1 /* SCTP kernel implementation
2 * Copyright (c) 1999-2000 Cisco, Inc.
3 * Copyright (c) 1999-2001 Motorola, Inc.
4 * Copyright (c) 2001-2003 International Business Machines Corp.
5 * Copyright (c) 2001 Intel Corp.
6 * Copyright (c) 2001 La Monte H.P. Yarroll
8 * This file is part of the SCTP kernel implementation
10 * This module provides the abstraction for an SCTP tranport representing
11 * a remote transport address. For local transport addresses, we just use
14 * This SCTP implementation is free software;
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)
20 * This SCTP implementation is distributed in the hope that it
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.
26 * You should have received a copy of the GNU General Public License
27 * along with GNU CC; see the file COPYING. If not, write to
28 * the Free Software Foundation, 59 Temple Place - Suite 330,
29 * Boston, MA 02111-1307, USA.
31 * Please send any bug reports or fixes you make to the
33 * lksctp developers <lksctp-developers@lists.sourceforge.net>
35 * Or submit a bug report through the following website:
36 * http://www.sf.net/projects/lksctp
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Karl Knutson <karl@athena.chicago.il.us>
41 * Jon Grimm <jgrimm@us.ibm.com>
42 * Xingang Guo <xingang.guo@intel.com>
43 * Hui Huang <hui.huang@nokia.com>
44 * Sridhar Samudrala <sri@us.ibm.com>
45 * Ardelle Fan <ardelle.fan@intel.com>
47 * Any bugs reported given to us we will try to fix... any fixes shared will
48 * be incorporated into the next SCTP release.
51 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
53 #include <linux/slab.h>
54 #include <linux/types.h>
55 #include <linux/random.h>
56 #include <net/sctp/sctp.h>
57 #include <net/sctp/sm.h>
59 /* 1st Level Abstractions. */
61 /* Initialize a new transport from provided memory. */
62 static struct sctp_transport
*sctp_transport_init(struct sctp_transport
*peer
,
63 const union sctp_addr
*addr
,
66 /* Copy in the address. */
68 peer
->af_specific
= sctp_get_af_specific(addr
->sa
.sa_family
);
69 memset(&peer
->saddr
, 0, sizeof(union sctp_addr
));
71 peer
->sack_generation
= 0;
73 /* From 6.3.1 RTO Calculation:
75 * C1) Until an RTT measurement has been made for a packet sent to the
76 * given destination transport address, set RTO to the protocol
77 * parameter 'RTO.Initial'.
79 peer
->rto
= msecs_to_jiffies(sctp_rto_initial
);
81 peer
->last_time_heard
= jiffies
;
82 peer
->last_time_ecne_reduced
= jiffies
;
84 peer
->param_flags
= SPP_HB_DISABLE
|
88 /* Initialize the default path max_retrans. */
89 peer
->pathmaxrxt
= sctp_max_retrans_path
;
91 INIT_LIST_HEAD(&peer
->transmitted
);
92 INIT_LIST_HEAD(&peer
->send_ready
);
93 INIT_LIST_HEAD(&peer
->transports
);
95 setup_timer(&peer
->T3_rtx_timer
, sctp_generate_t3_rtx_event
,
97 setup_timer(&peer
->hb_timer
, sctp_generate_heartbeat_event
,
99 setup_timer(&peer
->proto_unreach_timer
,
100 sctp_generate_proto_unreach_event
, (unsigned long)peer
);
102 /* Initialize the 64-bit random nonce sent with heartbeat. */
103 get_random_bytes(&peer
->hb_nonce
, sizeof(peer
->hb_nonce
));
105 atomic_set(&peer
->refcnt
, 1);
110 /* Allocate and initialize a new transport. */
111 struct sctp_transport
*sctp_transport_new(const union sctp_addr
*addr
,
114 struct sctp_transport
*transport
;
116 transport
= t_new(struct sctp_transport
, gfp
);
120 if (!sctp_transport_init(transport
, addr
, gfp
))
123 transport
->malloced
= 1;
124 SCTP_DBG_OBJCNT_INC(transport
);
135 /* This transport is no longer needed. Free up if possible, or
136 * delay until it last reference count.
138 void sctp_transport_free(struct sctp_transport
*transport
)
142 /* Try to delete the heartbeat timer. */
143 if (del_timer(&transport
->hb_timer
))
144 sctp_transport_put(transport
);
146 /* Delete the T3_rtx timer if it's active.
