]>
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 | ||
1da177e4 LT |
402 | /* All packets are sent to the network through this function from |
403 | * sctp_outq_tail(). | |
404 | * | |
405 | * The return value is a normal kernel error return value. | |
406 | */ | |
cea8768f | 407 | int sctp_packet_transmit(struct sctp_packet *packet, gfp_t gfp) |
1da177e4 LT |
408 | { |
409 | struct sctp_transport *tp = packet->transport; | |
410 | struct sctp_association *asoc = tp->asoc; | |
411 | struct sctphdr *sh; | |
90017acc | 412 | struct sk_buff *nskb = NULL, *head = NULL; |
79af02c2 | 413 | struct sctp_chunk *chunk, *tmp; |
1da177e4 LT |
414 | struct sock *sk; |
415 | int err = 0; | |
416 | int padding; /* How much padding do we need? */ | |
90017acc | 417 | int pkt_size; |
1da177e4 | 418 | __u8 has_data = 0; |
90017acc MRL |
419 | int gso = 0; |
420 | int pktcount = 0; | |
0438816e | 421 | struct dst_entry *dst; |
4cd57c80 | 422 | unsigned char *auth = NULL; /* pointer to auth in skb data */ |
1da177e4 | 423 | |
bb33381d | 424 | pr_debug("%s: packet:%p\n", __func__, packet); |
1da177e4 LT |
425 | |
426 | /* Do NOT generate a chunkless packet. */ | |
79af02c2 | 427 | if (list_empty(&packet->chunk_list)) |
1da177e4 LT |
428 | return err; |
429 | ||
430 | /* Set up convenience variables... */ | |
79af02c2 | 431 | chunk = list_entry(packet->chunk_list.next, struct sctp_chunk, list); |
1da177e4 LT |
432 | sk = chunk->skb->sk; |
433 | ||
90017acc MRL |
434 | /* Allocate the head skb, or main one if not in GSO */ |
435 | if (packet->size > tp->pathmtu && !packet->ipfragok) { | |
436 | if (sk_can_gso(sk)) { | |
437 | gso = 1; | |
438 | pkt_size = packet->overhead; | |
439 | } else { | |
440 | /* If this happens, we trash this packet and try | |
441 | * to build a new one, hopefully correct this | |
442 | * time. Application may notice this error. | |
443 | */ | |
444 | pr_err_once("Trying to GSO but underlying device doesn't support it."); | |
41001faf | 445 | goto err; |
90017acc MRL |
446 | } |
447 | } else { | |
448 | pkt_size = packet->size; | |
449 | } | |
450 | head = alloc_skb(pkt_size + MAX_HEADER, gfp); | |
451 | if (!head) | |
41001faf | 452 | goto err; |
90017acc MRL |
453 | if (gso) { |
454 | NAPI_GRO_CB(head)->last = head; | |
455 | skb_shinfo(head)->gso_type = sk->sk_gso_type; | |
456 | } | |
1da177e4 LT |
457 | |
458 | /* Make sure the outbound skb has enough header room reserved. */ | |
90017acc | 459 | skb_reserve(head, packet->overhead + MAX_HEADER); |
1da177e4 LT |
460 | |
461 | /* Set the owning socket so that we know where to get the | |
462 | * destination IP address. | |
463 | */ | |
90017acc | 464 | sctp_packet_set_owner_w(head, sk); |
1da177e4 | 465 | |
e0268868 | 466 | if (!sctp_transport_dst_check(tp)) { |
503b55fd SS |
467 | sctp_transport_route(tp, NULL, sctp_sk(sk)); |
468 | if (asoc && (asoc->param_flags & SPP_PMTUD_ENABLE)) { | |
02f3d4ce | 469 | sctp_assoc_sync_pmtu(sk, asoc); |
503b55fd SS |
470 | } |
471 | } | |
adf30907 | 472 | dst = dst_clone(tp->dst); |
41001faf XL |
473 | if (!dst) { |
474 | if (asoc) | |
475 | IP_INC_STATS(sock_net(asoc->base.sk), | |
476 | IPSTATS_MIB_OUTNOROUTES); | |
477 | goto nodst; | |
