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 | { | |
b7e10c25 RH |
72 | /* sctp_packet_transmit() relies on this to reset size to the |
73 | * current overhead after sending packets. | |
74 | */ | |
be297143 | 75 | packet->size = packet->overhead; |
b7e10c25 | 76 | |
be297143 WY |
77 | packet->has_cookie_echo = 0; |
78 | packet->has_sack = 0; | |
79 | packet->has_data = 0; | |
80 | packet->has_auth = 0; | |
81 | packet->ipfragok = 0; | |
82 | packet->auth = NULL; | |
83 | } | |
84 | ||
1da177e4 LT |
85 | /* Config a packet. |
86 | * This appears to be a followup set of initializations. | |
87 | */ | |
66b91d2c MRL |
88 | void sctp_packet_config(struct sctp_packet *packet, __u32 vtag, |
89 | int ecn_capable) | |
1da177e4 | 90 | { |
90017acc MRL |
91 | struct sctp_transport *tp = packet->transport; |
92 | struct sctp_association *asoc = tp->asoc; | |
df2729c3 | 93 | struct sock *sk; |
b7e10c25 | 94 | size_t overhead = sizeof(struct ipv6hdr) + sizeof(struct sctphdr); |
1da177e4 | 95 | |
bb33381d | 96 | pr_debug("%s: packet:%p vtag:0x%x\n", __func__, packet, vtag); |
1da177e4 | 97 | packet->vtag = vtag; |
1da177e4 | 98 | |
df2729c3 XL |
99 | /* do the following jobs only once for a flush schedule */ |
100 | if (!sctp_packet_empty(packet)) | |
101 | return; | |
90017acc | 102 | |
b7e10c25 | 103 | /* set packet max_size with pathmtu, then calculate overhead */ |
df2729c3 | 104 | packet->max_size = tp->pathmtu; |
b7e10c25 RH |
105 | if (asoc) { |
106 | struct sctp_sock *sp = sctp_sk(asoc->base.sk); | |
107 | struct sctp_af *af = sp->pf->af; | |
108 | ||
109 | overhead = af->net_header_len + | |
110 | af->ip_options_len(asoc->base.sk); | |
111 | overhead += sizeof(struct sctphdr); | |
112 | packet->overhead = overhead; | |
113 | packet->size = overhead; | |
114 | } else { | |
115 | packet->overhead = overhead; | |
116 | packet->size = overhead; | |
df2729c3 | 117 | return; |
b7e10c25 | 118 | } |
90017acc | 119 | |
df2729c3 XL |
120 | /* update dst or transport pathmtu if in need */ |
121 | sk = asoc->base.sk; | |
122 | if (!sctp_transport_dst_check(tp)) { | |
123 | sctp_transport_route(tp, NULL, sctp_sk(sk)); | |
124 | if (asoc->param_flags & SPP_PMTUD_ENABLE) | |
3ebfdf08 | 125 | sctp_assoc_sync_pmtu(asoc); |
df2729c3 XL |
126 | } else if (!sctp_transport_pmtu_check(tp)) { |
127 | if (asoc->param_flags & SPP_PMTUD_ENABLE) | |
3ebfdf08 | 128 | sctp_assoc_sync_pmtu(asoc); |
90017acc MRL |
129 | } |
130 | ||
df2729c3 XL |
131 | /* If there a is a prepend chunk stick it on the list before |
132 | * any other chunks get appended. | |
133 | */ | |
134 | if (ecn_capable) { | |
135 | struct sctp_chunk *chunk = sctp_get_ecne_prepend(asoc); | |
1da177e4 | 136 | |
1da177e4 LT |
137 | if (chunk) |
138 | sctp_packet_append_chunk(packet, chunk); | |
139 | } | |
df2729c3 XL |
140 | |
141 | if (!tp->dst) | |
142 | return; | |
143 | ||
144 | /* set packet max_size with gso_max_size if gso is enabled*/ | |
145 | rcu_read_lock(); | |
146 | if (__sk_dst_get(sk) != tp->dst) { | |
147 | dst_hold(tp->dst); | |
148 | sk_setup_caps(sk, tp->dst); | |
149 | } | |
150 | packet->max_size = sk_can_gso(sk) ? tp->dst->dev->gso_max_size | |
151 | : asoc->pathmtu; | |
152 | rcu_read_unlock(); | |
1da177e4 LT |
153 | } |
154 | ||
155 | /* Initialize the packet structure. */ | |
66b91d2c MRL |
156 | void sctp_packet_init(struct sctp_packet *packet, |
157 | struct sctp_transport *transport, | |
158 | __u16 sport, __u16 dport) | |
1da177e4 | 159 | { |
bb33381d | 160 | pr_debug("%s: packet:%p transport:%p\n", __func__, packet, transport); |
1da177e4 LT |
161 | |
162 | packet->transport = transport; | |
163 | packet->source_port = sport; | |
164 | packet->destination_port = dport; | |
79af02c2 | 165 | INIT_LIST_HEAD(&packet->chunk_list); |
b7e10c25 RH |
166 | /* The overhead will be calculated by sctp_packet_config() */ |
167 | packet->overhead = 0; | |
be297143 | 168 | sctp_packet_reset(packet); |
