tcp_ipv4.c 64.4 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
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
 * Version:	$Id: tcp_ipv4.c,v 1.240 2002/02/01 22:01:04 davem Exp $
Linus Torvalds's avatar
Linus Torvalds committed
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
 *
 *		IPv4 specific functions
 *
 *
 *		code split from:
 *		linux/ipv4/tcp.c
 *		linux/ipv4/tcp_input.c
 *		linux/ipv4/tcp_output.c
 *
 *		See tcp.c for author information
 *
 *	This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */

/*
 * Changes:
 *		David S. Miller	:	New socket lookup architecture.
 *					This code is dedicated to John Dyson.
 *		David S. Miller :	Change semantics of established hash,
 *					half is devoted to TIME_WAIT sockets
 *					and the rest go in the other half.
 *		Andi Kleen :		Add support for syncookies and fixed
 *					some bugs: ip options weren't passed to
35 36
 *					the TCP layer, missed a check for an
 *					ACK bit.
Linus Torvalds's avatar
Linus Torvalds committed
37 38 39 40 41 42 43 44 45
 *		Andi Kleen :		Implemented fast path mtu discovery.
 *	     				Fixed many serious bugs in the
 *					open_request handling and moved
 *					most of it into the af independent code.
 *					Added tail drop and some other bugfixes.
 *					Added new listen sematics.
 *		Mike McLagan	:	Routing by source
 *	Juan Jose Ciarlante:		ip_dynaddr bits
 *		Andi Kleen:		various fixes.
46 47
 *	Vitaly E. Lavrov	:	Transparent proxy revived after year
 *					coma.
Linus Torvalds's avatar
Linus Torvalds committed
48 49
 *	Andi Kleen		:	Fix new listen.
 *	Andi Kleen		:	Fix accept error reporting.
50 51 52
 *	YOSHIFUJI Hideaki @USAGI and:	Support IPV6_V6ONLY socket option, which
 *	Alexey Kuznetsov		allow both IPv4 and IPv6 sockets to bind
 *					a single port at the same time.
Linus Torvalds's avatar
Linus Torvalds committed
53 54 55
 */

#include <linux/config.h>
56

Linus Torvalds's avatar
Linus Torvalds committed
57 58
#include <linux/types.h>
#include <linux/fcntl.h>
59
#include <linux/module.h>
Linus Torvalds's avatar
Linus Torvalds committed
60 61
#include <linux/random.h>
#include <linux/cache.h>
62
#include <linux/jhash.h>
Linus Torvalds's avatar
Linus Torvalds committed
63
#include <linux/init.h>
64
#include <linux/times.h>
Linus Torvalds's avatar
Linus Torvalds committed
65 66 67 68 69

#include <net/icmp.h>
#include <net/tcp.h>
#include <net/ipv6.h>
#include <net/inet_common.h>
70
#include <net/xfrm.h>
Linus Torvalds's avatar
Linus Torvalds committed
71 72

#include <linux/inet.h>
73
#include <linux/ipv6.h>
Linus Torvalds's avatar
Linus Torvalds committed
74
#include <linux/stddef.h>
75 76
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
Linus Torvalds's avatar
Linus Torvalds committed
77 78

extern int sysctl_ip_dynaddr;
79
int sysctl_tcp_tw_reuse;
80
int sysctl_tcp_low_latency;
Linus Torvalds's avatar
Linus Torvalds committed
81 82 83 84

/* Check TCP sequence numbers in ICMP packets. */
#define ICMP_MIN_LENGTH 8

85
/* Socket used for sending RSTs */
Linus Torvalds's avatar
Linus Torvalds committed
86
static struct socket *tcp_socket;
Linus Torvalds's avatar
Linus Torvalds committed
87

88
void tcp_v4_send_check(struct sock *sk, struct tcphdr *th, int len,
Linus Torvalds's avatar
Linus Torvalds committed
89 90 91
		       struct sk_buff *skb);

struct tcp_hashinfo __cacheline_aligned tcp_hashinfo = {
92 93
	.__tcp_lhash_lock	=	RW_LOCK_UNLOCKED,
	.__tcp_lhash_users	=	ATOMIC_INIT(0),
94 95
	.__tcp_lhash_wait
	  = __WAIT_QUEUE_HEAD_INITIALIZER(tcp_hashinfo.__tcp_lhash_wait),
96
	.__tcp_portalloc_lock	=	SPIN_LOCK_UNLOCKED
Linus Torvalds's avatar
Linus Torvalds committed
97 98 99 100 101 102 103 104
};

/*
 * This array holds the first and last local port number.
 * For high-usage systems, use sysctl to change this to
 * 32768-61000
 */
int sysctl_local_port_range[2] = { 1024, 4999 };
105
int tcp_port_rover = 1024 - 1;
Linus Torvalds's avatar
Linus Torvalds committed
106 107 108 109

static __inline__ int tcp_hashfn(__u32 laddr, __u16 lport,
				 __u32 faddr, __u16 fport)
{
110 111 112
	int h = (laddr ^ lport) ^ (faddr ^ fport);
	h ^= h >> 16;
	h ^= h >> 8;
Linus Torvalds's avatar
Linus Torvalds committed
113 114 115 116 117
	return h & (tcp_ehash_size - 1);
}

static __inline__ int tcp_sk_hashfn(struct sock *sk)
{
118 119 120 121 122
	struct inet_opt *inet = inet_sk(sk);
	__u32 laddr = inet->rcv_saddr;
	__u16 lport = inet->num;
	__u32 faddr = inet->daddr;
	__u16 fport = inet->dport;
Linus Torvalds's avatar
Linus Torvalds committed
123 124 125 126 127 128 129 130 131 132

	return tcp_hashfn(laddr, lport, faddr, fport);
}

/* Allocate and initialize a new TCP local port bind bucket.
 * The bindhash mutex for snum's hash chain must be held here.
 */
struct tcp_bind_bucket *tcp_bucket_create(struct tcp_bind_hashbucket *head,
					  unsigned short snum)
{
133 134 135
	struct tcp_bind_bucket *tb = kmem_cache_alloc(tcp_bucket_cachep,
						      SLAB_ATOMIC);
	if (tb) {
Linus Torvalds's avatar
Linus Torvalds committed
136 137
		tb->port = snum;
		tb->fastreuse = 0;
138
		INIT_HLIST_HEAD(&tb->owners);
139
		hlist_add_head(&tb->node, &head->chain);
Linus Torvalds's avatar
Linus Torvalds committed
140 141 142 143
	}
	return tb;
}

144 145 146
/* Caller must hold hashbucket lock for this tb with local BH disabled */
void tcp_bucket_destroy(struct tcp_bind_bucket *tb)
{
147
	if (hlist_empty(&tb->owners)) {
148
		__hlist_del(&tb->node);
149 150 151 152
		kmem_cache_free(tcp_bucket_cachep, tb);
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
153 154 155
/* Caller must disable local BH processing. */
static __inline__ void __tcp_inherit_port(struct sock *sk, struct sock *child)
{
156 157
	struct tcp_bind_hashbucket *head =
				&tcp_bhash[tcp_bhashfn(inet_sk(child)->num)];
Linus Torvalds's avatar
Linus Torvalds committed
158 159 160
	struct tcp_bind_bucket *tb;

	spin_lock(&head->lock);
161
	tb = tcp_sk(sk)->bind_hash;
162
	sk_add_bind_node(child, &tb->owners);
163
	tcp_sk(child)->bind_hash = tb;
Linus Torvalds's avatar
Linus Torvalds committed
164 165 166
	spin_unlock(&head->lock);
}

167
inline void tcp_inherit_port(struct sock *sk, struct sock *child)
Linus Torvalds's avatar
Linus Torvalds committed
168 169 170 171 172 173
{
	local_bh_disable();
	__tcp_inherit_port(sk, child);
	local_bh_enable();
}

174 175
void tcp_bind_hash(struct sock *sk, struct tcp_bind_bucket *tb,
		   unsigned short snum)
176 177
{
	inet_sk(sk)->num = snum;
178
	sk_add_bind_node(sk, &tb->owners);
179
	tcp_sk(sk)->bind_hash = tb;
180
}
Linus Torvalds's avatar
Linus Torvalds committed
181 182

static inline int tcp_bind_conflict(struct sock *sk, struct tcp_bind_bucket *tb)
183
{
184
	const u32 sk_rcv_saddr = tcp_v4_rcv_saddr(sk);
185 186
	struct sock *sk2;
	struct hlist_node *node;
187
	int reuse = sk->sk_reuse;
188

189
	sk_for_each_bound(sk2, node, &tb->owners) {
190
		if (sk != sk2 &&
191
		    !tcp_v6_ipv6only(sk2) &&
192 193 194
		    (!sk->sk_bound_dev_if ||
		     !sk2->sk_bound_dev_if ||
		     sk->sk_bound_dev_if == sk2->sk_bound_dev_if)) {
195 196
			if (!reuse || !sk2->sk_reuse ||
			    sk2->sk_state == TCP_LISTEN) {
197 198 199
				const u32 sk2_rcv_saddr = tcp_v4_rcv_saddr(sk2);
				if (!sk2_rcv_saddr || !sk_rcv_saddr ||
				    sk2_rcv_saddr == sk_rcv_saddr)
Linus Torvalds's avatar
Linus Torvalds committed
200 201 202 203
					break;
			}
		}
	}
204
	return node != NULL;
205
}
Linus Torvalds's avatar
Linus Torvalds committed
206

Linus Torvalds's avatar
Linus Torvalds committed
207 208 209 210 211 212
/* Obtain a reference to a local port for the given sock,
 * if snum is zero it means select any available local port.
 */
static int tcp_v4_get_port(struct sock *sk, unsigned short snum)
{
	struct tcp_bind_hashbucket *head;
213
	struct hlist_node *node;
Linus Torvalds's avatar
Linus Torvalds committed
214 215 216 217
	struct tcp_bind_bucket *tb;
	int ret;

	local_bh_disable();
218
	if (!snum) {
Linus Torvalds's avatar
Linus Torvalds committed
219 220 221 222 223 224 225
		int low = sysctl_local_port_range[0];
		int high = sysctl_local_port_range[1];
		int remaining = (high - low) + 1;
		int rover;

		spin_lock(&tcp_portalloc_lock);
		rover = tcp_port_rover;
226 227 228
		do {
			rover++;
			if (rover < low || rover > high)
Linus Torvalds's avatar
Linus Torvalds committed
229 230 231
				rover = low;
			head = &tcp_bhash[tcp_bhashfn(rover)];
			spin_lock(&head->lock);
232
			tb_for_each(tb, node, &head->chain)
Linus Torvalds's avatar
Linus Torvalds committed
233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253
				if (tb->port == rover)
					goto next;
			break;
		next:
			spin_unlock(&head->lock);
		} while (--remaining > 0);
		tcp_port_rover = rover;
		spin_unlock(&tcp_portalloc_lock);

		/* Exhausted local port range during search? */
		ret = 1;
		if (remaining <= 0)
			goto fail;

		/* OK, here is the one we will use.  HEAD is
		 * non-NULL and we hold it's mutex.
		 */
		snum = rover;
	} else {
		head = &tcp_bhash[tcp_bhashfn(snum)];
		spin_lock(&head->lock);
254
		tb_for_each(tb, node, &head->chain)
Linus Torvalds's avatar
Linus Torvalds committed
255
			if (tb->port == snum)
256
				goto tb_found;
Linus Torvalds's avatar
Linus Torvalds committed
257
	}
258 259 260 261
	tb = NULL;
	goto tb_not_found;
tb_found:
	if (!hlist_empty(&tb->owners)) {
262
		if (sk->sk_reuse > 1)
263
			goto success;
264 265
		if (tb->fastreuse > 0 &&
		    sk->sk_reuse && sk->sk_state != TCP_LISTEN) {
Linus Torvalds's avatar
Linus Torvalds committed
266 267
			goto success;
		} else {
268
			ret = 1;
Linus Torvalds's avatar
Linus Torvalds committed
269
			if (tcp_bind_conflict(sk, tb))
270
				goto fail_unlock;
Linus Torvalds's avatar
Linus Torvalds committed
271 272
		}
	}
273
tb_not_found:
Linus Torvalds's avatar
Linus Torvalds committed
274
	ret = 1;
275 276
	if (!tb && (tb = tcp_bucket_create(head, snum)) == NULL)
		goto fail_unlock;
277
	if (hlist_empty(&tb->owners)) {
278
		if (sk->sk_reuse && sk->sk_state != TCP_LISTEN)
Linus Torvalds's avatar
Linus Torvalds committed
279 280 281
			tb->fastreuse = 1;
		else
			tb->fastreuse = 0;
282 283
	} else if (tb->fastreuse &&
		   (!sk->sk_reuse || sk->sk_state == TCP_LISTEN))
Linus Torvalds's avatar
Linus Torvalds committed
284 285
		tb->fastreuse = 0;
success:
286
	if (!tcp_sk(sk)->bind_hash)
287
		tcp_bind_hash(sk, tb, snum);
288
	BUG_TRAP(tcp_sk(sk)->bind_hash == tb);
Linus Torvalds's avatar
Linus Torvalds committed
289
 	ret = 0;
Linus Torvalds's avatar
Linus Torvalds committed
290 291 292 293 294 295 296 297 298 299 300

fail_unlock:
	spin_unlock(&head->lock);
fail:
	local_bh_enable();
	return ret;
}

/* Get rid of any references to a local port held by the
 * given sock.
 */
301
static void __tcp_put_port(struct sock *sk)
Linus Torvalds's avatar
Linus Torvalds committed
302
{
303 304
	struct inet_opt *inet = inet_sk(sk);
	struct tcp_bind_hashbucket *head = &tcp_bhash[tcp_bhashfn(inet->num)];
Linus Torvalds's avatar
Linus Torvalds committed
305 306 307
	struct tcp_bind_bucket *tb;

	spin_lock(&head->lock);
308
	tb = tcp_sk(sk)->bind_hash;
309
	__sk_del_bind_node(sk);
310
	tcp_sk(sk)->bind_hash = NULL;
311
	inet->num = 0;
312
	tcp_bucket_destroy(tb);
Linus Torvalds's avatar
Linus Torvalds committed
313 314 315
	spin_unlock(&head->lock);
}

316
void tcp_put_port(struct sock *sk)
Linus Torvalds's avatar
Linus Torvalds committed
317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334
{
	local_bh_disable();
	__tcp_put_port(sk);
	local_bh_enable();
}

/* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it can be very bad on SMP.
 * Look, when several writers sleep and reader wakes them up, all but one
 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
 * this, _but_ remember, it adds useless work on UP machines (wake up each
 * exclusive lock release). It should be ifdefed really.
 */

void tcp_listen_wlock(void)
{
	write_lock(&tcp_lhash_lock);

	if (atomic_read(&tcp_lhash_users)) {
335
		DEFINE_WAIT(wait);
Linus Torvalds's avatar
Linus Torvalds committed
336 337

		for (;;) {
338 339
			prepare_to_wait_exclusive(&tcp_lhash_wait,
						&wait, TASK_UNINTERRUPTIBLE);
340
			if (!atomic_read(&tcp_lhash_users))
Linus Torvalds's avatar
Linus Torvalds committed
341 342 343 344 345 346
				break;
			write_unlock_bh(&tcp_lhash_lock);
			schedule();
			write_lock_bh(&tcp_lhash_lock);
		}

347
		finish_wait(&tcp_lhash_wait, &wait);
Linus Torvalds's avatar
Linus Torvalds committed
348 349 350
	}
}

351
static __inline__ void __tcp_v4_hash(struct sock *sk, const int listen_possible)
Linus Torvalds's avatar
Linus Torvalds committed
352
{
353
	struct hlist_head *list;
Linus Torvalds's avatar
Linus Torvalds committed
354 355
	rwlock_t *lock;

356
	BUG_TRAP(sk_unhashed(sk));
357
	if (listen_possible && sk->sk_state == TCP_LISTEN) {
358
		list = &tcp_listening_hash[tcp_sk_listen_hashfn(sk)];
Linus Torvalds's avatar
Linus Torvalds committed
359 360 361
		lock = &tcp_lhash_lock;
		tcp_listen_wlock();
	} else {
362
		list = &tcp_ehash[(sk->sk_hashent = tcp_sk_hashfn(sk))].chain;
363
		lock = &tcp_ehash[sk->sk_hashent].lock;
Linus Torvalds's avatar
Linus Torvalds committed
364 365
		write_lock(lock);
	}
366
	__sk_add_node(sk, list);
367
	sock_prot_inc_use(sk->sk_prot);
Linus Torvalds's avatar
Linus Torvalds committed
368
	write_unlock(lock);
369
	if (listen_possible && sk->sk_state == TCP_LISTEN)
Linus Torvalds's avatar
Linus Torvalds committed
370 371 372 373 374
		wake_up(&tcp_lhash_wait);
}

static void tcp_v4_hash(struct sock *sk)
{
375
	if (sk->sk_state != TCP_CLOSE) {
Linus Torvalds's avatar
Linus Torvalds committed
376
		local_bh_disable();
377
		__tcp_v4_hash(sk, 1);
Linus Torvalds's avatar
Linus Torvalds committed
378 379 380 381 382 383 384 385
		local_bh_enable();
	}
}

void tcp_unhash(struct sock *sk)
{
	rwlock_t *lock;

386
	if (sk_unhashed(sk))
387 388
		goto ende;

389
	if (sk->sk_state == TCP_LISTEN) {
Linus Torvalds's avatar
Linus Torvalds committed
390 391 392 393
		local_bh_disable();
		tcp_listen_wlock();
		lock = &tcp_lhash_lock;
	} else {
394
		struct tcp_ehash_bucket *head = &tcp_ehash[sk->sk_hashent];
Linus Torvalds's avatar
Linus Torvalds committed
395 396 397 398
		lock = &head->lock;
		write_lock_bh(&head->lock);
	}

399
	if (__sk_del_node_init(sk))
400
		sock_prot_dec_use(sk->sk_prot);
Linus Torvalds's avatar
Linus Torvalds committed
401
	write_unlock_bh(lock);
402 403

 ende:
404
	if (sk->sk_state == TCP_LISTEN)
Linus Torvalds's avatar
Linus Torvalds committed
405 406 407 408 409 410 411 412 413
		wake_up(&tcp_lhash_wait);
}

/* Don't inline this cruft.  Here are some nice properties to
 * exploit here.  The BSD API does not allow a listening TCP
 * to specify the remote port nor the remote address for the
 * connection.  So always assume those are both wildcarded
 * during the search since they can never be otherwise.
 */
414
static struct sock *__tcp_v4_lookup_listener(struct hlist_head *head, u32 daddr,
415
					     unsigned short hnum, int dif)
Linus Torvalds's avatar
Linus Torvalds committed
416
{
417 418
	struct sock *result = NULL, *sk;
	struct hlist_node *node;
Linus Torvalds's avatar
Linus Torvalds committed
419 420
	int score, hiscore;

421
	hiscore=-1;
422
	sk_for_each(sk, node, head) {
423 424
		struct inet_opt *inet = inet_sk(sk);

425
		if (inet->num == hnum && !ipv6_only_sock(sk)) {
426
			__u32 rcv_saddr = inet->rcv_saddr;
Linus Torvalds's avatar
Linus Torvalds committed
427

428
			score = (sk->sk_family == PF_INET ? 1 : 0);
429
			if (rcv_saddr) {
Linus Torvalds's avatar
Linus Torvalds committed
430 431
				if (rcv_saddr != daddr)
					continue;
432
				score+=2;
Linus Torvalds's avatar
Linus Torvalds committed
433
			}
434 435
			if (sk->sk_bound_dev_if) {
				if (sk->sk_bound_dev_if != dif)
Linus Torvalds's avatar
Linus Torvalds committed
436
					continue;
437
				score+=2;
Linus Torvalds's avatar
Linus Torvalds committed
438
			}
439
			if (score == 5)
Linus Torvalds's avatar
Linus Torvalds committed
440 441 442 443 444 445 446 447 448 449 450
				return sk;
			if (score > hiscore) {
				hiscore = score;
				result = sk;
			}
		}
	}
	return result;
}

/* Optimize the common listener case. */
451 452
inline struct sock *tcp_v4_lookup_listener(u32 daddr, unsigned short hnum,
					   int dif)
Linus Torvalds's avatar
Linus Torvalds committed
453
{
454 455
	struct sock *sk = NULL;
	struct hlist_head *head;
Linus Torvalds's avatar
Linus Torvalds committed
456 457

	read_lock(&tcp_lhash_lock);
458 459 460
	head = &tcp_listening_hash[tcp_lhashfn(hnum)];
	if (!hlist_empty(head)) {
		struct inet_opt *inet = inet_sk((sk = __sk_head(head)));
461

462
		if (inet->num == hnum && !sk->sk_node.next &&
463
		    (!inet->rcv_saddr || inet->rcv_saddr == daddr) &&
464 465
		    (sk->sk_family == PF_INET || !ipv6_only_sock(sk)) &&
		    !sk->sk_bound_dev_if)
Linus Torvalds's avatar
Linus Torvalds committed
466
			goto sherry_cache;
467
		sk = __tcp_v4_lookup_listener(head, daddr, hnum, dif);
Linus Torvalds's avatar
Linus Torvalds committed
468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483
	}
	if (sk) {
sherry_cache:
		sock_hold(sk);
	}
	read_unlock(&tcp_lhash_lock);
	return sk;
}

/* Sockets in TCP_CLOSE state are _always_ taken out of the hash, so
 * we need not check it for TCP lookups anymore, thanks Alexey. -DaveM
 *
 * Local BH must be disabled here.
 */

static inline struct sock *__tcp_v4_lookup_established(u32 saddr, u16 sport,
484 485
						       u32 daddr, u16 hnum,
						       int dif)
Linus Torvalds's avatar
Linus Torvalds committed
486 487 488 489 490
{
	struct tcp_ehash_bucket *head;
	TCP_V4_ADDR_COOKIE(acookie, saddr, daddr)
	__u32 ports = TCP_COMBINED_PORTS(sport, hnum);
	struct sock *sk;
491
	struct hlist_node *node;
Linus Torvalds's avatar
Linus Torvalds committed
492 493 494
	/* Optimize here for direct hit, only listening connections can
	 * have wildcards anyways.
	 */
495
	int hash = tcp_hashfn(daddr, hnum, saddr, sport);
Linus Torvalds's avatar
Linus Torvalds committed
496 497
	head = &tcp_ehash[hash];
	read_lock(&head->lock);
498
	sk_for_each(sk, node, &head->chain) {
499
		if (TCP_IPV4_MATCH(sk, acookie, saddr, daddr, ports, dif))
Linus Torvalds's avatar
Linus Torvalds committed
500 501 502 503
			goto hit; /* You sunk my battleship! */
	}

	/* Must check for a TIME_WAIT'er before going to listener hash. */
504
	sk_for_each(sk, node, &(head + tcp_ehash_size)->chain) {
505
		if (TCP_IPV4_TW_MATCH(sk, acookie, saddr, daddr, ports, dif))
Linus Torvalds's avatar
Linus Torvalds committed
506
			goto hit;
507 508
	}
	sk = NULL;
509
out:
Linus Torvalds's avatar
Linus Torvalds committed
510
	read_unlock(&head->lock);
511
	return sk;
Linus Torvalds's avatar
Linus Torvalds committed
512 513
hit:
	sock_hold(sk);
514
	goto out;
Linus Torvalds's avatar
Linus Torvalds committed
515 516 517 518 519
}

static inline struct sock *__tcp_v4_lookup(u32 saddr, u16 sport,
					   u32 daddr, u16 hnum, int dif)
{
520 521
	struct sock *sk = __tcp_v4_lookup_established(saddr, sport,
						      daddr, hnum, dif);
Linus Torvalds's avatar
Linus Torvalds committed
522

523
	return sk ? : tcp_v4_lookup_listener(daddr, hnum, dif);
Linus Torvalds's avatar
Linus Torvalds committed
524 525
}

526 527
inline struct sock *tcp_v4_lookup(u32 saddr, u16 sport, u32 daddr,
				  u16 dport, int dif)
Linus Torvalds's avatar
Linus Torvalds committed
528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545
{
	struct sock *sk;

	local_bh_disable();
	sk = __tcp_v4_lookup(saddr, sport, daddr, ntohs(dport), dif);
	local_bh_enable();

	return sk;
}

static inline __u32 tcp_v4_init_sequence(struct sock *sk, struct sk_buff *skb)
{
	return secure_tcp_sequence_number(skb->nh.iph->daddr,
					  skb->nh.iph->saddr,
					  skb->h.th->dest,
					  skb->h.th->source);
}

546 547 548
/* called with local bh disabled */
static int __tcp_v4_check_established(struct sock *sk, __u16 lport,
				      struct tcp_tw_bucket **twp)
Linus Torvalds's avatar
Linus Torvalds committed
549
{
550 551 552
	struct inet_opt *inet = inet_sk(sk);
	u32 daddr = inet->rcv_saddr;
	u32 saddr = inet->daddr;
553
	int dif = sk->sk_bound_dev_if;
Linus Torvalds's avatar
Linus Torvalds committed
554
	TCP_V4_ADDR_COOKIE(acookie, saddr, daddr)
555 556
	__u32 ports = TCP_COMBINED_PORTS(inet->dport, lport);
	int hash = tcp_hashfn(daddr, lport, saddr, inet->dport);
Linus Torvalds's avatar
Linus Torvalds committed
557
	struct tcp_ehash_bucket *head = &tcp_ehash[hash];
558 559
	struct sock *sk2;
	struct hlist_node *node;
Linus Torvalds's avatar
Linus Torvalds committed
560 561
	struct tcp_tw_bucket *tw;

562
	write_lock(&head->lock);
Linus Torvalds's avatar
Linus Torvalds committed
563 564

	/* Check TIME-WAIT sockets first. */
565
	sk_for_each(sk2, node, &(head + tcp_ehash_size)->chain) {
566
		tw = (struct tcp_tw_bucket *)sk2;
Linus Torvalds's avatar
Linus Torvalds committed
567

568
		if (TCP_IPV4_TW_MATCH(sk2, acookie, saddr, daddr, ports, dif)) {
569
			struct tcp_opt *tp = tcp_sk(sk);
Linus Torvalds's avatar
Linus Torvalds committed
570 571 572 573 574 575 576 577 578 579 580 581 582 583 584

			/* With PAWS, it is safe from the viewpoint
			   of data integrity. Even without PAWS it
			   is safe provided sequence spaces do not
			   overlap i.e. at data rates <= 80Mbit/sec.

			   Actually, the idea is close to VJ's one,
			   only timestamp cache is held not per host,
			   but per port pair and TW bucket is used
			   as state holder.

			   If TW bucket has been already destroyed we
			   fall back to VJ's scheme and use initial
			   timestamp retrieved from peer table.
			 */
585
			if (tw->tw_ts_recent_stamp &&
586
			    (!twp || (sysctl_tcp_tw_reuse &&
587
				      xtime.tv_sec -
588
				      tw->tw_ts_recent_stamp > 1))) {
589
				if ((tp->write_seq =
590
						tw->tw_snd_nxt + 65535 + 2) == 0)
Linus Torvalds's avatar
Linus Torvalds committed
591
					tp->write_seq = 1;
592 593
				tp->ts_recent	    = tw->tw_ts_recent;
				tp->ts_recent_stamp = tw->tw_ts_recent_stamp;
Linus Torvalds's avatar
Linus Torvalds committed
594 595 596 597 598 599 600 601 602
				sock_hold(sk2);
				goto unique;
			} else
				goto not_unique;
		}
	}
	tw = NULL;

	/* And established part... */
603
	sk_for_each(sk2, node, &head->chain) {
604
		if (TCP_IPV4_MATCH(sk2, acookie, saddr, daddr, ports, dif))
Linus Torvalds's avatar
Linus Torvalds committed
605 606 607 608
			goto not_unique;
	}

unique:
609 610
	/* Must record num and sport now. Otherwise we will see
	 * in hash table socket with a funny identity. */
611 612
	inet->num = lport;
	inet->sport = htons(lport);
613
	sk->sk_hashent = hash;
614
	BUG_TRAP(sk_unhashed(sk));
615
	__sk_add_node(sk, &head->chain);
616
	sock_prot_inc_use(sk->sk_prot);
617
	write_unlock(&head->lock);
Linus Torvalds's avatar
Linus Torvalds committed
618

619 620 621 622
	if (twp) {
		*twp = tw;
		NET_INC_STATS_BH(TimeWaitRecycled);
	} else if (tw) {
Linus Torvalds's avatar
Linus Torvalds committed
623 624 625 626 627 628 629 630 631 632
		/* Silly. Should hash-dance instead... */
		tcp_tw_deschedule(tw);
		NET_INC_STATS_BH(TimeWaitRecycled);

		tcp_tw_put(tw);
	}

	return 0;

not_unique:
633
	write_unlock(&head->lock);
Linus Torvalds's avatar
Linus Torvalds committed
634 635 636
	return -EADDRNOTAVAIL;
}

