tcp_output.c 72.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 * INET		An implementation of the TCP/IP protocol suite for the LINUX
 *		operating system.  INET is implemented using the  BSD Socket
 *		interface as the means of communication with the user level.
 *
 *		Implementation of the Transmission Control Protocol(TCP).
 *
 * Version:	$Id: tcp_output.c,v 1.146 2002/02/01 22:01:04 davem Exp $
 *
10
 * Authors:	Ross Biro
L
Linus Torvalds 已提交
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
 *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 *		Mark Evans, <evansmp@uhura.aston.ac.uk>
 *		Corey Minyard <wf-rch!minyard@relay.EU.net>
 *		Florian La Roche, <flla@stud.uni-sb.de>
 *		Charles Hedrick, <hedrick@klinzhai.rutgers.edu>
 *		Linus Torvalds, <torvalds@cs.helsinki.fi>
 *		Alan Cox, <gw4pts@gw4pts.ampr.org>
 *		Matthew Dillon, <dillon@apollo.west.oic.com>
 *		Arnt Gulbrandsen, <agulbra@nvg.unit.no>
 *		Jorge Cwik, <jorge@laser.satlink.net>
 */

/*
 * Changes:	Pedro Roque	:	Retransmit queue handled by TCP.
 *				:	Fragmentation on mtu decrease
 *				:	Segment collapse on retransmit
 *				:	AF independence
 *
 *		Linus Torvalds	:	send_delayed_ack
 *		David S. Miller	:	Charge memory using the right skb
 *					during syn/ack processing.
 *		David S. Miller :	Output engine completely rewritten.
 *		Andrea Arcangeli:	SYNACK carry ts_recent in tsecr.
 *		Cacophonix Gaul :	draft-minshall-nagle-01
 *		J Hadi Salim	:	ECN support
 *
 */

#include <net/tcp.h>

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

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

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

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

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

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

64
static void update_send_head(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
65
{
66 67
	struct tcp_sock *tp = tcp_sk(sk);

68
	tcp_advance_send_head(sk, skb);
L
Linus Torvalds 已提交
69
	tp->snd_nxt = TCP_SKB_CB(skb)->end_seq;
70
	tcp_packets_out_inc(sk, skb);
L
Linus Torvalds 已提交
71 72 73 74 75 76 77 78
}

/* 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:
 */
79
static inline __u32 tcp_acceptable_seq(struct sock *sk)
L
Linus Torvalds 已提交
80
{
81 82
	struct tcp_sock *tp = tcp_sk(sk);

L
Linus Torvalds 已提交
83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
	if (!before(tp->snd_una+tp->snd_wnd, tp->snd_nxt))
		return tp->snd_nxt;
	else
		return tp->snd_una+tp->snd_wnd;
}

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

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

	return (__u16)mss;
}

/* RFC2861. Reset CWND after idle period longer RTO to "restart window".
 * This is the first part of cwnd validation mechanism. */
119
static void tcp_cwnd_restart(struct sock *sk, struct dst_entry *dst)
L
Linus Torvalds 已提交
120
{
121
	struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
122 123 124 125
	s32 delta = tcp_time_stamp - tp->lsndtime;
	u32 restart_cwnd = tcp_init_cwnd(tp, dst);
	u32 cwnd = tp->snd_cwnd;

126
	tcp_ca_event(sk, CA_EVENT_CWND_RESTART);
L
Linus Torvalds 已提交
127

128
	tp->snd_ssthresh = tcp_current_ssthresh(sk);
L
Linus Torvalds 已提交
129 130
	restart_cwnd = min(restart_cwnd, cwnd);

131
	while ((delta -= inet_csk(sk)->icsk_rto) > 0 && cwnd > restart_cwnd)
L
Linus Torvalds 已提交
132 133 134 135 136 137
		cwnd >>= 1;
	tp->snd_cwnd = max(cwnd, restart_cwnd);
	tp->snd_cwnd_stamp = tcp_time_stamp;
	tp->snd_cwnd_used = 0;
}

S
Stephen Hemminger 已提交
138 139
static void tcp_event_data_sent(struct tcp_sock *tp,
				struct sk_buff *skb, struct sock *sk)
L
Linus Torvalds 已提交
140
{
141 142
	struct inet_connection_sock *icsk = inet_csk(sk);
	const u32 now = tcp_time_stamp;
L
Linus Torvalds 已提交
143

144 145
	if (sysctl_tcp_slow_start_after_idle &&
	    (!tp->packets_out && (s32)(now - tp->lsndtime) > icsk->icsk_rto))
146
		tcp_cwnd_restart(sk, __sk_dst_get(sk));
L
Linus Torvalds 已提交
147 148 149 150 151 152

	tp->lsndtime = now;

	/* If it is a reply for ato after last received
	 * packet, enter pingpong mode.
	 */
153 154
	if ((u32)(now - icsk->icsk_ack.lrcvtime) < icsk->icsk_ack.ato)
		icsk->icsk_ack.pingpong = 1;
L
Linus Torvalds 已提交
155 156
}

S
Stephen Hemminger 已提交
157
static inline void tcp_event_ack_sent(struct sock *sk, unsigned int pkts)
L
Linus Torvalds 已提交
158
{
159 160
	tcp_dec_quickack_mode(sk, pkts);
	inet_csk_clear_xmit_timer(sk, ICSK_TIME_DACK);
L
Linus Torvalds 已提交
161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185
}

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

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

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

	/* NOTE: offering an initial window larger than 32767
186 187 188 189 190 191
	 * 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
192
	 */
193 194 195 196 197
	if (sysctl_tcp_workaround_signed_windows)
		(*rcv_wnd) = min(space, MAX_TCP_WINDOW);
	else
		(*rcv_wnd) = space;

L
Linus Torvalds 已提交
198 199 200
	(*rcv_wscale) = 0;
	if (wscale_ok) {
		/* Set window scaling on max possible window
201
		 * See RFC1323 for an explanation of the limit to 14
L
Linus Torvalds 已提交
202 203
		 */
		space = max_t(u32, sysctl_tcp_rmem[2], sysctl_rmem_max);
204
		space = min_t(u32, space, *window_clamp);
L
Linus Torvalds 已提交
205 206 207 208 209 210 211
		while (space > 65535 && (*rcv_wscale) < 14) {
			space >>= 1;
			(*rcv_wscale)++;
		}
	}

	/* Set initial window to value enough for senders,
212
	 * following RFC2414. Senders, not following this RFC,
L
Linus Torvalds 已提交
213 214 215
	 * will be satisfied with 2.
	 */
	if (mss > (1<<*rcv_wscale)) {
216 217
		int init_cwnd = 4;
		if (mss > 1460*3)
L
Linus Torvalds 已提交
218
			init_cwnd = 2;
219 220
		else if (mss > 1460)
			init_cwnd = 3;
L
Linus Torvalds 已提交
221 222 223 224 225 226 227 228 229 230 231 232 233
		if (*rcv_wnd > init_cwnd*mss)
			*rcv_wnd = init_cwnd*mss;
	}

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

/* Chose a new window to advertise, update state in tcp_sock for the
 * socket, and return result with RFC1323 scaling applied.  The return
 * value can be stuffed directly into th->window for an outgoing
 * frame.
 */
S
Stephen Hemminger 已提交
234
static u16 tcp_select_window(struct sock *sk)
L
Linus Torvalds 已提交
235 236 237 238 239 240
{
	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
241
	if (new_win < cur_win) {
L
Linus Torvalds 已提交
242 243 244 245 246 247 248 249 250 251 252 253 254 255 256
		/* Danger Will Robinson!
		 * Don't update rcv_wup/rcv_wnd here or else
		 * we will not be able to advertise a zero
		 * window in time.  --DaveM
		 *
		 * Relax Will Robinson.
		 */
		new_win = cur_win;
	}
	tp->rcv_wnd = new_win;
	tp->rcv_wup = tp->rcv_nxt;

	/* Make sure we do not exceed the maximum possible
	 * scaled window.
	 */
257
	if (!tp->rx_opt.rcv_wscale && sysctl_tcp_workaround_signed_windows)
L
Linus Torvalds 已提交
258 259 260 261 262 263 264 265 266 267 268 269 270 271
		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;
}

272
static void tcp_build_and_update_options(__be32 *ptr, struct tcp_sock *tp,
273
					 __u32 tstamp, __u8 **md5_hash)
S
Stephen Hemminger 已提交
274 275
{
	if (tp->rx_opt.tstamp_ok) {
276 277 278 279
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
			       (TCPOPT_TIMESTAMP << 8) |
			       TCPOLEN_TIMESTAMP);
S
Stephen Hemminger 已提交
280 281 282 283 284 285 286 287 288 289 290 291
		*ptr++ = htonl(tstamp);
		*ptr++ = htonl(tp->rx_opt.ts_recent);
	}
	if (tp->rx_opt.eff_sacks) {
		struct tcp_sack_block *sp = tp->rx_opt.dsack ? tp->duplicate_sack : tp->selective_acks;
		int this_sack;

		*ptr++ = htonl((TCPOPT_NOP  << 24) |
			       (TCPOPT_NOP  << 16) |
			       (TCPOPT_SACK <<  8) |
			       (TCPOLEN_SACK_BASE + (tp->rx_opt.eff_sacks *
						     TCPOLEN_SACK_PERBLOCK)));
S
Stephen Hemminger 已提交
292 293

		for (this_sack = 0; this_sack < tp->rx_opt.eff_sacks; this_sack++) {
S
Stephen Hemminger 已提交
294 295 296
			*ptr++ = htonl(sp[this_sack].start_seq);
			*ptr++ = htonl(sp[this_sack].end_seq);
		}
S
Stephen Hemminger 已提交
297

S
Stephen Hemminger 已提交
298 299 300 301 302
		if (tp->rx_opt.dsack) {
			tp->rx_opt.dsack = 0;
			tp->rx_opt.eff_sacks--;
		}
	}
303 304 305 306 307 308 309 310 311
#ifdef CONFIG_TCP_MD5SIG
	if (md5_hash) {
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
			       (TCPOPT_MD5SIG << 8) |
			       TCPOLEN_MD5SIG);
		*md5_hash = (__u8 *)ptr;
	}
#endif
S
Stephen Hemminger 已提交
312 313 314 315 316 317
}

