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 <net/sctp/sctp.h>
53 #include <net/sctp/sm.h>
55 /* 1st Level Abstractions. */
57 /* Initialize a new transport from provided memory. */
58 static struct sctp_transport
*sctp_transport_init(struct sctp_transport
*peer
,
59 const union sctp_addr
*addr
,
62 /* Copy in the address. */
64 peer
->af_specific
= sctp_get_af_specific(addr
->sa
.sa_family
);
68 memset(&peer
->saddr
, 0, sizeof(union sctp_addr
));
70 /* From 6.3.1 RTO Calculation:
72 * C1) Until an RTT measurement has been made for a packet sent to the
73 * given destination transport address, set RTO to the protocol
74 * parameter 'RTO.Initial'.
77 peer
->rto
= sctp_rto_initial
;
80 peer
->rto_pending
= 0;
82 peer
->last_time_heard
= jiffies
;
83 peer
->last_time_used
= jiffies
;
84 peer
->last_time_ecne_reduced
= jiffies
;
86 peer
->init_sent_count
= 0;
88 peer
->state
= SCTP_ACTIVE
;
91 /* Initialize the default path max_retrans. */
92 peer
->max_retrans
= sctp_max_retrans_path
;
93 peer
->error_count
= 0;
95 INIT_LIST_HEAD(&peer
->transmitted
);
96 INIT_LIST_HEAD(&peer
->send_ready
);
97 INIT_LIST_HEAD(&peer
->transports
);
99 /* Set up the retransmission timer. */
100 init_timer(&peer
->T3_rtx_timer
);
101 peer
->T3_rtx_timer
.function
= sctp_generate_t3_rtx_event
;
102 peer
->T3_rtx_timer
.data
= (unsigned long)peer
;
104 /* Set up the heartbeat timer. */
105 init_timer(&peer
->hb_timer
);
106 peer
->hb_timer
.function
= sctp_generate_heartbeat_event
;
107 peer
->hb_timer
.data
= (unsigned long)peer
;
109 atomic_set(&peer
->refcnt
, 1);
114 /* Initialize the state information for SFR-CACC */
115 peer
->cacc
.changeover_active
= 0;
116 peer
->cacc
.cycling_changeover
= 0;
117 peer
->cacc
.next_tsn_at_change
= 0;
118 peer
->cacc
.cacc_saw_newack
= 0;
123 /* Allocate and initialize a new transport. */
124 struct sctp_transport
*sctp_transport_new(const union sctp_addr
*addr
, int gfp
)
126 struct sctp_transport
*transport
;
128 transport
= t_new(struct sctp_transport
, gfp
);
132 if (!sctp_transport_init(transport
, addr
, gfp
))
135 transport
->malloced
= 1;
136 SCTP_DBG_OBJCNT_INC(transport
);
147 /* This transport is no longer needed. Free up if possible, or
148 * delay until it last reference count.
150 void sctp_transport_free(struct sctp_transport
*transport
)
154 /* Try to delete the heartbeat timer. */
155 if (del_timer(&transport
->hb_timer
))
156 sctp_transport_put(transport
);
158 /* Delete the T3_rtx timer if it's active.
159 * There is no point in not doing this now and letting
160 * structure hang around in memory since we know
161 * the tranport is going away.
163 if (timer_pending(&transport
->T3_rtx_timer
) &&
164 del_timer(&transport
->T3_rtx_timer
))
165 sctp_transport_put(transport
);
168 sctp_transport_put(transport
);
171 /* Destroy the transport data structure.
172 * Assumes there are no more users of this structure.
174 static void sctp_transport_destroy(struct sctp_transport
*transport
)
176 SCTP_ASSERT(transport
->dead
, "Transport is not dead", return);
179 sctp_association_put(transport
->asoc
);
181 sctp_packet_free(&transport
->packet
);
183 dst_release(transport
->dst
);
185 SCTP_DBG_OBJCNT_DEC(transport
);
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.
