tcp_ipv6.c 52.8 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
}

A
Andrey Ignatov 已提交
120 121 122 123 124 125 126 127 128 129 130 131 132 133 134
static int tcp_v6_pre_connect(struct sock *sk, struct sockaddr *uaddr,
			      int addr_len)
{
	/* This check is replicated from tcp_v6_connect() and intended to
	 * prevent BPF program called below from accessing bytes that are out
	 * of the bound specified by user in addr_len.
	 */
	if (addr_len < SIN6_LEN_RFC2133)
		return -EINVAL;

	sock_owned_by_me(sk);

	return BPF_CGROUP_RUN_PROG_INET6_CONNECT(sk, uaddr);
}

135
static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
L
Linus Torvalds 已提交
136 137 138
			  int addr_len)
{
	struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
139
	struct inet_sock *inet = inet_sk(sk);
140
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
141 142
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
143
	struct in6_addr *saddr = NULL, *final_p, final;
144
	struct ipv6_txoptions *opt;
145
	struct flowi6 fl6;
L
Linus Torvalds 已提交
146 147 148
	struct dst_entry *dst;
	int addr_type;
	int err;
149
	struct inet_timewait_death_row *tcp_death_row = &sock_net(sk)->ipv4.tcp_death_row;
L
Linus Torvalds 已提交
150

151
	if (addr_len < SIN6_LEN_RFC2133)
L
Linus Torvalds 已提交
152 153
		return -EINVAL;

154
	if (usin->sin6_family != AF_INET6)
E
Eric Dumazet 已提交
155
		return -EAFNOSUPPORT;
L
Linus Torvalds 已提交
156

157
	memset(&fl6, 0, sizeof(fl6));
L
Linus Torvalds 已提交
158 159

	if (np->sndflow) {
160 161 162
		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
163
			struct ip6_flowlabel *flowlabel;
164
			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
165
			if (!flowlabel)
L
Linus Torvalds 已提交
166 167 168 169 170 171
				return -EINVAL;
			fl6_sock_release(flowlabel);
		}
	}

	/*
172 173 174
	 *	connect() to INADDR_ANY means loopback (BSD'ism).
	 */

175 176 177 178 179 180 181
	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
182 183 184

	addr_type = ipv6_addr_type(&usin->sin6_addr);

W
Weilong Chen 已提交
185
	if (addr_type & IPV6_ADDR_MULTICAST)
L
Linus Torvalds 已提交
186 187 188 189 190 191 192 193
		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.
			 */
194
			if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id))
L
Linus Torvalds 已提交
195 196 197 198 199 200 201 202 203 204 205
				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 &&
206
	    !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) {
L
Linus Torvalds 已提交
207 208 209 210 211
		tp->rx_opt.ts_recent = 0;
		tp->rx_opt.ts_recent_stamp = 0;
		tp->write_seq = 0;
	}

212
	sk->sk_v6_daddr = usin->sin6_addr;
213
	np->flow_label = fl6.flowlabel;
L
Linus Torvalds 已提交
214 215 216 217 218

	/*
	 *	TCP over IPv4
	 */

219
	if (addr_type & IPV6_ADDR_MAPPED) {
220
		u32 exthdrlen = icsk->icsk_ext_hdr_len;
L
Linus Torvalds 已提交
221 222 223 224 225 226 227 228 229 230 231
		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];

232
		icsk->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
233
		sk->sk_backlog_rcv = tcp_v4_do_rcv;
234 235 236
#ifdef CONFIG_TCP_MD5SIG
		tp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif
L
Linus Torvalds 已提交
237 238 239 240

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

		if (err) {
241 242
			icsk->icsk_ext_hdr_len = exthdrlen;
			icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
243
			sk->sk_backlog_rcv = tcp_v6_do_rcv;
244 245 246
#ifdef CONFIG_TCP_MD5SIG
			tp->af_specific = &tcp_sock_ipv6_specific;
#endif
L
Linus Torvalds 已提交
247 248
			goto failure;
		}
249
		np->saddr = sk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
250 251 252 253

		return err;
	}

254 255
	if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
		saddr = &sk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
256

257
	fl6.flowi6_proto = IPPROTO_TCP;
258
	fl6.daddr = sk->sk_v6_daddr;
A
Alexey Dobriyan 已提交
259
	fl6.saddr = saddr ? *saddr : np->saddr;
260 261
	fl6.flowi6_oif = sk->sk_bound_dev_if;
	fl6.flowi6_mark = sk->sk_mark;
262 263
	fl6.fl6_dport = usin->sin6_port;
	fl6.fl6_sport = inet->inet_sport;
264
	fl6.flowi6_uid = sk->sk_uid;
L
Linus Torvalds 已提交
265

266
	opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
267
	final_p = fl6_update_dst(&fl6, opt, &final);
L
Linus Torvalds 已提交
268

269
	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
V
Venkat Yekkirala 已提交
270

271
	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
272 273
	if (IS_ERR(dst)) {
		err = PTR_ERR(dst);
L
Linus Torvalds 已提交
274
		goto failure;
275
	}
L
Linus Torvalds 已提交
276

277
	if (!saddr) {
278
		saddr = &fl6.saddr;
279
		sk->sk_v6_rcv_saddr = *saddr;
L
Linus Torvalds 已提交
280 281 282
	}

	/* set the source address */
A
Alexey Dobriyan 已提交
283
	np->saddr = *saddr;
E
Eric Dumazet 已提交
284
	inet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
285

H
Herbert Xu 已提交
286
	sk->sk_gso_type = SKB_GSO_TCPV6;
E
Eric Dumazet 已提交
287
	ip6_dst_store(sk, dst, NULL, NULL);
L
Linus Torvalds 已提交
288

289
	icsk->icsk_ext_hdr_len = 0;
290 291 292
	if (opt)
		icsk->icsk_ext_hdr_len = opt->opt_flen +
					 opt->opt_nflen;
L
Linus Torvalds 已提交
293 294 295

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

E
Eric Dumazet 已提交
296
	inet->inet_dport = usin->sin6_port;
L
Linus Torvalds 已提交
297 298

	tcp_set_state(sk, TCP_SYN_SENT);
299
	err = inet6_hash_connect(tcp_death_row, sk);
L
Linus Torvalds 已提交
300 301 302
	if (err)
		goto late_failure;

303
	sk_set_txhash(sk);
304

305 306
	if (likely(!tp->repair)) {
		if (!tp->write_seq)
307 308 309 310
			tp->write_seq = secure_tcpv6_seq(np->saddr.s6_addr32,
							 sk->sk_v6_daddr.s6_addr32,
							 inet->inet_sport,
							 inet->inet_dport);
311 312
		tp->tsoffset = secure_tcpv6_ts_off(sock_net(sk),
						   np->saddr.s6_addr32,
313
						   sk->sk_v6_daddr.s6_addr32);
314
	}
L
Linus Torvalds 已提交
315

W
Wei Wang 已提交
316 317 318 319 320
	if (tcp_fastopen_defer_connect(sk, &err))
		return err;
	if (err)
		goto late_failure;

L
Linus Torvalds 已提交
321 322 323 324 325 326 327 328 329
	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
330
	inet->inet_dport = 0;
L
Linus Torvalds 已提交
331 332 333 334
	sk->sk_route_caps = 0;
	return err;
}

335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351
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
352
static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
353
		u8 type, u8 code, int offset, __be32 info)
L
Linus Torvalds 已提交
354
{
W
Weilong Chen 已提交
355
	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
356
	const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
357 358
	struct net *net = dev_net(skb->dev);
	struct request_sock *fastopen;
L
Linus Torvalds 已提交
359
	struct ipv6_pinfo *np;
360
	struct tcp_sock *tp;
361
	__u32 seq, snd_una;
362
	struct sock *sk;
363
	bool fatal;
364
	int err;
L
Linus Torvalds 已提交
365