637 638
/*
 * Bind a port for a connect operation and hash it.
Linus Torvalds's avatar
Linus Torvalds committed
639
 */
640
static int tcp_v4_hash_connect(struct sock *sk)
Linus Torvalds's avatar
Linus Torvalds committed
641
{
642
	unsigned short snum = inet_sk(sk)->num;
643 644
 	struct tcp_bind_hashbucket *head;
 	struct tcp_bind_bucket *tb;
645 646 647
	int ret;

 	if (!snum) {
648 649 650 651
 		int rover;
 		int low = sysctl_local_port_range[0];
 		int high = sysctl_local_port_range[1];
 		int remaining = (high - low) + 1;
652
		struct hlist_node *node;
653
 		struct tcp_tw_bucket *tw = NULL;
654

655
 		local_bh_disable();
656

657 658 659 660 661 662 663 664 665 666 667 668 669 670 671
 		/* TODO. Actually it is not so bad idea to remove
 		 * tcp_portalloc_lock before next submission to Linus.
 		 * As soon as we touch this place at all it is time to think.
 		 *
 		 * Now it protects single _advisory_ variable tcp_port_rover,
 		 * hence it is mostly useless.
 		 * Code will work nicely if we just delete it, but
 		 * I am afraid in contented case it will work not better or
 		 * even worse: another cpu just will hit the same bucket
 		 * and spin there.
 		 * So some cpu salt could remove both contention and
 		 * memory pingpong. Any ideas how to do this in a nice way?
 		 */
 		spin_lock(&tcp_portalloc_lock);
 		rover = tcp_port_rover;
672

673 674 675 676 677
 		do {
 			rover++;
 			if ((rover < low) || (rover > high))
 				rover = low;
 			head = &tcp_bhash[tcp_bhashfn(rover)];
678 679
 			spin_lock(&head->lock);

680 681 682 683
 			/* Does not bother with rcv_saddr checks,
 			 * because the established check is already
 			 * unique enough.
 			 */
684
			tb_for_each(tb, node, &head->chain) {
685
 				if (tb->port == rover) {
686
 					BUG_TRAP(!hlist_empty(&tb->owners));
687 688
 					if (tb->fastreuse >= 0)
 						goto next_port;
689 690 691
 					if (!__tcp_v4_check_established(sk,
									rover,
									&tw))
692 693 694 695
 						goto ok;
 					goto next_port;
 				}
 			}
696

697 698 699 700 701 702 703
 			tb = tcp_bucket_create(head, rover);
 			if (!tb) {
 				spin_unlock(&head->lock);
 				break;
 			}
 			tb->fastreuse = -1;
 			goto ok;
704

705 706 707 708 709
 		next_port:
 			spin_unlock(&head->lock);
 		} while (--remaining > 0);
 		tcp_port_rover = rover;
 		spin_unlock(&tcp_portalloc_lock);
710

711
 		local_bh_enable();
712

713
 		return -EADDRNOTAVAIL;
714 715

ok:
716 717 718
 		/* All locks still held and bhs disabled */
 		tcp_port_rover = rover;
 		spin_unlock(&tcp_portalloc_lock);
719

720
 		tcp_bind_hash(sk, tb, rover);
721
		if (sk_unhashed(sk)) {
722
 			inet_sk(sk)->sport = htons(rover);
723 724 725 726 727 728 729 730
 			__tcp_v4_hash(sk, 0);
 		}
 		spin_unlock(&head->lock);

 		if (tw) {
 			tcp_tw_deschedule(tw);
 			tcp_tw_put(tw);
 		}
731 732 733

		ret = 0;
		goto out;
734
 	}
735

736
 	head  = &tcp_bhash[tcp_bhashfn(snum)];
737
 	tb  = tcp_sk(sk)->bind_hash;
Linus Torvalds's avatar
Linus Torvalds committed
738
	spin_lock_bh(&head->lock);
739
	if (sk_head(&tb->owners) == sk && !sk->sk_bind_node.next) {
740
		__tcp_v4_hash(sk, 0);
Linus Torvalds's avatar
Linus Torvalds committed
741 742 743
		spin_unlock_bh(&head->lock);
		return 0;
	} else {
744
		spin_unlock(&head->lock);
Linus Torvalds's avatar
Linus Torvalds committed
745
		/* No definite answer... Walk to established hash table */
746
		ret = __tcp_v4_check_established(sk, snum, NULL);
747
out:
748 749
		local_bh_enable();
		return ret;
Linus Torvalds's avatar
Linus Torvalds committed
750 751 752 753 754 755
	}
}

/* This will initiate an outgoing connection. */
int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
{
756 757
	struct inet_opt *inet = inet_sk(sk);
	struct tcp_opt *tp = tcp_sk(sk);
758
	struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
Linus Torvalds's avatar
Linus Torvalds committed
759 760 761 762 763 764
	struct rtable *rt;
	u32 daddr, nexthop;
	int tmp;
	int err;

	if (addr_len < sizeof(struct sockaddr_in))
765
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
766 767

	if (usin->sin_family != AF_INET)
768
		return -EAFNOSUPPORT;
Linus Torvalds's avatar
Linus Torvalds committed
769 770

	nexthop = daddr = usin->sin_addr.s_addr;
771
	if (inet->opt && inet->opt->srr) {
772
		if (!daddr)
Linus Torvalds's avatar
Linus Torvalds committed
773
			return -EINVAL;
774
		nexthop = inet->opt->faddr;
Linus Torvalds's avatar
Linus Torvalds committed
775 776
	}

777
	tmp = ip_route_connect(&rt, nexthop, inet->saddr,
778
			       RT_CONN_FLAGS(sk), sk->sk_bound_dev_if,
779
			       IPPROTO_TCP,
780
			       inet->sport, usin->sin_port, sk);
Linus Torvalds's avatar
Linus Torvalds committed
781 782 783
	if (tmp < 0)
		return tmp;

784
	if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
Linus Torvalds's avatar
Linus Torvalds committed
785 786 787 788
		ip_rt_put(rt);
		return -ENETUNREACH;
	}

789
	if (!inet->opt || !inet->opt->srr)
Linus Torvalds's avatar
Linus Torvalds committed
790 791
		daddr = rt->rt_dst;

792 793 794
	if (!inet->saddr)
		inet->saddr = rt->rt_src;
	inet->rcv_saddr = inet->saddr;
Linus Torvalds's avatar
Linus Torvalds committed
795

796
	if (tp->ts_recent_stamp && inet->daddr != daddr) {
Linus Torvalds's avatar
Linus Torvalds committed
797
		/* Reset inherited state */
798
		tp->ts_recent	    = 0;
Linus Torvalds's avatar
Linus Torvalds committed
799
		tp->ts_recent_stamp = 0;
800
		tp->write_seq	    = 0;
Linus Torvalds's avatar
Linus Torvalds committed
801 802 803
	}

	if (sysctl_tcp_tw_recycle &&
804
	    !tp->ts_recent_stamp && rt->rt_dst == daddr) {
Linus Torvalds's avatar
Linus Torvalds committed
805 806 807 808 809 810 811 812 813 814 815 816 817
		struct inet_peer *peer = rt_get_peer(rt);

		/* VJ's idea. We save last timestamp seen from
		 * the destination in peer table, when entering state TIME-WAIT
		 * and initialize ts_recent from it, when trying new connection.
		 */

		if (peer && peer->tcp_ts_stamp + TCP_PAWS_MSL >= xtime.tv_sec) {
			tp->ts_recent_stamp = peer->tcp_ts_stamp;
			tp->ts_recent = peer->tcp_ts;
		}
	}

818 819
	inet->dport = usin->sin_port;
	inet->daddr = daddr;
Linus Torvalds's avatar
Linus Torvalds committed
820

821 822 823 824 825 826 827 828 829
	tp->ext_header_len = 0;
	if (inet->opt)
		tp->ext_header_len = inet->opt->optlen;

	tp->mss_clamp = 536;

	/* Socket identity is still unknown (sport may be zero).
	 * However we set state to SYN-SENT and not releasing socket
	 * lock select source port, enter ourselves into the hash tables and
Steven Cole's avatar
Steven Cole committed
830
	 * complete initialization after this.
831 832 833 834 835 836
	 */
	tcp_set_state(sk, TCP_SYN_SENT);
	err = tcp_v4_hash_connect(sk);
	if (err)
		goto failure;

837
	err = ip_route_newports(&rt, inet->sport, inet->dport, sk);
838 839 840 841 842 843
	if (err)
		goto failure;

	/* OK, now commit destination to socket.  */
	__sk_dst_set(sk, &rt->u.dst);
	tcp_v4_setup_caps(sk, &rt->u.dst);
844
	tp->ext2_header_len = rt->u.dst.header_len;
845

Linus Torvalds's avatar
Linus Torvalds committed
846
	if (!tp->write_seq)
847 848 849
		tp->write_seq = secure_tcp_sequence_number(inet->saddr,
							   inet->daddr,
							   inet->sport,
850
							   usin->sin_port);
Linus Torvalds's avatar
Linus Torvalds committed
851

852
	inet->id = tp->write_seq ^ jiffies;
Linus Torvalds's avatar
Linus Torvalds committed
853

854
	err = tcp_connect(sk);
855
	rt = NULL;
856 857
	if (err)
		goto failure;
Linus Torvalds's avatar
Linus Torvalds committed
858

859
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
860

Linus Torvalds's avatar
Linus Torvalds committed
861
failure:
862
	/* This unhashes the socket and releases the local port, if necessary. */
863
	tcp_set_state(sk, TCP_CLOSE);
864
	ip_rt_put(rt);
865
	sk->sk_route_caps = 0;
866
	inet->dport = 0;
Linus Torvalds's avatar
Linus Torvalds committed
867 868 869 870 871
	return err;
}

static __inline__ int tcp_v4_iif(struct sk_buff *skb)
{
872
	return ((struct rtable *)skb->dst)->rt_iif;
Linus Torvalds's avatar
Linus Torvalds committed
873 874
}

875
static __inline__ u32 tcp_v4_synq_hash(u32 raddr, u16 rport, u32 rnd)
Linus Torvalds's avatar
Linus Torvalds committed
876
{
877
	return (jhash_2words(raddr, (u32) rport, rnd) & (TCP_SYNQ_HSIZE - 1));
Linus Torvalds's avatar
Linus Torvalds committed
878 879
}

880
static struct open_request *tcp_v4_search_req(struct tcp_opt *tp,
Linus Torvalds's avatar
Linus Torvalds committed
881 882 883
					      struct open_request ***prevp,
					      __u16 rport,
					      __u32 raddr, __u32 laddr)
Linus Torvalds's avatar
Linus Torvalds committed
884 885
{
	struct tcp_listen_opt *lopt = tp->listen_opt;
886
	struct open_request *req, **prev;
Linus Torvalds's avatar
Linus Torvalds committed
887

888
	for (prev = &lopt->syn_table[tcp_v4_synq_hash(raddr, rport, lopt->hash_rnd)];
Linus Torvalds's avatar
Linus Torvalds committed
889 890 891 892
	     (req = *prev) != NULL;
	     prev = &req->dl_next) {
		if (req->rmt_port == rport &&
		    req->af.v4_req.rmt_addr == raddr &&
Linus Torvalds's avatar
Linus Torvalds committed
893
		    req->af.v4_req.loc_addr == laddr &&
Linus Torvalds's avatar
Linus Torvalds committed
894
		    TCP_INET_FAMILY(req->class->family)) {
895
			BUG_TRAP(!req->sk);
Linus Torvalds's avatar
Linus Torvalds committed
896
			*prevp = prev;
897
			break;
Linus Torvalds's avatar
Linus Torvalds committed
898 899 900
		}
	}

901
	return req;
Linus Torvalds's avatar
Linus Torvalds committed
902 903 904 905
}

static void tcp_v4_synq_add(struct sock *sk, struct open_request *req)
{
906
	struct tcp_opt *tp = tcp_sk(sk);
Linus Torvalds's avatar
Linus Torvalds committed
907
	struct tcp_listen_opt *lopt = tp->listen_opt;
908
	u32 h = tcp_v4_synq_hash(req->af.v4_req.rmt_addr, req->rmt_port, lopt->hash_rnd);
Linus Torvalds's avatar
Linus Torvalds committed
909 910 911 912 913 914 915 916 917 918 919 920 921 922

	req->expires = jiffies + TCP_TIMEOUT_INIT;
	req->retrans = 0;
	req->sk = NULL;
	req->dl_next = lopt->syn_table[h];

	write_lock(&tp->syn_wait_lock);
	lopt->syn_table[h] = req;
	write_unlock(&tp->syn_wait_lock);

	tcp_synq_added(sk);
}


923
/*
Linus Torvalds's avatar
Linus Torvalds committed
924 925
 * This routine does path mtu discovery as defined in RFC1191.
 */
926
static inline void do_pmtu_discovery(struct sock *sk, struct iphdr *iph,
927
				     u32 mtu)
Linus Torvalds's avatar
Linus Torvalds committed
928 929
{
	struct dst_entry *dst;
930 931
	struct inet_opt *inet = inet_sk(sk);
	struct tcp_opt *tp = tcp_sk(sk);
Linus Torvalds's avatar
Linus Torvalds committed
932 933 934 935 936

	/* We are not interested in TCP_LISTEN and open_requests (SYN-ACKs
	 * send out by Linux are always <576bytes so they should go through
	 * unfragmented).
	 */
937
	if (sk->sk_state == TCP_LISTEN)
938
		return;
Linus Torvalds's avatar
Linus Torvalds committed
939 940 941 942 943 944 945 946 947 948

	/* We don't check in the destentry if pmtu discovery is forbidden
	 * on this route. We just assume that no packet_to_big packets
	 * are send back when pmtu discovery is not active.
     	 * There is a small race when the user changes this flag in the
	 * route, but I think that's acceptable.
	 */
	if ((dst = __sk_dst_check(sk, 0)) == NULL)
		return;

949
	dst->ops->update_pmtu(dst, mtu);
Linus Torvalds's avatar
Linus Torvalds committed
950 951 952 953

	/* Something is about to be wrong... Remember soft error
	 * for the case, if this connection will not able to recover.
	 */
954
	if (mtu < dst_pmtu(dst) && ip_dont_fragment(sk, dst))
955
		sk->sk_err_soft = EMSGSIZE;
Linus Torvalds's avatar
Linus Torvalds committed
956

957 958
	mtu = dst_pmtu(dst);

959
	if (inet->pmtudisc != IP_PMTUDISC_DONT &&
960 961
	    tp->pmtu_cookie > mtu) {
		tcp_sync_mss(sk, mtu);
Linus Torvalds's avatar
Linus Torvalds committed
962

963
		/* Resend the TCP packet because it's
Linus Torvalds's avatar
Linus Torvalds committed
964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987
		 * clear that the old packet has been
		 * dropped. This is the new "fast" path mtu
		 * discovery.
		 */
		tcp_simple_retransmit(sk);
	} /* else let the usual retransmit timer handle it */
}

