tcp_input.c 182.9 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
 *		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
 */

64 65
#define pr_fmt(fmt) "TCP: " fmt

L
Linus Torvalds 已提交
66
#include <linux/mm.h>
67
#include <linux/slab.h>
L
Linus Torvalds 已提交
68 69
#include <linux/module.h>
#include <linux/sysctl.h>
70
#include <linux/kernel.h>
71
#include <linux/prefetch.h>
72
#include <net/dst.h>
L
Linus Torvalds 已提交
73 74 75 76
#include <net/tcp.h>
#include <net/inet_common.h>
#include <linux/ipsec.h>
#include <asm/unaligned.h>
77
#include <linux/errqueue.h>
L
Linus Torvalds 已提交
78

79 80 81
int sysctl_tcp_timestamps __read_mostly = 1;
int sysctl_tcp_window_scaling __read_mostly = 1;
int sysctl_tcp_sack __read_mostly = 1;
Y
Yuchung Cheng 已提交
82
int sysctl_tcp_fack __read_mostly;
83
int sysctl_tcp_max_reordering __read_mostly = 300;
84 85
int sysctl_tcp_dsack __read_mostly = 1;
int sysctl_tcp_app_win __read_mostly = 31;
86
int sysctl_tcp_adv_win_scale __read_mostly = 1;
E
Eric Dumazet 已提交
87
EXPORT_SYMBOL(sysctl_tcp_adv_win_scale);
88
EXPORT_SYMBOL(sysctl_tcp_timestamps);
L
Linus Torvalds 已提交
89

E
Eric Dumazet 已提交
90
/* rfc5961 challenge ack rate limiting */
91
int sysctl_tcp_challenge_ack_limit = 1000;
E
Eric Dumazet 已提交
92

93 94 95
int sysctl_tcp_stdurg __read_mostly;
int sysctl_tcp_rfc1337 __read_mostly;
int sysctl_tcp_max_orphans __read_mostly = NR_FILE;
96
int sysctl_tcp_frto __read_mostly = 2;
97
int sysctl_tcp_min_rtt_wlen __read_mostly = 300;
98
int sysctl_tcp_moderate_rcvbuf __read_mostly = 1;
N
Nandita Dukkipati 已提交
99
int sysctl_tcp_early_retrans __read_mostly = 3;
100
int sysctl_tcp_invalid_ratelimit __read_mostly = HZ/2;
L
Linus Torvalds 已提交
101 102 103 104 105 106 107 108

#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				*/
109
#define FLAG_LOST_RETRANS	0x80 /* This ACK marks some retransmission lost */
L
Linus Torvalds 已提交
110
#define FLAG_SLOWPATH		0x100 /* Do not skip RFC checks for window update.*/
Y
Yuchung Cheng 已提交
111
#define FLAG_ORIG_SACK_ACKED	0x200 /* Never retransmitted data are (s)acked	*/
112
#define FLAG_SND_UNA_ADVANCED	0x400 /* Snd_una was changed (!= FLAG_DATA_ACKED) */
113
#define FLAG_DSACKING_ACK	0x800 /* SACK blocks contained D-SACK info */
114
#define FLAG_SACK_RENEGING	0x2000 /* snd_una advanced to a sacked seq */
115
#define FLAG_UPDATE_TS_RECENT	0x4000 /* tcp_replace_ts_recent() */
L
Linus Torvalds 已提交
116 117 118 119 120 121 122

#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)

#define TCP_REMNANT (TCP_FLAG_FIN|TCP_FLAG_URG|TCP_FLAG_SYN|TCP_FLAG_PSH)
123
#define TCP_HP_BITS (~(TCP_RESERVED_BITS|TCP_FLAG_PSH))
L
Linus Torvalds 已提交
124

125 126 127 128
#define REXMIT_NONE	0 /* no loss recovery to do */
#define REXMIT_LOST	1 /* retransmit packets marked lost */
#define REXMIT_NEW	2 /* FRTO-style transmit of unsent/new packets */

129 130
static void tcp_gro_dev_warn(struct sock *sk, const struct sk_buff *skb,
			     unsigned int len)
131 132 133 134 135 136 137 138 139 140
{
	static bool __once __read_mostly;

	if (!__once) {
		struct net_device *dev;

		__once = true;

		rcu_read_lock();
		dev = dev_get_by_index_rcu(sock_net(sk), skb->skb_iif);
141 142 143
		if (!dev || len >= dev->mtu)
			pr_warn("%s: Driver has suspect GRO implementation, TCP performance may be compromised.\n",
				dev ? dev->name : "Unknown driver");
144 145 146 147
		rcu_read_unlock();
	}
}

148
/* Adapt the MSS value used to make delayed ack decision to the
L
Linus Torvalds 已提交
149
 * real world.
150
 */
151
static void tcp_measure_rcv_mss(struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
152
{
153
	struct inet_connection_sock *icsk = inet_csk(sk);
154
	const unsigned int lss = icsk->icsk_ack.last_seg_size;
155
	unsigned int len;
L
Linus Torvalds 已提交
156

157
	icsk->icsk_ack.last_seg_size = 0;
L
Linus Torvalds 已提交
158 159 160 161

	/* skb->len may jitter because of SACKs, even if peer
	 * sends good full-sized frames.
	 */
162
	len = skb_shinfo(skb)->gso_size ? : skb->len;
163
	if (len >= icsk->icsk_ack.rcv_mss) {
164 165
		icsk->icsk_ack.rcv_mss = min_t(unsigned int, len,
					       tcp_sk(sk)->advmss);
166 167 168 169
		/* Account for possibly-removed options */
		if (unlikely(len > icsk->icsk_ack.rcv_mss +
				   MAX_TCP_OPTION_SPACE))
			tcp_gro_dev_warn(sk, skb, len);
L
Linus Torvalds 已提交
170 171 172 173 174 175
	} else {
		/* Otherwise, we make more careful check taking into account,
		 * that SACKs block is variable.
		 *
		 * "len" is invariant segment length, including TCP header.
		 */
176
		len += skb->data - skb_transport_header(skb);
177
		if (len >= TCP_MSS_DEFAULT + sizeof(struct tcphdr) ||
L
Linus Torvalds 已提交
178 179 180 181 182 183
		    /* 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) &&
184
		     !(tcp_flag_word(tcp_hdr(skb)) & TCP_REMNANT))) {
L
Linus Torvalds 已提交
185 186 187 188
			/* Subtract also invariant (if peer is RFC compliant),
			 * tcp header plus fixed timestamp option length.
			 * Resulting "len" is MSS free of SACK jitter.
			 */
189 190
			len -= tcp_sk(sk)->tcp_header_len;
			icsk->icsk_ack.last_seg_size = len;
L
Linus Torvalds 已提交
191
			if (len == lss) {
192
				icsk->icsk_ack.rcv_mss = len;
L
Linus Torvalds 已提交
193 194 195
				return;
			}
		}
196 197
		if (icsk->icsk_ack.pending & ICSK_ACK_PUSHED)
			icsk->icsk_ack.pending |= ICSK_ACK_PUSHED2;
198
		icsk->icsk_ack.pending |= ICSK_ACK_PUSHED;
L
Linus Torvalds 已提交
199 200 201
	}
}

202
static void tcp_incr_quickack(struct sock *sk)
L
Linus Torvalds 已提交
203
{
204
	struct inet_connection_sock *icsk = inet_csk(sk);
205
	unsigned int quickacks = tcp_sk(sk)->rcv_wnd / (2 * icsk->icsk_ack.rcv_mss);
L
Linus Torvalds 已提交
206

207 208
	if (quickacks == 0)
		quickacks = 2;
209 210
	if (quickacks > icsk->icsk_ack.quick)
		icsk->icsk_ack.quick = min(quickacks, TCP_MAX_QUICKACKS);
L
Linus Torvalds 已提交
211 212
}

213
static void tcp_enter_quickack_mode(struct sock *sk)
L
Linus Torvalds 已提交
214
{
215 216 217 218
	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
219 220 221 222 223 224
}

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

225
static bool tcp_in_quickack_mode(struct sock *sk)
L
Linus Torvalds 已提交
226
{
227
	const struct inet_connection_sock *icsk = inet_csk(sk);
228
	const struct dst_entry *dst = __sk_dst_get(sk);
E
Eric Dumazet 已提交
229

230 231
	return (dst && dst_metric(dst, RTAX_QUICKACK)) ||
		(icsk->icsk_ack.quick && !icsk->icsk_ack.pingpong);
L
Linus Torvalds 已提交
232 233
}

234
static void tcp_ecn_queue_cwr(struct tcp_sock *tp)
235
{
236
	if (tp->ecn_flags & TCP_ECN_OK)
237 238 239
		tp->ecn_flags |= TCP_ECN_QUEUE_CWR;
}

240
static void tcp_ecn_accept_cwr(struct tcp_sock *tp, const struct sk_buff *skb)
241 242 243 244 245
{
	if (tcp_hdr(skb)->cwr)
		tp->ecn_flags &= ~TCP_ECN_DEMAND_CWR;
}

246
static void tcp_ecn_withdraw_cwr(struct tcp_sock *tp)
247 248 249 250
{
	tp->ecn_flags &= ~TCP_ECN_DEMAND_CWR;
}

251
static void __tcp_ecn_check_ce(struct tcp_sock *tp, const struct sk_buff *skb)
252
{
253
	switch (TCP_SKB_CB(skb)->ip_dsfield & INET_ECN_MASK) {
254
	case INET_ECN_NOT_ECT:
255
		/* Funny extension: if ECT is not set on a segment,
256 257 258 259
		 * and we already seen ECT on a previous segment,
		 * it is probably a retransmit.
		 */
		if (tp->ecn_flags & TCP_ECN_SEEN)
260
			tcp_enter_quickack_mode((struct sock *)tp);
261 262
		break;
	case INET_ECN_CE:
263 264 265
		if (tcp_ca_needs_ecn((struct sock *)tp))
			tcp_ca_event((struct sock *)tp, CA_EVENT_ECN_IS_CE);

266 267 268 269 270
		if (!(tp->ecn_flags & TCP_ECN_DEMAND_CWR)) {
			/* Better not delay acks, sender can have a very low cwnd */
			tcp_enter_quickack_mode((struct sock *)tp);
			tp->ecn_flags |= TCP_ECN_DEMAND_CWR;
		}
271 272
		tp->ecn_flags |= TCP_ECN_SEEN;
		break;
273
	default:
274 275
		if (tcp_ca_needs_ecn((struct sock *)tp))
			tcp_ca_event((struct sock *)tp, CA_EVENT_ECN_NO_CE);
276
		tp->ecn_flags |= TCP_ECN_SEEN;
277
		break;
278 279 280
	}
}

281 282 283 284 285 286 287
static void tcp_ecn_check_ce(struct tcp_sock *tp, const struct sk_buff *skb)
{
	if (tp->ecn_flags & TCP_ECN_OK)
		__tcp_ecn_check_ce(tp, skb);
}

static void tcp_ecn_rcv_synack(struct tcp_sock *tp, const struct tcphdr *th)
288
{
289
	if ((tp->ecn_flags & TCP_ECN_OK) && (!th->ece || th->cwr))
290 291 292
		tp->ecn_flags &= ~TCP_ECN_OK;
}

293
static void tcp_ecn_rcv_syn(struct tcp_sock *tp, const struct tcphdr *th)
294
{
295
	if ((tp->ecn_flags & TCP_ECN_OK) && (!th->ece || !th->cwr))
296 297 298
		tp->ecn_flags &= ~TCP_ECN_OK;
}

299
static bool tcp_ecn_rcv_ecn_echo(const struct tcp_sock *tp, const struct tcphdr *th)
300
{
301
	if (th->ece && !th->syn && (tp->ecn_flags & TCP_ECN_OK))
E
Eric Dumazet 已提交
302 303
		return true;
	return false;
304 305
}

L
Linus Torvalds 已提交
306 307 308 309 310
/* Buffer size and advertised window tuning.
 *
 * 1. Tuning sk->sk_sndbuf, when connection enters established state.
 */

E
Eric Dumazet 已提交
311
static void tcp_sndbuf_expand(struct sock *sk)
L
Linus Torvalds 已提交
312
{
E
Eric Dumazet 已提交
313
	const struct tcp_sock *tp = tcp_sk(sk);
314
	const struct tcp_congestion_ops *ca_ops = inet_csk(sk)->icsk_ca_ops;
E
Eric Dumazet 已提交
315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334
	int sndmem, per_mss;
	u32 nr_segs;

	/* Worst case is non GSO/TSO : each frame consumes one skb
	 * and skb->head is kmalloced using power of two area of memory
	 */
	per_mss = max_t(u32, tp->rx_opt.mss_clamp, tp->mss_cache) +
		  MAX_TCP_HEADER +
		  SKB_DATA_ALIGN(sizeof(struct skb_shared_info));

	per_mss = roundup_pow_of_two(per_mss) +
		  SKB_DATA_ALIGN(sizeof(struct sk_buff));

	nr_segs = max_t(u32, TCP_INIT_CWND, tp->snd_cwnd);
	nr_segs = max_t(u32, nr_segs, tp->reordering + 1);

	/* Fast Recovery (RFC 5681 3.2) :
	 * Cubic needs 1.7 factor, rounded to 2 to include
	 * extra cushion (application might react slowly to POLLOUT)
	 */
335 336
	sndmem = ca_ops->sndbuf_expand ? ca_ops->sndbuf_expand(sk) : 2;
	sndmem *= nr_segs * per_mss;
L
Linus Torvalds 已提交
337

338 339
	if (sk->sk_sndbuf < sndmem)
		sk->sk_sndbuf = min(sndmem, sysctl_tcp_wmem[2]);
L
Linus Torvalds 已提交
340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362
}

/* 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
363
 * window and then starts to feed us spaghetti. But it should work
L
Linus Torvalds 已提交
364 365 366 367
 * in common situations. Otherwise, we have to rely on queue collapsing.
 */

/* Slow part of check#2. */
368
static int __tcp_grow_window(const struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
369
{
370
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
371
	/* Optimize this! */
372 373
	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
374 375 376

	while (tp->rcv_ssthresh <= window) {
		if (truesize <= skb->len)
377
			return 2 * inet_csk(sk)->icsk_ack.rcv_mss;
L
Linus Torvalds 已提交
378 379 380 381 382 383 384

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

385
static void tcp_grow_window(struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
386
{
387 388
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
389 390 391
	/* Check #1 */
	if (tp->rcv_ssthresh < tp->window_clamp &&
	    (int)tp->rcv_ssthresh < tcp_space(sk) &&
392
	    !tcp_under_memory_pressure(sk)) {
L
Linus Torvalds 已提交
393 394 395 396 397 398
		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)
399
			incr = 2 * tp->advmss;
L
Linus Torvalds 已提交
400
		else
401
			incr = __tcp_grow_window(sk, skb);
L
Linus Torvalds 已提交
402 403

		if (incr) {
404
			incr = max_t(int, incr, 2 * skb->len);
405 406
			tp->rcv_ssthresh = min(tp->rcv_ssthresh + incr,
					       tp->window_clamp);
407
			inet_csk(sk)->icsk_ack.quick |= 1;
L
Linus Torvalds 已提交
408 409 410 411 412 413 414
		}
	}
}

/* 3. Tuning rcvbuf, when connection enters established state. */
static void tcp_fixup_rcvbuf(struct sock *sk)
{
415 416
	u32 mss = tcp_sk(sk)->advmss;
	int rcvmem;
L
Linus Torvalds 已提交
417

418 419
	rcvmem = 2 * SKB_TRUESIZE(mss + MAX_TCP_HEADER) *
		 tcp_default_init_rwnd(mss);
420

E
Eric Dumazet 已提交
421 422 423 424 425 426
	/* Dynamic Right Sizing (DRS) has 2 to 3 RTT latency
	 * Allow enough cushion so that sender is not limited by our window
	 */
	if (sysctl_tcp_moderate_rcvbuf)
		rcvmem <<= 2;

427 428
	if (sk->sk_rcvbuf < rcvmem)
		sk->sk_rcvbuf = min(rcvmem, sysctl_tcp_rmem[2]);
L
Linus Torvalds 已提交
429 430
}

S
Stephen Hemminger 已提交
431
/* 4. Try to fixup all. It is made immediately after connection enters
L
Linus Torvalds 已提交
432 433
 *    established state.
 */
434
void tcp_init_buffer_space(struct sock *sk)
L
Linus Torvalds 已提交
435 436 437 438 439 440 441
{
	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))
E
Eric Dumazet 已提交
442
		tcp_sndbuf_expand(sk);
L
Linus Torvalds 已提交
443 444

	tp->rcvq_space.space = tp->rcv_wnd;
E
Eric Dumazet 已提交
445 446
	tp->rcvq_space.time = tcp_time_stamp;
	tp->rcvq_space.seq = tp->copied_seq;
L
Linus Torvalds 已提交
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469

	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. */
470
static void tcp_clamp_window(struct sock *sk)
L
Linus Torvalds 已提交
471
{
472
	struct tcp_sock *tp = tcp_sk(sk);
473
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
474

475
	icsk->icsk_ack.quick = 0;
L
Linus Torvalds 已提交
476

477 478
	if (sk->sk_rcvbuf < sysctl_tcp_rmem[2] &&
	    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) &&
479
	    !tcp_under_memory_pressure(sk) &&
480
	    sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)) {
481 482
		sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc),
				    sysctl_tcp_rmem[2]);
L
Linus Torvalds 已提交
483
	}
484
	if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf)
485
		tp->rcv_ssthresh = min(tp->window_clamp, 2U * tp->advmss);
L
Linus Torvalds 已提交
486 487
}

S
Stephen Hemminger 已提交
488 489 490 491 492 493 494 495 496
/* 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)
{
497
	const struct tcp_sock *tp = tcp_sk(sk);
S
Stephen Hemminger 已提交
498 499
	unsigned int hint = min_t(unsigned int, tp->advmss, tp->mss_cache);

500
	hint = min(hint, tp->rcv_wnd / 2);
501
	hint = min(hint, TCP_MSS_DEFAULT);
S
Stephen Hemminger 已提交
502 503 504 505
	hint = max(hint, TCP_MIN_MSS);

	inet_csk(sk)->icsk_ack.rcv_mss = hint;
}
E
Eric Dumazet 已提交
506
EXPORT_SYMBOL(tcp_initialize_rcv_mss);
S
Stephen Hemminger 已提交
507

L
Linus Torvalds 已提交
508 509 510 511
/* 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.
512
 * <http://public.lanl.gov/radiant/pubs.html#DRS>
L
Linus Torvalds 已提交
513 514
 *
 * More detail on this code can be found at
515
 * <http://staff.psc.edu/jheffner/>,
L
Linus Torvalds 已提交
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533
 * 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
534
		 * non-timestamp case, we do not smooth things out
S
Stephen Hemminger 已提交
535
		 * else with timestamps disabled convergence takes too
L
Linus Torvalds 已提交
536 537 538 539 540
		 * long.
		 */
		if (!win_dep) {
			m -= (new_sample >> 3);
			new_sample += m;
541 542 543 544 545
		} else {
			m <<= 3;
			if (m < new_sample)
				new_sample = m;
		}
L
Linus Torvalds 已提交
546
	} else {
S
Stephen Hemminger 已提交
547
		/* No previous measure. */
L
Linus Torvalds 已提交
548 549 550 551 552 553 554 555 556 557 558 559 560
		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;
561
	tcp_rcv_rtt_update(tp, tcp_time_stamp - tp->rcv_rtt_est.time, 1);
L
Linus Torvalds 已提交
562 563 564 565 566 567

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

568 569
static inline void tcp_rcv_rtt_measure_ts(struct sock *sk,
					  const struct sk_buff *skb)
L
Linus Torvalds 已提交
570
{
571
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
572 573
	if (tp->rx_opt.rcv_tsecr &&
	    (TCP_SKB_CB(skb)->end_seq -
574
	     TCP_SKB_CB(skb)->seq >= inet_csk(sk)->icsk_ack.rcv_mss))
L
Linus Torvalds 已提交
575 576 577 578 579 580 581 582 583 584 585
		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;
E
Eric Dumazet 已提交
586
	int copied;
587

L
Linus Torvalds 已提交
588
	time = tcp_time_stamp - tp->rcvq_space.time;
589
	if (time < (tp->rcv_rtt_est.rtt >> 3) || tp->rcv_rtt_est.rtt == 0)
L
Linus Torvalds 已提交
590
		return;
591

E
Eric Dumazet 已提交
592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608
	/* Number of bytes copied to user in last RTT */
	copied = tp->copied_seq - tp->rcvq_space.seq;
	if (copied <= tp->rcvq_space.space)
		goto new_measure;

	/* A bit of theory :
	 * copied = bytes received in previous RTT, our base window
	 * To cope with packet losses, we need a 2x factor
	 * To cope with slow start, and sender growing its cwin by 100 %
	 * every RTT, we need a 4x factor, because the ACK we are sending
	 * now is for the next RTT, not the current one :
	 * <prev RTT . ><current RTT .. ><next RTT .... >
	 */

	if (sysctl_tcp_moderate_rcvbuf &&
	    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK)) {
		int rcvwin, rcvmem, rcvbuf;
L
Linus Torvalds 已提交
609

E
Eric Dumazet 已提交
610 611 612 613
		/* minimal window to cope with packet losses, assuming
		 * steady state. Add some cushion because of small variations.
		 */
		rcvwin = (copied << 1) + 16 * tp->advmss;
L
Linus Torvalds 已提交
614

E
Eric Dumazet 已提交
615 616 617 618 619 620 621 622 623 624 625 626 627
		/* If rate increased by 25%,
		 *	assume slow start, rcvwin = 3 * copied
		 * If rate increased by 50%,
		 *	assume sender can use 2x growth, rcvwin = 4 * copied
		 */
		if (copied >=
		    tp->rcvq_space.space + (tp->rcvq_space.space >> 2)) {
			if (copied >=
			    tp->rcvq_space.space + (tp->rcvq_space.space >> 1))
				rcvwin <<= 1;
			else
				rcvwin += (rcvwin >> 1);
		}
L
Linus Torvalds 已提交
628

E
Eric Dumazet 已提交
629 630 631
		rcvmem = SKB_TRUESIZE(tp->advmss + MAX_TCP_HEADER);
		while (tcp_win_from_space(rcvmem) < tp->advmss)
			rcvmem += 128;
L
Linus Torvalds 已提交
632

E
Eric Dumazet 已提交
633 634 635
		rcvbuf = min(rcvwin / tp->advmss * rcvmem, sysctl_tcp_rmem[2]);
		if (rcvbuf > sk->sk_rcvbuf) {
			sk->sk_rcvbuf = rcvbuf;
L
Linus Torvalds 已提交
636

E
Eric Dumazet 已提交
637 638
			/* Make the window clamp follow along.  */
			tp->window_clamp = rcvwin;
L
Linus Torvalds 已提交
639 640
		}
	}
E
Eric Dumazet 已提交
641
	tp->rcvq_space.space = copied;
642

L
Linus Torvalds 已提交
643 644 645 646 647 648 649 650 651 652 653 654 655 656 657
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
 */
658
static void tcp_event_data_recv(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
659
{
660
	struct tcp_sock *tp = tcp_sk(sk);
661
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
662 663
	u32 now;

664
	inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
665

666
	tcp_measure_rcv_mss(sk, skb);
L
Linus Torvalds 已提交
667 668

	tcp_rcv_rtt_measure(tp);
669

L
Linus Torvalds 已提交
670 671
	now = tcp_time_stamp;

672
	if (!icsk->icsk_ack.ato) {
L
Linus Torvalds 已提交
673 674 675
		/* The _first_ data packet received, initialize
		 * delayed ACK engine.
		 */
676 677
		tcp_incr_quickack(sk);
		icsk->icsk_ack.ato = TCP_ATO_MIN;
L
Linus Torvalds 已提交
678
	} else {
679
		int m = now - icsk->icsk_ack.lrcvtime;
L
Linus Torvalds 已提交
680

681
		if (m <= TCP_ATO_MIN / 2) {
L
Linus Torvalds 已提交
682
			/* The fastest case is the first. */
683 684 685 686 687 688
			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
689
			/* Too long gap. Apparently sender failed to
L
Linus Torvalds 已提交
690 691
			 * restart window, so that we send ACKs quickly.
			 */
692
			tcp_incr_quickack(sk);
693
			sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
694 695
		}
	}
696
	icsk->icsk_ack.lrcvtime = now;
L
Linus Torvalds 已提交
697

698
	tcp_ecn_check_ce(tp, skb);
L
Linus Torvalds 已提交
699 700

	if (skb->len >= 128)
701
		tcp_grow_window(sk, skb);
L
Linus Torvalds 已提交
702 703 704 705 706 707 708 709 710 711 712
}

/* 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
 */
713
static void tcp_rtt_estimator(struct sock *sk, long mrtt_us)
L
Linus Torvalds 已提交
714
{
715
	struct tcp_sock *tp = tcp_sk(sk);
716 717
	long m = mrtt_us; /* RTT */
	u32 srtt = tp->srtt_us;
L
Linus Torvalds 已提交
718 719 720 721

	/*	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.
722
	 *	This is designed to be as fast as possible
L
Linus Torvalds 已提交
723 724 725 726 727 728 729
	 *	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
730
	 * too slowly, when it should be increased quickly, decrease too quickly
L
Linus Torvalds 已提交
731 732 733 734
	 * 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)
	 */
735 736 737
	if (srtt != 0) {
		m -= (srtt >> 3);	/* m is now error in rtt est */
		srtt += m;		/* rtt = 7/8 rtt + 1/8 new */
L
Linus Torvalds 已提交
738 739
		if (m < 0) {
			m = -m;		/* m is now abs(error) */
740
			m -= (tp->mdev_us >> 2);   /* similar update on mdev */
L
Linus Torvalds 已提交
741 742 743 744 745 746 747 748 749 750 751
			/* 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 {
752
			m -= (tp->mdev_us >> 2);   /* similar update on mdev */
L
Linus Torvalds 已提交
753
		}
754 755 756 757 758
		tp->mdev_us += m;		/* mdev = 3/4 mdev + 1/4 new */
		if (tp->mdev_us > tp->mdev_max_us) {
			tp->mdev_max_us = tp->mdev_us;
			if (tp->mdev_max_us > tp->rttvar_us)
				tp->rttvar_us = tp->mdev_max_us;
L
Linus Torvalds 已提交
759 760
		}
		if (after(tp->snd_una, tp->rtt_seq)) {
761 762
			if (tp->mdev_max_us < tp->rttvar_us)
				tp->rttvar_us -= (tp->rttvar_us - tp->mdev_max_us) >> 2;
L
Linus Torvalds 已提交
763
			tp->rtt_seq = tp->snd_nxt;
764
			tp->mdev_max_us = tcp_rto_min_us(sk);
L
Linus Torvalds 已提交
765 766 767
		}
	} else {
		/* no previous measure. */
768
		srtt = m << 3;		/* take the measured time to be rtt */
769 770 771
		tp->mdev_us = m << 1;	/* make sure rto = 3*rtt */
		tp->rttvar_us = max(tp->mdev_us, tcp_rto_min_us(sk));
		tp->mdev_max_us = tp->rttvar_us;
L
Linus Torvalds 已提交
772 773
		tp->rtt_seq = tp->snd_nxt;
	}
774
	tp->srtt_us = max(1U, srtt);
L
Linus Torvalds 已提交
775 776
}

E
Eric Dumazet 已提交
777 778 779 780 781 782
/* Set the sk_pacing_rate to allow proper sizing of TSO packets.
 * Note: TCP stack does not yet implement pacing.
 * FQ packet scheduler can be used to implement cheap but effective
 * TCP pacing, to smooth the burst on large writes when packets
 * in flight is significantly lower than cwnd (or rwin)
 */
783 784 785
int sysctl_tcp_pacing_ss_ratio __read_mostly = 200;
int sysctl_tcp_pacing_ca_ratio __read_mostly = 120;

E
Eric Dumazet 已提交
786 787 788 789 790 791
static void tcp_update_pacing_rate(struct sock *sk)
{
	const struct tcp_sock *tp = tcp_sk(sk);
	u64 rate;

	/* set sk_pacing_rate to 200 % of current rate (mss * cwnd / srtt) */
792 793 794 795 796 797 798 799 800 801 802 803 804 805
	rate = (u64)tp->mss_cache * ((USEC_PER_SEC / 100) << 3);

	/* current rate is (cwnd * mss) / srtt
	 * In Slow Start [1], set sk_pacing_rate to 200 % the current rate.
	 * In Congestion Avoidance phase, set it to 120 % the current rate.
	 *
	 * [1] : Normal Slow Start condition is (tp->snd_cwnd < tp->snd_ssthresh)
	 *	 If snd_cwnd >= (tp->snd_ssthresh / 2), we are approaching
	 *	 end of slow start and should slow down.
	 */
	if (tp->snd_cwnd < tp->snd_ssthresh / 2)
		rate *= sysctl_tcp_pacing_ss_ratio;
	else
		rate *= sysctl_tcp_pacing_ca_ratio;
E
Eric Dumazet 已提交
806 807 808

	rate *= max(tp->snd_cwnd, tp->packets_out);

