]>
Commit | Line | Data |
---|---|---|
60c778b2 | 1 | /* SCTP kernel implementation |
1da177e4 LT |
2 | * Copyright (c) 1999-2000 Cisco, Inc. |
3 | * Copyright (c) 1999-2001 Motorola, Inc. | |
4 | * Copyright (c) 2001-2003 International Business Machines Corp. | |
5 | * Copyright (c) 2001 Intel Corp. | |
6 | * Copyright (c) 2001 La Monte H.P. Yarroll | |
7 | * | |
60c778b2 | 8 | * This file is part of the SCTP kernel implementation |
1da177e4 LT |
9 | * |
10 | * This module provides the abstraction for an SCTP tranport representing | |
11 | * a remote transport address. For local transport addresses, we just use | |
12 | * union sctp_addr. | |
13 | * | |
60c778b2 | 14 | * This SCTP implementation is free software; |
1da177e4 LT |
15 | * you can redistribute it and/or modify it under the terms of |
16 | * the GNU General Public License as published by | |
17 | * the Free Software Foundation; either version 2, or (at your option) | |
18 | * any later version. | |
19 | * | |
60c778b2 | 20 | * This SCTP implementation is distributed in the hope that it |
1da177e4 LT |
21 | * will be useful, but WITHOUT ANY WARRANTY; without even the implied |
22 | * ************************ | |
23 | * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. | |
24 | * See the GNU General Public License for more details. | |
25 | * | |
26 | * You should have received a copy of the GNU General Public License | |
4b2f13a2 JK |
27 | * along with GNU CC; see the file COPYING. If not, see |
28 | * <http://www.gnu.org/licenses/>. | |
1da177e4 LT |
29 | * |
30 | * Please send any bug reports or fixes you make to the | |
31 | * email address(es): | |
91705c61 | 32 | * lksctp developers <linux-sctp@vger.kernel.org> |
1da177e4 | 33 | * |
1da177e4 LT |
34 | * Written or modified by: |
35 | * La Monte H.P. Yarroll <piggy@acm.org> | |
36 | * Karl Knutson <karl@athena.chicago.il.us> | |
37 | * Jon Grimm <jgrimm@us.ibm.com> | |
38 | * Xingang Guo <xingang.guo@intel.com> | |
39 | * Hui Huang <hui.huang@nokia.com> | |
40 | * Sridhar Samudrala <sri@us.ibm.com> | |
41 | * Ardelle Fan <ardelle.fan@intel.com> | |
1da177e4 LT |
42 | */ |
43 | ||
145ce502 JP |
44 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
45 | ||
5a0e3ad6 | 46 | #include <linux/slab.h> |
1da177e4 | 47 | #include <linux/types.h> |
ad8fec17 | 48 | #include <linux/random.h> |
1da177e4 LT |
49 | #include <net/sctp/sctp.h> |
50 | #include <net/sctp/sm.h> | |
51 | ||
52 | /* 1st Level Abstractions. */ | |
53 | ||
54 | /* Initialize a new transport from provided memory. */ | |
89bf3450 EB |
55 | static struct sctp_transport *sctp_transport_init(struct net *net, |
56 | struct sctp_transport *peer, | |
1da177e4 | 57 | const union sctp_addr *addr, |
dd0fc66f | 58 | gfp_t gfp) |
1da177e4 LT |
59 | { |
60 | /* Copy in the address. */ | |
61 | peer->ipaddr = *addr; | |
62 | peer->af_specific = sctp_get_af_specific(addr->sa.sa_family); | |
1da177e4 LT |
63 | memset(&peer->saddr, 0, sizeof(union sctp_addr)); |
64 | ||
4244854d NH |
65 | peer->sack_generation = 0; |
66 | ||
1da177e4 LT |
67 | /* From 6.3.1 RTO Calculation: |
68 | * | |
69 | * C1) Until an RTT measurement has been made for a packet sent to the | |
70 | * given destination transport address, set RTO to the protocol | |
71 | * parameter 'RTO.Initial'. | |
72 | */ | |
e1fc3b14 | 73 | peer->rto = msecs_to_jiffies(net->sctp.rto_initial); |
1da177e4 | 74 | |
8b0e1953 | 75 | peer->last_time_heard = 0; |
1da177e4 LT |
76 | peer->last_time_ecne_reduced = jiffies; |
77 | ||
52ccb8e9 FF |
78 | peer->param_flags = SPP_HB_DISABLE | |
79 | SPP_PMTUD_ENABLE | | |