1da177e4 | 169 | packet->vtag = 0; |
1da177e4 LT |
170 | } |
171 | ||
172 | /* Free a packet. */ | |
173 | void sctp_packet_free(struct sctp_packet *packet) | |
174 | { | |
79af02c2 | 175 | struct sctp_chunk *chunk, *tmp; |
1da177e4 | 176 | |
bb33381d | 177 | pr_debug("%s: packet:%p\n", __func__, packet); |
1da177e4 | 178 | |
79af02c2 DM |
179 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { |
180 | list_del_init(&chunk->list); | |
1da177e4 | 181 | sctp_chunk_free(chunk); |
79af02c2 | 182 | } |
1da177e4 LT |
183 | } |
184 | ||
185 | /* This routine tries to append the chunk to the offered packet. If adding | |
186 | * the chunk causes the packet to exceed the path MTU and COOKIE_ECHO chunk | |
187 | * is not present in the packet, it transmits the input packet. | |
188 | * Data can be bundled with a packet containing a COOKIE_ECHO chunk as long | |
189 | * as it can fit in the packet, but any more data that does not fit in this | |
190 | * packet can be sent only after receiving the COOKIE_ACK. | |
191 | */ | |
86b36f2a XL |
192 | enum sctp_xmit sctp_packet_transmit_chunk(struct sctp_packet *packet, |
193 | struct sctp_chunk *chunk, | |
194 | int one_packet, gfp_t gfp) | |
1da177e4 | 195 | { |
86b36f2a | 196 | enum sctp_xmit retval; |
1da177e4 | 197 | |
5b5e0928 | 198 | pr_debug("%s: packet:%p size:%zu chunk:%p size:%d\n", __func__, |
942b3235 | 199 | packet, packet->size, chunk, chunk->skb ? chunk->skb->len : -1); |
1da177e4 LT |
200 | |
201 | switch ((retval = (sctp_packet_append_chunk(packet, chunk)))) { | |
202 | case SCTP_XMIT_PMTU_FULL: | |
203 | if (!packet->has_cookie_echo) { | |
64519440 XL |
204 | int error = 0; |
205 | ||
cea8768f | 206 | error = sctp_packet_transmit(packet, gfp); |
1da177e4 LT |
207 | if (error < 0) |
208 | chunk->skb->sk->sk_err = -error; | |
209 | ||
210 | /* If we have an empty packet, then we can NOT ever | |
211 | * return PMTU_FULL. | |
212 | */ | |
2e3216cd VY |
213 | if (!one_packet) |
214 | retval = sctp_packet_append_chunk(packet, | |
215 | chunk); | |
1da177e4 LT |
216 | } |
217 | break; | |
218 | ||
219 | case SCTP_XMIT_RWND_FULL: | |
220 | case SCTP_XMIT_OK: | |
526cbef7 | 221 | case SCTP_XMIT_DELAY: |
1da177e4 | 222 | break; |
3ff50b79 | 223 | } |
1da177e4 LT |
224 | |
225 | return retval; | |
226 | } | |
227 | ||
4cd57c80 | 228 | /* Try to bundle an auth chunk into the packet. */ |
86b36f2a XL |
229 | static enum sctp_xmit sctp_packet_bundle_auth(struct sctp_packet *pkt, |
230 | struct sctp_chunk *chunk) | |
4cd57c80 VY |
231 | { |
232 | struct sctp_association *asoc = pkt->transport->asoc; | |
86b36f2a | 233 | enum sctp_xmit retval = SCTP_XMIT_OK; |
4cd57c80 | 234 | struct sctp_chunk *auth; |
4cd57c80 VY |
235 | |
236 | /* if we don't have an association, we can't do authentication */ | |
237 | if (!asoc) | |
238 | return retval; | |
239 | ||
240 | /* See if this is an auth chunk we are bundling or if | |
241 | * auth is already bundled. | |
242 | */ | |
4007cc88 | 243 | if (chunk->chunk_hdr->type == SCTP_CID_AUTH || pkt->has_auth) |
4cd57c80 VY |
244 | return retval; |
245 | ||
246 | /* if the peer did not request this chunk to be authenticated, | |
247 | * don't do it | |
248 | */ | |
249 | if (!chunk->auth) | |
250 | return retval; | |
251 | ||
1b1e0bc9 | 252 | auth = sctp_make_auth(asoc, chunk->shkey->key_id); |
4cd57c80 VY |
253 | if (!auth) |
254 | return retval; | |
255 | ||
1b1e0bc9 XL |
256 | auth->shkey = chunk->shkey; |
257 | sctp_auth_shkey_hold(auth->shkey); | |
258 | ||
ed106277 NH |
259 | retval = __sctp_packet_append_chunk(pkt, auth); |
260 | ||
261 | if (retval != SCTP_XMIT_OK) | |
262 | sctp_chunk_free(auth); | |
4cd57c80 VY |
263 | |
264 | return retval; | |
265 | } | |
266 | ||
1da177e4 | 267 | /* Try to bundle a SACK with the packet. */ |
86b36f2a XL |
268 | static enum sctp_xmit sctp_packet_bundle_sack(struct sctp_packet *pkt, |
269 | struct sctp_chunk *chunk) | |
1da177e4 | 270 | { |
