tcp_output.c 73.4 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 * 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).
 *
 * Version:	$Id: tcp_output.c,v 1.146 2002/02/01 22:01:04 davem Exp $
 *
10
 * Authors:	Ross Biro
L
Linus Torvalds 已提交
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 44
 *		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	:	Retransmit queue handled by TCP.
 *				:	Fragmentation on mtu decrease
 *				:	Segment collapse on retransmit
 *				:	AF independence
 *
 *		Linus Torvalds	:	send_delayed_ack
 *		David S. Miller	:	Charge memory using the right skb
 *					during syn/ack processing.
 *		David S. Miller :	Output engine completely rewritten.
 *		Andrea Arcangeli:	SYNACK carry ts_recent in tsecr.
 *		Cacophonix Gaul :	draft-minshall-nagle-01
 *		J Hadi Salim	:	ECN support
 *
 */

#include <net/tcp.h>

#include <linux/compiler.h>
#include <linux/module.h>

/* People can turn this off for buggy TCP's found in printers etc. */
45
int sysctl_tcp_retrans_collapse __read_mostly = 1;
L
Linus Torvalds 已提交
46

47 48 49
/* People can turn this on to  work with those rare, broken TCPs that
 * interpret the window field as a signed quantity.
 */
50
int sysctl_tcp_workaround_signed_windows __read_mostly = 0;
51

L
Linus Torvalds 已提交
52 53 54 55
/* This limits the percentage of the congestion window which we
 * will allow a single TSO frame to consume.  Building TSO frames
 * which are too large can cause TCP streams to be bursty.
 */
56
int sysctl_tcp_tso_win_divisor __read_mostly = 3;
L
Linus Torvalds 已提交
57

58 59
int sysctl_tcp_mtu_probing __read_mostly = 0;
int sysctl_tcp_base_mss __read_mostly = 512;
J
John Heffner 已提交
60

61
/* By default, RFC2861 behavior.  */
62
int sysctl_tcp_slow_start_after_idle __read_mostly = 1;
63

64 65 66 67 68 69 70 71 72 73 74 75
static inline void tcp_packets_out_inc(struct sock *sk,
				       const struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);
	int orig = tp->packets_out;

	tp->packets_out += tcp_skb_pcount(skb);
	if (!orig)
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
					  inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
}

76
static void update_send_head(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
77
{
78 79
	struct tcp_sock *tp = tcp_sk(sk);

80
	tcp_advance_send_head(sk, skb);
L
Linus Torvalds 已提交
81
	tp->snd_nxt = TCP_SKB_CB(skb)->end_seq;
82
	tcp_packets_out_inc(sk, skb);
L
Linus Torvalds 已提交
83 84 85 86 87 88 89 90
}

/* SND.NXT, if window was not shrunk.
 * If window has been shrunk, what should we make? It is not clear at all.
 * Using SND.UNA we will fail to open window, SND.NXT is out of window. :-(
 * Anything in between SND.UNA...SND.UNA+SND.WND also can be already
 * invalid. OK, let's make this for now:
 */
91
static inline __u32 tcp_acceptable_seq(struct sock *sk)
L
Linus Torvalds 已提交
92
{
93 94
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
	if (!before(tp->snd_una+tp->snd_wnd, tp->snd_nxt))
		return tp->snd_nxt;
	else
		return tp->snd_una+tp->snd_wnd;
}

/* Calculate mss to advertise in SYN segment.
 * RFC1122, RFC1063, draft-ietf-tcpimpl-pmtud-01 state that:
 *
 * 1. It is independent of path mtu.
 * 2. Ideally, it is maximal possible segment size i.e. 65535-40.
 * 3. For IPv4 it is reasonable to calculate it from maximal MTU of
 *    attached devices, because some buggy hosts are confused by
 *    large MSS.
 * 4. We do not make 3, we advertise MSS, calculated from first
 *    hop device mtu, but allow to raise it to ip_rt_min_advmss.
 *    This may be overridden via information stored in routing table.
 * 5. Value 65535 for MSS is valid in IPv6 and means "as large as possible,
 *    probably even Jumbo".
 */
static __u16 tcp_advertise_mss(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct dst_entry *dst = __sk_dst_get(sk);
	int mss = tp->advmss;

	if (dst && dst_metric(dst, RTAX_ADVMSS) < mss) {
		mss = dst_metric(dst, RTAX_ADVMSS);
		tp->advmss = mss;
	}

	return (__u16)mss;
}

/* RFC2861. Reset CWND after idle period longer RTO to "restart window".
 * This is the first part of cwnd validation mechanism. */
131
static void tcp_cwnd_restart(struct sock *sk, struct dst_entry *dst)
L
Linus Torvalds 已提交
132
{
133
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
134 135 136 137
	s32 delta = tcp_time_stamp - tp->lsndtime;
	u32 restart_cwnd = tcp_init_cwnd(tp, dst);
	u32 cwnd = tp->snd_cwnd;

138
	tcp_ca_event(sk, CA_EVENT_CWND_RESTART);
L
Linus Torvalds 已提交
139

140
	tp->snd_ssthresh = tcp_current_ssthresh(sk);
L
Linus Torvalds 已提交
141 142
	restart_cwnd = min(restart_cwnd, cwnd);

143
	while ((delta -= inet_csk(sk)->icsk_rto) > 0 && cwnd > restart_cwnd)
L
Linus Torvalds 已提交
144 145 146 147 148 149
		cwnd >>= 1;
	tp->snd_cwnd = max(cwnd, restart_cwnd);
	tp->snd_cwnd_stamp = tcp_time_stamp;
	tp->snd_cwnd_used = 0;
}

S
Stephen Hemminger 已提交
150 151
static void tcp_event_data_sent(struct tcp_sock *tp,
				struct sk_buff *skb, struct sock *sk)
L
Linus Torvalds 已提交
152
{
153 154
	struct inet_connection_sock *icsk = inet_csk(sk);
	const u32 now = tcp_time_stamp;
L
Linus Torvalds 已提交
155

156 157
	if (sysctl_tcp_slow_start_after_idle &&
	    (!tp->packets_out && (s32)(now - tp->lsndtime) > icsk->icsk_rto))
158
		tcp_cwnd_restart(sk, __sk_dst_get(sk));
L
Linus Torvalds 已提交
159 160 161 162 163 164

	tp->lsndtime = now;

	/* If it is a reply for ato after last received
	 * packet, enter pingpong mode.
	 */
165 166
	if ((u32)(now - icsk->icsk_ack.lrcvtime) < icsk->icsk_ack.ato)
		icsk->icsk_ack.pingpong = 1;
L
Linus Torvalds 已提交
167 168
}

S
Stephen Hemminger 已提交
169
static inline void tcp_event_ack_sent(struct sock *sk, unsigned int pkts)
L
Linus Torvalds 已提交
170
{
171 172
	tcp_dec_quickack_mode(sk, pkts);
	inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
L
Linus Torvalds 已提交
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197
}

/* Determine a window scaling and initial window to offer.
 * Based on the assumption that the given amount of space
 * will be offered. Store the results in the tp structure.
 * NOTE: for smooth operation initial space offering should
 * be a multiple of mss if possible. We assume here that mss >= 1.
 * This MUST be enforced by all callers.
 */
void tcp_select_initial_window(int __space, __u32 mss,
			       __u32 *rcv_wnd, __u32 *window_clamp,
			       int wscale_ok, __u8 *rcv_wscale)
{
	unsigned int space = (__space < 0 ? 0 : __space);

	/* If no clamp set the clamp to the max possible scaled window */
	if (*window_clamp == 0)
		(*window_clamp) = (65535 << 14);
	space = min(*window_clamp, space);

	/* Quantize space offering to a multiple of mss if possible. */
	if (space > mss)
		space = (space / mss) * mss;

	/* NOTE: offering an initial window larger than 32767
198 199 200 201 202 203
	 * will break some buggy TCP stacks. If the admin tells us
	 * it is likely we could be speaking with such a buggy stack
	 * we will truncate our initial window offering to 32K-1
	 * unless the remote has sent us a window scaling option,
	 * which we interpret as a sign the remote TCP is not
	 * misinterpreting the window field as a signed quantity.
L
Linus Torvalds 已提交
204
	 */
205 206 207 208 209
	if (sysctl_tcp_workaround_signed_windows)
		(*rcv_wnd) = min(space, MAX_TCP_WINDOW);
	else
		(*rcv_wnd) = space;

L
Linus Torvalds 已提交
210 211 212
	(*rcv_wscale) = 0;
	if (wscale_ok) {
		/* Set window scaling on max possible window
213
		 * See RFC1323 for an explanation of the limit to 14
L
Linus Torvalds 已提交
214 215
		 */
		space = max_t(u32, sysctl_tcp_rmem[2], sysctl_rmem_max);
216
		space = min_t(u32, space, *window_clamp);
L
Linus Torvalds 已提交
217 218 219 220 221 222 223
		while (space > 65535 && (*rcv_wscale) < 14) {
			space >>= 1;
			(*rcv_wscale)++;
		}
	}

	/* Set initial window to value enough for senders,
224
	 * following RFC2414. Senders, not following this RFC,
L
Linus Torvalds 已提交
225 226 227
	 * will be satisfied with 2.
	 */
	if (mss > (1<<*rcv_wscale)) {
228 229
		int init_cwnd = 4;
		if (mss > 1460*3)
L
Linus Torvalds 已提交
230
			init_cwnd = 2;
231 232
		else if (mss > 1460)
			init_cwnd = 3;
L
Linus Torvalds 已提交
233 234 235 236 237 238 239 240 241 242 243 244 245
		if (*rcv_wnd > init_cwnd*mss)
			*rcv_wnd = init_cwnd*mss;
	}

	/* Set the clamp no higher than max representable value */
	(*window_clamp) = min(65535U << (*rcv_wscale), *window_clamp);
}

/* Chose a new window to advertise, update state in tcp_sock for the
 * socket, and return result with RFC1323 scaling applied.  The return
 * value can be stuffed directly into th->window for an outgoing
 * frame.
 */
S
Stephen Hemminger 已提交
246
static u16 tcp_select_window(struct sock *sk)
L
Linus Torvalds 已提交
247 248 249 250 251 252
{
	struct tcp_sock *tp = tcp_sk(sk);
	u32 cur_win = tcp_receive_window(tp);
	u32 new_win = __tcp_select_window(sk);

	/* Never shrink the offered window */
S
Stephen Hemminger 已提交
253
	if (new_win < cur_win) {
L
Linus Torvalds 已提交
254 255 256 257 258 259 260 261 262 263 264 265 266 267 268
		/* Danger Will Robinson!
		 * Don't update rcv_wup/rcv_wnd here or else
		 * we will not be able to advertise a zero
		 * window in time.  --DaveM
		 *
		 * Relax Will Robinson.
		 */
		new_win = cur_win;
	}
	tp->rcv_wnd = new_win;
	tp->rcv_wup = tp->rcv_nxt;

	/* Make sure we do not exceed the maximum possible
	 * scaled window.
	 */
269
	if (!tp->rx_opt.rcv_wscale && sysctl_tcp_workaround_signed_windows)
L
Linus Torvalds 已提交
270 271 272 273 274 275 276 277 278 279 280 281 282 283
		new_win = min(new_win, MAX_TCP_WINDOW);
	else
		new_win = min(new_win, (65535U << tp->rx_opt.rcv_wscale));

	/* RFC1323 scaling applied */
	new_win >>= tp->rx_opt.rcv_wscale;

	/* If we advertise zero window, disable fast path. */
	if (new_win == 0)
		tp->pred_flags = 0;

	return new_win;
}

284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
static inline void TCP_ECN_send_synack(struct tcp_sock *tp,
				       struct sk_buff *skb)
{
	TCP_SKB_CB(skb)->flags &= ~TCPCB_FLAG_CWR;
	if (!(tp->ecn_flags&TCP_ECN_OK))
		TCP_SKB_CB(skb)->flags &= ~TCPCB_FLAG_ECE;
}

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

	tp->ecn_flags = 0;
	if (sysctl_tcp_ecn) {
		TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_ECE|TCPCB_FLAG_CWR;
		tp->ecn_flags = TCP_ECN_OK;
	}
}

static __inline__ void
TCP_ECN_make_synack(struct request_sock *req, struct tcphdr *th)
{
	if (inet_rsk(req)->ecn_ok)
		th->ece = 1;
}

static inline void TCP_ECN_send(struct sock *sk, struct sk_buff *skb,
				int tcp_header_len)
{
	struct tcp_sock *tp = tcp_sk(sk);

	if (tp->ecn_flags & TCP_ECN_OK) {
		/* Not-retransmitted data segment: set ECT and inject CWR. */
		if (skb->len != tcp_header_len &&
		    !before(TCP_SKB_CB(skb)->seq, tp->snd_nxt)) {
			INET_ECN_xmit(sk);
			if (tp->ecn_flags&TCP_ECN_QUEUE_CWR) {
				tp->ecn_flags &= ~TCP_ECN_QUEUE_CWR;
				tcp_hdr(skb)->cwr = 1;
				skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
			}
		} else {
			/* ACK or retransmitted segment: clear ECT|CE */
			INET_ECN_dontxmit(sk);
		}
		if (tp->ecn_flags & TCP_ECN_DEMAND_CWR)
			tcp_hdr(skb)->ece = 1;
	}
}

334
static void tcp_build_and_update_options(__be32 *ptr, struct tcp_sock *tp,
335
					 __u32 tstamp, __u8 **md5_hash)
S
Stephen Hemminger 已提交
336 337
{
	if (tp->rx_opt.tstamp_ok) {
338 339 340 341
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
			       (TCPOPT_TIMESTAMP << 8) |
			       TCPOLEN_TIMESTAMP);
S
Stephen Hemminger 已提交
342 343 344 345 346 347 348 349 350 351 352 353
		*ptr++ = htonl(tstamp);
		*ptr++ = htonl(tp->rx_opt.ts_recent);
	}
	if (tp->rx_opt.eff_sacks) {
		struct tcp_sack_block *sp = tp->rx_opt.dsack ? tp->duplicate_sack : tp->selective_acks;
		int this_sack;

