tcp_ipv6.c 50.6 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>
G
Glauber Costa 已提交
64
#include <net/tcp_memcontrol.h>
65
#include <net/busy_poll.h>
L
Linus Torvalds 已提交
66 67 68 69

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

70 71 72 73
#include <linux/crypto.h>
#include <linux/scatterlist.h>

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

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

79 80
static const struct inet_connection_sock_af_ops ipv6_mapped;
static const struct inet_connection_sock_af_ops ipv6_specific;
81
#ifdef CONFIG_TCP_MD5SIG
82 83
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
84 85
#else
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
86
						   const struct in6_addr *addr)
87 88 89
{
	return NULL;
}
90
#endif
L
Linus Torvalds 已提交
91

92 93 94 95
static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
{
	struct dst_entry *dst = skb_dst(skb);

96 97 98 99 100 101 102 103 104
	if (dst) {
		const struct rt6_info *rt = (const struct rt6_info *)dst;

		dst_hold(dst);
		sk->sk_rx_dst = dst;
		inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
		if (rt->rt6i_node)
			inet6_sk(sk)->rx_dst_cookie = rt->rt6i_node->fn_sernum;
	}
105 106
}

L
Linus Torvalds 已提交
107 108 109
static void tcp_v6_hash(struct sock *sk)
{
	if (sk->sk_state != TCP_CLOSE) {
110
		if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
L
Linus Torvalds 已提交
111 112 113 114
			tcp_prot.hash(sk);
			return;
		}
		local_bh_disable();
115
		__inet6_hash(sk, NULL);
L
Linus Torvalds 已提交
116 117 118 119
		local_bh_enable();
	}
}

120
static __u32 tcp_v6_init_sequence(const struct sk_buff *skb)
L
Linus Torvalds 已提交
121
{
122 123
	return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
					    ipv6_hdr(skb)->saddr.s6_addr32,
124 125
					    tcp_hdr(skb)->dest,
					    tcp_hdr(skb)->source);
L
Linus Torvalds 已提交
126 127
}

128
static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
L
Linus Torvalds 已提交
129 130 131
			  int addr_len)
{
	struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
132
	struct inet_sock *inet = inet_sk(sk);
133
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
134 135
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
136
	struct in6_addr *saddr = NULL, *final_p, final;
137
	struct rt6_info *rt;
138
	struct flowi6 fl6;
L
Linus Torvalds 已提交
139 140 141 142
	struct dst_entry *dst;
	int addr_type;
	int err;

143
	if (addr_len < SIN6_LEN_RFC2133)
L
Linus Torvalds 已提交
144 145
		return -EINVAL;

146
	if (usin->sin6_family != AF_INET6)
E
Eric Dumazet 已提交
147
		return -EAFNOSUPPORT;
L
Linus Torvalds 已提交
148

149
	memset(&fl6, 0, sizeof(fl6));
L
Linus Torvalds 已提交
150 151

	if (np->sndflow) {
152 153 154
		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
155
			struct ip6_flowlabel *flowlabel;
156
			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
L
Linus Torvalds 已提交
157 158 159 160 161 162 163
			if (flowlabel == NULL)
				return -EINVAL;
			fl6_sock_release(flowlabel);
		}
	}

	/*
164 165 166
	 *	connect() to INADDR_ANY means loopback (BSD'ism).
	 */

W
Weilong Chen 已提交
167
	if (ipv6_addr_any(&usin->sin6_addr))
168
		usin->sin6_addr.s6_addr[15] = 0x1;
L
Linus Torvalds 已提交
169 170 171

	addr_type = ipv6_addr_type(&usin->sin6_addr);

W
Weilong Chen 已提交
172
	if (addr_type & IPV6_ADDR_MULTICAST)
L
Linus Torvalds 已提交
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193
		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.
			 */
			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 &&
194
	    !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) {
L
Linus Torvalds 已提交
195 196 197 198 199
		tp->rx_opt.ts_recent = 0;
		tp->rx_opt.ts_recent_stamp = 0;
		tp->write_seq = 0;
	}

200
	sk->sk_v6_daddr = usin->sin6_addr;
201
	np->flow_label = fl6.flowlabel;
L
Linus Torvalds 已提交
202 203 204 205 206 207

	/*
	 *	TCP over IPv4
	 */

	if (addr_type == IPV6_ADDR_MAPPED) {
208
		u32 exthdrlen = icsk->icsk_ext_hdr_len;
L
Linus Torvalds 已提交
209 210 211 212 213 214 215 216 217 218 219
		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];

220
		icsk->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
221
		sk->sk_backlog_rcv = tcp_v4_do_rcv;
222 223 224
#ifdef CONFIG_TCP_MD5SIG
		tp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif
L
Linus Torvalds 已提交
225 226 227 228

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

		if (err) {
229 230
			icsk->icsk_ext_hdr_len = exthdrlen;
			icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
231
			sk->sk_backlog_rcv = tcp_v6_do_rcv;
232 233 234
#ifdef CONFIG_TCP_MD5SIG
			tp->af_specific = &tcp_sock_ipv6_specific;
#endif
L
Linus Torvalds 已提交
235 236
			goto failure;
		} else {
E
Eric Dumazet 已提交
237 238
			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
239
					       &sk->sk_v6_rcv_saddr);
L
Linus Torvalds 已提交
240 241 242 243 244
		}

		return err;
	}

245 246
	if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
		saddr = &sk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
247

248
	fl6.flowi6_proto = IPPROTO_TCP;
249
	fl6.daddr = sk->sk_v6_daddr;
A
Alexey Dobriyan 已提交
250
	fl6.saddr = saddr ? *saddr : np->saddr;
251 252
	fl6.flowi6_oif = sk->sk_bound_dev_if;
	fl6.flowi6_mark = sk->sk_mark;
253 254
	fl6.fl6_dport = usin->sin6_port;
	fl6.fl6_sport = inet->inet_sport;
L
Linus Torvalds 已提交
255

256
	final_p = fl6_update_dst(&fl6, np->opt, &final);
L
Linus Torvalds 已提交
257

258
	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
V
Venkat Yekkirala 已提交
259

260
	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
261 262
	if (IS_ERR(dst)) {
		err = PTR_ERR(dst);
L
Linus Torvalds 已提交
263
		goto failure;
264
	}
L
Linus Torvalds 已提交
265 266

	if (saddr == NULL) {
267
		saddr = &fl6.saddr;
268
		sk->sk_v6_rcv_saddr = *saddr;
L
Linus Torvalds 已提交
269 270 271
	}

	/* set the source address */
A
Alexey Dobriyan 已提交
272
	np->saddr = *saddr;
E
Eric Dumazet 已提交
273
	inet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
274

H
Herbert Xu 已提交
275
	sk->sk_gso_type = SKB_GSO_TCPV6;
276
	__ip6_dst_store(sk, dst, NULL, NULL);
L
Linus Torvalds 已提交
277

278 279 280
	rt = (struct rt6_info *) dst;
	if (tcp_death_row.sysctl_tw_recycle &&
	    !tp->rx_opt.ts_recent_stamp &&
281
	    ipv6_addr_equal(&rt->rt6i_dst.addr, &sk->sk_v6_daddr))
282
		tcp_fetch_timewait_stamp(sk, dst);
283

