blob: 9c77947c0597c1c2f646214584a8480e038105e9 [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) 2001-2003 Intel Corp.
6 *
Vlad Yasevich60c778b2008-01-11 09:57:09 -05007 * This file is part of the SCTP kernel implementation
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 *
9 * These functions implement the sctp_outq class. The outqueue handles
10 * bundling and queueing of outgoing SCTP chunks.
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 * Perry Melange <pmelange@null.cc.uic.edu>
36 * Xingang Guo <xingang.guo@intel.com>
37 * Hui Huang <hui.huang@nokia.com>
38 * Sridhar Samudrala <sri@us.ibm.com>
39 * Jon Grimm <jgrimm@us.ibm.com>
Linus Torvalds1da177e2005-04-16 15:20:36 -070040 */
41
Joe Perches145ce502010-08-24 13:21:08 +000042#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
43
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/types.h>
45#include <linux/list.h> /* For struct list_head */
46#include <linux/socket.h>
47#include <linux/ip.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090048#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#include <net/sock.h> /* For skb_set_owner_w */
50
51#include <net/sctp/sctp.h>
52#include <net/sctp/sm.h>
53
54/* Declare internal functions here. */
55static int sctp_acked(struct sctp_sackhdr *sack, __u32 tsn);
56static void sctp_check_transmitted(struct sctp_outq *q,
57 struct list_head *transmitted_queue,
58 struct sctp_transport *transport,
Nicolas Dichteledfee032012-10-03 05:43:22 +000059 union sctp_addr *saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -070060 struct sctp_sackhdr *sack,
Vlad Yasevichbfa0d982010-04-30 22:41:10 -040061 __u32 *highest_new_tsn);
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
63static void sctp_mark_missing(struct sctp_outq *q,
64 struct list_head *transmitted_queue,
65 struct sctp_transport *transport,
66 __u32 highest_new_tsn,
67 int count_of_newacks);
68
69static void sctp_generate_fwdtsn(struct sctp_outq *q, __u32 sack_ctsn);
70
Adrian Bunkabd0b1982008-07-22 14:20:45 -070071static int sctp_outq_flush(struct sctp_outq *q, int rtx_timeout);
72
Linus Torvalds1da177e2005-04-16 15:20:36 -070073/* Add data to the front of the queue. */
74static inline void sctp_outq_head_data(struct sctp_outq *q,
75 struct sctp_chunk *ch)
76{
David S. Miller79af02c2005-07-08 21:47:49 -070077 list_add(&ch->list, &q->out_chunk_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 q->out_qlen += ch->skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -070079}
80
81/* Take data from the front of the queue. */
82static inline struct sctp_chunk *sctp_outq_dequeue_data(struct sctp_outq *q)
83{
David S. Miller79af02c2005-07-08 21:47:49 -070084 struct sctp_chunk *ch = NULL;
85
86 if (!list_empty(&q->out_chunk_list)) {
87 struct list_head *entry = q->out_chunk_list.next;
88
89 ch = list_entry(entry, struct sctp_chunk, list);
90 list_del_init(entry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 q->out_qlen -= ch->skb->len;
David S. Miller79af02c2005-07-08 21:47:49 -070092 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 return ch;
94}
95/* Add data chunk to the end of the queue. */
96static inline void sctp_outq_tail_data(struct sctp_outq *q,
97 struct sctp_chunk *ch)
98{
David S. Miller79af02c2005-07-08 21:47:49 -070099 list_add_tail(&ch->list, &q->out_chunk_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 q->out_qlen += ch->skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101}
102
103/*
104 * SFR-CACC algorithm:
105 * D) If count_of_newacks is greater than or equal to 2
106 * and t was not sent to the current primary then the
107 * sender MUST NOT increment missing report count for t.
108 */
109static inline int sctp_cacc_skip_3_1_d(struct sctp_transport *primary,
110 struct sctp_transport *transport,
111 int count_of_newacks)
112{
wangweidongcb3f8372013-12-23 12:16:50 +0800113 if (count_of_newacks >= 2 && transport != primary)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114 return 1;
115 return 0;
116}
117
118/*
119 * SFR-CACC algorithm:
120 * F) If count_of_newacks is less than 2, let d be the
121 * destination to which t was sent. If cacc_saw_newack
122 * is 0 for destination d, then the sender MUST NOT
123 * increment missing report count for t.
124 */
125static inline int sctp_cacc_skip_3_1_f(struct sctp_transport *transport,
126 int count_of_newacks)
127{
Vlad Yasevichf246a7b2011-04-18 19:13:56 +0000128 if (count_of_newacks < 2 &&
129 (transport && !transport->cacc.cacc_saw_newack))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 return 1;
131 return 0;
132}
133
134/*
135 * SFR-CACC algorithm:
136 * 3.1) If CYCLING_CHANGEOVER is 0, the sender SHOULD
137 * execute steps C, D, F.
138 *
139 * C has been implemented in sctp_outq_sack
140 */
141static inline int sctp_cacc_skip_3_1(struct sctp_transport *primary,
142 struct sctp_transport *transport,
143 int count_of_newacks)
144{
145 if (!primary->cacc.cycling_changeover) {
146 if (sctp_cacc_skip_3_1_d(primary, transport, count_of_newacks))
147 return 1;
148 if (sctp_cacc_skip_3_1_f(transport, count_of_newacks))
149 return 1;
150 return 0;
151 }
152 return 0;
153}
154
155/*
156 * SFR-CACC algorithm:
157 * 3.2) Else if CYCLING_CHANGEOVER is 1, and t is less
158 * than next_tsn_at_change of the current primary, then
159 * the sender MUST NOT increment missing report count
160 * for t.
161 */
162static inline int sctp_cacc_skip_3_2(struct sctp_transport *primary, __u32 tsn)
163{
164 if (primary->cacc.cycling_changeover &&
165 TSN_lt(tsn, primary->cacc.next_tsn_at_change))
166 return 1;
167 return 0;
168}
169
170/*
171 * SFR-CACC algorithm:
172 * 3) If the missing report count for TSN t is to be
173 * incremented according to [RFC2960] and
174 * [SCTP_STEWART-2002], and CHANGEOVER_ACTIVE is set,
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300175 * then the sender MUST further execute steps 3.1 and
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 * 3.2 to determine if the missing report count for
177 * TSN t SHOULD NOT be incremented.
178 *
179 * 3.3) If 3.1 and 3.2 do not dictate that the missing
180 * report count for t should not be incremented, then
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300181 * the sender SHOULD increment missing report count for
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 * t (according to [RFC2960] and [SCTP_STEWART_2002]).
183 */
184static inline int sctp_cacc_skip(struct sctp_transport *primary,
185 struct sctp_transport *transport,
186 int count_of_newacks,
187 __u32 tsn)
188{
189 if (primary->cacc.changeover_active &&
Joe Perchesf64f9e72009-11-29 16:55:45 -0800190 (sctp_cacc_skip_3_1(primary, transport, count_of_newacks) ||
191 sctp_cacc_skip_3_2(primary, tsn)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 return 1;
193 return 0;
194}
195
196/* Initialize an existing sctp_outq. This does the boring stuff.
197 * You still need to define handlers if you really want to DO
198 * something with this structure...
199 */
200void sctp_outq_init(struct sctp_association *asoc, struct sctp_outq *q)
201{
Neil Hormanc5c77742013-06-12 14:26:44 -0400202 memset(q, 0, sizeof(struct sctp_outq));
203
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 q->asoc = asoc;
David S. Miller79af02c2005-07-08 21:47:49 -0700205 INIT_LIST_HEAD(&q->out_chunk_list);
206 INIT_LIST_HEAD(&q->control_chunk_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 INIT_LIST_HEAD(&q->retransmit);
208 INIT_LIST_HEAD(&q->sacked);
209 INIT_LIST_HEAD(&q->abandoned);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210}
211
212/* Free the outqueue structure and any related pending chunks.
213 */
Neil Horman2f94aab2013-01-17 11:15:08 +0000214static void __sctp_outq_teardown(struct sctp_outq *q)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
216 struct sctp_transport *transport;
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700217 struct list_head *lchunk, *temp;
David S. Miller79af02c2005-07-08 21:47:49 -0700218 struct sctp_chunk *chunk, *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219
220 /* Throw away unacknowledged chunks. */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -0700221 list_for_each_entry(transport, &q->asoc->peer.transport_addr_list,
222 transports) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 while ((lchunk = sctp_list_dequeue(&transport->transmitted)) != NULL) {
224 chunk = list_entry(lchunk, struct sctp_chunk,
225 transmitted_list);
226 /* Mark as part of a failed message. */
227 sctp_chunk_fail(chunk, q->error);
228 sctp_chunk_free(chunk);
229 }
230 }
231
232 /* Throw away chunks that have been gap ACKed. */
233 list_for_each_safe(lchunk, temp, &q->sacked) {
234 list_del_init(lchunk);
235 chunk = list_entry(lchunk, struct sctp_chunk,
236 transmitted_list);
237 sctp_chunk_fail(chunk, q->error);
238 sctp_chunk_free(chunk);
239 }
240
241 /* Throw away any chunks in the retransmit queue. */
242 list_for_each_safe(lchunk, temp, &q->retransmit) {
243 list_del_init(lchunk);
244 chunk = list_entry(lchunk, struct sctp_chunk,
245 transmitted_list);
246 sctp_chunk_fail(chunk, q->error);
247 sctp_chunk_free(chunk);
248 }
249
250 /* Throw away any chunks that are in the abandoned queue. */
251 list_for_each_safe(lchunk, temp, &q->abandoned) {
252 list_del_init(lchunk);
253 chunk = list_entry(lchunk, struct sctp_chunk,
254 transmitted_list);
255 sctp_chunk_fail(chunk, q->error);
256 sctp_chunk_free(chunk);
257 }
258
259 /* Throw away any leftover data chunks. */
260 while ((chunk = sctp_outq_dequeue_data(q)) != NULL) {
261
262 /* Mark as send failure. */
263 sctp_chunk_fail(chunk, q->error);
264 sctp_chunk_free(chunk);
265 }
266
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 /* Throw away any leftover control chunks. */
David S. Miller79af02c2005-07-08 21:47:49 -0700268 list_for_each_entry_safe(chunk, tmp, &q->control_chunk_list, list) {
269 list_del_init(&chunk->list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 sctp_chunk_free(chunk);
David S. Miller79af02c2005-07-08 21:47:49 -0700271 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272}
273
Neil Horman2f94aab2013-01-17 11:15:08 +0000274void sctp_outq_teardown(struct sctp_outq *q)
275{
276 __sctp_outq_teardown(q);
277 sctp_outq_init(q->asoc, q);
278}
279
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280/* Free the outqueue structure and any related pending chunks. */
281void sctp_outq_free(struct sctp_outq *q)
282{
283 /* Throw away leftover chunks. */
Neil Horman2f94aab2013-01-17 11:15:08 +0000284 __sctp_outq_teardown(q);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285}
286
287/* Put a new chunk in an sctp_outq. */
288int sctp_outq_tail(struct sctp_outq *q, struct sctp_chunk *chunk)
289{
Eric W. Biedermanb01a2402012-08-06 08:47:55 +0000290 struct net *net = sock_net(q->asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 int error = 0;
292
Daniel Borkmannbb333812013-06-28 19:49:40 +0200293 pr_debug("%s: outq:%p, chunk:%p[%s]\n", __func__, q, chunk,
294 chunk && chunk->chunk_hdr ?
