Merge remote-tracking branch 'asoc/topic/core' into asoc-next
[linux-2.6-block.git] / net / ipv6 / tcp_ipv6.c
CommitLineData
1da177e4
LT
1/*
2 * TCP over IPv6
1ab1457c 3 * Linux INET6 implementation
1da177e4
LT
4 *
5 * Authors:
1ab1457c 6 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4 7 *
1ab1457c 8 * Based on:
1da177e4
LT
9 * linux/net/ipv4/tcp.c
10 * linux/net/ipv4/tcp_input.c
11 * linux/net/ipv4/tcp_output.c
12 *
13 * Fixes:
14 * Hideaki YOSHIFUJI : sin6_scope_id support
15 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
16 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
17 * a single port at the same time.
18 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/tcp6 to seq_file.
19 *
20 * This program is free software; you can redistribute it and/or
21 * modify it under the terms of the GNU General Public License
22 * as published by the Free Software Foundation; either version
23 * 2 of the License, or (at your option) any later version.
24 */
25
eb4dea58 26#include <linux/bottom_half.h>
1da177e4 27#include <linux/module.h>
1da177e4
LT
28#include <linux/errno.h>
29#include <linux/types.h>
30#include <linux/socket.h>
31#include <linux/sockios.h>
32#include <linux/net.h>
33#include <linux/jiffies.h>
34#include <linux/in.h>
35#include <linux/in6.h>
36#include <linux/netdevice.h>
37#include <linux/init.h>
38#include <linux/jhash.h>
39#include <linux/ipsec.h>
40#include <linux/times.h>
5a0e3ad6 41#include <linux/slab.h>
4aa956d8 42#include <linux/uaccess.h>
1da177e4
LT
43#include <linux/ipv6.h>
44#include <linux/icmpv6.h>
45#include <linux/random.h>
46
47#include <net/tcp.h>
48#include <net/ndisc.h>
5324a040 49#include <net/inet6_hashtables.h>
8129765a 50#include <net/inet6_connection_sock.h>
1da177e4
LT
51#include <net/ipv6.h>
52#include <net/transp_v6.h>
53#include <net/addrconf.h>
54#include <net/ip6_route.h>
55#include <net/ip6_checksum.h>
56#include <net/inet_ecn.h>
57#include <net/protocol.h>
58#include <net/xfrm.h>
1da177e4
LT
59#include <net/snmp.h>
60#include <net/dsfield.h>
6d6ee43e 61#include <net/timewait_sock.h>
3d58b5fa 62#include <net/inet_common.h>
6e5714ea 63#include <net/secure_seq.h>
d1a4c0b3 64#include <net/tcp_memcontrol.h>
076bb0c8 65#include <net/busy_poll.h>
1da177e4 66
1da177e4
LT
67#include <linux/proc_fs.h>
68#include <linux/seq_file.h>
69
cfb6eeb4
YH
70#include <linux/crypto.h>
71#include <linux/scatterlist.h>
72
cfb6eeb4 73static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb);
6edafaaf
GJ
74static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
75 struct request_sock *req);
1da177e4
LT
76
77static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
1da177e4 78
3b401a81
SH
79static const struct inet_connection_sock_af_ops ipv6_mapped;
80static const struct inet_connection_sock_af_ops ipv6_specific;
a928630a 81#ifdef CONFIG_TCP_MD5SIG
b2e4b3de
SH
82static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
83static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
9501f972
YH
84#else
85static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
b71d1d42 86 const struct in6_addr *addr)
9501f972
YH
87{
88 return NULL;
89}
a928630a 90#endif
1da177e4 91
fae6ef87
NC
92static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
93{
94 struct dst_entry *dst = skb_dst(skb);
fae6ef87 95
ca777eff
ED
96 if (dst) {
97 const struct rt6_info *rt = (const struct rt6_info *)dst;
98
99 dst_hold(dst);
100 sk->sk_rx_dst = dst;
101 inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
b197df4f 102 inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt);
ca777eff 103 }
fae6ef87
NC
104}
105
cf533ea5 106static __u32 tcp_v6_init_sequence(const struct sk_buff *skb)
1da177e4 107{
0660e03f
ACM
108 return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
109 ipv6_hdr(skb)->saddr.s6_addr32,
aa8223c7
ACM
110 tcp_hdr(skb)->dest,
111 tcp_hdr(skb)->source);
1da177e4
LT
112}
113
1ab1457c 114static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
1da177e4
LT
115 int addr_len)
116{
117 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
1ab1457c 118 struct inet_sock *inet = inet_sk(sk);
d83d8461 119 struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4
LT
120 struct ipv6_pinfo *np = inet6_sk(sk);
121 struct tcp_sock *tp = tcp_sk(sk);
20c59de2 122 struct in6_addr *saddr = NULL, *final_p, final;
4c9483b2 123 struct flowi6 fl6;
1da177e4
LT
124 struct dst_entry *dst;
125 int addr_type;
126 int err;
127
1ab1457c 128 if (addr_len < SIN6_LEN_RFC2133)
1da177e4
LT
129 return -EINVAL;
130
1ab1457c 131 if (usin->sin6_family != AF_INET6)
a02cec21 132 return -EAFNOSUPPORT;
1da177e4 133
4c9483b2 134 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
135
136 if (np->sndflow) {
4c9483b2
DM
137 fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
138 IP6_ECN_flow_init(fl6.flowlabel);
139 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1da177e4 140 struct ip6_flowlabel *flowlabel;
4c9483b2 141 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
63159f29 142 if (!flowlabel)
1da177e4 143 return -EINVAL;
1da177e4
LT
144 fl6_sock_release(flowlabel);
145 }
146 }
147
148 /*
1ab1457c
YH
149 * connect() to INADDR_ANY means loopback (BSD'ism).
150 */
151
4c99aa40 152 if (ipv6_addr_any(&usin->sin6_addr))
1ab1457c 153 usin->sin6_addr.s6_addr[15] = 0x1;
1da177e4
LT
154
155 addr_type = ipv6_addr_type(&usin->sin6_addr);
156
4c99aa40 157 if (addr_type & IPV6_ADDR_MULTICAST)
1da177e4
LT
158 return -ENETUNREACH;
159
160 if (addr_type&IPV6_ADDR_LINKLOCAL) {
161 if (addr_len >= sizeof(struct sockaddr_in6) &&
162 usin->sin6_scope_id) {
163 /* If interface is set while binding, indices
164 * must coincide.
165 */
166 if (sk->sk_bound_dev_if &&
167 sk->sk_bound_dev_if != usin->sin6_scope_id)
168 return -EINVAL;
169
170 sk->sk_bound_dev_if = usin->sin6_scope_id;
171 }
172
173 /* Connect to link-local address requires an interface */
174 if (!sk->sk_bound_dev_if)
175 return -EINVAL;
176 }
177
178 if (tp->rx_opt.ts_recent_stamp &&
efe4208f 179 !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) {
1da177e4
LT
180 tp->rx_opt.ts_recent = 0;
181 tp->rx_opt.ts_recent_stamp = 0;
182 tp->write_seq = 0;
183 }
184
efe4208f 185 sk->sk_v6_daddr = usin->sin6_addr;
4c9483b2 186 np->flow_label = fl6.flowlabel;
1da177e4
LT
187
188 /*
189 * TCP over IPv4
190 */
191
192 if (addr_type == IPV6_ADDR_MAPPED) {
d83d8461 193 u32 exthdrlen = icsk->icsk_ext_hdr_len;
1da177e4
LT
194 struct sockaddr_in sin;
195
196 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
197
198 if (__ipv6_only_sock(sk))
199 return -ENETUNREACH;
200
201 sin.sin_family = AF_INET;
202 sin.sin_port = usin->sin6_port;
203 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
204
d83d8461 205 icsk->icsk_af_ops = &ipv6_mapped;
1da177e4 206 sk->sk_backlog_rcv = tcp_v4_do_rcv;
cfb6eeb4
YH
207#ifdef CONFIG_TCP_MD5SIG
208 tp->af_specific = &tcp_sock_ipv6_mapped_specific;
209#endif
1da177e4
LT
210
211 err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
212
213 if (err) {
d83d8461
ACM
214 icsk->icsk_ext_hdr_len = exthdrlen;
215 icsk->icsk_af_ops = &ipv6_specific;
1da177e4 216 sk->sk_backlog_rcv = tcp_v6_do_rcv;
cfb6eeb4
YH
217#ifdef CONFIG_TCP_MD5SIG
218 tp->af_specific = &tcp_sock_ipv6_specific;
219#endif
1da177e4 220 goto failure;
1da177e4 221 }
d1e559d0 222 np->saddr = sk->sk_v6_rcv_saddr;
1da177e4
LT
223
224 return err;
225 }
226
efe4208f
ED
227 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
228 saddr = &sk->sk_v6_rcv_saddr;
1da177e4 229
4c9483b2 230 fl6.flowi6_proto = IPPROTO_TCP;
efe4208f 231 fl6.daddr = sk->sk_v6_daddr;
4e3fd7a0 232 fl6.saddr = saddr ? *saddr : np->saddr;
4c9483b2
DM
233 fl6.flowi6_oif = sk->sk_bound_dev_if;
234 fl6.flowi6_mark = sk->sk_mark;
1958b856
DM
235 fl6.fl6_dport = usin->sin6_port;
236 fl6.fl6_sport = inet->inet_sport;
1da177e4 237
4c9483b2 238 final_p = fl6_update_dst(&fl6, np->opt, &final);
1da177e4 239
4c9483b2 240 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 241
0e0d44ab 242 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
68d0c6d3
DM
243 if (IS_ERR(dst)) {
244 err = PTR_ERR(dst);
1da177e4 245 goto failure;
14e50e57 246 }
1da177e4 247
63159f29 248 if (!saddr) {
4c9483b2 249 saddr = &fl6.saddr;
efe4208f 250 sk->sk_v6_rcv_saddr = *saddr;
1da177e4
LT
251 }
252
253 /* set the source address */
4e3fd7a0 254 np->saddr = *saddr;
c720c7e8 255 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
1da177e4 256
f83ef8c0 257 sk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 258 __ip6_dst_store(sk, dst, NULL, NULL);
1da177e4 259
493f377d
DM
260 if (tcp_death_row.sysctl_tw_recycle &&
261 !tp->rx_opt.ts_recent_stamp &&
