tcp_ipv4.c 62.8 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;
E
Eric Dumazet 已提交
89
EXPORT_SYMBOL(sysctl_tcp_low_latency);
L
Linus Torvalds 已提交
90

91
#ifdef CONFIG_TCP_MD5SIG
E
Eric Dumazet 已提交
92
static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
93
			       __be32 daddr, __be32 saddr, const struct tcphdr *th);
94 95
#endif

96
struct inet_hashinfo tcp_hashinfo;
E
Eric Dumazet 已提交
97
EXPORT_SYMBOL(tcp_hashinfo);
L
Linus Torvalds 已提交
98

99
static  __u32 tcp_v4_init_sequence(const struct sk_buff *skb)
L
Linus Torvalds 已提交
100
{
101 102
	return secure_tcp_sequence_number(ip_hdr(skb)->daddr,
					  ip_hdr(skb)->saddr,
103 104
					  tcp_hdr(skb)->dest,
					  tcp_hdr(skb)->source);
L
Linus Torvalds 已提交
105 106
}

107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123
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 &&
124
	    (!twp || (sysctl_tcp_tw_reuse &&
125
			     get_seconds() - tcptw->tw_ts_recent_stamp > 1))) {
126 127 128 129 130 131 132 133 134 135 136 137 138
		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
139 140 141
/* This will initiate an outgoing connection. */
int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
{
142
	struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
L
Linus Torvalds 已提交
143 144
	struct inet_sock *inet = inet_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
145
	__be16 orig_sport, orig_dport;
146
	__be32 daddr, nexthop;
147
	struct flowi4 *fl4;
148
	struct rtable *rt;
L
Linus Torvalds 已提交
149
	int err;
150
	struct ip_options_rcu *inet_opt;
L
Linus Torvalds 已提交
151 152 153 154 155 156 157 158

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

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

	nexthop = daddr = usin->sin_addr.s_addr;
159
	inet_opt = rcu_dereference_protected(inet->inet_opt,
160
					     lockdep_sock_is_held(sk));
161
	if (inet_opt && inet_opt->opt.srr) {
L
Linus Torvalds 已提交
162 163
		if (!daddr)
			return -EINVAL;
164
		nexthop = inet_opt->opt.faddr;
L
Linus Torvalds 已提交
165 166
	}

167 168
	orig_sport = inet->inet_sport;
	orig_dport = usin->sin_port;
169 170
	fl4 = &inet->cork.fl.u.ip4;
	rt = ip_route_connect(fl4, nexthop, inet->inet_saddr,
171 172
			      RT_CONN_FLAGS(sk), sk->sk_bound_dev_if,
			      IPPROTO_TCP,
173
			      orig_sport, orig_dport, sk);
174 175 176
	if (IS_ERR(rt)) {
		err = PTR_ERR(rt);
		if (err == -ENETUNREACH)
177
			IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
178
		return err;
179
	}
L
Linus Torvalds 已提交
180 181 182 183 184 185

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

186
	if (!inet_opt || !inet_opt->opt.srr)
187
		daddr = fl4->daddr;
L
Linus Torvalds 已提交
188

E
Eric Dumazet 已提交
189
	if (!inet->inet_saddr)
190
		inet->inet_saddr = fl4->saddr;
191
	sk_rcv_saddr_set(sk, inet->inet_saddr);
L
Linus Torvalds 已提交
192

E
Eric Dumazet 已提交
193
	if (tp->rx_opt.ts_recent_stamp && inet->inet_daddr != daddr) {
L
Linus Torvalds 已提交
194 195 196
		/* Reset inherited state */
		tp->rx_opt.ts_recent	   = 0;
		tp->rx_opt.ts_recent_stamp = 0;
P
Pavel Emelyanov 已提交
197 198
		if (likely(!tp->repair))
			tp->write_seq	   = 0;
L
Linus Torvalds 已提交
199 200
	}

201
	if (tcp_death_row.sysctl_tw_recycle &&
202 203
	    !tp->rx_opt.ts_recent_stamp && fl4->daddr == daddr)
		tcp_fetch_timewait_stamp(sk, &rt->dst);
L
Linus Torvalds 已提交
204

E
Eric Dumazet 已提交
205
	inet->inet_dport = usin->sin_port;
206
	sk_daddr_set(sk, daddr);
L
Linus Torvalds 已提交
207

208
	inet_csk(sk)->icsk_ext_hdr_len = 0;
209 210
	if (inet_opt)
		inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
L
Linus Torvalds 已提交
211

212
	tp->rx_opt.mss_clamp = TCP_MSS_DEFAULT;
L
Linus Torvalds 已提交
213 214 215 216 217 218 219

	/* 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);
220
	err = inet_hash_connect(&tcp_death_row, sk);
L
Linus Torvalds 已提交
221 222 223
	if (err)
		goto failure;

224
	sk_set_txhash(sk);
225

226
	rt = ip_route_newports(fl4, rt, orig_sport, orig_dport,
227 228 229 230
			       inet->inet_sport, inet->inet_dport, sk);
	if (IS_ERR(rt)) {
		err = PTR_ERR(rt);
		rt = NULL;
L
Linus Torvalds 已提交
231
		goto failure;
232
	}
L
Linus Torvalds 已提交
233
	/* OK, now commit destination to socket.  */
234
	sk->sk_gso_type = SKB_GSO_TCPV4;
235
	sk_setup_caps(sk, &rt->dst);
L
Linus Torvalds 已提交
236

P
Pavel Emelyanov 已提交
237
	if (!tp->write_seq && likely(!tp->repair))
E
Eric Dumazet 已提交
238 239 240
		tp->write_seq = secure_tcp_sequence_number(inet->inet_saddr,
							   inet->inet_daddr,
							   inet->inet_sport,
L
Linus Torvalds 已提交
241 242
							   usin->sin_port);

E
Eric Dumazet 已提交
243
	inet->inet_id = tp->write_seq ^ jiffies;
L
Linus Torvalds 已提交
244

A
Andrey Vagin 已提交
245
	err = tcp_connect(sk);
P
Pavel Emelyanov 已提交
246

L
Linus Torvalds 已提交
247 248 249 250 251 252 253
	rt = NULL;
	if (err)
		goto failure;

	return 0;

failure:
254 255 256 257
	/*
	 * This unhashes the socket and releases the local port,
	 * if necessary.
	 */
L
Linus Torvalds 已提交
258 259 260
	tcp_set_state(sk, TCP_CLOSE);
	ip_rt_put(rt);
	sk->sk_route_caps = 0;
E
Eric Dumazet 已提交
261
	inet->inet_dport = 0;
L
Linus Torvalds 已提交
262 263
	return err;
}
E
Eric Dumazet 已提交
264
EXPORT_SYMBOL(tcp_v4_connect);
L
Linus Torvalds 已提交
265 266

/*
267 268 269
 * 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
270
 */
271
void tcp_v4_mtu_reduced(struct sock *sk)
L
Linus Torvalds 已提交
272 273 274
{
	struct dst_entry *dst;
	struct inet_sock *inet = inet_sk(sk);
275
	u32 mtu = tcp_sk(sk)->mtu_info;
L
Linus Torvalds 已提交
276

277 278
	dst = inet_csk_update_pmtu(sk, mtu);
	if (!dst)
L
Linus Torvalds 已提交
279 280 281 282 283 284 285 286 287 288 289
		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 &&
290
	    ip_sk_accept_pmtu(sk) &&
291
	    inet_csk(sk)->icsk_pmtu_cookie > mtu) {
L
Linus Torvalds 已提交
292 293 294 295 296 297 298 299 300 301
		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 */
}
302
EXPORT_SYMBOL(tcp_v4_mtu_reduced);
L
Linus Torvalds 已提交
303

304 305 306 307
static void do_redirect(struct sk_buff *skb, struct sock *sk)
{
	struct dst_entry *dst = __sk_dst_check(sk, 0);

308
	if (dst)
309
		dst->ops->redirect(dst, sk, skb);
310 311
}

312 313

/* handle ICMP messages on TCP_NEW_SYN_RECV request sockets */
314
void tcp_req_err(struct sock *sk, u32 seq, bool abort)
315 316 317 318 319 320 321 322
{
	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) {
323
		__NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
324
	} else if (abort) {
325 326 327 328 329 330
		/*
		 * 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
331
		inet_csk_reqsk_queue_drop(req->rsk_listener, req);
332
		tcp_listendrop(req->rsk_listener);
333
	}
334
	reqsk_put(req);
335 336 337
}
EXPORT_SYMBOL(tcp_req_err);

L
Linus Torvalds 已提交
338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
/*
 * 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.
 *
 */

354
void tcp_v4_err(struct sk_buff *icmp_skb, u32 info)
L
Linus Torvalds 已提交
355
{
356
	const struct iphdr *iph = (const struct iphdr *)icmp_skb->data;
357
	struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2));
358
	struct inet_connection_sock *icsk;
L
Linus Torvalds 已提交
359 360
	struct tcp_sock *tp;
	struct inet_sock *inet;
361 362
	const int type = icmp_hdr(icmp_skb)->type;
	const int code = icmp_hdr(icmp_skb)->code;
L
Linus Torvalds 已提交
363
	struct sock *sk;
364
	struct sk_buff *skb;
365 366
	struct request_sock *fastopen;
	__u32 seq, snd_una;
367
	__u32 remaining;
L
Linus Torvalds 已提交
368
	int err;
369
	struct net *net = dev_net(icmp_skb->dev);
L
Linus Torvalds 已提交
370