/* Construct a tcp options header for a SYN or SYN_ACK packet.
 * If this is every changed make sure to change the definition of
 * MAX_SYN_SIZE to match the new maximum number of options that you
 * can generate.
318 319 320 321
 *
 * Note - that with the RFC2385 TCP option, we make room for the
 * 16 byte MD5 hash. This will be filled in later, so the pointer for the
 * location to be filled is passed back up.
S
Stephen Hemminger 已提交
322
 */
323
static void tcp_syn_build_options(__be32 *ptr, int mss, int ts, int sack,
S
Stephen Hemminger 已提交
324
				  int offer_wscale, int wscale, __u32 tstamp,
325
				  __u32 ts_recent, __u8 **md5_hash)
S
Stephen Hemminger 已提交
326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341
{
	/* We always get an MSS option.
	 * The option bytes which will be seen in normal data
	 * packets should timestamps be used, must be in the MSS
	 * advertised.  But we subtract them from tp->mss_cache so
	 * that calculations in tcp_sendmsg are simpler etc.
	 * So account for this fact here if necessary.  If we
	 * don't do this correctly, as a receiver we won't
	 * recognize data packets as being full sized when we
	 * should, and thus we won't abide by the delayed ACK
	 * rules correctly.
	 * SACKs don't matter, we never delay an ACK when we
	 * have any of those going out.
	 */
	*ptr++ = htonl((TCPOPT_MSS << 24) | (TCPOLEN_MSS << 16) | mss);
	if (ts) {
S
Stephen Hemminger 已提交
342
		if (sack)
343 344 345 346
			*ptr++ = htonl((TCPOPT_SACK_PERM << 24) |
				       (TCPOLEN_SACK_PERM << 16) |
				       (TCPOPT_TIMESTAMP << 8) |
				       TCPOLEN_TIMESTAMP);
S
Stephen Hemminger 已提交
347
		else
348 349 350 351
			*ptr++ = htonl((TCPOPT_NOP << 24) |
				       (TCPOPT_NOP << 16) |
				       (TCPOPT_TIMESTAMP << 8) |
				       TCPOLEN_TIMESTAMP);
S
Stephen Hemminger 已提交
352 353
		*ptr++ = htonl(tstamp);		/* TSVAL */
		*ptr++ = htonl(ts_recent);	/* TSECR */
S
Stephen Hemminger 已提交
354
	} else if (sack)
355 356 357 358
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
			       (TCPOPT_SACK_PERM << 8) |
			       TCPOLEN_SACK_PERM);
S
Stephen Hemminger 已提交
359
	if (offer_wscale)
360 361 362 363
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_WINDOW << 16) |
			       (TCPOLEN_WINDOW << 8) |
			       (wscale));
364 365 366 367 368 369 370 371 372 373 374 375 376 377
#ifdef CONFIG_TCP_MD5SIG
	/*
	 * If MD5 is enabled, then we set the option, and include the size
	 * (always 18). The actual MD5 hash is added just before the
	 * packet is sent.
	 */
	if (md5_hash) {
		*ptr++ = htonl((TCPOPT_NOP << 24) |
			       (TCPOPT_NOP << 16) |
			       (TCPOPT_MD5SIG << 8) |
			       TCPOLEN_MD5SIG);
		*md5_hash = (__u8 *) ptr;
	}
#endif
S
Stephen Hemminger 已提交
378
}
L
Linus Torvalds 已提交
379 380 381 382 383 384 385 386 387 388 389 390

/* 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.
 */
391
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
392
{
393 394 395 396 397
	const struct inet_connection_sock *icsk = inet_csk(sk);
	struct inet_sock *inet;
	struct tcp_sock *tp;
	struct tcp_skb_cb *tcb;
	int tcp_header_size;
398 399 400 401
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *md5;
	__u8 *md5_hash_location;
#endif
402 403 404 405 406 407 408 409 410
	struct tcphdr *th;
	int sysctl_flags;
	int err;

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

	/* If congestion control is doing timestamping, we must
	 * take such a timestamp before we potentially clone/copy.
	 */
411
	if (icsk->icsk_ca_ops->flags & TCP_CONG_RTT_STAMP)
412 413 414 415 416 417 418 419 420 421
		__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
422

423 424 425 426
	inet = inet_sk(sk);
	tp = tcp_sk(sk);
	tcb = TCP_SKB_CB(skb);
	tcp_header_size = tp->tcp_header_len;
L
Linus Torvalds 已提交
427 428 429 430 431

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

432 433 434
	sysctl_flags = 0;
	if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
		tcp_header_size = sizeof(struct tcphdr) + TCPOLEN_MSS;
S
Stephen Hemminger 已提交
435
		if (sysctl_tcp_timestamps) {
436 437
			tcp_header_size += TCPOLEN_TSTAMP_ALIGNED;
			sysctl_flags |= SYSCTL_FLAG_TSTAMPS;
L
Linus Torvalds 已提交
438
		}
439 440 441
		if (sysctl_tcp_window_scaling) {
			tcp_header_size += TCPOLEN_WSCALE_ALIGNED;
			sysctl_flags |= SYSCTL_FLAG_WSCALE;
L
Linus Torvalds 已提交
442
		}
443 444 445 446
		if (sysctl_tcp_sack) {
			sysctl_flags |= SYSCTL_FLAG_SACK;
			if (!(sysctl_flags & SYSCTL_FLAG_TSTAMPS))
				tcp_header_size += TCPOLEN_SACKPERM_ALIGNED;
L
Linus Torvalds 已提交
447
		}
448 449 450 451 452 453 454 455
	} else if (unlikely(tp->rx_opt.eff_sacks)) {
		/* A SACK is 2 pad bytes, a 2 byte header, plus
		 * 2 32-bit sequence numbers for each SACK block.
		 */
		tcp_header_size += (TCPOLEN_SACK_BASE_ALIGNED +
				    (tp->rx_opt.eff_sacks *
				     TCPOLEN_SACK_PERBLOCK));
	}
456

457 458 459
	if (tcp_packets_in_flight(tp) == 0)
		tcp_ca_event(sk, CA_EVENT_TX_START);

460 461 462 463 464 465 466 467 468 469
#ifdef CONFIG_TCP_MD5SIG
	/*
	 * Are we doing MD5 on this segment? If so - make
	 * room for it.
	 */
	md5 = tp->af_specific->md5_lookup(sk, sk);
	if (md5)
		tcp_header_size += TCPOLEN_MD5SIG_ALIGNED;
#endif

470 471
	skb_push(skb, tcp_header_size);
	skb_reset_transport_header(skb);
472
	skb_set_owner_w(skb, sk);
473 474

	/* Build TCP header and checksum it. */
475
	th = tcp_hdr(skb);
476 477 478 479
	th->source		= inet->sport;
	th->dest		= inet->dport;
	th->seq			= htonl(tcb->seq);
	th->ack_seq		= htonl(tp->rcv_nxt);
480
	*(((__be16 *)th) + 6)	= htons(((tcp_header_size >> 2) << 12) |
481 482 483 484 485 486
					tcb->flags);

	if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
		/* RFC1323: The window in SYN & SYN/ACK segments
		 * is never scaled.
		 */
487
		th->window	= htons(min(tp->rcv_wnd, 65535U));
488 489 490 491 492
	} else {
		th->window	= htons(tcp_select_window(sk));
	}
	th->check		= 0;
	th->urg_ptr		= 0;
L
Linus Torvalds 已提交
493

494 495 496 497 498
	if (unlikely(tp->urg_mode &&
		     between(tp->snd_up, tcb->seq+1, tcb->seq+0xFFFF))) {
		th->urg_ptr		= htons(tp->snd_up-tcb->seq);
		th->urg			= 1;
	}
L
Linus Torvalds 已提交
499

500
	if (unlikely(tcb->flags & TCPCB_FLAG_SYN)) {
501
		tcp_syn_build_options((__be32 *)(th + 1),
502 503 504 505 506 507
				      tcp_advertise_mss(sk),
				      (sysctl_flags & SYSCTL_FLAG_TSTAMPS),
				      (sysctl_flags & SYSCTL_FLAG_SACK),
				      (sysctl_flags & SYSCTL_FLAG_WSCALE),
				      tp->rx_opt.rcv_wscale,
				      tcb->when,
508 509 510 511 512 513
				      tp->rx_opt.ts_recent,

#ifdef CONFIG_TCP_MD5SIG
				      md5 ? &md5_hash_location :
#endif
				      NULL);
514
	} else {
515
		tcp_build_and_update_options((__be32 *)(th + 1),
516 517 518 519 520
					     tp, tcb->when,
#ifdef CONFIG_TCP_MD5SIG
					     md5 ? &md5_hash_location :
#endif
					     NULL);
521
		TCP_ECN_send(sk, skb, tcp_header_size);
522
	}
L
Linus Torvalds 已提交
523

524 525 526 527 528 529
#ifdef CONFIG_TCP_MD5SIG
	/* Calculate the MD5 hash, as we have all we need now */
	if (md5) {
		tp->af_specific->calc_md5_hash(md5_hash_location,
					       md5,
					       sk, NULL, NULL,
530
					       tcp_hdr(skb),
531 532 533 534 535
					       sk->sk_protocol,
					       skb->len);
	}
#endif

536
	icsk->icsk_af_ops->send_check(sk, skb->len, skb);
L
Linus Torvalds 已提交
537

538 539
	if (likely(tcb->flags & TCPCB_FLAG_ACK))
		tcp_event_ack_sent(sk, tcp_skb_pcount(skb));
L
Linus Torvalds 已提交
540

541 542
	if (skb->len != tcp_header_size)
		tcp_event_data_sent(tp, skb, sk);
L
Linus Torvalds 已提交
543

544 545
	if (after(tcb->end_seq, tp->snd_nxt) || tcb->seq == tcb->end_seq)
		TCP_INC_STATS(TCP_MIB_OUTSEGS);
L
Linus Torvalds 已提交
546

547
	err = icsk->icsk_af_ops->queue_xmit(skb, 0);
548
	if (likely(err <= 0))
549 550
		return err;

551
	tcp_enter_cwr(sk, 1);
552

553
	return net_xmit_eval(err);
L
Linus Torvalds 已提交
554 555 556 557 558 559 560

#undef SYSCTL_FLAG_TSTAMPS
#undef SYSCTL_FLAG_WSCALE
#undef SYSCTL_FLAG_SACK
}


561
/* This routine just queue's the buffer
L
Linus Torvalds 已提交
562 563 564 565 566 567 568 569 570 571 572
 *
 * 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);
573
	tcp_add_write_queue_tail(sk, skb);
L
Linus Torvalds 已提交
574 575 576
	sk_charge_skb(sk, skb);
}