86b36f2a | 271 | enum sctp_xmit retval = SCTP_XMIT_OK; |
1da177e4 LT |
272 | |
273 | /* If sending DATA and haven't aleady bundled a SACK, try to | |
274 | * bundle one in to the packet. | |
275 | */ | |
276 | if (sctp_chunk_is_data(chunk) && !pkt->has_sack && | |
277 | !pkt->has_cookie_echo) { | |
278 | struct sctp_association *asoc; | |
af87b823 | 279 | struct timer_list *timer; |
1da177e4 | 280 | asoc = pkt->transport->asoc; |
af87b823 | 281 | timer = &asoc->timers[SCTP_EVENT_TIMEOUT_SACK]; |
1da177e4 | 282 | |
af87b823 DG |
283 | /* If the SACK timer is running, we have a pending SACK */ |
284 | if (timer_pending(timer)) { | |
1da177e4 | 285 | struct sctp_chunk *sack; |
4244854d NH |
286 | |
287 | if (pkt->transport->sack_generation != | |
288 | pkt->transport->asoc->peer.sack_generation) | |
289 | return retval; | |
290 | ||
1da177e4 LT |
291 | asoc->a_rwnd = asoc->rwnd; |
292 | sack = sctp_make_sack(asoc); | |
293 | if (sack) { | |
ed106277 NH |
294 | retval = __sctp_packet_append_chunk(pkt, sack); |
295 | if (retval != SCTP_XMIT_OK) { | |
296 | sctp_chunk_free(sack); | |
297 | goto out; | |
298 | } | |
1da177e4 | 299 | asoc->peer.sack_needed = 0; |
af87b823 | 300 | if (del_timer(timer)) |
1da177e4 LT |
301 | sctp_association_put(asoc); |
302 | } | |
303 | } | |
304 | } | |
ed106277 | 305 | out: |
1da177e4 LT |
306 | return retval; |
307 | } | |
308 | ||
ed106277 | 309 | |
1da177e4 LT |
310 | /* Append a chunk to the offered packet reporting back any inability to do |
311 | * so. | |
312 | */ | |
86b36f2a XL |
313 | static enum sctp_xmit __sctp_packet_append_chunk(struct sctp_packet *packet, |
314 | struct sctp_chunk *chunk) | |
1da177e4 | 315 | { |
e2f036a9 | 316 | __u16 chunk_len = SCTP_PAD4(ntohs(chunk->chunk_hdr->length)); |
86b36f2a | 317 | enum sctp_xmit retval = SCTP_XMIT_OK; |
1da177e4 | 318 | |
e83963b7 VY |
319 | /* Check to see if this chunk will fit into the packet */ |
320 | retval = sctp_packet_will_fit(packet, chunk, chunk_len); | |
321 | if (retval != SCTP_XMIT_OK) | |
322 | goto finish; | |
1da177e4 | 323 | |
e83963b7 | 324 | /* We believe that this chunk is OK to add to the packet */ |
4cd57c80 | 325 | switch (chunk->chunk_hdr->type) { |
f7010e61 | 326 | case SCTP_CID_DATA: |
668c9beb | 327 | case SCTP_CID_I_DATA: |
e83963b7 VY |
328 | /* Account for the data being in the packet */ |
329 | sctp_packet_append_data(packet, chunk); | |
1da177e4 LT |
330 | /* Disallow SACK bundling after DATA. */ |
331 | packet->has_sack = 1; | |
4cd57c80 VY |
332 | /* Disallow AUTH bundling after DATA */ |
333 | packet->has_auth = 1; | |
334 | /* Let it be knows that packet has DATA in it */ | |
335 | packet->has_data = 1; | |
759af00e VY |
336 | /* timestamp the chunk for rtx purposes */ |
337 | chunk->sent_at = jiffies; | |
a6c2f792 XL |
338 | /* Mainly used for prsctp RTX policy */ |
339 | chunk->sent_count++; | |
4cd57c80 | 340 | break; |
f7010e61 | 341 | case SCTP_CID_COOKIE_ECHO: |
1da177e4 | 342 | packet->has_cookie_echo = 1; |
4cd57c80 VY |
343 | break; |
344 | ||
f7010e61 | 345 | case SCTP_CID_SACK: |
1da177e4 | 346 | packet->has_sack = 1; |
196d6759 MB |
347 | if (chunk->asoc) |
348 | chunk->asoc->stats.osacks++; | |
4cd57c80 VY |
349 | break; |
350 | ||
f7010e61 | 351 | case SCTP_CID_AUTH: |
4cd57c80 VY |
352 | packet->has_auth = 1; |
353 | packet->auth = chunk; | |
354 | break; | |
355 | } | |
1da177e4 LT |
356 | |
357 | /* It is OK to send this chunk. */ | |
79af02c2 | 358 | list_add_tail(&chunk->list, &packet->chunk_list); |
1da177e4 LT |
359 | packet->size += chunk_len; |
360 | chunk->transport = packet->transport; | |
361 | finish: | |
362 | return retval; | |
363 | } | |
364 | ||
ed106277 NH |
365 | /* Append a chunk to the offered packet reporting back any inability to do |
366 | * so. | |
367 | */ | |
86b36f2a XL |
368 | enum sctp_xmit sctp_packet_append_chunk(struct sctp_packet *packet, |
369 | struct sctp_chunk *chunk) | |
ed106277 | 370 | { |
