tcp_ipv6.c 51.7 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>
L
Linus Torvalds 已提交
66 67 68 69 70 71

#include <asm/uaccess.h>

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

72 73 74 75
#include <linux/crypto.h>
#include <linux/scatterlist.h>

static void	tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb);
76 77
static void	tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
				      struct request_sock *req);
L
Linus Torvalds 已提交
78 79 80

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

81 82
static const struct inet_connection_sock_af_ops ipv6_mapped;
static const struct inet_connection_sock_af_ops ipv6_specific;
83
#ifdef CONFIG_TCP_MD5SIG
84 85
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
86 87
#else
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
88
						   const struct in6_addr *addr)
89 90 91
{
	return NULL;
}
92
#endif
L
Linus Torvalds 已提交
93

94 95 96 97 98 99 100 101 102 103 104 105
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
106 107 108
static void tcp_v6_hash(struct sock *sk)
{
	if (sk->sk_state != TCP_CLOSE) {
109
		if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
L
Linus Torvalds 已提交
110 111 112 113
			tcp_prot.hash(sk);
			return;
		}
		local_bh_disable();
114
		__inet6_hash(sk, NULL);
L
Linus Torvalds 已提交
115 116 117 118
		local_bh_enable();
	}
}

119
static __u32 tcp_v6_init_sequence(const struct sk_buff *skb)
L
Linus Torvalds 已提交
120
{
121 122
	return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
					    ipv6_hdr(skb)->saddr.s6_addr32,
123 124
					    tcp_hdr(skb)->dest,
					    tcp_hdr(skb)->source);
L
Linus Torvalds 已提交
125 126
}

127
static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
L
Linus Torvalds 已提交
128 129 130
			  int addr_len)
{
	struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
131
	struct inet_sock *inet = inet_sk(sk);
132
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
133 134
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
135
	struct in6_addr *saddr = NULL, *final_p, final;
136
	struct rt6_info *rt;
137
	struct flowi6 fl6;
L
Linus Torvalds 已提交
138 139 140 141
	struct dst_entry *dst;
	int addr_type;
	int err;

142
	if (addr_len < SIN6_LEN_RFC2133)
L
Linus Torvalds 已提交
143 144
		return -EINVAL;

145
	if (usin->sin6_family != AF_INET6)
E
Eric Dumazet 已提交
146
		return -EAFNOSUPPORT;
L
Linus Torvalds 已提交
147

148
	memset(&fl6, 0, sizeof(fl6));
L
Linus Torvalds 已提交
149 150

	if (np->sndflow) {
151 152 153
		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
154
			struct ip6_flowlabel *flowlabel;
155
			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
L
Linus Torvalds 已提交
156 157
			if (flowlabel == NULL)
				return -EINVAL;
A
Alexey Dobriyan 已提交
158
			usin->sin6_addr = flowlabel->dst;
L
Linus Torvalds 已提交
159 160 161 162 163
			fl6_sock_release(flowlabel);
		}
	}

	/*
164 165 166 167 168
	 *	connect() to INADDR_ANY means loopback (BSD'ism).
	 */

	if(ipv6_addr_any(&usin->sin6_addr))
		usin->sin6_addr.s6_addr[15] = 0x1;
L
Linus Torvalds 已提交
169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199

	addr_type = ipv6_addr_type(&usin->sin6_addr);

	if(addr_type & IPV6_ADDR_MULTICAST)
		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 &&
	    !ipv6_addr_equal(&np->daddr, &usin->sin6_addr)) {
		tp->rx_opt.ts_recent = 0;
		tp->rx_opt.ts_recent_stamp = 0;
		tp->write_seq = 0;
	}

A
Alexey Dobriyan 已提交
200
	np->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
237 238 239
			ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
			ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
					       &np->rcv_saddr);
L
Linus Torvalds 已提交
240 241 242 243 244 245 246 247
		}

		return err;
	}

	if (!ipv6_addr_any(&np->rcv_saddr))
		saddr = &np->rcv_saddr;

248
	fl6.flowi6_proto = IPPROTO_TCP;
A
Alexey Dobriyan 已提交
249 250
	fl6.daddr = np->daddr;
	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, true);
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;
A
Alexey Dobriyan 已提交
268
		np->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 &&
281 282
	    ipv6_addr_equal(&rt->rt6i_dst.addr, &np->daddr))
		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
299 300
		tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
							     np->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
{
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);
389 390
	}

L
Linus Torvalds 已提交
391
	if (type == ICMPV6_PKT_TOOBIG) {
392 393 394
		tp->mtu_info = ntohl(info);
		if (!sock_owned_by_user(sk))
			tcp_v6_mtu_reduced(sk);
395 396 397
		else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED,
					   &tp->tsq_flags))
			sock_hold(sk);
L
Linus Torvalds 已提交
398 399 400 401 402
		goto out;
	}

	icmpv6_err_convert(type, code, &err);

403
	/* Might be for an request_sock */
L
Linus Torvalds 已提交
404
	switch (sk->sk_state) {
405
		struct request_sock *req, **prev;
L
Linus Torvalds 已提交
406 407 408 409
	case TCP_LISTEN:
		if (sock_owned_by_user(sk))
			goto out;

410 411
		req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
					   &hdr->saddr, inet6_iif(skb));
L
Linus Torvalds 已提交
412 413 414 415 416 417
		if (!req)
			goto out;

		/* ICMPs are not backlogged, hence we cannot get
		 * an established socket here.
		 */
418
		WARN_ON(req->sk != NULL);
L
Linus Torvalds 已提交
419

420
		if (seq != tcp_rsk(req)->snt_isn) {
421
			NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
422 423 424
			goto out;
		}

425
		inet_csk_reqsk_queue_drop(sk, req, prev);
426
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
427 428 429 430
		goto out;

	case TCP_SYN_SENT:
	case TCP_SYN_RECV:  /* Cannot happen.
431
			       It can, it SYNs are crossed. --ANK */
L
Linus Torvalds 已提交
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453
		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);
}


454 455
static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst,
			      struct flowi6 *fl6,
456
			      struct request_sock *req,
457 458
			      struct request_values *rvp,
			      u16 queue_mapping)
L
Linus Torvalds 已提交
459
{
460
	struct inet6_request_sock *treq = inet6_rsk(req);
L
Linus Torvalds 已提交
461 462
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct sk_buff * skb;
463
	int err = -ENOMEM;
L
Linus Torvalds 已提交
464

