tcp_input.c 179.9 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
L
Linus Torvalds 已提交
2 3 4 5 6 7 8
/*
 * 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).
 *
9
 * Authors:	Ross Biro
L
Linus Torvalds 已提交
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 43
 *		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
44
 *		Andrey Savochkin:	Fix RTT measurements in the presence of
L
Linus Torvalds 已提交
45 46 47 48 49 50 51
 *					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.
52
 *		Andi Kleen:		Add tcp_measure_rcv_mss to make
L
Linus Torvalds 已提交
53
 *					connections with MSS<min(MTU,ann. MSS)
54
 *					work without delayed acks.
L
Linus Torvalds 已提交
55 56 57 58 59 60 61 62 63 64
 *		Andi Kleen:		Process packets with PSH set in the
 *					fast path.
 *		J Hadi Salim:		ECN support
 *	 	Andrei Gurtov,
 *		Pasi Sarolahti,
 *		Panu Kuhlberg:		Experimental audit of TCP (re)transmission
 *					engine. Lots of bugs are found.
 *		Pasi Sarolahti:		F-RTO for dealing with spurious RTOs
 */

65 66
#define pr_fmt(fmt) "TCP: " fmt

L
Linus Torvalds 已提交
67
#include <linux/mm.h>
68
#include <linux/slab.h>
L
Linus Torvalds 已提交
69 70
#include <linux/module.h>
#include <linux/sysctl.h>
71
#include <linux/kernel.h>
72
#include <linux/prefetch.h>
73
#include <net/dst.h>
L
Linus Torvalds 已提交
74 75 76 77
#include <net/tcp.h>
#include <net/inet_common.h>
#include <linux/ipsec.h>
#include <asm/unaligned.h>
78
#include <linux/errqueue.h>
79
#include <trace/events/tcp.h>
80
#include <linux/static_key.h>
L
Linus Torvalds 已提交
81

82
int sysctl_tcp_max_orphans __read_mostly = NR_FILE;
L
Linus Torvalds 已提交
83 84 85 86 87 88 89 90

#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				*/
91
#define FLAG_LOST_RETRANS	0x80 /* This ACK marks some retransmission lost */
92
#define FLAG_SLOWPATH		0x100 /* Do not skip RFC checks for window update.*/
Y
Yuchung Cheng 已提交
93
#define FLAG_ORIG_SACK_ACKED	0x200 /* Never retransmitted data are (s)acked	*/
94
#define FLAG_SND_UNA_ADVANCED	0x400 /* Snd_una was changed (!= FLAG_DATA_ACKED) */
95
#define FLAG_DSACKING_ACK	0x800 /* SACK blocks contained D-SACK info */
96
#define FLAG_SET_XMIT_TIMER	0x1000 /* Set TLP or RTO timer */
97
#define FLAG_SACK_RENEGING	0x2000 /* snd_una advanced to a sacked seq */
98
#define FLAG_UPDATE_TS_RECENT	0x4000 /* tcp_replace_ts_recent() */
99
#define FLAG_NO_CHALLENGE_ACK	0x8000 /* do not call tcp_send_challenge_ack()	*/
L
Linus Torvalds 已提交
100 101 102

#define FLAG_ACKED		(FLAG_DATA_ACKED|FLAG_SYN_ACKED)
#define FLAG_NOT_DUP		(FLAG_DATA|FLAG_WIN_UPDATE|FLAG_ACKED)
103
#define FLAG_CA_ALERT		(FLAG_DATA_SACKED|FLAG_ECE|FLAG_DSACKING_ACK)
L
Linus Torvalds 已提交
104 105 106
#define FLAG_FORWARD_PROGRESS	(FLAG_ACKED|FLAG_DATA_SACKED)

#define TCP_REMNANT (TCP_FLAG_FIN|TCP_FLAG_URG|TCP_FLAG_SYN|TCP_FLAG_PSH)
107
#define TCP_HP_BITS (~(TCP_RESERVED_BITS|TCP_FLAG_PSH))
L
Linus Torvalds 已提交
108

109 110 111 112
#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 */

113 114
static void tcp_gro_dev_warn(struct sock *sk, const struct sk_buff *skb,
			     unsigned int len)
115 116 117 118 119 120 121 122 123 124
{
	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);
125 126 127
		if (!dev || len >= dev->mtu)
			pr_warn("%s: Driver has suspect GRO implementation, TCP performance may be compromised.\n",
				dev ? dev->name : "Unknown driver");
128 129 130 131
		rcu_read_unlock();
	}
}

132
/* Adapt the MSS value used to make delayed ack decision to the
L
Linus Torvalds 已提交
133
 * real world.
134
 */
135
static void tcp_measure_rcv_mss(struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
136
{
137
	struct inet_connection_sock *icsk = inet_csk(sk);
138
	const unsigned int lss = icsk->icsk_ack.last_seg_size;
139
	unsigned int len;
L
Linus Torvalds 已提交
140

141
	icsk->icsk_ack.last_seg_size = 0;
L
Linus Torvalds 已提交
142 143 144 145

	/* skb->len may jitter because of SACKs, even if peer
	 * sends good full-sized frames.
	 */
146
	len = skb_shinfo(skb)->gso_size ? : skb->len;
147
	if (len >= icsk->icsk_ack.rcv_mss) {
148 149
		icsk->icsk_ack.rcv_mss = min_t(unsigned int, len,
					       tcp_sk(sk)->advmss);
150 151 152 153
		/* 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
154 155 156 157 158 159
	} else {
		/* Otherwise, we make more careful check taking into account,
		 * that SACKs block is variable.
		 *
		 * "len" is invariant segment length, including TCP header.
		 */
160
		len += skb->data - skb_transport_header(skb);
161
		if (len >= TCP_MSS_DEFAULT + sizeof(struct tcphdr) ||
L
Linus Torvalds 已提交
162 163 164 165 166 167
		    /* 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) &&
168
		     !(tcp_flag_word(tcp_hdr(skb)) & TCP_REMNANT))) {
L
Linus Torvalds 已提交
169 170 171 172
			/* Subtract also invariant (if peer is RFC compliant),
			 * tcp header plus fixed timestamp option length.
			 * Resulting "len" is MSS free of SACK jitter.
			 */
173 174
			len -= tcp_sk(sk)->tcp_header_len;
			icsk->icsk_ack.last_seg_size = len;
L
Linus Torvalds 已提交
175
			if (len == lss) {
176
				icsk->icsk_ack.rcv_mss = len;
L
Linus Torvalds 已提交
177 178 179
				return;
			}
		}
180 181
		if (icsk->icsk_ack.pending & ICSK_ACK_PUSHED)
			icsk->icsk_ack.pending |= ICSK_ACK_PUSHED2;
182
		icsk->icsk_ack.pending |= ICSK_ACK_PUSHED;
L
Linus Torvalds 已提交
183 184 185
	}
}

186
static void tcp_incr_quickack(struct sock *sk)
L
Linus Torvalds 已提交
187
{
188
	struct inet_connection_sock *icsk = inet_csk(sk);
189
	unsigned int quickacks = tcp_sk(sk)->rcv_wnd / (2 * icsk->icsk_ack.rcv_mss);
L
Linus Torvalds 已提交
190

191 192
	if (quickacks == 0)
		quickacks = 2;
193 194
	if (quickacks > icsk->icsk_ack.quick)
		icsk->icsk_ack.quick = min(quickacks, TCP_MAX_QUICKACKS);
L
Linus Torvalds 已提交
195 196
}

197
static void tcp_enter_quickack_mode(struct sock *sk)
L
Linus Torvalds 已提交
198
{
199 200 201 202
	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
203 204 205 206 207 208
}

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

209
static bool tcp_in_quickack_mode(struct sock *sk)
L
Linus Torvalds 已提交
210
{
211
	const struct inet_connection_sock *icsk = inet_csk(sk);
212
	const struct dst_entry *dst = __sk_dst_get(sk);
E
Eric Dumazet 已提交
213

214 215
	return (dst && dst_metric(dst, RTAX_QUICKACK)) ||
		(icsk->icsk_ack.quick && !icsk->icsk_ack.pingpong);
L
Linus Torvalds 已提交
216 217
}

218
static void tcp_ecn_queue_cwr(struct tcp_sock *tp)
219
{
220
	if (tp->ecn_flags & TCP_ECN_OK)
221 222 223
		tp->ecn_flags |= TCP_ECN_QUEUE_CWR;
}

224
static void tcp_ecn_accept_cwr(struct tcp_sock *tp, const struct sk_buff *skb)
225 226 227 228 229
{
	if (tcp_hdr(skb)->cwr)
		tp->ecn_flags &= ~TCP_ECN_DEMAND_CWR;
}

230
static void tcp_ecn_withdraw_cwr(struct tcp_sock *tp)
231 232 233 234
{
	tp->ecn_flags &= ~TCP_ECN_DEMAND_CWR;
}

235
static void __tcp_ecn_check_ce(struct tcp_sock *tp, const struct sk_buff *skb)
236
{
237
	switch (TCP_SKB_CB(skb)->ip_dsfield & INET_ECN_MASK) {
238
	case INET_ECN_NOT_ECT:
239
		/* Funny extension: if ECT is not set on a segment,
240 241 242 243
		 * and we already seen ECT on a previous segment,
		 * it is probably a retransmit.
		 */
		if (tp->ecn_flags & TCP_ECN_SEEN)
244
			tcp_enter_quickack_mode((struct sock *)tp);
245 246
		break;
	case INET_ECN_CE:
247 248 249
		if (tcp_ca_needs_ecn((struct sock *)tp))
			tcp_ca_event((struct sock *)tp, CA_EVENT_ECN_IS_CE);

250 251 252 253 254
		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;
		}
255 256
		tp->ecn_flags |= TCP_ECN_SEEN;
		break;
257
	default:
258 259
		if (tcp_ca_needs_ecn((struct sock *)tp))
			tcp_ca_event((struct sock *)tp, CA_EVENT_ECN_NO_CE);
260
		tp->ecn_flags |= TCP_ECN_SEEN;
261
		break;
262 263 264
	}
}

265 266 267 268 269 270 271
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)
272
{
273
	if ((tp->ecn_flags & TCP_ECN_OK) && (!th->ece || th->cwr))
274 275 276
		tp->ecn_flags &= ~TCP_ECN_OK;
}

277
static void tcp_ecn_rcv_syn(struct tcp_sock *tp, const struct tcphdr *th)
278
{
279
	if ((tp->ecn_flags & TCP_ECN_OK) && (!th->ece || !th->cwr))
280 281 282
		tp->ecn_flags &= ~TCP_ECN_OK;
}

283
static bool tcp_ecn_rcv_ecn_echo(const struct tcp_sock *tp, const struct tcphdr *th)
284
{
285
	if (th->ece && !th->syn && (tp->ecn_flags & TCP_ECN_OK))
E
Eric Dumazet 已提交
286 287
		return true;
	return false;
288 289
}

L
Linus Torvalds 已提交
290 291 292 293 294
/* Buffer size and advertised window tuning.
 *
 * 1. Tuning sk->sk_sndbuf, when connection enters established state.
 */

E
Eric Dumazet 已提交
295
static void tcp_sndbuf_expand(struct sock *sk)
L
Linus Torvalds 已提交
296
{
E
Eric Dumazet 已提交
297
	const struct tcp_sock *tp = tcp_sk(sk);
298
	const struct tcp_congestion_ops *ca_ops = inet_csk(sk)->icsk_ca_ops;
E
Eric Dumazet 已提交
299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318
	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)
	 */
319 320
	sndmem = ca_ops->sndbuf_expand ? ca_ops->sndbuf_expand(sk) : 2;
	sndmem *= nr_segs * per_mss;
L
Linus Torvalds 已提交
321

322
	if (sk->sk_sndbuf < sndmem)
323
		sk->sk_sndbuf = min(sndmem, sock_net(sk)->ipv4.sysctl_tcp_wmem[2]);
L
Linus Torvalds 已提交
324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346
}

/* 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
347
 * window and then starts to feed us spaghetti. But it should work
L
Linus Torvalds 已提交
348 349 350 351
 * in common situations. Otherwise, we have to rely on queue collapsing.
 */

/* Slow part of check#2. */
352
static int __tcp_grow_window(const struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
353
{
354
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
355
	/* Optimize this! */
356
	int truesize = tcp_win_from_space(sk, skb->truesize) >> 1;
357
	int window = tcp_win_from_space(sk, sock_net(sk)->ipv4.sysctl_tcp_rmem[2]) >> 1;
L
Linus Torvalds 已提交
358 359 360

	while (tp->rcv_ssthresh <= window) {
		if (truesize <= skb->len)
361
			return 2 * inet_csk(sk)->icsk_ack.rcv_mss;
L
Linus Torvalds 已提交
362 363 364 365 366 367 368

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

369
static void tcp_grow_window(struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
370
{
371 372
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
373 374 375
	/* Check #1 */
	if (tp->rcv_ssthresh < tp->window_clamp &&
	    (int)tp->rcv_ssthresh < tcp_space(sk) &&
376
	    !tcp_under_memory_pressure(sk)) {
L
Linus Torvalds 已提交
377 378 379 380 381
		int incr;

		/* Check #2. Increase window, if skb with such overhead
		 * will fit to rcvbuf in future.
		 */
382
		if (tcp_win_from_space(sk, skb->truesize) <= skb->len)
383
			incr = 2 * tp->advmss;
L
Linus Torvalds 已提交
384
		else
385
			incr = __tcp_grow_window(sk, skb);
L
Linus Torvalds 已提交
386 387

		if (incr) {
388
			incr = max_t(int, incr, 2 * skb->len);
389 390
			tp->rcv_ssthresh = min(tp->rcv_ssthresh + incr,
					       tp->window_clamp);
391
			inet_csk(sk)->icsk_ack.quick |= 1;
L
Linus Torvalds 已提交
392 393 394 395 396 397 398
		}
	}
}

/* 3. Tuning rcvbuf, when connection enters established state. */
static void tcp_fixup_rcvbuf(struct sock *sk)
{
399 400
	u32 mss = tcp_sk(sk)->advmss;
	int rcvmem;
L
Linus Torvalds 已提交
401

402 403
	rcvmem = 2 * SKB_TRUESIZE(mss + MAX_TCP_HEADER) *
		 tcp_default_init_rwnd(mss);
404

E
Eric Dumazet 已提交
405 406 407
	/* Dynamic Right Sizing (DRS) has 2 to 3 RTT latency
	 * Allow enough cushion so that sender is not limited by our window
	 */
408
	if (sock_net(sk)->ipv4.sysctl_tcp_moderate_rcvbuf)
E
Eric Dumazet 已提交
409 410
		rcvmem <<= 2;

411
	if (sk->sk_rcvbuf < rcvmem)
412
		sk->sk_rcvbuf = min(rcvmem, sock_net(sk)->ipv4.sysctl_tcp_rmem[2]);
L
Linus Torvalds 已提交
413 414
}

S
Stephen Hemminger 已提交
415
/* 4. Try to fixup all. It is made immediately after connection enters
L
Linus Torvalds 已提交
416 417
 *    established state.
 */
418
void tcp_init_buffer_space(struct sock *sk)
L
Linus Torvalds 已提交
419
{
420
	int tcp_app_win = sock_net(sk)->ipv4.sysctl_tcp_app_win;
L
Linus Torvalds 已提交
421 422 423 424 425 426
	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
427
		tcp_sndbuf_expand(sk);
L
Linus Torvalds 已提交
428 429

	tp->rcvq_space.space = tp->rcv_wnd;
430
	tcp_mstamp_refresh(tp);
431
	tp->rcvq_space.time = tp->tcp_mstamp;
E
Eric Dumazet 已提交
432
	tp->rcvq_space.seq = tp->copied_seq;
L
Linus Torvalds 已提交
433 434 435 436 437 438

	maxwin = tcp_full_space(sk);

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

439
		if (tcp_app_win && maxwin > 4 * tp->advmss)
L
Linus Torvalds 已提交
440
			tp->window_clamp = max(maxwin -
441
					       (maxwin >> tcp_app_win),
L
Linus Torvalds 已提交
442 443 444 445
					       4 * tp->advmss);
	}

	/* Force reservation of one segment. */
446
	if (tcp_app_win &&
L
Linus Torvalds 已提交
447 448 449 450 451
	    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);
452
	tp->snd_cwnd_stamp = tcp_jiffies32;
L
Linus Torvalds 已提交
453 454 455
}

/* 5. Recalculate window clamp after socket hit its memory bounds. */
456
static void tcp_clamp_window(struct sock *sk)
L
Linus Torvalds 已提交
457
{
458
	struct tcp_sock *tp = tcp_sk(sk);
459
	struct inet_connection_sock *icsk = inet_csk(sk);
460
	struct net *net = sock_net(sk);
L
Linus Torvalds 已提交
461

462
	icsk->icsk_ack.quick = 0;
L
Linus Torvalds 已提交
463

464
	if (sk->sk_rcvbuf < net->ipv4.sysctl_tcp_rmem[2] &&
465
	    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK) &&
466
	    !tcp_under_memory_pressure(sk) &&
467
	    sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)) {
468
		sk->sk_rcvbuf = min(atomic_read(&sk->sk_rmem_alloc),
469
				    net->ipv4.sysctl_tcp_rmem[2]);
L
Linus Torvalds 已提交
470
	}
471
	if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf)
472
		tp->rcv_ssthresh = min(tp->window_clamp, 2U * tp->advmss);
L
Linus Torvalds 已提交
473 474
}

S
Stephen Hemminger 已提交
475 476 477 478 479 480 481 482 483
/* 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)
{
484
	const struct tcp_sock *tp = tcp_sk(sk);
S
Stephen Hemminger 已提交
485 486
	unsigned int hint = min_t(unsigned int, tp->advmss, tp->mss_cache);

487
	hint = min(hint, tp->rcv_wnd / 2);
488
	hint = min(hint, TCP_MSS_DEFAULT);
S
Stephen Hemminger 已提交
489 490 491 492
	hint = max(hint, TCP_MIN_MSS);

	inet_csk(sk)->icsk_ack.rcv_mss = hint;
}
E
Eric Dumazet 已提交
493
EXPORT_SYMBOL(tcp_initialize_rcv_mss);
S
Stephen Hemminger 已提交
494

L
Linus Torvalds 已提交
495 496 497 498
/* 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.
499
 * <http://public.lanl.gov/radiant/pubs.html#DRS>
L
Linus Torvalds 已提交
500 501
 *
 * More detail on this code can be found at
502
 * <http://staff.psc.edu/jheffner/>,
L
Linus Torvalds 已提交
503 504 505 506 507
 * 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)
{
508
	u32 new_sample = tp->rcv_rtt_est.rtt_us;
L
Linus Torvalds 已提交
509 510 511 512 513 514 515 516 517 518 519 520
	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
521
		 * non-timestamp case, we do not smooth things out
S
Stephen Hemminger 已提交
522
		 * else with timestamps disabled convergence takes too
L
Linus Torvalds 已提交
523 524 525 526 527
		 * long.
		 */
		if (!win_dep) {
			m -= (new_sample >> 3);
			new_sample += m;
528 529 530 531 532
		} else {
			m <<= 3;
			if (m < new_sample)
				new_sample = m;
		}
L
Linus Torvalds 已提交
533
	} else {
S
Stephen Hemminger 已提交
534
		/* No previous measure. */
L
Linus Torvalds 已提交
535 536 537
		new_sample = m << 3;
	}

538
	tp->rcv_rtt_est.rtt_us = new_sample;
L
Linus Torvalds 已提交
539 540 541 542
}

static inline void tcp_rcv_rtt_measure(struct tcp_sock *tp)
{
543 544
	u32 delta_us;

545
	if (tp->rcv_rtt_est.time == 0)
L
Linus Torvalds 已提交
546 547 548
		goto new_measure;
	if (before(tp->rcv_nxt, tp->rcv_rtt_est.seq))
		return;
549
	delta_us = tcp_stamp_us_delta(tp->tcp_mstamp, tp->rcv_rtt_est.time);
550
	tcp_rcv_rtt_update(tp, delta_us, 1);
L
Linus Torvalds 已提交
551 552 553

new_measure:
	tp->rcv_rtt_est.seq = tp->rcv_nxt + tp->rcv_wnd;
554
	tp->rcv_rtt_est.time = tp->tcp_mstamp;
L
Linus Torvalds 已提交
555 556
}

557 558
static inline void tcp_rcv_rtt_measure_ts(struct sock *sk,
					  const struct sk_buff *skb)
L
Linus Torvalds 已提交
559
{
560
	struct tcp_sock *tp = tcp_sk(sk);
561

L
Linus Torvalds 已提交
562 563
	if (tp->rx_opt.rcv_tsecr &&
	    (TCP_SKB_CB(skb)->end_seq -
564 565 566 567 568 569
	     TCP_SKB_CB(skb)->seq >= inet_csk(sk)->icsk_ack.rcv_mss)) {
		u32 delta = tcp_time_stamp(tp) - tp->rx_opt.rcv_tsecr;
		u32 delta_us = delta * (USEC_PER_SEC / TCP_TS_HZ);

		tcp_rcv_rtt_update(tp, delta_us, 0);
	}
L
Linus Torvalds 已提交
570 571 572 573 574 575 576 577 578 579
}

/*
 * 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
580
	int copied;
581

582
	time = tcp_stamp_us_delta(tp->tcp_mstamp, tp->rcvq_space.time);
583
	if (time < (tp->rcv_rtt_est.rtt_us >> 3) || tp->rcv_rtt_est.rtt_us == 0)
L
Linus Torvalds 已提交
584
		return;
585

E
Eric Dumazet 已提交
586 587 588 589 590 591 592 593 594 595 596 597 598 599
	/* 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 .... >
	 */

600
	if (sock_net(sk)->ipv4.sysctl_tcp_moderate_rcvbuf &&
E
Eric Dumazet 已提交
601 602
	    !(sk->sk_userlocks & SOCK_RCVBUF_LOCK)) {
		int rcvwin, rcvmem, rcvbuf;
L
Linus Torvalds 已提交
603

E
Eric Dumazet 已提交
604 605 606 607
		/* 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
608

E
Eric Dumazet 已提交
609 610 611 612 613 614 615 616 617 618 619 620 621
		/* 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
622

E
Eric Dumazet 已提交
623
		rcvmem = SKB_TRUESIZE(tp->advmss + MAX_TCP_HEADER);
624
		while (tcp_win_from_space(sk, rcvmem) < tp->advmss)
E
Eric Dumazet 已提交
625
			rcvmem += 128;
L
Linus Torvalds 已提交
626

627 628
		rcvbuf = min(rcvwin / tp->advmss * rcvmem,
			     sock_net(sk)->ipv4.sysctl_tcp_rmem[2]);
E
Eric Dumazet 已提交
629 630
		if (rcvbuf > sk->sk_rcvbuf) {
			sk->sk_rcvbuf = rcvbuf;
L
Linus Torvalds 已提交
631

E
Eric Dumazet 已提交
632 633
			/* Make the window clamp follow along.  */
			tp->window_clamp = rcvwin;
L
Linus Torvalds 已提交
634 635
		}
	}
E
Eric Dumazet 已提交
636
	tp->rcvq_space.space = copied;
637

L
Linus Torvalds 已提交
638 639
new_measure:
	tp->rcvq_space.seq = tp->copied_seq;
640
	tp->rcvq_space.time = tp->tcp_mstamp;
L
Linus Torvalds 已提交
641 642 643 644 645 646 647 648 649 650 651 652
}

/* 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
 */
653
static void tcp_event_data_recv(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
654
{
655
	struct tcp_sock *tp = tcp_sk(sk);
656
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
657 658
	u32 now;

659
	inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
660

661
	tcp_measure_rcv_mss(sk, skb);
L
Linus Torvalds 已提交
662 663

	tcp_rcv_rtt_measure(tp);
664

665
	now = tcp_jiffies32;
L
Linus Torvalds 已提交
666

667
	if (!icsk->icsk_ack.ato) {
L
Linus Torvalds 已提交
668 669 670
		/* The _first_ data packet received, initialize
		 * delayed ACK engine.
		 */
671 672
		tcp_incr_quickack(sk);
		icsk->icsk_ack.ato = TCP_ATO_MIN;
L
Linus Torvalds 已提交
673
	} else {
674
		int m = now - icsk->icsk_ack.lrcvtime;
L
Linus Torvalds 已提交
675

676
		if (m <= TCP_ATO_MIN / 2) {
L
Linus Torvalds 已提交
677
			/* The fastest case is the first. */
678 679 680 681 682 683
			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
684
			/* Too long gap. Apparently sender failed to
L
Linus Torvalds 已提交
685 686
			 * restart window, so that we send ACKs quickly.
			 */
687
			tcp_incr_quickack(sk);
688
			sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
689 690
		}
	}
691
	icsk->icsk_ack.lrcvtime = now;
L
Linus Torvalds 已提交
692

693
	tcp_ecn_check_ce(tp, skb);
L
Linus Torvalds 已提交
694 695

	if (skb->len >= 128)
696
		tcp_grow_window(sk, skb);
L
Linus Torvalds 已提交
697 698 699 700 701 702 703 704 705 706 707
}

