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 | ||
c25a18ba | 30 | static void dccp_skb_entail(struct sock *sk, struct sk_buff *skb) |
48918a4d HX |
31 | { |
32 | skb_set_owner_w(skb, sk); | |
33 | WARN_ON(sk->sk_send_head); | |
34 | sk->sk_send_head = skb; | |
35 | } | |
36 | ||
7c657876 ACM |
37 | /* |
38 | * All SKB's seen here are completely headerless. It is our | |
39 | * job to build the DCCP header, and pass the packet down to | |
40 | * IP so it can do the same plus pass the packet off to the | |
41 | * device. | |
42 | */ | |
48918a4d | 43 | static int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb) |
7c657876 ACM |
44 | { |
45 | if (likely(skb != NULL)) { | |
46 | const struct inet_sock *inet = inet_sk(sk); | |
57cca05a | 47 | const struct inet_connection_sock *icsk = inet_csk(sk); |
7c657876 ACM |
48 | struct dccp_sock *dp = dccp_sk(sk); |
49 | struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb); | |
50 | struct dccp_hdr *dh; | |
51 | /* XXX For now we're using only 48 bits sequence numbers */ | |
118b2c95 | 52 | const u32 dccp_header_size = sizeof(*dh) + |
7c657876 | 53 | sizeof(struct dccp_hdr_ext) + |
7690af3f | 54 | dccp_packet_hdr_len(dcb->dccpd_type); |
7c657876 ACM |
55 | int err, set_ack = 1; |
56 | u64 ackno = dp->dccps_gsr; | |
73f18fdb GR |
57 | /* |
58 | * Increment GSS here already in case the option code needs it. | |
59 | * Update GSS for real only if option processing below succeeds. | |
60 | */ | |
61 | dcb->dccpd_seq = ADD48(dp->dccps_gss, 1); | |
7c657876 | 62 | |
7c657876 ACM |
63 | switch (dcb->dccpd_type) { |
64 | case DCCP_PKT_DATA: | |
65 | set_ack = 0; | |
edc9e819 HX |
66 | /* fall through */ |
67 | case DCCP_PKT_DATAACK: | |
ee1a1592 | 68 | case DCCP_PKT_RESET: |
7c657876 | 69 | break; |
edc9e819 | 70 | |
afe00251 AB |
71 | case DCCP_PKT_REQUEST: |
72 | set_ack = 0; | |
73f18fdb GR |
73 | /* Use ISS on the first (non-retransmitted) Request. */ |
74 | if (icsk->icsk_retransmits == 0) | |
75 | dcb->dccpd_seq = dp->dccps_iss; | |
afe00251 AB |
76 | /* fall through */ |
77 | ||
7c657876 ACM |
78 | case DCCP_PKT_SYNC: |
79 | case DCCP_PKT_SYNCACK: | |
b0d045ca | 80 | ackno = dcb->dccpd_ack_seq; |
edc9e819 HX |
81 | /* fall through */ |
82 | default: | |
83 | /* | |
ee1a1592 GR |
84 | * Set owner/destructor: some skbs are allocated via |
85 | * alloc_skb (e.g. when retransmission may happen). | |
86 | * Only Data, DataAck, and Reset packets should come | |
87 | * through here with skb->sk set. | |
edc9e819 HX |
88 | */ |
89 | WARN_ON(skb->sk); | |
90 | skb_set_owner_w(skb, sk); | |
7c657876 ACM |
91 | break; |
92 | } | |
24117727 | 93 | |
2d0817d1 ACM |
94 | if (dccp_insert_options(sk, skb)) { |
95 | kfree_skb(skb); | |
96 | return -EPROTO; | |
97 | } | |
c9eaf173 | 98 | |
fda0fd6c | 99 | |
7c657876 | 100 | /* Build DCCP header and checksum it. */ |
9b42078e | 101 | dh = dccp_zeroed_hdr(skb, dccp_header_size); |
7c657876 | 102 | dh->dccph_type = dcb->dccpd_type; |
c720c7e8 ED |
103 | dh->dccph_sport = inet->inet_sport; |
104 | dh->dccph_dport = inet->inet_dport; | |
7c657876 ACM |
105 | dh->dccph_doff = (dccp_header_size + dcb->dccpd_opt_len) / 4; |
106 | dh->dccph_ccval = dcb->dccpd_ccval; | |
6f4e5fff | 107 | dh->dccph_cscov = dp->dccps_pcslen; |
7c657876 ACM |
108 | /* XXX For now we're using only 48 bits sequence numbers */ |
109 | dh->dccph_x = 1; | |
110 | ||
73f18fdb | 111 | dccp_update_gss(sk, dcb->dccpd_seq); |
