]>
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. */ | |
ed106277 NH |
60 | static sctp_xmit_t __sctp_packet_append_chunk(struct sctp_packet *packet, |
61 | struct sctp_chunk *chunk); | |
e83963b7 | 62 | static sctp_xmit_t sctp_packet_can_append_data(struct sctp_packet *packet, |
1da177e4 | 63 | struct sctp_chunk *chunk); |
e83963b7 VY |
64 | static void sctp_packet_append_data(struct sctp_packet *packet, |
65 | struct sctp_chunk *chunk); | |
66 | static sctp_xmit_t 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 | */ | |
84 | struct sctp_packet *sctp_packet_config(struct sctp_packet *packet, | |
85 | __u32 vtag, int ecn_capable) | |
86 | { | |
90017acc MRL |
87 | struct sctp_transport *tp = packet->transport; |
88 | struct sctp_association *asoc = tp->asoc; | |
1da177e4 | 89 | |
bb33381d | 90 | pr_debug("%s: packet:%p vtag:0x%x\n", __func__, packet, vtag); |
1da177e4 LT |
91 | |
92 | packet->vtag = vtag; | |
1da177e4 | 93 | |
90017acc MRL |
94 | if (asoc && tp->dst) { |
95 | struct sock *sk = asoc->base.sk; | |
96 | ||
97 | rcu_read_lock(); | |
98 | if (__sk_dst_get(sk) != tp->dst) { | |
99 | dst_hold(tp->dst); | |
100 | sk_setup_caps(sk, tp->dst); | |
101 | } | |
102 | ||
103 | if (sk_can_gso(sk)) { | |
104 | struct net_device *dev = tp->dst->dev; | |
105 | ||
106 | packet->max_size = dev->gso_max_size; | |
107 | } else { | |
108 | packet->max_size = asoc->pathmtu; | |
109 | } | |
110 | rcu_read_unlock(); | |
111 | ||
112 | } else { | |
113 | packet->max_size = tp->pathmtu; | |
114 | } | |
115 | ||
1da177e4 | 116 | if (ecn_capable && sctp_packet_empty(packet)) { |
90017acc | 117 | struct sctp_chunk *chunk; |
1da177e4 LT |
118 | |
119 | /* If there a is a prepend chunk stick it on the list before | |
d808ad9a YH |
120 | * any other chunks get appended. |
121 | */ | |
90017acc | 122 | chunk = sctp_get_ecne_prepend(asoc); |
1da177e4 LT |
123 | if (chunk) |
124 | sctp_packet_append_chunk(packet, chunk); | |
125 | } | |
126 | ||
127 | return packet; | |
128 | } | |
129 | ||
130 | /* Initialize the packet structure. */ | |
131 | struct sctp_packet *sctp_packet_init(struct sctp_packet *packet, | |
132 | struct sctp_transport *transport, | |
133 | __u16 sport, __u16 dport) | |
134 | { | |
135 | struct sctp_association *asoc = transport->asoc; | |
136 | size_t overhead; | |
137 | ||
bb33381d | 138 | pr_debug("%s: packet:%p transport:%p\n", __func__, packet, transport); |
1da177e4 LT |
139 | |
140 | packet->transport = transport; | |
141 | packet->source_port = sport; | |
142 | packet->destination_port = dport; | |
79af02c2 | 143 | INIT_LIST_HEAD(&packet->chunk_list); |
1da177e4 | 144 | if (asoc) { |
d808ad9a YH |
145 | struct sctp_sock *sp = sctp_sk(asoc->base.sk); |
146 | overhead = sp->pf->af->net_header_len; | |
1da177e4 LT |
147 | } else { |
148 | overhead = sizeof(struct ipv6hdr); | |
149 | } | |
150 | overhead += sizeof(struct sctphdr); | |
151 | packet->overhead = overhead; | |
be297143 | 152 | sctp_packet_reset(packet); |
1da177e4 | 153 | packet->vtag = 0; |
3e3251b3 | 154 | |
1da177e4 LT |
155 | return packet; |
156 | } | |
157 | ||
158 | /* Free a packet. */ | |
159 | void sctp_packet_free(struct sctp_packet *packet) | |
160 | { | |
79af02c2 | 161 | struct sctp_chunk *chunk, *tmp; |
1da177e4 | 162 | |
bb33381d | 163 | pr_debug("%s: packet:%p\n", __func__, packet); |
1da177e4 | 164 | |
79af02c2 DM |
165 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { |
166 | list_del_init(&chunk->list); | |
1da177e4 | 167 | sctp_chunk_free(chunk); |
79af02c2 | 168 | } |
1da177e4 LT |
169 | } |
170 | ||
171 | /* This routine tries to append the chunk to the offered packet. If adding | |
