tcp_input.c 167.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7
/*
 * INET		An implementation of the TCP/IP protocol suite for the LINUX
 *		operating system.  INET is implemented using the  BSD Socket
 *		interface as the means of communication with the user level.
 *
 *		Implementation of the Transmission Control Protocol(TCP).
 *
8
 * Authors:	Ross Biro
L
Linus Torvalds 已提交
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42
 *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 *		Mark Evans, <evansmp@uhura.aston.ac.uk>
 *		Corey Minyard <wf-rch!minyard@relay.EU.net>
 *		Florian La Roche, <flla@stud.uni-sb.de>
 *		Charles Hedrick, <hedrick@klinzhai.rutgers.edu>
 *		Linus Torvalds, <torvalds@cs.helsinki.fi>
 *		Alan Cox, <gw4pts@gw4pts.ampr.org>
 *		Matthew Dillon, <dillon@apollo.west.oic.com>
 *		Arnt Gulbrandsen, <agulbra@nvg.unit.no>
 *		Jorge Cwik, <jorge@laser.satlink.net>
 */

/*
 * Changes:
 *		Pedro Roque	:	Fast Retransmit/Recovery.
 *					Two receive queues.
 *					Retransmit queue handled by TCP.
 *					Better retransmit timer handling.
 *					New congestion avoidance.
 *					Header prediction.
 *					Variable renaming.
 *
 *		Eric		:	Fast Retransmit.
 *		Randy Scott	:	MSS option defines.
 *		Eric Schenk	:	Fixes to slow start algorithm.
 *		Eric Schenk	:	Yet another double ACK bug.
 *		Eric Schenk	:	Delayed ACK bug fixes.
 *		Eric Schenk	:	Floyd style fast retrans war avoidance.
 *		David S. Miller	:	Don't allow zero congestion window.
 *		Eric Schenk	:	Fix retransmitter so that it sends
 *					next packet on ack of previous packet.
 *		Andi Kleen	:	Moved open_request checking here
 *					and process RSTs for open_requests.
 *		Andi Kleen	:	Better prune_queue, and other fixes.
S
Stephen Hemminger 已提交
43
 *		Andrey Savochkin:	Fix RTT measurements in the presence of
L
Linus Torvalds 已提交
44 45 46 47 48 49 50
 *					timestamps.
 *		Andrey Savochkin:	Check sequence numbers correctly when
 *					removing SACKs due to in sequence incoming
 *					data segments.
 *		Andi Kleen:		Make sure we never ack data there is not
 *					enough room for. Also make this condition
 *					a fatal error if it might still happen.
51
 *		Andi Kleen:		Add tcp_measure_rcv_mss to make
L
Linus Torvalds 已提交
52
 *					connections with MSS<min(MTU,ann. MSS)
53
 *					work without delayed acks.
L
Linus Torvalds 已提交
54 55 56 57 58 59 60 61 62 63 64
 *		Andi Kleen:		Process packets with PSH set in the
 *					fast path.
 *		J Hadi Salim:		ECN support
 *	 	Andrei Gurtov,
 *		Pasi Sarolahti,
 *		Panu Kuhlberg:		Experimental audit of TCP (re)transmission
 *					engine. Lots of bugs are found.
 *		Pasi Sarolahti:		F-RTO for dealing with spurious RTOs
 */

#include <linux/mm.h>
65
#include <linux/slab.h>
L
Linus Torvalds 已提交
66 67
#include <linux/module.h>
#include <linux/sysctl.h>
68
#include <linux/kernel.h>
69
#include <net/dst.h>
L
Linus Torvalds 已提交
70 71 72 73
#include <net/tcp.h>
#include <net/inet_common.h>
#include <linux/ipsec.h>
#include <asm/unaligned.h>
C
Chris Leech 已提交
74
#include <net/netdma.h>
L
Linus Torvalds 已提交
75

76 77 78 79 80
int sysctl_tcp_timestamps __read_mostly = 1;
int sysctl_tcp_window_scaling __read_mostly = 1;
int sysctl_tcp_sack __read_mostly = 1;
int sysctl_tcp_fack __read_mostly = 1;
int sysctl_tcp_reordering __read_mostly = TCP_FASTRETRANS_THRESH;
81
int sysctl_tcp_ecn __read_mostly = 2;
82 83 84
int sysctl_tcp_dsack __read_mostly = 1;
int sysctl_tcp_app_win __read_mostly = 31;
int sysctl_tcp_adv_win_scale __read_mostly = 2;
L
Linus Torvalds 已提交
85

86 87 88
int sysctl_tcp_stdurg __read_mostly;
int sysctl_tcp_rfc1337 __read_mostly;
int sysctl_tcp_max_orphans __read_mostly = NR_FILE;
89
int sysctl_tcp_frto __read_mostly = 2;
90
int sysctl_tcp_frto_response __read_mostly;
91
int sysctl_tcp_nometrics_save __read_mostly;
L
Linus Torvalds 已提交
92

A
Andreas Petlund 已提交
93 94
int sysctl_tcp_thin_dupack __read_mostly;

95 96
int sysctl_tcp_moderate_rcvbuf __read_mostly = 1;
int sysctl_tcp_abc __read_mostly;
L
Linus Torvalds 已提交
97 98 99 100 101 102 103 104 105 106

#define FLAG_DATA		0x01 /* Incoming frame contained data.		*/
#define FLAG_WIN_UPDATE		0x02 /* Incoming ACK was a window update.	*/
#define FLAG_DATA_ACKED		0x04 /* This ACK acknowledged new data.		*/
#define FLAG_RETRANS_DATA_ACKED	0x08 /* "" "" some of which was retransmitted.	*/
#define FLAG_SYN_ACKED		0x10 /* This ACK acknowledged SYN.		*/
#define FLAG_DATA_SACKED	0x20 /* New SACK.				*/
#define FLAG_ECE		0x40 /* ECE in this ACK				*/
#define FLAG_DATA_LOST		0x80 /* SACK detected data lossage.		*/
#define FLAG_SLOWPATH		0x100 /* Do not skip RFC checks for window update.*/
I
Ilpo Järvinen 已提交
107
#define FLAG_ONLY_ORIG_SACKED	0x200 /* SACKs only non-rexmit sent before RTO */
108
#define FLAG_SND_UNA_ADVANCED	0x400 /* Snd_una was changed (!= FLAG_DATA_ACKED) */
109
#define FLAG_DSACKING_ACK	0x800 /* SACK blocks contained D-SACK info */
110
#define FLAG_NONHEAD_RETRANS_ACKED	0x1000 /* Non-head rexmitted data was ACKed */
111
#define FLAG_SACK_RENEGING	0x2000 /* snd_una advanced to a sacked seq */
L
Linus Torvalds 已提交
112 113 114 115 116

#define FLAG_ACKED		(FLAG_DATA_ACKED|FLAG_SYN_ACKED)
#define FLAG_NOT_DUP		(FLAG_DATA|FLAG_WIN_UPDATE|FLAG_ACKED)
#define FLAG_CA_ALERT		(FLAG_DATA_SACKED|FLAG_ECE)
#define FLAG_FORWARD_PROGRESS	(FLAG_ACKED|FLAG_DATA_SACKED)
117
#define FLAG_ANY_PROGRESS	(FLAG_FORWARD_PROGRESS|FLAG_SND_UNA_ADVANCED)
L
Linus Torvalds 已提交
118 119

#define TCP_REMNANT (TCP_FLAG_FIN|TCP_FLAG_URG|TCP_FLAG_SYN|TCP_FLAG_PSH)
120
#define TCP_HP_BITS (~(TCP_RESERVED_BITS|TCP_FLAG_PSH))
L
Linus Torvalds 已提交
121

122
/* Adapt the MSS value used to make delayed ack decision to the
L
Linus Torvalds 已提交
123
 * real world.
124
 */
125
static void tcp_measure_rcv_mss(struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
126
{
127
	struct inet_connection_sock *icsk = inet_csk(sk);
128
	const unsigned int lss = icsk->icsk_ack.last_seg_size;
129
	unsigned int len;
L
Linus Torvalds 已提交
130

131
	icsk->icsk_ack.last_seg_size = 0;
L
Linus Torvalds 已提交
132 133 134 135

	/* skb->len may jitter because of SACKs, even if peer
	 * sends good full-sized frames.
	 */
136
	len = skb_shinfo(skb)->gso_size ? : skb->len;
137 138
	if (len >= icsk->icsk_ack.rcv_mss) {
		icsk->icsk_ack.rcv_mss = len;
L
Linus Torvalds 已提交
139 140 141 142 143 144
	} else {
		/* Otherwise, we make more careful check taking into account,
		 * that SACKs block is variable.
		 *
		 * "len" is invariant segment length, including TCP header.
		 */
145
		len += skb->data - skb_transport_header(skb);
146
		if (len >= TCP_MSS_DEFAULT + sizeof(struct tcphdr) ||
L
Linus Torvalds 已提交
147 148 149 150 151 152
		    /* If PSH is not set, packet should be
		     * full sized, provided peer TCP is not badly broken.
		     * This observation (if it is correct 8)) allows
		     * to handle super-low mtu links fairly.
		     */
		    (len >= TCP_MIN_MSS + sizeof(struct tcphdr) &&
153
		     !(tcp_flag_word(tcp_hdr(skb)) & TCP_REMNANT))) {
L
Linus Torvalds 已提交
154 155 156 157
			/* Subtract also invariant (if peer is RFC compliant),
			 * tcp header plus fixed timestamp option length.
			 * Resulting "len" is MSS free of SACK jitter.
			 */
158 159
			len -= tcp_sk(sk)->tcp_header_len;
			icsk->icsk_ack.last_seg_size = len;
L
Linus Torvalds 已提交
160
			if (len == lss) {
161
				icsk->icsk_ack.rcv_mss = len;
L
Linus Torvalds 已提交
162 163 164
				return;
			}
		}
165 166
		if (icsk->icsk_ack.pending & ICSK_ACK_PUSHED)
			icsk->icsk_ack.pending |= ICSK_ACK_PUSHED2;
167
		icsk->icsk_ack.pending |= ICSK_ACK_PUSHED;
L
Linus Torvalds 已提交
168 169 170
	}
}

171
static void tcp_incr_quickack(struct sock *sk)
L
Linus Torvalds 已提交
172
{
173 174
	struct inet_connection_sock *icsk = inet_csk(sk);
	unsigned quickacks = tcp_sk(sk)->rcv_wnd / (2 * icsk->icsk_ack.rcv_mss);
L
Linus Torvalds 已提交
175

176 177
	if (quickacks == 0)
		quickacks = 2;
178 179
	if (quickacks > icsk->icsk_ack.quick)
		icsk->icsk_ack.quick = min(quickacks, TCP_MAX_QUICKACKS);
L
Linus Torvalds 已提交
180 181
}

182
void tcp_enter_quickack_mode(struct sock *sk)
L
Linus Torvalds 已提交
183
{
184 185 186 187
	struct inet_connection_sock *icsk = inet_csk(sk);
	tcp_incr_quickack(sk);
	icsk->icsk_ack.pingpong = 0;
	icsk->icsk_ack.ato = TCP_ATO_MIN;
L
Linus Torvalds 已提交
188 189 190 191 192 193
}

/* Send ACKs quickly, if "quick" count is not exhausted
 * and the session is not interactive.
 */

194
static inline int tcp_in_quickack_mode(const struct sock *sk)
L
Linus Torvalds 已提交
195
{
196 197
	const struct inet_connection_sock *icsk = inet_csk(sk);
	return icsk->icsk_ack.quick && !icsk->icsk_ack.pingpong;
L
Linus Torvalds 已提交
198 199
}

200 201
static inline void TCP_ECN_queue_cwr(struct tcp_sock *tp)
{
202
	if (tp->ecn_flags & TCP_ECN_OK)
203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218
		tp->ecn_flags |= TCP_ECN_QUEUE_CWR;
}

static inline void TCP_ECN_accept_cwr(struct tcp_sock *tp, struct sk_buff *skb)
{
	if (tcp_hdr(skb)->cwr)
		tp->ecn_flags &= ~TCP_ECN_DEMAND_CWR;
}

static inline void TCP_ECN_withdraw_cwr(struct tcp_sock *tp)
{
	tp->ecn_flags &= ~TCP_ECN_DEMAND_CWR;
}

static inline void TCP_ECN_check_ce(struct tcp_sock *tp, struct sk_buff *skb)
{
219
	if (tp->ecn_flags & TCP_ECN_OK) {
220 221 222 223 224 225 226 227 228 229 230 231
		if (INET_ECN_is_ce(TCP_SKB_CB(skb)->flags))
			tp->ecn_flags |= TCP_ECN_DEMAND_CWR;
		/* Funny extension: if ECT is not set on a segment,
		 * it is surely retransmit. It is not in ECN RFC,
		 * but Linux follows this rule. */
		else if (INET_ECN_is_not_ect((TCP_SKB_CB(skb)->flags)))
			tcp_enter_quickack_mode((struct sock *)tp);
	}
}

static inline void TCP_ECN_rcv_synack(struct tcp_sock *tp, struct tcphdr *th)
{
232
	if ((tp->ecn_flags & TCP_ECN_OK) && (!th->ece || th->cwr))
233 234 235 236 237
		tp->ecn_flags &= ~TCP_ECN_OK;
}

static inline void TCP_ECN_rcv_syn(struct tcp_sock *tp, struct tcphdr *th)
{
238
	if ((tp->ecn_flags & TCP_ECN_OK) && (!th->ece || !th->cwr))
239 240 241 242 243
		tp->ecn_flags &= ~TCP_ECN_OK;
}

static inline int TCP_ECN_rcv_ecn_echo(struct tcp_sock *tp, struct tcphdr *th)
{
244
	if (th->ece && !th->syn && (tp->ecn_flags & TCP_ECN_OK))
245 246 247 248
		return 1;
	return 0;
}

L
Linus Torvalds 已提交
249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283
/* Buffer size and advertised window tuning.
 *
 * 1. Tuning sk->sk_sndbuf, when connection enters established state.
 */

static void tcp_fixup_sndbuf(struct sock *sk)
{
	int sndmem = tcp_sk(sk)->rx_opt.mss_clamp + MAX_TCP_HEADER + 16 +
		     sizeof(struct sk_buff);

	if (sk->sk_sndbuf < 3 * sndmem)
		sk->sk_sndbuf = min(3 * sndmem, sysctl_tcp_wmem[2]);
}

/* 2. Tuning advertised window (window_clamp, rcv_ssthresh)
 *
 * All tcp_full_space() is split to two parts: "network" buffer, allocated
 * forward and advertised in receiver window (tp->rcv_wnd) and
 * "application buffer", required to isolate scheduling/application
 * latencies from network.
 * window_clamp is maximal advertised window. It can be less than
 * tcp_full_space(), in this case tcp_full_space() - window_clamp
 * is reserved for "application" buffer. The less window_clamp is
 * the smoother our behaviour from viewpoint of network, but the lower
 * throughput and the higher sensitivity of the connection to losses. 8)
 *
 * rcv_ssthresh is more strict window_clamp used at "slow start"
 * phase to predict further behaviour of this connection.
 * It is used for two goals:
 * - to enforce header prediction at sender, even when application
 *   requires some significant "application buffer". It is check #1.
 * - to prevent pruning of receive queue because of misprediction
 *   of receiver window. Check #2.
 *
 * The scheme does not work when sender sends good segments opening
S
Stephen Hemminger 已提交
284
 * window and then starts to feed us spaghetti. But it should work
L
Linus Torvalds 已提交
285 286 287 288
 * in common situations. Otherwise, we have to rely on queue collapsing.
 */

/* Slow part of check#2. */
289
static int __tcp_grow_window(const struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
290
{
291
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
292
	/* Optimize this! */
293 294
	int truesize = tcp_win_from_space(skb->truesize) >> 1;
	int window = tcp_win_from_space(sysctl_tcp_rmem[2]) >> 1;
L
Linus Torvalds 已提交
295 296 297

	while (tp->rcv_ssthresh <= window) {
		if (truesize <= skb->len)
298
			return 2 * inet_csk(sk)->icsk_ack.rcv_mss;
L
Linus Torvalds 已提交
299 300 301 302 303 304 305

		truesize >>= 1;
		window >>= 1;
	}
	return 0;
}

306
static void tcp_grow_window(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
307
{
308 309
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
310 311 312 313 314 315 316 317 318 319
	/* Check #1 */
	if (tp->rcv_ssthresh < tp->window_clamp &&
	    (int)tp->rcv_ssthresh < tcp_space(sk) &&
	    !tcp_memory_pressure) {
		int incr;

		/* Check #2. Increase window, if skb with such overhead
		 * will fit to rcvbuf in future.
		 */
		if (tcp_win_from_space(skb->truesize) <= skb->len)
320
			incr = 2 * tp->advmss;
L
Linus Torvalds 已提交
321
		else
322
			incr = __tcp_grow_window(sk, skb);
L
Linus Torvalds 已提交
323 324

		if (incr) {
325 326
			tp->rcv_ssthresh = min(tp->rcv_ssthresh + incr,
					       tp->window_clamp);
327
			inet_csk(sk)->icsk_ack.quick |= 1;
L
Linus Torvalds 已提交
328 329 330 331 332 333 334 335 336 337 338 339
		}
	}
}

/* 3. Tuning rcvbuf, when connection enters established state. */

static void tcp_fixup_rcvbuf(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	int rcvmem = tp->advmss + MAX_TCP_HEADER + 16 + sizeof(struct sk_buff);

	/* Try to select rcvbuf so that 4 mss-sized segments
S
Stephen Hemminger 已提交
340
	 * will fit to window and corresponding skbs will fit to our rcvbuf.
L
Linus Torvalds 已提交
341 342 343 344 345 346 347 348
	 * (was 3; 4 is minimum to allow fast retransmit to work.)
	 */
	while (tcp_win_from_space(rcvmem) < tp->advmss)
		rcvmem += 128;
	if (sk->sk_rcvbuf < 4 * rcvmem)
		sk->sk_rcvbuf = min(4 * rcvmem, sysctl_tcp_rmem[2]);
}

S
Stephen Hemminger 已提交
349
/* 4. Try to fixup all. It is made immediately after connection enters
L
Linus Torvalds 已提交
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385
 *    established state.
 */
static void tcp_init_buffer_space(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	int maxwin;

	if (!(sk->sk_userlocks & SOCK_RCVBUF_LOCK))
		tcp_fixup_rcvbuf(sk);
	if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK))
		tcp_fixup_sndbuf(sk);

	tp->rcvq_space.space = tp->rcv_wnd;

	maxwin = tcp_full_space(sk);

	if (tp->window_clamp >= maxwin) {
		tp->window_clamp = maxwin;

		if (sysctl_tcp_app_win && maxwin > 4 * tp->advmss)
			tp->window_clamp = max(maxwin -
					       (maxwin >> sysctl_tcp_app_win),
					       4 * tp->advmss);
	}

	/* Force reservation of one segment. */
	if (sysctl_tcp_app_win &&
	    tp->window_clamp > 2 * tp->advmss &&
	    tp->window_clamp + tp->advmss > maxwin)
		tp->window_clamp = max(2 * tp->advmss, maxwin - tp->advmss);

	tp->rcv_ssthresh = min(tp->rcv_ssthresh, tp->window_clamp);
	tp->snd_cwnd_stamp = tcp_time_stamp;
}

/* 5. Recalculate window clamp after socket hit its memory bounds. */
386
static void tcp_clamp_window(struct sock *sk)
L
Linus Torvalds 已提交
387
{
388
	struct tcp_sock *tp = tcp_sk(sk);
389
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
390

391
	icsk->icsk_ack.quick = 0;
L
Linus Torvalds 已提交
392

393 394 395 396 397 398
	if (sk->sk_rcvbuf < sysctl_tcp_rmem[2] &&
	    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) &&
	    !tcp_memory_pressure &&
	    atomic_read(&tcp_memory_allocated) < sysctl_tcp_mem[0]) {
		sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc),
				    sysctl_tcp_rmem[2]);
L
Linus Torvalds 已提交
399
	}
400
	if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf)
401
		tp->rcv_ssthresh = min(tp->window_clamp, 2U * tp->advmss);
L
Linus Torvalds 已提交
402 403
}

S
Stephen Hemminger 已提交
404 405 406 407 408 409 410 411 412 413 414 415
/* Initialize RCV_MSS value.
 * RCV_MSS is an our guess about MSS used by the peer.
 * We haven't any direct information about the MSS.
 * It's better to underestimate the RCV_MSS rather than overestimate.
 * Overestimations make us ACKing less frequently than needed.
 * Underestimations are more easy to detect and fix by tcp_measure_rcv_mss().
 */
void tcp_initialize_rcv_mss(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	unsigned int hint = min_t(unsigned int, tp->advmss, tp->mss_cache);

416
	hint = min(hint, tp->rcv_wnd / 2);
417
	hint = min(hint, TCP_MSS_DEFAULT);
S
Stephen Hemminger 已提交
418 419 420 421 422
	hint = max(hint, TCP_MIN_MSS);

	inet_csk(sk)->icsk_ack.rcv_mss = hint;
}

L
Linus Torvalds 已提交
423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448
/* Receiver "autotuning" code.
 *
 * The algorithm for RTT estimation w/o timestamps is based on
 * Dynamic Right-Sizing (DRS) by Wu Feng and Mike Fisk of LANL.
 * <http://www.lanl.gov/radiant/website/pubs/drs/lacsi2001.ps>
 *
 * More detail on this code can be found at
 * <http://www.psc.edu/~jheffner/senior_thesis.ps>,
 * though this reference is out of date.  A new paper
 * is pending.
 */
static void tcp_rcv_rtt_update(struct tcp_sock *tp, u32 sample, int win_dep)
{
	u32 new_sample = tp->rcv_rtt_est.rtt;
	long m = sample;

	if (m == 0)
		m = 1;

	if (new_sample != 0) {
		/* If we sample in larger samples in the non-timestamp
		 * case, we could grossly overestimate the RTT especially
		 * with chatty applications or bulk transfer apps which
		 * are stalled on filesystem I/O.
		 *
		 * Also, since we are only going for a minimum in the
S
Stephen Hemminger 已提交
449
		 * non-timestamp case, we do not smooth things out
S
Stephen Hemminger 已提交
450
		 * else with timestamps disabled convergence takes too
L
Linus Torvalds 已提交
451 452 453 454 455 456 457 458
		 * long.
		 */
		if (!win_dep) {
			m -= (new_sample >> 3);
			new_sample += m;
		} else if (m < new_sample)
			new_sample = m << 3;
	} else {
S
Stephen Hemminger 已提交
459
		/* No previous measure. */
L
Linus Torvalds 已提交
460 461 462 463 464 465 466 467 468 469 470 471 472
		new_sample = m << 3;
	}

	if (tp->rcv_rtt_est.rtt != new_sample)
		tp->rcv_rtt_est.rtt = new_sample;
}

static inline void tcp_rcv_rtt_measure(struct tcp_sock *tp)
{
	if (tp->rcv_rtt_est.time == 0)
		goto new_measure;
	if (before(tp->rcv_nxt, tp->rcv_rtt_est.seq))
		return;
473
	tcp_rcv_rtt_update(tp, jiffies - tp->rcv_rtt_est.time, 1);
L
Linus Torvalds 已提交
474 475 476 477 478 479

new_measure:
	tp->rcv_rtt_est.seq = tp->rcv_nxt + tp->rcv_wnd;
	tp->rcv_rtt_est.time = tcp_time_stamp;
}

480 481
static inline void tcp_rcv_rtt_measure_ts(struct sock *sk,
					  const struct sk_buff *skb)
L
Linus Torvalds 已提交
482
{
483
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
484 485
	if (tp->rx_opt.rcv_tsecr &&
	    (TCP_SKB_CB(skb)->end_seq -
486
	     TCP_SKB_CB(skb)->seq >= inet_csk(sk)->icsk_ack.rcv_mss))
L
Linus Torvalds 已提交
487 488 489 490 491 492 493 494 495 496 497 498
		tcp_rcv_rtt_update(tp, tcp_time_stamp - tp->rx_opt.rcv_tsecr, 0);
}

/*
 * This function should be called every time data is copied to user space.
 * It calculates the appropriate TCP receive buffer space.
 */
void tcp_rcv_space_adjust(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	int time;
	int space;
499

L
Linus Torvalds 已提交
500 501
	if (tp->rcvq_space.time == 0)
		goto new_measure;
502

L
Linus Torvalds 已提交
503
	time = tcp_time_stamp - tp->rcvq_space.time;
504
	if (time < (tp->rcv_rtt_est.rtt >> 3) || tp->rcv_rtt_est.rtt == 0)
L
Linus Torvalds 已提交
505
		return;
506

L
Linus Torvalds 已提交
507 508 509 510 511 512 513 514 515
	space = 2 * (tp->copied_seq - tp->rcvq_space.seq);

	space = max(tp->rcvq_space.space, space);

	if (tp->rcvq_space.space != space) {
		int rcvmem;

		tp->rcvq_space.space = space;

516 517
		if (sysctl_tcp_moderate_rcvbuf &&
		    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK)) {
L
Linus Torvalds 已提交
518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540
			int new_clamp = space;

			/* Receive space grows, normalize in order to
			 * take into account packet headers and sk_buff
			 * structure overhead.
			 */
			space /= tp->advmss;
			if (!space)
				space = 1;
			rcvmem = (tp->advmss + MAX_TCP_HEADER +
				  16 + sizeof(struct sk_buff));
			while (tcp_win_from_space(rcvmem) < tp->advmss)
				rcvmem += 128;
			space *= rcvmem;
			space = min(space, sysctl_tcp_rmem[2]);
			if (space > sk->sk_rcvbuf) {
				sk->sk_rcvbuf = space;

				/* Make the window clamp follow along.  */
				tp->window_clamp = new_clamp;
			}
		}
	}
541

L
Linus Torvalds 已提交
542 543 544 545 546 547 548 549 550 551 552 553 554 555 556
new_measure:
	tp->rcvq_space.seq = tp->copied_seq;
	tp->rcvq_space.time = tcp_time_stamp;
}

/* There is something which you must keep in mind when you analyze the
 * behavior of the tp->ato delayed ack timeout interval.  When a
 * connection starts up, we want to ack as quickly as possible.  The
 * problem is that "good" TCP's do slow start at the beginning of data
 * transmission.  The means that until we send the first few ACK's the
 * sender will sit on his end and only queue most of his data, because
 * he can only send snd_cwnd unacked packets at any given time.  For
 * each ACK we send, he increments snd_cwnd and transmits more of his
 * queue.  -DaveM
 */
557
static void tcp_event_data_recv(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
558
{
559
	struct tcp_sock *tp = tcp_sk(sk);
560
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
561 562
	u32 now;

563
	inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
564

565
	tcp_measure_rcv_mss(sk, skb);
L
Linus Torvalds 已提交
566 567

	tcp_rcv_rtt_measure(tp);
568

L
Linus Torvalds 已提交
569 570
	now = tcp_time_stamp;

571
	if (!icsk->icsk_ack.ato) {
L
Linus Torvalds 已提交
572 573 574
		/* The _first_ data packet received, initialize
		 * delayed ACK engine.
		 */
575 576
		tcp_incr_quickack(sk);
		icsk->icsk_ack.ato = TCP_ATO_MIN;
L
Linus Torvalds 已提交
577
	} else {
578
		int m = now - icsk->icsk_ack.lrcvtime;
L
Linus Torvalds 已提交
579

580
		if (m <= TCP_ATO_MIN / 2) {
L
Linus Torvalds 已提交
581
			/* The fastest case is the first. */
582 583 584 585 586 587
			icsk->icsk_ack.ato = (icsk->icsk_ack.ato >> 1) + TCP_ATO_MIN / 2;
		} else if (m < icsk->icsk_ack.ato) {
			icsk->icsk_ack.ato = (icsk->icsk_ack.ato >> 1) + m;
			if (icsk->icsk_ack.ato > icsk->icsk_rto)
				icsk->icsk_ack.ato = icsk->icsk_rto;
		} else if (m > icsk->icsk_rto) {
S
Stephen Hemminger 已提交
588
			/* Too long gap. Apparently sender failed to
L
Linus Torvalds 已提交
589 590
			 * restart window, so that we send ACKs quickly.
			 */
591
			tcp_incr_quickack(sk);
592
			sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
593 594
		}
	}
595
	icsk->icsk_ack.lrcvtime = now;
L
Linus Torvalds 已提交
596 597 598 599

	TCP_ECN_check_ce(tp, skb);

	if (skb->len >= 128)
600
		tcp_grow_window(sk, skb);
L
Linus Torvalds 已提交
601 602 603 604 605 606 607 608 609 610 611
}

/* Called to compute a smoothed rtt estimate. The data fed to this
 * routine either comes from timestamps, or from segments that were
 * known _not_ to have been retransmitted [see Karn/Partridge
 * Proceedings SIGCOMM 87]. The algorithm is from the SIGCOMM 88
 * piece by Van Jacobson.
 * NOTE: the next three routines used to be one big routine.
 * To save cycles in the RFC 1323 implementation it was better to break
 * it up into three procedures. -- erics
 */
612
static void tcp_rtt_estimator(struct sock *sk, const __u32 mrtt)
L
Linus Torvalds 已提交
613
{
614
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
615 616 617 618 619
	long m = mrtt; /* RTT */

	/*	The following amusing code comes from Jacobson's
	 *	article in SIGCOMM '88.  Note that rtt and mdev
	 *	are scaled versions of rtt and mean deviation.
620
	 *	This is designed to be as fast as possible
L
Linus Torvalds 已提交
621 622 623 624 625 626 627
	 *	m stands for "measurement".
	 *
	 *	On a 1990 paper the rto value is changed to:
	 *	RTO = rtt + 4 * mdev
	 *
	 * Funny. This algorithm seems to be very broken.
	 * These formulae increase RTO, when it should be decreased, increase
S
Stephen Hemminger 已提交
628
	 * too slowly, when it should be increased quickly, decrease too quickly
L
Linus Torvalds 已提交
629 630 631 632
	 * etc. I guess in BSD RTO takes ONE value, so that it is absolutely
	 * does not matter how to _calculate_ it. Seems, it was trap
	 * that VJ failed to avoid. 8)
	 */
S
Stephen Hemminger 已提交
633
	if (m == 0)
L
Linus Torvalds 已提交
634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661
		m = 1;
	if (tp->srtt != 0) {
		m -= (tp->srtt >> 3);	/* m is now error in rtt est */
		tp->srtt += m;		/* rtt = 7/8 rtt + 1/8 new */
		if (m < 0) {
			m = -m;		/* m is now abs(error) */
			m -= (tp->mdev >> 2);   /* similar update on mdev */
			/* This is similar to one of Eifel findings.
			 * Eifel blocks mdev updates when rtt decreases.
			 * This solution is a bit different: we use finer gain
			 * for mdev in this case (alpha*beta).
			 * Like Eifel it also prevents growth of rto,
			 * but also it limits too fast rto decreases,
			 * happening in pure Eifel.
			 */
			if (m > 0)
				m >>= 3;
		} else {
			m -= (tp->mdev >> 2);   /* similar update on mdev */
		}
		tp->mdev += m;	    	/* mdev = 3/4 mdev + 1/4 new */
		if (tp->mdev > tp->mdev_max) {
			tp->mdev_max = tp->mdev;
			if (tp->mdev_max > tp->rttvar)
				tp->rttvar = tp->mdev_max;
		}
		if (after(tp->snd_una, tp->rtt_seq)) {
			if (tp->mdev_max < tp->rttvar)
662
				tp->rttvar -= (tp->rttvar - tp->mdev_max) >> 2;
L
Linus Torvalds 已提交
663
			tp->rtt_seq = tp->snd_nxt;
664
			tp->mdev_max = tcp_rto_min(sk);
L
Linus Torvalds 已提交
665 666 667
		}
	} else {
		/* no previous measure. */
668 669
		tp->srtt = m << 3;	/* take the measured time to be rtt */
		tp->mdev = m << 1;	/* make sure rto = 3*rtt */
670
		tp->mdev_max = tp->rttvar = max(tp->mdev, tcp_rto_min(sk));
L
Linus Torvalds 已提交
671 672 673 674 675 676 677
		tp->rtt_seq = tp->snd_nxt;
	}
}

