Commit | Line | Data |
---|---|---|
7c657876 ACM |
1 | /* |
2 | * net/dccp/output.c | |
8109b02b | 3 | * |
7c657876 ACM |
4 | * An implementation of the DCCP protocol |
5 | * Arnaldo Carvalho de Melo <acme@conectiva.com.br> | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or | |
8 | * modify it under the terms of the GNU General Public License | |
9 | * as published by the Free Software Foundation; either version | |
10 | * 2 of the License, or (at your option) any later version. | |
11 | */ | |
12 | ||
7c657876 | 13 | #include <linux/dccp.h> |
48918a4d | 14 | #include <linux/kernel.h> |
7c657876 | 15 | #include <linux/skbuff.h> |
5a0e3ad6 | 16 | #include <linux/slab.h> |
7c657876 | 17 | |
14c85021 | 18 | #include <net/inet_sock.h> |
7c657876 ACM |
19 | #include <net/sock.h> |
20 | ||
ae31c339 | 21 | #include "ackvec.h" |
7c657876 ACM |
22 | #include "ccid.h" |
23 | #include "dccp.h" | |
24 | ||
25 | static inline void dccp_event_ack_sent(struct sock *sk) | |
26 | { | |
27 | inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK); | |
28 | } | |
29 | ||
8695e801 GR |
30 | /* enqueue @skb on sk_send_head for retransmission, return clone to send now */ |
31 | static struct sk_buff *dccp_skb_entail(struct sock *sk, struct sk_buff *skb) | |
48918a4d HX |
32 | { |
33 | skb_set_owner_w(skb, sk); | |
34 | WARN_ON(sk->sk_send_head); | |
35 | sk->sk_send_head = skb; | |
8695e801 | 36 | return skb_clone(sk->sk_send_head, gfp_any()); |
48918a4d HX |
37 | } |
38 | ||
7c657876 ACM |
39 | /* |
40 | * All SKB's seen here are completely headerless. It is our | |
41 | * job to build the DCCP header, and pass the packet down to | |
42 | * IP so it can do the same plus pass the packet off to the | |
43 | * device. | |
44 | */ | |
48918a4d | 45 | static int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb) |
7c657876 ACM |
46 | { |
47 | if (likely(skb != NULL)) { | |
d9d8da80 | 48 | struct inet_sock *inet = inet_sk(sk); |
57cca05a | 49 | const struct inet_connection_sock *icsk = inet_csk(sk); |
7c657876 ACM |
50 | struct dccp_sock *dp = dccp_sk(sk); |
51 | struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb); | |
52 | struct dccp_hdr *dh; | |
53 | /* XXX For now we're using only 48 bits sequence numbers */ | |
118b2c95 | 54 | const u32 dccp_header_size = sizeof(*dh) + |
7c657876 | 55 | sizeof(struct dccp_hdr_ext) + |
7690af3f | 56 | dccp_packet_hdr_len(dcb->dccpd_type); |
7c657876 ACM |
57 | int err, set_ack = 1; |
58 | u64 ackno = dp->dccps_gsr; | |
73f18fdb GR |
59 | /* |
60 | * Increment GSS here already in case the option code needs it. | |
61 | * Update GSS for real only if option processing below succeeds. | |
62 | */ | |
63 | dcb->dccpd_seq = ADD48(dp->dccps_gss, 1); | |
7c657876 | 64 | |
7c657876 ACM |
65 | switch (dcb->dccpd_type) { |
66 | case DCCP_PKT_DATA: | |
67 | set_ack = 0; | |
edc9e819 HX |
68 | /* fall through */ |
69 | case DCCP_PKT_DATAACK: | |
ee1a1592 | 70 | case DCCP_PKT_RESET: |
7c657876 | 71 | break; |
edc9e819 | 72 | |
afe00251 AB |
73 | case DCCP_PKT_REQUEST: |
74 | set_ack = 0; | |
73f18fdb GR |
75 | /* Use ISS on the first (non-retransmitted) Request. */ |
76 | if (icsk->icsk_retransmits == 0) | |
77 | dcb->dccpd_seq = dp->dccps_iss; | |
afe00251 AB |
78 | /* fall through */ |
79 | ||
7c657876 ACM |
80 | case DCCP_PKT_SYNC: |
81 | case DCCP_PKT_SYNCACK: | |
b0d045ca | 82 | ackno = dcb->dccpd_ack_seq; |
edc9e819 HX |
83 | /* fall through */ |
84 | default: | |
85 | /* | |
ee1a1592 GR |
86 | * Set owner/destructor: some skbs are allocated via |
87 | * alloc_skb (e.g. when retransmission may happen). | |
88 | * Only Data, DataAck, and Reset packets should come | |
