tcp_ipv6.c 51.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 *	TCP over IPv6
3
 *	Linux INET6 implementation
L
Linus Torvalds 已提交
4 5
 *
 *	Authors:
6
 *	Pedro Roque		<roque@di.fc.ul.pt>
L
Linus Torvalds 已提交
7
 *
8
 *	Based on:
L
Linus Torvalds 已提交
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
 *	linux/net/ipv4/tcp.c
 *	linux/net/ipv4/tcp_input.c
 *	linux/net/ipv4/tcp_output.c
 *
 *	Fixes:
 *	Hideaki YOSHIFUJI	:	sin6_scope_id support
 *	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.
 *	YOSHIFUJI Hideaki @USAGI:	convert /proc/net/tcp6 to seq_file.
 *
 *	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.
 */

H
Herbert Xu 已提交
26
#include <linux/bottom_half.h>
L
Linus Torvalds 已提交
27 28 29 30 31 32 33 34 35 36 37 38 39 40
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/types.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/net.h>
#include <linux/jiffies.h>
#include <linux/in.h>
#include <linux/in6.h>
#include <linux/netdevice.h>
#include <linux/init.h>
#include <linux/jhash.h>
#include <linux/ipsec.h>
#include <linux/times.h>
41
#include <linux/slab.h>
W
Wang Yufen 已提交
42
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
43 44 45 46 47 48
#include <linux/ipv6.h>
#include <linux/icmpv6.h>
#include <linux/random.h>

#include <net/tcp.h>
#include <net/ndisc.h>
49
#include <net/inet6_hashtables.h>
50
#include <net/inet6_connection_sock.h>
L
Linus Torvalds 已提交
51 52 53 54 55 56 57 58 59 60
#include <net/ipv6.h>
#include <net/transp_v6.h>
#include <net/addrconf.h>
#include <net/ip6_route.h>
#include <net/ip6_checksum.h>
#include <net/inet_ecn.h>
#include <net/protocol.h>
#include <net/xfrm.h>
#include <net/snmp.h>
#include <net/dsfield.h>
61
#include <net/timewait_sock.h>
62
#include <net/inet_common.h>
63
#include <net/secure_seq.h>
64
#include <net/busy_poll.h>
L
Linus Torvalds 已提交
65 66 67 68

#include <linux/proc_fs.h>
#include <linux/seq_file.h>

H
Herbert Xu 已提交
69
#include <crypto/hash.h>
70 71
#include <linux/scatterlist.h>

72 73
#include <trace/events/tcp.h>

74 75
static void	tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb);
static void	tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
76
				      struct request_sock *req);
L
Linus Torvalds 已提交
77 78 79

static int	tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);

80 81
static const struct inet_connection_sock_af_ops ipv6_mapped;
static const struct inet_connection_sock_af_ops ipv6_specific;
82
#ifdef CONFIG_TCP_MD5SIG
83 84
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
85
#else
86
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk,
87
						   const struct in6_addr *addr)
88 89 90
{
	return NULL;
}
91
#endif
L
Linus Torvalds 已提交
92

93 94 95 96
static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
{
	struct dst_entry *dst = skb_dst(skb);

E
Eric Dumazet 已提交
97
	if (dst && dst_hold_safe(dst)) {
98 99 100 101
		const struct rt6_info *rt = (const struct rt6_info *)dst;

		sk->sk_rx_dst = dst;
		inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
102
		inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt);
103
	}
104 105
}

106
static u32 tcp_v6_init_seq(const struct sk_buff *skb)
L
Linus Torvalds 已提交
107
{
108 109 110 111 112 113
	return secure_tcpv6_seq(ipv6_hdr(skb)->daddr.s6_addr32,
				ipv6_hdr(skb)->saddr.s6_addr32,
				tcp_hdr(skb)->dest,
				tcp_hdr(skb)->source);
}

114
static u32 tcp_v6_init_ts_off(const struct net *net, const struct sk_buff *skb)
115
{
116
	return secure_tcpv6_ts_off(net, ipv6_hdr(skb)->daddr.s6_addr32,
117
				   ipv6_hdr(skb)->saddr.s6_addr32);
L
Linus Torvalds 已提交
118 119
}

120
static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
L
Linus Torvalds 已提交
121 122 123
			  int addr_len)
{
	struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
124
	struct inet_sock *inet = inet_sk(sk);
125
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
126 127
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
128
	struct in6_addr *saddr = NULL, *final_p, final;
129
	struct ipv6_txoptions *opt;
130
	struct flowi6 fl6;
L
Linus Torvalds 已提交
131 132 133
	struct dst_entry *dst;
	int addr_type;
	int err;
134
	struct inet_timewait_death_row *tcp_death_row = &sock_net(sk)->ipv4.tcp_death_row;
L
Linus Torvalds 已提交
135

136
	if (addr_len < SIN6_LEN_RFC2133)
L
Linus Torvalds 已提交
137 138
		return -EINVAL;

139
	if (usin->sin6_family != AF_INET6)
E
Eric Dumazet 已提交
140
		return -EAFNOSUPPORT;
L
Linus Torvalds 已提交
141

142
	memset(&fl6, 0, sizeof(fl6));
L
Linus Torvalds 已提交
143 144

	if (np->sndflow) {
145 146 147
		fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
		IP6_ECN_flow_init(fl6.flowlabel);
		if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
L
Linus Torvalds 已提交
148
			struct ip6_flowlabel *flowlabel;
149
			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
150
			if (!flowlabel)
L
Linus Torvalds 已提交
151 152 153 154 155 156
				return -EINVAL;
			fl6_sock_release(flowlabel);
		}
	}

	/*
157 158 159
	 *	connect() to INADDR_ANY means loopback (BSD'ism).
	 */

160 161 162 163 164 165 166
	if (ipv6_addr_any(&usin->sin6_addr)) {
		if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr))
			ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
					       &usin->sin6_addr);
		else
			usin->sin6_addr = in6addr_loopback;
	}
L
Linus Torvalds 已提交
167 168 169

	addr_type = ipv6_addr_type(&usin->sin6_addr);

W
Weilong Chen 已提交
170
	if (addr_type & IPV6_ADDR_MULTICAST)
L
Linus Torvalds 已提交
171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
		return -ENETUNREACH;

	if (addr_type&IPV6_ADDR_LINKLOCAL) {
		if (addr_len >= sizeof(struct sockaddr_in6) &&
		    usin->sin6_scope_id) {
			/* If interface is set while binding, indices
			 * must coincide.
			 */
			if (sk->sk_bound_dev_if &&
			    sk->sk_bound_dev_if != usin->sin6_scope_id)
				return -EINVAL;

			sk->sk_bound_dev_if = usin->sin6_scope_id;
		}

		/* Connect to link-local address requires an interface */
		if (!sk->sk_bound_dev_if)
			return -EINVAL;
	}

	if (tp->rx_opt.ts_recent_stamp &&
192
	    !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) {
L
Linus Torvalds 已提交
193 194 195 196 197
		tp->rx_opt.ts_recent = 0;
		tp->rx_opt.ts_recent_stamp = 0;
		tp->write_seq = 0;
	}

198
	sk->sk_v6_daddr = usin->sin6_addr;
199
	np->flow_label = fl6.flowlabel;
L
Linus Torvalds 已提交
200 201 202 203 204

	/*
	 *	TCP over IPv4
	 */

205
	if (addr_type & IPV6_ADDR_MAPPED) {
206
		u32 exthdrlen = icsk->icsk_ext_hdr_len;
L
Linus Torvalds 已提交
207 208 209 210 211 212 213 214 215 216 217
		struct sockaddr_in sin;

		SOCK_DEBUG(sk, "connect: ipv4 mapped\n");

		if (__ipv6_only_sock(sk))
			return -ENETUNREACH;

		sin.sin_family = AF_INET;
		sin.sin_port = usin->sin6_port;
		sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];

218
		icsk->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
219
		sk->sk_backlog_rcv = tcp_v4_do_rcv;
220 221 222
#ifdef CONFIG_TCP_MD5SIG
		tp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif
L
Linus Torvalds 已提交
223 224 225 226

		err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));

		if (err) {
227 228
			icsk->icsk_ext_hdr_len = exthdrlen;
			icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
229
			sk->sk_backlog_rcv = tcp_v6_do_rcv;
230 231 232
#ifdef CONFIG_TCP_MD5SIG
			tp->af_specific = &tcp_sock_ipv6_specific;
#endif
L
Linus Torvalds 已提交
233 234
			goto failure;
		}
235
		np->saddr = sk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
236 237 238 239

		return err;
	}

240 241
	if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
		saddr = &sk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
242

243
	fl6.flowi6_proto = IPPROTO_TCP;
244
	fl6.daddr = sk->sk_v6_daddr;
A
Alexey Dobriyan 已提交
245
	fl6.saddr = saddr ? *saddr : np->saddr;
246 247
	fl6.flowi6_oif = sk->sk_bound_dev_if;
	fl6.flowi6_mark = sk->sk_mark;
248 249
	fl6.fl6_dport = usin->sin6_port;
	fl6.fl6_sport = inet->inet_sport;
250
	fl6.flowi6_uid = sk->sk_uid;
L
Linus Torvalds 已提交
251

252
	opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
253
	final_p = fl6_update_dst(&fl6, opt, &final);
L
Linus Torvalds 已提交
254

255
	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
V
Venkat Yekkirala 已提交
256