577
static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb, unsigned int mss_now)
578
{
579
	if (skb->len <= mss_now || !sk_can_gso(sk)) {
580 581 582
		/* Avoid the costly divide in the normal
		 * non-TSO case.
		 */
583 584 585
		skb_shinfo(skb)->gso_segs = 1;
		skb_shinfo(skb)->gso_size = 0;
		skb_shinfo(skb)->gso_type = 0;
586 587 588
	} else {
		unsigned int factor;

589 590
		factor = skb->len + (mss_now - 1);
		factor /= mss_now;
591 592
		skb_shinfo(skb)->gso_segs = factor;
		skb_shinfo(skb)->gso_size = mss_now;
593
		skb_shinfo(skb)->gso_type = sk->sk_gso_type;
L
Linus Torvalds 已提交
594 595 596 597 598
	}
}

/* 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
599
 * packet to the list.  This won't be called frequently, I hope.
L
Linus Torvalds 已提交
600 601
 * Remember, these are still headerless SKBs at this point.
 */
602
int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len, unsigned int mss_now)
L
Linus Torvalds 已提交
603 604 605
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *buff;
606
	int nsize, old_factor;
607
	int nlen;
L
Linus Torvalds 已提交
608 609
	u16 flags;

610
	BUG_ON(len > skb->len);
611

612
	clear_all_retrans_hints(tp);
L
Linus Torvalds 已提交
613 614 615 616 617 618 619 620 621 622 623 624 625
	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
626

627 628 629 630
	sk_charge_skb(sk, buff);
	nlen = skb->len - len - nsize;
	buff->truesize += nlen;
	skb->truesize -= nlen;
L
Linus Torvalds 已提交
631 632 633 634 635 636 637 638 639 640

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

	/* PSH and FIN should only be set in the second packet. */
	flags = TCP_SKB_CB(skb)->flags;
	TCP_SKB_CB(skb)->flags = flags & ~(TCPCB_FLAG_FIN|TCPCB_FLAG_PSH);
	TCP_SKB_CB(buff)->flags = flags;
641
	TCP_SKB_CB(buff)->sacked = TCP_SKB_CB(skb)->sacked;
L
Linus Torvalds 已提交
642 643
	TCP_SKB_CB(skb)->sacked &= ~TCPCB_AT_TAIL;

644
	if (!skb_shinfo(skb)->nr_frags && skb->ip_summed != CHECKSUM_PARTIAL) {
L
Linus Torvalds 已提交
645 646 647 648 649 650 651 652
		/* Copy and checksum data tail into the new buffer. */
		buff->csum = csum_partial_copy_nocheck(skb->data + len, skb_put(buff, nsize),
						       nsize, 0);

		skb_trim(skb, len);

		skb->csum = csum_block_sub(skb->csum, buff->csum, len);
	} else {
653
		skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
654 655 656 657 658 659 660 661 662
		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;
663
	buff->tstamp = skb->tstamp;
L
Linus Torvalds 已提交
664

665 666
	old_factor = tcp_skb_pcount(skb);

L
Linus Torvalds 已提交
667
	/* Fix up tso_factor for both original and new SKB.  */
668 669
	tcp_set_skb_tso_segs(sk, skb, mss_now);
	tcp_set_skb_tso_segs(sk, buff, mss_now);
L
Linus Torvalds 已提交
670

671 672 673
	/* If this packet has been sent out already, we must
	 * adjust the various packet counters.
	 */
674
	if (!before(tp->snd_nxt, TCP_SKB_CB(buff)->end_seq)) {
675 676
		int diff = old_factor - tcp_skb_pcount(skb) -
			tcp_skb_pcount(buff);
L
Linus Torvalds 已提交
677

678
		tp->packets_out -= diff;
679 680 681 682 683 684

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

685 686 687 688
		if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST) {
			tp->lost_out -= diff;
			tp->left_out -= diff;
		}
689

690
		if (diff > 0) {
691 692 693 694 695 696 697 698
			/* Adjust Reno SACK estimate. */
			if (!tp->rx_opt.sack_ok) {
				tp->sacked_out -= diff;
				if ((int)tp->sacked_out < 0)
					tp->sacked_out = 0;
				tcp_sync_left_out(tp);
			}

699 700 701
			tp->fackets_out -= diff;
			if ((int)tp->fackets_out < 0)
				tp->fackets_out = 0;
702 703 704 705 706 707 708 709
			/* SACK fastpath might overwrite it unless dealt with */
			if (tp->fastpath_skb_hint != NULL &&
			    after(TCP_SKB_CB(tp->fastpath_skb_hint)->seq,
				  TCP_SKB_CB(skb)->seq)) {
				tp->fastpath_cnt_hint -= diff;
				if ((int)tp->fastpath_cnt_hint < 0)
					tp->fastpath_cnt_hint = 0;
			}
710
		}
L
Linus Torvalds 已提交
711 712 713
	}

	/* Link BUFF into the send queue. */
714
	skb_header_release(buff);
715
	tcp_insert_write_queue_after(skb, buff, sk);
L
Linus Torvalds 已提交
716 717 718 719 720 721 722 723

	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.
 */
724
static void __pskb_trim_head(struct sk_buff *skb, int len)
L
Linus Torvalds 已提交
725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745
{
	int i, k, eat;

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

746
	skb_reset_tail_pointer(skb);
L
Linus Torvalds 已提交
747 748 749 750 751 752 753 754 755 756
	skb->data_len -= len;
	skb->len = skb->data_len;
}

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

757 758
	/* If len == headlen, we avoid __skb_pull to preserve alignment. */
	if (unlikely(len < skb_headlen(skb)))
L
Linus Torvalds 已提交
759
		__skb_pull(skb, len);
760 761
	else
		__pskb_trim_head(skb, len - skb_headlen(skb));
L
Linus Torvalds 已提交
762 763

	TCP_SKB_CB(skb)->seq += len;
764
	skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
765 766 767 768 769 770 771 772 773 774

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

	/* Any change of skb->len requires recalculation of tso
	 * factor and mss.
	 */
	if (tcp_skb_pcount(skb) > 1)
775
		tcp_set_skb_tso_segs(sk, skb, tcp_current_mss(sk, 1));
L
Linus Torvalds 已提交
776 777 778 779

	return 0;
}

J
John Heffner 已提交
780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830
/* Not accounting for SACKs here. */
int tcp_mtu_to_mss(struct sock *sk, int pmtu)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);
	int mss_now;

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

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

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

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

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

	return mss_now;
}

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

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

	return mtu;
}

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

	icsk->icsk_mtup.enabled = sysctl_tcp_mtu_probing > 1;
	icsk->icsk_mtup.search_high = tp->rx_opt.mss_clamp + sizeof(struct tcphdr) +
831
			       icsk->icsk_af_ops->net_header_len;
J
John Heffner 已提交
832 833 834 835
	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
836 837 838 839 840 841
/* 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
842
   It is minimum of user_mss and mss received with SYN.
L
Linus Torvalds 已提交
843 844
   It also does not include TCP options.

845
   inet_csk(sk)->icsk_pmtu_cookie is last pmtu, seen by this function.
L
Linus Torvalds 已提交
846 847 848 849 850 851 852 853 854

   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.

855 856
   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
857 858 859 860 861
 */

unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu)
{
	struct tcp_sock *tp = tcp_sk(sk);
862
	struct inet_connection_sock *icsk = inet_csk(sk);
J
John Heffner 已提交
863
	int mss_now;
L
Linus Torvalds 已提交
864

J
John Heffner 已提交
865 866
	if (icsk->icsk_mtup.search_high > pmtu)
		icsk->icsk_mtup.search_high = pmtu;
L
Linus Torvalds 已提交
867

J
John Heffner 已提交
868
	mss_now = tcp_mtu_to_mss(sk, pmtu);
L
Linus Torvalds 已提交
869 870 871 872 873 874

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

	/* And store cached results */
875
	icsk->icsk_pmtu_cookie = pmtu;
J
John Heffner 已提交
876 877
	if (icsk->icsk_mtup.enabled)
		mss_now = min(mss_now, tcp_mtu_to_mss(sk, icsk->icsk_mtup.search_low));
878
	tp->mss_cache = mss_now;
L
Linus Torvalds 已提交
879 880 881 882 883 884 885 886 887 888 889

	return mss_now;
}

/* Compute the current effective MSS, taking SACKs and IP options,
 * and even PMTU discovery events into account.
 *
 * LARGESEND note: !urg_mode is overkill, only frames up to snd_up
 * cannot be large. However, taking into account rare use of URG, this
 * is not a big flaw.
 */
890
unsigned int tcp_current_mss(struct sock *sk, int large_allowed)
L
Linus Torvalds 已提交
891 892 893
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct dst_entry *dst = __sk_dst_get(sk);
894 895 896 897 898 899
	u32 mss_now;
	u16 xmit_size_goal;
	int doing_tso = 0;

	mss_now = tp->mss_cache;

900
	if (large_allowed && sk_can_gso(sk) && !tp->urg_mode)
901
		doing_tso = 1;
L
Linus Torvalds 已提交
902 903 904

	if (dst) {
		u32 mtu = dst_mtu(dst);
905
		if (mtu != inet_csk(sk)->icsk_pmtu_cookie)
L
Linus Torvalds 已提交
906 907 908
			mss_now = tcp_sync_mss(sk, mtu);
	}

909 910 911
	if (tp->rx_opt.eff_sacks)
		mss_now -= (TCPOLEN_SACK_BASE_ALIGNED +
			    (tp->rx_opt.eff_sacks * TCPOLEN_SACK_PERBLOCK));
L
Linus Torvalds 已提交
912

913 914 915 916 917
#ifdef CONFIG_TCP_MD5SIG
	if (tp->af_specific->md5_lookup(sk, sk))
		mss_now -= TCPOLEN_MD5SIG_ALIGNED;
#endif

918
	xmit_size_goal = mss_now;
L
Linus Torvalds 已提交
919

920
	if (doing_tso) {
921 922
		xmit_size_goal = (65535 -
				  inet_csk(sk)->icsk_af_ops->net_header_len -
923 924
				  inet_csk(sk)->icsk_ext_hdr_len -
				  tp->tcp_header_len);
L
Linus Torvalds 已提交
925

926 927 928 929
		if (tp->max_window &&
		    (xmit_size_goal > (tp->max_window >> 1)))
			xmit_size_goal = max((tp->max_window >> 1),
					     68U - tp->tcp_header_len);
L
Linus Torvalds 已提交
930

