tcp_ipv6.c 49.4 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 *	TCP over IPv6
3
 *	Linux INET6 implementation
L
Linus Torvalds 已提交
4 5
 *
 *	Authors:
6
 *	Pedro Roque		<roque@di.fc.ul.pt>
L
Linus Torvalds 已提交
7
 *
8
 *	Based on:
L
Linus Torvalds 已提交
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
 *	linux/net/ipv4/tcp.c
 *	linux/net/ipv4/tcp_input.c
 *	linux/net/ipv4/tcp_output.c
 *
 *	Fixes:
 *	Hideaki YOSHIFUJI	:	sin6_scope_id support
 *	YOSHIFUJI Hideaki @USAGI and:	Support IPV6_V6ONLY socket option, which
 *	Alexey Kuznetsov		allow both IPv4 and IPv6 sockets to bind
 *					a single port at the same time.
 *	YOSHIFUJI Hideaki @USAGI:	convert /proc/net/tcp6 to seq_file.
 *
 *	This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */

H
Herbert Xu 已提交
26
#include <linux/bottom_half.h>
L
Linus Torvalds 已提交
27 28 29 30 31 32 33 34 35 36 37 38 39 40
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/types.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/net.h>
#include <linux/jiffies.h>
#include <linux/in.h>
#include <linux/in6.h>
#include <linux/netdevice.h>
#include <linux/init.h>
#include <linux/jhash.h>
#include <linux/ipsec.h>
#include <linux/times.h>
41
#include <linux/slab.h>
W
Wang Yufen 已提交
42
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
43 44 45 46 47 48
#include <linux/ipv6.h>
#include <linux/icmpv6.h>
#include <linux/random.h>

#include <net/tcp.h>
#include <net/ndisc.h>
49
#include <net/inet6_hashtables.h>
50
#include <net/inet6_connection_sock.h>
L
Linus Torvalds 已提交
51 52 53 54 55 56 57 58 59 60
#include <net/ipv6.h>
#include <net/transp_v6.h>
#include <net/addrconf.h>
#include <net/ip6_route.h>
#include <net/ip6_checksum.h>
#include <net/inet_ecn.h>
#include <net/protocol.h>
#include <net/xfrm.h>
#include <net/snmp.h>
#include <net/dsfield.h>
61
#include <net/timewait_sock.h>
62
#include <net/inet_common.h>
63
#include <net/secure_seq.h>
G
Glauber Costa 已提交
64
#include <net/tcp_memcontrol.h>
65
#include <net/busy_poll.h>
L
Linus Torvalds 已提交
66 67 68 69

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

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

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

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

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

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

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

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

107
static __u32 tcp_v6_init_sequence(const struct sk_buff *skb)
L
Linus Torvalds 已提交
108
{
109 110
	return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
					    ipv6_hdr(skb)->saddr.s6_addr32,
111 112
					    tcp_hdr(skb)->dest,
					    tcp_hdr(skb)->source);
L
Linus Torvalds 已提交
113 114
}

115
static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
L
Linus Torvalds 已提交
116 117 118
			  int addr_len)
{
	struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
119
	struct inet_sock *inet = inet_sk(sk);
120
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
121 122
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
123
	struct in6_addr *saddr = NULL, *final_p, final;
124
	struct rt6_info *rt;
125
	struct flowi6 fl6;
L
Linus Torvalds 已提交
126 127 128 129
	struct dst_entry *dst;
	int addr_type;
	int err;

130
	if (addr_len < SIN6_LEN_RFC2133)
L
Linus Torvalds 已提交
131 132
		return -EINVAL;

133
	if (usin->sin6_family != AF_INET6)
E
Eric Dumazet 已提交
134
		return -EAFNOSUPPORT;
L
Linus Torvalds 已提交
135

136
	memset(&fl6, 0, sizeof(fl6));
L
Linus Torvalds 已提交
137 138

	if (np->sndflow) {
139 140 141
		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
142
			struct ip6_flowlabel *flowlabel;
143
			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
144
			if (!flowlabel)
L
Linus Torvalds 已提交
145 146 147 148 149 150
				return -EINVAL;
			fl6_sock_release(flowlabel);
		}
	}

	/*
151 152 153
	 *	connect() to INADDR_ANY means loopback (BSD'ism).
	 */

W
Weilong Chen 已提交
154
	if (ipv6_addr_any(&usin->sin6_addr))
155
		usin->sin6_addr.s6_addr[15] = 0x1;
L
Linus Torvalds 已提交
156 157 158

	addr_type = ipv6_addr_type(&usin->sin6_addr);

W
Weilong Chen 已提交
159
	if (addr_type & IPV6_ADDR_MULTICAST)
L
Linus Torvalds 已提交
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180
		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 &&
181
	    !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) {
L
Linus Torvalds 已提交
182 183 184 185 186
		tp->rx_opt.ts_recent = 0;
		tp->rx_opt.ts_recent_stamp = 0;
		tp->write_seq = 0;
	}

187
	sk->sk_v6_daddr = usin->sin6_addr;
188
	np->flow_label = fl6.flowlabel;
L
Linus Torvalds 已提交
189 190 191 192 193 194

	/*
	 *	TCP over IPv4
	 */

	if (addr_type == IPV6_ADDR_MAPPED) {
195
		u32 exthdrlen = icsk->icsk_ext_hdr_len;
L
Linus Torvalds 已提交
196 197 198 199 200 201 202 203 204 205 206
		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];

207
		icsk->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
208
		sk->sk_backlog_rcv = tcp_v4_do_rcv;
209 210 211
#ifdef CONFIG_TCP_MD5SIG
		tp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif
L
Linus Torvalds 已提交
212 213 214 215

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

		if (err) {
216 217
			icsk->icsk_ext_hdr_len = exthdrlen;
			icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
218
			sk->sk_backlog_rcv = tcp_v6_do_rcv;
219 220 221
#ifdef CONFIG_TCP_MD5SIG
			tp->af_specific = &tcp_sock_ipv6_specific;
#endif
L
Linus Torvalds 已提交
222 223
			goto failure;
		}
224
		np->saddr = sk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
225 226 227 228

		return err;
	}

229 230
	if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
		saddr = &sk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
231

232
	fl6.flowi6_proto = IPPROTO_TCP;
233
	fl6.daddr = sk->sk_v6_daddr;
A
Alexey Dobriyan 已提交
234
	fl6.saddr = saddr ? *saddr : np->saddr;
235 236
	fl6.flowi6_oif = sk->sk_bound_dev_if;
	fl6.flowi6_mark = sk->sk_mark;
237 238
	fl6.fl6_dport = usin->sin6_port;
	fl6.fl6_sport = inet->inet_sport;
L
Linus Torvalds 已提交
239

240
	final_p = fl6_update_dst(&fl6, np->opt, &final);
L
Linus Torvalds 已提交
241

242
	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
V
Venkat Yekkirala 已提交
243

244
	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
245 246
	if (IS_ERR(dst)) {
		err = PTR_ERR(dst);
L
Linus Torvalds 已提交
247
		goto failure;
248
	}
L
Linus Torvalds 已提交
249

250
	if (!saddr) {
251
		saddr = &fl6.saddr;
252
		sk->sk_v6_rcv_saddr = *saddr;
L
Linus Torvalds 已提交
253 254 255
	}

