output.c 14.7 KB
Newer Older
1 2
/*
 *  net/dccp/output.c
3
 *
4 5 6 7 8 9 10 11 12 13
 *  An implementation of the DCCP protocol
 *  Arnaldo Carvalho de Melo <acme@conectiva.com.br>
 *
 *	This program is free software; you can redistribute it and/or
 *	modify it under the terms of the GNU General Public License
 *	as published by the Free Software Foundation; either version
 *	2 of the License, or (at your option) any later version.
 */

#include <linux/dccp.h>
14
#include <linux/kernel.h>
15 16
#include <linux/skbuff.h>

17
#include <net/inet_sock.h>
18 19
#include <net/sock.h>

20
#include "ackvec.h"
21 22 23 24 25 26 27 28
#include "ccid.h"
#include "dccp.h"

static inline void dccp_event_ack_sent(struct sock *sk)
{
	inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
}

29
static void dccp_skb_entail(struct sock *sk, struct sk_buff *skb)
30 31 32 33 34 35
{
	skb_set_owner_w(skb, sk);
	WARN_ON(sk->sk_send_head);
	sk->sk_send_head = skb;
}

36 37 38 39 40 41
/*
 * All SKB's seen here are completely headerless. It is our
 * job to build the DCCP header, and pass the packet down to
 * IP so it can do the same plus pass the packet off to the
 * device.
 */
42
static int dccp_transmit_skb(struct sock *sk, struct sk_buff *skb)
43 44 45
{
	if (likely(skb != NULL)) {
		const struct inet_sock *inet = inet_sk(sk);
46
		const struct inet_connection_sock *icsk = inet_csk(sk);
47 48 49 50
		struct dccp_sock *dp = dccp_sk(sk);
		struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
		struct dccp_hdr *dh;
		/* XXX For now we're using only 48 bits sequence numbers */
51
		const u32 dccp_header_size = sizeof(*dh) +
52
					     sizeof(struct dccp_hdr_ext) +
53
					  dccp_packet_hdr_len(dcb->dccpd_type);
54 55 56 57 58 59 60 61
		int err, set_ack = 1;
		u64 ackno = dp->dccps_gsr;

		dccp_inc_seqno(&dp->dccps_gss);

		switch (dcb->dccpd_type) {
		case DCCP_PKT_DATA:
			set_ack = 0;
62 63
			/* fall through */
		case DCCP_PKT_DATAACK:
64
			break;
65

66 67 68 69
		case DCCP_PKT_REQUEST:
			set_ack = 0;
			/* fall through */

70 71
		case DCCP_PKT_SYNC:
		case DCCP_PKT_SYNCACK:
G
Gerrit Renker 已提交
72
			ackno = dcb->dccpd_ack_seq;
73 74 75 76 77 78 79 80
			/* fall through */
		default:
			/*
			 * Only data packets should come through with skb->sk
			 * set.
			 */
			WARN_ON(skb->sk);
			skb_set_owner_w(skb, sk);
81 82
			break;
		}
83 84

		dcb->dccpd_seq = dp->dccps_gss;
85 86 87 88 89

		if (dccp_insert_options(sk, skb)) {
			kfree_skb(skb);
			return -EPROTO;
		}
90

91

92
		/* Build DCCP header and checksum it. */
93
		dh = dccp_zeroed_hdr(skb, dccp_header_size);
94 95 96 97 98
		dh->dccph_type	= dcb->dccpd_type;
		dh->dccph_sport	= inet->sport;
		dh->dccph_dport	= inet->dport;
		dh->dccph_doff	= (dccp_header_size + dcb->dccpd_opt_len) / 4;
		dh->dccph_ccval	= dcb->dccpd_ccval;
99
		dh->dccph_cscov = dp->dccps_pcslen;
100 101 102 103 104 105 106 107 108 109
		/* XXX For now we're using only 48 bits sequence numbers */
		dh->dccph_x	= 1;

		dp->dccps_awh = dp->dccps_gss;
		dccp_hdr_set_seq(dh, dp->dccps_gss);
		if (set_ack)
			dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), ackno);

		switch (dcb->dccpd_type) {
		case DCCP_PKT_REQUEST:
110
			dccp_hdr_request(skb)->dccph_req_service =
111
							dp->dccps_service;
112 113
			break;
		case DCCP_PKT_RESET:
114 115
			dccp_hdr_reset(skb)->dccph_reset_code =
							dcb->dccpd_reset_code;
116 117 118
			break;
		}