147 * There is no point in not doing this now and letting
148 * structure hang around in memory since we know
149 * the tranport is going away.
151 if (timer_pending(&transport
->T3_rtx_timer
) &&
152 del_timer(&transport
->T3_rtx_timer
))
153 sctp_transport_put(transport
);
155 /* Delete the ICMP proto unreachable timer if it's active. */
156 if (timer_pending(&transport
->proto_unreach_timer
) &&
157 del_timer(&transport
->proto_unreach_timer
))
158 sctp_association_put(transport
->asoc
);
160 sctp_transport_put(transport
);
163 /* Destroy the transport data structure.
164 * Assumes there are no more users of this structure.
166 static void sctp_transport_destroy(struct sctp_transport
*transport
)
168 SCTP_ASSERT(transport
->dead
, "Transport is not dead", return);
171 sctp_association_put(transport
->asoc
);
173 sctp_packet_free(&transport
->packet
);
175 dst_release(transport
->dst
);
177 SCTP_DBG_OBJCNT_DEC(transport
);
180 /* Start T3_rtx timer if it is not already running and update the heartbeat
181 * timer. This routine is called every time a DATA chunk is sent.
183 void sctp_transport_reset_timers(struct sctp_transport
*transport
)
185 /* RFC 2960 6.3.2 Retransmission Timer Rules
187 * R1) Every time a DATA chunk is sent to any address(including a
188 * retransmission), if the T3-rtx timer of that address is not running
189 * start it running so that it will expire after the RTO of that
193 if (!timer_pending(&transport
->T3_rtx_timer
))
194 if (!mod_timer(&transport
->T3_rtx_timer
,
195 jiffies
+ transport
->rto
))
196 sctp_transport_hold(transport
);
198 /* When a data chunk is sent, reset the heartbeat interval. */
199 if (!mod_timer(&transport
->hb_timer
,
200 sctp_transport_timeout(transport
)))
201 sctp_transport_hold(transport
);
204 /* This transport has been assigned to an association.
205 * Initialize fields from the association or from the sock itself.
206 * Register the reference count in the association.
208 void sctp_transport_set_owner(struct sctp_transport
*transport
,
209 struct sctp_association
*asoc
)
211 transport
->asoc
= asoc
;
212 sctp_association_hold(asoc
);
215 /* Initialize the pmtu of a transport. */
216 void sctp_transport_pmtu(struct sctp_transport
*transport
, struct sock
*sk
)
218 /* If we don't have a fresh route, look one up */
219 if (!transport
->dst
|| transport
->dst
->obsolete
> 1) {
220 dst_release(transport
->dst
);
221 transport
->af_specific
->get_dst(transport
, &transport
->saddr
,
225 if (transport
->dst
) {
226 transport
->pathmtu
= dst_mtu(transport
->dst
);
228 transport
->pathmtu
= SCTP_DEFAULT_MAXSEGMENT
;
231 void sctp_transport_update_pmtu(struct sctp_transport
*t
, u32 pmtu
)
233 struct dst_entry
*dst
;
235 if (unlikely(pmtu
< SCTP_DEFAULT_MINSEGMENT
)) {
236 pr_warn("%s: Reported pmtu %d too low, using default minimum of %d\n",
238 SCTP_DEFAULT_MINSEGMENT
);
239 /* Use default minimum segment size and disable
240 * pmtu discovery on this transport.
242 t
->pathmtu
= SCTP_DEFAULT_MINSEGMENT
;
247 dst
= sctp_transport_dst_check(t
);
249 dst
->ops
->update_pmtu(dst
, pmtu
);
252 /* Caches the dst entry and source address for a transport's destination
255 void sctp_transport_route(struct sctp_transport
*transport
,
256 union sctp_addr
*saddr
, struct sctp_sock
*opt
)
258 struct sctp_association
*asoc
= transport
->asoc
;
259 struct sctp_af
*af
= transport
->af_specific
;
261 af
->get_dst(transport
, saddr
, &transport
->fl
, sctp_opt2sk(opt
));
264 memcpy(&transport
->saddr
, saddr
, sizeof(union sctp_addr
));
266 af
->get_saddr(opt
, transport
, &transport
->fl
);
268 if ((transport
->param_flags
& SPP_PMTUD_DISABLE
) && transport
->pathmtu
) {
271 if (transport
->dst
) {
272 transport
->pathmtu
= dst_mtu(transport
->dst
);
274 /* Initialize sk->sk_rcv_saddr, if the transport is the
275 * association's active path for getsockname().