295 sctp_cname(SCTP_ST_CHUNK(chunk->chunk_hdr->type)) :
296 "illegal chunk");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297
298 /* If it is data, queue it up, otherwise, send it
299 * immediately.
300 */
Shan Weiec7b9512010-04-30 22:41:09 -0400301 if (sctp_chunk_is_data(chunk)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 /* Is it OK to queue data chunks? */
303 /* From 9. Termination of Association
304 *
305 * When either endpoint performs a shutdown, the
306 * association on each peer will stop accepting new
307 * data from its user and only deliver data in queue
308 * at the time of sending or receiving the SHUTDOWN
309 * chunk.
310 */
311 switch (q->asoc->state) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 case SCTP_STATE_CLOSED:
313 case SCTP_STATE_SHUTDOWN_PENDING:
314 case SCTP_STATE_SHUTDOWN_SENT:
315 case SCTP_STATE_SHUTDOWN_RECEIVED:
316 case SCTP_STATE_SHUTDOWN_ACK_SENT:
317 /* Cannot send after transport endpoint shutdown */
318 error = -ESHUTDOWN;
319 break;
320
321 default:
Daniel Borkmannbb333812013-06-28 19:49:40 +0200322 pr_debug("%s: outqueueing: outq:%p, chunk:%p[%s])\n",
323 __func__, q, chunk, chunk && chunk->chunk_hdr ?
324 sctp_cname(SCTP_ST_CHUNK(chunk->chunk_hdr->type)) :
325 "illegal chunk");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326
327 sctp_outq_tail_data(q, chunk);
328 if (chunk->chunk_hdr->flags & SCTP_DATA_UNORDERED)
Eric W. Biedermanb01a2402012-08-06 08:47:55 +0000329 SCTP_INC_STATS(net, SCTP_MIB_OUTUNORDERCHUNKS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 else
Eric W. Biedermanb01a2402012-08-06 08:47:55 +0000331 SCTP_INC_STATS(net, SCTP_MIB_OUTORDERCHUNKS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -0700333 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 } else {
David S. Miller79af02c2005-07-08 21:47:49 -0700335 list_add_tail(&chunk->list, &q->control_chunk_list);
Eric W. Biedermanb01a2402012-08-06 08:47:55 +0000336 SCTP_INC_STATS(net, SCTP_MIB_OUTCTRLCHUNKS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 }
338
339 if (error < 0)
340 return error;
341
342 if (!q->cork)
343 error = sctp_outq_flush(q, 0);
344
345 return error;
346}
347
348/* Insert a chunk into the sorted list based on the TSNs. The retransmit list
349 * and the abandoned list are in ascending order.
350 */
351static void sctp_insert_list(struct list_head *head, struct list_head *new)
352{
353 struct list_head *pos;
354 struct sctp_chunk *nchunk, *lchunk;
355 __u32 ntsn, ltsn;
356 int done = 0;
357
358 nchunk = list_entry(new, struct sctp_chunk, transmitted_list);
359 ntsn = ntohl(nchunk->subh.data_hdr->tsn);
360
361 list_for_each(pos, head) {
362 lchunk = list_entry(pos, struct sctp_chunk, transmitted_list);
363 ltsn = ntohl(lchunk->subh.data_hdr->tsn);
364 if (TSN_lt(ntsn, ltsn)) {
365 list_add(new, pos->prev);
366 done = 1;
367 break;
368 }
369 }
370 if (!done)
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900371 list_add_tail(new, head);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372}
373
374/* Mark all the eligible packets on a transport for retransmission. */
375void sctp_retransmit_mark(struct sctp_outq *q,
376 struct sctp_transport *transport,
Vlad Yasevichb6157d82007-10-24 15:59:16 -0400377 __u8 reason)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378{
379 struct list_head *lchunk, *ltemp;
380 struct sctp_chunk *chunk;
381
382 /* Walk through the specified transmitted queue. */
383 list_for_each_safe(lchunk, ltemp, &transport->transmitted) {
384 chunk = list_entry(lchunk, struct sctp_chunk,
385 transmitted_list);
386
387 /* If the chunk is abandoned, move it to abandoned list. */
388 if (sctp_chunk_abandoned(chunk)) {
389 list_del_init(lchunk);
390 sctp_insert_list(&q->abandoned, lchunk);
Vlad Yasevich8c4a2d42007-02-21 02:06:04 -0800391
392 /* If this chunk has not been previousely acked,
393 * stop considering it 'outstanding'. Our peer
394 * will most likely never see it since it will
395 * not be retransmitted
396 */
397 if (!chunk->tsn_gap_acked) {
Vlad Yasevich31b02e12009-09-04 18:21:00 -0400398 if (chunk->transport)
399 chunk->transport->flight_size -=
400 sctp_data_size(chunk);
Vlad Yasevich8c4a2d42007-02-21 02:06:04 -0800401 q->outstanding_bytes -= sctp_data_size(chunk);
Thomas Grafa76c0ad2011-12-19 04:11:40 +0000402 q->asoc->peer.rwnd += sctp_data_size(chunk);
Vlad Yasevich8c4a2d42007-02-21 02:06:04 -0800403 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 continue;
405 }
406
Vlad Yasevichb6157d82007-10-24 15:59:16 -0400407 /* If we are doing retransmission due to a timeout or pmtu
408 * discovery, only the chunks that are not yet acked should
409 * be added to the retransmit queue.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 */
Vlad Yasevichb6157d82007-10-24 15:59:16 -0400411 if ((reason == SCTP_RTXR_FAST_RTX &&
Neil Hormanc226ef92008-07-25 12:44:09 -0400412 (chunk->fast_retransmit == SCTP_NEED_FRTX)) ||
Vlad Yasevichb6157d82007-10-24 15:59:16 -0400413 (reason != SCTP_RTXR_FAST_RTX && !chunk->tsn_gap_acked)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 /* RFC 2960 6.2.1 Processing a Received SACK
415 *
416 * C) Any time a DATA chunk is marked for
417 * retransmission (via either T3-rtx timer expiration
418 * (Section 6.3.3) or via fast retransmit
419 * (Section 7.2.4)), add the data size of those
420 * chunks to the rwnd.
421 */
Thomas Grafa76c0ad2011-12-19 04:11:40 +0000422 q->asoc->peer.rwnd += sctp_data_size(chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 q->outstanding_bytes -= sctp_data_size(chunk);
Vlad Yasevich31b02e12009-09-04 18:21:00 -0400424 if (chunk->transport)
425 transport->flight_size -= sctp_data_size(chunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426
427 /* sctpimpguide-05 Section 2.8.2
428 * M5) If a T3-rtx timer expires, the
429 * 'TSN.Missing.Report' of all affected TSNs is set
430 * to 0.
431 */
432 chunk->tsn_missing_report = 0;
433
434 /* If a chunk that is being used for RTT measurement
435 * has to be retransmitted, we cannot use this chunk
436 * anymore for RTT measurements. Reset rto_pending so
437 * that a new RTT measurement is started when a new
438 * data chunk is sent.
439 */
440 if (chunk->rtt_in_progress) {
441 chunk->rtt_in_progress = 0;
442 transport->rto_pending = 0;
443 }
444
Xufeng Zhang6eabca52013-11-25 11:26:57 +0800445 chunk->resent = 1;
446
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 /* Move the chunk to the retransmit queue. The chunks
448 * on the retransmit queue are always kept in order.
449 */
450 list_del_init(lchunk);
451 sctp_insert_list(&q->retransmit, lchunk);
452 }
453 }
454
Daniel Borkmannbb333812013-06-28 19:49:40 +0200455 pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d, "
456 "flight_size:%d, pba:%d\n", __func__, transport, reason,
457 transport->cwnd, transport->ssthresh, transport->flight_size,
458 transport->partial_bytes_acked);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459}
460
461/* Mark all the eligible packets on a transport for retransmission and force
462 * one packet out.
463 */
464void sctp_retransmit(struct sctp_outq *q, struct sctp_transport *transport,
465 sctp_retransmit_reason_t reason)
466{
Eric W. Biedermanb01a2402012-08-06 08:47:55 +0000467 struct net *net = sock_net(q->asoc->base.sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
wangweidongcb3f8372013-12-23 12:16:50 +0800470 switch (reason) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 case SCTP_RTXR_T3_RTX:
Eric W. Biedermanb01a2402012-08-06 08:47:55 +0000472 SCTP_INC_STATS(net, SCTP_MIB_T3_RETRANSMITS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 sctp_transport_lower_cwnd(transport, SCTP_LOWER_CWND_T3_RTX);
474 /* Update the retran path if the T3-rtx timer has expired for
475 * the current retran path.