172 | * the chunk causes the packet to exceed the path MTU and COOKIE_ECHO chunk | |
173 | * is not present in the packet, it transmits the input packet. | |
174 | * Data can be bundled with a packet containing a COOKIE_ECHO chunk as long | |
175 | * as it can fit in the packet, but any more data that does not fit in this | |
176 | * packet can be sent only after receiving the COOKIE_ACK. | |
177 | */ | |
178 | sctp_xmit_t sctp_packet_transmit_chunk(struct sctp_packet *packet, | |
2e3216cd | 179 | struct sctp_chunk *chunk, |
cea8768f | 180 | int one_packet, gfp_t gfp) |
1da177e4 LT |
181 | { |
182 | sctp_xmit_t retval; | |
1da177e4 | 183 | |
3b55a537 | 184 | pr_debug("%s: packet:%p size:%Zu chunk:%p size:%d\n", __func__, |
942b3235 | 185 | packet, packet->size, chunk, chunk->skb ? chunk->skb->len : -1); |
1da177e4 LT |
186 | |
187 | switch ((retval = (sctp_packet_append_chunk(packet, chunk)))) { | |
188 | case SCTP_XMIT_PMTU_FULL: | |
189 | if (!packet->has_cookie_echo) { | |
64519440 XL |
190 | int error = 0; |
191 | ||
cea8768f | 192 | error = sctp_packet_transmit(packet, gfp); |
1da177e4 LT |
193 | if (error < 0) |
194 | chunk->skb->sk->sk_err = -error; | |
195 | ||
196 | /* If we have an empty packet, then we can NOT ever | |
197 | * return PMTU_FULL. | |
198 | */ | |
2e3216cd VY |
199 | if (!one_packet) |
200 | retval = sctp_packet_append_chunk(packet, | |
201 | chunk); | |
1da177e4 LT |
202 | } |
203 | break; | |
204 | ||
205 | case SCTP_XMIT_RWND_FULL: | |
206 | case SCTP_XMIT_OK: | |
526cbef7 | 207 | case SCTP_XMIT_DELAY: |
1da177e4 | 208 | break; |
3ff50b79 | 209 | } |
1da177e4 LT |
210 | |
211 | return retval; | |
212 | } | |
213 | ||
4cd57c80 VY |
214 | /* Try to bundle an auth chunk into the packet. */ |
215 | static sctp_xmit_t sctp_packet_bundle_auth(struct sctp_packet *pkt, | |
216 | struct sctp_chunk *chunk) | |
217 | { | |
218 | struct sctp_association *asoc = pkt->transport->asoc; | |
219 | struct sctp_chunk *auth; | |
220 | sctp_xmit_t retval = SCTP_XMIT_OK; | |
221 | ||
222 | /* if we don't have an association, we can't do authentication */ | |
223 | if (!asoc) | |
224 | return retval; | |
225 | ||
226 | /* See if this is an auth chunk we are bundling or if | |
227 | * auth is already bundled. | |
228 | */ | |
4007cc88 | 229 | if (chunk->chunk_hdr->type == SCTP_CID_AUTH || pkt->has_auth) |
4cd57c80 VY |
230 | return retval; |
231 | ||
232 | /* if the peer did not request this chunk to be authenticated, | |
233 | * don't do it | |
234 | */ | |
235 | if (!chunk->auth) | |
236 | return retval; | |
237 | ||
238 | auth = sctp_make_auth(asoc); | |
239 | if (!auth) | |
240 | return retval; | |
241 | ||
ed106277 NH |
242 | retval = __sctp_packet_append_chunk(pkt, auth); |
243 | ||
244 | if (retval != SCTP_XMIT_OK) | |
245 | sctp_chunk_free(auth); | |
4cd57c80 VY |
246 | |
247 | return retval; | |
248 | } | |
249 | ||
1da177e4 LT |
250 | /* Try to bundle a SACK with the packet. */ |
251 | static sctp_xmit_t sctp_packet_bundle_sack(struct sctp_packet *pkt, | |
252 | struct sctp_chunk *chunk) | |
253 | { | |
254 | sctp_xmit_t retval = SCTP_XMIT_OK; | |
255 | ||
256 | /* If sending DATA and haven't aleady bundled a SACK, try to | |
257 | * bundle one in to the packet. | |
258 | */ | |
259 | if (sctp_chunk_is_data(chunk) && !pkt->has_sack && | |
260 | !pkt->has_cookie_echo) { | |
261 | struct sctp_association *asoc; | |
af87b823 | 262 | struct timer_list *timer; |
1da177e4 | 263 | asoc = pkt->transport->asoc; |
af87b823 | 264 | timer = &asoc->timers[SCTP_EVENT_TIMEOUT_SACK]; |
1da177e4 | 265 | |
af87b823 DG |
266 | /* If the SACK timer is running, we have a pending SACK */ |
267 | if (timer_pending(timer)) { | |