86b36f2a | 371 | enum sctp_xmit retval = SCTP_XMIT_OK; |
ed106277 | 372 | |
bb33381d | 373 | pr_debug("%s: packet:%p chunk:%p\n", __func__, packet, chunk); |
ed106277 NH |
374 | |
375 | /* Data chunks are special. Before seeing what else we can | |
376 | * bundle into this packet, check to see if we are allowed to | |
377 | * send this DATA. | |
378 | */ | |
379 | if (sctp_chunk_is_data(chunk)) { | |
380 | retval = sctp_packet_can_append_data(packet, chunk); | |
381 | if (retval != SCTP_XMIT_OK) | |
382 | goto finish; | |
383 | } | |
384 | ||
385 | /* Try to bundle AUTH chunk */ | |
386 | retval = sctp_packet_bundle_auth(packet, chunk); | |
387 | if (retval != SCTP_XMIT_OK) | |
388 | goto finish; | |
389 | ||
390 | /* Try to bundle SACK chunk */ | |
391 | retval = sctp_packet_bundle_sack(packet, chunk); | |
392 | if (retval != SCTP_XMIT_OK) | |
393 | goto finish; | |
394 | ||
395 | retval = __sctp_packet_append_chunk(packet, chunk); | |
396 | ||
397 | finish: | |
398 | return retval; | |
399 | } | |
400 | ||
4c3a5bda TG |
401 | static void sctp_packet_release_owner(struct sk_buff *skb) |
402 | { | |
403 | sk_free(skb->sk); | |
404 | } | |
405 | ||
406 | static void sctp_packet_set_owner_w(struct sk_buff *skb, struct sock *sk) | |
407 | { | |
408 | skb_orphan(skb); | |
409 | skb->sk = sk; | |
410 | skb->destructor = sctp_packet_release_owner; | |
411 | ||
412 | /* | |
413 | * The data chunks have already been accounted for in sctp_sendmsg(), | |
414 | * therefore only reserve a single byte to keep socket around until | |
415 | * the packet has been transmitted. | |
416 | */ | |
14afee4b | 417 | refcount_inc(&sk->sk_wmem_alloc); |
4c3a5bda TG |
418 | } |
419 | ||
e4ff952a XL |
420 | static int sctp_packet_pack(struct sctp_packet *packet, |
421 | struct sk_buff *head, int gso, gfp_t gfp) | |
1da177e4 LT |
422 | { |
423 | struct sctp_transport *tp = packet->transport; | |
e4ff952a | 424 | struct sctp_auth_chunk *auth = NULL; |
79af02c2 | 425 | struct sctp_chunk *chunk, *tmp; |
e4ff952a XL |
426 | int pkt_count = 0, pkt_size; |
427 | struct sock *sk = head->sk; | |
428 | struct sk_buff *nskb; | |
ecc515d7 | 429 | int auth_len = 0; |
1da177e4 | 430 | |
90017acc | 431 | if (gso) { |
90017acc | 432 | skb_shinfo(head)->gso_type = sk->sk_gso_type; |
e4ff952a XL |
433 | NAPI_GRO_CB(head)->last = head; |
434 | } else { | |
435 | nskb = head; | |
436 | pkt_size = packet->size; | |
437 | goto merge; | |
90017acc | 438 | } |
1da177e4 | 439 | |
90017acc | 440 | do { |
e4ff952a XL |
441 | /* calculate the pkt_size and alloc nskb */ |
442 | pkt_size = packet->overhead; | |
443 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, | |
444 | list) { | |
445 | int padded = SCTP_PAD4(chunk->skb->len); | |
6eabca54 | 446 | |
e4ff952a XL |
447 | if (chunk == packet->auth) |
448 | auth_len = padded; | |
449 | else if (auth_len + padded + packet->overhead > | |
450 | tp->pathmtu) | |
451 | return 0; | |
452 | else if (pkt_size + padded > tp->pathmtu) | |
453 | break; | |
454 | pkt_size += padded; | |
90017acc | 455 | } |
e4ff952a XL |
456 | nskb = alloc_skb(pkt_size + MAX_HEADER, gfp); |
457 | if (!nskb) | |
458 | return 0; | |
459 | skb_reserve(nskb, packet->overhead + MAX_HEADER); | |
1da177e4 | 460 | |
e4ff952a XL |
461 | merge: |
462 | /* merge chunks into nskb and append nskb into head list */ | |
90017acc MRL |
463 | pkt_size -= packet->overhead; |
464 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { | |
e4ff952a XL |
465 | int padding; |
466 | ||
90017acc MRL |
467 | list_del_init(&chunk->list); |
468 | if (sctp_chunk_is_data(chunk)) { | |
cc6ac9bc XL |
469 | if (!sctp_chunk_retransmitted(chunk) && |
470 | !tp->rto_pending) { | |
90017acc MRL |
471 | chunk->rtt_in_progress = 1; |
472 | tp->rto_pending = 1; | |
473 | } | |
90017acc MRL |
474 | } |
475 | ||
e2f036a9 | 476 | padding = SCTP_PAD4(chunk->skb->len) - chunk->skb->len; |
90017acc | 477 | if (padding) |