89 | * through here with skb->sk set. | |
edc9e819 HX |
90 | */ |
91 | WARN_ON(skb->sk); | |
92 | skb_set_owner_w(skb, sk); | |
7c657876 ACM |
93 | break; |
94 | } | |
24117727 | 95 | |
2d0817d1 ACM |
96 | if (dccp_insert_options(sk, skb)) { |
97 | kfree_skb(skb); | |
98 | return -EPROTO; | |
99 | } | |
c9eaf173 | 100 | |
fda0fd6c | 101 | |
7c657876 | 102 | /* Build DCCP header and checksum it. */ |
9b42078e | 103 | dh = dccp_zeroed_hdr(skb, dccp_header_size); |
7c657876 | 104 | dh->dccph_type = dcb->dccpd_type; |
c720c7e8 ED |
105 | dh->dccph_sport = inet->inet_sport; |
106 | dh->dccph_dport = inet->inet_dport; | |
7c657876 ACM |
107 | dh->dccph_doff = (dccp_header_size + dcb->dccpd_opt_len) / 4; |
108 | dh->dccph_ccval = dcb->dccpd_ccval; | |
6f4e5fff | 109 | dh->dccph_cscov = dp->dccps_pcslen; |
7c657876 ACM |
110 | /* XXX For now we're using only 48 bits sequence numbers */ |
111 | dh->dccph_x = 1; | |
112 | ||
73f18fdb | 113 | dccp_update_gss(sk, dcb->dccpd_seq); |
7c657876 ACM |
114 | dccp_hdr_set_seq(dh, dp->dccps_gss); |
115 | if (set_ack) | |
116 | dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), ackno); | |
117 | ||
118 | switch (dcb->dccpd_type) { | |
119 | case DCCP_PKT_REQUEST: | |
7690af3f | 120 | dccp_hdr_request(skb)->dccph_req_service = |
67e6b629 | 121 | dp->dccps_service; |
73f18fdb GR |
122 | /* |
123 | * Limit Ack window to ISS <= P.ackno <= GSS, so that | |
124 | * only Responses to Requests we sent are considered. | |
125 | */ | |
126 | dp->dccps_awl = dp->dccps_iss; | |
7c657876 ACM |
127 | break; |
128 | case DCCP_PKT_RESET: | |
7690af3f ACM |
129 | dccp_hdr_reset(skb)->dccph_reset_code = |
130 | dcb->dccpd_reset_code; | |
7c657876 ACM |
131 | break; |
132 | } | |
133 | ||
bb296246 | 134 | icsk->icsk_af_ops->send_check(sk, skb); |
7c657876 | 135 | |
7ad07e7c | 136 | if (set_ack) |
7c657876 ACM |
137 | dccp_event_ack_sent(sk); |
138 | ||
139 | DCCP_INC_STATS(DCCP_MIB_OUTSEGS); | |
140 | ||
d9d8da80 | 141 | err = icsk->icsk_af_ops->queue_xmit(skb, &inet->cork.fl); |
b9df3cb8 | 142 | return net_xmit_eval(err); |
7c657876 ACM |
143 | } |
144 | return -ENOBUFS; | |
145 | } | |
146 | ||
6179983a | 147 | /** |
25985edc | 148 | * dccp_determine_ccmps - Find out about CCID-specific packet-size limits |
6179983a GR |
149 | * We only consider the HC-sender CCID for setting the CCMPS (RFC 4340, 14.), |
150 | * since the RX CCID is restricted to feedback packets (Acks), which are small | |
151 | * in comparison with the data traffic. A value of 0 means "no current CCMPS". | |
152 | */ | |
153 | static u32 dccp_determine_ccmps(const struct dccp_sock *dp) | |
154 | { | |
155 | const struct ccid *tx_ccid = dp->dccps_hc_tx_ccid; | |
156 | ||
157 | if (tx_ccid == NULL || tx_ccid->ccid_ops == NULL) | |
158 | return 0; | |
159 | return tx_ccid->ccid_ops->ccid_ccmps; | |
160 | } | |
161 | ||
7c657876 ACM |
162 | unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu) |
163 | { | |
d83d8461 | 164 | struct inet_connection_sock *icsk = inet_csk(sk); |
7c657876 | 165 | struct dccp_sock *dp = dccp_sk(sk); |
6179983a | 166 | u32 ccmps = dccp_determine_ccmps(dp); |
361a5c1d | 167 | u32 cur_mps = ccmps ? min(pmtu, ccmps) : pmtu; |
7c657876 | 168 | |
6179983a GR |
169 | /* Account for header lengths and IPv4/v6 option overhead */ |
170 | cur_mps -= (icsk->icsk_af_ops->net_header_len + icsk->icsk_ext_hdr_len + | |
171 | sizeof(struct dccp_hdr) + sizeof(struct dccp_hdr_ext)); | |
7c657876 ACM |
172 | |
173 | /* | |
361a5c1d GR |
174 | * Leave enough headroom for common DCCP header options. |
175 | * This only considers options which may appear on DCCP-Data packets, as | |
