tcp_output.c 75.2 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7
/*
 * INET		An implementation of the TCP/IP protocol suite for the LINUX
 *		operating system.  INET is implemented using the  BSD Socket
 *		interface as the means of communication with the user level.
 *
 *		Implementation of the Transmission Control Protocol(TCP).
 *
8
 * Authors:	Ross Biro
L
Linus Torvalds 已提交
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42
 *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 *		Mark Evans, <evansmp@uhura.aston.ac.uk>
 *		Corey Minyard <wf-rch!minyard@relay.EU.net>
 *		Florian La Roche, <flla@stud.uni-sb.de>
 *		Charles Hedrick, <hedrick@klinzhai.rutgers.edu>
 *		Linus Torvalds, <torvalds@cs.helsinki.fi>
 *		Alan Cox, <gw4pts@gw4pts.ampr.org>
 *		Matthew Dillon, <dillon@apollo.west.oic.com>
 *		Arnt Gulbrandsen, <agulbra@nvg.unit.no>
 *		Jorge Cwik, <jorge@laser.satlink.net>
 */

/*
 * Changes:	Pedro Roque	:	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. */
43
int sysctl_tcp_retrans_collapse __read_mostly = 1;
L
Linus Torvalds 已提交
44

45
/* People can turn this on to work with those rare, broken TCPs that
46 47
 * interpret the window field as a signed quantity.
 */
48
int sysctl_tcp_workaround_signed_windows __read_mostly = 0;
49

L
Linus Torvalds 已提交
50 51 52 53
/* 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.
 */
54
int sysctl_tcp_tso_win_divisor __read_mostly = 3;
L
Linus Torvalds 已提交
55

56 57
int sysctl_tcp_mtu_probing __read_mostly = 0;
int sysctl_tcp_base_mss __read_mostly = 512;
J
John Heffner 已提交
58

59
/* By default, RFC2861 behavior.  */
60
int sysctl_tcp_slow_start_after_idle __read_mostly = 1;
61

62
/* Account for new data that has been sent to the network. */
63
static void tcp_event_new_data_sent(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
64
{
65
	struct tcp_sock *tp = tcp_sk(sk);
66
	unsigned int prior_packets = tp->packets_out;
67

68
	tcp_advance_send_head(sk, skb);
L
Linus Torvalds 已提交
69
	tp->snd_nxt = TCP_SKB_CB(skb)->end_seq;
70 71 72 73

	/* Don't override Nagle indefinately with F-RTO */
	if (tp->frto_counter == 2)
		tp->frto_counter = 3;
74 75 76 77 78

	tp->packets_out += tcp_skb_pcount(skb);
	if (!prior_packets)
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
					  inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
L
Linus Torvalds 已提交
79 80 81 82 83 84 85 86
}

/* 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:
 */
87
static inline __u32 tcp_acceptable_seq(struct sock *sk)
L
Linus Torvalds 已提交
88
{
89 90
	struct tcp_sock *tp = tcp_sk(sk);

91
	if (!before(tcp_wnd_end(tp), tp->snd_nxt))
L
Linus Torvalds 已提交
92 93
		return tp->snd_nxt;
	else
94
		return tcp_wnd_end(tp);
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
}

/* 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. */
127
static void tcp_cwnd_restart(struct sock *sk, struct dst_entry *dst)
L
Linus Torvalds 已提交
128
{
129
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
130 131 132 133
	s32 delta = tcp_time_stamp - tp->lsndtime;
	u32 restart_cwnd = tcp_init_cwnd(tp, dst);
	u32 cwnd = tp->snd_cwnd;

134
	tcp_ca_event(sk, CA_EVENT_CWND_RESTART);
L
Linus Torvalds 已提交
135

136
	tp->snd_ssthresh = tcp_current_ssthresh(sk);
L
Linus Torvalds 已提交
137 138
	restart_cwnd = min(restart_cwnd, cwnd);

139
	while ((delta -= inet_csk(sk)->icsk_rto) > 0 && cwnd > restart_cwnd)
L
Linus Torvalds 已提交
140 141 142 143 144 145
		cwnd >>= 1;
	tp->snd_cwnd = max(cwnd, restart_cwnd);
	tp->snd_cwnd_stamp = tcp_time_stamp;
	tp->snd_cwnd_used = 0;
}

146
/* Congestion state accounting after a packet has been sent. */
S
Stephen Hemminger 已提交
147 148
static void tcp_event_data_sent(struct tcp_sock *tp,
				struct sk_buff *skb, struct sock *sk)
L
Linus Torvalds 已提交
149
{
150 151
	struct inet_connection_sock *icsk = inet_csk(sk);
	const u32 now = tcp_time_stamp;
L
Linus Torvalds 已提交
152

153 154
	if (sysctl_tcp_slow_start_after_idle &&
	    (!tp->packets_out && (s32)(now - tp->lsndtime) > icsk->icsk_rto))
155
		tcp_cwnd_restart(sk, __sk_dst_get(sk));
L
Linus Torvalds 已提交
156 157 158 159 160 161

	tp->lsndtime = now;

	/* If it is a reply for ato after last received
	 * packet, enter pingpong mode.
	 */
162 163
	if ((u32)(now - icsk->icsk_ack.lrcvtime) < icsk->icsk_ack.ato)
		icsk->icsk_ack.pingpong = 1;
L
Linus Torvalds 已提交
164 165
}

166
/* Account for an ACK we sent. */
S
Stephen Hemminger 已提交
167
static inline void tcp_event_ack_sent(struct sock *sk, unsigned int pkts)
L
Linus Torvalds 已提交
168
{
169 170
	tcp_dec_quickack_mode(sk, pkts);
	inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
L
Linus Torvalds 已提交
171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195
}

/* 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
196 197 198 199 200 201
	 * 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
202
	 */
203 204 205 206 207
	if (sysctl_tcp_workaround_signed_windows)
		(*rcv_wnd) = min(space, MAX_TCP_WINDOW);
	else
		(*rcv_wnd) = space;

L
Linus Torvalds 已提交
208 209 210
	(*rcv_wscale) = 0;
	if (wscale_ok) {
		/* Set window scaling on max possible window
211
		 * See RFC1323 for an explanation of the limit to 14
L
Linus Torvalds 已提交
212 213
		 */
		space = max_t(u32, sysctl_tcp_rmem[2], sysctl_rmem_max);
214
		space = min_t(u32, space, *window_clamp);
L
Linus Torvalds 已提交
215 216 217 218 219 220 221
		while (space > 65535 && (*rcv_wscale) < 14) {
			space >>= 1;
			(*rcv_wscale)++;
		}
	}

	/* Set initial window to value enough for senders,
222
	 * following RFC2414. Senders, not following this RFC,
L
Linus Torvalds 已提交
223 224
	 * will be satisfied with 2.
	 */
225
	if (mss > (1 << *rcv_wscale)) {
226
		int init_cwnd = 4;
227
		if (mss > 1460 * 3)
L
Linus Torvalds 已提交
228
			init_cwnd = 2;
229 230
		else if (mss > 1460)
			init_cwnd = 3;
231 232
		if (*rcv_wnd > init_cwnd * mss)
			*rcv_wnd = init_cwnd * mss;
L
Linus Torvalds 已提交
233 234 235 236 237 238 239 240 241 242 243
	}

	/* 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
244
static u16 tcp_select_window(struct sock *sk)
L
Linus Torvalds 已提交
245 246 247 248 249 250
{
	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
251
	if (new_win < cur_win) {
L
Linus Torvalds 已提交
252 253 254 255 256 257 258
		/* 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.
		 */
259
		new_win = ALIGN(cur_win, 1 << tp->rx_opt.rcv_wscale);
L
Linus Torvalds 已提交
260 261 262 263 264 265 266
	}
	tp->rcv_wnd = new_win;
	tp->rcv_wup = tp->rcv_nxt;

	/* Make sure we do not exceed the maximum possible
	 * scaled window.
	 */
267
	if (!tp->rx_opt.rcv_wscale && sysctl_tcp_workaround_signed_windows)
L
Linus Torvalds 已提交
268 269 270 271 272 273 274 275 276 277 278 279 280 281
		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;
}

282
/* Packet ECN state for a SYN-ACK */
283
static inline void TCP_ECN_send_synack(struct tcp_sock *tp, struct sk_buff *skb)
284 285
{
	TCP_SKB_CB(skb)->flags &= ~TCPCB_FLAG_CWR;
286
	if (!(tp->ecn_flags & TCP_ECN_OK))
287 288 289
		TCP_SKB_CB(skb)->flags &= ~TCPCB_FLAG_ECE;
}

290
/* Packet ECN state for a SYN.  */
291 292 293 294 295
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;
296
	if (sysctl_tcp_ecn == 1) {
297
		TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_ECE | TCPCB_FLAG_CWR;
298 299 300 301 302 303 304 305 306 307 308
		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;
}

309 310 311
/* Set up ECN state for a packet on a ESTABLISHED socket that is about to
 * be sent.
 */
312 313 314 315 316 317 318 319 320 321
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);
322
			if (tp->ecn_flags & TCP_ECN_QUEUE_CWR) {
323 324 325 326 327 328 329 330 331 332 333 334 335
				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;
	}
}

336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
/* Constructs common control bits of non-data skb. If SYN/FIN is present,
 * auto increment end seqno.
 */
static void tcp_init_nondata_skb(struct sk_buff *skb, u32 seq, u8 flags)
{
	skb->csum = 0;

	TCP_SKB_CB(skb)->flags = flags;
	TCP_SKB_CB(skb)->sacked = 0;

	skb_shinfo(skb)->gso_segs = 1;
	skb_shinfo(skb)->gso_size = 0;
	skb_shinfo(skb)->gso_type = 0;

	TCP_SKB_CB(skb)->seq = seq;
	if (flags & (TCPCB_FLAG_SYN | TCPCB_FLAG_FIN))
		seq++;
	TCP_SKB_CB(skb)->end_seq = seq;
}

I
Ilpo Järvinen 已提交
356 357 358 359 360
static inline int tcp_urg_mode(const struct tcp_sock *tp)
{
	return tp->snd_una != tp->snd_up;
}

A
Adam Langley 已提交
361 362 363
#define OPTION_SACK_ADVERTISE	(1 << 0)
#define OPTION_TS		(1 << 1)
#define OPTION_MD5		(1 << 2)
364
#define OPTION_WSCALE		(1 << 3)
A
Adam Langley 已提交
365 366 367 368 369 370 371 372 373

struct tcp_out_options {
	u8 options;		/* bit field of OPTION_* */
	u8 ws;			/* window scale, 0 to disable */
	u8 num_sack_blocks;	/* number of SACK blocks to include */
	u16 mss;		/* 0 to disable */
	__u32 tsval, tsecr;	/* need to include OPTION_TS */
};

