1 /* SCTP kernel reference 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 reference 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 * The SCTP reference 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 * The SCTP reference 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 #include <linux/types.h>
52 #include <linux/random.h>
53 #include <net/sctp/sctp.h>
54 #include <net/sctp/sm.h>
56 /* 1st Level Abstractions. */
58 /* Initialize a new transport from provided memory. */
59 static struct sctp_transport
*sctp_transport_init(struct sctp_transport
*peer
,
60 const union sctp_addr
*addr
,
63 /* Copy in the address. */
65 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'.
78 peer
->rto
= msecs_to_jiffies(sctp_rto_initial
);
81 peer
->rto_pending
= 0;
83 peer
->last_time_heard
= jiffies
;
84 peer
->last_time_used
= jiffies
;
85 peer
->last_time_ecne_reduced
= jiffies
;
87 peer
->init_sent_count
= 0;
89 peer
->param_flags
= SPP_HB_DISABLE
|
94 /* Initialize the default path max_retrans. */
95 peer
->pathmaxrxt
= sctp_max_retrans_path
;
96 peer
->error_count
= 0;
98 INIT_LIST_HEAD(&peer
->transmitted
);
99 INIT_LIST_HEAD(&peer
->send_ready
);
100 INIT_LIST_HEAD(&peer
->transports
);
102 /* Set up the retransmission timer. */
103 init_timer(&peer
->T3_rtx_timer
);
104 peer
->T3_rtx_timer
.function
= sctp_generate_t3_rtx_event
;
105 peer
->T3_rtx_timer
.data
= (unsigned long)peer
;
107 /* Set up the heartbeat timer. */
108 init_timer(&peer
->hb_timer
);
109 peer
->hb_timer
.function
= sctp_generate_heartbeat_event
;
110 peer
->hb_timer
.data
= (unsigned long)peer
;
112 /* Initialize the 64-bit random nonce sent with heartbeat. */
113 get_random_bytes(&peer
->hb_nonce
, sizeof(peer
->hb_nonce
));
115 atomic_set(&peer
->refcnt
, 1);
120 /* Initialize the state information for SFR-CACC */
121 peer
->cacc
.changeover_active
= 0;
122 peer
->cacc
.cycling_changeover
= 0;
123 peer
->cacc
.next_tsn_at_change
= 0;
124 peer
->cacc
.cacc_saw_newack
= 0;
129 /* Allocate and initialize a new transport. */
130 struct sctp_transport
*sctp_transport_new(const union sctp_addr
*addr
,
133 struct sctp_transport
*transport
;
135 transport
= t_new(struct sctp_transport
, gfp
);
139 if (!sctp_transport_init(transport
, addr
, gfp
))
142 transport
->malloced
= 1;
143 SCTP_DBG_OBJCNT_INC(transport
);
154 /* This transport is no longer needed. Free up if possible, or
155 * delay until it last reference count.
157 void sctp_transport_free(struct sctp_transport
*transport
)
161 /* Try to delete the heartbeat timer. */
162 if (del_timer(&transport
->hb_timer
))
163 sctp_transport_put(transport
);
165 /* Delete the T3_rtx timer if it's active.
166 * There is no point in not doing this now and letting
167 * structure hang around in memory since we know
168 * the tranport is going away.
170 if (timer_pending(&transport
->T3_rtx_timer
) &&
171 del_timer(&transport
->T3_rtx_timer
))
172 sctp_transport_put(transport
);
175 sctp_transport_put(transport
);
178 /* Destroy the transport data structure.
179 * Assumes there are no more users of this structure.
181 static void sctp_transport_destroy(struct sctp_transport
*transport
)
183 SCTP_ASSERT(transport
->dead
, "Transport is not dead", return);
186 sctp_association_put(transport
->asoc
);
188 sctp_packet_free(&transport
->packet
);
190 dst_release(transport
->dst
);
192 SCTP_DBG_OBJCNT_DEC(transport
);
195 /* Start T3_rtx timer if it is not already running and update the heartbeat
196 * timer. This routine is called every time a DATA chunk is sent.
198 void sctp_transport_reset_timers(struct sctp_transport
*transport
)
200 /* RFC 2960 6.3.2 Retransmission Timer Rules
202 * R1) Every time a DATA chunk is sent to any address(including a
203 * retransmission), if the T3-rtx timer of that address is not running
204 * start it running so that it will expire after the RTO of that
208 if (!timer_pending(&transport
->T3_rtx_timer
))
209 if (!mod_timer(&transport
->T3_rtx_timer
,
210 jiffies
+ transport
->rto
))
211 sctp_transport_hold(transport
);
213 /* When a data chunk is sent, reset the heartbeat interval. */
214 if (!mod_timer(&transport
->hb_timer
,
215 sctp_transport_timeout(transport
)))
216 sctp_transport_hold(transport
);
219 /* This transport has been assigned to an association.
