tcp_input.c 183.0 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);
L
Linus Torvalds 已提交
88

E
Eric Dumazet 已提交
89
/* rfc5961 challenge ack rate limiting */
90
int sysctl_tcp_challenge_ack_limit = 1000;
E
Eric Dumazet 已提交
91

92 93 94
int sysctl_tcp_stdurg __read_mostly;
int sysctl_tcp_rfc1337 __read_mostly;
int sysctl_tcp_max_orphans __read_mostly = NR_FILE;
95
int sysctl_tcp_frto __read_mostly = 2;
96
int sysctl_tcp_min_rtt_wlen __read_mostly = 300;
97
int sysctl_tcp_moderate_rcvbuf __read_mostly = 1;
N
Nandita Dukkipati 已提交
98
int sysctl_tcp_early_retrans __read_mostly = 3;
99
int sysctl_tcp_invalid_ratelimit __read_mostly = HZ/2;
L
Linus Torvalds 已提交
100 101 102 103 104 105 106 107

#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				*/
108
#define FLAG_LOST_RETRANS	0x80 /* This ACK marks some retransmission lost */
L
Linus Torvalds 已提交
109
#define FLAG_SLOWPATH		0x100 /* Do not skip RFC checks for window update.*/
Y
Yuchung Cheng 已提交
110
#define FLAG_ORIG_SACK_ACKED	0x200 /* Never retransmitted data are (s)acked	*/
111
#define FLAG_SND_UNA_ADVANCED	0x400 /* Snd_una was changed (!= FLAG_DATA_ACKED) */
112
#define FLAG_DSACKING_ACK	0x800 /* SACK blocks contained D-SACK info */
113
#define FLAG_SACK_RENEGING	0x2000 /* snd_una advanced to a sacked seq */
114
#define FLAG_UPDATE_TS_RECENT	0x4000 /* tcp_replace_ts_recent() */
L
Linus Torvalds 已提交
115 116 117 118 119 120 121

#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)
122
#define TCP_HP_BITS (~(TCP_RESERVED_BITS|TCP_FLAG_PSH))
L
Linus Torvalds 已提交
123

124 125 126 127
#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 */

128 129
static void tcp_gro_dev_warn(struct sock *sk, const struct sk_buff *skb,
			     unsigned int len)
130 131 132 133 134 135 136 137 138 139
{
	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);
140 141 142
		if (!dev || len >= dev->mtu)
			pr_warn("%s: Driver has suspect GRO implementation, TCP performance may be compromised.\n",
				dev ? dev->name : "Unknown driver");
143 144 145 146
		rcu_read_unlock();
	}
}

147
/* Adapt the MSS value used to make delayed ack decision to the
L
Linus Torvalds 已提交
148
 * real world.
149
 */
150
static void tcp_measure_rcv_mss(struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
151
{
152
	struct inet_connection_sock *icsk = inet_csk(sk);
153
	const unsigned int lss = icsk->icsk_ack.last_seg_size;
154
	unsigned int len;
L
Linus Torvalds 已提交
155

156
	icsk->icsk_ack.last_seg_size = 0;
L
Linus Torvalds 已提交
157 158 159 160

	/* skb->len may jitter because of SACKs, even if peer
	 * sends good full-sized frames.
	 */
161
	len = skb_shinfo(skb)->gso_size ? : skb->len;
162
	if (len >= icsk->icsk_ack.rcv_mss) {
163 164
		icsk->icsk_ack.rcv_mss = min_t(unsigned int, len,
					       tcp_sk(sk)->advmss);
165 166 167 168
		/* 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
169 170 171 172 173 174
	} else {
		/* Otherwise, we make more careful check taking into account,
		 * that SACKs block is variable.
		 *
		 * "len" is invariant segment length, including TCP header.
		 */
175
		len += skb->data - skb_transport_header(skb);
176
		if (len >= TCP_MSS_DEFAULT + sizeof(struct tcphdr) ||
L
Linus Torvalds 已提交
177 178 179 180 181 182
		    /* 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) &&
183
		     !(tcp_flag_word(tcp_hdr(skb)) & TCP_REMNANT))) {
L
Linus Torvalds 已提交
184 185 186 187
			/* Subtract also invariant (if peer is RFC compliant),
			 * tcp header plus fixed timestamp option length.
			 * Resulting "len" is MSS free of SACK jitter.
			 */
188 189
			len -= tcp_sk(sk)->tcp_header_len;
			icsk->icsk_ack.last_seg_size = len;
L
Linus Torvalds 已提交
190
			if (len == lss) {
191
				icsk->icsk_ack.rcv_mss = len;
L
Linus Torvalds 已提交
192 193 194
				return;
			}
		}
195 196
		if (icsk->icsk_ack.pending & ICSK_ACK_PUSHED)
			icsk->icsk_ack.pending |= ICSK_ACK_PUSHED2;
197
		icsk->icsk_ack.pending |= ICSK_ACK_PUSHED;
L
Linus Torvalds 已提交
198 199 200
	}
}

201
static void tcp_incr_quickack(struct sock *sk)
L
Linus Torvalds 已提交
202
{
203
	struct inet_connection_sock *icsk = inet_csk(sk);
204
	unsigned int quickacks = tcp_sk(sk)->rcv_wnd / (2 * icsk->icsk_ack.rcv_mss);
L
Linus Torvalds 已提交
205

206 207
	if (quickacks == 0)
		quickacks = 2;
208 209
	if (quickacks > icsk->icsk_ack.quick)
		icsk->icsk_ack.quick = min(quickacks, TCP_MAX_QUICKACKS);
L
Linus Torvalds 已提交
210 211
}

212
static void tcp_enter_quickack_mode(struct sock *sk)
L
Linus Torvalds 已提交
213
{
214 215 216 217
	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
218 219 220 221 222 223
}

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

224
static bool tcp_in_quickack_mode(struct sock *sk)
L
Linus Torvalds 已提交
225
{
226
	const struct inet_connection_sock *icsk = inet_csk(sk);
227
	const struct dst_entry *dst = __sk_dst_get(sk);
E
Eric Dumazet 已提交
228

229 230
	return (dst && dst_metric(dst, RTAX_QUICKACK)) ||
		(icsk->icsk_ack.quick && !icsk->icsk_ack.pingpong);
L
Linus Torvalds 已提交
231 232
}

233
static void tcp_ecn_queue_cwr(struct tcp_sock *tp)
234
{
235
	if (tp->ecn_flags & TCP_ECN_OK)
236 237 238
		tp->ecn_flags |= TCP_ECN_QUEUE_CWR;
}

239
static void tcp_ecn_accept_cwr(struct tcp_sock *tp, const struct sk_buff *skb)
240 241 242 243 244
{
	if (tcp_hdr(skb)->cwr)
		tp->ecn_flags &= ~TCP_ECN_DEMAND_CWR;
}

245
static void tcp_ecn_withdraw_cwr(struct tcp_sock *tp)
246 247 248 249
{
	tp->ecn_flags &= ~TCP_ECN_DEMAND_CWR;
}

250
static void __tcp_ecn_check_ce(struct tcp_sock *tp, const struct sk_buff *skb)
251
{
252
	switch (TCP_SKB_CB(skb)->ip_dsfield & INET_ECN_MASK) {
253
	case INET_ECN_NOT_ECT:
254
		/* Funny extension: if ECT is not set on a segment,
255 256 257 258
		 * and we already seen ECT on a previous segment,
		 * it is probably a retransmit.
		 */
		if (tp->ecn_flags & TCP_ECN_SEEN)
259
			tcp_enter_quickack_mode((struct sock *)tp);
260 261
		break;
	case INET_ECN_CE:
262 263 264
		if (tcp_ca_needs_ecn((struct sock *)tp))
			tcp_ca_event((struct sock *)tp, CA_EVENT_ECN_IS_CE);

265 266 267 268 269
		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;
		}
270 271
		tp->ecn_flags |= TCP_ECN_SEEN;
		break;
272
	default:
273 274
		if (tcp_ca_needs_ecn((struct sock *)tp))
			tcp_ca_event((struct sock *)tp, CA_EVENT_ECN_NO_CE);
275
		tp->ecn_flags |= TCP_ECN_SEEN;
276
		break;
277 278 279
	}
}

280 281 282 283 284 285 286
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)
287
{
288
	if ((tp->ecn_flags & TCP_ECN_OK) && (!th->ece || th->cwr))
289 290 291
		tp->ecn_flags &= ~TCP_ECN_OK;
}

292
static void tcp_ecn_rcv_syn(struct tcp_sock *tp, const struct tcphdr *th)
293
{
294
	if ((tp->ecn_flags & TCP_ECN_OK) && (!th->ece || !th->cwr))
295 296 297
		tp->ecn_flags &= ~TCP_ECN_OK;
}

298
static bool tcp_ecn_rcv_ecn_echo(const struct tcp_sock *tp, const struct tcphdr *th)
299
{
300
	if (th->ece && !th->syn && (tp->ecn_flags & TCP_ECN_OK))
E
Eric Dumazet 已提交
301 302
		return true;
	return false;
303 304
}

L
Linus Torvalds 已提交
305 306 307 308 309
/* Buffer size and advertised window tuning.
 *
 * 1. Tuning sk->sk_sndbuf, when connection enters established state.
 */

E
Eric Dumazet 已提交
310
static void tcp_sndbuf_expand(struct sock *sk)
L
Linus Torvalds 已提交
311
{
E
Eric Dumazet 已提交
312
	const struct tcp_sock *tp = tcp_sk(sk);
313
	const struct tcp_congestion_ops *ca_ops = inet_csk(sk)->icsk_ca_ops;
E
Eric Dumazet 已提交
314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
	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)
	 */
334 335
	sndmem = ca_ops->sndbuf_expand ? ca_ops->sndbuf_expand(sk) : 2;
	sndmem *= nr_segs * per_mss;
L
Linus Torvalds 已提交
336

337 338
	if (sk->sk_sndbuf < sndmem)
		sk->sk_sndbuf = min(sndmem, sysctl_tcp_wmem[2]);
L
Linus Torvalds 已提交
339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361
}

/* 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
362
 * window and then starts to feed us spaghetti. But it should work
L
Linus Torvalds 已提交
363 364 365 366
 * in common situations. Otherwise, we have to rely on queue collapsing.
 */

/* Slow part of check#2. */
367
static int __tcp_grow_window(const struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
368
{
369
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
370
	/* Optimize this! */
371 372
	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
373 374 375

	while (tp->rcv_ssthresh <= window) {
		if (truesize <= skb->len)
376
			return 2 * inet_csk(sk)->icsk_ack.rcv_mss;
L
Linus Torvalds 已提交
377 378 379 380 381 382 383

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

384
static void tcp_grow_window(struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
385
{
386 387
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
388 389 390
	/* Check #1 */
	if (tp->rcv_ssthresh < tp->window_clamp &&
	    (int)tp->rcv_ssthresh < tcp_space(sk) &&
391
	    !tcp_under_memory_pressure(sk)) {
L
Linus Torvalds 已提交
392 393 394 395 396 397
		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)
398
			incr = 2 * tp->advmss;
L
Linus Torvalds 已提交
399
		else
400
			incr = __tcp_grow_window(sk, skb);
L
Linus Torvalds 已提交
401 402

		if (incr) {
403
			incr = max_t(int, incr, 2 * skb->len);
404 405
			tp->rcv_ssthresh = min(tp->rcv_ssthresh + incr,
					       tp->window_clamp);
406
			inet_csk(sk)->icsk_ack.quick |= 1;
L
Linus Torvalds 已提交
407 408 409 410 411 412 413
		}
	}
}

/* 3. Tuning rcvbuf, when connection enters established state. */
static void tcp_fixup_rcvbuf(struct sock *sk)
{
414 415
	u32 mss = tcp_sk(sk)->advmss;
	int rcvmem;
L
Linus Torvalds 已提交
416

417 418
	rcvmem = 2 * SKB_TRUESIZE(mss + MAX_TCP_HEADER) *
		 tcp_default_init_rwnd(mss);
419

E
Eric Dumazet 已提交
420 421 422 423 424 425
	/* 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;

426 427
	if (sk->sk_rcvbuf < rcvmem)
		sk->sk_rcvbuf = min(rcvmem, sysctl_tcp_rmem[2]);
L
Linus Torvalds 已提交
428 429
}

S
Stephen Hemminger 已提交
430
/* 4. Try to fixup all. It is made immediately after connection enters
L
Linus Torvalds 已提交
431 432
 *    established state.
 */
433
void tcp_init_buffer_space(struct sock *sk)
L
Linus Torvalds 已提交
434 435 436 437 438 439 440
{
	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
441
		tcp_sndbuf_expand(sk);
L
Linus Torvalds 已提交
442 443

	tp->rcvq_space.space = tp->rcv_wnd;
444 445
	skb_mstamp_get(&tp->tcp_mstamp);
	tp->rcvq_space.time = tp->tcp_mstamp;
E
Eric Dumazet 已提交
446
	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
 * 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)
{
521
	u32 new_sample = tp->rcv_rtt_est.rtt_us;
L
Linus Torvalds 已提交
522 523 524 525 526 527 528 529 530 531 532 533
	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
		new_sample = m << 3;
	}

551
	tp->rcv_rtt_est.rtt_us = new_sample;
L
Linus Torvalds 已提交
552 553 554 555
}

static inline void tcp_rcv_rtt_measure(struct tcp_sock *tp)
{
556 557 558
	u32 delta_us;

	if (tp->rcv_rtt_est.time.v64 == 0)
L
Linus Torvalds 已提交
559 560 561
		goto new_measure;
	if (before(tp->rcv_nxt, tp->rcv_rtt_est.seq))
		return;
562 563
	delta_us = skb_mstamp_us_delta(&tp->tcp_mstamp, &tp->rcv_rtt_est.time);
	tcp_rcv_rtt_update(tp, delta_us, 1);
L
Linus Torvalds 已提交
564 565 566

new_measure:
	tp->rcv_rtt_est.seq = tp->rcv_nxt + tp->rcv_wnd;
567
	tp->rcv_rtt_est.time = tp->tcp_mstamp;
L
Linus Torvalds 已提交
568 569
}

570 571
static inline void tcp_rcv_rtt_measure_ts(struct sock *sk,
					  const struct sk_buff *skb)
L
Linus Torvalds 已提交
572
{
573
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
574 575
	if (tp->rx_opt.rcv_tsecr &&
	    (TCP_SKB_CB(skb)->end_seq -
576
	     TCP_SKB_CB(skb)->seq >= inet_csk(sk)->icsk_ack.rcv_mss))
577 578 579 580
		tcp_rcv_rtt_update(tp,
				   jiffies_to_usecs(tcp_time_stamp -
						    tp->rx_opt.rcv_tsecr),
				   0);
L
Linus Torvalds 已提交
581 582 583 584 585 586 587 588 589 590
}

/*
 * 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
591
	int copied;
592

593 594
	time = skb_mstamp_us_delta(&tp->tcp_mstamp, &tp->rcvq_space.time);
	if (time < (tp->rcv_rtt_est.rtt_us >> 3) || tp->rcv_rtt_est.rtt_us == 0)
L
Linus Torvalds 已提交
595
		return;
596

E
Eric Dumazet 已提交
597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613
	/* 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
614

E
Eric Dumazet 已提交
615 616 617 618
		/* 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
619

E
Eric Dumazet 已提交
620 621 622 623 624 625 626 627 628 629 630 631 632
		/* 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
633

E
Eric Dumazet 已提交
634 635 636
		rcvmem = SKB_TRUESIZE(tp->advmss + MAX_TCP_HEADER);
		while (tcp_win_from_space(rcvmem) < tp->advmss)
			rcvmem += 128;
L
Linus Torvalds 已提交
637

E
Eric Dumazet 已提交
638 639 640
		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
641

E
Eric Dumazet 已提交
642 643
			/* Make the window clamp follow along.  */
			tp->window_clamp = rcvwin;
L
Linus Torvalds 已提交
644 645
		}
	}
E
Eric Dumazet 已提交
646
	tp->rcvq_space.space = copied;
647

L
Linus Torvalds 已提交
648 649
new_measure:
	tp->rcvq_space.seq = tp->copied_seq;
650
	tp->rcvq_space.time = tp->tcp_mstamp;
L
Linus Torvalds 已提交
651 652 653 654 655 656 657 658 659 660 661 662
}

/* 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
 */
663
static void tcp_event_data_recv(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
664
{
665
	struct tcp_sock *tp = tcp_sk(sk);
666
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
667 668
	u32 now;

669
	inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
670

671
	tcp_measure_rcv_mss(sk, skb);
L
Linus Torvalds 已提交
672 673

	tcp_rcv_rtt_measure(tp);
674

L
Linus Torvalds 已提交
675 676
	now = tcp_time_stamp;

677
	if (!icsk->icsk_ack.ato) {
L
Linus Torvalds 已提交
678 679 680
		/* The _first_ data packet received, initialize
		 * delayed ACK engine.
		 */
681 682
		tcp_incr_quickack(sk);
		icsk->icsk_ack.ato = TCP_ATO_MIN;
L
Linus Torvalds 已提交
683
	} else {
684
		int m = now - icsk->icsk_ack.lrcvtime;
L
Linus Torvalds 已提交
685

686
		if (m <= TCP_ATO_MIN / 2) {
L
Linus Torvalds 已提交
687
			/* The fastest case is the first. */
688 689 690 691 692 693
			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
694
			/* Too long gap. Apparently sender failed to
L
Linus Torvalds 已提交
695 696
			 * restart window, so that we send ACKs quickly.
			 */
697
			tcp_incr_quickack(sk);
698
			sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
699 700
		}
	}
701
	icsk->icsk_ack.lrcvtime = now;
L
Linus Torvalds 已提交
702

703
	tcp_ecn_check_ce(tp, skb);
L
Linus Torvalds 已提交
704 705

	if (skb->len >= 128)
706
		tcp_grow_window(sk, skb);
L
Linus Torvalds 已提交
707 708 709 710 711 712 713 714 715 716 717
}

/* 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
 */
718
static void tcp_rtt_estimator(struct sock *sk, long mrtt_us)
L
Linus Torvalds 已提交
719
{
720
	struct tcp_sock *tp = tcp_sk(sk);
721 722
	long m = mrtt_us; /* RTT */
	u32 srtt = tp->srtt_us;
L
Linus Torvalds 已提交
723 724 725 726

	/*	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.
727
	 *	This is designed to be as fast as possible
L
Linus Torvalds 已提交
728 729 730 731 732 733 734
	 *	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
735
	 * too slowly, when it should be increased quickly, decrease too quickly
L
Linus Torvalds 已提交
736 737 738 739
	 * 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)
	 */
740 741 742
	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
743 744
		if (m < 0) {
			m = -m;		/* m is now abs(error) */
745
			m -= (tp->mdev_us >> 2);   /* similar update on mdev */
L
Linus Torvalds 已提交
746 747 748 749 750 751 752 753 754 755 756
			/* 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 {
757
			m -= (tp->mdev_us >> 2);   /* similar update on mdev */
L
Linus Torvalds 已提交
758
		}
759 760 761 762 763
		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
764 765
		}
		if (after(tp->snd_una, tp->rtt_seq)) {
766 767
			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
768
			tp->rtt_seq = tp->snd_nxt;
769
			tp->mdev_max_us = tcp_rto_min_us(sk);
L
Linus Torvalds 已提交
770 771 772
		}
	} else {
		/* no previous measure. */
773
		srtt = m << 3;		/* take the measured time to be rtt */
774 775 776
		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
777 778
		tp->rtt_seq = tp->snd_nxt;
	}
779
	tp->srtt_us = max(1U, srtt);
L
Linus Torvalds 已提交
780 781
}

E
Eric Dumazet 已提交
782 783 784 785 786 787
/* 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)
 */
788 789 790
int sysctl_tcp_pacing_ss_ratio __read_mostly = 200;
int sysctl_tcp_pacing_ca_ratio __read_mostly = 120;

E
Eric Dumazet 已提交
791 792 793 794 795 796
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) */
797 798 799 800 801 802 803 804 805 806 807 808 809 810
	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
811 812 813

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

814 815
	if (likely(tp->srtt_us))
		do_div(rate, tp->srtt_us);
E
Eric Dumazet 已提交
816

817 818 819 820 821 822
	/* 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
823 824
}

L
Linus Torvalds 已提交
825 826 827
/* Calculate rto without backoff.  This is the second half of Van Jacobson's
 * routine referred to above.
 */
828
static void tcp_set_rto(struct sock *sk)
L
Linus Torvalds 已提交
829
{
830
	const struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
831 832 833 834 835 836 837 838
	/* 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
839
	 *    ACKs in some circumstances.
L
Linus Torvalds 已提交
840
	 */
841
	inet_csk(sk)->icsk_rto = __tcp_set_rto(tp);
L
Linus Torvalds 已提交
842 843 844 845

	/* 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
846
	 *    with correct one. It is exactly, which we pretend to do.
L
Linus Torvalds 已提交
847 848
	 */

849 850 851
	/* NOTE: clamping at TCP_RTO_MIN is not required, current algo
	 * guarantees that rto is higher.
	 */
852
	tcp_bound_rto(sk);
L
Linus Torvalds 已提交
853 854
}

855
__u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst)
L
Linus Torvalds 已提交
856 857 858
{
	__u32 cwnd = (dst ? dst_metric(dst, RTAX_INITCWND) : 0);

859
	if (!cwnd)
860
		cwnd = TCP_INIT_CWND;
L
Linus Torvalds 已提交
861 862 863
	return min_t(__u32, cwnd, tp->snd_cwnd_clamp);
}

864 865 866 867
/*
 * Packet counting of FACK is based on in-order assumptions, therefore TCP
 * disables it when reordering is detected
 */
868
void tcp_disable_fack(struct tcp_sock *tp)
869
{
870 871 872
	/* RFC3517 uses different metric in lost marker => reset on change */
	if (tcp_is_fack(tp))
		tp->lost_skb_hint = NULL;
873
	tp->rx_opt.sack_ok &= ~TCP_FACK_ENABLED;
874 875
}

876
/* Take a notice that peer is sending D-SACKs */
877 878
static void tcp_dsack_seen(struct tcp_sock *tp)
{
879
	tp->rx_opt.sack_ok |= TCP_DSACK_SEEN;
880 881
}

882 883
static void tcp_update_reordering(struct sock *sk, const int metric,
				  const int ts)
L
Linus Torvalds 已提交
884
{
885
	struct tcp_sock *tp = tcp_sk(sk);
886
	int mib_idx;
887

888
	if (metric > tp->reordering) {
889
		tp->reordering = min(sysctl_tcp_max_reordering, metric);
L
Linus Torvalds 已提交
890 891

#if FASTRETRANS_DEBUG > 1
892 893 894 895 896 897
		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
898
#endif
899
		tcp_disable_fack(tp);
L
Linus Torvalds 已提交
900
	}
Y
Yuchung Cheng 已提交
901

Y
Yuchung Cheng 已提交
902
	tp->rack.reord = 1;
903 904 905 906 907 908 909 910 911 912 913 914