/*
 * This routine is called by the ICMP module when it gets some
 * sort of error condition.  If err < 0 then the socket should
 * be closed and the error returned to the user.  If err > 0
 * it's just the icmp type << 8 | icmp code.  After adjustment
 * header points to the first 8 bytes of the tcp header.  We need
 * to find the appropriate port.
 *
 * The locking strategy used here is very "optimistic". When
 * someone else accesses the socket the ICMP is just dropped
 * and for some paths there is no check at all.
 * A more general error queue to queue errors for later handling
 * is probably better.
 *
 */

Linus Torvalds's avatar
Linus Torvalds committed
988
void tcp_v4_err(struct sk_buff *skb, u32 info)
Linus Torvalds's avatar
Linus Torvalds committed
989
{
990 991
	struct iphdr *iph = (struct iphdr *)skb->data;
	struct tcphdr *th = (struct tcphdr *)(skb->data + (iph->ihl << 2));
Linus Torvalds's avatar
Linus Torvalds committed
992
	struct tcp_opt *tp;
993
	struct inet_opt *inet;
Linus Torvalds's avatar
Linus Torvalds committed
994 995 996 997 998 999
	int type = skb->h.icmph->type;
	int code = skb->h.icmph->code;
	struct sock *sk;
	__u32 seq;
	int err;

Linus Torvalds's avatar
Linus Torvalds committed
1000
	if (skb->len < (iph->ihl << 2) + 8) {
1001
		ICMP_INC_STATS_BH(IcmpInErrors);
Linus Torvalds's avatar
Linus Torvalds committed
1002 1003 1004
		return;
	}

1005 1006 1007
	sk = tcp_v4_lookup(iph->daddr, th->dest, iph->saddr,
			   th->source, tcp_v4_iif(skb));
	if (!sk) {
Linus Torvalds's avatar
Linus Torvalds committed
1008 1009 1010
		ICMP_INC_STATS_BH(IcmpInErrors);
		return;
	}
1011
	if (sk->sk_state == TCP_TIME_WAIT) {
1012
		tcp_tw_put((struct tcp_tw_bucket *)sk);
Linus Torvalds's avatar
Linus Torvalds committed
1013 1014 1015 1016 1017 1018 1019
		return;
	}

	bh_lock_sock(sk);
	/* If too many ICMPs get dropped on busy
	 * servers this needs to be solved differently.
	 */
1020
	if (sock_owned_by_user(sk))
Linus Torvalds's avatar
Linus Torvalds committed
1021 1022
		NET_INC_STATS_BH(LockDroppedIcmps);

1023
	if (sk->sk_state == TCP_CLOSE)
Linus Torvalds's avatar
Linus Torvalds committed
1024 1025
		goto out;

1026
	tp = tcp_sk(sk);
Linus Torvalds's avatar
Linus Torvalds committed
1027
	seq = ntohl(th->seq);
1028
	if (sk->sk_state != TCP_LISTEN &&
1029
	    !between(seq, tp->snd_una, tp->snd_nxt)) {
Linus Torvalds's avatar
Linus Torvalds committed
1030 1031 1032 1033 1034 1035 1036 1037 1038
		NET_INC_STATS(OutOfWindowIcmps);
		goto out;
	}

	switch (type) {
	case ICMP_SOURCE_QUENCH:
		/* This is deprecated, but if someone generated it,
		 * we have no reasons to ignore it.
		 */
1039
		if (!sock_owned_by_user(sk))
Linus Torvalds's avatar
Linus Torvalds committed
1040 1041 1042 1043
			tcp_enter_cwr(tp);
		goto out;
	case ICMP_PARAMETERPROB:
		err = EPROTO;
1044
		break;
Linus Torvalds's avatar
Linus Torvalds committed
1045 1046 1047 1048 1049
	case ICMP_DEST_UNREACH:
		if (code > NR_ICMP_UNREACH)
			goto out;

		if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */
1050
			if (!sock_owned_by_user(sk))
Linus Torvalds's avatar
Linus Torvalds committed
1051
				do_pmtu_discovery(sk, iph, info);
Linus Torvalds's avatar
Linus Torvalds committed
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063
			goto out;
		}

		err = icmp_err_convert[code].errno;
		break;
	case ICMP_TIME_EXCEEDED:
		err = EHOSTUNREACH;
		break;
	default:
		goto out;
	}

1064
	switch (sk->sk_state) {
Linus Torvalds's avatar
Linus Torvalds committed
1065 1066
		struct open_request *req, **prev;
	case TCP_LISTEN:
1067
		if (sock_owned_by_user(sk))
Linus Torvalds's avatar
Linus Torvalds committed
1068 1069
			goto out;

1070 1071
		req = tcp_v4_search_req(tp, &prev, th->dest,
					iph->daddr, iph->saddr);
Linus Torvalds's avatar
Linus Torvalds committed
1072 1073 1074 1075 1076 1077
		if (!req)
			goto out;

		/* ICMPs are not backlogged, hence we cannot get
		   an established socket here.
		 */
1078
		BUG_TRAP(!req->sk);
Linus Torvalds's avatar
Linus Torvalds committed
1079 1080 1081 1082 1083 1084

		if (seq != req->snt_isn) {
			NET_INC_STATS_BH(OutOfWindowIcmps);
			goto out;
		}

1085
		/*
Linus Torvalds's avatar
Linus Torvalds committed
1086 1087 1088
		 * Still in SYN_RECV, just remove it silently.
		 * There is no good way to pass the error to the newly
		 * created socket, and POSIX does not want network
1089 1090
		 * errors returned from accept().
		 */
Linus Torvalds's avatar
Linus Torvalds committed
1091 1092 1093 1094 1095 1096
		tcp_synq_drop(sk, req, prev);
		goto out;

	case TCP_SYN_SENT:
	case TCP_SYN_RECV:  /* Cannot happen.
			       It can f.e. if SYNs crossed.
1097
			     */
1098
		if (!sock_owned_by_user(sk)) {
Linus Torvalds's avatar
Linus Torvalds committed
1099
			TCP_INC_STATS_BH(TcpAttemptFails);
1100
			sk->sk_err = err;
Linus Torvalds's avatar
Linus Torvalds committed
1101

1102
			sk->sk_error_report(sk);
Linus Torvalds's avatar
Linus Torvalds committed
1103 1104 1105

			tcp_done(sk);
		} else {
1106
			sk->sk_err_soft = err;
Linus Torvalds's avatar
Linus Torvalds committed
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126
		}
		goto out;
	}

	/* If we've already connected we will keep trying
	 * until we time out, or the user gives up.
	 *
	 * rfc1122 4.2.3.9 allows to consider as hard errors
	 * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too,
	 * but it is obsoleted by pmtu discovery).
	 *
	 * Note, that in modern internet, where routing is unreliable
	 * and in each dark corner broken firewalls sit, sending random
	 * errors ordered by their masters even this two messages finally lose
	 * their original sense (even Linux sends invalid PORT_UNREACHs)
	 *
	 * Now we are in compliance with RFCs.
	 *							--ANK (980905)
	 */

1127
	inet = inet_sk(sk);
1128
	if (!sock_owned_by_user(sk) && inet->recverr) {
1129 1130
		sk->sk_err = err;
		sk->sk_error_report(sk);
Linus Torvalds's avatar
Linus Torvalds committed
1131
	} else	{ /* Only an error on timeout */
1132
		sk->sk_err_soft = err;
Linus Torvalds's avatar
Linus Torvalds committed
1133 1134 1135 1136 1137 1138 1139 1140
	}

out:
	bh_unlock_sock(sk);
	sock_put(sk);
}

/* This routine computes an IPv4 TCP checksum. */
1141
void tcp_v4_send_check(struct sock *sk, struct tcphdr *th, int len,
Linus Torvalds's avatar
Linus Torvalds committed
1142 1143
		       struct sk_buff *skb)
{
1144 1145
	struct inet_opt *inet = inet_sk(sk);

Linus Torvalds's avatar
Linus Torvalds committed
1146
	if (skb->ip_summed == CHECKSUM_HW) {
1147
		th->check = ~tcp_v4_check(th, len, inet->saddr, inet->daddr, 0);
Linus Torvalds's avatar
Linus Torvalds committed
1148 1149
		skb->csum = offsetof(struct tcphdr, check);
	} else {
1150
		th->check = tcp_v4_check(th, len, inet->saddr, inet->daddr,
1151 1152 1153
					 csum_partial((char *)th,
						      th->doff << 2,
						      skb->csum));
Linus Torvalds's avatar
Linus Torvalds committed
1154
	}
Linus Torvalds's avatar
Linus Torvalds committed
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
}

/*
 *	This routine will send an RST to the other tcp.
 *
 *	Someone asks: why I NEVER use socket parameters (TOS, TTL etc.)
 *		      for reset.
 *	Answer: if a packet caused RST, it is not for a socket
 *		existing in our system, if it is matched to a socket,
 *		it is just duplicate segment or bug in other side's TCP.
 *		So that we build reply only basing on parameters
 *		arrived with segment.
 *	Exception: precedence violation. We do not implement it in any case.
 */

static void tcp_v4_send_reset(struct sk_buff *skb)
{
	struct tcphdr *th = skb->h.th;
	struct tcphdr rth;
	struct ip_reply_arg arg;

	/* Never send a reset in response to a reset. */
	if (th->rst)
		return;

1180
	if (((struct rtable *)skb->dst)->rt_type != RTN_LOCAL)
Linus Torvalds's avatar
Linus Torvalds committed
1181 1182 1183
		return;

	/* Swap the send and the receive. */
1184 1185 1186 1187 1188
	memset(&rth, 0, sizeof(struct tcphdr));
	rth.dest   = th->source;
	rth.source = th->dest;
	rth.doff   = sizeof(struct tcphdr) / 4;
	rth.rst    = 1;
Linus Torvalds's avatar
Linus Torvalds committed
1189 1190 1191 1192 1193

	if (th->ack) {
		rth.seq = th->ack_seq;
	} else {
		rth.ack = 1;
1194 1195
		rth.ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin +
				    skb->len - (th->doff << 2));
Linus Torvalds's avatar
Linus Torvalds committed
1196 1197
	}

1198 1199
	memset(&arg, 0, sizeof arg);
	arg.iov[0].iov_base = (unsigned char *)&rth;
Linus Torvalds's avatar
Linus Torvalds committed
1200
	arg.iov[0].iov_len  = sizeof rth;
1201
	arg.csum = csum_tcpudp_nofold(skb->nh.iph->daddr,
Linus Torvalds's avatar
Linus Torvalds committed
1202
				      skb->nh.iph->saddr, /*XXX*/
1203 1204
				      sizeof(struct tcphdr), IPPROTO_TCP, 0);
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
Linus Torvalds's avatar
Linus Torvalds committed
1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215

	ip_send_reply(tcp_socket->sk, skb, &arg, sizeof rth);

	TCP_INC_STATS_BH(TcpOutSegs);
	TCP_INC_STATS_BH(TcpOutRsts);
}

/* The code following below sending ACKs in SYN-RECV and TIME-WAIT states
   outside socket context is ugly, certainly. What can I do?
 */

1216 1217
static void tcp_v4_send_ack(struct sk_buff *skb, u32 seq, u32 ack,
			    u32 win, u32 ts)
Linus Torvalds's avatar
Linus Torvalds committed
1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
{
	struct tcphdr *th = skb->h.th;
	struct {
		struct tcphdr th;
		u32 tsopt[3];
	} rep;
	struct ip_reply_arg arg;

	memset(&rep.th, 0, sizeof(struct tcphdr));
	memset(&arg, 0, sizeof arg);

1229
	arg.iov[0].iov_base = (unsigned char *)&rep;
Linus Torvalds's avatar
Linus Torvalds committed
1230 1231
	arg.iov[0].iov_len  = sizeof(rep.th);
	if (ts) {
1232 1233 1234
		rep.tsopt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				     (TCPOPT_TIMESTAMP << 8) |
				     TCPOLEN_TIMESTAMP);
Linus Torvalds's avatar
Linus Torvalds committed
1235 1236 1237 1238 1239 1240
		rep.tsopt[1] = htonl(tcp_time_stamp);
		rep.tsopt[2] = htonl(ts);
		arg.iov[0].iov_len = sizeof(rep);
	}

	/* Swap the send and the receive. */
1241 1242 1243 1244
	rep.th.dest    = th->source;
	rep.th.source  = th->dest;
	rep.th.doff    = arg.iov[0].iov_len / 4;
	rep.th.seq     = htonl(seq);
Linus Torvalds's avatar
Linus Torvalds committed
1245
	rep.th.ack_seq = htonl(ack);
1246 1247
	rep.th.ack     = 1;
	rep.th.window  = htons(win);
Linus Torvalds's avatar
Linus Torvalds committed
1248

1249
	arg.csum = csum_tcpudp_nofold(skb->nh.iph->daddr,
Linus Torvalds's avatar
Linus Torvalds committed
1250
				      skb->nh.iph->saddr, /*XXX*/
1251 1252
				      arg.iov[0].iov_len, IPPROTO_TCP, 0);
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
Linus Torvalds's avatar
Linus Torvalds committed
1253 1254 1255 1256 1257 1258 1259 1260 1261 1262

	ip_send_reply(tcp_socket->sk, skb, &arg, arg.iov[0].iov_len);

	TCP_INC_STATS_BH(TcpOutSegs);
}

static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_tw_bucket *tw = (struct tcp_tw_bucket *)sk;

1263 1264
	tcp_v4_send_ack(skb, tw->tw_snd_nxt, tw->tw_rcv_nxt,
			tw->tw_rcv_wnd >> tw->tw_rcv_wscale, tw->tw_ts_recent);
Linus Torvalds's avatar
Linus Torvalds committed
1265 1266 1267 1268 1269 1270

	tcp_tw_put(tw);
}

static void tcp_v4_or_send_ack(struct sk_buff *skb, struct open_request *req)
{
1271
	tcp_v4_send_ack(skb, req->snt_isn + 1, req->rcv_isn + 1, req->rcv_wnd,
Linus Torvalds's avatar
Linus Torvalds committed
1272 1273 1274
			req->ts_recent);
}

1275 1276
static struct dst_entry* tcp_v4_route_req(struct sock *sk,
					  struct open_request *req)
Linus Torvalds's avatar
Linus Torvalds committed
1277 1278
{
	struct rtable *rt;
1279
	struct ip_options *opt = req->af.v4_req.opt;
1280
	struct flowi fl = { .oif = sk->sk_bound_dev_if,
1281
			    .nl_u = { .ip4_u =
1282 1283 1284 1285 1286
				      { .daddr = ((opt && opt->srr) ?
						  opt->faddr :
						  req->af.v4_req.rmt_addr),
					.saddr = req->af.v4_req.loc_addr,
					.tos = RT_CONN_FLAGS(sk) } },
1287 1288 1289 1290
			    .proto = IPPROTO_TCP,
			    .uli_u = { .ports =
				       { .sport = inet_sk(sk)->sport,
					 .dport = req->rmt_port } } };
1291

