tcp_ipv6.c 51.0 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
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,
74
				      struct request_sock *req);
L
Linus Torvalds 已提交
75 76 77

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

78 79
static const struct inet_connection_sock_af_ops ipv6_mapped;
static const struct inet_connection_sock_af_ops ipv6_specific;
80
#ifdef CONFIG_TCP_MD5SIG
81 82
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
83
#else
84
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk,
85
						   const struct in6_addr *addr)
86 87 88
{
	return NULL;
}
89
#endif
L
Linus Torvalds 已提交
90

91 92 93 94
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
95
	if (dst && dst_hold_safe(dst)) {
96 97 98 99
		const struct rt6_info *rt = (const struct rt6_info *)dst;

		sk->sk_rx_dst = dst;
		inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
100
		inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt);
101
	}
102 103
}

104
static u32 tcp_v6_init_seq(const struct sk_buff *skb)
L
Linus Torvalds 已提交
105
{
106 107 108 109 110 111
	return secure_tcpv6_seq(ipv6_hdr(skb)->daddr.s6_addr32,
				ipv6_hdr(skb)->saddr.s6_addr32,
				tcp_hdr(skb)->dest,
				tcp_hdr(skb)->source);
}

112
static u32 tcp_v6_init_ts_off(const struct net *net, const struct sk_buff *skb)
113
{
114
	return secure_tcpv6_ts_off(net, ipv6_hdr(skb)->daddr.s6_addr32,
115
				   ipv6_hdr(skb)->saddr.s6_addr32);
L
Linus Torvalds 已提交
116 117
}

118
static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
L
Linus Torvalds 已提交
119 120 121
			  int addr_len)
{
	struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
122
	struct inet_sock *inet = inet_sk(sk);
123
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
124 125
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
126
	struct in6_addr *saddr = NULL, *final_p, final;
127
	struct ipv6_txoptions *opt;
128
	struct flowi6 fl6;
L
Linus Torvalds 已提交
129 130 131
	struct dst_entry *dst;
	int addr_type;
	int err;
132
	struct inet_timewait_death_row *tcp_death_row = &sock_net(sk)->ipv4.tcp_death_row;
L
Linus Torvalds 已提交
133

134
	if (addr_len < SIN6_LEN_RFC2133)
L
Linus Torvalds 已提交
135 136
		return -EINVAL;

137
	if (usin->sin6_family != AF_INET6)
E
Eric Dumazet 已提交
138
		return -EAFNOSUPPORT;
L
Linus Torvalds 已提交
139

140
	memset(&fl6, 0, sizeof(fl6));
L
Linus Torvalds 已提交
141 142

	if (np->sndflow) {
143 144 145
		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
146
			struct ip6_flowlabel *flowlabel;
147
			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
148
			if (!flowlabel)
L
Linus Torvalds 已提交
149 150 151 152 153 154
				return -EINVAL;
			fl6_sock_release(flowlabel);
		}
	}

	/*
155 156 157
	 *	connect() to INADDR_ANY means loopback (BSD'ism).
	 */

158 159 160 161 162 163 164
	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
165 166 167

	addr_type = ipv6_addr_type(&usin->sin6_addr);

W
Weilong Chen 已提交
168
	if (addr_type & IPV6_ADDR_MULTICAST)
L
Linus Torvalds 已提交
169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189
		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 &&
190
	    !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) {
L
Linus Torvalds 已提交
191 192 193 194 195
		tp->rx_opt.ts_recent = 0;
		tp->rx_opt.ts_recent_stamp = 0;
		tp->write_seq = 0;
	}

196
	sk->sk_v6_daddr = usin->sin6_addr;
197
	np->flow_label = fl6.flowlabel;
L
Linus Torvalds 已提交
198 199 200 201 202

	/*
	 *	TCP over IPv4
	 */

203
	if (addr_type & IPV6_ADDR_MAPPED) {
204
		u32 exthdrlen = icsk->icsk_ext_hdr_len;
L
Linus Torvalds 已提交
205 206 207 208 209 210 211 212 213 214 215
		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];

216
		icsk->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
217
		sk->sk_backlog_rcv = tcp_v4_do_rcv;
218 219 220
#ifdef CONFIG_TCP_MD5SIG
		tp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif
L
Linus Torvalds 已提交
221 222 223 224

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

		if (err) {
225 226
			icsk->icsk_ext_hdr_len = exthdrlen;
			icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
227
			sk->sk_backlog_rcv = tcp_v6_do_rcv;
228 229 230
#ifdef CONFIG_TCP_MD5SIG
			tp->af_specific = &tcp_sock_ipv6_specific;
#endif
L
Linus Torvalds 已提交
231 232
			goto failure;
		}
233
		np->saddr = sk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
234 235 236 237

		return err;
	}

238 239
	if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
		saddr = &sk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
240

241
	fl6.flowi6_proto = IPPROTO_TCP;
242
	fl6.daddr = sk->sk_v6_daddr;
A
Alexey Dobriyan 已提交
243
	fl6.saddr = saddr ? *saddr : np->saddr;
244 245
	fl6.flowi6_oif = sk->sk_bound_dev_if;
	fl6.flowi6_mark = sk->sk_mark;
246 247
	fl6.fl6_dport = usin->sin6_port;
	fl6.fl6_sport = inet->inet_sport;
248
	fl6.flowi6_uid = sk->sk_uid;
L
Linus Torvalds 已提交
249

250
	opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
251
	final_p = fl6_update_dst(&fl6, opt, &final);
L
Linus Torvalds 已提交
252

253
	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
V
Venkat Yekkirala 已提交
254

255
	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
256 257
	if (IS_ERR(dst)) {
		err = PTR_ERR(dst);
L
Linus Torvalds 已提交
258
		goto failure;
259
	}
L
Linus Torvalds 已提交
260

261
	if (!saddr) {
262
		saddr = &fl6.saddr;
263
		sk->sk_v6_rcv_saddr = *saddr;
L
Linus Torvalds 已提交
264 265 266
	}

	/* set the source address */
A
Alexey Dobriyan 已提交
267
	np->saddr = *saddr;
E
Eric Dumazet 已提交
268
	inet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
269

H
Herbert Xu 已提交
270
	sk->sk_gso_type = SKB_GSO_TCPV6;
E
Eric Dumazet 已提交
271
	ip6_dst_store(sk, dst, NULL, NULL);
L
Linus Torvalds 已提交
272

273
	icsk->icsk_ext_hdr_len = 0;
274 275 276
	if (opt)
		icsk->icsk_ext_hdr_len = opt->opt_flen +
					 opt->opt_nflen;
L
Linus Torvalds 已提交
277 278 279

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

E
Eric Dumazet 已提交
280
	inet->inet_dport = usin->sin6_port;
L
Linus Torvalds 已提交
281 282

	tcp_set_state(sk, TCP_SYN_SENT);
283
	err = inet6_hash_connect(tcp_death_row, sk);
L
Linus Torvalds 已提交
284 285 286
	if (err)
		goto late_failure;

287
	sk_set_txhash(sk);
288

289 290
	if (likely(!tp->repair)) {
		if (!tp->write_seq)
291 292 293 294
			tp->write_seq = secure_tcpv6_seq(np->saddr.s6_addr32,
							 sk->sk_v6_daddr.s6_addr32,
							 inet->inet_sport,
							 inet->inet_dport);
295 296
		tp->tsoffset = secure_tcpv6_ts_off(sock_net(sk),
						   np->saddr.s6_addr32,
297
						   sk->sk_v6_daddr.s6_addr32);
298
	}
L
Linus Torvalds 已提交
299

