1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
7 * This file is part of the SCTP kernel implementation
9 * These functions implement the sctp_outq class. The outqueue handles
10 * bundling and queueing of outgoing SCTP chunks.
12 * This SCTP implementation is free software;
13 * you can redistribute it and/or modify it under the terms of
14 * the GNU General Public License as published by
15 * the Free Software Foundation; either version 2, or (at your option)
18 * This SCTP implementation is distributed in the hope that it
19 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
20 * ************************
21 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
22 * See the GNU General Public License for more details.
24 * You should have received a copy of the GNU General Public License
25 * along with GNU CC; see the file COPYING. If not, see
26 * <http://www.gnu.org/licenses/>.
28 * Please send any bug reports or fixes you make to the
30 * lksctp developers <linux-sctp@vger.kernel.org>
32 * Written or modified by:
33 * La Monte H.P. Yarroll <piggy@acm.org>
34 * Karl Knutson <karl@athena.chicago.il.us>
35 * Perry Melange <pmelange@null.cc.uic.edu>
36 * Xingang Guo <xingang.guo@intel.com>
37 * Hui Huang <hui.huang@nokia.com>
38 * Sridhar Samudrala <sri@us.ibm.com>
39 * Jon Grimm <jgrimm@us.ibm.com>
42 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
44 #include <linux/types.h>
45 #include <linux/list.h> /* For struct list_head */
46 #include <linux/socket.h>
48 #include <linux/slab.h>
49 #include <net/sock.h> /* For skb_set_owner_w */
51 #include <net/sctp/sctp.h>
52 #include <net/sctp/sm.h>
54 /* Declare internal functions here. */
55 static int sctp_acked(struct sctp_sackhdr
*sack
, __u32 tsn
);
56 static void sctp_check_transmitted(struct sctp_outq
*q
,
57 struct list_head
*transmitted_queue
,
58 struct sctp_transport
*transport
,
59 union sctp_addr
*saddr
,
60 struct sctp_sackhdr
*sack
,
61 __u32
*highest_new_tsn
);
63 static void sctp_mark_missing(struct sctp_outq
*q
,
64 struct list_head
*transmitted_queue
,
65 struct sctp_transport
*transport
,
66 __u32 highest_new_tsn
,
67 int count_of_newacks
);
69 static void sctp_generate_fwdtsn(struct sctp_outq
*q
, __u32 sack_ctsn
);
71 static int sctp_outq_flush(struct sctp_outq
*q
, int rtx_timeout
);
73 /* Add data to the front of the queue. */
74 static inline void sctp_outq_head_data(struct sctp_outq
*q
,
75 struct sctp_chunk
*ch
)
77 list_add(&ch
->list
, &q
->out_chunk_list
);
78 q
->out_qlen
+= ch
->skb
->len
;
81 /* Take data from the front of the queue. */
82 static inline struct sctp_chunk
*sctp_outq_dequeue_data(struct sctp_outq
*q
)
84 struct sctp_chunk
*ch
= NULL
;
86 if (!list_empty(&q
->out_chunk_list
)) {
87 struct list_head
*entry
= q
->out_chunk_list
.next
;
89 ch
= list_entry(entry
, struct sctp_chunk
, list
);
91 q
->out_qlen
-= ch
->skb
->len
;
95 /* Add data chunk to the end of the queue. */
96 static inline void sctp_outq_tail_data(struct sctp_outq
*q
,
97 struct sctp_chunk
*ch
)
99 list_add_tail(&ch
->list
, &q
->out_chunk_list
);
100 q
->out_qlen
+= ch
->skb
->len
;
104 * SFR-CACC algorithm:
105 * D) If count_of_newacks is greater than or equal to 2
106 * and t was not sent to the current primary then the
107 * sender MUST NOT increment missing report count for t.
109 static inline int sctp_cacc_skip_3_1_d(struct sctp_transport
*primary
,
110 struct sctp_transport
*transport
,
111 int count_of_newacks
)
113 if (count_of_newacks
>= 2 && transport
!= primary
)
119 * SFR-CACC algorithm:
120 * F) If count_of_newacks is less than 2, let d be the
121 * destination to which t was sent. If cacc_saw_newack
122 * is 0 for destination d, then the sender MUST NOT
123 * increment missing report count for t.
125 static inline int sctp_cacc_skip_3_1_f(struct sctp_transport
*transport
,
126 int count_of_newacks
)
128 if (count_of_newacks
< 2 &&
129 (transport
&& !transport
->cacc
.cacc_saw_newack
))
135 * SFR-CACC algorithm:
136 * 3.1) If CYCLING_CHANGEOVER is 0, the sender SHOULD
137 * execute steps C, D, F.
139 * C has been implemented in sctp_outq_sack
141 static inline int sctp_cacc_skip_3_1(struct sctp_transport
*primary
,
142 struct sctp_transport
*transport
,
143 int count_of_newacks
)
145 if (!primary
->cacc
.cycling_changeover
) {
146 if (sctp_cacc_skip_3_1_d(primary
, transport
, count_of_newacks
))
148 if (sctp_cacc_skip_3_1_f(transport
, count_of_newacks
))
156 * SFR-CACC algorithm:
157 * 3.2) Else if CYCLING_CHANGEOVER is 1, and t is less
158 * than next_tsn_at_change of the current primary, then
159 * the sender MUST NOT increment missing report count
162 static inline int sctp_cacc_skip_3_2(struct sctp_transport
*primary
, __u32 tsn
)
164 if (primary
->cacc
.cycling_changeover
&&
165 TSN_lt(tsn
, primary
->cacc
.next_tsn_at_change
))
171 * SFR-CACC algorithm:
172 * 3) If the missing report count for TSN t is to be
173 * incremented according to [RFC2960] and
174 * [SCTP_STEWART-2002], and CHANGEOVER_ACTIVE is set,
175 * then the sender MUST further execute steps 3.1 and
176 * 3.2 to determine if the missing report count for
177 * TSN t SHOULD NOT be incremented.
179 * 3.3) If 3.1 and 3.2 do not dictate that the missing
180 * report count for t should not be incremented, then
181 * the sender SHOULD increment missing report count for
182 * t (according to [RFC2960] and [SCTP_STEWART_2002]).
184 static inline int sctp_cacc_skip(struct sctp_transport
*primary
,
185 struct sctp_transport
*transport
,
186 int count_of_newacks
,
189 if (primary
->cacc
.changeover_active
&&
190 (sctp_cacc_skip_3_1(primary
, transport
, count_of_newacks
) ||
191 sctp_cacc_skip_3_2(primary
, tsn
)))
196 /* Initialize an existing sctp_outq. This does the boring stuff.
197 * You still need to define handlers if you really want to DO
198 * something with this structure...
200 void sctp_outq_init(struct sctp_association
*asoc
, struct sctp_outq
*q
)
202 memset(q
, 0, sizeof(struct sctp_outq
));
205 INIT_LIST_HEAD(&q
->out_chunk_list
);
206 INIT_LIST_HEAD(&q
->control_chunk_list
);
207 INIT_LIST_HEAD(&q
->retransmit
);
208 INIT_LIST_HEAD(&q
->sacked
);
209 INIT_LIST_HEAD(&q
->abandoned
);
212 /* Free the outqueue structure and any related pending chunks.
