Root/
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 | * |
6 | * This file is part of the SCTP kernel implementation |
7 | * |
8 | * These functions handle output processing. |
9 | * |
10 | * This SCTP implementation is free software; |
11 | * you can redistribute it and/or modify it under the terms of |
12 | * the GNU General Public License as published by |
13 | * the Free Software Foundation; either version 2, or (at your option) |
14 | * any later version. |
15 | * |
16 | * This SCTP implementation is distributed in the hope that it |
17 | * will be useful, but WITHOUT ANY WARRANTY; without even the implied |
18 | * ************************ |
19 | * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. |
20 | * See the GNU General Public License for more details. |
21 | * |
22 | * You should have received a copy of the GNU General Public License |
23 | * along with GNU CC; see the file COPYING. If not, write to |
24 | * the Free Software Foundation, 59 Temple Place - Suite 330, |
25 | * Boston, MA 02111-1307, USA. |
26 | * |
27 | * Please send any bug reports or fixes you make to the |
28 | * email address(es): |
29 | * lksctp developers <lksctp-developers@lists.sourceforge.net> |
30 | * |
31 | * Or submit a bug report through the following website: |
32 | * http://www.sf.net/projects/lksctp |
33 | * |
34 | * Written or modified by: |
35 | * La Monte H.P. Yarroll <piggy@acm.org> |
36 | * Karl Knutson <karl@athena.chicago.il.us> |
37 | * Jon Grimm <jgrimm@austin.ibm.com> |
38 | * Sridhar Samudrala <sri@us.ibm.com> |
39 | * |
40 | * Any bugs reported given to us we will try to fix... any fixes shared will |
41 | * be incorporated into the next SCTP release. |
42 | */ |
43 | |
44 | #include <linux/types.h> |
45 | #include <linux/kernel.h> |
46 | #include <linux/wait.h> |
47 | #include <linux/time.h> |
48 | #include <linux/ip.h> |
49 | #include <linux/ipv6.h> |
50 | #include <linux/init.h> |
51 | #include <net/inet_ecn.h> |
52 | #include <net/ip.h> |
53 | #include <net/icmp.h> |
54 | #include <net/net_namespace.h> |
55 | |
56 | #include <linux/socket.h> /* for sa_family_t */ |
57 | #include <net/sock.h> |
58 | |
59 | #include <net/sctp/sctp.h> |
60 | #include <net/sctp/sm.h> |
61 | #include <net/sctp/checksum.h> |
62 | |
63 | /* Forward declarations for private helpers. */ |
64 | static sctp_xmit_t sctp_packet_can_append_data(struct sctp_packet *packet, |
65 | struct sctp_chunk *chunk); |
66 | static void sctp_packet_append_data(struct sctp_packet *packet, |
67 | struct sctp_chunk *chunk); |
68 | static sctp_xmit_t sctp_packet_will_fit(struct sctp_packet *packet, |
69 | struct sctp_chunk *chunk, |
70 | u16 chunk_len); |
71 | |
72 | static void sctp_packet_reset(struct sctp_packet *packet) |
73 | { |
74 | packet->size = packet->overhead; |
75 | packet->has_cookie_echo = 0; |
76 | packet->has_sack = 0; |
77 | packet->has_data = 0; |
78 | packet->has_auth = 0; |
79 | packet->ipfragok = 0; |
80 | packet->auth = NULL; |
81 | } |
82 | |
83 | /* Config a packet. |
84 | * This appears to be a followup set of initializations. |
85 | */ |
86 | struct sctp_packet *sctp_packet_config(struct sctp_packet *packet, |
87 | __u32 vtag, int ecn_capable) |
88 | { |
89 | struct sctp_chunk *chunk = NULL; |
90 | |
91 | SCTP_DEBUG_PRINTK("%s: packet:%p vtag:0x%x\n", __func__, |
92 | packet, vtag); |
93 | |
94 | sctp_packet_reset(packet); |
95 | packet->vtag = vtag; |
96 | |
97 | if (ecn_capable && sctp_packet_empty(packet)) { |
98 | chunk = sctp_get_ecne_prepend(packet->transport->asoc); |