284
	icsk->icsk_ext_hdr_len = 0;
L
Linus Torvalds 已提交
285
	if (np->opt)
286 287
		icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
					  np->opt->opt_nflen);
L
Linus Torvalds 已提交
288 289 290

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

E
Eric Dumazet 已提交
291
	inet->inet_dport = usin->sin6_port;
L
Linus Torvalds 已提交
292 293

	tcp_set_state(sk, TCP_SYN_SENT);
294
	err = inet6_hash_connect(&tcp_death_row, sk);
L
Linus Torvalds 已提交
295 296 297
	if (err)
		goto late_failure;

298 299
	ip6_set_txhash(sk);

A
Andrey Vagin 已提交
300
	if (!tp->write_seq && likely(!tp->repair))
L
Linus Torvalds 已提交
301
		tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
302
							     sk->sk_v6_daddr.s6_addr32,
E
Eric Dumazet 已提交
303 304
							     inet->inet_sport,
							     inet->inet_dport);
L
Linus Torvalds 已提交
305 306 307 308 309 310 311 312 313 314 315

	err = tcp_connect(sk);
	if (err)
		goto late_failure;

	return 0;

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

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

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

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

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

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

351
	sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr,
352
			th->dest, &hdr->saddr, th->source, skb->dev->ifindex);
L
Linus Torvalds 已提交
353 354

	if (sk == NULL) {
355 356
		ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
				   ICMP6_MIB_INERRORS);
L
Linus Torvalds 已提交
357 358 359 360
		return;
	}

	if (sk->sk_state == TCP_TIME_WAIT) {
361
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
362 363 364 365
		return;
	}

	bh_lock_sock(sk);
366
	if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG)
367
		NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
L
Linus Torvalds 已提交
368 369 370 371

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

372 373 374 375 376
	if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) {
		NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
		goto out;
	}

L
Linus Torvalds 已提交
377
	tp = tcp_sk(sk);
378
	seq = ntohl(th->seq);
379 380 381
	/* 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
382
	if (sk->sk_state != TCP_LISTEN &&
383
	    !between(seq, snd_una, tp->snd_nxt)) {
384
		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
385 386 387 388 389
		goto out;
	}

	np = inet6_sk(sk);

390 391 392
	if (type == NDISC_REDIRECT) {
		struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);

393
		if (dst)
394
			dst->ops->redirect(dst, sk, skb);
395
		goto out;
396 397
	}

L
Linus Torvalds 已提交
398
	if (type == ICMPV6_PKT_TOOBIG) {
399 400 401 402 403 404 405
		/* 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;

406 407 408
		if (!ip6_sk_accept_pmtu(sk))
			goto out;

409 410 411
		tp->mtu_info = ntohl(info);
		if (!sock_owned_by_user(sk))
			tcp_v6_mtu_reduced(sk);
412 413 414
		else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED,
					   &tp->tsq_flags))
			sock_hold(sk);
L
Linus Torvalds 已提交
415 416 417 418 419
		goto out;
	}

	icmpv6_err_convert(type, code, &err);

420
	/* Might be for an request_sock */
L
Linus Torvalds 已提交
421
	switch (sk->sk_state) {
422
		struct request_sock *req, **prev;
L
Linus Torvalds 已提交
423 424 425 426
	case TCP_LISTEN:
		if (sock_owned_by_user(sk))
			goto out;

E
Eric Dumazet 已提交
427
		/* Note : We use inet6_iif() here, not tcp_v6_iif() */
428 429
		req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
					   &hdr->saddr, inet6_iif(skb));
L
Linus Torvalds 已提交
430 431 432 433 434 435
		if (!req)
			goto out;

		/* ICMPs are not backlogged, hence we cannot get
		 * an established socket here.
		 */
436
		WARN_ON(req->sk != NULL);
L
Linus Torvalds 已提交
437

438
		if (seq != tcp_rsk(req)->snt_isn) {
439
			NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
440 441 442
			goto out;
		}

443
		inet_csk_reqsk_queue_drop(sk, req, prev);
444
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
445 446 447
		goto out;

	case TCP_SYN_SENT:
448 449 450 451 452 453 454
	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.
		 */
		if (fastopen && fastopen->sk == NULL)
			break;

L
Linus Torvalds 已提交
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476
		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);
}


477
static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst,
478
			      struct flowi *fl,
479
			      struct request_sock *req,
480 481
			      u16 queue_mapping,
			      struct tcp_fastopen_cookie *foc)
L
Linus Torvalds 已提交
482
{
483
	struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
484
	struct ipv6_pinfo *np = inet6_sk(sk);
485
	struct flowi6 *fl6 = &fl->u.ip6;
W
Weilong Chen 已提交
486
	struct sk_buff *skb;
487
	int err = -ENOMEM;
L
Linus Torvalds 已提交
488

489 490
	/* First, grab a route. */
	if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL)
491
		goto done;
492

493
	skb = tcp_make_synack(sk, dst, req, foc);
494

L
Linus Torvalds 已提交
495
	if (skb) {
496 497
		__tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr,
				    &ireq->ir_v6_rmt_addr);
L
Linus Torvalds 已提交
498

499
		fl6->daddr = ireq->ir_v6_rmt_addr;
500 501 502
		if (np->repflow && (ireq->pktopts != NULL))
			fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts));

503
		skb_set_queue_mapping(skb, queue_mapping);
504
		err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass);
505
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
506 507 508 509 510 511
	}

done:
	return err;
}

512

513
static void tcp_v6_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
514
{
515
	kfree_skb(inet_rsk(req)->pktopts);
L
Linus Torvalds 已提交
516 517
}

518 519
#ifdef CONFIG_TCP_MD5SIG
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
520
						   const struct in6_addr *addr)
521
{
E
Eric Dumazet 已提交
522
	return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
523 524 525 526 527
}

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

static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
						      struct request_sock *req)
{
534
	return tcp_v6_md5_do_lookup(sk, &inet_rsk(req)->ir_v6_rmt_addr);
535 536
}

W
Wang Yufen 已提交
537 538
static int tcp_v6_parse_md5_keys(struct sock *sk, char __user *optval,
				 int optlen)
539 540 541 542 543 544 545 546 547 548 549 550 551 552
{
	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
553
		if (ipv6_addr_v4mapped(&sin6->sin6_addr))
E
Eric Dumazet 已提交
554 555 556 557
			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);
558 559 560 561 562
	}

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

E
Eric Dumazet 已提交
563 564 565
	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);
566

E
Eric Dumazet 已提交
567 568
	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
			      AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
569 570
}

571
static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
572 573
					const struct in6_addr *daddr,
					const struct in6_addr *saddr, int nbytes)
574 575
{
	struct tcp6_pseudohdr *bp;
576
	struct scatterlist sg;
577

578 579
	bp = &hp->md5_blk.ip6;
	/* 1. TCP pseudo-header (RFC2460) */
A
Alexey Dobriyan 已提交
580 581
	bp->saddr = *saddr;
	bp->daddr = *daddr;
582
	bp->protocol = cpu_to_be32(IPPROTO_TCP);
A
Adam Langley 已提交
583
	bp->len = cpu_to_be32(nbytes);
584

585 586 587
	sg_init_one(&sg, bp, sizeof(*bp));
	return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
}
588