7c657876 ACM |
112 | dccp_hdr_set_seq(dh, dp->dccps_gss); |
113 | if (set_ack) | |
114 | dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), ackno); | |
115 | ||
116 | switch (dcb->dccpd_type) { | |
117 | case DCCP_PKT_REQUEST: | |
7690af3f | 118 | dccp_hdr_request(skb)->dccph_req_service = |
67e6b629 | 119 | dp->dccps_service; |
73f18fdb GR |
120 | /* |
121 | * Limit Ack window to ISS <= P.ackno <= GSS, so that | |
122 | * only Responses to Requests we sent are considered. | |
123 | */ | |
124 | dp->dccps_awl = dp->dccps_iss; | |
7c657876 ACM |
125 | break; |
126 | case DCCP_PKT_RESET: | |
7690af3f ACM |
127 | dccp_hdr_reset(skb)->dccph_reset_code = |
128 | dcb->dccpd_reset_code; | |
7c657876 ACM |
129 | break; |
130 | } | |
131 | ||
bb296246 | 132 | icsk->icsk_af_ops->send_check(sk, skb); |
7c657876 | 133 | |
7ad07e7c | 134 | if (set_ack) |
7c657876 ACM |
135 | dccp_event_ack_sent(sk); |
136 | ||
137 | DCCP_INC_STATS(DCCP_MIB_OUTSEGS); | |
138 | ||
4e15ed4d | 139 | err = icsk->icsk_af_ops->queue_xmit(skb); |
b9df3cb8 | 140 | return net_xmit_eval(err); |
7c657876 ACM |
141 | } |
142 | return -ENOBUFS; | |
143 | } | |
144 | ||
6179983a | 145 | /** |
25985edc | 146 | * dccp_determine_ccmps - Find out about CCID-specific packet-size limits |
6179983a GR |
147 | * We only consider the HC-sender CCID for setting the CCMPS (RFC 4340, 14.), |
148 | * since the RX CCID is restricted to feedback packets (Acks), which are small | |
149 | * in comparison with the data traffic. A value of 0 means "no current CCMPS". | |
150 | */ | |
151 | static u32 dccp_determine_ccmps(const struct dccp_sock *dp) | |
152 | { | |
153 | const struct ccid *tx_ccid = dp->dccps_hc_tx_ccid; | |
154 | ||
155 | if (tx_ccid == NULL || tx_ccid->ccid_ops == NULL) | |
156 | return 0; | |
157 | return tx_ccid->ccid_ops->ccid_ccmps; | |
158 | } | |
159 | ||
7c657876 ACM |
160 | unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu) |
161 | { | |
d83d8461 | 162 | struct inet_connection_sock *icsk = inet_csk(sk); |
7c657876 | 163 | struct dccp_sock *dp = dccp_sk(sk); |
6179983a | 164 | u32 ccmps = dccp_determine_ccmps(dp); |
361a5c1d | 165 | u32 cur_mps = ccmps ? min(pmtu, ccmps) : pmtu; |
7c657876 | 166 | |
6179983a GR |
167 | /* Account for header lengths and IPv4/v6 option overhead */ |
168 | cur_mps -= (icsk->icsk_af_ops->net_header_len + icsk->icsk_ext_hdr_len + | |
169 | sizeof(struct dccp_hdr) + sizeof(struct dccp_hdr_ext)); | |
7c657876 ACM |
170 | |
171 | /* | |
361a5c1d GR |
172 | * Leave enough headroom for common DCCP header options. |
173 | * This only considers options which may appear on DCCP-Data packets, as | |
174 | * per table 3 in RFC 4340, 5.8. When running out of space for other | |
175 | * options (eg. Ack Vector which can take up to 255 bytes), it is better | |
176 | * to schedule a separate Ack. Thus we leave headroom for the following: | |
177 | * - 1 byte for Slow Receiver (11.6) | |
178 | * - 6 bytes for Timestamp (13.1) | |
179 | * - 10 bytes for Timestamp Echo (13.3) | |
180 | * - 8 bytes for NDP count (7.7, when activated) | |
181 | * - 6 bytes for Data Checksum (9.3) | |
182 | * - %DCCPAV_MIN_OPTLEN bytes for Ack Vector size (11.4, when enabled) | |
7c657876 | 183 | */ |
361a5c1d GR |
184 | cur_mps -= roundup(1 + 6 + 10 + dp->dccps_send_ndp_count * 8 + 6 + |
185 | (dp->dccps_hc_rx_ackvec ? DCCPAV_MIN_OPTLEN : 0), 4); | |
7c657876 ACM |
186 | |
187 | /* And store cached results */ | |