	/* set the source address */
A
Alexey Dobriyan 已提交
256
	np->saddr = *saddr;
E
Eric Dumazet 已提交
257
	inet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
258

H
Herbert Xu 已提交
259
	sk->sk_gso_type = SKB_GSO_TCPV6;
260
	__ip6_dst_store(sk, dst, NULL, NULL);
L
Linus Torvalds 已提交
261

262 263 264
	rt = (struct rt6_info *) dst;
	if (tcp_death_row.sysctl_tw_recycle &&
	    !tp->rx_opt.ts_recent_stamp &&
265
	    ipv6_addr_equal(&rt->rt6i_dst.addr, &sk->sk_v6_daddr))
266
		tcp_fetch_timewait_stamp(sk, dst);
267

268
	icsk->icsk_ext_hdr_len = 0;
L
Linus Torvalds 已提交
269
	if (np->opt)
270 271
		icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
					  np->opt->opt_nflen);
L
Linus Torvalds 已提交
272 273 274

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

E
Eric Dumazet 已提交
275
	inet->inet_dport = usin->sin6_port;
L
Linus Torvalds 已提交
276 277

	tcp_set_state(sk, TCP_SYN_SENT);
278
	err = inet6_hash_connect(&tcp_death_row, sk);
L
Linus Torvalds 已提交
279 280 281
	if (err)
		goto late_failure;

282 283
	ip6_set_txhash(sk);

A
Andrey Vagin 已提交
284
	if (!tp->write_seq && likely(!tp->repair))
L
Linus Torvalds 已提交
285
		tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
286
							     sk->sk_v6_daddr.s6_addr32,
E
Eric Dumazet 已提交
287 288
							     inet->inet_sport,
							     inet->inet_dport);
L
Linus Torvalds 已提交
289 290 291 292 293 294 295 296 297 298 299

	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
300
	inet->inet_dport = 0;
L
Linus Torvalds 已提交
301 302 303 304
	sk->sk_route_caps = 0;
	return err;
}

305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321
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
322
static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
323
		u8 type, u8 code, int offset, __be32 info)
L
Linus Torvalds 已提交
324
{
W
Weilong Chen 已提交
325
	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
326
	const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
327 328
	struct net *net = dev_net(skb->dev);
	struct request_sock *fastopen;
L
Linus Torvalds 已提交
329
	struct ipv6_pinfo *np;
330
	struct tcp_sock *tp;
331
	__u32 seq, snd_una;
332 333
	struct sock *sk;
	int err;
L
Linus Torvalds 已提交
334

335 336 337 338
	sk = __inet6_lookup_established(net, &tcp_hashinfo,
					&hdr->daddr, th->dest,
					&hdr->saddr, ntohs(th->source),
					skb->dev->ifindex);
L
Linus Torvalds 已提交
339

340
	if (!sk) {
341 342
		ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
				   ICMP6_MIB_INERRORS);
L
Linus Torvalds 已提交
343 344 345 346
		return;
	}

	if (sk->sk_state == TCP_TIME_WAIT) {
347
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
348 349
		return;
	}
350 351 352
	seq = ntohl(th->seq);
	if (sk->sk_state == TCP_NEW_SYN_RECV)
		return tcp_req_err(sk, seq);
L
Linus Torvalds 已提交
353 354

	bh_lock_sock(sk);
355
	if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG)
356
		NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
L
Linus Torvalds 已提交
357 358 359 360

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

361 362 363 364 365
	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
366
	tp = tcp_sk(sk);
367 368 369
	/* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
	fastopen = tp->fastopen_rsk;
	snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
L
Linus Torvalds 已提交
370
	if (sk->sk_state != TCP_LISTEN &&
371
	    !between(seq, snd_una, tp->snd_nxt)) {
372
		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
373 374 375 376 377
		goto out;
	}

	np = inet6_sk(sk);

378 379 380
	if (type == NDISC_REDIRECT) {
		struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);

381
		if (dst)
382
			dst->ops->redirect(dst, sk, skb);
383
		goto out;
384 385
	}

L
Linus Torvalds 已提交
386
	if (type == ICMPV6_PKT_TOOBIG) {
387 388 389 390 391 392 393
		/* 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;

394 395 396
		if (!ip6_sk_accept_pmtu(sk))
			goto out;

397 398 399
		tp->mtu_info = ntohl(info);
		if (!sock_owned_by_user(sk))
			tcp_v6_mtu_reduced(sk);
400 401 402
		else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED,
					   &tp->tsq_flags))
			sock_hold(sk);
L
Linus Torvalds 已提交
403 404 405 406 407
		goto out;
	}

	icmpv6_err_convert(type, code, &err);

408
	/* Might be for an request_sock */
L
Linus Torvalds 已提交
409 410
	switch (sk->sk_state) {
	case TCP_SYN_SENT:
411 412 413 414
	case TCP_SYN_RECV:
		/* Only in fast or simultaneous open. If a fast open socket is
		 * is already accepted it is treated as a connected one below.
		 */
415
		if (fastopen && !fastopen->sk)
416 417
			break;

L
Linus Torvalds 已提交
418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439
		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);
}


440
static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst,
441
			      struct flowi *fl,
442
			      struct request_sock *req,
443 444
			      u16 queue_mapping,
			      struct tcp_fastopen_cookie *foc)
L
Linus Torvalds 已提交
445
{
446
	struct inet_request_sock *ireq = inet_rsk(req);
L
Linus Torvalds 已提交
447
	struct ipv6_pinfo *np = inet6_sk(sk);
448
	struct flowi6 *fl6 = &fl->u.ip6;
W
Weilong Chen 已提交
449
	struct sk_buff *skb;
450
	int err = -ENOMEM;
L
Linus Torvalds 已提交
451

452 453
	/* First, grab a route. */
	if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL)
454
		goto done;
455

456
	skb = tcp_make_synack(sk, dst, req, foc);
457

L
Linus Torvalds 已提交
458
	if (skb) {
459 460
		__tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr,
				    &ireq->ir_v6_rmt_addr);
L
Linus Torvalds 已提交
461

462
		fl6->daddr = ireq->ir_v6_rmt_addr;
463
		if (np->repflow && ireq->pktopts)
464 465
			fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts));

466
		skb_set_queue_mapping(skb, queue_mapping);
467
		err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass);
468
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
469 470 471 472 473 474
	}

done:
	return err;
}

475

476
static void tcp_v6_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
477
{
478
	kfree_skb(inet_rsk(req)->pktopts);
L
Linus Torvalds 已提交
479 480
}

481 482
#ifdef CONFIG_TCP_MD5SIG
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
483
						   const struct in6_addr *addr)
484
{
E
Eric Dumazet 已提交
485
	return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
486 487 488
}

static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
489
						const struct sock *addr_sk)
490
{
491
	return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr);
492 493
}

W
Wang Yufen 已提交
494 495
static int tcp_v6_parse_md5_keys(struct sock *sk, char __user *optval,
				 int optlen)
496 497 498 499 500 501 502 503 504 505 506 507 508 509
{
	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
510
		if (ipv6_addr_v4mapped(&sin6->sin6_addr))
E
Eric Dumazet 已提交
511 512 513 514
			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);
515 516 517 518 519
	}

