tcp_output.c 81.1 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
 *		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>
40
#include <linux/gfp.h>
L
Linus Torvalds 已提交
41 42 43
#include <linux/module.h>

/* People can turn this off for buggy TCP's found in printers etc. */
44
int sysctl_tcp_retrans_collapse __read_mostly = 1;
L
Linus Torvalds 已提交
45

46
/* People can turn this on to work with those rare, broken TCPs that
47 48
 * interpret the window field as a signed quantity.
 */
49
int sysctl_tcp_workaround_signed_windows __read_mostly = 0;
50

L
Linus Torvalds 已提交
51 52 53 54
/* 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.
 */
55
int sysctl_tcp_tso_win_divisor __read_mostly = 3;
L
Linus Torvalds 已提交
56

57 58
int sysctl_tcp_mtu_probing __read_mostly = 0;
int sysctl_tcp_base_mss __read_mostly = 512;
J
John Heffner 已提交
59

60
/* By default, RFC2861 behavior.  */
61
int sysctl_tcp_slow_start_after_idle __read_mostly = 1;
62

63
int sysctl_tcp_cookie_size __read_mostly = 0; /* TCP_COOKIE_MAX */
64
EXPORT_SYMBOL_GPL(sysctl_tcp_cookie_size);
65 66


67
/* Account for new data that has been sent to the network. */
68
static void tcp_event_new_data_sent(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
69
{
70
	struct tcp_sock *tp = tcp_sk(sk);
71
	unsigned int prior_packets = tp->packets_out;
72

73
	tcp_advance_send_head(sk, skb);
L
Linus Torvalds 已提交
74
	tp->snd_nxt = TCP_SKB_CB(skb)->end_seq;
75 76 77 78

	/* Don't override Nagle indefinately with F-RTO */
	if (tp->frto_counter == 2)
		tp->frto_counter = 3;
79 80 81 82 83

	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
84 85 86 87 88 89 90 91
}

/* 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:
 */
92
static inline __u32 tcp_acceptable_seq(struct sock *sk)
L
Linus Torvalds 已提交
93
{
94 95
	struct tcp_sock *tp = tcp_sk(sk);

96
	if (!before(tcp_wnd_end(tp), tp->snd_nxt))
L
Linus Torvalds 已提交
97 98
		return tp->snd_nxt;
	else
99
		return tcp_wnd_end(tp);
L
Linus Torvalds 已提交
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131
}

/* 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. */
132
static void tcp_cwnd_restart(struct sock *sk, struct dst_entry *dst)
L
Linus Torvalds 已提交
133
{
134
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
135 136 137 138
	s32 delta = tcp_time_stamp - tp->lsndtime;
	u32 restart_cwnd = tcp_init_cwnd(tp, dst);
	u32 cwnd = tp->snd_cwnd;

139
	tcp_ca_event(sk, CA_EVENT_CWND_RESTART);
L
Linus Torvalds 已提交
140

141
	tp->snd_ssthresh = tcp_current_ssthresh(sk);
L
Linus Torvalds 已提交
142 143
	restart_cwnd = min(restart_cwnd, cwnd);

144
	while ((delta -= inet_csk(sk)->icsk_rto) > 0 && cwnd > restart_cwnd)
L
Linus Torvalds 已提交
145 146 147 148 149 150
		cwnd >>= 1;
	tp->snd_cwnd = max(cwnd, restart_cwnd);
	tp->snd_cwnd_stamp = tcp_time_stamp;
	tp->snd_cwnd_used = 0;
}

151
/* Congestion state accounting after a packet has been sent. */
S
Stephen Hemminger 已提交
152 153
static void tcp_event_data_sent(struct tcp_sock *tp,
				struct sk_buff *skb, struct sock *sk)
L
Linus Torvalds 已提交
154
{
155 156
	struct inet_connection_sock *icsk = inet_csk(sk);
	const u32 now = tcp_time_stamp;
L
Linus Torvalds 已提交
157

158 159
	if (sysctl_tcp_slow_start_after_idle &&
	    (!tp->packets_out && (s32)(now - tp->lsndtime) > icsk->icsk_rto))
160
		tcp_cwnd_restart(sk, __sk_dst_get(sk));
L
Linus Torvalds 已提交
161 162 163 164 165 166

	tp->lsndtime = now;

	/* If it is a reply for ato after last received
	 * packet, enter pingpong mode.
	 */
167 168
	if ((u32)(now - icsk->icsk_ack.lrcvtime) < icsk->icsk_ack.ato)
		icsk->icsk_ack.pingpong = 1;
L
Linus Torvalds 已提交
169 170
}

171
/* Account for an ACK we sent. */
S
Stephen Hemminger 已提交
172
static inline void tcp_event_ack_sent(struct sock *sk, unsigned int pkts)
L
Linus Torvalds 已提交
173
{
174 175
	tcp_dec_quickack_mode(sk, pkts);
	inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
L
Linus Torvalds 已提交
176 177 178 179 180 181 182 183 184 185 186
}

/* 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,
187 188
			       int wscale_ok, __u8 *rcv_wscale,
			       __u32 init_rcv_wnd)
L
Linus Torvalds 已提交
189 190 191 192 193 194 195 196 197 198 199 200 201
{
	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
202 203 204 205 206 207
	 * 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
208
	 */
209 210 211 212 213
	if (sysctl_tcp_workaround_signed_windows)
		(*rcv_wnd) = min(space, MAX_TCP_WINDOW);
	else
		(*rcv_wnd) = space;

L
Linus Torvalds 已提交
214 215 216
	(*rcv_wscale) = 0;
	if (wscale_ok) {
		/* Set window scaling on max possible window
217
		 * See RFC1323 for an explanation of the limit to 14
L
Linus Torvalds 已提交
218 219
		 */
		space = max_t(u32, sysctl_tcp_rmem[2], sysctl_rmem_max);
220
		space = min_t(u32, space, *window_clamp);
L
Linus Torvalds 已提交
221 222 223 224 225 226 227
		while (space > 65535 && (*rcv_wscale) < 14) {
			space >>= 1;
			(*rcv_wscale)++;
		}
	}

	/* Set initial window to value enough for senders,
228
	 * following RFC2414. Senders, not following this RFC,
L
Linus Torvalds 已提交
229 230
	 * will be satisfied with 2.
	 */
231
	if (mss > (1 << *rcv_wscale)) {
232
		int init_cwnd = 4;
233
		if (mss > 1460 * 3)
L
Linus Torvalds 已提交
234
			init_cwnd = 2;
235 236
		else if (mss > 1460)
			init_cwnd = 3;
237 238 239 240 241 242 243
		/* when initializing use the value from init_rcv_wnd
		 * rather than the default from above
		 */
		if (init_rcv_wnd &&
		    (*rcv_wnd > init_rcv_wnd * mss))
			*rcv_wnd = init_rcv_wnd * mss;
		else if (*rcv_wnd > init_cwnd * mss)
244
			*rcv_wnd = init_cwnd * mss;
L
Linus Torvalds 已提交
245 246 247 248 249 250 251 252 253 254 255
	}

	/* 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
256
static u16 tcp_select_window(struct sock *sk)
L
Linus Torvalds 已提交
257 258 259 260 261 262
{
	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
263
	if (new_win < cur_win) {
L
Linus Torvalds 已提交
264 265 266 267 268 269 270
		/* 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.
		 */
271
		new_win = ALIGN(cur_win, 1 << tp->rx_opt.rcv_wscale);
L
Linus Torvalds 已提交
272 273 274 275 276 277 278
	}
	tp->rcv_wnd = new_win;
	tp->rcv_wup = tp->rcv_nxt;

	/* Make sure we do not exceed the maximum possible
	 * scaled window.
	 */
279
	if (!tp->rx_opt.rcv_wscale && sysctl_tcp_workaround_signed_windows)
L
Linus Torvalds 已提交
280 281 282 283 284 285 286 287 288 289 290 291 292 293
		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;
}

294
/* Packet ECN state for a SYN-ACK */
295
static inline void TCP_ECN_send_synack(struct tcp_sock *tp, struct sk_buff *skb)
296 297
{
	TCP_SKB_CB(skb)->flags &= ~TCPCB_FLAG_CWR;
298
	if (!(tp->ecn_flags & TCP_ECN_OK))
299 300 301
		TCP_SKB_CB(skb)->flags &= ~TCPCB_FLAG_ECE;
}

302
/* Packet ECN state for a SYN.  */
303 304 305 306 307
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;
308
	if (sysctl_tcp_ecn == 1) {
309
		TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_ECE | TCPCB_FLAG_CWR;
310 311 312 313 314 315 316 317 318 319 320
		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;
}

321 322 323
/* Set up ECN state for a packet on a ESTABLISHED socket that is about to
 * be sent.
 */
324 325 326 327 328 329 330 331 332 333
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);
334
			if (tp->ecn_flags & TCP_ECN_QUEUE_CWR) {
335 336 337 338 339 340 341 342 343 344 345 346 347
				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;
	}
}

348 349 350 351 352
/* 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)
{
353
	skb->ip_summed = CHECKSUM_PARTIAL;
354 355 356 357 358 359 360 361 362 363 364 365 366 367 368
	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
369 370 371 372 373
static inline int tcp_urg_mode(const struct tcp_sock *tp)
{
	return tp->snd_una != tp->snd_up;
}

A
Adam Langley 已提交
374 375 376
#define OPTION_SACK_ADVERTISE	(1 << 0)
#define OPTION_TS		(1 << 1)
#define OPTION_MD5		(1 << 2)
377
#define OPTION_WSCALE		(1 << 3)
378
#define OPTION_COOKIE_EXTENSION	(1 << 4)
A
Adam Langley 已提交
379 380 381 382 383

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 */
384
	u8 hash_size;		/* bytes in hash_location */
A
Adam Langley 已提交
385 386
	u16 mss;		/* 0 to disable */
	__u32 tsval, tsecr;	/* need to include OPTION_TS */
387
	__u8 *hash_location;	/* temporary pointer, overloaded */
A
Adam Langley 已提交
388 389
};