d83d8461 | 188 | icsk->icsk_pmtu_cookie = pmtu; |
6179983a | 189 | dp->dccps_mss_cache = cur_mps; |
7c657876 | 190 | |
6179983a | 191 | return cur_mps; |
7c657876 ACM |
192 | } |
193 | ||
f21e68ca ACM |
194 | EXPORT_SYMBOL_GPL(dccp_sync_mss); |
195 | ||
c530cfb1 ACM |
196 | void dccp_write_space(struct sock *sk) |
197 | { | |
43815482 | 198 | struct socket_wq *wq; |
c530cfb1 | 199 | |
43815482 ED |
200 | rcu_read_lock(); |
201 | wq = rcu_dereference(sk->sk_wq); | |
202 | if (wq_has_sleeper(wq)) | |
203 | wake_up_interruptible(&wq->wait); | |
c530cfb1 ACM |
204 | /* Should agree with poll, otherwise some programs break */ |
205 | if (sock_writeable(sk)) | |
8d8ad9d7 | 206 | sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT); |
c530cfb1 | 207 | |
43815482 | 208 | rcu_read_unlock(); |
c530cfb1 ACM |
209 | } |
210 | ||
d6809c12 | 211 | /** |
b1fcf55e | 212 | * dccp_wait_for_ccid - Await CCID send permission |
bc849872 | 213 | * @sk: socket to wait for |
b1fcf55e GR |
214 | * @delay: timeout in jiffies |
215 | * This is used by CCIDs which need to delay the send time in process context. | |
d6809c12 | 216 | */ |
b1fcf55e | 217 | static int dccp_wait_for_ccid(struct sock *sk, unsigned long delay) |
d6809c12 | 218 | { |
d6809c12 | 219 | DEFINE_WAIT(wait); |
b1fcf55e | 220 | long remaining; |
d6809c12 | 221 | |
b1fcf55e GR |
222 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
223 | sk->sk_write_pending++; | |
224 | release_sock(sk); | |
bc849872 | 225 | |
b1fcf55e | 226 | remaining = schedule_timeout(delay); |
d6809c12 | 227 | |
b1fcf55e GR |
228 | lock_sock(sk); |
229 | sk->sk_write_pending--; | |
aa395145 | 230 | finish_wait(sk_sleep(sk), &wait); |
b1fcf55e GR |
231 | |
232 | if (signal_pending(current) || sk->sk_err) | |
233 | return -1; | |
234 | return remaining; | |
d6809c12 ACM |
235 | } |
236 | ||
dc841e30 GR |
237 | /** |
238 | * dccp_xmit_packet - Send data packet under control of CCID | |
239 | * Transmits next-queued payload and informs CCID to account for the packet. | |
240 | */ | |
241 | static void dccp_xmit_packet(struct sock *sk) | |
242 | { | |
243 | int err, len; | |
244 | struct dccp_sock *dp = dccp_sk(sk); | |
871a2c16 | 245 | struct sk_buff *skb = dccp_qpolicy_pop(sk); |
dc841e30 GR |
246 | |
247 | if (unlikely(skb == NULL)) | |
248 | return; | |
249 | len = skb->len; | |
250 | ||
251 | if (sk->sk_state == DCCP_PARTOPEN) { | |
252 | const u32 cur_mps = dp->dccps_mss_cache - DCCP_FEATNEG_OVERHEAD; | |
253 | /* | |
254 | * See 8.1.5 - Handshake Completion. | |
255 | * | |
256 | * For robustness we resend Confirm options until the client has | |
257 | * entered OPEN. During the initial feature negotiation, the MPS | |
258 | * is smaller than usual, reduced by the Change/Confirm options. | |
259 | */ | |
260 | if (!list_empty(&dp->dccps_featneg) && len > cur_mps) { | |
261 | DCCP_WARN("Payload too large (%d) for featneg.\n", len); | |
262 | dccp_send_ack(sk); | |
263 | dccp_feat_list_purge(&dp->dccps_featneg); | |
264 | } | |
265 | ||
266 | inet_csk_schedule_ack(sk); | |
267 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, | |
268 | inet_csk(sk)->icsk_rto, | |
269 | DCCP_RTO_MAX); | |
270 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_DATAACK; | |
271 | } else if (dccp_ack_pending(sk)) { | |
272 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_DATAACK; | |
273 | } else { | |
274 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_DATA; | |
275 | } | |
