error count so
1145 * that the restarted association doesn't get torn
1146 * down on the next retransmission timer.
1147 */
1148 asoc->overall_error_count = 0;
1149
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150 } else {
1151 /* Add any peer addresses from the new association. */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07001152 list_for_each_entry(trans, &new->peer.transport_addr_list,
1153 transports) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 if (!sctp_assoc_lookup_paddr(asoc, &trans->ipaddr))
1155 sctp_assoc_add_peer(asoc, &trans->ipaddr,
Sridhar Samudralaad8fec12006-07-21 14:48:50 -07001156 GFP_ATOMIC, trans->state);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 }
1158
1159 asoc->ctsn_ack_point = asoc->next_tsn - 1;
1160 asoc->adv_peer_ack_point = asoc->ctsn_ack_point;
1161 if (!asoc->ssnmap) {
1162 /* Move the ssnmap. */
1163 asoc->ssnmap = new->ssnmap;
1164 new->ssnmap = NULL;
1165 }
Vlad Yasevich07d93962007-05-04 13:55:27 -07001166
1167 if (!asoc->assoc_id) {
1168 /* get a new association id since we don't have one
1169 * yet.
1170 */
1171 sctp_assoc_set_id(asoc, GFP_ATOMIC);
1172 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 }
Vlad Yasevicha29a5bd2007-09-16 19:31:35 -07001174
wangweidong9d2c8812013-12-06 09:36:30 +08001175 /* SCTP-AUTH: Save the peer parameters from the new associations
Vlad Yasevich730fc3d2007-09-16 19:32:11 -07001176 * and also move the association shared keys over
1177 */
1178 kfree(asoc->peer.peer_random);
1179 asoc->peer.peer_random = new->peer.peer_random;
1180 new->peer.peer_random = NULL;
1181
1182 kfree(asoc->peer.peer_chunks);
1183 asoc->peer.peer_chunks = new->peer.peer_chunks;
1184 new->peer.peer_chunks = NULL;
1185
1186 kfree(asoc->peer.peer_hmacs);
1187 asoc->peer.peer_hmacs = new->peer.peer_hmacs;
1188 new->peer.peer_hmacs = NULL;
1189
Vlad Yasevich730fc3d2007-09-16 19:32:11 -07001190 sctp_auth_asoc_init_active_key(asoc, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191}
1192
1193/* Update the retran path for sending a retransmitted packet.
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001194 * See also RFC4960, 6.4. Multi-Homed SCTP Endpoints:
1195 *
1196 * When there is outbound data to send and the primary path
1197 * becomes inactive (e.g., due to failures), or where the
1198 * SCTP user explicitly requests to send data to an
1199 * inactive destination transport address, before reporting
1200 * an error to its ULP, the SCTP endpoint should try to send
1201 * the data to an alternate active destination transport
1202 * address if one exists.
1203 *
1204 * When retransmitting data that timed out, if the endpoint
1205 * is multihomed, it should consider each source-destination
1206 * address pair in its retransmission selection policy.
1207 * When retransmitting timed-out data, the endpoint should
1208 * attempt to pick the most divergent source-destination
1209 * pair from the original source-destination pair to which
1210 * the packet was transmitted.
1211 *
1212 * Note: Rules for picking the most divergent source-destination
1213 * pair are an implementation decision and are not specified
1214 * within this document.
