tcp_ipv4.c 68.0 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>
C
Chris Leech 已提交
75
#include <net/netdma.h>
76
#include <net/secure_seq.h>
G
Glauber Costa 已提交
77
#include <net/tcp_memcontrol.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


93
#ifdef CONFIG_TCP_MD5SIG
E
Eric Dumazet 已提交
94
static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
95
			       __be32 daddr, __be32 saddr, const struct tcphdr *th);
96 97
#endif

98
struct inet_hashinfo tcp_hashinfo;
E
Eric Dumazet 已提交
99
EXPORT_SYMBOL(tcp_hashinfo);
L
Linus Torvalds 已提交
100

101
static inline __u32 tcp_v4_init_sequence(const struct sk_buff *skb)
L
Linus Torvalds 已提交
102
{
103 104
	return secure_tcp_sequence_number(ip_hdr(skb)->daddr,
					  ip_hdr(skb)->saddr,
105 106
					  tcp_hdr(skb)->dest,
					  tcp_hdr(skb)->source);
L
Linus Torvalds 已提交
107 108
}

109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126
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 &&
	    (twp == NULL || (sysctl_tcp_tw_reuse &&
127
			     get_seconds() - tcptw->tw_ts_recent_stamp > 1))) {
128 129 130 131 132 133 134 135 136 137 138 139 140
		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);

P
Pavel Emelyanov 已提交
141 142 143 144 145 146 147 148
static int tcp_repair_connect(struct sock *sk)
{
	tcp_connect_init(sk);
	tcp_finish_connect(sk, NULL);

	return 0;
}

L
Linus Torvalds 已提交
149 150 151
/* This will initiate an outgoing connection. */
int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
{
152
	struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
L
Linus Torvalds 已提交
153 154
	struct inet_sock *inet = inet_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
155
	__be16 orig_sport, orig_dport;
156
	__be32 daddr, nexthop;
157
	struct flowi4 *fl4;
158
	struct rtable *rt;
L
Linus Torvalds 已提交
159
	int err;
160
	struct ip_options_rcu *inet_opt;
L
Linus Torvalds 已提交
161 162 163 164 165 166 167 168

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

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

	nexthop = daddr = usin->sin_addr.s_addr;
169 170 171
	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
172 173
		if (!daddr)
			return -EINVAL;
174
		nexthop = inet_opt->opt.faddr;
L
Linus Torvalds 已提交
175 176
	}

177 178
	orig_sport = inet->inet_sport;
	orig_dport = usin->sin_port;
179 180
	fl4 = &inet->cork.fl.u.ip4;
	rt = ip_route_connect(fl4, nexthop, inet->inet_saddr,
181 182 183 184 185 186
			      RT_CONN_FLAGS(sk), sk->sk_bound_dev_if,
			      IPPROTO_TCP,
			      orig_sport, orig_dport, sk, true);
	if (IS_ERR(rt)) {
		err = PTR_ERR(rt);
		if (err == -ENETUNREACH)
187
			IP_INC_STATS_BH(sock_net(sk), IPSTATS_MIB_OUTNOROUTES);
188
		return err;
189
	}
L
Linus Torvalds 已提交
190 191 192 193 194 195

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

196
	if (!inet_opt || !inet_opt->opt.srr)
197
		daddr = fl4->daddr;
L
Linus Torvalds 已提交
198

E
Eric Dumazet 已提交
199
	if (!inet->inet_saddr)
200
		inet->inet_saddr = fl4->saddr;
E
Eric Dumazet 已提交
201
	inet->inet_rcv_saddr = inet->inet_saddr;
L
Linus Torvalds 已提交
202

E
Eric Dumazet 已提交
203
	if (tp->rx_opt.ts_recent_stamp && inet->inet_daddr != daddr) {
L
Linus Torvalds 已提交
204 205 206
		/* Reset inherited state */
		tp->rx_opt.ts_recent	   = 0;
		tp->rx_opt.ts_recent_stamp = 0;
P
Pavel Emelyanov 已提交
207 208
		if (likely(!tp->repair))
			tp->write_seq	   = 0;
L
Linus Torvalds 已提交
209 210
	}

211
	if (tcp_death_row.sysctl_tw_recycle &&
212
	    !tp->rx_opt.ts_recent_stamp && fl4->daddr == daddr) {
213
		struct inet_peer *peer = rt_get_peer(rt, fl4->daddr);
214 215 216 217 218
		/*
		 * VJ's idea. We save last timestamp seen from
		 * the destination in peer table, when entering state
		 * TIME-WAIT * and initialize rx_opt.ts_recent from it,
		 * when trying new connection.
L
Linus Torvalds 已提交
219
		 */
220 221 222 223 224 225
		if (peer) {
			inet_peer_refcheck(peer);
			if ((u32)get_seconds() - peer->tcp_ts_stamp <= TCP_PAWS_MSL) {
				tp->rx_opt.ts_recent_stamp = peer->tcp_ts_stamp;
				tp->rx_opt.ts_recent = peer->tcp_ts;
			}
L
Linus Torvalds 已提交
226 227 228
		}
	}

E
Eric Dumazet 已提交
229 230
	inet->inet_dport = usin->sin_port;
	inet->inet_daddr = daddr;
L
Linus Torvalds 已提交
231

232
	inet_csk(sk)->icsk_ext_hdr_len = 0;
233 234
	if (inet_opt)
		inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
L
Linus Torvalds 已提交
235

236
	tp->rx_opt.mss_clamp = TCP_MSS_DEFAULT;
L
Linus Torvalds 已提交
237 238 239 240 241 242 243

	/* 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);
244
	err = inet_hash_connect(&tcp_death_row, sk);
L
Linus Torvalds 已提交
245 246 247
	if (err)
		goto failure;

248
	rt = ip_route_newports(fl4, rt, orig_sport, orig_dport,
249 250 251 252
			       inet->inet_sport, inet->inet_dport, sk);
	if (IS_ERR(rt)) {
		err = PTR_ERR(rt);
		rt = NULL;
L
Linus Torvalds 已提交
253
		goto failure;
254
	}
L
Linus Torvalds 已提交
255
	/* OK, now commit destination to socket.  */
256
	sk->sk_gso_type = SKB_GSO_TCPV4;
257
	sk_setup_caps(sk, &rt->dst);
L
Linus Torvalds 已提交
258

P
Pavel Emelyanov 已提交
259
	if (!tp->write_seq && likely(!tp->repair))
E
Eric Dumazet 已提交
260 261 262
		tp->write_seq = secure_tcp_sequence_number(inet->inet_saddr,
							   inet->inet_daddr,
							   inet->inet_sport,
L
Linus Torvalds 已提交
263 264
							   usin->sin_port);

E
Eric Dumazet 已提交
265
	inet->inet_id = tp->write_seq ^ jiffies;
L
Linus Torvalds 已提交
266

P
Pavel Emelyanov 已提交
267 268 269 270 271
	if (likely(!tp->repair))
		err = tcp_connect(sk);
	else
		err = tcp_repair_connect(sk);

L
Linus Torvalds 已提交
272 273 274 275 276 277 278
	rt = NULL;
	if (err)
		goto failure;

	return 0;

failure:
279 280 281 282
	/*
	 * This unhashes the socket and releases the local port,
	 * if necessary.
	 */
L
Linus Torvalds 已提交
283 284 285
	tcp_set_state(sk, TCP_CLOSE);
	ip_rt_put(rt);
	sk->sk_route_caps = 0;
E
Eric Dumazet 已提交
286
	inet->inet_dport = 0;
L
Linus Torvalds 已提交
287 288
	return err;
}
E
Eric Dumazet 已提交
289
EXPORT_SYMBOL(tcp_v4_connect);
L
Linus Torvalds 已提交
290 291 292 293

/*
 * This routine does path mtu discovery as defined in RFC1191.
 */
294
static void do_pmtu_discovery(struct sock *sk, const struct iphdr *iph, u32 mtu)
L
Linus Torvalds 已提交
295 296 297 298 299 300 301 302 303 304 305 306 307 308
{
	struct dst_entry *dst;
	struct inet_sock *inet = inet_sk(sk);

	/* 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)
		return;

	/* We don't check in the destentry if pmtu discovery is forbidden
	 * on this route. We just assume that no packet_to_big packets
	 * are send back when pmtu discovery is not active.
309
	 * There is a small race when the user changes this flag in the
L
Linus Torvalds 已提交
310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
	 * route, but I think that's acceptable.
	 */
	if ((dst = __sk_dst_check(sk, 0)) == NULL)
		return;

	dst->ops->update_pmtu(dst, mtu);

	/* 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 &&
326
	    inet_csk(sk)->icsk_pmtu_cookie > mtu) {
L
Linus Torvalds 已提交
327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353
		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 */
}

/*
 * 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;
L
Linus Torvalds 已提交
365
	__u32 seq;
366
	__u32 remaining;
L
Linus Torvalds 已提交
367
	int err;
368
	struct net *net = dev_net(icmp_skb->dev);
L
Linus Torvalds 已提交
369

370
	if (icmp_skb->len < (iph->ihl << 2) + 8) {
371
		ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
L
Linus Torvalds 已提交
372 373 374
		return;
	}

375
	sk = inet_lookup(net, &tcp_hashinfo, iph->daddr, th->dest,
376
			iph->saddr, 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
383 384 385 386 387 388 389 390
		return;
	}

	bh_lock_sock(sk);
	/* If too many ICMPs get dropped on busy
	 * servers this needs to be solved differently.
	 */
	if (sock_owned_by_user(sk))
391
		NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
L
Linus Torvalds 已提交
392 393 394 395

	if (sk->sk_state == TCP_CLOSE)
		goto out;

396 397 398 399 400
	if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
		NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
		goto out;
	}

401
	icsk = inet_csk(sk);
L
Linus Torvalds 已提交
402 403 404 405
	tp = tcp_sk(sk);
	seq = ntohl(th->seq);
	if (sk->sk_state != TCP_LISTEN &&
	    !between(seq, tp->snd_una, tp->snd_nxt)) {
406
		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427
		goto out;
	}

	switch (type) {
	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) */
			if (!sock_owned_by_user(sk))
				do_pmtu_discovery(sk, iph, info);
			goto out;
		}

		err = icmp_err_convert[code].errno;
428 429 430 431 432 433 434 435
		/* 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 ||
		    !icsk->icsk_backoff)
			break;

436 437 438
		if (sock_owned_by_user(sk))
			break;

439
		icsk->icsk_backoff--;
440 441
		inet_csk(sk)->icsk_rto = (tp->srtt ? __tcp_set_rto(tp) :
			TCP_TIMEOUT_INIT) << icsk->icsk_backoff;
442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458
		tcp_bound_rto(sk);

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

		remaining = icsk->icsk_rto - min(icsk->icsk_rto,
				tcp_time_stamp - TCP_SKB_CB(skb)->when);

		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
459 460 461 462 463 464 465 466 467
		break;
	case ICMP_TIME_EXCEEDED:
		err = EHOSTUNREACH;
		break;
	default:
		goto out;
	}

	switch (sk->sk_state) {
468
		struct request_sock *req, **prev;
L
Linus Torvalds 已提交
469 470 471 472
	case TCP_LISTEN:
		if (sock_owned_by_user(sk))
			goto out;

473 474
		req = inet_csk_search_req(sk, &prev, th->dest,
					  iph->daddr, iph->saddr);
L
Linus Torvalds 已提交
475 476 477 478 479 480
		if (!req)
			goto out;