176 | * per table 3 in RFC 4340, 5.8. When running out of space for other | |
177 | * options (eg. Ack Vector which can take up to 255 bytes), it is better | |
178 | * to schedule a separate Ack. Thus we leave headroom for the following: | |
179 | * - 1 byte for Slow Receiver (11.6) | |
180 | * - 6 bytes for Timestamp (13.1) | |
181 | * - 10 bytes for Timestamp Echo (13.3) | |
182 | * - 8 bytes for NDP count (7.7, when activated) | |
183 | * - 6 bytes for Data Checksum (9.3) | |
184 | * - %DCCPAV_MIN_OPTLEN bytes for Ack Vector size (11.4, when enabled) | |
7c657876 | 185 | */ |
361a5c1d GR |
186 | cur_mps -= roundup(1 + 6 + 10 + dp->dccps_send_ndp_count * 8 + 6 + |
187 | (dp->dccps_hc_rx_ackvec ? DCCPAV_MIN_OPTLEN : 0), 4); | |
7c657876 ACM |
188 | |
189 | /* And store cached results */ | |
d83d8461 | 190 | icsk->icsk_pmtu_cookie = pmtu; |
6179983a | 191 | dp->dccps_mss_cache = cur_mps; |
7c657876 | 192 | |
6179983a | 193 | return cur_mps; |
7c657876 ACM |
194 | } |
195 | ||
f21e68ca ACM |
196 | EXPORT_SYMBOL_GPL(dccp_sync_mss); |
197 | ||
c530cfb1 ACM |
198 | void dccp_write_space(struct sock *sk) |
199 | { | |
43815482 | 200 | struct socket_wq *wq; |
c530cfb1 | 201 | |
43815482 ED |
202 | rcu_read_lock(); |
203 | wq = rcu_dereference(sk->sk_wq); | |
204 | if (wq_has_sleeper(wq)) | |
205 | wake_up_interruptible(&wq->wait); | |
c530cfb1 ACM |
206 | /* Should agree with poll, otherwise some programs break */ |
207 | if (sock_writeable(sk)) | |
8d8ad9d7 | 208 | sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT); |
c530cfb1 | 209 | |
43815482 | 210 | rcu_read_unlock(); |
c530cfb1 ACM |
211 | } |
212 | ||
d6809c12 | 213 | /** |
b1fcf55e | 214 | * dccp_wait_for_ccid - Await CCID send permission |
bc849872 | 215 | * @sk: socket to wait for |
b1fcf55e GR |
216 | * @delay: timeout in jiffies |
217 | * This is used by CCIDs which need to delay the send time in process context. | |
d6809c12 | 218 | */ |
b1fcf55e | 219 | static int dccp_wait_for_ccid(struct sock *sk, unsigned long delay) |
d6809c12 | 220 | { |
d6809c12 | 221 | DEFINE_WAIT(wait); |
b1fcf55e | 222 | long remaining; |
d6809c12 | 223 | |
b1fcf55e GR |
224 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
225 | sk->sk_write_pending++; | |
226 | release_sock(sk); | |
bc849872 | 227 | |
b1fcf55e | 228 | remaining = schedule_timeout(delay); |
d6809c12 | 229 | |
b1fcf55e GR |
230 | lock_sock(sk); |
231 | sk->sk_write_pending--; | |
aa395145 | 232 | finish_wait(sk_sleep(sk), &wait); |
b1fcf55e GR |
233 | |
234 | if (signal_pending(current) || sk->sk_err) | |
235 | return -1; | |
236 | return remaining; | |
d6809c12 ACM |
237 | } |
238 | ||
dc841e30 GR |
239 | /** |
240 | * dccp_xmit_packet - Send data packet under control of CCID | |
241 | * Transmits next-queued payload and informs CCID to account for the packet. | |
242 | */ | |
243 | static void dccp_xmit_packet(struct sock *sk) | |
244 | { | |
245 | int err, len; | |
246 | struct dccp_sock *dp = dccp_sk(sk); | |
871a2c16 | 247 | struct sk_buff *skb = dccp_qpolicy_pop(sk); |
dc841e30 GR |
248 | |
249 | if (unlikely(skb == NULL)) | |
250 | return; | |
251 | len = skb->len; | |
252 | ||
253 | if (sk->sk_state == DCCP_PARTOPEN) { | |
254 | const u32 cur_mps = dp->dccps_mss_cache - DCCP_FEATNEG_OVERHEAD; | |
255 | /* | |
256 | * See 8.1.5 - Handshake Completion. | |
257 | * | |
258 | * For robustness we resend Confirm options until the client has | |
259 | * entered OPEN. During the initial feature negotiation, the MPS | |
260 | * is smaller than usual, reduced by the Change/Confirm options. | |
261 | */ | |
262 | if (!list_empty(&dp->dccps_featneg) && len > cur_mps) { | |
263 | DCCP_WARN("Payload too large (%d) for featneg.\n", len); | |