99 | |
100 | /* If there a is a prepend chunk stick it on the list before |
101 | * any other chunks get appended. |
102 | */ |
103 | if (chunk) |
104 | sctp_packet_append_chunk(packet, chunk); |
105 | } |
106 | |
107 | return packet; |
108 | } |
109 | |
110 | /* Initialize the packet structure. */ |
111 | struct sctp_packet *sctp_packet_init(struct sctp_packet *packet, |
112 | struct sctp_transport *transport, |
113 | __u16 sport, __u16 dport) |
114 | { |
115 | struct sctp_association *asoc = transport->asoc; |
116 | size_t overhead; |
117 | |
118 | SCTP_DEBUG_PRINTK("%s: packet:%p transport:%p\n", __func__, |
119 | packet, transport); |
120 | |
121 | packet->transport = transport; |
122 | packet->source_port = sport; |
123 | packet->destination_port = dport; |
124 | INIT_LIST_HEAD(&packet->chunk_list); |
125 | if (asoc) { |
126 | struct sctp_sock *sp = sctp_sk(asoc->base.sk); |
127 | overhead = sp->pf->af->net_header_len; |
128 | } else { |
129 | overhead = sizeof(struct ipv6hdr); |
130 | } |
131 | overhead += sizeof(struct sctphdr); |
132 | packet->overhead = overhead; |
133 | sctp_packet_reset(packet); |
134 | packet->vtag = 0; |
135 | packet->malloced = 0; |
136 | return packet; |
137 | } |
138 | |
139 | /* Free a packet. */ |
140 | void sctp_packet_free(struct sctp_packet *packet) |
141 | { |
142 | struct sctp_chunk *chunk, *tmp; |
143 | |
144 | SCTP_DEBUG_PRINTK("%s: packet:%p\n", __func__, packet); |
145 | |
146 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { |
147 | list_del_init(&chunk->list); |
148 | sctp_chunk_free(chunk); |
149 | } |
150 | |
151 | if (packet->malloced) |
152 | kfree(packet); |
153 | } |
154 | |
155 | /* This routine tries to append the chunk to the offered packet. If adding |
156 | * the chunk causes the packet to exceed the path MTU and COOKIE_ECHO chunk |
157 | * is not present in the packet, it transmits the input packet. |
158 | * Data can be bundled with a packet containing a COOKIE_ECHO chunk as long |
159 | * as it can fit in the packet, but any more data that does not fit in this |
160 | * packet can be sent only after receiving the COOKIE_ACK. |
161 | */ |
162 | sctp_xmit_t sctp_packet_transmit_chunk(struct sctp_packet *packet, |
163 | struct sctp_chunk *chunk, |
164 | int one_packet) |
165 | { |
166 | sctp_xmit_t retval; |
167 | int error = 0; |
168 | |
169 | SCTP_DEBUG_PRINTK("%s: packet:%p chunk:%p\n", __func__, |
170 | packet, chunk); |
171 | |
172 | switch ((retval = (sctp_packet_append_chunk(packet, chunk)))) { |
173 | case SCTP_XMIT_PMTU_FULL: |
174 | if (!packet->has_cookie_echo) { |
175 | error = sctp_packet_transmit(packet); |
176 | if (error < 0) |
177 | chunk->skb->sk->sk_err = -error; |
178 | |
179 | /* If we have an empty packet, then we can NOT ever |
180 | * return PMTU_FULL. |
181 | */ |
182 | if (!one_packet) |
183 | retval = sctp_packet_append_chunk(packet, |
184 | chunk); |
185 | } |
186 | break; |
187 | |
188 | case SCTP_XMIT_RWND_FULL: |
189 | case SCTP_XMIT_OK: |
190 | case SCTP_XMIT_NAGLE_DELAY: |
191 | break; |
192 | } |
193 | |
194 | return retval; |
195 | } |
196 | |
197 | /* Try to bundle an auth chunk into the packet. */ |
198 | static sctp_xmit_t sctp_packet_bundle_auth(struct sctp_packet *pkt, |
199 | struct sctp_chunk *chunk) |
200 | { |
201 | struct sctp_association *asoc = pkt->transport->asoc; |
202 | struct sctp_chunk *auth; |
203 | sctp_xmit_t retval = SCTP_XMIT_OK; |
204 | |
205 | /* if we don't have an association, we can't do authentication */ |
206 | if (!asoc) |