257
	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
258 259
	if (IS_ERR(dst)) {
		err = PTR_ERR(dst);
L
Linus Torvalds 已提交
260
		goto failure;
261
	}
L
Linus Torvalds 已提交
262

263
	if (!saddr) {
264
		saddr = &fl6.saddr;
265
		sk->sk_v6_rcv_saddr = *saddr;
L
Linus Torvalds 已提交
266 267 268
	}

	/* set the source address */
A
Alexey Dobriyan 已提交
269
	np->saddr = *saddr;
E
Eric Dumazet 已提交
270
	inet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
271

H
Herbert Xu 已提交
272
	sk->sk_gso_type = SKB_GSO_TCPV6;
E
Eric Dumazet 已提交
273
	ip6_dst_store(sk, dst, NULL, NULL);
L
Linus Torvalds 已提交
274

275
	icsk->icsk_ext_hdr_len = 0;
276 277 278
	if (opt)
		icsk->icsk_ext_hdr_len = opt->opt_flen +
					 opt->opt_nflen;
L
Linus Torvalds 已提交
279 280 281

	tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);

E
Eric Dumazet 已提交
282
	inet->inet_dport = usin->sin6_port;
L
Linus Torvalds 已提交
283 284

	tcp_set_state(sk, TCP_SYN_SENT);
285
	err = inet6_hash_connect(tcp_death_row, sk);
L
Linus Torvalds 已提交
286 287 288
	if (err)
		goto late_failure;

289
	sk_set_txhash(sk);
290

291 292
	if (likely(!tp->repair)) {
		if (!tp->write_seq)
293 294 295 296
			tp->write_seq = secure_tcpv6_seq(np->saddr.s6_addr32,
							 sk->sk_v6_daddr.s6_addr32,
							 inet->inet_sport,
							 inet->inet_dport);
297 298
		tp->tsoffset = secure_tcpv6_ts_off(sock_net(sk),
						   np->saddr.s6_addr32,
299
						   sk->sk_v6_daddr.s6_addr32);
300
	}
L
Linus Torvalds 已提交
301

W
Wei Wang 已提交
302 303 304 305 306
	if (tcp_fastopen_defer_connect(sk, &err))
		return err;
	if (err)
		goto late_failure;

L
Linus Torvalds 已提交
307 308 309 310 311 312 313 314 315
	err = tcp_connect(sk);
	if (err)
		goto late_failure;

	return 0;

late_failure:
	tcp_set_state(sk, TCP_CLOSE);
failure:
E
Eric Dumazet 已提交
316
	inet->inet_dport = 0;
L
Linus Torvalds 已提交
317 318 319 320
	sk->sk_route_caps = 0;
	return err;
}

321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337
static void tcp_v6_mtu_reduced(struct sock *sk)
{
	struct dst_entry *dst;

	if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
		return;

	dst = inet6_csk_update_pmtu(sk, tcp_sk(sk)->mtu_info);
	if (!dst)
		return;

	if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
		tcp_sync_mss(sk, dst_mtu(dst));
		tcp_simple_retransmit(sk);
	}
}

L
Linus Torvalds 已提交
338
static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
339
		u8 type, u8 code, int offset, __be32 info)
L
Linus Torvalds 已提交
340
{
W
Weilong Chen 已提交
341
	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
342
	const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
343 344
	struct net *net = dev_net(skb->dev);
	struct request_sock *fastopen;
L
Linus Torvalds 已提交
345
	struct ipv6_pinfo *np;
346
	struct tcp_sock *tp;
347
	__u32 seq, snd_una;
348
	struct sock *sk;
349
	bool fatal;
350
	int err;
L
Linus Torvalds 已提交
351

352 353 354
	sk = __inet6_lookup_established(net, &tcp_hashinfo,
					&hdr->daddr, th->dest,
					&hdr->saddr, ntohs(th->source),
355
					skb->dev->ifindex, inet6_sdif(skb));
L
Linus Torvalds 已提交
356

357
	if (!sk) {
E
Eric Dumazet 已提交
358 359
		__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
				  ICMP6_MIB_INERRORS);
L
Linus Torvalds 已提交
360 361 362 363
		return;
	}

	if (sk->sk_state == TCP_TIME_WAIT) {
364
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
365 366
		return;
	}
367
	seq = ntohl(th->seq);
368
	fatal = icmpv6_err_convert(type, code, &err);
369
	if (sk->sk_state == TCP_NEW_SYN_RECV)
370
		return tcp_req_err(sk, seq, fatal);
L
Linus Torvalds 已提交
371 372

	bh_lock_sock(sk);
373
	if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG)
374
		__NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
L
Linus Torvalds 已提交
375 376 377 378

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

379
	if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) {
380
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
381 382 383
		goto out;
	}

L
Linus Torvalds 已提交
384
	tp = tcp_sk(sk);
385 386 387
	/* 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
388
	if (sk->sk_state != TCP_LISTEN &&
389
	    !between(seq, snd_una, tp->snd_nxt)) {
390
		__NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
391 392 393 394 395
		goto out;
	}

	np = inet6_sk(sk);

396
	if (type == NDISC_REDIRECT) {
397 398
		if (!sock_owned_by_user(sk)) {
			struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
399

400 401 402
			if (dst)
				dst->ops->redirect(dst, sk, skb);
		}
403
		goto out;
404 405
	}

L
Linus Torvalds 已提交
406
	if (type == ICMPV6_PKT_TOOBIG) {
407 408 409 410 411 412 413
		/* 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;

414 415 416
		if (!ip6_sk_accept_pmtu(sk))
			goto out;

417 418 419
		tp->mtu_info = ntohl(info);
		if (!sock_owned_by_user(sk))
			tcp_v6_mtu_reduced(sk);
420
		else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED,
421
					   &sk->sk_tsq_flags))
422
			sock_hold(sk);
L
Linus Torvalds 已提交
423 424 425 426
		goto out;
	}


427
	/* Might be for an request_sock */
L
Linus Torvalds 已提交
428 429
	switch (sk->sk_state) {
	case TCP_SYN_SENT:
430 431 432 433
	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.
		 */
434
		if (fastopen && !fastopen->sk)
435 436
			break;

L
Linus Torvalds 已提交
437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458
		if (!sock_owned_by_user(sk)) {
			sk->sk_err = err;
			sk->sk_error_report(sk);		/* Wake people up to see the error (see connect in sock.c) */

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

	if (!sock_owned_by_user(sk) && np->recverr) {
		sk->sk_err = err;
		sk->sk_error_report(sk);
	} else
		sk->sk_err_soft = err;

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


459
static int tcp_v6_send_synack(const struct sock *sk, struct dst_entry *dst,
460
			      struct flowi *fl,
461
			      struct request_sock *req,
462
			      struct tcp_fastopen_cookie *foc,
463
			      enum tcp_synack_type synack_type)
L
Linus Torvalds 已提交
464
{
465
	struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
466
	struct ipv6_pinfo *np = inet6_sk(sk);
467
	struct ipv6_txoptions *opt;
468
	struct flowi6 *fl6 = &fl->u.ip6;
W
Weilong Chen 已提交
469
	struct sk_buff *skb;
470
	int err = -ENOMEM;
L
Linus Torvalds 已提交
471

472
	/* First, grab a route. */
473 474
	if (!dst && (dst = inet6_csk_route_req(sk, fl6, req,
					       IPPROTO_TCP)) == NULL)
475
		goto done;
476

477
	skb = tcp_make_synack(sk, dst, req, foc, synack_type);
478

L
Linus Torvalds 已提交
479
	if (skb) {
480 481
		__tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr,
				    &ireq->ir_v6_rmt_addr);
L
Linus Torvalds 已提交
482

483
		fl6->daddr = ireq->ir_v6_rmt_addr;
484
		if (np->repflow && ireq->pktopts)
485 486
			fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts));

487
		rcu_read_lock();
488 489 490
		opt = ireq->ipv6_opt;
		if (!opt)
			opt = rcu_dereference(np->opt);
491
		err = ip6_xmit(sk, skb, fl6, sk->sk_mark, opt, np->tclass);
492
		rcu_read_unlock();
493
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
494 495 496 497 498 499
	}

done:
	return err;
}

500

501
static void tcp_v6_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
502
{
503
	kfree(inet_rsk(req)->ipv6_opt);
504
	kfree_skb(inet_rsk(req)->pktopts);
L
Linus Torvalds 已提交
505 506
}

507
#ifdef CONFIG_TCP_MD5SIG
508
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk,
509
						   const struct in6_addr *addr)
510
{
E
Eric Dumazet 已提交
511
	return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
512 513
}

514
static struct tcp_md5sig_key *tcp_v6_md5_lookup(const struct sock *sk,
515
						const struct sock *addr_sk)
516
{
517
	return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr);
518 519
}

520 521
static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
				 char __user *optval, int optlen)
522 523 524
{
	struct tcp_md5sig cmd;
	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
525
	u8 prefixlen;
526 527 528 529 530 531 532 533 534 535

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

	if (copy_from_user(&cmd, optval, sizeof(cmd)))
		return -EFAULT;

	if (sin6->sin6_family != AF_INET6)
		return -EINVAL;