80 | SPP_SACKDELAY_ENABLE; | |
1da177e4 LT |
81 | |
82 | /* Initialize the default path max_retrans. */ | |
e1fc3b14 EB |
83 | peer->pathmaxrxt = net->sctp.max_retrans_path; |
84 | peer->pf_retrans = net->sctp.pf_retrans; | |
1da177e4 LT |
85 | |
86 | INIT_LIST_HEAD(&peer->transmitted); | |
87 | INIT_LIST_HEAD(&peer->send_ready); | |
88 | INIT_LIST_HEAD(&peer->transports); | |
89 | ||
b24b8a24 | 90 | setup_timer(&peer->T3_rtx_timer, sctp_generate_t3_rtx_event, |
7b9438de | 91 | (unsigned long)peer); |
b24b8a24 | 92 | setup_timer(&peer->hb_timer, sctp_generate_heartbeat_event, |
7b9438de XL |
93 | (unsigned long)peer); |
94 | setup_timer(&peer->reconf_timer, sctp_generate_reconf_event, | |
95 | (unsigned long)peer); | |
50b5d6ad VY |
96 | setup_timer(&peer->proto_unreach_timer, |
97 | sctp_generate_proto_unreach_event, (unsigned long)peer); | |
1da177e4 | 98 | |
ad8fec17 SS |
99 | /* Initialize the 64-bit random nonce sent with heartbeat. */ |
100 | get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce)); | |
101 | ||
1da177e4 | 102 | atomic_set(&peer->refcnt, 1); |
1da177e4 LT |
103 | |
104 | return peer; | |
105 | } | |
106 | ||
107 | /* Allocate and initialize a new transport. */ | |
89bf3450 EB |
108 | struct sctp_transport *sctp_transport_new(struct net *net, |
109 | const union sctp_addr *addr, | |
dd0fc66f | 110 | gfp_t gfp) |
1da177e4 | 111 | { |
d808ad9a | 112 | struct sctp_transport *transport; |
1da177e4 | 113 | |
939cfa75 | 114 | transport = kzalloc(sizeof(*transport), gfp); |
1da177e4 LT |
115 | if (!transport) |
116 | goto fail; | |
117 | ||
89bf3450 | 118 | if (!sctp_transport_init(net, transport, addr, gfp)) |
1da177e4 LT |
119 | goto fail_init; |
120 | ||
1da177e4 LT |
121 | SCTP_DBG_OBJCNT_INC(transport); |
122 | ||
123 | return transport; | |
124 | ||
125 | fail_init: | |
126 | kfree(transport); | |
127 | ||
128 | fail: | |
129 | return NULL; | |
130 | } | |
131 | ||
132 | /* This transport is no longer needed. Free up if possible, or | |
133 | * delay until it last reference count. | |
134 | */ | |
135 | void sctp_transport_free(struct sctp_transport *transport) | |
136 | { | |
1da177e4 LT |
137 | /* Try to delete the heartbeat timer. */ |
138 | if (del_timer(&transport->hb_timer)) | |
139 | sctp_transport_put(transport); | |
140 | ||
141 | /* Delete the T3_rtx timer if it's active. | |
142 | * There is no point in not doing this now and letting | |
143 | * structure hang around in memory since we know | |
144 | * the tranport is going away. | |
145 | */ | |
25cc4ae9 | 146 | if (del_timer(&transport->T3_rtx_timer)) |
1da177e4 LT |
147 | sctp_transport_put(transport); |
148 | ||
7b9438de XL |
149 | if (del_timer(&transport->reconf_timer)) |
150 | sctp_transport_put(transport); | |
151 | ||
55fa0cfd | 152 | /* Delete the ICMP proto unreachable timer if it's active. */ |
25cc4ae9 | 153 | if (del_timer(&transport->proto_unreach_timer)) |
55fa0cfd | 154 | sctp_association_put(transport->asoc); |
1da177e4 LT |
155 | |
156 | sctp_transport_put(transport); | |
157 | } | |
158 | ||
45122ca2 | 159 | static void sctp_transport_destroy_rcu(struct rcu_head *head) |
1da177e4 | 160 | { |
45122ca2 | 161 | struct sctp_transport *transport; |
1da177e4 | 162 | |
45122ca2 | 163 | transport = container_of(head, struct sctp_transport, rcu); |
1da177e4 LT |
164 | |
165 | dst_release(transport->dst); | |
166 | kfree(transport); | |
167 | SCTP_DBG_OBJCNT_DEC(transport); | |
168 | } | |
169 | ||
45122ca2 TG |
170 | /* Destroy the transport data structure. |
171 | * Assumes there are no more users of this structure. | |
