tcp_ipv4.c 63.5 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
int sysctl_tcp_low_latency __read_mostly;
L
Linus Torvalds 已提交
88

89
#ifdef CONFIG_TCP_MD5SIG
E
Eric Dumazet 已提交
90
static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
91
			       __be32 daddr, __be32 saddr, const struct tcphdr *th);
92 93
#endif

94
struct inet_hashinfo tcp_hashinfo;
E
Eric Dumazet 已提交
95
EXPORT_SYMBOL(tcp_hashinfo);
L
Linus Torvalds 已提交
96

97
static u32 tcp_v4_init_sequence(const struct sk_buff *skb, u32 *tsoff)
L
Linus Torvalds 已提交
98
{
99 100
	return secure_tcp_sequence_number(ip_hdr(skb)->daddr,
					  ip_hdr(skb)->saddr,
101
					  tcp_hdr(skb)->dest,
102
					  tcp_hdr(skb)->source, tsoff);
L
Linus Torvalds 已提交
103 104
}

105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121
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 &&
122
	    (!twp || (sock_net(sk)->ipv4.sysctl_tcp_tw_reuse &&
123
			     get_seconds() - tcptw->tw_ts_recent_stamp > 1))) {
124 125 126 127 128 129 130 131 132 133 134 135 136
		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
137 138 139
/* This will initiate an outgoing connection. */
int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
{
140
	struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
L
Linus Torvalds 已提交
141 142
	struct inet_sock *inet = inet_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
143
	__be16 orig_sport, orig_dport;
144
	__be32 daddr, nexthop;
145
	struct flowi4 *fl4;
146
	struct rtable *rt;
L
Linus Torvalds 已提交
147
	int err;
148
	struct ip_options_rcu *inet_opt;
149
	struct inet_timewait_death_row *tcp_death_row = &sock_net(sk)->ipv4.tcp_death_row;
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,
240 241
							   usin->sin_port,
							   &tp->tsoffset);
L
Linus Torvalds 已提交
242

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
			} else {
446
				if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, &sk->sk_tsq_flags))
447 448
					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
	arg.uid = sock_net_uid(net, sk && sk_fullsock(sk) ? sk : NULL);
696
	local_bh_disable();
697 698
	ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk),
			      skb, &TCP_SKB_CB(skb)->header.h4.opt,
699 700
			      ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
			      &arg, arg.iov[0].iov_len);
L
Linus Torvalds 已提交
701

E
Eric Dumazet 已提交
702 703
	__TCP_INC_STATS(net, TCP_MIB_OUTSEGS);
	__TCP_INC_STATS(net, TCP_MIB_OUTRSTS);
704
	local_bh_enable();
705 706

#ifdef CONFIG_TCP_MD5SIG
707 708
out:
	rcu_read_unlock();
709
#endif
L
Linus Torvalds 已提交
710 711 712 713 714 715
}

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

716
static void tcp_v4_send_ack(const struct sock *sk,
717
			    struct sk_buff *skb, u32 seq, u32 ack,
718
			    u32 win, u32 tsval, u32 tsecr, int oif,
719
			    struct tcp_md5sig_key *key,
720
			    int reply_flags, u8 tos)
L
Linus Torvalds 已提交
721
{
722
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
723 724
	struct {
		struct tcphdr th;
725
		__be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2)
726
#ifdef CONFIG_TCP_MD5SIG
727
			   + (TCPOLEN_MD5SIG_ALIGNED >> 2)
728 729
#endif
			];
L
Linus Torvalds 已提交
730
	} rep;
731
	struct net *net = sock_net(sk);
L
Linus Torvalds 已提交
732 733 734
	struct ip_reply_arg arg;

	memset(&rep.th, 0, sizeof(struct tcphdr));
735
	memset(&arg, 0, sizeof(arg));
L
Linus Torvalds 已提交
736 737 738

	arg.iov[0].iov_base = (unsigned char *)&rep;
	arg.iov[0].iov_len  = sizeof(rep.th);
739
	if (tsecr) {
740 741 742
		rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				   (TCPOPT_TIMESTAMP << 8) |
				   TCPOLEN_TIMESTAMP);
743 744
		rep.opt[1] = htonl(tsval);
		rep.opt[2] = htonl(tsecr);
745
		arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED;
L
Linus Torvalds 已提交
746 747 748 749 750 751 752 753 754 755 756
	}

	/* 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);

757 758
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
759
		int offset = (tsecr) ? 3 : 0;
760 761 762 763 764 765 766 767

		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;

768
		tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset],
769 770
				    key, ip_hdr(skb)->saddr,
				    ip_hdr(skb)->daddr, &rep.th);
771 772
	}
#endif
773
	arg.flags = reply_flags;
774 775
	arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
				      ip_hdr(skb)->saddr, /* XXX */
L
Linus Torvalds 已提交
776 777
				      arg.iov[0].iov_len, IPPROTO_TCP, 0);
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
778 779
	if (oif)
		arg.bound_dev_if = oif;
780
	arg.tos = tos;
781
	arg.uid = sock_net_uid(net, sk_fullsock(sk) ? sk : NULL);
782
	local_bh_disable();
783 784
	ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk),
			      skb, &TCP_SKB_CB(skb)->header.h4.opt,
785 786
			      ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
			      &arg, arg.iov[0].iov_len);
L
Linus Torvalds 已提交
787

E
Eric Dumazet 已提交
788
	__TCP_INC_STATS(net, TCP_MIB_OUTSEGS);
789
	local_bh_enable();
L
Linus Torvalds 已提交
790 791 792 793
}

static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
794
	struct inet_timewait_sock *tw = inet_twsk(sk);
795
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
796

797
	tcp_v4_send_ack(sk, skb,
798
			tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
799
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
800
			tcp_time_stamp + tcptw->tw_ts_offset,
801 802
			tcptw->tw_ts_recent,
			tw->tw_bound_dev_if,
803
			tcp_twsk_md5_key(tcptw),
804 805
			tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0,
			tw->tw_tos
806
			);
L
Linus Torvalds 已提交
807

808
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
809 810
}

811
static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
812
				  struct request_sock *req)
L
Linus Torvalds 已提交
813
{
814 815 816
	/* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
	 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
	 */
817 818 819
	u32 seq = (sk->sk_state == TCP_LISTEN) ? tcp_rsk(req)->snt_isn + 1 :
					     tcp_sk(sk)->snd_nxt;

820 821 822 823 824
	/* 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:
	 */
825
	tcp_v4_send_ack(sk, skb, seq,
826 827
			tcp_rsk(req)->rcv_nxt,
			req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale,
828
			tcp_time_stamp + tcp_rsk(req)->ts_off,
829 830
			req->ts_recent,
			0,
E
Eric Dumazet 已提交
831 832
			tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&ip_hdr(skb)->daddr,
					  AF_INET),
833 834
			inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0,
			ip_hdr(skb)->tos);
L
Linus Torvalds 已提交
835 836 837
}

/*
838
 *	Send a SYN-ACK after having received a SYN.
839
 *	This still operates on a request_sock only, not on a big
L
Linus Torvalds 已提交
840 841
 *	socket.
 */
842
static int tcp_v4_send_synack(const struct sock *sk, struct dst_entry *dst,
843
			      struct flowi *fl,
844
			      struct request_sock *req,
845
			      struct tcp_fastopen_cookie *foc,
846
			      enum tcp_synack_type synack_type)
L
Linus Torvalds 已提交
847
{
848
	const struct inet_request_sock *ireq = inet_rsk(req);
849
	struct flowi4 fl4;
L
Linus Torvalds 已提交
850
	int err = -1;
851
	struct sk_buff *skb;
L
Linus Torvalds 已提交
852 853

	/* First, grab a route. */
854
	if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL)
855
		return -1;
L
Linus Torvalds 已提交
856