	/* 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
915 916
}

917
/* This must be called before lost_out is incremented */
918 919
static void tcp_verify_retransmit_hint(struct tcp_sock *tp, struct sk_buff *skb)
{
920
	if (!tp->retransmit_skb_hint ||
921 922
	    before(TCP_SKB_CB(skb)->seq,
		   TCP_SKB_CB(tp->retransmit_skb_hint)->seq))
923
		tp->retransmit_skb_hint = skb;
924 925
}

926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941
/* 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);
}

942 943 944 945 946 947
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);
948
		tcp_sum_lost(tp, skb);
949 950 951 952
		TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
	}
}

Y
Yuchung Cheng 已提交
953
void tcp_skb_mark_lost_uncond_verify(struct tcp_sock *tp, struct sk_buff *skb)
954 955 956
{
	tcp_verify_retransmit_hint(tp, skb);

957
	tcp_sum_lost(tp, skb);
958 959 960 961 962 963
	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
964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984
/* 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())
985
 * 3. Loss detection event of two flavors:
L
Linus Torvalds 已提交
986 987
 *	A. Scoreboard estimator decided the packet is lost.
 *	   A'. Reno "three dupacks" marks head of queue lost.
988 989
 *	   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
990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
 *	   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
1009 1010 1011 1012 1013 1014 1015
 *
 * 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
1016 1017 1018 1019 1020 1021 1022 1023 1024
 * 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
1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
 *
 * 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
1047
 * again, D-SACK block must not to go across snd_una (for the same reason as
I
Ilpo Järvinen 已提交
1048 1049 1050 1051 1052 1053 1054 1055 1056
 * 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
1057
 */
E
Eric Dumazet 已提交
1058 1059
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
1060 1061 1062
{
	/* 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
1063
		return false;
I
Ilpo Järvinen 已提交
1064 1065 1066

	/* Nasty start_seq wrap-around check (see comments above) */
	if (!before(start_seq, tp->snd_nxt))
E
Eric Dumazet 已提交
1067
		return false;
I
Ilpo Järvinen 已提交
1068

1069
	/* In outstanding window? ...This is valid exit for D-SACKs too.
I
Ilpo Järvinen 已提交
1070 1071 1072
	 * start_seq == snd_una is non-sensical (see comments above)
	 */
	if (after(start_seq, tp->snd_una))
E
Eric Dumazet 已提交
1073
		return true;
I
Ilpo Järvinen 已提交
1074 1075

	if (!is_dsack || !tp->undo_marker)
E
Eric Dumazet 已提交
1076
		return false;
I
Ilpo Järvinen 已提交
1077 1078

	/* ...Then it's D-SACK, and must reside below snd_una completely */
Z
Zheng Yan 已提交
1079
	if (after(end_seq, tp->snd_una))
E
Eric Dumazet 已提交
1080
		return false;
I
Ilpo Järvinen 已提交
1081 1082

	if (!before(start_seq, tp->undo_marker))
E
Eric Dumazet 已提交
1083
		return true;
I
Ilpo Järvinen 已提交
1084 1085 1086

	/* Too old */
	if (!after(end_seq, tp->undo_marker))
E
Eric Dumazet 已提交
1087
		return false;
I
Ilpo Järvinen 已提交
1088 1089 1090 1091 1092 1093 1094

	/* 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
1095 1096 1097
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)
1098
{
1099
	struct tcp_sock *tp = tcp_sk(sk);
1100 1101
	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
1102
	bool dup_sack = false;
1103 1104

	if (before(start_seq_0, TCP_SKB_CB(ack_skb)->ack_seq)) {
E
Eric Dumazet 已提交
1105
		dup_sack = true;
1106
		tcp_dsack_seen(tp);
1107
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPDSACKRECV);
1108
	} else if (num_sacks > 1) {
1109 1110
		u32 end_seq_1 = get_unaligned_be32(&sp[1].end_seq);
		u32 start_seq_1 = get_unaligned_be32(&sp[1].start_seq);
1111 1112 1113

		if (!after(end_seq_0, end_seq_1) &&
		    !before(start_seq_0, start_seq_1)) {
E
Eric Dumazet 已提交
1114
			dup_sack = true;
1115
			tcp_dsack_seen(tp);
1116
			NET_INC_STATS(sock_net(sk),
1117
					LINUX_MIB_TCPDSACKOFORECV);
1118 1119 1120 1121
		}
	}

	/* D-SACK for already forgotten data... Do dumb counting. */
Y
Yuchung Cheng 已提交
1122
	if (dup_sack && tp->undo_marker && tp->undo_retrans > 0 &&
1123 1124 1125 1126 1127 1128 1129
	    !after(end_seq_0, prior_snd_una) &&
	    after(end_seq_0, tp->undo_marker))
		tp->undo_retrans--;

	return dup_sack;
}

1130
struct tcp_sacktag_state {
1131 1132
	int	reord;
	int	fack_count;
1133 1134 1135 1136 1137 1138
	/* 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;
1139
	struct rate_sample *rate;
1140
	int	flag;
1141 1142
};

1143 1144 1145 1146 1147
/* 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).
1148 1149
 *
 * FIXME: this could be merged to shift decision code
1150
 */
1151
static int tcp_match_skb_to_sack(struct sock *sk, struct sk_buff *skb,
E
Eric Dumazet 已提交
1152
				  u32 start_seq, u32 end_seq)
1153
{
E
Eric Dumazet 已提交
1154 1155
	int err;
	bool in_sack;
1156
	unsigned int pkt_len;
1157
	unsigned int mss;
1158 1159 1160 1161 1162 1163

	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)) {
1164
		mss = tcp_skb_mss(skb);
1165 1166
		in_sack = !after(start_seq, TCP_SKB_CB(skb)->seq);

1167
		if (!in_sack) {
1168
			pkt_len = start_seq - TCP_SKB_CB(skb)->seq;
1169 1170 1171
			if (pkt_len < mss)
				pkt_len = mss;
		} else {
1172
			pkt_len = end_seq - TCP_SKB_CB(skb)->seq;
1173 1174 1175 1176 1177 1178 1179 1180 1181
			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;
1182
			if (!in_sack && new_len < pkt_len)
1183 1184 1185
				new_len += mss;
			pkt_len = new_len;
		}
1186 1187 1188 1189

		if (pkt_len >= skb->len && !in_sack)
			return 0;

1190
		err = tcp_fragment(sk, skb, pkt_len, mss, GFP_ATOMIC);
1191 1192 1193 1194 1195 1196 1197
		if (err < 0)
			return err;
	}

	return in_sack;
}

1198 1199 1200 1201
/* 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,
1202 1203
			  int dup_sack, int pcount,
			  const struct skb_mstamp *xmit_time)
1204
{
1205
	struct tcp_sock *tp = tcp_sk(sk);
1206
	int fack_count = state->fack_count;
1207 1208 1209

	/* Account D-SACK for retransmitted packet. */
	if (dup_sack && (sacked & TCPCB_RETRANS)) {
Y
Yuchung Cheng 已提交
1210
		if (tp->undo_marker && tp->undo_retrans > 0 &&
1211
		    after(end_seq, tp->undo_marker))
1212
			tp->undo_retrans--;
1213
		if (sacked & TCPCB_SACKED_ACKED)
1214
			state->reord = min(fack_count, state->reord);
1215 1216 1217
	}

	/* Nothing to do; acked frame is about to be dropped (was ACKed). */
1218
	if (!after(end_seq, tp->snd_una))
1219
		return sacked;
1220 1221

	if (!(sacked & TCPCB_SACKED_ACKED)) {
1222
		tcp_rack_advance(tp, sacked, end_seq, xmit_time);
1223

1224 1225 1226 1227 1228 1229
		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) {
1230
				sacked &= ~(TCPCB_LOST|TCPCB_SACKED_RETRANS);
1231 1232
				tp->lost_out -= pcount;
				tp->retrans_out -= pcount;
1233 1234 1235 1236 1237 1238
			}
		} else {
			if (!(sacked & TCPCB_RETRANS)) {
				/* New sack for not retransmitted frame,
				 * which was in hole. It is reordering.
				 */
1239
				if (before(start_seq,
1240
					   tcp_highest_sack_seq(tp)))
1241 1242
					state->reord = min(fack_count,
							   state->reord);
Y
Yuchung Cheng 已提交
1243 1244
				if (!after(end_seq, tp->high_seq))
					state->flag |= FLAG_ORIG_SACK_ACKED;
1245 1246 1247
				if (state->first_sackt.v64 == 0)
					state->first_sackt = *xmit_time;
				state->last_sackt = *xmit_time;
1248 1249 1250
			}

			if (sacked & TCPCB_LOST) {
1251
				sacked &= ~TCPCB_LOST;
1252
				tp->lost_out -= pcount;
1253 1254 1255
			}
		}

1256 1257
		sacked |= TCPCB_SACKED_ACKED;
		state->flag |= FLAG_DATA_SACKED;
1258
		tp->sacked_out += pcount;
Y
Yuchung Cheng 已提交
1259
		tp->delivered += pcount;  /* Out-of-order packets delivered */
1260

1261
		fack_count += pcount;
1262 1263

		/* Lost marker hint past SACKed? Tweak RFC3517 cnt */
1264
		if (!tcp_is_fack(tp) && tp->lost_skb_hint &&
1265
		    before(start_seq, TCP_SKB_CB(tp->lost_skb_hint)->seq))
1266
			tp->lost_cnt_hint += pcount;
1267 1268 1269 1270 1271 1272 1273 1274 1275

		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.
	 */
1276 1277
	if (dup_sack && (sacked & TCPCB_SACKED_RETRANS)) {
		sacked &= ~TCPCB_SACKED_RETRANS;
1278
		tp->retrans_out -= pcount;
1279 1280
	}

1281
	return sacked;
1282 1283
}

1284 1285 1286
/* 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
1287 1288 1289 1290
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)
1291 1292
{
	struct tcp_sock *tp = tcp_sk(sk);
1293
	struct sk_buff *prev = tcp_write_queue_prev(sk, skb);
1294 1295
	u32 start_seq = TCP_SKB_CB(skb)->seq;	/* start of newly-SACKed */
	u32 end_seq = start_seq + shifted;	/* end of newly-SACKed */
1296 1297 1298

	BUG_ON(!pcount);

1299 1300 1301 1302 1303 1304 1305
	/* 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
1306
			start_seq, end_seq, dup_sack, pcount,
1307
			&skb->skb_mstamp);
1308
	tcp_rate_skb_delivered(sk, skb, state->rate);
1309 1310

	if (skb == tp->lost_skb_hint)
1311 1312
		tp->lost_cnt_hint += pcount;

1313 1314 1315
	TCP_SKB_CB(prev)->end_seq += shifted;
	TCP_SKB_CB(skb)->seq += shifted;

1316 1317 1318
	tcp_skb_pcount_add(prev, pcount);
	BUG_ON(tcp_skb_pcount(skb) < pcount);
	tcp_skb_pcount_add(skb, -pcount);
1319 1320 1321 1322 1323 1324

	/* 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.
	 */
1325 1326
	if (!TCP_SKB_CB(prev)->tcp_gso_size)
		TCP_SKB_CB(prev)->tcp_gso_size = mss;
1327 1328

	/* CHECKME: To clear or not to clear? Mimics normal skb currently */
1329
	if (tcp_skb_pcount(skb) <= 1)
1330
		TCP_SKB_CB(skb)->tcp_gso_size = 0;
1331 1332 1333 1334 1335 1336

	/* 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));
1337
		NET_INC_STATS(sock_net(sk), LINUX_MIB_SACKSHIFTED);
E
Eric Dumazet 已提交
1338
		return false;
1339 1340 1341 1342
	}

	/* Whole SKB was eaten :-) */

1343 1344 1345 1346 1347 1348 1349
	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);
	}

1350
	TCP_SKB_CB(prev)->tcp_flags |= TCP_SKB_CB(skb)->tcp_flags;
1351
	TCP_SKB_CB(prev)->eor = TCP_SKB_CB(skb)->eor;
1352 1353 1354
	if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)
		TCP_SKB_CB(prev)->end_seq++;

1355 1356 1357
	if (skb == tcp_highest_sack(sk))
		tcp_advance_highest_sack(sk, skb);

1358
	tcp_skb_collapse_tstamp(prev, skb);
1359 1360 1361
	if (unlikely(TCP_SKB_CB(prev)->tx.delivered_mstamp.v64))
		TCP_SKB_CB(prev)->tx.delivered_mstamp.v64 = 0;

1362 1363 1364
	tcp_unlink_write_queue(skb, sk);
	sk_wmem_free_skb(sk, skb);

1365
	NET_INC_STATS(sock_net(sk), LINUX_MIB_SACKMERGED);
1366

E
Eric Dumazet 已提交
1367
	return true;
1368 1369 1370 1371 1372
}

/* I wish gso_size would have a bit more sane initialization than
 * something-or-zero which complicates things
 */
1373
static int tcp_skb_seglen(const struct sk_buff *skb)
1374
{
1375
	return tcp_skb_pcount(skb) == 1 ? skb->len : tcp_skb_mss(skb);
1376 1377 1378
}

/* Shifting pages past head area doesn't work */
1379
static int skb_can_shift(const struct sk_buff *skb)
1380 1381 1382 1383 1384 1385 1386 1387
{
	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,
1388
					  struct tcp_sacktag_state *state,
1389
					  u32 start_seq, u32 end_seq,
E
Eric Dumazet 已提交
1390
					  bool dup_sack)
1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403
{
	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 &&
1404
	    (TCP_SKB_CB(skb)->sacked & (TCPCB_LOST|TCPCB_SACKED_RETRANS)) == TCPCB_SACKED_RETRANS)
1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
		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;

1420 1421 1422
	if (!tcp_skb_can_collapse_to(prev))
		goto fallback;

1423 1424 1425 1426 1427 1428
	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);
1429
		mss = tcp_skb_seglen(skb);
1430 1431 1432 1433

		/* TODO: Fix DSACKs to not fragment already SACKed and we can
		 * drop this restriction as unnecessary
		 */
1434
		if (mss != tcp_skb_seglen(prev))
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 1469 1470 1471 1472 1473 1474
			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
		 */
1475
		if (mss != tcp_skb_seglen(prev))
1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487
			goto fallback;

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

1488 1489 1490 1491
	/* tcp_sacktag_one() won't SACK-tag ranges below snd_una */
	if (!after(TCP_SKB_CB(skb)->seq + len, tp->snd_una))
		goto fallback;

1492 1493
	if (!skb_shift(prev, skb, len))
		goto fallback;
I
Ilpo Järvinen 已提交
1494
	if (!tcp_shifted_skb(sk, skb, state, pcount, len, mss, dup_sack))
1495 1496 1497 1498 1499 1500 1501 1502 1503
		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);

1504 1505 1506
	if (!skb_can_shift(skb) ||
	    (skb == tcp_send_head(sk)) ||
	    ((TCP_SKB_CB(skb)->sacked & TCPCB_TAGBITS) != TCPCB_SACKED_ACKED) ||
1507
	    (mss != tcp_skb_seglen(skb)))
1508 1509 1510 1511 1512
		goto out;

	len = skb->len;
	if (skb_shift(prev, skb, len)) {
		pcount += tcp_skb_pcount(skb);
I
Ilpo Järvinen 已提交
1513
		tcp_shifted_skb(sk, skb, state, tcp_skb_pcount(skb), len, mss, 0);
1514 1515 1516
	}

out:
1517
	state->fack_count += pcount;
1518 1519 1520 1521 1522 1523
	return prev;

noop:
	return skb;

fallback:
1524
	NET_INC_STATS(sock_net(sk), LINUX_MIB_SACKSHIFTFALLBACK);
1525 1526 1527
	return NULL;
}

1528 1529
static struct sk_buff *tcp_sacktag_walk(struct sk_buff *skb, struct sock *sk,
					struct tcp_sack_block *next_dup,
1530
					struct tcp_sacktag_state *state,
1531
					u32 start_seq, u32 end_seq,
E
Eric Dumazet 已提交
1532
					bool dup_sack_in)
1533
{
1534 1535 1536
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *tmp;

1537 1538
	tcp_for_write_queue_from(skb, sk) {
		int in_sack = 0;
E
Eric Dumazet 已提交
1539
		bool dup_sack = dup_sack_in;
1540 1541 1542 1543 1544 1545 1546 1547

		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;

1548
		if (next_dup  &&
1549 1550 1551 1552 1553
		    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
1554
				dup_sack = true;
1555 1556
		}

1557 1558 1559 1560 1561
		/* 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) {
1562 1563
			tmp = tcp_shift_skb_data(sk, skb, state,
						 start_seq, end_seq, dup_sack);
1564
			if (tmp) {
1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577
				if (tmp != skb) {
					skb = tmp;
					continue;
				}

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

1578 1579 1580
		if (unlikely(in_sack < 0))
			break;

1581
		if (in_sack) {
1582 1583 1584 1585 1586 1587 1588
			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
1589
						tcp_skb_pcount(skb),
1590
						&skb->skb_mstamp);
1591
			tcp_rate_skb_delivered(sk, skb, state->rate);
1592

1593 1594 1595 1596 1597
			if (!before(TCP_SKB_CB(skb)->seq,
				    tcp_highest_sack_seq(tp)))
				tcp_advance_highest_sack(sk, skb);
		}

1598
		state->fack_count += tcp_skb_pcount(skb);
1599 1600 1601 1602 1603 1604 1605 1606
	}
	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,
1607 1608
					struct tcp_sacktag_state *state,
					u32 skip_to_seq)
1609 1610 1611 1612 1613
{
	tcp_for_write_queue_from(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;

I
Ilpo Järvinen 已提交
1614
		if (after(TCP_SKB_CB(skb)->end_seq, skip_to_seq))
1615
			break;
1616

1617
		state->fack_count += tcp_skb_pcount(skb);
1618 1619 1620 1621 1622 1623 1624
	}
	return skb;
}

static struct sk_buff *tcp_maybe_skipping_dsack(struct sk_buff *skb,
						struct sock *sk,
						struct tcp_sack_block *next_dup,
1625 1626
						struct tcp_sacktag_state *state,
						u32 skip_to_seq)
1627
{
1628
	if (!next_dup)
1629 1630 1631
		return skb;

	if (before(next_dup->start_seq, skip_to_seq)) {
1632 1633 1634 1635
		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);
1636 1637 1638 1639 1640
	}

	return skb;
}

1641
static int tcp_sack_cache_ok(const struct tcp_sock *tp, const struct tcp_sack_block *cache)
1642 1643 1644 1645
{
	return cache < tp->recv_sack_cache + ARRAY_SIZE(tp->recv_sack_cache);
}

L
Linus Torvalds 已提交
1646
static int
1647
tcp_sacktag_write_queue(struct sock *sk, const struct sk_buff *ack_skb,
1648
			u32 prior_snd_una, struct tcp_sacktag_state *state)
L
Linus Torvalds 已提交
1649 1650
{
	struct tcp_sock *tp = tcp_sk(sk);
1651 1652
	const unsigned char *ptr = (skb_transport_header(ack_skb) +
				    TCP_SKB_CB(ack_skb)->sacked);
1653
	struct tcp_sack_block_wire *sp_wire = (struct tcp_sack_block_wire *)(ptr+2);
1654
	struct tcp_sack_block sp[TCP_NUM_SACKS];
1655 1656
	struct tcp_sack_block *cache;
	struct sk_buff *skb;
1657
	int num_sacks = min(TCP_NUM_SACKS, (ptr[1] - TCPOLEN_SACK_BASE) >> 3);
1658
	int used_sacks;
E
Eric Dumazet 已提交
1659
	bool found_dup_sack = false;
1660
	int i, j;
1661
	int first_sack_index;
L
Linus Torvalds 已提交
1662

1663 1664
	state->flag = 0;
	state->reord = tp->packets_out;
1665

1666
	if (!tp->sacked_out) {
1667 1668
		if (WARN_ON(tp->fackets_out))
			tp->fackets_out = 0;
1669
		tcp_highest_sack_reset(sk);
1670
	}
L
Linus Torvalds 已提交
1671

1672
	found_dup_sack = tcp_check_dsack(sk, ack_skb, sp_wire,
1673
					 num_sacks, prior_snd_una);
1674
	if (found_dup_sack) {
1675
		state->flag |= FLAG_DSACKING_ACK;
1676 1677
		tp->delivered++; /* A spurious retransmission is delivered */
	}
1678 1679 1680 1681 1682 1683 1684 1685

	/* 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;

1686 1687 1688
	if (!tp->packets_out)
		goto out;

1689 1690 1691
	used_sacks = 0;
	first_sack_index = 0;
	for (i = 0; i < num_sacks; i++) {
E
Eric Dumazet 已提交
1692
		bool dup_sack = !i && found_dup_sack;
1693

1694 1695
		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);
1696 1697 1698 1699

		if (!tcp_is_sackblock_valid(tp, dup_sack,
					    sp[used_sacks].start_seq,
					    sp[used_sacks].end_seq)) {
1700 1701
			int mib_idx;

1702 1703
			if (dup_sack) {
				if (!tp->undo_marker)
1704
					mib_idx = LINUX_MIB_TCPDSACKIGNOREDNOUNDO;
1705
				else
1706
					mib_idx = LINUX_MIB_TCPDSACKIGNOREDOLD;
1707 1708 1709 1710 1711
			} 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;
1712
				mib_idx = LINUX_MIB_TCPSACKDISCARD;
1713
			}
1714

1715
			NET_INC_STATS(sock_net(sk), mib_idx);
1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727
			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++;
	}

1728 1729
	/* order SACK blocks to allow in order walk of the retrans queue */
	for (i = used_sacks - 1; i > 0; i--) {
1730 1731
		for (j = 0; j < i; j++) {
			if (after(sp[j].start_seq, sp[j + 1].start_seq)) {
1732
				swap(sp[j], sp[j + 1]);
1733

1734 1735
				/* Track where the first SACK block goes to */
				if (j == first_sack_index)
1736
					first_sack_index = j + 1;
1737 1738 1739 1740
			}
		}
	}

1741
	skb = tcp_write_queue_head(sk);
1742
	state->fack_count = 0;
1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753
	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++;
1754 1755
	}

1756
	while (i < used_sacks) {
1757 1758
		u32 start_seq = sp[i].start_seq;
		u32 end_seq = sp[i].end_seq;
E
Eric Dumazet 已提交
1759
		bool dup_sack = (found_dup_sack && (i == first_sack_index));
1760
		struct tcp_sack_block *next_dup = NULL;
1761

1762 1763
		if (found_dup_sack && ((i + 1) == first_sack_index))
			next_dup = &sp[i + 1];
L
Linus Torvalds 已提交
1764

1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775
		/* 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)) {
1776
				skb = tcp_sacktag_skip(skb, sk, state,
1777
						       start_seq);
1778
				skb = tcp_sacktag_walk(skb, sk, next_dup,
1779
						       state,
1780 1781
						       start_seq,
						       cache->start_seq,
1782
						       dup_sack);
1783
			}
1784

1785
			/* Rest of the block already fully processed? */
1786
			if (!after(end_seq, cache->end_seq))
1787
				goto advance_sp;
1788

1789
			skb = tcp_maybe_skipping_dsack(skb, sk, next_dup,
1790
						       state,
1791
						       cache->end_seq);
1792

