Commit | Line | Data |
---|---|---|
47505b8b | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
60c778b2 | 2 | /* SCTP kernel implementation |
1da177e4 LT |
3 | * Copyright (c) 1999-2000 Cisco, Inc. |
4 | * Copyright (c) 1999-2001 Motorola, Inc. | |
5 | * Copyright (c) 2001-2003 International Business Machines Corp. | |
6 | * Copyright (c) 2001 Intel Corp. | |
7 | * Copyright (c) 2001 La Monte H.P. Yarroll | |
8 | * | |
60c778b2 | 9 | * This file is part of the SCTP kernel implementation |
1da177e4 LT |
10 | * |
11 | * This module provides the abstraction for an SCTP tranport representing | |
12 | * a remote transport address. For local transport addresses, we just use | |
13 | * union sctp_addr. | |
14 | * | |
1da177e4 LT |
15 | * Please send any bug reports or fixes you make to the |
16 | * email address(es): | |
91705c61 | 17 | * lksctp developers <linux-sctp@vger.kernel.org> |
1da177e4 | 18 | * |
1da177e4 LT |
19 | * Written or modified by: |
20 | * La Monte H.P. Yarroll <piggy@acm.org> | |
21 | * Karl Knutson <karl@athena.chicago.il.us> | |
22 | * Jon Grimm <jgrimm@us.ibm.com> | |
23 | * Xingang Guo <xingang.guo@intel.com> | |
24 | * Hui Huang <hui.huang@nokia.com> | |
25 | * Sridhar Samudrala <sri@us.ibm.com> | |
26 | * Ardelle Fan <ardelle.fan@intel.com> | |
1da177e4 LT |
27 | */ |
28 | ||
145ce502 JP |
29 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
30 | ||
5a0e3ad6 | 31 | #include <linux/slab.h> |
1da177e4 | 32 | #include <linux/types.h> |
ad8fec17 | 33 | #include <linux/random.h> |
1da177e4 LT |
34 | #include <net/sctp/sctp.h> |
35 | #include <net/sctp/sm.h> | |
36 | ||
37 | /* 1st Level Abstractions. */ | |
38 | ||
39 | /* Initialize a new transport from provided memory. */ | |
89bf3450 EB |
40 | static struct sctp_transport *sctp_transport_init(struct net *net, |
41 | struct sctp_transport *peer, | |
1da177e4 | 42 | const union sctp_addr *addr, |
dd0fc66f | 43 | gfp_t gfp) |
1da177e4 LT |
44 | { |
45 | /* Copy in the address. */ | |
1da177e4 | 46 | peer->af_specific = sctp_get_af_specific(addr->sa.sa_family); |
4c31bc6b | 47 | memcpy(&peer->ipaddr, addr, peer->af_specific->sockaddr_len); |
1da177e4 LT |
48 | memset(&peer->saddr, 0, sizeof(union sctp_addr)); |
49 | ||
4244854d NH |
50 | peer->sack_generation = 0; |
51 | ||
1da177e4 LT |
52 | /* From 6.3.1 RTO Calculation: |
53 | * | |
54 | * C1) Until an RTT measurement has been made for a packet sent to the | |
55 | * given destination transport address, set RTO to the protocol | |
56 | * parameter 'RTO.Initial'. | |
57 | */ | |
e1fc3b14 | 58 | peer->rto = msecs_to_jiffies(net->sctp.rto_initial); |
1da177e4 | 59 | |
8b0e1953 | 60 | peer->last_time_heard = 0; |
1da177e4 LT |
61 | peer->last_time_ecne_reduced = jiffies; |
62 | ||
52ccb8e9 FF |
63 | peer->param_flags = SPP_HB_DISABLE | |
64 | SPP_PMTUD_ENABLE | | |
65 | SPP_SACKDELAY_ENABLE; | |
1da177e4 LT |
66 | |
67 | /* Initialize the default path max_retrans. */ | |
e1fc3b14 EB |
68 | peer->pathmaxrxt = net->sctp.max_retrans_path; |
69 | peer->pf_retrans = net->sctp.pf_retrans; | |
1da177e4 LT |
70 | |
71 | INIT_LIST_HEAD(&peer->transmitted); | |
72 | INIT_LIST_HEAD(&peer->send_ready); | |
73 | INIT_LIST_HEAD(&peer->transports); | |
74 | ||
9c3b5751 KC |
75 | timer_setup(&peer->T3_rtx_timer, sctp_generate_t3_rtx_event, 0); |
76 | timer_setup(&peer->hb_timer, sctp_generate_heartbeat_event, 0); | |
77 | timer_setup(&peer->reconf_timer, sctp_generate_reconf_event, 0); | |
78 | timer_setup(&peer->proto_unreach_timer, | |
79 | sctp_generate_proto_unreach_event, 0); | |
1da177e4 | 80 | |
ad8fec17 SS |
81 | /* Initialize the 64-bit random nonce sent with heartbeat. */ |