536 537 538 539 540 541 542 543 544 545
	if (optname == TCP_MD5SIG_EXT &&
	    cmd.tcpm_flags & TCP_MD5SIG_FLAG_PREFIX) {
		prefixlen = cmd.tcpm_prefixlen;
		if (prefixlen > 128 || (ipv6_addr_v4mapped(&sin6->sin6_addr) &&
					prefixlen > 32))
			return -EINVAL;
	} else {
		prefixlen = ipv6_addr_v4mapped(&sin6->sin6_addr) ? 32 : 128;
	}

546
	if (!cmd.tcpm_keylen) {
B
Brian Haley 已提交
547
		if (ipv6_addr_v4mapped(&sin6->sin6_addr))
E
Eric Dumazet 已提交
548
			return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
549
					      AF_INET, prefixlen);
E
Eric Dumazet 已提交
550
		return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
551
				      AF_INET6, prefixlen);
552 553 554 555 556
	}

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

E
Eric Dumazet 已提交
557 558
	if (ipv6_addr_v4mapped(&sin6->sin6_addr))
		return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
559
				      AF_INET, prefixlen, cmd.tcpm_key,
560
				      cmd.tcpm_keylen, GFP_KERNEL);
561

E
Eric Dumazet 已提交
562
	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
563 564
			      AF_INET6, prefixlen, cmd.tcpm_key,
			      cmd.tcpm_keylen, GFP_KERNEL);
565 566
}

567 568 569 570
static int tcp_v6_md5_hash_headers(struct tcp_md5sig_pool *hp,
				   const struct in6_addr *daddr,
				   const struct in6_addr *saddr,
				   const struct tcphdr *th, int nbytes)
571 572
{
	struct tcp6_pseudohdr *bp;
573
	struct scatterlist sg;
574
	struct tcphdr *_th;
575

576
	bp = hp->scratch;
577
	/* 1. TCP pseudo-header (RFC2460) */
A
Alexey Dobriyan 已提交
578 579
	bp->saddr = *saddr;
	bp->daddr = *daddr;
580
	bp->protocol = cpu_to_be32(IPPROTO_TCP);
A
Adam Langley 已提交
581
	bp->len = cpu_to_be32(nbytes);
582

583 584 585 586 587 588 589
	_th = (struct tcphdr *)(bp + 1);
	memcpy(_th, th, sizeof(*th));
	_th->check = 0;

	sg_init_one(&sg, bp, sizeof(*bp) + sizeof(*th));
	ahash_request_set_crypt(hp->md5_req, &sg, NULL,
				sizeof(*bp) + sizeof(*th));
H
Herbert Xu 已提交
590
	return crypto_ahash_update(hp->md5_req);
591
}
592

593
static int tcp_v6_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
594
			       const struct in6_addr *daddr, struct in6_addr *saddr,
E
Eric Dumazet 已提交
595
			       const struct tcphdr *th)
596 597
{
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
598
	struct ahash_request *req;
599 600 601 602

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
603
	req = hp->md5_req;
604

H
Herbert Xu 已提交
605
	if (crypto_ahash_init(req))
606
		goto clear_hash;
607
	if (tcp_v6_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2))
608 609 610
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
H
Herbert Xu 已提交
611 612
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
613 614 615 616
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
617

618 619 620 621
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
622
	return 1;
623 624
}

625 626
static int tcp_v6_md5_hash_skb(char *md5_hash,
			       const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
627 628
			       const struct sock *sk,
			       const struct sk_buff *skb)
629
{
630
	const struct in6_addr *saddr, *daddr;
631
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
632
	struct ahash_request *req;
E
Eric Dumazet 已提交
633
	const struct tcphdr *th = tcp_hdr(skb);
634

635 636
	if (sk) { /* valid for establish/request sockets */
		saddr = &sk->sk_v6_rcv_saddr;
637
		daddr = &sk->sk_v6_daddr;
638
	} else {
639
		const struct ipv6hdr *ip6h = ipv6_hdr(skb);
640 641
		saddr = &ip6h->saddr;
		daddr = &ip6h->daddr;
642
	}
643 644 645 646

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
647
	req = hp->md5_req;
648

H
Herbert Xu 已提交
649
	if (crypto_ahash_init(req))
650 651
		goto clear_hash;

652
	if (tcp_v6_md5_hash_headers(hp, daddr, saddr, th, skb->len))
653 654 655 656 657
		goto clear_hash;
	if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
H
Herbert Xu 已提交
658 659
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
660 661 662 663 664 665 666 667 668 669
		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;
670 671
}

672 673 674 675
#endif

static bool tcp_v6_inbound_md5_hash(const struct sock *sk,
				    const struct sk_buff *skb)
676
{
677
#ifdef CONFIG_TCP_MD5SIG
678
	const __u8 *hash_location = NULL;
679
	struct tcp_md5sig_key *hash_expected;
680
	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
E
Eric Dumazet 已提交
681
	const struct tcphdr *th = tcp_hdr(skb);
682 683 684 685
	int genhash;
	u8 newhash[16];

	hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
686
	hash_location = tcp_parse_md5sig_option(th);
687

688 689
	/* We've parsed the options - do we have a hash? */
	if (!hash_expected && !hash_location)
690
		return false;
691 692

	if (hash_expected && !hash_location) {
693
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
694
		return true;
695 696
	}

697
	if (!hash_expected && hash_location) {
698
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
699
		return true;
700 701 702
	}

	/* check the signature */
703 704
	genhash = tcp_v6_md5_hash_skb(newhash,
				      hash_expected,
705
				      NULL, skb);
706

707
	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
708
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE);
709 710 711 712
		net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n",
				     genhash ? "failed" : "mismatch",
				     &ip6h->saddr, ntohs(th->source),
				     &ip6h->daddr, ntohs(th->dest));
713
		return true;
714
	}
715
#endif
716
	return false;
717 718
}

719 720
static void tcp_v6_init_req(struct request_sock *req,
			    const struct sock *sk_listener,
721 722 723
			    struct sk_buff *skb)
{
	struct inet_request_sock *ireq = inet_rsk(req);
724
	const struct ipv6_pinfo *np = inet6_sk(sk_listener);
725 726 727 728 729

	ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
	ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;

	/* So that link locals have meaning */
730
	if (!sk_listener->sk_bound_dev_if &&
731
	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
E
Eric Dumazet 已提交
732
		ireq->ir_iif = tcp_v6_iif(skb);
733

734
	if (!TCP_SKB_CB(skb)->tcp_tw_isn &&
735
	    (ipv6_opt_accepted(sk_listener, skb, &TCP_SKB_CB(skb)->header.h6) ||
736
	     np->rxopt.bits.rxinfo ||
737 738
	     np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim ||
	     np->rxopt.bits.rxohlim || np->repflow)) {
739
		refcount_inc(&skb->users);
740 741 742 743
		ireq->pktopts = skb;
	}
}

744 745
static struct dst_entry *tcp_v6_route_req(const struct sock *sk,
					  struct flowi *fl,
746
					  const struct request_sock *req)
747
{
748
	return inet6_csk_route_req(sk, &fl->u.ip6, req, IPPROTO_TCP);
749 750
}

751
struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
752
	.family		=	AF_INET6,
753
	.obj_size	=	sizeof(struct tcp6_request_sock),
754
	.rtx_syn_ack	=	tcp_rtx_synack,
755 756
	.send_ack	=	tcp_v6_reqsk_send_ack,
	.destructor	=	tcp_v6_reqsk_destructor,
757
	.send_reset	=	tcp_v6_send_reset,
W
Wang Yufen 已提交
758
	.syn_ack_timeout =	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
759 760
};

761
static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
762 763
	.mss_clamp	=	IPV6_MIN_MTU - sizeof(struct tcphdr) -
				sizeof(struct ipv6hdr),
764
#ifdef CONFIG_TCP_MD5SIG
765
	.req_md5_lookup	=	tcp_v6_md5_lookup,
766
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
767
#endif
768
	.init_req	=	tcp_v6_init_req,
769 770 771
#ifdef CONFIG_SYN_COOKIES
	.cookie_init_seq =	cookie_v6_init_sequence,
#endif
772
	.route_req	=	tcp_v6_route_req,
773 774
	.init_seq	=	tcp_v6_init_seq,
	.init_ts_off	=	tcp_v6_init_ts_off,
775
	.send_synack	=	tcp_v6_send_synack,
776
};
777

778
static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 seq,
779 780
				 u32 ack, u32 win, u32 tsval, u32 tsecr,
				 int oif, struct tcp_md5sig_key *key, int rst,
781
				 u8 tclass, __be32 label)
L
Linus Torvalds 已提交
782
{
783 784
	const struct tcphdr *th = tcp_hdr(skb);
	struct tcphdr *t1;
L
Linus Torvalds 已提交
785
	struct sk_buff *buff;
786
	struct flowi6 fl6;
787
	struct net *net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
788
	struct sock *ctl_sk = net->ipv6.tcp_sk;
789
	unsigned int tot_len = sizeof(struct tcphdr);
E
Eric Dumazet 已提交
790
	struct dst_entry *dst;
791
	__be32 *topt;
L
Linus Torvalds 已提交
792

793
	if (tsecr)
794
		tot_len += TCPOLEN_TSTAMP_ALIGNED;
795 796 797 798 799 800
#ifdef CONFIG_TCP_MD5SIG
	if (key)
		tot_len += TCPOLEN_MD5SIG_ALIGNED;
#endif

	buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
L
Linus Torvalds 已提交
801
			 GFP_ATOMIC);
802
	if (!buff)
803
		return;
L
Linus Torvalds 已提交
804