214 static void __sctp_outq_teardown(struct sctp_outq
*q
)
216 struct sctp_transport
*transport
;
217 struct list_head
*lchunk
, *temp
;
218 struct sctp_chunk
*chunk
, *tmp
;
220 /* Throw away unacknowledged chunks. */
221 list_for_each_entry(transport
, &q
->asoc
->peer
.transport_addr_list
,
223 while ((lchunk
= sctp_list_dequeue(&transport
->transmitted
)) != NULL
) {
224 chunk
= list_entry(lchunk
, struct sctp_chunk
,
226 /* Mark as part of a failed message. */
227 sctp_chunk_fail(chunk
, q
->error
);
228 sctp_chunk_free(chunk
);
232 /* Throw away chunks that have been gap ACKed. */
233 list_for_each_safe(lchunk
, temp
, &q
->sacked
) {
234 list_del_init(lchunk
);
235 chunk
= list_entry(lchunk
, struct sctp_chunk
,
237 sctp_chunk_fail(chunk
, q
->error
);
238 sctp_chunk_free(chunk
);
241 /* Throw away any chunks in the retransmit queue. */
242 list_for_each_safe(lchunk
, temp
, &q
->retransmit
) {
243 list_del_init(lchunk
);
244 chunk
= list_entry(lchunk
, struct sctp_chunk
,
246 sctp_chunk_fail(chunk
, q
->error
);
247 sctp_chunk_free(chunk
);
250 /* Throw away any chunks that are in the abandoned queue. */
251 list_for_each_safe(lchunk
, temp
, &q
->abandoned
) {
252 list_del_init(lchunk
);
253 chunk
= list_entry(lchunk
, struct sctp_chunk
,
255 sctp_chunk_fail(chunk
, q
->error
);
256 sctp_chunk_free(chunk
);
259 /* Throw away any leftover data chunks. */
260 while ((chunk
= sctp_outq_dequeue_data(q
)) != NULL
) {
262 /* Mark as send failure. */
263 sctp_chunk_fail(chunk
, q
->error
);
264 sctp_chunk_free(chunk
);
267 /* Throw away any leftover control chunks. */
268 list_for_each_entry_safe(chunk
, tmp
, &q
->control_chunk_list
, list
) {
269 list_del_init(&chunk
->list
);
270 sctp_chunk_free(chunk
);
274 void sctp_outq_teardown(struct sctp_outq
*q
)
276 __sctp_outq_teardown(q
);
277 sctp_outq_init(q
->asoc
, q
);
280 /* Free the outqueue structure and any related pending chunks. */
281 void sctp_outq_free(struct sctp_outq
*q
)
283 /* Throw away leftover chunks. */
284 __sctp_outq_teardown(q
);
287 /* Put a new chunk in an sctp_outq. */
288 int sctp_outq_tail(struct sctp_outq
*q
, struct sctp_chunk
*chunk
)
290 struct net
*net
= sock_net(q
->asoc
->base
.sk
);
293 pr_debug("%s: outq:%p, chunk:%p[%s]\n", __func__
, q
, chunk
,
294 chunk
&& chunk
->chunk_hdr
?
295 sctp_cname(SCTP_ST_CHUNK(chunk
->chunk_hdr
->type
)) :
298 /* If it is data, queue it up, otherwise, send it
301 if (sctp_chunk_is_data(chunk
)) {
302 /* Is it OK to queue data chunks? */
303 /* From 9. Termination of Association
305 * When either endpoint performs a shutdown, the
306 * association on each peer will stop accepting new
307 * data from its user and only deliver data in queue
308 * at the time of sending or receiving the SHUTDOWN
311 switch (q
->asoc
->state
) {
312 case SCTP_STATE_CLOSED
:
313 case SCTP_STATE_SHUTDOWN_PENDING
:
314 case SCTP_STATE_SHUTDOWN_SENT
:
315 case SCTP_STATE_SHUTDOWN_RECEIVED
:
316 case SCTP_STATE_SHUTDOWN_ACK_SENT
:
317 /* Cannot send after transport endpoint shutdown */
322 pr_debug("%s: outqueueing: outq:%p, chunk:%p[%s])\n",
323 __func__
, q
, chunk
, chunk
&& chunk
->chunk_hdr
?
324 sctp_cname(SCTP_ST_CHUNK(chunk
->chunk_hdr
->type
)) :
327 sctp_outq_tail_data(q
, chunk
);
328 if (chunk
->chunk_hdr
->flags
& SCTP_DATA_UNORDERED
)
329 SCTP_INC_STATS(net
, SCTP_MIB_OUTUNORDERCHUNKS
);
331 SCTP_INC_STATS(net
, SCTP_MIB_OUTORDERCHUNKS
);
335 list_add_tail(&chunk
->list
, &q
->control_chunk_list
);
336 SCTP_INC_STATS(net
, SCTP_MIB_OUTCTRLCHUNKS
);
343 error
= sctp_outq_flush(q
, 0);
348 /* Insert a chunk into the sorted list based on the TSNs. The retransmit list
349 * and the abandoned list are in ascending order.
351 static void sctp_insert_list(struct list_head
*head
, struct list_head
*new)
353 struct list_head
*pos
;
354 struct sctp_chunk
*nchunk
, *lchunk
;
358 nchunk
= list_entry(new, struct sctp_chunk
, transmitted_list
);
359 ntsn
= ntohl(nchunk
->subh
.data_hdr
->tsn
);
361 list_for_each(pos
, head
) {
362 lchunk
= list_entry(pos
, struct sctp_chunk
, transmitted_list
);
363 ltsn
= ntohl(lchunk
->subh
.data_hdr
->tsn
);
364 if (TSN_lt(ntsn
, ltsn
)) {
365 list_add(new, pos
->prev
);
371 list_add_tail(new, head
);
374 /* Mark all the eligible packets on a transport for retransmission. */
375 void sctp_retransmit_mark(struct sctp_outq
*q
,
376 struct sctp_transport
*transport
,
379 struct list_head
*lchunk
, *ltemp
;
380 struct sctp_chunk
*chunk
;
382 /* Walk through the specified transmitted queue. */
383 list_for_each_safe(lchunk
, ltemp
, &transport
->transmitted
) {
384 chunk
= list_entry(lchunk
, struct sctp_chunk
,
387 /* If the chunk is abandoned, move it to abandoned list. */
388 if (sctp_chunk_abandoned(chunk
)) {
389 list_del_init(lchunk
);
390 sctp_insert_list(&q
->abandoned
, lchunk
);
392 /* If this chunk has not been previousely acked,
393 * stop considering it 'outstanding'. Our peer
394 * will most likely never see it since it will
395 * not be retransmitted
397 if (!chunk
->tsn_gap_acked
) {
398 if (chunk
->transport
)
399 chunk
->transport
->flight_size
-=
400 sctp_data_size(chunk
);
401 q
->outstanding_bytes
-= sctp_data_size(chunk
);
402 q
->asoc
->peer
.rwnd
+= sctp_data_size(chunk
);
407 /* If we are doing retransmission due to a timeout or pmtu
408 * discovery, only the chunks that are not yet acked should
409 * be added to the retransmit queue.
411 if ((reason
== SCTP_RTXR_FAST_RTX
&&
412 (chunk
->fast_retransmit
== SCTP_NEED_FRTX
)) ||
413 (reason
!= SCTP_RTXR_FAST_RTX
&& !chunk
->tsn_gap_acked
)) {
414 /* RFC 2960 6.2.1 Processing a Received SACK
416 * C) Any time a DATA chunk is marked for
417 * retransmission (via either T3-rtx timer expiration
418 * (Section 6.3.3) or via fast retransmit
419 * (Section 7.2.4)), add the data size of those
420 * chunks to the rwnd.
422 q
->asoc
->peer
.rwnd
+= sctp_data_size(chunk
);
423 q
->outstanding_bytes
-= sctp_data_size(chunk
);
424 if (chunk
->transport
)
425 transport
->flight_size
-= sctp_data_size(chunk
);
427 /* sctpimpguide-05 Section 2.8.2
428 * M5) If a T3-rtx timer expires, the
429 * 'TSN.Missing.Report' of all affected TSNs is set
432 chunk
->tsn_missing_report
= 0;
434 /* If a chunk that is being used for RTT measurement
435 * has to be retransmitted, we cannot use this chunk
436 * anymore for RTT measurements. Reset rto_pending so
437 * that a new RTT measurement is started when a new
438 * data chunk is sent.