476 */
477 if (transport == transport->asoc->peer.retran_path)
478 sctp_assoc_update_retran_path(transport->asoc);
Neil Horman58fbbed2008-02-29 11:40:56 -0800479 transport->asoc->rtx_data_chunks +=
480 transport->asoc->unack_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 break;
482 case SCTP_RTXR_FAST_RTX:
Eric W. Biedermanb01a2402012-08-06 08:47:55 +0000483 SCTP_INC_STATS(net, SCTP_MIB_FAST_RETRANSMITS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 sctp_transport_lower_cwnd(transport, SCTP_LOWER_CWND_FAST_RTX);
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700485 q->fast_rtx = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 break;
487 case SCTP_RTXR_PMTUD:
Eric W. Biedermanb01a2402012-08-06 08:47:55 +0000488 SCTP_INC_STATS(net, SCTP_MIB_PMTUD_RETRANSMITS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 break;
Vlad Yasevichb6157d82007-10-24 15:59:16 -0400490 case SCTP_RTXR_T1_RTX:
Eric W. Biedermanb01a2402012-08-06 08:47:55 +0000491 SCTP_INC_STATS(net, SCTP_MIB_T1_RETRANSMITS);
Neil Horman58fbbed2008-02-29 11:40:56 -0800492 transport->asoc->init_retries++;
Vlad Yasevichb6157d82007-10-24 15:59:16 -0400493 break;
Sridhar Samudralaac0b0462006-08-22 00:15:33 -0700494 default:
495 BUG();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 }
497
Vlad Yasevichb6157d82007-10-24 15:59:16 -0400498 sctp_retransmit_mark(q, transport, reason);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
500 /* PR-SCTP A5) Any time the T3-rtx timer expires, on any destination,
501 * the sender SHOULD try to advance the "Advanced.Peer.Ack.Point" by
502 * following the procedures outlined in C1 - C5.
503 */
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700504 if (reason == SCTP_RTXR_T3_RTX)
505 sctp_generate_fwdtsn(q, q->asoc->ctsn_ack_point);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700507 /* Flush the queues only on timeout, since fast_rtx is only
508 * triggered during sack processing and the queue
509 * will be flushed at the end.
510 */
511 if (reason != SCTP_RTXR_FAST_RTX)
512 error = sctp_outq_flush(q, /* rtx_timeout */ 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513
514 if (error)
515 q->asoc->base.sk->sk_err = -error;
516}
517
518/*
519 * Transmit DATA chunks on the retransmit queue. Upon return from
520 * sctp_outq_flush_rtx() the packet 'pkt' may contain chunks which
521 * need to be transmitted by the caller.
522 * We assume that pkt->transport has already been set.
523 *
524 * The return value is a normal kernel error return value.
525 */
526static int sctp_outq_flush_rtx(struct sctp_outq *q, struct sctp_packet *pkt,
527 int rtx_timeout, int *start_timer)
528{
529 struct list_head *lqueue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 struct sctp_transport *transport = pkt->transport;
531 sctp_xmit_t status;
532 struct sctp_chunk *chunk, *chunk1;
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700533 int fast_rtx;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 int error = 0;
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700535 int timer = 0;
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700536 int done = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 lqueue = &q->retransmit;
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700539 fast_rtx = q->fast_rtx;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700541 /* This loop handles time-out retransmissions, fast retransmissions,
542 * and retransmissions due to opening of whindow.
543 *
544 * RFC 2960 6.3.3 Handle T3-rtx Expiration
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 *
546 * E3) Determine how many of the earliest (i.e., lowest TSN)
547 * outstanding DATA chunks for the address for which the
548 * T3-rtx has expired will fit into a single packet, subject
549 * to the MTU constraint for the path corresponding to the
550 * destination transport address to which the retransmission
551 * is being sent (this may be different from the address for
552 * which the timer expires [see Section 6.4]). Call this value
553 * K. Bundle and retransmit those K DATA chunks in a single
554 * packet to the destination endpoint.
555 *
556 * [Just to be painfully clear, if we are retransmitting
557 * because a timeout just happened, we should send only ONE
558 * packet of retransmitted data.]
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700559 *
560 * For fast retransmissions we also send only ONE packet. However,
561 * if we are just flushing the queue due to open window, we'll
562 * try to send as much as possible.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 */
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700564 list_for_each_entry_safe(chunk, chunk1, lqueue, transmitted_list) {
Wei Yongjun4c6a6f42011-04-19 21:32:28 +0000565 /* If the chunk is abandoned, move it to abandoned list. */
566 if (sctp_chunk_abandoned(chunk)) {
567 list_del_init(&chunk->transmitted_list);
568 sctp_insert_list(&q->abandoned,
569 &chunk->transmitted_list);
570 continue;
571 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572
573 /* Make sure that Gap Acked TSNs are not retransmitted. A
574 * simple approach is just to move such TSNs out of the
575 * way and into a 'transmitted' queue and skip to the
576 * next chunk.
577 */
578 if (chunk->tsn_gap_acked) {
Wei Yongjun54a27922012-09-03 23:58:16 +0000579 list_move_tail(&chunk->transmitted_list,
580 &transport->transmitted);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 continue;
582 }
583
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700584 /* If we are doing fast retransmit, ignore non-fast_rtransmit
585 * chunks
586 */
587 if (fast_rtx && !chunk->fast_retransmit)
588 continue;
589
Wei Yongjunbc4f8412010-04-30 22:38:53 -0400590redo:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 /* Attempt to append this chunk to the packet. */
592 status = sctp_packet_append_chunk(pkt, chunk);
593
594 switch (status) {
595 case SCTP_XMIT_PMTU_FULL:
Wei Yongjunbc4f8412010-04-30 22:38:53 -0400596 if (!pkt->has_data && !pkt->has_cookie_echo) {
597 /* If this packet did not contain DATA then
598 * retransmission did not happen, so do it
599 * again. We'll ignore the error here since
600 * control chunks are already freed so there
601 * is nothing we can do.
602 */
603 sctp_packet_transmit(pkt);
604 goto redo;
605 }
606
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 /* Send this packet. */
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700608 error = sctp_packet_transmit(pkt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
610 /* If we are retransmitting, we should only
611 * send a single packet.
Vlad Yasevichf246a7b2011-04-18 19:13:56 +0000612 * Otherwise, try appending this chunk again.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613 */
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700614 if (rtx_timeout || fast_rtx)
615 done = 1;
Vlad Yasevichf246a7b2011-04-18 19:13:56 +0000616 else
617 goto redo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700619 /* Bundle next chunk in the next round. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 break;
621
622 case SCTP_XMIT_RWND_FULL:
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900623 /* Send this packet. */
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700624 error = sctp_packet_transmit(pkt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625
626 /* Stop sending DATA as there is no more room
627 * at the receiver.
628 */
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700629 done = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 break;
631
632 case SCTP_XMIT_NAGLE_DELAY:
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900633 /* Send this packet. */
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700634 error = sctp_packet_transmit(pkt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635
636 /* Stop sending DATA because of nagle delay. */
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700637 done = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 break;
639
640 default:
641 /* The append was successful, so add this chunk to
642 * the transmitted list.
643 */
Wei Yongjun54a27922012-09-03 23:58:16 +0000644 list_move_tail(&chunk->transmitted_list,
645 &transport->transmitted);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900647 /* Mark the chunk as ineligible for fast retransmit
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 * after it is retransmitted.
649 */
Neil Hormanc226ef92008-07-25 12:44:09 -0400650 if (chunk->fast_retransmit == SCTP_NEED_FRTX)
651 chunk->fast_retransmit = SCTP_DONT_FRTX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
Michele Baldessari196d6752012-12-01 04:49:42 +0000653 q->asoc->stats.rtxchunks++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 break;
Stephen Hemminger3ff50b72007-04-20 17:09:22 -0700655 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700657 /* Set the timer if there were no errors */
658 if (!error && !timer)
659 timer = 1;
660
Vlad Yasevich8b750ce2008-06-04 12:39:36 -0700661 if (done)
662 break;
663 }
664
665 /* If we are here due to a retransmit timeout or a fast
666 * retransmit and if there are any chunks left in the retransmit
667 * queue that could not fit in the PMTU sized packet, they need
668 * to be marked as ineligible for a subsequent fast retransmit.
669 */
670 if (rtx_timeout || fast_rtx) {
671 list_for_each_entry(chunk1, lqueue, transmitted_list) {
Neil Hormanc226ef92008-07-25 12:44:09 -0400672 if (chunk1->fast_retransmit == SCTP_NEED_FRTX)
673 chunk1->fast_retransmit = SCTP_DONT_FRTX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 }
675 }
676
Vlad Yasevich62aeaff2008-06-04 12:39:11 -0700677 *start_timer = timer;
678
679 /* Clear fast retransmit hint */
680 if (fast_rtx)
681 q->fast_rtx = 0;
682
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 return error;
684}
685
686/* Cork the outqueue so queued chunks are really queued. */
687int sctp_outq_uncork(struct sctp_outq *q)
688{
Vlad Yasevich7d54dc62007-11-09 11:43:41 -0500689 if (q->cork)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 q->cork = 0;
Daniel Borkmanndacda322013-04-16 11:07:14 +0000691
692 return sctp_outq_flush(q, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693}
694
Vlad Yasevich2e3216c2008-06-19 16:08:18 -0700695
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696/*
697 * Try to flush an outqueue.
698 *
699 * Description: Send everything in q which we legally can, subject to
700 * congestion limitations.
701 * * Note: This function can be called from multiple contexts so appropriate
702 * locking concerns must be made. Today we use the sock lock to protect
703 * this function.