371 372 373
	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
374
	if (!sk) {
E
Eric Dumazet 已提交
375
		__ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
L
Linus Torvalds 已提交
376 377 378
		return;
	}
	if (sk->sk_state == TCP_TIME_WAIT) {
379
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
380 381
		return;
	}
382 383
	seq = ntohl(th->seq);
	if (sk->sk_state == TCP_NEW_SYN_RECV)
384 385 386 387 388 389
		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
390 391 392 393

	bh_lock_sock(sk);
	/* If too many ICMPs get dropped on busy
	 * servers this needs to be solved differently.
394 395
	 * 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
396
	 */
397 398
	if (sock_owned_by_user(sk)) {
		if (!(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED))
399
			__NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
400
	}
L
Linus Torvalds 已提交
401 402 403
	if (sk->sk_state == TCP_CLOSE)
		goto out;

404
	if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
405
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
406 407 408
		goto out;
	}

409
	icsk = inet_csk(sk);
L
Linus Torvalds 已提交
410
	tp = tcp_sk(sk);
411 412 413
	/* 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
414
	if (sk->sk_state != TCP_LISTEN &&
415
	    !between(seq, snd_una, tp->snd_nxt)) {
416
		__NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
417 418 419 420
		goto out;
	}

	switch (type) {
421 422 423
	case ICMP_REDIRECT:
		do_redirect(icmp_skb, sk);
		goto out;
L
Linus Torvalds 已提交
424 425 426 427 428 429 430 431 432 433 434
	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) */
435 436 437 438 439 440 441
			/* 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;

442
			tp->mtu_info = info;
443
			if (!sock_owned_by_user(sk)) {
444
				tcp_v4_mtu_reduced(sk);
445 446 447 448
			} else {
				if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, &tp->tsq_flags))
					sock_hold(sk);
			}
L
Linus Torvalds 已提交
449 450 451 452
			goto out;
		}

		err = icmp_err_convert[code].errno;
453 454 455 456 457
		/* 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 ||
458
		    !icsk->icsk_backoff || fastopen)
459 460
			break;

461 462 463
		if (sock_owned_by_user(sk))
			break;

464
		icsk->icsk_backoff--;
465 466 467
		icsk->icsk_rto = tp->srtt_us ? __tcp_set_rto(tp) :
					       TCP_TIMEOUT_INIT;
		icsk->icsk_rto = inet_csk_rto_backoff(icsk, TCP_RTO_MAX);
468 469 470 471

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

E
Eric Dumazet 已提交
472 473 474
		remaining = icsk->icsk_rto -
			    min(icsk->icsk_rto,
				tcp_time_stamp - tcp_skb_timestamp(skb));
475 476 477 478 479 480 481 482 483 484

		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
485 486 487 488 489 490 491 492 493 494
		break;
	case ICMP_TIME_EXCEEDED:
		err = EHOSTUNREACH;
		break;
	default:
		goto out;
	}

	switch (sk->sk_state) {
	case TCP_SYN_SENT:
495 496 497 498
	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.
		 */
499
		if (fastopen && !fastopen->sk)
500 501
			break;

L
Linus Torvalds 已提交
502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542
		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);
}

543
void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr)
L
Linus Torvalds 已提交
544
{
545
	struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
546

547
	if (skb->ip_summed == CHECKSUM_PARTIAL) {
548
		th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0);
549
		skb->csum_start = skb_transport_header(skb) - skb->head;
A
Al Viro 已提交
550
		skb->csum_offset = offsetof(struct tcphdr, check);
L
Linus Torvalds 已提交
551
	} else {
552
		th->check = tcp_v4_check(skb->len, saddr, daddr,
553
					 csum_partial(th,
L
Linus Torvalds 已提交
554 555 556 557 558
						      th->doff << 2,
						      skb->csum));
	}
}

559
/* This routine computes an IPv4 TCP checksum. */
560
void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb)
561
{
562
	const struct inet_sock *inet = inet_sk(sk);
563 564 565

	__tcp_v4_send_check(skb, inet->inet_saddr, inet->inet_daddr);
}
E
Eric Dumazet 已提交
566
EXPORT_SYMBOL(tcp_v4_send_check);
567

L
Linus Torvalds 已提交
568 569 570 571 572 573 574 575 576 577 578 579 580
/*
 *	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.
 */

581
static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
582
{
583
	const struct tcphdr *th = tcp_hdr(skb);
584 585 586
	struct {
		struct tcphdr th;
#ifdef CONFIG_TCP_MD5SIG
587
		__be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)];
588 589
#endif
	} rep;
L
Linus Torvalds 已提交
590
	struct ip_reply_arg arg;
591
#ifdef CONFIG_TCP_MD5SIG
592
	struct tcp_md5sig_key *key = NULL;
593 594 595 596
	const __u8 *hash_location = NULL;
	unsigned char newhash[16];
	int genhash;
	struct sock *sk1 = NULL;
597
#endif
598
	struct net *net;
L
Linus Torvalds 已提交
599 600 601 602 603

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

604 605 606 607
	/* 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
608 609 610
		return;

	/* Swap the send and the receive. */
611 612 613 614 615
	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
616 617

	if (th->ack) {
618
		rep.th.seq = th->ack_seq;
L
Linus Torvalds 已提交
619
	} else {
620 621 622
		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
623 624
	}

625
	memset(&arg, 0, sizeof(arg));
626 627 628
	arg.iov[0].iov_base = (unsigned char *)&rep;
	arg.iov[0].iov_len  = sizeof(rep.th);

629
	net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
630
#ifdef CONFIG_TCP_MD5SIG
631
	rcu_read_lock();
632
	hash_location = tcp_parse_md5sig_option(th);
633
	if (sk && sk_fullsock(sk)) {
634 635 636
		key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)
					&ip_hdr(skb)->saddr, AF_INET);
	} else if (hash_location) {
637 638 639 640 641 642 643
		/*
		 * 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.
		 */
644 645
		sk1 = __inet_lookup_listener(net, &tcp_hashinfo, NULL, 0,
					     ip_hdr(skb)->saddr,
646
					     th->source, ip_hdr(skb)->daddr,
647 648 649
					     ntohs(th->source), inet_iif(skb));
		/* don't send rst if it can't find key */
		if (!sk1)
650 651
			goto out;

652 653 654
		key = tcp_md5_do_lookup(sk1, (union tcp_md5_addr *)
					&ip_hdr(skb)->saddr, AF_INET);
		if (!key)
655 656
			goto out;

657

658
		genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, skb);
659
		if (genhash || memcmp(hash_location, newhash, 16) != 0)
660 661
			goto out;

662 663
	}

664 665 666 667 668 669 670 671 672
	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;

673
		tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1],
674 675
				     key, ip_hdr(skb)->saddr,
				     ip_hdr(skb)->daddr, &rep.th);
676 677
	}
#endif
678 679
	arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
				      ip_hdr(skb)->saddr, /* XXX */
680
				      arg.iov[0].iov_len, IPPROTO_TCP, 0);
L
Linus Torvalds 已提交
681
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
682 683
	arg.flags = (sk && inet_sk_transparent(sk)) ? IP_REPLY_ARG_NOSRCCHECK : 0;

684
	/* When socket is gone, all binding information is lost.
A
Alexey Kuznetsov 已提交
685 686
	 * routing might fail in this case. No choice here, if we choose to force
	 * input interface, we will misroute in case of asymmetric route.
687
	 */
A
Alexey Kuznetsov 已提交
688 689
	if (sk)
		arg.bound_dev_if = sk->sk_bound_dev_if;
L
Linus Torvalds 已提交
690

691 692 693
	BUILD_BUG_ON(offsetof(struct sock, sk_bound_dev_if) !=
		     offsetof(struct inet_timewait_sock, tw_bound_dev_if));

694
	arg.tos = ip_hdr(skb)->tos;
695
	local_bh_disable();
696 697
	ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk),
			      skb, &TCP_SKB_CB(skb)->header.h4.opt,
698 699
			      ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
			      &arg, arg.iov[0].iov_len);
L
Linus Torvalds 已提交
700

E
Eric Dumazet 已提交
701 702
	__TCP_INC_STATS(net, TCP_MIB_OUTSEGS);
	__TCP_INC_STATS(net, TCP_MIB_OUTRSTS);
703
	local_bh_enable();
704 705

#ifdef CONFIG_TCP_MD5SIG
706 707
out:
	rcu_read_unlock();
708
#endif
L
Linus Torvalds 已提交
709 710 711 712 713 714
}

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

715 716
static void tcp_v4_send_ack(struct net *net,
			    struct sk_buff *skb, u32 seq, u32 ack,
717
			    u32 win, u32 tsval, u32 tsecr, int oif,
718
			    struct tcp_md5sig_key *key,
719
			    int reply_flags, u8 tos)
L
Linus Torvalds 已提交
720
{
721
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
722 723
	struct {
		struct tcphdr th;
724
		__be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2)
725
#ifdef CONFIG_TCP_MD5SIG
726
			   + (TCPOLEN_MD5SIG_ALIGNED >> 2)
727 728
#endif
			];
L
Linus Torvalds 已提交
729 730 731 732
	} rep;
	struct ip_reply_arg arg;

	memset(&rep.th, 0, sizeof(struct tcphdr));
733
	memset(&arg, 0, sizeof(arg));
L
Linus Torvalds 已提交
734 735 736

	arg.iov[0].iov_base = (unsigned char *)&rep;
	arg.iov[0].iov_len  = sizeof(rep.th);
737
	if (tsecr) {
738 739 740
		rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				   (TCPOPT_TIMESTAMP << 8) |
				   TCPOLEN_TIMESTAMP);
741 742
		rep.opt[1] = htonl(tsval);
		rep.opt[2] = htonl(tsecr);
743
		arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED;
L
Linus Torvalds 已提交
744 745 746 747 748 749 750 751 752 753 754
	}

