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 /* From 6.3.1 RTO Calculation:
73 * C1) Until an RTT measurement has been made for a packet sent to the
74 * given destination transport address, set RTO to the protocol
75 * parameter 'RTO.Initial'.
77 peer
->rto
= msecs_to_jiffies(sctp_rto_initial
);
79 peer
->last_time_heard
= jiffies
;
80 peer
->last_time_ecne_reduced
= jiffies
;
82 peer
->param_flags
= SPP_HB_DISABLE
|
86 /* Initialize the default path max_retrans. */
87 peer
->pathmaxrxt
= sctp_max_retrans_path
;
89 INIT_LIST_HEAD(&peer
->transmitted
);
90 INIT_LIST_HEAD(&peer
->send_ready
);
91 INIT_LIST_HEAD(&peer
->transports
);
93 setup_timer(&peer
->T3_rtx_timer
, sctp_generate_t3_rtx_event
,
95 setup_timer(&peer
->hb_timer
, sctp_generate_heartbeat_event
,
97 setup_timer(&peer
->proto_unreach_timer
,
98 sctp_generate_proto_unreach_event
, (unsigned long)peer
);
100 /* Initialize the 64-bit random nonce sent with heartbeat. */
101 get_random_bytes(&peer
->hb_nonce
, sizeof(peer
->hb_nonce
));
103 atomic_set(&peer
->refcnt
, 1);
108 /* Allocate and initialize a new transport. */
109 struct sctp_transport
*sctp_transport_new(const union sctp_addr
*addr
,
112 struct sctp_transport
*transport
;
114 transport
= t_new(struct sctp_transport
, gfp
);
118 if (!sctp_transport_init(transport
, addr
, gfp
))
121 transport
->malloced
= 1;
122 SCTP_DBG_OBJCNT_INC(transport
);
133 /* This transport is no longer needed. Free up if possible, or
134 * delay until it last reference count.
136 void sctp_transport_free(struct sctp_transport
*transport
)
140 /* Try to delete the heartbeat timer. */
141 if (del_timer(&transport
->hb_timer
))
142 sctp_transport_put(transport
);
144 /* Delete the T3_rtx timer if it's active.
145 * There is no point in not doing this now and letting
146 * structure hang around in memory since we know
147 * the tranport is going away.
149 if (timer_pending(&transport
->T3_rtx_timer
) &&
150 del_timer(&transport
->T3_rtx_timer
))
151 sctp_transport_put(transport
);
153 /* Delete the ICMP proto unreachable timer if it's active. */
154 if (timer_pending(&transport
->proto_unreach_timer
) &&
155 del_timer(&transport
->proto_unreach_timer
))
156 sctp_association_put(transport
->asoc
);
158 sctp_transport_put(transport
);
161 /* Destroy the transport data structure.
162 * Assumes there are no more users of this structure.
164 static void sctp_transport_destroy(struct sctp_transport
*transport
)
166 SCTP_ASSERT(transport
->dead
, "Transport is not dead", return);
169 sctp_association_put(transport
->asoc
);
171 sctp_packet_free(&transport
->packet
);
173 dst_release(transport
->dst
);
175 SCTP_DBG_OBJCNT_DEC(transport
);
178 /* Start T3_rtx timer if it is not already running and update the heartbeat
179 * timer. This routine is called every time a DATA chunk is sent.
181 void sctp_transport_reset_timers(struct sctp_transport
*transport
)
183 /* RFC 2960 6.3.2 Retransmission Timer Rules
185 * R1) Every time a DATA chunk is sent to any address(including a
186 * retransmission), if the T3-rtx timer of that address is not running
187 * start it running so that it will expire after the RTO of that
191 if (!timer_pending(&transport
->T3_rtx_timer
))
192 if (!mod_timer(&transport
->T3_rtx_timer
,
193 jiffies
+ transport
->rto
))
194 sctp_transport_hold(transport
);
196 /* When a data chunk is sent, reset the heartbeat interval. */
197 if (!mod_timer(&transport
->hb_timer
,
198 sctp_transport_timeout(transport
)))
199 sctp_transport_hold(transport
);
202 /* This transport has been assigned to an association.