fd0273d7 262 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr))
81166dd6 263 tcp_fetch_timewait_stamp(sk, dst);
493f377d 264
d83d8461 265 icsk->icsk_ext_hdr_len = 0;
1da177e4 266 if (np->opt)
d83d8461
ACM
267 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
268 np->opt->opt_nflen);
1da177e4
LT
269
270 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
271
c720c7e8 272 inet->inet_dport = usin->sin6_port;
1da177e4
LT
273
274 tcp_set_state(sk, TCP_SYN_SENT);
d8313f5c 275 err = inet6_hash_connect(&tcp_death_row, sk);
1da177e4
LT
276 if (err)
277 goto late_failure;
278
9e7ceb06
SP
279 ip6_set_txhash(sk);
280
2b916477 281 if (!tp->write_seq && likely(!tp->repair))
1da177e4 282 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
efe4208f 283 sk->sk_v6_daddr.s6_addr32,
c720c7e8
ED
284 inet->inet_sport,
285 inet->inet_dport);
1da177e4
LT
286
287 err = tcp_connect(sk);
288 if (err)
289 goto late_failure;
290
291 return 0;
292
293late_failure:
294 tcp_set_state(sk, TCP_CLOSE);
295 __sk_dst_reset(sk);
296failure:
c720c7e8 297 inet->inet_dport = 0;
1da177e4
LT
298 sk->sk_route_caps = 0;
299 return err;
300}
301
563d34d0
ED
302static void tcp_v6_mtu_reduced(struct sock *sk)
303{
304 struct dst_entry *dst;
305
306 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
307 return;
308
309 dst = inet6_csk_update_pmtu(sk, tcp_sk(sk)->mtu_info);
310 if (!dst)
311 return;
312
313 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
314 tcp_sync_mss(sk, dst_mtu(dst));
315 tcp_simple_retransmit(sk);
316 }
317}
318
1da177e4 319static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 320 u8 type, u8 code, int offset, __be32 info)
1da177e4 321{
4c99aa40 322 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
505cbfc5 323 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
2215089b
ED
324 struct net *net = dev_net(skb->dev);
325 struct request_sock *fastopen;
1da177e4 326 struct ipv6_pinfo *np;
1ab1457c 327 struct tcp_sock *tp;
0a672f74 328 __u32 seq, snd_una;
2215089b
ED
329 struct sock *sk;
330 int err;
1da177e4 331
2215089b
ED
332 sk = __inet6_lookup_established(net, &tcp_hashinfo,
333 &hdr->daddr, th->dest,
334 &hdr->saddr, ntohs(th->source),
335 skb->dev->ifindex);
1da177e4 336
2215089b 337 if (!sk) {
e41b5368
DL
338 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
339 ICMP6_MIB_INERRORS);
1da177e4
LT
340 return;
341 }
342
343 if (sk->sk_state == TCP_TIME_WAIT) {
9469c7b4 344 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
345 return;
346 }
2215089b
ED
347 seq = ntohl(th->seq);
348 if (sk->sk_state == TCP_NEW_SYN_RECV)
349 return tcp_req_err(sk, seq);
1da177e4
LT
350
351 bh_lock_sock(sk);
563d34d0 352 if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG)
de0744af 353 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
1da177e4
LT
354
355 if (sk->sk_state == TCP_CLOSE)
356 goto out;
357
e802af9c
SH
358 if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) {
359 NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
360 goto out;
361 }
362
1da177e4 363 tp = tcp_sk(sk);
0a672f74
YC
364 /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
365 fastopen = tp->fastopen_rsk;
366 snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
1da177e4 367 if (sk->sk_state != TCP_LISTEN &&
0a672f74 368 !between(seq, snd_una, tp->snd_nxt)) {
de0744af 369 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
1da177e4
LT
370 goto out;
371 }
372
373 np = inet6_sk(sk);
374
ec18d9a2
DM
375 if (type == NDISC_REDIRECT) {
376 struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
377
1ed5c48f 378 if (dst)
6700c270 379 dst->ops->redirect(dst, sk, skb);
50a75a89 380 goto out;
ec18d9a2
DM
381 }
382
1da177e4 383 if (type == ICMPV6_PKT_TOOBIG) {
0d4f0608
ED
384 /* We are not interested in TCP_LISTEN and open_requests
385 * (SYN-ACKs send out by Linux are always <576bytes so
386 * they should go through unfragmented).
387 */
388 if (sk->sk_state == TCP_LISTEN)
389 goto out;
390
93b36cf3
HFS
391 if (!ip6_sk_accept_pmtu(sk))
392 goto out;
393
563d34d0
ED
394 tp->mtu_info = ntohl(info);
395 if (!sock_owned_by_user(sk))
396 tcp_v6_mtu_reduced(sk);
d013ef2a
JA
397 else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED,
398 &tp->tsq_flags))
399 sock_hold(sk);
1da177e4
LT
400 goto out;
401 }
402
403 icmpv6_err_convert(type, code, &err);
404
60236fdd 405 /* Might be for an request_sock */
1da177e4 406 switch (sk->sk_state) {
1da177e4 407 case TCP_SYN_SENT:
0a672f74
YC
408 case TCP_SYN_RECV:
409 /* Only in fast or simultaneous open. If a fast open socket is
410 * is already accepted it is treated as a connected one below.
411 */
63159f29 412 if (fastopen && !fastopen->sk)
0a672f74
YC
413 break;
414
1da177e4 415 if (!sock_owned_by_user(sk)) {
1da177e4
LT
416 sk->sk_err = err;
417 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
418
419 tcp_done(sk);
420 } else
421 sk->sk_err_soft = err;
422 goto out;
423 }
424
425 if (!sock_owned_by_user(sk) && np->recverr) {
426 sk->sk_err = err;
427 sk->sk_error_report(sk);
428 } else
429 sk->sk_err_soft = err;
430
431out:
432 bh_unlock_sock(sk);
433 sock_put(sk);
434}
435
436
9f10d3f6 437static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst,
d6274bd8 438 struct flowi *fl,
3840a06e 439 struct request_sock *req,
3a19ce0e
DL
440 u16 queue_mapping,
441 struct tcp_fastopen_cookie *foc)
1da177e4 442{
634fb979 443 struct inet_request_sock *ireq = inet_rsk(req);
1da177e4 444 struct ipv6_pinfo *np = inet6_sk(sk);
d6274bd8 445 struct flowi6 *fl6 = &fl->u.ip6;
4c99aa40 446 struct sk_buff *skb;
9494218f 447 int err = -ENOMEM;
1da177e4 448
9f10d3f6
NC
449 /* First, grab a route. */
450 if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL)
fd80eb94 451 goto done;
9494218f 452
3a19ce0e 453 skb = tcp_make_synack(sk, dst, req, foc);
9494218f 454
1da177e4 455 if (skb) {
634fb979
ED
456 __tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr,
457 &ireq->ir_v6_rmt_addr);
1da177e4 458
634fb979 459 fl6->daddr = ireq->ir_v6_rmt_addr;
53b24b8f 460 if (np->repflow && ireq->pktopts)
df3687ff
FF
461 fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts));
462
fff32699 463 skb_set_queue_mapping(skb, queue_mapping);
43264e0b 464 err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass);
b9df3cb8 465 err = net_xmit_eval(err);
1da177e4
LT
466 }
467
468done:
1da177e4
LT
469 return err;
470}
471
72659ecc 472
60236fdd 473static void tcp_v6_reqsk_destructor(struct request_sock *req)
1da177e4 474{
634fb979 475 kfree_skb(inet_rsk(req)->pktopts);
1da177e4
LT
476}
477
cfb6eeb4
YH
478#ifdef CONFIG_TCP_MD5SIG
479static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
b71d1d42 480 const struct in6_addr *addr)
cfb6eeb4 481{
a915da9b 482 return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
cfb6eeb4
YH
483}
484
485static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
fd3a154a 486 const struct sock *addr_sk)
cfb6eeb4 487{
efe4208f 488 return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr);
cfb6eeb4
YH
489}
490
4aa956d8
WY
491static int tcp_v6_parse_md5_keys(struct sock *sk, char __user *optval,
492 int optlen)
cfb6eeb4
YH
493{
494 struct tcp_md5sig cmd;
495 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
cfb6eeb4
YH
496
497 if (optlen < sizeof(cmd))
498 return -EINVAL;
499
500 if (copy_from_user(&cmd, optval, sizeof(cmd)))
501 return -EFAULT;
502
503 if (sin6->sin6_family != AF_INET6)
504 return -EINVAL;
505
506 if (!cmd.tcpm_keylen) {
e773e4fa 507 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
a915da9b
ED
508 return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
509 AF_INET);
510 return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
511 AF_INET6);
cfb6eeb4
YH
512 }
513
514 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
515 return -EINVAL;
516
a915da9b
ED
517 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
518 return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
519 AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
cfb6eeb4 520
a915da9b
ED
521 return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
522 AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
cfb6eeb4
YH
523}
524
49a72dfb 525static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
b71d1d42
ED
526 const struct in6_addr *daddr,
527 const struct in6_addr *saddr, int nbytes)
cfb6eeb4 528{
cfb6eeb4 529 struct tcp6_pseudohdr *bp;
49a72dfb 530 struct scatterlist sg;
8d26d76d 531
cfb6eeb4 532 bp = &hp->md5_blk.ip6;
cfb6eeb4 533 /* 1. TCP pseudo-header (RFC2460) */
4e3fd7a0
AD
534 bp->saddr = *saddr;
535 bp->daddr = *daddr;
49a72dfb 536 bp->protocol = cpu_to_be32(IPPROTO_TCP);
00b1304c 537 bp->len = cpu_to_be32(nbytes);
cfb6eeb4 538
