tcp_ipv6.c 51.2 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>
L
Linus Torvalds 已提交
42 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/netdma.h>
63
#include <net/inet_common.h>
64
#include <net/secure_seq.h>
G
Glauber Costa 已提交
65
#include <net/tcp_memcontrol.h>
66
#include <net/busy_poll.h>
L
Linus Torvalds 已提交
67 68 69 70 71 72

#include <asm/uaccess.h>

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

73 74 75 76
#include <linux/crypto.h>
#include <linux/scatterlist.h>

static void	tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb);
77 78
static void	tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
				      struct request_sock *req);
L
Linus Torvalds 已提交
79 80 81

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

82 83
static const struct inet_connection_sock_af_ops ipv6_mapped;
static const struct inet_connection_sock_af_ops ipv6_specific;
84
#ifdef CONFIG_TCP_MD5SIG
85 86
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
87 88
#else
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
89
						   const struct in6_addr *addr)
90 91 92
{
	return NULL;
}
93
#endif
L
Linus Torvalds 已提交
94

95 96 97 98 99 100 101 102 103 104 105 106
static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
{
	struct dst_entry *dst = skb_dst(skb);
	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;
}

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;

A
Andrey Vagin 已提交
298
	if (!tp->write_seq && likely(!tp->repair))
L
Linus Torvalds 已提交
299
		tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
300
							     sk->sk_v6_daddr.s6_addr32,
E
Eric Dumazet 已提交
301 302
							     inet->inet_sport,
							     inet->inet_dport);
L
Linus Torvalds 已提交
303 304 305 306 307 308 309 310 311 312 313

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

	return 0;

late_failure:
	tcp_set_state(sk, TCP_CLOSE);
	__sk_dst_reset(sk);
failure:
E
Eric Dumazet 已提交
314
	inet->inet_dport = 0;
L
Linus Torvalds 已提交
315 316 317 318
	sk->sk_route_caps = 0;
	return err;
}

319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335
static void tcp_v6_mtu_reduced(struct sock *sk)
{
	struct dst_entry *dst;

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

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

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

L
Linus Torvalds 已提交
336
static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
337
		u8 type, u8 code, int offset, __be32 info)
L
Linus Torvalds 已提交
338
{
W
Weilong Chen 已提交
339
	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
340
	const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
L
Linus Torvalds 已提交
341 342 343
	struct ipv6_pinfo *np;
	struct sock *sk;
	int err;
344
	struct tcp_sock *tp;
L
Linus Torvalds 已提交
345
	__u32 seq;
346
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
347

348
	sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr,
349
			th->dest, &hdr->saddr, th->source, skb->dev->ifindex);
L
Linus Torvalds 已提交
350 351

	if (sk == NULL) {
352 353
		ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
				   ICMP6_MIB_INERRORS);
L
Linus Torvalds 已提交
354 355 356 357
		return;
	}

	if (sk->sk_state == TCP_TIME_WAIT) {
358
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
359 360 361 362
		return;
	}

	bh_lock_sock(sk);
363
	if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG)
364
		NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
L
Linus Torvalds 已提交
365 366 367 368

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

369 370 371 372 373
	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
374
	tp = tcp_sk(sk);
375
	seq = ntohl(th->seq);
L
Linus Torvalds 已提交
376 377
	if (sk->sk_state != TCP_LISTEN &&
	    !between(seq, tp->snd_una, tp->snd_nxt)) {
378
		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
379 380 381 382 383
		goto out;
	}

	np = inet6_sk(sk);

384 385 386
	if (type == NDISC_REDIRECT) {
		struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);

387
		if (dst)
388
			dst->ops->redirect(dst, sk, skb);
389
		goto out;
390 391
	}

L
Linus Torvalds 已提交
392
	if (type == ICMPV6_PKT_TOOBIG) {
393 394 395 396 397 398 399
		/* 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;

400 401 402
		if (!ip6_sk_accept_pmtu(sk))
			goto out;

403 404 405
		tp->mtu_info = ntohl(info);
		if (!sock_owned_by_user(sk))
			tcp_v6_mtu_reduced(sk);
406 407 408
		else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED,
					   &tp->tsq_flags))
			sock_hold(sk);
L
Linus Torvalds 已提交
409 410 411 412 413
		goto out;
	}

	icmpv6_err_convert(type, code, &err);

414
	/* Might be for an request_sock */
L
Linus Torvalds 已提交
415
	switch (sk->sk_state) {
416
		struct request_sock *req, **prev;
L
Linus Torvalds 已提交
417 418 419 420
	case TCP_LISTEN:
		if (sock_owned_by_user(sk))
			goto out;

421 422
		req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
					   &hdr->saddr, inet6_iif(skb));
L
Linus Torvalds 已提交
423 424 425 426 427 428
		if (!req)
			goto out;

		/* ICMPs are not backlogged, hence we cannot get
		 * an established socket here.
		 */
429
		WARN_ON(req->sk != NULL);
L
Linus Torvalds 已提交
430

431
		if (seq != tcp_rsk(req)->snt_isn) {
432
			NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
433 434 435
			goto out;
		}

436
		inet_csk_reqsk_queue_drop(sk, req, prev);
437
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
438 439 440 441
		goto out;

	case TCP_SYN_SENT:
	case TCP_SYN_RECV:  /* Cannot happen.
442
			       It can, it SYNs are crossed. --ANK */
L
Linus Torvalds 已提交
443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464
		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);
}


465 466
static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst,
			      struct flowi6 *fl6,
467
			      struct request_sock *req,
468
			      u16 queue_mapping)
L
Linus Torvalds 已提交
469
{
470
	struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
471
	struct ipv6_pinfo *np = inet6_sk(sk);
W
Weilong Chen 已提交
472
	struct sk_buff *skb;
473
	int err = -ENOMEM;
L
Linus Torvalds 已提交
474

475 476
	/* First, grab a route. */
	if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL)
477
		goto done;
478

C
Christoph Paasch 已提交
479
	skb = tcp_make_synack(sk, dst, req, NULL);
480

L
Linus Torvalds 已提交
481
	if (skb) {
482 483
		__tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr,
				    &ireq->ir_v6_rmt_addr);
L
Linus Torvalds 已提交
484