809 810
	if (likely(tp->srtt_us))
		do_div(rate, tp->srtt_us);
E
Eric Dumazet 已提交
811

812 813 814 815 816 817
	/* ACCESS_ONCE() is needed because sch_fq fetches sk_pacing_rate
	 * without any lock. We want to make sure compiler wont store
	 * intermediate values in this location.
	 */
	ACCESS_ONCE(sk->sk_pacing_rate) = min_t(u64, rate,
						sk->sk_max_pacing_rate);
E
Eric Dumazet 已提交
818 819
}

L
Linus Torvalds 已提交
820 821 822
/* Calculate rto without backoff.  This is the second half of Van Jacobson's
 * routine referred to above.
 */
823
static void tcp_set_rto(struct sock *sk)
L
Linus Torvalds 已提交
824
{
825
	const struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
826 827 828 829 830 831 832 833
	/* 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
834
	 *    ACKs in some circumstances.
L
Linus Torvalds 已提交
835
	 */
836
	inet_csk(sk)->icsk_rto = __tcp_set_rto(tp);
L
Linus Torvalds 已提交
837 838 839 840

	/* 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
841
	 *    with correct one. It is exactly, which we pretend to do.
L
Linus Torvalds 已提交
842 843
	 */

844 845 846
	/* NOTE: clamping at TCP_RTO_MIN is not required, current algo
	 * guarantees that rto is higher.
	 */
847
	tcp_bound_rto(sk);
L
Linus Torvalds 已提交
848 849
}

850
__u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst)
L
Linus Torvalds 已提交
851 852 853
{
	__u32 cwnd = (dst ? dst_metric(dst, RTAX_INITCWND) : 0);

854
	if (!cwnd)
855
		cwnd = TCP_INIT_CWND;
L
Linus Torvalds 已提交
856 857 858
	return min_t(__u32, cwnd, tp->snd_cwnd_clamp);
}

859 860 861 862
/*
 * Packet counting of FACK is based on in-order assumptions, therefore TCP
 * disables it when reordering is detected
 */
863
void tcp_disable_fack(struct tcp_sock *tp)
864
{
865 866 867
	/* RFC3517 uses different metric in lost marker => reset on change */
	if (tcp_is_fack(tp))
		tp->lost_skb_hint = NULL;
868
	tp->rx_opt.sack_ok &= ~TCP_FACK_ENABLED;
869 870
}

871
/* Take a notice that peer is sending D-SACKs */
872 873
static void tcp_dsack_seen(struct tcp_sock *tp)
{
874
	tp->rx_opt.sack_ok |= TCP_DSACK_SEEN;
875 876
}

877 878
static void tcp_update_reordering(struct sock *sk, const int metric,
				  const int ts)
L
Linus Torvalds 已提交
879
{
880
	struct tcp_sock *tp = tcp_sk(sk);
881
	int mib_idx;
882

883
	if (metric > tp->reordering) {
884
		tp->reordering = min(sysctl_tcp_max_reordering, metric);
L
Linus Torvalds 已提交
885 886

#if FASTRETRANS_DEBUG > 1
887 888 889 890 891 892
		pr_debug("Disorder%d %d %u f%u s%u rr%d\n",
			 tp->rx_opt.sack_ok, inet_csk(sk)->icsk_ca_state,
			 tp->reordering,
			 tp->fackets_out,
			 tp->sacked_out,
			 tp->undo_marker ? tp->undo_retrans : 0);
L
Linus Torvalds 已提交
893
#endif
894
		tcp_disable_fack(tp);
L
Linus Torvalds 已提交
895
	}
Y
Yuchung Cheng 已提交
896

Y
Yuchung Cheng 已提交
897
	tp->rack.reord = 1;
898 899 900 901 902 903 904 905 906 907 908 909

	/* This exciting event is worth to be remembered. 8) */
	if (ts)
		mib_idx = LINUX_MIB_TCPTSREORDER;
	else if (tcp_is_reno(tp))
		mib_idx = LINUX_MIB_TCPRENOREORDER;
	else if (tcp_is_fack(tp))
		mib_idx = LINUX_MIB_TCPFACKREORDER;
	else
		mib_idx = LINUX_MIB_TCPSACKREORDER;

	NET_INC_STATS(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
910 911
}

912
/* This must be called before lost_out is incremented */
913 914
static void tcp_verify_retransmit_hint(struct tcp_sock *tp, struct sk_buff *skb)
{
915
	if (!tp->retransmit_skb_hint ||
916 917
	    before(TCP_SKB_CB(skb)->seq,
		   TCP_SKB_CB(tp->retransmit_skb_hint)->seq))
918
		tp->retransmit_skb_hint = skb;
919 920
}

921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936
/* Sum the number of packets on the wire we have marked as lost.
 * There are two cases we care about here:
 * a) Packet hasn't been marked lost (nor retransmitted),
 *    and this is the first loss.
 * b) Packet has been marked both lost and retransmitted,
 *    and this means we think it was lost again.
 */
static void tcp_sum_lost(struct tcp_sock *tp, struct sk_buff *skb)
{
	__u8 sacked = TCP_SKB_CB(skb)->sacked;

	if (!(sacked & TCPCB_LOST) ||
	    ((sacked & TCPCB_LOST) && (sacked & TCPCB_SACKED_RETRANS)))
		tp->lost += tcp_skb_pcount(skb);
}

937 938 939 940 941 942
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);
943
		tcp_sum_lost(tp, skb);
944 945 946 947
		TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
	}
}

Y
Yuchung Cheng 已提交
948
void tcp_skb_mark_lost_uncond_verify(struct tcp_sock *tp, struct sk_buff *skb)
949 950 951
{
	tcp_verify_retransmit_hint(tp, skb);

952
	tcp_sum_lost(tp, skb);
953 954 955 956 957 958
	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
959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979
/* 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())
980
 * 3. Loss detection event of two flavors:
L
Linus Torvalds 已提交
981 982
 *	A. Scoreboard estimator decided the packet is lost.
 *	   A'. Reno "three dupacks" marks head of queue lost.
983 984
 *	   A''. Its FACK modification, head until snd.fack is lost.
 *	B. SACK arrives sacking SND.NXT at the moment, when the
L
Linus Torvalds 已提交
985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003
 *	   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
1004 1005 1006 1007 1008 1009 1010
 *
 * 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
1011 1012 1013 1014 1015 1016 1017 1018 1019
 * 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
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
 *
 * 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
1042
 * again, D-SACK block must not to go across snd_una (for the same reason as
I
Ilpo Järvinen 已提交
1043 1044 1045 1046 1047 1048 1049 1050 1051
 * 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
1052
 */
E
Eric Dumazet 已提交
1053 1054
static bool tcp_is_sackblock_valid(struct tcp_sock *tp, bool is_dsack,
				   u32 start_seq, u32 end_seq)
I
Ilpo Järvinen 已提交
1055 1056 1057
{
	/* Too far in future, or reversed (interpretation is ambiguous) */
	if (after(end_seq, tp->snd_nxt) || !before(start_seq, end_seq))
E
Eric Dumazet 已提交
1058
		return false;
I
Ilpo Järvinen 已提交
1059 1060 1061

	/* Nasty start_seq wrap-around check (see comments above) */
	if (!before(start_seq, tp->snd_nxt))
E
Eric Dumazet 已提交
1062
		return false;
I
Ilpo Järvinen 已提交
1063

1064
	/* In outstanding window? ...This is valid exit for D-SACKs too.
I
Ilpo Järvinen 已提交
1065 1066 1067
	 * start_seq == snd_una is non-sensical (see comments above)
	 */
	if (after(start_seq, tp->snd_una))
E
Eric Dumazet 已提交
1068
		return true;
I
Ilpo Järvinen 已提交
1069 1070

	if (!is_dsack || !tp->undo_marker)
E
Eric Dumazet 已提交
1071
		return false;
I
Ilpo Järvinen 已提交
1072 1073

	/* ...Then it's D-SACK, and must reside below snd_una completely */
Z
Zheng Yan 已提交
1074
	if (after(end_seq, tp->snd_una))
E
Eric Dumazet 已提交
1075
		return false;
I
Ilpo Järvinen 已提交
1076 1077

	if (!before(start_seq, tp->undo_marker))
E
Eric Dumazet 已提交
1078
		return true;
I
Ilpo Järvinen 已提交
1079 1080 1081

	/* Too old */
	if (!after(end_seq, tp->undo_marker))
E
Eric Dumazet 已提交
1082
		return false;
I
Ilpo Järvinen 已提交
1083 1084 1085 1086 1087 1088 1089

	/* 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);
}

E
Eric Dumazet 已提交
1090 1091 1092
static bool tcp_check_dsack(struct sock *sk, const struct sk_buff *ack_skb,
			    struct tcp_sack_block_wire *sp, int num_sacks,
			    u32 prior_snd_una)
1093
{
1094
	struct tcp_sock *tp = tcp_sk(sk);
1095 1096
	u32 start_seq_0 = get_unaligned_be32(&sp[0].start_seq);
	u32 end_seq_0 = get_unaligned_be32(&sp[0].end_seq);
E
Eric Dumazet 已提交
1097
	bool dup_sack = false;
1098 1099

	if (before(start_seq_0, TCP_SKB_CB(ack_skb)->ack_seq)) {
E
Eric Dumazet 已提交
1100
		dup_sack = true;
1101
		tcp_dsack_seen(tp);
1102
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPDSACKRECV);
1103
	} else if (num_sacks > 1) {
1104 1105
		u32 end_seq_1 = get_unaligned_be32(&sp[1].end_seq);
		u32 start_seq_1 = get_unaligned_be32(&sp[1].start_seq);
1106 1107 1108

		if (!after(end_seq_0, end_seq_1) &&
		    !before(start_seq_0, start_seq_1)) {
E
Eric Dumazet 已提交
1109
			dup_sack = true;
1110
			tcp_dsack_seen(tp);
1111
			NET_INC_STATS(sock_net(sk),
1112
					LINUX_MIB_TCPDSACKOFORECV);
1113 1114 1115 1116
		}
	}

	/* D-SACK for already forgotten data... Do dumb counting. */
Y
Yuchung Cheng 已提交
1117
	if (dup_sack && tp->undo_marker && tp->undo_retrans > 0 &&
1118 1119 1120 1121 1122 1123 1124
	    !after(end_seq_0, prior_snd_una) &&
	    after(end_seq_0, tp->undo_marker))
		tp->undo_retrans--;

	return dup_sack;
}

1125
struct tcp_sacktag_state {
1126 1127
	int	reord;
	int	fack_count;
1128 1129 1130 1131 1132 1133
	/* Timestamps for earliest and latest never-retransmitted segment
	 * that was SACKed. RTO needs the earliest RTT to stay conservative,
	 * but congestion control should still get an accurate delay signal.
	 */
	struct skb_mstamp first_sackt;
	struct skb_mstamp last_sackt;
1134
	struct rate_sample *rate;
1135
	int	flag;
1136 1137
};

1138 1139 1140 1141 1142
/* 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).
1143 1144
 *
 * FIXME: this could be merged to shift decision code
1145
 */
1146
static int tcp_match_skb_to_sack(struct sock *sk, struct sk_buff *skb,
E
Eric Dumazet 已提交
1147
				  u32 start_seq, u32 end_seq)
1148
{
E
Eric Dumazet 已提交
1149 1150
	int err;
	bool in_sack;
1151
	unsigned int pkt_len;
1152
	unsigned int mss;
1153 1154 1155 1156 1157 1158

	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)) {
1159
		mss = tcp_skb_mss(skb);
1160 1161
		in_sack = !after(start_seq, TCP_SKB_CB(skb)->seq);

1162
		if (!in_sack) {
1163
			pkt_len = start_seq - TCP_SKB_CB(skb)->seq;
1164 1165 1166
			if (pkt_len < mss)
				pkt_len = mss;
		} else {
1167
			pkt_len = end_seq - TCP_SKB_CB(skb)->seq;
1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178
			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;
1179
				if (new_len >= skb->len)
1180 1181 1182 1183
					return 0;
			}
			pkt_len = new_len;
		}
1184
		err = tcp_fragment(sk, skb, pkt_len, mss, GFP_ATOMIC);
1185 1186 1187 1188 1189 1190 1191
		if (err < 0)
			return err;
	}

	return in_sack;
}

1192 1193 1194 1195
/* Mark the given newly-SACKed range as such, adjusting counters and hints. */
static u8 tcp_sacktag_one(struct sock *sk,
			  struct tcp_sacktag_state *state, u8 sacked,
			  u32 start_seq, u32 end_seq,
1196 1197
			  int dup_sack, int pcount,
			  const struct skb_mstamp *xmit_time)
1198
{
1199
	struct tcp_sock *tp = tcp_sk(sk);
1200
	int fack_count = state->fack_count;
1201 1202 1203

	/* Account D-SACK for retransmitted packet. */
	if (dup_sack && (sacked & TCPCB_RETRANS)) {
Y
Yuchung Cheng 已提交
1204
		if (tp->undo_marker && tp->undo_retrans > 0 &&
1205
		    after(end_seq, tp->undo_marker))
1206
			tp->undo_retrans--;
1207
		if (sacked & TCPCB_SACKED_ACKED)
1208
			state->reord = min(fack_count, state->reord);
1209 1210 1211
	}

	/* Nothing to do; acked frame is about to be dropped (was ACKed). */
1212
	if (!after(end_seq, tp->snd_una))
1213
		return sacked;
1214 1215

	if (!(sacked & TCPCB_SACKED_ACKED)) {
1216
		tcp_rack_advance(tp, sacked, end_seq, xmit_time);
1217

1218 1219 1220 1221 1222 1223
		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) {
1224
				sacked &= ~(TCPCB_LOST|TCPCB_SACKED_RETRANS);
1225 1226
				tp->lost_out -= pcount;
				tp->retrans_out -= pcount;
1227 1228 1229 1230 1231 1232
			}
		} else {
			if (!(sacked & TCPCB_RETRANS)) {
				/* New sack for not retransmitted frame,
				 * which was in hole. It is reordering.
				 */
1233
				if (before(start_seq,
1234
					   tcp_highest_sack_seq(tp)))
1235 1236
					state->reord = min(fack_count,
							   state->reord);
Y
Yuchung Cheng 已提交
1237 1238
				if (!after(end_seq, tp->high_seq))
					state->flag |= FLAG_ORIG_SACK_ACKED;
1239 1240 1241
				if (state->first_sackt.v64 == 0)
					state->first_sackt = *xmit_time;
				state->last_sackt = *xmit_time;
1242 1243 1244
			}

			if (sacked & TCPCB_LOST) {
1245
				sacked &= ~TCPCB_LOST;
1246
				tp->lost_out -= pcount;
1247 1248 1249
			}
		}

1250 1251
		sacked |= TCPCB_SACKED_ACKED;
		state->flag |= FLAG_DATA_SACKED;
1252
		tp->sacked_out += pcount;
Y
Yuchung Cheng 已提交
1253
		tp->delivered += pcount;  /* Out-of-order packets delivered */
1254

1255
		fack_count += pcount;
1256 1257

		/* Lost marker hint past SACKed? Tweak RFC3517 cnt */
1258
		if (!tcp_is_fack(tp) && tp->lost_skb_hint &&
1259
		    before(start_seq, TCP_SKB_CB(tp->lost_skb_hint)->seq))
1260
			tp->lost_cnt_hint += pcount;
1261 1262 1263 1264 1265 1266 1267 1268 1269

		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.
	 */
1270 1271
	if (dup_sack && (sacked & TCPCB_SACKED_RETRANS)) {
		sacked &= ~TCPCB_SACKED_RETRANS;
1272
		tp->retrans_out -= pcount;
1273 1274
	}

1275
	return sacked;
1276 1277
}

1278 1279 1280
/* Shift newly-SACKed bytes from this skb to the immediately previous
 * already-SACKed sk_buff. Mark the newly-SACKed bytes as such.
 */
E
Eric Dumazet 已提交
1281 1282 1283 1284
static bool tcp_shifted_skb(struct sock *sk, struct sk_buff *skb,
			    struct tcp_sacktag_state *state,
			    unsigned int pcount, int shifted, int mss,
			    bool dup_sack)
1285 1286
{
	struct tcp_sock *tp = tcp_sk(sk);
1287
	struct sk_buff *prev = tcp_write_queue_prev(sk, skb);
1288 1289
	u32 start_seq = TCP_SKB_CB(skb)->seq;	/* start of newly-SACKed */
	u32 end_seq = start_seq + shifted;	/* end of newly-SACKed */
1290 1291 1292

	BUG_ON(!pcount);

1293 1294 1295 1296 1297 1298 1299
	/* Adjust counters and hints for the newly sacked sequence
	 * range but discard the return value since prev is already
	 * marked. We must tag the range first because the seq
	 * advancement below implicitly advances
	 * tcp_highest_sack_seq() when skb is highest_sack.
	 */
	tcp_sacktag_one(sk, state, TCP_SKB_CB(skb)->sacked,
Y
Yuchung Cheng 已提交
1300
			start_seq, end_seq, dup_sack, pcount,
1301
			&skb->skb_mstamp);
1302
	tcp_rate_skb_delivered(sk, skb, state->rate);
1303 1304

	if (skb == tp->lost_skb_hint)
1305 1306
		tp->lost_cnt_hint += pcount;

1307 1308 1309
	TCP_SKB_CB(prev)->end_seq += shifted;
	TCP_SKB_CB(skb)->seq += shifted;

1310 1311 1312
	tcp_skb_pcount_add(prev, pcount);
	BUG_ON(tcp_skb_pcount(skb) < pcount);
	tcp_skb_pcount_add(skb, -pcount);
1313 1314 1315 1316 1317 1318

	/* 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.
	 */
1319 1320
	if (!TCP_SKB_CB(prev)->tcp_gso_size)
		TCP_SKB_CB(prev)->tcp_gso_size = mss;
1321 1322

	/* CHECKME: To clear or not to clear? Mimics normal skb currently */
1323
	if (tcp_skb_pcount(skb) <= 1)
1324
		TCP_SKB_CB(skb)->tcp_gso_size = 0;
1325 1326 1327 1328 1329 1330

	/* 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));
1331
		NET_INC_STATS(sock_net(sk), LINUX_MIB_SACKSHIFTED);
E
Eric Dumazet 已提交
1332
		return false;
1333 1334 1335 1336
	}

	/* Whole SKB was eaten :-) */

1337 1338 1339 1340 1341 1342 1343
	if (skb == tp->retransmit_skb_hint)
		tp->retransmit_skb_hint = prev;
	if (skb == tp->lost_skb_hint) {
		tp->lost_skb_hint = prev;
		tp->lost_cnt_hint -= tcp_skb_pcount(prev);
	}

1344
	TCP_SKB_CB(prev)->tcp_flags |= TCP_SKB_CB(skb)->tcp_flags;
1345
	TCP_SKB_CB(prev)->eor = TCP_SKB_CB(skb)->eor;
1346 1347 1348
	if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)
		TCP_SKB_CB(prev)->end_seq++;

1349 1350 1351
	if (skb == tcp_highest_sack(sk))
		tcp_advance_highest_sack(sk, skb);

1352
	tcp_skb_collapse_tstamp(prev, skb);
1353 1354 1355
	if (unlikely(TCP_SKB_CB(prev)->tx.delivered_mstamp.v64))
		TCP_SKB_CB(prev)->tx.delivered_mstamp.v64 = 0;

1356 1357 1358
	tcp_unlink_write_queue(skb, sk);
	sk_wmem_free_skb(sk, skb);

1359
	NET_INC_STATS(sock_net(sk), LINUX_MIB_SACKMERGED);
1360

E
Eric Dumazet 已提交
1361
	return true;
1362 1363 1364 1365 1366
}

/* I wish gso_size would have a bit more sane initialization than
 * something-or-zero which complicates things
 */
1367
static int tcp_skb_seglen(const struct sk_buff *skb)
1368
{
1369
	return tcp_skb_pcount(skb) == 1 ? skb->len : tcp_skb_mss(skb);
1370 1371 1372
}

/* Shifting pages past head area doesn't work */
1373
static int skb_can_shift(const struct sk_buff *skb)
1374 1375 1376 1377 1378 1379 1380 1381
{
	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,
1382
					  struct tcp_sacktag_state *state,
1383
					  u32 start_seq, u32 end_seq,
E
Eric Dumazet 已提交
1384
					  bool dup_sack)
1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
{
	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 &&
1398
	    (TCP_SKB_CB(skb)->sacked & (TCPCB_LOST|TCPCB_SACKED_RETRANS)) == TCPCB_SACKED_RETRANS)
1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413
		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;

1414 1415 1416
	if (!tcp_skb_can_collapse_to(prev))
		goto fallback;

1417 1418 1419 1420 1421 1422
	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);
1423
		mss = tcp_skb_seglen(skb);
1424 1425 1426 1427

		/* TODO: Fix DSACKs to not fragment already SACKed and we can
		 * drop this restriction as unnecessary
		 */
1428
		if (mss != tcp_skb_seglen(prev))
1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468
			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
		 */
1469
		if (mss != tcp_skb_seglen(prev))
1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481
			goto fallback;

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

1482 1483 1484 1485
	/* tcp_sacktag_one() won't SACK-tag ranges below snd_una */
	if (!after(TCP_SKB_CB(skb)->seq + len, tp->snd_una))
		goto fallback;

1486 1487
	if (!skb_shift(prev, skb, len))
		goto fallback;
I
Ilpo Järvinen 已提交
1488
	if (!tcp_shifted_skb(sk, skb, state, pcount, len, mss, dup_sack))
1489 1490 1491 1492 1493 1494 1495 1496 1497
		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);

1498 1499 1500
	if (!skb_can_shift(skb) ||
	    (skb == tcp_send_head(sk)) ||
	    ((TCP_SKB_CB(skb)->sacked & TCPCB_TAGBITS) != TCPCB_SACKED_ACKED) ||
1501
	    (mss != tcp_skb_seglen(skb)))
1502 1503 1504 1505 1506
		goto out;

	len = skb->len;
	if (skb_shift(prev, skb, len)) {
		pcount += tcp_skb_pcount(skb);
I
Ilpo Järvinen 已提交
1507
		tcp_shifted_skb(sk, skb, state, tcp_skb_pcount(skb), len, mss, 0);
1508 1509 1510
	}

out:
1511
	state->fack_count += pcount;
1512 1513 1514 1515 1516 1517
	return prev;

noop:
	return skb;

fallback:
1518
	NET_INC_STATS(sock_net(sk), LINUX_MIB_SACKSHIFTFALLBACK);
1519 1520 1521
	return NULL;
}

1522 1523
static struct sk_buff *tcp_sacktag_walk(struct sk_buff *skb, struct sock *sk,
					struct tcp_sack_block *next_dup,
1524
					struct tcp_sacktag_state *state,
1525
					u32 start_seq, u32 end_seq,
E
Eric Dumazet 已提交
1526
					bool dup_sack_in)
1527
{
1528 1529 1530
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *tmp;

1531 1532
	tcp_for_write_queue_from(skb, sk) {
		int in_sack = 0;
E
Eric Dumazet 已提交
1533
		bool dup_sack = dup_sack_in;
1534 1535 1536 1537 1538 1539 1540 1541

		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;

1542
		if (next_dup  &&
1543 1544 1545 1546 1547
		    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)
E
Eric Dumazet 已提交
1548
				dup_sack = true;
1549 1550
		}

1551 1552 1553 1554 1555
		/* 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) {
1556 1557
			tmp = tcp_shift_skb_data(sk, skb, state,
						 start_seq, end_seq, dup_sack);
1558
			if (tmp) {
1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571
				if (tmp != skb) {
					skb = tmp;
					continue;
				}

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

1572 1573 1574
		if (unlikely(in_sack < 0))
			break;

1575
		if (in_sack) {
1576 1577 1578 1579 1580 1581 1582
			TCP_SKB_CB(skb)->sacked =
				tcp_sacktag_one(sk,
						state,
						TCP_SKB_CB(skb)->sacked,
						TCP_SKB_CB(skb)->seq,
						TCP_SKB_CB(skb)->end_seq,
						dup_sack,
Y
Yuchung Cheng 已提交
1583
						tcp_skb_pcount(skb),
1584
						&skb->skb_mstamp);
1585
			tcp_rate_skb_delivered(sk, skb, state->rate);
1586

1587 1588 1589 1590 1591
			if (!before(TCP_SKB_CB(skb)->seq,
				    tcp_highest_sack_seq(tp)))
				tcp_advance_highest_sack(sk, skb);
		}

1592
		state->fack_count += tcp_skb_pcount(skb);
1593 1594 1595 1596 1597 1598 1599 1600
	}
	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,
1601 1602
					struct tcp_sacktag_state *state,
					u32 skip_to_seq)
1603 1604 1605 1606 1607
{
	tcp_for_write_queue_from(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;

I
Ilpo Järvinen 已提交
1608
		if (after(TCP_SKB_CB(skb)->end_seq, skip_to_seq))
1609
			break;
1610

1611
		state->fack_count += tcp_skb_pcount(skb);
1612 1613 1614 1615 1616 1617 1618
	}
	return skb;
}

static struct sk_buff *tcp_maybe_skipping_dsack(struct sk_buff *skb,
						struct sock *sk,
						struct tcp_sack_block *next_dup,
1619 1620
						struct tcp_sacktag_state *state,
						u32 skip_to_seq)
1621
{
1622
	if (!next_dup)
1623 1624 1625
		return skb;

	if (before(next_dup->start_seq, skip_to_seq)) {
1626 1627 1628 1629
		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);
1630 1631 1632 1633 1634
	}

	return skb;
}

1635
static int tcp_sack_cache_ok(const struct tcp_sock *tp, const struct tcp_sack_block *cache)
1636 1637 1638 1639
{
	return cache < tp->recv_sack_cache + ARRAY_SIZE(tp->recv_sack_cache);
}

L
Linus Torvalds 已提交
1640
static int
1641
tcp_sacktag_write_queue(struct sock *sk, const struct sk_buff *ack_skb,
1642
			u32 prior_snd_una, struct tcp_sacktag_state *state)
L
Linus Torvalds 已提交
1643 1644
{
	struct tcp_sock *tp = tcp_sk(sk);
1645 1646
	const unsigned char *ptr = (skb_transport_header(ack_skb) +
				    TCP_SKB_CB(ack_skb)->sacked);
1647
	struct tcp_sack_block_wire *sp_wire = (struct tcp_sack_block_wire *)(ptr+2);
1648
	struct tcp_sack_block sp[TCP_NUM_SACKS];
1649 1650
	struct tcp_sack_block *cache;
	struct sk_buff *skb;
1651
	int num_sacks = min(TCP_NUM_SACKS, (ptr[1] - TCPOLEN_SACK_BASE) >> 3);
1652
	int used_sacks;
E
Eric Dumazet 已提交
1653
	bool found_dup_sack = false;
1654
	int i, j;
1655
	int first_sack_index;
L
Linus Torvalds 已提交
1656

1657 1658
	state->flag = 0;
	state->reord = tp->packets_out;
1659

1660
	if (!tp->sacked_out) {
1661 1662
		if (WARN_ON(tp->fackets_out))
			tp->fackets_out = 0;
1663
		tcp_highest_sack_reset(sk);
1664
	}
L
Linus Torvalds 已提交
1665

1666
	found_dup_sack = tcp_check_dsack(sk, ack_skb, sp_wire,
1667
					 num_sacks, prior_snd_una);
1668
	if (found_dup_sack) {
1669
		state->flag |= FLAG_DSACKING_ACK;
1670 1671
		tp->delivered++; /* A spurious retransmission is delivered */
	}
1672 1673 1674 1675 1676 1677 1678 1679

	/* 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;

1680 1681 1682
	if (!tp->packets_out)
		goto out;

1683 1684 1685
	used_sacks = 0;
	first_sack_index = 0;
	for (i = 0; i < num_sacks; i++) {
E
Eric Dumazet 已提交
1686
		bool dup_sack = !i && found_dup_sack;
1687

1688 1689
		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);
1690 1691 1692 1693

		if (!tcp_is_sackblock_valid(tp, dup_sack,
					    sp[used_sacks].start_seq,
					    sp[used_sacks].end_seq)) {
1694 1695
			int mib_idx;

1696 1697
			if (dup_sack) {
				if (!tp->undo_marker)
1698
					mib_idx = LINUX_MIB_TCPDSACKIGNOREDNOUNDO;
1699
				else
1700
					mib_idx = LINUX_MIB_TCPDSACKIGNOREDOLD;
1701 1702 1703 1704 1705
			} 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;
1706
				mib_idx = LINUX_MIB_TCPSACKDISCARD;
1707
			}
1708

1709
			NET_INC_STATS(sock_net(sk), mib_idx);
1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721
			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++;
	}

1722 1723
	/* order SACK blocks to allow in order walk of the retrans queue */
	for (i = used_sacks - 1; i > 0; i--) {
1724 1725
		for (j = 0; j < i; j++) {
			if (after(sp[j].start_seq, sp[j + 1].start_seq)) {
1726
				swap(sp[j], sp[j + 1]);
1727

1728 1729
				/* Track where the first SACK block goes to */
				if (j == first_sack_index)
1730
					first_sack_index = j + 1;
1731 1732 1733 1734
			}
		}
	}