49a72dfb
AL
539 sg_init_one(&sg, bp, sizeof(*bp));
540 return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
541}
c7da57a1 542
49a72dfb 543static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
b71d1d42 544 const struct in6_addr *daddr, struct in6_addr *saddr,
318cf7aa 545 const struct tcphdr *th)
49a72dfb
AL
546{
547 struct tcp_md5sig_pool *hp;
548 struct hash_desc *desc;
549
550 hp = tcp_get_md5sig_pool();
551 if (!hp)
552 goto clear_hash_noput;
553 desc = &hp->md5_desc;
554
555 if (crypto_hash_init(desc))
556 goto clear_hash;
557 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2))
558 goto clear_hash;
559 if (tcp_md5_hash_header(hp, th))
560 goto clear_hash;
561 if (tcp_md5_hash_key(hp, key))
562 goto clear_hash;
563 if (crypto_hash_final(desc, md5_hash))
cfb6eeb4 564 goto clear_hash;
cfb6eeb4 565
cfb6eeb4 566 tcp_put_md5sig_pool();
cfb6eeb4 567 return 0;
49a72dfb 568
cfb6eeb4
YH
569clear_hash:
570 tcp_put_md5sig_pool();
571clear_hash_noput:
572 memset(md5_hash, 0, 16);
49a72dfb 573 return 1;
cfb6eeb4
YH
574}
575
39f8e58e
ED
576static int tcp_v6_md5_hash_skb(char *md5_hash,
577 const struct tcp_md5sig_key *key,
318cf7aa 578 const struct sock *sk,
318cf7aa 579 const struct sk_buff *skb)
cfb6eeb4 580{
b71d1d42 581 const struct in6_addr *saddr, *daddr;
49a72dfb
AL
582 struct tcp_md5sig_pool *hp;
583 struct hash_desc *desc;
318cf7aa 584 const struct tcphdr *th = tcp_hdr(skb);
cfb6eeb4 585
39f8e58e
ED
586 if (sk) { /* valid for establish/request sockets */
587 saddr = &sk->sk_v6_rcv_saddr;
efe4208f 588 daddr = &sk->sk_v6_daddr;
49a72dfb 589 } else {
b71d1d42 590 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
49a72dfb
AL
591 saddr = &ip6h->saddr;
592 daddr = &ip6h->daddr;
cfb6eeb4 593 }
49a72dfb
AL
594
595 hp = tcp_get_md5sig_pool();
596 if (!hp)
597 goto clear_hash_noput;
598 desc = &hp->md5_desc;
599
600 if (crypto_hash_init(desc))
601 goto clear_hash;
602
603 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len))
604 goto clear_hash;
605 if (tcp_md5_hash_header(hp, th))
606 goto clear_hash;
607 if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
608 goto clear_hash;
609 if (tcp_md5_hash_key(hp, key))
610 goto clear_hash;
611 if (crypto_hash_final(desc, md5_hash))
612 goto clear_hash;
613
614 tcp_put_md5sig_pool();
615 return 0;
616
617clear_hash:
618 tcp_put_md5sig_pool();
619clear_hash_noput:
620 memset(md5_hash, 0, 16);
621 return 1;
cfb6eeb4
YH
622}
623
ff74e23f 624static bool tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
cfb6eeb4 625{
cf533ea5 626 const __u8 *hash_location = NULL;
cfb6eeb4 627 struct tcp_md5sig_key *hash_expected;
b71d1d42 628 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
318cf7aa 629 const struct tcphdr *th = tcp_hdr(skb);
cfb6eeb4 630 int genhash;
cfb6eeb4
YH
631 u8 newhash[16];
632
633 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
7d5d5525 634 hash_location = tcp_parse_md5sig_option(th);
cfb6eeb4 635
785957d3
DM
636 /* We've parsed the options - do we have a hash? */
637 if (!hash_expected && !hash_location)
ff74e23f 638 return false;
785957d3
DM
639
640 if (hash_expected && !hash_location) {
641 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
ff74e23f 642 return true;
cfb6eeb4
YH
643 }
644
785957d3
DM
645 if (!hash_expected && hash_location) {
646 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
ff74e23f 647 return true;
cfb6eeb4
YH
648 }
649
650 /* check the signature */
49a72dfb
AL
651 genhash = tcp_v6_md5_hash_skb(newhash,
652 hash_expected,
39f8e58e 653 NULL, skb);
49a72dfb 654
cfb6eeb4 655 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
e87cc472
JP
656 net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n",
657 genhash ? "failed" : "mismatch",
658 &ip6h->saddr, ntohs(th->source),
659 &ip6h->daddr, ntohs(th->dest));
ff74e23f 660 return true;
cfb6eeb4 661 }
ff74e23f 662 return false;
cfb6eeb4
YH
663}
664#endif
665
16bea70a
OP
666static void tcp_v6_init_req(struct request_sock *req, struct sock *sk,
667 struct sk_buff *skb)
668{
669 struct inet_request_sock *ireq = inet_rsk(req);
670 struct ipv6_pinfo *np = inet6_sk(sk);
671
672 ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
673 ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
674
16bea70a
OP
675 /* So that link locals have meaning */
676 if (!sk->sk_bound_dev_if &&
677 ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
870c3151 678 ireq->ir_iif = tcp_v6_iif(skb);
16bea70a 679
04317daf 680 if (!TCP_SKB_CB(skb)->tcp_tw_isn &&
a224772d
ED
681 (ipv6_opt_accepted(sk, skb, &TCP_SKB_CB(skb)->header.h6) ||
682 np->rxopt.bits.rxinfo ||
16bea70a
OP
683 np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim ||
684 np->rxopt.bits.rxohlim || np->repflow)) {
685 atomic_inc(&skb->users);
686 ireq->pktopts = skb;
687 }
688}
689
d94e0417
OP
690static struct dst_entry *tcp_v6_route_req(struct sock *sk, struct flowi *fl,
691 const struct request_sock *req,
692 bool *strict)
693{
694 if (strict)
695 *strict = true;
696 return inet6_csk_route_req(sk, &fl->u.ip6, req);
697}
698
c6aefafb 699struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
1da177e4 700 .family = AF_INET6,
2e6599cb 701 .obj_size = sizeof(struct tcp6_request_sock),
5db92c99 702 .rtx_syn_ack = tcp_rtx_synack,
60236fdd
ACM
703 .send_ack = tcp_v6_reqsk_send_ack,
704 .destructor = tcp_v6_reqsk_destructor,
72659ecc 705 .send_reset = tcp_v6_send_reset,
4aa956d8 706 .syn_ack_timeout = tcp_syn_ack_timeout,
1da177e4
LT
707};
708
b2e4b3de 709static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
2aec4a29
OP
710 .mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) -
711 sizeof(struct ipv6hdr),
16bea70a 712#ifdef CONFIG_TCP_MD5SIG
fd3a154a 713 .req_md5_lookup = tcp_v6_md5_lookup,
e3afe7b7 714 .calc_md5_hash = tcp_v6_md5_hash_skb,
b6332e6c 715#endif
16bea70a 716 .init_req = tcp_v6_init_req,
fb7b37a7
OP
717#ifdef CONFIG_SYN_COOKIES
718 .cookie_init_seq = cookie_v6_init_sequence,
719#endif
d94e0417 720 .route_req = tcp_v6_route_req,
936b8bdb 721 .init_seq = tcp_v6_init_sequence,
d6274bd8 722 .send_synack = tcp_v6_send_synack,
695da14e 723 .queue_hash_add = inet6_csk_reqsk_queue_hash_add,
16bea70a 724};
cfb6eeb4 725
0f85feae
ED
726static void tcp_v6_send_response(struct sock *sk, struct sk_buff *skb, u32 seq,
727 u32 ack, u32 win, u32 tsval, u32 tsecr,
728 int oif, struct tcp_md5sig_key *key, int rst,
729 u8 tclass, u32 label)
1da177e4 730{
cf533ea5
ED
731 const struct tcphdr *th = tcp_hdr(skb);
732 struct tcphdr *t1;
1da177e4 733 struct sk_buff *buff;
4c9483b2 734 struct flowi6 fl6;
0f85feae 735 struct net *net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
e5047992 736 struct sock *ctl_sk = net->ipv6.tcp_sk;
77c676da 737 unsigned int tot_len = sizeof(struct tcphdr);
adf30907 738 struct dst_entry *dst;
81ada62d 739 __be32 *topt;
1da177e4 740
ee684b6f 741 if (tsecr)
626e264d 742 tot_len += TCPOLEN_TSTAMP_ALIGNED;
cfb6eeb4 743#ifdef CONFIG_TCP_MD5SIG
cfb6eeb4
YH
744 if (key)
745 tot_len += TCPOLEN_MD5SIG_ALIGNED;
746#endif
747
cfb6eeb4 748 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1da177e4 749 GFP_ATOMIC);
63159f29 750 if (!buff)
1ab1457c 751 return;
1da177e4 752
cfb6eeb4 753 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1da177e4 754
cfb6eeb4 755 t1 = (struct tcphdr *) skb_push(buff, tot_len);
6651ffc8 756 skb_reset_transport_header(buff);
1da177e4
LT
757
758 /* Swap the send and the receive. */
759 memset(t1, 0, sizeof(*t1));
760 t1->dest = th->source;
761 t1->source = th->dest;
cfb6eeb4 762 t1->doff = tot_len / 4;
626e264d
IJ
763 t1->seq = htonl(seq);
764 t1->ack_seq = htonl(ack);
765 t1->ack = !rst || !th->ack;
766 t1->rst = rst;
767 t1->window = htons(win);
1da177e4 768
81ada62d
IJ
769 topt = (__be32 *)(t1 + 1);
770
ee684b6f 771 if (tsecr) {
626e264d
IJ
772 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
773 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
ee684b6f
AV
774 *topt++ = htonl(tsval);
775 *topt++ = htonl(tsecr);
626e264d
IJ
776 }
777
cfb6eeb4
YH
778#ifdef CONFIG_TCP_MD5SIG
779 if (key) {
81ada62d
IJ
780 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
781 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
782 tcp_v6_md5_hash_hdr((__u8 *)topt, key,
78e645cb
IJ
783 &ipv6_hdr(skb)->saddr,
784 &ipv6_hdr(skb)->daddr, t1);
cfb6eeb4
YH
785 }
786#endif
787
4c9483b2 788 memset(&fl6, 0, sizeof(fl6));
4e3fd7a0
AD
789 fl6.daddr = ipv6_hdr(skb)->saddr;
790 fl6.saddr = ipv6_hdr(skb)->daddr;
1d13a96c 791 fl6.flowlabel = label;
1da177e4 792
e5700aff
DM
793 buff->ip_summed = CHECKSUM_PARTIAL;
794 buff->csum = 0;
795
4c9483b2 796 __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
1da177e4 797