931
		xmit_size_goal -= (xmit_size_goal % mss_now);
L
Linus Torvalds 已提交
932
	}
933
	tp->xmit_size_goal = xmit_size_goal;
L
Linus Torvalds 已提交
934 935 936 937

	return mss_now;
}

938 939
/* Congestion window validation. (RFC2861) */

940
static void tcp_cwnd_validate(struct sock *sk)
941
{
942
	struct tcp_sock *tp = tcp_sk(sk);
943 944 945 946 947 948 949 950 951 952 953
	__u32 packets_out = tp->packets_out;

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

954 955
		if (sysctl_tcp_slow_start_after_idle &&
		    (s32)(tcp_time_stamp - tp->snd_cwnd_stamp) >= inet_csk(sk)->icsk_rto)
956 957 958 959
			tcp_cwnd_application_limited(sk);
	}
}

960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976
static unsigned int tcp_window_allows(struct tcp_sock *tp, struct sk_buff *skb, unsigned int mss_now, unsigned int cwnd)
{
	u32 window, cwnd_len;

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

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

	/* Don't be strict about the congestion window for the final FIN.  */
977 978
	if ((TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN) &&
	    tcp_skb_pcount(skb) == 1)
979 980 981 982 983 984 985 986 987 988 989 990 991
		return 1;

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

	return 0;
}

/* This must be invoked the first time we consider transmitting
 * SKB onto the wire.
 */
S
Stephen Hemminger 已提交
992
static int tcp_init_tso_segs(struct sock *sk, struct sk_buff *skb, unsigned int mss_now)
993 994 995
{
	int tso_segs = tcp_skb_pcount(skb);

996 997
	if (!tso_segs ||
	    (tso_segs > 1 &&
998
	     tcp_skb_mss(skb) != mss_now)) {
999
		tcp_set_skb_tso_segs(sk, skb, mss_now);
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
		tso_segs = tcp_skb_pcount(skb);
	}
	return tso_segs;
}

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

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

static inline int tcp_nagle_check(const struct tcp_sock *tp,
1020
				  const struct sk_buff *skb,
1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044
				  unsigned mss_now, int nonagle)
{
	return (skb->len < mss_now &&
		((nonagle&TCP_NAGLE_CORK) ||
		 (!nonagle &&
		  tp->packets_out &&
		  tcp_minshall_check(tp))));
}

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

1045 1046 1047 1048
	/* Don't use the nagle rule for urgent data (or for the final FIN).
	 * Nagle can be ignored during F-RTO too (see RFC4138).
	 */
	if (tp->urg_mode || (tp->frto_counter == 2) ||
1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068
	    (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN))
		return 1;

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

	return 0;
}

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

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

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

1069
/* This checks if the data bearing packet SKB (usually tcp_send_head(sk))
1070 1071 1072 1073 1074 1075 1076 1077 1078
 * 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;

1079
	tcp_init_tso_segs(sk, skb, cur_mss);
1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091

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

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

	return cwnd_quota;
}

1092
int tcp_may_send_now(struct sock *sk)
1093
{
1094
	struct tcp_sock *tp = tcp_sk(sk);
1095
	struct sk_buff *skb = tcp_send_head(sk);
1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110

	return (skb &&
		tcp_snd_test(sk, skb, tcp_current_mss(sk, 1),
			     (tcp_skb_is_last(sk, skb) ?
			      TCP_NAGLE_PUSH :
			      tp->nonagle)));
}

/* 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).
 */
1111
static int tso_fragment(struct sock *sk, struct sk_buff *skb, unsigned int len, unsigned int mss_now)
1112 1113 1114 1115 1116 1117
{
	struct sk_buff *buff;
	int nlen = skb->len - len;
	u16 flags;

	/* All of a TSO frame must be composed of paged data.  */
1118 1119
	if (skb->len != skb->data_len)
		return tcp_fragment(sk, skb, len, mss_now);
1120 1121 1122 1123 1124

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

1125 1126
	sk_charge_skb(sk, buff);
	buff->truesize += nlen;
1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141
	skb->truesize -= nlen;

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

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

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

1142
	buff->ip_summed = skb->ip_summed = CHECKSUM_PARTIAL;
1143 1144 1145
	skb_split(skb, buff, len);

	/* Fix up tso_factor for both original and new SKB.  */
1146 1147
	tcp_set_skb_tso_segs(sk, skb, mss_now);
	tcp_set_skb_tso_segs(sk, buff, mss_now);
1148 1149 1150

	/* Link BUFF into the send queue. */
	skb_header_release(buff);
1151
	tcp_insert_write_queue_after(skb, buff, sk);
1152 1153 1154 1155 1156 1157 1158 1159 1160

	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.
 */
1161
static int tcp_tso_should_defer(struct sock *sk, struct sk_buff *skb)
1162
{
1163
	struct tcp_sock *tp = tcp_sk(sk);
1164
	const struct inet_connection_sock *icsk = inet_csk(sk);
1165 1166 1167
	u32 send_win, cong_win, limit, in_flight;

	if (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN)
J
John Heffner 已提交
1168
		goto send_now;
1169

1170
	if (icsk->icsk_ca_state != TCP_CA_Open)
J
John Heffner 已提交
1171 1172 1173 1174 1175
		goto send_now;

	/* Defer for less than two clock ticks. */
	if (!tp->tso_deferred && ((jiffies<<1)>>1) - (tp->tso_deferred>>1) > 1)
		goto send_now;
1176

1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
	in_flight = tcp_packets_in_flight(tp);

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

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

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

	limit = min(send_win, cong_win);

1189 1190
	/* If a full-sized TSO skb can be sent, do it. */
	if (limit >= 65536)
J
John Heffner 已提交
1191
		goto send_now;
1192

1193 1194 1195 1196 1197 1198 1199 1200
	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
1201
			goto send_now;
1202 1203 1204 1205 1206 1207 1208
	} 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
1209
			goto send_now;
1210 1211 1212
	}

	/* Ok, it looks like it is advisable to defer.  */
J
John Heffner 已提交
1213 1214
	tp->tso_deferred = 1 | (jiffies<<1);

1215
	return 1;
J
John Heffner 已提交
1216 1217 1218 1219

send_now:
	tp->tso_deferred = 0;
	return 0;
1220 1221
}

J
John Heffner 已提交
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257
/* Create a new MTU probe if we are ready.
 * Returns 0 if we should wait to probe (no cwnd available),
 *         1 if a probe was sent,
 *         -1 otherwise */
static int tcp_mtu_probe(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct inet_connection_sock *icsk = inet_csk(sk);
	struct sk_buff *skb, *nskb, *next;
	int len;
	int probe_size;
	unsigned int pif;
	int copy;
	int mss_now;

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

	/* Very simple search strategy: just double the MSS. */
	mss_now = tcp_current_mss(sk, 0);
	probe_size = 2*tp->mss_cache;
	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? */
	len = 0;
1258
	if ((skb = tcp_send_head(sk)) == NULL)
J
John Heffner 已提交
1259 1260
		return -1;
	while ((len += skb->len) < probe_size && !tcp_skb_is_last(sk, skb))
1261
		skb = tcp_write_queue_next(sk, skb);
J
John Heffner 已提交
1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287
	if (len < probe_size)
		return -1;

	/* Receive window check. */
	if (after(TCP_SKB_CB(skb)->seq + probe_size, tp->snd_una + tp->snd_wnd)) {
		if (tp->snd_wnd < probe_size)
			return -1;
		else
			return 0;
	}

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

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

1288 1289 1290
	skb = tcp_send_head(sk);
	tcp_insert_write_queue_before(nskb, skb, sk);
	tcp_advance_send_head(sk, skb);
J
John Heffner 已提交
1291 1292 1293 1294 1295 1296

	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;
1297
	nskb->ip_summed = skb->ip_summed;
J
John Heffner 已提交
1298 1299 1300

	len = 0;
	while (len < probe_size) {
1301
		next = tcp_write_queue_next(sk, skb);
J
John Heffner 已提交
1302 1303 1304 1305 1306 1307

		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,
1308
					 skb_put(nskb, copy), copy, nskb->csum);
J
John Heffner 已提交
1309 1310 1311 1312 1313

		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;
1314
			tcp_unlink_write_queue(skb, sk);
J
John Heffner 已提交
1315 1316 1317
			sk_stream_free_skb(sk, skb);
		} else {
			TCP_SKB_CB(nskb)->flags |= TCP_SKB_CB(skb)->flags &
1318
						   ~(TCPCB_FLAG_FIN|TCPCB_FLAG_PSH);
J
John Heffner 已提交
1319 1320
			if (!skb_shinfo(skb)->nr_frags) {
				skb_pull(skb, copy);
1321
				if (skb->ip_summed != CHECKSUM_PARTIAL)
J
John Heffner 已提交
1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341
					skb->csum = csum_partial(skb->data, skb->len, 0);
			} else {
				__pskb_trim_head(skb, copy);
				tcp_set_skb_tso_segs(sk, skb, mss_now);
			}
			TCP_SKB_CB(skb)->seq += copy;
		}

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

	/* We're ready to send.  If this fails, the probe will
	 * be resegmented into mss-sized pieces by tcp_write_xmit(). */
	TCP_SKB_CB(nskb)->when = tcp_time_stamp;
	if (!tcp_transmit_skb(sk, nskb, 1, GFP_ATOMIC)) {
		/* Decrement cwnd here because we are sending
		* effectively two packets. */
		tp->snd_cwnd--;
1342
		update_send_head(sk, nskb);
J
John Heffner 已提交
1343 1344

		icsk->icsk_mtup.probe_size = tcp_mss_to_mtu(sk, nskb->len);
1345 1346
		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
1347 1348 1349 1350 1351 1352 1353 1354

		return 1;
	}

	return -1;
}


L
Linus Torvalds 已提交
1355 1356 1357 1358 1359 1360 1361
/* This routine writes packets to the network.  It advances the
 * send_head.  This happens as incoming acks open up the remote
 * window for us.
 *
 * Returns 1, if no segments are in flight and we have queued segments, but
 * cannot send anything now because of SWS or another problem.
 */
1362
static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle)
L
Linus Torvalds 已提交
1363 1364
{
	struct tcp_sock *tp = tcp_sk(sk);
1365
	struct sk_buff *skb;
1366 1367
	unsigned int tso_segs, sent_pkts;
	int cwnd_quota;
J
John Heffner 已提交
1368
	int result;
L
Linus Torvalds 已提交
1369 1370 1371 1372 1373

	/* 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.
	 */
1374 1375
	if (unlikely(sk->sk_state == TCP_CLOSE))
		return 0;
L
Linus Torvalds 已提交
1376