1da177e4 | 268 | struct sctp_chunk *sack; |
4244854d NH |
269 | |
270 | if (pkt->transport->sack_generation != | |
271 | pkt->transport->asoc->peer.sack_generation) | |
272 | return retval; | |
273 | ||
1da177e4 LT |
274 | asoc->a_rwnd = asoc->rwnd; |
275 | sack = sctp_make_sack(asoc); | |
276 | if (sack) { | |
ed106277 NH |
277 | retval = __sctp_packet_append_chunk(pkt, sack); |
278 | if (retval != SCTP_XMIT_OK) { | |
279 | sctp_chunk_free(sack); | |
280 | goto out; | |
281 | } | |
1da177e4 | 282 | asoc->peer.sack_needed = 0; |
af87b823 | 283 | if (del_timer(timer)) |
1da177e4 LT |
284 | sctp_association_put(asoc); |
285 | } | |
286 | } | |
287 | } | |
ed106277 | 288 | out: |
1da177e4 LT |
289 | return retval; |
290 | } | |
291 | ||
ed106277 | 292 | |
1da177e4 LT |
293 | /* Append a chunk to the offered packet reporting back any inability to do |
294 | * so. | |
295 | */ | |
ed106277 NH |
296 | static sctp_xmit_t __sctp_packet_append_chunk(struct sctp_packet *packet, |
297 | struct sctp_chunk *chunk) | |
1da177e4 LT |
298 | { |
299 | sctp_xmit_t retval = SCTP_XMIT_OK; | |
e2f036a9 | 300 | __u16 chunk_len = SCTP_PAD4(ntohs(chunk->chunk_hdr->length)); |
1da177e4 | 301 | |
e83963b7 VY |
302 | /* Check to see if this chunk will fit into the packet */ |
303 | retval = sctp_packet_will_fit(packet, chunk, chunk_len); | |
304 | if (retval != SCTP_XMIT_OK) | |
305 | goto finish; | |
1da177e4 | 306 | |
e83963b7 | 307 | /* We believe that this chunk is OK to add to the packet */ |
4cd57c80 | 308 | switch (chunk->chunk_hdr->type) { |
f7010e61 | 309 | case SCTP_CID_DATA: |
e83963b7 VY |
310 | /* Account for the data being in the packet */ |
311 | sctp_packet_append_data(packet, chunk); | |
1da177e4 LT |
312 | /* Disallow SACK bundling after DATA. */ |
313 | packet->has_sack = 1; | |
4cd57c80 VY |
314 | /* Disallow AUTH bundling after DATA */ |
315 | packet->has_auth = 1; | |
316 | /* Let it be knows that packet has DATA in it */ | |
317 | packet->has_data = 1; | |
759af00e VY |
318 | /* timestamp the chunk for rtx purposes */ |
319 | chunk->sent_at = jiffies; | |
a6c2f792 XL |
320 | /* Mainly used for prsctp RTX policy */ |
321 | chunk->sent_count++; | |
4cd57c80 | 322 | break; |
f7010e61 | 323 | case SCTP_CID_COOKIE_ECHO: |
1da177e4 | 324 | packet->has_cookie_echo = 1; |
4cd57c80 VY |
325 | break; |
326 | ||
f7010e61 | 327 | case SCTP_CID_SACK: |
1da177e4 | 328 | packet->has_sack = 1; |
196d6759 MB |
329 | if (chunk->asoc) |
330 | chunk->asoc->stats.osacks++; | |
4cd57c80 VY |
331 | break; |
332 | ||
f7010e61 | 333 | case SCTP_CID_AUTH: |
4cd57c80 VY |
334 | packet->has_auth = 1; |
335 | packet->auth = chunk; | |
336 | break; | |
337 | } | |
1da177e4 LT |
338 | |
339 | /* It is OK to send this chunk. */ | |
79af02c2 | 340 | list_add_tail(&chunk->list, &packet->chunk_list); |
1da177e4 LT |
341 | packet->size += chunk_len; |
342 | chunk->transport = packet->transport; | |
343 | finish: | |
344 | return retval; | |
345 | } | |
346 | ||
ed106277 NH |
347 | /* Append a chunk to the offered packet reporting back any inability to do |
348 | * so. | |
349 | */ | |
350 | sctp_xmit_t sctp_packet_append_chunk(struct sctp_packet *packet, | |
351 | struct sctp_chunk *chunk) | |
352 | { | |
353 | sctp_xmit_t retval = SCTP_XMIT_OK; | |
354 | ||
bb33381d | 355 | pr_debug("%s: packet:%p chunk:%p\n", __func__, packet, chunk); |
ed106277 NH |
356 | |
357 | /* Data chunks are special. Before seeing what else we can | |
358 | * bundle into this packet, check to see if we are allowed to | |
359 | * send this DATA. | |
360 | */ | |
361 | if (sctp_chunk_is_data(chunk)) { | |
362 | retval = sctp_packet_can_append_data(packet, chunk); | |
363 | if (retval != SCTP_XMIT_OK) | |
