tcp_ipv4.c 62.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
/*
 * 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).
 *
 *		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
 *					the TCP layer, missed a check for an
 *					ACK bit.
 *		Andi Kleen :		Implemented fast path mtu discovery.
 *	     				Fixed many serious bugs in the
37
 *					request_sock handling and moved
L
Linus Torvalds 已提交
38 39
 *					most of it into the af independent code.
 *					Added tail drop and some other bugfixes.
S
Stephen Hemminger 已提交
40
 *					Added new listen semantics.
L
Linus Torvalds 已提交
41 42 43 44 45 46 47 48 49 50 51 52
 *		Mike McLagan	:	Routing by source
 *	Juan Jose Ciarlante:		ip_dynaddr bits
 *		Andi Kleen:		various fixes.
 *	Vitaly E. Lavrov	:	Transparent proxy revived after year
 *					coma.
 *	Andi Kleen		:	Fix new listen.
 *	Andi Kleen		:	Fix accept error reporting.
 *	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.
 */

53
#define pr_fmt(fmt) "TCP: " fmt
L
Linus Torvalds 已提交
54

H
Herbert Xu 已提交
55
#include <linux/bottom_half.h>
L
Linus Torvalds 已提交
56 57 58 59 60 61 62 63
#include <linux/types.h>
#include <linux/fcntl.h>
#include <linux/module.h>
#include <linux/random.h>
#include <linux/cache.h>
#include <linux/jhash.h>
#include <linux/init.h>
#include <linux/times.h>
64
#include <linux/slab.h>
L
Linus Torvalds 已提交
65

66
#include <net/net_namespace.h>
L
Linus Torvalds 已提交
67
#include <net/icmp.h>
68
#include <net/inet_hashtables.h>
L
Linus Torvalds 已提交
69
#include <net/tcp.h>
70
#include <net/transp_v6.h>
L
Linus Torvalds 已提交
71 72
#include <net/ipv6.h>
#include <net/inet_common.h>
73
#include <net/timewait_sock.h>
L
Linus Torvalds 已提交
74
#include <net/xfrm.h>
75
#include <net/secure_seq.h>
76
#include <net/busy_poll.h>
L
Linus Torvalds 已提交
77 78 79 80 81 82 83

#include <linux/inet.h>
#include <linux/ipv6.h>
#include <linux/stddef.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>

H
Herbert Xu 已提交
84
#include <crypto/hash.h>
85 86
#include <linux/scatterlist.h>

87 88
int sysctl_tcp_tw_reuse __read_mostly;
int sysctl_tcp_low_latency __read_mostly;
L
Linus Torvalds 已提交
89

90
#ifdef CONFIG_TCP_MD5SIG
E
Eric Dumazet 已提交
91
static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
92
			       __be32 daddr, __be32 saddr, const struct tcphdr *th);
93 94
#endif

95
struct inet_hashinfo tcp_hashinfo;
E
Eric Dumazet 已提交
96
EXPORT_SYMBOL(tcp_hashinfo);
L
Linus Torvalds 已提交
97

98
static  __u32 tcp_v4_init_sequence(const struct sk_buff *skb)
L
Linus Torvalds 已提交
99
{
100 101
	return secure_tcp_sequence_number(ip_hdr(skb)->daddr,
					  ip_hdr(skb)->saddr,
102 103
					  tcp_hdr(skb)->dest,
					  tcp_hdr(skb)->source);
L
Linus Torvalds 已提交
104 105
}

106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122
int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp)
{
	const struct tcp_timewait_sock *tcptw = tcp_twsk(sktw);
	struct tcp_sock *tp = tcp_sk(sk);

	/* 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.
	 */
	if (tcptw->tw_ts_recent_stamp &&
123
	    (!twp || (sysctl_tcp_tw_reuse &&
124
			     get_seconds() - tcptw->tw_ts_recent_stamp > 1))) {
125 126 127 128 129 130 131 132 133 134 135 136 137
		tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2;
		if (tp->write_seq == 0)
			tp->write_seq = 1;
		tp->rx_opt.ts_recent	   = tcptw->tw_ts_recent;
		tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp;
		sock_hold(sktw);
		return 1;
	}

	return 0;
}
EXPORT_SYMBOL_GPL(tcp_twsk_unique);

L
Linus Torvalds 已提交
138 139 140
/* This will initiate an outgoing connection. */
int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
{
141
	struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
L
Linus Torvalds 已提交
142 143
	struct inet_sock *inet = inet_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
144
	__be16 orig_sport, orig_dport;
145
	__be32 daddr, nexthop;
146
	struct flowi4 *fl4;
147
	struct rtable *rt;
L
Linus Torvalds 已提交
148
	int err;
149
	struct ip_options_rcu *inet_opt;
L
Linus Torvalds 已提交
150 151 152 153 154 155 156 157

	if (addr_len < sizeof(struct sockaddr_in))
		return -EINVAL;

	if (usin->sin_family != AF_INET)
		return -EAFNOSUPPORT;

	nexthop = daddr = usin->sin_addr.s_addr;
158
	inet_opt = rcu_dereference_protected(inet->inet_opt,
159
					     lockdep_sock_is_held(sk));
160
	if (inet_opt && inet_opt->opt.srr) {
L
Linus Torvalds 已提交
161 162
		if (!daddr)
			return -EINVAL;
163
		nexthop = inet_opt->opt.faddr;
L
Linus Torvalds 已提交
164 165
	}

166 167
	orig_sport = inet->inet_sport;
	orig_dport = usin->sin_port;
168 169
	fl4 = &inet->cork.fl.u.ip4;
	rt = ip_route_connect(fl4, nexthop, inet->inet_saddr,
170 171
			      RT_CONN_FLAGS(sk), sk->sk_bound_dev_if,
			      IPPROTO_TCP,
172
			      orig_sport, orig_dport, sk);
173 174 175
	if (IS_ERR(rt)) {
		err = PTR_ERR(rt);
		if (err == -ENETUNREACH)
176
			IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
177
		return err;
178
	}
L
Linus Torvalds 已提交
179 180 181 182 183 184

	if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
		ip_rt_put(rt);
		return -ENETUNREACH;
	}

185
	if (!inet_opt || !inet_opt->opt.srr)
186
		daddr = fl4->daddr;
L
Linus Torvalds 已提交
187

E
Eric Dumazet 已提交
188
	if (!inet->inet_saddr)
189
		inet->inet_saddr = fl4->saddr;
190
	sk_rcv_saddr_set(sk, inet->inet_saddr);
L
Linus Torvalds 已提交
191

E
Eric Dumazet 已提交
192
	if (tp->rx_opt.ts_recent_stamp && inet->inet_daddr != daddr) {
L
Linus Torvalds 已提交
193 194 195
		/* Reset inherited state */
		tp->rx_opt.ts_recent	   = 0;
		tp->rx_opt.ts_recent_stamp = 0;
P
Pavel Emelyanov 已提交
196 197
		if (likely(!tp->repair))
			tp->write_seq	   = 0;
L
Linus Torvalds 已提交
198 199
	}

200
	if (tcp_death_row.sysctl_tw_recycle &&
201 202
	    !tp->rx_opt.ts_recent_stamp && fl4->daddr == daddr)
		tcp_fetch_timewait_stamp(sk, &rt->dst);
L
Linus Torvalds 已提交
203

E
Eric Dumazet 已提交
204
	inet->inet_dport = usin->sin_port;
205
	sk_daddr_set(sk, daddr);
L
Linus Torvalds 已提交
206

207
	inet_csk(sk)->icsk_ext_hdr_len = 0;
208 209
	if (inet_opt)
		inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
L
Linus Torvalds 已提交
210

211
	tp->rx_opt.mss_clamp = TCP_MSS_DEFAULT;
L
Linus Torvalds 已提交
212 213 214 215 216 217 218

	/* 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
	 * complete initialization after this.
	 */
	tcp_set_state(sk, TCP_SYN_SENT);
219
	err = inet_hash_connect(&tcp_death_row, sk);
L
Linus Torvalds 已提交
220 221 222
	if (err)
		goto failure;

223
	sk_set_txhash(sk);
224

225
	rt = ip_route_newports(fl4, rt, orig_sport, orig_dport,
226 227 228 229
			       inet->inet_sport, inet->inet_dport, sk);
	if (IS_ERR(rt)) {
		err = PTR_ERR(rt);
		rt = NULL;
L
Linus Torvalds 已提交
230
		goto failure;
231
	}
L
Linus Torvalds 已提交
232
	/* OK, now commit destination to socket.  */
233
	sk->sk_gso_type = SKB_GSO_TCPV4;
234
	sk_setup_caps(sk, &rt->dst);
L
Linus Torvalds 已提交
235

P
Pavel Emelyanov 已提交
236
	if (!tp->write_seq && likely(!tp->repair))
E
Eric Dumazet 已提交
237 238 239
		tp->write_seq = secure_tcp_sequence_number(inet->inet_saddr,
							   inet->inet_daddr,
							   inet->inet_sport,
L
Linus Torvalds 已提交
240 241
							   usin->sin_port);

E
Eric Dumazet 已提交
242
	inet->inet_id = tp->write_seq ^ jiffies;
L
Linus Torvalds 已提交
243

A
Andrey Vagin 已提交
244
	err = tcp_connect(sk);
P
Pavel Emelyanov 已提交
245

L
Linus Torvalds 已提交
246 247 248 249 250 251 252
	rt = NULL;
	if (err)
		goto failure;

	return 0;

failure:
253 254 255 256
	/*
	 * This unhashes the socket and releases the local port,
	 * if necessary.
	 */
L
Linus Torvalds 已提交
257 258 259
	tcp_set_state(sk, TCP_CLOSE);
	ip_rt_put(rt);
	sk->sk_route_caps = 0;
E
Eric Dumazet 已提交
260
	inet->inet_dport = 0;
L
Linus Torvalds 已提交
261 262
	return err;
}
E
Eric Dumazet 已提交
263
EXPORT_SYMBOL(tcp_v4_connect);
L
Linus Torvalds 已提交
264 265

/*
266 267 268
 * This routine reacts to ICMP_FRAG_NEEDED mtu indications as defined in RFC1191.
 * It can be called through tcp_release_cb() if socket was owned by user
 * at the time tcp_v4_err() was called to handle ICMP message.
L
Linus Torvalds 已提交
269
 */
270
void tcp_v4_mtu_reduced(struct sock *sk)
L
Linus Torvalds 已提交
271 272 273
{
	struct dst_entry *dst;
	struct inet_sock *inet = inet_sk(sk);
274
	u32 mtu = tcp_sk(sk)->mtu_info;
L
Linus Torvalds 已提交
275

276 277
	dst = inet_csk_update_pmtu(sk, mtu);
	if (!dst)
L
Linus Torvalds 已提交
278 279 280 281 282 283 284 285 286 287 288
		return;

	/* Something is about to be wrong... Remember soft error
	 * for the case, if this connection will not able to recover.
	 */
	if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst))
		sk->sk_err_soft = EMSGSIZE;

	mtu = dst_mtu(dst);

	if (inet->pmtudisc != IP_PMTUDISC_DONT &&
289
	    ip_sk_accept_pmtu(sk) &&
290
	    inet_csk(sk)->icsk_pmtu_cookie > mtu) {
L
Linus Torvalds 已提交
291 292 293 294 295 296 297 298 299 300
		tcp_sync_mss(sk, mtu);

		/* Resend the TCP packet because it's
		 * 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 */
}
301
EXPORT_SYMBOL(tcp_v4_mtu_reduced);
L
Linus Torvalds 已提交
302

303 304 305 306
static void do_redirect(struct sk_buff *skb, struct sock *sk)
{
	struct dst_entry *dst = __sk_dst_check(sk, 0);

307
	if (dst)
308
		dst->ops->redirect(dst, sk, skb);
309 310
}

311 312

/* handle ICMP messages on TCP_NEW_SYN_RECV request sockets */
313
void tcp_req_err(struct sock *sk, u32 seq, bool abort)
314 315 316 317 318 319 320 321
{
	struct request_sock *req = inet_reqsk(sk);
	struct net *net = sock_net(sk);