805
	skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
L
Linus Torvalds 已提交
806

807
	t1 = skb_push(buff, tot_len);
808
	skb_reset_transport_header(buff);
L
Linus Torvalds 已提交
809 810 811 812 813

	/* Swap the send and the receive. */
	memset(t1, 0, sizeof(*t1));
	t1->dest = th->source;
	t1->source = th->dest;
814
	t1->doff = tot_len / 4;
815 816 817 818 819
	t1->seq = htonl(seq);
	t1->ack_seq = htonl(ack);
	t1->ack = !rst || !th->ack;
	t1->rst = rst;
	t1->window = htons(win);
L
Linus Torvalds 已提交
820

821 822
	topt = (__be32 *)(t1 + 1);

823
	if (tsecr) {
824 825
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
826 827
		*topt++ = htonl(tsval);
		*topt++ = htonl(tsecr);
828 829
	}

830 831
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
832 833 834
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
		tcp_v6_md5_hash_hdr((__u8 *)topt, key,
835 836
				    &ipv6_hdr(skb)->saddr,
				    &ipv6_hdr(skb)->daddr, t1);
837 838 839
	}
#endif

840
	memset(&fl6, 0, sizeof(fl6));
A
Alexey Dobriyan 已提交
841 842
	fl6.daddr = ipv6_hdr(skb)->saddr;
	fl6.saddr = ipv6_hdr(skb)->daddr;
843
	fl6.flowlabel = label;
L
Linus Torvalds 已提交
844

845 846 847
	buff->ip_summed = CHECKSUM_PARTIAL;
	buff->csum = 0;

848
	__tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
L
Linus Torvalds 已提交
849

850
	fl6.flowi6_proto = IPPROTO_TCP;
851
	if (rt6_need_strict(&fl6.daddr) && !oif)
E
Eric Dumazet 已提交
852
		fl6.flowi6_oif = tcp_v6_iif(skb);
853 854 855 856 857 858
	else {
		if (!oif && netif_index_is_l3_master(net, skb->skb_iif))
			oif = skb->skb_iif;

		fl6.flowi6_oif = oif;
	}
859

860
	fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark);
861 862
	fl6.fl6_dport = t1->dest;
	fl6.fl6_sport = t1->source;
863
	fl6.flowi6_uid = sock_net_uid(net, sk && sk_fullsock(sk) ? sk : NULL);
864
	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
L
Linus Torvalds 已提交
865

866 867 868 869
	/* Pass a socket to ip6_dst_lookup either it is for RST
	 * Underlying function will use this to retrieve the network
	 * namespace
	 */
870
	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
871 872
	if (!IS_ERR(dst)) {
		skb_dst_set(buff, dst);
873
		ip6_xmit(ctl_sk, buff, &fl6, fl6.flowi6_mark, NULL, tclass);
874
		TCP_INC_STATS(net, TCP_MIB_OUTSEGS);
875
		if (rst)
876
			TCP_INC_STATS(net, TCP_MIB_OUTRSTS);
877
		return;
L
Linus Torvalds 已提交
878 879 880 881 882
	}

	kfree_skb(buff);
}

883
static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
884
{
885
	const struct tcphdr *th = tcp_hdr(skb);
886
	u32 seq = 0, ack_seq = 0;
887
	struct tcp_md5sig_key *key = NULL;
888 889 890 891 892 893 894
#ifdef CONFIG_TCP_MD5SIG
	const __u8 *hash_location = NULL;
	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
	unsigned char newhash[16];
	int genhash;
	struct sock *sk1 = NULL;
#endif
895
	int oif = 0;
L
Linus Torvalds 已提交
896

897
	if (th->rst)
L
Linus Torvalds 已提交
898 899
		return;

900 901 902 903
	/* 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 && !ipv6_unicast_destination(skb))
904
		return;
L
Linus Torvalds 已提交
905

906
#ifdef CONFIG_TCP_MD5SIG
907
	rcu_read_lock();
908
	hash_location = tcp_parse_md5sig_option(th);
909
	if (sk && sk_fullsock(sk)) {
910 911
		key = tcp_v6_md5_do_lookup(sk, &ipv6h->saddr);
	} else if (hash_location) {
912 913 914 915 916 917 918 919
		/*
		 * 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 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev),
920 921
					   &tcp_hashinfo, NULL, 0,
					   &ipv6h->saddr,
922
					   th->source, &ipv6h->daddr,
923 924
					   ntohs(th->source), tcp_v6_iif(skb),
					   tcp_v6_sdif(skb));
925
		if (!sk1)
926
			goto out;
927 928 929

		key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr);
		if (!key)
930
			goto out;
931

932
		genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, skb);
933
		if (genhash || memcmp(hash_location, newhash, 16) != 0)
934
			goto out;
935
	}
936 937
#endif

938 939 940 941 942
	if (th->ack)
		seq = ntohl(th->ack_seq);
	else
		ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len -
			  (th->doff << 2);
L
Linus Torvalds 已提交
943

944 945 946 947 948
	if (sk) {
		oif = sk->sk_bound_dev_if;
		trace_tcp_send_reset(sk, skb);
	}

949
	tcp_v6_send_response(sk, skb, seq, ack_seq, 0, 0, 0, oif, key, 1, 0, 0);
950 951

#ifdef CONFIG_TCP_MD5SIG
952 953
out:
	rcu_read_unlock();
954
#endif
955
}
L
Linus Torvalds 已提交
956

957
static void tcp_v6_send_ack(const struct sock *sk, struct sk_buff *skb, u32 seq,
958
			    u32 ack, u32 win, u32 tsval, u32 tsecr, int oif,
959
			    struct tcp_md5sig_key *key, u8 tclass,
960
			    __be32 label)
961
{
962 963
	tcp_v6_send_response(sk, skb, seq, ack, win, tsval, tsecr, oif, key, 0,
			     tclass, label);
L
Linus Torvalds 已提交
964 965 966 967
}

static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
968
	struct inet_timewait_sock *tw = inet_twsk(sk);
969
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
970

971
	tcp_v6_send_ack(sk, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
972
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
973
			tcp_time_stamp_raw() + tcptw->tw_ts_offset,
W
Wang Yufen 已提交
974
			tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw),
975
			tw->tw_tclass, cpu_to_be32(tw->tw_flowlabel));
L
Linus Torvalds 已提交
976

977
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
978 979
}

980
static void tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
981
				  struct request_sock *req)
L
Linus Torvalds 已提交
982
{
983 984 985
	/* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
	 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
	 */
986 987 988 989 990
	/* RFC 7323 2.3
	 * The window field (SEG.WND) of every outgoing segment, with the
	 * exception of <SYN> segments, MUST be right-shifted by
	 * Rcv.Wind.Shift bits:
	 */
991
	tcp_v6_send_ack(sk, skb, (sk->sk_state == TCP_LISTEN) ?
992
			tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt,
993 994
			tcp_rsk(req)->rcv_nxt,
			req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale,
995
			tcp_time_stamp_raw() + tcp_rsk(req)->ts_off,
996
			req->ts_recent, sk->sk_bound_dev_if,
997
			tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->saddr),
998
			0, 0);
L
Linus Torvalds 已提交
999 1000 1001
}


1002
static struct sock *tcp_v6_cookie_check(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
1003
{
1004
#ifdef CONFIG_SYN_COOKIES
1005
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1006

1007
	if (!th->syn)
1008
		sk = cookie_v6_check(sk, skb);
L
Linus Torvalds 已提交
1009 1010 1011 1012 1013 1014 1015 1016 1017 1018
#endif
	return sk;
}

static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
{
	if (skb->protocol == htons(ETH_P_IP))
		return tcp_v4_conn_request(sk, skb);

	if (!ipv6_unicast_destination(skb))
1019
		goto drop;
L
Linus Torvalds 已提交
1020

O
Octavian Purdila 已提交
1021 1022
	return tcp_conn_request(&tcp6_request_sock_ops,
				&tcp_request_sock_ipv6_ops, sk, skb);
L
Linus Torvalds 已提交
1023 1024

drop:
1025
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1026 1027 1028
	return 0; /* don't send reset */
}

1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
static void tcp_v6_restore_cb(struct sk_buff *skb)
{
	/* We need to move header back to the beginning if xfrm6_policy_check()
	 * and tcp_v6_fill_cb() are going to be called again.
	 * ip6_datagram_recv_specific_ctl() also expects IP6CB to be there.
	 */
	memmove(IP6CB(skb), &TCP_SKB_CB(skb)->header.h6,
		sizeof(struct inet6_skb_parm));
}

1039
static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
W
Weilong Chen 已提交
1040
					 struct request_sock *req,
1041 1042 1043
					 struct dst_entry *dst,
					 struct request_sock *req_unhash,
					 bool *own_req)
L
Linus Torvalds 已提交
1044
{
1045
	struct inet_request_sock *ireq;
1046 1047
	struct ipv6_pinfo *newnp;
	const struct ipv6_pinfo *np = inet6_sk(sk);
1048
	struct ipv6_txoptions *opt;
L
Linus Torvalds 已提交
1049 1050 1051 1052
	struct tcp6_sock *newtcp6sk;
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1053 1054 1055
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1056
	struct flowi6 fl6;
L
Linus Torvalds 已提交
1057 1058 1059 1060 1061 1062

	if (skb->protocol == htons(ETH_P_IP)) {
		/*
		 *	v6 mapped
		 */

1063 1064
		newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst,
					     req_unhash, own_req);
L
Linus Torvalds 已提交
1065