		/* ICMPs are not backlogged, hence we cannot get
		   an established socket here.
		 */
481
		WARN_ON(req->sk);
L
Linus Torvalds 已提交
482

483
		if (seq != tcp_rsk(req)->snt_isn) {
484
			NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
485 486 487 488 489 490 491 492 493
			goto out;
		}

		/*
		 * 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().
		 */
494
		inet_csk_reqsk_queue_drop(sk, req, prev);
L
Linus Torvalds 已提交
495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
		goto out;

	case TCP_SYN_SENT:
	case TCP_SYN_RECV:  /* Cannot happen.
			       It can f.e. if SYNs crossed.
			     */
		if (!sock_owned_by_user(sk)) {
			sk->sk_err = err;

			sk->sk_error_report(sk);

			tcp_done(sk);
		} else {
			sk->sk_err_soft = err;
		}
		goto out;
	}

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

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

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

542 543
static 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

568 569
int tcp_v4_gso_send_check(struct sk_buff *skb)
{
570
	const struct iphdr *iph;
571 572 573 574 575
	struct tcphdr *th;

	if (!pskb_may_pull(skb, sizeof(*th)))
		return -EINVAL;

576
	iph = ip_hdr(skb);
577
	th = tcp_hdr(skb);
578 579

	th->check = 0;
580
	skb->ip_summed = CHECKSUM_PARTIAL;
581
	__tcp_v4_send_check(skb, iph->saddr, iph->daddr);
582 583 584
	return 0;
}

L
Linus Torvalds 已提交
585 586 587 588 589 590 591 592 593 594 595 596 597
/*
 *	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.
 */

598
static void tcp_v4_send_reset(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
599
{
600
	const struct tcphdr *th = tcp_hdr(skb);
601 602 603
	struct {
		struct tcphdr th;
#ifdef CONFIG_TCP_MD5SIG
604
		__be32 opt[(TCPOLEN_MD5SIG_ALIGNED >> 2)];
605 606
#endif
	} rep;
L
Linus Torvalds 已提交
607
	struct ip_reply_arg arg;
608 609
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
610 611 612 613
	const __u8 *hash_location = NULL;
	unsigned char newhash[16];
	int genhash;
	struct sock *sk1 = NULL;
614
#endif
615
	struct net *net;
L
Linus Torvalds 已提交
616 617 618 619 620

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

E
Eric Dumazet 已提交
621
	if (skb_rtable(skb)->rt_type != RTN_LOCAL)
L
Linus Torvalds 已提交
622 623 624
		return;

	/* Swap the send and the receive. */
625 626 627 628 629
	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
630 631

	if (th->ack) {
632
		rep.th.seq = th->ack_seq;
L
Linus Torvalds 已提交
633
	} else {
634 635 636
		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
637 638
	}

639
	memset(&arg, 0, sizeof(arg));
640 641 642 643
	arg.iov[0].iov_base = (unsigned char *)&rep;
	arg.iov[0].iov_len  = sizeof(rep.th);

#ifdef CONFIG_TCP_MD5SIG
644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
	hash_location = tcp_parse_md5sig_option(th);
	if (!sk && hash_location) {
		/*
		 * 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.
		 */
		sk1 = __inet_lookup_listener(dev_net(skb_dst(skb)->dev),
					     &tcp_hashinfo, ip_hdr(skb)->daddr,
					     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;

		genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, NULL, skb);
		if (genhash || memcmp(hash_location, newhash, 16) != 0)
			goto release_sk1;
	} else {
		key = sk ? tcp_md5_do_lookup(sk, (union tcp_md5_addr *)
					     &ip_hdr(skb)->saddr,
					     AF_INET) : NULL;
	}

674 675 676 677 678 679 680 681 682
	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;

683
		tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[1],
684 685
				     key, ip_hdr(skb)->saddr,
				     ip_hdr(skb)->daddr, &rep.th);
686 687
	}
#endif
688 689
	arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
				      ip_hdr(skb)->saddr, /* XXX */
690
				      arg.iov[0].iov_len, IPPROTO_TCP, 0);
L
Linus Torvalds 已提交
691
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
692
	arg.flags = (sk && inet_sk(sk)->transparent) ? IP_REPLY_ARG_NOSRCCHECK : 0;
693 694 695 696 697
	/* When socket is gone, all binding information is lost.
	 * routing might fail in this case. using iif for oif to
	 * make sure we can deliver it
	 */
	arg.bound_dev_if = sk ? sk->sk_bound_dev_if : inet_iif(skb);
L
Linus Torvalds 已提交
698

E
Eric Dumazet 已提交
699
	net = dev_net(skb_dst(skb)->dev);
700
	arg.tos = ip_hdr(skb)->tos;
701 702
	ip_send_unicast_reply(net->ipv4.tcp_sock, skb, ip_hdr(skb)->saddr,
			      ip_hdr(skb)->daddr, &arg, arg.iov[0].iov_len);
L
Linus Torvalds 已提交
703

704 705
	TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
	TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS);
706 707 708 709 710 711 712 713

#ifdef CONFIG_TCP_MD5SIG
release_sk1:
	if (sk1) {
		rcu_read_unlock();
		sock_put(sk1);
	}
#endif
L
Linus Torvalds 已提交
714 715 716 717 718 719
}

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

720 721
static void tcp_v4_send_ack(struct sk_buff *skb, u32 seq, u32 ack,
			    u32 win, u32 ts, int oif,
722
			    struct tcp_md5sig_key *key,
723
			    int reply_flags, u8 tos)
L
Linus Torvalds 已提交
724
{
725
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
726 727
	struct {
		struct tcphdr th;
728
		__be32 opt[(TCPOLEN_TSTAMP_ALIGNED >> 2)
729
#ifdef CONFIG_TCP_MD5SIG
730
			   + (TCPOLEN_MD5SIG_ALIGNED >> 2)
731 732
#endif
			];
L
Linus Torvalds 已提交
733 734
	} rep;
	struct ip_reply_arg arg;
E
Eric Dumazet 已提交
735
	struct net *net = dev_net(skb_dst(skb)->dev);
L
Linus Torvalds 已提交
736 737

	memset(&rep.th, 0, sizeof(struct tcphdr));
738
	memset(&arg, 0, sizeof(arg));
L
Linus Torvalds 已提交
739 740 741 742

	arg.iov[0].iov_base = (unsigned char *)&rep;
	arg.iov[0].iov_len  = sizeof(rep.th);
	if (ts) {
743 744 745 746 747
		rep.opt[0] = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				   (TCPOPT_TIMESTAMP << 8) |
				   TCPOLEN_TIMESTAMP);
		rep.opt[1] = htonl(tcp_time_stamp);
		rep.opt[2] = htonl(ts);
748
		arg.iov[0].iov_len += TCPOLEN_TSTAMP_ALIGNED;
L
Linus Torvalds 已提交
749 750 751 752 753 754 755 756 757 758 759
	}

	/* 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);

760 761 762 763 764 765 766 767 768 769 770
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
		int offset = (ts) ? 3 : 0;

		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;

771
		tcp_v4_md5_hash_hdr((__u8 *) &rep.opt[offset],
772 773
				    key, ip_hdr(skb)->saddr,
				    ip_hdr(skb)->daddr, &rep.th);
774 775
	}
#endif
776
	arg.flags = reply_flags;
777 778
	arg.csum = csum_tcpudp_nofold(ip_hdr(skb)->daddr,
				      ip_hdr(skb)->saddr, /* XXX */
L
Linus Torvalds 已提交
779 780
				      arg.iov[0].iov_len, IPPROTO_TCP, 0);
	arg.csumoffset = offsetof(struct tcphdr, check) / 2;
781 782
	if (oif)
		arg.bound_dev_if = oif;
783
	arg.tos = tos;
784 785
	ip_send_unicast_reply(net->ipv4.tcp_sock, skb, ip_hdr(skb)->saddr,
			      ip_hdr(skb)->daddr, &arg, arg.iov[0].iov_len);
L
Linus Torvalds 已提交
786

787
	TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
L
Linus Torvalds 已提交
788 789 790 791
}

static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
792
	struct inet_timewait_sock *tw = inet_twsk(sk);
793
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
794

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

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

807
static void tcp_v4_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
808
				  struct request_sock *req)
L
Linus Torvalds 已提交
809
{
810
	tcp_v4_send_ack(skb, tcp_rsk(req)->snt_isn + 1,
811
			tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd,
812 813
			req->ts_recent,
			0,
E
Eric Dumazet 已提交
814 815
			tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&ip_hdr(skb)->daddr,
					  AF_INET),
816 817
			inet_rsk(req)->no_srccheck ? IP_REPLY_ARG_NOSRCCHECK : 0,
			ip_hdr(skb)->tos);
L
Linus Torvalds 已提交
818 819 820
}

/*
821
 *	Send a SYN-ACK after having received a SYN.
822
 *	This still operates on a request_sock only, not on a big
L
Linus Torvalds 已提交
823 824
 *	socket.
 */
825 826
static int tcp_v4_send_synack(struct sock *sk, struct dst_entry *dst,
			      struct request_sock *req,
827
			      struct request_values *rvp,
828 829
			      u16 queue_mapping,
			      bool nocache)
L
Linus Torvalds 已提交
830
{
831
	const struct inet_request_sock *ireq = inet_rsk(req);
832
	struct flowi4 fl4;
L
Linus Torvalds 已提交
833 834 835 836
	int err = -1;
	struct sk_buff * skb;

	/* First, grab a route. */
837
	if (!dst && (dst = inet_csk_route_req(sk, &fl4, req, nocache)) == NULL)
838
		return -1;
L
Linus Torvalds 已提交
839

840
	skb = tcp_make_synack(sk, dst, req, rvp);
L
Linus Torvalds 已提交
841 842

	if (skb) {
843
		__tcp_v4_send_check(skb, ireq->loc_addr, ireq->rmt_addr);
L
Linus Torvalds 已提交
844

845
		skb_set_queue_mapping(skb, queue_mapping);
846 847 848
		err = ip_build_and_send_pkt(skb, sk, ireq->loc_addr,
					    ireq->rmt_addr,
					    ireq->opt);
849
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
850 851 852 853 854
	}

	return err;
}

855
static int tcp_v4_rtx_synack(struct sock *sk, struct request_sock *req,
856
			      struct request_values *rvp)
857
{
858
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS);
859
	return tcp_v4_send_synack(sk, NULL, req, rvp, 0, false);
860 861
}

L
Linus Torvalds 已提交
862
/*
863
 *	IPv4 request_sock destructor.
L
Linus Torvalds 已提交
864
 */
865
static void tcp_v4_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
866
{
J
Jesper Juhl 已提交
867
	kfree(inet_rsk(req)->opt);
L
Linus Torvalds 已提交
868 869
}

870
/*
E
Eric Dumazet 已提交
871
 * Return true if a syncookie should be sent
872
 */
E
Eric Dumazet 已提交
873
bool tcp_syn_flood_action(struct sock *sk,
874 875
			 const struct sk_buff *skb,
			 const char *proto)
L
Linus Torvalds 已提交
876
{
877
	const char *msg = "Dropping request";
E
Eric Dumazet 已提交
878
	bool want_cookie = false;
879 880 881
	struct listen_sock *lopt;