440 if (chunk
->rtt_in_progress
) {
441 chunk
->rtt_in_progress
= 0;
442 transport
->rto_pending
= 0;
447 /* Move the chunk to the retransmit queue. The chunks
448 * on the retransmit queue are always kept in order.
450 list_del_init(lchunk
);
451 sctp_insert_list(&q
->retransmit
, lchunk
);
455 pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d, "
456 "flight_size:%d, pba:%d\n", __func__
, transport
, reason
,
457 transport
->cwnd
, transport
->ssthresh
, transport
->flight_size
,
458 transport
->partial_bytes_acked
);
461 /* Mark all the eligible packets on a transport for retransmission and force
464 void sctp_retransmit(struct sctp_outq
*q
, struct sctp_transport
*transport
,
465 sctp_retransmit_reason_t reason
)
467 struct net
*net
= sock_net(q
->asoc
->base
.sk
);
471 case SCTP_RTXR_T3_RTX
:
472 SCTP_INC_STATS(net
, SCTP_MIB_T3_RETRANSMITS
);
473 sctp_transport_lower_cwnd(transport
, SCTP_LOWER_CWND_T3_RTX
);
474 /* Update the retran path if the T3-rtx timer has expired for
475 * the current retran path.
477 if (transport
== transport
->asoc
->peer
.retran_path
)
478 sctp_assoc_update_retran_path(transport
->asoc
);
479 transport
->asoc
->rtx_data_chunks
+=
480 transport
->asoc
->unack_data
;
482 case SCTP_RTXR_FAST_RTX
:
483 SCTP_INC_STATS(net
, SCTP_MIB_FAST_RETRANSMITS
);
484 sctp_transport_lower_cwnd(transport
, SCTP_LOWER_CWND_FAST_RTX
);
487 case SCTP_RTXR_PMTUD
:
488 SCTP_INC_STATS(net
, SCTP_MIB_PMTUD_RETRANSMITS
);
490 case SCTP_RTXR_T1_RTX
:
491 SCTP_INC_STATS(net
, SCTP_MIB_T1_RETRANSMITS
);
492 transport
->asoc
->init_retries
++;
498 sctp_retransmit_mark(q
, transport
, reason
);
500 /* PR-SCTP A5) Any time the T3-rtx timer expires, on any destination,
501 * the sender SHOULD try to advance the "Advanced.Peer.Ack.Point" by
502 * following the procedures outlined in C1 - C5.
504 if (reason
== SCTP_RTXR_T3_RTX
)
505 sctp_generate_fwdtsn(q
, q
->asoc
->ctsn_ack_point
);
507 /* Flush the queues only on timeout, since fast_rtx is only
508 * triggered during sack processing and the queue
509 * will be flushed at the end.
511 if (reason
!= SCTP_RTXR_FAST_RTX
)
512 error
= sctp_outq_flush(q
, /* rtx_timeout */ 1);
515 q
->asoc
->base
.sk
->sk_err
= -error
;
519 * Transmit DATA chunks on the retransmit queue. Upon return from
520 * sctp_outq_flush_rtx() the packet 'pkt' may contain chunks which
521 * need to be transmitted by the caller.
522 * We assume that pkt->transport has already been set.
524 * The return value is a normal kernel error return value.
526 static int sctp_outq_flush_rtx(struct sctp_outq
*q
, struct sctp_packet
*pkt
,
527 int rtx_timeout
, int *start_timer
)
529 struct list_head
*lqueue
;
530 struct sctp_transport
*transport
= pkt
->transport
;
532 struct sctp_chunk
*chunk
, *chunk1
;
538 lqueue
= &q
->retransmit
;
539 fast_rtx
= q
->fast_rtx
;
541 /* This loop handles time-out retransmissions, fast retransmissions,
542 * and retransmissions due to opening of whindow.
544 * RFC 2960 6.3.3 Handle T3-rtx Expiration
546 * E3) Determine how many of the earliest (i.e., lowest TSN)
547 * outstanding DATA chunks for the address for which the
548 * T3-rtx has expired will fit into a single packet, subject
549 * to the MTU constraint for the path corresponding to the
550 * destination transport address to which the retransmission
551 * is being sent (this may be different from the address for
552 * which the timer expires [see Section 6.4]). Call this value
553 * K. Bundle and retransmit those K DATA chunks in a single
554 * packet to the destination endpoint.
556 * [Just to be painfully clear, if we are retransmitting
557 * because a timeout just happened, we should send only ONE
558 * packet of retransmitted data.]
560 * For fast retransmissions we also send only ONE packet. However,
561 * if we are just flushing the queue due to open window, we'll
562 * try to send as much as possible.
564 list_for_each_entry_safe(chunk
, chunk1
, lqueue
, transmitted_list
) {
565 /* If the chunk is abandoned, move it to abandoned list. */
566 if (sctp_chunk_abandoned(chunk
)) {
567 list_del_init(&chunk
->transmitted_list
);
568 sctp_insert_list(&q
->abandoned
,
569 &chunk
->transmitted_list
);
573 /* Make sure that Gap Acked TSNs are not retransmitted. A
574 * simple approach is just to move such TSNs out of the
575 * way and into a 'transmitted' queue and skip to the
578 if (chunk
->tsn_gap_acked
) {
579 list_move_tail(&chunk
->transmitted_list
,
580 &transport
->transmitted
);
584 /* If we are doing fast retransmit, ignore non-fast_rtransmit
587 if (fast_rtx
&& !chunk
->fast_retransmit
)
591 /* Attempt to append this chunk to the packet. */
592 status
= sctp_packet_append_chunk(pkt
, chunk
);
595 case SCTP_XMIT_PMTU_FULL
:
596 if (!pkt
->has_data
&& !pkt
->has_cookie_echo
) {
597 /* If this packet did not contain DATA then
598 * retransmission did not happen, so do it
599 * again. We'll ignore the error here since
600 * control chunks are already freed so there
601 * is nothing we can do.
603 sctp_packet_transmit(pkt
);
607 /* Send this packet. */
608 error
= sctp_packet_transmit(pkt
);
610 /* If we are retransmitting, we should only
611 * send a single packet.
612 * Otherwise, try appending this chunk again.
614 if (rtx_timeout
|| fast_rtx
)
619 /* Bundle next chunk in the next round. */
622 case SCTP_XMIT_RWND_FULL
:
623 /* Send this packet. */
624 error
= sctp_packet_transmit(pkt
);
626 /* Stop sending DATA as there is no more room
632 case SCTP_XMIT_NAGLE_DELAY
:
633 /* Send this packet. */
634 error
= sctp_packet_transmit(pkt
);
636 /* Stop sending DATA because of nagle delay. */
641 /* The append was successful, so add this chunk to
642 * the transmitted list.
644 list_move_tail(&chunk
->transmitted_list
,
645 &transport
->transmitted
);
647 /* Mark the chunk as ineligible for fast retransmit
648 * after it is retransmitted.
650 if (chunk
->fast_retransmit
== SCTP_NEED_FRTX
)
651 chunk
->fast_retransmit
= SCTP_DONT_FRTX
;
653 q
->asoc
->stats
.rtxchunks
++;
657 /* Set the timer if there were no errors */
658 if (!error
&& !timer
)
665 /* If we are here due to a retransmit timeout or a fast
666 * retransmit and if there are any chunks left in the retransmit
667 * queue that could not fit in the PMTU sized packet, they need
668 * to be marked as ineligible for a subsequent fast retransmit.