704 */
Adrian Bunkabd0b1982008-07-22 14:20:45 -0700705static int sctp_outq_flush(struct sctp_outq *q, int rtx_timeout)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706{
707 struct sctp_packet *packet;
708 struct sctp_packet singleton;
709 struct sctp_association *asoc = q->asoc;
710 __u16 sport = asoc->base.bind_addr.port;
711 __u16 dport = asoc->peer.port;
712 __u32 vtag = asoc->peer.i.init_tag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 struct sctp_transport *transport = NULL;
714 struct sctp_transport *new_transport;
David S. Miller79af02c2005-07-08 21:47:49 -0700715 struct sctp_chunk *chunk, *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 sctp_xmit_t status;
717 int error = 0;
718 int start_timer = 0;
Vlad Yasevich2e3216c2008-06-19 16:08:18 -0700719 int one_packet = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720
721 /* These transports have chunks to send. */
722 struct list_head transport_list;
723 struct list_head *ltransport;
724
725 INIT_LIST_HEAD(&transport_list);
726 packet = NULL;
727
728 /*
729 * 6.10 Bundling
730 * ...
731 * When bundling control chunks with DATA chunks, an
732 * endpoint MUST place control chunks first in the outbound
733 * SCTP packet. The transmitter MUST transmit DATA chunks
734 * within a SCTP packet in increasing order of TSN.
735 * ...
736 */
737
David S. Miller79af02c2005-07-08 21:47:49 -0700738 list_for_each_entry_safe(chunk, tmp, &q->control_chunk_list, list) {
Michio Honda8a07eb02011-04-26 20:19:36 +0900739 /* RFC 5061, 5.3
740 * F1) This means that until such time as the ASCONF
741 * containing the add is acknowledged, the sender MUST
742 * NOT use the new IP address as a source for ANY SCTP
743 * packet except on carrying an ASCONF Chunk.
744 */
745 if (asoc->src_out_of_asoc_ok &&
746 chunk->chunk_hdr->type != SCTP_CID_ASCONF)
747 continue;
748
David S. Miller79af02c2005-07-08 21:47:49 -0700749 list_del_init(&chunk->list);
750
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 /* Pick the right transport to use. */
752 new_transport = chunk->transport;
753
754 if (!new_transport) {
Vlad Yasevicha08de642007-12-20 14:11:47 -0800755 /*
756 * If we have a prior transport pointer, see if
757 * the destination address of the chunk
758 * matches the destination address of the
759 * current transport. If not a match, then
760 * try to look up the transport with a given
761 * destination address. We do this because
762 * after processing ASCONFs, we may have new
763 * transports created.
764 */
765 if (transport &&
766 sctp_cmp_addr_exact(&chunk->dest,
767 &transport->ipaddr))
768 new_transport = transport;
769 else
770 new_transport = sctp_assoc_lookup_paddr(asoc,
771 &chunk->dest);
772
773 /* if we still don't have a new transport, then
774 * use the current active path.
775 */
776 if (!new_transport)
777 new_transport = asoc->peer.active_path;
Sridhar Samudralaad8fec12006-07-21 14:48:50 -0700778 } else if ((new_transport->state == SCTP_INACTIVE) ||
Neil Horman5aa93bc2012-07-21 07:56:07 +0000779 (new_transport->state == SCTP_UNCONFIRMED) ||
780 (new_transport->state == SCTP_PF)) {
Frank Filz3f7a87d2005-06-20 13:14:57 -0700781 /* If the chunk is Heartbeat or Heartbeat Ack,
782 * send it to chunk->transport, even if it's
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 * inactive.
784 *
785 * 3.3.6 Heartbeat Acknowledgement:
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900786 * ...
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 * A HEARTBEAT ACK is always sent to the source IP
788 * address of the IP datagram containing the
789 * HEARTBEAT chunk to which this ack is responding.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +0900790 * ...
Vlad Yasevicha08de642007-12-20 14:11:47 -0800791 *
792 * ASCONF_ACKs also must be sent to the source.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 */
794 if (chunk->chunk_hdr->type != SCTP_CID_HEARTBEAT &&
Vlad Yasevicha08de642007-12-20 14:11:47 -0800795 chunk->chunk_hdr->type != SCTP_CID_HEARTBEAT_ACK &&
796 chunk->chunk_hdr->type != SCTP_CID_ASCONF_ACK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 new_transport = asoc->peer.active_path;
798 }
799
800 /* Are we switching transports?
801 * Take care of transport locks.
802 */
803 if (new_transport != transport) {
804 transport = new_transport;
805 if (list_empty(&transport->send_ready)) {
806 list_add_tail(&transport->send_ready,
807 &transport_list);
808 }
809 packet = &transport->packet;
810 sctp_packet_config(packet, vtag,
811 asoc->peer.ecn_capable);
812 }
813
814 switch (chunk->chunk_hdr->type) {
815 /*
816 * 6.10 Bundling
817 * ...
818 * An endpoint MUST NOT bundle INIT, INIT ACK or SHUTDOWN
819 * COMPLETE with any other chunks. [Send them immediately.]
820 */
821 case SCTP_CID_INIT:
822 case SCTP_CID_INIT_ACK:
823 case SCTP_CID_SHUTDOWN_COMPLETE:
824 sctp_packet_init(&singleton, transport, sport, dport);
825 sctp_packet_config(&singleton, vtag, 0);
826 sctp_packet_append_chunk(&singleton, chunk);
827 error = sctp_packet_transmit(&singleton);
828 if (error < 0)
829 return error;
830 break;
831
832 case SCTP_CID_ABORT:
Gui Jianfengf4ad85c2008-04-12 18:39:34 -0700833 if (sctp_test_T_bit(chunk)) {
834 packet->vtag = asoc->c.my_vtag;
835 }
Vlad Yasevich2e3216c2008-06-19 16:08:18 -0700836 /* The following chunks are "response" chunks, i.e.
837 * they are generated in response to something we
838 * received. If we are sending these, then we can
839 * send only 1 packet containing these chunks.
840 */
841 case SCTP_CID_HEARTBEAT_ACK:
842 case SCTP_CID_SHUTDOWN_ACK:
843 case SCTP_CID_COOKIE_ACK:
844 case SCTP_CID_COOKIE_ECHO:
845 case SCTP_CID_ERROR:
846 case SCTP_CID_ECN_CWR:
847 case SCTP_CID_ASCONF_ACK:
848 one_packet = 1;
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300849 /* Fall through */
Vlad Yasevich2e3216c2008-06-19 16:08:18 -0700850
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 case SCTP_CID_SACK:
852 case SCTP_CID_HEARTBEAT:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 case SCTP_CID_SHUTDOWN:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 case SCTP_CID_ECN_ECNE:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 case SCTP_CID_ASCONF:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 case SCTP_CID_FWD_TSN:
Vlad Yasevich2e3216c2008-06-19 16:08:18 -0700857 status = sctp_packet_transmit_chunk(packet, chunk,
858 one_packet);
859 if (status != SCTP_XMIT_OK) {
860 /* put the chunk back */
861 list_add(&chunk->list, &q->control_chunk_list);
Michele Baldessari196d6752012-12-01 04:49:42 +0000862 } else {
863 asoc->stats.octrlchunks++;
Wei Yongjunbd69b982010-04-30 21:42:43 -0400864 /* PR-SCTP C5) If a FORWARD TSN is sent, the
865 * sender MUST assure that at least one T3-rtx
866 * timer is running.
867 */
Michele Baldessari196d6752012-12-01 04:49:42 +0000868 if (chunk->chunk_hdr->type == SCTP_CID_FWD_TSN)
869 sctp_transport_reset_timers(transport);
Vlad Yasevich2e3216c2008-06-19 16:08:18 -0700870 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 break;
872
873 default:
874 /* We built a chunk with an illegal type! */
875 BUG();
Stephen Hemminger3ff50b72007-04-20 17:09:22 -0700876 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 }
878
Michio Honda8a07eb02011-04-26 20:19:36 +0900879 if (q->asoc->src_out_of_asoc_ok)
880 goto sctp_flush_out;
881
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 /* Is it OK to send data chunks? */
883 switch (asoc->state) {
884 case SCTP_STATE_COOKIE_ECHOED:
885 /* Only allow bundling when this packet has a COOKIE-ECHO
886 * chunk.
887 */
888 if (!packet || !packet->has_cookie_echo)
889 break;
890
891 /* fallthru */
892 case SCTP_STATE_ESTABLISHED:
893 case SCTP_STATE_SHUTDOWN_PENDING:
894 case SCTP_STATE_SHUTDOWN_RECEIVED:
895 /*
896 * RFC 2960 6.1 Transmission of DATA Chunks
897 *
898 * C) When the time comes for the sender to transmit,
899 * before sending new DATA chunks, the sender MUST
900 * first transmit any outstanding DATA chunks which
901 * are marked for retransmission (limited by the
902 * current cwnd).
903 */
904 if (!list_empty(&q->retransmit)) {
Michio Hondaf207c052011-06-16 10:54:23 +0900905 if (asoc->peer.retran_path->state == SCTP_UNCONFIRMED)
906 goto sctp_flush_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 if (transport == asoc->peer.retran_path)
908 goto retran;
909
910 /* Switch transports & prepare the packet. */
911
912 transport = asoc->peer.retran_path;
913
914 if (list_empty(&transport->send_ready)) {
915 list_add_tail(&transport->send_ready,
916 &transport_list);
917 }
918
919 packet = &transport->packet;
920 sctp_packet_config(packet, vtag,
921 asoc->peer.ecn_capable);
922 retran:
923 error = sctp_outq_flush_rtx(q, packet,
924 rtx_timeout, &start_timer);
925
926 if (start_timer)
Vlad Yasevichd9efc222010-04-30 22:41:09 -0400927 sctp_transport_reset_timers(transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928
929 /* This can happen on COOKIE-ECHO resend. Only
930 * one chunk can get bundled with a COOKIE-ECHO.