/* 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
 */
708
static void tcp_rtt_estimator(struct sock *sk, long mrtt_us)
L
Linus Torvalds 已提交
709
{
710
	struct tcp_sock *tp = tcp_sk(sk);
711 712
	long m = mrtt_us; /* RTT */
	u32 srtt = tp->srtt_us;
L
Linus Torvalds 已提交
713 714 715 716

	/*	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.
717
	 *	This is designed to be as fast as possible
L
Linus Torvalds 已提交
718 719 720 721 722 723 724
	 *	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
725
	 * too slowly, when it should be increased quickly, decrease too quickly
L
Linus Torvalds 已提交
726 727 728 729
	 * 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)
	 */
730 731 732
	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
733 734
		if (m < 0) {
			m = -m;		/* m is now abs(error) */
735
			m -= (tp->mdev_us >> 2);   /* similar update on mdev */
L
Linus Torvalds 已提交
736 737 738 739 740 741 742 743 744 745 746
			/* 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 {
747
			m -= (tp->mdev_us >> 2);   /* similar update on mdev */
L
Linus Torvalds 已提交
748
		}
749 750 751 752 753
		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
754 755
		}
		if (after(tp->snd_una, tp->rtt_seq)) {
756 757
			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
758
			tp->rtt_seq = tp->snd_nxt;
759
			tp->mdev_max_us = tcp_rto_min_us(sk);
L
Linus Torvalds 已提交
760 761 762
		}
	} else {
		/* no previous measure. */
763
		srtt = m << 3;		/* take the measured time to be rtt */
764 765 766
		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
767 768
		tp->rtt_seq = tp->snd_nxt;
	}
769
	tp->srtt_us = max(1U, srtt);
L
Linus Torvalds 已提交
770 771
}

E
Eric Dumazet 已提交
772 773 774 775 776 777
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) */
778 779 780 781 782 783 784 785 786 787 788
	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)
789
		rate *= sock_net(sk)->ipv4.sysctl_tcp_pacing_ss_ratio;
790
	else
791
		rate *= sock_net(sk)->ipv4.sysctl_tcp_pacing_ca_ratio;
E
Eric Dumazet 已提交
792 793 794

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

795 796
	if (likely(tp->srtt_us))
		do_div(rate, tp->srtt_us);
E
Eric Dumazet 已提交
797

798 799 800 801 802 803
	/* 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
804 805
}

L
Linus Torvalds 已提交
806 807 808
/* Calculate rto without backoff.  This is the second half of Van Jacobson's
 * routine referred to above.
 */
809
static void tcp_set_rto(struct sock *sk)
L
Linus Torvalds 已提交
810
{
811
	const struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
812 813 814 815 816 817 818 819
	/* 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
820
	 *    ACKs in some circumstances.
L
Linus Torvalds 已提交
821
	 */
822
	inet_csk(sk)->icsk_rto = __tcp_set_rto(tp);
L
Linus Torvalds 已提交
823 824 825 826

	/* 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
827
	 *    with correct one. It is exactly, which we pretend to do.
L
Linus Torvalds 已提交
828 829
	 */

830 831 832
	/* NOTE: clamping at TCP_RTO_MIN is not required, current algo
	 * guarantees that rto is higher.
	 */
833
	tcp_bound_rto(sk);
L
Linus Torvalds 已提交
834 835
}

836
__u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst)
L
Linus Torvalds 已提交
837 838 839
{
	__u32 cwnd = (dst ? dst_metric(dst, RTAX_INITCWND) : 0);

840
	if (!cwnd)
841
		cwnd = TCP_INIT_CWND;
L
Linus Torvalds 已提交
842 843 844
	return min_t(__u32, cwnd, tp->snd_cwnd_clamp);
}

845
/* Take a notice that peer is sending D-SACKs */
846 847
static void tcp_dsack_seen(struct tcp_sock *tp)
{
848
	tp->rx_opt.sack_ok |= TCP_DSACK_SEEN;
849
	tp->rack.dsack_seen = 1;
850 851
}

852 853 854 855 856 857
/* It's reordering when higher sequence was delivered (i.e. sacked) before
 * some lower never-retransmitted sequence ("low_seq"). The maximum reordering
 * distance is approximated in full-mss packet distance ("reordering").
 */
static void tcp_check_sack_reordering(struct sock *sk, const u32 low_seq,
				      const int ts)
L
Linus Torvalds 已提交
858
{
859
	struct tcp_sock *tp = tcp_sk(sk);
860 861
	const u32 mss = tp->mss_cache;
	u32 fack, metric;
862

863 864
	fack = tcp_highest_sack_seq(tp);
	if (!before(low_seq, fack))
865 866
		return;

867 868
	metric = fack - low_seq;
	if ((metric > tp->reordering * mss) && mss) {
L
Linus Torvalds 已提交
869
#if FASTRETRANS_DEBUG > 1
870 871 872
		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,
873
			 0,
874 875
			 tp->sacked_out,
			 tp->undo_marker ? tp->undo_retrans : 0);
L
Linus Torvalds 已提交
876
#endif
877 878
		tp->reordering = min_t(u32, (metric + mss - 1) / mss,
				       sock_net(sk)->ipv4.sysctl_tcp_max_reordering);
L
Linus Torvalds 已提交
879
	}
Y
Yuchung Cheng 已提交
880

Y
Yuchung Cheng 已提交
881
	tp->rack.reord = 1;
882
	/* This exciting event is worth to be remembered. 8) */
883 884
	NET_INC_STATS(sock_net(sk),
		      ts ? LINUX_MIB_TCPTSREORDER : LINUX_MIB_TCPSACKREORDER);
L
Linus Torvalds 已提交
885 886
}

887
/* This must be called before lost_out is incremented */
888 889
static void tcp_verify_retransmit_hint(struct tcp_sock *tp, struct sk_buff *skb)
{
890
	if (!tp->retransmit_skb_hint ||
891 892
	    before(TCP_SKB_CB(skb)->seq,
		   TCP_SKB_CB(tp->retransmit_skb_hint)->seq))
893
		tp->retransmit_skb_hint = skb;
894 895
}

896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911
/* 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);
}

912 913 914 915 916 917
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);
918
		tcp_sum_lost(tp, skb);
919 920 921 922
		TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
	}
}

Y
Yuchung Cheng 已提交
923
void tcp_skb_mark_lost_uncond_verify(struct tcp_sock *tp, struct sk_buff *skb)
924 925 926
{
	tcp_verify_retransmit_hint(tp, skb);

927
	tcp_sum_lost(tp, skb);
928 929 930 931 932 933
	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
934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954
/* 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())
955
 * 3. Loss detection event of two flavors:
L
Linus Torvalds 已提交
956 957
 *	A. Scoreboard estimator decided the packet is lost.
 *	   A'. Reno "three dupacks" marks head of queue lost.
958
 *	B. SACK arrives sacking SND.NXT at the moment, when the
L
Linus Torvalds 已提交
959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977
 *	   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
978 979 980 981 982 983 984
 *
 * 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
985 986 987 988 989 990 991 992 993
 * 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
994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
 *
 * 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
1016
 * again, D-SACK block must not to go across snd_una (for the same reason as
I
Ilpo Järvinen 已提交
1017 1018 1019 1020 1021 1022 1023 1024 1025
 * 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
1026
 */
E
Eric Dumazet 已提交
1027 1028
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
1029 1030 1031
{
	/* 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
1032
		return false;
I
Ilpo Järvinen 已提交
1033 1034 1035

	/* Nasty start_seq wrap-around check (see comments above) */
	if (!before(start_seq, tp->snd_nxt))
E
Eric Dumazet 已提交
1036
		return false;
I
Ilpo Järvinen 已提交
1037

1038
	/* In outstanding window? ...This is valid exit for D-SACKs too.
I
Ilpo Järvinen 已提交
1039 1040 1041
	 * start_seq == snd_una is non-sensical (see comments above)
	 */
	if (after(start_seq, tp->snd_una))
E
Eric Dumazet 已提交
1042
		return true;
I
Ilpo Järvinen 已提交
1043 1044

	if (!is_dsack || !tp->undo_marker)
E
Eric Dumazet 已提交
1045
		return false;
I
Ilpo Järvinen 已提交
1046 1047

	/* ...Then it's D-SACK, and must reside below snd_una completely */
Z
Zheng Yan 已提交
1048
	if (after(end_seq, tp->snd_una))
E
Eric Dumazet 已提交
1049
		return false;
I
Ilpo Järvinen 已提交
1050 1051

	if (!before(start_seq, tp->undo_marker))
E
Eric Dumazet 已提交
1052
		return true;
I
Ilpo Järvinen 已提交
1053 1054 1055

	/* Too old */
	if (!after(end_seq, tp->undo_marker))
E
Eric Dumazet 已提交
1056
		return false;
I
Ilpo Järvinen 已提交
1057 1058 1059 1060 1061 1062 1063

	/* 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
1064 1065 1066
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)
1067
{
1068
	struct tcp_sock *tp = tcp_sk(sk);
1069 1070
	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
1071
	bool dup_sack = false;
1072 1073

	if (before(start_seq_0, TCP_SKB_CB(ack_skb)->ack_seq)) {
E
Eric Dumazet 已提交
1074
		dup_sack = true;
1075
		tcp_dsack_seen(tp);
1076
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPDSACKRECV);
1077
	} else if (num_sacks > 1) {
1078 1079
		u32 end_seq_1 = get_unaligned_be32(&sp[1].end_seq);
		u32 start_seq_1 = get_unaligned_be32(&sp[1].start_seq);
1080 1081 1082

		if (!after(end_seq_0, end_seq_1) &&
		    !before(start_seq_0, start_seq_1)) {
E
Eric Dumazet 已提交
1083
			dup_sack = true;
1084
			tcp_dsack_seen(tp);
1085
			NET_INC_STATS(sock_net(sk),
1086
					LINUX_MIB_TCPDSACKOFORECV);
1087 1088 1089 1090
		}
	}

	/* D-SACK for already forgotten data... Do dumb counting. */
Y
Yuchung Cheng 已提交
1091
	if (dup_sack && tp->undo_marker && tp->undo_retrans > 0 &&
1092 1093 1094 1095 1096 1097 1098
	    !after(end_seq_0, prior_snd_una) &&
	    after(end_seq_0, tp->undo_marker))
		tp->undo_retrans--;

	return dup_sack;
}

1099
struct tcp_sacktag_state {
1100
	u32	reord;
1101 1102 1103 1104
	/* 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.
	 */
1105 1106
	u64	first_sackt;
	u64	last_sackt;
1107
	struct rate_sample *rate;
1108
	int	flag;
1109
	unsigned int mss_now;
1110 1111
};

1112 1113 1114 1115 1116
/* 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).
1117 1118
 *
 * FIXME: this could be merged to shift decision code
1119
 */
1120
static int tcp_match_skb_to_sack(struct sock *sk, struct sk_buff *skb,
E
Eric Dumazet 已提交
1121
				  u32 start_seq, u32 end_seq)
1122
{
E
Eric Dumazet 已提交
1123 1124
	int err;
	bool in_sack;
1125
	unsigned int pkt_len;
1126
	unsigned int mss;
1127 1128 1129 1130 1131 1132

	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)) {
1133
		mss = tcp_skb_mss(skb);
1134 1135
		in_sack = !after(start_seq, TCP_SKB_CB(skb)->seq);

1136
		if (!in_sack) {
1137
			pkt_len = start_seq - TCP_SKB_CB(skb)->seq;
1138 1139 1140
			if (pkt_len < mss)
				pkt_len = mss;
		} else {
1141
			pkt_len = end_seq - TCP_SKB_CB(skb)->seq;
1142 1143 1144 1145 1146 1147 1148 1149 1150
			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;
1151
			if (!in_sack && new_len < pkt_len)
1152 1153 1154
				new_len += mss;
			pkt_len = new_len;
		}
1155 1156 1157 1158

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

1159 1160
		err = tcp_fragment(sk, TCP_FRAG_IN_RTX_QUEUE, skb,
				   pkt_len, mss, GFP_ATOMIC);
1161 1162 1163 1164 1165 1166 1167
		if (err < 0)
			return err;
	}

	return in_sack;
}

1168 1169 1170 1171
/* 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,
1172
			  int dup_sack, int pcount,
1173
			  u64 xmit_time)
1174
{
1175
	struct tcp_sock *tp = tcp_sk(sk);
1176 1177 1178

	/* Account D-SACK for retransmitted packet. */
	if (dup_sack && (sacked & TCPCB_RETRANS)) {
Y
Yuchung Cheng 已提交
1179
		if (tp->undo_marker && tp->undo_retrans > 0 &&
1180
		    after(end_seq, tp->undo_marker))
1181
			tp->undo_retrans--;
1182 1183 1184
		if ((sacked & TCPCB_SACKED_ACKED) &&
		    before(start_seq, state->reord))
				state->reord = start_seq;
1185 1186 1187
	}

	/* Nothing to do; acked frame is about to be dropped (was ACKed). */
1188
	if (!after(end_seq, tp->snd_una))
1189
		return sacked;
1190 1191

	if (!(sacked & TCPCB_SACKED_ACKED)) {
1192
		tcp_rack_advance(tp, sacked, end_seq, xmit_time);
1193

1194 1195 1196 1197 1198 1199
		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) {
1200
				sacked &= ~(TCPCB_LOST|TCPCB_SACKED_RETRANS);
1201 1202
				tp->lost_out -= pcount;
				tp->retrans_out -= pcount;
1203 1204 1205 1206 1207 1208
			}
		} else {
			if (!(sacked & TCPCB_RETRANS)) {
				/* New sack for not retransmitted frame,
				 * which was in hole. It is reordering.
				 */
1209
				if (before(start_seq,
1210 1211 1212 1213
					   tcp_highest_sack_seq(tp)) &&
				    before(start_seq, state->reord))
					state->reord = start_seq;

Y
Yuchung Cheng 已提交
1214 1215
				if (!after(end_seq, tp->high_seq))
					state->flag |= FLAG_ORIG_SACK_ACKED;
1216 1217 1218
				if (state->first_sackt == 0)
					state->first_sackt = xmit_time;
				state->last_sackt = xmit_time;
1219 1220 1221
			}

			if (sacked & TCPCB_LOST) {
1222
				sacked &= ~TCPCB_LOST;
1223
				tp->lost_out -= pcount;
1224 1225 1226
			}
		}

1227 1228
		sacked |= TCPCB_SACKED_ACKED;
		state->flag |= FLAG_DATA_SACKED;
1229
		tp->sacked_out += pcount;
Y
Yuchung Cheng 已提交
1230
		tp->delivered += pcount;  /* Out-of-order packets delivered */
1231 1232

		/* Lost marker hint past SACKed? Tweak RFC3517 cnt */
Y
Yuchung Cheng 已提交
1233
		if (tp->lost_skb_hint &&
1234
		    before(start_seq, TCP_SKB_CB(tp->lost_skb_hint)->seq))
1235
			tp->lost_cnt_hint += pcount;
1236 1237 1238 1239 1240 1241
	}

	/* 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.
	 */
1242 1243
	if (dup_sack && (sacked & TCPCB_SACKED_RETRANS)) {
		sacked &= ~TCPCB_SACKED_RETRANS;
1244
		tp->retrans_out -= pcount;
1245 1246
	}

1247
	return sacked;
1248 1249
}

1250 1251 1252
/* Shift newly-SACKed bytes from this skb to the immediately previous
 * already-SACKed sk_buff. Mark the newly-SACKed bytes as such.
 */
1253 1254
static bool tcp_shifted_skb(struct sock *sk, struct sk_buff *prev,
			    struct sk_buff *skb,
E
Eric Dumazet 已提交
1255 1256 1257
			    struct tcp_sacktag_state *state,
			    unsigned int pcount, int shifted, int mss,
			    bool dup_sack)
1258 1259
{
	struct tcp_sock *tp = tcp_sk(sk);
1260 1261
	u32 start_seq = TCP_SKB_CB(skb)->seq;	/* start of newly-SACKed */
	u32 end_seq = start_seq + shifted;	/* end of newly-SACKed */
1262 1263 1264

	BUG_ON(!pcount);

1265 1266 1267 1268 1269 1270 1271
	/* 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
1272
			start_seq, end_seq, dup_sack, pcount,
1273
			skb->skb_mstamp);
1274
	tcp_rate_skb_delivered(sk, skb, state->rate);
1275 1276

	if (skb == tp->lost_skb_hint)
1277 1278
		tp->lost_cnt_hint += pcount;

1279 1280 1281
	TCP_SKB_CB(prev)->end_seq += shifted;
	TCP_SKB_CB(skb)->seq += shifted;

1282 1283 1284
	tcp_skb_pcount_add(prev, pcount);
	BUG_ON(tcp_skb_pcount(skb) < pcount);
	tcp_skb_pcount_add(skb, -pcount);
1285 1286 1287 1288 1289 1290

	/* 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.
	 */
1291 1292
	if (!TCP_SKB_CB(prev)->tcp_gso_size)
		TCP_SKB_CB(prev)->tcp_gso_size = mss;
1293 1294

	/* CHECKME: To clear or not to clear? Mimics normal skb currently */
1295
	if (tcp_skb_pcount(skb) <= 1)
1296
		TCP_SKB_CB(skb)->tcp_gso_size = 0;
1297 1298 1299 1300 1301 1302

	/* 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));
1303
		NET_INC_STATS(sock_net(sk), LINUX_MIB_SACKSHIFTED);
E
Eric Dumazet 已提交
1304
		return false;
1305 1306 1307 1308
	}

	/* Whole SKB was eaten :-) */

1309 1310 1311 1312 1313 1314 1315
	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);
	}

1316
	TCP_SKB_CB(prev)->tcp_flags |= TCP_SKB_CB(skb)->tcp_flags;
1317
	TCP_SKB_CB(prev)->eor = TCP_SKB_CB(skb)->eor;
1318 1319 1320
	if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)
		TCP_SKB_CB(prev)->end_seq++;

1321 1322 1323
	if (skb == tcp_highest_sack(sk))
		tcp_advance_highest_sack(sk, skb);

1324
	tcp_skb_collapse_tstamp(prev, skb);
1325 1326
	if (unlikely(TCP_SKB_CB(prev)->tx.delivered_mstamp))
		TCP_SKB_CB(prev)->tx.delivered_mstamp = 0;
1327

1328
	tcp_rtx_queue_unlink_and_free(skb, sk);
1329

1330
	NET_INC_STATS(sock_net(sk), LINUX_MIB_SACKMERGED);
1331

E
Eric Dumazet 已提交
1332
	return true;
1333 1334 1335 1336 1337
}

/* I wish gso_size would have a bit more sane initialization than
 * something-or-zero which complicates things
 */
1338
static int tcp_skb_seglen(const struct sk_buff *skb)
1339
{
1340
	return tcp_skb_pcount(skb) == 1 ? skb->len : tcp_skb_mss(skb);
1341 1342 1343
}

/* Shifting pages past head area doesn't work */
1344
static int skb_can_shift(const struct sk_buff *skb)
1345 1346 1347 1348 1349 1350 1351 1352
{
	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,
1353
					  struct tcp_sacktag_state *state,
1354
					  u32 start_seq, u32 end_seq,
E
Eric Dumazet 已提交
1355
					  bool dup_sack)
1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368
{
	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 &&
1369
	    (TCP_SKB_CB(skb)->sacked & (TCPCB_LOST|TCPCB_SACKED_RETRANS)) == TCPCB_SACKED_RETRANS)
1370 1371 1372 1373 1374 1375 1376 1377
		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 */
1378 1379
	prev = skb_rb_prev(skb);
	if (!prev)
1380 1381 1382 1383 1384
		goto fallback;

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

1385 1386 1387
	if (!tcp_skb_can_collapse_to(prev))
		goto fallback;

1388 1389 1390 1391 1392 1393
	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);
1394
		mss = tcp_skb_seglen(skb);
1395 1396 1397 1398

		/* TODO: Fix DSACKs to not fragment already SACKed and we can
		 * drop this restriction as unnecessary
		 */
1399
		if (mss != tcp_skb_seglen(prev))
1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439
			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
		 */
1440
		if (mss != tcp_skb_seglen(prev))
1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
			goto fallback;

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

1453 1454 1455 1456
	/* tcp_sacktag_one() won't SACK-tag ranges below snd_una */
	if (!after(TCP_SKB_CB(skb)->seq + len, tp->snd_una))
		goto fallback;

1457 1458
	if (!skb_shift(prev, skb, len))
		goto fallback;
1459
	if (!tcp_shifted_skb(sk, prev, skb, state, pcount, len, mss, dup_sack))
1460 1461 1462 1463 1464
		goto out;

	/* Hole filled allows collapsing with the next as well, this is very
	 * useful when hole on every nth skb pattern happens
	 */
1465 1466
	skb = skb_rb_next(prev);
	if (!skb)
1467 1468
		goto out;

1469 1470
	if (!skb_can_shift(skb) ||
	    ((TCP_SKB_CB(skb)->sacked & TCPCB_TAGBITS) != TCPCB_SACKED_ACKED) ||
1471
	    (mss != tcp_skb_seglen(skb)))
1472 1473 1474 1475 1476
		goto out;

	len = skb->len;
	if (skb_shift(prev, skb, len)) {
		pcount += tcp_skb_pcount(skb);
1477 1478
		tcp_shifted_skb(sk, prev, skb, state, tcp_skb_pcount(skb),
				len, mss, 0);
1479 1480 1481 1482 1483 1484 1485 1486 1487
	}

out:
	return prev;

noop:
	return skb;

fallback:
1488
	NET_INC_STATS(sock_net(sk), LINUX_MIB_SACKSHIFTFALLBACK);
1489 1490 1491
	return NULL;
}

1492 1493
static struct sk_buff *tcp_sacktag_walk(struct sk_buff *skb, struct sock *sk,
					struct tcp_sack_block *next_dup,
1494
					struct tcp_sacktag_state *state,
1495
					u32 start_seq, u32 end_seq,
E
Eric Dumazet 已提交
1496
					bool dup_sack_in)
1497
{
1498 1499 1500
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *tmp;

1501
	skb_rbtree_walk_from(skb) {
1502
		int in_sack = 0;
E
Eric Dumazet 已提交
1503
		bool dup_sack = dup_sack_in;
1504 1505 1506 1507 1508

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

1509
		if (next_dup  &&
1510 1511 1512 1513 1514
		    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
1515
				dup_sack = true;
1516 1517
		}

1518 1519 1520 1521 1522
		/* 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) {
1523 1524
			tmp = tcp_shift_skb_data(sk, skb, state,
						 start_seq, end_seq, dup_sack);
1525
			if (tmp) {
1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538
				if (tmp != skb) {
					skb = tmp;
					continue;
				}

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

1539 1540 1541
		if (unlikely(in_sack < 0))
			break;

1542
		if (in_sack) {
1543 1544 1545 1546 1547 1548 1549
			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
1550
						tcp_skb_pcount(skb),
1551
						skb->skb_mstamp);
1552
			tcp_rate_skb_delivered(sk, skb, state->rate);
1553 1554
			if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)
				list_del_init(&skb->tcp_tsorted_anchor);
1555

1556 1557 1558 1559
			if (!before(TCP_SKB_CB(skb)->seq,
				    tcp_highest_sack_seq(tp)))
				tcp_advance_highest_sack(sk, skb);
		}
1560 1561 1562 1563
	}
	return skb;
}

1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586
static struct sk_buff *tcp_sacktag_bsearch(struct sock *sk,
					   struct tcp_sacktag_state *state,
					   u32 seq)
{
	struct rb_node *parent, **p = &sk->tcp_rtx_queue.rb_node;
	struct sk_buff *skb;

	while (*p) {
		parent = *p;
		skb = rb_to_skb(parent);
		if (before(seq, TCP_SKB_CB(skb)->seq)) {
			p = &parent->rb_left;
			continue;
		}
		if (!before(seq, TCP_SKB_CB(skb)->end_seq)) {
			p = &parent->rb_right;
			continue;
		}
		return skb;
	}
	return NULL;
}

1587
static struct sk_buff *tcp_sacktag_skip(struct sk_buff *skb, struct sock *sk,
1588 1589
					struct tcp_sacktag_state *state,
					u32 skip_to_seq)
1590
{
1591 1592
	if (skb && after(TCP_SKB_CB(skb)->seq, skip_to_seq))
		return skb;
1593

1594
	return tcp_sacktag_bsearch(sk, state, skip_to_seq);
1595 1596 1597 1598 1599
}

static struct sk_buff *tcp_maybe_skipping_dsack(struct sk_buff *skb,
						struct sock *sk,
						struct tcp_sack_block *next_dup,
1600 1601
						struct tcp_sacktag_state *state,
						u32 skip_to_seq)
1602
{
1603
	if (!next_dup)
1604 1605 1606
		return skb;

	if (before(next_dup->start_seq, skip_to_seq)) {
1607 1608 1609 1610
		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);
1611 1612 1613 1614 1615
	}

	return skb;
}

1616
static int tcp_sack_cache_ok(const struct tcp_sock *tp, const struct tcp_sack_block *cache)
1617 1618 1619 1620
{
	return cache < tp->recv_sack_cache + ARRAY_SIZE(tp->recv_sack_cache);
}

