]>
Commit | Line | Data |
---|---|---|
60c778b2 | 1 | /* SCTP kernel implementation |
1da177e4 LT |
2 | * (C) Copyright IBM Corp. 2001, 2004 |
3 | * Copyright (c) 1999-2000 Cisco, Inc. | |
4 | * Copyright (c) 1999-2001 Motorola, Inc. | |
5 | * | |
60c778b2 | 6 | * This file is part of the SCTP kernel implementation |
1da177e4 LT |
7 | * |
8 | * These functions handle output processing. | |
9 | * | |
60c778b2 | 10 | * This SCTP implementation is free software; |
1da177e4 LT |
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 | * | |
60c778b2 | 16 | * This SCTP implementation is distributed in the hope that it |
1da177e4 LT |
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 | |
4b2f13a2 JK |
23 | * along with GNU CC; see the file COPYING. If not, see |
24 | * <http://www.gnu.org/licenses/>. | |
1da177e4 LT |
25 | * |
26 | * Please send any bug reports or fixes you make to the | |
27 | * email address(es): | |
91705c61 | 28 | * lksctp developers <linux-sctp@vger.kernel.org> |
1da177e4 | 29 | * |
1da177e4 LT |
30 | * Written or modified by: |
31 | * La Monte H.P. Yarroll <piggy@acm.org> | |
32 | * Karl Knutson <karl@athena.chicago.il.us> | |
33 | * Jon Grimm <jgrimm@austin.ibm.com> | |
34 | * Sridhar Samudrala <sri@us.ibm.com> | |
1da177e4 LT |
35 | */ |
36 | ||
145ce502 JP |
37 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
38 | ||
1da177e4 LT |
39 | #include <linux/types.h> |
40 | #include <linux/kernel.h> | |
41 | #include <linux/wait.h> | |
42 | #include <linux/time.h> | |
43 | #include <linux/ip.h> | |
44 | #include <linux/ipv6.h> | |
45 | #include <linux/init.h> | |
5a0e3ad6 | 46 | #include <linux/slab.h> |
1da177e4 | 47 | #include <net/inet_ecn.h> |
8d2f9e81 | 48 | #include <net/ip.h> |
1da177e4 | 49 | #include <net/icmp.h> |
7c73a6fa | 50 | #include <net/net_namespace.h> |
1da177e4 | 51 | |
1da177e4 LT |
52 | #include <linux/socket.h> /* for sa_family_t */ |
53 | #include <net/sock.h> | |
54 | ||
55 | #include <net/sctp/sctp.h> | |
56 | #include <net/sctp/sm.h> | |
9ad0977f | 57 | #include <net/sctp/checksum.h> |
1da177e4 LT |
58 | |
59 | /* Forward declarations for private helpers. */ | |
86b36f2a XL |
60 | static enum sctp_xmit __sctp_packet_append_chunk(struct sctp_packet *packet, |
61 | struct sctp_chunk *chunk); | |
62 | static enum sctp_xmit sctp_packet_can_append_data(struct sctp_packet *packet, | |
63 | struct sctp_chunk *chunk); | |
e83963b7 | 64 | static void sctp_packet_append_data(struct sctp_packet *packet, |
86b36f2a XL |
65 | struct sctp_chunk *chunk); |
66 | static enum sctp_xmit sctp_packet_will_fit(struct sctp_packet *packet, | |
67 | struct sctp_chunk *chunk, | |
68 | u16 chunk_len); | |
1da177e4 | 69 | |
be297143 WY |
70 | static void sctp_packet_reset(struct sctp_packet *packet) |
71 | { | |
72 | packet->size = packet->overhead; | |
73 | packet->has_cookie_echo = 0; | |
74 | packet->has_sack = 0; | |
75 | packet->has_data = 0; | |
76 | packet->has_auth = 0; | |
77 | packet->ipfragok = 0; | |
78 | packet->auth = NULL; | |
79 | } | |
80 | ||
1da177e4 LT |
81 | /* Config a packet. |
82 | * This appears to be a followup set of initializations. | |
83 | */ | |
66b91d2c MRL |
84 | void sctp_packet_config(struct sctp_packet *packet, __u32 vtag, |
85 | int ecn_capable) | |
1da177e4 | 86 | { |
90017acc MRL |
87 | struct sctp_transport *tp = packet->transport; |
88 | struct sctp_association *asoc = tp->asoc; | |
df2729c3 | 89 | struct sock *sk; |
1da177e4 | 90 | |
bb33381d | 91 | pr_debug("%s: packet:%p vtag:0x%x\n", __func__, packet, vtag); |
1da177e4 | 92 | packet->vtag = vtag; |
1da177e4 | 93 | |
df2729c3 XL |
94 | /* do the following jobs only once for a flush schedule */ |
95 | if (!sctp_packet_empty(packet)) | |
96 | return; | |
90017acc | 97 | |
df2729c3 XL |
98 | /* set packet max_size with pathmtu */ |
99 | packet->max_size = tp->pathmtu; | |
100 | if (!asoc) | |
101 | return; | |
90017acc | 102 | |
df2729c3 XL |
103 | /* update dst or transport pathmtu if in need */ |
104 | sk = asoc->base.sk; | |
105 | if (!sctp_transport_dst_check(tp)) { | |