	/* ICMPs are not backlogged, hence we cannot get
	 * an established socket here.
	 */
	if (seq != tcp_rsk(req)->snt_isn) {
322
		__NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
323
	} else if (abort) {
324 325 326 327 328 329
		/*
		 * 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
		 * errors returned from accept().
		 */
F
Fan Du 已提交
330
		inet_csk_reqsk_queue_drop(req->rsk_listener, req);
331
		tcp_listendrop(req->rsk_listener);
332
	}
333
	reqsk_put(req);
334 335 336
}
EXPORT_SYMBOL(tcp_req_err);

L
Linus Torvalds 已提交
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352
/*
 * 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.
 *
 */

353
void tcp_v4_err(struct sk_buff *icmp_skb, u32 info)
L
Linus Torvalds 已提交
354
{
355
	const struct iphdr *iph = (const struct iphdr *)icmp_skb->data;
356
	struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2));
357
	struct inet_connection_sock *icsk;
L
Linus Torvalds 已提交
358 359
	struct tcp_sock *tp;
	struct inet_sock *inet;
360 361
	const int type = icmp_hdr(icmp_skb)->type;
	const int code = icmp_hdr(icmp_skb)->code;
L
Linus Torvalds 已提交
362
	struct sock *sk;
363
	struct sk_buff *skb;
364 365
	struct request_sock *fastopen;
	__u32 seq, snd_una;
366
	__u32 remaining;
L
Linus Torvalds 已提交
367
	int err;
368
	struct net *net = dev_net(icmp_skb->dev);
L
Linus Torvalds 已提交
369

370 371 372
	sk = __inet_lookup_established(net, &tcp_hashinfo, iph->daddr,
				       th->dest, iph->saddr, ntohs(th->source),
				       inet_iif(icmp_skb));
L
Linus Torvalds 已提交
373
	if (!sk) {
E
Eric Dumazet 已提交
374
		__ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
L
Linus Torvalds 已提交
375 376 377
		return;
	}
	if (sk->sk_state == TCP_TIME_WAIT) {
378
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
379 380
		return;
	}
381 382
	seq = ntohl(th->seq);
	if (sk->sk_state == TCP_NEW_SYN_RECV)
383 384 385 386 387 388
		return tcp_req_err(sk, seq,
				  type == ICMP_PARAMETERPROB ||
				  type == ICMP_TIME_EXCEEDED ||
				  (type == ICMP_DEST_UNREACH &&
				   (code == ICMP_NET_UNREACH ||
				    code == ICMP_HOST_UNREACH)));
L
Linus Torvalds 已提交
389 390 391 392

	bh_lock_sock(sk);
	/* If too many ICMPs get dropped on busy
	 * servers this needs to be solved differently.
393 394
	 * We do take care of PMTU discovery (RFC1191) special case :
	 * we can receive locally generated ICMP messages while socket is held.
L
Linus Torvalds 已提交
395
	 */
396 397
	if (sock_owned_by_user(sk)) {
		if (!(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED))
398
			__NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
399
	}
L
Linus Torvalds 已提交
400 401 402
	if (sk->sk_state == TCP_CLOSE)
		goto out;

403
	if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
404
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
405 406 407
		goto out;
	}

408
	icsk = inet_csk(sk);
L
Linus Torvalds 已提交
409
	tp = tcp_sk(sk);
410 411 412
	/* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
	fastopen = tp->fastopen_rsk;
	snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
L
Linus Torvalds 已提交
413
	if (sk->sk_state != TCP_LISTEN &&
414
	    !between(seq, snd_una, tp->snd_nxt)) {
415
		__NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
416 417 418 419
		goto out;
	}

	switch (type) {
420 421 422
	case ICMP_REDIRECT:
		do_redirect(icmp_skb, sk);
		goto out;
L
Linus Torvalds 已提交
423 424 425 426 427 428 429 430 431 432 433
	case ICMP_SOURCE_QUENCH:
		/* Just silently ignore these. */
		goto out;
	case ICMP_PARAMETERPROB:
		err = EPROTO;
		break;
	case ICMP_DEST_UNREACH:
		if (code > NR_ICMP_UNREACH)
			goto out;

		if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */
434 435 436 437 438 439 440
			/* 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).
			 */
			if (sk->sk_state == TCP_LISTEN)
				goto out;

441
			tp->mtu_info = info;
442
			if (!sock_owned_by_user(sk)) {
443
				tcp_v4_mtu_reduced(sk);
444 445 446 447
			} else {
				if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, &tp->tsq_flags))
					sock_hold(sk);
			}
L
Linus Torvalds 已提交
448 449 450 451
			goto out;
		}

		err = icmp_err_convert[code].errno;
452 453 454 455 456
		/* check if icmp_skb allows revert of backoff
		 * (see draft-zimmermann-tcp-lcd) */
		if (code != ICMP_NET_UNREACH && code != ICMP_HOST_UNREACH)
			break;
		if (seq != tp->snd_una  || !icsk->icsk_retransmits ||
457
		    !icsk->icsk_backoff || fastopen)
458 459
			break;

460 461 462
		if (sock_owned_by_user(sk))
			break;

463
		icsk->icsk_backoff--;
464 465 466
		icsk->icsk_rto = tp->srtt_us ? __tcp_set_rto(tp) :
					       TCP_TIMEOUT_INIT;
		icsk->icsk_rto = inet_csk_rto_backoff(icsk, TCP_RTO_MAX);
467 468 469 470

		skb = tcp_write_queue_head(sk);
		BUG_ON(!skb);

E
Eric Dumazet 已提交
471 472 473
		remaining = icsk->icsk_rto -
			    min(icsk->icsk_rto,
				tcp_time_stamp - tcp_skb_timestamp(skb));
474 475 476 477 478 479 480 481 482 483

		if (remaining) {
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
						  remaining, TCP_RTO_MAX);
		} else {
			/* RTO revert clocked out retransmission.
			 * Will retransmit now */
			tcp_retransmit_timer(sk);
		}

L
Linus Torvalds 已提交
484 485 486 487 488 489 490 491 492 493
		break;
	case ICMP_TIME_EXCEEDED:
		err = EHOSTUNREACH;
		break;
	default:
		goto out;
	}

	switch (sk->sk_state) {
	case TCP_SYN_SENT:
494 495 496 497
	case TCP_SYN_RECV:
		/* Only in fast or simultaneous open. If a fast open socket is
		 * is already accepted it is treated as a connected one below.
		 */
498
		if (fastopen && !fastopen->sk)
499 500
			break;

L
Linus Torvalds 已提交
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
		if (!sock_owned_by_user(sk)) {
			sk->sk_err = err;

			sk->sk_error_report(sk);

			tcp_done(sk);
		} else {
			sk->sk_err_soft = err;
		}
		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)
	 */

	inet = inet_sk(sk);
	if (!sock_owned_by_user(sk) && inet->recverr) {
		sk->sk_err = err;
		sk->sk_error_report(sk);
	} else	{ /* Only an error on timeout */
		sk->sk_err_soft = err;
	}

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

542
void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr)
L
Linus Torvalds 已提交
543
{
544
	struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
545

546
	if (skb->ip_summed == CHECKSUM_PARTIAL) {
547
		th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0);
548
		skb->csum_start = skb_transport_header(skb) - skb->head;
A
Al Viro 已提交
549
		skb->csum_offset = offsetof(struct tcphdr, check);
L
Linus Torvalds 已提交
550
	} else {
551
		th->check = tcp_v4_check(skb->len, saddr, daddr,
552
					 csum_partial(th,
L
Linus Torvalds 已提交
553 554 555 556 557
						      th->doff << 2,
						      skb->csum));
	}
}

558
/* This routine computes an IPv4 TCP checksum. */
559
void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb)
560
{
561
	const struct inet_sock *inet = inet_sk(sk);
562 563 564

	__tcp_v4_send_check(skb, inet->inet_saddr, inet->inet_daddr);
}
E
Eric Dumazet 已提交
565
EXPORT_SYMBOL(tcp_v4_send_check);
566

L
Linus Torvalds 已提交
567 568 569 570 571 572 573 574 575 576 577 578 579
/*
 *	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.
 */

580
static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
581
{
582
	const struct tcphdr *th = tcp_hdr(skb);
583 584 585
	struct {
		struct tcphdr th;
#ifdef CONFIG_TCP_MD5SIG
586
		__be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)];
587 588
#endif
	} rep;
L
Linus Torvalds 已提交
589
	struct ip_reply_arg arg;
590
#ifdef CONFIG_TCP_MD5SIG
591
	struct tcp_md5sig_key *key = NULL;
592 593 594 595
	const __u8 *hash_location = NULL;
	unsigned char newhash[16];
	int genhash;
	struct sock *sk1 = NULL;
596
#endif
597
	struct net *net;
L
Linus Torvalds 已提交
598 599 600 601 602

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

603 604 605 606
	/* If sk not NULL, it means we did a successful lookup and incoming
	 * route had to be correct. prequeue might have dropped our dst.
	 */
	if (!sk && skb_rtable(skb)->rt_type != RTN_LOCAL)
L
Linus Torvalds 已提交
607 608 609
		return;

	/* Swap the send and the receive. */
610 611 612 613 614
	memset(&rep, 0, sizeof(rep));
	rep.th.dest   = th->source;
	rep.th.source = th->dest;
	rep.th.doff   = sizeof(struct tcphdr) / 4;
	rep.th.rst    = 1;
L
Linus Torvalds 已提交
615 616

	if (th->ack) {
617
		rep.th.seq = th->ack_seq;
L
Linus Torvalds 已提交
618
	} else {
619 620 621
		rep.th.ack = 1;
		rep.th.ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin +
				       skb->len - (th->doff << 2));
L
Linus Torvalds 已提交
622 623
	}

624
	memset(&arg, 0, sizeof(arg));
625 626 627
	arg.iov[0].iov_base = (unsigned char *)&rep;
	arg.iov[0].iov_len  = sizeof(rep.th);

628
	net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
629
#ifdef CONFIG_TCP_MD5SIG
630
	rcu_read_lock();
631
	hash_location = tcp_parse_md5sig_option(th);
632
	if (sk && sk_fullsock(sk)) {
633 634 635
		key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)
					&ip_hdr(skb)->saddr, AF_INET);
	} else if (hash_location) {
636 637 638 639 640 641 642
		/*
		 * active side is lost. Try to find listening socket through
		 * source port, and then find md5 key through listening socket.
		 * we are not loose security here:
		 * Incoming packet is checked with md5 hash with finding key,
		 * no RST generated if md5 hash doesn't match.
		 */
643 644
		sk1 = __inet_lookup_listener(net, &tcp_hashinfo, NULL, 0,
					     ip_hdr(skb)->saddr,
645
					     th->source, ip_hdr(skb)->daddr,
646 647 648
					     ntohs(th->source), inet_iif(skb));
		/* don't send rst if it can't find key */
		if (!sk1)
649 650
			goto out;

651 652 653
		key = tcp_md5_do_lookup(sk1, (union tcp_md5_addr *)
					&ip_hdr(skb)->saddr, AF_INET);
		if (!key)
654 655
			goto out;

656

657
		genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, skb);
658
		if (genhash || memcmp(hash_location, newhash, 16) != 0)
659 660
			goto out;

661 662
	}

663 664 665 666 667 668 669 670 671
	if (key) {
		rep.opt[0] = htonl((TCPOPT_NOP << 24) |
				   (TCPOPT_NOP << 16) |
				   (TCPOPT_MD5SIG << 8) |
				   TCPOLEN_MD5SIG);
		/* Update length and the length the header thinks exists */
		arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED;
		rep.th.doff = arg.iov[0].iov_len / 4;