		*ptr++ = htonl((TCPOPT_NOP  << 24) |
			       (TCPOPT_NOP  << 16) |
			       (TCPOPT_SACK <<  8) |
			       (TCPOLEN_SACK_BASE + (tp->rx_opt.eff_sacks *
						     TCPOLEN_SACK_PERBLOCK)));
S
Stephen Hemminger 已提交
354 355

		for (this_sack = 0; this_sack < tp->rx_opt.eff_sacks; this_sack++) {
S
Stephen Hemminger 已提交
356 357 358
			*ptr++ = htonl(sp[this_sack].start_seq);
			*ptr++ = htonl(sp[this_sack].end_seq);
		}
S
Stephen Hemminger 已提交
359

S
Stephen Hemminger 已提交
360 361 362 363 364
		if (tp->rx_opt.dsack) {
			tp->rx_opt.dsack = 0;
			tp->rx_opt.eff_sacks--;
		}
	}
365 366 367 368 369 370 371 372 373
#ifdef CONFIG_TCP_MD5SIG
	if (md5_hash) {
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
			       (TCPOPT_MD5SIG << 8) |
			       TCPOLEN_MD5SIG);
		*md5_hash = (__u8 *)ptr;
	}
#endif
S
Stephen Hemminger 已提交
374 375 376 377 378 379
}

/* Construct a tcp options header for a SYN or SYN_ACK packet.
 * If this is every changed make sure to change the definition of
 * MAX_SYN_SIZE to match the new maximum number of options that you
 * can generate.
380 381 382 383
 *
 * Note - that with the RFC2385 TCP option, we make room for the
 * 16 byte MD5 hash. This will be filled in later, so the pointer for the
 * location to be filled is passed back up.
S
Stephen Hemminger 已提交
384
 */
385
static void tcp_syn_build_options(__be32 *ptr, int mss, int ts, int sack,
S
Stephen Hemminger 已提交
386
				  int offer_wscale, int wscale, __u32 tstamp,
387
				  __u32 ts_recent, __u8 **md5_hash)
S
Stephen Hemminger 已提交
388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403
{
	/* We always get an MSS option.
	 * The option bytes which will be seen in normal data
	 * packets should timestamps be used, must be in the MSS
	 * advertised.  But we subtract them from tp->mss_cache so
	 * that calculations in tcp_sendmsg are simpler etc.
	 * So account for this fact here if necessary.  If we
	 * don't do this correctly, as a receiver we won't
	 * recognize data packets as being full sized when we
	 * should, and thus we won't abide by the delayed ACK
	 * rules correctly.
	 * SACKs don't matter, we never delay an ACK when we
	 * have any of those going out.
	 */
	*ptr++ = htonl((TCPOPT_MSS << 24) | (TCPOLEN_MSS << 16) | mss);
	if (ts) {
S
Stephen Hemminger 已提交
404
		if (sack)
405 406 407 408
			*ptr++ = htonl((TCPOPT_SACK_PERM << 24) |
				       (TCPOLEN_SACK_PERM << 16) |
				       (TCPOPT_TIMESTAMP << 8) |
				       TCPOLEN_TIMESTAMP);
S
Stephen Hemminger 已提交
409
		else
410 411 412 413
			*ptr++ = htonl((TCPOPT_NOP << 24) |
				       (TCPOPT_NOP << 16) |
				       (TCPOPT_TIMESTAMP << 8) |
				       TCPOLEN_TIMESTAMP);
S
Stephen Hemminger 已提交
414 415
		*ptr++ = htonl(tstamp);		/* TSVAL */
		*ptr++ = htonl(ts_recent);	/* TSECR */
S
Stephen Hemminger 已提交
416
	} else if (sack)
417 418 419 420
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
			       (TCPOPT_SACK_PERM << 8) |
			       TCPOLEN_SACK_PERM);
S
Stephen Hemminger 已提交
421
	if (offer_wscale)
422 423 424 425
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_WINDOW << 16) |
			       (TCPOLEN_WINDOW << 8) |
			       (wscale));
426 427 428 429 430 431 432 433 434 435 436 437 438 439
#ifdef CONFIG_TCP_MD5SIG
	/*
	 * If MD5 is enabled, then we set the option, and include the size
	 * (always 18). The actual MD5 hash is added just before the
	 * packet is sent.
	 */
	if (md5_hash) {
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
			       (TCPOPT_MD5SIG << 8) |
			       TCPOLEN_MD5SIG);
		*md5_hash = (__u8 *) ptr;
	}
#endif
S
Stephen Hemminger 已提交
440
}
L
Linus Torvalds 已提交
441 442 443 444 445 446 447 448 449 450 451 452

/* This routine actually transmits TCP packets queued in by
 * tcp_do_sendmsg().  This is used by both the initial
 * transmission and possible later retransmissions.
 * All SKB's seen here are completely headerless.  It is our
 * job to build the TCP header, and pass the packet down to
 * IP so it can do the same plus pass the packet off to the
 * device.
 *
 * We are working here with either a clone of the original
 * SKB, or a fresh unique copy made by the retransmit engine.
 */
453
static int tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, int clone_it, gfp_t gfp_mask)
L
Linus Torvalds 已提交
454
{
455 456 457 458 459
	const struct inet_connection_sock *icsk = inet_csk(sk);
	struct inet_sock *inet;
	struct tcp_sock *tp;
	struct tcp_skb_cb *tcb;
	int tcp_header_size;
460 461 462 463
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *md5;
	__u8 *md5_hash_location;
#endif
464 465 466 467 468 469 470 471 472
	struct tcphdr *th;
	int sysctl_flags;
	int err;

	BUG_ON(!skb || !tcp_skb_pcount(skb));

	/* If congestion control is doing timestamping, we must
	 * take such a timestamp before we potentially clone/copy.
	 */
473
	if (icsk->icsk_ca_ops->flags & TCP_CONG_RTT_STAMP)
474 475 476 477 478 479 480 481 482 483
		__net_timestamp(skb);

	if (likely(clone_it)) {
		if (unlikely(skb_cloned(skb)))
			skb = pskb_copy(skb, gfp_mask);
		else
			skb = skb_clone(skb, gfp_mask);
		if (unlikely(!skb))
			return -ENOBUFS;
	}
L
Linus Torvalds 已提交
484

485 486 487 488
	inet = inet_sk(sk);
	tp = tcp_sk(sk);
	tcb = TCP_SKB_CB(skb);
	tcp_header_size = tp->tcp_header_len;
L
Linus Torvalds 已提交
489 490 491 492 493

#define SYSCTL_FLAG_TSTAMPS	0x1
#define SYSCTL_FLAG_WSCALE	0x2
#define SYSCTL_FLAG_SACK	0x4

494 495 496
	sysctl_flags = 0;
	if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
		tcp_header_size = sizeof(struct tcphdr) + TCPOLEN_MSS;
S
Stephen Hemminger 已提交
497
		if (sysctl_tcp_timestamps) {
498 499
			tcp_header_size += TCPOLEN_TSTAMP_ALIGNED;
			sysctl_flags |= SYSCTL_FLAG_TSTAMPS;
L
Linus Torvalds 已提交
500
		}
501 502 503
		if (sysctl_tcp_window_scaling) {
			tcp_header_size += TCPOLEN_WSCALE_ALIGNED;
			sysctl_flags |= SYSCTL_FLAG_WSCALE;
L
Linus Torvalds 已提交
504
		}
505 506 507 508
		if (sysctl_tcp_sack) {
			sysctl_flags |= SYSCTL_FLAG_SACK;
			if (!(sysctl_flags & SYSCTL_FLAG_TSTAMPS))
				tcp_header_size += TCPOLEN_SACKPERM_ALIGNED;
L
Linus Torvalds 已提交
509
		}
510 511 512 513 514 515 516 517
	} else if (unlikely(tp->rx_opt.eff_sacks)) {
		/* A SACK is 2 pad bytes, a 2 byte header, plus
		 * 2 32-bit sequence numbers for each SACK block.
		 */
		tcp_header_size += (TCPOLEN_SACK_BASE_ALIGNED +
				    (tp->rx_opt.eff_sacks *
				     TCPOLEN_SACK_PERBLOCK));
	}
518

519 520 521
	if (tcp_packets_in_flight(tp) == 0)
		tcp_ca_event(sk, CA_EVENT_TX_START);

522 523 524 525 526 527 528 529 530 531
#ifdef CONFIG_TCP_MD5SIG
	/*
	 * Are we doing MD5 on this segment? If so - make
	 * room for it.
	 */
	md5 = tp->af_specific->md5_lookup(sk, sk);
	if (md5)
		tcp_header_size += TCPOLEN_MD5SIG_ALIGNED;
#endif

532 533
	skb_push(skb, tcp_header_size);
	skb_reset_transport_header(skb);
534
	skb_set_owner_w(skb, sk);
535 536

	/* Build TCP header and checksum it. */
537
	th = tcp_hdr(skb);
538 539 540 541
	th->source		= inet->sport;
	th->dest		= inet->dport;
	th->seq			= htonl(tcb->seq);
	th->ack_seq		= htonl(tp->rcv_nxt);
542
	*(((__be16 *)th) + 6)	= htons(((tcp_header_size >> 2) << 12) |
543 544 545 546 547 548
					tcb->flags);

	if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
		/* RFC1323: The window in SYN & SYN/ACK segments
		 * is never scaled.
		 */
549
		th->window	= htons(min(tp->rcv_wnd, 65535U));
550 551 552 553 554
	} else {
		th->window	= htons(tcp_select_window(sk));
	}
	th->check		= 0;
	th->urg_ptr		= 0;
L
Linus Torvalds 已提交
555

556 557 558 559 560
	if (unlikely(tp->urg_mode &&
		     between(tp->snd_up, tcb->seq+1, tcb->seq+0xFFFF))) {
		th->urg_ptr		= htons(tp->snd_up-tcb->seq);
		th->urg			= 1;
	}
L
Linus Torvalds 已提交
561

562
	if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
563
		tcp_syn_build_options((__be32 *)(th + 1),
564 565 566 567 568 569
				      tcp_advertise_mss(sk),
				      (sysctl_flags & SYSCTL_FLAG_TSTAMPS),
				      (sysctl_flags & SYSCTL_FLAG_SACK),
				      (sysctl_flags & SYSCTL_FLAG_WSCALE),
				      tp->rx_opt.rcv_wscale,
				      tcb->when,
570 571 572 573 574 575
				      tp->rx_opt.ts_recent,

#ifdef CONFIG_TCP_MD5SIG
				      md5 ? &md5_hash_location :
#endif
				      NULL);
576
	} else {
577
		tcp_build_and_update_options((__be32 *)(th + 1),
578 579 580 581 582
					     tp, tcb->when,
#ifdef CONFIG_TCP_MD5SIG
					     md5 ? &md5_hash_location :
#endif
					     NULL);
583
		TCP_ECN_send(sk, skb, tcp_header_size);
584
	}
L
Linus Torvalds 已提交
585

586 587 588 589 590 591
#ifdef CONFIG_TCP_MD5SIG
	/* Calculate the MD5 hash, as we have all we need now */
	if (md5) {
		tp->af_specific->calc_md5_hash(md5_hash_location,
					       md5,
					       sk, NULL, NULL,
592
					       tcp_hdr(skb),
593 594 595 596 597
					       sk->sk_protocol,
					       skb->len);
	}
#endif

598
	icsk->icsk_af_ops->send_check(sk, skb->len, skb);
L
Linus Torvalds 已提交
599

600 601
	if (likely(tcb->flags & TCPCB_FLAG_ACK))
		tcp_event_ack_sent(sk, tcp_skb_pcount(skb));
L
Linus Torvalds 已提交
602

603 604
	if (skb->len != tcp_header_size)
		tcp_event_data_sent(tp, skb, sk);
L
Linus Torvalds 已提交
605

606 607
	if (after(tcb->end_seq, tp->snd_nxt) || tcb->seq == tcb->end_seq)
		TCP_INC_STATS(TCP_MIB_OUTSEGS);
L
Linus Torvalds 已提交
608

609
	err = icsk->icsk_af_ops->queue_xmit(skb, 0);
610
	if (likely(err <= 0))
611 612
		return err;

613
	tcp_enter_cwr(sk, 1);
614

615
	return net_xmit_eval(err);
L
Linus Torvalds 已提交
616 617 618 619 620 621 622

#undef SYSCTL_FLAG_TSTAMPS
#undef SYSCTL_FLAG_WSCALE
#undef SYSCTL_FLAG_SACK
}


623
/* This routine just queue's the buffer
L
Linus Torvalds 已提交
624 625 626 627 628 629 630 631 632 633 634
 *
 * NOTE: probe0 timer is not checked, do not forget tcp_push_pending_frames,
 * otherwise socket can stall.
 */
static void tcp_queue_skb(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);

	/* Advance write_seq and place onto the write_queue. */
	tp->write_seq = TCP_SKB_CB(skb)->end_seq;
	skb_header_release(skb);
635
	tcp_add_write_queue_tail(sk, skb);
L
Linus Torvalds 已提交
636 637 638
	sk_charge_skb(sk, skb);
}

639
static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb, unsigned int mss_now)
640
{
641
	if (skb->len <= mss_now || !sk_can_gso(sk)) {
642 643 644
		/* Avoid the costly divide in the normal
		 * non-TSO case.
		 */
645 646 647
		skb_shinfo(skb)->gso_segs = 1;
		skb_shinfo(skb)->gso_size = 0;
		skb_shinfo(skb)->gso_type = 0;
648
	} else {
I
Ilpo Järvinen 已提交
649
		skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss_now);
650
		skb_shinfo(skb)->gso_size = mss_now;
651
		skb_shinfo(skb)->gso_type = sk->sk_gso_type;
L
Linus Torvalds 已提交
652 653 654
	}
}

I
Ilpo Järvinen 已提交
655
/* When a modification to fackets out becomes necessary, we need to check
656
 * skb is counted to fackets_out or not.
I
Ilpo Järvinen 已提交
657
 */
658
static void tcp_adjust_fackets_out(struct sock *sk, struct sk_buff *skb,
I
Ilpo Järvinen 已提交
659 660
				   int decr)
{
661 662
	struct tcp_sock *tp = tcp_sk(sk);