276 | ||
277 | err = dccp_transmit_skb(sk, skb); | |
278 | if (err) | |
279 | dccp_pr_debug("transmit_skb() returned err=%d\n", err); | |
280 | /* | |
281 | * Register this one as sent even if an error occurred. To the remote | |
282 | * end a local packet drop is indistinguishable from network loss, i.e. | |
283 | * any local drop will eventually be reported via receiver feedback. | |
284 | */ | |
285 | ccid_hc_tx_packet_sent(dp->dccps_hc_tx_ccid, sk, len); | |
d83447f0 GR |
286 | |
287 | /* | |
288 | * If the CCID needs to transfer additional header options out-of-band | |
289 | * (e.g. Ack Vectors or feature-negotiation options), it activates this | |
290 | * flag to schedule a Sync. The Sync will automatically incorporate all | |
291 | * currently pending header options, thus clearing the backlog. | |
292 | */ | |
293 | if (dp->dccps_sync_scheduled) | |
294 | dccp_send_sync(sk, dp->dccps_gsr, DCCP_PKT_SYNC); | |
dc841e30 GR |
295 | } |
296 | ||
b1fcf55e GR |
297 | /** |
298 | * dccp_flush_write_queue - Drain queue at end of connection | |
299 | * Since dccp_sendmsg queues packets without waiting for them to be sent, it may | |
300 | * happen that the TX queue is not empty at the end of a connection. We give the | |
301 | * HC-sender CCID a grace period of up to @time_budget jiffies. If this function | |
302 | * returns with a non-empty write queue, it will be purged later. | |
303 | */ | |
304 | void dccp_flush_write_queue(struct sock *sk, long *time_budget) | |
305 | { | |
306 | struct dccp_sock *dp = dccp_sk(sk); | |
307 | struct sk_buff *skb; | |
308 | long delay, rc; | |
309 | ||
310 | while (*time_budget > 0 && (skb = skb_peek(&sk->sk_write_queue))) { | |
311 | rc = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb); | |
312 | ||
313 | switch (ccid_packet_dequeue_eval(rc)) { | |
314 | case CCID_PACKET_WILL_DEQUEUE_LATER: | |
315 | /* | |
316 | * If the CCID determines when to send, the next sending | |
317 | * time is unknown or the CCID may not even send again | |
318 | * (e.g. remote host crashes or lost Ack packets). | |
319 | */ | |
320 | DCCP_WARN("CCID did not manage to send all packets\n"); | |
321 | return; | |
322 | case CCID_PACKET_DELAY: | |
323 | delay = msecs_to_jiffies(rc); | |
324 | if (delay > *time_budget) | |
325 | return; | |
326 | rc = dccp_wait_for_ccid(sk, delay); | |
327 | if (rc < 0) | |
328 | return; | |
329 | *time_budget -= (delay - rc); | |
330 | /* check again if we can send now */ | |
331 | break; | |
332 | case CCID_PACKET_SEND_AT_ONCE: | |
333 | dccp_xmit_packet(sk); | |
334 | break; | |
335 | case CCID_PACKET_ERR: | |
336 | skb_dequeue(&sk->sk_write_queue); | |
337 | kfree_skb(skb); | |
338 | dccp_pr_debug("packet discarded due to err=%ld\n", rc); | |
339 | } | |
340 | } | |
341 | } | |
342 | ||
343 | void dccp_write_xmit(struct sock *sk) | |
27258ee5 | 344 | { |
97e5848d IM |
345 | struct dccp_sock *dp = dccp_sk(sk); |
346 | struct sk_buff *skb; | |
97e5848d | 347 | |
871a2c16 | 348 | while ((skb = dccp_qpolicy_top(sk))) { |
dc841e30 | 349 | int rc = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb); |
d6809c12 | 350 | |
dc841e30 GR |
351 | switch (ccid_packet_dequeue_eval(rc)) { |
352 | case CCID_PACKET_WILL_DEQUEUE_LATER: | |
353 | return; | |
354 | case CCID_PACKET_DELAY: | |
b1fcf55e GR |
355 | sk_reset_timer(sk, &dp->dccps_xmit_timer, |
356 | jiffies + msecs_to_jiffies(rc)); | |
357 | return; | |
dc841e30 GR |