364 | goto finish; | |
365 | } | |
366 | ||
367 | /* Try to bundle AUTH chunk */ | |
368 | retval = sctp_packet_bundle_auth(packet, chunk); | |
369 | if (retval != SCTP_XMIT_OK) | |
370 | goto finish; | |
371 | ||
372 | /* Try to bundle SACK chunk */ | |
373 | retval = sctp_packet_bundle_sack(packet, chunk); | |
374 | if (retval != SCTP_XMIT_OK) | |
375 | goto finish; | |
376 | ||
377 | retval = __sctp_packet_append_chunk(packet, chunk); | |
378 | ||
379 | finish: | |
380 | return retval; | |
381 | } | |
382 | ||
4c3a5bda TG |
383 | static void sctp_packet_release_owner(struct sk_buff *skb) |
384 | { | |
385 | sk_free(skb->sk); | |
386 | } | |
387 | ||
388 | static void sctp_packet_set_owner_w(struct sk_buff *skb, struct sock *sk) | |
389 | { | |
390 | skb_orphan(skb); | |
391 | skb->sk = sk; | |
392 | skb->destructor = sctp_packet_release_owner; | |
393 | ||
394 | /* | |
395 | * The data chunks have already been accounted for in sctp_sendmsg(), | |
396 | * therefore only reserve a single byte to keep socket around until | |
397 | * the packet has been transmitted. | |
398 | */ | |
399 | atomic_inc(&sk->sk_wmem_alloc); | |
400 | } | |
401 | ||
e4ff952a XL |
402 | static int sctp_packet_pack(struct sctp_packet *packet, |
403 | struct sk_buff *head, int gso, gfp_t gfp) | |
1da177e4 LT |
404 | { |
405 | struct sctp_transport *tp = packet->transport; | |
e4ff952a | 406 | struct sctp_auth_chunk *auth = NULL; |
79af02c2 | 407 | struct sctp_chunk *chunk, *tmp; |
e4ff952a XL |
408 | int pkt_count = 0, pkt_size; |
409 | struct sock *sk = head->sk; | |
410 | struct sk_buff *nskb; | |
ecc515d7 | 411 | int auth_len = 0; |
1da177e4 | 412 | |
90017acc | 413 | if (gso) { |
90017acc | 414 | skb_shinfo(head)->gso_type = sk->sk_gso_type; |
e4ff952a XL |
415 | NAPI_GRO_CB(head)->last = head; |
416 | } else { | |
417 | nskb = head; | |
418 | pkt_size = packet->size; | |
419 | goto merge; | |
90017acc | 420 | } |
1da177e4 | 421 | |
90017acc | 422 | do { |
e4ff952a XL |
423 | /* calculate the pkt_size and alloc nskb */ |
424 | pkt_size = packet->overhead; | |
425 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, | |
426 | list) { | |
427 | int padded = SCTP_PAD4(chunk->skb->len); | |
6eabca54 | 428 | |
e4ff952a XL |
429 | if (chunk == packet->auth) |
430 | auth_len = padded; | |
431 | else if (auth_len + padded + packet->overhead > | |
432 | tp->pathmtu) | |
433 | return 0; | |
434 | else if (pkt_size + padded > tp->pathmtu) | |
435 | break; | |
436 | pkt_size += padded; | |
90017acc | 437 | } |
e4ff952a XL |
438 | nskb = alloc_skb(pkt_size + MAX_HEADER, gfp); |
439 | if (!nskb) | |
440 | return 0; | |
441 | skb_reserve(nskb, packet->overhead + MAX_HEADER); | |
1da177e4 | 442 | |
e4ff952a XL |
443 | merge: |
444 | /* merge chunks into nskb and append nskb into head list */ | |
90017acc MRL |
445 | pkt_size -= packet->overhead; |
446 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { | |
e4ff952a XL |
447 | int padding; |
448 | ||
90017acc MRL |
449 | list_del_init(&chunk->list); |
450 | if (sctp_chunk_is_data(chunk)) { | |
cc6ac9bc XL |
451 | if (!sctp_chunk_retransmitted(chunk) && |
452 | !tp->rto_pending) { | |
90017acc MRL |
453 | chunk->rtt_in_progress = 1; |
454 | tp->rto_pending = 1; | |
455 | } | |
90017acc MRL |
456 | } |
457 | ||
e2f036a9 | 458 | padding = SCTP_PAD4(chunk->skb->len) - chunk->skb->len; |
90017acc MRL |
459 | if (padding) |
460 | memset(skb_put(chunk->skb, padding), 0, padding); | |
461 | ||
90017acc | 462 | if (chunk == packet->auth) |
e4ff952a XL |
463 | auth = (struct sctp_auth_chunk *) |
464 | skb_tail_pointer(nskb); | |
90017acc | 465 | |
e4ff952a XL |
466 | memcpy(skb_put(nskb, chunk->skb->len), chunk->skb->data, |