191 void sctp_transport_reset_timers(struct sctp_transport
*transport
)
193 /* RFC 2960 6.3.2 Retransmission Timer Rules
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
201 if (!timer_pending(&transport
->T3_rtx_timer
))
202 if (!mod_timer(&transport
->T3_rtx_timer
,
203 jiffies
+ transport
->rto
))
204 sctp_transport_hold(transport
);
206 /* When a data chunk is sent, reset the heartbeat interval. */
207 if (!mod_timer(&transport
->hb_timer
,
208 sctp_transport_timeout(transport
)))
209 sctp_transport_hold(transport
);
212 /* This transport has been assigned to an association.
213 * Initialize fields from the association or from the sock itself.
214 * Register the reference count in the association.
216 void sctp_transport_set_owner(struct sctp_transport
*transport
,
217 struct sctp_association
*asoc
)
219 transport
->asoc
= asoc
;
220 sctp_association_hold(asoc
);
223 /* Initialize the pmtu of a transport. */
224 void sctp_transport_pmtu(struct sctp_transport
*transport
)
226 struct dst_entry
*dst
;
228 dst
= transport
->af_specific
->get_dst(NULL
, &transport
->ipaddr
, NULL
);
231 transport
->pmtu
= dst_mtu(dst
);
234 transport
->pmtu
= SCTP_DEFAULT_MAXSEGMENT
;
237 /* Caches the dst entry and source address for a transport's destination
240 void sctp_transport_route(struct sctp_transport
*transport
,
241 union sctp_addr
*saddr
, struct sctp_sock
*opt
)
243 struct sctp_association
*asoc
= transport
->asoc
;
244 struct sctp_af
*af
= transport
->af_specific
;
245 union sctp_addr
*daddr
= &transport
->ipaddr
;
246 struct dst_entry
*dst
;
248 dst
= af
->get_dst(asoc
, daddr
, saddr
);
251 memcpy(&transport
->saddr
, saddr
, sizeof(union sctp_addr
));
253 af
->get_saddr(asoc
, dst
, daddr
, &transport
->saddr
);
255 transport
->dst
= dst
;
257 transport
->pmtu
= dst_mtu(dst
);
259 /* Initialize sk->sk_rcv_saddr, if the transport is the
260 * association's active path for getsockname().
262 if (asoc
&& (transport
== asoc
->peer
.active_path
))
263 af
->to_sk_saddr(&transport
->saddr
, asoc
->base
.sk
);
265 transport
->pmtu
= SCTP_DEFAULT_MAXSEGMENT
;
268 /* Hold a reference to a transport. */
269 void sctp_transport_hold(struct sctp_transport
*transport
)
271 atomic_inc(&transport
->refcnt
);
274 /* Release a reference to a transport and clean up
275 * if there are no more references.
277 void sctp_transport_put(struct sctp_transport
*transport
)
279 if (atomic_dec_and_test(&transport
->refcnt
))
280 sctp_transport_destroy(transport
);
283 /* Update transport's RTO based on the newly calculated RTT. */
284 void sctp_transport_update_rto(struct sctp_transport
*tp
, __u32 rtt
)
286 /* Check for valid transport. */
287 SCTP_ASSERT(tp
, "NULL transport", return);
289 /* We should not be doing any RTO updates unless rto_pending is set. */
290 SCTP_ASSERT(tp
->rto_pending
, "rto_pending not set", return);
292 if (tp
->rttvar
|| tp
->srtt
) {
293 /* 6.3.1 C3) When a new RTT measurement R' is made, set
294 * RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
295 * SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
298 /* Note: The above algorithm has been rewritten to
299 * express rto_beta and rto_alpha as inverse powers
301 * For example, assuming the default value of RTO.Alpha of
302 * 1/8, rto_alpha would be expressed as 3.
304 tp
->rttvar
= tp
->rttvar
- (tp
->rttvar
>> sctp_rto_beta
)
305 + ((abs(tp
->srtt
- rtt
)) >> sctp_rto_beta
);
306 tp
->srtt
= tp
->srtt
- (tp
->srtt
>> sctp_rto_alpha
)
307 + (rtt
>> sctp_rto_alpha
);
309 /* 6.3.1 C2) When the first RTT measurement R is made, set
310 * SRTT <- R, RTTVAR <- R/2.