366 367 368
	sk = __inet6_lookup_established(net, &tcp_hashinfo,
					&hdr->daddr, th->dest,
					&hdr->saddr, ntohs(th->source),
369
					skb->dev->ifindex, inet6_sdif(skb));
L
Linus Torvalds 已提交
370

371
	if (!sk) {
E
Eric Dumazet 已提交
372 373
		__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
				  ICMP6_MIB_INERRORS);
L
Linus Torvalds 已提交
374 375 376 377
		return;
	}

	if (sk->sk_state == TCP_TIME_WAIT) {
378
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
379 380
		return;
	}
381
	seq = ntohl(th->seq);
382
	fatal = icmpv6_err_convert(type, code, &err);
383
	if (sk->sk_state == TCP_NEW_SYN_RECV)
384
		return tcp_req_err(sk, seq, fatal);
L
Linus Torvalds 已提交
385 386

	bh_lock_sock(sk);
387
	if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG)
388
		__NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
L
Linus Torvalds 已提交
389 390 391 392

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

393
	if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) {
394
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
395 396 397
		goto out;
	}

L
Linus Torvalds 已提交
398
	tp = tcp_sk(sk);
399 400 401
	/* 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
402
	if (sk->sk_state != TCP_LISTEN &&
403
	    !between(seq, snd_una, tp->snd_nxt)) {
404
		__NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
405 406 407 408 409
		goto out;
	}

	np = inet6_sk(sk);

410
	if (type == NDISC_REDIRECT) {
411 412
		if (!sock_owned_by_user(sk)) {
			struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
413

414 415 416
			if (dst)
				dst->ops->redirect(dst, sk, skb);
		}
417
		goto out;
418 419
	}

L
Linus Torvalds 已提交
420
	if (type == ICMPV6_PKT_TOOBIG) {
421 422 423 424 425 426 427
		/* 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;

428 429 430
		if (!ip6_sk_accept_pmtu(sk))
			goto out;

431 432 433
		tp->mtu_info = ntohl(info);
		if (!sock_owned_by_user(sk))
			tcp_v6_mtu_reduced(sk);
434
		else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED,
435
					   &sk->sk_tsq_flags))
436
			sock_hold(sk);
L
Linus Torvalds 已提交
437 438 439 440
		goto out;
	}


441
	/* Might be for an request_sock */
L
Linus Torvalds 已提交
442 443
	switch (sk->sk_state) {
	case TCP_SYN_SENT:
444 445 446 447
	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.
		 */
448
		if (fastopen && !fastopen->sk)
449 450
			break;

L
Linus Torvalds 已提交
451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472
		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);
}


473
static int tcp_v6_send_synack(const struct sock *sk, struct dst_entry *dst,
474
			      struct flowi *fl,
475
			      struct request_sock *req,
476
			      struct tcp_fastopen_cookie *foc,
477
			      enum tcp_synack_type synack_type)
L
Linus Torvalds 已提交
478
{
479
	struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
480
	struct ipv6_pinfo *np = inet6_sk(sk);
481
	struct ipv6_txoptions *opt;
482
	struct flowi6 *fl6 = &fl->u.ip6;
W
Weilong Chen 已提交
483
	struct sk_buff *skb;
484
	int err = -ENOMEM;
L
Linus Torvalds 已提交
485

486
	/* First, grab a route. */
487 488
	if (!dst && (dst = inet6_csk_route_req(sk, fl6, req,
					       IPPROTO_TCP)) == NULL)
489
		goto done;
490

491
	skb = tcp_make_synack(sk, dst, req, foc, synack_type);
492

L
Linus Torvalds 已提交
493
	if (skb) {
494 495
		__tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr,
				    &ireq->ir_v6_rmt_addr);
L
Linus Torvalds 已提交
496

497
		fl6->daddr = ireq->ir_v6_rmt_addr;
498
		if (np->repflow && ireq->pktopts)
499 500
			fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts));

501
		rcu_read_lock();
502 503 504
		opt = ireq->ipv6_opt;
		if (!opt)
			opt = rcu_dereference(np->opt);
505
		err = ip6_xmit(sk, skb, fl6, sk->sk_mark, opt, np->tclass);
506
		rcu_read_unlock();
507
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
508 509 510 511 512 513
	}

done:
	return err;
}

514

515
static void tcp_v6_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
516
{
517
	kfree(inet_rsk(req)->ipv6_opt);
518
	kfree_skb(inet_rsk(req)->pktopts);
L
Linus Torvalds 已提交
519 520
}

521
#ifdef CONFIG_TCP_MD5SIG
522
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk,
523
						   const struct in6_addr *addr)
524
{
E
Eric Dumazet 已提交
525
	return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
526 527
}

528
static struct tcp_md5sig_key *tcp_v6_md5_lookup(const struct sock *sk,
529
						const struct sock *addr_sk)
530
{
531
	return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr);
532 533
}

534 535
static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
				 char __user *optval, int optlen)
536 537 538
{
	struct tcp_md5sig cmd;
	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
539
	u8 prefixlen;
540 541 542 543 544 545 546 547 548 549

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

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

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

550 551 552 553 554 555 556 557 558 559
	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;
	}

560
	if (!cmd.tcpm_keylen) {
B
Brian Haley 已提交
561
		if (ipv6_addr_v4mapped(&sin6->sin6_addr))
E
Eric Dumazet 已提交
562
			return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
563
					      AF_INET, prefixlen);
E
Eric Dumazet 已提交
564
		return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
565
				      AF_INET6, prefixlen);
566 567 568 569 570
	}

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

E
Eric Dumazet 已提交
571 572
	if (ipv6_addr_v4mapped(&sin6->sin6_addr))
		return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
573
				      AF_INET, prefixlen, cmd.tcpm_key,
574
				      cmd.tcpm_keylen, GFP_KERNEL);
575

E
Eric Dumazet 已提交
576
	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
577 578
			      AF_INET6, prefixlen, cmd.tcpm_key,
			      cmd.tcpm_keylen, GFP_KERNEL);
579 580
}

581 582 583 584
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)
585 586
{
	struct tcp6_pseudohdr *bp;
587
	struct scatterlist sg;
588
	struct tcphdr *_th;
589

590
	bp = hp->scratch;
591
	/* 1. TCP pseudo-header (RFC2460) */
A
Alexey Dobriyan 已提交
592 593
	bp->saddr = *saddr;
	bp->daddr = *daddr;
594
	bp->protocol = cpu_to_be32(IPPROTO_TCP);
A
Adam Langley 已提交
595
	bp->len = cpu_to_be32(nbytes);
596

597 598 599 600 601 602 603
	_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
604
	return crypto_ahash_update(hp->md5_req);
605
}
606

607
static int tcp_v6_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
608
			       const struct in6_addr *daddr, struct in6_addr *saddr,
E
Eric Dumazet 已提交
609
			       const struct tcphdr *th)
610 611
{
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
612
	struct ahash_request *req;
613 614 615 616

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
617
	req = hp->md5_req;
618

H
Herbert Xu 已提交
619
	if (crypto_ahash_init(req))
620
		goto clear_hash;
621
	if (tcp_v6_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2))
622 623 624
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
H
Herbert Xu 已提交
625 626
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
627 628 629 630
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
631

632 633 634 635
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
636
	return 1;
637 638
}