277 if (asoc
&& (!asoc
->peer
.primary_path
||
278 (transport
== asoc
->peer
.active_path
)))
279 opt
->pf
->af
->to_sk_saddr(&transport
->saddr
,
282 transport
->pathmtu
= SCTP_DEFAULT_MAXSEGMENT
;
285 /* Hold a reference to a transport. */
286 void sctp_transport_hold(struct sctp_transport
*transport
)
288 atomic_inc(&transport
->refcnt
);
291 /* Release a reference to a transport and clean up
292 * if there are no more references.
294 void sctp_transport_put(struct sctp_transport
*transport
)
296 if (atomic_dec_and_test(&transport
->refcnt
))
297 sctp_transport_destroy(transport
);
300 /* Update transport's RTO based on the newly calculated RTT. */
301 void sctp_transport_update_rto(struct sctp_transport
*tp
, __u32 rtt
)
303 /* Check for valid transport. */
304 SCTP_ASSERT(tp
, "NULL transport", return);
306 /* We should not be doing any RTO updates unless rto_pending is set. */
307 SCTP_ASSERT(tp
->rto_pending
, "rto_pending not set", return);
309 if (tp
->rttvar
|| tp
->srtt
) {
310 /* 6.3.1 C3) When a new RTT measurement R' is made, set
311 * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
312 * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
315 /* Note: The above algorithm has been rewritten to
316 * express rto_beta and rto_alpha as inverse powers
318 * For example, assuming the default value of RTO.Alpha of
319 * 1/8, rto_alpha would be expressed as 3.
321 tp
->rttvar
= tp
->rttvar
- (tp
->rttvar
>> sctp_rto_beta
)
322 + ((abs(tp
->srtt
- rtt
)) >> sctp_rto_beta
);
323 tp
->srtt
= tp
->srtt
- (tp
->srtt
>> sctp_rto_alpha
)
324 + (rtt
>> sctp_rto_alpha
);
326 /* 6.3.1 C2) When the first RTT measurement R is made, set
327 * SRTT <- R, RTTVAR <- R/2.
330 tp
->rttvar
= rtt
>> 1;
333 /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
334 * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
337 tp
->rttvar
= SCTP_CLOCK_GRANULARITY
;
339 /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
340 tp
->rto
= tp
->srtt
+ (tp
->rttvar
<< 2);
342 /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
343 * seconds then it is rounded up to RTO.Min seconds.
345 if (tp
->rto
< tp
->asoc
->rto_min
)
346 tp
->rto
= tp
->asoc
->rto_min
;
348 /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
349 * at least RTO.max seconds.
351 if (tp
->rto
> tp
->asoc
->rto_max
)
352 tp
->rto
= tp
->asoc
->rto_max
;
356 /* Reset rto_pending so that a new RTT measurement is started when a
357 * new data chunk is sent.
361 SCTP_DEBUG_PRINTK("%s: transport: %p, rtt: %d, srtt: %d "
362 "rttvar: %d, rto: %ld\n", __func__
,
363 tp
, rtt
, tp
->srtt
, tp
->rttvar
, tp
->rto
);
366 /* This routine updates the transport's cwnd and partial_bytes_acked
367 * parameters based on the bytes acked in the received SACK.
369 void sctp_transport_raise_cwnd(struct sctp_transport
*transport
,
370 __u32 sack_ctsn
, __u32 bytes_acked
)
372 struct sctp_association
*asoc
= transport
->asoc
;
373 __u32 cwnd
, ssthresh
, flight_size
, pba
, pmtu
;
375 cwnd
= transport
->cwnd
;
376 flight_size
= transport
->flight_size
;
378 /* See if we need to exit Fast Recovery first */
379 if (asoc
->fast_recovery
&&
380 TSN_lte(asoc
->fast_recovery_exit
, sack_ctsn
))
381 asoc
->fast_recovery
= 0;
383 /* The appropriate cwnd increase algorithm is performed if, and only
384 * if the cumulative TSN whould advanced and the congestion window is
385 * being fully utilized.