L
Linus Torvalds 已提交
882

883
#ifdef CONFIG_SYN_COOKIES
884
	if (sysctl_tcp_syncookies) {
885
		msg = "Sending cookies";
E
Eric Dumazet 已提交
886
		want_cookie = true;
887 888
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPREQQFULLDOCOOKIES);
	} else
889
#endif
890 891 892 893 894
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPREQQFULLDROP);

	lopt = inet_csk(sk)->icsk_accept_queue.listen_opt;
	if (!lopt->synflood_warned) {
		lopt->synflood_warned = 1;
895
		pr_info("%s: Possible SYN flooding on port %d. %s.  Check SNMP counters.\n",
896 897 898
			proto, ntohs(tcp_hdr(skb)->dest), msg);
	}
	return want_cookie;
899
}
900
EXPORT_SYMBOL(tcp_syn_flood_action);
L
Linus Torvalds 已提交
901 902

/*
903
 * Save and compile IPv4 options into the request_sock if needed.
L
Linus Torvalds 已提交
904
 */
905 906
static struct ip_options_rcu *tcp_v4_save_options(struct sock *sk,
						  struct sk_buff *skb)
L
Linus Torvalds 已提交
907
{
908 909
	const struct ip_options *opt = &(IPCB(skb)->opt);
	struct ip_options_rcu *dopt = NULL;
L
Linus Torvalds 已提交
910 911

	if (opt && opt->optlen) {
912 913
		int opt_size = sizeof(*dopt) + opt->optlen;

L
Linus Torvalds 已提交
914 915
		dopt = kmalloc(opt_size, GFP_ATOMIC);
		if (dopt) {
916
			if (ip_options_echo(&dopt->opt, skb)) {
L
Linus Torvalds 已提交
917 918 919 920 921 922 923 924
				kfree(dopt);
				dopt = NULL;
			}
		}
	}
	return dopt;
}

925 926 927 928 929 930 931 932
#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.  */
E
Eric Dumazet 已提交
933 934 935
struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
					 const union tcp_md5_addr *addr,
					 int family)
936 937
{
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
938 939 940
	struct tcp_md5sig_key *key;
	struct hlist_node *pos;
	unsigned int size = sizeof(struct in_addr);
941
	struct tcp_md5sig_info *md5sig;
942

943 944
	/* caller either holds rcu_read_lock() or socket lock */
	md5sig = rcu_dereference_check(tp->md5sig_info,
945 946
				       sock_owned_by_user(sk) ||
				       lockdep_is_held(&sk->sk_lock.slock));
947
	if (!md5sig)
948
		return NULL;
E
Eric Dumazet 已提交
949 950 951 952
#if IS_ENABLED(CONFIG_IPV6)
	if (family == AF_INET6)
		size = sizeof(struct in6_addr);
#endif
953
	hlist_for_each_entry_rcu(key, pos, &md5sig->head, node) {
E
Eric Dumazet 已提交
954 955 956 957
		if (key->family != family)
			continue;
		if (!memcmp(&key->addr, addr, size))
			return key;
958 959 960
	}
	return NULL;
}
E
Eric Dumazet 已提交
961
EXPORT_SYMBOL(tcp_md5_do_lookup);
962 963 964 965

struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
					 struct sock *addr_sk)
{
E
Eric Dumazet 已提交
966 967 968 969
	union tcp_md5_addr *addr;

	addr = (union tcp_md5_addr *)&inet_sk(addr_sk)->inet_daddr;
	return tcp_md5_do_lookup(sk, addr, AF_INET);
970 971 972
}
EXPORT_SYMBOL(tcp_v4_md5_lookup);

A
Adrian Bunk 已提交
973 974
static struct tcp_md5sig_key *tcp_v4_reqsk_md5_lookup(struct sock *sk,
						      struct request_sock *req)
975
{
E
Eric Dumazet 已提交
976 977 978 979
	union tcp_md5_addr *addr;

	addr = (union tcp_md5_addr *)&inet_rsk(req)->rmt_addr;
	return tcp_md5_do_lookup(sk, addr, AF_INET);
980 981 982
}

/* This can be called on a newly created socket, from other files */
E
Eric Dumazet 已提交
983 984
int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
		   int family, const u8 *newkey, u8 newkeylen, gfp_t gfp)
985 986
{
	/* Add Key to the list */
987
	struct tcp_md5sig_key *key;
988
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
989
	struct tcp_md5sig_info *md5sig;
990

E
Eric Dumazet 已提交
991
	key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&addr, AF_INET);
992 993
	if (key) {
		/* Pre-existing entry - just update that one. */
E
Eric Dumazet 已提交
994
		memcpy(key->key, newkey, newkeylen);
995
		key->keylen = newkeylen;
E
Eric Dumazet 已提交
996 997
		return 0;
	}
998

999 1000
	md5sig = rcu_dereference_protected(tp->md5sig_info,
					   sock_owned_by_user(sk));
E
Eric Dumazet 已提交
1001 1002 1003
	if (!md5sig) {
		md5sig = kmalloc(sizeof(*md5sig), gfp);
		if (!md5sig)
1004 1005
			return -ENOMEM;

E
Eric Dumazet 已提交
1006 1007
		sk_nocaps_add(sk, NETIF_F_GSO_MASK);
		INIT_HLIST_HEAD(&md5sig->head);
1008
		rcu_assign_pointer(tp->md5sig_info, md5sig);
E
Eric Dumazet 已提交
1009
	}
1010

1011
	key = sock_kmalloc(sk, sizeof(*key), gfp);
E
Eric Dumazet 已提交
1012 1013 1014
	if (!key)
		return -ENOMEM;
	if (hlist_empty(&md5sig->head) && !tcp_alloc_md5sig_pool(sk)) {
1015
		sock_kfree_s(sk, key, sizeof(*key));
E
Eric Dumazet 已提交
1016
		return -ENOMEM;
1017
	}
E
Eric Dumazet 已提交
1018 1019 1020 1021 1022 1023 1024 1025

	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);
1026 1027
	return 0;
}
E
Eric Dumazet 已提交
1028
EXPORT_SYMBOL(tcp_md5_do_add);
1029

E
Eric Dumazet 已提交
1030
int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family)
1031 1032
{
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
1033
	struct tcp_md5sig_key *key;
1034
	struct tcp_md5sig_info *md5sig;
E
Eric Dumazet 已提交
1035 1036 1037 1038 1039

	key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&addr, AF_INET);
	if (!key)
		return -ENOENT;
	hlist_del_rcu(&key->node);
1040
	atomic_sub(sizeof(*key), &sk->sk_omem_alloc);
E
Eric Dumazet 已提交
1041
	kfree_rcu(key, rcu);
1042 1043 1044
	md5sig = rcu_dereference_protected(tp->md5sig_info,
					   sock_owned_by_user(sk));
	if (hlist_empty(&md5sig->head))
E
Eric Dumazet 已提交
1045 1046
		tcp_free_md5sig_pool();
	return 0;
1047
}
E
Eric Dumazet 已提交
1048
EXPORT_SYMBOL(tcp_md5_do_del);
1049

E
Eric Dumazet 已提交
1050
void tcp_clear_md5_list(struct sock *sk)
1051 1052
{
	struct tcp_sock *tp = tcp_sk(sk);
E
Eric Dumazet 已提交
1053 1054
	struct tcp_md5sig_key *key;
	struct hlist_node *pos, *n;
1055
	struct tcp_md5sig_info *md5sig;
1056

1057 1058 1059
	md5sig = rcu_dereference_protected(tp->md5sig_info, 1);

	if (!hlist_empty(&md5sig->head))
1060
		tcp_free_md5sig_pool();
1061
	hlist_for_each_entry_safe(key, pos, n, &md5sig->head, node) {
E
Eric Dumazet 已提交
1062
		hlist_del_rcu(&key->node);
1063
		atomic_sub(sizeof(*key), &sk->sk_omem_alloc);
E
Eric Dumazet 已提交
1064
		kfree_rcu(key, rcu);
1065 1066 1067
	}
}

1068 1069
static int tcp_v4_parse_md5_keys(struct sock *sk, char __user *optval,
				 int optlen)
1070 1071 1072 1073 1074 1075 1076
{
	struct tcp_md5sig cmd;
	struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.tcpm_addr;

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

1077
	if (copy_from_user(&cmd, optval, sizeof(cmd)))
1078 1079 1080 1081 1082
		return -EFAULT;

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

1083
	if (!cmd.tcpm_key || !cmd.tcpm_keylen)
E
Eric Dumazet 已提交
1084 1085
		return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin->sin_addr.s_addr,
				      AF_INET);
1086 1087 1088 1089

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

E
Eric Dumazet 已提交
1090 1091 1092
	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);
1093 1094
}

1095 1096
static int tcp_v4_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
					__be32 daddr, __be32 saddr, int nbytes)
1097 1098
{
	struct tcp4_pseudohdr *bp;
1099
	struct scatterlist sg;
1100 1101 1102 1103

	bp = &hp->md5_blk.ip4;

	/*
1104
	 * 1. the TCP pseudo-header (in the order: source IP address,
1105 1106 1107 1108 1109 1110
	 * destination IP address, zero-padded protocol number, and
	 * segment length)
	 */
	bp->saddr = saddr;
	bp->daddr = daddr;
	bp->pad = 0;
1111
	bp->protocol = IPPROTO_TCP;
1112
	bp->len = cpu_to_be16(nbytes);
1113

1114 1115 1116 1117
	sg_init_one(&sg, bp, sizeof(*bp));
	return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
}

E
Eric Dumazet 已提交
1118
static int tcp_v4_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
1119
			       __be32 daddr, __be32 saddr, const struct tcphdr *th)
1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
{
	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))
1138 1139 1140 1141
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
1142

1143 1144 1145 1146
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
1147
	return 1;
1148 1149
}

1150
int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
1151 1152
			const struct sock *sk, const struct request_sock *req,
			const struct sk_buff *skb)
1153
{
1154 1155
	struct tcp_md5sig_pool *hp;
	struct hash_desc *desc;
E
Eric Dumazet 已提交
1156
	const struct tcphdr *th = tcp_hdr(skb);
1157 1158 1159
	__be32 saddr, daddr;

	if (sk) {
E
Eric Dumazet 已提交
1160 1161
		saddr = inet_sk(sk)->inet_saddr;
		daddr = inet_sk(sk)->inet_daddr;
1162 1163 1164
	} else if (req) {
		saddr = inet_rsk(req)->loc_addr;
		daddr = inet_rsk(req)->rmt_addr;
1165
	} else {
1166 1167 1168
		const struct iphdr *iph = ip_hdr(skb);
		saddr = iph->saddr;
		daddr = iph->daddr;
1169
	}
1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197

	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;
1198
}
1199
EXPORT_SYMBOL(tcp_v4_md5_hash_skb);
1200

E
Eric Dumazet 已提交
1201
static bool tcp_v4_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
1202 1203 1204 1205 1206 1207 1208 1209 1210
{
	/*
	 * 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.
	 */
1211
	const __u8 *hash_location = NULL;
1212
	struct tcp_md5sig_key *hash_expected;
1213
	const struct iphdr *iph = ip_hdr(skb);
1214
	const struct tcphdr *th = tcp_hdr(skb);
1215 1216 1217
	int genhash;
	unsigned char newhash[16];