589
static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
590
			       const struct in6_addr *daddr, struct in6_addr *saddr,
E
Eric Dumazet 已提交
591
			       const struct tcphdr *th)
592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609
{
	struct tcp_md5sig_pool *hp;
	struct hash_desc *desc;

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

	if (crypto_hash_init(desc))
		goto clear_hash;
	if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2))
		goto clear_hash;
	if (tcp_md5_hash_header(hp, th))
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
	if (crypto_hash_final(desc, md5_hash))
610 611 612 613
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
614

615 616 617 618
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
619
	return 1;
620 621
}

622
static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
623 624 625
			       const struct sock *sk,
			       const struct request_sock *req,
			       const struct sk_buff *skb)
626
{
627
	const struct in6_addr *saddr, *daddr;
628 629
	struct tcp_md5sig_pool *hp;
	struct hash_desc *desc;
E
Eric Dumazet 已提交
630
	const struct tcphdr *th = tcp_hdr(skb);
631 632 633

	if (sk) {
		saddr = &inet6_sk(sk)->saddr;
634
		daddr = &sk->sk_v6_daddr;
635
	} else if (req) {
636 637
		saddr = &inet_rsk(req)->ir_v6_loc_addr;
		daddr = &inet_rsk(req)->ir_v6_rmt_addr;
638
	} else {
639
		const struct ipv6hdr *ip6h = ipv6_hdr(skb);
640 641
		saddr = &ip6h->saddr;
		daddr = &ip6h->daddr;
642
	}
643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670

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

	if (crypto_hash_init(desc))
		goto clear_hash;

	if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len))
		goto clear_hash;
	if (tcp_md5_hash_header(hp, th))
		goto clear_hash;
	if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
		goto clear_hash;
	if (tcp_md5_hash_key(hp, key))
		goto clear_hash;
	if (crypto_hash_final(desc, md5_hash))
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;

clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
	return 1;
671 672
}

673 674
static int __tcp_v6_inbound_md5_hash(struct sock *sk,
				     const struct sk_buff *skb)
675
{
676
	const __u8 *hash_location = NULL;
677
	struct tcp_md5sig_key *hash_expected;
678
	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
E
Eric Dumazet 已提交
679
	const struct tcphdr *th = tcp_hdr(skb);
680 681 682 683
	int genhash;
	u8 newhash[16];

	hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
684
	hash_location = tcp_parse_md5sig_option(th);
685

686 687 688 689 690 691
	/* We've parsed the options - do we have a hash? */
	if (!hash_expected && !hash_location)
		return 0;

	if (hash_expected && !hash_location) {
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
692 693 694
		return 1;
	}

695 696
	if (!hash_expected && hash_location) {
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
697 698 699 700
		return 1;
	}

	/* check the signature */
701 702 703 704
	genhash = tcp_v6_md5_hash_skb(newhash,
				      hash_expected,
				      NULL, NULL, skb);

705
	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
706 707 708 709
		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));
710 711 712 713
		return 1;
	}
	return 0;
}
714 715 716 717 718 719 720 721 722 723 724 725

static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
{
	int ret;

	rcu_read_lock();
	ret = __tcp_v6_inbound_md5_hash(sk, skb);
	rcu_read_unlock();

	return ret;
}

726 727
#endif

728 729 730 731 732 733 734 735 736 737 738 739 740 741
static void tcp_v6_init_req(struct request_sock *req, struct sock *sk,
			    struct sk_buff *skb)
{
	struct inet_request_sock *ireq = inet_rsk(req);
	struct ipv6_pinfo *np = inet6_sk(sk);

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

	ireq->ir_iif = sk->sk_bound_dev_if;

	/* So that link locals have meaning */
	if (!sk->sk_bound_dev_if &&
	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
E
Eric Dumazet 已提交
742
		ireq->ir_iif = tcp_v6_iif(skb);
743

744
	if (!TCP_SKB_CB(skb)->tcp_tw_isn &&
745 746
	    (ipv6_opt_accepted(sk, skb, &TCP_SKB_CB(skb)->header.h6) ||
	     np->rxopt.bits.rxinfo ||
747 748 749 750 751 752 753
	     np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim ||
	     np->rxopt.bits.rxohlim || np->repflow)) {
		atomic_inc(&skb->users);
		ireq->pktopts = skb;
	}
}

754 755 756 757 758 759 760 761 762
static struct dst_entry *tcp_v6_route_req(struct sock *sk, struct flowi *fl,
					  const struct request_sock *req,
					  bool *strict)
{
	if (strict)
		*strict = true;
	return inet6_csk_route_req(sk, &fl->u.ip6, req);
}

763
struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
764
	.family		=	AF_INET6,
765
	.obj_size	=	sizeof(struct tcp6_request_sock),
766
	.rtx_syn_ack	=	tcp_rtx_synack,
767 768
	.send_ack	=	tcp_v6_reqsk_send_ack,
	.destructor	=	tcp_v6_reqsk_destructor,
769
	.send_reset	=	tcp_v6_send_reset,
W
Wang Yufen 已提交
770
	.syn_ack_timeout =	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
771 772
};

773
static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
774 775
	.mss_clamp	=	IPV6_MIN_MTU - sizeof(struct tcphdr) -
				sizeof(struct ipv6hdr),
776
#ifdef CONFIG_TCP_MD5SIG
777
	.md5_lookup	=	tcp_v6_reqsk_md5_lookup,
778
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
779
#endif
780
	.init_req	=	tcp_v6_init_req,
781 782 783
#ifdef CONFIG_SYN_COOKIES
	.cookie_init_seq =	cookie_v6_init_sequence,
#endif
784
	.route_req	=	tcp_v6_route_req,
785
	.init_seq	=	tcp_v6_init_sequence,
786
	.send_synack	=	tcp_v6_send_synack,
787
	.queue_hash_add =	inet6_csk_reqsk_queue_hash_add,
788
};
789

790 791 792 793
static void tcp_v6_send_response(struct sock *sk, struct sk_buff *skb, u32 seq,
				 u32 ack, u32 win, u32 tsval, u32 tsecr,
				 int oif, struct tcp_md5sig_key *key, int rst,
				 u8 tclass, u32 label)
L
Linus Torvalds 已提交
794
{
795 796
	const struct tcphdr *th = tcp_hdr(skb);
	struct tcphdr *t1;
L
Linus Torvalds 已提交
797
	struct sk_buff *buff;
798
	struct flowi6 fl6;
799
	struct net *net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
800
	struct sock *ctl_sk = net->ipv6.tcp_sk;
801
	unsigned int tot_len = sizeof(struct tcphdr);
E
Eric Dumazet 已提交
802
	struct dst_entry *dst;
803
	__be32 *topt;
L
Linus Torvalds 已提交
804

805
	if (tsecr)
806
		tot_len += TCPOLEN_TSTAMP_ALIGNED;
807 808 809 810 811 812
#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
813
			 GFP_ATOMIC);
814 815
	if (buff == NULL)
		return;
L
Linus Torvalds 已提交
816

817
	skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
L
Linus Torvalds 已提交
818

819
	t1 = (struct tcphdr *) skb_push(buff, tot_len);
820
	skb_reset_transport_header(buff);
L
Linus Torvalds 已提交
821 822 823 824 825