485
		fl6->daddr = ireq->ir_v6_rmt_addr;
486
		skb_set_queue_mapping(skb, queue_mapping);
487
		err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass);
488
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
489 490 491 492 493 494
	}

done:
	return err;
}

C
Christoph Paasch 已提交
495
static int tcp_v6_rtx_synack(struct sock *sk, struct request_sock *req)
496
{
497
	struct flowi6 fl6;
498
	int res;
499

C
Christoph Paasch 已提交
500
	res = tcp_v6_send_synack(sk, NULL, &fl6, req, 0);
501 502 503
	if (!res)
		TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS);
	return res;
504 505
}

506
static void tcp_v6_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
507
{
508
	kfree_skb(inet_rsk(req)->pktopts);
L
Linus Torvalds 已提交
509 510
}

511 512
#ifdef CONFIG_TCP_MD5SIG
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
513
						   const struct in6_addr *addr)
514
{
E
Eric Dumazet 已提交
515
	return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
516 517 518 519 520
}

static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
						struct sock *addr_sk)
{
521
	return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr);
522 523 524 525 526
}

static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
						      struct request_sock *req)
{
527
	return tcp_v6_md5_do_lookup(sk, &inet_rsk(req)->ir_v6_rmt_addr);
528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545
}

static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval,
				  int optlen)
{
	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
546
		if (ipv6_addr_v4mapped(&sin6->sin6_addr))
E
Eric Dumazet 已提交
547 548 549 550
			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);
551 552 553 554 555
	}

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

E
Eric Dumazet 已提交
556 557 558
	if (ipv6_addr_v4mapped(&sin6->sin6_addr))
		return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
				      AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
559

E
Eric Dumazet 已提交
560 561
	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
			      AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
562 563
}

564
static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
565 566
					const struct in6_addr *daddr,
					const struct in6_addr *saddr, int nbytes)
567 568
{
	struct tcp6_pseudohdr *bp;
569
	struct scatterlist sg;
570

571 572
	bp = &hp->md5_blk.ip6;
	/* 1. TCP pseudo-header (RFC2460) */
A
Alexey Dobriyan 已提交
573 574
	bp->saddr = *saddr;
	bp->daddr = *daddr;
575
	bp->protocol = cpu_to_be32(IPPROTO_TCP);
A
Adam Langley 已提交
576
	bp->len = cpu_to_be32(nbytes);
577

578 579 580
	sg_init_one(&sg, bp, sizeof(*bp));
	return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
}
581

582
static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
583
			       const struct in6_addr *daddr, struct in6_addr *saddr,
E
Eric Dumazet 已提交
584
			       const struct tcphdr *th)
585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602
{
	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))
603 604 605 606
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
607

608 609 610 611
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
612
	return 1;
613 614
}

615
static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
616 617 618
			       const struct sock *sk,
			       const struct request_sock *req,
			       const struct sk_buff *skb)
619
{
620
	const struct in6_addr *saddr, *daddr;
621 622
	struct tcp_md5sig_pool *hp;
	struct hash_desc *desc;
E
Eric Dumazet 已提交
623
	const struct tcphdr *th = tcp_hdr(skb);
624 625 626

	if (sk) {
		saddr = &inet6_sk(sk)->saddr;
627
		daddr = &sk->sk_v6_daddr;
628
	} else if (req) {
629 630
		saddr = &inet_rsk(req)->ir_v6_loc_addr;
		daddr = &inet_rsk(req)->ir_v6_rmt_addr;
631
	} else {
632
		const struct ipv6hdr *ip6h = ipv6_hdr(skb);
633 634
		saddr = &ip6h->saddr;
		daddr = &ip6h->daddr;
635
	}
636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663

	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;
664 665
}

E
Eric Dumazet 已提交
666
static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
667
{
668
	const __u8 *hash_location = NULL;
669
	struct tcp_md5sig_key *hash_expected;
670
	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
E
Eric Dumazet 已提交
671
	const struct tcphdr *th = tcp_hdr(skb);
672 673 674 675
	int genhash;
	u8 newhash[16];

	hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
676
	hash_location = tcp_parse_md5sig_option(th);
677

678 679 680 681 682 683
	/* 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);
684 685 686
		return 1;
	}

687 688
	if (!hash_expected && hash_location) {
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
689 690 691 692
		return 1;
	}

	/* check the signature */
693 694 695 696
	genhash = tcp_v6_md5_hash_skb(newhash,
				      hash_expected,
				      NULL, NULL, skb);

697
	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
698 699 700 701
		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));
702 703 704 705 706 707
		return 1;
	}
	return 0;
}
#endif

708
struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
709
	.family		=	AF_INET6,
710
	.obj_size	=	sizeof(struct tcp6_request_sock),
711
	.rtx_syn_ack	=	tcp_v6_rtx_synack,
712 713
	.send_ack	=	tcp_v6_reqsk_send_ack,
	.destructor	=	tcp_v6_reqsk_destructor,
714 715
	.send_reset	=	tcp_v6_send_reset,
	.syn_ack_timeout = 	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
716 717
};

718
#ifdef CONFIG_TCP_MD5SIG
719
static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
720
	.md5_lookup	=	tcp_v6_reqsk_md5_lookup,
721
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
722
};
723
#endif
724

725
static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win,
726 727
				 u32 tsval, u32 tsecr,
				 struct tcp_md5sig_key *key, int rst, u8 tclass)
L
Linus Torvalds 已提交
728
{
729 730
	const struct tcphdr *th = tcp_hdr(skb);
	struct tcphdr *t1;
L
Linus Torvalds 已提交
731
	struct sk_buff *buff;
732
	struct flowi6 fl6;
E
Eric Dumazet 已提交
733
	struct net *net = dev_net(skb_dst(skb)->dev);
734
	struct sock *ctl_sk = net->ipv6.tcp_sk;
735
	unsigned int tot_len = sizeof(struct tcphdr);
E
Eric Dumazet 已提交
736
	struct dst_entry *dst;
737
	__be32 *topt;
L
Linus Torvalds 已提交
738

739
	if (tsecr)
740
		tot_len += TCPOLEN_TSTAMP_ALIGNED;
741 742 743 744 745 746
#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
747
			 GFP_ATOMIC);
748 749
	if (buff == NULL)
		return;
L
Linus Torvalds 已提交
750

