ccid2.c 18.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
/*
 *  Copyright (c) 2005, 2006 Andrea Bittau <a.bittau@cs.ucl.ac.uk>
 *
 *  Changes to meet Linux coding standards, and DCCP infrastructure fixes.
 *
 *  Copyright (c) 2006 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.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

/*
24
 * This implementation should follow RFC 4341
25
 */
26
#include <linux/slab.h>
27
#include "../feat.h"
28 29 30
#include "ccid2.h"


31
#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
32 33
static int ccid2_debug;
#define ccid2_pr_debug(format, a...)	DCCP_PR_DEBUG(ccid2_debug, format, ##a)
34
#else
35
#define ccid2_pr_debug(format, a...)
36 37
#endif

38
static int ccid2_hc_tx_alloc_seq(struct ccid2_hc_tx_sock *hc)
39 40 41 42 43
{
	struct ccid2_seq *seqp;
	int i;

	/* check if we have space to preserve the pointer to the buffer */
44 45
	if (hc->tx_seqbufc >= (sizeof(hc->tx_seqbuf) /
			       sizeof(struct ccid2_seq *)))
46 47 48
		return -ENOMEM;

	/* allocate buffer and initialize linked list */
G
Gerrit Renker 已提交
49
	seqp = kmalloc(CCID2_SEQBUF_LEN * sizeof(struct ccid2_seq), gfp_any());
50 51 52
	if (seqp == NULL)
		return -ENOMEM;

G
Gerrit Renker 已提交
53
	for (i = 0; i < (CCID2_SEQBUF_LEN - 1); i++) {
54 55 56
		seqp[i].ccid2s_next = &seqp[i + 1];
		seqp[i + 1].ccid2s_prev = &seqp[i];
	}
G
Gerrit Renker 已提交
57 58
	seqp[CCID2_SEQBUF_LEN - 1].ccid2s_next = seqp;
	seqp->ccid2s_prev = &seqp[CCID2_SEQBUF_LEN - 1];
59 60

	/* This is the first allocation.  Initiate the head and tail.  */
61 62
	if (hc->tx_seqbufc == 0)
		hc->tx_seqh = hc->tx_seqt = seqp;
63 64
	else {
		/* link the existing list with the one we just created */
65 66
		hc->tx_seqh->ccid2s_next = seqp;
		seqp->ccid2s_prev = hc->tx_seqh;
67

68 69
		hc->tx_seqt->ccid2s_prev = &seqp[CCID2_SEQBUF_LEN - 1];
		seqp[CCID2_SEQBUF_LEN - 1].ccid2s_next = hc->tx_seqt;
70 71 72
	}

	/* store the original pointer to the buffer so we can free it */
73 74
	hc->tx_seqbuf[hc->tx_seqbufc] = seqp;
	hc->tx_seqbufc++;
75 76 77 78

	return 0;
}

79
static int ccid2_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
80
{
G
Gerrit Renker 已提交
81 82 83
	if (ccid2_cwnd_network_limited(ccid2_hc_tx_sk(sk)))
		return CCID_PACKET_WILL_DEQUEUE_LATER;
	return CCID_PACKET_SEND_AT_ONCE;
84 85
}

86
static void ccid2_change_l_ack_ratio(struct sock *sk, u32 val)
87 88
{
	struct dccp_sock *dp = dccp_sk(sk);
89
	u32 max_ratio = DIV_ROUND_UP(ccid2_hc_tx_sk(sk)->tx_cwnd, 2);
90

91
	/*
92 93 94 95
	 * Ensure that Ack Ratio does not exceed ceil(cwnd/2), which is (2) from
	 * RFC 4341, 6.1.2. We ignore the statement that Ack Ratio 2 is always
	 * acceptable since this causes starvation/deadlock whenever cwnd < 2.
	 * The same problem arises when Ack Ratio is 0 (ie. Ack Ratio disabled).
96
	 */
97 98 99
	if (val == 0 || val > max_ratio) {
		DCCP_WARN("Limiting Ack Ratio (%u) to %u\n", val, max_ratio);
		val = max_ratio;
100
	}
101 102
	if (val > DCCPF_ACK_RATIO_MAX)
		val = DCCPF_ACK_RATIO_MAX;
103