857
	skb = tcp_make_synack(sk, dst, req, foc, synack_type);
L
Linus Torvalds 已提交
858 859

	if (skb) {
860
		__tcp_v4_send_check(skb, ireq->ir_loc_addr, ireq->ir_rmt_addr);
L
Linus Torvalds 已提交
861

862 863
		err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr,
					    ireq->ir_rmt_addr,
864
					    ireq->opt);
865
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
866 867 868 869 870 871
	}

	return err;
}

/*
872
 *	IPv4 request_sock destructor.
L
Linus Torvalds 已提交
873
 */
874
static void tcp_v4_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
875
{
J
Jesper Juhl 已提交
876
	kfree(inet_rsk(req)->opt);
L
Linus Torvalds 已提交
877 878
}

879 880 881 882 883 884 885 886
#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.  */
887
struct tcp_md5sig_key *tcp_md5_do_lookup(const struct sock *sk,
E
Eric Dumazet 已提交
888 889
					 const union tcp_md5_addr *addr,
					 int family)
890
{
891
	const struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
892 893
	struct tcp_md5sig_key *key;
	unsigned int size = sizeof(struct in_addr);
894
	const struct tcp_md5sig_info *md5sig;
895

896 897
	/* caller either holds rcu_read_lock() or socket lock */
	md5sig = rcu_dereference_check(tp->md5sig_info,
898
				       lockdep_sock_is_held(sk));
899
	if (!md5sig)
900
		return NULL;
E
Eric Dumazet 已提交
901 902 903 904
#if IS_ENABLED(CONFIG_IPV6)
	if (family == AF_INET6)
		size = sizeof(struct in6_addr);
#endif
905
	hlist_for_each_entry_rcu(key, &md5sig->head, node) {
E
Eric Dumazet 已提交
906 907 908 909
		if (key->family != family)
			continue;
		if (!memcmp(&key->addr, addr, size))
			return key;
910 911 912
	}
	return NULL;
}
E
Eric Dumazet 已提交
913
EXPORT_SYMBOL(tcp_md5_do_lookup);
914

915
struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk,
916
					 const struct sock *addr_sk)
917
{
918
	const union tcp_md5_addr *addr;
E
Eric Dumazet 已提交
919

920
	addr = (const union tcp_md5_addr *)&addr_sk->sk_daddr;
E
Eric Dumazet 已提交
921
	return tcp_md5_do_lookup(sk, addr, AF_INET);
922 923 924 925
}
EXPORT_SYMBOL(tcp_v4_md5_lookup);

/* This can be called on a newly created socket, from other files */
E
Eric Dumazet 已提交
926 927
int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
		   int family, const u8 *newkey, u8 newkeylen, gfp_t gfp)
928 929
{
	/* Add Key to the list */
930
	struct tcp_md5sig_key *key;
931
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
932
	struct tcp_md5sig_info *md5sig;
933

934
	key = tcp_md5_do_lookup(sk, addr, family);
935 936
	if (key) {
		/* Pre-existing entry - just update that one. */
E
Eric Dumazet 已提交
937
		memcpy(key->key, newkey, newkeylen);
938
		key->keylen = newkeylen;
E
Eric Dumazet 已提交
939 940
		return 0;
	}
941

942
	md5sig = rcu_dereference_protected(tp->md5sig_info,
943
					   lockdep_sock_is_held(sk));
E
Eric Dumazet 已提交
944 945 946
	if (!md5sig) {
		md5sig = kmalloc(sizeof(*md5sig), gfp);
		if (!md5sig)
947 948
			return -ENOMEM;

E
Eric Dumazet 已提交
949 950
		sk_nocaps_add(sk, NETIF_F_GSO_MASK);
		INIT_HLIST_HEAD(&md5sig->head);
951
		rcu_assign_pointer(tp->md5sig_info, md5sig);
E
Eric Dumazet 已提交
952
	}
953

954
	key = sock_kmalloc(sk, sizeof(*key), gfp);
E
Eric Dumazet 已提交
955 956
	if (!key)
		return -ENOMEM;
957
	if (!tcp_alloc_md5sig_pool()) {
958
		sock_kfree_s(sk, key, sizeof(*key));
E
Eric Dumazet 已提交
959
		return -ENOMEM;
960
	}
E
Eric Dumazet 已提交
961 962 963 964 965 966 967 968

	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);
969 970
	return 0;
}
E
Eric Dumazet 已提交
971
EXPORT_SYMBOL(tcp_md5_do_add);
972

E
Eric Dumazet 已提交
973
int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family)
974
{
E
Eric Dumazet 已提交
975 976
	struct tcp_md5sig_key *key;

977
	key = tcp_md5_do_lookup(sk, addr, family);
E
Eric Dumazet 已提交
978 979 980
	if (!key)
		return -ENOENT;
	hlist_del_rcu(&key->node);
981
	atomic_sub(sizeof(*key), &sk->sk_omem_alloc);
E
Eric Dumazet 已提交
982 983
	kfree_rcu(key, rcu);
	return 0;
984
}
E
Eric Dumazet 已提交
985
EXPORT_SYMBOL(tcp_md5_do_del);
986

987
static void tcp_clear_md5_list(struct sock *sk)
988 989
{
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
990
	struct tcp_md5sig_key *key;
991
	struct hlist_node *n;
992
	struct tcp_md5sig_info *md5sig;
993

994 995
	md5sig = rcu_dereference_protected(tp->md5sig_info, 1);

996
	hlist_for_each_entry_safe(key, n, &md5sig->head, node) {
E
Eric Dumazet 已提交
997
		hlist_del_rcu(&key->node);
998
		atomic_sub(sizeof(*key), &sk->sk_omem_alloc);
E
Eric Dumazet 已提交
999
		kfree_rcu(key, rcu);
1000 1001 1002
	}
}

1003 1004
static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval,
				 int optlen)
1005 1006 1007 1008 1009 1010 1011
{
	struct tcp_md5sig cmd;
	struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr;

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

1012
	if (copy_from_user(&cmd, optval, sizeof(cmd)))
1013 1014 1015 1016 1017
		return -EFAULT;

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

1018
	if (!cmd.tcpm_keylen)
E
Eric Dumazet 已提交
1019 1020
		return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr,
				      AF_INET);
1021 1022 1023 1024

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

E
Eric Dumazet 已提交
1025 1026 1027
	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);
1028 1029
}

1030 1031 1032
static int tcp_v4_md5_hash_headers(struct tcp_md5sig_pool *hp,
				   __be32 daddr, __be32 saddr,
				   const struct tcphdr *th, int nbytes)
1033 1034
{
	struct tcp4_pseudohdr *bp;
1035
	struct scatterlist sg;
1036
	struct tcphdr *_th;
1037

1038
	bp = hp->scratch;
1039 1040 1041
	bp->saddr = saddr;
	bp->daddr = daddr;
	bp->pad = 0;
1042
	bp->protocol = IPPROTO_TCP;
1043
	bp->len = cpu_to_be16(nbytes);
1044

1045 1046 1047 1048 1049 1050 1051
	_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
1052
	return crypto_ahash_update(hp->md5_req);
1053 1054
}

E
Eric Dumazet 已提交
1055
static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
1056
			       __be32 daddr, __be32 saddr, const struct tcphdr *th)
1057 1058
{
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
1059
	struct ahash_request *req;
1060 1061 1062 1063

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
1064
	req = hp->md5_req;
1065

H
Herbert Xu 已提交
1066
	if (crypto_ahash_init(req))
1067
		goto clear_hash;
1068
	if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2))
1069 1070 1071
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
H
Herbert Xu 已提交
1072 1073
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
1074 1075 1076 1077
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
1078