374 375 376
/* Write previously computed TCP options to the packet.
 *
 * Beware: Something in the Internet is very sensitive to the ordering of
377 378 379 380 381 382 383 384 385 386
 * TCP options, we learned this through the hard way, so be careful here.
 * Luckily we can at least blame others for their non-compliance but from
 * inter-operatibility perspective it seems that we're somewhat stuck with
 * the ordering which we have been using if we want to keep working with
 * those broken things (not that it currently hurts anybody as there isn't
 * particular reason why the ordering would need to be changed).
 *
 * At least SACK_PERM as the first option is known to lead to a disaster
 * (but it may well be that other scenarios fail similarly).
 */
A
Adam Langley 已提交
387 388 389 390
static void tcp_options_write(__be32 *ptr, struct tcp_sock *tp,
			      const struct tcp_out_options *opts,
			      __u8 **md5_hash) {
	if (unlikely(OPTION_MD5 & opts->options)) {
391 392
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
A
Adam Langley 已提交
393 394 395 396 397 398
			       (TCPOPT_MD5SIG << 8) |
			       TCPOLEN_MD5SIG);
		*md5_hash = (__u8 *)ptr;
		ptr += 4;
	} else {
		*md5_hash = NULL;
S
Stephen Hemminger 已提交
399
	}
A
Adam Langley 已提交
400

401 402 403 404 405 406
	if (unlikely(opts->mss)) {
		*ptr++ = htonl((TCPOPT_MSS << 24) |
			       (TCPOLEN_MSS << 16) |
			       opts->mss);
	}

A
Adam Langley 已提交
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
	if (likely(OPTION_TS & opts->options)) {
		if (unlikely(OPTION_SACK_ADVERTISE & opts->options)) {
			*ptr++ = htonl((TCPOPT_SACK_PERM << 24) |
				       (TCPOLEN_SACK_PERM << 16) |
				       (TCPOPT_TIMESTAMP << 8) |
				       TCPOLEN_TIMESTAMP);
		} else {
			*ptr++ = htonl((TCPOPT_NOP << 24) |
				       (TCPOPT_NOP << 16) |
				       (TCPOPT_TIMESTAMP << 8) |
				       TCPOLEN_TIMESTAMP);
		}
		*ptr++ = htonl(opts->tsval);
		*ptr++ = htonl(opts->tsecr);
	}

	if (unlikely(OPTION_SACK_ADVERTISE & opts->options &&
		     !(OPTION_TS & opts->options))) {
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
			       (TCPOPT_SACK_PERM << 8) |
			       TCPOLEN_SACK_PERM);
	}

431
	if (unlikely(OPTION_WSCALE & opts->options)) {
A
Adam Langley 已提交
432 433 434 435 436 437 438 439 440
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_WINDOW << 16) |
			       (TCPOLEN_WINDOW << 8) |
			       opts->ws);
	}

	if (unlikely(opts->num_sack_blocks)) {
		struct tcp_sack_block *sp = tp->rx_opt.dsack ?
			tp->duplicate_sack : tp->selective_acks;
S
Stephen Hemminger 已提交
441 442 443 444 445
		int this_sack;

		*ptr++ = htonl((TCPOPT_NOP  << 24) |
			       (TCPOPT_NOP  << 16) |
			       (TCPOPT_SACK <<  8) |
A
Adam Langley 已提交
446
			       (TCPOLEN_SACK_BASE + (opts->num_sack_blocks *
S
Stephen Hemminger 已提交
447
						     TCPOLEN_SACK_PERBLOCK)));
S
Stephen Hemminger 已提交
448

A
Adam Langley 已提交
449 450
		for (this_sack = 0; this_sack < opts->num_sack_blocks;
		     ++this_sack) {
S
Stephen Hemminger 已提交
451 452 453
			*ptr++ = htonl(sp[this_sack].start_seq);
			*ptr++ = htonl(sp[this_sack].end_seq);
		}
S
Stephen Hemminger 已提交
454

455
		tp->rx_opt.dsack = 0;
S
Stephen Hemminger 已提交
456
	}
A
Adam Langley 已提交
457 458
}

459 460 461
/* Compute TCP options for SYN packets. This is not the final
 * network wire format yet.
 */
A
Adam Langley 已提交
462 463 464 465 466
static unsigned tcp_syn_options(struct sock *sk, struct sk_buff *skb,
				struct tcp_out_options *opts,
				struct tcp_md5sig_key **md5) {
	struct tcp_sock *tp = tcp_sk(sk);
	unsigned size = 0;
467
	struct dst_entry *dst = __sk_dst_get(sk);
A
Adam Langley 已提交
468

469
#ifdef CONFIG_TCP_MD5SIG
A
Adam Langley 已提交
470 471 472 473
	*md5 = tp->af_specific->md5_lookup(sk, sk);
	if (*md5) {
		opts->options |= OPTION_MD5;
		size += TCPOLEN_MD5SIG_ALIGNED;
474
	}
A
Adam Langley 已提交
475 476
#else
	*md5 = NULL;
477
#endif
A
Adam Langley 已提交
478 479 480 481 482 483 484 485 486 487 488 489 490

	/* 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.  */
	opts->mss = tcp_advertise_mss(sk);
	size += TCPOLEN_MSS_ALIGNED;

491 492 493
	if (likely(sysctl_tcp_timestamps &&
		   !dst_feature(dst, RTAX_FEATURE_NO_TSTAMP) &&
		   *md5 == NULL)) {
A
Adam Langley 已提交
494 495 496 497 498
		opts->options |= OPTION_TS;
		opts->tsval = TCP_SKB_CB(skb)->when;
		opts->tsecr = tp->rx_opt.ts_recent;
		size += TCPOLEN_TSTAMP_ALIGNED;
	}
499 500
	if (likely(sysctl_tcp_window_scaling &&
		   !dst_feature(dst, RTAX_FEATURE_NO_WSCALE))) {
A
Adam Langley 已提交
501
		opts->ws = tp->rx_opt.rcv_wscale;
502 503
		opts->options |= OPTION_WSCALE;
		size += TCPOLEN_WSCALE_ALIGNED;
A
Adam Langley 已提交
504
	}
505 506
	if (likely(sysctl_tcp_sack &&
		   !dst_feature(dst, RTAX_FEATURE_NO_SACK))) {
A
Adam Langley 已提交
507
		opts->options |= OPTION_SACK_ADVERTISE;
508
		if (unlikely(!(OPTION_TS & opts->options)))
A
Adam Langley 已提交
509 510 511 512
			size += TCPOLEN_SACKPERM_ALIGNED;
	}

	return size;
S
Stephen Hemminger 已提交
513 514
}

515
/* Set up TCP options for SYN-ACKs. */
A
Adam Langley 已提交
516 517 518 519 520 521 522 523 524
static unsigned tcp_synack_options(struct sock *sk,
				   struct request_sock *req,
				   unsigned mss, struct sk_buff *skb,
				   struct tcp_out_options *opts,
				   struct tcp_md5sig_key **md5) {
	unsigned size = 0;
	struct inet_request_sock *ireq = inet_rsk(req);
	char doing_ts;

525
#ifdef CONFIG_TCP_MD5SIG
A
Adam Langley 已提交
526 527 528 529
	*md5 = tcp_rsk(req)->af_specific->md5_lookup(sk, req);
	if (*md5) {
		opts->options |= OPTION_MD5;
		size += TCPOLEN_MD5SIG_ALIGNED;
530
	}
A
Adam Langley 已提交
531 532
#else
	*md5 = NULL;
533
#endif
A
Adam Langley 已提交
534 535 536 537 538 539 540 541 542 543 544 545

	/* we can't fit any SACK blocks in a packet with MD5 + TS
	   options. There was discussion about disabling SACK rather than TS in
	   order to fit in better with old, buggy kernels, but that was deemed
	   to be unnecessary. */
	doing_ts = ireq->tstamp_ok && !(*md5 && ireq->sack_ok);

	opts->mss = mss;
	size += TCPOLEN_MSS_ALIGNED;

	if (likely(ireq->wscale_ok)) {
		opts->ws = ireq->rcv_wscale;
546 547
		opts->options |= OPTION_WSCALE;
		size += TCPOLEN_WSCALE_ALIGNED;
A
Adam Langley 已提交
548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563
	}
	if (likely(doing_ts)) {
		opts->options |= OPTION_TS;
		opts->tsval = TCP_SKB_CB(skb)->when;
		opts->tsecr = req->ts_recent;
		size += TCPOLEN_TSTAMP_ALIGNED;
	}
	if (likely(ireq->sack_ok)) {
		opts->options |= OPTION_SACK_ADVERTISE;
		if (unlikely(!doing_ts))
			size += TCPOLEN_SACKPERM_ALIGNED;
	}

	return size;
}

564 565 566
/* Compute TCP options for ESTABLISHED sockets. This is not the
 * final wire format yet.
 */
A
Adam Langley 已提交
567 568 569 570 571 572
static unsigned tcp_established_options(struct sock *sk, struct sk_buff *skb,
					struct tcp_out_options *opts,
					struct tcp_md5sig_key **md5) {
	struct tcp_skb_cb *tcb = skb ? TCP_SKB_CB(skb) : NULL;
	struct tcp_sock *tp = tcp_sk(sk);
	unsigned size = 0;
573
	unsigned int eff_sacks;
A
Adam Langley 已提交
574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591

#ifdef CONFIG_TCP_MD5SIG
	*md5 = tp->af_specific->md5_lookup(sk, sk);
	if (unlikely(*md5)) {
		opts->options |= OPTION_MD5;
		size += TCPOLEN_MD5SIG_ALIGNED;
	}
#else
	*md5 = NULL;
#endif

	if (likely(tp->rx_opt.tstamp_ok)) {
		opts->options |= OPTION_TS;
		opts->tsval = tcb ? tcb->when : 0;
		opts->tsecr = tp->rx_opt.ts_recent;
		size += TCPOLEN_TSTAMP_ALIGNED;
	}

592 593
	eff_sacks = tp->rx_opt.num_sacks + tp->rx_opt.dsack;
	if (unlikely(eff_sacks)) {
A
Adam Langley 已提交
594 595
		const unsigned remaining = MAX_TCP_OPTION_SPACE - size;
		opts->num_sack_blocks =
596
			min_t(unsigned, eff_sacks,
A
Adam Langley 已提交
597 598 599 600 601 602 603
			      (remaining - TCPOLEN_SACK_BASE_ALIGNED) /
			      TCPOLEN_SACK_PERBLOCK);
		size += TCPOLEN_SACK_BASE_ALIGNED +
			opts->num_sack_blocks * TCPOLEN_SACK_PERBLOCK;
	}

	return size;
S
Stephen Hemminger 已提交
604
}
L
Linus Torvalds 已提交
605 606 607 608 609 610 611 612 613 614 615 616

/* 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.
 */
617 618
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
619
{
620 621 622 623
	const struct inet_connection_sock *icsk = inet_csk(sk);
	struct inet_sock *inet;
	struct tcp_sock *tp;
	struct tcp_skb_cb *tcb;
A
Adam Langley 已提交
624 625
	struct tcp_out_options opts;
	unsigned tcp_options_size, tcp_header_size;
626 627
	struct tcp_md5sig_key *md5;
	__u8 *md5_hash_location;
628 629 630 631 632 633 634 635
	struct tcphdr *th;
	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.
	 */
636
	if (icsk->icsk_ca_ops->flags & TCP_CONG_RTT_STAMP)
637 638 639 640 641 642 643 644 645 646
		__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
647