467 | chunk->skb->len); | |
1da177e4 | 468 | |
90017acc MRL |
469 | pr_debug("*** Chunk:%p[%s] %s 0x%x, length:%d, chunk->skb->len:%d, rtt_in_progress:%d\n", |
470 | chunk, | |
471 | sctp_cname(SCTP_ST_CHUNK(chunk->chunk_hdr->type)), | |
472 | chunk->has_tsn ? "TSN" : "No TSN", | |
473 | chunk->has_tsn ? ntohl(chunk->subh.data_hdr->tsn) : 0, | |
474 | ntohs(chunk->chunk_hdr->length), chunk->skb->len, | |
475 | chunk->rtt_in_progress); | |
476 | ||
e2f036a9 | 477 | pkt_size -= SCTP_PAD4(chunk->skb->len); |
1da177e4 | 478 | |
f1533cce | 479 | if (!sctp_chunk_is_data(chunk) && chunk != packet->auth) |
90017acc MRL |
480 | sctp_chunk_free(chunk); |
481 | ||
482 | if (!pkt_size) | |
483 | break; | |
484 | } | |
485 | ||
e4ff952a XL |
486 | if (auth) { |
487 | sctp_auth_calculate_hmac(tp->asoc, nskb, auth, gfp); | |
488 | /* free auth if no more chunks, or add it back */ | |
489 | if (list_empty(&packet->chunk_list)) | |
490 | sctp_chunk_free(packet->auth); | |
491 | else | |
1aa25ec2 XL |
492 | list_add(&packet->auth->list, |
493 | &packet->chunk_list); | |
f1533cce MRL |
494 | } |
495 | ||
e4ff952a XL |
496 | if (gso) { |
497 | if (skb_gro_receive(&head, nskb)) { | |
498 | kfree_skb(nskb); | |
499 | return 0; | |
500 | } | |
501 | if (WARN_ON_ONCE(skb_shinfo(head)->gso_segs >= | |
502 | sk->sk_gso_max_segs)) | |
503 | return 0; | |
41001faf | 504 | } |
e4ff952a XL |
505 | |
506 | pkt_count++; | |
90017acc | 507 | } while (!list_empty(&packet->chunk_list)); |
4cd57c80 | 508 | |
e4ff952a XL |
509 | if (gso) { |
510 | memset(head->cb, 0, max(sizeof(struct inet_skb_parm), | |
511 | sizeof(struct inet6_skb_parm))); | |
512 | skb_shinfo(head)->gso_segs = pkt_count; | |
513 | skb_shinfo(head)->gso_size = GSO_BY_FRAGS; | |
514 | rcu_read_lock(); | |
515 | if (skb_dst(head) != tp->dst) { | |
516 | dst_hold(tp->dst); | |
517 | sk_setup_caps(sk, tp->dst); | |
8dc92f7e | 518 | } |
e4ff952a XL |
519 | rcu_read_unlock(); |
520 | goto chksum; | |
8dc92f7e | 521 | } |
1da177e4 | 522 | |
e4ff952a XL |
523 | if (sctp_checksum_disable) |
524 | return 1; | |
1da177e4 | 525 | |
e4ff952a XL |
526 | if (!(skb_dst(head)->dev->features & NETIF_F_SCTP_CRC) || |
527 | dst_xfrm(skb_dst(head)) || packet->ipfragok) { | |
528 | struct sctphdr *sh = | |
529 | (struct sctphdr *)skb_transport_header(head); | |
1da177e4 | 530 | |
e4ff952a XL |
531 | sh->checksum = sctp_compute_cksum(head, 0); |
532 | } else { | |
533 | chksum: | |
534 | head->ip_summed = CHECKSUM_PARTIAL; | |
535 | head->csum_start = skb_transport_header(head) - head->head; | |
536 | head->csum_offset = offsetof(struct sctphdr, checksum); | |
1da177e4 LT |
537 | } |
538 | ||
e4ff952a XL |
539 | return pkt_count; |
540 | } | |
541 | ||
542 | /* All packets are sent to the network through this function from | |
543 | * sctp_outq_tail(). | |
544 | * | |
545 | * The return value is always 0 for now. | |
546 | */ | |
547 | int sctp_packet_transmit(struct sctp_packet *packet, gfp_t gfp) | |
548 | { | |
549 | struct sctp_transport *tp = packet->transport; | |
550 | struct sctp_association *asoc = tp->asoc; | |
551 | struct sctp_chunk *chunk, *tmp; | |
552 | int pkt_count, gso = 0; | |
553 | struct dst_entry *dst; | |
554 | struct sk_buff *head; | |
555 | struct sctphdr *sh; | |
556 | struct sock *sk; | |
1da177e4 | 557 | |
e4ff952a XL |
558 | pr_debug("%s: packet:%p\n", __func__, packet); |
559 | if (list_empty(&packet->chunk_list)) | |
560 | return 0; | |
561 | chunk = list_entry(packet->chunk_list.next, struct sctp_chunk, list); | |
562 | sk = chunk->skb->sk; | |
1da177e4 | 563 | |
e4ff952a XL |
564 | /* check gso */ |
565 | if (packet->size > tp->pathmtu && !packet->ipfragok) { | |