	/* Swap the send and the receive. */
	memset(t1, 0, sizeof(*t1));
	t1->dest = th->source;
	t1->source = th->dest;
826
	t1->doff = tot_len / 4;
827 828 829 830 831
	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
832

833 834
	topt = (__be32 *)(t1 + 1);

835
	if (tsecr) {
836 837
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
838 839
		*topt++ = htonl(tsval);
		*topt++ = htonl(tsecr);
840 841
	}

842 843
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
844 845 846
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
		tcp_v6_md5_hash_hdr((__u8 *)topt, key,
847 848
				    &ipv6_hdr(skb)->saddr,
				    &ipv6_hdr(skb)->daddr, t1);
849 850 851
	}
#endif

852
	memset(&fl6, 0, sizeof(fl6));
A
Alexey Dobriyan 已提交
853 854
	fl6.daddr = ipv6_hdr(skb)->saddr;
	fl6.saddr = ipv6_hdr(skb)->daddr;
855
	fl6.flowlabel = label;
L
Linus Torvalds 已提交
856

857 858 859
	buff->ip_summed = CHECKSUM_PARTIAL;
	buff->csum = 0;

860
	__tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
L
Linus Torvalds 已提交
861

862
	fl6.flowi6_proto = IPPROTO_TCP;
863
	if (rt6_need_strict(&fl6.daddr) && !oif)
E
Eric Dumazet 已提交
864
		fl6.flowi6_oif = tcp_v6_iif(skb);
W
Wang Yufen 已提交
865 866
	else
		fl6.flowi6_oif = oif;
867
	fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark);
868 869
	fl6.fl6_dport = t1->dest;
	fl6.fl6_sport = t1->source;
870
	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
L
Linus Torvalds 已提交
871

872 873 874 875
	/* Pass a socket to ip6_dst_lookup either it is for RST
	 * Underlying function will use this to retrieve the network
	 * namespace
	 */
876
	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
877 878
	if (!IS_ERR(dst)) {
		skb_dst_set(buff, dst);
879
		ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass);
880 881 882 883
		TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
		if (rst)
			TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS);
		return;
L
Linus Torvalds 已提交
884 885 886 887 888
	}

	kfree_skb(buff);
}

889
static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
890
{
891
	const struct tcphdr *th = tcp_hdr(skb);
892
	u32 seq = 0, ack_seq = 0;
893
	struct tcp_md5sig_key *key = NULL;
894 895 896 897 898 899 900
#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
901
	int oif;
L
Linus Torvalds 已提交
902

903
	if (th->rst)
L
Linus Torvalds 已提交
904 905
		return;

906 907 908 909
	/* 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))
910
		return;
L
Linus Torvalds 已提交
911

912
#ifdef CONFIG_TCP_MD5SIG
913 914 915 916 917 918 919 920 921 922
	hash_location = tcp_parse_md5sig_option(th);
	if (!sk && hash_location) {
		/*
		 * active side is lost. Try to find listening socket through
		 * source port, and then find md5 key through listening socket.
		 * we are not loose security here:
		 * Incoming packet is checked with md5 hash with finding key,
		 * no RST generated if md5 hash doesn't match.
		 */
		sk1 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev),
923 924
					   &tcp_hashinfo, &ipv6h->saddr,
					   th->source, &ipv6h->daddr,
E
Eric Dumazet 已提交
925
					   ntohs(th->source), tcp_v6_iif(skb));
926 927 928 929 930 931 932 933 934 935 936 937 938 939
		if (!sk1)
			return;

		rcu_read_lock();
		key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr);
		if (!key)
			goto release_sk1;

		genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, NULL, skb);
		if (genhash || memcmp(hash_location, newhash, 16) != 0)
			goto release_sk1;
	} else {
		key = sk ? tcp_v6_md5_do_lookup(sk, &ipv6h->saddr) : NULL;
	}
940 941
#endif

942 943 944 945 946
	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
947

W
Wang Yufen 已提交
948
	oif = sk ? sk->sk_bound_dev_if : 0;
949
	tcp_v6_send_response(sk, skb, seq, ack_seq, 0, 0, 0, oif, key, 1, 0, 0);
950 951 952 953 954 955 956 957

#ifdef CONFIG_TCP_MD5SIG
release_sk1:
	if (sk1) {
		rcu_read_unlock();
		sock_put(sk1);
	}
#endif
958
}
L
Linus Torvalds 已提交
959

960 961
static void tcp_v6_send_ack(struct sock *sk, struct sk_buff *skb, u32 seq,
			    u32 ack, u32 win, u32 tsval, u32 tsecr, int oif,
962 963
			    struct tcp_md5sig_key *key, u8 tclass,
			    u32 label)
964
{
965 966
	tcp_v6_send_response(sk, skb, seq, ack, win, tsval, tsecr, oif, key, 0,
			     tclass, label);
L
Linus Torvalds 已提交
967 968 969 970
}

static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
971
	struct inet_timewait_sock *tw = inet_twsk(sk);
972
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
973

974
	tcp_v6_send_ack(sk, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
975
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
976
			tcp_time_stamp + tcptw->tw_ts_offset,
W
Wang Yufen 已提交
977
			tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw),
978
			tw->tw_tclass, (tw->tw_flowlabel << 12));
L
Linus Torvalds 已提交
979

980
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
981 982
}

983 984
static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
				  struct request_sock *req)
L
Linus Torvalds 已提交
985
{
986 987 988
	/* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
	 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
	 */
989
	tcp_v6_send_ack(sk, skb, (sk->sk_state == TCP_LISTEN) ?
990
			tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt,
991 992
			tcp_rsk(req)->rcv_nxt, req->rcv_wnd,
			tcp_time_stamp, req->ts_recent, sk->sk_bound_dev_if,
993 994
			tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr),
			0, 0);
L
Linus Torvalds 已提交
995 996 997
}


W
Weilong Chen 已提交
998
static struct sock *tcp_v6_hnd_req(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
999
{
1000
	struct request_sock *req, **prev;
1001
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1002 1003 1004
	struct sock *nsk;

	/* Find possible connection requests. */
1005
	req = inet6_csk_search_req(sk, &prev, th->source,
1006
				   &ipv6_hdr(skb)->saddr,
E
Eric Dumazet 已提交
1007
				   &ipv6_hdr(skb)->daddr, tcp_v6_iif(skb));
L
Linus Torvalds 已提交
1008
	if (req)
1009
		return tcp_check_req(sk, skb, req, prev, false);
L
Linus Torvalds 已提交
1010

1011
	nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
E
Eric Dumazet 已提交
1012 1013 1014
					 &ipv6_hdr(skb)->saddr, th->source,
					 &ipv6_hdr(skb)->daddr, ntohs(th->dest),
					 tcp_v6_iif(skb));
L
Linus Torvalds 已提交
1015 1016 1017 1018 1019 1020

	if (nsk) {
		if (nsk->sk_state != TCP_TIME_WAIT) {
			bh_lock_sock(nsk);
			return nsk;
		}
1021
		inet_twsk_put(inet_twsk(nsk));
L
Linus Torvalds 已提交
1022 1023 1024
		return NULL;
	}

1025
#ifdef CONFIG_SYN_COOKIES
1026
	if (!th->syn)
1027
		sk = cookie_v6_check(sk, skb);
L
Linus Torvalds 已提交
1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
#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))
1038
		goto drop;
L
Linus Torvalds 已提交
1039