478 | } | |
90017acc | 479 | skb_dst_set(head, dst); |
503b55fd | 480 | |
1da177e4 | 481 | /* Build the SCTP header. */ |
90017acc MRL |
482 | sh = (struct sctphdr *)skb_push(head, sizeof(struct sctphdr)); |
483 | skb_reset_transport_header(head); | |
1da177e4 LT |
484 | sh->source = htons(packet->source_port); |
485 | sh->dest = htons(packet->destination_port); | |
486 | ||
487 | /* From 6.8 Adler-32 Checksum Calculation: | |
488 | * After the packet is constructed (containing the SCTP common | |
489 | * header and one or more control or DATA chunks), the | |
490 | * transmitter shall: | |
491 | * | |
492 | * 1) Fill in the proper Verification Tag in the SCTP common | |
493 | * header and initialize the checksum field to 0's. | |
494 | */ | |
495 | sh->vtag = htonl(packet->vtag); | |
496 | sh->checksum = 0; | |
497 | ||
bb33381d DB |
498 | pr_debug("***sctp_transmit_packet***\n"); |
499 | ||
90017acc MRL |
500 | do { |
501 | /* Set up convenience variables... */ | |
502 | chunk = list_entry(packet->chunk_list.next, struct sctp_chunk, list); | |
503 | pktcount++; | |
1da177e4 | 504 | |
90017acc MRL |
505 | /* Calculate packet size, so it fits in PMTU. Leave |
506 | * other chunks for the next packets. | |
507 | */ | |
508 | if (gso) { | |
509 | pkt_size = packet->overhead; | |
510 | list_for_each_entry(chunk, &packet->chunk_list, list) { | |
e2f036a9 | 511 | int padded = SCTP_PAD4(chunk->skb->len); |
90017acc MRL |
512 | |
513 | if (pkt_size + padded > tp->pathmtu) | |
514 | break; | |
515 | pkt_size += padded; | |
d8dd1578 | 516 | } |
6eabca54 | 517 | |
90017acc MRL |
518 | /* Allocate a new skb. */ |
519 | nskb = alloc_skb(pkt_size + MAX_HEADER, gfp); | |
520 | if (!nskb) | |
521 | goto nomem; | |
1da177e4 | 522 | |
90017acc MRL |
523 | /* Make sure the outbound skb has enough header |
524 | * room reserved. | |
525 | */ | |
526 | skb_reserve(nskb, packet->overhead + MAX_HEADER); | |
527 | } else { | |
528 | nskb = head; | |
529 | } | |
1da177e4 | 530 | |
90017acc MRL |
531 | /** |
532 | * 3.2 Chunk Field Descriptions | |
533 | * | |
534 | * The total length of a chunk (including Type, Length and | |
535 | * Value fields) MUST be a multiple of 4 bytes. If the length | |
536 | * of the chunk is not a multiple of 4 bytes, the sender MUST | |
537 | * pad the chunk with all zero bytes and this padding is not | |
538 | * included in the chunk length field. The sender should | |
539 | * never pad with more than 3 bytes. | |
540 | * | |
e2f036a9 | 541 | * [This whole comment explains SCTP_PAD4() below.] |
4cd57c80 | 542 | */ |
4cd57c80 | 543 | |
90017acc MRL |
544 | pkt_size -= packet->overhead; |
545 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { | |
546 | list_del_init(&chunk->list); | |
547 | if (sctp_chunk_is_data(chunk)) { | |
548 | /* 6.3.1 C4) When data is in flight and when allowed | |
549 | * by rule C5, a new RTT measurement MUST be made each | |
550 | * round trip. Furthermore, new RTT measurements | |
551 | * SHOULD be made no more than once per round-trip | |
552 | * for a given destination transport address. | |
553 | */ | |
554 | ||
555 | if (!chunk->resent && !tp->rto_pending) { | |
556 | chunk->rtt_in_progress = 1; | |