751
	skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
L
Linus Torvalds 已提交
752

753
	t1 = (struct tcphdr *) skb_push(buff, tot_len);
754
	skb_reset_transport_header(buff);
L
Linus Torvalds 已提交
755 756 757 758 759

	/* Swap the send and the receive. */
	memset(t1, 0, sizeof(*t1));
	t1->dest = th->source;
	t1->source = th->dest;
760
	t1->doff = tot_len / 4;
761 762 763 764 765
	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
766

767 768
	topt = (__be32 *)(t1 + 1);

769
	if (tsecr) {
770 771
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
772 773
		*topt++ = htonl(tsval);
		*topt++ = htonl(tsecr);
774 775
	}

776 777
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
778 779 780
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
		tcp_v6_md5_hash_hdr((__u8 *)topt, key,
781 782
				    &ipv6_hdr(skb)->saddr,
				    &ipv6_hdr(skb)->daddr, t1);
783 784 785
	}
#endif

786
	memset(&fl6, 0, sizeof(fl6));
A
Alexey Dobriyan 已提交
787 788
	fl6.daddr = ipv6_hdr(skb)->saddr;
	fl6.saddr = ipv6_hdr(skb)->daddr;
L
Linus Torvalds 已提交
789

790 791 792
	buff->ip_summed = CHECKSUM_PARTIAL;
	buff->csum = 0;

793
	__tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
L
Linus Torvalds 已提交
794

795
	fl6.flowi6_proto = IPPROTO_TCP;
A
Alexey Kuznetsov 已提交
796 797
	if (ipv6_addr_type(&fl6.daddr) & IPV6_ADDR_LINKLOCAL)
		fl6.flowi6_oif = inet6_iif(skb);
798 799
	fl6.fl6_dport = t1->dest;
	fl6.fl6_sport = t1->source;
800
	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
L
Linus Torvalds 已提交
801

802 803 804 805
	/* Pass a socket to ip6_dst_lookup either it is for RST
	 * Underlying function will use this to retrieve the network
	 * namespace
	 */
806
	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
807 808
	if (!IS_ERR(dst)) {
		skb_dst_set(buff, dst);
809
		ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass);
810 811 812 813
		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
814 815 816 817 818
	}

	kfree_skb(buff);
}

819
static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
820
{
821
	const struct tcphdr *th = tcp_hdr(skb);
822
	u32 seq = 0, ack_seq = 0;
823
	struct tcp_md5sig_key *key = NULL;
824 825 826 827 828 829 830
#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
L
Linus Torvalds 已提交
831

832
	if (th->rst)
L
Linus Torvalds 已提交
833 834
		return;

835 836
	if (!ipv6_unicast_destination(skb))
		return;
L
Linus Torvalds 已提交
837

838
#ifdef CONFIG_TCP_MD5SIG
839 840 841 842 843 844 845 846 847 848
	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),
849 850
					   &tcp_hashinfo, &ipv6h->saddr,
					   th->source, &ipv6h->daddr,
851 852 853 854 855 856 857 858 859 860 861 862 863 864 865
					   ntohs(th->source), inet6_iif(skb));
		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;
	}
866 867
#endif

868 869 870 871 872
	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
873

874
	tcp_v6_send_response(skb, seq, ack_seq, 0, 0, 0, key, 1, 0);
875 876 877 878 879 880 881 882

#ifdef CONFIG_TCP_MD5SIG
release_sk1:
	if (sk1) {
		rcu_read_unlock();
		sock_put(sk1);
	}
#endif
883
}
L
Linus Torvalds 已提交
884

885 886
static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack,
			    u32 win, u32 tsval, u32 tsecr,
887
			    struct tcp_md5sig_key *key, u8 tclass)
888
{
889
	tcp_v6_send_response(skb, seq, ack, win, tsval, tsecr, key, 0, tclass);
L
Linus Torvalds 已提交
890 891 892 893
}

static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
894
	struct inet_timewait_sock *tw = inet_twsk(sk);
895
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
896

897
	tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
898
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
899
			tcp_time_stamp + tcptw->tw_ts_offset,
900 901
			tcptw->tw_ts_recent, tcp_twsk_md5_key(tcptw),
			tw->tw_tclass);
L
Linus Torvalds 已提交
902

903
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
904 905
}

906 907
static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
				  struct request_sock *req)
L
Linus Torvalds 已提交
908
{
909 910
	tcp_v6_send_ack(skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1,
			req->rcv_wnd, tcp_time_stamp, req->ts_recent,
911
			tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr), 0);
L
Linus Torvalds 已提交
912 913 914
}


W
Weilong Chen 已提交
915
static struct sock *tcp_v6_hnd_req(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
916
{
917
	struct request_sock *req, **prev;
918
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
919 920 921
	struct sock *nsk;

	/* Find possible connection requests. */
922
	req = inet6_csk_search_req(sk, &prev, th->source,
923 924
				   &ipv6_hdr(skb)->saddr,
				   &ipv6_hdr(skb)->daddr, inet6_iif(skb));
L
Linus Torvalds 已提交
925
	if (req)
926
		return tcp_check_req(sk, skb, req, prev, false);
L
Linus Torvalds 已提交
927

928
	nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
929 930
			&ipv6_hdr(skb)->saddr, th->source,
			&ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb));
L
Linus Torvalds 已提交
931 932 933 934 935 936

	if (nsk) {
		if (nsk->sk_state != TCP_TIME_WAIT) {
			bh_lock_sock(nsk);
			return nsk;
		}
937
		inet_twsk_put(inet_twsk(nsk));
L
Linus Torvalds 已提交
938 939 940
		return NULL;
	}

941
#ifdef CONFIG_SYN_COOKIES
942
	if (!th->syn)
943
		sk = cookie_v6_check(sk, skb);
L
Linus Torvalds 已提交
944 945 946 947 948 949 950 951 952
#endif
	return sk;
}