82 | get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce)); | |
83 | ||
a4b2b58e | 84 | refcount_set(&peer->refcnt, 1); |
1da177e4 LT |
85 | |
86 | return peer; | |
87 | } | |
88 | ||
89 | /* Allocate and initialize a new transport. */ | |
89bf3450 EB |
90 | struct sctp_transport *sctp_transport_new(struct net *net, |
91 | const union sctp_addr *addr, | |
dd0fc66f | 92 | gfp_t gfp) |
1da177e4 | 93 | { |
d808ad9a | 94 | struct sctp_transport *transport; |
1da177e4 | 95 | |
939cfa75 | 96 | transport = kzalloc(sizeof(*transport), gfp); |
1da177e4 LT |
97 | if (!transport) |
98 | goto fail; | |
99 | ||
89bf3450 | 100 | if (!sctp_transport_init(net, transport, addr, gfp)) |
1da177e4 LT |
101 | goto fail_init; |
102 | ||
1da177e4 LT |
103 | SCTP_DBG_OBJCNT_INC(transport); |
104 | ||
105 | return transport; | |
106 | ||
107 | fail_init: | |
108 | kfree(transport); | |
109 | ||
110 | fail: | |
111 | return NULL; | |
112 | } | |
113 | ||
114 | /* This transport is no longer needed. Free up if possible, or | |
115 | * delay until it last reference count. | |
116 | */ | |
117 | void sctp_transport_free(struct sctp_transport *transport) | |
118 | { | |
1da177e4 LT |
119 | /* Try to delete the heartbeat timer. */ |
120 | if (del_timer(&transport->hb_timer)) | |
121 | sctp_transport_put(transport); | |
122 | ||
123 | /* Delete the T3_rtx timer if it's active. | |
124 | * There is no point in not doing this now and letting | |
125 | * structure hang around in memory since we know | |
126 | * the tranport is going away. | |
127 | */ | |
25cc4ae9 | 128 | if (del_timer(&transport->T3_rtx_timer)) |
1da177e4 LT |
129 | sctp_transport_put(transport); |
130 | ||
7b9438de XL |
131 | if (del_timer(&transport->reconf_timer)) |
132 | sctp_transport_put(transport); | |
133 | ||
55fa0cfd | 134 | /* Delete the ICMP proto unreachable timer if it's active. */ |
25cc4ae9 | 135 | if (del_timer(&transport->proto_unreach_timer)) |
55fa0cfd | 136 | sctp_association_put(transport->asoc); |
1da177e4 LT |
137 | |
138 | sctp_transport_put(transport); | |
139 | } | |
140 | ||
45122ca2 | 141 | static void sctp_transport_destroy_rcu(struct rcu_head *head) |
1da177e4 | 142 | { |
45122ca2 | 143 | struct sctp_transport *transport; |
1da177e4 | 144 | |
45122ca2 | 145 | transport = container_of(head, struct sctp_transport, rcu); |
1da177e4 LT |
146 | |
147 | dst_release(transport->dst); | |
148 | kfree(transport); | |
149 | SCTP_DBG_OBJCNT_DEC(transport); | |
150 | } | |
151 | ||
45122ca2 TG |
152 | /* Destroy the transport data structure. |
153 | * Assumes there are no more users of this structure. | |
154 | */ | |
155 | static void sctp_transport_destroy(struct sctp_transport *transport) | |
156 | { | |
a4b2b58e | 157 | if (unlikely(refcount_read(&transport->refcnt))) { |
bb33381d DB |
158 | WARN(1, "Attempt to destroy undead transport %p!\n", transport); |
159 | return; | |
160 | } | |
45122ca2 | 161 | |
8c98653f DB |
162 | sctp_packet_free(&transport->packet); |
163 | ||
164 | if (transport->asoc) | |
165 | sctp_association_put(transport->asoc); | |
771085d6 DB |
166 | |
167 | call_rcu(&transport->rcu, sctp_transport_destroy_rcu); | |
45122ca2 TG |
168 | } |
169 | ||
1da177e4 LT |
170 | /* Start T3_rtx timer if it is not already running and update the heartbeat |
171 | * timer. This routine is called every time a DATA chunk is sent. | |
172 | */ | |
ba6f5e33 | 173 | void sctp_transport_reset_t3_rtx(struct sctp_transport *transport) |
1da177e4 LT |
174 | { |
175 | /* RFC 2960 6.3.2 Retransmission Timer Rules | |
176 | * | |
177 | * R1) Every time a DATA chunk is sent to any address(including a | |