119
		icsk->icsk_af_ops->send_check(sk, 0, skb);
120

121
		if (set_ack)
122 123 124 125
			dccp_event_ack_sent(sk);

		DCCP_INC_STATS(DCCP_MIB_OUTSEGS);

H
Herbert Xu 已提交
126
		memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
127
		err = icsk->icsk_af_ops->queue_xmit(skb, 0);
128
		return net_xmit_eval(err);
129 130 131 132 133 134
	}
	return -ENOBUFS;
}

unsigned int dccp_sync_mss(struct sock *sk, u32 pmtu)
{
135
	struct inet_connection_sock *icsk = inet_csk(sk);
136
	struct dccp_sock *dp = dccp_sk(sk);
137
	int mss_now = (pmtu - icsk->icsk_af_ops->net_header_len -
138
		       sizeof(struct dccp_hdr) - sizeof(struct dccp_hdr_ext));
139 140

	/* Now subtract optional transport overhead */
141
	mss_now -= icsk->icsk_ext_hdr_len;
142 143 144 145 146 147 148 149 150 151 152 153

	/*
	 * FIXME: this should come from the CCID infrastructure, where, say,
	 * TFRC will say it wants TIMESTAMPS, ELAPSED time, etc, for now lets
	 * put a rough estimate for NDP + TIMESTAMP + TIMESTAMP_ECHO + ELAPSED
	 * TIME + TFRC_OPT_LOSS_EVENT_RATE + TFRC_OPT_RECEIVE_RATE + padding to
	 * make it a multiple of 4
	 */

	mss_now -= ((5 + 6 + 10 + 6 + 6 + 6 + 3) / 4) * 4;

	/* And store cached results */
154
	icsk->icsk_pmtu_cookie = pmtu;
155 156 157 158 159
	dp->dccps_mss_cache = mss_now;

	return mss_now;
}

160 161
EXPORT_SYMBOL_GPL(dccp_sync_mss);

162 163 164 165 166 167 168 169 170 171 172 173 174
void dccp_write_space(struct sock *sk)
{
	read_lock(&sk->sk_callback_lock);

	if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
		wake_up_interruptible(sk->sk_sleep);
	/* Should agree with poll, otherwise some programs break */
	if (sock_writeable(sk))
		sk_wake_async(sk, 2, POLL_OUT);

	read_unlock(&sk->sk_callback_lock);
}

175 176 177 178
/**
 * dccp_wait_for_ccid - Wait for ccid to tell us we can send a packet
 * @sk: socket to wait for
 */
I
Ian McDonald 已提交
179
static int dccp_wait_for_ccid(struct sock *sk, struct sk_buff *skb)
180 181 182
{
	struct dccp_sock *dp = dccp_sk(sk);
	DEFINE_WAIT(wait);
I
Ian McDonald 已提交
183
	unsigned long delay;
184 185 186 187 188
	int rc;

	while (1) {
		prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);

I
Ian McDonald 已提交
189
		if (sk->sk_err)
190 191 192 193
			goto do_error;
		if (signal_pending(current))
			goto do_interrupted;