104 105 106
	if (val == dp->dccps_l_ack_ratio)
		return;

107
	ccid2_pr_debug("changing local ack ratio to %u\n", val);
108 109 110 111 112 113
	dp->dccps_l_ack_ratio = val;
}

static void ccid2_hc_tx_rto_expire(unsigned long data)
{
	struct sock *sk = (struct sock *)data;
114
	struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
G
Gerrit Renker 已提交
115
	const bool sender_was_blocked = ccid2_cwnd_network_limited(hc);
116 117 118

	bh_lock_sock(sk);
	if (sock_owned_by_user(sk)) {
119
		sk_reset_timer(sk, &hc->tx_rtotimer, jiffies + HZ / 5);
120 121 122 123 124 125
		goto out;
	}

	ccid2_pr_debug("RTO_EXPIRE\n");

	/* back-off timer */
126
	hc->tx_rto <<= 1;
127 128
	if (hc->tx_rto > DCCP_RTO_MAX)
		hc->tx_rto = DCCP_RTO_MAX;
129

130
	/* adjust pipe, cwnd etc */
131 132 133
	hc->tx_ssthresh = hc->tx_cwnd / 2;
	if (hc->tx_ssthresh < 2)
		hc->tx_ssthresh = 2;
G
Gerrit Renker 已提交
134 135
	hc->tx_cwnd	= 1;
	hc->tx_pipe	= 0;
136 137

	/* clear state about stuff we sent */
138 139
	hc->tx_seqt = hc->tx_seqh;
	hc->tx_packets_acked = 0;
140 141

	/* clear ack ratio state. */
142 143
	hc->tx_rpseq    = 0;
	hc->tx_rpdupack = -1;
144
	ccid2_change_l_ack_ratio(sk, 1);
G
Gerrit Renker 已提交
145 146 147 148 149 150

	/* if we were blocked before, we may now send cwnd=1 packet */
	if (sender_was_blocked)
		tasklet_schedule(&dccp_sk(sk)->dccps_xmitlet);
	/* restart backed-off timer */
	sk_reset_timer(sk, &hc->tx_rtotimer, jiffies + hc->tx_rto);
151 152
out:
	bh_unlock_sock(sk);
153
	sock_put(sk);
154 155
}

156
static void ccid2_hc_tx_packet_sent(struct sock *sk, unsigned int len)
157 158
{
	struct dccp_sock *dp = dccp_sk(sk);
159
	struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
160
	struct ccid2_seq *next;
161

162
	hc->tx_pipe++;
163

164 165
	hc->tx_seqh->ccid2s_seq   = dp->dccps_gss;
	hc->tx_seqh->ccid2s_acked = 0;
166
	hc->tx_seqh->ccid2s_sent  = ccid2_time_stamp;
167

168
	next = hc->tx_seqh->ccid2s_next;
169
	/* check if we need to alloc more space */
170 171
	if (next == hc->tx_seqt) {
		if (ccid2_hc_tx_alloc_seq(hc)) {
G
Gerrit Renker 已提交
172 173 174 175
			DCCP_CRIT("packet history - out of memory!");
			/* FIXME: find a more graceful way to bail out */
			return;
		}
176 177
		next = hc->tx_seqh->ccid2s_next;
		BUG_ON(next == hc->tx_seqt);
178
	}
179
	hc->tx_seqh = next;
180