203 * Initialize fields from the association or from the sock itself.
204 * Register the reference count in the association.
206 void sctp_transport_set_owner(struct sctp_transport
*transport
,
207 struct sctp_association
*asoc
)
209 transport
->asoc
= asoc
;
210 sctp_association_hold(asoc
);
213 /* Initialize the pmtu of a transport. */
214 void sctp_transport_pmtu(struct sctp_transport
*transport
)
216 struct dst_entry
*dst
;
218 dst
= transport
->af_specific
->get_dst(NULL
, &transport
->ipaddr
, NULL
);
221 transport
->pathmtu
= dst_mtu(dst
);
224 transport
->pathmtu
= SCTP_DEFAULT_MAXSEGMENT
;
227 /* this is a complete rip-off from __sk_dst_check
228 * the cookie is always 0 since this is how it's used in the
231 static struct dst_entry
*sctp_transport_dst_check(struct sctp_transport
*t
)
233 struct dst_entry
*dst
= t
->dst
;
235 if (dst
&& dst
->obsolete
&& dst
->ops
->check(dst
, 0) == NULL
) {
244 void sctp_transport_update_pmtu(struct sctp_transport
*t
, u32 pmtu
)
246 struct dst_entry
*dst
;
248 if (unlikely(pmtu
< SCTP_DEFAULT_MINSEGMENT
)) {
249 pr_warn("%s: Reported pmtu %d too low, using default minimum of %d\n",
251 SCTP_DEFAULT_MINSEGMENT
);
252 /* Use default minimum segment size and disable
253 * pmtu discovery on this transport.
255 t
->pathmtu
= SCTP_DEFAULT_MINSEGMENT
;
260 dst
= sctp_transport_dst_check(t
);
262 dst
->ops
->update_pmtu(dst
, pmtu
);
265 /* Caches the dst entry and source address for a transport's destination
268 void sctp_transport_route(struct sctp_transport
*transport
,
269 union sctp_addr
*saddr
, struct sctp_sock
*opt
)
271 struct sctp_association
*asoc
= transport
->asoc
;
272 struct sctp_af
*af
= transport
->af_specific
;
273 union sctp_addr
*daddr
= &transport
->ipaddr
;
274 struct dst_entry
*dst
;
276 dst
= af
->get_dst(asoc
, daddr
, saddr
);
279 memcpy(&transport
->saddr
, saddr
, sizeof(union sctp_addr
));
281 af
->get_saddr(opt
, asoc
, dst
, daddr
, &transport
->saddr
);
283 transport
->dst
= dst
;
284 if ((transport
->param_flags
& SPP_PMTUD_DISABLE
) && transport
->pathmtu
) {
288 transport
->pathmtu
= dst_mtu(dst
);
290 /* Initialize sk->sk_rcv_saddr, if the transport is the
291 * association's active path for getsockname().
293 if (asoc
&& (!asoc
->peer
.primary_path
||
294 (transport
== asoc
->peer
.active_path
)))
295 opt
->pf
->af
->to_sk_saddr(&transport
->saddr
,
298 transport
->pathmtu
= SCTP_DEFAULT_MAXSEGMENT
;
301 /* Hold a reference to a transport. */
302 void sctp_transport_hold(struct sctp_transport
*transport
)
304 atomic_inc(&transport
->refcnt
);
307 /* Release a reference to a transport and clean up
308 * if there are no more references.