1377
	sent_pkts = 0;
J
John Heffner 已提交
1378 1379 1380 1381 1382 1383 1384 1385

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

1386
	while ((skb = tcp_send_head(sk))) {
1387 1388
		unsigned int limit;

1389
		tso_segs = tcp_init_tso_segs(sk, skb, mss_now);
1390
		BUG_ON(!tso_segs);
1391

1392 1393 1394 1395 1396 1397 1398
		cwnd_quota = tcp_cwnd_test(tp, skb);
		if (!cwnd_quota)
			break;

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

1399 1400 1401 1402 1403 1404
		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 {
1405
			if (tcp_tso_should_defer(sk, skb))
1406 1407
				break;
		}
1408

1409
		limit = mss_now;
1410
		if (tso_segs > 1) {
1411 1412
			limit = tcp_window_allows(tp, skb,
						  mss_now, cwnd_quota);
1413 1414 1415

			if (skb->len < limit) {
				unsigned int trim = skb->len % mss_now;
1416

1417 1418 1419
				if (trim)
					limit = skb->len - trim;
			}
1420
		}
L
Linus Torvalds 已提交
1421

1422 1423 1424 1425
		if (skb->len > limit &&
		    unlikely(tso_fragment(sk, skb, limit, mss_now)))
			break;

1426
		TCP_SKB_CB(skb)->when = tcp_time_stamp;
1427

1428
		if (unlikely(tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC)))
1429
			break;
L
Linus Torvalds 已提交
1430

1431 1432 1433
		/* Advance the send_head.  This one is sent out.
		 * This call will increment packets_out.
		 */
1434
		update_send_head(sk, skb);
L
Linus Torvalds 已提交
1435

1436
		tcp_minshall_update(tp, mss_now, skb);
1437
		sent_pkts++;
1438
	}
L
Linus Torvalds 已提交
1439

1440
	if (likely(sent_pkts)) {
1441
		tcp_cwnd_validate(sk);
1442
		return 0;
L
Linus Torvalds 已提交
1443
	}
1444
	return !tp->packets_out && tcp_send_head(sk);
L
Linus Torvalds 已提交
1445 1446
}

1447 1448 1449 1450
/* 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.
 */
1451 1452
void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
			       int nonagle)
1453
{
1454
	struct sk_buff *skb = tcp_send_head(sk);
1455 1456

	if (skb) {
1457
		if (tcp_write_xmit(sk, cur_mss, nonagle))
1458
			tcp_check_probe_timer(sk);
1459 1460 1461
	}
}

1462 1463 1464 1465 1466 1467
/* Send _single_ skb sitting at the send head. This function requires
 * true push pending frames to setup probe timer etc.
 */
void tcp_push_one(struct sock *sk, unsigned int mss_now)
{
	struct tcp_sock *tp = tcp_sk(sk);
1468
	struct sk_buff *skb = tcp_send_head(sk);
1469 1470 1471 1472
	unsigned int tso_segs, cwnd_quota;

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

1473
	tso_segs = tcp_init_tso_segs(sk, skb, mss_now);
1474 1475 1476
	cwnd_quota = tcp_snd_test(sk, skb, mss_now, TCP_NAGLE_PUSH);

	if (likely(cwnd_quota)) {
1477 1478
		unsigned int limit;

1479 1480
		BUG_ON(!tso_segs);

1481
		limit = mss_now;
1482
		if (tso_segs > 1) {
1483 1484
			limit = tcp_window_allows(tp, skb,
						  mss_now, cwnd_quota);
1485 1486 1487 1488 1489 1490 1491 1492 1493

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

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

1494 1495 1496 1497
		if (skb->len > limit &&
		    unlikely(tso_fragment(sk, skb, limit, mss_now)))
			return;

1498 1499 1500
		/* Send it out now. */
		TCP_SKB_CB(skb)->when = tcp_time_stamp;

1501
		if (likely(!tcp_transmit_skb(sk, skb, 1, sk->sk_allocation))) {
1502 1503
			update_send_head(sk, skb);
			tcp_cwnd_validate(sk);
1504 1505 1506 1507 1508
			return;
		}
	}
}

L
Linus Torvalds 已提交
1509 1510
/* This function returns the amount that we can raise the
 * usable window based on the following constraints
1511
 *
L
Linus Torvalds 已提交
1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529
 * 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.
1530
 *
L
Linus Torvalds 已提交
1531 1532
 * Of course, if the sender implements sender side SWS prevention
 * then this will not be a problem.
1533
 *
L
Linus Torvalds 已提交
1534
 * BSD seems to make the following compromise:
1535
 *
L
Linus Torvalds 已提交
1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562
 *	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)
{
1563
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1564
	struct tcp_sock *tp = tcp_sk(sk);
S
Stephen Hemminger 已提交
1565
	/* MSS for the peer's data.  Previous versions used mss_clamp
L
Linus Torvalds 已提交
1566 1567 1568 1569 1570
	 * 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
	 */
1571
	int mss = icsk->icsk_ack.rcv_mss;
L
Linus Torvalds 已提交
1572 1573 1574 1575 1576
	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)
1577
		mss = full_space;
L
Linus Torvalds 已提交
1578 1579

	if (free_space < full_space/2) {
1580
		icsk->icsk_ack.quick = 0;
L
Linus Torvalds 已提交
1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616

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

		if (free_space < mss)
			return 0;
	}

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

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

		/* Advertise enough space so that it won't get scaled away.
		 * Import case: prevent zero window announcement if
		 * 1<<rcv_wscale > mss.
		 */
		if (((window >> tp->rx_opt.rcv_wscale) << tp->rx_opt.rcv_wscale) != window)
			window = (((window >> tp->rx_opt.rcv_wscale) + 1)
				  << tp->rx_opt.rcv_wscale);
	} else {
		/* Get the largest window that is a nice multiple of mss.
		 * Window clamp already applied above.
		 * If our current window offering is within 1 mss of the
		 * free space we just keep it. This prevents the divide
		 * and multiply from happening most of the time.
		 * We also don't do any window rounding when the free space
		 * is too small.
		 */
		if (window <= free_space - mss || window > free_space)
			window = (free_space/mss)*mss;
1617
		else if (mss == full_space &&
1618
			 free_space > window + full_space/2)
1619
			window = free_space;
L
Linus Torvalds 已提交
1620 1621 1622 1623 1624 1625 1626 1627 1628
	}

	return window;
}

/* Attempt to collapse two adjacent SKB's during retransmission. */
static void tcp_retrans_try_collapse(struct sock *sk, struct sk_buff *skb, int mss_now)
{
	struct tcp_sock *tp = tcp_sk(sk);
1629
	struct sk_buff *next_skb = tcp_write_queue_next(sk, skb);
L
Linus Torvalds 已提交
1630 1631 1632 1633 1634 1635 1636 1637 1638

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

		/* Also punt if next skb has been SACK'd. */
S
Stephen Hemminger 已提交
1639
		if (TCP_SKB_CB(next_skb)->sacked & TCPCB_SACKED_ACKED)
L
Linus Torvalds 已提交
1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656
			return;

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

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

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

1657 1658 1659 1660
		/* changing transmit queue under us so clear hints */
		clear_all_retrans_hints(tp);

		/* Ok.	We will be able to collapse the packet. */
1661
		tcp_unlink_write_queue(next_skb, sk);
L
Linus Torvalds 已提交
1662

1663 1664 1665
		skb_copy_from_linear_data(next_skb,
					  skb_put(skb, next_skb_size),
					  next_skb_size);
L
Linus Torvalds 已提交
1666

1667 1668
		if (next_skb->ip_summed == CHECKSUM_PARTIAL)
			skb->ip_summed = CHECKSUM_PARTIAL;
L
Linus Torvalds 已提交
1669

1670
		if (skb->ip_summed != CHECKSUM_PARTIAL)
L
Linus Torvalds 已提交
1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705
			skb->csum = csum_block_add(skb->csum, next_skb->csum, skb_size);

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

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

		/* All done, get rid of second SKB and account for it so
		 * packet counting does not break.
		 */
		TCP_SKB_CB(skb)->sacked |= TCP_SKB_CB(next_skb)->sacked&(TCPCB_EVER_RETRANS|TCPCB_AT_TAIL);
		if (TCP_SKB_CB(next_skb)->sacked&TCPCB_SACKED_RETRANS)
			tp->retrans_out -= tcp_skb_pcount(next_skb);
		if (TCP_SKB_CB(next_skb)->sacked&TCPCB_LOST) {
			tp->lost_out -= tcp_skb_pcount(next_skb);
			tp->left_out -= tcp_skb_pcount(next_skb);
		}
		/* Reno case is special. Sigh... */
		if (!tp->rx_opt.sack_ok && tp->sacked_out) {
			tcp_dec_pcount_approx(&tp->sacked_out, next_skb);
			tp->left_out -= tcp_skb_pcount(next_skb);
		}

		/* Not quite right: it can be > snd.fack, but
		 * it is better to underestimate fackets.
		 */
		tcp_dec_pcount_approx(&tp->fackets_out, next_skb);
		tcp_packets_out_dec(tp, next_skb);
		sk_stream_free_skb(sk, next_skb);
	}
}

/* Do a simple retransmit without using the backoff mechanisms in
1706
 * tcp_timer. This is used for path mtu discovery.
L
Linus Torvalds 已提交
1707
 * The socket is already locked here.
1708
 */
L
Linus Torvalds 已提交
1709 1710
void tcp_simple_retransmit(struct sock *sk)
{
1711
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1712 1713 1714 1715 1716
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
	unsigned int mss = tcp_current_mss(sk, 0);
	int lost = 0;

1717 1718 1719
	tcp_for_write_queue(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
1720
		if (skb->len > mss &&
L
Linus Torvalds 已提交
1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733
		    !(TCP_SKB_CB(skb)->sacked&TCPCB_SACKED_ACKED)) {
			if (TCP_SKB_CB(skb)->sacked&TCPCB_SACKED_RETRANS) {
				TCP_SKB_CB(skb)->sacked &= ~TCPCB_SACKED_RETRANS;
				tp->retrans_out -= tcp_skb_pcount(skb);
			}
			if (!(TCP_SKB_CB(skb)->sacked&TCPCB_LOST)) {
				TCP_SKB_CB(skb)->sacked |= TCPCB_LOST;
				tp->lost_out += tcp_skb_pcount(skb);
				lost = 1;
			}
		}
	}