465 466
	/* First, grab a route. */
	if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL)
467
		goto done;
468

469
	skb = tcp_make_synack(sk, dst, req, rvp, NULL);
470

L
Linus Torvalds 已提交
471
	if (skb) {
472
		__tcp_v6_send_check(skb, &treq->loc_addr, &treq->rmt_addr);
L
Linus Torvalds 已提交
473

474
		fl6->daddr = treq->rmt_addr;
475
		skb_set_queue_mapping(skb, queue_mapping);
476
		err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass);
477
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
478 479 480 481 482 483
	}

done:
	return err;
}

484 485 486
static int tcp_v6_rtx_synack(struct sock *sk, struct request_sock *req,
			     struct request_values *rvp)
{
487
	struct flowi6 fl6;
488
	int res;
489

490 491 492 493
	res = tcp_v6_send_synack(sk, NULL, &fl6, req, rvp, 0);
	if (!res)
		TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS);
	return res;
494 495
}

496
static void tcp_v6_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
497
{
498
	kfree_skb(inet6_rsk(req)->pktopts);
L
Linus Torvalds 已提交
499 500
}

501 502
#ifdef CONFIG_TCP_MD5SIG
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
503
						   const struct in6_addr *addr)
504
{
E
Eric Dumazet 已提交
505
	return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535
}

static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
						struct sock *addr_sk)
{
	return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr);
}

static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
						      struct request_sock *req)
{
	return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr);
}

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
536
		if (ipv6_addr_v4mapped(&sin6->sin6_addr))
E
Eric Dumazet 已提交
537 538 539 540
			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);
541 542 543 544 545
	}

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

E
Eric Dumazet 已提交
546 547 548
	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);
549

E
Eric Dumazet 已提交
550 551
	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
			      AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
552 553
}

554
static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
555 556
					const struct in6_addr *daddr,
					const struct in6_addr *saddr, int nbytes)
557 558
{
	struct tcp6_pseudohdr *bp;
559
	struct scatterlist sg;
560

561 562
	bp = &hp->md5_blk.ip6;
	/* 1. TCP pseudo-header (RFC2460) */
A
Alexey Dobriyan 已提交
563 564
	bp->saddr = *saddr;
	bp->daddr = *daddr;
565
	bp->protocol = cpu_to_be32(IPPROTO_TCP);
A
Adam Langley 已提交
566
	bp->len = cpu_to_be32(nbytes);
567

568 569 570
	sg_init_one(&sg, bp, sizeof(*bp));
	return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
}
571

572
static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
573
			       const struct in6_addr *daddr, struct in6_addr *saddr,
E
Eric Dumazet 已提交
574
			       const struct tcphdr *th)
575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592
{
	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))
593 594 595 596
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
597

598 599 600 601
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
602
	return 1;
603 604
}

605
static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
606 607 608
			       const struct sock *sk,
			       const struct request_sock *req,
			       const struct sk_buff *skb)
609
{
610
	const struct in6_addr *saddr, *daddr;
611 612
	struct tcp_md5sig_pool *hp;
	struct hash_desc *desc;
E
Eric Dumazet 已提交
613
	const struct tcphdr *th = tcp_hdr(skb);
614 615 616 617

	if (sk) {
		saddr = &inet6_sk(sk)->saddr;
		daddr = &inet6_sk(sk)->daddr;
618
	} else if (req) {
619 620
		saddr = &inet6_rsk(req)->loc_addr;
		daddr = &inet6_rsk(req)->rmt_addr;
621
	} else {
622
		const struct ipv6hdr *ip6h = ipv6_hdr(skb);
623 624
		saddr = &ip6h->saddr;
		daddr = &ip6h->daddr;
625
	}
626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653

	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;
654 655
}

E
Eric Dumazet 已提交
656
static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
657
{
658
	const __u8 *hash_location = NULL;
659
	struct tcp_md5sig_key *hash_expected;
660
	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
E
Eric Dumazet 已提交
661
	const struct tcphdr *th = tcp_hdr(skb);
662 663 664 665
	int genhash;
	u8 newhash[16];

	hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
666
	hash_location = tcp_parse_md5sig_option(th);
667

668 669 670 671 672 673
	/* 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);
674 675 676
		return 1;
	}

677 678
	if (!hash_expected && hash_location) {
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
679 680 681 682
		return 1;
	}

	/* check the signature */
683 684 685 686
	genhash = tcp_v6_md5_hash_skb(newhash,
				      hash_expected,
				      NULL, NULL, skb);

687
	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
688 689 690 691
		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));
692 693 694 695 696 697
		return 1;
	}
	return 0;
}
#endif

698
struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
699
	.family		=	AF_INET6,
700
	.obj_size	=	sizeof(struct tcp6_request_sock),
701
	.rtx_syn_ack	=	tcp_v6_rtx_synack,
702 703
	.send_ack	=	tcp_v6_reqsk_send_ack,
	.destructor	=	tcp_v6_reqsk_destructor,
704 705
	.send_reset	=	tcp_v6_send_reset,
	.syn_ack_timeout = 	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
706 707
};

708
#ifdef CONFIG_TCP_MD5SIG
709
static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
710
	.md5_lookup	=	tcp_v6_reqsk_md5_lookup,
711
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
712
};
713
#endif
714

715
static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win,
716
				 u32 ts, struct tcp_md5sig_key *key, int rst, u8 tclass)
L
Linus Torvalds 已提交
717
{
718 719
	const struct tcphdr *th = tcp_hdr(skb);
	struct tcphdr *t1;
L
Linus Torvalds 已提交
720
	struct sk_buff *buff;
721
	struct flowi6 fl6;
E
Eric Dumazet 已提交
722
	struct net *net = dev_net(skb_dst(skb)->dev);
723
	struct sock *ctl_sk = net->ipv6.tcp_sk;
724
	unsigned int tot_len = sizeof(struct tcphdr);
E
Eric Dumazet 已提交
725
	struct dst_entry *dst;
726
	__be32 *topt;
L
Linus Torvalds 已提交
727