566 | if (!sk_can_gso(sk)) { | |
567 | pr_err_once("Trying to GSO but underlying device doesn't support it."); | |
568 | goto out; | |
1da177e4 | 569 | } |
e4ff952a XL |
570 | gso = 1; |
571 | } | |
572 | ||
573 | /* alloc head skb */ | |
574 | head = alloc_skb((gso ? packet->overhead : packet->size) + | |
575 | MAX_HEADER, gfp); | |
576 | if (!head) | |
577 | goto out; | |
578 | skb_reserve(head, packet->overhead + MAX_HEADER); | |
579 | sctp_packet_set_owner_w(head, sk); | |
580 | ||
581 | /* set sctp header */ | |
582 | sh = (struct sctphdr *)skb_push(head, sizeof(struct sctphdr)); | |
583 | skb_reset_transport_header(head); | |
584 | sh->source = htons(packet->source_port); | |
585 | sh->dest = htons(packet->destination_port); | |
586 | sh->vtag = htonl(packet->vtag); | |
587 | sh->checksum = 0; | |
588 | ||
589 | /* update dst if in need */ | |
590 | if (!sctp_transport_dst_check(tp)) { | |
591 | sctp_transport_route(tp, NULL, sctp_sk(sk)); | |
592 | if (asoc && asoc->param_flags & SPP_PMTUD_ENABLE) | |
593 | sctp_assoc_sync_pmtu(sk, asoc); | |
1da177e4 | 594 | } |
e4ff952a XL |
595 | dst = dst_clone(tp->dst); |
596 | if (!dst) { | |
597 | IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES); | |
598 | kfree_skb(head); | |
599 | goto out; | |
600 | } | |
601 | skb_dst_set(head, dst); | |
1da177e4 | 602 | |
e4ff952a XL |
603 | /* pack up chunks */ |
604 | pkt_count = sctp_packet_pack(packet, head, gso, gfp); | |
605 | if (!pkt_count) { | |
606 | kfree_skb(head); | |
607 | goto out; | |
608 | } | |
90017acc MRL |
609 | pr_debug("***sctp_transmit_packet*** skb->len:%d\n", head->len); |
610 | ||
e4ff952a XL |
611 | /* start autoclose timer */ |
612 | if (packet->has_data && sctp_state(asoc, ESTABLISHED) && | |
613 | asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE]) { | |
614 | struct timer_list *timer = | |
615 | &asoc->timers[SCTP_EVENT_TIMEOUT_AUTOCLOSE]; | |
616 | unsigned long timeout = | |
617 | asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE]; | |
1da177e4 | 618 | |
e4ff952a XL |
619 | if (!mod_timer(timer, jiffies + timeout)) |
620 | sctp_association_hold(asoc); | |
621 | } | |
90017acc | 622 | |
e4ff952a XL |
623 | /* sctp xmit */ |
624 | tp->af_specific->ecn_capable(sk); | |
625 | if (asoc) { | |
626 | asoc->stats.opackets += pkt_count; | |
627 | if (asoc->peer.last_sent_to != tp) | |
628 | asoc->peer.last_sent_to = tp; | |
90017acc MRL |
629 | } |
630 | head->ignore_df = packet->ipfragok; | |
631 | tp->af_specific->sctp_xmit(head, tp); | |
1da177e4 | 632 | |
e4ff952a | 633 | out: |
79af02c2 DM |
634 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { |
635 | list_del_init(&chunk->list); | |
1da177e4 | 636 | if (!sctp_chunk_is_data(chunk)) |
d808ad9a | 637 | sctp_chunk_free(chunk); |
1da177e4 | 638 | } |
41001faf | 639 | sctp_packet_reset(packet); |
e4ff952a | 640 | return 0; |
1da177e4 LT |
641 | } |
642 | ||
643 | /******************************************************************** | |
644 | * 2nd Level Abstractions | |
645 | ********************************************************************/ | |
646 | ||
e83963b7 VY |
647 | /* This private function check to see if a chunk can be added */ |
648 | static sctp_xmit_t sctp_packet_can_append_data(struct sctp_packet *packet, | |
1da177e4 LT |
649 | struct sctp_chunk *chunk) |
650 | { | |
e83963b7 | 651 | size_t datasize, rwnd, inflight, flight_size; |
1da177e4 | 652 | struct sctp_transport *transport = packet->transport; |
1da177e4 | 653 | struct sctp_association *asoc = transport->asoc; |
1da177e4 LT |
654 | struct sctp_outq *q = &asoc->outqueue; |
655 | ||
656 | /* RFC 2960 6.1 Transmission of DATA Chunks | |