1793
			/* ...tail remains todo... */
1794
			if (tcp_highest_sack_seq(tp) == cache->end_seq) {
1795
				/* ...but better entrypoint exists! */
1796
				skb = tcp_highest_sack(sk);
1797
				if (!skb)
1798
					break;
1799
				state->fack_count = tp->fackets_out;
1800 1801
				cache++;
				goto walk;
1802 1803
			}

1804
			skb = tcp_sacktag_skip(skb, sk, state, cache->end_seq);
1805 1806 1807 1808
			/* Check overlap against next cached too (past this one already) */
			cache++;
			continue;
		}
L
Linus Torvalds 已提交
1809

1810 1811
		if (!before(start_seq, tcp_highest_sack_seq(tp))) {
			skb = tcp_highest_sack(sk);
1812
			if (!skb)
1813
				break;
1814
			state->fack_count = tp->fackets_out;
L
Linus Torvalds 已提交
1815
		}
1816
		skb = tcp_sacktag_skip(skb, sk, state, start_seq);
1817 1818

walk:
1819
		skb = tcp_sacktag_walk(skb, sk, next_dup, state,
1820
				       start_seq, end_seq, dup_sack);
1821

1822 1823
advance_sp:
		i++;
L
Linus Torvalds 已提交
1824 1825
	}

1826 1827 1828 1829 1830 1831 1832 1833
	/* 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];

1834
	if ((state->reord < tp->fackets_out) &&
Y
Yuchung Cheng 已提交
1835
	    ((inet_csk(sk)->icsk_ca_state != TCP_CA_Loss) || tp->undo_marker))
1836
		tcp_update_reordering(sk, tp->fackets_out - state->reord, 0);
L
Linus Torvalds 已提交
1837

1838
	tcp_verify_left_out(tp);
1839 1840
out:

L
Linus Torvalds 已提交
1841
#if FASTRETRANS_DEBUG > 0
1842 1843 1844 1845
	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
1846
#endif
1847
	return state->flag;
L
Linus Torvalds 已提交
1848 1849
}

1850
/* Limits sacked_out so that sum with lost_out isn't ever larger than
E
Eric Dumazet 已提交
1851
 * packets_out. Returns false if sacked_out adjustement wasn't necessary.
1852
 */
E
Eric Dumazet 已提交
1853
static bool tcp_limit_reno_sacked(struct tcp_sock *tp)
1854 1855 1856 1857 1858 1859 1860 1861
{
	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
1862
		return true;
1863
	}
E
Eric Dumazet 已提交
1864
	return false;
1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875
}

/* 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);
1876 1877 1878 1879 1880 1881 1882
}

/* 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
1883 1884
	u32 prior_sacked = tp->sacked_out;

1885 1886
	tp->sacked_out++;
	tcp_check_reno_reordering(sk, 0);
Y
Yuchung Cheng 已提交
1887 1888
	if (tp->sacked_out > prior_sacked)
		tp->delivered++; /* Some out-of-order packet is delivered */
1889
	tcp_verify_left_out(tp);
1890 1891 1892 1893 1894 1895 1896 1897 1898 1899
}

/* 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
1900
		tp->delivered += max_t(int, acked - tp->sacked_out, 1);
1901
		if (acked - 1 >= tp->sacked_out)
1902 1903
			tp->sacked_out = 0;
		else
1904
			tp->sacked_out -= acked - 1;
1905 1906
	}
	tcp_check_reno_reordering(sk, acked);
1907
	tcp_verify_left_out(tp);
1908 1909 1910 1911 1912 1913 1914
}

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

Y
Yuchung Cheng 已提交
1915
void tcp_clear_retrans(struct tcp_sock *tp)
L
Linus Torvalds 已提交
1916 1917 1918 1919
{
	tp->retrans_out = 0;
	tp->lost_out = 0;
	tp->undo_marker = 0;
Y
Yuchung Cheng 已提交
1920
	tp->undo_retrans = -1;
Y
Yuchung Cheng 已提交
1921 1922
	tp->fackets_out = 0;
	tp->sacked_out = 0;
L
Linus Torvalds 已提交
1923 1924
}

Y
Yuchung Cheng 已提交
1925
static inline void tcp_init_undo(struct tcp_sock *tp)
1926
{
Y
Yuchung Cheng 已提交
1927 1928 1929
	tp->undo_marker = tp->snd_una;
	/* Retransmission still in flight may cause DSACKs later. */
	tp->undo_retrans = tp->retrans_out ? : -1;
1930 1931
}

1932
/* Enter Loss state. If we detect SACK reneging, forget all SACK information
L
Linus Torvalds 已提交
1933 1934 1935
 * and reset tags completely, otherwise preserve SACKs. If receiver
 * dropped its ofo queue, we will know this due to reneging detection.
 */
1936
void tcp_enter_loss(struct sock *sk)
L
Linus Torvalds 已提交
1937
{
1938
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1939
	struct tcp_sock *tp = tcp_sk(sk);
1940
	struct net *net = sock_net(sk);
L
Linus Torvalds 已提交
1941
	struct sk_buff *skb;
1942
	bool new_recovery = icsk->icsk_ca_state < TCP_CA_Recovery;
1943
	bool is_reneg;			/* is receiver reneging on SACKs? */
1944
	bool mark_lost;
L
Linus Torvalds 已提交
1945 1946

	/* Reduce ssthresh if it has not yet been made inside this window. */
Y
Yuchung Cheng 已提交
1947 1948
	if (icsk->icsk_ca_state <= TCP_CA_Disorder ||
	    !after(tp->high_seq, tp->snd_una) ||
1949 1950 1951 1952
	    (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
1953
		tcp_init_undo(tp);
L
Linus Torvalds 已提交
1954 1955 1956 1957 1958
	}
	tp->snd_cwnd	   = 1;
	tp->snd_cwnd_cnt   = 0;
	tp->snd_cwnd_stamp = tcp_time_stamp;

Y
Yuchung Cheng 已提交
1959 1960
	tp->retrans_out = 0;
	tp->lost_out = 0;
1961 1962 1963

	if (tcp_is_reno(tp))
		tcp_reset_reno_sack(tp);
L
Linus Torvalds 已提交
1964

1965 1966 1967
	skb = tcp_write_queue_head(sk);
	is_reneg = skb && (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED);
	if (is_reneg) {
1968
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPSACKRENEGING);
1969 1970
		tp->sacked_out = 0;
		tp->fackets_out = 0;
1971
	}
1972
	tcp_clear_all_retrans_hints(tp);
L
Linus Torvalds 已提交
1973

1974 1975 1976
	tcp_for_write_queue(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
1977

1978 1979 1980 1981
		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
1982
		TCP_SKB_CB(skb)->sacked &= (~TCPCB_TAGBITS)|TCPCB_SACKED_ACKED;
1983
		if (mark_lost) {
L
Linus Torvalds 已提交
1984 1985 1986 1987 1988
			TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_ACKED;
			TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
			tp->lost_out += tcp_skb_pcount(skb);
		}
	}
1989
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
1990

1991 1992 1993 1994
	/* 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 &&
1995
	    tp->sacked_out >= net->ipv4.sysctl_tcp_reordering)
1996
		tp->reordering = min_t(unsigned int, tp->reordering,
1997
				       net->ipv4.sysctl_tcp_reordering);
1998
	tcp_set_ca_state(sk, TCP_CA_Loss);
L
Linus Torvalds 已提交
1999
	tp->high_seq = tp->snd_nxt;
2000
	tcp_ecn_queue_cwr(tp);
Y
Yuchung Cheng 已提交
2001

2002 2003 2004 2005 2006 2007 2008 2009
	/* 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
2010
	 */
2011 2012 2013
	tp->frto = sysctl_tcp_frto &&
		   (new_recovery || icsk->icsk_retransmits) &&
		   !inet_csk(sk)->icsk_mtup.probe_size;
L
Linus Torvalds 已提交
2014 2015
}

2016 2017 2018 2019
/* 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).
 *
2020 2021 2022 2023 2024
 * 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.
2025
 */
E
Eric Dumazet 已提交
2026
static bool tcp_check_sack_reneging(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2027
{
2028
	if (flag & FLAG_SACK_RENEGING) {
2029 2030 2031
		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
2032

2033
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
2034
					  delay, TCP_RTO_MAX);
E
Eric Dumazet 已提交
2035
		return true;
L
Linus Torvalds 已提交
2036
	}
E
Eric Dumazet 已提交
2037
	return false;
L
Linus Torvalds 已提交
2038 2039
}

2040
static inline int tcp_fackets_out(const struct tcp_sock *tp)
L
Linus Torvalds 已提交
2041
{
2042
	return tcp_is_reno(tp) ? tp->sacked_out + 1 : tp->fackets_out;
L
Linus Torvalds 已提交
2043 2044
}

2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059
/* 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.
 */
2060
static inline int tcp_dupack_heuristics(const struct tcp_sock *tp)
2061 2062 2063 2064
{
	return tcp_is_fack(tp) ? tp->fackets_out : tp->sacked_out + 1;
}

L
Linus Torvalds 已提交
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 2106 2107 2108 2109 2110 2111
/* 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.
 *
2112
 *		Essentially, we have now a few algorithms detecting
L
Linus Torvalds 已提交
2113 2114
 *		lost packets.
 *
2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129
 *		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
2130 2131
 *		FACK (Disabled by default. Subsumbed by RACK):
 *		It is the simplest heuristics. As soon as we decided
L
Linus Torvalds 已提交
2132 2133 2134 2135 2136 2137 2138 2139
 *		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.
 *
2140 2141 2142
 *		If the receiver does not support SACK:
 *
 *		NewReno (RFC6582): in Recovery we assume that one segment
L
Linus Torvalds 已提交
2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171
 *		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
2172
static bool tcp_time_to_recover(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2173
{
2174
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2175 2176 2177

	/* Trick#1: The loss is proven. */
	if (tp->lost_out)
E
Eric Dumazet 已提交
2178
		return true;
L
Linus Torvalds 已提交
2179 2180

	/* Not-A-Trick#2 : Classic rule... */
2181
	if (tcp_dupack_heuristics(tp) > tp->reordering)
E
Eric Dumazet 已提交
2182
		return true;
L
Linus Torvalds 已提交
2183

E
Eric Dumazet 已提交
2184
	return false;
L
Linus Torvalds 已提交
2185 2186
}

2187 2188 2189 2190 2191
/* 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.
2192
 */
I
Ilpo Järvinen 已提交
2193
static void tcp_mark_head_lost(struct sock *sk, int packets, int mark_head)
L
Linus Torvalds 已提交
2194
{
2195
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2196
	struct sk_buff *skb;
2197
	int cnt, oldcnt, lost;
2198
	unsigned int mss;
2199 2200
	/* 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
2201

2202
	WARN_ON(packets > tp->packets_out);
2203 2204 2205
	if (tp->lost_skb_hint) {
		skb = tp->lost_skb_hint;
		cnt = tp->lost_cnt_hint;
I
Ilpo Järvinen 已提交
2206 2207 2208
		/* Head already handled? */
		if (mark_head && skb != tcp_write_queue_head(sk))
			return;
2209
	} else {
2210
		skb = tcp_write_queue_head(sk);
2211 2212
		cnt = 0;
	}
L
Linus Torvalds 已提交
2213

2214 2215 2216
	tcp_for_write_queue_from(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
2217 2218 2219 2220
		/* TODO: do this better */
		/* this is not the most efficient way to do this... */
		tp->lost_skb_hint = skb;
		tp->lost_cnt_hint = cnt;
2221

2222
		if (after(TCP_SKB_CB(skb)->end_seq, loss_high))
2223 2224 2225
			break;

		oldcnt = cnt;
2226
		if (tcp_is_fack(tp) || tcp_is_reno(tp) ||
2227 2228 2229
		    (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
			cnt += tcp_skb_pcount(skb);

2230
		if (cnt > packets) {
2231
			if ((tcp_is_sack(tp) && !tcp_is_fack(tp)) ||
2232
			    (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED) ||
2233
			    (oldcnt >= packets))
2234 2235
				break;

2236
			mss = tcp_skb_mss(skb);
2237 2238 2239 2240
			/* 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)
2241 2242 2243 2244
				break;
			cnt = packets;
		}

2245
		tcp_skb_mark_lost(tp, skb);
I
Ilpo Järvinen 已提交
2246 2247 2248

		if (mark_head)
			break;
L
Linus Torvalds 已提交
2249
	}
2250
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2251 2252 2253 2254
}

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

2255
static void tcp_update_scoreboard(struct sock *sk, int fast_rexmit)
L
Linus Torvalds 已提交
2256
{
2257 2258
	struct tcp_sock *tp = tcp_sk(sk);

2259
	if (tcp_is_reno(tp)) {
I
Ilpo Järvinen 已提交
2260
		tcp_mark_head_lost(sk, 1, 1);
2261
	} else if (tcp_is_fack(tp)) {
L
Linus Torvalds 已提交
2262 2263 2264
		int lost = tp->fackets_out - tp->reordering;
		if (lost <= 0)
			lost = 1;
I
Ilpo Järvinen 已提交
2265
		tcp_mark_head_lost(sk, lost, 0);
L
Linus Torvalds 已提交
2266
	} else {
2267
		int sacked_upto = tp->sacked_out - tp->reordering;
I
Ilpo Järvinen 已提交
2268 2269 2270 2271
		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
2272 2273 2274
	}
}

2275 2276 2277 2278 2279 2280
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);
}

2281 2282 2283 2284 2285 2286 2287 2288 2289 2290
/* 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
2291 2292 2293
/* Nothing was retransmitted or returned timestamp is less
 * than timestamp of the first retransmission.
 */
2294
static inline bool tcp_packet_delayed(const struct tcp_sock *tp)
L
Linus Torvalds 已提交
2295 2296
{
	return !tp->retrans_stamp ||
2297
	       tcp_tsopt_ecr_before(tp, tp->retrans_stamp);
L
Linus Torvalds 已提交
2298 2299 2300 2301
}

/* Undo procedures. */

2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330
/* 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
2331
#if FASTRETRANS_DEBUG > 1
2332
static void DBGUNDO(struct sock *sk, const char *msg)
L
Linus Torvalds 已提交
2333
{
2334
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2335
	struct inet_sock *inet = inet_sk(sk);
2336

2337
	if (sk->sk_family == AF_INET) {
2338 2339 2340 2341 2342 2343
		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);
2344
	}
E
Eric Dumazet 已提交
2345
#if IS_ENABLED(CONFIG_IPV6)
2346
	else if (sk->sk_family == AF_INET6) {
2347 2348
		pr_debug("Undo %s %pI6/%u c%u l%u ss%u/%u p%u\n",
			 msg,
2349
			 &sk->sk_v6_daddr, ntohs(inet->inet_dport),
2350 2351 2352
			 tp->snd_cwnd, tcp_left_out(tp),
			 tp->snd_ssthresh, tp->prior_ssthresh,
			 tp->packets_out);
2353 2354
	}
#endif
L
Linus Torvalds 已提交
2355 2356 2357 2358 2359
}
#else
#define DBGUNDO(x...) do { } while (0)
#endif

2360
static void tcp_undo_cwnd_reduction(struct sock *sk, bool unmark_loss)
L
Linus Torvalds 已提交
2361
{
2362 2363
	struct tcp_sock *tp = tcp_sk(sk);

Y
Yuchung Cheng 已提交
2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375
	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
2376
	if (tp->prior_ssthresh) {
2377 2378
		const struct inet_connection_sock *icsk = inet_csk(sk);

2379
		tp->snd_cwnd = icsk->icsk_ca_ops->undo_cwnd(sk);
L
Linus Torvalds 已提交
2380

2381
		if (tp->prior_ssthresh > tp->snd_ssthresh) {
L
Linus Torvalds 已提交
2382
			tp->snd_ssthresh = tp->prior_ssthresh;
2383
			tcp_ecn_withdraw_cwr(tp);
L
Linus Torvalds 已提交
2384 2385 2386
		}
	}
	tp->snd_cwnd_stamp = tcp_time_stamp;
2387
	tp->undo_marker = 0;
L
Linus Torvalds 已提交
2388 2389
}

2390
static inline bool tcp_may_undo(const struct tcp_sock *tp)
L
Linus Torvalds 已提交
2391
{
2392
	return tp->undo_marker && (!tp->undo_retrans || tcp_packet_delayed(tp));
L
Linus Torvalds 已提交
2393 2394 2395
}

/* People celebrate: "We love our President!" */
E
Eric Dumazet 已提交
2396
static bool tcp_try_undo_recovery(struct sock *sk)
L
Linus Torvalds 已提交
2397
{
2398 2399
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2400
	if (tcp_may_undo(tp)) {
2401 2402
		int mib_idx;

L
Linus Torvalds 已提交
2403 2404 2405
		/* Happy end! We did not retransmit anything
		 * or our original transmission succeeded.
		 */
2406
		DBGUNDO(sk, inet_csk(sk)->icsk_ca_state == TCP_CA_Loss ? "loss" : "retrans");
2407
		tcp_undo_cwnd_reduction(sk, false);
2408
		if (inet_csk(sk)->icsk_ca_state == TCP_CA_Loss)
2409
			mib_idx = LINUX_MIB_TCPLOSSUNDO;
L
Linus Torvalds 已提交
2410
		else
2411 2412
			mib_idx = LINUX_MIB_TCPFULLUNDO;

2413
		NET_INC_STATS(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
2414
	}
2415
	if (tp->snd_una == tp->high_seq && tcp_is_reno(tp)) {
L
Linus Torvalds 已提交
2416 2417 2418
		/* 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. */
2419 2420
		if (!tcp_any_retrans_done(sk))
			tp->retrans_stamp = 0;
E
Eric Dumazet 已提交
2421
		return true;
L
Linus Torvalds 已提交
2422
	}
2423
	tcp_set_ca_state(sk, TCP_CA_Open);
E
Eric Dumazet 已提交
2424
	return false;
L
Linus Torvalds 已提交
2425 2426 2427
}

/* Try to undo cwnd reduction, because D-SACKs acked all retransmitted data */
2428
static bool tcp_try_undo_dsack(struct sock *sk)
L
Linus Torvalds 已提交
2429
{
2430 2431
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2432
	if (tp->undo_marker && !tp->undo_retrans) {
2433
		DBGUNDO(sk, "D-SACK");
2434
		tcp_undo_cwnd_reduction(sk, false);
2435
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPDSACKUNDO);
2436
		return true;
L
Linus Torvalds 已提交
2437
	}
2438
	return false;
L
Linus Torvalds 已提交
2439 2440
}

Y
Yuchung Cheng 已提交
2441 2442
/* 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
2443
{
2444 2445
	struct tcp_sock *tp = tcp_sk(sk);

Y
Yuchung Cheng 已提交
2446
	if (frto_undo || tcp_may_undo(tp)) {
2447
		tcp_undo_cwnd_reduction(sk, true);
2448

2449
		DBGUNDO(sk, "partial loss");
2450
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPLOSSUNDO);
Y
Yuchung Cheng 已提交
2451
		if (frto_undo)
2452
			NET_INC_STATS(sock_net(sk),
2453
					LINUX_MIB_TCPSPURIOUSRTOS);
2454
		inet_csk(sk)->icsk_retransmits = 0;
Y
Yuchung Cheng 已提交
2455
		if (frto_undo || tcp_is_sack(tp))
2456
			tcp_set_ca_state(sk, TCP_CA_Open);
E
Eric Dumazet 已提交
2457
		return true;
L
Linus Torvalds 已提交
2458
	}
E
Eric Dumazet 已提交
2459
	return false;
L
Linus Torvalds 已提交
2460 2461
}

2462
/* The cwnd reduction in CWR and Recovery uses the PRR algorithm in RFC 6937.
2463 2464 2465 2466
 * 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.
2467 2468 2469
 *   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.
2470
 */
2471
static void tcp_init_cwnd_reduction(struct sock *sk)
2472 2473 2474 2475
{
	struct tcp_sock *tp = tcp_sk(sk);

	tp->high_seq = tp->snd_nxt;
N
Nandita Dukkipati 已提交
2476
	tp->tlp_high_seq = 0;
2477 2478 2479 2480
	tp->snd_cwnd_cnt = 0;
	tp->prior_cwnd = tp->snd_cwnd;
	tp->prr_delivered = 0;
	tp->prr_out = 0;
2481
	tp->snd_ssthresh = inet_csk(sk)->icsk_ca_ops->ssthresh(sk);
2482
	tcp_ecn_queue_cwr(tp);
2483 2484
}

2485
void tcp_cwnd_reduction(struct sock *sk, int newly_acked_sacked, int flag)
2486 2487 2488 2489 2490
{
	struct tcp_sock *tp = tcp_sk(sk);
	int sndcnt = 0;
	int delta = tp->snd_ssthresh - tcp_packets_in_flight(tp);

2491 2492 2493
	if (newly_acked_sacked <= 0 || WARN_ON_ONCE(!tp->prior_cwnd))
		return;

2494
	tp->prr_delivered += newly_acked_sacked;
2495
	if (delta < 0) {
2496 2497 2498
		u64 dividend = (u64)tp->snd_ssthresh * tp->prr_delivered +
			       tp->prior_cwnd - 1;
		sndcnt = div_u64(dividend, tp->prior_cwnd) - tp->prr_out;
2499 2500
	} else if ((flag & FLAG_RETRANS_DATA_ACKED) &&
		   !(flag & FLAG_LOST_RETRANS)) {
2501 2502 2503
		sndcnt = min_t(int, delta,
			       max_t(int, tp->prr_delivered - tp->prr_out,
				     newly_acked_sacked) + 1);
2504 2505
	} else {
		sndcnt = min(delta, newly_acked_sacked);
2506
	}
2507 2508
	/* Force a fast retransmit upon entering fast recovery */
	sndcnt = max(sndcnt, (tp->prr_out ? 0 : 1));
2509 2510 2511
	tp->snd_cwnd = tcp_packets_in_flight(tp) + sndcnt;
}

2512
static inline void tcp_end_cwnd_reduction(struct sock *sk)
L
Linus Torvalds 已提交
2513
{
2514
	struct tcp_sock *tp = tcp_sk(sk);
2515

2516 2517 2518
	if (inet_csk(sk)->icsk_ca_ops->cong_control)
		return;

2519 2520 2521 2522 2523
	/* 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;
2524
	}
2525
	tcp_ca_event(sk, CA_EVENT_COMPLETE_CWR);
L
Linus Torvalds 已提交
2526 2527
}

2528
/* Enter CWR state. Disable cwnd undo since congestion is proven with ECN */
2529
void tcp_enter_cwr(struct sock *sk)
Y
Yuchung Cheng 已提交
2530 2531 2532 2533
{
	struct tcp_sock *tp = tcp_sk(sk);

	tp->prior_ssthresh = 0;
2534
	if (inet_csk(sk)->icsk_ca_state < TCP_CA_CWR) {
Y
Yuchung Cheng 已提交
2535
		tp->undo_marker = 0;
2536
		tcp_init_cwnd_reduction(sk);
Y
Yuchung Cheng 已提交
2537 2538 2539
		tcp_set_ca_state(sk, TCP_CA_CWR);
	}
}
2540
EXPORT_SYMBOL(tcp_enter_cwr);
Y
Yuchung Cheng 已提交
2541