/* Calculate rto without backoff.  This is the second half of Van Jacobson's
 * routine referred to above.
 */
678
static inline void tcp_set_rto(struct sock *sk)
L
Linus Torvalds 已提交
679
{
680
	const struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
681 682 683 684 685 686 687 688
	/* Old crap is replaced with new one. 8)
	 *
	 * More seriously:
	 * 1. If rtt variance happened to be less 50msec, it is hallucination.
	 *    It cannot be less due to utterly erratic ACK generation made
	 *    at least by solaris and freebsd. "Erratic ACKs" has _nothing_
	 *    to do with delayed acks, because at cwnd>2 true delack timeout
	 *    is invisible. Actually, Linux-2.4 also generates erratic
S
Stephen Hemminger 已提交
689
	 *    ACKs in some circumstances.
L
Linus Torvalds 已提交
690
	 */
691
	inet_csk(sk)->icsk_rto = __tcp_set_rto(tp);
L
Linus Torvalds 已提交
692 693 694 695

	/* 2. Fixups made earlier cannot be right.
	 *    If we do not estimate RTO correctly without them,
	 *    all the algo is pure shit and should be replaced
S
Stephen Hemminger 已提交
696
	 *    with correct one. It is exactly, which we pretend to do.
L
Linus Torvalds 已提交
697 698
	 */

699 700 701
	/* NOTE: clamping at TCP_RTO_MIN is not required, current algo
	 * guarantees that rto is higher.
	 */
702
	tcp_bound_rto(sk);
L
Linus Torvalds 已提交
703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718
}

/* Save metrics learned by this TCP session.
   This function is called only, when TCP finishes successfully
   i.e. when it enters TIME-WAIT or goes from LAST-ACK to CLOSE.
 */
void tcp_update_metrics(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct dst_entry *dst = __sk_dst_get(sk);

	if (sysctl_tcp_nometrics_save)
		return;

	dst_confirm(dst);

719
	if (dst && (dst->flags & DST_HOST)) {
720
		const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
721
		int m;
S
Stephen Hemminger 已提交
722
		unsigned long rtt;
L
Linus Torvalds 已提交
723

724
		if (icsk->icsk_backoff || !tp->srtt) {
L
Linus Torvalds 已提交
725 726 727 728 729
			/* This session failed to estimate rtt. Why?
			 * Probably, no packets returned in time.
			 * Reset our results.
			 */
			if (!(dst_metric_locked(dst, RTAX_RTT)))
730
				dst->metrics[RTAX_RTT - 1] = 0;
L
Linus Torvalds 已提交
731 732 733
			return;
		}

S
Stephen Hemminger 已提交
734 735
		rtt = dst_metric_rtt(dst, RTAX_RTT);
		m = rtt - tp->srtt;
L
Linus Torvalds 已提交
736 737 738 739 740 741 742

		/* If newly calculated rtt larger than stored one,
		 * store new one. Otherwise, use EWMA. Remember,
		 * rtt overestimation is always better than underestimation.
		 */
		if (!(dst_metric_locked(dst, RTAX_RTT))) {
			if (m <= 0)
S
Stephen Hemminger 已提交
743
				set_dst_metric_rtt(dst, RTAX_RTT, tp->srtt);
L
Linus Torvalds 已提交
744
			else
S
Stephen Hemminger 已提交
745
				set_dst_metric_rtt(dst, RTAX_RTT, rtt - (m >> 3));
L
Linus Torvalds 已提交
746 747 748
		}

		if (!(dst_metric_locked(dst, RTAX_RTTVAR))) {
S
Stephen Hemminger 已提交
749
			unsigned long var;
L
Linus Torvalds 已提交
750 751 752 753 754 755 756 757
			if (m < 0)
				m = -m;

			/* Scale deviation to rttvar fixed point */
			m >>= 1;
			if (m < tp->mdev)
				m = tp->mdev;

S
Stephen Hemminger 已提交
758 759 760
			var = dst_metric_rtt(dst, RTAX_RTTVAR);
			if (m >= var)
				var = m;
L
Linus Torvalds 已提交
761
			else
S
Stephen Hemminger 已提交
762 763 764
				var -= (var - m) >> 2;

			set_dst_metric_rtt(dst, RTAX_RTTVAR, var);
L
Linus Torvalds 已提交
765 766
		}

I
Ilpo Järvinen 已提交
767
		if (tcp_in_initial_slowstart(tp)) {
L
Linus Torvalds 已提交
768 769 770 771 772 773 774
			/* Slow start still did not finish. */
			if (dst_metric(dst, RTAX_SSTHRESH) &&
			    !dst_metric_locked(dst, RTAX_SSTHRESH) &&
			    (tp->snd_cwnd >> 1) > dst_metric(dst, RTAX_SSTHRESH))
				dst->metrics[RTAX_SSTHRESH-1] = tp->snd_cwnd >> 1;
			if (!dst_metric_locked(dst, RTAX_CWND) &&
			    tp->snd_cwnd > dst_metric(dst, RTAX_CWND))
775
				dst->metrics[RTAX_CWND - 1] = tp->snd_cwnd;
L
Linus Torvalds 已提交
776
		} else if (tp->snd_cwnd > tp->snd_ssthresh &&
777
			   icsk->icsk_ca_state == TCP_CA_Open) {
L
Linus Torvalds 已提交
778 779 780 781 782
			/* Cong. avoidance phase, cwnd is reliable. */
			if (!dst_metric_locked(dst, RTAX_SSTHRESH))
				dst->metrics[RTAX_SSTHRESH-1] =
					max(tp->snd_cwnd >> 1, tp->snd_ssthresh);
			if (!dst_metric_locked(dst, RTAX_CWND))
783
				dst->metrics[RTAX_CWND-1] = (dst_metric(dst, RTAX_CWND) + tp->snd_cwnd) >> 1;
L
Linus Torvalds 已提交
784 785 786 787 788
		} else {
			/* Else slow start did not finish, cwnd is non-sense,
			   ssthresh may be also invalid.
			 */
			if (!dst_metric_locked(dst, RTAX_CWND))
789 790
				dst->metrics[RTAX_CWND-1] = (dst_metric(dst, RTAX_CWND) + tp->snd_ssthresh) >> 1;
			if (dst_metric(dst, RTAX_SSTHRESH) &&
L
Linus Torvalds 已提交
791
			    !dst_metric_locked(dst, RTAX_SSTHRESH) &&
792
			    tp->snd_ssthresh > dst_metric(dst, RTAX_SSTHRESH))
L
Linus Torvalds 已提交
793 794 795 796
				dst->metrics[RTAX_SSTHRESH-1] = tp->snd_ssthresh;
		}

		if (!dst_metric_locked(dst, RTAX_REORDERING)) {
797
			if (dst_metric(dst, RTAX_REORDERING) < tp->reordering &&
L
Linus Torvalds 已提交
798 799 800 801 802 803
			    tp->reordering != sysctl_tcp_reordering)
				dst->metrics[RTAX_REORDERING-1] = tp->reordering;
		}
	}
}

804 805 806 807 808 809 810 811 812
/* Numbers are taken from RFC3390.
 *
 * John Heffner states:
 *
 *	The RFC specifies a window of no more than 4380 bytes
 *	unless 2*MSS > 4380.  Reading the pseudocode in the RFC
 *	is a bit misleading because they use a clamp at 4380 bytes
 *	rather than use a multiplier in the relevant range.
 */
L
Linus Torvalds 已提交
813 814 815 816
__u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst)
{
	__u32 cwnd = (dst ? dst_metric(dst, RTAX_INITCWND) : 0);

817 818 819 820 821 822
	if (!cwnd) {
		if (tp->mss_cache > 1460)
			cwnd = 2;
		else
			cwnd = (tp->mss_cache > 1095) ? 3 : 4;
	}
L
Linus Torvalds 已提交
823 824 825
	return min_t(__u32, cwnd, tp->snd_cwnd_clamp);
}

S
Stephen Hemminger 已提交
826
/* Set slow start threshold and cwnd not falling to slow start */
827
void tcp_enter_cwr(struct sock *sk, const int set_ssthresh)
S
Stephen Hemminger 已提交
828 829
{
	struct tcp_sock *tp = tcp_sk(sk);
830
	const struct inet_connection_sock *icsk = inet_csk(sk);
S
Stephen Hemminger 已提交
831 832 833

	tp->prior_ssthresh = 0;
	tp->bytes_acked = 0;
834
	if (icsk->icsk_ca_state < TCP_CA_CWR) {
S
Stephen Hemminger 已提交
835
		tp->undo_marker = 0;
836 837
		if (set_ssthresh)
			tp->snd_ssthresh = icsk->icsk_ca_ops->ssthresh(sk);
S
Stephen Hemminger 已提交
838 839 840 841 842 843 844 845 846 847 848
		tp->snd_cwnd = min(tp->snd_cwnd,
				   tcp_packets_in_flight(tp) + 1U);
		tp->snd_cwnd_cnt = 0;
		tp->high_seq = tp->snd_nxt;
		tp->snd_cwnd_stamp = tcp_time_stamp;
		TCP_ECN_queue_cwr(tp);

		tcp_set_ca_state(sk, TCP_CA_CWR);
	}
}

849 850 851 852 853 854
/*
 * Packet counting of FACK is based on in-order assumptions, therefore TCP
 * disables it when reordering is detected
 */
static void tcp_disable_fack(struct tcp_sock *tp)
{
855 856 857
	/* RFC3517 uses different metric in lost marker => reset on change */
	if (tcp_is_fack(tp))
		tp->lost_skb_hint = NULL;
858 859 860
	tp->rx_opt.sack_ok &= ~2;
}

861
/* Take a notice that peer is sending D-SACKs */
862 863 864 865 866
static void tcp_dsack_seen(struct tcp_sock *tp)
{
	tp->rx_opt.sack_ok |= 4;
}

L
Linus Torvalds 已提交
867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887
/* Initialize metrics on socket. */

static void tcp_init_metrics(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct dst_entry *dst = __sk_dst_get(sk);

	if (dst == NULL)
		goto reset;

	dst_confirm(dst);

	if (dst_metric_locked(dst, RTAX_CWND))
		tp->snd_cwnd_clamp = dst_metric(dst, RTAX_CWND);
	if (dst_metric(dst, RTAX_SSTHRESH)) {
		tp->snd_ssthresh = dst_metric(dst, RTAX_SSTHRESH);
		if (tp->snd_ssthresh > tp->snd_cwnd_clamp)
			tp->snd_ssthresh = tp->snd_cwnd_clamp;
	}
	if (dst_metric(dst, RTAX_REORDERING) &&
	    tp->reordering != dst_metric(dst, RTAX_REORDERING)) {
888
		tcp_disable_fack(tp);
L
Linus Torvalds 已提交
889 890 891 892 893 894
		tp->reordering = dst_metric(dst, RTAX_REORDERING);
	}

	if (dst_metric(dst, RTAX_RTT) == 0)
		goto reset;

S
Stephen Hemminger 已提交
895
	if (!tp->srtt && dst_metric_rtt(dst, RTAX_RTT) < (TCP_TIMEOUT_INIT << 3))
L
Linus Torvalds 已提交
896 897 898 899 900 901 902 903
		goto reset;

	/* Initial rtt is determined from SYN,SYN-ACK.
	 * The segment is small and rtt may appear much
	 * less than real one. Use per-dst memory
	 * to make it more realistic.
	 *
	 * A bit of theory. RTT is time passed after "normal" sized packet
S
Stephen Hemminger 已提交
904
	 * is sent until it is ACKed. In normal circumstances sending small
L
Linus Torvalds 已提交
905 906 907 908 909 910 911
	 * packets force peer to delay ACKs and calculation is correct too.
	 * The algorithm is adaptive and, provided we follow specs, it
	 * NEVER underestimate RTT. BUT! If peer tries to make some clever
	 * tricks sort of "quick acks" for time long enough to decrease RTT
	 * to low value, and then abruptly stops to do it and starts to delay
	 * ACKs, wait for troubles.
	 */
S
Stephen Hemminger 已提交
912 913
	if (dst_metric_rtt(dst, RTAX_RTT) > tp->srtt) {
		tp->srtt = dst_metric_rtt(dst, RTAX_RTT);
L
Linus Torvalds 已提交
914 915
		tp->rtt_seq = tp->snd_nxt;
	}
S
Stephen Hemminger 已提交
916 917
	if (dst_metric_rtt(dst, RTAX_RTTVAR) > tp->mdev) {
		tp->mdev = dst_metric_rtt(dst, RTAX_RTTVAR);
918
		tp->mdev_max = tp->rttvar = max(tp->mdev, tcp_rto_min(sk));
L
Linus Torvalds 已提交
919
	}
920 921
	tcp_set_rto(sk);
	if (inet_csk(sk)->icsk_rto < TCP_TIMEOUT_INIT && !tp->rx_opt.saw_tstamp)
L
Linus Torvalds 已提交
922
		goto reset;
I
Ilpo Järvinen 已提交
923 924

cwnd:
L
Linus Torvalds 已提交
925 926 927 928 929 930 931 932 933 934 935 936
	tp->snd_cwnd = tcp_init_cwnd(tp, dst);
	tp->snd_cwnd_stamp = tcp_time_stamp;
	return;

reset:
	/* Play conservative. If timestamps are not
	 * supported, TCP will fail to recalculate correct
	 * rtt, if initial rto is too small. FORGET ALL AND RESET!
	 */
	if (!tp->rx_opt.saw_tstamp && tp->srtt) {
		tp->srtt = 0;
		tp->mdev = tp->mdev_max = tp->rttvar = TCP_TIMEOUT_INIT;
937
		inet_csk(sk)->icsk_rto = TCP_TIMEOUT_INIT;
L
Linus Torvalds 已提交
938
	}
I
Ilpo Järvinen 已提交
939
	goto cwnd;
L
Linus Torvalds 已提交
940 941
}

942 943
static void tcp_update_reordering(struct sock *sk, const int metric,
				  const int ts)
L
Linus Torvalds 已提交
944
{
945
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
946
	if (metric > tp->reordering) {
947 948
		int mib_idx;

L
Linus Torvalds 已提交
949 950 951 952
		tp->reordering = min(TCP_MAX_REORDERING, metric);

		/* This exciting event is worth to be remembered. 8) */
		if (ts)
953
			mib_idx = LINUX_MIB_TCPTSREORDER;
954
		else if (tcp_is_reno(tp))
955
			mib_idx = LINUX_MIB_TCPRENOREORDER;
956
		else if (tcp_is_fack(tp))
957
			mib_idx = LINUX_MIB_TCPFACKREORDER;
L
Linus Torvalds 已提交
958
		else
959 960
			mib_idx = LINUX_MIB_TCPSACKREORDER;

961
		NET_INC_STATS_BH(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
962 963
#if FASTRETRANS_DEBUG > 1
		printk(KERN_DEBUG "Disorder%d %d %u f%u s%u rr%d\n",
964
		       tp->rx_opt.sack_ok, inet_csk(sk)->icsk_ca_state,
L
Linus Torvalds 已提交
965 966 967 968 969
		       tp->reordering,
		       tp->fackets_out,
		       tp->sacked_out,
		       tp->undo_marker ? tp->undo_retrans : 0);
#endif
970
		tcp_disable_fack(tp);
L
Linus Torvalds 已提交
971 972 973
	}
}

974
/* This must be called before lost_out is incremented */
975 976
static void tcp_verify_retransmit_hint(struct tcp_sock *tp, struct sk_buff *skb)
{
977
	if ((tp->retransmit_skb_hint == NULL) ||
978 979
	    before(TCP_SKB_CB(skb)->seq,
		   TCP_SKB_CB(tp->retransmit_skb_hint)->seq))
980 981 982 983 984
		tp->retransmit_skb_hint = skb;

	if (!tp->lost_out ||
	    after(TCP_SKB_CB(skb)->end_seq, tp->retransmit_high))
		tp->retransmit_high = TCP_SKB_CB(skb)->end_seq;
985 986
}

987 988 989 990 991 992 993 994 995 996
static void tcp_skb_mark_lost(struct tcp_sock *tp, struct sk_buff *skb)
{
	if (!(TCP_SKB_CB(skb)->sacked & (TCPCB_LOST|TCPCB_SACKED_ACKED))) {
		tcp_verify_retransmit_hint(tp, skb);

		tp->lost_out += tcp_skb_pcount(skb);
		TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
	}
}

997 998
static void tcp_skb_mark_lost_uncond_verify(struct tcp_sock *tp,
					    struct sk_buff *skb)
999 1000 1001 1002 1003 1004 1005 1006 1007
{
	tcp_verify_retransmit_hint(tp, skb);

	if (!(TCP_SKB_CB(skb)->sacked & (TCPCB_LOST|TCPCB_SACKED_ACKED))) {
		tp->lost_out += tcp_skb_pcount(skb);
		TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
	}
}

L
Linus Torvalds 已提交
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054
/* This procedure tags the retransmission queue when SACKs arrive.
 *
 * We have three tag bits: SACKED(S), RETRANS(R) and LOST(L).
 * Packets in queue with these bits set are counted in variables
 * sacked_out, retrans_out and lost_out, correspondingly.
 *
 * Valid combinations are:
 * Tag  InFlight	Description
 * 0	1		- orig segment is in flight.
 * S	0		- nothing flies, orig reached receiver.
 * L	0		- nothing flies, orig lost by net.
 * R	2		- both orig and retransmit are in flight.
 * L|R	1		- orig is lost, retransmit is in flight.
 * S|R  1		- orig reached receiver, retrans is still in flight.
 * (L|S|R is logically valid, it could occur when L|R is sacked,
 *  but it is equivalent to plain S and code short-curcuits it to S.
 *  L|S is logically invalid, it would mean -1 packet in flight 8))
 *
 * These 6 states form finite state machine, controlled by the following events:
 * 1. New ACK (+SACK) arrives. (tcp_sacktag_write_queue())
 * 2. Retransmission. (tcp_retransmit_skb(), tcp_xmit_retransmit_queue())
 * 3. Loss detection event of one of three flavors:
 *	A. Scoreboard estimator decided the packet is lost.
 *	   A'. Reno "three dupacks" marks head of queue lost.
 *	   A''. Its FACK modfication, head until snd.fack is lost.
 *	B. SACK arrives sacking data transmitted after never retransmitted
 *	   hole was sent out.
 *	C. SACK arrives sacking SND.NXT at the moment, when the
 *	   segment was retransmitted.
 * 4. D-SACK added new rule: D-SACK changes any tag to S.
 *
 * It is pleasant to note, that state diagram turns out to be commutative,
 * so that we are allowed not to be bothered by order of our actions,
 * when multiple events arrive simultaneously. (see the function below).
 *
 * Reordering detection.
 * --------------------
 * Reordering metric is maximal distance, which a packet can be displaced
 * in packet stream. With SACKs we can estimate it:
 *
 * 1. SACK fills old hole and the corresponding segment was not
 *    ever retransmitted -> reordering. Alas, we cannot use it
 *    when segment was retransmitted.
 * 2. The last flaw is solved with D-SACK. D-SACK arrives
 *    for retransmitted and already SACKed segment -> reordering..
 * Both of these heuristics are not used in Loss state, when we cannot
 * account for retransmits accurately.
I
Ilpo Järvinen 已提交
1055 1056 1057 1058 1059 1060 1061
 *
 * SACK block validation.
 * ----------------------
 *
 * SACK block range validation checks that the received SACK block fits to
 * the expected sequence limits, i.e., it is between SND.UNA and SND.NXT.
 * Note that SND.UNA is not included to the range though being valid because
1062 1063 1064 1065 1066 1067 1068 1069 1070
 * it means that the receiver is rather inconsistent with itself reporting
 * SACK reneging when it should advance SND.UNA. Such SACK block this is
 * perfectly valid, however, in light of RFC2018 which explicitly states
 * that "SACK block MUST reflect the newest segment.  Even if the newest
 * segment is going to be discarded ...", not that it looks very clever
 * in case of head skb. Due to potentional receiver driven attacks, we
 * choose to avoid immediate execution of a walk in write queue due to
 * reneging and defer head skb's loss recovery to standard loss recovery
 * procedure that will eventually trigger (nothing forbids us doing this).
I
Ilpo Järvinen 已提交
1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
 *
 * Implements also blockage to start_seq wrap-around. Problem lies in the
 * fact that though start_seq (s) is before end_seq (i.e., not reversed),
 * there's no guarantee that it will be before snd_nxt (n). The problem
 * happens when start_seq resides between end_seq wrap (e_w) and snd_nxt
 * wrap (s_w):
 *
 *         <- outs wnd ->                          <- wrapzone ->
 *         u     e      n                         u_w   e_w  s n_w
 *         |     |      |                          |     |   |  |
 * |<------------+------+----- TCP seqno space --------------+---------->|
 * ...-- <2^31 ->|                                           |<--------...
 * ...---- >2^31 ------>|                                    |<--------...
 *
 * Current code wouldn't be vulnerable but it's better still to discard such
 * crazy SACK blocks. Doing this check for start_seq alone closes somewhat
 * similar case (end_seq after snd_nxt wrap) as earlier reversed check in
 * snd_nxt wrap -> snd_una region will then become "well defined", i.e.,
 * equal to the ideal case (infinite seqno space without wrap caused issues).
 *
 * With D-SACK the lower bound is extended to cover sequence space below
 * SND.UNA down to undo_marker, which is the last point of interest. Yet
1093
 * again, D-SACK block must not to go across snd_una (for the same reason as
I
Ilpo Järvinen 已提交
1094 1095 1096 1097 1098 1099 1100 1101 1102
 * for the normal SACK blocks, explained above). But there all simplicity
 * ends, TCP might receive valid D-SACKs below that. As long as they reside
 * fully below undo_marker they do not affect behavior in anyway and can
 * therefore be safely ignored. In rare cases (which are more or less
 * theoretical ones), the D-SACK will nicely cross that boundary due to skb
 * fragmentation and packet reordering past skb's retransmission. To consider
 * them correctly, the acceptable range must be extended even more though
 * the exact amount is rather hard to quantify. However, tp->max_window can
 * be used as an exaggerated estimate.
L
Linus Torvalds 已提交
1103
 */
I
Ilpo Järvinen 已提交
1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114
static int tcp_is_sackblock_valid(struct tcp_sock *tp, int is_dsack,
				  u32 start_seq, u32 end_seq)
{
	/* Too far in future, or reversed (interpretation is ambiguous) */
	if (after(end_seq, tp->snd_nxt) || !before(start_seq, end_seq))
		return 0;

	/* Nasty start_seq wrap-around check (see comments above) */
	if (!before(start_seq, tp->snd_nxt))
		return 0;

1115
	/* In outstanding window? ...This is valid exit for D-SACKs too.
I
Ilpo Järvinen 已提交
1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140
	 * start_seq == snd_una is non-sensical (see comments above)
	 */
	if (after(start_seq, tp->snd_una))
		return 1;

	if (!is_dsack || !tp->undo_marker)
		return 0;

	/* ...Then it's D-SACK, and must reside below snd_una completely */
	if (!after(end_seq, tp->snd_una))
		return 0;

	if (!before(start_seq, tp->undo_marker))
		return 1;

	/* Too old */
	if (!after(end_seq, tp->undo_marker))
		return 0;

	/* Undo_marker boundary crossing (overestimates a lot). Known already:
	 *   start_seq < undo_marker and end_seq >= undo_marker.
	 */
	return !before(start_seq, end_seq - tp->max_window);
}

1141 1142 1143
/* Check for lost retransmit. This superb idea is borrowed from "ratehalving".
 * Event "C". Later note: FACK people cheated me again 8), we have to account
 * for reordering! Ugly, but should help.
1144 1145 1146
 *
 * Search retransmitted skbs from write_queue that were sent when snd_nxt was
 * less than what is now known to be received by the other end (derived from
1147 1148
 * highest SACK block). Also calculate the lowest snd_nxt among the remaining
 * retransmitted skbs to avoid some costly processing per ACKs.
1149
 */
1150
static void tcp_mark_lost_retrans(struct sock *sk)
1151
{
1152
	const struct inet_connection_sock *icsk = inet_csk(sk);
1153 1154
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
1155
	int cnt = 0;
1156
	u32 new_low_seq = tp->snd_nxt;
1157
	u32 received_upto = tcp_highest_sack_seq(tp);
1158 1159 1160 1161

	if (!tcp_is_fack(tp) || !tp->retrans_out ||
	    !after(received_upto, tp->lost_retrans_low) ||
	    icsk->icsk_ca_state != TCP_CA_Recovery)
1162
		return;
1163 1164 1165 1166 1167 1168

	tcp_for_write_queue(skb, sk) {
		u32 ack_seq = TCP_SKB_CB(skb)->ack_seq;

		if (skb == tcp_send_head(sk))
			break;
1169
		if (cnt == tp->retrans_out)
1170 1171 1172 1173
			break;
		if (!after(TCP_SKB_CB(skb)->end_seq, tp->snd_una))
			continue;

1174 1175 1176
		if (!(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS))
			continue;

1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
		/* TODO: We would like to get rid of tcp_is_fack(tp) only
		 * constraint here (see above) but figuring out that at
		 * least tp->reordering SACK blocks reside between ack_seq
		 * and received_upto is not easy task to do cheaply with
		 * the available datastructures.
		 *
		 * Whether FACK should check here for tp->reordering segs
		 * in-between one could argue for either way (it would be
		 * rather simple to implement as we could count fack_count
		 * during the walk and do tp->fackets_out - fack_count).
		 */
		if (after(received_upto, ack_seq)) {
1189 1190 1191
			TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS;
			tp->retrans_out -= tcp_skb_pcount(skb);

1192
			tcp_skb_mark_lost_uncond_verify(tp, skb);
1193
			NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPLOSTRETRANSMIT);
1194
		} else {
1195
			if (before(ack_seq, new_low_seq))
1196
				new_low_seq = ack_seq;
1197
			cnt += tcp_skb_pcount(skb);
1198 1199
		}
	}
1200 1201 1202

	if (tp->retrans_out)
		tp->lost_retrans_low = new_low_seq;
1203
}
I
Ilpo Järvinen 已提交
1204

1205
static int tcp_check_dsack(struct sock *sk, struct sk_buff *ack_skb,
1206 1207 1208
			   struct tcp_sack_block_wire *sp, int num_sacks,
			   u32 prior_snd_una)
{
1209
	struct tcp_sock *tp = tcp_sk(sk);
1210 1211
	u32 start_seq_0 = get_unaligned_be32(&sp[0].start_seq);
	u32 end_seq_0 = get_unaligned_be32(&sp[0].end_seq);
1212 1213 1214 1215
	int dup_sack = 0;

	if (before(start_seq_0, TCP_SKB_CB(ack_skb)->ack_seq)) {
		dup_sack = 1;
1216
		tcp_dsack_seen(tp);
1217
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPDSACKRECV);
1218
	} else if (num_sacks > 1) {
1219 1220
		u32 end_seq_1 = get_unaligned_be32(&sp[1].end_seq);
		u32 start_seq_1 = get_unaligned_be32(&sp[1].start_seq);
1221 1222 1223 1224

		if (!after(end_seq_0, end_seq_1) &&
		    !before(start_seq_0, start_seq_1)) {
			dup_sack = 1;
1225
			tcp_dsack_seen(tp);
1226 1227
			NET_INC_STATS_BH(sock_net(sk),
					LINUX_MIB_TCPDSACKOFORECV);
1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
		}
	}

	/* D-SACK for already forgotten data... Do dumb counting. */
	if (dup_sack &&
	    !after(end_seq_0, prior_snd_una) &&
	    after(end_seq_0, tp->undo_marker))
		tp->undo_retrans--;

	return dup_sack;
}

1240 1241 1242 1243 1244 1245
struct tcp_sacktag_state {
	int reord;
	int fack_count;
	int flag;
};

1246 1247 1248 1249 1250
/* Check if skb is fully within the SACK block. In presence of GSO skbs,
 * the incoming SACK may not exactly match but we can find smaller MSS
 * aligned portion of it that matches. Therefore we might need to fragment
 * which may fail and creates some hassle (caller must handle error case
 * returns).
1251 1252
 *
 * FIXME: this could be merged to shift decision code
1253
 */
1254 1255
static int tcp_match_skb_to_sack(struct sock *sk, struct sk_buff *skb,
				 u32 start_seq, u32 end_seq)
1256 1257 1258
{
	int in_sack, err;
	unsigned int pkt_len;
1259
	unsigned int mss;
1260 1261 1262 1263 1264 1265

	in_sack = !after(start_seq, TCP_SKB_CB(skb)->seq) &&
		  !before(end_seq, TCP_SKB_CB(skb)->end_seq);

	if (tcp_skb_pcount(skb) > 1 && !in_sack &&
	    after(TCP_SKB_CB(skb)->end_seq, start_seq)) {
1266
		mss = tcp_skb_mss(skb);
1267 1268
		in_sack = !after(start_seq, TCP_SKB_CB(skb)->seq);

1269
		if (!in_sack) {
1270
			pkt_len = start_seq - TCP_SKB_CB(skb)->seq;
1271 1272 1273
			if (pkt_len < mss)
				pkt_len = mss;
		} else {
1274
			pkt_len = end_seq - TCP_SKB_CB(skb)->seq;
1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291
			if (pkt_len < mss)
				return -EINVAL;
		}

		/* Round if necessary so that SACKs cover only full MSSes
		 * and/or the remaining small portion (if present)
		 */
		if (pkt_len > mss) {
			unsigned int new_len = (pkt_len / mss) * mss;
			if (!in_sack && new_len < pkt_len) {
				new_len += mss;
				if (new_len > skb->len)
					return 0;
			}
			pkt_len = new_len;
		}
		err = tcp_fragment(sk, skb, pkt_len, mss);
1292 1293 1294 1295 1296 1297 1298
		if (err < 0)
			return err;
	}

	return in_sack;
}

1299 1300 1301
static u8 tcp_sacktag_one(struct sk_buff *skb, struct sock *sk,
			  struct tcp_sacktag_state *state,
			  int dup_sack, int pcount)
1302
{
1303
	struct tcp_sock *tp = tcp_sk(sk);
1304
	u8 sacked = TCP_SKB_CB(skb)->sacked;
1305
	int fack_count = state->fack_count;
1306 1307 1308 1309 1310

	/* Account D-SACK for retransmitted packet. */
	if (dup_sack && (sacked & TCPCB_RETRANS)) {
		if (after(TCP_SKB_CB(skb)->end_seq, tp->undo_marker))
			tp->undo_retrans--;
1311
		if (sacked & TCPCB_SACKED_ACKED)
1312
			state->reord = min(fack_count, state->reord);
1313 1314 1315 1316
	}

	/* Nothing to do; acked frame is about to be dropped (was ACKed). */
	if (!after(TCP_SKB_CB(skb)->end_seq, tp->snd_una))
1317
		return sacked;
1318 1319 1320 1321 1322 1323 1324 1325

	if (!(sacked & TCPCB_SACKED_ACKED)) {
		if (sacked & TCPCB_SACKED_RETRANS) {
			/* If the segment is not tagged as lost,
			 * we do not clear RETRANS, believing
			 * that retransmission is still in flight.
			 */
			if (sacked & TCPCB_LOST) {
1326
				sacked &= ~(TCPCB_LOST|TCPCB_SACKED_RETRANS);
1327 1328
				tp->lost_out -= pcount;
				tp->retrans_out -= pcount;
1329 1330 1331 1332 1333 1334 1335 1336
			}
		} else {
			if (!(sacked & TCPCB_RETRANS)) {
				/* New sack for not retransmitted frame,
				 * which was in hole. It is reordering.
				 */
				if (before(TCP_SKB_CB(skb)->seq,
					   tcp_highest_sack_seq(tp)))
1337 1338
					state->reord = min(fack_count,
							   state->reord);
1339 1340 1341

				/* SACK enhanced F-RTO (RFC4138; Appendix B) */
				if (!after(TCP_SKB_CB(skb)->end_seq, tp->frto_highmark))
1342
					state->flag |= FLAG_ONLY_ORIG_SACKED;
1343 1344 1345
			}

			if (sacked & TCPCB_LOST) {
1346
				sacked &= ~TCPCB_LOST;
1347
				tp->lost_out -= pcount;
1348 1349 1350
			}
		}

1351 1352
		sacked |= TCPCB_SACKED_ACKED;
		state->flag |= FLAG_DATA_SACKED;
1353
		tp->sacked_out += pcount;
1354

1355
		fack_count += pcount;
1356 1357 1358 1359 1360

		/* Lost marker hint past SACKed? Tweak RFC3517 cnt */
		if (!tcp_is_fack(tp) && (tp->lost_skb_hint != NULL) &&
		    before(TCP_SKB_CB(skb)->seq,
			   TCP_SKB_CB(tp->lost_skb_hint)->seq))
1361
			tp->lost_cnt_hint += pcount;
1362 1363 1364 1365 1366 1367 1368 1369 1370

		if (fack_count > tp->fackets_out)
			tp->fackets_out = fack_count;
	}

	/* D-SACK. We can detect redundant retransmission in S|R and plain R
	 * frames and clear it. undo_retrans is decreased above, L|R frames
	 * are accounted above as well.
	 */
1371 1372
	if (dup_sack && (sacked & TCPCB_SACKED_RETRANS)) {
		sacked &= ~TCPCB_SACKED_RETRANS;
1373
		tp->retrans_out -= pcount;
1374 1375
	}