672
		tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1],
673 674
				     key, ip_hdr(skb)->saddr,
				     ip_hdr(skb)->daddr, &rep.th);
675 676
	}
#endif
677 678
	arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
				      ip_hdr(skb)->saddr, /* XXX */
679
				      arg.iov[0].iov_len, IPPROTO_TCP, 0);
L
Linus Torvalds 已提交
680
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
681 682
	arg.flags = (sk && inet_sk_transparent(sk)) ? IP_REPLY_ARG_NOSRCCHECK : 0;

683
	/* When socket is gone, all binding information is lost.
A
Alexey Kuznetsov 已提交
684 685
	 * routing might fail in this case. No choice here, if we choose to force
	 * input interface, we will misroute in case of asymmetric route.
686
	 */
A
Alexey Kuznetsov 已提交
687 688
	if (sk)
		arg.bound_dev_if = sk->sk_bound_dev_if;
L
Linus Torvalds 已提交
689

690 691 692
	BUILD_BUG_ON(offsetof(struct sock, sk_bound_dev_if) !=
		     offsetof(struct inet_timewait_sock, tw_bound_dev_if));

693
	arg.tos = ip_hdr(skb)->tos;
694
	local_bh_disable();
695 696
	ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk),
			      skb, &TCP_SKB_CB(skb)->header.h4.opt,
697 698
			      ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
			      &arg, arg.iov[0].iov_len);
L
Linus Torvalds 已提交
699

E
Eric Dumazet 已提交
700 701
	__TCP_INC_STATS(net, TCP_MIB_OUTSEGS);
	__TCP_INC_STATS(net, TCP_MIB_OUTRSTS);
702
	local_bh_enable();
703 704

#ifdef CONFIG_TCP_MD5SIG
705 706
out:
	rcu_read_unlock();
707
#endif
L
Linus Torvalds 已提交
708 709 710 711 712 713
}

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

714 715
static void tcp_v4_send_ack(struct net *net,
			    struct sk_buff *skb, u32 seq, u32 ack,
716
			    u32 win, u32 tsval, u32 tsecr, int oif,
717
			    struct tcp_md5sig_key *key,
718
			    int reply_flags, u8 tos)
L
Linus Torvalds 已提交
719
{
720
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
721 722
	struct {
		struct tcphdr th;
723
		__be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2)
724
#ifdef CONFIG_TCP_MD5SIG
725
			   + (TCPOLEN_MD5SIG_ALIGNED >> 2)
726 727
#endif
			];
L
Linus Torvalds 已提交
728 729 730 731
	} rep;
	struct ip_reply_arg arg;

	memset(&rep.th, 0, sizeof(struct tcphdr));
732
	memset(&arg, 0, sizeof(arg));
L
Linus Torvalds 已提交
733 734 735

	arg.iov[0].iov_base = (unsigned char *)&rep;
	arg.iov[0].iov_len  = sizeof(rep.th);
736
	if (tsecr) {
737 738 739
		rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				   (TCPOPT_TIMESTAMP << 8) |
				   TCPOLEN_TIMESTAMP);
740 741
		rep.opt[1] = htonl(tsval);
		rep.opt[2] = htonl(tsecr);
742
		arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED;
L
Linus Torvalds 已提交
743 744 745 746 747 748 749 750 751 752 753
	}

	/* Swap the send and the receive. */
	rep.th.dest    = th->source;
	rep.th.source  = th->dest;
	rep.th.doff    = arg.iov[0].iov_len / 4;
	rep.th.seq     = htonl(seq);
	rep.th.ack_seq = htonl(ack);
	rep.th.ack     = 1;
	rep.th.window  = htons(win);

754 755
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
756
		int offset = (tsecr) ? 3 : 0;
757 758 759 760 761 762 763 764

		rep.opt[offset++] = htonl((TCPOPT_NOP << 24) |
					  (TCPOPT_NOP << 16) |
					  (TCPOPT_MD5SIG << 8) |
					  TCPOLEN_MD5SIG);
		arg.iov[0].iov_len += TCPOLEN_MD5SIG_ALIGNED;
		rep.th.doff = arg.iov[0].iov_len/4;

765
		tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset],
766 767
				    key, ip_hdr(skb)->saddr,
				    ip_hdr(skb)->daddr, &rep.th);
768 769
	}
#endif
770
	arg.flags = reply_flags;
771 772
	arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
				      ip_hdr(skb)->saddr, /* XXX */
L
Linus Torvalds 已提交
773 774
				      arg.iov[0].iov_len, IPPROTO_TCP, 0);
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
775 776
	if (oif)
		arg.bound_dev_if = oif;
777
	arg.tos = tos;
778
	local_bh_disable();
779 780
	ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk),
			      skb, &TCP_SKB_CB(skb)->header.h4.opt,
781 782
			      ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
			      &arg, arg.iov[0].iov_len);
L
Linus Torvalds 已提交
783

E
Eric Dumazet 已提交
784
	__TCP_INC_STATS(net, TCP_MIB_OUTSEGS);
785
	local_bh_enable();
L
Linus Torvalds 已提交
786 787 788 789
}

static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
790
	struct inet_timewait_sock *tw = inet_twsk(sk);
791
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
792

793 794
	tcp_v4_send_ack(sock_net(sk), skb,
			tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
795
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
796
			tcp_time_stamp + tcptw->tw_ts_offset,
797 798
			tcptw->tw_ts_recent,
			tw->tw_bound_dev_if,
799
			tcp_twsk_md5_key(tcptw),
800 801
			tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0,
			tw->tw_tos
802
			);
L
Linus Torvalds 已提交
803

804
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
805 806
}

807
static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
808
				  struct request_sock *req)
L
Linus Torvalds 已提交
809
{
810 811 812
	/* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
	 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
	 */
813 814 815
	u32 seq = (sk->sk_state == TCP_LISTEN) ? tcp_rsk(req)->snt_isn + 1 :
					     tcp_sk(sk)->snd_nxt;

816 817 818 819 820
	/* RFC 7323 2.3
	 * The window field (SEG.WND) of every outgoing segment, with the
	 * exception of <SYN> segments, MUST be right-shifted by
	 * Rcv.Wind.Shift bits:
	 */
821
	tcp_v4_send_ack(sock_net(sk), skb, seq,
822 823
			tcp_rsk(req)->rcv_nxt,
			req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale,
824
			tcp_time_stamp,
825 826
			req->ts_recent,
			0,
E
Eric Dumazet 已提交
827 828
			tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&ip_hdr(skb)->daddr,
					  AF_INET),
829 830
			inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0,
			ip_hdr(skb)->tos);
L
Linus Torvalds 已提交
831 832 833
}

/*
834
 *	Send a SYN-ACK after having received a SYN.
835
 *	This still operates on a request_sock only, not on a big
L
Linus Torvalds 已提交
836 837
 *	socket.
 */
838
static int tcp_v4_send_synack(const struct sock *sk, struct dst_entry *dst,
839
			      struct flowi *fl,
840
			      struct request_sock *req,
841
			      struct tcp_fastopen_cookie *foc,
842
			      enum tcp_synack_type synack_type)
L
Linus Torvalds 已提交
843
{
844
	const struct inet_request_sock *ireq = inet_rsk(req);
845
	struct flowi4 fl4;
L
Linus Torvalds 已提交
846
	int err = -1;
847
	struct sk_buff *skb;
L
Linus Torvalds 已提交
848 849

	/* First, grab a route. */
850
	if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL)
851
		return -1;
L
Linus Torvalds 已提交
852

853
	skb = tcp_make_synack(sk, dst, req, foc, synack_type);
L
Linus Torvalds 已提交
854 855

	if (skb) {
856
		__tcp_v4_send_check(skb, ireq->ir_loc_addr, ireq->ir_rmt_addr);
L
Linus Torvalds 已提交
857

858 859
		err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr,
					    ireq->ir_rmt_addr,
860
					    ireq->opt);
861
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
862 863 864 865 866 867
	}

	return err;
}

/*
868
 *	IPv4 request_sock destructor.
L
Linus Torvalds 已提交
869
 */
870
static void tcp_v4_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
871
{
J
Jesper Juhl 已提交
872
	kfree(inet_rsk(req)->opt);
L
Linus Torvalds 已提交
873 874
}

875 876 877 878 879 880 881 882
#ifdef CONFIG_TCP_MD5SIG
/*
 * RFC2385 MD5 checksumming requires a mapping of
 * IP address->MD5 Key.
 * We need to maintain these in the sk structure.
 */

/* Find the Key structure for an address.  */
883
struct tcp_md5sig_key *tcp_md5_do_lookup(const struct sock *sk,
E
Eric Dumazet 已提交
884 885
					 const union tcp_md5_addr *addr,
					 int family)
886
{
887
	const struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
888 889
	struct tcp_md5sig_key *key;
	unsigned int size = sizeof(struct in_addr);
890
	const struct tcp_md5sig_info *md5sig;
891

892 893
	/* caller either holds rcu_read_lock() or socket lock */
	md5sig = rcu_dereference_check(tp->md5sig_info,
894
				       lockdep_sock_is_held(sk));
895
	if (!md5sig)
896
		return NULL;
E
Eric Dumazet 已提交
897 898 899 900
#if IS_ENABLED(CONFIG_IPV6)
	if (family == AF_INET6)
		size = sizeof(struct in6_addr);
#endif
901
	hlist_for_each_entry_rcu(key, &md5sig->head, node) {
E
Eric Dumazet 已提交
902 903 904 905
		if (key->family != family)
			continue;
		if (!memcmp(&key->addr, addr, size))
			return key;
906 907 908
	}
	return NULL;
}
E
Eric Dumazet 已提交
909
EXPORT_SYMBOL(tcp_md5_do_lookup);
910

911
struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk,
912
					 const struct sock *addr_sk)
913
{
914
	const union tcp_md5_addr *addr;
E
Eric Dumazet 已提交
915

916
	addr = (const union tcp_md5_addr *)&addr_sk->sk_daddr;
E
Eric Dumazet 已提交
917
	return tcp_md5_do_lookup(sk, addr, AF_INET);
918 919 920 921
}
EXPORT_SYMBOL(tcp_v4_md5_lookup);

/* This can be called on a newly created socket, from other files */
E
Eric Dumazet 已提交
922 923
int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
		   int family, const u8 *newkey, u8 newkeylen, gfp_t gfp)
924 925
{
	/* Add Key to the list */
926
	struct tcp_md5sig_key *key;
927
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
928
	struct tcp_md5sig_info *md5sig;
929

930
	key = tcp_md5_do_lookup(sk, addr, family);
931 932
	if (key) {
		/* Pre-existing entry - just update that one. */
E
Eric Dumazet 已提交
933
		memcpy(key->key, newkey, newkeylen);
934
		key->keylen = newkeylen;
E
Eric Dumazet 已提交
935 936
		return 0;
	}
937

938
	md5sig = rcu_dereference_protected(tp->md5sig_info,
939
					   lockdep_sock_is_held(sk));
E
Eric Dumazet 已提交
940 941 942
	if (!md5sig) {
		md5sig = kmalloc(sizeof(*md5sig), gfp);
		if (!md5sig)
943 944
			return -ENOMEM;

E
Eric Dumazet 已提交
945 946
		sk_nocaps_add(sk, NETIF_F_GSO_MASK);
		INIT_HLIST_HEAD(&md5sig->head);
947
		rcu_assign_pointer(tp->md5sig_info, md5sig);
E
Eric Dumazet 已提交
948
	}
949

950
	key = sock_kmalloc(sk, sizeof(*key), gfp);
E
Eric Dumazet 已提交
951 952
	if (!key)
		return -ENOMEM;
953
	if (!tcp_alloc_md5sig_pool()) {
954
		sock_kfree_s(sk, key, sizeof(*key));
E
Eric Dumazet 已提交
955
		return -ENOMEM;
956
	}
E
Eric Dumazet 已提交
957 958 959 960 961 962 963 964

	memcpy(key->key, newkey, newkeylen);
	key->keylen = newkeylen;
	key->family = family;
	memcpy(&key->addr, addr,
	       (family == AF_INET6) ? sizeof(struct in6_addr) :
				      sizeof(struct in_addr));
	hlist_add_head_rcu(&key->node, &md5sig->head);
965 966
	return 0;
}
E
Eric Dumazet 已提交
967
EXPORT_SYMBOL(tcp_md5_do_add);
968