1079 1080 1081 1082
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
1083
	return 1;
1084 1085
}

1086 1087
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
1088
			const struct sk_buff *skb)
1089
{
1090
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
1091
	struct ahash_request *req;
E
Eric Dumazet 已提交
1092
	const struct tcphdr *th = tcp_hdr(skb);
1093 1094
	__be32 saddr, daddr;

1095 1096 1097
	if (sk) { /* valid for establish/request sockets */
		saddr = sk->sk_rcv_saddr;
		daddr = sk->sk_daddr;
1098
	} else {
1099 1100 1101
		const struct iphdr *iph = ip_hdr(skb);
		saddr = iph->saddr;
		daddr = iph->daddr;
1102
	}
1103 1104 1105 1106

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
1107
	req = hp->md5_req;
1108

H
Herbert Xu 已提交
1109
	if (crypto_ahash_init(req))
1110 1111
		goto clear_hash;

1112
	if (tcp_v4_md5_hash_headers(hp, daddr, saddr, th, skb->len))
1113 1114 1115 1116 1117
		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
1118 1119
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
1120 1121 1122 1123 1124 1125 1126 1127 1128 1129
		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;
1130
}
1131
EXPORT_SYMBOL(tcp_v4_md5_hash_skb);
1132

1133 1134
#endif

1135
/* Called with rcu_read_lock() */
1136
static bool tcp_v4_inbound_md5_hash(const struct sock *sk,
1137
				    const struct sk_buff *skb)
1138
{
1139
#ifdef CONFIG_TCP_MD5SIG
1140 1141 1142 1143 1144 1145 1146 1147
	/*
	 * 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.
	 */
1148
	const __u8 *hash_location = NULL;
1149
	struct tcp_md5sig_key *hash_expected;
1150
	const struct iphdr *iph = ip_hdr(skb);
1151
	const struct tcphdr *th = tcp_hdr(skb);
1152 1153 1154
	int genhash;
	unsigned char newhash[16];

E
Eric Dumazet 已提交
1155 1156
	hash_expected = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&iph->saddr,
					  AF_INET);
1157
	hash_location = tcp_parse_md5sig_option(th);
1158 1159 1160

	/* We've parsed the options - do we have a hash? */
	if (!hash_expected && !hash_location)
E
Eric Dumazet 已提交
1161
		return false;
1162 1163

	if (hash_expected && !hash_location) {
1164
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
E
Eric Dumazet 已提交
1165
		return true;
1166 1167 1168
	}

	if (!hash_expected && hash_location) {
1169
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
E
Eric Dumazet 已提交
1170
		return true;
1171 1172 1173 1174 1175
	}

	/* Okay, so this is hash_expected and hash_location -
	 * so we need to calculate the checksum.
	 */
1176 1177
	genhash = tcp_v4_md5_hash_skb(newhash,
				      hash_expected,
1178
				      NULL, skb);
1179 1180

	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
1181
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE);
1182 1183 1184 1185 1186
		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
1187
		return true;
1188
	}
E
Eric Dumazet 已提交
1189
	return false;
1190
#endif
1191 1192
	return false;
}
1193

1194 1195
static void tcp_v4_init_req(struct request_sock *req,
			    const struct sock *sk_listener,
1196 1197 1198 1199
			    struct sk_buff *skb)
{
	struct inet_request_sock *ireq = inet_rsk(req);

1200 1201
	sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr);
	sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr);
1202 1203 1204
	ireq->opt = tcp_v4_save_options(skb);
}

1205 1206
static struct dst_entry *tcp_v4_route_req(const struct sock *sk,
					  struct flowi *fl,
1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221
					  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;
}

1222
struct request_sock_ops tcp_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
1223
	.family		=	PF_INET,
1224
	.obj_size	=	sizeof(struct tcp_request_sock),
1225
	.rtx_syn_ack	=	tcp_rtx_synack,
1226 1227
	.send_ack	=	tcp_v4_reqsk_send_ack,
	.destructor	=	tcp_v4_reqsk_destructor,
L
Linus Torvalds 已提交
1228
	.send_reset	=	tcp_v4_send_reset,
S
stephen hemminger 已提交
1229
	.syn_ack_timeout =	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
1230 1231
};

1232
static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = {
1233
	.mss_clamp	=	TCP_MSS_DEFAULT,
1234
#ifdef CONFIG_TCP_MD5SIG
1235
	.req_md5_lookup	=	tcp_v4_md5_lookup,
1236
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1237
#endif
1238
	.init_req	=	tcp_v4_init_req,
1239 1240 1241
#ifdef CONFIG_SYN_COOKIES
	.cookie_init_seq =	cookie_v4_init_sequence,
#endif
1242
	.route_req	=	tcp_v4_route_req,
1243
	.init_seq	=	tcp_v4_init_sequence,
1244
	.send_synack	=	tcp_v4_send_synack,
1245
};
1246

L
Linus Torvalds 已提交
1247 1248 1249
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
1250
	if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
L
Linus Torvalds 已提交
1251 1252
		goto drop;

O
Octavian Purdila 已提交
1253 1254
	return tcp_conn_request(&tcp_request_sock_ops,
				&tcp_request_sock_ipv4_ops, sk, skb);
L
Linus Torvalds 已提交
1255 1256

drop:
1257
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1258 1259
	return 0;
}
E
Eric Dumazet 已提交
1260
EXPORT_SYMBOL(tcp_v4_conn_request);
L
Linus Torvalds 已提交
1261 1262 1263 1264 1265 1266


/*
 * The three way handshake has completed - we got a valid synack -
 * now create the new socket.
 */
1267
struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
1268
				  struct request_sock *req,
1269 1270 1271
				  struct dst_entry *dst,
				  struct request_sock *req_unhash,
				  bool *own_req)
L
Linus Torvalds 已提交
1272
{
1273
	struct inet_request_sock *ireq;
L
Linus Torvalds 已提交
1274 1275 1276
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1277 1278 1279
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1280
	struct ip_options_rcu *inet_opt;
L
Linus Torvalds 已提交
1281 1282 1283 1284 1285 1286

	if (sk_acceptq_is_full(sk))
		goto exit_overflow;

	newsk = tcp_create_openreq_child(sk, req, skb);
	if (!newsk)
1287
		goto exit_nonewsk;
L
Linus Torvalds 已提交
1288

1289
	newsk->sk_gso_type = SKB_GSO_TCPV4;
1290
	inet_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1291 1292 1293

	newtp		      = tcp_sk(newsk);
	newinet		      = inet_sk(newsk);
1294
	ireq		      = inet_rsk(req);
1295 1296
	sk_daddr_set(newsk, ireq->ir_rmt_addr);
	sk_rcv_saddr_set(newsk, ireq->ir_loc_addr);
1297
	newsk->sk_bound_dev_if = ireq->ir_iif;
1298
	newinet->inet_saddr	      = ireq->ir_loc_addr;
1299 1300
	inet_opt	      = ireq->opt;
	rcu_assign_pointer(newinet->inet_opt, inet_opt);
1301
	ireq->opt	      = NULL;
1302
	newinet->mc_index     = inet_iif(skb);
1303
	newinet->mc_ttl	      = ip_hdr(skb)->ttl;
1304
	newinet->rcv_tos      = ip_hdr(skb)->tos;
1305
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
1306 1307
	if (inet_opt)
		inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
E
Eric Dumazet 已提交
1308
	newinet->inet_id = newtp->write_seq ^ jiffies;
L
Linus Torvalds 已提交
1309

E
Eric Dumazet 已提交
1310 1311 1312 1313 1314 1315 1316
	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() */
	}
1317 1318
	sk_setup_caps(newsk, dst);

1319 1320
	tcp_ca_openreq_child(newsk, dst);