313 tp
->rttvar
= rtt
>> 1;
316 /* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
317 * adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
320 tp
->rttvar
= SCTP_CLOCK_GRANULARITY
;
322 /* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
323 tp
->rto
= tp
->srtt
+ (tp
->rttvar
<< 2);
325 /* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
326 * seconds then it is rounded up to RTO.Min seconds.
328 if (tp
->rto
< tp
->asoc
->rto_min
)
329 tp
->rto
= tp
->asoc
->rto_min
;
331 /* 6.3.1 C7) A maximum value may be placed on RTO provided it is
332 * at least RTO.max seconds.
334 if (tp
->rto
> tp
->asoc
->rto_max
)
335 tp
->rto
= tp
->asoc
->rto_max
;
339 /* Reset rto_pending so that a new RTT measurement is started when a
340 * new data chunk is sent.
344 SCTP_DEBUG_PRINTK("%s: transport: %p, rtt: %d, srtt: %d "
345 "rttvar: %d, rto: %d\n", __FUNCTION__
,
346 tp
, rtt
, tp
->srtt
, tp
->rttvar
, tp
->rto
);
349 /* This routine updates the transport's cwnd and partial_bytes_acked
350 * parameters based on the bytes acked in the received SACK.
352 void sctp_transport_raise_cwnd(struct sctp_transport
*transport
,
353 __u32 sack_ctsn
, __u32 bytes_acked
)
355 __u32 cwnd
, ssthresh
, flight_size
, pba
, pmtu
;
357 cwnd
= transport
->cwnd
;
358 flight_size
= transport
->flight_size
;
360 /* The appropriate cwnd increase algorithm is performed if, and only
361 * if the cumulative TSN has advanced and the congestion window is
362 * being fully utilized.
364 if ((transport
->asoc
->ctsn_ack_point
>= sack_ctsn
) ||
365 (flight_size
< cwnd
))
368 ssthresh
= transport
->ssthresh
;
369 pba
= transport
->partial_bytes_acked
;
370 pmtu
= transport
->asoc
->pmtu
;
372 if (cwnd
<= ssthresh
) {
373 /* RFC 2960 7.2.1, sctpimpguide-05 2.14.2 When cwnd is less
374 * than or equal to ssthresh an SCTP endpoint MUST use the
375 * slow start algorithm to increase cwnd only if the current
376 * congestion window is being fully utilized and an incoming
377 * SACK advances the Cumulative TSN Ack Point. Only when these
378 * two conditions are met can the cwnd be increased otherwise
379 * the cwnd MUST not be increased. If these conditions are met
380 * then cwnd MUST be increased by at most the lesser of
381 * 1) the total size of the previously outstanding DATA
382 * chunk(s) acknowledged, and 2) the destination's path MTU.
384 if (bytes_acked
> pmtu
)
388 SCTP_DEBUG_PRINTK("%s: SLOW START: transport: %p, "
389 "bytes_acked: %d, cwnd: %d, ssthresh: %d, "
390 "flight_size: %d, pba: %d\n",
392 transport
, bytes_acked
, cwnd
,
393 ssthresh
, flight_size
, pba
);
395 /* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
396 * upon each SACK arrival that advances the Cumulative TSN Ack
397 * Point, increase partial_bytes_acked by the total number of
398 * bytes of all new chunks acknowledged in that SACK including
399 * chunks acknowledged by the new Cumulative TSN Ack and by
402 * When partial_bytes_acked is equal to or greater than cwnd
403 * and before the arrival of the SACK the sender had cwnd or
404 * more bytes of data outstanding (i.e., before arrival of the
405 * SACK, flightsize was greater than or equal to cwnd),
406 * increase cwnd by MTU, and reset partial_bytes_acked to
407 * (partial_bytes_acked - cwnd).