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

E
Eric Dumazet 已提交
520 521 522
	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);
523

E
Eric Dumazet 已提交
524 525
	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
			      AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
526 527
}

528
static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
529 530
					const struct in6_addr *daddr,
					const struct in6_addr *saddr, int nbytes)
531 532
{
	struct tcp6_pseudohdr *bp;
533
	struct scatterlist sg;
534

535 536
	bp = &hp->md5_blk.ip6;
	/* 1. TCP pseudo-header (RFC2460) */
A
Alexey Dobriyan 已提交
537 538
	bp->saddr = *saddr;
	bp->daddr = *daddr;
539
	bp->protocol = cpu_to_be32(IPPROTO_TCP);
A
Adam Langley 已提交
540
	bp->len = cpu_to_be32(nbytes);
541

542 543 544
	sg_init_one(&sg, bp, sizeof(*bp));
	return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
}
545

546
static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
547
			       const struct in6_addr *daddr, struct in6_addr *saddr,
E
Eric Dumazet 已提交
548
			       const struct tcphdr *th)
549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566
{
	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))
567 568 569 570
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
571

572 573 574 575
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
576
	return 1;
577 578
}

579 580
static int tcp_v6_md5_hash_skb(char *md5_hash,
			       const struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
581 582
			       const struct sock *sk,
			       const struct sk_buff *skb)
583
{
584
	const struct in6_addr *saddr, *daddr;
585 586
	struct tcp_md5sig_pool *hp;
	struct hash_desc *desc;
E
Eric Dumazet 已提交
587
	const struct tcphdr *th = tcp_hdr(skb);
588

589 590
	if (sk) { /* valid for establish/request sockets */
		saddr = &sk->sk_v6_rcv_saddr;
591
		daddr = &sk->sk_v6_daddr;
592
	} else {
593
		const struct ipv6hdr *ip6h = ipv6_hdr(skb);
594 595
		saddr = &ip6h->saddr;
		daddr = &ip6h->daddr;
596
	}
597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624

	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;
625 626
}

627
static bool tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
628
{
629
	const __u8 *hash_location = NULL;
630
	struct tcp_md5sig_key *hash_expected;
631
	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
E
Eric Dumazet 已提交
632
	const struct tcphdr *th = tcp_hdr(skb);
633 634 635 636
	int genhash;
	u8 newhash[16];

	hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
637
	hash_location = tcp_parse_md5sig_option(th);
638

639 640
	/* We've parsed the options - do we have a hash? */
	if (!hash_expected && !hash_location)
641
		return false;
642 643 644

	if (hash_expected && !hash_location) {
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
645
		return true;
646 647
	}

648 649
	if (!hash_expected && hash_location) {
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
650
		return true;
651 652 653
	}

	/* check the signature */
654 655
	genhash = tcp_v6_md5_hash_skb(newhash,
				      hash_expected,
656
				      NULL, skb);
657

658
	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
659 660 661 662
		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));
663
		return true;
664
	}
665
	return false;
666 667 668
}
#endif

669 670 671 672 673 674 675 676 677 678 679 680
static void tcp_v6_init_req(struct request_sock *req, struct sock *sk,
			    struct sk_buff *skb)
{
	struct inet_request_sock *ireq = inet_rsk(req);
	struct ipv6_pinfo *np = inet6_sk(sk);

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

	/* So that link locals have meaning */
	if (!sk->sk_bound_dev_if &&
	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
E
Eric Dumazet 已提交
681
		ireq->ir_iif = tcp_v6_iif(skb);
682

683
	if (!TCP_SKB_CB(skb)->tcp_tw_isn &&
684 685
	    (ipv6_opt_accepted(sk, skb, &TCP_SKB_CB(skb)->header.h6) ||
	     np->rxopt.bits.rxinfo ||
686 687 688 689 690 691 692
	     np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim ||
	     np->rxopt.bits.rxohlim || np->repflow)) {
		atomic_inc(&skb->users);
		ireq->pktopts = skb;
	}
}

693 694 695 696 697 698 699 700 701
static struct dst_entry *tcp_v6_route_req(struct sock *sk, struct flowi *fl,
					  const struct request_sock *req,
					  bool *strict)
{
	if (strict)
		*strict = true;
	return inet6_csk_route_req(sk, &fl->u.ip6, req);
}

702
struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
703
	.family		=	AF_INET6,
704
	.obj_size	=	sizeof(struct tcp6_request_sock),
705
	.rtx_syn_ack	=	tcp_rtx_synack,
706 707
	.send_ack	=	tcp_v6_reqsk_send_ack,
	.destructor	=	tcp_v6_reqsk_destructor,
708
	.send_reset	=	tcp_v6_send_reset,
W
Wang Yufen 已提交
709
	.syn_ack_timeout =	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
710 711
};

712
static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
713 714
	.mss_clamp	=	IPV6_MIN_MTU - sizeof(struct tcphdr) -
				sizeof(struct ipv6hdr),
715
#ifdef CONFIG_TCP_MD5SIG
716
	.req_md5_lookup	=	tcp_v6_md5_lookup,
717
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
718
#endif
719
	.init_req	=	tcp_v6_init_req,
720 721 722
#ifdef CONFIG_SYN_COOKIES
	.cookie_init_seq =	cookie_v6_init_sequence,
#endif
723
	.route_req	=	tcp_v6_route_req,
724
	.init_seq	=	tcp_v6_init_sequence,
725
	.send_synack	=	tcp_v6_send_synack,
726
	.queue_hash_add =	inet6_csk_reqsk_queue_hash_add,
727
};
728

729 730 731 732
static void tcp_v6_send_response(struct sock *sk, struct sk_buff *skb, u32 seq,
				 u32 ack, u32 win, u32 tsval, u32 tsecr,
				 int oif, struct tcp_md5sig_key *key, int rst,
				 u8 tclass, u32 label)
L
Linus Torvalds 已提交
733
{
734 735
	const struct tcphdr *th = tcp_hdr(skb);
	struct tcphdr *t1;
L
Linus Torvalds 已提交
736
	struct sk_buff *buff;
737
	struct flowi6 fl6;
738
	struct net *net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
739
	struct sock *ctl_sk = net->ipv6.tcp_sk;
740
	unsigned int tot_len = sizeof(struct tcphdr);
E
Eric Dumazet 已提交
741
	struct dst_entry *dst;
742
	__be32 *topt;
L
Linus Torvalds 已提交
743

744
	if (tsecr)
745
		tot_len += TCPOLEN_TSTAMP_ALIGNED;
746 747 748 749 750 751
#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
752
			 GFP_ATOMIC);
753
	if (!buff)
754
		return;
L
Linus Torvalds 已提交
755

756
	skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
L
Linus Torvalds 已提交
757

758
	t1 = (struct tcphdr *) skb_push(buff, tot_len);
759
	skb_reset_transport_header(buff);
L
Linus Torvalds 已提交
760 761 762 763 764

	/* Swap the send and the receive. */
	memset(t1, 0, sizeof(*t1));
	t1->dest = th->source;
	t1->source = th->dest;
765
	t1->doff = tot_len / 4;
766 767 768 769 770
	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
771