181
	ccid2_pr_debug("cwnd=%d pipe=%d\n", hc->tx_cwnd, hc->tx_pipe);
182

183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202
	/*
	 * FIXME: The code below is broken and the variables have been removed
	 * from the socket struct. The `ackloss' variable was always set to 0,
	 * and with arsent there are several problems:
	 *  (i) it doesn't just count the number of Acks, but all sent packets;
	 *  (ii) it is expressed in # of packets, not # of windows, so the
	 *  comparison below uses the wrong formula: Appendix A of RFC 4341
	 *  comes up with the number K = cwnd / (R^2 - R) of consecutive windows
	 *  of data with no lost or marked Ack packets. If arsent were the # of
	 *  consecutive Acks received without loss, then Ack Ratio needs to be
	 *  decreased by 1 when
	 *	      arsent >=  K * cwnd / R  =  cwnd^2 / (R^3 - R^2)
	 *  where cwnd / R is the number of Acks received per window of data
	 *  (cf. RFC 4341, App. A). The problems are that
	 *  - arsent counts other packets as well;
	 *  - the comparison uses a formula different from RFC 4341;
	 *  - computing a cubic/quadratic equation each time is too complicated.
	 *  Hence a different algorithm is needed.
	 */
#if 0
203
	/* Ack Ratio.  Need to maintain a concept of how many windows we sent */
204
	hc->tx_arsent++;
205
	/* We had an ack loss in this window... */
206 207 208 209
	if (hc->tx_ackloss) {
		if (hc->tx_arsent >= hc->tx_cwnd) {
			hc->tx_arsent  = 0;
			hc->tx_ackloss = 0;
210
		}
211 212
	} else {
		/* No acks lost up to now... */
213 214 215
		/* decrease ack ratio if enough packets were sent */
		if (dp->dccps_l_ack_ratio > 1) {
			/* XXX don't calculate denominator each time */
216 217
			int denom = dp->dccps_l_ack_ratio * dp->dccps_l_ack_ratio -
				    dp->dccps_l_ack_ratio;
218

219
			denom = hc->tx_cwnd * hc->tx_cwnd / denom;
220

221
			if (hc->tx_arsent >= denom) {
222
				ccid2_change_l_ack_ratio(sk, dp->dccps_l_ack_ratio - 1);
223
				hc->tx_arsent = 0;
224
			}
225 226
		} else {
			/* we can't increase ack ratio further [1] */
227
			hc->tx_arsent = 0; /* or maybe set it to cwnd*/
228 229
		}
	}
230
#endif
231

232
	sk_reset_timer(sk, &hc->tx_rtotimer, jiffies + hc->tx_rto);
233

234
#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
235
	do {
236
		struct ccid2_seq *seqp = hc->tx_seqt;
237

238
		while (seqp != hc->tx_seqh) {
239
			ccid2_pr_debug("out seq=%llu acked=%d time=%u\n",
240
				       (unsigned long long)seqp->ccid2s_seq,
R
Randy Dunlap 已提交
241
				       seqp->ccid2s_acked, seqp->ccid2s_sent);
242 243
			seqp = seqp->ccid2s_next;
		}
244
	} while (0);
245 246 247 248
	ccid2_pr_debug("=========\n");
#endif
}

249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267
/**
 * ccid2_rtt_estimator - Sample RTT and compute RTO using RFC2988 algorithm
 * This code is almost identical with TCP's tcp_rtt_estimator(), since
 * - it has a higher sampling frequency (recommended by RFC 1323),
 * - the RTO does not collapse into RTT due to RTTVAR going towards zero,
 * - it is simple (cf. more complex proposals such as Eifel timer or research
 *   which suggests that the gain should be set according to window size),
 * - in tests it was found to work well with CCID2 [gerrit].
 */
static void ccid2_rtt_estimator(struct sock *sk, const long mrtt)
{
	struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
	long m = mrtt ? : 1;

	if (hc->tx_srtt == 0) {
		/* First measurement m */
		hc->tx_srtt = m << 3;
		hc->tx_mdev = m << 1;

268
		hc->tx_mdev_max = max(hc->tx_mdev, tcp_rto_min(sk));
269
		hc->tx_rttvar   = hc->tx_mdev_max;
270

271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310
		hc->tx_rtt_seq  = dccp_sk(sk)->dccps_gss;
	} else {
		/* Update scaled SRTT as SRTT += 1/8 * (m - SRTT) */
		m -= (hc->tx_srtt >> 3);
		hc->tx_srtt += m;