264 | dccp_send_ack(sk); | |
265 | dccp_feat_list_purge(&dp->dccps_featneg); | |
266 | } | |
267 | ||
268 | inet_csk_schedule_ack(sk); | |
269 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, | |
270 | inet_csk(sk)->icsk_rto, | |
271 | DCCP_RTO_MAX); | |
272 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_DATAACK; | |
273 | } else if (dccp_ack_pending(sk)) { | |
274 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_DATAACK; | |
275 | } else { | |
276 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_DATA; | |
277 | } | |
278 | ||
279 | err = dccp_transmit_skb(sk, skb); | |
280 | if (err) | |
281 | dccp_pr_debug("transmit_skb() returned err=%d\n", err); | |
282 | /* | |
283 | * Register this one as sent even if an error occurred. To the remote | |
284 | * end a local packet drop is indistinguishable from network loss, i.e. | |
285 | * any local drop will eventually be reported via receiver feedback. | |
286 | */ | |
287 | ccid_hc_tx_packet_sent(dp->dccps_hc_tx_ccid, sk, len); | |
d83447f0 GR |
288 | |
289 | /* | |
290 | * If the CCID needs to transfer additional header options out-of-band | |
291 | * (e.g. Ack Vectors or feature-negotiation options), it activates this | |
292 | * flag to schedule a Sync. The Sync will automatically incorporate all | |
293 | * currently pending header options, thus clearing the backlog. | |
294 | */ | |
295 | if (dp->dccps_sync_scheduled) | |
296 | dccp_send_sync(sk, dp->dccps_gsr, DCCP_PKT_SYNC); | |
dc841e30 GR |
297 | } |
298 | ||
b1fcf55e GR |
299 | /** |
300 | * dccp_flush_write_queue - Drain queue at end of connection | |
301 | * Since dccp_sendmsg queues packets without waiting for them to be sent, it may | |
302 | * happen that the TX queue is not empty at the end of a connection. We give the | |
303 | * HC-sender CCID a grace period of up to @time_budget jiffies. If this function | |
304 | * returns with a non-empty write queue, it will be purged later. | |
305 | */ | |
306 | void dccp_flush_write_queue(struct sock *sk, long *time_budget) | |
307 | { | |
308 | struct dccp_sock *dp = dccp_sk(sk); | |
309 | struct sk_buff *skb; | |
310 | long delay, rc; | |
311 | ||
312 | while (*time_budget > 0 && (skb = skb_peek(&sk->sk_write_queue))) { | |
313 | rc = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb); | |
314 | ||
315 | switch (ccid_packet_dequeue_eval(rc)) { | |
316 | case CCID_PACKET_WILL_DEQUEUE_LATER: | |
317 | /* | |
318 | * If the CCID determines when to send, the next sending | |
319 | * time is unknown or the CCID may not even send again | |
320 | * (e.g. remote host crashes or lost Ack packets). | |
321 | */ | |
322 | DCCP_WARN("CCID did not manage to send all packets\n"); | |
323 | return; | |
324 | case CCID_PACKET_DELAY: | |
325 | delay = msecs_to_jiffies(rc); | |
326 | if (delay > *time_budget) | |
327 | return; | |
328 | rc = dccp_wait_for_ccid(sk, delay); | |
329 | if (rc < 0) | |
330 | return; | |
331 | *time_budget -= (delay - rc); | |
332 | /* check again if we can send now */ | |
333 | break; | |
334 | case CCID_PACKET_SEND_AT_ONCE: | |
335 | dccp_xmit_packet(sk); | |
336 | break; | |
337 | case CCID_PACKET_ERR: | |
338 | skb_dequeue(&sk->sk_write_queue); | |
339 | kfree_skb(skb); | |
340 | dccp_pr_debug("packet discarded due to err=%ld\n", rc); | |
341 | } | |
342 | } | |
343 | } | |
344 | ||
345 | void dccp_write_xmit(struct sock *sk) | |
27258ee5 | 346 | { |
97e5848d IM |
347 | struct dccp_sock *dp = dccp_sk(sk); |
348 | struct sk_buff *skb; | |
97e5848d | 349 | |
871a2c16 | 350 | while ((skb = dccp_qpolicy_top(sk))) { |
dc841e30 | 351 | int rc = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb); |