L
Linus Torvalds 已提交
1321
	tcp_sync_mss(newsk, dst_mtu(dst));
1322
	newtp->advmss = dst_metric_advmss(dst);
1323 1324 1325 1326
	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
1327 1328
	tcp_initialize_rcv_mss(newsk);

1329 1330
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
E
Eric Dumazet 已提交
1331 1332
	key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&newinet->inet_daddr,
				AF_INET);
1333
	if (key) {
1334 1335 1336 1337 1338 1339
		/*
		 * 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
1340 1341
		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
1342
		sk_nocaps_add(newsk, NETIF_F_GSO_MASK);
1343 1344 1345
	}
#endif

1346 1347
	if (__inet_inherit_port(sk, newsk) < 0)
		goto put_and_exit;
1348
	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
E
Eric Dumazet 已提交
1349
	if (*own_req)
1350
		tcp_move_syn(newtp, req);
L
Linus Torvalds 已提交
1351 1352 1353 1354

	return newsk;

exit_overflow:
1355
	NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1356 1357
exit_nonewsk:
	dst_release(dst);
L
Linus Torvalds 已提交
1358
exit:
1359
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1360
	return NULL;
1361
put_and_exit:
1362 1363
	inet_csk_prepare_forced_close(newsk);
	tcp_done(newsk);
1364
	goto exit;
L
Linus Torvalds 已提交
1365
}
E
Eric Dumazet 已提交
1366
EXPORT_SYMBOL(tcp_v4_syn_recv_sock);
L
Linus Torvalds 已提交
1367

1368
static struct sock *tcp_v4_cookie_check(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
1369
{
1370
#ifdef CONFIG_SYN_COOKIES
1371
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1372

1373
	if (!th->syn)
C
Cong Wang 已提交
1374
		sk = cookie_v4_check(sk, skb);
L
Linus Torvalds 已提交
1375 1376 1377 1378 1379
#endif
	return sk;
}

/* The socket must have it's spinlock held when we get
1380
 * here, unless it is a TCP_LISTEN socket.
L
Linus Torvalds 已提交
1381 1382 1383 1384 1385 1386 1387 1388
 *
 * 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)
{
1389 1390
	struct sock *rsk;

L
Linus Torvalds 已提交
1391
	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1392 1393
		struct dst_entry *dst = sk->sk_rx_dst;

1394
		sock_rps_save_rxhash(sk, skb);
1395
		sk_mark_napi_id(sk, skb);
1396
		if (dst) {
E
Eric Dumazet 已提交
1397
			if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif ||
1398
			    !dst->ops->check(dst, 0)) {
1399 1400 1401 1402
				dst_release(dst);
				sk->sk_rx_dst = NULL;
			}
		}
1403
		tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
L
Linus Torvalds 已提交
1404 1405 1406
		return 0;
	}

E
Eric Dumazet 已提交
1407
	if (tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1408 1409 1410
		goto csum_err;

	if (sk->sk_state == TCP_LISTEN) {
1411 1412
		struct sock *nsk = tcp_v4_cookie_check(sk, skb);

L
Linus Torvalds 已提交
1413 1414 1415
		if (!nsk)
			goto discard;
		if (nsk != sk) {
1416
			sock_rps_save_rxhash(nsk, skb);
1417
			sk_mark_napi_id(nsk, skb);
1418 1419
			if (tcp_child_process(sk, nsk, skb)) {
				rsk = nsk;
L
Linus Torvalds 已提交
1420
				goto reset;
1421
			}
L
Linus Torvalds 已提交
1422 1423
			return 0;
		}
1424
	} else
1425
		sock_rps_save_rxhash(sk, skb);
1426

1427
	if (tcp_rcv_state_process(sk, skb)) {
1428
		rsk = sk;
L
Linus Torvalds 已提交
1429
		goto reset;
1430
	}
L
Linus Torvalds 已提交
1431 1432 1433
	return 0;

reset:
1434
	tcp_v4_send_reset(rsk, skb);
L
Linus Torvalds 已提交
1435 1436 1437 1438 1439 1440 1441 1442 1443 1444
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:
1445 1446
	TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS);
	TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1447 1448
	goto discard;
}
E
Eric Dumazet 已提交
1449
EXPORT_SYMBOL(tcp_v4_do_rcv);
L
Linus Torvalds 已提交
1450

1451
void tcp_v4_early_demux(struct sk_buff *skb)
D
David S. Miller 已提交
1452 1453 1454 1455 1456 1457
{
	const struct iphdr *iph;
	const struct tcphdr *th;
	struct sock *sk;

	if (skb->pkt_type != PACKET_HOST)
1458
		return;
D
David S. Miller 已提交
1459

1460
	if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr)))
1461
		return;
D
David S. Miller 已提交
1462 1463

	iph = ip_hdr(skb);
1464
	th = tcp_hdr(skb);
D
David S. Miller 已提交
1465 1466

	if (th->doff < sizeof(struct tcphdr) / 4)
1467
		return;
D
David S. Miller 已提交
1468

1469
	sk = __inet_lookup_established(dev_net(skb->dev), &tcp_hashinfo,
D
David S. Miller 已提交
1470
				       iph->saddr, th->source,
1471
				       iph->daddr, ntohs(th->dest),
E
Eric Dumazet 已提交
1472
				       skb->skb_iif);
D
David S. Miller 已提交
1473 1474 1475
	if (sk) {
		skb->sk = sk;
		skb->destructor = sock_edemux;
1476
		if (sk_fullsock(sk)) {
1477
			struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst);
E
Eric Dumazet 已提交
1478

D
David S. Miller 已提交
1479 1480
			if (dst)
				dst = dst_check(dst, 0);
1481
			if (dst &&
E
Eric Dumazet 已提交
1482
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
1483
				skb_dst_set_noref(skb, dst);
D
David S. Miller 已提交
1484 1485 1486 1487
		}
	}
}

E
Eric Dumazet 已提交
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
/* 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;

1506 1507 1508 1509 1510 1511 1512 1513 1514
	/* 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
1515
		skb_dst_force_safe(skb);
1516

E
Eric Dumazet 已提交
1517 1518
	__skb_queue_tail(&tp->ucopy.prequeue, skb);
	tp->ucopy.memory += skb->truesize;
E
Eric Dumazet 已提交
1519 1520
	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
1521 1522 1523
		struct sk_buff *skb1;

		BUG_ON(sock_owned_by_user(sk));
E
Eric Dumazet 已提交
1524 1525
		__NET_ADD_STATS(sock_net(sk), LINUX_MIB_TCPPREQUEUEDROPPED,
				skb_queue_len(&tp->ucopy.prequeue));
E
Eric Dumazet 已提交
1526

E
Eric Dumazet 已提交
1527
		while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL)
E
Eric Dumazet 已提交
1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542
			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
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 1568 1569 1570
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);

1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585
int tcp_filter(struct sock *sk, struct sk_buff *skb)
{
	struct tcphdr *th = (struct tcphdr *)skb->data;
	unsigned int eaten = skb->len;
	int err;

	err = sk_filter_trim_cap(sk, skb, th->doff * 4);
	if (!err) {
		eaten -= skb->len;
		TCP_SKB_CB(skb)->end_seq -= eaten;
	}
	return err;
}
EXPORT_SYMBOL(tcp_filter);

L
Linus Torvalds 已提交
1586 1587 1588 1589 1590 1591
/*
 *	From tcp_input.c
 */

int tcp_v4_rcv(struct sk_buff *skb)
{
1592
	struct net *net = dev_net(skb->dev);
1593
	const struct iphdr *iph;
1594
	const struct tcphdr *th;
1595
	bool refcounted;
L
Linus Torvalds 已提交
1596 1597 1598 1599 1600 1601 1602
	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
1603
	__TCP_INC_STATS(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1604 1605 1606 1607