207 | return retval; |
208 | |
209 | /* See if this is an auth chunk we are bundling or if |
210 | * auth is already bundled. |
211 | */ |
212 | if (chunk->chunk_hdr->type == SCTP_CID_AUTH || pkt->has_auth) |
213 | return retval; |
214 | |
215 | /* if the peer did not request this chunk to be authenticated, |
216 | * don't do it |
217 | */ |
218 | if (!chunk->auth) |
219 | return retval; |
220 | |
221 | auth = sctp_make_auth(asoc); |
222 | if (!auth) |
223 | return retval; |
224 | |
225 | retval = sctp_packet_append_chunk(pkt, auth); |
226 | |
227 | return retval; |
228 | } |
229 | |
230 | /* Try to bundle a SACK with the packet. */ |
231 | static sctp_xmit_t sctp_packet_bundle_sack(struct sctp_packet *pkt, |
232 | struct sctp_chunk *chunk) |
233 | { |
234 | sctp_xmit_t retval = SCTP_XMIT_OK; |
235 | |
236 | /* If sending DATA and haven't aleady bundled a SACK, try to |
237 | * bundle one in to the packet. |
238 | */ |
239 | if (sctp_chunk_is_data(chunk) && !pkt->has_sack && |
240 | !pkt->has_cookie_echo) { |
241 | struct sctp_association *asoc; |
242 | struct timer_list *timer; |
243 | asoc = pkt->transport->asoc; |
244 | timer = &asoc->timers[SCTP_EVENT_TIMEOUT_SACK]; |
245 | |
246 | /* If the SACK timer is running, we have a pending SACK */ |
247 | if (timer_pending(timer)) { |
248 | struct sctp_chunk *sack; |
249 | asoc->a_rwnd = asoc->rwnd; |
250 | sack = sctp_make_sack(asoc); |
251 | if (sack) { |
252 | retval = sctp_packet_append_chunk(pkt, sack); |
253 | asoc->peer.sack_needed = 0; |
254 | if (del_timer(timer)) |
255 | sctp_association_put(asoc); |
256 | } |
257 | } |
258 | } |
259 | return retval; |
260 | } |
261 | |
262 | /* Append a chunk to the offered packet reporting back any inability to do |
263 | * so. |
264 | */ |
265 | sctp_xmit_t sctp_packet_append_chunk(struct sctp_packet *packet, |
266 | struct sctp_chunk *chunk) |
267 | { |
268 | sctp_xmit_t retval = SCTP_XMIT_OK; |
269 | __u16 chunk_len = WORD_ROUND(ntohs(chunk->chunk_hdr->length)); |
270 | |
271 | SCTP_DEBUG_PRINTK("%s: packet:%p chunk:%p\n", __func__, packet, |
272 | chunk); |
273 | |
274 | /* Data chunks are special. Before seeing what else we can |
275 | * bundle into this packet, check to see if we are allowed to |
276 | * send this DATA. |
277 | */ |
278 | if (sctp_chunk_is_data(chunk)) { |
279 | retval = sctp_packet_can_append_data(packet, chunk); |
280 | if (retval != SCTP_XMIT_OK) |
281 | goto finish; |
282 | } |
283 | |
284 | /* Try to bundle AUTH chunk */ |
285 | retval = sctp_packet_bundle_auth(packet, chunk); |
286 | if (retval != SCTP_XMIT_OK) |
287 | goto finish; |
288 | |
289 | /* Try to bundle SACK chunk */ |
290 | retval = sctp_packet_bundle_sack(packet, chunk); |
291 | if (retval != SCTP_XMIT_OK) |
292 | goto finish; |
293 | |
294 | /* Check to see if this chunk will fit into the packet */ |
295 | retval = sctp_packet_will_fit(packet, chunk, chunk_len); |
296 | if (retval != SCTP_XMIT_OK) |
297 | goto finish; |
298 | |
299 | /* We believe that this chunk is OK to add to the packet */ |
300 | switch (chunk->chunk_hdr->type) { |
301 | case SCTP_CID_DATA: |
302 | /* Account for the data being in the packet */ |
303 | sctp_packet_append_data(packet, chunk); |
304 | /* Disallow SACK bundling after DATA. */ |
305 | packet->has_sack = 1; |
306 | /* Disallow AUTH bundling after DATA */ |
307 | packet->has_auth = 1; |
308 | /* Let it be knows that packet has DATA in it */ |
309 | packet->has_data = 1; |
310 | /* timestamp the chunk for rtx purposes */ |
311 | chunk->sent_at = jiffies; |