d6809c12 | 352 | |
dc841e30 GR |
353 | switch (ccid_packet_dequeue_eval(rc)) { |
354 | case CCID_PACKET_WILL_DEQUEUE_LATER: | |
355 | return; | |
356 | case CCID_PACKET_DELAY: | |
b1fcf55e GR |
357 | sk_reset_timer(sk, &dp->dccps_xmit_timer, |
358 | jiffies + msecs_to_jiffies(rc)); | |
359 | return; | |
dc841e30 GR |
360 | case CCID_PACKET_SEND_AT_ONCE: |
361 | dccp_xmit_packet(sk); | |
362 | break; | |
363 | case CCID_PACKET_ERR: | |
871a2c16 | 364 | dccp_qpolicy_drop(sk, skb); |
dc841e30 | 365 | dccp_pr_debug("packet discarded due to err=%d\n", rc); |
f6282f4d | 366 | } |
97e5848d | 367 | } |
27258ee5 ACM |
368 | } |
369 | ||
59435444 GR |
370 | /** |
371 | * dccp_retransmit_skb - Retransmit Request, Close, or CloseReq packets | |
372 | * There are only four retransmittable packet types in DCCP: | |
373 | * - Request in client-REQUEST state (sec. 8.1.1), | |
374 | * - CloseReq in server-CLOSEREQ state (sec. 8.3), | |
375 | * - Close in node-CLOSING state (sec. 8.3), | |
376 | * - Acks in client-PARTOPEN state (sec. 8.1.5, handled by dccp_delack_timer()). | |
377 | * This function expects sk->sk_send_head to contain the original skb. | |
378 | */ | |
379 | int dccp_retransmit_skb(struct sock *sk) | |
7c657876 | 380 | { |
59435444 GR |
381 | WARN_ON(sk->sk_send_head == NULL); |
382 | ||
57cca05a | 383 | if (inet_csk(sk)->icsk_af_ops->rebuild_header(sk) != 0) |
7c657876 ACM |
384 | return -EHOSTUNREACH; /* Routing failure or similar. */ |
385 | ||
59435444 GR |
386 | /* this count is used to distinguish original and retransmitted skb */ |
387 | inet_csk(sk)->icsk_retransmits++; | |
388 | ||
389 | return dccp_transmit_skb(sk, skb_clone(sk->sk_send_head, GFP_ATOMIC)); | |
7c657876 ACM |
390 | } |
391 | ||
392 | struct sk_buff *dccp_make_response(struct sock *sk, struct dst_entry *dst, | |
393 | struct request_sock *req) | |
394 | { | |
395 | struct dccp_hdr *dh; | |
67e6b629 | 396 | struct dccp_request_sock *dreq; |
118b2c95 | 397 | const u32 dccp_header_size = sizeof(struct dccp_hdr) + |
7c657876 ACM |
398 | sizeof(struct dccp_hdr_ext) + |
399 | sizeof(struct dccp_hdr_response); | |
118b2c95 | 400 | struct sk_buff *skb = sock_wmalloc(sk, sk->sk_prot->max_header, 1, |
7c657876 ACM |
401 | GFP_ATOMIC); |
402 | if (skb == NULL) | |
403 | return NULL; | |
404 | ||
405 | /* Reserve space for headers. */ | |
118b2c95 | 406 | skb_reserve(skb, sk->sk_prot->max_header); |
7c657876 | 407 | |
adf30907 | 408 | skb_dst_set(skb, dst_clone(dst)); |
7c657876 | 409 | |
67e6b629 | 410 | dreq = dccp_rsk(req); |
e11d9d30 GR |
411 | if (inet_rsk(req)->acked) /* increase ISS upon retransmission */ |
412 | dccp_inc_seqno(&dreq->dreq_iss); | |
7c657876 | 413 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_RESPONSE; |
67e6b629 | 414 | DCCP_SKB_CB(skb)->dccpd_seq = dreq->dreq_iss; |
2d0817d1 | 415 | |
0c116839 GR |
416 | /* Resolve feature dependencies resulting from choice of CCID */ |
417 | if (dccp_feat_server_ccid_dependencies(dreq)) | |
418 | goto response_failed; | |
419 | ||
420 | if (dccp_insert_options_rsk(dreq, skb)) | |
421 | goto response_failed; | |
7c657876 | 422 | |
09dbc389 | 423 | /* Build and checksum header */ |
9b42078e | 424 | dh = dccp_zeroed_hdr(skb, dccp_header_size); |
7c657876 | 425 | |
944f7502 | 426 | dh->dccph_sport = inet_rsk(req)->loc_port; |
7c657876 | 427 | dh->dccph_dport = inet_rsk(req)->rmt_port; |
7690af3f ACM |
428 | dh->dccph_doff = (dccp_header_size + |
429 | DCCP_SKB_CB(skb)->dccpd_opt_len) / 4; | |
7c657876 ACM |
430 | dh->dccph_type = DCCP_PKT_RESPONSE; |
431 | dh->dccph_x = 1; | |
67e6b629 ACM |
432 | dccp_hdr_set_seq(dh, dreq->dreq_iss); |