/* FIXME: this is substantially similar to the ipv4 code.
 * Can some kind of merge be done? -- erics
 */
static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
{
953 954
	struct tcp_options_received tmp_opt;
	struct request_sock *req;
955
	struct inet_request_sock *ireq;
L
Linus Torvalds 已提交
956 957
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
958
	__u32 isn = TCP_SKB_CB(skb)->when;
959
	struct dst_entry *dst = NULL;
960
	struct flowi6 fl6;
E
Eric Dumazet 已提交
961
	bool want_cookie = false;
L
Linus Torvalds 已提交
962 963 964 965 966

	if (skb->protocol == htons(ETH_P_IP))
		return tcp_v4_conn_request(sk, skb);

	if (!ipv6_unicast_destination(skb))
967
		goto drop;
L
Linus Torvalds 已提交
968

969 970
	if ((sysctl_tcp_syncookies == 2 ||
	     inet_csk_reqsk_queue_is_full(sk)) && !isn) {
971 972 973
		want_cookie = tcp_syn_flood_action(sk, skb, "TCPv6");
		if (!want_cookie)
			goto drop;
L
Linus Torvalds 已提交
974 975
	}

976 977
	if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1) {
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
L
Linus Torvalds 已提交
978
		goto drop;
979
	}
L
Linus Torvalds 已提交
980

981
	req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
L
Linus Torvalds 已提交
982 983 984
	if (req == NULL)
		goto drop;

985 986 987 988
#ifdef CONFIG_TCP_MD5SIG
	tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
#endif

L
Linus Torvalds 已提交
989 990 991
	tcp_clear_options(&tmp_opt);
	tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
	tmp_opt.user_mss = tp->rx_opt.user_mss;
C
Christoph Paasch 已提交
992
	tcp_parse_options(skb, &tmp_opt, 0, NULL);
L
Linus Torvalds 已提交
993

994
	if (want_cookie && !tmp_opt.saw_tstamp)
995 996
		tcp_clear_options(&tmp_opt);

L
Linus Torvalds 已提交
997 998 999
	tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
	tcp_openreq_init(req, &tmp_opt, skb);

1000 1001 1002
	ireq = inet_rsk(req);
	ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
	ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
1003
	if (!want_cookie || tmp_opt.tstamp_ok)
1004
		TCP_ECN_create_request(req, skb, sock_net(sk));
1005

1006
	ireq->ir_iif = sk->sk_bound_dev_if;
1007 1008 1009

	/* So that link locals have meaning */
	if (!sk->sk_bound_dev_if &&
1010 1011
	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
		ireq->ir_iif = inet6_iif(skb);
1012

1013
	if (!isn) {
1014 1015 1016 1017
		if (ipv6_opt_accepted(sk, skb) ||
		    np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
		    np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
			atomic_inc(&skb->users);
1018
			ireq->pktopts = skb;
1019
		}
1020 1021

		if (want_cookie) {
1022 1023
			isn = cookie_v6_init_sequence(sk, skb, &req->mss);
			req->cookie_ts = tmp_opt.tstamp_ok;
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
			goto have_isn;
		}

		/* VJ's idea. We save last timestamp seen
		 * from the destination in peer table, when entering
		 * state TIME-WAIT, and check against it before
		 * accepting new connection request.
		 *
		 * If "isn" is not zero, this request hit alive
		 * timewait bucket, so that all the necessary checks
		 * are made in the function processing timewait state.
		 */
		if (tmp_opt.saw_tstamp &&
		    tcp_death_row.sysctl_tw_recycle &&
1038 1039
		    (dst = inet6_csk_route_req(sk, &fl6, req)) != NULL) {
			if (!tcp_peer_is_proven(req, dst, true)) {
1040 1041 1042 1043 1044 1045 1046 1047
				NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED);
				goto drop_and_release;
			}
		}
		/* Kill the following clause, if you dislike this way. */
		else if (!sysctl_tcp_syncookies &&
			 (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) <
			  (sysctl_max_syn_backlog >> 2)) &&
1048
			 !tcp_peer_is_proven(req, dst, false)) {
1049 1050 1051 1052 1053 1054 1055 1056
			/* Without syncookies last quarter of
			 * backlog is filled with destinations,
			 * proven to be alive.
			 * It means that we continue to communicate
			 * to destinations, already remembered
			 * to the moment of synflood.
			 */
			LIMIT_NETDEBUG(KERN_DEBUG "TCP: drop open request from %pI6/%u\n",
1057
				       &ireq->ir_v6_rmt_addr, ntohs(tcp_hdr(skb)->source));
1058
			goto drop_and_release;
1059
		}
1060 1061

		isn = tcp_v6_init_sequence(skb);
1062
	}
1063
have_isn:
1064
	tcp_rsk(req)->snt_isn = isn;
L
Linus Torvalds 已提交
1065

1066 1067
	if (security_inet_conn_request(sk, skb, req))
		goto drop_and_release;
1068

1069
	if (tcp_v6_send_synack(sk, dst, &fl6, req,
1070
			       skb_get_queue_mapping(skb)) ||
1071
	    want_cookie)
1072
		goto drop_and_free;
L
Linus Torvalds 已提交
1073

1074
	tcp_rsk(req)->snt_synack = tcp_time_stamp;
1075
	tcp_rsk(req)->listener = NULL;
1076 1077
	inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
	return 0;
L
Linus Torvalds 已提交
1078

1079 1080
drop_and_release:
	dst_release(dst);
1081 1082
drop_and_free:
	reqsk_free(req);
L
Linus Torvalds 已提交
1083
drop:
1084
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1085 1086 1087
	return 0; /* don't send reset */
}

W
Weilong Chen 已提交
1088 1089 1090
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
1091
{
1092
	struct inet_request_sock *ireq;
L
Linus Torvalds 已提交
1093 1094 1095 1096 1097
	struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
	struct tcp6_sock *newtcp6sk;
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1098 1099 1100
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1101
	struct flowi6 fl6;
L
Linus Torvalds 已提交
1102 1103 1104 1105 1106 1107 1108 1109

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

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

1110
		if (newsk == NULL)
L
Linus Torvalds 已提交
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121
			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));

1122
		ipv6_addr_set_v4mapped(newinet->inet_daddr, &newsk->sk_v6_daddr);
L
Linus Torvalds 已提交
1123

E
Eric Dumazet 已提交
1124
		ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr);
L
Linus Torvalds 已提交
1125