310 void sctp_transport_put(struct sctp_transport
*transport
)
312 if (atomic_dec_and_test(&transport
->refcnt
))
313 sctp_transport_destroy(transport
);
316 /* Update transport's RTO based on the newly calculated RTT. */
317 void sctp_transport_update_rto(struct sctp_transport
*tp
, __u32 rtt
)
319 /* Check for valid transport. */
320 SCTP_ASSERT(tp
, "NULL transport", return);
322 /* We should not be doing any RTO updates unless rto_pending is set. */
323 SCTP_ASSERT(tp
->rto_pending
, "rto_pending not set", return);
325 if (tp
->rttvar
|| tp
->srtt
) {
326 /* 6.3.1 C3) When a new RTT measurement R' is made, set
327 * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
328 * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
331 /* Note: The above algorithm has been rewritten to
332 * express rto_beta and rto_alpha as inverse powers
334 * For example, assuming the default value of RTO.Alpha of
335 * 1/8, rto_alpha would be expressed as 3.
337 tp
->rttvar
= tp
->rttvar
- (tp
->rttvar
>> sctp_rto_beta
)
338 + ((abs(tp
->srtt
- rtt
)) >> sctp_rto_beta
);
339 tp
->srtt
= tp
->srtt
- (tp
->srtt
>> sctp_rto_alpha
)
340 + (rtt
>> sctp_rto_alpha
);
342 /* 6.3.1 C2) When the first RTT measurement R is made, set
343 * SRTT <- R, RTTVAR <- R/2.
346 tp
->rttvar
= rtt
>> 1;
349 /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
350 * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
353 tp
->rttvar
= SCTP_CLOCK_GRANULARITY
;
355 /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
356 tp
->rto
= tp
->srtt
+ (tp
->rttvar
<< 2);
358 /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
359 * seconds then it is rounded up to RTO.Min seconds.
361 if (tp
->rto
< tp
->asoc
->rto_min
)
362 tp
->rto
= tp
->asoc
->rto_min
;
364 /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
365 * at least RTO.max seconds.
367 if (tp
->rto
> tp
->asoc
->rto_max
)
368 tp
->rto
= tp
->asoc
->rto_max
;
372 /* Reset rto_pending so that a new RTT measurement is started when a
373 * new data chunk is sent.
377 SCTP_DEBUG_PRINTK("%s: transport: %p, rtt: %d, srtt: %d "
378 "rttvar: %d, rto: %ld\n", __func__
,
379 tp
, rtt
, tp
->srtt
, tp
->rttvar
, tp
->rto
);
382 /* This routine updates the transport's cwnd and partial_bytes_acked
383 * parameters based on the bytes acked in the received SACK.
385 void sctp_transport_raise_cwnd(struct sctp_transport
*transport
,
386 __u32 sack_ctsn
, __u32 bytes_acked
)
388 struct sctp_association
*asoc
= transport
->asoc
;
389 __u32 cwnd
, ssthresh
, flight_size
, pba
, pmtu
;
391 cwnd
= transport
->cwnd
;
392 flight_size
= transport
->flight_size
;
394 /* See if we need to exit Fast Recovery first */
395 if (asoc
->fast_recovery
&&
396 TSN_lte(asoc
->fast_recovery_exit
, sack_ctsn
))
397 asoc
->fast_recovery
= 0;
399 /* The appropriate cwnd increase algorithm is performed if, and only
400 * if the cumulative TSN whould advanced and the congestion window is
401 * being fully utilized.
403 if (TSN_lte(sack_ctsn
, transport
->asoc
->ctsn_ack_point
) ||
404 (flight_size
< cwnd
))
407 ssthresh
= transport
->ssthresh
;
408 pba
= transport
->partial_bytes_acked
;
409 pmtu
= transport
->asoc
->pathmtu
;
411 if (cwnd
<= ssthresh
) {
413 * o When cwnd is less than or equal to ssthresh, an SCTP
414 * endpoint MUST use the slow-start algorithm to increase
415 * cwnd only if the current congestion window is being fully
416 * utilized, an incoming SACK advances the Cumulative TSN
417 * Ack Point, and the data sender is not in Fast Recovery.