W
Wei Wang 已提交
300 301 302 303 304
	if (tcp_fastopen_defer_connect(sk, &err))
		return err;
	if (err)
		goto late_failure;

L
Linus Torvalds 已提交
305 306 307 308 309 310 311 312 313
	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
314
	inet->inet_dport = 0;
L
Linus Torvalds 已提交
315 316 317 318
	sk->sk_route_caps = 0;
	return err;
}

319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335
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
336
static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
337
		u8 type, u8 code, int offset, __be32 info)
L
Linus Torvalds 已提交
338
{
W
Weilong Chen 已提交
339
	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
340
	const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
341 342
	struct net *net = dev_net(skb->dev);
	struct request_sock *fastopen;
L
Linus Torvalds 已提交
343
	struct ipv6_pinfo *np;
344
	struct tcp_sock *tp;
345
	__u32 seq, snd_una;
346
	struct sock *sk;
347
	bool fatal;
348
	int err;
L
Linus Torvalds 已提交
349

350 351 352 353
	sk = __inet6_lookup_established(net, &tcp_hashinfo,
					&hdr->daddr, th->dest,
					&hdr->saddr, ntohs(th->source),
					skb->dev->ifindex);
L
Linus Torvalds 已提交
354

355
	if (!sk) {
E
Eric Dumazet 已提交
356 357
		__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
				  ICMP6_MIB_INERRORS);
L
Linus Torvalds 已提交
358 359 360 361
		return;
	}

	if (sk->sk_state == TCP_TIME_WAIT) {
362
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
363 364
		return;
	}
365
	seq = ntohl(th->seq);
366
	fatal = icmpv6_err_convert(type, code, &err);
367
	if (sk->sk_state == TCP_NEW_SYN_RECV)
368
		return tcp_req_err(sk, seq, fatal);
L
Linus Torvalds 已提交
369 370

	bh_lock_sock(sk);
371
	if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG)
372
		__NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
L
Linus Torvalds 已提交
373 374 375 376

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

377
	if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) {
378
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
379 380 381
		goto out;
	}

L
Linus Torvalds 已提交
382
	tp = tcp_sk(sk);
383 384 385
	/* 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
386
	if (sk->sk_state != TCP_LISTEN &&
387
	    !between(seq, snd_una, tp->snd_nxt)) {
388
		__NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
389 390 391 392 393
		goto out;
	}

	np = inet6_sk(sk);

394
	if (type == NDISC_REDIRECT) {
395 396
		if (!sock_owned_by_user(sk)) {
			struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
397

398 399 400
			if (dst)
				dst->ops->redirect(dst, sk, skb);
		}
401
		goto out;
402 403
	}

L
Linus Torvalds 已提交
404
	if (type == ICMPV6_PKT_TOOBIG) {
405 406 407 408 409 410 411
		/* 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;

412 413 414
		if (!ip6_sk_accept_pmtu(sk))
			goto out;

415 416 417
		tp->mtu_info = ntohl(info);
		if (!sock_owned_by_user(sk))
			tcp_v6_mtu_reduced(sk);
418
		else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED,
419
					   &sk->sk_tsq_flags))
420
			sock_hold(sk);
L
Linus Torvalds 已提交
421 422 423 424
		goto out;
	}


425
	/* Might be for an request_sock */
L
Linus Torvalds 已提交
426 427
	switch (sk->sk_state) {
	case TCP_SYN_SENT:
428 429 430 431
	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.
		 */
432
		if (fastopen && !fastopen->sk)
433 434
			break;

L
Linus Torvalds 已提交
435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456
		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);
}


457
static int tcp_v6_send_synack(const struct sock *sk, struct dst_entry *dst,
458
			      struct flowi *fl,
459
			      struct request_sock *req,
460
			      struct tcp_fastopen_cookie *foc,
461
			      enum tcp_synack_type synack_type)
L
Linus Torvalds 已提交
462
{
463
	struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
464
	struct ipv6_pinfo *np = inet6_sk(sk);
465
	struct ipv6_txoptions *opt;
466
	struct flowi6 *fl6 = &fl->u.ip6;
W
Weilong Chen 已提交
467
	struct sk_buff *skb;
468
	int err = -ENOMEM;
L
Linus Torvalds 已提交
469

470
	/* First, grab a route. */
471 472
	if (!dst && (dst = inet6_csk_route_req(sk, fl6, req,
					       IPPROTO_TCP)) == NULL)
473
		goto done;
474

475
	skb = tcp_make_synack(sk, dst, req, foc, synack_type);
476

L
Linus Torvalds 已提交
477
	if (skb) {
478 479
		__tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr,
				    &ireq->ir_v6_rmt_addr);
L
Linus Torvalds 已提交
480

481
		fl6->daddr = ireq->ir_v6_rmt_addr;
482
		if (np->repflow && ireq->pktopts)
483 484
			fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts));

485
		rcu_read_lock();
486 487 488
		opt = ireq->ipv6_opt;
		if (!opt)
			opt = rcu_dereference(np->opt);
489
		err = ip6_xmit(sk, skb, fl6, sk->sk_mark, opt, np->tclass);
490
		rcu_read_unlock();
491
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
492 493 494 495 496 497
	}

done:
	return err;
}

498

499
static void tcp_v6_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
500
{
501
	kfree(inet_rsk(req)->ipv6_opt);
502
	kfree_skb(inet_rsk(req)->pktopts);
L
Linus Torvalds 已提交
503 504
}

505
#ifdef CONFIG_TCP_MD5SIG
506
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk,
507
						   const struct in6_addr *addr)
508
{
E
Eric Dumazet 已提交
509
	return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
510 511
}

512
static struct tcp_md5sig_key *tcp_v6_md5_lookup(const struct sock *sk,
513
						const struct sock *addr_sk)
514
{
515
	return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr);
516 517
}

W
Wang Yufen 已提交
518 519
static int tcp_v6_parse_md5_keys(struct sock *sk, char __user *optval,
				 int optlen)
520 521 522 523 524 525 526 527 528 529 530 531 532 533
{
	struct tcp_md5sig cmd;
	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;

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

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

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

	if (!cmd.tcpm_keylen) {
B
Brian Haley 已提交
534
		if (ipv6_addr_v4mapped(&sin6->sin6_addr))
E
Eric Dumazet 已提交
535 536 537 538
			return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
					      AF_INET);
		return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
				      AF_INET6);
539 540 541 542 543
	}

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

E
Eric Dumazet 已提交
544 545 546
	if (ipv6_addr_v4mapped(&sin6->sin6_addr))
		return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
				      AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
547

E
Eric Dumazet 已提交
548 549
	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
			      AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
550 551
}

552 553 554 555
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)
556 557
{
	struct tcp6_pseudohdr *bp;
558
	struct scatterlist sg;
559
	struct tcphdr *_th;
560

561
	bp = hp->scratch;
562
	/* 1. TCP pseudo-header (RFC2460) */
A
Alexey Dobriyan 已提交
563 564
	bp->saddr = *saddr;
	bp->daddr = *daddr;
565
	bp->protocol = cpu_to_be32(IPPROTO_TCP);
A
Adam Langley 已提交
566
	bp->len = cpu_to_be32(nbytes);
567