670 if (rtx_timeout
|| fast_rtx
) {
671 list_for_each_entry(chunk1
, lqueue
, transmitted_list
) {
672 if (chunk1
->fast_retransmit
== SCTP_NEED_FRTX
)
673 chunk1
->fast_retransmit
= SCTP_DONT_FRTX
;
677 *start_timer
= timer
;
679 /* Clear fast retransmit hint */
686 /* Cork the outqueue so queued chunks are really queued. */
687 int sctp_outq_uncork(struct sctp_outq
*q
)
692 return sctp_outq_flush(q
, 0);
697 * Try to flush an outqueue.
699 * Description: Send everything in q which we legally can, subject to
700 * congestion limitations.
701 * * Note: This function can be called from multiple contexts so appropriate
702 * locking concerns must be made. Today we use the sock lock to protect
705 static int sctp_outq_flush(struct sctp_outq
*q
, int rtx_timeout
)
707 struct sctp_packet
*packet
;
708 struct sctp_packet singleton
;
709 struct sctp_association
*asoc
= q
->asoc
;
710 __u16 sport
= asoc
->base
.bind_addr
.port
;
711 __u16 dport
= asoc
->peer
.port
;
712 __u32 vtag
= asoc
->peer
.i
.init_tag
;
713 struct sctp_transport
*transport
= NULL
;
714 struct sctp_transport
*new_transport
;
715 struct sctp_chunk
*chunk
, *tmp
;
721 /* These transports have chunks to send. */
722 struct list_head transport_list
;
723 struct list_head
*ltransport
;
725 INIT_LIST_HEAD(&transport_list
);
731 * When bundling control chunks with DATA chunks, an
732 * endpoint MUST place control chunks first in the outbound
733 * SCTP packet. The transmitter MUST transmit DATA chunks
734 * within a SCTP packet in increasing order of TSN.
738 list_for_each_entry_safe(chunk
, tmp
, &q
->control_chunk_list
, list
) {
740 * F1) This means that until such time as the ASCONF
741 * containing the add is acknowledged, the sender MUST
742 * NOT use the new IP address as a source for ANY SCTP
743 * packet except on carrying an ASCONF Chunk.
745 if (asoc
->src_out_of_asoc_ok
&&
746 chunk
->chunk_hdr
->type
!= SCTP_CID_ASCONF
)
749 list_del_init(&chunk
->list
);
751 /* Pick the right transport to use. */
752 new_transport
= chunk
->transport
;
754 if (!new_transport
) {
756 * If we have a prior transport pointer, see if
757 * the destination address of the chunk
758 * matches the destination address of the
759 * current transport. If not a match, then
760 * try to look up the transport with a given
761 * destination address. We do this because
762 * after processing ASCONFs, we may have new
763 * transports created.
766 sctp_cmp_addr_exact(&chunk
->dest
,
768 new_transport
= transport
;
770 new_transport
= sctp_assoc_lookup_paddr(asoc
,
773 /* if we still don't have a new transport, then
774 * use the current active path.
777 new_transport
= asoc
->peer
.active_path
;
778 } else if ((new_transport
->state
== SCTP_INACTIVE
) ||
779 (new_transport
->state
== SCTP_UNCONFIRMED
) ||
780 (new_transport
->state
== SCTP_PF
)) {
781 /* If the chunk is Heartbeat or Heartbeat Ack,
782 * send it to chunk->transport, even if it's
785 * 3.3.6 Heartbeat Acknowledgement:
787 * A HEARTBEAT ACK is always sent to the source IP
788 * address of the IP datagram containing the
789 * HEARTBEAT chunk to which this ack is responding.
792 * ASCONF_ACKs also must be sent to the source.
794 if (chunk
->chunk_hdr
->type
!= SCTP_CID_HEARTBEAT
&&
795 chunk
->chunk_hdr
->type
!= SCTP_CID_HEARTBEAT_ACK
&&
796 chunk
->chunk_hdr
->type
!= SCTP_CID_ASCONF_ACK
)
797 new_transport
= asoc
->peer
.active_path
;
800 /* Are we switching transports?
801 * Take care of transport locks.
803 if (new_transport
!= transport
) {
804 transport
= new_transport
;
805 if (list_empty(&transport
->send_ready
)) {
806 list_add_tail(&transport
->send_ready
,
809 packet
= &transport
->packet
;
810 sctp_packet_config(packet
, vtag
,
811 asoc
->peer
.ecn_capable
);
814 switch (chunk
->chunk_hdr
->type
) {
818 * An endpoint MUST NOT bundle INIT, INIT ACK or SHUTDOWN
819 * COMPLETE with any other chunks. [Send them immediately.]
822 case SCTP_CID_INIT_ACK
:
823 case SCTP_CID_SHUTDOWN_COMPLETE
:
824 sctp_packet_init(&singleton
, transport
, sport
, dport
);
825 sctp_packet_config(&singleton
, vtag
, 0);
826 sctp_packet_append_chunk(&singleton
, chunk
);
827 error
= sctp_packet_transmit(&singleton
);
833 if (sctp_test_T_bit(chunk
)) {
834 packet
->vtag
= asoc
->c
.my_vtag
;
836 /* The following chunks are "response" chunks, i.e.
837 * they are generated in response to something we
838 * received. If we are sending these, then we can
839 * send only 1 packet containing these chunks.
841 case SCTP_CID_HEARTBEAT_ACK
:
842 case SCTP_CID_SHUTDOWN_ACK
:
843 case SCTP_CID_COOKIE_ACK
:
844 case SCTP_CID_COOKIE_ECHO
:
846 case SCTP_CID_ECN_CWR
:
847 case SCTP_CID_ASCONF_ACK
:
852 case SCTP_CID_HEARTBEAT
:
853 case SCTP_CID_SHUTDOWN
:
854 case SCTP_CID_ECN_ECNE
:
855 case SCTP_CID_ASCONF
:
856 case SCTP_CID_FWD_TSN
:
857 status
= sctp_packet_transmit_chunk(packet
, chunk
,
859 if (status
!= SCTP_XMIT_OK
) {
860 /* put the chunk back */
861 list_add(&chunk
->list
, &q
->control_chunk_list
);
863 asoc
->stats
.octrlchunks
++;
864 /* PR-SCTP C5) If a FORWARD TSN is sent, the
865 * sender MUST assure that at least one T3-rtx
868 if (chunk
->chunk_hdr
->type
== SCTP_CID_FWD_TSN
)
869 sctp_transport_reset_timers(transport
);
874 /* We built a chunk with an illegal type! */
879 if (q
->asoc
->src_out_of_asoc_ok
)
882 /* Is it OK to send data chunks? */
883 switch (asoc
->state
) {
884 case SCTP_STATE_COOKIE_ECHOED
:
885 /* Only allow bundling when this packet has a COOKIE-ECHO
888 if (!packet
|| !packet
->has_cookie_echo
)
892 case SCTP_STATE_ESTABLISHED
:
893 case SCTP_STATE_SHUTDOWN_PENDING
:
894 case SCTP_STATE_SHUTDOWN_RECEIVED
:
896 * RFC 2960 6.1 Transmission of DATA Chunks
898 * C) When the time comes for the sender to transmit,
899 * before sending new DATA chunks, the sender MUST
900 * first transmit any outstanding DATA chunks which
901 * are marked for retransmission (limited by the
904 if (!list_empty(&q
->retransmit
)) {
905 if (asoc
->peer
.retran_path
->state
== SCTP_UNCONFIRMED
)
907 if (transport
== asoc
->peer
.retran_path
)
910 /* Switch transports & prepare the packet. */
912 transport
= asoc
->peer
.retran_path
;
914 if (list_empty(&transport
->send_ready
)) {
915 list_add_tail(&transport
->send_ready
,
919 packet
= &transport
->packet
;
920 sctp_packet_config(packet
, vtag
,
921 asoc
->peer
.ecn_capable
);
923 error
= sctp_outq_flush_rtx(q
, packet
,
924 rtx_timeout
, &start_timer
);
927 sctp_transport_reset_timers(transport
);
929 /* This can happen on COOKIE-ECHO resend. Only
930 * one chunk can get bundled with a COOKIE-ECHO.