557 | tp->rto_pending = 1; | |
558 | } | |
559 | ||
560 | has_data = 1; | |
561 | } | |
562 | ||
e2f036a9 | 563 | padding = SCTP_PAD4(chunk->skb->len) - chunk->skb->len; |
90017acc MRL |
564 | if (padding) |
565 | memset(skb_put(chunk->skb, padding), 0, padding); | |
566 | ||
567 | /* if this is the auth chunk that we are adding, | |
568 | * store pointer where it will be added and put | |
569 | * the auth into the packet. | |
570 | */ | |
571 | if (chunk == packet->auth) | |
572 | auth = skb_tail_pointer(nskb); | |
573 | ||
574 | memcpy(skb_put(nskb, chunk->skb->len), | |
503b55fd | 575 | chunk->skb->data, chunk->skb->len); |
1da177e4 | 576 | |
90017acc MRL |
577 | pr_debug("*** Chunk:%p[%s] %s 0x%x, length:%d, chunk->skb->len:%d, rtt_in_progress:%d\n", |
578 | chunk, | |
579 | sctp_cname(SCTP_ST_CHUNK(chunk->chunk_hdr->type)), | |
580 | chunk->has_tsn ? "TSN" : "No TSN", | |
581 | chunk->has_tsn ? ntohl(chunk->subh.data_hdr->tsn) : 0, | |
582 | ntohs(chunk->chunk_hdr->length), chunk->skb->len, | |
583 | chunk->rtt_in_progress); | |
584 | ||
585 | /* If this is a control chunk, this is our last | |
586 | * reference. Free data chunks after they've been | |
587 | * acknowledged or have failed. | |
588 | * Re-queue auth chunks if needed. | |
589 | */ | |
e2f036a9 | 590 | pkt_size -= SCTP_PAD4(chunk->skb->len); |
1da177e4 | 591 | |
f1533cce | 592 | if (!sctp_chunk_is_data(chunk) && chunk != packet->auth) |
90017acc MRL |
593 | sctp_chunk_free(chunk); |
594 | ||
595 | if (!pkt_size) | |
596 | break; | |
597 | } | |
598 | ||
599 | /* SCTP-AUTH, Section 6.2 | |
600 | * The sender MUST calculate the MAC as described in RFC2104 [2] | |
601 | * using the hash function H as described by the MAC Identifier and | |
602 | * the shared association key K based on the endpoint pair shared key | |
603 | * described by the shared key identifier. The 'data' used for the | |
604 | * computation of the AUTH-chunk is given by the AUTH chunk with its | |
605 | * HMAC field set to zero (as shown in Figure 6) followed by all | |
606 | * chunks that are placed after the AUTH chunk in the SCTP packet. | |
607 | */ | |
608 | if (auth) | |
609 | sctp_auth_calculate_hmac(asoc, nskb, | |
610 | (struct sctp_auth_chunk *)auth, | |
611 | gfp); | |
612 | ||
f1533cce MRL |
613 | if (packet->auth) { |
614 | if (!list_empty(&packet->chunk_list)) { | |
615 | /* We will generate more packets, so re-queue | |
616 | * auth chunk. | |
617 | */ | |
1aa25ec2 XL |
618 | list_add(&packet->auth->list, |
619 | &packet->chunk_list); | |
f1533cce MRL |
620 | } else { |
621 | sctp_chunk_free(packet->auth); | |
622 | packet->auth = NULL; | |
623 | } | |
624 | } | |
625 | ||
90017acc MRL |
626 | if (!gso) |
627 | break; | |
628 | ||
41001faf XL |
629 | if (skb_gro_receive(&head, nskb)) { |
630 | kfree_skb(nskb); | |
90017acc | 631 | goto nomem; |
41001faf | 632 | } |
90017acc MRL |
633 | nskb = NULL; |
634 | if (WARN_ON_ONCE(skb_shinfo(head)->gso_segs >= | |
635 | sk->sk_gso_max_segs)) | |
636 | goto nomem; | |
637 | } while (!list_empty(&packet->chunk_list)); | |
4cd57c80 VY |
638 | |
639 | /* 2) Calculate the Adler-32 checksum of the whole packet, | |
640 | * including the SCTP common header and all the | |