	/* 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);

755 756
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
757
		int offset = (tsecr) ? 3 : 0;
758 759 760 761 762 763 764 765

		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;

766
		tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset],
767 768
				    key, ip_hdr(skb)->saddr,
				    ip_hdr(skb)->daddr, &rep.th);
769 770
	}
#endif
771
	arg.flags = reply_flags;
772 773
	arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
				      ip_hdr(skb)->saddr, /* XXX */
L
Linus Torvalds 已提交
774 775
				      arg.iov[0].iov_len, IPPROTO_TCP, 0);
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
776 777
	if (oif)
		arg.bound_dev_if = oif;
778
	arg.tos = tos;
779
	local_bh_disable();
780 781
	ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk),
			      skb, &TCP_SKB_CB(skb)->header.h4.opt,
782 783
			      ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
			      &arg, arg.iov[0].iov_len);
L
Linus Torvalds 已提交
784

E
Eric Dumazet 已提交
785
	__TCP_INC_STATS(net, TCP_MIB_OUTSEGS);
786
	local_bh_enable();
L
Linus Torvalds 已提交
787 788 789 790
}

static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
791
	struct inet_timewait_sock *tw = inet_twsk(sk);
792
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
793

794 795
	tcp_v4_send_ack(sock_net(sk), skb,
			tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
796
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
797
			tcp_time_stamp + tcptw->tw_ts_offset,
798 799
			tcptw->tw_ts_recent,
			tw->tw_bound_dev_if,
800
			tcp_twsk_md5_key(tcptw),
801 802
			tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0,
			tw->tw_tos
803
			);
L
Linus Torvalds 已提交
804

805
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
806 807
}

808
static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
809
				  struct request_sock *req)
L
Linus Torvalds 已提交
810
{
811 812 813
	/* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
	 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
	 */
814 815 816
	u32 seq = (sk->sk_state == TCP_LISTEN) ? tcp_rsk(req)->snt_isn + 1 :
					     tcp_sk(sk)->snd_nxt;

817 818 819 820 821
	/* 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:
	 */
822
	tcp_v4_send_ack(sock_net(sk), skb, seq,
823 824
			tcp_rsk(req)->rcv_nxt,
			req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale,
825
			tcp_time_stamp,
826 827
			req->ts_recent,
			0,
E
Eric Dumazet 已提交
828 829
			tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&ip_hdr(skb)->daddr,
					  AF_INET),
830 831
			inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0,
			ip_hdr(skb)->tos);
L
Linus Torvalds 已提交
832 833 834
}

/*
835
 *	Send a SYN-ACK after having received a SYN.
836
 *	This still operates on a request_sock only, not on a big
L
Linus Torvalds 已提交
837 838
 *	socket.
 */
839
static int tcp_v4_send_synack(const struct sock *sk, struct dst_entry *dst,
840
			      struct flowi *fl,
841
			      struct request_sock *req,
842
			      struct tcp_fastopen_cookie *foc,
843
			      enum tcp_synack_type synack_type)
L
Linus Torvalds 已提交
844
{
845
	const struct inet_request_sock *ireq = inet_rsk(req);
846
	struct flowi4 fl4;
L
Linus Torvalds 已提交
847
	int err = -1;
848
	struct sk_buff *skb;
L
Linus Torvalds 已提交
849 850

	/* First, grab a route. */
851
	if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL)
852
		return -1;
L
Linus Torvalds 已提交
853

854
	skb = tcp_make_synack(sk, dst, req, foc, synack_type);
L
Linus Torvalds 已提交
855 856

	if (skb) {
857
		__tcp_v4_send_check(skb, ireq->ir_loc_addr, ireq->ir_rmt_addr);
L
Linus Torvalds 已提交
858

859 860
		err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr,
					    ireq->ir_rmt_addr,
861
					    ireq->opt);
862
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
863 864 865 866 867 868
	}

	return err;
}

/*
869
 *	IPv4 request_sock destructor.
L
Linus Torvalds 已提交
870
 */
871
static void tcp_v4_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
872
{
J
Jesper Juhl 已提交
873
	kfree(inet_rsk(req)->opt);
L
Linus Torvalds 已提交
874 875
}

876 877 878 879 880 881 882 883
#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.  */
884
struct tcp_md5sig_key *tcp_md5_do_lookup(const struct sock *sk,
E
Eric Dumazet 已提交
885 886
					 const union tcp_md5_addr *addr,
					 int family)
887
{
888
	const struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
889 890
	struct tcp_md5sig_key *key;
	unsigned int size = sizeof(struct in_addr);
891
	const struct tcp_md5sig_info *md5sig;
892

893 894
	/* caller either holds rcu_read_lock() or socket lock */
	md5sig = rcu_dereference_check(tp->md5sig_info,
895
				       lockdep_sock_is_held(sk));
896
	if (!md5sig)
897
		return NULL;
E
Eric Dumazet 已提交
898 899 900 901
#if IS_ENABLED(CONFIG_IPV6)
	if (family == AF_INET6)
		size = sizeof(struct in6_addr);
#endif
902
	hlist_for_each_entry_rcu(key, &md5sig->head, node) {
E
Eric Dumazet 已提交
903 904 905 906
		if (key->family != family)
			continue;
		if (!memcmp(&key->addr, addr, size))
			return key;
907 908 909
	}
	return NULL;
}
E
Eric Dumazet 已提交
910
EXPORT_SYMBOL(tcp_md5_do_lookup);
911

912
struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk,
913
					 const struct sock *addr_sk)
914
{
915
	const union tcp_md5_addr *addr;
E
Eric Dumazet 已提交
916

917
	addr = (const union tcp_md5_addr *)&addr_sk->sk_daddr;
E
Eric Dumazet 已提交
918
	return tcp_md5_do_lookup(sk, addr, AF_INET);
919 920 921 922
}
EXPORT_SYMBOL(tcp_v4_md5_lookup);

/* This can be called on a newly created socket, from other files */
E
Eric Dumazet 已提交
923 924
int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
		   int family, const u8 *newkey, u8 newkeylen, gfp_t gfp)
925 926
{
	/* Add Key to the list */
927
	struct tcp_md5sig_key *key;
928
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
929
	struct tcp_md5sig_info *md5sig;
930

931
	key = tcp_md5_do_lookup(sk, addr, family);
932 933
	if (key) {
		/* Pre-existing entry - just update that one. */
E
Eric Dumazet 已提交
934
		memcpy(key->key, newkey, newkeylen);
935
		key->keylen = newkeylen;
E
Eric Dumazet 已提交
936 937
		return 0;
	}
938

939
	md5sig = rcu_dereference_protected(tp->md5sig_info,
940
					   lockdep_sock_is_held(sk));
E
Eric Dumazet 已提交
941 942 943
	if (!md5sig) {
		md5sig = kmalloc(sizeof(*md5sig), gfp);
		if (!md5sig)
944 945
			return -ENOMEM;

E
Eric Dumazet 已提交
946 947
		sk_nocaps_add(sk, NETIF_F_GSO_MASK);
		INIT_HLIST_HEAD(&md5sig->head);
948
		rcu_assign_pointer(tp->md5sig_info, md5sig);
E
Eric Dumazet 已提交
949
	}
950

951
	key = sock_kmalloc(sk, sizeof(*key), gfp);
E
Eric Dumazet 已提交
952 953
	if (!key)
		return -ENOMEM;
954
	if (!tcp_alloc_md5sig_pool()) {
955
		sock_kfree_s(sk, key, sizeof(*key));
E
Eric Dumazet 已提交
956
		return -ENOMEM;
957
	}
E
Eric Dumazet 已提交
958 959 960 961 962 963 964 965

	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);
966 967
	return 0;
}
E
Eric Dumazet 已提交
968
EXPORT_SYMBOL(tcp_md5_do_add);
969

E
Eric Dumazet 已提交
970
int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family)
971
{
E
Eric Dumazet 已提交
972 973
	struct tcp_md5sig_key *key;

974
	key = tcp_md5_do_lookup(sk, addr, family);
E
Eric Dumazet 已提交
975 976 977
	if (!key)
		return -ENOENT;
	hlist_del_rcu(&key->node);
978
	atomic_sub(sizeof(*key), &sk->sk_omem_alloc);
E
Eric Dumazet 已提交
979 980
	kfree_rcu(key, rcu);
	return 0;
981
}
E
Eric Dumazet 已提交
982
EXPORT_SYMBOL(tcp_md5_do_del);
983

984
static void tcp_clear_md5_list(struct sock *sk)
985 986
{
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
987
	struct tcp_md5sig_key *key;
988
	struct hlist_node *n;
989
	struct tcp_md5sig_info *md5sig;
990

991 992
	md5sig = rcu_dereference_protected(tp->md5sig_info, 1);

993
	hlist_for_each_entry_safe(key, n, &md5sig->head, node) {
E
Eric Dumazet 已提交
994
		hlist_del_rcu(&key->node);
995
		atomic_sub(sizeof(*key), &sk->sk_omem_alloc);
E
Eric Dumazet 已提交
996
		kfree_rcu(key, rcu);
997 998 999
	}
}