1292
	if (ip_route_output_flow(&rt, &fl, sk, 0)) {
1293
		IP_INC_STATS_BH(OutNoRoutes);
Linus Torvalds's avatar
Linus Torvalds committed
1294 1295 1296 1297
		return NULL;
	}
	if (opt && opt->is_strictroute && rt->rt_dst != rt->rt_gateway) {
		ip_rt_put(rt);
1298
		IP_INC_STATS_BH(OutNoRoutes);
Linus Torvalds's avatar
Linus Torvalds committed
1299 1300 1301 1302 1303 1304
		return NULL;
	}
	return &rt->u.dst;
}

/*
1305
 *	Send a SYN-ACK after having received an ACK.
Linus Torvalds's avatar
Linus Torvalds committed
1306 1307
 *	This still operates on a open_request only, not on a big
 *	socket.
1308
 */
Linus Torvalds's avatar
Linus Torvalds committed
1309 1310 1311 1312 1313 1314 1315
static int tcp_v4_send_synack(struct sock *sk, struct open_request *req,
			      struct dst_entry *dst)
{
	int err = -1;
	struct sk_buff * skb;

	/* First, grab a route. */
1316
	if (!dst && (dst = tcp_v4_route_req(sk, req)) == NULL)
Linus Torvalds's avatar
Linus Torvalds committed
1317 1318 1319 1320 1321 1322 1323 1324
		goto out;

	skb = tcp_make_synack(sk, dst, req);

	if (skb) {
		struct tcphdr *th = skb->h.th;

		th->check = tcp_v4_check(th, skb->len,
1325 1326 1327 1328
					 req->af.v4_req.loc_addr,
					 req->af.v4_req.rmt_addr,
					 csum_partial((char *)th, skb->len,
						      skb->csum));
Linus Torvalds's avatar
Linus Torvalds committed
1329 1330

		err = ip_build_and_send_pkt(skb, sk, req->af.v4_req.loc_addr,
1331 1332
					    req->af.v4_req.rmt_addr,
					    req->af.v4_req.opt);
Linus Torvalds's avatar
Linus Torvalds committed
1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343
		if (err == NET_XMIT_CN)
			err = 0;
	}

out:
	dst_release(dst);
	return err;
}

/*
 *	IPv4 open_request destructor.
1344
 */
Linus Torvalds's avatar
Linus Torvalds committed
1345 1346 1347 1348 1349 1350 1351 1352 1353
static void tcp_v4_or_free(struct open_request *req)
{
	if (req->af.v4_req.opt)
		kfree(req->af.v4_req.opt);
}

static inline void syn_flood_warning(struct sk_buff *skb)
{
	static unsigned long warntime;
1354

1355
	if (time_after(jiffies, (warntime + HZ * 60))) {
Linus Torvalds's avatar
Linus Torvalds committed
1356
		warntime = jiffies;
1357 1358
		printk(KERN_INFO
		       "possible SYN flooding on port %d. Sending cookies.\n",
Linus Torvalds's avatar
Linus Torvalds committed
1359 1360 1361 1362
		       ntohs(skb->h.th->dest));
	}
}

1363 1364
/*
 * Save and compile IPv4 options into the open_request if needed.
Linus Torvalds's avatar
Linus Torvalds committed
1365
 */
1366 1367
static inline struct ip_options *tcp_v4_save_options(struct sock *sk,
						     struct sk_buff *skb)
Linus Torvalds's avatar
Linus Torvalds committed
1368 1369
{
	struct ip_options *opt = &(IPCB(skb)->opt);
1370
	struct ip_options *dopt = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
1371 1372

	if (opt && opt->optlen) {
1373
		int opt_size = optlength(opt);
Linus Torvalds's avatar
Linus Torvalds committed
1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384
		dopt = kmalloc(opt_size, GFP_ATOMIC);
		if (dopt) {
			if (ip_options_echo(dopt, skb)) {
				kfree(dopt);
				dopt = NULL;
			}
		}
	}
	return dopt;
}

1385
/*
Linus Torvalds's avatar
Linus Torvalds committed
1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
 * Maximum number of SYN_RECV sockets in queue per LISTEN socket.
 * One SYN_RECV socket costs about 80bytes on a 32bit machine.
 * It would be better to replace it with a global counter for all sockets
 * but then some measure against one socket starving all other sockets
 * would be needed.
 *
 * It was 128 by default. Experiments with real servers show, that
 * it is absolutely not enough even at 100conn/sec. 256 cures most
 * of problems. This value is adjusted to 128 for very small machines
 * (<=32Mb of memory) and to 1024 on normal or better ones (>=256Mb).
 * Further increasing requires to change hash table size.
 */
1398
int sysctl_max_syn_backlog = 256;
Linus Torvalds's avatar
Linus Torvalds committed
1399 1400

struct or_calltable or_ipv4 = {
1401 1402 1403 1404 1405
	.family		=	PF_INET,
	.rtx_syn_ack	=	tcp_v4_send_synack,
	.send_ack	=	tcp_v4_or_send_ack,
	.destructor	=	tcp_v4_or_free,
	.send_reset	=	tcp_v4_send_reset,
Linus Torvalds's avatar
Linus Torvalds committed
1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422
};

int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_opt tp;
	struct open_request *req;
	__u32 saddr = skb->nh.iph->saddr;
	__u32 daddr = skb->nh.iph->daddr;
	__u32 isn = TCP_SKB_CB(skb)->when;
	struct dst_entry *dst = NULL;
#ifdef CONFIG_SYN_COOKIES
	int want_cookie = 0;
#else
#define want_cookie 0 /* Argh, why doesn't gcc optimize this :( */
#endif

	/* Never answer to SYNs send to broadcast or multicast */
1423 1424 1425
	if (((struct rtable *)skb->dst)->rt_flags &
	    (RTCF_BROADCAST | RTCF_MULTICAST))
		goto drop;
Linus Torvalds's avatar
Linus Torvalds committed
1426 1427 1428 1429 1430 1431 1432 1433

	/* TW buckets are converted to open requests without
	 * limitations, they conserve resources and peer is
	 * evidently real one.
	 */
	if (tcp_synq_is_full(sk) && !isn) {
#ifdef CONFIG_SYN_COOKIES
		if (sysctl_tcp_syncookies) {
1434
			want_cookie = 1;
Linus Torvalds's avatar
Linus Torvalds committed
1435 1436 1437 1438 1439 1440 1441 1442 1443 1444
		} else
#endif
		goto drop;
	}

	/* Accept backlog is full. If we have already queued enough
	 * of warm entries in syn queue, drop request. It is better than
	 * clogging syn queue with openreqs with exponentially increasing
	 * timeout.
	 */
1445
	if (sk_acceptq_is_full(sk) && tcp_synq_young(sk) > 1)
Linus Torvalds's avatar
Linus Torvalds committed
1446 1447 1448
		goto drop;

	req = tcp_openreq_alloc();
1449
	if (!req)
Linus Torvalds's avatar
Linus Torvalds committed
1450 1451 1452 1453
		goto drop;

	tcp_clear_options(&tp);
	tp.mss_clamp = 536;
1454
	tp.user_mss  = tcp_sk(sk)->user_mss;
Linus Torvalds's avatar
Linus Torvalds committed
1455 1456 1457 1458 1459 1460 1461 1462

	tcp_parse_options(skb, &tp, 0);

	if (want_cookie) {
		tcp_clear_options(&tp);
		tp.saw_tstamp = 0;
	}

1463 1464 1465 1466 1467 1468 1469 1470 1471
	if (tp.saw_tstamp && !tp.rcv_tsval) {
		/* Some OSes (unknown ones, but I see them on web server, which
		 * contains information interesting only for windows'
		 * users) do not send their stamp in SYN. It is easy case.
		 * We simply do not advertise TS support.
		 */
		tp.saw_tstamp = 0;
		tp.tstamp_ok  = 0;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487
	tp.tstamp_ok = tp.saw_tstamp;

	tcp_openreq_init(req, &tp, skb);

	req->af.v4_req.loc_addr = daddr;
	req->af.v4_req.rmt_addr = saddr;
	req->af.v4_req.opt = tcp_v4_save_options(sk, skb);
	req->class = &or_ipv4;
	if (!want_cookie)
		TCP_ECN_create_request(req, skb->h.th);

	if (want_cookie) {
#ifdef CONFIG_SYN_COOKIES
		syn_flood_warning(skb);
#endif
		isn = cookie_v4_init_sequence(sk, skb, &req->mss);
1488
	} else if (!isn) {
Linus Torvalds's avatar
Linus Torvalds committed
1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502
		struct inet_peer *peer = NULL;

		/* VJ's idea. We save last timestamp seen
		 * from the destination in peer table, when entering
		 * state TIME-WAIT, and check against it before
		 * accepting new connection request.
		 *
		 * If "isn" is not zero, this request hit alive
		 * timewait bucket, so that all the necessary checks
		 * are made in the function processing timewait state.
		 */
		if (tp.saw_tstamp &&
		    sysctl_tcp_tw_recycle &&
		    (dst = tcp_v4_route_req(sk, req)) != NULL &&
1503
		    (peer = rt_get_peer((struct rtable *)dst)) != NULL &&
Linus Torvalds's avatar
Linus Torvalds committed
1504 1505
		    peer->v4daddr == saddr) {
			if (xtime.tv_sec < peer->tcp_ts_stamp + TCP_PAWS_MSL &&
1506 1507
			    (s32)(peer->tcp_ts - req->ts_recent) >
							TCP_PAWS_WINDOW) {
Linus Torvalds's avatar
Linus Torvalds committed
1508 1509 1510 1511 1512 1513 1514
				NET_INC_STATS_BH(PAWSPassiveRejected);
				dst_release(dst);
				goto drop_and_free;
			}
		}
		/* Kill the following clause, if you dislike this way. */
		else if (!sysctl_tcp_syncookies &&
1515 1516
			 (sysctl_max_syn_backlog - tcp_synq_len(sk) <
			  (sysctl_max_syn_backlog >> 2)) &&
Linus Torvalds's avatar
Linus Torvalds committed
1517
			 (!peer || !peer->tcp_ts_stamp) &&
1518
			 (!dst || !dst_metric(dst, RTAX_RTT))) {
Linus Torvalds's avatar
Linus Torvalds committed
1519
			/* Without syncookies last quarter of
1520 1521
			 * backlog is filled with destinations,
			 * proven to be alive.
Linus Torvalds's avatar
Linus Torvalds committed
1522 1523 1524 1525 1526
			 * It means that we continue to communicate
			 * to destinations, already remembered
			 * to the moment of synflood.
			 */
			NETDEBUG(if (net_ratelimit()) \
1527 1528 1529 1530 1531
					printk(KERN_DEBUG "TCP: drop open "
							  "request from %u.%u."
							  "%u.%u/%u\n", \
					       NIPQUAD(saddr),
					       ntohs(skb->h.th->source)));
Linus Torvalds's avatar
Linus Torvalds committed
1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
			dst_release(dst);
			goto drop_and_free;
		}

		isn = tcp_v4_init_sequence(sk, skb);
	}
	req->snt_isn = isn;

	if (tcp_v4_send_synack(sk, req, dst))
		goto drop_and_free;

	if (want_cookie) {
1544
	   	tcp_openreq_free(req);
Linus Torvalds's avatar
Linus Torvalds committed
1545 1546 1547 1548 1549 1550
	} else {
		tcp_v4_synq_add(sk, req);
	}
	return 0;

drop_and_free:
1551
	tcp_openreq_free(req);
Linus Torvalds's avatar
Linus Torvalds committed
1552 1553 1554 1555 1556 1557
drop:
	TCP_INC_STATS_BH(TcpAttemptFails);
	return 0;
}


1558 1559 1560
/*
 * The three way handshake has completed - we got a valid synack -
 * now create the new socket.
Linus Torvalds's avatar
Linus Torvalds committed
1561
 */
1562 1563 1564
struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
				  struct open_request *req,
				  struct dst_entry *dst)
Linus Torvalds's avatar
Linus Torvalds committed
1565
{
1566
	struct inet_opt *newinet;
Linus Torvalds's avatar
Linus Torvalds committed
1567 1568 1569
	struct tcp_opt *newtp;
	struct sock *newsk;

1570
	if (sk_acceptq_is_full(sk))
Linus Torvalds's avatar
Linus Torvalds committed
1571 1572
		goto exit_overflow;

1573
	if (!dst && (dst = tcp_v4_route_req(sk, req)) == NULL)
Linus Torvalds's avatar
Linus Torvalds committed
1574 1575 1576 1577 1578 1579
		goto exit;

	newsk = tcp_create_openreq_child(sk, req, skb);
	if (!newsk)
		goto exit;

1580
	newsk->sk_dst_cache = dst;
1581
	tcp_v4_setup_caps(newsk, dst);
Linus Torvalds's avatar
Linus Torvalds committed
1582

1583 1584 1585 1586 1587 1588 1589 1590 1591
	newtp		      = tcp_sk(newsk);
	newinet		      = inet_sk(newsk);
	newinet->daddr	      = req->af.v4_req.rmt_addr;
	newinet->rcv_saddr    = req->af.v4_req.loc_addr;
	newinet->saddr	      = req->af.v4_req.loc_addr;
	newinet->opt	      = req->af.v4_req.opt;
	req->af.v4_req.opt    = NULL;
	newinet->mc_index     = tcp_v4_iif(skb);
	newinet->mc_ttl	      = skb->nh.iph->ttl;
Linus Torvalds's avatar
Linus Torvalds committed
1592
	newtp->ext_header_len = 0;
1593 1594
	if (newinet->opt)
		newtp->ext_header_len = newinet->opt->optlen;
1595
	newtp->ext2_header_len = dst->header_len;
1596
	newinet->id = newtp->write_seq ^ jiffies;
Linus Torvalds's avatar
Linus Torvalds committed
1597

1598
	tcp_sync_mss(newsk, dst_pmtu(dst));
1599
	newtp->advmss = dst_metric(dst, RTAX_ADVMSS);
Linus Torvalds's avatar
Linus Torvalds committed
1600 1601
	tcp_initialize_rcv_mss(newsk);

1602
	__tcp_v4_hash(newsk, 0);
Linus Torvalds's avatar
Linus Torvalds committed
1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614
	__tcp_inherit_port(sk, newsk);

	return newsk;

exit_overflow:
	NET_INC_STATS_BH(ListenOverflows);
exit:
	NET_INC_STATS_BH(ListenDrops);
	dst_release(dst);
	return NULL;
}