390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416
/* The sysctl int routines are generic, so check consistency here.
 */
static u8 tcp_cookie_size_check(u8 desired)
{
	if (desired > 0) {
		/* previously specified */
		return desired;
	}
	if (sysctl_tcp_cookie_size <= 0) {
		/* no default specified */
		return 0;
	}
	if (sysctl_tcp_cookie_size <= TCP_COOKIE_MIN) {
		/* value too small, specify minimum */
		return TCP_COOKIE_MIN;
	}
	if (sysctl_tcp_cookie_size >= TCP_COOKIE_MAX) {
		/* value too large, specify maximum */
		return TCP_COOKIE_MAX;
	}
	if (0x1 & sysctl_tcp_cookie_size) {
		/* 8-bit multiple, illegal, fix it */
		return (u8)(sysctl_tcp_cookie_size + 0x1);
	}
	return (u8)sysctl_tcp_cookie_size;
}

417 418 419
/* Write previously computed TCP options to the packet.
 *
 * Beware: Something in the Internet is very sensitive to the ordering of
420 421 422 423 424 425 426 427 428 429
 * 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
430
static void tcp_options_write(__be32 *ptr, struct tcp_sock *tp,
431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457
			      struct tcp_out_options *opts)
{
	u8 options = opts->options;	/* mungable copy */

	/* Having both authentication and cookies for security is redundant,
	 * and there's certainly not enough room.  Instead, the cookie-less
	 * extension variant is proposed.
	 *
	 * Consider the pessimal case with authentication.  The options
	 * could look like:
	 *   COOKIE|MD5(20) + MSS(4) + SACK|TS(12) + WSCALE(4) == 40
	 */
	if (unlikely(OPTION_MD5 & options)) {
		if (unlikely(OPTION_COOKIE_EXTENSION & options)) {
			*ptr++ = htonl((TCPOPT_COOKIE << 24) |
				       (TCPOLEN_COOKIE_BASE << 16) |
				       (TCPOPT_MD5SIG << 8) |
				       TCPOLEN_MD5SIG);
		} else {
			*ptr++ = htonl((TCPOPT_NOP << 24) |
				       (TCPOPT_NOP << 16) |
				       (TCPOPT_MD5SIG << 8) |
				       TCPOLEN_MD5SIG);
		}
		options &= ~OPTION_COOKIE_EXTENSION;
		/* overload cookie hash location */
		opts->hash_location = (__u8 *)ptr;
A
Adam Langley 已提交
458
		ptr += 4;
S
Stephen Hemminger 已提交
459
	}
A
Adam Langley 已提交
460

461 462 463 464 465 466
	if (unlikely(opts->mss)) {
		*ptr++ = htonl((TCPOPT_MSS << 24) |
			       (TCPOLEN_MSS << 16) |
			       opts->mss);
	}

467 468
	if (likely(OPTION_TS & options)) {
		if (unlikely(OPTION_SACK_ADVERTISE & options)) {
A
Adam Langley 已提交
469 470 471 472
			*ptr++ = htonl((TCPOPT_SACK_PERM << 24) |
				       (TCPOLEN_SACK_PERM << 16) |
				       (TCPOPT_TIMESTAMP << 8) |
				       TCPOLEN_TIMESTAMP);
473
			options &= ~OPTION_SACK_ADVERTISE;
A
Adam Langley 已提交
474 475 476 477 478 479 480 481 482 483
		} else {
			*ptr++ = htonl((TCPOPT_NOP << 24) |
				       (TCPOPT_NOP << 16) |
				       (TCPOPT_TIMESTAMP << 8) |
				       TCPOLEN_TIMESTAMP);
		}
		*ptr++ = htonl(opts->tsval);
		*ptr++ = htonl(opts->tsecr);
	}

484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
	/* Specification requires after timestamp, so do it now.
	 *
	 * Consider the pessimal case without authentication.  The options
	 * could look like:
	 *   MSS(4) + SACK|TS(12) + COOKIE(20) + WSCALE(4) == 40
	 */
	if (unlikely(OPTION_COOKIE_EXTENSION & options)) {
		__u8 *cookie_copy = opts->hash_location;
		u8 cookie_size = opts->hash_size;

		/* 8-bit multiple handled in tcp_cookie_size_check() above,
		 * and elsewhere.
		 */
		if (0x2 & cookie_size) {
			__u8 *p = (__u8 *)ptr;

			/* 16-bit multiple */
			*p++ = TCPOPT_COOKIE;
			*p++ = TCPOLEN_COOKIE_BASE + cookie_size;
			*p++ = *cookie_copy++;
			*p++ = *cookie_copy++;
			ptr++;
			cookie_size -= 2;
		} else {
			/* 32-bit multiple */
			*ptr++ = htonl(((TCPOPT_NOP << 24) |
					(TCPOPT_NOP << 16) |
					(TCPOPT_COOKIE << 8) |
					TCPOLEN_COOKIE_BASE) +
				       cookie_size);
		}

		if (cookie_size > 0) {
			memcpy(ptr, cookie_copy, cookie_size);
			ptr += (cookie_size / 4);
		}
	}

	if (unlikely(OPTION_SACK_ADVERTISE & options)) {
A
Adam Langley 已提交
523 524 525 526 527 528
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
			       (TCPOPT_SACK_PERM << 8) |
			       TCPOLEN_SACK_PERM);
	}

529
	if (unlikely(OPTION_WSCALE & options)) {
A
Adam Langley 已提交
530 531 532 533 534 535 536 537 538
		*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
539 540 541 542 543
		int this_sack;

		*ptr++ = htonl((TCPOPT_NOP  << 24) |
			       (TCPOPT_NOP  << 16) |
			       (TCPOPT_SACK <<  8) |
A
Adam Langley 已提交
544
			       (TCPOLEN_SACK_BASE + (opts->num_sack_blocks *
S
Stephen Hemminger 已提交
545
						     TCPOLEN_SACK_PERBLOCK)));
S
Stephen Hemminger 已提交
546

A
Adam Langley 已提交
547 548
		for (this_sack = 0; this_sack < opts->num_sack_blocks;
		     ++this_sack) {
S
Stephen Hemminger 已提交
549 550 551
			*ptr++ = htonl(sp[this_sack].start_seq);
			*ptr++ = htonl(sp[this_sack].end_seq);
		}
S
Stephen Hemminger 已提交
552

553
		tp->rx_opt.dsack = 0;
S
Stephen Hemminger 已提交
554
	}
A
Adam Langley 已提交
555 556
}

557 558 559
/* Compute TCP options for SYN packets. This is not the final
 * network wire format yet.
 */
A
Adam Langley 已提交
560 561 562 563
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);
564 565 566 567 568
	struct tcp_cookie_values *cvp = tp->cookie_values;
	unsigned remaining = MAX_TCP_OPTION_SPACE;
	u8 cookie_size = (!tp->rx_opt.cookie_out_never && cvp != NULL) ?
			 tcp_cookie_size_check(cvp->cookie_desired) :
			 0;
A
Adam Langley 已提交
569

570
#ifdef CONFIG_TCP_MD5SIG
A
Adam Langley 已提交
571 572 573
	*md5 = tp->af_specific->md5_lookup(sk, sk);
	if (*md5) {
		opts->options |= OPTION_MD5;
574
		remaining -= TCPOLEN_MD5SIG_ALIGNED;
575
	}
A
Adam Langley 已提交
576 577
#else
	*md5 = NULL;
578
#endif
A
Adam Langley 已提交
579 580 581 582 583 584 585 586 587 588 589

	/* 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);
590
	remaining -= TCPOLEN_MSS_ALIGNED;
A
Adam Langley 已提交
591

592
	if (likely(sysctl_tcp_timestamps && *md5 == NULL)) {
A
Adam Langley 已提交
593 594 595
		opts->options |= OPTION_TS;
		opts->tsval = TCP_SKB_CB(skb)->when;
		opts->tsecr = tp->rx_opt.ts_recent;
596
		remaining -= TCPOLEN_TSTAMP_ALIGNED;
A
Adam Langley 已提交
597
	}
598
	if (likely(sysctl_tcp_window_scaling)) {
A
Adam Langley 已提交
599
		opts->ws = tp->rx_opt.rcv_wscale;
600
		opts->options |= OPTION_WSCALE;
601
		remaining -= TCPOLEN_WSCALE_ALIGNED;
A
Adam Langley 已提交
602
	}
603
	if (likely(sysctl_tcp_sack)) {
A
Adam Langley 已提交
604
		opts->options |= OPTION_SACK_ADVERTISE;
605
		if (unlikely(!(OPTION_TS & opts->options)))
606
			remaining -= TCPOLEN_SACKPERM_ALIGNED;
A
Adam Langley 已提交
607 608
	}

609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655
	/* Note that timestamps are required by the specification.
	 *
	 * Odd numbers of bytes are prohibited by the specification, ensuring
	 * that the cookie is 16-bit aligned, and the resulting cookie pair is
	 * 32-bit aligned.
	 */
	if (*md5 == NULL &&
	    (OPTION_TS & opts->options) &&
	    cookie_size > 0) {
		int need = TCPOLEN_COOKIE_BASE + cookie_size;

		if (0x2 & need) {
			/* 32-bit multiple */
			need += 2; /* NOPs */

			if (need > remaining) {
				/* try shrinking cookie to fit */
				cookie_size -= 2;
				need -= 4;
			}
		}
		while (need > remaining && TCP_COOKIE_MIN <= cookie_size) {
			cookie_size -= 4;
			need -= 4;
		}
		if (TCP_COOKIE_MIN <= cookie_size) {
			opts->options |= OPTION_COOKIE_EXTENSION;
			opts->hash_location = (__u8 *)&cvp->cookie_pair[0];
			opts->hash_size = cookie_size;

			/* Remember for future incarnations. */
			cvp->cookie_desired = cookie_size;

			if (cvp->cookie_desired != cvp->cookie_pair_size) {
				/* Currently use random bytes as a nonce,
				 * assuming these are completely unpredictable
				 * by hostile users of the same system.
				 */
				get_random_bytes(&cvp->cookie_pair[0],
						 cookie_size);
				cvp->cookie_pair_size = cookie_size;
			}

			remaining -= need;
		}
	}
	return MAX_TCP_OPTION_SPACE - remaining;
S
Stephen Hemminger 已提交
656 657
}