639 640
static int tcp_v6_md5_hash_skb(char *md5_hash,
			       const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
641 642
			       const struct sock *sk,
			       const struct sk_buff *skb)
643
{
644
	const struct in6_addr *saddr, *daddr;
645
	struct tcp_md5sig_pool *hp;
H
Herbert Xu 已提交
646
	struct ahash_request *req;
E
Eric Dumazet 已提交
647
	const struct tcphdr *th = tcp_hdr(skb);
648

649 650
	if (sk) { /* valid for establish/request sockets */
		saddr = &sk->sk_v6_rcv_saddr;
651
		daddr = &sk->sk_v6_daddr;
652
	} else {
653
		const struct ipv6hdr *ip6h = ipv6_hdr(skb);
654 655
		saddr = &ip6h->saddr;
		daddr = &ip6h->daddr;
656
	}
657 658 659 660

	hp = tcp_get_md5sig_pool();
	if (!hp)
		goto clear_hash_noput;
H
Herbert Xu 已提交
661
	req = hp->md5_req;
662

H
Herbert Xu 已提交
663
	if (crypto_ahash_init(req))
664 665
		goto clear_hash;

666
	if (tcp_v6_md5_hash_headers(hp, daddr, saddr, th, skb->len))
667 668 669 670 671
		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
672 673
	ahash_request_set_crypt(req, NULL, md5_hash, 0);
	if (crypto_ahash_final(req))
674 675 676 677 678 679 680 681 682 683
		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;
684 685
}

686 687 688 689
#endif

static bool tcp_v6_inbound_md5_hash(const struct sock *sk,
				    const struct sk_buff *skb)
690
{
691
#ifdef CONFIG_TCP_MD5SIG
692
	const __u8 *hash_location = NULL;
693
	struct tcp_md5sig_key *hash_expected;
694
	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
E
Eric Dumazet 已提交
695
	const struct tcphdr *th = tcp_hdr(skb);
696 697 698 699
	int genhash;
	u8 newhash[16];

	hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
700
	hash_location = tcp_parse_md5sig_option(th);
701

702 703
	/* We've parsed the options - do we have a hash? */
	if (!hash_expected && !hash_location)
704
		return false;
705 706

	if (hash_expected && !hash_location) {
707
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
708
		return true;
709 710
	}

711
	if (!hash_expected && hash_location) {
712
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
713
		return true;
714 715 716
	}

	/* check the signature */
717 718
	genhash = tcp_v6_md5_hash_skb(newhash,
				      hash_expected,
719
				      NULL, skb);
720

721
	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
722
		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE);
723 724 725 726
		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));
727
		return true;
728
	}
729
#endif
730
	return false;
731 732
}

733 734
static void tcp_v6_init_req(struct request_sock *req,
			    const struct sock *sk_listener,
735 736 737
			    struct sk_buff *skb)
{
	struct inet_request_sock *ireq = inet_rsk(req);
738
	const struct ipv6_pinfo *np = inet6_sk(sk_listener);
739 740 741 742 743

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

	/* So that link locals have meaning */
744
	if (!sk_listener->sk_bound_dev_if &&
745
	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
E
Eric Dumazet 已提交
746
		ireq->ir_iif = tcp_v6_iif(skb);
747

748
	if (!TCP_SKB_CB(skb)->tcp_tw_isn &&
749
	    (ipv6_opt_accepted(sk_listener, skb, &TCP_SKB_CB(skb)->header.h6) ||
750
	     np->rxopt.bits.rxinfo ||
751 752
	     np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim ||
	     np->rxopt.bits.rxohlim || np->repflow)) {
753
		refcount_inc(&skb->users);
754 755 756 757
		ireq->pktopts = skb;
	}
}

758 759
static struct dst_entry *tcp_v6_route_req(const struct sock *sk,
					  struct flowi *fl,
760
					  const struct request_sock *req)
761
{
762
	return inet6_csk_route_req(sk, &fl->u.ip6, req, IPPROTO_TCP);
763 764
}

765
struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
766
	.family		=	AF_INET6,
767
	.obj_size	=	sizeof(struct tcp6_request_sock),
768
	.rtx_syn_ack	=	tcp_rtx_synack,
769 770
	.send_ack	=	tcp_v6_reqsk_send_ack,
	.destructor	=	tcp_v6_reqsk_destructor,
771
	.send_reset	=	tcp_v6_send_reset,
W
Wang Yufen 已提交
772
	.syn_ack_timeout =	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
773 774
};

775
static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
776 777
	.mss_clamp	=	IPV6_MIN_MTU - sizeof(struct tcphdr) -
				sizeof(struct ipv6hdr),
778
#ifdef CONFIG_TCP_MD5SIG
779
	.req_md5_lookup	=	tcp_v6_md5_lookup,
780
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
781
#endif
782
	.init_req	=	tcp_v6_init_req,
783 784 785
#ifdef CONFIG_SYN_COOKIES
	.cookie_init_seq =	cookie_v6_init_sequence,
#endif
786
	.route_req	=	tcp_v6_route_req,
787 788
	.init_seq	=	tcp_v6_init_seq,
	.init_ts_off	=	tcp_v6_init_ts_off,
789
	.send_synack	=	tcp_v6_send_synack,
790
};
791

792
static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 seq,
793 794
				 u32 ack, u32 win, u32 tsval, u32 tsecr,
				 int oif, struct tcp_md5sig_key *key, int rst,
795
				 u8 tclass, __be32 label)
L
Linus Torvalds 已提交
796
{
797 798
	const struct tcphdr *th = tcp_hdr(skb);
	struct tcphdr *t1;
L
Linus Torvalds 已提交
799
	struct sk_buff *buff;
800
	struct flowi6 fl6;
801
	struct net *net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
802
	struct sock *ctl_sk = net->ipv6.tcp_sk;
803
	unsigned int tot_len = sizeof(struct tcphdr);
E
Eric Dumazet 已提交
804
	struct dst_entry *dst;
805
	__be32 *topt;
L
Linus Torvalds 已提交
806

807
	if (tsecr)
808
		tot_len += TCPOLEN_TSTAMP_ALIGNED;
809 810 811 812 813 814
#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
815
			 GFP_ATOMIC);
816
	if (!buff)
817
		return;
L
Linus Torvalds 已提交
818

819
	skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
L
Linus Torvalds 已提交
820

821
	t1 = skb_push(buff, tot_len);
822
	skb_reset_transport_header(buff);
L
Linus Torvalds 已提交
823 824 825 826 827

	/* Swap the send and the receive. */
	memset(t1, 0, sizeof(*t1));
	t1->dest = th->source;
	t1->source = th->dest;
828
	t1->doff = tot_len / 4;
829 830 831 832 833
	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
834

835 836
	topt = (__be32 *)(t1 + 1);

837
	if (tsecr) {
838 839
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
840 841
		*topt++ = htonl(tsval);
		*topt++ = htonl(tsecr);
842 843
	}

844 845
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
846 847 848
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
		tcp_v6_md5_hash_hdr((__u8 *)topt, key,
849 850
				    &ipv6_hdr(skb)->saddr,
				    &ipv6_hdr(skb)->daddr, t1);
851 852 853
	}
#endif

854
	memset(&fl6, 0, sizeof(fl6));
A
Alexey Dobriyan 已提交
855 856
	fl6.daddr = ipv6_hdr(skb)->saddr;
	fl6.saddr = ipv6_hdr(skb)->daddr;
857
	fl6.flowlabel = label;
L
Linus Torvalds 已提交
858

859 860 861
	buff->ip_summed = CHECKSUM_PARTIAL;
	buff->csum = 0;

862
	__tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
L
Linus Torvalds 已提交
863