E
Eric Dumazet 已提交
1218 1219
	hash_expected = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&iph->saddr,
					  AF_INET);
1220
	hash_location = tcp_parse_md5sig_option(th);
1221 1222 1223

	/* We've parsed the options - do we have a hash? */
	if (!hash_expected && !hash_location)
E
Eric Dumazet 已提交
1224
		return false;
1225 1226

	if (hash_expected && !hash_location) {
1227
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
E
Eric Dumazet 已提交
1228
		return true;
1229 1230 1231
	}

	if (!hash_expected && hash_location) {
1232
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
E
Eric Dumazet 已提交
1233
		return true;
1234 1235 1236 1237 1238
	}

	/* Okay, so this is hash_expected and hash_location -
	 * so we need to calculate the checksum.
	 */
1239 1240 1241
	genhash = tcp_v4_md5_hash_skb(newhash,
				      hash_expected,
				      NULL, NULL, skb);
1242 1243

	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
1244 1245 1246 1247 1248
		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
1249
		return true;
1250
	}
E
Eric Dumazet 已提交
1251
	return false;
1252 1253 1254 1255
}

#endif

1256
struct request_sock_ops tcp_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
1257
	.family		=	PF_INET,
1258
	.obj_size	=	sizeof(struct tcp_request_sock),
1259
	.rtx_syn_ack	=	tcp_v4_rtx_synack,
1260 1261
	.send_ack	=	tcp_v4_reqsk_send_ack,
	.destructor	=	tcp_v4_reqsk_destructor,
L
Linus Torvalds 已提交
1262
	.send_reset	=	tcp_v4_send_reset,
1263
	.syn_ack_timeout = 	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
1264 1265
};

1266
#ifdef CONFIG_TCP_MD5SIG
1267
static const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = {
1268
	.md5_lookup	=	tcp_v4_reqsk_md5_lookup,
1269
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1270
};
1271
#endif
1272

L
Linus Torvalds 已提交
1273 1274
int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
{
1275
	struct tcp_extend_values tmp_ext;
L
Linus Torvalds 已提交
1276
	struct tcp_options_received tmp_opt;
1277
	const u8 *hash_location;
1278
	struct request_sock *req;
1279
	struct inet_request_sock *ireq;
1280
	struct tcp_sock *tp = tcp_sk(sk);
1281
	struct dst_entry *dst = NULL;
1282 1283
	__be32 saddr = ip_hdr(skb)->saddr;
	__be32 daddr = ip_hdr(skb)->daddr;
L
Linus Torvalds 已提交
1284
	__u32 isn = TCP_SKB_CB(skb)->when;
E
Eric Dumazet 已提交
1285
	bool want_cookie = false;
L
Linus Torvalds 已提交
1286 1287

	/* Never answer to SYNs send to broadcast or multicast */
E
Eric Dumazet 已提交
1288
	if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
L
Linus Torvalds 已提交
1289 1290 1291 1292 1293 1294
		goto drop;

	/* TW buckets are converted to open requests without
	 * limitations, they conserve resources and peer is
	 * evidently real one.
	 */
1295
	if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
1296 1297 1298
		want_cookie = tcp_syn_flood_action(sk, skb, "TCP");
		if (!want_cookie)
			goto drop;
L
Linus Torvalds 已提交
1299 1300 1301 1302 1303 1304 1305
	}

	/* Accept backlog is full. If we have already queued enough
	 * of warm entries in syn queue, drop request. It is better than
	 * clogging syn queue with openreqs with exponentially increasing
	 * timeout.
	 */
1306
	if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
L
Linus Torvalds 已提交
1307 1308
		goto drop;

1309
	req = inet_reqsk_alloc(&tcp_request_sock_ops);
L
Linus Torvalds 已提交
1310 1311 1312
	if (!req)
		goto drop;

1313 1314 1315 1316
#ifdef CONFIG_TCP_MD5SIG
	tcp_rsk(req)->af_specific = &tcp_request_sock_ipv4_ops;
#endif

L
Linus Torvalds 已提交
1317
	tcp_clear_options(&tmp_opt);
1318
	tmp_opt.mss_clamp = TCP_MSS_DEFAULT;
1319
	tmp_opt.user_mss  = tp->rx_opt.user_mss;
1320
	tcp_parse_options(skb, &tmp_opt, &hash_location, 0);
1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335

	if (tmp_opt.cookie_plus > 0 &&
	    tmp_opt.saw_tstamp &&
	    !tp->rx_opt.cookie_out_never &&
	    (sysctl_tcp_cookie_size > 0 ||
	     (tp->cookie_values != NULL &&
	      tp->cookie_values->cookie_desired > 0))) {
		u8 *c;
		u32 *mess = &tmp_ext.cookie_bakery[COOKIE_DIGEST_WORDS];
		int l = tmp_opt.cookie_plus - TCPOLEN_COOKIE_BASE;

		if (tcp_cookie_generator(&tmp_ext.cookie_bakery[0]) != 0)
			goto drop_and_release;

		/* Secret recipe starts with IP addresses */
1336 1337
		*mess++ ^= (__force u32)daddr;
		*mess++ ^= (__force u32)saddr;
L
Linus Torvalds 已提交
1338

1339 1340 1341 1342 1343
		/* plus variable length Initiator Cookie */
		c = (u8 *)mess;
		while (l-- > 0)
			*c++ ^= *hash_location++;

E
Eric Dumazet 已提交
1344
		want_cookie = false;	/* not our kind of cookie */
1345 1346 1347 1348 1349 1350 1351 1352 1353 1354
		tmp_ext.cookie_out_never = 0; /* false */
		tmp_ext.cookie_plus = tmp_opt.cookie_plus;
	} else if (!tp->rx_opt.cookie_in_always) {
		/* redundant indications, but ensure initialization. */
		tmp_ext.cookie_out_never = 1; /* true */
		tmp_ext.cookie_plus = 0;
	} else {
		goto drop_and_release;
	}
	tmp_ext.cookie_in_always = tp->rx_opt.cookie_in_always;
L
Linus Torvalds 已提交
1355

1356
	if (want_cookie && !tmp_opt.saw_tstamp)
L
Linus Torvalds 已提交
1357 1358 1359 1360 1361
		tcp_clear_options(&tmp_opt);

	tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
	tcp_openreq_init(req, &tmp_opt, skb);

1362 1363 1364 1365 1366 1367
	ireq = inet_rsk(req);
	ireq->loc_addr = daddr;
	ireq->rmt_addr = saddr;
	ireq->no_srccheck = inet_sk(sk)->transparent;
	ireq->opt = tcp_v4_save_options(sk, skb);

1368
	if (security_inet_conn_request(sk, skb, req))
1369
		goto drop_and_free;
1370

1371
	if (!want_cookie || tmp_opt.tstamp_ok)
1372
		TCP_ECN_create_request(req, skb);
L
Linus Torvalds 已提交
1373 1374 1375

	if (want_cookie) {
		isn = cookie_v4_init_sequence(sk, skb, &req->mss);
1376
		req->cookie_ts = tmp_opt.tstamp_ok;
L
Linus Torvalds 已提交
1377 1378
	} else if (!isn) {
		struct inet_peer *peer = NULL;
1379
		struct flowi4 fl4;
L
Linus Torvalds 已提交
1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390

		/* VJ's idea. We save last timestamp seen
		 * from the destination in peer table, when entering
		 * state TIME-WAIT, and check against it before
		 * accepting new connection request.
		 *
		 * If "isn" is not zero, this request hit alive
		 * timewait bucket, so that all the necessary checks
		 * are made in the function processing timewait state.
		 */
		if (tmp_opt.saw_tstamp &&
1391
		    tcp_death_row.sysctl_tw_recycle &&
1392
		    (dst = inet_csk_route_req(sk, &fl4, req, want_cookie)) != NULL &&
1393 1394
		    fl4.daddr == saddr &&
		    (peer = rt_get_peer((struct rtable *)dst, fl4.daddr)) != NULL) {
1395
			inet_peer_refcheck(peer);
E
Eric Dumazet 已提交
1396
			if ((u32)get_seconds() - peer->tcp_ts_stamp < TCP_PAWS_MSL &&
L
Linus Torvalds 已提交
1397 1398
			    (s32)(peer->tcp_ts - req->ts_recent) >
							TCP_PAWS_WINDOW) {
1399
				NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED);
1400
				goto drop_and_release;
L
Linus Torvalds 已提交
1401 1402 1403 1404
			}
		}
		/* Kill the following clause, if you dislike this way. */
		else if (!sysctl_tcp_syncookies &&
1405
			 (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) <
L
Linus Torvalds 已提交
1406 1407
			  (sysctl_max_syn_backlog >> 2)) &&
			 (!peer || !peer->tcp_ts_stamp) &&
1408
			 !tcp_peer_is_proven(req, dst)) {
L
Linus Torvalds 已提交
1409 1410 1411 1412 1413 1414 1415
			/* Without syncookies last quarter of
			 * backlog is filled with destinations,
			 * proven to be alive.
			 * It means that we continue to communicate
			 * to destinations, already remembered
			 * to the moment of synflood.
			 */
1416
			LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("drop open request from %pI4/%u\n"),
1417
				       &saddr, ntohs(tcp_hdr(skb)->source));
1418
			goto drop_and_release;
L
Linus Torvalds 已提交
1419 1420
		}

1421
		isn = tcp_v4_init_sequence(skb);
L
Linus Torvalds 已提交
1422
	}
1423
	tcp_rsk(req)->snt_isn = isn;
1424
	tcp_rsk(req)->snt_synack = tcp_time_stamp;
L
Linus Torvalds 已提交
1425

1426
	if (tcp_v4_send_synack(sk, dst, req,
1427
			       (struct request_values *)&tmp_ext,
1428 1429
			       skb_get_queue_mapping(skb),
			       want_cookie) ||
1430
	    want_cookie)
L
Linus Torvalds 已提交
1431 1432
		goto drop_and_free;

1433
	inet_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
L
Linus Torvalds 已提交
1434 1435
	return 0;

1436 1437
drop_and_release:
	dst_release(dst);
L
Linus Torvalds 已提交
1438
drop_and_free:
1439
	reqsk_free(req);
L
Linus Torvalds 已提交
1440 1441 1442
drop:
	return 0;
}
E
Eric Dumazet 已提交
1443
EXPORT_SYMBOL(tcp_v4_conn_request);
L
Linus Torvalds 已提交
1444 1445 1446 1447 1448 1449 1450


/*
 * The three way handshake has completed - we got a valid synack -
 * now create the new socket.
 */
struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
1451
				  struct request_sock *req,
L
Linus Torvalds 已提交
1452 1453
				  struct dst_entry *dst)
{
1454
	struct inet_request_sock *ireq;
L
Linus Torvalds 已提交
1455 1456 1457
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1458 1459 1460
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1461
	struct ip_options_rcu *inet_opt;
L
Linus Torvalds 已提交
1462 1463 1464 1465 1466 1467

	if (sk_acceptq_is_full(sk))
		goto exit_overflow;

	newsk = tcp_create_openreq_child(sk, req, skb);
	if (!newsk)
1468
		goto exit_nonewsk;
L
Linus Torvalds 已提交
1469