1066
		if (!newsk)
L
Linus Torvalds 已提交
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077
			return NULL;

		newtcp6sk = (struct tcp6_sock *)newsk;
		inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;

		newinet = inet_sk(newsk);
		newnp = inet6_sk(newsk);
		newtp = tcp_sk(newsk);

		memcpy(newnp, np, sizeof(struct ipv6_pinfo));

1078
		newnp->saddr = newsk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
1079

1080
		inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
1081
		newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1082 1083 1084 1085
#ifdef CONFIG_TCP_MD5SIG
		newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif

1086
		newnp->ipv6_mc_list = NULL;
1087 1088
		newnp->ipv6_ac_list = NULL;
		newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1089 1090
		newnp->pktoptions  = NULL;
		newnp->opt	   = NULL;
E
Eric Dumazet 已提交
1091
		newnp->mcast_oif   = tcp_v6_iif(skb);
1092
		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
1093
		newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1094 1095
		if (np->repflow)
			newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1096

1097 1098 1099 1100
		/*
		 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
		 * here, tcp_create_openreq_child now does this for us, see the comment in
		 * that function for the gory details. -acme
L
Linus Torvalds 已提交
1101 1102 1103
		 */

		/* It is tricky place. Until this moment IPv4 tcp
1104
		   worked with IPv6 icsk.icsk_af_ops.
L
Linus Torvalds 已提交
1105 1106
		   Sync it now.
		 */
1107
		tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
1108 1109 1110 1111

		return newsk;
	}

1112
	ireq = inet_rsk(req);
L
Linus Torvalds 已提交
1113 1114 1115 1116

	if (sk_acceptq_is_full(sk))
		goto out_overflow;

1117
	if (!dst) {
1118
		dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_TCP);
1119
		if (!dst)
L
Linus Torvalds 已提交
1120
			goto out;
1121
	}
L
Linus Torvalds 已提交
1122 1123

	newsk = tcp_create_openreq_child(sk, req, skb);
1124
	if (!newsk)
1125
		goto out_nonewsk;
L
Linus Torvalds 已提交
1126

1127 1128 1129 1130 1131
	/*
	 * No need to charge this sock to the relevant IPv6 refcnt debug socks
	 * count here, tcp_create_openreq_child now does this for us, see the
	 * comment in that function for the gory details. -acme
	 */
L
Linus Torvalds 已提交
1132

1133
	newsk->sk_gso_type = SKB_GSO_TCPV6;
E
Eric Dumazet 已提交
1134
	ip6_dst_store(newsk, dst, NULL, NULL);
1135
	inet6_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1136 1137 1138 1139 1140 1141 1142 1143 1144 1145

	newtcp6sk = (struct tcp6_sock *)newsk;
	inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;

	newtp = tcp_sk(newsk);
	newinet = inet_sk(newsk);
	newnp = inet6_sk(newsk);

	memcpy(newnp, np, sizeof(struct ipv6_pinfo));

1146 1147 1148 1149
	newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr;
	newnp->saddr = ireq->ir_v6_loc_addr;
	newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr;
	newsk->sk_bound_dev_if = ireq->ir_iif;
L
Linus Torvalds 已提交
1150

1151
	/* Now IPv6 options...
L
Linus Torvalds 已提交
1152 1153 1154

	   First: no IPv4 options.
	 */
1155
	newinet->inet_opt = NULL;
1156
	newnp->ipv6_mc_list = NULL;
1157
	newnp->ipv6_ac_list = NULL;
1158
	newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1159 1160 1161 1162 1163 1164

	/* Clone RX bits */
	newnp->rxopt.all = np->rxopt.all;

	newnp->pktoptions = NULL;
	newnp->opt	  = NULL;
E
Eric Dumazet 已提交
1165
	newnp->mcast_oif  = tcp_v6_iif(skb);
1166
	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1167
	newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1168 1169
	if (np->repflow)
		newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1170 1171 1172 1173 1174 1175 1176

	/* Clone native IPv6 options from listening socket (if any)

	   Yes, keeping reference count would be much more clever,
	   but we make one more one thing there: reattach optmem
	   to newsk.
	 */
1177 1178 1179
	opt = ireq->ipv6_opt;
	if (!opt)
		opt = rcu_dereference(np->opt);
1180 1181 1182 1183
	if (opt) {
		opt = ipv6_dup_options(newsk, opt);
		RCU_INIT_POINTER(newnp->opt, opt);
	}
1184
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
1185 1186 1187
	if (opt)
		inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
						    opt->opt_flen;
L
Linus Torvalds 已提交
1188

1189 1190
	tcp_ca_openreq_child(newsk, dst);

L
Linus Torvalds 已提交
1191
	tcp_sync_mss(newsk, dst_mtu(dst));
E
Eric Dumazet 已提交
1192
	newtp->advmss = tcp_mss_clamp(tcp_sk(sk), dst_metric_advmss(dst));
1193

L
Linus Torvalds 已提交
1194 1195
	tcp_initialize_rcv_mss(newsk);

E
Eric Dumazet 已提交
1196 1197
	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
1198

1199 1200
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
W
Wang Yufen 已提交
1201
	key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr);
1202
	if (key) {
1203 1204 1205 1206 1207
		/* 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.
		 */
1208
		tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
1209
			       AF_INET6, 128, key->key, key->keylen,
1210
			       sk_gfp_mask(sk, GFP_ATOMIC));
1211 1212 1213
	}
#endif

1214
	if (__inet_inherit_port(sk, newsk) < 0) {
1215 1216
		inet_csk_prepare_forced_close(newsk);
		tcp_done(newsk);
1217 1218
		goto out;
	}
1219
	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
E
Eric Dumazet 已提交
1220
	if (*own_req) {
1221
		tcp_move_syn(newtp, req);
E
Eric Dumazet 已提交
1222 1223 1224 1225

		/* Clone pktoptions received with SYN, if we own the req */
		if (ireq->pktopts) {
			newnp->pktoptions = skb_clone(ireq->pktopts,
1226
						      sk_gfp_mask(sk, GFP_ATOMIC));
E
Eric Dumazet 已提交
1227 1228
			consume_skb(ireq->pktopts);
			ireq->pktopts = NULL;
1229 1230
			if (newnp->pktoptions) {
				tcp_v6_restore_cb(newnp->pktoptions);
E
Eric Dumazet 已提交
1231
				skb_set_owner_r(newnp->pktoptions, newsk);
1232
			}
E
Eric Dumazet 已提交
1233
		}
E
Eric Dumazet 已提交
1234
	}
L
Linus Torvalds 已提交
1235 1236 1237 1238

	return newsk;

out_overflow:
1239
	__NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1240
out_nonewsk:
L
Linus Torvalds 已提交
1241
	dst_release(dst);
1242
out:
1243
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1244 1245 1246 1247
	return NULL;
}

/* The socket must have it's spinlock held when we get
1248
 * here, unless it is a TCP_LISTEN socket.
L
Linus Torvalds 已提交
1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287
 *
 * 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.
 */
static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
{
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp;
	struct sk_buff *opt_skb = NULL;

	/* Imagine: socket is IPv6. IPv4 packet arrives,
	   goes to IPv4 receive handler and backlogged.
	   From backlog it always goes here. Kerboom...
	   Fortunately, tcp_rcv_established and rcv_established
	   handle them correctly, but it is not case with
	   tcp_v6_hnd_req and tcp_v6_send_reset().   --ANK
	 */

	if (skb->protocol == htons(ETH_P_IP))
		return tcp_v4_do_rcv(sk, skb);

	/*
	 *	socket locking is here for SMP purposes as backlog rcv
	 *	is currently called with bh processing disabled.
	 */

	/* Do Stevens' IPV6_PKTOPTIONS.

	   Yes, guys, it is the only place in our code, where we
	   may make it not affecting IPv4.
	   The rest of code is protocol independent,
	   and I do not like idea to uglify IPv4.

	   Actually, all the idea behind IPV6_PKTOPTIONS
	   looks not very well thought. For now we latch
	   options, received in the last packet, enqueued
	   by tcp. Feel free to propose better solution.
1288
					       --ANK (980728)
L
Linus Torvalds 已提交
1289 1290
	 */
	if (np->rxopt.all)
1291
		opt_skb = skb_clone(skb, sk_gfp_mask(sk, GFP_ATOMIC));
L
Linus Torvalds 已提交
1292 1293

	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
E
Eric Dumazet 已提交
1294 1295
		struct dst_entry *dst = sk->sk_rx_dst;

1296
		sock_rps_save_rxhash(sk, skb);
1297
		sk_mark_napi_id(sk, skb);
E
Eric Dumazet 已提交
1298 1299 1300 1301 1302 1303 1304 1305
		if (dst) {
			if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif ||
			    dst->ops->check(dst, np->rx_dst_cookie) == NULL) {
				dst_release(dst);
				sk->sk_rx_dst = NULL;
			}
		}

1306
		tcp_rcv_established(sk, skb, tcp_hdr(skb));
L
Linus Torvalds 已提交
1307 1308 1309 1310 1311
		if (opt_skb)
			goto ipv6_pktoptions;
		return 0;
	}

E
Eric Dumazet 已提交
1312
	if (tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1313 1314
		goto csum_err;

1315
	if (sk->sk_state == TCP_LISTEN) {
1316 1317
		struct sock *nsk = tcp_v6_cookie_check(sk, skb);

L
Linus Torvalds 已提交
1318 1319 1320
		if (!nsk)
			goto discard;

W
Weilong Chen 已提交
1321
		if (nsk != sk) {
L
Linus Torvalds 已提交
1322 1323 1324 1325 1326 1327
			if (tcp_child_process(sk, nsk, skb))
				goto reset;
			if (opt_skb)
				__kfree_skb(opt_skb);
			return 0;
		}
1328
	} else
1329
		sock_rps_save_rxhash(sk, skb);
L
Linus Torvalds 已提交
1330

1331
	if (tcp_rcv_state_process(sk, skb))
L
Linus Torvalds 已提交
1332 1333 1334 1335 1336 1337
		goto reset;
	if (opt_skb)
		goto ipv6_pktoptions;
	return 0;

reset:
1338
	tcp_v6_send_reset(sk, skb);
L
Linus Torvalds 已提交
1339 1340 1341 1342 1343 1344
discard:
	if (opt_skb)
		__kfree_skb(opt_skb);
	kfree_skb(skb);
	return 0;
csum_err:
1345 1346
	TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS);
	TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360
	goto discard;


ipv6_pktoptions:
	/* Do you ask, what is it?