1126
		newsk->sk_v6_rcv_saddr = newnp->saddr;
L
Linus Torvalds 已提交
1127

1128
		inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
1129
		newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1130 1131 1132 1133
#ifdef CONFIG_TCP_MD5SIG
		newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif

1134 1135
		newnp->ipv6_ac_list = NULL;
		newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1136 1137
		newnp->pktoptions  = NULL;
		newnp->opt	   = NULL;
1138
		newnp->mcast_oif   = inet6_iif(skb);
1139
		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
1140
		newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1141

1142 1143 1144 1145
		/*
		 * 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
1146 1147 1148
		 */

		/* It is tricky place. Until this moment IPv4 tcp
1149
		   worked with IPv6 icsk.icsk_af_ops.
L
Linus Torvalds 已提交
1150 1151
		   Sync it now.
		 */
1152
		tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
1153 1154 1155 1156

		return newsk;
	}

1157
	ireq = inet_rsk(req);
L
Linus Torvalds 已提交
1158 1159 1160 1161

	if (sk_acceptq_is_full(sk))
		goto out_overflow;

1162
	if (!dst) {
1163
		dst = inet6_csk_route_req(sk, &fl6, req);
1164
		if (!dst)
L
Linus Torvalds 已提交
1165
			goto out;
1166
	}
L
Linus Torvalds 已提交
1167 1168 1169

	newsk = tcp_create_openreq_child(sk, req, skb);
	if (newsk == NULL)
1170
		goto out_nonewsk;
L
Linus Torvalds 已提交
1171

1172 1173 1174 1175 1176
	/*
	 * 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
1177

1178
	newsk->sk_gso_type = SKB_GSO_TCPV6;
1179
	__ip6_dst_store(newsk, dst, NULL, NULL);
1180
	inet6_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1181 1182 1183 1184 1185 1186 1187 1188 1189 1190

	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));

1191 1192 1193 1194
	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
1195

1196
	/* Now IPv6 options...
L
Linus Torvalds 已提交
1197 1198 1199

	   First: no IPv4 options.
	 */
1200
	newinet->inet_opt = NULL;
1201
	newnp->ipv6_ac_list = NULL;
1202
	newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1203 1204 1205 1206 1207 1208

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

	/* Clone pktoptions received with SYN */
	newnp->pktoptions = NULL;
1209 1210
	if (ireq->pktopts != NULL) {
		newnp->pktoptions = skb_clone(ireq->pktopts,
1211
					      sk_gfp_atomic(sk, GFP_ATOMIC));
1212 1213
		consume_skb(ireq->pktopts);
		ireq->pktopts = NULL;
L
Linus Torvalds 已提交
1214 1215 1216 1217
		if (newnp->pktoptions)
			skb_set_owner_r(newnp->pktoptions, newsk);
	}
	newnp->opt	  = NULL;
1218
	newnp->mcast_oif  = inet6_iif(skb);
1219
	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1220
	newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1221 1222 1223 1224 1225 1226 1227

	/* 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.
	 */
1228 1229
	if (np->opt)
		newnp->opt = ipv6_dup_options(newsk, np->opt);
L
Linus Torvalds 已提交
1230

1231
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
L
Linus Torvalds 已提交
1232
	if (newnp->opt)
1233 1234
		inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
						     newnp->opt->opt_flen);
L
Linus Torvalds 已提交
1235

J
John Heffner 已提交
1236
	tcp_mtup_init(newsk);
L
Linus Torvalds 已提交
1237
	tcp_sync_mss(newsk, dst_mtu(dst));
1238
	newtp->advmss = dst_metric_advmss(dst);
1239 1240 1241 1242
	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
1243 1244
	tcp_initialize_rcv_mss(newsk);

E
Eric Dumazet 已提交
1245 1246
	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
1247

1248 1249
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
1250
	if ((key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr)) != NULL) {
1251 1252 1253 1254 1255
		/* 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.
		 */
1256
		tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
1257 1258
			       AF_INET6, key->key, key->keylen,
			       sk_gfp_atomic(sk, GFP_ATOMIC));
1259 1260 1261
	}
#endif

1262
	if (__inet_inherit_port(sk, newsk) < 0) {
1263 1264
		inet_csk_prepare_forced_close(newsk);
		tcp_done(newsk);
1265 1266
		goto out;
	}
1267
	__inet6_hash(newsk, NULL);
L
Linus Torvalds 已提交
1268 1269 1270 1271

	return newsk;

out_overflow:
1272
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1273
out_nonewsk:
L
Linus Torvalds 已提交
1274
	dst_release(dst);
1275 1276
out:
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1277 1278 1279
	return NULL;
}

1280
static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
L
Linus Torvalds 已提交
1281
{
1282
	if (skb->ip_summed == CHECKSUM_COMPLETE) {
H
Herbert Xu 已提交
1283
		if (!tcp_v6_check(skb->len, &ipv6_hdr(skb)->saddr,
1284
				  &ipv6_hdr(skb)->daddr, skb->csum)) {
1285
			skb->ip_summed = CHECKSUM_UNNECESSARY;
L
Linus Torvalds 已提交
1286
			return 0;
1287
		}
L
Linus Torvalds 已提交
1288
	}
1289

H
Herbert Xu 已提交
1290
	skb->csum = ~csum_unfold(tcp_v6_check(skb->len,
1291 1292
					      &ipv6_hdr(skb)->saddr,
					      &ipv6_hdr(skb)->daddr, 0));
1293

L
Linus Torvalds 已提交
1294
	if (skb->len <= 76) {
1295
		return __skb_checksum_complete(skb);
L
Linus Torvalds 已提交
1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324
	}
	return 0;
}

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

1325 1326 1327 1328 1329
#ifdef CONFIG_TCP_MD5SIG
	if (tcp_v6_inbound_md5_hash (sk, skb))
		goto discard;
#endif

1330
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348
		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.
1349
					       --ANK (980728)
L
Linus Torvalds 已提交
1350 1351
	 */
	if (np->rxopt.all)
1352
		opt_skb = skb_clone(skb, sk_gfp_atomic(sk, GFP_ATOMIC));
L
Linus Torvalds 已提交
1353 1354