864
	fl6.flowi6_proto = IPPROTO_TCP;
865
	if (rt6_need_strict(&fl6.daddr) && !oif)
E
Eric Dumazet 已提交
866
		fl6.flowi6_oif = tcp_v6_iif(skb);
867 868 869 870 871 872
	else {
		if (!oif && netif_index_is_l3_master(net, skb->skb_iif))
			oif = skb->skb_iif;

		fl6.flowi6_oif = oif;
	}
873

874
	fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark);
875 876
	fl6.fl6_dport = t1->dest;
	fl6.fl6_sport = t1->source;
877
	fl6.flowi6_uid = sock_net_uid(net, sk && sk_fullsock(sk) ? sk : NULL);
878
	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
L
Linus Torvalds 已提交
879

880 881 882 883
	/* Pass a socket to ip6_dst_lookup either it is for RST
	 * Underlying function will use this to retrieve the network
	 * namespace
	 */
884
	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
885 886
	if (!IS_ERR(dst)) {
		skb_dst_set(buff, dst);
887
		ip6_xmit(ctl_sk, buff, &fl6, fl6.flowi6_mark, NULL, tclass);
888
		TCP_INC_STATS(net, TCP_MIB_OUTSEGS);
889
		if (rst)
890
			TCP_INC_STATS(net, TCP_MIB_OUTRSTS);
891
		return;
L
Linus Torvalds 已提交
892 893 894 895 896
	}

	kfree_skb(buff);
}

897
static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
898
{
899
	const struct tcphdr *th = tcp_hdr(skb);
900
	u32 seq = 0, ack_seq = 0;
901
	struct tcp_md5sig_key *key = NULL;
902 903 904 905 906 907 908
#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
909
	int oif = 0;
L
Linus Torvalds 已提交
910

911
	if (th->rst)
L
Linus Torvalds 已提交
912 913
		return;

914 915 916 917
	/* 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))
918
		return;
L
Linus Torvalds 已提交
919

920
#ifdef CONFIG_TCP_MD5SIG
921
	rcu_read_lock();
922
	hash_location = tcp_parse_md5sig_option(th);
923
	if (sk && sk_fullsock(sk)) {
924 925
		key = tcp_v6_md5_do_lookup(sk, &ipv6h->saddr);
	} else if (hash_location) {
926 927 928 929 930 931 932 933
		/*
		 * 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),
934 935
					   &tcp_hashinfo, NULL, 0,
					   &ipv6h->saddr,
936
					   th->source, &ipv6h->daddr,
937 938
					   ntohs(th->source), tcp_v6_iif(skb),
					   tcp_v6_sdif(skb));
939
		if (!sk1)
940
			goto out;
941 942 943

		key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr);
		if (!key)
944
			goto out;
945

946
		genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, skb);
947
		if (genhash || memcmp(hash_location, newhash, 16) != 0)
948
			goto out;
949
	}
950 951
#endif

952 953 954 955 956
	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
957

958 959
	if (sk) {
		oif = sk->sk_bound_dev_if;
960 961
		if (sk_fullsock(sk))
			trace_tcp_send_reset(sk, skb);
962 963
	}

964
	tcp_v6_send_response(sk, skb, seq, ack_seq, 0, 0, 0, oif, key, 1, 0, 0);
965 966

#ifdef CONFIG_TCP_MD5SIG
967 968
out:
	rcu_read_unlock();
969
#endif
970
}
L
Linus Torvalds 已提交
971

972
static void tcp_v6_send_ack(const struct sock *sk, struct sk_buff *skb, u32 seq,
973
			    u32 ack, u32 win, u32 tsval, u32 tsecr, int oif,
974
			    struct tcp_md5sig_key *key, u8 tclass,
975
			    __be32 label)
976
{
977 978
	tcp_v6_send_response(sk, skb, seq, ack, win, tsval, tsecr, oif, key, 0,
			     tclass, label);
L
Linus Torvalds 已提交
979 980 981 982
}

static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
983
	struct inet_timewait_sock *tw = inet_twsk(sk);
984
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
985

986
	tcp_v6_send_ack(sk, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
987
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
988
			tcp_time_stamp_raw() + tcptw->tw_ts_offset,
W
Wang Yufen 已提交
989
			tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw),
990
			tw->tw_tclass, cpu_to_be32(tw->tw_flowlabel));
L
Linus Torvalds 已提交
991

992
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
993 994
}

995
static void tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
996
				  struct request_sock *req)
L
Linus Torvalds 已提交
997
{
998 999 1000
	/* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
	 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
	 */
1001 1002 1003 1004 1005
	/* 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:
	 */
1006
	tcp_v6_send_ack(sk, skb, (sk->sk_state == TCP_LISTEN) ?
1007
			tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt,
1008 1009
			tcp_rsk(req)->rcv_nxt,
			req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale,
1010
			tcp_time_stamp_raw() + tcp_rsk(req)->ts_off,
1011
			req->ts_recent, sk->sk_bound_dev_if,
1012
			tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->saddr),
1013
			0, 0);
L
Linus Torvalds 已提交
1014 1015 1016
}


1017
static struct sock *tcp_v6_cookie_check(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
1018
{
1019
#ifdef CONFIG_SYN_COOKIES
1020
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1021

1022
	if (!th->syn)
1023
		sk = cookie_v6_check(sk, skb);
L
Linus Torvalds 已提交
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
#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))
1034
		goto drop;
L
Linus Torvalds 已提交
1035

O
Octavian Purdila 已提交
1036 1037
	return tcp_conn_request(&tcp6_request_sock_ops,
				&tcp_request_sock_ipv6_ops, sk, skb);
L
Linus Torvalds 已提交
1038 1039

drop:
1040
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1041 1042 1043
	return 0; /* don't send reset */
}

1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
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));
}

1054
static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
W
Weilong Chen 已提交
1055
					 struct request_sock *req,
1056 1057 1058
					 struct dst_entry *dst,
					 struct request_sock *req_unhash,
					 bool *own_req)
L
Linus Torvalds 已提交
1059
{
1060
	struct inet_request_sock *ireq;
1061 1062
	struct ipv6_pinfo *newnp;
	const struct ipv6_pinfo *np = inet6_sk(sk);
1063
	struct ipv6_txoptions *opt;
L
Linus Torvalds 已提交
1064 1065 1066 1067
	struct tcp6_sock *newtcp6sk;
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1068 1069 1070
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1071
	struct flowi6 fl6;
L
Linus Torvalds 已提交
1072 1073 1074 1075 1076 1077

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

1078 1079
		newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst,
					     req_unhash, own_req);
L
Linus Torvalds 已提交
1080

1081
		if (!newsk)
L
Linus Torvalds 已提交
1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
			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));

1093
		newnp->saddr = newsk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
1094

1095
		inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
1096
		newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1097 1098 1099 1100
#ifdef CONFIG_TCP_MD5SIG
		newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif

1101
		newnp->ipv6_mc_list = NULL;
1102 1103
		newnp->ipv6_ac_list = NULL;
		newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1104 1105
		newnp->pktoptions  = NULL;
		newnp->opt	   = NULL;
E
Eric Dumazet 已提交
1106
		newnp->mcast_oif   = tcp_v6_iif(skb);
1107
		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
1108
		newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1109 1110
		if (np->repflow)
			newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1111

1112 1113 1114 1115
		/*
		 * 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
1116 1117 1118
		 */

		/* It is tricky place. Until this moment IPv4 tcp
1119
		   worked with IPv6 icsk.icsk_af_ops.
L
Linus Torvalds 已提交
1120 1121
		   Sync it now.
		 */
1122
		tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
1123 1124 1125 1126

		return newsk;
	}

1127
	ireq = inet_rsk(req);
L
Linus Torvalds 已提交
1128 1129 1130 1131

	if (sk_acceptq_is_full(sk))
		goto out_overflow;