1000 1001
static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval,
				 int optlen)
1002 1003 1004 1005 1006 1007 1008
{
	struct tcp_md5sig cmd;
	struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr;

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

1009
	if (copy_from_user(&cmd, optval, sizeof(cmd)))
1010 1011 1012 1013 1014
		return -EFAULT;

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

1015
	if (!cmd.tcpm_keylen)
E
Eric Dumazet 已提交
1016 1017
		return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr,
				      AF_INET);
1018 1019 1020 1021

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

E
Eric Dumazet 已提交
1022 1023 1024
	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);
1025 1026
}

1027 1028 1029
static int tcp_v4_md5_hash_headers(struct tcp_md5sig_pool *hp,
				   __be32 daddr, __be32 saddr,
				   const struct tcphdr *th, int nbytes)
1030 1031
{
	struct tcp4_pseudohdr *bp;
1032
	struct scatterlist sg;
1033
	struct tcphdr *_th;
1034

1035
	bp = hp->scratch;
1036 1037 1038
	bp->saddr = saddr;
	bp->daddr = daddr;
	bp->pad = 0;
1039
	bp->protocol = IPPROTO_TCP;
1040
	bp->len = cpu_to_be16(nbytes);
1041

1042 1043 1044 1045 1046 1047 1048
	_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
1049
	return crypto_ahash_update(hp->md5_req);
1050 1051
}

E
Eric Dumazet 已提交
1052
static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
1053
			       __be32 daddr, __be32 saddr, const struct tcphdr *th)
1054 1055
{
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
1056
	struct ahash_request *req;
1057 1058 1059 1060

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
1061
	req = hp->md5_req;
1062

H
Herbert Xu 已提交
1063
	if (crypto_ahash_init(req))
1064
		goto clear_hash;
1065
	if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2))
1066 1067 1068
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
H
Herbert Xu 已提交
1069 1070
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
1071 1072 1073 1074
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
1075

1076 1077 1078 1079
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
1080
	return 1;
1081 1082
}

1083 1084
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
1085
			const struct sk_buff *skb)
1086
{
1087
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
1088
	struct ahash_request *req;
E
Eric Dumazet 已提交
1089
	const struct tcphdr *th = tcp_hdr(skb);
1090 1091
	__be32 saddr, daddr;

1092 1093 1094
	if (sk) { /* valid for establish/request sockets */
		saddr = sk->sk_rcv_saddr;
		daddr = sk->sk_daddr;
1095
	} else {
1096 1097 1098
		const struct iphdr *iph = ip_hdr(skb);
		saddr = iph->saddr;
		daddr = iph->daddr;
1099
	}
1100 1101 1102 1103

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
1104
	req = hp->md5_req;
1105

H
Herbert Xu 已提交
1106
	if (crypto_ahash_init(req))
1107 1108
		goto clear_hash;

1109
	if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, skb->len))
1110 1111 1112 1113 1114
		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
1115 1116
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
1117 1118 1119 1120 1121 1122 1123 1124 1125 1126
		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;
1127
}
1128
EXPORT_SYMBOL(tcp_v4_md5_hash_skb);
1129

1130 1131
#endif

1132
/* Called with rcu_read_lock() */
1133
static bool tcp_v4_inbound_md5_hash(const struct sock *sk,
1134
				    const struct sk_buff *skb)
1135
{
1136
#ifdef CONFIG_TCP_MD5SIG
1137 1138 1139 1140 1141 1142 1143 1144
	/*
	 * 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.
	 */
1145
	const __u8 *hash_location = NULL;
1146
	struct tcp_md5sig_key *hash_expected;
1147
	const struct iphdr *iph = ip_hdr(skb);
1148
	const struct tcphdr *th = tcp_hdr(skb);
1149 1150 1151
	int genhash;
	unsigned char newhash[16];

E
Eric Dumazet 已提交
1152 1153
	hash_expected = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&iph->saddr,
					  AF_INET);
1154
	hash_location = tcp_parse_md5sig_option(th);
1155 1156 1157

	/* We've parsed the options - do we have a hash? */
	if (!hash_expected && !hash_location)
E
Eric Dumazet 已提交
1158
		return false;
1159 1160

	if (hash_expected && !hash_location) {
1161
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
E
Eric Dumazet 已提交
1162
		return true;
1163 1164 1165
	}

	if (!hash_expected && hash_location) {
1166
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
E
Eric Dumazet 已提交
1167
		return true;
1168 1169 1170 1171 1172
	}

	/* Okay, so this is hash_expected and hash_location -
	 * so we need to calculate the checksum.
	 */
1173 1174
	genhash = tcp_v4_md5_hash_skb(newhash,
				      hash_expected,
1175
				      NULL, skb);
1176 1177

	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
1178
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE);
1179 1180 1181 1182 1183
		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
1184
		return true;
1185
	}
E
Eric Dumazet 已提交
1186
	return false;
1187
#endif
1188 1189
	return false;
}
1190

1191 1192
static void tcp_v4_init_req(struct request_sock *req,
			    const struct sock *sk_listener,
1193 1194 1195 1196
			    struct sk_buff *skb)
{
	struct inet_request_sock *ireq = inet_rsk(req);

1197 1198
	sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr);
	sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr);
1199 1200 1201
	ireq->opt = tcp_v4_save_options(skb);
}

1202 1203
static struct dst_entry *tcp_v4_route_req(const struct sock *sk,
					  struct flowi *fl,
1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218
					  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;
}

1219
struct request_sock_ops tcp_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
1220
	.family		=	PF_INET,
1221
	.obj_size	=	sizeof(struct tcp_request_sock),
1222
	.rtx_syn_ack	=	tcp_rtx_synack,
1223 1224
	.send_ack	=	tcp_v4_reqsk_send_ack,
	.destructor	=	tcp_v4_reqsk_destructor,
L
Linus Torvalds 已提交
1225
	.send_reset	=	tcp_v4_send_reset,
S
stephen hemminger 已提交
1226
	.syn_ack_timeout =	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
1227 1228
};

1229
static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = {
1230
	.mss_clamp	=	TCP_MSS_DEFAULT,
1231
#ifdef CONFIG_TCP_MD5SIG
1232
	.req_md5_lookup	=	tcp_v4_md5_lookup,
1233
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1234
#endif
1235
	.init_req	=	tcp_v4_init_req,
1236 1237 1238
#ifdef CONFIG_SYN_COOKIES
	.cookie_init_seq =	cookie_v4_init_sequence,
#endif
1239
	.route_req	=	tcp_v4_route_req,
1240
	.init_seq	=	tcp_v4_init_sequence,
1241
	.send_synack	=	tcp_v4_send_synack,
1242
};
1243

L
Linus Torvalds 已提交
1244 1245 1246
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
1247
	if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
L
Linus Torvalds 已提交
1248 1249
		goto drop;

O
Octavian Purdila 已提交
1250 1251
	return tcp_conn_request(&tcp_request_sock_ops,
				&tcp_request_sock_ipv4_ops, sk, skb);
L
Linus Torvalds 已提交
1252 1253

drop:
1254
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1255 1256
	return 0;
}
E
Eric Dumazet 已提交
1257
EXPORT_SYMBOL(tcp_v4_conn_request);
L
Linus Torvalds 已提交
1258 1259 1260 1261 1262 1263


/*
 * The three way handshake has completed - we got a valid synack -
 * now create the new socket.
 */
1264
struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
1265
				  struct request_sock *req,
1266 1267 1268
				  struct dst_entry *dst,
				  struct request_sock *req_unhash,
				  bool *own_req)
L
Linus Torvalds 已提交
1269
{
1270
	struct inet_request_sock *ireq;
L
Linus Torvalds 已提交
1271 1272 1273
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1274 1275 1276
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1277
	struct ip_options_rcu *inet_opt;
L
Linus Torvalds 已提交
1278 1279 1280 1281 1282 1283

	if (sk_acceptq_is_full(sk))
		goto exit_overflow;

	newsk = tcp_create_openreq_child(sk, req, skb);
	if (!newsk)
1284
		goto exit_nonewsk;
L
Linus Torvalds 已提交
1285

1286
	newsk->sk_gso_type = SKB_GSO_TCPV4;
1287
	inet_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1288 1289 1290

	newtp		      = tcp_sk(newsk);
	newinet		      = inet_sk(newsk);
1291
	ireq		      = inet_rsk(req);
1292 1293
	sk_daddr_set(newsk, ireq->ir_rmt_addr);
	sk_rcv_saddr_set(newsk, ireq->ir_loc_addr);
1294
	newsk->sk_bound_dev_if = ireq->ir_iif;
1295
	newinet->inet_saddr	      = ireq->ir_loc_addr;
1296 1297
	inet_opt	      = ireq->opt;
	rcu_assign_pointer(newinet->inet_opt, inet_opt);
1298
	ireq->opt	      = NULL;
1299
	newinet->mc_index     = inet_iif(skb);
1300
	newinet->mc_ttl	      = ip_hdr(skb)->ttl;
1301
	newinet->rcv_tos      = ip_hdr(skb)->tos;
1302
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
1303 1304
	if (inet_opt)
		inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
E
Eric Dumazet 已提交
1305
	newinet->inet_id = newtp->write_seq ^ jiffies;
L
Linus Torvalds 已提交
1306