663
	if (!tp->sacked_out || tcp_is_reno(tp))
I
Ilpo Järvinen 已提交
664 665
		return;

666
	if (!before(tcp_highest_sack_seq(tp), TCP_SKB_CB(skb)->seq))
I
Ilpo Järvinen 已提交
667 668 669
		tp->fackets_out -= decr;
}

L
Linus Torvalds 已提交
670 671
/* Function to create two new TCP segments.  Shrinks the given segment
 * to the specified size and appends a new segment with the rest of the
672
 * packet to the list.  This won't be called frequently, I hope.
L
Linus Torvalds 已提交
673 674
 * Remember, these are still headerless SKBs at this point.
 */
675
int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len, unsigned int mss_now)
L
Linus Torvalds 已提交
676 677 678
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *buff;
679
	int nsize, old_factor;
680
	int nlen;
L
Linus Torvalds 已提交
681 682
	u16 flags;

683
	BUG_ON(len > skb->len);
684

685
	tcp_clear_retrans_hints_partial(tp);
L
Linus Torvalds 已提交
686 687 688 689 690 691 692 693 694 695 696 697 698
	nsize = skb_headlen(skb) - len;
	if (nsize < 0)
		nsize = 0;

	if (skb_cloned(skb) &&
	    skb_is_nonlinear(skb) &&
	    pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
		return -ENOMEM;

	/* Get a new skb... force flag on. */
	buff = sk_stream_alloc_skb(sk, nsize, GFP_ATOMIC);
	if (buff == NULL)
		return -ENOMEM; /* We'll just try again later. */
H
Herbert Xu 已提交
699

700 701 702 703
	sk_charge_skb(sk, buff);
	nlen = skb->len - len - nsize;
	buff->truesize += nlen;
	skb->truesize -= nlen;
L
Linus Torvalds 已提交
704 705 706 707 708 709

	/* Correct the sequence numbers. */
	TCP_SKB_CB(buff)->seq = TCP_SKB_CB(skb)->seq + len;
	TCP_SKB_CB(buff)->end_seq = TCP_SKB_CB(skb)->end_seq;
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(buff)->seq;

710 711
	if (tcp_is_sack(tp) && tp->sacked_out && (skb == tp->highest_sack))
		tp->highest_sack = buff;
712

L
Linus Torvalds 已提交
713 714 715 716
	/* PSH and FIN should only be set in the second packet. */
	flags = TCP_SKB_CB(skb)->flags;
	TCP_SKB_CB(skb)->flags = flags & ~(TCPCB_FLAG_FIN|TCPCB_FLAG_PSH);
	TCP_SKB_CB(buff)->flags = flags;
717
	TCP_SKB_CB(buff)->sacked = TCP_SKB_CB(skb)->sacked;
L
Linus Torvalds 已提交
718 719
	TCP_SKB_CB(skb)->sacked &= ~TCPCB_AT_TAIL;

720
	if (!skb_shinfo(skb)->nr_frags && skb->ip_summed != CHECKSUM_PARTIAL) {
L
Linus Torvalds 已提交
721 722 723 724 725 726 727 728
		/* Copy and checksum data tail into the new buffer. */
		buff->csum = csum_partial_copy_nocheck(skb->data + len, skb_put(buff, nsize),
						       nsize, 0);

		skb_trim(skb, len);

		skb->csum = csum_block_sub(skb->csum, buff->csum, len);
	} else {
729
		skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
730 731 732 733 734 735 736 737 738
		skb_split(skb, buff, len);
	}

	buff->ip_summed = skb->ip_summed;

	/* Looks stupid, but our code really uses when of
	 * skbs, which it never sent before. --ANK
	 */
	TCP_SKB_CB(buff)->when = TCP_SKB_CB(skb)->when;
739
	buff->tstamp = skb->tstamp;
L
Linus Torvalds 已提交
740

741 742
	old_factor = tcp_skb_pcount(skb);

L
Linus Torvalds 已提交
743
	/* Fix up tso_factor for both original and new SKB.  */
744 745
	tcp_set_skb_tso_segs(sk, skb, mss_now);
	tcp_set_skb_tso_segs(sk, buff, mss_now);
L
Linus Torvalds 已提交
746

747 748 749
	/* If this packet has been sent out already, we must
	 * adjust the various packet counters.
	 */
750
	if (!before(tp->snd_nxt, TCP_SKB_CB(buff)->end_seq)) {
751 752
		int diff = old_factor - tcp_skb_pcount(skb) -
			tcp_skb_pcount(buff);
L
Linus Torvalds 已提交
753

754
		tp->packets_out -= diff;
755 756 757 758 759 760

		if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)
			tp->sacked_out -= diff;
		if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS)
			tp->retrans_out -= diff;

761
		if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST)
762
			tp->lost_out -= diff;
763

I
Ilpo Järvinen 已提交
764 765 766 767
		/* Adjust Reno SACK estimate. */
		if (tcp_is_reno(tp) && diff > 0) {
			tcp_dec_pcount_approx_int(&tp->sacked_out, diff);
			tcp_verify_left_out(tp);
768
		}
769
		tcp_adjust_fackets_out(sk, skb, diff);
L
Linus Torvalds 已提交
770 771 772
	}

	/* Link BUFF into the send queue. */
773
	skb_header_release(buff);
774
	tcp_insert_write_queue_after(skb, buff, sk);
L
Linus Torvalds 已提交
775 776 777 778 779 780 781 782

	return 0;
}

/* This is similar to __pskb_pull_head() (it will go to core/skbuff.c
 * eventually). The difference is that pulled data not copied, but
 * immediately discarded.
 */
783
static void __pskb_trim_head(struct sk_buff *skb, int len)
L
Linus Torvalds 已提交
784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804
{
	int i, k, eat;

	eat = len;
	k = 0;
	for (i=0; i<skb_shinfo(skb)->nr_frags; i++) {
		if (skb_shinfo(skb)->frags[i].size <= eat) {
			put_page(skb_shinfo(skb)->frags[i].page);
			eat -= skb_shinfo(skb)->frags[i].size;
		} else {
			skb_shinfo(skb)->frags[k] = skb_shinfo(skb)->frags[i];
			if (eat) {
				skb_shinfo(skb)->frags[k].page_offset += eat;
				skb_shinfo(skb)->frags[k].size -= eat;
				eat = 0;
			}
			k++;
		}
	}
	skb_shinfo(skb)->nr_frags = k;

805
	skb_reset_tail_pointer(skb);
L
Linus Torvalds 已提交
806 807 808 809 810 811 812 813 814 815
	skb->data_len -= len;
	skb->len = skb->data_len;
}

int tcp_trim_head(struct sock *sk, struct sk_buff *skb, u32 len)
{
	if (skb_cloned(skb) &&
	    pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
		return -ENOMEM;

816 817
	/* If len == headlen, we avoid __skb_pull to preserve alignment. */
	if (unlikely(len < skb_headlen(skb)))
L
Linus Torvalds 已提交
818
		__skb_pull(skb, len);
819 820
	else
		__pskb_trim_head(skb, len - skb_headlen(skb));
L
Linus Torvalds 已提交
821 822

	TCP_SKB_CB(skb)->seq += len;
823
	skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
824 825 826 827 828 829 830 831 832 833

	skb->truesize	     -= len;
	sk->sk_wmem_queued   -= len;
	sk->sk_forward_alloc += len;
	sock_set_flag(sk, SOCK_QUEUE_SHRUNK);

	/* Any change of skb->len requires recalculation of tso
	 * factor and mss.
	 */
	if (tcp_skb_pcount(skb) > 1)
834
		tcp_set_skb_tso_segs(sk, skb, tcp_current_mss(sk, 1));
L
Linus Torvalds 已提交
835 836 837 838

	return 0;
}

J
John Heffner 已提交
839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889
/* Not accounting for SACKs here. */
int tcp_mtu_to_mss(struct sock *sk, int pmtu)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);
	int mss_now;

	/* Calculate base mss without TCP options:
	   It is MMS_S - sizeof(tcphdr) of rfc1122
	 */
	mss_now = pmtu - icsk->icsk_af_ops->net_header_len - sizeof(struct tcphdr);

	/* Clamp it (mss_clamp does not include tcp options) */
	if (mss_now > tp->rx_opt.mss_clamp)
		mss_now = tp->rx_opt.mss_clamp;

	/* Now subtract optional transport overhead */
	mss_now -= icsk->icsk_ext_hdr_len;

	/* Then reserve room for full set of TCP options and 8 bytes of data */
	if (mss_now < 48)
		mss_now = 48;

	/* Now subtract TCP options size, not including SACKs */
	mss_now -= tp->tcp_header_len - sizeof(struct tcphdr);

	return mss_now;
}

/* Inverse of above */
int tcp_mss_to_mtu(struct sock *sk, int mss)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);
	int mtu;

	mtu = mss +
	      tp->tcp_header_len +
	      icsk->icsk_ext_hdr_len +
	      icsk->icsk_af_ops->net_header_len;

	return mtu;
}

void tcp_mtup_init(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);

	icsk->icsk_mtup.enabled = sysctl_tcp_mtu_probing > 1;
	icsk->icsk_mtup.search_high = tp->rx_opt.mss_clamp + sizeof(struct tcphdr) +
890
			       icsk->icsk_af_ops->net_header_len;
J
John Heffner 已提交
891 892 893 894
	icsk->icsk_mtup.search_low = tcp_mss_to_mtu(sk, sysctl_tcp_base_mss);
	icsk->icsk_mtup.probe_size = 0;
}

L
Linus Torvalds 已提交
895 896 897 898 899 900
/* This function synchronize snd mss to current pmtu/exthdr set.

   tp->rx_opt.user_mss is mss set by user by TCP_MAXSEG. It does NOT counts
   for TCP options, but includes only bare TCP header.

   tp->rx_opt.mss_clamp is mss negotiated at connection setup.
S
Stephen Hemminger 已提交
901
   It is minimum of user_mss and mss received with SYN.
L
Linus Torvalds 已提交
902 903
   It also does not include TCP options.

904
   inet_csk(sk)->icsk_pmtu_cookie is last pmtu, seen by this function.
L
Linus Torvalds 已提交
905 906 907 908 909 910 911 912 913

   tp->mss_cache is current effective sending mss, including
   all tcp options except for SACKs. It is evaluated,
   taking into account current pmtu, but never exceeds
   tp->rx_opt.mss_clamp.

   NOTE1. rfc1122 clearly states that advertised MSS
   DOES NOT include either tcp or ip options.

914 915
   NOTE2. inet_csk(sk)->icsk_pmtu_cookie and tp->mss_cache
   are READ ONLY outside this function.		--ANK (980731)
L
Linus Torvalds 已提交
916 917 918 919 920
 */

unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu)
{
	struct tcp_sock *tp = tcp_sk(sk);
921
	struct inet_connection_sock *icsk = inet_csk(sk);
J
John Heffner 已提交
922
	int mss_now;
L
Linus Torvalds 已提交
923

J
John Heffner 已提交
924 925
	if (icsk->icsk_mtup.search_high > pmtu)
		icsk->icsk_mtup.search_high = pmtu;
L
Linus Torvalds 已提交
926

J
John Heffner 已提交
927
	mss_now = tcp_mtu_to_mss(sk, pmtu);
L
Linus Torvalds 已提交
928 929 930 931 932 933

	/* Bound mss with half of window */
	if (tp->max_window && mss_now > (tp->max_window>>1))
		mss_now = max((tp->max_window>>1), 68U - tp->tcp_header_len);

	/* And store cached results */
934
	icsk->icsk_pmtu_cookie = pmtu;
J
John Heffner 已提交
935 936
	if (icsk->icsk_mtup.enabled)
		mss_now = min(mss_now, tcp_mtu_to_mss(sk, icsk->icsk_mtup.search_low));
937
	tp->mss_cache = mss_now;
L
Linus Torvalds 已提交
938 939 940 941 942 943 944 945 946 947 948

	return mss_now;
}

/* Compute the current effective MSS, taking SACKs and IP options,
 * and even PMTU discovery events into account.
 *
 * LARGESEND note: !urg_mode is overkill, only frames up to snd_up
 * cannot be large. However, taking into account rare use of URG, this
 * is not a big flaw.
 */
949
unsigned int tcp_current_mss(struct sock *sk, int large_allowed)
L
Linus Torvalds 已提交
950 951 952
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct dst_entry *dst = __sk_dst_get(sk);
953 954 955 956 957 958
	u32 mss_now;
	u16 xmit_size_goal;
	int doing_tso = 0;

	mss_now = tp->mss_cache;

959
	if (large_allowed && sk_can_gso(sk) && !tp->urg_mode)
960
		doing_tso = 1;
L
Linus Torvalds 已提交
961 962 963

	if (dst) {
		u32 mtu = dst_mtu(dst);
964
		if (mtu != inet_csk(sk)->icsk_pmtu_cookie)
L
Linus Torvalds 已提交
965 966 967
			mss_now = tcp_sync_mss(sk, mtu);
	}

968 969 970
	if (tp->rx_opt.eff_sacks)
		mss_now -= (TCPOLEN_SACK_BASE_ALIGNED +
			    (tp->rx_opt.eff_sacks * TCPOLEN_SACK_PERBLOCK));
L
Linus Torvalds 已提交
971

972 973 974 975 976
#ifdef CONFIG_TCP_MD5SIG
	if (tp->af_specific->md5_lookup(sk, sk))
		mss_now -= TCPOLEN_MD5SIG_ALIGNED;
#endif

977
	xmit_size_goal = mss_now;
L
Linus Torvalds 已提交
978

979
	if (doing_tso) {
980 981
		xmit_size_goal = (65535 -
				  inet_csk(sk)->icsk_af_ops->net_header_len -
982 983
				  inet_csk(sk)->icsk_ext_hdr_len -
				  tp->tcp_header_len);
L
Linus Torvalds 已提交
984

985 986 987 988
		if (tp->max_window &&
		    (xmit_size_goal > (tp->max_window >> 1)))
			xmit_size_goal = max((tp->max_window >> 1),
					     68U - tp->tcp_header_len);
L
Linus Torvalds 已提交
989