L
Linus Torvalds 已提交
1621
static int
1622
tcp_sacktag_write_queue(struct sock *sk, const struct sk_buff *ack_skb,
1623
			u32 prior_snd_una, struct tcp_sacktag_state *state)
L
Linus Torvalds 已提交
1624 1625
{
	struct tcp_sock *tp = tcp_sk(sk);
1626 1627
	const unsigned char *ptr = (skb_transport_header(ack_skb) +
				    TCP_SKB_CB(ack_skb)->sacked);
1628
	struct tcp_sack_block_wire *sp_wire = (struct tcp_sack_block_wire *)(ptr+2);
1629
	struct tcp_sack_block sp[TCP_NUM_SACKS];
1630 1631
	struct tcp_sack_block *cache;
	struct sk_buff *skb;
1632
	int num_sacks = min(TCP_NUM_SACKS, (ptr[1] - TCPOLEN_SACK_BASE) >> 3);
1633
	int used_sacks;
E
Eric Dumazet 已提交
1634
	bool found_dup_sack = false;
1635
	int i, j;
1636
	int first_sack_index;
L
Linus Torvalds 已提交
1637

1638
	state->flag = 0;
1639
	state->reord = tp->snd_nxt;
1640

1641
	if (!tp->sacked_out)
1642
		tcp_highest_sack_reset(sk);
L
Linus Torvalds 已提交
1643

1644
	found_dup_sack = tcp_check_dsack(sk, ack_skb, sp_wire,
1645
					 num_sacks, prior_snd_una);
1646
	if (found_dup_sack) {
1647
		state->flag |= FLAG_DSACKING_ACK;
1648 1649
		tp->delivered++; /* A spurious retransmission is delivered */
	}
1650 1651 1652 1653 1654 1655 1656 1657

	/* 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;

1658 1659 1660
	if (!tp->packets_out)
		goto out;

1661 1662 1663
	used_sacks = 0;
	first_sack_index = 0;
	for (i = 0; i < num_sacks; i++) {
E
Eric Dumazet 已提交
1664
		bool dup_sack = !i && found_dup_sack;
1665

1666 1667
		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);
1668 1669 1670 1671

		if (!tcp_is_sackblock_valid(tp, dup_sack,
					    sp[used_sacks].start_seq,
					    sp[used_sacks].end_seq)) {
1672 1673
			int mib_idx;

1674 1675
			if (dup_sack) {
				if (!tp->undo_marker)
1676
					mib_idx = LINUX_MIB_TCPDSACKIGNOREDNOUNDO;
1677
				else
1678
					mib_idx = LINUX_MIB_TCPDSACKIGNOREDOLD;
1679 1680 1681 1682 1683
			} 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;
1684
				mib_idx = LINUX_MIB_TCPSACKDISCARD;
1685
			}
1686

1687
			NET_INC_STATS(sock_net(sk), mib_idx);
1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699
			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++;
	}

1700 1701
	/* order SACK blocks to allow in order walk of the retrans queue */
	for (i = used_sacks - 1; i > 0; i--) {
1702 1703
		for (j = 0; j < i; j++) {
			if (after(sp[j].start_seq, sp[j + 1].start_seq)) {
1704
				swap(sp[j], sp[j + 1]);
1705

1706 1707
				/* Track where the first SACK block goes to */
				if (j == first_sack_index)
1708
					first_sack_index = j + 1;
1709 1710 1711 1712
			}
		}
	}

1713 1714
	state->mss_now = tcp_current_mss(sk);
	skb = NULL;
1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725
	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++;
1726 1727
	}

1728
	while (i < used_sacks) {
1729 1730
		u32 start_seq = sp[i].start_seq;
		u32 end_seq = sp[i].end_seq;
E
Eric Dumazet 已提交
1731
		bool dup_sack = (found_dup_sack && (i == first_sack_index));
1732
		struct tcp_sack_block *next_dup = NULL;
1733

1734 1735
		if (found_dup_sack && ((i + 1) == first_sack_index))
			next_dup = &sp[i + 1];
L
Linus Torvalds 已提交
1736

1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747
		/* 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)) {
1748
				skb = tcp_sacktag_skip(skb, sk, state,
1749
						       start_seq);
1750
				skb = tcp_sacktag_walk(skb, sk, next_dup,
1751
						       state,
1752 1753
						       start_seq,
						       cache->start_seq,
1754
						       dup_sack);
1755
			}
1756

1757
			/* Rest of the block already fully processed? */
1758
			if (!after(end_seq, cache->end_seq))
1759
				goto advance_sp;
1760

1761
			skb = tcp_maybe_skipping_dsack(skb, sk, next_dup,
1762
						       state,
1763
						       cache->end_seq);
1764

1765
			/* ...tail remains todo... */
1766
			if (tcp_highest_sack_seq(tp) == cache->end_seq) {
1767
				/* ...but better entrypoint exists! */
1768
				skb = tcp_highest_sack(sk);
1769
				if (!skb)
1770
					break;
1771 1772
				cache++;
				goto walk;
1773 1774
			}

1775
			skb = tcp_sacktag_skip(skb, sk, state, cache->end_seq);
1776 1777 1778 1779
			/* Check overlap against next cached too (past this one already) */
			cache++;
			continue;
		}
L
Linus Torvalds 已提交
1780

1781 1782
		if (!before(start_seq, tcp_highest_sack_seq(tp))) {
			skb = tcp_highest_sack(sk);
1783
			if (!skb)
1784
				break;
L
Linus Torvalds 已提交
1785
		}
1786
		skb = tcp_sacktag_skip(skb, sk, state, start_seq);
1787 1788

walk:
1789
		skb = tcp_sacktag_walk(skb, sk, next_dup, state,
1790
				       start_seq, end_seq, dup_sack);
1791

1792 1793
advance_sp:
		i++;
L
Linus Torvalds 已提交
1794 1795
	}

1796 1797 1798 1799 1800 1801 1802 1803
	/* 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];

1804 1805
	if (inet_csk(sk)->icsk_ca_state != TCP_CA_Loss || tp->undo_marker)
		tcp_check_sack_reordering(sk, state->reord, 0);
L
Linus Torvalds 已提交
1806

1807
	tcp_verify_left_out(tp);
1808 1809
out:

L
Linus Torvalds 已提交
1810
#if FASTRETRANS_DEBUG > 0
1811 1812 1813 1814
	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
1815
#endif
1816
	return state->flag;
L
Linus Torvalds 已提交
1817 1818
}

1819
/* Limits sacked_out so that sum with lost_out isn't ever larger than
E
Eric Dumazet 已提交
1820
 * packets_out. Returns false if sacked_out adjustement wasn't necessary.
1821
 */
E
Eric Dumazet 已提交
1822
static bool tcp_limit_reno_sacked(struct tcp_sock *tp)
1823 1824 1825 1826 1827 1828 1829 1830
{
	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
1831
		return true;
1832
	}
E
Eric Dumazet 已提交
1833
	return false;
1834 1835 1836 1837 1838 1839 1840 1841 1842
}

/* 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);
1843 1844 1845 1846 1847 1848 1849

	if (!tcp_limit_reno_sacked(tp))
		return;

	tp->reordering = min_t(u32, tp->packets_out + addend,
			       sock_net(sk)->ipv4.sysctl_tcp_max_reordering);
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRENOREORDER);
1850 1851 1852 1853 1854 1855 1856
}

/* 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
1857 1858
	u32 prior_sacked = tp->sacked_out;

1859 1860
	tp->sacked_out++;
	tcp_check_reno_reordering(sk, 0);
Y
Yuchung Cheng 已提交
1861 1862
	if (tp->sacked_out > prior_sacked)
		tp->delivered++; /* Some out-of-order packet is delivered */
1863
	tcp_verify_left_out(tp);
1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
}

/* 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
1874
		tp->delivered += max_t(int, acked - tp->sacked_out, 1);
1875
		if (acked - 1 >= tp->sacked_out)
1876 1877
			tp->sacked_out = 0;
		else
1878
			tp->sacked_out -= acked - 1;
1879 1880
	}
	tcp_check_reno_reordering(sk, acked);
1881
	tcp_verify_left_out(tp);
1882 1883 1884 1885 1886 1887 1888
}

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

Y
Yuchung Cheng 已提交
1889
void tcp_clear_retrans(struct tcp_sock *tp)
L
Linus Torvalds 已提交
1890 1891 1892 1893
{
	tp->retrans_out = 0;
	tp->lost_out = 0;
	tp->undo_marker = 0;
Y
Yuchung Cheng 已提交
1894
	tp->undo_retrans = -1;
Y
Yuchung Cheng 已提交
1895
	tp->sacked_out = 0;
L
Linus Torvalds 已提交
1896 1897
}

Y
Yuchung Cheng 已提交
1898
static inline void tcp_init_undo(struct tcp_sock *tp)
1899
{
Y
Yuchung Cheng 已提交
1900 1901 1902
	tp->undo_marker = tp->snd_una;
	/* Retransmission still in flight may cause DSACKs later. */
	tp->undo_retrans = tp->retrans_out ? : -1;
1903 1904
}

1905
/* Enter Loss state. If we detect SACK reneging, forget all SACK information
L
Linus Torvalds 已提交
1906 1907 1908
 * and reset tags completely, otherwise preserve SACKs. If receiver
 * dropped its ofo queue, we will know this due to reneging detection.
 */
1909
void tcp_enter_loss(struct sock *sk)
L
Linus Torvalds 已提交
1910
{
1911
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1912
	struct tcp_sock *tp = tcp_sk(sk);
1913
	struct net *net = sock_net(sk);
L
Linus Torvalds 已提交
1914
	struct sk_buff *skb;
1915
	bool new_recovery = icsk->icsk_ca_state < TCP_CA_Recovery;
1916
	bool is_reneg;			/* is receiver reneging on SACKs? */
1917
	bool mark_lost;
L
Linus Torvalds 已提交
1918 1919

	/* Reduce ssthresh if it has not yet been made inside this window. */
Y
Yuchung Cheng 已提交
1920 1921
	if (icsk->icsk_ca_state <= TCP_CA_Disorder ||
	    !after(tp->high_seq, tp->snd_una) ||
1922 1923
	    (icsk->icsk_ca_state == TCP_CA_Loss && !icsk->icsk_retransmits)) {
		tp->prior_ssthresh = tcp_current_ssthresh(sk);
1924
		tp->prior_cwnd = tp->snd_cwnd;
1925 1926
		tp->snd_ssthresh = icsk->icsk_ca_ops->ssthresh(sk);
		tcp_ca_event(sk, CA_EVENT_LOSS);
Y
Yuchung Cheng 已提交
1927
		tcp_init_undo(tp);
L
Linus Torvalds 已提交
1928 1929 1930
	}
	tp->snd_cwnd	   = 1;
	tp->snd_cwnd_cnt   = 0;
1931
	tp->snd_cwnd_stamp = tcp_jiffies32;
L
Linus Torvalds 已提交
1932

Y
Yuchung Cheng 已提交
1933 1934
	tp->retrans_out = 0;
	tp->lost_out = 0;
1935 1936 1937

	if (tcp_is_reno(tp))
		tcp_reset_reno_sack(tp);
L
Linus Torvalds 已提交
1938

1939
	skb = tcp_rtx_queue_head(sk);
1940 1941
	is_reneg = skb && (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED);
	if (is_reneg) {
1942
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPSACKRENEGING);
1943
		tp->sacked_out = 0;
1944
	}
1945
	tcp_clear_all_retrans_hints(tp);
L
Linus Torvalds 已提交
1946

1947
	skb_rbtree_walk_from(skb) {
1948 1949 1950 1951
		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
1952
		TCP_SKB_CB(skb)->sacked &= (~TCPCB_TAGBITS)|TCPCB_SACKED_ACKED;
1953
		if (mark_lost) {
L
Linus Torvalds 已提交
1954 1955 1956 1957 1958
			TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_ACKED;
			TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
			tp->lost_out += tcp_skb_pcount(skb);
		}
	}
1959
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
1960

1961 1962 1963 1964
	/* 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 &&
1965
	    tp->sacked_out >= net->ipv4.sysctl_tcp_reordering)
1966
		tp->reordering = min_t(unsigned int, tp->reordering,
1967
				       net->ipv4.sysctl_tcp_reordering);
1968
	tcp_set_ca_state(sk, TCP_CA_Loss);
L
Linus Torvalds 已提交
1969
	tp->high_seq = tp->snd_nxt;
1970
	tcp_ecn_queue_cwr(tp);
Y
Yuchung Cheng 已提交
1971

1972 1973 1974 1975 1976 1977 1978 1979
	/* 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
1980
	 */
E
Eric Dumazet 已提交
1981
	tp->frto = net->ipv4.sysctl_tcp_frto &&
1982 1983
		   (new_recovery || icsk->icsk_retransmits) &&
		   !inet_csk(sk)->icsk_mtup.probe_size;
L
Linus Torvalds 已提交
1984 1985
}

1986 1987 1988 1989
/* 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).
 *
1990 1991 1992 1993 1994
 * 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.
1995
 */
E
Eric Dumazet 已提交
1996
static bool tcp_check_sack_reneging(struct sock *sk, int flag)
L
Linus Torvalds 已提交
1997
{
1998
	if (flag & FLAG_SACK_RENEGING) {
1999 2000 2001
		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
2002

2003
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
2004
					  delay, TCP_RTO_MAX);
E
Eric Dumazet 已提交
2005
		return true;
L
Linus Torvalds 已提交
2006
	}
E
Eric Dumazet 已提交
2007
	return false;
L
Linus Torvalds 已提交
2008 2009
}

2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
/* 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).
 *
 * 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.
 */
2021
static inline int tcp_dupack_heuristics(const struct tcp_sock *tp)
2022
{
Y
Yuchung Cheng 已提交
2023
	return tp->sacked_out + 1;
2024 2025
}

Y
Yuchung Cheng 已提交
2026
/* Linux NewReno/SACK/ECN state machine.
L
Linus Torvalds 已提交
2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072
 * --------------------------------------
 *
 * "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.
 *
2073
 *		Essentially, we have now a few algorithms detecting
L
Linus Torvalds 已提交
2074 2075
 *		lost packets.
 *
2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093
 *		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.
 *
 *		If the receiver does not support SACK:
 *
 *		NewReno (RFC6582): in Recovery we assume that one segment
L
Linus Torvalds 已提交
2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122
 *		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
2123
static bool tcp_time_to_recover(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2124
{
2125
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2126 2127 2128

	/* Trick#1: The loss is proven. */
	if (tp->lost_out)
E
Eric Dumazet 已提交
2129
		return true;
L
Linus Torvalds 已提交
2130 2131

	/* Not-A-Trick#2 : Classic rule... */
2132
	if (tcp_dupack_heuristics(tp) > tp->reordering)
E
Eric Dumazet 已提交
2133
		return true;
L
Linus Torvalds 已提交
2134

E
Eric Dumazet 已提交
2135
	return false;
L
Linus Torvalds 已提交
2136 2137
}

2138
/* Detect loss in event "A" above by marking head of queue up as lost.
Y
Yuchung Cheng 已提交
2139
 * For non-SACK(Reno) senders, the first "packets" number of segments
2140 2141 2142
 * 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.
2143
 */
I
Ilpo Järvinen 已提交
2144
static void tcp_mark_head_lost(struct sock *sk, int packets, int mark_head)
L
Linus Torvalds 已提交
2145
{
2146
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2147
	struct sk_buff *skb;
2148
	int cnt, oldcnt, lost;
2149
	unsigned int mss;
2150 2151
	/* 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
2152

2153
	WARN_ON(packets > tp->packets_out);
2154 2155
	skb = tp->lost_skb_hint;
	if (skb) {
I
Ilpo Järvinen 已提交
2156
		/* Head already handled? */
2157
		if (mark_head && after(TCP_SKB_CB(skb)->seq, tp->snd_una))
I
Ilpo Järvinen 已提交
2158
			return;
2159
		cnt = tp->lost_cnt_hint;
2160
	} else {
2161
		skb = tcp_rtx_queue_head(sk);
2162 2163
		cnt = 0;
	}
L
Linus Torvalds 已提交
2164

2165
	skb_rbtree_walk_from(skb) {
2166 2167 2168 2169
		/* TODO: do this better */
		/* this is not the most efficient way to do this... */
		tp->lost_skb_hint = skb;
		tp->lost_cnt_hint = cnt;
2170

2171
		if (after(TCP_SKB_CB(skb)->end_seq, loss_high))
2172 2173 2174
			break;

		oldcnt = cnt;
Y
Yuchung Cheng 已提交
2175
		if (tcp_is_reno(tp) ||
2176 2177 2178
		    (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
			cnt += tcp_skb_pcount(skb);

2179
		if (cnt > packets) {
Y
Yuchung Cheng 已提交
2180
			if (tcp_is_sack(tp) ||
2181
			    (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED) ||
2182
			    (oldcnt >= packets))
2183 2184
				break;

2185
			mss = tcp_skb_mss(skb);
2186 2187 2188
			/* If needed, chop off the prefix to mark as lost. */
			lost = (packets - oldcnt) * mss;
			if (lost < skb->len &&
2189 2190
			    tcp_fragment(sk, TCP_FRAG_IN_RTX_QUEUE, skb,
					 lost, mss, GFP_ATOMIC) < 0)
2191 2192 2193 2194
				break;
			cnt = packets;
		}

2195
		tcp_skb_mark_lost(tp, skb);
I
Ilpo Järvinen 已提交
2196 2197 2198

		if (mark_head)
			break;
L
Linus Torvalds 已提交
2199
	}
2200
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2201 2202 2203 2204
}

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

2205
static void tcp_update_scoreboard(struct sock *sk, int fast_rexmit)
L
Linus Torvalds 已提交
2206
{
2207 2208
	struct tcp_sock *tp = tcp_sk(sk);

2209
	if (tcp_is_reno(tp)) {
I
Ilpo Järvinen 已提交
2210
		tcp_mark_head_lost(sk, 1, 1);
L
Linus Torvalds 已提交
2211
	} else {
2212
		int sacked_upto = tp->sacked_out - tp->reordering;
I
Ilpo Järvinen 已提交
2213 2214 2215 2216
		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
2217 2218 2219
	}
}

2220 2221 2222 2223 2224 2225
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);
}

2226 2227 2228 2229 2230 2231 2232 2233 2234 2235
/* 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
2236 2237 2238
/* Nothing was retransmitted or returned timestamp is less
 * than timestamp of the first retransmission.
 */
2239
static inline bool tcp_packet_delayed(const struct tcp_sock *tp)
L
Linus Torvalds 已提交
2240 2241
{
	return !tp->retrans_stamp ||
2242
	       tcp_tsopt_ecr_before(tp, tp->retrans_stamp);
L
Linus Torvalds 已提交
2243 2244 2245 2246
}

/* Undo procedures. */

2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268
/* 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;

2269
	skb = tcp_rtx_queue_head(sk);
2270 2271 2272 2273 2274 2275
	if (unlikely(skb && TCP_SKB_CB(skb)->sacked & TCPCB_EVER_RETRANS))
		return true;

	return false;
}

2276
static void DBGUNDO(struct sock *sk, const char *msg)
L
Linus Torvalds 已提交
2277
{
J
Joe Perches 已提交
2278
#if FASTRETRANS_DEBUG > 1
2279
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2280
	struct inet_sock *inet = inet_sk(sk);
2281

2282
	if (sk->sk_family == AF_INET) {
2283 2284 2285 2286 2287 2288
		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);
2289
	}
E
Eric Dumazet 已提交
2290
#if IS_ENABLED(CONFIG_IPV6)
2291
	else if (sk->sk_family == AF_INET6) {
2292 2293
		pr_debug("Undo %s %pI6/%u c%u l%u ss%u/%u p%u\n",
			 msg,
2294
			 &sk->sk_v6_daddr, ntohs(inet->inet_dport),
2295 2296 2297
			 tp->snd_cwnd, tcp_left_out(tp),
			 tp->snd_ssthresh, tp->prior_ssthresh,
			 tp->packets_out);
2298 2299
	}
#endif
L
Linus Torvalds 已提交
2300
#endif
J
Joe Perches 已提交
2301
}
L
Linus Torvalds 已提交
2302

2303
static void tcp_undo_cwnd_reduction(struct sock *sk, bool unmark_loss)
L
Linus Torvalds 已提交
2304
{
2305 2306
	struct tcp_sock *tp = tcp_sk(sk);

Y
Yuchung Cheng 已提交
2307 2308 2309
	if (unmark_loss) {
		struct sk_buff *skb;

2310
		skb_rbtree_walk(skb, &sk->tcp_rtx_queue) {
Y
Yuchung Cheng 已提交
2311 2312 2313 2314 2315 2316
			TCP_SKB_CB(skb)->sacked &= ~TCPCB_LOST;
		}
		tp->lost_out = 0;
		tcp_clear_all_retrans_hints(tp);
	}

L
Linus Torvalds 已提交
2317
	if (tp->prior_ssthresh) {
2318 2319
		const struct inet_connection_sock *icsk = inet_csk(sk);

2320
		tp->snd_cwnd = icsk->icsk_ca_ops->undo_cwnd(sk);
L
Linus Torvalds 已提交
2321

2322
		if (tp->prior_ssthresh > tp->snd_ssthresh) {
L
Linus Torvalds 已提交
2323
			tp->snd_ssthresh = tp->prior_ssthresh;
2324
			tcp_ecn_withdraw_cwr(tp);
L
Linus Torvalds 已提交
2325 2326
		}
	}
2327
	tp->snd_cwnd_stamp = tcp_jiffies32;
2328
	tp->undo_marker = 0;
L
Linus Torvalds 已提交
2329 2330
}

2331
static inline bool tcp_may_undo(const struct tcp_sock *tp)
L
Linus Torvalds 已提交
2332
{
2333
	return tp->undo_marker && (!tp->undo_retrans || tcp_packet_delayed(tp));
L
Linus Torvalds 已提交
2334 2335 2336
}

/* People celebrate: "We love our President!" */
E
Eric Dumazet 已提交
2337
static bool tcp_try_undo_recovery(struct sock *sk)
L
Linus Torvalds 已提交
2338
{
2339 2340
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2341
	if (tcp_may_undo(tp)) {
2342 2343
		int mib_idx;

L
Linus Torvalds 已提交
2344 2345 2346
		/* Happy end! We did not retransmit anything
		 * or our original transmission succeeded.
		 */
2347
		DBGUNDO(sk, inet_csk(sk)->icsk_ca_state == TCP_CA_Loss ? "loss" : "retrans");
2348
		tcp_undo_cwnd_reduction(sk, false);
2349
		if (inet_csk(sk)->icsk_ca_state == TCP_CA_Loss)
2350
			mib_idx = LINUX_MIB_TCPLOSSUNDO;
L
Linus Torvalds 已提交
2351
		else
2352 2353
			mib_idx = LINUX_MIB_TCPFULLUNDO;

2354
		NET_INC_STATS(sock_net(sk), mib_idx);
2355 2356
	} else if (tp->rack.reo_wnd_persist) {
		tp->rack.reo_wnd_persist--;
L
Linus Torvalds 已提交
2357
	}
2358
	if (tp->snd_una == tp->high_seq && tcp_is_reno(tp)) {
L
Linus Torvalds 已提交
2359 2360 2361
		/* 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. */
2362 2363
		if (!tcp_any_retrans_done(sk))
			tp->retrans_stamp = 0;
E
Eric Dumazet 已提交
2364
		return true;
L
Linus Torvalds 已提交
2365
	}
2366
	tcp_set_ca_state(sk, TCP_CA_Open);
E
Eric Dumazet 已提交
2367
	return false;
L
Linus Torvalds 已提交
2368 2369 2370
}

/* Try to undo cwnd reduction, because D-SACKs acked all retransmitted data */
2371
static bool tcp_try_undo_dsack(struct sock *sk)
L
Linus Torvalds 已提交
2372
{
2373 2374
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
2375
	if (tp->undo_marker && !tp->undo_retrans) {
2376 2377
		tp->rack.reo_wnd_persist = min(TCP_RACK_RECOVERY_THRESH,
					       tp->rack.reo_wnd_persist + 1);
2378
		DBGUNDO(sk, "D-SACK");
2379
		tcp_undo_cwnd_reduction(sk, false);
2380
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPDSACKUNDO);
2381
		return true;
L
Linus Torvalds 已提交
2382
	}
2383
	return false;
L
Linus Torvalds 已提交
2384 2385
}

Y
Yuchung Cheng 已提交
2386 2387
/* 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
2388
{
2389 2390
	struct tcp_sock *tp = tcp_sk(sk);

Y
Yuchung Cheng 已提交
2391
	if (frto_undo || tcp_may_undo(tp)) {
2392
		tcp_undo_cwnd_reduction(sk, true);
2393