	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
E
Eric Dumazet 已提交
1355 1356
		struct dst_entry *dst = sk->sk_rx_dst;

1357
		sock_rps_save_rxhash(sk, skb);
E
Eric Dumazet 已提交
1358 1359 1360 1361 1362 1363 1364 1365
		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;
			}
		}

1366
		tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
L
Linus Torvalds 已提交
1367 1368 1369 1370 1371
		if (opt_skb)
			goto ipv6_pktoptions;
		return 0;
	}

1372
	if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1373 1374
		goto csum_err;

1375
	if (sk->sk_state == TCP_LISTEN) {
L
Linus Torvalds 已提交
1376 1377 1378 1379 1380 1381 1382 1383 1384
		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
1385
		if (nsk != sk) {
1386
			sock_rps_save_rxhash(nsk, skb);
L
Linus Torvalds 已提交
1387 1388 1389 1390 1391 1392
			if (tcp_child_process(sk, nsk, skb))
				goto reset;
			if (opt_skb)
				__kfree_skb(opt_skb);
			return 0;
		}
1393
	} else
1394
		sock_rps_save_rxhash(sk, skb);
L
Linus Torvalds 已提交
1395

1396
	if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
L
Linus Torvalds 已提交
1397 1398 1399 1400 1401 1402
		goto reset;
	if (opt_skb)
		goto ipv6_pktoptions;
	return 0;

reset:
1403
	tcp_v6_send_reset(sk, skb);
L
Linus Torvalds 已提交
1404 1405 1406 1407 1408 1409
discard:
	if (opt_skb)
		__kfree_skb(opt_skb);
	kfree_skb(skb);
	return 0;
csum_err:
1410
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS);
1411
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425
	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))) {
1426
		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
1427
			np->mcast_oif = inet6_iif(opt_skb);
1428
		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1429
			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1430
		if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
1431
			np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
L
Linus Torvalds 已提交
1432 1433 1434 1435 1436 1437 1438 1439 1440
		if (ipv6_opt_accepted(sk, opt_skb)) {
			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);
		}
	}

1441
	kfree_skb(opt_skb);
L
Linus Torvalds 已提交
1442 1443 1444
	return 0;
}

1445
static int tcp_v6_rcv(struct sk_buff *skb)
L
Linus Torvalds 已提交
1446
{
1447
	const struct tcphdr *th;
1448
	const struct ipv6hdr *hdr;
L
Linus Torvalds 已提交
1449 1450
	struct sock *sk;
	int ret;
1451
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
1452 1453 1454 1455 1456 1457 1458

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

	/*
	 *	Count it even if it's bad.
	 */
1459
	TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1460 1461 1462 1463

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

1464
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1465 1466 1467 1468 1469 1470

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

1471
	if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
1472
		goto csum_error;
L
Linus Torvalds 已提交
1473

1474
	th = tcp_hdr(skb);
1475
	hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
1476 1477 1478 1479 1480
	TCP_SKB_CB(skb)->seq = ntohl(th->seq);
	TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
				    skb->len - th->doff*4);
	TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
	TCP_SKB_CB(skb)->when = 0;
1481
	TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
L
Linus Torvalds 已提交
1482 1483
	TCP_SKB_CB(skb)->sacked = 0;

1484
	sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest);
L
Linus Torvalds 已提交
1485 1486 1487 1488 1489 1490 1491
	if (!sk)
		goto no_tcp_socket;

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

1492 1493 1494 1495 1496
	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
1497 1498 1499
	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;

1500
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1501 1502
		goto discard_and_relse;

1503
	sk_mark_napi_id(sk, skb);
L
Linus Torvalds 已提交
1504 1505
	skb->dev = NULL;

1506
	bh_lock_sock_nested(sk);
L
Linus Torvalds 已提交
1507 1508
	ret = 0;
	if (!sock_owned_by_user(sk)) {
C
Chris Leech 已提交
1509
#ifdef CONFIG_NET_DMA
1510
		struct tcp_sock *tp = tcp_sk(sk);
1511
		if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
1512
			tp->ucopy.dma_chan = net_dma_find_channel();
1513 1514 1515
		if (tp->ucopy.dma_chan)
			ret = tcp_v6_do_rcv(sk, skb);
		else
C
Chris Leech 已提交
1516 1517 1518 1519 1520
#endif
		{
			if (!tcp_prequeue(sk, skb))
				ret = tcp_v6_do_rcv(sk, skb);
		}
1521 1522
	} else if (unlikely(sk_add_backlog(sk, skb,
					   sk->sk_rcvbuf + sk->sk_sndbuf))) {
Z
Zhu Yi 已提交
1523
		bh_unlock_sock(sk);
1524
		NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
Z
Zhu Yi 已提交
1525 1526
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1527 1528 1529 1530 1531 1532 1533 1534 1535 1536
	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;

	if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1537 1538
csum_error:
		TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS);
L
Linus Torvalds 已提交
1539
bad_packet:
1540
		TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1541
	} else {
1542
		tcp_v6_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554
	}

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)) {
1555
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1556 1557 1558
		goto discard_it;
	}

1559
	if (skb->len < (th->doff<<2)) {
1560
		inet_twsk_put(inet_twsk(sk));
1561 1562 1563 1564 1565
		goto bad_packet;
	}
	if (tcp_checksum_complete(skb)) {
		inet_twsk_put(inet_twsk(sk));
		goto csum_error;
L
Linus Torvalds 已提交
1566 1567
	}

1568
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1569 1570 1571 1572
	case TCP_TW_SYN:
	{
		struct sock *sk2;

1573
		sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1574
					    &ipv6_hdr(skb)->saddr, th->source,
1575
					    &ipv6_hdr(skb)->daddr,
1576
					    ntohs(th->dest), inet6_iif(skb));
L
Linus Torvalds 已提交
1577
		if (sk2 != NULL) {
1578 1579 1580
			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
1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595
			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;
	case TCP_TW_SUCCESS:;
	}
	goto discard_it;
}

E
Eric Dumazet 已提交
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
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;

	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;
1623

E
Eric Dumazet 已提交
1624
			if (dst)
E
Eric Dumazet 已提交
1625
				dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie);
E
Eric Dumazet 已提交
1626
			if (dst &&
1627
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
E
Eric Dumazet 已提交
1628 1629 1630 1631 1632
				skb_dst_set_noref(skb, dst);
		}
	}
}