	   1. skb was enqueued by tcp.
	   2. skb is added to tail of read queue, rather than out of order.
	   3. socket is not in passive state.
	   4. Finally, it really contains options, which user wants to receive.
	 */
	tp = tcp_sk(sk);
	if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
	    !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
1361
		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
E
Eric Dumazet 已提交
1362
			np->mcast_oif = tcp_v6_iif(opt_skb);
1363
		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1364
			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1365
		if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
1366
			np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
1367 1368
		if (np->repflow)
			np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
1369
		if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->header.h6)) {
L
Linus Torvalds 已提交
1370
			skb_set_owner_r(opt_skb, sk);
1371
			tcp_v6_restore_cb(opt_skb);
L
Linus Torvalds 已提交
1372 1373 1374 1375 1376 1377 1378
			opt_skb = xchg(&np->pktoptions, opt_skb);
		} else {
			__kfree_skb(opt_skb);
			opt_skb = xchg(&np->pktoptions, NULL);
		}
	}

1379
	kfree_skb(opt_skb);
L
Linus Torvalds 已提交
1380 1381 1382
	return 0;
}

1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402
static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
			   const struct tcphdr *th)
{
	/* This is tricky: we move IP6CB at its correct location into
	 * TCP_SKB_CB(). It must be done after xfrm6_policy_check(), because
	 * _decode_session6() uses IP6CB().
	 * barrier() makes sure compiler won't play aliasing games.
	 */
	memmove(&TCP_SKB_CB(skb)->header.h6, IP6CB(skb),
		sizeof(struct inet6_skb_parm));
	barrier();

	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)->tcp_flags = tcp_flag_byte(th);
	TCP_SKB_CB(skb)->tcp_tw_isn = 0;
	TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
	TCP_SKB_CB(skb)->sacked = 0;
1403 1404
	TCP_SKB_CB(skb)->has_rxtstamp =
			skb->tstamp || skb_hwtstamps(skb)->hwtstamp;
1405 1406
}

1407
static int tcp_v6_rcv(struct sk_buff *skb)
L
Linus Torvalds 已提交
1408
{
1409
	int sdif = inet6_sdif(skb);
1410
	const struct tcphdr *th;
1411
	const struct ipv6hdr *hdr;
1412
	bool refcounted;
L
Linus Torvalds 已提交
1413 1414
	struct sock *sk;
	int ret;
1415
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
1416 1417 1418 1419 1420 1421 1422

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

	/*
	 *	Count it even if it's bad.
	 */
E
Eric Dumazet 已提交
1423
	__TCP_INC_STATS(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1424 1425 1426 1427

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

1428
	th = (const struct tcphdr *)skb->data;
L
Linus Torvalds 已提交
1429

1430
	if (unlikely(th->doff < sizeof(struct tcphdr)/4))
L
Linus Torvalds 已提交
1431 1432 1433 1434
		goto bad_packet;
	if (!pskb_may_pull(skb, th->doff*4))
		goto discard_it;

1435
	if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo))
1436
		goto csum_error;
L
Linus Torvalds 已提交
1437

1438
	th = (const struct tcphdr *)skb->data;
1439
	hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
1440

1441
lookup:
1442
	sk = __inet6_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th),
1443
				th->source, th->dest, inet6_iif(skb), sdif,
1444
				&refcounted);
L
Linus Torvalds 已提交
1445 1446 1447 1448 1449 1450 1451
	if (!sk)
		goto no_tcp_socket;

process:
	if (sk->sk_state == TCP_TIME_WAIT)
		goto do_time_wait;

1452 1453
	if (sk->sk_state == TCP_NEW_SYN_RECV) {
		struct request_sock *req = inet_reqsk(sk);
1454
		struct sock *nsk;
1455 1456 1457

		sk = req->rsk_listener;
		if (tcp_v6_inbound_md5_hash(sk, skb)) {
1458
			sk_drops_add(sk, skb);
1459 1460 1461
			reqsk_put(req);
			goto discard_it;
		}
1462
		if (unlikely(sk->sk_state != TCP_LISTEN)) {
1463
			inet_csk_reqsk_queue_drop_and_put(sk, req);
1464 1465
			goto lookup;
		}
1466
		sock_hold(sk);
1467
		refcounted = true;
E
Eric Dumazet 已提交
1468
		nsk = NULL;
1469 1470 1471 1472
		if (!tcp_filter(sk, skb)) {
			th = (const struct tcphdr *)skb->data;
			hdr = ipv6_hdr(skb);
			tcp_v6_fill_cb(skb, hdr, th);
E
Eric Dumazet 已提交
1473
			nsk = tcp_check_req(sk, skb, req, false);
1474
		}
1475 1476
		if (!nsk) {
			reqsk_put(req);
1477
			goto discard_and_relse;
1478 1479 1480 1481 1482 1483
		}
		if (nsk == sk) {
			reqsk_put(req);
			tcp_v6_restore_cb(skb);
		} else if (tcp_child_process(sk, nsk, skb)) {
			tcp_v6_send_reset(nsk, skb);
1484
			goto discard_and_relse;
1485
		} else {
1486
			sock_put(sk);
1487 1488 1489
			return 0;
		}
	}
1490
	if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) {
1491
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
1492 1493 1494
		goto discard_and_relse;
	}

L
Linus Torvalds 已提交
1495 1496 1497
	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;

1498 1499 1500
	if (tcp_v6_inbound_md5_hash(sk, skb))
		goto discard_and_relse;

1501
	if (tcp_filter(sk, skb))
L
Linus Torvalds 已提交
1502
		goto discard_and_relse;
1503 1504
	th = (const struct tcphdr *)skb->data;
	hdr = ipv6_hdr(skb);
1505
	tcp_v6_fill_cb(skb, hdr, th);
L
Linus Torvalds 已提交
1506 1507 1508

	skb->dev = NULL;

1509 1510 1511 1512 1513 1514 1515
	if (sk->sk_state == TCP_LISTEN) {
		ret = tcp_v6_do_rcv(sk, skb);
		goto put_and_return;
	}

	sk_incoming_cpu_update(sk);

1516
	bh_lock_sock_nested(sk);
1517
	tcp_segs_in(tcp_sk(sk), skb);
L
Linus Torvalds 已提交
1518 1519
	ret = 0;
	if (!sock_owned_by_user(sk)) {
F
Florian Westphal 已提交
1520
		ret = tcp_v6_do_rcv(sk, skb);
E
Eric Dumazet 已提交
1521
	} else if (tcp_add_backlog(sk, skb)) {
Z
Zhu Yi 已提交
1522 1523
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1524 1525
	bh_unlock_sock(sk);

1526
put_and_return:
1527 1528
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1529 1530 1531 1532 1533 1534
	return ret ? -1 : 0;

no_tcp_socket:
	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
		goto discard_it;

1535 1536
	tcp_v6_fill_cb(skb, hdr, th);

E
Eric Dumazet 已提交
1537
	if (tcp_checksum_complete(skb)) {
1538
csum_error:
E
Eric Dumazet 已提交
1539
		__TCP_INC_STATS(net, TCP_MIB_CSUMERRORS);
L
Linus Torvalds 已提交
1540
bad_packet:
E
Eric Dumazet 已提交
1541
		__TCP_INC_STATS(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1542
	} else {
1543
		tcp_v6_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1544 1545 1546 1547 1548 1549 1550
	}

discard_it:
	kfree_skb(skb);
	return 0;

discard_and_relse:
1551
	sk_drops_add(sk, skb);
1552 1553
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1554 1555 1556 1557
	goto discard_it;

do_time_wait:
	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1558
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1559 1560 1561
		goto discard_it;
	}

1562 1563
	tcp_v6_fill_cb(skb, hdr, th);

1564 1565 1566
	if (tcp_checksum_complete(skb)) {
		inet_twsk_put(inet_twsk(sk));
		goto csum_error;
L
Linus Torvalds 已提交
1567 1568
	}

1569
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1570 1571 1572 1573
	case TCP_TW_SYN:
	{
		struct sock *sk2;