412 pba
= ((cwnd
< pba
) ? (pba
- cwnd
) : 0);
414 SCTP_DEBUG_PRINTK("%s: CONGESTION AVOIDANCE: "
415 "transport: %p, bytes_acked: %d, cwnd: %d, "
416 "ssthresh: %d, flight_size: %d, pba: %d\n",
418 transport
, bytes_acked
, cwnd
,
419 ssthresh
, flight_size
, pba
);
422 transport
->cwnd
= cwnd
;
423 transport
->partial_bytes_acked
= pba
;
426 /* This routine is used to lower the transport's cwnd when congestion is
429 void sctp_transport_lower_cwnd(struct sctp_transport
*transport
,
430 sctp_lower_cwnd_t reason
)
433 case SCTP_LOWER_CWND_T3_RTX
:
434 /* RFC 2960 Section 7.2.3, sctpimpguide
435 * When the T3-rtx timer expires on an address, SCTP should
436 * perform slow start by:
437 * ssthresh = max(cwnd/2, 4*MTU)
439 * partial_bytes_acked = 0
441 transport
->ssthresh
= max(transport
->cwnd
/2,
442 4*transport
->asoc
->pmtu
);
443 transport
->cwnd
= transport
->asoc
->pmtu
;
446 case SCTP_LOWER_CWND_FAST_RTX
:
447 /* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
448 * destination address(es) to which the missing DATA chunks
449 * were last sent, according to the formula described in
452 * RFC 2960 7.2.3, sctpimpguide Upon detection of packet
453 * losses from SACK (see Section 7.2.4), An endpoint
454 * should do the following:
455 * ssthresh = max(cwnd/2, 4*MTU)
457 * partial_bytes_acked = 0
459 transport
->ssthresh
= max(transport
->cwnd
/2,
460 4*transport
->asoc
->pmtu
);
461 transport
->cwnd
= transport
->ssthresh
;
464 case SCTP_LOWER_CWND_ECNE
:
465 /* RFC 2481 Section 6.1.2.
466 * If the sender receives an ECN-Echo ACK packet
467 * then the sender knows that congestion was encountered in the
468 * network on the path from the sender to the receiver. The
469 * indication of congestion should be treated just as a
470 * congestion loss in non-ECN Capable TCP. That is, the TCP
471 * source halves the congestion window "cwnd" and reduces the
472 * slow start threshold "ssthresh".
473 * A critical condition is that TCP does not react to
474 * congestion indications more than once every window of
475 * data (or more loosely more than once every round-trip time).
477 if ((jiffies
- transport
->last_time_ecne_reduced
) >
479 transport
->ssthresh
= max(transport
->cwnd
/2,
480 4*transport
->asoc
->pmtu
);
481 transport
->cwnd
= transport
->ssthresh
;
482 transport
->last_time_ecne_reduced
= jiffies
;
486 case SCTP_LOWER_CWND_INACTIVE
:
487 /* RFC 2960 Section 7.2.1, sctpimpguide
488 * When the endpoint does not transmit data on a given
489 * transport address, the cwnd of the transport address
490 * should be adjusted to max(cwnd/2, 4*MTU) per RTO.
491 * NOTE: Although the draft recommends that this check needs
492 * to be done every RTO interval, we do it every hearbeat
495 if ((jiffies
- transport
->last_time_used
) > transport
->rto
)
496 transport
->cwnd
= max(transport
->cwnd
/2,
497 4*transport
->asoc
->pmtu
);
501 transport
->partial_bytes_acked
= 0;
502 SCTP_DEBUG_PRINTK("%s: transport: %p reason: %d cwnd: "
503 "%d ssthresh: %d\n", __FUNCTION__
,
505 transport
->cwnd
, transport
->ssthresh
);
508 /* What is the next timeout value for this transport? */
509 unsigned long sctp_transport_timeout(struct sctp_transport
*t
)
511 unsigned long timeout
;
512 timeout
= t
->hb_interval
+ t
->rto
+ sctp_jitter(t
->rto
);