106 | sctp_transport_route(tp, NULL, sctp_sk(sk)); | |
107 | if (asoc->param_flags & SPP_PMTUD_ENABLE) | |
3ebfdf08 | 108 | sctp_assoc_sync_pmtu(asoc); |
df2729c3 XL |
109 | } else if (!sctp_transport_pmtu_check(tp)) { |
110 | if (asoc->param_flags & SPP_PMTUD_ENABLE) | |
3ebfdf08 | 111 | sctp_assoc_sync_pmtu(asoc); |
90017acc MRL |
112 | } |
113 | ||
df2729c3 XL |
114 | /* If there a is a prepend chunk stick it on the list before |
115 | * any other chunks get appended. | |
116 | */ | |
117 | if (ecn_capable) { | |
118 | struct sctp_chunk *chunk = sctp_get_ecne_prepend(asoc); | |
1da177e4 | 119 | |
1da177e4 LT |
120 | if (chunk) |
121 | sctp_packet_append_chunk(packet, chunk); | |
122 | } | |
df2729c3 XL |
123 | |
124 | if (!tp->dst) | |
125 | return; | |
126 | ||
127 | /* set packet max_size with gso_max_size if gso is enabled*/ | |
128 | rcu_read_lock(); | |
129 | if (__sk_dst_get(sk) != tp->dst) { | |
130 | dst_hold(tp->dst); | |
131 | sk_setup_caps(sk, tp->dst); | |
132 | } | |
133 | packet->max_size = sk_can_gso(sk) ? tp->dst->dev->gso_max_size | |
134 | : asoc->pathmtu; | |
135 | rcu_read_unlock(); | |
1da177e4 LT |
136 | } |
137 | ||
138 | /* Initialize the packet structure. */ | |
66b91d2c MRL |
139 | void sctp_packet_init(struct sctp_packet *packet, |
140 | struct sctp_transport *transport, | |
141 | __u16 sport, __u16 dport) | |
1da177e4 LT |
142 | { |
143 | struct sctp_association *asoc = transport->asoc; | |
144 | size_t overhead; | |
145 | ||
bb33381d | 146 | pr_debug("%s: packet:%p transport:%p\n", __func__, packet, transport); |
1da177e4 LT |
147 | |
148 | packet->transport = transport; | |
149 | packet->source_port = sport; | |
150 | packet->destination_port = dport; | |
79af02c2 | 151 | INIT_LIST_HEAD(&packet->chunk_list); |
1da177e4 | 152 | if (asoc) { |
d808ad9a YH |
153 | struct sctp_sock *sp = sctp_sk(asoc->base.sk); |
154 | overhead = sp->pf->af->net_header_len; | |
1da177e4 LT |
155 | } else { |
156 | overhead = sizeof(struct ipv6hdr); | |
157 | } | |
158 | overhead += sizeof(struct sctphdr); | |
159 | packet->overhead = overhead; | |
be297143 | 160 | sctp_packet_reset(packet); |
1da177e4 | 161 | packet->vtag = 0; |
1da177e4 LT |
162 | } |
163 | ||
164 | /* Free a packet. */ | |
165 | void sctp_packet_free(struct sctp_packet *packet) | |
166 | { | |
79af02c2 | 167 | struct sctp_chunk *chunk, *tmp; |
1da177e4 | 168 | |
bb33381d | 169 | pr_debug("%s: packet:%p\n", __func__, packet); |
1da177e4 | 170 | |
79af02c2 DM |
171 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { |
172 | list_del_init(&chunk->list); | |
1da177e4 | 173 | sctp_chunk_free(chunk); |
79af02c2 | 174 | } |
1da177e4 LT |
175 | } |
176 | ||
177 | /* This routine tries to append the chunk to the offered packet. If adding | |
178 | * the chunk causes the packet to exceed the path MTU and COOKIE_ECHO chunk | |
179 | * is not present in the packet, it transmits the input packet. | |
180 | * Data can be bundled with a packet containing a COOKIE_ECHO chunk as long | |
181 | * as it can fit in the packet, but any more data that does not fit in this | |
182 | * packet can be sent only after receiving the COOKIE_ACK. | |
183 | */ | |
86b36f2a XL |
184 | enum sctp_xmit sctp_packet_transmit_chunk(struct sctp_packet *packet, |
185 | struct sctp_chunk *chunk, | |
186 | int one_packet, gfp_t gfp) | |
1da177e4 | 187 | { |
86b36f2a | 188 | enum sctp_xmit retval; |
1da177e4 | 189 | |
5b5e0928 | 190 | pr_debug("%s: packet:%p size:%zu chunk:%p size:%d\n", __func__, |
942b3235 | 191 | packet, packet->size, chunk, chunk->skb ? chunk->skb->len : -1); |
1da177e4 LT |
192 | |
193 | switch ((retval = (sctp_packet_append_chunk(packet, chunk)))) { | |
194 | case SCTP_XMIT_PMTU_FULL: | |
195 | if (!packet->has_cookie_echo) { | |
64519440 XL |
196 | int error = 0; |
197 | ||
cea8768f | 198 | error = sctp_packet_transmit(packet, gfp); |
1da177e4 LT |
199 | if (error < 0) |
200 | chunk->skb->sk->sk_err = -error; | |
201 | ||
202 | /* If we have an empty packet, then we can NOT ever | |