2394
		DBGUNDO(sk, "partial loss");
2395
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPLOSSUNDO);
Y
Yuchung Cheng 已提交
2396
		if (frto_undo)
2397
			NET_INC_STATS(sock_net(sk),
2398
					LINUX_MIB_TCPSPURIOUSRTOS);
2399
		inet_csk(sk)->icsk_retransmits = 0;
Y
Yuchung Cheng 已提交
2400
		if (frto_undo || tcp_is_sack(tp))
2401
			tcp_set_ca_state(sk, TCP_CA_Open);
E
Eric Dumazet 已提交
2402
		return true;
L
Linus Torvalds 已提交
2403
	}
E
Eric Dumazet 已提交
2404
	return false;
L
Linus Torvalds 已提交
2405 2406
}

2407
/* The cwnd reduction in CWR and Recovery uses the PRR algorithm in RFC 6937.
2408 2409 2410 2411
 * 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.
2412 2413 2414
 *   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.
2415
 */
2416
static void tcp_init_cwnd_reduction(struct sock *sk)
2417 2418 2419 2420
{
	struct tcp_sock *tp = tcp_sk(sk);

	tp->high_seq = tp->snd_nxt;
N
Nandita Dukkipati 已提交
2421
	tp->tlp_high_seq = 0;
2422 2423 2424 2425
	tp->snd_cwnd_cnt = 0;
	tp->prior_cwnd = tp->snd_cwnd;
	tp->prr_delivered = 0;
	tp->prr_out = 0;
2426
	tp->snd_ssthresh = inet_csk(sk)->icsk_ca_ops->ssthresh(sk);
2427
	tcp_ecn_queue_cwr(tp);
2428 2429
}

2430
void tcp_cwnd_reduction(struct sock *sk, int newly_acked_sacked, int flag)
2431 2432 2433 2434 2435
{
	struct tcp_sock *tp = tcp_sk(sk);
	int sndcnt = 0;
	int delta = tp->snd_ssthresh - tcp_packets_in_flight(tp);

2436 2437 2438
	if (newly_acked_sacked <= 0 || WARN_ON_ONCE(!tp->prior_cwnd))
		return;

2439
	tp->prr_delivered += newly_acked_sacked;
2440
	if (delta < 0) {
2441 2442 2443
		u64 dividend = (u64)tp->snd_ssthresh * tp->prr_delivered +
			       tp->prior_cwnd - 1;
		sndcnt = div_u64(dividend, tp->prior_cwnd) - tp->prr_out;
2444 2445
	} else if ((flag & FLAG_RETRANS_DATA_ACKED) &&
		   !(flag & FLAG_LOST_RETRANS)) {
2446 2447 2448
		sndcnt = min_t(int, delta,
			       max_t(int, tp->prr_delivered - tp->prr_out,
				     newly_acked_sacked) + 1);
2449 2450
	} else {
		sndcnt = min(delta, newly_acked_sacked);
2451
	}
2452 2453
	/* Force a fast retransmit upon entering fast recovery */
	sndcnt = max(sndcnt, (tp->prr_out ? 0 : 1));
2454 2455 2456
	tp->snd_cwnd = tcp_packets_in_flight(tp) + sndcnt;
}

2457
static inline void tcp_end_cwnd_reduction(struct sock *sk)
L
Linus Torvalds 已提交
2458
{
2459
	struct tcp_sock *tp = tcp_sk(sk);
2460

2461 2462 2463
	if (inet_csk(sk)->icsk_ca_ops->cong_control)
		return;

2464
	/* Reset cwnd to ssthresh in CWR or Recovery (unless it's undone) */
2465 2466
	if (tp->snd_ssthresh < TCP_INFINITE_SSTHRESH &&
	    (inet_csk(sk)->icsk_ca_state == TCP_CA_CWR || tp->undo_marker)) {
2467
		tp->snd_cwnd = tp->snd_ssthresh;
2468
		tp->snd_cwnd_stamp = tcp_jiffies32;
2469
	}
2470
	tcp_ca_event(sk, CA_EVENT_COMPLETE_CWR);
L
Linus Torvalds 已提交
2471 2472
}

2473
/* Enter CWR state. Disable cwnd undo since congestion is proven with ECN */
2474
void tcp_enter_cwr(struct sock *sk)
Y
Yuchung Cheng 已提交
2475 2476 2477 2478
{
	struct tcp_sock *tp = tcp_sk(sk);

	tp->prior_ssthresh = 0;
2479
	if (inet_csk(sk)->icsk_ca_state < TCP_CA_CWR) {
Y
Yuchung Cheng 已提交
2480
		tp->undo_marker = 0;
2481
		tcp_init_cwnd_reduction(sk);
Y
Yuchung Cheng 已提交
2482 2483 2484
		tcp_set_ca_state(sk, TCP_CA_CWR);
	}
}
2485
EXPORT_SYMBOL(tcp_enter_cwr);
Y
Yuchung Cheng 已提交
2486

2487 2488 2489 2490 2491
static void tcp_try_keep_open(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	int state = TCP_CA_Open;

2492
	if (tcp_left_out(tp) || tcp_any_retrans_done(sk))
2493 2494 2495 2496 2497 2498 2499 2500
		state = TCP_CA_Disorder;

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

2501
static void tcp_try_to_open(struct sock *sk, int flag)
L
Linus Torvalds 已提交
2502
{
2503 2504
	struct tcp_sock *tp = tcp_sk(sk);

2505 2506
	tcp_verify_left_out(tp);

Y
Yuchung Cheng 已提交
2507
	if (!tcp_any_retrans_done(sk))
L
Linus Torvalds 已提交
2508 2509
		tp->retrans_stamp = 0;

2510
	if (flag & FLAG_ECE)
2511
		tcp_enter_cwr(sk);
L
Linus Torvalds 已提交
2512

2513
	if (inet_csk(sk)->icsk_ca_state != TCP_CA_CWR) {
2514
		tcp_try_keep_open(sk);
L
Linus Torvalds 已提交
2515 2516 2517
	}
}

J
John Heffner 已提交
2518 2519 2520 2521 2522 2523
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;
2524
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMTUPFAIL);
J
John Heffner 已提交
2525 2526
}

2527
static void tcp_mtup_probe_success(struct sock *sk)
J
John Heffner 已提交
2528 2529 2530 2531 2532 2533 2534 2535 2536 2537
{
	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;
2538
	tp->snd_cwnd_stamp = tcp_jiffies32;
2539
	tp->snd_ssthresh = tcp_current_ssthresh(sk);
J
John Heffner 已提交
2540 2541 2542 2543

	icsk->icsk_mtup.search_low = icsk->icsk_mtup.probe_size;
	icsk->icsk_mtup.probe_size = 0;
	tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
2544
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMTUPSUCCESS);
J
John Heffner 已提交
2545 2546
}

2547 2548 2549 2550 2551 2552 2553 2554 2555
/* 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
2556
	unsigned int mss = tcp_current_mss(sk);
2557

2558
	skb_rbtree_walk(skb, &sk->tcp_rtx_queue) {
2559
		if (tcp_skb_seglen(skb) > mss &&
2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570
		    !(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);

2571
	if (!tp->lost_out)
2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592
		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
2593
EXPORT_SYMBOL(tcp_simple_retransmit);
2594

2595
void tcp_enter_recovery(struct sock *sk, bool ece_ack)
2596 2597 2598 2599 2600 2601 2602 2603 2604
{
	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;

2605
	NET_INC_STATS(sock_net(sk), mib_idx);
2606 2607

	tp->prior_ssthresh = 0;
Y
Yuchung Cheng 已提交
2608
	tcp_init_undo(tp);
2609

2610
	if (!tcp_in_cwnd_reduction(sk)) {
2611 2612
		if (!ece_ack)
			tp->prior_ssthresh = tcp_current_ssthresh(sk);
2613
		tcp_init_cwnd_reduction(sk);
2614 2615 2616 2617
	}
	tcp_set_ca_state(sk, TCP_CA_Recovery);
}

2618 2619 2620
/* Process an ACK in CA_Loss state. Move to CA_Open if lost data are
 * recovered or spurious. Otherwise retransmits more on partial ACKs.
 */
2621 2622
static void tcp_process_loss(struct sock *sk, int flag, bool is_dupack,
			     int *rexmit)
2623 2624
{
	struct tcp_sock *tp = tcp_sk(sk);
Y
Yuchung Cheng 已提交
2625
	bool recovered = !before(tp->snd_una, tp->high_seq);
2626

2627 2628 2629 2630
	if ((flag & FLAG_SND_UNA_ADVANCED) &&
	    tcp_try_undo_loss(sk, false))
		return;

2631 2632 2633 2634 2635 2636 2637 2638 2639 2640
	/* 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;
2641

2642
	if (tp->frto) { /* F-RTO RFC5682 sec 3.1 (sack enhanced version). */
2643 2644 2645
		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
2646 2647
		} else if (flag & FLAG_SND_UNA_ADVANCED && !recovered) {
			tp->high_seq = tp->snd_nxt;
2648 2649 2650 2651
			/* Step 2.b. Try send new data (but deferred until cwnd
			 * is updated in tcp_ack()). Otherwise fall back to
			 * the conventional recovery.
			 */
2652
			if (!tcp_write_queue_empty(sk) &&
2653 2654 2655 2656
			    after(tcp_wnd_end(tp), tp->snd_nxt)) {
				*rexmit = REXMIT_NEW;
				return;
			}
Y
Yuchung Cheng 已提交
2657 2658 2659 2660 2661 2662
			tp->frto = 0;
		}
	}

	if (recovered) {
		/* F-RTO RFC5682 sec 3.1 step 2.a and 1st part of step 3.a */
2663 2664 2665
		tcp_try_undo_recovery(sk);
		return;
	}
Y
Yuchung Cheng 已提交
2666 2667 2668 2669 2670 2671 2672 2673 2674
	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);
	}
2675
	*rexmit = REXMIT_LOST;
2676 2677
}

Y
Yuchung Cheng 已提交
2678
/* Undo during fast recovery after partial ACK. */
2679
static bool tcp_try_undo_partial(struct sock *sk, u32 prior_snd_una)
Y
Yuchung Cheng 已提交
2680 2681 2682
{
	struct tcp_sock *tp = tcp_sk(sk);

2683
	if (tp->undo_marker && tcp_packet_delayed(tp)) {
Y
Yuchung Cheng 已提交
2684
		/* Plain luck! Hole if filled with delayed
2685
		 * packet, rather than with a retransmit. Check reordering.
Y
Yuchung Cheng 已提交
2686
		 */
2687
		tcp_check_sack_reordering(sk, prior_snd_una, 1);
2688 2689 2690 2691 2692 2693

		/* 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.
		 */
2694
		if (tp->retrans_out)
2695 2696
			return true;

Y
Yuchung Cheng 已提交
2697 2698 2699
		if (!tcp_any_retrans_done(sk))
			tp->retrans_stamp = 0;

2700 2701
		DBGUNDO(sk, "partial recovery");
		tcp_undo_cwnd_reduction(sk, true);
2702
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPPARTIALUNDO);
2703 2704
		tcp_try_keep_open(sk);
		return true;
Y
Yuchung Cheng 已提交
2705
	}
2706
	return false;
Y
Yuchung Cheng 已提交
2707 2708
}

2709
static void tcp_rack_identify_loss(struct sock *sk, int *ack_flag)
2710 2711 2712 2713
{
	struct tcp_sock *tp = tcp_sk(sk);

	/* Use RACK to detect loss */
2714
	if (sock_net(sk)->ipv4.sysctl_tcp_recovery & TCP_RACK_LOSS_DETECTION) {
2715 2716
		u32 prior_retrans = tp->retrans_out;

2717
		tcp_rack_mark_lost(sk);
2718 2719 2720 2721 2722
		if (prior_retrans > tp->retrans_out)
			*ack_flag |= FLAG_LOST_RETRANS;
	}
}

2723 2724 2725 2726 2727 2728 2729 2730
static bool tcp_force_fast_retransmit(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);

	return after(tcp_highest_sack_seq(tp),
		     tp->snd_una + tp->reordering * tp->mss_cache);
}

L
Linus Torvalds 已提交
2731 2732 2733 2734 2735
/* 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.
 *
2736 2737 2738
 * 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
2739 2740 2741 2742
 *
 * It does _not_ decide what to send, it is made in function
 * tcp_xmit_retransmit_queue().
 */
2743
static void tcp_fastretrans_alert(struct sock *sk, const u32 prior_snd_una,
2744
				  bool is_dupack, int *ack_flag, int *rexmit)
L
Linus Torvalds 已提交
2745
{
2746
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2747
	struct tcp_sock *tp = tcp_sk(sk);
2748
	int fast_rexmit = 0, flag = *ack_flag;
Y
Yuchung Cheng 已提交
2749
	bool do_lost = is_dupack || ((flag & FLAG_DATA_SACKED) &&
2750
				     tcp_force_fast_retransmit(sk));
L
Linus Torvalds 已提交
2751

2752
	if (!tp->packets_out && tp->sacked_out)
L
Linus Torvalds 已提交
2753 2754
		tp->sacked_out = 0;

2755
	/* Now state machine starts.
L
Linus Torvalds 已提交
2756
	 * A. ECE, hence prohibit cwnd undoing, the reduction is required. */
2757
	if (flag & FLAG_ECE)
L
Linus Torvalds 已提交
2758 2759 2760
		tp->prior_ssthresh = 0;

	/* B. In all the states check for reneging SACKs. */
2761
	if (tcp_check_sack_reneging(sk, flag))
L
Linus Torvalds 已提交
2762 2763
		return;

2764
	/* C. Check consistency of the current state. */
2765
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
2766

2767
	/* D. Check state exit conditions. State can be terminated
L
Linus Torvalds 已提交
2768
	 *    when high_seq is ACKed. */
2769
	if (icsk->icsk_ca_state == TCP_CA_Open) {
2770
		WARN_ON(tp->retrans_out != 0);
L
Linus Torvalds 已提交
2771 2772
		tp->retrans_stamp = 0;
	} else if (!before(tp->snd_una, tp->high_seq)) {
2773
		switch (icsk->icsk_ca_state) {
L
Linus Torvalds 已提交
2774 2775 2776 2777
		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) {
2778
				tcp_end_cwnd_reduction(sk);
2779
				tcp_set_ca_state(sk, TCP_CA_Open);
L
Linus Torvalds 已提交
2780 2781 2782 2783
			}
			break;

		case TCP_CA_Recovery:
2784
			if (tcp_is_reno(tp))
L
Linus Torvalds 已提交
2785
				tcp_reset_reno_sack(tp);
2786
			if (tcp_try_undo_recovery(sk))
L
Linus Torvalds 已提交
2787
				return;
2788
			tcp_end_cwnd_reduction(sk);
L
Linus Torvalds 已提交
2789 2790 2791 2792
			break;
		}
	}

2793
	/* E. Process state. */
2794
	switch (icsk->icsk_ca_state) {
L
Linus Torvalds 已提交
2795
	case TCP_CA_Recovery:
2796
		if (!(flag & FLAG_SND_UNA_ADVANCED)) {
2797
			if (tcp_is_reno(tp) && is_dupack)
2798
				tcp_add_reno_sack(sk);
2799
		} else {
2800
			if (tcp_try_undo_partial(sk, prior_snd_una))
2801 2802 2803
				return;
			/* Partial ACK arrived. Force fast retransmit. */
			do_lost = tcp_is_reno(tp) ||
2804
				  tcp_force_fast_retransmit(sk);
2805
		}
2806 2807 2808 2809
		if (tcp_try_undo_dsack(sk)) {
			tcp_try_keep_open(sk);
			return;
		}
2810
		tcp_rack_identify_loss(sk, ack_flag);
L
Linus Torvalds 已提交
2811 2812
		break;
	case TCP_CA_Loss:
2813
		tcp_process_loss(sk, flag, is_dupack, rexmit);
2814
		tcp_rack_identify_loss(sk, ack_flag);
2815 2816
		if (!(icsk->icsk_ca_state == TCP_CA_Open ||
		      (*ack_flag & FLAG_LOST_RETRANS)))
L
Linus Torvalds 已提交
2817
			return;
2818
		/* Change state if cwnd is undone or retransmits are lost */
2819
		/* fall through */
L
Linus Torvalds 已提交
2820
	default:
2821
		if (tcp_is_reno(tp)) {
2822
			if (flag & FLAG_SND_UNA_ADVANCED)
L
Linus Torvalds 已提交
2823 2824
				tcp_reset_reno_sack(tp);
			if (is_dupack)
2825
				tcp_add_reno_sack(sk);
L
Linus Torvalds 已提交
2826 2827
		}

2828
		if (icsk->icsk_ca_state <= TCP_CA_Disorder)
2829
			tcp_try_undo_dsack(sk);
L
Linus Torvalds 已提交
2830

2831
		tcp_rack_identify_loss(sk, ack_flag);
2832
		if (!tcp_time_to_recover(sk, flag)) {
2833
			tcp_try_to_open(sk, flag);
L
Linus Torvalds 已提交
2834 2835 2836
			return;
		}

J
John Heffner 已提交
2837 2838 2839
		/* MTU probe failure: don't reduce cwnd */
		if (icsk->icsk_ca_state < TCP_CA_CWR &&
		    icsk->icsk_mtup.probe_size &&
2840
		    tp->snd_una == tp->mtu_probe.probe_seq_start) {
J
John Heffner 已提交
2841 2842 2843 2844 2845 2846 2847
			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
2848
		/* Otherwise enter Recovery state */
2849
		tcp_enter_recovery(sk, (flag & FLAG_ECE));
2850
		fast_rexmit = 1;
L
Linus Torvalds 已提交
2851 2852
	}

2853
	if (do_lost)
2854
		tcp_update_scoreboard(sk, fast_rexmit);
2855
	*rexmit = REXMIT_LOST;
L
Linus Torvalds 已提交
2856 2857
}

2858 2859
static void tcp_update_rtt_min(struct sock *sk, u32 rtt_us)
{
2860
	u32 wlen = sock_net(sk)->ipv4.sysctl_tcp_min_rtt_wlen * HZ;
2861 2862
	struct tcp_sock *tp = tcp_sk(sk);

2863
	minmax_running_min(&tp->rtt_min, wlen, tcp_jiffies32,
2864
			   rtt_us ? : jiffies_to_usecs(1));
2865 2866
}

2867 2868 2869
static bool tcp_ack_update_rtt(struct sock *sk, const int flag,
			       long seq_rtt_us, long sack_rtt_us,
			       long ca_rtt_us, struct rate_sample *rs)
2870
{
2871 2872 2873 2874 2875 2876 2877
	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).
	 */
2878 2879
	if (seq_rtt_us < 0)
		seq_rtt_us = sack_rtt_us;
Y
Yuchung Cheng 已提交
2880

L
Linus Torvalds 已提交
2881 2882 2883 2884 2885 2886
	/* 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.
	 */
2887
	if (seq_rtt_us < 0 && tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr &&
2888 2889 2890 2891 2892 2893
	    flag & FLAG_ACKED) {
		u32 delta = tcp_time_stamp(tp) - tp->rx_opt.rcv_tsecr;
		u32 delta_us = delta * (USEC_PER_SEC / TCP_TS_HZ);

		seq_rtt_us = ca_rtt_us = delta_us;
	}
2894
	rs->rtt_us = ca_rtt_us; /* RTT of last (S)ACKed packet (or -1) */
2895
	if (seq_rtt_us < 0)
Y
Yuchung Cheng 已提交
2896
		return false;
L
Linus Torvalds 已提交
2897

2898 2899 2900 2901 2902
	/* 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);
2903
	tcp_rtt_estimator(sk, seq_rtt_us);
2904
	tcp_set_rto(sk);
L
Linus Torvalds 已提交
2905

2906 2907
	/* RFC6298: only reset backoff on valid RTT measurement. */
	inet_csk(sk)->icsk_backoff = 0;
Y
Yuchung Cheng 已提交
2908
	return true;
L
Linus Torvalds 已提交
2909 2910
}

2911
/* Compute time elapsed between (last) SYNACK and the ACK completing 3WHS. */
Y
Yuchung Cheng 已提交
2912
void tcp_synack_rtt_meas(struct sock *sk, struct request_sock *req)
2913
{
2914
	struct rate_sample rs;
Y
Yuchung Cheng 已提交
2915
	long rtt_us = -1L;
2916

2917 2918
	if (req && !req->num_retrans && tcp_rsk(req)->snt_synack)
		rtt_us = tcp_stamp_us_delta(tcp_clock_us(), tcp_rsk(req)->snt_synack);
Y
Yuchung Cheng 已提交
2919

2920
	tcp_ack_update_rtt(sk, FLAG_SYN_ACKED, rtt_us, -1L, rtt_us, &rs);
2921 2922
}

Y
Yuchung Cheng 已提交
2923

2924
static void tcp_cong_avoid(struct sock *sk, u32 ack, u32 acked)
L
Linus Torvalds 已提交
2925
{
2926
	const struct inet_connection_sock *icsk = inet_csk(sk);
2927 2928

	icsk->icsk_ca_ops->cong_avoid(sk, ack, acked);
2929
	tcp_sk(sk)->snd_cwnd_stamp = tcp_jiffies32;
L
Linus Torvalds 已提交
2930 2931 2932 2933 2934
}

/* Restart timer after forward progress on connection.
 * RFC2988 recommends to restart timer to now+rto.
 */
2935
void tcp_rearm_rto(struct sock *sk)
L
Linus Torvalds 已提交
2936
{
N
Nandita Dukkipati 已提交
2937
	const struct inet_connection_sock *icsk = inet_csk(sk);
2938
	struct tcp_sock *tp = tcp_sk(sk);
2939

2940 2941 2942 2943 2944 2945
	/* 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
2946
	if (!tp->packets_out) {
2947
		inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS);
L
Linus Torvalds 已提交
2948
	} else {
2949 2950
		u32 rto = inet_csk(sk)->icsk_rto;
		/* Offset the time elapsed after installing regular RTO */
Y
Yuchung Cheng 已提交
2951
		if (icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
N
Nandita Dukkipati 已提交
2952
		    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
2953
			s64 delta_us = tcp_rto_delta_us(sk);
E
Eric Dumazet 已提交
2954
			/* delta_us may not be positive if the socket is locked
N
Nandita Dukkipati 已提交
2955
			 * when the retrans timer fires and is rescheduled.
2956
			 */
2957
			rto = usecs_to_jiffies(max_t(int, delta_us, 1));
2958 2959 2960
		}
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, rto,
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2961
	}
2962 2963
}

2964 2965 2966 2967 2968 2969 2970
/* Try to schedule a loss probe; if that doesn't work, then schedule an RTO. */
static void tcp_set_xmit_timer(struct sock *sk)
{
	if (!tcp_schedule_loss_probe(sk))
		tcp_rearm_rto(sk);
}

2971
/* If we get here, the whole TSO packet has not been acked. */
2972
static u32 tcp_tso_acked(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
2973 2974
{
	struct tcp_sock *tp = tcp_sk(sk);
2975
	u32 packets_acked;
L
Linus Torvalds 已提交
2976

2977
	BUG_ON(!after(TCP_SKB_CB(skb)->end_seq, tp->snd_una));
L
Linus Torvalds 已提交
2978 2979

	packets_acked = tcp_skb_pcount(skb);
2980
	if (tcp_trim_head(sk, skb, tp->snd_una - TCP_SKB_CB(skb)->seq))
L
Linus Torvalds 已提交
2981 2982 2983 2984 2985
		return 0;
	packets_acked -= tcp_skb_pcount(skb);

	if (packets_acked) {
		BUG_ON(tcp_skb_pcount(skb) == 0);
2986
		BUG_ON(!before(TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq));
L
Linus Torvalds 已提交
2987 2988
	}

2989
	return packets_acked;
L
Linus Torvalds 已提交
2990 2991
}

2992 2993 2994 2995 2996 2997
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 */
2998
	if (likely(!TCP_SKB_CB(skb)->txstamp_ack))
2999 3000 3001
		return;

	shinfo = skb_shinfo(skb);
3002
	if (!before(shinfo->tskey, prior_snd_una) &&
3003 3004 3005 3006 3007
	    before(shinfo->tskey, tcp_sk(sk)->snd_una)) {
		tcp_skb_tsorted_save(skb) {
			__skb_tstamp_tx(skb, NULL, sk, SCM_TSTAMP_ACK);
		} tcp_skb_tsorted_restore(skb);
	}
3008 3009
}

3010 3011 3012 3013
/* 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.
 */
3014 3015
static int tcp_clean_rtx_queue(struct sock *sk, u32 prior_fack,
			       u32 prior_snd_una,
3016
			       struct tcp_sacktag_state *sack)
L
Linus Torvalds 已提交
3017
{
3018
	const struct inet_connection_sock *icsk = inet_csk(sk);
3019
	u64 first_ackt, last_ackt;
3020 3021
	struct tcp_sock *tp = tcp_sk(sk);
	u32 prior_sacked = tp->sacked_out;
3022
	u32 reord = tp->snd_nxt; /* lowest acked un-retx un-sacked seq */
3023
	struct sk_buff *skb, *next;
3024
	bool fully_acked = true;
3025
	long sack_rtt_us = -1L;
3026
	long seq_rtt_us = -1L;
3027
	long ca_rtt_us = -1L;
3028
	u32 pkts_acked = 0;
3029
	u32 last_in_flight = 0;
3030
	bool rtt_update;
3031 3032
	int flag = 0;