641 | * chunks. | |
642 | * | |
643 | * Note: Adler-32 is no longer applicable, as has been replaced | |
644 | * by CRC32-C as described in <draft-ietf-tsvwg-sctpcsum-02.txt>. | |
90017acc MRL |
645 | * |
646 | * If it's a GSO packet, it's postponed to sctp_skb_segment. | |
4cd57c80 | 647 | */ |
90017acc MRL |
648 | if (!sctp_checksum_disable || gso) { |
649 | if (!gso && (!(dst->dev->features & NETIF_F_SCTP_CRC) || | |
650 | dst_xfrm(dst) || packet->ipfragok)) { | |
651 | sh->checksum = sctp_compute_cksum(head, 0); | |
b73c43f8 | 652 | } else { |
25985edc | 653 | /* no need to seed pseudo checksum for SCTP */ |
90017acc MRL |
654 | head->ip_summed = CHECKSUM_PARTIAL; |
655 | head->csum_start = skb_transport_header(head) - head->head; | |
656 | head->csum_offset = offsetof(struct sctphdr, checksum); | |
8dc92f7e JB |
657 | } |
658 | } | |
1da177e4 | 659 | |
1da177e4 LT |
660 | /* IP layer ECN support |
661 | * From RFC 2481 | |
662 | * "The ECN-Capable Transport (ECT) bit would be set by the | |
663 | * data sender to indicate that the end-points of the | |
664 | * transport protocol are ECN-capable." | |
665 | * | |
666 | * Now setting the ECT bit all the time, as it should not cause | |
667 | * any problems protocol-wise even if our peer ignores it. | |
668 | * | |
669 | * Note: The works for IPv6 layer checks this bit too later | |
670 | * in transmission. See IP6_ECN_flow_xmit(). | |
671 | */ | |
90017acc | 672 | tp->af_specific->ecn_capable(sk); |
1da177e4 LT |
673 | |
674 | /* Set up the IP options. */ | |
675 | /* BUG: not implemented | |
676 | * For v4 this all lives somewhere in sk->sk_opt... | |
677 | */ | |
678 | ||
679 | /* Dump that on IP! */ | |
196d6759 | 680 | if (asoc) { |
90017acc | 681 | asoc->stats.opackets += pktcount; |
196d6759 MB |
682 | if (asoc->peer.last_sent_to != tp) |
683 | /* Considering the multiple CPU scenario, this is a | |
684 | * "correcter" place for last_sent_to. --xguo | |
685 | */ | |
686 | asoc->peer.last_sent_to = tp; | |
1da177e4 LT |
687 | } |
688 | ||
689 | if (has_data) { | |
690 | struct timer_list *timer; | |
691 | unsigned long timeout; | |
692 | ||
1da177e4 | 693 | /* Restart the AUTOCLOSE timer when sending data. */ |
9f70f46b NH |
694 | if (sctp_state(asoc, ESTABLISHED) && |
695 | asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE]) { | |
1da177e4 LT |
696 | timer = &asoc->timers[SCTP_EVENT_TIMEOUT_AUTOCLOSE]; |
697 | timeout = asoc->timeouts[SCTP_EVENT_TIMEOUT_AUTOCLOSE]; | |
698 | ||
699 | if (!mod_timer(timer, jiffies + timeout)) | |
700 | sctp_association_hold(asoc); | |
701 | } | |
702 | } | |
703 | ||
90017acc MRL |
704 | pr_debug("***sctp_transmit_packet*** skb->len:%d\n", head->len); |
705 | ||
706 | if (gso) { | |
707 | /* Cleanup our debris for IP stacks */ | |
708 | memset(head->cb, 0, max(sizeof(struct inet_skb_parm), | |
709 | sizeof(struct inet6_skb_parm))); | |
1da177e4 | 710 | |
90017acc MRL |
711 | skb_shinfo(head)->gso_segs = pktcount; |
712 | skb_shinfo(head)->gso_size = GSO_BY_FRAGS; | |
713 | ||
714 | /* We have to refresh this in case we are xmiting to | |
715 | * more than one transport at a time | |
716 | */ | |
717 | rcu_read_lock(); | |