312 | break; |
313 | case SCTP_CID_COOKIE_ECHO: |
314 | packet->has_cookie_echo = 1; |
315 | break; |
316 | |
317 | case SCTP_CID_SACK: |
318 | packet->has_sack = 1; |
319 | break; |
320 | |
321 | case SCTP_CID_AUTH: |
322 | packet->has_auth = 1; |
323 | packet->auth = chunk; |
324 | break; |
325 | } |
326 | |
327 | /* It is OK to send this chunk. */ |
328 | list_add_tail(&chunk->list, &packet->chunk_list); |
329 | packet->size += chunk_len; |
330 | chunk->transport = packet->transport; |
331 | finish: |
332 | return retval; |
333 | } |
334 | |
335 | /* All packets are sent to the network through this function from |
336 | * sctp_outq_tail(). |
337 | * |
338 | * The return value is a normal kernel error return value. |
339 | */ |
340 | int sctp_packet_transmit(struct sctp_packet *packet) |
341 | { |
342 | struct sctp_transport *tp = packet->transport; |
343 | struct sctp_association *asoc = tp->asoc; |
344 | struct sctphdr *sh; |
345 | struct sk_buff *nskb; |
346 | struct sctp_chunk *chunk, *tmp; |
347 | struct sock *sk; |
348 | int err = 0; |
349 | int padding; /* How much padding do we need? */ |
350 | __u8 has_data = 0; |
351 | struct dst_entry *dst = tp->dst; |
352 | unsigned char *auth = NULL; /* pointer to auth in skb data */ |
353 | __u32 cksum_buf_len = sizeof(struct sctphdr); |
354 | |
355 | SCTP_DEBUG_PRINTK("%s: packet:%p\n", __func__, packet); |
356 | |
357 | /* Do NOT generate a chunkless packet. */ |
358 | if (list_empty(&packet->chunk_list)) |
359 | return err; |
360 | |
361 | /* Set up convenience variables... */ |
362 | chunk = list_entry(packet->chunk_list.next, struct sctp_chunk, list); |
363 | sk = chunk->skb->sk; |
364 | |
365 | /* Allocate the new skb. */ |
366 | nskb = alloc_skb(packet->size + LL_MAX_HEADER, GFP_ATOMIC); |
367 | if (!nskb) |
368 | goto nomem; |
369 | |
370 | /* Make sure the outbound skb has enough header room reserved. */ |
371 | skb_reserve(nskb, packet->overhead + LL_MAX_HEADER); |
372 | |
373 | /* Set the owning socket so that we know where to get the |
374 | * destination IP address. |
375 | */ |
376 | skb_set_owner_w(nskb, sk); |
377 | |
378 | /* The 'obsolete' field of dst is set to 2 when a dst is freed. */ |
379 | if (!dst || (dst->obsolete > 1)) { |
380 | dst_release(dst); |
381 | sctp_transport_route(tp, NULL, sctp_sk(sk)); |
382 | if (asoc && (asoc->param_flags & SPP_PMTUD_ENABLE)) { |
383 | sctp_assoc_sync_pmtu(asoc); |
384 | } |
385 | } |
386 | dst = dst_clone(tp->dst); |
387 | skb_dst_set(nskb, dst); |
388 | if (!dst) |
389 | goto no_route; |
390 | |
391 | /* Build the SCTP header. */ |
392 | sh = (struct sctphdr *)skb_push(nskb, sizeof(struct sctphdr)); |
393 | skb_reset_transport_header(nskb); |
394 | sh->source = htons(packet->source_port); |
395 | sh->dest = htons(packet->destination_port); |
396 | |
397 | /* From 6.8 Adler-32 Checksum Calculation: |
398 | * After the packet is constructed (containing the SCTP common |
399 | * header and one or more control or DATA chunks), the |
400 | * transmitter shall: |
401 | * |
402 | * 1) Fill in the proper Verification Tag in the SCTP common |
403 | * header and initialize the checksum field to 0's. |
404 | */ |
405 | sh->vtag = htonl(packet->vtag); |
406 | sh->checksum = 0; |
407 | |
408 | /** |
409 | * 6.10 Bundling |
410 | * |
411 | * An endpoint bundles chunks by simply including multiple |
412 | * chunks in one outbound SCTP packet. ... |
413 | */ |
414 | |
415 | /** |
416 | * 3.2 Chunk Field Descriptions |
417 | * |