3033
	first_ackt = 0;
L
Linus Torvalds 已提交
3034

3035
	for (skb = skb_rb_first(&sk->tcp_rtx_queue); skb; skb = next) {
3036
		struct tcp_skb_cb *scb = TCP_SKB_CB(skb);
3037
		const u32 start_seq = scb->seq;
3038
		u8 sacked = scb->sacked;
3039
		u32 acked_pcount;
L
Linus Torvalds 已提交
3040

3041
		tcp_ack_tstamp(sk, skb, prior_snd_una);
3042

3043
		/* Determine how many packets and what bytes were acked, tso and else */
L
Linus Torvalds 已提交
3044
		if (after(scb->end_seq, tp->snd_una)) {
3045 3046 3047 3048
			if (tcp_skb_pcount(skb) == 1 ||
			    !after(tp->snd_una, scb->seq))
				break;

3049 3050
			acked_pcount = tcp_tso_acked(sk, skb);
			if (!acked_pcount)
3051
				break;
E
Eric Dumazet 已提交
3052
			fully_acked = false;
3053
		} else {
3054
			acked_pcount = tcp_skb_pcount(skb);
L
Linus Torvalds 已提交
3055 3056
		}

3057
		if (unlikely(sacked & TCPCB_RETRANS)) {
3058
			if (sacked & TCPCB_SACKED_RETRANS)
3059
				tp->retrans_out -= acked_pcount;
3060
			flag |= FLAG_RETRANS_DATA_ACKED;
3061
		} else if (!(sacked & TCPCB_SACKED_ACKED)) {
3062
			last_ackt = skb->skb_mstamp;
3063 3064
			WARN_ON_ONCE(last_ackt == 0);
			if (!first_ackt)
3065 3066
				first_ackt = last_ackt;

3067
			last_in_flight = TCP_SKB_CB(skb)->tx.in_flight;
3068 3069
			if (before(start_seq, reord))
				reord = start_seq;
3070 3071
			if (!after(scb->end_seq, tp->high_seq))
				flag |= FLAG_ORIG_SACK_ACKED;
3072
		}
3073

Y
Yuchung Cheng 已提交
3074
		if (sacked & TCPCB_SACKED_ACKED) {
3075
			tp->sacked_out -= acked_pcount;
Y
Yuchung Cheng 已提交
3076 3077 3078
		} else if (tcp_is_sack(tp)) {
			tp->delivered += acked_pcount;
			if (!tcp_skb_spurious_retrans(tp, skb))
3079
				tcp_rack_advance(tp, sacked, scb->end_seq,
3080
						 skb->skb_mstamp);
Y
Yuchung Cheng 已提交
3081
		}
3082
		if (sacked & TCPCB_LOST)
3083
			tp->lost_out -= acked_pcount;
3084

3085 3086
		tp->packets_out -= acked_pcount;
		pkts_acked += acked_pcount;
3087
		tcp_rate_skb_delivered(sk, skb, sack->rate);
3088

3089 3090 3091 3092 3093 3094 3095
		/* 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.
		 */
3096
		if (likely(!(scb->tcp_flags & TCPHDR_SYN))) {
3097 3098 3099 3100 3101 3102
			flag |= FLAG_DATA_ACKED;
		} else {
			flag |= FLAG_SYN_ACKED;
			tp->retrans_stamp = 0;
		}

3103 3104 3105
		if (!fully_acked)
			break;

3106
		next = skb_rb_next(skb);
3107
		if (unlikely(skb == tp->retransmit_skb_hint))
3108
			tp->retransmit_skb_hint = NULL;
3109
		if (unlikely(skb == tp->lost_skb_hint))
3110
			tp->lost_skb_hint = NULL;
3111
		tcp_rtx_queue_unlink_and_free(skb, sk);
L
Linus Torvalds 已提交
3112 3113
	}

3114 3115 3116
	if (!skb)
		tcp_chrono_stop(sk, TCP_CHRONO_BUSY);

I
Ilpo Järvinen 已提交
3117 3118 3119
	if (likely(between(tp->snd_up, prior_snd_una, tp->snd_una)))
		tp->snd_up = tp->snd_una;

3120 3121 3122
	if (skb && (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
		flag |= FLAG_SACK_RENEGING;

3123 3124 3125
	if (likely(first_ackt) && !(flag & FLAG_RETRANS_DATA_ACKED)) {
		seq_rtt_us = tcp_stamp_us_delta(tp->tcp_mstamp, first_ackt);
		ca_rtt_us = tcp_stamp_us_delta(tp->tcp_mstamp, last_ackt);
3126
	}
3127 3128 3129
	if (sack->first_sackt) {
		sack_rtt_us = tcp_stamp_us_delta(tp->tcp_mstamp, sack->first_sackt);
		ca_rtt_us = tcp_stamp_us_delta(tp->tcp_mstamp, sack->last_sackt);
3130
	}
3131
	rtt_update = tcp_ack_update_rtt(sk, flag, seq_rtt_us, sack_rtt_us,
3132
					ca_rtt_us, sack->rate);
Y
Yuchung Cheng 已提交
3133

3134
	if (flag & FLAG_ACKED) {
3135
		flag |= FLAG_SET_XMIT_TIMER;  /* set TLP or RTO timer */
3136 3137 3138 3139 3140
		if (unlikely(icsk->icsk_mtup.probe_size &&
			     !after(tp->mtu_probe.probe_seq_end, tp->snd_una))) {
			tcp_mtup_probe_success(sk);
		}

3141 3142 3143
		if (tcp_is_reno(tp)) {
			tcp_remove_reno_sacks(sk, pkts_acked);
		} else {
3144 3145
			int delta;

3146
			/* Non-retransmitted hole got filled? That's reordering */
3147 3148
			if (before(reord, prior_fack))
				tcp_check_sack_reordering(sk, reord, 0);
3149

Y
Yuchung Cheng 已提交
3150
			delta = prior_sacked - tp->sacked_out;
3151
			tp->lost_cnt_hint -= min(tp->lost_cnt_hint, delta);
3152
		}
3153
	} else if (skb && rtt_update && sack_rtt_us >= 0 &&
3154
		   sack_rtt_us > tcp_stamp_us_delta(tp->tcp_mstamp, skb->skb_mstamp)) {
3155 3156 3157 3158
		/* 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.
		 */
3159
		flag |= FLAG_SET_XMIT_TIMER;  /* set TLP or RTO timer */
L
Linus Torvalds 已提交
3160 3161
	}

3162 3163
	if (icsk->icsk_ca_ops->pkts_acked) {
		struct ack_sample sample = { .pkts_acked = pkts_acked,
3164
					     .rtt_us = sack->rate->rtt_us,
3165
					     .in_flight = last_in_flight };
3166 3167 3168

		icsk->icsk_ca_ops->pkts_acked(sk, &sample);
	}
3169

L
Linus Torvalds 已提交
3170
#if FASTRETRANS_DEBUG > 0
3171 3172 3173
	WARN_ON((int)tp->sacked_out < 0);
	WARN_ON((int)tp->lost_out < 0);
	WARN_ON((int)tp->retrans_out < 0);
3174
	if (!tp->packets_out && tcp_is_sack(tp)) {
S
Stephen Hemminger 已提交
3175
		icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3176
		if (tp->lost_out) {
3177 3178
			pr_debug("Leak l=%u %d\n",
				 tp->lost_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3179 3180 3181
			tp->lost_out = 0;
		}
		if (tp->sacked_out) {
3182 3183
			pr_debug("Leak s=%u %d\n",
				 tp->sacked_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3184 3185 3186
			tp->sacked_out = 0;
		}
		if (tp->retrans_out) {
3187 3188
			pr_debug("Leak r=%u %d\n",
				 tp->retrans_out, icsk->icsk_ca_state);
L
Linus Torvalds 已提交
3189 3190 3191 3192
			tp->retrans_out = 0;
		}
	}
#endif
3193
	return flag;
L
Linus Torvalds 已提交
3194 3195 3196 3197
}

static void tcp_ack_probe(struct sock *sk)
{
3198
	struct inet_connection_sock *icsk = inet_csk(sk);
3199 3200
	struct sk_buff *head = tcp_send_head(sk);
	const struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
3201 3202

	/* Was it a usable window open? */
3203 3204 3205
	if (!head)
		return;
	if (!after(TCP_SKB_CB(head)->end_seq, tcp_wnd_end(tp))) {
3206 3207
		icsk->icsk_backoff = 0;
		inet_csk_clear_xmit_timer(sk, ICSK_TIME_PROBE0);
L
Linus Torvalds 已提交
3208 3209 3210 3211
		/* Socket must be waked up by subsequent tcp_data_snd_check().
		 * This function is not for random using!
		 */
	} else {
3212
		unsigned long when = tcp_probe0_when(sk, TCP_RTO_MAX);
3213

3214
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
3215
					  when, TCP_RTO_MAX);
L
Linus Torvalds 已提交
3216 3217 3218
	}
}

3219
static inline bool tcp_ack_is_dubious(const struct sock *sk, const int flag)
L
Linus Torvalds 已提交
3220
{
E
Eric Dumazet 已提交
3221 3222
	return !(flag & FLAG_NOT_DUP) || (flag & FLAG_CA_ALERT) ||
		inet_csk(sk)->icsk_ca_state != TCP_CA_Open;
L
Linus Torvalds 已提交
3223 3224
}

3225
/* Decide wheather to run the increase function of congestion control. */
3226
static inline bool tcp_may_raise_cwnd(const struct sock *sk, const int flag)
L
Linus Torvalds 已提交
3227
{
3228 3229
	/* If reordering is high then always grow cwnd whenever data is
	 * delivered regardless of its ordering. Otherwise stay conservative
3230
	 * and only grow cwnd on in-order delivery (RFC5681). A stretched ACK w/
3231 3232 3233
	 * new SACK or ECE mark may first advance cwnd here and later reduce
	 * cwnd in tcp_fastretrans_alert() based on more states.
	 */
3234
	if (tcp_sk(sk)->reordering > sock_net(sk)->ipv4.sysctl_tcp_reordering)
3235 3236
		return flag & FLAG_FORWARD_PROGRESS;

3237
	return flag & FLAG_DATA_ACKED;
L
Linus Torvalds 已提交
3238 3239
}

Y
Yuchung Cheng 已提交
3240 3241 3242 3243 3244 3245
/* 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,
3246
			     int flag, const struct rate_sample *rs)
Y
Yuchung Cheng 已提交
3247
{
3248 3249 3250 3251 3252 3253 3254
	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
3255 3256 3257 3258 3259 3260 3261 3262 3263 3264
	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
3265 3266 3267
/* Check that window update is acceptable.
 * The function assumes that snd_una<=ack<=snd_next.
 */
3268
static inline bool tcp_may_update_window(const struct tcp_sock *tp,
3269 3270
					const u32 ack, const u32 ack_seq,
					const u32 nwin)
L
Linus Torvalds 已提交
3271
{
E
Eric Dumazet 已提交
3272
	return	after(ack, tp->snd_una) ||
L
Linus Torvalds 已提交
3273
		after(ack_seq, tp->snd_wl1) ||
E
Eric Dumazet 已提交
3274
		(ack_seq == tp->snd_wl1 && nwin > tp->snd_wnd);
L
Linus Torvalds 已提交
3275 3276
}

3277 3278 3279 3280 3281
/* 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;

3282
	sock_owned_by_me((struct sock *)tp);
3283 3284 3285 3286
	tp->bytes_acked += delta;
	tp->snd_una = ack;
}

3287 3288 3289 3290 3291
/* 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;

3292
	sock_owned_by_me((struct sock *)tp);
3293 3294 3295 3296
	tp->bytes_received += delta;
	tp->rcv_nxt = seq;
}

L
Linus Torvalds 已提交
3297 3298 3299 3300 3301
/* 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.
 */
3302
static int tcp_ack_update_window(struct sock *sk, const struct sk_buff *skb, u32 ack,
3303
				 u32 ack_seq)
L
Linus Torvalds 已提交
3304
{
3305
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
3306
	int flag = 0;
3307
	u32 nwin = ntohs(tcp_hdr(skb)->window);
L
Linus Torvalds 已提交
3308

3309
	if (likely(!tcp_hdr(skb)->syn))
L
Linus Torvalds 已提交
3310 3311 3312 3313
		nwin <<= tp->rx_opt.snd_wscale;

	if (tcp_may_update_window(tp, ack, ack_seq, nwin)) {
		flag |= FLAG_WIN_UPDATE;
3314
		tcp_update_wl(tp, ack_seq);
L
Linus Torvalds 已提交
3315 3316 3317 3318

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

3319 3320 3321 3322 3323 3324
			/* Note, it is the only place, where
			 * fast path is recovered for sending TCP.
			 */
			tp->pred_flags = 0;
			tcp_fast_path_check(sk);

3325
			if (!tcp_write_queue_empty(sk))
3326 3327
				tcp_slow_start_after_idle_check(sk);

L
Linus Torvalds 已提交
3328 3329
			if (nwin > tp->max_window) {
				tp->max_window = nwin;
3330
				tcp_sync_mss(sk, inet_csk(sk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
3331 3332 3333 3334
			}
		}
	}

3335
	tcp_snd_una_update(tp, ack);
L
Linus Torvalds 已提交
3336 3337 3338 3339

	return flag;
}

3340 3341 3342 3343
static bool __tcp_oow_rate_limited(struct net *net, int mib_idx,
				   u32 *last_oow_ack_time)
{
	if (*last_oow_ack_time) {
3344
		s32 elapsed = (s32)(tcp_jiffies32 - *last_oow_ack_time);
3345

3346
		if (0 <= elapsed && elapsed < net->ipv4.sysctl_tcp_invalid_ratelimit) {
3347 3348 3349 3350 3351
			NET_INC_STATS(net, mib_idx);
			return true;	/* rate-limited: don't send yet! */
		}
	}

3352
	*last_oow_ack_time = tcp_jiffies32;
3353 3354 3355 3356

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

3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369
/* 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)
3370
		return false;
3371

3372
	return __tcp_oow_rate_limited(net, mib_idx, last_oow_ack_time);
3373 3374
}

3375
/* RFC 5961 7 [ACK Throttling] */
3376
static void tcp_send_challenge_ack(struct sock *sk, const struct sk_buff *skb)
3377 3378 3379 3380
{
	/* unprotected vars, we dont care of overwrites */
	static u32 challenge_timestamp;
	static unsigned int challenge_count;
3381
	struct tcp_sock *tp = tcp_sk(sk);
3382
	struct net *net = sock_net(sk);
3383
	u32 count, now;
3384 3385

	/* First check our per-socket dupack rate limit. */
3386
	if (__tcp_oow_rate_limited(net,
3387 3388
				   LINUX_MIB_TCPACKSKIPPEDCHALLENGE,
				   &tp->last_oow_ack_time))
3389
		return;
3390

3391
	/* Then check host-wide RFC 5961 rate limit. */
3392
	now = jiffies / HZ;
3393
	if (now != challenge_timestamp) {
3394 3395
		u32 ack_limit = net->ipv4.sysctl_tcp_challenge_ack_limit;
		u32 half = (ack_limit + 1) >> 1;
3396

3397
		challenge_timestamp = now;
3398
		WRITE_ONCE(challenge_count, half + prandom_u32_max(ack_limit));
3399
	}
3400 3401 3402
	count = READ_ONCE(challenge_count);
	if (count > 0) {
		WRITE_ONCE(challenge_count, count - 1);
3403
		NET_INC_STATS(net, LINUX_MIB_TCPCHALLENGEACK);
3404 3405 3406 3407
		tcp_send_ack(sk);
	}
}

3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428
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
3429
/* This routine deals with acks during a TLP episode.
3430 3431
 * 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
3432 3433 3434 3435 3436 3437
 * 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);

3438
	if (before(ack, tp->tlp_high_seq))
N
Nandita Dukkipati 已提交
3439 3440
		return;

3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451
	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);
3452
		NET_INC_STATS(sock_net(sk),
3453
				LINUX_MIB_TCPLOSSPROBERECOVERY);
3454 3455 3456
	} 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
3457 3458 3459 3460
		tp->tlp_high_seq = 0;
	}
}

3461 3462 3463 3464 3465 3466 3467 3468
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);
}

3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489
/* 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
3490
/* This routine deals with incoming acks, but not outgoing ones. */
3491
static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
L
Linus Torvalds 已提交
3492
{
3493
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
3494
	struct tcp_sock *tp = tcp_sk(sk);
3495
	struct tcp_sacktag_state sack_state;
3496
	struct rate_sample rs = { .prior_delivered = 0 };
L
Linus Torvalds 已提交
3497 3498 3499
	u32 prior_snd_una = tp->snd_una;
	u32 ack_seq = TCP_SKB_CB(skb)->seq;
	u32 ack = TCP_SKB_CB(skb)->ack_seq;
3500
	bool is_dupack = false;
3501
	int prior_packets = tp->packets_out;
3502 3503
	u32 delivered = tp->delivered;
	u32 lost = tp->lost;
3504
	int rexmit = REXMIT_NONE; /* Flag to (re)transmit to recover losses */
3505
	u32 prior_fack;
3506

3507
	sack_state.first_sackt = 0;
3508
	sack_state.rate = &rs;
L
Linus Torvalds 已提交
3509

3510 3511
	/* We very likely will need to access rtx queue. */
	prefetch(sk->tcp_rtx_queue.rb_node);
3512

3513
	/* If the ack is older than previous acks
L
Linus Torvalds 已提交
3514 3515
	 * then we can probably ignore it.
	 */
3516 3517 3518
	if (before(ack, prior_snd_una)) {
		/* RFC 5961 5.2 [Blind Data Injection Attack].[Mitigation] */
		if (before(ack, prior_snd_una - tp->max_window)) {
3519 3520
			if (!(flag & FLAG_NO_CHALLENGE_ACK))
				tcp_send_challenge_ack(sk, skb);
3521 3522
			return -1;
		}
L
Linus Torvalds 已提交
3523
		goto old_ack;
3524
	}
L
Linus Torvalds 已提交
3525

3526 3527 3528 3529 3530 3531
	/* 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;

3532
	if (after(ack, prior_snd_una)) {
3533
		flag |= FLAG_SND_UNA_ADVANCED;
3534 3535
		icsk->icsk_retransmits = 0;
	}
3536

3537
	prior_fack = tcp_highest_sack_seq(tp);
3538
	rs.prior_in_flight = tcp_packets_in_flight(tp);
3539

3540 3541 3542 3543 3544 3545
	/* 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);

3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558
	if (!(flag & FLAG_SLOWPATH) && after(ack, prior_snd_una)) {
		/* Window is constant, pure forward advance.
		 * No more checks are required.
		 * Note, we use the fact that SND.UNA>=SND.WL2.
		 */
		tcp_update_wl(tp, ack_seq);
		tcp_snd_una_update(tp, ack);
		flag |= FLAG_WIN_UPDATE;

		tcp_in_ack_event(sk, CA_ACK_WIN_UPDATE);

		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPHPACKS);
	} else {
3559
		u32 ack_ev_flags = CA_ACK_SLOWPATH;
L
Linus Torvalds 已提交
3560

3561 3562 3563 3564
		if (ack_seq != TCP_SKB_CB(skb)->end_seq)
			flag |= FLAG_DATA;
		else
			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPPUREACKS);
L
Linus Torvalds 已提交
3565

3566
		flag |= tcp_ack_update_window(sk, skb, ack, ack_seq);
L
Linus Torvalds 已提交
3567

3568 3569 3570
		if (TCP_SKB_CB(skb)->sacked)
			flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una,
							&sack_state);
3571

3572 3573 3574 3575 3576 3577 3578
		if (tcp_ecn_rcv_ecn_echo(tp, tcp_hdr(skb))) {
			flag |= FLAG_ECE;
			ack_ev_flags |= CA_ACK_ECE;
		}

		if (flag & FLAG_WIN_UPDATE)
			ack_ev_flags |= CA_ACK_WIN_UPDATE;
L
Linus Torvalds 已提交
3579

3580 3581
		tcp_in_ack_event(sk, ack_ev_flags);
	}
L
Linus Torvalds 已提交
3582 3583 3584 3585 3586

	/* We passed data and got it acked, remove any soft error
	 * log. Something worked...
	 */
	sk->sk_err_soft = 0;
3587
	icsk->icsk_probes_out = 0;
3588
	tp->rcv_tstamp = tcp_jiffies32;
L
Linus Torvalds 已提交
3589 3590 3591 3592
	if (!prior_packets)
		goto no_queue;

	/* See if we can take anything off of the retransmit queue. */
3593
	flag |= tcp_clean_rtx_queue(sk, prior_fack, prior_snd_una, &sack_state);
3594

3595 3596
	tcp_rack_update_reo_wnd(sk, &rs);

3597 3598 3599 3600 3601 3602
	if (tp->tlp_high_seq)
		tcp_process_tlp_ack(sk, ack, flag);
	/* If needed, reset TLP/RTO timer; RACK may later override this. */
	if (flag & FLAG_SET_XMIT_TIMER)
		tcp_set_xmit_timer(sk);

3603
	if (tcp_ack_is_dubious(sk, flag)) {
3604
		is_dupack = !(flag & (FLAG_SND_UNA_ADVANCED | FLAG_NOT_DUP));
3605 3606
		tcp_fastretrans_alert(sk, prior_snd_una, is_dupack, &flag,
				      &rexmit);
L
Linus Torvalds 已提交
3607
	}
N
Nandita Dukkipati 已提交
3608

3609 3610
	if ((flag & FLAG_FORWARD_PROGRESS) || !(flag & FLAG_NOT_DUP))
		sk_dst_confirm(sk);
N
Nandita Dukkipati 已提交
3611

3612 3613
	delivered = tp->delivered - delivered;	/* freshly ACKed or SACKed */
	lost = tp->lost - lost;			/* freshly marked lost */
3614
	tcp_rate_gen(sk, delivered, lost, sack_state.rate);
3615
	tcp_cong_control(sk, ack, delivered, flag, sack_state.rate);
3616
	tcp_xmit_recovery(sk, rexmit);
L
Linus Torvalds 已提交
3617 3618 3619
	return 1;

no_queue:
3620 3621
	/* If data was DSACKed, see if we can undo a cwnd reduction. */
	if (flag & FLAG_DSACKING_ACK)
3622 3623
		tcp_fastretrans_alert(sk, prior_snd_una, is_dupack, &flag,
				      &rexmit);
L
Linus Torvalds 已提交
3624 3625 3626 3627
	/* 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.
	 */
3628
	tcp_ack_probe(sk);
N
Nandita Dukkipati 已提交
3629 3630 3631

	if (tp->tlp_high_seq)
		tcp_process_tlp_ack(sk, ack, flag);
L
Linus Torvalds 已提交
3632 3633
	return 1;

3634 3635 3636 3637
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
3638
old_ack:
3639 3640 3641
	/* 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.
	 */
3642
	if (TCP_SKB_CB(skb)->sacked) {
Y
Yuchung Cheng 已提交
3643
		flag |= tcp_sacktag_write_queue(sk, skb, prior_snd_una,
3644
						&sack_state);
3645 3646
		tcp_fastretrans_alert(sk, prior_snd_una, is_dupack, &flag,
				      &rexmit);
3647
		tcp_xmit_recovery(sk, rexmit);
3648
	}
L
Linus Torvalds 已提交
3649

3650
	SOCK_DEBUG(sk, "Ack %u before %u:%u\n", ack, tp->snd_una, tp->snd_nxt);
L
Linus Torvalds 已提交
3651 3652 3653
	return 0;
}

3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670
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;
}

3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685
static void smc_parse_options(const struct tcphdr *th,
			      struct tcp_options_received *opt_rx,
			      const unsigned char *ptr,
			      int opsize)
{
#if IS_ENABLED(CONFIG_SMC)
	if (static_branch_unlikely(&tcp_have_smc)) {
		if (th->syn && !(opsize & 1) &&
		    opsize >= TCPOLEN_EXP_SMC_BASE &&
		    get_unaligned_be32(ptr) == TCPOPT_SMC_MAGIC)
			opt_rx->smc_ok = 1;
	}
#endif
}

L
Linus Torvalds 已提交
3686 3687 3688 3689
/* 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.
 */
3690 3691
void tcp_parse_options(const struct net *net,
		       const struct sk_buff *skb,
C
Christoph Paasch 已提交
3692
		       struct tcp_options_received *opt_rx, int estab,
Y
Yuchung Cheng 已提交
3693
		       struct tcp_fastopen_cookie *foc)
L
Linus Torvalds 已提交
3694
{
3695 3696
	const unsigned char *ptr;
	const struct tcphdr *th = tcp_hdr(skb);
3697
	int length = (th->doff * 4) - sizeof(struct tcphdr);
L
Linus Torvalds 已提交
3698

3699
	ptr = (const unsigned char *)(th + 1);
L
Linus Torvalds 已提交
3700 3701
	opt_rx->saw_tstamp = 0;