E
Eric Dumazet 已提交
1307 1308 1309 1310 1311 1312 1313
	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() */
	}
1314 1315
	sk_setup_caps(newsk, dst);

1316 1317
	tcp_ca_openreq_child(newsk, dst);

L
Linus Torvalds 已提交
1318
	tcp_sync_mss(newsk, dst_mtu(dst));
1319
	newtp->advmss = dst_metric_advmss(dst);
1320 1321 1322 1323
	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
1324 1325
	tcp_initialize_rcv_mss(newsk);

1326 1327
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
E
Eric Dumazet 已提交
1328 1329
	key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&newinet->inet_daddr,
				AF_INET);
1330
	if (key) {
1331 1332 1333 1334 1335 1336
		/*
		 * 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
1337 1338
		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
1339
		sk_nocaps_add(newsk, NETIF_F_GSO_MASK);
1340 1341 1342
	}
#endif

1343 1344
	if (__inet_inherit_port(sk, newsk) < 0)
		goto put_and_exit;
1345
	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
E
Eric Dumazet 已提交
1346
	if (*own_req)
1347
		tcp_move_syn(newtp, req);
L
Linus Torvalds 已提交
1348 1349 1350 1351

	return newsk;

exit_overflow:
1352
	NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1353 1354
exit_nonewsk:
	dst_release(dst);
L
Linus Torvalds 已提交
1355
exit:
1356
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1357
	return NULL;
1358
put_and_exit:
1359 1360
	inet_csk_prepare_forced_close(newsk);
	tcp_done(newsk);
1361
	goto exit;
L
Linus Torvalds 已提交
1362
}
E
Eric Dumazet 已提交
1363
EXPORT_SYMBOL(tcp_v4_syn_recv_sock);
L
Linus Torvalds 已提交
1364

1365
static struct sock *tcp_v4_cookie_check(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
1366
{
1367
#ifdef CONFIG_SYN_COOKIES
1368
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1369

1370
	if (!th->syn)
C
Cong Wang 已提交
1371
		sk = cookie_v4_check(sk, skb);
L
Linus Torvalds 已提交
1372 1373 1374 1375 1376
#endif
	return sk;
}

/* The socket must have it's spinlock held when we get
1377
 * here, unless it is a TCP_LISTEN socket.
L
Linus Torvalds 已提交
1378 1379 1380 1381 1382 1383 1384 1385
 *
 * 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)
{
1386 1387
	struct sock *rsk;

L
Linus Torvalds 已提交
1388
	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1389 1390
		struct dst_entry *dst = sk->sk_rx_dst;

1391
		sock_rps_save_rxhash(sk, skb);
1392
		sk_mark_napi_id(sk, skb);
1393
		if (dst) {
E
Eric Dumazet 已提交
1394
			if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif ||
1395
			    !dst->ops->check(dst, 0)) {
1396 1397 1398 1399
				dst_release(dst);
				sk->sk_rx_dst = NULL;
			}
		}
1400
		tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
L
Linus Torvalds 已提交
1401 1402 1403
		return 0;
	}

E
Eric Dumazet 已提交
1404
	if (tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1405 1406 1407
		goto csum_err;

	if (sk->sk_state == TCP_LISTEN) {
1408 1409
		struct sock *nsk = tcp_v4_cookie_check(sk, skb);

L
Linus Torvalds 已提交
1410 1411 1412
		if (!nsk)
			goto discard;
		if (nsk != sk) {
1413
			sock_rps_save_rxhash(nsk, skb);
1414
			sk_mark_napi_id(nsk, skb);
1415 1416
			if (tcp_child_process(sk, nsk, skb)) {
				rsk = nsk;
L
Linus Torvalds 已提交
1417
				goto reset;
1418
			}
L
Linus Torvalds 已提交
1419 1420
			return 0;
		}
1421
	} else
1422
		sock_rps_save_rxhash(sk, skb);
1423

1424
	if (tcp_rcv_state_process(sk, skb)) {
1425
		rsk = sk;
L
Linus Torvalds 已提交
1426
		goto reset;
1427
	}
L
Linus Torvalds 已提交
1428 1429 1430
	return 0;

reset:
1431
	tcp_v4_send_reset(rsk, skb);
L
Linus Torvalds 已提交
1432 1433 1434 1435 1436 1437 1438 1439 1440 1441
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:
1442 1443
	TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS);
	TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1444 1445
	goto discard;
}
E
Eric Dumazet 已提交
1446
EXPORT_SYMBOL(tcp_v4_do_rcv);
L
Linus Torvalds 已提交
1447

1448
void tcp_v4_early_demux(struct sk_buff *skb)
D
David S. Miller 已提交
1449 1450 1451 1452 1453 1454
{
	const struct iphdr *iph;
	const struct tcphdr *th;
	struct sock *sk;

	if (skb->pkt_type != PACKET_HOST)
1455
		return;
D
David S. Miller 已提交
1456

1457
	if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr)))
1458
		return;
D
David S. Miller 已提交
1459 1460

	iph = ip_hdr(skb);
1461
	th = tcp_hdr(skb);
D
David S. Miller 已提交
1462 1463

	if (th->doff < sizeof(struct tcphdr) / 4)
1464
		return;
D
David S. Miller 已提交
1465

1466
	sk = __inet_lookup_established(dev_net(skb->dev), &tcp_hashinfo,
D
David S. Miller 已提交
1467
				       iph->saddr, th->source,
1468
				       iph->daddr, ntohs(th->dest),
E
Eric Dumazet 已提交
1469
				       skb->skb_iif);
D
David S. Miller 已提交
1470 1471 1472
	if (sk) {
		skb->sk = sk;
		skb->destructor = sock_edemux;
1473
		if (sk_fullsock(sk)) {
1474
			struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst);
E
Eric Dumazet 已提交
1475

D
David S. Miller 已提交
1476 1477
			if (dst)
				dst = dst_check(dst, 0);
1478
			if (dst &&
E
Eric Dumazet 已提交
1479
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
1480
				skb_dst_set_noref(skb, dst);
D
David S. Miller 已提交
1481 1482 1483 1484
		}
	}
}

E
Eric Dumazet 已提交
1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502
/* 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;

1503 1504 1505 1506 1507 1508 1509 1510 1511
	/* 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
1512
		skb_dst_force_safe(skb);
1513

E
Eric Dumazet 已提交
1514 1515
	__skb_queue_tail(&tp->ucopy.prequeue, skb);
	tp->ucopy.memory += skb->truesize;
E
Eric Dumazet 已提交
1516 1517
	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
1518 1519 1520
		struct sk_buff *skb1;

		BUG_ON(sock_owned_by_user(sk));
E
Eric Dumazet 已提交
1521 1522
		__NET_ADD_STATS(sock_net(sk), LINUX_MIB_TCPPREQUEUEDROPPED,
				skb_queue_len(&tp->ucopy.prequeue));
E
Eric Dumazet 已提交
1523

E
Eric Dumazet 已提交
1524
		while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL)
E
Eric Dumazet 已提交
1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539
			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
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 1567
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
1568 1569 1570 1571 1572 1573
/*
 *	From tcp_input.c
 */

int tcp_v4_rcv(struct sk_buff *skb)
{
1574
	struct net *net = dev_net(skb->dev);
1575
	const struct iphdr *iph;
1576
	const struct tcphdr *th;
1577
	bool refcounted;
L
Linus Torvalds 已提交
1578 1579 1580 1581 1582 1583 1584
	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
1585
	__TCP_INC_STATS(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1586 1587 1588 1589

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

1590
	th = (const struct tcphdr *)skb->data;
L
Linus Torvalds 已提交
1591

1592
	if (unlikely(th->doff < sizeof(struct tcphdr) / 4))
L
Linus Torvalds 已提交
1593 1594 1595 1596 1597 1598
		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
1599
	 * provided case of th->doff==0 is eliminated.
L
Linus Torvalds 已提交
1600
	 * So, we defer the checks. */
1601 1602

	if (skb_checksum_init(skb, IPPROTO_TCP, inet_compute_pseudo))
1603
		goto csum_error;
L
Linus Torvalds 已提交
1604

1605
	th = (const struct tcphdr *)skb->data;
1606
	iph = ip_hdr(skb);
1607 1608 1609 1610 1611 1612 1613
	/* 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
1614 1615 1616 1617
	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);
1618
	TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
1619
	TCP_SKB_CB(skb)->tcp_tw_isn = 0;
1620
	TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph);
L
Linus Torvalds 已提交
1621 1622
	TCP_SKB_CB(skb)->sacked	 = 0;

1623
lookup:
1624
	sk = __inet_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th), th->source,
1625
			       th->dest, &refcounted);
L
Linus Torvalds 已提交
1626 1627 1628
	if (!sk)
		goto no_tcp_socket;

E
Eric Dumazet 已提交
1629 1630 1631 1632
process:
	if (sk->sk_state == TCP_TIME_WAIT)
		goto do_time_wait;