728 729
	if (ts)
		tot_len += TCPOLEN_TSTAMP_ALIGNED;
730 731 732 733 734 735
#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
736
			 GFP_ATOMIC);
737 738
	if (buff == NULL)
		return;
L
Linus Torvalds 已提交
739

740
	skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
L
Linus Torvalds 已提交
741

742
	t1 = (struct tcphdr *) skb_push(buff, tot_len);
743
	skb_reset_transport_header(buff);
L
Linus Torvalds 已提交
744 745 746 747 748

	/* Swap the send and the receive. */
	memset(t1, 0, sizeof(*t1));
	t1->dest = th->source;
	t1->source = th->dest;
749
	t1->doff = tot_len / 4;
750 751 752 753 754
	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
755

756 757
	topt = (__be32 *)(t1 + 1);

758 759 760 761 762 763 764
	if (ts) {
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
		*topt++ = htonl(tcp_time_stamp);
		*topt++ = htonl(ts);
	}

765 766
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
767 768 769
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
		tcp_v6_md5_hash_hdr((__u8 *)topt, key,
770 771
				    &ipv6_hdr(skb)->saddr,
				    &ipv6_hdr(skb)->daddr, t1);
772 773 774
	}
#endif

775
	memset(&fl6, 0, sizeof(fl6));
A
Alexey Dobriyan 已提交
776 777
	fl6.daddr = ipv6_hdr(skb)->saddr;
	fl6.saddr = ipv6_hdr(skb)->daddr;
L
Linus Torvalds 已提交
778

779 780 781
	buff->ip_summed = CHECKSUM_PARTIAL;
	buff->csum = 0;

782
	__tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
L
Linus Torvalds 已提交
783

784
	fl6.flowi6_proto = IPPROTO_TCP;
A
Alexey Kuznetsov 已提交
785 786
	if (ipv6_addr_type(&fl6.daddr) & IPV6_ADDR_LINKLOCAL)
		fl6.flowi6_oif = inet6_iif(skb);
787 788
	fl6.fl6_dport = t1->dest;
	fl6.fl6_sport = t1->source;
789
	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
L
Linus Torvalds 已提交
790

791 792 793 794
	/* Pass a socket to ip6_dst_lookup either it is for RST
	 * Underlying function will use this to retrieve the network
	 * namespace
	 */
795
	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL, false);
796 797
	if (!IS_ERR(dst)) {
		skb_dst_set(buff, dst);
798
		ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass);
799 800 801 802
		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
803 804 805 806 807
	}

	kfree_skb(buff);
}

808
static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
809
{
810
	const struct tcphdr *th = tcp_hdr(skb);
811
	u32 seq = 0, ack_seq = 0;
812
	struct tcp_md5sig_key *key = NULL;
813 814 815 816 817 818 819
#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
820

821
	if (th->rst)
L
Linus Torvalds 已提交
822 823
		return;

824 825
	if (!ipv6_unicast_destination(skb))
		return;
L
Linus Torvalds 已提交
826

827
#ifdef CONFIG_TCP_MD5SIG
828 829 830 831 832 833 834 835 836 837
	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),
838 839
					   &tcp_hashinfo, &ipv6h->saddr,
					   th->source, &ipv6h->daddr,
840 841 842 843 844 845 846 847 848 849 850 851 852 853 854
					   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;
	}
855 856
#endif

857 858 859 860 861
	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
862

863
	tcp_v6_send_response(skb, seq, ack_seq, 0, 0, key, 1, 0);
864 865 866 867 868 869 870 871

#ifdef CONFIG_TCP_MD5SIG
release_sk1:
	if (sk1) {
		rcu_read_unlock();
		sock_put(sk1);
	}
#endif
872
}
L
Linus Torvalds 已提交
873

874
static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts,
875
			    struct tcp_md5sig_key *key, u8 tclass)
876
{
877
	tcp_v6_send_response(skb, seq, ack, win, ts, key, 0, tclass);
L
Linus Torvalds 已提交
878 879 880 881
}

static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
882
	struct inet_timewait_sock *tw = inet_twsk(sk);
883
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
884

885
	tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
886
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
887 888
			tcptw->tw_ts_recent, tcp_twsk_md5_key(tcptw),
			tw->tw_tclass);
L
Linus Torvalds 已提交
889

890
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
891 892
}

893 894
static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
				  struct request_sock *req)
L
Linus Torvalds 已提交
895
{
896
	tcp_v6_send_ack(skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, req->ts_recent,
897
			tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr), 0);
L
Linus Torvalds 已提交
898 899 900 901 902
}


static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
{
903
	struct request_sock *req, **prev;
904
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
905 906 907
	struct sock *nsk;

	/* Find possible connection requests. */
908
	req = inet6_csk_search_req(sk, &prev, th->source,
909 910
				   &ipv6_hdr(skb)->saddr,
				   &ipv6_hdr(skb)->daddr, inet6_iif(skb));
L
Linus Torvalds 已提交
911
	if (req)
912
		return tcp_check_req(sk, skb, req, prev, false);
L
Linus Torvalds 已提交
913

914
	nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
915 916
			&ipv6_hdr(skb)->saddr, th->source,
			&ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb));
L
Linus Torvalds 已提交
917 918 919 920 921 922

	if (nsk) {
		if (nsk->sk_state != TCP_TIME_WAIT) {
			bh_lock_sock(nsk);
			return nsk;
		}
923
		inet_twsk_put(inet_twsk(nsk));
L
Linus Torvalds 已提交
924 925 926
		return NULL;
	}

927
#ifdef CONFIG_SYN_COOKIES
928
	if (!th->syn)
929
		sk = cookie_v6_check(sk, skb);
L
Linus Torvalds 已提交
930 931 932 933 934 935 936 937 938
#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)
{
939
	struct tcp_extend_values tmp_ext;
940
	struct tcp_options_received tmp_opt;
941
	const u8 *hash_location;
942
	struct request_sock *req;
943
	struct inet6_request_sock *treq;
L
Linus Torvalds 已提交
944 945
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
946
	__u32 isn = TCP_SKB_CB(skb)->when;
947
	struct dst_entry *dst = NULL;
948
	struct flowi6 fl6;
E
Eric Dumazet 已提交
949
	bool want_cookie = false;
L
Linus Torvalds 已提交
950 951 952 953 954