387 if (TSN_lte(sack_ctsn
, transport
->asoc
->ctsn_ack_point
) ||
388 (flight_size
< cwnd
))
391 ssthresh
= transport
->ssthresh
;
392 pba
= transport
->partial_bytes_acked
;
393 pmtu
= transport
->asoc
->pathmtu
;
395 if (cwnd
<= ssthresh
) {
397 * o When cwnd is less than or equal to ssthresh, an SCTP
398 * endpoint MUST use the slow-start algorithm to increase
399 * cwnd only if the current congestion window is being fully
400 * utilized, an incoming SACK advances the Cumulative TSN
401 * Ack Point, and the data sender is not in Fast Recovery.
402 * Only when these three conditions are met can the cwnd be
403 * increased; otherwise, the cwnd MUST not be increased.
404 * If these conditions are met, then cwnd MUST be increased
405 * by, at most, the lesser of 1) the total size of the
406 * previously outstanding DATA chunk(s) acknowledged, and
407 * 2) the destination's path MTU. This upper bound protects
408 * against the ACK-Splitting attack outlined in [SAVAGE99].
410 if (asoc
->fast_recovery
)
413 if (bytes_acked
> pmtu
)
417 SCTP_DEBUG_PRINTK("%s: SLOW START: transport: %p, "
418 "bytes_acked: %d, cwnd: %d, ssthresh: %d, "
419 "flight_size: %d, pba: %d\n",
421 transport
, bytes_acked
, cwnd
,
422 ssthresh
, flight_size
, pba
);
424 /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
425 * upon each SACK arrival that advances the Cumulative TSN Ack
426 * Point, increase partial_bytes_acked by the total number of
427 * bytes of all new chunks acknowledged in that SACK including
428 * chunks acknowledged by the new Cumulative TSN Ack and by
431 * When partial_bytes_acked is equal to or greater than cwnd
432 * and before the arrival of the SACK the sender had cwnd or
433 * more bytes of data outstanding (i.e., before arrival of the
434 * SACK, flightsize was greater than or equal to cwnd),
435 * increase cwnd by MTU, and reset partial_bytes_acked to
436 * (partial_bytes_acked - cwnd).
441 pba
= ((cwnd
< pba
) ? (pba
- cwnd
) : 0);
443 SCTP_DEBUG_PRINTK("%s: CONGESTION AVOIDANCE: "
444 "transport: %p, bytes_acked: %d, cwnd: %d, "
445 "ssthresh: %d, flight_size: %d, pba: %d\n",
447 transport
, bytes_acked
, cwnd
,
448 ssthresh
, flight_size
, pba
);
451 transport
->cwnd
= cwnd
;
452 transport
->partial_bytes_acked
= pba
;
455 /* This routine is used to lower the transport's cwnd when congestion is
458 void sctp_transport_lower_cwnd(struct sctp_transport
*transport
,
459 sctp_lower_cwnd_t reason
)
461 struct sctp_association
*asoc
= transport
->asoc
;
464 case SCTP_LOWER_CWND_T3_RTX
:
465 /* RFC 2960 Section 7.2.3, sctpimpguide
466 * When the T3-rtx timer expires on an address, SCTP should
467 * perform slow start by:
468 * ssthresh = max(cwnd/2, 4*MTU)
470 * partial_bytes_acked = 0
472 transport
->ssthresh
= max(transport
->cwnd
/2,
474 transport
->cwnd
= asoc
->pathmtu
;
476 /* T3-rtx also clears fast recovery */
477 asoc
->fast_recovery
= 0;
480 case SCTP_LOWER_CWND_FAST_RTX
:
481 /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
482 * destination address(es) to which the missing DATA chunks
483 * were last sent, according to the formula described in
486 * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
487 * losses from SACK (see Section 7.2.4), An endpoint
488 * should do the following:
489 * ssthresh = max(cwnd/2, 4*MTU)
491 * partial_bytes_acked = 0
493 if (asoc
->fast_recovery
)
496 /* Mark Fast recovery */
497 asoc
->fast_recovery
= 1;
498 asoc
->fast_recovery_exit
= asoc
->next_tsn
- 1;
500 transport
->ssthresh
= max(transport
->cwnd
/2,
502 transport
->cwnd
= transport
->ssthresh
;
505 case SCTP_LOWER_CWND_ECNE
:
506 /* RFC 2481 Section 6.1.2.
507 * If the sender receives an ECN-Echo ACK packet
508 * then the sender knows that congestion was encountered in the
509 * network on the path from the sender to the receiver. The
510 * indication of congestion should be treated just as a
511 * congestion loss in non-ECN Capable TCP. That is, the TCP
512 * source halves the congestion window "cwnd" and reduces the
513 * slow start threshold "ssthresh".