648 649 650
	inet = inet_sk(sk);
	tp = tcp_sk(sk);
	tcb = TCP_SKB_CB(skb);
A
Adam Langley 已提交
651
	memset(&opts, 0, sizeof(opts));
L
Linus Torvalds 已提交
652

A
Adam Langley 已提交
653 654 655 656 657 658
	if (unlikely(tcb->flags & TCPCB_FLAG_SYN))
		tcp_options_size = tcp_syn_options(sk, skb, &opts, &md5);
	else
		tcp_options_size = tcp_established_options(sk, skb, &opts,
							   &md5);
	tcp_header_size = tcp_options_size + sizeof(struct tcphdr);
659

660 661 662
	if (tcp_packets_in_flight(tp) == 0)
		tcp_ca_event(sk, CA_EVENT_TX_START);

663 664
	skb_push(skb, tcp_header_size);
	skb_reset_transport_header(skb);
665
	skb_set_owner_w(skb, sk);
666 667

	/* Build TCP header and checksum it. */
668
	th = tcp_hdr(skb);
E
Eric Dumazet 已提交
669 670
	th->source		= inet->inet_sport;
	th->dest		= inet->inet_dport;
671 672
	th->seq			= htonl(tcb->seq);
	th->ack_seq		= htonl(tp->rcv_nxt);
673
	*(((__be16 *)th) + 6)	= htons(((tcp_header_size >> 2) << 12) |
674 675 676 677 678 679
					tcb->flags);

	if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
		/* RFC1323: The window in SYN & SYN/ACK segments
		 * is never scaled.
		 */
680
		th->window	= htons(min(tp->rcv_wnd, 65535U));
681 682 683 684 685
	} else {
		th->window	= htons(tcp_select_window(sk));
	}
	th->check		= 0;
	th->urg_ptr		= 0;
L
Linus Torvalds 已提交
686

I
Ilpo Järvinen 已提交
687
	/* The urg_mode check is necessary during a below snd_una win probe */
688 689 690 691 692 693 694 695
	if (unlikely(tcp_urg_mode(tp) && before(tcb->seq, tp->snd_up))) {
		if (before(tp->snd_up, tcb->seq + 0x10000)) {
			th->urg_ptr = htons(tp->snd_up - tcb->seq);
			th->urg = 1;
		} else if (after(tcb->seq + 0xFFFF, tp->snd_nxt)) {
			th->urg_ptr = 0xFFFF;
			th->urg = 1;
		}
696
	}
L
Linus Torvalds 已提交
697

A
Adam Langley 已提交
698 699
	tcp_options_write((__be32 *)(th + 1), tp, &opts, &md5_hash_location);
	if (likely((tcb->flags & TCPCB_FLAG_SYN) == 0))
700
		TCP_ECN_send(sk, skb, tcp_header_size);
L
Linus Torvalds 已提交
701

702 703 704
#ifdef CONFIG_TCP_MD5SIG
	/* Calculate the MD5 hash, as we have all we need now */
	if (md5) {
A
Adam Langley 已提交
705
		sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
706
		tp->af_specific->calc_md5_hash(md5_hash_location,
707
					       md5, sk, NULL, skb);
708 709 710
	}
#endif

711
	icsk->icsk_af_ops->send_check(sk, skb->len, skb);
L
Linus Torvalds 已提交
712

713 714
	if (likely(tcb->flags & TCPCB_FLAG_ACK))
		tcp_event_ack_sent(sk, tcp_skb_pcount(skb));
L
Linus Torvalds 已提交
715

716 717
	if (skb->len != tcp_header_size)
		tcp_event_data_sent(tp, skb, sk);
L
Linus Torvalds 已提交
718

719
	if (after(tcb->end_seq, tp->snd_nxt) || tcb->seq == tcb->end_seq)
P
Pavel Emelyanov 已提交
720
		TCP_INC_STATS(sock_net(sk), TCP_MIB_OUTSEGS);
L
Linus Torvalds 已提交
721

722
	err = icsk->icsk_af_ops->queue_xmit(skb, 0);
723
	if (likely(err <= 0))
724 725
		return err;

726
	tcp_enter_cwr(sk, 1);
727

728
	return net_xmit_eval(err);
L
Linus Torvalds 已提交
729 730
}

731
/* This routine just queues the buffer for sending.
L
Linus Torvalds 已提交
732 733 734 735 736 737 738 739 740 741 742
 *
 * 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);
743
	tcp_add_write_queue_tail(sk, skb);
744 745
	sk->sk_wmem_queued += skb->truesize;
	sk_mem_charge(sk, skb->truesize);
L
Linus Torvalds 已提交
746 747
}

748
/* Initialize TSO segments for a packet. */
749 750
static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb,
				 unsigned int mss_now)
751
{
752 753
	if (skb->len <= mss_now || !sk_can_gso(sk) ||
	    skb->ip_summed == CHECKSUM_NONE) {
754 755 756
		/* Avoid the costly divide in the normal
		 * non-TSO case.
		 */
757 758 759
		skb_shinfo(skb)->gso_segs = 1;
		skb_shinfo(skb)->gso_size = 0;
		skb_shinfo(skb)->gso_type = 0;
760
	} else {
I
Ilpo Järvinen 已提交
761
		skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss_now);
762
		skb_shinfo(skb)->gso_size = mss_now;
763
		skb_shinfo(skb)->gso_type = sk->sk_gso_type;
L
Linus Torvalds 已提交
764 765 766
	}
}

I
Ilpo Järvinen 已提交
767
/* When a modification to fackets out becomes necessary, we need to check
768
 * skb is counted to fackets_out or not.
I
Ilpo Järvinen 已提交
769
 */
770
static void tcp_adjust_fackets_out(struct sock *sk, struct sk_buff *skb,
I
Ilpo Järvinen 已提交
771 772
				   int decr)
{
773 774
	struct tcp_sock *tp = tcp_sk(sk);

775
	if (!tp->sacked_out || tcp_is_reno(tp))
I
Ilpo Järvinen 已提交
776 777
		return;

778
	if (after(tcp_highest_sack_seq(tp), TCP_SKB_CB(skb)->seq))
I
Ilpo Järvinen 已提交
779 780 781
		tp->fackets_out -= decr;
}

782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805
/* Pcount in the middle of the write queue got changed, we need to do various
 * tweaks to fix counters
 */
static void tcp_adjust_pcount(struct sock *sk, struct sk_buff *skb, int decr)
{
	struct tcp_sock *tp = tcp_sk(sk);

	tp->packets_out -= decr;

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

	/* Reno case is special. Sigh... */
	if (tcp_is_reno(tp) && decr > 0)
		tp->sacked_out -= min_t(u32, tp->sacked_out, decr);

	tcp_adjust_fackets_out(sk, skb, decr);

	if (tp->lost_skb_hint &&
	    before(TCP_SKB_CB(skb)->seq, TCP_SKB_CB(tp->lost_skb_hint)->seq) &&
806
	    (tcp_is_fack(tp) || (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)))
807 808 809 810 811
		tp->lost_cnt_hint -= decr;

	tcp_verify_left_out(tp);
}

L
Linus Torvalds 已提交
812 813
/* 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
814
 * packet to the list.  This won't be called frequently, I hope.
L
Linus Torvalds 已提交
815 816
 * Remember, these are still headerless SKBs at this point.
 */
817 818
int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len,
		 unsigned int mss_now)
L
Linus Torvalds 已提交
819 820 821
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *buff;
822
	int nsize, old_factor;
823
	int nlen;
824
	u8 flags;
L
Linus Torvalds 已提交
825

826
	BUG_ON(len > skb->len);
827

L
Linus Torvalds 已提交
828 829 830 831 832 833 834 835 836 837 838 839 840
	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
841

842 843
	sk->sk_wmem_queued += buff->truesize;
	sk_mem_charge(sk, buff->truesize);
844 845 846
	nlen = skb->len - len - nsize;
	buff->truesize += nlen;
	skb->truesize -= nlen;
L
Linus Torvalds 已提交
847 848 849 850 851 852 853 854

	/* 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;
855
	TCP_SKB_CB(skb)->flags = flags & ~(TCPCB_FLAG_FIN | TCPCB_FLAG_PSH);
L
Linus Torvalds 已提交
856
	TCP_SKB_CB(buff)->flags = flags;
857
	TCP_SKB_CB(buff)->sacked = TCP_SKB_CB(skb)->sacked;
L
Linus Torvalds 已提交
858

859
	if (!skb_shinfo(skb)->nr_frags && skb->ip_summed != CHECKSUM_PARTIAL) {
L
Linus Torvalds 已提交
860
		/* Copy and checksum data tail into the new buffer. */
861 862
		buff->csum = csum_partial_copy_nocheck(skb->data + len,
						       skb_put(buff, nsize),
L
Linus Torvalds 已提交
863 864 865 866 867 868
						       nsize, 0);

		skb_trim(skb, len);

		skb->csum = csum_block_sub(skb->csum, buff->csum, len);
	} else {
869
		skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
870 871 872 873 874 875 876 877 878
		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;
879
	buff->tstamp = skb->tstamp;
L
Linus Torvalds 已提交
880

881 882
	old_factor = tcp_skb_pcount(skb);

L
Linus Torvalds 已提交
883
	/* Fix up tso_factor for both original and new SKB.  */
884 885
	tcp_set_skb_tso_segs(sk, skb, mss_now);
	tcp_set_skb_tso_segs(sk, buff, mss_now);
L
Linus Torvalds 已提交
886

887 888 889
	/* If this packet has been sent out already, we must
	 * adjust the various packet counters.
	 */
890
	if (!before(tp->snd_nxt, TCP_SKB_CB(buff)->end_seq)) {
891 892
		int diff = old_factor - tcp_skb_pcount(skb) -
			tcp_skb_pcount(buff);
L
Linus Torvalds 已提交
893

894 895
		if (diff)
			tcp_adjust_pcount(sk, skb, diff);
L
Linus Torvalds 已提交
896 897 898
	}

	/* Link BUFF into the send queue. */
899
	skb_header_release(buff);
900
	tcp_insert_write_queue_after(skb, buff, sk);
L
Linus Torvalds 已提交
901 902 903 904 905 906 907 908

	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.
 */
909
static void __pskb_trim_head(struct sk_buff *skb, int len)
L
Linus Torvalds 已提交
910 911 912 913 914
{
	int i, k, eat;

	eat = len;
	k = 0;
915
	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
L
Linus Torvalds 已提交
916 917 918 919 920 921 922 923 924 925 926 927 928 929 930
		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;

931
	skb_reset_tail_pointer(skb);
L
Linus Torvalds 已提交
932 933 934 935
	skb->data_len -= len;
	skb->len = skb->data_len;
}

936
/* Remove acked data from a packet in the transmit queue. */
L
Linus Torvalds 已提交
937 938
int tcp_trim_head(struct sock *sk, struct sk_buff *skb, u32 len)
{
939
	if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
L
Linus Torvalds 已提交
940 941
		return -ENOMEM;

942 943
	/* If len == headlen, we avoid __skb_pull to preserve alignment. */
	if (unlikely(len < skb_headlen(skb)))
L
Linus Torvalds 已提交
944
		__skb_pull(skb, len);
945 946
	else
		__pskb_trim_head(skb, len - skb_headlen(skb));
L
Linus Torvalds 已提交
947 948

	TCP_SKB_CB(skb)->seq += len;
949
	skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
950 951 952

	skb->truesize	     -= len;
	sk->sk_wmem_queued   -= len;
953
	sk_mem_uncharge(sk, len);
L
Linus Torvalds 已提交
954 955 956 957 958 959
	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)
I
Ilpo Järvinen 已提交
960
		tcp_set_skb_tso_segs(sk, skb, tcp_current_mss(sk));
L
Linus Torvalds 已提交
961 962 963 964

	return 0;
}

965
/* Calculate MSS. Not accounting for SACKs here.  */
J
John Heffner 已提交
966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
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;
}

1009
/* MTU probing init per socket */
J
John Heffner 已提交
1010 1011 1012 1013 1014 1015 1016
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) +
1017
			       icsk->icsk_af_ops->net_header_len;
J
John Heffner 已提交
1018 1019 1020 1021
	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
1022 1023 1024 1025 1026 1027
/* 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
1028
   It is minimum of user_mss and mss received with SYN.
L
Linus Torvalds 已提交
1029 1030
   It also does not include TCP options.