931 */
932 if (packet->has_cookie_echo)
933 goto sctp_flush_out;
934
935 /* Don't send new data if there is still data
936 * waiting to retransmit.
937 */
938 if (!list_empty(&q->retransmit))
939 goto sctp_flush_out;
940 }
941
Vlad Yasevich46d5a802009-11-23 15:54:00 -0500942 /* Apply Max.Burst limitation to the current transport in
943 * case it will be used for new data. We are going to
944 * rest it before we return, but we want to apply the limit
945 * to the currently queued data.
946 */
947 if (transport)
948 sctp_transport_burst_limited(transport);
949
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 /* Finally, transmit new packets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 while ((chunk = sctp_outq_dequeue_data(q)) != NULL) {
952 /* RFC 2960 6.5 Every DATA chunk MUST carry a valid
953 * stream identifier.
954 */
955 if (chunk->sinfo.sinfo_stream >=
956 asoc->c.sinit_num_ostreams) {
957
958 /* Mark as failed send. */
959 sctp_chunk_fail(chunk, SCTP_ERROR_INV_STRM);
960 sctp_chunk_free(chunk);
961 continue;
962 }
963
964 /* Has this chunk expired? */
965 if (sctp_chunk_abandoned(chunk)) {
966 sctp_chunk_fail(chunk, 0);
967 sctp_chunk_free(chunk);
968 continue;
969 }
970
971 /* If there is a specified transport, use it.
972 * Otherwise, we want to use the active path.
973 */
974 new_transport = chunk->transport;
Frank Filz3f7a87d2005-06-20 13:14:57 -0700975 if (!new_transport ||
Sridhar Samudralaad8fec12006-07-21 14:48:50 -0700976 ((new_transport->state == SCTP_INACTIVE) ||
Neil Horman5aa93bc2012-07-21 07:56:07 +0000977 (new_transport->state == SCTP_UNCONFIRMED) ||
978 (new_transport->state == SCTP_PF)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 new_transport = asoc->peer.active_path;
Michio Hondaf207c052011-06-16 10:54:23 +0900980 if (new_transport->state == SCTP_UNCONFIRMED)
981 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
983 /* Change packets if necessary. */
984 if (new_transport != transport) {
985 transport = new_transport;
986
987 /* Schedule to have this transport's
988 * packet flushed.
989 */
990 if (list_empty(&transport->send_ready)) {
991 list_add_tail(&transport->send_ready,
992 &transport_list);
993 }
994
995 packet = &transport->packet;
996 sctp_packet_config(packet, vtag,
997 asoc->peer.ecn_capable);
Vlad Yasevich46d5a802009-11-23 15:54:00 -0500998 /* We've switched transports, so apply the
999 * Burst limit to the new transport.
1000 */
1001 sctp_transport_burst_limited(transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 }
1003
Daniel Borkmannbb333812013-06-28 19:49:40 +02001004 pr_debug("%s: outq:%p, chunk:%p[%s], tx-tsn:0x%x skb->head:%p "
1005 "skb->users:%d\n",
1006 __func__, q, chunk, chunk && chunk->chunk_hdr ?
1007 sctp_cname(SCTP_ST_CHUNK(chunk->chunk_hdr->type)) :
1008 "illegal chunk", ntohl(chunk->subh.data_hdr->tsn),
1009 chunk->skb ? chunk->skb->head : NULL, chunk->skb ?
1010 atomic_read(&chunk->skb->users) : -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
1012 /* Add the chunk to the packet. */
Vlad Yasevich2e3216c2008-06-19 16:08:18 -07001013 status = sctp_packet_transmit_chunk(packet, chunk, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
1015 switch (status) {
1016 case SCTP_XMIT_PMTU_FULL:
1017 case SCTP_XMIT_RWND_FULL:
1018 case SCTP_XMIT_NAGLE_DELAY:
1019 /* We could not append this chunk, so put
1020 * the chunk back on the output queue.
1021 */
Daniel Borkmannbb333812013-06-28 19:49:40 +02001022 pr_debug("%s: could not transmit tsn:0x%x, status:%d\n",
1023 __func__, ntohl(chunk->subh.data_hdr->tsn),
1024 status);
1025
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 sctp_outq_head_data(q, chunk);
1027 goto sctp_flush_out;
1028 break;
1029
1030 case SCTP_XMIT_OK:
Wei Yongjunb93d6472009-11-23 15:53:56 -05001031 /* The sender is in the SHUTDOWN-PENDING state,
1032 * The sender MAY set the I-bit in the DATA
1033 * chunk header.
1034 */
1035 if (asoc->state == SCTP_STATE_SHUTDOWN_PENDING)
1036 chunk->chunk_hdr->flags |= SCTP_DATA_SACK_IMM;
Michele Baldessari196d6752012-12-01 04:49:42 +00001037 if (chunk->chunk_hdr->flags & SCTP_DATA_UNORDERED)
1038 asoc->stats.ouodchunks++;
1039 else
1040 asoc->stats.oodchunks++;
Wei Yongjunb93d6472009-11-23 15:53:56 -05001041
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 break;
1043
1044 default:
1045 BUG();
1046 }
1047
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001048 /* BUG: We assume that the sctp_packet_transmit()
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 * call below will succeed all the time and add the
1050 * chunk to the transmitted list and restart the
1051 * timers.
1052 * It is possible that the call can fail under OOM
1053 * conditions.
1054 *
1055 * Is this really a problem? Won't this behave
1056 * like a lost TSN?
1057 */
1058 list_add_tail(&chunk->transmitted_list,
1059 &transport->transmitted);
1060
Vlad Yasevichd9efc222010-04-30 22:41:09 -04001061 sctp_transport_reset_timers(transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 /* Only let one DATA chunk get bundled with a
1064 * COOKIE-ECHO chunk.
1065 */
1066 if (packet->has_cookie_echo)
1067 goto sctp_flush_out;
1068 }
1069 break;
1070
1071 default:
1072 /* Do nothing. */
1073 break;
1074 }
1075
1076sctp_flush_out:
1077
1078 /* Before returning, examine all the transports touched in
1079 * this call. Right now, we bluntly force clear all the
1080 * transports. Things might change after we implement Nagle.
1081 * But such an examination is still required.
1082 *
1083 * --xguo
1084 */
wangweidongcb3f8372013-12-23 12:16:50 +08001085 while ((ltransport = sctp_list_dequeue(&transport_list)) != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 struct sctp_transport *t = list_entry(ltransport,
1087 struct sctp_transport,
1088 send_ready);
1089 packet = &t->packet;
1090 if (!sctp_packet_empty(packet))
1091 error = sctp_packet_transmit(packet);
Vlad Yasevich46d5a802009-11-23 15:54:00 -05001092
1093 /* Clear the burst limited state, if any */
1094 sctp_transport_burst_reset(t);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 }
1096
1097 return error;
1098}
1099
1100/* Update unack_data based on the incoming SACK chunk */
1101static void sctp_sack_update_unack_data(struct sctp_association *assoc,
1102 struct sctp_sackhdr *sack)
1103{
1104 sctp_sack_variable_t *frags;
1105 __u16 unack_data;
1106 int i;
1107
1108 unack_data = assoc->next_tsn - assoc->ctsn_ack_point - 1;
1109
1110 frags = sack->variable;
1111 for (i = 0; i < ntohs(sack->num_gap_ack_blocks); i++) {
1112 unack_data -= ((ntohs(frags[i].gab.end) -
1113 ntohs(frags[i].gab.start) + 1));
1114 }
1115
1116 assoc->unack_data = unack_data;
1117}
1118
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119/* This is where we REALLY process a SACK.
1120 *
1121 * Process the SACK against the outqueue. Mostly, this just frees
1122 * things off the transmitted queue.
1123 */
Nicolas Dichteledfee032012-10-03 05:43:22 +00001124int sctp_outq_sack(struct sctp_outq *q, struct sctp_chunk *chunk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125{
1126 struct sctp_association *asoc = q->asoc;
Nicolas Dichteledfee032012-10-03 05:43:22 +00001127 struct sctp_sackhdr *sack = chunk->subh.sack_hdr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 struct sctp_transport *transport;
1129 struct sctp_chunk *tchunk = NULL;
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07001130 struct list_head *lchunk, *transport_list, *temp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 sctp_sack_variable_t *frags = sack->variable;
1132 __u32 sack_ctsn, ctsn, tsn;
1133 __u32 highest_tsn, highest_new_tsn;
1134 __u32 sack_a_rwnd;
Eric Dumazet95c96172012-04-15 05:58:06 +00001135 unsigned int outstanding;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 struct sctp_transport *primary = asoc->peer.primary_path;
1137 int count_of_newacks = 0;
Vlad Yasevich2cd9b822008-06-19 17:59:13 -04001138 int gap_ack_blocks;
Vlad Yasevichea862c82010-04-30 22:41:10 -04001139 u8 accum_moved = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140
1141 /* Grab the association's destination address list. */
1142 transport_list = &asoc->peer.transport_addr_list;
1143
1144 sack_ctsn = ntohl(sack->cum_tsn_ack);
Vlad Yasevich2cd9b822008-06-19 17:59:13 -04001145 gap_ack_blocks = ntohs(sack->num_gap_ack_blocks);
Michele Baldessari196d6752012-12-01 04:49:42 +00001146 asoc->stats.gapcnt += gap_ack_blocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 /*
1148 * SFR-CACC algorithm:
1149 * On receipt of a SACK the sender SHOULD execute the
1150 * following statements.
1151 *
1152 * 1) If the cumulative ack in the SACK passes next tsn_at_change
1153 * on the current primary, the CHANGEOVER_ACTIVE flag SHOULD be
1154 * cleared. The CYCLING_CHANGEOVER flag SHOULD also be cleared for
1155 * all destinations.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156 * 2) If the SACK contains gap acks and the flag CHANGEOVER_ACTIVE
1157 * is set the receiver of the SACK MUST take the following actions:
1158 *
1159 * A) Initialize the cacc_saw_newack to 0 for all destination
1160 * addresses.