1574
		sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1575
					    skb, __tcp_hdrlen(th),
1576
					    &ipv6_hdr(skb)->saddr, th->source,
1577
					    &ipv6_hdr(skb)->daddr,
1578 1579
					    ntohs(th->dest), tcp_v6_iif(skb),
					    sdif);
1580
		if (sk2) {
1581
			struct inet_timewait_sock *tw = inet_twsk(sk);
1582
			inet_twsk_deschedule_put(tw);
L
Linus Torvalds 已提交
1583
			sk = sk2;
1584
			tcp_v6_restore_cb(skb);
1585
			refcounted = false;
L
Linus Torvalds 已提交
1586 1587 1588
			goto process;
		}
	}
1589 1590
		/* to ACK */
		/* fall through */
L
Linus Torvalds 已提交
1591 1592 1593 1594
	case TCP_TW_ACK:
		tcp_v6_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
1595 1596 1597
		tcp_v6_send_reset(sk, skb);
		inet_twsk_deschedule_put(inet_twsk(sk));
		goto discard_it;
W
Wang Yufen 已提交
1598 1599
	case TCP_TW_SUCCESS:
		;
L
Linus Torvalds 已提交
1600 1601 1602 1603
	}
	goto discard_it;
}

E
Eric Dumazet 已提交
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
static void tcp_v6_early_demux(struct sk_buff *skb)
{
	const struct ipv6hdr *hdr;
	const struct tcphdr *th;
	struct sock *sk;

	if (skb->pkt_type != PACKET_HOST)
		return;

	if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr)))
		return;

	hdr = ipv6_hdr(skb);
	th = tcp_hdr(skb);

	if (th->doff < sizeof(struct tcphdr) / 4)
		return;

E
Eric Dumazet 已提交
1622
	/* Note : We use inet6_iif() here, not tcp_v6_iif() */
E
Eric Dumazet 已提交
1623 1624 1625
	sk = __inet6_lookup_established(dev_net(skb->dev), &tcp_hashinfo,
					&hdr->saddr, th->source,
					&hdr->daddr, ntohs(th->dest),
1626
					inet6_iif(skb), inet6_sdif(skb));
E
Eric Dumazet 已提交
1627 1628 1629
	if (sk) {
		skb->sk = sk;
		skb->destructor = sock_edemux;
1630
		if (sk_fullsock(sk)) {
1631
			struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst);
1632

E
Eric Dumazet 已提交
1633
			if (dst)
E
Eric Dumazet 已提交
1634
				dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie);
E
Eric Dumazet 已提交
1635
			if (dst &&
1636
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
E
Eric Dumazet 已提交
1637 1638 1639 1640 1641
				skb_dst_set_noref(skb, dst);
		}
	}
}

1642 1643 1644
static struct timewait_sock_ops tcp6_timewait_sock_ops = {
	.twsk_obj_size	= sizeof(struct tcp6_timewait_sock),
	.twsk_unique	= tcp_twsk_unique,
W
Wang Yufen 已提交
1645
	.twsk_destructor = tcp_twsk_destructor,
1646 1647
};

1648
static const struct inet_connection_sock_af_ops ipv6_specific = {
1649 1650 1651
	.queue_xmit	   = inet6_csk_xmit,
	.send_check	   = tcp_v6_send_check,
	.rebuild_header	   = inet6_sk_rebuild_header,
E
Eric Dumazet 已提交
1652
	.sk_rx_dst_set	   = inet6_sk_rx_dst_set,
1653 1654 1655
	.conn_request	   = tcp_v6_conn_request,
	.syn_recv_sock	   = tcp_v6_syn_recv_sock,
	.net_header_len	   = sizeof(struct ipv6hdr),
1656
	.net_frag_header_len = sizeof(struct frag_hdr),
1657 1658 1659 1660
	.setsockopt	   = ipv6_setsockopt,
	.getsockopt	   = ipv6_getsockopt,
	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in6),
1661
#ifdef CONFIG_COMPAT
1662 1663
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1664
#endif
1665
	.mtu_reduced	   = tcp_v6_mtu_reduced,
L
Linus Torvalds 已提交
1666 1667
};

1668
#ifdef CONFIG_TCP_MD5SIG
1669
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1670
	.md5_lookup	=	tcp_v6_md5_lookup,
1671
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
1672 1673
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1674
#endif
1675

L
Linus Torvalds 已提交
1676 1677 1678
/*
 *	TCP over IPv4 via INET6 API
 */
1679
static const struct inet_connection_sock_af_ops ipv6_mapped = {
1680 1681 1682
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
1683
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1684 1685 1686 1687 1688 1689 1690
	.conn_request	   = tcp_v6_conn_request,
	.syn_recv_sock	   = tcp_v6_syn_recv_sock,
	.net_header_len	   = sizeof(struct iphdr),
	.setsockopt	   = ipv6_setsockopt,
	.getsockopt	   = ipv6_getsockopt,
	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in6),
1691
#ifdef CONFIG_COMPAT
1692 1693
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1694
#endif
1695
	.mtu_reduced	   = tcp_v4_mtu_reduced,
L
Linus Torvalds 已提交
1696 1697
};

1698
#ifdef CONFIG_TCP_MD5SIG
1699
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1700
	.md5_lookup	=	tcp_v4_md5_lookup,
1701
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1702 1703
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1704
#endif
1705

L
Linus Torvalds 已提交
1706 1707 1708 1709 1710
/* NOTE: A lot of things set to zero explicitly by call to
 *       sk_alloc() so need not be done here.
 */
static int tcp_v6_init_sock(struct sock *sk)
{
1711
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1712

1713
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1714

1715
	icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
1716

1717
#ifdef CONFIG_TCP_MD5SIG
1718
	tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
1719 1720
#endif

L
Linus Torvalds 已提交
1721 1722 1723
	return 0;
}

1724
static void tcp_v6_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1725 1726
{
	tcp_v4_destroy_sock(sk);
1727
	inet6_destroy_sock(sk);
L
Linus Torvalds 已提交
1728 1729
}

1730
#ifdef CONFIG_PROC_FS
L
Linus Torvalds 已提交
1731
/* Proc filesystem TCPv6 sock list dumping. */
1732
static void get_openreq6(struct seq_file *seq,
E
Eric Dumazet 已提交
1733
			 const struct request_sock *req, int i)
L
Linus Torvalds 已提交
1734
{
1735
	long ttd = req->rsk_timer.expires - jiffies;
1736 1737
	const struct in6_addr *src = &inet_rsk(req)->ir_v6_loc_addr;
	const struct in6_addr *dest = &inet_rsk(req)->ir_v6_rmt_addr;
L
Linus Torvalds 已提交
1738 1739 1740 1741 1742 1743

	if (ttd < 0)
		ttd = 0;

	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1744
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1745 1746 1747
		   i,
		   src->s6_addr32[0], src->s6_addr32[1],
		   src->s6_addr32[2], src->s6_addr32[3],
E
Eric Dumazet 已提交
1748
		   inet_rsk(req)->ir_num,
L
Linus Torvalds 已提交
1749 1750
		   dest->s6_addr32[0], dest->s6_addr32[1],
		   dest->s6_addr32[2], dest->s6_addr32[3],
1751
		   ntohs(inet_rsk(req)->ir_rmt_port),
L
Linus Torvalds 已提交
1752
		   TCP_SYN_RECV,
W
Weilong Chen 已提交
1753
		   0, 0, /* could print option size, but that is af dependent. */
1754 1755
		   1,   /* timers active (only the expire timer) */
		   jiffies_to_clock_t(ttd),
1756
		   req->num_timeout,
E
Eric Dumazet 已提交
1757 1758
		   from_kuid_munged(seq_user_ns(seq),
				    sock_i_uid(req->rsk_listener)),
1759
		   0,  /* non standard timer */
L
Linus Torvalds 已提交
1760 1761 1762 1763 1764 1765
		   0, /* open_requests have no inode */
		   0, req);
}

static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
{
1766
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1767 1768 1769
	__u16 destp, srcp;
	int timer_active;
	unsigned long timer_expires;
1770 1771
	const struct inet_sock *inet = inet_sk(sp);
	const struct tcp_sock *tp = tcp_sk(sp);
1772
	const struct inet_connection_sock *icsk = inet_csk(sp);
1773
	const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq;
1774 1775
	int rx_queue;
	int state;
L
Linus Torvalds 已提交
1776

1777 1778
	dest  = &sp->sk_v6_daddr;
	src   = &sp->sk_v6_rcv_saddr;
E
Eric Dumazet 已提交
1779 1780
	destp = ntohs(inet->inet_dport);
	srcp  = ntohs(inet->inet_sport);
1781

1782
	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
1783
	    icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
1784
	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
L
Linus Torvalds 已提交
1785
		timer_active	= 1;
1786 1787
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
1788
		timer_active	= 4;
1789
		timer_expires	= icsk->icsk_timeout;
L
Linus Torvalds 已提交
1790 1791 1792 1793 1794 1795 1796 1797
	} else if (timer_pending(&sp->sk_timer)) {
		timer_active	= 2;
		timer_expires	= sp->sk_timer.expires;
	} else {
		timer_active	= 0;
		timer_expires = jiffies;
	}

1798 1799 1800 1801 1802 1803 1804 1805 1806
	state = sk_state_load(sp);
	if (state == TCP_LISTEN)
		rx_queue = sp->sk_ack_backlog;
	else
		/* Because we don't lock the socket,
		 * we might find a transient negative value.
		 */
		rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0);