b080db58 | 478 | skb_put_zero(chunk->skb, padding); |
90017acc | 479 | |
90017acc | 480 | if (chunk == packet->auth) |
e4ff952a XL |
481 | auth = (struct sctp_auth_chunk *) |
482 | skb_tail_pointer(nskb); | |
90017acc | 483 | |
59ae1d12 | 484 | skb_put_data(nskb, chunk->skb->data, chunk->skb->len); |
1da177e4 | 485 | |
90017acc MRL |
486 | pr_debug("*** Chunk:%p[%s] %s 0x%x, length:%d, chunk->skb->len:%d, rtt_in_progress:%d\n", |
487 | chunk, | |
488 | sctp_cname(SCTP_ST_CHUNK(chunk->chunk_hdr->type)), | |
489 | chunk->has_tsn ? "TSN" : "No TSN", | |
490 | chunk->has_tsn ? ntohl(chunk->subh.data_hdr->tsn) : 0, | |
491 | ntohs(chunk->chunk_hdr->length), chunk->skb->len, | |
492 | chunk->rtt_in_progress); | |
493 | ||
e2f036a9 | 494 | pkt_size -= SCTP_PAD4(chunk->skb->len); |
1da177e4 | 495 | |
f1533cce | 496 | if (!sctp_chunk_is_data(chunk) && chunk != packet->auth) |
90017acc MRL |
497 | sctp_chunk_free(chunk); |
498 | ||
499 | if (!pkt_size) | |
500 | break; | |
501 | } | |
502 | ||
e4ff952a | 503 | if (auth) { |
1b1e0bc9 XL |
504 | sctp_auth_calculate_hmac(tp->asoc, nskb, auth, |
505 | packet->auth->shkey, gfp); | |
e4ff952a XL |
506 | /* free auth if no more chunks, or add it back */ |
507 | if (list_empty(&packet->chunk_list)) | |
508 | sctp_chunk_free(packet->auth); | |
509 | else | |
1aa25ec2 XL |
510 | list_add(&packet->auth->list, |
511 | &packet->chunk_list); | |
f1533cce MRL |
512 | } |
513 | ||
e4ff952a XL |
514 | if (gso) { |
515 | if (skb_gro_receive(&head, nskb)) { | |
516 | kfree_skb(nskb); | |
517 | return 0; | |
518 | } | |
519 | if (WARN_ON_ONCE(skb_shinfo(head)->gso_segs >= | |
520 | sk->sk_gso_max_segs)) | |
521 | return 0; | |
41001faf | 522 | } |
e4ff952a XL |
523 | |
524 | pkt_count++; | |
90017acc | 525 | } while (!list_empty(&packet->chunk_list)); |
4cd57c80 | 526 | |
e4ff952a XL |
527 | if (gso) { |
528 | memset(head->cb, 0, max(sizeof(struct inet_skb_parm), | |
529 | sizeof(struct inet6_skb_parm))); | |
530 | skb_shinfo(head)->gso_segs = pkt_count; | |
531 | skb_shinfo(head)->gso_size = GSO_BY_FRAGS; | |
532 | rcu_read_lock(); | |
533 | if (skb_dst(head) != tp->dst) { | |
534 | dst_hold(tp->dst); | |
535 | sk_setup_caps(sk, tp->dst); | |
8dc92f7e | 536 | } |
e4ff952a XL |
537 | rcu_read_unlock(); |
538 | goto chksum; | |
8dc92f7e | 539 | } |
1da177e4 | 540 | |
e4ff952a XL |
541 | if (sctp_checksum_disable) |
542 | return 1; | |
1da177e4 | 543 | |
e4ff952a XL |
544 | if (!(skb_dst(head)->dev->features & NETIF_F_SCTP_CRC) || |
545 | dst_xfrm(skb_dst(head)) || packet->ipfragok) { | |
546 | struct sctphdr *sh = | |
547 | (struct sctphdr *)skb_transport_header(head); | |
1da177e4 | 548 | |
e4ff952a XL |
549 | sh->checksum = sctp_compute_cksum(head, 0); |
550 | } else { | |
551 | chksum: | |
552 | head->ip_summed = CHECKSUM_PARTIAL; | |
dba00306 | 553 | head->csum_not_inet = 1; |
e4ff952a XL |
554 | head->csum_start = skb_transport_header(head) - head->head; |
555 | head->csum_offset = offsetof(struct sctphdr, checksum); | |
1da177e4 LT |
556 | } |
557 | ||
e4ff952a XL |
558 | return pkt_count; |
559 | } | |
560 | ||
561 | /* All packets are sent to the network through this function from | |
562 | * sctp_outq_tail(). | |
563 | * | |
564 | * The return value is always 0 for now. | |
565 | */ | |
566 | int sctp_packet_transmit(struct sctp_packet *packet, gfp_t gfp) | |
567 | { | |
568 | struct sctp_transport *tp = packet->transport; | |
569 | struct sctp_association *asoc = tp->asoc; | |
570 | struct sctp_chunk *chunk, *tmp; | |
571 | int pkt_count, gso = 0; | |
572 | struct dst_entry *dst; | |
573 | struct sk_buff *head; | |
574 | struct sctphdr *sh; | |
575 | struct sock *sk; | |
1da177e4 | 576 | |
e4ff952a XL |
577 | pr_debug("%s: packet:%p\n", __func__, packet); |
578 | if (list_empty(&packet->chunk_list)) | |
579 | return 0; | |
580 | chunk = list_entry(packet->chunk_list.next, struct sctp_chunk, list); | |
581 | sk = chunk->skb->sk; | |
1da177e4 | 582 | |