1031
   inet_csk(sk)->icsk_pmtu_cookie is last pmtu, seen by this function.
L
Linus Torvalds 已提交
1032 1033 1034 1035 1036 1037 1038 1039 1040

   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.

1041 1042
   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
1043 1044 1045 1046
 */
unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu)
{
	struct tcp_sock *tp = tcp_sk(sk);
1047
	struct inet_connection_sock *icsk = inet_csk(sk);
J
John Heffner 已提交
1048
	int mss_now;
L
Linus Torvalds 已提交
1049

J
John Heffner 已提交
1050 1051
	if (icsk->icsk_mtup.search_high > pmtu)
		icsk->icsk_mtup.search_high = pmtu;
L
Linus Torvalds 已提交
1052

J
John Heffner 已提交
1053
	mss_now = tcp_mtu_to_mss(sk, pmtu);
1054
	mss_now = tcp_bound_to_half_wnd(tp, mss_now);
L
Linus Torvalds 已提交
1055 1056

	/* And store cached results */
1057
	icsk->icsk_pmtu_cookie = pmtu;
J
John Heffner 已提交
1058 1059
	if (icsk->icsk_mtup.enabled)
		mss_now = min(mss_now, tcp_mtu_to_mss(sk, icsk->icsk_mtup.search_low));
1060
	tp->mss_cache = mss_now;
L
Linus Torvalds 已提交
1061 1062 1063 1064 1065 1066 1067

	return mss_now;
}

/* Compute the current effective MSS, taking SACKs and IP options,
 * and even PMTU discovery events into account.
 */
I
Ilpo Järvinen 已提交
1068
unsigned int tcp_current_mss(struct sock *sk)
L
Linus Torvalds 已提交
1069 1070 1071
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct dst_entry *dst = __sk_dst_get(sk);
1072
	u32 mss_now;
A
Adam Langley 已提交
1073 1074 1075
	unsigned header_len;
	struct tcp_out_options opts;
	struct tcp_md5sig_key *md5;
1076 1077 1078

	mss_now = tp->mss_cache;

L
Linus Torvalds 已提交
1079 1080
	if (dst) {
		u32 mtu = dst_mtu(dst);
1081
		if (mtu != inet_csk(sk)->icsk_pmtu_cookie)
L
Linus Torvalds 已提交
1082 1083 1084
			mss_now = tcp_sync_mss(sk, mtu);
	}

A
Adam Langley 已提交
1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
	header_len = tcp_established_options(sk, NULL, &opts, &md5) +
		     sizeof(struct tcphdr);
	/* The mss_cache is sized based on tp->tcp_header_len, which assumes
	 * some common options. If this is an odd packet (because we have SACK
	 * blocks etc) then our calculated header_len will be different, and
	 * we have to adjust mss_now correspondingly */
	if (header_len != tp->tcp_header_len) {
		int delta = (int) header_len - tp->tcp_header_len;
		mss_now -= delta;
	}
1095

L
Linus Torvalds 已提交
1096 1097 1098
	return mss_now;
}

1099
/* Congestion window validation. (RFC2861) */
1100
static void tcp_cwnd_validate(struct sock *sk)
1101
{
1102
	struct tcp_sock *tp = tcp_sk(sk);
1103

1104
	if (tp->packets_out >= tp->snd_cwnd) {
1105 1106 1107 1108 1109 1110 1111 1112
		/* 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;

1113 1114
		if (sysctl_tcp_slow_start_after_idle &&
		    (s32)(tcp_time_stamp - tp->snd_cwnd_stamp) >= inet_csk(sk)->icsk_rto)
1115 1116 1117 1118
			tcp_cwnd_application_limited(sk);
	}
}

1119 1120 1121 1122
/* Returns the portion of skb which can be sent right away without
 * introducing MSS oddities to segment boundaries. In rare cases where
 * mss_now != mss_cache, we will request caller to create a small skb
 * per input skb which could be mostly avoided here (if desired).
1123 1124 1125 1126 1127 1128 1129
 *
 * We explicitly want to create a request for splitting write queue tail
 * to a small skb for Nagle purposes while avoiding unnecessary modulos,
 * thus all the complexity (cwnd_len is always MSS multiple which we
 * return whenever allowed by the other factors). Basically we need the
 * modulo only when the receiver window alone is the limiting factor or
 * when we would be allowed to send the split-due-to-Nagle skb fully.
1130 1131
 */
static unsigned int tcp_mss_split_point(struct sock *sk, struct sk_buff *skb,
1132
					unsigned int mss_now, unsigned int cwnd)
1133
{
1134 1135
	struct tcp_sock *tp = tcp_sk(sk);
	u32 needed, window, cwnd_len;
1136

1137
	window = tcp_wnd_end(tp) - TCP_SKB_CB(skb)->seq;
1138
	cwnd_len = mss_now * cwnd;
1139 1140 1141 1142

	if (likely(cwnd_len <= window && skb != tcp_write_queue_tail(sk)))
		return cwnd_len;

1143 1144
	needed = min(skb->len, window);

1145
	if (cwnd_len <= needed)
1146 1147 1148
		return cwnd_len;

	return needed - needed % mss_now;
1149 1150 1151 1152 1153
}

/* 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.
 */
1154 1155
static inline unsigned int tcp_cwnd_test(struct tcp_sock *tp,
					 struct sk_buff *skb)
1156 1157 1158 1159
{
	u32 in_flight, cwnd;

	/* Don't be strict about the congestion window for the final FIN.  */
1160 1161
	if ((TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN) &&
	    tcp_skb_pcount(skb) == 1)
1162 1163 1164 1165 1166 1167 1168 1169 1170 1171
		return 1;

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

	return 0;
}

1172 1173
/* Intialize TSO state of a skb.
 * This must be invoked the first time we consider transmitting
1174 1175
 * SKB onto the wire.
 */
1176 1177
static int tcp_init_tso_segs(struct sock *sk, struct sk_buff *skb,
			     unsigned int mss_now)
1178 1179 1180
{
	int tso_segs = tcp_skb_pcount(skb);

1181
	if (!tso_segs || (tso_segs > 1 && tcp_skb_mss(skb) != mss_now)) {
1182
		tcp_set_skb_tso_segs(sk, skb, mss_now);
1183 1184 1185 1186 1187
		tso_segs = tcp_skb_pcount(skb);
	}
	return tso_segs;
}

1188
/* Minshall's variant of the Nagle send check. */
1189 1190
static inline int tcp_minshall_check(const struct tcp_sock *tp)
{
1191
	return after(tp->snd_sml, tp->snd_una) &&
1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
		!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,
1203
				  const struct sk_buff *skb,
1204 1205 1206
				  unsigned mss_now, int nonagle)
{
	return (skb->len < mss_now &&
1207 1208
		((nonagle & TCP_NAGLE_CORK) ||
		 (!nonagle && tp->packets_out && tcp_minshall_check(tp))));
1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
}

/* 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;

1226 1227 1228
	/* Don't use the nagle rule for urgent data (or for the final FIN).
	 * Nagle can be ignored during F-RTO too (see RFC4138).
	 */
I
Ilpo Järvinen 已提交
1229
	if (tcp_urg_mode(tp) || (tp->frto_counter == 2) ||
1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
	    (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? */
1240 1241
static inline int tcp_snd_wnd_test(struct tcp_sock *tp, struct sk_buff *skb,
				   unsigned int cur_mss)
1242 1243 1244 1245 1246 1247
{
	u32 end_seq = TCP_SKB_CB(skb)->end_seq;

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

1248
	return !after(end_seq, tcp_wnd_end(tp));
1249 1250
}

1251
/* This checks if the data bearing packet SKB (usually tcp_send_head(sk))
1252 1253 1254 1255 1256 1257 1258 1259 1260
 * 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;

1261
	tcp_init_tso_segs(sk, skb, cur_mss);
1262 1263 1264 1265 1266

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

	cwnd_quota = tcp_cwnd_test(tp, skb);
1267
	if (cwnd_quota && !tcp_snd_wnd_test(tp, skb, cur_mss))
1268 1269 1270 1271 1272
		cwnd_quota = 0;

	return cwnd_quota;
}

1273
/* Test if sending is allowed right now. */
1274
int tcp_may_send_now(struct sock *sk)
1275
{
1276
	struct tcp_sock *tp = tcp_sk(sk);
1277
	struct sk_buff *skb = tcp_send_head(sk);
1278 1279

	return (skb &&
I
Ilpo Järvinen 已提交
1280
		tcp_snd_test(sk, skb, tcp_current_mss(sk),
1281
			     (tcp_skb_is_last(sk, skb) ?
1282
			      tp->nonagle : TCP_NAGLE_PUSH)));
1283 1284 1285 1286 1287 1288 1289 1290 1291
}

/* 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).
 */
1292 1293
static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len,
			unsigned int mss_now)
1294 1295 1296
{
	struct sk_buff *buff;
	int nlen = skb->len - len;
1297
	u8 flags;
1298 1299

	/* All of a TSO frame must be composed of paged data.  */
1300 1301
	if (skb->len != skb->data_len)
		return tcp_fragment(sk, skb, len, mss_now);
1302

1303
	buff = sk_stream_alloc_skb(sk, 0, GFP_ATOMIC);
1304 1305 1306
	if (unlikely(buff == NULL))
		return -ENOMEM;

1307 1308
	sk->sk_wmem_queued += buff->truesize;
	sk_mem_charge(sk, buff->truesize);
1309
	buff->truesize += nlen;
1310 1311 1312 1313 1314 1315 1316 1317 1318
	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;
1319
	TCP_SKB_CB(skb)->flags = flags & ~(TCPCB_FLAG_FIN | TCPCB_FLAG_PSH);
1320 1321 1322 1323 1324
	TCP_SKB_CB(buff)->flags = flags;