1215 *
1216 * Our basic strategy is to round-robin transports in priorities
Denys Vlasenko2103d6b2015-09-28 14:34:04 +02001217 * according to sctp_trans_score() e.g., if no such
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001218 * transport with state SCTP_ACTIVE exists, round-robin through
1219 * SCTP_UNKNOWN, etc. You get the picture.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 */
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001221static u8 sctp_trans_score(const struct sctp_transport *trans)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222{
Denys Vlasenko2103d6b2015-09-28 14:34:04 +02001223 switch (trans->state) {
1224 case SCTP_ACTIVE:
1225 return 3; /* best case */
1226 case SCTP_UNKNOWN:
1227 return 2;
1228 case SCTP_PF:
1229 return 1;
1230 default: /* case SCTP_INACTIVE */
1231 return 0; /* worst case */
1232 }
Frank Filz3f7a87d2005-06-20 13:14:57 -07001233}
1234
Daniel Borkmanna7288c42014-06-11 18:19:31 +02001235static struct sctp_transport *sctp_trans_elect_tie(struct sctp_transport *trans1,
1236 struct sctp_transport *trans2)
1237{
1238 if (trans1->error_count > trans2->error_count) {
1239 return trans2;
1240 } else if (trans1->error_count == trans2->error_count &&
1241 ktime_after(trans2->last_time_heard,
1242 trans1->last_time_heard)) {
1243 return trans2;
1244 } else {
1245 return trans1;
1246 }
1247}
1248
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001249static struct sctp_transport *sctp_trans_elect_best(struct sctp_transport *curr,
1250 struct sctp_transport *best)
1251{
Daniel Borkmanna7288c42014-06-11 18:19:31 +02001252 u8 score_curr, score_best;
1253
Daniel Borkmannea4f19c2014-08-22 13:03:29 +02001254 if (best == NULL || curr == best)
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001255 return curr;
1256
Daniel Borkmanna7288c42014-06-11 18:19:31 +02001257 score_curr = sctp_trans_score(curr);
1258 score_best = sctp_trans_score(best);
1259
1260 /* First, try a score-based selection if both transport states
1261 * differ. If we're in a tie, lets try to make a more clever
1262 * decision here based on error counts and last time heard.
1263 */
1264 if (score_curr > score_best)
1265 return curr;
1266 else if (score_curr == score_best)
Xin Long39d2ade2016-03-10 15:31:57 +08001267 return sctp_trans_elect_tie(best, curr);
Daniel Borkmanna7288c42014-06-11 18:19:31 +02001268 else
1269 return best;
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001270}
1271
1272void sctp_assoc_update_retran_path(struct sctp_association *asoc)
1273{
1274 struct sctp_transport *trans = asoc->peer.retran_path;
1275 struct sctp_transport *trans_next = NULL;
1276
1277 /* We're done as we only have the one and only path. */
1278 if (asoc->peer.transport_count == 1)
1279 return;
1280 /* If active_path and retran_path are the same and active,
1281 * then this is the only active path. Use it.
1282 */
1283 if (asoc->peer.active_path == asoc->peer.retran_path &&
1284 asoc->peer.active_path->state == SCTP_ACTIVE)
1285 return;
1286
1287 /* Iterate from retran_path's successor back to retran_path. */
1288 for (trans = list_next_entry(trans, transports); 1;
1289 trans = list_next_entry(trans, transports)) {
1290 /* Manually skip the head element. */
1291 if (&trans->transports == &asoc->peer.transport_addr_list)
1292 continue;
1293 if (trans->state == SCTP_UNCONFIRMED)
1294 continue;
1295 trans_next = sctp_trans_elect_best(trans, trans_next);
1296 /* Active is good enough for immediate return. */
1297 if (trans_next->state == SCTP_ACTIVE)
1298 break;
1299 /* We've reached the end, time to update path. */
1300 if (trans == asoc->peer.retran_path)
1301 break;
1302 }
1303
Daniel Borkmann433131b2014-03-13 14:45:26 +01001304 asoc->peer.retran_path = trans_next;
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001305
1306 pr_debug("%s: association:%p updated new path to addr:%pISpc\n",
1307 __func__, asoc, &asoc->peer.retran_path->ipaddr.sa);
1308}
1309
Daniel Borkmannb82e8f32014-06-11 18:19:29 +02001310static void sctp_select_active_and_retran_path(struct sctp_association *asoc)
1311{
1312 struct sctp_transport *trans, *trans_pri = NULL, *trans_sec = NULL;
Daniel Borkmanna7288c42014-06-11 18:19:31 +02001313 struct sctp_transport *trans_pf = NULL;
Daniel Borkmannb82e8f32014-06-11 18:19:29 +02001314
1315 /* Look for the two most recently used active transports. */
1316 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
1317 transports) {
Daniel Borkmanna7288c42014-06-11 18:19:31 +02001318 /* Skip uninteresting transports. */
Daniel Borkmannb82e8f32014-06-11 18:19:29 +02001319 if (trans->state == SCTP_INACTIVE ||
Daniel Borkmanna7288c42014-06-11 18:19:31 +02001320 trans->state == SCTP_UNCONFIRMED)
Daniel Borkmannb82e8f32014-06-11 18:19:29 +02001321 continue;
Daniel Borkmanna7288c42014-06-11 18:19:31 +02001322 /* Keep track of the best PF transport from our
1323 * list in case we don't find an active one.