514 * A critical condition is that TCP does not react to
515 * congestion indications more than once every window of
516 * data (or more loosely more than once every round-trip time).
518 if (time_after(jiffies
, transport
->last_time_ecne_reduced
+
520 transport
->ssthresh
= max(transport
->cwnd
/2,
522 transport
->cwnd
= transport
->ssthresh
;
523 transport
->last_time_ecne_reduced
= jiffies
;
527 case SCTP_LOWER_CWND_INACTIVE
:
528 /* RFC 2960 Section 7.2.1, sctpimpguide
529 * When the endpoint does not transmit data on a given
530 * transport address, the cwnd of the transport address
531 * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
532 * NOTE: Although the draft recommends that this check needs
533 * to be done every RTO interval, we do it every hearbeat
536 transport
->cwnd
= max(transport
->cwnd
/2,
541 transport
->partial_bytes_acked
= 0;
542 SCTP_DEBUG_PRINTK("%s: transport: %p reason: %d cwnd: "
543 "%d ssthresh: %d\n", __func__
,
545 transport
->cwnd
, transport
->ssthresh
);
548 /* Apply Max.Burst limit to the congestion window:
549 * sctpimpguide-05 2.14.2
550 * D) When the time comes for the sender to
551 * transmit new DATA chunks, the protocol parameter Max.Burst MUST
552 * first be applied to limit how many new DATA chunks may be sent.
553 * The limit is applied by adjusting cwnd as follows:
554 * if ((flightsize+ Max.Burst * MTU) < cwnd)
555 * cwnd = flightsize + Max.Burst * MTU
558 void sctp_transport_burst_limited(struct sctp_transport
*t
)
560 struct sctp_association
*asoc
= t
->asoc
;
561 u32 old_cwnd
= t
->cwnd
;
564 if (t
->burst_limited
)
567 max_burst_bytes
= t
->flight_size
+ (asoc
->max_burst
* asoc
->pathmtu
);
568 if (max_burst_bytes
< old_cwnd
) {
569 t
->cwnd
= max_burst_bytes
;
570 t
->burst_limited
= old_cwnd
;
574 /* Restore the old cwnd congestion window, after the burst had it's
577 void sctp_transport_burst_reset(struct sctp_transport
*t
)
579 if (t
->burst_limited
) {
580 t
->cwnd
= t
->burst_limited
;
581 t
->burst_limited
= 0;
585 /* What is the next timeout value for this transport? */
586 unsigned long sctp_transport_timeout(struct sctp_transport
*t
)
588 unsigned long timeout
;
589 timeout
= t
->rto
+ sctp_jitter(t
->rto
);
590 if (t
->state
!= SCTP_UNCONFIRMED
)
591 timeout
+= t
->hbinterval
;
596 /* Reset transport variables to their initial values */
597 void sctp_transport_reset(struct sctp_transport
*t
)
599 struct sctp_association
*asoc
= t
->asoc
;
601 /* RFC 2960 (bis), Section 5.2.4
602 * All the congestion control parameters (e.g., cwnd, ssthresh)
603 * related to this peer MUST be reset to their initial values
604 * (see Section 6.2.1)
606 t
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
, 2*asoc
->pathmtu
, 4380));
607 t
->burst_limited
= 0;
608 t
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
609 t
->rto
= asoc
->rto_initial
;
614 /* Reset these additional varibles so that we have a clean
617 t
->partial_bytes_acked
= 0;
623 /* Initialize the state information for SFR-CACC */
624 t
->cacc
.changeover_active
= 0;
625 t
->cacc
.cycling_changeover
= 0;
626 t
->cacc
.next_tsn_at_change
= 0;
627 t
->cacc
.cacc_saw_newack
= 0;
630 /* Schedule retransmission on the given transport */
631 void sctp_transport_immediate_rtx(struct sctp_transport
*t
)
633 /* Stop pending T3_rtx_timer */
634 if (timer_pending(&t
->T3_rtx_timer
)) {
635 (void)del_timer(&t
->T3_rtx_timer
);
636 sctp_transport_put(t
);
638 sctp_retransmit(&t
->asoc
->outqueue
, t
, SCTP_RTXR_T3_RTX
);
639 if (!timer_pending(&t
->T3_rtx_timer
)) {
640 if (!mod_timer(&t
->T3_rtx_timer
, jiffies
+ t
->rto
))
641 sctp_transport_hold(t
);