e4ff952a XL |
583 | /* check gso */ |
584 | if (packet->size > tp->pathmtu && !packet->ipfragok) { | |
585 | if (!sk_can_gso(sk)) { | |
586 | pr_err_once("Trying to GSO but underlying device doesn't support it."); | |
587 | goto out; | |
1da177e4 | 588 | } |
e4ff952a XL |
589 | gso = 1; |
590 | } | |
591 | ||
592 | /* alloc head skb */ | |
593 | head = alloc_skb((gso ? packet->overhead : packet->size) + | |
594 | MAX_HEADER, gfp); | |
595 | if (!head) | |
596 | goto out; | |
597 | skb_reserve(head, packet->overhead + MAX_HEADER); | |
598 | sctp_packet_set_owner_w(head, sk); | |
599 | ||
600 | /* set sctp header */ | |
d58ff351 | 601 | sh = skb_push(head, sizeof(struct sctphdr)); |
e4ff952a XL |
602 | skb_reset_transport_header(head); |
603 | sh->source = htons(packet->source_port); | |
604 | sh->dest = htons(packet->destination_port); | |
605 | sh->vtag = htonl(packet->vtag); | |
606 | sh->checksum = 0; | |
607 | ||
df2729c3 | 608 | /* drop packet if no dst */ |
e4ff952a XL |
609 | dst = dst_clone(tp->dst); |
610 | if (!dst) { | |
611 | IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES); | |
612 | kfree_skb(head); | |
613 | goto out; | |
614 | } | |
615 | skb_dst_set(head, dst); | |
1da177e4 | 616 | |
e4ff952a XL |
617 | /* pack up chunks */ |
618 | pkt_count = sctp_packet_pack(packet, head, gso, gfp); | |
619 | if (!pkt_count) { | |
620 | kfree_skb(head); | |
621 | goto out; | |
622 | } | |
90017acc MRL |
623 | pr_debug("***sctp_transmit_packet*** skb->len:%d\n", head->len); |
624 | ||
e4ff952a XL |
625 | /* start autoclose timer */ |
626 | if (packet->has_data && sctp_state(asoc, ESTABLISHED) && | |
627 | asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE]) { | |
628 | struct timer_list *timer = | |
629 | &asoc->timers[SCTP_EVENT_TIMEOUT_AUTOCLOSE]; | |
630 | unsigned long timeout = | |
631 | asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE]; | |
1da177e4 | 632 | |
e4ff952a XL |
633 | if (!mod_timer(timer, jiffies + timeout)) |
634 | sctp_association_hold(asoc); | |
635 | } | |
90017acc | 636 | |
e4ff952a XL |
637 | /* sctp xmit */ |
638 | tp->af_specific->ecn_capable(sk); | |
639 | if (asoc) { | |
640 | asoc->stats.opackets += pkt_count; | |
641 | if (asoc->peer.last_sent_to != tp) | |
642 | asoc->peer.last_sent_to = tp; | |
90017acc MRL |
643 | } |
644 | head->ignore_df = packet->ipfragok; | |
486a43db | 645 | if (tp->dst_pending_confirm) |
c86a773c JA |
646 | skb_set_dst_pending_confirm(head, 1); |
647 | /* neighbour should be confirmed on successful transmission or | |
648 | * positive error | |
649 | */ | |
486a43db XL |
650 | if (tp->af_specific->sctp_xmit(head, tp) >= 0 && |
651 | tp->dst_pending_confirm) | |
c86a773c | 652 | tp->dst_pending_confirm = 0; |
1da177e4 | 653 | |
e4ff952a | 654 | out: |
79af02c2 DM |
655 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { |
656 | list_del_init(&chunk->list); | |
1da177e4 | 657 | if (!sctp_chunk_is_data(chunk)) |
d808ad9a | 658 | sctp_chunk_free(chunk); |
1da177e4 | 659 | } |
41001faf | 660 | sctp_packet_reset(packet); |
e4ff952a | 661 | return 0; |
1da177e4 LT |
662 | } |
663 | ||
664 | /******************************************************************** | |
665 | * 2nd Level Abstractions | |
666 | ********************************************************************/ | |
667 | ||
e83963b7 | 668 | /* This private function check to see if a chunk can be added */ |
86b36f2a XL |
669 | static enum sctp_xmit sctp_packet_can_append_data(struct sctp_packet *packet, |
670 | struct sctp_chunk *chunk) | |
1da177e4 | 671 | { |
e83963b7 | 672 | size_t datasize, rwnd, inflight, flight_size; |
1da177e4 | 673 | struct sctp_transport *transport = packet->transport; |
1da177e4 | 674 | struct sctp_association *asoc = transport->asoc; |
1da177e4 LT |
675 | struct sctp_outq *q = &asoc->outqueue; |
676 | ||
677 | /* RFC 2960 6.1 Transmission of DATA Chunks | |
678 | * | |