1615
static struct sock *tcp_v4_hnd_req(struct sock *sk, struct sk_buff *skb)
Linus Torvalds's avatar
Linus Torvalds committed
1616 1617
{
	struct tcphdr *th = skb->h.th;
Linus Torvalds's avatar
Linus Torvalds committed
1618
	struct iphdr *iph = skb->nh.iph;
1619
	struct tcp_opt *tp = tcp_sk(sk);
Linus Torvalds's avatar
Linus Torvalds committed
1620
	struct sock *nsk;
1621
	struct open_request **prev;
Linus Torvalds's avatar
Linus Torvalds committed
1622
	/* Find possible connection requests. */
1623 1624
	struct open_request *req = tcp_v4_search_req(tp, &prev, th->source,
						     iph->saddr, iph->daddr);
Linus Torvalds's avatar
Linus Torvalds committed
1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
	if (req)
		return tcp_check_req(sk, skb, req, prev);

	nsk = __tcp_v4_lookup_established(skb->nh.iph->saddr,
					  th->source,
					  skb->nh.iph->daddr,
					  ntohs(th->dest),
					  tcp_v4_iif(skb));

	if (nsk) {
1635
		if (nsk->sk_state != TCP_TIME_WAIT) {
Linus Torvalds's avatar
Linus Torvalds committed
1636 1637 1638
			bh_lock_sock(nsk);
			return nsk;
		}
1639
		tcp_tw_put((struct tcp_tw_bucket *)nsk);
Linus Torvalds's avatar
Linus Torvalds committed
1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652
		return NULL;
	}

#ifdef CONFIG_SYN_COOKIES
	if (!th->rst && !th->syn && th->ack)
		sk = cookie_v4_check(sk, skb, &(IPCB(skb)->opt));
#endif
	return sk;
}

static int tcp_v4_checksum_init(struct sk_buff *skb)
{
	if (skb->ip_summed == CHECKSUM_HW) {
Linus Torvalds's avatar
Linus Torvalds committed
1653
		skb->ip_summed = CHECKSUM_UNNECESSARY;
1654 1655
		if (!tcp_v4_check(skb->h.th, skb->len, skb->nh.iph->saddr,
				  skb->nh.iph->daddr, skb->csum))
Linus Torvalds's avatar
Linus Torvalds committed
1656 1657
			return 0;

1658 1659
		NETDEBUG(if (net_ratelimit())
				printk(KERN_DEBUG "hw tcp v4 csum failed\n"));
Linus Torvalds's avatar
Linus Torvalds committed
1660 1661 1662
		skb->ip_summed = CHECKSUM_NONE;
	}
	if (skb->len <= 76) {
1663
		if (tcp_v4_check(skb->h.th, skb->len, skb->nh.iph->saddr,
Linus Torvalds's avatar
Linus Torvalds committed
1664 1665
				 skb->nh.iph->daddr,
				 skb_checksum(skb, 0, skb->len, 0)))
Linus Torvalds's avatar
Linus Torvalds committed
1666 1667 1668
			return -1;
		skb->ip_summed = CHECKSUM_UNNECESSARY;
	} else {
1669 1670 1671
		skb->csum = ~tcp_v4_check(skb->h.th, skb->len,
					  skb->nh.iph->saddr,
					  skb->nh.iph->daddr, 0);
Linus Torvalds's avatar
Linus Torvalds committed
1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
	}
	return 0;
}


/* The socket must have it's spinlock held when we get
 * here.
 *
 * We have a potential double-lock case here, so even when
 * doing backlog processing we use the BH locking scheme.
 * This is because we cannot sleep with the original spinlock
 * held.
 */
int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
{
1687
	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
Linus Torvalds's avatar
Linus Torvalds committed
1688 1689 1690 1691
		TCP_CHECK_TIMER(sk);
		if (tcp_rcv_established(sk, skb, skb->h.th, skb->len))
			goto reset;
		TCP_CHECK_TIMER(sk);
1692
		return 0;
Linus Torvalds's avatar
Linus Torvalds committed
1693 1694
	}

1695
	if (skb->len < (skb->h.th->doff << 2) || tcp_checksum_complete(skb))
Linus Torvalds's avatar
Linus Torvalds committed
1696 1697
		goto csum_err;

1698
	if (sk->sk_state == TCP_LISTEN) {
Linus Torvalds's avatar
Linus Torvalds committed
1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720
		struct sock *nsk = tcp_v4_hnd_req(sk, skb);
		if (!nsk)
			goto discard;

		if (nsk != sk) {
			if (tcp_child_process(sk, nsk, skb))
				goto reset;
			return 0;
		}
	}

	TCP_CHECK_TIMER(sk);
	if (tcp_rcv_state_process(sk, skb, skb->h.th, skb->len))
		goto reset;
	TCP_CHECK_TIMER(sk);
	return 0;

reset:
	tcp_v4_send_reset(skb);
discard:
	kfree_skb(skb);
	/* Be careful here. If this function gets more complicated and
1721
	 * gcc suffers from register pressure on the x86, sk (in %ebx)
Linus Torvalds's avatar
Linus Torvalds committed
1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735
	 * might be destroyed here. This current version compiles correctly,
	 * but you have been warned.
	 */
	return 0;

csum_err:
	TCP_INC_STATS_BH(TcpInErrs);
	goto discard;
}

/*
 *	From tcp_input.c
 */

Linus Torvalds's avatar
Linus Torvalds committed
1736
int tcp_v4_rcv(struct sk_buff *skb)
Linus Torvalds's avatar
Linus Torvalds committed
1737 1738 1739 1740 1741
{
	struct tcphdr *th;
	struct sock *sk;
	int ret;

1742
	if (skb->pkt_type != PACKET_HOST)
Linus Torvalds's avatar
Linus Torvalds committed
1743 1744 1745 1746 1747
		goto discard_it;

	/* Count it even if it's bad */
	TCP_INC_STATS_BH(TcpInSegs);

Linus Torvalds's avatar
Linus Torvalds committed
1748 1749 1750 1751 1752
	if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
		goto discard_it;

	th = skb->h.th;

1753
	if (th->doff < sizeof(struct tcphdr) / 4)
Linus Torvalds's avatar
Linus Torvalds committed
1754
		goto bad_packet;
1755
	if (!pskb_may_pull(skb, th->doff * 4))
Linus Torvalds's avatar
Linus Torvalds committed
1756 1757
		goto discard_it;

Linus Torvalds's avatar
Linus Torvalds committed
1758 1759 1760 1761
	/* An explanation is required here, I think.
	 * Packet length and doff are validated by header prediction,
	 * provided case of th->doff==0 is elimineted.
	 * So, we defer the checks. */
Linus Torvalds's avatar
Linus Torvalds committed
1762
	if ((skb->ip_summed != CHECKSUM_UNNECESSARY &&
Linus Torvalds's avatar
Linus Torvalds committed
1763 1764 1765
	     tcp_v4_checksum_init(skb) < 0))
		goto bad_packet;

Linus Torvalds's avatar
Linus Torvalds committed
1766
	th = skb->h.th;
Linus Torvalds's avatar
Linus Torvalds committed
1767 1768
	TCP_SKB_CB(skb)->seq = ntohl(th->seq);
	TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1769
				    skb->len - th->doff * 4);
Linus Torvalds's avatar
Linus Torvalds committed
1770
	TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1771 1772 1773
	TCP_SKB_CB(skb)->when	 = 0;
	TCP_SKB_CB(skb)->flags	 = skb->nh.iph->tos;
	TCP_SKB_CB(skb)->sacked	 = 0;
Linus Torvalds's avatar
Linus Torvalds committed
1774 1775

	sk = __tcp_v4_lookup(skb->nh.iph->saddr, th->source,
1776 1777
			     skb->nh.iph->daddr, ntohs(th->dest),
			     tcp_v4_iif(skb));
Linus Torvalds's avatar
Linus Torvalds committed
1778 1779 1780 1781 1782

	if (!sk)
		goto no_tcp_socket;

process:
1783
	if (sk->sk_state == TCP_TIME_WAIT)
Linus Torvalds's avatar
Linus Torvalds committed
1784 1785
		goto do_time_wait;

1786
	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1787 1788
		goto discard_and_relse;

1789 1790 1791
	if (sk_filter(sk, skb, 0))
		goto discard_and_relse;

Linus Torvalds's avatar
Linus Torvalds committed
1792 1793 1794 1795
	skb->dev = NULL;

	bh_lock_sock(sk);
	ret = 0;
1796
	if (!sock_owned_by_user(sk)) {
Linus Torvalds's avatar
Linus Torvalds committed
1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807
		if (!tcp_prequeue(sk, skb))
			ret = tcp_v4_do_rcv(sk, skb);
	} else
		sk_add_backlog(sk, skb);
	bh_unlock_sock(sk);

	sock_put(sk);

	return ret;

no_tcp_socket:
1808
	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1809 1810
		goto discard_it;

1811
	if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) {
Linus Torvalds's avatar
Linus Torvalds committed
1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827
bad_packet:
		TCP_INC_STATS_BH(TcpInErrs);
	} else {
		tcp_v4_send_reset(skb);
	}

discard_it:
	/* Discard frame. */
	kfree_skb(skb);
  	return 0;

discard_and_relse:
	sock_put(sk);
	goto discard_it;

do_time_wait:
1828 1829 1830 1831
	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
		tcp_tw_put((struct tcp_tw_bucket *) sk);
		goto discard_it;
	}
1832

1833
	if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) {
Linus Torvalds's avatar
Linus Torvalds committed
1834
		TCP_INC_STATS_BH(TcpInErrs);
1835 1836
		tcp_tw_put((struct tcp_tw_bucket *) sk);
		goto discard_it;
Linus Torvalds's avatar
Linus Torvalds committed
1837
	}
1838 1839 1840 1841 1842 1843 1844
	switch (tcp_timewait_state_process((struct tcp_tw_bucket *)sk,
					   skb, th, skb->len)) {
	case TCP_TW_SYN: {
		struct sock *sk2 = tcp_v4_lookup_listener(skb->nh.iph->daddr,
							  ntohs(th->dest),
							  tcp_v4_iif(skb));
		if (sk2) {
Linus Torvalds's avatar
Linus Torvalds committed
1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
			tcp_tw_deschedule((struct tcp_tw_bucket *)sk);
			tcp_tw_put((struct tcp_tw_bucket *)sk);
			sk = sk2;
			goto process;
		}
		/* Fall through to ACK */
	}
	case TCP_TW_ACK:
		tcp_v4_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
		goto no_tcp_socket;
	case TCP_TW_SUCCESS:;
	}
	goto discard_it;
}

/* With per-bucket locks this operation is not-atomic, so that
 * this version is not worse.
 */
static void __tcp_v4_rehash(struct sock *sk)
{
1867 1868
	sk->sk_prot->unhash(sk);
	sk->sk_prot->hash(sk);
Linus Torvalds's avatar
Linus Torvalds committed
1869 1870 1871 1872
}

static int tcp_v4_reselect_saddr(struct sock *sk)
{
1873
	struct inet_opt *inet = inet_sk(sk);
Linus Torvalds's avatar
Linus Torvalds committed
1874 1875
	int err;
	struct rtable *rt;
1876
	__u32 old_saddr = inet->saddr;
Linus Torvalds's avatar
Linus Torvalds committed
1877
	__u32 new_saddr;
1878
	__u32 daddr = inet->daddr;
Linus Torvalds's avatar
Linus Torvalds committed
1879

1880 1881
	if (inet->opt && inet->opt->srr)
		daddr = inet->opt->faddr;
Linus Torvalds's avatar
Linus Torvalds committed
1882 1883 1884

	/* Query new route. */
	err = ip_route_connect(&rt, daddr, 0,
1885 1886
			       RT_TOS(inet->tos) | sk->sk_localroute,
			       sk->sk_bound_dev_if,
1887
			       IPPROTO_TCP,
1888
			       inet->sport, inet->dport, sk);
Linus Torvalds's avatar
Linus Torvalds committed
1889 1890 1891 1892
	if (err)
		return err;

	__sk_dst_set(sk, &rt->u.dst);
1893
	tcp_v4_setup_caps(sk, &rt->u.dst);
1894
	tcp_sk(sk)->ext2_header_len = rt->u.dst.header_len;
Linus Torvalds's avatar
Linus Torvalds committed
1895 1896 1897 1898 1899 1900 1901

	new_saddr = rt->rt_src;

	if (new_saddr == old_saddr)
		return 0;

	if (sysctl_ip_dynaddr > 1) {
1902 1903
		printk(KERN_INFO "tcp_v4_rebuild_header(): shifting inet->"
				 "saddr from %d.%d.%d.%d to %d.%d.%d.%d\n",
1904
		       NIPQUAD(old_saddr),
Linus Torvalds's avatar
Linus Torvalds committed
1905 1906 1907
		       NIPQUAD(new_saddr));
	}

1908 1909
	inet->saddr = new_saddr;
	inet->rcv_saddr = new_saddr;
Linus Torvalds's avatar
Linus Torvalds committed
1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923

	/* XXX The only one ugly spot where we need to
	 * XXX really change the sockets identity after
	 * XXX it has entered the hashes. -DaveM
	 *
	 * Besides that, it does not check for connection
	 * uniqueness. Wait for troubles.
	 */
	__tcp_v4_rehash(sk);
	return 0;
}

int tcp_v4_rebuild_header(struct sock *sk)
{
1924
	struct inet_opt *inet = inet_sk(sk);
Linus Torvalds's avatar
Linus Torvalds committed
1925 1926 1927 1928 1929
	struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
	u32 daddr;
	int err;

	/* Route is OK, nothing to do. */
1930
	if (rt)
Linus Torvalds's avatar
Linus Torvalds committed
1931 1932 1933
		return 0;

	/* Reroute. */
1934
	daddr = inet->daddr;
1935 1936
	if (inet->opt && inet->opt->srr)
		daddr = inet->opt->faddr;
Linus Torvalds's avatar
Linus Torvalds committed
1937

1938
	{
1939
		struct flowi fl = { .oif = sk->sk_bound_dev_if,
1940
				    .nl_u = { .ip4_u =
1941 1942 1943
					      { .daddr = daddr,
						.saddr = inet->saddr,
						.tos = RT_CONN_FLAGS(sk) } },
1944 1945 1946 1947
				    .proto = IPPROTO_TCP,
				    .uli_u = { .ports =
					       { .sport = inet->sport,
						 .dport = inet->dport } } };
1948
						
1949
		err = ip_route_output_flow(&rt, &fl, sk, 0);
1950
	}
Linus Torvalds's avatar
Linus Torvalds committed
1951 1952
	if (!err) {
		__sk_dst_set(sk, &rt->u.dst);
1953
		tcp_v4_setup_caps(sk, &rt->u.dst);
1954
		tcp_sk(sk)->ext2_header_len = rt->u.dst.header_len;
Linus Torvalds's avatar
Linus Torvalds committed
1955 1956 1957 1958
		return 0;
	}