4c9483b2 798 fl6.flowi6_proto = IPPROTO_TCP;
a36dbdb2 799 if (rt6_need_strict(&fl6.daddr) && !oif)
870c3151 800 fl6.flowi6_oif = tcp_v6_iif(skb);
9c76a114
WY
801 else
802 fl6.flowi6_oif = oif;
e110861f 803 fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark);
1958b856
DM
804 fl6.fl6_dport = t1->dest;
805 fl6.fl6_sport = t1->source;
4c9483b2 806 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
1da177e4 807
c20121ae
DL
808 /* Pass a socket to ip6_dst_lookup either it is for RST
809 * Underlying function will use this to retrieve the network
810 * namespace
811 */
0e0d44ab 812 dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
68d0c6d3
DM
813 if (!IS_ERR(dst)) {
814 skb_dst_set(buff, dst);
b903d324 815 ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass);
68d0c6d3
DM
816 TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
817 if (rst)
818 TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS);
819 return;
1da177e4
LT
820 }
821
822 kfree_skb(buff);
823}
824
626e264d 825static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
1da177e4 826{
cf533ea5 827 const struct tcphdr *th = tcp_hdr(skb);
626e264d 828 u32 seq = 0, ack_seq = 0;
fa3e5b4e 829 struct tcp_md5sig_key *key = NULL;
658ddaaf
SL
830#ifdef CONFIG_TCP_MD5SIG
831 const __u8 *hash_location = NULL;
832 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
833 unsigned char newhash[16];
834 int genhash;
835 struct sock *sk1 = NULL;
836#endif
9c76a114 837 int oif;
1da177e4 838
626e264d 839 if (th->rst)
1da177e4
LT
840 return;
841
c3658e8d
ED
842 /* If sk not NULL, it means we did a successful lookup and incoming
843 * route had to be correct. prequeue might have dropped our dst.
844 */
845 if (!sk && !ipv6_unicast_destination(skb))
626e264d 846 return;
1da177e4 847
cfb6eeb4 848#ifdef CONFIG_TCP_MD5SIG
658ddaaf
SL
849 hash_location = tcp_parse_md5sig_option(th);
850 if (!sk && hash_location) {
851 /*
852 * active side is lost. Try to find listening socket through
853 * source port, and then find md5 key through listening socket.
854 * we are not loose security here:
855 * Incoming packet is checked with md5 hash with finding key,
856 * no RST generated if md5 hash doesn't match.
857 */
858 sk1 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev),
5ba24953
TH
859 &tcp_hashinfo, &ipv6h->saddr,
860 th->source, &ipv6h->daddr,
870c3151 861 ntohs(th->source), tcp_v6_iif(skb));
658ddaaf
SL
862 if (!sk1)
863 return;
864
865 rcu_read_lock();
866 key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr);
867 if (!key)
868 goto release_sk1;
869
39f8e58e 870 genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, skb);
658ddaaf
SL
871 if (genhash || memcmp(hash_location, newhash, 16) != 0)
872 goto release_sk1;
873 } else {
874 key = sk ? tcp_v6_md5_do_lookup(sk, &ipv6h->saddr) : NULL;
875 }
cfb6eeb4
YH
876#endif
877
626e264d
IJ
878 if (th->ack)
879 seq = ntohl(th->ack_seq);
880 else
881 ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len -
882 (th->doff << 2);
1da177e4 883
9c76a114 884 oif = sk ? sk->sk_bound_dev_if : 0;
0f85feae 885 tcp_v6_send_response(sk, skb, seq, ack_seq, 0, 0, 0, oif, key, 1, 0, 0);
658ddaaf
SL
886
887#ifdef CONFIG_TCP_MD5SIG
888release_sk1:
889 if (sk1) {
890 rcu_read_unlock();
891 sock_put(sk1);
892 }
893#endif
626e264d 894}
1da177e4 895
0f85feae
ED
896static void tcp_v6_send_ack(struct sock *sk, struct sk_buff *skb, u32 seq,
897 u32 ack, u32 win, u32 tsval, u32 tsecr, int oif,
1d13a96c
FF
898 struct tcp_md5sig_key *key, u8 tclass,
899 u32 label)
626e264d 900{
0f85feae
ED
901 tcp_v6_send_response(sk, skb, seq, ack, win, tsval, tsecr, oif, key, 0,
902 tclass, label);
1da177e4
LT
903}
904
905static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
906{
8feaf0c0 907 struct inet_timewait_sock *tw = inet_twsk(sk);
cfb6eeb4 908 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
1da177e4 909
0f85feae 910 tcp_v6_send_ack(sk, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
8feaf0c0 911 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
ee684b6f 912 tcp_time_stamp + tcptw->tw_ts_offset,
9c76a114 913 tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw),
21858cd0 914 tw->tw_tclass, cpu_to_be32(tw->tw_flowlabel));
1da177e4 915
8feaf0c0 916 inet_twsk_put(tw);
1da177e4
LT
917}
918
6edafaaf
GJ
919static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
920 struct request_sock *req)
1da177e4 921{
3a19ce0e
DL
922 /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
923 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
924 */
0f85feae 925 tcp_v6_send_ack(sk, skb, (sk->sk_state == TCP_LISTEN) ?
3a19ce0e 926 tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt,
0f85feae
ED
927 tcp_rsk(req)->rcv_nxt, req->rcv_wnd,
928 tcp_time_stamp, req->ts_recent, sk->sk_bound_dev_if,
1d13a96c
FF
929 tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr),
930 0, 0);
1da177e4
LT
931}
932
933
4c99aa40 934static struct sock *tcp_v6_hnd_req(struct sock *sk, struct sk_buff *skb)
1da177e4 935{
aa8223c7 936 const struct tcphdr *th = tcp_hdr(skb);
52452c54 937 struct request_sock *req;
1da177e4
LT
938 struct sock *nsk;
939
940 /* Find possible connection requests. */
52452c54 941 req = inet6_csk_search_req(sk, th->source,
0660e03f 942 &ipv6_hdr(skb)->saddr,
870c3151 943 &ipv6_hdr(skb)->daddr, tcp_v6_iif(skb));
fa76ce73
ED
944 if (req) {
945 nsk = tcp_check_req(sk, skb, req, false);
3257d8b1 946 if (!nsk || nsk == sk)
b357a364 947 reqsk_put(req);
fa76ce73
ED
948 return nsk;
949 }
3b1e0a65 950 nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
870c3151
ED
951 &ipv6_hdr(skb)->saddr, th->source,
952 &ipv6_hdr(skb)->daddr, ntohs(th->dest),
953 tcp_v6_iif(skb));
1da177e4
LT
954
955 if (nsk) {
956 if (nsk->sk_state != TCP_TIME_WAIT) {
957 bh_lock_sock(nsk);
958 return nsk;
959 }
9469c7b4 960 inet_twsk_put(inet_twsk(nsk));
1da177e4
LT
961 return NULL;
962 }
963
c6aefafb 964#ifdef CONFIG_SYN_COOKIES
af9b4738 965 if (!th->syn)
c6aefafb 966 sk = cookie_v6_check(sk, skb);
1da177e4
LT
967#endif
968 return sk;
969}
970
1da177e4
LT
971static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
972{
1da177e4
LT
973 if (skb->protocol == htons(ETH_P_IP))
974 return tcp_v4_conn_request(sk, skb);
975
976 if (!ipv6_unicast_destination(skb))
1ab1457c 977 goto drop;
1da177e4 978
1fb6f159
OP
979 return tcp_conn_request(&tcp6_request_sock_ops,
980 &tcp_request_sock_ipv6_ops, sk, skb);
1da177e4
LT
981
982drop:
5f1e942c 983 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
1da177e4
LT
984 return 0; /* don't send reset */
985}
986
4c99aa40
WC
987static struct sock *tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
988 struct request_sock *req,
989 struct dst_entry *dst)
1da177e4 990{
634fb979 991 struct inet_request_sock *ireq;
1da177e4
LT
992 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
993 struct tcp6_sock *newtcp6sk;
994 struct inet_sock *newinet;
995 struct tcp_sock *newtp;
996 struct sock *newsk;
cfb6eeb4
YH
997#ifdef CONFIG_TCP_MD5SIG
998 struct tcp_md5sig_key *key;
999#endif
3840a06e 1000 struct flowi6 fl6;
1da177e4
LT
1001
1002 if (skb->protocol == htons(ETH_P_IP)) {
1003 /*
1004 * v6 mapped
1005 */
1006
1007 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1008
63159f29 1009 if (!newsk)
1da177e4
LT
1010 return NULL;
1011
1012 newtcp6sk = (struct tcp6_sock *)newsk;
1013 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1014
1015 newinet = inet_sk(newsk);
1016 newnp = inet6_sk(newsk);
1017 newtp = tcp_sk(newsk);
1018
1019 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1020
d1e559d0 1021 newnp->saddr = newsk->sk_v6_rcv_saddr;
1da177e4 1022
8292a17a 1023 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1da177e4 1024 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
cfb6eeb4
YH
1025#ifdef CONFIG_TCP_MD5SIG
1026 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1027#endif
1028
676a1184
YZ
1029 newnp->ipv6_ac_list = NULL;
1030 newnp->ipv6_fl_list = NULL;
1da177e4
LT
1031 newnp->pktoptions = NULL;
1032 newnp->opt = NULL;
870c3151 1033 newnp->mcast_oif = tcp_v6_iif(skb);
0660e03f 1034 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1397ed35 1035 newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
df3687ff
FF
1036 if (np->repflow)
1037 newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
1da177e4 1038
e6848976
ACM
1039 /*
1040 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1041 * here, tcp_create_openreq_child now does this for us, see the comment in
1042 * that function for the gory details. -acme
1da177e4 1043 */
1da177e4
LT
1044
1045 /* It is tricky place. Until this moment IPv4 tcp
8292a17a 1046 worked with IPv6 icsk.icsk_af_ops.