657 | * | |
658 | * A) At any given time, the data sender MUST NOT transmit new data to | |
659 | * any destination transport address if its peer's rwnd indicates | |
660 | * that the peer has no buffer space (i.e. rwnd is 0, see Section | |
661 | * 6.2.1). However, regardless of the value of rwnd (including if it | |
662 | * is 0), the data sender can always have one DATA chunk in flight to | |
663 | * the receiver if allowed by cwnd (see rule B below). This rule | |
664 | * allows the sender to probe for a change in rwnd that the sender | |
665 | * missed due to the SACK having been lost in transit from the data | |
666 | * receiver to the data sender. | |
667 | */ | |
668 | ||
669 | rwnd = asoc->peer.rwnd; | |
e83963b7 VY |
670 | inflight = q->outstanding_bytes; |
671 | flight_size = transport->flight_size; | |
1da177e4 LT |
672 | |
673 | datasize = sctp_data_size(chunk); | |
674 | ||
723189fa DL |
675 | if (datasize > rwnd && inflight > 0) |
676 | /* We have (at least) one data chunk in flight, | |
677 | * so we can't fall back to rule 6.1 B). | |
678 | */ | |
679 | return SCTP_XMIT_RWND_FULL; | |
1da177e4 | 680 | |
1da177e4 LT |
681 | /* RFC 2960 6.1 Transmission of DATA Chunks |
682 | * | |
683 | * B) At any given time, the sender MUST NOT transmit new data | |
684 | * to a given transport address if it has cwnd or more bytes | |
685 | * of data outstanding to that transport address. | |
686 | */ | |
687 | /* RFC 7.2.4 & the Implementers Guide 2.8. | |
688 | * | |
689 | * 3) ... | |
690 | * When a Fast Retransmit is being performed the sender SHOULD | |
691 | * ignore the value of cwnd and SHOULD NOT delay retransmission. | |
692 | */ | |
723189fa DL |
693 | if (chunk->fast_retransmit != SCTP_NEED_FRTX && |
694 | flight_size >= transport->cwnd) | |
695 | return SCTP_XMIT_RWND_FULL; | |
1da177e4 LT |
696 | |
697 | /* Nagle's algorithm to solve small-packet problem: | |
698 | * Inhibit the sending of new chunks when new outgoing data arrives | |
699 | * if any previously transmitted data on the connection remains | |
700 | * unacknowledged. | |
701 | */ | |
1da177e4 | 702 | |
723189fa DL |
703 | if (sctp_sk(asoc->base.sk)->nodelay) |
704 | /* Nagle disabled */ | |
705 | return SCTP_XMIT_OK; | |
706 | ||
707 | if (!sctp_packet_empty(packet)) | |
708 | /* Append to packet */ | |
709 | return SCTP_XMIT_OK; | |
710 | ||
711 | if (inflight == 0) | |
712 | /* Nothing unacked */ | |
713 | return SCTP_XMIT_OK; | |
714 | ||
715 | if (!sctp_state(asoc, ESTABLISHED)) | |
716 | return SCTP_XMIT_OK; | |
717 | ||
718 | /* Check whether this chunk and all the rest of pending data will fit | |
719 | * or delay in hopes of bundling a full sized packet. | |
720 | */ | |
e43569e6 MRL |
721 | if (chunk->skb->len + q->out_qlen > |
722 | transport->pathmtu - packet->overhead - sizeof(sctp_data_chunk_t) - 4) | |
723189fa DL |
723 | /* Enough data queued to fill a packet */ |
724 | return SCTP_XMIT_OK; | |
725 | ||
726 | /* Don't delay large message writes that may have been fragmented */ | |
727 | if (!chunk->msg->can_delay) | |
728 | return SCTP_XMIT_OK; | |
729 | ||
730 | /* Defer until all data acked or packet full */ | |
526cbef7 | 731 | return SCTP_XMIT_DELAY; |
e83963b7 VY |
732 | } |
733 | ||
734 | /* This private function does management things when adding DATA chunk */ | |
735 | static void sctp_packet_append_data(struct sctp_packet *packet, | |
736 | struct sctp_chunk *chunk) | |
737 | { | |
738 | struct sctp_transport *transport = packet->transport; | |
739 | size_t datasize = sctp_data_size(chunk); | |
740 | struct sctp_association *asoc = transport->asoc; | |
741 | u32 rwnd = asoc->peer.rwnd; | |
742 | ||
1da177e4 LT |