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

	if (!ipv6_unicast_destination(skb))
955
		goto drop;
L
Linus Torvalds 已提交
956

957
	if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
958 959 960
		want_cookie = tcp_syn_flood_action(sk, skb, "TCPv6");
		if (!want_cookie)
			goto drop;
L
Linus Torvalds 已提交
961 962
	}

963 964
	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
965
		goto drop;
966
	}
L
Linus Torvalds 已提交
967

968
	req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
L
Linus Torvalds 已提交
969 970 971
	if (req == NULL)
		goto drop;

972 973 974 975
#ifdef CONFIG_TCP_MD5SIG
	tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
#endif

L
Linus Torvalds 已提交
976 977 978
	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;
Y
Yuchung Cheng 已提交
979
	tcp_parse_options(skb, &tmp_opt, &hash_location, 0, NULL);
980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995

	if (tmp_opt.cookie_plus > 0 &&
	    tmp_opt.saw_tstamp &&
	    !tp->rx_opt.cookie_out_never &&
	    (sysctl_tcp_cookie_size > 0 ||
	     (tp->cookie_values != NULL &&
	      tp->cookie_values->cookie_desired > 0))) {
		u8 *c;
		u32 *d;
		u32 *mess = &tmp_ext.cookie_bakery[COOKIE_DIGEST_WORDS];
		int l = tmp_opt.cookie_plus - TCPOLEN_COOKIE_BASE;

		if (tcp_cookie_generator(&tmp_ext.cookie_bakery[0]) != 0)
			goto drop_and_free;

		/* Secret recipe starts with IP addresses */
996
		d = (__force u32 *)&ipv6_hdr(skb)->daddr.s6_addr32[0];
997 998 999 1000
		*mess++ ^= *d++;
		*mess++ ^= *d++;
		*mess++ ^= *d++;
		*mess++ ^= *d++;
1001
		d = (__force u32 *)&ipv6_hdr(skb)->saddr.s6_addr32[0];
1002 1003 1004 1005 1006 1007 1008 1009 1010
		*mess++ ^= *d++;
		*mess++ ^= *d++;
		*mess++ ^= *d++;
		*mess++ ^= *d++;

		/* plus variable length Initiator Cookie */
		c = (u8 *)mess;
		while (l-- > 0)
			*c++ ^= *hash_location++;
L
Linus Torvalds 已提交
1011

E
Eric Dumazet 已提交
1012
		want_cookie = false;	/* not our kind of cookie */
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
		tmp_ext.cookie_out_never = 0; /* false */
		tmp_ext.cookie_plus = tmp_opt.cookie_plus;
	} else if (!tp->rx_opt.cookie_in_always) {
		/* redundant indications, but ensure initialization. */
		tmp_ext.cookie_out_never = 1; /* true */
		tmp_ext.cookie_plus = 0;
	} else {
		goto drop_and_free;
	}
	tmp_ext.cookie_in_always = tp->rx_opt.cookie_in_always;
L
Linus Torvalds 已提交
1023

1024
	if (want_cookie && !tmp_opt.saw_tstamp)
1025 1026
		tcp_clear_options(&tmp_opt);

L
Linus Torvalds 已提交
1027 1028 1029
	tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
	tcp_openreq_init(req, &tmp_opt, skb);

1030
	treq = inet6_rsk(req);
A
Alexey Dobriyan 已提交
1031 1032
	treq->rmt_addr = ipv6_hdr(skb)->saddr;
	treq->loc_addr = ipv6_hdr(skb)->daddr;
1033
	if (!want_cookie || tmp_opt.tstamp_ok)
1034
		TCP_ECN_create_request(req, skb, sock_net(sk));
1035

1036 1037 1038 1039 1040 1041 1042
	treq->iif = sk->sk_bound_dev_if;

	/* So that link locals have meaning */
	if (!sk->sk_bound_dev_if &&
	    ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL)
		treq->iif = inet6_iif(skb);

1043
	if (!isn) {
1044 1045 1046 1047 1048 1049
		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);
			treq->pktopts = skb;
		}
1050 1051

		if (want_cookie) {
1052 1053
			isn = cookie_v6_init_sequence(sk, skb, &req->mss);
			req->cookie_ts = tmp_opt.tstamp_ok;
1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
			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 &&
1068 1069
		    (dst = inet6_csk_route_req(sk, &fl6, req)) != NULL) {
			if (!tcp_peer_is_proven(req, dst, true)) {
1070 1071 1072 1073 1074 1075 1076 1077
				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)) &&
1078
			 !tcp_peer_is_proven(req, dst, false)) {
1079 1080 1081 1082 1083 1084 1085 1086 1087 1088
			/* 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",
				       &treq->rmt_addr, ntohs(tcp_hdr(skb)->source));
			goto drop_and_release;
1089
		}
1090 1091

		isn = tcp_v6_init_sequence(skb);
1092
	}
1093
have_isn:
1094
	tcp_rsk(req)->snt_isn = isn;
L
Linus Torvalds 已提交
1095

1096 1097
	if (security_inet_conn_request(sk, skb, req))
		goto drop_and_release;
1098

1099
	if (tcp_v6_send_synack(sk, dst, &fl6, req,
1100 1101
			       (struct request_values *)&tmp_ext,
			       skb_get_queue_mapping(skb)) ||
1102
	    want_cookie)
1103
		goto drop_and_free;
L
Linus Torvalds 已提交
1104

1105
	tcp_rsk(req)->snt_synack = tcp_time_stamp;
1106
	tcp_rsk(req)->listener = NULL;
1107 1108
	inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
	return 0;
L
Linus Torvalds 已提交
1109

1110 1111
drop_and_release:
	dst_release(dst);
1112 1113
drop_and_free:
	reqsk_free(req);
L
Linus Torvalds 已提交
1114
drop:
1115
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1116 1117 1118 1119
	return 0; /* don't send reset */
}

static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
1120
					  struct request_sock *req,
L
Linus Torvalds 已提交
1121 1122
					  struct dst_entry *dst)
{
1123
	struct inet6_request_sock *treq;
L
Linus Torvalds 已提交
1124 1125 1126 1127 1128
	struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
	struct tcp6_sock *newtcp6sk;
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1129 1130 1131
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1132
	struct flowi6 fl6;
L
Linus Torvalds 已提交
1133 1134 1135 1136 1137 1138 1139 1140