2542 2543 2544 2545 2546
static void tcp_try_keep_open(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	int state = TCP_CA_Open;

2547
	if (tcp_left_out(tp) || tcp_any_retrans_done(sk))
2548 2549 2550 2551 2552 2553 2554 2555
		state = TCP_CA_Disorder;

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

2556
static void tcp_try_to_open(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2557
{
2558 2559
	struct tcp_sock *tp = tcp_sk(sk);

2560 2561
	tcp_verify_left_out(tp);

Y
Yuchung Cheng 已提交
2562
	if (!tcp_any_retrans_done(sk))
L
Linus Torvalds 已提交
2563 2564
		tp->retrans_stamp = 0;

2565
	if (flag & FLAG_ECE)
2566
		tcp_enter_cwr(sk);
L
Linus Torvalds 已提交
2567

2568
	if (inet_csk(sk)->icsk_ca_state != TCP_CA_CWR) {
2569
		tcp_try_keep_open(sk);
L
Linus Torvalds 已提交
2570 2571 2572
	}
}

J
John Heffner 已提交
2573 2574 2575 2576 2577 2578
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;
2579
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMTUPFAIL);
J
John Heffner 已提交
2580 2581
}

2582
static void tcp_mtup_probe_success(struct sock *sk)
J
John Heffner 已提交
2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593
{
	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;
2594
	tp->snd_ssthresh = tcp_current_ssthresh(sk);
J
John Heffner 已提交
2595 2596 2597 2598

	icsk->icsk_mtup.search_low = icsk->icsk_mtup.probe_size;
	icsk->icsk_mtup.probe_size = 0;
	tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
2599
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMTUPSUCCESS);
J
John Heffner 已提交
2600 2601
}

2602 2603 2604 2605 2606 2607 2608 2609 2610
/* 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
2611
	unsigned int mss = tcp_current_mss(sk);
2612 2613 2614 2615 2616
	u32 prior_lost = tp->lost_out;

	tcp_for_write_queue(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
2617
		if (tcp_skb_seglen(skb) > mss &&
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 2645 2646 2647 2648 2649 2650
		    !(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
2651
EXPORT_SYMBOL(tcp_simple_retransmit);
2652

2653
void tcp_enter_recovery(struct sock *sk, bool ece_ack)
2654 2655 2656 2657 2658 2659 2660 2661 2662
{
	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;

2663
	NET_INC_STATS(sock_net(sk), mib_idx);
2664 2665

	tp->prior_ssthresh = 0;
Y
Yuchung Cheng 已提交
2666
	tcp_init_undo(tp);
2667

2668
	if (!tcp_in_cwnd_reduction(sk)) {
2669 2670
		if (!ece_ack)
			tp->prior_ssthresh = tcp_current_ssthresh(sk);
2671
		tcp_init_cwnd_reduction(sk);
2672 2673 2674 2675
	}
	tcp_set_ca_state(sk, TCP_CA_Recovery);
}

2676 2677 2678
/* Process an ACK in CA_Loss state. Move to CA_Open if lost data are
 * recovered or spurious. Otherwise retransmits more on partial ACKs.
 */
2679 2680
static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack,
			     int *rexmit)
2681 2682
{
	struct tcp_sock *tp = tcp_sk(sk);
Y
Yuchung Cheng 已提交
2683
	bool recovered = !before(tp->snd_una, tp->high_seq);
2684

2685 2686 2687 2688
	if ((flag & FLAG_SND_UNA_ADVANCED) &&
	    tcp_try_undo_loss(sk, false))
		return;

2689 2690 2691 2692 2693 2694 2695 2696 2697 2698
	/* 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;
2699

2700
	if (tp->frto) { /* F-RTO RFC5682 sec 3.1 (sack enhanced version). */
2701 2702 2703
		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
2704 2705
		} else if (flag & FLAG_SND_UNA_ADVANCED && !recovered) {
			tp->high_seq = tp->snd_nxt;
2706 2707 2708 2709 2710 2711 2712 2713 2714
			/* 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
2715 2716 2717 2718 2719 2720
			tp->frto = 0;
		}
	}

	if (recovered) {
		/* F-RTO RFC5682 sec 3.1 step 2.a and 1st part of step 3.a */
2721 2722 2723
		tcp_try_undo_recovery(sk);
		return;
	}
Y
Yuchung Cheng 已提交
2724 2725 2726 2727 2728 2729 2730 2731 2732
	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);
	}
2733
	*rexmit = REXMIT_LOST;
2734 2735
}

Y
Yuchung Cheng 已提交
2736
/* Undo during fast recovery after partial ACK. */
2737
static bool tcp_try_undo_partial(struct sock *sk, const int acked)
Y
Yuchung Cheng 已提交
2738 2739 2740
{
	struct tcp_sock *tp = tcp_sk(sk);

2741
	if (tp->undo_marker && tcp_packet_delayed(tp)) {
Y
Yuchung Cheng 已提交
2742 2743 2744
		/* Plain luck! Hole if filled with delayed
		 * packet, rather than with a retransmit.
		 */
2745 2746 2747 2748 2749 2750 2751
		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.
		 */
2752
		if (tp->retrans_out)
2753 2754
			return true;

Y
Yuchung Cheng 已提交
2755 2756 2757
		if (!tcp_any_retrans_done(sk))
			tp->retrans_stamp = 0;

2758 2759
		DBGUNDO(sk, "partial recovery");
		tcp_undo_cwnd_reduction(sk, true);
2760
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPPARTIALUNDO);
2761 2762
		tcp_try_keep_open(sk);
		return true;
Y
Yuchung Cheng 已提交
2763
	}
2764
	return false;
Y
Yuchung Cheng 已提交
2765 2766
}

2767
static void tcp_rack_identify_loss(struct sock *sk, int *ack_flag)
2768 2769 2770 2771
{
	struct tcp_sock *tp = tcp_sk(sk);

	/* Use RACK to detect loss */
2772
	if (sysctl_tcp_recovery & TCP_RACK_LOSS_DETECTION) {
2773 2774
		u32 prior_retrans = tp->retrans_out;

2775
		tcp_rack_mark_lost(sk);
2776 2777 2778 2779 2780
		if (prior_retrans > tp->retrans_out)
			*ack_flag |= FLAG_LOST_RETRANS;
	}
}

L
Linus Torvalds 已提交
2781 2782 2783 2784 2785
/* 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.
 *
2786 2787 2788
 * 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
2789 2790 2791 2792
 *
 * It does _not_ decide what to send, it is made in function
 * tcp_xmit_retransmit_queue().
 */
2793
static void tcp_fastretrans_alert(struct sock *sk, const int acked,
2794
				  bool is_dupack, int *ack_flag, int *rexmit)
L
Linus Torvalds 已提交
2795
{
2796
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2797
	struct tcp_sock *tp = tcp_sk(sk);
2798
	int fast_rexmit = 0, flag = *ack_flag;
Y
Yuchung Cheng 已提交
2799
	bool do_lost = is_dupack || ((flag & FLAG_DATA_SACKED) &&
2800
				    (tcp_fackets_out(tp) > tp->reordering));
L
Linus Torvalds 已提交
2801

2802
	if (WARN_ON(!tp->packets_out && tp->sacked_out))
L
Linus Torvalds 已提交
2803
		tp->sacked_out = 0;
I
Ilpo Järvinen 已提交
2804
	if (WARN_ON(!tp->sacked_out && tp->fackets_out))
L
Linus Torvalds 已提交
2805 2806
		tp->fackets_out = 0;

2807
	/* Now state machine starts.
L
Linus Torvalds 已提交
2808
	 * A. ECE, hence prohibit cwnd undoing, the reduction is required. */
2809
	if (flag & FLAG_ECE)
L
Linus Torvalds 已提交
2810 2811 2812
		tp->prior_ssthresh = 0;

	/* B. In all the states check for reneging SACKs. */
2813
	if (tcp_check_sack_reneging(sk, flag))
L
Linus Torvalds 已提交
2814 2815
		return;

2816
	/* C. Check consistency of the current state. */
2817
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2818

2819
	/* D. Check state exit conditions. State can be terminated
L
Linus Torvalds 已提交
2820
	 *    when high_seq is ACKed. */
2821
	if (icsk->icsk_ca_state == TCP_CA_Open) {
2822
		WARN_ON(tp->retrans_out != 0);
L
Linus Torvalds 已提交
2823 2824
		tp->retrans_stamp = 0;
	} else if (!before(tp->snd_una, tp->high_seq)) {
2825
		switch (icsk->icsk_ca_state) {
L
Linus Torvalds 已提交
2826 2827 2828 2829
		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) {
2830
				tcp_end_cwnd_reduction(sk);
2831
				tcp_set_ca_state(sk, TCP_CA_Open);
L
Linus Torvalds 已提交
2832 2833 2834 2835
			}
			break;

		case TCP_CA_Recovery:
2836
			if (tcp_is_reno(tp))
L
Linus Torvalds 已提交
2837
				tcp_reset_reno_sack(tp);
2838
			if (tcp_try_undo_recovery(sk))
L
Linus Torvalds 已提交
2839
				return;
2840
			tcp_end_cwnd_reduction(sk);
L
Linus Torvalds 已提交
2841 2842 2843 2844
			break;
		}
	}

2845
	/* E. Process state. */
2846
	switch (icsk->icsk_ca_state) {
L
Linus Torvalds 已提交
2847
	case TCP_CA_Recovery:
2848
		if (!(flag & FLAG_SND_UNA_ADVANCED)) {
2849
			if (tcp_is_reno(tp) && is_dupack)
2850
				tcp_add_reno_sack(sk);
2851
		} else {
2852
			if (tcp_try_undo_partial(sk, acked))
2853 2854 2855 2856 2857
				return;
			/* Partial ACK arrived. Force fast retransmit. */
			do_lost = tcp_is_reno(tp) ||
				  tcp_fackets_out(tp) > tp->reordering;
		}
2858 2859 2860 2861
		if (tcp_try_undo_dsack(sk)) {
			tcp_try_keep_open(sk);
			return;
		}
2862
		tcp_rack_identify_loss(sk, ack_flag);
L
Linus Torvalds 已提交
2863 2864
		break;
	case TCP_CA_Loss:
2865
		tcp_process_loss(sk, flag, is_dupack, rexmit);
2866
		tcp_rack_identify_loss(sk, ack_flag);
2867 2868
		if (!(icsk->icsk_ca_state == TCP_CA_Open ||
		      (*ack_flag & FLAG_LOST_RETRANS)))
L
Linus Torvalds 已提交
2869
			return;
2870
		/* Change state if cwnd is undone or retransmits are lost */
L
Linus Torvalds 已提交
2871
	default:
2872
		if (tcp_is_reno(tp)) {
2873
			if (flag & FLAG_SND_UNA_ADVANCED)
L
Linus Torvalds 已提交
2874 2875
				tcp_reset_reno_sack(tp);
			if (is_dupack)
2876
				tcp_add_reno_sack(sk);
L
Linus Torvalds 已提交
2877 2878
		}

2879
		if (icsk->icsk_ca_state <= TCP_CA_Disorder)
2880
			tcp_try_undo_dsack(sk);
L
Linus Torvalds 已提交
2881

2882
		tcp_rack_identify_loss(sk, ack_flag);
2883
		if (!tcp_time_to_recover(sk, flag)) {
2884
			tcp_try_to_open(sk, flag);
L
Linus Torvalds 已提交
2885 2886 2887
			return;
		}

J
John Heffner 已提交
2888 2889 2890
		/* MTU probe failure: don't reduce cwnd */
		if (icsk->icsk_ca_state < TCP_CA_CWR &&
		    icsk->icsk_mtup.probe_size &&
2891
		    tp->snd_una == tp->mtu_probe.probe_seq_start) {
J
John Heffner 已提交
2892 2893 2894 2895 2896 2897 2898
			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
2899
		/* Otherwise enter Recovery state */
2900
		tcp_enter_recovery(sk, (flag & FLAG_ECE));
2901
		fast_rexmit = 1;
L
Linus Torvalds 已提交
2902 2903
	}

2904
	if (do_lost)
2905
		tcp_update_scoreboard(sk, fast_rexmit);
2906
	*rexmit = REXMIT_LOST;
L
Linus Torvalds 已提交
2907 2908
}

2909 2910
static void tcp_update_rtt_min(struct sock *sk, u32 rtt_us)
{
2911 2912 2913 2914 2915
	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));
2916 2917
}

Y
Yuchung Cheng 已提交
2918
static inline bool tcp_ack_update_rtt(struct sock *sk, const int flag,
2919 2920
				      long seq_rtt_us, long sack_rtt_us,
				      long ca_rtt_us)
2921
{
2922 2923 2924 2925 2926 2927 2928
	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).
	 */
2929 2930
	if (seq_rtt_us < 0)
		seq_rtt_us = sack_rtt_us;
Y
Yuchung Cheng 已提交
2931

L
Linus Torvalds 已提交
2932 2933 2934 2935 2936 2937
	/* 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.
	 */
2938
	if (seq_rtt_us < 0 && tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr &&
2939
	    flag & FLAG_ACKED)
2940 2941
		seq_rtt_us = ca_rtt_us = jiffies_to_usecs(tcp_time_stamp -
							  tp->rx_opt.rcv_tsecr);
2942
	if (seq_rtt_us < 0)
Y
Yuchung Cheng 已提交
2943
		return false;
L
Linus Torvalds 已提交
2944

2945 2946 2947 2948 2949
	/* 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);
2950
	tcp_rtt_estimator(sk, seq_rtt_us);
2951
	tcp_set_rto(sk);
L
Linus Torvalds 已提交
2952

2953 2954
	/* RFC6298: only reset backoff on valid RTT measurement. */
	inet_csk(sk)->icsk_backoff = 0;
Y
Yuchung Cheng 已提交
2955
	return true;
L
Linus Torvalds 已提交
2956 2957
}

2958
/* Compute time elapsed between (last) SYNACK and the ACK completing 3WHS. */
Y
Yuchung Cheng 已提交
2959
void tcp_synack_rtt_meas(struct sock *sk, struct request_sock *req)
2960
{
Y
Yuchung Cheng 已提交
2961
	long rtt_us = -1L;
2962

Y
Yuchung Cheng 已提交
2963 2964
	if (req && !req->num_retrans && tcp_rsk(req)->snt_synack.v64) {
		struct skb_mstamp now;
2965

Y
Yuchung Cheng 已提交
2966 2967 2968 2969
		skb_mstamp_get(&now);
		rtt_us = skb_mstamp_us_delta(&now, &tcp_rsk(req)->snt_synack);
	}

2970
	tcp_ack_update_rtt(sk, FLAG_SYN_ACKED, rtt_us, -1L, rtt_us);
2971 2972
}

Y
Yuchung Cheng 已提交
2973

2974
static void tcp_cong_avoid(struct sock *sk, u32 ack, u32 acked)
L
Linus Torvalds 已提交
2975
{
2976
	const struct inet_connection_sock *icsk = inet_csk(sk);
2977 2978

	icsk->icsk_ca_ops->cong_avoid(sk, ack, acked);
2979
	tcp_sk(sk)->snd_cwnd_stamp = tcp_time_stamp;
L
Linus Torvalds 已提交
2980 2981 2982 2983 2984
}

/* Restart timer after forward progress on connection.
 * RFC2988 recommends to restart timer to now+rto.
 */
2985
void tcp_rearm_rto(struct sock *sk)
L
Linus Torvalds 已提交
2986
{
N
Nandita Dukkipati 已提交
2987
	const struct inet_connection_sock *icsk = inet_csk(sk);
2988
	struct tcp_sock *tp = tcp_sk(sk);
2989

2990 2991 2992 2993 2994 2995
	/* 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
2996
	if (!tp->packets_out) {
2997
		inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS);
L
Linus Torvalds 已提交
2998
	} else {
2999 3000
		u32 rto = inet_csk(sk)->icsk_rto;
		/* Offset the time elapsed after installing regular RTO */
Y
Yuchung Cheng 已提交
3001
		if (icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
N
Nandita Dukkipati 已提交
3002
		    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
3003
			struct sk_buff *skb = tcp_write_queue_head(sk);
E
Eric Dumazet 已提交
3004 3005
			const u32 rto_time_stamp =
				tcp_skb_timestamp(skb) + rto;
3006 3007
			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
3008
			 * when the retrans timer fires and is rescheduled.
3009 3010 3011 3012 3013 3014
			 */
			if (delta > 0)
				rto = delta;
		}
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, rto,
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
3015
	}
3016 3017
}

3018
/* If we get here, the whole TSO packet has not been acked. */
3019
static u32 tcp_tso_acked(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
3020 3021
{
	struct tcp_sock *tp = tcp_sk(sk);
3022
	u32 packets_acked;
L
Linus Torvalds 已提交
3023

3024
	BUG_ON(!after(TCP_SKB_CB(skb)->end_seq, tp->snd_una));
L
Linus Torvalds 已提交
3025 3026

	packets_acked = tcp_skb_pcount(skb);
3027
	if (tcp_trim_head(sk, skb, tp->snd_una - TCP_SKB_CB(skb)->seq))
L
Linus Torvalds 已提交
3028 3029 3030 3031 3032
		return 0;
	packets_acked -= tcp_skb_pcount(skb);

	if (packets_acked) {
		BUG_ON(tcp_skb_pcount(skb) == 0);
3033
		BUG_ON(!before(TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq));
L
Linus Torvalds 已提交
3034 3035
	}

3036
	return packets_acked;
L
Linus Torvalds 已提交
3037 3038
}

3039 3040 3041 3042 3043 3044
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 */
3045
	if (likely(!TCP_SKB_CB(skb)->txstamp_ack))
3046 3047 3048
		return;

	shinfo = skb_shinfo(skb);
3049
	if (!before(shinfo->tskey, prior_snd_una) &&
3050
	    before(shinfo->tskey, tcp_sk(sk)->snd_una))
3051 3052 3053
		__skb_tstamp_tx(skb, NULL, sk, SCM_TSTAMP_ACK);
}

3054 3055 3056 3057
/* 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
3058
static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets,
3059
			       u32 prior_snd_una, int *acked,
3060
			       struct tcp_sacktag_state *sack)
L
Linus Torvalds 已提交
3061
{
3062
	const struct inet_connection_sock *icsk = inet_csk(sk);
3063
	struct skb_mstamp first_ackt, last_ackt;
3064
	struct tcp_sock *tp = tcp_sk(sk);
3065
	struct skb_mstamp *now = &tp->tcp_mstamp;
3066 3067
	u32 prior_sacked = tp->sacked_out;
	u32 reord = tp->packets_out;
3068
	bool fully_acked = true;
3069
	long sack_rtt_us = -1L;
3070
	long seq_rtt_us = -1L;
3071
	long ca_rtt_us = -1L;
3072
	struct sk_buff *skb;
3073
	u32 pkts_acked = 0;
3074
	u32 last_in_flight = 0;
3075
	bool rtt_update;
3076 3077 3078
	int flag = 0;

	first_ackt.v64 = 0;
L
Linus Torvalds 已提交
3079

3080
	while ((skb = tcp_write_queue_head(sk)) && skb != tcp_send_head(sk)) {
3081
		struct tcp_skb_cb *scb = TCP_SKB_CB(skb);
3082
		u8 sacked = scb->sacked;
3083
		u32 acked_pcount;
L
Linus Torvalds 已提交
3084

3085
		tcp_ack_tstamp(sk, skb, prior_snd_una);
3086

3087
		/* Determine how many packets and what bytes were acked, tso and else */
L
Linus Torvalds 已提交
3088
		if (after(scb->end_seq, tp->snd_una)) {
3089 3090 3091 3092
			if (tcp_skb_pcount(skb) == 1 ||
			    !after(tp->snd_una, scb->seq))
				break;

3093 3094
			acked_pcount = tcp_tso_acked(sk, skb);
			if (!acked_pcount)
3095
				break;
E
Eric Dumazet 已提交
3096
			fully_acked = false;
3097
		} else {
3098 3099
			/* Speedup tcp_unlink_write_queue() and next loop */
			prefetchw(skb->next);
3100
			acked_pcount = tcp_skb_pcount(skb);
L
Linus Torvalds 已提交
3101 3102
		}

3103
		if (unlikely(sacked & TCPCB_RETRANS)) {
3104
			if (sacked & TCPCB_SACKED_RETRANS)
3105
				tp->retrans_out -= acked_pcount;
3106
			flag |= FLAG_RETRANS_DATA_ACKED;
3107
		} else if (!(sacked & TCPCB_SACKED_ACKED)) {
3108
			last_ackt = skb->skb_mstamp;
E
Eric Dumazet 已提交
3109
			WARN_ON_ONCE(last_ackt.v64 == 0);
3110 3111 3112
			if (!first_ackt.v64)
				first_ackt = last_ackt;

3113
			last_in_flight = TCP_SKB_CB(skb)->tx.in_flight;
3114 3115 3116
			reord = min(pkts_acked, reord);
			if (!after(scb->end_seq, tp->high_seq))
				flag |= FLAG_ORIG_SACK_ACKED;
3117
		}
3118

Y
Yuchung Cheng 已提交
3119
		if (sacked & TCPCB_SACKED_ACKED) {
3120
			tp->sacked_out -= acked_pcount;
Y
Yuchung Cheng 已提交
3121 3122 3123
		} else if (tcp_is_sack(tp)) {
			tp->delivered += acked_pcount;
			if (!tcp_skb_spurious_retrans(tp, skb))
3124
				tcp_rack_advance(tp, sacked, scb->end_seq,
3125
						 &skb->skb_mstamp);
Y
Yuchung Cheng 已提交
3126
		}
3127
		if (sacked & TCPCB_LOST)
3128
			tp->lost_out -= acked_pcount;
3129

3130 3131
		tp->packets_out -= acked_pcount;
		pkts_acked += acked_pcount;
3132
		tcp_rate_skb_delivered(sk, skb, sack->rate);
3133

3134 3135 3136 3137 3138 3139 3140
		/* 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.
		 */
3141
		if (likely(!(scb->tcp_flags & TCPHDR_SYN))) {
3142 3143 3144 3145 3146 3147
			flag |= FLAG_DATA_ACKED;
		} else {
			flag |= FLAG_SYN_ACKED;
			tp->retrans_stamp = 0;
		}

3148 3149 3150
		if (!fully_acked)
			break;

3151
		tcp_unlink_write_queue(skb, sk);
3152
		sk_wmem_free_skb(sk, skb);
3153
		if (unlikely(skb == tp->retransmit_skb_hint))
3154
			tp->retransmit_skb_hint = NULL;
3155
		if (unlikely(skb == tp->lost_skb_hint))
3156
			tp->lost_skb_hint = NULL;
L
Linus Torvalds 已提交
3157 3158
	}

3159 3160 3161
	if (!skb)
		tcp_chrono_stop(sk, TCP_CHRONO_BUSY);

I
Ilpo Järvinen 已提交
3162 3163 3164
	if (likely(between(tp->snd_up, prior_snd_una, tp->snd_una)))
		tp->snd_up = tp->snd_una;