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

1608
	th = (const struct tcphdr *)skb->data;
L
Linus Torvalds 已提交
1609

1610
	if (unlikely(th->doff < sizeof(struct tcphdr) / 4))
L
Linus Torvalds 已提交
1611 1612 1613 1614 1615 1616
		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
1617
	 * provided case of th->doff==0 is eliminated.
L
Linus Torvalds 已提交
1618
	 * So, we defer the checks. */
1619 1620

	if (skb_checksum_init(skb, IPPROTO_TCP, inet_compute_pseudo))
1621
		goto csum_error;
L
Linus Torvalds 已提交
1622

1623
	th = (const struct tcphdr *)skb->data;
1624
	iph = ip_hdr(skb);
1625 1626 1627 1628 1629 1630 1631
	/* 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
1632 1633 1634 1635
	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);
1636
	TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
1637
	TCP_SKB_CB(skb)->tcp_tw_isn = 0;
1638
	TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph);
L
Linus Torvalds 已提交
1639 1640
	TCP_SKB_CB(skb)->sacked	 = 0;

1641
lookup:
1642
	sk = __inet_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th), th->source,
1643
			       th->dest, &refcounted);
L
Linus Torvalds 已提交
1644 1645 1646
	if (!sk)
		goto no_tcp_socket;

E
Eric Dumazet 已提交
1647 1648 1649 1650
process:
	if (sk->sk_state == TCP_TIME_WAIT)
		goto do_time_wait;

1651 1652
	if (sk->sk_state == TCP_NEW_SYN_RECV) {
		struct request_sock *req = inet_reqsk(sk);
1653
		struct sock *nsk;
1654 1655

		sk = req->rsk_listener;
1656
		if (unlikely(tcp_v4_inbound_md5_hash(sk, skb))) {
1657
			sk_drops_add(sk, skb);
1658 1659 1660
			reqsk_put(req);
			goto discard_it;
		}
1661
		if (unlikely(sk->sk_state != TCP_LISTEN)) {
1662
			inet_csk_reqsk_queue_drop_and_put(sk, req);
1663 1664
			goto lookup;
		}
1665 1666 1667
		/* We own a reference on the listener, increase it again
		 * as we might lose it too soon.
		 */
1668
		sock_hold(sk);
1669
		refcounted = true;
1670
		nsk = tcp_check_req(sk, skb, req, false);
1671 1672
		if (!nsk) {
			reqsk_put(req);
1673
			goto discard_and_relse;
1674 1675 1676 1677 1678
		}
		if (nsk == sk) {
			reqsk_put(req);
		} else if (tcp_child_process(sk, nsk, skb)) {
			tcp_v4_send_reset(nsk, skb);
1679
			goto discard_and_relse;
1680
		} else {
1681
			sock_put(sk);
1682 1683 1684
			return 0;
		}
	}
1685
	if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
1686
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
1687
		goto discard_and_relse;
1688
	}
1689

L
Linus Torvalds 已提交
1690 1691
	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;
1692 1693 1694 1695

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

1696
	nf_reset(skb);
L
Linus Torvalds 已提交
1697

1698
	if (tcp_filter(sk, skb))
L
Linus Torvalds 已提交
1699
		goto discard_and_relse;
1700 1701
	th = (const struct tcphdr *)skb->data;
	iph = ip_hdr(skb);
L
Linus Torvalds 已提交
1702 1703 1704

	skb->dev = NULL;

1705 1706 1707 1708 1709 1710 1711
	if (sk->sk_state == TCP_LISTEN) {
		ret = tcp_v4_do_rcv(sk, skb);
		goto put_and_return;
	}

	sk_incoming_cpu_update(sk);

1712
	bh_lock_sock_nested(sk);
1713
	tcp_segs_in(tcp_sk(sk), skb);
L
Linus Torvalds 已提交
1714 1715
	ret = 0;
	if (!sock_owned_by_user(sk)) {
D
Dan Williams 已提交
1716
		if (!tcp_prequeue(sk, skb))
L
Linus Torvalds 已提交
1717
			ret = tcp_v4_do_rcv(sk, skb);
E
Eric Dumazet 已提交
1718
	} else if (tcp_add_backlog(sk, skb)) {
Z
Zhu Yi 已提交
1719 1720
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1721 1722
	bh_unlock_sock(sk);

1723
put_and_return:
1724 1725
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1726 1727 1728 1729 1730 1731 1732

	return ret;

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

E
Eric Dumazet 已提交
1733
	if (tcp_checksum_complete(skb)) {
1734
csum_error:
E
Eric Dumazet 已提交
1735
		__TCP_INC_STATS(net, TCP_MIB_CSUMERRORS);
L
Linus Torvalds 已提交
1736
bad_packet:
E
Eric Dumazet 已提交
1737
		__TCP_INC_STATS(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1738
	} else {
1739
		tcp_v4_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1740 1741 1742 1743 1744
	}

discard_it:
	/* Discard frame. */
	kfree_skb(skb);
1745
	return 0;
L
Linus Torvalds 已提交
1746 1747

discard_and_relse:
1748
	sk_drops_add(sk, skb);
1749 1750
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1751 1752 1753 1754
	goto discard_it;

do_time_wait:
	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1755
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1756 1757 1758
		goto discard_it;
	}

1759 1760 1761
	if (tcp_checksum_complete(skb)) {
		inet_twsk_put(inet_twsk(sk));
		goto csum_error;
L
Linus Torvalds 已提交
1762
	}
1763
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1764
	case TCP_TW_SYN: {
1765
		struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev),
1766 1767
							&tcp_hashinfo, skb,
							__tcp_hdrlen(th),
1768
							iph->saddr, th->source,
1769
							iph->daddr, th->dest,
1770
							inet_iif(skb));
L
Linus Torvalds 已提交
1771
		if (sk2) {
1772
			inet_twsk_deschedule_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1773
			sk = sk2;
1774
			refcounted = false;
L
Linus Torvalds 已提交
1775 1776 1777 1778 1779 1780 1781 1782
			goto process;
		}
		/* Fall through to ACK */
	}
	case TCP_TW_ACK:
		tcp_v4_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
1783 1784 1785
		tcp_v4_send_reset(sk, skb);
		inet_twsk_deschedule_put(inet_twsk(sk));
		goto discard_it;
L
Linus Torvalds 已提交
1786 1787 1788 1789 1790
	case TCP_TW_SUCCESS:;
	}
	goto discard_it;
}

1791 1792 1793 1794 1795
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
1796

1797
void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
E
Eric Dumazet 已提交
1798 1799 1800
{
	struct dst_entry *dst = skb_dst(skb);

E
Eric Dumazet 已提交
1801
	if (dst && dst_hold_safe(dst)) {
1802 1803 1804
		sk->sk_rx_dst = dst;
		inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
	}
E
Eric Dumazet 已提交
1805
}
1806
EXPORT_SYMBOL(inet_sk_rx_dst_set);
E
Eric Dumazet 已提交
1807

1808
const struct inet_connection_sock_af_ops ipv4_specific = {
1809 1810 1811
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
E
Eric Dumazet 已提交
1812
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1813 1814 1815 1816 1817 1818 1819
	.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),
1820
	.bind_conflict	   = inet_csk_bind_conflict,
1821
#ifdef CONFIG_COMPAT
1822 1823
	.compat_setsockopt = compat_ip_setsockopt,
	.compat_getsockopt = compat_ip_getsockopt,
1824
#endif
1825
	.mtu_reduced	   = tcp_v4_mtu_reduced,
L
Linus Torvalds 已提交
1826
};
E
Eric Dumazet 已提交
1827
EXPORT_SYMBOL(ipv4_specific);
L
Linus Torvalds 已提交
1828