178 | * retransmission), if the T3-rtx timer of that address is not running | |
179 | * start it running so that it will expire after the RTO of that | |
180 | * address. | |
181 | */ | |
182 | ||
d9efc223 | 183 | if (!timer_pending(&transport->T3_rtx_timer)) |
1da177e4 LT |
184 | if (!mod_timer(&transport->T3_rtx_timer, |
185 | jiffies + transport->rto)) | |
186 | sctp_transport_hold(transport); | |
ba6f5e33 MRL |
187 | } |
188 | ||
189 | void sctp_transport_reset_hb_timer(struct sctp_transport *transport) | |
190 | { | |
191 | unsigned long expires; | |
1da177e4 LT |
192 | |
193 | /* When a data chunk is sent, reset the heartbeat interval. */ | |
ba6f5e33 | 194 | expires = jiffies + sctp_transport_timeout(transport); |
d1f20c03 MK |
195 | if ((time_before(transport->hb_timer.expires, expires) || |
196 | !timer_pending(&transport->hb_timer)) && | |
ba6f5e33 MRL |
197 | !mod_timer(&transport->hb_timer, |
198 | expires + prandom_u32_max(transport->rto))) | |
199 | sctp_transport_hold(transport); | |
1da177e4 LT |
200 | } |
201 | ||
7b9438de XL |
202 | void sctp_transport_reset_reconf_timer(struct sctp_transport *transport) |
203 | { | |
204 | if (!timer_pending(&transport->reconf_timer)) | |
205 | if (!mod_timer(&transport->reconf_timer, | |
206 | jiffies + transport->rto)) | |
207 | sctp_transport_hold(transport); | |
208 | } | |
209 | ||
1da177e4 LT |
210 | /* This transport has been assigned to an association. |
211 | * Initialize fields from the association or from the sock itself. | |
212 | * Register the reference count in the association. | |
213 | */ | |
214 | void sctp_transport_set_owner(struct sctp_transport *transport, | |
215 | struct sctp_association *asoc) | |
216 | { | |
217 | transport->asoc = asoc; | |
218 | sctp_association_hold(asoc); | |
219 | } | |
220 | ||
221 | /* Initialize the pmtu of a transport. */ | |
9914ae3c | 222 | void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk) |
1da177e4 | 223 | { |
da0420be | 224 | /* If we don't have a fresh route, look one up */ |
f5b0a874 | 225 | if (!transport->dst || transport->dst->obsolete) { |
c86a773c | 226 | sctp_transport_dst_release(transport); |
da0420be | 227 | transport->af_specific->get_dst(transport, &transport->saddr, |
8663c938 | 228 | &transport->fl, sk); |
da0420be | 229 | } |
1da177e4 | 230 | |
6e91b578 MRL |
231 | if (transport->param_flags & SPP_PMTUD_DISABLE) { |
232 | struct sctp_association *asoc = transport->asoc; | |
233 | ||
234 | if (!transport->pathmtu && asoc && asoc->pathmtu) | |
235 | transport->pathmtu = asoc->pathmtu; | |
236 | if (transport->pathmtu) | |
237 | return; | |
238 | } | |
239 | ||
6ff0f871 MRL |
240 | if (transport->dst) |
241 | transport->pathmtu = sctp_dst_mtu(transport->dst); | |
242 | else | |
52ccb8e9 | 243 | transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT; |
1da177e4 LT |
244 | } |
245 | ||
b6c5734d | 246 | bool sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu) |
c910b47e | 247 | { |
3ebfdf08 | 248 | struct dst_entry *dst = sctp_transport_dst_check(t); |
d7ab5cdc | 249 | struct sock *sk = t->asoc->base.sk; |
b6c5734d | 250 | bool change = true; |
c910b47e VY |
251 | |
252 | if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) { | |
b6c5734d MRL |
253 | pr_warn_ratelimited("%s: Reported pmtu %d too low, using default minimum of %d\n", |
254 | __func__, pmtu, SCTP_DEFAULT_MINSEGMENT); | |
255 | /* Use default minimum segment instead */ | |
256 | pmtu = SCTP_DEFAULT_MINSEGMENT; | |
c910b47e | 257 | } |
b6c5734d | 258 | pmtu = SCTP_TRUNC4(pmtu); |
c910b47e | 259 | |
02f3d4ce | 260 | if (dst) { |
d7ab5cdc XL |
261 | struct sctp_pf *pf = sctp_get_pf_specific(dst->ops->family); |