1735
	skb = tcp_write_queue_head(sk);
1736
	state->fack_count = 0;
1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747
	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++;
1748 1749
	}

1750
	while (i < used_sacks) {
1751 1752
		u32 start_seq = sp[i].start_seq;
		u32 end_seq = sp[i].end_seq;
E
Eric Dumazet 已提交
1753
		bool dup_sack = (found_dup_sack && (i == first_sack_index));
1754
		struct tcp_sack_block *next_dup = NULL;
1755

1756 1757
		if (found_dup_sack && ((i + 1) == first_sack_index))
			next_dup = &sp[i + 1];
L
Linus Torvalds 已提交
1758

1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769
		/* 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)) {
1770
				skb = tcp_sacktag_skip(skb, sk, state,
1771
						       start_seq);
1772
				skb = tcp_sacktag_walk(skb, sk, next_dup,
1773
						       state,
1774 1775
						       start_seq,
						       cache->start_seq,
1776
						       dup_sack);
1777
			}
1778

1779
			/* Rest of the block already fully processed? */
1780
			if (!after(end_seq, cache->end_seq))
1781
				goto advance_sp;
1782

1783
			skb = tcp_maybe_skipping_dsack(skb, sk, next_dup,
1784
						       state,
1785
						       cache->end_seq);
1786

1787
			/* ...tail remains todo... */
1788
			if (tcp_highest_sack_seq(tp) == cache->end_seq) {
1789
				/* ...but better entrypoint exists! */
1790
				skb = tcp_highest_sack(sk);
1791
				if (!skb)
1792
					break;
1793
				state->fack_count = tp->fackets_out;
1794 1795
				cache++;
				goto walk;
1796 1797
			}

1798
			skb = tcp_sacktag_skip(skb, sk, state, cache->end_seq);
1799 1800 1801 1802
			/* Check overlap against next cached too (past this one already) */
			cache++;
			continue;
		}
L
Linus Torvalds 已提交
1803

1804 1805
		if (!before(start_seq, tcp_highest_sack_seq(tp))) {
			skb = tcp_highest_sack(sk);
1806
			if (!skb)
1807
				break;
1808
			state->fack_count = tp->fackets_out;
L
Linus Torvalds 已提交
1809
		}
1810
		skb = tcp_sacktag_skip(skb, sk, state, start_seq);
1811 1812

walk:
1813
		skb = tcp_sacktag_walk(skb, sk, next_dup, state,
1814
				       start_seq, end_seq, dup_sack);
1815

1816 1817
advance_sp:
		i++;
L
Linus Torvalds 已提交
1818 1819
	}

1820 1821 1822 1823 1824 1825 1826 1827
	/* 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];

1828
	if ((state->reord < tp->fackets_out) &&
Y
Yuchung Cheng 已提交
1829
	    ((inet_csk(sk)->icsk_ca_state != TCP_CA_Loss) || tp->undo_marker))
1830
		tcp_update_reordering(sk, tp->fackets_out - state->reord, 0);
L
Linus Torvalds 已提交
1831

1832
	tcp_verify_left_out(tp);
1833 1834
out:

L
Linus Torvalds 已提交
1835
#if FASTRETRANS_DEBUG > 0
1836 1837 1838 1839
	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
1840
#endif
1841
	return state->flag;
L
Linus Torvalds 已提交
1842 1843
}

1844
/* Limits sacked_out so that sum with lost_out isn't ever larger than
E
Eric Dumazet 已提交
1845
 * packets_out. Returns false if sacked_out adjustement wasn't necessary.
1846
 */
E
Eric Dumazet 已提交
1847
static bool tcp_limit_reno_sacked(struct tcp_sock *tp)
1848 1849 1850 1851 1852 1853 1854 1855
{
	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;
E
Eric Dumazet 已提交
1856
		return true;
1857
	}
E
Eric Dumazet 已提交
1858
	return false;
1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869
}

/* 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);
1870 1871 1872 1873 1874 1875 1876
}

/* 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);
Y
Yuchung Cheng 已提交
1877 1878
	u32 prior_sacked = tp->sacked_out;

1879 1880
	tp->sacked_out++;
	tcp_check_reno_reordering(sk, 0);
Y
Yuchung Cheng 已提交
1881 1882
	if (tp->sacked_out > prior_sacked)
		tp->delivered++; /* Some out-of-order packet is delivered */
1883
	tcp_verify_left_out(tp);
1884 1885 1886 1887 1888 1889 1890 1891 1892 1893
}

/* 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. */
Y
Yuchung Cheng 已提交
1894
		tp->delivered += max_t(int, acked - tp->sacked_out, 1);
1895
		if (acked - 1 >= tp->sacked_out)
1896 1897
			tp->sacked_out = 0;
		else
1898
			tp->sacked_out -= acked - 1;
1899 1900
	}
	tcp_check_reno_reordering(sk, acked);
1901
	tcp_verify_left_out(tp);
1902 1903 1904 1905 1906 1907 1908
}

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

Y
Yuchung Cheng 已提交
1909
void tcp_clear_retrans(struct tcp_sock *tp)
L
Linus Torvalds 已提交
1910 1911 1912 1913
{
	tp->retrans_out = 0;
	tp->lost_out = 0;
	tp->undo_marker = 0;
Y
Yuchung Cheng 已提交
1914
	tp->undo_retrans = -1;
Y
Yuchung Cheng 已提交
1915 1916
	tp->fackets_out = 0;
	tp->sacked_out = 0;
L
Linus Torvalds 已提交
1917 1918
}

Y
Yuchung Cheng 已提交
1919
static inline void tcp_init_undo(struct tcp_sock *tp)
1920
{
Y
Yuchung Cheng 已提交
1921 1922 1923
	tp->undo_marker = tp->snd_una;
	/* Retransmission still in flight may cause DSACKs later. */
	tp->undo_retrans = tp->retrans_out ? : -1;
1924 1925
}

1926
/* Enter Loss state. If we detect SACK reneging, forget all SACK information
L
Linus Torvalds 已提交
1927 1928 1929
 * and reset tags completely, otherwise preserve SACKs. If receiver
 * dropped its ofo queue, we will know this due to reneging detection.
 */
1930
void tcp_enter_loss(struct sock *sk)
L
Linus Torvalds 已提交
1931
{
1932
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1933
	struct tcp_sock *tp = tcp_sk(sk);
1934
	struct net *net = sock_net(sk);
L
Linus Torvalds 已提交
1935
	struct sk_buff *skb;
1936
	bool new_recovery = icsk->icsk_ca_state < TCP_CA_Recovery;
1937
	bool is_reneg;			/* is receiver reneging on SACKs? */
1938
	bool mark_lost;
L
Linus Torvalds 已提交
1939 1940

	/* Reduce ssthresh if it has not yet been made inside this window. */
Y
Yuchung Cheng 已提交
1941 1942
	if (icsk->icsk_ca_state <= TCP_CA_Disorder ||
	    !after(tp->high_seq, tp->snd_una) ||
1943 1944 1945 1946
	    (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);
Y
Yuchung Cheng 已提交
1947
		tcp_init_undo(tp);
L
Linus Torvalds 已提交
1948 1949 1950 1951 1952
	}
	tp->snd_cwnd	   = 1;
	tp->snd_cwnd_cnt   = 0;
	tp->snd_cwnd_stamp = tcp_time_stamp;

Y
Yuchung Cheng 已提交
1953 1954
	tp->retrans_out = 0;
	tp->lost_out = 0;
1955 1956 1957

	if (tcp_is_reno(tp))
		tcp_reset_reno_sack(tp);
L
Linus Torvalds 已提交
1958

1959 1960 1961
	skb = tcp_write_queue_head(sk);
	is_reneg = skb && (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED);
	if (is_reneg) {
1962
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPSACKRENEGING);
1963 1964
		tp->sacked_out = 0;
		tp->fackets_out = 0;
1965
	}
1966
	tcp_clear_all_retrans_hints(tp);
L
Linus Torvalds 已提交
1967

1968 1969 1970
	tcp_for_write_queue(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
1971

1972 1973 1974 1975
		mark_lost = (!(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED) ||
			     is_reneg);
		if (mark_lost)
			tcp_sum_lost(tp, skb);
L
Linus Torvalds 已提交
1976
		TCP_SKB_CB(skb)->sacked &= (~TCPCB_TAGBITS)|TCPCB_SACKED_ACKED;
1977
		if (mark_lost) {
L
Linus Torvalds 已提交
1978 1979 1980 1981 1982
			TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_ACKED;
			TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
			tp->lost_out += tcp_skb_pcount(skb);
		}
	}
1983
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
1984

1985 1986 1987 1988
	/* Timeout in disordered state after receiving substantial DUPACKs
	 * suggests that the degree of reordering is over-estimated.
	 */
	if (icsk->icsk_ca_state <= TCP_CA_Disorder &&
1989
	    tp->sacked_out >= net->ipv4.sysctl_tcp_reordering)
1990
		tp->reordering = min_t(unsigned int, tp->reordering,
1991
				       net->ipv4.sysctl_tcp_reordering);
1992
	tcp_set_ca_state(sk, TCP_CA_Loss);
L
Linus Torvalds 已提交
1993
	tp->high_seq = tp->snd_nxt;
1994
	tcp_ecn_queue_cwr(tp);
Y
Yuchung Cheng 已提交
1995

1996 1997 1998 1999 2000 2001 2002 2003
	/* F-RTO RFC5682 sec 3.1 step 1: retransmit SND.UNA if no previous
	 * loss recovery is underway except recurring timeout(s) on
	 * the same SND.UNA (sec 3.2). Disable F-RTO on path MTU probing
	 *
	 * In theory F-RTO can be used repeatedly during loss recovery.
	 * In practice this interacts badly with broken middle-boxes that
	 * falsely raise the receive window, which results in repeated
	 * timeouts and stop-and-go behavior.
Y
Yuchung Cheng 已提交
2004
	 */
2005 2006 2007
	tp->frto = sysctl_tcp_frto &&
		   (new_recovery || icsk->icsk_retransmits) &&
		   !inet_csk(sk)->icsk_mtup.probe_size;
L
Linus Torvalds 已提交
2008 2009
}

2010 2011 2012 2013
/* 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).
 *
2014 2015 2016 2017 2018
 * To avoid big spurious retransmission bursts due to transient SACK
 * scoreboard oddities that look like reneging, we give the receiver a
 * little time (max(RTT/2, 10ms)) to send us some more ACKs that will
 * restore sanity to the SACK scoreboard. If the apparent reneging
 * persists until this RTO then we'll clear the SACK scoreboard.
2019
 */
E
Eric Dumazet 已提交
2020
static bool tcp_check_sack_reneging(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2021
{
2022
	if (flag & FLAG_SACK_RENEGING) {
2023 2024 2025
		struct tcp_sock *tp = tcp_sk(sk);
		unsigned long delay = max(usecs_to_jiffies(tp->srtt_us >> 4),
					  msecs_to_jiffies(10));
L
Linus Torvalds 已提交
2026

2027
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
2028
					  delay, TCP_RTO_MAX);
E
Eric Dumazet 已提交
2029
		return true;
L
Linus Torvalds 已提交
2030
	}
E
Eric Dumazet 已提交
2031
	return false;
L
Linus Torvalds 已提交
2032 2033
}

2034
static inline int tcp_fackets_out(const struct tcp_sock *tp)
L
Linus Torvalds 已提交
2035
{
2036
	return tcp_is_reno(tp) ? tp->sacked_out + 1 : tp->fackets_out;
L
Linus Torvalds 已提交
2037 2038
}

2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053
/* 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.
 */
2054
static inline int tcp_dupack_heuristics(const struct tcp_sock *tp)
2055 2056 2057 2058
{
	return tcp_is_fack(tp) ? tp->fackets_out : tp->sacked_out + 1;
}

L
Linus Torvalds 已提交
2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105
/* 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.
 *
2106
 *		Essentially, we have now a few algorithms detecting
L
Linus Torvalds 已提交
2107 2108
 *		lost packets.
 *
2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123
 *		If the receiver supports SACK:
 *
 *		RFC6675/3517: It is the conventional algorithm. A packet is
 *		considered lost if the number of higher sequence packets
 *		SACKed is greater than or equal the DUPACK thoreshold
 *		(reordering). This is implemented in tcp_mark_head_lost and
 *		tcp_update_scoreboard.
 *
 *		RACK (draft-ietf-tcpm-rack-01): it is a newer algorithm
 *		(2017-) that checks timing instead of counting DUPACKs.
 *		Essentially a packet is considered lost if it's not S/ACKed
 *		after RTT + reordering_window, where both metrics are
 *		dynamically measured and adjusted. This is implemented in
 *		tcp_rack_mark_lost.
 *
Y
Yuchung Cheng 已提交
2124 2125
 *		FACK (Disabled by default. Subsumbed by RACK):
 *		It is the simplest heuristics. As soon as we decided
L
Linus Torvalds 已提交
2126 2127 2128 2129 2130 2131 2132 2133
 *		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.
 *
2134 2135 2136
 *		If the receiver does not support SACK:
 *
 *		NewReno (RFC6582): in Recovery we assume that one segment
L
Linus Torvalds 已提交
2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165
 *		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.
 *
 * 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?
 */
E
Eric Dumazet 已提交
2166
static bool tcp_time_to_recover(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2167
{
2168
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2169 2170 2171

	/* Trick#1: The loss is proven. */
	if (tp->lost_out)
E
Eric Dumazet 已提交
2172
		return true;
L
Linus Torvalds 已提交
2173 2174

	/* Not-A-Trick#2 : Classic rule... */
2175
	if (tcp_dupack_heuristics(tp) > tp->reordering)
E
Eric Dumazet 已提交
2176
		return true;
L
Linus Torvalds 已提交
2177

E
Eric Dumazet 已提交
2178
	return false;
L
Linus Torvalds 已提交
2179 2180
}

2181 2182 2183 2184 2185
/* Detect loss in event "A" above by marking head of queue up as lost.
 * For FACK or non-SACK(Reno) senders, the first "packets" number of segments
 * are considered lost. For RFC3517 SACK, a segment is considered lost if it
 * has at least tp->reordering SACKed seqments above it; "packets" refers to
 * the maximum SACKed segments to pass before reaching this limit.
2186
 */
I
Ilpo Järvinen 已提交
2187
static void tcp_mark_head_lost(struct sock *sk, int packets, int mark_head)
L
Linus Torvalds 已提交
2188
{
2189
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2190
	struct sk_buff *skb;
2191
	int cnt, oldcnt, lost;
2192
	unsigned int mss;
2193 2194
	/* Use SACK to deduce losses of new sequences sent during recovery */
	const u32 loss_high = tcp_is_sack(tp) ?  tp->snd_nxt : tp->high_seq;
L
Linus Torvalds 已提交
2195

2196
	WARN_ON(packets > tp->packets_out);
2197 2198 2199
	if (tp->lost_skb_hint) {
		skb = tp->lost_skb_hint;
		cnt = tp->lost_cnt_hint;
I
Ilpo Järvinen 已提交
2200 2201 2202
		/* Head already handled? */
		if (mark_head && skb != tcp_write_queue_head(sk))
			return;
2203
	} else {
2204
		skb = tcp_write_queue_head(sk);
2205 2206
		cnt = 0;
	}
L
Linus Torvalds 已提交
2207

2208 2209 2210
	tcp_for_write_queue_from(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
2211 2212 2213 2214
		/* TODO: do this better */
		/* this is not the most efficient way to do this... */
		tp->lost_skb_hint = skb;
		tp->lost_cnt_hint = cnt;
2215

2216
		if (after(TCP_SKB_CB(skb)->end_seq, loss_high))
2217 2218 2219
			break;

		oldcnt = cnt;
2220
		if (tcp_is_fack(tp) || tcp_is_reno(tp) ||
2221 2222 2223
		    (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
			cnt += tcp_skb_pcount(skb);

2224
		if (cnt > packets) {
2225
			if ((tcp_is_sack(tp) && !tcp_is_fack(tp)) ||
2226
			    (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED) ||
2227
			    (oldcnt >= packets))
2228 2229
				break;

2230
			mss = tcp_skb_mss(skb);
2231 2232 2233 2234
			/* If needed, chop off the prefix to mark as lost. */
			lost = (packets - oldcnt) * mss;
			if (lost < skb->len &&
			    tcp_fragment(sk, skb, lost, mss, GFP_ATOMIC) < 0)
2235 2236 2237 2238
				break;
			cnt = packets;
		}

2239
		tcp_skb_mark_lost(tp, skb);
I
Ilpo Järvinen 已提交
2240 2241 2242

		if (mark_head)
			break;
L
Linus Torvalds 已提交
2243
	}
2244
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2245 2246 2247 2248
}

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

2249
static void tcp_update_scoreboard(struct sock *sk, int fast_rexmit)
L
Linus Torvalds 已提交
2250
{
2251 2252
	struct tcp_sock *tp = tcp_sk(sk);

2253
	if (tcp_is_reno(tp)) {
I
Ilpo Järvinen 已提交
2254
		tcp_mark_head_lost(sk, 1, 1);
2255
	} else if (tcp_is_fack(tp)) {
L
Linus Torvalds 已提交
2256 2257 2258
		int lost = tp->fackets_out - tp->reordering;
		if (lost <= 0)
			lost = 1;
I
Ilpo Järvinen 已提交
2259
		tcp_mark_head_lost(sk, lost, 0);
L
Linus Torvalds 已提交
2260
	} else {
2261
		int sacked_upto = tp->sacked_out - tp->reordering;
I
Ilpo Järvinen 已提交
2262 2263 2264 2265
		if (sacked_upto >= 0)
			tcp_mark_head_lost(sk, sacked_upto, 0);
		else if (fast_rexmit)
			tcp_mark_head_lost(sk, 1, 1);
L
Linus Torvalds 已提交
2266 2267 2268
	}
}

2269 2270 2271 2272 2273 2274
static bool tcp_tsopt_ecr_before(const struct tcp_sock *tp, u32 when)
{
	return tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr &&
	       before(tp->rx_opt.rcv_tsecr, when);
}

2275 2276 2277 2278 2279 2280 2281 2282 2283 2284
/* skb is spurious retransmitted if the returned timestamp echo
 * reply is prior to the skb transmission time
 */
static bool tcp_skb_spurious_retrans(const struct tcp_sock *tp,
				     const struct sk_buff *skb)
{
	return (TCP_SKB_CB(skb)->sacked & TCPCB_RETRANS) &&
	       tcp_tsopt_ecr_before(tp, tcp_skb_timestamp(skb));
}

L
Linus Torvalds 已提交
2285 2286 2287
/* Nothing was retransmitted or returned timestamp is less
 * than timestamp of the first retransmission.
 */
2288
static inline bool tcp_packet_delayed(const struct tcp_sock *tp)
L
Linus Torvalds 已提交
2289 2290
{
	return !tp->retrans_stamp ||
2291
	       tcp_tsopt_ecr_before(tp, tp->retrans_stamp);
L
Linus Torvalds 已提交
2292 2293 2294 2295
}

/* Undo procedures. */

2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324
/* 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 bool tcp_any_retrans_done(const struct sock *sk)
{
	const struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;

	if (tp->retrans_out)
		return true;

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

	return false;
}

L
Linus Torvalds 已提交
2325
#if FASTRETRANS_DEBUG > 1
2326
static void DBGUNDO(struct sock *sk, const char *msg)
L
Linus Torvalds 已提交
2327
{
2328
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2329
	struct inet_sock *inet = inet_sk(sk);
2330

2331
	if (sk->sk_family == AF_INET) {
2332 2333 2334 2335 2336 2337
		pr_debug("Undo %s %pI4/%u c%u l%u ss%u/%u p%u\n",
			 msg,
			 &inet->inet_daddr, ntohs(inet->inet_dport),
			 tp->snd_cwnd, tcp_left_out(tp),
			 tp->snd_ssthresh, tp->prior_ssthresh,
			 tp->packets_out);
2338
	}
E
Eric Dumazet 已提交
2339
#if IS_ENABLED(CONFIG_IPV6)
2340
	else if (sk->sk_family == AF_INET6) {
2341 2342
		pr_debug("Undo %s %pI6/%u c%u l%u ss%u/%u p%u\n",
			 msg,
2343
			 &sk->sk_v6_daddr, ntohs(inet->inet_dport),
2344 2345 2346
			 tp->snd_cwnd, tcp_left_out(tp),
			 tp->snd_ssthresh, tp->prior_ssthresh,
			 tp->packets_out);
2347 2348
	}
#endif
L
Linus Torvalds 已提交
2349 2350 2351 2352 2353
}
#else
#define DBGUNDO(x...) do { } while (0)
#endif

2354
static void tcp_undo_cwnd_reduction(struct sock *sk, bool unmark_loss)
L
Linus Torvalds 已提交
2355
{
2356 2357
	struct tcp_sock *tp = tcp_sk(sk);

Y
Yuchung Cheng 已提交
2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369
	if (unmark_loss) {
		struct sk_buff *skb;

		tcp_for_write_queue(skb, sk) {
			if (skb == tcp_send_head(sk))
				break;
			TCP_SKB_CB(skb)->sacked &= ~TCPCB_LOST;
		}
		tp->lost_out = 0;
		tcp_clear_all_retrans_hints(tp);
	}

L
Linus Torvalds 已提交
2370
	if (tp->prior_ssthresh) {
2371 2372
		const struct inet_connection_sock *icsk = inet_csk(sk);

2373
		tp->snd_cwnd = icsk->icsk_ca_ops->undo_cwnd(sk);
L
Linus Torvalds 已提交
2374

2375
		if (tp->prior_ssthresh > tp->snd_ssthresh) {
L
Linus Torvalds 已提交
2376
			tp->snd_ssthresh = tp->prior_ssthresh;
2377
			tcp_ecn_withdraw_cwr(tp);
L
Linus Torvalds 已提交
2378 2379 2380
		}
	}
	tp->snd_cwnd_stamp = tcp_time_stamp;
2381
	tp->undo_marker = 0;
L
Linus Torvalds 已提交
2382 2383
}

2384
static inline bool tcp_may_undo(const struct tcp_sock *tp)
L
Linus Torvalds 已提交
2385
{
2386
	return tp->undo_marker && (!tp->undo_retrans || tcp_packet_delayed(tp));
L
Linus Torvalds 已提交
2387 2388 2389
}

/* People celebrate: "We love our President!" */
E
Eric Dumazet 已提交
2390
static bool tcp_try_undo_recovery(struct sock *sk)
L
Linus Torvalds 已提交
2391
{
2392 2393
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2394
	if (tcp_may_undo(tp)) {
2395 2396
		int mib_idx;

L
Linus Torvalds 已提交
2397 2398 2399
		/* Happy end! We did not retransmit anything
		 * or our original transmission succeeded.
		 */
2400
		DBGUNDO(sk, inet_csk(sk)->icsk_ca_state == TCP_CA_Loss ? "loss" : "retrans");
2401
		tcp_undo_cwnd_reduction(sk, false);
2402
		if (inet_csk(sk)->icsk_ca_state == TCP_CA_Loss)
2403
			mib_idx = LINUX_MIB_TCPLOSSUNDO;
L
Linus Torvalds 已提交
2404
		else
2405 2406
			mib_idx = LINUX_MIB_TCPFULLUNDO;

2407
		NET_INC_STATS(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
2408
	}
2409
	if (tp->snd_una == tp->high_seq && tcp_is_reno(tp)) {
L
Linus Torvalds 已提交
2410 2411 2412
		/* 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. */
2413 2414
		if (!tcp_any_retrans_done(sk))
			tp->retrans_stamp = 0;
E
Eric Dumazet 已提交
2415
		return true;
L
Linus Torvalds 已提交
2416
	}
2417
	tcp_set_ca_state(sk, TCP_CA_Open);
E
Eric Dumazet 已提交
2418
	return false;
L
Linus Torvalds 已提交
2419 2420 2421
}

/* Try to undo cwnd reduction, because D-SACKs acked all retransmitted data */
2422
static bool tcp_try_undo_dsack(struct sock *sk)
L
Linus Torvalds 已提交
2423
{
2424 2425
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2426
	if (tp->undo_marker && !tp->undo_retrans) {
2427
		DBGUNDO(sk, "D-SACK");
2428
		tcp_undo_cwnd_reduction(sk, false);
2429
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPDSACKUNDO);
2430
		return true;
L
Linus Torvalds 已提交
2431
	}
2432
	return false;
L
Linus Torvalds 已提交
2433 2434
}

Y
Yuchung Cheng 已提交
2435 2436
/* Undo during loss recovery after partial ACK or using F-RTO. */
static bool tcp_try_undo_loss(struct sock *sk, bool frto_undo)
L
Linus Torvalds 已提交
2437
{
2438 2439
	struct tcp_sock *tp = tcp_sk(sk);

Y
Yuchung Cheng 已提交
2440
	if (frto_undo || tcp_may_undo(tp)) {
2441
		tcp_undo_cwnd_reduction(sk, true);
2442

2443
		DBGUNDO(sk, "partial loss");
2444
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPLOSSUNDO);
Y
Yuchung Cheng 已提交
2445
		if (frto_undo)
2446
			NET_INC_STATS(sock_net(sk),
2447
					LINUX_MIB_TCPSPURIOUSRTOS);
2448
		inet_csk(sk)->icsk_retransmits = 0;
Y
Yuchung Cheng 已提交
2449
		if (frto_undo || tcp_is_sack(tp))
2450
			tcp_set_ca_state(sk, TCP_CA_Open);
E
Eric Dumazet 已提交
2451
		return true;
L
Linus Torvalds 已提交
2452
	}
E
Eric Dumazet 已提交
2453
	return false;
L
Linus Torvalds 已提交
2454 2455
}

2456
/* The cwnd reduction in CWR and Recovery uses the PRR algorithm in RFC 6937.
2457 2458 2459 2460
 * It computes the number of packets to send (sndcnt) based on packets newly
 * delivered:
 *   1) If the packets in flight is larger than ssthresh, PRR spreads the
 *	cwnd reductions across a full RTT.
2461 2462 2463
 *   2) Otherwise PRR uses packet conservation to send as much as delivered.
 *      But when the retransmits are acked without further losses, PRR
 *      slow starts cwnd up to ssthresh to speed up the recovery.
2464
 */
2465
static void tcp_init_cwnd_reduction(struct sock *sk)
2466 2467 2468 2469
{
	struct tcp_sock *tp = tcp_sk(sk);

	tp->high_seq = tp->snd_nxt;
N
Nandita Dukkipati 已提交
2470
	tp->tlp_high_seq = 0;
2471 2472 2473 2474
	tp->snd_cwnd_cnt = 0;
	tp->prior_cwnd = tp->snd_cwnd;
	tp->prr_delivered = 0;
	tp->prr_out = 0;
2475
	tp->snd_ssthresh = inet_csk(sk)->icsk_ca_ops->ssthresh(sk);
2476
	tcp_ecn_queue_cwr(tp);
2477 2478
}

2479
void tcp_cwnd_reduction(struct sock *sk, int newly_acked_sacked, int flag)
2480 2481 2482 2483 2484
{
	struct tcp_sock *tp = tcp_sk(sk);
	int sndcnt = 0;
	int delta = tp->snd_ssthresh - tcp_packets_in_flight(tp);

2485 2486 2487
	if (newly_acked_sacked <= 0 || WARN_ON_ONCE(!tp->prior_cwnd))
		return;

2488
	tp->prr_delivered += newly_acked_sacked;
2489
	if (delta < 0) {
2490 2491 2492
		u64 dividend = (u64)tp->snd_ssthresh * tp->prr_delivered +
			       tp->prior_cwnd - 1;
		sndcnt = div_u64(dividend, tp->prior_cwnd) - tp->prr_out;
2493 2494
	} else if ((flag & FLAG_RETRANS_DATA_ACKED) &&
		   !(flag & FLAG_LOST_RETRANS)) {
2495 2496 2497
		sndcnt = min_t(int, delta,
			       max_t(int, tp->prr_delivered - tp->prr_out,
				     newly_acked_sacked) + 1);
2498 2499
	} else {
		sndcnt = min(delta, newly_acked_sacked);
2500
	}
2501 2502
	/* Force a fast retransmit upon entering fast recovery */
	sndcnt = max(sndcnt, (tp->prr_out ? 0 : 1));
2503 2504 2505
	tp->snd_cwnd = tcp_packets_in_flight(tp) + sndcnt;
}

2506
static inline void tcp_end_cwnd_reduction(struct sock *sk)
L
Linus Torvalds 已提交
2507
{
2508
	struct tcp_sock *tp = tcp_sk(sk);
2509

2510 2511 2512
	if (inet_csk(sk)->icsk_ca_ops->cong_control)
		return;

2513 2514 2515 2516 2517
	/* Reset cwnd to ssthresh in CWR or Recovery (unless it's undone) */
	if (inet_csk(sk)->icsk_ca_state == TCP_CA_CWR ||
	    (tp->undo_marker && tp->snd_ssthresh < TCP_INFINITE_SSTHRESH)) {
		tp->snd_cwnd = tp->snd_ssthresh;
		tp->snd_cwnd_stamp = tcp_time_stamp;
2518
	}
2519
	tcp_ca_event(sk, CA_EVENT_COMPLETE_CWR);
L
Linus Torvalds 已提交
2520 2521
}

