tcp_ipv4.c 61.0 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>
G
Glauber Costa 已提交
76
#include <net/tcp_memcontrol.h>
77
#include <net/busy_poll.h>
L
Linus Torvalds 已提交
78 79 80 81 82 83 84

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

85 86 87
#include <linux/crypto.h>
#include <linux/scatterlist.h>

88 89
int sysctl_tcp_tw_reuse __read_mostly;
int sysctl_tcp_low_latency __read_mostly;
E
Eric Dumazet 已提交
90
EXPORT_SYMBOL(sysctl_tcp_low_latency);
L
Linus Torvalds 已提交
91

92
#ifdef CONFIG_TCP_MD5SIG
E
Eric Dumazet 已提交
93
static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
94
			       __be32 daddr, __be32 saddr, const struct tcphdr *th);
95 96
#endif

97
struct inet_hashinfo tcp_hashinfo;
E
Eric Dumazet 已提交
98
EXPORT_SYMBOL(tcp_hashinfo);
L
Linus Torvalds 已提交
99

100
static  __u32 tcp_v4_init_sequence(const struct sk_buff *skb)
L
Linus Torvalds 已提交
101
{
102 103
	return secure_tcp_sequence_number(ip_hdr(skb)->daddr,
					  ip_hdr(skb)->saddr,
104 105
					  tcp_hdr(skb)->dest,
					  tcp_hdr(skb)->source);
L
Linus Torvalds 已提交
106 107
}

108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124
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 &&
125
	    (!twp || (sysctl_tcp_tw_reuse &&
126
			     get_seconds() - tcptw->tw_ts_recent_stamp > 1))) {
127 128 129 130 131 132 133 134 135 136 137 138 139
		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
140 141 142
/* This will initiate an outgoing connection. */
int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
{
143
	struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
L
Linus Torvalds 已提交
144 145
	struct inet_sock *inet = inet_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
146
	__be16 orig_sport, orig_dport;
147
	__be32 daddr, nexthop;
148
	struct flowi4 *fl4;
149
	struct rtable *rt;
L
Linus Torvalds 已提交
150
	int err;
151
	struct ip_options_rcu *inet_opt;
L
Linus Torvalds 已提交
152 153 154 155 156 157 158 159

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

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

	nexthop = daddr = usin->sin_addr.s_addr;
160 161 162
	inet_opt = rcu_dereference_protected(inet->inet_opt,
					     sock_owned_by_user(sk));
	if (inet_opt && inet_opt->opt.srr) {
L
Linus Torvalds 已提交
163 164
		if (!daddr)
			return -EINVAL;
165
		nexthop = inet_opt->opt.faddr;
L
Linus Torvalds 已提交
166 167
	}

168 169
	orig_sport = inet->inet_sport;
	orig_dport = usin->sin_port;
170 171
	fl4 = &inet->cork.fl.u.ip4;
	rt = ip_route_connect(fl4, nexthop, inet->inet_saddr,
172 173
			      RT_CONN_FLAGS(sk), sk->sk_bound_dev_if,
			      IPPROTO_TCP,
174
			      orig_sport, orig_dport, sk);
175 176 177
	if (IS_ERR(rt)) {
		err = PTR_ERR(rt);
		if (err == -ENETUNREACH)
178
			IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
179
		return err;
180
	}
L
Linus Torvalds 已提交
181 182 183 184 185 186

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

187
	if (!inet_opt || !inet_opt->opt.srr)
188
		daddr = fl4->daddr;
L
Linus Torvalds 已提交
189

E
Eric Dumazet 已提交
190
	if (!inet->inet_saddr)
191
		inet->inet_saddr = fl4->saddr;
192
	sk_rcv_saddr_set(sk, inet->inet_saddr);
L
Linus Torvalds 已提交
193

E
Eric Dumazet 已提交
194
	if (tp->rx_opt.ts_recent_stamp && inet->inet_daddr != daddr) {
L
Linus Torvalds 已提交
195 196 197
		/* Reset inherited state */
		tp->rx_opt.ts_recent	   = 0;
		tp->rx_opt.ts_recent_stamp = 0;
P
Pavel Emelyanov 已提交
198 199
		if (likely(!tp->repair))
			tp->write_seq	   = 0;
L
Linus Torvalds 已提交
200 201
	}

202
	if (tcp_death_row.sysctl_tw_recycle &&
203 204
	    !tp->rx_opt.ts_recent_stamp && fl4->daddr == daddr)
		tcp_fetch_timewait_stamp(sk, &rt->dst);
L
Linus Torvalds 已提交
205

E
Eric Dumazet 已提交
206
	inet->inet_dport = usin->sin_port;
207
	sk_daddr_set(sk, daddr);
L
Linus Torvalds 已提交
208

209
	inet_csk(sk)->icsk_ext_hdr_len = 0;
210 211
	if (inet_opt)
		inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
L
Linus Torvalds 已提交
212

213
	tp->rx_opt.mss_clamp = TCP_MSS_DEFAULT;
L
Linus Torvalds 已提交
214 215 216 217 218 219 220

	/* 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);
221
	err = inet_hash_connect(&tcp_death_row, sk);
L
Linus Torvalds 已提交
222 223 224
	if (err)
		goto failure;

225
	sk_set_txhash(sk);
226

227
	rt = ip_route_newports(fl4, rt, orig_sport, orig_dport,
228 229 230 231
			       inet->inet_sport, inet->inet_dport, sk);
	if (IS_ERR(rt)) {
		err = PTR_ERR(rt);
		rt = NULL;
L
Linus Torvalds 已提交
232
		goto failure;
233
	}
L
Linus Torvalds 已提交
234
	/* OK, now commit destination to socket.  */
235
	sk->sk_gso_type = SKB_GSO_TCPV4;
236
	sk_setup_caps(sk, &rt->dst);
L
Linus Torvalds 已提交
237

P
Pavel Emelyanov 已提交
238
	if (!tp->write_seq && likely(!tp->repair))
E
Eric Dumazet 已提交
239 240 241
		tp->write_seq = secure_tcp_sequence_number(inet->inet_saddr,
							   inet->inet_daddr,
							   inet->inet_sport,
L
Linus Torvalds 已提交
242 243
							   usin->sin_port);

E
Eric Dumazet 已提交
244
	inet->inet_id = tp->write_seq ^ jiffies;
L
Linus Torvalds 已提交
245

A
Andrey Vagin 已提交
246
	err = tcp_connect(sk);
P
Pavel Emelyanov 已提交
247

L
Linus Torvalds 已提交
248 249 250 251 252 253 254
	rt = NULL;
	if (err)
		goto failure;

	return 0;

failure:
255 256 257 258
	/*
	 * This unhashes the socket and releases the local port,
	 * if necessary.
	 */
L
Linus Torvalds 已提交
259 260 261
	tcp_set_state(sk, TCP_CLOSE);
	ip_rt_put(rt);
	sk->sk_route_caps = 0;
E
Eric Dumazet 已提交
262
	inet->inet_dport = 0;
L
Linus Torvalds 已提交
263 264
	return err;
}
E
Eric Dumazet 已提交
265
EXPORT_SYMBOL(tcp_v4_connect);
L
Linus Torvalds 已提交
266 267

/*
268 269 270
 * 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
271
 */
272
void tcp_v4_mtu_reduced(struct sock *sk)
L
Linus Torvalds 已提交
273 274 275
{
	struct dst_entry *dst;
	struct inet_sock *inet = inet_sk(sk);
276
	u32 mtu = tcp_sk(sk)->mtu_info;
L
Linus Torvalds 已提交
277

278 279
	dst = inet_csk_update_pmtu(sk, mtu);
	if (!dst)
L
Linus Torvalds 已提交
280 281 282 283 284 285 286 287 288 289 290
		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 &&
291
	    ip_sk_accept_pmtu(sk) &&
292
	    inet_csk(sk)->icsk_pmtu_cookie > mtu) {
L
Linus Torvalds 已提交
293 294 295 296 297 298 299 300 301 302
		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 */
}
303
EXPORT_SYMBOL(tcp_v4_mtu_reduced);
L
Linus Torvalds 已提交
304

305 306 307 308
static void do_redirect(struct sk_buff *skb, struct sock *sk)
{
	struct dst_entry *dst = __sk_dst_check(sk, 0);

309
	if (dst)
310
		dst->ops->redirect(dst, sk, skb);
311 312
}

313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333

/* handle ICMP messages on TCP_NEW_SYN_RECV request sockets */
void tcp_req_err(struct sock *sk, u32 seq)
{
	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.
	 */
	WARN_ON(req->sk);

	if (seq != tcp_rsk(req)->snt_isn) {
		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
	} else {
		/*
		 * 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
334
		inet_csk_reqsk_queue_drop(req->rsk_listener, req);
335
		NET_INC_STATS_BH(net, LINUX_MIB_LISTENDROPS);
336
	}
337
	reqsk_put(req);
338 339 340
}
EXPORT_SYMBOL(tcp_req_err);

L
Linus Torvalds 已提交
341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356
/*
 * 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.
 *
 */