E
Eric Dumazet 已提交
969
int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family)
970
{
E
Eric Dumazet 已提交
971 972
	struct tcp_md5sig_key *key;

973
	key = tcp_md5_do_lookup(sk, addr, family);
E
Eric Dumazet 已提交
974 975 976
	if (!key)
		return -ENOENT;
	hlist_del_rcu(&key->node);
977
	atomic_sub(sizeof(*key), &sk->sk_omem_alloc);
E
Eric Dumazet 已提交
978 979
	kfree_rcu(key, rcu);
	return 0;
980
}
E
Eric Dumazet 已提交
981
EXPORT_SYMBOL(tcp_md5_do_del);
982

983
static void tcp_clear_md5_list(struct sock *sk)
984 985
{
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
986
	struct tcp_md5sig_key *key;
987
	struct hlist_node *n;
988
	struct tcp_md5sig_info *md5sig;
989

990 991
	md5sig = rcu_dereference_protected(tp->md5sig_info, 1);

992
	hlist_for_each_entry_safe(key, n, &md5sig->head, node) {
E
Eric Dumazet 已提交
993
		hlist_del_rcu(&key->node);
994
		atomic_sub(sizeof(*key), &sk->sk_omem_alloc);
E
Eric Dumazet 已提交
995
		kfree_rcu(key, rcu);
996 997 998
	}
}

999 1000
static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval,
				 int optlen)
1001 1002 1003 1004 1005 1006 1007
{
	struct tcp_md5sig cmd;
	struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr;

	if (optlen < sizeof(cmd))
		return -EINVAL;

1008
	if (copy_from_user(&cmd, optval, sizeof(cmd)))
1009 1010 1011 1012 1013
		return -EFAULT;

	if (sin->sin_family != AF_INET)
		return -EINVAL;

1014
	if (!cmd.tcpm_keylen)
E
Eric Dumazet 已提交
1015 1016
		return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr,
				      AF_INET);
1017 1018 1019 1020

	if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
		return -EINVAL;

E
Eric Dumazet 已提交
1021 1022 1023
	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr,
			      AF_INET, cmd.tcpm_key, cmd.tcpm_keylen,
			      GFP_KERNEL);
1024 1025
}

1026 1027 1028
static int tcp_v4_md5_hash_headers(struct tcp_md5sig_pool *hp,
				   __be32 daddr, __be32 saddr,
				   const struct tcphdr *th, int nbytes)
1029 1030
{
	struct tcp4_pseudohdr *bp;
1031
	struct scatterlist sg;
1032
	struct tcphdr *_th;
1033

1034
	bp = hp->scratch;
1035 1036 1037
	bp->saddr = saddr;
	bp->daddr = daddr;
	bp->pad = 0;
1038
	bp->protocol = IPPROTO_TCP;
1039
	bp->len = cpu_to_be16(nbytes);
1040

1041 1042 1043 1044 1045 1046 1047
	_th = (struct tcphdr *)(bp + 1);
	memcpy(_th, th, sizeof(*th));
	_th->check = 0;

	sg_init_one(&sg, bp, sizeof(*bp) + sizeof(*th));
	ahash_request_set_crypt(hp->md5_req, &sg, NULL,
				sizeof(*bp) + sizeof(*th));
H
Herbert Xu 已提交
1048
	return crypto_ahash_update(hp->md5_req);
1049 1050
}

E
Eric Dumazet 已提交
1051
static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
1052
			       __be32 daddr, __be32 saddr, const struct tcphdr *th)
1053 1054
{
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
1055
	struct ahash_request *req;
1056 1057 1058 1059

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
1060
	req = hp->md5_req;
1061

H
Herbert Xu 已提交
1062
	if (crypto_ahash_init(req))
1063
		goto clear_hash;
1064
	if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2))
1065 1066 1067
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
H
Herbert Xu 已提交
1068 1069
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
1070 1071 1072 1073
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
1074

1075 1076 1077 1078
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
1079
	return 1;
1080 1081
}

1082 1083
int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key,
			const struct sock *sk,
E
Eric Dumazet 已提交
1084
			const struct sk_buff *skb)
1085
{
1086
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
1087
	struct ahash_request *req;
E
Eric Dumazet 已提交
1088
	const struct tcphdr *th = tcp_hdr(skb);
1089 1090
	__be32 saddr, daddr;

1091 1092 1093
	if (sk) { /* valid for establish/request sockets */
		saddr = sk->sk_rcv_saddr;
		daddr = sk->sk_daddr;
1094
	} else {
1095 1096 1097
		const struct iphdr *iph = ip_hdr(skb);
		saddr = iph->saddr;
		daddr = iph->daddr;
1098
	}
1099 1100 1101 1102

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
1103
	req = hp->md5_req;
1104

H
Herbert Xu 已提交
1105
	if (crypto_ahash_init(req))
1106 1107
		goto clear_hash;

1108
	if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, skb->len))
1109 1110 1111 1112 1113
		goto clear_hash;
	if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
H
Herbert Xu 已提交
1114 1115
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
1116 1117 1118 1119 1120 1121 1122 1123 1124 1125
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;

clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
	return 1;
1126
}
1127
EXPORT_SYMBOL(tcp_v4_md5_hash_skb);
1128

1129 1130
#endif

1131
/* Called with rcu_read_lock() */
1132
static bool tcp_v4_inbound_md5_hash(const struct sock *sk,
1133
				    const struct sk_buff *skb)
1134
{
1135
#ifdef CONFIG_TCP_MD5SIG
1136 1137 1138 1139 1140 1141 1142 1143
	/*
	 * This gets called for each TCP segment that arrives
	 * so we want to be efficient.
	 * We have 3 drop cases:
	 * o No MD5 hash and one expected.
	 * o MD5 hash and we're not expecting one.
	 * o MD5 hash and its wrong.
	 */
1144
	const __u8 *hash_location = NULL;
1145
	struct tcp_md5sig_key *hash_expected;
1146
	const struct iphdr *iph = ip_hdr(skb);
1147
	const struct tcphdr *th = tcp_hdr(skb);
1148 1149 1150
	int genhash;
	unsigned char newhash[16];

E
Eric Dumazet 已提交
1151 1152
	hash_expected = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&iph->saddr,
					  AF_INET);
1153
	hash_location = tcp_parse_md5sig_option(th);
1154 1155 1156

	/* We've parsed the options - do we have a hash? */
	if (!hash_expected && !hash_location)
E
Eric Dumazet 已提交
1157
		return false;
1158 1159

	if (hash_expected && !hash_location) {
1160
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
E
Eric Dumazet 已提交
1161
		return true;
1162 1163 1164
	}

	if (!hash_expected && hash_location) {
1165
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
E
Eric Dumazet 已提交
1166
		return true;
1167 1168 1169 1170 1171
	}

	/* Okay, so this is hash_expected and hash_location -
	 * so we need to calculate the checksum.
	 */
1172 1173
	genhash = tcp_v4_md5_hash_skb(newhash,
				      hash_expected,
1174
				      NULL, skb);
1175 1176

	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
1177
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE);
1178 1179 1180 1181 1182
		net_info_ratelimited("MD5 Hash failed for (%pI4, %d)->(%pI4, %d)%s\n",
				     &iph->saddr, ntohs(th->source),
				     &iph->daddr, ntohs(th->dest),
				     genhash ? " tcp_v4_calc_md5_hash failed"
				     : "");
E
Eric Dumazet 已提交
1183
		return true;
1184
	}
E
Eric Dumazet 已提交
1185
	return false;
1186
#endif
1187 1188
	return false;
}
1189

1190 1191
static void tcp_v4_init_req(struct request_sock *req,
			    const struct sock *sk_listener,
1192 1193 1194 1195
			    struct sk_buff *skb)
{
	struct inet_request_sock *ireq = inet_rsk(req);

1196 1197
	sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr);
	sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr);
1198 1199 1200
	ireq->opt = tcp_v4_save_options(skb);
}

1201 1202
static struct dst_entry *tcp_v4_route_req(const struct sock *sk,
					  struct flowi *fl,
1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217
					  const struct request_sock *req,
					  bool *strict)
{
	struct dst_entry *dst = inet_csk_route_req(sk, &fl->u.ip4, req);

	if (strict) {
		if (fl->u.ip4.daddr == inet_rsk(req)->ir_rmt_addr)
			*strict = true;
		else
			*strict = false;
	}

	return dst;
}

1218
struct request_sock_ops tcp_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
1219
	.family		=	PF_INET,
1220
	.obj_size	=	sizeof(struct tcp_request_sock),
1221
	.rtx_syn_ack	=	tcp_rtx_synack,
1222 1223
	.send_ack	=	tcp_v4_reqsk_send_ack,
	.destructor	=	tcp_v4_reqsk_destructor,
L
Linus Torvalds 已提交
1224
	.send_reset	=	tcp_v4_send_reset,
S
stephen hemminger 已提交
1225
	.syn_ack_timeout =	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
1226 1227
};

1228
static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = {
1229
	.mss_clamp	=	TCP_MSS_DEFAULT,
1230
#ifdef CONFIG_TCP_MD5SIG
1231
	.req_md5_lookup	=	tcp_v4_md5_lookup,
1232
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1233
#endif
1234
	.init_req	=	tcp_v4_init_req,
1235 1236 1237
#ifdef CONFIG_SYN_COOKIES
	.cookie_init_seq =	cookie_v4_init_sequence,
#endif
1238
	.route_req	=	tcp_v4_route_req,
1239
	.init_seq	=	tcp_v4_init_sequence,
1240
	.send_synack	=	tcp_v4_send_synack,
1241
};
1242

L
Linus Torvalds 已提交
1243 1244 1245
int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
{
	/* Never answer to SYNs send to broadcast or multicast */
E
Eric Dumazet 已提交
1246
	if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
L
Linus Torvalds 已提交
1247 1248
		goto drop;

O
Octavian Purdila 已提交
1249 1250
	return tcp_conn_request(&tcp_request_sock_ops,
				&tcp_request_sock_ipv4_ops, sk, skb);
L
Linus Torvalds 已提交
1251 1252

drop:
1253
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1254 1255
	return 0;
}
E
Eric Dumazet 已提交
1256
EXPORT_SYMBOL(tcp_v4_conn_request);
L
Linus Torvalds 已提交
1257 1258 1259 1260 1261 1262


/*
 * The three way handshake has completed - we got a valid synack -
 * now create the new socket.
 */
1263
struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
1264
				  struct request_sock *req,
1265 1266 1267
				  struct dst_entry *dst,
				  struct request_sock *req_unhash,
				  bool *own_req)
L
Linus Torvalds 已提交
1268
{
1269
	struct inet_request_sock *ireq;
L
Linus Torvalds 已提交
1270 1271 1272
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1273 1274 1275
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1276
	struct ip_options_rcu *inet_opt;
L
Linus Torvalds 已提交
1277 1278 1279 1280 1281 1282

	if (sk_acceptq_is_full(sk))
		goto exit_overflow;

	newsk = tcp_create_openreq_child(sk, req, skb);
	if (!newsk)
1283
		goto exit_nonewsk;
L
Linus Torvalds 已提交
1284