Vlad Yasevichab5216a2008-06-19 18:17:24 -04001161 *
1162 * Only bother if changeover_active is set. Otherwise, this is
1163 * totally suboptimal to do on every SACK.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 */
Vlad Yasevichab5216a2008-06-19 18:17:24 -04001165 if (primary->cacc.changeover_active) {
1166 u8 clear_cycling = 0;
1167
1168 if (TSN_lte(primary->cacc.next_tsn_at_change, sack_ctsn)) {
1169 primary->cacc.changeover_active = 0;
1170 clear_cycling = 1;
1171 }
1172
1173 if (clear_cycling || gap_ack_blocks) {
1174 list_for_each_entry(transport, transport_list,
1175 transports) {
1176 if (clear_cycling)
1177 transport->cacc.cycling_changeover = 0;
1178 if (gap_ack_blocks)
1179 transport->cacc.cacc_saw_newack = 0;
1180 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 }
1182 }
1183
1184 /* Get the highest TSN in the sack. */
1185 highest_tsn = sack_ctsn;
Vlad Yasevich2cd9b822008-06-19 17:59:13 -04001186 if (gap_ack_blocks)
1187 highest_tsn += ntohs(frags[gap_ack_blocks - 1].gab.end);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188
Vlad Yasevichbfa0d982010-04-30 22:41:10 -04001189 if (TSN_lt(asoc->highest_sacked, highest_tsn))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190 asoc->highest_sacked = highest_tsn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191
Vlad Yasevichbfa0d982010-04-30 22:41:10 -04001192 highest_new_tsn = sack_ctsn;
Vlad Yasevich2cd9b822008-06-19 17:59:13 -04001193
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 /* Run through the retransmit queue. Credit bytes received
1195 * and free those chunks that we can.
1196 */
Nicolas Dichteledfee032012-10-03 05:43:22 +00001197 sctp_check_transmitted(q, &q->retransmit, NULL, NULL, sack, &highest_new_tsn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198
1199 /* Run through the transmitted queue.
1200 * Credit bytes received and free those chunks which we can.
1201 *
1202 * This is a MASSIVE candidate for optimization.
1203 */
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07001204 list_for_each_entry(transport, transport_list, transports) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 sctp_check_transmitted(q, &transport->transmitted,
Nicolas Dichteledfee032012-10-03 05:43:22 +00001206 transport, &chunk->source, sack,
1207 &highest_new_tsn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208 /*
1209 * SFR-CACC algorithm:
1210 * C) Let count_of_newacks be the number of
1211 * destinations for which cacc_saw_newack is set.
1212 */
1213 if (transport->cacc.cacc_saw_newack)
wangweidongcb3f8372013-12-23 12:16:50 +08001214 count_of_newacks++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 }
1216
Vlad Yasevichea862c82010-04-30 22:41:10 -04001217 /* Move the Cumulative TSN Ack Point if appropriate. */
1218 if (TSN_lt(asoc->ctsn_ack_point, sack_ctsn)) {
1219 asoc->ctsn_ack_point = sack_ctsn;
1220 accum_moved = 1;
1221 }
1222
Vlad Yasevich2cd9b822008-06-19 17:59:13 -04001223 if (gap_ack_blocks) {
Vlad Yasevichea862c82010-04-30 22:41:10 -04001224
1225 if (asoc->fast_recovery && accum_moved)
1226 highest_new_tsn = highest_tsn;
1227
Vlad Yasevich2cd9b822008-06-19 17:59:13 -04001228 list_for_each_entry(transport, transport_list, transports)
1229 sctp_mark_missing(q, &transport->transmitted, transport,
1230 highest_new_tsn, count_of_newacks);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 }
1232
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 /* Update unack_data field in the assoc. */
1234 sctp_sack_update_unack_data(asoc, sack);
1235
1236 ctsn = asoc->ctsn_ack_point;
1237
1238 /* Throw away stuff rotting on the sack queue. */
1239 list_for_each_safe(lchunk, temp, &q->sacked) {
1240 tchunk = list_entry(lchunk, struct sctp_chunk,
1241 transmitted_list);
1242 tsn = ntohl(tchunk->subh.data_hdr->tsn);
Vlad Yasevich5f9646c2008-02-05 14:23:44 -05001243 if (TSN_lte(tsn, ctsn)) {
1244 list_del_init(&tchunk->transmitted_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 sctp_chunk_free(tchunk);
Vlad Yasevich5f9646c2008-02-05 14:23:44 -05001246 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247 }
1248
1249 /* ii) Set rwnd equal to the newly received a_rwnd minus the
1250 * number of bytes still outstanding after processing the
1251 * Cumulative TSN Ack and the Gap Ack Blocks.
1252 */
1253
1254 sack_a_rwnd = ntohl(sack->a_rwnd);
1255 outstanding = q->outstanding_bytes;
1256
1257 if (outstanding < sack_a_rwnd)
1258 sack_a_rwnd -= outstanding;
1259 else
1260 sack_a_rwnd = 0;
1261
1262 asoc->peer.rwnd = sack_a_rwnd;
1263
1264 sctp_generate_fwdtsn(q, sack_ctsn);
1265
Daniel Borkmannbb333812013-06-28 19:49:40 +02001266 pr_debug("%s: sack cumulative tsn ack:0x%x\n", __func__, sack_ctsn);
1267 pr_debug("%s: cumulative tsn ack of assoc:%p is 0x%x, "
1268 "advertised peer ack point:0x%x\n", __func__, asoc, ctsn,
1269 asoc->adv_peer_ack_point);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270
Vlad Yasevich619a60e2014-01-02 14:39:44 -05001271 return sctp_outq_is_empty(q);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272}
1273
Vlad Yasevich619a60e2014-01-02 14:39:44 -05001274/* Is the outqueue empty?
1275 * The queue is empty when we have not pending data, no in-flight data
1276 * and nothing pending retransmissions.
1277 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278int sctp_outq_is_empty(const struct sctp_outq *q)
1279{
Vlad Yasevich619a60e2014-01-02 14:39:44 -05001280 return q->out_qlen == 0 && q->outstanding_bytes == 0 &&
1281 list_empty(&q->retransmit);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282}
1283
1284/********************************************************************
1285 * 2nd Level Abstractions
1286 ********************************************************************/
1287
1288/* Go through a transport's transmitted list or the association's retransmit
1289 * list and move chunks that are acked by the Cumulative TSN Ack to q->sacked.
1290 * The retransmit list will not have an associated transport.
1291 *
1292 * I added coherent debug information output. --xguo
1293 *
1294 * Instead of printing 'sacked' or 'kept' for each TSN on the
1295 * transmitted_queue, we print a range: SACKED: TSN1-TSN2, TSN3, TSN4-TSN5.
1296 * KEPT TSN6-TSN7, etc.
1297 */
1298static void sctp_check_transmitted(struct sctp_outq *q,
1299 struct list_head *transmitted_queue,
1300 struct sctp_transport *transport,
Nicolas Dichteledfee032012-10-03 05:43:22 +00001301 union sctp_addr *saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302 struct sctp_sackhdr *sack,
Vlad Yasevichbfa0d982010-04-30 22:41:10 -04001303 __u32 *highest_new_tsn_in_sack)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304{
1305 struct list_head *lchunk;
1306 struct sctp_chunk *tchunk;
1307 struct list_head tlist;
1308 __u32 tsn;
1309 __u32 sack_ctsn;
1310 __u32 rtt;
1311 __u8 restart_timer = 0;
1312 int bytes_acked = 0;
Vlad Yasevich31b02e12009-09-04 18:21:00 -04001313 int migrate_bytes = 0;
Daniel Borkmann8c2f4142013-07-09 16:17:04 +02001314 bool forward_progress = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 sack_ctsn = ntohl(sack->cum_tsn_ack);
1317
1318 INIT_LIST_HEAD(&tlist);
1319
1320 /* The while loop will skip empty transmitted queues. */
1321 while (NULL != (lchunk = sctp_list_dequeue(transmitted_queue))) {
1322 tchunk = list_entry(lchunk, struct sctp_chunk,
1323 transmitted_list);
1324
1325 if (sctp_chunk_abandoned(tchunk)) {
1326 /* Move the chunk to abandoned list. */
1327 sctp_insert_list(&q->abandoned, lchunk);
Vlad Yasevich8c4a2d42007-02-21 02:06:04 -08001328
1329 /* If this chunk has not been acked, stop
1330 * considering it as 'outstanding'.
1331 */
1332 if (!tchunk->tsn_gap_acked) {
Vlad Yasevich31b02e12009-09-04 18:21:00 -04001333 if (tchunk->transport)
1334 tchunk->transport->flight_size -=
1335 sctp_data_size(tchunk);
Vlad Yasevich8c4a2d42007-02-21 02:06:04 -08001336 q->outstanding_bytes -= sctp_data_size(tchunk);
1337 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 continue;
1339 }
1340
1341 tsn = ntohl(tchunk->subh.data_hdr->tsn);
1342 if (sctp_acked(sack, tsn)) {
1343 /* If this queue is the retransmit queue, the
1344 * retransmit timer has already reclaimed
1345 * the outstanding bytes for this chunk, so only
1346 * count bytes associated with a transport.
1347 */
1348 if (transport) {
1349 /* If this chunk is being used for RTT
1350 * measurement, calculate the RTT and update
1351 * the RTO using this value.
1352 *
1353 * 6.3.1 C5) Karn's algorithm: RTT measurements
1354 * MUST NOT be made using packets that were
1355 * retransmitted (and thus for which it is
1356 * ambiguous whether the reply was for the
1357 * first instance of the packet or a later
1358 * instance).