		/* Similarly, update scaled mdev with regard to |m| */
		if (m < 0) {
			m = -m;
			m -= (hc->tx_mdev >> 2);
			/*
			 * This neutralises RTO increase when RTT < SRTT - mdev
			 * (see P. Sarolahti, A. Kuznetsov,"Congestion Control
			 * in Linux TCP", USENIX 2002, pp. 49-62).
			 */
			if (m > 0)
				m >>= 3;
		} else {
			m -= (hc->tx_mdev >> 2);
		}
		hc->tx_mdev += m;

		if (hc->tx_mdev > hc->tx_mdev_max) {
			hc->tx_mdev_max = hc->tx_mdev;
			if (hc->tx_mdev_max > hc->tx_rttvar)
				hc->tx_rttvar = hc->tx_mdev_max;
		}

		/*
		 * Decay RTTVAR at most once per flight, exploiting that
		 *  1) pipe <= cwnd <= Sequence_Window = W  (RFC 4340, 7.5.2)
		 *  2) AWL = GSS-W+1 <= GAR <= GSS          (RFC 4340, 7.5.1)
		 * GAR is a useful bound for FlightSize = pipe.
		 * AWL is probably too low here, as it over-estimates pipe.
		 */
		if (after48(dccp_sk(sk)->dccps_gar, hc->tx_rtt_seq)) {
			if (hc->tx_mdev_max < hc->tx_rttvar)
				hc->tx_rttvar -= (hc->tx_rttvar -
						  hc->tx_mdev_max) >> 2;
			hc->tx_rtt_seq  = dccp_sk(sk)->dccps_gss;
311
			hc->tx_mdev_max = tcp_rto_min(sk);
312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329
		}
	}

	/*
	 * Set RTO from SRTT and RTTVAR
	 * As in TCP, 4 * RTTVAR >= TCP_RTO_MIN, giving a minimum RTO of 200 ms.
	 * This agrees with RFC 4341, 5:
	 *	"Because DCCP does not retransmit data, DCCP does not require
	 *	 TCP's recommended minimum timeout of one second".
	 */
	hc->tx_rto = (hc->tx_srtt >> 3) + hc->tx_rttvar;

	if (hc->tx_rto > DCCP_RTO_MAX)
		hc->tx_rto = DCCP_RTO_MAX;
}

static void ccid2_new_ack(struct sock *sk, struct ccid2_seq *seqp,
			  unsigned int *maxincr)
330
{
331
	struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
332

333 334 335 336 337
	if (hc->tx_cwnd < hc->tx_ssthresh) {
		if (*maxincr > 0 && ++hc->tx_packets_acked == 2) {
			hc->tx_cwnd += 1;
			*maxincr    -= 1;
			hc->tx_packets_acked = 0;
338
		}
339 340 341
	} else if (++hc->tx_packets_acked >= hc->tx_cwnd) {
			hc->tx_cwnd += 1;
			hc->tx_packets_acked = 0;
342
	}
343 344 345 346 347 348 349 350
	/*
	 * FIXME: RTT is sampled several times per acknowledgment (for each
	 * entry in the Ack Vector), instead of once per Ack (as in TCP SACK).
	 * This causes the RTT to be over-estimated, since the older entries
	 * in the Ack Vector have earlier sending times.
	 * The cleanest solution is to not use the ccid2s_sent field at all
	 * and instead use DCCP timestamps: requires changes in other places.
	 */
351
	ccid2_rtt_estimator(sk, ccid2_time_stamp - seqp->ccid2s_sent);
352 353 354 355
}

static void ccid2_congestion_event(struct sock *sk, struct ccid2_seq *seqp)
{
356
	struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
357

358
	if ((s32)(seqp->ccid2s_sent - hc->tx_last_cong) < 0) {
359 360
		ccid2_pr_debug("Multiple losses in an RTT---treating as one\n");
		return;
361
	}
362

363
	hc->tx_last_cong = ccid2_time_stamp;
364

365 366
	hc->tx_cwnd      = hc->tx_cwnd / 2 ? : 1U;
	hc->tx_ssthresh  = max(hc->tx_cwnd, 2U);
367 368

	/* Avoid spurious timeouts resulting from Ack Ratio > cwnd */
369 370
	if (dccp_sk(sk)->dccps_l_ack_ratio > hc->tx_cwnd)
		ccid2_change_l_ack_ratio(sk, hc->tx_cwnd);
371 372
}