1132
	if (!dst) {
1133
		dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_TCP);
1134
		if (!dst)
L
Linus Torvalds 已提交
1135
			goto out;
1136
	}
L
Linus Torvalds 已提交
1137 1138

	newsk = tcp_create_openreq_child(sk, req, skb);
1139
	if (!newsk)
1140
		goto out_nonewsk;
L
Linus Torvalds 已提交
1141

1142 1143 1144 1145 1146
	/*
	 * 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
1147

1148
	newsk->sk_gso_type = SKB_GSO_TCPV6;
E
Eric Dumazet 已提交
1149
	ip6_dst_store(newsk, dst, NULL, NULL);
1150
	inet6_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1151 1152 1153 1154 1155 1156 1157 1158 1159 1160

	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));

1161 1162 1163 1164
	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
1165

1166
	/* Now IPv6 options...
L
Linus Torvalds 已提交
1167 1168 1169

	   First: no IPv4 options.
	 */
1170
	newinet->inet_opt = NULL;
1171
	newnp->ipv6_mc_list = NULL;
1172
	newnp->ipv6_ac_list = NULL;
1173
	newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1174 1175 1176 1177 1178 1179

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

	newnp->pktoptions = NULL;
	newnp->opt	  = NULL;
E
Eric Dumazet 已提交
1180
	newnp->mcast_oif  = tcp_v6_iif(skb);
1181
	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1182
	newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1183 1184
	if (np->repflow)
		newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1185 1186 1187 1188 1189 1190 1191

	/* 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.
	 */
1192 1193 1194
	opt = ireq->ipv6_opt;
	if (!opt)
		opt = rcu_dereference(np->opt);
1195 1196 1197 1198
	if (opt) {
		opt = ipv6_dup_options(newsk, opt);
		RCU_INIT_POINTER(newnp->opt, opt);
	}
1199
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
1200 1201 1202
	if (opt)
		inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
						    opt->opt_flen;
L
Linus Torvalds 已提交
1203

1204 1205
	tcp_ca_openreq_child(newsk, dst);

L
Linus Torvalds 已提交
1206
	tcp_sync_mss(newsk, dst_mtu(dst));
E
Eric Dumazet 已提交
1207
	newtp->advmss = tcp_mss_clamp(tcp_sk(sk), dst_metric_advmss(dst));
1208

L
Linus Torvalds 已提交
1209 1210
	tcp_initialize_rcv_mss(newsk);

E
Eric Dumazet 已提交
1211 1212
	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
1213

1214 1215
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
W
Wang Yufen 已提交
1216
	key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr);
1217
	if (key) {
1218 1219 1220 1221 1222
		/* 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.
		 */
1223
		tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
1224
			       AF_INET6, 128, key->key, key->keylen,
1225
			       sk_gfp_mask(sk, GFP_ATOMIC));
1226 1227 1228
	}
#endif

1229
	if (__inet_inherit_port(sk, newsk) < 0) {
1230 1231
		inet_csk_prepare_forced_close(newsk);
		tcp_done(newsk);
1232 1233
		goto out;
	}
1234
	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
E
Eric Dumazet 已提交
1235
	if (*own_req) {
1236
		tcp_move_syn(newtp, req);
E
Eric Dumazet 已提交
1237 1238 1239 1240

		/* Clone pktoptions received with SYN, if we own the req */
		if (ireq->pktopts) {
			newnp->pktoptions = skb_clone(ireq->pktopts,
1241
						      sk_gfp_mask(sk, GFP_ATOMIC));
E
Eric Dumazet 已提交
1242 1243
			consume_skb(ireq->pktopts);
			ireq->pktopts = NULL;
1244 1245
			if (newnp->pktoptions) {
				tcp_v6_restore_cb(newnp->pktoptions);
E
Eric Dumazet 已提交
1246
				skb_set_owner_r(newnp->pktoptions, newsk);
1247
			}
E
Eric Dumazet 已提交
1248
		}
E
Eric Dumazet 已提交
1249
	}
L
Linus Torvalds 已提交
1250 1251 1252 1253

	return newsk;

out_overflow:
1254
	__NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1255
out_nonewsk:
L
Linus Torvalds 已提交
1256
	dst_release(dst);
1257
out:
1258
	tcp_listendrop(sk);
L
Linus Torvalds 已提交
1259 1260 1261 1262
	return NULL;
}

/* The socket must have it's spinlock held when we get
1263
 * here, unless it is a TCP_LISTEN socket.
L
Linus Torvalds 已提交
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302
 *
 * 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.
1303
					       --ANK (980728)
L
Linus Torvalds 已提交
1304 1305
	 */
	if (np->rxopt.all)
1306
		opt_skb = skb_clone(skb, sk_gfp_mask(sk, GFP_ATOMIC));
L
Linus Torvalds 已提交
1307 1308

	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
E
Eric Dumazet 已提交
1309 1310
		struct dst_entry *dst = sk->sk_rx_dst;

1311
		sock_rps_save_rxhash(sk, skb);
1312
		sk_mark_napi_id(sk, skb);
E
Eric Dumazet 已提交
1313 1314 1315 1316 1317 1318 1319 1320
		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;
			}
		}

1321
		tcp_rcv_established(sk, skb, tcp_hdr(skb));
L
Linus Torvalds 已提交
1322 1323 1324 1325 1326
		if (opt_skb)
			goto ipv6_pktoptions;
		return 0;
	}

E
Eric Dumazet 已提交
1327
	if (tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1328 1329
		goto csum_err;

1330
	if (sk->sk_state == TCP_LISTEN) {
1331 1332
		struct sock *nsk = tcp_v6_cookie_check(sk, skb);

L
Linus Torvalds 已提交
1333 1334 1335
		if (!nsk)
			goto discard;

W
Weilong Chen 已提交
1336
		if (nsk != sk) {
L
Linus Torvalds 已提交
1337 1338 1339 1340 1341 1342
			if (tcp_child_process(sk, nsk, skb))
				goto reset;
			if (opt_skb)
				__kfree_skb(opt_skb);
			return 0;
		}
1343
	} else
1344
		sock_rps_save_rxhash(sk, skb);
L
Linus Torvalds 已提交
1345

1346
	if (tcp_rcv_state_process(sk, skb))
L
Linus Torvalds 已提交
1347 1348 1349 1350 1351 1352
		goto reset;
	if (opt_skb)
		goto ipv6_pktoptions;
	return 0;

reset:
1353
	tcp_v6_send_reset(sk, skb);
L
Linus Torvalds 已提交
1354 1355 1356 1357 1358 1359
discard:
	if (opt_skb)
		__kfree_skb(opt_skb);
	kfree_skb(skb);
	return 0;
csum_err:
1360 1361
	TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS);
	TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375
	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))) {
1376
		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
E
Eric Dumazet 已提交
1377
			np->mcast_oif = tcp_v6_iif(opt_skb);
1378
		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1379
			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1380
		if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
1381
			np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
1382 1383
		if (np->repflow)
			np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
1384
		if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->header.h6)) {
L
Linus Torvalds 已提交
1385
			skb_set_owner_r(opt_skb, sk);
1386
			tcp_v6_restore_cb(opt_skb);
L
Linus Torvalds 已提交
1387 1388 1389 1390 1391 1392 1393
			opt_skb = xchg(&np->pktoptions, opt_skb);
		} else {
			__kfree_skb(opt_skb);
			opt_skb = xchg(&np->pktoptions, NULL);
		}
	}