1324 */
1325 if (trans->state == SCTP_PF) {
1326 trans_pf = sctp_trans_elect_best(trans, trans_pf);
1327 continue;
1328 }
1329 /* For active transports, pick the most recent ones. */
Daniel Borkmannb82e8f32014-06-11 18:19:29 +02001330 if (trans_pri == NULL ||
Daniel Borkmanne575235f2014-06-11 18:19:30 +02001331 ktime_after(trans->last_time_heard,
1332 trans_pri->last_time_heard)) {
Daniel Borkmannb82e8f32014-06-11 18:19:29 +02001333 trans_sec = trans_pri;
1334 trans_pri = trans;
1335 } else if (trans_sec == NULL ||
Daniel Borkmanne575235f2014-06-11 18:19:30 +02001336 ktime_after(trans->last_time_heard,
1337 trans_sec->last_time_heard)) {
Daniel Borkmannb82e8f32014-06-11 18:19:29 +02001338 trans_sec = trans;
1339 }
1340 }
1341
1342 /* RFC 2960 6.4 Multi-Homed SCTP Endpoints
1343 *
1344 * By default, an endpoint should always transmit to the primary
1345 * path, unless the SCTP user explicitly specifies the
1346 * destination transport address (and possibly source transport
1347 * address) to use. [If the primary is active but not most recent,
1348 * bump the most recently used transport.]
1349 */
1350 if ((asoc->peer.primary_path->state == SCTP_ACTIVE ||
1351 asoc->peer.primary_path->state == SCTP_UNKNOWN) &&
1352 asoc->peer.primary_path != trans_pri) {
1353 trans_sec = trans_pri;
1354 trans_pri = asoc->peer.primary_path;
1355 }
1356
1357 /* We did not find anything useful for a possible retransmission
1358 * path; either primary path that we found is the the same as
1359 * the current one, or we didn't generally find an active one.
1360 */
1361 if (trans_sec == NULL)
1362 trans_sec = trans_pri;
1363
1364 /* If we failed to find a usable transport, just camp on the
Daniel Borkmannaa4a83e2014-08-22 13:03:30 +02001365 * active or pick a PF iff it's the better choice.
Daniel Borkmannb82e8f32014-06-11 18:19:29 +02001366 */
1367 if (trans_pri == NULL) {
Daniel Borkmannaa4a83e2014-08-22 13:03:30 +02001368 trans_pri = sctp_trans_elect_best(asoc->peer.active_path, trans_pf);
1369 trans_sec = trans_pri;
Daniel Borkmannb82e8f32014-06-11 18:19:29 +02001370 }
1371
1372 /* Set the active and retran transports. */
1373 asoc->peer.active_path = trans_pri;
1374 asoc->peer.retran_path = trans_sec;
1375}
1376
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001377struct sctp_transport *
1378sctp_assoc_choose_alter_transport(struct sctp_association *asoc,
1379 struct sctp_transport *last_sent_to)
Frank Filz3f7a87d2005-06-20 13:14:57 -07001380{
Wei Yongjun9919b452009-05-12 21:52:51 +08001381 /* If this is the first time packet is sent, use the active path,
1382 * else use the retran path. If the last packet was sent over the
Frank Filz3f7a87d2005-06-20 13:14:57 -07001383 * retran path, update the retran path and use it.
1384 */
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001385 if (last_sent_to == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 return asoc->peer.active_path;
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001387 } else {
Wei Yongjun9919b452009-05-12 21:52:51 +08001388 if (last_sent_to == asoc->peer.retran_path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 sctp_assoc_update_retran_path(asoc);
Daniel Borkmann4c47af42014-02-20 20:51:06 +01001390
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 return asoc->peer.retran_path;
1392 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393}
1394
1395/* Update the association's pmtu and frag_point by going through all the
1396 * transports. This routine is called when a transport's PMTU has changed.