1633 1634
	if (sk->sk_state == TCP_NEW_SYN_RECV) {
		struct request_sock *req = inet_reqsk(sk);
1635
		struct sock *nsk;
1636 1637

		sk = req->rsk_listener;
1638
		if (unlikely(tcp_v4_inbound_md5_hash(sk, skb))) {
1639
			sk_drops_add(sk, skb);
1640 1641 1642
			reqsk_put(req);
			goto discard_it;
		}
1643
		if (unlikely(sk->sk_state != TCP_LISTEN)) {
1644
			inet_csk_reqsk_queue_drop_and_put(sk, req);
1645 1646
			goto lookup;
		}
1647 1648 1649
		/* We own a reference on the listener, increase it again
		 * as we might lose it too soon.
		 */
1650
		sock_hold(sk);
1651
		refcounted = true;
1652
		nsk = tcp_check_req(sk, skb, req, false);
1653 1654
		if (!nsk) {
			reqsk_put(req);
1655
			goto discard_and_relse;
1656 1657 1658 1659 1660
		}
		if (nsk == sk) {
			reqsk_put(req);
		} else if (tcp_child_process(sk, nsk, skb)) {
			tcp_v4_send_reset(nsk, skb);
1661
			goto discard_and_relse;
1662
		} else {
1663
			sock_put(sk);
1664 1665 1666
			return 0;
		}
	}
1667
	if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
1668
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
1669
		goto discard_and_relse;
1670
	}
1671

L
Linus Torvalds 已提交
1672 1673
	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;
1674 1675 1676 1677

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

1678
	nf_reset(skb);
L
Linus Torvalds 已提交
1679

1680
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1681 1682 1683 1684
		goto discard_and_relse;

	skb->dev = NULL;

1685 1686 1687 1688 1689 1690 1691
	if (sk->sk_state == TCP_LISTEN) {
		ret = tcp_v4_do_rcv(sk, skb);
		goto put_and_return;
	}

	sk_incoming_cpu_update(sk);

1692
	bh_lock_sock_nested(sk);
1693
	tcp_segs_in(tcp_sk(sk), skb);
L
Linus Torvalds 已提交
1694 1695
	ret = 0;
	if (!sock_owned_by_user(sk)) {
D
Dan Williams 已提交
1696
		if (!tcp_prequeue(sk, skb))
L
Linus Torvalds 已提交
1697
			ret = tcp_v4_do_rcv(sk, skb);
E
Eric Dumazet 已提交
1698
	} else if (tcp_add_backlog(sk, skb)) {
Z
Zhu Yi 已提交
1699 1700
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1701 1702
	bh_unlock_sock(sk);

1703
put_and_return:
1704 1705
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1706 1707 1708 1709 1710 1711 1712

	return ret;

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

E
Eric Dumazet 已提交
1713
	if (tcp_checksum_complete(skb)) {
1714
csum_error:
E
Eric Dumazet 已提交
1715
		__TCP_INC_STATS(net, TCP_MIB_CSUMERRORS);
L
Linus Torvalds 已提交
1716
bad_packet:
E
Eric Dumazet 已提交
1717
		__TCP_INC_STATS(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1718
	} else {
1719
		tcp_v4_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1720 1721 1722 1723 1724
	}

discard_it:
	/* Discard frame. */
	kfree_skb(skb);
1725
	return 0;
L
Linus Torvalds 已提交
1726 1727

discard_and_relse:
1728
	sk_drops_add(sk, skb);
1729 1730
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1731 1732 1733 1734
	goto discard_it;

do_time_wait:
	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1735
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1736 1737 1738
		goto discard_it;
	}

1739 1740 1741
	if (tcp_checksum_complete(skb)) {
		inet_twsk_put(inet_twsk(sk));
		goto csum_error;
L
Linus Torvalds 已提交
1742
	}
1743
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1744
	case TCP_TW_SYN: {
1745
		struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev),
1746 1747
							&tcp_hashinfo, skb,
							__tcp_hdrlen(th),
1748
							iph->saddr, th->source,
1749
							iph->daddr, th->dest,
1750
							inet_iif(skb));
L
Linus Torvalds 已提交
1751
		if (sk2) {
1752
			inet_twsk_deschedule_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1753
			sk = sk2;
1754
			refcounted = false;
L
Linus Torvalds 已提交
1755 1756 1757 1758 1759 1760 1761 1762
			goto process;
		}
		/* Fall through to ACK */
	}
	case TCP_TW_ACK:
		tcp_v4_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
1763 1764 1765
		tcp_v4_send_reset(sk, skb);
		inet_twsk_deschedule_put(inet_twsk(sk));
		goto discard_it;
L
Linus Torvalds 已提交
1766 1767 1768 1769 1770
	case TCP_TW_SUCCESS:;
	}
	goto discard_it;
}

1771 1772 1773 1774 1775
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
1776

1777
void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
E
Eric Dumazet 已提交
1778 1779 1780
{
	struct dst_entry *dst = skb_dst(skb);

E
Eric Dumazet 已提交
1781
	if (dst && dst_hold_safe(dst)) {
1782 1783 1784
		sk->sk_rx_dst = dst;
		inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
	}
E
Eric Dumazet 已提交
1785
}
1786
EXPORT_SYMBOL(inet_sk_rx_dst_set);
E
Eric Dumazet 已提交
1787

1788
const struct inet_connection_sock_af_ops ipv4_specific = {
1789 1790 1791
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
E
Eric Dumazet 已提交
1792
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1793 1794 1795 1796 1797 1798 1799
	.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),
1800
	.bind_conflict	   = inet_csk_bind_conflict,
1801
#ifdef CONFIG_COMPAT
1802 1803
	.compat_setsockopt = compat_ip_setsockopt,
	.compat_getsockopt = compat_ip_getsockopt,
1804
#endif
1805
	.mtu_reduced	   = tcp_v4_mtu_reduced,
L
Linus Torvalds 已提交
1806
};
E
Eric Dumazet 已提交
1807
EXPORT_SYMBOL(ipv4_specific);
L
Linus Torvalds 已提交
1808

1809
#ifdef CONFIG_TCP_MD5SIG
1810
static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = {
1811
	.md5_lookup		= tcp_v4_md5_lookup,
1812
	.calc_md5_hash		= tcp_v4_md5_hash_skb,
1813 1814
	.md5_parse		= tcp_v4_parse_md5_keys,
};
1815
#endif
1816

L
Linus Torvalds 已提交
1817 1818 1819 1820 1821
/* 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)
{
1822
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1823

1824
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1825

1826
	icsk->icsk_af_ops = &ipv4_specific;
1827

1828
#ifdef CONFIG_TCP_MD5SIG
1829
	tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific;
1830
#endif
L
Linus Torvalds 已提交
1831 1832 1833 1834

	return 0;
}

1835
void tcp_v4_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1836 1837 1838 1839 1840
{
	struct tcp_sock *tp = tcp_sk(sk);

	tcp_clear_xmit_timers(sk);

1841
	tcp_cleanup_congestion_control(sk);
1842

L
Linus Torvalds 已提交
1843
	/* Cleanup up the write buffer. */
1844
	tcp_write_queue_purge(sk);
L
Linus Torvalds 已提交
1845 1846

	/* Cleans up our, hopefully empty, out_of_order_queue. */
1847
	skb_rbtree_purge(&tp->out_of_order_queue);
L
Linus Torvalds 已提交
1848

1849 1850 1851
#ifdef CONFIG_TCP_MD5SIG
	/* Clean up the MD5 key list, if any */
	if (tp->md5sig_info) {
E
Eric Dumazet 已提交
1852
		tcp_clear_md5_list(sk);
1853
		kfree_rcu(tp->md5sig_info, rcu);
1854 1855 1856
		tp->md5sig_info = NULL;
	}
#endif
C
Chris Leech 已提交
1857

L
Linus Torvalds 已提交
1858 1859 1860 1861
	/* Clean prequeue, it must be empty really */
	__skb_queue_purge(&tp->ucopy.prequeue);

	/* Clean up a referenced TCP bind bucket. */
1862
	if (inet_csk(sk)->icsk_bind_hash)
1863
		inet_put_port(sk);
L
Linus Torvalds 已提交
1864

1865
	BUG_ON(tp->fastopen_rsk);
1866

1867 1868
	/* If socket is aborted during connect operation */
	tcp_free_fastopen_req(tp);
1869
	tcp_saved_syn_free(tp);
1870

E
Eric Dumazet 已提交
1871
	local_bh_disable();
1872
	sk_sockets_allocated_dec(sk);
E
Eric Dumazet 已提交
1873
	local_bh_enable();
L
Linus Torvalds 已提交
1874 1875 1876 1877 1878 1879
}
EXPORT_SYMBOL(tcp_v4_destroy_sock);

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

1880 1881 1882 1883 1884
/*
 * 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
1885 1886
static void *listening_get_next(struct seq_file *seq, void *cur)
{
J
Jianjun Kong 已提交
1887
	struct tcp_iter_state *st = seq->private;
1888
	struct net *net = seq_file_net(seq);
1889 1890 1891
	struct inet_listen_hashbucket *ilb;
	struct inet_connection_sock *icsk;
	struct sock *sk = cur;
L
Linus Torvalds 已提交
1892 1893

	if (!sk) {
1894
get_head:
1895
		ilb = &tcp_hashinfo.listening_hash[st->bucket];
1896
		spin_lock_bh(&ilb->lock);
1897
		sk = sk_head(&ilb->head);
1898
		st->offset = 0;
L
Linus Torvalds 已提交
1899 1900
		goto get_sk;
	}
1901
	ilb = &tcp_hashinfo.listening_hash[st->bucket];
L
Linus Torvalds 已提交
1902
	++st->num;
1903
	++st->offset;
L
Linus Torvalds 已提交
1904