1376
	return sacked;
1377 1378
}

1379
static int tcp_shifted_skb(struct sock *sk, struct sk_buff *skb,
1380
			   struct tcp_sacktag_state *state,
I
Ilpo Järvinen 已提交
1381 1382
			   unsigned int pcount, int shifted, int mss,
			   int dup_sack)
1383 1384
{
	struct tcp_sock *tp = tcp_sk(sk);
1385
	struct sk_buff *prev = tcp_write_queue_prev(sk, skb);
1386 1387 1388

	BUG_ON(!pcount);

1389 1390 1391 1392 1393
	/* Tweak before seqno plays */
	if (!tcp_is_fack(tp) && tcp_is_sack(tp) && tp->lost_skb_hint &&
	    !before(TCP_SKB_CB(tp->lost_skb_hint)->seq, TCP_SKB_CB(skb)->seq))
		tp->lost_cnt_hint += pcount;

1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416
	TCP_SKB_CB(prev)->end_seq += shifted;
	TCP_SKB_CB(skb)->seq += shifted;

	skb_shinfo(prev)->gso_segs += pcount;
	BUG_ON(skb_shinfo(skb)->gso_segs < pcount);
	skb_shinfo(skb)->gso_segs -= pcount;

	/* When we're adding to gso_segs == 1, gso_size will be zero,
	 * in theory this shouldn't be necessary but as long as DSACK
	 * code can come after this skb later on it's better to keep
	 * setting gso_size to something.
	 */
	if (!skb_shinfo(prev)->gso_size) {
		skb_shinfo(prev)->gso_size = mss;
		skb_shinfo(prev)->gso_type = sk->sk_gso_type;
	}

	/* CHECKME: To clear or not to clear? Mimics normal skb currently */
	if (skb_shinfo(skb)->gso_segs <= 1) {
		skb_shinfo(skb)->gso_size = 0;
		skb_shinfo(skb)->gso_type = 0;
	}

1417
	/* We discard results */
I
Ilpo Järvinen 已提交
1418
	tcp_sacktag_one(skb, sk, state, dup_sack, pcount);
1419 1420 1421 1422 1423 1424

	/* Difference in this won't matter, both ACKed by the same cumul. ACK */
	TCP_SKB_CB(prev)->sacked |= (TCP_SKB_CB(skb)->sacked & TCPCB_EVER_RETRANS);

	if (skb->len > 0) {
		BUG_ON(!tcp_skb_pcount(skb));
1425
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_SACKSHIFTED);
1426 1427 1428 1429 1430
		return 0;
	}

	/* Whole SKB was eaten :-) */

1431 1432 1433 1434 1435 1436 1437 1438 1439
	if (skb == tp->retransmit_skb_hint)
		tp->retransmit_skb_hint = prev;
	if (skb == tp->scoreboard_skb_hint)
		tp->scoreboard_skb_hint = prev;
	if (skb == tp->lost_skb_hint) {
		tp->lost_skb_hint = prev;
		tp->lost_cnt_hint -= tcp_skb_pcount(prev);
	}

1440 1441 1442 1443 1444 1445 1446
	TCP_SKB_CB(skb)->flags |= TCP_SKB_CB(prev)->flags;
	if (skb == tcp_highest_sack(sk))
		tcp_advance_highest_sack(sk, skb);

	tcp_unlink_write_queue(skb, sk);
	sk_wmem_free_skb(sk, skb);

1447 1448
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_SACKMERGED);

1449 1450 1451 1452 1453 1454
	return 1;
}

/* I wish gso_size would have a bit more sane initialization than
 * something-or-zero which complicates things
 */
1455
static int tcp_skb_seglen(struct sk_buff *skb)
1456
{
1457
	return tcp_skb_pcount(skb) == 1 ? skb->len : tcp_skb_mss(skb);
1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469
}

/* Shifting pages past head area doesn't work */
static int skb_can_shift(struct sk_buff *skb)
{
	return !skb_headlen(skb) && skb_is_nonlinear(skb);
}

/* Try collapsing SACK blocks spanning across multiple skbs to a single
 * skb.
 */
static struct sk_buff *tcp_shift_skb_data(struct sock *sk, struct sk_buff *skb,
1470
					  struct tcp_sacktag_state *state,
1471
					  u32 start_seq, u32 end_seq,
1472
					  int dup_sack)
1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *prev;
	int mss;
	int pcount = 0;
	int len;
	int in_sack;

	if (!sk_can_gso(sk))
		goto fallback;

	/* Normally R but no L won't result in plain S */
	if (!dup_sack &&
1486
	    (TCP_SKB_CB(skb)->sacked & (TCPCB_LOST|TCPCB_SACKED_RETRANS)) == TCPCB_SACKED_RETRANS)
1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507
		goto fallback;
	if (!skb_can_shift(skb))
		goto fallback;
	/* This frame is about to be dropped (was ACKed). */
	if (!after(TCP_SKB_CB(skb)->end_seq, tp->snd_una))
		goto fallback;

	/* Can only happen with delayed DSACK + discard craziness */
	if (unlikely(skb == tcp_write_queue_head(sk)))
		goto fallback;
	prev = tcp_write_queue_prev(sk, skb);

	if ((TCP_SKB_CB(prev)->sacked & TCPCB_TAGBITS) != TCPCB_SACKED_ACKED)
		goto fallback;

	in_sack = !after(start_seq, TCP_SKB_CB(skb)->seq) &&
		  !before(end_seq, TCP_SKB_CB(skb)->end_seq);

	if (in_sack) {
		len = skb->len;
		pcount = tcp_skb_pcount(skb);
1508
		mss = tcp_skb_seglen(skb);
1509 1510 1511 1512

		/* TODO: Fix DSACKs to not fragment already SACKed and we can
		 * drop this restriction as unnecessary
		 */
1513
		if (mss != tcp_skb_seglen(prev))
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553
			goto fallback;
	} else {
		if (!after(TCP_SKB_CB(skb)->end_seq, start_seq))
			goto noop;
		/* CHECKME: This is non-MSS split case only?, this will
		 * cause skipped skbs due to advancing loop btw, original
		 * has that feature too
		 */
		if (tcp_skb_pcount(skb) <= 1)
			goto noop;

		in_sack = !after(start_seq, TCP_SKB_CB(skb)->seq);
		if (!in_sack) {
			/* TODO: head merge to next could be attempted here
			 * if (!after(TCP_SKB_CB(skb)->end_seq, end_seq)),
			 * though it might not be worth of the additional hassle
			 *
			 * ...we can probably just fallback to what was done
			 * previously. We could try merging non-SACKed ones
			 * as well but it probably isn't going to buy off
			 * because later SACKs might again split them, and
			 * it would make skb timestamp tracking considerably
			 * harder problem.
			 */
			goto fallback;
		}

		len = end_seq - TCP_SKB_CB(skb)->seq;
		BUG_ON(len < 0);
		BUG_ON(len > skb->len);

		/* MSS boundaries should be honoured or else pcount will
		 * severely break even though it makes things bit trickier.
		 * Optimize common case to avoid most of the divides
		 */
		mss = tcp_skb_mss(skb);

		/* TODO: Fix DSACKs to not fragment already SACKed and we can
		 * drop this restriction as unnecessary
		 */
1554
		if (mss != tcp_skb_seglen(prev))
1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568
			goto fallback;

		if (len == mss) {
			pcount = 1;
		} else if (len < mss) {
			goto noop;
		} else {
			pcount = len / mss;
			len = pcount * mss;
		}
	}

	if (!skb_shift(prev, skb, len))
		goto fallback;
I
Ilpo Järvinen 已提交
1569
	if (!tcp_shifted_skb(sk, skb, state, pcount, len, mss, dup_sack))
1570 1571 1572 1573 1574 1575 1576 1577 1578
		goto out;

	/* Hole filled allows collapsing with the next as well, this is very
	 * useful when hole on every nth skb pattern happens
	 */
	if (prev == tcp_write_queue_tail(sk))
		goto out;
	skb = tcp_write_queue_next(sk, prev);

1579 1580 1581
	if (!skb_can_shift(skb) ||
	    (skb == tcp_send_head(sk)) ||
	    ((TCP_SKB_CB(skb)->sacked & TCPCB_TAGBITS) != TCPCB_SACKED_ACKED) ||
1582
	    (mss != tcp_skb_seglen(skb)))
1583 1584 1585 1586 1587
		goto out;

	len = skb->len;
	if (skb_shift(prev, skb, len)) {
		pcount += tcp_skb_pcount(skb);
I
Ilpo Järvinen 已提交
1588
		tcp_shifted_skb(sk, skb, state, tcp_skb_pcount(skb), len, mss, 0);
1589 1590 1591
	}

out:
1592
	state->fack_count += pcount;
1593 1594 1595 1596 1597 1598
	return prev;

noop:
	return skb;

fallback:
1599
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_SACKSHIFTFALLBACK);
1600 1601 1602
	return NULL;
}

1603 1604
static struct sk_buff *tcp_sacktag_walk(struct sk_buff *skb, struct sock *sk,
					struct tcp_sack_block *next_dup,
1605
					struct tcp_sacktag_state *state,
1606
					u32 start_seq, u32 end_seq,
1607
					int dup_sack_in)
1608
{
1609 1610 1611
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *tmp;

1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631
	tcp_for_write_queue_from(skb, sk) {
		int in_sack = 0;
		int dup_sack = dup_sack_in;

		if (skb == tcp_send_head(sk))
			break;

		/* queue is in-order => we can short-circuit the walk early */
		if (!before(TCP_SKB_CB(skb)->seq, end_seq))
			break;

		if ((next_dup != NULL) &&
		    before(TCP_SKB_CB(skb)->seq, next_dup->end_seq)) {
			in_sack = tcp_match_skb_to_sack(sk, skb,
							next_dup->start_seq,
							next_dup->end_seq);
			if (in_sack > 0)
				dup_sack = 1;
		}

1632 1633 1634 1635 1636
		/* skb reference here is a bit tricky to get right, since
		 * shifting can eat and free both this skb and the next,
		 * so not even _safe variant of the loop is enough.
		 */
		if (in_sack <= 0) {
1637 1638
			tmp = tcp_shift_skb_data(sk, skb, state,
						 start_seq, end_seq, dup_sack);
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652
			if (tmp != NULL) {
				if (tmp != skb) {
					skb = tmp;
					continue;
				}

				in_sack = 0;
			} else {
				in_sack = tcp_match_skb_to_sack(sk, skb,
								start_seq,
								end_seq);
			}
		}

1653 1654 1655
		if (unlikely(in_sack < 0))
			break;

1656
		if (in_sack) {
1657 1658 1659 1660
			TCP_SKB_CB(skb)->sacked = tcp_sacktag_one(skb, sk,
								  state,
								  dup_sack,
								  tcp_skb_pcount(skb));
1661

1662 1663 1664 1665 1666
			if (!before(TCP_SKB_CB(skb)->seq,
				    tcp_highest_sack_seq(tp)))
				tcp_advance_highest_sack(sk, skb);
		}

1667
		state->fack_count += tcp_skb_pcount(skb);
1668 1669 1670 1671 1672 1673 1674 1675
	}
	return skb;
}

/* Avoid all extra work that is being done by sacktag while walking in
 * a normal way
 */
static struct sk_buff *tcp_sacktag_skip(struct sk_buff *skb, struct sock *sk,
1676 1677
					struct tcp_sacktag_state *state,
					u32 skip_to_seq)
1678 1679 1680 1681 1682
{
	tcp_for_write_queue_from(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;

I
Ilpo Järvinen 已提交
1683
		if (after(TCP_SKB_CB(skb)->end_seq, skip_to_seq))
1684
			break;
1685

1686
		state->fack_count += tcp_skb_pcount(skb);
1687 1688 1689 1690 1691 1692 1693
	}
	return skb;
}

static struct sk_buff *tcp_maybe_skipping_dsack(struct sk_buff *skb,
						struct sock *sk,
						struct tcp_sack_block *next_dup,
1694 1695
						struct tcp_sacktag_state *state,
						u32 skip_to_seq)
1696 1697 1698 1699 1700
{
	if (next_dup == NULL)
		return skb;

	if (before(next_dup->start_seq, skip_to_seq)) {
1701 1702 1703 1704
		skb = tcp_sacktag_skip(skb, sk, state, next_dup->start_seq);
		skb = tcp_sacktag_walk(skb, sk, NULL, state,
				       next_dup->start_seq, next_dup->end_seq,
				       1);
1705 1706 1707 1708 1709 1710 1711 1712 1713 1714
	}

	return skb;
}

static int tcp_sack_cache_ok(struct tcp_sock *tp, struct tcp_sack_block *cache)
{
	return cache < tp->recv_sack_cache + ARRAY_SIZE(tp->recv_sack_cache);
}

L
Linus Torvalds 已提交
1715
static int
1716 1717
tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb,
			u32 prior_snd_una)
L
Linus Torvalds 已提交
1718
{
1719
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1720
	struct tcp_sock *tp = tcp_sk(sk);
1721 1722
	unsigned char *ptr = (skb_transport_header(ack_skb) +
			      TCP_SKB_CB(ack_skb)->sacked);
1723
	struct tcp_sack_block_wire *sp_wire = (struct tcp_sack_block_wire *)(ptr+2);
1724
	struct tcp_sack_block sp[TCP_NUM_SACKS];
1725
	struct tcp_sack_block *cache;
1726
	struct tcp_sacktag_state state;
1727
	struct sk_buff *skb;
1728
	int num_sacks = min(TCP_NUM_SACKS, (ptr[1] - TCPOLEN_SACK_BASE) >> 3);
1729
	int used_sacks;
1730
	int found_dup_sack = 0;
1731
	int i, j;
1732
	int first_sack_index;
L
Linus Torvalds 已提交
1733

1734 1735 1736
	state.flag = 0;
	state.reord = tp->packets_out;

1737
	if (!tp->sacked_out) {
1738 1739
		if (WARN_ON(tp->fackets_out))
			tp->fackets_out = 0;
1740
		tcp_highest_sack_reset(sk);
1741
	}
L
Linus Torvalds 已提交
1742

1743
	found_dup_sack = tcp_check_dsack(sk, ack_skb, sp_wire,
1744 1745
					 num_sacks, prior_snd_una);
	if (found_dup_sack)
1746
		state.flag |= FLAG_DSACKING_ACK;
1747 1748 1749 1750 1751 1752 1753 1754

	/* Eliminate too old ACKs, but take into
	 * account more or less fresh ones, they can
	 * contain valid SACK info.
	 */
	if (before(TCP_SKB_CB(ack_skb)->ack_seq, prior_snd_una - tp->max_window))
		return 0;

1755 1756 1757
	if (!tp->packets_out)
		goto out;

1758 1759 1760 1761 1762
	used_sacks = 0;
	first_sack_index = 0;
	for (i = 0; i < num_sacks; i++) {
		int dup_sack = !i && found_dup_sack;

1763 1764
		sp[used_sacks].start_seq = get_unaligned_be32(&sp_wire[i].start_seq);
		sp[used_sacks].end_seq = get_unaligned_be32(&sp_wire[i].end_seq);
1765 1766 1767 1768

		if (!tcp_is_sackblock_valid(tp, dup_sack,
					    sp[used_sacks].start_seq,
					    sp[used_sacks].end_seq)) {
1769 1770
			int mib_idx;

1771 1772
			if (dup_sack) {
				if (!tp->undo_marker)
1773
					mib_idx = LINUX_MIB_TCPDSACKIGNOREDNOUNDO;
1774
				else
1775
					mib_idx = LINUX_MIB_TCPDSACKIGNOREDOLD;
1776 1777 1778 1779 1780
			} else {
				/* Don't count olds caused by ACK reordering */
				if ((TCP_SKB_CB(ack_skb)->ack_seq != tp->snd_una) &&
				    !after(sp[used_sacks].end_seq, tp->snd_una))
					continue;
1781
				mib_idx = LINUX_MIB_TCPSACKDISCARD;
1782
			}
1783

1784
			NET_INC_STATS_BH(sock_net(sk), mib_idx);
1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796
			if (i == 0)
				first_sack_index = -1;
			continue;
		}

		/* Ignore very old stuff early */
		if (!after(sp[used_sacks].end_seq, prior_snd_una))
			continue;

		used_sacks++;
	}

1797 1798
	/* order SACK blocks to allow in order walk of the retrans queue */
	for (i = used_sacks - 1; i > 0; i--) {
1799 1800
		for (j = 0; j < i; j++) {
			if (after(sp[j].start_seq, sp[j + 1].start_seq)) {
1801
				swap(sp[j], sp[j + 1]);
1802

1803 1804
				/* Track where the first SACK block goes to */
				if (j == first_sack_index)
1805
					first_sack_index = j + 1;
1806 1807 1808 1809
			}
		}
	}

1810
	skb = tcp_write_queue_head(sk);
1811
	state.fack_count = 0;
1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822
	i = 0;

	if (!tp->sacked_out) {
		/* It's already past, so skip checking against it */
		cache = tp->recv_sack_cache + ARRAY_SIZE(tp->recv_sack_cache);
	} else {
		cache = tp->recv_sack_cache;
		/* Skip empty blocks in at head of the cache */
		while (tcp_sack_cache_ok(tp, cache) && !cache->start_seq &&
		       !cache->end_seq)
			cache++;
1823 1824
	}

1825
	while (i < used_sacks) {
1826 1827
		u32 start_seq = sp[i].start_seq;
		u32 end_seq = sp[i].end_seq;
1828
		int dup_sack = (found_dup_sack && (i == first_sack_index));
1829
		struct tcp_sack_block *next_dup = NULL;
1830

1831 1832
		if (found_dup_sack && ((i + 1) == first_sack_index))
			next_dup = &sp[i + 1];
L
Linus Torvalds 已提交
1833 1834 1835

		/* Event "B" in the comment above. */
		if (after(end_seq, tp->high_seq))
1836
			state.flag |= FLAG_DATA_LOST;
L
Linus Torvalds 已提交
1837

1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848
		/* Skip too early cached blocks */
		while (tcp_sack_cache_ok(tp, cache) &&
		       !before(start_seq, cache->end_seq))
			cache++;

		/* Can skip some work by looking recv_sack_cache? */
		if (tcp_sack_cache_ok(tp, cache) && !dup_sack &&
		    after(end_seq, cache->start_seq)) {

			/* Head todo? */
			if (before(start_seq, cache->start_seq)) {
1849 1850
				skb = tcp_sacktag_skip(skb, sk, &state,
						       start_seq);
1851
				skb = tcp_sacktag_walk(skb, sk, next_dup,
1852
						       &state,
1853 1854
						       start_seq,
						       cache->start_seq,
1855
						       dup_sack);
1856
			}
1857

1858
			/* Rest of the block already fully processed? */
1859
			if (!after(end_seq, cache->end_seq))
1860
				goto advance_sp;
1861

1862
			skb = tcp_maybe_skipping_dsack(skb, sk, next_dup,
1863 1864
						       &state,
						       cache->end_seq);
1865

1866
			/* ...tail remains todo... */
1867
			if (tcp_highest_sack_seq(tp) == cache->end_seq) {
1868
				/* ...but better entrypoint exists! */
1869 1870 1871
				skb = tcp_highest_sack(sk);
				if (skb == NULL)
					break;
1872
				state.fack_count = tp->fackets_out;
1873 1874
				cache++;
				goto walk;
1875 1876
			}

1877
			skb = tcp_sacktag_skip(skb, sk, &state, cache->end_seq);
1878 1879 1880 1881
			/* Check overlap against next cached too (past this one already) */
			cache++;
			continue;
		}
L
Linus Torvalds 已提交
1882

1883 1884 1885 1886
		if (!before(start_seq, tcp_highest_sack_seq(tp))) {
			skb = tcp_highest_sack(sk);
			if (skb == NULL)
				break;
1887
			state.fack_count = tp->fackets_out;
L
Linus Torvalds 已提交
1888
		}
1889
		skb = tcp_sacktag_skip(skb, sk, &state, start_seq);
1890 1891

walk:
1892 1893
		skb = tcp_sacktag_walk(skb, sk, next_dup, &state,
				       start_seq, end_seq, dup_sack);
1894

1895
advance_sp:
1896 1897 1898 1899
		/* SACK enhanced FRTO (RFC4138, Appendix B): Clearing correct
		 * due to in-order walk
		 */
		if (after(end_seq, tp->frto_highmark))
1900
			state.flag &= ~FLAG_ONLY_ORIG_SACKED;
1901 1902

		i++;
L
Linus Torvalds 已提交
1903 1904
	}

1905 1906 1907 1908 1909 1910 1911 1912
	/* Clear the head of the cache sack blocks so we can skip it next time */
	for (i = 0; i < ARRAY_SIZE(tp->recv_sack_cache) - used_sacks; i++) {
		tp->recv_sack_cache[i].start_seq = 0;
		tp->recv_sack_cache[i].end_seq = 0;
	}
	for (j = 0; j < used_sacks; j++)
		tp->recv_sack_cache[i++] = sp[j];

1913
	tcp_mark_lost_retrans(sk);
L
Linus Torvalds 已提交
1914

1915 1916
	tcp_verify_left_out(tp);

1917
	if ((state.reord < tp->fackets_out) &&
1918
	    ((icsk->icsk_ca_state != TCP_CA_Loss) || tp->undo_marker) &&
1919
	    (!tp->frto_highmark || after(tp->snd_una, tp->frto_highmark)))
1920
		tcp_update_reordering(sk, tp->fackets_out - state.reord, 0);
L
Linus Torvalds 已提交
1921

1922 1923
out:

L
Linus Torvalds 已提交
1924
#if FASTRETRANS_DEBUG > 0
1925 1926 1927 1928
	WARN_ON((int)tp->sacked_out < 0);
	WARN_ON((int)tp->lost_out < 0);
	WARN_ON((int)tp->retrans_out < 0);
	WARN_ON((int)tcp_packets_in_flight(tp) < 0);
L
Linus Torvalds 已提交
1929
#endif
1930
	return state.flag;
L
Linus Torvalds 已提交
1931 1932
}

1933 1934
/* Limits sacked_out so that sum with lost_out isn't ever larger than
 * packets_out. Returns zero if sacked_out adjustement wasn't necessary.
1935
 */
1936
static int tcp_limit_reno_sacked(struct tcp_sock *tp)
1937 1938 1939 1940 1941 1942 1943 1944
{
	u32 holes;

	holes = max(tp->lost_out, 1U);
	holes = min(holes, tp->packets_out);

	if ((tp->sacked_out + holes) > tp->packets_out) {
		tp->sacked_out = tp->packets_out - holes;
1945
		return 1;
1946
	}
1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958
	return 0;
}

/* If we receive more dupacks than we expected counting segments
 * in assumption of absent reordering, interpret this as reordering.
 * The only another reason could be bug in receiver TCP.
 */
static void tcp_check_reno_reordering(struct sock *sk, const int addend)
{
	struct tcp_sock *tp = tcp_sk(sk);
	if (tcp_limit_reno_sacked(tp))
		tcp_update_reordering(sk, tp->packets_out + addend, 0);
1959 1960 1961 1962 1963 1964 1965 1966 1967
}

/* Emulate SACKs for SACKless connection: account for a new dupack. */

static void tcp_add_reno_sack(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	tp->sacked_out++;
	tcp_check_reno_reordering(sk, 0);
1968
	tcp_verify_left_out(tp);
1969 1970 1971 1972 1973 1974 1975 1976 1977 1978
}

/* Account for ACK, ACKing some data in Reno Recovery phase. */

static void tcp_remove_reno_sacks(struct sock *sk, int acked)
{
	struct tcp_sock *tp = tcp_sk(sk);

	if (acked > 0) {
		/* One ACK acked hole. The rest eat duplicate ACKs. */
1979
		if (acked - 1 >= tp->sacked_out)
1980 1981
			tp->sacked_out = 0;
		else
1982
			tp->sacked_out -= acked - 1;
1983 1984
	}
	tcp_check_reno_reordering(sk, acked);
1985
	tcp_verify_left_out(tp);
1986 1987 1988 1989 1990 1991 1992
}

static inline void tcp_reset_reno_sack(struct tcp_sock *tp)
{
	tp->sacked_out = 0;
}

1993 1994 1995 1996 1997
static int tcp_is_sackfrto(const struct tcp_sock *tp)
{
	return (sysctl_tcp_frto == 0x2) && !tcp_is_reno(tp);
}

1998 1999 2000
/* F-RTO can only be used if TCP has never retransmitted anything other than
 * head (SACK enhanced variant from Appendix B of RFC4138 is more robust here)
 */
2001
int tcp_use_frto(struct sock *sk)
2002 2003
{
	const struct tcp_sock *tp = tcp_sk(sk);
2004
	const struct inet_connection_sock *icsk = inet_csk(sk);
2005 2006
	struct sk_buff *skb;

2007
	if (!sysctl_tcp_frto)
2008
		return 0;
2009

2010 2011 2012 2013
	/* MTU probe and F-RTO won't really play nicely along currently */
	if (icsk->icsk_mtup.probe_size)
		return 0;

2014
	if (tcp_is_sackfrto(tp))
I
Ilpo Järvinen 已提交
2015 2016
		return 1;

2017 2018 2019 2020
	/* Avoid expensive walking of rexmit queue if possible */
	if (tp->retrans_out > 1)
		return 0;

2021
	skb = tcp_write_queue_head(sk);
2022 2023
	if (tcp_skb_is_last(sk, skb))
		return 1;
2024 2025 2026 2027
	skb = tcp_write_queue_next(sk, skb);	/* Skips head */
	tcp_for_write_queue_from(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
2028
		if (TCP_SKB_CB(skb)->sacked & TCPCB_RETRANS)
2029 2030
			return 0;
		/* Short-circuit when first non-SACKed skb has been checked */
2031
		if (!(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
2032 2033 2034
			break;
	}
	return 1;
2035 2036
}

2037 2038
/* RTO occurred, but do not yet enter Loss state. Instead, defer RTO
 * recovery a bit and use heuristics in tcp_process_frto() to detect if
2039 2040 2041 2042 2043
 * the RTO was spurious. Only clear SACKED_RETRANS of the head here to
 * keep retrans_out counting accurate (with SACK F-RTO, other than head
 * may still have that bit set); TCPCB_LOST and remaining SACKED_RETRANS
 * bits are handled if the Loss state is really to be entered (in
 * tcp_enter_frto_loss).
2044 2045 2046 2047
 *
 * Do like tcp_enter_loss() would; when RTO expires the second time it
 * does:
 *  "Reduce ssthresh if it has not yet been made inside this window."
L
Linus Torvalds 已提交
2048 2049 2050
 */
void tcp_enter_frto(struct sock *sk)
{
2051
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2052 2053 2054
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;

2055
	if ((!tp->frto_counter && icsk->icsk_ca_state <= TCP_CA_Disorder) ||
2056
	    tp->snd_una == tp->high_seq ||
2057 2058
	    ((icsk->icsk_ca_state == TCP_CA_Loss || tp->frto_counter) &&
	     !icsk->icsk_retransmits)) {
2059
		tp->prior_ssthresh = tcp_current_ssthresh(sk);
2060
		/* Our state is too optimistic in ssthresh() call because cwnd
2061
		 * is not reduced until tcp_enter_frto_loss() when previous F-RTO
2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084
		 * recovery has not yet completed. Pattern would be this: RTO,
		 * Cumulative ACK, RTO (2xRTO for the same segment does not end
		 * up here twice).
		 * RFC4138 should be more specific on what to do, even though
		 * RTO is quite unlikely to occur after the first Cumulative ACK
		 * due to back-off and complexity of triggering events ...
		 */
		if (tp->frto_counter) {
			u32 stored_cwnd;
			stored_cwnd = tp->snd_cwnd;
			tp->snd_cwnd = 2;
			tp->snd_ssthresh = icsk->icsk_ca_ops->ssthresh(sk);
			tp->snd_cwnd = stored_cwnd;
		} else {
			tp->snd_ssthresh = icsk->icsk_ca_ops->ssthresh(sk);
		}
		/* ... in theory, cong.control module could do "any tricks" in
		 * ssthresh(), which means that ca_state, lost bits and lost_out
		 * counter would have to be faked before the call occurs. We
		 * consider that too expensive, unlikely and hacky, so modules
		 * using these in ssthresh() must deal these incompatibility
		 * issues if they receives CA_EVENT_FRTO and frto_counter != 0
		 */
2085
		tcp_ca_event(sk, CA_EVENT_FRTO);
L
Linus Torvalds 已提交
2086 2087 2088 2089 2090
	}

	tp->undo_marker = tp->snd_una;
	tp->undo_retrans = 0;

2091
	skb = tcp_write_queue_head(sk);
2092 2093
	if (TCP_SKB_CB(skb)->sacked & TCPCB_RETRANS)
		tp->undo_marker = 0;
2094
	if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS) {
2095
		TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS;
2096
		tp->retrans_out -= tcp_skb_pcount(skb);
L
Linus Torvalds 已提交
2097
	}
2098
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2099

2100 2101 2102
	/* Too bad if TCP was application limited */
	tp->snd_cwnd = min(tp->snd_cwnd, tcp_packets_in_flight(tp) + 1);

I
Ilpo Järvinen 已提交
2103 2104 2105
	/* Earlier loss recovery underway (see RFC4138; Appendix B).
	 * The last condition is necessary at least in tp->frto_counter case.
	 */
2106
	if (tcp_is_sackfrto(tp) && (tp->frto_counter ||
I
Ilpo Järvinen 已提交
2107 2108 2109 2110 2111 2112
	    ((1 << icsk->icsk_ca_state) & (TCPF_CA_Recovery|TCPF_CA_Loss))) &&
	    after(tp->high_seq, tp->snd_una)) {
		tp->frto_highmark = tp->high_seq;
	} else {
		tp->frto_highmark = tp->snd_nxt;
	}
2113 2114
	tcp_set_ca_state(sk, TCP_CA_Disorder);
	tp->high_seq = tp->snd_nxt;
2115
	tp->frto_counter = 1;
L
Linus Torvalds 已提交
2116 2117 2118 2119 2120 2121
}

/* Enter Loss state after F-RTO was applied. Dupack arrived after RTO,
 * which indicates that we should follow the traditional RTO recovery,
 * i.e. mark everything lost and do go-back-N retransmission.
 */
2122
static void tcp_enter_frto_loss(struct sock *sk, int allowed_segments, int flag)
L
Linus Torvalds 已提交
2123 2124 2125 2126 2127
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;

	tp->lost_out = 0;
2128
	tp->retrans_out = 0;
2129
	if (tcp_is_reno(tp))
2130
		tcp_reset_reno_sack(tp);
L
Linus Torvalds 已提交
2131

2132 2133 2134
	tcp_for_write_queue(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
2135 2136

		TCP_SKB_CB(skb)->sacked &= ~TCPCB_LOST;
2137 2138 2139 2140
		/*
		 * Count the retransmission made on RTO correctly (only when
		 * waiting for the first ACK and did not get it)...
		 */
2141
		if ((tp->frto_counter == 1) && !(flag & FLAG_DATA_ACKED)) {
2142 2143 2144
			/* For some reason this R-bit might get cleared? */
			if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS)
				tp->retrans_out += tcp_skb_pcount(skb);
2145 2146 2147
			/* ...enter this if branch just for the first segment */
			flag |= FLAG_DATA_ACKED;
		} else {
2148 2149
			if (TCP_SKB_CB(skb)->sacked & TCPCB_RETRANS)
				tp->undo_marker = 0;
2150
			TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS;
2151
		}
L
Linus Torvalds 已提交
2152

2153 2154 2155 2156 2157 2158 2159 2160 2161 2162
		/* Marking forward transmissions that were made after RTO lost
		 * can cause unnecessary retransmissions in some scenarios,
		 * SACK blocks will mitigate that in some but not in all cases.
		 * We used to not mark them but it was causing break-ups with
		 * receivers that do only in-order receival.
		 *
		 * TODO: we could detect presence of such receiver and select
		 * different behavior per flow.
		 */
		if (!(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)) {
2163 2164
			TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
			tp->lost_out += tcp_skb_pcount(skb);
2165
			tp->retransmit_high = TCP_SKB_CB(skb)->end_seq;
L
Linus Torvalds 已提交
2166 2167
		}
	}
2168
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2169