172 | */ | |
173 | static void sctp_transport_destroy(struct sctp_transport *transport) | |
174 | { | |
47faa1e4 | 175 | if (unlikely(atomic_read(&transport->refcnt))) { |
bb33381d DB |
176 | WARN(1, "Attempt to destroy undead transport %p!\n", transport); |
177 | return; | |
178 | } | |
45122ca2 | 179 | |
8c98653f DB |
180 | sctp_packet_free(&transport->packet); |
181 | ||
182 | if (transport->asoc) | |
183 | sctp_association_put(transport->asoc); | |
771085d6 DB |
184 | |
185 | call_rcu(&transport->rcu, sctp_transport_destroy_rcu); | |
45122ca2 TG |
186 | } |
187 | ||
1da177e4 LT |
188 | /* Start T3_rtx timer if it is not already running and update the heartbeat |
189 | * timer. This routine is called every time a DATA chunk is sent. | |
190 | */ | |
ba6f5e33 | 191 | void sctp_transport_reset_t3_rtx(struct sctp_transport *transport) |
1da177e4 LT |
192 | { |
193 | /* RFC 2960 6.3.2 Retransmission Timer Rules | |
194 | * | |
195 | * R1) Every time a DATA chunk is sent to any address(including a | |
196 | * retransmission), if the T3-rtx timer of that address is not running | |
197 | * start it running so that it will expire after the RTO of that | |
198 | * address. | |
199 | */ | |
200 | ||
d9efc223 | 201 | if (!timer_pending(&transport->T3_rtx_timer)) |
1da177e4 LT |
202 | if (!mod_timer(&transport->T3_rtx_timer, |
203 | jiffies + transport->rto)) | |
204 | sctp_transport_hold(transport); | |
ba6f5e33 MRL |
205 | } |
206 | ||
207 | void sctp_transport_reset_hb_timer(struct sctp_transport *transport) | |
208 | { | |
209 | unsigned long expires; | |
1da177e4 LT |
210 | |
211 | /* When a data chunk is sent, reset the heartbeat interval. */ | |
ba6f5e33 MRL |
212 | expires = jiffies + sctp_transport_timeout(transport); |
213 | if (time_before(transport->hb_timer.expires, expires) && | |
214 | !mod_timer(&transport->hb_timer, | |
215 | expires + prandom_u32_max(transport->rto))) | |
216 | sctp_transport_hold(transport); | |
1da177e4 LT |
217 | } |
218 | ||
7b9438de XL |
219 | void sctp_transport_reset_reconf_timer(struct sctp_transport *transport) |
220 | { | |
221 | if (!timer_pending(&transport->reconf_timer)) | |
222 | if (!mod_timer(&transport->reconf_timer, | |
223 | jiffies + transport->rto)) | |
224 | sctp_transport_hold(transport); | |
225 | } | |
226 | ||
1da177e4 LT |
227 | /* This transport has been assigned to an association. |
228 | * Initialize fields from the association or from the sock itself. | |
229 | * Register the reference count in the association. | |
230 | */ | |
231 | void sctp_transport_set_owner(struct sctp_transport *transport, | |
232 | struct sctp_association *asoc) | |
233 | { | |
234 | transport->asoc = asoc; | |
235 | sctp_association_hold(asoc); | |
236 | } | |
237 | ||
238 | /* Initialize the pmtu of a transport. */ | |
9914ae3c | 239 | void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk) |
1da177e4 | 240 | { |
da0420be | 241 | /* If we don't have a fresh route, look one up */ |
f5b0a874 | 242 | if (!transport->dst || transport->dst->obsolete) { |
c86a773c | 243 | sctp_transport_dst_release(transport); |
da0420be | 244 | transport->af_specific->get_dst(transport, &transport->saddr, |
8663c938 | 245 | &transport->fl, sk); |
da0420be | 246 | } |
1da177e4 | 247 | |
da0420be | 248 | if (transport->dst) { |
e2f036a9 | 249 | transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst)); |
1da177e4 | 250 | } else |
52ccb8e9 | 251 | transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT; |
1da177e4 LT |
252 | } |
253 | ||
3ebfdf08 | 254 | void sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu) |