220 * Initialize fields from the association or from the sock itself.
221 * Register the reference count in the association.
223 void sctp_transport_set_owner(struct sctp_transport
*transport
,
224 struct sctp_association
*asoc
)
226 transport
->asoc
= asoc
;
227 sctp_association_hold(asoc
);
230 /* Initialize the pmtu of a transport. */
231 void sctp_transport_pmtu(struct sctp_transport
*transport
)
233 struct dst_entry
*dst
;
235 dst
= transport
->af_specific
->get_dst(NULL
, &transport
->ipaddr
, NULL
);
238 transport
->pathmtu
= dst_mtu(dst
);
241 transport
->pathmtu
= SCTP_DEFAULT_MAXSEGMENT
;
244 /* this is a complete rip-off from __sk_dst_check
245 * the cookie is always 0 since this is how it's used in the
248 static struct dst_entry
*sctp_transport_dst_check(struct sctp_transport
*t
)
250 struct dst_entry
*dst
= t
->dst
;
252 if (dst
&& dst
->obsolete
&& dst
->ops
->check(dst
, 0) == NULL
) {
261 void sctp_transport_update_pmtu(struct sctp_transport
*t
, u32 pmtu
)
263 struct dst_entry
*dst
;
265 if (unlikely(pmtu
< SCTP_DEFAULT_MINSEGMENT
)) {
266 printk(KERN_WARNING
"%s: Reported pmtu %d too low, "
267 "using default minimum of %d\n",
269 SCTP_DEFAULT_MINSEGMENT
);
270 /* Use default minimum segment size and disable
271 * pmtu discovery on this transport.
273 t
->pathmtu
= SCTP_DEFAULT_MINSEGMENT
;
274 t
->param_flags
= (t
->param_flags
& ~SPP_PMTUD
) |
280 dst
= sctp_transport_dst_check(t
);
282 dst
->ops
->update_pmtu(dst
, pmtu
);
285 /* Caches the dst entry and source address for a transport's destination
288 void sctp_transport_route(struct sctp_transport
*transport
,
289 union sctp_addr
*saddr
, struct sctp_sock
*opt
)
291 struct sctp_association
*asoc
= transport
->asoc
;
292 struct sctp_af
*af
= transport
->af_specific
;
293 union sctp_addr
*daddr
= &transport
->ipaddr
;
294 struct dst_entry
*dst
;
296 dst
= af
->get_dst(asoc
, daddr
, saddr
);
299 memcpy(&transport
->saddr
, saddr
, sizeof(union sctp_addr
));
301 af
->get_saddr(asoc
, dst
, daddr
, &transport
->saddr
);
303 transport
->dst
= dst
;
304 if ((transport
->param_flags
& SPP_PMTUD_DISABLE
) && transport
->pathmtu
) {
308 transport
->pathmtu
= dst_mtu(dst
);
310 /* Initialize sk->sk_rcv_saddr, if the transport is the
311 * association's active path for getsockname().
313 if (asoc
&& (transport
== asoc
->peer
.active_path
))
314 opt
->pf
->af
->to_sk_saddr(&transport
->saddr
,
317 transport
->pathmtu
= SCTP_DEFAULT_MAXSEGMENT
;
320 /* Hold a reference to a transport. */
321 void sctp_transport_hold(struct sctp_transport
*transport
)
323 atomic_inc(&transport
->refcnt
);
326 /* Release a reference to a transport and clean up
327 * if there are no more references.
329 void sctp_transport_put(struct sctp_transport
*transport
)
331 if (atomic_dec_and_test(&transport
->refcnt
))
332 sctp_transport_destroy(transport
);
335 /* Update transport's RTO based on the newly calculated RTT. */
336 void sctp_transport_update_rto(struct sctp_transport
*tp
, __u32 rtt
)
338 /* Check for valid transport. */
339 SCTP_ASSERT(tp
, "NULL transport", return);
341 /* We should not be doing any RTO updates unless rto_pending is set. */
342 SCTP_ASSERT(tp
->rto_pending
, "rto_pending not set", return);
344 if (tp
->rttvar
|| tp
->srtt
) {
345 /* 6.3.1 C3) When a new RTT measurement R' is made, set
346 * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
347 * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
350 /* Note: The above algorithm has been rewritten to
351 * express rto_beta and rto_alpha as inverse powers
353 * For example, assuming the default value of RTO.Alpha of
354 * 1/8, rto_alpha would be expressed as 3.