568 569 570 571 572 573 574
	_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
575
	return crypto_ahash_update(hp->md5_req);
576
}
577

578
static int tcp_v6_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
579
			       const struct in6_addr *daddr, struct in6_addr *saddr,
E
Eric Dumazet 已提交
580
			       const struct tcphdr *th)
581 582
{
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
583
	struct ahash_request *req;
584 585 586 587

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
588
	req = hp->md5_req;
589

H
Herbert Xu 已提交
590
	if (crypto_ahash_init(req))
591
		goto clear_hash;
592
	if (tcp_v6_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2))
593 594 595
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
H
Herbert Xu 已提交
596 597
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
598 599 600 601
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
602

603 604 605 606
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
607
	return 1;
608 609
}

610 611
static int tcp_v6_md5_hash_skb(char *md5_hash,
			       const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
612 613
			       const struct sock *sk,
			       const struct sk_buff *skb)
614
{
615
	const struct in6_addr *saddr, *daddr;
616
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
617
	struct ahash_request *req;
E
Eric Dumazet 已提交
618
	const struct tcphdr *th = tcp_hdr(skb);
619

620 621
	if (sk) { /* valid for establish/request sockets */
		saddr = &sk->sk_v6_rcv_saddr;
622
		daddr = &sk->sk_v6_daddr;
623
	} else {
624
		const struct ipv6hdr *ip6h = ipv6_hdr(skb);
625 626
		saddr = &ip6h->saddr;
		daddr = &ip6h->daddr;
627
	}
628 629 630 631

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
632
	req = hp->md5_req;
633

H
Herbert Xu 已提交
634
	if (crypto_ahash_init(req))
635 636
		goto clear_hash;

637
	if (tcp_v6_md5_hash_headers(hp, daddr, saddr, th, skb->len))
638 639 640 641 642
		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
643 644
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
645 646 647 648 649 650 651 652 653 654
		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;
655 656
}

657 658 659 660
#endif

static bool tcp_v6_inbound_md5_hash(const struct sock *sk,
				    const struct sk_buff *skb)
661
{
662
#ifdef CONFIG_TCP_MD5SIG
663
	const __u8 *hash_location = NULL;
664
	struct tcp_md5sig_key *hash_expected;
665
	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
E
Eric Dumazet 已提交
666
	const struct tcphdr *th = tcp_hdr(skb);
667 668 669 670
	int genhash;
	u8 newhash[16];

	hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
671
	hash_location = tcp_parse_md5sig_option(th);
672

673 674
	/* We've parsed the options - do we have a hash? */
	if (!hash_expected && !hash_location)
675
		return false;
676 677

	if (hash_expected && !hash_location) {
678
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
679
		return true;
680 681
	}

682
	if (!hash_expected && hash_location) {
683
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
684
		return true;
685 686 687
	}

	/* check the signature */
688 689
	genhash = tcp_v6_md5_hash_skb(newhash,
				      hash_expected,
690
				      NULL, skb);
691

692
	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
693
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE);
694 695 696 697
		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));
698
		return true;
699
	}
700
#endif
701
	return false;
702 703
}

704 705
static void tcp_v6_init_req(struct request_sock *req,
			    const struct sock *sk_listener,
706 707 708
			    struct sk_buff *skb)
{
	struct inet_request_sock *ireq = inet_rsk(req);
709
	const struct ipv6_pinfo *np = inet6_sk(sk_listener);
710 711 712 713 714

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

	/* So that link locals have meaning */
715
	if (!sk_listener->sk_bound_dev_if &&
716
	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
E
Eric Dumazet 已提交
717
		ireq->ir_iif = tcp_v6_iif(skb);
718

719
	if (!TCP_SKB_CB(skb)->tcp_tw_isn &&
720
	    (ipv6_opt_accepted(sk_listener, skb, &TCP_SKB_CB(skb)->header.h6) ||
721
	     np->rxopt.bits.rxinfo ||
722 723 724 725 726 727 728
	     np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim ||
	     np->rxopt.bits.rxohlim || np->repflow)) {
		atomic_inc(&skb->users);
		ireq->pktopts = skb;
	}
}

729 730
static struct dst_entry *tcp_v6_route_req(const struct sock *sk,
					  struct flowi *fl,
731
					  const struct request_sock *req)
732
{
733
	return inet6_csk_route_req(sk, &fl->u.ip6, req, IPPROTO_TCP);
734 735
}

736
struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
737
	.family		=	AF_INET6,
738
	.obj_size	=	sizeof(struct tcp6_request_sock),
739
	.rtx_syn_ack	=	tcp_rtx_synack,
740 741
	.send_ack	=	tcp_v6_reqsk_send_ack,
	.destructor	=	tcp_v6_reqsk_destructor,
742
	.send_reset	=	tcp_v6_send_reset,
W
Wang Yufen 已提交
743
	.syn_ack_timeout =	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
744 745
};

746
static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
747 748
	.mss_clamp	=	IPV6_MIN_MTU - sizeof(struct tcphdr) -
				sizeof(struct ipv6hdr),
749
#ifdef CONFIG_TCP_MD5SIG
750
	.req_md5_lookup	=	tcp_v6_md5_lookup,
751
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
752
#endif
753
	.init_req	=	tcp_v6_init_req,
754 755 756
#ifdef CONFIG_SYN_COOKIES
	.cookie_init_seq =	cookie_v6_init_sequence,
#endif
757
	.route_req	=	tcp_v6_route_req,
758 759
	.init_seq	=	tcp_v6_init_seq,
	.init_ts_off	=	tcp_v6_init_ts_off,
760
	.send_synack	=	tcp_v6_send_synack,
761
};
762

763
static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 seq,
764 765
				 u32 ack, u32 win, u32 tsval, u32 tsecr,
				 int oif, struct tcp_md5sig_key *key, int rst,
766
				 u8 tclass, __be32 label)
L
Linus Torvalds 已提交
767
{
768 769
	const struct tcphdr *th = tcp_hdr(skb);
	struct tcphdr *t1;
L
Linus Torvalds 已提交
770
	struct sk_buff *buff;
771
	struct flowi6 fl6;
772
	struct net *net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
773
	struct sock *ctl_sk = net->ipv6.tcp_sk;
774
	unsigned int tot_len = sizeof(struct tcphdr);
E
Eric Dumazet 已提交
775
	struct dst_entry *dst;
776
	__be32 *topt;
L
Linus Torvalds 已提交
777

778
	if (tsecr)
779
		tot_len += TCPOLEN_TSTAMP_ALIGNED;
780 781 782 783 784 785
#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
786
			 GFP_ATOMIC);
787
	if (!buff)
788
		return;
L
Linus Torvalds 已提交
789

790
	skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
L
Linus Torvalds 已提交
791

792
	t1 = (struct tcphdr *) skb_push(buff, tot_len);
793
	skb_reset_transport_header(buff);
L
Linus Torvalds 已提交
794 795 796 797 798

	/* Swap the send and the receive. */
	memset(t1, 0, sizeof(*t1));
	t1->dest = th->source;
	t1->source = th->dest;
799
	t1->doff = tot_len / 4;
800 801 802 803 804
	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
805

806 807
	topt = (__be32 *)(t1 + 1);

808
	if (tsecr) {
809 810
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
811 812
		*topt++ = htonl(tsval);
		*topt++ = htonl(tsecr);
813 814
	}