357
void tcp_v4_err(struct sk_buff *icmp_skb, u32 info)
L
Linus Torvalds 已提交
358
{
359
	const struct iphdr *iph = (const struct iphdr *)icmp_skb->data;
360
	struct tcphdr *th = (struct tcphdr *)(icmp_skb->data + (iph->ihl << 2));
361
	struct inet_connection_sock *icsk;
L
Linus Torvalds 已提交
362 363
	struct tcp_sock *tp;
	struct inet_sock *inet;
364 365
	const int type = icmp_hdr(icmp_skb)->type;
	const int code = icmp_hdr(icmp_skb)->code;
L
Linus Torvalds 已提交
366
	struct sock *sk;
367
	struct sk_buff *skb;
368 369
	struct request_sock *fastopen;
	__u32 seq, snd_una;
370
	__u32 remaining;
L
Linus Torvalds 已提交
371
	int err;
372
	struct net *net = dev_net(icmp_skb->dev);
L
Linus Torvalds 已提交
373

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

	bh_lock_sock(sk);
	/* If too many ICMPs get dropped on busy
	 * servers this needs to be solved differently.
392 393
	 * 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
394
	 */
395 396 397 398
	if (sock_owned_by_user(sk)) {
		if (!(type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED))
			NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
	}
L
Linus Torvalds 已提交
399 400 401
	if (sk->sk_state == TCP_CLOSE)
		goto out;

402 403 404 405 406
	if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
		NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
		goto out;
	}

407
	icsk = inet_csk(sk);
L
Linus Torvalds 已提交
408
	tp = tcp_sk(sk);
409 410 411
	/* 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
412
	if (sk->sk_state != TCP_LISTEN &&
413
	    !between(seq, snd_una, tp->snd_nxt)) {
414
		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
415 416 417 418
		goto out;
	}

	switch (type) {
419 420 421
	case ICMP_REDIRECT:
		do_redirect(icmp_skb, sk);
		goto out;
L
Linus Torvalds 已提交
422 423 424 425 426 427 428 429 430 431 432
	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) */
433 434 435 436 437 438 439
			/* 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;

440
			tp->mtu_info = info;
441
			if (!sock_owned_by_user(sk)) {
442
				tcp_v4_mtu_reduced(sk);
443 444 445 446
			} else {
				if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, &tp->tsq_flags))
					sock_hold(sk);
			}
L
Linus Torvalds 已提交
447 448 449 450
			goto out;
		}

		err = icmp_err_convert[code].errno;
451 452 453 454 455
		/* 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 ||
456
		    !icsk->icsk_backoff || fastopen)
457 458
			break;

459 460 461
		if (sock_owned_by_user(sk))
			break;

462
		icsk->icsk_backoff--;
463 464 465
		icsk->icsk_rto = tp->srtt_us ? __tcp_set_rto(tp) :
					       TCP_TIMEOUT_INIT;
		icsk->icsk_rto = inet_csk_rto_backoff(icsk, TCP_RTO_MAX);
466 467 468 469

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

E
Eric Dumazet 已提交
470 471 472
		remaining = icsk->icsk_rto -
			    min(icsk->icsk_rto,
				tcp_time_stamp - tcp_skb_timestamp(skb));
473 474 475 476 477 478 479 480 481 482

		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
483 484 485 486 487 488 489 490 491 492
		break;
	case ICMP_TIME_EXCEEDED:
		err = EHOSTUNREACH;
		break;
	default:
		goto out;
	}

	switch (sk->sk_state) {
	case TCP_SYN_SENT:
493 494 495 496
	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.
		 */
497
		if (fastopen && !fastopen->sk)
498 499
			break;

L
Linus Torvalds 已提交
500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540
		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);
}

541
void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr)
L
Linus Torvalds 已提交
542
{
543
	struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
544

545
	if (skb->ip_summed == CHECKSUM_PARTIAL) {
546
		th->check = ~tcp_v4_check(skb->len, saddr, daddr, 0);
547
		skb->csum_start = skb_transport_header(skb) - skb->head;
A
Al Viro 已提交
548
		skb->csum_offset = offsetof(struct tcphdr, check);
L
Linus Torvalds 已提交
549
	} else {
550
		th->check = tcp_v4_check(skb->len, saddr, daddr,
551
					 csum_partial(th,
L
Linus Torvalds 已提交
552 553 554 555 556
						      th->doff << 2,
						      skb->csum));
	}
}

557
/* This routine computes an IPv4 TCP checksum. */
558
void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb)
559
{
560
	const struct inet_sock *inet = inet_sk(sk);
561 562 563

	__tcp_v4_send_check(skb, inet->inet_saddr, inet->inet_daddr);
}
E
Eric Dumazet 已提交
564
EXPORT_SYMBOL(tcp_v4_send_check);
565

L
Linus Torvalds 已提交
566 567 568 569 570 571 572 573 574 575 576 577 578
/*
 *	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.
 */

579
static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
580
{
581
	const struct tcphdr *th = tcp_hdr(skb);
582 583 584
	struct {
		struct tcphdr th;
#ifdef CONFIG_TCP_MD5SIG
585
		__be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)];
586 587
#endif
	} rep;
L
Linus Torvalds 已提交
588
	struct ip_reply_arg arg;
589
#ifdef CONFIG_TCP_MD5SIG
590
	struct tcp_md5sig_key *key = NULL;
591 592 593 594
	const __u8 *hash_location = NULL;
	unsigned char newhash[16];
	int genhash;
	struct sock *sk1 = NULL;
595
#endif
596
	struct net *net;
L
Linus Torvalds 已提交
597 598 599 600 601

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

602 603 604 605
	/* 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
606 607 608
		return;

	/* Swap the send and the receive. */
609 610 611 612 613
	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
614 615

	if (th->ack) {
616
		rep.th.seq = th->ack_seq;
L
Linus Torvalds 已提交
617
	} else {
618 619 620
		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
621 622
	}

623
	memset(&arg, 0, sizeof(arg));
624 625 626
	arg.iov[0].iov_base = (unsigned char *)&rep;
	arg.iov[0].iov_len  = sizeof(rep.th);

627
	net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
628
#ifdef CONFIG_TCP_MD5SIG
629
	hash_location = tcp_parse_md5sig_option(th);
630
	if (sk && sk_fullsock(sk)) {
631 632 633
		key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)
					&ip_hdr(skb)->saddr, AF_INET);
	} else if (hash_location) {
634 635 636 637 638 639 640
		/*
		 * 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.
		 */
641
		sk1 = __inet_lookup_listener(net,
642 643
					     &tcp_hashinfo, ip_hdr(skb)->saddr,
					     th->source, ip_hdr(skb)->daddr,
644 645 646 647 648 649 650 651 652 653
					     ntohs(th->source), inet_iif(skb));
		/* don't send rst if it can't find key */
		if (!sk1)
			return;
		rcu_read_lock();
		key = tcp_md5_do_lookup(sk1, (union tcp_md5_addr *)
					&ip_hdr(skb)->saddr, AF_INET);
		if (!key)
			goto release_sk1;

654
		genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, skb);
655 656 657 658
		if (genhash || memcmp(hash_location, newhash, 16) != 0)
			goto release_sk1;
	}

659 660 661 662 663 664 665 666 667
	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;

668
		tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1],
669 670
				     key, ip_hdr(skb)->saddr,
				     ip_hdr(skb)->daddr, &rep.th);
671 672
	}
#endif
673 674
	arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
				      ip_hdr(skb)->saddr, /* XXX */
675
				      arg.iov[0].iov_len, IPPROTO_TCP, 0);
L
Linus Torvalds 已提交
676
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
677 678
	arg.flags = (sk && inet_sk_transparent(sk)) ? IP_REPLY_ARG_NOSRCCHECK : 0;

679
	/* When socket is gone, all binding information is lost.
A
Alexey Kuznetsov 已提交
680 681
	 * routing might fail in this case. No choice here, if we choose to force
	 * input interface, we will misroute in case of asymmetric route.
682
	 */
A
Alexey Kuznetsov 已提交
683 684
	if (sk)
		arg.bound_dev_if = sk->sk_bound_dev_if;
L
Linus Torvalds 已提交
685

686 687 688
	BUILD_BUG_ON(offsetof(struct sock, sk_bound_dev_if) !=
		     offsetof(struct inet_timewait_sock, tw_bound_dev_if));

689
	arg.tos = ip_hdr(skb)->tos;
690 691
	ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk),
			      skb, &TCP_SKB_CB(skb)->header.h4.opt,
692 693
			      ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
			      &arg, arg.iov[0].iov_len);
L
Linus Torvalds 已提交
694

695 696
	TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
	TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS);
697 698 699 700 701 702 703 704

#ifdef CONFIG_TCP_MD5SIG
release_sk1:
	if (sk1) {
		rcu_read_unlock();
		sock_put(sk1);
	}
#endif
L
Linus Torvalds 已提交
705 706 707 708 709 710
}