2522
/* Enter CWR state. Disable cwnd undo since congestion is proven with ECN */
2523
void tcp_enter_cwr(struct sock *sk)
Y
Yuchung Cheng 已提交
2524 2525 2526 2527
{
	struct tcp_sock *tp = tcp_sk(sk);

	tp->prior_ssthresh = 0;
2528
	if (inet_csk(sk)->icsk_ca_state < TCP_CA_CWR) {
Y
Yuchung Cheng 已提交
2529
		tp->undo_marker = 0;
2530
		tcp_init_cwnd_reduction(sk);
Y
Yuchung Cheng 已提交
2531 2532 2533
		tcp_set_ca_state(sk, TCP_CA_CWR);
	}
}
2534
EXPORT_SYMBOL(tcp_enter_cwr);
Y
Yuchung Cheng 已提交
2535

2536 2537 2538 2539 2540
static void tcp_try_keep_open(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	int state = TCP_CA_Open;

2541
	if (tcp_left_out(tp) || tcp_any_retrans_done(sk))
2542 2543 2544 2545 2546 2547 2548 2549
		state = TCP_CA_Disorder;

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

2550
static void tcp_try_to_open(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2551
{
2552 2553
	struct tcp_sock *tp = tcp_sk(sk);

2554 2555
	tcp_verify_left_out(tp);

Y
Yuchung Cheng 已提交
2556
	if (!tcp_any_retrans_done(sk))
L
Linus Torvalds 已提交
2557 2558
		tp->retrans_stamp = 0;

2559
	if (flag & FLAG_ECE)
2560
		tcp_enter_cwr(sk);
L
Linus Torvalds 已提交
2561

2562
	if (inet_csk(sk)->icsk_ca_state != TCP_CA_CWR) {
2563
		tcp_try_keep_open(sk);
L
Linus Torvalds 已提交
2564 2565 2566
	}
}

J
John Heffner 已提交
2567 2568 2569 2570 2571 2572
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;
2573
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMTUPFAIL);
J
John Heffner 已提交
2574 2575
}

2576
static void tcp_mtup_probe_success(struct sock *sk)
J
John Heffner 已提交
2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587
{
	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;
2588
	tp->snd_ssthresh = tcp_current_ssthresh(sk);
J
John Heffner 已提交
2589 2590 2591 2592

	icsk->icsk_mtup.search_low = icsk->icsk_mtup.probe_size;
	icsk->icsk_mtup.probe_size = 0;
	tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
2593
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMTUPSUCCESS);
J
John Heffner 已提交
2594 2595
}

2596 2597 2598 2599 2600 2601 2602 2603 2604
/* 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
2605
	unsigned int mss = tcp_current_mss(sk);
2606 2607 2608 2609 2610
	u32 prior_lost = tp->lost_out;

	tcp_for_write_queue(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
2611
		if (tcp_skb_seglen(skb) > mss &&
2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644
		    !(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);
}
E
Eric Dumazet 已提交
2645
EXPORT_SYMBOL(tcp_simple_retransmit);
2646

2647
void tcp_enter_recovery(struct sock *sk, bool ece_ack)
2648 2649 2650 2651 2652 2653 2654 2655 2656
{
	struct tcp_sock *tp = tcp_sk(sk);
	int mib_idx;

	if (tcp_is_reno(tp))
		mib_idx = LINUX_MIB_TCPRENORECOVERY;
	else
		mib_idx = LINUX_MIB_TCPSACKRECOVERY;

2657
	NET_INC_STATS(sock_net(sk), mib_idx);
2658 2659

	tp->prior_ssthresh = 0;
Y
Yuchung Cheng 已提交
2660
	tcp_init_undo(tp);
2661

2662
	if (!tcp_in_cwnd_reduction(sk)) {
2663 2664
		if (!ece_ack)
			tp->prior_ssthresh = tcp_current_ssthresh(sk);
2665
		tcp_init_cwnd_reduction(sk);
2666 2667 2668 2669
	}
	tcp_set_ca_state(sk, TCP_CA_Recovery);
}

2670 2671 2672
/* Process an ACK in CA_Loss state. Move to CA_Open if lost data are
 * recovered or spurious. Otherwise retransmits more on partial ACKs.
 */
2673 2674
static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack,
			     int *rexmit)
2675 2676
{
	struct tcp_sock *tp = tcp_sk(sk);
Y
Yuchung Cheng 已提交
2677
	bool recovered = !before(tp->snd_una, tp->high_seq);
2678

2679 2680 2681 2682
	if ((flag & FLAG_SND_UNA_ADVANCED) &&
	    tcp_try_undo_loss(sk, false))
		return;

2683 2684 2685 2686 2687 2688 2689 2690 2691 2692
	/* The ACK (s)acks some never-retransmitted data meaning not all
	 * the data packets before the timeout were lost. Therefore we
	 * undo the congestion window and state. This is essentially
	 * the operation in F-RTO (RFC5682 section 3.1 step 3.b). Since
	 * a retransmitted skb is permantly marked, we can apply such an
	 * operation even if F-RTO was not used.
	 */
	if ((flag & FLAG_ORIG_SACK_ACKED) &&
	    tcp_try_undo_loss(sk, tp->undo_marker))
		return;
2693

2694
	if (tp->frto) { /* F-RTO RFC5682 sec 3.1 (sack enhanced version). */
2695 2696 2697
		if (after(tp->snd_nxt, tp->high_seq)) {
			if (flag & FLAG_DATA_SACKED || is_dupack)
				tp->frto = 0; /* Step 3.a. loss was real */
Y
Yuchung Cheng 已提交
2698 2699
		} else if (flag & FLAG_SND_UNA_ADVANCED && !recovered) {
			tp->high_seq = tp->snd_nxt;
2700 2701 2702 2703 2704 2705 2706 2707 2708
			/* Step 2.b. Try send new data (but deferred until cwnd
			 * is updated in tcp_ack()). Otherwise fall back to
			 * the conventional recovery.
			 */
			if (tcp_send_head(sk) &&
			    after(tcp_wnd_end(tp), tp->snd_nxt)) {
				*rexmit = REXMIT_NEW;
				return;
			}
Y
Yuchung Cheng 已提交
2709 2710 2711 2712 2713 2714
			tp->frto = 0;
		}
	}

	if (recovered) {
		/* F-RTO RFC5682 sec 3.1 step 2.a and 1st part of step 3.a */
2715 2716 2717
		tcp_try_undo_recovery(sk);
		return;
	}
Y
Yuchung Cheng 已提交
2718 2719 2720 2721 2722 2723 2724 2725 2726
	if (tcp_is_reno(tp)) {
		/* A Reno DUPACK means new data in F-RTO step 2.b above are
		 * delivered. Lower inflight to clock out (re)tranmissions.
		 */
		if (after(tp->snd_nxt, tp->high_seq) && is_dupack)
			tcp_add_reno_sack(sk);
		else if (flag & FLAG_SND_UNA_ADVANCED)
			tcp_reset_reno_sack(tp);
	}
2727
	*rexmit = REXMIT_LOST;
2728 2729
}

Y
Yuchung Cheng 已提交
2730
/* Undo during fast recovery after partial ACK. */
2731
static bool tcp_try_undo_partial(struct sock *sk, const int acked)
Y
Yuchung Cheng 已提交
2732 2733 2734
{
	struct tcp_sock *tp = tcp_sk(sk);

2735
	if (tp->undo_marker && tcp_packet_delayed(tp)) {
Y
Yuchung Cheng 已提交
2736 2737 2738
		/* Plain luck! Hole if filled with delayed
		 * packet, rather than with a retransmit.
		 */
2739 2740 2741 2742 2743 2744 2745
		tcp_update_reordering(sk, tcp_fackets_out(tp) + acked, 1);

		/* We are getting evidence that the reordering degree is higher
		 * than we realized. If there are no retransmits out then we
		 * can undo. Otherwise we clock out new packets but do not
		 * mark more packets lost or retransmit more.
		 */
2746
		if (tp->retrans_out)
2747 2748
			return true;

Y
Yuchung Cheng 已提交
2749 2750 2751
		if (!tcp_any_retrans_done(sk))
			tp->retrans_stamp = 0;

2752 2753
		DBGUNDO(sk, "partial recovery");
		tcp_undo_cwnd_reduction(sk, true);
2754
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPPARTIALUNDO);
2755 2756
		tcp_try_keep_open(sk);
		return true;
Y
Yuchung Cheng 已提交
2757
	}
2758
	return false;
Y
Yuchung Cheng 已提交
2759 2760
}

2761
static void tcp_rack_identify_loss(struct sock *sk, int *ack_flag)
2762 2763 2764 2765
{
	struct tcp_sock *tp = tcp_sk(sk);

	/* Use RACK to detect loss */
2766
	if (sysctl_tcp_recovery & TCP_RACK_LOSS_DETECTION) {
2767 2768
		u32 prior_retrans = tp->retrans_out;

2769
		tcp_rack_mark_lost(sk);
2770 2771 2772 2773 2774
		if (prior_retrans > tp->retrans_out)
			*ack_flag |= FLAG_LOST_RETRANS;
	}
}

L
Linus Torvalds 已提交
2775 2776 2777 2778 2779
/* 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.
 *
2780 2781 2782
 * Besides that it updates the congestion state when packet loss or ECN
 * is detected. But it does not reduce the cwnd, it is done by the
 * congestion control later.
L
Linus Torvalds 已提交
2783 2784 2785 2786
 *
 * It does _not_ decide what to send, it is made in function
 * tcp_xmit_retransmit_queue().
 */
2787
static void tcp_fastretrans_alert(struct sock *sk, const int acked,
2788
				  bool is_dupack, int *ack_flag, int *rexmit)
L
Linus Torvalds 已提交
2789
{
2790
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2791
	struct tcp_sock *tp = tcp_sk(sk);
2792
	int fast_rexmit = 0, flag = *ack_flag;
Y
Yuchung Cheng 已提交
2793
	bool do_lost = is_dupack || ((flag & FLAG_DATA_SACKED) &&
2794
				    (tcp_fackets_out(tp) > tp->reordering));
L
Linus Torvalds 已提交
2795

2796
	if (WARN_ON(!tp->packets_out && tp->sacked_out))
L
Linus Torvalds 已提交
2797
		tp->sacked_out = 0;
I
Ilpo Järvinen 已提交
2798
	if (WARN_ON(!tp->sacked_out && tp->fackets_out))
L
Linus Torvalds 已提交
2799 2800
		tp->fackets_out = 0;

2801
	/* Now state machine starts.
L
Linus Torvalds 已提交
2802
	 * A. ECE, hence prohibit cwnd undoing, the reduction is required. */
2803
	if (flag & FLAG_ECE)
L
Linus Torvalds 已提交
2804 2805 2806
		tp->prior_ssthresh = 0;

	/* B. In all the states check for reneging SACKs. */
2807
	if (tcp_check_sack_reneging(sk, flag))
L
Linus Torvalds 已提交
2808 2809
		return;

2810
	/* C. Check consistency of the current state. */
2811
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2812

2813
	/* D. Check state exit conditions. State can be terminated
L
Linus Torvalds 已提交
2814
	 *    when high_seq is ACKed. */
2815
	if (icsk->icsk_ca_state == TCP_CA_Open) {
2816
		WARN_ON(tp->retrans_out != 0);
L
Linus Torvalds 已提交
2817 2818
		tp->retrans_stamp = 0;
	} else if (!before(tp->snd_una, tp->high_seq)) {
2819
		switch (icsk->icsk_ca_state) {
L
Linus Torvalds 已提交
2820 2821 2822 2823
		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) {
2824
				tcp_end_cwnd_reduction(sk);
2825
				tcp_set_ca_state(sk, TCP_CA_Open);
L
Linus Torvalds 已提交
2826 2827 2828 2829
			}
			break;

		case TCP_CA_Recovery:
2830
			if (tcp_is_reno(tp))
L
Linus Torvalds 已提交
2831
				tcp_reset_reno_sack(tp);
2832
			if (tcp_try_undo_recovery(sk))
L
Linus Torvalds 已提交
2833
				return;
2834
			tcp_end_cwnd_reduction(sk);
L
Linus Torvalds 已提交
2835 2836 2837 2838
			break;
		}
	}

2839
	/* E. Process state. */
2840
	switch (icsk->icsk_ca_state) {
L
Linus Torvalds 已提交
2841
	case TCP_CA_Recovery:
2842
		if (!(flag & FLAG_SND_UNA_ADVANCED)) {
2843
			if (tcp_is_reno(tp) && is_dupack)
2844
				tcp_add_reno_sack(sk);
2845
		} else {
2846
			if (tcp_try_undo_partial(sk, acked))
2847 2848 2849 2850 2851
				return;
			/* Partial ACK arrived. Force fast retransmit. */
			do_lost = tcp_is_reno(tp) ||
				  tcp_fackets_out(tp) > tp->reordering;
		}
2852 2853 2854 2855
		if (tcp_try_undo_dsack(sk)) {
			tcp_try_keep_open(sk);
			return;
		}
2856
		tcp_rack_identify_loss(sk, ack_flag);
L
Linus Torvalds 已提交
2857 2858
		break;
	case TCP_CA_Loss:
2859
		tcp_process_loss(sk, flag, is_dupack, rexmit);
2860
		tcp_rack_identify_loss(sk, ack_flag);
2861 2862
		if (!(icsk->icsk_ca_state == TCP_CA_Open ||
		      (*ack_flag & FLAG_LOST_RETRANS)))
L
Linus Torvalds 已提交
2863
			return;
2864
		/* Change state if cwnd is undone or retransmits are lost */
L
Linus Torvalds 已提交
2865
	default:
2866
		if (tcp_is_reno(tp)) {
2867
			if (flag & FLAG_SND_UNA_ADVANCED)
L
Linus Torvalds 已提交
2868 2869
				tcp_reset_reno_sack(tp);
			if (is_dupack)
2870
				tcp_add_reno_sack(sk);
L
Linus Torvalds 已提交
2871 2872
		}

2873
		if (icsk->icsk_ca_state <= TCP_CA_Disorder)
2874
			tcp_try_undo_dsack(sk);
L
Linus Torvalds 已提交
2875

2876
		tcp_rack_identify_loss(sk, ack_flag);
2877
		if (!tcp_time_to_recover(sk, flag)) {
2878
			tcp_try_to_open(sk, flag);
L
Linus Torvalds 已提交
2879 2880 2881
			return;
		}

J
John Heffner 已提交
2882 2883 2884
		/* MTU probe failure: don't reduce cwnd */
		if (icsk->icsk_ca_state < TCP_CA_CWR &&
		    icsk->icsk_mtup.probe_size &&
2885
		    tp->snd_una == tp->mtu_probe.probe_seq_start) {
J
John Heffner 已提交
2886 2887 2888 2889 2890 2891 2892
			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
2893
		/* Otherwise enter Recovery state */
2894
		tcp_enter_recovery(sk, (flag & FLAG_ECE));
2895
		fast_rexmit = 1;
L
Linus Torvalds 已提交
2896 2897
	}

2898
	if (do_lost)
2899
		tcp_update_scoreboard(sk, fast_rexmit);
2900
	*rexmit = REXMIT_LOST;
L
Linus Torvalds 已提交
2901 2902
}

2903 2904
static void tcp_update_rtt_min(struct sock *sk, u32 rtt_us)
{
2905 2906 2907 2908 2909
	struct tcp_sock *tp = tcp_sk(sk);
	u32 wlen = sysctl_tcp_min_rtt_wlen * HZ;

	minmax_running_min(&tp->rtt_min, wlen, tcp_time_stamp,
			   rtt_us ? : jiffies_to_usecs(1));
2910 2911
}

Y
Yuchung Cheng 已提交
2912
static inline bool tcp_ack_update_rtt(struct sock *sk, const int flag,
2913 2914
				      long seq_rtt_us, long sack_rtt_us,
				      long ca_rtt_us)
2915
{
2916 2917 2918 2919 2920 2921 2922
	const struct tcp_sock *tp = tcp_sk(sk);

	/* Prefer RTT measured from ACK's timing to TS-ECR. This is because
	 * broken middle-boxes or peers may corrupt TS-ECR fields. But
	 * Karn's algorithm forbids taking RTT if some retransmitted data
	 * is acked (RFC6298).
	 */
2923 2924
	if (seq_rtt_us < 0)
		seq_rtt_us = sack_rtt_us;
Y
Yuchung Cheng 已提交
2925

L
Linus Torvalds 已提交
2926 2927 2928 2929 2930 2931
	/* 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.
	 */
2932
	if (seq_rtt_us < 0 && tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr &&
2933
	    flag & FLAG_ACKED)
2934 2935
		seq_rtt_us = ca_rtt_us = jiffies_to_usecs(tcp_time_stamp -
							  tp->rx_opt.rcv_tsecr);
2936
	if (seq_rtt_us < 0)
Y
Yuchung Cheng 已提交
2937
		return false;
L
Linus Torvalds 已提交
2938

2939 2940 2941 2942 2943
	/* ca_rtt_us >= 0 is counting on the invariant that ca_rtt_us is
	 * always taken together with ACK, SACK, or TS-opts. Any negative
	 * values will be skipped with the seq_rtt_us < 0 check above.
	 */
	tcp_update_rtt_min(sk, ca_rtt_us);
2944
	tcp_rtt_estimator(sk, seq_rtt_us);
2945
	tcp_set_rto(sk);
L
Linus Torvalds 已提交
2946

2947 2948
	/* RFC6298: only reset backoff on valid RTT measurement. */
	inet_csk(sk)->icsk_backoff = 0;
Y
Yuchung Cheng 已提交
2949
	return true;
L
Linus Torvalds 已提交
2950 2951
}

2952
/* Compute time elapsed between (last) SYNACK and the ACK completing 3WHS. */
Y
Yuchung Cheng 已提交
2953
void tcp_synack_rtt_meas(struct sock *sk, struct request_sock *req)
2954
{
Y
Yuchung Cheng 已提交
2955
	long rtt_us = -1L;
2956

Y
Yuchung Cheng 已提交
2957 2958
	if (req && !req->num_retrans && tcp_rsk(req)->snt_synack.v64) {
		struct skb_mstamp now;
2959

Y
Yuchung Cheng 已提交
2960 2961 2962 2963
		skb_mstamp_get(&now);
		rtt_us = skb_mstamp_us_delta(&now, &tcp_rsk(req)->snt_synack);
	}

2964
	tcp_ack_update_rtt(sk, FLAG_SYN_ACKED, rtt_us, -1L, rtt_us);
2965 2966
}

Y
Yuchung Cheng 已提交
2967

2968
static void tcp_cong_avoid(struct sock *sk, u32 ack, u32 acked)
L
Linus Torvalds 已提交
2969
{
2970
	const struct inet_connection_sock *icsk = inet_csk(sk);
2971 2972

	icsk->icsk_ca_ops->cong_avoid(sk, ack, acked);
2973
	tcp_sk(sk)->snd_cwnd_stamp = tcp_time_stamp;
L
Linus Torvalds 已提交
2974 2975 2976 2977 2978
}

/* Restart timer after forward progress on connection.
 * RFC2988 recommends to restart timer to now+rto.
 */
2979
void tcp_rearm_rto(struct sock *sk)
L
Linus Torvalds 已提交
2980
{
N
Nandita Dukkipati 已提交
2981
	const struct inet_connection_sock *icsk = inet_csk(sk);
2982
	struct tcp_sock *tp = tcp_sk(sk);
2983

2984 2985 2986 2987 2988 2989
	/* If the retrans timer is currently being used by Fast Open
	 * for SYN-ACK retrans purpose, stay put.
	 */
	if (tp->fastopen_rsk)
		return;

L
Linus Torvalds 已提交
2990
	if (!tp->packets_out) {
2991
		inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS);
L
Linus Torvalds 已提交
2992
	} else {
2993 2994
		u32 rto = inet_csk(sk)->icsk_rto;
		/* Offset the time elapsed after installing regular RTO */
Y
Yuchung Cheng 已提交
2995
		if (icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
N
Nandita Dukkipati 已提交
2996
		    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
2997
			struct sk_buff *skb = tcp_write_queue_head(sk);
E
Eric Dumazet 已提交
2998 2999
			const u32 rto_time_stamp =
				tcp_skb_timestamp(skb) + rto;
3000 3001
			s32 delta = (s32)(rto_time_stamp - tcp_time_stamp);
			/* delta may not be positive if the socket is locked
N
Nandita Dukkipati 已提交
3002
			 * when the retrans timer fires and is rescheduled.
3003 3004 3005 3006 3007 3008
			 */
			if (delta > 0)
				rto = delta;
		}
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, rto,
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
3009
	}
3010 3011
}

3012
/* If we get here, the whole TSO packet has not been acked. */
3013
static u32 tcp_tso_acked(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
3014 3015
{
	struct tcp_sock *tp = tcp_sk(sk);
3016
	u32 packets_acked;
L
Linus Torvalds 已提交
3017

3018
	BUG_ON(!after(TCP_SKB_CB(skb)->end_seq, tp->snd_una));
L
Linus Torvalds 已提交
3019 3020

	packets_acked = tcp_skb_pcount(skb);
3021
	if (tcp_trim_head(sk, skb, tp->snd_una - TCP_SKB_CB(skb)->seq))
L
Linus Torvalds 已提交
3022 3023 3024 3025 3026
		return 0;
	packets_acked -= tcp_skb_pcount(skb);

	if (packets_acked) {
		BUG_ON(tcp_skb_pcount(skb) == 0);
3027
		BUG_ON(!before(TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq));
L
Linus Torvalds 已提交
3028 3029
	}

3030
	return packets_acked;
L
Linus Torvalds 已提交
3031 3032
}

3033 3034 3035 3036 3037 3038
static void tcp_ack_tstamp(struct sock *sk, struct sk_buff *skb,
			   u32 prior_snd_una)
{
	const struct skb_shared_info *shinfo;

	/* Avoid cache line misses to get skb_shinfo() and shinfo->tx_flags */
3039
	if (likely(!TCP_SKB_CB(skb)->txstamp_ack))
3040 3041 3042
		return;

	shinfo = skb_shinfo(skb);
3043
	if (!before(shinfo->tskey, prior_snd_una) &&
3044
	    before(shinfo->tskey, tcp_sk(sk)->snd_una))
3045 3046 3047
		__skb_tstamp_tx(skb, NULL, sk, SCM_TSTAMP_ACK);
}

3048 3049 3050 3051
/* 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
3052
static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets,
3053
			       u32 prior_snd_una, int *acked,
3054
			       struct tcp_sacktag_state *sack)
L
Linus Torvalds 已提交
3055
{
3056
	const struct inet_connection_sock *icsk = inet_csk(sk);
3057
	struct skb_mstamp first_ackt, last_ackt;
3058
	struct tcp_sock *tp = tcp_sk(sk);
3059
	struct skb_mstamp *now = &tp->tcp_mstamp;
3060 3061
	u32 prior_sacked = tp->sacked_out;
	u32 reord = tp->packets_out;
3062
	bool fully_acked = true;
3063
	long sack_rtt_us = -1L;
3064
	long seq_rtt_us = -1L;
3065
	long ca_rtt_us = -1L;
3066
	struct sk_buff *skb;
3067
	u32 pkts_acked = 0;
3068
	u32 last_in_flight = 0;
3069
	bool rtt_update;
3070 3071 3072
	int flag = 0;

	first_ackt.v64 = 0;
L
Linus Torvalds 已提交
3073

3074
	while ((skb = tcp_write_queue_head(sk)) && skb != tcp_send_head(sk)) {
3075
		struct tcp_skb_cb *scb = TCP_SKB_CB(skb);
3076
		u8 sacked = scb->sacked;
3077
		u32 acked_pcount;
L
Linus Torvalds 已提交
3078

3079
		tcp_ack_tstamp(sk, skb, prior_snd_una);
3080

3081
		/* Determine how many packets and what bytes were acked, tso and else */
L
Linus Torvalds 已提交
3082
		if (after(scb->end_seq, tp->snd_una)) {
3083 3084 3085 3086
			if (tcp_skb_pcount(skb) == 1 ||
			    !after(tp->snd_una, scb->seq))
				break;

3087 3088
			acked_pcount = tcp_tso_acked(sk, skb);
			if (!acked_pcount)
3089
				break;
E
Eric Dumazet 已提交
3090
			fully_acked = false;
3091
		} else {
3092 3093
			/* Speedup tcp_unlink_write_queue() and next loop */
			prefetchw(skb->next);
3094
			acked_pcount = tcp_skb_pcount(skb);
L
Linus Torvalds 已提交
3095 3096
		}

3097
		if (unlikely(sacked & TCPCB_RETRANS)) {
3098
			if (sacked & TCPCB_SACKED_RETRANS)
3099
				tp->retrans_out -= acked_pcount;
3100
			flag |= FLAG_RETRANS_DATA_ACKED;
3101
		} else if (!(sacked & TCPCB_SACKED_ACKED)) {
3102
			last_ackt = skb->skb_mstamp;
E
Eric Dumazet 已提交
3103
			WARN_ON_ONCE(last_ackt.v64 == 0);
3104 3105 3106
			if (!first_ackt.v64)
				first_ackt = last_ackt;

3107
			last_in_flight = TCP_SKB_CB(skb)->tx.in_flight;
3108 3109 3110
			reord = min(pkts_acked, reord);
			if (!after(scb->end_seq, tp->high_seq))
				flag |= FLAG_ORIG_SACK_ACKED;
3111
		}
3112

Y
Yuchung Cheng 已提交
3113
		if (sacked & TCPCB_SACKED_ACKED) {
3114
			tp->sacked_out -= acked_pcount;
Y
Yuchung Cheng 已提交
3115 3116 3117
		} else if (tcp_is_sack(tp)) {
			tp->delivered += acked_pcount;
			if (!tcp_skb_spurious_retrans(tp, skb))
3118
				tcp_rack_advance(tp, sacked, scb->end_seq,
3119
						 &skb->skb_mstamp);
Y
Yuchung Cheng 已提交
3120
		}
3121
		if (sacked & TCPCB_LOST)
3122
			tp->lost_out -= acked_pcount;
3123

3124 3125
		tp->packets_out -= acked_pcount;
		pkts_acked += acked_pcount;
3126
		tcp_rate_skb_delivered(sk, skb, sack->rate);
3127

3128 3129 3130 3131 3132 3133 3134
		/* 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.
		 */
3135
		if (likely(!(scb->tcp_flags & TCPHDR_SYN))) {
3136 3137 3138 3139 3140 3141
			flag |= FLAG_DATA_ACKED;
		} else {
			flag |= FLAG_SYN_ACKED;
			tp->retrans_stamp = 0;
		}

3142 3143 3144
		if (!fully_acked)
			break;

3145
		tcp_unlink_write_queue(skb, sk);
3146
		sk_wmem_free_skb(sk, skb);
3147
		if (unlikely(skb == tp->retransmit_skb_hint))
3148
			tp->retransmit_skb_hint = NULL;
3149
		if (unlikely(skb == tp->lost_skb_hint))
3150
			tp->lost_skb_hint = NULL;
L
Linus Torvalds 已提交
3151 3152
	}

3153 3154 3155
	if (!skb)
		tcp_chrono_stop(sk, TCP_CHRONO_BUSY);

I
Ilpo Järvinen 已提交
3156 3157 3158
	if (likely(between(tp->snd_up, prior_snd_una, tp->snd_una)))
		tp->snd_up = tp->snd_una;

3159 3160 3161
	if (skb && (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
		flag |= FLAG_SACK_RENEGING;

3162
	if (likely(first_ackt.v64) && !(flag & FLAG_RETRANS_DATA_ACKED)) {
3163 3164
		seq_rtt_us = skb_mstamp_us_delta(now, &first_ackt);
		ca_rtt_us = skb_mstamp_us_delta(now, &last_ackt);
3165 3166
	}
	if (sack->first_sackt.v64) {
3167 3168
		sack_rtt_us = skb_mstamp_us_delta(now, &sack->first_sackt);
		ca_rtt_us = skb_mstamp_us_delta(now, &sack->last_sackt);
3169
	}
3170
	sack->rate->rtt_us = ca_rtt_us; /* RTT of last (S)ACKed packet, or -1 */
3171 3172
	rtt_update = tcp_ack_update_rtt(sk, flag, seq_rtt_us, sack_rtt_us,
					ca_rtt_us);
Y
Yuchung Cheng 已提交
3173

3174
	if (flag & FLAG_ACKED) {
3175
		tcp_rearm_rto(sk);
3176 3177 3178 3179 3180
		if (unlikely(icsk->icsk_mtup.probe_size &&
			     !after(tp->mtu_probe.probe_seq_end, tp->snd_una))) {
			tcp_mtup_probe_success(sk);
		}

3181 3182 3183
		if (tcp_is_reno(tp)) {
			tcp_remove_reno_sacks(sk, pkts_acked);
		} else {
3184 3185
			int delta;

3186 3187 3188
			/* Non-retransmitted hole got filled? That's reordering */
			if (reord < prior_fackets)
				tcp_update_reordering(sk, tp->fackets_out - reord, 0);
3189

3190 3191 3192
			delta = tcp_is_fack(tp) ? pkts_acked :
						  prior_sacked - tp->sacked_out;
			tp->lost_cnt_hint -= min(tp->lost_cnt_hint, delta);
3193 3194
		}