1394
	kfree_skb(opt_skb);
L
Linus Torvalds 已提交
1395 1396 1397
	return 0;
}

1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
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;
1418 1419
	TCP_SKB_CB(skb)->has_rxtstamp =
			skb->tstamp || skb_hwtstamps(skb)->hwtstamp;
1420 1421
}

1422
static int tcp_v6_rcv(struct sk_buff *skb)
L
Linus Torvalds 已提交
1423
{
1424
	int sdif = inet6_sdif(skb);
1425
	const struct tcphdr *th;
1426
	const struct ipv6hdr *hdr;
1427
	bool refcounted;
L
Linus Torvalds 已提交
1428 1429
	struct sock *sk;
	int ret;
1430
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
1431 1432 1433 1434 1435 1436 1437

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

	/*
	 *	Count it even if it's bad.
	 */
E
Eric Dumazet 已提交
1438
	__TCP_INC_STATS(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1439 1440 1441 1442

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

1443
	th = (const struct tcphdr *)skb->data;
L
Linus Torvalds 已提交
1444

1445
	if (unlikely(th->doff < sizeof(struct tcphdr)/4))
L
Linus Torvalds 已提交
1446 1447 1448 1449
		goto bad_packet;
	if (!pskb_may_pull(skb, th->doff*4))
		goto discard_it;

1450
	if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo))
1451
		goto csum_error;
L
Linus Torvalds 已提交
1452

1453
	th = (const struct tcphdr *)skb->data;
1454
	hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
1455

1456
lookup:
1457
	sk = __inet6_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th),
1458
				th->source, th->dest, inet6_iif(skb), sdif,
1459
				&refcounted);
L
Linus Torvalds 已提交
1460 1461 1462 1463 1464 1465 1466
	if (!sk)
		goto no_tcp_socket;

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

1467 1468
	if (sk->sk_state == TCP_NEW_SYN_RECV) {
		struct request_sock *req = inet_reqsk(sk);
1469
		bool req_stolen = false;
1470
		struct sock *nsk;
1471 1472 1473

		sk = req->rsk_listener;
		if (tcp_v6_inbound_md5_hash(sk, skb)) {
1474
			sk_drops_add(sk, skb);
1475 1476 1477
			reqsk_put(req);
			goto discard_it;
		}
1478
		if (unlikely(sk->sk_state != TCP_LISTEN)) {
1479
			inet_csk_reqsk_queue_drop_and_put(sk, req);
1480 1481
			goto lookup;
		}
1482
		sock_hold(sk);
1483
		refcounted = true;
E
Eric Dumazet 已提交
1484
		nsk = NULL;
1485 1486 1487 1488
		if (!tcp_filter(sk, skb)) {
			th = (const struct tcphdr *)skb->data;
			hdr = ipv6_hdr(skb);
			tcp_v6_fill_cb(skb, hdr, th);
1489
			nsk = tcp_check_req(sk, skb, req, false, &req_stolen);
1490
		}
1491 1492
		if (!nsk) {
			reqsk_put(req);
1493 1494 1495 1496 1497 1498 1499 1500 1501 1502
			if (req_stolen) {
				/* Another cpu got exclusive access to req
				 * and created a full blown socket.
				 * Try to feed this packet to this socket
				 * instead of discarding it.
				 */
				tcp_v6_restore_cb(skb);
				sock_put(sk);
				goto lookup;
			}
1503
			goto discard_and_relse;
1504 1505 1506 1507 1508 1509
		}
		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);
1510
			goto discard_and_relse;
1511
		} else {
1512
			sock_put(sk);
1513 1514 1515
			return 0;
		}
	}
1516
	if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) {
1517
		__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
1518 1519 1520
		goto discard_and_relse;
	}

L
Linus Torvalds 已提交
1521 1522 1523
	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;

1524 1525 1526
	if (tcp_v6_inbound_md5_hash(sk, skb))
		goto discard_and_relse;

1527
	if (tcp_filter(sk, skb))
L
Linus Torvalds 已提交
1528
		goto discard_and_relse;
1529 1530
	th = (const struct tcphdr *)skb->data;
	hdr = ipv6_hdr(skb);
1531
	tcp_v6_fill_cb(skb, hdr, th);
L
Linus Torvalds 已提交
1532 1533 1534

	skb->dev = NULL;

1535 1536 1537 1538 1539 1540 1541
	if (sk->sk_state == TCP_LISTEN) {
		ret = tcp_v6_do_rcv(sk, skb);
		goto put_and_return;
	}

	sk_incoming_cpu_update(sk);

1542
	bh_lock_sock_nested(sk);
1543
	tcp_segs_in(tcp_sk(sk), skb);
L
Linus Torvalds 已提交
1544 1545
	ret = 0;
	if (!sock_owned_by_user(sk)) {
F
Florian Westphal 已提交
1546
		ret = tcp_v6_do_rcv(sk, skb);
E
Eric Dumazet 已提交
1547
	} else if (tcp_add_backlog(sk, skb)) {
Z
Zhu Yi 已提交
1548 1549
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1550 1551
	bh_unlock_sock(sk);

1552
put_and_return:
1553 1554
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1555 1556 1557 1558 1559 1560
	return ret ? -1 : 0;

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

1561 1562
	tcp_v6_fill_cb(skb, hdr, th);

E
Eric Dumazet 已提交
1563
	if (tcp_checksum_complete(skb)) {
1564
csum_error:
E
Eric Dumazet 已提交
1565
		__TCP_INC_STATS(net, TCP_MIB_CSUMERRORS);
L
Linus Torvalds 已提交
1566
bad_packet:
E
Eric Dumazet 已提交
1567
		__TCP_INC_STATS(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1568
	} else {
1569
		tcp_v6_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1570 1571 1572 1573 1574 1575 1576
	}

discard_it:
	kfree_skb(skb);
	return 0;

discard_and_relse:
1577
	sk_drops_add(sk, skb);
1578 1579
	if (refcounted)
		sock_put(sk);
L
Linus Torvalds 已提交
1580 1581 1582 1583
	goto discard_it;

do_time_wait:
	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1584
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1585 1586 1587
		goto discard_it;
	}

1588 1589
	tcp_v6_fill_cb(skb, hdr, th);

1590 1591 1592
	if (tcp_checksum_complete(skb)) {
		inet_twsk_put(inet_twsk(sk));
		goto csum_error;
L
Linus Torvalds 已提交
1593 1594
	}

1595
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1596 1597 1598 1599
	case TCP_TW_SYN:
	{
		struct sock *sk2;

1600
		sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1601
					    skb, __tcp_hdrlen(th),
1602
					    &ipv6_hdr(skb)->saddr, th->source,
1603
					    &ipv6_hdr(skb)->daddr,
1604 1605
					    ntohs(th->dest), tcp_v6_iif(skb),
					    sdif);
1606
		if (sk2) {
1607
			struct inet_timewait_sock *tw = inet_twsk(sk);
1608
			inet_twsk_deschedule_put(tw);
L
Linus Torvalds 已提交
1609
			sk = sk2;
1610
			tcp_v6_restore_cb(skb);
1611
			refcounted = false;
L
Linus Torvalds 已提交
1612 1613 1614
			goto process;
		}
	}
1615 1616
		/* to ACK */
		/* fall through */
L
Linus Torvalds 已提交
1617 1618 1619 1620
	case TCP_TW_ACK:
		tcp_v6_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
1621 1622 1623
		tcp_v6_send_reset(sk, skb);
		inet_twsk_deschedule_put(inet_twsk(sk));
		goto discard_it;
W
Wang Yufen 已提交
1624 1625
	case TCP_TW_SUCCESS:
		;
L
Linus Torvalds 已提交
1626 1627 1628 1629
	}
	goto discard_it;
}