658
/* Set up TCP options for SYN-ACKs. */
A
Adam Langley 已提交
659 660 661 662
static unsigned tcp_synack_options(struct sock *sk,
				   struct request_sock *req,
				   unsigned mss, struct sk_buff *skb,
				   struct tcp_out_options *opts,
663 664 665
				   struct tcp_md5sig_key **md5,
				   struct tcp_extend_values *xvp)
{
A
Adam Langley 已提交
666
	struct inet_request_sock *ireq = inet_rsk(req);
667 668 669 670 671
	unsigned remaining = MAX_TCP_OPTION_SPACE;
	u8 cookie_plus = (xvp != NULL && !xvp->cookie_out_never) ?
			 xvp->cookie_plus :
			 0;
	bool doing_ts = ireq->tstamp_ok;
A
Adam Langley 已提交
672

673
#ifdef CONFIG_TCP_MD5SIG
A
Adam Langley 已提交
674 675 676
	*md5 = tcp_rsk(req)->af_specific->md5_lookup(sk, req);
	if (*md5) {
		opts->options |= OPTION_MD5;
677 678 679 680 681 682 683 684
		remaining -= TCPOLEN_MD5SIG_ALIGNED;

		/* 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->sack_ok;
685
	}
A
Adam Langley 已提交
686 687
#else
	*md5 = NULL;
688
#endif
A
Adam Langley 已提交
689

690
	/* We always send an MSS option. */
A
Adam Langley 已提交
691
	opts->mss = mss;
692
	remaining -= TCPOLEN_MSS_ALIGNED;
A
Adam Langley 已提交
693 694 695

	if (likely(ireq->wscale_ok)) {
		opts->ws = ireq->rcv_wscale;
696
		opts->options |= OPTION_WSCALE;
697
		remaining -= TCPOLEN_WSCALE_ALIGNED;
A
Adam Langley 已提交
698 699 700 701 702
	}
	if (likely(doing_ts)) {
		opts->options |= OPTION_TS;
		opts->tsval = TCP_SKB_CB(skb)->when;
		opts->tsecr = req->ts_recent;
703
		remaining -= TCPOLEN_TSTAMP_ALIGNED;
A
Adam Langley 已提交
704 705 706 707
	}
	if (likely(ireq->sack_ok)) {
		opts->options |= OPTION_SACK_ADVERTISE;
		if (unlikely(!doing_ts))
708
			remaining -= TCPOLEN_SACKPERM_ALIGNED;
A
Adam Langley 已提交
709 710
	}

711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733
	/* Similar rationale to tcp_syn_options() applies here, too.
	 * If the <SYN> options fit, the same options should fit now!
	 */
	if (*md5 == NULL &&
	    doing_ts &&
	    cookie_plus > TCPOLEN_COOKIE_BASE) {
		int need = cookie_plus; /* has TCPOLEN_COOKIE_BASE */

		if (0x2 & need) {
			/* 32-bit multiple */
			need += 2; /* NOPs */
		}
		if (need <= remaining) {
			opts->options |= OPTION_COOKIE_EXTENSION;
			opts->hash_size = cookie_plus - TCPOLEN_COOKIE_BASE;
			remaining -= need;
		} else {
			/* There's no error return, so flag it. */
			xvp->cookie_out_never = 1; /* true */
			opts->hash_size = 0;
		}
	}
	return MAX_TCP_OPTION_SPACE - remaining;
A
Adam Langley 已提交
734 735
}

736 737 738
/* Compute TCP options for ESTABLISHED sockets. This is not the
 * final wire format yet.
 */
A
Adam Langley 已提交
739 740 741 742 743 744
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;
745
	unsigned int eff_sacks;
A
Adam Langley 已提交
746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763

#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;
	}

764 765
	eff_sacks = tp->rx_opt.num_sacks + tp->rx_opt.dsack;
	if (unlikely(eff_sacks)) {
A
Adam Langley 已提交
766 767
		const unsigned remaining = MAX_TCP_OPTION_SPACE - size;
		opts->num_sack_blocks =
768
			min_t(unsigned, eff_sacks,
A
Adam Langley 已提交
769 770 771 772 773 774 775
			      (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
776
}
L
Linus Torvalds 已提交
777 778 779 780 781 782 783 784 785 786 787 788

/* 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.
 */
789 790
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
791
{
792 793 794 795
	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
796 797
	struct tcp_out_options opts;
	unsigned tcp_options_size, tcp_header_size;
798
	struct tcp_md5sig_key *md5;
799 800 801 802 803 804 805 806
	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.
	 */
807
	if (icsk->icsk_ca_ops->flags & TCP_CONG_RTT_STAMP)
808 809 810 811 812 813 814 815 816 817
		__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
818

819 820 821
	inet = inet_sk(sk);
	tp = tcp_sk(sk);
	tcb = TCP_SKB_CB(skb);
A
Adam Langley 已提交
822
	memset(&opts, 0, sizeof(opts));
L
Linus Torvalds 已提交
823

A
Adam Langley 已提交
824 825 826 827 828 829
	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);
830

831 832 833
	if (tcp_packets_in_flight(tp) == 0)
		tcp_ca_event(sk, CA_EVENT_TX_START);

834 835
	skb_push(skb, tcp_header_size);
	skb_reset_transport_header(skb);
836
	skb_set_owner_w(skb, sk);
837 838

	/* Build TCP header and checksum it. */
839
	th = tcp_hdr(skb);
E
Eric Dumazet 已提交
840 841
	th->source		= inet->inet_sport;
	th->dest		= inet->inet_dport;
842 843
	th->seq			= htonl(tcb->seq);
	th->ack_seq		= htonl(tp->rcv_nxt);
844
	*(((__be16 *)th) + 6)	= htons(((tcp_header_size >> 2) << 12) |
845 846 847 848 849 850
					tcb->flags);

	if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
		/* RFC1323: The window in SYN & SYN/ACK segments
		 * is never scaled.
		 */
851
		th->window	= htons(min(tp->rcv_wnd, 65535U));
852 853 854 855 856
	} else {
		th->window	= htons(tcp_select_window(sk));
	}
	th->check		= 0;
	th->urg_ptr		= 0;
L
Linus Torvalds 已提交
857

I
Ilpo Järvinen 已提交
858
	/* The urg_mode check is necessary during a below snd_una win probe */
859 860 861 862 863 864 865 866
	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;
		}
867
	}
L
Linus Torvalds 已提交
868

869
	tcp_options_write((__be32 *)(th + 1), tp, &opts);
A
Adam Langley 已提交
870
	if (likely((tcb->flags & TCPCB_FLAG_SYN) == 0))
871
		TCP_ECN_send(sk, skb, tcp_header_size);
L
Linus Torvalds 已提交
872

873 874 875
#ifdef CONFIG_TCP_MD5SIG
	/* Calculate the MD5 hash, as we have all we need now */
	if (md5) {
A
Adam Langley 已提交
876
		sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
877
		tp->af_specific->calc_md5_hash(opts.hash_location,
878
					       md5, sk, NULL, skb);
879 880 881
	}
#endif

882
	icsk->icsk_af_ops->send_check(sk, skb);
L
Linus Torvalds 已提交
883

884 885
	if (likely(tcb->flags & TCPCB_FLAG_ACK))
		tcp_event_ack_sent(sk, tcp_skb_pcount(skb));
L
Linus Torvalds 已提交
886

887 888
	if (skb->len != tcp_header_size)
		tcp_event_data_sent(tp, skb, sk);
L
Linus Torvalds 已提交
889

890
	if (after(tcb->end_seq, tp->snd_nxt) || tcb->seq == tcb->end_seq)
P
Pavel Emelyanov 已提交
891
		TCP_INC_STATS(sock_net(sk), TCP_MIB_OUTSEGS);
L
Linus Torvalds 已提交
892

893
	err = icsk->icsk_af_ops->queue_xmit(skb);
894
	if (likely(err <= 0))
895 896
		return err;

897
	tcp_enter_cwr(sk, 1);
898

899
	return net_xmit_eval(err);
L
Linus Torvalds 已提交
900 901
}

902
/* This routine just queues the buffer for sending.
L
Linus Torvalds 已提交
903 904 905 906 907 908 909 910 911 912 913
 *
 * 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);
914
	tcp_add_write_queue_tail(sk, skb);
915 916
	sk->sk_wmem_queued += skb->truesize;
	sk_mem_charge(sk, skb->truesize);
L
Linus Torvalds 已提交
917 918
}

919
/* Initialize TSO segments for a packet. */
920 921
static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb,
				 unsigned int mss_now)
922
{
923 924
	if (skb->len <= mss_now || !sk_can_gso(sk) ||
	    skb->ip_summed == CHECKSUM_NONE) {
925 926 927
		/* Avoid the costly divide in the normal
		 * non-TSO case.
		 */
928 929 930
		skb_shinfo(skb)->gso_segs = 1;
		skb_shinfo(skb)->gso_size = 0;
		skb_shinfo(skb)->gso_type = 0;
931
	} else {
I
Ilpo Järvinen 已提交
932
		skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss_now);
933
		skb_shinfo(skb)->gso_size = mss_now;
934
		skb_shinfo(skb)->gso_type = sk->sk_gso_type;
L
Linus Torvalds 已提交
935 936 937
	}
}

I
Ilpo Järvinen 已提交
938
/* When a modification to fackets out becomes necessary, we need to check
939
 * skb is counted to fackets_out or not.
I
Ilpo Järvinen 已提交
940
 */
941
static void tcp_adjust_fackets_out(struct sock *sk, struct sk_buff *skb,
I
Ilpo Järvinen 已提交
942 943
				   int decr)
{
944 945
	struct tcp_sock *tp = tcp_sk(sk);

946
	if (!tp->sacked_out || tcp_is_reno(tp))
I
Ilpo Järvinen 已提交
947 948
		return;

949
	if (after(tcp_highest_sack_seq(tp), TCP_SKB_CB(skb)->seq))
I
Ilpo Järvinen 已提交
950 951 952
		tp->fackets_out -= decr;
}

953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976
/* 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) &&
977
	    (tcp_is_fack(tp) || (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)))
978 979 980 981 982
		tp->lost_cnt_hint -= decr;

	tcp_verify_left_out(tp);
}

L
Linus Torvalds 已提交
983 984
/* 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
985
 * packet to the list.  This won't be called frequently, I hope.
L
Linus Torvalds 已提交
986 987
 * Remember, these are still headerless SKBs at this point.
 */
988 989
int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len,
		 unsigned int mss_now)
L
Linus Torvalds 已提交
990 991 992
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *buff;
993
	int nsize, old_factor;
994
	int nlen;
995
	u8 flags;
L
Linus Torvalds 已提交
996