O
Octavian Purdila 已提交
1040 1041
	return tcp_conn_request(&tcp6_request_sock_ops,
				&tcp_request_sock_ipv6_ops, sk, skb);
L
Linus Torvalds 已提交
1042 1043

drop:
1044
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1045 1046 1047
	return 0; /* don't send reset */
}

W
Weilong Chen 已提交
1048 1049 1050
static struct sock *tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
					 struct request_sock *req,
					 struct dst_entry *dst)
L
Linus Torvalds 已提交
1051
{
1052
	struct inet_request_sock *ireq;
L
Linus Torvalds 已提交
1053 1054 1055 1056 1057
	struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
	struct tcp6_sock *newtcp6sk;
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1058 1059 1060
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1061
	struct flowi6 fl6;
L
Linus Torvalds 已提交
1062 1063 1064 1065 1066 1067 1068 1069

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

		newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);

1070
		if (newsk == NULL)
L
Linus Torvalds 已提交
1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081
			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));

1082
		ipv6_addr_set_v4mapped(newinet->inet_daddr, &newsk->sk_v6_daddr);
L
Linus Torvalds 已提交
1083

E
Eric Dumazet 已提交
1084
		ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr);
L
Linus Torvalds 已提交
1085

1086
		newsk->sk_v6_rcv_saddr = newnp->saddr;
L
Linus Torvalds 已提交
1087

1088
		inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
1089
		newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1090 1091 1092 1093
#ifdef CONFIG_TCP_MD5SIG
		newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif

1094 1095
		newnp->ipv6_ac_list = NULL;
		newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1096 1097
		newnp->pktoptions  = NULL;
		newnp->opt	   = NULL;
E
Eric Dumazet 已提交
1098
		newnp->mcast_oif   = tcp_v6_iif(skb);
1099
		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
1100
		newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1101 1102
		if (np->repflow)
			newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1103

1104 1105 1106 1107
		/*
		 * 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
1108 1109 1110
		 */

		/* It is tricky place. Until this moment IPv4 tcp
1111
		   worked with IPv6 icsk.icsk_af_ops.
L
Linus Torvalds 已提交
1112 1113
		   Sync it now.
		 */
1114
		tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
1115 1116 1117 1118

		return newsk;
	}

1119
	ireq = inet_rsk(req);
L
Linus Torvalds 已提交
1120 1121 1122 1123

	if (sk_acceptq_is_full(sk))
		goto out_overflow;

1124
	if (!dst) {
1125
		dst = inet6_csk_route_req(sk, &fl6, req);
1126
		if (!dst)
L
Linus Torvalds 已提交
1127
			goto out;
1128
	}
L
Linus Torvalds 已提交
1129 1130 1131

	newsk = tcp_create_openreq_child(sk, req, skb);
	if (newsk == NULL)
1132
		goto out_nonewsk;
L
Linus Torvalds 已提交
1133

1134 1135 1136 1137 1138
	/*
	 * 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
1139

1140
	newsk->sk_gso_type = SKB_GSO_TCPV6;
1141
	__ip6_dst_store(newsk, dst, NULL, NULL);
1142
	inet6_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152

	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));

1153 1154 1155 1156
	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
1157

1158 1159
	ip6_set_txhash(newsk);

1160
	/* Now IPv6 options...
L
Linus Torvalds 已提交
1161 1162 1163

	   First: no IPv4 options.
	 */
1164
	newinet->inet_opt = NULL;
1165
	newnp->ipv6_ac_list = NULL;
1166
	newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1167 1168 1169 1170 1171 1172

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

	/* Clone pktoptions received with SYN */
	newnp->pktoptions = NULL;
1173 1174
	if (ireq->pktopts != NULL) {
		newnp->pktoptions = skb_clone(ireq->pktopts,
1175
					      sk_gfp_atomic(sk, GFP_ATOMIC));
1176 1177
		consume_skb(ireq->pktopts);
		ireq->pktopts = NULL;
L
Linus Torvalds 已提交
1178 1179 1180 1181
		if (newnp->pktoptions)
			skb_set_owner_r(newnp->pktoptions, newsk);
	}
	newnp->opt	  = NULL;
E
Eric Dumazet 已提交
1182
	newnp->mcast_oif  = tcp_v6_iif(skb);
1183
	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1184
	newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1185 1186
	if (np->repflow)
		newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1187 1188 1189 1190 1191 1192 1193

	/* 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.
	 */
1194 1195
	if (np->opt)
		newnp->opt = ipv6_dup_options(newsk, np->opt);
L
Linus Torvalds 已提交
1196

1197
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
L
Linus Torvalds 已提交
1198
	if (newnp->opt)
1199 1200
		inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
						     newnp->opt->opt_flen);
L
Linus Torvalds 已提交
1201

1202 1203
	tcp_ca_openreq_child(newsk, dst);

L
Linus Torvalds 已提交
1204
	tcp_sync_mss(newsk, dst_mtu(dst));
1205
	newtp->advmss = dst_metric_advmss(dst);
1206 1207 1208 1209
	if (tcp_sk(sk)->rx_opt.user_mss &&
	    tcp_sk(sk)->rx_opt.user_mss < newtp->advmss)
		newtp->advmss = tcp_sk(sk)->rx_opt.user_mss;

L
Linus Torvalds 已提交
1210 1211
	tcp_initialize_rcv_mss(newsk);

E
Eric Dumazet 已提交
1212 1213
	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
1214

1215 1216
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
W
Wang Yufen 已提交
1217 1218
	key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr);
	if (key != NULL) {
1219 1220 1221 1222 1223
		/* 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.
		 */
1224
		tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
1225 1226
			       AF_INET6, key->key, key->keylen,
			       sk_gfp_atomic(sk, GFP_ATOMIC));
1227 1228 1229
	}
#endif

1230
	if (__inet_inherit_port(sk, newsk) < 0) {
1231 1232
		inet_csk_prepare_forced_close(newsk);
		tcp_done(newsk);
1233 1234
		goto out;
	}
1235
	__inet6_hash(newsk, NULL);
L
Linus Torvalds 已提交
1236 1237 1238 1239

	return newsk;

out_overflow:
1240
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1241
out_nonewsk:
L
Linus Torvalds 已提交
1242
	dst_release(dst);
1243 1244
out:
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272
	return NULL;
}

/* The socket must have it's spinlock held when we get
 * here.
 *
 * We have a potential double-lock case here, so even when
 * doing backlog processing we use the BH locking scheme.
 * This is because we cannot sleep with the original spinlock
 * held.
 */
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);

1273
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291
		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.
1292
					       --ANK (980728)
L
Linus Torvalds 已提交
1293 1294
	 */
	if (np->rxopt.all)
1295
		opt_skb = skb_clone(skb, sk_gfp_atomic(sk, GFP_ATOMIC));
L
Linus Torvalds 已提交
1296 1297

	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
E
Eric Dumazet 已提交
1298 1299
		struct dst_entry *dst = sk->sk_rx_dst;

1300
		sock_rps_save_rxhash(sk, skb);
1301
		sk_mark_napi_id(sk, skb);
E
Eric Dumazet 已提交
1302 1303 1304 1305 1306 1307 1308 1309
		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;
			}
		}

