tcp_ipv6.c 51.6 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 *	TCP over IPv6
3
 *	Linux INET6 implementation
L
Linus Torvalds 已提交
4 5
 *
 *	Authors:
6
 *	Pedro Roque		<roque@di.fc.ul.pt>
L
Linus Torvalds 已提交
7
 *
8
 *	Based on:
L
Linus Torvalds 已提交
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
 *	linux/net/ipv4/tcp.c
 *	linux/net/ipv4/tcp_input.c
 *	linux/net/ipv4/tcp_output.c
 *
 *	Fixes:
 *	Hideaki YOSHIFUJI	:	sin6_scope_id support
 *	YOSHIFUJI Hideaki @USAGI and:	Support IPV6_V6ONLY socket option, which
 *	Alexey Kuznetsov		allow both IPv4 and IPv6 sockets to bind
 *					a single port at the same time.
 *	YOSHIFUJI Hideaki @USAGI:	convert /proc/net/tcp6 to seq_file.
 *
 *	This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */

H
Herbert Xu 已提交
26
#include <linux/bottom_half.h>
L
Linus Torvalds 已提交
27 28 29 30 31 32 33 34 35 36 37 38 39 40
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/types.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/net.h>
#include <linux/jiffies.h>
#include <linux/in.h>
#include <linux/in6.h>
#include <linux/netdevice.h>
#include <linux/init.h>
#include <linux/jhash.h>
#include <linux/ipsec.h>
#include <linux/times.h>
41
#include <linux/slab.h>
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

static int	tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
80
static void	__tcp_v6_send_check(struct sk_buff *skb,
81 82
				    const struct in6_addr *saddr,
				    const struct in6_addr *daddr);
L
Linus Torvalds 已提交
83

84 85
static const struct inet_connection_sock_af_ops ipv6_mapped;
static const struct inet_connection_sock_af_ops ipv6_specific;
86
#ifdef CONFIG_TCP_MD5SIG
87 88
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
89 90
#else
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
91
						   const struct in6_addr *addr)
92 93 94
{
	return NULL;
}
95
#endif
L
Linus Torvalds 已提交
96 97 98 99

static void tcp_v6_hash(struct sock *sk)
{
	if (sk->sk_state != TCP_CLOSE) {
100
		if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
L
Linus Torvalds 已提交
101 102 103 104
			tcp_prot.hash(sk);
			return;
		}
		local_bh_disable();
105
		__inet6_hash(sk, NULL);
L
Linus Torvalds 已提交
106 107 108 109
		local_bh_enable();
	}
}

H
Herbert Xu 已提交
110
static __inline__ __sum16 tcp_v6_check(int len,
111 112
				   const struct in6_addr *saddr,
				   const struct in6_addr *daddr,
113
				   __wsum base)
L
Linus Torvalds 已提交
114 115 116 117
{
	return csum_ipv6_magic(saddr, daddr, len, IPPROTO_TCP, base);
}

118
static __u32 tcp_v6_init_sequence(const struct sk_buff *skb)
L
Linus Torvalds 已提交
119
{
120 121
	return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
					    ipv6_hdr(skb)->saddr.s6_addr32,
122 123
					    tcp_hdr(skb)->dest,
					    tcp_hdr(skb)->source);
L
Linus Torvalds 已提交
124 125
}

126
static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
L
Linus Torvalds 已提交
127 128 129
			  int addr_len)
{
	struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
130
	struct inet_sock *inet = inet_sk(sk);
131
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
132 133
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
134
	struct in6_addr *saddr = NULL, *final_p, final;
135
	struct rt6_info *rt;
136
	struct flowi6 fl6;
L
Linus Torvalds 已提交
137 138 139 140
	struct dst_entry *dst;
	int addr_type;
	int err;

141
	if (addr_len < SIN6_LEN_RFC2133)
L
Linus Torvalds 已提交
142 143
		return -EINVAL;

144
	if (usin->sin6_family != AF_INET6)
E
Eric Dumazet 已提交
145
		return -EAFNOSUPPORT;
L
Linus Torvalds 已提交
146

147
	memset(&fl6, 0, sizeof(fl6));
L
Linus Torvalds 已提交
148 149

	if (np->sndflow) {
150 151 152
		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
153
			struct ip6_flowlabel *flowlabel;
154
			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
L
Linus Torvalds 已提交
155 156
			if (flowlabel == NULL)
				return -EINVAL;
A
Alexey Dobriyan 已提交
157
			usin->sin6_addr = flowlabel->dst;
L
Linus Torvalds 已提交
158 159 160 161 162
			fl6_sock_release(flowlabel);
		}
	}

	/*
163 164 165 166 167
	 *	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
168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198

	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
199
	np->daddr = usin->sin6_addr;
200
	np->flow_label = fl6.flowlabel;
L
Linus Torvalds 已提交
201 202 203 204 205 206

	/*
	 *	TCP over IPv4
	 */

	if (addr_type == IPV6_ADDR_MAPPED) {
207
		u32 exthdrlen = icsk->icsk_ext_hdr_len;
L
Linus Torvalds 已提交
208 209 210 211 212 213 214 215 216 217 218
		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];

219
		icsk->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
220
		sk->sk_backlog_rcv = tcp_v4_do_rcv;
221 222 223
#ifdef CONFIG_TCP_MD5SIG
		tp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif
L
Linus Torvalds 已提交
224 225 226 227

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

		if (err) {
228 229
			icsk->icsk_ext_hdr_len = exthdrlen;
			icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
230
			sk->sk_backlog_rcv = tcp_v6_do_rcv;
231 232 233
#ifdef CONFIG_TCP_MD5SIG
			tp->af_specific = &tcp_sock_ipv6_specific;
#endif
L
Linus Torvalds 已提交
234 235
			goto failure;
		} else {
E
Eric Dumazet 已提交
236 237 238
			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
239 240 241 242 243 244 245 246
		}

		return err;
	}

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

247
	fl6.flowi6_proto = IPPROTO_TCP;
A
Alexey Dobriyan 已提交
248 249
	fl6.daddr = np->daddr;
	fl6.saddr = saddr ? *saddr : np->saddr;
250 251
	fl6.flowi6_oif = sk->sk_bound_dev_if;
	fl6.flowi6_mark = sk->sk_mark;
252 253
	fl6.fl6_dport = usin->sin6_port;
	fl6.fl6_sport = inet->inet_sport;
L
Linus Torvalds 已提交
254

255
	final_p = fl6_update_dst(&fl6, np->opt, &final);
L
Linus Torvalds 已提交
256

257
	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
V
Venkat Yekkirala 已提交
258

259
	dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
260 261
	if (IS_ERR(dst)) {
		err = PTR_ERR(dst);
L
Linus Torvalds 已提交
262
		goto failure;
263
	}
L
Linus Torvalds 已提交
264 265

	if (saddr == NULL) {
266
		saddr = &fl6.saddr;
A
Alexey Dobriyan 已提交
267
		np->rcv_saddr = *saddr;
L
Linus Torvalds 已提交
268 269 270
	}

	/* set the source address */
A
Alexey Dobriyan 已提交
271
	np->saddr = *saddr;
E
Eric Dumazet 已提交
272
	inet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
273

H
Herbert Xu 已提交
274
	sk->sk_gso_type = SKB_GSO_TCPV6;
275
	__ip6_dst_store(sk, dst, NULL, NULL);
L
Linus Torvalds 已提交
276