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

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

1141
		if (newsk == NULL)
L
Linus Torvalds 已提交
1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
			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));

E
Eric Dumazet 已提交
1153
		ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr);
L
Linus Torvalds 已提交
1154

E
Eric Dumazet 已提交
1155
		ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr);
L
Linus Torvalds 已提交
1156

A
Alexey Dobriyan 已提交
1157
		newnp->rcv_saddr = newnp->saddr;
L
Linus Torvalds 已提交
1158

1159
		inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
1160
		newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1161 1162 1163 1164
#ifdef CONFIG_TCP_MD5SIG
		newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif

1165 1166
		newnp->ipv6_ac_list = NULL;
		newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1167 1168
		newnp->pktoptions  = NULL;
		newnp->opt	   = NULL;
1169
		newnp->mcast_oif   = inet6_iif(skb);
1170
		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
1171
		newnp->rcv_tclass  = ipv6_get_dsfield(ipv6_hdr(skb));
L
Linus Torvalds 已提交
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 1179
		 */

		/* It is tricky place. Until this moment IPv4 tcp
1180
		   worked with IPv6 icsk.icsk_af_ops.
L
Linus Torvalds 已提交
1181 1182
		   Sync it now.
		 */
1183
		tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
1184 1185 1186 1187

		return newsk;
	}

1188
	treq = inet6_rsk(req);
L
Linus Torvalds 已提交
1189 1190 1191 1192

	if (sk_acceptq_is_full(sk))
		goto out_overflow;

1193
	if (!dst) {
1194
		dst = inet6_csk_route_req(sk, &fl6, req);
1195
		if (!dst)
L
Linus Torvalds 已提交
1196
			goto out;
1197
	}
L
Linus Torvalds 已提交
1198 1199 1200

	newsk = tcp_create_openreq_child(sk, req, skb);
	if (newsk == NULL)
1201
		goto out_nonewsk;
L
Linus Torvalds 已提交
1202

1203 1204 1205 1206 1207
	/*
	 * 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
1208

1209
	newsk->sk_gso_type = SKB_GSO_TCPV6;
1210
	__ip6_dst_store(newsk, dst, NULL, NULL);
1211
	inet6_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1212 1213 1214 1215 1216 1217 1218 1219 1220 1221

	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));

A
Alexey Dobriyan 已提交
1222 1223 1224
	newnp->daddr = treq->rmt_addr;
	newnp->saddr = treq->loc_addr;
	newnp->rcv_saddr = treq->loc_addr;
1225
	newsk->sk_bound_dev_if = treq->iif;
L
Linus Torvalds 已提交
1226

1227
	/* Now IPv6 options...
L
Linus Torvalds 已提交
1228 1229 1230

	   First: no IPv4 options.
	 */
1231
	newinet->inet_opt = NULL;
1232
	newnp->ipv6_ac_list = NULL;
1233
	newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1234 1235 1236 1237 1238 1239

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

	/* Clone pktoptions received with SYN */
	newnp->pktoptions = NULL;
1240
	if (treq->pktopts != NULL) {
1241 1242
		newnp->pktoptions = skb_clone(treq->pktopts,
					      sk_gfp_atomic(sk, GFP_ATOMIC));
1243
		consume_skb(treq->pktopts);
1244
		treq->pktopts = NULL;
L
Linus Torvalds 已提交
1245 1246 1247 1248
		if (newnp->pktoptions)
			skb_set_owner_r(newnp->pktoptions, newsk);
	}
	newnp->opt	  = NULL;
1249
	newnp->mcast_oif  = inet6_iif(skb);
1250
	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1251
	newnp->rcv_tclass = ipv6_get_dsfield(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1252 1253 1254 1255 1256 1257 1258

	/* 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.
	 */
1259 1260
	if (np->opt)
		newnp->opt = ipv6_dup_options(newsk, np->opt);
L
Linus Torvalds 已提交
1261

1262
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
L
Linus Torvalds 已提交
1263
	if (newnp->opt)
1264 1265
		inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
						     newnp->opt->opt_flen);
L
Linus Torvalds 已提交
1266

J
John Heffner 已提交
1267
	tcp_mtup_init(newsk);
L
Linus Torvalds 已提交
1268
	tcp_sync_mss(newsk, dst_mtu(dst));
1269
	newtp->advmss = dst_metric_advmss(dst);
1270 1271 1272 1273
	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
1274
	tcp_initialize_rcv_mss(newsk);
1275
	tcp_synack_rtt_meas(newsk, req);
1276
	newtp->total_retrans = req->num_retrans;
L
Linus Torvalds 已提交
1277

E
Eric Dumazet 已提交
1278 1279
	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
1280

1281 1282 1283 1284 1285 1286 1287 1288
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
	if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) {
		/* 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.
		 */
E
Eric Dumazet 已提交
1289
		tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newnp->daddr,
1290 1291
			       AF_INET6, key->key, key->keylen,
			       sk_gfp_atomic(sk, GFP_ATOMIC));
1292 1293 1294
	}
#endif

1295
	if (__inet_inherit_port(sk, newsk) < 0) {
1296 1297
		inet_csk_prepare_forced_close(newsk);
		tcp_done(newsk);
1298 1299
		goto out;
	}
1300
	__inet6_hash(newsk, NULL);
L
Linus Torvalds 已提交
1301 1302 1303 1304

	return newsk;

out_overflow:
1305
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1306
out_nonewsk:
L
Linus Torvalds 已提交
1307
	dst_release(dst);
1308 1309
out:
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1310 1311 1312
	return NULL;
}

1313
static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
L
Linus Torvalds 已提交
1314
{
1315
	if (skb->ip_summed == CHECKSUM_COMPLETE) {
H
Herbert Xu 已提交
1316
		if (!tcp_v6_check(skb->len, &ipv6_hdr(skb)->saddr,
1317
				  &ipv6_hdr(skb)->daddr, skb->csum)) {
1318
			skb->ip_summed = CHECKSUM_UNNECESSARY;
L
Linus Torvalds 已提交
1319
			return 0;
1320
		}
L
Linus Torvalds 已提交
1321
	}
1322

H
Herbert Xu 已提交
1323
	skb->csum = ~csum_unfold(tcp_v6_check(skb->len,
1324 1325
					      &ipv6_hdr(skb)->saddr,
					      &ipv6_hdr(skb)->daddr, 0));
1326