373 374 375 376 377 378 379 380 381 382 383 384 385 386
static int ccid2_hc_tx_parse_options(struct sock *sk, u8 packet_type,
				     u8 option, u8 *optval, u8 optlen)
{
	struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);

	switch (option) {
	case DCCPO_ACK_VECTOR_0:
	case DCCPO_ACK_VECTOR_1:
		return dccp_ackvec_parsed_add(&hc->tx_av_chunks, optval, optlen,
					      option - DCCPO_ACK_VECTOR_0);
	}
	return 0;
}

387 388 389
static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
{
	struct dccp_sock *dp = dccp_sk(sk);
390
	struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
G
Gerrit Renker 已提交
391
	const bool sender_was_blocked = ccid2_cwnd_network_limited(hc);
392
	struct dccp_ackvec_parsed *avp;
393 394 395 396 397 398 399 400 401 402 403 404 405
	u64 ackno, seqno;
	struct ccid2_seq *seqp;
	int done = 0;
	unsigned int maxincr = 0;

	/* check reverse path congestion */
	seqno = DCCP_SKB_CB(skb)->dccpd_seq;

	/* XXX this whole "algorithm" is broken.  Need to fix it to keep track
	 * of the seqnos of the dupacks so that rpseq and rpdupack are correct
	 * -sorbo.
	 */
	/* need to bootstrap */
406 407 408
	if (hc->tx_rpdupack == -1) {
		hc->tx_rpdupack = 0;
		hc->tx_rpseq    = seqno;
409
	} else {
410
		/* check if packet is consecutive */
411 412
		if (dccp_delta_seqno(hc->tx_rpseq, seqno) == 1)
			hc->tx_rpseq = seqno;
413
		/* it's a later packet */
414 415
		else if (after48(seqno, hc->tx_rpseq)) {
			hc->tx_rpdupack++;
416 417

			/* check if we got enough dupacks */
418 419 420
			if (hc->tx_rpdupack >= NUMDUPACK) {
				hc->tx_rpdupack = -1; /* XXX lame */
				hc->tx_rpseq    = 0;
421

422
				ccid2_change_l_ack_ratio(sk, 2 * dp->dccps_l_ack_ratio);
423 424 425 426 427
			}
		}
	}

	/* check forward path congestion */
428
	if (dccp_packet_without_ack(skb))
429 430
		return;

431 432 433
	/* still didn't send out new data packets */
	if (hc->tx_seqh == hc->tx_seqt)
		goto done;
434 435

	ackno = DCCP_SKB_CB(skb)->dccpd_ack_seq;
436 437
	if (after48(ackno, hc->tx_high_ack))
		hc->tx_high_ack = ackno;
A
Andrea Bittau 已提交
438

439
	seqp = hc->tx_seqt;
A
Andrea Bittau 已提交
440 441
	while (before48(seqp->ccid2s_seq, ackno)) {
		seqp = seqp->ccid2s_next;
442 443
		if (seqp == hc->tx_seqh) {
			seqp = hc->tx_seqh->ccid2s_prev;
A
Andrea Bittau 已提交
444 445 446
			break;
		}
	}
447

448 449 450 451
	/*
	 * In slow-start, cwnd can increase up to a maximum of Ack Ratio/2
	 * packets per acknowledgement. Rounding up avoids that cwnd is not
	 * advanced when Ack Ratio is 1 and gives a slight edge otherwise.
452
	 */
453
	if (hc->tx_cwnd < hc->tx_ssthresh)
454
		maxincr = DIV_ROUND_UP(dp->dccps_l_ack_ratio, 2);
455 456

	/* go through all ack vectors */
457
	list_for_each_entry(avp, &hc->tx_av_chunks, node) {
458
		/* go through this ack vector */
459 460 461
		for (; avp->len--; avp->vec++) {
			u64 ackno_end_rl = SUB48(ackno,
						 dccp_ackvec_runlen(avp->vec));
462