3165 3166 3167
	if (skb && (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
		flag |= FLAG_SACK_RENEGING;

3168
	if (likely(first_ackt.v64) && !(flag & FLAG_RETRANS_DATA_ACKED)) {
3169 3170
		seq_rtt_us = skb_mstamp_us_delta(now, &first_ackt);
		ca_rtt_us = skb_mstamp_us_delta(now, &last_ackt);
3171 3172
	}
	if (sack->first_sackt.v64) {
3173 3174
		sack_rtt_us = skb_mstamp_us_delta(now, &sack->first_sackt);
		ca_rtt_us = skb_mstamp_us_delta(now, &sack->last_sackt);
3175
	}
3176
	sack->rate->rtt_us = ca_rtt_us; /* RTT of last (S)ACKed packet, or -1 */
3177 3178
	rtt_update = tcp_ack_update_rtt(sk, flag, seq_rtt_us, sack_rtt_us,
					ca_rtt_us);
Y
Yuchung Cheng 已提交
3179

3180
	if (flag & FLAG_ACKED) {
3181
		tcp_rearm_rto(sk);
3182 3183 3184 3185 3186
		if (unlikely(icsk->icsk_mtup.probe_size &&
			     !after(tp->mtu_probe.probe_seq_end, tp->snd_una))) {
			tcp_mtup_probe_success(sk);
		}

3187 3188 3189
		if (tcp_is_reno(tp)) {
			tcp_remove_reno_sacks(sk, pkts_acked);
		} else {
3190 3191
			int delta;

3192 3193 3194
			/* Non-retransmitted hole got filled? That's reordering */
			if (reord < prior_fackets)
				tcp_update_reordering(sk, tp->fackets_out - reord, 0);
3195

3196 3197 3198
			delta = tcp_is_fack(tp) ? pkts_acked :
						  prior_sacked - tp->sacked_out;
			tp->lost_cnt_hint -= min(tp->lost_cnt_hint, delta);
3199 3200
		}

I
Ilpo Järvinen 已提交
3201
		tp->fackets_out -= min(pkts_acked, tp->fackets_out);
3202

3203
	} else if (skb && rtt_update && sack_rtt_us >= 0 &&
3204
		   sack_rtt_us > skb_mstamp_us_delta(now, &skb->skb_mstamp)) {
3205 3206 3207 3208 3209
		/* 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
3210 3211
	}

3212 3213
	if (icsk->icsk_ca_ops->pkts_acked) {
		struct ack_sample sample = { .pkts_acked = pkts_acked,
3214 3215
					     .rtt_us = ca_rtt_us,
					     .in_flight = last_in_flight };
3216 3217 3218

		icsk->icsk_ca_ops->pkts_acked(sk, &sample);
	}
3219

L
Linus Torvalds 已提交
3220
#if FASTRETRANS_DEBUG > 0
3221 3222 3223
	WARN_ON((int)tp->sacked_out < 0);
	WARN_ON((int)tp->lost_out < 0);
	WARN_ON((int)tp->retrans_out < 0);
3224
	if (!tp->packets_out && tcp_is_sack(tp)) {
S
Stephen Hemminger 已提交
3225
		icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3226
		if (tp->lost_out) {
3227 3228
			pr_debug("Leak l=%u %d\n",
				 tp->lost_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3229 3230 3231
			tp->lost_out = 0;
		}
		if (tp->sacked_out) {
3232 3233
			pr_debug("Leak s=%u %d\n",
				 tp->sacked_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3234 3235 3236
			tp->sacked_out = 0;
		}
		if (tp->retrans_out) {
3237 3238
			pr_debug("Leak r=%u %d\n",
				 tp->retrans_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3239 3240 3241 3242
			tp->retrans_out = 0;
		}
	}
#endif
3243
	*acked = pkts_acked;
3244
	return flag;
L
Linus Torvalds 已提交
3245 3246 3247 3248
}

static void tcp_ack_probe(struct sock *sk)
{
3249 3250
	const struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3251 3252 3253

	/* Was it a usable window open? */

3254
	if (!after(TCP_SKB_CB(tcp_send_head(sk))->end_seq, tcp_wnd_end(tp))) {
3255 3256
		icsk->icsk_backoff = 0;
		inet_csk_clear_xmit_timer(sk, ICSK_TIME_PROBE0);
L
Linus Torvalds 已提交
3257 3258 3259 3260
		/* Socket must be waked up by subsequent tcp_data_snd_check().
		 * This function is not for random using!
		 */
	} else {
3261
		unsigned long when = tcp_probe0_when(sk, TCP_RTO_MAX);
3262

3263
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
3264
					  when, TCP_RTO_MAX);
L
Linus Torvalds 已提交
3265 3266 3267
	}
}

3268
static inline bool tcp_ack_is_dubious(const struct sock *sk, const int flag)
L
Linus Torvalds 已提交
3269
{
E
Eric Dumazet 已提交
3270 3271
	return !(flag & FLAG_NOT_DUP) || (flag & FLAG_CA_ALERT) ||
		inet_csk(sk)->icsk_ca_state != TCP_CA_Open;
L
Linus Torvalds 已提交
3272 3273
}

3274
/* Decide wheather to run the increase function of congestion control. */
3275
static inline bool tcp_may_raise_cwnd(const struct sock *sk, const int flag)
L
Linus Torvalds 已提交
3276
{
3277 3278
	/* If reordering is high then always grow cwnd whenever data is
	 * delivered regardless of its ordering. Otherwise stay conservative
3279
	 * and only grow cwnd on in-order delivery (RFC5681). A stretched ACK w/
3280 3281 3282
	 * new SACK or ECE mark may first advance cwnd here and later reduce
	 * cwnd in tcp_fastretrans_alert() based on more states.
	 */
3283
	if (tcp_sk(sk)->reordering > sock_net(sk)->ipv4.sysctl_tcp_reordering)
3284 3285
		return flag & FLAG_FORWARD_PROGRESS;

3286
	return flag & FLAG_DATA_ACKED;
L
Linus Torvalds 已提交
3287 3288
}

Y
Yuchung Cheng 已提交
3289 3290 3291 3292 3293 3294
/* 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,
3295
			     int flag, const struct rate_sample *rs)
Y
Yuchung Cheng 已提交
3296
{
3297 3298 3299 3300 3301 3302 3303
	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
3304 3305 3306 3307 3308 3309 3310 3311 3312 3313
	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
3314 3315 3316
/* Check that window update is acceptable.
 * The function assumes that snd_una<=ack<=snd_next.
 */
3317
static inline bool tcp_may_update_window(const struct tcp_sock *tp,
3318 3319
					const u32 ack, const u32 ack_seq,
					const u32 nwin)
L
Linus Torvalds 已提交
3320
{
E
Eric Dumazet 已提交
3321
	return	after(ack, tp->snd_una) ||
L
Linus Torvalds 已提交
3322
		after(ack_seq, tp->snd_wl1) ||
E
Eric Dumazet 已提交
3323
		(ack_seq == tp->snd_wl1 && nwin > tp->snd_wnd);
L
Linus Torvalds 已提交
3324 3325
}

3326 3327 3328 3329 3330
/* 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;

3331
	sock_owned_by_me((struct sock *)tp);
3332 3333 3334 3335
	tp->bytes_acked += delta;
	tp->snd_una = ack;
}

3336 3337 3338 3339 3340
/* 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;

3341
	sock_owned_by_me((struct sock *)tp);
3342 3343 3344 3345
	tp->bytes_received += delta;
	tp->rcv_nxt = seq;
}

L
Linus Torvalds 已提交
3346 3347 3348 3349 3350
/* 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.
 */
3351
static int tcp_ack_update_window(struct sock *sk, const struct sk_buff *skb, u32 ack,
3352
				 u32 ack_seq)
L
Linus Torvalds 已提交
3353
{
3354
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
3355
	int flag = 0;
3356
	u32 nwin = ntohs(tcp_hdr(skb)->window);
L
Linus Torvalds 已提交
3357

3358
	if (likely(!tcp_hdr(skb)->syn))
L
Linus Torvalds 已提交
3359 3360 3361 3362
		nwin <<= tp->rx_opt.snd_wscale;

	if (tcp_may_update_window(tp, ack, ack_seq, nwin)) {
		flag |= FLAG_WIN_UPDATE;
3363
		tcp_update_wl(tp, ack_seq);
L
Linus Torvalds 已提交
3364 3365 3366 3367 3368 3369 3370

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

			/* Note, it is the only place, where
			 * fast path is recovered for sending TCP.
			 */
3371
			tp->pred_flags = 0;
3372
			tcp_fast_path_check(sk);
L
Linus Torvalds 已提交
3373

3374 3375 3376
			if (tcp_send_head(sk))
				tcp_slow_start_after_idle_check(sk);

L
Linus Torvalds 已提交
3377 3378
			if (nwin > tp->max_window) {
				tp->max_window = nwin;
3379
				tcp_sync_mss(sk, inet_csk(sk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
3380 3381 3382 3383
			}
		}
	}

3384
	tcp_snd_una_update(tp, ack);
L
Linus Torvalds 已提交
3385 3386 3387 3388

	return flag;
}

3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405
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! */
}

3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418
/* 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)
3419
		return false;
3420

3421
	return __tcp_oow_rate_limited(net, mib_idx, last_oow_ack_time);
3422 3423
}

3424
/* RFC 5961 7 [ACK Throttling] */
3425
static void tcp_send_challenge_ack(struct sock *sk, const struct sk_buff *skb)
3426 3427 3428 3429
{
	/* unprotected vars, we dont care of overwrites */
	static u32 challenge_timestamp;
	static unsigned int challenge_count;
3430
	struct tcp_sock *tp = tcp_sk(sk);
3431
	u32 count, now;
3432 3433

	/* First check our per-socket dupack rate limit. */
3434 3435 3436
	if (__tcp_oow_rate_limited(sock_net(sk),
				   LINUX_MIB_TCPACKSKIPPEDCHALLENGE,
				   &tp->last_oow_ack_time))
3437
		return;
3438

3439
	/* Then check host-wide RFC 5961 rate limit. */
3440
	now = jiffies / HZ;
3441
	if (now != challenge_timestamp) {
3442 3443
		u32 half = (sysctl_tcp_challenge_ack_limit + 1) >> 1;

3444
		challenge_timestamp = now;
3445 3446
		WRITE_ONCE(challenge_count, half +
			   prandom_u32_max(sysctl_tcp_challenge_ack_limit));
3447
	}
3448 3449 3450
	count = READ_ONCE(challenge_count);
	if (count > 0) {
		WRITE_ONCE(challenge_count, count - 1);
3451
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPCHALLENGEACK);
3452 3453 3454 3455
		tcp_send_ack(sk);
	}
}

3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476
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
3477
/* This routine deals with acks during a TLP episode.
3478 3479
 * 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
3480 3481 3482 3483 3484 3485
 * 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);

3486
	if (before(ack, tp->tlp_high_seq))
N
Nandita Dukkipati 已提交
3487 3488
		return;

3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499
	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);
3500
		NET_INC_STATS(sock_net(sk),
3501
				LINUX_MIB_TCPLOSSPROBERECOVERY);
3502 3503 3504
	} 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
3505 3506 3507 3508
		tp->tlp_high_seq = 0;
	}
}

3509 3510 3511 3512 3513 3514 3515 3516
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);
}

3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537
/* 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
3538
/* This routine deals with incoming acks, but not outgoing ones. */
3539
static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
L
Linus Torvalds 已提交
3540
{
3541
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3542
	struct tcp_sock *tp = tcp_sk(sk);
3543
	struct tcp_sacktag_state sack_state;
3544
	struct rate_sample rs = { .prior_delivered = 0 };
L
Linus Torvalds 已提交
3545 3546 3547
	u32 prior_snd_una = tp->snd_una;
	u32 ack_seq = TCP_SKB_CB(skb)->seq;
	u32 ack = TCP_SKB_CB(skb)->ack_seq;
3548
	bool is_dupack = false;
3549
	u32 prior_fackets;
3550
	int prior_packets = tp->packets_out;
3551 3552
	u32 delivered = tp->delivered;
	u32 lost = tp->lost;
3553
	int acked = 0; /* Number of packets newly acked */
3554
	int rexmit = REXMIT_NONE; /* Flag to (re)transmit to recover losses */
3555

3556
	sack_state.first_sackt.v64 = 0;
3557
	sack_state.rate = &rs;
L
Linus Torvalds 已提交
3558

3559 3560 3561
	/* We very likely will need to access write queue head. */
	prefetchw(sk->sk_write_queue.next);

3562
	/* If the ack is older than previous acks
L
Linus Torvalds 已提交
3563 3564
	 * then we can probably ignore it.
	 */
3565 3566 3567
	if (before(ack, prior_snd_una)) {
		/* RFC 5961 5.2 [Blind Data Injection Attack].[Mitigation] */
		if (before(ack, prior_snd_una - tp->max_window)) {
3568
			tcp_send_challenge_ack(sk, skb);
3569 3570
			return -1;
		}
L
Linus Torvalds 已提交
3571
		goto old_ack;
3572
	}
L
Linus Torvalds 已提交
3573

3574 3575 3576 3577 3578 3579
	/* 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
3580
	if (icsk->icsk_pending == ICSK_TIME_LOSS_PROBE)
3581 3582
		tcp_rearm_rto(sk);

3583
	if (after(ack, prior_snd_una)) {
3584
		flag |= FLAG_SND_UNA_ADVANCED;
3585 3586
		icsk->icsk_retransmits = 0;
	}
3587

3588
	prior_fackets = tp->fackets_out;
3589
	rs.prior_in_flight = tcp_packets_in_flight(tp);
3590

3591 3592 3593 3594 3595 3596
	/* 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);

3597
	if (!(flag & FLAG_SLOWPATH) && after(ack, prior_snd_una)) {
L
Linus Torvalds 已提交
3598 3599 3600 3601
		/* Window is constant, pure forward advance.
		 * No more checks are required.
		 * Note, we use the fact that SND.UNA>=SND.WL2.
		 */
3602
		tcp_update_wl(tp, ack_seq);
3603
		tcp_snd_una_update(tp, ack);
L
Linus Torvalds 已提交
3604 3605
		flag |= FLAG_WIN_UPDATE;

3606
		tcp_in_ack_event(sk, CA_ACK_WIN_UPDATE);
3607

3608
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPHPACKS);
L
Linus Torvalds 已提交
3609
	} else {
3610 3611
		u32 ack_ev_flags = CA_ACK_SLOWPATH;

L
Linus Torvalds 已提交
3612 3613 3614
		if (ack_seq != TCP_SKB_CB(skb)->end_seq)
			flag |= FLAG_DATA;
		else
3615
			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPPUREACKS);
L
Linus Torvalds 已提交
3616

3617
		flag |= tcp_ack_update_window(sk, skb, ack, ack_seq);
L
Linus Torvalds 已提交
3618 3619

		if (TCP_SKB_CB(skb)->sacked)
Y
Yuchung Cheng 已提交
3620
			flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una,
3621
							&sack_state);
L
Linus Torvalds 已提交
3622

3623
		if (tcp_ecn_rcv_ecn_echo(tp, tcp_hdr(skb))) {
L
Linus Torvalds 已提交
3624
			flag |= FLAG_ECE;
3625 3626 3627 3628 3629
			ack_ev_flags |= CA_ACK_ECE;
		}

		if (flag & FLAG_WIN_UPDATE)
			ack_ev_flags |= CA_ACK_WIN_UPDATE;
L
Linus Torvalds 已提交
3630

3631
		tcp_in_ack_event(sk, ack_ev_flags);
L
Linus Torvalds 已提交
3632 3633 3634 3635 3636 3637
	}

	/* We passed data and got it acked, remove any soft error
	 * log. Something worked...
	 */
	sk->sk_err_soft = 0;
3638
	icsk->icsk_probes_out = 0;
L
Linus Torvalds 已提交
3639 3640 3641 3642 3643
	tp->rcv_tstamp = tcp_time_stamp;
	if (!prior_packets)
		goto no_queue;

	/* See if we can take anything off of the retransmit queue. */
3644
	flag |= tcp_clean_rtx_queue(sk, prior_fackets, prior_snd_una, &acked,
3645
				    &sack_state);
3646

3647
	if (tcp_ack_is_dubious(sk, flag)) {
3648
		is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP));
3649
		tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit);
L
Linus Torvalds 已提交
3650
	}
N
Nandita Dukkipati 已提交
3651 3652 3653
	if (tp->tlp_high_seq)
		tcp_process_tlp_ack(sk, ack, flag);

3654 3655
	if ((flag & FLAG_FORWARD_PROGRESS) || !(flag & FLAG_NOT_DUP))
		sk_dst_confirm(sk);
N
Nandita Dukkipati 已提交
3656 3657 3658

	if (icsk->icsk_pending == ICSK_TIME_RETRANS)
		tcp_schedule_loss_probe(sk);
3659 3660
	delivered = tp->delivered - delivered;	/* freshly ACKed or SACKed */
	lost = tp->lost - lost;			/* freshly marked lost */
3661
	tcp_rate_gen(sk, delivered, lost, sack_state.rate);
3662
	tcp_cong_control(sk, ack, delivered, flag, sack_state.rate);
3663
	tcp_xmit_recovery(sk, rexmit);
L
Linus Torvalds 已提交
3664 3665 3666
	return 1;

no_queue:
3667 3668
	/* If data was DSACKed, see if we can undo a cwnd reduction. */
	if (flag & FLAG_DSACKING_ACK)
3669
		tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit);
L
Linus Torvalds 已提交
3670 3671 3672 3673
	/* 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.
	 */
3674
	if (tcp_send_head(sk))
L
Linus Torvalds 已提交
3675
		tcp_ack_probe(sk);
N
Nandita Dukkipati 已提交
3676 3677 3678

	if (tp->tlp_high_seq)
		tcp_process_tlp_ack(sk, ack, flag);
L
Linus Torvalds 已提交
3679 3680
	return 1;

3681 3682 3683 3684
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
3685
old_ack:
3686 3687 3688
	/* 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.
	 */
3689
	if (TCP_SKB_CB(skb)->sacked) {
Y
Yuchung Cheng 已提交
3690
		flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una,
3691
						&sack_state);
3692
		tcp_fastretrans_alert(sk, acked, is_dupack, &flag, &rexmit);
3693
		tcp_xmit_recovery(sk, rexmit);
3694
	}
L
Linus Torvalds 已提交
3695

3696
	SOCK_DEBUG(sk, "Ack %u before %u:%u\n", ack, tp->snd_una, tp->snd_nxt);
L
Linus Torvalds 已提交
3697 3698 3699
	return 0;
}

3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716
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
3717 3718 3719 3720
/* 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
3721 3722
void tcp_parse_options(const struct sk_buff *skb,
		       struct tcp_options_received *opt_rx, int estab,
Y
Yuchung Cheng 已提交
3723
		       struct tcp_fastopen_cookie *foc)
L
Linus Torvalds 已提交
3724
{
3725 3726
	const unsigned char *ptr;
	const struct tcphdr *th = tcp_hdr(skb);
3727
	int length = (th->doff * 4) - sizeof(struct tcphdr);
L
Linus Torvalds 已提交
3728

3729
	ptr = (const unsigned char *)(th + 1);
L
Linus Torvalds 已提交
3730 3731
	opt_rx->saw_tstamp = 0;

S
Stephen Hemminger 已提交
3732
	while (length > 0) {
3733
		int opcode = *ptr++;
L
Linus Torvalds 已提交
3734 3735 3736
		int opsize;

		switch (opcode) {
3737 3738 3739 3740 3741 3742 3743 3744
		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
3745
				return;
3746 3747 3748 3749 3750
			if (opsize > length)
				return;	/* don't parse partial options */
			switch (opcode) {
			case TCPOPT_MSS:
				if (opsize == TCPOLEN_MSS && th->syn && !estab) {
3751
					u16 in_mss = get_unaligned_be16(ptr);
3752 3753 3754 3755 3756
					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
3757
					}
3758 3759 3760 3761
				}
				break;
			case TCPOPT_WINDOW:
				if (opsize == TCPOLEN_WINDOW && th->syn &&
3762
				    !estab && sysctl_tcp_window_scaling) {
3763 3764
					__u8 snd_wscale = *(__u8 *)ptr;
					opt_rx->wscale_ok = 1;
3765 3766
					if (snd_wscale > TCP_MAX_WSCALE) {
						net_info_ratelimited("%s: Illegal window scaling value %d > %u received\n",
3767
								     __func__,
3768 3769 3770
								     snd_wscale,
								     TCP_MAX_WSCALE);
						snd_wscale = TCP_MAX_WSCALE;
L
Linus Torvalds 已提交
3771
					}
3772 3773 3774 3775 3776 3777
					opt_rx->snd_wscale = snd_wscale;
				}
				break;
			case TCPOPT_TIMESTAMP:
				if ((opsize == TCPOLEN_TIMESTAMP) &&
				    ((estab && opt_rx->tstamp_ok) ||
3778
				     (!estab && sysctl_tcp_timestamps))) {
3779
					opt_rx->saw_tstamp = 1;
3780 3781
					opt_rx->rcv_tsval = get_unaligned_be32(ptr);
					opt_rx->rcv_tsecr = get_unaligned_be32(ptr + 4);
3782 3783 3784 3785
				}
				break;
			case TCPOPT_SACK_PERM:
				if (opsize == TCPOLEN_SACK_PERM && th->syn &&
3786
				    !estab && sysctl_tcp_sack) {
3787
					opt_rx->sack_ok = TCP_SACK_SEEN;
3788 3789 3790
					tcp_sack_reset(opt_rx);
				}
				break;
L
Linus Torvalds 已提交
3791

3792 3793 3794 3795 3796 3797 3798
			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;
3799
#ifdef CONFIG_TCP_MD5SIG
3800 3801 3802 3803 3804 3805
			case TCPOPT_MD5SIG:
				/*
				 * The MD5 Hash has already been
				 * checked (see tcp_v{4,6}_do_rcv()).
				 */
				break;
3806
#endif
3807 3808 3809 3810 3811 3812
			case TCPOPT_FASTOPEN:
				tcp_parse_fastopen_option(
					opsize - TCPOLEN_FASTOPEN_BASE,
					ptr, th->syn, foc, false);
				break;

Y
Yuchung Cheng 已提交
3813 3814
			case TCPOPT_EXP:
				/* Fast Open option shares code 254 using a
3815
				 * 16 bits magic number.
Y
Yuchung Cheng 已提交
3816
				 */
3817 3818 3819 3820 3821 3822
				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
3823 3824 3825
				break;

			}
3826 3827
			ptr += opsize-2;
			length -= opsize;
3828
		}
L
Linus Torvalds 已提交
3829 3830
	}
}
E
Eric Dumazet 已提交
3831
EXPORT_SYMBOL(tcp_parse_options);
L
Linus Torvalds 已提交
3832

E
Eric Dumazet 已提交
3833
static bool tcp_parse_aligned_timestamp(struct tcp_sock *tp, const struct tcphdr *th)
3834
{
3835
	const __be32 *ptr = (const __be32 *)(th + 1);
3836 3837 3838 3839 3840 3841 3842