772 773
	topt = (__be32 *)(t1 + 1);

774
	if (tsecr) {
775 776
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
777 778
		*topt++ = htonl(tsval);
		*topt++ = htonl(tsecr);
779 780
	}

781 782
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
783 784 785
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
		tcp_v6_md5_hash_hdr((__u8 *)topt, key,
786 787
				    &ipv6_hdr(skb)->saddr,
				    &ipv6_hdr(skb)->daddr, t1);
788 789 790
	}
#endif

791
	memset(&fl6, 0, sizeof(fl6));
A
Alexey Dobriyan 已提交
792 793
	fl6.daddr = ipv6_hdr(skb)->saddr;
	fl6.saddr = ipv6_hdr(skb)->daddr;
794
	fl6.flowlabel = label;
L
Linus Torvalds 已提交
795

796 797 798
	buff->ip_summed = CHECKSUM_PARTIAL;
	buff->csum = 0;

799
	__tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
L
Linus Torvalds 已提交
800

801
	fl6.flowi6_proto = IPPROTO_TCP;
802
	if (rt6_need_strict(&fl6.daddr) && !oif)
E
Eric Dumazet 已提交
803
		fl6.flowi6_oif = tcp_v6_iif(skb);
W
Wang Yufen 已提交
804 805
	else
		fl6.flowi6_oif = oif;
806
	fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark);
807 808
	fl6.fl6_dport = t1->dest;
	fl6.fl6_sport = t1->source;
809
	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
L
Linus Torvalds 已提交
810

811 812 813 814
	/* Pass a socket to ip6_dst_lookup either it is for RST
	 * Underlying function will use this to retrieve the network
	 * namespace
	 */
815
	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
816 817
	if (!IS_ERR(dst)) {
		skb_dst_set(buff, dst);
818
		ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass);
819 820 821 822
		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
823 824 825 826 827
	}

	kfree_skb(buff);
}

828
static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
829
{
830
	const struct tcphdr *th = tcp_hdr(skb);
831
	u32 seq = 0, ack_seq = 0;
832
	struct tcp_md5sig_key *key = NULL;
833 834 835 836 837 838 839
#ifdef CONFIG_TCP_MD5SIG
	const __u8 *hash_location = NULL;
	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
	unsigned char newhash[16];
	int genhash;
	struct sock *sk1 = NULL;
#endif
W
Wang Yufen 已提交
840
	int oif;
L
Linus Torvalds 已提交
841

842
	if (th->rst)
L
Linus Torvalds 已提交
843 844
		return;

845 846 847 848
	/* If sk not NULL, it means we did a successful lookup and incoming
	 * route had to be correct. prequeue might have dropped our dst.
	 */
	if (!sk && !ipv6_unicast_destination(skb))
849
		return;
L
Linus Torvalds 已提交
850

851
#ifdef CONFIG_TCP_MD5SIG
852 853 854 855 856 857 858 859 860 861
	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),
862 863
					   &tcp_hashinfo, &ipv6h->saddr,
					   th->source, &ipv6h->daddr,
E
Eric Dumazet 已提交
864
					   ntohs(th->source), tcp_v6_iif(skb));
865 866 867 868 869 870 871 872
		if (!sk1)
			return;

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

873
		genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, skb);
874 875 876 877 878
		if (genhash || memcmp(hash_location, newhash, 16) != 0)
			goto release_sk1;
	} else {
		key = sk ? tcp_v6_md5_do_lookup(sk, &ipv6h->saddr) : NULL;
	}
879 880
#endif

881 882 883 884 885
	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
886

W
Wang Yufen 已提交
887
	oif = sk ? sk->sk_bound_dev_if : 0;
888
	tcp_v6_send_response(sk, skb, seq, ack_seq, 0, 0, 0, oif, key, 1, 0, 0);
889 890 891 892 893 894 895 896

#ifdef CONFIG_TCP_MD5SIG
release_sk1:
	if (sk1) {
		rcu_read_unlock();
		sock_put(sk1);
	}
#endif
897
}
L
Linus Torvalds 已提交
898

899 900
static void tcp_v6_send_ack(struct sock *sk, struct sk_buff *skb, u32 seq,
			    u32 ack, u32 win, u32 tsval, u32 tsecr, int oif,
901 902
			    struct tcp_md5sig_key *key, u8 tclass,
			    u32 label)
903
{
904 905
	tcp_v6_send_response(sk, skb, seq, ack, win, tsval, tsecr, oif, key, 0,
			     tclass, label);
L
Linus Torvalds 已提交
906 907 908 909
}

static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
910
	struct inet_timewait_sock *tw = inet_twsk(sk);
911
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
912

913
	tcp_v6_send_ack(sk, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
914
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
915
			tcp_time_stamp + tcptw->tw_ts_offset,
W
Wang Yufen 已提交
916
			tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw),
917
			tw->tw_tclass, (tw->tw_flowlabel << 12));
L
Linus Torvalds 已提交
918

919
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
920 921
}

922 923
static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
				  struct request_sock *req)
L
Linus Torvalds 已提交
924
{
925 926 927
	/* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
	 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
	 */
928
	tcp_v6_send_ack(sk, skb, (sk->sk_state == TCP_LISTEN) ?
929
			tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt,
930 931
			tcp_rsk(req)->rcv_nxt, req->rcv_wnd,
			tcp_time_stamp, req->ts_recent, sk->sk_bound_dev_if,
932 933
			tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr),
			0, 0);
L
Linus Torvalds 已提交
934 935 936
}


W
Weilong Chen 已提交
937
static struct sock *tcp_v6_hnd_req(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
938
{
939
	const struct tcphdr *th = tcp_hdr(skb);
940
	struct request_sock *req;
L
Linus Torvalds 已提交
941 942 943
	struct sock *nsk;

	/* Find possible connection requests. */
944
	req = inet6_csk_search_req(sk, th->source,
945
				   &ipv6_hdr(skb)->saddr,
E
Eric Dumazet 已提交
946
				   &ipv6_hdr(skb)->daddr, tcp_v6_iif(skb));
947 948
	if (req) {
		nsk = tcp_check_req(sk, skb, req, false);
949 950
		if (!nsk)
			reqsk_put(req);
951 952
		return nsk;
	}
953
	nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
E
Eric Dumazet 已提交
954 955 956
					 &ipv6_hdr(skb)->saddr, th->source,
					 &ipv6_hdr(skb)->daddr, ntohs(th->dest),
					 tcp_v6_iif(skb));
L
Linus Torvalds 已提交
957 958 959 960 961 962

	if (nsk) {
		if (nsk->sk_state != TCP_TIME_WAIT) {
			bh_lock_sock(nsk);
			return nsk;
		}
963
		inet_twsk_put(inet_twsk(nsk));
L
Linus Torvalds 已提交
964 965 966
		return NULL;
	}