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

1325
	buff->ip_summed = skb->ip_summed = CHECKSUM_PARTIAL;
1326 1327 1328
	skb_split(skb, buff, len);

	/* Fix up tso_factor for both original and new SKB.  */
1329 1330
	tcp_set_skb_tso_segs(sk, skb, mss_now);
	tcp_set_skb_tso_segs(sk, buff, mss_now);
1331 1332 1333

	/* Link BUFF into the send queue. */
	skb_header_release(buff);
1334
	tcp_insert_write_queue_after(skb, buff, sk);
1335 1336 1337 1338 1339 1340 1341 1342 1343

	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.
 */
1344
static int tcp_tso_should_defer(struct sock *sk, struct sk_buff *skb)
1345
{
1346
	struct tcp_sock *tp = tcp_sk(sk);
1347
	const struct inet_connection_sock *icsk = inet_csk(sk);
1348 1349 1350
	u32 send_win, cong_win, limit, in_flight;

	if (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN)
J
John Heffner 已提交
1351
		goto send_now;
1352

1353
	if (icsk->icsk_ca_state != TCP_CA_Open)
J
John Heffner 已提交
1354 1355 1356
		goto send_now;

	/* Defer for less than two clock ticks. */
I
Ilpo Järvinen 已提交
1357
	if (tp->tso_deferred &&
1358
	    (((u32)jiffies << 1) >> 1) - (tp->tso_deferred >> 1) > 1)
J
John Heffner 已提交
1359
		goto send_now;
1360

1361 1362
	in_flight = tcp_packets_in_flight(tp);

1363
	BUG_ON(tcp_skb_pcount(skb) <= 1 || (tp->snd_cwnd <= in_flight));
1364

1365
	send_win = tcp_wnd_end(tp) - TCP_SKB_CB(skb)->seq;
1366 1367 1368 1369 1370 1371

	/* 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);

1372
	/* If a full-sized TSO skb can be sent, do it. */
1373
	if (limit >= sk->sk_gso_max_size)
J
John Heffner 已提交
1374
		goto send_now;
1375

1376 1377 1378 1379
	/* Middle in queue won't get any more data, full sendable already? */
	if ((skb != tcp_write_queue_tail(sk)) && (limit >= skb->len))
		goto send_now;

1380 1381 1382 1383 1384 1385 1386 1387
	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
1388
			goto send_now;
1389 1390 1391 1392 1393 1394 1395
	} 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
1396
			goto send_now;
1397 1398 1399
	}

	/* Ok, it looks like it is advisable to defer.  */
1400
	tp->tso_deferred = 1 | (jiffies << 1);
J
John Heffner 已提交
1401

1402
	return 1;
J
John Heffner 已提交
1403 1404 1405 1406

send_now:
	tp->tso_deferred = 0;
	return 0;
1407 1408
}

J
John Heffner 已提交
1409
/* Create a new MTU probe if we are ready.
1410 1411 1412 1413
 * MTU probe is regularly attempting to increase the path MTU by
 * deliberately sending larger packets.  This discovers routing
 * changes resulting in larger path MTUs.
 *
J
John Heffner 已提交
1414 1415
 * Returns 0 if we should wait to probe (no cwnd available),
 *         1 if a probe was sent,
1416 1417
 *         -1 otherwise
 */
J
John Heffner 已提交
1418 1419 1420 1421 1422 1423 1424
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;
1425
	int size_needed;
J
John Heffner 已提交
1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
	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 ||
1437
	    tp->rx_opt.num_sacks || tp->rx_opt.dsack)
J
John Heffner 已提交
1438 1439 1440
		return -1;

	/* Very simple search strategy: just double the MSS. */
I
Ilpo Järvinen 已提交
1441
	mss_now = tcp_current_mss(sk);
1442
	probe_size = 2 * tp->mss_cache;
1443
	size_needed = probe_size + (tp->reordering + 1) * tp->mss_cache;
J
John Heffner 已提交
1444 1445 1446 1447 1448 1449
	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? */
1450
	if (tp->write_seq - tp->snd_nxt < size_needed)
J
John Heffner 已提交
1451 1452
		return -1;

1453 1454
	if (tp->snd_wnd < size_needed)
		return -1;
1455
	if (after(tp->snd_nxt + size_needed, tcp_wnd_end(tp)))
1456
		return 0;
J
John Heffner 已提交
1457

1458 1459 1460
	/* Do we need to wait to drain cwnd? With none in flight, don't stall */
	if (tcp_packets_in_flight(tp) + 2 > tp->snd_cwnd) {
		if (!tcp_packets_in_flight(tp))
J
John Heffner 已提交
1461 1462 1463 1464 1465 1466 1467 1468
			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;
1469 1470
	sk->sk_wmem_queued += nskb->truesize;
	sk_mem_charge(sk, nskb->truesize);
J
John Heffner 已提交
1471

1472
	skb = tcp_send_head(sk);
J
John Heffner 已提交
1473 1474 1475 1476 1477 1478

	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;
1479
	nskb->ip_summed = skb->ip_summed;
J
John Heffner 已提交
1480

1481 1482
	tcp_insert_write_queue_before(nskb, skb, sk);

J
John Heffner 已提交
1483
	len = 0;
1484
	tcp_for_write_queue_from_safe(skb, next, sk) {
J
John Heffner 已提交
1485 1486 1487 1488 1489
		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,
1490 1491
							    skb_put(nskb, copy),
							    copy, nskb->csum);
J
John Heffner 已提交
1492 1493 1494 1495 1496

		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;
1497
			tcp_unlink_write_queue(skb, sk);
1498
			sk_wmem_free_skb(sk, skb);
J
John Heffner 已提交
1499 1500
		} else {
			TCP_SKB_CB(nskb)->flags |= TCP_SKB_CB(skb)->flags &
1501
						   ~(TCPCB_FLAG_FIN|TCPCB_FLAG_PSH);
J
John Heffner 已提交
1502 1503
			if (!skb_shinfo(skb)->nr_frags) {
				skb_pull(skb, copy);
1504
				if (skb->ip_summed != CHECKSUM_PARTIAL)
1505 1506
					skb->csum = csum_partial(skb->data,
								 skb->len, 0);
J
John Heffner 已提交
1507 1508 1509 1510 1511 1512 1513 1514
			} else {
				__pskb_trim_head(skb, copy);
				tcp_set_skb_tso_segs(sk, skb, mss_now);
			}
			TCP_SKB_CB(skb)->seq += copy;
		}

		len += copy;
1515 1516 1517

		if (len >= probe_size)
			break;
J
John Heffner 已提交
1518 1519 1520 1521 1522 1523 1524 1525
	}
	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
1526
		 * effectively two packets. */
J
John Heffner 已提交
1527
		tp->snd_cwnd--;
1528
		tcp_event_new_data_sent(sk, nskb);
J
John Heffner 已提交
1529 1530

		icsk->icsk_mtup.probe_size = tcp_mss_to_mtu(sk, nskb->len);
1531 1532
		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
1533 1534 1535 1536 1537 1538 1539

		return 1;
	}

	return -1;
}

L
Linus Torvalds 已提交
1540 1541 1542 1543
/* This routine writes packets to the network.  It advances the
 * send_head.  This happens as incoming acks open up the remote
 * window for us.
 *
1544 1545 1546 1547
 * LARGESEND note: !tcp_urg_mode is overkill, only frames between
 * snd_up-64k-mss .. snd_up cannot be large. However, taking into
 * account rare use of URG, this is not a big flaw.
 *
L
Linus Torvalds 已提交
1548 1549 1550
 * Returns 1, if no segments are in flight and we have queued segments, but
 * cannot send anything now because of SWS or another problem.
 */
1551 1552
static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle,
			  int push_one, gfp_t gfp)
L
Linus Torvalds 已提交
1553 1554
{
	struct tcp_sock *tp = tcp_sk(sk);
1555
	struct sk_buff *skb;
1556 1557
	unsigned int tso_segs, sent_pkts;
	int cwnd_quota;
J
John Heffner 已提交
1558
	int result;
L
Linus Torvalds 已提交
1559

1560
	sent_pkts = 0;
J
John Heffner 已提交
1561

1562 1563 1564 1565 1566 1567 1568 1569
	if (!push_one) {
		/* Do MTU probing. */
		result = tcp_mtu_probe(sk);
		if (!result) {
			return 0;
		} else if (result > 0) {
			sent_pkts = 1;
		}
J
John Heffner 已提交
1570 1571
	}

1572
	while ((skb = tcp_send_head(sk))) {
1573 1574
		unsigned int limit;

1575
		tso_segs = tcp_init_tso_segs(sk, skb, mss_now);
1576
		BUG_ON(!tso_segs);
1577

1578 1579 1580 1581 1582 1583 1584
		cwnd_quota = tcp_cwnd_test(tp, skb);
		if (!cwnd_quota)
			break;

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

1585 1586 1587 1588 1589 1590
		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 {
1591
			if (!push_one && tcp_tso_should_defer(sk, skb))
1592 1593
				break;
		}
1594

1595
		limit = mss_now;
1596
		if (tso_segs > 1 && !tcp_urg_mode(tp))
1597 1598
			limit = tcp_mss_split_point(sk, skb, mss_now,
						    cwnd_quota);
L
Linus Torvalds 已提交
1599

1600 1601 1602 1603
		if (skb->len > limit &&
		    unlikely(tso_fragment(sk, skb, limit, mss_now)))
			break;

1604
		TCP_SKB_CB(skb)->when = tcp_time_stamp;
1605

1606
		if (unlikely(tcp_transmit_skb(sk, skb, 1, gfp)))
1607
			break;
L
Linus Torvalds 已提交
1608

1609 1610 1611
		/* Advance the send_head.  This one is sent out.
		 * This call will increment packets_out.
		 */
1612
		tcp_event_new_data_sent(sk, skb);
L
Linus Torvalds 已提交
1613

1614
		tcp_minshall_update(tp, mss_now, skb);
1615
		sent_pkts++;
1616 1617 1618

		if (push_one)
			break;
1619
	}
L
Linus Torvalds 已提交
1620

1621
	if (likely(sent_pkts)) {
1622
		tcp_cwnd_validate(sk);
1623
		return 0;
L
Linus Torvalds 已提交
1624
	}
1625
	return !tp->packets_out && tcp_send_head(sk);
L
Linus Torvalds 已提交
1626 1627
}

1628 1629 1630 1631
/* 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.
 */
1632 1633
void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
			       int nonagle)
1634
{
1635
	struct sk_buff *skb = tcp_send_head(sk);
1636

1637 1638 1639 1640 1641 1642 1643 1644 1645 1646
	if (!skb)
		return;

	/* 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.
	 */
	if (unlikely(sk->sk_state == TCP_CLOSE))
		return;

1647
	if (tcp_write_xmit(sk, cur_mss, nonagle, 0, GFP_ATOMIC))
1648
		tcp_check_probe_timer(sk);
1649 1650
}