	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;
3843 3844 3845 3846
		if (*ptr)
			tp->rx_opt.rcv_tsecr = ntohl(*ptr) - tp->tsoffset;
		else
			tp->rx_opt.rcv_tsecr = 0;
E
Eric Dumazet 已提交
3847
		return true;
3848
	}
E
Eric Dumazet 已提交
3849
	return false;
3850 3851
}

L
Linus Torvalds 已提交
3852 3853 3854
/* 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
3855
static bool tcp_fast_parse_options(const struct sk_buff *skb,
C
Christoph Paasch 已提交
3856
				   const struct tcphdr *th, struct tcp_sock *tp)
L
Linus Torvalds 已提交
3857
{
3858 3859 3860 3861
	/* 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
3862
		tp->rx_opt.saw_tstamp = 0;
E
Eric Dumazet 已提交
3863
		return false;
L
Linus Torvalds 已提交
3864
	} else if (tp->rx_opt.tstamp_ok &&
3865
		   th->doff == ((sizeof(*th) + TCPOLEN_TSTAMP_ALIGNED) / 4)) {
3866
		if (tcp_parse_aligned_timestamp(tp, th))
E
Eric Dumazet 已提交
3867
			return true;
L
Linus Torvalds 已提交
3868
	}
3869

C
Christoph Paasch 已提交
3870
	tcp_parse_options(skb, &tp->rx_opt, 1, NULL);
3871
	if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr)
3872 3873
		tp->rx_opt.rcv_tsecr -= tp->tsoffset;

E
Eric Dumazet 已提交
3874
	return true;
L
Linus Torvalds 已提交
3875 3876
}

3877 3878 3879 3880
#ifdef CONFIG_TCP_MD5SIG
/*
 * Parse MD5 Signature option
 */
3881
const u8 *tcp_parse_md5sig_option(const struct tcphdr *th)
3882
{
3883 3884
	int length = (th->doff << 2) - sizeof(*th);
	const u8 *ptr = (const u8 *)(th + 1);
3885 3886 3887 3888 3889 3890 3891 3892 3893

	/* 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
3894
		switch (opcode) {
3895 3896 3897 3898 3899 3900 3901 3902 3903 3904
		case TCPOPT_EOL:
			return NULL;
		case TCPOPT_NOP:
			length--;
			continue;
		default:
			opsize = *ptr++;
			if (opsize < 2 || opsize > length)
				return NULL;
			if (opcode == TCPOPT_MD5SIG)
3905
				return opsize == TCPOLEN_MD5SIG ? ptr : NULL;
3906 3907 3908 3909 3910 3911
		}
		ptr += opsize - 2;
		length -= opsize;
	}
	return NULL;
}
E
Eric Dumazet 已提交
3912
EXPORT_SYMBOL(tcp_parse_md5sig_option);
3913 3914
#endif

L
Linus Torvalds 已提交
3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937
/* 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) ]
 */

3938
static int tcp_disordered_ack(const struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
3939
{
3940 3941
	const struct tcp_sock *tp = tcp_sk(sk);
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954
	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. */
3955
		(s32)(tp->rx_opt.ts_recent - tp->rx_opt.rcv_tsval) <= (inet_csk(sk)->icsk_rto * 1024) / HZ);
L
Linus Torvalds 已提交
3956 3957
}

3958
static inline bool tcp_paws_discard(const struct sock *sk,
3959
				   const struct sk_buff *skb)
L
Linus Torvalds 已提交
3960
{
3961
	const struct tcp_sock *tp = tcp_sk(sk);
I
Ilpo Järvinen 已提交
3962 3963 3964

	return !tcp_paws_check(&tp->rx_opt, TCP_PAWS_WINDOW) &&
	       !tcp_disordered_ack(sk, skb);
L
Linus Torvalds 已提交
3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979
}

/* 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)
 */

3980
static inline bool tcp_sequence(const struct tcp_sock *tp, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
3981 3982 3983 3984 3985 3986
{
	return	!before(end_seq, tp->rcv_wup) &&
		!after(seq, tp->rcv_nxt + tcp_receive_window(tp));
}

/* When we get a reset we do this. */
3987
void tcp_reset(struct sock *sk)
L
Linus Torvalds 已提交
3988 3989 3990
{
	/* We want the right error as BSD sees it (and indeed as we do). */
	switch (sk->sk_state) {
3991 3992 3993 3994 3995 3996 3997 3998 3999 4000
	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
4001
	}
T
Tom Marshall 已提交
4002 4003
	/* This barrier is coupled with smp_rmb() in tcp_poll() */
	smp_wmb();
L
Linus Torvalds 已提交
4004

4005 4006
	tcp_done(sk);

L
Linus Torvalds 已提交
4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024
	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.
 */
4025
void tcp_fin(struct sock *sk)
L
Linus Torvalds 已提交
4026 4027 4028
{
	struct tcp_sock *tp = tcp_sk(sk);

4029
	inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
4030 4031 4032 4033 4034

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

	switch (sk->sk_state) {
4035 4036 4037 4038
	case TCP_SYN_RECV:
	case TCP_ESTABLISHED:
		/* Move to CLOSE_WAIT */
		tcp_set_state(sk, TCP_CLOSE_WAIT);
4039
		inet_csk(sk)->icsk_ack.pingpong = 1;
4040
		break;
L
Linus Torvalds 已提交
4041

4042 4043 4044 4045 4046 4047 4048 4049 4050
	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
4051

4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068
	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
4069
		pr_err("%s: Impossible, sk->sk_state=%d\n",
4070
		       __func__, sk->sk_state);
4071
		break;
4072
	}
L
Linus Torvalds 已提交
4073 4074 4075 4076

	/* It _is_ possible, that we have something out-of-order _after_ FIN.
	 * Probably, we should reset in this case. For now drop them.
	 */
4077
	skb_rbtree_purge(&tp->out_of_order_queue);
4078
	if (tcp_is_sack(tp))
L
Linus Torvalds 已提交
4079
		tcp_sack_reset(&tp->rx_opt);
4080
	sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
4081 4082 4083 4084 4085 4086 4087

	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)
4088
			sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_HUP);
L
Linus Torvalds 已提交
4089
		else
4090
			sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
L
Linus Torvalds 已提交
4091 4092 4093
	}
}

E
Eric Dumazet 已提交
4094
static inline bool tcp_sack_extend(struct tcp_sack_block *sp, u32 seq,
4095
				  u32 end_seq)
L
Linus Torvalds 已提交
4096 4097 4098 4099 4100 4101
{
	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
4102
		return true;
L
Linus Torvalds 已提交
4103
	}
E
Eric Dumazet 已提交
4104
	return false;
L
Linus Torvalds 已提交
4105 4106
}

4107
static void tcp_dsack_set(struct sock *sk, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
4108
{
4109 4110
	struct tcp_sock *tp = tcp_sk(sk);

4111
	if (tcp_is_sack(tp) && sysctl_tcp_dsack) {
4112 4113
		int mib_idx;

L
Linus Torvalds 已提交
4114
		if (before(seq, tp->rcv_nxt))
4115
			mib_idx = LINUX_MIB_TCPDSACKOLDSENT;
L
Linus Torvalds 已提交
4116
		else
4117 4118
			mib_idx = LINUX_MIB_TCPDSACKOFOSENT;

4119
		NET_INC_STATS(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
4120 4121 4122 4123 4124 4125 4126

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

4127
static void tcp_dsack_extend(struct sock *sk, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
4128
{
4129 4130
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
4131
	if (!tp->rx_opt.dsack)
4132
		tcp_dsack_set(sk, seq, end_seq);
L
Linus Torvalds 已提交
4133 4134 4135 4136
	else
		tcp_sack_extend(tp->duplicate_sack, seq, end_seq);
}

4137
static void tcp_send_dupack(struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
4138 4139 4140 4141 4142
{
	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)) {
4143
		NET_INC_STATS(sock_net(sk), LINUX_MIB_DELAYEDACKLOST);
4144
		tcp_enter_quickack_mode(sk);
L
Linus Torvalds 已提交
4145

4146
		if (tcp_is_sack(tp) && sysctl_tcp_dsack) {
L
Linus Torvalds 已提交
4147 4148 4149 4150
			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;
4151
			tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, end_seq);
L
Linus Torvalds 已提交
4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164
		}
	}

	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];
4165
	struct tcp_sack_block *swalk = sp + 1;
L
Linus Torvalds 已提交
4166 4167 4168 4169

	/* See if the recent change to the first SACK eats into
	 * or hits the sequence space of other SACK blocks, if so coalesce.
	 */
4170
	for (this_sack = 1; this_sack < tp->rx_opt.num_sacks;) {
L
Linus Torvalds 已提交
4171 4172 4173 4174 4175 4176 4177
		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--;
4178 4179
			for (i = this_sack; i < tp->rx_opt.num_sacks; i++)
				sp[i] = sp[i + 1];
L
Linus Torvalds 已提交
4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195
			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;

4196
	for (this_sack = 0; this_sack < cur_sacks; this_sack++, sp++) {
L
Linus Torvalds 已提交
4197 4198
		if (tcp_sack_extend(sp, seq, end_seq)) {
			/* Rotate this_sack to the first one. */
4199
			for (; this_sack > 0; this_sack--, sp--)
4200
				swap(*sp, *(sp - 1));
L
Linus Torvalds 已提交
4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212
			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.
	 */
4213
	if (this_sack >= TCP_NUM_SACKS) {
L
Linus Torvalds 已提交
4214 4215 4216 4217
		this_sack--;
		tp->rx_opt.num_sacks--;
		sp--;
	}
S
Stephen Hemminger 已提交
4218
	for (; this_sack > 0; this_sack--, sp--)
4219
		*sp = *(sp - 1);
L
Linus Torvalds 已提交
4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236

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. */
4237
	if (RB_EMPTY_ROOT(&tp->out_of_order_queue)) {
L
Linus Torvalds 已提交
4238 4239 4240 4241
		tp->rx_opt.num_sacks = 0;
		return;
	}

4242
	for (this_sack = 0; this_sack < num_sacks;) {
L
Linus Torvalds 已提交
4243 4244 4245 4246 4247
		/* 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! */
4248
			WARN_ON(before(tp->rcv_nxt, sp->end_seq));
L
Linus Torvalds 已提交
4249 4250

			/* Zap this SACK, by moving forward any other SACKS. */
W
Weilong Chen 已提交
4251
			for (i = this_sack+1; i < num_sacks; i++)
L
Linus Torvalds 已提交
4252 4253 4254 4255 4256 4257 4258
				tp->selective_acks[i-1] = tp->selective_acks[i];
			num_sacks--;
			continue;
		}
		this_sack++;
		sp++;
	}
4259
	tp->rx_opt.num_sacks = num_sacks;
L
Linus Torvalds 已提交
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 4287 4288 4289 4290 4291 4292
/**
 * 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);
4293
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRCVCOALESCE);
4294 4295 4296 4297 4298 4299
	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;
}

4300 4301 4302 4303 4304 4305
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
4306 4307 4308 4309 4310 4311 4312
/* 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;
4313
	bool fin, fragstolen, eaten;
4314
	struct sk_buff *skb, *tail;
4315
	struct rb_node *p;
L
Linus Torvalds 已提交
4316

4317 4318 4319
	p = rb_first(&tp->out_of_order_queue);
	while (p) {
		skb = rb_entry(p, struct sk_buff, rbnode);
L
Linus Torvalds 已提交
4320 4321 4322 4323 4324 4325 4326
		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;
4327
			tcp_dsack_extend(sk, TCP_SKB_CB(skb)->seq, dsack);
L
Linus Torvalds 已提交
4328
		}
4329 4330
		p = rb_next(p);
		rb_erase(&skb->rbnode, &tp->out_of_order_queue);
L
Linus Torvalds 已提交
4331

4332
		if (unlikely(!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt))) {
4333
			SOCK_DEBUG(sk, "ofo packet was already received\n");
4334
			tcp_drop(sk, skb);
L
Linus Torvalds 已提交
4335 4336 4337 4338 4339 4340
			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);

4341 4342
		tail = skb_peek_tail(&sk->sk_receive_queue);
		eaten = tail && tcp_try_coalesce(sk, tail, skb, &fragstolen);
4343
		tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq);
4344
		fin = TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN;
4345 4346
		if (!eaten)
			__skb_queue_tail(&sk->sk_receive_queue, skb);
4347
		else
4348
			kfree_skb_partial(skb, fragstolen);
4349 4350 4351 4352 4353 4354 4355 4356

		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
4357 4358 4359
	}
}

E
Eric Dumazet 已提交
4360
static bool tcp_prune_ofo_queue(struct sock *sk);
L
Linus Torvalds 已提交
4361 4362
static int tcp_prune_queue(struct sock *sk);

4363 4364
static int tcp_try_rmem_schedule(struct sock *sk, struct sk_buff *skb,
				 unsigned int size)
4365 4366
{
	if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
4367
	    !sk_rmem_schedule(sk, skb, size)) {
4368 4369 4370 4371

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

4372
		while (!sk_rmem_schedule(sk, skb, size)) {
4373 4374
			if (!tcp_prune_ofo_queue(sk))
				return -1;
4375 4376 4377 4378 4379
		}
	}
	return 0;
}

E
Eric Dumazet 已提交
4380 4381 4382
static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);
4383
	struct rb_node **p, *q, *parent;
E
Eric Dumazet 已提交
4384 4385
	struct sk_buff *skb1;
	u32 seq, end_seq;
4386
	bool fragstolen;
E
Eric Dumazet 已提交
4387

4388
	tcp_ecn_check_ce(tp, skb);
E
Eric Dumazet 已提交
4389

4390
	if (unlikely(tcp_try_rmem_schedule(sk, skb, skb->truesize))) {
4391
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFODROP);
4392
		tcp_drop(sk, skb);
E
Eric Dumazet 已提交
4393 4394 4395 4396 4397 4398 4399
		return;
	}

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

4400
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFOQUEUE);
4401 4402
	seq = TCP_SKB_CB(skb)->seq;
	end_seq = TCP_SKB_CB(skb)->end_seq;
E
Eric Dumazet 已提交
4403
	SOCK_DEBUG(sk, "out of order segment: rcv_next %X seq %X - %X\n",
4404
		   tp->rcv_nxt, seq, end_seq);
E
Eric Dumazet 已提交
4405

4406 4407
	p = &tp->out_of_order_queue.rb_node;
	if (RB_EMPTY_ROOT(&tp->out_of_order_queue)) {
E
Eric Dumazet 已提交
4408 4409 4410
		/* Initial out of order segment, build 1 SACK. */
		if (tcp_is_sack(tp)) {
			tp->rx_opt.num_sacks = 1;
4411 4412
			tp->selective_acks[0].start_seq = seq;
			tp->selective_acks[0].end_seq = end_seq;
E
Eric Dumazet 已提交
4413
		}
4414 4415 4416
		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
4417 4418 4419
		goto end;
	}

4420 4421 4422 4423 4424 4425 4426 4427 4428 4429
	/* 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;
	}
4430 4431 4432 4433 4434 4435
	/* 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;
	}
4436 4437 4438 4439 4440 4441 4442 4443 4444

	/* 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
4445
		}
4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470
		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);
4471
				goto merge_right;
4472 4473 4474
			}
		} else if (tcp_try_coalesce(sk, skb1, skb, &fragstolen)) {
			goto coalesce_done;
E
Eric Dumazet 已提交
4475
		}
4476
		p = &parent->rb_right;
E
Eric Dumazet 已提交
4477
	}
4478
insert:
4479 4480 4481
	/* 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
4482

4483
merge_right:
4484 4485 4486
	/* 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
4487 4488 4489 4490 4491 4492 4493 4494

		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;
		}
4495
		rb_erase(&skb1->rbnode, &tp->out_of_order_queue);
E
Eric Dumazet 已提交
4496 4497
		tcp_dsack_extend(sk, TCP_SKB_CB(skb1)->seq,
				 TCP_SKB_CB(skb1)->end_seq);
4498
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFOMERGE);
4499
		tcp_drop(sk, skb1);
E
Eric Dumazet 已提交
4500
	}
4501 4502 4503
	/* 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
4504 4505 4506 4507 4508

add_sack:
	if (tcp_is_sack(tp))
		tcp_sack_new_ofo_skb(sk, seq, end_seq);
end:
4509 4510
	if (skb) {
		tcp_grow_window(sk, skb);
4511
		skb_condense(skb);
E
Eric Dumazet 已提交
4512
		skb_set_owner_r(skb, sk);
4513
	}
E
Eric Dumazet 已提交
4514 4515
}

4516
static int __must_check tcp_queue_rcv(struct sock *sk, struct sk_buff *skb, int hdrlen,
4517 4518 4519 4520 4521 4522 4523 4524
		  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;
4525
	tcp_rcv_nxt_update(tcp_sk(sk), TCP_SKB_CB(skb)->end_seq);
4526 4527 4528 4529 4530 4531
	if (!eaten) {
		__skb_queue_tail(&sk->sk_receive_queue, skb);
		skb_set_owner_r(skb, sk);
	}
	return eaten;
}
E
Eric Dumazet 已提交
4532

4533 4534
int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size)
{
4535
	struct sk_buff *skb;
4536 4537
	int err = -ENOMEM;
	int data_len = 0;
4538 4539
	bool fragstolen;

4540 4541 4542
	if (size == 0)
		return 0;

4543 4544 4545 4546 4547 4548 4549 4550 4551
	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);
4552 4553 4554
	if (!skb)
		goto err;

4555 4556 4557 4558
	skb_put(skb, size - data_len);
	skb->data_len = data_len;
	skb->len = size;

4559
	if (tcp_try_rmem_schedule(sk, skb, skb->truesize))
4560 4561
		goto err_free;

4562 4563
	err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size);
	if (err)
4564 4565 4566 4567 4568 4569
		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;

4570
	if (tcp_queue_rcv(sk, skb, 0, &fragstolen)) {
4571 4572 4573 4574 4575 4576 4577 4578
		WARN_ON_ONCE(fragstolen); /* should not happen */
		__kfree_skb(skb);
	}
	return size;

err_free:
	kfree_skb(skb);
err:
4579 4580
	return err;

4581 4582
}

L
Linus Torvalds 已提交
4583 4584 4585
static void tcp_data_queue(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);
4586
	bool fragstolen = false;
4587
	int eaten = -1;
L
Linus Torvalds 已提交
4588

4589 4590 4591 4592
	if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq) {
		__kfree_skb(skb);
		return;
	}
E
Eric Dumazet 已提交
4593
	skb_dst_drop(skb);
4594
	__skb_pull(skb, tcp_hdr(skb)->doff * 4);
L
Linus Torvalds 已提交
4595

4596
	tcp_ecn_accept_cwr(tp, skb);
L
Linus Torvalds 已提交
4597

4598
	tp->rx_opt.dsack = 0;
L
Linus Torvalds 已提交
4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612

	/*  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,
4613
					  tp->ucopy.len);
L
Linus Torvalds 已提交
4614 4615 4616

			__set_current_state(TASK_RUNNING);

4617
			if (!skb_copy_datagram_msg(skb, 0, tp->ucopy.msg, chunk)) {
L
Linus Torvalds 已提交
4618 4619
				tp->ucopy.len -= chunk;
				tp->copied_seq += chunk;
4620
				eaten = (chunk == skb->len);
L
Linus Torvalds 已提交
4621 4622 4623 4624 4625 4626
				tcp_rcv_space_adjust(sk);
			}
		}

		if (eaten <= 0) {
queue_and_out:
4627 4628 4629 4630 4631 4632
			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;
			}
4633
			eaten = tcp_queue_rcv(sk, skb, 0, &fragstolen);
L
Linus Torvalds 已提交
4634
		}
4635
		tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq);
S
Stephen Hemminger 已提交
4636
		if (skb->len)
4637
			tcp_event_data_recv(sk, skb);
4638
		if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)
4639
			tcp_fin(sk);
L
Linus Torvalds 已提交
4640

4641
		if (!RB_EMPTY_ROOT(&tp->out_of_order_queue)) {
L
Linus Torvalds 已提交
4642 4643 4644 4645 4646
			tcp_ofo_queue(sk);

			/* RFC2581. 4.2. SHOULD send immediate ACK, when
			 * gap in queue is filled.
			 */
4647
			if (RB_EMPTY_ROOT(&tp->out_of_order_queue))
4648
				inet_csk(sk)->icsk_ack.pingpong = 0;
L
Linus Torvalds 已提交
4649 4650 4651 4652 4653
		}

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

4654
		tcp_fast_path_check(sk);
L
Linus Torvalds 已提交
4655

4656 4657
		if (eaten > 0)
			kfree_skb_partial(skb, fragstolen);
4658
		if (!sock_flag(sk, SOCK_DEAD))
4659
			sk->sk_data_ready(sk);
L
Linus Torvalds 已提交
4660 4661 4662 4663 4664
		return;
	}

	if (!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) {
		/* A retransmit, 2nd most common case.  Force an immediate ack. */
4665
		NET_INC_STATS(sock_net(sk), LINUX_MIB_DELAYEDACKLOST);
4666
		tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq);
L
Linus Torvalds 已提交
4667 4668

out_of_window:
4669 4670
		tcp_enter_quickack_mode(sk);
		inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
4671
drop:
4672
		tcp_drop(sk, skb);
L
Linus Torvalds 已提交
4673 4674 4675 4676 4677 4678 4679
		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;

4680
	tcp_enter_quickack_mode(sk);
L
Linus Torvalds 已提交
4681 4682 4683 4684 4685 4686 4687

	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);

4688
		tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, tp->rcv_nxt);
4689

L
Linus Torvalds 已提交
4690 4691 4692 4693 4694 4695 4696 4697
		/* 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
4698
	tcp_data_queue_ofo(sk, skb);
L
Linus Torvalds 已提交
4699 4700
}

4701 4702 4703 4704 4705 4706 4707 4708
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);
}

4709
static struct sk_buff *tcp_collapse_one(struct sock *sk, struct sk_buff *skb,
4710 4711
					struct sk_buff_head *list,
					struct rb_root *root)
4712
{
4713
	struct sk_buff *next = tcp_skb_next(skb, list);
4714

4715 4716 4717 4718
	if (list)
		__skb_unlink(skb, list);
	else
		rb_erase(&skb->rbnode, root);
4719 4720

	__kfree_skb(skb);
4721
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRCVCOLLAPSED);
4722 4723 4724 4725

	return next;
}

4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744
/* 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
4745 4746
/* Collapse contiguous sequence of skbs head..tail with
 * sequence numbers start..end.
4747
 *
4748
 * If tail is NULL, this means until the end of the queue.
4749
 *
L
Linus Torvalds 已提交
4750 4751 4752 4753
 * Segments with FIN/SYN are not collapsed (only because this
 * simplifies code)
 */
static void
4754 4755
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
4756
{
4757 4758
	struct sk_buff *skb = head, *n;
	struct sk_buff_head tmp;
4759
	bool end_of_skbs;
L
Linus Torvalds 已提交
4760