967
#ifdef CONFIG_SYN_COOKIES
968
	if (!th->syn)
969
		sk = cookie_v6_check(sk, skb);
L
Linus Torvalds 已提交
970 971 972 973 974 975 976 977 978 979
#endif
	return sk;
}

static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
{
	if (skb->protocol == htons(ETH_P_IP))
		return tcp_v4_conn_request(sk, skb);

	if (!ipv6_unicast_destination(skb))
980
		goto drop;
L
Linus Torvalds 已提交
981

O
Octavian Purdila 已提交
982 983
	return tcp_conn_request(&tcp6_request_sock_ops,
				&tcp_request_sock_ipv6_ops, sk, skb);
L
Linus Torvalds 已提交
984 985

drop:
986
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
987 988 989
	return 0; /* don't send reset */
}

W
Weilong Chen 已提交
990 991 992
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
993
{
994
	struct inet_request_sock *ireq;
L
Linus Torvalds 已提交
995 996 997 998 999
	struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
	struct tcp6_sock *newtcp6sk;
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1000 1001 1002
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1003
	struct flowi6 fl6;
L
Linus Torvalds 已提交
1004 1005 1006 1007 1008 1009 1010 1011

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

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

1012
		if (!newsk)
L
Linus Torvalds 已提交
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023
			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));

1024
		newnp->saddr = newsk->sk_v6_rcv_saddr;
L
Linus Torvalds 已提交
1025

1026
		inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
1027
		newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1028 1029 1030 1031
#ifdef CONFIG_TCP_MD5SIG
		newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif

1032 1033
		newnp->ipv6_ac_list = NULL;
		newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1034 1035
		newnp->pktoptions  = NULL;
		newnp->opt	   = NULL;
E
Eric Dumazet 已提交
1036
		newnp->mcast_oif   = tcp_v6_iif(skb);
1037
		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
1038
		newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1039 1040
		if (np->repflow)
			newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1041

1042 1043 1044 1045
		/*
		 * 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
1046 1047 1048
		 */

		/* It is tricky place. Until this moment IPv4 tcp
1049
		   worked with IPv6 icsk.icsk_af_ops.
L
Linus Torvalds 已提交
1050 1051
		   Sync it now.
		 */
1052
		tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
1053 1054 1055 1056

		return newsk;
	}

1057
	ireq = inet_rsk(req);
L
Linus Torvalds 已提交
1058 1059 1060 1061

	if (sk_acceptq_is_full(sk))
		goto out_overflow;

1062
	if (!dst) {
1063
		dst = inet6_csk_route_req(sk, &fl6, req);
1064
		if (!dst)
L
Linus Torvalds 已提交
1065
			goto out;
1066
	}
L
Linus Torvalds 已提交
1067 1068

	newsk = tcp_create_openreq_child(sk, req, skb);
1069
	if (!newsk)
1070
		goto out_nonewsk;
L
Linus Torvalds 已提交
1071

1072 1073 1074 1075 1076
	/*
	 * 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
1077

1078
	newsk->sk_gso_type = SKB_GSO_TCPV6;
1079
	__ip6_dst_store(newsk, dst, NULL, NULL);
1080
	inet6_sk_rx_dst_set(newsk, skb);
L
Linus Torvalds 已提交
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090

	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));

1091 1092 1093 1094
	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
1095

1096 1097
	ip6_set_txhash(newsk);

1098
	/* Now IPv6 options...
L
Linus Torvalds 已提交
1099 1100 1101

	   First: no IPv4 options.
	 */
1102
	newinet->inet_opt = NULL;
1103
	newnp->ipv6_ac_list = NULL;
1104
	newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1105 1106 1107 1108 1109 1110

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

	/* Clone pktoptions received with SYN */
	newnp->pktoptions = NULL;
1111
	if (ireq->pktopts) {
1112
		newnp->pktoptions = skb_clone(ireq->pktopts,
1113
					      sk_gfp_atomic(sk, GFP_ATOMIC));
1114 1115
		consume_skb(ireq->pktopts);
		ireq->pktopts = NULL;
L
Linus Torvalds 已提交
1116 1117 1118 1119
		if (newnp->pktoptions)
			skb_set_owner_r(newnp->pktoptions, newsk);
	}
	newnp->opt	  = NULL;
E
Eric Dumazet 已提交
1120
	newnp->mcast_oif  = tcp_v6_iif(skb);
1121
	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1122
	newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1123 1124
	if (np->repflow)
		newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1125 1126 1127 1128 1129 1130 1131

	/* 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.
	 */
1132 1133
	if (np->opt)
		newnp->opt = ipv6_dup_options(newsk, np->opt);
L
Linus Torvalds 已提交
1134

1135
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
L
Linus Torvalds 已提交
1136
	if (newnp->opt)
1137 1138
		inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
						     newnp->opt->opt_flen);
L
Linus Torvalds 已提交
1139

1140 1141
	tcp_ca_openreq_child(newsk, dst);

L
Linus Torvalds 已提交
1142
	tcp_sync_mss(newsk, dst_mtu(dst));
1143
	newtp->advmss = dst_metric_advmss(dst);
1144 1145 1146 1147
	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
1148 1149
	tcp_initialize_rcv_mss(newsk);

E
Eric Dumazet 已提交
1150 1151
	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
1152

1153 1154
#ifdef CONFIG_TCP_MD5SIG
	/* Copy over the MD5 key from the original socket */
W
Wang Yufen 已提交
1155
	key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr);
1156
	if (key) {
1157 1158 1159 1160 1161
		/* 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.
		 */
1162
		tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
1163 1164
			       AF_INET6, key->key, key->keylen,
			       sk_gfp_atomic(sk, GFP_ATOMIC));
1165 1166 1167
	}
#endif

1168
	if (__inet_inherit_port(sk, newsk) < 0) {
1169 1170
		inet_csk_prepare_forced_close(newsk);
		tcp_done(newsk);
1171 1172
		goto out;
	}
E
Eric Dumazet 已提交
1173
	__inet_hash(newsk, NULL);
L
Linus Torvalds 已提交
1174 1175 1176 1177

	return newsk;

out_overflow:
1178
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1179
out_nonewsk:
L
Linus Torvalds 已提交
1180
	dst_release(dst);
1181 1182
out:
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210
	return NULL;
}

/* The socket must have it's spinlock held when we get
 * here.
 *
 * We have a potential double-lock case here, so even when
 * doing backlog processing we use the BH locking scheme.
 * This is because we cannot sleep with the original spinlock
 * held.
 */
static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
{
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp;
	struct sk_buff *opt_skb = NULL;

	/* Imagine: socket is IPv6. IPv4 packet arrives,
	   goes to IPv4 receive handler and backlogged.
	   From backlog it always goes here. Kerboom...
	   Fortunately, tcp_rcv_established and rcv_established
	   handle them correctly, but it is not case with
	   tcp_v6_hnd_req and tcp_v6_send_reset().   --ANK
	 */

	if (skb->protocol == htons(ETH_P_IP))
		return tcp_v4_do_rcv(sk, skb);

1211
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
		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.
1230
					       --ANK (980728)
L
Linus Torvalds 已提交
1231 1232
	 */
	if (np->rxopt.all)
1233
		opt_skb = skb_clone(skb, sk_gfp_atomic(sk, GFP_ATOMIC));
L
Linus Torvalds 已提交
1234 1235

	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
E
Eric Dumazet 已提交
1236 1237
		struct dst_entry *dst = sk->sk_rx_dst;