S
Stephen Hemminger 已提交
3702
	while (length > 0) {
3703
		int opcode = *ptr++;
L
Linus Torvalds 已提交
3704 3705 3706
		int opsize;

		switch (opcode) {
3707 3708 3709 3710 3711 3712 3713 3714
		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
3715
				return;
3716 3717 3718 3719 3720
			if (opsize > length)
				return;	/* don't parse partial options */
			switch (opcode) {
			case TCPOPT_MSS:
				if (opsize == TCPOLEN_MSS && th->syn && !estab) {
3721
					u16 in_mss = get_unaligned_be16(ptr);
3722 3723 3724 3725 3726
					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
3727
					}
3728 3729 3730 3731
				}
				break;
			case TCPOPT_WINDOW:
				if (opsize == TCPOLEN_WINDOW && th->syn &&
3732
				    !estab && net->ipv4.sysctl_tcp_window_scaling) {
3733 3734
					__u8 snd_wscale = *(__u8 *)ptr;
					opt_rx->wscale_ok = 1;
3735 3736
					if (snd_wscale > TCP_MAX_WSCALE) {
						net_info_ratelimited("%s: Illegal window scaling value %d > %u received\n",
3737
								     __func__,
3738 3739 3740
								     snd_wscale,
								     TCP_MAX_WSCALE);
						snd_wscale = TCP_MAX_WSCALE;
L
Linus Torvalds 已提交
3741
					}
3742 3743 3744 3745 3746 3747
					opt_rx->snd_wscale = snd_wscale;
				}
				break;
			case TCPOPT_TIMESTAMP:
				if ((opsize == TCPOLEN_TIMESTAMP) &&
				    ((estab && opt_rx->tstamp_ok) ||
3748
				     (!estab && net->ipv4.sysctl_tcp_timestamps))) {
3749
					opt_rx->saw_tstamp = 1;
3750 3751
					opt_rx->rcv_tsval = get_unaligned_be32(ptr);
					opt_rx->rcv_tsecr = get_unaligned_be32(ptr + 4);
3752 3753 3754 3755
				}
				break;
			case TCPOPT_SACK_PERM:
				if (opsize == TCPOLEN_SACK_PERM && th->syn &&
E
Eric Dumazet 已提交
3756
				    !estab && net->ipv4.sysctl_tcp_sack) {
3757
					opt_rx->sack_ok = TCP_SACK_SEEN;
3758 3759 3760
					tcp_sack_reset(opt_rx);
				}
				break;
L
Linus Torvalds 已提交
3761

3762 3763 3764 3765 3766 3767 3768
			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;
3769
#ifdef CONFIG_TCP_MD5SIG
3770 3771 3772 3773 3774 3775
			case TCPOPT_MD5SIG:
				/*
				 * The MD5 Hash has already been
				 * checked (see tcp_v{4,6}_do_rcv()).
				 */
				break;
3776
#endif
3777 3778 3779 3780 3781 3782
			case TCPOPT_FASTOPEN:
				tcp_parse_fastopen_option(
					opsize - TCPOLEN_FASTOPEN_BASE,
					ptr, th->syn, foc, false);
				break;

Y
Yuchung Cheng 已提交
3783 3784
			case TCPOPT_EXP:
				/* Fast Open option shares code 254 using a
3785
				 * 16 bits magic number.
Y
Yuchung Cheng 已提交
3786
				 */
3787 3788 3789 3790 3791 3792
				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);
3793 3794 3795
				else
					smc_parse_options(th, opt_rx, ptr,
							  opsize);
Y
Yuchung Cheng 已提交
3796 3797 3798
				break;

			}
3799 3800
			ptr += opsize-2;
			length -= opsize;
3801
		}
L
Linus Torvalds 已提交
3802 3803
	}
}
E
Eric Dumazet 已提交
3804
EXPORT_SYMBOL(tcp_parse_options);
L
Linus Torvalds 已提交
3805

E
Eric Dumazet 已提交
3806
static bool tcp_parse_aligned_timestamp(struct tcp_sock *tp, const struct tcphdr *th)
3807
{
3808
	const __be32 *ptr = (const __be32 *)(th + 1);
3809 3810 3811 3812 3813 3814 3815

	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;
3816 3817 3818 3819
		if (*ptr)
			tp->rx_opt.rcv_tsecr = ntohl(*ptr) - tp->tsoffset;
		else
			tp->rx_opt.rcv_tsecr = 0;
E
Eric Dumazet 已提交
3820
		return true;
3821
	}
E
Eric Dumazet 已提交
3822
	return false;
3823 3824
}

L
Linus Torvalds 已提交
3825 3826 3827
/* Fast parse options. This hopes to only see timestamps.
 * If it is wrong it falls back on tcp_parse_options().
 */
3828 3829
static bool tcp_fast_parse_options(const struct net *net,
				   const struct sk_buff *skb,
C
Christoph Paasch 已提交
3830
				   const struct tcphdr *th, struct tcp_sock *tp)
L
Linus Torvalds 已提交
3831
{
3832 3833 3834 3835
	/* 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
3836
		tp->rx_opt.saw_tstamp = 0;
E
Eric Dumazet 已提交
3837
		return false;
L
Linus Torvalds 已提交
3838
	} else if (tp->rx_opt.tstamp_ok &&
3839
		   th->doff == ((sizeof(*th) + TCPOLEN_TSTAMP_ALIGNED) / 4)) {
3840
		if (tcp_parse_aligned_timestamp(tp, th))
E
Eric Dumazet 已提交
3841
			return true;
L
Linus Torvalds 已提交
3842
	}
3843

3844
	tcp_parse_options(net, skb, &tp->rx_opt, 1, NULL);
3845
	if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr)
3846 3847
		tp->rx_opt.rcv_tsecr -= tp->tsoffset;

E
Eric Dumazet 已提交
3848
	return true;
L
Linus Torvalds 已提交
3849 3850
}

3851 3852 3853 3854
#ifdef CONFIG_TCP_MD5SIG
/*
 * Parse MD5 Signature option
 */
3855
const u8 *tcp_parse_md5sig_option(const struct tcphdr *th)
3856
{
3857 3858
	int length = (th->doff << 2) - sizeof(*th);
	const u8 *ptr = (const u8 *)(th + 1);
3859 3860 3861 3862 3863 3864 3865 3866 3867

	/* 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
3868
		switch (opcode) {
3869 3870 3871 3872 3873 3874 3875 3876 3877 3878
		case TCPOPT_EOL:
			return NULL;
		case TCPOPT_NOP:
			length--;
			continue;
		default:
			opsize = *ptr++;
			if (opsize < 2 || opsize > length)
				return NULL;
			if (opcode == TCPOPT_MD5SIG)
3879
				return opsize == TCPOLEN_MD5SIG ? ptr : NULL;
3880 3881 3882 3883 3884 3885
		}
		ptr += opsize - 2;
		length -= opsize;
	}
	return NULL;
}
E
Eric Dumazet 已提交
3886
EXPORT_SYMBOL(tcp_parse_md5sig_option);
3887 3888
#endif

L
Linus Torvalds 已提交
3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911
/* 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) ]
 */

3912
static int tcp_disordered_ack(const struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
3913
{
3914 3915
	const struct tcp_sock *tp = tcp_sk(sk);
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928
	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. */
3929
		(s32)(tp->rx_opt.ts_recent - tp->rx_opt.rcv_tsval) <= (inet_csk(sk)->icsk_rto * 1024) / HZ);
L
Linus Torvalds 已提交
3930 3931
}

3932
static inline bool tcp_paws_discard(const struct sock *sk,
3933
				   const struct sk_buff *skb)
L
Linus Torvalds 已提交
3934
{
3935
	const struct tcp_sock *tp = tcp_sk(sk);
I
Ilpo Järvinen 已提交
3936 3937 3938

	return !tcp_paws_check(&tp->rx_opt, TCP_PAWS_WINDOW) &&
	       !tcp_disordered_ack(sk, skb);
L
Linus Torvalds 已提交
3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953
}

/* 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)
 */

3954
static inline bool tcp_sequence(const struct tcp_sock *tp, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
3955 3956 3957 3958 3959 3960
{
	return	!before(end_seq, tp->rcv_wup) &&
		!after(seq, tp->rcv_nxt + tcp_receive_window(tp));
}

/* When we get a reset we do this. */
3961
void tcp_reset(struct sock *sk)
L
Linus Torvalds 已提交
3962
{
3963 3964
	trace_tcp_receive_reset(sk);

L
Linus Torvalds 已提交
3965 3966
	/* We want the right error as BSD sees it (and indeed as we do). */
	switch (sk->sk_state) {
3967 3968 3969 3970 3971 3972 3973 3974 3975 3976
	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
3977
	}
T
Tom Marshall 已提交
3978 3979
	/* This barrier is coupled with smp_rmb() in tcp_poll() */
	smp_wmb();
L
Linus Torvalds 已提交
3980

3981 3982
	tcp_done(sk);

L
Linus Torvalds 已提交
3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000
	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.
 */
4001
void tcp_fin(struct sock *sk)
L
Linus Torvalds 已提交
4002 4003 4004
{
	struct tcp_sock *tp = tcp_sk(sk);

4005
	inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
4006 4007 4008 4009 4010

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

	switch (sk->sk_state) {
4011 4012 4013 4014
	case TCP_SYN_RECV:
	case TCP_ESTABLISHED:
		/* Move to CLOSE_WAIT */
		tcp_set_state(sk, TCP_CLOSE_WAIT);
4015
		inet_csk(sk)->icsk_ack.pingpong = 1;
4016
		break;
L
Linus Torvalds 已提交
4017

4018 4019 4020 4021 4022 4023 4024 4025 4026
	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
4027

4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044
	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
4045
		pr_err("%s: Impossible, sk->sk_state=%d\n",
4046
		       __func__, sk->sk_state);
4047
		break;
4048
	}
L
Linus Torvalds 已提交
4049 4050 4051 4052

	/* It _is_ possible, that we have something out-of-order _after_ FIN.
	 * Probably, we should reset in this case. For now drop them.
	 */
4053
	skb_rbtree_purge(&tp->out_of_order_queue);
4054
	if (tcp_is_sack(tp))
L
Linus Torvalds 已提交
4055
		tcp_sack_reset(&tp->rx_opt);
4056
	sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
4057 4058 4059 4060 4061 4062 4063

	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)
4064
			sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_HUP);
L
Linus Torvalds 已提交
4065
		else
4066
			sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
L
Linus Torvalds 已提交
4067 4068 4069
	}
}

E
Eric Dumazet 已提交
4070
static inline bool tcp_sack_extend(struct tcp_sack_block *sp, u32 seq,
4071
				  u32 end_seq)
L
Linus Torvalds 已提交
4072 4073 4074 4075 4076 4077
{
	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
4078
		return true;
L
Linus Torvalds 已提交
4079
	}
E
Eric Dumazet 已提交
4080
	return false;
L
Linus Torvalds 已提交
4081 4082
}

4083
static void tcp_dsack_set(struct sock *sk, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
4084
{
4085 4086
	struct tcp_sock *tp = tcp_sk(sk);

E
Eric Dumazet 已提交
4087
	if (tcp_is_sack(tp) && sock_net(sk)->ipv4.sysctl_tcp_dsack) {
4088 4089
		int mib_idx;

L
Linus Torvalds 已提交
4090
		if (before(seq, tp->rcv_nxt))
4091
			mib_idx = LINUX_MIB_TCPDSACKOLDSENT;
L
Linus Torvalds 已提交
4092
		else
4093 4094
			mib_idx = LINUX_MIB_TCPDSACKOFOSENT;

4095
		NET_INC_STATS(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
4096 4097 4098 4099 4100 4101 4102

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

4103
static void tcp_dsack_extend(struct sock *sk, u32 seq, u32 end_seq)
L
Linus Torvalds 已提交
4104
{
4105 4106
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
4107
	if (!tp->rx_opt.dsack)
4108
		tcp_dsack_set(sk, seq, end_seq);
L
Linus Torvalds 已提交
4109 4110 4111 4112
	else
		tcp_sack_extend(tp->duplicate_sack, seq, end_seq);
}

4113
static void tcp_send_dupack(struct sock *sk, const struct sk_buff *skb)
L
Linus Torvalds 已提交
4114 4115 4116 4117 4118
{
	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)) {
4119
		NET_INC_STATS(sock_net(sk), LINUX_MIB_DELAYEDACKLOST);
4120
		tcp_enter_quickack_mode(sk);
L
Linus Torvalds 已提交
4121

E
Eric Dumazet 已提交
4122
		if (tcp_is_sack(tp) && sock_net(sk)->ipv4.sysctl_tcp_dsack) {
L
Linus Torvalds 已提交
4123 4124 4125 4126
			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;
4127
			tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, end_seq);
L
Linus Torvalds 已提交
4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140
		}
	}

	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];
4141
	struct tcp_sack_block *swalk = sp + 1;
L
Linus Torvalds 已提交
4142 4143 4144 4145

	/* See if the recent change to the first SACK eats into
	 * or hits the sequence space of other SACK blocks, if so coalesce.
	 */
4146
	for (this_sack = 1; this_sack < tp->rx_opt.num_sacks;) {
L
Linus Torvalds 已提交
4147 4148 4149 4150 4151 4152 4153
		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--;
4154 4155
			for (i = this_sack; i < tp->rx_opt.num_sacks; i++)
				sp[i] = sp[i + 1];
L
Linus Torvalds 已提交
4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171
			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;

4172
	for (this_sack = 0; this_sack < cur_sacks; this_sack++, sp++) {
L
Linus Torvalds 已提交
4173 4174
		if (tcp_sack_extend(sp, seq, end_seq)) {
			/* Rotate this_sack to the first one. */
4175
			for (; this_sack > 0; this_sack--, sp--)
4176
				swap(*sp, *(sp - 1));
L
Linus Torvalds 已提交
4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188
			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.
	 */
4189
	if (this_sack >= TCP_NUM_SACKS) {
L
Linus Torvalds 已提交
4190 4191 4192 4193
		this_sack--;
		tp->rx_opt.num_sacks--;
		sp--;
	}
S
Stephen Hemminger 已提交
4194
	for (; this_sack > 0; this_sack--, sp--)
4195
		*sp = *(sp - 1);
L
Linus Torvalds 已提交
4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212

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. */
4213
	if (RB_EMPTY_ROOT(&tp->out_of_order_queue)) {
L
Linus Torvalds 已提交
4214 4215 4216 4217
		tp->rx_opt.num_sacks = 0;
		return;
	}

4218
	for (this_sack = 0; this_sack < num_sacks;) {
L
Linus Torvalds 已提交
4219 4220 4221 4222 4223
		/* 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! */
4224
			WARN_ON(before(tp->rcv_nxt, sp->end_seq));
L
Linus Torvalds 已提交
4225 4226

			/* Zap this SACK, by moving forward any other SACKS. */
W
Weilong Chen 已提交
4227
			for (i = this_sack+1; i < num_sacks; i++)
L
Linus Torvalds 已提交
4228 4229 4230 4231 4232 4233 4234
				tp->selective_acks[i-1] = tp->selective_acks[i];
			num_sacks--;
			continue;
		}
		this_sack++;
		sp++;
	}
4235
	tp->rx_opt.num_sacks = num_sacks;
L
Linus Torvalds 已提交
4236 4237
}

4238 4239 4240
/**
 * tcp_try_coalesce - try to merge skb to prior one
 * @sk: socket
4241
 * @dest: destination queue
4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269
 * @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);
4270
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRCVCOALESCE);
4271 4272 4273
	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;
4274 4275 4276

	if (TCP_SKB_CB(from)->has_rxtstamp) {
		TCP_SKB_CB(to)->has_rxtstamp = true;
E
Eric Dumazet 已提交
4277
		to->tstamp = from->tstamp;
4278 4279
	}

4280 4281 4282
	return true;
}

4283 4284 4285 4286 4287 4288
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
4289 4290 4291 4292 4293 4294 4295
/* 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;
4296
	bool fin, fragstolen, eaten;
4297
	struct sk_buff *skb, *tail;
4298
	struct rb_node *p;
L
Linus Torvalds 已提交
4299

4300 4301
	p = rb_first(&tp->out_of_order_queue);
	while (p) {
4302
		skb = rb_to_skb(p);
L
Linus Torvalds 已提交
4303 4304 4305 4306 4307 4308 4309
		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;
4310
			tcp_dsack_extend(sk, TCP_SKB_CB(skb)->seq, dsack);
L
Linus Torvalds 已提交
4311
		}
4312 4313
		p = rb_next(p);
		rb_erase(&skb->rbnode, &tp->out_of_order_queue);
L
Linus Torvalds 已提交
4314

4315
		if (unlikely(!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt))) {
4316
			SOCK_DEBUG(sk, "ofo packet was already received\n");
4317
			tcp_drop(sk, skb);
L
Linus Torvalds 已提交
4318 4319 4320 4321 4322 4323
			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);

4324
		tail = skb_peek_tail(&sk->sk_receive_queue);
E
Eric Dumazet 已提交
4325
		eaten = tail && tcp_try_coalesce(sk, tail, skb, &fragstolen);
4326
		tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq);
4327
		fin = TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN;
4328 4329
		if (!eaten)
			__skb_queue_tail(&sk->sk_receive_queue, skb);
4330
		else
4331
			kfree_skb_partial(skb, fragstolen);
4332 4333 4334 4335 4336 4337 4338 4339

		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
4340 4341 4342
	}
}

E
Eric Dumazet 已提交
4343
static bool tcp_prune_ofo_queue(struct sock *sk);
L
Linus Torvalds 已提交
4344 4345
static int tcp_prune_queue(struct sock *sk);

4346 4347
static int tcp_try_rmem_schedule(struct sock *sk, struct sk_buff *skb,
				 unsigned int size)
4348 4349
{
	if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
4350
	    !sk_rmem_schedule(sk, skb, size)) {
4351 4352 4353 4354

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

4355
		while (!sk_rmem_schedule(sk, skb, size)) {
4356 4357
			if (!tcp_prune_ofo_queue(sk))
				return -1;
4358 4359 4360 4361 4362
		}
	}
	return 0;
}

E
Eric Dumazet 已提交
4363 4364 4365
static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);
4366
	struct rb_node **p, *parent;
E
Eric Dumazet 已提交
4367 4368
	struct sk_buff *skb1;
	u32 seq, end_seq;
4369
	bool fragstolen;
E
Eric Dumazet 已提交
4370

4371
	tcp_ecn_check_ce(tp, skb);
E
Eric Dumazet 已提交
4372

4373
	if (unlikely(tcp_try_rmem_schedule(sk, skb, skb->truesize))) {
4374
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFODROP);
4375
		tcp_drop(sk, skb);
E
Eric Dumazet 已提交
4376 4377 4378
		return;
	}

4379 4380
	/* Disable header prediction. */
	tp->pred_flags = 0;
E
Eric Dumazet 已提交
4381 4382
	inet_csk_schedule_ack(sk);

4383
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFOQUEUE);
4384 4385
	seq = TCP_SKB_CB(skb)->seq;
	end_seq = TCP_SKB_CB(skb)->end_seq;
E
Eric Dumazet 已提交
4386
	SOCK_DEBUG(sk, "out of order segment: rcv_next %X seq %X - %X\n",
4387
		   tp->rcv_nxt, seq, end_seq);
E
Eric Dumazet 已提交
4388

4389 4390
	p = &tp->out_of_order_queue.rb_node;
	if (RB_EMPTY_ROOT(&tp->out_of_order_queue)) {
E
Eric Dumazet 已提交
4391 4392 4393
		/* Initial out of order segment, build 1 SACK. */
		if (tcp_is_sack(tp)) {
			tp->rx_opt.num_sacks = 1;
4394 4395
			tp->selective_acks[0].start_seq = seq;
			tp->selective_acks[0].end_seq = end_seq;
E
Eric Dumazet 已提交
4396
		}
4397 4398 4399
		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
4400 4401 4402
		goto end;
	}

4403 4404 4405
	/* 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.
	 */
E
Eric Dumazet 已提交
4406
	if (tcp_try_coalesce(sk, tp->ooo_last_skb,
4407
			     skb, &fragstolen)) {
4408 4409 4410 4411 4412 4413
coalesce_done:
		tcp_grow_window(sk, skb);
		kfree_skb_partial(skb, fragstolen);
		skb = NULL;
		goto add_sack;
	}
4414 4415 4416 4417 4418 4419
	/* 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;
	}
4420 4421 4422 4423 4424

	/* Find place to insert this segment. Handle overlaps on the way. */
	parent = NULL;
	while (*p) {
		parent = *p;
4425
		skb1 = rb_to_skb(parent);
4426 4427 4428
		if (before(seq, TCP_SKB_CB(skb1)->seq)) {
			p = &parent->rb_left;
			continue;
E
Eric Dumazet 已提交
4429
		}
4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454
		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);
4455
				goto merge_right;
4456
			}
E
Eric Dumazet 已提交
4457
		} else if (tcp_try_coalesce(sk, skb1,
4458
					    skb, &fragstolen)) {
4459
			goto coalesce_done;
E
Eric Dumazet 已提交
4460
		}
4461
		p = &parent->rb_right;
E
Eric Dumazet 已提交
4462
	}
4463
insert:
4464 4465 4466
	/* 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
4467

4468
merge_right:
4469
	/* Remove other segments covered by skb. */
4470
	while ((skb1 = skb_rb_next(skb)) != NULL) {
E
Eric Dumazet 已提交
4471 4472 4473 4474 4475 4476 4477
		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;
		}
4478
		rb_erase(&skb1->rbnode, &tp->out_of_order_queue);
E
Eric Dumazet 已提交
4479 4480
		tcp_dsack_extend(sk, TCP_SKB_CB(skb1)->seq,
				 TCP_SKB_CB(skb1)->end_seq);
4481
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFOMERGE);
4482
		tcp_drop(sk, skb1);
E
Eric Dumazet 已提交
4483
	}
4484
	/* If there is no skb after us, we are the last_skb ! */
4485
	if (!skb1)
4486
		tp->ooo_last_skb = skb;
E
Eric Dumazet 已提交
4487 4488 4489 4490 4491

add_sack:
	if (tcp_is_sack(tp))
		tcp_sack_new_ofo_skb(sk, seq, end_seq);
end:
4492 4493
	if (skb) {
		tcp_grow_window(sk, skb);
4494
		skb_condense(skb);
E
Eric Dumazet 已提交
4495
		skb_set_owner_r(skb, sk);
4496
	}
E
Eric Dumazet 已提交
4497 4498
}

4499
static int __must_check tcp_queue_rcv(struct sock *sk, struct sk_buff *skb, int hdrlen,
4500 4501 4502 4503 4504 4505 4506
		  bool *fragstolen)
{
	int eaten;
	struct sk_buff *tail = skb_peek_tail(&sk->sk_receive_queue);

	__skb_pull(skb, hdrlen);
	eaten = (tail &&
E
Eric Dumazet 已提交
4507
		 tcp_try_coalesce(sk, tail,
4508
				  skb, fragstolen)) ? 1 : 0;
4509
	tcp_rcv_nxt_update(tcp_sk(sk), TCP_SKB_CB(skb)->end_seq);
4510 4511 4512 4513 4514 4515
	if (!eaten) {
		__skb_queue_tail(&sk->sk_receive_queue, skb);
		skb_set_owner_r(skb, sk);
	}
	return eaten;
}
E
Eric Dumazet 已提交
4516

4517 4518
int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size)
{
4519
	struct sk_buff *skb;
4520 4521
	int err = -ENOMEM;
	int data_len = 0;
4522 4523
	bool fragstolen;

4524 4525 4526
	if (size == 0)
		return 0;

4527 4528 4529 4530 4531 4532 4533 4534 4535
	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);
4536 4537 4538
	if (!skb)
		goto err;

4539 4540 4541 4542
	skb_put(skb, size - data_len);
	skb->data_len = data_len;
	skb->len = size;

4543
	if (tcp_try_rmem_schedule(sk, skb, skb->truesize))
4544 4545
		goto err_free;

4546 4547
	err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size);
	if (err)
4548 4549 4550 4551 4552 4553
		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;

4554
	if (tcp_queue_rcv(sk, skb, 0, &fragstolen)) {
4555 4556 4557 4558 4559 4560 4561 4562
		WARN_ON_ONCE(fragstolen); /* should not happen */
		__kfree_skb(skb);
	}
	return size;

err_free:
	kfree_skb(skb);
err:
4563 4564
	return err;

4565 4566
}

L
Linus Torvalds 已提交
4567 4568 4569
static void tcp_data_queue(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
4570 4571
	bool fragstolen;
	int eaten;
L
Linus Torvalds 已提交
4572

4573 4574 4575 4576
	if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq) {
		__kfree_skb(skb);
		return;
	}
E
Eric Dumazet 已提交
4577
	skb_dst_drop(skb);
4578
	__skb_pull(skb, tcp_hdr(skb)->doff * 4);
L
Linus Torvalds 已提交
4579

4580
	tcp_ecn_accept_cwr(tp, skb);
L
Linus Torvalds 已提交
4581

4582
	tp->rx_opt.dsack = 0;
L
Linus Torvalds 已提交
4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593

	/*  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. */
queue_and_out:
E
Eric Dumazet 已提交
4594 4595 4596 4597 4598
		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;