1734 1735
	clear_all_retrans_hints(tp);

L
Linus Torvalds 已提交
1736 1737 1738 1739 1740
	if (!lost)
		return;

	tcp_sync_left_out(tp);

1741
	/* Don't muck with the congestion window here.
L
Linus Torvalds 已提交
1742 1743 1744 1745
	 * Reason is that we do not increase amount of _data_
	 * in network, but units changed and effective
	 * cwnd/ssthresh really reduced now.
	 */
1746
	if (icsk->icsk_ca_state != TCP_CA_Loss) {
L
Linus Torvalds 已提交
1747
		tp->high_seq = tp->snd_nxt;
1748
		tp->snd_ssthresh = tcp_current_ssthresh(sk);
L
Linus Torvalds 已提交
1749 1750
		tp->prior_ssthresh = 0;
		tp->undo_marker = 0;
1751
		tcp_set_ca_state(sk, TCP_CA_Loss);
L
Linus Torvalds 已提交
1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762
	}
	tcp_xmit_retransmit_queue(sk);
}

/* This retransmits one SKB.  Policy decisions and retransmit queue
 * state updates are done by the caller.  Returns non-zero if an
 * error occurred which prevented the send.
 */
int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);
J
John Heffner 已提交
1763
	struct inet_connection_sock *icsk = inet_csk(sk);
1764
	unsigned int cur_mss = tcp_current_mss(sk, 0);
L
Linus Torvalds 已提交
1765 1766
	int err;

J
John Heffner 已提交
1767 1768 1769 1770 1771
	/* Inconslusive MTU probe */
	if (icsk->icsk_mtup.probe_size) {
		icsk->icsk_mtup.probe_size = 0;
	}

L
Linus Torvalds 已提交
1772
	/* Do not sent more than we queued. 1/4 is reserved for possible
S
Stephen Hemminger 已提交
1773
	 * copying overhead: fragmentation, tunneling, mangling etc.
L
Linus Torvalds 已提交
1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795
	 */
	if (atomic_read(&sk->sk_wmem_alloc) >
	    min(sk->sk_wmem_queued + (sk->sk_wmem_queued >> 2), sk->sk_sndbuf))
		return -EAGAIN;

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

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

	if (skb->len > cur_mss) {
1796
		if (tcp_fragment(sk, skb, cur_mss, cur_mss))
L
Linus Torvalds 已提交
1797 1798 1799 1800
			return -ENOMEM; /* We'll try again later. */
	}

	/* Collapse two adjacent packets if worthwhile and we can. */
S
Stephen Hemminger 已提交
1801 1802 1803 1804 1805 1806 1807
	if (!(TCP_SKB_CB(skb)->flags & TCPCB_FLAG_SYN) &&
	    (skb->len < (cur_mss >> 1)) &&
	    (tcp_write_queue_next(sk, skb) != tcp_send_head(sk)) &&
	    (!tcp_skb_is_last(sk, skb)) &&
	    (skb_shinfo(skb)->nr_frags == 0 && skb_shinfo(tcp_write_queue_next(sk, skb))->nr_frags == 0) &&
	    (tcp_skb_pcount(skb) == 1 && tcp_skb_pcount(tcp_write_queue_next(sk, skb)) == 1) &&
	    (sysctl_tcp_retrans_collapse != 0))
L
Linus Torvalds 已提交
1808 1809
		tcp_retrans_try_collapse(sk, skb, cur_mss);

1810
	if (inet_csk(sk)->icsk_af_ops->rebuild_header(sk))
L
Linus Torvalds 已提交
1811 1812 1813 1814 1815 1816
		return -EHOSTUNREACH; /* Routing failure or similar. */

	/* Some Solaris stacks overoptimize and ignore the FIN on a
	 * retransmit when old data is attached.  So strip it off
	 * since it is cheap to do so and saves bytes on the network.
	 */
S
Stephen Hemminger 已提交
1817 1818 1819
	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
1820 1821
		if (!pskb_trim(skb, 0)) {
			TCP_SKB_CB(skb)->seq = TCP_SKB_CB(skb)->end_seq - 1;
1822 1823 1824
			skb_shinfo(skb)->gso_segs = 1;
			skb_shinfo(skb)->gso_size = 0;
			skb_shinfo(skb)->gso_type = 0;
L
Linus Torvalds 已提交
1825 1826 1827 1828 1829 1830 1831 1832 1833 1834
			skb->ip_summed = CHECKSUM_NONE;
			skb->csum = 0;
		}
	}

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

1835
	err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875

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

		tp->total_retrans++;

#if FASTRETRANS_DEBUG > 0
		if (TCP_SKB_CB(skb)->sacked&TCPCB_SACKED_RETRANS) {
			if (net_ratelimit())
				printk(KERN_DEBUG "retrans_out leaked.\n");
		}
#endif
		TCP_SKB_CB(skb)->sacked |= TCPCB_RETRANS;
		tp->retrans_out += tcp_skb_pcount(skb);

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

		tp->undo_retrans++;

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

/* This gets called after a retransmit timeout, and the initially
 * retransmitted data is acknowledged.  It tries to continue
 * resending the rest of the retransmit queue, until either
 * we've sent it all or the congestion window limit is reached.
 * If doing SACK, the first ACK which comes back for a timeout
 * based retransmit packet might feed us FACK information again.
 * If so, we use it to avoid unnecessarily retransmissions.
 */
void tcp_xmit_retransmit_queue(struct sock *sk)
{
1876
	const struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1877 1878
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *skb;
1879 1880 1881 1882 1883 1884
	int packet_cnt;

	if (tp->retransmit_skb_hint) {
		skb = tp->retransmit_skb_hint;
		packet_cnt = tp->retransmit_cnt_hint;
	}else{
1885
		skb = tcp_write_queue_head(sk);
1886 1887
		packet_cnt = 0;
	}
L
Linus Torvalds 已提交
1888 1889

	/* First pass: retransmit lost packets. */
1890
	if (tp->lost_out) {
1891
		tcp_for_write_queue_from(skb, sk) {
L
Linus Torvalds 已提交
1892 1893
			__u8 sacked = TCP_SKB_CB(skb)->sacked;

1894 1895
			if (skb == tcp_send_head(sk))
				break;
1896 1897 1898 1899
			/* we could do better than to assign each time */
			tp->retransmit_skb_hint = skb;
			tp->retransmit_cnt_hint = packet_cnt;

L
Linus Torvalds 已提交
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909
			/* 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;

1910
			if (sacked & TCPCB_LOST) {
L
Linus Torvalds 已提交
1911
				if (!(sacked&(TCPCB_SACKED_ACKED|TCPCB_SACKED_RETRANS))) {
1912 1913
					if (tcp_retransmit_skb(sk, skb)) {
						tp->retransmit_skb_hint = NULL;
L
Linus Torvalds 已提交
1914
						return;
1915
					}
1916
					if (icsk->icsk_ca_state != TCP_CA_Loss)
L
Linus Torvalds 已提交
1917 1918 1919 1920
						NET_INC_STATS_BH(LINUX_MIB_TCPFASTRETRANS);
					else
						NET_INC_STATS_BH(LINUX_MIB_TCPSLOWSTARTRETRANS);

1921
					if (skb == tcp_write_queue_head(sk))
1922
						inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
1923 1924
									  inet_csk(sk)->icsk_rto,
									  TCP_RTO_MAX);
L
Linus Torvalds 已提交
1925 1926
				}

1927 1928
				packet_cnt += tcp_skb_pcount(skb);
				if (packet_cnt >= tp->lost_out)
L
Linus Torvalds 已提交
1929 1930 1931 1932 1933 1934 1935 1936
					break;
			}
		}
	}

	/* OK, demanded retransmission is finished. */

	/* Forward retransmissions are possible only during Recovery. */
1937
	if (icsk->icsk_ca_state != TCP_CA_Recovery)
L
Linus Torvalds 已提交
1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950
		return;

	/* No forward retransmissions in Reno are possible. */
	if (!tp->rx_opt.sack_ok)
		return;

	/* Yeah, we have to make difficult choice between forward transmission
	 * and retransmission... Both ways have their merits...
	 *
	 * For now we do not retransmit anything, while we have some new
	 * segments to send.
	 */

1951
	if (tcp_may_send_now(sk))
L
Linus Torvalds 已提交
1952 1953
		return;

1954 1955 1956 1957
	if (tp->forward_skb_hint) {
		skb = tp->forward_skb_hint;
		packet_cnt = tp->forward_cnt_hint;
	} else{
1958
		skb = tcp_write_queue_head(sk);
1959 1960 1961
		packet_cnt = 0;
	}

1962 1963 1964
	tcp_for_write_queue_from(skb, sk) {
		if (skb == tcp_send_head(sk))
			break;
1965 1966
		tp->forward_cnt_hint = packet_cnt;
		tp->forward_skb_hint = skb;
L
Linus Torvalds 已提交
1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983

		/* Similar to the retransmit loop above we
		 * can pretend that the retransmitted SKB
		 * we send out here will be composed of one
		 * real MSS sized packet because tcp_retransmit_skb()
		 * will fragment it if necessary.
		 */
		if (++packet_cnt > tp->fackets_out)
			break;

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

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

		/* Ok, retransmit it. */
1984 1985
		if (tcp_retransmit_skb(sk, skb)) {
			tp->forward_skb_hint = NULL;
L
Linus Torvalds 已提交
1986
			break;
1987
		}
L
Linus Torvalds 已提交
1988

1989
		if (skb == tcp_write_queue_head(sk))
1990 1991 1992
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
						  inet_csk(sk)->icsk_rto,
						  TCP_RTO_MAX);
L
Linus Torvalds 已提交
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

		NET_INC_STATS_BH(LINUX_MIB_TCPFORWARDRETRANS);
	}
}


/* Send a fin.  The caller locks the socket for us.  This cannot be
 * allowed to fail queueing a FIN frame under any circumstances.
 */
void tcp_send_fin(struct sock *sk)
{
2004
	struct tcp_sock *tp = tcp_sk(sk);
2005
	struct sk_buff *skb = tcp_write_queue_tail(sk);
L
Linus Torvalds 已提交
2006
	int mss_now;
2007

L
Linus Torvalds 已提交
2008 2009 2010 2011 2012 2013
	/* Optimization, tack on the FIN if we have a queue of
	 * unsent frames.  But be careful about outgoing SACKS
	 * and IP options.
	 */
	mss_now = tcp_current_mss(sk, 1);