1238
		sock_rps_save_rxhash(sk, skb);
1239
		sk_mark_napi_id(sk, skb);
E
Eric Dumazet 已提交
1240 1241 1242 1243 1244 1245 1246 1247
		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;
			}
		}

1248
		tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
L
Linus Torvalds 已提交
1249 1250 1251 1252 1253
		if (opt_skb)
			goto ipv6_pktoptions;
		return 0;
	}

1254
	if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1255 1256
		goto csum_err;

1257
	if (sk->sk_state == TCP_LISTEN) {
L
Linus Torvalds 已提交
1258 1259 1260 1261 1262 1263 1264 1265 1266
		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
1267
		if (nsk != sk) {
1268
			sock_rps_save_rxhash(nsk, skb);
1269
			sk_mark_napi_id(sk, skb);
L
Linus Torvalds 已提交
1270 1271 1272 1273 1274 1275
			if (tcp_child_process(sk, nsk, skb))
				goto reset;
			if (opt_skb)
				__kfree_skb(opt_skb);
			return 0;
		}
1276
	} else
1277
		sock_rps_save_rxhash(sk, skb);
L
Linus Torvalds 已提交
1278

1279
	if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
L
Linus Torvalds 已提交
1280 1281 1282 1283 1284 1285
		goto reset;
	if (opt_skb)
		goto ipv6_pktoptions;
	return 0;

reset:
1286
	tcp_v6_send_reset(sk, skb);
L
Linus Torvalds 已提交
1287 1288 1289 1290 1291 1292
discard:
	if (opt_skb)
		__kfree_skb(opt_skb);
	kfree_skb(skb);
	return 0;
csum_err:
1293
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS);
1294
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308
	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))) {
1309
		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
E
Eric Dumazet 已提交
1310
			np->mcast_oif = tcp_v6_iif(opt_skb);
1311
		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1312
			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1313
		if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
1314
			np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
1315 1316
		if (np->repflow)
			np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
1317
		if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->header.h6)) {
L
Linus Torvalds 已提交
1318 1319 1320 1321 1322 1323 1324 1325
			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);
		}
	}

1326
	kfree_skb(opt_skb);
L
Linus Torvalds 已提交
1327 1328 1329
	return 0;
}

1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351
static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
			   const struct tcphdr *th)
{
	/* This is tricky: we move IP6CB at its correct location into
	 * TCP_SKB_CB(). It must be done after xfrm6_policy_check(), because
	 * _decode_session6() uses IP6CB().
	 * barrier() makes sure compiler won't play aliasing games.
	 */
	memmove(&TCP_SKB_CB(skb)->header.h6, IP6CB(skb),
		sizeof(struct inet6_skb_parm));
	barrier();

	TCP_SKB_CB(skb)->seq = ntohl(th->seq);
	TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
				    skb->len - th->doff*4);
	TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
	TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
	TCP_SKB_CB(skb)->tcp_tw_isn = 0;
	TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
	TCP_SKB_CB(skb)->sacked = 0;
}

1352 1353 1354 1355 1356 1357 1358 1359 1360
static void tcp_v6_restore_cb(struct sk_buff *skb)
{
	/* We need to move header back to the beginning if xfrm6_policy_check()
	 * and tcp_v6_fill_cb() are going to be called again.
	 */
	memmove(IP6CB(skb), &TCP_SKB_CB(skb)->header.h6,
		sizeof(struct inet6_skb_parm));
}

1361
static int tcp_v6_rcv(struct sk_buff *skb)
L
Linus Torvalds 已提交
1362
{
1363
	const struct tcphdr *th;
1364
	const struct ipv6hdr *hdr;
L
Linus Torvalds 已提交
1365 1366
	struct sock *sk;
	int ret;
1367
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
1368 1369 1370 1371 1372 1373 1374

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

	/*
	 *	Count it even if it's bad.
	 */
1375
	TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1376 1377 1378 1379

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

1380
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1381 1382 1383 1384 1385 1386

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

1387
	if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo))
1388
		goto csum_error;
L
Linus Torvalds 已提交
1389

1390
	th = tcp_hdr(skb);
1391
	hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
1392

E
Eric Dumazet 已提交
1393
	sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest,
1394
				inet6_iif(skb));
L
Linus Torvalds 已提交
1395 1396 1397 1398 1399 1400 1401
	if (!sk)
		goto no_tcp_socket;

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

1402 1403 1404 1405 1406
	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
1407 1408 1409
	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;

1410 1411
	tcp_v6_fill_cb(skb, hdr, th);

1412 1413 1414 1415 1416
#ifdef CONFIG_TCP_MD5SIG
	if (tcp_v6_inbound_md5_hash(sk, skb))
		goto discard_and_relse;
#endif

1417
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1418 1419
		goto discard_and_relse;

E
Eric Dumazet 已提交
1420
	sk_incoming_cpu_update(sk);
L
Linus Torvalds 已提交
1421 1422
	skb->dev = NULL;

1423
	bh_lock_sock_nested(sk);
L
Linus Torvalds 已提交
1424 1425
	ret = 0;
	if (!sock_owned_by_user(sk)) {
D
Dan Williams 已提交
1426
		if (!tcp_prequeue(sk, skb))
1427
			ret = tcp_v6_do_rcv(sk, skb);
1428 1429
	} else if (unlikely(sk_add_backlog(sk, skb,
					   sk->sk_rcvbuf + sk->sk_sndbuf))) {
Z
Zhu Yi 已提交
1430
		bh_unlock_sock(sk);
1431
		NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
Z
Zhu Yi 已提交
1432 1433
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1434 1435 1436 1437 1438 1439 1440 1441 1442
	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;

1443 1444
	tcp_v6_fill_cb(skb, hdr, th);

L
Linus Torvalds 已提交
1445
	if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1446 1447
csum_error:
		TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS);
L
Linus Torvalds 已提交
1448
bad_packet:
1449
		TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1450
	} else {
1451
		tcp_v6_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463
	}

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)) {
1464
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1465 1466 1467
		goto discard_it;
	}

1468 1469
	tcp_v6_fill_cb(skb, hdr, th);

1470
	if (skb->len < (th->doff<<2)) {
1471
		inet_twsk_put(inet_twsk(sk));
1472 1473 1474 1475 1476
		goto bad_packet;
	}
	if (tcp_checksum_complete(skb)) {
		inet_twsk_put(inet_twsk(sk));
		goto csum_error;
L
Linus Torvalds 已提交
1477 1478
	}

1479
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1480 1481 1482 1483
	case TCP_TW_SYN:
	{
		struct sock *sk2;

1484
		sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1485
					    &ipv6_hdr(skb)->saddr, th->source,
1486
					    &ipv6_hdr(skb)->daddr,
E
Eric Dumazet 已提交
1487
					    ntohs(th->dest), tcp_v6_iif(skb));
1488
		if (sk2) {
1489
			struct inet_timewait_sock *tw = inet_twsk(sk);
1490
			inet_twsk_deschedule(tw);
1491
			inet_twsk_put(tw);
L
Linus Torvalds 已提交
1492
			sk = sk2;
1493
			tcp_v6_restore_cb(skb);
L
Linus Torvalds 已提交
1494 1495 1496 1497 1498 1499 1500 1501
			goto process;
		}
		/* Fall through to ACK */
	}
	case TCP_TW_ACK:
		tcp_v6_timewait_ack(sk, skb);
		break;
	case TCP_TW_RST:
1502
		tcp_v6_restore_cb(skb);
L
Linus Torvalds 已提交
1503
		goto no_tcp_socket;
W
Wang Yufen 已提交
1504 1505
	case TCP_TW_SUCCESS:
		;
L
Linus Torvalds 已提交
1506 1507 1508 1509
	}
	goto discard_it;
}