277 278 279
	rt = (struct rt6_info *) dst;
	if (tcp_death_row.sysctl_tw_recycle &&
	    !tp->rx_opt.ts_recent_stamp &&
280 281
	    ipv6_addr_equal(&rt->rt6i_dst.addr, &np->daddr))
		tcp_fetch_timewait_stamp(sk, dst);
282

283
	icsk->icsk_ext_hdr_len = 0;
L
Linus Torvalds 已提交
284
	if (np->opt)
285 286
		icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
					  np->opt->opt_nflen);
L
Linus Torvalds 已提交
287 288 289

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

E
Eric Dumazet 已提交
290
	inet->inet_dport = usin->sin6_port;
L
Linus Torvalds 已提交
291 292

	tcp_set_state(sk, TCP_SYN_SENT);
293
	err = inet6_hash_connect(&tcp_death_row, sk);
L
Linus Torvalds 已提交
294 295 296 297 298 299
	if (err)
		goto late_failure;

	if (!tp->write_seq)
		tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
							     np->daddr.s6_addr32,
E
Eric Dumazet 已提交
300 301
							     inet->inet_sport,
							     inet->inet_dport);
L
Linus Torvalds 已提交
302 303 304 305 306 307 308 309 310 311 312

	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
313
	inet->inet_dport = 0;
L
Linus Torvalds 已提交
314 315 316 317 318
	sk->sk_route_caps = 0;
	return err;
}

static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
319
		u8 type, u8 code, int offset, __be32 info)
L
Linus Torvalds 已提交
320
{
321
	const struct ipv6hdr *hdr = (const struct ipv6hdr*)skb->data;
322
	const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
L
Linus Torvalds 已提交
323 324 325
	struct ipv6_pinfo *np;
	struct sock *sk;
	int err;
326
	struct tcp_sock *tp;
L
Linus Torvalds 已提交
327
	__u32 seq;
328
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
329

330
	sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr,
331
			th->dest, &hdr->saddr, th->source, skb->dev->ifindex);
L
Linus Torvalds 已提交
332 333

	if (sk == NULL) {
334 335
		ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
				   ICMP6_MIB_INERRORS);
L
Linus Torvalds 已提交
336 337 338 339
		return;
	}

	if (sk->sk_state == TCP_TIME_WAIT) {
340
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
341 342 343 344 345
		return;
	}

	bh_lock_sock(sk);
	if (sock_owned_by_user(sk))
346
		NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
L
Linus Torvalds 已提交
347 348 349 350

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

351 352 353 354 355
	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
356
	tp = tcp_sk(sk);
357
	seq = ntohl(th->seq);
L
Linus Torvalds 已提交
358 359
	if (sk->sk_state != TCP_LISTEN &&
	    !between(seq, tp->snd_una, tp->snd_nxt)) {
360
		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
361 362 363 364 365
		goto out;
	}

	np = inet6_sk(sk);

366 367 368
	if (type == NDISC_REDIRECT) {
		struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);

369
		if (dst)
370
			dst->ops->redirect(dst, sk, skb);
371 372
	}

L
Linus Torvalds 已提交
373
	if (type == ICMPV6_PKT_TOOBIG) {
374
		struct dst_entry *dst;
L
Linus Torvalds 已提交
375 376 377 378 379 380

		if (sock_owned_by_user(sk))
			goto out;
		if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
			goto out;

381 382 383
		dst = inet6_csk_update_pmtu(sk, ntohl(info));
		if (!dst)
			goto out;
384

385
		if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
L
Linus Torvalds 已提交
386 387
			tcp_sync_mss(sk, dst_mtu(dst));
			tcp_simple_retransmit(sk);
388
		}
L
Linus Torvalds 已提交
389 390 391 392 393
		goto out;
	}

	icmpv6_err_convert(type, code, &err);

394
	/* Might be for an request_sock */
L
Linus Torvalds 已提交
395
	switch (sk->sk_state) {
396
		struct request_sock *req, **prev;
L
Linus Torvalds 已提交
397 398 399 400
	case TCP_LISTEN:
		if (sock_owned_by_user(sk))
			goto out;

401 402
		req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
					   &hdr->saddr, inet6_iif(skb));
L
Linus Torvalds 已提交
403 404 405 406 407 408
		if (!req)
			goto out;

		/* ICMPs are not backlogged, hence we cannot get
		 * an established socket here.
		 */
409
		WARN_ON(req->sk != NULL);
L
Linus Torvalds 已提交
410

411
		if (seq != tcp_rsk(req)->snt_isn) {
412
			NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
L
Linus Torvalds 已提交
413 414 415
			goto out;
		}

416
		inet_csk_reqsk_queue_drop(sk, req, prev);
L
Linus Torvalds 已提交
417 418 419 420
		goto out;

	case TCP_SYN_SENT:
	case TCP_SYN_RECV:  /* Cannot happen.
421
			       It can, it SYNs are crossed. --ANK */
L
Linus Torvalds 已提交
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443
		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);
}


444 445
static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst,
			      struct flowi6 *fl6,
446
			      struct request_sock *req,
447 448
			      struct request_values *rvp,
			      u16 queue_mapping)
L
Linus Torvalds 已提交
449
{
450
	struct inet6_request_sock *treq = inet6_rsk(req);
L
Linus Torvalds 已提交
451 452
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct sk_buff * skb;
453
	int err = -ENOMEM;
L
Linus Torvalds 已提交
454

455 456
	/* First, grab a route. */
	if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL)
457
		goto done;
458

459
	skb = tcp_make_synack(sk, dst, req, rvp);
460

L
Linus Torvalds 已提交
461
	if (skb) {
462
		__tcp_v6_send_check(skb, &treq->loc_addr, &treq->rmt_addr);
L
Linus Torvalds 已提交
463

464
		fl6->daddr = treq->rmt_addr;
465
		skb_set_queue_mapping(skb, queue_mapping);
466
		err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass);
467
		err = net_xmit_eval(err);
L
Linus Torvalds 已提交
468 469 470 471 472 473
	}

done:
	return err;
}

474 475 476
static int tcp_v6_rtx_synack(struct sock *sk, struct request_sock *req,
			     struct request_values *rvp)
{
477 478
	struct flowi6 fl6;

479
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS);
480
	return tcp_v6_send_synack(sk, NULL, &fl6, req, rvp, 0);
481 482
}

483
static void tcp_v6_reqsk_destructor(struct request_sock *req)
L
Linus Torvalds 已提交
484
{
485
	kfree_skb(inet6_rsk(req)->pktopts);
L
Linus Torvalds 已提交
486 487
}

488 489
#ifdef CONFIG_TCP_MD5SIG
static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
490
						   const struct in6_addr *addr)
491
{
E
Eric Dumazet 已提交
492
	return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
}

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
523
		if (ipv6_addr_v4mapped(&sin6->sin6_addr))
E
Eric Dumazet 已提交
524 525 526 527
			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);
528 529 530 531 532
	}

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

E
Eric Dumazet 已提交
533 534 535
	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);
536

E
Eric Dumazet 已提交
537 538
	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
			      AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
539 540
}

541
static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
542 543
					const struct in6_addr *daddr,
					const struct in6_addr *saddr, int nbytes)
544 545
{
	struct tcp6_pseudohdr *bp;
546
	struct scatterlist sg;
547