194
		rc = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb);
195 196
		if (rc <= 0)
			break;
197
		dccp_pr_debug("delayed send by %d msec\n", rc);
198 199 200
		delay = msecs_to_jiffies(rc);
		sk->sk_write_pending++;
		release_sock(sk);
I
Ian McDonald 已提交
201
		schedule_timeout(delay);
202 203 204 205 206 207 208 209 210 211 212
		lock_sock(sk);
		sk->sk_write_pending--;
	}
out:
	finish_wait(sk->sk_sleep, &wait);
	return rc;

do_error:
	rc = -EPIPE;
	goto out;
do_interrupted:
I
Ian McDonald 已提交
213
	rc = -EINTR;
214 215 216
	goto out;
}

I
Ian McDonald 已提交
217
void dccp_write_xmit(struct sock *sk, int block)
218
{
I
Ian McDonald 已提交
219 220 221 222
	struct dccp_sock *dp = dccp_sk(sk);
	struct sk_buff *skb;

	while ((skb = skb_peek(&sk->sk_write_queue))) {
223
		int err = ccid_hc_tx_send_packet(dp->dccps_hc_tx_ccid, sk, skb);
224

I
Ian McDonald 已提交
225 226 227 228 229
		if (err > 0) {
			if (!block) {
				sk_reset_timer(sk, &dp->dccps_xmit_timer,
						msecs_to_jiffies(err)+jiffies);
				break;
I
Ian McDonald 已提交
230 231
			} else
				err = dccp_wait_for_ccid(sk, skb);
232
			if (err && err != -EINTR)
233
				DCCP_BUG("err=%d after dccp_wait_for_ccid", err);
I
Ian McDonald 已提交
234
		}
235

I
Ian McDonald 已提交
236 237 238 239
		skb_dequeue(&sk->sk_write_queue);
		if (err == 0) {
			struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
			const int len = skb->len;
240

I
Ian McDonald 已提交
241 242 243 244
			if (sk->sk_state == DCCP_PARTOPEN) {
				/* See 8.1.5.  Handshake Completion */
				inet_csk_schedule_ack(sk);
				inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
245 246
						  inet_csk(sk)->icsk_rto,
						  DCCP_RTO_MAX);
I
Ian McDonald 已提交
247 248 249 250 251 252 253 254
				dcb->dccpd_type = DCCP_PKT_DATAACK;
			} else if (dccp_ack_pending(sk))
				dcb->dccpd_type = DCCP_PKT_DATAACK;
			else
				dcb->dccpd_type = DCCP_PKT_DATA;

			err = dccp_transmit_skb(sk, skb);
			ccid_hc_tx_packet_sent(dp->dccps_hc_tx_ccid, sk, 0, len);
255 256 257
			if (err)
				DCCP_BUG("err=%d after ccid_hc_tx_packet_sent",
					 err);
258
		} else {
259
			dccp_pr_debug("packet discarded due to err=%d\n", err);
P
Patrick McHardy 已提交
260
			kfree_skb(skb);
261
		}
I
Ian McDonald 已提交
262
	}
263 264
}

265 266
int dccp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
{
267
	if (inet_csk(sk)->icsk_af_ops->rebuild_header(sk) != 0)
268 269 270 271 272 273 274 275 276 277 278
		return -EHOSTUNREACH; /* Routing failure or similar. */

	return dccp_transmit_skb(sk, (skb_cloned(skb) ?
				      pskb_copy(skb, GFP_ATOMIC):
				      skb_clone(skb, GFP_ATOMIC)));
}

struct sk_buff *dccp_make_response(struct sock *sk, struct dst_entry *dst,
				   struct request_sock *req)
{
	struct dccp_hdr *dh;
279
	struct dccp_request_sock *dreq;
280
	const u32 dccp_header_size = sizeof(struct dccp_hdr) +
281 282
				     sizeof(struct dccp_hdr_ext) +
				     sizeof(struct dccp_hdr_response);
283
	struct sk_buff *skb = sock_wmalloc(sk, sk->sk_prot->max_header, 1,
284 285 286 287 288
					   GFP_ATOMIC);
	if (skb == NULL)
		return NULL;