1397 */
David S. Miller02f3d4c2012-07-16 03:57:14 -07001398void sctp_assoc_sync_pmtu(struct sock *sk, struct sctp_association *asoc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399{
1400 struct sctp_transport *t;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 __u32 pmtu = 0;
1402
1403 if (!asoc)
1404 return;
1405
1406 /* Get the lowest pmtu of all the transports. */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07001407 list_for_each_entry(t, &asoc->peer.transport_addr_list,
1408 transports) {
Vlad Yasevich8a479492007-06-07 14:21:05 -04001409 if (t->pmtu_pending && t->dst) {
Marcelo Ricardo Leitner3822a5f2016-03-19 12:17:20 -03001410 sctp_transport_update_pmtu(sk, t,
1411 WORD_TRUNC(dst_mtu(t->dst)));
Vlad Yasevich8a479492007-06-07 14:21:05 -04001412 t->pmtu_pending = 0;
1413 }
Frank Filz52ccb8e2005-12-22 11:36:46 -08001414 if (!pmtu || (t->pathmtu < pmtu))
1415 pmtu = t->pathmtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416 }
1417
1418 if (pmtu) {
Frank Filz52ccb8e2005-12-22 11:36:46 -08001419 asoc->pathmtu = pmtu;
Vlad Yasevichf68b2e02009-09-04 18:21:00 -04001420 asoc->frag_point = sctp_frag_point(asoc, pmtu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 }
1422
Daniel Borkmannbb333812013-06-28 19:49:40 +02001423 pr_debug("%s: asoc:%p, pmtu:%d, frag_point:%d\n", __func__, asoc,
1424 asoc->pathmtu, asoc->frag_point);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425}
1426
1427/* Should we send a SACK to update our peer? */
wangweidongce4a03d2013-12-06 09:36:29 +08001428static inline bool sctp_peer_needs_update(struct sctp_association *asoc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429{
Eric W. Biedermane1fc3b12012-08-07 07:29:57 +00001430 struct net *net = sock_net(asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 switch (asoc->state) {
1432 case SCTP_STATE_ESTABLISHED:
1433 case SCTP_STATE_SHUTDOWN_PENDING:
1434 case SCTP_STATE_SHUTDOWN_RECEIVED:
1435 case SCTP_STATE_SHUTDOWN_SENT:
1436 if ((asoc->rwnd > asoc->a_rwnd) &&
Vlad Yasevich90f2f532009-11-23 15:53:57 -05001437 ((asoc->rwnd - asoc->a_rwnd) >= max_t(__u32,
Eric W. Biedermane1fc3b12012-08-07 07:29:57 +00001438 (asoc->base.sk->sk_rcvbuf >> net->sctp.rwnd_upd_shift),
Vlad Yasevich90f2f532009-11-23 15:53:57 -05001439 asoc->pathmtu)))
wangweidongce4a03d2013-12-06 09:36:29 +08001440 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 break;
1442 default:
1443 break;
1444 }
wangweidongce4a03d2013-12-06 09:36:29 +08001445 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446}
1447
Daniel Borkmann362d5202014-04-14 21:45:17 +02001448/* Increase asoc's rwnd by len and send any window update SACK if needed. */
1449void sctp_assoc_rwnd_increase(struct sctp_association *asoc, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450{
1451 struct sctp_chunk *sack;
1452 struct timer_list *timer;
1453
Daniel Borkmann362d5202014-04-14 21:45:17 +02001454 if (asoc->rwnd_over) {
1455 if (asoc->rwnd_over >= len) {
1456 asoc->rwnd_over -= len;
1457 } else {
1458 asoc->rwnd += (len - asoc->rwnd_over);
1459 asoc->rwnd_over = 0;
1460 }
1461 } else {
1462 asoc->rwnd += len;
1463 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464
Daniel Borkmann362d5202014-04-14 21:45:17 +02001465 /* If we had window pressure, start recovering it
1466 * once our rwnd had reached the accumulated pressure
1467 * threshold. The idea is to recover slowly, but up
1468 * to the initial advertised window.