203 | * return PMTU_FULL. | |
204 | */ | |
2e3216cd VY |
205 | if (!one_packet) |
206 | retval = sctp_packet_append_chunk(packet, | |
207 | chunk); | |
1da177e4 LT |
208 | } |
209 | break; | |
210 | ||
211 | case SCTP_XMIT_RWND_FULL: | |
212 | case SCTP_XMIT_OK: | |
526cbef7 | 213 | case SCTP_XMIT_DELAY: |
1da177e4 | 214 | break; |
3ff50b79 | 215 | } |
1da177e4 LT |
216 | |
217 | return retval; | |
218 | } | |
219 | ||
4cd57c80 | 220 | /* Try to bundle an auth chunk into the packet. */ |
86b36f2a XL |
221 | static enum sctp_xmit sctp_packet_bundle_auth(struct sctp_packet *pkt, |
222 | struct sctp_chunk *chunk) | |
4cd57c80 VY |
223 | { |
224 | struct sctp_association *asoc = pkt->transport->asoc; | |
86b36f2a | 225 | enum sctp_xmit retval = SCTP_XMIT_OK; |
4cd57c80 | 226 | struct sctp_chunk *auth; |
4cd57c80 VY |
227 | |
228 | /* if we don't have an association, we can't do authentication */ | |
229 | if (!asoc) | |
230 | return retval; | |
231 | ||
232 | /* See if this is an auth chunk we are bundling or if | |
233 | * auth is already bundled. | |
234 | */ | |
4007cc88 | 235 | if (chunk->chunk_hdr->type == SCTP_CID_AUTH || pkt->has_auth) |
4cd57c80 VY |
236 | return retval; |
237 | ||
238 | /* if the peer did not request this chunk to be authenticated, | |
239 | * don't do it | |
240 | */ | |
241 | if (!chunk->auth) | |
242 | return retval; | |
243 | ||
244 | auth = sctp_make_auth(asoc); | |
245 | if (!auth) | |
246 | return retval; | |
247 | ||
ed106277 NH |
248 | retval = __sctp_packet_append_chunk(pkt, auth); |
249 | ||
250 | if (retval != SCTP_XMIT_OK) | |
251 | sctp_chunk_free(auth); | |
4cd57c80 VY |
252 | |
253 | return retval; | |
254 | } | |
255 | ||
1da177e4 | 256 | /* Try to bundle a SACK with the packet. */ |
86b36f2a XL |
257 | static enum sctp_xmit sctp_packet_bundle_sack(struct sctp_packet *pkt, |
258 | struct sctp_chunk *chunk) | |
1da177e4 | 259 | { |
86b36f2a | 260 | enum sctp_xmit retval = SCTP_XMIT_OK; |
1da177e4 LT |
261 | |
262 | /* If sending DATA and haven't aleady bundled a SACK, try to | |
263 | * bundle one in to the packet. | |
264 | */ | |
265 | if (sctp_chunk_is_data(chunk) && !pkt->has_sack && | |
266 | !pkt->has_cookie_echo) { | |
267 | struct sctp_association *asoc; | |
af87b823 | 268 | struct timer_list *timer; |
1da177e4 | 269 | asoc = pkt->transport->asoc; |
af87b823 | 270 | timer = &asoc->timers[SCTP_EVENT_TIMEOUT_SACK]; |
1da177e4 | 271 | |
af87b823 DG |
272 | /* If the SACK timer is running, we have a pending SACK */ |
273 | if (timer_pending(timer)) { | |
1da177e4 | 274 | struct sctp_chunk *sack; |
4244854d NH |
275 | |
276 | if (pkt->transport->sack_generation != | |
277 | pkt->transport->asoc->peer.sack_generation) | |
278 | return retval; | |
279 | ||
1da177e4 LT |
280 | asoc->a_rwnd = asoc->rwnd; |
281 | sack = sctp_make_sack(asoc); | |
282 | if (sack) { | |
ed106277 NH |
283 | retval = __sctp_packet_append_chunk(pkt, sack); |
284 | if (retval != SCTP_XMIT_OK) { | |
285 | sctp_chunk_free(sack); | |
286 | goto out; | |
287 | } | |
1da177e4 | 288 | asoc->peer.sack_needed = 0; |
af87b823 | 289 | if (del_timer(timer)) |
1da177e4 LT |
290 | sctp_association_put(asoc); |
291 | } | |
292 | } | |
293 | } | |
ed106277 | 294 | out: |
1da177e4 LT |
295 | return retval; |
296 | } | |
297 | ||
ed106277 | 298 | |
1da177e4 LT |
299 | /* Append a chunk to the offered packet reporting back any inability to do |
300 | * so. | |
301 | */ | |
86b36f2a XL |
302 | static enum sctp_xmit __sctp_packet_append_chunk(struct sctp_packet *packet, |
303 | struct sctp_chunk *chunk) | |
1da177e4 | 304 | { |
e2f036a9 | 305 | __u16 chunk_len = SCTP_PAD4(ntohs(chunk->chunk_hdr->length)); |
86b36f2a | 306 | enum sctp_xmit retval = SCTP_XMIT_OK; |
1da177e4 | 307 | |
e83963b7 VY |
308 | /* Check to see if this chunk will fit into the packet */ |
309 | retval = sctp_packet_will_fit(packet, chunk, chunk_len); | |
310 | if (retval != SCTP_XMIT_OK) | |
311 | goto finish; | |
1da177e4 | 312 | |
e83963b7 | 313 | /* We believe that this chunk is OK to add to the packet */ |