1633 1634 1635 1636 1637 1638
static struct timewait_sock_ops tcp6_timewait_sock_ops = {
	.twsk_obj_size	= sizeof(struct tcp6_timewait_sock),
	.twsk_unique	= tcp_twsk_unique,
	.twsk_destructor= tcp_twsk_destructor,
};

1639
static const struct inet_connection_sock_af_ops ipv6_specific = {
1640 1641 1642
	.queue_xmit	   = inet6_csk_xmit,
	.send_check	   = tcp_v6_send_check,
	.rebuild_header	   = inet6_sk_rebuild_header,
E
Eric Dumazet 已提交
1643
	.sk_rx_dst_set	   = inet6_sk_rx_dst_set,
1644 1645 1646
	.conn_request	   = tcp_v6_conn_request,
	.syn_recv_sock	   = tcp_v6_syn_recv_sock,
	.net_header_len	   = sizeof(struct ipv6hdr),
1647
	.net_frag_header_len = sizeof(struct frag_hdr),
1648 1649 1650 1651
	.setsockopt	   = ipv6_setsockopt,
	.getsockopt	   = ipv6_getsockopt,
	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in6),
1652
	.bind_conflict	   = inet6_csk_bind_conflict,
1653
#ifdef CONFIG_COMPAT
1654 1655
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1656
#endif
L
Linus Torvalds 已提交
1657 1658
};

1659
#ifdef CONFIG_TCP_MD5SIG
1660
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1661
	.md5_lookup	=	tcp_v6_md5_lookup,
1662
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
1663 1664
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1665
#endif
1666

L
Linus Torvalds 已提交
1667 1668 1669 1670
/*
 *	TCP over IPv4 via INET6 API
 */

1671
static const struct inet_connection_sock_af_ops ipv6_mapped = {
1672 1673 1674
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
1675
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1676 1677 1678 1679 1680 1681 1682
	.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),
1683
	.bind_conflict	   = inet6_csk_bind_conflict,
1684
#ifdef CONFIG_COMPAT
1685 1686
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1687
#endif
L
Linus Torvalds 已提交
1688 1689
};

1690
#ifdef CONFIG_TCP_MD5SIG
1691
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1692
	.md5_lookup	=	tcp_v4_md5_lookup,
1693
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1694 1695
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1696
#endif
1697

L
Linus Torvalds 已提交
1698 1699 1700 1701 1702
/* 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)
{
1703
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1704

1705
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1706

1707
	icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
1708

1709
#ifdef CONFIG_TCP_MD5SIG
1710
	tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
1711 1712
#endif

L
Linus Torvalds 已提交
1713 1714 1715
	return 0;
}

1716
static void tcp_v6_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1717 1718
{
	tcp_v4_destroy_sock(sk);
1719
	inet6_destroy_sock(sk);
L
Linus Torvalds 已提交
1720 1721
}

1722
#ifdef CONFIG_PROC_FS
L
Linus Torvalds 已提交
1723
/* Proc filesystem TCPv6 sock list dumping. */
1724
static void get_openreq6(struct seq_file *seq,
1725
			 const struct sock *sk, struct request_sock *req, int i, kuid_t uid)
L
Linus Torvalds 已提交
1726 1727
{
	int ttd = req->expires - jiffies;
1728 1729
	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
1730 1731 1732 1733 1734 1735

	if (ttd < 0)
		ttd = 0;

	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1736
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1737 1738 1739
		   i,
		   src->s6_addr32[0], src->s6_addr32[1],
		   src->s6_addr32[2], src->s6_addr32[3],
E
Eric Dumazet 已提交
1740
		   inet_rsk(req)->ir_num,
L
Linus Torvalds 已提交
1741 1742
		   dest->s6_addr32[0], dest->s6_addr32[1],
		   dest->s6_addr32[2], dest->s6_addr32[3],
1743
		   ntohs(inet_rsk(req)->ir_rmt_port),
L
Linus Torvalds 已提交
1744
		   TCP_SYN_RECV,
W
Weilong Chen 已提交
1745
		   0, 0, /* could print option size, but that is af dependent. */
1746 1747
		   1,   /* timers active (only the expire timer) */
		   jiffies_to_clock_t(ttd),
1748
		   req->num_timeout,
1749
		   from_kuid_munged(seq_user_ns(seq), uid),
1750
		   0,  /* non standard timer */
L
Linus Torvalds 已提交
1751 1752 1753 1754 1755 1756
		   0, /* open_requests have no inode */
		   0, req);
}

static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
{
1757
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1758 1759 1760
	__u16 destp, srcp;
	int timer_active;
	unsigned long timer_expires;
1761 1762
	const struct inet_sock *inet = inet_sk(sp);
	const struct tcp_sock *tp = tcp_sk(sp);
1763
	const struct inet_connection_sock *icsk = inet_csk(sp);
L
Linus Torvalds 已提交
1764

1765 1766
	dest  = &sp->sk_v6_daddr;
	src   = &sp->sk_v6_rcv_saddr;
E
Eric Dumazet 已提交
1767 1768
	destp = ntohs(inet->inet_dport);
	srcp  = ntohs(inet->inet_sport);
1769 1770

	if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
L
Linus Torvalds 已提交
1771
		timer_active	= 1;
1772 1773
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
1774
		timer_active	= 4;
1775
		timer_expires	= icsk->icsk_timeout;
L
Linus Torvalds 已提交
1776 1777 1778 1779 1780 1781 1782 1783 1784 1785
	} 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 "
1786
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
L
Linus Torvalds 已提交
1787 1788 1789 1790 1791
		   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,
1792
		   sp->sk_state,
1793 1794
		   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
1795
		   timer_active,
1796
		   jiffies_delta_to_clock_t(timer_expires - jiffies),
1797
		   icsk->icsk_retransmits,
1798
		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1799
		   icsk->icsk_probes_out,
L
Linus Torvalds 已提交
1800 1801
		   sock_i_ino(sp),
		   atomic_read(&sp->sk_refcnt), sp,
1802 1803
		   jiffies_to_clock_t(icsk->icsk_rto),
		   jiffies_to_clock_t(icsk->icsk_ack.ato),
W
Weilong Chen 已提交
1804
		   (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
I
Ilpo Järvinen 已提交
1805 1806
		   tp->snd_cwnd,
		   tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh
L
Linus Torvalds 已提交
1807 1808 1809
		   );
}