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

711
static void tcp_v4_send_ack(struct sk_buff *skb, u32 seq, u32 ack,
712
			    u32 win, u32 tsval, u32 tsecr, int oif,
713
			    struct tcp_md5sig_key *key,
714
			    int reply_flags, u8 tos)
L
Linus Torvalds 已提交
715
{
716
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
717 718
	struct {
		struct tcphdr th;
719
		__be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2)
720
#ifdef CONFIG_TCP_MD5SIG
721
			   + (TCPOLEN_MD5SIG_ALIGNED >> 2)
722 723
#endif
			];
L
Linus Torvalds 已提交
724 725
	} rep;
	struct ip_reply_arg arg;
E
Eric Dumazet 已提交
726
	struct net *net = dev_net(skb_dst(skb)->dev);
L
Linus Torvalds 已提交
727 728

	memset(&rep.th, 0, sizeof(struct tcphdr));
729
	memset(&arg, 0, sizeof(arg));
L
Linus Torvalds 已提交
730 731 732

	arg.iov[0].iov_base = (unsigned char *)&rep;
	arg.iov[0].iov_len  = sizeof(rep.th);
733
	if (tsecr) {
734 735 736
		rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				   (TCPOPT_TIMESTAMP << 8) |
				   TCPOLEN_TIMESTAMP);
737 738
		rep.opt[1] = htonl(tsval);
		rep.opt[2] = htonl(tsecr);
739
		arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED;
L
Linus Torvalds 已提交
740 741 742 743 744 745 746 747 748 749 750
	}

	/* 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);

751 752
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
753
		int offset = (tsecr) ? 3 : 0;
754 755 756 757 758 759 760 761

		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;

762
		tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset],
763 764
				    key, ip_hdr(skb)->saddr,
				    ip_hdr(skb)->daddr, &rep.th);
765 766
	}
#endif
767
	arg.flags = reply_flags;
768 769
	arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
				      ip_hdr(skb)->saddr, /* XXX */
L
Linus Torvalds 已提交
770 771
				      arg.iov[0].iov_len, IPPROTO_TCP, 0);
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
772 773
	if (oif)
		arg.bound_dev_if = oif;
774
	arg.tos = tos;
775 776
	ip_send_unicast_reply(*this_cpu_ptr(net->ipv4.tcp_sk),
			      skb, &TCP_SKB_CB(skb)->header.h4.opt,
777 778
			      ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
			      &arg, arg.iov[0].iov_len);
L
Linus Torvalds 已提交
779

780
	TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
L
Linus Torvalds 已提交
781 782 783 784
}

static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
785
	struct inet_timewait_sock *tw = inet_twsk(sk);
786
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
787

788
	tcp_v4_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
789
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
790
			tcp_time_stamp + tcptw->tw_ts_offset,
791 792
			tcptw->tw_ts_recent,
			tw->tw_bound_dev_if,
793
			tcp_twsk_md5_key(tcptw),
794 795
			tw->tw_transparent ? IP_REPLY_ARG_NOSRCCHECK : 0,
			tw->tw_tos
796
			);
L
Linus Torvalds 已提交
797

798
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
799 800
}

801
static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
802
				  struct request_sock *req)
L
Linus Torvalds 已提交
803
{
804 805 806 807 808
	/* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
	 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
	 */
	tcp_v4_send_ack(skb, (sk->sk_state == TCP_LISTEN) ?
			tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt,
809
			tcp_rsk(req)->rcv_nxt, req->rsk_rcv_wnd,
810
			tcp_time_stamp,
811 812
			req->ts_recent,
			0,
E
Eric Dumazet 已提交
813 814
			tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&ip_hdr(skb)->daddr,
					  AF_INET),
815 816
			inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0,
			ip_hdr(skb)->tos);
L
Linus Torvalds 已提交
817 818 819
}

/*
820
 *	Send a SYN-ACK after having received a SYN.
821
 *	This still operates on a request_sock only, not on a big
L
Linus Torvalds 已提交
822 823
 *	socket.
 */
824
static int tcp_v4_send_synack(const struct sock *sk, struct dst_entry *dst,
825
			      struct flowi *fl,
826
			      struct request_sock *req,
827 828
			      struct tcp_fastopen_cookie *foc,
				  bool attach_req)
L
Linus Torvalds 已提交
829
{
830
	const struct inet_request_sock *ireq = inet_rsk(req);
831
	struct flowi4 fl4;
L
Linus Torvalds 已提交
832
	int err = -1;
833
	struct sk_buff *skb;
L
Linus Torvalds 已提交
834 835

	/* First, grab a route. */
836
	if (!dst && (dst = inet_csk_route_req(sk, &fl4, req)) == NULL)
837
		return -1;
L
Linus Torvalds 已提交
838

839
	skb = tcp_make_synack(sk, dst, req, foc, attach_req);
L
Linus Torvalds 已提交
840 841

	if (skb) {
842
		__tcp_v4_send_check(skb, ireq->ir_loc_addr, ireq->ir_rmt_addr);
L
Linus Torvalds 已提交
843

844 845
		err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr,
					    ireq->ir_rmt_addr,
846
					    ireq->opt);
847
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
848 849 850 851 852 853
	}

	return err;
}

/*
854
 *	IPv4 request_sock destructor.
L
Linus Torvalds 已提交
855
 */
856
static void tcp_v4_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
857
{
J
Jesper Juhl 已提交
858
	kfree(inet_rsk(req)->opt);
L
Linus Torvalds 已提交
859 860 861
}


862 863 864 865 866 867 868 869
#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.  */
870
struct tcp_md5sig_key *tcp_md5_do_lookup(const struct sock *sk,
E
Eric Dumazet 已提交
871 872
					 const union tcp_md5_addr *addr,
					 int family)
873
{
874
	const struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
875 876
	struct tcp_md5sig_key *key;
	unsigned int size = sizeof(struct in_addr);
877
	const struct tcp_md5sig_info *md5sig;
878

879 880
	/* caller either holds rcu_read_lock() or socket lock */
	md5sig = rcu_dereference_check(tp->md5sig_info,
881
				       sock_owned_by_user(sk) ||
882
				       lockdep_is_held((spinlock_t *)&sk->sk_lock.slock));
883
	if (!md5sig)
884
		return NULL;
E
Eric Dumazet 已提交
885 886 887 888
#if IS_ENABLED(CONFIG_IPV6)
	if (family == AF_INET6)
		size = sizeof(struct in6_addr);
#endif
889
	hlist_for_each_entry_rcu(key, &md5sig->head, node) {
E
Eric Dumazet 已提交
890 891 892 893
		if (key->family != family)
			continue;
		if (!memcmp(&key->addr, addr, size))
			return key;
894 895 896
	}
	return NULL;
}
E
Eric Dumazet 已提交
897
EXPORT_SYMBOL(tcp_md5_do_lookup);
898

899
struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk,
900
					 const struct sock *addr_sk)
901
{
902
	const union tcp_md5_addr *addr;
E
Eric Dumazet 已提交
903

904
	addr = (const union tcp_md5_addr *)&addr_sk->sk_daddr;
E
Eric Dumazet 已提交
905
	return tcp_md5_do_lookup(sk, addr, AF_INET);
906 907 908 909
}
EXPORT_SYMBOL(tcp_v4_md5_lookup);

/* This can be called on a newly created socket, from other files */
E
Eric Dumazet 已提交
910 911
int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
		   int family, const u8 *newkey, u8 newkeylen, gfp_t gfp)
912 913
{
	/* Add Key to the list */
914
	struct tcp_md5sig_key *key;
915
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
916
	struct tcp_md5sig_info *md5sig;
917

918
	key = tcp_md5_do_lookup(sk, addr, family);
919 920
	if (key) {
		/* Pre-existing entry - just update that one. */
E
Eric Dumazet 已提交
921
		memcpy(key->key, newkey, newkeylen);
922
		key->keylen = newkeylen;
E
Eric Dumazet 已提交
923 924
		return 0;
	}
925

926
	md5sig = rcu_dereference_protected(tp->md5sig_info,
E
Eric Dumazet 已提交
927 928
					   sock_owned_by_user(sk) ||
					   lockdep_is_held(&sk->sk_lock.slock));
E
Eric Dumazet 已提交
929 930 931
	if (!md5sig) {
		md5sig = kmalloc(sizeof(*md5sig), gfp);
		if (!md5sig)
932 933
			return -ENOMEM;

E
Eric Dumazet 已提交
934 935
		sk_nocaps_add(sk, NETIF_F_GSO_MASK);
		INIT_HLIST_HEAD(&md5sig->head);
936
		rcu_assign_pointer(tp->md5sig_info, md5sig);
E
Eric Dumazet 已提交
937
	}
938

939
	key = sock_kmalloc(sk, sizeof(*key), gfp);
E
Eric Dumazet 已提交
940 941
	if (!key)
		return -ENOMEM;
942
	if (!tcp_alloc_md5sig_pool()) {
943
		sock_kfree_s(sk, key, sizeof(*key));
E
Eric Dumazet 已提交
944
		return -ENOMEM;
945
	}
E
Eric Dumazet 已提交
946 947 948 949 950 951 952 953

	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);
954 955
	return 0;
}
E
Eric Dumazet 已提交
956
EXPORT_SYMBOL(tcp_md5_do_add);
957

E
Eric Dumazet 已提交
958
int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family)
959
{
E
Eric Dumazet 已提交
960 961
	struct tcp_md5sig_key *key;

962
	key = tcp_md5_do_lookup(sk, addr, family);
E
Eric Dumazet 已提交
963 964 965
	if (!key)
		return -ENOENT;
	hlist_del_rcu(&key->node);
966
	atomic_sub(sizeof(*key), &sk->sk_omem_alloc);
E
Eric Dumazet 已提交
967 968
	kfree_rcu(key, rcu);
	return 0;
969
}
E
Eric Dumazet 已提交
970
EXPORT_SYMBOL(tcp_md5_do_del);
971