548 549
	bp = &hp->md5_blk.ip6;
	/* 1. TCP pseudo-header (RFC2460) */
A
Alexey Dobriyan 已提交
550 551
	bp->saddr = *saddr;
	bp->daddr = *daddr;
552
	bp->protocol = cpu_to_be32(IPPROTO_TCP);
A
Adam Langley 已提交
553
	bp->len = cpu_to_be32(nbytes);
554

555 556 557
	sg_init_one(&sg, bp, sizeof(*bp));
	return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
}
558

559
static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
560
			       const struct in6_addr *daddr, struct in6_addr *saddr,
E
Eric Dumazet 已提交
561
			       const struct tcphdr *th)
562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579
{
	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))
580 581 582 583
		goto clear_hash;

	tcp_put_md5sig_pool();
	return 0;
584

585 586 587 588
clear_hash:
	tcp_put_md5sig_pool();
clear_hash_noput:
	memset(md5_hash, 0, 16);
589
	return 1;
590 591
}

592
static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
E
Eric Dumazet 已提交
593 594 595
			       const struct sock *sk,
			       const struct request_sock *req,
			       const struct sk_buff *skb)
596
{
597
	const struct in6_addr *saddr, *daddr;
598 599
	struct tcp_md5sig_pool *hp;
	struct hash_desc *desc;
E
Eric Dumazet 已提交
600
	const struct tcphdr *th = tcp_hdr(skb);
601 602 603 604

	if (sk) {
		saddr = &inet6_sk(sk)->saddr;
		daddr = &inet6_sk(sk)->daddr;
605
	} else if (req) {
606 607
		saddr = &inet6_rsk(req)->loc_addr;
		daddr = &inet6_rsk(req)->rmt_addr;
608
	} else {
609
		const struct ipv6hdr *ip6h = ipv6_hdr(skb);
610 611
		saddr = &ip6h->saddr;
		daddr = &ip6h->daddr;
612
	}
613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640

	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;
641 642
}

E
Eric Dumazet 已提交
643
static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
644
{
645
	const __u8 *hash_location = NULL;
646
	struct tcp_md5sig_key *hash_expected;
647
	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
E
Eric Dumazet 已提交
648
	const struct tcphdr *th = tcp_hdr(skb);
649 650 651 652
	int genhash;
	u8 newhash[16];

	hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
653
	hash_location = tcp_parse_md5sig_option(th);
654

655 656 657 658 659 660
	/* 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);
661 662 663
		return 1;
	}

664 665
	if (!hash_expected && hash_location) {
		NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
666 667 668 669
		return 1;
	}

	/* check the signature */
670 671 672 673
	genhash = tcp_v6_md5_hash_skb(newhash,
				      hash_expected,
				      NULL, NULL, skb);

674
	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
675 676 677 678
		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));
679 680 681 682 683 684
		return 1;
	}
	return 0;
}
#endif

685
struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
L
Linus Torvalds 已提交
686
	.family		=	AF_INET6,
687
	.obj_size	=	sizeof(struct tcp6_request_sock),
688
	.rtx_syn_ack	=	tcp_v6_rtx_synack,
689 690
	.send_ack	=	tcp_v6_reqsk_send_ack,
	.destructor	=	tcp_v6_reqsk_destructor,
691 692
	.send_reset	=	tcp_v6_send_reset,
	.syn_ack_timeout = 	tcp_syn_ack_timeout,
L
Linus Torvalds 已提交
693 694
};

695
#ifdef CONFIG_TCP_MD5SIG
696
static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
697
	.md5_lookup	=	tcp_v6_reqsk_md5_lookup,
698
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
699
};
700
#endif
701

702
static void __tcp_v6_send_check(struct sk_buff *skb,
703
				const struct in6_addr *saddr, const struct in6_addr *daddr)
L
Linus Torvalds 已提交
704
{
705
	struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
706

707
	if (skb->ip_summed == CHECKSUM_PARTIAL) {
708
		th->check = ~tcp_v6_check(skb->len, saddr, daddr, 0);
709
		skb->csum_start = skb_transport_header(skb) - skb->head;
A
Al Viro 已提交
710
		skb->csum_offset = offsetof(struct tcphdr, check);
L
Linus Torvalds 已提交
711
	} else {
712 713 714
		th->check = tcp_v6_check(skb->len, saddr, daddr,
					 csum_partial(th, th->doff << 2,
						      skb->csum));
L
Linus Torvalds 已提交
715 716 717
	}
}

718
static void tcp_v6_send_check(struct sock *sk, struct sk_buff *skb)
719 720 721 722 723 724
{
	struct ipv6_pinfo *np = inet6_sk(sk);

	__tcp_v6_send_check(skb, &np->saddr, &np->daddr);
}

725 726
static int tcp_v6_gso_send_check(struct sk_buff *skb)
{
727
	const struct ipv6hdr *ipv6h;
728 729 730 731 732
	struct tcphdr *th;

	if (!pskb_may_pull(skb, sizeof(*th)))
		return -EINVAL;

733
	ipv6h = ipv6_hdr(skb);
734
	th = tcp_hdr(skb);
735 736

	th->check = 0;
737
	skb->ip_summed = CHECKSUM_PARTIAL;
738
	__tcp_v6_send_check(skb, &ipv6h->saddr, &ipv6h->daddr);
739 740
	return 0;
}
L
Linus Torvalds 已提交
741

H
Herbert Xu 已提交
742 743
static struct sk_buff **tcp6_gro_receive(struct sk_buff **head,
					 struct sk_buff *skb)
H
Herbert Xu 已提交
744
{
745
	const struct ipv6hdr *iph = skb_gro_network_header(skb);
H
Herbert Xu 已提交
746 747 748

	switch (skb->ip_summed) {
	case CHECKSUM_COMPLETE:
749
		if (!tcp_v6_check(skb_gro_len(skb), &iph->saddr, &iph->daddr,
H
Herbert Xu 已提交
750 751 752 753 754 755 756 757 758 759 760 761 762 763
				  skb->csum)) {
			skb->ip_summed = CHECKSUM_UNNECESSARY;
			break;
		}

		/* fall through */
	case CHECKSUM_NONE:
		NAPI_GRO_CB(skb)->flush = 1;
		return NULL;
	}

	return tcp_gro_receive(head, skb);
}

H
Herbert Xu 已提交
764
static int tcp6_gro_complete(struct sk_buff *skb)
H
Herbert Xu 已提交
765
{
766
	const struct ipv6hdr *iph = ipv6_hdr(skb);
H
Herbert Xu 已提交
767 768 769 770 771 772 773 774 775
	struct tcphdr *th = tcp_hdr(skb);

	th->check = ~tcp_v6_check(skb->len - skb_transport_offset(skb),
				  &iph->saddr, &iph->daddr, 0);
	skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;

	return tcp_gro_complete(skb);
}

776
static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win,
777
				 u32 ts, struct tcp_md5sig_key *key, int rst, u8 tclass)
L
Linus Torvalds 已提交
778
{
779 780
	const struct tcphdr *th = tcp_hdr(skb);
	struct tcphdr *t1;
L
Linus Torvalds 已提交
781
	struct sk_buff *buff;
782
	struct flowi6 fl6;
E
Eric Dumazet 已提交
783
	struct net *net = dev_net(skb_dst(skb)->dev);
784
	struct sock *ctl_sk = net->ipv6.tcp_sk;
785
	unsigned int tot_len = sizeof(struct tcphdr);
E
Eric Dumazet 已提交
786
	struct dst_entry *dst;
787
	__be32 *topt;
L
Linus Torvalds 已提交
788