1810
static void get_timewait6_sock(struct seq_file *seq,
1811
			       struct inet_timewait_sock *tw, int i)
L
Linus Torvalds 已提交
1812
{
1813
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1814
	__u16 destp, srcp;
1815
	s32 delta = tw->tw_ttd - inet_tw_time_stamp();
L
Linus Torvalds 已提交
1816

1817 1818
	dest = &tw->tw_v6_daddr;
	src  = &tw->tw_v6_rcv_saddr;
L
Linus Torvalds 已提交
1819 1820 1821 1822 1823
	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
1824
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1825 1826 1827 1828 1829 1830
		   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,
1831
		   3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
L
Linus Torvalds 已提交
1832 1833 1834 1835 1836 1837
		   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
1838
	struct sock *sk = v;
L
Linus Torvalds 已提交
1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853

	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
1854 1855 1856 1857
		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
1858 1859 1860 1861 1862 1863 1864 1865 1866
		break;
	case TCP_SEQ_STATE_OPENREQ:
		get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
		break;
	}
out:
	return 0;
}

1867 1868 1869 1870 1871 1872 1873 1874
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
1875 1876 1877
static struct tcp_seq_afinfo tcp6_seq_afinfo = {
	.name		= "tcp6",
	.family		= AF_INET6,
1878
	.seq_fops	= &tcp6_afinfo_seq_fops,
1879 1880 1881
	.seq_ops	= {
		.show		= tcp6_seq_show,
	},
L
Linus Torvalds 已提交
1882 1883
};

1884
int __net_init tcp6_proc_init(struct net *net)
L
Linus Torvalds 已提交
1885
{
1886
	return tcp_proc_register(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1887 1888
}

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

E
Eric Dumazet 已提交
1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905
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
1906 1907 1908 1909 1910 1911
struct proto tcpv6_prot = {
	.name			= "TCPv6",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v6_connect,
	.disconnect		= tcp_disconnect,
1912
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
1913 1914 1915 1916 1917 1918 1919
	.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,
1920 1921
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
1922
	.backlog_rcv		= tcp_v6_do_rcv,
E
Eric Dumazet 已提交
1923
	.release_cb		= tcp_release_cb,
1924
	.mtu_reduced		= tcp_v6_mtu_reduced,
L
Linus Torvalds 已提交
1925
	.hash			= tcp_v6_hash,
1926 1927
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
1928
	.enter_memory_pressure	= tcp_enter_memory_pressure,
1929
	.stream_memory_free	= tcp_stream_memory_free,
L
Linus Torvalds 已提交
1930 1931 1932
	.sockets_allocated	= &tcp_sockets_allocated,
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
1933
	.orphan_count		= &tcp_orphan_count,
1934
	.sysctl_mem		= sysctl_tcp_mem,
L
Linus Torvalds 已提交
1935 1936 1937 1938
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp6_sock),
1939
	.slab_flags		= SLAB_DESTROY_BY_RCU,
1940
	.twsk_prot		= &tcp6_timewait_sock_ops,
1941
	.rsk_prot		= &tcp6_request_sock_ops,
1942
	.h.hashinfo		= &tcp_hashinfo,
1943
	.no_autobind		= true,
1944 1945 1946 1947
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
#endif
A
Andrew Morton 已提交
1948
#ifdef CONFIG_MEMCG_KMEM
G
Glauber Costa 已提交
1949 1950
	.proto_cgroup		= tcp_proto_cgroup,
#endif
E
Eric Dumazet 已提交
1951
	.clear_sk		= tcp_v6_clear_sk,
L
Linus Torvalds 已提交
1952 1953
};

1954
static const struct inet6_protocol tcpv6_protocol = {
E
Eric Dumazet 已提交
1955
	.early_demux	=	tcp_v6_early_demux,
L
Linus Torvalds 已提交
1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966
	.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,
	.no_check	=	0,
1967 1968
	.flags		=	INET_PROTOSW_PERMANENT |
				INET_PROTOSW_ICSK,
L
Linus Torvalds 已提交
1969 1970
};

1971
static int __net_init tcpv6_net_init(struct net *net)
1972
{
1973 1974
	return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
				    SOCK_RAW, IPPROTO_TCP, net);
1975 1976
}

1977
static void __net_exit tcpv6_net_exit(struct net *net)
1978
{
1979
	inet_ctl_sock_destroy(net->ipv6.tcp_sk);
E
Eric W. Biederman 已提交
1980 1981
}

1982
static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
E
Eric W. Biederman 已提交
1983 1984
{
	inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6);
1985 1986 1987
}

static struct pernet_operations tcpv6_net_ops = {
E
Eric W. Biederman 已提交
1988 1989 1990
	.init	    = tcpv6_net_init,
	.exit	    = tcpv6_net_exit,
	.exit_batch = tcpv6_net_exit_batch,
1991 1992
};

1993
int __init tcpv6_init(void)
L
Linus Torvalds 已提交
1994
{
1995 1996
	int ret;

1997 1998
	ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
	if (ret)
1999
		goto out;
2000

L
Linus Torvalds 已提交
2001
	/* register inet6 protocol */
2002 2003 2004 2005
	ret = inet6_register_protosw(&tcpv6_protosw);
	if (ret)
		goto out_tcpv6_protocol;

2006
	ret = register_pernet_subsys(&tcpv6_net_ops);
2007 2008 2009 2010
	if (ret)
		goto out_tcpv6_protosw;
out:
	return ret;
2011

2012 2013
out_tcpv6_protosw:
	inet6_unregister_protosw(&tcpv6_protosw);
2014 2015
out_tcpv6_protocol:
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2016 2017 2018
	goto out;
}

2019
void tcpv6_exit(void)
2020
{
2021
	unregister_pernet_subsys(&tcpv6_net_ops);
2022 2023
	inet6_unregister_protosw(&tcpv6_protosw);
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
L
Linus Torvalds 已提交
2024
}