1310
		tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
L
Linus Torvalds 已提交
1311 1312 1313 1314 1315
		if (opt_skb)
			goto ipv6_pktoptions;
		return 0;
	}

1316
	if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1317 1318
		goto csum_err;

1319
	if (sk->sk_state == TCP_LISTEN) {
L
Linus Torvalds 已提交
1320 1321 1322 1323 1324 1325 1326 1327 1328
		struct sock *nsk = tcp_v6_hnd_req(sk, skb);
		if (!nsk)
			goto discard;

		/*
		 * Queue it on the new socket if the new socket is active,
		 * otherwise we just shortcircuit this and continue with
		 * the new socket..
		 */
W
Weilong Chen 已提交
1329
		if (nsk != sk) {
1330
			sock_rps_save_rxhash(nsk, skb);
1331
			sk_mark_napi_id(sk, skb);
L
Linus Torvalds 已提交
1332 1333 1334 1335 1336 1337
			if (tcp_child_process(sk, nsk, skb))
				goto reset;
			if (opt_skb)
				__kfree_skb(opt_skb);
			return 0;
		}
1338
	} else
1339
		sock_rps_save_rxhash(sk, skb);
L
Linus Torvalds 已提交
1340

1341
	if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
L
Linus Torvalds 已提交
1342 1343 1344 1345 1346 1347
		goto reset;
	if (opt_skb)
		goto ipv6_pktoptions;
	return 0;

reset:
1348
	tcp_v6_send_reset(sk, skb);
L
Linus Torvalds 已提交
1349 1350 1351 1352 1353 1354
discard:
	if (opt_skb)
		__kfree_skb(opt_skb);
	kfree_skb(skb);
	return 0;
csum_err:
1355
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS);
1356
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370
	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))) {
1371
		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
E
Eric Dumazet 已提交
1372
			np->mcast_oif = tcp_v6_iif(opt_skb);
1373
		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1374
			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1375
		if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
1376
			np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
1377 1378
		if (np->repflow)
			np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
1379
		if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->header.h6)) {
L
Linus Torvalds 已提交
1380 1381 1382 1383 1384 1385 1386 1387
			skb_set_owner_r(opt_skb, sk);
			opt_skb = xchg(&np->pktoptions, opt_skb);
		} else {
			__kfree_skb(opt_skb);
			opt_skb = xchg(&np->pktoptions, NULL);
		}
	}

1388
	kfree_skb(opt_skb);
L
Linus Torvalds 已提交
1389 1390 1391
	return 0;
}

1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413
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;
}

1414
static int tcp_v6_rcv(struct sk_buff *skb)
L
Linus Torvalds 已提交
1415
{
1416
	const struct tcphdr *th;
1417
	const struct ipv6hdr *hdr;
L
Linus Torvalds 已提交
1418 1419
	struct sock *sk;
	int ret;
1420
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
1421 1422 1423 1424 1425 1426 1427

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

	/*
	 *	Count it even if it's bad.
	 */
1428
	TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1429 1430 1431 1432

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

1433
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1434 1435 1436 1437 1438 1439

	if (th->doff < sizeof(struct tcphdr)/4)
		goto bad_packet;
	if (!pskb_may_pull(skb, th->doff*4))
		goto discard_it;

1440
	if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo))
1441
		goto csum_error;
L
Linus Torvalds 已提交
1442

1443
	th = tcp_hdr(skb);
1444
	hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
1445

E
Eric Dumazet 已提交
1446
	sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest,
1447
				inet6_iif(skb));
L
Linus Torvalds 已提交
1448 1449 1450 1451 1452 1453 1454
	if (!sk)
		goto no_tcp_socket;

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

1455 1456 1457 1458 1459
	if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) {
		NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
		goto discard_and_relse;
	}

L
Linus Torvalds 已提交
1460 1461 1462
	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;

1463 1464
	tcp_v6_fill_cb(skb, hdr, th);

1465 1466 1467 1468 1469
#ifdef CONFIG_TCP_MD5SIG
	if (tcp_v6_inbound_md5_hash(sk, skb))
		goto discard_and_relse;
#endif

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

E
Eric Dumazet 已提交
1473
	sk_incoming_cpu_update(sk);
L
Linus Torvalds 已提交
1474 1475
	skb->dev = NULL;

1476
	bh_lock_sock_nested(sk);
L
Linus Torvalds 已提交
1477 1478
	ret = 0;
	if (!sock_owned_by_user(sk)) {
D
Dan Williams 已提交
1479
		if (!tcp_prequeue(sk, skb))
1480
			ret = tcp_v6_do_rcv(sk, skb);
1481 1482
	} else if (unlikely(sk_add_backlog(sk, skb,
					   sk->sk_rcvbuf + sk->sk_sndbuf))) {
Z
Zhu Yi 已提交
1483
		bh_unlock_sock(sk);
1484
		NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
Z
Zhu Yi 已提交
1485 1486
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1487 1488 1489 1490 1491 1492 1493 1494 1495
	bh_unlock_sock(sk);

	sock_put(sk);
	return ret ? -1 : 0;

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

1496 1497
	tcp_v6_fill_cb(skb, hdr, th);

L
Linus Torvalds 已提交
1498
	if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1499 1500
csum_error:
		TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS);
L
Linus Torvalds 已提交
1501
bad_packet:
1502
		TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1503
	} else {
1504
		tcp_v6_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516
	}

discard_it:
	kfree_skb(skb);
	return 0;

discard_and_relse:
	sock_put(sk);
	goto discard_it;

do_time_wait:
	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1517
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1518 1519 1520
		goto discard_it;
	}

1521 1522
	tcp_v6_fill_cb(skb, hdr, th);

1523
	if (skb->len < (th->doff<<2)) {
1524
		inet_twsk_put(inet_twsk(sk));
1525 1526 1527 1528 1529
		goto bad_packet;
	}
	if (tcp_checksum_complete(skb)) {
		inet_twsk_put(inet_twsk(sk));
		goto csum_error;
L
Linus Torvalds 已提交
1530 1531
	}

1532
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1533 1534 1535 1536
	case TCP_TW_SYN:
	{
		struct sock *sk2;

1537
		sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1538
					    &ipv6_hdr(skb)->saddr, th->source,
1539
					    &ipv6_hdr(skb)->daddr,
E
Eric Dumazet 已提交
1540
					    ntohs(th->dest), tcp_v6_iif(skb));
L
Linus Torvalds 已提交
1541
		if (sk2 != NULL) {
1542 1543 1544
			struct inet_timewait_sock *tw = inet_twsk(sk);
			inet_twsk_deschedule(tw, &tcp_death_row);
			inet_twsk_put(tw);
L
Linus Torvalds 已提交
1545 1546 1547 1548 1549 1550 1551 1552 1553 1554
			sk = sk2;
			goto process;
		}
		/* Fall through to ACK */
	}
	case TCP_TW_ACK:
		tcp_v6_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
		goto no_tcp_socket;
W
Wang Yufen 已提交
1555 1556
	case TCP_TW_SUCCESS:
		;
L
Linus Torvalds 已提交
1557 1558 1559 1560
	}
	goto discard_it;
}