I
Ilpo Järvinen 已提交
3195
		tp->fackets_out -= min(pkts_acked, tp->fackets_out);
3196

3197
	} else if (skb && rtt_update && sack_rtt_us >= 0 &&
3198
		   sack_rtt_us > skb_mstamp_us_delta(now, &skb->skb_mstamp)) {
3199 3200 3201 3202 3203
		/* Do not re-arm RTO if the sack RTT is measured from data sent
		 * after when the head was last (re)transmitted. Otherwise the
		 * timeout may continue to extend in loss recovery.
		 */
		tcp_rearm_rto(sk);
L
Linus Torvalds 已提交
3204 3205
	}

3206 3207
	if (icsk->icsk_ca_ops->pkts_acked) {
		struct ack_sample sample = { .pkts_acked = pkts_acked,
3208 3209
					     .rtt_us = ca_rtt_us,
					     .in_flight = last_in_flight };
3210 3211 3212

		icsk->icsk_ca_ops->pkts_acked(sk, &sample);
	}
3213

L
Linus Torvalds 已提交
3214
#if FASTRETRANS_DEBUG > 0
3215 3216 3217
	WARN_ON((int)tp->sacked_out < 0);
	WARN_ON((int)tp->lost_out < 0);
	WARN_ON((int)tp->retrans_out < 0);
3218
	if (!tp->packets_out && tcp_is_sack(tp)) {
S
Stephen Hemminger 已提交
3219
		icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3220
		if (tp->lost_out) {
3221 3222
			pr_debug("Leak l=%u %d\n",
				 tp->lost_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3223 3224 3225
			tp->lost_out = 0;
		}
		if (tp->sacked_out) {
3226 3227
			pr_debug("Leak s=%u %d\n",
				 tp->sacked_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3228 3229 3230
			tp->sacked_out = 0;
		}
		if (tp->retrans_out) {
3231 3232
			pr_debug("Leak r=%u %d\n",
				 tp->retrans_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3233 3234 3235 3236
			tp->retrans_out = 0;
		}
	}
#endif
3237
	*acked = pkts_acked;
3238
	return flag;
L
Linus Torvalds 已提交
3239 3240 3241 3242
}

static void tcp_ack_probe(struct sock *sk)
{
3243 3244
	const struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3245 3246 3247

	/* Was it a usable window open? */

3248
	if (!after(TCP_SKB_CB(tcp_send_head(sk))->end_seq, tcp_wnd_end(tp))) {
3249 3250
		icsk->icsk_backoff = 0;
		inet_csk_clear_xmit_timer(sk, ICSK_TIME_PROBE0);
L
Linus Torvalds 已提交
3251 3252 3253 3254
		/* Socket must be waked up by subsequent tcp_data_snd_check().
		 * This function is not for random using!
		 */
	} else {
3255
		unsigned long when = tcp_probe0_when(sk, TCP_RTO_MAX);
3256

3257
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
3258
					  when, TCP_RTO_MAX);
L
Linus Torvalds 已提交
3259 3260 3261
	}
}

3262
static inline bool tcp_ack_is_dubious(const struct sock *sk, const int flag)
L
Linus Torvalds 已提交
3263
{
E
Eric Dumazet 已提交
3264 3265
	return !(flag & FLAG_NOT_DUP) || (flag & FLAG_CA_ALERT) ||
		inet_csk(sk)->icsk_ca_state != TCP_CA_Open;
L
Linus Torvalds 已提交
3266 3267
}

3268
/* Decide wheather to run the increase function of congestion control. */
3269
static inline bool tcp_may_raise_cwnd(const struct sock *sk, const int flag)
L
Linus Torvalds 已提交
3270
{
3271 3272
	/* If reordering is high then always grow cwnd whenever data is
	 * delivered regardless of its ordering. Otherwise stay conservative
3273
	 * and only grow cwnd on in-order delivery (RFC5681). A stretched ACK w/
3274 3275 3276
	 * new SACK or ECE mark may first advance cwnd here and later reduce
	 * cwnd in tcp_fastretrans_alert() based on more states.
	 */
3277
	if (tcp_sk(sk)->reordering > sock_net(sk)->ipv4.sysctl_tcp_reordering)
3278 3279
		return flag & FLAG_FORWARD_PROGRESS;

3280
	return flag & FLAG_DATA_ACKED;
L
Linus Torvalds 已提交
3281 3282
}

Y
Yuchung Cheng 已提交
3283 3284 3285 3286 3287 3288
/* The "ultimate" congestion control function that aims to replace the rigid
 * cwnd increase and decrease control (tcp_cong_avoid,tcp_*cwnd_reduction).
 * It's called toward the end of processing an ACK with precise rate
 * information. All transmission or retransmission are delayed afterwards.
 */
static void tcp_cong_control(struct sock *sk, u32 ack, u32 acked_sacked,
3289
			     int flag, const struct rate_sample *rs)
Y
Yuchung Cheng 已提交
3290
{
3291 3292 3293 3294 3295 3296 3297
	const struct inet_connection_sock *icsk = inet_csk(sk);

	if (icsk->icsk_ca_ops->cong_control) {
		icsk->icsk_ca_ops->cong_control(sk, rs);
		return;
	}

Y
Yuchung Cheng 已提交
3298 3299 3300 3301 3302 3303 3304 3305 3306 3307
	if (tcp_in_cwnd_reduction(sk)) {
		/* Reduce cwnd if state mandates */
		tcp_cwnd_reduction(sk, acked_sacked, flag);
	} else if (tcp_may_raise_cwnd(sk, flag)) {
		/* Advance cwnd if state allows */
		tcp_cong_avoid(sk, ack, acked_sacked);
	}
	tcp_update_pacing_rate(sk);
}

L
Linus Torvalds 已提交
3308 3309 3310
/* Check that window update is acceptable.
 * The function assumes that snd_una<=ack<=snd_next.
 */
3311
static inline bool tcp_may_update_window(const struct tcp_sock *tp,
3312 3313
					const u32 ack, const u32 ack_seq,
					const u32 nwin)
L
Linus Torvalds 已提交
3314
{
E
Eric Dumazet 已提交
3315
	return	after(ack, tp->snd_una) ||
L
Linus Torvalds 已提交
3316
		after(ack_seq, tp->snd_wl1) ||
E
Eric Dumazet 已提交
3317
		(ack_seq == tp->snd_wl1 && nwin > tp->snd_wnd);
L
Linus Torvalds 已提交
3318 3319
}

3320 3321 3322 3323 3324
/* If we update tp->snd_una, also update tp->bytes_acked */
static void tcp_snd_una_update(struct tcp_sock *tp, u32 ack)
{
	u32 delta = ack - tp->snd_una;

3325
	sock_owned_by_me((struct sock *)tp);
3326 3327 3328 3329
	tp->bytes_acked += delta;
	tp->snd_una = ack;
}

3330 3331 3332 3333 3334
/* If we update tp->rcv_nxt, also update tp->bytes_received */
static void tcp_rcv_nxt_update(struct tcp_sock *tp, u32 seq)
{
	u32 delta = seq - tp->rcv_nxt;

3335
	sock_owned_by_me((struct sock *)tp);
3336 3337 3338 3339
	tp->bytes_received += delta;
	tp->rcv_nxt = seq;
}

L
Linus Torvalds 已提交
3340 3341 3342 3343 3344
/* 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.
 */
3345
static int tcp_ack_update_window(struct sock *sk, const struct sk_buff *skb, u32 ack,
3346
				 u32 ack_seq)
L
Linus Torvalds 已提交
3347
{
3348
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
3349
	int flag = 0;
3350
	u32 nwin = ntohs(tcp_hdr(skb)->window);
L
Linus Torvalds 已提交
3351

3352
	if (likely(!tcp_hdr(skb)->syn))
L
Linus Torvalds 已提交
3353 3354 3355 3356
		nwin <<= tp->rx_opt.snd_wscale;

	if (tcp_may_update_window(tp, ack, ack_seq, nwin)) {
		flag |= FLAG_WIN_UPDATE;
3357
		tcp_update_wl(tp, ack_seq);
L
Linus Torvalds 已提交
3358 3359 3360 3361 3362 3363 3364

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

			/* Note, it is the only place, where
			 * fast path is recovered for sending TCP.
			 */
3365
			tp->pred_flags = 0;
3366
			tcp_fast_path_check(sk);
L
Linus Torvalds 已提交
3367

3368 3369 3370
			if (tcp_send_head(sk))
				tcp_slow_start_after_idle_check(sk);

L
Linus Torvalds 已提交
3371 3372
			if (nwin > tp->max_window) {
				tp->max_window = nwin;
3373
				tcp_sync_mss(sk, inet_csk(sk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
3374 3375 3376 3377
			}
		}
	}

3378
	tcp_snd_una_update(tp, ack);
L
Linus Torvalds 已提交
3379 3380 3381 3382

	return flag;
}

3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399
static bool __tcp_oow_rate_limited(struct net *net, int mib_idx,
				   u32 *last_oow_ack_time)
{
	if (*last_oow_ack_time) {
		s32 elapsed = (s32)(tcp_time_stamp - *last_oow_ack_time);

		if (0 <= elapsed && elapsed < sysctl_tcp_invalid_ratelimit) {
			NET_INC_STATS(net, mib_idx);
			return true;	/* rate-limited: don't send yet! */
		}
	}

	*last_oow_ack_time = tcp_time_stamp;

	return false;	/* not rate-limited: go ahead, send dupack now! */
}

3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412
/* Return true if we're currently rate-limiting out-of-window ACKs and
 * thus shouldn't send a dupack right now. We rate-limit dupacks in
 * response to out-of-window SYNs or ACKs to mitigate ACK loops or DoS
 * attacks that send repeated SYNs or ACKs for the same connection. To
 * do this, we do not send a duplicate SYNACK or ACK if the remote
 * endpoint is sending out-of-window SYNs or pure ACKs at a high rate.
 */
bool tcp_oow_rate_limited(struct net *net, const struct sk_buff *skb,
			  int mib_idx, u32 *last_oow_ack_time)
{
	/* Data packets without SYNs are not likely part of an ACK loop. */
	if ((TCP_SKB_CB(skb)->seq != TCP_SKB_CB(skb)->end_seq) &&
	    !tcp_hdr(skb)->syn)
3413
		return false;
3414

3415
	return __tcp_oow_rate_limited(net, mib_idx, last_oow_ack_time);
3416 3417
}

3418
/* RFC 5961 7 [ACK Throttling] */
3419
static void tcp_send_challenge_ack(struct sock *sk, const struct sk_buff *skb)
3420 3421 3422 3423
{
	/* unprotected vars, we dont care of overwrites */
	static u32 challenge_timestamp;
	static unsigned int challenge_count;
3424
	struct tcp_sock *tp = tcp_sk(sk);
3425
	u32 count, now;
3426 3427

	/* First check our per-socket dupack rate limit. */
3428 3429 3430
	if (__tcp_oow_rate_limited(sock_net(sk),
				   LINUX_MIB_TCPACKSKIPPEDCHALLENGE,
				   &tp->last_oow_ack_time))
3431
		return;
3432

3433
	/* Then check host-wide RFC 5961 rate limit. */
3434
	now = jiffies / HZ;
3435
	if (now != challenge_timestamp) {
3436 3437
		u32 half = (sysctl_tcp_challenge_ack_limit + 1) >> 1;

3438
		challenge_timestamp = now;
3439 3440
		WRITE_ONCE(challenge_count, half +
			   prandom_u32_max(sysctl_tcp_challenge_ack_limit));
3441
	}
3442 3443 3444
	count = READ_ONCE(challenge_count);
	if (count > 0) {
		WRITE_ONCE(challenge_count, count - 1);
3445
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPCHALLENGEACK);
3446 3447 3448 3449
		tcp_send_ack(sk);
	}
}

3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470
static void tcp_store_ts_recent(struct tcp_sock *tp)
{
	tp->rx_opt.ts_recent = tp->rx_opt.rcv_tsval;
	tp->rx_opt.ts_recent_stamp = get_seconds();
}

static 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.
		 */

		if (tcp_paws_check(&tp->rx_opt, 0))
			tcp_store_ts_recent(tp);
	}
}

N
Nandita Dukkipati 已提交
3471
/* This routine deals with acks during a TLP episode.
3472 3473
 * We mark the end of a TLP episode on receiving TLP dupack or when
 * ack is after tlp_high_seq.
N
Nandita Dukkipati 已提交
3474 3475 3476 3477 3478 3479
 * Ref: loss detection algorithm in draft-dukkipati-tcpm-tcp-loss-probe.
 */
static void tcp_process_tlp_ack(struct sock *sk, u32 ack, int flag)
{
	struct tcp_sock *tp = tcp_sk(sk);

3480
	if (before(ack, tp->tlp_high_seq))
N
Nandita Dukkipati 已提交
3481 3482
		return;

3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493
	if (flag & FLAG_DSACKING_ACK) {
		/* This DSACK means original and TLP probe arrived; no loss */
		tp->tlp_high_seq = 0;
	} else if (after(ack, tp->tlp_high_seq)) {
		/* ACK advances: there was a loss, so reduce cwnd. Reset
		 * tlp_high_seq in tcp_init_cwnd_reduction()
		 */
		tcp_init_cwnd_reduction(sk);
		tcp_set_ca_state(sk, TCP_CA_CWR);
		tcp_end_cwnd_reduction(sk);
		tcp_try_keep_open(sk);
3494
		NET_INC_STATS(sock_net(sk),
3495
				LINUX_MIB_TCPLOSSPROBERECOVERY);
3496 3497 3498
	} else if (!(flag & (FLAG_SND_UNA_ADVANCED |
			     FLAG_NOT_DUP | FLAG_DATA_SACKED))) {
		/* Pure dupack: original and TLP probe arrived; no loss */
N
Nandita Dukkipati 已提交
3499 3500 3501 3502
		tp->tlp_high_seq = 0;
	}
}

3503 3504 3505 3506 3507 3508 3509 3510
static inline void tcp_in_ack_event(struct sock *sk, u32 flags)
{
	const struct inet_connection_sock *icsk = inet_csk(sk);

	if (icsk->icsk_ca_ops->in_ack_event)
		icsk->icsk_ca_ops->in_ack_event(sk, flags);
}

3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531
/* Congestion control has updated the cwnd already. So if we're in
 * loss recovery then now we do any new sends (for FRTO) or
 * retransmits (for CA_Loss or CA_recovery) that make sense.
 */
static void tcp_xmit_recovery(struct sock *sk, int rexmit)
{
	struct tcp_sock *tp = tcp_sk(sk);

	if (rexmit == REXMIT_NONE)
		return;

	if (unlikely(rexmit == 2)) {
		__tcp_push_pending_frames(sk, tcp_current_mss(sk),
					  TCP_NAGLE_OFF);
		if (after(tp->snd_nxt, tp->high_seq))
			return;
		tp->frto = 0;
	}
	tcp_xmit_retransmit_queue(sk);
}

L
Linus Torvalds 已提交
3532
/* This routine deals with incoming acks, but not outgoing ones. */
3533
static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
L
Linus Torvalds 已提交
3534
{
3535
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3536
	struct tcp_sock *tp = tcp_sk(sk);
3537
	struct tcp_sacktag_state sack_state;
3538
	struct rate_sample rs = { .prior_delivered = 0 };
L
Linus Torvalds 已提交
3539 3540 3541
	u32 prior_snd_una = tp->snd_una;
	u32 ack_seq = TCP_SKB_CB(skb)->seq;
	u32 ack = TCP_SKB_CB(skb)->ack_seq;
3542
	bool is_dupack = false;
3543
	u32 prior_fackets;
3544
	int prior_packets = tp->packets_out;
3545 3546
	u32 delivered = tp->delivered;
	u32 lost = tp->lost;
3547
	int acked = 0; /* Number of packets newly acked */
3548
	int rexmit = REXMIT_NONE; /* Flag to (re)transmit to recover losses */
3549

3550
	sack_state.first_sackt.v64 = 0;
3551
	sack_state.rate = &rs;
L
Linus Torvalds 已提交
3552

3553 3554 3555
	/* We very likely will need to access write queue head. */
	prefetchw(sk->sk_write_queue.next);

3556
	/* If the ack is older than previous acks
L
Linus Torvalds 已提交
3557 3558
	 * then we can probably ignore it.
	 */
3559 3560 3561
	if (before(ack, prior_snd_una)) {
		/* RFC 5961 5.2 [Blind Data Injection Attack].[Mitigation] */
		if (before(ack, prior_snd_una - tp->max_window)) {
3562
			tcp_send_challenge_ack(sk, skb);
3563 3564
			return -1;
		}
L
Linus Torvalds 已提交
3565
		goto old_ack;
3566
	}
L
Linus Torvalds 已提交
3567

3568 3569 3570 3571 3572 3573
	/* 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;

Y
Yuchung Cheng 已提交
3574
	if (icsk->icsk_pending == ICSK_TIME_LOSS_PROBE)
3575 3576
		tcp_rearm_rto(sk);

3577
	if (after(ack, prior_snd_una)) {
3578
		flag |= FLAG_SND_UNA_ADVANCED;
3579 3580
		icsk->icsk_retransmits = 0;
	}
3581

3582
	prior_fackets = tp->fackets_out;
3583
	rs.prior_in_flight = tcp_packets_in_flight(tp);
3584

3585 3586 3587 3588 3589 3590
	/* ts_recent update must be made after we are sure that the packet
	 * is in window.
	 */
	if (flag & FLAG_UPDATE_TS_RECENT)
		tcp_replace_ts_recent(tp, TCP_SKB_CB(skb)->seq);

3591
	if (!(flag & FLAG_SLOWPATH) && after(ack, prior_snd_una)) {
L
Linus Torvalds 已提交
3592 3593 3594 3595
		/* Window is constant, pure forward advance.
		 * No more checks are required.
		 * Note, we use the fact that SND.UNA>=SND.WL2.
		 */
3596
		tcp_update_wl(tp, ack_seq);
3597
		tcp_snd_una_update(tp, ack);
L
Linus Torvalds 已提交
3598 3599
		flag |= FLAG_WIN_UPDATE;

3600
		tcp_in_ack_event(sk, CA_ACK_WIN_UPDATE);
3601

3602
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPHPACKS);
L
Linus Torvalds 已提交
3603
	} else {
3604 3605
		u32 ack_ev_flags = CA_ACK_SLOWPATH;

L
Linus Torvalds 已提交
3606 3607 3608
		if (ack_seq != TCP_SKB_CB(skb)->end_seq)
			flag |= FLAG_DATA;
		else
3609
			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPPUREACKS);
L
Linus Torvalds 已提交
3610

3611
		flag |= tcp_ack_update_window(sk, skb, ack, ack_seq);
L
Linus Torvalds 已提交
3612 3613

		if (TCP_SKB_CB(skb)->sacked)
Y
Yuchung Cheng 已提交
3614
			flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una,
3615
							&sack_state);
L
Linus Torvalds 已提交
3616

3617
		if (tcp_ecn_rcv_ecn_echo(tp, tcp_hdr(skb))) {
L
Linus Torvalds 已提交
3618
			flag |= FLAG_ECE;
3619 3620 3621 3622 3623
			ack_ev_flags |= CA_ACK_ECE;
		}

		if (flag & FLAG_WIN_UPDATE)
			ack_ev_flags |= CA_ACK_WIN_UPDATE;
L
Linus Torvalds 已提交
3624

3625
		tcp_in_ack_event(sk, ack_ev_flags);
L
Linus Torvalds 已提交
3626 3627 3628 3629 3630 3631
	}

	/* We passed data and got it acked, remove any soft error
	 * log. Something worked...
	 */
	sk->sk_err_soft = 0;
3632
	icsk->icsk_probes_out = 0;
L
Linus Torvalds 已提交
3633 3634 3635 3636 3637
	tp->rcv_tstamp = tcp_time_stamp;
	if (!prior_packets)
		goto no_queue;

	/* See if we can take anything off of the retransmit queue. */
3638
	flag |= tcp_clean_rtx_queue(sk, prior_fackets, prior_snd_una, &acked,
3639
				    &sack_state);
3640

3641
	if (tcp_ack_is_dubious(sk, flag)) {
3642
		is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP));
3643
		tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit);
L
Linus Torvalds 已提交
3644
	}
N
Nandita Dukkipati 已提交
3645 3646 3647
	if (tp->tlp_high_seq)
		tcp_process_tlp_ack(sk, ack, flag);

3648 3649
	if ((flag & FLAG_FORWARD_PROGRESS) || !(flag & FLAG_NOT_DUP))
		sk_dst_confirm(sk);
N
Nandita Dukkipati 已提交
3650 3651 3652

	if (icsk->icsk_pending == ICSK_TIME_RETRANS)
		tcp_schedule_loss_probe(sk);
3653 3654
	delivered = tp->delivered - delivered;	/* freshly ACKed or SACKed */
	lost = tp->lost - lost;			/* freshly marked lost */
3655
	tcp_rate_gen(sk, delivered, lost, sack_state.rate);
3656
	tcp_cong_control(sk, ack, delivered, flag, sack_state.rate);
3657
	tcp_xmit_recovery(sk, rexmit);
L
Linus Torvalds 已提交
3658 3659 3660
	return 1;

no_queue:
3661 3662
	/* If data was DSACKed, see if we can undo a cwnd reduction. */
	if (flag & FLAG_DSACKING_ACK)
3663
		tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit);
L
Linus Torvalds 已提交
3664 3665 3666 3667
	/* 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.
	 */
3668
	if (tcp_send_head(sk))
L
Linus Torvalds 已提交
3669
		tcp_ack_probe(sk);
N
Nandita Dukkipati 已提交
3670 3671 3672

	if (tp->tlp_high_seq)
		tcp_process_tlp_ack(sk, ack, flag);
L
Linus Torvalds 已提交
3673 3674
	return 1;

3675 3676 3677 3678
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
3679
old_ack:
3680 3681 3682
	/* If data was SACKed, tag it and see if we should send more data.
	 * If data was DSACKed, see if we can undo a cwnd reduction.
	 */
3683
	if (TCP_SKB_CB(skb)->sacked) {
Y
Yuchung Cheng 已提交
3684
		flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una,
3685
						&sack_state);
3686
		tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit);
3687
		tcp_xmit_recovery(sk, rexmit);
3688
	}
L
Linus Torvalds 已提交
3689

3690
	SOCK_DEBUG(sk, "Ack %u before %u:%u\n", ack, tp->snd_una, tp->snd_nxt);
L
Linus Torvalds 已提交
3691 3692 3693
	return 0;
}

3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710
static void tcp_parse_fastopen_option(int len, const unsigned char *cookie,
				      bool syn, struct tcp_fastopen_cookie *foc,
				      bool exp_opt)
{
	/* Valid only in SYN or SYN-ACK with an even length.  */
	if (!foc || !syn || len < 0 || (len & 1))
		return;

	if (len >= TCP_FASTOPEN_COOKIE_MIN &&
	    len <= TCP_FASTOPEN_COOKIE_MAX)
		memcpy(foc->val, cookie, len);
	else if (len != 0)
		len = -1;
	foc->len = len;
	foc->exp = exp_opt;
}

L
Linus Torvalds 已提交
3711 3712 3713 3714
/* 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.
 */
C
Christoph Paasch 已提交
3715 3716
void tcp_parse_options(const struct sk_buff *skb,
		       struct tcp_options_received *opt_rx, int estab,
Y
Yuchung Cheng 已提交
3717
		       struct tcp_fastopen_cookie *foc)
L
Linus Torvalds 已提交
3718
{
3719 3720
	const unsigned char *ptr;
	const struct tcphdr *th = tcp_hdr(skb);
3721
	int length = (th->doff * 4) - sizeof(struct tcphdr);
L
Linus Torvalds 已提交
3722

3723
	ptr = (const unsigned char *)(th + 1);
L
Linus Torvalds 已提交
3724 3725
	opt_rx->saw_tstamp = 0;

S
Stephen Hemminger 已提交
3726
	while (length > 0) {
3727
		int opcode = *ptr++;
L
Linus Torvalds 已提交
3728 3729 3730
		int opsize;

		switch (opcode) {
3731 3732 3733 3734 3735 3736 3737 3738
		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
3739
				return;
3740 3741 3742 3743 3744
			if (opsize > length)
				return;	/* don't parse partial options */
			switch (opcode) {
			case TCPOPT_MSS:
				if (opsize == TCPOLEN_MSS && th->syn && !estab) {
3745
					u16 in_mss = get_unaligned_be16(ptr);
3746 3747 3748 3749 3750
					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
3751
					}
3752 3753 3754 3755
				}
				break;
			case TCPOPT_WINDOW:
				if (opsize == TCPOLEN_WINDOW && th->syn &&
3756
				    !estab && sysctl_tcp_window_scaling) {
3757 3758
					__u8 snd_wscale = *(__u8 *)ptr;
					opt_rx->wscale_ok = 1;
3759 3760
					if (snd_wscale > TCP_MAX_WSCALE) {
						net_info_ratelimited("%s: Illegal window scaling value %d > %u received\n",
3761
								     __func__,
3762 3763 3764
								     snd_wscale,
								     TCP_MAX_WSCALE);
						snd_wscale = TCP_MAX_WSCALE;
L
Linus Torvalds 已提交
3765
					}
3766 3767 3768 3769 3770 3771
					opt_rx->snd_wscale = snd_wscale;
				}
				break;
			case TCPOPT_TIMESTAMP:
				if ((opsize == TCPOLEN_TIMESTAMP) &&
				    ((estab && opt_rx->tstamp_ok) ||
3772
				     (!estab && sysctl_tcp_timestamps))) {
3773
					opt_rx->saw_tstamp = 1;
3774 3775
					opt_rx->rcv_tsval = get_unaligned_be32(ptr);
					opt_rx->rcv_tsecr = get_unaligned_be32(ptr + 4);
3776 3777 3778 3779
				}
				break;
			case TCPOPT_SACK_PERM:
				if (opsize == TCPOLEN_SACK_PERM && th->syn &&
3780
				    !estab && sysctl_tcp_sack) {
3781
					opt_rx->sack_ok = TCP_SACK_SEEN;
3782 3783 3784
					tcp_sack_reset(opt_rx);
				}
				break;
L
Linus Torvalds 已提交
3785

3786 3787 3788 3789 3790 3791 3792
			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;
3793
#ifdef CONFIG_TCP_MD5SIG
3794 3795 3796 3797 3798 3799
			case TCPOPT_MD5SIG:
				/*
				 * The MD5 Hash has already been
				 * checked (see tcp_v{4,6}_do_rcv()).
				 */
				break;
3800
#endif
3801 3802 3803 3804 3805 3806
			case TCPOPT_FASTOPEN:
				tcp_parse_fastopen_option(
					opsize - TCPOLEN_FASTOPEN_BASE,
					ptr, th->syn, foc, false);
				break;

Y
Yuchung Cheng 已提交
3807 3808
			case TCPOPT_EXP:
				/* Fast Open option shares code 254 using a
3809
				 * 16 bits magic number.
Y
Yuchung Cheng 已提交
3810
				 */
3811 3812 3813 3814 3815 3816
				if (opsize >= TCPOLEN_EXP_FASTOPEN_BASE &&
				    get_unaligned_be16(ptr) ==
				    TCPOPT_FASTOPEN_MAGIC)
					tcp_parse_fastopen_option(opsize -
						TCPOLEN_EXP_FASTOPEN_BASE,
						ptr + 2, th->syn, foc, true);
Y
Yuchung Cheng 已提交
3817 3818 3819
				break;

			}
3820 3821
			ptr += opsize-2;
			length -= opsize;
3822
		}
L
Linus Torvalds 已提交
3823 3824
	}
}
E
Eric Dumazet 已提交
3825
EXPORT_SYMBOL(tcp_parse_options);
L
Linus Torvalds 已提交
3826

E
Eric Dumazet 已提交
3827
static bool tcp_parse_aligned_timestamp(struct tcp_sock *tp, const struct tcphdr *th)
3828
{
3829
	const __be32 *ptr = (const __be32 *)(th + 1);
3830 3831 3832 3833 3834 3835 3836

	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;
3837 3838 3839 3840
		if (*ptr)
			tp->rx_opt.rcv_tsecr = ntohl(*ptr) - tp->tsoffset;
		else
			tp->rx_opt.rcv_tsecr = 0;
E
Eric Dumazet 已提交
3841
		return true;
3842
	}
E
Eric Dumazet 已提交
3843
	return false;
3844 3845
}

L
Linus Torvalds 已提交
3846 3847 3848
/* Fast parse options. This hopes to only see timestamps.
 * If it is wrong it falls back on tcp_parse_options().
 */
E
Eric Dumazet 已提交
3849
static bool tcp_fast_parse_options(const struct sk_buff *skb,
C
Christoph Paasch 已提交
3850
				   const struct tcphdr *th, struct tcp_sock *tp)
L
Linus Torvalds 已提交
3851
{
3852 3853 3854 3855
	/* 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
3856
		tp->rx_opt.saw_tstamp = 0;
E
Eric Dumazet 已提交
3857
		return false;
L
Linus Torvalds 已提交
3858
	} else if (tp->rx_opt.tstamp_ok &&
3859
		   th->doff == ((sizeof(*th) + TCPOLEN_TSTAMP_ALIGNED) / 4)) {
3860
		if (tcp_parse_aligned_timestamp(tp, th))
E
Eric Dumazet 已提交
3861
			return true;
L
Linus Torvalds 已提交
3862
	}
3863