679 | * A) At any given time, the data sender MUST NOT transmit new data to | |
680 | * any destination transport address if its peer's rwnd indicates | |
681 | * that the peer has no buffer space (i.e. rwnd is 0, see Section | |
682 | * 6.2.1). However, regardless of the value of rwnd (including if it | |
683 | * is 0), the data sender can always have one DATA chunk in flight to | |
684 | * the receiver if allowed by cwnd (see rule B below). This rule | |
685 | * allows the sender to probe for a change in rwnd that the sender | |
686 | * missed due to the SACK having been lost in transit from the data | |
687 | * receiver to the data sender. | |
688 | */ | |
689 | ||
690 | rwnd = asoc->peer.rwnd; | |
e83963b7 VY |
691 | inflight = q->outstanding_bytes; |
692 | flight_size = transport->flight_size; | |
1da177e4 LT |
693 | |
694 | datasize = sctp_data_size(chunk); | |
695 | ||
723189fa DL |
696 | if (datasize > rwnd && inflight > 0) |
697 | /* We have (at least) one data chunk in flight, | |
698 | * so we can't fall back to rule 6.1 B). | |
699 | */ | |
700 | return SCTP_XMIT_RWND_FULL; | |
1da177e4 | 701 | |
1da177e4 LT |
702 | /* RFC 2960 6.1 Transmission of DATA Chunks |
703 | * | |
704 | * B) At any given time, the sender MUST NOT transmit new data | |
705 | * to a given transport address if it has cwnd or more bytes | |
706 | * of data outstanding to that transport address. | |
707 | */ | |
708 | /* RFC 7.2.4 & the Implementers Guide 2.8. | |
709 | * | |
710 | * 3) ... | |
711 | * When a Fast Retransmit is being performed the sender SHOULD | |
712 | * ignore the value of cwnd and SHOULD NOT delay retransmission. | |
713 | */ | |
723189fa DL |
714 | if (chunk->fast_retransmit != SCTP_NEED_FRTX && |
715 | flight_size >= transport->cwnd) | |
716 | return SCTP_XMIT_RWND_FULL; | |
1da177e4 LT |
717 | |
718 | /* Nagle's algorithm to solve small-packet problem: | |
719 | * Inhibit the sending of new chunks when new outgoing data arrives | |
720 | * if any previously transmitted data on the connection remains | |
721 | * unacknowledged. | |
722 | */ | |
1da177e4 | 723 | |
4ea0c32f | 724 | if ((sctp_sk(asoc->base.sk)->nodelay || inflight == 0) && |
f9ba3501 | 725 | !asoc->force_delay) |
4ea0c32f | 726 | /* Nothing unacked */ |
723189fa DL |
727 | return SCTP_XMIT_OK; |
728 | ||
729 | if (!sctp_packet_empty(packet)) | |
730 | /* Append to packet */ | |
731 | return SCTP_XMIT_OK; | |
732 | ||
723189fa DL |
733 | if (!sctp_state(asoc, ESTABLISHED)) |
734 | return SCTP_XMIT_OK; | |
735 | ||
736 | /* Check whether this chunk and all the rest of pending data will fit | |
737 | * or delay in hopes of bundling a full sized packet. | |
738 | */ | |
9f8d3147 | 739 | if (chunk->skb->len + q->out_qlen > transport->pathmtu - |
668c9beb | 740 | packet->overhead - sctp_datachk_len(&chunk->asoc->stream) - 4) |
723189fa DL |
741 | /* Enough data queued to fill a packet */ |
742 | return SCTP_XMIT_OK; | |
743 | ||
744 | /* Don't delay large message writes that may have been fragmented */ | |
745 | if (!chunk->msg->can_delay) | |
746 | return SCTP_XMIT_OK; | |
747 | ||
748 | /* Defer until all data acked or packet full */ | |
526cbef7 | 749 | return SCTP_XMIT_DELAY; |
e83963b7 VY |
750 | } |
751 | ||
752 | /* This private function does management things when adding DATA chunk */ | |
753 | static void sctp_packet_append_data(struct sctp_packet *packet, | |
754 | struct sctp_chunk *chunk) | |
755 | { | |
756 | struct sctp_transport *transport = packet->transport; | |
757 | size_t datasize = sctp_data_size(chunk); | |
758 | struct sctp_association *asoc = transport->asoc; | |
759 | u32 rwnd = asoc->peer.rwnd; | |
760 | ||
1da177e4 LT |
761 | /* Keep track of how many bytes are in flight over this transport. */ |
762 | transport->flight_size += datasize; | |
763 | ||
764 | /* Keep track of how many bytes are in flight to the receiver. */ | |
765 | asoc->outqueue.outstanding_bytes += datasize; | |
766 | ||