433 | dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dreq->dreq_isr); | |
434 | dccp_hdr_response(skb)->dccph_resp_service = dreq->dreq_service; | |
7c657876 | 435 | |
6f4e5fff GR |
436 | dccp_csum_outgoing(skb); |
437 | ||
e11d9d30 GR |
438 | /* We use `acked' to remember that a Response was already sent. */ |
439 | inet_rsk(req)->acked = 1; | |
7c657876 ACM |
440 | DCCP_INC_STATS(DCCP_MIB_OUTSEGS); |
441 | return skb; | |
0c116839 GR |
442 | response_failed: |
443 | kfree_skb(skb); | |
444 | return NULL; | |
7c657876 ACM |
445 | } |
446 | ||
f21e68ca ACM |
447 | EXPORT_SYMBOL_GPL(dccp_make_response); |
448 | ||
e356d37a | 449 | /* answer offending packet in @rcv_skb with Reset from control socket @ctl */ |
7630f026 | 450 | struct sk_buff *dccp_ctl_make_reset(struct sock *sk, struct sk_buff *rcv_skb) |
e356d37a GR |
451 | { |
452 | struct dccp_hdr *rxdh = dccp_hdr(rcv_skb), *dh; | |
453 | struct dccp_skb_cb *dcb = DCCP_SKB_CB(rcv_skb); | |
454 | const u32 dccp_hdr_reset_len = sizeof(struct dccp_hdr) + | |
455 | sizeof(struct dccp_hdr_ext) + | |
456 | sizeof(struct dccp_hdr_reset); | |
457 | struct dccp_hdr_reset *dhr; | |
458 | struct sk_buff *skb; | |
459 | ||
7630f026 | 460 | skb = alloc_skb(sk->sk_prot->max_header, GFP_ATOMIC); |
e356d37a GR |
461 | if (skb == NULL) |
462 | return NULL; | |
463 | ||
7630f026 | 464 | skb_reserve(skb, sk->sk_prot->max_header); |
e356d37a GR |
465 | |
466 | /* Swap the send and the receive. */ | |
467 | dh = dccp_zeroed_hdr(skb, dccp_hdr_reset_len); | |
468 | dh->dccph_type = DCCP_PKT_RESET; | |
469 | dh->dccph_sport = rxdh->dccph_dport; | |
470 | dh->dccph_dport = rxdh->dccph_sport; | |
471 | dh->dccph_doff = dccp_hdr_reset_len / 4; | |
472 | dh->dccph_x = 1; | |
473 | ||
474 | dhr = dccp_hdr_reset(skb); | |
475 | dhr->dccph_reset_code = dcb->dccpd_reset_code; | |
476 | ||
477 | switch (dcb->dccpd_reset_code) { | |
478 | case DCCP_RESET_CODE_PACKET_ERROR: | |
479 | dhr->dccph_reset_data[0] = rxdh->dccph_type; | |
480 | break; | |
481 | case DCCP_RESET_CODE_OPTION_ERROR: /* fall through */ | |
482 | case DCCP_RESET_CODE_MANDATORY_ERROR: | |
483 | memcpy(dhr->dccph_reset_data, dcb->dccpd_reset_data, 3); | |
484 | break; | |
485 | } | |
486 | /* | |
487 | * From RFC 4340, 8.3.1: | |
488 | * If P.ackno exists, set R.seqno := P.ackno + 1. | |
489 | * Else set R.seqno := 0. | |
490 | */ | |
491 | if (dcb->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ) | |
492 | dccp_hdr_set_seq(dh, ADD48(dcb->dccpd_ack_seq, 1)); | |
493 | dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dcb->dccpd_seq); | |
494 | ||
495 | dccp_csum_outgoing(skb); | |
496 | return skb; | |
497 | } | |
498 | ||
499 | EXPORT_SYMBOL_GPL(dccp_ctl_make_reset); | |
500 | ||
ee1a1592 | 501 | /* send Reset on established socket, to close or abort the connection */ |
017487d7 ACM |
502 | int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code) |
503 | { | |
ee1a1592 | 504 | struct sk_buff *skb; |
017487d7 ACM |
505 | /* |
506 | * FIXME: what if rebuild_header fails? | |
507 | * Should we be doing a rebuild_header here? | |
508 | */ | |
f53dc67c | 509 | int err = inet_csk(sk)->icsk_af_ops->rebuild_header(sk); |
017487d7 | 510 | |
ee1a1592 GR |
511 | if (err != 0) |
512 | return err; | |
513 | ||
514 | skb = sock_wmalloc(sk, sk->sk_prot->max_header, 1, GFP_ATOMIC); | |
515 | if (skb == NULL) | |
516 | return -ENOBUFS; | |
517 | ||
518 | /* Reserve space for headers and prepare control bits. */ | |
519 | skb_reserve(skb, sk->sk_prot->max_header); | |
520 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_RESET; | |
521 | DCCP_SKB_CB(skb)->dccpd_reset_code = code; | |
017487d7 | 522 | |