815 816
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
817 818 819
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
		tcp_v6_md5_hash_hdr((__u8 *)topt, key,
820 821
				    &ipv6_hdr(skb)->saddr,
				    &ipv6_hdr(skb)->daddr, t1);
822 823 824
	}
#endif

825
	memset(&fl6, 0, sizeof(fl6));
A
Alexey Dobriyan 已提交
826 827
	fl6.daddr = ipv6_hdr(skb)->saddr;
	fl6.saddr = ipv6_hdr(skb)->daddr;
828
	fl6.flowlabel = label;
L
Linus Torvalds 已提交
829

830 831 832
	buff->ip_summed = CHECKSUM_PARTIAL;
	buff->csum = 0;

833
	__tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
L
Linus Torvalds 已提交
834

835
	fl6.flowi6_proto = IPPROTO_TCP;
836
	if (rt6_need_strict(&fl6.daddr) && !oif)
E
Eric Dumazet 已提交
837
		fl6.flowi6_oif = tcp_v6_iif(skb);
838 839 840 841 842 843
	else {
		if (!oif && netif_index_is_l3_master(net, skb->skb_iif))
			oif = skb->skb_iif;

		fl6.flowi6_oif = oif;
	}
844

845
	fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark);
846 847
	fl6.fl6_dport = t1->dest;
	fl6.fl6_sport = t1->source;
848
	fl6.flowi6_uid = sock_net_uid(net, sk && sk_fullsock(sk) ? sk : NULL);
849
	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
L
Linus Torvalds 已提交
850

851 852 853 854
	/* Pass a socket to ip6_dst_lookup either it is for RST
	 * Underlying function will use this to retrieve the network
	 * namespace
	 */
855
	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
856 857
	if (!IS_ERR(dst)) {
		skb_dst_set(buff, dst);
858
		ip6_xmit(ctl_sk, buff, &fl6, fl6.flowi6_mark, NULL, tclass);
859
		TCP_INC_STATS(net, TCP_MIB_OUTSEGS);
860
		if (rst)
861
			TCP_INC_STATS(net, TCP_MIB_OUTRSTS);
862
		return;
L
Linus Torvalds 已提交
863 864 865 866 867
	}

	kfree_skb(buff);
}

868
static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
869
{
870
	const struct tcphdr *th = tcp_hdr(skb);
871
	u32 seq = 0, ack_seq = 0;
872
	struct tcp_md5sig_key *key = NULL;
873 874 875 876 877 878 879
#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
W
Wang Yufen 已提交
880
	int oif;
L
Linus Torvalds 已提交
881

882
	if (th->rst)
L
Linus Torvalds 已提交
883 884
		return;

885 886 887 888
	/* 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))
889
		return;
L
Linus Torvalds 已提交
890

891
#ifdef CONFIG_TCP_MD5SIG
892
	rcu_read_lock();
893
	hash_location = tcp_parse_md5sig_option(th);
894
	if (sk && sk_fullsock(sk)) {
895 896
		key = tcp_v6_md5_do_lookup(sk, &ipv6h->saddr);
	} else if (hash_location) {
897 898 899 900 901 902 903 904
		/*
		 * 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),
905 906
					   &tcp_hashinfo, NULL, 0,
					   &ipv6h->saddr,
907
					   th->source, &ipv6h->daddr,
E
Eric Dumazet 已提交
908
					   ntohs(th->source), tcp_v6_iif(skb));
909
		if (!sk1)
910
			goto out;
911 912 913

		key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr);
		if (!key)
914
			goto out;
915

916
		genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, skb);
917
		if (genhash || memcmp(hash_location, newhash, 16) != 0)
918
			goto out;
919
	}
920 921
#endif

922 923 924 925 926
	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
927

W
Wang Yufen 已提交
928
	oif = sk ? sk->sk_bound_dev_if : 0;
929
	tcp_v6_send_response(sk, skb, seq, ack_seq, 0, 0, 0, oif, key, 1, 0, 0);
930 931

#ifdef CONFIG_TCP_MD5SIG
932 933
out:
	rcu_read_unlock();
934
#endif
935
}
L
Linus Torvalds 已提交
936

937
static void tcp_v6_send_ack(const struct sock *sk, struct sk_buff *skb, u32 seq,
938
			    u32 ack, u32 win, u32 tsval, u32 tsecr, int oif,
939
			    struct tcp_md5sig_key *key, u8 tclass,
940
			    __be32 label)
941
{
942 943
	tcp_v6_send_response(sk, skb, seq, ack, win, tsval, tsecr, oif, key, 0,
			     tclass, label);
L
Linus Torvalds 已提交
944 945 946 947
}

static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
948
	struct inet_timewait_sock *tw = inet_twsk(sk);
949
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
950

951
	tcp_v6_send_ack(sk, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
952
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
953
			tcp_time_stamp_raw() + tcptw->tw_ts_offset,
W
Wang Yufen 已提交
954
			tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw),
955
			tw->tw_tclass, cpu_to_be32(tw->tw_flowlabel));
L
Linus Torvalds 已提交
956

957
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
958 959
}

960
static void tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
961
				  struct request_sock *req)
L
Linus Torvalds 已提交
962
{
963 964 965
	/* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
	 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
	 */
966 967 968 969 970
	/* 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:
	 */
971
	tcp_v6_send_ack(sk, skb, (sk->sk_state == TCP_LISTEN) ?
972
			tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt,
973 974
			tcp_rsk(req)->rcv_nxt,
			req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale,
975
			tcp_time_stamp_raw() + tcp_rsk(req)->ts_off,
976
			req->ts_recent, sk->sk_bound_dev_if,
977 978
			tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr),
			0, 0);
L
Linus Torvalds 已提交
979 980 981
}


982
static struct sock *tcp_v6_cookie_check(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
983
{
984
#ifdef CONFIG_SYN_COOKIES
985
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
986

987
	if (!th->syn)
988
		sk = cookie_v6_check(sk, skb);
L
Linus Torvalds 已提交
989 990 991 992 993 994 995 996 997 998
#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))
999
		goto drop;
L
Linus Torvalds 已提交
1000

O
Octavian Purdila 已提交
1001 1002
	return tcp_conn_request(&tcp6_request_sock_ops,
				&tcp_request_sock_ipv6_ops, sk, skb);
L
Linus Torvalds 已提交
1003 1004

drop:
1005
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1006 1007 1008
	return 0; /* don't send reset */
}

1009 1010 1011 1012 1013 1014 1015 1016 1017 1018
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));
}

1019
static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
W
Weilong Chen 已提交
1020
					 struct request_sock *req,
1021 1022 1023
					 struct dst_entry *dst,
					 struct request_sock *req_unhash,
					 bool *own_req)
L
Linus Torvalds 已提交
1024
{
1025
	struct inet_request_sock *ireq;
1026 1027
	struct ipv6_pinfo *newnp;
	const struct ipv6_pinfo *np = inet6_sk(sk);
1028
	struct ipv6_txoptions *opt;
L
Linus Torvalds 已提交
1029 1030 1031 1032
	struct tcp6_sock *newtcp6sk;
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1033 1034 1035
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1036
	struct flowi6 fl6;
L
Linus Torvalds 已提交
1037 1038 1039 1040 1041 1042

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

1043 1044
		newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst,
					     req_unhash, own_req);
L
Linus Torvalds 已提交
1045

1046
		if (!newsk)
L
Linus Torvalds 已提交
1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057
			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));