972
static void tcp_clear_md5_list(struct sock *sk)
973 974
{
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
975
	struct tcp_md5sig_key *key;
976
	struct hlist_node *n;
977
	struct tcp_md5sig_info *md5sig;
978

979 980
	md5sig = rcu_dereference_protected(tp->md5sig_info, 1);

981
	hlist_for_each_entry_safe(key, n, &md5sig->head, node) {
E
Eric Dumazet 已提交
982
		hlist_del_rcu(&key->node);
983
		atomic_sub(sizeof(*key), &sk->sk_omem_alloc);
E
Eric Dumazet 已提交
984
		kfree_rcu(key, rcu);
985 986 987
	}
}

988 989
static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval,
				 int optlen)
990 991 992 993 994 995 996
{
	struct tcp_md5sig cmd;
	struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr;

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

997
	if (copy_from_user(&cmd, optval, sizeof(cmd)))
998 999 1000 1001 1002
		return -EFAULT;

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

1003
	if (!cmd.tcpm_keylen)
E
Eric Dumazet 已提交
1004 1005
		return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr,
				      AF_INET);
1006 1007 1008 1009

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

E
Eric Dumazet 已提交
1010 1011 1012
	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);
1013 1014
}

1015 1016
static int tcp_v4_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
					__be32 daddr, __be32 saddr, int nbytes)
1017 1018
{
	struct tcp4_pseudohdr *bp;
1019
	struct scatterlist sg;
1020 1021 1022 1023

	bp = &hp->md5_blk.ip4;

	/*
1024
	 * 1. the TCP pseudo-header (in the order: source IP address,
1025 1026 1027 1028 1029 1030
	 * destination IP address, zero-padded protocol number, and
	 * segment length)
	 */
	bp->saddr = saddr;
	bp->daddr = daddr;
	bp->pad = 0;
1031
	bp->protocol = IPPROTO_TCP;
1032
	bp->len = cpu_to_be16(nbytes);
1033

1034 1035 1036 1037
	sg_init_one(&sg, bp, sizeof(*bp));
	return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
}

E
Eric Dumazet 已提交
1038
static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
1039
			       __be32 daddr, __be32 saddr, const struct tcphdr *th)
1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057
{
	struct tcp_md5sig_pool *hp;
	struct hash_desc *desc;

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
	desc = &hp->md5_desc;

	if (crypto_hash_init(desc))
		goto clear_hash;
	if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2))
		goto clear_hash;
	if (tcp_md5_hash_header(hp, th))
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
	if (crypto_hash_final(desc, md5_hash))
1058 1059 1060 1061
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
1062

1063 1064 1065 1066
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
1067
	return 1;
1068 1069
}

1070 1071
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
1072
			const struct sk_buff *skb)
1073
{
1074 1075
	struct tcp_md5sig_pool *hp;
	struct hash_desc *desc;
E
Eric Dumazet 已提交
1076
	const struct tcphdr *th = tcp_hdr(skb);
1077 1078
	__be32 saddr, daddr;

1079 1080 1081
	if (sk) { /* valid for establish/request sockets */
		saddr = sk->sk_rcv_saddr;
		daddr = sk->sk_daddr;
1082
	} else {
1083 1084 1085
		const struct iphdr *iph = ip_hdr(skb);
		saddr = iph->saddr;
		daddr = iph->daddr;
1086
	}
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
	desc = &hp->md5_desc;

	if (crypto_hash_init(desc))
		goto clear_hash;

	if (tcp_v4_md5_hash_pseudoheader(hp, daddr, saddr, skb->len))
		goto clear_hash;
	if (tcp_md5_hash_header(hp, th))
		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;
	if (crypto_hash_final(desc, md5_hash))
		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;
1115
}
1116
EXPORT_SYMBOL(tcp_v4_md5_hash_skb);
1117

1118 1119
#endif

1120
/* Called with rcu_read_lock() */
1121
static bool tcp_v4_inbound_md5_hash(const struct sock *sk,
1122
				    const struct sk_buff *skb)
1123
{
1124
#ifdef CONFIG_TCP_MD5SIG
1125 1126 1127 1128 1129 1130 1131 1132
	/*
	 * 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.
	 */
1133
	const __u8 *hash_location = NULL;
1134
	struct tcp_md5sig_key *hash_expected;
1135
	const struct iphdr *iph = ip_hdr(skb);
1136
	const struct tcphdr *th = tcp_hdr(skb);
1137 1138 1139
	int genhash;
	unsigned char newhash[16];

E
Eric Dumazet 已提交
1140 1141
	hash_expected = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&iph->saddr,
					  AF_INET);
1142
	hash_location = tcp_parse_md5sig_option(th);
1143 1144 1145

	/* We've parsed the options - do we have a hash? */
	if (!hash_expected && !hash_location)
E
Eric Dumazet 已提交
1146
		return false;
1147 1148

	if (hash_expected && !hash_location) {
1149
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
E
Eric Dumazet 已提交
1150
		return true;
1151 1152 1153
	}

	if (!hash_expected && hash_location) {
1154
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
E
Eric Dumazet 已提交
1155
		return true;
1156 1157 1158 1159 1160
	}

	/* Okay, so this is hash_expected and hash_location -
	 * so we need to calculate the checksum.
	 */
1161 1162
	genhash = tcp_v4_md5_hash_skb(newhash,
				      hash_expected,
1163
				      NULL, skb);
1164 1165

	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
1166 1167 1168 1169 1170
		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
1171
		return true;
1172
	}
E
Eric Dumazet 已提交
1173
	return false;
1174
#endif
1175 1176
	return false;
}
1177

1178 1179
static void tcp_v4_init_req(struct request_sock *req,
			    const struct sock *sk_listener,
1180 1181 1182 1183
			    struct sk_buff *skb)
{
	struct inet_request_sock *ireq = inet_rsk(req);

1184 1185 1186
	sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr);
	sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr);
	ireq->no_srccheck = inet_sk(sk_listener)->transparent;
1187 1188 1189
	ireq->opt = tcp_v4_save_options(skb);
}

1190 1191
static struct dst_entry *tcp_v4_route_req(const struct sock *sk,
					  struct flowi *fl,
1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206
					  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;
}

1207
struct request_sock_ops tcp_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
1208
	.family		=	PF_INET,
1209
	.obj_size	=	sizeof(struct tcp_request_sock),
1210
	.rtx_syn_ack	=	tcp_rtx_synack,
1211 1212
	.send_ack	=	tcp_v4_reqsk_send_ack,
	.destructor	=	tcp_v4_reqsk_destructor,
L
Linus Torvalds 已提交
1213
	.send_reset	=	tcp_v4_send_reset,
S
stephen hemminger 已提交
1214
	.syn_ack_timeout =	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
1215 1216
};

1217
static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = {
1218
	.mss_clamp	=	TCP_MSS_DEFAULT,
1219
#ifdef CONFIG_TCP_MD5SIG
1220
	.req_md5_lookup	=	tcp_v4_md5_lookup,
1221
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1222
#endif
1223
	.init_req	=	tcp_v4_init_req,
1224 1225 1226
#ifdef CONFIG_SYN_COOKIES
	.cookie_init_seq =	cookie_v4_init_sequence,
#endif
1227
	.route_req	=	tcp_v4_route_req,
1228
	.init_seq	=	tcp_v4_init_sequence,
1229
	.send_synack	=	tcp_v4_send_synack,
1230
};
1231

L
Linus Torvalds 已提交
1232 1233 1234
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
1235
	if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
L
Linus Torvalds 已提交
1236 1237
		goto drop;

O
Octavian Purdila 已提交
1238 1239
	return tcp_conn_request(&tcp_request_sock_ops,
				&tcp_request_sock_ipv4_ops, sk, skb);
L
Linus Torvalds 已提交
1240 1241

drop:
1242
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1243 1244
	return 0;
}
E
Eric Dumazet 已提交
1245
EXPORT_SYMBOL(tcp_v4_conn_request);
L
Linus Torvalds 已提交
1246 1247 1248 1249 1250 1251


/*
 * The three way handshake has completed - we got a valid synack -
 * now create the new socket.
 */
1252
struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
1253
				  struct request_sock *req,
1254 1255 1256
				  struct dst_entry *dst,
				  struct request_sock *req_unhash,
				  bool *own_req)
L
Linus Torvalds 已提交
1257
{
1258
	struct inet_request_sock *ireq;
L
Linus Torvalds 已提交
1259 1260 1261
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1262 1263 1264
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1265
	struct ip_options_rcu *inet_opt;
L
Linus Torvalds 已提交
1266 1267 1268 1269 1270 1271

	if (sk_acceptq_is_full(sk))
		goto exit_overflow;

	newsk = tcp_create_openreq_child(sk, req, skb);
	if (!newsk)
1272
		goto exit_nonewsk;
L
Linus Torvalds 已提交
1273