	/* Reserve space for headers. */
289
	skb_reserve(skb, sk->sk_prot->max_header);
290 291 292

	skb->dst = dst_clone(dst);

293
	dreq = dccp_rsk(req);
294 295
	if (inet_rsk(req)->acked)	/* increase ISS upon retransmission */
		dccp_inc_seqno(&dreq->dreq_iss);
296
	DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_RESPONSE;
297
	DCCP_SKB_CB(skb)->dccpd_seq  = dreq->dreq_iss;
298 299 300 301 302

	if (dccp_insert_options(sk, skb)) {
		kfree_skb(skb);
		return NULL;
	}
303

304
	/* Build and checksum header */
305
	dh = dccp_zeroed_hdr(skb, dccp_header_size);
306 307 308

	dh->dccph_sport	= inet_sk(sk)->sport;
	dh->dccph_dport	= inet_rsk(req)->rmt_port;
309 310
	dh->dccph_doff	= (dccp_header_size +
			   DCCP_SKB_CB(skb)->dccpd_opt_len) / 4;
311 312
	dh->dccph_type	= DCCP_PKT_RESPONSE;
	dh->dccph_x	= 1;
313 314 315
	dccp_hdr_set_seq(dh, dreq->dreq_iss);
	dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dreq->dreq_isr);
	dccp_hdr_response(skb)->dccph_resp_service = dreq->dreq_service;
316

317 318
	dccp_csum_outgoing(skb);

319 320
	/* We use `acked' to remember that a Response was already sent. */
	inet_rsk(req)->acked = 1;
321 322 323 324
	DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
	return skb;
}

325 326
EXPORT_SYMBOL_GPL(dccp_make_response);

327 328
static struct sk_buff *dccp_make_reset(struct sock *sk, struct dst_entry *dst,
				       const enum dccp_reset_codes code)
329 330 331
{
	struct dccp_hdr *dh;
	struct dccp_sock *dp = dccp_sk(sk);
332
	const u32 dccp_header_size = sizeof(struct dccp_hdr) +
333 334
				     sizeof(struct dccp_hdr_ext) +
				     sizeof(struct dccp_hdr_reset);
335
	struct sk_buff *skb = sock_wmalloc(sk, sk->sk_prot->max_header, 1,
336 337 338 339 340
					   GFP_ATOMIC);
	if (skb == NULL)
		return NULL;

	/* Reserve space for headers. */
341
	skb_reserve(skb, sk->sk_prot->max_header);
342 343 344 345 346 347 348 349

	skb->dst = dst_clone(dst);

	dccp_inc_seqno(&dp->dccps_gss);

	DCCP_SKB_CB(skb)->dccpd_reset_code = code;
	DCCP_SKB_CB(skb)->dccpd_type	   = DCCP_PKT_RESET;
	DCCP_SKB_CB(skb)->dccpd_seq	   = dp->dccps_gss;
350 351 352 353 354

	if (dccp_insert_options(sk, skb)) {
		kfree_skb(skb);
		return NULL;
	}
355

356
	dh = dccp_zeroed_hdr(skb, dccp_header_size);
357 358 359

	dh->dccph_sport	= inet_sk(sk)->sport;
	dh->dccph_dport	= inet_sk(sk)->dport;
360 361
	dh->dccph_doff	= (dccp_header_size +
			   DCCP_SKB_CB(skb)->dccpd_opt_len) / 4;
362 363 364 365 366 367
	dh->dccph_type	= DCCP_PKT_RESET;
	dh->dccph_x	= 1;
	dccp_hdr_set_seq(dh, dp->dccps_gss);
	dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), dp->dccps_gsr);

	dccp_hdr_reset(skb)->dccph_reset_code = code;
368
	inet_csk(sk)->icsk_af_ops->send_check(sk, 0, skb);
369 370 371 372 373

	DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
	return skb;
}

374 375 376 377 378 379 380 381 382 383 384 385 386
int dccp_send_reset(struct sock *sk, enum dccp_reset_codes code)
{
	/*
	 * FIXME: what if rebuild_header fails?
	 * Should we be doing a rebuild_header here?
	 */
	int err = inet_sk_rebuild_header(sk);

	if (err == 0) {
		struct sk_buff *skb = dccp_make_reset(sk, sk->sk_dst_cache,
						      code);
		if (skb != NULL) {
			memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
387
			err = inet_csk(sk)->icsk_af_ops->queue_xmit(skb, 0);
388
			return net_xmit_eval(err);
389 390 391 392 393 394
		}
	}

	return err;
}

395 396 397 398 399
/*
 * Do all connect socket setups that can be done AF independent.
 */
static inline void dccp_connect_init(struct sock *sk)
{
400
	struct dccp_sock *dp = dccp_sk(sk);
401 402 403 404 405
	struct dst_entry *dst = __sk_dst_get(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);

	sk->sk_err = 0;
	sock_reset_flag(sk, SOCK_DONE);
406

407 408
	dccp_sync_mss(sk, dst_mtu(dst));

409
	/*
410 411 412 413 414 415
	 * SWL and AWL are initially adjusted so that they are not less than
	 * the initial Sequence Numbers received and sent, respectively:
	 *	SWL := max(GSR + 1 - floor(W/4), ISR),
	 *	AWL := max(GSS - W' + 1, ISS).
	 * These adjustments MUST be applied only at the beginning of the
	 * connection.
416
	 */
417
	dccp_update_gss(sk, dp->dccps_iss);
418
	dccp_set_seqno(&dp->dccps_awl, max48(dp->dccps_awl, dp->dccps_iss));
419

420 421 422 423
	/* S.GAR - greatest valid acknowledgement number received on a non-Sync;
	 *         initialized to S.ISS (sec. 8.5)                            */
	dp->dccps_gar = dp->dccps_iss;

424 425 426 427 428 429 430 431 432 433
	icsk->icsk_retransmits = 0;
}

int dccp_connect(struct sock *sk)
{
	struct sk_buff *skb;
	struct inet_connection_sock *icsk = inet_csk(sk);

	dccp_connect_init(sk);

434
	skb = alloc_skb(sk->sk_prot->max_header, sk->sk_allocation);
435 436 437 438
	if (unlikely(skb == NULL))
		return -ENOBUFS;

	/* Reserve space for headers. */
439
	skb_reserve(skb, sk->sk_prot->max_header);
440 441 442

	DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_REQUEST;

443
	dccp_skb_entail(sk, skb);
444 445 446 447
	dccp_transmit_skb(sk, skb_clone(skb, GFP_KERNEL));
	DCCP_INC_STATS(DCCP_MIB_ACTIVEOPENS);

	/* Timer for repeating the REQUEST until an answer. */
448 449
	inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
				  icsk->icsk_rto, DCCP_RTO_MAX);
450 451 452
	return 0;
}

453 454
EXPORT_SYMBOL_GPL(dccp_connect);

455 456 457 458
void dccp_send_ack(struct sock *sk)
{
	/* If we have been reset, we may not send again. */
	if (sk->sk_state != DCCP_CLOSED) {
459 460
		struct sk_buff *skb = alloc_skb(sk->sk_prot->max_header,
						GFP_ATOMIC);
461 462 463 464

		if (skb == NULL) {
			inet_csk_schedule_ack(sk);
			inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN;
465 466 467
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
						  TCP_DELACK_MAX,
						  DCCP_RTO_MAX);
468 469 470 471
			return;
		}

		/* Reserve space for headers */
472
		skb_reserve(skb, sk->sk_prot->max_header);
473 474 475 476 477 478 479
		DCCP_SKB_CB(skb)->dccpd_type = DCCP_PKT_ACK;
		dccp_transmit_skb(sk, skb);
	}
}

EXPORT_SYMBOL_GPL(dccp_send_ack);