2014
	if (tcp_send_head(sk) != NULL) {
L
Linus Torvalds 已提交
2015 2016 2017 2018 2019 2020
		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 (;;) {
2021
			skb = alloc_skb_fclone(MAX_TCP_HEADER, GFP_KERNEL);
L
Linus Torvalds 已提交
2022 2023 2024 2025 2026 2027 2028 2029 2030 2031
			if (skb)
				break;
			yield();
		}

		/* Reserve space for headers and prepare control bits. */
		skb_reserve(skb, MAX_TCP_HEADER);
		skb->csum = 0;
		TCP_SKB_CB(skb)->flags = (TCPCB_FLAG_ACK | TCPCB_FLAG_FIN);
		TCP_SKB_CB(skb)->sacked = 0;
2032 2033 2034
		skb_shinfo(skb)->gso_segs = 1;
		skb_shinfo(skb)->gso_size = 0;
		skb_shinfo(skb)->gso_type = 0;
L
Linus Torvalds 已提交
2035 2036 2037 2038 2039 2040

		/* FIN eats a sequence byte, write_seq advanced by tcp_queue_skb(). */
		TCP_SKB_CB(skb)->seq = tp->write_seq;
		TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq + 1;
		tcp_queue_skb(sk, skb);
	}
2041
	__tcp_push_pending_frames(sk, mss_now, TCP_NAGLE_OFF);
L
Linus Torvalds 已提交
2042 2043 2044 2045 2046
}

/* 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
2047
 * by RFC 2525, section 2.17.  -DaveM
L
Linus Torvalds 已提交
2048
 */
A
Al Viro 已提交
2049
void tcp_send_active_reset(struct sock *sk, gfp_t priority)
L
Linus Torvalds 已提交
2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064
{
	struct sk_buff *skb;

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

	/* Reserve space for headers and prepare control bits. */
	skb_reserve(skb, MAX_TCP_HEADER);
	skb->csum = 0;
	TCP_SKB_CB(skb)->flags = (TCPCB_FLAG_ACK | TCPCB_FLAG_RST);
	TCP_SKB_CB(skb)->sacked = 0;
2065 2066 2067
	skb_shinfo(skb)->gso_segs = 1;
	skb_shinfo(skb)->gso_size = 0;
	skb_shinfo(skb)->gso_type = 0;
L
Linus Torvalds 已提交
2068 2069

	/* Send it off. */
2070
	TCP_SKB_CB(skb)->seq = tcp_acceptable_seq(sk);
L
Linus Torvalds 已提交
2071 2072
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2073
	if (tcp_transmit_skb(sk, skb, 0, priority))
L
Linus Torvalds 已提交
2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085
		NET_INC_STATS(LINUX_MIB_TCPABORTFAILED);
}

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

2086
	skb = tcp_write_queue_head(sk);
L
Linus Torvalds 已提交
2087 2088 2089 2090 2091 2092 2093 2094 2095
	if (skb == NULL || !(TCP_SKB_CB(skb)->flags&TCPCB_FLAG_SYN)) {
		printk(KERN_DEBUG "tcp_send_synack: wrong queue state\n");
		return -EFAULT;
	}
	if (!(TCP_SKB_CB(skb)->flags&TCPCB_FLAG_ACK)) {
		if (skb_cloned(skb)) {
			struct sk_buff *nskb = skb_copy(skb, GFP_ATOMIC);
			if (nskb == NULL)
				return -ENOMEM;
2096
			tcp_unlink_write_queue(skb, sk);
L
Linus Torvalds 已提交
2097
			skb_header_release(nskb);
2098
			__tcp_add_write_queue_head(sk, nskb);
L
Linus Torvalds 已提交
2099 2100 2101 2102 2103 2104 2105 2106 2107
			sk_stream_free_skb(sk, skb);
			sk_charge_skb(sk, nskb);
			skb = nskb;
		}

		TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_ACK;
		TCP_ECN_send_synack(tcp_sk(sk), skb);
	}
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2108
	return tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
2109 2110 2111 2112 2113 2114
}

/*
 * Prepare a SYN-ACK.
 */
struct sk_buff * tcp_make_synack(struct sock *sk, struct dst_entry *dst,
2115
				 struct request_sock *req)
L
Linus Torvalds 已提交
2116
{
2117
	struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
2118 2119 2120 2121
	struct tcp_sock *tp = tcp_sk(sk);
	struct tcphdr *th;
	int tcp_header_size;
	struct sk_buff *skb;
2122 2123 2124 2125
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *md5;
	__u8 *md5_hash_location;
#endif
L
Linus Torvalds 已提交
2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136

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

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

	skb->dst = dst_clone(dst);

	tcp_header_size = (sizeof(struct tcphdr) + TCPOLEN_MSS +
2137 2138
			   (ireq->tstamp_ok ? TCPOLEN_TSTAMP_ALIGNED : 0) +
			   (ireq->wscale_ok ? TCPOLEN_WSCALE_ALIGNED : 0) +
L
Linus Torvalds 已提交
2139
			   /* SACK_PERM is in the place of NOP NOP of TS */
2140
			   ((ireq->sack_ok && !ireq->tstamp_ok) ? TCPOLEN_SACKPERM_ALIGNED : 0));
2141 2142 2143 2144 2145 2146 2147

#ifdef CONFIG_TCP_MD5SIG
	/* Are we doing MD5 on this segment? If so - make room for it */
	md5 = tcp_rsk(req)->af_specific->md5_lookup(sk, req);
	if (md5)
		tcp_header_size += TCPOLEN_MD5SIG_ALIGNED;
#endif
2148 2149
	skb_push(skb, tcp_header_size);
	skb_reset_transport_header(skb);
L
Linus Torvalds 已提交
2150

2151
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
2152 2153 2154 2155 2156
	memset(th, 0, sizeof(struct tcphdr));
	th->syn = 1;
	th->ack = 1;
	TCP_ECN_make_synack(req, th);
	th->source = inet_sk(sk)->sport;
2157 2158
	th->dest = ireq->rmt_port;
	TCP_SKB_CB(skb)->seq = tcp_rsk(req)->snt_isn;
L
Linus Torvalds 已提交
2159 2160
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq + 1;
	TCP_SKB_CB(skb)->sacked = 0;
2161 2162 2163
	skb_shinfo(skb)->gso_segs = 1;
	skb_shinfo(skb)->gso_size = 0;
	skb_shinfo(skb)->gso_type = 0;
L
Linus Torvalds 已提交
2164
	th->seq = htonl(TCP_SKB_CB(skb)->seq);
2165
	th->ack_seq = htonl(tcp_rsk(req)->rcv_isn + 1);
L
Linus Torvalds 已提交
2166
	if (req->rcv_wnd == 0) { /* ignored for retransmitted syns */
2167
		__u8 rcv_wscale;
L
Linus Torvalds 已提交
2168 2169 2170
		/* 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 */
2171
		tcp_select_initial_window(tcp_full_space(sk),
2172
			dst_metric(dst, RTAX_ADVMSS) - (ireq->tstamp_ok ? TCPOLEN_TSTAMP_ALIGNED : 0),
L
Linus Torvalds 已提交
2173 2174
			&req->rcv_wnd,
			&req->window_clamp,
2175
			ireq->wscale_ok,
L
Linus Torvalds 已提交
2176
			&rcv_wscale);
2177
		ireq->rcv_wscale = rcv_wscale;
L
Linus Torvalds 已提交
2178 2179 2180
	}

	/* RFC1323: The window in SYN & SYN/ACK segments is never scaled. */
2181
	th->window = htons(min(req->rcv_wnd, 65535U));
L
Linus Torvalds 已提交
2182 2183

	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2184
	tcp_syn_build_options((__be32 *)(th + 1), dst_metric(dst, RTAX_ADVMSS), ireq->tstamp_ok,
2185
			      ireq->sack_ok, ireq->wscale_ok, ireq->rcv_wscale,
L
Linus Torvalds 已提交
2186
			      TCP_SKB_CB(skb)->when,
2187 2188 2189 2190 2191 2192 2193
			      req->ts_recent,
			      (
#ifdef CONFIG_TCP_MD5SIG
			       md5 ? &md5_hash_location :
#endif
			       NULL)
			      );
L
Linus Torvalds 已提交
2194 2195 2196 2197

	skb->csum = 0;
	th->doff = (tcp_header_size >> 2);
	TCP_INC_STATS(TCP_MIB_OUTSEGS);
2198 2199 2200 2201 2202 2203 2204

#ifdef CONFIG_TCP_MD5SIG
	/* Okay, we have all we need - do the md5 hash if needed */
	if (md5) {
		tp->af_specific->calc_md5_hash(md5_hash_location,
					       md5,
					       NULL, dst, req,
2205
					       tcp_hdr(skb), sk->sk_protocol,
2206 2207 2208 2209
					       skb->len);
	}
#endif

L
Linus Torvalds 已提交
2210 2211 2212
	return skb;
}

2213
/*
L
Linus Torvalds 已提交
2214
 * Do all connect socket setups that can be done AF independent.
2215
 */
S
Stephen Hemminger 已提交
2216
static void tcp_connect_init(struct sock *sk)
L
Linus Torvalds 已提交
2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227
{
	struct dst_entry *dst = __sk_dst_get(sk);
	struct tcp_sock *tp = tcp_sk(sk);
	__u8 rcv_wscale;

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

2228 2229 2230 2231 2232
#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
2233 2234 2235 2236
	/* 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
2237
	tcp_mtup_init(sk);
L
Linus Torvalds 已提交
2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264
	tcp_sync_mss(sk, dst_mtu(dst));

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

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

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

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

2265 2266
	inet_csk(sk)->icsk_rto = TCP_TIMEOUT_INIT;
	inet_csk(sk)->icsk_retransmits = 0;
L
Linus Torvalds 已提交
2267 2268 2269 2270 2271
	tcp_clear_retrans(tp);
}

/*
 * Build a SYN and send it off.
2272
 */
L
Linus Torvalds 已提交
2273 2274 2275 2276 2277 2278 2279
int tcp_connect(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	struct sk_buff *buff;

	tcp_connect_init(sk);

2280
	buff = alloc_skb_fclone(MAX_TCP_HEADER + 15, sk->sk_allocation);
L
Linus Torvalds 已提交
2281 2282 2283 2284 2285 2286 2287
	if (unlikely(buff == NULL))
		return -ENOBUFS;