1285
	newsk->sk_gso_type = SKB_GSO_TCPV4;
1286
	inet_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1287 1288 1289

	newtp		      = tcp_sk(newsk);
	newinet		      = inet_sk(newsk);
1290
	ireq		      = inet_rsk(req);
1291 1292
	sk_daddr_set(newsk, ireq->ir_rmt_addr);
	sk_rcv_saddr_set(newsk, ireq->ir_loc_addr);
1293
	newsk->sk_bound_dev_if = ireq->ir_iif;
1294
	newinet->inet_saddr	      = ireq->ir_loc_addr;
1295 1296
	inet_opt	      = ireq->opt;
	rcu_assign_pointer(newinet->inet_opt, inet_opt);
1297
	ireq->opt	      = NULL;
1298
	newinet->mc_index     = inet_iif(skb);
1299
	newinet->mc_ttl	      = ip_hdr(skb)->ttl;
1300
	newinet->rcv_tos      = ip_hdr(skb)->tos;
1301
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
1302 1303
	if (inet_opt)
		inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
E
Eric Dumazet 已提交
1304
	newinet->inet_id = newtp->write_seq ^ jiffies;
L
Linus Torvalds 已提交
1305

E
Eric Dumazet 已提交
1306 1307 1308 1309 1310 1311 1312
	if (!dst) {
		dst = inet_csk_route_child_sock(sk, newsk, req);
		if (!dst)
			goto put_and_exit;
	} else {
		/* syncookie case : see end of cookie_v4_check() */
	}
1313 1314
	sk_setup_caps(newsk, dst);

1315 1316
	tcp_ca_openreq_child(newsk, dst);

L
Linus Torvalds 已提交
1317
	tcp_sync_mss(newsk, dst_mtu(dst));
1318
	newtp->advmss = dst_metric_advmss(dst);
1319 1320 1321 1322
	if (tcp_sk(sk)->rx_opt.user_mss &&
	    tcp_sk(sk)->rx_opt.user_mss < newtp->advmss)
		newtp->advmss = tcp_sk(sk)->rx_opt.user_mss;

L
Linus Torvalds 已提交
1323 1324
	tcp_initialize_rcv_mss(newsk);

1325 1326
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
E
Eric Dumazet 已提交
1327 1328
	key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&newinet->inet_daddr,
				AF_INET);
1329
	if (key) {
1330 1331 1332 1333 1334 1335
		/*
		 * We're using one, so create a matching key
		 * on the newsk structure. If we fail to get
		 * memory, then we end up not copying the key
		 * across. Shucks.
		 */
E
Eric Dumazet 已提交
1336 1337
		tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newinet->inet_daddr,
			       AF_INET, key->key, key->keylen, GFP_ATOMIC);
E
Eric Dumazet 已提交
1338
		sk_nocaps_add(newsk, NETIF_F_GSO_MASK);
1339 1340 1341
	}
#endif

1342 1343
	if (__inet_inherit_port(sk, newsk) < 0)
		goto put_and_exit;
1344
	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
E
Eric Dumazet 已提交
1345
	if (*own_req)
1346
		tcp_move_syn(newtp, req);
L
Linus Torvalds 已提交
1347 1348 1349 1350

	return newsk;

exit_overflow:
1351
	NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1352 1353
exit_nonewsk:
	dst_release(dst);
L
Linus Torvalds 已提交
1354
exit:
1355
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1356
	return NULL;
1357
put_and_exit:
1358 1359
	inet_csk_prepare_forced_close(newsk);
	tcp_done(newsk);
1360
	goto exit;
L
Linus Torvalds 已提交
1361
}
E
Eric Dumazet 已提交
1362
EXPORT_SYMBOL(tcp_v4_syn_recv_sock);
L
Linus Torvalds 已提交
1363

1364
static struct sock *tcp_v4_cookie_check(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
1365
{
1366
#ifdef CONFIG_SYN_COOKIES
1367
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1368

1369
	if (!th->syn)
C
Cong Wang 已提交
1370
		sk = cookie_v4_check(sk, skb);
L
Linus Torvalds 已提交
1371 1372 1373 1374 1375
#endif
	return sk;
}

/* The socket must have it's spinlock held when we get
1376
 * here, unless it is a TCP_LISTEN socket.
L
Linus Torvalds 已提交
1377 1378 1379 1380 1381 1382 1383 1384
 *
 * 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)
{
1385 1386
	struct sock *rsk;

L
Linus Torvalds 已提交
1387
	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1388 1389
		struct dst_entry *dst = sk->sk_rx_dst;

1390
		sock_rps_save_rxhash(sk, skb);
1391
		sk_mark_napi_id(sk, skb);
1392
		if (dst) {
E
Eric Dumazet 已提交
1393
			if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif ||
1394
			    !dst->ops->check(dst, 0)) {
1395 1396 1397 1398
				dst_release(dst);
				sk->sk_rx_dst = NULL;
			}
		}
1399
		tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
L
Linus Torvalds 已提交
1400 1401 1402
		return 0;
	}

E
Eric Dumazet 已提交
1403
	if (tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1404 1405 1406
		goto csum_err;

	if (sk->sk_state == TCP_LISTEN) {
1407 1408
		struct sock *nsk = tcp_v4_cookie_check(sk, skb);

L
Linus Torvalds 已提交
1409 1410 1411
		if (!nsk)
			goto discard;
		if (nsk != sk) {
1412
			sock_rps_save_rxhash(nsk, skb);
1413
			sk_mark_napi_id(nsk, skb);
1414 1415
			if (tcp_child_process(sk, nsk, skb)) {
				rsk = nsk;
L
Linus Torvalds 已提交
1416
				goto reset;
1417
			}
L
Linus Torvalds 已提交
1418 1419
			return 0;
		}
1420
	} else
1421
		sock_rps_save_rxhash(sk, skb);
1422

1423
	if (tcp_rcv_state_process(sk, skb)) {
1424
		rsk = sk;
L
Linus Torvalds 已提交
1425
		goto reset;
1426
	}
L
Linus Torvalds 已提交
1427 1428 1429
	return 0;

reset:
1430
	tcp_v4_send_reset(rsk, skb);
L
Linus Torvalds 已提交
1431 1432 1433 1434 1435 1436 1437 1438 1439 1440
discard:
	kfree_skb(skb);
	/* Be careful here. If this function gets more complicated and
	 * gcc suffers from register pressure on the x86, sk (in %ebx)
	 * might be destroyed here. This current version compiles correctly,
	 * but you have been warned.
	 */
	return 0;

csum_err:
1441 1442
	TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS);
	TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1443 1444
	goto discard;
}
E
Eric Dumazet 已提交
1445
EXPORT_SYMBOL(tcp_v4_do_rcv);
L
Linus Torvalds 已提交
1446

1447
void tcp_v4_early_demux(struct sk_buff *skb)
D
David S. Miller 已提交
1448 1449 1450 1451 1452 1453
{
	const struct iphdr *iph;
	const struct tcphdr *th;
	struct sock *sk;

	if (skb->pkt_type != PACKET_HOST)
1454
		return;
D
David S. Miller 已提交
1455

1456
	if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr)))
1457
		return;
D
David S. Miller 已提交
1458 1459

	iph = ip_hdr(skb);
1460
	th = tcp_hdr(skb);
D
David S. Miller 已提交
1461 1462

	if (th->doff < sizeof(struct tcphdr) / 4)
1463
		return;
D
David S. Miller 已提交
1464

1465
	sk = __inet_lookup_established(dev_net(skb->dev), &tcp_hashinfo,
D
David S. Miller 已提交
1466
				       iph->saddr, th->source,
1467
				       iph->daddr, ntohs(th->dest),
E
Eric Dumazet 已提交
1468
				       skb->skb_iif);
D
David S. Miller 已提交
1469 1470 1471
	if (sk) {
		skb->sk = sk;
		skb->destructor = sock_edemux;
1472
		if (sk_fullsock(sk)) {
1473
			struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst);
E
Eric Dumazet 已提交
1474

D
David S. Miller 已提交
1475 1476
			if (dst)
				dst = dst_check(dst, 0);
1477
			if (dst &&
E
Eric Dumazet 已提交
1478
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
1479
				skb_dst_set_noref(skb, dst);
D
David S. Miller 已提交
1480 1481 1482 1483
		}
	}
}

E
Eric Dumazet 已提交
1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501
/* Packet is added to VJ-style prequeue for processing in process
 * context, if a reader task is waiting. Apparently, this exciting
 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
 * failed somewhere. Latency? Burstiness? Well, at least now we will
 * see, why it failed. 8)8)				  --ANK
 *
 */
bool tcp_prequeue(struct sock *sk, struct sk_buff *skb)
{
	struct tcp_sock *tp = tcp_sk(sk);

	if (sysctl_tcp_low_latency || !tp->ucopy.task)
		return false;

	if (skb->len <= tcp_hdrlen(skb) &&
	    skb_queue_len(&tp->ucopy.prequeue) == 0)
		return false;

1502 1503 1504 1505 1506 1507 1508 1509 1510
	/* Before escaping RCU protected region, we need to take care of skb
	 * dst. Prequeue is only enabled for established sockets.
	 * For such sockets, we might need the skb dst only to set sk->sk_rx_dst
	 * Instead of doing full sk_rx_dst validity here, let's perform
	 * an optimistic check.
	 */
	if (likely(sk->sk_rx_dst))
		skb_dst_drop(skb);
	else
E
Eric Dumazet 已提交
1511
		skb_dst_force_safe(skb);
1512

E
Eric Dumazet 已提交
1513 1514
	__skb_queue_tail(&tp->ucopy.prequeue, skb);
	tp->ucopy.memory += skb->truesize;
E
Eric Dumazet 已提交
1515 1516
	if (skb_queue_len(&tp->ucopy.prequeue) >= 32 ||
	    tp->ucopy.memory + atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) {
E
Eric Dumazet 已提交
1517 1518 1519
		struct sk_buff *skb1;

		BUG_ON(sock_owned_by_user(sk));
E
Eric Dumazet 已提交
1520 1521
		__NET_ADD_STATS(sock_net(sk), LINUX_MIB_TCPPREQUEUEDROPPED,
				skb_queue_len(&tp->ucopy.prequeue));
E
Eric Dumazet 已提交
1522

E
Eric Dumazet 已提交
1523
		while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL)
E
Eric Dumazet 已提交
1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538
			sk_backlog_rcv(sk, skb1);

		tp->ucopy.memory = 0;
	} else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
		wake_up_interruptible_sync_poll(sk_sleep(sk),
					   POLLIN | POLLRDNORM | POLLRDBAND);
		if (!inet_csk_ack_scheduled(sk))
			inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
						  (3 * tcp_rto_min(sk)) / 4,
						  TCP_RTO_MAX);
	}
	return true;
}
EXPORT_SYMBOL(tcp_prequeue);

E
Eric Dumazet 已提交
1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566
bool tcp_add_backlog(struct sock *sk, struct sk_buff *skb)
{
	u32 limit = sk->sk_rcvbuf + sk->sk_sndbuf;

	/* Only socket owner can try to collapse/prune rx queues
	 * to reduce memory overhead, so add a little headroom here.
	 * Few sockets backlog are possibly concurrently non empty.
	 */
	limit += 64*1024;

	/* In case all data was pulled from skb frags (in __pskb_pull_tail()),
	 * we can fix skb->truesize to its real value to avoid future drops.
	 * This is valid because skb is not yet charged to the socket.
	 * It has been noticed pure SACK packets were sometimes dropped
	 * (if cooked by drivers without copybreak feature).
	 */
	if (!skb->data_len)
		skb->truesize = SKB_TRUESIZE(skb_end_offset(skb));

	if (unlikely(sk_add_backlog(sk, skb, limit))) {
		bh_unlock_sock(sk);
		__NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPBACKLOGDROP);
		return true;
	}
	return false;
}
EXPORT_SYMBOL(tcp_add_backlog);

L
Linus Torvalds 已提交
1567 1568 1569 1570 1571 1572
/*
 *	From tcp_input.c
 */

int tcp_v4_rcv(struct sk_buff *skb)
{
1573
	struct net *net = dev_net(skb->dev);
1574
	const struct iphdr *iph;
1575
	const struct tcphdr *th;
1576
	bool refcounted;
L
Linus Torvalds 已提交
1577 1578 1579 1580 1581 1582 1583
	struct sock *sk;
	int ret;

	if (skb->pkt_type != PACKET_HOST)
		goto discard_it;