1651 1652 1653 1654 1655
/* 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)
{
1656
	struct sk_buff *skb = tcp_send_head(sk);
1657 1658 1659

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

1660
	tcp_write_xmit(sk, mss_now, TCP_NAGLE_PUSH, 1, sk->sk_allocation);
1661 1662
}

L
Linus Torvalds 已提交
1663 1664
/* This function returns the amount that we can raise the
 * usable window based on the following constraints
1665
 *
L
Linus Torvalds 已提交
1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683
 * 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.
1684
 *
L
Linus Torvalds 已提交
1685 1686
 * Of course, if the sender implements sender side SWS prevention
 * then this will not be a problem.
1687
 *
L
Linus Torvalds 已提交
1688
 * BSD seems to make the following compromise:
1689
 *
L
Linus Torvalds 已提交
1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
 *	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)
{
1717
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1718
	struct tcp_sock *tp = tcp_sk(sk);
S
Stephen Hemminger 已提交
1719
	/* MSS for the peer's data.  Previous versions used mss_clamp
L
Linus Torvalds 已提交
1720 1721 1722 1723 1724
	 * 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
	 */
1725
	int mss = icsk->icsk_ack.rcv_mss;
L
Linus Torvalds 已提交
1726 1727 1728 1729 1730
	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)
1731
		mss = full_space;
L
Linus Torvalds 已提交
1732

1733
	if (free_space < (full_space >> 1)) {
1734
		icsk->icsk_ack.quick = 0;
L
Linus Torvalds 已提交
1735 1736

		if (tcp_memory_pressure)
1737 1738
			tp->rcv_ssthresh = min(tp->rcv_ssthresh,
					       4U * tp->advmss);
L
Linus Torvalds 已提交
1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770

		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)
1771
			window = (free_space / mss) * mss;
1772
		else if (mss == full_space &&
1773
			 free_space > window + (full_space >> 1))
1774
			window = free_space;
L
Linus Torvalds 已提交
1775 1776 1777 1778 1779
	}

	return window;
}

1780 1781
/* Collapses two adjacent SKB's during retransmission. */
static void tcp_collapse_retrans(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
1782 1783
{
	struct tcp_sock *tp = tcp_sk(sk);
1784
	struct sk_buff *next_skb = tcp_write_queue_next(sk, skb);
1785
	int skb_size, next_skb_size;
L
Linus Torvalds 已提交
1786

1787 1788
	skb_size = skb->len;
	next_skb_size = next_skb->len;
L
Linus Torvalds 已提交
1789

1790
	BUG_ON(tcp_skb_pcount(skb) != 1 || tcp_skb_pcount(next_skb) != 1);
1791

1792
	tcp_highest_sack_combine(sk, next_skb, skb);
L
Linus Torvalds 已提交
1793

1794
	tcp_unlink_write_queue(next_skb, sk);
L
Linus Torvalds 已提交
1795

1796 1797
	skb_copy_from_linear_data(next_skb, skb_put(skb, next_skb_size),
				  next_skb_size);
L
Linus Torvalds 已提交
1798

1799 1800
	if (next_skb->ip_summed == CHECKSUM_PARTIAL)
		skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
1801

1802 1803
	if (skb->ip_summed != CHECKSUM_PARTIAL)
		skb->csum = csum_block_add(skb->csum, next_skb->csum, skb_size);
L
Linus Torvalds 已提交
1804

1805 1806
	/* Update sequence range on original skb. */
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(next_skb)->end_seq;
L
Linus Torvalds 已提交
1807

1808 1809
	/* Merge over control information. This moves PSH/FIN etc. over */
	TCP_SKB_CB(skb)->flags |= TCP_SKB_CB(next_skb)->flags;
1810 1811 1812 1813 1814 1815 1816

	/* 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;

	/* changed transmit queue under us so clear hints */
1817 1818 1819
	tcp_clear_retrans_hints_partial(tp);
	if (next_skb == tp->retransmit_skb_hint)
		tp->retransmit_skb_hint = skb;
1820

1821 1822
	tcp_adjust_pcount(sk, next_skb, tcp_skb_pcount(next_skb));

1823
	sk_wmem_free_skb(sk, next_skb);
L
Linus Torvalds 已提交
1824 1825
}

1826
/* Check if coalescing SKBs is legal. */
1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844
static int tcp_can_collapse(struct sock *sk, struct sk_buff *skb)
{
	if (tcp_skb_pcount(skb) > 1)
		return 0;
	/* TODO: SACK collapsing could be used to remove this condition */
	if (skb_shinfo(skb)->nr_frags != 0)
		return 0;
	if (skb_cloned(skb))
		return 0;
	if (skb == tcp_send_head(sk))
		return 0;
	/* Some heurestics for collapsing over SACK'd could be invented */
	if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)
		return 0;

	return 1;
}

1845 1846 1847
/* Collapse packets in the retransmit queue to make to create
 * less packets on the wire. This is only done on retransmission.
 */
1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885
static void tcp_retrans_try_collapse(struct sock *sk, struct sk_buff *to,
				     int space)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb = to, *tmp;
	int first = 1;

	if (!sysctl_tcp_retrans_collapse)
		return;
	if (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_SYN)
		return;

	tcp_for_write_queue_from_safe(skb, tmp, sk) {
		if (!tcp_can_collapse(sk, skb))
			break;

		space -= skb->len;

		if (first) {
			first = 0;
			continue;
		}

		if (space < 0)
			break;
		/* Punt if not enough space exists in the first SKB for
		 * the data in the second
		 */
		if (skb->len > skb_tailroom(to))
			break;

		if (after(TCP_SKB_CB(skb)->end_seq, tcp_wnd_end(tp)))
			break;

		tcp_collapse_retrans(sk, to);
	}
}

L
Linus Torvalds 已提交
1886 1887 1888 1889 1890 1891 1892
/* 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
1893
	struct inet_connection_sock *icsk = inet_csk(sk);
1894
	unsigned int cur_mss;
L
Linus Torvalds 已提交
1895 1896
	int err;

J
John Heffner 已提交
1897 1898 1899 1900 1901
	/* Inconslusive MTU probe */
	if (icsk->icsk_mtup.probe_size) {
		icsk->icsk_mtup.probe_size = 0;
	}

L
Linus Torvalds 已提交
1902
	/* Do not sent more than we queued. 1/4 is reserved for possible
S
Stephen Hemminger 已提交
1903
	 * copying overhead: fragmentation, tunneling, mangling etc.
L
Linus Torvalds 已提交
1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915
	 */
	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;
	}

1916 1917 1918
	if (inet_csk(sk)->icsk_af_ops->rebuild_header(sk))
		return -EHOSTUNREACH; /* Routing failure or similar. */

I
Ilpo Järvinen 已提交
1919
	cur_mss = tcp_current_mss(sk);
1920

L
Linus Torvalds 已提交
1921 1922 1923 1924 1925
	/* 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.
	 */
1926 1927
	if (!before(TCP_SKB_CB(skb)->seq, tcp_wnd_end(tp)) &&
	    TCP_SKB_CB(skb)->seq != tp->snd_una)
L
Linus Torvalds 已提交
1928 1929 1930
		return -EAGAIN;

	if (skb->len > cur_mss) {
1931
		if (tcp_fragment(sk, skb, cur_mss, cur_mss))
L
Linus Torvalds 已提交
1932
			return -ENOMEM; /* We'll try again later. */
1933
	} else {
1934 1935 1936 1937 1938 1939
		int oldpcount = tcp_skb_pcount(skb);

		if (unlikely(oldpcount > 1)) {
			tcp_init_tso_segs(sk, skb, cur_mss);
			tcp_adjust_pcount(sk, skb, oldpcount - tcp_skb_pcount(skb));
		}
L
Linus Torvalds 已提交
1940 1941
	}

1942
	tcp_retrans_try_collapse(sk, skb, cur_mss);
L
Linus Torvalds 已提交
1943 1944 1945 1946 1947

	/* 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
1948 1949 1950
	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
1951
		if (!pskb_trim(skb, 0)) {
1952 1953 1954
			/* Reuse, even though it does some unnecessary work */
			tcp_init_nondata_skb(skb, TCP_SKB_CB(skb)->end_seq - 1,
					     TCP_SKB_CB(skb)->flags);
L
Linus Torvalds 已提交
1955 1956 1957 1958 1959 1960 1961 1962 1963
			skb->ip_summed = CHECKSUM_NONE;
		}
	}

	/* 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;

1964
	err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
1965 1966 1967

	if (err == 0) {
		/* Update global TCP statistics. */
P
Pavel Emelyanov 已提交
1968
		TCP_INC_STATS(sock_net(sk), TCP_MIB_RETRANSSEGS);
L
Linus Torvalds 已提交
1969 1970 1971 1972

		tp->total_retrans++;

#if FASTRETRANS_DEBUG > 0
1973
		if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS) {
L
Linus Torvalds 已提交
1974 1975 1976 1977
			if (net_ratelimit())
				printk(KERN_DEBUG "retrans_out leaked.\n");
		}
#endif
1978 1979
		if (!tp->retrans_out)
			tp->lost_retrans_low = tp->snd_nxt;
L
Linus Torvalds 已提交
1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
		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;
}

1997 1998 1999
/* Check if we forward retransmits are possible in the current
 * window/congestion state.
 */
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026
static int tcp_can_forward_retransmit(struct sock *sk)
{
	const struct inet_connection_sock *icsk = inet_csk(sk);
	struct tcp_sock *tp = tcp_sk(sk);

	/* Forward retransmissions are possible only during Recovery. */
	if (icsk->icsk_ca_state != TCP_CA_Recovery)
		return 0;

	/* No forward retransmissions in Reno are possible. */
	if (tcp_is_reno(tp))
		return 0;

	/* 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
	 * segments to send. In the other cases, follow rule 3 for
	 * NextSeg() specified in RFC3517.
	 */

	if (tcp_may_send_now(sk))
		return 0;

	return 1;
}

L
Linus Torvalds 已提交
2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
/* 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)
{
2037
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2038 2039
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
2040
	struct sk_buff *hole = NULL;
2041
	u32 last_lost;
2042
	int mib_idx;
2043
	int fwd_rexmitting = 0;
2044

2045 2046 2047
	if (!tp->lost_out)
		tp->retransmit_high = tp->snd_una;

2048
	if (tp->retransmit_skb_hint) {
2049
		skb = tp->retransmit_skb_hint;
2050 2051 2052 2053
		last_lost = TCP_SKB_CB(skb)->end_seq;
		if (after(last_lost, tp->retransmit_high))
			last_lost = tp->retransmit_high;
	} else {
2054
		skb = tcp_write_queue_head(sk);
2055 2056
		last_lost = tp->snd_una;
	}
L
Linus Torvalds 已提交
2057

2058 2059
	tcp_for_write_queue_from(skb, sk) {
		__u8 sacked = TCP_SKB_CB(skb)->sacked;
L
Linus Torvalds 已提交
2060