789 790
	if (ts)
		tot_len += TCPOLEN_TSTAMP_ALIGNED;
791 792 793 794 795 796
#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
797
			 GFP_ATOMIC);
798 799
	if (buff == NULL)
		return;
L
Linus Torvalds 已提交
800

801
	skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
L
Linus Torvalds 已提交
802

803
	t1 = (struct tcphdr *) skb_push(buff, tot_len);
804
	skb_reset_transport_header(buff);
L
Linus Torvalds 已提交
805 806 807 808 809

	/* Swap the send and the receive. */
	memset(t1, 0, sizeof(*t1));
	t1->dest = th->source;
	t1->source = th->dest;
810
	t1->doff = tot_len / 4;
811 812 813 814 815
	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
816

817 818
	topt = (__be32 *)(t1 + 1);

819 820 821 822 823 824 825
	if (ts) {
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
		*topt++ = htonl(tcp_time_stamp);
		*topt++ = htonl(ts);
	}

826 827
#ifdef CONFIG_TCP_MD5SIG
	if (key) {
828 829 830
		*topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
				(TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
		tcp_v6_md5_hash_hdr((__u8 *)topt, key,
831 832
				    &ipv6_hdr(skb)->saddr,
				    &ipv6_hdr(skb)->daddr, t1);
833 834 835
	}
#endif

836
	memset(&fl6, 0, sizeof(fl6));
A
Alexey Dobriyan 已提交
837 838
	fl6.daddr = ipv6_hdr(skb)->saddr;
	fl6.saddr = ipv6_hdr(skb)->daddr;
L
Linus Torvalds 已提交
839

840 841 842
	buff->ip_summed = CHECKSUM_PARTIAL;
	buff->csum = 0;

843
	__tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
L
Linus Torvalds 已提交
844

845 846
	fl6.flowi6_proto = IPPROTO_TCP;
	fl6.flowi6_oif = inet6_iif(skb);
847 848
	fl6.fl6_dport = t1->dest;
	fl6.fl6_sport = t1->source;
849
	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
L
Linus Torvalds 已提交
850

851 852 853 854
	/* Pass a socket to ip6_dst_lookup either it is for RST
	 * Underlying function will use this to retrieve the network
	 * namespace
	 */
855
	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL, false);
856 857
	if (!IS_ERR(dst)) {
		skb_dst_set(buff, dst);
858
		ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass);
859 860 861 862
		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
863 864 865 866 867
	}

	kfree_skb(buff);
}

868
static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
L
Linus Torvalds 已提交
869
{
870
	const struct tcphdr *th = tcp_hdr(skb);
871
	u32 seq = 0, ack_seq = 0;
872
	struct tcp_md5sig_key *key = NULL;
873 874 875 876 877 878 879
#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
880

881
	if (th->rst)
L
Linus Torvalds 已提交
882 883
		return;

884 885
	if (!ipv6_unicast_destination(skb))
		return;
L
Linus Torvalds 已提交
886

887
#ifdef CONFIG_TCP_MD5SIG
888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913
	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),
					   &tcp_hashinfo, &ipv6h->daddr,
					   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;
	}
914 915
#endif

916 917 918 919 920
	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
921

922
	tcp_v6_send_response(skb, seq, ack_seq, 0, 0, key, 1, 0);
923 924 925 926 927 928 929 930

#ifdef CONFIG_TCP_MD5SIG
release_sk1:
	if (sk1) {
		rcu_read_unlock();
		sock_put(sk1);
	}
#endif
931
}
L
Linus Torvalds 已提交
932

933
static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts,
934
			    struct tcp_md5sig_key *key, u8 tclass)
935
{
936
	tcp_v6_send_response(skb, seq, ack, win, ts, key, 0, tclass);
L
Linus Torvalds 已提交
937 938 939 940
}

static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
{
941
	struct inet_timewait_sock *tw = inet_twsk(sk);
942
	struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
L
Linus Torvalds 已提交
943

944
	tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
945
			tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
946 947
			tcptw->tw_ts_recent, tcp_twsk_md5_key(tcptw),
			tw->tw_tclass);
L
Linus Torvalds 已提交
948

949
	inet_twsk_put(tw);
L
Linus Torvalds 已提交
950 951
}

952 953
static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
				  struct request_sock *req)
L
Linus Torvalds 已提交
954
{
955
	tcp_v6_send_ack(skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, req->ts_recent,
956
			tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr), 0);
L
Linus Torvalds 已提交
957 958 959 960 961
}


static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
{
962
	struct request_sock *req, **prev;
963
	const struct tcphdr *th = tcp_hdr(skb);
L
Linus Torvalds 已提交
964 965 966
	struct sock *nsk;

	/* Find possible connection requests. */
967
	req = inet6_csk_search_req(sk, &prev, th->source,
968 969
				   &ipv6_hdr(skb)->saddr,
				   &ipv6_hdr(skb)->daddr, inet6_iif(skb));
L
Linus Torvalds 已提交
970 971 972
	if (req)
		return tcp_check_req(sk, skb, req, prev);

973
	nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
974 975
			&ipv6_hdr(skb)->saddr, th->source,
			&ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb));
L
Linus Torvalds 已提交
976 977 978 979 980 981

	if (nsk) {
		if (nsk->sk_state != TCP_TIME_WAIT) {
			bh_lock_sock(nsk);
			return nsk;
		}
982
		inet_twsk_put(inet_twsk(nsk));
L
Linus Torvalds 已提交
983 984 985
		return NULL;
	}

986
#ifdef CONFIG_SYN_COOKIES
987
	if (!th->syn)
988
		sk = cookie_v6_check(sk, skb);
L
Linus Torvalds 已提交
989 990 991 992 993 994 995 996 997
#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)
{
998
	struct tcp_extend_values tmp_ext;
999
	struct tcp_options_received tmp_opt;
1000
	const u8 *hash_location;
1001
	struct request_sock *req;
1002
	struct inet6_request_sock *treq;
L
Linus Torvalds 已提交
1003 1004
	struct ipv6_pinfo *np = inet6_sk(sk);
	struct tcp_sock *tp = tcp_sk(sk);
1005
	__u32 isn = TCP_SKB_CB(skb)->when;
1006
	struct dst_entry *dst = NULL;
1007
	struct flowi6 fl6;
E
Eric Dumazet 已提交
1008
	bool want_cookie = false;
L
Linus Torvalds 已提交
1009 1010 1011 1012 1013

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

	if (!ipv6_unicast_destination(skb))
1014
		goto drop;
L
Linus Torvalds 已提交
1015

1016
	if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
1017 1018 1019
		want_cookie = tcp_syn_flood_action(sk, skb, "TCPv6");
		if (!want_cookie)
			goto drop;
L
Linus Torvalds 已提交
1020 1021
	}

1022
	if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
L
Linus Torvalds 已提交
1023 1024
		goto drop;

1025
	req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
L
Linus Torvalds 已提交
1026 1027 1028
	if (req == NULL)
		goto drop;

1029 1030 1031 1032
#ifdef CONFIG_TCP_MD5SIG
	tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
#endif

L
Linus Torvalds 已提交
1033 1034 1035
	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
1036
	tcp_parse_options(skb, &tmp_opt, &hash_location, 0, NULL);
1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052