E
Eric Dumazet 已提交
1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
static void tcp_v6_early_demux(struct sk_buff *skb)
{
	const struct ipv6hdr *hdr;
	const struct tcphdr *th;
	struct sock *sk;

	if (skb->pkt_type != PACKET_HOST)
		return;

	if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr)))
		return;

	hdr = ipv6_hdr(skb);
	th = tcp_hdr(skb);

	if (th->doff < sizeof(struct tcphdr) / 4)
		return;

E
Eric Dumazet 已提交
1528
	/* Note : We use inet6_iif() here, not tcp_v6_iif() */
E
Eric Dumazet 已提交
1529 1530 1531 1532 1533 1534 1535
	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;
1536
		if (sk_fullsock(sk)) {
1537
			struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst);
1538

E
Eric Dumazet 已提交
1539
			if (dst)
E
Eric Dumazet 已提交
1540
				dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie);
E
Eric Dumazet 已提交
1541
			if (dst &&
1542
			    inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
E
Eric Dumazet 已提交
1543 1544 1545 1546 1547
				skb_dst_set_noref(skb, dst);
		}
	}
}

1548 1549 1550
static struct timewait_sock_ops tcp6_timewait_sock_ops = {
	.twsk_obj_size	= sizeof(struct tcp6_timewait_sock),
	.twsk_unique	= tcp_twsk_unique,
W
Wang Yufen 已提交
1551
	.twsk_destructor = tcp_twsk_destructor,
1552 1553
};

1554
static const struct inet_connection_sock_af_ops ipv6_specific = {
1555 1556 1557
	.queue_xmit	   = inet6_csk_xmit,
	.send_check	   = tcp_v6_send_check,
	.rebuild_header	   = inet6_sk_rebuild_header,
E
Eric Dumazet 已提交
1558
	.sk_rx_dst_set	   = inet6_sk_rx_dst_set,
1559 1560 1561
	.conn_request	   = tcp_v6_conn_request,
	.syn_recv_sock	   = tcp_v6_syn_recv_sock,
	.net_header_len	   = sizeof(struct ipv6hdr),
1562
	.net_frag_header_len = sizeof(struct frag_hdr),
1563 1564 1565 1566
	.setsockopt	   = ipv6_setsockopt,
	.getsockopt	   = ipv6_getsockopt,
	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in6),
1567
	.bind_conflict	   = inet6_csk_bind_conflict,
1568
#ifdef CONFIG_COMPAT
1569 1570
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1571
#endif
1572
	.mtu_reduced	   = tcp_v6_mtu_reduced,
L
Linus Torvalds 已提交
1573 1574
};

1575
#ifdef CONFIG_TCP_MD5SIG
1576
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1577
	.md5_lookup	=	tcp_v6_md5_lookup,
1578
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
1579 1580
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1581
#endif
1582

L
Linus Torvalds 已提交
1583 1584 1585
/*
 *	TCP over IPv4 via INET6 API
 */
1586
static const struct inet_connection_sock_af_ops ipv6_mapped = {
1587 1588 1589
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
1590
	.sk_rx_dst_set	   = inet_sk_rx_dst_set,
1591 1592 1593 1594 1595 1596 1597
	.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),
1598
	.bind_conflict	   = inet6_csk_bind_conflict,
1599
#ifdef CONFIG_COMPAT
1600 1601
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1602
#endif
1603
	.mtu_reduced	   = tcp_v4_mtu_reduced,
L
Linus Torvalds 已提交
1604 1605
};

1606
#ifdef CONFIG_TCP_MD5SIG
1607
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1608
	.md5_lookup	=	tcp_v4_md5_lookup,
1609
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1610 1611
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1612
#endif
1613

L
Linus Torvalds 已提交
1614 1615 1616 1617 1618
/* 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)
{
1619
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1620

1621
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1622

1623
	icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
1624

1625
#ifdef CONFIG_TCP_MD5SIG
1626
	tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
1627 1628
#endif

L
Linus Torvalds 已提交
1629 1630 1631
	return 0;
}

1632
static void tcp_v6_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1633 1634
{
	tcp_v4_destroy_sock(sk);
1635
	inet6_destroy_sock(sk);
L
Linus Torvalds 已提交
1636 1637
}

1638
#ifdef CONFIG_PROC_FS
L
Linus Torvalds 已提交
1639
/* Proc filesystem TCPv6 sock list dumping. */
1640
static void get_openreq6(struct seq_file *seq,
1641
			 struct request_sock *req, int i, kuid_t uid)
L
Linus Torvalds 已提交
1642
{
1643
	long ttd = req->rsk_timer.expires - jiffies;
1644 1645
	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
1646 1647 1648 1649 1650 1651

	if (ttd < 0)
		ttd = 0;

	seq_printf(seq,
		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1652
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1653 1654 1655
		   i,
		   src->s6_addr32[0], src->s6_addr32[1],
		   src->s6_addr32[2], src->s6_addr32[3],
E
Eric Dumazet 已提交
1656
		   inet_rsk(req)->ir_num,
L
Linus Torvalds 已提交
1657 1658
		   dest->s6_addr32[0], dest->s6_addr32[1],
		   dest->s6_addr32[2], dest->s6_addr32[3],
1659
		   ntohs(inet_rsk(req)->ir_rmt_port),
L
Linus Torvalds 已提交
1660
		   TCP_SYN_RECV,
W
Weilong Chen 已提交
1661
		   0, 0, /* could print option size, but that is af dependent. */
1662 1663
		   1,   /* timers active (only the expire timer) */
		   jiffies_to_clock_t(ttd),
1664
		   req->num_timeout,
1665
		   from_kuid_munged(seq_user_ns(seq), uid),
1666
		   0,  /* non standard timer */
L
Linus Torvalds 已提交
1667 1668 1669 1670 1671 1672
		   0, /* open_requests have no inode */
		   0, req);
}

static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
{
1673
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1674 1675 1676
	__u16 destp, srcp;
	int timer_active;
	unsigned long timer_expires;
1677 1678
	const struct inet_sock *inet = inet_sk(sp);
	const struct tcp_sock *tp = tcp_sk(sp);
1679
	const struct inet_connection_sock *icsk = inet_csk(sp);
1680
	struct fastopen_queue *fastopenq = icsk->icsk_accept_queue.fastopenq;
L
Linus Torvalds 已提交
1681