2170
	tp->snd_cwnd = tcp_packets_in_flight(tp) + allowed_segments;
L
Linus Torvalds 已提交
2171 2172 2173
	tp->snd_cwnd_cnt = 0;
	tp->snd_cwnd_stamp = tcp_time_stamp;
	tp->frto_counter = 0;
2174
	tp->bytes_acked = 0;
L
Linus Torvalds 已提交
2175 2176

	tp->reordering = min_t(unsigned int, tp->reordering,
2177
			       sysctl_tcp_reordering);
2178
	tcp_set_ca_state(sk, TCP_CA_Loss);
2179
	tp->high_seq = tp->snd_nxt;
L
Linus Torvalds 已提交
2180
	TCP_ECN_queue_cwr(tp);
2181

2182
	tcp_clear_all_retrans_hints(tp);
L
Linus Torvalds 已提交
2183 2184
}

2185
static void tcp_clear_retrans_partial(struct tcp_sock *tp)
L
Linus Torvalds 已提交
2186 2187 2188 2189 2190 2191 2192 2193
{
	tp->retrans_out = 0;
	tp->lost_out = 0;

	tp->undo_marker = 0;
	tp->undo_retrans = 0;
}

2194 2195 2196 2197 2198 2199 2200 2201
void tcp_clear_retrans(struct tcp_sock *tp)
{
	tcp_clear_retrans_partial(tp);

	tp->fackets_out = 0;
	tp->sacked_out = 0;
}

L
Linus Torvalds 已提交
2202 2203 2204 2205 2206 2207
/* Enter Loss state. If "how" is not zero, forget all SACK information
 * and reset tags completely, otherwise preserve SACKs. If receiver
 * dropped its ofo queue, we will know this due to reneging detection.
 */
void tcp_enter_loss(struct sock *sk, int how)
{
2208
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2209 2210 2211 2212
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;

	/* Reduce ssthresh if it has not yet been made inside this window. */
2213 2214 2215 2216 2217
	if (icsk->icsk_ca_state <= TCP_CA_Disorder || tp->snd_una == tp->high_seq ||
	    (icsk->icsk_ca_state == TCP_CA_Loss && !icsk->icsk_retransmits)) {
		tp->prior_ssthresh = tcp_current_ssthresh(sk);
		tp->snd_ssthresh = icsk->icsk_ca_ops->ssthresh(sk);
		tcp_ca_event(sk, CA_EVENT_LOSS);
L
Linus Torvalds 已提交
2218 2219 2220 2221 2222
	}
	tp->snd_cwnd	   = 1;
	tp->snd_cwnd_cnt   = 0;
	tp->snd_cwnd_stamp = tcp_time_stamp;

2223
	tp->bytes_acked = 0;
2224 2225 2226 2227
	tcp_clear_retrans_partial(tp);

	if (tcp_is_reno(tp))
		tcp_reset_reno_sack(tp);
L
Linus Torvalds 已提交
2228

2229 2230 2231
	if (!how) {
		/* Push undo marker, if it was plain RTO and nothing
		 * was retransmitted. */
L
Linus Torvalds 已提交
2232
		tp->undo_marker = tp->snd_una;
2233
	} else {
2234 2235
		tp->sacked_out = 0;
		tp->fackets_out = 0;
2236
	}
2237
	tcp_clear_all_retrans_hints(tp);
L
Linus Torvalds 已提交
2238

2239 2240 2241
	tcp_for_write_queue(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
2242

2243
		if (TCP_SKB_CB(skb)->sacked & TCPCB_RETRANS)
L
Linus Torvalds 已提交
2244 2245 2246 2247 2248 2249
			tp->undo_marker = 0;
		TCP_SKB_CB(skb)->sacked &= (~TCPCB_TAGBITS)|TCPCB_SACKED_ACKED;
		if (!(TCP_SKB_CB(skb)->sacked&TCPCB_SACKED_ACKED) || how) {
			TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_ACKED;
			TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
			tp->lost_out += tcp_skb_pcount(skb);
2250
			tp->retransmit_high = TCP_SKB_CB(skb)->end_seq;
L
Linus Torvalds 已提交
2251 2252
		}
	}
2253
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2254 2255

	tp->reordering = min_t(unsigned int, tp->reordering,
2256
			       sysctl_tcp_reordering);
2257
	tcp_set_ca_state(sk, TCP_CA_Loss);
L
Linus Torvalds 已提交
2258 2259
	tp->high_seq = tp->snd_nxt;
	TCP_ECN_queue_cwr(tp);
2260
	/* Abort F-RTO algorithm if one is in progress */
2261
	tp->frto_counter = 0;
L
Linus Torvalds 已提交
2262 2263
}

2264 2265 2266 2267 2268 2269 2270
/* If ACK arrived pointing to a remembered SACK, it means that our
 * remembered SACKs do not reflect real state of receiver i.e.
 * receiver _host_ is heavily congested (or buggy).
 *
 * Do processing similar to RTO timeout.
 */
static int tcp_check_sack_reneging(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2271
{
2272
	if (flag & FLAG_SACK_RENEGING) {
2273
		struct inet_connection_sock *icsk = inet_csk(sk);
2274
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPSACKRENEGING);
L
Linus Torvalds 已提交
2275 2276

		tcp_enter_loss(sk, 1);
2277
		icsk->icsk_retransmits++;
2278
		tcp_retransmit_skb(sk, tcp_write_queue_head(sk));
2279
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
2280
					  icsk->icsk_rto, TCP_RTO_MAX);
L
Linus Torvalds 已提交
2281 2282 2283 2284 2285 2286 2287
		return 1;
	}
	return 0;
}

static inline int tcp_fackets_out(struct tcp_sock *tp)
{
2288
	return tcp_is_reno(tp) ? tp->sacked_out + 1 : tp->fackets_out;
L
Linus Torvalds 已提交
2289 2290
}

2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305
/* Heurestics to calculate number of duplicate ACKs. There's no dupACKs
 * counter when SACK is enabled (without SACK, sacked_out is used for
 * that purpose).
 *
 * Instead, with FACK TCP uses fackets_out that includes both SACKed
 * segments up to the highest received SACK block so far and holes in
 * between them.
 *
 * With reordering, holes may still be in flight, so RFC3517 recovery
 * uses pure sacked_out (total number of SACKed segments) even though
 * it violates the RFC that uses duplicate ACKs, often these are equal
 * but when e.g. out-of-window ACKs or packet duplication occurs,
 * they differ. Since neither occurs due to loss, TCP should really
 * ignore them.
 */
2306
static inline int tcp_dupack_heuristics(struct tcp_sock *tp)
2307 2308 2309 2310
{
	return tcp_is_fack(tp) ? tp->fackets_out : tp->sacked_out + 1;
}

2311
static inline int tcp_skb_timedout(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
2312
{
2313
	return (tcp_time_stamp - TCP_SKB_CB(skb)->when > inet_csk(sk)->icsk_rto);
L
Linus Torvalds 已提交
2314 2315
}

2316
static inline int tcp_head_timedout(struct sock *sk)
L
Linus Torvalds 已提交
2317
{
2318 2319
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2320
	return tp->packets_out &&
2321
	       tcp_skb_timedout(sk, tcp_write_queue_head(sk));
L
Linus Torvalds 已提交
2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416
}

/* Linux NewReno/SACK/FACK/ECN state machine.
 * --------------------------------------
 *
 * "Open"	Normal state, no dubious events, fast path.
 * "Disorder"   In all the respects it is "Open",
 *		but requires a bit more attention. It is entered when
 *		we see some SACKs or dupacks. It is split of "Open"
 *		mainly to move some processing from fast path to slow one.
 * "CWR"	CWND was reduced due to some Congestion Notification event.
 *		It can be ECN, ICMP source quench, local device congestion.
 * "Recovery"	CWND was reduced, we are fast-retransmitting.
 * "Loss"	CWND was reduced due to RTO timeout or SACK reneging.
 *
 * tcp_fastretrans_alert() is entered:
 * - each incoming ACK, if state is not "Open"
 * - when arrived ACK is unusual, namely:
 *	* SACK
 *	* Duplicate ACK.
 *	* ECN ECE.
 *
 * Counting packets in flight is pretty simple.
 *
 *	in_flight = packets_out - left_out + retrans_out
 *
 *	packets_out is SND.NXT-SND.UNA counted in packets.
 *
 *	retrans_out is number of retransmitted segments.
 *
 *	left_out is number of segments left network, but not ACKed yet.
 *
 *		left_out = sacked_out + lost_out
 *
 *     sacked_out: Packets, which arrived to receiver out of order
 *		   and hence not ACKed. With SACKs this number is simply
 *		   amount of SACKed data. Even without SACKs
 *		   it is easy to give pretty reliable estimate of this number,
 *		   counting duplicate ACKs.
 *
 *       lost_out: Packets lost by network. TCP has no explicit
 *		   "loss notification" feedback from network (for now).
 *		   It means that this number can be only _guessed_.
 *		   Actually, it is the heuristics to predict lossage that
 *		   distinguishes different algorithms.
 *
 *	F.e. after RTO, when all the queue is considered as lost,
 *	lost_out = packets_out and in_flight = retrans_out.
 *
 *		Essentially, we have now two algorithms counting
 *		lost packets.
 *
 *		FACK: It is the simplest heuristics. As soon as we decided
 *		that something is lost, we decide that _all_ not SACKed
 *		packets until the most forward SACK are lost. I.e.
 *		lost_out = fackets_out - sacked_out and left_out = fackets_out.
 *		It is absolutely correct estimate, if network does not reorder
 *		packets. And it loses any connection to reality when reordering
 *		takes place. We use FACK by default until reordering
 *		is suspected on the path to this destination.
 *
 *		NewReno: when Recovery is entered, we assume that one segment
 *		is lost (classic Reno). While we are in Recovery and
 *		a partial ACK arrives, we assume that one more packet
 *		is lost (NewReno). This heuristics are the same in NewReno
 *		and SACK.
 *
 *  Imagine, that's all! Forget about all this shamanism about CWND inflation
 *  deflation etc. CWND is real congestion window, never inflated, changes
 *  only according to classic VJ rules.
 *
 * Really tricky (and requiring careful tuning) part of algorithm
 * is hidden in functions tcp_time_to_recover() and tcp_xmit_retransmit_queue().
 * The first determines the moment _when_ we should reduce CWND and,
 * hence, slow down forward transmission. In fact, it determines the moment
 * when we decide that hole is caused by loss, rather than by a reorder.
 *
 * tcp_xmit_retransmit_queue() decides, _what_ we should retransmit to fill
 * holes, caused by lost packets.
 *
 * And the most logically complicated part of algorithm is undo
 * heuristics. We detect false retransmits due to both too early
 * fast retransmit (reordering) and underestimated RTO, analyzing
 * timestamps and D-SACKs. When we detect that some segments were
 * retransmitted by mistake and CWND reduction was wrong, we undo
 * window reduction and abort recovery phase. This logic is hidden
 * inside several functions named tcp_try_undo_<something>.
 */

/* This function decides, when we should leave Disordered state
 * and enter Recovery phase, reducing congestion window.
 *
 * Main question: may we further continue forward transmission
 * with the same cwnd?
 */
2417
static int tcp_time_to_recover(struct sock *sk)
L
Linus Torvalds 已提交
2418
{
2419
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2420 2421
	__u32 packets_out;

2422
	/* Do not perform any recovery during F-RTO algorithm */
2423 2424 2425
	if (tp->frto_counter)
		return 0;

L
Linus Torvalds 已提交
2426 2427 2428 2429 2430
	/* Trick#1: The loss is proven. */
	if (tp->lost_out)
		return 1;

	/* Not-A-Trick#2 : Classic rule... */
2431
	if (tcp_dupack_heuristics(tp) > tp->reordering)
L
Linus Torvalds 已提交
2432 2433 2434 2435 2436
		return 1;

	/* Trick#3 : when we use RFC2988 timer restart, fast
	 * retransmit can be triggered by timeout of queue head.
	 */
2437
	if (tcp_is_fack(tp) && tcp_head_timedout(sk))
L
Linus Torvalds 已提交
2438 2439 2440 2441 2442 2443 2444 2445
		return 1;

	/* Trick#4: It is still not OK... But will it be useful to delay
	 * recovery more?
	 */
	packets_out = tp->packets_out;
	if (packets_out <= tp->reordering &&
	    tp->sacked_out >= max_t(__u32, packets_out/2, sysctl_tcp_reordering) &&
2446
	    !tcp_may_send_now(sk)) {
L
Linus Torvalds 已提交
2447 2448 2449 2450 2451 2452
		/* We have nothing to send. This connection is limited
		 * either by receiver window or by application.
		 */
		return 1;
	}

A
Andreas Petlund 已提交
2453 2454 2455 2456 2457 2458 2459 2460 2461 2462
	/* If a thin stream is detected, retransmit after first
	 * received dupack. Employ only if SACK is supported in order
	 * to avoid possible corner-case series of spurious retransmissions
	 * Use only if there are no unsent data.
	 */
	if ((tp->thin_dupack || sysctl_tcp_thin_dupack) &&
	    tcp_stream_is_thin(tp) && tcp_dupack_heuristics(tp) > 1 &&
	    tcp_is_sack(tp) && !tcp_send_head(sk))
		return 1;

L
Linus Torvalds 已提交
2463 2464 2465
	return 0;
}

2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503
/* New heuristics: it is possible only after we switched to restart timer
 * each time when something is ACKed. Hence, we can detect timed out packets
 * during fast retransmit without falling to slow start.
 *
 * Usefulness of this as is very questionable, since we should know which of
 * the segments is the next to timeout which is relatively expensive to find
 * in general case unless we add some data structure just for that. The
 * current approach certainly won't find the right one too often and when it
 * finally does find _something_ it usually marks large part of the window
 * right away (because a retransmission with a larger timestamp blocks the
 * loop from advancing). -ij
 */
static void tcp_timeout_skbs(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;

	if (!tcp_is_fack(tp) || !tcp_head_timedout(sk))
		return;

	skb = tp->scoreboard_skb_hint;
	if (tp->scoreboard_skb_hint == NULL)
		skb = tcp_write_queue_head(sk);

	tcp_for_write_queue_from(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
		if (!tcp_skb_timedout(sk, skb))
			break;

		tcp_skb_mark_lost(tp, skb);
	}

	tp->scoreboard_skb_hint = skb;

	tcp_verify_left_out(tp);
}

2504 2505 2506
/* Mark head of queue up as lost. With RFC3517 SACK, the packets is
 * is against sacked "cnt", otherwise it's against facked "cnt"
 */
2507
static void tcp_mark_head_lost(struct sock *sk, int packets)
L
Linus Torvalds 已提交
2508
{
2509
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2510
	struct sk_buff *skb;
2511 2512 2513
	int cnt, oldcnt;
	int err;
	unsigned int mss;
L
Linus Torvalds 已提交
2514

2515 2516 2517
	if (packets == 0)
		return;

2518
	WARN_ON(packets > tp->packets_out);
2519 2520 2521 2522
	if (tp->lost_skb_hint) {
		skb = tp->lost_skb_hint;
		cnt = tp->lost_cnt_hint;
	} else {
2523
		skb = tcp_write_queue_head(sk);
2524 2525
		cnt = 0;
	}
L
Linus Torvalds 已提交
2526

2527 2528 2529
	tcp_for_write_queue_from(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
2530 2531 2532 2533
		/* TODO: do this better */
		/* this is not the most efficient way to do this... */
		tp->lost_skb_hint = skb;
		tp->lost_cnt_hint = cnt;
2534

2535 2536 2537 2538
		if (after(TCP_SKB_CB(skb)->end_seq, tp->high_seq))
			break;

		oldcnt = cnt;
2539
		if (tcp_is_fack(tp) || tcp_is_reno(tp) ||
2540 2541 2542
		    (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
			cnt += tcp_skb_pcount(skb);

2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553
		if (cnt > packets) {
			if (tcp_is_sack(tp) || (oldcnt >= packets))
				break;

			mss = skb_shinfo(skb)->gso_size;
			err = tcp_fragment(sk, skb, (packets - oldcnt) * mss, mss);
			if (err < 0)
				break;
			cnt = packets;
		}

2554
		tcp_skb_mark_lost(tp, skb);
L
Linus Torvalds 已提交
2555
	}
2556
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2557 2558 2559 2560
}

/* Account newly detected lost packet(s) */

2561
static void tcp_update_scoreboard(struct sock *sk, int fast_rexmit)
L
Linus Torvalds 已提交
2562
{
2563 2564
	struct tcp_sock *tp = tcp_sk(sk);

2565
	if (tcp_is_reno(tp)) {
2566
		tcp_mark_head_lost(sk, 1);
2567
	} else if (tcp_is_fack(tp)) {
L
Linus Torvalds 已提交
2568 2569 2570
		int lost = tp->fackets_out - tp->reordering;
		if (lost <= 0)
			lost = 1;
2571
		tcp_mark_head_lost(sk, lost);
L
Linus Torvalds 已提交
2572
	} else {
2573
		int sacked_upto = tp->sacked_out - tp->reordering;
2574 2575 2576
		if (sacked_upto < fast_rexmit)
			sacked_upto = fast_rexmit;
		tcp_mark_head_lost(sk, sacked_upto);
L
Linus Torvalds 已提交
2577 2578
	}

2579
	tcp_timeout_skbs(sk);
L
Linus Torvalds 已提交
2580 2581 2582 2583 2584 2585 2586 2587
}

/* CWND moderation, preventing bursts due to too big ACKs
 * in dubious situations.
 */
static inline void tcp_moderate_cwnd(struct tcp_sock *tp)
{
	tp->snd_cwnd = min(tp->snd_cwnd,
2588
			   tcp_packets_in_flight(tp) + tcp_max_burst(tp));
L
Linus Torvalds 已提交
2589 2590 2591
	tp->snd_cwnd_stamp = tcp_time_stamp;
}

2592 2593 2594 2595 2596 2597 2598 2599 2600 2601
/* Lower bound on congestion window is slow start threshold
 * unless congestion avoidance choice decides to overide it.
 */
static inline u32 tcp_cwnd_min(const struct sock *sk)
{
	const struct tcp_congestion_ops *ca_ops = inet_csk(sk)->icsk_ca_ops;

	return ca_ops->min_cwnd ? ca_ops->min_cwnd(sk) : tcp_sk(sk)->snd_ssthresh;
}

L
Linus Torvalds 已提交
2602
/* Decrease cwnd each second ack. */
2603
static void tcp_cwnd_down(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2604
{
2605
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2606 2607
	int decr = tp->snd_cwnd_cnt + 1;

2608 2609 2610
	if ((flag & (FLAG_ANY_PROGRESS | FLAG_DSACKING_ACK)) ||
	    (tcp_is_reno(tp) && !(flag & FLAG_NOT_DUP))) {
		tp->snd_cwnd_cnt = decr & 1;
2611
		decr >>= 1;
L
Linus Torvalds 已提交
2612

2613 2614
		if (decr && tp->snd_cwnd > tcp_cwnd_min(sk))
			tp->snd_cwnd -= decr;
L
Linus Torvalds 已提交
2615

2616
		tp->snd_cwnd = min(tp->snd_cwnd, tcp_packets_in_flight(tp) + 1);
2617 2618
		tp->snd_cwnd_stamp = tcp_time_stamp;
	}
L
Linus Torvalds 已提交
2619 2620 2621 2622 2623 2624 2625 2626 2627
}

/* Nothing was retransmitted or returned timestamp is less
 * than timestamp of the first retransmission.
 */
static inline int tcp_packet_delayed(struct tcp_sock *tp)
{
	return !tp->retrans_stamp ||
		(tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr &&
2628
		 before(tp->rx_opt.rcv_tsecr, tp->retrans_stamp));
L
Linus Torvalds 已提交
2629 2630 2631 2632 2633
}

/* Undo procedures. */

#if FASTRETRANS_DEBUG > 1
2634
static void DBGUNDO(struct sock *sk, const char *msg)
L
Linus Torvalds 已提交
2635
{
2636
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2637
	struct inet_sock *inet = inet_sk(sk);
2638

2639
	if (sk->sk_family == AF_INET) {
2640
		printk(KERN_DEBUG "Undo %s %pI4/%u c%u l%u ss%u/%u p%u\n",
2641
		       msg,
2642
		       &inet->inet_daddr, ntohs(inet->inet_dport),
2643 2644 2645 2646 2647 2648 2649
		       tp->snd_cwnd, tcp_left_out(tp),
		       tp->snd_ssthresh, tp->prior_ssthresh,
		       tp->packets_out);
	}
#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
	else if (sk->sk_family == AF_INET6) {
		struct ipv6_pinfo *np = inet6_sk(sk);
H
Harvey Harrison 已提交
2650
		printk(KERN_DEBUG "Undo %s %pI6/%u c%u l%u ss%u/%u p%u\n",
2651
		       msg,
2652
		       &np->daddr, ntohs(inet->inet_dport),
2653 2654 2655 2656 2657
		       tp->snd_cwnd, tcp_left_out(tp),
		       tp->snd_ssthresh, tp->prior_ssthresh,
		       tp->packets_out);
	}
#endif
L
Linus Torvalds 已提交
2658 2659 2660 2661 2662
}
#else
#define DBGUNDO(x...) do { } while (0)
#endif

2663
static void tcp_undo_cwr(struct sock *sk, const int undo)
L
Linus Torvalds 已提交
2664
{
2665 2666
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2667
	if (tp->prior_ssthresh) {
2668 2669 2670 2671
		const struct inet_connection_sock *icsk = inet_csk(sk);

		if (icsk->icsk_ca_ops->undo_cwnd)
			tp->snd_cwnd = icsk->icsk_ca_ops->undo_cwnd(sk);
L
Linus Torvalds 已提交
2672
		else
2673
			tp->snd_cwnd = max(tp->snd_cwnd, tp->snd_ssthresh << 1);
L
Linus Torvalds 已提交
2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687

		if (undo && tp->prior_ssthresh > tp->snd_ssthresh) {
			tp->snd_ssthresh = tp->prior_ssthresh;
			TCP_ECN_withdraw_cwr(tp);
		}
	} else {
		tp->snd_cwnd = max(tp->snd_cwnd, tp->snd_ssthresh);
	}
	tcp_moderate_cwnd(tp);
	tp->snd_cwnd_stamp = tcp_time_stamp;
}

static inline int tcp_may_undo(struct tcp_sock *tp)
{
2688
	return tp->undo_marker && (!tp->undo_retrans || tcp_packet_delayed(tp));
L
Linus Torvalds 已提交
2689 2690 2691
}

/* People celebrate: "We love our President!" */
2692
static int tcp_try_undo_recovery(struct sock *sk)
L
Linus Torvalds 已提交
2693
{
2694 2695
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2696
	if (tcp_may_undo(tp)) {
2697 2698
		int mib_idx;

L
Linus Torvalds 已提交
2699 2700 2701
		/* Happy end! We did not retransmit anything
		 * or our original transmission succeeded.
		 */
2702
		DBGUNDO(sk, inet_csk(sk)->icsk_ca_state == TCP_CA_Loss ? "loss" : "retrans");
2703 2704
		tcp_undo_cwr(sk, 1);
		if (inet_csk(sk)->icsk_ca_state == TCP_CA_Loss)
2705
			mib_idx = LINUX_MIB_TCPLOSSUNDO;
L
Linus Torvalds 已提交
2706
		else
2707 2708
			mib_idx = LINUX_MIB_TCPFULLUNDO;

2709
		NET_INC_STATS_BH(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
2710 2711
		tp->undo_marker = 0;
	}
2712
	if (tp->snd_una == tp->high_seq && tcp_is_reno(tp)) {
L
Linus Torvalds 已提交
2713 2714 2715 2716 2717 2718
		/* Hold old state until something *above* high_seq
		 * is ACKed. For Reno it is MUST to prevent false
		 * fast retransmits (RFC2582). SACK TCP is safe. */
		tcp_moderate_cwnd(tp);
		return 1;
	}
2719
	tcp_set_ca_state(sk, TCP_CA_Open);
L
Linus Torvalds 已提交
2720 2721 2722 2723
	return 0;
}

/* Try to undo cwnd reduction, because D-SACKs acked all retransmitted data */
2724
static void tcp_try_undo_dsack(struct sock *sk)
L
Linus Torvalds 已提交
2725
{
2726 2727
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2728
	if (tp->undo_marker && !tp->undo_retrans) {
2729
		DBGUNDO(sk, "D-SACK");
2730
		tcp_undo_cwr(sk, 1);
L
Linus Torvalds 已提交
2731
		tp->undo_marker = 0;
2732
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPDSACKUNDO);
L
Linus Torvalds 已提交
2733 2734 2735
	}
}

2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764
/* We can clear retrans_stamp when there are no retransmissions in the
 * window. It would seem that it is trivially available for us in
 * tp->retrans_out, however, that kind of assumptions doesn't consider
 * what will happen if errors occur when sending retransmission for the
 * second time. ...It could the that such segment has only
 * TCPCB_EVER_RETRANS set at the present time. It seems that checking
 * the head skb is enough except for some reneging corner cases that
 * are not worth the effort.
 *
 * Main reason for all this complexity is the fact that connection dying
 * time now depends on the validity of the retrans_stamp, in particular,
 * that successive retransmissions of a segment must not advance
 * retrans_stamp under any conditions.
 */
static int tcp_any_retrans_done(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;

	if (tp->retrans_out)
		return 1;

	skb = tcp_write_queue_head(sk);
	if (unlikely(skb && TCP_SKB_CB(skb)->sacked & TCPCB_EVER_RETRANS))
		return 1;

	return 0;
}

L
Linus Torvalds 已提交
2765 2766
/* Undo during fast recovery after partial ACK. */

2767
static int tcp_try_undo_partial(struct sock *sk, int acked)
L
Linus Torvalds 已提交
2768
{
2769
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2770
	/* Partial ACK arrived. Force Hoe's retransmit. */
2771
	int failed = tcp_is_reno(tp) || (tcp_fackets_out(tp) > tp->reordering);
L
Linus Torvalds 已提交
2772 2773 2774 2775 2776

	if (tcp_may_undo(tp)) {
		/* Plain luck! Hole if filled with delayed
		 * packet, rather than with a retransmit.
		 */
2777
		if (!tcp_any_retrans_done(sk))
L
Linus Torvalds 已提交
2778 2779
			tp->retrans_stamp = 0;

2780
		tcp_update_reordering(sk, tcp_fackets_out(tp) + acked, 1);
L
Linus Torvalds 已提交
2781

2782
		DBGUNDO(sk, "Hoe");
2783
		tcp_undo_cwr(sk, 0);
2784
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPPARTIALUNDO);
L
Linus Torvalds 已提交
2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795

		/* So... Do not make Hoe's retransmit yet.
		 * If the first packet was delayed, the rest
		 * ones are most probably delayed as well.
		 */
		failed = 0;
	}
	return failed;
}

/* Undo during loss recovery after partial ACK. */
2796
static int tcp_try_undo_loss(struct sock *sk)
L
Linus Torvalds 已提交
2797
{
2798 2799
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2800 2801
	if (tcp_may_undo(tp)) {
		struct sk_buff *skb;
2802 2803 2804
		tcp_for_write_queue(skb, sk) {
			if (skb == tcp_send_head(sk))
				break;
L
Linus Torvalds 已提交
2805 2806
			TCP_SKB_CB(skb)->sacked &= ~TCPCB_LOST;
		}
2807

2808
		tcp_clear_all_retrans_hints(tp);
2809

2810
		DBGUNDO(sk, "partial loss");
L
Linus Torvalds 已提交
2811
		tp->lost_out = 0;
2812
		tcp_undo_cwr(sk, 1);
2813
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPLOSSUNDO);
2814
		inet_csk(sk)->icsk_retransmits = 0;
L
Linus Torvalds 已提交
2815
		tp->undo_marker = 0;
2816
		if (tcp_is_sack(tp))
2817
			tcp_set_ca_state(sk, TCP_CA_Open);
L
Linus Torvalds 已提交
2818 2819 2820 2821 2822
		return 1;
	}
	return 0;
}

2823
static inline void tcp_complete_cwr(struct sock *sk)
L
Linus Torvalds 已提交
2824
{
2825
	struct tcp_sock *tp = tcp_sk(sk);
2826
	tp->snd_cwnd = min(tp->snd_cwnd, tp->snd_ssthresh);
L
Linus Torvalds 已提交
2827
	tp->snd_cwnd_stamp = tcp_time_stamp;
2828
	tcp_ca_event(sk, CA_EVENT_COMPLETE_CWR);
L
Linus Torvalds 已提交
2829 2830
}

2831 2832 2833 2834 2835
static void tcp_try_keep_open(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	int state = TCP_CA_Open;

2836
	if (tcp_left_out(tp) || tcp_any_retrans_done(sk) || tp->undo_marker)
2837 2838 2839 2840 2841 2842 2843 2844
		state = TCP_CA_Disorder;

	if (inet_csk(sk)->icsk_ca_state != state) {
		tcp_set_ca_state(sk, state);
		tp->high_seq = tp->snd_nxt;
	}
}

2845
static void tcp_try_to_open(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2846
{
2847 2848
	struct tcp_sock *tp = tcp_sk(sk);

2849 2850
	tcp_verify_left_out(tp);

2851
	if (!tp->frto_counter && !tcp_any_retrans_done(sk))
L
Linus Torvalds 已提交
2852 2853
		tp->retrans_stamp = 0;

2854
	if (flag & FLAG_ECE)
2855
		tcp_enter_cwr(sk, 1);
L
Linus Torvalds 已提交
2856

2857
	if (inet_csk(sk)->icsk_ca_state != TCP_CA_CWR) {
2858
		tcp_try_keep_open(sk);
L
Linus Torvalds 已提交
2859 2860
		tcp_moderate_cwnd(tp);
	} else {
2861
		tcp_cwnd_down(sk, flag);
L
Linus Torvalds 已提交
2862 2863 2864
	}
}

J
John Heffner 已提交
2865 2866 2867 2868 2869 2870 2871 2872
static void tcp_mtup_probe_failed(struct sock *sk)
{
	struct inet_connection_sock *icsk = inet_csk(sk);

	icsk->icsk_mtup.search_high = icsk->icsk_mtup.probe_size - 1;
	icsk->icsk_mtup.probe_size = 0;
}

2873
static void tcp_mtup_probe_success(struct sock *sk)
J
John Heffner 已提交
2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);

	/* FIXME: breaks with very large cwnd */
	tp->prior_ssthresh = tcp_current_ssthresh(sk);
	tp->snd_cwnd = tp->snd_cwnd *
		       tcp_mss_to_mtu(sk, tp->mss_cache) /
		       icsk->icsk_mtup.probe_size;
	tp->snd_cwnd_cnt = 0;
	tp->snd_cwnd_stamp = tcp_time_stamp;
	tp->rcv_ssthresh = tcp_current_ssthresh(sk);

	icsk->icsk_mtup.search_low = icsk->icsk_mtup.probe_size;
	icsk->icsk_mtup.probe_size = 0;
	tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
}

2892 2893 2894 2895 2896 2897 2898 2899 2900
/* Do a simple retransmit without using the backoff mechanisms in
 * tcp_timer. This is used for path mtu discovery.
 * The socket is already locked here.
 */
void tcp_simple_retransmit(struct sock *sk)
{
	const struct inet_connection_sock *icsk = inet_csk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
I
Ilpo Järvinen 已提交
2901
	unsigned int mss = tcp_current_mss(sk);
2902 2903 2904 2905 2906
	u32 prior_lost = tp->lost_out;

	tcp_for_write_queue(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
2907
		if (tcp_skb_seglen(skb) > mss &&
2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941
		    !(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)) {
			if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS) {
				TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS;
				tp->retrans_out -= tcp_skb_pcount(skb);
			}
			tcp_skb_mark_lost_uncond_verify(tp, skb);
		}
	}

	tcp_clear_retrans_hints_partial(tp);

	if (prior_lost == tp->lost_out)
		return;

	if (tcp_is_reno(tp))
		tcp_limit_reno_sacked(tp);

	tcp_verify_left_out(tp);

	/* Don't muck with the congestion window here.
	 * Reason is that we do not increase amount of _data_
	 * in network, but units changed and effective
	 * cwnd/ssthresh really reduced now.
	 */
	if (icsk->icsk_ca_state != TCP_CA_Loss) {
		tp->high_seq = tp->snd_nxt;
		tp->snd_ssthresh = tcp_current_ssthresh(sk);
		tp->prior_ssthresh = 0;
		tp->undo_marker = 0;
		tcp_set_ca_state(sk, TCP_CA_Loss);
	}
	tcp_xmit_retransmit_queue(sk);
}