C
Christoph Paasch 已提交
3864
	tcp_parse_options(skb, &tp->rx_opt, 1, NULL);
3865
	if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr)
3866 3867
		tp->rx_opt.rcv_tsecr -= tp->tsoffset;

E
Eric Dumazet 已提交
3868
	return true;
L
Linus Torvalds 已提交
3869 3870
}

3871 3872 3873 3874
#ifdef CONFIG_TCP_MD5SIG
/*
 * Parse MD5 Signature option
 */
3875
const u8 *tcp_parse_md5sig_option(const struct tcphdr *th)
3876
{
3877 3878
	int length = (th->doff << 2) - sizeof(*th);
	const u8 *ptr = (const u8 *)(th + 1);
3879 3880 3881 3882 3883 3884 3885 3886 3887

	/* 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;

W
Weilong Chen 已提交
3888
		switch (opcode) {
3889 3890 3891 3892 3893 3894 3895 3896 3897 3898
		case TCPOPT_EOL:
			return NULL;
		case TCPOPT_NOP:
			length--;
			continue;
		default:
			opsize = *ptr++;
			if (opsize < 2 || opsize > length)
				return NULL;
			if (opcode == TCPOPT_MD5SIG)
3899
				return opsize == TCPOLEN_MD5SIG ? ptr : NULL;
3900 3901 3902 3903 3904 3905
		}
		ptr += opsize - 2;
		length -= opsize;
	}
	return NULL;
}
E
Eric Dumazet 已提交
3906
EXPORT_SYMBOL(tcp_parse_md5sig_option);
3907 3908
#endif

L
Linus Torvalds 已提交
3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931
/* 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) ]
 */

3932
static int tcp_disordered_ack(const struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
3933
{
3934 3935
	const struct tcp_sock *tp = tcp_sk(sk);
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948
	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. */
3949
		(s32)(tp->rx_opt.ts_recent - tp->rx_opt.rcv_tsval) <= (inet_csk(sk)->icsk_rto * 1024) / HZ);
L
Linus Torvalds 已提交
3950 3951
}

3952
static inline bool tcp_paws_discard(const struct sock *sk,
3953
				   const struct sk_buff *skb)
L
Linus Torvalds 已提交
3954
{
3955
	const struct tcp_sock *tp = tcp_sk(sk);
I
Ilpo Järvinen 已提交
3956 3957 3958

	return !tcp_paws_check(&tp->rx_opt, TCP_PAWS_WINDOW) &&
	       !tcp_disordered_ack(sk, skb);
L
Linus Torvalds 已提交
3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973
}

/* 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)
 */

3974
static inline bool tcp_sequence(const struct tcp_sock *tp, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
3975 3976 3977 3978 3979 3980
{
	return	!before(end_seq, tp->rcv_wup) &&
		!after(seq, tp->rcv_nxt + tcp_receive_window(tp));
}

/* When we get a reset we do this. */
3981
void tcp_reset(struct sock *sk)
L
Linus Torvalds 已提交
3982 3983 3984
{
	/* We want the right error as BSD sees it (and indeed as we do). */
	switch (sk->sk_state) {
3985 3986 3987 3988 3989 3990 3991 3992 3993 3994
	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
3995
	}
T
Tom Marshall 已提交
3996 3997
	/* This barrier is coupled with smp_rmb() in tcp_poll() */
	smp_wmb();
L
Linus Torvalds 已提交
3998

3999 4000
	tcp_done(sk);

L
Linus Torvalds 已提交
4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018
	if (!sock_flag(sk, SOCK_DEAD))
		sk->sk_error_report(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.
 */
4019
void tcp_fin(struct sock *sk)
L
Linus Torvalds 已提交
4020 4021 4022
{
	struct tcp_sock *tp = tcp_sk(sk);

4023
	inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
4024 4025 4026 4027 4028

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

	switch (sk->sk_state) {
4029 4030 4031 4032
	case TCP_SYN_RECV:
	case TCP_ESTABLISHED:
		/* Move to CLOSE_WAIT */
		tcp_set_state(sk, TCP_CLOSE_WAIT);
4033
		inet_csk(sk)->icsk_ack.pingpong = 1;
4034
		break;
L
Linus Torvalds 已提交
4035

4036 4037 4038 4039 4040 4041 4042 4043 4044
	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
4045

4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062
	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.
		 */
J
Joe Perches 已提交
4063
		pr_err("%s: Impossible, sk->sk_state=%d\n",
4064
		       __func__, sk->sk_state);
4065
		break;
4066
	}
L
Linus Torvalds 已提交
4067 4068 4069 4070

	/* It _is_ possible, that we have something out-of-order _after_ FIN.
	 * Probably, we should reset in this case. For now drop them.
	 */
4071
	skb_rbtree_purge(&tp->out_of_order_queue);
4072
	if (tcp_is_sack(tp))
L
Linus Torvalds 已提交
4073
		tcp_sack_reset(&tp->rx_opt);
4074
	sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
4075 4076 4077 4078 4079 4080 4081

	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)
4082
			sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_HUP);
L
Linus Torvalds 已提交
4083
		else
4084
			sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
L
Linus Torvalds 已提交
4085 4086 4087
	}
}

E
Eric Dumazet 已提交
4088
static inline bool tcp_sack_extend(struct tcp_sack_block *sp, u32 seq,
4089
				  u32 end_seq)
L
Linus Torvalds 已提交
4090 4091 4092 4093 4094 4095
{
	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;
E
Eric Dumazet 已提交
4096
		return true;
L
Linus Torvalds 已提交
4097
	}
E
Eric Dumazet 已提交
4098
	return false;
L
Linus Torvalds 已提交
4099 4100
}

4101
static void tcp_dsack_set(struct sock *sk, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
4102
{
4103 4104
	struct tcp_sock *tp = tcp_sk(sk);

4105
	if (tcp_is_sack(tp) && sysctl_tcp_dsack) {
4106 4107
		int mib_idx;

L
Linus Torvalds 已提交
4108
		if (before(seq, tp->rcv_nxt))
4109
			mib_idx = LINUX_MIB_TCPDSACKOLDSENT;
L
Linus Torvalds 已提交
4110
		else
4111 4112
			mib_idx = LINUX_MIB_TCPDSACKOFOSENT;

4113
		NET_INC_STATS(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
4114 4115 4116 4117 4118 4119 4120

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

4121
static void tcp_dsack_extend(struct sock *sk, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
4122
{
4123 4124
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
4125
	if (!tp->rx_opt.dsack)
4126
		tcp_dsack_set(sk, seq, end_seq);
L
Linus Torvalds 已提交
4127 4128 4129 4130
	else
		tcp_sack_extend(tp->duplicate_sack, seq, end_seq);
}

4131
static void tcp_send_dupack(struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
4132 4133 4134 4135 4136
{
	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)) {
4137
		NET_INC_STATS(sock_net(sk), LINUX_MIB_DELAYEDACKLOST);
4138
		tcp_enter_quickack_mode(sk);
L
Linus Torvalds 已提交
4139

4140
		if (tcp_is_sack(tp) && sysctl_tcp_dsack) {
L
Linus Torvalds 已提交
4141 4142 4143 4144
			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;
4145
			tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, end_seq);
L
Linus Torvalds 已提交
4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158
		}
	}

	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];
4159
	struct tcp_sack_block *swalk = sp + 1;
L
Linus Torvalds 已提交
4160 4161 4162 4163

	/* See if the recent change to the first SACK eats into
	 * or hits the sequence space of other SACK blocks, if so coalesce.
	 */
4164
	for (this_sack = 1; this_sack < tp->rx_opt.num_sacks;) {
L
Linus Torvalds 已提交
4165 4166 4167 4168 4169 4170 4171
		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--;
4172 4173
			for (i = this_sack; i < tp->rx_opt.num_sacks; i++)
				sp[i] = sp[i + 1];
L
Linus Torvalds 已提交
4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189
			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;

4190
	for (this_sack = 0; this_sack < cur_sacks; this_sack++, sp++) {
L
Linus Torvalds 已提交
4191 4192
		if (tcp_sack_extend(sp, seq, end_seq)) {
			/* Rotate this_sack to the first one. */
4193
			for (; this_sack > 0; this_sack--, sp--)
4194
				swap(*sp, *(sp - 1));
L
Linus Torvalds 已提交
4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206
			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.
	 */
4207
	if (this_sack >= TCP_NUM_SACKS) {
L
Linus Torvalds 已提交
4208 4209 4210 4211
		this_sack--;
		tp->rx_opt.num_sacks--;
		sp--;
	}
S
Stephen Hemminger 已提交
4212
	for (; this_sack > 0; this_sack--, sp--)
4213
		*sp = *(sp - 1);
L
Linus Torvalds 已提交
4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230

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. */
4231
	if (RB_EMPTY_ROOT(&tp->out_of_order_queue)) {
L
Linus Torvalds 已提交
4232 4233 4234 4235
		tp->rx_opt.num_sacks = 0;
		return;
	}

4236
	for (this_sack = 0; this_sack < num_sacks;) {
L
Linus Torvalds 已提交
4237 4238 4239 4240 4241
		/* 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! */
4242
			WARN_ON(before(tp->rcv_nxt, sp->end_seq));
L
Linus Torvalds 已提交
4243 4244

			/* Zap this SACK, by moving forward any other SACKS. */
W
Weilong Chen 已提交
4245
			for (i = this_sack+1; i < num_sacks; i++)
L
Linus Torvalds 已提交
4246 4247 4248 4249 4250 4251 4252
				tp->selective_acks[i-1] = tp->selective_acks[i];
			num_sacks--;
			continue;
		}
		this_sack++;
		sp++;
	}
4253
	tp->rx_opt.num_sacks = num_sacks;
L
Linus Torvalds 已提交
4254 4255
}

4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286
/**
 * tcp_try_coalesce - try to merge skb to prior one
 * @sk: socket
 * @to: prior buffer
 * @from: buffer to add in queue
 * @fragstolen: pointer to boolean
 *
 * Before queueing skb @from after @to, try to merge them
 * to reduce overall memory use and queue lengths, if cost is small.
 * Packets in ofo or receive queues can stay a long time.
 * Better try to coalesce them right now to avoid future collapses.
 * Returns true if caller should free @from instead of queueing it
 */
static bool tcp_try_coalesce(struct sock *sk,
			     struct sk_buff *to,
			     struct sk_buff *from,
			     bool *fragstolen)
{
	int delta;

	*fragstolen = false;

	/* Its possible this segment overlaps with prior segment in queue */
	if (TCP_SKB_CB(from)->seq != TCP_SKB_CB(to)->end_seq)
		return false;

	if (!skb_try_coalesce(to, from, fragstolen, &delta))
		return false;

	atomic_add(delta, &sk->sk_rmem_alloc);
	sk_mem_charge(sk, delta);
4287
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRCVCOALESCE);
4288 4289 4290 4291 4292 4293
	TCP_SKB_CB(to)->end_seq = TCP_SKB_CB(from)->end_seq;
	TCP_SKB_CB(to)->ack_seq = TCP_SKB_CB(from)->ack_seq;
	TCP_SKB_CB(to)->tcp_flags |= TCP_SKB_CB(from)->tcp_flags;
	return true;
}

4294 4295 4296 4297 4298 4299
static void tcp_drop(struct sock *sk, struct sk_buff *skb)
{
	sk_drops_add(sk, skb);
	__kfree_skb(skb);
}

L
Linus Torvalds 已提交
4300 4301 4302 4303 4304 4305 4306
/* 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;
4307
	bool fin, fragstolen, eaten;
4308
	struct sk_buff *skb, *tail;
4309
	struct rb_node *p;
L
Linus Torvalds 已提交
4310

4311 4312 4313
	p = rb_first(&tp->out_of_order_queue);
	while (p) {
		skb = rb_entry(p, struct sk_buff, rbnode);
L
Linus Torvalds 已提交
4314 4315 4316 4317 4318 4319 4320
		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;
4321
			tcp_dsack_extend(sk, TCP_SKB_CB(skb)->seq, dsack);
L
Linus Torvalds 已提交
4322
		}
4323 4324
		p = rb_next(p);
		rb_erase(&skb->rbnode, &tp->out_of_order_queue);
L
Linus Torvalds 已提交
4325

4326
		if (unlikely(!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt))) {
4327
			SOCK_DEBUG(sk, "ofo packet was already received\n");
4328
			tcp_drop(sk, skb);
L
Linus Torvalds 已提交
4329 4330 4331 4332 4333 4334
			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);

4335 4336
		tail = skb_peek_tail(&sk->sk_receive_queue);
		eaten = tail && tcp_try_coalesce(sk, tail, skb, &fragstolen);
4337
		tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq);
4338
		fin = TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN;
4339 4340
		if (!eaten)
			__skb_queue_tail(&sk->sk_receive_queue, skb);
4341
		else
4342
			kfree_skb_partial(skb, fragstolen);
4343 4344 4345 4346 4347 4348 4349 4350

		if (unlikely(fin)) {
			tcp_fin(sk);
			/* tcp_fin() purges tp->out_of_order_queue,
			 * so we must end this loop right now.
			 */
			break;
		}
L
Linus Torvalds 已提交
4351 4352 4353
	}
}

E
Eric Dumazet 已提交
4354
static bool tcp_prune_ofo_queue(struct sock *sk);
L
Linus Torvalds 已提交
4355 4356
static int tcp_prune_queue(struct sock *sk);

4357 4358
static int tcp_try_rmem_schedule(struct sock *sk, struct sk_buff *skb,
				 unsigned int size)
4359 4360
{
	if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
4361
	    !sk_rmem_schedule(sk, skb, size)) {
4362 4363 4364 4365

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

4366
		while (!sk_rmem_schedule(sk, skb, size)) {
4367 4368
			if (!tcp_prune_ofo_queue(sk))
				return -1;
4369 4370 4371 4372 4373
		}
	}
	return 0;
}

E
Eric Dumazet 已提交
4374 4375 4376
static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);
4377
	struct rb_node **p, *q, *parent;
E
Eric Dumazet 已提交
4378 4379
	struct sk_buff *skb1;
	u32 seq, end_seq;
4380
	bool fragstolen;
E
Eric Dumazet 已提交
4381

4382
	tcp_ecn_check_ce(tp, skb);
E
Eric Dumazet 已提交
4383

4384
	if (unlikely(tcp_try_rmem_schedule(sk, skb, skb->truesize))) {
4385
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFODROP);
4386
		tcp_drop(sk, skb);
E
Eric Dumazet 已提交
4387 4388 4389 4390 4391 4392 4393
		return;
	}

	/* Disable header prediction. */
	tp->pred_flags = 0;
	inet_csk_schedule_ack(sk);

4394
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFOQUEUE);
4395 4396
	seq = TCP_SKB_CB(skb)->seq;
	end_seq = TCP_SKB_CB(skb)->end_seq;
E
Eric Dumazet 已提交
4397
	SOCK_DEBUG(sk, "out of order segment: rcv_next %X seq %X - %X\n",
4398
		   tp->rcv_nxt, seq, end_seq);
E
Eric Dumazet 已提交
4399

4400 4401
	p = &tp->out_of_order_queue.rb_node;
	if (RB_EMPTY_ROOT(&tp->out_of_order_queue)) {
E
Eric Dumazet 已提交
4402 4403 4404
		/* Initial out of order segment, build 1 SACK. */
		if (tcp_is_sack(tp)) {
			tp->rx_opt.num_sacks = 1;
4405 4406
			tp->selective_acks[0].start_seq = seq;
			tp->selective_acks[0].end_seq = end_seq;
E
Eric Dumazet 已提交
4407
		}
4408 4409 4410
		rb_link_node(&skb->rbnode, NULL, p);
		rb_insert_color(&skb->rbnode, &tp->out_of_order_queue);
		tp->ooo_last_skb = skb;
E
Eric Dumazet 已提交
4411 4412 4413
		goto end;
	}

4414 4415 4416 4417 4418 4419 4420 4421 4422 4423
	/* In the typical case, we are adding an skb to the end of the list.
	 * Use of ooo_last_skb avoids the O(Log(N)) rbtree lookup.
	 */
	if (tcp_try_coalesce(sk, tp->ooo_last_skb, skb, &fragstolen)) {
coalesce_done:
		tcp_grow_window(sk, skb);
		kfree_skb_partial(skb, fragstolen);
		skb = NULL;
		goto add_sack;
	}
4424 4425 4426 4427 4428 4429
	/* Can avoid an rbtree lookup if we are adding skb after ooo_last_skb */
	if (!before(seq, TCP_SKB_CB(tp->ooo_last_skb)->end_seq)) {
		parent = &tp->ooo_last_skb->rbnode;
		p = &parent->rb_right;
		goto insert;
	}
4430 4431 4432 4433 4434 4435 4436 4437 4438

	/* Find place to insert this segment. Handle overlaps on the way. */
	parent = NULL;
	while (*p) {
		parent = *p;
		skb1 = rb_entry(parent, struct sk_buff, rbnode);
		if (before(seq, TCP_SKB_CB(skb1)->seq)) {
			p = &parent->rb_left;
			continue;
E
Eric Dumazet 已提交
4439
		}
4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464
		if (before(seq, TCP_SKB_CB(skb1)->end_seq)) {
			if (!after(end_seq, TCP_SKB_CB(skb1)->end_seq)) {
				/* All the bits are present. Drop. */
				NET_INC_STATS(sock_net(sk),
					      LINUX_MIB_TCPOFOMERGE);
				__kfree_skb(skb);
				skb = NULL;
				tcp_dsack_set(sk, seq, end_seq);
				goto add_sack;
			}
			if (after(seq, TCP_SKB_CB(skb1)->seq)) {
				/* Partial overlap. */
				tcp_dsack_set(sk, seq, TCP_SKB_CB(skb1)->end_seq);
			} else {
				/* skb's seq == skb1's seq and skb covers skb1.
				 * Replace skb1 with skb.
				 */
				rb_replace_node(&skb1->rbnode, &skb->rbnode,
						&tp->out_of_order_queue);
				tcp_dsack_extend(sk,
						 TCP_SKB_CB(skb1)->seq,
						 TCP_SKB_CB(skb1)->end_seq);
				NET_INC_STATS(sock_net(sk),
					      LINUX_MIB_TCPOFOMERGE);
				__kfree_skb(skb1);
4465
				goto merge_right;
4466 4467 4468
			}
		} else if (tcp_try_coalesce(sk, skb1, skb, &fragstolen)) {
			goto coalesce_done;
E
Eric Dumazet 已提交
4469
		}
4470
		p = &parent->rb_right;
E
Eric Dumazet 已提交
4471
	}
4472
insert:
4473 4474 4475
	/* Insert segment into RB tree. */
	rb_link_node(&skb->rbnode, parent, p);
	rb_insert_color(&skb->rbnode, &tp->out_of_order_queue);
E
Eric Dumazet 已提交
4476

4477
merge_right:
4478 4479 4480
	/* Remove other segments covered by skb. */
	while ((q = rb_next(&skb->rbnode)) != NULL) {
		skb1 = rb_entry(q, struct sk_buff, rbnode);
E
Eric Dumazet 已提交
4481 4482 4483 4484 4485 4486 4487 4488

		if (!after(end_seq, TCP_SKB_CB(skb1)->seq))
			break;
		if (before(end_seq, TCP_SKB_CB(skb1)->end_seq)) {
			tcp_dsack_extend(sk, TCP_SKB_CB(skb1)->seq,
					 end_seq);
			break;
		}
4489
		rb_erase(&skb1->rbnode, &tp->out_of_order_queue);
E
Eric Dumazet 已提交
4490 4491
		tcp_dsack_extend(sk, TCP_SKB_CB(skb1)->seq,
				 TCP_SKB_CB(skb1)->end_seq);
4492
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFOMERGE);
4493
		tcp_drop(sk, skb1);
E
Eric Dumazet 已提交
4494
	}
4495 4496 4497
	/* If there is no skb after us, we are the last_skb ! */
	if (!q)
		tp->ooo_last_skb = skb;
E
Eric Dumazet 已提交
4498 4499 4500 4501 4502

add_sack:
	if (tcp_is_sack(tp))
		tcp_sack_new_ofo_skb(sk, seq, end_seq);
end:
4503 4504
	if (skb) {
		tcp_grow_window(sk, skb);
4505
		skb_condense(skb);
E
Eric Dumazet 已提交
4506
		skb_set_owner_r(skb, sk);
4507
	}
E
Eric Dumazet 已提交
4508 4509
}

4510
static int __must_check tcp_queue_rcv(struct sock *sk, struct sk_buff *skb, int hdrlen,
4511 4512 4513 4514 4515 4516 4517 4518
		  bool *fragstolen)
{
	int eaten;
	struct sk_buff *tail = skb_peek_tail(&sk->sk_receive_queue);

	__skb_pull(skb, hdrlen);
	eaten = (tail &&
		 tcp_try_coalesce(sk, tail, skb, fragstolen)) ? 1 : 0;
4519
	tcp_rcv_nxt_update(tcp_sk(sk), TCP_SKB_CB(skb)->end_seq);
4520 4521 4522 4523 4524 4525
	if (!eaten) {
		__skb_queue_tail(&sk->sk_receive_queue, skb);
		skb_set_owner_r(skb, sk);
	}
	return eaten;
}
E
Eric Dumazet 已提交
4526

4527 4528
int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size)
{
4529
	struct sk_buff *skb;
4530 4531
	int err = -ENOMEM;
	int data_len = 0;
4532 4533
	bool fragstolen;

4534 4535 4536
	if (size == 0)
		return 0;

4537 4538 4539 4540 4541 4542 4543 4544 4545
	if (size > PAGE_SIZE) {
		int npages = min_t(size_t, size >> PAGE_SHIFT, MAX_SKB_FRAGS);

		data_len = npages << PAGE_SHIFT;
		size = data_len + (size & ~PAGE_MASK);
	}
	skb = alloc_skb_with_frags(size - data_len, data_len,
				   PAGE_ALLOC_COSTLY_ORDER,
				   &err, sk->sk_allocation);
4546 4547 4548
	if (!skb)
		goto err;

4549 4550 4551 4552
	skb_put(skb, size - data_len);
	skb->data_len = data_len;
	skb->len = size;

4553
	if (tcp_try_rmem_schedule(sk, skb, skb->truesize))
4554 4555
		goto err_free;

4556 4557
	err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size);
	if (err)
4558 4559 4560 4561 4562 4563
		goto err_free;

	TCP_SKB_CB(skb)->seq = tcp_sk(sk)->rcv_nxt;
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq + size;
	TCP_SKB_CB(skb)->ack_seq = tcp_sk(sk)->snd_una - 1;

4564
	if (tcp_queue_rcv(sk, skb, 0, &fragstolen)) {
4565 4566 4567 4568 4569 4570 4571 4572
		WARN_ON_ONCE(fragstolen); /* should not happen */
		__kfree_skb(skb);
	}
	return size;

err_free:
	kfree_skb(skb);
err:
4573 4574
	return err;

4575 4576
}

L
Linus Torvalds 已提交
4577 4578 4579
static void tcp_data_queue(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);
4580
	bool fragstolen = false;
4581
	int eaten = -1;
L
Linus Torvalds 已提交
4582

4583 4584 4585 4586
	if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq) {
		__kfree_skb(skb);
		return;
	}
E
Eric Dumazet 已提交
4587
	skb_dst_drop(skb);
4588
	__skb_pull(skb, tcp_hdr(skb)->doff * 4);
L
Linus Torvalds 已提交
4589

4590
	tcp_ecn_accept_cwr(tp, skb);
L
Linus Torvalds 已提交
4591

4592
	tp->rx_opt.dsack = 0;
L
Linus Torvalds 已提交
4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606

	/*  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,
4607
					  tp->ucopy.len);
L
Linus Torvalds 已提交
4608 4609 4610

			__set_current_state(TASK_RUNNING);

4611
			if (!skb_copy_datagram_msg(skb, 0, tp->ucopy.msg, chunk)) {
L
Linus Torvalds 已提交
4612 4613
				tp->ucopy.len -= chunk;
				tp->copied_seq += chunk;
4614
				eaten = (chunk == skb->len);
L
Linus Torvalds 已提交
4615 4616 4617 4618 4619 4620
				tcp_rcv_space_adjust(sk);
			}
		}

		if (eaten <= 0) {
queue_and_out:
4621 4622 4623 4624 4625 4626
			if (eaten < 0) {
				if (skb_queue_len(&sk->sk_receive_queue) == 0)
					sk_forced_mem_schedule(sk, skb->truesize);
				else if (tcp_try_rmem_schedule(sk, skb, skb->truesize))
					goto drop;
			}
4627
			eaten = tcp_queue_rcv(sk, skb, 0, &fragstolen);
L
Linus Torvalds 已提交
4628
		}
4629
		tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq);
S
Stephen Hemminger 已提交
4630
		if (skb->len)
4631
			tcp_event_data_recv(sk, skb);
4632
		if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)
4633
			tcp_fin(sk);
L
Linus Torvalds 已提交
4634

4635
		if (!RB_EMPTY_ROOT(&tp->out_of_order_queue)) {
L
Linus Torvalds 已提交
4636 4637 4638 4639 4640
			tcp_ofo_queue(sk);

			/* RFC2581. 4.2. SHOULD send immediate ACK, when
			 * gap in queue is filled.
			 */
4641
			if (RB_EMPTY_ROOT(&tp->out_of_order_queue))
4642
				inet_csk(sk)->icsk_ack.pingpong = 0;
L
Linus Torvalds 已提交
4643 4644 4645 4646 4647
		}

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

4648
		tcp_fast_path_check(sk);
L
Linus Torvalds 已提交
4649

4650 4651
		if (eaten > 0)
			kfree_skb_partial(skb, fragstolen);
4652
		if (!sock_flag(sk, SOCK_DEAD))
4653
			sk->sk_data_ready(sk);
L
Linus Torvalds 已提交
4654 4655 4656 4657 4658
		return;
	}

	if (!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) {
		/* A retransmit, 2nd most common case.  Force an immediate ack. */
4659
		NET_INC_STATS(sock_net(sk), LINUX_MIB_DELAYEDACKLOST);
4660
		tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq);
L
Linus Torvalds 已提交
4661 4662

out_of_window:
4663 4664
		tcp_enter_quickack_mode(sk);
		inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
4665
drop:
4666
		tcp_drop(sk, skb);
L
Linus Torvalds 已提交
4667 4668 4669 4670 4671 4672 4673
		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;

4674
	tcp_enter_quickack_mode(sk);
L
Linus Torvalds 已提交
4675 4676 4677 4678 4679 4680 4681

	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);

4682
		tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, tp->rcv_nxt);
4683

L
Linus Torvalds 已提交
4684 4685 4686 4687 4688 4689 4690 4691
		/* 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;
	}

E
Eric Dumazet 已提交
4692
	tcp_data_queue_ofo(sk, skb);
L
Linus Torvalds 已提交
4693 4694
}

4695 4696 4697 4698 4699 4700 4701 4702
static struct sk_buff *tcp_skb_next(struct sk_buff *skb, struct sk_buff_head *list)
{
	if (list)
		return !skb_queue_is_last(list, skb) ? skb->next : NULL;

	return rb_entry_safe(rb_next(&skb->rbnode), struct sk_buff, rbnode);
}

4703
static struct sk_buff *tcp_collapse_one(struct sock *sk, struct sk_buff *skb,
4704 4705
					struct sk_buff_head *list,
					struct rb_root *root)
4706
{
4707
	struct sk_buff *next = tcp_skb_next(skb, list);
4708

4709 4710 4711 4712
	if (list)
		__skb_unlink(skb, list);
	else
		rb_erase(&skb->rbnode, root);
4713 4714

	__kfree_skb(skb);
4715
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRCVCOLLAPSED);
4716 4717 4718 4719

	return next;
}

4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738
/* Insert skb into rb tree, ordered by TCP_SKB_CB(skb)->seq */
static void tcp_rbtree_insert(struct rb_root *root, struct sk_buff *skb)
{
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent = NULL;
	struct sk_buff *skb1;

	while (*p) {
		parent = *p;
		skb1 = rb_entry(parent, struct sk_buff, rbnode);
		if (before(TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb1)->seq))
			p = &parent->rb_left;
		else
			p = &parent->rb_right;
	}
	rb_link_node(&skb->rbnode, parent, p);
	rb_insert_color(&skb->rbnode, root);
}

L
Linus Torvalds 已提交
4739 4740
/* Collapse contiguous sequence of skbs head..tail with
 * sequence numbers start..end.
4741
 *
4742
 * If tail is NULL, this means until the end of the queue.
4743
 *
L
Linus Torvalds 已提交
4744 4745 4746 4747
 * Segments with FIN/SYN are not collapsed (only because this
 * simplifies code)
 */
static void
4748 4749
tcp_collapse(struct sock *sk, struct sk_buff_head *list, struct rb_root *root,
	     struct sk_buff *head, struct sk_buff *tail, u32 start, u32 end)
L
Linus Torvalds 已提交
4750
{
4751 4752
	struct sk_buff *skb = head, *n;
	struct sk_buff_head tmp;
4753
	bool end_of_skbs;
L
Linus Torvalds 已提交
4754