	/* Routing failed... */
1959
	sk->sk_route_caps = 0;
Linus Torvalds's avatar
Linus Torvalds committed
1960 1961

	if (!sysctl_ip_dynaddr ||
1962 1963
	    sk->sk_state != TCP_SYN_SENT ||
	    (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
Linus Torvalds's avatar
Linus Torvalds committed
1964
	    (err = tcp_v4_reselect_saddr(sk)) != 0)
1965
		sk->sk_err_soft = -err;
Linus Torvalds's avatar
Linus Torvalds committed
1966

Linus Torvalds's avatar
Linus Torvalds committed
1967 1968 1969 1970 1971 1972
	return err;
}

static void v4_addr2sockaddr(struct sock *sk, struct sockaddr * uaddr)
{
	struct sockaddr_in *sin = (struct sockaddr_in *) uaddr;
1973
	struct inet_opt *inet = inet_sk(sk);
Linus Torvalds's avatar
Linus Torvalds committed
1974 1975

	sin->sin_family		= AF_INET;
1976 1977
	sin->sin_addr.s_addr	= inet->daddr;
	sin->sin_port		= inet->dport;
Linus Torvalds's avatar
Linus Torvalds committed
1978 1979 1980 1981 1982 1983 1984 1985 1986 1987
}

/* VJ's idea. Save last timestamp seen from this destination
 * and hold it at least for normal timewait interval to use for duplicate
 * segment detection in subsequent connections, before they enter synchronized
 * state.
 */

int tcp_v4_remember_stamp(struct sock *sk)
{
1988
	struct inet_opt *inet = inet_sk(sk);
1989
	struct tcp_opt *tp = tcp_sk(sk);
1990
	struct rtable *rt = (struct rtable *)__sk_dst_get(sk);
Linus Torvalds's avatar
Linus Torvalds committed
1991 1992 1993
	struct inet_peer *peer = NULL;
	int release_it = 0;

1994
	if (!rt || rt->rt_dst != inet->daddr) {
1995
		peer = inet_getpeer(inet->daddr, 1);
Linus Torvalds's avatar
Linus Torvalds committed
1996 1997
		release_it = 1;
	} else {
1998
		if (!rt->peer)
Linus Torvalds's avatar
Linus Torvalds committed
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
			rt_bind_peer(rt, 1);
		peer = rt->peer;
	}

	if (peer) {
		if ((s32)(peer->tcp_ts - tp->ts_recent) <= 0 ||
		    (peer->tcp_ts_stamp + TCP_PAWS_MSL < xtime.tv_sec &&
		     peer->tcp_ts_stamp <= tp->ts_recent_stamp)) {
			peer->tcp_ts_stamp = tp->ts_recent_stamp;
			peer->tcp_ts = tp->ts_recent;
		}
		if (release_it)
			inet_putpeer(peer);
		return 1;
	}

	return 0;
}

int tcp_v4_tw_remember_stamp(struct tcp_tw_bucket *tw)
{
	struct inet_peer *peer = NULL;

2022
	peer = inet_getpeer(tw->tw_daddr, 1);
Linus Torvalds's avatar
Linus Torvalds committed
2023 2024

	if (peer) {
2025
		if ((s32)(peer->tcp_ts - tw->tw_ts_recent) <= 0 ||
Linus Torvalds's avatar
Linus Torvalds committed
2026
		    (peer->tcp_ts_stamp + TCP_PAWS_MSL < xtime.tv_sec &&
2027 2028 2029
		     peer->tcp_ts_stamp <= tw->tw_ts_recent_stamp)) {
			peer->tcp_ts_stamp = tw->tw_ts_recent_stamp;
			peer->tcp_ts = tw->tw_ts_recent;
Linus Torvalds's avatar
Linus Torvalds committed
2030 2031 2032 2033 2034 2035 2036 2037 2038
		}
		inet_putpeer(peer);
		return 1;
	}

	return 0;
}

struct tcp_func ipv4_specific = {
2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049
	.queue_xmit	=	ip_queue_xmit,
	.send_check	=	tcp_v4_send_check,
	.rebuild_header	=	tcp_v4_rebuild_header,
	.conn_request	=	tcp_v4_conn_request,
	.syn_recv_sock	=	tcp_v4_syn_recv_sock,
	.remember_stamp	=	tcp_v4_remember_stamp,
	.net_header_len	=	sizeof(struct iphdr),
	.setsockopt	=	ip_setsockopt,
	.getsockopt	=	ip_getsockopt,
	.addr2sockaddr	=	v4_addr2sockaddr,
	.sockaddr_len	=	sizeof(struct sockaddr_in),
Linus Torvalds's avatar
Linus Torvalds committed
2050 2051 2052 2053 2054 2055 2056
};

/* NOTE: A lot of things set to zero explicitly by call to
 *       sk_alloc() so need not be done here.
 */
static int tcp_v4_init_sock(struct sock *sk)
{
2057
	struct tcp_opt *tp = tcp_sk(sk);
Linus Torvalds's avatar
Linus Torvalds committed
2058 2059 2060 2061 2062 2063 2064

	skb_queue_head_init(&tp->out_of_order_queue);
	tcp_init_xmit_timers(sk);
	tcp_prequeue_init(tp);

	tp->rto  = TCP_TIMEOUT_INIT;
	tp->mdev = TCP_TIMEOUT_INIT;
2065

Linus Torvalds's avatar
Linus Torvalds committed
2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081
	/* So many TCP implementations out there (incorrectly) count the
	 * initial SYN frame in their delayed-ACK and congestion control
	 * algorithms that we must have the following bandaid to talk
	 * efficiently to them.  -DaveM
	 */
	tp->snd_cwnd = 2;

	/* See draft-stevens-tcpca-spec-01 for discussion of the
	 * initialization of these values.
	 */
	tp->snd_ssthresh = 0x7fffffff;	/* Infinity */
	tp->snd_cwnd_clamp = ~0;
	tp->mss_cache = 536;

	tp->reordering = sysctl_tcp_reordering;

2082
	sk->sk_state = TCP_CLOSE;
Linus Torvalds's avatar
Linus Torvalds committed
2083

2084
	sk->sk_write_space = sk_stream_write_space;
2085
	sk->sk_use_write_queue = 1;
Linus Torvalds's avatar
Linus Torvalds committed
2086

2087
	tp->af_specific = &ipv4_specific;
Linus Torvalds's avatar
Linus Torvalds committed
2088

2089 2090
	sk->sk_sndbuf = tcp_prot.sysctl_wmem[1];
	sk->sk_rcvbuf = tcp_prot.sysctl_rmem[1];
Linus Torvalds's avatar
Linus Torvalds committed
2091

2092
	atomic_inc(&tcp_prot.sockets_allocated);
Linus Torvalds's avatar
Linus Torvalds committed
2093 2094 2095 2096 2097 2098

	return 0;
}

static int tcp_v4_destroy_sock(struct sock *sk)
{
2099
	struct tcp_opt *tp = tcp_sk(sk);
Linus Torvalds's avatar
Linus Torvalds committed
2100 2101 2102 2103 2104 2105

	tcp_clear_xmit_timers(sk);

	/* Cleanup up the write buffer. */
  	tcp_writequeue_purge(sk);

Linus Torvalds's avatar
Linus Torvalds committed
2106
	/* Cleans up our, hopefully empty, out_of_order_queue. */
Linus Torvalds's avatar
Linus Torvalds committed
2107 2108 2109 2110 2111 2112
  	__skb_queue_purge(&tp->out_of_order_queue);

	/* Clean prequeue, it must be empty really */
	__skb_queue_purge(&tp->ucopy.prequeue);

	/* Clean up a referenced TCP bind bucket. */
2113
	if (tp->bind_hash)
Linus Torvalds's avatar
Linus Torvalds committed
2114 2115
		tcp_put_port(sk);

2116
	atomic_dec(&tcp_prot.sockets_allocated);
Linus Torvalds's avatar
Linus Torvalds committed
2117 2118 2119 2120

	return 0;
}

2121
#ifdef CONFIG_PROC_FS
Linus Torvalds's avatar
Linus Torvalds committed
2122
/* Proc filesystem TCP sock list dumping. */
2123

2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135
static inline struct tcp_tw_bucket *tw_head(struct hlist_head *head)
{
	return hlist_empty(head) ? NULL :
		list_entry(head->first, struct tcp_tw_bucket, tw_node);
}

static inline struct tcp_tw_bucket *tw_next(struct tcp_tw_bucket *tw)
{
	return tw->tw_node.next ?
		hlist_entry(tw->tw_node.next, typeof(*tw), tw_node) : NULL;
}

2136 2137 2138
static void *listening_get_next(struct seq_file *seq, void *cur)
{
	struct tcp_opt *tp;
2139
	struct hlist_node *node;
2140 2141 2142
	struct sock *sk = cur;
	struct tcp_iter_state* st = seq->private;

2143 2144 2145 2146 2147 2148
	if (!sk) {
		st->bucket = 0;
		sk = sk_head(&tcp_listening_hash[0]);
		goto get_sk;
	}

2149 2150
	++st->num;

2151 2152 2153 2154 2155 2156 2157
	if (st->state == TCP_SEQ_STATE_OPENREQ) {
		struct open_request *req = cur;

	       	tp = tcp_sk(st->syn_wait_sk);
		req = req->dl_next;
		while (1) {
			while (req) {
2158
				if (req->class->family == st->family) {
2159 2160 2161 2162 2163 2164 2165 2166 2167 2168
					cur = req;
					goto out;
				}
				req = req->dl_next;
			}
			if (++st->sbucket >= TCP_SYNQ_HSIZE)
				break;
get_req:
			req = tp->listen_opt->syn_table[st->sbucket];
		}
2169
		sk	  = sk_next(st->syn_wait_sk);
2170 2171 2172
		st->state = TCP_SEQ_STATE_LISTENING;
		read_unlock_bh(&tp->syn_wait_lock);
	} else
2173
		sk = sk_next(sk);
2174
get_sk:
2175
	sk_for_each_from(sk, node) {
2176
		if (sk->sk_family == st->family) {
2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191
			cur = sk;
			goto out;
		}
	       	tp = tcp_sk(sk);
		read_lock_bh(&tp->syn_wait_lock);
		if (tp->listen_opt && tp->listen_opt->qlen) {
			st->uid		= sock_i_uid(sk);
			st->syn_wait_sk = sk;
			st->state	= TCP_SEQ_STATE_OPENREQ;
			st->sbucket	= 0;
			goto get_req;
		}
		read_unlock_bh(&tp->syn_wait_lock);
	}
	if (++st->bucket < TCP_LHTABLE_SIZE) {
2192
		sk = sk_head(&tcp_listening_hash[st->bucket]);
2193 2194 2195 2196 2197 2198 2199 2200 2201
		goto get_sk;
	}
	cur = NULL;
out:
	return cur;
}

static void *listening_get_idx(struct seq_file *seq, loff_t *pos)
{
2202
	void *rc = listening_get_next(seq, NULL);
2203

2204 2205 2206 2207 2208
	while (rc && *pos) {
		rc = listening_get_next(seq, rc);
		--*pos;
	}
	return rc;
2209 2210 2211 2212 2213 2214 2215 2216 2217
}

static void *established_get_first(struct seq_file *seq)
{
	struct tcp_iter_state* st = seq->private;
	void *rc = NULL;

	for (st->bucket = 0; st->bucket < tcp_ehash_size; ++st->bucket) {
		struct sock *sk;
2218
		struct hlist_node *node;
2219 2220 2221
		struct tcp_tw_bucket *tw;
	       
		read_lock(&tcp_ehash[st->bucket].lock);
2222 2223
		sk_for_each(sk, node, &tcp_ehash[st->bucket].chain) {
			if (sk->sk_family != st->family) {
2224
				continue;
2225
			}
2226 2227 2228 2229
			rc = sk;
			goto out;
		}
		st->state = TCP_SEQ_STATE_TIME_WAIT;
2230 2231 2232
		tw_for_each(tw, node,
			    &tcp_ehash[st->bucket + tcp_ehash_size].chain) {
			if (tw->tw_family != st->family) {
2233
				continue;
2234
			}
2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248
			rc = tw;
			goto out;
		}
		read_unlock(&tcp_ehash[st->bucket].lock);
		st->state = TCP_SEQ_STATE_ESTABLISHED;
	}
out:
	return rc;
}

static void *established_get_next(struct seq_file *seq, void *cur)
{
	struct sock *sk = cur;
	struct tcp_tw_bucket *tw;
2249
	struct hlist_node *node;
2250 2251
	struct tcp_iter_state* st = seq->private;

2252 2253
	++st->num;

2254 2255
	if (st->state == TCP_SEQ_STATE_TIME_WAIT) {
		tw = cur;
2256
		tw = tw_next(tw);
2257
get_tw:
2258
		while (tw && tw->tw_family != st->family) {
2259
			tw = tw_next(tw);
2260 2261 2262 2263 2264 2265 2266 2267 2268
		}
		if (tw) {
			cur = tw;
			goto out;
		}
		read_unlock(&tcp_ehash[st->bucket].lock);
		st->state = TCP_SEQ_STATE_ESTABLISHED;
		if (++st->bucket < tcp_ehash_size) {
			read_lock(&tcp_ehash[st->bucket].lock);
2269
			sk = sk_head(&tcp_ehash[st->bucket].chain);
2270 2271 2272 2273 2274
		} else {
			cur = NULL;
			goto out;
		}
	} else
2275
		sk = sk_next(sk);
2276

2277 2278 2279
	sk_for_each_from(sk, node) {
		if (sk->sk_family == st->family)
			goto found;
2280
	}
2281 2282 2283 2284 2285

	st->state = TCP_SEQ_STATE_TIME_WAIT;
	tw = tw_head(&tcp_ehash[st->bucket + tcp_ehash_size].chain);
	goto get_tw;
found:
2286 2287 2288 2289 2290 2291 2292 2293 2294
	cur = sk;
out:
	return cur;
}

static void *established_get_idx(struct seq_file *seq, loff_t pos)
{
	void *rc = established_get_first(seq);

2295 2296 2297 2298 2299
	while (rc && pos) {
		rc = established_get_next(seq, rc);
		--pos;
	}		
	return rc;
2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322
}

static void *tcp_get_idx(struct seq_file *seq, loff_t pos)
{
	void *rc;
	struct tcp_iter_state* st = seq->private;

	tcp_listen_lock();
	st->state = TCP_SEQ_STATE_LISTENING;
	rc	  = listening_get_idx(seq, &pos);

	if (!rc) {
		tcp_listen_unlock();
		local_bh_disable();
		st->state = TCP_SEQ_STATE_ESTABLISHED;
		rc	  = established_get_idx(seq, pos);
	}

	return rc;
}

static void *tcp_seq_start(struct seq_file *seq, loff_t *pos)
{
2323
	struct tcp_iter_state* st = seq->private;
2324
	st->state = TCP_SEQ_STATE_LISTENING;
2325
	st->num = 0;
2326
	return *pos ? tcp_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2327 2328 2329 2330 2331 2332 2333
}

static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
	void *rc = NULL;
	struct tcp_iter_state* st;