	/* Count it even if it's bad */
E
Eric Dumazet 已提交
1584
	__TCP_INC_STATS(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1585 1586 1587 1588

	if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
		goto discard_it;

1589
	th = (const struct tcphdr *)skb->data;
L
Linus Torvalds 已提交
1590

1591
	if (unlikely(th->doff < sizeof(struct tcphdr) / 4))
L
Linus Torvalds 已提交
1592 1593 1594 1595 1596 1597
		goto bad_packet;
	if (!pskb_may_pull(skb, th->doff * 4))
		goto discard_it;

	/* An explanation is required here, I think.
	 * Packet length and doff are validated by header prediction,
S
Stephen Hemminger 已提交
1598
	 * provided case of th->doff==0 is eliminated.
L
Linus Torvalds 已提交
1599
	 * So, we defer the checks. */
1600 1601

	if (skb_checksum_init(skb, IPPROTO_TCP, inet_compute_pseudo))
1602
		goto csum_error;
L
Linus Torvalds 已提交
1603

1604
	th = (const struct tcphdr *)skb->data;
1605
	iph = ip_hdr(skb);
1606 1607 1608 1609 1610 1611 1612
	/* This is tricky : We move IPCB at its correct location into TCP_SKB_CB()
	 * barrier() makes sure compiler wont play fool^Waliasing games.
	 */
	memmove(&TCP_SKB_CB(skb)->header.h4, IPCB(skb),
		sizeof(struct inet_skb_parm));
	barrier();

L
Linus Torvalds 已提交
1613 1614 1615 1616
	TCP_SKB_CB(skb)->seq = ntohl(th->seq);
	TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
				    skb->len - th->doff * 4);
	TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1617
	TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
1618
	TCP_SKB_CB(skb)->tcp_tw_isn = 0;
1619
	TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph);
L
Linus Torvalds 已提交
1620 1621
	TCP_SKB_CB(skb)->sacked	 = 0;

1622
lookup:
1623
	sk = __inet_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th), th->source,
1624
			       th->dest, &refcounted);
L
Linus Torvalds 已提交
1625 1626 1627
	if (!sk)
		goto no_tcp_socket;

E
Eric Dumazet 已提交
1628 1629 1630 1631
process:
	if (sk->sk_state == TCP_TIME_WAIT)
		goto do_time_wait;

1632 1633
	if (sk->sk_state == TCP_NEW_SYN_RECV) {
		struct request_sock *req = inet_reqsk(sk);
1634
		struct sock *nsk;
1635 1636

		sk = req->rsk_listener;
1637
		if (unlikely(tcp_v4_inbound_md5_hash(sk, skb))) {
1638
			sk_drops_add(sk, skb);
1639 1640 1641
			reqsk_put(req);
			goto discard_it;
		}
1642
		if (unlikely(sk->sk_state != TCP_LISTEN)) {
1643
			inet_csk_reqsk_queue_drop_and_put(sk, req);
1644 1645
			goto lookup;
		}
1646 1647 1648
		/* We own a reference on the listener, increase it again
		 * as we might lose it too soon.
		 */
1649
		sock_hold(sk);
1650
		refcounted = true;
1651
		nsk = tcp_check_req(sk, skb, req, false);
1652 1653
		if (!nsk) {
			reqsk_put(req);
1654
			goto discard_and_relse;
1655 1656 1657 1658 1659
		}
		if (nsk == sk) {
			reqsk_put(req);
		} else if (tcp_child_process(sk, nsk, skb)) {
			tcp_v4_send_reset(nsk, skb);
1660
			goto discard_and_relse;
1661
		} else {
1662
			sock_put(sk);
1663 1664 1665
			return 0;
		}
	}
1666
	if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
1667
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
1668
		goto discard_and_relse;
1669
	}
1670

L
Linus Torvalds 已提交
1671 1672
	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;
1673 1674 1675 1676

	if (tcp_v4_inbound_md5_hash(sk, skb))
		goto discard_and_relse;

1677
	nf_reset(skb);
L
Linus Torvalds 已提交
1678

1679
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1680 1681 1682 1683
		goto discard_and_relse;

	skb->dev = NULL;

1684 1685 1686 1687 1688 1689 1690
	if (sk->sk_state == TCP_LISTEN) {
		ret = tcp_v4_do_rcv(sk, skb);
		goto put_and_return;
	}

	sk_incoming_cpu_update(sk);

1691
	bh_lock_sock_nested(sk);
1692
	tcp_segs_in(tcp_sk(sk), skb);
L
Linus Torvalds 已提交
1693 1694
	ret = 0;
	if (!sock_owned_by_user(sk)) {
D
Dan Williams 已提交
1695
		if (!tcp_prequeue(sk, skb))
L
Linus Torvalds 已提交
1696
			ret = tcp_v4_do_rcv(sk, skb);
E
Eric Dumazet 已提交
1697
	} else if (tcp_add_backlog(sk, skb)) {
Z
Zhu Yi 已提交
1698 1699
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1700 1701
	bh_unlock_sock(sk);

1702
put_and_return:
1703 1704
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1705 1706 1707 1708 1709 1710 1711

	return ret;

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

E
Eric Dumazet 已提交
1712
	if (tcp_checksum_complete(skb)) {
1713
csum_error:
E
Eric Dumazet 已提交
1714
		__TCP_INC_STATS(net, TCP_MIB_CSUMERRORS);
L
Linus Torvalds 已提交
1715
bad_packet:
E
Eric Dumazet 已提交
1716
		__TCP_INC_STATS(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1717
	} else {
1718
		tcp_v4_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1719 1720 1721 1722 1723
	}

discard_it:
	/* Discard frame. */
	kfree_skb(skb);
1724
	return 0;
L
Linus Torvalds 已提交
1725 1726

discard_and_relse:
1727
	sk_drops_add(sk, skb);
1728 1729
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1730 1731 1732 1733
	goto discard_it;

do_time_wait:
	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1734
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1735 1736 1737
		goto discard_it;
	}

1738 1739 1740
	if (tcp_checksum_complete(skb)) {
		inet_twsk_put(inet_twsk(sk));
		goto csum_error;
L
Linus Torvalds 已提交
1741
	}
1742
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1743
	case TCP_TW_SYN: {
1744
		struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev),
1745 1746
							&tcp_hashinfo, skb,
							__tcp_hdrlen(th),
1747
							iph->saddr, th->source,
1748
							iph->daddr, th->dest,
1749
							inet_iif(skb));
L
Linus Torvalds 已提交
1750
		if (sk2) {
1751
			inet_twsk_deschedule_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1752
			sk = sk2;
1753
			refcounted = false;
L
Linus Torvalds 已提交
1754 1755 1756 1757 1758 1759 1760 1761
			goto process;
		}
		/* Fall through to ACK */
	}
	case TCP_TW_ACK:
		tcp_v4_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
1762 1763 1764
		tcp_v4_send_reset(sk, skb);
		inet_twsk_deschedule_put(inet_twsk(sk));
		goto discard_it;
L
Linus Torvalds 已提交
1765 1766 1767 1768 1769
	case TCP_TW_SUCCESS:;
	}
	goto discard_it;
}

1770 1771 1772 1773 1774
static struct timewait_sock_ops tcp_timewait_sock_ops = {
	.twsk_obj_size	= sizeof(struct tcp_timewait_sock),
	.twsk_unique	= tcp_twsk_unique,
	.twsk_destructor= tcp_twsk_destructor,
};
L
Linus Torvalds 已提交
1775

1776
void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
E
Eric Dumazet 已提交
1777 1778 1779
{
	struct dst_entry *dst = skb_dst(skb);

E
Eric Dumazet 已提交
1780
	if (dst && dst_hold_safe(dst)) {
1781 1782 1783
		sk->sk_rx_dst = dst;
		inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
	}
E
Eric Dumazet 已提交
1784
}
1785
EXPORT_SYMBOL(inet_sk_rx_dst_set);
E
Eric Dumazet 已提交
1786

1787
const struct inet_connection_sock_af_ops ipv4_specific = {
1788 1789 1790
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
E
Eric Dumazet 已提交
1791
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1792 1793 1794 1795 1796 1797 1798
	.conn_request	   = tcp_v4_conn_request,
	.syn_recv_sock	   = tcp_v4_syn_recv_sock,
	.net_header_len	   = sizeof(struct iphdr),
	.setsockopt	   = ip_setsockopt,
	.getsockopt	   = ip_getsockopt,
	.addr2sockaddr	   = inet_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in),
1799
	.bind_conflict	   = inet_csk_bind_conflict,
1800
#ifdef CONFIG_COMPAT
1801 1802
	.compat_setsockopt = compat_ip_setsockopt,
	.compat_getsockopt = compat_ip_getsockopt,
1803
#endif
1804
	.mtu_reduced	   = tcp_v4_mtu_reduced,
L
Linus Torvalds 已提交
1805
};
E
Eric Dumazet 已提交
1806
EXPORT_SYMBOL(ipv4_specific);
L
Linus Torvalds 已提交
1807

1808
#ifdef CONFIG_TCP_MD5SIG
1809
static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = {
1810
	.md5_lookup		= tcp_v4_md5_lookup,
1811
	.calc_md5_hash		= tcp_v4_md5_hash_skb,
1812 1813
	.md5_parse		= tcp_v4_parse_md5_keys,
};
1814
#endif
1815

L
Linus Torvalds 已提交
1816 1817 1818 1819 1820
/* 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)
{
1821
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1822

1823
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1824

1825
	icsk->icsk_af_ops = &ipv4_specific;
1826

1827
#ifdef CONFIG_TCP_MD5SIG
1828
	tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific;
1829
#endif
L
Linus Torvalds 已提交
1830 1831 1832 1833

	return 0;
}

1834
void tcp_v4_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1835 1836 1837 1838 1839
{
	struct tcp_sock *tp = tcp_sk(sk);

	tcp_clear_xmit_timers(sk);

1840
	tcp_cleanup_congestion_control(sk);
1841

L
Linus Torvalds 已提交
1842
	/* Cleanup up the write buffer. */
1843
	tcp_write_queue_purge(sk);
L
Linus Torvalds 已提交
1844 1845

	/* Cleans up our, hopefully empty, out_of_order_queue. */
1846
	skb_rbtree_purge(&tp->out_of_order_queue);
L
Linus Torvalds 已提交
1847

1848 1849 1850
#ifdef CONFIG_TCP_MD5SIG
	/* Clean up the MD5 key list, if any */
	if (tp->md5sig_info) {
E
Eric Dumazet 已提交
1851
		tcp_clear_md5_list(sk);
1852
		kfree_rcu(tp->md5sig_info, rcu);
1853 1854 1855
		tp->md5sig_info = NULL;
	}
#endif
C
Chris Leech 已提交
1856

L
Linus Torvalds 已提交
1857 1858 1859 1860
	/* Clean prequeue, it must be empty really */
	__skb_queue_purge(&tp->ucopy.prequeue);

	/* Clean up a referenced TCP bind bucket. */
1861
	if (inet_csk(sk)->icsk_bind_hash)
1862
		inet_put_port(sk);
L
Linus Torvalds 已提交
1863

1864
	BUG_ON(tp->fastopen_rsk);
1865

1866 1867
	/* If socket is aborted during connect operation */
	tcp_free_fastopen_req(tp);
1868
	tcp_saved_syn_free(tp);
1869

E
Eric Dumazet 已提交
1870
	local_bh_disable();
1871
	sk_sockets_allocated_dec(sk);
E
Eric Dumazet 已提交
1872
	local_bh_enable();
L
Linus Torvalds 已提交
1873 1874 1875 1876 1877 1878
}
EXPORT_SYMBOL(tcp_v4_destroy_sock);