L
Linus Torvalds 已提交
2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952
/* Process an event, which can update packets-in-flight not trivially.
 * Main goal of this function is to calculate new estimate for left_out,
 * taking into account both packets sitting in receiver's buffer and
 * packets lost by network.
 *
 * Besides that it does CWND reduction, when packet loss is detected
 * and changes state of machine.
 *
 * It does _not_ decide what to send, it is made in function
 * tcp_xmit_retransmit_queue().
 */
2953
static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked, int flag)
L
Linus Torvalds 已提交
2954
{
2955
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2956
	struct tcp_sock *tp = tcp_sk(sk);
2957 2958
	int is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP));
	int do_lost = is_dupack || ((flag & FLAG_DATA_SACKED) &&
2959
				    (tcp_fackets_out(tp) > tp->reordering));
2960
	int fast_rexmit = 0, mib_idx;
L
Linus Torvalds 已提交
2961

2962
	if (WARN_ON(!tp->packets_out && tp->sacked_out))
L
Linus Torvalds 已提交
2963
		tp->sacked_out = 0;
I
Ilpo Järvinen 已提交
2964
	if (WARN_ON(!tp->sacked_out && tp->fackets_out))
L
Linus Torvalds 已提交
2965 2966
		tp->fackets_out = 0;

2967
	/* Now state machine starts.
L
Linus Torvalds 已提交
2968
	 * A. ECE, hence prohibit cwnd undoing, the reduction is required. */
2969
	if (flag & FLAG_ECE)
L
Linus Torvalds 已提交
2970 2971 2972
		tp->prior_ssthresh = 0;

	/* B. In all the states check for reneging SACKs. */
2973
	if (tcp_check_sack_reneging(sk, flag))
L
Linus Torvalds 已提交
2974 2975 2976
		return;

	/* C. Process data loss notification, provided it is valid. */
2977
	if (tcp_is_fack(tp) && (flag & FLAG_DATA_LOST) &&
L
Linus Torvalds 已提交
2978
	    before(tp->snd_una, tp->high_seq) &&
2979
	    icsk->icsk_ca_state != TCP_CA_Open &&
L
Linus Torvalds 已提交
2980
	    tp->fackets_out > tp->reordering) {
2981
		tcp_mark_head_lost(sk, tp->fackets_out - tp->reordering);
2982
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPLOSS);
L
Linus Torvalds 已提交
2983 2984
	}

2985 2986
	/* D. Check consistency of the current state. */
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2987 2988 2989

	/* E. Check state exit conditions. State can be terminated
	 *    when high_seq is ACKed. */
2990
	if (icsk->icsk_ca_state == TCP_CA_Open) {
2991
		WARN_ON(tp->retrans_out != 0);
L
Linus Torvalds 已提交
2992 2993
		tp->retrans_stamp = 0;
	} else if (!before(tp->snd_una, tp->high_seq)) {
2994
		switch (icsk->icsk_ca_state) {
L
Linus Torvalds 已提交
2995
		case TCP_CA_Loss:
2996
			icsk->icsk_retransmits = 0;
2997
			if (tcp_try_undo_recovery(sk))
L
Linus Torvalds 已提交
2998 2999 3000 3001 3002 3003 3004
				return;
			break;

		case TCP_CA_CWR:
			/* CWR is to be held something *above* high_seq
			 * is ACKed for CWR bit to reach receiver. */
			if (tp->snd_una != tp->high_seq) {
3005 3006
				tcp_complete_cwr(sk);
				tcp_set_ca_state(sk, TCP_CA_Open);
L
Linus Torvalds 已提交
3007 3008 3009 3010
			}
			break;

		case TCP_CA_Disorder:
3011
			tcp_try_undo_dsack(sk);
L
Linus Torvalds 已提交
3012 3013 3014
			if (!tp->undo_marker ||
			    /* For SACK case do not Open to allow to undo
			     * catching for all duplicate ACKs. */
3015
			    tcp_is_reno(tp) || tp->snd_una != tp->high_seq) {
L
Linus Torvalds 已提交
3016
				tp->undo_marker = 0;
3017
				tcp_set_ca_state(sk, TCP_CA_Open);
L
Linus Torvalds 已提交
3018 3019 3020 3021
			}
			break;

		case TCP_CA_Recovery:
3022
			if (tcp_is_reno(tp))
L
Linus Torvalds 已提交
3023
				tcp_reset_reno_sack(tp);
3024
			if (tcp_try_undo_recovery(sk))
L
Linus Torvalds 已提交
3025
				return;
3026
			tcp_complete_cwr(sk);
L
Linus Torvalds 已提交
3027 3028 3029 3030 3031
			break;
		}
	}

	/* F. Process state. */
3032
	switch (icsk->icsk_ca_state) {
L
Linus Torvalds 已提交
3033
	case TCP_CA_Recovery:
3034
		if (!(flag & FLAG_SND_UNA_ADVANCED)) {
3035
			if (tcp_is_reno(tp) && is_dupack)
3036
				tcp_add_reno_sack(sk);
3037 3038
		} else
			do_lost = tcp_try_undo_partial(sk, pkts_acked);
L
Linus Torvalds 已提交
3039 3040
		break;
	case TCP_CA_Loss:
3041
		if (flag & FLAG_DATA_ACKED)
3042
			icsk->icsk_retransmits = 0;
3043 3044
		if (tcp_is_reno(tp) && flag & FLAG_SND_UNA_ADVANCED)
			tcp_reset_reno_sack(tp);
3045
		if (!tcp_try_undo_loss(sk)) {
L
Linus Torvalds 已提交
3046 3047 3048 3049
			tcp_moderate_cwnd(tp);
			tcp_xmit_retransmit_queue(sk);
			return;
		}
3050
		if (icsk->icsk_ca_state != TCP_CA_Open)
L
Linus Torvalds 已提交
3051 3052 3053
			return;
		/* Loss is undone; fall through to processing in Open state. */
	default:
3054
		if (tcp_is_reno(tp)) {
3055
			if (flag & FLAG_SND_UNA_ADVANCED)
L
Linus Torvalds 已提交
3056 3057
				tcp_reset_reno_sack(tp);
			if (is_dupack)
3058
				tcp_add_reno_sack(sk);
L
Linus Torvalds 已提交
3059 3060
		}

3061
		if (icsk->icsk_ca_state == TCP_CA_Disorder)
3062
			tcp_try_undo_dsack(sk);
L
Linus Torvalds 已提交
3063

3064 3065
		if (!tcp_time_to_recover(sk)) {
			tcp_try_to_open(sk, flag);
L
Linus Torvalds 已提交
3066 3067 3068
			return;
		}

J
John Heffner 已提交
3069 3070 3071
		/* MTU probe failure: don't reduce cwnd */
		if (icsk->icsk_ca_state < TCP_CA_CWR &&
		    icsk->icsk_mtup.probe_size &&
3072
		    tp->snd_una == tp->mtu_probe.probe_seq_start) {
J
John Heffner 已提交
3073 3074 3075 3076 3077 3078 3079
			tcp_mtup_probe_failed(sk);
			/* Restores the reduction we did in tcp_mtup_probe() */
			tp->snd_cwnd++;
			tcp_simple_retransmit(sk);
			return;
		}

L
Linus Torvalds 已提交
3080 3081
		/* Otherwise enter Recovery state */

3082
		if (tcp_is_reno(tp))
3083
			mib_idx = LINUX_MIB_TCPRENORECOVERY;
L
Linus Torvalds 已提交
3084
		else
3085 3086
			mib_idx = LINUX_MIB_TCPSACKRECOVERY;

3087
		NET_INC_STATS_BH(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
3088 3089 3090 3091 3092 3093

		tp->high_seq = tp->snd_nxt;
		tp->prior_ssthresh = 0;
		tp->undo_marker = tp->snd_una;
		tp->undo_retrans = tp->retrans_out;

3094
		if (icsk->icsk_ca_state < TCP_CA_CWR) {
3095
			if (!(flag & FLAG_ECE))
3096 3097
				tp->prior_ssthresh = tcp_current_ssthresh(sk);
			tp->snd_ssthresh = icsk->icsk_ca_ops->ssthresh(sk);
L
Linus Torvalds 已提交
3098 3099 3100
			TCP_ECN_queue_cwr(tp);
		}

3101
		tp->bytes_acked = 0;
L
Linus Torvalds 已提交
3102
		tp->snd_cwnd_cnt = 0;
3103
		tcp_set_ca_state(sk, TCP_CA_Recovery);
3104
		fast_rexmit = 1;
L
Linus Torvalds 已提交
3105 3106
	}

3107 3108
	if (do_lost || (tcp_is_fack(tp) && tcp_head_timedout(sk)))
		tcp_update_scoreboard(sk, fast_rexmit);
3109
	tcp_cwnd_down(sk, flag);
L
Linus Torvalds 已提交
3110 3111 3112
	tcp_xmit_retransmit_queue(sk);
}

3113 3114 3115 3116 3117 3118 3119
static void tcp_valid_rtt_meas(struct sock *sk, u32 seq_rtt)
{
	tcp_rtt_estimator(sk, seq_rtt);
	tcp_set_rto(sk);
	inet_csk(sk)->icsk_backoff = 0;
}

L
Linus Torvalds 已提交
3120
/* Read draft-ietf-tcplw-high-performance before mucking
S
Stephen Hemminger 已提交
3121
 * with this code. (Supersedes RFC1323)
L
Linus Torvalds 已提交
3122
 */
3123
static void tcp_ack_saw_tstamp(struct sock *sk, int flag)
L
Linus Torvalds 已提交
3124 3125 3126 3127 3128 3129 3130 3131 3132 3133
{
	/* RTTM Rule: A TSecr value received in a segment is used to
	 * update the averaged RTT measurement only if the segment
	 * acknowledges some new data, i.e., only if it advances the
	 * left edge of the send window.
	 *
	 * See draft-ietf-tcplw-high-performance-00, section 3.3.
	 * 1998/04/10 Andrey V. Savochkin <saw@msu.ru>
	 *
	 * Changed: reset backoff as soon as we see the first valid sample.
S
Stephen Hemminger 已提交
3134
	 * If we do not, we get strongly overestimated rto. With timestamps
L
Linus Torvalds 已提交
3135 3136 3137 3138 3139
	 * samples are accepted even from very old segments: f.e., when rtt=1
	 * increases to 8, we retransmit 5 times and after 8 seconds delayed
	 * answer arrives rto becomes 120 seconds! If at least one of segments
	 * in window is lost... Voila.	 			--ANK (010210)
	 */
3140
	struct tcp_sock *tp = tcp_sk(sk);
3141 3142

	tcp_valid_rtt_meas(sk, tcp_time_stamp - tp->rx_opt.rcv_tsecr);
L
Linus Torvalds 已提交
3143 3144
}

3145
static void tcp_ack_no_tstamp(struct sock *sk, u32 seq_rtt, int flag)
L
Linus Torvalds 已提交
3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158
{
	/* We don't have a timestamp. Can only use
	 * packets that are not retransmitted to determine
	 * rtt estimates. Also, we must not reset the
	 * backoff for rto until we get a non-retransmitted
	 * packet. This allows us to deal with a situation
	 * where the network delay has increased suddenly.
	 * I.e. Karn's algorithm. (SIGCOMM '87, p5.)
	 */

	if (flag & FLAG_RETRANS_DATA_ACKED)
		return;

3159
	tcp_valid_rtt_meas(sk, seq_rtt);
L
Linus Torvalds 已提交
3160 3161
}

3162
static inline void tcp_ack_update_rtt(struct sock *sk, const int flag,
3163
				      const s32 seq_rtt)
L
Linus Torvalds 已提交
3164
{
3165
	const struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
3166 3167
	/* Note that peer MAY send zero echo. In this case it is ignored. (rfc1323) */
	if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr)
3168
		tcp_ack_saw_tstamp(sk, flag);
L
Linus Torvalds 已提交
3169
	else if (seq_rtt >= 0)
3170
		tcp_ack_no_tstamp(sk, seq_rtt, flag);
L
Linus Torvalds 已提交
3171 3172
}

3173
static void tcp_cong_avoid(struct sock *sk, u32 ack, u32 in_flight)
L
Linus Torvalds 已提交
3174
{
3175
	const struct inet_connection_sock *icsk = inet_csk(sk);
3176
	icsk->icsk_ca_ops->cong_avoid(sk, ack, in_flight);
3177
	tcp_sk(sk)->snd_cwnd_stamp = tcp_time_stamp;
L
Linus Torvalds 已提交
3178 3179 3180 3181 3182
}

/* Restart timer after forward progress on connection.
 * RFC2988 recommends to restart timer to now+rto.
 */
3183
static void tcp_rearm_rto(struct sock *sk)
L
Linus Torvalds 已提交
3184
{
3185 3186
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
3187
	if (!tp->packets_out) {
3188
		inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS);
L
Linus Torvalds 已提交
3189
	} else {
3190 3191
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
					  inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
L
Linus Torvalds 已提交
3192 3193 3194
	}
}

3195
/* If we get here, the whole TSO packet has not been acked. */
3196
static u32 tcp_tso_acked(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
3197 3198
{
	struct tcp_sock *tp = tcp_sk(sk);
3199
	u32 packets_acked;
L
Linus Torvalds 已提交
3200

3201
	BUG_ON(!after(TCP_SKB_CB(skb)->end_seq, tp->snd_una));
L
Linus Torvalds 已提交
3202 3203

	packets_acked = tcp_skb_pcount(skb);
3204
	if (tcp_trim_head(sk, skb, tp->snd_una - TCP_SKB_CB(skb)->seq))
L
Linus Torvalds 已提交
3205 3206 3207 3208 3209
		return 0;
	packets_acked -= tcp_skb_pcount(skb);

	if (packets_acked) {
		BUG_ON(tcp_skb_pcount(skb) == 0);
3210
		BUG_ON(!before(TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq));
L
Linus Torvalds 已提交
3211 3212
	}

3213
	return packets_acked;
L
Linus Torvalds 已提交
3214 3215
}

3216 3217 3218 3219
/* Remove acknowledged frames from the retransmission queue. If our packet
 * is before the ack sequence we can discard it as it's confirmed to have
 * arrived at the other end.
 */
I
Ilpo Järvinen 已提交
3220 3221
static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets,
			       u32 prior_snd_una)
L
Linus Torvalds 已提交
3222 3223
{
	struct tcp_sock *tp = tcp_sk(sk);
3224
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3225
	struct sk_buff *skb;
3226
	u32 now = tcp_time_stamp;
3227
	int fully_acked = 1;
3228
	int flag = 0;
3229
	u32 pkts_acked = 0;
3230
	u32 reord = tp->packets_out;
3231
	u32 prior_sacked = tp->sacked_out;
3232
	s32 seq_rtt = -1;
3233
	s32 ca_seq_rtt = -1;
3234
	ktime_t last_ackt = net_invalid_timestamp();
L
Linus Torvalds 已提交
3235

3236
	while ((skb = tcp_write_queue_head(sk)) && skb != tcp_send_head(sk)) {
3237
		struct tcp_skb_cb *scb = TCP_SKB_CB(skb);
3238
		u32 acked_pcount;
3239
		u8 sacked = scb->sacked;
L
Linus Torvalds 已提交
3240

3241
		/* Determine how many packets and what bytes were acked, tso and else */
L
Linus Torvalds 已提交
3242
		if (after(scb->end_seq, tp->snd_una)) {
3243 3244 3245 3246
			if (tcp_skb_pcount(skb) == 1 ||
			    !after(tp->snd_una, scb->seq))
				break;

3247 3248
			acked_pcount = tcp_tso_acked(sk, skb);
			if (!acked_pcount)
3249 3250 3251 3252
				break;

			fully_acked = 0;
		} else {
3253
			acked_pcount = tcp_skb_pcount(skb);
L
Linus Torvalds 已提交
3254 3255
		}

3256 3257
		if (sacked & TCPCB_RETRANS) {
			if (sacked & TCPCB_SACKED_RETRANS)
3258
				tp->retrans_out -= acked_pcount;
3259 3260 3261
			flag |= FLAG_RETRANS_DATA_ACKED;
			ca_seq_rtt = -1;
			seq_rtt = -1;
3262
			if ((flag & FLAG_DATA_ACKED) || (acked_pcount > 1))
3263
				flag |= FLAG_NONHEAD_RETRANS_ACKED;
3264
		} else {
3265 3266
			ca_seq_rtt = now - scb->when;
			last_ackt = skb->tstamp;
3267
			if (seq_rtt < 0) {
3268
				seq_rtt = ca_seq_rtt;
3269
			}
3270
			if (!(sacked & TCPCB_SACKED_ACKED))
3271
				reord = min(pkts_acked, reord);
3272
		}
3273 3274

		if (sacked & TCPCB_SACKED_ACKED)
3275
			tp->sacked_out -= acked_pcount;
3276
		if (sacked & TCPCB_LOST)
3277
			tp->lost_out -= acked_pcount;
3278

3279 3280
		tp->packets_out -= acked_pcount;
		pkts_acked += acked_pcount;
3281

3282 3283 3284 3285 3286 3287 3288
		/* Initial outgoing SYN's get put onto the write_queue
		 * just like anything else we transmit.  It is not
		 * true data, and if we misinform our callers that
		 * this ACK acks real data, we will erroneously exit
		 * connection startup slow start one packet too
		 * quickly.  This is severely frowned upon behavior.
		 */
C
Changli Gao 已提交
3289
		if (!(scb->flags & TCPHDR_SYN)) {
3290 3291 3292 3293 3294 3295
			flag |= FLAG_DATA_ACKED;
		} else {
			flag |= FLAG_SYN_ACKED;
			tp->retrans_stamp = 0;
		}

3296 3297 3298
		if (!fully_acked)
			break;

3299
		tcp_unlink_write_queue(skb, sk);
3300
		sk_wmem_free_skb(sk, skb);
3301
		tp->scoreboard_skb_hint = NULL;
3302 3303
		if (skb == tp->retransmit_skb_hint)
			tp->retransmit_skb_hint = NULL;
3304 3305
		if (skb == tp->lost_skb_hint)
			tp->lost_skb_hint = NULL;
L
Linus Torvalds 已提交
3306 3307
	}

I
Ilpo Järvinen 已提交
3308 3309 3310
	if (likely(between(tp->snd_up, prior_snd_una, tp->snd_una)))
		tp->snd_up = tp->snd_una;

3311 3312 3313
	if (skb && (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
		flag |= FLAG_SACK_RENEGING;

3314
	if (flag & FLAG_ACKED) {
3315 3316 3317
		const struct tcp_congestion_ops *ca_ops
			= inet_csk(sk)->icsk_ca_ops;

3318 3319 3320 3321 3322
		if (unlikely(icsk->icsk_mtup.probe_size &&
			     !after(tp->mtu_probe.probe_seq_end, tp->snd_una))) {
			tcp_mtup_probe_success(sk);
		}

3323
		tcp_ack_update_rtt(sk, flag, seq_rtt);
3324
		tcp_rearm_rto(sk);
3325

3326 3327 3328
		if (tcp_is_reno(tp)) {
			tcp_remove_reno_sacks(sk, pkts_acked);
		} else {
3329 3330
			int delta;

3331 3332 3333
			/* Non-retransmitted hole got filled? That's reordering */
			if (reord < prior_fackets)
				tcp_update_reordering(sk, tp->fackets_out - reord, 0);
3334

3335 3336 3337
			delta = tcp_is_fack(tp) ? pkts_acked :
						  prior_sacked - tp->sacked_out;
			tp->lost_cnt_hint -= min(tp->lost_cnt_hint, delta);
3338 3339
		}

I
Ilpo Järvinen 已提交
3340
		tp->fackets_out -= min(pkts_acked, tp->fackets_out);
3341

3342 3343 3344 3345
		if (ca_ops->pkts_acked) {
			s32 rtt_us = -1;

			/* Is the ACK triggering packet unambiguous? */
3346
			if (!(flag & FLAG_RETRANS_DATA_ACKED)) {
3347 3348 3349 3350 3351 3352
				/* High resolution needed and available? */
				if (ca_ops->flags & TCP_CONG_RTT_STAMP &&
				    !ktime_equal(last_ackt,
						 net_invalid_timestamp()))
					rtt_us = ktime_us_delta(ktime_get_real(),
								last_ackt);
3353 3354
				else if (ca_seq_rtt > 0)
					rtt_us = jiffies_to_usecs(ca_seq_rtt);
3355
			}
3356

3357 3358
			ca_ops->pkts_acked(sk, pkts_acked, rtt_us);
		}
L
Linus Torvalds 已提交
3359 3360 3361
	}

#if FASTRETRANS_DEBUG > 0
3362 3363 3364
	WARN_ON((int)tp->sacked_out < 0);
	WARN_ON((int)tp->lost_out < 0);
	WARN_ON((int)tp->retrans_out < 0);
3365
	if (!tp->packets_out && tcp_is_sack(tp)) {
S
Stephen Hemminger 已提交
3366
		icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3367 3368
		if (tp->lost_out) {
			printk(KERN_DEBUG "Leak l=%u %d\n",
3369
			       tp->lost_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3370 3371 3372 3373
			tp->lost_out = 0;
		}
		if (tp->sacked_out) {
			printk(KERN_DEBUG "Leak s=%u %d\n",
3374
			       tp->sacked_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3375 3376 3377 3378
			tp->sacked_out = 0;
		}
		if (tp->retrans_out) {
			printk(KERN_DEBUG "Leak r=%u %d\n",
3379
			       tp->retrans_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3380 3381 3382 3383
			tp->retrans_out = 0;
		}
	}
#endif
3384
	return flag;
L
Linus Torvalds 已提交
3385 3386 3387 3388
}

static void tcp_ack_probe(struct sock *sk)
{
3389 3390
	const struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3391 3392 3393

	/* Was it a usable window open? */

3394
	if (!after(TCP_SKB_CB(tcp_send_head(sk))->end_seq, tcp_wnd_end(tp))) {
3395 3396
		icsk->icsk_backoff = 0;
		inet_csk_clear_xmit_timer(sk, ICSK_TIME_PROBE0);
L
Linus Torvalds 已提交
3397 3398 3399 3400
		/* Socket must be waked up by subsequent tcp_data_snd_check().
		 * This function is not for random using!
		 */
	} else {
3401
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
3402 3403
					  min(icsk->icsk_rto << icsk->icsk_backoff, TCP_RTO_MAX),
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
3404 3405 3406
	}
}

3407
static inline int tcp_ack_is_dubious(const struct sock *sk, const int flag)
L
Linus Torvalds 已提交
3408 3409
{
	return (!(flag & FLAG_NOT_DUP) || (flag & FLAG_CA_ALERT) ||
3410
		inet_csk(sk)->icsk_ca_state != TCP_CA_Open);
L
Linus Torvalds 已提交
3411 3412
}

3413
static inline int tcp_may_raise_cwnd(const struct sock *sk, const int flag)
L
Linus Torvalds 已提交
3414
{
3415
	const struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
3416
	return (!(flag & FLAG_ECE) || tp->snd_cwnd < tp->snd_ssthresh) &&
3417
		!((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_Recovery | TCPF_CA_CWR));
L
Linus Torvalds 已提交
3418 3419 3420 3421 3422
}

/* Check that window update is acceptable.
 * The function assumes that snd_una<=ack<=snd_next.
 */
3423 3424 3425
static inline int tcp_may_update_window(const struct tcp_sock *tp,
					const u32 ack, const u32 ack_seq,
					const u32 nwin)
L
Linus Torvalds 已提交
3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436
{
	return (after(ack, tp->snd_una) ||
		after(ack_seq, tp->snd_wl1) ||
		(ack_seq == tp->snd_wl1 && nwin > tp->snd_wnd));
}

/* Update our send window.
 *
 * Window update algorithm, described in RFC793/RFC1122 (used in linux-2.2
 * and in FreeBSD. NetBSD's one is even worse.) is wrong.
 */
3437 3438
static int tcp_ack_update_window(struct sock *sk, struct sk_buff *skb, u32 ack,
				 u32 ack_seq)
L
Linus Torvalds 已提交
3439
{
3440
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
3441
	int flag = 0;
3442
	u32 nwin = ntohs(tcp_hdr(skb)->window);
L
Linus Torvalds 已提交
3443

3444
	if (likely(!tcp_hdr(skb)->syn))
L
Linus Torvalds 已提交
3445 3446 3447 3448
		nwin <<= tp->rx_opt.snd_wscale;

	if (tcp_may_update_window(tp, ack, ack_seq, nwin)) {
		flag |= FLAG_WIN_UPDATE;
3449
		tcp_update_wl(tp, ack_seq);
L
Linus Torvalds 已提交
3450 3451 3452 3453 3454 3455 3456

		if (tp->snd_wnd != nwin) {
			tp->snd_wnd = nwin;

			/* Note, it is the only place, where
			 * fast path is recovered for sending TCP.
			 */
3457
			tp->pred_flags = 0;
3458
			tcp_fast_path_check(sk);
L
Linus Torvalds 已提交
3459 3460 3461

			if (nwin > tp->max_window) {
				tp->max_window = nwin;
3462
				tcp_sync_mss(sk, inet_csk(sk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
3463 3464 3465 3466 3467 3468 3469 3470 3471
			}
		}
	}

	tp->snd_una = ack;

	return flag;
}

3472 3473 3474 3475 3476 3477
/* A very conservative spurious RTO response algorithm: reduce cwnd and
 * continue in congestion avoidance.
 */
static void tcp_conservative_spur_to_response(struct tcp_sock *tp)
{
	tp->snd_cwnd = min(tp->snd_cwnd, tp->snd_ssthresh);
3478
	tp->snd_cwnd_cnt = 0;
3479
	tp->bytes_acked = 0;
3480
	TCP_ECN_queue_cwr(tp);
3481 3482 3483
	tcp_moderate_cwnd(tp);
}

3484 3485 3486 3487 3488 3489 3490 3491
/* A conservative spurious RTO response algorithm: reduce cwnd using
 * rate halving and continue in congestion avoidance.
 */
static void tcp_ratehalving_spur_to_response(struct sock *sk)
{
	tcp_enter_cwr(sk, 0);
}

3492
static void tcp_undo_spur_to_response(struct sock *sk, int flag)
3493
{
3494
	if (flag & FLAG_ECE)
3495 3496 3497
		tcp_ratehalving_spur_to_response(sk);
	else
		tcp_undo_cwr(sk, 1);
3498 3499
}

3500 3501
/* F-RTO spurious RTO detection algorithm (RFC4138)
 *
3502 3503 3504
 * F-RTO affects during two new ACKs following RTO (well, almost, see inline
 * comments). State (ACK number) is kept in frto_counter. When ACK advances
 * window (but not to or beyond highest sequence sent before RTO):
3505 3506 3507 3508 3509
 *   On First ACK,  send two new segments out.
 *   On Second ACK, RTO was likely spurious. Do spurious response (response
 *                  algorithm is not part of the F-RTO detection algorithm
 *                  given in RFC4138 but can be selected separately).
 * Otherwise (basically on duplicate ACK), RTO was (likely) caused by a loss
3510 3511 3512
 * and TCP falls back to conventional RTO recovery. F-RTO allows overriding
 * of Nagle, this is done using frto_counter states 2 and 3, when a new data
 * segment of any size sent during F-RTO, state 2 is upgraded to 3.
3513 3514 3515 3516
 *
 * Rationale: if the RTO was spurious, new ACKs should arrive from the
 * original window even after we transmit two new data segments.
 *
I
Ilpo Järvinen 已提交
3517 3518 3519 3520
 * SACK version:
 *   on first step, wait until first cumulative ACK arrives, then move to
 *   the second step. In second step, the next ACK decides.
 *
3521 3522 3523 3524 3525 3526 3527 3528 3529
 * F-RTO is implemented (mainly) in four functions:
 *   - tcp_use_frto() is used to determine if TCP is can use F-RTO
 *   - tcp_enter_frto() prepares TCP state on RTO if F-RTO is used, it is
 *     called when tcp_use_frto() showed green light
 *   - tcp_process_frto() handles incoming ACKs during F-RTO algorithm
 *   - tcp_enter_frto_loss() is called if there is not enough evidence
 *     to prove that the RTO is indeed spurious. It transfers the control
 *     from F-RTO to the conventional RTO recovery
 */
3530
static int tcp_process_frto(struct sock *sk, int flag)
L
Linus Torvalds 已提交
3531 3532
{
	struct tcp_sock *tp = tcp_sk(sk);
3533

3534
	tcp_verify_left_out(tp);
3535

3536
	/* Duplicate the behavior from Loss state (fastretrans_alert) */
3537
	if (flag & FLAG_DATA_ACKED)
3538 3539
		inet_csk(sk)->icsk_retransmits = 0;

3540 3541 3542 3543
	if ((flag & FLAG_NONHEAD_RETRANS_ACKED) ||
	    ((tp->frto_counter >= 2) && (flag & FLAG_RETRANS_DATA_ACKED)))
		tp->undo_marker = 0;

3544
	if (!before(tp->snd_una, tp->frto_highmark)) {
3545
		tcp_enter_frto_loss(sk, (tp->frto_counter == 1 ? 2 : 3), flag);
3546
		return 1;
3547 3548
	}

3549
	if (!tcp_is_sackfrto(tp)) {
I
Ilpo Järvinen 已提交
3550 3551 3552 3553
		/* RFC4138 shortcoming in step 2; should also have case c):
		 * ACK isn't duplicate nor advances window, e.g., opposite dir
		 * data, winupdate
		 */
3554
		if (!(flag & FLAG_ANY_PROGRESS) && (flag & FLAG_NOT_DUP))
I
Ilpo Järvinen 已提交
3555 3556
			return 1;

3557
		if (!(flag & FLAG_DATA_ACKED)) {
I
Ilpo Järvinen 已提交
3558 3559 3560 3561 3562
			tcp_enter_frto_loss(sk, (tp->frto_counter == 1 ? 0 : 3),
					    flag);
			return 1;
		}
	} else {
3563
		if (!(flag & FLAG_DATA_ACKED) && (tp->frto_counter == 1)) {
I
Ilpo Järvinen 已提交
3564 3565 3566 3567 3568
			/* Prevent sending of new data. */
			tp->snd_cwnd = min(tp->snd_cwnd,
					   tcp_packets_in_flight(tp));
			return 1;
		}
3569

3570
		if ((tp->frto_counter >= 2) &&
3571 3572 3573
		    (!(flag & FLAG_FORWARD_PROGRESS) ||
		     ((flag & FLAG_DATA_SACKED) &&
		      !(flag & FLAG_ONLY_ORIG_SACKED)))) {
I
Ilpo Järvinen 已提交
3574
			/* RFC4138 shortcoming (see comment above) */
3575 3576
			if (!(flag & FLAG_FORWARD_PROGRESS) &&
			    (flag & FLAG_NOT_DUP))
I
Ilpo Järvinen 已提交
3577 3578 3579 3580 3581
				return 1;

			tcp_enter_frto_loss(sk, 3, flag);
			return 1;
		}
L
Linus Torvalds 已提交
3582 3583 3584
	}

	if (tp->frto_counter == 1) {
3585
		/* tcp_may_send_now needs to see updated state */
L
Linus Torvalds 已提交
3586
		tp->snd_cwnd = tcp_packets_in_flight(tp) + 2;
3587
		tp->frto_counter = 2;
3588 3589 3590 3591

		if (!tcp_may_send_now(sk))
			tcp_enter_frto_loss(sk, 2, flag);

3592
		return 1;
3593
	} else {
3594 3595
		switch (sysctl_tcp_frto_response) {
		case 2:
3596
			tcp_undo_spur_to_response(sk, flag);
3597 3598 3599 3600 3601 3602 3603
			break;
		case 1:
			tcp_conservative_spur_to_response(tp);
			break;
		default:
			tcp_ratehalving_spur_to_response(sk);
			break;
3604
		}
3605
		tp->frto_counter = 0;
3606
		tp->undo_marker = 0;
3607
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPSPURIOUSRTOS);
L
Linus Torvalds 已提交
3608
	}
3609
	return 0;
L
Linus Torvalds 已提交
3610 3611 3612 3613 3614
}

/* This routine deals with incoming acks, but not outgoing ones. */
static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag)
{
3615
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3616 3617 3618 3619 3620
	struct tcp_sock *tp = tcp_sk(sk);
	u32 prior_snd_una = tp->snd_una;
	u32 ack_seq = TCP_SKB_CB(skb)->seq;
	u32 ack = TCP_SKB_CB(skb)->ack_seq;
	u32 prior_in_flight;
3621
	u32 prior_fackets;
L
Linus Torvalds 已提交
3622
	int prior_packets;
3623
	int frto_cwnd = 0;
L
Linus Torvalds 已提交
3624