990
		xmit_size_goal -= (xmit_size_goal % mss_now);
L
Linus Torvalds 已提交
991
	}
992
	tp->xmit_size_goal = xmit_size_goal;
L
Linus Torvalds 已提交
993 994 995 996

	return mss_now;
}

997 998
/* Congestion window validation. (RFC2861) */

999
static void tcp_cwnd_validate(struct sock *sk)
1000
{
1001
	struct tcp_sock *tp = tcp_sk(sk);
1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012
	__u32 packets_out = tp->packets_out;

	if (packets_out >= tp->snd_cwnd) {
		/* Network is feed fully. */
		tp->snd_cwnd_used = 0;
		tp->snd_cwnd_stamp = tcp_time_stamp;
	} else {
		/* Network starves. */
		if (tp->packets_out > tp->snd_cwnd_used)
			tp->snd_cwnd_used = tp->packets_out;

1013 1014
		if (sysctl_tcp_slow_start_after_idle &&
		    (s32)(tcp_time_stamp - tp->snd_cwnd_stamp) >= inet_csk(sk)->icsk_rto)
1015 1016 1017 1018
			tcp_cwnd_application_limited(sk);
	}
}

1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
static unsigned int tcp_window_allows(struct tcp_sock *tp, struct sk_buff *skb, unsigned int mss_now, unsigned int cwnd)
{
	u32 window, cwnd_len;

	window = (tp->snd_una + tp->snd_wnd - TCP_SKB_CB(skb)->seq);
	cwnd_len = mss_now * cwnd;
	return min(window, cwnd_len);
}

/* Can at least one segment of SKB be sent right now, according to the
 * congestion window rules?  If so, return how many segments are allowed.
 */
static inline unsigned int tcp_cwnd_test(struct tcp_sock *tp, struct sk_buff *skb)
{
	u32 in_flight, cwnd;

	/* Don't be strict about the congestion window for the final FIN.  */
1036 1037
	if ((TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN) &&
	    tcp_skb_pcount(skb) == 1)
1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
		return 1;

	in_flight = tcp_packets_in_flight(tp);
	cwnd = tp->snd_cwnd;
	if (in_flight < cwnd)
		return (cwnd - in_flight);

	return 0;
}

/* This must be invoked the first time we consider transmitting
 * SKB onto the wire.
 */
S
Stephen Hemminger 已提交
1051
static int tcp_init_tso_segs(struct sock *sk, struct sk_buff *skb, unsigned int mss_now)
1052 1053 1054
{
	int tso_segs = tcp_skb_pcount(skb);

1055 1056
	if (!tso_segs ||
	    (tso_segs > 1 &&
1057
	     tcp_skb_mss(skb) != mss_now)) {
1058
		tcp_set_skb_tso_segs(sk, skb, mss_now);
1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
		tso_segs = tcp_skb_pcount(skb);
	}
	return tso_segs;
}

static inline int tcp_minshall_check(const struct tcp_sock *tp)
{
	return after(tp->snd_sml,tp->snd_una) &&
		!after(tp->snd_sml, tp->snd_nxt);
}

/* Return 0, if packet can be sent now without violation Nagle's rules:
 * 1. It is full sized.
 * 2. Or it contains FIN. (already checked by caller)
 * 3. Or TCP_NODELAY was set.
 * 4. Or TCP_CORK is not set, and all sent packets are ACKed.
 *    With Minshall's modification: all sent small packets are ACKed.
 */

static inline int tcp_nagle_check(const struct tcp_sock *tp,
1079
				  const struct sk_buff *skb,
1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103
				  unsigned mss_now, int nonagle)
{
	return (skb->len < mss_now &&
		((nonagle&TCP_NAGLE_CORK) ||
		 (!nonagle &&
		  tp->packets_out &&
		  tcp_minshall_check(tp))));
}

/* Return non-zero if the Nagle test allows this packet to be
 * sent now.
 */
static inline int tcp_nagle_test(struct tcp_sock *tp, struct sk_buff *skb,
				 unsigned int cur_mss, int nonagle)
{
	/* Nagle rule does not apply to frames, which sit in the middle of the
	 * write_queue (they have no chances to get new data).
	 *
	 * This is implemented in the callers, where they modify the 'nonagle'
	 * argument based upon the location of SKB in the send queue.
	 */
	if (nonagle & TCP_NAGLE_PUSH)
		return 1;

1104 1105 1106 1107
	/* Don't use the nagle rule for urgent data (or for the final FIN).
	 * Nagle can be ignored during F-RTO too (see RFC4138).
	 */
	if (tp->urg_mode || (tp->frto_counter == 2) ||
1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127
	    (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN))
		return 1;

	if (!tcp_nagle_check(tp, skb, cur_mss, nonagle))
		return 1;

	return 0;
}

/* Does at least the first segment of SKB fit into the send window? */
static inline int tcp_snd_wnd_test(struct tcp_sock *tp, struct sk_buff *skb, unsigned int cur_mss)
{
	u32 end_seq = TCP_SKB_CB(skb)->end_seq;

	if (skb->len > cur_mss)
		end_seq = TCP_SKB_CB(skb)->seq + cur_mss;

	return !after(end_seq, tp->snd_una + tp->snd_wnd);
}

1128
/* This checks if the data bearing packet SKB (usually tcp_send_head(sk))
1129 1130 1131 1132 1133 1134 1135 1136 1137
 * should be put on the wire right now.  If so, it returns the number of
 * packets allowed by the congestion window.
 */
static unsigned int tcp_snd_test(struct sock *sk, struct sk_buff *skb,
				 unsigned int cur_mss, int nonagle)
{
	struct tcp_sock *tp = tcp_sk(sk);
	unsigned int cwnd_quota;

1138
	tcp_init_tso_segs(sk, skb, cur_mss);
1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150

	if (!tcp_nagle_test(tp, skb, cur_mss, nonagle))
		return 0;

	cwnd_quota = tcp_cwnd_test(tp, skb);
	if (cwnd_quota &&
	    !tcp_snd_wnd_test(tp, skb, cur_mss))
		cwnd_quota = 0;

	return cwnd_quota;
}

1151
int tcp_may_send_now(struct sock *sk)
1152
{
1153
	struct tcp_sock *tp = tcp_sk(sk);
1154
	struct sk_buff *skb = tcp_send_head(sk);
1155 1156 1157 1158

	return (skb &&
		tcp_snd_test(sk, skb, tcp_current_mss(sk, 1),
			     (tcp_skb_is_last(sk, skb) ?
1159
			      tp->nonagle : TCP_NAGLE_PUSH)));
1160 1161 1162 1163 1164 1165 1166 1167 1168
}

/* Trim TSO SKB to LEN bytes, put the remaining data into a new packet
 * which is put after SKB on the list.  It is very much like
 * tcp_fragment() except that it may make several kinds of assumptions
 * in order to speed up the splitting operation.  In particular, we
 * know that all the data is in scatter-gather pages, and that the
 * packet has never been sent out before (and thus is not cloned).
 */
1169
static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len, unsigned int mss_now)
1170 1171 1172 1173 1174 1175
{
	struct sk_buff *buff;
	int nlen = skb->len - len;
	u16 flags;

	/* All of a TSO frame must be composed of paged data.  */
1176 1177
	if (skb->len != skb->data_len)
		return tcp_fragment(sk, skb, len, mss_now);
1178 1179 1180 1181 1182

	buff = sk_stream_alloc_pskb(sk, 0, 0, GFP_ATOMIC);
	if (unlikely(buff == NULL))
		return -ENOMEM;

1183 1184
	sk_charge_skb(sk, buff);
	buff->truesize += nlen;
1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
	skb->truesize -= nlen;

	/* Correct the sequence numbers. */
	TCP_SKB_CB(buff)->seq = TCP_SKB_CB(skb)->seq + len;
	TCP_SKB_CB(buff)->end_seq = TCP_SKB_CB(skb)->end_seq;
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(buff)->seq;

	/* PSH and FIN should only be set in the second packet. */
	flags = TCP_SKB_CB(skb)->flags;
	TCP_SKB_CB(skb)->flags = flags & ~(TCPCB_FLAG_FIN|TCPCB_FLAG_PSH);
	TCP_SKB_CB(buff)->flags = flags;

	/* This packet was never sent out yet, so no SACK bits. */
	TCP_SKB_CB(buff)->sacked = 0;

1200
	buff->ip_summed = skb->ip_summed = CHECKSUM_PARTIAL;
1201 1202 1203
	skb_split(skb, buff, len);

	/* Fix up tso_factor for both original and new SKB.  */
1204 1205
	tcp_set_skb_tso_segs(sk, skb, mss_now);
	tcp_set_skb_tso_segs(sk, buff, mss_now);
1206 1207 1208

	/* Link BUFF into the send queue. */
	skb_header_release(buff);
1209
	tcp_insert_write_queue_after(skb, buff, sk);
1210 1211 1212 1213 1214 1215 1216 1217 1218

	return 0;
}

/* Try to defer sending, if possible, in order to minimize the amount
 * of TSO splitting we do.  View it as a kind of TSO Nagle test.
 *
 * This algorithm is from John Heffner.
 */
1219
static int tcp_tso_should_defer(struct sock *sk, struct sk_buff *skb)
1220
{
1221
	struct tcp_sock *tp = tcp_sk(sk);
1222
	const struct inet_connection_sock *icsk = inet_csk(sk);
1223 1224 1225
	u32 send_win, cong_win, limit, in_flight;

	if (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN)
J
John Heffner 已提交
1226
		goto send_now;
1227

1228
	if (icsk->icsk_ca_state != TCP_CA_Open)
J
John Heffner 已提交
1229 1230 1231 1232 1233
		goto send_now;

	/* Defer for less than two clock ticks. */
	if (!tp->tso_deferred && ((jiffies<<1)>>1) - (tp->tso_deferred>>1) > 1)
		goto send_now;
1234

1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246
	in_flight = tcp_packets_in_flight(tp);

	BUG_ON(tcp_skb_pcount(skb) <= 1 ||
	       (tp->snd_cwnd <= in_flight));

	send_win = (tp->snd_una + tp->snd_wnd) - TCP_SKB_CB(skb)->seq;

	/* From in_flight test above, we know that cwnd > in_flight.  */
	cong_win = (tp->snd_cwnd - in_flight) * tp->mss_cache;

	limit = min(send_win, cong_win);

1247 1248
	/* If a full-sized TSO skb can be sent, do it. */
	if (limit >= 65536)
J
John Heffner 已提交
1249
		goto send_now;
1250

1251 1252 1253 1254 1255 1256 1257 1258
	if (sysctl_tcp_tso_win_divisor) {
		u32 chunk = min(tp->snd_wnd, tp->snd_cwnd * tp->mss_cache);

		/* If at least some fraction of a window is available,
		 * just use it.
		 */
		chunk /= sysctl_tcp_tso_win_divisor;
		if (limit >= chunk)
J
John Heffner 已提交
1259
			goto send_now;
1260 1261 1262 1263 1264 1265 1266
	} else {
		/* Different approach, try not to defer past a single
		 * ACK.  Receiver should ACK every other full sized
		 * frame, so if we have space for more than 3 frames
		 * then send now.
		 */
		if (limit > tcp_max_burst(tp) * tp->mss_cache)
J
John Heffner 已提交
1267
			goto send_now;
1268 1269 1270
	}

	/* Ok, it looks like it is advisable to defer.  */
J
John Heffner 已提交
1271 1272
	tp->tso_deferred = 1 | (jiffies<<1);

1273
	return 1;
J
John Heffner 已提交
1274 1275 1276 1277

send_now:
	tp->tso_deferred = 0;
	return 0;
1278 1279
}

J
John Heffner 已提交
1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
/* Create a new MTU probe if we are ready.
 * Returns 0 if we should wait to probe (no cwnd available),
 *         1 if a probe was sent,
 *         -1 otherwise */
static int tcp_mtu_probe(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);
	struct sk_buff *skb, *nskb, *next;
	int len;
	int probe_size;
1291
	int size_needed;
J
John Heffner 已提交
1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309
	unsigned int pif;
	int copy;
	int mss_now;

	/* Not currently probing/verifying,
	 * not in recovery,
	 * have enough cwnd, and
	 * not SACKing (the variable headers throw things off) */
	if (!icsk->icsk_mtup.enabled ||
	    icsk->icsk_mtup.probe_size ||
	    inet_csk(sk)->icsk_ca_state != TCP_CA_Open ||
	    tp->snd_cwnd < 11 ||
	    tp->rx_opt.eff_sacks)
		return -1;

	/* Very simple search strategy: just double the MSS. */
	mss_now = tcp_current_mss(sk, 0);
	probe_size = 2*tp->mss_cache;
1310
	size_needed = probe_size + (tp->reordering + 1) * tp->mss_cache;
J
John Heffner 已提交
1311 1312 1313 1314 1315 1316
	if (probe_size > tcp_mtu_to_mss(sk, icsk->icsk_mtup.search_high)) {
		/* TODO: set timer for probe_converge_event */
		return -1;
	}

	/* Have enough data in the send queue to probe? */
1317
	if (tp->write_seq - tp->snd_nxt < size_needed)
J
John Heffner 已提交
1318 1319
		return -1;

1320 1321 1322 1323
	if (tp->snd_wnd < size_needed)
		return -1;
	if (after(tp->snd_nxt + size_needed, tp->snd_una + tp->snd_wnd))
		return 0;
J
John Heffner 已提交
1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339

	/* Do we need to wait to drain cwnd? */
	pif = tcp_packets_in_flight(tp);
	if (pif + 2 > tp->snd_cwnd) {
		/* With no packets in flight, don't stall. */
		if (pif == 0)
			return -1;
		else
			return 0;
	}

	/* We're allowed to probe.  Build it now. */
	if ((nskb = sk_stream_alloc_skb(sk, probe_size, GFP_ATOMIC)) == NULL)
		return -1;
	sk_charge_skb(sk, nskb);

1340 1341
	skb = tcp_send_head(sk);
	tcp_insert_write_queue_before(nskb, skb, sk);
J
John Heffner 已提交
1342 1343 1344 1345 1346 1347