E
Eric Dumazet 已提交
1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
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
1648
	/* Note : We use inet6_iif() here, not tcp_v6_iif() */
E
Eric Dumazet 已提交
1649 1650 1651
	sk = __inet6_lookup_established(dev_net(skb->dev), &tcp_hashinfo,
					&hdr->saddr, th->source,
					&hdr->daddr, ntohs(th->dest),
1652
					inet6_iif(skb), inet6_sdif(skb));
E
Eric Dumazet 已提交
1653 1654 1655
	if (sk) {
		skb->sk = sk;
		skb->destructor = sock_edemux;
1656
		if (sk_fullsock(sk)) {
1657
			struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst);
1658

E
Eric Dumazet 已提交
1659
			if (dst)
E
Eric Dumazet 已提交
1660
				dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie);
E
Eric Dumazet 已提交
1661
			if (dst &&
1662
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
E
Eric Dumazet 已提交
1663 1664 1665 1666 1667
				skb_dst_set_noref(skb, dst);
		}
	}
}

1668 1669 1670
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
1671
	.twsk_destructor = tcp_twsk_destructor,
1672 1673
};

1674
static const struct inet_connection_sock_af_ops ipv6_specific = {
1675 1676 1677
	.queue_xmit	   = inet6_csk_xmit,
	.send_check	   = tcp_v6_send_check,
	.rebuild_header	   = inet6_sk_rebuild_header,
E
Eric Dumazet 已提交
1678
	.sk_rx_dst_set	   = inet6_sk_rx_dst_set,
1679 1680 1681
	.conn_request	   = tcp_v6_conn_request,
	.syn_recv_sock	   = tcp_v6_syn_recv_sock,
	.net_header_len	   = sizeof(struct ipv6hdr),
1682
	.net_frag_header_len = sizeof(struct frag_hdr),
1683 1684 1685 1686
	.setsockopt	   = ipv6_setsockopt,
	.getsockopt	   = ipv6_getsockopt,
	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in6),
1687
#ifdef CONFIG_COMPAT
1688 1689
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1690
#endif
1691
	.mtu_reduced	   = tcp_v6_mtu_reduced,
L
Linus Torvalds 已提交
1692 1693
};

1694
#ifdef CONFIG_TCP_MD5SIG
1695
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1696
	.md5_lookup	=	tcp_v6_md5_lookup,
1697
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
1698 1699
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1700
#endif
1701

L
Linus Torvalds 已提交
1702 1703 1704
/*
 *	TCP over IPv4 via INET6 API
 */
1705
static const struct inet_connection_sock_af_ops ipv6_mapped = {
1706 1707 1708
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
1709
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1710 1711 1712 1713 1714 1715 1716
	.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),
1717
#ifdef CONFIG_COMPAT
1718 1719
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1720
#endif
1721
	.mtu_reduced	   = tcp_v4_mtu_reduced,
L
Linus Torvalds 已提交
1722 1723
};

1724
#ifdef CONFIG_TCP_MD5SIG
1725
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1726
	.md5_lookup	=	tcp_v4_md5_lookup,
1727
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1728 1729
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1730
#endif
1731

L
Linus Torvalds 已提交
1732 1733 1734 1735 1736
/* 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)
{
1737
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1738

1739
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1740

1741
	icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
1742

1743
#ifdef CONFIG_TCP_MD5SIG
1744
	tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
1745 1746
#endif

L
Linus Torvalds 已提交
1747 1748 1749
	return 0;
}

1750
static void tcp_v6_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1751 1752
{
	tcp_v4_destroy_sock(sk);
1753
	inet6_destroy_sock(sk);
L
Linus Torvalds 已提交
1754 1755
}

1756
#ifdef CONFIG_PROC_FS
L
Linus Torvalds 已提交
1757
/* Proc filesystem TCPv6 sock list dumping. */
1758
static void get_openreq6(struct seq_file *seq,
E
Eric Dumazet 已提交
1759
			 const struct request_sock *req, int i)
L
Linus Torvalds 已提交
1760
{
1761
	long ttd = req->rsk_timer.expires - jiffies;
1762 1763
	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
1764 1765 1766 1767 1768 1769

	if (ttd < 0)
		ttd = 0;

	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1770
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1771 1772 1773
		   i,
		   src->s6_addr32[0], src->s6_addr32[1],
		   src->s6_addr32[2], src->s6_addr32[3],
E
Eric Dumazet 已提交
1774
		   inet_rsk(req)->ir_num,
L
Linus Torvalds 已提交
1775 1776
		   dest->s6_addr32[0], dest->s6_addr32[1],
		   dest->s6_addr32[2], dest->s6_addr32[3],
1777
		   ntohs(inet_rsk(req)->ir_rmt_port),
L
Linus Torvalds 已提交
1778
		   TCP_SYN_RECV,
W
Weilong Chen 已提交
1779
		   0, 0, /* could print option size, but that is af dependent. */
1780 1781
		   1,   /* timers active (only the expire timer) */
		   jiffies_to_clock_t(ttd),
1782
		   req->num_timeout,
E
Eric Dumazet 已提交
1783 1784
		   from_kuid_munged(seq_user_ns(seq),
				    sock_i_uid(req->rsk_listener)),
1785
		   0,  /* non standard timer */
L
Linus Torvalds 已提交
1786 1787 1788 1789 1790 1791
		   0, /* open_requests have no inode */
		   0, req);
}

static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
{
1792
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1793 1794 1795
	__u16 destp, srcp;
	int timer_active;
	unsigned long timer_expires;
1796 1797
	const struct inet_sock *inet = inet_sk(sp);
	const struct tcp_sock *tp = tcp_sk(sp);
1798
	const struct inet_connection_sock *icsk = inet_csk(sp);
1799
	const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq;
1800 1801
	int rx_queue;
	int state;
L
Linus Torvalds 已提交
1802

1803 1804
	dest  = &sp->sk_v6_daddr;
	src   = &sp->sk_v6_rcv_saddr;
E
Eric Dumazet 已提交
1805 1806
	destp = ntohs(inet->inet_dport);
	srcp  = ntohs(inet->inet_sport);
1807

1808
	if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
1809
	    icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
1810
	    icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
L
Linus Torvalds 已提交
1811
		timer_active	= 1;
1812 1813
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
1814
		timer_active	= 4;
1815
		timer_expires	= icsk->icsk_timeout;
L
Linus Torvalds 已提交
1816 1817 1818 1819 1820 1821 1822 1823
	} else if (timer_pending(&sp->sk_timer)) {
		timer_active	= 2;
		timer_expires	= sp->sk_timer.expires;
	} else {
		timer_active	= 0;
		timer_expires = jiffies;
	}

1824
	state = inet_sk_state_load(sp);
1825 1826 1827 1828 1829 1830 1831 1832
	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
1833 1834
	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1835
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
L
Linus Torvalds 已提交
1836 1837 1838 1839 1840
		   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,
1841 1842 1843
		   state,
		   tp->write_seq - tp->snd_una,
		   rx_queue,
L
Linus Torvalds 已提交
1844
		   timer_active,
1845
		   jiffies_delta_to_clock_t(timer_expires - jiffies),
1846
		   icsk->icsk_retransmits,
1847
		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1848
		   icsk->icsk_probes_out,
L
Linus Torvalds 已提交
1849
		   sock_i_ino(sp),
1850
		   refcount_read(&sp->sk_refcnt), sp,
1851 1852
		   jiffies_to_clock_t(icsk->icsk_rto),
		   jiffies_to_clock_t(icsk->icsk_ack.ato),
W
Weilong Chen 已提交
1853
		   (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
I
Ilpo Järvinen 已提交
1854
		   tp->snd_cwnd,
1855
		   state == TCP_LISTEN ?
1856
			fastopenq->max_qlen :
1857
			(tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)
L
Linus Torvalds 已提交
1858 1859 1860
		   );
}