1470
	newsk->sk_gso_type = SKB_GSO_TCPV4;
L
Linus Torvalds 已提交
1471 1472 1473

	newtp		      = tcp_sk(newsk);
	newinet		      = inet_sk(newsk);
1474
	ireq		      = inet_rsk(req);
E
Eric Dumazet 已提交
1475 1476 1477
	newinet->inet_daddr   = ireq->rmt_addr;
	newinet->inet_rcv_saddr = ireq->loc_addr;
	newinet->inet_saddr	      = ireq->loc_addr;
1478 1479
	inet_opt	      = ireq->opt;
	rcu_assign_pointer(newinet->inet_opt, inet_opt);
1480
	ireq->opt	      = NULL;
1481
	newinet->mc_index     = inet_iif(skb);
1482
	newinet->mc_ttl	      = ip_hdr(skb)->ttl;
1483
	newinet->rcv_tos      = ip_hdr(skb)->tos;
1484
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
1485 1486
	if (inet_opt)
		inet_csk(newsk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
E
Eric Dumazet 已提交
1487
	newinet->inet_id = newtp->write_seq ^ jiffies;
L
Linus Torvalds 已提交
1488

E
Eric Dumazet 已提交
1489 1490 1491 1492 1493 1494 1495
	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() */
	}
1496 1497
	sk_setup_caps(newsk, dst);

J
John Heffner 已提交
1498
	tcp_mtup_init(newsk);
L
Linus Torvalds 已提交
1499
	tcp_sync_mss(newsk, dst_mtu(dst));
1500
	newtp->advmss = dst_metric_advmss(dst);
1501 1502 1503 1504
	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
1505
	tcp_initialize_rcv_mss(newsk);
1506 1507 1508 1509
	if (tcp_rsk(req)->snt_synack)
		tcp_valid_rtt_meas(newsk,
		    tcp_time_stamp - tcp_rsk(req)->snt_synack);
	newtp->total_retrans = req->retrans;
L
Linus Torvalds 已提交
1510

1511 1512
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
E
Eric Dumazet 已提交
1513 1514
	key = tcp_md5_do_lookup(sk, (union tcp_md5_addr *)&newinet->inet_daddr,
				AF_INET);
E
Eric Dumazet 已提交
1515
	if (key != NULL) {
1516 1517 1518 1519 1520 1521
		/*
		 * 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
1522 1523
		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
1524
		sk_nocaps_add(newsk, NETIF_F_GSO_MASK);
1525 1526 1527
	}
#endif

1528 1529
	if (__inet_inherit_port(sk, newsk) < 0)
		goto put_and_exit;
1530
	__inet_hash_nolisten(newsk, NULL);
L
Linus Torvalds 已提交
1531 1532 1533 1534

	return newsk;

exit_overflow:
1535
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1536 1537
exit_nonewsk:
	dst_release(dst);
L
Linus Torvalds 已提交
1538
exit:
1539
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1540
	return NULL;
1541
put_and_exit:
1542
	tcp_clear_xmit_timers(newsk);
1543
	tcp_cleanup_congestion_control(newsk);
1544
	bh_unlock_sock(newsk);
1545 1546
	sock_put(newsk);
	goto exit;
L
Linus Torvalds 已提交
1547
}
E
Eric Dumazet 已提交
1548
EXPORT_SYMBOL(tcp_v4_syn_recv_sock);
L
Linus Torvalds 已提交
1549 1550 1551

static struct sock *tcp_v4_hnd_req(struct sock *sk, struct sk_buff *skb)
{
1552
	struct tcphdr *th = tcp_hdr(skb);
1553
	const struct iphdr *iph = ip_hdr(skb);
L
Linus Torvalds 已提交
1554
	struct sock *nsk;
1555
	struct request_sock **prev;
L
Linus Torvalds 已提交
1556
	/* Find possible connection requests. */
1557 1558
	struct request_sock *req = inet_csk_search_req(sk, &prev, th->source,
						       iph->saddr, iph->daddr);
L
Linus Torvalds 已提交
1559 1560 1561
	if (req)
		return tcp_check_req(sk, skb, req, prev);

1562
	nsk = inet_lookup_established(sock_net(sk), &tcp_hashinfo, iph->saddr,
1563
			th->source, iph->daddr, th->dest, inet_iif(skb));
L
Linus Torvalds 已提交
1564 1565 1566 1567 1568 1569

	if (nsk) {
		if (nsk->sk_state != TCP_TIME_WAIT) {
			bh_lock_sock(nsk);
			return nsk;
		}
1570
		inet_twsk_put(inet_twsk(nsk));
L
Linus Torvalds 已提交
1571 1572 1573 1574
		return NULL;
	}

#ifdef CONFIG_SYN_COOKIES
1575
	if (!th->syn)
L
Linus Torvalds 已提交
1576 1577 1578 1579 1580
		sk = cookie_v4_check(sk, skb, &(IPCB(skb)->opt));
#endif
	return sk;
}

1581
static __sum16 tcp_v4_checksum_init(struct sk_buff *skb)
L
Linus Torvalds 已提交
1582
{
1583 1584
	const struct iphdr *iph = ip_hdr(skb);

1585
	if (skb->ip_summed == CHECKSUM_COMPLETE) {
1586 1587
		if (!tcp_v4_check(skb->len, iph->saddr,
				  iph->daddr, skb->csum)) {
1588
			skb->ip_summed = CHECKSUM_UNNECESSARY;
L
Linus Torvalds 已提交
1589
			return 0;
1590
		}
L
Linus Torvalds 已提交
1591
	}
1592

1593
	skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
1594 1595
				       skb->len, IPPROTO_TCP, 0);

L
Linus Torvalds 已提交
1596
	if (skb->len <= 76) {
1597
		return __skb_checksum_complete(skb);
L
Linus Torvalds 已提交
1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612
	}
	return 0;
}


/* The socket must have it's spinlock held when we get
 * here.
 *
 * We have a potential double-lock case here, so even when
 * doing backlog processing we use the BH locking scheme.
 * This is because we cannot sleep with the original spinlock
 * held.
 */
int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
{
1613 1614 1615 1616 1617 1618 1619 1620
	struct sock *rsk;
#ifdef CONFIG_TCP_MD5SIG
	/*
	 * We really want to reject the packet as early as possible
	 * if:
	 *  o We're expecting an MD5'd packet and this is no MD5 tcp option
	 *  o There is an MD5 option and we're not expecting one
	 */
1621
	if (tcp_v4_inbound_md5_hash(sk, skb))
1622 1623 1624
		goto discard;
#endif

L
Linus Torvalds 已提交
1625
	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1626
		sock_rps_save_rxhash(sk, skb);
1627
		if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len)) {
1628
			rsk = sk;
L
Linus Torvalds 已提交
1629
			goto reset;
1630
		}
L
Linus Torvalds 已提交
1631 1632 1633
		return 0;
	}

1634
	if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1635 1636 1637 1638 1639 1640 1641 1642
		goto csum_err;

	if (sk->sk_state == TCP_LISTEN) {
		struct sock *nsk = tcp_v4_hnd_req(sk, skb);
		if (!nsk)
			goto discard;

		if (nsk != sk) {
1643
			sock_rps_save_rxhash(nsk, skb);
1644 1645
			if (tcp_child_process(sk, nsk, skb)) {
				rsk = nsk;
L
Linus Torvalds 已提交
1646
				goto reset;
1647
			}
L
Linus Torvalds 已提交
1648 1649
			return 0;
		}
1650
	} else
1651
		sock_rps_save_rxhash(sk, skb);
1652

1653
	if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) {
1654
		rsk = sk;
L
Linus Torvalds 已提交
1655
		goto reset;
1656
	}
L
Linus Torvalds 已提交
1657 1658 1659
	return 0;

reset:
1660
	tcp_v4_send_reset(rsk, skb);
L
Linus Torvalds 已提交
1661 1662 1663 1664 1665 1666 1667 1668 1669 1670
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:
1671
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1672 1673
	goto discard;
}
E
Eric Dumazet 已提交
1674
EXPORT_SYMBOL(tcp_v4_do_rcv);
L
Linus Torvalds 已提交
1675

1676
void tcp_v4_early_demux(struct sk_buff *skb)
D
David S. Miller 已提交
1677 1678 1679 1680
{
	struct net *net = dev_net(skb->dev);
	const struct iphdr *iph;
	const struct tcphdr *th;
1681
	struct net_device *dev;
D
David S. Miller 已提交
1682 1683 1684
	struct sock *sk;

	if (skb->pkt_type != PACKET_HOST)
1685
		return;
D
David S. Miller 已提交
1686 1687

	if (!pskb_may_pull(skb, ip_hdrlen(skb) + sizeof(struct tcphdr)))
1688
		return;
D
David S. Miller 已提交
1689 1690 1691 1692 1693

	iph = ip_hdr(skb);
	th = (struct tcphdr *) ((char *)iph + ip_hdrlen(skb));

	if (th->doff < sizeof(struct tcphdr) / 4)
1694
		return;
D
David S. Miller 已提交
1695 1696

	if (!pskb_may_pull(skb, ip_hdrlen(skb) + th->doff * 4))
1697
		return;
D
David S. Miller 已提交
1698

1699
	dev = skb->dev;
D
David S. Miller 已提交
1700 1701
	sk = __inet_lookup_established(net, &tcp_hashinfo,
				       iph->saddr, th->source,
1702
				       iph->daddr, ntohs(th->dest),
1703
				       dev->ifindex);
D
David S. Miller 已提交
1704 1705 1706 1707 1708 1709 1710 1711
	if (sk) {
		skb->sk = sk;
		skb->destructor = sock_edemux;
		if (sk->sk_state != TCP_TIME_WAIT) {
			struct dst_entry *dst = sk->sk_rx_dst;
			if (dst)
				dst = dst_check(dst, 0);
			if (dst) {
1712 1713
				struct rtable *rt = (struct rtable *) dst;

1714
				if (rt->rt_iif == dev->ifindex)
1715
					skb_dst_set_noref(skb, dst);
D
David S. Miller 已提交
1716 1717 1718 1719 1720
			}
		}
	}
}

L
Linus Torvalds 已提交
1721 1722 1723 1724 1725 1726
/*
 *	From tcp_input.c
 */

int tcp_v4_rcv(struct sk_buff *skb)
{
1727
	const struct iphdr *iph;
1728
	const struct tcphdr *th;
L
Linus Torvalds 已提交
1729 1730
	struct sock *sk;
	int ret;
1731
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
1732 1733 1734 1735 1736

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

	/* Count it even if it's bad */
1737
	TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1738 1739 1740 1741

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

1742
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1743 1744 1745 1746 1747 1748 1749 1750

	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
1751
	 * provided case of th->doff==0 is eliminated.
L
Linus Torvalds 已提交
1752
	 * So, we defer the checks. */
1753
	if (!skb_csum_unnecessary(skb) && tcp_v4_checksum_init(skb))
L
Linus Torvalds 已提交
1754 1755
		goto bad_packet;

1756
	th = tcp_hdr(skb);
1757
	iph = ip_hdr(skb);
L
Linus Torvalds 已提交
1758 1759 1760 1761 1762
	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);
	TCP_SKB_CB(skb)->when	 = 0;
1763
	TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph);
L
Linus Torvalds 已提交
1764 1765
	TCP_SKB_CB(skb)->sacked	 = 0;

1766
	sk = __inet_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest);
L
Linus Torvalds 已提交
1767 1768 1769
	if (!sk)
		goto no_tcp_socket;

E
Eric Dumazet 已提交
1770 1771 1772 1773
process:
	if (sk->sk_state == TCP_TIME_WAIT)
		goto do_time_wait;