997
	BUG_ON(len > skb->len);
998

L
Linus Torvalds 已提交
999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
	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
1012

1013 1014
	sk->sk_wmem_queued += buff->truesize;
	sk_mem_charge(sk, buff->truesize);
1015 1016 1017
	nlen = skb->len - len - nsize;
	buff->truesize += nlen;
	skb->truesize -= nlen;
L
Linus Torvalds 已提交
1018 1019 1020 1021 1022 1023 1024 1025

	/* 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;
1026
	TCP_SKB_CB(skb)->flags = flags & ~(TCPCB_FLAG_FIN | TCPCB_FLAG_PSH);
L
Linus Torvalds 已提交
1027
	TCP_SKB_CB(buff)->flags = flags;
1028
	TCP_SKB_CB(buff)->sacked = TCP_SKB_CB(skb)->sacked;
L
Linus Torvalds 已提交
1029

1030
	if (!skb_shinfo(skb)->nr_frags && skb->ip_summed != CHECKSUM_PARTIAL) {
L
Linus Torvalds 已提交
1031
		/* Copy and checksum data tail into the new buffer. */
1032 1033
		buff->csum = csum_partial_copy_nocheck(skb->data + len,
						       skb_put(buff, nsize),
L
Linus Torvalds 已提交
1034 1035 1036 1037 1038 1039
						       nsize, 0);

		skb_trim(skb, len);

		skb->csum = csum_block_sub(skb->csum, buff->csum, len);
	} else {
1040
		skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
1041 1042 1043 1044 1045 1046 1047 1048 1049
		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;
1050
	buff->tstamp = skb->tstamp;
L
Linus Torvalds 已提交
1051

1052 1053
	old_factor = tcp_skb_pcount(skb);

L
Linus Torvalds 已提交
1054
	/* Fix up tso_factor for both original and new SKB.  */
1055 1056
	tcp_set_skb_tso_segs(sk, skb, mss_now);
	tcp_set_skb_tso_segs(sk, buff, mss_now);
L
Linus Torvalds 已提交
1057

1058 1059 1060
	/* If this packet has been sent out already, we must
	 * adjust the various packet counters.
	 */
1061
	if (!before(tp->snd_nxt, TCP_SKB_CB(buff)->end_seq)) {
1062 1063
		int diff = old_factor - tcp_skb_pcount(skb) -
			tcp_skb_pcount(buff);
L
Linus Torvalds 已提交
1064

1065 1066
		if (diff)
			tcp_adjust_pcount(sk, skb, diff);
L
Linus Torvalds 已提交
1067 1068 1069
	}

	/* Link BUFF into the send queue. */
1070
	skb_header_release(buff);
1071
	tcp_insert_write_queue_after(skb, buff, sk);
L
Linus Torvalds 已提交
1072 1073 1074 1075 1076 1077 1078 1079

	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.
 */
1080
static void __pskb_trim_head(struct sk_buff *skb, int len)
L
Linus Torvalds 已提交
1081 1082 1083 1084 1085
{
	int i, k, eat;

	eat = len;
	k = 0;
1086
	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
L
Linus Torvalds 已提交
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101
		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;

1102
	skb_reset_tail_pointer(skb);
L
Linus Torvalds 已提交
1103 1104 1105 1106
	skb->data_len -= len;
	skb->len = skb->data_len;
}

1107
/* Remove acked data from a packet in the transmit queue. */
L
Linus Torvalds 已提交
1108 1109
int tcp_trim_head(struct sock *sk, struct sk_buff *skb, u32 len)
{
1110
	if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
L
Linus Torvalds 已提交
1111 1112
		return -ENOMEM;

1113 1114
	/* If len == headlen, we avoid __skb_pull to preserve alignment. */
	if (unlikely(len < skb_headlen(skb)))
L
Linus Torvalds 已提交
1115
		__skb_pull(skb, len);
1116 1117
	else
		__pskb_trim_head(skb, len - skb_headlen(skb));
L
Linus Torvalds 已提交
1118 1119

	TCP_SKB_CB(skb)->seq += len;
1120
	skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
1121 1122 1123

	skb->truesize	     -= len;
	sk->sk_wmem_queued   -= len;
1124
	sk_mem_uncharge(sk, len);
L
Linus Torvalds 已提交
1125 1126 1127 1128 1129 1130
	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
1131
		tcp_set_skb_tso_segs(sk, skb, tcp_current_mss(sk));
L
Linus Torvalds 已提交
1132 1133 1134 1135

	return 0;
}

1136
/* Calculate MSS. Not accounting for SACKs here.  */
J
John Heffner 已提交
1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179
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;
}

1180
/* MTU probing init per socket */
J
John Heffner 已提交
1181 1182 1183 1184 1185 1186 1187
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) +
1188
			       icsk->icsk_af_ops->net_header_len;
J
John Heffner 已提交
1189 1190 1191 1192
	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
1193 1194 1195 1196 1197 1198
/* 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
1199
   It is minimum of user_mss and mss received with SYN.
L
Linus Torvalds 已提交
1200 1201
   It also does not include TCP options.

1202
   inet_csk(sk)->icsk_pmtu_cookie is last pmtu, seen by this function.
L
Linus Torvalds 已提交
1203 1204 1205 1206 1207 1208 1209 1210 1211

   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.

1212 1213
   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
1214 1215 1216 1217
 */
unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu)
{
	struct tcp_sock *tp = tcp_sk(sk);
1218
	struct inet_connection_sock *icsk = inet_csk(sk);
J
John Heffner 已提交
1219
	int mss_now;
L
Linus Torvalds 已提交
1220

J
John Heffner 已提交
1221 1222
	if (icsk->icsk_mtup.search_high > pmtu)
		icsk->icsk_mtup.search_high = pmtu;
L
Linus Torvalds 已提交
1223

J
John Heffner 已提交
1224
	mss_now = tcp_mtu_to_mss(sk, pmtu);
1225
	mss_now = tcp_bound_to_half_wnd(tp, mss_now);
L
Linus Torvalds 已提交
1226 1227

	/* And store cached results */
1228
	icsk->icsk_pmtu_cookie = pmtu;
J
John Heffner 已提交
1229 1230
	if (icsk->icsk_mtup.enabled)
		mss_now = min(mss_now, tcp_mtu_to_mss(sk, icsk->icsk_mtup.search_low));
1231
	tp->mss_cache = mss_now;
L
Linus Torvalds 已提交
1232 1233 1234 1235 1236 1237 1238

	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
1239
unsigned int tcp_current_mss(struct sock *sk)
L
Linus Torvalds 已提交
1240 1241 1242
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct dst_entry *dst = __sk_dst_get(sk);
1243
	u32 mss_now;
A
Adam Langley 已提交
1244 1245 1246
	unsigned header_len;
	struct tcp_out_options opts;
	struct tcp_md5sig_key *md5;
1247 1248 1249

	mss_now = tp->mss_cache;

L
Linus Torvalds 已提交
1250 1251
	if (dst) {
		u32 mtu = dst_mtu(dst);
1252
		if (mtu != inet_csk(sk)->icsk_pmtu_cookie)
L
Linus Torvalds 已提交
1253 1254 1255
			mss_now = tcp_sync_mss(sk, mtu);
	}

A
Adam Langley 已提交
1256 1257 1258 1259 1260 1261 1262 1263 1264 1265
	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;
	}
1266

L
Linus Torvalds 已提交
1267 1268 1269
	return mss_now;
}

1270
/* Congestion window validation. (RFC2861) */
1271
static void tcp_cwnd_validate(struct sock *sk)
1272
{
1273
	struct tcp_sock *tp = tcp_sk(sk);
1274

1275
	if (tp->packets_out >= tp->snd_cwnd) {
1276 1277 1278 1279 1280 1281 1282 1283
		/* 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;

1284 1285
		if (sysctl_tcp_slow_start_after_idle &&
		    (s32)(tcp_time_stamp - tp->snd_cwnd_stamp) >= inet_csk(sk)->icsk_rto)
1286 1287 1288 1289
			tcp_cwnd_application_limited(sk);
	}
}

1290 1291 1292 1293
/* 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).
1294 1295 1296 1297 1298 1299 1300
 *
 * 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.
1301 1302
 */
static unsigned int tcp_mss_split_point(struct sock *sk, struct sk_buff *skb,
1303
					unsigned int mss_now, unsigned int cwnd)
1304
{
1305 1306
	struct tcp_sock *tp = tcp_sk(sk);
	u32 needed, window, cwnd_len;
1307

1308
	window = tcp_wnd_end(tp) - TCP_SKB_CB(skb)->seq;
1309
	cwnd_len = mss_now * cwnd;
1310 1311 1312 1313

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

1314 1315
	needed = min(skb->len, window);

1316
	if (cwnd_len <= needed)
1317 1318 1319
		return cwnd_len;

	return needed - needed % mss_now;
1320 1321 1322 1323 1324
}

/* 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.
 */
1325 1326
static inline unsigned int tcp_cwnd_test(struct tcp_sock *tp,
					 struct sk_buff *skb)
1327 1328 1329 1330
{
	u32 in_flight, cwnd;

	/* Don't be strict about the congestion window for the final FIN.  */
1331 1332
	if ((TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN) &&
	    tcp_skb_pcount(skb) == 1)
1333 1334 1335 1336 1337 1338 1339 1340 1341 1342
		return 1;

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

	return 0;
}

1343 1344
/* Intialize TSO state of a skb.
 * This must be invoked the first time we consider transmitting
1345 1346
 * SKB onto the wire.
 */
1347 1348
static int tcp_init_tso_segs(struct sock *sk, struct sk_buff *skb,
			     unsigned int mss_now)
1349 1350 1351
{
	int tso_segs = tcp_skb_pcount(skb);

1352
	if (!tso_segs || (tso_segs > 1 && tcp_skb_mss(skb) != mss_now)) {
1353
		tcp_set_skb_tso_segs(sk, skb, mss_now);
1354 1355 1356 1357 1358
		tso_segs = tcp_skb_pcount(skb);
	}
	return tso_segs;
}