418 | * The total length of a chunk (including Type, Length and |
419 | * Value fields) MUST be a multiple of 4 bytes. If the length |
420 | * of the chunk is not a multiple of 4 bytes, the sender MUST |
421 | * pad the chunk with all zero bytes and this padding is not |
422 | * included in the chunk length field. The sender should |
423 | * never pad with more than 3 bytes. |
424 | * |
425 | * [This whole comment explains WORD_ROUND() below.] |
426 | */ |
427 | SCTP_DEBUG_PRINTK("***sctp_transmit_packet***\n"); |
428 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { |
429 | list_del_init(&chunk->list); |
430 | if (sctp_chunk_is_data(chunk)) { |
431 | |
432 | if (!chunk->has_tsn) { |
433 | sctp_chunk_assign_ssn(chunk); |
434 | sctp_chunk_assign_tsn(chunk); |
435 | |
436 | /* 6.3.1 C4) When data is in flight and when allowed |
437 | * by rule C5, a new RTT measurement MUST be made each |
438 | * round trip. Furthermore, new RTT measurements |
439 | * SHOULD be made no more than once per round-trip |
440 | * for a given destination transport address. |
441 | */ |
442 | |
443 | if (!tp->rto_pending) { |
444 | chunk->rtt_in_progress = 1; |
445 | tp->rto_pending = 1; |
446 | } |
447 | } else |
448 | chunk->resent = 1; |
449 | |
450 | has_data = 1; |
451 | } |
452 | |
453 | padding = WORD_ROUND(chunk->skb->len) - chunk->skb->len; |
454 | if (padding) |
455 | memset(skb_put(chunk->skb, padding), 0, padding); |
456 | |
457 | /* if this is the auth chunk that we are adding, |
458 | * store pointer where it will be added and put |
459 | * the auth into the packet. |
460 | */ |
461 | if (chunk == packet->auth) |
462 | auth = skb_tail_pointer(nskb); |
463 | |
464 | cksum_buf_len += chunk->skb->len; |
465 | memcpy(skb_put(nskb, chunk->skb->len), |
466 | chunk->skb->data, chunk->skb->len); |
467 | |
468 | SCTP_DEBUG_PRINTK("%s %p[%s] %s 0x%x, %s %d, %s %d, %s %d\n", |
469 | "*** Chunk", chunk, |
470 | sctp_cname(SCTP_ST_CHUNK( |
471 | chunk->chunk_hdr->type)), |
472 | chunk->has_tsn ? "TSN" : "No TSN", |
473 | chunk->has_tsn ? |
474 | ntohl(chunk->subh.data_hdr->tsn) : 0, |
475 | "length", ntohs(chunk->chunk_hdr->length), |
476 | "chunk->skb->len", chunk->skb->len, |
477 | "rtt_in_progress", chunk->rtt_in_progress); |
478 | |
479 | /* |
480 | * If this is a control chunk, this is our last |
481 | * reference. Free data chunks after they've been |
482 | * acknowledged or have failed. |
483 | */ |
484 | if (!sctp_chunk_is_data(chunk)) |
485 | sctp_chunk_free(chunk); |
486 | } |
487 | |
488 | /* SCTP-AUTH, Section 6.2 |
489 | * The sender MUST calculate the MAC as described in RFC2104 [2] |
490 | * using the hash function H as described by the MAC Identifier and |
491 | * the shared association key K based on the endpoint pair shared key |
492 | * described by the shared key identifier. The 'data' used for the |
493 | * computation of the AUTH-chunk is given by the AUTH chunk with its |
494 | * HMAC field set to zero (as shown in Figure 6) followed by all |
495 | * chunks that are placed after the AUTH chunk in the SCTP packet. |
496 | */ |
497 | if (auth) |
498 | sctp_auth_calculate_hmac(asoc, nskb, |
499 | (struct sctp_auth_chunk *)auth, |
500 | GFP_ATOMIC); |
501 | |
502 | /* 2) Calculate the Adler-32 checksum of the whole packet, |
503 | * including the SCTP common header and all the |
504 | * chunks. |
505 | * |
506 | * Note: Adler-32 is no longer applicable, as has been replaced |
507 | * by CRC32-C as described in <draft-ietf-tsvwg-sctpcsum-02.txt>. |
508 | */ |
509 | if (!sctp_checksum_disable && |