1274
	newsk->sk_gso_type = SKB_GSO_TCPV4;
1275
	inet_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1276 1277 1278

	newtp		      = tcp_sk(newsk);
	newinet		      = inet_sk(newsk);
1279
	ireq		      = inet_rsk(req);
1280 1281
	sk_daddr_set(newsk, ireq->ir_rmt_addr);
	sk_rcv_saddr_set(newsk, ireq->ir_loc_addr);
1282
	newsk->sk_bound_dev_if = ireq->ir_iif;
1283
	newinet->inet_saddr	      = ireq->ir_loc_addr;
1284 1285
	inet_opt	      = ireq->opt;
	rcu_assign_pointer(newinet->inet_opt, inet_opt);
1286
	ireq->opt	      = NULL;
1287
	newinet->mc_index     = inet_iif(skb);
1288
	newinet->mc_ttl	      = ip_hdr(skb)->ttl;
1289
	newinet->rcv_tos      = ip_hdr(skb)->tos;
1290
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
1291 1292
	if (inet_opt)
		inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
E
Eric Dumazet 已提交
1293
	newinet->inet_id = newtp->write_seq ^ jiffies;
L
Linus Torvalds 已提交
1294

E
Eric Dumazet 已提交
1295 1296 1297 1298 1299 1300 1301
	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() */
	}
1302 1303
	sk_setup_caps(newsk, dst);

1304 1305
	tcp_ca_openreq_child(newsk, dst);

L
Linus Torvalds 已提交
1306
	tcp_sync_mss(newsk, dst_mtu(dst));
1307
	newtp->advmss = dst_metric_advmss(dst);
1308 1309 1310 1311
	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
1312 1313
	tcp_initialize_rcv_mss(newsk);

1314 1315
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
E
Eric Dumazet 已提交
1316 1317
	key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&newinet->inet_daddr,
				AF_INET);
1318
	if (key) {
1319 1320 1321 1322 1323 1324
		/*
		 * 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
1325 1326
		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
1327
		sk_nocaps_add(newsk, NETIF_F_GSO_MASK);
1328 1329 1330
	}
#endif

1331 1332
	if (__inet_inherit_port(sk, newsk) < 0)
		goto put_and_exit;
1333
	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
E
Eric Dumazet 已提交
1334
	if (*own_req)
1335
		tcp_move_syn(newtp, req);
L
Linus Torvalds 已提交
1336 1337 1338 1339

	return newsk;

exit_overflow:
1340
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1341 1342
exit_nonewsk:
	dst_release(dst);
L
Linus Torvalds 已提交
1343
exit:
1344
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1345
	return NULL;
1346
put_and_exit:
1347 1348
	inet_csk_prepare_forced_close(newsk);
	tcp_done(newsk);
1349
	goto exit;
L
Linus Torvalds 已提交
1350
}
E
Eric Dumazet 已提交
1351
EXPORT_SYMBOL(tcp_v4_syn_recv_sock);
L
Linus Torvalds 已提交
1352

1353
static struct sock *tcp_v4_cookie_check(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
1354
{
1355
#ifdef CONFIG_SYN_COOKIES
1356
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1357

1358
	if (!th->syn)
C
Cong Wang 已提交
1359
		sk = cookie_v4_check(sk, skb);
L
Linus Torvalds 已提交
1360 1361 1362 1363 1364
#endif
	return sk;
}

/* The socket must have it's spinlock held when we get
1365
 * here, unless it is a TCP_LISTEN socket.
L
Linus Torvalds 已提交
1366 1367 1368 1369 1370 1371 1372 1373
 *
 * 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)
{
1374 1375
	struct sock *rsk;

L
Linus Torvalds 已提交
1376
	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1377 1378
		struct dst_entry *dst = sk->sk_rx_dst;

1379
		sock_rps_save_rxhash(sk, skb);
1380
		sk_mark_napi_id(sk, skb);
1381
		if (dst) {
E
Eric Dumazet 已提交
1382
			if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif ||
1383
			    !dst->ops->check(dst, 0)) {
1384 1385 1386 1387
				dst_release(dst);
				sk->sk_rx_dst = NULL;
			}
		}
1388
		tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
L
Linus Torvalds 已提交
1389 1390 1391
		return 0;
	}

E
Eric Dumazet 已提交
1392
	if (tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1393 1394 1395
		goto csum_err;

	if (sk->sk_state == TCP_LISTEN) {
1396 1397
		struct sock *nsk = tcp_v4_cookie_check(sk, skb);

L
Linus Torvalds 已提交
1398 1399 1400
		if (!nsk)
			goto discard;
		if (nsk != sk) {
1401
			sock_rps_save_rxhash(nsk, skb);
1402
			sk_mark_napi_id(nsk, skb);
1403 1404
			if (tcp_child_process(sk, nsk, skb)) {
				rsk = nsk;
L
Linus Torvalds 已提交
1405
				goto reset;
1406
			}
L
Linus Torvalds 已提交
1407 1408
			return 0;
		}
1409
	} else
1410
		sock_rps_save_rxhash(sk, skb);
1411

1412
	if (tcp_rcv_state_process(sk, skb)) {
1413
		rsk = sk;
L
Linus Torvalds 已提交
1414
		goto reset;
1415
	}
L
Linus Torvalds 已提交
1416 1417 1418
	return 0;

reset:
1419
	tcp_v4_send_reset(rsk, skb);
L
Linus Torvalds 已提交
1420 1421 1422 1423 1424 1425 1426 1427 1428 1429
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:
1430
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS);
1431
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1432 1433
	goto discard;
}
E
Eric Dumazet 已提交
1434
EXPORT_SYMBOL(tcp_v4_do_rcv);
L
Linus Torvalds 已提交
1435

1436
void tcp_v4_early_demux(struct sk_buff *skb)
D
David S. Miller 已提交
1437 1438 1439 1440 1441 1442
{
	const struct iphdr *iph;
	const struct tcphdr *th;
	struct sock *sk;

	if (skb->pkt_type != PACKET_HOST)
1443
		return;
D
David S. Miller 已提交
1444

1445
	if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr)))
1446
		return;
D
David S. Miller 已提交
1447 1448

	iph = ip_hdr(skb);
1449
	th = tcp_hdr(skb);
D
David S. Miller 已提交
1450 1451

	if (th->doff < sizeof(struct tcphdr) / 4)
1452
		return;
D
David S. Miller 已提交
1453

1454
	sk = __inet_lookup_established(dev_net(skb->dev), &tcp_hashinfo,
D
David S. Miller 已提交
1455
				       iph->saddr, th->source,
1456
				       iph->daddr, ntohs(th->dest),
E
Eric Dumazet 已提交
1457
				       skb->skb_iif);
D
David S. Miller 已提交
1458 1459 1460
	if (sk) {
		skb->sk = sk;
		skb->destructor = sock_edemux;
1461
		if (sk_fullsock(sk)) {
1462
			struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst);
E
Eric Dumazet 已提交
1463

D
David S. Miller 已提交
1464 1465
			if (dst)
				dst = dst_check(dst, 0);
1466
			if (dst &&
E
Eric Dumazet 已提交
1467
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
1468
				skb_dst_set_noref(skb, dst);
D
David S. Miller 已提交
1469 1470 1471 1472
		}
	}
}

E
Eric Dumazet 已提交
1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490
/* 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;

1491 1492 1493 1494 1495 1496 1497 1498 1499
	/* 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
1500
		skb_dst_force_safe(skb);
1501

E
Eric Dumazet 已提交
1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
	__skb_queue_tail(&tp->ucopy.prequeue, skb);
	tp->ucopy.memory += skb->truesize;
	if (tp->ucopy.memory > sk->sk_rcvbuf) {
		struct sk_buff *skb1;

		BUG_ON(sock_owned_by_user(sk));

		while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
			sk_backlog_rcv(sk, skb1);
			NET_INC_STATS_BH(sock_net(sk),
					 LINUX_MIB_TCPPREQUEUEDROPPED);
		}

		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);

L
Linus Torvalds 已提交
1528 1529 1530 1531 1532 1533
/*
 *	From tcp_input.c
 */

int tcp_v4_rcv(struct sk_buff *skb)
{
1534
	const struct iphdr *iph;
1535
	const struct tcphdr *th;
L
Linus Torvalds 已提交
1536 1537
	struct sock *sk;
	int ret;
1538
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
1539 1540 1541 1542 1543

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

	/* Count it even if it's bad */
1544
	TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1545 1546 1547 1548

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

1549
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1550 1551 1552 1553 1554 1555 1556 1557

	if (th->doff < sizeof(struct tcphdr) / 4)
		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
1558
	 * provided case of th->doff==0 is eliminated.
L
Linus Torvalds 已提交
1559
	 * So, we defer the checks. */
1560 1561

	if (skb_checksum_init(skb, IPPROTO_TCP, inet_compute_pseudo))
1562
		goto csum_error;
L
Linus Torvalds 已提交
1563

1564
	th = tcp_hdr(skb);
1565
	iph = ip_hdr(skb);
1566 1567 1568 1569 1570 1571 1572
	/* 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
1573 1574 1575 1576
	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);
1577
	TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
1578
	TCP_SKB_CB(skb)->tcp_tw_isn = 0;
1579
	TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph);
L
Linus Torvalds 已提交
1580 1581
	TCP_SKB_CB(skb)->sacked	 = 0;

1582
lookup:
1583
	sk = __inet_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest);
L
Linus Torvalds 已提交
1584 1585 1586
	if (!sk)
		goto no_tcp_socket;

E
Eric Dumazet 已提交
1587 1588 1589 1590
process:
	if (sk->sk_state == TCP_TIME_WAIT)
		goto do_time_wait;