743 | /* Keep track of how many bytes are in flight over this transport. */ |
744 | transport->flight_size += datasize; | |
745 | ||
746 | /* Keep track of how many bytes are in flight to the receiver. */ | |
747 | asoc->outqueue.outstanding_bytes += datasize; | |
748 | ||
a76c0adf | 749 | /* Update our view of the receiver's rwnd. */ |
1da177e4 LT |
750 | if (datasize < rwnd) |
751 | rwnd -= datasize; | |
752 | else | |
753 | rwnd = 0; | |
754 | ||
755 | asoc->peer.rwnd = rwnd; | |
d8dd1578 NH |
756 | sctp_chunk_assign_tsn(chunk); |
757 | sctp_chunk_assign_ssn(chunk); | |
e83963b7 VY |
758 | } |
759 | ||
760 | static sctp_xmit_t sctp_packet_will_fit(struct sctp_packet *packet, | |
761 | struct sctp_chunk *chunk, | |
762 | u16 chunk_len) | |
763 | { | |
7303a147 | 764 | size_t psize, pmtu, maxsize; |
e83963b7 VY |
765 | sctp_xmit_t retval = SCTP_XMIT_OK; |
766 | ||
767 | psize = packet->size; | |
90017acc MRL |
768 | if (packet->transport->asoc) |
769 | pmtu = packet->transport->asoc->pathmtu; | |
770 | else | |
771 | pmtu = packet->transport->pathmtu; | |
e83963b7 VY |
772 | |
773 | /* Decide if we need to fragment or resubmit later. */ | |
90017acc MRL |
774 | if (psize + chunk_len > pmtu) { |
775 | /* It's OK to fragment at IP level if any one of the following | |
e83963b7 | 776 | * is true: |
90017acc MRL |
777 | * 1. The packet is empty (meaning this chunk is greater |
778 | * the MTU) | |
779 | * 2. The packet doesn't have any data in it yet and data | |
780 | * requires authentication. | |
e83963b7 | 781 | */ |
90017acc | 782 | if (sctp_packet_empty(packet) || |
e83963b7 VY |
783 | (!packet->has_data && chunk->auth)) { |
784 | /* We no longer do re-fragmentation. | |
785 | * Just fragment at the IP layer, if we | |
786 | * actually hit this condition | |
787 | */ | |
788 | packet->ipfragok = 1; | |
90017acc | 789 | goto out; |
e83963b7 | 790 | } |
90017acc | 791 | |
7303a147 MRL |
792 | /* Similarly, if this chunk was built before a PMTU |
793 | * reduction, we have to fragment it at IP level now. So | |
794 | * if the packet already contains something, we need to | |
795 | * flush. | |
796 | */ | |
797 | maxsize = pmtu - packet->overhead; | |
798 | if (packet->auth) | |
e2f036a9 | 799 | maxsize -= SCTP_PAD4(packet->auth->skb->len); |
7303a147 MRL |
800 | if (chunk_len > maxsize) |
801 | retval = SCTP_XMIT_PMTU_FULL; | |
802 | ||
90017acc MRL |
803 | /* It is also okay to fragment if the chunk we are |
804 | * adding is a control chunk, but only if current packet | |
805 | * is not a GSO one otherwise it causes fragmentation of | |
806 | * a large frame. So in this case we allow the | |
807 | * fragmentation by forcing it to be in a new packet. | |
808 | */ | |
809 | if (!sctp_chunk_is_data(chunk) && packet->has_data) | |
810 | retval = SCTP_XMIT_PMTU_FULL; | |
811 | ||
812 | if (psize + chunk_len > packet->max_size) | |
813 | /* Hit GSO/PMTU limit, gotta flush */ | |
814 | retval = SCTP_XMIT_PMTU_FULL; | |
815 | ||
816 | if (!packet->transport->burst_limited && | |
817 | psize + chunk_len > (packet->transport->cwnd >> 1)) | |
818 | /* Do not allow a single GSO packet to use more | |
819 | * than half of cwnd. | |
820 | */ | |
821 | retval = SCTP_XMIT_PMTU_FULL; | |
822 | ||
823 | if (packet->transport->burst_limited && | |
824 | psize + chunk_len > (packet->transport->burst_limited >> 1)) | |
825 | /* Do not allow a single GSO packet to use more | |
826 | * than half of original cwnd. | |
827 | */ | |
828 | retval = SCTP_XMIT_PMTU_FULL; | |
829 | /* Otherwise it will fit in the GSO packet */ | |
e83963b7 | 830 | } |
1da177e4 | 831 | |
90017acc | 832 | out: |
1da177e4 LT |
833 | return retval; |
834 | } |