4cd57c80 | 314 | switch (chunk->chunk_hdr->type) { |
f7010e61 | 315 | case SCTP_CID_DATA: |
e83963b7 VY |
316 | /* Account for the data being in the packet */ |
317 | sctp_packet_append_data(packet, chunk); | |
1da177e4 LT |
318 | /* Disallow SACK bundling after DATA. */ |
319 | packet->has_sack = 1; | |
4cd57c80 VY |
320 | /* Disallow AUTH bundling after DATA */ |
321 | packet->has_auth = 1; | |
322 | /* Let it be knows that packet has DATA in it */ | |
323 | packet->has_data = 1; | |
759af00e VY |
324 | /* timestamp the chunk for rtx purposes */ |
325 | chunk->sent_at = jiffies; | |
a6c2f792 XL |
326 | /* Mainly used for prsctp RTX policy */ |
327 | chunk->sent_count++; | |
4cd57c80 | 328 | break; |
f7010e61 | 329 | case SCTP_CID_COOKIE_ECHO: |
1da177e4 | 330 | packet->has_cookie_echo = 1; |
4cd57c80 VY |
331 | break; |
332 | ||
f7010e61 | 333 | case SCTP_CID_SACK: |
1da177e4 | 334 | packet->has_sack = 1; |
196d6759 MB |
335 | if (chunk->asoc) |
336 | chunk->asoc->stats.osacks++; | |
4cd57c80 VY |
337 | break; |
338 | ||
f7010e61 | 339 | case SCTP_CID_AUTH: |
4cd57c80 VY |
340 | packet->has_auth = 1; |
341 | packet->auth = chunk; | |
342 | break; | |
343 | } | |
1da177e4 LT |
344 | |
345 | /* It is OK to send this chunk. */ | |
79af02c2 | 346 | list_add_tail(&chunk->list, &packet->chunk_list); |
1da177e4 LT |
347 | packet->size += chunk_len; |
348 | chunk->transport = packet->transport; | |
349 | finish: | |
350 | return retval; | |
351 | } | |
352 | ||
ed106277 NH |
353 | /* Append a chunk to the offered packet reporting back any inability to do |
354 | * so. | |
355 | */ | |
86b36f2a XL |
356 | enum sctp_xmit sctp_packet_append_chunk(struct sctp_packet *packet, |
357 | struct sctp_chunk *chunk) | |
ed106277 | 358 | { |
86b36f2a | 359 | enum sctp_xmit retval = SCTP_XMIT_OK; |
ed106277 | 360 | |
bb33381d | 361 | pr_debug("%s: packet:%p chunk:%p\n", __func__, packet, chunk); |
ed106277 NH |
362 | |
363 | /* Data chunks are special. Before seeing what else we can | |
364 | * bundle into this packet, check to see if we are allowed to | |
365 | * send this DATA. | |
366 | */ | |
367 | if (sctp_chunk_is_data(chunk)) { | |
368 | retval = sctp_packet_can_append_data(packet, chunk); | |
369 | if (retval != SCTP_XMIT_OK) | |
370 | goto finish; | |
371 | } | |
372 | ||
373 | /* Try to bundle AUTH chunk */ | |
374 | retval = sctp_packet_bundle_auth(packet, chunk); | |
375 | if (retval != SCTP_XMIT_OK) | |
376 | goto finish; | |
377 | ||
378 | /* Try to bundle SACK chunk */ | |
379 | retval = sctp_packet_bundle_sack(packet, chunk); | |
380 | if (retval != SCTP_XMIT_OK) | |
381 | goto finish; | |
382 | ||
383 | retval = __sctp_packet_append_chunk(packet, chunk); | |
384 | ||
385 | finish: | |
386 | return retval; | |
387 | } | |
388 | ||
e4ff952a XL |
389 | static int sctp_packet_pack(struct sctp_packet *packet, |
390 | struct sk_buff *head, int gso, gfp_t gfp) | |
1da177e4 LT |
391 | { |
392 | struct sctp_transport *tp = packet->transport; | |
e4ff952a | 393 | struct sctp_auth_chunk *auth = NULL; |
79af02c2 | 394 | struct sctp_chunk *chunk, *tmp; |
e4ff952a XL |
395 | int pkt_count = 0, pkt_size; |
396 | struct sock *sk = head->sk; | |
397 | struct sk_buff *nskb; | |
ecc515d7 | 398 | int auth_len = 0; |
1da177e4 | 399 | |
90017acc | 400 | if (gso) { |
90017acc | 401 | skb_shinfo(head)->gso_type = sk->sk_gso_type; |
e4ff952a XL |
402 | NAPI_GRO_CB(head)->last = head; |
403 | } else { | |
404 | nskb = head; | |
405 | pkt_size = packet->size; | |
406 | goto merge; | |
90017acc | 407 | } |
1da177e4 | 408 | |
90017acc | 409 | do { |
e4ff952a XL |
410 | /* calculate the pkt_size and alloc nskb */ |
411 | pkt_size = packet->overhead; | |
412 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, | |
413 | list) { | |
414 | int padded = SCTP_PAD4(chunk->skb->len); | |
6eabca54 | 415 | |
e4ff952a XL |
416 | if (chunk == packet->auth) |
417 | auth_len = padded; | |
418 | else if (auth_len + padded + packet->overhead > | |
419 | tp->pathmtu) | |
420 | return 0; | |