510 | !(dst->dev->features & (NETIF_F_NO_CSUM | NETIF_F_SCTP_CSUM))) { |
511 | __u32 crc32 = sctp_start_cksum((__u8 *)sh, cksum_buf_len); |
512 | |
513 | /* 3) Put the resultant value into the checksum field in the |
514 | * common header, and leave the rest of the bits unchanged. |
515 | */ |
516 | sh->checksum = sctp_end_cksum(crc32); |
517 | } else { |
518 | if (dst->dev->features & NETIF_F_SCTP_CSUM) { |
519 | /* no need to seed psuedo checksum for SCTP */ |
520 | nskb->ip_summed = CHECKSUM_PARTIAL; |
521 | nskb->csum_start = (skb_transport_header(nskb) - |
522 | nskb->head); |
523 | nskb->csum_offset = offsetof(struct sctphdr, checksum); |
524 | } else { |
525 | nskb->ip_summed = CHECKSUM_UNNECESSARY; |
526 | } |
527 | } |
528 | |
529 | /* IP layer ECN support |
530 | * From RFC 2481 |
531 | * "The ECN-Capable Transport (ECT) bit would be set by the |
532 | * data sender to indicate that the end-points of the |
533 | * transport protocol are ECN-capable." |
534 | * |
535 | * Now setting the ECT bit all the time, as it should not cause |
536 | * any problems protocol-wise even if our peer ignores it. |
537 | * |
538 | * Note: The works for IPv6 layer checks this bit too later |
539 | * in transmission. See IP6_ECN_flow_xmit(). |
540 | */ |
541 | (*tp->af_specific->ecn_capable)(nskb->sk); |
542 | |
543 | /* Set up the IP options. */ |
544 | /* BUG: not implemented |
545 | * For v4 this all lives somewhere in sk->sk_opt... |
546 | */ |
547 | |
548 | /* Dump that on IP! */ |
549 | if (asoc && asoc->peer.last_sent_to != tp) { |
550 | /* Considering the multiple CPU scenario, this is a |
551 | * "correcter" place for last_sent_to. --xguo |
552 | */ |
553 | asoc->peer.last_sent_to = tp; |
554 | } |
555 | |
556 | if (has_data) { |
557 | struct timer_list *timer; |
558 | unsigned long timeout; |
559 | |
560 | tp->last_time_used = jiffies; |
561 | |
562 | /* Restart the AUTOCLOSE timer when sending data. */ |
563 | if (sctp_state(asoc, ESTABLISHED) && asoc->autoclose) { |
564 | timer = &asoc->timers[SCTP_EVENT_TIMEOUT_AUTOCLOSE]; |
565 | timeout = asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE]; |
566 | |
567 | if (!mod_timer(timer, jiffies + timeout)) |
568 | sctp_association_hold(asoc); |
569 | } |
570 | } |
571 | |
572 | SCTP_DEBUG_PRINTK("***sctp_transmit_packet*** skb len %d\n", |
573 | nskb->len); |
574 | |
575 | nskb->local_df = packet->ipfragok; |
576 | (*tp->af_specific->sctp_xmit)(nskb, tp); |
577 | |
578 | out: |
579 | sctp_packet_reset(packet); |
580 | return err; |
581 | no_route: |
582 | kfree_skb(nskb); |
583 | IP_INC_STATS_BH(&init_net, IPSTATS_MIB_OUTNOROUTES); |
584 | |
585 | /* FIXME: Returning the 'err' will effect all the associations |
586 | * associated with a socket, although only one of the paths of the |
587 | * association is unreachable. |
588 | * The real failure of a transport or association can be passed on |
589 | * to the user via notifications. So setting this error may not be |
590 | * required. |
591 | */ |
592 | /* err = -EHOSTUNREACH; */ |
593 | err: |
594 | /* Control chunks are unreliable so just drop them. DATA chunks |
595 | * will get resent or dropped later. |
596 | */ |
597 | |
598 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { |
599 | list_del_init(&chunk->list); |
600 | if (!sctp_chunk_is_data(chunk)) |
601 | sctp_chunk_free(chunk); |
602 | } |
603 | goto out; |
604 | nomem: |
605 | err = -ENOMEM; |
606 | goto err; |
607 | } |
608 | |
609 | /******************************************************************** |