463
			ccid2_pr_debug("ackvec %llu |%u,%u|\n",
R
Randy Dunlap 已提交
464
				       (unsigned long long)ackno,
465 466
				       dccp_ackvec_state(avp->vec) >> 6,
				       dccp_ackvec_runlen(avp->vec));
467 468 469 470 471
			/* if the seqno we are analyzing is larger than the
			 * current ackno, then move towards the tail of our
			 * seqnos.
			 */
			while (after48(seqp->ccid2s_seq, ackno)) {
472
				if (seqp == hc->tx_seqt) {
473 474 475 476 477 478 479 480 481 482 483 484
					done = 1;
					break;
				}
				seqp = seqp->ccid2s_prev;
			}
			if (done)
				break;

			/* check all seqnos in the range of the vector
			 * run length
			 */
			while (between48(seqp->ccid2s_seq,ackno_end_rl,ackno)) {
485
				const u8 state = dccp_ackvec_state(avp->vec);
486 487

				/* new packet received or marked */
488
				if (state != DCCPAV_NOT_RECEIVED &&
489
				    !seqp->ccid2s_acked) {
490
					if (state == DCCPAV_ECN_MARKED)
491
						ccid2_congestion_event(sk,
492
								       seqp);
493
					else
494 495 496 497 498
						ccid2_new_ack(sk, seqp,
							      &maxincr);

					seqp->ccid2s_acked = 1;
					ccid2_pr_debug("Got ack for %llu\n",
R
Randy Dunlap 已提交
499
						       (unsigned long long)seqp->ccid2s_seq);
500
					hc->tx_pipe--;
501
				}
502
				if (seqp == hc->tx_seqt) {
503 504 505
					done = 1;
					break;
				}
506
				seqp = seqp->ccid2s_prev;
507 508 509 510
			}
			if (done)
				break;

511
			ackno = SUB48(ackno_end_rl, 1);
512 513 514 515 516 517 518 519
		}
		if (done)
			break;
	}

	/* The state about what is acked should be correct now
	 * Check for NUMDUPACK
	 */
520 521
	seqp = hc->tx_seqt;
	while (before48(seqp->ccid2s_seq, hc->tx_high_ack)) {
A
Andrea Bittau 已提交
522
		seqp = seqp->ccid2s_next;
523 524
		if (seqp == hc->tx_seqh) {
			seqp = hc->tx_seqh->ccid2s_prev;
A
Andrea Bittau 已提交
525 526 527
			break;
		}
	}
528 529 530 531
	done = 0;
	while (1) {
		if (seqp->ccid2s_acked) {
			done++;
532
			if (done == NUMDUPACK)
533 534
				break;
		}
535
		if (seqp == hc->tx_seqt)
536 537 538 539 540 541 542
			break;
		seqp = seqp->ccid2s_prev;
	}

	/* If there are at least 3 acknowledgements, anything unacknowledged
	 * below the last sequence number is considered lost
	 */
543
	if (done == NUMDUPACK) {
544 545 546 547 548
		struct ccid2_seq *last_acked = seqp;

		/* check for lost packets */
		while (1) {
			if (!seqp->ccid2s_acked) {
549
				ccid2_pr_debug("Packet lost: %llu\n",
R
Randy Dunlap 已提交
550
					       (unsigned long long)seqp->ccid2s_seq);
551 552 553 554
				/* XXX need to traverse from tail -> head in
				 * order to detect multiple congestion events in
				 * one ack vector.
				 */
555
				ccid2_congestion_event(sk, seqp);
556
				hc->tx_pipe--;
557
			}
558
			if (seqp == hc->tx_seqt)
559 560 561 562
				break;
			seqp = seqp->ccid2s_prev;
		}

563
		hc->tx_seqt = last_acked;
564 565 566
	}

	/* trim acked packets in tail */
567 568
	while (hc->tx_seqt != hc->tx_seqh) {
		if (!hc->tx_seqt->ccid2s_acked)
569 570
			break;

571
		hc->tx_seqt = hc->tx_seqt->ccid2s_next;
572
	}
573 574 575 576 577 578

	/* restart RTO timer if not all outstanding data has been acked */
	if (hc->tx_pipe == 0)
		sk_stop_timer(sk, &hc->tx_rtotimer);
	else
		sk_reset_timer(sk, &hc->tx_rtotimer, jiffies + hc->tx_rto);
579
done:
G
Gerrit Renker 已提交
580 581 582
	/* check if incoming Acks allow pending packets to be sent */
	if (sender_was_blocked && !ccid2_cwnd_network_limited(hc))
		tasklet_schedule(&dccp_sk(sk)->dccps_xmitlet);
583
	dccp_ackvec_parsed_cleanup(&hc->tx_av_chunks);
584 585
}