1359 */
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001360 if (!tchunk->tsn_gap_acked &&
Xufeng Zhang6eabca52013-11-25 11:26:57 +08001361 !tchunk->resent &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 tchunk->rtt_in_progress) {
Vlad Yasevich4c9f5d52006-06-17 22:56:08 -07001363 tchunk->rtt_in_progress = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 rtt = jiffies - tchunk->sent_at;
1365 sctp_transport_update_rto(transport,
1366 rtt);
1367 }
1368 }
Vlad Yasevich31b02e12009-09-04 18:21:00 -04001369
1370 /* If the chunk hasn't been marked as ACKED,
1371 * mark it and account bytes_acked if the
1372 * chunk had a valid transport (it will not
1373 * have a transport if ASCONF had deleted it
1374 * while DATA was outstanding).
1375 */
1376 if (!tchunk->tsn_gap_acked) {
1377 tchunk->tsn_gap_acked = 1;
Chang Xiangzhongd6c41612013-11-21 22:56:28 +01001378 if (TSN_lt(*highest_new_tsn_in_sack, tsn))
1379 *highest_new_tsn_in_sack = tsn;
Vlad Yasevich31b02e12009-09-04 18:21:00 -04001380 bytes_acked += sctp_data_size(tchunk);
1381 if (!tchunk->transport)
1382 migrate_bytes += sctp_data_size(tchunk);
Daniel Borkmann8c2f4142013-07-09 16:17:04 +02001383 forward_progress = true;
Vlad Yasevich31b02e12009-09-04 18:21:00 -04001384 }
1385
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001386 if (TSN_lte(tsn, sack_ctsn)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 /* RFC 2960 6.3.2 Retransmission Timer Rules
1388 *
1389 * R3) Whenever a SACK is received
1390 * that acknowledges the DATA chunk
1391 * with the earliest outstanding TSN
1392 * for that address, restart T3-rtx
1393 * timer for that address with its
1394 * current RTO.
1395 */
1396 restart_timer = 1;
Daniel Borkmann8c2f4142013-07-09 16:17:04 +02001397 forward_progress = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398
1399 if (!tchunk->tsn_gap_acked) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 /*
1401 * SFR-CACC algorithm:
1402 * 2) If the SACK contains gap acks
1403 * and the flag CHANGEOVER_ACTIVE is
1404 * set the receiver of the SACK MUST
1405 * take the following action:
1406 *
1407 * B) For each TSN t being acked that
1408 * has not been acked in any SACK so
1409 * far, set cacc_saw_newack to 1 for
1410 * the destination that the TSN was
1411 * sent to.
1412 */
1413 if (transport &&
1414 sack->num_gap_ack_blocks &&
1415 q->asoc->peer.primary_path->cacc.
1416 changeover_active)
1417 transport->cacc.cacc_saw_newack
1418 = 1;
1419 }
1420
1421 list_add_tail(&tchunk->transmitted_list,
1422 &q->sacked);
1423 } else {
1424 /* RFC2960 7.2.4, sctpimpguide-05 2.8.2
1425 * M2) Each time a SACK arrives reporting
1426 * 'Stray DATA chunk(s)' record the highest TSN
1427 * reported as newly acknowledged, call this
1428 * value 'HighestTSNinSack'. A newly
1429 * acknowledged DATA chunk is one not
1430 * previously acknowledged in a SACK.
1431 *
1432 * When the SCTP sender of data receives a SACK
1433 * chunk that acknowledges, for the first time,
1434 * the receipt of a DATA chunk, all the still
1435 * unacknowledged DATA chunks whose TSN is
1436 * older than that newly acknowledged DATA
1437 * chunk, are qualified as 'Stray DATA chunks'.
1438 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 list_add_tail(lchunk, &tlist);
1440 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 } else {
1442 if (tchunk->tsn_gap_acked) {
Daniel Borkmannbb333812013-06-28 19:49:40 +02001443 pr_debug("%s: receiver reneged on data TSN:0x%x\n",
1444 __func__, tsn);
1445
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 tchunk->tsn_gap_acked = 0;
1447
Vlad Yasevich31b02e12009-09-04 18:21:00 -04001448 if (tchunk->transport)
1449 bytes_acked -= sctp_data_size(tchunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450
1451 /* RFC 2960 6.3.2 Retransmission Timer Rules
1452 *
1453 * R4) Whenever a SACK is received missing a
1454 * TSN that was previously acknowledged via a
1455 * Gap Ack Block, start T3-rtx for the
1456 * destination address to which the DATA
1457 * chunk was originally
1458 * transmitted if it is not already running.
1459 */
1460 restart_timer = 1;
1461 }
1462
1463 list_add_tail(lchunk, &tlist);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 }
1465 }
1466
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 if (transport) {
1468 if (bytes_acked) {
Thomas Graff8d96052011-07-07 00:28:35 +00001469 struct sctp_association *asoc = transport->asoc;
1470
Vlad Yasevich31b02e12009-09-04 18:21:00 -04001471 /* We may have counted DATA that was migrated
1472 * to this transport due to DEL-IP operation.
1473 * Subtract those bytes, since the were never
1474 * send on this transport and shouldn't be
1475 * credited to this transport.
1476 */
1477 bytes_acked -= migrate_bytes;
1478
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 /* 8.2. When an outstanding TSN is acknowledged,
1480 * the endpoint shall clear the error counter of
1481 * the destination transport address to which the
1482 * DATA chunk was last sent.
1483 * The association's overall error counter is
1484 * also cleared.
1485 */
1486 transport->error_count = 0;
1487 transport->asoc->overall_error_count = 0;
Daniel Borkmann8c2f4142013-07-09 16:17:04 +02001488 forward_progress = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489
Thomas Graff8d96052011-07-07 00:28:35 +00001490 /*
1491 * While in SHUTDOWN PENDING, we may have started
1492 * the T5 shutdown guard timer after reaching the
1493 * retransmission limit. Stop that timer as soon
1494 * as the receiver acknowledged any data.
1495 */
1496 if (asoc->state == SCTP_STATE_SHUTDOWN_PENDING &&
1497 del_timer(&asoc->timers
1498 [SCTP_EVENT_TIMEOUT_T5_SHUTDOWN_GUARD]))
1499 sctp_association_put(asoc);
1500
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 /* Mark the destination transport address as
1502 * active if it is not so marked.
1503 */
Nicolas Dichteledfee032012-10-03 05:43:22 +00001504 if ((transport->state == SCTP_INACTIVE ||
1505 transport->state == SCTP_UNCONFIRMED) &&
1506 sctp_cmp_addr_exact(&transport->ipaddr, saddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 sctp_assoc_control_transport(
1508 transport->asoc,
1509 transport,
1510 SCTP_TRANSPORT_UP,
1511 SCTP_RECEIVED_SACK);
1512 }
1513
1514 sctp_transport_raise_cwnd(transport, sack_ctsn,
1515 bytes_acked);
1516
1517 transport->flight_size -= bytes_acked;
Gui Jianfeng8b73a072008-04-17 14:22:18 -07001518 if (transport->flight_size == 0)
1519 transport->partial_bytes_acked = 0;
Vlad Yasevich31b02e12009-09-04 18:21:00 -04001520 q->outstanding_bytes -= bytes_acked + migrate_bytes;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 } else {
1522 /* RFC 2960 6.1, sctpimpguide-06 2.15.2
1523 * When a sender is doing zero window probing, it
1524 * should not timeout the association if it continues
1525 * to receive new packets from the receiver. The
1526 * reason is that the receiver MAY keep its window
1527 * closed for an indefinite time.
1528 * A sender is doing zero window probing when the
1529 * receiver's advertised window is zero, and there is
1530 * only one data chunk in flight to the receiver.
Thomas Graff8d96052011-07-07 00:28:35 +00001531 *
1532 * Allow the association to timeout while in SHUTDOWN
1533 * PENDING or SHUTDOWN RECEIVED in case the receiver
1534 * stays in zero window mode forever.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 */
1536 if (!q->asoc->peer.rwnd &&
1537 !list_empty(&tlist) &&
Thomas Graff8d96052011-07-07 00:28:35 +00001538 (sack_ctsn+2 == q->asoc->next_tsn) &&
1539 q->asoc->state < SCTP_STATE_SHUTDOWN_PENDING) {
Daniel Borkmannbb333812013-06-28 19:49:40 +02001540 pr_debug("%s: sack received for zero window "
1541 "probe:%u\n", __func__, sack_ctsn);
1542
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 q->asoc->overall_error_count = 0;
1544 transport->error_count = 0;
1545 }
1546 }
1547
1548 /* RFC 2960 6.3.2 Retransmission Timer Rules
1549 *
1550 * R2) Whenever all outstanding data sent to an address have
1551 * been acknowledged, turn off the T3-rtx timer of that
1552 * address.
1553 */
1554 if (!transport->flight_size) {
Ying Xue25cc4ae2013-02-03 20:32:57 +00001555 if (del_timer(&transport->T3_rtx_timer))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 sctp_transport_put(transport);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 } else if (restart_timer) {
1558 if (!mod_timer(&transport->T3_rtx_timer,
1559 jiffies + transport->rto))
1560 sctp_transport_hold(transport);
1561 }
Daniel Borkmann8c2f4142013-07-09 16:17:04 +02001562
1563 if (forward_progress) {
1564 if (transport->dst)
1565 dst_confirm(transport->dst);
1566 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 }
1568
1569 list_splice(&tlist, transmitted_queue);
1570}
1571
1572/* Mark chunks as missing and consequently may get retransmitted. */
1573static void sctp_mark_missing(struct sctp_outq *q,
1574 struct list_head *transmitted_queue,
1575 struct sctp_transport *transport,
1576 __u32 highest_new_tsn_in_sack,
1577 int count_of_newacks)
1578{
1579 struct sctp_chunk *chunk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 __u32 tsn;
1581 char do_fast_retransmit = 0;
Vlad Yasevichea862c82010-04-30 22:41:10 -04001582 struct sctp_association *asoc = q->asoc;
1583 struct sctp_transport *primary = asoc->peer.primary_path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584
Robert P. J. Day9dbc15f2008-04-12 18:54:24 -07001585 list_for_each_entry(chunk, transmitted_queue, transmitted_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 tsn = ntohl(chunk->subh.data_hdr->tsn);
1588
1589 /* RFC 2960 7.2.4, sctpimpguide-05 2.8.2 M3) Examine all
1590 * 'Unacknowledged TSN's', if the TSN number of an
1591 * 'Unacknowledged TSN' is smaller than the 'HighestTSNinSack'
1592 * value, increment the 'TSN.Missing.Report' count on that
1593 * chunk if it has NOT been fast retransmitted or marked for
1594 * fast retransmit already.