421 | else if (pkt_size + padded > tp->pathmtu) | |
422 | break; | |
423 | pkt_size += padded; | |
90017acc | 424 | } |
e4ff952a XL |
425 | nskb = alloc_skb(pkt_size + MAX_HEADER, gfp); |
426 | if (!nskb) | |
427 | return 0; | |
428 | skb_reserve(nskb, packet->overhead + MAX_HEADER); | |
1da177e4 | 429 | |
e4ff952a XL |
430 | merge: |
431 | /* merge chunks into nskb and append nskb into head list */ | |
90017acc MRL |
432 | pkt_size -= packet->overhead; |
433 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { | |
e4ff952a XL |
434 | int padding; |
435 | ||
90017acc MRL |
436 | list_del_init(&chunk->list); |
437 | if (sctp_chunk_is_data(chunk)) { | |
cc6ac9bc XL |
438 | if (!sctp_chunk_retransmitted(chunk) && |
439 | !tp->rto_pending) { | |
90017acc MRL |
440 | chunk->rtt_in_progress = 1; |
441 | tp->rto_pending = 1; | |
442 | } | |
90017acc MRL |
443 | } |
444 | ||
e2f036a9 | 445 | padding = SCTP_PAD4(chunk->skb->len) - chunk->skb->len; |
90017acc | 446 | if (padding) |
b080db58 | 447 | skb_put_zero(chunk->skb, padding); |
90017acc | 448 | |
90017acc | 449 | if (chunk == packet->auth) |
e4ff952a XL |
450 | auth = (struct sctp_auth_chunk *) |
451 | skb_tail_pointer(nskb); | |
90017acc | 452 | |
59ae1d12 | 453 | skb_put_data(nskb, chunk->skb->data, chunk->skb->len); |
1da177e4 | 454 | |
90017acc MRL |
455 | pr_debug("*** Chunk:%p[%s] %s 0x%x, length:%d, chunk->skb->len:%d, rtt_in_progress:%d\n", |
456 | chunk, | |
457 | sctp_cname(SCTP_ST_CHUNK(chunk->chunk_hdr->type)), | |
458 | chunk->has_tsn ? "TSN" : "No TSN", | |
459 | chunk->has_tsn ? ntohl(chunk->subh.data_hdr->tsn) : 0, | |
460 | ntohs(chunk->chunk_hdr->length), chunk->skb->len, | |
461 | chunk->rtt_in_progress); | |
462 | ||
e2f036a9 | 463 | pkt_size -= SCTP_PAD4(chunk->skb->len); |
1da177e4 | 464 | |
f1533cce | 465 | if (!sctp_chunk_is_data(chunk) && chunk != packet->auth) |
90017acc MRL |
466 | sctp_chunk_free(chunk); |
467 | ||
468 | if (!pkt_size) | |
469 | break; | |
470 | } | |
471 | ||
e4ff952a XL |
472 | if (auth) { |
473 | sctp_auth_calculate_hmac(tp->asoc, nskb, auth, gfp); | |
474 | /* free auth if no more chunks, or add it back */ | |
475 | if (list_empty(&packet->chunk_list)) | |
476 | sctp_chunk_free(packet->auth); | |
477 | else | |
1aa25ec2 XL |
478 | list_add(&packet->auth->list, |
479 | &packet->chunk_list); | |
f1533cce MRL |
480 | } |
481 | ||
e4ff952a XL |
482 | if (gso) { |
483 | if (skb_gro_receive(&head, nskb)) { | |
484 | kfree_skb(nskb); | |
485 | return 0; | |
486 | } | |
487 | if (WARN_ON_ONCE(skb_shinfo(head)->gso_segs >= | |
488 | sk->sk_gso_max_segs)) | |
489 | return 0; | |
41001faf | 490 | } |
e4ff952a XL |
491 | |
492 | pkt_count++; | |
90017acc | 493 | } while (!list_empty(&packet->chunk_list)); |
4cd57c80 | 494 | |
e4ff952a XL |
495 | if (gso) { |
496 | memset(head->cb, 0, max(sizeof(struct inet_skb_parm), | |
497 | sizeof(struct inet6_skb_parm))); | |
498 | skb_shinfo(head)->gso_segs = pkt_count; | |
499 | skb_shinfo(head)->gso_size = GSO_BY_FRAGS; | |
500 | rcu_read_lock(); | |
501 | if (skb_dst(head) != tp->dst) { | |
502 | dst_hold(tp->dst); | |
503 | sk_setup_caps(sk, tp->dst); | |
8dc92f7e | 504 | } |
e4ff952a XL |
505 | rcu_read_unlock(); |
506 | goto chksum; | |
8dc92f7e | 507 | } |
1da177e4 | 508 | |
e4ff952a XL |
509 | if (sctp_checksum_disable) |
510 | return 1; | |
1da177e4 | 511 | |
e4ff952a XL |
512 | if (!(skb_dst(head)->dev->features & NETIF_F_SCTP_CRC) || |
513 | dst_xfrm(skb_dst(head)) || packet->ipfragok) { | |
514 | struct sctphdr *sh = | |
515 | (struct sctphdr *)skb_transport_header(head); | |
1da177e4 | 516 | |
e4ff952a XL |
517 | sh->checksum = sctp_compute_cksum(head, 0); |
518 | } else { | |
519 | chksum: | |
520 | head->ip_summed = CHECKSUM_PARTIAL; | |
dba00306 | 521 | head->csum_not_inet = 1; |
e4ff952a XL |
522 | head->csum_start = skb_transport_header(head) - head->head; |
523 | head->csum_offset = offsetof(struct sctphdr, checksum); | |
1da177e4 LT |
524 | } |
525 | ||
e4ff952a XL |
526 | return pkt_count; |
527 | } | |
528 | ||