356 tp
->rttvar
= tp
->rttvar
- (tp
->rttvar
>> sctp_rto_beta
)
357 + ((abs(tp
->srtt
- rtt
)) >> sctp_rto_beta
);
358 tp
->srtt
= tp
->srtt
- (tp
->srtt
>> sctp_rto_alpha
)
359 + (rtt
>> sctp_rto_alpha
);
361 /* 6.3.1 C2) When the first RTT measurement R is made, set
362 * SRTT <- R, RTTVAR <- R/2.
365 tp
->rttvar
= rtt
>> 1;
368 /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
369 * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
372 tp
->rttvar
= SCTP_CLOCK_GRANULARITY
;
374 /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
375 tp
->rto
= tp
->srtt
+ (tp
->rttvar
<< 2);
377 /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
378 * seconds then it is rounded up to RTO.Min seconds.
380 if (tp
->rto
< tp
->asoc
->rto_min
)
381 tp
->rto
= tp
->asoc
->rto_min
;
383 /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
384 * at least RTO.max seconds.
386 if (tp
->rto
> tp
->asoc
->rto_max
)
387 tp
->rto
= tp
->asoc
->rto_max
;
391 /* Reset rto_pending so that a new RTT measurement is started when a
392 * new data chunk is sent.
396 SCTP_DEBUG_PRINTK("%s: transport: %p, rtt: %d, srtt: %d "
397 "rttvar: %d, rto: %ld\n", __FUNCTION__
,
398 tp
, rtt
, tp
->srtt
, tp
->rttvar
, tp
->rto
);
401 /* This routine updates the transport's cwnd and partial_bytes_acked
402 * parameters based on the bytes acked in the received SACK.
404 void sctp_transport_raise_cwnd(struct sctp_transport
*transport
,
405 __u32 sack_ctsn
, __u32 bytes_acked
)
407 __u32 cwnd
, ssthresh
, flight_size
, pba
, pmtu
;
409 cwnd
= transport
->cwnd
;
410 flight_size
= transport
->flight_size
;
412 /* The appropriate cwnd increase algorithm is performed if, and only
413 * if the cumulative TSN has advanced and the congestion window is
414 * being fully utilized.
416 if ((transport
->asoc
->ctsn_ack_point
>= sack_ctsn
) ||
417 (flight_size
< cwnd
))
420 ssthresh
= transport
->ssthresh
;
421 pba
= transport
->partial_bytes_acked
;
422 pmtu
= transport
->asoc
->pathmtu
;
424 if (cwnd
<= ssthresh
) {
425 /* RFC 2960 7.2.1, sctpimpguide-05 2.14.2 When cwnd is less
426 * than or equal to ssthresh an SCTP endpoint MUST use the
427 * slow start algorithm to increase cwnd only if the current
428 * congestion window is being fully utilized and an incoming
429 * SACK advances the Cumulative TSN Ack Point. Only when these
430 * two conditions are met can the cwnd be increased otherwise
431 * the cwnd MUST not be increased. If these conditions are met
432 * then cwnd MUST be increased by at most the lesser of
433 * 1) the total size of the previously outstanding DATA
434 * chunk(s) acknowledged, and 2) the destination's path MTU.
436 if (bytes_acked
> pmtu
)
440 SCTP_DEBUG_PRINTK("%s: SLOW START: transport: %p, "
441 "bytes_acked: %d, cwnd: %d, ssthresh: %d, "
442 "flight_size: %d, pba: %d\n",
444 transport
, bytes_acked
, cwnd
,
445 ssthresh
, flight_size
, pba
);
447 /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
448 * upon each SACK arrival that advances the Cumulative TSN Ack
449 * Point, increase partial_bytes_acked by the total number of
450 * bytes of all new chunks acknowledged in that SACK including
451 * chunks acknowledged by the new Cumulative TSN Ack and by
454 * When partial_bytes_acked is equal to or greater than cwnd
455 * and before the arrival of the SACK the sender had cwnd or
456 * more bytes of data outstanding (i.e., before arrival of the
457 * SACK, flightsize was greater than or equal to cwnd),
458 * increase cwnd by MTU, and reset partial_bytes_acked to
459 * (partial_bytes_acked - cwnd).