1682 1683
	dest  = &sp->sk_v6_daddr;
	src   = &sp->sk_v6_rcv_saddr;
E
Eric Dumazet 已提交
1684 1685
	destp = ntohs(inet->inet_dport);
	srcp  = ntohs(inet->inet_sport);
1686 1687

	if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
L
Linus Torvalds 已提交
1688
		timer_active	= 1;
1689 1690
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
1691
		timer_active	= 4;
1692
		timer_expires	= icsk->icsk_timeout;
L
Linus Torvalds 已提交
1693 1694 1695 1696 1697 1698 1699 1700 1701 1702
	} 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 "
1703
		   "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
L
Linus Torvalds 已提交
1704 1705 1706 1707 1708
		   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,
1709
		   sp->sk_state,
1710 1711
		   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
1712
		   timer_active,
1713
		   jiffies_delta_to_clock_t(timer_expires - jiffies),
1714
		   icsk->icsk_retransmits,
1715
		   from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1716
		   icsk->icsk_probes_out,
L
Linus Torvalds 已提交
1717 1718
		   sock_i_ino(sp),
		   atomic_read(&sp->sk_refcnt), sp,
1719 1720
		   jiffies_to_clock_t(icsk->icsk_rto),
		   jiffies_to_clock_t(icsk->icsk_ack.ato),
W
Weilong Chen 已提交
1721
		   (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
I
Ilpo Järvinen 已提交
1722
		   tp->snd_cwnd,
1723 1724 1725
		   sp->sk_state == TCP_LISTEN ?
			(fastopenq ? fastopenq->max_qlen : 0) :
			(tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)
L
Linus Torvalds 已提交
1726 1727 1728
		   );
}

1729
static void get_timewait6_sock(struct seq_file *seq,
1730
			       struct inet_timewait_sock *tw, int i)
L
Linus Torvalds 已提交
1731
{
1732
	long delta = tw->tw_timer.expires - jiffies;
1733
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1734 1735
	__u16 destp, srcp;

1736 1737
	dest = &tw->tw_v6_daddr;
	src  = &tw->tw_v6_rcv_saddr;
L
Linus Torvalds 已提交
1738 1739 1740 1741 1742
	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
1743
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1744 1745 1746 1747 1748 1749
		   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,
1750
		   3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
L
Linus Torvalds 已提交
1751 1752 1753 1754 1755 1756
		   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
1757
	struct sock *sk = v;
L
Linus Torvalds 已提交
1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772

	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
1773 1774 1775 1776
		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
1777 1778
		break;
	case TCP_SEQ_STATE_OPENREQ:
1779
		get_openreq6(seq, v, st->num, st->uid);
L
Linus Torvalds 已提交
1780 1781 1782 1783 1784 1785
		break;
	}
out:
	return 0;
}

1786 1787 1788 1789 1790 1791 1792 1793
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
1794 1795 1796
static struct tcp_seq_afinfo tcp6_seq_afinfo = {
	.name		= "tcp6",
	.family		= AF_INET6,
1797
	.seq_fops	= &tcp6_afinfo_seq_fops,
1798 1799 1800
	.seq_ops	= {
		.show		= tcp6_seq_show,
	},
L
Linus Torvalds 已提交
1801 1802
};

1803
int __net_init tcp6_proc_init(struct net *net)
L
Linus Torvalds 已提交
1804
{
1805
	return tcp_proc_register(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1806 1807
}

1808
void tcp6_proc_exit(struct net *net)
L
Linus Torvalds 已提交
1809
{
1810
	tcp_proc_unregister(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1811 1812 1813
}
#endif

E
Eric Dumazet 已提交
1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824
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
1825 1826 1827 1828 1829 1830
struct proto tcpv6_prot = {
	.name			= "TCPv6",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v6_connect,
	.disconnect		= tcp_disconnect,
1831
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
1832 1833 1834 1835 1836 1837 1838
	.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,
1839 1840
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
1841
	.backlog_rcv		= tcp_v6_do_rcv,
E
Eric Dumazet 已提交
1842
	.release_cb		= tcp_release_cb,
E
Eric Dumazet 已提交
1843
	.hash			= inet_hash,
1844 1845
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
1846
	.enter_memory_pressure	= tcp_enter_memory_pressure,
1847
	.stream_memory_free	= tcp_stream_memory_free,
L
Linus Torvalds 已提交
1848 1849 1850
	.sockets_allocated	= &tcp_sockets_allocated,
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
1851
	.orphan_count		= &tcp_orphan_count,
1852
	.sysctl_mem		= sysctl_tcp_mem,
L
Linus Torvalds 已提交
1853 1854 1855 1856
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp6_sock),
1857
	.slab_flags		= SLAB_DESTROY_BY_RCU,
1858
	.twsk_prot		= &tcp6_timewait_sock_ops,
1859
	.rsk_prot		= &tcp6_request_sock_ops,
1860
	.h.hashinfo		= &tcp_hashinfo,
1861
	.no_autobind		= true,
1862 1863 1864 1865
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
#endif
A
Andrew Morton 已提交
1866
#ifdef CONFIG_MEMCG_KMEM
G
Glauber Costa 已提交
1867 1868
	.proto_cgroup		= tcp_proto_cgroup,
#endif
E
Eric Dumazet 已提交
1869
	.clear_sk		= tcp_v6_clear_sk,
L
Linus Torvalds 已提交
1870 1871
};

1872
static const struct inet6_protocol tcpv6_protocol = {
E
Eric Dumazet 已提交
1873
	.early_demux	=	tcp_v6_early_demux,
L
Linus Torvalds 已提交
1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
	.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,
1884 1885
	.flags		=	INET_PROTOSW_PERMANENT |
				INET_PROTOSW_ICSK,
L
Linus Torvalds 已提交
1886 1887
};

1888
static int __net_init tcpv6_net_init(struct net *net)
1889
{
1890 1891
	return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
				    SOCK_RAW, IPPROTO_TCP, net);
1892 1893
}

1894
static void __net_exit tcpv6_net_exit(struct net *net)
1895
{
1896
	inet_ctl_sock_destroy(net->ipv6.tcp_sk);
E
Eric W. Biederman 已提交
1897 1898
}

1899
static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
E
Eric W. Biederman 已提交
1900 1901
{
	inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6);
1902 1903 1904
}

static struct pernet_operations tcpv6_net_ops = {
E
Eric W. Biederman 已提交
1905 1906 1907
	.init	    = tcpv6_net_init,
	.exit	    = tcpv6_net_exit,
	.exit_batch = tcpv6_net_exit_batch,
1908 1909
};

1910
int __init tcpv6_init(void)
L
Linus Torvalds 已提交
1911
{
1912 1913
	int ret;

1914 1915
	ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
	if (ret)
1916
		goto out;
1917

L
Linus Torvalds 已提交
1918
	/* register inet6 protocol */
1919 1920 1921 1922
	ret = inet6_register_protosw(&tcpv6_protosw);
	if (ret)
		goto out_tcpv6_protocol;

1923
	ret = register_pernet_subsys(&tcpv6_net_ops);
1924 1925 1926 1927
	if (ret)
		goto out_tcpv6_protosw;
out:
	return ret;
1928

1929 1930
out_tcpv6_protosw:
	inet6_unregister_protosw(&tcpv6_protosw);
1931 1932
out_tcpv6_protocol:
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
1933 1934 1935
	goto out;
}

1936
void tcpv6_exit(void)
1937
{
1938
	unregister_pernet_subsys(&tcpv6_net_ops);
1939 1940
	inet6_unregister_protosw(&tcpv6_protosw);
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
L
Linus Torvalds 已提交
1941
}