L
Linus Torvalds 已提交
1327
	if (skb->len <= 76) {
1328
		return __skb_checksum_complete(skb);
L
Linus Torvalds 已提交
1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
	}
	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);

1358 1359 1360 1361 1362
#ifdef CONFIG_TCP_MD5SIG
	if (tcp_v6_inbound_md5_hash (sk, skb))
		goto discard;
#endif

1363
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381
		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.
1382
					       --ANK (980728)
L
Linus Torvalds 已提交
1383 1384
	 */
	if (np->rxopt.all)
1385
		opt_skb = skb_clone(skb, sk_gfp_atomic(sk, GFP_ATOMIC));
L
Linus Torvalds 已提交
1386 1387

	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
E
Eric Dumazet 已提交
1388 1389
		struct dst_entry *dst = sk->sk_rx_dst;

1390
		sock_rps_save_rxhash(sk, skb);
E
Eric Dumazet 已提交
1391 1392 1393 1394 1395 1396 1397 1398
		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;
			}
		}

1399
		if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len))
L
Linus Torvalds 已提交
1400 1401 1402 1403 1404 1405
			goto reset;
		if (opt_skb)
			goto ipv6_pktoptions;
		return 0;
	}

1406
	if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1407 1408
		goto csum_err;

1409
	if (sk->sk_state == TCP_LISTEN) {
L
Linus Torvalds 已提交
1410 1411 1412 1413 1414 1415 1416 1417 1418
		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..
		 */
1419
		if(nsk != sk) {
1420
			sock_rps_save_rxhash(nsk, skb);
L
Linus Torvalds 已提交
1421 1422 1423 1424 1425 1426
			if (tcp_child_process(sk, nsk, skb))
				goto reset;
			if (opt_skb)
				__kfree_skb(opt_skb);
			return 0;
		}
1427
	} else
1428
		sock_rps_save_rxhash(sk, skb);
L
Linus Torvalds 已提交
1429

1430
	if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
L
Linus Torvalds 已提交
1431 1432 1433 1434 1435 1436
		goto reset;
	if (opt_skb)
		goto ipv6_pktoptions;
	return 0;

reset:
1437
	tcp_v6_send_reset(sk, skb);
L
Linus Torvalds 已提交
1438 1439 1440 1441 1442 1443
discard:
	if (opt_skb)
		__kfree_skb(opt_skb);
	kfree_skb(skb);
	return 0;
csum_err:
1444
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458
	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))) {
1459
		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
1460
			np->mcast_oif = inet6_iif(opt_skb);
1461
		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1462
			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1463
		if (np->rxopt.bits.rxtclass)
1464
			np->rcv_tclass = ipv6_get_dsfield(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1465 1466 1467 1468 1469 1470 1471 1472 1473
		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);
		}
	}

1474
	kfree_skb(opt_skb);
L
Linus Torvalds 已提交
1475 1476 1477
	return 0;
}

1478
static int tcp_v6_rcv(struct sk_buff *skb)
L
Linus Torvalds 已提交
1479
{
1480
	const struct tcphdr *th;
1481
	const struct ipv6hdr *hdr;
L
Linus Torvalds 已提交
1482 1483
	struct sock *sk;
	int ret;
1484
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
1485 1486 1487 1488 1489 1490 1491

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

	/*
	 *	Count it even if it's bad.
	 */
1492
	TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1493 1494 1495 1496

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

1497
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1498 1499 1500 1501 1502 1503

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

1504
	if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
L
Linus Torvalds 已提交
1505 1506
		goto bad_packet;

1507
	th = tcp_hdr(skb);
1508
	hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
1509 1510 1511 1512 1513
	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;
1514
	TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
L
Linus Torvalds 已提交
1515 1516
	TCP_SKB_CB(skb)->sacked = 0;

1517
	sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest);
L
Linus Torvalds 已提交
1518 1519 1520 1521 1522 1523 1524
	if (!sk)
		goto no_tcp_socket;

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

1525 1526 1527 1528 1529
	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
1530 1531 1532
	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;

1533
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1534 1535 1536 1537
		goto discard_and_relse;

	skb->dev = NULL;

1538
	bh_lock_sock_nested(sk);
L
Linus Torvalds 已提交
1539 1540
	ret = 0;
	if (!sock_owned_by_user(sk)) {
C
Chris Leech 已提交
1541
#ifdef CONFIG_NET_DMA
1542
		struct tcp_sock *tp = tcp_sk(sk);
1543
		if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
1544
			tp->ucopy.dma_chan = net_dma_find_channel();
1545 1546 1547
		if (tp->ucopy.dma_chan)
			ret = tcp_v6_do_rcv(sk, skb);
		else
C
Chris Leech 已提交
1548 1549 1550 1551 1552
#endif
		{
			if (!tcp_prequeue(sk, skb))
				ret = tcp_v6_do_rcv(sk, skb);
		}
1553 1554
	} else if (unlikely(sk_add_backlog(sk, skb,
					   sk->sk_rcvbuf + sk->sk_sndbuf))) {
Z
Zhu Yi 已提交
1555
		bh_unlock_sock(sk);
1556
		NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
Z
Zhu Yi 已提交
1557 1558
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569
	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)) {
bad_packet:
1570
		TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1571
	} else {
1572
		tcp_v6_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589
	}

discard_it:

	/*
	 *	Discard frame
	 */

	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)) {
1590
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1591 1592 1593 1594
		goto discard_it;
	}

	if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1595
		TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1596
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1597 1598 1599
		goto discard_it;
	}

1600
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1601 1602 1603 1604
	case TCP_TW_SYN:
	{
		struct sock *sk2;

1605
		sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1606
					    &ipv6_hdr(skb)->saddr, th->source,
1607
					    &ipv6_hdr(skb)->daddr,
1608
					    ntohs(th->dest), inet6_iif(skb));
L
Linus Torvalds 已提交
1609
		if (sk2 != NULL) {
1610 1611 1612
			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
1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627
			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
1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654
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;
1655

E
Eric Dumazet 已提交
1656
			if (dst)
E
Eric Dumazet 已提交
1657
				dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie);
E
Eric Dumazet 已提交
1658
			if (dst &&
1659
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
E
Eric Dumazet 已提交
1660 1661 1662 1663 1664
				skb_dst_set_noref(skb, dst);
		}
	}
}