ee1a1592 | 523 | return dccp_transmit_skb(sk, skb); |
017487d7 ACM |
524 | } |
525 | ||
7c657876 ACM |
526 | /* |
527 | * Do all connect socket setups that can be done AF independent. | |
528 | */ | |
93344af4 | 529 | int dccp_connect(struct sock *sk) |
7c657876 | 530 | { |
93344af4 | 531 | struct sk_buff *skb; |
f21e68ca | 532 | struct dccp_sock *dp = dccp_sk(sk); |
7c657876 ACM |
533 | struct dst_entry *dst = __sk_dst_get(sk); |
534 | struct inet_connection_sock *icsk = inet_csk(sk); | |
535 | ||
536 | sk->sk_err = 0; | |
537 | sock_reset_flag(sk, SOCK_DONE); | |
c9eaf173 | 538 | |
7c657876 ACM |
539 | dccp_sync_mss(sk, dst_mtu(dst)); |
540 | ||
9eca0a47 GR |
541 | /* do not connect if feature negotiation setup fails */ |
542 | if (dccp_feat_finalise_settings(dccp_sk(sk))) | |
543 | return -EPROTO; | |
544 | ||
93344af4 GR |
545 | /* Initialise GAR as per 8.5; AWL/AWH are set in dccp_transmit_skb() */ |
546 | dp->dccps_gar = dp->dccps_iss; | |
7c657876 | 547 | |
118b2c95 | 548 | skb = alloc_skb(sk->sk_prot->max_header, sk->sk_allocation); |
7c657876 ACM |
549 | if (unlikely(skb == NULL)) |
550 | return -ENOBUFS; | |
551 | ||
552 | /* Reserve space for headers. */ | |
118b2c95 | 553 | skb_reserve(skb, sk->sk_prot->max_header); |
7c657876 ACM |
554 | |
555 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_REQUEST; | |
7c657876 | 556 | |
8695e801 | 557 | dccp_transmit_skb(sk, dccp_skb_entail(sk, skb)); |
7c657876 ACM |
558 | DCCP_INC_STATS(DCCP_MIB_ACTIVEOPENS); |
559 | ||
560 | /* Timer for repeating the REQUEST until an answer. */ | |
93344af4 | 561 | icsk->icsk_retransmits = 0; |
27258ee5 ACM |
562 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, |
563 | icsk->icsk_rto, DCCP_RTO_MAX); | |
7c657876 ACM |
564 | return 0; |
565 | } | |
566 | ||
f21e68ca ACM |
567 | EXPORT_SYMBOL_GPL(dccp_connect); |
568 | ||
7c657876 ACM |
569 | void dccp_send_ack(struct sock *sk) |
570 | { | |
571 | /* If we have been reset, we may not send again. */ | |
572 | if (sk->sk_state != DCCP_CLOSED) { | |
118b2c95 ACM |
573 | struct sk_buff *skb = alloc_skb(sk->sk_prot->max_header, |
574 | GFP_ATOMIC); | |
7c657876 ACM |
575 | |
576 | if (skb == NULL) { | |
577 | inet_csk_schedule_ack(sk); | |
578 | inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN; | |
7690af3f ACM |
579 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, |
580 | TCP_DELACK_MAX, | |
581 | DCCP_RTO_MAX); | |
7c657876 ACM |
582 | return; |
583 | } | |
584 | ||
585 | /* Reserve space for headers */ | |
118b2c95 | 586 | skb_reserve(skb, sk->sk_prot->max_header); |
7c657876 | 587 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_ACK; |
7c657876 ACM |
588 | dccp_transmit_skb(sk, skb); |
589 | } | |
590 | } | |
591 | ||
592 | EXPORT_SYMBOL_GPL(dccp_send_ack); | |
593 | ||
1e2f0e5e | 594 | #if 0 |
727ecc5f | 595 | /* FIXME: Is this still necessary (11.3) - currently nowhere used by DCCP. */ |
7c657876 ACM |
596 | void dccp_send_delayed_ack(struct sock *sk) |
597 | { | |
598 | struct inet_connection_sock *icsk = inet_csk(sk); | |
599 | /* | |
600 | * FIXME: tune this timer. elapsed time fixes the skew, so no problem | |
601 | * with using 2s, and active senders also piggyback the ACK into a | |
602 | * DATAACK packet, so this is really for quiescent senders. | |
603 | */ | |
604 | unsigned long timeout = jiffies + 2 * HZ; | |
605 | ||
606 | /* Use new timeout only if there wasn't a older one earlier. */ | |
607 | if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) { | |
608 | /* If delack timer was blocked or is about to expire, | |
609 | * send ACK now. | |
610 | * | |
611 | * FIXME: check the "about to expire" part | |
612 | */ | |
613 | if (icsk->icsk_ack.blocked) { | |