3625
	/* If the ack is older than previous acks
L
Linus Torvalds 已提交
3626 3627 3628 3629 3630
	 * then we can probably ignore it.
	 */
	if (before(ack, prior_snd_una))
		goto old_ack;

3631 3632 3633 3634 3635 3636
	/* If the ack includes data we haven't sent yet, discard
	 * this segment (RFC793 Section 3.9).
	 */
	if (after(ack, tp->snd_nxt))
		goto invalid_ack;

3637 3638 3639
	if (after(ack, prior_snd_una))
		flag |= FLAG_SND_UNA_ADVANCED;

3640 3641 3642 3643 3644
	if (sysctl_tcp_abc) {
		if (icsk->icsk_ca_state < TCP_CA_CWR)
			tp->bytes_acked += ack - prior_snd_una;
		else if (icsk->icsk_ca_state == TCP_CA_Loss)
			/* we assume just one segment left network */
3645 3646
			tp->bytes_acked += min(ack - prior_snd_una,
					       tp->mss_cache);
3647
	}
3648

3649
	prior_fackets = tp->fackets_out;
3650
	prior_in_flight = tcp_packets_in_flight(tp);
3651

3652
	if (!(flag & FLAG_SLOWPATH) && after(ack, prior_snd_una)) {
L
Linus Torvalds 已提交
3653 3654 3655 3656
		/* Window is constant, pure forward advance.
		 * No more checks are required.
		 * Note, we use the fact that SND.UNA>=SND.WL2.
		 */
3657
		tcp_update_wl(tp, ack_seq);
L
Linus Torvalds 已提交
3658 3659 3660
		tp->snd_una = ack;
		flag |= FLAG_WIN_UPDATE;

3661
		tcp_ca_event(sk, CA_EVENT_FAST_ACK);
3662

3663
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPHPACKS);
L
Linus Torvalds 已提交
3664 3665 3666 3667
	} else {
		if (ack_seq != TCP_SKB_CB(skb)->end_seq)
			flag |= FLAG_DATA;
		else
3668
			NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPPUREACKS);
L
Linus Torvalds 已提交
3669

3670
		flag |= tcp_ack_update_window(sk, skb, ack, ack_seq);
L
Linus Torvalds 已提交
3671 3672 3673 3674

		if (TCP_SKB_CB(skb)->sacked)
			flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una);

3675
		if (TCP_ECN_rcv_ecn_echo(tp, tcp_hdr(skb)))
L
Linus Torvalds 已提交
3676 3677
			flag |= FLAG_ECE;

3678
		tcp_ca_event(sk, CA_EVENT_SLOW_ACK);
L
Linus Torvalds 已提交
3679 3680 3681 3682 3683 3684
	}

	/* We passed data and got it acked, remove any soft error
	 * log. Something worked...
	 */
	sk->sk_err_soft = 0;
3685
	icsk->icsk_probes_out = 0;
L
Linus Torvalds 已提交
3686 3687 3688 3689 3690 3691
	tp->rcv_tstamp = tcp_time_stamp;
	prior_packets = tp->packets_out;
	if (!prior_packets)
		goto no_queue;

	/* See if we can take anything off of the retransmit queue. */
I
Ilpo Järvinen 已提交
3692
	flag |= tcp_clean_rtx_queue(sk, prior_fackets, prior_snd_una);
L
Linus Torvalds 已提交
3693

3694 3695
	if (tp->frto_counter)
		frto_cwnd = tcp_process_frto(sk, flag);
3696 3697 3698
	/* Guarantee sacktag reordering detection against wrap-arounds */
	if (before(tp->frto_highmark, tp->snd_una))
		tp->frto_highmark = 0;
L
Linus Torvalds 已提交
3699

3700
	if (tcp_ack_is_dubious(sk, flag)) {
S
Stephen Hemminger 已提交
3701
		/* Advance CWND, if state allows this. */
3702 3703
		if ((flag & FLAG_DATA_ACKED) && !frto_cwnd &&
		    tcp_may_raise_cwnd(sk, flag))
3704
			tcp_cong_avoid(sk, ack, prior_in_flight);
3705 3706
		tcp_fastretrans_alert(sk, prior_packets - tp->packets_out,
				      flag);
L
Linus Torvalds 已提交
3707
	} else {
3708
		if ((flag & FLAG_DATA_ACKED) && !frto_cwnd)
3709
			tcp_cong_avoid(sk, ack, prior_in_flight);
L
Linus Torvalds 已提交
3710 3711
	}

3712
	if ((flag & FLAG_FORWARD_PROGRESS) || !(flag & FLAG_NOT_DUP))
E
Eric Dumazet 已提交
3713
		dst_confirm(__sk_dst_get(sk));
L
Linus Torvalds 已提交
3714 3715 3716 3717 3718 3719 3720 3721

	return 1;

no_queue:
	/* If this ack opens up a zero window, clear backoff.  It was
	 * being used to time the probes, and is probably far higher than
	 * it needs to be for normal retransmission.
	 */
3722
	if (tcp_send_head(sk))
L
Linus Torvalds 已提交
3723 3724 3725
		tcp_ack_probe(sk);
	return 1;

3726 3727 3728 3729
invalid_ack:
	SOCK_DEBUG(sk, "Ack %u after %u:%u\n", ack, tp->snd_una, tp->snd_nxt);
	return -1;

L
Linus Torvalds 已提交
3730
old_ack:
3731
	if (TCP_SKB_CB(skb)->sacked) {
L
Linus Torvalds 已提交
3732
		tcp_sacktag_write_queue(sk, skb, prior_snd_una);
3733 3734 3735
		if (icsk->icsk_ca_state == TCP_CA_Open)
			tcp_try_keep_open(sk);
	}
L
Linus Torvalds 已提交
3736

3737
	SOCK_DEBUG(sk, "Ack %u before %u:%u\n", ack, tp->snd_una, tp->snd_nxt);
L
Linus Torvalds 已提交
3738 3739 3740 3741 3742 3743 3744
	return 0;
}

/* Look for tcp options. Normally only called on SYN and SYNACK packets.
 * But, this can also be called on packets in the established flow when
 * the fast version below fails.
 */
3745
void tcp_parse_options(struct sk_buff *skb, struct tcp_options_received *opt_rx,
3746
		       u8 **hvpp, int estab)
L
Linus Torvalds 已提交
3747 3748
{
	unsigned char *ptr;
3749
	struct tcphdr *th = tcp_hdr(skb);
3750
	int length = (th->doff * 4) - sizeof(struct tcphdr);
L
Linus Torvalds 已提交
3751 3752 3753 3754

	ptr = (unsigned char *)(th + 1);
	opt_rx->saw_tstamp = 0;

S
Stephen Hemminger 已提交
3755
	while (length > 0) {
3756
		int opcode = *ptr++;
L
Linus Torvalds 已提交
3757 3758 3759
		int opsize;

		switch (opcode) {
3760 3761 3762 3763 3764 3765 3766 3767
		case TCPOPT_EOL:
			return;
		case TCPOPT_NOP:	/* Ref: RFC 793 section 3.1 */
			length--;
			continue;
		default:
			opsize = *ptr++;
			if (opsize < 2) /* "silly options" */
L
Linus Torvalds 已提交
3768
				return;
3769 3770 3771 3772 3773
			if (opsize > length)
				return;	/* don't parse partial options */
			switch (opcode) {
			case TCPOPT_MSS:
				if (opsize == TCPOLEN_MSS && th->syn && !estab) {
3774
					u16 in_mss = get_unaligned_be16(ptr);
3775 3776 3777 3778 3779
					if (in_mss) {
						if (opt_rx->user_mss &&
						    opt_rx->user_mss < in_mss)
							in_mss = opt_rx->user_mss;
						opt_rx->mss_clamp = in_mss;
L
Linus Torvalds 已提交
3780
					}
3781 3782 3783 3784
				}
				break;
			case TCPOPT_WINDOW:
				if (opsize == TCPOLEN_WINDOW && th->syn &&
3785
				    !estab && sysctl_tcp_window_scaling) {
3786 3787 3788 3789 3790 3791 3792 3793
					__u8 snd_wscale = *(__u8 *)ptr;
					opt_rx->wscale_ok = 1;
					if (snd_wscale > 14) {
						if (net_ratelimit())
							printk(KERN_INFO "tcp_parse_options: Illegal window "
							       "scaling value %d >14 received.\n",
							       snd_wscale);
						snd_wscale = 14;
L
Linus Torvalds 已提交
3794
					}
3795 3796 3797 3798 3799 3800
					opt_rx->snd_wscale = snd_wscale;
				}
				break;
			case TCPOPT_TIMESTAMP:
				if ((opsize == TCPOLEN_TIMESTAMP) &&
				    ((estab && opt_rx->tstamp_ok) ||
3801
				     (!estab && sysctl_tcp_timestamps))) {
3802
					opt_rx->saw_tstamp = 1;
3803 3804
					opt_rx->rcv_tsval = get_unaligned_be32(ptr);
					opt_rx->rcv_tsecr = get_unaligned_be32(ptr + 4);
3805 3806 3807 3808
				}
				break;
			case TCPOPT_SACK_PERM:
				if (opsize == TCPOLEN_SACK_PERM && th->syn &&
3809
				    !estab && sysctl_tcp_sack) {
3810 3811 3812 3813
					opt_rx->sack_ok = 1;
					tcp_sack_reset(opt_rx);
				}
				break;
L
Linus Torvalds 已提交
3814

3815 3816 3817 3818 3819 3820 3821
			case TCPOPT_SACK:
				if ((opsize >= (TCPOLEN_SACK_BASE + TCPOLEN_SACK_PERBLOCK)) &&
				   !((opsize - TCPOLEN_SACK_BASE) % TCPOLEN_SACK_PERBLOCK) &&
				   opt_rx->sack_ok) {
					TCP_SKB_CB(skb)->sacked = (ptr - 2) - (unsigned char *)th;
				}
				break;
3822
#ifdef CONFIG_TCP_MD5SIG
3823 3824 3825 3826 3827 3828
			case TCPOPT_MD5SIG:
				/*
				 * The MD5 Hash has already been
				 * checked (see tcp_v{4,6}_do_rcv()).
				 */
				break;
3829
#endif
3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847
			case TCPOPT_COOKIE:
				/* This option is variable length.
				 */
				switch (opsize) {
				case TCPOLEN_COOKIE_BASE:
					/* not yet implemented */
					break;
				case TCPOLEN_COOKIE_PAIR:
					/* not yet implemented */
					break;
				case TCPOLEN_COOKIE_MIN+0:
				case TCPOLEN_COOKIE_MIN+2:
				case TCPOLEN_COOKIE_MIN+4:
				case TCPOLEN_COOKIE_MIN+6:
				case TCPOLEN_COOKIE_MAX:
					/* 16-bit multiple */
					opt_rx->cookie_plus = opsize;
					*hvpp = ptr;
3848
					break;
3849 3850 3851
				default:
					/* ignore option */
					break;
3852
				}
3853
				break;
3854
			}
3855

3856 3857
			ptr += opsize-2;
			length -= opsize;
3858
		}
L
Linus Torvalds 已提交
3859 3860 3861
	}
}

3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877
static int tcp_parse_aligned_timestamp(struct tcp_sock *tp, struct tcphdr *th)
{
	__be32 *ptr = (__be32 *)(th + 1);

	if (*ptr == htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16)
			  | (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP)) {
		tp->rx_opt.saw_tstamp = 1;
		++ptr;
		tp->rx_opt.rcv_tsval = ntohl(*ptr);
		++ptr;
		tp->rx_opt.rcv_tsecr = ntohl(*ptr);
		return 1;
	}
	return 0;
}

L
Linus Torvalds 已提交
3878 3879 3880
/* Fast parse options. This hopes to only see timestamps.
 * If it is wrong it falls back on tcp_parse_options().
 */
S
Stephen Hemminger 已提交
3881
static int tcp_fast_parse_options(struct sk_buff *skb, struct tcphdr *th,
3882
				  struct tcp_sock *tp, u8 **hvpp)
L
Linus Torvalds 已提交
3883
{
3884 3885 3886 3887
	/* In the spirit of fast parsing, compare doff directly to constant
	 * values.  Because equality is used, short doff can be ignored here.
	 */
	if (th->doff == (sizeof(*th) / 4)) {
L
Linus Torvalds 已提交
3888 3889 3890
		tp->rx_opt.saw_tstamp = 0;
		return 0;
	} else if (tp->rx_opt.tstamp_ok &&
3891
		   th->doff == ((sizeof(*th) + TCPOLEN_TSTAMP_ALIGNED) / 4)) {
3892
		if (tcp_parse_aligned_timestamp(tp, th))
L
Linus Torvalds 已提交
3893 3894
			return 1;
	}
3895
	tcp_parse_options(skb, &tp->rx_opt, hvpp, 1);
L
Linus Torvalds 已提交
3896 3897 3898
	return 1;
}

3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935
#ifdef CONFIG_TCP_MD5SIG
/*
 * Parse MD5 Signature option
 */
u8 *tcp_parse_md5sig_option(struct tcphdr *th)
{
	int length = (th->doff << 2) - sizeof (*th);
	u8 *ptr = (u8*)(th + 1);

	/* If the TCP option is too short, we can short cut */
	if (length < TCPOLEN_MD5SIG)
		return NULL;

	while (length > 0) {
		int opcode = *ptr++;
		int opsize;

		switch(opcode) {
		case TCPOPT_EOL:
			return NULL;
		case TCPOPT_NOP:
			length--;
			continue;
		default:
			opsize = *ptr++;
			if (opsize < 2 || opsize > length)
				return NULL;
			if (opcode == TCPOPT_MD5SIG)
				return ptr;
		}
		ptr += opsize - 2;
		length -= opsize;
	}
	return NULL;
}
#endif

L
Linus Torvalds 已提交
3936 3937 3938
static inline void tcp_store_ts_recent(struct tcp_sock *tp)
{
	tp->rx_opt.ts_recent = tp->rx_opt.rcv_tsval;
3939
	tp->rx_opt.ts_recent_stamp = get_seconds();
L
Linus Torvalds 已提交
3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951
}

static inline void tcp_replace_ts_recent(struct tcp_sock *tp, u32 seq)
{
	if (tp->rx_opt.saw_tstamp && !after(seq, tp->rcv_wup)) {
		/* PAWS bug workaround wrt. ACK frames, the PAWS discard
		 * extra check below makes sure this can only happen
		 * for pure ACK frames.  -DaveM
		 *
		 * Not only, also it occurs for expired timestamps.
		 */

I
Ilpo Järvinen 已提交
3952
		if (tcp_paws_check(&tp->rx_opt, 0))
L
Linus Torvalds 已提交
3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979
			tcp_store_ts_recent(tp);
	}
}

/* Sorry, PAWS as specified is broken wrt. pure-ACKs -DaveM
 *
 * It is not fatal. If this ACK does _not_ change critical state (seqs, window)
 * it can pass through stack. So, the following predicate verifies that
 * this segment is not used for anything but congestion avoidance or
 * fast retransmit. Moreover, we even are able to eliminate most of such
 * second order effects, if we apply some small "replay" window (~RTO)
 * to timestamp space.
 *
 * All these measures still do not guarantee that we reject wrapped ACKs
 * on networks with high bandwidth, when sequence space is recycled fastly,
 * but it guarantees that such events will be very rare and do not affect
 * connection seriously. This doesn't look nice, but alas, PAWS is really
 * buggy extension.
 *
 * [ Later note. Even worse! It is buggy for segments _with_ data. RFC
 * states that events when retransmit arrives after original data are rare.
 * It is a blatant lie. VJ forgot about fast retransmit! 8)8) It is
 * the biggest problem on large power networks even with minor reordering.
 * OK, let's give it small replay window. If peer clock is even 1hz, it is safe
 * up to bandwidth of 18Gigabit/sec. 8) ]
 */

3980
static int tcp_disordered_ack(const struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
3981
{
3982
	struct tcp_sock *tp = tcp_sk(sk);
3983
	struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996
	u32 seq = TCP_SKB_CB(skb)->seq;
	u32 ack = TCP_SKB_CB(skb)->ack_seq;

	return (/* 1. Pure ACK with correct sequence number. */
		(th->ack && seq == TCP_SKB_CB(skb)->end_seq && seq == tp->rcv_nxt) &&

		/* 2. ... and duplicate ACK. */
		ack == tp->snd_una &&

		/* 3. ... and does not update window. */
		!tcp_may_update_window(tp, ack, seq, ntohs(th->window) << tp->rx_opt.snd_wscale) &&

		/* 4. ... and sits in replay window. */
3997
		(s32)(tp->rx_opt.ts_recent - tp->rx_opt.rcv_tsval) <= (inet_csk(sk)->icsk_rto * 1024) / HZ);
L
Linus Torvalds 已提交
3998 3999
}

4000 4001
static inline int tcp_paws_discard(const struct sock *sk,
				   const struct sk_buff *skb)
L
Linus Torvalds 已提交
4002
{
4003
	const struct tcp_sock *tp = tcp_sk(sk);
I
Ilpo Järvinen 已提交
4004 4005 4006

	return !tcp_paws_check(&tp->rx_opt, TCP_PAWS_WINDOW) &&
	       !tcp_disordered_ack(sk, skb);
L
Linus Torvalds 已提交
4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032
}

/* Check segment sequence number for validity.
 *
 * Segment controls are considered valid, if the segment
 * fits to the window after truncation to the window. Acceptability
 * of data (and SYN, FIN, of course) is checked separately.
 * See tcp_data_queue(), for example.
 *
 * Also, controls (RST is main one) are accepted using RCV.WUP instead
 * of RCV.NXT. Peer still did not advance his SND.UNA when we
 * delayed ACK, so that hisSND.UNA<=ourRCV.WUP.
 * (borrowed from freebsd)
 */

static inline int tcp_sequence(struct tcp_sock *tp, u32 seq, u32 end_seq)
{
	return	!before(end_seq, tp->rcv_wup) &&
		!after(seq, tp->rcv_nxt + tcp_receive_window(tp));
}

/* When we get a reset we do this. */
static void tcp_reset(struct sock *sk)
{
	/* We want the right error as BSD sees it (and indeed as we do). */
	switch (sk->sk_state) {
4033 4034 4035 4036 4037 4038 4039 4040 4041 4042
	case TCP_SYN_SENT:
		sk->sk_err = ECONNREFUSED;
		break;
	case TCP_CLOSE_WAIT:
		sk->sk_err = EPIPE;
		break;
	case TCP_CLOSE:
		return;
	default:
		sk->sk_err = ECONNRESET;
L
Linus Torvalds 已提交
4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068
	}

	if (!sock_flag(sk, SOCK_DEAD))
		sk->sk_error_report(sk);

	tcp_done(sk);
}

/*
 * 	Process the FIN bit. This now behaves as it is supposed to work
 *	and the FIN takes effect when it is validly part of sequence
 *	space. Not before when we get holes.
 *
 *	If we are ESTABLISHED, a received fin moves us to CLOSE-WAIT
 *	(and thence onto LAST-ACK and finally, CLOSE, we never enter
 *	TIME-WAIT)
 *
 *	If we are in FINWAIT-1, a received FIN indicates simultaneous
 *	close and we go into CLOSING (and later onto TIME-WAIT)
 *
 *	If we are in FINWAIT-2, a received FIN moves us to TIME-WAIT.
 */
static void tcp_fin(struct sk_buff *skb, struct sock *sk, struct tcphdr *th)
{
	struct tcp_sock *tp = tcp_sk(sk);

4069
	inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
4070 4071 4072 4073 4074

	sk->sk_shutdown |= RCV_SHUTDOWN;
	sock_set_flag(sk, SOCK_DONE);

	switch (sk->sk_state) {
4075 4076 4077 4078 4079 4080
	case TCP_SYN_RECV:
	case TCP_ESTABLISHED:
		/* Move to CLOSE_WAIT */
		tcp_set_state(sk, TCP_CLOSE_WAIT);
		inet_csk(sk)->icsk_ack.pingpong = 1;
		break;
L
Linus Torvalds 已提交
4081

4082 4083 4084 4085 4086 4087 4088 4089 4090
	case TCP_CLOSE_WAIT:
	case TCP_CLOSING:
		/* Received a retransmission of the FIN, do
		 * nothing.
		 */
		break;
	case TCP_LAST_ACK:
		/* RFC793: Remain in the LAST-ACK state. */
		break;
L
Linus Torvalds 已提交
4091

4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109
	case TCP_FIN_WAIT1:
		/* This case occurs when a simultaneous close
		 * happens, we must ack the received FIN and
		 * enter the CLOSING state.
		 */
		tcp_send_ack(sk);
		tcp_set_state(sk, TCP_CLOSING);
		break;
	case TCP_FIN_WAIT2:
		/* Received a FIN -- send ACK and enter TIME_WAIT. */
		tcp_send_ack(sk);
		tcp_time_wait(sk, TCP_TIME_WAIT, 0);
		break;
	default:
		/* Only TCP_LISTEN and TCP_CLOSE are left, in these
		 * cases we should never reach this piece of code.
		 */
		printk(KERN_ERR "%s: Impossible, sk->sk_state=%d\n",
4110
		       __func__, sk->sk_state);
4111
		break;
4112
	}
L
Linus Torvalds 已提交
4113 4114 4115 4116 4117

	/* It _is_ possible, that we have something out-of-order _after_ FIN.
	 * Probably, we should reset in this case. For now drop them.
	 */
	__skb_queue_purge(&tp->out_of_order_queue);
4118
	if (tcp_is_sack(tp))
L
Linus Torvalds 已提交
4119
		tcp_sack_reset(&tp->rx_opt);
4120
	sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
4121 4122 4123 4124 4125 4126 4127

	if (!sock_flag(sk, SOCK_DEAD)) {
		sk->sk_state_change(sk);

		/* Do not send POLL_HUP for half duplex close. */
		if (sk->sk_shutdown == SHUTDOWN_MASK ||
		    sk->sk_state == TCP_CLOSE)
4128
			sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_HUP);
L
Linus Torvalds 已提交
4129
		else
4130
			sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
L
Linus Torvalds 已提交
4131 4132 4133
	}
}

4134 4135
static inline int tcp_sack_extend(struct tcp_sack_block *sp, u32 seq,
				  u32 end_seq)
L
Linus Torvalds 已提交
4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146
{
	if (!after(seq, sp->end_seq) && !after(sp->start_seq, end_seq)) {
		if (before(seq, sp->start_seq))
			sp->start_seq = seq;
		if (after(end_seq, sp->end_seq))
			sp->end_seq = end_seq;
		return 1;
	}
	return 0;
}

4147
static void tcp_dsack_set(struct sock *sk, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
4148
{
4149 4150
	struct tcp_sock *tp = tcp_sk(sk);

4151
	if (tcp_is_sack(tp) && sysctl_tcp_dsack) {
4152 4153
		int mib_idx;

L
Linus Torvalds 已提交
4154
		if (before(seq, tp->rcv_nxt))
4155
			mib_idx = LINUX_MIB_TCPDSACKOLDSENT;
L
Linus Torvalds 已提交
4156
		else
4157 4158
			mib_idx = LINUX_MIB_TCPDSACKOFOSENT;

4159
		NET_INC_STATS_BH(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
4160 4161 4162 4163 4164 4165 4166

		tp->rx_opt.dsack = 1;
		tp->duplicate_sack[0].start_seq = seq;
		tp->duplicate_sack[0].end_seq = end_seq;
	}
}

4167
static void tcp_dsack_extend(struct sock *sk, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
4168
{
4169 4170
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
4171
	if (!tp->rx_opt.dsack)
4172
		tcp_dsack_set(sk, seq, end_seq);
L
Linus Torvalds 已提交
4173 4174 4175 4176 4177 4178 4179 4180 4181 4182
	else
		tcp_sack_extend(tp->duplicate_sack, seq, end_seq);
}

static void tcp_send_dupack(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);

	if (TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(skb)->seq &&
	    before(TCP_SKB_CB(skb)->seq, tp->rcv_nxt)) {
4183
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_DELAYEDACKLOST);
4184
		tcp_enter_quickack_mode(sk);
L
Linus Torvalds 已提交
4185

4186
		if (tcp_is_sack(tp) && sysctl_tcp_dsack) {
L
Linus Torvalds 已提交
4187 4188 4189 4190
			u32 end_seq = TCP_SKB_CB(skb)->end_seq;

			if (after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt))
				end_seq = tp->rcv_nxt;
4191
			tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, end_seq);
L
Linus Torvalds 已提交
4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204
		}
	}

	tcp_send_ack(sk);
}

/* These routines update the SACK block as out-of-order packets arrive or
 * in-order packets close up the sequence space.
 */
static void tcp_sack_maybe_coalesce(struct tcp_sock *tp)
{
	int this_sack;
	struct tcp_sack_block *sp = &tp->selective_acks[0];
4205
	struct tcp_sack_block *swalk = sp + 1;
L
Linus Torvalds 已提交
4206 4207 4208 4209

	/* See if the recent change to the first SACK eats into
	 * or hits the sequence space of other SACK blocks, if so coalesce.
	 */
4210
	for (this_sack = 1; this_sack < tp->rx_opt.num_sacks;) {
L
Linus Torvalds 已提交
4211 4212 4213 4214 4215 4216 4217
		if (tcp_sack_extend(sp, swalk->start_seq, swalk->end_seq)) {
			int i;

			/* Zap SWALK, by moving every further SACK up by one slot.
			 * Decrease num_sacks.
			 */
			tp->rx_opt.num_sacks--;
4218 4219
			for (i = this_sack; i < tp->rx_opt.num_sacks; i++)
				sp[i] = sp[i + 1];
L
Linus Torvalds 已提交
4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235
			continue;
		}
		this_sack++, swalk++;
	}
}

static void tcp_sack_new_ofo_skb(struct sock *sk, u32 seq, u32 end_seq)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct tcp_sack_block *sp = &tp->selective_acks[0];
	int cur_sacks = tp->rx_opt.num_sacks;
	int this_sack;

	if (!cur_sacks)
		goto new_sack;

4236
	for (this_sack = 0; this_sack < cur_sacks; this_sack++, sp++) {
L
Linus Torvalds 已提交
4237 4238
		if (tcp_sack_extend(sp, seq, end_seq)) {
			/* Rotate this_sack to the first one. */
4239
			for (; this_sack > 0; this_sack--, sp--)
4240
				swap(*sp, *(sp - 1));
L
Linus Torvalds 已提交
4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252
			if (cur_sacks > 1)
				tcp_sack_maybe_coalesce(tp);
			return;
		}
	}

	/* Could not find an adjacent existing SACK, build a new one,
	 * put it at the front, and shift everyone else down.  We
	 * always know there is at least one SACK present already here.
	 *
	 * If the sack array is full, forget about the last one.
	 */
4253
	if (this_sack >= TCP_NUM_SACKS) {
L
Linus Torvalds 已提交
4254 4255 4256 4257
		this_sack--;
		tp->rx_opt.num_sacks--;
		sp--;
	}
S
Stephen Hemminger 已提交
4258
	for (; this_sack > 0; this_sack--, sp--)
4259
		*sp = *(sp - 1);
L
Linus Torvalds 已提交
4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276

new_sack:
	/* Build the new head SACK, and we're done. */
	sp->start_seq = seq;
	sp->end_seq = end_seq;
	tp->rx_opt.num_sacks++;
}

/* RCV.NXT advances, some SACKs should be eaten. */

static void tcp_sack_remove(struct tcp_sock *tp)
{
	struct tcp_sack_block *sp = &tp->selective_acks[0];
	int num_sacks = tp->rx_opt.num_sacks;
	int this_sack;

	/* Empty ofo queue, hence, all the SACKs are eaten. Clear. */
4277
	if (skb_queue_empty(&tp->out_of_order_queue)) {
L
Linus Torvalds 已提交
4278 4279 4280 4281
		tp->rx_opt.num_sacks = 0;
		return;
	}

4282
	for (this_sack = 0; this_sack < num_sacks;) {
L
Linus Torvalds 已提交
4283 4284 4285 4286 4287
		/* Check if the start of the sack is covered by RCV.NXT. */
		if (!before(tp->rcv_nxt, sp->start_seq)) {
			int i;

			/* RCV.NXT must cover all the block! */
4288
			WARN_ON(before(tp->rcv_nxt, sp->end_seq));
L
Linus Torvalds 已提交
4289 4290 4291 4292 4293 4294 4295 4296 4297 4298

			/* Zap this SACK, by moving forward any other SACKS. */
			for (i=this_sack+1; i < num_sacks; i++)
				tp->selective_acks[i-1] = tp->selective_acks[i];
			num_sacks--;
			continue;
		}
		this_sack++;
		sp++;
	}
4299
	tp->rx_opt.num_sacks = num_sacks;
L
Linus Torvalds 已提交
4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318
}

/* This one checks to see if we can put data from the
 * out_of_order queue into the receive_queue.
 */
static void tcp_ofo_queue(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	__u32 dsack_high = tp->rcv_nxt;
	struct sk_buff *skb;

	while ((skb = skb_peek(&tp->out_of_order_queue)) != NULL) {
		if (after(TCP_SKB_CB(skb)->seq, tp->rcv_nxt))
			break;

		if (before(TCP_SKB_CB(skb)->seq, dsack_high)) {
			__u32 dsack = dsack_high;
			if (before(TCP_SKB_CB(skb)->end_seq, dsack_high))
				dsack_high = TCP_SKB_CB(skb)->end_seq;
4319
			tcp_dsack_extend(sk, TCP_SKB_CB(skb)->seq, dsack);
L
Linus Torvalds 已提交
4320 4321 4322
		}

		if (!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) {
4323
			SOCK_DEBUG(sk, "ofo packet was already received\n");
D
David S. Miller 已提交
4324
			__skb_unlink(skb, &tp->out_of_order_queue);
L
Linus Torvalds 已提交
4325 4326 4327 4328 4329 4330 4331
			__kfree_skb(skb);
			continue;
		}
		SOCK_DEBUG(sk, "ofo requeuing : rcv_next %X seq %X - %X\n",
			   tp->rcv_nxt, TCP_SKB_CB(skb)->seq,
			   TCP_SKB_CB(skb)->end_seq);

D
David S. Miller 已提交
4332
		__skb_unlink(skb, &tp->out_of_order_queue);
L
Linus Torvalds 已提交
4333 4334
		__skb_queue_tail(&sk->sk_receive_queue, skb);
		tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq;
4335 4336
		if (tcp_hdr(skb)->fin)
			tcp_fin(skb, sk, tcp_hdr(skb));
L
Linus Torvalds 已提交
4337 4338 4339
	}
}

4340
static int tcp_prune_ofo_queue(struct sock *sk);
L
Linus Torvalds 已提交
4341 4342
static int tcp_prune_queue(struct sock *sk);

4343 4344 4345 4346 4347 4348 4349 4350 4351
static inline int tcp_try_rmem_schedule(struct sock *sk, unsigned int size)
{
	if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
	    !sk_rmem_schedule(sk, size)) {

		if (tcp_prune_queue(sk) < 0)
			return -1;

		if (!sk_rmem_schedule(sk, size)) {
4352 4353 4354
			if (!tcp_prune_ofo_queue(sk))
				return -1;

4355 4356 4357 4358 4359 4360 4361
			if (!sk_rmem_schedule(sk, size))
				return -1;
		}
	}
	return 0;
}

L
Linus Torvalds 已提交
4362 4363
static void tcp_data_queue(struct sock *sk, struct sk_buff *skb)
{
4364
	struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
4365 4366 4367 4368 4369 4370
	struct tcp_sock *tp = tcp_sk(sk);
	int eaten = -1;

	if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq)
		goto drop;

E
Eric Dumazet 已提交
4371
	skb_dst_drop(skb);
4372
	__skb_pull(skb, th->doff * 4);
L
Linus Torvalds 已提交
4373 4374 4375

	TCP_ECN_accept_cwr(tp, skb);

4376
	tp->rx_opt.dsack = 0;
L
Linus Torvalds 已提交
4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390

	/*  Queue data for delivery to the user.
	 *  Packets in sequence go to the receive queue.
	 *  Out of sequence packets to the out_of_order_queue.
	 */
	if (TCP_SKB_CB(skb)->seq == tp->rcv_nxt) {
		if (tcp_receive_window(tp) == 0)
			goto out_of_window;