1774 1775
	if (unlikely(iph->ttl < inet_sk(sk)->min_ttl)) {
		NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
1776
		goto discard_and_relse;
1777
	}
1778

L
Linus Torvalds 已提交
1779 1780
	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;
1781
	nf_reset(skb);
L
Linus Torvalds 已提交
1782

1783
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1784 1785 1786 1787
		goto discard_and_relse;

	skb->dev = NULL;

1788
	bh_lock_sock_nested(sk);
L
Linus Torvalds 已提交
1789 1790
	ret = 0;
	if (!sock_owned_by_user(sk)) {
C
Chris Leech 已提交
1791 1792 1793
#ifdef CONFIG_NET_DMA
		struct tcp_sock *tp = tcp_sk(sk);
		if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
1794
			tp->ucopy.dma_chan = net_dma_find_channel();
C
Chris Leech 已提交
1795
		if (tp->ucopy.dma_chan)
L
Linus Torvalds 已提交
1796
			ret = tcp_v4_do_rcv(sk, skb);
C
Chris Leech 已提交
1797 1798 1799 1800
		else
#endif
		{
			if (!tcp_prequeue(sk, skb))
S
Shan Wei 已提交
1801
				ret = tcp_v4_do_rcv(sk, skb);
C
Chris Leech 已提交
1802
		}
1803 1804
	} else if (unlikely(sk_add_backlog(sk, skb,
					   sk->sk_rcvbuf + sk->sk_sndbuf))) {
Z
Zhu Yi 已提交
1805
		bh_unlock_sock(sk);
1806
		NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
Z
Zhu Yi 已提交
1807 1808
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820
	bh_unlock_sock(sk);

	sock_put(sk);

	return ret;

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

	if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) {
bad_packet:
1821
		TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1822
	} else {
1823
		tcp_v4_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1824 1825 1826 1827 1828
	}

discard_it:
	/* Discard frame. */
	kfree_skb(skb);
1829
	return 0;
L
Linus Torvalds 已提交
1830 1831 1832 1833 1834 1835 1836

discard_and_relse:
	sock_put(sk);
	goto discard_it;

do_time_wait:
	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1837
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1838 1839 1840 1841
		goto discard_it;
	}

	if (skb->len < (th->doff << 2) || tcp_checksum_complete(skb)) {
1842
		TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1843
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1844 1845
		goto discard_it;
	}
1846
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1847
	case TCP_TW_SYN: {
1848
		struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev),
1849
							&tcp_hashinfo,
1850
							iph->daddr, th->dest,
1851
							inet_iif(skb));
L
Linus Torvalds 已提交
1852
		if (sk2) {
1853 1854
			inet_twsk_deschedule(inet_twsk(sk), &tcp_death_row);
			inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869
			sk = sk2;
			goto process;
		}
		/* Fall through to ACK */
	}
	case TCP_TW_ACK:
		tcp_v4_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
		goto no_tcp_socket;
	case TCP_TW_SUCCESS:;
	}
	goto discard_it;
}

1870
struct inet_peer *tcp_v4_get_peer(struct sock *sk)
L
Linus Torvalds 已提交
1871
{
1872
	struct rtable *rt = (struct rtable *) __sk_dst_get(sk);
L
Linus Torvalds 已提交
1873 1874
	struct inet_sock *inet = inet_sk(sk);

1875 1876 1877 1878 1879 1880 1881
	/* If we don't have a valid cached route, or we're doing IP
	 * options which make the IPv4 header destination address
	 * different from our peer's, do not bother with this.
	 */
	if (!rt || inet->cork.fl.u.ip4.daddr != inet->inet_daddr)
		return NULL;
	return rt_get_peer_create(rt, inet->inet_daddr);
L
Linus Torvalds 已提交
1882
}
1883
EXPORT_SYMBOL(tcp_v4_get_peer);
L
Linus Torvalds 已提交
1884

1885 1886 1887 1888 1889
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
1890

1891
const struct inet_connection_sock_af_ops ipv4_specific = {
1892 1893 1894 1895 1896
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
	.conn_request	   = tcp_v4_conn_request,
	.syn_recv_sock	   = tcp_v4_syn_recv_sock,
1897
	.get_peer	   = tcp_v4_get_peer,
1898 1899 1900 1901 1902
	.net_header_len	   = sizeof(struct iphdr),
	.setsockopt	   = ip_setsockopt,
	.getsockopt	   = ip_getsockopt,
	.addr2sockaddr	   = inet_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in),
1903
	.bind_conflict	   = inet_csk_bind_conflict,
1904
#ifdef CONFIG_COMPAT
1905 1906
	.compat_setsockopt = compat_ip_setsockopt,
	.compat_getsockopt = compat_ip_getsockopt,
1907
#endif
L
Linus Torvalds 已提交
1908
};
E
Eric Dumazet 已提交
1909
EXPORT_SYMBOL(ipv4_specific);
L
Linus Torvalds 已提交
1910

1911
#ifdef CONFIG_TCP_MD5SIG
1912
static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = {
1913
	.md5_lookup		= tcp_v4_md5_lookup,
1914
	.calc_md5_hash		= tcp_v4_md5_hash_skb,
1915 1916
	.md5_parse		= tcp_v4_parse_md5_keys,
};
1917
#endif
1918

L
Linus Torvalds 已提交
1919 1920 1921 1922 1923
/* 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)
{
1924
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1925

1926
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1927

1928
	icsk->icsk_af_ops = &ipv4_specific;
1929

1930
#ifdef CONFIG_TCP_MD5SIG
1931
	tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific;
1932
#endif
L
Linus Torvalds 已提交
1933 1934 1935 1936

	return 0;
}

1937
void tcp_v4_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1938 1939 1940 1941 1942
{
	struct tcp_sock *tp = tcp_sk(sk);

	tcp_clear_xmit_timers(sk);

1943
	tcp_cleanup_congestion_control(sk);
1944

L
Linus Torvalds 已提交
1945
	/* Cleanup up the write buffer. */
1946
	tcp_write_queue_purge(sk);
L
Linus Torvalds 已提交
1947 1948

	/* Cleans up our, hopefully empty, out_of_order_queue. */
1949
	__skb_queue_purge(&tp->out_of_order_queue);
L
Linus Torvalds 已提交
1950

1951 1952 1953
#ifdef CONFIG_TCP_MD5SIG
	/* Clean up the MD5 key list, if any */
	if (tp->md5sig_info) {
E
Eric Dumazet 已提交
1954
		tcp_clear_md5_list(sk);
1955
		kfree_rcu(tp->md5sig_info, rcu);
1956 1957 1958 1959
		tp->md5sig_info = NULL;
	}
#endif

C
Chris Leech 已提交
1960 1961
#ifdef CONFIG_NET_DMA
	/* Cleans up our sk_async_wait_queue */
1962
	__skb_queue_purge(&sk->sk_async_wait_queue);
C
Chris Leech 已提交
1963 1964
#endif

L
Linus Torvalds 已提交
1965 1966 1967 1968
	/* Clean prequeue, it must be empty really */
	__skb_queue_purge(&tp->ucopy.prequeue);

	/* Clean up a referenced TCP bind bucket. */
1969
	if (inet_csk(sk)->icsk_bind_hash)
1970
		inet_put_port(sk);
L
Linus Torvalds 已提交
1971 1972 1973 1974 1975 1976 1977 1978 1979

	/*
	 * If sendmsg cached page exists, toss it.
	 */
	if (sk->sk_sndmsg_page) {
		__free_page(sk->sk_sndmsg_page);
		sk->sk_sndmsg_page = NULL;
	}

1980 1981 1982 1983 1984 1985 1986
	/* TCP Cookie Transactions */
	if (tp->cookie_values != NULL) {
		kref_put(&tp->cookie_values->kref,
			 tcp_cookie_values_release);
		tp->cookie_values = NULL;
	}

1987
	sk_sockets_allocated_dec(sk);
G
Glauber Costa 已提交
1988
	sock_release_memcg(sk);
L
Linus Torvalds 已提交
1989 1990 1991 1992 1993 1994
}
EXPORT_SYMBOL(tcp_v4_destroy_sock);

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

1995
static inline struct inet_timewait_sock *tw_head(struct hlist_nulls_head *head)
L
Linus Torvalds 已提交
1996
{
1997
	return hlist_nulls_empty(head) ? NULL :
1998
		list_entry(head->first, struct inet_timewait_sock, tw_node);
L
Linus Torvalds 已提交
1999 2000
}

2001
static inline struct inet_timewait_sock *tw_next(struct inet_timewait_sock *tw)
L
Linus Torvalds 已提交
2002
{
2003 2004
	return !is_a_nulls(tw->tw_node.next) ?
		hlist_nulls_entry(tw->tw_node.next, typeof(*tw), tw_node) : NULL;
L
Linus Torvalds 已提交
2005 2006
}

2007 2008 2009 2010 2011
/*
 * 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
2012 2013
static void *listening_get_next(struct seq_file *seq, void *cur)
{
2014
	struct inet_connection_sock *icsk;
2015
	struct hlist_nulls_node *node;
L
Linus Torvalds 已提交
2016
	struct sock *sk = cur;
2017
	struct inet_listen_hashbucket *ilb;
J
Jianjun Kong 已提交
2018
	struct tcp_iter_state *st = seq->private;
2019
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
2020 2021

	if (!sk) {
2022
		ilb = &tcp_hashinfo.listening_hash[st->bucket];
2023
		spin_lock_bh(&ilb->lock);
2024
		sk = sk_nulls_head(&ilb->head);
2025
		st->offset = 0;
L
Linus Torvalds 已提交
2026 2027
		goto get_sk;
	}
2028
	ilb = &tcp_hashinfo.listening_hash[st->bucket];
L
Linus Torvalds 已提交
2029
	++st->num;
2030
	++st->offset;
L
Linus Torvalds 已提交
2031 2032

	if (st->state == TCP_SEQ_STATE_OPENREQ) {
2033
		struct request_sock *req = cur;
L
Linus Torvalds 已提交
2034