L
Linus Torvalds 已提交
1807 1808
	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1809
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
L
Linus Torvalds 已提交
1810 1811 1812 1813 1814
		   i,
		   src->s6_addr32[0], src->s6_addr32[1],
		   src->s6_addr32[2], src->s6_addr32[3], srcp,
		   dest->s6_addr32[0], dest->s6_addr32[1],
		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1815 1816 1817
		   state,
		   tp->write_seq - tp->snd_una,
		   rx_queue,
L
Linus Torvalds 已提交
1818
		   timer_active,
1819
		   jiffies_delta_to_clock_t(timer_expires - jiffies),
1820
		   icsk->icsk_retransmits,
1821
		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1822
		   icsk->icsk_probes_out,
L
Linus Torvalds 已提交
1823
		   sock_i_ino(sp),
1824
		   refcount_read(&sp->sk_refcnt), sp,
1825 1826
		   jiffies_to_clock_t(icsk->icsk_rto),
		   jiffies_to_clock_t(icsk->icsk_ack.ato),
W
Weilong Chen 已提交
1827
		   (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
I
Ilpo Järvinen 已提交
1828
		   tp->snd_cwnd,
1829
		   state == TCP_LISTEN ?
1830
			fastopenq->max_qlen :
1831
			(tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)
L
Linus Torvalds 已提交
1832 1833 1834
		   );
}

1835
static void get_timewait6_sock(struct seq_file *seq,
1836
			       struct inet_timewait_sock *tw, int i)
L
Linus Torvalds 已提交
1837
{
1838
	long delta = tw->tw_timer.expires - jiffies;
1839
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1840 1841
	__u16 destp, srcp;

1842 1843
	dest = &tw->tw_v6_daddr;
	src  = &tw->tw_v6_rcv_saddr;
L
Linus Torvalds 已提交
1844 1845 1846 1847 1848
	destp = ntohs(tw->tw_dport);
	srcp  = ntohs(tw->tw_sport);

	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
D
Dan Rosenberg 已提交
1849
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1850 1851 1852 1853 1854 1855
		   i,
		   src->s6_addr32[0], src->s6_addr32[1],
		   src->s6_addr32[2], src->s6_addr32[3], srcp,
		   dest->s6_addr32[0], dest->s6_addr32[1],
		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
		   tw->tw_substate, 0, 0,
1856
		   3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
1857
		   refcount_read(&tw->tw_refcnt), tw);
L
Linus Torvalds 已提交
1858 1859 1860 1861 1862
}

static int tcp6_seq_show(struct seq_file *seq, void *v)
{
	struct tcp_iter_state *st;
E
Eric Dumazet 已提交
1863
	struct sock *sk = v;
L
Linus Torvalds 已提交
1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875

	if (v == SEQ_START_TOKEN) {
		seq_puts(seq,
			 "  sl  "
			 "local_address                         "
			 "remote_address                        "
			 "st tx_queue rx_queue tr tm->when retrnsmt"
			 "   uid  timeout inode\n");
		goto out;
	}
	st = seq->private;

1876 1877 1878
	if (sk->sk_state == TCP_TIME_WAIT)
		get_timewait6_sock(seq, v, st->num);
	else if (sk->sk_state == TCP_NEW_SYN_RECV)
E
Eric Dumazet 已提交
1879
		get_openreq6(seq, v, st->num);
1880 1881
	else
		get_tcp6_sock(seq, v, st->num);
L
Linus Torvalds 已提交
1882 1883 1884 1885
out:
	return 0;
}

1886 1887 1888 1889 1890 1891 1892 1893
static const struct file_operations tcp6_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
1894 1895 1896
static struct tcp_seq_afinfo tcp6_seq_afinfo = {
	.name		= "tcp6",
	.family		= AF_INET6,
1897
	.seq_fops	= &tcp6_afinfo_seq_fops,
1898 1899 1900
	.seq_ops	= {
		.show		= tcp6_seq_show,
	},
L
Linus Torvalds 已提交
1901 1902
};

1903
int __net_init tcp6_proc_init(struct net *net)
L
Linus Torvalds 已提交
1904
{
1905
	return tcp_proc_register(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1906 1907
}

1908
void tcp6_proc_exit(struct net *net)
L
Linus Torvalds 已提交
1909
{
1910
	tcp_proc_unregister(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1911 1912 1913 1914 1915 1916 1917 1918 1919
}
#endif

struct proto tcpv6_prot = {
	.name			= "TCPv6",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v6_connect,
	.disconnect		= tcp_disconnect,
1920
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
1921 1922 1923 1924 1925 1926
	.ioctl			= tcp_ioctl,
	.init			= tcp_v6_init_sock,
	.destroy		= tcp_v6_destroy_sock,
	.shutdown		= tcp_shutdown,
	.setsockopt		= tcp_setsockopt,
	.getsockopt		= tcp_getsockopt,
1927
	.keepalive		= tcp_set_keepalive,
L
Linus Torvalds 已提交
1928
	.recvmsg		= tcp_recvmsg,
1929 1930
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
1931
	.backlog_rcv		= tcp_v6_do_rcv,
E
Eric Dumazet 已提交
1932
	.release_cb		= tcp_release_cb,
1933
	.hash			= inet6_hash,
1934 1935
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
1936
	.enter_memory_pressure	= tcp_enter_memory_pressure,
1937
	.leave_memory_pressure	= tcp_leave_memory_pressure,
1938
	.stream_memory_free	= tcp_stream_memory_free,
L
Linus Torvalds 已提交
1939 1940 1941
	.sockets_allocated	= &tcp_sockets_allocated,
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
1942
	.orphan_count		= &tcp_orphan_count,
1943
	.sysctl_mem		= sysctl_tcp_mem,
1944 1945
	.sysctl_wmem_offset	= offsetof(struct net, ipv4.sysctl_tcp_wmem),
	.sysctl_rmem_offset	= offsetof(struct net, ipv4.sysctl_tcp_rmem),
L
Linus Torvalds 已提交
1946 1947
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp6_sock),
1948
	.slab_flags		= SLAB_TYPESAFE_BY_RCU,
1949
	.twsk_prot		= &tcp6_timewait_sock_ops,
1950
	.rsk_prot		= &tcp6_request_sock_ops,
1951
	.h.hashinfo		= &tcp_hashinfo,
1952
	.no_autobind		= true,
1953 1954 1955
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
G
Glauber Costa 已提交
1956
#endif
1957
	.diag_destroy		= tcp_abort,
L
Linus Torvalds 已提交
1958 1959
};

1960 1961 1962
/* thinking of making this const? Don't.
 * early_demux can change based on sysctl.
 */
1963
static struct inet6_protocol tcpv6_protocol = {
E
Eric Dumazet 已提交
1964
	.early_demux	=	tcp_v6_early_demux,
1965
	.early_demux_handler =  tcp_v6_early_demux,
L
Linus Torvalds 已提交
1966 1967 1968 1969 1970 1971 1972 1973 1974 1975
	.handler	=	tcp_v6_rcv,
	.err_handler	=	tcp_v6_err,
	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
};

static struct inet_protosw tcpv6_protosw = {
	.type		=	SOCK_STREAM,
	.protocol	=	IPPROTO_TCP,
	.prot		=	&tcpv6_prot,
	.ops		=	&inet6_stream_ops,
1976 1977
	.flags		=	INET_PROTOSW_PERMANENT |
				INET_PROTOSW_ICSK,
L
Linus Torvalds 已提交
1978 1979
};

1980
static int __net_init tcpv6_net_init(struct net *net)
1981
{
1982 1983
	return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
				    SOCK_RAW, IPPROTO_TCP, net);
1984 1985
}

1986
static void __net_exit tcpv6_net_exit(struct net *net)
1987
{
1988
	inet_ctl_sock_destroy(net->ipv6.tcp_sk);
E
Eric W. Biederman 已提交
1989 1990
}

1991
static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
E
Eric W. Biederman 已提交
1992
{
1993
	inet_twsk_purge(&tcp_hashinfo, AF_INET6);
1994 1995 1996
}

static struct pernet_operations tcpv6_net_ops = {
E
Eric W. Biederman 已提交
1997 1998 1999
	.init	    = tcpv6_net_init,
	.exit	    = tcpv6_net_exit,
	.exit_batch = tcpv6_net_exit_batch,
2000 2001
};

2002
int __init tcpv6_init(void)
L
Linus Torvalds 已提交
2003
{
2004 2005
	int ret;

2006 2007
	ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
	if (ret)
2008
		goto out;
2009

L
Linus Torvalds 已提交
2010
	/* register inet6 protocol */
2011 2012 2013 2014
	ret = inet6_register_protosw(&tcpv6_protosw);
	if (ret)
		goto out_tcpv6_protocol;

2015
	ret = register_pernet_subsys(&tcpv6_net_ops);
2016 2017 2018 2019
	if (ret)
		goto out_tcpv6_protosw;
out:
	return ret;
2020

2021 2022
out_tcpv6_protosw:
	inet6_unregister_protosw(&tcpv6_protosw);
2023 2024
out_tcpv6_protocol:
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2025 2026 2027
	goto out;
}

2028
void tcpv6_exit(void)
2029
{
2030
	unregister_pernet_subsys(&tcpv6_net_ops);
2031 2032
	inet6_unregister_protosw(&tcpv6_protosw);
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
L
Linus Torvalds 已提交
2033
}