		/* Ok. In sequence. In window. */
		if (tp->ucopy.task == current &&
		    tp->copied_seq == tp->rcv_nxt && tp->ucopy.len &&
		    sock_owned_by_user(sk) && !tp->urg_data) {
			int chunk = min_t(unsigned int, skb->len,
4391
					  tp->ucopy.len);
L
Linus Torvalds 已提交
4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407

			__set_current_state(TASK_RUNNING);

			local_bh_enable();
			if (!skb_copy_datagram_iovec(skb, 0, tp->ucopy.iov, chunk)) {
				tp->ucopy.len -= chunk;
				tp->copied_seq += chunk;
				eaten = (chunk == skb->len && !th->fin);
				tcp_rcv_space_adjust(sk);
			}
			local_bh_disable();
		}

		if (eaten <= 0) {
queue_and_out:
			if (eaten < 0 &&
4408 4409 4410
			    tcp_try_rmem_schedule(sk, skb->truesize))
				goto drop;

4411
			skb_set_owner_r(skb, sk);
L
Linus Torvalds 已提交
4412 4413 4414
			__skb_queue_tail(&sk->sk_receive_queue, skb);
		}
		tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq;
S
Stephen Hemminger 已提交
4415
		if (skb->len)
4416
			tcp_event_data_recv(sk, skb);
S
Stephen Hemminger 已提交
4417
		if (th->fin)
L
Linus Torvalds 已提交
4418 4419
			tcp_fin(skb, sk, th);

4420
		if (!skb_queue_empty(&tp->out_of_order_queue)) {
L
Linus Torvalds 已提交
4421 4422 4423 4424 4425
			tcp_ofo_queue(sk);

			/* RFC2581. 4.2. SHOULD send immediate ACK, when
			 * gap in queue is filled.
			 */
4426
			if (skb_queue_empty(&tp->out_of_order_queue))
4427
				inet_csk(sk)->icsk_ack.pingpong = 0;
L
Linus Torvalds 已提交
4428 4429 4430 4431 4432
		}

		if (tp->rx_opt.num_sacks)
			tcp_sack_remove(tp);

4433
		tcp_fast_path_check(sk);
L
Linus Torvalds 已提交
4434 4435 4436 4437 4438 4439 4440 4441 4442 4443

		if (eaten > 0)
			__kfree_skb(skb);
		else if (!sock_flag(sk, SOCK_DEAD))
			sk->sk_data_ready(sk, 0);
		return;
	}

	if (!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) {
		/* A retransmit, 2nd most common case.  Force an immediate ack. */
4444
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_DELAYEDACKLOST);
4445
		tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq);
L
Linus Torvalds 已提交
4446 4447

out_of_window:
4448 4449
		tcp_enter_quickack_mode(sk);
		inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
4450 4451 4452 4453 4454 4455 4456 4457 4458
drop:
		__kfree_skb(skb);
		return;
	}

	/* Out of window. F.e. zero window probe. */
	if (!before(TCP_SKB_CB(skb)->seq, tp->rcv_nxt + tcp_receive_window(tp)))
		goto out_of_window;

4459
	tcp_enter_quickack_mode(sk);
L
Linus Torvalds 已提交
4460 4461 4462 4463 4464 4465 4466

	if (before(TCP_SKB_CB(skb)->seq, tp->rcv_nxt)) {
		/* Partial packet, seq < rcv_next < end_seq */
		SOCK_DEBUG(sk, "partial packet: rcv_next %X seq %X - %X\n",
			   tp->rcv_nxt, TCP_SKB_CB(skb)->seq,
			   TCP_SKB_CB(skb)->end_seq);

4467
		tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, tp->rcv_nxt);
4468

L
Linus Torvalds 已提交
4469 4470 4471 4472 4473 4474 4475 4476 4477 4478
		/* If window is closed, drop tail of packet. But after
		 * remembering D-SACK for its head made in previous line.
		 */
		if (!tcp_receive_window(tp))
			goto out_of_window;
		goto queue_and_out;
	}

	TCP_ECN_check_ce(tp, skb);

4479 4480
	if (tcp_try_rmem_schedule(sk, skb->truesize))
		goto drop;
L
Linus Torvalds 已提交
4481 4482 4483

	/* Disable header prediction. */
	tp->pred_flags = 0;
4484
	inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
4485 4486 4487 4488

	SOCK_DEBUG(sk, "out of order segment: rcv_next %X seq %X - %X\n",
		   tp->rcv_nxt, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq);

4489
	skb_set_owner_r(skb, sk);
L
Linus Torvalds 已提交
4490 4491 4492

	if (!skb_peek(&tp->out_of_order_queue)) {
		/* Initial out of order segment, build 1 SACK. */
4493
		if (tcp_is_sack(tp)) {
L
Linus Torvalds 已提交
4494 4495 4496 4497 4498
			tp->rx_opt.num_sacks = 1;
			tp->selective_acks[0].start_seq = TCP_SKB_CB(skb)->seq;
			tp->selective_acks[0].end_seq =
						TCP_SKB_CB(skb)->end_seq;
		}
4499
		__skb_queue_head(&tp->out_of_order_queue, skb);
L
Linus Torvalds 已提交
4500
	} else {
4501
		struct sk_buff *skb1 = skb_peek_tail(&tp->out_of_order_queue);
L
Linus Torvalds 已提交
4502 4503 4504 4505
		u32 seq = TCP_SKB_CB(skb)->seq;
		u32 end_seq = TCP_SKB_CB(skb)->end_seq;

		if (seq == TCP_SKB_CB(skb1)->end_seq) {
4506
			__skb_queue_after(&tp->out_of_order_queue, skb1, skb);
L
Linus Torvalds 已提交
4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517

			if (!tp->rx_opt.num_sacks ||
			    tp->selective_acks[0].end_seq != seq)
				goto add_sack;

			/* Common case: data arrive in order after hole. */
			tp->selective_acks[0].end_seq = end_seq;
			return;
		}

		/* Find place to insert this segment. */
4518
		while (1) {
L
Linus Torvalds 已提交
4519 4520
			if (!after(TCP_SKB_CB(skb1)->seq, seq))
				break;
4521 4522 4523 4524 4525 4526
			if (skb_queue_is_first(&tp->out_of_order_queue, skb1)) {
				skb1 = NULL;
				break;
			}
			skb1 = skb_queue_prev(&tp->out_of_order_queue, skb1);
		}
L
Linus Torvalds 已提交
4527 4528

		/* Do skb overlap to previous one? */
4529
		if (skb1 && before(seq, TCP_SKB_CB(skb1)->end_seq)) {
L
Linus Torvalds 已提交
4530 4531 4532
			if (!after(end_seq, TCP_SKB_CB(skb1)->end_seq)) {
				/* All the bits are present. Drop. */
				__kfree_skb(skb);
4533
				tcp_dsack_set(sk, seq, end_seq);
L
Linus Torvalds 已提交
4534 4535 4536 4537
				goto add_sack;
			}
			if (after(seq, TCP_SKB_CB(skb1)->seq)) {
				/* Partial overlap. */
4538
				tcp_dsack_set(sk, seq,
4539
					      TCP_SKB_CB(skb1)->end_seq);
L
Linus Torvalds 已提交
4540
			} else {
4541 4542 4543 4544 4545 4546 4547
				if (skb_queue_is_first(&tp->out_of_order_queue,
						       skb1))
					skb1 = NULL;
				else
					skb1 = skb_queue_prev(
						&tp->out_of_order_queue,
						skb1);
L
Linus Torvalds 已提交
4548 4549
			}
		}
4550 4551 4552 4553
		if (!skb1)
			__skb_queue_head(&tp->out_of_order_queue, skb);
		else
			__skb_queue_after(&tp->out_of_order_queue, skb1, skb);
4554

L
Linus Torvalds 已提交
4555
		/* And clean segments covered by new one as whole. */
4556 4557
		while (!skb_queue_is_last(&tp->out_of_order_queue, skb)) {
			skb1 = skb_queue_next(&tp->out_of_order_queue, skb);
4558

4559 4560 4561
			if (!after(end_seq, TCP_SKB_CB(skb1)->seq))
				break;
			if (before(end_seq, TCP_SKB_CB(skb1)->end_seq)) {
4562
				tcp_dsack_extend(sk, TCP_SKB_CB(skb1)->seq,
4563 4564
						 end_seq);
				break;
4565
			}
4566 4567 4568 4569
			__skb_unlink(skb1, &tp->out_of_order_queue);
			tcp_dsack_extend(sk, TCP_SKB_CB(skb1)->seq,
					 TCP_SKB_CB(skb1)->end_seq);
			__kfree_skb(skb1);
L
Linus Torvalds 已提交
4570 4571 4572
		}

add_sack:
4573
		if (tcp_is_sack(tp))
L
Linus Torvalds 已提交
4574 4575 4576 4577
			tcp_sack_new_ofo_skb(sk, seq, end_seq);
	}
}

4578 4579 4580
static struct sk_buff *tcp_collapse_one(struct sock *sk, struct sk_buff *skb,
					struct sk_buff_head *list)
{
4581 4582 4583 4584
	struct sk_buff *next = NULL;

	if (!skb_queue_is_last(list, skb))
		next = skb_queue_next(list, skb);
4585 4586 4587 4588 4589 4590 4591 4592

	__skb_unlink(skb, list);
	__kfree_skb(skb);
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPRCVCOLLAPSED);

	return next;
}

L
Linus Torvalds 已提交
4593 4594
/* Collapse contiguous sequence of skbs head..tail with
 * sequence numbers start..end.
4595 4596 4597
 *
 * If tail is NULL, this means until the end of the list.
 *
L
Linus Torvalds 已提交
4598 4599 4600 4601
 * Segments with FIN/SYN are not collapsed (only because this
 * simplifies code)
 */
static void
D
David S. Miller 已提交
4602 4603 4604
tcp_collapse(struct sock *sk, struct sk_buff_head *list,
	     struct sk_buff *head, struct sk_buff *tail,
	     u32 start, u32 end)
L
Linus Torvalds 已提交
4605
{
4606 4607
	struct sk_buff *skb, *n;
	bool end_of_skbs;
L
Linus Torvalds 已提交
4608

S
Stephen Hemminger 已提交
4609
	/* First, check that queue is collapsible and find
L
Linus Torvalds 已提交
4610
	 * the point where collapsing can be useful. */
4611 4612 4613 4614 4615 4616
	skb = head;
restart:
	end_of_skbs = true;
	skb_queue_walk_from_safe(list, skb, n) {
		if (skb == tail)
			break;
L
Linus Torvalds 已提交
4617 4618
		/* No new bits? It is possible on ofo queue. */
		if (!before(start, TCP_SKB_CB(skb)->end_seq)) {
4619
			skb = tcp_collapse_one(sk, skb, list);
4620 4621 4622
			if (!skb)
				break;
			goto restart;
L
Linus Torvalds 已提交
4623 4624 4625 4626 4627 4628 4629
		}

		/* The first skb to collapse is:
		 * - not SYN/FIN and
		 * - bloated or contains data before "start" or
		 *   overlaps to the next one.
		 */
4630
		if (!tcp_hdr(skb)->syn && !tcp_hdr(skb)->fin &&
L
Linus Torvalds 已提交
4631
		    (tcp_win_from_space(skb->truesize) > skb->len ||
4632 4633
		     before(TCP_SKB_CB(skb)->seq, start))) {
			end_of_skbs = false;
L
Linus Torvalds 已提交
4634
			break;
4635 4636 4637 4638 4639 4640 4641 4642 4643 4644
		}

		if (!skb_queue_is_last(list, skb)) {
			struct sk_buff *next = skb_queue_next(list, skb);
			if (next != tail &&
			    TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(next)->seq) {
				end_of_skbs = false;
				break;
			}
		}
L
Linus Torvalds 已提交
4645 4646 4647 4648

		/* Decided to skip this, advance start seq. */
		start = TCP_SKB_CB(skb)->end_seq;
	}
4649
	if (end_of_skbs || tcp_hdr(skb)->syn || tcp_hdr(skb)->fin)
L
Linus Torvalds 已提交
4650 4651 4652 4653
		return;

	while (before(start, end)) {
		struct sk_buff *nskb;
4654
		unsigned int header = skb_headroom(skb);
L
Linus Torvalds 已提交
4655 4656 4657 4658 4659
		int copy = SKB_MAX_ORDER(header, 0);

		/* Too big header? This can happen with IPv6. */
		if (copy < 0)
			return;
4660 4661 4662
		if (end - start < copy)
			copy = end - start;
		nskb = alloc_skb(copy + header, GFP_ATOMIC);
L
Linus Torvalds 已提交
4663 4664
		if (!nskb)
			return;
4665

4666
		skb_set_mac_header(nskb, skb_mac_header(skb) - skb->head);
4667 4668 4669 4670
		skb_set_network_header(nskb, (skb_network_header(skb) -
					      skb->head));
		skb_set_transport_header(nskb, (skb_transport_header(skb) -
						skb->head));
L
Linus Torvalds 已提交
4671 4672 4673 4674
		skb_reserve(nskb, header);
		memcpy(nskb->head, skb->head, header);
		memcpy(nskb->cb, skb->cb, sizeof(skb->cb));
		TCP_SKB_CB(nskb)->seq = TCP_SKB_CB(nskb)->end_seq = start;
4675
		__skb_queue_before(list, skb, nskb);
4676
		skb_set_owner_r(nskb, sk);
L
Linus Torvalds 已提交
4677 4678 4679 4680 4681 4682

		/* Copy data, releasing collapsed skbs. */
		while (copy > 0) {
			int offset = start - TCP_SKB_CB(skb)->seq;
			int size = TCP_SKB_CB(skb)->end_seq - start;

4683
			BUG_ON(offset < 0);
L
Linus Torvalds 已提交
4684 4685 4686 4687 4688 4689 4690 4691 4692
			if (size > 0) {
				size = min(copy, size);
				if (skb_copy_bits(skb, offset, skb_put(nskb, size), size))
					BUG();
				TCP_SKB_CB(nskb)->end_seq += size;
				copy -= size;
				start += size;
			}
			if (!before(start, TCP_SKB_CB(skb)->end_seq)) {
4693
				skb = tcp_collapse_one(sk, skb, list);
4694 4695
				if (!skb ||
				    skb == tail ||
4696 4697
				    tcp_hdr(skb)->syn ||
				    tcp_hdr(skb)->fin)
L
Linus Torvalds 已提交
4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721
					return;
			}
		}
	}
}

/* Collapse ofo queue. Algorithm: select contiguous sequence of skbs
 * and tcp_collapse() them until all the queue is collapsed.
 */
static void tcp_collapse_ofo_queue(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb = skb_peek(&tp->out_of_order_queue);
	struct sk_buff *head;
	u32 start, end;

	if (skb == NULL)
		return;

	start = TCP_SKB_CB(skb)->seq;
	end = TCP_SKB_CB(skb)->end_seq;
	head = skb;

	for (;;) {
4722 4723 4724 4725 4726
		struct sk_buff *next = NULL;

		if (!skb_queue_is_last(&tp->out_of_order_queue, skb))
			next = skb_queue_next(&tp->out_of_order_queue, skb);
		skb = next;
L
Linus Torvalds 已提交
4727 4728 4729

		/* Segment is terminated when we see gap or when
		 * we are at the end of all the queue. */
4730
		if (!skb ||
L
Linus Torvalds 已提交
4731 4732
		    after(TCP_SKB_CB(skb)->seq, end) ||
		    before(TCP_SKB_CB(skb)->end_seq, start)) {
D
David S. Miller 已提交
4733 4734
			tcp_collapse(sk, &tp->out_of_order_queue,
				     head, skb, start, end);
L
Linus Torvalds 已提交
4735
			head = skb;
4736
			if (!skb)
L
Linus Torvalds 已提交
4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749
				break;
			/* Start new segment */
			start = TCP_SKB_CB(skb)->seq;
			end = TCP_SKB_CB(skb)->end_seq;
		} else {
			if (before(TCP_SKB_CB(skb)->seq, start))
				start = TCP_SKB_CB(skb)->seq;
			if (after(TCP_SKB_CB(skb)->end_seq, end))
				end = TCP_SKB_CB(skb)->end_seq;
		}
	}
}

4750 4751
/*
 * Purge the out-of-order queue.
4752
 * Return true if queue was pruned.
4753
 */
4754
static int tcp_prune_ofo_queue(struct sock *sk)
4755 4756
{
	struct tcp_sock *tp = tcp_sk(sk);
4757
	int res = 0;
4758 4759

	if (!skb_queue_empty(&tp->out_of_order_queue)) {
4760
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_OFOPRUNED);
4761 4762 4763 4764 4765 4766 4767 4768 4769 4770
		__skb_queue_purge(&tp->out_of_order_queue);

		/* Reset SACK state.  A conforming SACK implementation will
		 * do the same at a timeout based retransmit.  When a connection
		 * is in a sad state like this, we care only about integrity
		 * of the connection not performance.
		 */
		if (tp->rx_opt.sack_ok)
			tcp_sack_reset(&tp->rx_opt);
		sk_mem_reclaim(sk);
4771
		res = 1;
4772
	}
4773
	return res;
4774 4775
}

L
Linus Torvalds 已提交
4776 4777 4778 4779 4780 4781 4782 4783 4784
/* Reduce allocated memory if we can, trying to get
 * the socket within its memory limits again.
 *
 * Return less than zero if we should start dropping frames
 * until the socket owning process reads some of the data
 * to stabilize the situation.
 */
static int tcp_prune_queue(struct sock *sk)
{
4785
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
4786 4787 4788

	SOCK_DEBUG(sk, "prune_queue: c=%x\n", tp->copied_seq);

4789
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PRUNECALLED);
L
Linus Torvalds 已提交
4790 4791

	if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
4792
		tcp_clamp_window(sk);
L
Linus Torvalds 已提交
4793 4794 4795 4796
	else if (tcp_memory_pressure)
		tp->rcv_ssthresh = min(tp->rcv_ssthresh, 4U * tp->advmss);

	tcp_collapse_ofo_queue(sk);
4797 4798 4799 4800 4801
	if (!skb_queue_empty(&sk->sk_receive_queue))
		tcp_collapse(sk, &sk->sk_receive_queue,
			     skb_peek(&sk->sk_receive_queue),
			     NULL,
			     tp->copied_seq, tp->rcv_nxt);
4802
	sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
4803 4804 4805 4806 4807 4808 4809

	if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf)
		return 0;

	/* Collapsing did not help, destructive actions follow.
	 * This must not ever occur. */

4810
	tcp_prune_ofo_queue(sk);
L
Linus Torvalds 已提交
4811 4812 4813 4814 4815 4816 4817 4818

	if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf)
		return 0;

	/* If we are really being abused, tell the caller to silently
	 * drop receive data on the floor.  It will get retransmitted
	 * and hopefully then we'll have sufficient space.
	 */
4819
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_RCVPRUNED);
L
Linus Torvalds 已提交
4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833

	/* Massive buffer overcommit. */
	tp->pred_flags = 0;
	return -1;
}

/* RFC2861, slow part. Adjust cwnd, after it was not full during one rto.
 * As additional protections, we do not touch cwnd in retransmission phases,
 * and if application hit its sndbuf limit recently.
 */
void tcp_cwnd_application_limited(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);

4834
	if (inet_csk(sk)->icsk_ca_state == TCP_CA_Open &&
L
Linus Torvalds 已提交
4835 4836
	    sk->sk_socket && !test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
		/* Limited by application or receiver window. */
4837 4838
		u32 init_win = tcp_init_cwnd(tp, __sk_dst_get(sk));
		u32 win_used = max(tp->snd_cwnd_used, init_win);
L
Linus Torvalds 已提交
4839
		if (win_used < tp->snd_cwnd) {
4840
			tp->snd_ssthresh = tcp_current_ssthresh(sk);
L
Linus Torvalds 已提交
4841 4842 4843 4844 4845 4846 4847
			tp->snd_cwnd = (tp->snd_cwnd + win_used) >> 1;
		}
		tp->snd_cwnd_used = 0;
	}
	tp->snd_cwnd_stamp = tcp_time_stamp;
}

4848
static int tcp_should_expand_sndbuf(struct sock *sk)
4849
{
4850 4851
	struct tcp_sock *tp = tcp_sk(sk);

4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871
	/* If the user specified a specific send buffer setting, do
	 * not modify it.
	 */
	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
		return 0;

	/* If we are under global TCP memory pressure, do not expand.  */
	if (tcp_memory_pressure)
		return 0;

	/* If we are under soft global TCP memory pressure, do not expand.  */
	if (atomic_read(&tcp_memory_allocated) >= sysctl_tcp_mem[0])
		return 0;

	/* If we filled the congestion window, do not expand.  */
	if (tp->packets_out >= tp->snd_cwnd)
		return 0;

	return 1;
}
L
Linus Torvalds 已提交
4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882

/* When incoming ACK allowed to free some skb from write_queue,
 * we remember this event in flag SOCK_QUEUE_SHRUNK and wake up socket
 * on the exit from tcp input handler.
 *
 * PROBLEM: sndbuf expansion does not work well with largesend.
 */
static void tcp_new_space(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);

4883
	if (tcp_should_expand_sndbuf(sk)) {
4884
		int sndmem = max_t(u32, tp->rx_opt.mss_clamp, tp->mss_cache) +
I
Ilpo Järvinen 已提交
4885 4886
			MAX_TCP_HEADER + 16 + sizeof(struct sk_buff);
		int demanded = max_t(unsigned int, tp->snd_cwnd,
4887 4888
				     tp->reordering + 1);
		sndmem *= 2 * demanded;
L
Linus Torvalds 已提交
4889 4890 4891 4892 4893 4894 4895 4896
		if (sndmem > sk->sk_sndbuf)
			sk->sk_sndbuf = min(sndmem, sysctl_tcp_wmem[2]);
		tp->snd_cwnd_stamp = tcp_time_stamp;
	}

	sk->sk_write_space(sk);
}

S
Stephen Hemminger 已提交
4897
static void tcp_check_space(struct sock *sk)
L
Linus Torvalds 已提交
4898 4899 4900 4901 4902 4903 4904 4905 4906
{
	if (sock_flag(sk, SOCK_QUEUE_SHRUNK)) {
		sock_reset_flag(sk, SOCK_QUEUE_SHRUNK);
		if (sk->sk_socket &&
		    test_bit(SOCK_NOSPACE, &sk->sk_socket->flags))
			tcp_new_space(sk);
	}
}

4907
static inline void tcp_data_snd_check(struct sock *sk)
L
Linus Torvalds 已提交
4908
{
4909
	tcp_push_pending_frames(sk);
L
Linus Torvalds 已提交
4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920
	tcp_check_space(sk);
}

/*
 * Check if sending an ack is needed.
 */
static void __tcp_ack_snd_check(struct sock *sk, int ofo_possible)
{
	struct tcp_sock *tp = tcp_sk(sk);

	    /* More than one full frame received... */
4921
	if (((tp->rcv_nxt - tp->rcv_wup) > inet_csk(sk)->icsk_ack.rcv_mss &&
L
Linus Torvalds 已提交
4922 4923 4924
	     /* ... and right edge of window advances far enough.
	      * (tcp_recvmsg() will send ACK otherwise). Or...
	      */
4925
	     __tcp_select_window(sk) >= tp->rcv_wnd) ||
L
Linus Torvalds 已提交
4926
	    /* We ACK each frame or... */
4927
	    tcp_in_quickack_mode(sk) ||
L
Linus Torvalds 已提交
4928
	    /* We have out of order data. */
4929
	    (ofo_possible && skb_peek(&tp->out_of_order_queue))) {
L
Linus Torvalds 已提交
4930 4931 4932 4933 4934 4935 4936 4937
		/* Then ack it now */
		tcp_send_ack(sk);
	} else {
		/* Else, send delayed ack. */
		tcp_send_delayed_ack(sk);
	}
}

S
Stephen Hemminger 已提交
4938
static inline void tcp_ack_snd_check(struct sock *sk)
L
Linus Torvalds 已提交
4939
{
4940
	if (!inet_csk_ack_scheduled(sk)) {
L
Linus Torvalds 已提交
4941 4942 4943 4944 4945 4946 4947 4948
		/* We sent a data segment already. */
		return;
	}
	__tcp_ack_snd_check(sk, 1);
}

/*
 *	This routine is only called when we have urgent data
S
Stephen Hemminger 已提交
4949
 *	signaled. Its the 'slow' part of tcp_urg. It could be
L
Linus Torvalds 已提交
4950 4951 4952 4953 4954 4955
 *	moved inline now as tcp_urg is only called from one
 *	place. We handle URGent data wrong. We have to - as
 *	BSD still doesn't use the correction from RFC961.
 *	For 1003.1g we should support a new option TCP_STDURG to permit
 *	either form (or just set the sysctl tcp_stdurg).
 */
4956

4957
static void tcp_check_urg(struct sock *sk, struct tcphdr *th)
L
Linus Torvalds 已提交
4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993
{
	struct tcp_sock *tp = tcp_sk(sk);
	u32 ptr = ntohs(th->urg_ptr);

	if (ptr && !sysctl_tcp_stdurg)
		ptr--;
	ptr += ntohl(th->seq);

	/* Ignore urgent data that we've already seen and read. */
	if (after(tp->copied_seq, ptr))
		return;

	/* Do not replay urg ptr.
	 *
	 * NOTE: interesting situation not covered by specs.
	 * Misbehaving sender may send urg ptr, pointing to segment,
	 * which we already have in ofo queue. We are not able to fetch
	 * such data and will stay in TCP_URG_NOTYET until will be eaten
	 * by recvmsg(). Seems, we are not obliged to handle such wicked
	 * situations. But it is worth to think about possibility of some
	 * DoSes using some hypothetical application level deadlock.
	 */
	if (before(ptr, tp->rcv_nxt))
		return;

	/* Do we already have a newer (or duplicate) urgent pointer? */
	if (tp->urg_data && !after(ptr, tp->urg_seq))
		return;

	/* Tell the world about our new urgent pointer. */
	sk_send_sigurg(sk);

	/* We may be adding urgent data when the last byte read was
	 * urgent. To do this requires some care. We cannot just ignore
	 * tp->copied_seq since we would read the last urgent byte again
	 * as data, nor can we alter copied_seq until this data arrives
S
Stephen Hemminger 已提交
4994
	 * or we break the semantics of SIOCATMARK (and thus sockatmark())
L
Linus Torvalds 已提交
4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005
	 *
	 * NOTE. Double Dutch. Rendering to plain English: author of comment
	 * above did something sort of 	send("A", MSG_OOB); send("B", MSG_OOB);
	 * and expect that both A and B disappear from stream. This is _wrong_.
	 * Though this happens in BSD with high probability, this is occasional.
	 * Any application relying on this is buggy. Note also, that fix "works"
	 * only in this artificial test. Insert some normal data between A and B and we will
	 * decline of BSD again. Verdict: it is better to remove to trap
	 * buggy users.
	 */
	if (tp->urg_seq == tp->copied_seq && tp->urg_data &&
5006
	    !sock_flag(sk, SOCK_URGINLINE) && tp->copied_seq != tp->rcv_nxt) {
L
Linus Torvalds 已提交
5007 5008 5009
		struct sk_buff *skb = skb_peek(&sk->sk_receive_queue);
		tp->copied_seq++;
		if (skb && !before(tp->copied_seq, TCP_SKB_CB(skb)->end_seq)) {
D
David S. Miller 已提交
5010
			__skb_unlink(skb, &sk->sk_receive_queue);
L
Linus Torvalds 已提交
5011 5012 5013 5014
			__kfree_skb(skb);
		}
	}

5015 5016
	tp->urg_data = TCP_URG_NOTYET;
	tp->urg_seq = ptr;
L
Linus Torvalds 已提交
5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028

	/* Disable header prediction. */
	tp->pred_flags = 0;
}

/* This is the 'fast' part of urgent handling. */
static void tcp_urg(struct sock *sk, struct sk_buff *skb, struct tcphdr *th)
{
	struct tcp_sock *tp = tcp_sk(sk);

	/* Check if we get a new urgent pointer - normally not. */
	if (th->urg)
5029
		tcp_check_urg(sk, th);
L
Linus Torvalds 已提交
5030 5031 5032 5033 5034 5035

	/* Do we wait for any urgent data? - normally not... */
	if (tp->urg_data == TCP_URG_NOTYET) {
		u32 ptr = tp->urg_seq - ntohl(th->seq) + (th->doff * 4) -
			  th->syn;

5036
		/* Is the urgent pointer pointing into this packet? */
L
Linus Torvalds 已提交
5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054
		if (ptr < skb->len) {
			u8 tmp;
			if (skb_copy_bits(skb, ptr, &tmp, 1))
				BUG();
			tp->urg_data = TCP_URG_VALID | tmp;
			if (!sock_flag(sk, SOCK_DEAD))
				sk->sk_data_ready(sk, 0);
		}
	}
}

static int tcp_copy_to_iovec(struct sock *sk, struct sk_buff *skb, int hlen)
{
	struct tcp_sock *tp = tcp_sk(sk);
	int chunk = skb->len - hlen;
	int err;

	local_bh_enable();
5055
	if (skb_csum_unnecessary(skb))
L
Linus Torvalds 已提交
5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070
		err = skb_copy_datagram_iovec(skb, hlen, tp->ucopy.iov, chunk);
	else
		err = skb_copy_and_csum_datagram_iovec(skb, hlen,
						       tp->ucopy.iov);

	if (!err) {
		tp->ucopy.len -= chunk;
		tp->copied_seq += chunk;
		tcp_rcv_space_adjust(sk);
	}

	local_bh_disable();
	return err;
}

5071 5072
static __sum16 __tcp_checksum_complete_user(struct sock *sk,
					    struct sk_buff *skb)
L
Linus Torvalds 已提交
5073
{
5074
	__sum16 result;
L
Linus Torvalds 已提交
5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085

	if (sock_owned_by_user(sk)) {
		local_bh_enable();
		result = __tcp_checksum_complete(skb);
		local_bh_disable();
	} else {
		result = __tcp_checksum_complete(skb);
	}
	return result;
}

5086 5087
static inline int tcp_checksum_complete_user(struct sock *sk,
					     struct sk_buff *skb)
L
Linus Torvalds 已提交
5088
{
5089
	return !skb_csum_unnecessary(skb) &&
5090
	       __tcp_checksum_complete_user(sk, skb);
L
Linus Torvalds 已提交
5091 5092
}

C
Chris Leech 已提交
5093
#ifdef CONFIG_NET_DMA
5094 5095
static int tcp_dma_try_early_copy(struct sock *sk, struct sk_buff *skb,
				  int hlen)
C
Chris Leech 已提交
5096 5097 5098 5099 5100 5101 5102
{
	struct tcp_sock *tp = tcp_sk(sk);
	int chunk = skb->len - hlen;
	int dma_cookie;
	int copied_early = 0;

	if (tp->ucopy.wakeup)
5103
		return 0;
C
Chris Leech 已提交
5104 5105

	if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
5106
		tp->ucopy.dma_chan = dma_find_channel(DMA_MEMCPY);
C
Chris Leech 已提交
5107

5108
	if (tp->ucopy.dma_chan && skb_csum_unnecessary(skb)) {
C
Chris Leech 已提交
5109 5110

		dma_cookie = dma_skb_copy_datagram_iovec(tp->ucopy.dma_chan,
5111 5112 5113
							 skb, hlen,
							 tp->ucopy.iov, chunk,
							 tp->ucopy.pinned_list);
C
Chris Leech 已提交
5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125

		if (dma_cookie < 0)
			goto out;

		tp->ucopy.dma_cookie = dma_cookie;
		copied_early = 1;

		tp->ucopy.len -= chunk;
		tp->copied_seq += chunk;
		tcp_rcv_space_adjust(sk);

		if ((tp->ucopy.len == 0) ||
5126
		    (tcp_flag_word(tcp_hdr(skb)) & TCP_FLAG_PSH) ||
C
Chris Leech 已提交
5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139
		    (atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1))) {
			tp->ucopy.wakeup = 1;
			sk->sk_data_ready(sk, 0);
		}
	} else if (chunk > 0) {
		tp->ucopy.wakeup = 1;
		sk->sk_data_ready(sk, 0);
	}
out:
	return copied_early;
}
#endif /* CONFIG_NET_DMA */

5140 5141 5142 5143 5144 5145
/* Does PAWS and seqno based validation of an incoming segment, flags will
 * play significant role here.
 */
static int tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
			      struct tcphdr *th, int syn_inerr)
{
5146
	u8 *hash_location;
5147 5148 5149
	struct tcp_sock *tp = tcp_sk(sk);