1058
		newnp->saddr = newsk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
1059

1060
		inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
1061
		newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1062 1063 1064 1065
#ifdef CONFIG_TCP_MD5SIG
		newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif

1066
		newnp->ipv6_mc_list = NULL;
1067 1068
		newnp->ipv6_ac_list = NULL;
		newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1069 1070
		newnp->pktoptions  = NULL;
		newnp->opt	   = NULL;
E
Eric Dumazet 已提交
1071
		newnp->mcast_oif   = tcp_v6_iif(skb);
1072
		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
1073
		newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1074 1075
		if (np->repflow)
			newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1076

1077 1078 1079 1080
		/*
		 * 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
1081 1082 1083
		 */

		/* It is tricky place. Until this moment IPv4 tcp
1084
		   worked with IPv6 icsk.icsk_af_ops.
L
Linus Torvalds 已提交
1085 1086
		   Sync it now.
		 */
1087
		tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
1088 1089 1090 1091

		return newsk;
	}

1092
	ireq = inet_rsk(req);
L
Linus Torvalds 已提交
1093 1094 1095 1096

	if (sk_acceptq_is_full(sk))
		goto out_overflow;

1097
	if (!dst) {
1098
		dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_TCP);
1099
		if (!dst)
L
Linus Torvalds 已提交
1100
			goto out;
1101
	}
L
Linus Torvalds 已提交
1102 1103

	newsk = tcp_create_openreq_child(sk, req, skb);
1104
	if (!newsk)
1105
		goto out_nonewsk;
L
Linus Torvalds 已提交
1106

1107 1108 1109 1110 1111
	/*
	 * 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
1112

1113
	newsk->sk_gso_type = SKB_GSO_TCPV6;
E
Eric Dumazet 已提交
1114
	ip6_dst_store(newsk, dst, NULL, NULL);
1115
	inet6_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1116 1117 1118 1119 1120 1121 1122 1123 1124 1125

	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));

1126 1127 1128 1129
	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
1130

1131
	/* Now IPv6 options...
L
Linus Torvalds 已提交
1132 1133 1134

	   First: no IPv4 options.
	 */
1135
	newinet->inet_opt = NULL;
1136
	newnp->ipv6_mc_list = NULL;
1137
	newnp->ipv6_ac_list = NULL;
1138
	newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1139 1140 1141 1142 1143 1144

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

	newnp->pktoptions = NULL;
	newnp->opt	  = NULL;
E
Eric Dumazet 已提交
1145
	newnp->mcast_oif  = tcp_v6_iif(skb);
1146
	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1147
	newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1148 1149
	if (np->repflow)
		newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1150 1151 1152 1153 1154 1155 1156

	/* 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.
	 */
1157 1158 1159
	opt = ireq->ipv6_opt;
	if (!opt)
		opt = rcu_dereference(np->opt);
1160 1161 1162 1163
	if (opt) {
		opt = ipv6_dup_options(newsk, opt);
		RCU_INIT_POINTER(newnp->opt, opt);
	}
1164
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
1165 1166 1167
	if (opt)
		inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
						    opt->opt_flen;
L
Linus Torvalds 已提交
1168

1169 1170
	tcp_ca_openreq_child(newsk, dst);

L
Linus Torvalds 已提交
1171
	tcp_sync_mss(newsk, dst_mtu(dst));
E
Eric Dumazet 已提交
1172
	newtp->advmss = tcp_mss_clamp(tcp_sk(sk), dst_metric_advmss(dst));
1173

L
Linus Torvalds 已提交
1174 1175
	tcp_initialize_rcv_mss(newsk);

E
Eric Dumazet 已提交
1176 1177
	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
1178

1179 1180
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
W
Wang Yufen 已提交
1181
	key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr);
1182
	if (key) {
1183 1184 1185 1186 1187
		/* 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.
		 */
1188
		tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
1189
			       AF_INET6, key->key, key->keylen,
1190
			       sk_gfp_mask(sk, GFP_ATOMIC));
1191 1192 1193
	}
#endif

1194
	if (__inet_inherit_port(sk, newsk) < 0) {
1195 1196
		inet_csk_prepare_forced_close(newsk);
		tcp_done(newsk);
1197 1198
		goto out;
	}
1199
	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
E
Eric Dumazet 已提交
1200
	if (*own_req) {
1201
		tcp_move_syn(newtp, req);
E
Eric Dumazet 已提交
1202 1203 1204 1205

		/* Clone pktoptions received with SYN, if we own the req */
		if (ireq->pktopts) {
			newnp->pktoptions = skb_clone(ireq->pktopts,
1206
						      sk_gfp_mask(sk, GFP_ATOMIC));
E
Eric Dumazet 已提交
1207 1208
			consume_skb(ireq->pktopts);
			ireq->pktopts = NULL;
1209 1210
			if (newnp->pktoptions) {
				tcp_v6_restore_cb(newnp->pktoptions);
E
Eric Dumazet 已提交
1211
				skb_set_owner_r(newnp->pktoptions, newsk);
1212
			}
E
Eric Dumazet 已提交
1213
		}
E
Eric Dumazet 已提交
1214
	}
L
Linus Torvalds 已提交
1215 1216 1217 1218

	return newsk;

out_overflow:
1219
	__NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1220
out_nonewsk:
L
Linus Torvalds 已提交
1221
	dst_release(dst);
1222
out:
1223
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1224 1225 1226 1227
	return NULL;
}

/* The socket must have it's spinlock held when we get
1228
 * here, unless it is a TCP_LISTEN socket.
L
Linus Torvalds 已提交
1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
 *
 * 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);

1252
	if (tcp_filter(sk, skb))
L
Linus Torvalds 已提交
1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
		goto discard;

	/*
	 *	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.
1271
					       --ANK (980728)
L
Linus Torvalds 已提交
1272 1273
	 */
	if (np->rxopt.all)
1274
		opt_skb = skb_clone(skb, sk_gfp_mask(sk, GFP_ATOMIC));
L
Linus Torvalds 已提交
1275 1276

	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
E
Eric Dumazet 已提交
1277 1278
		struct dst_entry *dst = sk->sk_rx_dst;

1279
		sock_rps_save_rxhash(sk, skb);
1280
		sk_mark_napi_id(sk, skb);
E
Eric Dumazet 已提交
1281 1282 1283 1284 1285 1286 1287 1288
		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;
			}
		}

1289
		tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
L
Linus Torvalds 已提交
1290 1291 1292 1293 1294
		if (opt_skb)
			goto ipv6_pktoptions;
		return 0;
	}

E
Eric Dumazet 已提交
1295
	if (tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1296 1297
		goto csum_err;

1298
	if (sk->sk_state == TCP_LISTEN) {
1299 1300
		struct sock *nsk = tcp_v6_cookie_check(sk, skb);

L
Linus Torvalds 已提交
1301 1302 1303
		if (!nsk)
			goto discard;

W
Weilong Chen 已提交
1304
		if (nsk != sk) {
L
Linus Torvalds 已提交
1305 1306 1307 1308 1309 1310
			if (tcp_child_process(sk, nsk, skb))
				goto reset;
			if (opt_skb)
				__kfree_skb(opt_skb);
			return 0;
		}
1311
	} else
1312
		sock_rps_save_rxhash(sk, skb);
L
Linus Torvalds 已提交
1313