1829
#ifdef CONFIG_TCP_MD5SIG
1830
static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = {
1831
	.md5_lookup		= tcp_v4_md5_lookup,
1832
	.calc_md5_hash		= tcp_v4_md5_hash_skb,
1833 1834
	.md5_parse		= tcp_v4_parse_md5_keys,
};
1835
#endif
1836

L
Linus Torvalds 已提交
1837 1838 1839 1840 1841
/* 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)
{
1842
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1843

1844
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1845

1846
	icsk->icsk_af_ops = &ipv4_specific;
1847

1848
#ifdef CONFIG_TCP_MD5SIG
1849
	tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific;
1850
#endif
L
Linus Torvalds 已提交
1851 1852 1853 1854

	return 0;
}

1855
void tcp_v4_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1856 1857 1858 1859 1860
{
	struct tcp_sock *tp = tcp_sk(sk);

	tcp_clear_xmit_timers(sk);

1861
	tcp_cleanup_congestion_control(sk);
1862

L
Linus Torvalds 已提交
1863
	/* Cleanup up the write buffer. */
1864
	tcp_write_queue_purge(sk);
L
Linus Torvalds 已提交
1865 1866

	/* Cleans up our, hopefully empty, out_of_order_queue. */
1867
	skb_rbtree_purge(&tp->out_of_order_queue);
L
Linus Torvalds 已提交
1868

1869 1870 1871
#ifdef CONFIG_TCP_MD5SIG
	/* Clean up the MD5 key list, if any */
	if (tp->md5sig_info) {
E
Eric Dumazet 已提交
1872
		tcp_clear_md5_list(sk);
1873
		kfree_rcu(tp->md5sig_info, rcu);
1874 1875 1876
		tp->md5sig_info = NULL;
	}
#endif
C
Chris Leech 已提交
1877

L
Linus Torvalds 已提交
1878 1879 1880 1881
	/* Clean prequeue, it must be empty really */
	__skb_queue_purge(&tp->ucopy.prequeue);

	/* Clean up a referenced TCP bind bucket. */
1882
	if (inet_csk(sk)->icsk_bind_hash)
1883
		inet_put_port(sk);
L
Linus Torvalds 已提交
1884

1885
	BUG_ON(tp->fastopen_rsk);
1886

1887 1888
	/* If socket is aborted during connect operation */
	tcp_free_fastopen_req(tp);
1889
	tcp_saved_syn_free(tp);
1890

E
Eric Dumazet 已提交
1891
	local_bh_disable();
1892
	sk_sockets_allocated_dec(sk);
E
Eric Dumazet 已提交
1893
	local_bh_enable();
L
Linus Torvalds 已提交
1894 1895 1896 1897 1898 1899
}
EXPORT_SYMBOL(tcp_v4_destroy_sock);

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

1900 1901 1902 1903 1904
/*
 * 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
1905 1906
static void *listening_get_next(struct seq_file *seq, void *cur)
{
J
Jianjun Kong 已提交
1907
	struct tcp_iter_state *st = seq->private;
1908
	struct net *net = seq_file_net(seq);
1909 1910
	struct inet_listen_hashbucket *ilb;
	struct sock *sk = cur;
L
Linus Torvalds 已提交
1911 1912

	if (!sk) {
1913
get_head:
1914
		ilb = &tcp_hashinfo.listening_hash[st->bucket];
1915
		spin_lock(&ilb->lock);
1916
		sk = sk_head(&ilb->head);
1917
		st->offset = 0;
L
Linus Torvalds 已提交
1918 1919
		goto get_sk;
	}
1920
	ilb = &tcp_hashinfo.listening_hash[st->bucket];
L
Linus Torvalds 已提交
1921
	++st->num;
1922
	++st->offset;
L
Linus Torvalds 已提交
1923

1924
	sk = sk_next(sk);
L
Linus Torvalds 已提交
1925
get_sk:
1926
	sk_for_each_from(sk) {
1927 1928
		if (!net_eq(sock_net(sk), net))
			continue;
1929 1930
		if (sk->sk_family == st->family)
			return sk;
L
Linus Torvalds 已提交
1931
	}
1932
	spin_unlock(&ilb->lock);
1933
	st->offset = 0;
1934 1935 1936
	if (++st->bucket < INET_LHTABLE_SIZE)
		goto get_head;
	return NULL;
L
Linus Torvalds 已提交
1937 1938 1939 1940
}

static void *listening_get_idx(struct seq_file *seq, loff_t *pos)
{
1941 1942 1943 1944 1945 1946
	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
1947 1948 1949 1950 1951 1952 1953 1954

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

E
Eric Dumazet 已提交
1955
static inline bool empty_bucket(const struct tcp_iter_state *st)
1956
{
E
Eric Dumazet 已提交
1957
	return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain);
1958 1959
}

1960 1961 1962 1963
/*
 * 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
1964 1965
static void *established_get_first(struct seq_file *seq)
{
J
Jianjun Kong 已提交
1966
	struct tcp_iter_state *st = seq->private;
1967
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
1968 1969
	void *rc = NULL;

1970 1971
	st->offset = 0;
	for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) {
L
Linus Torvalds 已提交
1972
		struct sock *sk;
1973
		struct hlist_nulls_node *node;
1974
		spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket);
L
Linus Torvalds 已提交
1975

1976 1977 1978 1979
		/* Lockless fast path for the common case of empty buckets */
		if (empty_bucket(st))
			continue;

1980
		spin_lock_bh(lock);
1981
		sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) {
1982
			if (sk->sk_family != st->family ||
1983
			    !net_eq(sock_net(sk), net)) {
L
Linus Torvalds 已提交
1984 1985 1986 1987 1988
				continue;
			}
			rc = sk;
			goto out;
		}
1989
		spin_unlock_bh(lock);
L
Linus Torvalds 已提交
1990 1991 1992 1993 1994 1995 1996 1997
	}
out:
	return rc;
}

static void *established_get_next(struct seq_file *seq, void *cur)
{
	struct sock *sk = cur;
1998
	struct hlist_nulls_node *node;
J
Jianjun Kong 已提交
1999
	struct tcp_iter_state *st = seq->private;
2000
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
2001 2002

	++st->num;
2003
	++st->offset;
L
Linus Torvalds 已提交
2004

E
Eric Dumazet 已提交
2005
	sk = sk_nulls_next(sk);
L
Linus Torvalds 已提交
2006

2007
	sk_nulls_for_each_from(sk, node) {
2008
		if (sk->sk_family == st->family && net_eq(sock_net(sk), net))
E
Eric Dumazet 已提交
2009
			return sk;
L
Linus Torvalds 已提交
2010 2011
	}

E
Eric Dumazet 已提交
2012 2013 2014
	spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
	++st->bucket;
	return established_get_first(seq);
L
Linus Torvalds 已提交
2015 2016 2017 2018
}

static void *established_get_idx(struct seq_file *seq, loff_t pos)
{
2019 2020 2021 2022 2023
	struct tcp_iter_state *st = seq->private;
	void *rc;

	st->bucket = 0;
	rc = established_get_first(seq);
L
Linus Torvalds 已提交
2024 2025 2026 2027

	while (rc && pos) {
		rc = established_get_next(seq, rc);
		--pos;
2028
	}
L
Linus Torvalds 已提交
2029 2030 2031 2032 2033 2034
	return rc;
}

static void *tcp_get_idx(struct seq_file *seq, loff_t pos)
{
	void *rc;
J
Jianjun Kong 已提交
2035
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047

	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;
}

2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065
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
2066
		st->state = TCP_SEQ_STATE_ESTABLISHED;
2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080
		/* 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
2081 2082
static void *tcp_seq_start(struct seq_file *seq, loff_t *pos)
{
J
Jianjun Kong 已提交
2083
	struct tcp_iter_state *st = seq->private;
2084 2085 2086 2087 2088 2089 2090 2091
	void *rc;