358 | case CCID_PACKET_SEND_AT_ONCE: |
359 | dccp_xmit_packet(sk); | |
360 | break; | |
361 | case CCID_PACKET_ERR: | |
871a2c16 | 362 | dccp_qpolicy_drop(sk, skb); |
dc841e30 | 363 | dccp_pr_debug("packet discarded due to err=%d\n", rc); |
f6282f4d | 364 | } |
97e5848d | 365 | } |
27258ee5 ACM |
366 | } |
367 | ||
59435444 GR |
368 | /** |
369 | * dccp_retransmit_skb - Retransmit Request, Close, or CloseReq packets | |
370 | * There are only four retransmittable packet types in DCCP: | |
371 | * - Request in client-REQUEST state (sec. 8.1.1), | |
372 | * - CloseReq in server-CLOSEREQ state (sec. 8.3), | |
373 | * - Close in node-CLOSING state (sec. 8.3), | |
374 | * - Acks in client-PARTOPEN state (sec. 8.1.5, handled by dccp_delack_timer()). | |
375 | * This function expects sk->sk_send_head to contain the original skb. | |
376 | */ | |
377 | int dccp_retransmit_skb(struct sock *sk) | |
7c657876 | 378 | { |
59435444 GR |
379 | WARN_ON(sk->sk_send_head == NULL); |
380 | ||
57cca05a | 381 | if (inet_csk(sk)->icsk_af_ops->rebuild_header(sk) != 0) |
7c657876 ACM |
382 | return -EHOSTUNREACH; /* Routing failure or similar. */ |
383 | ||
59435444 GR |
384 | /* this count is used to distinguish original and retransmitted skb */ |
385 | inet_csk(sk)->icsk_retransmits++; | |
386 | ||
387 | return dccp_transmit_skb(sk, skb_clone(sk->sk_send_head, GFP_ATOMIC)); | |
7c657876 ACM |
388 | } |
389 | ||
390 | struct sk_buff *dccp_make_response(struct sock *sk, struct dst_entry *dst, | |
391 | struct request_sock *req) | |
392 | { | |
393 | struct dccp_hdr *dh; | |
67e6b629 | 394 | struct dccp_request_sock *dreq; |
118b2c95 | 395 | const u32 dccp_header_size = sizeof(struct dccp_hdr) + |
7c657876 ACM |
396 | sizeof(struct dccp_hdr_ext) + |
397 | sizeof(struct dccp_hdr_response); | |
118b2c95 | 398 | struct sk_buff *skb = sock_wmalloc(sk, sk->sk_prot->max_header, 1, |
7c657876 ACM |
399 | GFP_ATOMIC); |
400 | if (skb == NULL) | |
401 | return NULL; | |
402 | ||
403 | /* Reserve space for headers. */ | |
118b2c95 | 404 | skb_reserve(skb, sk->sk_prot->max_header); |
7c657876 | 405 | |
adf30907 | 406 | skb_dst_set(skb, dst_clone(dst)); |
7c657876 | 407 | |
67e6b629 | 408 | dreq = dccp_rsk(req); |
e11d9d30 GR |
409 | if (inet_rsk(req)->acked) /* increase ISS upon retransmission */ |
410 | dccp_inc_seqno(&dreq->dreq_iss); | |
7c657876 | 411 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_RESPONSE; |
67e6b629 | 412 | DCCP_SKB_CB(skb)->dccpd_seq = dreq->dreq_iss; |
2d0817d1 | 413 | |
0c116839 GR |
414 | /* Resolve feature dependencies resulting from choice of CCID */ |
415 | if (dccp_feat_server_ccid_dependencies(dreq)) | |
416 | goto response_failed; | |
417 | ||
418 | if (dccp_insert_options_rsk(dreq, skb)) | |
419 | goto response_failed; | |
7c657876 | 420 | |
09dbc389 | 421 | /* Build and checksum header */ |
9b42078e | 422 | dh = dccp_zeroed_hdr(skb, dccp_header_size); |
7c657876 | 423 | |
944f7502 | 424 | dh->dccph_sport = inet_rsk(req)->loc_port; |
7c657876 | 425 | dh->dccph_dport = inet_rsk(req)->rmt_port; |
7690af3f ACM |
426 | dh->dccph_doff = (dccp_header_size + |
427 | DCCP_SKB_CB(skb)->dccpd_opt_len) / 4; | |
7c657876 ACM |
428 | dh->dccph_type = DCCP_PKT_RESPONSE; |
429 | dh->dccph_x = 1; | |
67e6b629 ACM |
430 | dccp_hdr_set_seq(dh, dreq->dreq_iss); |
431 | dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dreq->dreq_isr); | |
432 | dccp_hdr_response(skb)->dccph_resp_service = dreq->dreq_service; | |