932 if (packet
->has_cookie_echo
)
935 /* Don't send new data if there is still data
936 * waiting to retransmit.
938 if (!list_empty(&q
->retransmit
))
942 /* Apply Max.Burst limitation to the current transport in
943 * case it will be used for new data. We are going to
944 * rest it before we return, but we want to apply the limit
945 * to the currently queued data.
948 sctp_transport_burst_limited(transport
);
950 /* Finally, transmit new packets. */
951 while ((chunk
= sctp_outq_dequeue_data(q
)) != NULL
) {
952 /* RFC 2960 6.5 Every DATA chunk MUST carry a valid
955 if (chunk
->sinfo
.sinfo_stream
>=
956 asoc
->c
.sinit_num_ostreams
) {
958 /* Mark as failed send. */
959 sctp_chunk_fail(chunk
, SCTP_ERROR_INV_STRM
);
960 sctp_chunk_free(chunk
);
964 /* Has this chunk expired? */
965 if (sctp_chunk_abandoned(chunk
)) {
966 sctp_chunk_fail(chunk
, 0);
967 sctp_chunk_free(chunk
);
971 /* If there is a specified transport, use it.
972 * Otherwise, we want to use the active path.
974 new_transport
= chunk
->transport
;
975 if (!new_transport
||
976 ((new_transport
->state
== SCTP_INACTIVE
) ||
977 (new_transport
->state
== SCTP_UNCONFIRMED
) ||
978 (new_transport
->state
== SCTP_PF
)))
979 new_transport
= asoc
->peer
.active_path
;
980 if (new_transport
->state
== SCTP_UNCONFIRMED
)
983 /* Change packets if necessary. */
984 if (new_transport
!= transport
) {
985 transport
= new_transport
;
987 /* Schedule to have this transport's
990 if (list_empty(&transport
->send_ready
)) {
991 list_add_tail(&transport
->send_ready
,
995 packet
= &transport
->packet
;
996 sctp_packet_config(packet
, vtag
,
997 asoc
->peer
.ecn_capable
);
998 /* We've switched transports, so apply the
999 * Burst limit to the new transport.
1001 sctp_transport_burst_limited(transport
);
1004 pr_debug("%s: outq:%p, chunk:%p[%s], tx-tsn:0x%x skb->head:%p "
1006 __func__
, q
, chunk
, chunk
&& chunk
->chunk_hdr
?
1007 sctp_cname(SCTP_ST_CHUNK(chunk
->chunk_hdr
->type
)) :
1008 "illegal chunk", ntohl(chunk
->subh
.data_hdr
->tsn
),
1009 chunk
->skb
? chunk
->skb
->head
: NULL
, chunk
->skb
?
1010 atomic_read(&chunk
->skb
->users
) : -1);
1012 /* Add the chunk to the packet. */
1013 status
= sctp_packet_transmit_chunk(packet
, chunk
, 0);
1016 case SCTP_XMIT_PMTU_FULL
:
1017 case SCTP_XMIT_RWND_FULL
:
1018 case SCTP_XMIT_NAGLE_DELAY
:
1019 /* We could not append this chunk, so put
1020 * the chunk back on the output queue.
1022 pr_debug("%s: could not transmit tsn:0x%x, status:%d\n",
1023 __func__
, ntohl(chunk
->subh
.data_hdr
->tsn
),
1026 sctp_outq_head_data(q
, chunk
);
1027 goto sctp_flush_out
;
1031 /* The sender is in the SHUTDOWN-PENDING state,
1032 * The sender MAY set the I-bit in the DATA
1035 if (asoc
->state
== SCTP_STATE_SHUTDOWN_PENDING
)
1036 chunk
->chunk_hdr
->flags
|= SCTP_DATA_SACK_IMM
;
1037 if (chunk
->chunk_hdr
->flags
& SCTP_DATA_UNORDERED
)
1038 asoc
->stats
.ouodchunks
++;
1040 asoc
->stats
.oodchunks
++;
1048 /* BUG: We assume that the sctp_packet_transmit()
1049 * call below will succeed all the time and add the
1050 * chunk to the transmitted list and restart the
1052 * It is possible that the call can fail under OOM
1055 * Is this really a problem? Won't this behave
1058 list_add_tail(&chunk
->transmitted_list
,
1059 &transport
->transmitted
);
1061 sctp_transport_reset_timers(transport
);
1063 /* Only let one DATA chunk get bundled with a
1064 * COOKIE-ECHO chunk.
1066 if (packet
->has_cookie_echo
)
1067 goto sctp_flush_out
;
1078 /* Before returning, examine all the transports touched in
1079 * this call. Right now, we bluntly force clear all the
1080 * transports. Things might change after we implement Nagle.
1081 * But such an examination is still required.
1085 while ((ltransport
= sctp_list_dequeue(&transport_list
)) != NULL
) {
1086 struct sctp_transport
*t
= list_entry(ltransport
,
1087 struct sctp_transport
,
1089 packet
= &t
->packet
;
1090 if (!sctp_packet_empty(packet
))
1091 error
= sctp_packet_transmit(packet
);
1093 /* Clear the burst limited state, if any */
1094 sctp_transport_burst_reset(t
);
1100 /* Update unack_data based on the incoming SACK chunk */
1101 static void sctp_sack_update_unack_data(struct sctp_association
*assoc
,
1102 struct sctp_sackhdr
*sack
)
1104 sctp_sack_variable_t
*frags
;
1108 unack_data
= assoc
->next_tsn
- assoc
->ctsn_ack_point
- 1;
1110 frags
= sack
->variable
;
1111 for (i
= 0; i
< ntohs(sack
->num_gap_ack_blocks
); i
++) {
1112 unack_data
-= ((ntohs(frags
[i
].gab
.end
) -
1113 ntohs(frags
[i
].gab
.start
) + 1));
1116 assoc
->unack_data
= unack_data
;
1119 /* This is where we REALLY process a SACK.
1121 * Process the SACK against the outqueue. Mostly, this just frees
1122 * things off the transmitted queue.
1124 int sctp_outq_sack(struct sctp_outq
*q
, struct sctp_chunk
*chunk
)
1126 struct sctp_association
*asoc
= q
->asoc
;
1127 struct sctp_sackhdr
*sack
= chunk
->subh
.sack_hdr
;
1128 struct sctp_transport
*transport
;
1129 struct sctp_chunk
*tchunk
= NULL
;
1130 struct list_head
*lchunk
, *transport_list
, *temp
;
1131 sctp_sack_variable_t
*frags
= sack
->variable
;
1132 __u32 sack_ctsn
, ctsn
, tsn
;
1133 __u32 highest_tsn
, highest_new_tsn
;
1135 unsigned int outstanding
;
1136 struct sctp_transport
*primary
= asoc
->peer
.primary_path
;
1137 int count_of_newacks
= 0;
1141 /* Grab the association's destination address list. */
1142 transport_list
= &asoc
->peer
.transport_addr_list
;
1144 sack_ctsn
= ntohl(sack
->cum_tsn_ack
);
1145 gap_ack_blocks
= ntohs(sack
->num_gap_ack_blocks
);
1146 asoc
->stats
.gapcnt
+= gap_ack_blocks
;
1148 * SFR-CACC algorithm:
1149 * On receipt of a SACK the sender SHOULD execute the
1150 * following statements.
1152 * 1) If the cumulative ack in the SACK passes next tsn_at_change
1153 * on the current primary, the CHANGEOVER_ACTIVE flag SHOULD be
1154 * cleared. The CYCLING_CHANGEOVER flag SHOULD also be cleared for
1156 * 2) If the SACK contains gap acks and the flag CHANGEOVER_ACTIVE
1157 * is set the receiver of the SACK MUST take the following actions:
1159 * A) Initialize the cacc_saw_newack to 0 for all destination
1162 * Only bother if changeover_active is set. Otherwise, this is
1163 * totally suboptimal to do on every SACK.