	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 */
1053
		d = (__force u32 *)&ipv6_hdr(skb)->daddr.s6_addr32[0];
1054 1055 1056 1057
		*mess++ ^= *d++;
		*mess++ ^= *d++;
		*mess++ ^= *d++;
		*mess++ ^= *d++;
1058
		d = (__force u32 *)&ipv6_hdr(skb)->saddr.s6_addr32[0];
1059 1060 1061 1062 1063 1064 1065 1066 1067
		*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
1068

E
Eric Dumazet 已提交
1069
		want_cookie = false;	/* not our kind of cookie */
1070 1071 1072 1073 1074 1075 1076 1077 1078 1079
		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
1080

1081
	if (want_cookie && !tmp_opt.saw_tstamp)
1082 1083
		tcp_clear_options(&tmp_opt);

L
Linus Torvalds 已提交
1084 1085 1086
	tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
	tcp_openreq_init(req, &tmp_opt, skb);

1087
	treq = inet6_rsk(req);
A
Alexey Dobriyan 已提交
1088 1089
	treq->rmt_addr = ipv6_hdr(skb)->saddr;
	treq->loc_addr = ipv6_hdr(skb)->daddr;
1090
	if (!want_cookie || tmp_opt.tstamp_ok)
1091
		TCP_ECN_create_request(req, skb);
1092

1093 1094 1095 1096 1097 1098 1099
	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);

1100
	if (!isn) {
1101 1102 1103 1104 1105 1106
		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;
		}
1107 1108

		if (want_cookie) {
1109 1110
			isn = cookie_v6_init_sequence(sk, skb, &req->mss);
			req->cookie_ts = tmp_opt.tstamp_ok;
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124
			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 &&
1125 1126
		    (dst = inet6_csk_route_req(sk, &fl6, req)) != NULL) {
			if (!tcp_peer_is_proven(req, dst, true)) {
1127 1128 1129 1130 1131 1132 1133 1134
				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)) &&
1135
			 !tcp_peer_is_proven(req, dst, false)) {
1136 1137 1138 1139 1140 1141 1142 1143 1144 1145
			/* 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;
1146
		}
1147 1148

		isn = tcp_v6_init_sequence(skb);
1149
	}
1150
have_isn:
1151
	tcp_rsk(req)->snt_isn = isn;
1152
	tcp_rsk(req)->snt_synack = tcp_time_stamp;
L
Linus Torvalds 已提交
1153

1154 1155
	if (security_inet_conn_request(sk, skb, req))
		goto drop_and_release;
1156

1157
	if (tcp_v6_send_synack(sk, dst, &fl6, req,
1158 1159
			       (struct request_values *)&tmp_ext,
			       skb_get_queue_mapping(skb)) ||
1160
	    want_cookie)
1161
		goto drop_and_free;
L
Linus Torvalds 已提交
1162

1163 1164
	inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
	return 0;
L
Linus Torvalds 已提交
1165

1166 1167
drop_and_release:
	dst_release(dst);
1168 1169
drop_and_free:
	reqsk_free(req);
L
Linus Torvalds 已提交
1170 1171 1172 1173 1174
drop:
	return 0; /* don't send reset */
}

static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
1175
					  struct request_sock *req,
L
Linus Torvalds 已提交
1176 1177
					  struct dst_entry *dst)
{
1178
	struct inet6_request_sock *treq;
L
Linus Torvalds 已提交
1179 1180 1181 1182 1183
	struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
	struct tcp6_sock *newtcp6sk;
	struct inet_sock *newinet;
	struct tcp_sock *newtp;
	struct sock *newsk;
1184 1185 1186
#ifdef CONFIG_TCP_MD5SIG
	struct tcp_md5sig_key *key;
#endif
1187
	struct flowi6 fl6;
L
Linus Torvalds 已提交
1188 1189 1190 1191 1192 1193 1194 1195

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

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

1196
		if (newsk == NULL)
L
Linus Torvalds 已提交
1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207
			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
1208
		ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr);
L
Linus Torvalds 已提交
1209

E
Eric Dumazet 已提交
1210
		ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr);
L
Linus Torvalds 已提交
1211

A
Alexey Dobriyan 已提交
1212
		newnp->rcv_saddr = newnp->saddr;
L
Linus Torvalds 已提交
1213

1214
		inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
L
Linus Torvalds 已提交
1215
		newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1216 1217 1218 1219
#ifdef CONFIG_TCP_MD5SIG
		newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
#endif

1220 1221
		newnp->ipv6_ac_list = NULL;
		newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1222 1223
		newnp->pktoptions  = NULL;
		newnp->opt	   = NULL;
1224
		newnp->mcast_oif   = inet6_iif(skb);
1225
		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
1226
		newnp->rcv_tclass  = ipv6_tclass(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1227

1228 1229 1230 1231
		/*
		 * 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
1232 1233 1234
		 */

		/* It is tricky place. Until this moment IPv4 tcp
1235
		   worked with IPv6 icsk.icsk_af_ops.
L
Linus Torvalds 已提交
1236 1237
		   Sync it now.
		 */
1238
		tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
L
Linus Torvalds 已提交
1239 1240 1241 1242

		return newsk;
	}

1243
	treq = inet6_rsk(req);
L
Linus Torvalds 已提交
1244 1245 1246 1247

	if (sk_acceptq_is_full(sk))
		goto out_overflow;

1248
	if (!dst) {
1249
		dst = inet6_csk_route_req(sk, &fl6, req);
1250
		if (!dst)
L
Linus Torvalds 已提交
1251
			goto out;
1252
	}
L
Linus Torvalds 已提交
1253 1254 1255

	newsk = tcp_create_openreq_child(sk, req, skb);
	if (newsk == NULL)
1256
		goto out_nonewsk;
L
Linus Torvalds 已提交
1257

1258 1259 1260 1261 1262
	/*
	 * 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
1263

1264
	newsk->sk_gso_type = SKB_GSO_TCPV6;
1265
	__ip6_dst_store(newsk, dst, NULL, NULL);
L
Linus Torvalds 已提交
1266 1267 1268 1269 1270 1271 1272 1273 1274 1275

	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
1276 1277 1278
	newnp->daddr = treq->rmt_addr;
	newnp->saddr = treq->loc_addr;
	newnp->rcv_saddr = treq->loc_addr;
1279
	newsk->sk_bound_dev_if = treq->iif;
L
Linus Torvalds 已提交
1280

1281
	/* Now IPv6 options...
L
Linus Torvalds 已提交
1282 1283 1284

	   First: no IPv4 options.
	 */
1285
	newinet->inet_opt = NULL;
1286
	newnp->ipv6_ac_list = NULL;
1287
	newnp->ipv6_fl_list = NULL;
L
Linus Torvalds 已提交
1288 1289 1290 1291 1292 1293

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

	/* Clone pktoptions received with SYN */
	newnp->pktoptions = NULL;
1294 1295
	if (treq->pktopts != NULL) {
		newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC);
1296
		consume_skb(treq->pktopts);
1297
		treq->pktopts = NULL;
L
Linus Torvalds 已提交
1298 1299 1300 1301
		if (newnp->pktoptions)
			skb_set_owner_r(newnp->pktoptions, newsk);
	}
	newnp->opt	  = NULL;
1302
	newnp->mcast_oif  = inet6_iif(skb);
1303
	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1304
	newnp->rcv_tclass = ipv6_tclass(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1305 1306 1307 1308 1309 1310 1311

	/* 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.
	 */
1312 1313
	if (np->opt)
		newnp->opt = ipv6_dup_options(newsk, np->opt);
L
Linus Torvalds 已提交
1314