7c657876 | 433 | |
6f4e5fff GR |
434 | dccp_csum_outgoing(skb); |
435 | ||
e11d9d30 GR |
436 | /* We use `acked' to remember that a Response was already sent. */ |
437 | inet_rsk(req)->acked = 1; | |
7c657876 ACM |
438 | DCCP_INC_STATS(DCCP_MIB_OUTSEGS); |
439 | return skb; | |
0c116839 GR |
440 | response_failed: |
441 | kfree_skb(skb); | |
442 | return NULL; | |
7c657876 ACM |
443 | } |
444 | ||
f21e68ca ACM |
445 | EXPORT_SYMBOL_GPL(dccp_make_response); |
446 | ||
e356d37a | 447 | /* answer offending packet in @rcv_skb with Reset from control socket @ctl */ |
7630f026 | 448 | struct sk_buff *dccp_ctl_make_reset(struct sock *sk, struct sk_buff *rcv_skb) |
e356d37a GR |
449 | { |
450 | struct dccp_hdr *rxdh = dccp_hdr(rcv_skb), *dh; | |
451 | struct dccp_skb_cb *dcb = DCCP_SKB_CB(rcv_skb); | |
452 | const u32 dccp_hdr_reset_len = sizeof(struct dccp_hdr) + | |
453 | sizeof(struct dccp_hdr_ext) + | |
454 | sizeof(struct dccp_hdr_reset); | |
455 | struct dccp_hdr_reset *dhr; | |
456 | struct sk_buff *skb; | |
457 | ||
7630f026 | 458 | skb = alloc_skb(sk->sk_prot->max_header, GFP_ATOMIC); |
e356d37a GR |
459 | if (skb == NULL) |
460 | return NULL; | |
461 | ||
7630f026 | 462 | skb_reserve(skb, sk->sk_prot->max_header); |
e356d37a GR |
463 | |
464 | /* Swap the send and the receive. */ | |
465 | dh = dccp_zeroed_hdr(skb, dccp_hdr_reset_len); | |
466 | dh->dccph_type = DCCP_PKT_RESET; | |
467 | dh->dccph_sport = rxdh->dccph_dport; | |
468 | dh->dccph_dport = rxdh->dccph_sport; | |
469 | dh->dccph_doff = dccp_hdr_reset_len / 4; | |
470 | dh->dccph_x = 1; | |
471 | ||
472 | dhr = dccp_hdr_reset(skb); | |
473 | dhr->dccph_reset_code = dcb->dccpd_reset_code; | |
474 | ||
475 | switch (dcb->dccpd_reset_code) { | |
476 | case DCCP_RESET_CODE_PACKET_ERROR: | |
477 | dhr->dccph_reset_data[0] = rxdh->dccph_type; | |
478 | break; | |
479 | case DCCP_RESET_CODE_OPTION_ERROR: /* fall through */ | |
480 | case DCCP_RESET_CODE_MANDATORY_ERROR: | |
481 | memcpy(dhr->dccph_reset_data, dcb->dccpd_reset_data, 3); | |
482 | break; | |
483 | } | |
484 | /* | |
485 | * From RFC 4340, 8.3.1: | |
486 | * If P.ackno exists, set R.seqno := P.ackno + 1. | |
487 | * Else set R.seqno := 0. | |
488 | */ | |
489 | if (dcb->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ) | |
490 | dccp_hdr_set_seq(dh, ADD48(dcb->dccpd_ack_seq, 1)); | |
491 | dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dcb->dccpd_seq); | |
492 | ||
493 | dccp_csum_outgoing(skb); | |
494 | return skb; | |
495 | } | |
496 | ||
497 | EXPORT_SYMBOL_GPL(dccp_ctl_make_reset); | |
498 | ||
ee1a1592 | 499 | /* send Reset on established socket, to close or abort the connection */ |
017487d7 ACM |
500 | int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code) |
501 | { | |
ee1a1592 | 502 | struct sk_buff *skb; |
017487d7 ACM |
503 | /* |
504 | * FIXME: what if rebuild_header fails? | |
505 | * Should we be doing a rebuild_header here? | |
506 | */ | |
f53dc67c | 507 | int err = inet_csk(sk)->icsk_af_ops->rebuild_header(sk); |
017487d7 | 508 | |
ee1a1592 GR |
509 | if (err != 0) |
510 | return err; | |
511 | ||
512 | skb = sock_wmalloc(sk, sk->sk_prot->max_header, 1, GFP_ATOMIC); | |
513 | if (skb == NULL) | |
514 | return -ENOBUFS; | |
515 | ||
516 | /* Reserve space for headers and prepare control bits. */ | |
517 | skb_reserve(skb, sk->sk_prot->max_header); | |
518 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_RESET; | |