529 | /* All packets are sent to the network through this function from | |
530 | * sctp_outq_tail(). | |
531 | * | |
532 | * The return value is always 0 for now. | |
533 | */ | |
534 | int sctp_packet_transmit(struct sctp_packet *packet, gfp_t gfp) | |
535 | { | |
536 | struct sctp_transport *tp = packet->transport; | |
537 | struct sctp_association *asoc = tp->asoc; | |
538 | struct sctp_chunk *chunk, *tmp; | |
539 | int pkt_count, gso = 0; | |
540 | struct dst_entry *dst; | |
541 | struct sk_buff *head; | |
542 | struct sctphdr *sh; | |
543 | struct sock *sk; | |
1da177e4 | 544 | |
e4ff952a XL |
545 | pr_debug("%s: packet:%p\n", __func__, packet); |
546 | if (list_empty(&packet->chunk_list)) | |
547 | return 0; | |
548 | chunk = list_entry(packet->chunk_list.next, struct sctp_chunk, list); | |
549 | sk = chunk->skb->sk; | |
1da177e4 | 550 | |
e4ff952a XL |
551 | /* check gso */ |
552 | if (packet->size > tp->pathmtu && !packet->ipfragok) { | |
553 | if (!sk_can_gso(sk)) { | |
554 | pr_err_once("Trying to GSO but underlying device doesn't support it."); | |
555 | goto out; | |
1da177e4 | 556 | } |
e4ff952a XL |
557 | gso = 1; |
558 | } | |
559 | ||
560 | /* alloc head skb */ | |
561 | head = alloc_skb((gso ? packet->overhead : packet->size) + | |
562 | MAX_HEADER, gfp); | |
563 | if (!head) | |
564 | goto out; | |
565 | skb_reserve(head, packet->overhead + MAX_HEADER); | |
6f43e071 | 566 | skb_set_owner_w(head, sk); |
e4ff952a XL |
567 | |
568 | /* set sctp header */ | |
d58ff351 | 569 | sh = skb_push(head, sizeof(struct sctphdr)); |
e4ff952a XL |
570 | skb_reset_transport_header(head); |
571 | sh->source = htons(packet->source_port); | |
572 | sh->dest = htons(packet->destination_port); | |
573 | sh->vtag = htonl(packet->vtag); | |
574 | sh->checksum = 0; | |
575 | ||
df2729c3 | 576 | /* drop packet if no dst */ |
e4ff952a XL |
577 | dst = dst_clone(tp->dst); |
578 | if (!dst) { | |
579 | IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES); | |
580 | kfree_skb(head); | |
581 | goto out; | |
582 | } | |
583 | skb_dst_set(head, dst); | |
1da177e4 | 584 | |
e4ff952a XL |
585 | /* pack up chunks */ |
586 | pkt_count = sctp_packet_pack(packet, head, gso, gfp); | |
587 | if (!pkt_count) { | |
588 | kfree_skb(head); | |
589 | goto out; | |
590 | } | |
90017acc MRL |
591 | pr_debug("***sctp_transmit_packet*** skb->len:%d\n", head->len); |
592 | ||
e4ff952a XL |
593 | /* start autoclose timer */ |
594 | if (packet->has_data && sctp_state(asoc, ESTABLISHED) && | |
595 | asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE]) { | |
596 | struct timer_list *timer = | |
597 | &asoc->timers[SCTP_EVENT_TIMEOUT_AUTOCLOSE]; | |
598 | unsigned long timeout = | |
599 | asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE]; | |
1da177e4 | 600 | |
e4ff952a XL |
601 | if (!mod_timer(timer, jiffies + timeout)) |
602 | sctp_association_hold(asoc); | |
603 | } | |
90017acc | 604 | |
e4ff952a XL |
605 | /* sctp xmit */ |
606 | tp->af_specific->ecn_capable(sk); | |
607 | if (asoc) { | |
608 | asoc->stats.opackets += pkt_count; | |
609 | if (asoc->peer.last_sent_to != tp) | |
610 | asoc->peer.last_sent_to = tp; | |
90017acc MRL |
611 | } |
612 | head->ignore_df = packet->ipfragok; | |
486a43db | 613 | if (tp->dst_pending_confirm) |
c86a773c JA |
614 | skb_set_dst_pending_confirm(head, 1); |
615 | /* neighbour should be confirmed on successful transmission or | |
616 | * positive error | |
617 | */ | |
486a43db XL |
618 | if (tp->af_specific->sctp_xmit(head, tp) >= 0 && |
619 | tp->dst_pending_confirm) | |
c86a773c | 620 | tp->dst_pending_confirm = 0; |
1da177e4 | 621 | |
e4ff952a | 622 | out: |
79af02c2 DM |
623 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { |
624 | list_del_init(&chunk->list); | |
1da177e4 | 625 | if (!sctp_chunk_is_data(chunk)) |
d808ad9a | 626 | sctp_chunk_free(chunk); |
1da177e4 | 627 | } |
41001faf | 628 | sctp_packet_reset(packet); |
e4ff952a | 629 | return 0; |
1da177e4 LT |
630 | } |
631 | ||
632 | /******************************************************************** | |