1905
	sk = sk_next(sk);
L
Linus Torvalds 已提交
1906
get_sk:
1907
	sk_for_each_from(sk) {
1908 1909
		if (!net_eq(sock_net(sk), net))
			continue;
1910 1911
		if (sk->sk_family == st->family)
			return sk;
1912
		icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1913
	}
1914
	spin_unlock_bh(&ilb->lock);
1915
	st->offset = 0;
1916 1917 1918
	if (++st->bucket < INET_LHTABLE_SIZE)
		goto get_head;
	return NULL;
L
Linus Torvalds 已提交
1919 1920 1921 1922
}

static void *listening_get_idx(struct seq_file *seq, loff_t *pos)
{
1923 1924 1925 1926 1927 1928
	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
1929 1930 1931 1932 1933 1934 1935 1936

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

E
Eric Dumazet 已提交
1937
static inline bool empty_bucket(const struct tcp_iter_state *st)
1938
{
E
Eric Dumazet 已提交
1939
	return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain);
1940 1941
}

1942 1943 1944 1945
/*
 * 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
1946 1947
static void *established_get_first(struct seq_file *seq)
{
J
Jianjun Kong 已提交
1948
	struct tcp_iter_state *st = seq->private;
1949
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
1950 1951
	void *rc = NULL;

1952 1953
	st->offset = 0;
	for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) {
L
Linus Torvalds 已提交
1954
		struct sock *sk;
1955
		struct hlist_nulls_node *node;
1956
		spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket);
L
Linus Torvalds 已提交
1957

1958 1959 1960 1961
		/* Lockless fast path for the common case of empty buckets */
		if (empty_bucket(st))
			continue;

1962
		spin_lock_bh(lock);
1963
		sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) {
1964
			if (sk->sk_family != st->family ||
1965
			    !net_eq(sock_net(sk), net)) {
L
Linus Torvalds 已提交
1966 1967 1968 1969 1970
				continue;
			}
			rc = sk;
			goto out;
		}
1971
		spin_unlock_bh(lock);
L
Linus Torvalds 已提交
1972 1973 1974 1975 1976 1977 1978 1979
	}
out:
	return rc;
}

static void *established_get_next(struct seq_file *seq, void *cur)
{
	struct sock *sk = cur;
1980
	struct hlist_nulls_node *node;
J
Jianjun Kong 已提交
1981
	struct tcp_iter_state *st = seq->private;
1982
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
1983 1984

	++st->num;
1985
	++st->offset;
L
Linus Torvalds 已提交
1986

E
Eric Dumazet 已提交
1987
	sk = sk_nulls_next(sk);
L
Linus Torvalds 已提交
1988

1989
	sk_nulls_for_each_from(sk, node) {
1990
		if (sk->sk_family == st->family && net_eq(sock_net(sk), net))
E
Eric Dumazet 已提交
1991
			return sk;
L
Linus Torvalds 已提交
1992 1993
	}

E
Eric Dumazet 已提交
1994 1995 1996
	spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
	++st->bucket;
	return established_get_first(seq);
L
Linus Torvalds 已提交
1997 1998 1999 2000
}

static void *established_get_idx(struct seq_file *seq, loff_t pos)
{
2001 2002 2003 2004 2005
	struct tcp_iter_state *st = seq->private;
	void *rc;

	st->bucket = 0;
	rc = established_get_first(seq);
L
Linus Torvalds 已提交
2006 2007 2008 2009

	while (rc && pos) {
		rc = established_get_next(seq, rc);
		--pos;
2010
	}
L
Linus Torvalds 已提交
2011 2012 2013 2014 2015 2016
	return rc;
}

static void *tcp_get_idx(struct seq_file *seq, loff_t pos)
{
	void *rc;
J
Jianjun Kong 已提交
2017
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029

	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;
}

2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047
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
2048
		st->state = TCP_SEQ_STATE_ESTABLISHED;
2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062
		/* 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
2063 2064
static void *tcp_seq_start(struct seq_file *seq, loff_t *pos)
{
J
Jianjun Kong 已提交
2065
	struct tcp_iter_state *st = seq->private;
2066 2067 2068 2069 2070 2071 2072 2073
	void *rc;

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

L
Linus Torvalds 已提交
2074 2075
	st->state = TCP_SEQ_STATE_LISTENING;
	st->num = 0;
2076 2077 2078 2079 2080 2081 2082
	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
2083 2084 2085 2086
}

static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
2087
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099
	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;
2100 2101
			st->bucket = 0;
			st->offset = 0;
L
Linus Torvalds 已提交
2102 2103 2104 2105 2106 2107 2108 2109 2110
			rc	  = established_get_first(seq);
		}
		break;
	case TCP_SEQ_STATE_ESTABLISHED:
		rc = established_get_next(seq, v);
		break;
	}
out:
	++*pos;
2111
	st->last_pos = *pos;
L
Linus Torvalds 已提交
2112 2113 2114 2115 2116
	return rc;
}

static void tcp_seq_stop(struct seq_file *seq, void *v)
{
J
Jianjun Kong 已提交
2117
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2118 2119 2120 2121

	switch (st->state) {
	case TCP_SEQ_STATE_LISTENING:
		if (v != SEQ_START_TOKEN)
2122
			spin_unlock_bh(&tcp_hashinfo.listening_hash[st->bucket].lock);
L
Linus Torvalds 已提交
2123 2124 2125
		break;
	case TCP_SEQ_STATE_ESTABLISHED:
		if (v)
2126
			spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
L
Linus Torvalds 已提交
2127 2128 2129 2130
		break;
	}
}

2131
int tcp_seq_open(struct inode *inode, struct file *file)
L
Linus Torvalds 已提交
2132
{
A
Al Viro 已提交
2133
	struct tcp_seq_afinfo *afinfo = PDE_DATA(inode);
L
Linus Torvalds 已提交
2134
	struct tcp_iter_state *s;
2135
	int err;
L
Linus Torvalds 已提交
2136

2137 2138 2139 2140
	err = seq_open_net(inode, file, &afinfo->seq_ops,
			  sizeof(struct tcp_iter_state));
	if (err < 0)
		return err;
2141

2142
	s = ((struct seq_file *)file->private_data)->private;
L
Linus Torvalds 已提交
2143
	s->family		= afinfo->family;
S
stephen hemminger 已提交
2144
	s->last_pos		= 0;
2145 2146
	return 0;
}
2147
EXPORT_SYMBOL(tcp_seq_open);
2148

2149
int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo)
L
Linus Torvalds 已提交
2150 2151 2152 2153
{
	int rc = 0;
	struct proc_dir_entry *p;

2154 2155 2156 2157
	afinfo->seq_ops.start		= tcp_seq_start;
	afinfo->seq_ops.next		= tcp_seq_next;
	afinfo->seq_ops.stop		= tcp_seq_stop;

2158
	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
2159
			     afinfo->seq_fops, afinfo);
2160
	if (!p)
L
Linus Torvalds 已提交
2161 2162 2163
		rc = -ENOMEM;
	return rc;
}
E
Eric Dumazet 已提交
2164
EXPORT_SYMBOL(tcp_proc_register);
L
Linus Torvalds 已提交
2165

2166
void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo)
L
Linus Torvalds 已提交
2167
{
2168
	remove_proc_entry(afinfo->name, net->proc_net);
L
Linus Torvalds 已提交
2169
}
E
Eric Dumazet 已提交
2170
EXPORT_SYMBOL(tcp_proc_unregister);
L
Linus Torvalds 已提交
2171

2172
static void get_openreq4(const struct request_sock *req,
E
Eric Dumazet 已提交
2173
			 struct seq_file *f, int i)
L
Linus Torvalds 已提交
2174
{
2175
	const struct inet_request_sock *ireq = inet_rsk(req);
2176
	long delta = req->rsk_timer.expires - jiffies;
L
Linus Torvalds 已提交
2177

2178
	seq_printf(f, "%4d: %08X:%04X %08X:%04X"
2179
		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %u %d %pK",
L
Linus Torvalds 已提交
2180
		i,
2181
		ireq->ir_loc_addr,
2182
		ireq->ir_num,
2183 2184
		ireq->ir_rmt_addr,
		ntohs(ireq->ir_rmt_port),
L
Linus Torvalds 已提交
2185 2186 2187
		TCP_SYN_RECV,
		0, 0, /* could print option size, but that is af dependent. */
		1,    /* timers active (only the expire timer) */
2188
		jiffies_delta_to_clock_t(delta),
2189
		req->num_timeout,
E
Eric Dumazet 已提交
2190 2191
		from_kuid_munged(seq_user_ns(f),
				 sock_i_uid(req->rsk_listener)),
L
Linus Torvalds 已提交
2192 2193
		0,  /* non standard timer */
		0, /* open_requests have no inode */
2194
		0,
2195
		req);
L
Linus Torvalds 已提交
2196 2197
}

2198
static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i)
L
Linus Torvalds 已提交
2199 2200 2201
{
	int timer_active;
	unsigned long timer_expires;
2202
	const struct tcp_sock *tp = tcp_sk(sk);
2203
	const struct inet_connection_sock *icsk = inet_csk(sk);
2204
	const struct inet_sock *inet = inet_sk(sk);
2205
	const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq;
E
Eric Dumazet 已提交
2206 2207 2208 2209
	__be32 dest = inet->inet_daddr;
	__be32 src = inet->inet_rcv_saddr;
	__u16 destp = ntohs(inet->inet_dport);
	__u16 srcp = ntohs(inet->inet_sport);
2210
	int rx_queue;
2211
	int state;
L
Linus Torvalds 已提交
2212