480
/* FIXME: Is this still necessary (11.3) - currently nowhere used by DCCP. */
481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510
void dccp_send_delayed_ack(struct sock *sk)
{
	struct inet_connection_sock *icsk = inet_csk(sk);
	/*
	 * FIXME: tune this timer. elapsed time fixes the skew, so no problem
	 * with using 2s, and active senders also piggyback the ACK into a
	 * DATAACK packet, so this is really for quiescent senders.
	 */
	unsigned long timeout = jiffies + 2 * HZ;

	/* Use new timeout only if there wasn't a older one earlier. */
	if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) {
		/* If delack timer was blocked or is about to expire,
		 * send ACK now.
		 *
		 * FIXME: check the "about to expire" part
		 */
		if (icsk->icsk_ack.blocked) {
			dccp_send_ack(sk);
			return;
		}

		if (!time_before(timeout, icsk->icsk_ack.timeout))
			timeout = icsk->icsk_ack.timeout;
	}
	icsk->icsk_ack.pending |= ICSK_ACK_SCHED | ICSK_ACK_TIMER;
	icsk->icsk_ack.timeout = timeout;
	sk_reset_timer(sk, &icsk->icsk_delack_timer, timeout);
}

G
Gerrit Renker 已提交
511
void dccp_send_sync(struct sock *sk, const u64 ackno,
512
		    const enum dccp_pkt_type pkt_type)
513 514 515 516 517 518
{
	/*
	 * We are not putting this on the write queue, so
	 * dccp_transmit_skb() will set the ownership to this
	 * sock.
	 */
519
	struct sk_buff *skb = alloc_skb(sk->sk_prot->max_header, GFP_ATOMIC);
520

G
Gerrit Renker 已提交
521
	if (skb == NULL) {
522
		/* FIXME: how to make sure the sync is sent? */
G
Gerrit Renker 已提交
523
		DCCP_CRIT("could not send %s", dccp_packet_name(pkt_type));
524
		return;
G
Gerrit Renker 已提交
525
	}
526 527

	/* Reserve space for headers and prepare control bits. */
528
	skb_reserve(skb, sk->sk_prot->max_header);
529
	DCCP_SKB_CB(skb)->dccpd_type = pkt_type;
G
Gerrit Renker 已提交
530
	DCCP_SKB_CB(skb)->dccpd_ack_seq = ackno;
531 532 533 534

	dccp_transmit_skb(sk, skb);
}

535 536
EXPORT_SYMBOL_GPL(dccp_send_sync);

537 538 539 540
/*
 * Send a DCCP_PKT_CLOSE/CLOSEREQ. The caller locks the socket for us. This
 * cannot be allowed to fail queueing a DCCP_PKT_CLOSE/CLOSEREQ frame under
 * any circumstances.
541
 */
542
void dccp_send_close(struct sock *sk, const int active)
543 544 545
{
	struct dccp_sock *dp = dccp_sk(sk);
	struct sk_buff *skb;
A
Al Viro 已提交
546
	const gfp_t prio = active ? GFP_KERNEL : GFP_ATOMIC;
547

548 549 550
	skb = alloc_skb(sk->sk_prot->max_header, prio);
	if (skb == NULL)
		return;
551 552 553

	/* Reserve space for headers and prepare control bits. */
	skb_reserve(skb, sk->sk_prot->max_header);
554 555
	DCCP_SKB_CB(skb)->dccpd_type = dp->dccps_role == DCCP_ROLE_CLIENT ?
					DCCP_PKT_CLOSE : DCCP_PKT_CLOSEREQ;
556

557
	if (active) {
I
Ian McDonald 已提交
558
		dccp_write_xmit(sk, 1);
559
		dccp_skb_entail(sk, skb);
560
		dccp_transmit_skb(sk, skb_clone(skb, prio));
I
Ian McDonald 已提交
561
		/* FIXME do we need a retransmit timer here? */
562 563
	} else
		dccp_transmit_skb(sk, skb);
564
}