2035
		icsk = inet_csk(st->syn_wait_sk);
L
Linus Torvalds 已提交
2036 2037 2038
		req = req->dl_next;
		while (1) {
			while (req) {
2039
				if (req->rsk_ops->family == st->family) {
L
Linus Torvalds 已提交
2040 2041 2042 2043 2044
					cur = req;
					goto out;
				}
				req = req->dl_next;
			}
2045
			if (++st->sbucket >= icsk->icsk_accept_queue.listen_opt->nr_table_entries)
L
Linus Torvalds 已提交
2046 2047
				break;
get_req:
2048
			req = icsk->icsk_accept_queue.listen_opt->syn_table[st->sbucket];
L
Linus Torvalds 已提交
2049
		}
E
Eric Dumazet 已提交
2050
		sk	  = sk_nulls_next(st->syn_wait_sk);
L
Linus Torvalds 已提交
2051
		st->state = TCP_SEQ_STATE_LISTENING;
2052
		read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
L
Linus Torvalds 已提交
2053
	} else {
2054
		icsk = inet_csk(sk);
2055 2056
		read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
		if (reqsk_queue_len(&icsk->icsk_accept_queue))
L
Linus Torvalds 已提交
2057
			goto start_req;
2058
		read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
E
Eric Dumazet 已提交
2059
		sk = sk_nulls_next(sk);
L
Linus Torvalds 已提交
2060 2061
	}
get_sk:
2062
	sk_nulls_for_each_from(sk, node) {
2063 2064 2065
		if (!net_eq(sock_net(sk), net))
			continue;
		if (sk->sk_family == st->family) {
L
Linus Torvalds 已提交
2066 2067 2068
			cur = sk;
			goto out;
		}
2069
		icsk = inet_csk(sk);
2070 2071
		read_lock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
		if (reqsk_queue_len(&icsk->icsk_accept_queue)) {
L
Linus Torvalds 已提交
2072 2073 2074 2075 2076 2077 2078
start_req:
			st->uid		= sock_i_uid(sk);
			st->syn_wait_sk = sk;
			st->state	= TCP_SEQ_STATE_OPENREQ;
			st->sbucket	= 0;
			goto get_req;
		}
2079
		read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
L
Linus Torvalds 已提交
2080
	}
2081
	spin_unlock_bh(&ilb->lock);
2082
	st->offset = 0;
2083
	if (++st->bucket < INET_LHTABLE_SIZE) {
2084 2085
		ilb = &tcp_hashinfo.listening_hash[st->bucket];
		spin_lock_bh(&ilb->lock);
2086
		sk = sk_nulls_head(&ilb->head);
L
Linus Torvalds 已提交
2087 2088 2089 2090 2091 2092 2093 2094 2095
		goto get_sk;
	}
	cur = NULL;
out:
	return cur;
}

static void *listening_get_idx(struct seq_file *seq, loff_t *pos)
{
2096 2097 2098 2099 2100 2101
	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
2102 2103 2104 2105 2106 2107 2108 2109

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

E
Eric Dumazet 已提交
2110
static inline bool empty_bucket(struct tcp_iter_state *st)
2111
{
2112 2113
	return hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].chain) &&
		hlist_nulls_empty(&tcp_hashinfo.ehash[st->bucket].twchain);
2114 2115
}

2116 2117 2118 2119
/*
 * 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
2120 2121
static void *established_get_first(struct seq_file *seq)
{
J
Jianjun Kong 已提交
2122
	struct tcp_iter_state *st = seq->private;
2123
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
2124 2125
	void *rc = NULL;

2126 2127
	st->offset = 0;
	for (; st->bucket <= tcp_hashinfo.ehash_mask; ++st->bucket) {
L
Linus Torvalds 已提交
2128
		struct sock *sk;
2129
		struct hlist_nulls_node *node;
2130
		struct inet_timewait_sock *tw;
2131
		spinlock_t *lock = inet_ehash_lockp(&tcp_hashinfo, st->bucket);
L
Linus Torvalds 已提交
2132

2133 2134 2135 2136
		/* Lockless fast path for the common case of empty buckets */
		if (empty_bucket(st))
			continue;

2137
		spin_lock_bh(lock);
2138
		sk_nulls_for_each(sk, node, &tcp_hashinfo.ehash[st->bucket].chain) {
2139
			if (sk->sk_family != st->family ||
2140
			    !net_eq(sock_net(sk), net)) {
L
Linus Torvalds 已提交
2141 2142 2143 2144 2145 2146
				continue;
			}
			rc = sk;
			goto out;
		}
		st->state = TCP_SEQ_STATE_TIME_WAIT;
2147
		inet_twsk_for_each(tw, node,
2148
				   &tcp_hashinfo.ehash[st->bucket].twchain) {
2149
			if (tw->tw_family != st->family ||
2150
			    !net_eq(twsk_net(tw), net)) {
L
Linus Torvalds 已提交
2151 2152 2153 2154 2155
				continue;
			}
			rc = tw;
			goto out;
		}
2156
		spin_unlock_bh(lock);
L
Linus Torvalds 已提交
2157 2158 2159 2160 2161 2162 2163 2164 2165
		st->state = TCP_SEQ_STATE_ESTABLISHED;
	}
out:
	return rc;
}

static void *established_get_next(struct seq_file *seq, void *cur)
{
	struct sock *sk = cur;
2166
	struct inet_timewait_sock *tw;
2167
	struct hlist_nulls_node *node;
J
Jianjun Kong 已提交
2168
	struct tcp_iter_state *st = seq->private;
2169
	struct net *net = seq_file_net(seq);
L
Linus Torvalds 已提交
2170 2171

	++st->num;
2172
	++st->offset;
L
Linus Torvalds 已提交
2173 2174 2175 2176 2177

	if (st->state == TCP_SEQ_STATE_TIME_WAIT) {
		tw = cur;
		tw = tw_next(tw);
get_tw:
2178
		while (tw && (tw->tw_family != st->family || !net_eq(twsk_net(tw), net))) {
L
Linus Torvalds 已提交
2179 2180 2181 2182 2183 2184
			tw = tw_next(tw);
		}
		if (tw) {
			cur = tw;
			goto out;
		}
2185
		spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
L
Linus Torvalds 已提交
2186 2187
		st->state = TCP_SEQ_STATE_ESTABLISHED;

2188
		/* Look for next non empty bucket */
2189
		st->offset = 0;
2190
		while (++st->bucket <= tcp_hashinfo.ehash_mask &&
2191 2192
				empty_bucket(st))
			;
2193
		if (st->bucket > tcp_hashinfo.ehash_mask)
2194 2195
			return NULL;

2196
		spin_lock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
2197
		sk = sk_nulls_head(&tcp_hashinfo.ehash[st->bucket].chain);
L
Linus Torvalds 已提交
2198
	} else
2199
		sk = sk_nulls_next(sk);
L
Linus Torvalds 已提交
2200

2201
	sk_nulls_for_each_from(sk, node) {
2202
		if (sk->sk_family == st->family && net_eq(sock_net(sk), net))
L
Linus Torvalds 已提交
2203 2204 2205 2206
			goto found;
	}

	st->state = TCP_SEQ_STATE_TIME_WAIT;
2207
	tw = tw_head(&tcp_hashinfo.ehash[st->bucket].twchain);
L
Linus Torvalds 已提交
2208 2209 2210 2211 2212 2213 2214 2215 2216
	goto get_tw;
found:
	cur = sk;
out:
	return cur;
}

static void *established_get_idx(struct seq_file *seq, loff_t pos)
{
2217 2218 2219 2220 2221
	struct tcp_iter_state *st = seq->private;
	void *rc;

	st->bucket = 0;
	rc = established_get_first(seq);
L
Linus Torvalds 已提交
2222 2223 2224 2225

	while (rc && pos) {
		rc = established_get_next(seq, rc);
		--pos;
2226
	}
L
Linus Torvalds 已提交
2227 2228 2229 2230 2231 2232
	return rc;
}

static void *tcp_get_idx(struct seq_file *seq, loff_t pos)
{
	void *rc;
J
Jianjun Kong 已提交
2233
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245

	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;
}

2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280
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_OPENREQ:
	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;
		/* Fallthrough */
	case TCP_SEQ_STATE_ESTABLISHED:
	case TCP_SEQ_STATE_TIME_WAIT:
		st->state = 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
2281 2282
static void *tcp_seq_start(struct seq_file *seq, loff_t *pos)
{
J
Jianjun Kong 已提交
2283
	struct tcp_iter_state *st = seq->private;
2284 2285 2286 2287 2288 2289 2290 2291
	void *rc;

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

L
Linus Torvalds 已提交
2292 2293
	st->state = TCP_SEQ_STATE_LISTENING;
	st->num = 0;
2294 2295 2296 2297 2298 2299 2300
	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
2301 2302 2303 2304
}

static void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
2305
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318
	void *rc = NULL;

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

	switch (st->state) {
	case TCP_SEQ_STATE_OPENREQ:
	case TCP_SEQ_STATE_LISTENING:
		rc = listening_get_next(seq, v);
		if (!rc) {
			st->state = TCP_SEQ_STATE_ESTABLISHED;
2319 2320
			st->bucket = 0;
			st->offset = 0;
L
Linus Torvalds 已提交
2321 2322 2323 2324 2325 2326 2327 2328 2329 2330
			rc	  = established_get_first(seq);
		}
		break;
	case TCP_SEQ_STATE_ESTABLISHED:
	case TCP_SEQ_STATE_TIME_WAIT:
		rc = established_get_next(seq, v);
		break;
	}
out:
	++*pos;
2331
	st->last_pos = *pos;
L
Linus Torvalds 已提交
2332 2333 2334 2335 2336
	return rc;
}

static void tcp_seq_stop(struct seq_file *seq, void *v)
{
J
Jianjun Kong 已提交
2337
	struct tcp_iter_state *st = seq->private;
L
Linus Torvalds 已提交
2338 2339 2340 2341

	switch (st->state) {
	case TCP_SEQ_STATE_OPENREQ:
		if (v) {
2342 2343
			struct inet_connection_sock *icsk = inet_csk(st->syn_wait_sk);
			read_unlock_bh(&icsk->icsk_accept_queue.syn_wait_lock);
L
Linus Torvalds 已提交
2344 2345 2346
		}
	case TCP_SEQ_STATE_LISTENING:
		if (v != SEQ_START_TOKEN)
2347
			spin_unlock_bh(&tcp_hashinfo.listening_hash[st->bucket].lock);
L
Linus Torvalds 已提交
2348 2349 2350 2351
		break;
	case TCP_SEQ_STATE_TIME_WAIT:
	case TCP_SEQ_STATE_ESTABLISHED:
		if (v)
2352
			spin_unlock_bh(inet_ehash_lockp(&tcp_hashinfo, st->bucket));
L
Linus Torvalds 已提交
2353 2354 2355 2356
		break;
	}
}

2357
int tcp_seq_open(struct inode *inode, struct file *file)
L
Linus Torvalds 已提交
2358 2359 2360
{
	struct tcp_seq_afinfo *afinfo = PDE(inode)->data;
	struct tcp_iter_state *s;
2361
	int err;
L
Linus Torvalds 已提交
2362

2363 2364 2365 2366
	err = seq_open_net(inode, file, &afinfo->seq_ops,
			  sizeof(struct tcp_iter_state));
	if (err < 0)
		return err;
2367

2368
	s = ((struct seq_file *)file->private_data)->private;
L
Linus Torvalds 已提交
2369
	s->family		= afinfo->family;
2370
	s->last_pos 		= 0;
2371 2372
	return 0;
}
2373
EXPORT_SYMBOL(tcp_seq_open);
2374

2375
int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo)
L
Linus Torvalds 已提交
2376 2377 2378 2379
{
	int rc = 0;
	struct proc_dir_entry *p;

2380 2381 2382 2383
	afinfo->seq_ops.start		= tcp_seq_start;
	afinfo->seq_ops.next		= tcp_seq_next;
	afinfo->seq_ops.stop		= tcp_seq_stop;

2384
	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
2385
			     afinfo->seq_fops, afinfo);
2386
	if (!p)
L
Linus Torvalds 已提交
2387 2388 2389
		rc = -ENOMEM;
	return rc;
}
E
Eric Dumazet 已提交
2390
EXPORT_SYMBOL(tcp_proc_register);
L
Linus Torvalds 已提交
2391

2392
void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo)
L
Linus Torvalds 已提交
2393
{
2394
	proc_net_remove(net, afinfo->name);
L
Linus Torvalds 已提交
2395
}
E
Eric Dumazet 已提交
2396
EXPORT_SYMBOL(tcp_proc_unregister);
L
Linus Torvalds 已提交
2397

