bnx2x: Add new 57840 device IDs
[GitHub/mt8127/android_kernel_alcatel_ttab.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>
1da177e4
LT
42
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>
18134bed 62#include <net/netdma.h>
3d58b5fa 63#include <net/inet_common.h>
6e5714ea 64#include <net/secure_seq.h>
d1a4c0b3 65#include <net/tcp_memcontrol.h>
1da177e4
LT
66
67#include <asm/uaccess.h>
68
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
71
cfb6eeb4
YH
72#include <linux/crypto.h>
73#include <linux/scatterlist.h>
74
cfb6eeb4 75static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb);
6edafaaf
GJ
76static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
77 struct request_sock *req);
1da177e4
LT
78
79static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
8ad50d96 80static void __tcp_v6_send_check(struct sk_buff *skb,
b71d1d42
ED
81 const struct in6_addr *saddr,
82 const struct in6_addr *daddr);
1da177e4 83
3b401a81
SH
84static const struct inet_connection_sock_af_ops ipv6_mapped;
85static const struct inet_connection_sock_af_ops ipv6_specific;
a928630a 86#ifdef CONFIG_TCP_MD5SIG
b2e4b3de
SH
87static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
88static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
9501f972
YH
89#else
90static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
b71d1d42 91 const struct in6_addr *addr)
9501f972
YH
92{
93 return NULL;
94}
a928630a 95#endif
1da177e4 96
1da177e4
LT
97static void tcp_v6_hash(struct sock *sk)
98{
99 if (sk->sk_state != TCP_CLOSE) {
8292a17a 100 if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
1da177e4
LT
101 tcp_prot.hash(sk);
102 return;
103 }
104 local_bh_disable();
9327f705 105 __inet6_hash(sk, NULL);
1da177e4
LT
106 local_bh_enable();
107 }
108}
109
684f2176 110static __inline__ __sum16 tcp_v6_check(int len,
b71d1d42
ED
111 const struct in6_addr *saddr,
112 const struct in6_addr *daddr,
868c86bc 113 __wsum base)
1da177e4
LT
114{
115 return csum_ipv6_magic(saddr, daddr, len, IPPROTO_TCP, base);
116}
117
cf533ea5 118static __u32 tcp_v6_init_sequence(const struct sk_buff *skb)
1da177e4 119{
0660e03f
ACM
120 return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
121 ipv6_hdr(skb)->saddr.s6_addr32,
aa8223c7
ACM
122 tcp_hdr(skb)->dest,
123 tcp_hdr(skb)->source);
1da177e4
LT
124}
125
1ab1457c 126static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
1da177e4
LT
127 int addr_len)
128{
129 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
1ab1457c 130 struct inet_sock *inet = inet_sk(sk);
d83d8461 131 struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4
LT
132 struct ipv6_pinfo *np = inet6_sk(sk);
133 struct tcp_sock *tp = tcp_sk(sk);
20c59de2 134 struct in6_addr *saddr = NULL, *final_p, final;
493f377d 135 struct rt6_info *rt;
4c9483b2 136 struct flowi6 fl6;
1da177e4
LT
137 struct dst_entry *dst;
138 int addr_type;
139 int err;
140
1ab1457c 141 if (addr_len < SIN6_LEN_RFC2133)
1da177e4
LT
142 return -EINVAL;
143
1ab1457c 144 if (usin->sin6_family != AF_INET6)
a02cec21 145 return -EAFNOSUPPORT;
1da177e4 146
4c9483b2 147 memset(&fl6, 0, sizeof(fl6));
1da177e4
LT
148
149 if (np->sndflow) {
4c9483b2
DM
150 fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
151 IP6_ECN_flow_init(fl6.flowlabel);
152 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1da177e4 153 struct ip6_flowlabel *flowlabel;
4c9483b2 154 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1da177e4
LT
155 if (flowlabel == NULL)
156 return -EINVAL;
4e3fd7a0 157 usin->sin6_addr = flowlabel->dst;
1da177e4
LT
158 fl6_sock_release(flowlabel);
159 }
160 }
161
162 /*
1ab1457c
YH
163 * connect() to INADDR_ANY means loopback (BSD'ism).