1591 1592 1593 1594 1595 1596 1597
	if (sk->sk_state == TCP_NEW_SYN_RECV) {
		struct request_sock *req = inet_reqsk(sk);
		struct sock *nsk = NULL;

		sk = req->rsk_listener;
		if (tcp_v4_inbound_md5_hash(sk, skb))
			goto discard_and_relse;
1598
		if (likely(sk->sk_state == TCP_LISTEN)) {
1599
			nsk = tcp_check_req(sk, skb, req, false);
1600
		} else {
1601
			inet_csk_reqsk_queue_drop_and_put(sk, req);
1602 1603
			goto lookup;
		}
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617
		if (!nsk) {
			reqsk_put(req);
			goto discard_it;
		}
		if (nsk == sk) {
			sock_hold(sk);
			reqsk_put(req);
		} else if (tcp_child_process(sk, nsk, skb)) {
			tcp_v4_send_reset(nsk, skb);
			goto discard_it;
		} else {
			return 0;
		}
	}
1618 1619
	if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
		NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
1620
		goto discard_and_relse;
1621
	}
1622

L
Linus Torvalds 已提交
1623 1624
	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;
1625 1626 1627 1628

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

1629
	nf_reset(skb);
L
Linus Torvalds 已提交
1630

1631
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1632 1633 1634 1635
		goto discard_and_relse;

	skb->dev = NULL;

1636 1637 1638 1639 1640 1641 1642
	if (sk->sk_state == TCP_LISTEN) {
		ret = tcp_v4_do_rcv(sk, skb);
		goto put_and_return;
	}

	sk_incoming_cpu_update(sk);

1643
	bh_lock_sock_nested(sk);
1644
	tcp_sk(sk)->segs_in += max_t(u16, 1, skb_shinfo(skb)->gso_segs);
L
Linus Torvalds 已提交
1645 1646
	ret = 0;
	if (!sock_owned_by_user(sk)) {
D
Dan Williams 已提交
1647
		if (!tcp_prequeue(sk, skb))
L
Linus Torvalds 已提交
1648
			ret = tcp_v4_do_rcv(sk, skb);
1649 1650
	} else if (unlikely(sk_add_backlog(sk, skb,
					   sk->sk_rcvbuf + sk->sk_sndbuf))) {
Z
Zhu Yi 已提交
1651
		bh_unlock_sock(sk);
1652
		NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
Z
Zhu Yi 已提交
1653 1654
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1655 1656
	bh_unlock_sock(sk);

1657
put_and_return:
L
Linus Torvalds 已提交
1658 1659 1660 1661 1662 1663 1664 1665
	sock_put(sk);

	return ret;

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

E
Eric Dumazet 已提交
1666
	if (tcp_checksum_complete(skb)) {
1667 1668
csum_error:
		TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS);
L
Linus Torvalds 已提交
1669
bad_packet:
1670
		TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1671
	} else {
1672
		tcp_v4_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1673 1674 1675 1676 1677
	}

discard_it:
	/* Discard frame. */
	kfree_skb(skb);
1678
	return 0;
L
Linus Torvalds 已提交
1679 1680 1681 1682 1683 1684 1685

discard_and_relse:
	sock_put(sk);
	goto discard_it;

do_time_wait:
	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1686
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1687 1688 1689
		goto discard_it;
	}

1690 1691 1692
	if (tcp_checksum_complete(skb)) {
		inet_twsk_put(inet_twsk(sk));
		goto csum_error;
L
Linus Torvalds 已提交
1693
	}
1694
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1695
	case TCP_TW_SYN: {
1696
		struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev),
1697
							&tcp_hashinfo,
1698
							iph->saddr, th->source,
1699
							iph->daddr, th->dest,
1700
							inet_iif(skb));
L
Linus Torvalds 已提交
1701
		if (sk2) {
1702
			inet_twsk_deschedule_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1703 1704 1705 1706 1707 1708 1709 1710 1711
			sk = sk2;
			goto process;
		}
		/* Fall through to ACK */
	}
	case TCP_TW_ACK:
		tcp_v4_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
1712 1713 1714
		tcp_v4_send_reset(sk, skb);
		inet_twsk_deschedule_put(inet_twsk(sk));
		goto discard_it;
L
Linus Torvalds 已提交
1715 1716 1717 1718 1719
	case TCP_TW_SUCCESS:;
	}
	goto discard_it;
}

1720 1721 1722 1723 1724
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
1725

1726
void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
E
Eric Dumazet 已提交
1727 1728 1729
{
	struct dst_entry *dst = skb_dst(skb);

E
Eric Dumazet 已提交
1730
	if (dst && dst_hold_safe(dst)) {
1731 1732 1733
		sk->sk_rx_dst = dst;
		inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
	}
E
Eric Dumazet 已提交
1734
}
1735
EXPORT_SYMBOL(inet_sk_rx_dst_set);
E
Eric Dumazet 已提交
1736

1737
const struct inet_connection_sock_af_ops ipv4_specific = {
1738 1739 1740
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
E
Eric Dumazet 已提交
1741
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1742 1743 1744 1745 1746 1747 1748
	.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),
1749
	.bind_conflict	   = inet_csk_bind_conflict,
1750
#ifdef CONFIG_COMPAT
1751 1752
	.compat_setsockopt = compat_ip_setsockopt,
	.compat_getsockopt = compat_ip_getsockopt,
1753
#endif
1754
	.mtu_reduced	   = tcp_v4_mtu_reduced,
L
Linus Torvalds 已提交
1755
};
E
Eric Dumazet 已提交
1756
EXPORT_SYMBOL(ipv4_specific);
L
Linus Torvalds 已提交
1757

1758
#ifdef CONFIG_TCP_MD5SIG
1759
static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = {
1760
	.md5_lookup		= tcp_v4_md5_lookup,
1761
	.calc_md5_hash		= tcp_v4_md5_hash_skb,
1762 1763
	.md5_parse		= tcp_v4_parse_md5_keys,
};
1764
#endif
1765

L
Linus Torvalds 已提交
1766 1767 1768 1769 1770
/* 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)
{
1771
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1772

1773
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1774

1775
	icsk->icsk_af_ops = &ipv4_specific;
1776

1777
#ifdef CONFIG_TCP_MD5SIG
1778
	tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific;
1779
#endif
L
Linus Torvalds 已提交
1780 1781 1782 1783

	return 0;
}

1784
void tcp_v4_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1785 1786 1787 1788 1789
{
	struct tcp_sock *tp = tcp_sk(sk);

	tcp_clear_xmit_timers(sk);

1790
	tcp_cleanup_congestion_control(sk);
1791

L
Linus Torvalds 已提交
1792
	/* Cleanup up the write buffer. */
1793
	tcp_write_queue_purge(sk);
L
Linus Torvalds 已提交
1794 1795

	/* Cleans up our, hopefully empty, out_of_order_queue. */
1796
	__skb_queue_purge(&tp->out_of_order_queue);
L
Linus Torvalds 已提交
1797

1798 1799 1800
#ifdef CONFIG_TCP_MD5SIG
	/* Clean up the MD5 key list, if any */
	if (tp->md5sig_info) {
E
Eric Dumazet 已提交
1801
		tcp_clear_md5_list(sk);
1802
		kfree_rcu(tp->md5sig_info, rcu);
1803 1804 1805
		tp->md5sig_info = NULL;
	}
#endif
C
Chris Leech 已提交
1806

L
Linus Torvalds 已提交
1807 1808 1809 1810
	/* Clean prequeue, it must be empty really */
	__skb_queue_purge(&tp->ucopy.prequeue);

	/* Clean up a referenced TCP bind bucket. */
1811
	if (inet_csk(sk)->icsk_bind_hash)
1812
		inet_put_port(sk);
L
Linus Torvalds 已提交
1813

1814
	BUG_ON(tp->fastopen_rsk);
1815

1816 1817
	/* If socket is aborted during connect operation */
	tcp_free_fastopen_req(tp);
1818
	tcp_saved_syn_free(tp);
1819

1820
	sk_sockets_allocated_dec(sk);
G
Glauber Costa 已提交
1821
	sock_release_memcg(sk);
L
Linus Torvalds 已提交
1822 1823 1824 1825 1826 1827
}
EXPORT_SYMBOL(tcp_v4_destroy_sock);

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

1828 1829 1830 1831 1832
/*
 * 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
1833 1834
static void *listening_get_next(struct seq_file *seq, void *cur)
{
1835
	struct inet_connection_sock *icsk;
1836
	struct hlist_nulls_node *node;
L
Linus Torvalds 已提交
1837
	struct sock *sk = cur;
1838
	struct inet_listen_hashbucket *ilb;
J
Jianjun Kong 已提交
1839
	struct tcp_iter_state *st = seq->private;
1840
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
1841 1842

	if (!sk) {
1843
		ilb = &tcp_hashinfo.listening_hash[st->bucket];
1844
		spin_lock_bh(&ilb->lock);
1845
		sk = sk_nulls_head(&ilb->head);
1846
		st->offset = 0;
L
Linus Torvalds 已提交
1847 1848
		goto get_sk;
	}
1849
	ilb = &tcp_hashinfo.listening_hash[st->bucket];
L
Linus Torvalds 已提交
1850
	++st->num;
1851
	++st->offset;
L
Linus Torvalds 已提交
1852