519 | DCCP_SKB_CB(skb)->dccpd_reset_code = code; | |
017487d7 | 520 | |
ee1a1592 | 521 | return dccp_transmit_skb(sk, skb); |
017487d7 ACM |
522 | } |
523 | ||
7c657876 ACM |
524 | /* |
525 | * Do all connect socket setups that can be done AF independent. | |
526 | */ | |
93344af4 | 527 | int dccp_connect(struct sock *sk) |
7c657876 | 528 | { |
93344af4 | 529 | struct sk_buff *skb; |
f21e68ca | 530 | struct dccp_sock *dp = dccp_sk(sk); |
7c657876 ACM |
531 | struct dst_entry *dst = __sk_dst_get(sk); |
532 | struct inet_connection_sock *icsk = inet_csk(sk); | |
533 | ||
534 | sk->sk_err = 0; | |
535 | sock_reset_flag(sk, SOCK_DONE); | |
c9eaf173 | 536 | |
7c657876 ACM |
537 | dccp_sync_mss(sk, dst_mtu(dst)); |
538 | ||
9eca0a47 GR |
539 | /* do not connect if feature negotiation setup fails */ |
540 | if (dccp_feat_finalise_settings(dccp_sk(sk))) | |
541 | return -EPROTO; | |
542 | ||
93344af4 GR |
543 | /* Initialise GAR as per 8.5; AWL/AWH are set in dccp_transmit_skb() */ |
544 | dp->dccps_gar = dp->dccps_iss; | |
7c657876 | 545 | |
118b2c95 | 546 | skb = alloc_skb(sk->sk_prot->max_header, sk->sk_allocation); |
7c657876 ACM |
547 | if (unlikely(skb == NULL)) |
548 | return -ENOBUFS; | |
549 | ||
550 | /* Reserve space for headers. */ | |
118b2c95 | 551 | skb_reserve(skb, sk->sk_prot->max_header); |
7c657876 ACM |
552 | |
553 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_REQUEST; | |
7c657876 | 554 | |
48918a4d | 555 | dccp_skb_entail(sk, skb); |
7c657876 ACM |
556 | dccp_transmit_skb(sk, skb_clone(skb, GFP_KERNEL)); |
557 | DCCP_INC_STATS(DCCP_MIB_ACTIVEOPENS); | |
558 | ||
559 | /* Timer for repeating the REQUEST until an answer. */ | |
93344af4 | 560 | icsk->icsk_retransmits = 0; |
27258ee5 ACM |
561 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, |
562 | icsk->icsk_rto, DCCP_RTO_MAX); | |
7c657876 ACM |
563 | return 0; |
564 | } | |
565 | ||
f21e68ca ACM |
566 | EXPORT_SYMBOL_GPL(dccp_connect); |
567 | ||
7c657876 ACM |
568 | void dccp_send_ack(struct sock *sk) |
569 | { | |
570 | /* If we have been reset, we may not send again. */ | |
571 | if (sk->sk_state != DCCP_CLOSED) { | |
118b2c95 ACM |
572 | struct sk_buff *skb = alloc_skb(sk->sk_prot->max_header, |
573 | GFP_ATOMIC); | |
7c657876 ACM |
574 | |
575 | if (skb == NULL) { | |
576 | inet_csk_schedule_ack(sk); | |
577 | inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN; | |
7690af3f ACM |
578 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, |
579 | TCP_DELACK_MAX, | |
580 | DCCP_RTO_MAX); | |
7c657876 ACM |
581 | return; |
582 | } | |
583 | ||
584 | /* Reserve space for headers */ | |
118b2c95 | 585 | skb_reserve(skb, sk->sk_prot->max_header); |
7c657876 | 586 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_ACK; |
7c657876 ACM |
587 | dccp_transmit_skb(sk, skb); |
588 | } | |
589 | } | |
590 | ||
591 | EXPORT_SYMBOL_GPL(dccp_send_ack); | |
592 | ||
1e2f0e5e | 593 | #if 0 |
727ecc5f | 594 | /* FIXME: Is this still necessary (11.3) - currently nowhere used by DCCP. */ |
7c657876 ACM |
595 | void dccp_send_delayed_ack(struct sock *sk) |
596 | { | |
597 | struct inet_connection_sock *icsk = inet_csk(sk); | |
598 | /* | |
599 | * FIXME: tune this timer. elapsed time fixes the skew, so no problem | |
600 | * with using 2s, and active senders also piggyback the ACK into a | |
601 | * DATAACK packet, so this is really for quiescent senders. | |
602 | */ | |
603 | unsigned long timeout = jiffies + 2 * HZ; | |