2398
static void get_openreq4(const struct sock *sk, const struct request_sock *req,
2399
			 struct seq_file *f, int i, int uid, int *len)
L
Linus Torvalds 已提交
2400
{
2401
	const struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
2402 2403
	int ttd = req->expires - jiffies;

2404
	seq_printf(f, "%4d: %08X:%04X %08X:%04X"
D
Dan Rosenberg 已提交
2405
		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %u %d %pK%n",
L
Linus Torvalds 已提交
2406
		i,
2407
		ireq->loc_addr,
E
Eric Dumazet 已提交
2408
		ntohs(inet_sk(sk)->inet_sport),
2409 2410
		ireq->rmt_addr,
		ntohs(ireq->rmt_port),
L
Linus Torvalds 已提交
2411 2412 2413 2414 2415 2416 2417 2418 2419
		TCP_SYN_RECV,
		0, 0, /* could print option size, but that is af dependent. */
		1,    /* timers active (only the expire timer) */
		jiffies_to_clock_t(ttd),
		req->retrans,
		uid,
		0,  /* non standard timer */
		0, /* open_requests have no inode */
		atomic_read(&sk->sk_refcnt),
2420 2421
		req,
		len);
L
Linus Torvalds 已提交
2422 2423
}

2424
static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i, int *len)
L
Linus Torvalds 已提交
2425 2426 2427
{
	int timer_active;
	unsigned long timer_expires;
2428
	const struct tcp_sock *tp = tcp_sk(sk);
2429
	const struct inet_connection_sock *icsk = inet_csk(sk);
2430
	const struct inet_sock *inet = inet_sk(sk);
E
Eric Dumazet 已提交
2431 2432 2433 2434
	__be32 dest = inet->inet_daddr;
	__be32 src = inet->inet_rcv_saddr;
	__u16 destp = ntohs(inet->inet_dport);
	__u16 srcp = ntohs(inet->inet_sport);
2435
	int rx_queue;
L
Linus Torvalds 已提交
2436

2437
	if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
L
Linus Torvalds 已提交
2438
		timer_active	= 1;
2439 2440
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
2441
		timer_active	= 4;
2442
		timer_expires	= icsk->icsk_timeout;
2443
	} else if (timer_pending(&sk->sk_timer)) {
L
Linus Torvalds 已提交
2444
		timer_active	= 2;
2445
		timer_expires	= sk->sk_timer.expires;
L
Linus Torvalds 已提交
2446 2447 2448 2449 2450
	} else {
		timer_active	= 0;
		timer_expires = jiffies;
	}

2451 2452 2453 2454 2455 2456 2457 2458
	if (sk->sk_state == TCP_LISTEN)
		rx_queue = sk->sk_ack_backlog;
	else
		/*
		 * because we dont lock socket, we might find a transient negative value
		 */
		rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0);

2459
	seq_printf(f, "%4d: %08X:%04X %08X:%04X %02X %08X:%08X %02X:%08lX "
D
Dan Rosenberg 已提交
2460
			"%08X %5d %8d %lu %d %pK %lu %lu %u %u %d%n",
2461
		i, src, srcp, dest, destp, sk->sk_state,
2462
		tp->write_seq - tp->snd_una,
2463
		rx_queue,
L
Linus Torvalds 已提交
2464 2465
		timer_active,
		jiffies_to_clock_t(timer_expires - jiffies),
2466
		icsk->icsk_retransmits,
2467
		sock_i_uid(sk),
2468
		icsk->icsk_probes_out,
2469 2470
		sock_i_ino(sk),
		atomic_read(&sk->sk_refcnt), sk,
2471 2472
		jiffies_to_clock_t(icsk->icsk_rto),
		jiffies_to_clock_t(icsk->icsk_ack.ato),
2473
		(icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
L
Linus Torvalds 已提交
2474
		tp->snd_cwnd,
I
Ilpo Järvinen 已提交
2475
		tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh,
2476
		len);
L
Linus Torvalds 已提交
2477 2478
}

2479
static void get_timewait4_sock(const struct inet_timewait_sock *tw,
2480
			       struct seq_file *f, int i, int *len)
L
Linus Torvalds 已提交
2481
{
2482
	__be32 dest, src;
L
Linus Torvalds 已提交
2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493
	__u16 destp, srcp;
	int ttd = tw->tw_ttd - jiffies;

	if (ttd < 0)
		ttd = 0;

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

2494
	seq_printf(f, "%4d: %08X:%04X %08X:%04X"
D
Dan Rosenberg 已提交
2495
		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK%n",
L
Linus Torvalds 已提交
2496 2497
		i, src, srcp, dest, destp, tw->tw_substate, 0, 0,
		3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
2498
		atomic_read(&tw->tw_refcnt), tw, len);
L
Linus Torvalds 已提交
2499 2500 2501 2502 2503 2504
}

#define TMPSZ 150

static int tcp4_seq_show(struct seq_file *seq, void *v)
{
J
Jianjun Kong 已提交
2505
	struct tcp_iter_state *st;
2506
	int len;
L
Linus Torvalds 已提交
2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519

	if (v == SEQ_START_TOKEN) {
		seq_printf(seq, "%-*s\n", TMPSZ - 1,
			   "  sl  local_address rem_address   st tx_queue "
			   "rx_queue tr tm->when retrnsmt   uid  timeout "
			   "inode");
		goto out;
	}
	st = seq->private;

	switch (st->state) {
	case TCP_SEQ_STATE_LISTENING:
	case TCP_SEQ_STATE_ESTABLISHED:
2520
		get_tcp4_sock(v, seq, st->num, &len);
L
Linus Torvalds 已提交
2521 2522
		break;
	case TCP_SEQ_STATE_OPENREQ:
2523
		get_openreq4(st->syn_wait_sk, v, seq, st->num, st->uid, &len);
L
Linus Torvalds 已提交
2524 2525
		break;
	case TCP_SEQ_STATE_TIME_WAIT:
2526
		get_timewait4_sock(v, seq, st->num, &len);
L
Linus Torvalds 已提交
2527 2528
		break;
	}
2529
	seq_printf(seq, "%*s\n", TMPSZ - 1 - len, "");
L
Linus Torvalds 已提交
2530 2531 2532 2533
out:
	return 0;
}

2534 2535 2536 2537 2538 2539 2540 2541
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
2542 2543 2544
static struct tcp_seq_afinfo tcp4_seq_afinfo = {
	.name		= "tcp",
	.family		= AF_INET,
2545
	.seq_fops	= &tcp_afinfo_seq_fops,
2546 2547 2548
	.seq_ops	= {
		.show		= tcp4_seq_show,
	},
L
Linus Torvalds 已提交
2549 2550
};

2551
static int __net_init tcp4_proc_init_net(struct net *net)
2552 2553 2554 2555
{
	return tcp_proc_register(net, &tcp4_seq_afinfo);
}

2556
static void __net_exit tcp4_proc_exit_net(struct net *net)
2557 2558 2559 2560 2561 2562 2563 2564 2565
{
	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
2566 2567
int __init tcp4_proc_init(void)
{
2568
	return register_pernet_subsys(&tcp4_net_ops);
L
Linus Torvalds 已提交
2569 2570 2571 2572
}

void tcp4_proc_exit(void)
{
2573
	unregister_pernet_subsys(&tcp4_net_ops);
L
Linus Torvalds 已提交
2574 2575 2576
}
#endif /* CONFIG_PROC_FS */

H
Herbert Xu 已提交
2577 2578
struct sk_buff **tcp4_gro_receive(struct sk_buff **head, struct sk_buff *skb)
{
2579
	const struct iphdr *iph = skb_gro_network_header(skb);
H
Herbert Xu 已提交
2580 2581 2582

	switch (skb->ip_summed) {
	case CHECKSUM_COMPLETE:
2583
		if (!tcp_v4_check(skb_gro_len(skb), iph->saddr, iph->daddr,
H
Herbert Xu 已提交
2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599
				  skb->csum)) {
			skb->ip_summed = CHECKSUM_UNNECESSARY;
			break;
		}

		/* fall through */
	case CHECKSUM_NONE:
		NAPI_GRO_CB(skb)->flush = 1;
		return NULL;
	}

	return tcp_gro_receive(head, skb);
}

int tcp4_gro_complete(struct sk_buff *skb)
{
2600
	const struct iphdr *iph = ip_hdr(skb);
H
Herbert Xu 已提交
2601 2602 2603 2604 2605 2606 2607 2608 2609
	struct tcphdr *th = tcp_hdr(skb);

	th->check = ~tcp_v4_check(skb->len - skb_transport_offset(skb),
				  iph->saddr, iph->daddr, 0);
	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;

	return tcp_gro_complete(skb);
}

L
Linus Torvalds 已提交
2610 2611 2612 2613 2614 2615
struct proto tcp_prot = {
	.name			= "TCP",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v4_connect,
	.disconnect		= tcp_disconnect,
2616
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
2617 2618 2619 2620 2621 2622 2623
	.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,
2624 2625
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
2626
	.backlog_rcv		= tcp_v4_do_rcv,
2627 2628 2629
	.hash			= inet_hash,
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
2630 2631
	.enter_memory_pressure	= tcp_enter_memory_pressure,
	.sockets_allocated	= &tcp_sockets_allocated,
2632
	.orphan_count		= &tcp_orphan_count,
L
Linus Torvalds 已提交
2633 2634 2635 2636 2637 2638
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp_sock),
2639
	.slab_flags		= SLAB_DESTROY_BY_RCU,
2640
	.twsk_prot		= &tcp_timewait_sock_ops,
2641
	.rsk_prot		= &tcp_request_sock_ops,
2642
	.h.hashinfo		= &tcp_hashinfo,
2643
	.no_autobind		= true,
2644 2645 2646 2647
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
#endif
G
Glauber Costa 已提交
2648 2649 2650 2651 2652
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
	.init_cgroup		= tcp_init_cgroup,
	.destroy_cgroup		= tcp_destroy_cgroup,
	.proto_cgroup		= tcp_proto_cgroup,
#endif
L
Linus Torvalds 已提交
2653
};
E
Eric Dumazet 已提交
2654
EXPORT_SYMBOL(tcp_prot);
L
Linus Torvalds 已提交
2655

2656 2657 2658 2659 2660 2661 2662 2663 2664
static int __net_init tcp_sk_init(struct net *net)
{
	return inet_ctl_sock_create(&net->ipv4.tcp_sock,
				    PF_INET, SOCK_RAW, IPPROTO_TCP, net);
}

static void __net_exit tcp_sk_exit(struct net *net)
{
	inet_ctl_sock_destroy(net->ipv4.tcp_sock);
E
Eric W. Biederman 已提交
2665 2666 2667 2668 2669
}

static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list)
{
	inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET);
2670 2671 2672
}

static struct pernet_operations __net_initdata tcp_sk_ops = {
E
Eric W. Biederman 已提交
2673 2674 2675
       .init	   = tcp_sk_init,
       .exit	   = tcp_sk_exit,
       .exit_batch = tcp_sk_exit_batch,
2676 2677
};

2678
void __init tcp_v4_init(void)
L
Linus Torvalds 已提交
2679
{
2680
	inet_hashinfo_init(&tcp_hashinfo);
2681
	if (register_pernet_subsys(&tcp_sk_ops))
L
Linus Torvalds 已提交
2682 2683
		panic("Failed to create the TCP control socket.\n");
}