1665 1666 1667 1668 1669 1670
static struct timewait_sock_ops tcp6_timewait_sock_ops = {
	.twsk_obj_size	= sizeof(struct tcp6_timewait_sock),
	.twsk_unique	= tcp_twsk_unique,
	.twsk_destructor= tcp_twsk_destructor,
};

1671
static const struct inet_connection_sock_af_ops ipv6_specific = {
1672 1673 1674
	.queue_xmit	   = inet6_csk_xmit,
	.send_check	   = tcp_v6_send_check,
	.rebuild_header	   = inet6_sk_rebuild_header,
E
Eric Dumazet 已提交
1675
	.sk_rx_dst_set	   = inet6_sk_rx_dst_set,
1676 1677 1678
	.conn_request	   = tcp_v6_conn_request,
	.syn_recv_sock	   = tcp_v6_syn_recv_sock,
	.net_header_len	   = sizeof(struct ipv6hdr),
1679
	.net_frag_header_len = sizeof(struct frag_hdr),
1680 1681 1682 1683
	.setsockopt	   = ipv6_setsockopt,
	.getsockopt	   = ipv6_getsockopt,
	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in6),
1684
	.bind_conflict	   = inet6_csk_bind_conflict,
1685
#ifdef CONFIG_COMPAT
1686 1687
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1688
#endif
L
Linus Torvalds 已提交
1689 1690
};

1691
#ifdef CONFIG_TCP_MD5SIG
1692
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1693
	.md5_lookup	=	tcp_v6_md5_lookup,
1694
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
1695 1696
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1697
#endif
1698

L
Linus Torvalds 已提交
1699 1700 1701 1702
/*
 *	TCP over IPv4 via INET6 API
 */

1703
static const struct inet_connection_sock_af_ops ipv6_mapped = {
1704 1705 1706
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
1707
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1708 1709 1710 1711 1712 1713 1714
	.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),
1715
	.bind_conflict	   = inet6_csk_bind_conflict,
1716
#ifdef CONFIG_COMPAT
1717 1718
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1719
#endif
L
Linus Torvalds 已提交
1720 1721
};

1722
#ifdef CONFIG_TCP_MD5SIG
1723
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1724
	.md5_lookup	=	tcp_v4_md5_lookup,
1725
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1726 1727
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1728
#endif
1729

L
Linus Torvalds 已提交
1730 1731 1732 1733 1734
/* 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)
{
1735
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1736

1737
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1738

1739
	icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
1740

1741
#ifdef CONFIG_TCP_MD5SIG
1742
	tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
1743 1744
#endif

L
Linus Torvalds 已提交
1745 1746 1747
	return 0;
}

1748
static void tcp_v6_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1749 1750
{
	tcp_v4_destroy_sock(sk);
1751
	inet6_destroy_sock(sk);
L
Linus Torvalds 已提交
1752 1753
}

1754
#ifdef CONFIG_PROC_FS
L
Linus Torvalds 已提交
1755
/* Proc filesystem TCPv6 sock list dumping. */
1756
static void get_openreq6(struct seq_file *seq,
1757
			 const struct sock *sk, struct request_sock *req, int i, kuid_t uid)
L
Linus Torvalds 已提交
1758 1759
{
	int ttd = req->expires - jiffies;
1760 1761
	const struct in6_addr *src = &inet6_rsk(req)->loc_addr;
	const struct in6_addr *dest = &inet6_rsk(req)->rmt_addr;
L
Linus Torvalds 已提交
1762 1763 1764 1765 1766 1767

	if (ttd < 0)
		ttd = 0;

	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
D
Dan Rosenberg 已提交
1768
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1769 1770 1771
		   i,
		   src->s6_addr32[0], src->s6_addr32[1],
		   src->s6_addr32[2], src->s6_addr32[3],
1772
		   ntohs(inet_rsk(req)->loc_port),
L
Linus Torvalds 已提交
1773 1774
		   dest->s6_addr32[0], dest->s6_addr32[1],
		   dest->s6_addr32[2], dest->s6_addr32[3],
1775
		   ntohs(inet_rsk(req)->rmt_port),
L
Linus Torvalds 已提交
1776 1777
		   TCP_SYN_RECV,
		   0,0, /* could print option size, but that is af dependent. */
1778 1779
		   1,   /* timers active (only the expire timer) */
		   jiffies_to_clock_t(ttd),
1780
		   req->num_timeout,
1781
		   from_kuid_munged(seq_user_ns(seq), uid),
1782
		   0,  /* non standard timer */
L
Linus Torvalds 已提交
1783 1784 1785 1786 1787 1788
		   0, /* open_requests have no inode */
		   0, req);
}

static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
{
1789
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1790 1791 1792
	__u16 destp, srcp;
	int timer_active;
	unsigned long timer_expires;
1793 1794
	const struct inet_sock *inet = inet_sk(sp);
	const struct tcp_sock *tp = tcp_sk(sp);
1795
	const struct inet_connection_sock *icsk = inet_csk(sp);
1796
	const struct ipv6_pinfo *np = inet6_sk(sp);
L
Linus Torvalds 已提交
1797 1798 1799

	dest  = &np->daddr;
	src   = &np->rcv_saddr;
E
Eric Dumazet 已提交
1800 1801
	destp = ntohs(inet->inet_dport);
	srcp  = ntohs(inet->inet_sport);
1802 1803

	if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
L
Linus Torvalds 已提交
1804
		timer_active	= 1;
1805 1806
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
1807
		timer_active	= 4;
1808
		timer_expires	= icsk->icsk_timeout;
L
Linus Torvalds 已提交
1809 1810 1811 1812 1813 1814 1815 1816 1817 1818
	} 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 "
D
Dan Rosenberg 已提交
1819
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %lu %lu %u %u %d\n",
L
Linus Torvalds 已提交
1820 1821 1822 1823 1824
		   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,
1825
		   sp->sk_state,
1826 1827
		   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
1828
		   timer_active,
1829
		   jiffies_delta_to_clock_t(timer_expires - jiffies),
1830
		   icsk->icsk_retransmits,
1831
		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1832
		   icsk->icsk_probes_out,
L
Linus Torvalds 已提交
1833 1834
		   sock_i_ino(sp),
		   atomic_read(&sp->sk_refcnt), sp,
1835 1836
		   jiffies_to_clock_t(icsk->icsk_rto),
		   jiffies_to_clock_t(icsk->icsk_ack.ato),
1837
		   (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
I
Ilpo Järvinen 已提交
1838 1839
		   tp->snd_cwnd,
		   tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh
L
Linus Torvalds 已提交
1840 1841 1842
		   );
}