1165 if (primary
->cacc
.changeover_active
) {
1166 u8 clear_cycling
= 0;
1168 if (TSN_lte(primary
->cacc
.next_tsn_at_change
, sack_ctsn
)) {
1169 primary
->cacc
.changeover_active
= 0;
1173 if (clear_cycling
|| gap_ack_blocks
) {
1174 list_for_each_entry(transport
, transport_list
,
1177 transport
->cacc
.cycling_changeover
= 0;
1179 transport
->cacc
.cacc_saw_newack
= 0;
1184 /* Get the highest TSN in the sack. */
1185 highest_tsn
= sack_ctsn
;
1187 highest_tsn
+= ntohs(frags
[gap_ack_blocks
- 1].gab
.end
);
1189 if (TSN_lt(asoc
->highest_sacked
, highest_tsn
))
1190 asoc
->highest_sacked
= highest_tsn
;
1192 highest_new_tsn
= sack_ctsn
;
1194 /* Run through the retransmit queue. Credit bytes received
1195 * and free those chunks that we can.
1197 sctp_check_transmitted(q
, &q
->retransmit
, NULL
, NULL
, sack
, &highest_new_tsn
);
1199 /* Run through the transmitted queue.
1200 * Credit bytes received and free those chunks which we can.
1202 * This is a MASSIVE candidate for optimization.
1204 list_for_each_entry(transport
, transport_list
, transports
) {
1205 sctp_check_transmitted(q
, &transport
->transmitted
,
1206 transport
, &chunk
->source
, sack
,
1209 * SFR-CACC algorithm:
1210 * C) Let count_of_newacks be the number of
1211 * destinations for which cacc_saw_newack is set.
1213 if (transport
->cacc
.cacc_saw_newack
)
1217 /* Move the Cumulative TSN Ack Point if appropriate. */
1218 if (TSN_lt(asoc
->ctsn_ack_point
, sack_ctsn
)) {
1219 asoc
->ctsn_ack_point
= sack_ctsn
;
1223 if (gap_ack_blocks
) {
1225 if (asoc
->fast_recovery
&& accum_moved
)
1226 highest_new_tsn
= highest_tsn
;
1228 list_for_each_entry(transport
, transport_list
, transports
)
1229 sctp_mark_missing(q
, &transport
->transmitted
, transport
,
1230 highest_new_tsn
, count_of_newacks
);
1233 /* Update unack_data field in the assoc. */
1234 sctp_sack_update_unack_data(asoc
, sack
);
1236 ctsn
= asoc
->ctsn_ack_point
;
1238 /* Throw away stuff rotting on the sack queue. */
1239 list_for_each_safe(lchunk
, temp
, &q
->sacked
) {
1240 tchunk
= list_entry(lchunk
, struct sctp_chunk
,
1242 tsn
= ntohl(tchunk
->subh
.data_hdr
->tsn
);
1243 if (TSN_lte(tsn
, ctsn
)) {
1244 list_del_init(&tchunk
->transmitted_list
);
1245 sctp_chunk_free(tchunk
);
1249 /* ii) Set rwnd equal to the newly received a_rwnd minus the
1250 * number of bytes still outstanding after processing the
1251 * Cumulative TSN Ack and the Gap Ack Blocks.
1254 sack_a_rwnd
= ntohl(sack
->a_rwnd
);
1255 outstanding
= q
->outstanding_bytes
;
1257 if (outstanding
< sack_a_rwnd
)
1258 sack_a_rwnd
-= outstanding
;
1262 asoc
->peer
.rwnd
= sack_a_rwnd
;
1264 sctp_generate_fwdtsn(q
, sack_ctsn
);
1266 pr_debug("%s: sack cumulative tsn ack:0x%x\n", __func__
, sack_ctsn
);
1267 pr_debug("%s: cumulative tsn ack of assoc:%p is 0x%x, "
1268 "advertised peer ack point:0x%x\n", __func__
, asoc
, ctsn
,
1269 asoc
->adv_peer_ack_point
);
1271 return sctp_outq_is_empty(q
);
1274 /* Is the outqueue empty?
1275 * The queue is empty when we have not pending data, no in-flight data
1276 * and nothing pending retransmissions.
1278 int sctp_outq_is_empty(const struct sctp_outq
*q
)
1280 return q
->out_qlen
== 0 && q
->outstanding_bytes
== 0 &&
1281 list_empty(&q
->retransmit
);
1284 /********************************************************************
1285 * 2nd Level Abstractions
1286 ********************************************************************/
1288 /* Go through a transport's transmitted list or the association's retransmit
1289 * list and move chunks that are acked by the Cumulative TSN Ack to q->sacked.
1290 * The retransmit list will not have an associated transport.
1292 * I added coherent debug information output. --xguo
1294 * Instead of printing 'sacked' or 'kept' for each TSN on the
1295 * transmitted_queue, we print a range: SACKED: TSN1-TSN2, TSN3, TSN4-TSN5.
1296 * KEPT TSN6-TSN7, etc.
1298 static void sctp_check_transmitted(struct sctp_outq
*q
,
1299 struct list_head
*transmitted_queue
,
1300 struct sctp_transport
*transport
,
1301 union sctp_addr
*saddr
,
1302 struct sctp_sackhdr
*sack
,
1303 __u32
*highest_new_tsn_in_sack
)
1305 struct list_head
*lchunk
;
1306 struct sctp_chunk
*tchunk
;
1307 struct list_head tlist
;
1311 __u8 restart_timer
= 0;
1312 int bytes_acked
= 0;
1313 int migrate_bytes
= 0;
1314 bool forward_progress
= false;
1316 sack_ctsn
= ntohl(sack
->cum_tsn_ack
);
1318 INIT_LIST_HEAD(&tlist
);
1320 /* The while loop will skip empty transmitted queues. */
1321 while (NULL
!= (lchunk
= sctp_list_dequeue(transmitted_queue
))) {
1322 tchunk
= list_entry(lchunk
, struct sctp_chunk
,
1325 if (sctp_chunk_abandoned(tchunk
)) {
1326 /* Move the chunk to abandoned list. */
1327 sctp_insert_list(&q
->abandoned
, lchunk
);
1329 /* If this chunk has not been acked, stop
1330 * considering it as 'outstanding'.
1332 if (!tchunk
->tsn_gap_acked
) {
1333 if (tchunk
->transport
)
1334 tchunk
->transport
->flight_size
-=
1335 sctp_data_size(tchunk
);
1336 q
->outstanding_bytes
-= sctp_data_size(tchunk
);
1341 tsn
= ntohl(tchunk
->subh
.data_hdr
->tsn
);
1342 if (sctp_acked(sack
, tsn
)) {
1343 /* If this queue is the retransmit queue, the
1344 * retransmit timer has already reclaimed
1345 * the outstanding bytes for this chunk, so only
1346 * count bytes associated with a transport.
1349 /* If this chunk is being used for RTT
1350 * measurement, calculate the RTT and update
1351 * the RTO using this value.
1353 * 6.3.1 C5) Karn's algorithm: RTT measurements
1354 * MUST NOT be made using packets that were
1355 * retransmitted (and thus for which it is
1356 * ambiguous whether the reply was for the
1357 * first instance of the packet or a later
1360 if (!tchunk
->tsn_gap_acked
&&
1362 tchunk
->rtt_in_progress
) {
1363 tchunk
->rtt_in_progress
= 0;
1364 rtt
= jiffies
- tchunk
->sent_at
;
1365 sctp_transport_update_rto(transport
,
1370 /* If the chunk hasn't been marked as ACKED,
1371 * mark it and account bytes_acked if the
1372 * chunk had a valid transport (it will not
1373 * have a transport if ASCONF had deleted it
1374 * while DATA was outstanding).