1853
	sk = sk_nulls_next(sk);
L
Linus Torvalds 已提交
1854
get_sk:
1855
	sk_nulls_for_each_from(sk, node) {
1856 1857 1858
		if (!net_eq(sock_net(sk), net))
			continue;
		if (sk->sk_family == st->family) {
L
Linus Torvalds 已提交
1859 1860 1861
			cur = sk;
			goto out;
		}
1862
		icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1863
	}
1864
	spin_unlock_bh(&ilb->lock);
1865
	st->offset = 0;
1866
	if (++st->bucket < INET_LHTABLE_SIZE) {
1867 1868
		ilb = &tcp_hashinfo.listening_hash[st->bucket];
		spin_lock_bh(&ilb->lock);
1869
		sk = sk_nulls_head(&ilb->head);
L
Linus Torvalds 已提交
1870 1871 1872 1873 1874 1875 1876 1877 1878
		goto get_sk;
	}
	cur = NULL;
out:
	return cur;
}

static void *listening_get_idx(struct seq_file *seq, loff_t *pos)
{
1879 1880 1881 1882 1883 1884
	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
1885 1886 1887 1888 1889 1890 1891 1892

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

E
Eric Dumazet 已提交
1893
static inline bool empty_bucket(const struct tcp_iter_state *st)
1894
{
E
Eric Dumazet 已提交
1895
	return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain);
1896 1897
}

1898 1899 1900 1901
/*
 * 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
1902 1903
static void *established_get_first(struct seq_file *seq)
{
J
Jianjun Kong 已提交
1904
	struct tcp_iter_state *st = seq->private;
1905
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
1906 1907
	void *rc = NULL;

1908 1909
	st->offset = 0;
	for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) {
L
Linus Torvalds 已提交
1910
		struct sock *sk;
1911
		struct hlist_nulls_node *node;
1912
		spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket);
L
Linus Torvalds 已提交
1913

1914 1915 1916 1917
		/* Lockless fast path for the common case of empty buckets */
		if (empty_bucket(st))
			continue;

1918
		spin_lock_bh(lock);
1919
		sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) {
1920
			if (sk->sk_family != st->family ||
1921
			    !net_eq(sock_net(sk), net)) {
L
Linus Torvalds 已提交
1922 1923 1924 1925 1926
				continue;
			}
			rc = sk;
			goto out;
		}
1927
		spin_unlock_bh(lock);
L
Linus Torvalds 已提交
1928 1929 1930 1931 1932 1933 1934 1935
	}
out:
	return rc;
}

static void *established_get_next(struct seq_file *seq, void *cur)
{
	struct sock *sk = cur;
1936
	struct hlist_nulls_node *node;
J
Jianjun Kong 已提交
1937
	struct tcp_iter_state *st = seq->private;
1938
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
1939 1940

	++st->num;
1941
	++st->offset;
L
Linus Torvalds 已提交
1942

E
Eric Dumazet 已提交
1943
	sk = sk_nulls_next(sk);
L
Linus Torvalds 已提交
1944

1945
	sk_nulls_for_each_from(sk, node) {
1946
		if (sk->sk_family == st->family && net_eq(sock_net(sk), net))
E
Eric Dumazet 已提交
1947
			return sk;
L
Linus Torvalds 已提交
1948 1949
	}

E
Eric Dumazet 已提交
1950 1951 1952
	spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
	++st->bucket;
	return established_get_first(seq);
L
Linus Torvalds 已提交
1953 1954 1955 1956
}

static void *established_get_idx(struct seq_file *seq, loff_t pos)
{
1957 1958 1959 1960 1961
	struct tcp_iter_state *st = seq->private;
	void *rc;

	st->bucket = 0;
	rc = established_get_first(seq);
L
Linus Torvalds 已提交
1962 1963 1964 1965

	while (rc && pos) {
		rc = established_get_next(seq, rc);
		--pos;
1966
	}
L
Linus Torvalds 已提交
1967 1968 1969 1970 1971 1972
	return rc;
}

static void *tcp_get_idx(struct seq_file *seq, loff_t pos)
{
	void *rc;
J
Jianjun Kong 已提交
1973
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985

	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;
}

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
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
2004
		st->state = TCP_SEQ_STATE_ESTABLISHED;
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
		/* 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
2019 2020
static void *tcp_seq_start(struct seq_file *seq, loff_t *pos)
{
J
Jianjun Kong 已提交
2021
	struct tcp_iter_state *st = seq->private;
2022 2023 2024 2025 2026 2027 2028 2029
	void *rc;

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

L
Linus Torvalds 已提交
2030 2031
	st->state = TCP_SEQ_STATE_LISTENING;
	st->num = 0;
2032 2033 2034 2035 2036 2037 2038
	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
2039 2040 2041 2042
}

static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
2043
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
	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;
2056 2057
			st->bucket = 0;
			st->offset = 0;
L
Linus Torvalds 已提交
2058 2059 2060 2061 2062 2063 2064 2065 2066
			rc	  = established_get_first(seq);
		}
		break;
	case TCP_SEQ_STATE_ESTABLISHED:
		rc = established_get_next(seq, v);
		break;
	}
out:
	++*pos;
2067
	st->last_pos = *pos;
L
Linus Torvalds 已提交
2068 2069 2070 2071 2072
	return rc;
}

static void tcp_seq_stop(struct seq_file *seq, void *v)
{
J
Jianjun Kong 已提交
2073
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2074 2075 2076 2077

	switch (st->state) {
	case TCP_SEQ_STATE_LISTENING:
		if (v != SEQ_START_TOKEN)
2078
			spin_unlock_bh(&tcp_hashinfo.listening_hash[st->bucket].lock);
L
Linus Torvalds 已提交
2079 2080 2081
		break;
	case TCP_SEQ_STATE_ESTABLISHED:
		if (v)
2082
			spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
L
Linus Torvalds 已提交
2083 2084 2085 2086
		break;
	}
}

2087
int tcp_seq_open(struct inode *inode, struct file *file)
L
Linus Torvalds 已提交
2088
{
A
Al Viro 已提交
2089
	struct tcp_seq_afinfo *afinfo = PDE_DATA(inode);
L
Linus Torvalds 已提交
2090
	struct tcp_iter_state *s;
2091
	int err;
L
Linus Torvalds 已提交
2092

2093 2094 2095 2096
	err = seq_open_net(inode, file, &afinfo->seq_ops,
			  sizeof(struct tcp_iter_state));
	if (err < 0)
		return err;
2097

2098
	s = ((struct seq_file *)file->private_data)->private;
L
Linus Torvalds 已提交
2099
	s->family		= afinfo->family;
S
stephen hemminger 已提交
2100
	s->last_pos		= 0;
2101 2102
	return 0;
}
2103
EXPORT_SYMBOL(tcp_seq_open);
2104

2105
int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo)
L
Linus Torvalds 已提交
2106 2107 2108 2109
{
	int rc = 0;
	struct proc_dir_entry *p;

2110 2111 2112 2113
	afinfo->seq_ops.start		= tcp_seq_start;
	afinfo->seq_ops.next		= tcp_seq_next;
	afinfo->seq_ops.stop		= tcp_seq_stop;

2114
	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
2115
			     afinfo->seq_fops, afinfo);
2116
	if (!p)
L
Linus Torvalds 已提交
2117 2118 2119
		rc = -ENOMEM;
	return rc;
}
E
Eric Dumazet 已提交
2120
EXPORT_SYMBOL(tcp_proc_register);
L
Linus Torvalds 已提交
2121

2122
void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo)
L
Linus Torvalds 已提交
2123
{
2124
	remove_proc_entry(afinfo->name, net->proc_net);
L
Linus Torvalds 已提交
2125
}
E
Eric Dumazet 已提交
2126
EXPORT_SYMBOL(tcp_proc_unregister);
L
Linus Torvalds 已提交
2127

2128
static void get_openreq4(const struct request_sock *req,
E
Eric Dumazet 已提交
2129
			 struct seq_file *f, int i)
L
Linus Torvalds 已提交
2130
{
2131
	const struct inet_request_sock *ireq = inet_rsk(req);
2132
	long delta = req->rsk_timer.expires - jiffies;
L
Linus Torvalds 已提交
2133

2134
	seq_printf(f, "%4d: %08X:%04X %08X:%04X"
2135
		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %u %d %pK",
L
Linus Torvalds 已提交
2136
		i,
2137
		ireq->ir_loc_addr,
2138
		ireq->ir_num,
2139 2140
		ireq->ir_rmt_addr,
		ntohs(ireq->ir_rmt_port),
L
Linus Torvalds 已提交
2141 2142 2143
		TCP_SYN_RECV,
		0, 0, /* could print option size, but that is af dependent. */
		1,    /* timers active (only the expire timer) */
2144
		jiffies_delta_to_clock_t(delta),
2145
		req->num_timeout,
E
Eric Dumazet 已提交
2146 2147
		from_kuid_munged(seq_user_ns(f),
				 sock_i_uid(req->rsk_listener)),
L
Linus Torvalds 已提交
2148 2149
		0,  /* non standard timer */
		0, /* open_requests have no inode */
2150
		0,
2151
		req);
L
Linus Torvalds 已提交
2152 2153
}