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

L
Linus Torvalds 已提交
2092 2093
	st->state = TCP_SEQ_STATE_LISTENING;
	st->num = 0;
2094 2095 2096 2097 2098 2099 2100
	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
2101 2102 2103 2104
}

static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
2105
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117
	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;
2118 2119
			st->bucket = 0;
			st->offset = 0;
L
Linus Torvalds 已提交
2120 2121 2122 2123 2124 2125 2126 2127 2128
			rc	  = established_get_first(seq);
		}
		break;
	case TCP_SEQ_STATE_ESTABLISHED:
		rc = established_get_next(seq, v);
		break;
	}
out:
	++*pos;
2129
	st->last_pos = *pos;
L
Linus Torvalds 已提交
2130 2131 2132 2133 2134
	return rc;
}

static void tcp_seq_stop(struct seq_file *seq, void *v)
{
J
Jianjun Kong 已提交
2135
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2136 2137 2138 2139

	switch (st->state) {
	case TCP_SEQ_STATE_LISTENING:
		if (v != SEQ_START_TOKEN)
2140
			spin_unlock(&tcp_hashinfo.listening_hash[st->bucket].lock);
L
Linus Torvalds 已提交
2141 2142 2143
		break;
	case TCP_SEQ_STATE_ESTABLISHED:
		if (v)
2144
			spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
L
Linus Torvalds 已提交
2145 2146 2147 2148
		break;
	}
}

2149
int tcp_seq_open(struct inode *inode, struct file *file)
L
Linus Torvalds 已提交
2150
{
A
Al Viro 已提交
2151
	struct tcp_seq_afinfo *afinfo = PDE_DATA(inode);
L
Linus Torvalds 已提交
2152
	struct tcp_iter_state *s;
2153
	int err;
L
Linus Torvalds 已提交
2154

2155 2156 2157 2158
	err = seq_open_net(inode, file, &afinfo->seq_ops,
			  sizeof(struct tcp_iter_state));
	if (err < 0)
		return err;
2159

2160
	s = ((struct seq_file *)file->private_data)->private;
L
Linus Torvalds 已提交
2161
	s->family		= afinfo->family;
S
stephen hemminger 已提交
2162
	s->last_pos		= 0;
2163 2164
	return 0;
}
2165
EXPORT_SYMBOL(tcp_seq_open);
2166

2167
int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo)
L
Linus Torvalds 已提交
2168 2169 2170 2171
{
	int rc = 0;
	struct proc_dir_entry *p;

2172 2173 2174 2175
	afinfo->seq_ops.start		= tcp_seq_start;
	afinfo->seq_ops.next		= tcp_seq_next;
	afinfo->seq_ops.stop		= tcp_seq_stop;

2176
	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
2177
			     afinfo->seq_fops, afinfo);
2178
	if (!p)
L
Linus Torvalds 已提交
2179 2180 2181
		rc = -ENOMEM;
	return rc;
}
E
Eric Dumazet 已提交
2182
EXPORT_SYMBOL(tcp_proc_register);
L
Linus Torvalds 已提交
2183

2184
void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo)
L
Linus Torvalds 已提交
2185
{
2186
	remove_proc_entry(afinfo->name, net->proc_net);
L
Linus Torvalds 已提交
2187
}
E
Eric Dumazet 已提交
2188
EXPORT_SYMBOL(tcp_proc_unregister);
L
Linus Torvalds 已提交
2189

2190
static void get_openreq4(const struct request_sock *req,
E
Eric Dumazet 已提交
2191
			 struct seq_file *f, int i)
L
Linus Torvalds 已提交
2192
{
2193
	const struct inet_request_sock *ireq = inet_rsk(req);
2194
	long delta = req->rsk_timer.expires - jiffies;
L
Linus Torvalds 已提交
2195

2196
	seq_printf(f, "%4d: %08X:%04X %08X:%04X"
2197
		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %u %d %pK",
L
Linus Torvalds 已提交
2198
		i,
2199
		ireq->ir_loc_addr,
2200
		ireq->ir_num,
2201 2202
		ireq->ir_rmt_addr,
		ntohs(ireq->ir_rmt_port),
L
Linus Torvalds 已提交
2203 2204 2205
		TCP_SYN_RECV,
		0, 0, /* could print option size, but that is af dependent. */
		1,    /* timers active (only the expire timer) */
2206
		jiffies_delta_to_clock_t(delta),
2207
		req->num_timeout,
E
Eric Dumazet 已提交
2208 2209
		from_kuid_munged(seq_user_ns(f),
				 sock_i_uid(req->rsk_listener)),
L
Linus Torvalds 已提交
2210 2211
		0,  /* non standard timer */
		0, /* open_requests have no inode */
2212
		0,
2213
		req);
L
Linus Torvalds 已提交
2214 2215
}

2216
static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i)
L
Linus Torvalds 已提交
2217 2218 2219
{
	int timer_active;
	unsigned long timer_expires;
2220
	const struct tcp_sock *tp = tcp_sk(sk);
2221
	const struct inet_connection_sock *icsk = inet_csk(sk);
2222
	const struct inet_sock *inet = inet_sk(sk);
2223
	const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq;
E
Eric Dumazet 已提交
2224 2225 2226 2227
	__be32 dest = inet->inet_daddr;
	__be32 src = inet->inet_rcv_saddr;
	__u16 destp = ntohs(inet->inet_dport);
	__u16 srcp = ntohs(inet->inet_sport);
2228
	int rx_queue;
2229
	int state;
L
Linus Torvalds 已提交
2230

N
Nandita Dukkipati 已提交
2231 2232 2233
	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
2234
		timer_active	= 1;
2235 2236
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
2237
		timer_active	= 4;
2238
		timer_expires	= icsk->icsk_timeout;
2239
	} else if (timer_pending(&sk->sk_timer)) {
L
Linus Torvalds 已提交
2240
		timer_active	= 2;
2241
		timer_expires	= sk->sk_timer.expires;
L
Linus Torvalds 已提交
2242 2243 2244 2245 2246
	} else {
		timer_active	= 0;
		timer_expires = jiffies;
	}

2247 2248
	state = sk_state_load(sk);
	if (state == TCP_LISTEN)
2249 2250
		rx_queue = sk->sk_ack_backlog;
	else
2251 2252
		/* Because we don't lock the socket,
		 * we might find a transient negative value.
2253 2254 2255
		 */
		rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0);

2256
	seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX "
2257
			"%08X %5u %8d %lu %d %pK %lu %lu %u %u %d",
2258
		i, src, srcp, dest, destp, state,
2259
		tp->write_seq - tp->snd_una,
2260
		rx_queue,
L
Linus Torvalds 已提交
2261
		timer_active,
2262
		jiffies_delta_to_clock_t(timer_expires - jiffies),
2263
		icsk->icsk_retransmits,
2264
		from_kuid_munged(seq_user_ns(f), sock_i_uid(sk)),
2265
		icsk->icsk_probes_out,
2266 2267
		sock_i_ino(sk),
		atomic_read(&sk->sk_refcnt), sk,
2268 2269
		jiffies_to_clock_t(icsk->icsk_rto),
		jiffies_to_clock_t(icsk->icsk_ack.ato),
2270
		(icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
L
Linus Torvalds 已提交
2271
		tp->snd_cwnd,
2272 2273
		state == TCP_LISTEN ?
		    fastopenq->max_qlen :
2274
		    (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh));
L
Linus Torvalds 已提交
2275 2276
}

2277
static void get_timewait4_sock(const struct inet_timewait_sock *tw,
2278
			       struct seq_file *f, int i)
L
Linus Torvalds 已提交
2279
{
2280
	long delta = tw->tw_timer.expires - jiffies;
2281
	__be32 dest, src;
L
Linus Torvalds 已提交
2282 2283 2284 2285 2286 2287 2288
	__u16 destp, srcp;