586
static int ccid2_hc_tx_init(struct ccid *ccid, struct sock *sk)
587
{
588
	struct ccid2_hc_tx_sock *hc = ccid_priv(ccid);
589 590
	struct dccp_sock *dp = dccp_sk(sk);
	u32 max_ratio;
591

592
	/* RFC 4341, 5: initialise ssthresh to arbitrarily high (max) value */
593
	hc->tx_ssthresh = ~0U;
594

595 596
	/* Use larger initial windows (RFC 4341, section 5). */
	hc->tx_cwnd = rfc3390_bytes_to_packets(dp->dccps_mss_cache);
597 598

	/* Make sure that Ack Ratio is enabled and within bounds. */
599
	max_ratio = DIV_ROUND_UP(hc->tx_cwnd, 2);
600 601 602
	if (dp->dccps_l_ack_ratio == 0 || dp->dccps_l_ack_ratio > max_ratio)
		dp->dccps_l_ack_ratio = max_ratio;

603
	/* XXX init ~ to window size... */
604
	if (ccid2_hc_tx_alloc_seq(hc))
605
		return -ENOMEM;
606

607
	hc->tx_rto	 = DCCP_TIMEOUT_INIT;
608
	hc->tx_rpdupack  = -1;
609
	hc->tx_last_cong = ccid2_time_stamp;
610
	setup_timer(&hc->tx_rtotimer, ccid2_hc_tx_rto_expire,
611
			(unsigned long)sk);
612
	INIT_LIST_HEAD(&hc->tx_av_chunks);
613 614 615 616 617
	return 0;
}

static void ccid2_hc_tx_exit(struct sock *sk)
{
618
	struct ccid2_hc_tx_sock *hc = ccid2_hc_tx_sk(sk);
619
	int i;
620

621
	sk_stop_timer(sk, &hc->tx_rtotimer);
622

623 624 625
	for (i = 0; i < hc->tx_seqbufc; i++)
		kfree(hc->tx_seqbuf[i]);
	hc->tx_seqbufc = 0;
626 627 628 629 630
}

static void ccid2_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb)
{
	const struct dccp_sock *dp = dccp_sk(sk);
631
	struct ccid2_hc_rx_sock *hc = ccid2_hc_rx_sk(sk);
632 633 634 635

	switch (DCCP_SKB_CB(skb)->dccpd_type) {
	case DCCP_PKT_DATA:
	case DCCP_PKT_DATAACK:
636 637
		hc->rx_data++;
		if (hc->rx_data >= dp->dccps_r_ack_ratio) {
638
			dccp_send_ack(sk);
639
			hc->rx_data = 0;
640 641 642 643 644
		}
		break;
	}
}

645
struct ccid_operations ccid2_ops = {
646 647 648 649 650 651 652 653 654 655 656
	.ccid_id		  = DCCPC_CCID2,
	.ccid_name		  = "TCP-like",
	.ccid_hc_tx_obj_size	  = sizeof(struct ccid2_hc_tx_sock),
	.ccid_hc_tx_init	  = ccid2_hc_tx_init,
	.ccid_hc_tx_exit	  = ccid2_hc_tx_exit,
	.ccid_hc_tx_send_packet	  = ccid2_hc_tx_send_packet,
	.ccid_hc_tx_packet_sent	  = ccid2_hc_tx_packet_sent,
	.ccid_hc_tx_parse_options = ccid2_hc_tx_parse_options,
	.ccid_hc_tx_packet_recv	  = ccid2_hc_tx_packet_recv,
	.ccid_hc_rx_obj_size	  = sizeof(struct ccid2_hc_rx_sock),
	.ccid_hc_rx_packet_recv	  = ccid2_hc_rx_packet_recv,
657 658
};

659
#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
660
module_param(ccid2_debug, bool, 0644);
661
MODULE_PARM_DESC(ccid2_debug, "Enable CCID-2 debug messages");
662
#endif