262 | union sctp_addr addr; | |
263 | ||
264 | pf->af->from_sk(&addr, sk); | |
265 | pf->to_sk_daddr(&t->ipaddr, sk); | |
266 | dst->ops->update_pmtu(dst, sk, NULL, pmtu); | |
267 | pf->to_sk_daddr(&addr, sk); | |
268 | ||
02f3d4ce | 269 | dst = sctp_transport_dst_check(t); |
02f3d4ce | 270 | } |
3ebfdf08 | 271 | |
b6c5734d | 272 | if (!dst) { |
d7ab5cdc | 273 | t->af_specific->get_dst(t, &t->saddr, &t->fl, sk); |
b6c5734d MRL |
274 | dst = t->dst; |
275 | } | |
276 | ||
277 | if (dst) { | |
278 | /* Re-fetch, as under layers may have a higher minimum size */ | |
a6592547 | 279 | pmtu = sctp_dst_mtu(dst); |
b6c5734d MRL |
280 | change = t->pathmtu != pmtu; |
281 | } | |
282 | t->pathmtu = pmtu; | |
283 | ||
284 | return change; | |
c910b47e VY |
285 | } |
286 | ||
1da177e4 LT |
287 | /* Caches the dst entry and source address for a transport's destination |
288 | * address. | |
289 | */ | |
290 | void sctp_transport_route(struct sctp_transport *transport, | |
291 | union sctp_addr *saddr, struct sctp_sock *opt) | |
292 | { | |
293 | struct sctp_association *asoc = transport->asoc; | |
294 | struct sctp_af *af = transport->af_specific; | |
1da177e4 | 295 | |
6e91b578 | 296 | sctp_transport_dst_release(transport); |
8663c938 | 297 | af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt)); |
1da177e4 LT |
298 | |
299 | if (saddr) | |
300 | memcpy(&transport->saddr, saddr, sizeof(union sctp_addr)); | |
301 | else | |
8663c938 | 302 | af->get_saddr(opt, transport, &transport->fl); |
1da177e4 | 303 | |
6e91b578 MRL |
304 | sctp_transport_pmtu(transport, sctp_opt2sk(opt)); |
305 | ||
306 | /* Initialize sk->sk_rcv_saddr, if the transport is the | |
307 | * association's active path for getsockname(). | |
308 | */ | |
309 | if (transport->dst && asoc && | |
310 | (!asoc->peer.primary_path || transport == asoc->peer.active_path)) | |
311 | opt->pf->to_sk_saddr(&transport->saddr, asoc->base.sk); | |
1da177e4 LT |
312 | } |
313 | ||
314 | /* Hold a reference to a transport. */ | |
1eed6779 | 315 | int sctp_transport_hold(struct sctp_transport *transport) |
1da177e4 | 316 | { |
a4b2b58e | 317 | return refcount_inc_not_zero(&transport->refcnt); |
1da177e4 LT |
318 | } |
319 | ||
320 | /* Release a reference to a transport and clean up | |
321 | * if there are no more references. | |
322 | */ | |
323 | void sctp_transport_put(struct sctp_transport *transport) | |
324 | { | |
a4b2b58e | 325 | if (refcount_dec_and_test(&transport->refcnt)) |
1da177e4 LT |
326 | sctp_transport_destroy(transport); |
327 | } | |
328 | ||
329 | /* Update transport's RTO based on the newly calculated RTT. */ | |
330 | void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt) | |
331 | { | |
bb33381d DB |
332 | if (unlikely(!tp->rto_pending)) |
333 | /* We should not be doing any RTO updates unless rto_pending is set. */ | |
334 | pr_debug("%s: rto_pending not set on transport %p!\n", __func__, tp); | |
1da177e4 LT |
335 | |
336 | if (tp->rttvar || tp->srtt) { | |
e1fc3b14 | 337 | struct net *net = sock_net(tp->asoc->base.sk); |
1da177e4 LT |
338 | /* 6.3.1 C3) When a new RTT measurement R' is made, set |
339 | * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'| | |
340 | * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R' | |
341 | */ | |
342 | ||
343 | /* Note: The above algorithm has been rewritten to | |
344 | * express rto_beta and rto_alpha as inverse powers | |
345 | * of two. | |
346 | * For example, assuming the default value of RTO.Alpha of | |
347 | * 1/8, rto_alpha would be expressed as 3. | |
348 | */ | |
e1fc3b14 | 349 | tp->rttvar = tp->rttvar - (tp->rttvar >> net->sctp.rto_beta) |