1376 if (!tchunk
->tsn_gap_acked
) {
1377 tchunk
->tsn_gap_acked
= 1;
1378 if (TSN_lt(*highest_new_tsn_in_sack
, tsn
))
1379 *highest_new_tsn_in_sack
= tsn
;
1380 bytes_acked
+= sctp_data_size(tchunk
);
1381 if (!tchunk
->transport
)
1382 migrate_bytes
+= sctp_data_size(tchunk
);
1383 forward_progress
= true;
1386 if (TSN_lte(tsn
, sack_ctsn
)) {
1387 /* RFC 2960 6.3.2 Retransmission Timer Rules
1389 * R3) Whenever a SACK is received
1390 * that acknowledges the DATA chunk
1391 * with the earliest outstanding TSN
1392 * for that address, restart T3-rtx
1393 * timer for that address with its
1397 forward_progress
= true;
1399 if (!tchunk
->tsn_gap_acked
) {
1401 * SFR-CACC algorithm:
1402 * 2) If the SACK contains gap acks
1403 * and the flag CHANGEOVER_ACTIVE is
1404 * set the receiver of the SACK MUST
1405 * take the following action:
1407 * B) For each TSN t being acked that
1408 * has not been acked in any SACK so
1409 * far, set cacc_saw_newack to 1 for
1410 * the destination that the TSN was
1414 sack
->num_gap_ack_blocks
&&
1415 q
->asoc
->peer
.primary_path
->cacc
.
1417 transport
->cacc
.cacc_saw_newack
1421 list_add_tail(&tchunk
->transmitted_list
,
1424 /* RFC2960 7.2.4, sctpimpguide-05 2.8.2
1425 * M2) Each time a SACK arrives reporting
1426 * 'Stray DATA chunk(s)' record the highest TSN
1427 * reported as newly acknowledged, call this
1428 * value 'HighestTSNinSack'. A newly
1429 * acknowledged DATA chunk is one not
1430 * previously acknowledged in a SACK.
1432 * When the SCTP sender of data receives a SACK
1433 * chunk that acknowledges, for the first time,
1434 * the receipt of a DATA chunk, all the still
1435 * unacknowledged DATA chunks whose TSN is
1436 * older than that newly acknowledged DATA
1437 * chunk, are qualified as 'Stray DATA chunks'.
1439 list_add_tail(lchunk
, &tlist
);
1442 if (tchunk
->tsn_gap_acked
) {
1443 pr_debug("%s: receiver reneged on data TSN:0x%x\n",
1446 tchunk
->tsn_gap_acked
= 0;
1448 if (tchunk
->transport
)
1449 bytes_acked
-= sctp_data_size(tchunk
);
1451 /* RFC 2960 6.3.2 Retransmission Timer Rules
1453 * R4) Whenever a SACK is received missing a
1454 * TSN that was previously acknowledged via a
1455 * Gap Ack Block, start T3-rtx for the
1456 * destination address to which the DATA
1457 * chunk was originally
1458 * transmitted if it is not already running.
1463 list_add_tail(lchunk
, &tlist
);
1469 struct sctp_association
*asoc
= transport
->asoc
;
1471 /* We may have counted DATA that was migrated
1472 * to this transport due to DEL-IP operation.
1473 * Subtract those bytes, since the were never
1474 * send on this transport and shouldn't be
1475 * credited to this transport.
1477 bytes_acked
-= migrate_bytes
;
1479 /* 8.2. When an outstanding TSN is acknowledged,
1480 * the endpoint shall clear the error counter of
1481 * the destination transport address to which the
1482 * DATA chunk was last sent.
1483 * The association's overall error counter is
1486 transport
->error_count
= 0;
1487 transport
->asoc
->overall_error_count
= 0;
1488 forward_progress
= true;
1491 * While in SHUTDOWN PENDING, we may have started
1492 * the T5 shutdown guard timer after reaching the
1493 * retransmission limit. Stop that timer as soon
1494 * as the receiver acknowledged any data.
1496 if (asoc
->state
== SCTP_STATE_SHUTDOWN_PENDING
&&
1497 del_timer(&asoc
->timers
1498 [SCTP_EVENT_TIMEOUT_T5_SHUTDOWN_GUARD
]))
1499 sctp_association_put(asoc
);
1501 /* Mark the destination transport address as
1502 * active if it is not so marked.
1504 if ((transport
->state
== SCTP_INACTIVE
||
1505 transport
->state
== SCTP_UNCONFIRMED
) &&
1506 sctp_cmp_addr_exact(&transport
->ipaddr
, saddr
)) {
1507 sctp_assoc_control_transport(
1511 SCTP_RECEIVED_SACK
);
1514 sctp_transport_raise_cwnd(transport
, sack_ctsn
,
1517 transport
->flight_size
-= bytes_acked
;
1518 if (transport
->flight_size
== 0)
1519 transport
->partial_bytes_acked
= 0;
1520 q
->outstanding_bytes
-= bytes_acked
+ migrate_bytes
;
1522 /* RFC 2960 6.1, sctpimpguide-06 2.15.2
1523 * When a sender is doing zero window probing, it
1524 * should not timeout the association if it continues
1525 * to receive new packets from the receiver. The
1526 * reason is that the receiver MAY keep its window
1527 * closed for an indefinite time.
1528 * A sender is doing zero window probing when the
1529 * receiver's advertised window is zero, and there is
1530 * only one data chunk in flight to the receiver.
1532 * Allow the association to timeout while in SHUTDOWN
1533 * PENDING or SHUTDOWN RECEIVED in case the receiver
1534 * stays in zero window mode forever.
1536 if (!q
->asoc
->peer
.rwnd
&&
1537 !list_empty(&tlist
) &&
1538 (sack_ctsn
+2 == q
->asoc
->next_tsn
) &&
1539 q
->asoc
->state
< SCTP_STATE_SHUTDOWN_PENDING
) {
1540 pr_debug("%s: sack received for zero window "
1541 "probe:%u\n", __func__
, sack_ctsn
);
1543 q
->asoc
->overall_error_count
= 0;
1544 transport
->error_count
= 0;
1548 /* RFC 2960 6.3.2 Retransmission Timer Rules
1550 * R2) Whenever all outstanding data sent to an address have
1551 * been acknowledged, turn off the T3-rtx timer of that
1554 if (!transport
->flight_size
) {
1555 if (del_timer(&transport
->T3_rtx_timer
))
1556 sctp_transport_put(transport
);
1557 } else if (restart_timer
) {
1558 if (!mod_timer(&transport
->T3_rtx_timer
,
1559 jiffies
+ transport
->rto
))
1560 sctp_transport_hold(transport
);
1563 if (forward_progress
) {
1565 dst_confirm(transport
->dst
);
1569 list_splice(&tlist
, transmitted_queue
);
1572 /* Mark chunks as missing and consequently may get retransmitted. */
1573 static void sctp_mark_missing(struct sctp_outq
*q
,
1574 struct list_head
*transmitted_queue
,
1575 struct sctp_transport
*transport
,
1576 __u32 highest_new_tsn_in_sack
,
1577 int count_of_newacks
)
1579 struct sctp_chunk
*chunk
;
1581 char do_fast_retransmit
= 0;
1582 struct sctp_association
*asoc
= q
->asoc
;
1583 struct sctp_transport
*primary
= asoc
->peer
.primary_path
;
1585 list_for_each_entry(chunk
, transmitted_queue
, transmitted_list
) {
1587 tsn
= ntohl(chunk
->subh
.data_hdr
->tsn
);
1589 /* RFC 2960 7.2.4, sctpimpguide-05 2.8.2 M3) Examine all
1590 * 'Unacknowledged TSN's', if the TSN number of an
1591 * 'Unacknowledged TSN' is smaller than the 'HighestTSNinSack'
1592 * value, increment the 'TSN.Missing.Report' count on that
1593 * chunk if it has NOT been fast retransmitted or marked for
1594 * fast retransmit already.