718 | if (__sk_dst_get(sk) != tp->dst) { | |
719 | dst_hold(tp->dst); | |
720 | sk_setup_caps(sk, tp->dst); | |
721 | } | |
722 | rcu_read_unlock(); | |
723 | } | |
724 | head->ignore_df = packet->ipfragok; | |
725 | tp->af_specific->sctp_xmit(head, tp); | |
41001faf | 726 | goto out; |
1da177e4 | 727 | |
41001faf XL |
728 | nomem: |
729 | if (packet->auth && list_empty(&packet->auth->list)) | |
730 | sctp_chunk_free(packet->auth); | |
1da177e4 | 731 | |
41001faf | 732 | nodst: |
1da177e4 LT |
733 | /* FIXME: Returning the 'err' will effect all the associations |
734 | * associated with a socket, although only one of the paths of the | |
735 | * association is unreachable. | |
736 | * The real failure of a transport or association can be passed on | |
737 | * to the user via notifications. So setting this error may not be | |
738 | * required. | |
739 | */ | |
740 | /* err = -EHOSTUNREACH; */ | |
41001faf | 741 | kfree_skb(head); |
1da177e4 | 742 | |
41001faf | 743 | err: |
79af02c2 DM |
744 | list_for_each_entry_safe(chunk, tmp, &packet->chunk_list, list) { |
745 | list_del_init(&chunk->list); | |
1da177e4 | 746 | if (!sctp_chunk_is_data(chunk)) |
d808ad9a | 747 | sctp_chunk_free(chunk); |
1da177e4 | 748 | } |
41001faf XL |
749 | |
750 | out: | |
751 | sctp_packet_reset(packet); | |
752 | return err; | |
1da177e4 LT |
753 | } |
754 | ||
755 | /******************************************************************** | |
756 | * 2nd Level Abstractions | |
757 | ********************************************************************/ | |
758 | ||
e83963b7 VY |
759 | /* This private function check to see if a chunk can be added */ |
760 | static sctp_xmit_t sctp_packet_can_append_data(struct sctp_packet *packet, | |
1da177e4 LT |
761 | struct sctp_chunk *chunk) |
762 | { | |
e83963b7 | 763 | size_t datasize, rwnd, inflight, flight_size; |
1da177e4 | 764 | struct sctp_transport *transport = packet->transport; |
1da177e4 | 765 | struct sctp_association *asoc = transport->asoc; |
1da177e4 LT |
766 | struct sctp_outq *q = &asoc->outqueue; |
767 | ||
768 | /* RFC 2960 6.1 Transmission of DATA Chunks | |
769 | * | |
770 | * A) At any given time, the data sender MUST NOT transmit new data to | |
771 | * any destination transport address if its peer's rwnd indicates | |
772 | * that the peer has no buffer space (i.e. rwnd is 0, see Section | |
773 | * 6.2.1). However, regardless of the value of rwnd (including if it | |
774 | * is 0), the data sender can always have one DATA chunk in flight to | |
775 | * the receiver if allowed by cwnd (see rule B below). This rule | |
776 | * allows the sender to probe for a change in rwnd that the sender | |
777 | * missed due to the SACK having been lost in transit from the data | |
778 | * receiver to the data sender. | |
779 | */ | |
780 | ||
781 | rwnd = asoc->peer.rwnd; | |
e83963b7 VY |
782 | inflight = q->outstanding_bytes; |
783 | flight_size = transport->flight_size; | |
1da177e4 LT |
784 | |
785 | datasize = sctp_data_size(chunk); | |
786 | ||
723189fa DL |
787 | if (datasize > rwnd && inflight > 0) |
788 | /* We have (at least) one data chunk in flight, | |
789 | * so we can't fall back to rule 6.1 B). | |
790 | */ | |