79211c8e | 350 | + (((__u32)abs((__s64)tp->srtt - (__s64)rtt)) >> net->sctp.rto_beta); |
e1fc3b14 EB |
351 | tp->srtt = tp->srtt - (tp->srtt >> net->sctp.rto_alpha) |
352 | + (rtt >> net->sctp.rto_alpha); | |
1da177e4 LT |
353 | } else { |
354 | /* 6.3.1 C2) When the first RTT measurement R is made, set | |
355 | * SRTT <- R, RTTVAR <- R/2. | |
356 | */ | |
357 | tp->srtt = rtt; | |
358 | tp->rttvar = rtt >> 1; | |
359 | } | |
360 | ||
361 | /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then | |
362 | * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY. | |
363 | */ | |
364 | if (tp->rttvar == 0) | |
365 | tp->rttvar = SCTP_CLOCK_GRANULARITY; | |
366 | ||
367 | /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */ | |
368 | tp->rto = tp->srtt + (tp->rttvar << 2); | |
369 | ||
370 | /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min | |
371 | * seconds then it is rounded up to RTO.Min seconds. | |
372 | */ | |
373 | if (tp->rto < tp->asoc->rto_min) | |
374 | tp->rto = tp->asoc->rto_min; | |
375 | ||
376 | /* 6.3.1 C7) A maximum value may be placed on RTO provided it is | |
377 | * at least RTO.max seconds. | |
378 | */ | |
379 | if (tp->rto > tp->asoc->rto_max) | |
380 | tp->rto = tp->asoc->rto_max; | |
381 | ||
196d6759 | 382 | sctp_max_rto(tp->asoc, tp); |
1da177e4 LT |
383 | tp->rtt = rtt; |
384 | ||
385 | /* Reset rto_pending so that a new RTT measurement is started when a | |
386 | * new data chunk is sent. | |
387 | */ | |
388 | tp->rto_pending = 0; | |
389 | ||
bb33381d DB |
390 | pr_debug("%s: transport:%p, rtt:%d, srtt:%d rttvar:%d, rto:%ld\n", |
391 | __func__, tp, rtt, tp->srtt, tp->rttvar, tp->rto); | |
1da177e4 LT |
392 | } |
393 | ||
394 | /* This routine updates the transport's cwnd and partial_bytes_acked | |
395 | * parameters based on the bytes acked in the received SACK. | |
396 | */ | |
397 | void sctp_transport_raise_cwnd(struct sctp_transport *transport, | |
398 | __u32 sack_ctsn, __u32 bytes_acked) | |
399 | { | |
cf9b4812 | 400 | struct sctp_association *asoc = transport->asoc; |
1da177e4 LT |
401 | __u32 cwnd, ssthresh, flight_size, pba, pmtu; |
402 | ||
403 | cwnd = transport->cwnd; | |
404 | flight_size = transport->flight_size; | |
405 | ||
a6465234 | 406 | /* See if we need to exit Fast Recovery first */ |
cf9b4812 VY |
407 | if (asoc->fast_recovery && |
408 | TSN_lte(asoc->fast_recovery_exit, sack_ctsn)) | |
409 | asoc->fast_recovery = 0; | |
a6465234 | 410 | |
1da177e4 LT |
411 | ssthresh = transport->ssthresh; |
412 | pba = transport->partial_bytes_acked; | |
52ccb8e9 | 413 | pmtu = transport->asoc->pathmtu; |
1da177e4 LT |
414 | |
415 | if (cwnd <= ssthresh) { | |
a6465234 VY |
416 | /* RFC 4960 7.2.1 |
417 | * o When cwnd is less than or equal to ssthresh, an SCTP | |
418 | * endpoint MUST use the slow-start algorithm to increase | |
419 | * cwnd only if the current congestion window is being fully | |
420 | * utilized, an incoming SACK advances the Cumulative TSN | |
421 | * Ack Point, and the data sender is not in Fast Recovery. | |
422 | * Only when these three conditions are met can the cwnd be | |
423 | * increased; otherwise, the cwnd MUST not be increased. | |
424 | * If these conditions are met, then cwnd MUST be increased | |
425 | * by, at most, the lesser of 1) the total size of the | |
426 | * previously outstanding DATA chunk(s) acknowledged, and | |
427 | * 2) the destination's path MTU. This upper bound protects | |
428 | * against the ACK-Splitting attack outlined in [SAVAGE99]. | |
1da177e4 | 429 | */ |
cf9b4812 | 430 | if (asoc->fast_recovery) |
a6465234 VY |
431 | return; |
432 | ||