1469 */
1470 if (asoc->rwnd_press && asoc->rwnd >= asoc->rwnd_press) {
1471 int change = min(asoc->pathmtu, asoc->rwnd_press);
1472 asoc->rwnd += change;
1473 asoc->rwnd_press -= change;
1474 }
Vlad Yasevich4d3c46e2009-09-04 18:20:59 -04001475
Daniel Borkmann362d5202014-04-14 21:45:17 +02001476 pr_debug("%s: asoc:%p rwnd increased by %d to (%u, %u) - %u\n",
1477 __func__, asoc, len, asoc->rwnd, asoc->rwnd_over,
1478 asoc->a_rwnd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479
1480 /* Send a window update SACK if the rwnd has increased by at least the
1481 * minimum of the association's PMTU and half of the receive buffer.
1482 * The algorithm used is similar to the one described in
1483 * Section 4.2.3.3 of RFC 1122.
1484 */
Daniel Borkmann362d5202014-04-14 21:45:17 +02001485 if (sctp_peer_needs_update(asoc)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 asoc->a_rwnd = asoc->rwnd;
Daniel Borkmannbb333812013-06-28 19:49:40 +02001487
1488 pr_debug("%s: sending window update SACK- asoc:%p rwnd:%u "
1489 "a_rwnd:%u\n", __func__, asoc, asoc->rwnd,
1490 asoc->a_rwnd);
1491
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 sack = sctp_make_sack(asoc);
1493 if (!sack)
1494 return;
1495
1496 asoc->peer.sack_needed = 0;
1497
Marcelo Ricardo Leitnercea87682016-03-10 18:33:07 -03001498 sctp_outq_tail(&asoc->outqueue, sack, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499
1500 /* Stop the SACK timer. */
1501 timer = &asoc->timers[SCTP_EVENT_TIMEOUT_SACK];
Ying Xue25cc4ae2013-02-03 20:32:57 +00001502 if (del_timer(timer))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503 sctp_association_put(asoc);
1504 }
1505}
1506
Daniel Borkmann362d5202014-04-14 21:45:17 +02001507/* Decrease asoc's rwnd by len. */
1508void sctp_assoc_rwnd_decrease(struct sctp_association *asoc, unsigned int len)
1509{
1510 int rx_count;
1511 int over = 0;
1512
1513 if (unlikely(!asoc->rwnd || asoc->rwnd_over))
1514 pr_debug("%s: association:%p has asoc->rwnd:%u, "
1515 "asoc->rwnd_over:%u!\n", __func__, asoc,
1516 asoc->rwnd, asoc->rwnd_over);
1517
1518 if (asoc->ep->rcvbuf_policy)
1519 rx_count = atomic_read(&asoc->rmem_alloc);
1520 else
1521 rx_count = atomic_read(&asoc->base.sk->sk_rmem_alloc);
1522
1523 /* If we've reached or overflowed our receive buffer, announce
1524 * a 0 rwnd if rwnd would still be positive. Store the
1525 * the potential pressure overflow so that the window can be restored
1526 * back to original value.
1527 */
1528 if (rx_count >= asoc->base.sk->sk_rcvbuf)
1529 over = 1;
1530
1531 if (asoc->rwnd >= len) {
1532 asoc->rwnd -= len;
1533 if (over) {
1534 asoc->rwnd_press += asoc->rwnd;
1535 asoc->rwnd = 0;
1536 }
1537 } else {
1538 asoc->rwnd_over = len - asoc->rwnd;
1539 asoc->rwnd = 0;
1540 }
1541
1542 pr_debug("%s: asoc:%p rwnd decreased by %d to (%u, %u, %u)\n",
1543 __func__, asoc, len, asoc->rwnd, asoc->rwnd_over,
1544 asoc->rwnd_press);
1545}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546
1547/* Build the bind address list for the association based on info from the
1548 * local endpoint and the remote peer.