S
Stephen Hemminger 已提交
4755
	/* First, check that queue is collapsible and find
4756 4757
	 * the point where collapsing can be useful.
	 */
4758
restart:
4759 4760 4761
	for (end_of_skbs = true; skb != NULL && skb != tail; skb = n) {
		n = tcp_skb_next(skb, list);

L
Linus Torvalds 已提交
4762 4763
		/* No new bits? It is possible on ofo queue. */
		if (!before(start, TCP_SKB_CB(skb)->end_seq)) {
4764
			skb = tcp_collapse_one(sk, skb, list, root);
4765 4766 4767
			if (!skb)
				break;
			goto restart;
L
Linus Torvalds 已提交
4768 4769 4770 4771 4772 4773 4774
		}

		/* The first skb to collapse is:
		 * - not SYN/FIN and
		 * - bloated or contains data before "start" or
		 *   overlaps to the next one.
		 */
4775
		if (!(TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)) &&
L
Linus Torvalds 已提交
4776
		    (tcp_win_from_space(skb->truesize) > skb->len ||
4777 4778
		     before(TCP_SKB_CB(skb)->seq, start))) {
			end_of_skbs = false;
L
Linus Torvalds 已提交
4779
			break;
4780 4781
		}

4782 4783 4784 4785
		if (n && n != tail &&
		    TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(n)->seq) {
			end_of_skbs = false;
			break;
4786
		}
L
Linus Torvalds 已提交
4787 4788 4789 4790

		/* Decided to skip this, advance start seq. */
		start = TCP_SKB_CB(skb)->end_seq;
	}
4791 4792
	if (end_of_skbs ||
	    (TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)))
L
Linus Torvalds 已提交
4793 4794
		return;

4795 4796
	__skb_queue_head_init(&tmp);

L
Linus Torvalds 已提交
4797
	while (before(start, end)) {
4798
		int copy = min_t(int, SKB_MAX_ORDER(0, 0), end - start);
L
Linus Torvalds 已提交
4799 4800
		struct sk_buff *nskb;

4801
		nskb = alloc_skb(copy, GFP_ATOMIC);
L
Linus Torvalds 已提交
4802
		if (!nskb)
4803
			break;
4804

L
Linus Torvalds 已提交
4805 4806
		memcpy(nskb->cb, skb->cb, sizeof(skb->cb));
		TCP_SKB_CB(nskb)->seq = TCP_SKB_CB(nskb)->end_seq = start;
4807 4808 4809 4810
		if (list)
			__skb_queue_before(list, skb, nskb);
		else
			__skb_queue_tail(&tmp, nskb); /* defer rbtree insertion */
4811
		skb_set_owner_r(nskb, sk);
L
Linus Torvalds 已提交
4812 4813 4814 4815 4816 4817

		/* 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;

4818
			BUG_ON(offset < 0);
L
Linus Torvalds 已提交
4819 4820 4821 4822 4823 4824 4825 4826 4827
			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)) {
4828
				skb = tcp_collapse_one(sk, skb, list, root);
4829 4830
				if (!skb ||
				    skb == tail ||
4831
				    (TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)))
4832
					goto end;
L
Linus Torvalds 已提交
4833 4834 4835
			}
		}
	}
4836 4837 4838
end:
	skb_queue_walk_safe(&tmp, skb, n)
		tcp_rbtree_insert(root, skb);
L
Linus Torvalds 已提交
4839 4840 4841 4842 4843 4844 4845 4846
}

/* 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);
4847 4848
	struct sk_buff *skb, *head;
	struct rb_node *p;
L
Linus Torvalds 已提交
4849 4850
	u32 start, end;

4851 4852 4853 4854 4855 4856 4857 4858 4859 4860
	p = rb_first(&tp->out_of_order_queue);
	skb = rb_entry_safe(p, struct sk_buff, rbnode);
new_range:
	if (!skb) {
		p = rb_last(&tp->out_of_order_queue);
		/* Note: This is possible p is NULL here. We do not
		 * use rb_entry_safe(), as ooo_last_skb is valid only
		 * if rbtree is not empty.
		 */
		tp->ooo_last_skb = rb_entry(p, struct sk_buff, rbnode);
L
Linus Torvalds 已提交
4861
		return;
4862
	}
L
Linus Torvalds 已提交
4863 4864
	start = TCP_SKB_CB(skb)->seq;
	end = TCP_SKB_CB(skb)->end_seq;
4865

4866 4867
	for (head = skb;;) {
		skb = tcp_skb_next(skb, NULL);
L
Linus Torvalds 已提交
4868

4869 4870 4871
		/* Range is terminated when we see a gap or when
		 * we are at the queue end.
		 */
4872
		if (!skb ||
L
Linus Torvalds 已提交
4873 4874
		    after(TCP_SKB_CB(skb)->seq, end) ||
		    before(TCP_SKB_CB(skb)->end_seq, start)) {
4875
			tcp_collapse(sk, NULL, &tp->out_of_order_queue,
D
David S. Miller 已提交
4876
				     head, skb, start, end);
4877 4878 4879 4880
			goto new_range;
		}

		if (unlikely(before(TCP_SKB_CB(skb)->seq, start)))
L
Linus Torvalds 已提交
4881
			start = TCP_SKB_CB(skb)->seq;
4882
		if (after(TCP_SKB_CB(skb)->end_seq, end))
L
Linus Torvalds 已提交
4883 4884 4885 4886
			end = TCP_SKB_CB(skb)->end_seq;
	}
}

4887
/*
4888 4889 4890 4891 4892 4893 4894 4895
 * Clean the out-of-order queue to make room.
 * We drop high sequences packets to :
 * 1) Let a chance for holes to be filled.
 * 2) not add too big latencies if thousands of packets sit there.
 *    (But if application shrinks SO_RCVBUF, we could still end up
 *     freeing whole queue here)
 *
 * Return true if queue has shrunk.
4896
 */
E
Eric Dumazet 已提交
4897
static bool tcp_prune_ofo_queue(struct sock *sk)
4898 4899
{
	struct tcp_sock *tp = tcp_sk(sk);
4900
	struct rb_node *node, *prev;
4901

4902
	if (RB_EMPTY_ROOT(&tp->out_of_order_queue))
4903
		return false;
4904

4905
	NET_INC_STATS(sock_net(sk), LINUX_MIB_OFOPRUNED);
4906 4907 4908 4909 4910
	node = &tp->ooo_last_skb->rbnode;
	do {
		prev = rb_prev(node);
		rb_erase(node, &tp->out_of_order_queue);
		tcp_drop(sk, rb_entry(node, struct sk_buff, rbnode));
4911
		sk_mem_reclaim(sk);
4912 4913 4914
		if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
		    !tcp_under_memory_pressure(sk))
			break;
4915 4916 4917
		node = prev;
	} while (node);
	tp->ooo_last_skb = rb_entry(prev, struct sk_buff, rbnode);
4918 4919 4920 4921 4922 4923 4924 4925 4926

	/* 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);
	return true;
4927 4928
}

L
Linus Torvalds 已提交
4929 4930 4931 4932 4933 4934 4935 4936 4937
/* 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)
{
4938
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
4939 4940 4941

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

4942
	NET_INC_STATS(sock_net(sk), LINUX_MIB_PRUNECALLED);
L
Linus Torvalds 已提交
4943 4944

	if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
4945
		tcp_clamp_window(sk);
4946
	else if (tcp_under_memory_pressure(sk))
L
Linus Torvalds 已提交
4947 4948 4949
		tp->rcv_ssthresh = min(tp->rcv_ssthresh, 4U * tp->advmss);

	tcp_collapse_ofo_queue(sk);
4950
	if (!skb_queue_empty(&sk->sk_receive_queue))
4951
		tcp_collapse(sk, &sk->sk_receive_queue, NULL,
4952 4953 4954
			     skb_peek(&sk->sk_receive_queue),
			     NULL,
			     tp->copied_seq, tp->rcv_nxt);
4955
	sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
4956 4957 4958 4959 4960 4961 4962

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

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

4963
	tcp_prune_ofo_queue(sk);
L
Linus Torvalds 已提交
4964 4965 4966 4967 4968 4969 4970 4971

	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.
	 */
4972
	NET_INC_STATS(sock_net(sk), LINUX_MIB_RCVPRUNED);
L
Linus Torvalds 已提交
4973 4974 4975 4976 4977 4978

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

E
Eric Dumazet 已提交
4979
static bool tcp_should_expand_sndbuf(const struct sock *sk)
4980
{
4981
	const struct tcp_sock *tp = tcp_sk(sk);
4982

4983 4984 4985 4986
	/* If the user specified a specific send buffer setting, do
	 * not modify it.
	 */
	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
E
Eric Dumazet 已提交
4987
		return false;
4988 4989

	/* If we are under global TCP memory pressure, do not expand.  */
4990
	if (tcp_under_memory_pressure(sk))
E
Eric Dumazet 已提交
4991
		return false;
4992 4993

	/* If we are under soft global TCP memory pressure, do not expand.  */
4994
	if (sk_memory_allocated(sk) >= sk_prot_mem_limits(sk, 0))
E
Eric Dumazet 已提交
4995
		return false;
4996 4997

	/* If we filled the congestion window, do not expand.  */
4998
	if (tcp_packets_in_flight(tp) >= tp->snd_cwnd)
E
Eric Dumazet 已提交
4999
		return false;
5000

E
Eric Dumazet 已提交
5001
	return true;
5002
}
L
Linus Torvalds 已提交
5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013

/* 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);

5014
	if (tcp_should_expand_sndbuf(sk)) {
E
Eric Dumazet 已提交
5015
		tcp_sndbuf_expand(sk);
L
Linus Torvalds 已提交
5016 5017 5018 5019 5020 5021
		tp->snd_cwnd_stamp = tcp_time_stamp;
	}

	sk->sk_write_space(sk);
}

S
Stephen Hemminger 已提交
5022
static void tcp_check_space(struct sock *sk)
L
Linus Torvalds 已提交
5023 5024 5025
{
	if (sock_flag(sk, SOCK_QUEUE_SHRUNK)) {
		sock_reset_flag(sk, SOCK_QUEUE_SHRUNK);
5026
		/* pairs with tcp_poll() */
5027
		smp_mb();
L
Linus Torvalds 已提交
5028
		if (sk->sk_socket &&
5029
		    test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
L
Linus Torvalds 已提交
5030
			tcp_new_space(sk);
5031 5032 5033
			if (!test_bit(SOCK_NOSPACE, &sk->sk_socket->flags))
				tcp_chrono_stop(sk, TCP_CHRONO_SNDBUF_LIMITED);
		}
L
Linus Torvalds 已提交
5034 5035 5036
	}
}

5037
static inline void tcp_data_snd_check(struct sock *sk)
L
Linus Torvalds 已提交
5038
{
5039
	tcp_push_pending_frames(sk);
L
Linus Torvalds 已提交
5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050
	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... */
5051
	if (((tp->rcv_nxt - tp->rcv_wup) > inet_csk(sk)->icsk_ack.rcv_mss &&
L
Linus Torvalds 已提交
5052 5053 5054
	     /* ... and right edge of window advances far enough.
	      * (tcp_recvmsg() will send ACK otherwise). Or...
	      */
5055
	     __tcp_select_window(sk) >= tp->rcv_wnd) ||
L
Linus Torvalds 已提交
5056
	    /* We ACK each frame or... */
5057
	    tcp_in_quickack_mode(sk) ||
L
Linus Torvalds 已提交
5058
	    /* We have out of order data. */
5059
	    (ofo_possible && !RB_EMPTY_ROOT(&tp->out_of_order_queue))) {
L
Linus Torvalds 已提交
5060 5061 5062 5063 5064 5065 5066 5067
		/* Then ack it now */
		tcp_send_ack(sk);
	} else {
		/* Else, send delayed ack. */
		tcp_send_delayed_ack(sk);
	}
}

S
Stephen Hemminger 已提交
5068
static inline void tcp_ack_snd_check(struct sock *sk)
L
Linus Torvalds 已提交
5069
{
5070
	if (!inet_csk_ack_scheduled(sk)) {
L
Linus Torvalds 已提交
5071 5072 5073 5074 5075 5076 5077 5078
		/* 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
5079
 *	signaled. Its the 'slow' part of tcp_urg. It could be
L
Linus Torvalds 已提交
5080 5081 5082 5083 5084 5085
 *	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).
 */
5086

5087
static void tcp_check_urg(struct sock *sk, const struct tcphdr *th)
L
Linus Torvalds 已提交
5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123
{
	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
5124
	 * or we break the semantics of SIOCATMARK (and thus sockatmark())
L
Linus Torvalds 已提交
5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135
	 *
	 * 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 &&
5136
	    !sock_flag(sk, SOCK_URGINLINE) && tp->copied_seq != tp->rcv_nxt) {
L
Linus Torvalds 已提交
5137 5138 5139
		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
5140
			__skb_unlink(skb, &sk->sk_receive_queue);
L
Linus Torvalds 已提交
5141 5142 5143 5144
			__kfree_skb(skb);
		}
	}

5145 5146
	tp->urg_data = TCP_URG_NOTYET;
	tp->urg_seq = ptr;
L
Linus Torvalds 已提交
5147 5148 5149 5150 5151 5152

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

/* This is the 'fast' part of urgent handling. */
5153
static void tcp_urg(struct sock *sk, struct sk_buff *skb, const struct tcphdr *th)
L
Linus Torvalds 已提交
5154 5155 5156 5157 5158
{
	struct tcp_sock *tp = tcp_sk(sk);

	/* Check if we get a new urgent pointer - normally not. */
	if (th->urg)
5159
		tcp_check_urg(sk, th);
L
Linus Torvalds 已提交
5160 5161 5162 5163 5164 5165

	/* 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;

5166
		/* Is the urgent pointer pointing into this packet? */
L
Linus Torvalds 已提交
5167 5168 5169 5170 5171 5172
		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))
5173
				sk->sk_data_ready(sk);
L
Linus Torvalds 已提交
5174 5175 5176 5177 5178 5179 5180 5181 5182 5183
		}
	}
}

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;

5184
	if (skb_csum_unnecessary(skb))
5185
		err = skb_copy_datagram_msg(skb, hlen, tp->ucopy.msg, chunk);
L
Linus Torvalds 已提交
5186
	else
5187
		err = skb_copy_and_csum_datagram_msg(skb, hlen, tp->ucopy.msg);
L
Linus Torvalds 已提交
5188 5189 5190 5191 5192 5193 5194 5195 5196 5197

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

	return err;
}

5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214
/* Accept RST for rcv_nxt - 1 after a FIN.
 * When tcp connections are abruptly terminated from Mac OSX (via ^C), a
 * FIN is sent followed by a RST packet. The RST is sent with the same
 * sequence number as the FIN, and thus according to RFC 5961 a challenge
 * ACK should be sent. However, Mac OSX rate limits replies to challenge
 * ACKs on the closed socket. In addition middleboxes can drop either the
 * challenge ACK or a subsequent RST.
 */
static bool tcp_reset_check(const struct sock *sk, const struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);

	return unlikely(TCP_SKB_CB(skb)->seq == (tp->rcv_nxt - 1) &&
			(1 << sk->sk_state) & (TCPF_CLOSE_WAIT | TCPF_LAST_ACK |
					       TCPF_CLOSING));
}

5215 5216 5217
/* Does PAWS and seqno based validation of an incoming segment, flags will
 * play significant role here.
 */
E
Eric Dumazet 已提交
5218 5219
static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
				  const struct tcphdr *th, int syn_inerr)
5220 5221
{
	struct tcp_sock *tp = tcp_sk(sk);
5222
	bool rst_seq_match = false;
5223 5224

	/* RFC1323: H1. Apply PAWS check first. */
C
Christoph Paasch 已提交
5225
	if (tcp_fast_parse_options(skb, th, tp) && tp->rx_opt.saw_tstamp &&
5226 5227
	    tcp_paws_discard(sk, skb)) {
		if (!th->rst) {
5228
			NET_INC_STATS(sock_net(sk), LINUX_MIB_PAWSESTABREJECTED);
5229 5230 5231 5232
			if (!tcp_oow_rate_limited(sock_net(sk), skb,
						  LINUX_MIB_TCPACKSKIPPEDPAWS,
						  &tp->last_oow_ack_time))
				tcp_send_dupack(sk, skb);
5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245
			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)".
		 */
5246 5247 5248
		if (!th->rst) {
			if (th->syn)
				goto syn_challenge;
5249 5250 5251 5252
			if (!tcp_oow_rate_limited(sock_net(sk), skb,
						  LINUX_MIB_TCPACKSKIPPEDSEQ,
						  &tp->last_oow_ack_time))
				tcp_send_dupack(sk, skb);
5253 5254
		} else if (tcp_reset_check(sk, skb)) {
			tcp_reset(sk);
5255
		}
5256 5257 5258 5259 5260
		goto discard;
	}

	/* Step 2: check RST bit */
	if (th->rst) {
5261 5262 5263 5264
		/* RFC 5961 3.2 (extend to match against (RCV.NXT - 1) after a
		 * FIN and SACK too if available):
		 * If seq num matches RCV.NXT or (RCV.NXT - 1) after a FIN, or
		 * the right-most SACK block,
5265
		 * then
E
Eric Dumazet 已提交
5266 5267 5268 5269
		 *     RESET the connection
		 * else
		 *     Send a challenge ACK
		 */
5270 5271
		if (TCP_SKB_CB(skb)->seq == tp->rcv_nxt ||
		    tcp_reset_check(sk, skb)) {
5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289
			rst_seq_match = true;
		} else if (tcp_is_sack(tp) && tp->rx_opt.num_sacks > 0) {
			struct tcp_sack_block *sp = &tp->selective_acks[0];
			int max_sack = sp[0].end_seq;
			int this_sack;

			for (this_sack = 1; this_sack < tp->rx_opt.num_sacks;
			     ++this_sack) {
				max_sack = after(sp[this_sack].end_seq,
						 max_sack) ?
					sp[this_sack].end_seq : max_sack;
			}

			if (TCP_SKB_CB(skb)->seq == max_sack)
				rst_seq_match = true;
		}

		if (rst_seq_match)
E
Eric Dumazet 已提交
5290
			tcp_reset(sk);
5291 5292 5293 5294 5295 5296 5297
		else {
			/* Disable TFO if RST is out-of-order
			 * and no data has been received
			 * for current active TFO socket
			 */
			if (tp->syn_fastopen && !tp->data_segs_in &&
			    sk->sk_state == TCP_ESTABLISHED)
5298
				tcp_fastopen_active_disable(sk);
5299
			tcp_send_challenge_ack(sk, skb);
5300
		}
5301 5302 5303 5304 5305
		goto discard;
	}

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

E
Eric Dumazet 已提交
5306
	/* step 4: Check for a SYN
5307
	 * RFC 5961 4.2 : Send a challenge ack
E
Eric Dumazet 已提交
5308 5309
	 */
	if (th->syn) {
5310
syn_challenge:
5311
		if (syn_inerr)
5312 5313
			TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPSYNCHALLENGE);
5314
		tcp_send_challenge_ack(sk, skb);
E
Eric Dumazet 已提交
5315
		goto discard;
5316 5317
	}

E
Eric Dumazet 已提交
5318
	return true;
5319 5320

discard:
5321
	tcp_drop(sk, skb);
E
Eric Dumazet 已提交
5322
	return false;
5323 5324
}

L
Linus Torvalds 已提交
5325
/*
5326
 *	TCP receive function for the ESTABLISHED state.
L
Linus Torvalds 已提交
5327
 *
5328
 *	It is split into a fast path and a slow path. The fast path is
L
Linus Torvalds 已提交
5329 5330
 * 	disabled when:
 *	- A zero window was announced from us - zero window probing
5331
 *        is only handled properly in the slow path.
L
Linus Torvalds 已提交
5332 5333 5334 5335
 *	- Out of order segments arrived.
 *	- Urgent data is expected.
 *	- There is no buffer space left
 *	- Unexpected TCP flags/window values/header lengths are received
5336
 *	  (detected by checking the TCP header against pred_flags)
L
Linus Torvalds 已提交
5337 5338 5339 5340 5341
 *	- 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.
 *
5342
 *	When these conditions are not satisfied it drops into a standard
L
Linus Torvalds 已提交
5343 5344
 *	receive procedure patterned after RFC793 to handle all cases.
 *	The first three cases are guaranteed by proper pred_flags setting,
5345
 *	the rest is checked inline. Fast processing is turned on in
L
Linus Torvalds 已提交
5346 5347
 *	tcp_data_queue when everything is OK.
 */
5348 5349
void tcp_rcv_established(struct sock *sk, struct sk_buff *skb,
			 const struct tcphdr *th, unsigned int len)
L
Linus Torvalds 已提交
5350 5351 5352
{
	struct tcp_sock *tp = tcp_sk(sk);

E
Eric Dumazet 已提交
5353
	skb_mstamp_get(&tp->tcp_mstamp);
5354
	if (unlikely(!sk->sk_rx_dst))
E
Eric Dumazet 已提交
5355
		inet_csk(sk)->icsk_af_ops->sk_rx_dst_set(sk, skb);
L
Linus Torvalds 已提交
5356 5357
	/*
	 *	Header prediction.
5358
	 *	The code loosely follows the one in the famous
L
Linus Torvalds 已提交
5359
	 *	"30 instruction TCP receive" Van Jacobson mail.
5360 5361
	 *
	 *	Van's trick is to deposit buffers into socket queue
L
Linus Torvalds 已提交
5362 5363 5364 5365
	 *	on a device interrupt, to call tcp_recv function
	 *	on the receive process context and checksum and copy
	 *	the buffer to user space. smart...
	 *
5366
	 *	Our current scheme is not silly either but we take the
L
Linus Torvalds 已提交
5367 5368 5369 5370 5371 5372 5373
	 *	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
5374
	 *	if header_prediction is to be made
L
Linus Torvalds 已提交
5375 5376
	 *	'S' will always be tp->tcp_header_len >> 2
	 *	'?' will be 0 for the fast path, otherwise pred_flags is 0 to
5377
	 *  turn it off	(when there are holes in the receive
L
Linus Torvalds 已提交
5378 5379 5380 5381 5382
	 *	 space for instance)
	 *	PSH flag is ignored.
	 */

	if ((tcp_flag_word(th) & TCP_HP_BITS) == tp->pred_flags &&
5383 5384
	    TCP_SKB_CB(skb)->seq == tp->rcv_nxt &&
	    !after(TCP_SKB_CB(skb)->ack_seq, tp->snd_nxt)) {
L
Linus Torvalds 已提交
5385 5386 5387 5388 5389 5390 5391 5392 5393 5394
		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! */
5395
			if (!tcp_parse_aligned_timestamp(tp, th))
L
Linus Torvalds 已提交
5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424
				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);
5425
				__kfree_skb(skb);
5426
				tcp_data_snd_check(sk);
5427
				return;
L
Linus Torvalds 已提交
5428
			} else { /* Header too small */
5429
				TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
5430 5431 5432 5433
				goto discard;
			}
		} else {
			int eaten = 0;
5434
			bool fragstolen = false;
L
Linus Torvalds 已提交
5435

D
Dan Williams 已提交
5436 5437 5438 5439 5440
			if (tp->ucopy.task == current &&
			    tp->copied_seq == tp->rcv_nxt &&
			    len - tcp_header_len <= tp->ucopy.len &&
			    sock_owned_by_user(sk)) {
				__set_current_state(TASK_RUNNING);
L
Linus Torvalds 已提交
5441

D
Dan Williams 已提交
5442
				if (!tcp_copy_to_iovec(sk, skb, tcp_header_len)) {
L
Linus Torvalds 已提交
5443 5444 5445 5446 5447 5448 5449 5450 5451 5452
					/* 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);

5453
					tcp_rcv_rtt_measure_ts(sk, skb);
L
Linus Torvalds 已提交
5454 5455

					__skb_pull(skb, tcp_header_len);
5456
					tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq);
5457
					NET_INC_STATS(sock_net(sk),
5458
							LINUX_MIB_TCPHPHITSTOUSER);
D
Dan Williams 已提交
5459
					eaten = 1;
L
Linus Torvalds 已提交
5460 5461 5462
				}
			}
			if (!eaten) {
5463
				if (tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
5464 5465
					goto csum_error;

5466 5467 5468
				if ((int)skb->truesize > sk->sk_forward_alloc)
					goto step5;

L
Linus Torvalds 已提交
5469 5470 5471 5472 5473 5474 5475 5476 5477
				/* 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);

5478
				tcp_rcv_rtt_measure_ts(sk, skb);
L
Linus Torvalds 已提交
5479

5480
				NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPHPHITS);
L
Linus Torvalds 已提交
5481 5482

				/* Bulk data transfer: receiver */
5483 5484
				eaten = tcp_queue_rcv(sk, skb, tcp_header_len,
						      &fragstolen);
L
Linus Torvalds 已提交
5485 5486
			}

5487
			tcp_event_data_recv(sk, skb);
L
Linus Torvalds 已提交
5488 5489 5490 5491

			if (TCP_SKB_CB(skb)->ack_seq != tp->snd_una) {
				/* Well, only one small jumplet in fast path... */
				tcp_ack(sk, skb, FLAG_DATA);
5492
				tcp_data_snd_check(sk);
5493
				if (!inet_csk_ack_scheduled(sk))
L
Linus Torvalds 已提交
5494 5495 5496
					goto no_ack;
			}

D
Dan Williams 已提交
5497
			__tcp_ack_snd_check(sk, 0);
L
Linus Torvalds 已提交
5498 5499
no_ack:
			if (eaten)
5500
				kfree_skb_partial(skb, fragstolen);
5501
			sk->sk_data_ready(sk);
5502
			return;
L
Linus Torvalds 已提交
5503 5504 5505 5506
		}
	}

slow_path:
5507
	if (len < (th->doff << 2) || tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
5508 5509
		goto csum_error;

5510
	if (!th->ack && !th->rst && !th->syn)
5511 5512
		goto discard;

L
Linus Torvalds 已提交
5513 5514 5515 5516
	/*
	 *	Standard slow path.
	 */

E
Eric Dumazet 已提交
5517
	if (!tcp_validate_incoming(sk, skb, th, 1))
5518
		return;
L
Linus Torvalds 已提交
5519 5520

step5:
5521
	if (tcp_ack(sk, skb, FLAG_SLOWPATH | FLAG_UPDATE_TS_RECENT) < 0)
5522
		goto discard;
L
Linus Torvalds 已提交
5523

5524
	tcp_rcv_rtt_measure_ts(sk, skb);
L
Linus Torvalds 已提交
5525 5526 5527 5528 5529 5530 5531

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

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

5532
	tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
5533
	tcp_ack_snd_check(sk);
5534
	return;
L
Linus Torvalds 已提交
5535 5536

csum_error:
5537 5538
	TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS);
	TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
5539 5540

discard:
5541
	tcp_drop(sk, skb);
L
Linus Torvalds 已提交
5542
}
E
Eric Dumazet 已提交
5543
EXPORT_SYMBOL(tcp_rcv_established);
L
Linus Torvalds 已提交
5544

P
Pavel Emelyanov 已提交
5545 5546 5547 5548 5549 5550
void tcp_finish_connect(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);

	tcp_set_state(sk, TCP_ESTABLISHED);
5551
	icsk->icsk_ack.lrcvtime = tcp_time_stamp;
P
Pavel Emelyanov 已提交
5552

5553
	if (skb) {
E
Eric Dumazet 已提交
5554
		icsk->icsk_af_ops->sk_rx_dst_set(sk, skb);
P
Pavel Emelyanov 已提交
5555
		security_inet_conn_established(sk, skb);
D
David S. Miller 已提交
5556
	}
P
Pavel Emelyanov 已提交
5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581

	/* Make sure socket is routed, for correct metrics.  */
	icsk->icsk_af_ops->rebuild_header(sk);

	tcp_init_metrics(sk);

	tcp_init_congestion_control(sk);

	/* 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))
		inet_csk_reset_keepalive_timer(sk, keepalive_time_when(tp));

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

}

5582 5583 5584 5585
static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack,
				    struct tcp_fastopen_cookie *cookie)
{
	struct tcp_sock *tp = tcp_sk(sk);
5586
	struct sk_buff *data = tp->syn_data ? tcp_write_queue_head(sk) : NULL;
5587 5588
	u16 mss = tp->rx_opt.mss_clamp, try_exp = 0;
	bool syn_drop = false;
5589 5590 5591 5592 5593 5594 5595

	if (mss == tp->rx_opt.user_mss) {
		struct tcp_options_received opt;

		/* Get original SYNACK MSS value if user MSS sets mss_clamp */
		tcp_clear_options(&opt);
		opt.user_mss = opt.mss_clamp = 0;
C
Christoph Paasch 已提交
5596
		tcp_parse_options(synack, &opt, 0, NULL);
5597 5598 5599
		mss = opt.mss_clamp;
	}