c910b47e | 255 | { |
3ebfdf08 | 256 | struct dst_entry *dst = sctp_transport_dst_check(t); |
c910b47e VY |
257 | |
258 | if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) { | |
145ce502 | 259 | pr_warn("%s: Reported pmtu %d too low, using default minimum of %d\n", |
3ebfdf08 | 260 | __func__, pmtu, SCTP_DEFAULT_MINSEGMENT); |
c910b47e VY |
261 | /* Use default minimum segment size and disable |
262 | * pmtu discovery on this transport. | |
263 | */ | |
264 | t->pathmtu = SCTP_DEFAULT_MINSEGMENT; | |
c910b47e VY |
265 | } else { |
266 | t->pathmtu = pmtu; | |
267 | } | |
268 | ||
02f3d4ce | 269 | if (dst) { |
3ebfdf08 | 270 | dst->ops->update_pmtu(dst, t->asoc->base.sk, NULL, pmtu); |
02f3d4ce | 271 | dst = sctp_transport_dst_check(t); |
02f3d4ce | 272 | } |
3ebfdf08 XL |
273 | |
274 | if (!dst) | |
275 | t->af_specific->get_dst(t, &t->saddr, &t->fl, t->asoc->base.sk); | |
c910b47e VY |
276 | } |
277 | ||
1da177e4 LT |
278 | /* Caches the dst entry and source address for a transport's destination |
279 | * address. | |
280 | */ | |
281 | void sctp_transport_route(struct sctp_transport *transport, | |
282 | union sctp_addr *saddr, struct sctp_sock *opt) | |
283 | { | |
284 | struct sctp_association *asoc = transport->asoc; | |
285 | struct sctp_af *af = transport->af_specific; | |
1da177e4 | 286 | |
8663c938 | 287 | af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt)); |
1da177e4 LT |
288 | |
289 | if (saddr) | |
290 | memcpy(&transport->saddr, saddr, sizeof(union sctp_addr)); | |
291 | else | |
8663c938 | 292 | af->get_saddr(opt, transport, &transport->fl); |
1da177e4 | 293 | |
52ccb8e9 FF |
294 | if ((transport->param_flags & SPP_PMTUD_DISABLE) && transport->pathmtu) { |
295 | return; | |
296 | } | |
da0420be | 297 | if (transport->dst) { |
e2f036a9 | 298 | transport->pathmtu = SCTP_TRUNC4(dst_mtu(transport->dst)); |
1da177e4 LT |
299 | |
300 | /* Initialize sk->sk_rcv_saddr, if the transport is the | |
301 | * association's active path for getsockname(). | |
d808ad9a | 302 | */ |
a78102e7 VY |
303 | if (asoc && (!asoc->peer.primary_path || |
304 | (transport == asoc->peer.active_path))) | |
299ee123 JG |
305 | opt->pf->to_sk_saddr(&transport->saddr, |
306 | asoc->base.sk); | |
1da177e4 | 307 | } else |
52ccb8e9 | 308 | transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT; |
1da177e4 LT |
309 | } |
310 | ||
311 | /* Hold a reference to a transport. */ | |
1eed6779 | 312 | int sctp_transport_hold(struct sctp_transport *transport) |
1da177e4 | 313 | { |
1eed6779 | 314 | return atomic_add_unless(&transport->refcnt, 1, 0); |
1da177e4 LT |
315 | } |
316 | ||
317 | /* Release a reference to a transport and clean up | |
318 | * if there are no more references. | |
319 | */ | |
320 | void sctp_transport_put(struct sctp_transport *transport) | |
321 | { | |
322 | if (atomic_dec_and_test(&transport->refcnt)) | |
323 | sctp_transport_destroy(transport); | |
324 | } | |
325 | ||
326 | /* Update transport's RTO based on the newly calculated RTT. */ | |
327 | void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt) | |
328 | { | |
bb33381d DB |
329 | if (unlikely(!tp->rto_pending)) |
330 | /* We should not be doing any RTO updates unless rto_pending is set. */ | |
331 | pr_debug("%s: rto_pending not set on transport %p!\n", __func__, tp); | |
1da177e4 LT |
332 | |
333 | if (tp->rttvar || tp->srtt) { | |
e1fc3b14 | 334 | struct net *net = sock_net(tp->asoc->base.sk); |
1da177e4 LT |
335 | /* 6.3.1 C3) When a new RTT measurement R' is made, set |
336 | * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'| | |