	/* RFC1323: H1. Apply PAWS check first. */
5150 5151
	if (tcp_fast_parse_options(skb, th, tp, &hash_location) &&
	    tp->rx_opt.saw_tstamp &&
5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202
	    tcp_paws_discard(sk, skb)) {
		if (!th->rst) {
			NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSESTABREJECTED);
			tcp_send_dupack(sk, skb);
			goto discard;
		}
		/* Reset is accepted even if it did not pass PAWS. */
	}

	/* Step 1: check sequence number */
	if (!tcp_sequence(tp, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq)) {
		/* RFC793, page 37: "In all states except SYN-SENT, all reset
		 * (RST) segments are validated by checking their SEQ-fields."
		 * And page 69: "If an incoming segment is not acceptable,
		 * an acknowledgment should be sent in reply (unless the RST
		 * bit is set, if so drop the segment and return)".
		 */
		if (!th->rst)
			tcp_send_dupack(sk, skb);
		goto discard;
	}

	/* Step 2: check RST bit */
	if (th->rst) {
		tcp_reset(sk);
		goto discard;
	}

	/* ts_recent update must be made after we are sure that the packet
	 * is in window.
	 */
	tcp_replace_ts_recent(tp, TCP_SKB_CB(skb)->seq);

	/* step 3: check security and precedence [ignored] */

	/* step 4: Check for a SYN in window. */
	if (th->syn && !before(TCP_SKB_CB(skb)->seq, tp->rcv_nxt)) {
		if (syn_inerr)
			TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPABORTONSYN);
		tcp_reset(sk);
		return -1;
	}

	return 1;

discard:
	__kfree_skb(skb);
	return 0;
}

L
Linus Torvalds 已提交
5203
/*
5204
 *	TCP receive function for the ESTABLISHED state.
L
Linus Torvalds 已提交
5205
 *
5206
 *	It is split into a fast path and a slow path. The fast path is
L
Linus Torvalds 已提交
5207 5208
 * 	disabled when:
 *	- A zero window was announced from us - zero window probing
5209
 *        is only handled properly in the slow path.
L
Linus Torvalds 已提交
5210 5211 5212 5213
 *	- Out of order segments arrived.
 *	- Urgent data is expected.
 *	- There is no buffer space left
 *	- Unexpected TCP flags/window values/header lengths are received
5214
 *	  (detected by checking the TCP header against pred_flags)
L
Linus Torvalds 已提交
5215 5216 5217 5218 5219
 *	- Data is sent in both directions. Fast path only supports pure senders
 *	  or pure receivers (this means either the sequence number or the ack
 *	  value must stay constant)
 *	- Unexpected TCP option.
 *
5220
 *	When these conditions are not satisfied it drops into a standard
L
Linus Torvalds 已提交
5221 5222
 *	receive procedure patterned after RFC793 to handle all cases.
 *	The first three cases are guaranteed by proper pred_flags setting,
5223
 *	the rest is checked inline. Fast processing is turned on in
L
Linus Torvalds 已提交
5224 5225 5226 5227 5228 5229
 *	tcp_data_queue when everything is OK.
 */
int tcp_rcv_established(struct sock *sk, struct sk_buff *skb,
			struct tcphdr *th, unsigned len)
{
	struct tcp_sock *tp = tcp_sk(sk);
5230
	int res;
L
Linus Torvalds 已提交
5231 5232 5233

	/*
	 *	Header prediction.
5234
	 *	The code loosely follows the one in the famous
L
Linus Torvalds 已提交
5235
	 *	"30 instruction TCP receive" Van Jacobson mail.
5236 5237
	 *
	 *	Van's trick is to deposit buffers into socket queue
L
Linus Torvalds 已提交
5238 5239 5240 5241
	 *	on a device interrupt, to call tcp_recv function
	 *	on the receive process context and checksum and copy
	 *	the buffer to user space. smart...
	 *
5242
	 *	Our current scheme is not silly either but we take the
L
Linus Torvalds 已提交
5243 5244 5245 5246 5247 5248 5249
	 *	extra cost of the net_bh soft interrupt processing...
	 *	We do checksum and copy also but from device to kernel.
	 */

	tp->rx_opt.saw_tstamp = 0;

	/*	pred_flags is 0xS?10 << 16 + snd_wnd
S
Stephen Hemminger 已提交
5250
	 *	if header_prediction is to be made
L
Linus Torvalds 已提交
5251 5252
	 *	'S' will always be tp->tcp_header_len >> 2
	 *	'?' will be 0 for the fast path, otherwise pred_flags is 0 to
5253
	 *  turn it off	(when there are holes in the receive
L
Linus Torvalds 已提交
5254 5255 5256 5257 5258
	 *	 space for instance)
	 *	PSH flag is ignored.
	 */

	if ((tcp_flag_word(th) & TCP_HP_BITS) == tp->pred_flags &&
5259 5260
	    TCP_SKB_CB(skb)->seq == tp->rcv_nxt &&
	    !after(TCP_SKB_CB(skb)->ack_seq, tp->snd_nxt)) {
L
Linus Torvalds 已提交
5261 5262 5263 5264 5265 5266 5267 5268 5269 5270
		int tcp_header_len = tp->tcp_header_len;

		/* Timestamp header prediction: tcp_header_len
		 * is automatically equal to th->doff*4 due to pred_flags
		 * match.
		 */

		/* Check timestamp */
		if (tcp_header_len == sizeof(struct tcphdr) + TCPOLEN_TSTAMP_ALIGNED) {
			/* No? Slow path! */
5271
			if (!tcp_parse_aligned_timestamp(tp, th))
L
Linus Torvalds 已提交
5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300
				goto slow_path;

			/* If PAWS failed, check it more carefully in slow path */
			if ((s32)(tp->rx_opt.rcv_tsval - tp->rx_opt.ts_recent) < 0)
				goto slow_path;

			/* DO NOT update ts_recent here, if checksum fails
			 * and timestamp was corrupted part, it will result
			 * in a hung connection since we will drop all
			 * future packets due to the PAWS test.
			 */
		}

		if (len <= tcp_header_len) {
			/* Bulk data transfer: sender */
			if (len == tcp_header_len) {
				/* Predicted packet is in window by definition.
				 * seq == rcv_nxt and rcv_wup <= rcv_nxt.
				 * Hence, check seq<=rcv_wup reduces to:
				 */
				if (tcp_header_len ==
				    (sizeof(struct tcphdr) + TCPOLEN_TSTAMP_ALIGNED) &&
				    tp->rcv_nxt == tp->rcv_wup)
					tcp_store_ts_recent(tp);

				/* We know that such packets are checksummed
				 * on entry.
				 */
				tcp_ack(sk, skb, 0);
5301
				__kfree_skb(skb);
5302
				tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
5303 5304
				return 0;
			} else { /* Header too small */
5305
				TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
5306 5307 5308 5309
				goto discard;
			}
		} else {
			int eaten = 0;
C
Chris Leech 已提交
5310
			int copied_early = 0;
L
Linus Torvalds 已提交
5311

C
Chris Leech 已提交
5312 5313 5314 5315 5316 5317 5318 5319
			if (tp->copied_seq == tp->rcv_nxt &&
			    len - tcp_header_len <= tp->ucopy.len) {
#ifdef CONFIG_NET_DMA
				if (tcp_dma_try_early_copy(sk, skb, tcp_header_len)) {
					copied_early = 1;
					eaten = 1;
				}
#endif
5320 5321
				if (tp->ucopy.task == current &&
				    sock_owned_by_user(sk) && !copied_early) {
C
Chris Leech 已提交
5322
					__set_current_state(TASK_RUNNING);
L
Linus Torvalds 已提交
5323

C
Chris Leech 已提交
5324 5325 5326 5327
					if (!tcp_copy_to_iovec(sk, skb, tcp_header_len))
						eaten = 1;
				}
				if (eaten) {
L
Linus Torvalds 已提交
5328 5329 5330 5331 5332 5333 5334 5335 5336 5337
					/* Predicted packet is in window by definition.
					 * seq == rcv_nxt and rcv_wup <= rcv_nxt.
					 * Hence, check seq<=rcv_wup reduces to:
					 */
					if (tcp_header_len ==
					    (sizeof(struct tcphdr) +
					     TCPOLEN_TSTAMP_ALIGNED) &&
					    tp->rcv_nxt == tp->rcv_wup)
						tcp_store_ts_recent(tp);

5338
					tcp_rcv_rtt_measure_ts(sk, skb);
L
Linus Torvalds 已提交
5339 5340 5341

					__skb_pull(skb, tcp_header_len);
					tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq;
5342
					NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPHPHITSTOUSER);
L
Linus Torvalds 已提交
5343
				}
C
Chris Leech 已提交
5344 5345
				if (copied_early)
					tcp_cleanup_rbuf(sk, skb->len);
L
Linus Torvalds 已提交
5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359
			}
			if (!eaten) {
				if (tcp_checksum_complete_user(sk, skb))
					goto csum_error;

				/* Predicted packet is in window by definition.
				 * seq == rcv_nxt and rcv_wup <= rcv_nxt.
				 * Hence, check seq<=rcv_wup reduces to:
				 */
				if (tcp_header_len ==
				    (sizeof(struct tcphdr) + TCPOLEN_TSTAMP_ALIGNED) &&
				    tp->rcv_nxt == tp->rcv_wup)
					tcp_store_ts_recent(tp);

5360
				tcp_rcv_rtt_measure_ts(sk, skb);
L
Linus Torvalds 已提交
5361 5362 5363 5364

				if ((int)skb->truesize > sk->sk_forward_alloc)
					goto step5;

5365
				NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPHPHITS);
L
Linus Torvalds 已提交
5366 5367

				/* Bulk data transfer: receiver */
5368
				__skb_pull(skb, tcp_header_len);
L
Linus Torvalds 已提交
5369
				__skb_queue_tail(&sk->sk_receive_queue, skb);
5370
				skb_set_owner_r(skb, sk);
L
Linus Torvalds 已提交
5371 5372 5373
				tp->rcv_nxt = TCP_SKB_CB(skb)->end_seq;
			}

5374
			tcp_event_data_recv(sk, skb);
L
Linus Torvalds 已提交
5375 5376 5377 5378

			if (TCP_SKB_CB(skb)->ack_seq != tp->snd_una) {
				/* Well, only one small jumplet in fast path... */
				tcp_ack(sk, skb, FLAG_DATA);
5379
				tcp_data_snd_check(sk);
5380
				if (!inet_csk_ack_scheduled(sk))
L
Linus Torvalds 已提交
5381 5382 5383
					goto no_ack;
			}

5384 5385
			if (!copied_early || tp->rcv_nxt != tp->rcv_wup)
				__tcp_ack_snd_check(sk, 0);
L
Linus Torvalds 已提交
5386
no_ack:
C
Chris Leech 已提交
5387 5388 5389 5390 5391
#ifdef CONFIG_NET_DMA
			if (copied_early)
				__skb_queue_tail(&sk->sk_async_wait_queue, skb);
			else
#endif
L
Linus Torvalds 已提交
5392 5393 5394 5395 5396 5397 5398 5399 5400
			if (eaten)
				__kfree_skb(skb);
			else
				sk->sk_data_ready(sk, 0);
			return 0;
		}
	}

slow_path:
5401
	if (len < (th->doff << 2) || tcp_checksum_complete_user(sk, skb))
L
Linus Torvalds 已提交
5402 5403 5404 5405 5406 5407
		goto csum_error;

	/*
	 *	Standard slow path.
	 */

5408 5409 5410
	res = tcp_validate_incoming(sk, skb, th, 1);
	if (res <= 0)
		return -res;
L
Linus Torvalds 已提交
5411 5412

step5:
5413 5414
	if (th->ack && tcp_ack(sk, skb, FLAG_SLOWPATH) < 0)
		goto discard;
L
Linus Torvalds 已提交
5415

5416
	tcp_rcv_rtt_measure_ts(sk, skb);
L
Linus Torvalds 已提交
5417 5418 5419 5420 5421 5422 5423

	/* Process urgent data. */
	tcp_urg(sk, skb, th);

	/* step 7: process the segment text */
	tcp_data_queue(sk, skb);

5424
	tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
5425 5426 5427 5428
	tcp_ack_snd_check(sk);
	return 0;

csum_error:
5429
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
5430 5431 5432 5433 5434 5435 5436 5437 5438

discard:
	__kfree_skb(skb);
	return 0;
}

static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
					 struct tcphdr *th, unsigned len)
{
5439
	u8 *hash_location;
5440
	struct inet_connection_sock *icsk = inet_csk(sk);
5441 5442 5443
	struct tcp_sock *tp = tcp_sk(sk);
	struct tcp_cookie_values *cvp = tp->cookie_values;
	int saved_clamp = tp->rx_opt.mss_clamp;
L
Linus Torvalds 已提交
5444

5445
	tcp_parse_options(skb, &tp->rx_opt, &hash_location, 0);
L
Linus Torvalds 已提交
5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464

	if (th->ack) {
		/* rfc793:
		 * "If the state is SYN-SENT then
		 *    first check the ACK bit
		 *      If the ACK bit is set
		 *	  If SEG.ACK =< ISS, or SEG.ACK > SND.NXT, send
		 *        a reset (unless the RST bit is set, if so drop
		 *        the segment and return)"
		 *
		 *  We do not send data with SYN, so that RFC-correct
		 *  test reduces to:
		 */
		if (TCP_SKB_CB(skb)->ack_seq != tp->snd_nxt)
			goto reset_and_undo;

		if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr &&
		    !between(tp->rx_opt.rcv_tsecr, tp->retrans_stamp,
			     tcp_time_stamp)) {
5465
			NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSACTIVEREJECTED);
L
Linus Torvalds 已提交
5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513
			goto reset_and_undo;
		}

		/* Now ACK is acceptable.
		 *
		 * "If the RST bit is set
		 *    If the ACK was acceptable then signal the user "error:
		 *    connection reset", drop the segment, enter CLOSED state,
		 *    delete TCB, and return."
		 */

		if (th->rst) {
			tcp_reset(sk);
			goto discard;
		}

		/* rfc793:
		 *   "fifth, if neither of the SYN or RST bits is set then
		 *    drop the segment and return."
		 *
		 *    See note below!
		 *                                        --ANK(990513)
		 */
		if (!th->syn)
			goto discard_and_undo;

		/* rfc793:
		 *   "If the SYN bit is on ...
		 *    are acceptable then ...
		 *    (our SYN has been ACKed), change the connection
		 *    state to ESTABLISHED..."
		 */

		TCP_ECN_rcv_synack(tp, th);

		tp->snd_wl1 = TCP_SKB_CB(skb)->seq;
		tcp_ack(sk, skb, FLAG_SLOWPATH);

		/* Ok.. it's good. Set up sequence numbers and
		 * move to established.
		 */
		tp->rcv_nxt = TCP_SKB_CB(skb)->seq + 1;
		tp->rcv_wup = TCP_SKB_CB(skb)->seq + 1;

		/* RFC1323: The window in SYN & SYN/ACK segments is
		 * never scaled.
		 */
		tp->snd_wnd = ntohs(th->window);
5514
		tcp_init_wl(tp, TCP_SKB_CB(skb)->seq);
L
Linus Torvalds 已提交
5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530

		if (!tp->rx_opt.wscale_ok) {
			tp->rx_opt.snd_wscale = tp->rx_opt.rcv_wscale = 0;
			tp->window_clamp = min(tp->window_clamp, 65535U);
		}

		if (tp->rx_opt.saw_tstamp) {
			tp->rx_opt.tstamp_ok	   = 1;
			tp->tcp_header_len =
				sizeof(struct tcphdr) + TCPOLEN_TSTAMP_ALIGNED;
			tp->advmss	    -= TCPOLEN_TSTAMP_ALIGNED;
			tcp_store_ts_recent(tp);
		} else {
			tp->tcp_header_len = sizeof(struct tcphdr);
		}

5531 5532
		if (tcp_is_sack(tp) && sysctl_tcp_fack)
			tcp_enable_fack(tp);
L
Linus Torvalds 已提交
5533

J
John Heffner 已提交
5534
		tcp_mtup_init(sk);
5535
		tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
5536 5537 5538 5539 5540 5541
		tcp_initialize_rcv_mss(sk);

		/* Remember, tcp_poll() does not lock socket!
		 * Change state from SYN-SENT only after copied_seq
		 * is initialized. */
		tp->copied_seq = tp->rcv_nxt;
5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566

		if (cvp != NULL &&
		    cvp->cookie_pair_size > 0 &&
		    tp->rx_opt.cookie_plus > 0) {
			int cookie_size = tp->rx_opt.cookie_plus
					- TCPOLEN_COOKIE_BASE;
			int cookie_pair_size = cookie_size
					     + cvp->cookie_desired;

			/* A cookie extension option was sent and returned.
			 * Note that each incoming SYNACK replaces the
			 * Responder cookie.  The initial exchange is most
			 * fragile, as protection against spoofing relies
			 * entirely upon the sequence and timestamp (above).
			 * This replacement strategy allows the correct pair to
			 * pass through, while any others will be filtered via
			 * Responder verification later.
			 */
			if (sizeof(cvp->cookie_pair) >= cookie_pair_size) {
				memcpy(&cvp->cookie_pair[cvp->cookie_desired],
				       hash_location, cookie_size);
				cvp->cookie_pair_size = cookie_pair_size;
			}
		}

R
Ralf Baechle 已提交
5567
		smp_mb();
L
Linus Torvalds 已提交
5568 5569
		tcp_set_state(sk, TCP_ESTABLISHED);

5570 5571
		security_inet_conn_established(sk, skb);

L
Linus Torvalds 已提交
5572
		/* Make sure socket is routed, for correct metrics.  */
5573
		icsk->icsk_af_ops->rebuild_header(sk);
L
Linus Torvalds 已提交
5574 5575 5576

		tcp_init_metrics(sk);

5577
		tcp_init_congestion_control(sk);
5578

L
Linus Torvalds 已提交
5579 5580 5581 5582 5583 5584 5585 5586
		/* Prevent spurious tcp_cwnd_restart() on first data
		 * packet.
		 */
		tp->lsndtime = tcp_time_stamp;

		tcp_init_buffer_space(sk);

		if (sock_flag(sk, SOCK_KEEPOPEN))
5587
			inet_csk_reset_keepalive_timer(sk, keepalive_time_when(tp));
L
Linus Torvalds 已提交
5588 5589 5590 5591 5592 5593 5594 5595

		if (!tp->rx_opt.snd_wscale)
			__tcp_fast_path_on(tp, tp->snd_wnd);
		else
			tp->pred_flags = 0;

		if (!sock_flag(sk, SOCK_DEAD)) {
			sk->sk_state_change(sk);
5596
			sk_wake_async(sk, SOCK_WAKE_IO, POLL_OUT);
L
Linus Torvalds 已提交
5597 5598
		}

5599 5600 5601
		if (sk->sk_write_pending ||
		    icsk->icsk_accept_queue.rskq_defer_accept ||
		    icsk->icsk_ack.pingpong) {
L
Linus Torvalds 已提交
5602 5603 5604 5605 5606 5607 5608
			/* Save one ACK. Data will be ready after
			 * several ticks, if write_pending is set.
			 *
			 * It may be deleted, but with this feature tcpdumps
			 * look so _wonderfully_ clever, that I was not able
			 * to stand against the temptation 8)     --ANK
			 */
5609
			inet_csk_schedule_ack(sk);
5610 5611
			icsk->icsk_ack.lrcvtime = tcp_time_stamp;
			icsk->icsk_ack.ato	 = TCP_ATO_MIN;
5612 5613
			tcp_incr_quickack(sk);
			tcp_enter_quickack_mode(sk);
5614 5615
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
						  TCP_DELACK_MAX, TCP_RTO_MAX);
L
Linus Torvalds 已提交
5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638

discard:
			__kfree_skb(skb);
			return 0;
		} else {
			tcp_send_ack(sk);
		}
		return -1;
	}

	/* No ACK in the segment */

	if (th->rst) {
		/* rfc793:
		 * "If the RST bit is set
		 *
		 *      Otherwise (no ACK) drop the segment and return."
		 */

		goto discard_and_undo;
	}

	/* PAWS check. */
5639
	if (tp->rx_opt.ts_recent_stamp && tp->rx_opt.saw_tstamp &&
I
Ilpo Järvinen 已提交
5640
	    tcp_paws_reject(&tp->rx_opt, 0))
L
Linus Torvalds 已提交
5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670
		goto discard_and_undo;

	if (th->syn) {
		/* We see SYN without ACK. It is attempt of
		 * simultaneous connect with crossed SYNs.
		 * Particularly, it can be connect to self.
		 */
		tcp_set_state(sk, TCP_SYN_RECV);

		if (tp->rx_opt.saw_tstamp) {
			tp->rx_opt.tstamp_ok = 1;
			tcp_store_ts_recent(tp);
			tp->tcp_header_len =
				sizeof(struct tcphdr) + TCPOLEN_TSTAMP_ALIGNED;
		} else {
			tp->tcp_header_len = sizeof(struct tcphdr);
		}

		tp->rcv_nxt = TCP_SKB_CB(skb)->seq + 1;
		tp->rcv_wup = TCP_SKB_CB(skb)->seq + 1;

		/* RFC1323: The window in SYN & SYN/ACK segments is
		 * never scaled.
		 */
		tp->snd_wnd    = ntohs(th->window);
		tp->snd_wl1    = TCP_SKB_CB(skb)->seq;
		tp->max_window = tp->snd_wnd;

		TCP_ECN_rcv_syn(tp, th);

J
John Heffner 已提交
5671
		tcp_mtup_init(sk);
5672
		tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707
		tcp_initialize_rcv_mss(sk);

		tcp_send_synack(sk);
#if 0
		/* Note, we could accept data and URG from this segment.
		 * There are no obstacles to make this.
		 *
		 * However, if we ignore data in ACKless segments sometimes,
		 * we have no reasons to accept it sometimes.
		 * Also, seems the code doing it in step6 of tcp_rcv_state_process
		 * is not flawless. So, discard packet for sanity.
		 * Uncomment this return to process the data.
		 */
		return -1;
#else
		goto discard;
#endif
	}
	/* "fifth, if neither of the SYN or RST bits is set then
	 * drop the segment and return."
	 */

discard_and_undo:
	tcp_clear_options(&tp->rx_opt);
	tp->rx_opt.mss_clamp = saved_clamp;
	goto discard;

reset_and_undo:
	tcp_clear_options(&tp->rx_opt);
	tp->rx_opt.mss_clamp = saved_clamp;
	return 1;
}

/*
 *	This function implements the receiving procedure of RFC 793 for
5708
 *	all states except ESTABLISHED and TIME_WAIT.
L
Linus Torvalds 已提交
5709 5710 5711
 *	It's called from both tcp_v4_rcv and tcp_v6_rcv and should be
 *	address independent.
 */
5712

L
Linus Torvalds 已提交
5713 5714 5715 5716
int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
			  struct tcphdr *th, unsigned len)
{
	struct tcp_sock *tp = tcp_sk(sk);
5717
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
5718
	int queued = 0;
5719
	int res;
L
Linus Torvalds 已提交
5720 5721 5722 5723 5724 5725 5726 5727

	tp->rx_opt.saw_tstamp = 0;

	switch (sk->sk_state) {
	case TCP_CLOSE:
		goto discard;

	case TCP_LISTEN:
S
Stephen Hemminger 已提交
5728
		if (th->ack)
L
Linus Torvalds 已提交
5729 5730
			return 1;

S
Stephen Hemminger 已提交
5731
		if (th->rst)
L
Linus Torvalds 已提交
5732 5733
			goto discard;

S
Stephen Hemminger 已提交
5734
		if (th->syn) {
5735
			if (icsk->icsk_af_ops->conn_request(sk, skb) < 0)
L
Linus Torvalds 已提交
5736 5737
				return 1;

5738 5739
			/* Now we have several options: In theory there is
			 * nothing else in the frame. KA9Q has an option to
L
Linus Torvalds 已提交
5740
			 * send data with the syn, BSD accepts data with the
5741 5742 5743
			 * syn up to the [to be] advertised window and
			 * Solaris 2.1 gives you a protocol error. For now
			 * we just ignore it, that fits the spec precisely
L
Linus Torvalds 已提交
5744 5745 5746
			 * and avoids incompatibilities. It would be nice in
			 * future to drop through and process the data.
			 *
5747
			 * Now that TTCP is starting to be used we ought to
L
Linus Torvalds 已提交
5748 5749
			 * queue this data.
			 * But, this leaves one open to an easy denial of
5750
			 * service attack, and SYN cookies can't defend
L
Linus Torvalds 已提交
5751
			 * against this problem. So, we drop the data
5752 5753
			 * in the interest of security over speed unless
			 * it's still in use.
L
Linus Torvalds 已提交
5754
			 */
5755 5756
			kfree_skb(skb);
			return 0;
L
Linus Torvalds 已提交
5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767
		}
		goto discard;

	case TCP_SYN_SENT:
		queued = tcp_rcv_synsent_state_process(sk, skb, th, len);
		if (queued >= 0)
			return queued;

		/* Do step6 onward by hand. */
		tcp_urg(sk, skb, th);
		__kfree_skb(skb);
5768
		tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
5769 5770 5771
		return 0;
	}

5772 5773 5774
	res = tcp_validate_incoming(sk, skb, th, 0);
	if (res <= 0)
		return -res;
L
Linus Torvalds 已提交
5775 5776 5777

	/* step 5: check the ACK field */
	if (th->ack) {
5778
		int acceptable = tcp_ack(sk, skb, FLAG_SLOWPATH) > 0;
L
Linus Torvalds 已提交
5779

S
Stephen Hemminger 已提交
5780
		switch (sk->sk_state) {
L
Linus Torvalds 已提交
5781 5782 5783
		case TCP_SYN_RECV:
			if (acceptable) {
				tp->copied_seq = tp->rcv_nxt;
R
Ralf Baechle 已提交
5784
				smp_mb();
L
Linus Torvalds 已提交
5785 5786 5787 5788 5789 5790 5791 5792
				tcp_set_state(sk, TCP_ESTABLISHED);
				sk->sk_state_change(sk);

				/* Note, that this wakeup is only for marginal
				 * crossed SYN case. Passively open sockets
				 * are not waked up, because sk->sk_sleep ==
				 * NULL and sk->sk_socket == NULL.
				 */
5793 5794
				if (sk->sk_socket)
					sk_wake_async(sk,
5795
						      SOCK_WAKE_IO, POLL_OUT);
L
Linus Torvalds 已提交
5796 5797 5798 5799

				tp->snd_una = TCP_SKB_CB(skb)->ack_seq;
				tp->snd_wnd = ntohs(th->window) <<
					      tp->rx_opt.snd_wscale;
5800
				tcp_init_wl(tp, TCP_SKB_CB(skb)->seq);
L
Linus Torvalds 已提交
5801 5802 5803

				/* tcp_ack considers this ACK as duplicate
				 * and does not calculate rtt.
D
Damian Lukowski 已提交
5804
				 * Force it here.
L
Linus Torvalds 已提交
5805
				 */
D
Damian Lukowski 已提交
5806
				tcp_ack_update_rtt(sk, 0, 0);
L
Linus Torvalds 已提交
5807 5808 5809 5810 5811 5812 5813

				if (tp->rx_opt.tstamp_ok)
					tp->advmss -= TCPOLEN_TSTAMP_ALIGNED;

				/* Make sure socket is routed, for
				 * correct metrics.
				 */
5814
				icsk->icsk_af_ops->rebuild_header(sk);
L
Linus Torvalds 已提交
5815 5816 5817

				tcp_init_metrics(sk);

5818
				tcp_init_congestion_control(sk);
5819

L
Linus Torvalds 已提交
5820 5821 5822 5823 5824
				/* Prevent spurious tcp_cwnd_restart() on
				 * first data packet.
				 */
				tp->lsndtime = tcp_time_stamp;

J
John Heffner 已提交
5825
				tcp_mtup_init(sk);
L
Linus Torvalds 已提交
5826 5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837
				tcp_initialize_rcv_mss(sk);
				tcp_init_buffer_space(sk);
				tcp_fast_path_on(tp);
			} else {
				return 1;
			}
			break;

		case TCP_FIN_WAIT1:
			if (tp->snd_una == tp->write_seq) {
				tcp_set_state(sk, TCP_FIN_WAIT2);
				sk->sk_shutdown |= SEND_SHUTDOWN;
E
Eric Dumazet 已提交
5838
				dst_confirm(__sk_dst_get(sk));
L
Linus Torvalds 已提交
5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849

				if (!sock_flag(sk, SOCK_DEAD))
					/* Wake up lingering close() */
					sk->sk_state_change(sk);
				else {
					int tmo;

					if (tp->linger2 < 0 ||
					    (TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(skb)->seq &&
					     after(TCP_SKB_CB(skb)->end_seq - th->fin, tp->rcv_nxt))) {
						tcp_done(sk);
5850
						NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPABORTONDATA);
L
Linus Torvalds 已提交
5851 5852 5853
						return 1;
					}

5854
					tmo = tcp_fin_time(sk);
L
Linus Torvalds 已提交
5855
					if (tmo > TCP_TIMEWAIT_LEN) {
5856
						inet_csk_reset_keepalive_timer(sk, tmo - TCP_TIMEWAIT_LEN);
L
Linus Torvalds 已提交
5857 5858 5859 5860 5861 5862 5863
					} else if (th->fin || sock_owned_by_user(sk)) {
						/* Bad case. We could lose such FIN otherwise.
						 * It is not a big problem, but it looks confusing
						 * and not so rare event. We still can lose it now,
						 * if it spins in bh_lock_sock(), but it is really
						 * marginal case.
						 */
5864
						inet_csk_reset_keepalive_timer(sk, tmo);
L
Linus Torvalds 已提交
5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903
					} else {
						tcp_time_wait(sk, TCP_FIN_WAIT2, tmo);
						goto discard;
					}
				}
			}
			break;

		case TCP_CLOSING:
			if (tp->snd_una == tp->write_seq) {
				tcp_time_wait(sk, TCP_TIME_WAIT, 0);
				goto discard;
			}
			break;

		case TCP_LAST_ACK:
			if (tp->snd_una == tp->write_seq) {
				tcp_update_metrics(sk);
				tcp_done(sk);
				goto discard;
			}
			break;
		}
	} else
		goto discard;

	/* step 6: check the URG bit */
	tcp_urg(sk, skb, th);

	/* step 7: process the segment text */
	switch (sk->sk_state) {
	case TCP_CLOSE_WAIT:
	case TCP_CLOSING:
	case TCP_LAST_ACK:
		if (!before(TCP_SKB_CB(skb)->seq, tp->rcv_nxt))
			break;
	case TCP_FIN_WAIT1:
	case TCP_FIN_WAIT2:
		/* RFC 793 says to queue data in these states,
5904
		 * RFC 1122 says we MUST send a reset.
L
Linus Torvalds 已提交
5905 5906 5907 5908 5909
		 * BSD 4.4 also does reset.
		 */
		if (sk->sk_shutdown & RCV_SHUTDOWN) {
			if (TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(skb)->seq &&
			    after(TCP_SKB_CB(skb)->end_seq - th->fin, tp->rcv_nxt)) {
5910
				NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPABORTONDATA);
L
Linus Torvalds 已提交
5911 5912 5913 5914 5915
				tcp_reset(sk);
				return 1;
			}
		}
		/* Fall through */
5916
	case TCP_ESTABLISHED:
L
Linus Torvalds 已提交
5917 5918 5919 5920 5921 5922 5923
		tcp_data_queue(sk, skb);
		queued = 1;
		break;
	}

	/* tcp_data could move socket to TIME-WAIT */
	if (sk->sk_state != TCP_CLOSE) {
5924
		tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
5925 5926 5927
		tcp_ack_snd_check(sk);
	}

5928
	if (!queued) {
L
Linus Torvalds 已提交
5929 5930 5931 5932 5933 5934 5935 5936
discard:
		__kfree_skb(skb);
	}
	return 0;
}

EXPORT_SYMBOL(sysctl_tcp_ecn);
EXPORT_SYMBOL(sysctl_tcp_reordering);
5937
EXPORT_SYMBOL(sysctl_tcp_adv_win_scale);
L
Linus Torvalds 已提交
5938
EXPORT_SYMBOL(tcp_parse_options);
5939 5940 5941
#ifdef CONFIG_TCP_MD5SIG
EXPORT_SYMBOL(tcp_parse_md5sig_option);
#endif
L
Linus Torvalds 已提交
5942 5943
EXPORT_SYMBOL(tcp_rcv_established);
EXPORT_SYMBOL(tcp_rcv_state_process);
S
Stephen Hemminger 已提交
5944
EXPORT_SYMBOL(tcp_initialize_rcv_mss);