4ccbd0b0 MRL |
433 | /* The appropriate cwnd increase algorithm is performed |
434 | * if, and only if the congestion window is being fully | |
435 | * utilized. Note that RFC4960 Errata 3.22 removed the | |
436 | * other condition on ctsn moving. | |
437 | */ | |
438 | if (flight_size < cwnd) | |
439 | return; | |
440 | ||
1da177e4 LT |
441 | if (bytes_acked > pmtu) |
442 | cwnd += pmtu; | |
443 | else | |
444 | cwnd += bytes_acked; | |
bb33381d DB |
445 | |
446 | pr_debug("%s: slow start: transport:%p, bytes_acked:%d, " | |
447 | "cwnd:%d, ssthresh:%d, flight_size:%d, pba:%d\n", | |
448 | __func__, transport, bytes_acked, cwnd, ssthresh, | |
449 | flight_size, pba); | |
1da177e4 LT |
450 | } else { |
451 | /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh, | |
e56f777a MRL |
452 | * upon each SACK arrival, increase partial_bytes_acked |
453 | * by the total number of bytes of all new chunks | |
454 | * acknowledged in that SACK including chunks | |
455 | * acknowledged by the new Cumulative TSN Ack and by Gap | |
456 | * Ack Blocks. (updated by RFC4960 Errata 3.22) | |
1da177e4 | 457 | * |
4ccbd0b0 MRL |
458 | * When partial_bytes_acked is greater than cwnd and |
459 | * before the arrival of the SACK the sender had less | |
460 | * bytes of data outstanding than cwnd (i.e., before | |
461 | * arrival of the SACK, flightsize was less than cwnd), | |
462 | * reset partial_bytes_acked to cwnd. (RFC 4960 Errata | |
463 | * 3.26) | |
464 | * | |
d0b53f40 MRL |
465 | * When partial_bytes_acked is equal to or greater than |
466 | * cwnd and before the arrival of the SACK the sender | |
467 | * had cwnd or more bytes of data outstanding (i.e., | |
468 | * before arrival of the SACK, flightsize was greater | |
469 | * than or equal to cwnd), partial_bytes_acked is reset | |
470 | * to (partial_bytes_acked - cwnd). Next, cwnd is | |
471 | * increased by MTU. (RFC 4960 Errata 3.12) | |
1da177e4 LT |
472 | */ |
473 | pba += bytes_acked; | |
4ccbd0b0 MRL |
474 | if (pba > cwnd && flight_size < cwnd) |
475 | pba = cwnd; | |
476 | if (pba >= cwnd && flight_size >= cwnd) { | |
d0b53f40 | 477 | pba = pba - cwnd; |
1da177e4 | 478 | cwnd += pmtu; |
1da177e4 | 479 | } |
bb33381d DB |
480 | |
481 | pr_debug("%s: congestion avoidance: transport:%p, " | |
482 | "bytes_acked:%d, cwnd:%d, ssthresh:%d, " | |
483 | "flight_size:%d, pba:%d\n", __func__, | |
484 | transport, bytes_acked, cwnd, ssthresh, | |
485 | flight_size, pba); | |
1da177e4 LT |
486 | } |
487 | ||
488 | transport->cwnd = cwnd; | |
489 | transport->partial_bytes_acked = pba; | |
490 | } | |
491 | ||
492 | /* This routine is used to lower the transport's cwnd when congestion is | |
493 | * detected. | |
494 | */ | |
495 | void sctp_transport_lower_cwnd(struct sctp_transport *transport, | |
233e7936 | 496 | enum sctp_lower_cwnd reason) |
1da177e4 | 497 | { |
cf9b4812 VY |
498 | struct sctp_association *asoc = transport->asoc; |
499 | ||
1da177e4 LT |
500 | switch (reason) { |
501 | case SCTP_LOWER_CWND_T3_RTX: | |
502 | /* RFC 2960 Section 7.2.3, sctpimpguide | |
503 | * When the T3-rtx timer expires on an address, SCTP should | |
504 | * perform slow start by: | |
505 | * ssthresh = max(cwnd/2, 4*MTU) | |
506 | * cwnd = 1*MTU | |
507 | * partial_bytes_acked = 0 | |
508 | */ | |
509 | transport->ssthresh = max(transport->cwnd/2, | |
cf9b4812 VY |
510 | 4*asoc->pathmtu); |
511 | transport->cwnd = asoc->pathmtu; | |
33ce8281 | 512 | |
cf9b4812 VY |
513 | /* T3-rtx also clears fast recovery */ |
514 | asoc->fast_recovery = 0; | |
1da177e4 LT |
515 | break; |
516 | ||
517 | case SCTP_LOWER_CWND_FAST_RTX: | |