337 | * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R' | |
338 | */ | |
339 | ||
340 | /* Note: The above algorithm has been rewritten to | |
341 | * express rto_beta and rto_alpha as inverse powers | |
342 | * of two. | |
343 | * For example, assuming the default value of RTO.Alpha of | |
344 | * 1/8, rto_alpha would be expressed as 3. | |
345 | */ | |
e1fc3b14 | 346 | tp->rttvar = tp->rttvar - (tp->rttvar >> net->sctp.rto_beta) |
79211c8e | 347 | + (((__u32)abs((__s64)tp->srtt - (__s64)rtt)) >> net->sctp.rto_beta); |
e1fc3b14 EB |
348 | tp->srtt = tp->srtt - (tp->srtt >> net->sctp.rto_alpha) |
349 | + (rtt >> net->sctp.rto_alpha); | |
1da177e4 LT |
350 | } else { |
351 | /* 6.3.1 C2) When the first RTT measurement R is made, set | |
352 | * SRTT <- R, RTTVAR <- R/2. | |
353 | */ | |
354 | tp->srtt = rtt; | |
355 | tp->rttvar = rtt >> 1; | |
356 | } | |
357 | ||
358 | /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then | |
359 | * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY. | |
360 | */ | |
361 | if (tp->rttvar == 0) | |
362 | tp->rttvar = SCTP_CLOCK_GRANULARITY; | |
363 | ||
364 | /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */ | |
365 | tp->rto = tp->srtt + (tp->rttvar << 2); | |
366 | ||
367 | /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min | |
368 | * seconds then it is rounded up to RTO.Min seconds. | |
369 | */ | |
370 | if (tp->rto < tp->asoc->rto_min) | |
371 | tp->rto = tp->asoc->rto_min; | |
372 | ||
373 | /* 6.3.1 C7) A maximum value may be placed on RTO provided it is | |
374 | * at least RTO.max seconds. | |
375 | */ | |
376 | if (tp->rto > tp->asoc->rto_max) | |
377 | tp->rto = tp->asoc->rto_max; | |
378 | ||
196d6759 | 379 | sctp_max_rto(tp->asoc, tp); |
1da177e4 LT |
380 | tp->rtt = rtt; |
381 | ||
382 | /* Reset rto_pending so that a new RTT measurement is started when a | |
383 | * new data chunk is sent. | |
384 | */ | |
385 | tp->rto_pending = 0; | |
386 | ||
bb33381d DB |
387 | pr_debug("%s: transport:%p, rtt:%d, srtt:%d rttvar:%d, rto:%ld\n", |
388 | __func__, tp, rtt, tp->srtt, tp->rttvar, tp->rto); | |
1da177e4 LT |
389 | } |
390 | ||
391 | /* This routine updates the transport's cwnd and partial_bytes_acked | |
392 | * parameters based on the bytes acked in the received SACK. | |
393 | */ | |
394 | void sctp_transport_raise_cwnd(struct sctp_transport *transport, | |
395 | __u32 sack_ctsn, __u32 bytes_acked) | |
396 | { | |
cf9b4812 | 397 | struct sctp_association *asoc = transport->asoc; |
1da177e4 LT |
398 | __u32 cwnd, ssthresh, flight_size, pba, pmtu; |
399 | ||
400 | cwnd = transport->cwnd; | |
401 | flight_size = transport->flight_size; | |
402 | ||
a6465234 | 403 | /* See if we need to exit Fast Recovery first */ |
cf9b4812 VY |
404 | if (asoc->fast_recovery && |
405 | TSN_lte(asoc->fast_recovery_exit, sack_ctsn)) | |
406 | asoc->fast_recovery = 0; | |
a6465234 | 407 | |
1da177e4 | 408 | /* The appropriate cwnd increase algorithm is performed if, and only |
a6465234 | 409 | * if the cumulative TSN whould advanced and the congestion window is |
1da177e4 LT |
410 | * being fully utilized. |
411 | */ | |
a6465234 | 412 | if (TSN_lte(sack_ctsn, transport->asoc->ctsn_ack_point) || |
1da177e4 LT |
413 | (flight_size < cwnd)) |
414 | return; | |
415 | ||
416 | ssthresh = transport->ssthresh; | |
417 | pba = transport->partial_bytes_acked; | |
52ccb8e9 | 418 | pmtu = transport->asoc->pathmtu; |
1da177e4 LT |
419 | |
420 | if (cwnd <= ssthresh) { | |
a6465234 VY |
421 | /* RFC 4960 7.2.1 |
422 | * o When cwnd is less than or equal to ssthresh, an SCTP | |