1314
	if (tcp_rcv_state_process(sk, skb))
L
Linus Torvalds 已提交
1315 1316 1317 1318 1319 1320
		goto reset;
	if (opt_skb)
		goto ipv6_pktoptions;
	return 0;

reset:
1321
	tcp_v6_send_reset(sk, skb);
L
Linus Torvalds 已提交
1322 1323 1324 1325 1326 1327
discard:
	if (opt_skb)
		__kfree_skb(opt_skb);
	kfree_skb(skb);
	return 0;
csum_err:
1328 1329
	TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS);
	TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343
	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))) {
1344
		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
E
Eric Dumazet 已提交
1345
			np->mcast_oif = tcp_v6_iif(opt_skb);
1346
		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1347
			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1348
		if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
1349
			np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
1350 1351
		if (np->repflow)
			np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
1352
		if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->header.h6)) {
L
Linus Torvalds 已提交
1353
			skb_set_owner_r(opt_skb, sk);
1354
			tcp_v6_restore_cb(opt_skb);
L
Linus Torvalds 已提交
1355 1356 1357 1358 1359 1360 1361
			opt_skb = xchg(&np->pktoptions, opt_skb);
		} else {
			__kfree_skb(opt_skb);
			opt_skb = xchg(&np->pktoptions, NULL);
		}
	}

1362
	kfree_skb(opt_skb);
L
Linus Torvalds 已提交
1363 1364 1365
	return 0;
}

1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387
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;
}

1388
static int tcp_v6_rcv(struct sk_buff *skb)
L
Linus Torvalds 已提交
1389
{
1390
	const struct tcphdr *th;
1391
	const struct ipv6hdr *hdr;
1392
	bool refcounted;
L
Linus Torvalds 已提交
1393 1394
	struct sock *sk;
	int ret;
1395
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
1396 1397 1398 1399 1400 1401 1402

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

	/*
	 *	Count it even if it's bad.
	 */
E
Eric Dumazet 已提交
1403
	__TCP_INC_STATS(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1404 1405 1406 1407

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

1408
	th = (const struct tcphdr *)skb->data;
L
Linus Torvalds 已提交
1409

1410
	if (unlikely(th->doff < sizeof(struct tcphdr)/4))
L
Linus Torvalds 已提交
1411 1412 1413 1414
		goto bad_packet;
	if (!pskb_may_pull(skb, th->doff*4))
		goto discard_it;

1415
	if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo))
1416
		goto csum_error;
L
Linus Torvalds 已提交
1417

1418
	th = (const struct tcphdr *)skb->data;
1419
	hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
1420

1421
lookup:
1422
	sk = __inet6_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th),
1423 1424
				th->source, th->dest, inet6_iif(skb),
				&refcounted);
L
Linus Torvalds 已提交
1425 1426 1427 1428 1429 1430 1431
	if (!sk)
		goto no_tcp_socket;

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

1432 1433
	if (sk->sk_state == TCP_NEW_SYN_RECV) {
		struct request_sock *req = inet_reqsk(sk);
1434
		struct sock *nsk;
1435 1436 1437 1438

		sk = req->rsk_listener;
		tcp_v6_fill_cb(skb, hdr, th);
		if (tcp_v6_inbound_md5_hash(sk, skb)) {
1439
			sk_drops_add(sk, skb);
1440 1441 1442
			reqsk_put(req);
			goto discard_it;
		}
1443
		if (unlikely(sk->sk_state != TCP_LISTEN)) {
1444
			inet_csk_reqsk_queue_drop_and_put(sk, req);
1445 1446
			goto lookup;
		}
1447
		sock_hold(sk);
1448
		refcounted = true;
1449
		nsk = tcp_check_req(sk, skb, req, false);
1450 1451
		if (!nsk) {
			reqsk_put(req);
1452
			goto discard_and_relse;
1453 1454 1455 1456 1457 1458
		}
		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);
1459
			goto discard_and_relse;
1460
		} else {
1461
			sock_put(sk);
1462 1463 1464
			return 0;
		}
	}
1465
	if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) {
1466
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
1467 1468 1469
		goto discard_and_relse;
	}

L
Linus Torvalds 已提交
1470 1471 1472
	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;

1473 1474
	tcp_v6_fill_cb(skb, hdr, th);

1475 1476 1477
	if (tcp_v6_inbound_md5_hash(sk, skb))
		goto discard_and_relse;

1478
	if (tcp_filter(sk, skb))
L
Linus Torvalds 已提交
1479
		goto discard_and_relse;
1480 1481
	th = (const struct tcphdr *)skb->data;
	hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
1482 1483 1484

	skb->dev = NULL;

1485 1486 1487 1488 1489 1490 1491
	if (sk->sk_state == TCP_LISTEN) {
		ret = tcp_v6_do_rcv(sk, skb);
		goto put_and_return;
	}

	sk_incoming_cpu_update(sk);

1492
	bh_lock_sock_nested(sk);
1493
	tcp_segs_in(tcp_sk(sk), skb);
L
Linus Torvalds 已提交
1494 1495
	ret = 0;
	if (!sock_owned_by_user(sk)) {
D
Dan Williams 已提交
1496
		if (!tcp_prequeue(sk, skb))
1497
			ret = tcp_v6_do_rcv(sk, skb);
E
Eric Dumazet 已提交
1498
	} else if (tcp_add_backlog(sk, skb)) {
Z
Zhu Yi 已提交
1499 1500
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1501 1502
	bh_unlock_sock(sk);

1503
put_and_return:
1504 1505
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1506 1507 1508 1509 1510 1511
	return ret ? -1 : 0;

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

1512 1513
	tcp_v6_fill_cb(skb, hdr, th);

E
Eric Dumazet 已提交
1514
	if (tcp_checksum_complete(skb)) {
1515
csum_error:
E
Eric Dumazet 已提交
1516
		__TCP_INC_STATS(net, TCP_MIB_CSUMERRORS);
L
Linus Torvalds 已提交
1517
bad_packet:
E
Eric Dumazet 已提交
1518
		__TCP_INC_STATS(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1519
	} else {
1520
		tcp_v6_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1521 1522 1523 1524 1525 1526 1527
	}

discard_it:
	kfree_skb(skb);
	return 0;

discard_and_relse:
1528
	sk_drops_add(sk, skb);
1529 1530
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1531 1532 1533 1534
	goto discard_it;

do_time_wait:
	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1535
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1536 1537 1538
		goto discard_it;
	}

1539 1540
	tcp_v6_fill_cb(skb, hdr, th);

1541 1542 1543
	if (tcp_checksum_complete(skb)) {
		inet_twsk_put(inet_twsk(sk));
		goto csum_error;
L
Linus Torvalds 已提交
1544 1545
	}

1546
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1547 1548 1549 1550
	case TCP_TW_SYN:
	{
		struct sock *sk2;

1551
		sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1552
					    skb, __tcp_hdrlen(th),
1553
					    &ipv6_hdr(skb)->saddr, th->source,
1554
					    &ipv6_hdr(skb)->daddr,
E
Eric Dumazet 已提交
1555
					    ntohs(th->dest), tcp_v6_iif(skb));
1556
		if (sk2) {
1557
			struct inet_timewait_sock *tw = inet_twsk(sk);
1558
			inet_twsk_deschedule_put(tw);
L
Linus Torvalds 已提交
1559
			sk = sk2;
1560
			tcp_v6_restore_cb(skb);
1561
			refcounted = false;
L
Linus Torvalds 已提交
1562 1563 1564 1565 1566 1567 1568 1569
			goto process;
		}
		/* Fall through to ACK */
	}
	case TCP_TW_ACK:
		tcp_v6_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
1570
		tcp_v6_restore_cb(skb);
1571 1572 1573
		tcp_v6_send_reset(sk, skb);
		inet_twsk_deschedule_put(inet_twsk(sk));
		goto discard_it;
W
Wang Yufen 已提交
1574 1575
	case TCP_TW_SUCCESS:
		;
L
Linus Torvalds 已提交
1576 1577 1578 1579
	}
	goto discard_it;
}