464 pba
= ((cwnd
< pba
) ? (pba
- cwnd
) : 0);
466 SCTP_DEBUG_PRINTK("%s: CONGESTION AVOIDANCE: "
467 "transport: %p, bytes_acked: %d, cwnd: %d, "
468 "ssthresh: %d, flight_size: %d, pba: %d\n",
470 transport
, bytes_acked
, cwnd
,
471 ssthresh
, flight_size
, pba
);
474 transport
->cwnd
= cwnd
;
475 transport
->partial_bytes_acked
= pba
;
478 /* This routine is used to lower the transport's cwnd when congestion is
481 void sctp_transport_lower_cwnd(struct sctp_transport
*transport
,
482 sctp_lower_cwnd_t reason
)
485 case SCTP_LOWER_CWND_T3_RTX
:
486 /* RFC 2960 Section 7.2.3, sctpimpguide
487 * When the T3-rtx timer expires on an address, SCTP should
488 * perform slow start by:
489 * ssthresh = max(cwnd/2, 4*MTU)
491 * partial_bytes_acked = 0
493 transport
->ssthresh
= max(transport
->cwnd
/2,
494 4*transport
->asoc
->pathmtu
);
495 transport
->cwnd
= transport
->asoc
->pathmtu
;
498 case SCTP_LOWER_CWND_FAST_RTX
:
499 /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
500 * destination address(es) to which the missing DATA chunks
501 * were last sent, according to the formula described in
504 * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
505 * losses from SACK (see Section 7.2.4), An endpoint
506 * should do the following:
507 * ssthresh = max(cwnd/2, 4*MTU)
509 * partial_bytes_acked = 0
511 transport
->ssthresh
= max(transport
->cwnd
/2,
512 4*transport
->asoc
->pathmtu
);
513 transport
->cwnd
= transport
->ssthresh
;
516 case SCTP_LOWER_CWND_ECNE
:
517 /* RFC 2481 Section 6.1.2.
518 * If the sender receives an ECN-Echo ACK packet
519 * then the sender knows that congestion was encountered in the
520 * network on the path from the sender to the receiver. The
521 * indication of congestion should be treated just as a
522 * congestion loss in non-ECN Capable TCP. That is, the TCP
523 * source halves the congestion window "cwnd" and reduces the
524 * slow start threshold "ssthresh".
525 * A critical condition is that TCP does not react to
526 * congestion indications more than once every window of
527 * data (or more loosely more than once every round-trip time).
529 if ((jiffies
- transport
->last_time_ecne_reduced
) >
531 transport
->ssthresh
= max(transport
->cwnd
/2,
532 4*transport
->asoc
->pathmtu
);
533 transport
->cwnd
= transport
->ssthresh
;
534 transport
->last_time_ecne_reduced
= jiffies
;
538 case SCTP_LOWER_CWND_INACTIVE
:
539 /* RFC 2960 Section 7.2.1, sctpimpguide
540 * When the endpoint does not transmit data on a given
541 * transport address, the cwnd of the transport address
542 * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
543 * NOTE: Although the draft recommends that this check needs
544 * to be done every RTO interval, we do it every hearbeat
547 if ((jiffies
- transport
->last_time_used
) > transport
->rto
)
548 transport
->cwnd
= max(transport
->cwnd
/2,
549 4*transport
->asoc
->pathmtu
);
553 transport
->partial_bytes_acked
= 0;
554 SCTP_DEBUG_PRINTK("%s: transport: %p reason: %d cwnd: "
555 "%d ssthresh: %d\n", __FUNCTION__
,
557 transport
->cwnd
, transport
->ssthresh
);
560 /* What is the next timeout value for this transport? */
561 unsigned long sctp_transport_timeout(struct sctp_transport
*t
)
563 unsigned long timeout
;
564 timeout
= t
->rto
+ sctp_jitter(t
->rto
);
565 if (t
->state
!= SCTP_UNCONFIRMED
)
566 timeout
+= t
->hbinterval
;
571 /* Reset transport variables to their initial values */
572 void sctp_transport_reset(struct sctp_transport
*t
)
574 struct sctp_association
*asoc
= t
->asoc
;
576 /* RFC 2960 (bis), Section 5.2.4
577 * All the congestion control parameters (e.g., cwnd, ssthresh)
578 * related to this peer MUST be reset to their initial values
579 * (see Section 6.2.1)
581 t
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
, 2*asoc
->pathmtu
, 4380));
582 t
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
583 t
->rto
= asoc
->rto_initial
;
588 /* Reset these additional varibles so that we have a clean
591 t
->partial_bytes_acked
= 0;
596 /* Initialize the state information for SFR-CACC */
597 t
->cacc
.changeover_active
= 0;
598 t
->cacc
.cycling_changeover
= 0;
599 t
->cacc
.next_tsn_at_change
= 0;
600 t
->cacc
.cacc_saw_newack
= 0;