423 | * endpoint MUST use the slow-start algorithm to increase | |
424 | * cwnd only if the current congestion window is being fully | |
425 | * utilized, an incoming SACK advances the Cumulative TSN | |
426 | * Ack Point, and the data sender is not in Fast Recovery. | |
427 | * Only when these three conditions are met can the cwnd be | |
428 | * increased; otherwise, the cwnd MUST not be increased. | |
429 | * If these conditions are met, then cwnd MUST be increased | |
430 | * by, at most, the lesser of 1) the total size of the | |
431 | * previously outstanding DATA chunk(s) acknowledged, and | |
432 | * 2) the destination's path MTU. This upper bound protects | |
433 | * against the ACK-Splitting attack outlined in [SAVAGE99]. | |
1da177e4 | 434 | */ |
cf9b4812 | 435 | if (asoc->fast_recovery) |
a6465234 VY |
436 | return; |
437 | ||
1da177e4 LT |
438 | if (bytes_acked > pmtu) |
439 | cwnd += pmtu; | |
440 | else | |
441 | cwnd += bytes_acked; | |
bb33381d DB |
442 | |
443 | pr_debug("%s: slow start: transport:%p, bytes_acked:%d, " | |
444 | "cwnd:%d, ssthresh:%d, flight_size:%d, pba:%d\n", | |
445 | __func__, transport, bytes_acked, cwnd, ssthresh, | |
446 | flight_size, pba); | |
1da177e4 LT |
447 | } else { |
448 | /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh, | |
449 | * upon each SACK arrival that advances the Cumulative TSN Ack | |
450 | * Point, increase partial_bytes_acked by the total number of | |
451 | * bytes of all new chunks acknowledged in that SACK including | |
452 | * chunks acknowledged by the new Cumulative TSN Ack and by | |
453 | * Gap Ack Blocks. | |
454 | * | |
455 | * When partial_bytes_acked is equal to or greater than cwnd | |
456 | * and before the arrival of the SACK the sender had cwnd or | |
457 | * more bytes of data outstanding (i.e., before arrival of the | |
458 | * SACK, flightsize was greater than or equal to cwnd), | |
459 | * increase cwnd by MTU, and reset partial_bytes_acked to | |
460 | * (partial_bytes_acked - cwnd). | |
461 | */ | |
462 | pba += bytes_acked; | |
463 | if (pba >= cwnd) { | |
464 | cwnd += pmtu; | |
465 | pba = ((cwnd < pba) ? (pba - cwnd) : 0); | |
466 | } | |
bb33381d DB |
467 | |
468 | pr_debug("%s: congestion avoidance: transport:%p, " | |
469 | "bytes_acked:%d, cwnd:%d, ssthresh:%d, " | |
470 | "flight_size:%d, pba:%d\n", __func__, | |
471 | transport, bytes_acked, cwnd, ssthresh, | |
472 | flight_size, pba); | |
1da177e4 LT |
473 | } |
474 | ||
475 | transport->cwnd = cwnd; | |
476 | transport->partial_bytes_acked = pba; | |
477 | } | |
478 | ||
479 | /* This routine is used to lower the transport's cwnd when congestion is | |
480 | * detected. | |
481 | */ | |
482 | void sctp_transport_lower_cwnd(struct sctp_transport *transport, | |
483 | sctp_lower_cwnd_t reason) | |
484 | { | |
cf9b4812 VY |
485 | struct sctp_association *asoc = transport->asoc; |
486 | ||
1da177e4 LT |
487 | switch (reason) { |
488 | case SCTP_LOWER_CWND_T3_RTX: | |
489 | /* RFC 2960 Section 7.2.3, sctpimpguide | |
490 | * When the T3-rtx timer expires on an address, SCTP should | |
491 | * perform slow start by: | |
492 | * ssthresh = max(cwnd/2, 4*MTU) | |
493 | * cwnd = 1*MTU | |
494 | * partial_bytes_acked = 0 | |
495 | */ | |
496 | transport->ssthresh = max(transport->cwnd/2, | |
cf9b4812 VY |
497 | 4*asoc->pathmtu); |
498 | transport->cwnd = asoc->pathmtu; | |
33ce8281 | 499 | |
cf9b4812 VY |
500 | /* T3-rtx also clears fast recovery */ |
501 | asoc->fast_recovery = 0; | |
1da177e4 LT |
502 | break; |
503 | ||
504 | case SCTP_LOWER_CWND_FAST_RTX: | |