1315
	inet_csk(newsk)->icsk_ext_hdr_len = 0;
L
Linus Torvalds 已提交
1316
	if (newnp->opt)
1317 1318
		inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
						     newnp->opt->opt_flen);
L
Linus Torvalds 已提交
1319

J
John Heffner 已提交
1320
	tcp_mtup_init(newsk);
L
Linus Torvalds 已提交
1321
	tcp_sync_mss(newsk, dst_mtu(dst));
1322
	newtp->advmss = dst_metric_advmss(dst);
1323 1324 1325 1326
	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
1327
	tcp_initialize_rcv_mss(newsk);
1328 1329 1330 1331
	if (tcp_rsk(req)->snt_synack)
		tcp_valid_rtt_meas(newsk,
		    tcp_time_stamp - tcp_rsk(req)->snt_synack);
	newtp->total_retrans = req->retrans;
L
Linus Torvalds 已提交
1332

E
Eric Dumazet 已提交
1333 1334
	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
L
Linus Torvalds 已提交
1335

1336 1337 1338 1339 1340 1341 1342 1343
#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
1344 1345
		tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newnp->daddr,
			       AF_INET6, key->key, key->keylen, GFP_ATOMIC);
1346 1347 1348
	}
#endif

1349 1350 1351 1352
	if (__inet_inherit_port(sk, newsk) < 0) {
		sock_put(newsk);
		goto out;
	}
1353
	__inet6_hash(newsk, NULL);
L
Linus Torvalds 已提交
1354 1355 1356 1357

	return newsk;

out_overflow:
1358
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1359
out_nonewsk:
L
Linus Torvalds 已提交
1360
	dst_release(dst);
1361 1362
out:
	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
L
Linus Torvalds 已提交
1363 1364 1365
	return NULL;
}

1366
static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
L
Linus Torvalds 已提交
1367
{
1368
	if (skb->ip_summed == CHECKSUM_COMPLETE) {
H
Herbert Xu 已提交
1369
		if (!tcp_v6_check(skb->len, &ipv6_hdr(skb)->saddr,
1370
				  &ipv6_hdr(skb)->daddr, skb->csum)) {
1371
			skb->ip_summed = CHECKSUM_UNNECESSARY;
L
Linus Torvalds 已提交
1372
			return 0;
1373
		}
L
Linus Torvalds 已提交
1374
	}
1375

H
Herbert Xu 已提交
1376
	skb->csum = ~csum_unfold(tcp_v6_check(skb->len,
1377 1378
					      &ipv6_hdr(skb)->saddr,
					      &ipv6_hdr(skb)->daddr, 0));
1379

L
Linus Torvalds 已提交
1380
	if (skb->len <= 76) {
1381
		return __skb_checksum_complete(skb);
L
Linus Torvalds 已提交
1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410
	}
	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);

1411 1412 1413 1414 1415
#ifdef CONFIG_TCP_MD5SIG
	if (tcp_v6_inbound_md5_hash (sk, skb))
		goto discard;
#endif

1416
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434
		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.
1435
					       --ANK (980728)
L
Linus Torvalds 已提交
1436 1437 1438 1439 1440
	 */
	if (np->rxopt.all)
		opt_skb = skb_clone(skb, GFP_ATOMIC);

	if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1441
		sock_rps_save_rxhash(sk, skb);
1442
		if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len))
L
Linus Torvalds 已提交
1443 1444 1445 1446 1447 1448
			goto reset;
		if (opt_skb)
			goto ipv6_pktoptions;
		return 0;
	}

1449
	if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
L
Linus Torvalds 已提交
1450 1451
		goto csum_err;

1452
	if (sk->sk_state == TCP_LISTEN) {
L
Linus Torvalds 已提交
1453 1454 1455 1456 1457 1458 1459 1460 1461
		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..
		 */
1462
		if(nsk != sk) {
1463
			sock_rps_save_rxhash(nsk, skb);
L
Linus Torvalds 已提交
1464 1465 1466 1467 1468 1469
			if (tcp_child_process(sk, nsk, skb))
				goto reset;
			if (opt_skb)
				__kfree_skb(opt_skb);
			return 0;
		}
1470
	} else
1471
		sock_rps_save_rxhash(sk, skb);
L
Linus Torvalds 已提交
1472

1473
	if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
L
Linus Torvalds 已提交
1474 1475 1476 1477 1478 1479
		goto reset;
	if (opt_skb)
		goto ipv6_pktoptions;
	return 0;

reset:
1480
	tcp_v6_send_reset(sk, skb);
L
Linus Torvalds 已提交
1481 1482 1483 1484 1485 1486
discard:
	if (opt_skb)
		__kfree_skb(opt_skb);
	kfree_skb(skb);
	return 0;
csum_err:
1487
	TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501
	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))) {
1502
		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
1503
			np->mcast_oif = inet6_iif(opt_skb);
1504
		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1505
			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1506 1507
		if (np->rxopt.bits.rxtclass)
			np->rcv_tclass = ipv6_tclass(ipv6_hdr(skb));
L
Linus Torvalds 已提交
1508 1509 1510 1511 1512 1513 1514 1515 1516
		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);
		}
	}

1517
	kfree_skb(opt_skb);
L
Linus Torvalds 已提交
1518 1519 1520
	return 0;
}

1521
static int tcp_v6_rcv(struct sk_buff *skb)
L
Linus Torvalds 已提交
1522
{
1523
	const struct tcphdr *th;
1524
	const struct ipv6hdr *hdr;
L
Linus Torvalds 已提交
1525 1526
	struct sock *sk;
	int ret;
1527
	struct net *net = dev_net(skb->dev);
L
Linus Torvalds 已提交
1528 1529 1530 1531 1532 1533 1534

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

	/*
	 *	Count it even if it's bad.
	 */
1535
	TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
L
Linus Torvalds 已提交
1536 1537 1538 1539

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

1540
	th = tcp_hdr(skb);
L
Linus Torvalds 已提交
1541 1542 1543 1544 1545 1546

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

1547
	if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
L
Linus Torvalds 已提交
1548 1549
		goto bad_packet;

1550
	th = tcp_hdr(skb);
1551
	hdr = ipv6_hdr(skb);
L
Linus Torvalds 已提交
1552 1553 1554 1555 1556
	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;
1557
	TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
L
Linus Torvalds 已提交
1558 1559
	TCP_SKB_CB(skb)->sacked = 0;

1560
	sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest);
L
Linus Torvalds 已提交
1561 1562 1563 1564 1565 1566 1567
	if (!sk)
		goto no_tcp_socket;

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

1568 1569 1570 1571 1572
	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
1573 1574 1575
	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
		goto discard_and_relse;

1576
	if (sk_filter(sk, skb))
L
Linus Torvalds 已提交
1577 1578 1579 1580
		goto discard_and_relse;

	skb->dev = NULL;

1581
	bh_lock_sock_nested(sk);
L
Linus Torvalds 已提交
1582 1583
	ret = 0;
	if (!sock_owned_by_user(sk)) {
C
Chris Leech 已提交
1584
#ifdef CONFIG_NET_DMA
1585
		struct tcp_sock *tp = tcp_sk(sk);
1586
		if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
1587
			tp->ucopy.dma_chan = net_dma_find_channel();
1588 1589 1590
		if (tp->ucopy.dma_chan)
			ret = tcp_v6_do_rcv(sk, skb);
		else
C
Chris Leech 已提交
1591 1592 1593 1594 1595
#endif
		{
			if (!tcp_prequeue(sk, skb))
				ret = tcp_v6_do_rcv(sk, skb);
		}
1596 1597
	} else if (unlikely(sk_add_backlog(sk, skb,
					   sk->sk_rcvbuf + sk->sk_sndbuf))) {
Z
Zhu Yi 已提交
1598
		bh_unlock_sock(sk);
1599
		NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
Z
Zhu Yi 已提交
1600 1601
		goto discard_and_relse;
	}
L
Linus Torvalds 已提交
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612
	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:
1613
		TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
L
Linus Torvalds 已提交
1614
	} else {
1615
		tcp_v6_send_reset(NULL, skb);
L
Linus Torvalds 已提交
1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632
	}