604 | ||
605 | /* Use new timeout only if there wasn't a older one earlier. */ | |
606 | if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) { | |
607 | /* If delack timer was blocked or is about to expire, | |
608 | * send ACK now. | |
609 | * | |
610 | * FIXME: check the "about to expire" part | |
611 | */ | |
612 | if (icsk->icsk_ack.blocked) { | |
613 | dccp_send_ack(sk); | |
614 | return; | |
615 | } | |
616 | ||
617 | if (!time_before(timeout, icsk->icsk_ack.timeout)) | |
618 | timeout = icsk->icsk_ack.timeout; | |
619 | } | |
620 | icsk->icsk_ack.pending |= ICSK_ACK_SCHED | ICSK_ACK_TIMER; | |
621 | icsk->icsk_ack.timeout = timeout; | |
622 | sk_reset_timer(sk, &icsk->icsk_delack_timer, timeout); | |
623 | } | |
1e2f0e5e | 624 | #endif |
7c657876 | 625 | |
b0d045ca | 626 | void dccp_send_sync(struct sock *sk, const u64 ackno, |
e92ae93a | 627 | const enum dccp_pkt_type pkt_type) |
7c657876 ACM |
628 | { |
629 | /* | |
630 | * We are not putting this on the write queue, so | |
631 | * dccp_transmit_skb() will set the ownership to this | |
632 | * sock. | |
633 | */ | |
118b2c95 | 634 | struct sk_buff *skb = alloc_skb(sk->sk_prot->max_header, GFP_ATOMIC); |
7c657876 | 635 | |
b0d045ca | 636 | if (skb == NULL) { |
7c657876 | 637 | /* FIXME: how to make sure the sync is sent? */ |
b0d045ca | 638 | DCCP_CRIT("could not send %s", dccp_packet_name(pkt_type)); |
7c657876 | 639 | return; |
b0d045ca | 640 | } |
7c657876 ACM |
641 | |
642 | /* Reserve space for headers and prepare control bits. */ | |
118b2c95 | 643 | skb_reserve(skb, sk->sk_prot->max_header); |
e92ae93a | 644 | DCCP_SKB_CB(skb)->dccpd_type = pkt_type; |
b0d045ca | 645 | DCCP_SKB_CB(skb)->dccpd_ack_seq = ackno; |
7c657876 | 646 | |
d83447f0 GR |
647 | /* |
648 | * Clear the flag in case the Sync was scheduled for out-of-band data, | |
649 | * such as carrying a long Ack Vector. | |
650 | */ | |
651 | dccp_sk(sk)->dccps_sync_scheduled = 0; | |
652 | ||
7c657876 ACM |
653 | dccp_transmit_skb(sk, skb); |
654 | } | |
655 | ||
b61fafc4 ACM |
656 | EXPORT_SYMBOL_GPL(dccp_send_sync); |
657 | ||
7690af3f ACM |
658 | /* |
659 | * Send a DCCP_PKT_CLOSE/CLOSEREQ. The caller locks the socket for us. This | |
660 | * cannot be allowed to fail queueing a DCCP_PKT_CLOSE/CLOSEREQ frame under | |
661 | * any circumstances. | |
7c657876 | 662 | */ |
7ad07e7c | 663 | void dccp_send_close(struct sock *sk, const int active) |
7c657876 ACM |
664 | { |
665 | struct dccp_sock *dp = dccp_sk(sk); | |
666 | struct sk_buff *skb; | |
7d877f3b | 667 | const gfp_t prio = active ? GFP_KERNEL : GFP_ATOMIC; |
7c657876 | 668 | |
7ad07e7c ACM |
669 | skb = alloc_skb(sk->sk_prot->max_header, prio); |
670 | if (skb == NULL) | |
671 | return; | |
7c657876 ACM |
672 | |
673 | /* Reserve space for headers and prepare control bits. */ | |
674 | skb_reserve(skb, sk->sk_prot->max_header); | |
b8599d20 GR |
675 | if (dp->dccps_role == DCCP_ROLE_SERVER && !dp->dccps_server_timewait) |
676 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_CLOSEREQ; | |
677 | else | |
678 | DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_CLOSE; | |
7c657876 | 679 | |
7ad07e7c | 680 | if (active) { |
48918a4d | 681 | dccp_skb_entail(sk, skb); |
7ad07e7c | 682 | dccp_transmit_skb(sk, skb_clone(skb, prio)); |
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); | |
7ad07e7c ACM |
695 | } else |
696 | dccp_transmit_skb(sk, skb); | |
7c657876 | 697 | } |