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

	TCP_SKB_CB(buff)->flags = TCPCB_FLAG_SYN;
2288
	TCP_ECN_send_syn(sk, buff);
L
Linus Torvalds 已提交
2289
	TCP_SKB_CB(buff)->sacked = 0;
2290 2291 2292
	skb_shinfo(buff)->gso_segs = 1;
	skb_shinfo(buff)->gso_size = 0;
	skb_shinfo(buff)->gso_type = 0;
L
Linus Torvalds 已提交
2293
	buff->csum = 0;
2294
	tp->snd_nxt = tp->write_seq;
L
Linus Torvalds 已提交
2295 2296 2297 2298 2299 2300 2301
	TCP_SKB_CB(buff)->seq = tp->write_seq++;
	TCP_SKB_CB(buff)->end_seq = tp->write_seq;

	/* Send it off. */
	TCP_SKB_CB(buff)->when = tcp_time_stamp;
	tp->retrans_stamp = TCP_SKB_CB(buff)->when;
	skb_header_release(buff);
2302
	__tcp_add_write_queue_tail(sk, buff);
L
Linus Torvalds 已提交
2303 2304
	sk_charge_skb(sk, buff);
	tp->packets_out += tcp_skb_pcount(buff);
2305
	tcp_transmit_skb(sk, buff, 1, GFP_KERNEL);
2306 2307 2308 2309 2310 2311

	/* We change tp->snd_nxt after the tcp_transmit_skb() call
	 * in order to make this packet get counted in tcpOutSegs.
	 */
	tp->snd_nxt = tp->write_seq;
	tp->pushed_seq = tp->write_seq;
L
Linus Torvalds 已提交
2312 2313 2314
	TCP_INC_STATS(TCP_MIB_ACTIVEOPENS);

	/* Timer for repeating the SYN until an answer. */
2315 2316
	inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
				  inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
L
Linus Torvalds 已提交
2317 2318 2319 2320 2321 2322 2323 2324 2325
	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)
{
2326 2327
	struct inet_connection_sock *icsk = inet_csk(sk);
	int ato = icsk->icsk_ack.ato;
L
Linus Torvalds 已提交
2328 2329 2330
	unsigned long timeout;

	if (ato > TCP_DELACK_MIN) {
2331
		const struct tcp_sock *tp = tcp_sk(sk);
L
Linus Torvalds 已提交
2332 2333
		int max_ato = HZ/2;

2334
		if (icsk->icsk_ack.pingpong || (icsk->icsk_ack.pending & ICSK_ACK_PUSHED))
L
Linus Torvalds 已提交
2335 2336 2337 2338 2339
			max_ato = TCP_DELACK_MAX;

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

		/* If some rtt estimate is known, use it to bound delayed ack.
2340
		 * Do not use inet_csk(sk)->icsk_rto here, use results of rtt measurements
L
Linus Torvalds 已提交
2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356
		 * directly.
		 */
		if (tp->srtt) {
			int rtt = max(tp->srtt>>3, TCP_DELACK_MIN);

			if (rtt < max_ato)
				max_ato = rtt;
		}

		ato = min(ato, max_ato);
	}

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

	/* Use new timeout only if there wasn't a older one earlier. */
2357
	if (icsk->icsk_ack.pending & ICSK_ACK_TIMER) {
L
Linus Torvalds 已提交
2358 2359 2360
		/* If delack timer was blocked or is about to expire,
		 * send ACK now.
		 */
2361 2362
		if (icsk->icsk_ack.blocked ||
		    time_before_eq(icsk->icsk_ack.timeout, jiffies + (ato >> 2))) {
L
Linus Torvalds 已提交
2363 2364 2365 2366
			tcp_send_ack(sk);
			return;
		}

2367 2368
		if (!time_before(timeout, icsk->icsk_ack.timeout))
			timeout = icsk->icsk_ack.timeout;
L
Linus Torvalds 已提交
2369
	}
2370 2371 2372
	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
2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387
}

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

		/* We are not putting this on the write queue, so
		 * tcp_transmit_skb() will set the ownership to this
		 * sock.
		 */
		buff = alloc_skb(MAX_TCP_HEADER, GFP_ATOMIC);
		if (buff == NULL) {
2388 2389
			inet_csk_schedule_ack(sk);
			inet_csk(sk)->icsk_ack.ato = TCP_ATO_MIN;
2390 2391
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
						  TCP_DELACK_MAX, TCP_RTO_MAX);
L
Linus Torvalds 已提交
2392 2393 2394 2395 2396 2397 2398 2399
			return;
		}

		/* Reserve space for headers and prepare control bits. */
		skb_reserve(buff, MAX_TCP_HEADER);
		buff->csum = 0;
		TCP_SKB_CB(buff)->flags = TCPCB_FLAG_ACK;
		TCP_SKB_CB(buff)->sacked = 0;
2400 2401 2402
		skb_shinfo(buff)->gso_segs = 1;
		skb_shinfo(buff)->gso_size = 0;
		skb_shinfo(buff)->gso_type = 0;
L
Linus Torvalds 已提交
2403 2404

		/* Send it off, this clears delayed acks for us. */
2405
		TCP_SKB_CB(buff)->seq = TCP_SKB_CB(buff)->end_seq = tcp_acceptable_seq(sk);
L
Linus Torvalds 已提交
2406
		TCP_SKB_CB(buff)->when = tcp_time_stamp;
2407
		tcp_transmit_skb(sk, buff, 0, GFP_ATOMIC);
L
Linus Torvalds 已提交
2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428
	}
}

/* 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);
2429
	if (skb == NULL)
L
Linus Torvalds 已提交
2430 2431 2432 2433 2434 2435 2436
		return -1;

	/* Reserve space for headers and set control bits. */
	skb_reserve(skb, MAX_TCP_HEADER);
	skb->csum = 0;
	TCP_SKB_CB(skb)->flags = TCPCB_FLAG_ACK;
	TCP_SKB_CB(skb)->sacked = urgent;
2437 2438 2439
	skb_shinfo(skb)->gso_segs = 1;
	skb_shinfo(skb)->gso_size = 0;
	skb_shinfo(skb)->gso_type = 0;
L
Linus Torvalds 已提交
2440 2441 2442 2443 2444 2445 2446 2447

	/* Use a previous sequence.  This should cause the other
	 * end to send an ack.  Don't queue or clone SKB, just
	 * send it.
	 */
	TCP_SKB_CB(skb)->seq = urgent ? tp->snd_una : tp->snd_una - 1;
	TCP_SKB_CB(skb)->end_seq = TCP_SKB_CB(skb)->seq;
	TCP_SKB_CB(skb)->when = tcp_time_stamp;
2448
	return tcp_transmit_skb(sk, skb, 0, GFP_ATOMIC);
L
Linus Torvalds 已提交
2449 2450 2451 2452 2453 2454 2455 2456
}

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

2457
		if ((skb = tcp_send_head(sk)) != NULL &&
L
Linus Torvalds 已提交
2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473
		    before(TCP_SKB_CB(skb)->seq, tp->snd_una+tp->snd_wnd)) {
			int err;
			unsigned int mss = tcp_current_mss(sk, 0);
			unsigned int seg_size = tp->snd_una+tp->snd_wnd-TCP_SKB_CB(skb)->seq;

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

			/* We are probing the opening of a window
			 * but the window size is != 0
			 * must have been a result SWS avoidance ( sender )
			 */
			if (seg_size < TCP_SKB_CB(skb)->end_seq - TCP_SKB_CB(skb)->seq ||
			    skb->len > mss) {
				seg_size = min(seg_size, mss);
				TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
2474
				if (tcp_fragment(sk, skb, seg_size, mss))
L
Linus Torvalds 已提交
2475 2476
					return -1;
			} else if (!tcp_skb_pcount(skb))
2477
				tcp_set_skb_tso_segs(sk, skb, mss);
L
Linus Torvalds 已提交
2478 2479 2480

			TCP_SKB_CB(skb)->flags |= TCPCB_FLAG_PSH;
			TCP_SKB_CB(skb)->when = tcp_time_stamp;
2481
			err = tcp_transmit_skb(sk, skb, 1, GFP_ATOMIC);
L
Linus Torvalds 已提交
2482
			if (!err) {
2483
				update_send_head(sk, skb);
L
Linus Torvalds 已提交
2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500
			}
			return err;
		} else {
			if (tp->urg_mode &&
			    between(tp->snd_up, tp->snd_una+1, tp->snd_una+0xFFFF))
				tcp_xmit_probe_skb(sk, TCPCB_URG);
			return tcp_xmit_probe_skb(sk, 0);
		}
	}
	return -1;
}

/* A window probe timeout has occurred.  If window is not closed send
 * a partial packet else a zero probe.
 */
void tcp_send_probe0(struct sock *sk)
{
2501
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
2502 2503 2504 2505 2506
	struct tcp_sock *tp = tcp_sk(sk);
	int err;

	err = tcp_write_wakeup(sk);

2507
	if (tp->packets_out || !tcp_send_head(sk)) {
L
Linus Torvalds 已提交
2508
		/* Cancel probe timer, if it is not required. */
2509
		icsk->icsk_probes_out = 0;
2510
		icsk->icsk_backoff = 0;
L
Linus Torvalds 已提交
2511 2512 2513 2514
		return;
	}

	if (err <= 0) {
2515 2516
		if (icsk->icsk_backoff < sysctl_tcp_retries2)
			icsk->icsk_backoff++;
2517
		icsk->icsk_probes_out++;
2518
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
2519 2520
					  min(icsk->icsk_rto << icsk->icsk_backoff, TCP_RTO_MAX),
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2521 2522
	} else {
		/* If packet was not sent due to local congestion,
2523
		 * do not backoff and do not remember icsk_probes_out.
L
Linus Torvalds 已提交
2524 2525 2526 2527
		 * Let local senders to fight for local resources.
		 *
		 * Use accumulated backoff yet.
		 */
2528 2529
		if (!icsk->icsk_probes_out)
			icsk->icsk_probes_out = 1;
2530
		inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
2531
					  min(icsk->icsk_rto << icsk->icsk_backoff,
2532 2533
					      TCP_RESOURCE_PROBE_INTERVAL),
					  TCP_RTO_MAX);
L
Linus Torvalds 已提交
2534 2535 2536 2537 2538 2539 2540
	}
}

EXPORT_SYMBOL(tcp_connect);
EXPORT_SYMBOL(tcp_make_synack);
EXPORT_SYMBOL(tcp_simple_retransmit);
EXPORT_SYMBOL(tcp_sync_mss);
2541
EXPORT_SYMBOL(sysctl_tcp_tso_win_divisor);
J
John Heffner 已提交
2542
EXPORT_SYMBOL(tcp_mtup_init);