791 | return SCTP_XMIT_RWND_FULL; | |
1da177e4 | 792 | |
1da177e4 LT |
793 | /* RFC 2960 6.1 Transmission of DATA Chunks |
794 | * | |
795 | * B) At any given time, the sender MUST NOT transmit new data | |
796 | * to a given transport address if it has cwnd or more bytes | |
797 | * of data outstanding to that transport address. | |
798 | */ | |
799 | /* RFC 7.2.4 & the Implementers Guide 2.8. | |
800 | * | |
801 | * 3) ... | |
802 | * When a Fast Retransmit is being performed the sender SHOULD | |
803 | * ignore the value of cwnd and SHOULD NOT delay retransmission. | |
804 | */ | |
723189fa DL |
805 | if (chunk->fast_retransmit != SCTP_NEED_FRTX && |
806 | flight_size >= transport->cwnd) | |
807 | return SCTP_XMIT_RWND_FULL; | |
1da177e4 LT |
808 | |
809 | /* Nagle's algorithm to solve small-packet problem: | |
810 | * Inhibit the sending of new chunks when new outgoing data arrives | |
811 | * if any previously transmitted data on the connection remains | |
812 | * unacknowledged. | |
813 | */ | |
1da177e4 | 814 | |
723189fa DL |
815 | if (sctp_sk(asoc->base.sk)->nodelay) |
816 | /* Nagle disabled */ | |
817 | return SCTP_XMIT_OK; | |
818 | ||
819 | if (!sctp_packet_empty(packet)) | |
820 | /* Append to packet */ | |
821 | return SCTP_XMIT_OK; | |
822 | ||
823 | if (inflight == 0) | |
824 | /* Nothing unacked */ | |
825 | return SCTP_XMIT_OK; | |
826 | ||
827 | if (!sctp_state(asoc, ESTABLISHED)) | |
828 | return SCTP_XMIT_OK; | |
829 | ||
830 | /* Check whether this chunk and all the rest of pending data will fit | |
831 | * or delay in hopes of bundling a full sized packet. | |
832 | */ | |
e43569e6 MRL |
833 | if (chunk->skb->len + q->out_qlen > |
834 | transport->pathmtu - packet->overhead - sizeof(sctp_data_chunk_t) - 4) | |
723189fa DL |
835 | /* Enough data queued to fill a packet */ |
836 | return SCTP_XMIT_OK; | |
837 | ||
838 | /* Don't delay large message writes that may have been fragmented */ | |
839 | if (!chunk->msg->can_delay) | |
840 | return SCTP_XMIT_OK; | |
841 | ||
842 | /* Defer until all data acked or packet full */ | |
526cbef7 | 843 | return SCTP_XMIT_DELAY; |
e83963b7 VY |
844 | } |
845 | ||
846 | /* This private function does management things when adding DATA chunk */ | |
847 | static void sctp_packet_append_data(struct sctp_packet *packet, | |
848 | struct sctp_chunk *chunk) | |
849 | { | |
850 | struct sctp_transport *transport = packet->transport; | |
851 | size_t datasize = sctp_data_size(chunk); | |
852 | struct sctp_association *asoc = transport->asoc; | |
853 | u32 rwnd = asoc->peer.rwnd; | |
854 | ||
1da177e4 LT |
855 | /* Keep track of how many bytes are in flight over this transport. */ |
856 | transport->flight_size += datasize; | |
857 | ||
858 | /* Keep track of how many bytes are in flight to the receiver. */ | |
859 | asoc->outqueue.outstanding_bytes += datasize; | |
860 | ||
a76c0adf | 861 | /* Update our view of the receiver's rwnd. */ |
1da177e4 LT |
862 | if (datasize < rwnd) |
863 | rwnd -= datasize; | |
864 | else | |
865 | rwnd = 0; | |
866 | ||
867 | asoc->peer.rwnd = rwnd; | |
868 | /* Has been accepted for transmission. */ | |
869 | if (!asoc->peer.prsctp_capable) | |
870 | chunk->msg->can_abandon = 0; | |