610 | * 2nd Level Abstractions |
611 | ********************************************************************/ |
612 | |
613 | /* This private function check to see if a chunk can be added */ |
614 | static sctp_xmit_t sctp_packet_can_append_data(struct sctp_packet *packet, |
615 | struct sctp_chunk *chunk) |
616 | { |
617 | sctp_xmit_t retval = SCTP_XMIT_OK; |
618 | size_t datasize, rwnd, inflight, flight_size; |
619 | struct sctp_transport *transport = packet->transport; |
620 | __u32 max_burst_bytes; |
621 | struct sctp_association *asoc = transport->asoc; |
622 | struct sctp_outq *q = &asoc->outqueue; |
623 | |
624 | /* RFC 2960 6.1 Transmission of DATA Chunks |
625 | * |
626 | * A) At any given time, the data sender MUST NOT transmit new data to |
627 | * any destination transport address if its peer's rwnd indicates |
628 | * that the peer has no buffer space (i.e. rwnd is 0, see Section |
629 | * 6.2.1). However, regardless of the value of rwnd (including if it |
630 | * is 0), the data sender can always have one DATA chunk in flight to |
631 | * the receiver if allowed by cwnd (see rule B below). This rule |
632 | * allows the sender to probe for a change in rwnd that the sender |
633 | * missed due to the SACK having been lost in transit from the data |
634 | * receiver to the data sender. |
635 | */ |
636 | |
637 | rwnd = asoc->peer.rwnd; |
638 | inflight = q->outstanding_bytes; |
639 | flight_size = transport->flight_size; |
640 | |
641 | datasize = sctp_data_size(chunk); |
642 | |
643 | if (datasize > rwnd) { |
644 | if (inflight > 0) { |
645 | /* We have (at least) one data chunk in flight, |
646 | * so we can't fall back to rule 6.1 B). |
647 | */ |
648 | retval = SCTP_XMIT_RWND_FULL; |
649 | goto finish; |
650 | } |
651 | } |
652 | |
653 | /* sctpimpguide-05 2.14.2 |
654 | * D) When the time comes for the sender to |
655 | * transmit new DATA chunks, the protocol parameter Max.Burst MUST |
656 | * first be applied to limit how many new DATA chunks may be sent. |
657 | * The limit is applied by adjusting cwnd as follows: |
658 | * if ((flightsize + Max.Burst * MTU) < cwnd) |
659 | * cwnd = flightsize + Max.Burst * MTU |
660 | */ |
661 | max_burst_bytes = asoc->max_burst * asoc->pathmtu; |
662 | if ((flight_size + max_burst_bytes) < transport->cwnd) { |
663 | transport->cwnd = flight_size + max_burst_bytes; |
664 | SCTP_DEBUG_PRINTK("%s: cwnd limited by max_burst: " |
665 | "transport: %p, cwnd: %d, " |
666 | "ssthresh: %d, flight_size: %d, " |
667 | "pba: %d\n", |
668 | __func__, transport, |
669 | transport->cwnd, |
670 | transport->ssthresh, |
671 | transport->flight_size, |
672 | transport->partial_bytes_acked); |
673 | } |
674 | |
675 | /* RFC 2960 6.1 Transmission of DATA Chunks |
676 | * |
677 | * B) At any given time, the sender MUST NOT transmit new data |
678 | * to a given transport address if it has cwnd or more bytes |
679 | * of data outstanding to that transport address. |
680 | */ |
681 | /* RFC 7.2.4 & the Implementers Guide 2.8. |
682 | * |
683 | * 3) ... |
684 | * When a Fast Retransmit is being performed the sender SHOULD |
685 | * ignore the value of cwnd and SHOULD NOT delay retransmission. |
686 | */ |
687 | if (chunk->fast_retransmit != SCTP_NEED_FRTX) |
688 | if (flight_size >= transport->cwnd) { |
689 | retval = SCTP_XMIT_RWND_FULL; |
690 | goto finish; |
691 | } |
692 | |
693 | /* Nagle's algorithm to solve small-packet problem: |
694 | * Inhibit the sending of new chunks when new outgoing data arrives |