	TCP_SKB_CB(nskb)->seq = TCP_SKB_CB(skb)->seq;
	TCP_SKB_CB(nskb)->end_seq = TCP_SKB_CB(skb)->seq + probe_size;
	TCP_SKB_CB(nskb)->flags = TCPCB_FLAG_ACK;
	TCP_SKB_CB(nskb)->sacked = 0;
	nskb->csum = 0;
1348
	nskb->ip_summed = skb->ip_summed;
J
John Heffner 已提交
1349 1350 1351

	len = 0;
	while (len < probe_size) {
1352
		next = tcp_write_queue_next(sk, skb);
J
John Heffner 已提交
1353 1354 1355 1356 1357 1358

		copy = min_t(int, skb->len, probe_size - len);
		if (nskb->ip_summed)
			skb_copy_bits(skb, 0, skb_put(nskb, copy), copy);
		else
			nskb->csum = skb_copy_and_csum_bits(skb, 0,
1359
					 skb_put(nskb, copy), copy, nskb->csum);
J
John Heffner 已提交
1360 1361 1362 1363 1364

		if (skb->len <= copy) {
			/* We've eaten all the data from this skb.
			 * Throw it away. */
			TCP_SKB_CB(nskb)->flags |= TCP_SKB_CB(skb)->flags;
1365
			tcp_unlink_write_queue(skb, sk);
J
John Heffner 已提交
1366 1367 1368
			sk_stream_free_skb(sk, skb);
		} else {
			TCP_SKB_CB(nskb)->flags |= TCP_SKB_CB(skb)->flags &
1369
						   ~(TCPCB_FLAG_FIN|TCPCB_FLAG_PSH);
J
John Heffner 已提交
1370 1371
			if (!skb_shinfo(skb)->nr_frags) {
				skb_pull(skb, copy);
1372
				if (skb->ip_summed != CHECKSUM_PARTIAL)
J
John Heffner 已提交
1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392
					skb->csum = csum_partial(skb->data, skb->len, 0);
			} else {
				__pskb_trim_head(skb, copy);
				tcp_set_skb_tso_segs(sk, skb, mss_now);
			}
			TCP_SKB_CB(skb)->seq += copy;
		}

		len += copy;
		skb = next;
	}
	tcp_init_tso_segs(sk, nskb, nskb->len);

	/* We're ready to send.  If this fails, the probe will
	 * be resegmented into mss-sized pieces by tcp_write_xmit(). */
	TCP_SKB_CB(nskb)->when = tcp_time_stamp;
	if (!tcp_transmit_skb(sk, nskb, 1, GFP_ATOMIC)) {
		/* Decrement cwnd here because we are sending
		* effectively two packets. */
		tp->snd_cwnd--;
1393
		update_send_head(sk, nskb);
J
John Heffner 已提交
1394 1395

		icsk->icsk_mtup.probe_size = tcp_mss_to_mtu(sk, nskb->len);
1396 1397
		tp->mtu_probe.probe_seq_start = TCP_SKB_CB(nskb)->seq;
		tp->mtu_probe.probe_seq_end = TCP_SKB_CB(nskb)->end_seq;
J
John Heffner 已提交
1398 1399 1400 1401 1402 1403 1404 1405

		return 1;
	}

	return -1;
}


L
Linus Torvalds 已提交
1406 1407 1408 1409 1410 1411 1412
/* This routine writes packets to the network.  It advances the
 * send_head.  This happens as incoming acks open up the remote
 * window for us.
 *
 * Returns 1, if no segments are in flight and we have queued segments, but
 * cannot send anything now because of SWS or another problem.
 */
1413
static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle)
L
Linus Torvalds 已提交
1414 1415
{
	struct tcp_sock *tp = tcp_sk(sk);
1416
	struct sk_buff *skb;
1417 1418
	unsigned int tso_segs, sent_pkts;
	int cwnd_quota;
J
John Heffner 已提交
1419
	int result;
L
Linus Torvalds 已提交
1420 1421 1422 1423 1424

	/* If we are closed, the bytes will have to remain here.
	 * In time closedown will finish, we empty the write queue and all
	 * will be happy.
	 */
1425 1426
	if (unlikely(sk->sk_state == TCP_CLOSE))
		return 0;
L
Linus Torvalds 已提交
1427

1428
	sent_pkts = 0;
J
John Heffner 已提交
1429 1430 1431 1432 1433 1434 1435 1436

	/* Do MTU probing. */
	if ((result = tcp_mtu_probe(sk)) == 0) {
		return 0;
	} else if (result > 0) {
		sent_pkts = 1;
	}

1437
	while ((skb = tcp_send_head(sk))) {
1438 1439
		unsigned int limit;

1440
		tso_segs = tcp_init_tso_segs(sk, skb, mss_now);
1441
		BUG_ON(!tso_segs);
1442

1443 1444 1445 1446 1447 1448 1449
		cwnd_quota = tcp_cwnd_test(tp, skb);
		if (!cwnd_quota)
			break;

		if (unlikely(!tcp_snd_wnd_test(tp, skb, mss_now)))
			break;

1450 1451 1452 1453 1454 1455
		if (tso_segs == 1) {
			if (unlikely(!tcp_nagle_test(tp, skb, mss_now,
						     (tcp_skb_is_last(sk, skb) ?
						      nonagle : TCP_NAGLE_PUSH))))
				break;
		} else {
1456
			if (tcp_tso_should_defer(sk, skb))
1457 1458
				break;
		}
1459

1460
		limit = mss_now;
1461
		if (tso_segs > 1) {
1462 1463
			limit = tcp_window_allows(tp, skb,
						  mss_now, cwnd_quota);
1464 1465 1466

			if (skb->len < limit) {
				unsigned int trim = skb->len % mss_now;
1467

1468 1469 1470
				if (trim)
					limit = skb->len - trim;
			}
1471
		}
L
Linus Torvalds 已提交
1472

1473 1474 1475 1476
		if (skb->len > limit &&
		    unlikely(tso_fragment(sk, skb, limit, mss_now)))
			break;

1477
		TCP_SKB_CB(skb)->when = tcp_time_stamp;
1478

1479
		if (unlikely(tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC)))
1480
			break;
L
Linus Torvalds 已提交
1481

1482 1483 1484
		/* Advance the send_head.  This one is sent out.
		 * This call will increment packets_out.
		 */
1485
		update_send_head(sk, skb);
L
Linus Torvalds 已提交
1486

1487
		tcp_minshall_update(tp, mss_now, skb);
1488
		sent_pkts++;
1489
	}
L
Linus Torvalds 已提交
1490

1491
	if (likely(sent_pkts)) {
1492
		tcp_cwnd_validate(sk);
1493
		return 0;
L
Linus Torvalds 已提交
1494
	}
1495
	return !tp->packets_out && tcp_send_head(sk);
L
Linus Torvalds 已提交
1496 1497
}

1498 1499 1500 1501
/* Push out any pending frames which were held back due to
 * TCP_CORK or attempt at coalescing tiny packets.
 * The socket must be locked by the caller.
 */
1502 1503
void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
			       int nonagle)
1504
{
1505
	struct sk_buff *skb = tcp_send_head(sk);
1506 1507

	if (skb) {
1508
		if (tcp_write_xmit(sk, cur_mss, nonagle))
1509
			tcp_check_probe_timer(sk);
1510 1511 1512
	}
}

1513 1514 1515 1516 1517 1518
/* Send _single_ skb sitting at the send head. This function requires
 * true push pending frames to setup probe timer etc.
 */
void tcp_push_one(struct sock *sk, unsigned int mss_now)
{
	struct tcp_sock *tp = tcp_sk(sk);
1519
	struct sk_buff *skb = tcp_send_head(sk);
1520 1521 1522 1523
	unsigned int tso_segs, cwnd_quota;

	BUG_ON(!skb || skb->len < mss_now);

1524
	tso_segs = tcp_init_tso_segs(sk, skb, mss_now);
1525 1526 1527
	cwnd_quota = tcp_snd_test(sk, skb, mss_now, TCP_NAGLE_PUSH);

	if (likely(cwnd_quota)) {
1528 1529
		unsigned int limit;

1530 1531
		BUG_ON(!tso_segs);

1532
		limit = mss_now;
1533
		if (tso_segs > 1) {
1534 1535
			limit = tcp_window_allows(tp, skb,
						  mss_now, cwnd_quota);
1536 1537 1538 1539 1540 1541 1542 1543 1544

			if (skb->len < limit) {
				unsigned int trim = skb->len % mss_now;

				if (trim)
					limit = skb->len - trim;
			}
		}

1545 1546 1547 1548
		if (skb->len > limit &&
		    unlikely(tso_fragment(sk, skb, limit, mss_now)))
			return;

1549 1550 1551
		/* Send it out now. */
		TCP_SKB_CB(skb)->when = tcp_time_stamp;

1552
		if (likely(!tcp_transmit_skb(sk, skb, 1, sk->sk_allocation))) {
1553 1554
			update_send_head(sk, skb);
			tcp_cwnd_validate(sk);
1555 1556 1557 1558 1559
			return;
		}
	}
}

L
Linus Torvalds 已提交
1560 1561
/* This function returns the amount that we can raise the
 * usable window based on the following constraints
1562
 *
L
Linus Torvalds 已提交
1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580
 * 1. The window can never be shrunk once it is offered (RFC 793)
 * 2. We limit memory per socket
 *
 * RFC 1122:
 * "the suggested [SWS] avoidance algorithm for the receiver is to keep
 *  RECV.NEXT + RCV.WIN fixed until:
 *  RCV.BUFF - RCV.USER - RCV.WINDOW >= min(1/2 RCV.BUFF, MSS)"
 *
 * i.e. don't raise the right edge of the window until you can raise
 * it at least MSS bytes.
 *
 * Unfortunately, the recommended algorithm breaks header prediction,
 * since header prediction assumes th->window stays fixed.
 *
 * Strictly speaking, keeping th->window fixed violates the receiver
 * side SWS prevention criteria. The problem is that under this rule
 * a stream of single byte packets will cause the right side of the
 * window to always advance by a single byte.
1581
 *
L
Linus Torvalds 已提交
1582 1583
 * Of course, if the sender implements sender side SWS prevention
 * then this will not be a problem.
1584
 *
L
Linus Torvalds 已提交
1585
 * BSD seems to make the following compromise:
1586
 *
L
Linus Torvalds 已提交
1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
 *	If the free space is less than the 1/4 of the maximum
 *	space available and the free space is less than 1/2 mss,
 *	then set the window to 0.
 *	[ Actually, bsd uses MSS and 1/4 of maximal _window_ ]
 *	Otherwise, just prevent the window from shrinking
 *	and from being larger than the largest representable value.
 *
 * This prevents incremental opening of the window in the regime
 * where TCP is limited by the speed of the reader side taking
 * data out of the TCP receive queue. It does nothing about
 * those cases where the window is constrained on the sender side
 * because the pipeline is full.
 *
 * BSD also seems to "accidentally" limit itself to windows that are a
 * multiple of MSS, at least until the free space gets quite small.
 * This would appear to be a side effect of the mbuf implementation.
 * Combining these two algorithms results in the observed behavior
 * of having a fixed window size at almost all times.
 *
 * Below we obtain similar behavior by forcing the offered window to
 * a multiple of the mss when it is feasible to do so.
 *
 * Note, we don't "adjust" for TIMESTAMP or SACK option bytes.
 * Regular options like TIMESTAMP are taken into account.
 */
u32 __tcp_select_window(struct sock *sk)
{
1614
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1615
	struct tcp_sock *tp = tcp_sk(sk);
S
Stephen Hemminger 已提交
1616
	/* MSS for the peer's data.  Previous versions used mss_clamp
L
Linus Torvalds 已提交
1617 1618 1619 1620 1621
	 * here.  I don't know if the value based on our guesses
	 * of peer's MSS is better for the performance.  It's more correct
	 * but may be worse for the performance because of rcv_mss
	 * fluctuations.  --SAW  1998/11/1
	 */
1622
	int mss = icsk->icsk_ack.rcv_mss;
L
Linus Torvalds 已提交
1623 1624 1625 1626 1627
	int free_space = tcp_space(sk);
	int full_space = min_t(int, tp->window_clamp, tcp_full_space(sk));
	int window;

	if (mss > full_space)
1628
		mss = full_space;
L
Linus Torvalds 已提交
1629 1630

	if (free_space < full_space/2) {
1631
		icsk->icsk_ack.quick = 0;
L
Linus Torvalds 已提交
1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667

		if (tcp_memory_pressure)
			tp->rcv_ssthresh = min(tp->rcv_ssthresh, 4U*tp->advmss);

		if (free_space < mss)
			return 0;
	}

	if (free_space > tp->rcv_ssthresh)
		free_space = tp->rcv_ssthresh;

	/* Don't do rounding if we are using window scaling, since the
	 * scaled window will not line up with the MSS boundary anyway.
	 */
	window = tp->rcv_wnd;
	if (tp->rx_opt.rcv_wscale) {
		window = free_space;

		/* Advertise enough space so that it won't get scaled away.
		 * Import case: prevent zero window announcement if
		 * 1<<rcv_wscale > mss.
		 */
		if (((window >> tp->rx_opt.rcv_wscale) << tp->rx_opt.rcv_wscale) != window)
			window = (((window >> tp->rx_opt.rcv_wscale) + 1)
				  << tp->rx_opt.rcv_wscale);
	} else {
		/* Get the largest window that is a nice multiple of mss.
		 * Window clamp already applied above.
		 * If our current window offering is within 1 mss of the
		 * free space we just keep it. This prevents the divide
		 * and multiply from happening most of the time.
		 * We also don't do any window rounding when the free space
		 * is too small.
		 */
		if (window <= free_space - mss || window > free_space)
			window = (free_space/mss)*mss;
1668
		else if (mss == full_space &&
1669
			 free_space > window + full_space/2)
1670
			window = free_space;
L
Linus Torvalds 已提交
1671 1672 1673 1674 1675 1676 1677 1678 1679
	}

	return window;
}

/* Attempt to collapse two adjacent SKB's during retransmission. */
static void tcp_retrans_try_collapse(struct sock *sk, struct sk_buff *skb, int mss_now)
{
	struct tcp_sock *tp = tcp_sk(sk);
1680
	struct sk_buff *next_skb = tcp_write_queue_next(sk, skb);
L
Linus Torvalds 已提交
1681 1682 1683 1684 1685 1686 1687 1688 1689