1359
/* Minshall's variant of the Nagle send check. */
1360 1361
static inline int tcp_minshall_check(const struct tcp_sock *tp)
{
1362
	return after(tp->snd_sml, tp->snd_una) &&
1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373
		!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,
1374
				  const struct sk_buff *skb,
1375 1376 1377
				  unsigned mss_now, int nonagle)
{
	return (skb->len < mss_now &&
1378 1379
		((nonagle & TCP_NAGLE_CORK) ||
		 (!nonagle && tp->packets_out && tcp_minshall_check(tp))));
1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396
}

/* 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;

1397 1398 1399
	/* 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
1400
	if (tcp_urg_mode(tp) || (tp->frto_counter == 2) ||
1401 1402 1403 1404 1405 1406 1407 1408 1409 1410
	    (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? */
1411 1412
static inline int tcp_snd_wnd_test(struct tcp_sock *tp, struct sk_buff *skb,
				   unsigned int cur_mss)
1413 1414 1415 1416 1417 1418
{
	u32 end_seq = TCP_SKB_CB(skb)->end_seq;

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

1419
	return !after(end_seq, tcp_wnd_end(tp));
1420 1421
}

1422
/* This checks if the data bearing packet SKB (usually tcp_send_head(sk))
1423 1424 1425 1426 1427 1428 1429 1430 1431
 * 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;

1432
	tcp_init_tso_segs(sk, skb, cur_mss);
1433 1434 1435 1436 1437

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

	cwnd_quota = tcp_cwnd_test(tp, skb);
1438
	if (cwnd_quota && !tcp_snd_wnd_test(tp, skb, cur_mss))
1439 1440 1441 1442 1443
		cwnd_quota = 0;

	return cwnd_quota;
}

1444
/* Test if sending is allowed right now. */
1445
int tcp_may_send_now(struct sock *sk)
1446
{
1447
	struct tcp_sock *tp = tcp_sk(sk);
1448
	struct sk_buff *skb = tcp_send_head(sk);
1449 1450

	return (skb &&
I
Ilpo Järvinen 已提交
1451
		tcp_snd_test(sk, skb, tcp_current_mss(sk),
1452
			     (tcp_skb_is_last(sk, skb) ?
1453
			      tp->nonagle : TCP_NAGLE_PUSH)));
1454 1455 1456 1457 1458 1459 1460 1461 1462
}

/* 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).
 */
1463 1464
static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len,
			unsigned int mss_now)
1465 1466 1467
{
	struct sk_buff *buff;
	int nlen = skb->len - len;
1468
	u8 flags;
1469 1470

	/* All of a TSO frame must be composed of paged data.  */
1471 1472
	if (skb->len != skb->data_len)
		return tcp_fragment(sk, skb, len, mss_now);
1473

1474
	buff = sk_stream_alloc_skb(sk, 0, GFP_ATOMIC);
1475 1476 1477
	if (unlikely(buff == NULL))
		return -ENOMEM;

1478 1479
	sk->sk_wmem_queued += buff->truesize;
	sk_mem_charge(sk, buff->truesize);
1480
	buff->truesize += nlen;
1481 1482 1483 1484 1485 1486 1487 1488 1489
	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;
1490
	TCP_SKB_CB(skb)->flags = flags & ~(TCPCB_FLAG_FIN | TCPCB_FLAG_PSH);
1491 1492 1493 1494 1495
	TCP_SKB_CB(buff)->flags = flags;

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

1496
	buff->ip_summed = skb->ip_summed = CHECKSUM_PARTIAL;
1497 1498 1499
	skb_split(skb, buff, len);

	/* Fix up tso_factor for both original and new SKB.  */
1500 1501
	tcp_set_skb_tso_segs(sk, skb, mss_now);
	tcp_set_skb_tso_segs(sk, buff, mss_now);
1502 1503 1504

	/* Link BUFF into the send queue. */
	skb_header_release(buff);
1505
	tcp_insert_write_queue_after(skb, buff, sk);
1506 1507 1508 1509 1510 1511 1512 1513 1514

	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.
 */
1515
static int tcp_tso_should_defer(struct sock *sk, struct sk_buff *skb)
1516
{
1517
	struct tcp_sock *tp = tcp_sk(sk);
1518
	const struct inet_connection_sock *icsk = inet_csk(sk);
1519 1520 1521
	u32 send_win, cong_win, limit, in_flight;

	if (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN)
J
John Heffner 已提交
1522
		goto send_now;
1523

1524
	if (icsk->icsk_ca_state != TCP_CA_Open)
J
John Heffner 已提交
1525 1526 1527
		goto send_now;

	/* Defer for less than two clock ticks. */
I
Ilpo Järvinen 已提交
1528
	if (tp->tso_deferred &&
1529
	    (((u32)jiffies << 1) >> 1) - (tp->tso_deferred >> 1) > 1)
J
John Heffner 已提交
1530
		goto send_now;
1531

1532 1533
	in_flight = tcp_packets_in_flight(tp);

1534
	BUG_ON(tcp_skb_pcount(skb) <= 1 || (tp->snd_cwnd <= in_flight));
1535

1536
	send_win = tcp_wnd_end(tp) - TCP_SKB_CB(skb)->seq;
1537 1538 1539 1540 1541 1542

	/* 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);

1543
	/* If a full-sized TSO skb can be sent, do it. */
1544
	if (limit >= sk->sk_gso_max_size)
J
John Heffner 已提交
1545
		goto send_now;
1546

1547 1548 1549 1550
	/* 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;

1551 1552 1553 1554 1555 1556 1557 1558
	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
1559
			goto send_now;
1560 1561 1562 1563 1564 1565 1566
	} 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
1567
			goto send_now;
1568 1569 1570
	}

	/* Ok, it looks like it is advisable to defer.  */
1571
	tp->tso_deferred = 1 | (jiffies << 1);
J
John Heffner 已提交
1572

1573
	return 1;
J
John Heffner 已提交
1574 1575 1576 1577

send_now:
	tp->tso_deferred = 0;
	return 0;
1578 1579
}

J
John Heffner 已提交
1580
/* Create a new MTU probe if we are ready.
1581 1582 1583 1584
 * 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
1585 1586
 * Returns 0 if we should wait to probe (no cwnd available),
 *         1 if a probe was sent,
1587 1588
 *         -1 otherwise
 */
J
John Heffner 已提交
1589 1590 1591 1592 1593 1594 1595
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;
1596
	int size_needed;
J
John Heffner 已提交
1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607
	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 ||
1608
	    tp->rx_opt.num_sacks || tp->rx_opt.dsack)
J
John Heffner 已提交
1609 1610 1611
		return -1;

	/* Very simple search strategy: just double the MSS. */
I
Ilpo Järvinen 已提交
1612
	mss_now = tcp_current_mss(sk);
1613
	probe_size = 2 * tp->mss_cache;
1614
	size_needed = probe_size + (tp->reordering + 1) * tp->mss_cache;
J
John Heffner 已提交
1615 1616 1617 1618 1619 1620
	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? */
1621
	if (tp->write_seq - tp->snd_nxt < size_needed)
J
John Heffner 已提交
1622 1623
		return -1;

1624 1625
	if (tp->snd_wnd < size_needed)
		return -1;
1626
	if (after(tp->snd_nxt + size_needed, tcp_wnd_end(tp)))
1627
		return 0;
J
John Heffner 已提交
1628

1629 1630 1631
	/* 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
1632 1633 1634 1635 1636 1637 1638 1639
			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;
1640 1641
	sk->sk_wmem_queued += nskb->truesize;
	sk_mem_charge(sk, nskb->truesize);
J
John Heffner 已提交
1642

1643
	skb = tcp_send_head(sk);
J
John Heffner 已提交
1644 1645 1646 1647 1648 1649

	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;
1650
	nskb->ip_summed = skb->ip_summed;
J
John Heffner 已提交
1651

1652 1653
	tcp_insert_write_queue_before(nskb, skb, sk);

J
John Heffner 已提交
1654
	len = 0;
1655
	tcp_for_write_queue_from_safe(skb, next, sk) {
J
John Heffner 已提交
1656 1657 1658 1659 1660
		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,
1661 1662
							    skb_put(nskb, copy),
							    copy, nskb->csum);
J
John Heffner 已提交
1663 1664 1665 1666 1667

		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;
1668
			tcp_unlink_write_queue(skb, sk);
1669
			sk_wmem_free_skb(sk, skb);
J
John Heffner 已提交
1670 1671
		} else {
			TCP_SKB_CB(nskb)->flags |= TCP_SKB_CB(skb)->flags &
1672
						   ~(TCPCB_FLAG_FIN|TCPCB_FLAG_PSH);
J
John Heffner 已提交
1673 1674
			if (!skb_shinfo(skb)->nr_frags) {
				skb_pull(skb, copy);
1675
				if (skb->ip_summed != CHECKSUM_PARTIAL)
1676 1677
					skb->csum = csum_partial(skb->data,
								 skb->len, 0);
J
John Heffner 已提交
1678 1679 1680 1681 1682 1683 1684 1685
			} else {
				__pskb_trim_head(skb, copy);
				tcp_set_skb_tso_segs(sk, skb, mss_now);
			}
			TCP_SKB_CB(skb)->seq += copy;
		}

		len += copy;
1686 1687 1688

		if (len >= probe_size)
			break;
J
John Heffner 已提交
1689 1690 1691 1692 1693 1694 1695 1696
	}
	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
1697
		 * effectively two packets. */
J
John Heffner 已提交
1698
		tp->snd_cwnd--;
1699
		tcp_event_new_data_sent(sk, nskb);
J
John Heffner 已提交
1700 1701

		icsk->icsk_mtup.probe_size = tcp_mss_to_mtu(sk, nskb->len);
1702 1703
		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
1704 1705 1706 1707 1708 1709 1710

		return 1;
	}

	return -1;
}

L
Linus Torvalds 已提交
1711 1712 1713 1714
/* This routine writes packets to the network.  It advances the
 * send_head.  This happens as incoming acks open up the remote
 * window for us.
 *
1715 1716 1717 1718
 * 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
1719 1720 1721
 * Returns 1, if no segments are in flight and we have queued segments, but
 * cannot send anything now because of SWS or another problem.
 */
1722 1723
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
1724 1725
{
	struct tcp_sock *tp = tcp_sk(sk);
1726
	struct sk_buff *skb;
1727 1728
	unsigned int tso_segs, sent_pkts;
	int cwnd_quota;
J
John Heffner 已提交
1729
	int result;
L
Linus Torvalds 已提交
1730