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

2289
	seq_printf(f, "%4d: %08X:%04X %08X:%04X"
2290
		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK",
L
Linus Torvalds 已提交
2291
		i, src, srcp, dest, destp, tw->tw_substate, 0, 0,
2292
		3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
2293
		atomic_read(&tw->tw_refcnt), tw);
L
Linus Torvalds 已提交
2294 2295 2296 2297 2298 2299
}

#define TMPSZ 150

static int tcp4_seq_show(struct seq_file *seq, void *v)
{
J
Jianjun Kong 已提交
2300
	struct tcp_iter_state *st;
E
Eric Dumazet 已提交
2301
	struct sock *sk = v;
L
Linus Torvalds 已提交
2302

2303
	seq_setwidth(seq, TMPSZ - 1);
L
Linus Torvalds 已提交
2304
	if (v == SEQ_START_TOKEN) {
2305
		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
L
Linus Torvalds 已提交
2306 2307 2308 2309 2310 2311
			   "rx_queue tr tm->when retrnsmt   uid  timeout "
			   "inode");
		goto out;
	}
	st = seq->private;

2312 2313 2314
	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
2315
		get_openreq4(v, seq, st->num);
2316 2317
	else
		get_tcp4_sock(v, seq, st->num);
L
Linus Torvalds 已提交
2318
out:
2319
	seq_pad(seq, '\n');
L
Linus Torvalds 已提交
2320 2321 2322
	return 0;
}

2323 2324 2325 2326 2327 2328 2329 2330
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
2331 2332 2333
static struct tcp_seq_afinfo tcp4_seq_afinfo = {
	.name		= "tcp",
	.family		= AF_INET,
2334
	.seq_fops	= &tcp_afinfo_seq_fops,
2335 2336 2337
	.seq_ops	= {
		.show		= tcp4_seq_show,
	},
L
Linus Torvalds 已提交
2338 2339
};

2340
static int __net_init tcp4_proc_init_net(struct net *net)
2341 2342 2343 2344
{
	return tcp_proc_register(net, &tcp4_seq_afinfo);
}

2345
static void __net_exit tcp4_proc_exit_net(struct net *net)
2346 2347 2348 2349 2350 2351 2352 2353 2354
{
	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
2355 2356
int __init tcp4_proc_init(void)
{
2357
	return register_pernet_subsys(&tcp4_net_ops);
L
Linus Torvalds 已提交
2358 2359 2360 2361
}

void tcp4_proc_exit(void)
{
2362
	unregister_pernet_subsys(&tcp4_net_ops);
L
Linus Torvalds 已提交
2363 2364 2365 2366 2367 2368 2369 2370 2371
}
#endif /* CONFIG_PROC_FS */

struct proto tcp_prot = {
	.name			= "TCP",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v4_connect,
	.disconnect		= tcp_disconnect,
2372
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
2373 2374 2375 2376 2377 2378 2379
	.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,
2380 2381
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
2382
	.backlog_rcv		= tcp_v4_do_rcv,
E
Eric Dumazet 已提交
2383
	.release_cb		= tcp_release_cb,
2384 2385 2386
	.hash			= inet_hash,
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
2387
	.enter_memory_pressure	= tcp_enter_memory_pressure,
2388
	.stream_memory_free	= tcp_stream_memory_free,
L
Linus Torvalds 已提交
2389
	.sockets_allocated	= &tcp_sockets_allocated,
2390
	.orphan_count		= &tcp_orphan_count,
L
Linus Torvalds 已提交
2391 2392
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
2393
	.sysctl_mem		= sysctl_tcp_mem,
L
Linus Torvalds 已提交
2394 2395 2396 2397
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp_sock),
2398
	.slab_flags		= SLAB_DESTROY_BY_RCU,
2399
	.twsk_prot		= &tcp_timewait_sock_ops,
2400
	.rsk_prot		= &tcp_request_sock_ops,
2401
	.h.hashinfo		= &tcp_hashinfo,
2402
	.no_autobind		= true,
2403 2404 2405
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
G
Glauber Costa 已提交
2406
#endif
2407
	.diag_destroy		= tcp_abort,
L
Linus Torvalds 已提交
2408
};
E
Eric Dumazet 已提交
2409
EXPORT_SYMBOL(tcp_prot);
L
Linus Torvalds 已提交
2410

2411 2412 2413 2414 2415 2416 2417 2418 2419
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);
}

2420 2421
static int __net_init tcp_sk_init(struct net *net)
{
2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434
	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;
2435
		sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
2436 2437
		*per_cpu_ptr(net->ipv4.tcp_sk, cpu) = sk;
	}
2438

2439
	net->ipv4.sysctl_tcp_ecn = 2;
2440 2441
	net->ipv4.sysctl_tcp_ecn_fallback = 1;

F
Fan Du 已提交
2442
	net->ipv4.sysctl_tcp_base_mss = TCP_BASE_MSS;
2443
	net->ipv4.sysctl_tcp_probe_threshold = TCP_PROBE_THRESHOLD;
2444
	net->ipv4.sysctl_tcp_probe_interval = TCP_PROBE_INTERVAL;
2445

2446
	net->ipv4.sysctl_tcp_keepalive_time = TCP_KEEPALIVE_TIME;
2447
	net->ipv4.sysctl_tcp_keepalive_probes = TCP_KEEPALIVE_PROBES;
2448
	net->ipv4.sysctl_tcp_keepalive_intvl = TCP_KEEPALIVE_INTVL;
2449

2450
	net->ipv4.sysctl_tcp_syn_retries = TCP_SYN_RETRIES;
2451
	net->ipv4.sysctl_tcp_synack_retries = TCP_SYNACK_RETRIES;
2452
	net->ipv4.sysctl_tcp_syncookies = 1;
2453
	net->ipv4.sysctl_tcp_reordering = TCP_FASTRETRANS_THRESH;
2454
	net->ipv4.sysctl_tcp_retries1 = TCP_RETR1;
2455
	net->ipv4.sysctl_tcp_retries2 = TCP_RETR2;
2456
	net->ipv4.sysctl_tcp_orphan_retries = 0;
2457
	net->ipv4.sysctl_tcp_fin_timeout = TCP_FIN_TIMEOUT;
2458
	net->ipv4.sysctl_tcp_notsent_lowat = UINT_MAX;
2459
	net->ipv4.sysctl_tcp_tw_reuse = 0;
2460

2461 2462 2463 2464
	net->ipv4.tcp_death_row.sysctl_tw_recycle = 0;
	net->ipv4.tcp_death_row.sysctl_max_tw_buckets = (tcp_hashinfo.ehash_mask + 1) / 2;
	net->ipv4.tcp_death_row.hashinfo = &tcp_hashinfo;

2465
	return 0;
2466 2467 2468 2469
fail:
	tcp_sk_exit(net);

	return res;
E
Eric W. Biederman 已提交
2470 2471 2472 2473
}

static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list)
{
2474
	inet_twsk_purge(&tcp_hashinfo, AF_INET);
2475 2476 2477
}

static struct pernet_operations __net_initdata tcp_sk_ops = {
E
Eric W. Biederman 已提交
2478 2479 2480
       .init	   = tcp_sk_init,
       .exit	   = tcp_sk_exit,
       .exit_batch = tcp_sk_exit_batch,
2481 2482
};

2483
void __init tcp_v4_init(void)
L
Linus Torvalds 已提交
2484
{
2485
	if (register_pernet_subsys(&tcp_sk_ops))
L
Linus Torvalds 已提交
2486 2487
		panic("Failed to create the TCP control socket.\n");
}