	/* The first test we must make is that neither of these two
	 * SKB's are still referenced by someone else.
	 */
	if (!skb_cloned(skb) && !skb_cloned(next_skb)) {
		int skb_size = skb->len, next_skb_size = next_skb->len;
		u16 flags = TCP_SKB_CB(skb)->flags;

		/* Also punt if next skb has been SACK'd. */
S
Stephen Hemminger 已提交
1690
		if (TCP_SKB_CB(next_skb)->sacked & TCPCB_SACKED_ACKED)
L
Linus Torvalds 已提交
1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707
			return;

		/* Next skb is out of window. */
		if (after(TCP_SKB_CB(next_skb)->end_seq, tp->snd_una+tp->snd_wnd))
			return;

		/* Punt if not enough space exists in the first SKB for
		 * the data in the second, or the total combined payload
		 * would exceed the MSS.
		 */
		if ((next_skb_size > skb_tailroom(skb)) ||
		    ((skb_size + next_skb_size) > mss_now))
			return;

		BUG_ON(tcp_skb_pcount(skb) != 1 ||
		       tcp_skb_pcount(next_skb) != 1);

1708
		if (WARN_ON(tcp_is_sack(tp) && tp->sacked_out &&
1709
		    (next_skb == tp->highest_sack)))
1710 1711
			return;

1712
		/* Ok.	We will be able to collapse the packet. */
1713
		tcp_unlink_write_queue(next_skb, sk);
L
Linus Torvalds 已提交
1714

1715 1716 1717
		skb_copy_from_linear_data(next_skb,
					  skb_put(skb, next_skb_size),
					  next_skb_size);
L
Linus Torvalds 已提交
1718

1719 1720
		if (next_skb->ip_summed == CHECKSUM_PARTIAL)
			skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
1721

1722
		if (skb->ip_summed != CHECKSUM_PARTIAL)
L
Linus Torvalds 已提交
1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737
			skb->csum = csum_block_add(skb->csum, next_skb->csum, skb_size);

		/* Update sequence range on original skb. */
		TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(next_skb)->end_seq;

		/* Merge over control information. */
		flags |= TCP_SKB_CB(next_skb)->flags; /* This moves PSH/FIN etc. over */
		TCP_SKB_CB(skb)->flags = flags;

		/* All done, get rid of second SKB and account for it so
		 * packet counting does not break.
		 */
		TCP_SKB_CB(skb)->sacked |= TCP_SKB_CB(next_skb)->sacked&(TCPCB_EVER_RETRANS|TCPCB_AT_TAIL);
		if (TCP_SKB_CB(next_skb)->sacked&TCPCB_SACKED_RETRANS)
			tp->retrans_out -= tcp_skb_pcount(next_skb);
1738
		if (TCP_SKB_CB(next_skb)->sacked&TCPCB_LOST)
L
Linus Torvalds 已提交
1739 1740
			tp->lost_out -= tcp_skb_pcount(next_skb);
		/* Reno case is special. Sigh... */
1741
		if (tcp_is_reno(tp) && tp->sacked_out)
L
Linus Torvalds 已提交
1742 1743
			tcp_dec_pcount_approx(&tp->sacked_out, next_skb);

1744
		tcp_adjust_fackets_out(sk, next_skb, tcp_skb_pcount(next_skb));
1745
		tp->packets_out -= tcp_skb_pcount(next_skb);
1746 1747 1748 1749

		/* changed transmit queue under us so clear hints */
		tcp_clear_retrans_hints_partial(tp);

L
Linus Torvalds 已提交
1750 1751 1752 1753 1754
		sk_stream_free_skb(sk, next_skb);
	}
}

/* Do a simple retransmit without using the backoff mechanisms in
1755
 * tcp_timer. This is used for path mtu discovery.
L
Linus Torvalds 已提交
1756
 * The socket is already locked here.
1757
 */
L
Linus Torvalds 已提交
1758 1759
void tcp_simple_retransmit(struct sock *sk)
{
1760
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1761 1762 1763 1764 1765
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
	unsigned int mss = tcp_current_mss(sk, 0);
	int lost = 0;

1766 1767 1768
	tcp_for_write_queue(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
1769
		if (skb->len > mss &&
L
Linus Torvalds 已提交
1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782
		    !(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);
			}
			if (!(TCP_SKB_CB(skb)->sacked&TCPCB_LOST)) {
				TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
				tp->lost_out += tcp_skb_pcount(skb);
				lost = 1;
			}
		}
	}

1783
	tcp_clear_all_retrans_hints(tp);
1784

L
Linus Torvalds 已提交
1785 1786 1787
	if (!lost)
		return;

1788
	tcp_verify_left_out(tp);
L
Linus Torvalds 已提交
1789

1790
	/* Don't muck with the congestion window here.
L
Linus Torvalds 已提交
1791 1792 1793 1794
	 * Reason is that we do not increase amount of _data_
	 * in network, but units changed and effective
	 * cwnd/ssthresh really reduced now.
	 */
1795
	if (icsk->icsk_ca_state != TCP_CA_Loss) {
L
Linus Torvalds 已提交
1796
		tp->high_seq = tp->snd_nxt;
1797
		tp->snd_ssthresh = tcp_current_ssthresh(sk);
L
Linus Torvalds 已提交
1798 1799
		tp->prior_ssthresh = 0;
		tp->undo_marker = 0;
1800
		tcp_set_ca_state(sk, TCP_CA_Loss);
L
Linus Torvalds 已提交
1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811
	}
	tcp_xmit_retransmit_queue(sk);
}

/* This retransmits one SKB.  Policy decisions and retransmit queue
 * state updates are done by the caller.  Returns non-zero if an
 * error occurred which prevented the send.
 */
int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);
J
John Heffner 已提交
1812
	struct inet_connection_sock *icsk = inet_csk(sk);
1813
	unsigned int cur_mss = tcp_current_mss(sk, 0);
L
Linus Torvalds 已提交
1814 1815
	int err;

J
John Heffner 已提交
1816 1817 1818 1819 1820
	/* Inconslusive MTU probe */
	if (icsk->icsk_mtup.probe_size) {
		icsk->icsk_mtup.probe_size = 0;
	}

L
Linus Torvalds 已提交
1821
	/* Do not sent more than we queued. 1/4 is reserved for possible
S
Stephen Hemminger 已提交
1822
	 * copying overhead: fragmentation, tunneling, mangling etc.
L
Linus Torvalds 已提交
1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844
	 */
	if (atomic_read(&sk->sk_wmem_alloc) >
	    min(sk->sk_wmem_queued + (sk->sk_wmem_queued >> 2), sk->sk_sndbuf))
		return -EAGAIN;

	if (before(TCP_SKB_CB(skb)->seq, tp->snd_una)) {
		if (before(TCP_SKB_CB(skb)->end_seq, tp->snd_una))
			BUG();
		if (tcp_trim_head(sk, skb, tp->snd_una - TCP_SKB_CB(skb)->seq))
			return -ENOMEM;
	}

	/* If receiver has shrunk his window, and skb is out of
	 * new window, do not retransmit it. The exception is the
	 * case, when window is shrunk to zero. In this case
	 * our retransmit serves as a zero window probe.
	 */
	if (!before(TCP_SKB_CB(skb)->seq, tp->snd_una+tp->snd_wnd)
	    && TCP_SKB_CB(skb)->seq != tp->snd_una)
		return -EAGAIN;

	if (skb->len > cur_mss) {
1845
		if (tcp_fragment(sk, skb, cur_mss, cur_mss))
L
Linus Torvalds 已提交
1846 1847 1848 1849
			return -ENOMEM; /* We'll try again later. */
	}

	/* Collapse two adjacent packets if worthwhile and we can. */
S
Stephen Hemminger 已提交
1850 1851 1852 1853 1854 1855 1856
	if (!(TCP_SKB_CB(skb)->flags & TCPCB_FLAG_SYN) &&
	    (skb->len < (cur_mss >> 1)) &&
	    (tcp_write_queue_next(sk, skb) != tcp_send_head(sk)) &&
	    (!tcp_skb_is_last(sk, skb)) &&
	    (skb_shinfo(skb)->nr_frags == 0 && skb_shinfo(tcp_write_queue_next(sk, skb))->nr_frags == 0) &&
	    (tcp_skb_pcount(skb) == 1 && tcp_skb_pcount(tcp_write_queue_next(sk, skb)) == 1) &&
	    (sysctl_tcp_retrans_collapse != 0))
L
Linus Torvalds 已提交
1857 1858
		tcp_retrans_try_collapse(sk, skb, cur_mss);

1859
	if (inet_csk(sk)->icsk_af_ops->rebuild_header(sk))
L
Linus Torvalds 已提交
1860 1861 1862 1863 1864 1865
		return -EHOSTUNREACH; /* Routing failure or similar. */

	/* Some Solaris stacks overoptimize and ignore the FIN on a
	 * retransmit when old data is attached.  So strip it off
	 * since it is cheap to do so and saves bytes on the network.
	 */
S
Stephen Hemminger 已提交
1866 1867 1868
	if (skb->len > 0 &&
	    (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN) &&
	    tp->snd_una == (TCP_SKB_CB(skb)->end_seq - 1)) {
L
Linus Torvalds 已提交
1869 1870
		if (!pskb_trim(skb, 0)) {
			TCP_SKB_CB(skb)->seq = TCP_SKB_CB(skb)->end_seq - 1;
1871 1872 1873
			skb_shinfo(skb)->gso_segs = 1;
			skb_shinfo(skb)->gso_size = 0;
			skb_shinfo(skb)->gso_type = 0;
L
Linus Torvalds 已提交
1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
			skb->ip_summed = CHECKSUM_NONE;
			skb->csum = 0;
		}
	}

	/* Make a copy, if the first transmission SKB clone we made
	 * is still in somebody's hands, else make a clone.
	 */
	TCP_SKB_CB(skb)->when = tcp_time_stamp;

1884
	err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897

	if (err == 0) {
		/* Update global TCP statistics. */
		TCP_INC_STATS(TCP_MIB_RETRANSSEGS);

		tp->total_retrans++;

#if FASTRETRANS_DEBUG > 0
		if (TCP_SKB_CB(skb)->sacked&TCPCB_SACKED_RETRANS) {
			if (net_ratelimit())
				printk(KERN_DEBUG "retrans_out leaked.\n");
		}
#endif
1898 1899
		if (!tp->retrans_out)
			tp->lost_retrans_low = tp->snd_nxt;
L
Linus Torvalds 已提交
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926
		TCP_SKB_CB(skb)->sacked |= TCPCB_RETRANS;
		tp->retrans_out += tcp_skb_pcount(skb);

		/* Save stamp of the first retransmit. */
		if (!tp->retrans_stamp)
			tp->retrans_stamp = TCP_SKB_CB(skb)->when;

		tp->undo_retrans++;

		/* snd_nxt is stored to detect loss of retransmitted segment,
		 * see tcp_input.c tcp_sacktag_write_queue().
		 */
		TCP_SKB_CB(skb)->ack_seq = tp->snd_nxt;
	}
	return err;
}

/* This gets called after a retransmit timeout, and the initially
 * retransmitted data is acknowledged.  It tries to continue
 * resending the rest of the retransmit queue, until either
 * we've sent it all or the congestion window limit is reached.
 * If doing SACK, the first ACK which comes back for a timeout
 * based retransmit packet might feed us FACK information again.
 * If so, we use it to avoid unnecessarily retransmissions.
 */
void tcp_xmit_retransmit_queue(struct sock *sk)
{
1927
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1928 1929
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
1930 1931 1932 1933 1934 1935
	int packet_cnt;

	if (tp->retransmit_skb_hint) {
		skb = tp->retransmit_skb_hint;
		packet_cnt = tp->retransmit_cnt_hint;
	}else{
1936
		skb = tcp_write_queue_head(sk);
1937 1938
		packet_cnt = 0;
	}
L
Linus Torvalds 已提交
1939 1940

	/* First pass: retransmit lost packets. */
1941
	if (tp->lost_out) {
1942
		tcp_for_write_queue_from(skb, sk) {
L
Linus Torvalds 已提交
1943 1944
			__u8 sacked = TCP_SKB_CB(skb)->sacked;

1945 1946
			if (skb == tcp_send_head(sk))
				break;
1947 1948 1949 1950
			/* we could do better than to assign each time */
			tp->retransmit_skb_hint = skb;
			tp->retransmit_cnt_hint = packet_cnt;

L
Linus Torvalds 已提交
1951 1952 1953 1954 1955 1956 1957 1958 1959 1960
			/* Assume this retransmit will generate
			 * only one packet for congestion window
			 * calculation purposes.  This works because
			 * tcp_retransmit_skb() will chop up the
			 * packet to be MSS sized and all the
			 * packet counting works out.
			 */
			if (tcp_packets_in_flight(tp) >= tp->snd_cwnd)
				return;

1961
			if (sacked & TCPCB_LOST) {
L
Linus Torvalds 已提交
1962
				if (!(sacked&(TCPCB_SACKED_ACKED|TCPCB_SACKED_RETRANS))) {
1963 1964
					if (tcp_retransmit_skb(sk, skb)) {
						tp->retransmit_skb_hint = NULL;
L
Linus Torvalds 已提交
1965
						return;
1966
					}
1967
					if (icsk->icsk_ca_state != TCP_CA_Loss)
L
Linus Torvalds 已提交
1968 1969 1970 1971
						NET_INC_STATS_BH(LINUX_MIB_TCPFASTRETRANS);
					else
						NET_INC_STATS_BH(LINUX_MIB_TCPSLOWSTARTRETRANS);

1972
					if (skb == tcp_write_queue_head(sk))
1973
						inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
1974 1975
									  inet_csk(sk)->icsk_rto,
									  TCP_RTO_MAX);
L
Linus Torvalds 已提交
1976 1977
				}

1978 1979
				packet_cnt += tcp_skb_pcount(skb);
				if (packet_cnt >= tp->lost_out)
L
Linus Torvalds 已提交
1980 1981 1982 1983 1984 1985 1986 1987
					break;
			}
		}
	}