E
Eric Dumazet 已提交
1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578
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
1579
	/* Note : We use inet6_iif() here, not tcp_v6_iif() */
E
Eric Dumazet 已提交
1580 1581 1582 1583 1584 1585 1586 1587 1588
	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;
		if (sk->sk_state != TCP_TIME_WAIT) {
			struct dst_entry *dst = sk->sk_rx_dst;
1589

E
Eric Dumazet 已提交
1590
			if (dst)
E
Eric Dumazet 已提交
1591
				dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie);
E
Eric Dumazet 已提交
1592
			if (dst &&
1593
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
E
Eric Dumazet 已提交
1594 1595 1596 1597 1598
				skb_dst_set_noref(skb, dst);
		}
	}
}

1599 1600 1601
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
1602
	.twsk_destructor = tcp_twsk_destructor,
1603 1604
};

1605
static const struct inet_connection_sock_af_ops ipv6_specific = {
1606 1607 1608
	.queue_xmit	   = inet6_csk_xmit,
	.send_check	   = tcp_v6_send_check,
	.rebuild_header	   = inet6_sk_rebuild_header,
E
Eric Dumazet 已提交
1609
	.sk_rx_dst_set	   = inet6_sk_rx_dst_set,
1610 1611 1612
	.conn_request	   = tcp_v6_conn_request,
	.syn_recv_sock	   = tcp_v6_syn_recv_sock,
	.net_header_len	   = sizeof(struct ipv6hdr),
1613
	.net_frag_header_len = sizeof(struct frag_hdr),
1614 1615 1616 1617
	.setsockopt	   = ipv6_setsockopt,
	.getsockopt	   = ipv6_getsockopt,
	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in6),
1618
	.bind_conflict	   = inet6_csk_bind_conflict,
1619
#ifdef CONFIG_COMPAT
1620 1621
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1622
#endif
1623
	.mtu_reduced	   = tcp_v6_mtu_reduced,
L
Linus Torvalds 已提交
1624 1625
};

1626
#ifdef CONFIG_TCP_MD5SIG
1627
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1628
	.md5_lookup	=	tcp_v6_md5_lookup,
1629
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
1630 1631
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1632
#endif
1633

L
Linus Torvalds 已提交
1634 1635 1636
/*
 *	TCP over IPv4 via INET6 API
 */
1637
static const struct inet_connection_sock_af_ops ipv6_mapped = {
1638 1639 1640
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
1641
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1642 1643 1644 1645 1646 1647 1648
	.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),
1649
	.bind_conflict	   = inet6_csk_bind_conflict,
1650
#ifdef CONFIG_COMPAT
1651 1652
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1653
#endif
1654
	.mtu_reduced	   = tcp_v4_mtu_reduced,
L
Linus Torvalds 已提交
1655 1656
};

1657
#ifdef CONFIG_TCP_MD5SIG
1658
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1659
	.md5_lookup	=	tcp_v4_md5_lookup,
1660
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1661 1662
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1663
#endif
1664

L
Linus Torvalds 已提交
1665 1666 1667 1668 1669
/* 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)
{
1670
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1671

1672
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1673

1674
	icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
1675

1676
#ifdef CONFIG_TCP_MD5SIG
1677
	tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
1678 1679
#endif

L
Linus Torvalds 已提交
1680 1681 1682
	return 0;
}

1683
static void tcp_v6_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1684 1685
{
	tcp_v4_destroy_sock(sk);
1686
	inet6_destroy_sock(sk);
L
Linus Torvalds 已提交
1687 1688
}

1689
#ifdef CONFIG_PROC_FS
L
Linus Torvalds 已提交
1690
/* Proc filesystem TCPv6 sock list dumping. */
1691
static void get_openreq6(struct seq_file *seq,
1692
			 const struct sock *sk, struct request_sock *req, int i, kuid_t uid)
L
Linus Torvalds 已提交
1693 1694
{
	int ttd = req->expires - jiffies;
1695 1696
	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
1697 1698 1699 1700 1701 1702

	if (ttd < 0)
		ttd = 0;

	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1703
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1704 1705 1706
		   i,
		   src->s6_addr32[0], src->s6_addr32[1],
		   src->s6_addr32[2], src->s6_addr32[3],
E
Eric Dumazet 已提交
1707
		   inet_rsk(req)->ir_num,
L
Linus Torvalds 已提交
1708 1709
		   dest->s6_addr32[0], dest->s6_addr32[1],
		   dest->s6_addr32[2], dest->s6_addr32[3],
1710
		   ntohs(inet_rsk(req)->ir_rmt_port),
L
Linus Torvalds 已提交
1711
		   TCP_SYN_RECV,
W
Weilong Chen 已提交
1712
		   0, 0, /* could print option size, but that is af dependent. */
1713 1714
		   1,   /* timers active (only the expire timer) */
		   jiffies_to_clock_t(ttd),
1715
		   req->num_timeout,
1716
		   from_kuid_munged(seq_user_ns(seq), uid),
1717
		   0,  /* non standard timer */
L
Linus Torvalds 已提交
1718 1719 1720 1721 1722 1723
		   0, /* open_requests have no inode */
		   0, req);
}

static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
{
1724
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1725 1726 1727
	__u16 destp, srcp;
	int timer_active;
	unsigned long timer_expires;
1728 1729
	const struct inet_sock *inet = inet_sk(sp);
	const struct tcp_sock *tp = tcp_sk(sp);
1730
	const struct inet_connection_sock *icsk = inet_csk(sp);
1731
	struct fastopen_queue *fastopenq = icsk->icsk_accept_queue.fastopenq;
L
Linus Torvalds 已提交
1732

1733 1734
	dest  = &sp->sk_v6_daddr;
	src   = &sp->sk_v6_rcv_saddr;
E
Eric Dumazet 已提交
1735 1736
	destp = ntohs(inet->inet_dport);
	srcp  = ntohs(inet->inet_sport);
1737 1738

	if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
L
Linus Torvalds 已提交
1739
		timer_active	= 1;
1740 1741
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
1742
		timer_active	= 4;
1743
		timer_expires	= icsk->icsk_timeout;
L
Linus Torvalds 已提交
1744 1745 1746 1747 1748 1749 1750 1751 1752 1753
	} else if (timer_pending(&sp->sk_timer)) {
		timer_active	= 2;
		timer_expires	= sp->sk_timer.expires;
	} else {
		timer_active	= 0;
		timer_expires = jiffies;
	}

	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1754
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
L
Linus Torvalds 已提交
1755 1756 1757 1758 1759
		   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,
1760
		   sp->sk_state,
1761 1762
		   tp->write_seq-tp->snd_una,
		   (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
L
Linus Torvalds 已提交
1763
		   timer_active,
1764
		   jiffies_delta_to_clock_t(timer_expires - jiffies),
1765
		   icsk->icsk_retransmits,
1766
		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1767
		   icsk->icsk_probes_out,
L
Linus Torvalds 已提交
1768 1769
		   sock_i_ino(sp),
		   atomic_read(&sp->sk_refcnt), sp,
1770 1771
		   jiffies_to_clock_t(icsk->icsk_rto),
		   jiffies_to_clock_t(icsk->icsk_ack.ato),
W
Weilong Chen 已提交
1772
		   (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
I
Ilpo Järvinen 已提交
1773
		   tp->snd_cwnd,
1774 1775 1776
		   sp->sk_state == TCP_LISTEN ?
			(fastopenq ? fastopenq->max_qlen : 0) :
			(tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)
L
Linus Torvalds 已提交
1777 1778 1779
		   );
}