1731
	sent_pkts = 0;
J
John Heffner 已提交
1732

1733 1734 1735 1736 1737 1738 1739 1740
	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
1741 1742
	}

1743
	while ((skb = tcp_send_head(sk))) {
1744 1745
		unsigned int limit;

1746
		tso_segs = tcp_init_tso_segs(sk, skb, mss_now);
1747
		BUG_ON(!tso_segs);
1748

1749 1750 1751 1752 1753 1754 1755
		cwnd_quota = tcp_cwnd_test(tp, skb);
		if (!cwnd_quota)
			break;

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

1756 1757 1758 1759 1760 1761
		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 {
1762
			if (!push_one && tcp_tso_should_defer(sk, skb))
1763 1764
				break;
		}
1765

1766
		limit = mss_now;
1767
		if (tso_segs > 1 && !tcp_urg_mode(tp))
1768 1769
			limit = tcp_mss_split_point(sk, skb, mss_now,
						    cwnd_quota);
L
Linus Torvalds 已提交
1770

1771 1772 1773 1774
		if (skb->len > limit &&
		    unlikely(tso_fragment(sk, skb, limit, mss_now)))
			break;

1775
		TCP_SKB_CB(skb)->when = tcp_time_stamp;
1776

1777
		if (unlikely(tcp_transmit_skb(sk, skb, 1, gfp)))
1778
			break;
L
Linus Torvalds 已提交
1779

1780 1781 1782
		/* Advance the send_head.  This one is sent out.
		 * This call will increment packets_out.
		 */
1783
		tcp_event_new_data_sent(sk, skb);
L
Linus Torvalds 已提交
1784

1785
		tcp_minshall_update(tp, mss_now, skb);
1786
		sent_pkts++;
1787 1788 1789

		if (push_one)
			break;
1790
	}
L
Linus Torvalds 已提交
1791

1792
	if (likely(sent_pkts)) {
1793
		tcp_cwnd_validate(sk);
1794
		return 0;
L
Linus Torvalds 已提交
1795
	}
1796
	return !tp->packets_out && tcp_send_head(sk);
L
Linus Torvalds 已提交
1797 1798
}

1799 1800 1801 1802
/* 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.
 */
1803 1804
void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
			       int nonagle)
1805
{
1806 1807 1808 1809 1810 1811 1812
	/* 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;

1813
	if (tcp_write_xmit(sk, cur_mss, nonagle, 0, GFP_ATOMIC))
1814
		tcp_check_probe_timer(sk);
1815 1816
}

1817 1818 1819 1820 1821
/* 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)
{
1822
	struct sk_buff *skb = tcp_send_head(sk);
1823 1824 1825

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

1826
	tcp_write_xmit(sk, mss_now, TCP_NAGLE_PUSH, 1, sk->sk_allocation);
1827 1828
}

L
Linus Torvalds 已提交
1829 1830
/* This function returns the amount that we can raise the
 * usable window based on the following constraints
1831
 *
L
Linus Torvalds 已提交
1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849
 * 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.
1850
 *
L
Linus Torvalds 已提交
1851 1852
 * Of course, if the sender implements sender side SWS prevention
 * then this will not be a problem.
1853
 *
L
Linus Torvalds 已提交
1854
 * BSD seems to make the following compromise:
1855
 *
L
Linus Torvalds 已提交
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
 *	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)
{
1883
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1884
	struct tcp_sock *tp = tcp_sk(sk);
S
Stephen Hemminger 已提交
1885
	/* MSS for the peer's data.  Previous versions used mss_clamp
L
Linus Torvalds 已提交
1886 1887 1888 1889 1890
	 * 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
	 */
1891
	int mss = icsk->icsk_ack.rcv_mss;
L
Linus Torvalds 已提交
1892 1893 1894 1895 1896
	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)
1897
		mss = full_space;
L
Linus Torvalds 已提交
1898

1899
	if (free_space < (full_space >> 1)) {
1900
		icsk->icsk_ack.quick = 0;
L
Linus Torvalds 已提交
1901 1902

		if (tcp_memory_pressure)
1903 1904
			tp->rcv_ssthresh = min(tp->rcv_ssthresh,
					       4U * tp->advmss);
L
Linus Torvalds 已提交
1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936

		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)
1937
			window = (free_space / mss) * mss;
1938
		else if (mss == full_space &&
1939
			 free_space > window + (full_space >> 1))
1940
			window = free_space;
L
Linus Torvalds 已提交
1941 1942 1943 1944 1945
	}

	return window;
}

1946 1947
/* 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
1948 1949
{
	struct tcp_sock *tp = tcp_sk(sk);
1950
	struct sk_buff *next_skb = tcp_write_queue_next(sk, skb);
1951
	int skb_size, next_skb_size;
L
Linus Torvalds 已提交
1952

1953 1954
	skb_size = skb->len;
	next_skb_size = next_skb->len;
L
Linus Torvalds 已提交
1955

1956
	BUG_ON(tcp_skb_pcount(skb) != 1 || tcp_skb_pcount(next_skb) != 1);
1957

1958
	tcp_highest_sack_combine(sk, next_skb, skb);
L
Linus Torvalds 已提交
1959

1960
	tcp_unlink_write_queue(next_skb, sk);
L
Linus Torvalds 已提交
1961

1962 1963
	skb_copy_from_linear_data(next_skb, skb_put(skb, next_skb_size),
				  next_skb_size);
L
Linus Torvalds 已提交
1964

1965 1966
	if (next_skb->ip_summed == CHECKSUM_PARTIAL)
		skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
1967

1968 1969
	if (skb->ip_summed != CHECKSUM_PARTIAL)
		skb->csum = csum_block_add(skb->csum, next_skb->csum, skb_size);
L
Linus Torvalds 已提交
1970

1971 1972
	/* 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
1973

1974 1975
	/* Merge over control information. This moves PSH/FIN etc. over */
	TCP_SKB_CB(skb)->flags |= TCP_SKB_CB(next_skb)->flags;
1976 1977 1978 1979 1980 1981 1982

	/* 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 */
1983 1984 1985
	tcp_clear_retrans_hints_partial(tp);
	if (next_skb == tp->retransmit_skb_hint)
		tp->retransmit_skb_hint = skb;
1986

1987 1988
	tcp_adjust_pcount(sk, next_skb, tcp_skb_pcount(next_skb));

1989
	sk_wmem_free_skb(sk, next_skb);
L
Linus Torvalds 已提交
1990 1991
}

1992
/* Check if coalescing SKBs is legal. */
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
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;
}

2011 2012 2013
/* Collapse packets in the retransmit queue to make to create
 * less packets on the wire. This is only done on retransmission.
 */
2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051
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
2052 2053 2054 2055 2056 2057 2058
/* 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
2059
	struct inet_connection_sock *icsk = inet_csk(sk);
2060
	unsigned int cur_mss;
L
Linus Torvalds 已提交
2061 2062
	int err;

J
John Heffner 已提交
2063 2064 2065 2066 2067
	/* Inconslusive MTU probe */
	if (icsk->icsk_mtup.probe_size) {
		icsk->icsk_mtup.probe_size = 0;
	}

L
Linus Torvalds 已提交
2068
	/* Do not sent more than we queued. 1/4 is reserved for possible
S
Stephen Hemminger 已提交
2069
	 * copying overhead: fragmentation, tunneling, mangling etc.
L
Linus Torvalds 已提交
2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081
	 */
	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;
	}

2082 2083 2084
	if (inet_csk(sk)->icsk_af_ops->rebuild_header(sk))
		return -EHOSTUNREACH; /* Routing failure or similar. */

I
Ilpo Järvinen 已提交
2085
	cur_mss = tcp_current_mss(sk);
2086

L
Linus Torvalds 已提交
2087 2088 2089 2090 2091
	/* 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.
	 */
2092 2093
	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
2094 2095 2096
		return -EAGAIN;

	if (skb->len > cur_mss) {
2097
		if (tcp_fragment(sk, skb, cur_mss, cur_mss))
L
Linus Torvalds 已提交
2098
			return -ENOMEM; /* We'll try again later. */
2099
	} else {
2100 2101 2102 2103 2104 2105
		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
2106 2107
	}

2108
	tcp_retrans_try_collapse(sk, skb, cur_mss);
L
Linus Torvalds 已提交
2109 2110 2111 2112 2113

	/* 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
2114 2115 2116
	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
2117
		if (!pskb_trim(skb, 0)) {
2118 2119 2120
			/* 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
2121 2122 2123 2124 2125 2126 2127 2128 2129
			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;

2130
	err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
2131 2132 2133

	if (err == 0) {
		/* Update global TCP statistics. */
P
Pavel Emelyanov 已提交
2134
		TCP_INC_STATS(sock_net(sk), TCP_MIB_RETRANSSEGS);
L
Linus Torvalds 已提交
2135 2136 2137 2138

		tp->total_retrans++;

#if FASTRETRANS_DEBUG > 0
2139
		if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_RETRANS) {
L
Linus Torvalds 已提交
2140 2141 2142 2143
			if (net_ratelimit())
				printk(KERN_DEBUG "retrans_out leaked.\n");
		}
#endif
2144 2145
		if (!tp->retrans_out)
			tp->lost_retrans_low = tp->snd_nxt;
L
Linus Torvalds 已提交
2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162
		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;
}

2163 2164 2165
/* Check if we forward retransmits are possible in the current
 * window/congestion state.
 */
2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192
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
2193 2194 2195 2196 2197 2198 2199 2200 2201 2202
/* 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)
{
2203
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2204 2205
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
2206
	struct sk_buff *hole = NULL;
2207
	u32 last_lost;
2208
	int mib_idx;
2209
	int fwd_rexmitting = 0;
2210

2211 2212 2213
	if (!tp->lost_out)
		tp->retransmit_high = tp->snd_una;