1861
static void get_timewait6_sock(struct seq_file *seq,
1862
			       struct inet_timewait_sock *tw, int i)
L
Linus Torvalds 已提交
1863
{
1864
	long delta = tw->tw_timer.expires - jiffies;
1865
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1866 1867
	__u16 destp, srcp;

1868 1869
	dest = &tw->tw_v6_daddr;
	src  = &tw->tw_v6_rcv_saddr;
L
Linus Torvalds 已提交
1870 1871 1872 1873 1874
	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
1875
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1876 1877 1878 1879 1880 1881
		   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,
1882
		   3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
1883
		   refcount_read(&tw->tw_refcnt), tw);
L
Linus Torvalds 已提交
1884 1885 1886 1887 1888
}

static int tcp6_seq_show(struct seq_file *seq, void *v)
{
	struct tcp_iter_state *st;
E
Eric Dumazet 已提交
1889
	struct sock *sk = v;
L
Linus Torvalds 已提交
1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901

	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;

1902 1903 1904
	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
1905
		get_openreq6(seq, v, st->num);
1906 1907
	else
		get_tcp6_sock(seq, v, st->num);
L
Linus Torvalds 已提交
1908 1909 1910 1911
out:
	return 0;
}

1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924
static const struct seq_operations tcp6_seq_ops = {
	.show		= tcp6_seq_show,
	.start		= tcp_seq_start,
	.next		= tcp_seq_next,
	.stop		= tcp_seq_stop,
};

static int tcp6_seq_open(struct inode *inode, struct file *file)
{
	return seq_open_net(inode, file, &tcp6_seq_ops,
			  sizeof(struct tcp_iter_state));
}

1925
static const struct file_operations tcp6_afinfo_seq_fops = {
1926
	.open    = tcp6_seq_open,
1927 1928 1929 1930 1931
	.read    = seq_read,
	.llseek  = seq_lseek,
	.release = seq_release_net
};

L
Linus Torvalds 已提交
1932 1933 1934 1935
static struct tcp_seq_afinfo tcp6_seq_afinfo = {
	.family		= AF_INET6,
};

1936
int __net_init tcp6_proc_init(struct net *net)
L
Linus Torvalds 已提交
1937
{
1938 1939 1940 1941
	if (!proc_create_data("tcp6", 0444, net->proc_net,
			&tcp6_afinfo_seq_fops, &tcp6_seq_afinfo))
		return -ENOMEM;
	return 0;
L
Linus Torvalds 已提交
1942 1943
}

1944
void tcp6_proc_exit(struct net *net)
L
Linus Torvalds 已提交
1945
{
1946
	remove_proc_entry("tcp6", net->proc_net);
L
Linus Torvalds 已提交
1947 1948 1949 1950 1951 1952 1953
}
#endif

struct proto tcpv6_prot = {
	.name			= "TCPv6",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
A
Andrey Ignatov 已提交
1954
	.pre_connect		= tcp_v6_pre_connect,
L
Linus Torvalds 已提交
1955 1956
	.connect		= tcp_v6_connect,
	.disconnect		= tcp_disconnect,
1957
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
1958 1959 1960 1961 1962 1963
	.ioctl			= tcp_ioctl,
	.init			= tcp_v6_init_sock,
	.destroy		= tcp_v6_destroy_sock,
	.shutdown		= tcp_shutdown,
	.setsockopt		= tcp_setsockopt,
	.getsockopt		= tcp_getsockopt,
1964
	.keepalive		= tcp_set_keepalive,
L
Linus Torvalds 已提交
1965
	.recvmsg		= tcp_recvmsg,
1966 1967
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
1968
	.backlog_rcv		= tcp_v6_do_rcv,
E
Eric Dumazet 已提交
1969
	.release_cb		= tcp_release_cb,
1970
	.hash			= inet6_hash,
1971 1972
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
1973
	.enter_memory_pressure	= tcp_enter_memory_pressure,
1974
	.leave_memory_pressure	= tcp_leave_memory_pressure,
1975
	.stream_memory_free	= tcp_stream_memory_free,
L
Linus Torvalds 已提交
1976 1977 1978
	.sockets_allocated	= &tcp_sockets_allocated,
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
1979
	.orphan_count		= &tcp_orphan_count,
1980
	.sysctl_mem		= sysctl_tcp_mem,
1981 1982
	.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
1983 1984
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp6_sock),
1985
	.slab_flags		= SLAB_TYPESAFE_BY_RCU,
1986
	.twsk_prot		= &tcp6_timewait_sock_ops,
1987
	.rsk_prot		= &tcp6_request_sock_ops,
1988
	.h.hashinfo		= &tcp_hashinfo,
1989
	.no_autobind		= true,
1990 1991 1992
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
G
Glauber Costa 已提交
1993
#endif
1994
	.diag_destroy		= tcp_abort,
L
Linus Torvalds 已提交
1995 1996
};

1997 1998 1999
/* thinking of making this const? Don't.
 * early_demux can change based on sysctl.
 */
2000
static struct inet6_protocol tcpv6_protocol = {
E
Eric Dumazet 已提交
2001
	.early_demux	=	tcp_v6_early_demux,
2002
	.early_demux_handler =  tcp_v6_early_demux,
L
Linus Torvalds 已提交
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
	.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,
2013 2014
	.flags		=	INET_PROTOSW_PERMANENT |
				INET_PROTOSW_ICSK,
L
Linus Torvalds 已提交
2015 2016
};

2017
static int __net_init tcpv6_net_init(struct net *net)
2018
{
2019 2020
	return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
				    SOCK_RAW, IPPROTO_TCP, net);
2021 2022
}

2023
static void __net_exit tcpv6_net_exit(struct net *net)
2024
{
2025
	inet_ctl_sock_destroy(net->ipv6.tcp_sk);
E
Eric W. Biederman 已提交
2026 2027
}

2028
static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
E
Eric W. Biederman 已提交
2029
{
2030
	inet_twsk_purge(&tcp_hashinfo, AF_INET6);
2031 2032 2033
}

static struct pernet_operations tcpv6_net_ops = {
E
Eric W. Biederman 已提交
2034 2035 2036
	.init	    = tcpv6_net_init,
	.exit	    = tcpv6_net_exit,
	.exit_batch = tcpv6_net_exit_batch,
2037 2038
};

2039
int __init tcpv6_init(void)
L
Linus Torvalds 已提交
2040
{
2041 2042
	int ret;

2043 2044
	ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
	if (ret)
2045
		goto out;
2046

L
Linus Torvalds 已提交
2047
	/* register inet6 protocol */
2048 2049 2050 2051
	ret = inet6_register_protosw(&tcpv6_protosw);
	if (ret)
		goto out_tcpv6_protocol;

2052
	ret = register_pernet_subsys(&tcpv6_net_ops);
2053 2054 2055 2056
	if (ret)
		goto out_tcpv6_protosw;
out:
	return ret;
2057

2058 2059
out_tcpv6_protosw:
	inet6_unregister_protosw(&tcpv6_protosw);
2060 2061
out_tcpv6_protocol:
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2062 2063 2064
	goto out;
}

2065
void tcpv6_exit(void)
2066
{
2067
	unregister_pernet_subsys(&tcpv6_net_ops);
2068 2069
	inet6_unregister_protosw(&tcpv6_protosw);
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
L
Linus Torvalds 已提交
2070
}