S
Stephen Hemminger 已提交
4761
	/* First, check that queue is collapsible and find
4762 4763
	 * the point where collapsing can be useful.
	 */
4764
restart:
4765 4766 4767
	for (end_of_skbs = true; skb != NULL && skb != tail; skb = n) {
		n = tcp_skb_next(skb, list);

L
Linus Torvalds 已提交
4768 4769
		/* No new bits? It is possible on ofo queue. */
		if (!before(start, TCP_SKB_CB(skb)->end_seq)) {
4770
			skb = tcp_collapse_one(sk, skb, list, root);
4771 4772 4773
			if (!skb)
				break;
			goto restart;
L
Linus Torvalds 已提交
4774 4775 4776 4777 4778 4779 4780
		}

		/* The first skb to collapse is:
		 * - not SYN/FIN and
		 * - bloated or contains data before "start" or
		 *   overlaps to the next one.
		 */
4781
		if (!(TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)) &&
L
Linus Torvalds 已提交
4782
		    (tcp_win_from_space(skb->truesize) > skb->len ||
4783 4784
		     before(TCP_SKB_CB(skb)->seq, start))) {
			end_of_skbs = false;
L
Linus Torvalds 已提交
4785
			break;
4786 4787
		}

4788 4789 4790 4791
		if (n && n != tail &&
		    TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(n)->seq) {
			end_of_skbs = false;
			break;
4792
		}
L
Linus Torvalds 已提交
4793 4794 4795 4796

		/* Decided to skip this, advance start seq. */
		start = TCP_SKB_CB(skb)->end_seq;
	}
4797 4798
	if (end_of_skbs ||
	    (TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)))
L
Linus Torvalds 已提交
4799 4800
		return;

4801 4802
	__skb_queue_head_init(&tmp);

L
Linus Torvalds 已提交
4803
	while (before(start, end)) {
4804
		int copy = min_t(int, SKB_MAX_ORDER(0, 0), end - start);
L
Linus Torvalds 已提交
4805 4806
		struct sk_buff *nskb;

4807
		nskb = alloc_skb(copy, GFP_ATOMIC);
L
Linus Torvalds 已提交
4808
		if (!nskb)
4809
			break;
4810

L
Linus Torvalds 已提交
4811 4812
		memcpy(nskb->cb, skb->cb, sizeof(skb->cb));
		TCP_SKB_CB(nskb)->seq = TCP_SKB_CB(nskb)->end_seq = start;
4813 4814 4815 4816
		if (list)
			__skb_queue_before(list, skb, nskb);
		else
			__skb_queue_tail(&tmp, nskb); /* defer rbtree insertion */
4817
		skb_set_owner_r(nskb, sk);
L
Linus Torvalds 已提交
4818 4819 4820 4821 4822 4823

		/* 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;

4824
			BUG_ON(offset < 0);
L
Linus Torvalds 已提交
4825 4826 4827 4828 4829 4830 4831 4832 4833
			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)) {
4834
				skb = tcp_collapse_one(sk, skb, list, root);
4835 4836
				if (!skb ||
				    skb == tail ||
4837
				    (TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)))
4838
					goto end;
L
Linus Torvalds 已提交
4839 4840 4841
			}
		}
	}
4842 4843 4844
end:
	skb_queue_walk_safe(&tmp, skb, n)
		tcp_rbtree_insert(root, skb);
L
Linus Torvalds 已提交
4845 4846 4847 4848 4849 4850 4851 4852
}

/* 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);
4853 4854
	struct sk_buff *skb, *head;
	struct rb_node *p;
L
Linus Torvalds 已提交
4855 4856
	u32 start, end;

4857 4858 4859 4860 4861 4862 4863 4864 4865 4866
	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
4867
		return;
4868
	}
L
Linus Torvalds 已提交
4869 4870
	start = TCP_SKB_CB(skb)->seq;
	end = TCP_SKB_CB(skb)->end_seq;
4871

4872 4873
	for (head = skb;;) {
		skb = tcp_skb_next(skb, NULL);
L
Linus Torvalds 已提交
4874

4875 4876 4877
		/* Range is terminated when we see a gap or when
		 * we are at the queue end.
		 */
4878
		if (!skb ||
L
Linus Torvalds 已提交
4879 4880
		    after(TCP_SKB_CB(skb)->seq, end) ||
		    before(TCP_SKB_CB(skb)->end_seq, start)) {
4881
			tcp_collapse(sk, NULL, &tp->out_of_order_queue,
D
David S. Miller 已提交
4882
				     head, skb, start, end);
4883 4884 4885 4886
			goto new_range;
		}

		if (unlikely(before(TCP_SKB_CB(skb)->seq, start)))
L
Linus Torvalds 已提交
4887
			start = TCP_SKB_CB(skb)->seq;
4888
		if (after(TCP_SKB_CB(skb)->end_seq, end))
L
Linus Torvalds 已提交
4889 4890 4891 4892
			end = TCP_SKB_CB(skb)->end_seq;
	}
}

4893
/*
4894 4895 4896 4897 4898 4899 4900 4901
 * 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.
4902
 */
E
Eric Dumazet 已提交
4903
static bool tcp_prune_ofo_queue(struct sock *sk)
4904 4905
{
	struct tcp_sock *tp = tcp_sk(sk);
4906
	struct rb_node *node, *prev;
4907

4908
	if (RB_EMPTY_ROOT(&tp->out_of_order_queue))
4909
		return false;
4910

4911
	NET_INC_STATS(sock_net(sk), LINUX_MIB_OFOPRUNED);
4912 4913 4914 4915 4916
	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));
4917
		sk_mem_reclaim(sk);
4918 4919 4920
		if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
		    !tcp_under_memory_pressure(sk))
			break;
4921 4922 4923
		node = prev;
	} while (node);
	tp->ooo_last_skb = rb_entry(prev, struct sk_buff, rbnode);
4924 4925 4926 4927 4928 4929 4930 4931 4932

	/* 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;
4933 4934
}

L
Linus Torvalds 已提交
4935 4936 4937 4938 4939 4940 4941 4942 4943
/* 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)
{
4944
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
4945 4946 4947

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

4948
	NET_INC_STATS(sock_net(sk), LINUX_MIB_PRUNECALLED);
L
Linus Torvalds 已提交
4949 4950

	if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
4951
		tcp_clamp_window(sk);
4952
	else if (tcp_under_memory_pressure(sk))
L
Linus Torvalds 已提交
4953 4954 4955
		tp->rcv_ssthresh = min(tp->rcv_ssthresh, 4U * tp->advmss);

	tcp_collapse_ofo_queue(sk);
4956
	if (!skb_queue_empty(&sk->sk_receive_queue))
4957
		tcp_collapse(sk, &sk->sk_receive_queue, NULL,
4958 4959 4960
			     skb_peek(&sk->sk_receive_queue),
			     NULL,
			     tp->copied_seq, tp->rcv_nxt);
4961
	sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
4962 4963 4964 4965 4966 4967 4968

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

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

4969
	tcp_prune_ofo_queue(sk);
L
Linus Torvalds 已提交
4970 4971 4972 4973 4974 4975 4976 4977

	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.
	 */
4978
	NET_INC_STATS(sock_net(sk), LINUX_MIB_RCVPRUNED);
L
Linus Torvalds 已提交
4979 4980 4981 4982 4983 4984

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

E
Eric Dumazet 已提交
4985
static bool tcp_should_expand_sndbuf(const struct sock *sk)
4986
{
4987
	const struct tcp_sock *tp = tcp_sk(sk);
4988

4989 4990 4991 4992
	/* 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
4993
		return false;
4994 4995

	/* If we are under global TCP memory pressure, do not expand.  */
4996
	if (tcp_under_memory_pressure(sk))
E
Eric Dumazet 已提交
4997
		return false;
4998 4999

	/* If we are under soft global TCP memory pressure, do not expand.  */
5000
	if (sk_memory_allocated(sk) >= sk_prot_mem_limits(sk, 0))
E
Eric Dumazet 已提交
5001
		return false;
5002 5003

	/* If we filled the congestion window, do not expand.  */
5004
	if (tcp_packets_in_flight(tp) >= tp->snd_cwnd)
E
Eric Dumazet 已提交
5005
		return false;
5006

E
Eric Dumazet 已提交
5007
	return true;
5008
}
L
Linus Torvalds 已提交
5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019

/* 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);

5020
	if (tcp_should_expand_sndbuf(sk)) {
E
Eric Dumazet 已提交
5021
		tcp_sndbuf_expand(sk);
L
Linus Torvalds 已提交
5022 5023 5024 5025 5026 5027
		tp->snd_cwnd_stamp = tcp_time_stamp;
	}

	sk->sk_write_space(sk);
}

S
Stephen Hemminger 已提交
5028
static void tcp_check_space(struct sock *sk)
L
Linus Torvalds 已提交
5029 5030 5031
{
	if (sock_flag(sk, SOCK_QUEUE_SHRUNK)) {
		sock_reset_flag(sk, SOCK_QUEUE_SHRUNK);
5032
		/* pairs with tcp_poll() */
5033
		smp_mb();
L
Linus Torvalds 已提交
5034
		if (sk->sk_socket &&
5035
		    test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
L
Linus Torvalds 已提交
5036
			tcp_new_space(sk);
5037 5038 5039
			if (!test_bit(SOCK_NOSPACE, &sk->sk_socket->flags))
				tcp_chrono_stop(sk, TCP_CHRONO_SNDBUF_LIMITED);
		}
L
Linus Torvalds 已提交
5040 5041 5042
	}
}

5043
static inline void tcp_data_snd_check(struct sock *sk)
L
Linus Torvalds 已提交
5044
{
5045
	tcp_push_pending_frames(sk);
L
Linus Torvalds 已提交
5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056
	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... */
5057
	if (((tp->rcv_nxt - tp->rcv_wup) > inet_csk(sk)->icsk_ack.rcv_mss &&
L
Linus Torvalds 已提交
5058 5059 5060
	     /* ... and right edge of window advances far enough.
	      * (tcp_recvmsg() will send ACK otherwise). Or...
	      */
5061
	     __tcp_select_window(sk) >= tp->rcv_wnd) ||
L
Linus Torvalds 已提交
5062
	    /* We ACK each frame or... */
5063
	    tcp_in_quickack_mode(sk) ||
L
Linus Torvalds 已提交
5064
	    /* We have out of order data. */
5065
	    (ofo_possible && !RB_EMPTY_ROOT(&tp->out_of_order_queue))) {
L
Linus Torvalds 已提交
5066 5067 5068 5069 5070 5071 5072 5073
		/* Then ack it now */
		tcp_send_ack(sk);
	} else {
		/* Else, send delayed ack. */
		tcp_send_delayed_ack(sk);
	}
}

S
Stephen Hemminger 已提交
5074
static inline void tcp_ack_snd_check(struct sock *sk)
L
Linus Torvalds 已提交
5075
{
5076
	if (!inet_csk_ack_scheduled(sk)) {
L
Linus Torvalds 已提交
5077 5078 5079 5080 5081 5082 5083 5084
		/* 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
5085
 *	signaled. Its the 'slow' part of tcp_urg. It could be
L
Linus Torvalds 已提交
5086 5087 5088 5089 5090 5091
 *	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).
 */
5092

5093
static void tcp_check_urg(struct sock *sk, const struct tcphdr *th)
L
Linus Torvalds 已提交
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 5124 5125 5126 5127 5128 5129
{
	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
5130
	 * or we break the semantics of SIOCATMARK (and thus sockatmark())
L
Linus Torvalds 已提交
5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141
	 *
	 * 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 &&
5142
	    !sock_flag(sk, SOCK_URGINLINE) && tp->copied_seq != tp->rcv_nxt) {
L
Linus Torvalds 已提交
5143 5144 5145
		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
5146
			__skb_unlink(skb, &sk->sk_receive_queue);
L
Linus Torvalds 已提交
5147 5148 5149 5150
			__kfree_skb(skb);
		}
	}

5151 5152
	tp->urg_data = TCP_URG_NOTYET;
	tp->urg_seq = ptr;
L
Linus Torvalds 已提交
5153 5154 5155 5156 5157 5158

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

/* This is the 'fast' part of urgent handling. */
5159
static void tcp_urg(struct sock *sk, struct sk_buff *skb, const struct tcphdr *th)
L
Linus Torvalds 已提交
5160 5161 5162 5163 5164
{
	struct tcp_sock *tp = tcp_sk(sk);

	/* Check if we get a new urgent pointer - normally not. */
	if (th->urg)
5165
		tcp_check_urg(sk, th);
L
Linus Torvalds 已提交
5166 5167 5168 5169 5170 5171

	/* 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;

5172
		/* Is the urgent pointer pointing into this packet? */
L
Linus Torvalds 已提交
5173 5174 5175 5176 5177 5178
		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))
5179
				sk->sk_data_ready(sk);
L
Linus Torvalds 已提交
5180 5181 5182 5183 5184 5185 5186 5187 5188 5189
		}
	}
}

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;

5190
	if (skb_csum_unnecessary(skb))
5191
		err = skb_copy_datagram_msg(skb, hlen, tp->ucopy.msg, chunk);
L
Linus Torvalds 已提交
5192
	else
5193
		err = skb_copy_and_csum_datagram_msg(skb, hlen, tp->ucopy.msg);
L
Linus Torvalds 已提交
5194 5195 5196 5197 5198 5199 5200 5201 5202 5203

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

	return err;
}

5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220
/* 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));
}

5221 5222 5223
/* Does PAWS and seqno based validation of an incoming segment, flags will
 * play significant role here.
 */
E
Eric Dumazet 已提交
5224 5225
static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
				  const struct tcphdr *th, int syn_inerr)
5226 5227
{
	struct tcp_sock *tp = tcp_sk(sk);
5228
	bool rst_seq_match = false;
5229 5230

	/* RFC1323: H1. Apply PAWS check first. */
C
Christoph Paasch 已提交
5231
	if (tcp_fast_parse_options(skb, th, tp) && tp->rx_opt.saw_tstamp &&
5232 5233
	    tcp_paws_discard(sk, skb)) {
		if (!th->rst) {
5234
			NET_INC_STATS(sock_net(sk), LINUX_MIB_PAWSESTABREJECTED);
5235 5236 5237 5238
			if (!tcp_oow_rate_limited(sock_net(sk), skb,
						  LINUX_MIB_TCPACKSKIPPEDPAWS,
						  &tp->last_oow_ack_time))
				tcp_send_dupack(sk, skb);
5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251
			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)".
		 */
5252 5253 5254
		if (!th->rst) {
			if (th->syn)
				goto syn_challenge;
5255 5256 5257 5258
			if (!tcp_oow_rate_limited(sock_net(sk), skb,
						  LINUX_MIB_TCPACKSKIPPEDSEQ,
						  &tp->last_oow_ack_time))
				tcp_send_dupack(sk, skb);
5259 5260
		} else if (tcp_reset_check(sk, skb)) {
			tcp_reset(sk);
5261
		}
5262 5263 5264 5265 5266
		goto discard;
	}

	/* Step 2: check RST bit */
	if (th->rst) {
5267 5268 5269 5270
		/* 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,
5271
		 * then
E
Eric Dumazet 已提交
5272 5273 5274 5275
		 *     RESET the connection
		 * else
		 *     Send a challenge ACK
		 */
5276 5277
		if (TCP_SKB_CB(skb)->seq == tp->rcv_nxt ||
		    tcp_reset_check(sk, skb)) {
5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295
			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
5296
			tcp_reset(sk);
5297 5298 5299 5300 5301 5302 5303
		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)
5304
				tcp_fastopen_active_disable(sk);
5305
			tcp_send_challenge_ack(sk, skb);
5306
		}
5307 5308 5309 5310 5311
		goto discard;
	}

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

E
Eric Dumazet 已提交
5312
	/* step 4: Check for a SYN
5313
	 * RFC 5961 4.2 : Send a challenge ack
E
Eric Dumazet 已提交
5314 5315
	 */
	if (th->syn) {
5316
syn_challenge:
5317
		if (syn_inerr)
5318 5319
			TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPSYNCHALLENGE);
5320
		tcp_send_challenge_ack(sk, skb);
E
Eric Dumazet 已提交
5321
		goto discard;
5322 5323
	}

E
Eric Dumazet 已提交
5324
	return true;
5325 5326

discard:
5327
	tcp_drop(sk, skb);
E
Eric Dumazet 已提交
5328
	return false;
5329 5330
}

L
Linus Torvalds 已提交
5331
/*
5332
 *	TCP receive function for the ESTABLISHED state.
L
Linus Torvalds 已提交
5333
 *
5334
 *	It is split into a fast path and a slow path. The fast path is
L
Linus Torvalds 已提交
5335 5336
 * 	disabled when:
 *	- A zero window was announced from us - zero window probing
5337
 *        is only handled properly in the slow path.
L
Linus Torvalds 已提交
5338 5339 5340 5341
 *	- Out of order segments arrived.
 *	- Urgent data is expected.
 *	- There is no buffer space left
 *	- Unexpected TCP flags/window values/header lengths are received
5342
 *	  (detected by checking the TCP header against pred_flags)
L
Linus Torvalds 已提交
5343 5344 5345 5346 5347
 *	- 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.
 *
5348
 *	When these conditions are not satisfied it drops into a standard
L
Linus Torvalds 已提交
5349 5350
 *	receive procedure patterned after RFC793 to handle all cases.
 *	The first three cases are guaranteed by proper pred_flags setting,
5351
 *	the rest is checked inline. Fast processing is turned on in
L
Linus Torvalds 已提交
5352 5353
 *	tcp_data_queue when everything is OK.
 */
5354 5355
void tcp_rcv_established(struct sock *sk, struct sk_buff *skb,
			 const struct tcphdr *th, unsigned int len)
L
Linus Torvalds 已提交
5356 5357 5358
{
	struct tcp_sock *tp = tcp_sk(sk);

E
Eric Dumazet 已提交
5359
	skb_mstamp_get(&tp->tcp_mstamp);
5360
	if (unlikely(!sk->sk_rx_dst))
E
Eric Dumazet 已提交
5361
		inet_csk(sk)->icsk_af_ops->sk_rx_dst_set(sk, skb);
L
Linus Torvalds 已提交
5362 5363
	/*
	 *	Header prediction.
5364
	 *	The code loosely follows the one in the famous
L
Linus Torvalds 已提交
5365
	 *	"30 instruction TCP receive" Van Jacobson mail.
5366 5367
	 *
	 *	Van's trick is to deposit buffers into socket queue
L
Linus Torvalds 已提交
5368 5369 5370 5371
	 *	on a device interrupt, to call tcp_recv function
	 *	on the receive process context and checksum and copy
	 *	the buffer to user space. smart...
	 *
5372
	 *	Our current scheme is not silly either but we take the
L
Linus Torvalds 已提交
5373 5374 5375 5376 5377 5378 5379
	 *	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
5380
	 *	if header_prediction is to be made
L
Linus Torvalds 已提交
5381 5382
	 *	'S' will always be tp->tcp_header_len >> 2
	 *	'?' will be 0 for the fast path, otherwise pred_flags is 0 to
5383
	 *  turn it off	(when there are holes in the receive
L
Linus Torvalds 已提交
5384 5385 5386 5387 5388
	 *	 space for instance)
	 *	PSH flag is ignored.
	 */

	if ((tcp_flag_word(th) & TCP_HP_BITS) == tp->pred_flags &&
5389 5390
	    TCP_SKB_CB(skb)->seq == tp->rcv_nxt &&
	    !after(TCP_SKB_CB(skb)->ack_seq, tp->snd_nxt)) {
L
Linus Torvalds 已提交
5391 5392 5393 5394 5395 5396 5397 5398 5399 5400
		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! */
5401
			if (!tcp_parse_aligned_timestamp(tp, th))
L
Linus Torvalds 已提交
5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430
				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);
5431
				__kfree_skb(skb);
5432
				tcp_data_snd_check(sk);
5433
				return;
L
Linus Torvalds 已提交
5434
			} else { /* Header too small */
5435
				TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
5436 5437 5438 5439
				goto discard;
			}
		} else {
			int eaten = 0;
5440
			bool fragstolen = false;
L
Linus Torvalds 已提交
5441

D
Dan Williams 已提交
5442 5443 5444 5445 5446
			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
5447

D
Dan Williams 已提交
5448
				if (!tcp_copy_to_iovec(sk, skb, tcp_header_len)) {
L
Linus Torvalds 已提交
5449 5450 5451 5452 5453 5454 5455 5456 5457 5458
					/* 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);

5459
					tcp_rcv_rtt_measure_ts(sk, skb);
L
Linus Torvalds 已提交
5460 5461

					__skb_pull(skb, tcp_header_len);
5462
					tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq);
5463
					NET_INC_STATS(sock_net(sk),
5464
							LINUX_MIB_TCPHPHITSTOUSER);
D
Dan Williams 已提交
5465
					eaten = 1;
L
Linus Torvalds 已提交
5466 5467 5468
				}
			}
			if (!eaten) {
5469
				if (tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
5470 5471
					goto csum_error;

5472 5473 5474
				if ((int)skb->truesize > sk->sk_forward_alloc)
					goto step5;

L
Linus Torvalds 已提交
5475 5476 5477 5478 5479 5480 5481 5482 5483
				/* 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);

5484
				tcp_rcv_rtt_measure_ts(sk, skb);
L
Linus Torvalds 已提交
5485

5486
				NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPHPHITS);
L
Linus Torvalds 已提交
5487 5488

				/* Bulk data transfer: receiver */
5489 5490
				eaten = tcp_queue_rcv(sk, skb, tcp_header_len,
						      &fragstolen);
L
Linus Torvalds 已提交
5491 5492
			}

5493
			tcp_event_data_recv(sk, skb);
L
Linus Torvalds 已提交
5494 5495 5496 5497

			if (TCP_SKB_CB(skb)->ack_seq != tp->snd_una) {
				/* Well, only one small jumplet in fast path... */
				tcp_ack(sk, skb, FLAG_DATA);
5498
				tcp_data_snd_check(sk);
5499
				if (!inet_csk_ack_scheduled(sk))
L
Linus Torvalds 已提交
5500 5501 5502
					goto no_ack;
			}

D
Dan Williams 已提交
5503
			__tcp_ack_snd_check(sk, 0);
L
Linus Torvalds 已提交
5504 5505
no_ack:
			if (eaten)
5506
				kfree_skb_partial(skb, fragstolen);
5507
			sk->sk_data_ready(sk);
5508
			return;
L
Linus Torvalds 已提交
5509 5510 5511 5512
		}
	}

slow_path:
5513
	if (len < (th->doff << 2) || tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
5514 5515
		goto csum_error;

5516
	if (!th->ack && !th->rst && !th->syn)
5517 5518
		goto discard;

L
Linus Torvalds 已提交
5519 5520 5521 5522
	/*
	 *	Standard slow path.
	 */

E
Eric Dumazet 已提交
5523
	if (!tcp_validate_incoming(sk, skb, th, 1))
5524
		return;
L
Linus Torvalds 已提交
5525 5526

step5:
5527
	if (tcp_ack(sk, skb, FLAG_SLOWPATH | FLAG_UPDATE_TS_RECENT) < 0)
5528
		goto discard;
L
Linus Torvalds 已提交
5529