2214
	if (tp->retransmit_skb_hint) {
2215
		skb = tp->retransmit_skb_hint;
2216 2217 2218 2219
		last_lost = TCP_SKB_CB(skb)->end_seq;
		if (after(last_lost, tp->retransmit_high))
			last_lost = tp->retransmit_high;
	} else {
2220
		skb = tcp_write_queue_head(sk);
2221 2222
		last_lost = tp->snd_una;
	}
L
Linus Torvalds 已提交
2223

2224 2225
	tcp_for_write_queue_from(skb, sk) {
		__u8 sacked = TCP_SKB_CB(skb)->sacked;
L
Linus Torvalds 已提交
2226

2227 2228 2229
		if (skb == tcp_send_head(sk))
			break;
		/* we could do better than to assign each time */
2230 2231
		if (hole == NULL)
			tp->retransmit_skb_hint = skb;
2232 2233 2234 2235 2236 2237 2238 2239 2240 2241

		/* 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
2242

2243 2244 2245 2246 2247
		if (fwd_rexmitting) {
begin_fwd:
			if (!before(TCP_SKB_CB(skb)->seq, tcp_highest_sack_seq(tp)))
				break;
			mib_idx = LINUX_MIB_TCPFORWARDRETRANS;
2248

2249
		} else if (!before(TCP_SKB_CB(skb)->seq, tp->retransmit_high)) {
2250
			tp->retransmit_high = last_lost;
2251 2252 2253 2254 2255 2256 2257 2258 2259
			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
2260

2261
		} else if (!(sacked & TCPCB_LOST)) {
2262
			if (hole == NULL && !(sacked & (TCPCB_SACKED_RETRANS|TCPCB_SACKED_ACKED)))
2263 2264
				hole = skb;
			continue;
L
Linus Torvalds 已提交
2265

2266
		} else {
2267
			last_lost = TCP_SKB_CB(skb)->end_seq;
2268 2269 2270 2271 2272
			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
2273

2274
		if (sacked & (TCPCB_SACKED_ACKED|TCPCB_SACKED_RETRANS))
L
Linus Torvalds 已提交
2275 2276
			continue;

2277
		if (tcp_retransmit_skb(sk, skb))
2278 2279
			return;
		NET_INC_STATS_BH(sock_net(sk), mib_idx);
L
Linus Torvalds 已提交
2280

2281
		if (skb == tcp_write_queue_head(sk))
2282 2283 2284
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
						  inet_csk(sk)->icsk_rto,
						  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2285 2286 2287 2288 2289 2290 2291 2292
	}
}

/* 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)
{
2293
	struct tcp_sock *tp = tcp_sk(sk);
2294
	struct sk_buff *skb = tcp_write_queue_tail(sk);
L
Linus Torvalds 已提交
2295
	int mss_now;
2296

L
Linus Torvalds 已提交
2297 2298 2299 2300
	/* 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
2301
	mss_now = tcp_current_mss(sk);
L
Linus Torvalds 已提交
2302

2303
	if (tcp_send_head(sk) != NULL) {
L
Linus Torvalds 已提交
2304 2305 2306 2307 2308 2309
		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 (;;) {
2310 2311
			skb = alloc_skb_fclone(MAX_TCP_HEADER,
					       sk->sk_allocation);
L
Linus Torvalds 已提交
2312 2313 2314 2315 2316 2317 2318 2319
			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(). */
2320 2321
		tcp_init_nondata_skb(skb, tp->write_seq,
				     TCPCB_FLAG_ACK | TCPCB_FLAG_FIN);
L
Linus Torvalds 已提交
2322 2323
		tcp_queue_skb(sk, skb);
	}
2324
	__tcp_push_pending_frames(sk, mss_now, TCP_NAGLE_OFF);
L
Linus Torvalds 已提交
2325 2326 2327 2328 2329
}

/* 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
2330
 * by RFC 2525, section 2.17.  -DaveM
L
Linus Torvalds 已提交
2331
 */
A
Al Viro 已提交
2332
void tcp_send_active_reset(struct sock *sk, gfp_t priority)
L
Linus Torvalds 已提交
2333 2334 2335 2336 2337 2338
{
	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
2339
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTFAILED);
L
Linus Torvalds 已提交
2340 2341 2342 2343 2344
		return;
	}

	/* Reserve space for headers and prepare control bits. */
	skb_reserve(skb, MAX_TCP_HEADER);
2345 2346
	tcp_init_nondata_skb(skb, tcp_acceptable_seq(sk),
			     TCPCB_FLAG_ACK | TCPCB_FLAG_RST);
L
Linus Torvalds 已提交
2347 2348
	/* Send it off. */
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2349
	if (tcp_transmit_skb(sk, skb, 0, priority))
P
Pavel Emelyanov 已提交
2350
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPABORTFAILED);
2351

P
Pavel Emelyanov 已提交
2352
	TCP_INC_STATS(sock_net(sk), TCP_MIB_OUTRSTS);
L
Linus Torvalds 已提交
2353 2354
}

2355 2356
/* 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
2357 2358 2359 2360 2361 2362
 * 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)
{
2363
	struct sk_buff *skb;
L
Linus Torvalds 已提交
2364

2365
	skb = tcp_write_queue_head(sk);
2366
	if (skb == NULL || !(TCP_SKB_CB(skb)->flags & TCPCB_FLAG_SYN)) {
L
Linus Torvalds 已提交
2367 2368 2369
		printk(KERN_DEBUG "tcp_send_synack: wrong queue state\n");
		return -EFAULT;
	}
2370
	if (!(TCP_SKB_CB(skb)->flags & TCPCB_FLAG_ACK)) {
L
Linus Torvalds 已提交
2371 2372 2373 2374
		if (skb_cloned(skb)) {
			struct sk_buff *nskb = skb_copy(skb, GFP_ATOMIC);
			if (nskb == NULL)
				return -ENOMEM;
2375
			tcp_unlink_write_queue(skb, sk);
L
Linus Torvalds 已提交
2376
			skb_header_release(nskb);
2377
			__tcp_add_write_queue_head(sk, nskb);
2378 2379 2380
			sk_wmem_free_skb(sk, skb);
			sk->sk_wmem_queued += nskb->truesize;
			sk_mem_charge(sk, nskb->truesize);
L
Linus Torvalds 已提交
2381 2382 2383 2384 2385 2386 2387
			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;
2388
	return tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
2389 2390
}

2391
/* Prepare a SYN-ACK. */
2392
struct sk_buff *tcp_make_synack(struct sock *sk, struct dst_entry *dst,
2393 2394
				struct request_sock *req,
				struct request_values *rvp)
L
Linus Torvalds 已提交
2395
{
2396
	struct tcp_out_options opts;
2397
	struct tcp_extend_values *xvp = tcp_xv(rvp);
2398
	struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
2399
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
2400
	const struct tcp_cookie_values *cvp = tp->cookie_values;
L
Linus Torvalds 已提交
2401 2402
	struct tcphdr *th;
	struct sk_buff *skb;
2403
	struct tcp_md5sig_key *md5;
2404
	int tcp_header_size;
2405
	int mss;
E
Eric Dumazet 已提交
2406
	int s_data_desired = 0;
L
Linus Torvalds 已提交
2407

E
Eric Dumazet 已提交
2408 2409 2410
	if (cvp != NULL && cvp->s_data_constant && cvp->s_data_desired)
		s_data_desired = cvp->s_data_desired;
	skb = sock_wmalloc(sk, MAX_TCP_HEADER + 15 + s_data_desired, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
2411 2412 2413 2414 2415 2416
	if (skb == NULL)
		return NULL;

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

E
Eric Dumazet 已提交
2417
	skb_dst_set(skb, dst_clone(dst));
L
Linus Torvalds 已提交
2418

2419 2420 2421 2422
	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
2423 2424 2425 2426 2427 2428
	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),
2429
			mss - (ireq->tstamp_ok ? TCPOLEN_TSTAMP_ALIGNED : 0),
A
Adam Langley 已提交
2430 2431 2432
			&req->rcv_wnd,
			&req->window_clamp,
			ireq->wscale_ok,
2433 2434
			&rcv_wscale,
			dst_metric(dst, RTAX_INITRWND));
A
Adam Langley 已提交
2435 2436 2437 2438
		ireq->rcv_wscale = rcv_wscale;
	}

	memset(&opts, 0, sizeof(opts));
2439 2440 2441 2442 2443
#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
2444
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2445
	tcp_header_size = tcp_synack_options(sk, req, mss,
2446 2447
					     skb, &opts, &md5, xvp)
			+ sizeof(*th);
2448

2449 2450
	skb_push(skb, tcp_header_size);
	skb_reset_transport_header(skb);
L
Linus Torvalds 已提交
2451

2452
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
2453 2454 2455 2456
	memset(th, 0, sizeof(struct tcphdr));
	th->syn = 1;
	th->ack = 1;
	TCP_ECN_make_synack(req, th);
2457
	th->source = ireq->loc_port;
2458
	th->dest = ireq->rmt_port;
2459 2460 2461 2462 2463
	/* 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);
2464 2465

	if (OPTION_COOKIE_EXTENSION & opts.options) {
E
Eric Dumazet 已提交
2466 2467
		if (s_data_desired) {
			u8 *buf = skb_put(skb, s_data_desired);
2468 2469

			/* copy data directly from the listening socket. */
E
Eric Dumazet 已提交
2470 2471
			memcpy(buf, cvp->s_data_payload, s_data_desired);
			TCP_SKB_CB(skb)->end_seq += s_data_desired;
2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488
		}

		if (opts.hash_size > 0) {
			__u32 workspace[SHA_WORKSPACE_WORDS];
			u32 *mess = &xvp->cookie_bakery[COOKIE_DIGEST_WORDS];
			u32 *tail = &mess[COOKIE_MESSAGE_WORDS-1];

			/* Secret recipe depends on the Timestamp, (future)
			 * Sequence and Acknowledgment Numbers, Initiator
			 * Cookie, and others handled by IP variant caller.
			 */
			*tail-- ^= opts.tsval;
			*tail-- ^= tcp_rsk(req)->rcv_isn + 1;
			*tail-- ^= TCP_SKB_CB(skb)->seq + 1;

			/* recommended */
			*tail-- ^= ((th->dest << 16) | th->source);
D
David S. Miller 已提交
2489
			*tail-- ^= (u32)(unsigned long)cvp; /* per sockopt */
2490 2491 2492 2493 2494 2495 2496 2497 2498

			sha_transform((__u32 *)&xvp->cookie_bakery[0],
				      (char *)mess,
				      &workspace[0]);
			opts.hash_location =
				(__u8 *)&xvp->cookie_bakery[0];
		}
	}