4599
		eaten = tcp_queue_rcv(sk, skb, 0, &fragstolen);
4600
		tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq);
S
Stephen Hemminger 已提交
4601
		if (skb->len)
4602
			tcp_event_data_recv(sk, skb);
4603
		if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)
4604
			tcp_fin(sk);
L
Linus Torvalds 已提交
4605

4606
		if (!RB_EMPTY_ROOT(&tp->out_of_order_queue)) {
L
Linus Torvalds 已提交
4607 4608 4609 4610 4611
			tcp_ofo_queue(sk);

			/* RFC2581. 4.2. SHOULD send immediate ACK, when
			 * gap in queue is filled.
			 */
4612
			if (RB_EMPTY_ROOT(&tp->out_of_order_queue))
4613
				inet_csk(sk)->icsk_ack.pingpong = 0;
L
Linus Torvalds 已提交
4614 4615 4616 4617 4618
		}

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

4619 4620
		tcp_fast_path_check(sk);

4621 4622
		if (eaten > 0)
			kfree_skb_partial(skb, fragstolen);
4623
		if (!sock_flag(sk, SOCK_DEAD))
4624
			sk->sk_data_ready(sk);
L
Linus Torvalds 已提交
4625 4626 4627 4628 4629
		return;
	}

	if (!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) {
		/* A retransmit, 2nd most common case.  Force an immediate ack. */
4630
		NET_INC_STATS(sock_net(sk), LINUX_MIB_DELAYEDACKLOST);
4631
		tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, TCP_SKB_CB(skb)->end_seq);
L
Linus Torvalds 已提交
4632 4633

out_of_window:
4634 4635
		tcp_enter_quickack_mode(sk);
		inet_csk_schedule_ack(sk);
L
Linus Torvalds 已提交
4636
drop:
4637
		tcp_drop(sk, skb);
L
Linus Torvalds 已提交
4638 4639 4640 4641 4642 4643 4644
		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;

4645
	tcp_enter_quickack_mode(sk);
L
Linus Torvalds 已提交
4646 4647 4648 4649 4650 4651 4652

	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);

4653
		tcp_dsack_set(sk, TCP_SKB_CB(skb)->seq, tp->rcv_nxt);
4654

L
Linus Torvalds 已提交
4655 4656 4657 4658 4659 4660 4661 4662
		/* 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
4663
	tcp_data_queue_ofo(sk, skb);
L
Linus Torvalds 已提交
4664 4665
}

4666 4667 4668 4669 4670
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;

4671
	return skb_rb_next(skb);
4672 4673
}

4674
static struct sk_buff *tcp_collapse_one(struct sock *sk, struct sk_buff *skb,
4675 4676
					struct sk_buff_head *list,
					struct rb_root *root)
4677
{
4678
	struct sk_buff *next = tcp_skb_next(skb, list);
4679

4680 4681 4682 4683
	if (list)
		__skb_unlink(skb, list);
	else
		rb_erase(&skb->rbnode, root);
4684 4685

	__kfree_skb(skb);
4686
	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRCVCOLLAPSED);
4687 4688 4689 4690

	return next;
}

4691
/* Insert skb into rb tree, ordered by TCP_SKB_CB(skb)->seq */
4692
void tcp_rbtree_insert(struct rb_root *root, struct sk_buff *skb)
4693 4694 4695 4696 4697 4698 4699
{
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent = NULL;
	struct sk_buff *skb1;

	while (*p) {
		parent = *p;
4700
		skb1 = rb_to_skb(parent);
4701 4702 4703 4704 4705 4706 4707 4708 4709
		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
4710 4711
/* Collapse contiguous sequence of skbs head..tail with
 * sequence numbers start..end.
4712
 *
4713
 * If tail is NULL, this means until the end of the queue.
4714
 *
L
Linus Torvalds 已提交
4715 4716 4717 4718
 * Segments with FIN/SYN are not collapsed (only because this
 * simplifies code)
 */
static void
4719 4720
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
4721
{
4722 4723
	struct sk_buff *skb = head, *n;
	struct sk_buff_head tmp;
4724
	bool end_of_skbs;
L
Linus Torvalds 已提交
4725

S
Stephen Hemminger 已提交
4726
	/* First, check that queue is collapsible and find
4727 4728
	 * the point where collapsing can be useful.
	 */
4729
restart:
4730 4731 4732
	for (end_of_skbs = true; skb != NULL && skb != tail; skb = n) {
		n = tcp_skb_next(skb, list);

L
Linus Torvalds 已提交
4733 4734
		/* No new bits? It is possible on ofo queue. */
		if (!before(start, TCP_SKB_CB(skb)->end_seq)) {
4735
			skb = tcp_collapse_one(sk, skb, list, root);
4736 4737 4738
			if (!skb)
				break;
			goto restart;
L
Linus Torvalds 已提交
4739 4740 4741 4742 4743 4744 4745
		}

		/* The first skb to collapse is:
		 * - not SYN/FIN and
		 * - bloated or contains data before "start" or
		 *   overlaps to the next one.
		 */
4746
		if (!(TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)) &&
4747
		    (tcp_win_from_space(sk, skb->truesize) > skb->len ||
4748 4749
		     before(TCP_SKB_CB(skb)->seq, start))) {
			end_of_skbs = false;
L
Linus Torvalds 已提交
4750
			break;
4751 4752
		}

4753 4754 4755 4756
		if (n && n != tail &&
		    TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(n)->seq) {
			end_of_skbs = false;
			break;
4757
		}
L
Linus Torvalds 已提交
4758 4759 4760 4761

		/* Decided to skip this, advance start seq. */
		start = TCP_SKB_CB(skb)->end_seq;
	}
4762 4763
	if (end_of_skbs ||
	    (TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)))
L
Linus Torvalds 已提交
4764 4765
		return;

4766 4767
	__skb_queue_head_init(&tmp);

L
Linus Torvalds 已提交
4768
	while (before(start, end)) {
4769
		int copy = min_t(int, SKB_MAX_ORDER(0, 0), end - start);
L
Linus Torvalds 已提交
4770 4771
		struct sk_buff *nskb;

4772
		nskb = alloc_skb(copy, GFP_ATOMIC);
L
Linus Torvalds 已提交
4773
		if (!nskb)
4774
			break;
4775

L
Linus Torvalds 已提交
4776 4777
		memcpy(nskb->cb, skb->cb, sizeof(skb->cb));
		TCP_SKB_CB(nskb)->seq = TCP_SKB_CB(nskb)->end_seq = start;
4778 4779 4780 4781
		if (list)
			__skb_queue_before(list, skb, nskb);
		else
			__skb_queue_tail(&tmp, nskb); /* defer rbtree insertion */
4782
		skb_set_owner_r(nskb, sk);
L
Linus Torvalds 已提交
4783 4784 4785 4786 4787 4788

		/* 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;

4789
			BUG_ON(offset < 0);
L
Linus Torvalds 已提交
4790 4791 4792 4793 4794 4795 4796 4797 4798
			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)) {
4799
				skb = tcp_collapse_one(sk, skb, list, root);
4800 4801
				if (!skb ||
				    skb == tail ||
4802
				    (TCP_SKB_CB(skb)->tcp_flags & (TCPHDR_SYN | TCPHDR_FIN)))
4803
					goto end;
L
Linus Torvalds 已提交
4804 4805 4806
			}
		}
	}
4807 4808 4809
end:
	skb_queue_walk_safe(&tmp, skb, n)
		tcp_rbtree_insert(root, skb);
L
Linus Torvalds 已提交
4810 4811 4812 4813 4814 4815 4816 4817
}

/* 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);
4818
	struct sk_buff *skb, *head;
L
Linus Torvalds 已提交
4819 4820
	u32 start, end;

4821
	skb = skb_rb_first(&tp->out_of_order_queue);
4822 4823
new_range:
	if (!skb) {
4824
		tp->ooo_last_skb = skb_rb_last(&tp->out_of_order_queue);
L
Linus Torvalds 已提交
4825
		return;
4826
	}
L
Linus Torvalds 已提交
4827 4828
	start = TCP_SKB_CB(skb)->seq;
	end = TCP_SKB_CB(skb)->end_seq;
4829

4830
	for (head = skb;;) {
4831
		skb = skb_rb_next(skb);
L
Linus Torvalds 已提交
4832

4833 4834 4835
		/* Range is terminated when we see a gap or when
		 * we are at the queue end.
		 */
4836
		if (!skb ||
L
Linus Torvalds 已提交
4837 4838
		    after(TCP_SKB_CB(skb)->seq, end) ||
		    before(TCP_SKB_CB(skb)->end_seq, start)) {
4839
			tcp_collapse(sk, NULL, &tp->out_of_order_queue,
D
David S. Miller 已提交
4840
				     head, skb, start, end);
4841 4842 4843 4844
			goto new_range;
		}

		if (unlikely(before(TCP_SKB_CB(skb)->seq, start)))
L
Linus Torvalds 已提交
4845
			start = TCP_SKB_CB(skb)->seq;
4846
		if (after(TCP_SKB_CB(skb)->end_seq, end))
L
Linus Torvalds 已提交
4847 4848 4849 4850
			end = TCP_SKB_CB(skb)->end_seq;
	}
}

4851
/*
4852 4853 4854 4855 4856 4857 4858 4859
 * 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.
4860
 */
E
Eric Dumazet 已提交
4861
static bool tcp_prune_ofo_queue(struct sock *sk)
4862 4863
{
	struct tcp_sock *tp = tcp_sk(sk);
4864
	struct rb_node *node, *prev;
4865

4866
	if (RB_EMPTY_ROOT(&tp->out_of_order_queue))
4867
		return false;
4868

4869
	NET_INC_STATS(sock_net(sk), LINUX_MIB_OFOPRUNED);
4870 4871 4872 4873
	node = &tp->ooo_last_skb->rbnode;
	do {
		prev = rb_prev(node);
		rb_erase(node, &tp->out_of_order_queue);
4874
		tcp_drop(sk, rb_to_skb(node));
4875
		sk_mem_reclaim(sk);
4876 4877 4878
		if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
		    !tcp_under_memory_pressure(sk))
			break;
4879 4880
		node = prev;
	} while (node);
4881
	tp->ooo_last_skb = rb_to_skb(prev);
4882 4883 4884 4885 4886 4887 4888 4889 4890

	/* 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;
4891 4892
}

L
Linus Torvalds 已提交
4893 4894 4895 4896 4897 4898 4899 4900 4901
/* 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)
{
4902
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
4903 4904 4905

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

4906
	NET_INC_STATS(sock_net(sk), LINUX_MIB_PRUNECALLED);
L
Linus Torvalds 已提交
4907 4908

	if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
4909
		tcp_clamp_window(sk);
4910
	else if (tcp_under_memory_pressure(sk))
L
Linus Torvalds 已提交
4911 4912 4913
		tp->rcv_ssthresh = min(tp->rcv_ssthresh, 4U * tp->advmss);

	tcp_collapse_ofo_queue(sk);
4914
	if (!skb_queue_empty(&sk->sk_receive_queue))
4915
		tcp_collapse(sk, &sk->sk_receive_queue, NULL,
4916 4917 4918
			     skb_peek(&sk->sk_receive_queue),
			     NULL,
			     tp->copied_seq, tp->rcv_nxt);
4919
	sk_mem_reclaim(sk);
L
Linus Torvalds 已提交
4920 4921 4922 4923 4924 4925 4926

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

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

4927
	tcp_prune_ofo_queue(sk);
L
Linus Torvalds 已提交
4928 4929 4930 4931 4932 4933 4934 4935

	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.
	 */
4936
	NET_INC_STATS(sock_net(sk), LINUX_MIB_RCVPRUNED);
L
Linus Torvalds 已提交
4937 4938

	/* Massive buffer overcommit. */
4939
	tp->pred_flags = 0;
L
Linus Torvalds 已提交
4940 4941 4942
	return -1;
}

E
Eric Dumazet 已提交
4943
static bool tcp_should_expand_sndbuf(const struct sock *sk)
4944
{
4945
	const struct tcp_sock *tp = tcp_sk(sk);
4946

4947 4948 4949 4950
	/* 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
4951
		return false;
4952 4953

	/* If we are under global TCP memory pressure, do not expand.  */
4954
	if (tcp_under_memory_pressure(sk))
E
Eric Dumazet 已提交
4955
		return false;
4956 4957

	/* If we are under soft global TCP memory pressure, do not expand.  */
4958
	if (sk_memory_allocated(sk) >= sk_prot_mem_limits(sk, 0))
E
Eric Dumazet 已提交
4959
		return false;
4960 4961

	/* If we filled the congestion window, do not expand.  */
4962
	if (tcp_packets_in_flight(tp) >= tp->snd_cwnd)
E
Eric Dumazet 已提交
4963
		return false;
4964

E
Eric Dumazet 已提交
4965
	return true;
4966
}
L
Linus Torvalds 已提交
4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977

/* 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);

4978
	if (tcp_should_expand_sndbuf(sk)) {
E
Eric Dumazet 已提交
4979
		tcp_sndbuf_expand(sk);
4980
		tp->snd_cwnd_stamp = tcp_jiffies32;
L
Linus Torvalds 已提交
4981 4982 4983 4984 4985
	}

	sk->sk_write_space(sk);
}

S
Stephen Hemminger 已提交
4986
static void tcp_check_space(struct sock *sk)
L
Linus Torvalds 已提交
4987 4988 4989
{
	if (sock_flag(sk, SOCK_QUEUE_SHRUNK)) {
		sock_reset_flag(sk, SOCK_QUEUE_SHRUNK);
4990
		/* pairs with tcp_poll() */
4991
		smp_mb();
L
Linus Torvalds 已提交
4992
		if (sk->sk_socket &&
4993
		    test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
L
Linus Torvalds 已提交
4994
			tcp_new_space(sk);
4995 4996 4997
			if (!test_bit(SOCK_NOSPACE, &sk->sk_socket->flags))
				tcp_chrono_stop(sk, TCP_CHRONO_SNDBUF_LIMITED);
		}
L
Linus Torvalds 已提交
4998 4999 5000
	}
}

5001
static inline void tcp_data_snd_check(struct sock *sk)
L
Linus Torvalds 已提交
5002
{
5003
	tcp_push_pending_frames(sk);
L
Linus Torvalds 已提交
5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014
	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... */
5015
	if (((tp->rcv_nxt - tp->rcv_wup) > inet_csk(sk)->icsk_ack.rcv_mss &&
L
Linus Torvalds 已提交
5016 5017 5018
	     /* ... and right edge of window advances far enough.
	      * (tcp_recvmsg() will send ACK otherwise). Or...
	      */
5019
	     __tcp_select_window(sk) >= tp->rcv_wnd) ||
L
Linus Torvalds 已提交
5020
	    /* We ACK each frame or... */
5021
	    tcp_in_quickack_mode(sk) ||
L
Linus Torvalds 已提交
5022
	    /* We have out of order data. */
5023
	    (ofo_possible && !RB_EMPTY_ROOT(&tp->out_of_order_queue))) {
L
Linus Torvalds 已提交
5024 5025 5026 5027 5028 5029 5030 5031
		/* Then ack it now */
		tcp_send_ack(sk);
	} else {
		/* Else, send delayed ack. */
		tcp_send_delayed_ack(sk);
	}
}

S
Stephen Hemminger 已提交
5032
static inline void tcp_ack_snd_check(struct sock *sk)
L
Linus Torvalds 已提交
5033
{
5034
	if (!inet_csk_ack_scheduled(sk)) {
L
Linus Torvalds 已提交
5035 5036 5037 5038 5039 5040 5041 5042
		/* 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
5043
 *	signaled. Its the 'slow' part of tcp_urg. It could be
L
Linus Torvalds 已提交
5044 5045 5046 5047 5048 5049
 *	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).
 */
5050

5051
static void tcp_check_urg(struct sock *sk, const struct tcphdr *th)
L
Linus Torvalds 已提交
5052 5053 5054 5055
{
	struct tcp_sock *tp = tcp_sk(sk);
	u32 ptr = ntohs(th->urg_ptr);

5056
	if (ptr && !sock_net(sk)->ipv4.sysctl_tcp_stdurg)
L
Linus Torvalds 已提交
5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087
		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
5088
	 * or we break the semantics of SIOCATMARK (and thus sockatmark())
L
Linus Torvalds 已提交
5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099
	 *
	 * 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 &&
5100
	    !sock_flag(sk, SOCK_URGINLINE) && tp->copied_seq != tp->rcv_nxt) {
L
Linus Torvalds 已提交
5101 5102 5103
		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
5104
			__skb_unlink(skb, &sk->sk_receive_queue);
L
Linus Torvalds 已提交
5105 5106 5107 5108
			__kfree_skb(skb);
		}
	}

5109 5110
	tp->urg_data = TCP_URG_NOTYET;
	tp->urg_seq = ptr;
5111 5112 5113

	/* Disable header prediction. */
	tp->pred_flags = 0;
L
Linus Torvalds 已提交
5114 5115 5116
}

/* This is the 'fast' part of urgent handling. */
5117
static void tcp_urg(struct sock *sk, struct sk_buff *skb, const struct tcphdr *th)
L
Linus Torvalds 已提交
5118 5119 5120 5121 5122
{
	struct tcp_sock *tp = tcp_sk(sk);

	/* Check if we get a new urgent pointer - normally not. */
	if (th->urg)
5123
		tcp_check_urg(sk, th);
L
Linus Torvalds 已提交
5124 5125 5126 5127 5128 5129

	/* 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;

5130
		/* Is the urgent pointer pointing into this packet? */
L
Linus Torvalds 已提交
5131 5132 5133 5134 5135 5136
		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))
5137
				sk->sk_data_ready(sk);
L
Linus Torvalds 已提交
5138 5139 5140 5141
		}
	}
}

5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157 5158
/* 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));
}

5159 5160 5161
/* Does PAWS and seqno based validation of an incoming segment, flags will
 * play significant role here.
 */
E
Eric Dumazet 已提交
5162 5163
static bool tcp_validate_incoming(struct sock *sk, struct sk_buff *skb,
				  const struct tcphdr *th, int syn_inerr)
5164 5165
{
	struct tcp_sock *tp = tcp_sk(sk);
5166
	bool rst_seq_match = false;
5167 5168

	/* RFC1323: H1. Apply PAWS check first. */
5169 5170
	if (tcp_fast_parse_options(sock_net(sk), skb, th, tp) &&
	    tp->rx_opt.saw_tstamp &&
5171 5172
	    tcp_paws_discard(sk, skb)) {
		if (!th->rst) {
5173
			NET_INC_STATS(sock_net(sk), LINUX_MIB_PAWSESTABREJECTED);
5174 5175 5176 5177
			if (!tcp_oow_rate_limited(sock_net(sk), skb,
						  LINUX_MIB_TCPACKSKIPPEDPAWS,
						  &tp->last_oow_ack_time))
				tcp_send_dupack(sk, skb);
5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190
			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)".
		 */
5191 5192 5193
		if (!th->rst) {
			if (th->syn)
				goto syn_challenge;
5194 5195 5196 5197
			if (!tcp_oow_rate_limited(sock_net(sk), skb,
						  LINUX_MIB_TCPACKSKIPPEDSEQ,
						  &tp->last_oow_ack_time))
				tcp_send_dupack(sk, skb);
5198 5199
		} else if (tcp_reset_check(sk, skb)) {
			tcp_reset(sk);
5200
		}
5201 5202 5203 5204 5205
		goto discard;
	}

	/* Step 2: check RST bit */
	if (th->rst) {
5206 5207 5208 5209
		/* 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,
5210
		 * then
E
Eric Dumazet 已提交
5211 5212 5213 5214
		 *     RESET the connection
		 * else
		 *     Send a challenge ACK
		 */
5215 5216
		if (TCP_SKB_CB(skb)->seq == tp->rcv_nxt ||
		    tcp_reset_check(sk, skb)) {
5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234
			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
5235
			tcp_reset(sk);
5236 5237 5238 5239 5240 5241 5242
		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)
5243
				tcp_fastopen_active_disable(sk);
5244
			tcp_send_challenge_ack(sk, skb);
5245
		}
5246 5247 5248 5249 5250
		goto discard;
	}

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

E
Eric Dumazet 已提交
5251
	/* step 4: Check for a SYN
5252
	 * RFC 5961 4.2 : Send a challenge ack
E
Eric Dumazet 已提交
5253 5254
	 */
	if (th->syn) {
5255
syn_challenge:
5256
		if (syn_inerr)
5257 5258
			TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPSYNCHALLENGE);
5259
		tcp_send_challenge_ack(sk, skb);
E
Eric Dumazet 已提交
5260
		goto discard;
5261 5262
	}

E
Eric Dumazet 已提交
5263
	return true;
5264 5265

discard:
5266
	tcp_drop(sk, skb);
E
Eric Dumazet 已提交
5267
	return false;
5268 5269
}

L
Linus Torvalds 已提交
5270
/*
5271
 *	TCP receive function for the ESTABLISHED state.
5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291
 *
 *	It is split into a fast path and a slow path. The fast path is
 * 	disabled when:
 *	- A zero window was announced from us - zero window probing
 *        is only handled properly in the slow path.
 *	- Out of order segments arrived.
 *	- Urgent data is expected.
 *	- There is no buffer space left
 *	- Unexpected TCP flags/window values/header lengths are received
 *	  (detected by checking the TCP header against pred_flags)
 *	- 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.
 *
 *	When these conditions are not satisfied it drops into a standard
 *	receive procedure patterned after RFC793 to handle all cases.
 *	The first three cases are guaranteed by proper pred_flags setting,
 *	the rest is checked inline. Fast processing is turned on in
 *	tcp_data_queue when everything is OK.
L
Linus Torvalds 已提交
5292
 */
5293
void tcp_rcv_established(struct sock *sk, struct sk_buff *skb,
5294
			 const struct tcphdr *th)
L
Linus Torvalds 已提交
5295
{
5296
	unsigned int len = skb->len;
L
Linus Torvalds 已提交
5297 5298
	struct tcp_sock *tp = tcp_sk(sk);

5299
	tcp_mstamp_refresh(tp);
5300
	if (unlikely(!sk->sk_rx_dst))
E
Eric Dumazet 已提交
5301
		inet_csk(sk)->icsk_af_ops->sk_rx_dst_set(sk, skb);
5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315
	/*
	 *	Header prediction.
	 *	The code loosely follows the one in the famous
	 *	"30 instruction TCP receive" Van Jacobson mail.
	 *
	 *	Van's trick is to deposit buffers into socket queue
	 *	on a device interrupt, to call tcp_recv function
	 *	on the receive process context and checksum and copy
	 *	the buffer to user space. smart...
	 *
	 *	Our current scheme is not silly either but we take the
	 *	extra cost of the net_bh soft interrupt processing...
	 *	We do checksum and copy also but from device to kernel.
	 */
L
Linus Torvalds 已提交
5316 5317 5318

	tp->rx_opt.saw_tstamp = 0;

5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424
	/*	pred_flags is 0xS?10 << 16 + snd_wnd
	 *	if header_prediction is to be made
	 *	'S' will always be tp->tcp_header_len >> 2
	 *	'?' will be 0 for the fast path, otherwise pred_flags is 0 to
	 *  turn it off	(when there are holes in the receive
	 *	 space for instance)
	 *	PSH flag is ignored.
	 */

	if ((tcp_flag_word(th) & TCP_HP_BITS) == tp->pred_flags &&
	    TCP_SKB_CB(skb)->seq == tp->rcv_nxt &&
	    !after(TCP_SKB_CB(skb)->ack_seq, tp->snd_nxt)) {
		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! */
			if (!tcp_parse_aligned_timestamp(tp, th))
				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);
				__kfree_skb(skb);
				tcp_data_snd_check(sk);
				return;
			} else { /* Header too small */
				TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
				goto discard;
			}
		} else {
			int eaten = 0;
			bool fragstolen = false;

			if (tcp_checksum_complete(skb))
				goto csum_error;

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

			/* 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);

			tcp_rcv_rtt_measure_ts(sk, skb);

			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPHPHITS);

			/* Bulk data transfer: receiver */
			eaten = tcp_queue_rcv(sk, skb, tcp_header_len,
					      &fragstolen);

			tcp_event_data_recv(sk, skb);

			if (TCP_SKB_CB(skb)->ack_seq != tp->snd_una) {
				/* Well, only one small jumplet in fast path... */
				tcp_ack(sk, skb, FLAG_DATA);
				tcp_data_snd_check(sk);
				if (!inet_csk_ack_scheduled(sk))
					goto no_ack;
			}

			__tcp_ack_snd_check(sk, 0);
no_ack:
			if (eaten)
				kfree_skb_partial(skb, fragstolen);
			sk->sk_data_ready(sk);
			return;
		}
	}

slow_path:
5425
	if (len < (th->doff << 2) || tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
5426 5427
		goto csum_error;

5428
	if (!th->ack && !th->rst && !th->syn)
5429 5430
		goto discard;

5431 5432 5433 5434
	/*
	 *	Standard slow path.
	 */

E
Eric Dumazet 已提交
5435
	if (!tcp_validate_incoming(sk, skb, th, 1))
5436
		return;
L
Linus Torvalds 已提交
5437