418 * Only when these three conditions are met can the cwnd be
419 * increased; otherwise, the cwnd MUST not be increased.
420 * If these conditions are met, then cwnd MUST be increased
421 * by, at most, the lesser of 1) the total size of the
422 * previously outstanding DATA chunk(s) acknowledged, and
423 * 2) the destination's path MTU. This upper bound protects
424 * against the ACK-Splitting attack outlined in [SAVAGE99].
426 if (asoc
->fast_recovery
)
429 if (bytes_acked
> pmtu
)
433 SCTP_DEBUG_PRINTK("%s: SLOW START: transport: %p, "
434 "bytes_acked: %d, cwnd: %d, ssthresh: %d, "
435 "flight_size: %d, pba: %d\n",
437 transport
, bytes_acked
, cwnd
,
438 ssthresh
, flight_size
, pba
);
440 /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
441 * upon each SACK arrival that advances the Cumulative TSN Ack
442 * Point, increase partial_bytes_acked by the total number of
443 * bytes of all new chunks acknowledged in that SACK including
444 * chunks acknowledged by the new Cumulative TSN Ack and by
447 * When partial_bytes_acked is equal to or greater than cwnd
448 * and before the arrival of the SACK the sender had cwnd or
449 * more bytes of data outstanding (i.e., before arrival of the
450 * SACK, flightsize was greater than or equal to cwnd),
451 * increase cwnd by MTU, and reset partial_bytes_acked to
452 * (partial_bytes_acked - cwnd).
457 pba
= ((cwnd
< pba
) ? (pba
- cwnd
) : 0);
459 SCTP_DEBUG_PRINTK("%s: CONGESTION AVOIDANCE: "
460 "transport: %p, bytes_acked: %d, cwnd: %d, "
461 "ssthresh: %d, flight_size: %d, pba: %d\n",
463 transport
, bytes_acked
, cwnd
,
464 ssthresh
, flight_size
, pba
);
467 transport
->cwnd
= cwnd
;
468 transport
->partial_bytes_acked
= pba
;
471 /* This routine is used to lower the transport's cwnd when congestion is
474 void sctp_transport_lower_cwnd(struct sctp_transport
*transport
,
475 sctp_lower_cwnd_t reason
)
477 struct sctp_association
*asoc
= transport
->asoc
;
480 case SCTP_LOWER_CWND_T3_RTX
:
481 /* RFC 2960 Section 7.2.3, sctpimpguide
482 * When the T3-rtx timer expires on an address, SCTP should
483 * perform slow start by:
484 * ssthresh = max(cwnd/2, 4*MTU)
486 * partial_bytes_acked = 0
488 transport
->ssthresh
= max(transport
->cwnd
/2,
490 transport
->cwnd
= asoc
->pathmtu
;
492 /* T3-rtx also clears fast recovery */
493 asoc
->fast_recovery
= 0;
496 case SCTP_LOWER_CWND_FAST_RTX
:
497 /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
498 * destination address(es) to which the missing DATA chunks
499 * were last sent, according to the formula described in
502 * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
503 * losses from SACK (see Section 7.2.4), An endpoint
504 * should do the following:
505 * ssthresh = max(cwnd/2, 4*MTU)
507 * partial_bytes_acked = 0
509 if (asoc
->fast_recovery
)
512 /* Mark Fast recovery */
513 asoc
->fast_recovery
= 1;
514 asoc
->fast_recovery_exit
= asoc
->next_tsn
- 1;
516 transport
->ssthresh
= max(transport
->cwnd
/2,
518 transport
->cwnd
= transport
->ssthresh
;
521 case SCTP_LOWER_CWND_ECNE
:
522 /* RFC 2481 Section 6.1.2.