	/* OK, demanded retransmission is finished. */

	/* Forward retransmissions are possible only during Recovery. */
1988
	if (icsk->icsk_ca_state != TCP_CA_Recovery)
L
Linus Torvalds 已提交
1989 1990 1991
		return;

	/* No forward retransmissions in Reno are possible. */
1992
	if (tcp_is_reno(tp))
L
Linus Torvalds 已提交
1993 1994 1995 1996 1997 1998
		return;

	/* Yeah, we have to make difficult choice between forward transmission
	 * and retransmission... Both ways have their merits...
	 *
	 * For now we do not retransmit anything, while we have some new
1999 2000
	 * segments to send. In the other cases, follow rule 3 for
	 * NextSeg() specified in RFC3517.
L
Linus Torvalds 已提交
2001 2002
	 */

2003
	if (tcp_may_send_now(sk))
L
Linus Torvalds 已提交
2004 2005
		return;

2006 2007 2008 2009 2010
	/* If nothing is SACKed, highest_sack in the loop won't be valid */
	if (!tp->sacked_out)
		return;

	if (tp->forward_skb_hint)
2011
		skb = tp->forward_skb_hint;
2012
	else
2013
		skb = tcp_write_queue_head(sk);
2014

2015 2016 2017
	tcp_for_write_queue_from(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
2018
		tp->forward_skb_hint = skb;
L
Linus Torvalds 已提交
2019

2020
		if (after(TCP_SKB_CB(skb)->seq, tcp_highest_sack_seq(tp)))
L
Linus Torvalds 已提交
2021 2022 2023 2024 2025 2026 2027 2028 2029
			break;

		if (tcp_packets_in_flight(tp) >= tp->snd_cwnd)
			break;

		if (TCP_SKB_CB(skb)->sacked & TCPCB_TAGBITS)
			continue;

		/* Ok, retransmit it. */
2030 2031
		if (tcp_retransmit_skb(sk, skb)) {
			tp->forward_skb_hint = NULL;
L
Linus Torvalds 已提交
2032
			break;
2033
		}
L
Linus Torvalds 已提交
2034

2035
		if (skb == tcp_write_queue_head(sk))
2036 2037 2038
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
						  inet_csk(sk)->icsk_rto,
						  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049

		NET_INC_STATS_BH(LINUX_MIB_TCPFORWARDRETRANS);
	}
}


/* Send a fin.  The caller locks the socket for us.  This cannot be
 * allowed to fail queueing a FIN frame under any circumstances.
 */
void tcp_send_fin(struct sock *sk)
{
2050
	struct tcp_sock *tp = tcp_sk(sk);
2051
	struct sk_buff *skb = tcp_write_queue_tail(sk);
L
Linus Torvalds 已提交
2052
	int mss_now;
2053

L
Linus Torvalds 已提交
2054 2055 2056 2057 2058 2059
	/* Optimization, tack on the FIN if we have a queue of
	 * unsent frames.  But be careful about outgoing SACKS
	 * and IP options.
	 */
	mss_now = tcp_current_mss(sk, 1);

2060
	if (tcp_send_head(sk) != NULL) {
L
Linus Torvalds 已提交
2061 2062 2063 2064 2065 2066
		TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_FIN;
		TCP_SKB_CB(skb)->end_seq++;
		tp->write_seq++;
	} else {
		/* Socket is locked, keep trying until memory is available. */
		for (;;) {
2067
			skb = alloc_skb_fclone(MAX_TCP_HEADER, GFP_KERNEL);
L
Linus Torvalds 已提交
2068 2069 2070 2071 2072 2073 2074 2075 2076 2077
			if (skb)
				break;
			yield();
		}

		/* Reserve space for headers and prepare control bits. */
		skb_reserve(skb, MAX_TCP_HEADER);
		skb->csum = 0;
		TCP_SKB_CB(skb)->flags = (TCPCB_FLAG_ACK | TCPCB_FLAG_FIN);
		TCP_SKB_CB(skb)->sacked = 0;
2078 2079 2080
		skb_shinfo(skb)->gso_segs = 1;
		skb_shinfo(skb)->gso_size = 0;
		skb_shinfo(skb)->gso_type = 0;
L
Linus Torvalds 已提交
2081 2082 2083 2084 2085 2086

		/* FIN eats a sequence byte, write_seq advanced by tcp_queue_skb(). */
		TCP_SKB_CB(skb)->seq = tp->write_seq;
		TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq + 1;
		tcp_queue_skb(sk, skb);
	}
2087
	__tcp_push_pending_frames(sk, mss_now, TCP_NAGLE_OFF);
L
Linus Torvalds 已提交
2088 2089 2090 2091 2092
}

/* We get here when a process closes a file descriptor (either due to
 * an explicit close() or as a byproduct of exit()'ing) and there
 * was unread data in the receive queue.  This behavior is recommended
2093
 * by RFC 2525, section 2.17.  -DaveM
L
Linus Torvalds 已提交
2094
 */
A
Al Viro 已提交
2095
void tcp_send_active_reset(struct sock *sk, gfp_t priority)
L
Linus Torvalds 已提交
2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110
{
	struct sk_buff *skb;

	/* NOTE: No TCP options attached and we never retransmit this. */
	skb = alloc_skb(MAX_TCP_HEADER, priority);
	if (!skb) {
		NET_INC_STATS(LINUX_MIB_TCPABORTFAILED);
		return;
	}

	/* Reserve space for headers and prepare control bits. */
	skb_reserve(skb, MAX_TCP_HEADER);
	skb->csum = 0;
	TCP_SKB_CB(skb)->flags = (TCPCB_FLAG_ACK | TCPCB_FLAG_RST);
	TCP_SKB_CB(skb)->sacked = 0;
2111 2112 2113
	skb_shinfo(skb)->gso_segs = 1;
	skb_shinfo(skb)->gso_size = 0;
	skb_shinfo(skb)->gso_type = 0;
L
Linus Torvalds 已提交
2114 2115

	/* Send it off. */
2116
	TCP_SKB_CB(skb)->seq = tcp_acceptable_seq(sk);
L
Linus Torvalds 已提交
2117 2118
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2119
	if (tcp_transmit_skb(sk, skb, 0, priority))
L
Linus Torvalds 已提交
2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131
		NET_INC_STATS(LINUX_MIB_TCPABORTFAILED);
}

/* WARNING: This routine must only be called when we have already sent
 * a SYN packet that crossed the incoming SYN that caused this routine
 * to get called. If this assumption fails then the initial rcv_wnd
 * and rcv_wscale values will not be correct.
 */
int tcp_send_synack(struct sock *sk)
{
	struct sk_buff* skb;

2132
	skb = tcp_write_queue_head(sk);
L
Linus Torvalds 已提交
2133 2134 2135 2136 2137 2138 2139 2140 2141
	if (skb == NULL || !(TCP_SKB_CB(skb)->flags&TCPCB_FLAG_SYN)) {
		printk(KERN_DEBUG "tcp_send_synack: wrong queue state\n");
		return -EFAULT;
	}
	if (!(TCP_SKB_CB(skb)->flags&TCPCB_FLAG_ACK)) {
		if (skb_cloned(skb)) {
			struct sk_buff *nskb = skb_copy(skb, GFP_ATOMIC);
			if (nskb == NULL)
				return -ENOMEM;
2142
			tcp_unlink_write_queue(skb, sk);
L
Linus Torvalds 已提交
2143
			skb_header_release(nskb);
2144
			__tcp_add_write_queue_head(sk, nskb);
L
Linus Torvalds 已提交
2145 2146 2147 2148 2149 2150 2151 2152 2153
			sk_stream_free_skb(sk, skb);
			sk_charge_skb(sk, nskb);
			skb = nskb;
		}

		TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_ACK;
		TCP_ECN_send_synack(tcp_sk(sk), skb);
	}
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2154
	return tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
2155 2156 2157 2158 2159 2160
}

/*
 * Prepare a SYN-ACK.
 */
struct sk_buff * tcp_make_synack(struct sock *sk, struct dst_entry *dst,
2161
				 struct request_sock *req)
L
Linus Torvalds 已提交
2162
{
2163
	struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
2164 2165 2166 2167
	struct tcp_sock *tp = tcp_sk(sk);
	struct tcphdr *th;
	int tcp_header_size;
	struct sk_buff *skb;
2168 2169 2170 2171
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *md5;
	__u8 *md5_hash_location;
#endif
L
Linus Torvalds 已提交
2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182

	skb = sock_wmalloc(sk, MAX_TCP_HEADER + 15, 1, GFP_ATOMIC);
	if (skb == NULL)
		return NULL;

	/* Reserve space for headers. */
	skb_reserve(skb, MAX_TCP_HEADER);

	skb->dst = dst_clone(dst);

	tcp_header_size = (sizeof(struct tcphdr) + TCPOLEN_MSS +
2183 2184
			   (ireq->tstamp_ok ? TCPOLEN_TSTAMP_ALIGNED : 0) +
			   (ireq->wscale_ok ? TCPOLEN_WSCALE_ALIGNED : 0) +
L
Linus Torvalds 已提交
2185
			   /* SACK_PERM is in the place of NOP NOP of TS */
2186
			   ((ireq->sack_ok && !ireq->tstamp_ok) ? TCPOLEN_SACKPERM_ALIGNED : 0));
2187 2188 2189 2190 2191 2192 2193

#ifdef CONFIG_TCP_MD5SIG
	/* Are we doing MD5 on this segment? If so - make room for it */
	md5 = tcp_rsk(req)->af_specific->md5_lookup(sk, req);
	if (md5)
		tcp_header_size += TCPOLEN_MD5SIG_ALIGNED;
#endif
2194 2195
	skb_push(skb, tcp_header_size);
	skb_reset_transport_header(skb);
L
Linus Torvalds 已提交
2196

2197
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
2198 2199 2200 2201 2202
	memset(th, 0, sizeof(struct tcphdr));
	th->syn = 1;
	th->ack = 1;
	TCP_ECN_make_synack(req, th);
	th->source = inet_sk(sk)->sport;
2203 2204
	th->dest = ireq->rmt_port;
	TCP_SKB_CB(skb)->seq = tcp_rsk(req)->snt_isn;
L
Linus Torvalds 已提交
2205 2206
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq + 1;
	TCP_SKB_CB(skb)->sacked = 0;
2207 2208 2209
	skb_shinfo(skb)->gso_segs = 1;
	skb_shinfo(skb)->gso_size = 0;
	skb_shinfo(skb)->gso_type = 0;
L
Linus Torvalds 已提交
2210
	th->seq = htonl(TCP_SKB_CB(skb)->seq);
2211
	th->ack_seq = htonl(tcp_rsk(req)->rcv_isn + 1);
L
Linus Torvalds 已提交
2212
	if (req->rcv_wnd == 0) { /* ignored for retransmitted syns */
2213
		__u8 rcv_wscale;
L
Linus Torvalds 已提交
2214 2215 2216
		/* Set this up on the first call only */
		req->window_clamp = tp->window_clamp ? : dst_metric(dst, RTAX_WINDOW);
		/* tcp_full_space because it is guaranteed to be the first packet */
2217
		tcp_select_initial_window(tcp_full_space(sk),
2218
			dst_metric(dst, RTAX_ADVMSS) - (ireq->tstamp_ok ? TCPOLEN_TSTAMP_ALIGNED : 0),
L
Linus Torvalds 已提交
2219 2220
			&req->rcv_wnd,
			&req->window_clamp,
2221
			ireq->wscale_ok,
L
Linus Torvalds 已提交
2222
			&rcv_wscale);
2223
		ireq->rcv_wscale = rcv_wscale;
L
Linus Torvalds 已提交
2224 2225 2226
	}

	/* RFC1323: The window in SYN & SYN/ACK segments is never scaled. */
2227
	th->window = htons(min(req->rcv_wnd, 65535U));
L
Linus Torvalds 已提交
2228 2229

	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2230
	tcp_syn_build_options((__be32 *)(th + 1), dst_metric(dst, RTAX_ADVMSS), ireq->tstamp_ok,
2231
			      ireq->sack_ok, ireq->wscale_ok, ireq->rcv_wscale,
L
Linus Torvalds 已提交
2232
			      TCP_SKB_CB(skb)->when,
2233 2234 2235 2236 2237 2238 2239
			      req->ts_recent,
			      (
#ifdef CONFIG_TCP_MD5SIG
			       md5 ? &md5_hash_location :
#endif
			       NULL)
			      );
L
Linus Torvalds 已提交
2240 2241 2242 2243

	skb->csum = 0;
	th->doff = (tcp_header_size >> 2);
	TCP_INC_STATS(TCP_MIB_OUTSEGS);
2244 2245 2246 2247 2248 2249 2250

#ifdef CONFIG_TCP_MD5SIG
	/* Okay, we have all we need - do the md5 hash if needed */
	if (md5) {
		tp->af_specific->calc_md5_hash(md5_hash_location,
					       md5,
					       NULL, dst, req,
2251
					       tcp_hdr(skb), sk->sk_protocol,
2252 2253 2254 2255
					       skb->len);
	}
#endif

L
Linus Torvalds 已提交
2256 2257 2258
	return skb;
}

2259
/*
L
Linus Torvalds 已提交
2260
 * Do all connect socket setups that can be done AF independent.
2261
 */
S
Stephen Hemminger 已提交
2262
static void tcp_connect_init(struct sock *sk)
L
Linus Torvalds 已提交
2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273
{
	struct dst_entry *dst = __sk_dst_get(sk);
	struct tcp_sock *tp = tcp_sk(sk);
	__u8 rcv_wscale;

	/* We'll fix this up when we get a response from the other end.
	 * See tcp_input.c:tcp_rcv_state_process case TCP_SYN_SENT.
	 */
	tp->tcp_header_len = sizeof(struct tcphdr) +
		(sysctl_tcp_timestamps ? TCPOLEN_TSTAMP_ALIGNED : 0);

2274 2275 2276 2277 2278
#ifdef CONFIG_TCP_MD5SIG
	if (tp->af_specific->md5_lookup(sk, sk) != NULL)
		tp->tcp_header_len += TCPOLEN_MD5SIG_ALIGNED;
#endif