1da177e4
LT
1047 Sync it now.
1048 */
d83d8461 1049 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1da177e4
LT
1050
1051 return newsk;
1052 }
1053
634fb979 1054 ireq = inet_rsk(req);
1da177e4
LT
1055
1056 if (sk_acceptq_is_full(sk))
1057 goto out_overflow;
1058
493f377d 1059 if (!dst) {
3840a06e 1060 dst = inet6_csk_route_req(sk, &fl6, req);
493f377d 1061 if (!dst)
1da177e4 1062 goto out;
1ab1457c 1063 }
1da177e4
LT
1064
1065 newsk = tcp_create_openreq_child(sk, req, skb);
63159f29 1066 if (!newsk)
093d2823 1067 goto out_nonewsk;
1da177e4 1068
e6848976
ACM
1069 /*
1070 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1071 * count here, tcp_create_openreq_child now does this for us, see the
1072 * comment in that function for the gory details. -acme
1073 */
1da177e4 1074
59eed279 1075 newsk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 1076 __ip6_dst_store(newsk, dst, NULL, NULL);
fae6ef87 1077 inet6_sk_rx_dst_set(newsk, skb);
1da177e4
LT
1078
1079 newtcp6sk = (struct tcp6_sock *)newsk;
1080 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1081
1082 newtp = tcp_sk(newsk);
1083 newinet = inet_sk(newsk);
1084 newnp = inet6_sk(newsk);
1085
1086 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1087
634fb979
ED
1088 newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr;
1089 newnp->saddr = ireq->ir_v6_loc_addr;
1090 newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr;
1091 newsk->sk_bound_dev_if = ireq->ir_iif;
1da177e4 1092
b73c3d0e
TH
1093 ip6_set_txhash(newsk);
1094
1ab1457c 1095 /* Now IPv6 options...
1da177e4
LT
1096
1097 First: no IPv4 options.
1098 */
f6d8bd05 1099 newinet->inet_opt = NULL;
676a1184 1100 newnp->ipv6_ac_list = NULL;
d35690be 1101 newnp->ipv6_fl_list = NULL;
1da177e4
LT
1102
1103 /* Clone RX bits */
1104 newnp->rxopt.all = np->rxopt.all;
1105
1106 /* Clone pktoptions received with SYN */
1107 newnp->pktoptions = NULL;
53b24b8f 1108 if (ireq->pktopts) {
634fb979 1109 newnp->pktoptions = skb_clone(ireq->pktopts,
99a1dec7 1110 sk_gfp_atomic(sk, GFP_ATOMIC));
634fb979
ED
1111 consume_skb(ireq->pktopts);
1112 ireq->pktopts = NULL;
1da177e4
LT
1113 if (newnp->pktoptions)
1114 skb_set_owner_r(newnp->pktoptions, newsk);
1115 }
1116 newnp->opt = NULL;
870c3151 1117 newnp->mcast_oif = tcp_v6_iif(skb);
0660e03f 1118 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1397ed35 1119 newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
df3687ff
FF
1120 if (np->repflow)
1121 newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
1da177e4
LT
1122
1123 /* Clone native IPv6 options from listening socket (if any)
1124
1125 Yes, keeping reference count would be much more clever,
1126 but we make one more one thing there: reattach optmem
1127 to newsk.
1128 */
43264e0b
RL
1129 if (np->opt)
1130 newnp->opt = ipv6_dup_options(newsk, np->opt);
1da177e4 1131
d83d8461 1132 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1da177e4 1133 if (newnp->opt)
d83d8461
ACM
1134 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1135 newnp->opt->opt_flen);
1da177e4 1136
81164413
DB
1137 tcp_ca_openreq_child(newsk, dst);
1138
1da177e4 1139 tcp_sync_mss(newsk, dst_mtu(dst));
0dbaee3b 1140 newtp->advmss = dst_metric_advmss(dst);
d135c522
NC
1141 if (tcp_sk(sk)->rx_opt.user_mss &&
1142 tcp_sk(sk)->rx_opt.user_mss < newtp->advmss)
1143 newtp->advmss = tcp_sk(sk)->rx_opt.user_mss;
1144
1da177e4
LT
1145 tcp_initialize_rcv_mss(newsk);
1146
c720c7e8
ED
1147 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
1148 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
1da177e4 1149
cfb6eeb4
YH
1150#ifdef CONFIG_TCP_MD5SIG
1151 /* Copy over the MD5 key from the original socket */
4aa956d8 1152 key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr);
53b24b8f 1153 if (key) {
cfb6eeb4
YH
1154 /* We're using one, so create a matching key
1155 * on the newsk structure. If we fail to get
1156 * memory, then we end up not copying the key
1157 * across. Shucks.
1158 */
efe4208f 1159 tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
99a1dec7
MG
1160 AF_INET6, key->key, key->keylen,
1161 sk_gfp_atomic(sk, GFP_ATOMIC));
cfb6eeb4
YH
1162 }
1163#endif
1164
093d2823 1165 if (__inet_inherit_port(sk, newsk) < 0) {
e337e24d
CP
1166 inet_csk_prepare_forced_close(newsk);
1167 tcp_done(newsk);
093d2823
BS
1168 goto out;
1169 }
77a6a471 1170 __inet_hash(newsk, NULL);
1da177e4
LT
1171
1172 return newsk;
1173
1174out_overflow:
de0744af 1175 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
093d2823 1176out_nonewsk:
1da177e4 1177 dst_release(dst);
093d2823
BS
1178out:
1179 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
1da177e4
LT
1180 return NULL;
1181}
1182
1da177e4
LT
1183/* The socket must have it's spinlock held when we get
1184 * here.
1185 *
1186 * We have a potential double-lock case here, so even when
1187 * doing backlog processing we use the BH locking scheme.
1188 * This is because we cannot sleep with the original spinlock
1189 * held.
1190 */
1191static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1192{
1193 struct ipv6_pinfo *np = inet6_sk(sk);
1194 struct tcp_sock *tp;
1195 struct sk_buff *opt_skb = NULL;
1196
1197 /* Imagine: socket is IPv6. IPv4 packet arrives,
1198 goes to IPv4 receive handler and backlogged.
1199 From backlog it always goes here. Kerboom...
1200 Fortunately, tcp_rcv_established and rcv_established
1201 handle them correctly, but it is not case with
1202 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1203 */
1204
1205 if (skb->protocol == htons(ETH_P_IP))
1206 return tcp_v4_do_rcv(sk, skb);
1207
fda9ef5d 1208 if (sk_filter(sk, skb))
1da177e4
LT
1209 goto discard;
1210
1211 /*
1212 * socket locking is here for SMP purposes as backlog rcv
1213 * is currently called with bh processing disabled.
1214 */
1215
1216 /* Do Stevens' IPV6_PKTOPTIONS.
1217
1218 Yes, guys, it is the only place in our code, where we
1219 may make it not affecting IPv4.
1220 The rest of code is protocol independent,
1221 and I do not like idea to uglify IPv4.
1222
1223 Actually, all the idea behind IPV6_PKTOPTIONS
1224 looks not very well thought. For now we latch
1225 options, received in the last packet, enqueued
1226 by tcp. Feel free to propose better solution.
1ab1457c 1227 --ANK (980728)
1da177e4
LT
1228 */
1229 if (np->rxopt.all)
99a1dec7 1230 opt_skb = skb_clone(skb, sk_gfp_atomic(sk, GFP_ATOMIC));
1da177e4
LT
1231
1232 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
5d299f3d
ED
1233 struct dst_entry *dst = sk->sk_rx_dst;
1234
bdeab991 1235 sock_rps_save_rxhash(sk, skb);
3d97379a 1236 sk_mark_napi_id(sk, skb);
5d299f3d
ED
1237 if (dst) {
1238 if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif ||
1239 dst->ops->check(dst, np->rx_dst_cookie) == NULL) {
1240 dst_release(dst);
1241 sk->sk_rx_dst = NULL;
1242 }
1243 }
1244
c995ae22 1245 tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
1da177e4
LT
1246 if (opt_skb)
1247 goto ipv6_pktoptions;
1248 return 0;
1249 }
1250
12e25e10 1251 if (tcp_checksum_complete(skb))
1da177e4
LT
1252 goto csum_err;
1253
1ab1457c 1254 if (sk->sk_state == TCP_LISTEN) {
1da177e4
LT
1255 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1256 if (!nsk)
1257 goto discard;
1258
1259 /*
1260 * Queue it on the new socket if the new socket is active,
1261 * otherwise we just shortcircuit this and continue with
1262 * the new socket..