E
Eric Dumazet 已提交
1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597
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
1598
	/* Note : We use inet6_iif() here, not tcp_v6_iif() */
E
Eric Dumazet 已提交
1599 1600 1601 1602 1603 1604 1605
	sk = __inet6_lookup_established(dev_net(skb->dev), &tcp_hashinfo,
					&hdr->saddr, th->source,
					&hdr->daddr, ntohs(th->dest),
					inet6_iif(skb));
	if (sk) {
		skb->sk = sk;
		skb->destructor = sock_edemux;
1606
		if (sk_fullsock(sk)) {
1607
			struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst);
1608

E
Eric Dumazet 已提交
1609
			if (dst)
E
Eric Dumazet 已提交
1610
				dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie);
E
Eric Dumazet 已提交
1611
			if (dst &&
1612
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
E
Eric Dumazet 已提交
1613 1614 1615 1616 1617
				skb_dst_set_noref(skb, dst);
		}
	}
}

1618 1619 1620
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
1621
	.twsk_destructor = tcp_twsk_destructor,
1622 1623
};

1624
static const struct inet_connection_sock_af_ops ipv6_specific = {
1625 1626 1627
	.queue_xmit	   = inet6_csk_xmit,
	.send_check	   = tcp_v6_send_check,
	.rebuild_header	   = inet6_sk_rebuild_header,
E
Eric Dumazet 已提交
1628
	.sk_rx_dst_set	   = inet6_sk_rx_dst_set,
1629 1630 1631
	.conn_request	   = tcp_v6_conn_request,
	.syn_recv_sock	   = tcp_v6_syn_recv_sock,
	.net_header_len	   = sizeof(struct ipv6hdr),
1632
	.net_frag_header_len = sizeof(struct frag_hdr),
1633 1634 1635 1636
	.setsockopt	   = ipv6_setsockopt,
	.getsockopt	   = ipv6_getsockopt,
	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in6),
1637
#ifdef CONFIG_COMPAT
1638 1639
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1640
#endif
1641
	.mtu_reduced	   = tcp_v6_mtu_reduced,
L
Linus Torvalds 已提交
1642 1643
};

1644
#ifdef CONFIG_TCP_MD5SIG
1645
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1646
	.md5_lookup	=	tcp_v6_md5_lookup,
1647
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
1648 1649
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1650
#endif
1651

L
Linus Torvalds 已提交
1652 1653 1654
/*
 *	TCP over IPv4 via INET6 API
 */
1655
static const struct inet_connection_sock_af_ops ipv6_mapped = {
1656 1657 1658
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
1659
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1660 1661 1662 1663 1664 1665 1666
	.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),
1667
#ifdef CONFIG_COMPAT
1668 1669
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1670
#endif
1671
	.mtu_reduced	   = tcp_v4_mtu_reduced,
L
Linus Torvalds 已提交
1672 1673
};

1674
#ifdef CONFIG_TCP_MD5SIG
1675
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1676
	.md5_lookup	=	tcp_v4_md5_lookup,
1677
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1678 1679
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1680
#endif
1681

L
Linus Torvalds 已提交
1682 1683 1684 1685 1686
/* 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)
{
1687
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1688

1689
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1690

1691
	icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
1692

1693
#ifdef CONFIG_TCP_MD5SIG
1694
	tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
1695 1696
#endif

L
Linus Torvalds 已提交
1697 1698 1699
	return 0;
}

1700
static void tcp_v6_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1701 1702
{
	tcp_v4_destroy_sock(sk);
1703
	inet6_destroy_sock(sk);
L
Linus Torvalds 已提交
1704 1705
}

1706
#ifdef CONFIG_PROC_FS
L
Linus Torvalds 已提交
1707
/* Proc filesystem TCPv6 sock list dumping. */
1708
static void get_openreq6(struct seq_file *seq,
E
Eric Dumazet 已提交
1709
			 const struct request_sock *req, int i)
L
Linus Torvalds 已提交
1710
{
1711
	long ttd = req->rsk_timer.expires - jiffies;
1712 1713
	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
1714 1715 1716 1717 1718 1719

	if (ttd < 0)
		ttd = 0;

	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1720
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1721 1722 1723
		   i,
		   src->s6_addr32[0], src->s6_addr32[1],
		   src->s6_addr32[2], src->s6_addr32[3],
E
Eric Dumazet 已提交
1724
		   inet_rsk(req)->ir_num,
L
Linus Torvalds 已提交
1725 1726
		   dest->s6_addr32[0], dest->s6_addr32[1],
		   dest->s6_addr32[2], dest->s6_addr32[3],
1727
		   ntohs(inet_rsk(req)->ir_rmt_port),
L
Linus Torvalds 已提交
1728
		   TCP_SYN_RECV,
W
Weilong Chen 已提交
1729
		   0, 0, /* could print option size, but that is af dependent. */
1730 1731
		   1,   /* timers active (only the expire timer) */
		   jiffies_to_clock_t(ttd),
1732
		   req->num_timeout,
E
Eric Dumazet 已提交
1733 1734
		   from_kuid_munged(seq_user_ns(seq),
				    sock_i_uid(req->rsk_listener)),
1735
		   0,  /* non standard timer */
L
Linus Torvalds 已提交
1736 1737 1738 1739 1740 1741
		   0, /* open_requests have no inode */
		   0, req);
}

static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
{
1742
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1743 1744 1745
	__u16 destp, srcp;
	int timer_active;
	unsigned long timer_expires;
1746 1747
	const struct inet_sock *inet = inet_sk(sp);
	const struct tcp_sock *tp = tcp_sk(sp);
1748
	const struct inet_connection_sock *icsk = inet_csk(sp);
1749
	const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq;
1750 1751
	int rx_queue;
	int state;
L
Linus Torvalds 已提交
1752

1753 1754
	dest  = &sp->sk_v6_daddr;
	src   = &sp->sk_v6_rcv_saddr;
E
Eric Dumazet 已提交
1755 1756
	destp = ntohs(inet->inet_dport);
	srcp  = ntohs(inet->inet_sport);
1757

1758
	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
1759
	    icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
1760
	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
L
Linus Torvalds 已提交
1761
		timer_active	= 1;
1762 1763
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
1764
		timer_active	= 4;
1765
		timer_expires	= icsk->icsk_timeout;
L
Linus Torvalds 已提交
1766 1767 1768 1769 1770 1771 1772 1773
	} else if (timer_pending(&sp->sk_timer)) {
		timer_active	= 2;
		timer_expires	= sp->sk_timer.expires;
	} else {
		timer_active	= 0;
		timer_expires = jiffies;
	}