518 | /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the | |
519 | * destination address(es) to which the missing DATA chunks | |
520 | * were last sent, according to the formula described in | |
521 | * Section 7.2.3. | |
d808ad9a YH |
522 | * |
523 | * RFC 2960 7.2.3, sctpimpguide Upon detection of packet | |
1da177e4 LT |
524 | * losses from SACK (see Section 7.2.4), An endpoint |
525 | * should do the following: | |
526 | * ssthresh = max(cwnd/2, 4*MTU) | |
527 | * cwnd = ssthresh | |
528 | * partial_bytes_acked = 0 | |
529 | */ | |
cf9b4812 | 530 | if (asoc->fast_recovery) |
a6465234 VY |
531 | return; |
532 | ||
533 | /* Mark Fast recovery */ | |
cf9b4812 VY |
534 | asoc->fast_recovery = 1; |
535 | asoc->fast_recovery_exit = asoc->next_tsn - 1; | |
a6465234 | 536 | |
1da177e4 | 537 | transport->ssthresh = max(transport->cwnd/2, |
cf9b4812 | 538 | 4*asoc->pathmtu); |
1da177e4 LT |
539 | transport->cwnd = transport->ssthresh; |
540 | break; | |
541 | ||
542 | case SCTP_LOWER_CWND_ECNE: | |
543 | /* RFC 2481 Section 6.1.2. | |
544 | * If the sender receives an ECN-Echo ACK packet | |
545 | * then the sender knows that congestion was encountered in the | |
546 | * network on the path from the sender to the receiver. The | |
547 | * indication of congestion should be treated just as a | |
548 | * congestion loss in non-ECN Capable TCP. That is, the TCP | |
549 | * source halves the congestion window "cwnd" and reduces the | |
550 | * slow start threshold "ssthresh". | |
551 | * A critical condition is that TCP does not react to | |
552 | * congestion indications more than once every window of | |
553 | * data (or more loosely more than once every round-trip time). | |
554 | */ | |
f61f6f82 WY |
555 | if (time_after(jiffies, transport->last_time_ecne_reduced + |
556 | transport->rtt)) { | |
1da177e4 | 557 | transport->ssthresh = max(transport->cwnd/2, |
cf9b4812 | 558 | 4*asoc->pathmtu); |
1da177e4 LT |
559 | transport->cwnd = transport->ssthresh; |
560 | transport->last_time_ecne_reduced = jiffies; | |
561 | } | |
562 | break; | |
563 | ||
564 | case SCTP_LOWER_CWND_INACTIVE: | |
565 | /* RFC 2960 Section 7.2.1, sctpimpguide | |
566 | * When the endpoint does not transmit data on a given | |
567 | * transport address, the cwnd of the transport address | |
568 | * should be adjusted to max(cwnd/2, 4*MTU) per RTO. | |
569 | * NOTE: Although the draft recommends that this check needs | |
570 | * to be done every RTO interval, we do it every hearbeat | |
571 | * interval. | |
572 | */ | |
245cba7e | 573 | transport->cwnd = max(transport->cwnd/2, |
cf9b4812 | 574 | 4*asoc->pathmtu); |
a02d036c MRL |
575 | /* RFC 4960 Errata 3.27.2: also adjust sshthresh */ |
576 | transport->ssthresh = transport->cwnd; | |
1da177e4 | 577 | break; |
3ff50b79 | 578 | } |
1da177e4 LT |
579 | |
580 | transport->partial_bytes_acked = 0; | |
bb33381d DB |
581 | |
582 | pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d\n", | |
583 | __func__, transport, reason, transport->cwnd, | |
584 | transport->ssthresh); | |
1da177e4 LT |
585 | } |
586 | ||
46d5a808 VY |
587 | /* Apply Max.Burst limit to the congestion window: |
588 | * sctpimpguide-05 2.14.2 | |
589 | * D) When the time comes for the sender to | |
590 | * transmit new DATA chunks, the protocol parameter Max.Burst MUST | |
591 | * first be applied to limit how many new DATA chunks may be sent. | |
592 | * The limit is applied by adjusting cwnd as follows: | |
593 | * if ((flightsize+ Max.Burst * MTU) < cwnd) | |
594 | * cwnd = flightsize + Max.Burst * MTU | |
595 | */ | |
596 | ||
597 | void sctp_transport_burst_limited(struct sctp_transport *t) | |
598 | { | |