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)) {
1633
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1634 1635 1636 1637
		goto discard_it;
	}

	if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1638
		TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1639
		inet_twsk_put(inet_twsk(sk));
L
Linus Torvalds 已提交
1640 1641 1642
		goto discard_it;
	}

1643
	switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
L
Linus Torvalds 已提交
1644 1645 1646 1647
	case TCP_TW_SYN:
	{
		struct sock *sk2;

1648
		sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1649
					    &ipv6_hdr(skb)->daddr,
1650
					    ntohs(th->dest), inet6_iif(skb));
L
Linus Torvalds 已提交
1651
		if (sk2 != NULL) {
1652 1653 1654
			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
1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669
			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;
}

1670 1671 1672 1673 1674 1675
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,
};

1676
static const struct inet_connection_sock_af_ops ipv6_specific = {
1677 1678 1679 1680 1681 1682
	.queue_xmit	   = inet6_csk_xmit,
	.send_check	   = tcp_v6_send_check,
	.rebuild_header	   = inet6_sk_rebuild_header,
	.conn_request	   = tcp_v6_conn_request,
	.syn_recv_sock	   = tcp_v6_syn_recv_sock,
	.net_header_len	   = sizeof(struct ipv6hdr),
1683
	.net_frag_header_len = sizeof(struct frag_hdr),
1684 1685 1686 1687
	.setsockopt	   = ipv6_setsockopt,
	.getsockopt	   = ipv6_getsockopt,
	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
	.sockaddr_len	   = sizeof(struct sockaddr_in6),
1688
	.bind_conflict	   = inet6_csk_bind_conflict,
1689
#ifdef CONFIG_COMPAT
1690 1691
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1692
#endif
L
Linus Torvalds 已提交
1693 1694
};

1695
#ifdef CONFIG_TCP_MD5SIG
1696
static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1697
	.md5_lookup	=	tcp_v6_md5_lookup,
1698
	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
1699 1700
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1701
#endif
1702

L
Linus Torvalds 已提交
1703 1704 1705 1706
/*
 *	TCP over IPv4 via INET6 API
 */

1707
static const struct inet_connection_sock_af_ops ipv6_mapped = {
1708 1709 1710 1711 1712 1713 1714 1715 1716 1717
	.queue_xmit	   = ip_queue_xmit,
	.send_check	   = tcp_v4_send_check,
	.rebuild_header	   = inet_sk_rebuild_header,
	.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),
1718
	.bind_conflict	   = inet6_csk_bind_conflict,
1719
#ifdef CONFIG_COMPAT
1720 1721
	.compat_setsockopt = compat_ipv6_setsockopt,
	.compat_getsockopt = compat_ipv6_getsockopt,
1722
#endif
L
Linus Torvalds 已提交
1723 1724
};

1725
#ifdef CONFIG_TCP_MD5SIG
1726
static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1727
	.md5_lookup	=	tcp_v4_md5_lookup,
1728
	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
1729 1730
	.md5_parse	=	tcp_v6_parse_md5_keys,
};
1731
#endif
1732

L
Linus Torvalds 已提交
1733 1734 1735 1736 1737
/* 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)
{
1738
	struct inet_connection_sock *icsk = inet_csk(sk);
L
Linus Torvalds 已提交
1739

1740
	tcp_init_sock(sk);
L
Linus Torvalds 已提交
1741

1742
	icsk->icsk_af_ops = &ipv6_specific;
L
Linus Torvalds 已提交
1743

1744
#ifdef CONFIG_TCP_MD5SIG
1745
	tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
1746 1747
#endif

L
Linus Torvalds 已提交
1748 1749 1750
	return 0;
}

1751
static void tcp_v6_destroy_sock(struct sock *sk)
L
Linus Torvalds 已提交
1752 1753
{
	tcp_v4_destroy_sock(sk);
1754
	inet6_destroy_sock(sk);
L
Linus Torvalds 已提交
1755 1756
}

1757
#ifdef CONFIG_PROC_FS
L
Linus Torvalds 已提交
1758
/* Proc filesystem TCPv6 sock list dumping. */
1759
static void get_openreq6(struct seq_file *seq,
1760
			 const struct sock *sk, struct request_sock *req, int i, int uid)
L
Linus Torvalds 已提交
1761 1762
{
	int ttd = req->expires - jiffies;
1763 1764
	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
1765 1766 1767 1768 1769 1770

	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
1771
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1772 1773 1774
		   i,
		   src->s6_addr32[0], src->s6_addr32[1],
		   src->s6_addr32[2], src->s6_addr32[3],
1775
		   ntohs(inet_rsk(req)->loc_port),
L
Linus Torvalds 已提交
1776 1777
		   dest->s6_addr32[0], dest->s6_addr32[1],
		   dest->s6_addr32[2], dest->s6_addr32[3],
1778
		   ntohs(inet_rsk(req)->rmt_port),
L
Linus Torvalds 已提交
1779 1780
		   TCP_SYN_RECV,
		   0,0, /* could print option size, but that is af dependent. */
1781 1782
		   1,   /* timers active (only the expire timer) */
		   jiffies_to_clock_t(ttd),
L
Linus Torvalds 已提交
1783 1784
		   req->retrans,
		   uid,
1785
		   0,  /* non standard timer */
L
Linus Torvalds 已提交
1786 1787 1788 1789 1790 1791
		   0, /* open_requests have no inode */
		   0, req);
}

static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
{
1792
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1793 1794 1795
	__u16 destp, srcp;
	int timer_active;
	unsigned long timer_expires;
1796 1797
	const struct inet_sock *inet = inet_sk(sp);
	const struct tcp_sock *tp = tcp_sk(sp);
1798
	const struct inet_connection_sock *icsk = inet_csk(sp);
1799
	const struct ipv6_pinfo *np = inet6_sk(sp);
L
Linus Torvalds 已提交
1800 1801 1802

	dest  = &np->daddr;
	src   = &np->rcv_saddr;
E
Eric Dumazet 已提交
1803 1804
	destp = ntohs(inet->inet_dport);
	srcp  = ntohs(inet->inet_sport);
1805 1806

	if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
L
Linus Torvalds 已提交
1807
		timer_active	= 1;
1808 1809
		timer_expires	= icsk->icsk_timeout;
	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
L
Linus Torvalds 已提交
1810
		timer_active	= 4;
1811
		timer_expires	= icsk->icsk_timeout;
L
Linus Torvalds 已提交
1812 1813 1814 1815 1816 1817 1818 1819 1820 1821
	} 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
1822
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %lu %lu %u %u %d\n",
L
Linus Torvalds 已提交
1823 1824 1825 1826 1827
		   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,
1828
		   sp->sk_state,
1829 1830
		   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
1831 1832
		   timer_active,
		   jiffies_to_clock_t(timer_expires - jiffies),
1833
		   icsk->icsk_retransmits,
L
Linus Torvalds 已提交
1834
		   sock_i_uid(sp),
1835
		   icsk->icsk_probes_out,
L
Linus Torvalds 已提交
1836 1837
		   sock_i_ino(sp),
		   atomic_read(&sp->sk_refcnt), sp,
1838 1839
		   jiffies_to_clock_t(icsk->icsk_rto),
		   jiffies_to_clock_t(icsk->icsk_ack.ato),
1840
		   (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
I
Ilpo Järvinen 已提交
1841 1842
		   tp->snd_cwnd,
		   tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh
L
Linus Torvalds 已提交
1843 1844 1845
		   );
}