1549 */
Alexey Dobriyan3182cd82005-07-11 20:57:47 -07001550int sctp_assoc_set_bind_addr_from_ep(struct sctp_association *asoc,
Vlad Yasevich409b95a2009-11-10 08:57:34 +00001551 sctp_scope_t scope, gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553 int flags;
1554
1555 /* Use scoping rules to determine the subset of addresses from
1556 * the endpoint.
1557 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 flags = (PF_INET6 == asoc->base.sk->sk_family) ? SCTP_ADDR6_ALLOWED : 0;
1559 if (asoc->peer.ipv4_address)
1560 flags |= SCTP_ADDR4_PEERSUPP;
1561 if (asoc->peer.ipv6_address)
1562 flags |= SCTP_ADDR6_PEERSUPP;
1563
Eric W. Biederman4db67e82012-08-06 08:42:04 +00001564 return sctp_bind_addr_copy(sock_net(asoc->base.sk),
1565 &asoc->base.bind_addr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 &asoc->ep->base.bind_addr,
1567 scope, gfp, flags);
1568}
1569
1570/* Build the association's bind address list from the cookie. */
1571int sctp_assoc_set_bind_addr_from_cookie(struct sctp_association *asoc,
Alexey Dobriyan3182cd82005-07-11 20:57:47 -07001572 struct sctp_cookie *cookie,
Al Virodd0fc662005-10-07 07:46:04 +01001573 gfp_t gfp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574{
1575 int var_size2 = ntohs(cookie->peer_init->chunk_hdr.length);
1576 int var_size3 = cookie->raw_addr_list_len;
1577 __u8 *raw = (__u8 *)cookie->peer_init + var_size2;
1578
1579 return sctp_raw_to_bind_addrs(&asoc->base.bind_addr, raw, var_size3,
1580 asoc->ep->base.bind_addr.port, gfp);
1581}
1582
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001583/* Lookup laddr in the bind address list of an association. */
1584int sctp_assoc_lookup_laddr(struct sctp_association *asoc,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 const union sctp_addr *laddr)
1586{
Vlad Yasevich559cf712007-09-16 16:03:28 -07001587 int found = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 if ((asoc->base.bind_addr.port == ntohs(laddr->v4.sin_port)) &&
1590 sctp_bind_addr_match(&asoc->base.bind_addr, laddr,
Vlad Yasevich559cf712007-09-16 16:03:28 -07001591 sctp_sk(asoc->base.sk)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 found = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 return found;
1595}
Vlad Yasevich07d93962007-05-04 13:55:27 -07001596
1597/* Set an association id for a given association */
1598int sctp_assoc_set_id(struct sctp_association *asoc, gfp_t gfp)
1599{
Mel Gormand0164ad2015-11-06 16:28:21 -08001600 bool preload = gfpflags_allow_blocking(gfp);
Tejun Heo94960e82013-02-27 17:05:00 -08001601 int ret;
Vlad Yasevichc6ba68a2009-06-01 12:41:15 -04001602
1603 /* If the id is already assigned, keep it. */
1604 if (asoc->assoc_id)
Tejun Heo94960e82013-02-27 17:05:00 -08001605 return 0;
Vlad Yasevich07d93962007-05-04 13:55:27 -07001606
Tejun Heo94960e82013-02-27 17:05:00 -08001607 if (preload)
1608 idr_preload(gfp);
Vlad Yasevich07d93962007-05-04 13:55:27 -07001609 spin_lock_bh(&sctp_assocs_id_lock);
Jeff Layton713e00a2013-04-29 16:21:22 -07001610 /* 0 is not a valid assoc_id, must be >= 1 */
1611 ret = idr_alloc_cyclic(&sctp_assocs_id, asoc, 1, 0, GFP_NOWAIT);
Vlad Yasevich07d93962007-05-04 13:55:27 -07001612 spin_unlock_bh(&sctp_assocs_id_lock);