1780
static void get_timewait6_sock(struct seq_file *seq,
1781
			       struct inet_timewait_sock *tw, int i)
L
Linus Torvalds 已提交
1782
{
1783
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1784
	__u16 destp, srcp;
1785
	s32 delta = tw->tw_ttd - inet_tw_time_stamp();
L
Linus Torvalds 已提交
1786

1787 1788
	dest = &tw->tw_v6_daddr;
	src  = &tw->tw_v6_rcv_saddr;
L
Linus Torvalds 已提交
1789 1790 1791 1792 1793
	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
1794
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1795 1796 1797 1798 1799 1800
		   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,
1801
		   3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
L
Linus Torvalds 已提交
1802 1803 1804 1805 1806 1807
		   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
1808
	struct sock *sk = v;
L
Linus Torvalds 已提交
1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823

	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;

	switch (st->state) {
	case TCP_SEQ_STATE_LISTENING:
	case TCP_SEQ_STATE_ESTABLISHED:
E
Eric Dumazet 已提交
1824 1825 1826 1827
		if (sk->sk_state == TCP_TIME_WAIT)
			get_timewait6_sock(seq, v, st->num);
		else
			get_tcp6_sock(seq, v, st->num);
L
Linus Torvalds 已提交
1828 1829 1830 1831 1832 1833 1834 1835 1836
		break;
	case TCP_SEQ_STATE_OPENREQ:
		get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
		break;
	}
out:
	return 0;
}

1837 1838 1839 1840 1841 1842 1843 1844
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
1845 1846 1847
static struct tcp_seq_afinfo tcp6_seq_afinfo = {
	.name		= "tcp6",
	.family		= AF_INET6,
1848
	.seq_fops	= &tcp6_afinfo_seq_fops,
1849 1850 1851
	.seq_ops	= {
		.show		= tcp6_seq_show,
	},
L
Linus Torvalds 已提交
1852 1853
};

1854
int __net_init tcp6_proc_init(struct net *net)
L
Linus Torvalds 已提交
1855
{
1856
	return tcp_proc_register(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1857 1858
}

1859
void tcp6_proc_exit(struct net *net)
L
Linus Torvalds 已提交
1860
{
1861
	tcp_proc_unregister(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1862 1863 1864
}
#endif

E
Eric Dumazet 已提交
1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875
static void tcp_v6_clear_sk(struct sock *sk, int size)
{
	struct inet_sock *inet = inet_sk(sk);

	/* we do not want to clear pinet6 field, because of RCU lookups */
	sk_prot_clear_nulls(sk, offsetof(struct inet_sock, pinet6));

	size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
	memset(&inet->pinet6 + 1, 0, size);
}

L
Linus Torvalds 已提交
1876 1877 1878 1879 1880 1881
struct proto tcpv6_prot = {
	.name			= "TCPv6",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v6_connect,
	.disconnect		= tcp_disconnect,
1882
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
1883 1884 1885 1886 1887 1888 1889
	.ioctl			= tcp_ioctl,
	.init			= tcp_v6_init_sock,
	.destroy		= tcp_v6_destroy_sock,
	.shutdown		= tcp_shutdown,
	.setsockopt		= tcp_setsockopt,
	.getsockopt		= tcp_getsockopt,
	.recvmsg		= tcp_recvmsg,
1890 1891
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
1892
	.backlog_rcv		= tcp_v6_do_rcv,
E
Eric Dumazet 已提交
1893
	.release_cb		= tcp_release_cb,
L
Linus Torvalds 已提交
1894
	.hash			= tcp_v6_hash,
1895 1896
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
1897
	.enter_memory_pressure	= tcp_enter_memory_pressure,
1898
	.stream_memory_free	= tcp_stream_memory_free,
L
Linus Torvalds 已提交
1899 1900 1901
	.sockets_allocated	= &tcp_sockets_allocated,
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
1902
	.orphan_count		= &tcp_orphan_count,
1903
	.sysctl_mem		= sysctl_tcp_mem,
L
Linus Torvalds 已提交
1904 1905 1906 1907
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp6_sock),
1908
	.slab_flags		= SLAB_DESTROY_BY_RCU,
1909
	.twsk_prot		= &tcp6_timewait_sock_ops,
1910
	.rsk_prot		= &tcp6_request_sock_ops,
1911
	.h.hashinfo		= &tcp_hashinfo,
1912
	.no_autobind		= true,
1913 1914 1915 1916
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
#endif
A
Andrew Morton 已提交
1917
#ifdef CONFIG_MEMCG_KMEM
G
Glauber Costa 已提交
1918 1919
	.proto_cgroup		= tcp_proto_cgroup,
#endif
E
Eric Dumazet 已提交
1920
	.clear_sk		= tcp_v6_clear_sk,
L
Linus Torvalds 已提交
1921 1922
};

1923
static const struct inet6_protocol tcpv6_protocol = {
E
Eric Dumazet 已提交
1924
	.early_demux	=	tcp_v6_early_demux,
L
Linus Torvalds 已提交
1925 1926 1927 1928 1929 1930 1931 1932 1933 1934
	.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,
1935 1936
	.flags		=	INET_PROTOSW_PERMANENT |
				INET_PROTOSW_ICSK,
L
Linus Torvalds 已提交
1937 1938
};

1939
static int __net_init tcpv6_net_init(struct net *net)
1940
{
1941 1942
	return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
				    SOCK_RAW, IPPROTO_TCP, net);
1943 1944
}

1945
static void __net_exit tcpv6_net_exit(struct net *net)
1946
{
1947
	inet_ctl_sock_destroy(net->ipv6.tcp_sk);
E
Eric W. Biederman 已提交
1948 1949
}

1950
static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
E
Eric W. Biederman 已提交
1951 1952
{
	inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6);
1953 1954 1955
}

static struct pernet_operations tcpv6_net_ops = {
E
Eric W. Biederman 已提交
1956 1957 1958
	.init	    = tcpv6_net_init,
	.exit	    = tcpv6_net_exit,
	.exit_batch = tcpv6_net_exit_batch,
1959 1960
};

1961
int __init tcpv6_init(void)
L
Linus Torvalds 已提交
1962
{
1963 1964
	int ret;

1965 1966
	ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
	if (ret)
1967
		goto out;
1968

L
Linus Torvalds 已提交
1969
	/* register inet6 protocol */
1970 1971 1972 1973
	ret = inet6_register_protosw(&tcpv6_protosw);
	if (ret)
		goto out_tcpv6_protocol;

1974
	ret = register_pernet_subsys(&tcpv6_net_ops);
1975 1976 1977 1978
	if (ret)
		goto out_tcpv6_protosw;
out:
	return ret;
1979

1980 1981
out_tcpv6_protosw:
	inet6_unregister_protosw(&tcpv6_protosw);
1982 1983
out_tcpv6_protocol:
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
1984 1985 1986
	goto out;
}

1987
void tcpv6_exit(void)
1988
{
1989
	unregister_pernet_subsys(&tcpv6_net_ops);
1990 1991
	inet6_unregister_protosw(&tcpv6_protosw);
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
L
Linus Torvalds 已提交
1992
}