1774 1775 1776 1777 1778 1779 1780 1781 1782
	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
1783 1784
	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1785
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
L
Linus Torvalds 已提交
1786 1787 1788 1789 1790
		   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,
1791 1792 1793
		   state,
		   tp->write_seq - tp->snd_una,
		   rx_queue,
L
Linus Torvalds 已提交
1794
		   timer_active,
1795
		   jiffies_delta_to_clock_t(timer_expires - jiffies),
1796
		   icsk->icsk_retransmits,
1797
		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1798
		   icsk->icsk_probes_out,
L
Linus Torvalds 已提交
1799 1800
		   sock_i_ino(sp),
		   atomic_read(&sp->sk_refcnt), sp,
1801 1802
		   jiffies_to_clock_t(icsk->icsk_rto),
		   jiffies_to_clock_t(icsk->icsk_ack.ato),
W
Weilong Chen 已提交
1803
		   (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
I
Ilpo Järvinen 已提交
1804
		   tp->snd_cwnd,
1805
		   state == TCP_LISTEN ?
1806
			fastopenq->max_qlen :
1807
			(tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)
L
Linus Torvalds 已提交
1808 1809 1810
		   );
}

1811
static void get_timewait6_sock(struct seq_file *seq,
1812
			       struct inet_timewait_sock *tw, int i)
L
Linus Torvalds 已提交
1813
{
1814
	long delta = tw->tw_timer.expires - jiffies;
1815
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1816 1817
	__u16 destp, srcp;

1818 1819
	dest = &tw->tw_v6_daddr;
	src  = &tw->tw_v6_rcv_saddr;
L
Linus Torvalds 已提交
1820 1821 1822 1823 1824
	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
1825
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1826 1827 1828 1829 1830 1831
		   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,
1832
		   3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
L
Linus Torvalds 已提交
1833 1834 1835 1836 1837 1838
		   atomic_read(&tw->tw_refcnt), tw);
}

static int tcp6_seq_show(struct seq_file *seq, void *v)
{
	struct tcp_iter_state *st;
E
Eric Dumazet 已提交
1839
	struct sock *sk = v;
L
Linus Torvalds 已提交
1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851

	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;

1852 1853 1854
	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
1855
		get_openreq6(seq, v, st->num);
1856 1857
	else
		get_tcp6_sock(seq, v, st->num);
L
Linus Torvalds 已提交
1858 1859 1860 1861
out:
	return 0;
}

1862 1863 1864 1865 1866 1867 1868 1869
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
1870 1871 1872
static struct tcp_seq_afinfo tcp6_seq_afinfo = {
	.name		= "tcp6",
	.family		= AF_INET6,
1873
	.seq_fops	= &tcp6_afinfo_seq_fops,
1874 1875 1876
	.seq_ops	= {
		.show		= tcp6_seq_show,
	},
L
Linus Torvalds 已提交
1877 1878
};

1879
int __net_init tcp6_proc_init(struct net *net)
L
Linus Torvalds 已提交
1880
{
1881
	return tcp_proc_register(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1882 1883
}

1884
void tcp6_proc_exit(struct net *net)
L
Linus Torvalds 已提交
1885
{
1886
	tcp_proc_unregister(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1887 1888 1889 1890 1891 1892 1893 1894 1895
}
#endif

struct proto tcpv6_prot = {
	.name			= "TCPv6",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v6_connect,
	.disconnect		= tcp_disconnect,
1896
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
1897 1898 1899 1900 1901 1902
	.ioctl			= tcp_ioctl,
	.init			= tcp_v6_init_sock,
	.destroy		= tcp_v6_destroy_sock,
	.shutdown		= tcp_shutdown,
	.setsockopt		= tcp_setsockopt,
	.getsockopt		= tcp_getsockopt,
1903
	.keepalive		= tcp_set_keepalive,
L
Linus Torvalds 已提交
1904
	.recvmsg		= tcp_recvmsg,
1905 1906
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
1907
	.backlog_rcv		= tcp_v6_do_rcv,
E
Eric Dumazet 已提交
1908
	.release_cb		= tcp_release_cb,
1909
	.hash			= inet6_hash,
1910 1911
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
1912
	.enter_memory_pressure	= tcp_enter_memory_pressure,
1913
	.leave_memory_pressure	= tcp_leave_memory_pressure,
1914
	.stream_memory_free	= tcp_stream_memory_free,
L
Linus Torvalds 已提交
1915 1916 1917
	.sockets_allocated	= &tcp_sockets_allocated,
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
1918
	.orphan_count		= &tcp_orphan_count,
1919
	.sysctl_mem		= sysctl_tcp_mem,
L
Linus Torvalds 已提交
1920 1921 1922 1923
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp6_sock),
1924
	.slab_flags		= SLAB_TYPESAFE_BY_RCU,
1925
	.twsk_prot		= &tcp6_timewait_sock_ops,
1926
	.rsk_prot		= &tcp6_request_sock_ops,
1927
	.h.hashinfo		= &tcp_hashinfo,
1928
	.no_autobind		= true,
1929 1930 1931
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
G
Glauber Costa 已提交
1932
#endif
1933
	.diag_destroy		= tcp_abort,
L
Linus Torvalds 已提交
1934 1935
};

1936
static struct inet6_protocol tcpv6_protocol = {
E
Eric Dumazet 已提交
1937
	.early_demux	=	tcp_v6_early_demux,
1938
	.early_demux_handler =  tcp_v6_early_demux,
L
Linus Torvalds 已提交
1939 1940 1941 1942 1943 1944 1945 1946 1947 1948
	.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,
1949 1950
	.flags		=	INET_PROTOSW_PERMANENT |
				INET_PROTOSW_ICSK,
L
Linus Torvalds 已提交
1951 1952
};

1953
static int __net_init tcpv6_net_init(struct net *net)
1954
{
1955 1956
	return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
				    SOCK_RAW, IPPROTO_TCP, net);
1957 1958
}

1959
static void __net_exit tcpv6_net_exit(struct net *net)
1960
{
1961
	inet_ctl_sock_destroy(net->ipv6.tcp_sk);
E
Eric W. Biederman 已提交
1962 1963
}

1964
static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
E
Eric W. Biederman 已提交
1965
{
1966
	inet_twsk_purge(&tcp_hashinfo, AF_INET6);
1967 1968 1969
}

static struct pernet_operations tcpv6_net_ops = {
E
Eric W. Biederman 已提交
1970 1971 1972
	.init	    = tcpv6_net_init,
	.exit	    = tcpv6_net_exit,
	.exit_batch = tcpv6_net_exit_batch,
1973 1974
};

1975
int __init tcpv6_init(void)
L
Linus Torvalds 已提交
1976
{
1977 1978
	int ret;

1979 1980
	ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
	if (ret)
1981
		goto out;
1982

L
Linus Torvalds 已提交
1983
	/* register inet6 protocol */
1984 1985 1986 1987
	ret = inet6_register_protosw(&tcpv6_protosw);
	if (ret)
		goto out_tcpv6_protocol;

1988
	ret = register_pernet_subsys(&tcpv6_net_ops);
1989 1990 1991 1992
	if (ret)
		goto out_tcpv6_protosw;
out:
	return ret;
1993

1994 1995
out_tcpv6_protosw:
	inet6_unregister_protosw(&tcpv6_protosw);
1996 1997
out_tcpv6_protocol:
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
1998 1999 2000
	goto out;
}

2001
void tcpv6_exit(void)
2002
{
2003
	unregister_pernet_subsys(&tcpv6_net_ops);
2004 2005
	inet6_unregister_protosw(&tcpv6_protosw);
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
L
Linus Torvalds 已提交
2006
}