1846
static void get_timewait6_sock(struct seq_file *seq,
1847
			       struct inet_timewait_sock *tw, int i)
L
Linus Torvalds 已提交
1848
{
1849
	const struct in6_addr *dest, *src;
L
Linus Torvalds 已提交
1850
	__u16 destp, srcp;
1851
	const struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
L
Linus Torvalds 已提交
1852 1853 1854 1855 1856
	int ttd = tw->tw_ttd - jiffies;

	if (ttd < 0)
		ttd = 0;

1857 1858
	dest = &tw6->tw_v6_daddr;
	src  = &tw6->tw_v6_rcv_saddr;
L
Linus Torvalds 已提交
1859 1860 1861 1862 1863
	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
1864
		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
L
Linus Torvalds 已提交
1865 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 1900 1901 1902 1903 1904 1905
		   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,
		   3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
		   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;
}

1906 1907 1908 1909 1910 1911 1912 1913
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
1914 1915 1916
static struct tcp_seq_afinfo tcp6_seq_afinfo = {
	.name		= "tcp6",
	.family		= AF_INET6,
1917
	.seq_fops	= &tcp6_afinfo_seq_fops,
1918 1919 1920
	.seq_ops	= {
		.show		= tcp6_seq_show,
	},
L
Linus Torvalds 已提交
1921 1922
};

1923
int __net_init tcp6_proc_init(struct net *net)
L
Linus Torvalds 已提交
1924
{
1925
	return tcp_proc_register(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1926 1927
}

1928
void tcp6_proc_exit(struct net *net)
L
Linus Torvalds 已提交
1929
{
1930
	tcp_proc_unregister(net, &tcp6_seq_afinfo);
L
Linus Torvalds 已提交
1931 1932 1933 1934 1935 1936 1937 1938 1939
}
#endif

struct proto tcpv6_prot = {
	.name			= "TCPv6",
	.owner			= THIS_MODULE,
	.close			= tcp_close,
	.connect		= tcp_v6_connect,
	.disconnect		= tcp_disconnect,
1940
	.accept			= inet_csk_accept,
L
Linus Torvalds 已提交
1941 1942 1943 1944 1945 1946 1947
	.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,
1948 1949
	.sendmsg		= tcp_sendmsg,
	.sendpage		= tcp_sendpage,
L
Linus Torvalds 已提交
1950
	.backlog_rcv		= tcp_v6_do_rcv,
E
Eric Dumazet 已提交
1951
	.release_cb		= tcp_release_cb,
L
Linus Torvalds 已提交
1952
	.hash			= tcp_v6_hash,
1953 1954
	.unhash			= inet_unhash,
	.get_port		= inet_csk_get_port,
L
Linus Torvalds 已提交
1955 1956 1957 1958
	.enter_memory_pressure	= tcp_enter_memory_pressure,
	.sockets_allocated	= &tcp_sockets_allocated,
	.memory_allocated	= &tcp_memory_allocated,
	.memory_pressure	= &tcp_memory_pressure,
1959
	.orphan_count		= &tcp_orphan_count,
L
Linus Torvalds 已提交
1960 1961 1962 1963
	.sysctl_wmem		= sysctl_tcp_wmem,
	.sysctl_rmem		= sysctl_tcp_rmem,
	.max_header		= MAX_TCP_HEADER,
	.obj_size		= sizeof(struct tcp6_sock),
1964
	.slab_flags		= SLAB_DESTROY_BY_RCU,
1965
	.twsk_prot		= &tcp6_timewait_sock_ops,
1966
	.rsk_prot		= &tcp6_request_sock_ops,
1967
	.h.hashinfo		= &tcp_hashinfo,
1968
	.no_autobind		= true,
1969 1970 1971 1972
#ifdef CONFIG_COMPAT
	.compat_setsockopt	= compat_tcp_setsockopt,
	.compat_getsockopt	= compat_tcp_getsockopt,
#endif
G
Glauber Costa 已提交
1973 1974 1975
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
	.proto_cgroup		= tcp_proto_cgroup,
#endif
L
Linus Torvalds 已提交
1976 1977
};

1978
static const struct inet6_protocol tcpv6_protocol = {
L
Linus Torvalds 已提交
1979 1980
	.handler	=	tcp_v6_rcv,
	.err_handler	=	tcp_v6_err,
1981
	.gso_send_check	=	tcp_v6_gso_send_check,
H
Herbert Xu 已提交
1982
	.gso_segment	=	tcp_tso_segment,
H
Herbert Xu 已提交
1983 1984
	.gro_receive	=	tcp6_gro_receive,
	.gro_complete	=	tcp6_gro_complete,
L
Linus Torvalds 已提交
1985 1986 1987 1988 1989 1990 1991 1992 1993
	.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,
1994 1995
	.flags		=	INET_PROTOSW_PERMANENT |
				INET_PROTOSW_ICSK,
L
Linus Torvalds 已提交
1996 1997
};

1998
static int __net_init tcpv6_net_init(struct net *net)
1999
{
2000 2001
	return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
				    SOCK_RAW, IPPROTO_TCP, net);
2002 2003
}

2004
static void __net_exit tcpv6_net_exit(struct net *net)
2005
{
2006
	inet_ctl_sock_destroy(net->ipv6.tcp_sk);
E
Eric W. Biederman 已提交
2007 2008
}

2009
static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
E
Eric W. Biederman 已提交
2010 2011
{
	inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6);
2012 2013 2014
}

static struct pernet_operations tcpv6_net_ops = {
E
Eric W. Biederman 已提交
2015 2016 2017
	.init	    = tcpv6_net_init,
	.exit	    = tcpv6_net_exit,
	.exit_batch = tcpv6_net_exit_batch,
2018 2019
};

2020
int __init tcpv6_init(void)
L
Linus Torvalds 已提交
2021
{
2022 2023 2024 2025 2026 2027
	int ret;

	ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
	if (ret)
		goto out;

L
Linus Torvalds 已提交
2028
	/* register inet6 protocol */
2029 2030 2031 2032
	ret = inet6_register_protosw(&tcpv6_protosw);
	if (ret)
		goto out_tcpv6_protocol;

2033
	ret = register_pernet_subsys(&tcpv6_net_ops);
2034 2035 2036 2037
	if (ret)
		goto out_tcpv6_protosw;
out:
	return ret;
2038

2039 2040 2041 2042 2043 2044 2045
out_tcpv6_protocol:
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
out_tcpv6_protosw:
	inet6_unregister_protosw(&tcpv6_protosw);
	goto out;
}

2046
void tcpv6_exit(void)
2047
{
2048
	unregister_pernet_subsys(&tcpv6_net_ops);
2049 2050
	inet6_unregister_protosw(&tcpv6_protosw);
	inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
L
Linus Torvalds 已提交
2051
}