1263 */
4c99aa40 1264 if (nsk != sk) {
bdeab991 1265 sock_rps_save_rxhash(nsk, skb);
3d97379a 1266 sk_mark_napi_id(sk, skb);
1da177e4
LT
1267 if (tcp_child_process(sk, nsk, skb))
1268 goto reset;
1269 if (opt_skb)
1270 __kfree_skb(opt_skb);
1271 return 0;
1272 }
47482f13 1273 } else
bdeab991 1274 sock_rps_save_rxhash(sk, skb);
1da177e4 1275
aa8223c7 1276 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
1da177e4 1277 goto reset;
1da177e4
LT
1278 if (opt_skb)
1279 goto ipv6_pktoptions;
1280 return 0;
1281
1282reset:
cfb6eeb4 1283 tcp_v6_send_reset(sk, skb);
1da177e4
LT
1284discard:
1285 if (opt_skb)
1286 __kfree_skb(opt_skb);
1287 kfree_skb(skb);
1288 return 0;
1289csum_err:
6a5dc9e5 1290 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS);
63231bdd 1291 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
1da177e4
LT
1292 goto discard;
1293
1294
1295ipv6_pktoptions:
1296 /* Do you ask, what is it?
1297
1298 1. skb was enqueued by tcp.
1299 2. skb is added to tail of read queue, rather than out of order.
1300 3. socket is not in passive state.
1301 4. Finally, it really contains options, which user wants to receive.
1302 */
1303 tp = tcp_sk(sk);
1304 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1305 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
333fad53 1306 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
870c3151 1307 np->mcast_oif = tcp_v6_iif(opt_skb);
333fad53 1308 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
0660e03f 1309 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
82e9f105 1310 if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
1397ed35 1311 np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
df3687ff
FF
1312 if (np->repflow)
1313 np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
a224772d 1314 if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->header.h6)) {
1da177e4
LT
1315 skb_set_owner_r(opt_skb, sk);
1316 opt_skb = xchg(&np->pktoptions, opt_skb);
1317 } else {
1318 __kfree_skb(opt_skb);
1319 opt_skb = xchg(&np->pktoptions, NULL);
1320 }
1321 }
1322
800d55f1 1323 kfree_skb(opt_skb);
1da177e4
LT
1324 return 0;
1325}
1326
2dc49d16
ND
1327static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
1328 const struct tcphdr *th)
1329{
1330 /* This is tricky: we move IP6CB at its correct location into
1331 * TCP_SKB_CB(). It must be done after xfrm6_policy_check(), because
1332 * _decode_session6() uses IP6CB().
1333 * barrier() makes sure compiler won't play aliasing games.
1334 */
1335 memmove(&TCP_SKB_CB(skb)->header.h6, IP6CB(skb),
1336 sizeof(struct inet6_skb_parm));
1337 barrier();
1338
1339 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1340 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1341 skb->len - th->doff*4);
1342 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1343 TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
1344 TCP_SKB_CB(skb)->tcp_tw_isn = 0;
1345 TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
1346 TCP_SKB_CB(skb)->sacked = 0;
1347}
1348
4ad19de8
AK
1349static void tcp_v6_restore_cb(struct sk_buff *skb)
1350{
1351 /* We need to move header back to the beginning if xfrm6_policy_check()
1352 * and tcp_v6_fill_cb() are going to be called again.
1353 */
1354 memmove(IP6CB(skb), &TCP_SKB_CB(skb)->header.h6,
1355 sizeof(struct inet6_skb_parm));
1356}
1357
e5bbef20 1358static int tcp_v6_rcv(struct sk_buff *skb)
1da177e4 1359{
cf533ea5 1360 const struct tcphdr *th;
b71d1d42 1361 const struct ipv6hdr *hdr;
1da177e4
LT
1362 struct sock *sk;
1363 int ret;
a86b1e30 1364 struct net *net = dev_net(skb->dev);
1da177e4
LT
1365
1366 if (skb->pkt_type != PACKET_HOST)
1367 goto discard_it;
1368
1369 /*
1370 * Count it even if it's bad.
1371 */
63231bdd 1372 TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
1da177e4
LT
1373
1374 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1375 goto discard_it;
1376
aa8223c7 1377 th = tcp_hdr(skb);
1da177e4
LT
1378
1379 if (th->doff < sizeof(struct tcphdr)/4)
1380 goto bad_packet;
1381 if (!pskb_may_pull(skb, th->doff*4))
1382 goto discard_it;
1383
e4f45b7f 1384 if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo))
6a5dc9e5 1385 goto csum_error;
1da177e4 1386
aa8223c7 1387 th = tcp_hdr(skb);
e802af9c 1388 hdr = ipv6_hdr(skb);
1da177e4 1389
870c3151 1390 sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest,
2dc49d16 1391 inet6_iif(skb));
1da177e4
LT
1392 if (!sk)
1393 goto no_tcp_socket;
1394
1395process:
1396 if (sk->sk_state == TCP_TIME_WAIT)
1397 goto do_time_wait;
1398
e802af9c
SH
1399 if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) {
1400 NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
1401 goto discard_and_relse;
1402 }
1403
1da177e4
LT
1404 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1405 goto discard_and_relse;
1406
2dc49d16
ND
1407 tcp_v6_fill_cb(skb, hdr, th);
1408
9ea88a15
DP
1409#ifdef CONFIG_TCP_MD5SIG
1410 if (tcp_v6_inbound_md5_hash(sk, skb))
1411 goto discard_and_relse;
1412#endif
1413
fda9ef5d 1414 if (sk_filter(sk, skb))
1da177e4
LT
1415 goto discard_and_relse;
1416
2c8c56e1 1417 sk_incoming_cpu_update(sk);
1da177e4
LT
1418 skb->dev = NULL;
1419
293b9c42 1420 bh_lock_sock_nested(sk);
2efd055c 1421 tcp_sk(sk)->segs_in += max_t(u16, 1, skb_shinfo(skb)->gso_segs);
1da177e4
LT
1422 ret = 0;
1423 if (!sock_owned_by_user(sk)) {
7bced397 1424 if (!tcp_prequeue(sk, skb))
1ab1457c 1425 ret = tcp_v6_do_rcv(sk, skb);
da882c1f
ED
1426 } else if (unlikely(sk_add_backlog(sk, skb,
1427 sk->sk_rcvbuf + sk->sk_sndbuf))) {
6b03a53a 1428 bh_unlock_sock(sk);
6cce09f8 1429 NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
6b03a53a
ZY
1430 goto discard_and_relse;
1431 }
1da177e4
LT
1432 bh_unlock_sock(sk);
1433
1434 sock_put(sk);
1435 return ret ? -1 : 0;
1436
1437no_tcp_socket:
1438 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1439 goto discard_it;
1440
2dc49d16
ND
1441 tcp_v6_fill_cb(skb, hdr, th);
1442
12e25e10 1443 if (tcp_checksum_complete(skb)) {
6a5dc9e5
ED
1444csum_error:
1445 TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS);
1da177e4 1446bad_packet:
63231bdd 1447 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1da177e4 1448 } else {
cfb6eeb4 1449 tcp_v6_send_reset(NULL, skb);
1da177e4
LT
1450 }
1451
1452discard_it:
1da177e4
LT
1453 kfree_skb(skb);
1454 return 0;
1455
1456discard_and_relse:
1457 sock_put(sk);
1458 goto discard_it;
1459
1460do_time_wait:
1461 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
9469c7b4 1462 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1463 goto discard_it;
1464 }
1465
2dc49d16
ND
1466 tcp_v6_fill_cb(skb, hdr, th);
1467
6a5dc9e5
ED
1468 if (tcp_checksum_complete(skb)) {
1469 inet_twsk_put(inet_twsk(sk));
1470 goto csum_error;
1da177e4
LT
1471 }
1472
9469c7b4 1473 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1da177e4
LT
1474 case TCP_TW_SYN:
1475 {
1476 struct sock *sk2;
1477
c346dca1 1478 sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
5ba24953 1479 &ipv6_hdr(skb)->saddr, th->source,
0660e03f 1480 &ipv6_hdr(skb)->daddr,
870c3151 1481 ntohs(th->dest), tcp_v6_iif(skb));
53b24b8f 1482 if (sk2) {
295ff7ed 1483 struct inet_timewait_sock *tw = inet_twsk(sk);
789f558c 1484 inet_twsk_deschedule(tw);
295ff7ed 1485 inet_twsk_put(tw);
1da177e4 1486 sk = sk2;
4ad19de8 1487 tcp_v6_restore_cb(skb);
1da177e4
LT
1488 goto process;
1489 }
1490 /* Fall through to ACK */
1491 }
1492 case TCP_TW_ACK:
1493 tcp_v6_timewait_ack(sk, skb);
1494 break;
1495 case TCP_TW_RST:
4ad19de8 1496 tcp_v6_restore_cb(skb);
1da177e4 1497 goto no_tcp_socket;
4aa956d8
WY
1498 case TCP_TW_SUCCESS:
1499 ;
1da177e4
LT
1500 }
1501 goto discard_it;
1502}
1503
c7109986
ED
1504static void tcp_v6_early_demux(struct sk_buff *skb)
1505{
1506 const struct ipv6hdr *hdr;
1507 const struct tcphdr *th;
1508 struct sock *sk;
1509
1510 if (skb->pkt_type != PACKET_HOST)
1511 return;
1512
1513 if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr)))