Tejun Heo94960e82013-02-27 17:05:00 -08001613 if (preload)
1614 idr_preload_end();
1615 if (ret < 0)
1616 return ret;
Vlad Yasevich07d93962007-05-04 13:55:27 -07001617
Tejun Heo94960e82013-02-27 17:05:00 -08001618 asoc->assoc_id = (sctp_assoc_t)ret;
1619 return 0;
Vlad Yasevich07d93962007-05-04 13:55:27 -07001620}
Vlad Yasevicha08de642007-12-20 14:11:47 -08001621
Wei Yongjun8b4472c2011-05-24 21:48:02 +00001622/* Free the ASCONF queue */
1623static void sctp_assoc_free_asconf_queue(struct sctp_association *asoc)
1624{
1625 struct sctp_chunk *asconf;
1626 struct sctp_chunk *tmp;
1627
1628 list_for_each_entry_safe(asconf, tmp, &asoc->addip_chunk_list, list) {
1629 list_del_init(&asconf->list);
1630 sctp_chunk_free(asconf);
1631 }
1632}
1633
Vlad Yasevicha08de642007-12-20 14:11:47 -08001634/* Free asconf_ack cache */
1635static void sctp_assoc_free_asconf_acks(struct sctp_association *asoc)
1636{
1637 struct sctp_chunk *ack;
1638 struct sctp_chunk *tmp;
1639
1640 list_for_each_entry_safe(ack, tmp, &asoc->asconf_ack_list,
1641 transmitted_list) {
1642 list_del_init(&ack->transmitted_list);
1643 sctp_chunk_free(ack);
1644 }
1645}
1646
1647/* Clean up the ASCONF_ACK queue */
1648void sctp_assoc_clean_asconf_ack_cache(const struct sctp_association *asoc)
1649{
1650 struct sctp_chunk *ack;
1651 struct sctp_chunk *tmp;
1652
Lucas De Marchi25985ed2011-03-30 22:57:33 -03001653 /* We can remove all the entries from the queue up to
Vlad Yasevicha08de642007-12-20 14:11:47 -08001654 * the "Peer-Sequence-Number".
1655 */
1656 list_for_each_entry_safe(ack, tmp, &asoc->asconf_ack_list,
1657 transmitted_list) {
1658 if (ack->subh.addip_hdr->serial ==
1659 htonl(asoc->peer.addip_serial))
1660 break;
1661
1662 list_del_init(&ack->transmitted_list);
1663 sctp_chunk_free(ack);
1664 }
1665}
1666
1667/* Find the ASCONF_ACK whose serial number matches ASCONF */
1668struct sctp_chunk *sctp_assoc_lookup_asconf_ack(
1669 const struct sctp_association *asoc,
1670 __be32 serial)
1671{
Wei Yongjuna8699812008-02-05 23:35:04 +09001672 struct sctp_chunk *ack;
Vlad Yasevicha08de642007-12-20 14:11:47 -08001673
1674 /* Walk through the list of cached ASCONF-ACKs and find the
1675 * ack chunk whose serial number matches that of the request.
1676 */
1677 list_for_each_entry(ack, &asoc->asconf_ack_list, transmitted_list) {
Daniel Borkmannb69040d2014-10-09 22:55:32 +02001678 if (sctp_chunk_pending(ack))
1679 continue;
Vlad Yasevicha08de642007-12-20 14:11:47 -08001680 if (ack->subh.addip_hdr->serial == serial) {
1681 sctp_chunk_hold(ack);
Wei Yongjuna8699812008-02-05 23:35:04 +09001682 return ack;
Vlad Yasevicha08de642007-12-20 14:11:47 -08001683 }
1684 }
1685
Wei Yongjuna8699812008-02-05 23:35:04 +09001686 return NULL;
Vlad Yasevicha08de642007-12-20 14:11:47 -08001687}
Wei Yongjuna000c012011-05-29 23:23:36 +00001688
1689void sctp_asconf_queue_teardown(struct sctp_association *asoc)
1690{
1691 /* Free any cached ASCONF_ACK chunk. */
1692 sctp_assoc_free_asconf_acks(asoc);
1693
1694 /* Free the ASCONF queue. */
1695 sctp_assoc_free_asconf_queue(asoc);
1696
1697 /* Free any cached ASCONF chunk. */
1698 if (asoc->addip_last_asconf)
1699 sctp_chunk_free(asoc->addip_last_asconf);
1700}