5438 5439
step5:
	if (tcp_ack(sk, skb, FLAG_SLOWPATH | FLAG_UPDATE_TS_RECENT) < 0)
5440
		goto discard;
L
Linus Torvalds 已提交
5441

5442
	tcp_rcv_rtt_measure_ts(sk, skb);
L
Linus Torvalds 已提交
5443 5444 5445 5446 5447 5448 5449

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

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

5450
	tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
5451
	tcp_ack_snd_check(sk);
5452
	return;
L
Linus Torvalds 已提交
5453 5454

csum_error:
5455 5456
	TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS);
	TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
5457 5458

discard:
5459
	tcp_drop(sk, skb);
L
Linus Torvalds 已提交
5460
}
E
Eric Dumazet 已提交
5461
EXPORT_SYMBOL(tcp_rcv_established);
L
Linus Torvalds 已提交
5462

P
Pavel Emelyanov 已提交
5463 5464 5465 5466 5467 5468
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);
5469
	icsk->icsk_ack.lrcvtime = tcp_jiffies32;
P
Pavel Emelyanov 已提交
5470

5471
	if (skb) {
E
Eric Dumazet 已提交
5472
		icsk->icsk_af_ops->sk_rx_dst_set(sk, skb);
P
Pavel Emelyanov 已提交
5473
		security_inet_conn_established(sk, skb);
D
David S. Miller 已提交
5474
	}
P
Pavel Emelyanov 已提交
5475

5476
	tcp_init_transfer(sk, BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB);
P
Pavel Emelyanov 已提交
5477 5478 5479 5480

	/* Prevent spurious tcp_cwnd_restart() on first data
	 * packet.
	 */
5481
	tp->lsndtime = tcp_jiffies32;
P
Pavel Emelyanov 已提交
5482 5483 5484

	if (sock_flag(sk, SOCK_KEEPOPEN))
		inet_csk_reset_keepalive_timer(sk, keepalive_time_when(tp));
5485 5486 5487 5488 5489

	if (!tp->rx_opt.snd_wscale)
		__tcp_fast_path_on(tp, tp->snd_wnd);
	else
		tp->pred_flags = 0;
P
Pavel Emelyanov 已提交
5490 5491
}

5492 5493 5494 5495
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);
5496
	struct sk_buff *data = tp->syn_data ? tcp_rtx_queue_head(sk) : NULL;
5497 5498
	u16 mss = tp->rx_opt.mss_clamp, try_exp = 0;
	bool syn_drop = false;
5499 5500 5501 5502 5503 5504 5505

	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;
5506
		tcp_parse_options(sock_net(sk), synack, &opt, 0, NULL);
5507 5508 5509
		mss = opt.mss_clamp;
	}

5510 5511
	if (!tp->syn_fastopen) {
		/* Ignore an unsolicited cookie */
5512
		cookie->len = -1;
5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526
	} 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;
	}
5527

5528
	tcp_fastopen_cache_set(sk, mss, cookie, syn_drop, try_exp);
5529 5530

	if (data) { /* Retransmit unacked data in SYN */
5531 5532
		skb_rbtree_walk_from(data) {
			if (__tcp_retransmit_skb(sk, data, 1))
5533 5534
				break;
		}
5535
		tcp_rearm_rto(sk);
5536
		NET_INC_STATS(sock_net(sk),
5537
				LINUX_MIB_TCPFASTOPENACTIVEFAIL);
5538 5539
		return true;
	}
Y
Yuchung Cheng 已提交
5540
	tp->syn_data_acked = tp->syn_data;
5541
	if (tp->syn_data_acked)
5542
		NET_INC_STATS(sock_net(sk),
5543
				LINUX_MIB_TCPFASTOPENACTIVE);
5544 5545 5546

	tcp_fastopen_add_skb(sk, synack);

5547 5548 5549
	return false;
}

5550 5551 5552 5553 5554 5555 5556 5557 5558 5559
static void smc_check_reset_syn(struct tcp_sock *tp)
{
#if IS_ENABLED(CONFIG_SMC)
	if (static_branch_unlikely(&tcp_have_smc)) {
		if (tp->syn_smc && !tp->rx_opt.smc_ok)
			tp->syn_smc = 0;
	}
#endif
}

L
Linus Torvalds 已提交
5560
static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
5561
					 const struct tcphdr *th)
L
Linus Torvalds 已提交
5562
{
5563
	struct inet_connection_sock *icsk = inet_csk(sk);
5564
	struct tcp_sock *tp = tcp_sk(sk);
5565
	struct tcp_fastopen_cookie foc = { .len = -1 };
5566
	int saved_clamp = tp->rx_opt.mss_clamp;
5567
	bool fastopen_fail;
L
Linus Torvalds 已提交
5568

5569
	tcp_parse_options(sock_net(sk), skb, &tp->rx_opt, 0, &foc);
5570
	if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr)
5571
		tp->rx_opt.rcv_tsecr -= tp->tsoffset;
L
Linus Torvalds 已提交
5572 5573 5574 5575 5576 5577 5578 5579 5580 5581

	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)"
		 */
5582 5583
		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
5584 5585 5586 5587
			goto reset_and_undo;

		if (tp->rx_opt.saw_tstamp && tp->rx_opt.rcv_tsecr &&
		    !between(tp->rx_opt.rcv_tsecr, tp->retrans_stamp,
5588
			     tcp_time_stamp(tp))) {
5589
			NET_INC_STATS(sock_net(sk),
5590
					LINUX_MIB_PAWSACTIVEREJECTED);
L
Linus Torvalds 已提交
5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623
			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..."
		 */

5624
		tcp_ecn_rcv_synack(tp, th);
L
Linus Torvalds 已提交
5625

5626
		tcp_init_wl(tp, TCP_SKB_CB(skb)->seq);
5627
		tcp_ack(sk, skb, FLAG_SLOWPATH);
L
Linus Torvalds 已提交
5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654

		/* 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);
		}

5655
		tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
5656 5657 5658 5659 5660 5661
		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;
5662

5663 5664
		smc_check_reset_syn(tp);

R
Ralf Baechle 已提交
5665
		smp_mb();
L
Linus Torvalds 已提交
5666

P
Pavel Emelyanov 已提交
5667
		tcp_finish_connect(sk, skb);
L
Linus Torvalds 已提交
5668

5669 5670
		fastopen_fail = (tp->syn_fastopen || tp->syn_data) &&
				tcp_rcv_fastopen_synack(sk, skb, &foc);
5671

5672 5673 5674 5675 5676 5677
		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;
5678 5679 5680
		if (sk->sk_write_pending ||
		    icsk->icsk_accept_queue.rskq_defer_accept ||
		    icsk->icsk_ack.pingpong) {
L
Linus Torvalds 已提交
5681 5682 5683 5684 5685 5686 5687
			/* 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
			 */
5688 5689
			inet_csk_schedule_ack(sk);
			tcp_enter_quickack_mode(sk);
5690 5691
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
						  TCP_DELACK_MAX, TCP_RTO_MAX);
L
Linus Torvalds 已提交
5692 5693

discard:
5694
			tcp_drop(sk, skb);
L
Linus Torvalds 已提交
5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714
			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. */
5715
	if (tp->rx_opt.ts_recent_stamp && tp->rx_opt.saw_tstamp &&
I
Ilpo Järvinen 已提交
5716
	    tcp_paws_reject(&tp->rx_opt, 0))
L
Linus Torvalds 已提交
5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735
		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;
5736
		tp->copied_seq = tp->rcv_nxt;
L
Linus Torvalds 已提交
5737 5738 5739 5740 5741 5742 5743 5744 5745
		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;

5746
		tcp_ecn_rcv_syn(tp, th);
L
Linus Torvalds 已提交
5747

J
John Heffner 已提交
5748
		tcp_mtup_init(sk);
5749
		tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
5750 5751 5752 5753 5754
		tcp_initialize_rcv_mss(sk);

		tcp_send_synack(sk);
#if 0
		/* Note, we could accept data and URG from this segment.
5755 5756 5757
		 * 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
5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785 5786
		 *
		 * 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
5787
 *	all states except ESTABLISHED and TIME_WAIT.
L
Linus Torvalds 已提交
5788 5789 5790
 *	It's called from both tcp_v4_rcv and tcp_v6_rcv and should be
 *	address independent.
 */
5791

5792
int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
5793 5794
{
	struct tcp_sock *tp = tcp_sk(sk);
5795
	struct inet_connection_sock *icsk = inet_csk(sk);
5796
	const struct tcphdr *th = tcp_hdr(skb);
5797
	struct request_sock *req;
L
Linus Torvalds 已提交
5798
	int queued = 0;
5799
	bool acceptable;
L
Linus Torvalds 已提交
5800 5801 5802 5803 5804 5805

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

	case TCP_LISTEN:
S
Stephen Hemminger 已提交
5806
		if (th->ack)
L
Linus Torvalds 已提交
5807 5808
			return 1;

S
Stephen Hemminger 已提交
5809
		if (th->rst)
L
Linus Torvalds 已提交
5810 5811
			goto discard;

S
Stephen Hemminger 已提交
5812
		if (th->syn) {
E
Eric Dumazet 已提交
5813 5814
			if (th->fin)
				goto discard;
5815 5816 5817 5818 5819 5820
			/* 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
5821

5822 5823
			if (!acceptable)
				return 1;
5824
			consume_skb(skb);
5825
			return 0;
L
Linus Torvalds 已提交
5826 5827 5828 5829
		}
		goto discard;

	case TCP_SYN_SENT:
5830
		tp->rx_opt.saw_tstamp = 0;
5831
		tcp_mstamp_refresh(tp);
5832
		queued = tcp_rcv_synsent_state_process(sk, skb, th);
L
Linus Torvalds 已提交
5833 5834 5835 5836 5837 5838
		if (queued >= 0)
			return queued;

		/* Do step6 onward by hand. */
		tcp_urg(sk, skb, th);
		__kfree_skb(skb);
5839
		tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
5840 5841 5842
		return 0;
	}

5843
	tcp_mstamp_refresh(tp);
5844
	tp->rx_opt.saw_tstamp = 0;
5845
	req = tp->fastopen_rsk;
5846
	if (req) {
5847
		WARN_ON_ONCE(sk->sk_state != TCP_SYN_RECV &&
5848 5849
		    sk->sk_state != TCP_FIN_WAIT1);

5850
		if (!tcp_check_req(sk, skb, req, true))
5851
			goto discard;
5852
	}
5853

5854
	if (!th->ack && !th->rst && !th->syn)
5855 5856
		goto discard;

5857
	if (!tcp_validate_incoming(sk, skb, th, 0))
E
Eric Dumazet 已提交
5858
		return 0;
L
Linus Torvalds 已提交
5859 5860

	/* step 5: check the ACK field */
5861 5862
	acceptable = tcp_ack(sk, skb, FLAG_SLOWPATH |
				      FLAG_UPDATE_TS_RECENT |
5863
				      FLAG_NO_CHALLENGE_ACK) > 0;
5864

5865 5866 5867 5868 5869 5870
	if (!acceptable) {
		if (sk->sk_state == TCP_SYN_RECV)
			return 1;	/* send one RST */
		tcp_send_challenge_ack(sk, skb);
		goto discard;
	}
5871 5872
	switch (sk->sk_state) {
	case TCP_SYN_RECV:
Y
Yuchung Cheng 已提交
5873 5874 5875
		if (!tp->srtt_us)
			tcp_synack_rtt_meas(sk, req);

5876 5877 5878 5879
		/* Once we leave TCP_SYN_RECV, we no longer need req
		 * so release it.
		 */
		if (req) {
5880
			inet_csk(sk)->icsk_retransmits = 0;
5881
			reqsk_fastopen_remove(sk, req, false);
5882 5883 5884 5885 5886 5887 5888 5889 5890
			/* 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.
			 */
			tcp_rearm_rto(sk);
5891
		} else {
5892
			tcp_init_transfer(sk, BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB);
5893 5894 5895 5896 5897
			tp->copied_seq = tp->rcv_nxt;
		}
		smp_mb();
		tcp_set_state(sk, TCP_ESTABLISHED);
		sk->sk_state_change(sk);
L
Linus Torvalds 已提交
5898

5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912
		/* 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;

5913 5914
		if (!inet_csk(sk)->icsk_ca_ops->cong_control)
			tcp_update_pacing_rate(sk);
5915

5916
		/* Prevent spurious tcp_cwnd_restart() on first data packet */
5917
		tp->lsndtime = tcp_jiffies32;
5918 5919

		tcp_initialize_rcv_mss(sk);
5920
		tcp_fast_path_on(tp);
5921 5922
		break;

5923 5924 5925
	case TCP_FIN_WAIT1: {
		int tmo;

5926 5927 5928 5929 5930
		/* 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.
		 */
5931
		if (req) {
5932 5933 5934 5935
			/* We no longer need the request sock. */
			reqsk_fastopen_remove(sk, req, false);
			tcp_rearm_rto(sk);
		}
5936 5937
		if (tp->snd_una != tp->write_seq)
			break;
5938

5939 5940
		tcp_set_state(sk, TCP_FIN_WAIT2);
		sk->sk_shutdown |= SEND_SHUTDOWN;
5941

5942
		sk_dst_confirm(sk);
5943

5944 5945 5946 5947 5948
		if (!sock_flag(sk, SOCK_DEAD)) {
			/* Wake up lingering close() */
			sk->sk_state_change(sk);
			break;
		}
L
Linus Torvalds 已提交
5949

5950 5951 5952 5953 5954 5955 5956 5957 5958
		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)
5959
				tcp_fastopen_active_disable(sk);
5960
			tcp_done(sk);
5961
			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONDATA);
5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978
			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;
5979 5980
		}
		break;
5981
	}
L
Linus Torvalds 已提交
5982

5983 5984 5985 5986 5987 5988
	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
5989

5990 5991 5992 5993 5994
	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
5995
		}
5996
		break;
5997
	}
L
Linus Torvalds 已提交
5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008

	/* 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;
6009
		/* fall through */
L
Linus Torvalds 已提交
6010 6011 6012
	case TCP_FIN_WAIT1:
	case TCP_FIN_WAIT2:
		/* RFC 793 says to queue data in these states,
6013
		 * RFC 1122 says we MUST send a reset.
L
Linus Torvalds 已提交
6014 6015 6016 6017 6018
		 * 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)) {
6019
				NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTONDATA);
L
Linus Torvalds 已提交
6020 6021 6022 6023 6024
				tcp_reset(sk);
				return 1;
			}
		}
		/* Fall through */
6025
	case TCP_ESTABLISHED:
L
Linus Torvalds 已提交
6026 6027 6028 6029 6030 6031 6032
		tcp_data_queue(sk, skb);
		queued = 1;
		break;
	}

	/* tcp_data could move socket to TIME-WAIT */
	if (sk->sk_state != TCP_CLOSE) {
6033
		tcp_data_snd_check(sk);
L
Linus Torvalds 已提交
6034 6035 6036
		tcp_ack_snd_check(sk);
	}

6037
	if (!queued) {
L
Linus Torvalds 已提交
6038
discard:
6039
		tcp_drop(sk, skb);
L
Linus Torvalds 已提交
6040 6041 6042 6043
	}
	return 0;
}
EXPORT_SYMBOL(tcp_rcv_state_process);
O
Octavian Purdila 已提交
6044 6045 6046 6047 6048 6049

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)
6050 6051
		net_dbg_ratelimited("drop open request from %pI4/%u\n",
				    &ireq->ir_rmt_addr, port);
6052 6053
#if IS_ENABLED(CONFIG_IPV6)
	else if (family == AF_INET6)
6054 6055
		net_dbg_ratelimited("drop open request from %pI6/%u\n",
				    &ireq->ir_v6_rmt_addr, port);
6056
#endif
O
Octavian Purdila 已提交
6057 6058
}

6059 6060 6061 6062 6063
/* 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
6064
 * TCP ECN negotiation.
6065 6066
 *
 * Exception: tcp_ca wants ECN. This is required for DCTCP
6067 6068 6069
 * congestion control: Linux DCTCP asserts ECT on all packets,
 * including SYN, which is most optimal solution; however,
 * others, such as FreeBSD do not.
6070 6071 6072
 */
static void tcp_ecn_create_request(struct request_sock *req,
				   const struct sk_buff *skb,
6073 6074
				   const struct sock *listen_sk,
				   const struct dst_entry *dst)
6075 6076 6077 6078
{
	const struct tcphdr *th = tcp_hdr(skb);
	const struct net *net = sock_net(listen_sk);
	bool th_ecn = th->ece && th->cwr;
6079
	bool ect, ecn_ok;
6080
	u32 ecn_ok_dst;
6081 6082 6083 6084 6085

	if (!th_ecn)
		return;

	ect = !INET_ECN_is_not_ect(TCP_SKB_CB(skb)->ip_dsfield);
6086 6087
	ecn_ok_dst = dst_feature(dst, DST_FEATURE_ECN_MASK);
	ecn_ok = net->ipv4.sysctl_tcp_ecn || ecn_ok_dst;
6088

6089
	if ((!ect && ecn_ok) || tcp_ca_needs_ecn(listen_sk) ||
6090 6091
	    (ecn_ok_dst & DST_FEATURE_ECN_CA) ||
	    tcp_bpf_ca_needs_ecn((struct sock *)req))
6092 6093 6094
		inet_rsk(req)->ecn_ok = 1;
}

6095 6096 6097 6098 6099 6100
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);

6101
	req->rsk_rcv_wnd = 0;		/* So that tcp_send_synack() knows! */
6102 6103 6104
	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;
6105
	tcp_rsk(req)->snt_synack = tcp_clock_us();
6106 6107 6108 6109 6110 6111 6112 6113 6114 6115 6116 6117
	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);
6118 6119 6120
#if IS_ENABLED(CONFIG_SMC)
	ireq->smc_ok = rx_opt->smc_ok;
#endif
6121 6122
}

E
Eric Dumazet 已提交
6123
struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
6124 6125
				      struct sock *sk_listener,
				      bool attach_listener)
E
Eric Dumazet 已提交
6126
{
6127 6128
	struct request_sock *req = reqsk_alloc(ops, sk_listener,
					       attach_listener);
E
Eric Dumazet 已提交
6129 6130 6131 6132 6133

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

		kmemcheck_annotate_bitfield(ireq, flags);
E
Eric Dumazet 已提交
6134
		ireq->ireq_opt = NULL;
6135 6136 6137
#if IS_ENABLED(CONFIG_IPV6)
		ireq->pktopts = NULL;
#endif
E
Eric Dumazet 已提交
6138 6139 6140
		atomic64_set(&ireq->ir_cookie, 0);
		ireq->ireq_state = TCP_NEW_SYN_RECV;
		write_pnet(&ireq->ireq_net, sock_net(sk_listener));
6141
		ireq->ireq_family = sk_listener->sk_family;
E
Eric Dumazet 已提交
6142 6143 6144 6145 6146 6147
	}

	return req;
}
EXPORT_SYMBOL(inet_reqsk_alloc);

6148 6149 6150
/*
 * Return true if a syncookie should be sent
 */
6151
static bool tcp_syn_flood_action(const struct sock *sk,
6152 6153 6154
				 const struct sk_buff *skb,
				 const char *proto)
{
6155
	struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue;
6156 6157
	const char *msg = "Dropping request";
	bool want_cookie = false;
6158
	struct net *net = sock_net(sk);
6159 6160

#ifdef CONFIG_SYN_COOKIES
6161
	if (net->ipv4.sysctl_tcp_syncookies) {
6162 6163
		msg = "Sending cookies";
		want_cookie = true;
6164
		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPREQQFULLDOCOOKIES);
6165 6166
	} else
#endif
6167
		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPREQQFULLDROP);
6168

6169
	if (!queue->synflood_warned &&
6170
	    net->ipv4.sysctl_tcp_syncookies != 2 &&
6171
	    xchg(&queue->synflood_warned, 1) == 0)
6172 6173
		pr_info("%s: Possible SYN flooding on port %d. %s.  Check SNMP counters.\n",
			proto, ntohs(tcp_hdr(skb)->dest), msg);
6174

6175 6176 6177
	return want_cookie;
}

6178 6179 6180 6181 6182 6183 6184 6185 6186 6187 6188 6189 6190 6191 6192 6193 6194
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
6195 6196 6197 6198
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)
{
6199 6200
	struct tcp_fastopen_cookie foc = { .len = -1 };
	__u32 isn = TCP_SKB_CB(skb)->tcp_tw_isn;
O
Octavian Purdila 已提交
6201 6202
	struct tcp_options_received tmp_opt;
	struct tcp_sock *tp = tcp_sk(sk);
6203
	struct net *net = sock_net(sk);
6204 6205 6206
	struct sock *fastopen_sk = NULL;
	struct request_sock *req;
	bool want_cookie = false;
6207
	struct dst_entry *dst;
O
Octavian Purdila 已提交
6208 6209 6210 6211 6212 6213
	struct flowi fl;

	/* TW buckets are converted to open requests without
	 * limitations, they conserve resources and peer is
	 * evidently real one.
	 */
6214
	if ((net->ipv4.sysctl_tcp_syncookies == 2 ||
O
Octavian Purdila 已提交
6215 6216 6217 6218 6219 6220
	     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;
	}

6221
	if (sk_acceptq_is_full(sk)) {
6222
		NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
O
Octavian Purdila 已提交
6223 6224 6225
		goto drop;
	}

6226
	req = inet_reqsk_alloc(rsk_ops, sk, !want_cookie);
O
Octavian Purdila 已提交
6227 6228 6229 6230
	if (!req)
		goto drop;

	tcp_rsk(req)->af_specific = af_ops;
6231
	tcp_rsk(req)->ts_off = 0;
O
Octavian Purdila 已提交
6232 6233 6234 6235

	tcp_clear_options(&tmp_opt);
	tmp_opt.mss_clamp = af_ops->mss_clamp;
	tmp_opt.user_mss  = tp->rx_opt.user_mss;
6236 6237
	tcp_parse_options(sock_net(sk), skb, &tmp_opt, 0,
			  want_cookie ? NULL : &foc);
O
Octavian Purdila 已提交
6238 6239 6240 6241 6242 6243

	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);
6244
	inet_rsk(req)->no_srccheck = inet_sk(sk)->transparent;
O
Octavian Purdila 已提交
6245

6246
	/* Note: tcp_v6_init_req() might override ir_iif for link locals */
6247
	inet_rsk(req)->ir_iif = inet_request_bound_dev_if(sk, skb);
6248

O
Octavian Purdila 已提交
6249 6250 6251 6252 6253
	af_ops->init_req(req, sk, skb);

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

6254
	if (tmp_opt.tstamp_ok)
6255
		tcp_rsk(req)->ts_off = af_ops->init_ts_off(net, skb);
6256

6257 6258 6259 6260
	dst = af_ops->route_req(sk, &fl, req);
	if (!dst)
		goto drop_and_free;

6261
	if (!want_cookie && !isn) {
O
Octavian Purdila 已提交
6262
		/* Kill the following clause, if you dislike this way. */
6263 6264 6265 6266
		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
6267 6268 6269 6270 6271 6272 6273 6274 6275 6276 6277 6278
			/* 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;
		}

6279
		isn = af_ops->init_seq(skb);
O
Octavian Purdila 已提交
6280 6281
	}

6282 6283 6284 6285 6286 6287 6288 6289 6290
	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
6291
	tcp_rsk(req)->snt_isn = isn;
6292
	tcp_rsk(req)->txhash = net_tx_rndhash();
O
Octavian Purdila 已提交
6293
	tcp_openreq_init_rwin(req, sk, dst);
6294 6295
	if (!want_cookie) {
		tcp_reqsk_record_syn(sk, req, skb);
6296
		fastopen_sk = tcp_try_fastopen(sk, skb, req, &foc, dst);
6297
	}
6298
	if (fastopen_sk) {
6299
		af_ops->send_synack(fastopen_sk, dst, &fl, req,
6300
				    &foc, TCP_SYNACK_FASTOPEN);
6301 6302 6303 6304
		/* 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);
6305 6306
		sock_put(fastopen_sk);
	} else {
6307
		tcp_rsk(req)->tfo_listener = false;
6308
		if (!want_cookie)
6309 6310
			inet_csk_reqsk_queue_hash_add(sk, req,
				tcp_timeout_init((struct sock *)req));
6311 6312 6313
		af_ops->send_synack(sk, dst, &fl, req, &foc,
				    !want_cookie ? TCP_SYNACK_NORMAL :
						   TCP_SYNACK_COOKIE);
6314 6315 6316 6317
		if (want_cookie) {
			reqsk_free(req);
			return 0;
		}
O
Octavian Purdila 已提交
6318
	}
6319
	reqsk_put(req);
O
Octavian Purdila 已提交
6320 6321 6322 6323 6324 6325 6326
	return 0;

drop_and_release:
	dst_release(dst);
drop_and_free:
	reqsk_free(req);
drop:
6327
	tcp_listendrop(sk);
O
Octavian Purdila 已提交
6328 6329 6330
	return 0;
}
EXPORT_SYMBOL(tcp_conn_request);