N
Nandita Dukkipati 已提交
2213 2214 2215
	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
2216
		timer_active	= 1;
2217 2218
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
2219
		timer_active	= 4;
2220
		timer_expires	= icsk->icsk_timeout;
2221
	} else if (timer_pending(&sk->sk_timer)) {
L
Linus Torvalds 已提交
2222
		timer_active	= 2;
2223
		timer_expires	= sk->sk_timer.expires;
L
Linus Torvalds 已提交
2224 2225 2226 2227 2228
	} else {
		timer_active	= 0;
		timer_expires = jiffies;
	}

2229 2230
	state = sk_state_load(sk);
	if (state == TCP_LISTEN)
2231 2232
		rx_queue = sk->sk_ack_backlog;
	else
2233 2234
		/* Because we don't lock the socket,
		 * we might find a transient negative value.
2235 2236 2237
		 */
		rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0);

2238
	seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX "
2239
			"%08X %5u %8d %lu %d %pK %lu %lu %u %u %d",
2240
		i, src, srcp, dest, destp, state,
2241
		tp->write_seq - tp->snd_una,
2242
		rx_queue,
L
Linus Torvalds 已提交
2243
		timer_active,
2244
		jiffies_delta_to_clock_t(timer_expires - jiffies),
2245
		icsk->icsk_retransmits,
2246
		from_kuid_munged(seq_user_ns(f), sock_i_uid(sk)),
2247
		icsk->icsk_probes_out,
2248 2249
		sock_i_ino(sk),
		atomic_read(&sk->sk_refcnt), sk,
2250 2251
		jiffies_to_clock_t(icsk->icsk_rto),
		jiffies_to_clock_t(icsk->icsk_ack.ato),
2252
		(icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
L
Linus Torvalds 已提交
2253
		tp->snd_cwnd,
2254 2255
		state == TCP_LISTEN ?
		    fastopenq->max_qlen :
2256
		    (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh));
L
Linus Torvalds 已提交
2257 2258
}

2259
static void get_timewait4_sock(const struct inet_timewait_sock *tw,
2260
			       struct seq_file *f, int i)
L
Linus Torvalds 已提交
2261
{
2262
	long delta = tw->tw_timer.expires - jiffies;
2263
	__be32 dest, src;
L
Linus Torvalds 已提交
2264 2265 2266 2267 2268 2269 2270
	__u16 destp, srcp;

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

2271
	seq_printf(f, "%4d: %08X:%04X %08X:%04X"
2272
		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK",
L
Linus Torvalds 已提交
2273
		i, src, srcp, dest, destp, tw->tw_substate, 0, 0,
2274
		3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
2275
		atomic_read(&tw->tw_refcnt), tw);
L
Linus Torvalds 已提交
2276 2277 2278 2279 2280 2281
}

#define TMPSZ 150

static int tcp4_seq_show(struct seq_file *seq, void *v)
{
J
Jianjun Kong 已提交
2282
	struct tcp_iter_state *st;
E
Eric Dumazet 已提交
2283
	struct sock *sk = v;
L
Linus Torvalds 已提交
2284

2285
	seq_setwidth(seq, TMPSZ - 1);
L
Linus Torvalds 已提交
2286
	if (v == SEQ_START_TOKEN) {
2287
		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
L
Linus Torvalds 已提交
2288 2289 2290 2291 2292 2293
			   "rx_queue tr tm->when retrnsmt   uid  timeout "
			   "inode");
		goto out;
	}
	st = seq->private;

2294 2295 2296
	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
2297
		get_openreq4(v, seq, st->num);
2298 2299
	else
		get_tcp4_sock(v, seq, st->num);
L
Linus Torvalds 已提交
2300
out:
2301
	seq_pad(seq, '\n');
L
Linus Torvalds 已提交
2302 2303 2304
	return 0;
}

2305 2306 2307 2308 2309 2310 2311 2312
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
2313 2314 2315
static struct tcp_seq_afinfo tcp4_seq_afinfo = {
	.name		= "tcp",
	.family		= AF_INET,
2316
	.seq_fops	= &tcp_afinfo_seq_fops,
2317 2318 2319
	.seq_ops	= {
		.show		= tcp4_seq_show,
	},
L
Linus Torvalds 已提交
2320 2321
};

2322
static int __net_init tcp4_proc_init_net(struct net *net)
2323 2324 2325 2326
{
	return tcp_proc_register(net, &tcp4_seq_afinfo);
}

2327
static void __net_exit tcp4_proc_exit_net(struct net *net)
2328 2329 2330 2331 2332 2333 2334 2335 2336
{
	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
2337 2338
int __init tcp4_proc_init(void)
{
2339
	return register_pernet_subsys(&tcp4_net_ops);
L
Linus Torvalds 已提交
2340 2341 2342 2343
}

void tcp4_proc_exit(void)
{
2344
	unregister_pernet_subsys(&tcp4_net_ops);
L
Linus Torvalds 已提交
2345 2346 2347 2348 2349 2350 2351 2352 2353
}
#endif /* CONFIG_PROC_FS */

struct proto tcp_prot = {
	.name			= "TCP",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v4_connect,
	.disconnect		= tcp_disconnect,
2354
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
2355 2356 2357 2358 2359 2360 2361
	.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,
2362 2363
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
2364
	.backlog_rcv		= tcp_v4_do_rcv,
E
Eric Dumazet 已提交
2365
	.release_cb		= tcp_release_cb,
2366 2367 2368
	.hash			= inet_hash,
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
2369
	.enter_memory_pressure	= tcp_enter_memory_pressure,
2370
	.stream_memory_free	= tcp_stream_memory_free,
L
Linus Torvalds 已提交
2371
	.sockets_allocated	= &tcp_sockets_allocated,
2372
	.orphan_count		= &tcp_orphan_count,
L
Linus Torvalds 已提交
2373 2374
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
2375
	.sysctl_mem		= sysctl_tcp_mem,
L
Linus Torvalds 已提交
2376 2377 2378 2379
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp_sock),
2380
	.slab_flags		= SLAB_DESTROY_BY_RCU,
2381
	.twsk_prot		= &tcp_timewait_sock_ops,
2382
	.rsk_prot		= &tcp_request_sock_ops,
2383
	.h.hashinfo		= &tcp_hashinfo,
2384
	.no_autobind		= true,
2385 2386 2387
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
G
Glauber Costa 已提交
2388
#endif
2389
	.diag_destroy		= tcp_abort,
L
Linus Torvalds 已提交
2390
};
E
Eric Dumazet 已提交
2391
EXPORT_SYMBOL(tcp_prot);
L
Linus Torvalds 已提交
2392

2393 2394 2395 2396 2397 2398 2399 2400 2401
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);
}

2402 2403
static int __net_init tcp_sk_init(struct net *net)
{
2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416
	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;
2417
		sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
2418 2419
		*per_cpu_ptr(net->ipv4.tcp_sk, cpu) = sk;
	}
2420

2421
	net->ipv4.sysctl_tcp_ecn = 2;
2422 2423
	net->ipv4.sysctl_tcp_ecn_fallback = 1;

F
Fan Du 已提交
2424
	net->ipv4.sysctl_tcp_base_mss = TCP_BASE_MSS;
2425
	net->ipv4.sysctl_tcp_probe_threshold = TCP_PROBE_THRESHOLD;
2426
	net->ipv4.sysctl_tcp_probe_interval = TCP_PROBE_INTERVAL;
2427

2428
	net->ipv4.sysctl_tcp_keepalive_time = TCP_KEEPALIVE_TIME;
2429
	net->ipv4.sysctl_tcp_keepalive_probes = TCP_KEEPALIVE_PROBES;
2430
	net->ipv4.sysctl_tcp_keepalive_intvl = TCP_KEEPALIVE_INTVL;
2431

2432
	net->ipv4.sysctl_tcp_syn_retries = TCP_SYN_RETRIES;
2433
	net->ipv4.sysctl_tcp_synack_retries = TCP_SYNACK_RETRIES;
2434
	net->ipv4.sysctl_tcp_syncookies = 1;
2435
	net->ipv4.sysctl_tcp_reordering = TCP_FASTRETRANS_THRESH;
2436
	net->ipv4.sysctl_tcp_retries1 = TCP_RETR1;
2437
	net->ipv4.sysctl_tcp_retries2 = TCP_RETR2;
2438
	net->ipv4.sysctl_tcp_orphan_retries = 0;
2439
	net->ipv4.sysctl_tcp_fin_timeout = TCP_FIN_TIMEOUT;
2440
	net->ipv4.sysctl_tcp_notsent_lowat = UINT_MAX;
2441

2442
	return 0;
2443 2444 2445 2446
fail:
	tcp_sk_exit(net);

	return res;
E
Eric W. Biederman 已提交
2447 2448 2449 2450 2451
}

static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list)
{
	inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET);
2452 2453 2454
}

static struct pernet_operations __net_initdata tcp_sk_ops = {
E
Eric W. Biederman 已提交
2455 2456 2457
       .init	   = tcp_sk_init,
       .exit	   = tcp_sk_exit,
       .exit_batch = tcp_sk_exit_batch,
2458 2459
};

2460
void __init tcp_v4_init(void)
L
Linus Torvalds 已提交
2461
{
2462
	inet_hashinfo_init(&tcp_hashinfo);
2463
	if (register_pernet_subsys(&tcp_sk_ops))
L
Linus Torvalds 已提交
2464 2465
		panic("Failed to create the TCP control socket.\n");
}