#ifdef CONFIG_PROC_FS
/* Proc filesystem TCP sock list dumping. */

1879 1880 1881 1882 1883
/*
 * Get next listener socket follow cur.  If cur is NULL, get first socket
 * starting from bucket given in st->bucket; when st->bucket is zero the
 * very first socket in the hash table is returned.
 */
L
Linus Torvalds 已提交
1884 1885
static void *listening_get_next(struct seq_file *seq, void *cur)
{
J
Jianjun Kong 已提交
1886
	struct tcp_iter_state *st = seq->private;
1887
	struct net *net = seq_file_net(seq);
1888 1889
	struct inet_listen_hashbucket *ilb;
	struct sock *sk = cur;
L
Linus Torvalds 已提交
1890 1891

	if (!sk) {
1892
get_head:
1893
		ilb = &tcp_hashinfo.listening_hash[st->bucket];
1894
		spin_lock(&ilb->lock);
1895
		sk = sk_head(&ilb->head);
1896
		st->offset = 0;
L
Linus Torvalds 已提交
1897 1898
		goto get_sk;
	}
1899
	ilb = &tcp_hashinfo.listening_hash[st->bucket];
L
Linus Torvalds 已提交
1900
	++st->num;
1901
	++st->offset;
L
Linus Torvalds 已提交
1902

1903
	sk = sk_next(sk);
L
Linus Torvalds 已提交
1904
get_sk:
1905
	sk_for_each_from(sk) {
1906 1907
		if (!net_eq(sock_net(sk), net))
			continue;
1908 1909
		if (sk->sk_family == st->family)
			return sk;
L
Linus Torvalds 已提交
1910
	}
1911
	spin_unlock(&ilb->lock);
1912
	st->offset = 0;
1913 1914 1915
	if (++st->bucket < INET_LHTABLE_SIZE)
		goto get_head;
	return NULL;
L
Linus Torvalds 已提交
1916 1917 1918 1919
}

static void *listening_get_idx(struct seq_file *seq, loff_t *pos)
{
1920 1921 1922 1923 1924 1925
	struct tcp_iter_state *st = seq->private;
	void *rc;

	st->bucket = 0;
	st->offset = 0;
	rc = listening_get_next(seq, NULL);
L
Linus Torvalds 已提交
1926 1927 1928 1929 1930 1931 1932 1933

	while (rc && *pos) {
		rc = listening_get_next(seq, rc);
		--*pos;
	}
	return rc;
}

E
Eric Dumazet 已提交
1934
static inline bool empty_bucket(const struct tcp_iter_state *st)
1935
{
E
Eric Dumazet 已提交
1936
	return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain);
1937 1938
}

1939 1940 1941 1942
/*
 * Get first established socket starting from bucket given in st->bucket.
 * If st->bucket is zero, the very first socket in the hash is returned.
 */
L
Linus Torvalds 已提交
1943 1944
static void *established_get_first(struct seq_file *seq)
{
J
Jianjun Kong 已提交
1945
	struct tcp_iter_state *st = seq->private;
1946
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
1947 1948
	void *rc = NULL;

1949 1950
	st->offset = 0;
	for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) {
L
Linus Torvalds 已提交
1951
		struct sock *sk;
1952
		struct hlist_nulls_node *node;
1953
		spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket);
L
Linus Torvalds 已提交
1954

1955 1956 1957 1958
		/* Lockless fast path for the common case of empty buckets */
		if (empty_bucket(st))
			continue;

1959
		spin_lock_bh(lock);
1960
		sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) {
1961
			if (sk->sk_family != st->family ||
1962
			    !net_eq(sock_net(sk), net)) {
L
Linus Torvalds 已提交
1963 1964 1965 1966 1967
				continue;
			}
			rc = sk;
			goto out;
		}
1968
		spin_unlock_bh(lock);
L
Linus Torvalds 已提交
1969 1970 1971 1972 1973 1974 1975 1976
	}
out:
	return rc;
}

static void *established_get_next(struct seq_file *seq, void *cur)
{
	struct sock *sk = cur;
1977
	struct hlist_nulls_node *node;
J
Jianjun Kong 已提交
1978
	struct tcp_iter_state *st = seq->private;
1979
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
1980 1981

	++st->num;
1982
	++st->offset;
L
Linus Torvalds 已提交
1983

E
Eric Dumazet 已提交
1984
	sk = sk_nulls_next(sk);
L
Linus Torvalds 已提交
1985

1986
	sk_nulls_for_each_from(sk, node) {
1987
		if (sk->sk_family == st->family && net_eq(sock_net(sk), net))
E
Eric Dumazet 已提交
1988
			return sk;
L
Linus Torvalds 已提交
1989 1990
	}

E
Eric Dumazet 已提交
1991 1992 1993
	spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
	++st->bucket;
	return established_get_first(seq);
L
Linus Torvalds 已提交
1994 1995 1996 1997
}

static void *established_get_idx(struct seq_file *seq, loff_t pos)
{
1998 1999 2000 2001 2002
	struct tcp_iter_state *st = seq->private;
	void *rc;

	st->bucket = 0;
	rc = established_get_first(seq);
L
Linus Torvalds 已提交
2003 2004 2005 2006

	while (rc && pos) {
		rc = established_get_next(seq, rc);
		--pos;
2007
	}
L
Linus Torvalds 已提交
2008 2009 2010 2011 2012 2013
	return rc;
}

static void *tcp_get_idx(struct seq_file *seq, loff_t pos)
{
	void *rc;
J
Jianjun Kong 已提交
2014
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026

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

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

	return rc;
}

2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044
static void *tcp_seek_last_pos(struct seq_file *seq)
{
	struct tcp_iter_state *st = seq->private;
	int offset = st->offset;
	int orig_num = st->num;
	void *rc = NULL;

	switch (st->state) {
	case TCP_SEQ_STATE_LISTENING:
		if (st->bucket >= INET_LHTABLE_SIZE)
			break;
		st->state = TCP_SEQ_STATE_LISTENING;
		rc = listening_get_next(seq, NULL);
		while (offset-- && rc)
			rc = listening_get_next(seq, rc);
		if (rc)
			break;
		st->bucket = 0;
E
Eric Dumazet 已提交
2045
		st->state = TCP_SEQ_STATE_ESTABLISHED;
2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059
		/* Fallthrough */
	case TCP_SEQ_STATE_ESTABLISHED:
		if (st->bucket > tcp_hashinfo.ehash_mask)
			break;
		rc = established_get_first(seq);
		while (offset-- && rc)
			rc = established_get_next(seq, rc);
	}

	st->num = orig_num;

	return rc;
}

L
Linus Torvalds 已提交
2060 2061
static void *tcp_seq_start(struct seq_file *seq, loff_t *pos)
{
J
Jianjun Kong 已提交
2062
	struct tcp_iter_state *st = seq->private;
2063 2064 2065 2066 2067 2068 2069 2070
	void *rc;

	if (*pos && *pos == st->last_pos) {
		rc = tcp_seek_last_pos(seq);
		if (rc)
			goto out;
	}

L
Linus Torvalds 已提交
2071 2072
	st->state = TCP_SEQ_STATE_LISTENING;
	st->num = 0;
2073 2074 2075 2076 2077 2078 2079
	st->bucket = 0;
	st->offset = 0;
	rc = *pos ? tcp_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;

out:
	st->last_pos = *pos;
	return rc;
L
Linus Torvalds 已提交
2080 2081 2082 2083
}

static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
2084
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096
	void *rc = NULL;

	if (v == SEQ_START_TOKEN) {
		rc = tcp_get_idx(seq, 0);
		goto out;
	}

	switch (st->state) {
	case TCP_SEQ_STATE_LISTENING:
		rc = listening_get_next(seq, v);
		if (!rc) {
			st->state = TCP_SEQ_STATE_ESTABLISHED;
2097 2098
			st->bucket = 0;
			st->offset = 0;
L
Linus Torvalds 已提交
2099 2100 2101 2102 2103 2104 2105 2106 2107
			rc	  = established_get_first(seq);
		}
		break;
	case TCP_SEQ_STATE_ESTABLISHED:
		rc = established_get_next(seq, v);
		break;
	}
out:
	++*pos;
2108
	st->last_pos = *pos;
L
Linus Torvalds 已提交
2109 2110 2111 2112 2113
	return rc;
}

static void tcp_seq_stop(struct seq_file *seq, void *v)
{
J
Jianjun Kong 已提交
2114
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2115 2116 2117 2118

	switch (st->state) {
	case TCP_SEQ_STATE_LISTENING:
		if (v != SEQ_START_TOKEN)
2119
			spin_unlock(&tcp_hashinfo.listening_hash[st->bucket].lock);
L
Linus Torvalds 已提交
2120 2121 2122
		break;
	case TCP_SEQ_STATE_ESTABLISHED:
		if (v)
2123
			spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
L
Linus Torvalds 已提交
2124 2125 2126 2127
		break;
	}
}

2128
int tcp_seq_open(struct inode *inode, struct file *file)
L
Linus Torvalds 已提交
2129
{
A
Al Viro 已提交
2130
	struct tcp_seq_afinfo *afinfo = PDE_DATA(inode);
L
Linus Torvalds 已提交
2131
	struct tcp_iter_state *s;
2132
	int err;
L
Linus Torvalds 已提交
2133

2134 2135 2136 2137
	err = seq_open_net(inode, file, &afinfo->seq_ops,
			  sizeof(struct tcp_iter_state));
	if (err < 0)
		return err;
2138

2139
	s = ((struct seq_file *)file->private_data)->private;
L
Linus Torvalds 已提交
2140
	s->family		= afinfo->family;
S
stephen hemminger 已提交
2141
	s->last_pos		= 0;
2142 2143
	return 0;
}
2144
EXPORT_SYMBOL(tcp_seq_open);
2145

2146
int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo)
L
Linus Torvalds 已提交
2147 2148 2149 2150
{
	int rc = 0;
	struct proc_dir_entry *p;

2151 2152 2153 2154
	afinfo->seq_ops.start		= tcp_seq_start;
	afinfo->seq_ops.next		= tcp_seq_next;
	afinfo->seq_ops.stop		= tcp_seq_stop;

2155
	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
2156
			     afinfo->seq_fops, afinfo);
2157
	if (!p)
L
Linus Torvalds 已提交
2158 2159 2160
		rc = -ENOMEM;
	return rc;
}
E
Eric Dumazet 已提交
2161
EXPORT_SYMBOL(tcp_proc_register);
L
Linus Torvalds 已提交
2162

2163
void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo)
L
Linus Torvalds 已提交
2164
{
2165
	remove_proc_entry(afinfo->name, net->proc_net);
L
Linus Torvalds 已提交
2166
}
E
Eric Dumazet 已提交
2167
EXPORT_SYMBOL(tcp_proc_unregister);
L
Linus Torvalds 已提交
2168

2169
static void get_openreq4(const struct request_sock *req,
E
Eric Dumazet 已提交
2170
			 struct seq_file *f, int i)
L
Linus Torvalds 已提交
2171
{
2172
	const struct inet_request_sock *ireq = inet_rsk(req);
2173
	long delta = req->rsk_timer.expires - jiffies;
L
Linus Torvalds 已提交
2174

2175
	seq_printf(f, "%4d: %08X:%04X %08X:%04X"
2176
		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %u %d %pK",
L
Linus Torvalds 已提交
2177
		i,
2178
		ireq->ir_loc_addr,
2179
		ireq->ir_num,
2180 2181
		ireq->ir_rmt_addr,
		ntohs(ireq->ir_rmt_port),
L
Linus Torvalds 已提交
2182 2183 2184
		TCP_SYN_RECV,
		0, 0, /* could print option size, but that is af dependent. */
		1,    /* timers active (only the expire timer) */
2185
		jiffies_delta_to_clock_t(delta),
2186
		req->num_timeout,
E
Eric Dumazet 已提交
2187 2188
		from_kuid_munged(seq_user_ns(f),
				 sock_i_uid(req->rsk_listener)),
L
Linus Torvalds 已提交
2189 2190
		0,  /* non standard timer */
		0, /* open_requests have no inode */
2191
		0,
2192
		req);
L
Linus Torvalds 已提交
2193 2194
}