599 | struct sctp_association *asoc = t->asoc; | |
600 | u32 old_cwnd = t->cwnd; | |
601 | u32 max_burst_bytes; | |
602 | ||
78ac814f | 603 | if (t->burst_limited || asoc->max_burst == 0) |
46d5a808 VY |
604 | return; |
605 | ||
606 | max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu); | |
607 | if (max_burst_bytes < old_cwnd) { | |
608 | t->cwnd = max_burst_bytes; | |
609 | t->burst_limited = old_cwnd; | |
610 | } | |
611 | } | |
612 | ||
613 | /* Restore the old cwnd congestion window, after the burst had it's | |
614 | * desired effect. | |
615 | */ | |
616 | void sctp_transport_burst_reset(struct sctp_transport *t) | |
617 | { | |
618 | if (t->burst_limited) { | |
619 | t->cwnd = t->burst_limited; | |
620 | t->burst_limited = 0; | |
621 | } | |
622 | } | |
623 | ||
1da177e4 | 624 | /* What is the next timeout value for this transport? */ |
8f61059a | 625 | unsigned long sctp_transport_timeout(struct sctp_transport *trans) |
1da177e4 | 626 | { |
8f61059a | 627 | /* RTO + timer slack +/- 50% of RTO */ |
ba6f5e33 | 628 | unsigned long timeout = trans->rto >> 1; |
8f61059a DB |
629 | |
630 | if (trans->state != SCTP_UNCONFIRMED && | |
631 | trans->state != SCTP_PF) | |
632 | timeout += trans->hbinterval; | |
633 | ||
1d88ba1e | 634 | return max_t(unsigned long, timeout, HZ / 5); |
1da177e4 | 635 | } |
749bf921 VY |
636 | |
637 | /* Reset transport variables to their initial values */ | |
638 | void sctp_transport_reset(struct sctp_transport *t) | |
639 | { | |
640 | struct sctp_association *asoc = t->asoc; | |
641 | ||
642 | /* RFC 2960 (bis), Section 5.2.4 | |
643 | * All the congestion control parameters (e.g., cwnd, ssthresh) | |
644 | * related to this peer MUST be reset to their initial values | |
645 | * (see Section 6.2.1) | |
646 | */ | |
647 | t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380)); | |
46d5a808 | 648 | t->burst_limited = 0; |
289f4249 | 649 | t->ssthresh = asoc->peer.i.a_rwnd; |
5fdd4bae | 650 | t->rto = asoc->rto_initial; |
196d6759 | 651 | sctp_max_rto(asoc, t); |
749bf921 VY |
652 | t->rtt = 0; |
653 | t->srtt = 0; | |
654 | t->rttvar = 0; | |
655 | ||
b564d62e | 656 | /* Reset these additional variables so that we have a clean slate. */ |
749bf921 VY |
657 | t->partial_bytes_acked = 0; |
658 | t->flight_size = 0; | |
659 | t->error_count = 0; | |
660 | t->rto_pending = 0; | |
faee47cd | 661 | t->hb_sent = 0; |
749bf921 VY |
662 | |
663 | /* Initialize the state information for SFR-CACC */ | |
664 | t->cacc.changeover_active = 0; | |
665 | t->cacc.cycling_changeover = 0; | |
666 | t->cacc.next_tsn_at_change = 0; | |
667 | t->cacc.cacc_saw_newack = 0; | |
668 | } | |
ddc4bbee MH |
669 | |
670 | /* Schedule retransmission on the given transport */ | |
671 | void sctp_transport_immediate_rtx(struct sctp_transport *t) | |
672 | { | |
673 | /* Stop pending T3_rtx_timer */ | |
25cc4ae9 | 674 | if (del_timer(&t->T3_rtx_timer)) |
ddc4bbee | 675 | sctp_transport_put(t); |
25cc4ae9 | 676 | |
ddc4bbee MH |
677 | sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX); |
678 | if (!timer_pending(&t->T3_rtx_timer)) { | |
679 | if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto)) | |
680 | sctp_transport_hold(t); | |
681 | } | |
ddc4bbee | 682 | } |
c86a773c JA |
683 | |
684 | /* Drop dst */ | |
685 | void sctp_transport_dst_release(struct sctp_transport *t) | |
686 | { | |
687 | dst_release(t->dst); | |
688 | t->dst = NULL; | |
689 | t->dst_pending_confirm = 0; | |
690 | } | |
691 | ||
692 | /* Schedule neighbour confirm */ | |
693 | void sctp_transport_dst_confirm(struct sctp_transport *t) | |
694 | { | |
695 | t->dst_pending_confirm = 1; | |
696 | } |