1596 if (chunk
->fast_retransmit
== SCTP_CAN_FRTX
&&
1597 !chunk
->tsn_gap_acked
&&
1598 TSN_lt(tsn
, highest_new_tsn_in_sack
)) {
1600 /* SFR-CACC may require us to skip marking
1601 * this chunk as missing.
1603 if (!transport
|| !sctp_cacc_skip(primary
,
1605 count_of_newacks
, tsn
)) {
1606 chunk
->tsn_missing_report
++;
1608 pr_debug("%s: tsn:0x%x missing counter:%d\n",
1609 __func__
, tsn
, chunk
->tsn_missing_report
);
1613 * M4) If any DATA chunk is found to have a
1614 * 'TSN.Missing.Report'
1615 * value larger than or equal to 3, mark that chunk for
1616 * retransmission and start the fast retransmit procedure.
1619 if (chunk
->tsn_missing_report
>= 3) {
1620 chunk
->fast_retransmit
= SCTP_NEED_FRTX
;
1621 do_fast_retransmit
= 1;
1626 if (do_fast_retransmit
)
1627 sctp_retransmit(q
, transport
, SCTP_RTXR_FAST_RTX
);
1629 pr_debug("%s: transport:%p, cwnd:%d, ssthresh:%d, "
1630 "flight_size:%d, pba:%d\n", __func__
, transport
,
1631 transport
->cwnd
, transport
->ssthresh
,
1632 transport
->flight_size
, transport
->partial_bytes_acked
);
1636 /* Is the given TSN acked by this packet? */
1637 static int sctp_acked(struct sctp_sackhdr
*sack
, __u32 tsn
)
1640 sctp_sack_variable_t
*frags
;
1642 __u32 ctsn
= ntohl(sack
->cum_tsn_ack
);
1644 if (TSN_lte(tsn
, ctsn
))
1647 /* 3.3.4 Selective Acknowledgement (SACK) (3):
1650 * These fields contain the Gap Ack Blocks. They are repeated
1651 * for each Gap Ack Block up to the number of Gap Ack Blocks
1652 * defined in the Number of Gap Ack Blocks field. All DATA
1653 * chunks with TSNs greater than or equal to (Cumulative TSN
1654 * Ack + Gap Ack Block Start) and less than or equal to
1655 * (Cumulative TSN Ack + Gap Ack Block End) of each Gap Ack
1656 * Block are assumed to have been received correctly.
1659 frags
= sack
->variable
;
1661 for (i
= 0; i
< ntohs(sack
->num_gap_ack_blocks
); ++i
) {
1662 if (TSN_lte(ntohs(frags
[i
].gab
.start
), gap
) &&
1663 TSN_lte(gap
, ntohs(frags
[i
].gab
.end
)))
1672 static inline int sctp_get_skip_pos(struct sctp_fwdtsn_skip
*skiplist
,
1673 int nskips
, __be16 stream
)
1677 for (i
= 0; i
< nskips
; i
++) {
1678 if (skiplist
[i
].stream
== stream
)
1684 /* Create and add a fwdtsn chunk to the outq's control queue if needed. */
1685 static void sctp_generate_fwdtsn(struct sctp_outq
*q
, __u32 ctsn
)
1687 struct sctp_association
*asoc
= q
->asoc
;
1688 struct sctp_chunk
*ftsn_chunk
= NULL
;
1689 struct sctp_fwdtsn_skip ftsn_skip_arr
[10];
1693 struct sctp_chunk
*chunk
;
1694 struct list_head
*lchunk
, *temp
;
1696 if (!asoc
->peer
.prsctp_capable
)
1699 /* PR-SCTP C1) Let SackCumAck be the Cumulative TSN ACK carried in the
1702 * If (Advanced.Peer.Ack.Point < SackCumAck), then update
1703 * Advanced.Peer.Ack.Point to be equal to SackCumAck.
1705 if (TSN_lt(asoc
->adv_peer_ack_point
, ctsn
))
1706 asoc
->adv_peer_ack_point
= ctsn
;
1708 /* PR-SCTP C2) Try to further advance the "Advanced.Peer.Ack.Point"
1709 * locally, that is, to move "Advanced.Peer.Ack.Point" up as long as
1710 * the chunk next in the out-queue space is marked as "abandoned" as
1711 * shown in the following example:
1713 * Assuming that a SACK arrived with the Cumulative TSN ACK 102
1714 * and the Advanced.Peer.Ack.Point is updated to this value:
1716 * out-queue at the end of ==> out-queue after Adv.Ack.Point
1717 * normal SACK processing local advancement
1719 * Adv.Ack.Pt-> 102 acked 102 acked
1720 * 103 abandoned 103 abandoned
1721 * 104 abandoned Adv.Ack.P-> 104 abandoned
1723 * 106 acked 106 acked
1726 * In this example, the data sender successfully advanced the
1727 * "Advanced.Peer.Ack.Point" from 102 to 104 locally.
1729 list_for_each_safe(lchunk
, temp
, &q
->abandoned
) {
1730 chunk
= list_entry(lchunk
, struct sctp_chunk
,
1732 tsn
= ntohl(chunk
->subh
.data_hdr
->tsn
);
1734 /* Remove any chunks in the abandoned queue that are acked by
1737 if (TSN_lte(tsn
, ctsn
)) {
1738 list_del_init(lchunk
);
1739 sctp_chunk_free(chunk
);
1741 if (TSN_lte(tsn
, asoc
->adv_peer_ack_point
+1)) {
1742 asoc
->adv_peer_ack_point
= tsn
;
1743 if (chunk
->chunk_hdr
->flags
&
1744 SCTP_DATA_UNORDERED
)
1746 skip_pos
= sctp_get_skip_pos(&ftsn_skip_arr
[0],
1748 chunk
->subh
.data_hdr
->stream
);
1749 ftsn_skip_arr
[skip_pos
].stream
=
1750 chunk
->subh
.data_hdr
->stream
;
1751 ftsn_skip_arr
[skip_pos
].ssn
=
1752 chunk
->subh
.data_hdr
->ssn
;
1753 if (skip_pos
== nskips
)
1762 /* PR-SCTP C3) If, after step C1 and C2, the "Advanced.Peer.Ack.Point"
1763 * is greater than the Cumulative TSN ACK carried in the received
1764 * SACK, the data sender MUST send the data receiver a FORWARD TSN
1765 * chunk containing the latest value of the
1766 * "Advanced.Peer.Ack.Point".
1768 * C4) For each "abandoned" TSN the sender of the FORWARD TSN SHOULD
1769 * list each stream and sequence number in the forwarded TSN. This
1770 * information will enable the receiver to easily find any
1771 * stranded TSN's waiting on stream reorder queues. Each stream
1772 * SHOULD only be reported once; this means that if multiple
1773 * abandoned messages occur in the same stream then only the
1774 * highest abandoned stream sequence number is reported. If the
1775 * total size of the FORWARD TSN does NOT fit in a single MTU then
1776 * the sender of the FORWARD TSN SHOULD lower the
1777 * Advanced.Peer.Ack.Point to the last TSN that will fit in a
1780 if (asoc
->adv_peer_ack_point
> ctsn
)
1781 ftsn_chunk
= sctp_make_fwdtsn(asoc
, asoc
->adv_peer_ack_point
,
1782 nskips
, &ftsn_skip_arr
[0]);
1785 list_add_tail(&ftsn_chunk
->list
, &q
->control_chunk_list
);
1786 SCTP_INC_STATS(sock_net(asoc
->base
.sk
), SCTP_MIB_OUTCTRLCHUNKS
);