2195
static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i)
L
Linus Torvalds 已提交
2196 2197 2198
{
	int timer_active;
	unsigned long timer_expires;
2199
	const struct tcp_sock *tp = tcp_sk(sk);
2200
	const struct inet_connection_sock *icsk = inet_csk(sk);
2201
	const struct inet_sock *inet = inet_sk(sk);
2202
	const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq;
E
Eric Dumazet 已提交
2203 2204 2205 2206
	__be32 dest = inet->inet_daddr;
	__be32 src = inet->inet_rcv_saddr;
	__u16 destp = ntohs(inet->inet_dport);
	__u16 srcp = ntohs(inet->inet_sport);
2207
	int rx_queue;
2208
	int state;
L
Linus Torvalds 已提交
2209

N
Nandita Dukkipati 已提交
2210 2211 2212
	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
	    icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
L
Linus Torvalds 已提交
2213
		timer_active	= 1;
2214 2215
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
2216
		timer_active	= 4;
2217
		timer_expires	= icsk->icsk_timeout;
2218
	} else if (timer_pending(&sk->sk_timer)) {
L
Linus Torvalds 已提交
2219
		timer_active	= 2;
2220
		timer_expires	= sk->sk_timer.expires;
L
Linus Torvalds 已提交
2221 2222 2223 2224 2225
	} else {
		timer_active	= 0;
		timer_expires = jiffies;
	}

2226 2227
	state = sk_state_load(sk);
	if (state == TCP_LISTEN)
2228 2229
		rx_queue = sk->sk_ack_backlog;
	else
2230 2231
		/* Because we don't lock the socket,
		 * we might find a transient negative value.
2232 2233 2234
		 */
		rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0);

2235
	seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX "
2236
			"%08X %5u %8d %lu %d %pK %lu %lu %u %u %d",
2237
		i, src, srcp, dest, destp, state,
2238
		tp->write_seq - tp->snd_una,
2239
		rx_queue,
L
Linus Torvalds 已提交
2240
		timer_active,
2241
		jiffies_delta_to_clock_t(timer_expires - jiffies),
2242
		icsk->icsk_retransmits,
2243
		from_kuid_munged(seq_user_ns(f), sock_i_uid(sk)),
2244
		icsk->icsk_probes_out,
2245 2246
		sock_i_ino(sk),
		atomic_read(&sk->sk_refcnt), sk,
2247 2248
		jiffies_to_clock_t(icsk->icsk_rto),
		jiffies_to_clock_t(icsk->icsk_ack.ato),
2249
		(icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
L
Linus Torvalds 已提交
2250
		tp->snd_cwnd,
2251 2252
		state == TCP_LISTEN ?
		    fastopenq->max_qlen :
2253
		    (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh));
L
Linus Torvalds 已提交
2254 2255
}

2256
static void get_timewait4_sock(const struct inet_timewait_sock *tw,
2257
			       struct seq_file *f, int i)
L
Linus Torvalds 已提交
2258
{
2259
	long delta = tw->tw_timer.expires - jiffies;
2260
	__be32 dest, src;
L
Linus Torvalds 已提交
2261 2262 2263 2264 2265 2266 2267
	__u16 destp, srcp;

	dest  = tw->tw_daddr;
	src   = tw->tw_rcv_saddr;
	destp = ntohs(tw->tw_dport);
	srcp  = ntohs(tw->tw_sport);

2268
	seq_printf(f, "%4d: %08X:%04X %08X:%04X"
2269
		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK",
L
Linus Torvalds 已提交
2270
		i, src, srcp, dest, destp, tw->tw_substate, 0, 0,
2271
		3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
2272
		atomic_read(&tw->tw_refcnt), tw);
L
Linus Torvalds 已提交
2273 2274 2275 2276 2277 2278
}

#define TMPSZ 150

static int tcp4_seq_show(struct seq_file *seq, void *v)
{
J
Jianjun Kong 已提交
2279
	struct tcp_iter_state *st;
E
Eric Dumazet 已提交
2280
	struct sock *sk = v;
L
Linus Torvalds 已提交
2281

2282
	seq_setwidth(seq, TMPSZ - 1);
L
Linus Torvalds 已提交
2283
	if (v == SEQ_START_TOKEN) {
2284
		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
L
Linus Torvalds 已提交
2285 2286 2287 2288 2289 2290
			   "rx_queue tr tm->when retrnsmt   uid  timeout "
			   "inode");
		goto out;
	}
	st = seq->private;

2291 2292 2293
	if (sk->sk_state == TCP_TIME_WAIT)
		get_timewait4_sock(v, seq, st->num);
	else if (sk->sk_state == TCP_NEW_SYN_RECV)
E
Eric Dumazet 已提交
2294
		get_openreq4(v, seq, st->num);
2295 2296
	else
		get_tcp4_sock(v, seq, st->num);
L
Linus Torvalds 已提交
2297
out:
2298
	seq_pad(seq, '\n');
L
Linus Torvalds 已提交
2299 2300 2301
	return 0;
}

2302 2303 2304 2305 2306 2307 2308 2309
static const struct file_operations tcp_afinfo_seq_fops = {
	.owner   = THIS_MODULE,
	.open    = tcp_seq_open,
	.read    = seq_read,
	.llseek  = seq_lseek,
	.release = seq_release_net
};

L
Linus Torvalds 已提交
2310 2311 2312
static struct tcp_seq_afinfo tcp4_seq_afinfo = {
	.name		= "tcp",
	.family		= AF_INET,
2313
	.seq_fops	= &tcp_afinfo_seq_fops,
2314 2315 2316
	.seq_ops	= {
		.show		= tcp4_seq_show,
	},
L
Linus Torvalds 已提交
2317 2318
};

2319
static int __net_init tcp4_proc_init_net(struct net *net)
2320 2321 2322 2323
{
	return tcp_proc_register(net, &tcp4_seq_afinfo);
}

2324
static void __net_exit tcp4_proc_exit_net(struct net *net)
2325 2326 2327 2328 2329 2330 2331 2332 2333
{
	tcp_proc_unregister(net, &tcp4_seq_afinfo);
}

static struct pernet_operations tcp4_net_ops = {
	.init = tcp4_proc_init_net,
	.exit = tcp4_proc_exit_net,
};

L
Linus Torvalds 已提交
2334 2335
int __init tcp4_proc_init(void)
{
2336
	return register_pernet_subsys(&tcp4_net_ops);
L
Linus Torvalds 已提交
2337 2338 2339 2340
}

void tcp4_proc_exit(void)
{
2341
	unregister_pernet_subsys(&tcp4_net_ops);
L
Linus Torvalds 已提交
2342 2343 2344 2345 2346 2347 2348 2349 2350
}
#endif /* CONFIG_PROC_FS */

struct proto tcp_prot = {
	.name			= "TCP",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v4_connect,
	.disconnect		= tcp_disconnect,
2351
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
2352 2353 2354 2355 2356 2357 2358
	.ioctl			= tcp_ioctl,
	.init			= tcp_v4_init_sock,
	.destroy		= tcp_v4_destroy_sock,
	.shutdown		= tcp_shutdown,
	.setsockopt		= tcp_setsockopt,
	.getsockopt		= tcp_getsockopt,
	.recvmsg		= tcp_recvmsg,
2359 2360
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
2361
	.backlog_rcv		= tcp_v4_do_rcv,
E
Eric Dumazet 已提交
2362
	.release_cb		= tcp_release_cb,
2363 2364 2365
	.hash			= inet_hash,
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
2366
	.enter_memory_pressure	= tcp_enter_memory_pressure,
2367
	.stream_memory_free	= tcp_stream_memory_free,
L
Linus Torvalds 已提交
2368
	.sockets_allocated	= &tcp_sockets_allocated,
2369
	.orphan_count		= &tcp_orphan_count,
L
Linus Torvalds 已提交
2370 2371
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
2372
	.sysctl_mem		= sysctl_tcp_mem,
L
Linus Torvalds 已提交
2373 2374 2375 2376
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp_sock),
2377
	.slab_flags		= SLAB_DESTROY_BY_RCU,
2378
	.twsk_prot		= &tcp_timewait_sock_ops,
2379
	.rsk_prot		= &tcp_request_sock_ops,
2380
	.h.hashinfo		= &tcp_hashinfo,
2381
	.no_autobind		= true,
2382 2383 2384
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
G
Glauber Costa 已提交
2385
#endif
2386
	.diag_destroy		= tcp_abort,
L
Linus Torvalds 已提交
2387
};
E
Eric Dumazet 已提交
2388
EXPORT_SYMBOL(tcp_prot);
L
Linus Torvalds 已提交
2389

2390 2391 2392 2393 2394 2395 2396 2397 2398
static void __net_exit tcp_sk_exit(struct net *net)
{
	int cpu;

	for_each_possible_cpu(cpu)
		inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.tcp_sk, cpu));
	free_percpu(net->ipv4.tcp_sk);
}

2399 2400
static int __net_init tcp_sk_init(struct net *net)
{
2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413
	int res, cpu;

	net->ipv4.tcp_sk = alloc_percpu(struct sock *);
	if (!net->ipv4.tcp_sk)
		return -ENOMEM;

	for_each_possible_cpu(cpu) {
		struct sock *sk;

		res = inet_ctl_sock_create(&sk, PF_INET, SOCK_RAW,
					   IPPROTO_TCP, net);
		if (res)
			goto fail;
2414
		sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
2415 2416
		*per_cpu_ptr(net->ipv4.tcp_sk, cpu) = sk;
	}
2417

2418
	net->ipv4.sysctl_tcp_ecn = 2;
2419 2420
	net->ipv4.sysctl_tcp_ecn_fallback = 1;

F
Fan Du 已提交
2421
	net->ipv4.sysctl_tcp_base_mss = TCP_BASE_MSS;
2422
	net->ipv4.sysctl_tcp_probe_threshold = TCP_PROBE_THRESHOLD;
2423
	net->ipv4.sysctl_tcp_probe_interval = TCP_PROBE_INTERVAL;
2424

2425
	net->ipv4.sysctl_tcp_keepalive_time = TCP_KEEPALIVE_TIME;
2426
	net->ipv4.sysctl_tcp_keepalive_probes = TCP_KEEPALIVE_PROBES;
2427
	net->ipv4.sysctl_tcp_keepalive_intvl = TCP_KEEPALIVE_INTVL;
2428

2429
	net->ipv4.sysctl_tcp_syn_retries = TCP_SYN_RETRIES;
2430
	net->ipv4.sysctl_tcp_synack_retries = TCP_SYNACK_RETRIES;
2431
	net->ipv4.sysctl_tcp_syncookies = 1;
2432
	net->ipv4.sysctl_tcp_reordering = TCP_FASTRETRANS_THRESH;
2433
	net->ipv4.sysctl_tcp_retries1 = TCP_RETR1;
2434
	net->ipv4.sysctl_tcp_retries2 = TCP_RETR2;
2435
	net->ipv4.sysctl_tcp_orphan_retries = 0;
2436
	net->ipv4.sysctl_tcp_fin_timeout = TCP_FIN_TIMEOUT;
2437
	net->ipv4.sysctl_tcp_notsent_lowat = UINT_MAX;
2438

2439
	return 0;
2440 2441 2442 2443
fail:
	tcp_sk_exit(net);

	return res;
E
Eric W. Biederman 已提交
2444 2445 2446 2447 2448
}

static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list)
{
	inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET);
2449 2450 2451
}

static struct pernet_operations __net_initdata tcp_sk_ops = {
E
Eric W. Biederman 已提交
2452 2453 2454
       .init	   = tcp_sk_init,
       .exit	   = tcp_sk_exit,
       .exit_batch = tcp_sk_exit_batch,
2455 2456
};

2457
void __init tcp_v4_init(void)
L
Linus Torvalds 已提交
2458
{
2459
	inet_hashinfo_init(&tcp_hashinfo);
2460
	if (register_pernet_subsys(&tcp_sk_ops))
L
Linus Torvalds 已提交
2461 2462
		panic("Failed to create the TCP control socket.\n");
}