d8dd1578 NH |
871 | sctp_chunk_assign_tsn(chunk); |
872 | sctp_chunk_assign_ssn(chunk); | |
e83963b7 VY |
873 | } |
874 | ||
875 | static sctp_xmit_t sctp_packet_will_fit(struct sctp_packet *packet, | |
876 | struct sctp_chunk *chunk, | |
877 | u16 chunk_len) | |
878 | { | |
7303a147 | 879 | size_t psize, pmtu, maxsize; |
e83963b7 VY |
880 | sctp_xmit_t retval = SCTP_XMIT_OK; |
881 | ||
882 | psize = packet->size; | |
90017acc MRL |
883 | if (packet->transport->asoc) |
884 | pmtu = packet->transport->asoc->pathmtu; | |
885 | else | |
886 | pmtu = packet->transport->pathmtu; | |
e83963b7 VY |
887 | |
888 | /* Decide if we need to fragment or resubmit later. */ | |
90017acc MRL |
889 | if (psize + chunk_len > pmtu) { |
890 | /* It's OK to fragment at IP level if any one of the following | |
e83963b7 | 891 | * is true: |
90017acc MRL |
892 | * 1. The packet is empty (meaning this chunk is greater |
893 | * the MTU) | |
894 | * 2. The packet doesn't have any data in it yet and data | |
895 | * requires authentication. | |
e83963b7 | 896 | */ |
90017acc | 897 | if (sctp_packet_empty(packet) || |
e83963b7 VY |
898 | (!packet->has_data && chunk->auth)) { |
899 | /* We no longer do re-fragmentation. | |
900 | * Just fragment at the IP layer, if we | |
901 | * actually hit this condition | |
902 | */ | |
903 | packet->ipfragok = 1; | |
90017acc | 904 | goto out; |
e83963b7 | 905 | } |
90017acc | 906 | |
7303a147 MRL |
907 | /* Similarly, if this chunk was built before a PMTU |
908 | * reduction, we have to fragment it at IP level now. So | |
909 | * if the packet already contains something, we need to | |
910 | * flush. | |
911 | */ | |
912 | maxsize = pmtu - packet->overhead; | |
913 | if (packet->auth) | |
e2f036a9 | 914 | maxsize -= SCTP_PAD4(packet->auth->skb->len); |
7303a147 MRL |
915 | if (chunk_len > maxsize) |
916 | retval = SCTP_XMIT_PMTU_FULL; | |
917 | ||
90017acc MRL |
918 | /* It is also okay to fragment if the chunk we are |
919 | * adding is a control chunk, but only if current packet | |
920 | * is not a GSO one otherwise it causes fragmentation of | |
921 | * a large frame. So in this case we allow the | |
922 | * fragmentation by forcing it to be in a new packet. | |
923 | */ | |
924 | if (!sctp_chunk_is_data(chunk) && packet->has_data) | |
925 | retval = SCTP_XMIT_PMTU_FULL; | |
926 | ||
927 | if (psize + chunk_len > packet->max_size) | |
928 | /* Hit GSO/PMTU limit, gotta flush */ | |
929 | retval = SCTP_XMIT_PMTU_FULL; | |
930 | ||
931 | if (!packet->transport->burst_limited && | |
932 | psize + chunk_len > (packet->transport->cwnd >> 1)) | |
933 | /* Do not allow a single GSO packet to use more | |
934 | * than half of cwnd. | |
935 | */ | |
936 | retval = SCTP_XMIT_PMTU_FULL; | |
937 | ||
938 | if (packet->transport->burst_limited && | |
939 | psize + chunk_len > (packet->transport->burst_limited >> 1)) | |
940 | /* Do not allow a single GSO packet to use more | |
941 | * than half of original cwnd. | |
942 | */ | |
943 | retval = SCTP_XMIT_PMTU_FULL; | |
944 | /* Otherwise it will fit in the GSO packet */ | |
e83963b7 | 945 | } |
1da177e4 | 946 | |
90017acc | 947 | out: |
1da177e4 LT |
948 | return retval; |
949 | } |