614 | dccp_send_ack(sk); | |
615 | return; | |
616 | } | |
617 | ||
618 | if (!time_before(timeout, icsk->icsk_ack.timeout)) | |
619 | timeout = icsk->icsk_ack.timeout; | |
620 | } | |
621 | icsk->icsk_ack.pending |= ICSK_ACK_SCHED | ICSK_ACK_TIMER; | |
622 | icsk->icsk_ack.timeout = timeout; | |
623 | sk_reset_timer(sk, &icsk->icsk_delack_timer, timeout); | |
624 | } | |
1e2f0e5e | 625 | #endif |
7c657876 | 626 | |
b0d045ca | 627 | void dccp_send_sync(struct sock *sk, const u64 ackno, |
e92ae93a | 628 | const enum dccp_pkt_type pkt_type) |
7c657876 ACM |
629 | { |
630 | /* | |
631 | * We are not putting this on the write queue, so | |
632 | * dccp_transmit_skb() will set the ownership to this | |
633 | * sock. | |
634 | */ | |
118b2c95 | 635 | struct sk_buff *skb = alloc_skb(sk->sk_prot->max_header, GFP_ATOMIC); |
7c657876 | 636 | |
b0d045ca | 637 | if (skb == NULL) { |
7c657876 | 638 | /* FIXME: how to make sure the sync is sent? */ |
b0d045ca | 639 | DCCP_CRIT("could not send %s", dccp_packet_name(pkt_type)); |
7c657876 | 640 | return; |
b0d045ca | 641 | } |
7c657876 ACM |
642 | |
643 | /* Reserve space for headers and prepare control bits. */ | |
118b2c95 | 644 | skb_reserve(skb, sk->sk_prot->max_header); |
e92ae93a | 645 | DCCP_SKB_CB(skb)->dccpd_type = pkt_type; |
b0d045ca | 646 | DCCP_SKB_CB(skb)->dccpd_ack_seq = ackno; |
7c657876 | 647 | |
d83447f0 GR |
648 | /* |
649 | * Clear the flag in case the Sync was scheduled for out-of-band data, | |
650 | * such as carrying a long Ack Vector. | |
651 | */ | |
652 | dccp_sk(sk)->dccps_sync_scheduled = 0; | |
653 | ||
7c657876 ACM |
654 | dccp_transmit_skb(sk, skb); |
655 | } | |
656 | ||
b61fafc4 ACM |
657 | EXPORT_SYMBOL_GPL(dccp_send_sync); |
658 | ||
7690af3f ACM |
659 | /* |
660 | * Send a DCCP_PKT_CLOSE/CLOSEREQ. The caller locks the socket for us. This | |
661 | * cannot be allowed to fail queueing a DCCP_PKT_CLOSE/CLOSEREQ frame under | |
662 | * any circumstances. | |
7c657876 | 663 | */ |
7ad07e7c | 664 | void dccp_send_close(struct sock *sk, const int active) |
7c657876 ACM |
665 | { |
666 | struct dccp_sock *dp = dccp_sk(sk); | |
667 | struct sk_buff *skb; | |
7d877f3b | 668 | const gfp_t prio = active ? GFP_KERNEL : GFP_ATOMIC; |
7c657876 | 669 | |
7ad07e7c ACM |
670 | skb = alloc_skb(sk->sk_prot->max_header, prio); |
671 | if (skb == NULL) | |
672 | return; | |
7c657876 ACM |
673 | |
674 | /* Reserve space for headers and prepare control bits. */ | |
675 | skb_reserve(skb, sk->sk_prot->max_header); | |
b8599d20 GR |
676 | if (dp->dccps_role == DCCP_ROLE_SERVER && !dp->dccps_server_timewait) |
677 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_CLOSEREQ; | |
678 | else | |
679 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_CLOSE; | |
7c657876 | 680 | |
7ad07e7c | 681 | if (active) { |
8695e801 | 682 | skb = dccp_skb_entail(sk, skb); |
92d31920 GR |
683 | /* |
684 | * Retransmission timer for active-close: RFC 4340, 8.3 requires | |
685 | * to retransmit the Close/CloseReq until the CLOSING/CLOSEREQ | |
686 | * state can be left. The initial timeout is 2 RTTs. | |
687 | * Since RTT measurement is done by the CCIDs, there is no easy | |
688 | * way to get an RTT sample. The fallback RTT from RFC 4340, 3.4 | |
689 | * is too low (200ms); we use a high value to avoid unnecessary | |
690 | * retransmissions when the link RTT is > 0.2 seconds. | |
691 | * FIXME: Let main module sample RTTs and use that instead. | |
692 | */ | |
693 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, | |
694 | DCCP_TIMEOUT_INIT, DCCP_RTO_MAX); | |
8695e801 GR |
695 | } |
696 | dccp_transmit_skb(sk, skb); | |
7c657876 | 697 | } |