5530
	tcp_rcv_rtt_measure_ts(sk, skb);
L
Linus Torvalds 已提交
5531 5532 5533 5534 5535 5536 5537

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

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

5538
	tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
5539
	tcp_ack_snd_check(sk);
5540
	return;
L
Linus Torvalds 已提交
5541 5542

csum_error:
5543 5544
	TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS);
	TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
5545 5546

discard:
5547
	tcp_drop(sk, skb);
L
Linus Torvalds 已提交
5548
}
E
Eric Dumazet 已提交
5549
EXPORT_SYMBOL(tcp_rcv_established);
L
Linus Torvalds 已提交
5550

P
Pavel Emelyanov 已提交
5551 5552 5553 5554 5555 5556
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);
5557
	icsk->icsk_ack.lrcvtime = tcp_time_stamp;
P
Pavel Emelyanov 已提交
5558

5559
	if (skb) {
E
Eric Dumazet 已提交
5560
		icsk->icsk_af_ops->sk_rx_dst_set(sk, skb);
P
Pavel Emelyanov 已提交
5561
		security_inet_conn_established(sk, skb);
D
David S. Miller 已提交
5562
	}
P
Pavel Emelyanov 已提交
5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573

	/* 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.
	 */
5574
	tp->lsndtime = tcp_jiffies32;
P
Pavel Emelyanov 已提交
5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587

	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;

}

5588 5589 5590 5591
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);
5592
	struct sk_buff *data = tp->syn_data ? tcp_write_queue_head(sk) : NULL;
5593 5594
	u16 mss = tp->rx_opt.mss_clamp, try_exp = 0;
	bool syn_drop = false;
5595 5596 5597 5598 5599 5600 5601

	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
5602
		tcp_parse_options(synack, &opt, 0, NULL);
5603 5604 5605
		mss = opt.mss_clamp;
	}

5606 5607
	if (!tp->syn_fastopen) {
		/* Ignore an unsolicited cookie */
5608
		cookie->len = -1;
5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622
	} 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;
	}
5623

5624
	tcp_fastopen_cache_set(sk, mss, cookie, syn_drop, try_exp);
5625 5626

	if (data) { /* Retransmit unacked data in SYN */
5627 5628
		tcp_for_write_queue_from(data, sk) {
			if (data == tcp_send_head(sk) ||
5629
			    __tcp_retransmit_skb(sk, data, 1))
5630 5631
				break;
		}
5632
		tcp_rearm_rto(sk);
5633
		NET_INC_STATS(sock_net(sk),
5634
				LINUX_MIB_TCPFASTOPENACTIVEFAIL);
5635 5636
		return true;
	}
Y
Yuchung Cheng 已提交
5637
	tp->syn_data_acked = tp->syn_data;
5638
	if (tp->syn_data_acked)
5639
		NET_INC_STATS(sock_net(sk),
5640
				LINUX_MIB_TCPFASTOPENACTIVE);
5641 5642 5643

	tcp_fastopen_add_skb(sk, synack);

5644 5645 5646
	return false;
}

L
Linus Torvalds 已提交
5647
static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
5648
					 const struct tcphdr *th)
L
Linus Torvalds 已提交
5649
{
5650
	struct inet_connection_sock *icsk = inet_csk(sk);
5651
	struct tcp_sock *tp = tcp_sk(sk);
5652
	struct tcp_fastopen_cookie foc = { .len = -1 };
5653
	int saved_clamp = tp->rx_opt.mss_clamp;
5654
	bool fastopen_fail;
L
Linus Torvalds 已提交
5655

C
Christoph Paasch 已提交
5656
	tcp_parse_options(skb, &tp->rx_opt, 0, &foc);
5657
	if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr)
5658
		tp->rx_opt.rcv_tsecr -= tp->tsoffset;
L
Linus Torvalds 已提交
5659 5660 5661 5662 5663 5664 5665 5666 5667 5668

	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)"
		 */
5669 5670
		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
5671 5672 5673 5674 5675
			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)) {
5676
			NET_INC_STATS(sock_net(sk),
5677
					LINUX_MIB_PAWSACTIVEREJECTED);
L
Linus Torvalds 已提交
5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710
			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..."
		 */

5711
		tcp_ecn_rcv_synack(tp, th);
L
Linus Torvalds 已提交
5712

5713
		tcp_init_wl(tp, TCP_SKB_CB(skb)->seq);
L
Linus Torvalds 已提交
5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735 5736 5737 5738 5739 5740 5741
		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);
		}

5742 5743
		if (tcp_is_sack(tp) && sysctl_tcp_fack)
			tcp_enable_fack(tp);
L
Linus Torvalds 已提交
5744

J
John Heffner 已提交
5745
		tcp_mtup_init(sk);
5746
		tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
5747 5748 5749 5750 5751 5752
		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;
5753

R
Ralf Baechle 已提交
5754
		smp_mb();
L
Linus Torvalds 已提交
5755

P
Pavel Emelyanov 已提交
5756
		tcp_finish_connect(sk, skb);
L
Linus Torvalds 已提交
5757

5758 5759
		fastopen_fail = (tp->syn_fastopen || tp->syn_data) &&
				tcp_rcv_fastopen_synack(sk, skb, &foc);
5760

5761 5762 5763 5764 5765 5766
		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;
5767 5768 5769
		if (sk->sk_write_pending ||
		    icsk->icsk_accept_queue.rskq_defer_accept ||
		    icsk->icsk_ack.pingpong) {
L
Linus Torvalds 已提交
5770 5771 5772 5773 5774 5775 5776
			/* 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
			 */
5777 5778
			inet_csk_schedule_ack(sk);
			tcp_enter_quickack_mode(sk);
5779 5780
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
						  TCP_DELACK_MAX, TCP_RTO_MAX);
L
Linus Torvalds 已提交
5781 5782

discard:
5783
			tcp_drop(sk, skb);
L
Linus Torvalds 已提交
5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803
			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. */
5804
	if (tp->rx_opt.ts_recent_stamp && tp->rx_opt.saw_tstamp &&
I
Ilpo Järvinen 已提交
5805
	    tcp_paws_reject(&tp->rx_opt, 0))
L
Linus Torvalds 已提交
5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824
		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;
5825
		tp->copied_seq = tp->rcv_nxt;
L
Linus Torvalds 已提交
5826 5827 5828 5829 5830 5831 5832 5833 5834
		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;

5835
		tcp_ecn_rcv_syn(tp, th);
L
Linus Torvalds 已提交
5836

J
John Heffner 已提交
5837
		tcp_mtup_init(sk);
5838
		tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
5839 5840 5841 5842 5843
		tcp_initialize_rcv_mss(sk);

		tcp_send_synack(sk);
#if 0
		/* Note, we could accept data and URG from this segment.
5844 5845 5846
		 * 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
5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875
		 *
		 * 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
5876
 *	all states except ESTABLISHED and TIME_WAIT.
L
Linus Torvalds 已提交
5877 5878 5879
 *	It's called from both tcp_v4_rcv and tcp_v6_rcv and should be
 *	address independent.
 */
5880

5881
int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
5882 5883
{
	struct tcp_sock *tp = tcp_sk(sk);
5884
	struct inet_connection_sock *icsk = inet_csk(sk);
5885
	const struct tcphdr *th = tcp_hdr(skb);
5886
	struct request_sock *req;
L
Linus Torvalds 已提交
5887
	int queued = 0;
5888
	bool acceptable;
L
Linus Torvalds 已提交
5889 5890 5891 5892 5893 5894

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

	case TCP_LISTEN:
S
Stephen Hemminger 已提交
5895
		if (th->ack)
L
Linus Torvalds 已提交
5896 5897
			return 1;

S
Stephen Hemminger 已提交
5898
		if (th->rst)
L
Linus Torvalds 已提交
5899 5900
			goto discard;

S
Stephen Hemminger 已提交
5901
		if (th->syn) {
E
Eric Dumazet 已提交
5902 5903
			if (th->fin)
				goto discard;
5904 5905 5906 5907 5908 5909
			/* 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
5910

5911 5912
			if (!acceptable)
				return 1;
5913
			consume_skb(skb);
5914
			return 0;
L
Linus Torvalds 已提交
5915 5916 5917 5918
		}
		goto discard;

	case TCP_SYN_SENT:
5919
		tp->rx_opt.saw_tstamp = 0;
E
Eric Dumazet 已提交
5920
		skb_mstamp_get(&tp->tcp_mstamp);
5921
		queued = tcp_rcv_synsent_state_process(sk, skb, th);
L
Linus Torvalds 已提交
5922 5923 5924 5925 5926 5927
		if (queued >= 0)
			return queued;

		/* Do step6 onward by hand. */
		tcp_urg(sk, skb, th);
		__kfree_skb(skb);
5928
		tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
5929 5930 5931
		return 0;
	}

E
Eric Dumazet 已提交
5932
	skb_mstamp_get(&tp->tcp_mstamp);
5933
	tp->rx_opt.saw_tstamp = 0;
5934
	req = tp->fastopen_rsk;
5935
	if (req) {
5936
		WARN_ON_ONCE(sk->sk_state != TCP_SYN_RECV &&
5937 5938
		    sk->sk_state != TCP_FIN_WAIT1);

5939
		if (!tcp_check_req(sk, skb, req, true))
5940
			goto discard;
5941
	}
5942

5943
	if (!th->ack && !th->rst && !th->syn)
5944 5945
		goto discard;

5946
	if (!tcp_validate_incoming(sk, skb, th, 0))
E
Eric Dumazet 已提交
5947
		return 0;
L
Linus Torvalds 已提交
5948 5949

	/* step 5: check the ACK field */
5950 5951
	acceptable = tcp_ack(sk, skb, FLAG_SLOWPATH |
				      FLAG_UPDATE_TS_RECENT) > 0;
5952

5953 5954
	switch (sk->sk_state) {
	case TCP_SYN_RECV:
5955 5956
		if (!acceptable)
			return 1;
L
Linus Torvalds 已提交
5957

Y
Yuchung Cheng 已提交
5958 5959 5960
		if (!tp->srtt_us)
			tcp_synack_rtt_meas(sk, req);

5961 5962 5963 5964
		/* Once we leave TCP_SYN_RECV, we no longer need req
		 * so release it.
		 */
		if (req) {
5965
			inet_csk(sk)->icsk_retransmits = 0;
5966 5967 5968 5969 5970 5971 5972 5973
			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
5974
			tcp_init_buffer_space(sk);
5975 5976 5977 5978
		}
		smp_mb();
		tcp_set_state(sk, TCP_ESTABLISHED);
		sk->sk_state_change(sk);
L
Linus Torvalds 已提交
5979

5980 5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999 6000 6001
		/* 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.
6002
			 */
6003 6004 6005
			tcp_rearm_rto(sk);
		} else
			tcp_init_metrics(sk);
L
Linus Torvalds 已提交
6006

6007 6008
		if (!inet_csk(sk)->icsk_ca_ops->cong_control)
			tcp_update_pacing_rate(sk);
6009

6010
		/* Prevent spurious tcp_cwnd_restart() on first data packet */
6011
		tp->lsndtime = tcp_jiffies32;
6012 6013 6014

		tcp_initialize_rcv_mss(sk);
		tcp_fast_path_on(tp);
6015 6016
		break;

6017 6018 6019
	case TCP_FIN_WAIT1: {
		int tmo;

6020 6021 6022 6023 6024
		/* 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.
		 */
6025
		if (req) {
6026 6027 6028 6029 6030
			/* 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.
6031
			 */
6032 6033 6034 6035 6036 6037
			if (!acceptable)
				return 1;
			/* We no longer need the request sock. */
			reqsk_fastopen_remove(sk, req, false);
			tcp_rearm_rto(sk);
		}
6038 6039
		if (tp->snd_una != tp->write_seq)
			break;
6040

6041 6042
		tcp_set_state(sk, TCP_FIN_WAIT2);
		sk->sk_shutdown |= SEND_SHUTDOWN;
6043

6044
		sk_dst_confirm(sk);
6045

6046 6047 6048 6049 6050
		if (!sock_flag(sk, SOCK_DEAD)) {
			/* Wake up lingering close() */
			sk->sk_state_change(sk);
			break;
		}
L
Linus Torvalds 已提交
6051

6052 6053 6054 6055 6056 6057 6058 6059 6060
		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)
6061
				tcp_fastopen_active_disable(sk);
6062
			tcp_done(sk);
6063
			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONDATA);
6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080
			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;
6081 6082
		}
		break;
6083
	}
L
Linus Torvalds 已提交
6084

6085 6086 6087 6088 6089 6090
	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
6091

6092 6093 6094 6095 6096
	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
6097
		}
6098
		break;
6099
	}
L
Linus Torvalds 已提交
6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110 6111 6112 6113

	/* 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,
6114
		 * RFC 1122 says we MUST send a reset.
L
Linus Torvalds 已提交
6115 6116 6117 6118 6119
		 * 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)) {
6120
				NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONDATA);
L
Linus Torvalds 已提交
6121 6122 6123 6124 6125
				tcp_reset(sk);
				return 1;
			}
		}
		/* Fall through */
6126
	case TCP_ESTABLISHED:
L
Linus Torvalds 已提交
6127 6128 6129 6130 6131 6132 6133
		tcp_data_queue(sk, skb);
		queued = 1;
		break;
	}

	/* tcp_data could move socket to TIME-WAIT */
	if (sk->sk_state != TCP_CLOSE) {
6134
		tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
6135 6136 6137
		tcp_ack_snd_check(sk);
	}

6138
	if (!queued) {
L
Linus Torvalds 已提交
6139
discard:
6140
		tcp_drop(sk, skb);
L
Linus Torvalds 已提交
6141 6142 6143 6144
	}
	return 0;
}
EXPORT_SYMBOL(tcp_rcv_state_process);
O
Octavian Purdila 已提交
6145 6146 6147 6148 6149 6150

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)
6151 6152
		net_dbg_ratelimited("drop open request from %pI4/%u\n",
				    &ireq->ir_rmt_addr, port);
6153 6154
#if IS_ENABLED(CONFIG_IPV6)
	else if (family == AF_INET6)
6155 6156
		net_dbg_ratelimited("drop open request from %pI6/%u\n",
				    &ireq->ir_v6_rmt_addr, port);
6157
#endif
O
Octavian Purdila 已提交
6158 6159
}

6160 6161 6162 6163 6164
/* 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
6165
 * TCP ECN negotiation.
6166 6167
 *
 * Exception: tcp_ca wants ECN. This is required for DCTCP
6168 6169 6170
 * congestion control: Linux DCTCP asserts ECT on all packets,
 * including SYN, which is most optimal solution; however,
 * others, such as FreeBSD do not.
6171 6172 6173
 */
static void tcp_ecn_create_request(struct request_sock *req,
				   const struct sk_buff *skb,
6174 6175
				   const struct sock *listen_sk,
				   const struct dst_entry *dst)
6176 6177 6178 6179
{
	const struct tcphdr *th = tcp_hdr(skb);
	const struct net *net = sock_net(listen_sk);
	bool th_ecn = th->ece && th->cwr;
6180
	bool ect, ecn_ok;
6181
	u32 ecn_ok_dst;
6182 6183 6184 6185 6186

	if (!th_ecn)
		return;

	ect = !INET_ECN_is_not_ect(TCP_SKB_CB(skb)->ip_dsfield);
6187 6188
	ecn_ok_dst = dst_feature(dst, DST_FEATURE_ECN_MASK);
	ecn_ok = net->ipv4.sysctl_tcp_ecn || ecn_ok_dst;
6189

6190 6191
	if ((!ect && ecn_ok) || tcp_ca_needs_ecn(listen_sk) ||
	    (ecn_ok_dst & DST_FEATURE_ECN_CA))
6192 6193 6194
		inet_rsk(req)->ecn_ok = 1;
}

6195 6196 6197 6198 6199 6200
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);

6201
	req->rsk_rcv_wnd = 0;		/* So that tcp_send_synack() knows! */
6202 6203 6204
	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
6205
	skb_mstamp_get(&tcp_rsk(req)->snt_synack);
6206 6207 6208 6209 6210 6211 6212 6213 6214 6215 6216 6217 6218 6219
	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
6220
struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
6221 6222
				      struct sock *sk_listener,
				      bool attach_listener)
E
Eric Dumazet 已提交
6223
{
6224 6225
	struct request_sock *req = reqsk_alloc(ops, sk_listener,
					       attach_listener);
E
Eric Dumazet 已提交
6226 6227 6228 6229 6230 6231

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

		kmemcheck_annotate_bitfield(ireq, flags);
		ireq->opt = NULL;
6232 6233 6234
#if IS_ENABLED(CONFIG_IPV6)
		ireq->pktopts = NULL;
#endif
E
Eric Dumazet 已提交
6235 6236 6237
		atomic64_set(&ireq->ir_cookie, 0);
		ireq->ireq_state = TCP_NEW_SYN_RECV;
		write_pnet(&ireq->ireq_net, sock_net(sk_listener));
6238
		ireq->ireq_family = sk_listener->sk_family;
E
Eric Dumazet 已提交
6239 6240 6241 6242 6243 6244
	}

	return req;
}
EXPORT_SYMBOL(inet_reqsk_alloc);

6245 6246 6247
/*
 * Return true if a syncookie should be sent
 */
6248
static bool tcp_syn_flood_action(const struct sock *sk,
6249 6250 6251
				 const struct sk_buff *skb,
				 const char *proto)
{
6252
	struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue;
6253 6254
	const char *msg = "Dropping request";
	bool want_cookie = false;
6255
	struct net *net = sock_net(sk);
6256 6257

#ifdef CONFIG_SYN_COOKIES
6258
	if (net->ipv4.sysctl_tcp_syncookies) {
6259 6260
		msg = "Sending cookies";
		want_cookie = true;
6261
		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPREQQFULLDOCOOKIES);
6262 6263
	} else
#endif
6264
		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPREQQFULLDROP);
6265

6266
	if (!queue->synflood_warned &&
6267
	    net->ipv4.sysctl_tcp_syncookies != 2 &&
6268
	    xchg(&queue->synflood_warned, 1) == 0)
6269 6270
		pr_info("%s: Possible SYN flooding on port %d. %s.  Check SNMP counters.\n",
			proto, ntohs(tcp_hdr(skb)->dest), msg);
6271

6272 6273 6274
	return want_cookie;
}

6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 6285 6286 6287 6288 6289 6290 6291
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
6292 6293 6294 6295
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)
{
6296 6297
	struct tcp_fastopen_cookie foc = { .len = -1 };
	__u32 isn = TCP_SKB_CB(skb)->tcp_tw_isn;
O
Octavian Purdila 已提交
6298 6299
	struct tcp_options_received tmp_opt;
	struct tcp_sock *tp = tcp_sk(sk);
6300
	struct net *net = sock_net(sk);
6301
	struct sock *fastopen_sk = NULL;
O
Octavian Purdila 已提交
6302
	struct dst_entry *dst = NULL;
6303 6304
	struct request_sock *req;
	bool want_cookie = false;
O
Octavian Purdila 已提交
6305 6306 6307 6308 6309 6310
	struct flowi fl;

	/* TW buckets are converted to open requests without
	 * limitations, they conserve resources and peer is
	 * evidently real one.
	 */
6311
	if ((net->ipv4.sysctl_tcp_syncookies == 2 ||
O
Octavian Purdila 已提交
6312 6313 6314 6315 6316 6317
	     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;
	}

6318
	if (sk_acceptq_is_full(sk)) {
6319
		NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
O
Octavian Purdila 已提交
6320 6321 6322
		goto drop;
	}

6323
	req = inet_reqsk_alloc(rsk_ops, sk, !want_cookie);
O
Octavian Purdila 已提交
6324 6325 6326 6327
	if (!req)
		goto drop;

	tcp_rsk(req)->af_specific = af_ops;
6328
	tcp_rsk(req)->ts_off = 0;
O
Octavian Purdila 已提交
6329 6330 6331 6332 6333 6334 6335 6336 6337 6338 6339

	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);
6340
	inet_rsk(req)->no_srccheck = inet_sk(sk)->transparent;
O
Octavian Purdila 已提交
6341

6342
	/* Note: tcp_v6_init_req() might override ir_iif for link locals */
6343
	inet_rsk(req)->ir_iif = inet_request_bound_dev_if(sk, skb);
6344

O
Octavian Purdila 已提交
6345 6346 6347 6348 6349
	af_ops->init_req(req, sk, skb);

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

6350 6351
	if (tmp_opt.tstamp_ok)
		tcp_rsk(req)->ts_off = af_ops->init_ts_off(skb);
6352

6353
	if (!want_cookie && !isn) {
O
Octavian Purdila 已提交
6354
		/* Kill the following clause, if you dislike this way. */
6355 6356 6357 6358
		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
6359 6360 6361 6362 6363 6364 6365 6366 6367 6368 6369 6370
			/* 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;
		}

6371
		isn = af_ops->init_seq(skb);
O
Octavian Purdila 已提交
6372 6373
	}
	if (!dst) {
6374
		dst = af_ops->route_req(sk, &fl, req);
O
Octavian Purdila 已提交
6375 6376 6377 6378
		if (!dst)
			goto drop_and_free;
	}

6379 6380 6381 6382 6383 6384 6385 6386 6387
	tcp_ecn_create_request(req, skb, sk, dst);

	if (want_cookie) {
		isn = cookie_init_sequence(af_ops, sk, skb, &req->mss);
		req->cookie_ts = tmp_opt.tstamp_ok;
		if (!tmp_opt.tstamp_ok)
			inet_rsk(req)->ecn_ok = 0;
	}

O
Octavian Purdila 已提交
6388
	tcp_rsk(req)->snt_isn = isn;
6389
	tcp_rsk(req)->txhash = net_tx_rndhash();
O
Octavian Purdila 已提交
6390
	tcp_openreq_init_rwin(req, sk, dst);
6391 6392
	if (!want_cookie) {
		tcp_reqsk_record_syn(sk, req, skb);
6393
		fastopen_sk = tcp_try_fastopen(sk, skb, req, &foc, dst);
6394
	}
6395
	if (fastopen_sk) {
6396
		af_ops->send_synack(fastopen_sk, dst, &fl, req,
6397
				    &foc, TCP_SYNACK_FASTOPEN);
6398 6399 6400 6401
		/* 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);
6402 6403
		sock_put(fastopen_sk);
	} else {
6404
		tcp_rsk(req)->tfo_listener = false;
6405 6406
		if (!want_cookie)
			inet_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
6407 6408 6409
		af_ops->send_synack(sk, dst, &fl, req, &foc,
				    !want_cookie ? TCP_SYNACK_NORMAL :
						   TCP_SYNACK_COOKIE);
6410 6411 6412 6413
		if (want_cookie) {
			reqsk_free(req);
			return 0;
		}
O
Octavian Purdila 已提交
6414
	}
6415
	reqsk_put(req);
O
Octavian Purdila 已提交
6416 6417 6418 6419 6420 6421 6422
	return 0;

drop_and_release:
	dst_release(dst);
drop_and_free:
	reqsk_free(req);
drop:
6423
	tcp_listendrop(sk);
O
Octavian Purdila 已提交
6424 6425 6426
	return 0;
}
EXPORT_SYMBOL(tcp_conn_request);