2061 2062 2063
		if (skb == tcp_send_head(sk))
			break;
		/* we could do better than to assign each time */
2064 2065
		if (hole == NULL)
			tp->retransmit_skb_hint = skb;
2066 2067 2068 2069 2070 2071 2072 2073 2074 2075

		/* 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;
L
Linus Torvalds 已提交
2076

2077 2078 2079 2080 2081
		if (fwd_rexmitting) {
begin_fwd:
			if (!before(TCP_SKB_CB(skb)->seq, tcp_highest_sack_seq(tp)))
				break;
			mib_idx = LINUX_MIB_TCPFORWARDRETRANS;
2082

2083
		} else if (!before(TCP_SKB_CB(skb)->seq, tp->retransmit_high)) {
2084
			tp->retransmit_high = last_lost;
2085 2086 2087 2088 2089 2090 2091 2092 2093
			if (!tcp_can_forward_retransmit(sk))
				break;
			/* Backtrack if necessary to non-L'ed skb */
			if (hole != NULL) {
				skb = hole;
				hole = NULL;
			}
			fwd_rexmitting = 1;
			goto begin_fwd;
L
Linus Torvalds 已提交
2094

2095
		} else if (!(sacked & TCPCB_LOST)) {
2096
			if (hole == NULL && !(sacked & (TCPCB_SACKED_RETRANS|TCPCB_SACKED_ACKED)))
2097 2098
				hole = skb;
			continue;
L
Linus Torvalds 已提交
2099

2100
		} else {
2101
			last_lost = TCP_SKB_CB(skb)->end_seq;
2102 2103 2104 2105 2106
			if (icsk->icsk_ca_state != TCP_CA_Loss)
				mib_idx = LINUX_MIB_TCPFASTRETRANS;
			else
				mib_idx = LINUX_MIB_TCPSLOWSTARTRETRANS;
		}
L
Linus Torvalds 已提交
2107

2108
		if (sacked & (TCPCB_SACKED_ACKED|TCPCB_SACKED_RETRANS))
L
Linus Torvalds 已提交
2109 2110
			continue;

2111
		if (tcp_retransmit_skb(sk, skb))
2112 2113
			return;
		NET_INC_STATS_BH(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
2114

2115
		if (skb == tcp_write_queue_head(sk))
2116 2117 2118
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
						  inet_csk(sk)->icsk_rto,
						  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2119 2120 2121 2122 2123 2124 2125 2126
	}
}

/* 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)
{
2127
	struct tcp_sock *tp = tcp_sk(sk);
2128
	struct sk_buff *skb = tcp_write_queue_tail(sk);
L
Linus Torvalds 已提交
2129
	int mss_now;
2130

L
Linus Torvalds 已提交
2131 2132 2133 2134
	/* Optimization, tack on the FIN if we have a queue of
	 * unsent frames.  But be careful about outgoing SACKS
	 * and IP options.
	 */
I
Ilpo Järvinen 已提交
2135
	mss_now = tcp_current_mss(sk);
L
Linus Torvalds 已提交
2136

2137
	if (tcp_send_head(sk) != NULL) {
L
Linus Torvalds 已提交
2138 2139 2140 2141 2142 2143
		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 (;;) {
2144 2145
			skb = alloc_skb_fclone(MAX_TCP_HEADER,
					       sk->sk_allocation);
L
Linus Torvalds 已提交
2146 2147 2148 2149 2150 2151 2152 2153
			if (skb)
				break;
			yield();
		}

		/* Reserve space for headers and prepare control bits. */
		skb_reserve(skb, MAX_TCP_HEADER);
		/* FIN eats a sequence byte, write_seq advanced by tcp_queue_skb(). */
2154 2155
		tcp_init_nondata_skb(skb, tp->write_seq,
				     TCPCB_FLAG_ACK | TCPCB_FLAG_FIN);
L
Linus Torvalds 已提交
2156 2157
		tcp_queue_skb(sk, skb);
	}
2158
	__tcp_push_pending_frames(sk, mss_now, TCP_NAGLE_OFF);
L
Linus Torvalds 已提交
2159 2160 2161 2162 2163
}

/* 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
2164
 * by RFC 2525, section 2.17.  -DaveM
L
Linus Torvalds 已提交
2165
 */
A
Al Viro 已提交
2166
void tcp_send_active_reset(struct sock *sk, gfp_t priority)
L
Linus Torvalds 已提交
2167 2168 2169 2170 2171 2172
{
	struct sk_buff *skb;

	/* NOTE: No TCP options attached and we never retransmit this. */
	skb = alloc_skb(MAX_TCP_HEADER, priority);
	if (!skb) {
P
Pavel Emelyanov 已提交
2173
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTFAILED);
L
Linus Torvalds 已提交
2174 2175 2176 2177 2178
		return;
	}

	/* Reserve space for headers and prepare control bits. */
	skb_reserve(skb, MAX_TCP_HEADER);
2179 2180
	tcp_init_nondata_skb(skb, tcp_acceptable_seq(sk),
			     TCPCB_FLAG_ACK | TCPCB_FLAG_RST);
L
Linus Torvalds 已提交
2181 2182
	/* Send it off. */
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2183
	if (tcp_transmit_skb(sk, skb, 0, priority))
P
Pavel Emelyanov 已提交
2184
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTFAILED);
2185

P
Pavel Emelyanov 已提交
2186
	TCP_INC_STATS(sock_net(sk), TCP_MIB_OUTRSTS);
L
Linus Torvalds 已提交
2187 2188
}

2189 2190
/* Send a crossed SYN-ACK during socket establishment.
 * WARNING: This routine must only be called when we have already sent
L
Linus Torvalds 已提交
2191 2192 2193 2194 2195 2196
 * 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)
{
2197
	struct sk_buff *skb;
L
Linus Torvalds 已提交
2198

2199
	skb = tcp_write_queue_head(sk);
2200
	if (skb == NULL || !(TCP_SKB_CB(skb)->flags & TCPCB_FLAG_SYN)) {
L
Linus Torvalds 已提交
2201 2202 2203
		printk(KERN_DEBUG "tcp_send_synack: wrong queue state\n");
		return -EFAULT;
	}
2204
	if (!(TCP_SKB_CB(skb)->flags & TCPCB_FLAG_ACK)) {
L
Linus Torvalds 已提交
2205 2206 2207 2208
		if (skb_cloned(skb)) {
			struct sk_buff *nskb = skb_copy(skb, GFP_ATOMIC);
			if (nskb == NULL)
				return -ENOMEM;
2209
			tcp_unlink_write_queue(skb, sk);
L
Linus Torvalds 已提交
2210
			skb_header_release(nskb);
2211
			__tcp_add_write_queue_head(sk, nskb);
2212 2213 2214
			sk_wmem_free_skb(sk, skb);
			sk->sk_wmem_queued += nskb->truesize;
			sk_mem_charge(sk, nskb->truesize);
L
Linus Torvalds 已提交
2215 2216 2217 2218 2219 2220 2221
			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;
2222
	return tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
2223 2224
}

2225
/* Prepare a SYN-ACK. */
2226 2227
struct sk_buff *tcp_make_synack(struct sock *sk, struct dst_entry *dst,
				struct request_sock *req)
L
Linus Torvalds 已提交
2228
{
2229
	struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
2230 2231 2232
	struct tcp_sock *tp = tcp_sk(sk);
	struct tcphdr *th;
	int tcp_header_size;
A
Adam Langley 已提交
2233
	struct tcp_out_options opts;
L
Linus Torvalds 已提交
2234
	struct sk_buff *skb;
2235 2236
	struct tcp_md5sig_key *md5;
	__u8 *md5_hash_location;
2237
	int mss;
L
Linus Torvalds 已提交
2238 2239 2240 2241 2242 2243 2244 2245

	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);

E
Eric Dumazet 已提交
2246
	skb_dst_set(skb, dst_clone(dst));
L
Linus Torvalds 已提交
2247

2248 2249 2250 2251
	mss = dst_metric(dst, RTAX_ADVMSS);
	if (tp->rx_opt.user_mss && tp->rx_opt.user_mss < mss)
		mss = tp->rx_opt.user_mss;

A
Adam Langley 已提交
2252 2253 2254 2255 2256 2257
	if (req->rcv_wnd == 0) { /* ignored for retransmitted syns */
		__u8 rcv_wscale;
		/* 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 */
		tcp_select_initial_window(tcp_full_space(sk),
2258
			mss - (ireq->tstamp_ok ? TCPOLEN_TSTAMP_ALIGNED : 0),
A
Adam Langley 已提交
2259 2260 2261 2262 2263 2264 2265 2266
			&req->rcv_wnd,
			&req->window_clamp,
			ireq->wscale_ok,
			&rcv_wscale);
		ireq->rcv_wscale = rcv_wscale;
	}

	memset(&opts, 0, sizeof(opts));
2267 2268 2269 2270 2271
#ifdef CONFIG_SYN_COOKIES
	if (unlikely(req->cookie_ts))
		TCP_SKB_CB(skb)->when = cookie_init_timestamp(req);
	else
#endif
A
Adam Langley 已提交
2272
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2273
	tcp_header_size = tcp_synack_options(sk, req, mss,
A
Adam Langley 已提交
2274 2275
					     skb, &opts, &md5) +
			  sizeof(struct tcphdr);
2276

2277 2278
	skb_push(skb, tcp_header_size);
	skb_reset_transport_header(skb);
L
Linus Torvalds 已提交
2279

2280
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
2281 2282 2283 2284
	memset(th, 0, sizeof(struct tcphdr));
	th->syn = 1;
	th->ack = 1;
	TCP_ECN_make_synack(req, th);
2285
	th->source = ireq->loc_port;
2286
	th->dest = ireq->rmt_port;
2287 2288 2289 2290 2291
	/* Setting of flags are superfluous here for callers (and ECE is
	 * not even correctly set)
	 */
	tcp_init_nondata_skb(skb, tcp_rsk(req)->snt_isn,
			     TCPCB_FLAG_SYN | TCPCB_FLAG_ACK);
L
Linus Torvalds 已提交
2292
	th->seq = htonl(TCP_SKB_CB(skb)->seq);
2293
	th->ack_seq = htonl(tcp_rsk(req)->rcv_isn + 1);
L
Linus Torvalds 已提交
2294 2295

	/* RFC1323: The window in SYN & SYN/ACK segments is never scaled. */
2296
	th->window = htons(min(req->rcv_wnd, 65535U));
A
Adam Langley 已提交
2297
	tcp_options_write((__be32 *)(th + 1), tp, &opts, &md5_hash_location);
L
Linus Torvalds 已提交
2298
	th->doff = (tcp_header_size >> 2);
P
Pavel Emelyanov 已提交
2299
	TCP_INC_STATS(sock_net(sk), TCP_MIB_OUTSEGS);
2300 2301 2302 2303

#ifdef CONFIG_TCP_MD5SIG
	/* Okay, we have all we need - do the md5 hash if needed */
	if (md5) {
2304
		tcp_rsk(req)->af_specific->calc_md5_hash(md5_hash_location,
2305
					       md5, NULL, req, skb);
2306 2307 2308
	}
#endif