1514 return;
1515
1516 hdr = ipv6_hdr(skb);
1517 th = tcp_hdr(skb);
1518
1519 if (th->doff < sizeof(struct tcphdr) / 4)
1520 return;
1521
870c3151 1522 /* Note : We use inet6_iif() here, not tcp_v6_iif() */
c7109986
ED
1523 sk = __inet6_lookup_established(dev_net(skb->dev), &tcp_hashinfo,
1524 &hdr->saddr, th->source,
1525 &hdr->daddr, ntohs(th->dest),
1526 inet6_iif(skb));
1527 if (sk) {
1528 skb->sk = sk;
1529 skb->destructor = sock_edemux;
f7e4eb03 1530 if (sk_fullsock(sk)) {
d0c294c5 1531 struct dst_entry *dst = READ_ONCE(sk->sk_rx_dst);
f3f12135 1532
c7109986 1533 if (dst)
5d299f3d 1534 dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie);
c7109986 1535 if (dst &&
f3f12135 1536 inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
c7109986
ED
1537 skb_dst_set_noref(skb, dst);
1538 }
1539 }
1540}
1541
ccb7c410
DM
1542static struct timewait_sock_ops tcp6_timewait_sock_ops = {
1543 .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
1544 .twsk_unique = tcp_twsk_unique,
4aa956d8 1545 .twsk_destructor = tcp_twsk_destructor,
ccb7c410
DM
1546};
1547
3b401a81 1548static const struct inet_connection_sock_af_ops ipv6_specific = {
543d9cfe
ACM
1549 .queue_xmit = inet6_csk_xmit,
1550 .send_check = tcp_v6_send_check,
1551 .rebuild_header = inet6_sk_rebuild_header,
5d299f3d 1552 .sk_rx_dst_set = inet6_sk_rx_dst_set,
543d9cfe
ACM
1553 .conn_request = tcp_v6_conn_request,
1554 .syn_recv_sock = tcp_v6_syn_recv_sock,
543d9cfe 1555 .net_header_len = sizeof(struct ipv6hdr),
67469601 1556 .net_frag_header_len = sizeof(struct frag_hdr),
543d9cfe
ACM
1557 .setsockopt = ipv6_setsockopt,
1558 .getsockopt = ipv6_getsockopt,
1559 .addr2sockaddr = inet6_csk_addr2sockaddr,
1560 .sockaddr_len = sizeof(struct sockaddr_in6),
ab1e0a13 1561 .bind_conflict = inet6_csk_bind_conflict,
3fdadf7d 1562#ifdef CONFIG_COMPAT
543d9cfe
ACM
1563 .compat_setsockopt = compat_ipv6_setsockopt,
1564 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1565#endif
4fab9071 1566 .mtu_reduced = tcp_v6_mtu_reduced,
1da177e4
LT
1567};
1568
cfb6eeb4 1569#ifdef CONFIG_TCP_MD5SIG
b2e4b3de 1570static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
cfb6eeb4 1571 .md5_lookup = tcp_v6_md5_lookup,
49a72dfb 1572 .calc_md5_hash = tcp_v6_md5_hash_skb,
cfb6eeb4 1573 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1574};
a928630a 1575#endif
cfb6eeb4 1576
1da177e4
LT
1577/*
1578 * TCP over IPv4 via INET6 API
1579 */
3b401a81 1580static const struct inet_connection_sock_af_ops ipv6_mapped = {
543d9cfe
ACM
1581 .queue_xmit = ip_queue_xmit,
1582 .send_check = tcp_v4_send_check,
1583 .rebuild_header = inet_sk_rebuild_header,
63d02d15 1584 .sk_rx_dst_set = inet_sk_rx_dst_set,
543d9cfe
ACM
1585 .conn_request = tcp_v6_conn_request,
1586 .syn_recv_sock = tcp_v6_syn_recv_sock,
543d9cfe
ACM
1587 .net_header_len = sizeof(struct iphdr),
1588 .setsockopt = ipv6_setsockopt,
1589 .getsockopt = ipv6_getsockopt,
1590 .addr2sockaddr = inet6_csk_addr2sockaddr,
1591 .sockaddr_len = sizeof(struct sockaddr_in6),
ab1e0a13 1592 .bind_conflict = inet6_csk_bind_conflict,
3fdadf7d 1593#ifdef CONFIG_COMPAT
543d9cfe
ACM
1594 .compat_setsockopt = compat_ipv6_setsockopt,
1595 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1596#endif
4fab9071 1597 .mtu_reduced = tcp_v4_mtu_reduced,
1da177e4
LT
1598};
1599
cfb6eeb4 1600#ifdef CONFIG_TCP_MD5SIG
b2e4b3de 1601static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
cfb6eeb4 1602 .md5_lookup = tcp_v4_md5_lookup,
49a72dfb 1603 .calc_md5_hash = tcp_v4_md5_hash_skb,
cfb6eeb4 1604 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1605};
a928630a 1606#endif
cfb6eeb4 1607
1da177e4
LT
1608/* NOTE: A lot of things set to zero explicitly by call to
1609 * sk_alloc() so need not be done here.
1610 */
1611static int tcp_v6_init_sock(struct sock *sk)
1612{
6687e988 1613 struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4 1614
900f65d3 1615 tcp_init_sock(sk);
1da177e4 1616
8292a17a 1617 icsk->icsk_af_ops = &ipv6_specific;
1da177e4 1618
cfb6eeb4 1619#ifdef CONFIG_TCP_MD5SIG
ac807fa8 1620 tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
cfb6eeb4
YH
1621#endif
1622
1da177e4
LT
1623 return 0;
1624}
1625
7d06b2e0 1626static void tcp_v6_destroy_sock(struct sock *sk)
1da177e4 1627{
1da177e4 1628 tcp_v4_destroy_sock(sk);
7d06b2e0 1629 inet6_destroy_sock(sk);
1da177e4
LT
1630}
1631
952a10be 1632#ifdef CONFIG_PROC_FS
1da177e4 1633/* Proc filesystem TCPv6 sock list dumping. */
1ab1457c 1634static void get_openreq6(struct seq_file *seq,
d4f06873 1635 struct request_sock *req, int i, kuid_t uid)
1da177e4 1636{
fa76ce73 1637 long ttd = req->rsk_timer.expires - jiffies;
634fb979
ED
1638 const struct in6_addr *src = &inet_rsk(req)->ir_v6_loc_addr;
1639 const struct in6_addr *dest = &inet_rsk(req)->ir_v6_rmt_addr;
1da177e4
LT
1640
1641 if (ttd < 0)
1642 ttd = 0;
1643
1da177e4
LT
1644 seq_printf(seq,
1645 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
d14c5ab6 1646 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
1da177e4
LT
1647 i,
1648 src->s6_addr32[0], src->s6_addr32[1],
1649 src->s6_addr32[2], src->s6_addr32[3],
b44084c2 1650 inet_rsk(req)->ir_num,
1da177e4
LT
1651 dest->s6_addr32[0], dest->s6_addr32[1],
1652 dest->s6_addr32[2], dest->s6_addr32[3],
634fb979 1653 ntohs(inet_rsk(req)->ir_rmt_port),
1da177e4 1654 TCP_SYN_RECV,
4c99aa40 1655 0, 0, /* could print option size, but that is af dependent. */
1ab1457c
YH
1656 1, /* timers active (only the expire timer) */
1657 jiffies_to_clock_t(ttd),
e6c022a4 1658 req->num_timeout,
a7cb5a49 1659 from_kuid_munged(seq_user_ns(seq), uid),
1ab1457c 1660 0, /* non standard timer */
1da177e4
LT
1661 0, /* open_requests have no inode */
1662 0, req);
1663}
1664
1665static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1666{
b71d1d42 1667 const struct in6_addr *dest, *src;
1da177e4
LT
1668 __u16 destp, srcp;
1669 int timer_active;
1670 unsigned long timer_expires;
cf533ea5
ED
1671 const struct inet_sock *inet = inet_sk(sp);
1672 const struct tcp_sock *tp = tcp_sk(sp);
463c84b9 1673 const struct inet_connection_sock *icsk = inet_csk(sp);
0a672f74 1674 struct fastopen_queue *fastopenq = icsk->icsk_accept_queue.fastopenq;
1da177e4 1675
efe4208f
ED
1676 dest = &sp->sk_v6_daddr;
1677 src = &sp->sk_v6_rcv_saddr;
c720c7e8
ED
1678 destp = ntohs(inet->inet_dport);
1679 srcp = ntohs(inet->inet_sport);
463c84b9
ACM
1680
1681 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
1da177e4 1682 timer_active = 1;
463c84b9
ACM
1683 timer_expires = icsk->icsk_timeout;
1684 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
1da177e4 1685 timer_active = 4;
463c84b9 1686 timer_expires = icsk->icsk_timeout;
1da177e4
LT
1687 } else if (timer_pending(&sp->sk_timer)) {
1688 timer_active = 2;
1689 timer_expires = sp->sk_timer.expires;
1690 } else {
1691 timer_active = 0;
1692 timer_expires = jiffies;
1693 }
1694
1695 seq_printf(seq,
1696 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
d14c5ab6 1697 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
1da177e4
LT
1698 i,
1699 src->s6_addr32[0], src->s6_addr32[1],
1700 src->s6_addr32[2], src->s6_addr32[3], srcp,
1701 dest->s6_addr32[0], dest->s6_addr32[1],
1702 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1ab1457c 1703 sp->sk_state,
47da8ee6
SS
1704 tp->write_seq-tp->snd_una,
1705 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
1da177e4 1706 timer_active,
a399a805 1707 jiffies_delta_to_clock_t(timer_expires - jiffies),
463c84b9 1708 icsk->icsk_retransmits,
a7cb5a49 1709 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
6687e988 1710 icsk->icsk_probes_out,
1da177e4
LT
1711 sock_i_ino(sp),
1712 atomic_read(&sp->sk_refcnt), sp,
7be87351
SH
1713 jiffies_to_clock_t(icsk->icsk_rto),
1714 jiffies_to_clock_t(icsk->icsk_ack.ato),
4c99aa40 1715 (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
0b6a05c1 1716 tp->snd_cwnd,
0a672f74
YC
1717 sp->sk_state == TCP_LISTEN ?