633 | * 2nd Level Abstractions | |
634 | ********************************************************************/ | |
635 | ||
e83963b7 | 636 | /* This private function check to see if a chunk can be added */ |
86b36f2a XL |
637 | static enum sctp_xmit sctp_packet_can_append_data(struct sctp_packet *packet, |
638 | struct sctp_chunk *chunk) | |
1da177e4 | 639 | { |
e83963b7 | 640 | size_t datasize, rwnd, inflight, flight_size; |
1da177e4 | 641 | struct sctp_transport *transport = packet->transport; |
1da177e4 | 642 | struct sctp_association *asoc = transport->asoc; |
1da177e4 LT |
643 | struct sctp_outq *q = &asoc->outqueue; |
644 | ||
645 | /* RFC 2960 6.1 Transmission of DATA Chunks | |
646 | * | |
647 | * A) At any given time, the data sender MUST NOT transmit new data to | |
648 | * any destination transport address if its peer's rwnd indicates | |
649 | * that the peer has no buffer space (i.e. rwnd is 0, see Section | |
650 | * 6.2.1). However, regardless of the value of rwnd (including if it | |
651 | * is 0), the data sender can always have one DATA chunk in flight to | |
652 | * the receiver if allowed by cwnd (see rule B below). This rule | |
653 | * allows the sender to probe for a change in rwnd that the sender | |
654 | * missed due to the SACK having been lost in transit from the data | |
655 | * receiver to the data sender. | |
656 | */ | |
657 | ||
658 | rwnd = asoc->peer.rwnd; | |
e83963b7 VY |
659 | inflight = q->outstanding_bytes; |
660 | flight_size = transport->flight_size; | |
1da177e4 LT |
661 | |
662 | datasize = sctp_data_size(chunk); | |
663 | ||
723189fa DL |
664 | if (datasize > rwnd && inflight > 0) |
665 | /* We have (at least) one data chunk in flight, | |
666 | * so we can't fall back to rule 6.1 B). | |
667 | */ | |
668 | return SCTP_XMIT_RWND_FULL; | |
1da177e4 | 669 | |
1da177e4 LT |
670 | /* RFC 2960 6.1 Transmission of DATA Chunks |
671 | * | |
672 | * B) At any given time, the sender MUST NOT transmit new data | |
673 | * to a given transport address if it has cwnd or more bytes | |
674 | * of data outstanding to that transport address. | |
675 | */ | |
676 | /* RFC 7.2.4 & the Implementers Guide 2.8. | |
677 | * | |
678 | * 3) ... | |
679 | * When a Fast Retransmit is being performed the sender SHOULD | |
680 | * ignore the value of cwnd and SHOULD NOT delay retransmission. | |
681 | */ | |
723189fa DL |
682 | if (chunk->fast_retransmit != SCTP_NEED_FRTX && |
683 | flight_size >= transport->cwnd) | |
684 | return SCTP_XMIT_RWND_FULL; | |
1da177e4 LT |
685 | |
686 | /* Nagle's algorithm to solve small-packet problem: | |
687 | * Inhibit the sending of new chunks when new outgoing data arrives | |
688 | * if any previously transmitted data on the connection remains | |
689 | * unacknowledged. | |
690 | */ | |
1da177e4 | 691 | |
4ea0c32f | 692 | if ((sctp_sk(asoc->base.sk)->nodelay || inflight == 0) && |
f9ba3501 | 693 | !asoc->force_delay) |
4ea0c32f | 694 | /* Nothing unacked */ |
723189fa DL |
695 | return SCTP_XMIT_OK; |
696 | ||
697 | if (!sctp_packet_empty(packet)) | |
698 | /* Append to packet */ | |
699 | return SCTP_XMIT_OK; | |
700 | ||
723189fa DL |
701 | if (!sctp_state(asoc, ESTABLISHED)) |
702 | return SCTP_XMIT_OK; | |
703 | ||
704 | /* Check whether this chunk and all the rest of pending data will fit | |
705 | * or delay in hopes of bundling a full sized packet. | |
706 | */ | |
9f8d3147 XL |
707 | if (chunk->skb->len + q->out_qlen > transport->pathmtu - |
708 | packet->overhead - sizeof(struct sctp_data_chunk) - 4) | |
723189fa DL |
709 | /* Enough data queued to fill a packet */ |
710 | return SCTP_XMIT_OK; | |
711 | ||
712 | /* Don't delay large message writes that may have been fragmented */ | |
713 | if (!chunk->msg->can_delay) | |
714 | return SCTP_XMIT_OK; | |
715 | ||
716 | /* Defer until all data acked or packet full */ | |
526cbef7 | 717 | return SCTP_XMIT_DELAY; |
e83963b7 VY |
718 | } |
719 | ||
720 | /* This private function does management things when adding DATA chunk */ | |