L
Linus Torvalds 已提交
2279 2280 2281 2282
	/* If user gave his TCP_MAXSEG, record it to clamp */
	if (tp->rx_opt.user_mss)
		tp->rx_opt.mss_clamp = tp->rx_opt.user_mss;
	tp->max_window = 0;
J
John Heffner 已提交
2283
	tcp_mtup_init(sk);
L
Linus Torvalds 已提交
2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310
	tcp_sync_mss(sk, dst_mtu(dst));

	if (!tp->window_clamp)
		tp->window_clamp = dst_metric(dst, RTAX_WINDOW);
	tp->advmss = dst_metric(dst, RTAX_ADVMSS);
	tcp_initialize_rcv_mss(sk);

	tcp_select_initial_window(tcp_full_space(sk),
				  tp->advmss - (tp->rx_opt.ts_recent_stamp ? tp->tcp_header_len - sizeof(struct tcphdr) : 0),
				  &tp->rcv_wnd,
				  &tp->window_clamp,
				  sysctl_tcp_window_scaling,
				  &rcv_wscale);

	tp->rx_opt.rcv_wscale = rcv_wscale;
	tp->rcv_ssthresh = tp->rcv_wnd;

	sk->sk_err = 0;
	sock_reset_flag(sk, SOCK_DONE);
	tp->snd_wnd = 0;
	tcp_init_wl(tp, tp->write_seq, 0);
	tp->snd_una = tp->write_seq;
	tp->snd_sml = tp->write_seq;
	tp->rcv_nxt = 0;
	tp->rcv_wup = 0;
	tp->copied_seq = 0;

2311 2312
	inet_csk(sk)->icsk_rto = TCP_TIMEOUT_INIT;
	inet_csk(sk)->icsk_retransmits = 0;
L
Linus Torvalds 已提交
2313 2314 2315 2316 2317
	tcp_clear_retrans(tp);
}

/*
 * Build a SYN and send it off.
2318
 */
L
Linus Torvalds 已提交
2319 2320 2321 2322 2323 2324 2325
int tcp_connect(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *buff;

	tcp_connect_init(sk);

2326
	buff = alloc_skb_fclone(MAX_TCP_HEADER + 15, sk->sk_allocation);
L
Linus Torvalds 已提交
2327 2328 2329 2330 2331 2332 2333
	if (unlikely(buff == NULL))
		return -ENOBUFS;

	/* Reserve space for headers. */
	skb_reserve(buff, MAX_TCP_HEADER);

	TCP_SKB_CB(buff)->flags = TCPCB_FLAG_SYN;
2334
	TCP_ECN_send_syn(sk, buff);
L
Linus Torvalds 已提交
2335
	TCP_SKB_CB(buff)->sacked = 0;
2336 2337 2338
	skb_shinfo(buff)->gso_segs = 1;
	skb_shinfo(buff)->gso_size = 0;
	skb_shinfo(buff)->gso_type = 0;
L
Linus Torvalds 已提交
2339
	buff->csum = 0;
2340
	tp->snd_nxt = tp->write_seq;
L
Linus Torvalds 已提交
2341 2342 2343 2344 2345 2346 2347
	TCP_SKB_CB(buff)->seq = tp->write_seq++;
	TCP_SKB_CB(buff)->end_seq = tp->write_seq;

	/* Send it off. */
	TCP_SKB_CB(buff)->when = tcp_time_stamp;
	tp->retrans_stamp = TCP_SKB_CB(buff)->when;
	skb_header_release(buff);
2348
	__tcp_add_write_queue_tail(sk, buff);
L
Linus Torvalds 已提交
2349 2350
	sk_charge_skb(sk, buff);
	tp->packets_out += tcp_skb_pcount(buff);
2351
	tcp_transmit_skb(sk, buff, 1, GFP_KERNEL);
2352 2353 2354 2355 2356 2357

	/* We change tp->snd_nxt after the tcp_transmit_skb() call
	 * in order to make this packet get counted in tcpOutSegs.
	 */
	tp->snd_nxt = tp->write_seq;
	tp->pushed_seq = tp->write_seq;
L
Linus Torvalds 已提交
2358 2359 2360
	TCP_INC_STATS(TCP_MIB_ACTIVEOPENS);

	/* Timer for repeating the SYN until an answer. */
2361 2362
	inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
				  inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
L
Linus Torvalds 已提交
2363 2364 2365 2366 2367 2368 2369 2370 2371
	return 0;
}

/* Send out a delayed ack, the caller does the policy checking
 * to see if we should even be here.  See tcp_input.c:tcp_ack_snd_check()
 * for details.
 */
void tcp_send_delayed_ack(struct sock *sk)
{
2372 2373
	struct inet_connection_sock *icsk = inet_csk(sk);
	int ato = icsk->icsk_ack.ato;
L
Linus Torvalds 已提交
2374 2375 2376
	unsigned long timeout;

	if (ato > TCP_DELACK_MIN) {
2377
		const struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2378 2379
		int max_ato = HZ/2;

2380
		if (icsk->icsk_ack.pingpong || (icsk->icsk_ack.pending & ICSK_ACK_PUSHED))
L
Linus Torvalds 已提交
2381 2382 2383 2384 2385
			max_ato = TCP_DELACK_MAX;

		/* Slow path, intersegment interval is "high". */

		/* If some rtt estimate is known, use it to bound delayed ack.
2386
		 * Do not use inet_csk(sk)->icsk_rto here, use results of rtt measurements
L
Linus Torvalds 已提交
2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402
		 * directly.
		 */
		if (tp->srtt) {
			int rtt = max(tp->srtt>>3, TCP_DELACK_MIN);

			if (rtt < max_ato)
				max_ato = rtt;
		}

		ato = min(ato, max_ato);
	}

	/* Stay within the limit we were given */
	timeout = jiffies + ato;

	/* Use new timeout only if there wasn't a older one earlier. */
2403
	if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) {
L
Linus Torvalds 已提交
2404 2405 2406
		/* If delack timer was blocked or is about to expire,
		 * send ACK now.
		 */
2407 2408
		if (icsk->icsk_ack.blocked ||
		    time_before_eq(icsk->icsk_ack.timeout, jiffies + (ato >> 2))) {
L
Linus Torvalds 已提交
2409 2410 2411 2412
			tcp_send_ack(sk);
			return;
		}

2413 2414
		if (!time_before(timeout, icsk->icsk_ack.timeout))
			timeout = icsk->icsk_ack.timeout;
L
Linus Torvalds 已提交
2415
	}
2416 2417 2418
	icsk->icsk_ack.pending |= ICSK_ACK_SCHED | ICSK_ACK_TIMER;
	icsk->icsk_ack.timeout = timeout;
	sk_reset_timer(sk, &icsk->icsk_delack_timer, timeout);
L
Linus Torvalds 已提交
2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433
}

/* This routine sends an ack and also updates the window. */
void tcp_send_ack(struct sock *sk)
{
	/* If we have been reset, we may not send again. */
	if (sk->sk_state != TCP_CLOSE) {
		struct sk_buff *buff;

		/* We are not putting this on the write queue, so
		 * tcp_transmit_skb() will set the ownership to this
		 * sock.
		 */
		buff = alloc_skb(MAX_TCP_HEADER, GFP_ATOMIC);
		if (buff == NULL) {
2434 2435
			inet_csk_schedule_ack(sk);
			inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN;
2436 2437
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
						  TCP_DELACK_MAX, TCP_RTO_MAX);
L
Linus Torvalds 已提交
2438 2439 2440 2441 2442 2443 2444 2445
			return;
		}

		/* Reserve space for headers and prepare control bits. */
		skb_reserve(buff, MAX_TCP_HEADER);
		buff->csum = 0;
		TCP_SKB_CB(buff)->flags = TCPCB_FLAG_ACK;
		TCP_SKB_CB(buff)->sacked = 0;
2446 2447 2448
		skb_shinfo(buff)->gso_segs = 1;
		skb_shinfo(buff)->gso_size = 0;
		skb_shinfo(buff)->gso_type = 0;
L
Linus Torvalds 已提交
2449 2450

		/* Send it off, this clears delayed acks for us. */
2451
		TCP_SKB_CB(buff)->seq = TCP_SKB_CB(buff)->end_seq = tcp_acceptable_seq(sk);
L
Linus Torvalds 已提交
2452
		TCP_SKB_CB(buff)->when = tcp_time_stamp;
2453
		tcp_transmit_skb(sk, buff, 0, GFP_ATOMIC);
L
Linus Torvalds 已提交
2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474
	}
}

/* This routine sends a packet with an out of date sequence
 * number. It assumes the other end will try to ack it.
 *
 * Question: what should we make while urgent mode?
 * 4.4BSD forces sending single byte of data. We cannot send
 * out of window data, because we have SND.NXT==SND.MAX...
 *
 * Current solution: to send TWO zero-length segments in urgent mode:
 * one is with SEG.SEQ=SND.UNA to deliver urgent pointer, another is
 * out-of-date with SND.UNA-1 to probe window.
 */
static int tcp_xmit_probe_skb(struct sock *sk, int urgent)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;

	/* We don't queue it, tcp_transmit_skb() sets ownership. */
	skb = alloc_skb(MAX_TCP_HEADER, GFP_ATOMIC);
2475
	if (skb == NULL)
L
Linus Torvalds 已提交
2476 2477 2478 2479 2480 2481 2482
		return -1;

	/* Reserve space for headers and set control bits. */
	skb_reserve(skb, MAX_TCP_HEADER);
	skb->csum = 0;
	TCP_SKB_CB(skb)->flags = TCPCB_FLAG_ACK;
	TCP_SKB_CB(skb)->sacked = urgent;
2483 2484 2485
	skb_shinfo(skb)->gso_segs = 1;
	skb_shinfo(skb)->gso_size = 0;
	skb_shinfo(skb)->gso_type = 0;
L
Linus Torvalds 已提交
2486 2487 2488 2489 2490 2491 2492 2493

	/* Use a previous sequence.  This should cause the other
	 * end to send an ack.  Don't queue or clone SKB, just
	 * send it.
	 */
	TCP_SKB_CB(skb)->seq = urgent ? tp->snd_una : tp->snd_una - 1;
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2494
	return tcp_transmit_skb(sk, skb, 0, GFP_ATOMIC);
L
Linus Torvalds 已提交
2495 2496 2497 2498 2499 2500 2501 2502
}

int tcp_write_wakeup(struct sock *sk)
{
	if (sk->sk_state != TCP_CLOSE) {
		struct tcp_sock *tp = tcp_sk(sk);
		struct sk_buff *skb;

2503
		if ((skb = tcp_send_head(sk)) != NULL &&
L
Linus Torvalds 已提交
2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519
		    before(TCP_SKB_CB(skb)->seq, tp->snd_una+tp->snd_wnd)) {
			int err;
			unsigned int mss = tcp_current_mss(sk, 0);
			unsigned int seg_size = tp->snd_una+tp->snd_wnd-TCP_SKB_CB(skb)->seq;

			if (before(tp->pushed_seq, TCP_SKB_CB(skb)->end_seq))
				tp->pushed_seq = TCP_SKB_CB(skb)->end_seq;

			/* We are probing the opening of a window
			 * but the window size is != 0
			 * must have been a result SWS avoidance ( sender )
			 */
			if (seg_size < TCP_SKB_CB(skb)->end_seq - TCP_SKB_CB(skb)->seq ||
			    skb->len > mss) {
				seg_size = min(seg_size, mss);
				TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
2520
				if (tcp_fragment(sk, skb, seg_size, mss))
L
Linus Torvalds 已提交
2521 2522
					return -1;
			} else if (!tcp_skb_pcount(skb))
2523
				tcp_set_skb_tso_segs(sk, skb, mss);
L
Linus Torvalds 已提交
2524 2525 2526

			TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
			TCP_SKB_CB(skb)->when = tcp_time_stamp;
2527
			err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
2528
			if (!err) {
2529
				update_send_head(sk, skb);
L
Linus Torvalds 已提交
2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546
			}
			return err;
		} else {
			if (tp->urg_mode &&
			    between(tp->snd_up, tp->snd_una+1, tp->snd_una+0xFFFF))
				tcp_xmit_probe_skb(sk, TCPCB_URG);
			return tcp_xmit_probe_skb(sk, 0);
		}
	}
	return -1;
}

/* A window probe timeout has occurred.  If window is not closed send
 * a partial packet else a zero probe.
 */
void tcp_send_probe0(struct sock *sk)
{
2547
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2548 2549 2550 2551 2552
	struct tcp_sock *tp = tcp_sk(sk);
	int err;

	err = tcp_write_wakeup(sk);

2553
	if (tp->packets_out || !tcp_send_head(sk)) {
L
Linus Torvalds 已提交
2554
		/* Cancel probe timer, if it is not required. */
2555
		icsk->icsk_probes_out = 0;
2556
		icsk->icsk_backoff = 0;
L
Linus Torvalds 已提交
2557 2558 2559 2560
		return;
	}

	if (err <= 0) {
2561 2562
		if (icsk->icsk_backoff < sysctl_tcp_retries2)
			icsk->icsk_backoff++;
2563
		icsk->icsk_probes_out++;
2564
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
2565 2566
					  min(icsk->icsk_rto << icsk->icsk_backoff, TCP_RTO_MAX),
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2567 2568
	} else {
		/* If packet was not sent due to local congestion,
2569
		 * do not backoff and do not remember icsk_probes_out.
L
Linus Torvalds 已提交
2570 2571 2572 2573
		 * Let local senders to fight for local resources.
		 *
		 * Use accumulated backoff yet.
		 */
2574 2575
		if (!icsk->icsk_probes_out)
			icsk->icsk_probes_out = 1;
2576
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
2577
					  min(icsk->icsk_rto << icsk->icsk_backoff,
2578 2579
					      TCP_RESOURCE_PROBE_INTERVAL),
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2580 2581 2582 2583 2584 2585 2586
	}
}

EXPORT_SYMBOL(tcp_connect);
EXPORT_SYMBOL(tcp_make_synack);
EXPORT_SYMBOL(tcp_simple_retransmit);
EXPORT_SYMBOL(tcp_sync_mss);
2587
EXPORT_SYMBOL(sysctl_tcp_tso_win_divisor);
J
John Heffner 已提交
2588
EXPORT_SYMBOL(tcp_mtup_init);