523 * If the sender receives an ECN-Echo ACK packet
524 * then the sender knows that congestion was encountered in the
525 * network on the path from the sender to the receiver. The
526 * indication of congestion should be treated just as a
527 * congestion loss in non-ECN Capable TCP. That is, the TCP
528 * source halves the congestion window "cwnd" and reduces the
529 * slow start threshold "ssthresh".
530 * A critical condition is that TCP does not react to
531 * congestion indications more than once every window of
532 * data (or more loosely more than once every round-trip time).
534 if (time_after(jiffies
, transport
->last_time_ecne_reduced
+
536 transport
->ssthresh
= max(transport
->cwnd
/2,
538 transport
->cwnd
= transport
->ssthresh
;
539 transport
->last_time_ecne_reduced
= jiffies
;
543 case SCTP_LOWER_CWND_INACTIVE
:
544 /* RFC 2960 Section 7.2.1, sctpimpguide
545 * When the endpoint does not transmit data on a given
546 * transport address, the cwnd of the transport address
547 * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
548 * NOTE: Although the draft recommends that this check needs
549 * to be done every RTO interval, we do it every hearbeat
552 transport
->cwnd
= max(transport
->cwnd
/2,
557 transport
->partial_bytes_acked
= 0;
558 SCTP_DEBUG_PRINTK("%s: transport: %p reason: %d cwnd: "
559 "%d ssthresh: %d\n", __func__
,
561 transport
->cwnd
, transport
->ssthresh
);
564 /* Apply Max.Burst limit to the congestion window:
565 * sctpimpguide-05 2.14.2
566 * D) When the time comes for the sender to
567 * transmit new DATA chunks, the protocol parameter Max.Burst MUST
568 * first be applied to limit how many new DATA chunks may be sent.
569 * The limit is applied by adjusting cwnd as follows:
570 * if ((flightsize+ Max.Burst * MTU) < cwnd)
571 * cwnd = flightsize + Max.Burst * MTU
574 void sctp_transport_burst_limited(struct sctp_transport
*t
)
576 struct sctp_association
*asoc
= t
->asoc
;
577 u32 old_cwnd
= t
->cwnd
;
580 if (t
->burst_limited
)
583 max_burst_bytes
= t
->flight_size
+ (asoc
->max_burst
* asoc
->pathmtu
);
584 if (max_burst_bytes
< old_cwnd
) {
585 t
->cwnd
= max_burst_bytes
;
586 t
->burst_limited
= old_cwnd
;
590 /* Restore the old cwnd congestion window, after the burst had it's
593 void sctp_transport_burst_reset(struct sctp_transport
*t
)
595 if (t
->burst_limited
) {
596 t
->cwnd
= t
->burst_limited
;
597 t
->burst_limited
= 0;
601 /* What is the next timeout value for this transport? */
602 unsigned long sctp_transport_timeout(struct sctp_transport
*t
)
604 unsigned long timeout
;
605 timeout
= t
->rto
+ sctp_jitter(t
->rto
);
606 if (t
->state
!= SCTP_UNCONFIRMED
)
607 timeout
+= t
->hbinterval
;
612 /* Reset transport variables to their initial values */
613 void sctp_transport_reset(struct sctp_transport
*t
)
615 struct sctp_association
*asoc
= t
->asoc
;
617 /* RFC 2960 (bis), Section 5.2.4
618 * All the congestion control parameters (e.g., cwnd, ssthresh)
619 * related to this peer MUST be reset to their initial values
620 * (see Section 6.2.1)
622 t
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
, 2*asoc
->pathmtu
, 4380));
623 t
->burst_limited
= 0;
624 t
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
625 t
->rto
= asoc
->rto_initial
;
630 /* Reset these additional varibles so that we have a clean
633 t
->partial_bytes_acked
= 0;
639 /* Initialize the state information for SFR-CACC */
640 t
->cacc
.changeover_active
= 0;
641 t
->cacc
.cycling_changeover
= 0;
642 t
->cacc
.next_tsn_at_change
= 0;
643 t
->cacc
.cacc_saw_newack
= 0;