505 | /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the | |
506 | * destination address(es) to which the missing DATA chunks | |
507 | * were last sent, according to the formula described in | |
508 | * Section 7.2.3. | |
d808ad9a YH |
509 | * |
510 | * RFC 2960 7.2.3, sctpimpguide Upon detection of packet | |
1da177e4 LT |
511 | * losses from SACK (see Section 7.2.4), An endpoint |
512 | * should do the following: | |
513 | * ssthresh = max(cwnd/2, 4*MTU) | |
514 | * cwnd = ssthresh | |
515 | * partial_bytes_acked = 0 | |
516 | */ | |
cf9b4812 | 517 | if (asoc->fast_recovery) |
a6465234 VY |
518 | return; |
519 | ||
520 | /* Mark Fast recovery */ | |
cf9b4812 VY |
521 | asoc->fast_recovery = 1; |
522 | asoc->fast_recovery_exit = asoc->next_tsn - 1; | |
a6465234 | 523 | |
1da177e4 | 524 | transport->ssthresh = max(transport->cwnd/2, |
cf9b4812 | 525 | 4*asoc->pathmtu); |
1da177e4 LT |
526 | transport->cwnd = transport->ssthresh; |
527 | break; | |
528 | ||
529 | case SCTP_LOWER_CWND_ECNE: | |
530 | /* RFC 2481 Section 6.1.2. | |
531 | * If the sender receives an ECN-Echo ACK packet | |
532 | * then the sender knows that congestion was encountered in the | |
533 | * network on the path from the sender to the receiver. The | |
534 | * indication of congestion should be treated just as a | |
535 | * congestion loss in non-ECN Capable TCP. That is, the TCP | |
536 | * source halves the congestion window "cwnd" and reduces the | |
537 | * slow start threshold "ssthresh". | |
538 | * A critical condition is that TCP does not react to | |
539 | * congestion indications more than once every window of | |
540 | * data (or more loosely more than once every round-trip time). | |
541 | */ | |
f61f6f82 WY |
542 | if (time_after(jiffies, transport->last_time_ecne_reduced + |
543 | transport->rtt)) { | |
1da177e4 | 544 | transport->ssthresh = max(transport->cwnd/2, |
cf9b4812 | 545 | 4*asoc->pathmtu); |
1da177e4 LT |
546 | transport->cwnd = transport->ssthresh; |
547 | transport->last_time_ecne_reduced = jiffies; | |
548 | } | |
549 | break; | |
550 | ||
551 | case SCTP_LOWER_CWND_INACTIVE: | |
552 | /* RFC 2960 Section 7.2.1, sctpimpguide | |
553 | * When the endpoint does not transmit data on a given | |
554 | * transport address, the cwnd of the transport address | |
555 | * should be adjusted to max(cwnd/2, 4*MTU) per RTO. | |
556 | * NOTE: Although the draft recommends that this check needs | |
557 | * to be done every RTO interval, we do it every hearbeat | |
558 | * interval. | |
559 | */ | |
245cba7e | 560 | transport->cwnd = max(transport->cwnd/2, |
cf9b4812 | 561 | 4*asoc->pathmtu); |
1da177e4 | 562 | break; |
3ff50b79 | 563 | } |
1da177e4 LT |
564 | |
565 | transport->partial_bytes_acked = 0; | |
bb33381d DB |
566 | |
567 | pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d\n", | |
568 | __func__, transport, reason, transport->cwnd, | |
569 | transport->ssthresh); | |
1da177e4 LT |
570 | } |
571 | ||
46d5a808 VY |
572 | /* Apply Max.Burst limit to the congestion window: |
573 | * sctpimpguide-05 2.14.2 | |
574 | * D) When the time comes for the sender to | |
575 | * transmit new DATA chunks, the protocol parameter Max.Burst MUST | |
576 | * first be applied to limit how many new DATA chunks may be sent. | |
577 | * The limit is applied by adjusting cwnd as follows: | |
578 | * if ((flightsize+ Max.Burst * MTU) < cwnd) | |
579 | * cwnd = flightsize + Max.Burst * MTU | |
580 | */ | |
581 | ||
582 | void sctp_transport_burst_limited(struct sctp_transport *t) | |
583 | { | |
584 | struct sctp_association *asoc = t->asoc; | |