1718 (fastopenq ? fastopenq->max_qlen : 0) :
1719 (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)
1da177e4
LT
1720 );
1721}
1722
1ab1457c 1723static void get_timewait6_sock(struct seq_file *seq,
8feaf0c0 1724 struct inet_timewait_sock *tw, int i)
1da177e4 1725{
789f558c 1726 long delta = tw->tw_timer.expires - jiffies;
b71d1d42 1727 const struct in6_addr *dest, *src;
1da177e4 1728 __u16 destp, srcp;
1da177e4 1729
efe4208f
ED
1730 dest = &tw->tw_v6_daddr;
1731 src = &tw->tw_v6_rcv_saddr;
1da177e4
LT
1732 destp = ntohs(tw->tw_dport);
1733 srcp = ntohs(tw->tw_sport);
1734
1735 seq_printf(seq,
1736 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
71338aa7 1737 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
1da177e4
LT
1738 i,
1739 src->s6_addr32[0], src->s6_addr32[1],
1740 src->s6_addr32[2], src->s6_addr32[3], srcp,
1741 dest->s6_addr32[0], dest->s6_addr32[1],
1742 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1743 tw->tw_substate, 0, 0,
a399a805 1744 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
1da177e4
LT
1745 atomic_read(&tw->tw_refcnt), tw);
1746}
1747
1da177e4
LT
1748static int tcp6_seq_show(struct seq_file *seq, void *v)
1749{
1750 struct tcp_iter_state *st;
05dbc7b5 1751 struct sock *sk = v;
1da177e4
LT
1752
1753 if (v == SEQ_START_TOKEN) {
1754 seq_puts(seq,
1755 " sl "
1756 "local_address "
1757 "remote_address "
1758 "st tx_queue rx_queue tr tm->when retrnsmt"
1759 " uid timeout inode\n");
1760 goto out;
1761 }
1762 st = seq->private;
1763
1764 switch (st->state) {
1765 case TCP_SEQ_STATE_LISTENING:
1766 case TCP_SEQ_STATE_ESTABLISHED:
05dbc7b5
ED
1767 if (sk->sk_state == TCP_TIME_WAIT)
1768 get_timewait6_sock(seq, v, st->num);
1769 else
1770 get_tcp6_sock(seq, v, st->num);
1da177e4
LT
1771 break;
1772 case TCP_SEQ_STATE_OPENREQ:
d4f06873 1773 get_openreq6(seq, v, st->num, st->uid);
1da177e4 1774 break;
1da177e4
LT
1775 }
1776out:
1777 return 0;
1778}
1779
73cb88ec
AV
1780static const struct file_operations tcp6_afinfo_seq_fops = {
1781 .owner = THIS_MODULE,
1782 .open = tcp_seq_open,
1783 .read = seq_read,
1784 .llseek = seq_lseek,
1785 .release = seq_release_net
1786};
1787
1da177e4 1788static struct tcp_seq_afinfo tcp6_seq_afinfo = {
1da177e4
LT
1789 .name = "tcp6",
1790 .family = AF_INET6,
73cb88ec 1791 .seq_fops = &tcp6_afinfo_seq_fops,
9427c4b3
DL
1792 .seq_ops = {
1793 .show = tcp6_seq_show,
1794 },
1da177e4
LT
1795};
1796
2c8c1e72 1797int __net_init tcp6_proc_init(struct net *net)
1da177e4 1798{
6f8b13bc 1799 return tcp_proc_register(net, &tcp6_seq_afinfo);
1da177e4
LT
1800}
1801
6f8b13bc 1802void tcp6_proc_exit(struct net *net)
1da177e4 1803{
6f8b13bc 1804 tcp_proc_unregister(net, &tcp6_seq_afinfo);
1da177e4
LT
1805}
1806#endif
1807
f77d6021
ED
1808static void tcp_v6_clear_sk(struct sock *sk, int size)
1809{
1810 struct inet_sock *inet = inet_sk(sk);
1811
1812 /* we do not want to clear pinet6 field, because of RCU lookups */
1813 sk_prot_clear_nulls(sk, offsetof(struct inet_sock, pinet6));
1814
1815 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1816 memset(&inet->pinet6 + 1, 0, size);
1817}
1818
1da177e4
LT
1819struct proto tcpv6_prot = {
1820 .name = "TCPv6",
1821 .owner = THIS_MODULE,
1822 .close = tcp_close,
1823 .connect = tcp_v6_connect,
1824 .disconnect = tcp_disconnect,
463c84b9 1825 .accept = inet_csk_accept,
1da177e4
LT
1826 .ioctl = tcp_ioctl,
1827 .init = tcp_v6_init_sock,
1828 .destroy = tcp_v6_destroy_sock,
1829 .shutdown = tcp_shutdown,
1830 .setsockopt = tcp_setsockopt,
1831 .getsockopt = tcp_getsockopt,
1da177e4 1832 .recvmsg = tcp_recvmsg,
7ba42910
CG
1833 .sendmsg = tcp_sendmsg,
1834 .sendpage = tcp_sendpage,
1da177e4 1835 .backlog_rcv = tcp_v6_do_rcv,
46d3ceab 1836 .release_cb = tcp_release_cb,
77a6a471 1837 .hash = inet_hash,
ab1e0a13
ACM
1838 .unhash = inet_unhash,
1839 .get_port = inet_csk_get_port,
1da177e4 1840 .enter_memory_pressure = tcp_enter_memory_pressure,
c9bee3b7 1841 .stream_memory_free = tcp_stream_memory_free,
1da177e4
LT
1842 .sockets_allocated = &tcp_sockets_allocated,
1843 .memory_allocated = &tcp_memory_allocated,
1844 .memory_pressure = &tcp_memory_pressure,
0a5578cf 1845 .orphan_count = &tcp_orphan_count,
a4fe34bf 1846 .sysctl_mem = sysctl_tcp_mem,
1da177e4
LT
1847 .sysctl_wmem = sysctl_tcp_wmem,
1848 .sysctl_rmem = sysctl_tcp_rmem,
1849 .max_header = MAX_TCP_HEADER,
1850 .obj_size = sizeof(struct tcp6_sock),
3ab5aee7 1851 .slab_flags = SLAB_DESTROY_BY_RCU,
6d6ee43e 1852 .twsk_prot = &tcp6_timewait_sock_ops,
60236fdd 1853 .rsk_prot = &tcp6_request_sock_ops,
39d8cda7 1854 .h.hashinfo = &tcp_hashinfo,
7ba42910 1855 .no_autobind = true,
543d9cfe
ACM
1856#ifdef CONFIG_COMPAT
1857 .compat_setsockopt = compat_tcp_setsockopt,
1858 .compat_getsockopt = compat_tcp_getsockopt,
1859#endif
c255a458 1860#ifdef CONFIG_MEMCG_KMEM
d1a4c0b3
GC
1861 .proto_cgroup = tcp_proto_cgroup,
1862#endif
f77d6021 1863 .clear_sk = tcp_v6_clear_sk,
1da177e4
LT
1864};
1865
41135cc8 1866static const struct inet6_protocol tcpv6_protocol = {
c7109986 1867 .early_demux = tcp_v6_early_demux,
1da177e4
LT
1868 .handler = tcp_v6_rcv,
1869 .err_handler = tcp_v6_err,
1870 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1871};
1872
1da177e4
LT
1873static struct inet_protosw tcpv6_protosw = {
1874 .type = SOCK_STREAM,
1875 .protocol = IPPROTO_TCP,
1876 .prot = &tcpv6_prot,
1877 .ops = &inet6_stream_ops,
d83d8461
ACM
1878 .flags = INET_PROTOSW_PERMANENT |
1879 INET_PROTOSW_ICSK,
1da177e4
LT
1880};
1881
2c8c1e72 1882static int __net_init tcpv6_net_init(struct net *net)
93ec926b 1883{
5677242f
DL
1884 return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
1885 SOCK_RAW, IPPROTO_TCP, net);
93ec926b
DL
1886}
1887
2c8c1e72 1888static void __net_exit tcpv6_net_exit(struct net *net)
93ec926b 1889{
5677242f 1890 inet_ctl_sock_destroy(net->ipv6.tcp_sk);
b099ce26
EB
1891}
1892
2c8c1e72 1893static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
b099ce26
EB
1894{
1895 inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6);
93ec926b
DL
1896}
1897
1898static struct pernet_operations tcpv6_net_ops = {
b099ce26
EB
1899 .init = tcpv6_net_init,
1900 .exit = tcpv6_net_exit,
1901 .exit_batch = tcpv6_net_exit_batch,
93ec926b
DL
1902};
1903
7f4e4868 1904int __init tcpv6_init(void)
1da177e4 1905{
7f4e4868
DL
1906 int ret;
1907
3336288a
VY
1908 ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
1909 if (ret)
c6b641a4 1910 goto out;
3336288a 1911
1da177e4 1912 /* register inet6 protocol */
7f4e4868
DL
1913 ret = inet6_register_protosw(&tcpv6_protosw);
1914 if (ret)
1915 goto out_tcpv6_protocol;
1916
93ec926b 1917 ret = register_pernet_subsys(&tcpv6_net_ops);
7f4e4868
DL
1918 if (ret)
1919 goto out_tcpv6_protosw;
1920out:
1921 return ret;
ae0f7d5f 1922
7f4e4868
DL
1923out_tcpv6_protosw:
1924 inet6_unregister_protosw(&tcpv6_protosw);
3336288a
VY
1925out_tcpv6_protocol:
1926 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
7f4e4868
DL
1927 goto out;
1928}
1929
09f7709f 1930void tcpv6_exit(void)
7f4e4868 1931{
93ec926b 1932 unregister_pernet_subsys(&tcpv6_net_ops);
7f4e4868
DL
1933 inet6_unregister_protosw(&tcpv6_protosw);
1934 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
1da177e4 1935}