1843
static void get_timewait6_sock(struct seq_file *seq,
1844
			       struct inet_timewait_sock *tw, int i)
L
Linus Torvalds 已提交
1845
{
1846
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1847
	__u16 destp, srcp;
1848
	const struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
1849
	long delta = tw->tw_ttd - jiffies;
L
Linus Torvalds 已提交
1850

1851 1852
	dest = &tw6->tw_v6_daddr;
	src  = &tw6->tw_v6_rcv_saddr;
L
Linus Torvalds 已提交
1853 1854 1855 1856 1857
	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
1858
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1859 1860 1861 1862 1863 1864
		   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,
1865
		   3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
L
Linus Torvalds 已提交
1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899
		   atomic_read(&tw->tw_refcnt), tw);
}

static int tcp6_seq_show(struct seq_file *seq, void *v)
{
	struct tcp_iter_state *st;

	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:
		get_tcp6_sock(seq, v, st->num);
		break;
	case TCP_SEQ_STATE_OPENREQ:
		get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
		break;
	case TCP_SEQ_STATE_TIME_WAIT:
		get_timewait6_sock(seq, v, st->num);
		break;
	}
out:
	return 0;
}

1900 1901 1902 1903 1904 1905 1906 1907
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
1908 1909 1910
static struct tcp_seq_afinfo tcp6_seq_afinfo = {
	.name		= "tcp6",
	.family		= AF_INET6,
1911
	.seq_fops	= &tcp6_afinfo_seq_fops,
1912 1913 1914
	.seq_ops	= {
		.show		= tcp6_seq_show,
	},
L
Linus Torvalds 已提交
1915 1916
};

1917
int __net_init tcp6_proc_init(struct net *net)
L
Linus Torvalds 已提交
1918
{
1919
	return tcp_proc_register(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1920 1921
}

1922
void tcp6_proc_exit(struct net *net)
L
Linus Torvalds 已提交
1923
{
1924
	tcp_proc_unregister(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1925 1926 1927 1928 1929 1930 1931 1932 1933
}
#endif

struct proto tcpv6_prot = {
	.name			= "TCPv6",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v6_connect,
	.disconnect		= tcp_disconnect,
1934
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
1935 1936 1937 1938 1939 1940 1941
	.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,
1942 1943
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
1944
	.backlog_rcv		= tcp_v6_do_rcv,
E
Eric Dumazet 已提交
1945
	.release_cb		= tcp_release_cb,
1946
	.mtu_reduced		= tcp_v6_mtu_reduced,
L
Linus Torvalds 已提交
1947
	.hash			= tcp_v6_hash,
1948 1949
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
1950 1951 1952 1953
	.enter_memory_pressure	= tcp_enter_memory_pressure,
	.sockets_allocated	= &tcp_sockets_allocated,
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
1954
	.orphan_count		= &tcp_orphan_count,
L
Linus Torvalds 已提交
1955 1956 1957 1958
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp6_sock),
1959
	.slab_flags		= SLAB_DESTROY_BY_RCU,
1960
	.twsk_prot		= &tcp6_timewait_sock_ops,
1961
	.rsk_prot		= &tcp6_request_sock_ops,
1962
	.h.hashinfo		= &tcp_hashinfo,
1963
	.no_autobind		= true,
1964 1965 1966 1967
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
#endif
A
Andrew Morton 已提交
1968
#ifdef CONFIG_MEMCG_KMEM
G
Glauber Costa 已提交
1969 1970
	.proto_cgroup		= tcp_proto_cgroup,
#endif
L
Linus Torvalds 已提交
1971 1972
};

1973
static const struct inet6_protocol tcpv6_protocol = {
E
Eric Dumazet 已提交
1974
	.early_demux	=	tcp_v6_early_demux,
L
Linus Torvalds 已提交
1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985
	.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,
1986 1987
	.flags		=	INET_PROTOSW_PERMANENT |
				INET_PROTOSW_ICSK,
L
Linus Torvalds 已提交
1988 1989
};

1990
static int __net_init tcpv6_net_init(struct net *net)
1991
{
1992 1993
	return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
				    SOCK_RAW, IPPROTO_TCP, net);
1994 1995
}

1996
static void __net_exit tcpv6_net_exit(struct net *net)
1997
{
1998
	inet_ctl_sock_destroy(net->ipv6.tcp_sk);
E
Eric W. Biederman 已提交
1999 2000
}

2001
static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
E
Eric W. Biederman 已提交
2002 2003
{
	inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6);
2004 2005 2006
}

static struct pernet_operations tcpv6_net_ops = {
E
Eric W. Biederman 已提交
2007 2008 2009
	.init	    = tcpv6_net_init,
	.exit	    = tcpv6_net_exit,
	.exit_batch = tcpv6_net_exit_batch,
2010 2011
};

2012
int __init tcpv6_init(void)
L
Linus Torvalds 已提交
2013
{
2014 2015
	int ret;

2016 2017
	ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
	if (ret)
2018
		goto out;
2019

L
Linus Torvalds 已提交
2020
	/* register inet6 protocol */
2021 2022 2023 2024
	ret = inet6_register_protosw(&tcpv6_protosw);
	if (ret)
		goto out_tcpv6_protocol;

2025
	ret = register_pernet_subsys(&tcpv6_net_ops);
2026 2027 2028 2029
	if (ret)
		goto out_tcpv6_protosw;
out:
	return ret;
2030

2031 2032
out_tcpv6_protosw:
	inet6_unregister_protosw(&tcpv6_protosw);
2033 2034
out_tcpv6_protocol:
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2035 2036 2037
	goto out;
}

2038
void tcpv6_exit(void)
2039
{
2040
	unregister_pernet_subsys(&tcpv6_net_ops);
2041 2042
	inet6_unregister_protosw(&tcpv6_protosw);
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
L
Linus Torvalds 已提交
2043
}