2154
static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i)
L
Linus Torvalds 已提交
2155 2156 2157
{
	int timer_active;
	unsigned long timer_expires;
2158
	const struct tcp_sock *tp = tcp_sk(sk);
2159
	const struct inet_connection_sock *icsk = inet_csk(sk);
2160
	const struct inet_sock *inet = inet_sk(sk);
2161
	const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq;
E
Eric Dumazet 已提交
2162 2163 2164 2165
	__be32 dest = inet->inet_daddr;
	__be32 src = inet->inet_rcv_saddr;
	__u16 destp = ntohs(inet->inet_dport);
	__u16 srcp = ntohs(inet->inet_sport);
2166
	int rx_queue;
2167
	int state;
L
Linus Torvalds 已提交
2168

N
Nandita Dukkipati 已提交
2169 2170 2171
	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
2172
		timer_active	= 1;
2173 2174
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
2175
		timer_active	= 4;
2176
		timer_expires	= icsk->icsk_timeout;
2177
	} else if (timer_pending(&sk->sk_timer)) {
L
Linus Torvalds 已提交
2178
		timer_active	= 2;
2179
		timer_expires	= sk->sk_timer.expires;
L
Linus Torvalds 已提交
2180 2181 2182 2183 2184
	} else {
		timer_active	= 0;
		timer_expires = jiffies;
	}

2185 2186
	state = sk_state_load(sk);
	if (state == TCP_LISTEN)
2187 2188
		rx_queue = sk->sk_ack_backlog;
	else
2189 2190
		/* Because we don't lock the socket,
		 * we might find a transient negative value.
2191 2192 2193
		 */
		rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0);

2194
	seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX "
2195
			"%08X %5u %8d %lu %d %pK %lu %lu %u %u %d",
2196
		i, src, srcp, dest, destp, state,
2197
		tp->write_seq - tp->snd_una,
2198
		rx_queue,
L
Linus Torvalds 已提交
2199
		timer_active,
2200
		jiffies_delta_to_clock_t(timer_expires - jiffies),
2201
		icsk->icsk_retransmits,
2202
		from_kuid_munged(seq_user_ns(f), sock_i_uid(sk)),
2203
		icsk->icsk_probes_out,
2204 2205
		sock_i_ino(sk),
		atomic_read(&sk->sk_refcnt), sk,
2206 2207
		jiffies_to_clock_t(icsk->icsk_rto),
		jiffies_to_clock_t(icsk->icsk_ack.ato),
2208
		(icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
L
Linus Torvalds 已提交
2209
		tp->snd_cwnd,
2210 2211
		state == TCP_LISTEN ?
		    fastopenq->max_qlen :
2212
		    (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh));
L
Linus Torvalds 已提交
2213 2214
}

2215
static void get_timewait4_sock(const struct inet_timewait_sock *tw,
2216
			       struct seq_file *f, int i)
L
Linus Torvalds 已提交
2217
{
2218
	long delta = tw->tw_timer.expires - jiffies;
2219
	__be32 dest, src;
L
Linus Torvalds 已提交
2220 2221 2222 2223 2224 2225 2226
	__u16 destp, srcp;

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

2227
	seq_printf(f, "%4d: %08X:%04X %08X:%04X"
2228
		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK",
L
Linus Torvalds 已提交
2229
		i, src, srcp, dest, destp, tw->tw_substate, 0, 0,
2230
		3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
2231
		atomic_read(&tw->tw_refcnt), tw);
L
Linus Torvalds 已提交
2232 2233 2234 2235 2236 2237
}

#define TMPSZ 150

static int tcp4_seq_show(struct seq_file *seq, void *v)
{
J
Jianjun Kong 已提交
2238
	struct tcp_iter_state *st;
E
Eric Dumazet 已提交
2239
	struct sock *sk = v;
L
Linus Torvalds 已提交
2240

2241
	seq_setwidth(seq, TMPSZ - 1);
L
Linus Torvalds 已提交
2242
	if (v == SEQ_START_TOKEN) {
2243
		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
L
Linus Torvalds 已提交
2244 2245 2246 2247 2248 2249
			   "rx_queue tr tm->when retrnsmt   uid  timeout "
			   "inode");
		goto out;
	}
	st = seq->private;

2250 2251 2252
	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
2253
		get_openreq4(v, seq, st->num);
2254 2255
	else
		get_tcp4_sock(v, seq, st->num);
L
Linus Torvalds 已提交
2256
out:
2257
	seq_pad(seq, '\n');
L
Linus Torvalds 已提交
2258 2259 2260
	return 0;
}

2261 2262 2263 2264 2265 2266 2267 2268
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
2269 2270 2271
static struct tcp_seq_afinfo tcp4_seq_afinfo = {
	.name		= "tcp",
	.family		= AF_INET,
2272
	.seq_fops	= &tcp_afinfo_seq_fops,
2273 2274 2275
	.seq_ops	= {
		.show		= tcp4_seq_show,
	},
L
Linus Torvalds 已提交
2276 2277
};

2278
static int __net_init tcp4_proc_init_net(struct net *net)
2279 2280 2281 2282
{
	return tcp_proc_register(net, &tcp4_seq_afinfo);
}

2283
static void __net_exit tcp4_proc_exit_net(struct net *net)
2284 2285 2286 2287 2288 2289 2290 2291 2292
{
	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
2293 2294
int __init tcp4_proc_init(void)
{
2295
	return register_pernet_subsys(&tcp4_net_ops);
L
Linus Torvalds 已提交
2296 2297 2298 2299
}

void tcp4_proc_exit(void)
{
2300
	unregister_pernet_subsys(&tcp4_net_ops);
L
Linus Torvalds 已提交
2301 2302 2303 2304 2305 2306 2307 2308 2309
}
#endif /* CONFIG_PROC_FS */

struct proto tcp_prot = {
	.name			= "TCP",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v4_connect,
	.disconnect		= tcp_disconnect,
2310
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
2311 2312 2313 2314 2315 2316 2317
	.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,
2318 2319
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
2320
	.backlog_rcv		= tcp_v4_do_rcv,
E
Eric Dumazet 已提交
2321
	.release_cb		= tcp_release_cb,
2322 2323 2324
	.hash			= inet_hash,
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
2325
	.enter_memory_pressure	= tcp_enter_memory_pressure,
2326
	.stream_memory_free	= tcp_stream_memory_free,
L
Linus Torvalds 已提交
2327
	.sockets_allocated	= &tcp_sockets_allocated,
2328
	.orphan_count		= &tcp_orphan_count,
L
Linus Torvalds 已提交
2329 2330
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
2331
	.sysctl_mem		= sysctl_tcp_mem,
L
Linus Torvalds 已提交
2332 2333 2334 2335
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp_sock),
2336
	.slab_flags		= SLAB_DESTROY_BY_RCU,
2337
	.twsk_prot		= &tcp_timewait_sock_ops,
2338
	.rsk_prot		= &tcp_request_sock_ops,
2339
	.h.hashinfo		= &tcp_hashinfo,
2340
	.no_autobind		= true,
2341 2342 2343 2344
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
#endif
A
Andrew Morton 已提交
2345
#ifdef CONFIG_MEMCG_KMEM
G
Glauber Costa 已提交
2346 2347 2348 2349
	.init_cgroup		= tcp_init_cgroup,
	.destroy_cgroup		= tcp_destroy_cgroup,
	.proto_cgroup		= tcp_proto_cgroup,
#endif
2350
	.diag_destroy		= tcp_abort,
L
Linus Torvalds 已提交
2351
};
E
Eric Dumazet 已提交
2352
EXPORT_SYMBOL(tcp_prot);
L
Linus Torvalds 已提交
2353

2354 2355 2356 2357 2358 2359 2360 2361 2362
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);
}

2363 2364
static int __net_init tcp_sk_init(struct net *net)
{
2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379
	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;
		*per_cpu_ptr(net->ipv4.tcp_sk, cpu) = sk;
	}
2380

2381
	net->ipv4.sysctl_tcp_ecn = 2;
2382 2383
	net->ipv4.sysctl_tcp_ecn_fallback = 1;

F
Fan Du 已提交
2384
	net->ipv4.sysctl_tcp_base_mss = TCP_BASE_MSS;
2385
	net->ipv4.sysctl_tcp_probe_threshold = TCP_PROBE_THRESHOLD;
2386
	net->ipv4.sysctl_tcp_probe_interval = TCP_PROBE_INTERVAL;
2387

2388 2389
	net->ipv4.sysctl_tcp_keepalive_time = TCP_KEEPALIVE_TIME;

2390
	return 0;
2391 2392 2393 2394
fail:
	tcp_sk_exit(net);

	return res;
E
Eric W. Biederman 已提交
2395 2396 2397 2398 2399
}

static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list)
{
	inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET);
2400 2401 2402
}

static struct pernet_operations __net_initdata tcp_sk_ops = {
E
Eric W. Biederman 已提交
2403 2404 2405
       .init	   = tcp_sk_init,
       .exit	   = tcp_sk_exit,
       .exit_batch = tcp_sk_exit_batch,
2406 2407
};

2408
void __init tcp_v4_init(void)
L
Linus Torvalds 已提交
2409
{
2410
	inet_hashinfo_init(&tcp_hashinfo);
2411
	if (register_pernet_subsys(&tcp_sk_ops))
L
Linus Torvalds 已提交
2412 2413
		panic("Failed to create the TCP control socket.\n");
}