5600 5601
	if (!tp->syn_fastopen) {
		/* Ignore an unsolicited cookie */
5602
		cookie->len = -1;
5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616
	} else if (tp->total_retrans) {
		/* SYN timed out and the SYN-ACK neither has a cookie nor
		 * acknowledges data. Presumably the remote received only
		 * the retransmitted (regular) SYNs: either the original
		 * SYN-data or the corresponding SYN-ACK was dropped.
		 */
		syn_drop = (cookie->len < 0 && data);
	} else if (cookie->len < 0 && !tp->syn_data) {
		/* We requested a cookie but didn't get it. If we did not use
		 * the (old) exp opt format then try so next time (try_exp=1).
		 * Otherwise we go back to use the RFC7413 opt (try_exp=2).
		 */
		try_exp = tp->syn_fastopen_exp ? 2 : 1;
	}
5617

5618
	tcp_fastopen_cache_set(sk, mss, cookie, syn_drop, try_exp);
5619 5620

	if (data) { /* Retransmit unacked data in SYN */
5621 5622
		tcp_for_write_queue_from(data, sk) {
			if (data == tcp_send_head(sk) ||
5623
			    __tcp_retransmit_skb(sk, data, 1))
5624 5625
				break;
		}
5626
		tcp_rearm_rto(sk);
5627
		NET_INC_STATS(sock_net(sk),
5628
				LINUX_MIB_TCPFASTOPENACTIVEFAIL);
5629 5630
		return true;
	}
Y
Yuchung Cheng 已提交
5631
	tp->syn_data_acked = tp->syn_data;
5632
	if (tp->syn_data_acked)
5633
		NET_INC_STATS(sock_net(sk),
5634
				LINUX_MIB_TCPFASTOPENACTIVE);
5635 5636 5637

	tcp_fastopen_add_skb(sk, synack);

5638 5639 5640
	return false;
}

L
Linus Torvalds 已提交
5641
static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
5642
					 const struct tcphdr *th)
L
Linus Torvalds 已提交
5643
{
5644
	struct inet_connection_sock *icsk = inet_csk(sk);
5645
	struct tcp_sock *tp = tcp_sk(sk);
5646
	struct tcp_fastopen_cookie foc = { .len = -1 };
5647
	int saved_clamp = tp->rx_opt.mss_clamp;
5648
	bool fastopen_fail;
L
Linus Torvalds 已提交
5649

C
Christoph Paasch 已提交
5650
	tcp_parse_options(skb, &tp->rx_opt, 0, &foc);
5651
	if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr)
5652
		tp->rx_opt.rcv_tsecr -= tp->tsoffset;
L
Linus Torvalds 已提交
5653 5654 5655 5656 5657 5658 5659 5660 5661 5662

	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)"
		 */
5663 5664
		if (!after(TCP_SKB_CB(skb)->ack_seq, tp->snd_una) ||
		    after(TCP_SKB_CB(skb)->ack_seq, tp->snd_nxt))
L
Linus Torvalds 已提交
5665 5666 5667 5668 5669
			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)) {
5670
			NET_INC_STATS(sock_net(sk),
5671
					LINUX_MIB_PAWSACTIVEREJECTED);
L
Linus Torvalds 已提交
5672 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
			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..."
		 */

5705
		tcp_ecn_rcv_synack(tp, th);
L
Linus Torvalds 已提交
5706

5707
		tcp_init_wl(tp, TCP_SKB_CB(skb)->seq);
L
Linus Torvalds 已提交
5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735
		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);

		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);
		}

5736 5737
		if (tcp_is_sack(tp) && sysctl_tcp_fack)
			tcp_enable_fack(tp);
L
Linus Torvalds 已提交
5738

J
John Heffner 已提交
5739
		tcp_mtup_init(sk);
5740
		tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
5741 5742 5743 5744 5745 5746
		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;
5747

R
Ralf Baechle 已提交
5748
		smp_mb();
L
Linus Torvalds 已提交
5749

P
Pavel Emelyanov 已提交
5750
		tcp_finish_connect(sk, skb);
L
Linus Torvalds 已提交
5751

5752 5753
		fastopen_fail = (tp->syn_fastopen || tp->syn_data) &&
				tcp_rcv_fastopen_synack(sk, skb, &foc);
5754

5755 5756 5757 5758 5759 5760
		if (!sock_flag(sk, SOCK_DEAD)) {
			sk->sk_state_change(sk);
			sk_wake_async(sk, SOCK_WAKE_IO, POLL_OUT);
		}
		if (fastopen_fail)
			return -1;
5761 5762 5763
		if (sk->sk_write_pending ||
		    icsk->icsk_accept_queue.rskq_defer_accept ||
		    icsk->icsk_ack.pingpong) {
L
Linus Torvalds 已提交
5764 5765 5766 5767 5768 5769 5770
			/* 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
			 */
5771 5772
			inet_csk_schedule_ack(sk);
			tcp_enter_quickack_mode(sk);
5773 5774
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
						  TCP_DELACK_MAX, TCP_RTO_MAX);
L
Linus Torvalds 已提交
5775 5776

discard:
5777
			tcp_drop(sk, skb);
L
Linus Torvalds 已提交
5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797
			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. */
5798
	if (tp->rx_opt.ts_recent_stamp && tp->rx_opt.saw_tstamp &&
I
Ilpo Järvinen 已提交
5799
	    tcp_paws_reject(&tp->rx_opt, 0))
L
Linus Torvalds 已提交
5800 5801 5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818
		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;
5819
		tp->copied_seq = tp->rcv_nxt;
L
Linus Torvalds 已提交
5820 5821 5822 5823 5824 5825 5826 5827 5828
		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;

5829
		tcp_ecn_rcv_syn(tp, th);
L
Linus Torvalds 已提交
5830

J
John Heffner 已提交
5831
		tcp_mtup_init(sk);
5832
		tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
5833 5834 5835 5836 5837
		tcp_initialize_rcv_mss(sk);

		tcp_send_synack(sk);
#if 0
		/* Note, we could accept data and URG from this segment.
5838 5839 5840
		 * There are no obstacles to make this (except that we must
		 * either change tcp_recvmsg() to prevent it from returning data
		 * before 3WHS completes per RFC793, or employ TCP Fast Open).
L
Linus Torvalds 已提交
5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869
		 *
		 * 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
5870
 *	all states except ESTABLISHED and TIME_WAIT.
L
Linus Torvalds 已提交
5871 5872 5873
 *	It's called from both tcp_v4_rcv and tcp_v6_rcv and should be
 *	address independent.
 */
5874

5875
int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
5876 5877
{
	struct tcp_sock *tp = tcp_sk(sk);
5878
	struct inet_connection_sock *icsk = inet_csk(sk);
5879
	const struct tcphdr *th = tcp_hdr(skb);
5880
	struct request_sock *req;
L
Linus Torvalds 已提交
5881
	int queued = 0;
5882
	bool acceptable;
L
Linus Torvalds 已提交
5883 5884 5885 5886 5887 5888

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

	case TCP_LISTEN:
S
Stephen Hemminger 已提交
5889
		if (th->ack)
L
Linus Torvalds 已提交
5890 5891
			return 1;

S
Stephen Hemminger 已提交
5892
		if (th->rst)
L
Linus Torvalds 已提交
5893 5894
			goto discard;

S
Stephen Hemminger 已提交
5895
		if (th->syn) {
E
Eric Dumazet 已提交
5896 5897
			if (th->fin)
				goto discard;
5898 5899 5900 5901 5902 5903
			/* It is possible that we process SYN packets from backlog,
			 * so we need to make sure to disable BH right there.
			 */
			local_bh_disable();
			acceptable = icsk->icsk_af_ops->conn_request(sk, skb) >= 0;
			local_bh_enable();
L
Linus Torvalds 已提交
5904

5905 5906
			if (!acceptable)
				return 1;
5907
			consume_skb(skb);
5908
			return 0;
L
Linus Torvalds 已提交
5909 5910 5911 5912
		}
		goto discard;

	case TCP_SYN_SENT:
5913
		tp->rx_opt.saw_tstamp = 0;
E
Eric Dumazet 已提交
5914
		skb_mstamp_get(&tp->tcp_mstamp);
5915
		queued = tcp_rcv_synsent_state_process(sk, skb, th);
L
Linus Torvalds 已提交
5916 5917 5918 5919 5920 5921
		if (queued >= 0)
			return queued;

		/* Do step6 onward by hand. */
		tcp_urg(sk, skb, th);
		__kfree_skb(skb);
5922
		tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
5923 5924 5925
		return 0;
	}

E
Eric Dumazet 已提交
5926
	skb_mstamp_get(&tp->tcp_mstamp);
5927
	tp->rx_opt.saw_tstamp = 0;
5928
	req = tp->fastopen_rsk;
5929
	if (req) {
5930
		WARN_ON_ONCE(sk->sk_state != TCP_SYN_RECV &&
5931 5932
		    sk->sk_state != TCP_FIN_WAIT1);

5933
		if (!tcp_check_req(sk, skb, req, true))
5934
			goto discard;
5935
	}
5936

5937
	if (!th->ack && !th->rst && !th->syn)
5938 5939
		goto discard;

5940
	if (!tcp_validate_incoming(sk, skb, th, 0))
E
Eric Dumazet 已提交
5941
		return 0;
L
Linus Torvalds 已提交
5942 5943

	/* step 5: check the ACK field */
5944 5945
	acceptable = tcp_ack(sk, skb, FLAG_SLOWPATH |
				      FLAG_UPDATE_TS_RECENT) > 0;
5946

5947 5948
	switch (sk->sk_state) {
	case TCP_SYN_RECV:
5949 5950
		if (!acceptable)
			return 1;
L
Linus Torvalds 已提交
5951

Y
Yuchung Cheng 已提交
5952 5953 5954
		if (!tp->srtt_us)
			tcp_synack_rtt_meas(sk, req);

5955 5956 5957 5958
		/* Once we leave TCP_SYN_RECV, we no longer need req
		 * so release it.
		 */
		if (req) {
5959
			inet_csk(sk)->icsk_retransmits = 0;
5960 5961 5962 5963 5964 5965 5966 5967
			reqsk_fastopen_remove(sk, req, false);
		} else {
			/* Make sure socket is routed, for correct metrics. */
			icsk->icsk_af_ops->rebuild_header(sk);
			tcp_init_congestion_control(sk);

			tcp_mtup_init(sk);
			tp->copied_seq = tp->rcv_nxt;
E
Eric Dumazet 已提交
5968
			tcp_init_buffer_space(sk);
5969 5970 5971 5972
		}
		smp_mb();
		tcp_set_state(sk, TCP_ESTABLISHED);
		sk->sk_state_change(sk);
L
Linus Torvalds 已提交
5973

5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995
		/* 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.
		 */
		if (sk->sk_socket)
			sk_wake_async(sk, SOCK_WAKE_IO, POLL_OUT);

		tp->snd_una = TCP_SKB_CB(skb)->ack_seq;
		tp->snd_wnd = ntohs(th->window) << tp->rx_opt.snd_wscale;
		tcp_init_wl(tp, TCP_SKB_CB(skb)->seq);

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

		if (req) {
			/* Re-arm the timer because data may have been sent out.
			 * This is similar to the regular data transmission case
			 * when new data has just been ack'ed.
			 *
			 * (TFO) - we could try to be more aggressive and
			 * retransmitting any data sooner based on when they
			 * are sent out.
5996
			 */
5997 5998 5999
			tcp_rearm_rto(sk);
		} else
			tcp_init_metrics(sk);
L
Linus Torvalds 已提交
6000

6001 6002
		if (!inet_csk(sk)->icsk_ca_ops->cong_control)
			tcp_update_pacing_rate(sk);
6003

6004 6005 6006 6007 6008
		/* Prevent spurious tcp_cwnd_restart() on first data packet */
		tp->lsndtime = tcp_time_stamp;

		tcp_initialize_rcv_mss(sk);
		tcp_fast_path_on(tp);
6009 6010
		break;

6011 6012 6013
	case TCP_FIN_WAIT1: {
		int tmo;

6014 6015 6016 6017 6018
		/* If we enter the TCP_FIN_WAIT1 state and we are a
		 * Fast Open socket and this is the first acceptable
		 * ACK we have received, this would have acknowledged
		 * our SYNACK so stop the SYNACK timer.
		 */
6019
		if (req) {
6020 6021 6022 6023 6024
			/* Return RST if ack_seq is invalid.
			 * Note that RFC793 only says to generate a
			 * DUPACK for it but for TCP Fast Open it seems
			 * better to treat this case like TCP_SYN_RECV
			 * above.
6025
			 */
6026 6027 6028 6029 6030 6031
			if (!acceptable)
				return 1;
			/* We no longer need the request sock. */
			reqsk_fastopen_remove(sk, req, false);
			tcp_rearm_rto(sk);
		}
6032 6033
		if (tp->snd_una != tp->write_seq)
			break;
6034

6035 6036
		tcp_set_state(sk, TCP_FIN_WAIT2);
		sk->sk_shutdown |= SEND_SHUTDOWN;
6037

6038
		sk_dst_confirm(sk);
6039

6040 6041 6042 6043 6044
		if (!sock_flag(sk, SOCK_DEAD)) {
			/* Wake up lingering close() */
			sk->sk_state_change(sk);
			break;
		}
L
Linus Torvalds 已提交
6045

6046 6047 6048 6049 6050 6051 6052 6053 6054
		if (tp->linger2 < 0) {
			tcp_done(sk);
			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONDATA);
			return 1;
		}
		if (TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(skb)->seq &&
		    after(TCP_SKB_CB(skb)->end_seq - th->fin, tp->rcv_nxt)) {
			/* Receive out of order FIN after close() */
			if (tp->syn_fastopen && th->fin)
6055
				tcp_fastopen_active_disable(sk);
6056
			tcp_done(sk);
6057
			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONDATA);
6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074
			return 1;
		}

		tmo = tcp_fin_time(sk);
		if (tmo > TCP_TIMEWAIT_LEN) {
			inet_csk_reset_keepalive_timer(sk, tmo - TCP_TIMEWAIT_LEN);
		} 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.
			 */
			inet_csk_reset_keepalive_timer(sk, tmo);
		} else {
			tcp_time_wait(sk, TCP_FIN_WAIT2, tmo);
			goto discard;
6075 6076
		}
		break;
6077
	}
L
Linus Torvalds 已提交
6078

6079 6080 6081 6082 6083 6084
	case TCP_CLOSING:
		if (tp->snd_una == tp->write_seq) {
			tcp_time_wait(sk, TCP_TIME_WAIT, 0);
			goto discard;
		}
		break;
L
Linus Torvalds 已提交
6085

6086 6087 6088 6089 6090
	case TCP_LAST_ACK:
		if (tp->snd_una == tp->write_seq) {
			tcp_update_metrics(sk);
			tcp_done(sk);
			goto discard;
L
Linus Torvalds 已提交
6091
		}
6092
		break;
6093
	}
L
Linus Torvalds 已提交
6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107

	/* 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,
6108
		 * RFC 1122 says we MUST send a reset.
L
Linus Torvalds 已提交
6109 6110 6111 6112 6113
		 * 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)) {
6114
				NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONDATA);
L
Linus Torvalds 已提交
6115 6116 6117 6118 6119
				tcp_reset(sk);
				return 1;
			}
		}
		/* Fall through */
6120
	case TCP_ESTABLISHED:
L
Linus Torvalds 已提交
6121 6122 6123 6124 6125 6126 6127
		tcp_data_queue(sk, skb);
		queued = 1;
		break;
	}

	/* tcp_data could move socket to TIME-WAIT */
	if (sk->sk_state != TCP_CLOSE) {
6128
		tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
6129 6130 6131
		tcp_ack_snd_check(sk);
	}

6132
	if (!queued) {
L
Linus Torvalds 已提交
6133
discard:
6134
		tcp_drop(sk, skb);
L
Linus Torvalds 已提交
6135 6136 6137 6138
	}
	return 0;
}
EXPORT_SYMBOL(tcp_rcv_state_process);
O
Octavian Purdila 已提交
6139 6140 6141 6142 6143 6144

static inline void pr_drop_req(struct request_sock *req, __u16 port, int family)
{
	struct inet_request_sock *ireq = inet_rsk(req);

	if (family == AF_INET)
6145 6146
		net_dbg_ratelimited("drop open request from %pI4/%u\n",
				    &ireq->ir_rmt_addr, port);
6147 6148
#if IS_ENABLED(CONFIG_IPV6)
	else if (family == AF_INET6)
6149 6150
		net_dbg_ratelimited("drop open request from %pI6/%u\n",
				    &ireq->ir_v6_rmt_addr, port);
6151
#endif
O
Octavian Purdila 已提交
6152 6153
}

6154 6155 6156 6157 6158
/* RFC3168 : 6.1.1 SYN packets must not have ECT/ECN bits set
 *
 * If we receive a SYN packet with these bits set, it means a
 * network is playing bad games with TOS bits. In order to
 * avoid possible false congestion notifications, we disable
S
stephen hemminger 已提交
6159
 * TCP ECN negotiation.
6160 6161
 *
 * Exception: tcp_ca wants ECN. This is required for DCTCP
6162 6163 6164
 * congestion control: Linux DCTCP asserts ECT on all packets,
 * including SYN, which is most optimal solution; however,
 * others, such as FreeBSD do not.
6165 6166 6167
 */
static void tcp_ecn_create_request(struct request_sock *req,
				   const struct sk_buff *skb,
6168 6169
				   const struct sock *listen_sk,
				   const struct dst_entry *dst)
6170 6171 6172 6173
{
	const struct tcphdr *th = tcp_hdr(skb);
	const struct net *net = sock_net(listen_sk);
	bool th_ecn = th->ece && th->cwr;
6174
	bool ect, ecn_ok;
6175
	u32 ecn_ok_dst;
6176 6177 6178 6179 6180

	if (!th_ecn)
		return;

	ect = !INET_ECN_is_not_ect(TCP_SKB_CB(skb)->ip_dsfield);
6181 6182
	ecn_ok_dst = dst_feature(dst, DST_FEATURE_ECN_MASK);
	ecn_ok = net->ipv4.sysctl_tcp_ecn || ecn_ok_dst;
6183

6184 6185
	if ((!ect && ecn_ok) || tcp_ca_needs_ecn(listen_sk) ||
	    (ecn_ok_dst & DST_FEATURE_ECN_CA))
6186 6187 6188
		inet_rsk(req)->ecn_ok = 1;
}

6189 6190 6191 6192 6193 6194
static void tcp_openreq_init(struct request_sock *req,
			     const struct tcp_options_received *rx_opt,
			     struct sk_buff *skb, const struct sock *sk)
{
	struct inet_request_sock *ireq = inet_rsk(req);

6195
	req->rsk_rcv_wnd = 0;		/* So that tcp_send_synack() knows! */
6196 6197 6198
	req->cookie_ts = 0;
	tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
	tcp_rsk(req)->rcv_nxt = TCP_SKB_CB(skb)->seq + 1;
Y
Yuchung Cheng 已提交
6199
	skb_mstamp_get(&tcp_rsk(req)->snt_synack);
6200 6201 6202 6203 6204 6205 6206 6207 6208 6209 6210 6211 6212 6213
	tcp_rsk(req)->last_oow_ack_time = 0;
	req->mss = rx_opt->mss_clamp;
	req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
	ireq->tstamp_ok = rx_opt->tstamp_ok;
	ireq->sack_ok = rx_opt->sack_ok;
	ireq->snd_wscale = rx_opt->snd_wscale;
	ireq->wscale_ok = rx_opt->wscale_ok;
	ireq->acked = 0;
	ireq->ecn_ok = 0;
	ireq->ir_rmt_port = tcp_hdr(skb)->source;
	ireq->ir_num = ntohs(tcp_hdr(skb)->dest);
	ireq->ir_mark = inet_request_mark(sk, skb);
}

E
Eric Dumazet 已提交
6214
struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
6215 6216
				      struct sock *sk_listener,
				      bool attach_listener)
E
Eric Dumazet 已提交
6217
{
6218 6219
	struct request_sock *req = reqsk_alloc(ops, sk_listener,
					       attach_listener);
E
Eric Dumazet 已提交
6220 6221 6222 6223 6224 6225

	if (req) {
		struct inet_request_sock *ireq = inet_rsk(req);

		kmemcheck_annotate_bitfield(ireq, flags);
		ireq->opt = NULL;
6226 6227 6228
#if IS_ENABLED(CONFIG_IPV6)
		ireq->pktopts = NULL;
#endif
E
Eric Dumazet 已提交
6229 6230 6231
		atomic64_set(&ireq->ir_cookie, 0);
		ireq->ireq_state = TCP_NEW_SYN_RECV;
		write_pnet(&ireq->ireq_net, sock_net(sk_listener));
6232
		ireq->ireq_family = sk_listener->sk_family;
E
Eric Dumazet 已提交
6233 6234 6235 6236 6237 6238
	}

	return req;
}
EXPORT_SYMBOL(inet_reqsk_alloc);

6239 6240 6241
/*
 * Return true if a syncookie should be sent
 */
6242
static bool tcp_syn_flood_action(const struct sock *sk,
6243 6244 6245
				 const struct sk_buff *skb,
				 const char *proto)
{
6246
	struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue;
6247 6248
	const char *msg = "Dropping request";
	bool want_cookie = false;
6249
	struct net *net = sock_net(sk);
6250 6251

#ifdef CONFIG_SYN_COOKIES
6252
	if (net->ipv4.sysctl_tcp_syncookies) {
6253 6254
		msg = "Sending cookies";
		want_cookie = true;
6255
		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPREQQFULLDOCOOKIES);
6256 6257
	} else
#endif
6258
		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPREQQFULLDROP);
6259

6260
	if (!queue->synflood_warned &&
6261
	    net->ipv4.sysctl_tcp_syncookies != 2 &&
6262
	    xchg(&queue->synflood_warned, 1) == 0)
6263 6264
		pr_info("%s: Possible SYN flooding on port %d. %s.  Check SNMP counters.\n",
			proto, ntohs(tcp_hdr(skb)->dest), msg);
6265

6266 6267 6268
	return want_cookie;
}

6269 6270 6271 6272 6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 6285
static void tcp_reqsk_record_syn(const struct sock *sk,
				 struct request_sock *req,
				 const struct sk_buff *skb)
{
	if (tcp_sk(sk)->save_syn) {
		u32 len = skb_network_header_len(skb) + tcp_hdrlen(skb);
		u32 *copy;

		copy = kmalloc(len + sizeof(u32), GFP_ATOMIC);
		if (copy) {
			copy[0] = len;
			memcpy(&copy[1], skb_network_header(skb), len);
			req->saved_syn = copy;
		}
	}
}

O
Octavian Purdila 已提交
6286 6287 6288 6289
int tcp_conn_request(struct request_sock_ops *rsk_ops,
		     const struct tcp_request_sock_ops *af_ops,
		     struct sock *sk, struct sk_buff *skb)
{
6290 6291
	struct tcp_fastopen_cookie foc = { .len = -1 };
	__u32 isn = TCP_SKB_CB(skb)->tcp_tw_isn;
O
Octavian Purdila 已提交
6292 6293
	struct tcp_options_received tmp_opt;
	struct tcp_sock *tp = tcp_sk(sk);
6294
	struct net *net = sock_net(sk);
6295
	struct sock *fastopen_sk = NULL;
O
Octavian Purdila 已提交
6296
	struct dst_entry *dst = NULL;
6297 6298
	struct request_sock *req;
	bool want_cookie = false;
O
Octavian Purdila 已提交
6299 6300 6301 6302 6303 6304
	struct flowi fl;

	/* TW buckets are converted to open requests without
	 * limitations, they conserve resources and peer is
	 * evidently real one.
	 */
6305
	if ((net->ipv4.sysctl_tcp_syncookies == 2 ||
O
Octavian Purdila 已提交
6306 6307 6308 6309 6310 6311
	     inet_csk_reqsk_queue_is_full(sk)) && !isn) {
		want_cookie = tcp_syn_flood_action(sk, skb, rsk_ops->slab_name);
		if (!want_cookie)
			goto drop;
	}

6312
	if (sk_acceptq_is_full(sk)) {
6313
		NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
O
Octavian Purdila 已提交
6314 6315 6316
		goto drop;
	}

6317
	req = inet_reqsk_alloc(rsk_ops, sk, !want_cookie);
O
Octavian Purdila 已提交
6318 6319 6320 6321
	if (!req)
		goto drop;

	tcp_rsk(req)->af_specific = af_ops;
6322
	tcp_rsk(req)->ts_off = 0;
O
Octavian Purdila 已提交
6323 6324 6325 6326 6327 6328 6329 6330 6331 6332 6333

	tcp_clear_options(&tmp_opt);
	tmp_opt.mss_clamp = af_ops->mss_clamp;
	tmp_opt.user_mss  = tp->rx_opt.user_mss;
	tcp_parse_options(skb, &tmp_opt, 0, want_cookie ? NULL : &foc);

	if (want_cookie && !tmp_opt.saw_tstamp)
		tcp_clear_options(&tmp_opt);

	tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
	tcp_openreq_init(req, &tmp_opt, skb, sk);
6334
	inet_rsk(req)->no_srccheck = inet_sk(sk)->transparent;
O
Octavian Purdila 已提交
6335

6336
	/* Note: tcp_v6_init_req() might override ir_iif for link locals */
6337
	inet_rsk(req)->ir_iif = inet_request_bound_dev_if(sk, skb);
6338

O
Octavian Purdila 已提交
6339 6340 6341 6342 6343
	af_ops->init_req(req, sk, skb);

	if (security_inet_conn_request(sk, skb, req))
		goto drop_and_free;

6344
	if (isn && tmp_opt.tstamp_ok)
6345
		af_ops->init_seq_tsoff(skb, &tcp_rsk(req)->ts_off);
6346

6347
	if (!want_cookie && !isn) {
O
Octavian Purdila 已提交
6348
		/* Kill the following clause, if you dislike this way. */
6349 6350 6351 6352
		if (!net->ipv4.sysctl_tcp_syncookies &&
		    (net->ipv4.sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) <
		     (net->ipv4.sysctl_max_syn_backlog >> 2)) &&
		    !tcp_peer_is_proven(req, dst)) {
O
Octavian Purdila 已提交
6353 6354 6355 6356 6357 6358 6359 6360 6361 6362 6363 6364
			/* Without syncookies last quarter of
			 * backlog is filled with destinations,
			 * proven to be alive.
			 * It means that we continue to communicate
			 * to destinations, already remembered
			 * to the moment of synflood.
			 */
			pr_drop_req(req, ntohs(tcp_hdr(skb)->source),
				    rsk_ops->family);
			goto drop_and_release;
		}

6365
		isn = af_ops->init_seq_tsoff(skb, &tcp_rsk(req)->ts_off);
O
Octavian Purdila 已提交
6366 6367
	}
	if (!dst) {
6368
		dst = af_ops->route_req(sk, &fl, req);
O
Octavian Purdila 已提交
6369 6370 6371 6372
		if (!dst)
			goto drop_and_free;
	}

6373 6374 6375 6376
	tcp_ecn_create_request(req, skb, sk, dst);

	if (want_cookie) {
		isn = cookie_init_sequence(af_ops, sk, skb, &req->mss);
6377
		tcp_rsk(req)->ts_off = 0;
6378 6379 6380 6381 6382
		req->cookie_ts = tmp_opt.tstamp_ok;
		if (!tmp_opt.tstamp_ok)
			inet_rsk(req)->ecn_ok = 0;
	}

O
Octavian Purdila 已提交
6383
	tcp_rsk(req)->snt_isn = isn;
6384
	tcp_rsk(req)->txhash = net_tx_rndhash();
O
Octavian Purdila 已提交
6385
	tcp_openreq_init_rwin(req, sk, dst);
6386 6387
	if (!want_cookie) {
		tcp_reqsk_record_syn(sk, req, skb);
6388
		fastopen_sk = tcp_try_fastopen(sk, skb, req, &foc, dst);
6389
	}
6390
	if (fastopen_sk) {
6391
		af_ops->send_synack(fastopen_sk, dst, &fl, req,
6392
				    &foc, TCP_SYNACK_FASTOPEN);
6393 6394 6395 6396
		/* Add the child socket directly into the accept queue */
		inet_csk_reqsk_queue_add(sk, req, fastopen_sk);
		sk->sk_data_ready(sk);
		bh_unlock_sock(fastopen_sk);
6397 6398
		sock_put(fastopen_sk);
	} else {
6399
		tcp_rsk(req)->tfo_listener = false;
6400 6401
		if (!want_cookie)
			inet_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
6402 6403 6404
		af_ops->send_synack(sk, dst, &fl, req, &foc,
				    !want_cookie ? TCP_SYNACK_NORMAL :
						   TCP_SYNACK_COOKIE);
6405 6406 6407 6408
		if (want_cookie) {
			reqsk_free(req);
			return 0;
		}
O
Octavian Purdila 已提交
6409
	}
6410
	reqsk_put(req);
O
Octavian Purdila 已提交
6411 6412 6413 6414 6415 6416 6417
	return 0;

drop_and_release:
	dst_release(dst);
drop_and_free:
	reqsk_free(req);
drop:
6418
	tcp_listendrop(sk);
O
Octavian Purdila 已提交
6419 6420 6421
	return 0;
}
EXPORT_SYMBOL(tcp_conn_request);