2334
	if (v == SEQ_START_TOKEN) {
2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371
		rc = tcp_get_idx(seq, 0);
		goto out;
	}
	st = seq->private;

	switch (st->state) {
	case TCP_SEQ_STATE_OPENREQ:
	case TCP_SEQ_STATE_LISTENING:
		rc = listening_get_next(seq, v);
		if (!rc) {
			tcp_listen_unlock();
			local_bh_disable();
			st->state = TCP_SEQ_STATE_ESTABLISHED;
			rc	  = established_get_first(seq);
		}
		break;
	case TCP_SEQ_STATE_ESTABLISHED:
	case TCP_SEQ_STATE_TIME_WAIT:
		rc = established_get_next(seq, v);
		break;
	}
out:
	++*pos;
	return rc;
}

static void tcp_seq_stop(struct seq_file *seq, void *v)
{
	struct tcp_iter_state* st = seq->private;

	switch (st->state) {
	case TCP_SEQ_STATE_OPENREQ:
		if (v) {
			struct tcp_opt *tp = tcp_sk(st->syn_wait_sk);
			read_unlock_bh(&tp->syn_wait_lock);
		}
	case TCP_SEQ_STATE_LISTENING:
2372
		if (v != SEQ_START_TOKEN)
2373
			tcp_listen_unlock();
2374 2375 2376 2377 2378 2379 2380 2381 2382 2383
		break;
	case TCP_SEQ_STATE_TIME_WAIT:
	case TCP_SEQ_STATE_ESTABLISHED:
		if (v)
			read_unlock(&tcp_ehash[st->bucket].lock);
		local_bh_enable();
		break;
	}
}

2384 2385 2386 2387
static int tcp_seq_open(struct inode *inode, struct file *file)
{
	struct tcp_seq_afinfo *afinfo = PDE(inode)->data;
	struct seq_file *seq;
2388 2389
	struct tcp_iter_state *s;
	int rc;
2390

2391 2392 2393 2394
	if (unlikely(afinfo == NULL))
		return -EINVAL;

	s = kmalloc(sizeof(*s), GFP_KERNEL);
2395
	if (!s)
2396
		return -ENOMEM;
2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428
	memset(s, 0, sizeof(*s));
	s->family		= afinfo->family;
	s->seq_ops.start	= tcp_seq_start;
	s->seq_ops.next		= tcp_seq_next;
	s->seq_ops.show		= afinfo->seq_show;
	s->seq_ops.stop		= tcp_seq_stop;

	rc = seq_open(file, &s->seq_ops);
	if (rc)
		goto out_kfree;
	seq	     = file->private_data;
	seq->private = s;
out:
	return rc;
out_kfree:
	kfree(s);
	goto out;
}

int tcp_proc_register(struct tcp_seq_afinfo *afinfo)
{
	int rc = 0;
	struct proc_dir_entry *p;

	if (!afinfo)
		return -EINVAL;
	afinfo->seq_fops->owner		= afinfo->owner;
	afinfo->seq_fops->open		= tcp_seq_open;
	afinfo->seq_fops->read		= seq_read;
	afinfo->seq_fops->llseek	= seq_lseek;
	afinfo->seq_fops->release	= seq_release_private;
	
2429 2430
	p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops);
	if (p)
2431
		p->data = afinfo;
2432
	else
2433 2434 2435 2436 2437 2438 2439 2440
		rc = -ENOMEM;
	return rc;
}

void tcp_proc_unregister(struct tcp_seq_afinfo *afinfo)
{
	if (!afinfo)
		return;
2441
	proc_net_remove(afinfo->name);
2442 2443 2444 2445 2446
	memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops)); 
}

static void get_openreq4(struct sock *sk, struct open_request *req,
			 char *tmpbuf, int i, int uid)
Linus Torvalds's avatar
Linus Torvalds committed
2447 2448 2449 2450
{
	int ttd = req->expires - jiffies;

	sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
2451
		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %u %d %p",
Linus Torvalds's avatar
Linus Torvalds committed
2452 2453
		i,
		req->af.v4_req.loc_addr,
2454
		ntohs(inet_sk(sk)->sport),
Linus Torvalds's avatar
Linus Torvalds committed
2455 2456 2457
		req->af.v4_req.rmt_addr,
		ntohs(req->rmt_port),
		TCP_SYN_RECV,
2458 2459
		0, 0, /* could print option size, but that is af dependent. */
		1,    /* timers active (only the expire timer) */
2460
		jiffies_to_clock_t(ttd),
Linus Torvalds's avatar
Linus Torvalds committed
2461 2462
		req->retrans,
		uid,
2463
		0,  /* non standard timer */
Linus Torvalds's avatar
Linus Torvalds committed
2464
		0, /* open_requests have no inode */
2465
		atomic_read(&sk->sk_refcnt),
2466
		req);
Linus Torvalds's avatar
Linus Torvalds committed
2467 2468
}

2469
static void get_tcp4_sock(struct sock *sp, char *tmpbuf, int i)
Linus Torvalds's avatar
Linus Torvalds committed
2470 2471 2472
{
	int timer_active;
	unsigned long timer_expires;
2473
	struct tcp_opt *tp = tcp_sk(sp);
2474
	struct inet_opt *inet = inet_sk(sp);
2475 2476 2477 2478
	unsigned int dest = inet->daddr;
	unsigned int src = inet->rcv_saddr;
	__u16 destp = ntohs(inet->dport);
	__u16 srcp = ntohs(inet->sport);
Linus Torvalds's avatar
Linus Torvalds committed
2479 2480 2481 2482 2483 2484 2485

	if (tp->pending == TCP_TIME_RETRANS) {
		timer_active	= 1;
		timer_expires	= tp->timeout;
	} else if (tp->pending == TCP_TIME_PROBE0) {
		timer_active	= 4;
		timer_expires	= tp->timeout;
2486
	} else if (timer_pending(&sp->sk_timer)) {
Linus Torvalds's avatar
Linus Torvalds committed
2487
		timer_active	= 2;
2488
		timer_expires	= sp->sk_timer.expires;
Linus Torvalds's avatar
Linus Torvalds committed
2489 2490 2491 2492 2493
	} else {
		timer_active	= 0;
		timer_expires = jiffies;
	}

2494 2495
	sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX "
			"%08X %5d %8d %lu %d %p %u %u %u %u %d",
2496
		i, src, srcp, dest, destp, sp->sk_state,
2497
		tp->write_seq - tp->snd_una, tp->rcv_nxt - tp->copied_seq,
2498 2499
		timer_active,
		jiffies_to_clock_t(timer_expires - jiffies),
Linus Torvalds's avatar
Linus Torvalds committed
2500 2501 2502 2503
		tp->retransmits,
		sock_i_uid(sp),
		tp->probes_out,
		sock_i_ino(sp),
2504
		atomic_read(&sp->sk_refcnt), sp,
2505 2506 2507
		tp->rto, tp->ack.ato, (tp->ack.quick << 1) | tp->ack.pingpong,
		tp->snd_cwnd,
		tp->snd_ssthresh >= 0xFFFF ? -1 : tp->snd_ssthresh);
Linus Torvalds's avatar
Linus Torvalds committed
2508 2509
}

2510
static void get_timewait4_sock(struct tcp_tw_bucket *tw, char *tmpbuf, int i)
Linus Torvalds's avatar
Linus Torvalds committed
2511 2512 2513
{
	unsigned int dest, src;
	__u16 destp, srcp;
2514
	int ttd = tw->tw_ttd - jiffies;
Linus Torvalds's avatar
Linus Torvalds committed
2515 2516 2517 2518

	if (ttd < 0)
		ttd = 0;

2519 2520 2521 2522
	dest  = tw->tw_daddr;
	src   = tw->tw_rcv_saddr;
	destp = ntohs(tw->tw_dport);
	srcp  = ntohs(tw->tw_sport);
Linus Torvalds's avatar
Linus Torvalds committed
2523 2524

	sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
2525
		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p",
2526
		i, src, srcp, dest, destp, tw->tw_substate, 0, 0,
2527
		3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
2528
		atomic_read(&tw->tw_refcnt), tw);
Linus Torvalds's avatar
Linus Torvalds committed
2529 2530 2531 2532
}

#define TMPSZ 150

2533
static int tcp4_seq_show(struct seq_file *seq, void *v)
Linus Torvalds's avatar
Linus Torvalds committed
2534
{
2535
	struct tcp_iter_state* st;
2536
	char tmpbuf[TMPSZ + 1];
Linus Torvalds's avatar
Linus Torvalds committed
2537

2538
	if (v == SEQ_START_TOKEN) {
2539 2540 2541 2542 2543
		seq_printf(seq, "%-*s\n", TMPSZ - 1,
			   "  sl  local_address rem_address   st tx_queue "
			   "rx_queue tr tm->when retrnsmt   uid  timeout "
			   "inode");
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
2544
	}
2545
	st = seq->private;
Linus Torvalds's avatar
Linus Torvalds committed
2546

2547 2548 2549
	switch (st->state) {
	case TCP_SEQ_STATE_LISTENING:
	case TCP_SEQ_STATE_ESTABLISHED:
2550
		get_tcp4_sock(v, tmpbuf, st->num);
2551 2552
		break;
	case TCP_SEQ_STATE_OPENREQ:
2553
		get_openreq4(st->syn_wait_sk, v, tmpbuf, st->num, st->uid);
2554 2555
		break;
	case TCP_SEQ_STATE_TIME_WAIT:
2556
		get_timewait4_sock(v, tmpbuf, st->num);
2557 2558 2559 2560 2561 2562
		break;
	}
	seq_printf(seq, "%-*s\n", TMPSZ - 1, tmpbuf);
out:
	return 0;
}
Linus Torvalds's avatar
Linus Torvalds committed
2563

2564 2565 2566 2567 2568 2569 2570
static struct file_operations tcp4_seq_fops;
static struct tcp_seq_afinfo tcp4_seq_afinfo = {
	.owner		= THIS_MODULE,
	.name		= "tcp",
	.family		= AF_INET,
	.seq_show	= tcp4_seq_show,
	.seq_fops	= &tcp4_seq_fops,
2571
};
Linus Torvalds's avatar
Linus Torvalds committed
2572

2573
int __init tcp4_proc_init(void)
2574
{
2575
	return tcp_proc_register(&tcp4_seq_afinfo);
2576 2577
}

2578
void tcp4_proc_exit(void)
2579
{
2580
	tcp_proc_unregister(&tcp4_seq_afinfo);
Linus Torvalds's avatar
Linus Torvalds committed
2581
}
2582
#endif /* CONFIG_PROC_FS */
Linus Torvalds's avatar
Linus Torvalds committed
2583 2584

struct proto tcp_prot = {
2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604
	.name			= "TCP",
	.close			= tcp_close,
	.connect		= tcp_v4_connect,
	.disconnect		= tcp_disconnect,
	.accept			= tcp_accept,
	.ioctl			= tcp_ioctl,
	.init			= tcp_v4_init_sock,
	.destroy		= tcp_v4_destroy_sock,
	.shutdown		= tcp_shutdown,
	.setsockopt		= tcp_setsockopt,
	.getsockopt		= tcp_getsockopt,
	.sendmsg		= tcp_sendmsg,
	.recvmsg		= tcp_recvmsg,
	.backlog_rcv		= tcp_v4_do_rcv,
	.hash			= tcp_v4_hash,
	.unhash			= tcp_unhash,
	.get_port		= tcp_v4_get_port,
	.enter_memory_pressure	= tcp_enter_memory_pressure,
	.sysctl_wmem		= { 4 * 1024, 16 * 1024, 128 * 1024 },
	.sysctl_rmem		= { 4 * 1024, 87380, 87380 * 2 },
Linus Torvalds's avatar
Linus Torvalds committed
2605 2606 2607 2608 2609 2610
};



void __init tcp_v4_init(struct net_proto_family *ops)
{
James Morris's avatar
James Morris committed
2611
	int err = sock_create_kern(PF_INET, SOCK_RAW, IPPROTO_TCP, &tcp_socket);
Linus Torvalds's avatar
Linus Torvalds committed
2612
	if (err < 0)
Linus Torvalds's avatar
Linus Torvalds committed
2613
		panic("Failed to create the TCP control socket.\n");
2614
	tcp_socket->sk->sk_allocation   = GFP_ATOMIC;
2615
	inet_sk(tcp_socket->sk)->uc_ttl = -1;
Linus Torvalds's avatar
Linus Torvalds committed
2616 2617 2618 2619 2620

	/* Unhash it so that IP input processing does not even
	 * see it, we do not wish this socket to see incoming
	 * packets.
	 */
2621
	tcp_socket->sk->sk_prot->unhash(tcp_socket->sk);
Linus Torvalds's avatar
Linus Torvalds committed
2622
}
2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651

EXPORT_SYMBOL(ipv4_specific);
EXPORT_SYMBOL(tcp_bind_hash);
EXPORT_SYMBOL(tcp_bucket_create);
EXPORT_SYMBOL(tcp_hashinfo);
EXPORT_SYMBOL(tcp_inherit_port);
EXPORT_SYMBOL(tcp_listen_wlock);
EXPORT_SYMBOL(tcp_port_rover);
EXPORT_SYMBOL(tcp_prot);
EXPORT_SYMBOL(tcp_put_port);
EXPORT_SYMBOL(tcp_unhash);
EXPORT_SYMBOL(tcp_v4_conn_request);
EXPORT_SYMBOL(tcp_v4_connect);
EXPORT_SYMBOL(tcp_v4_do_rcv);
EXPORT_SYMBOL(tcp_v4_lookup_listener);
EXPORT_SYMBOL(tcp_v4_rebuild_header);
EXPORT_SYMBOL(tcp_v4_remember_stamp);
EXPORT_SYMBOL(tcp_v4_send_check);
EXPORT_SYMBOL(tcp_v4_syn_recv_sock);

#ifdef CONFIG_PROC_FS
EXPORT_SYMBOL(tcp_proc_register);
EXPORT_SYMBOL(tcp_proc_unregister);
#endif
#ifdef CONFIG_SYSCTL
EXPORT_SYMBOL(sysctl_local_port_range);
EXPORT_SYMBOL(sysctl_max_syn_backlog);
EXPORT_SYMBOL(sysctl_tcp_low_latency);
#endif