721 | static void sctp_packet_append_data(struct sctp_packet *packet, | |
722 | struct sctp_chunk *chunk) | |
723 | { | |
724 | struct sctp_transport *transport = packet->transport; | |
725 | size_t datasize = sctp_data_size(chunk); | |
726 | struct sctp_association *asoc = transport->asoc; | |
727 | u32 rwnd = asoc->peer.rwnd; | |
728 | ||
1da177e4 LT |
729 | /* Keep track of how many bytes are in flight over this transport. */ |
730 | transport->flight_size += datasize; | |
731 | ||
732 | /* Keep track of how many bytes are in flight to the receiver. */ | |
733 | asoc->outqueue.outstanding_bytes += datasize; | |
734 | ||
a76c0adf | 735 | /* Update our view of the receiver's rwnd. */ |
1da177e4 LT |
736 | if (datasize < rwnd) |
737 | rwnd -= datasize; | |
738 | else | |
739 | rwnd = 0; | |
740 | ||
741 | asoc->peer.rwnd = rwnd; | |
d8dd1578 NH |
742 | sctp_chunk_assign_tsn(chunk); |
743 | sctp_chunk_assign_ssn(chunk); | |
e83963b7 VY |
744 | } |
745 | ||
86b36f2a XL |
746 | static enum sctp_xmit sctp_packet_will_fit(struct sctp_packet *packet, |
747 | struct sctp_chunk *chunk, | |
748 | u16 chunk_len) | |
e83963b7 | 749 | { |
86b36f2a | 750 | enum sctp_xmit retval = SCTP_XMIT_OK; |
7303a147 | 751 | size_t psize, pmtu, maxsize; |
e83963b7 VY |
752 | |
753 | psize = packet->size; | |
90017acc MRL |
754 | if (packet->transport->asoc) |
755 | pmtu = packet->transport->asoc->pathmtu; | |
756 | else | |
757 | pmtu = packet->transport->pathmtu; | |
e83963b7 VY |
758 | |
759 | /* Decide if we need to fragment or resubmit later. */ | |
90017acc MRL |
760 | if (psize + chunk_len > pmtu) { |
761 | /* It's OK to fragment at IP level if any one of the following | |
e83963b7 | 762 | * is true: |
90017acc MRL |
763 | * 1. The packet is empty (meaning this chunk is greater |
764 | * the MTU) | |
765 | * 2. The packet doesn't have any data in it yet and data | |
766 | * requires authentication. | |
e83963b7 | 767 | */ |
90017acc | 768 | if (sctp_packet_empty(packet) || |
e83963b7 VY |
769 | (!packet->has_data && chunk->auth)) { |
770 | /* We no longer do re-fragmentation. | |
771 | * Just fragment at the IP layer, if we | |
772 | * actually hit this condition | |
773 | */ | |
774 | packet->ipfragok = 1; | |
90017acc | 775 | goto out; |
e83963b7 | 776 | } |
90017acc | 777 | |
7303a147 MRL |
778 | /* Similarly, if this chunk was built before a PMTU |
779 | * reduction, we have to fragment it at IP level now. So | |
780 | * if the packet already contains something, we need to | |
781 | * flush. | |
782 | */ | |
783 | maxsize = pmtu - packet->overhead; | |
784 | if (packet->auth) | |
e2f036a9 | 785 | maxsize -= SCTP_PAD4(packet->auth->skb->len); |
7303a147 MRL |
786 | if (chunk_len > maxsize) |
787 | retval = SCTP_XMIT_PMTU_FULL; | |
788 | ||
90017acc MRL |
789 | /* It is also okay to fragment if the chunk we are |
790 | * adding is a control chunk, but only if current packet | |
791 | * is not a GSO one otherwise it causes fragmentation of | |
792 | * a large frame. So in this case we allow the | |
793 | * fragmentation by forcing it to be in a new packet. | |
794 | */ | |
795 | if (!sctp_chunk_is_data(chunk) && packet->has_data) | |
796 | retval = SCTP_XMIT_PMTU_FULL; | |
797 | ||
798 | if (psize + chunk_len > packet->max_size) | |
799 | /* Hit GSO/PMTU limit, gotta flush */ | |
800 | retval = SCTP_XMIT_PMTU_FULL; | |
801 | ||
802 | if (!packet->transport->burst_limited && | |
803 | psize + chunk_len > (packet->transport->cwnd >> 1)) | |
804 | /* Do not allow a single GSO packet to use more | |
805 | * than half of cwnd. | |
806 | */ | |
807 | retval = SCTP_XMIT_PMTU_FULL; | |
808 | ||
809 | if (packet->transport->burst_limited && | |
810 | psize + chunk_len > (packet->transport->burst_limited >> 1)) | |
811 | /* Do not allow a single GSO packet to use more | |
812 | * than half of original cwnd. | |
813 | */ | |
814 | retval = SCTP_XMIT_PMTU_FULL; | |
815 | /* Otherwise it will fit in the GSO packet */ | |
e83963b7 | 816 | } |
1da177e4 | 817 | |
90017acc | 818 | out: |
1da177e4 LT |
819 | return retval; |
820 | } |