L
Linus Torvalds 已提交
2309 2310 2311
	return skb;
}

2312
/* Do all connect socket setups that can be done AF independent. */
S
Stephen Hemminger 已提交
2313
static void tcp_connect_init(struct sock *sk)
L
Linus Torvalds 已提交
2314 2315 2316 2317 2318 2319 2320 2321 2322
{
	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) +
2323 2324 2325
		(sysctl_tcp_timestamps &&
		(!dst_feature(dst, RTAX_FEATURE_NO_TSTAMP) ?
		  TCPOLEN_TSTAMP_ALIGNED : 0));
L
Linus Torvalds 已提交
2326

2327 2328 2329 2330 2331
#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
2332 2333 2334 2335
	/* 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
2336
	tcp_mtup_init(sk);
L
Linus Torvalds 已提交
2337 2338 2339 2340 2341
	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);
2342 2343 2344
	if (tp->rx_opt.user_mss && tp->rx_opt.user_mss < tp->advmss)
		tp->advmss = tp->rx_opt.user_mss;

L
Linus Torvalds 已提交
2345 2346 2347 2348 2349 2350
	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,
2351 2352
				  (sysctl_tcp_window_scaling &&
				   !dst_feature(dst, RTAX_FEATURE_NO_WSCALE)),
L
Linus Torvalds 已提交
2353 2354 2355 2356 2357 2358 2359 2360
				  &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;
2361
	tcp_init_wl(tp, 0);
L
Linus Torvalds 已提交
2362 2363
	tp->snd_una = tp->write_seq;
	tp->snd_sml = tp->write_seq;
I
Ilpo Järvinen 已提交
2364
	tp->snd_up = tp->write_seq;
L
Linus Torvalds 已提交
2365 2366 2367 2368
	tp->rcv_nxt = 0;
	tp->rcv_wup = 0;
	tp->copied_seq = 0;

2369 2370
	inet_csk(sk)->icsk_rto = TCP_TIMEOUT_INIT;
	inet_csk(sk)->icsk_retransmits = 0;
L
Linus Torvalds 已提交
2371 2372 2373
	tcp_clear_retrans(tp);
}

2374
/* Build a SYN and send it off. */
L
Linus Torvalds 已提交
2375 2376 2377 2378 2379 2380 2381
int tcp_connect(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *buff;

	tcp_connect_init(sk);

2382
	buff = alloc_skb_fclone(MAX_TCP_HEADER + 15, sk->sk_allocation);
L
Linus Torvalds 已提交
2383 2384 2385 2386 2387 2388
	if (unlikely(buff == NULL))
		return -ENOBUFS;

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

2389
	tp->snd_nxt = tp->write_seq;
2390 2391
	tcp_init_nondata_skb(buff, tp->write_seq++, TCPCB_FLAG_SYN);
	TCP_ECN_send_syn(sk, buff);
L
Linus Torvalds 已提交
2392 2393 2394 2395 2396

	/* Send it off. */
	TCP_SKB_CB(buff)->when = tcp_time_stamp;
	tp->retrans_stamp = TCP_SKB_CB(buff)->when;
	skb_header_release(buff);
2397
	__tcp_add_write_queue_tail(sk, buff);
2398 2399
	sk->sk_wmem_queued += buff->truesize;
	sk_mem_charge(sk, buff->truesize);
L
Linus Torvalds 已提交
2400
	tp->packets_out += tcp_skb_pcount(buff);
2401
	tcp_transmit_skb(sk, buff, 1, sk->sk_allocation);
2402 2403 2404 2405 2406 2407

	/* 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;
P
Pavel Emelyanov 已提交
2408
	TCP_INC_STATS(sock_net(sk), TCP_MIB_ACTIVEOPENS);
L
Linus Torvalds 已提交
2409 2410

	/* Timer for repeating the SYN until an answer. */
2411 2412
	inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
				  inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
L
Linus Torvalds 已提交
2413 2414 2415 2416 2417 2418 2419 2420 2421
	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)
{
2422 2423
	struct inet_connection_sock *icsk = inet_csk(sk);
	int ato = icsk->icsk_ack.ato;
L
Linus Torvalds 已提交
2424 2425 2426
	unsigned long timeout;

	if (ato > TCP_DELACK_MIN) {
2427
		const struct tcp_sock *tp = tcp_sk(sk);
2428
		int max_ato = HZ / 2;
L
Linus Torvalds 已提交
2429

2430 2431
		if (icsk->icsk_ack.pingpong ||
		    (icsk->icsk_ack.pending & ICSK_ACK_PUSHED))
L
Linus Torvalds 已提交
2432 2433 2434 2435 2436
			max_ato = TCP_DELACK_MAX;

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

		/* If some rtt estimate is known, use it to bound delayed ack.
2437
		 * Do not use inet_csk(sk)->icsk_rto here, use results of rtt measurements
L
Linus Torvalds 已提交
2438 2439 2440
		 * directly.
		 */
		if (tp->srtt) {
2441
			int rtt = max(tp->srtt >> 3, TCP_DELACK_MIN);
L
Linus Torvalds 已提交
2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453

			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. */
2454
	if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) {
L
Linus Torvalds 已提交
2455 2456 2457
		/* If delack timer was blocked or is about to expire,
		 * send ACK now.
		 */
2458 2459
		if (icsk->icsk_ack.blocked ||
		    time_before_eq(icsk->icsk_ack.timeout, jiffies + (ato >> 2))) {
L
Linus Torvalds 已提交
2460 2461 2462 2463
			tcp_send_ack(sk);
			return;
		}

2464 2465
		if (!time_before(timeout, icsk->icsk_ack.timeout))
			timeout = icsk->icsk_ack.timeout;
L
Linus Torvalds 已提交
2466
	}
2467 2468 2469
	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
2470 2471 2472 2473 2474
}

/* This routine sends an ack and also updates the window. */
void tcp_send_ack(struct sock *sk)
{
2475
	struct sk_buff *buff;
L
Linus Torvalds 已提交
2476

2477 2478 2479
	/* If we have been reset, we may not send again. */
	if (sk->sk_state == TCP_CLOSE)
		return;
L
Linus Torvalds 已提交
2480

2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491
	/* 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) {
		inet_csk_schedule_ack(sk);
		inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN;
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
					  TCP_DELACK_MAX, TCP_RTO_MAX);
		return;
L
Linus Torvalds 已提交
2492
	}
2493 2494 2495

	/* Reserve space for headers and prepare control bits. */
	skb_reserve(buff, MAX_TCP_HEADER);
2496
	tcp_init_nondata_skb(buff, tcp_acceptable_seq(sk), TCPCB_FLAG_ACK);
2497 2498 2499 2500

	/* Send it off, this clears delayed acks for us. */
	TCP_SKB_CB(buff)->when = tcp_time_stamp;
	tcp_transmit_skb(sk, buff, 0, GFP_ATOMIC);
L
Linus Torvalds 已提交
2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520
}

/* 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);
2521
	if (skb == NULL)
L
Linus Torvalds 已提交
2522 2523 2524 2525 2526 2527 2528 2529
		return -1;

	/* Reserve space for headers and set control bits. */
	skb_reserve(skb, MAX_TCP_HEADER);
	/* Use a previous sequence.  This should cause the other
	 * end to send an ack.  Don't queue or clone SKB, just
	 * send it.
	 */
2530
	tcp_init_nondata_skb(skb, tp->snd_una - !urgent, TCPCB_FLAG_ACK);
L
Linus Torvalds 已提交
2531
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2532
	return tcp_transmit_skb(sk, skb, 0, GFP_ATOMIC);
L
Linus Torvalds 已提交
2533 2534
}

2535
/* Initiate keepalive or window probe from timer. */
L
Linus Torvalds 已提交
2536 2537
int tcp_write_wakeup(struct sock *sk)
{
2538 2539
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
L
Linus Torvalds 已提交
2540

2541 2542 2543 2544 2545 2546
	if (sk->sk_state == TCP_CLOSE)
		return -1;

	if ((skb = tcp_send_head(sk)) != NULL &&
	    before(TCP_SKB_CB(skb)->seq, tcp_wnd_end(tp))) {
		int err;
I
Ilpo Järvinen 已提交
2547
		unsigned int mss = tcp_current_mss(sk);
2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559
		unsigned int seg_size = tcp_wnd_end(tp) - 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);
L
Linus Torvalds 已提交
2560
			TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572
			if (tcp_fragment(sk, skb, seg_size, mss))
				return -1;
		} else if (!tcp_skb_pcount(skb))
			tcp_set_skb_tso_segs(sk, skb, mss);

		TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
		TCP_SKB_CB(skb)->when = tcp_time_stamp;
		err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
		if (!err)
			tcp_event_new_data_sent(sk, skb);
		return err;
	} else {
I
Ilpo Järvinen 已提交
2573
		if (between(tp->snd_up, tp->snd_una + 1, tp->snd_una + 0xFFFF))
2574 2575
			tcp_xmit_probe_skb(sk, 1);
		return tcp_xmit_probe_skb(sk, 0);
L
Linus Torvalds 已提交
2576 2577 2578 2579 2580 2581 2582 2583
	}
}

/* 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)
{
2584
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2585 2586 2587 2588 2589
	struct tcp_sock *tp = tcp_sk(sk);
	int err;

	err = tcp_write_wakeup(sk);

2590
	if (tp->packets_out || !tcp_send_head(sk)) {
L
Linus Torvalds 已提交
2591
		/* Cancel probe timer, if it is not required. */
2592
		icsk->icsk_probes_out = 0;
2593
		icsk->icsk_backoff = 0;
L
Linus Torvalds 已提交
2594 2595 2596 2597
		return;
	}

	if (err <= 0) {
2598 2599
		if (icsk->icsk_backoff < sysctl_tcp_retries2)
			icsk->icsk_backoff++;
2600
		icsk->icsk_probes_out++;
2601
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
2602 2603
					  min(icsk->icsk_rto << icsk->icsk_backoff, TCP_RTO_MAX),
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2604 2605
	} else {
		/* If packet was not sent due to local congestion,
2606
		 * do not backoff and do not remember icsk_probes_out.
L
Linus Torvalds 已提交
2607 2608 2609 2610
		 * Let local senders to fight for local resources.
		 *
		 * Use accumulated backoff yet.
		 */
2611 2612
		if (!icsk->icsk_probes_out)
			icsk->icsk_probes_out = 1;
2613
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
2614
					  min(icsk->icsk_rto << icsk->icsk_backoff,
2615 2616
					      TCP_RESOURCE_PROBE_INTERVAL),
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2617 2618 2619
	}
}

2620
EXPORT_SYMBOL(tcp_select_initial_window);
L
Linus Torvalds 已提交
2621 2622 2623 2624
EXPORT_SYMBOL(tcp_connect);
EXPORT_SYMBOL(tcp_make_synack);
EXPORT_SYMBOL(tcp_simple_retransmit);
EXPORT_SYMBOL(tcp_sync_mss);
J
John Heffner 已提交
2625
EXPORT_SYMBOL(tcp_mtup_init);