a76c0adf | 767 | /* Update our view of the receiver's rwnd. */ |
1da177e4 LT |
768 | if (datasize < rwnd) |
769 | rwnd -= datasize; | |
770 | else | |
771 | rwnd = 0; | |
772 | ||
773 | asoc->peer.rwnd = rwnd; | |
d8dd1578 | 774 | sctp_chunk_assign_tsn(chunk); |
668c9beb | 775 | asoc->stream.si->assign_number(chunk); |
e83963b7 VY |
776 | } |
777 | ||
86b36f2a XL |
778 | static enum sctp_xmit sctp_packet_will_fit(struct sctp_packet *packet, |
779 | struct sctp_chunk *chunk, | |
780 | u16 chunk_len) | |
e83963b7 | 781 | { |
86b36f2a | 782 | enum sctp_xmit retval = SCTP_XMIT_OK; |
7303a147 | 783 | size_t psize, pmtu, maxsize; |
e83963b7 | 784 | |
1b1e0bc9 XL |
785 | /* Don't bundle in this packet if this chunk's auth key doesn't |
786 | * match other chunks already enqueued on this packet. Also, | |
787 | * don't bundle the chunk with auth key if other chunks in this | |
788 | * packet don't have auth key. | |
789 | */ | |
790 | if ((packet->auth && chunk->shkey != packet->auth->shkey) || | |
791 | (!packet->auth && chunk->shkey && | |
792 | chunk->chunk_hdr->type != SCTP_CID_AUTH)) | |
793 | return SCTP_XMIT_PMTU_FULL; | |
794 | ||
e83963b7 | 795 | psize = packet->size; |
90017acc MRL |
796 | if (packet->transport->asoc) |
797 | pmtu = packet->transport->asoc->pathmtu; | |
798 | else | |
799 | pmtu = packet->transport->pathmtu; | |
e83963b7 VY |
800 | |
801 | /* Decide if we need to fragment or resubmit later. */ | |
90017acc MRL |
802 | if (psize + chunk_len > pmtu) { |
803 | /* It's OK to fragment at IP level if any one of the following | |
e83963b7 | 804 | * is true: |
90017acc MRL |
805 | * 1. The packet is empty (meaning this chunk is greater |
806 | * the MTU) | |
807 | * 2. The packet doesn't have any data in it yet and data | |
808 | * requires authentication. | |
e83963b7 | 809 | */ |
90017acc | 810 | if (sctp_packet_empty(packet) || |
e83963b7 VY |
811 | (!packet->has_data && chunk->auth)) { |
812 | /* We no longer do re-fragmentation. | |
813 | * Just fragment at the IP layer, if we | |
814 | * actually hit this condition | |
815 | */ | |
816 | packet->ipfragok = 1; | |
90017acc | 817 | goto out; |
e83963b7 | 818 | } |
90017acc | 819 | |
7303a147 MRL |
820 | /* Similarly, if this chunk was built before a PMTU |
821 | * reduction, we have to fragment it at IP level now. So | |
822 | * if the packet already contains something, we need to | |
823 | * flush. | |
824 | */ | |
825 | maxsize = pmtu - packet->overhead; | |
826 | if (packet->auth) | |
e2f036a9 | 827 | maxsize -= SCTP_PAD4(packet->auth->skb->len); |
7303a147 MRL |
828 | if (chunk_len > maxsize) |
829 | retval = SCTP_XMIT_PMTU_FULL; | |
830 | ||
90017acc MRL |
831 | /* It is also okay to fragment if the chunk we are |
832 | * adding is a control chunk, but only if current packet | |
833 | * is not a GSO one otherwise it causes fragmentation of | |
834 | * a large frame. So in this case we allow the | |
835 | * fragmentation by forcing it to be in a new packet. | |
836 | */ | |
837 | if (!sctp_chunk_is_data(chunk) && packet->has_data) | |
838 | retval = SCTP_XMIT_PMTU_FULL; | |
839 | ||
840 | if (psize + chunk_len > packet->max_size) | |
841 | /* Hit GSO/PMTU limit, gotta flush */ | |
842 | retval = SCTP_XMIT_PMTU_FULL; | |
843 | ||
844 | if (!packet->transport->burst_limited && | |
845 | psize + chunk_len > (packet->transport->cwnd >> 1)) | |
846 | /* Do not allow a single GSO packet to use more | |
847 | * than half of cwnd. | |
848 | */ | |
849 | retval = SCTP_XMIT_PMTU_FULL; | |
850 | ||
851 | if (packet->transport->burst_limited && | |
852 | psize + chunk_len > (packet->transport->burst_limited >> 1)) | |
853 | /* Do not allow a single GSO packet to use more | |
854 | * than half of original cwnd. | |
855 | */ | |
856 | retval = SCTP_XMIT_PMTU_FULL; | |
857 | /* Otherwise it will fit in the GSO packet */ | |
e83963b7 | 858 | } |
1da177e4 | 859 | |
90017acc | 860 | out: |
1da177e4 LT |
861 | return retval; |
862 | } |