585 | u32 old_cwnd = t->cwnd; | |
586 | u32 max_burst_bytes; | |
587 | ||
78ac814f | 588 | if (t->burst_limited || asoc->max_burst == 0) |
46d5a808 VY |
589 | return; |
590 | ||
591 | max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu); | |
592 | if (max_burst_bytes < old_cwnd) { | |
593 | t->cwnd = max_burst_bytes; | |
594 | t->burst_limited = old_cwnd; | |
595 | } | |
596 | } | |
597 | ||
598 | /* Restore the old cwnd congestion window, after the burst had it's | |
599 | * desired effect. | |
600 | */ | |
601 | void sctp_transport_burst_reset(struct sctp_transport *t) | |
602 | { | |
603 | if (t->burst_limited) { | |
604 | t->cwnd = t->burst_limited; | |
605 | t->burst_limited = 0; | |
606 | } | |
607 | } | |
608 | ||
1da177e4 | 609 | /* What is the next timeout value for this transport? */ |
8f61059a | 610 | unsigned long sctp_transport_timeout(struct sctp_transport *trans) |
1da177e4 | 611 | { |
8f61059a | 612 | /* RTO + timer slack +/- 50% of RTO */ |
ba6f5e33 | 613 | unsigned long timeout = trans->rto >> 1; |
8f61059a DB |
614 | |
615 | if (trans->state != SCTP_UNCONFIRMED && | |
616 | trans->state != SCTP_PF) | |
617 | timeout += trans->hbinterval; | |
618 | ||
ba6f5e33 | 619 | return timeout; |
1da177e4 | 620 | } |
749bf921 VY |
621 | |
622 | /* Reset transport variables to their initial values */ | |
623 | void sctp_transport_reset(struct sctp_transport *t) | |
624 | { | |
625 | struct sctp_association *asoc = t->asoc; | |
626 | ||
627 | /* RFC 2960 (bis), Section 5.2.4 | |
628 | * All the congestion control parameters (e.g., cwnd, ssthresh) | |
629 | * related to this peer MUST be reset to their initial values | |
630 | * (see Section 6.2.1) | |
631 | */ | |
632 | t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380)); | |
46d5a808 | 633 | t->burst_limited = 0; |
289f4249 | 634 | t->ssthresh = asoc->peer.i.a_rwnd; |
5fdd4bae | 635 | t->rto = asoc->rto_initial; |
196d6759 | 636 | sctp_max_rto(asoc, t); |
749bf921 VY |
637 | t->rtt = 0; |
638 | t->srtt = 0; | |
639 | t->rttvar = 0; | |
640 | ||
b564d62e | 641 | /* Reset these additional variables so that we have a clean slate. */ |
749bf921 VY |
642 | t->partial_bytes_acked = 0; |
643 | t->flight_size = 0; | |
644 | t->error_count = 0; | |
645 | t->rto_pending = 0; | |
faee47cd | 646 | t->hb_sent = 0; |
749bf921 VY |
647 | |
648 | /* Initialize the state information for SFR-CACC */ | |
649 | t->cacc.changeover_active = 0; | |
650 | t->cacc.cycling_changeover = 0; | |
651 | t->cacc.next_tsn_at_change = 0; | |
652 | t->cacc.cacc_saw_newack = 0; | |
653 | } | |
ddc4bbee MH |
654 | |
655 | /* Schedule retransmission on the given transport */ | |
656 | void sctp_transport_immediate_rtx(struct sctp_transport *t) | |
657 | { | |
658 | /* Stop pending T3_rtx_timer */ | |
25cc4ae9 | 659 | if (del_timer(&t->T3_rtx_timer)) |
ddc4bbee | 660 | sctp_transport_put(t); |
25cc4ae9 | 661 | |
ddc4bbee MH |
662 | sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX); |
663 | if (!timer_pending(&t->T3_rtx_timer)) { | |
664 | if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto)) | |
665 | sctp_transport_hold(t); | |
666 | } | |
ddc4bbee | 667 | } |
c86a773c JA |
668 | |
669 | /* Drop dst */ | |
670 | void sctp_transport_dst_release(struct sctp_transport *t) | |
671 | { | |
672 | dst_release(t->dst); | |
673 | t->dst = NULL; | |
674 | t->dst_pending_confirm = 0; | |
675 | } | |
676 | ||
677 | /* Schedule neighbour confirm */ | |
678 | void sctp_transport_dst_confirm(struct sctp_transport *t) | |
679 | { | |
680 | t->dst_pending_confirm = 1; | |
681 | } |