1595 */
Neil Hormanc226ef92008-07-25 12:44:09 -04001596 if (chunk->fast_retransmit == SCTP_CAN_FRTX &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 !chunk->tsn_gap_acked &&
1598 TSN_lt(tsn, highest_new_tsn_in_sack)) {
1599
1600 /* SFR-CACC may require us to skip marking
1601 * this chunk as missing.
1602 */
Vlad Yasevichf246a7b2011-04-18 19:13:56 +00001603 if (!transport || !sctp_cacc_skip(primary,
1604 chunk->transport,
1605 count_of_newacks, tsn)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 chunk->tsn_missing_report++;
1607
Daniel Borkmannbb333812013-06-28 19:49:40 +02001608 pr_debug("%s: tsn:0x%x missing counter:%d\n",
1609 __func__, tsn, chunk->tsn_missing_report);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610 }
1611 }
1612 /*
1613 * M4) If any DATA chunk is found to have a
1614 * 'TSN.Missing.Report'
Vlad Yasevich27852c22006-02-02 16:57:31 -08001615 * value larger than or equal to 3, mark that chunk for
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 * retransmission and start the fast retransmit procedure.
1617 */
1618
Vlad Yasevich27852c22006-02-02 16:57:31 -08001619 if (chunk->tsn_missing_report >= 3) {
Neil Hormanc226ef92008-07-25 12:44:09 -04001620 chunk->fast_retransmit = SCTP_NEED_FRTX;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 do_fast_retransmit = 1;
1622 }
1623 }
1624
1625 if (transport) {
1626 if (do_fast_retransmit)
1627 sctp_retransmit(q, transport, SCTP_RTXR_FAST_RTX);
1628
Daniel Borkmannbb333812013-06-28 19:49:40 +02001629 pr_debug("%s: transport:%p, cwnd:%d, ssthresh:%d, "
1630 "flight_size:%d, pba:%d\n", __func__, transport,
1631 transport->cwnd, transport->ssthresh,
1632 transport->flight_size, transport->partial_bytes_acked);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633 }
1634}
1635
1636/* Is the given TSN acked by this packet? */
1637static int sctp_acked(struct sctp_sackhdr *sack, __u32 tsn)
1638{
1639 int i;
1640 sctp_sack_variable_t *frags;
1641 __u16 gap;
1642 __u32 ctsn = ntohl(sack->cum_tsn_ack);
1643
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001644 if (TSN_lte(tsn, ctsn))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 goto pass;
1646
1647 /* 3.3.4 Selective Acknowledgement (SACK) (3):
1648 *
1649 * Gap Ack Blocks:
1650 * These fields contain the Gap Ack Blocks. They are repeated
1651 * for each Gap Ack Block up to the number of Gap Ack Blocks
1652 * defined in the Number of Gap Ack Blocks field. All DATA
1653 * chunks with TSNs greater than or equal to (Cumulative TSN
1654 * Ack + Gap Ack Block Start) and less than or equal to
1655 * (Cumulative TSN Ack + Gap Ack Block End) of each Gap Ack
1656 * Block are assumed to have been received correctly.
1657 */
1658
1659 frags = sack->variable;
1660 gap = tsn - ctsn;
1661 for (i = 0; i < ntohs(sack->num_gap_ack_blocks); ++i) {
1662 if (TSN_lte(ntohs(frags[i].gab.start), gap) &&
1663 TSN_lte(gap, ntohs(frags[i].gab.end)))
1664 goto pass;
1665 }
1666
1667 return 0;
1668pass:
1669 return 1;
1670}
1671
1672static inline int sctp_get_skip_pos(struct sctp_fwdtsn_skip *skiplist,
Al Viro9f81bcd2006-11-20 17:26:34 -08001673 int nskips, __be16 stream)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674{
1675 int i;
1676
1677 for (i = 0; i < nskips; i++) {
1678 if (skiplist[i].stream == stream)
1679 return i;
1680 }
1681 return i;
1682}
1683
1684/* Create and add a fwdtsn chunk to the outq's control queue if needed. */
1685static void sctp_generate_fwdtsn(struct sctp_outq *q, __u32 ctsn)
1686{
1687 struct sctp_association *asoc = q->asoc;
1688 struct sctp_chunk *ftsn_chunk = NULL;
1689 struct sctp_fwdtsn_skip ftsn_skip_arr[10];
1690 int nskips = 0;
1691 int skip_pos = 0;
1692 __u32 tsn;
1693 struct sctp_chunk *chunk;
1694 struct list_head *lchunk, *temp;
1695
Wei Yongjun76595022009-03-12 09:49:19 +00001696 if (!asoc->peer.prsctp_capable)
1697 return;
1698
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 /* PR-SCTP C1) Let SackCumAck be the Cumulative TSN ACK carried in the
1700 * received SACK.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001701 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 * If (Advanced.Peer.Ack.Point < SackCumAck), then update
1703 * Advanced.Peer.Ack.Point to be equal to SackCumAck.
1704 */
1705 if (TSN_lt(asoc->adv_peer_ack_point, ctsn))
1706 asoc->adv_peer_ack_point = ctsn;
1707
1708 /* PR-SCTP C2) Try to further advance the "Advanced.Peer.Ack.Point"
1709 * locally, that is, to move "Advanced.Peer.Ack.Point" up as long as
1710 * the chunk next in the out-queue space is marked as "abandoned" as
1711 * shown in the following example:
1712 *
1713 * Assuming that a SACK arrived with the Cumulative TSN ACK 102
1714 * and the Advanced.Peer.Ack.Point is updated to this value:
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001715 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 * out-queue at the end of ==> out-queue after Adv.Ack.Point
1717 * normal SACK processing local advancement
1718 * ... ...
1719 * Adv.Ack.Pt-> 102 acked 102 acked
1720 * 103 abandoned 103 abandoned
1721 * 104 abandoned Adv.Ack.P-> 104 abandoned
1722 * 105 105
1723 * 106 acked 106 acked
1724 * ... ...
1725 *
1726 * In this example, the data sender successfully advanced the
1727 * "Advanced.Peer.Ack.Point" from 102 to 104 locally.
1728 */
1729 list_for_each_safe(lchunk, temp, &q->abandoned) {
1730 chunk = list_entry(lchunk, struct sctp_chunk,
1731 transmitted_list);
1732 tsn = ntohl(chunk->subh.data_hdr->tsn);
1733
1734 /* Remove any chunks in the abandoned queue that are acked by
1735 * the ctsn.
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001736 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737 if (TSN_lte(tsn, ctsn)) {
1738 list_del_init(lchunk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 sctp_chunk_free(chunk);
1740 } else {
1741 if (TSN_lte(tsn, asoc->adv_peer_ack_point+1)) {
1742 asoc->adv_peer_ack_point = tsn;
1743 if (chunk->chunk_hdr->flags &
1744 SCTP_DATA_UNORDERED)
1745 continue;
1746 skip_pos = sctp_get_skip_pos(&ftsn_skip_arr[0],
1747 nskips,
1748 chunk->subh.data_hdr->stream);
1749 ftsn_skip_arr[skip_pos].stream =
1750 chunk->subh.data_hdr->stream;
1751 ftsn_skip_arr[skip_pos].ssn =
1752 chunk->subh.data_hdr->ssn;
1753 if (skip_pos == nskips)
1754 nskips++;
1755 if (nskips == 10)
1756 break;
1757 } else
1758 break;
1759 }
1760 }
1761
1762 /* PR-SCTP C3) If, after step C1 and C2, the "Advanced.Peer.Ack.Point"
1763 * is greater than the Cumulative TSN ACK carried in the received
1764 * SACK, the data sender MUST send the data receiver a FORWARD TSN
1765 * chunk containing the latest value of the
1766 * "Advanced.Peer.Ack.Point".
1767 *
1768 * C4) For each "abandoned" TSN the sender of the FORWARD TSN SHOULD
1769 * list each stream and sequence number in the forwarded TSN. This
1770 * information will enable the receiver to easily find any
1771 * stranded TSN's waiting on stream reorder queues. Each stream
1772 * SHOULD only be reported once; this means that if multiple
1773 * abandoned messages occur in the same stream then only the
1774 * highest abandoned stream sequence number is reported. If the
1775 * total size of the FORWARD TSN does NOT fit in a single MTU then
1776 * the sender of the FORWARD TSN SHOULD lower the
1777 * Advanced.Peer.Ack.Point to the last TSN that will fit in a
1778 * single MTU.
1779 */
1780 if (asoc->adv_peer_ack_point > ctsn)
1781 ftsn_chunk = sctp_make_fwdtsn(asoc, asoc->adv_peer_ack_point,
YOSHIFUJI Hideakid808ad92007-02-09 23:25:18 +09001782 nskips, &ftsn_skip_arr[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783
1784 if (ftsn_chunk) {
David S. Miller79af02c2005-07-08 21:47:49 -07001785 list_add_tail(&ftsn_chunk->list, &q->control_chunk_list);
Eric W. Biedermanb01a2402012-08-06 08:47:55 +00001786 SCTP_INC_STATS(sock_net(asoc->base.sk), SCTP_MIB_OUTCTRLCHUNKS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787 }
1788}