695 | * if any previously transmitted data on the connection remains |
696 | * unacknowledged. |
697 | */ |
698 | if (!sctp_sk(asoc->base.sk)->nodelay && sctp_packet_empty(packet) && |
699 | inflight && sctp_state(asoc, ESTABLISHED)) { |
700 | unsigned max = transport->pathmtu - packet->overhead; |
701 | unsigned len = chunk->skb->len + q->out_qlen; |
702 | |
703 | /* Check whether this chunk and all the rest of pending |
704 | * data will fit or delay in hopes of bundling a full |
705 | * sized packet. |
706 | * Don't delay large message writes that may have been |
707 | * fragmeneted into small peices. |
708 | */ |
709 | if ((len < max) && (chunk->msg->msg_size < max)) { |
710 | retval = SCTP_XMIT_NAGLE_DELAY; |
711 | goto finish; |
712 | } |
713 | } |
714 | |
715 | finish: |
716 | return retval; |
717 | } |
718 | |
719 | /* This private function does management things when adding DATA chunk */ |
720 | static void sctp_packet_append_data(struct sctp_packet *packet, |
721 | struct sctp_chunk *chunk) |
722 | { |
723 | struct sctp_transport *transport = packet->transport; |
724 | size_t datasize = sctp_data_size(chunk); |
725 | struct sctp_association *asoc = transport->asoc; |
726 | u32 rwnd = asoc->peer.rwnd; |
727 | |
728 | /* Keep track of how many bytes are in flight over this transport. */ |
729 | transport->flight_size += datasize; |
730 | |
731 | /* Keep track of how many bytes are in flight to the receiver. */ |
732 | asoc->outqueue.outstanding_bytes += datasize; |
733 | |
734 | /* Update our view of the receiver's rwnd. Include sk_buff overhead |
735 | * while updating peer.rwnd so that it reduces the chances of a |
736 | * receiver running out of receive buffer space even when receive |
737 | * window is still open. This can happen when a sender is sending |
738 | * sending small messages. |
739 | */ |
740 | datasize += sizeof(struct sk_buff); |
741 | if (datasize < rwnd) |
742 | rwnd -= datasize; |
743 | else |
744 | rwnd = 0; |
745 | |
746 | asoc->peer.rwnd = rwnd; |
747 | /* Has been accepted for transmission. */ |
748 | if (!asoc->peer.prsctp_capable) |
749 | chunk->msg->can_abandon = 0; |
750 | } |
751 | |
752 | static sctp_xmit_t sctp_packet_will_fit(struct sctp_packet *packet, |
753 | struct sctp_chunk *chunk, |
754 | u16 chunk_len) |
755 | { |
756 | size_t psize; |
757 | size_t pmtu; |
758 | int too_big; |
759 | sctp_xmit_t retval = SCTP_XMIT_OK; |
760 | |
761 | psize = packet->size; |
762 | pmtu = ((packet->transport->asoc) ? |
763 | (packet->transport->asoc->pathmtu) : |
764 | (packet->transport->pathmtu)); |
765 | |
766 | too_big = (psize + chunk_len > pmtu); |
767 | |
768 | /* Decide if we need to fragment or resubmit later. */ |
769 | if (too_big) { |
770 | /* It's OK to fragmet at IP level if any one of the following |
771 | * is true: |
772 | * 1. The packet is empty (meaning this chunk is greater |
773 | * the MTU) |
774 | * 2. The chunk we are adding is a control chunk |
775 | * 3. The packet doesn't have any data in it yet and data |
776 | * requires authentication. |
777 | */ |
778 | if (sctp_packet_empty(packet) || !sctp_chunk_is_data(chunk) || |
779 | (!packet->has_data && chunk->auth)) { |
780 | /* We no longer do re-fragmentation. |
781 | * Just fragment at the IP layer, if we |
782 | * actually hit this condition |
783 | */ |
784 | packet->ipfragok = 1; |
785 | } else { |
786 | retval = SCTP_XMIT_PMTU_FULL; |
787 | } |
788 | } |
789 | |
790 | return retval; |
791 | } |
792 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9