L
Linus Torvalds 已提交
2499
	th->seq = htonl(TCP_SKB_CB(skb)->seq);
2500
	th->ack_seq = htonl(tcp_rsk(req)->rcv_isn + 1);
L
Linus Torvalds 已提交
2501 2502

	/* RFC1323: The window in SYN & SYN/ACK segments is never scaled. */
2503
	th->window = htons(min(req->rcv_wnd, 65535U));
2504
	tcp_options_write((__be32 *)(th + 1), tp, &opts);
L
Linus Torvalds 已提交
2505
	th->doff = (tcp_header_size >> 2);
P
Pavel Emelyanov 已提交
2506
	TCP_INC_STATS(sock_net(sk), TCP_MIB_OUTSEGS);
2507 2508 2509 2510

#ifdef CONFIG_TCP_MD5SIG
	/* Okay, we have all we need - do the md5 hash if needed */
	if (md5) {
2511
		tcp_rsk(req)->af_specific->calc_md5_hash(opts.hash_location,
2512
					       md5, NULL, req, skb);
2513 2514 2515
	}
#endif

L
Linus Torvalds 已提交
2516 2517 2518
	return skb;
}

2519
/* Do all connect socket setups that can be done AF independent. */
S
Stephen Hemminger 已提交
2520
static void tcp_connect_init(struct sock *sk)
L
Linus Torvalds 已提交
2521 2522 2523 2524 2525 2526 2527 2528 2529
{
	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) +
2530
		(sysctl_tcp_timestamps ? TCPOLEN_TSTAMP_ALIGNED : 0);
L
Linus Torvalds 已提交
2531

2532 2533 2534 2535 2536
#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
2537 2538 2539 2540
	/* 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
2541
	tcp_mtup_init(sk);
L
Linus Torvalds 已提交
2542 2543 2544 2545 2546
	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);
2547 2548 2549
	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
2550 2551 2552 2553 2554 2555
	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,
2556
				  sysctl_tcp_window_scaling,
2557 2558
				  &rcv_wscale,
				  dst_metric(dst, RTAX_INITRWND));
L
Linus Torvalds 已提交
2559 2560 2561 2562 2563 2564 2565

	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;
2566
	tcp_init_wl(tp, 0);
L
Linus Torvalds 已提交
2567 2568
	tp->snd_una = tp->write_seq;
	tp->snd_sml = tp->write_seq;
I
Ilpo Järvinen 已提交
2569
	tp->snd_up = tp->write_seq;
L
Linus Torvalds 已提交
2570 2571 2572 2573
	tp->rcv_nxt = 0;
	tp->rcv_wup = 0;
	tp->copied_seq = 0;

2574 2575
	inet_csk(sk)->icsk_rto = TCP_TIMEOUT_INIT;
	inet_csk(sk)->icsk_retransmits = 0;
L
Linus Torvalds 已提交
2576 2577 2578
	tcp_clear_retrans(tp);
}

2579
/* Build a SYN and send it off. */
L
Linus Torvalds 已提交
2580 2581 2582 2583 2584 2585 2586
int tcp_connect(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *buff;

	tcp_connect_init(sk);

2587
	buff = alloc_skb_fclone(MAX_TCP_HEADER + 15, sk->sk_allocation);
L
Linus Torvalds 已提交
2588 2589 2590 2591 2592 2593
	if (unlikely(buff == NULL))
		return -ENOBUFS;

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

2594
	tp->snd_nxt = tp->write_seq;
2595 2596
	tcp_init_nondata_skb(buff, tp->write_seq++, TCPCB_FLAG_SYN);
	TCP_ECN_send_syn(sk, buff);
L
Linus Torvalds 已提交
2597 2598 2599 2600 2601

	/* Send it off. */
	TCP_SKB_CB(buff)->when = tcp_time_stamp;
	tp->retrans_stamp = TCP_SKB_CB(buff)->when;
	skb_header_release(buff);
2602
	__tcp_add_write_queue_tail(sk, buff);
2603 2604
	sk->sk_wmem_queued += buff->truesize;
	sk_mem_charge(sk, buff->truesize);
L
Linus Torvalds 已提交
2605
	tp->packets_out += tcp_skb_pcount(buff);
2606
	tcp_transmit_skb(sk, buff, 1, sk->sk_allocation);
2607 2608 2609 2610 2611 2612

	/* 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
2613
	TCP_INC_STATS(sock_net(sk), TCP_MIB_ACTIVEOPENS);
L
Linus Torvalds 已提交
2614 2615

	/* Timer for repeating the SYN until an answer. */
2616 2617
	inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
				  inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
L
Linus Torvalds 已提交
2618 2619 2620 2621 2622 2623 2624 2625 2626
	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)
{
2627 2628
	struct inet_connection_sock *icsk = inet_csk(sk);
	int ato = icsk->icsk_ack.ato;
L
Linus Torvalds 已提交
2629 2630 2631
	unsigned long timeout;

	if (ato > TCP_DELACK_MIN) {
2632
		const struct tcp_sock *tp = tcp_sk(sk);
2633
		int max_ato = HZ / 2;
L
Linus Torvalds 已提交
2634

2635 2636
		if (icsk->icsk_ack.pingpong ||
		    (icsk->icsk_ack.pending & ICSK_ACK_PUSHED))
L
Linus Torvalds 已提交
2637 2638 2639 2640 2641
			max_ato = TCP_DELACK_MAX;

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

		/* If some rtt estimate is known, use it to bound delayed ack.
2642
		 * Do not use inet_csk(sk)->icsk_rto here, use results of rtt measurements
L
Linus Torvalds 已提交
2643 2644 2645
		 * directly.
		 */
		if (tp->srtt) {
2646
			int rtt = max(tp->srtt >> 3, TCP_DELACK_MIN);
L
Linus Torvalds 已提交
2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658

			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. */
2659
	if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) {
L
Linus Torvalds 已提交
2660 2661 2662
		/* If delack timer was blocked or is about to expire,
		 * send ACK now.
		 */
2663 2664
		if (icsk->icsk_ack.blocked ||
		    time_before_eq(icsk->icsk_ack.timeout, jiffies + (ato >> 2))) {
L
Linus Torvalds 已提交
2665 2666 2667 2668
			tcp_send_ack(sk);
			return;
		}

2669 2670
		if (!time_before(timeout, icsk->icsk_ack.timeout))
			timeout = icsk->icsk_ack.timeout;
L
Linus Torvalds 已提交
2671
	}
2672 2673 2674
	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
2675 2676 2677 2678 2679
}

/* This routine sends an ack and also updates the window. */
void tcp_send_ack(struct sock *sk)
{
2680
	struct sk_buff *buff;
L
Linus Torvalds 已提交
2681

2682 2683 2684
	/* If we have been reset, we may not send again. */
	if (sk->sk_state == TCP_CLOSE)
		return;
L
Linus Torvalds 已提交
2685

2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696
	/* 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
2697
	}
2698 2699 2700

	/* Reserve space for headers and prepare control bits. */
	skb_reserve(buff, MAX_TCP_HEADER);
2701
	tcp_init_nondata_skb(buff, tcp_acceptable_seq(sk), TCPCB_FLAG_ACK);
2702 2703 2704 2705

	/* 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
2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725
}

/* 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);
2726
	if (skb == NULL)
L
Linus Torvalds 已提交
2727 2728 2729 2730 2731 2732 2733 2734
		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.
	 */
2735
	tcp_init_nondata_skb(skb, tp->snd_una - !urgent, TCPCB_FLAG_ACK);
L
Linus Torvalds 已提交
2736
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2737
	return tcp_transmit_skb(sk, skb, 0, GFP_ATOMIC);
L
Linus Torvalds 已提交
2738 2739
}

2740
/* Initiate keepalive or window probe from timer. */
L
Linus Torvalds 已提交
2741 2742
int tcp_write_wakeup(struct sock *sk)
{
2743 2744
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
L
Linus Torvalds 已提交
2745

2746 2747 2748 2749 2750 2751
	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
2752
		unsigned int mss = tcp_current_mss(sk);
2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764
		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
2765
			TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777
			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
2778
		if (between(tp->snd_up, tp->snd_una + 1, tp->snd_una + 0xFFFF))
2779 2780
			tcp_xmit_probe_skb(sk, 1);
		return tcp_xmit_probe_skb(sk, 0);
L
Linus Torvalds 已提交
2781 2782 2783 2784 2785 2786 2787 2788
	}
}

/* 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)
{
2789
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2790 2791 2792 2793 2794
	struct tcp_sock *tp = tcp_sk(sk);
	int err;

	err = tcp_write_wakeup(sk);

2795
	if (tp->packets_out || !tcp_send_head(sk)) {
L
Linus Torvalds 已提交
2796
		/* Cancel probe timer, if it is not required. */
2797
		icsk->icsk_probes_out = 0;
2798
		icsk->icsk_backoff = 0;
L
Linus Torvalds 已提交
2799 2800 2801 2802
		return;
	}

	if (err <= 0) {
2803 2804
		if (icsk->icsk_backoff < sysctl_tcp_retries2)
			icsk->icsk_backoff++;
2805
		icsk->icsk_probes_out++;
2806
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
2807 2808
					  min(icsk->icsk_rto << icsk->icsk_backoff, TCP_RTO_MAX),
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2809 2810
	} else {
		/* If packet was not sent due to local congestion,
2811
		 * do not backoff and do not remember icsk_probes_out.
L
Linus Torvalds 已提交
2812 2813 2814 2815
		 * Let local senders to fight for local resources.
		 *
		 * Use accumulated backoff yet.
		 */
2816 2817
		if (!icsk->icsk_probes_out)
			icsk->icsk_probes_out = 1;
2818
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
2819
					  min(icsk->icsk_rto << icsk->icsk_backoff,
2820 2821
					      TCP_RESOURCE_PROBE_INTERVAL),
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2822 2823 2824
	}
}

2825
EXPORT_SYMBOL(tcp_select_initial_window);
L
Linus Torvalds 已提交
2826 2827 2828 2829
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
2830
EXPORT_SYMBOL(tcp_mtup_init);