164 */
165
166 if(ipv6_addr_any(&usin->sin6_addr))
167 usin->sin6_addr.s6_addr[15] = 0x1;
1da177e4
LT
168
169 addr_type = ipv6_addr_type(&usin->sin6_addr);
170
171 if(addr_type & IPV6_ADDR_MULTICAST)
172 return -ENETUNREACH;
173
174 if (addr_type&IPV6_ADDR_LINKLOCAL) {
175 if (addr_len >= sizeof(struct sockaddr_in6) &&
176 usin->sin6_scope_id) {
177 /* If interface is set while binding, indices
178 * must coincide.
179 */
180 if (sk->sk_bound_dev_if &&
181 sk->sk_bound_dev_if != usin->sin6_scope_id)
182 return -EINVAL;
183
184 sk->sk_bound_dev_if = usin->sin6_scope_id;
185 }
186
187 /* Connect to link-local address requires an interface */
188 if (!sk->sk_bound_dev_if)
189 return -EINVAL;
190 }
191
192 if (tp->rx_opt.ts_recent_stamp &&
193 !ipv6_addr_equal(&np->daddr, &usin->sin6_addr)) {
194 tp->rx_opt.ts_recent = 0;
195 tp->rx_opt.ts_recent_stamp = 0;
196 tp->write_seq = 0;
197 }
198
4e3fd7a0 199 np->daddr = usin->sin6_addr;
4c9483b2 200 np->flow_label = fl6.flowlabel;
1da177e4
LT
201
202 /*
203 * TCP over IPv4
204 */
205
206 if (addr_type == IPV6_ADDR_MAPPED) {
d83d8461 207 u32 exthdrlen = icsk->icsk_ext_hdr_len;
1da177e4
LT
208 struct sockaddr_in sin;
209
210 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
211
212 if (__ipv6_only_sock(sk))
213 return -ENETUNREACH;
214
215 sin.sin_family = AF_INET;
216 sin.sin_port = usin->sin6_port;
217 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
218
d83d8461 219 icsk->icsk_af_ops = &ipv6_mapped;
1da177e4 220 sk->sk_backlog_rcv = tcp_v4_do_rcv;
cfb6eeb4
YH
221#ifdef CONFIG_TCP_MD5SIG
222 tp->af_specific = &tcp_sock_ipv6_mapped_specific;
223#endif
1da177e4
LT
224
225 err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
226
227 if (err) {
d83d8461
ACM
228 icsk->icsk_ext_hdr_len = exthdrlen;
229 icsk->icsk_af_ops = &ipv6_specific;
1da177e4 230 sk->sk_backlog_rcv = tcp_v6_do_rcv;
cfb6eeb4
YH
231#ifdef CONFIG_TCP_MD5SIG
232 tp->af_specific = &tcp_sock_ipv6_specific;
233#endif
1da177e4
LT
234 goto failure;
235 } else {
c720c7e8
ED
236 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr);
237 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr,
238 &np->rcv_saddr);
1da177e4
LT
239 }
240
241 return err;
242 }
243
244 if (!ipv6_addr_any(&np->rcv_saddr))
245 saddr = &np->rcv_saddr;
246
4c9483b2 247 fl6.flowi6_proto = IPPROTO_TCP;
4e3fd7a0
AD
248 fl6.daddr = np->daddr;
249 fl6.saddr = saddr ? *saddr : np->saddr;
4c9483b2
DM
250 fl6.flowi6_oif = sk->sk_bound_dev_if;
251 fl6.flowi6_mark = sk->sk_mark;
1958b856
DM
252 fl6.fl6_dport = usin->sin6_port;
253 fl6.fl6_sport = inet->inet_sport;
1da177e4 254
4c9483b2 255 final_p = fl6_update_dst(&fl6, np->opt, &final);
1da177e4 256
4c9483b2 257 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
beb8d13b 258
4c9483b2 259 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
68d0c6d3
DM
260 if (IS_ERR(dst)) {
261 err = PTR_ERR(dst);
1da177e4 262 goto failure;
14e50e57 263 }
1da177e4
LT
264
265 if (saddr == NULL) {
4c9483b2 266 saddr = &fl6.saddr;
4e3fd7a0 267 np->rcv_saddr = *saddr;
1da177e4
LT
268 }
269
270 /* set the source address */
4e3fd7a0 271 np->saddr = *saddr;
c720c7e8 272 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
1da177e4 273
f83ef8c0 274 sk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 275 __ip6_dst_store(sk, dst, NULL, NULL);
1da177e4 276
493f377d
DM
277 rt = (struct rt6_info *) dst;
278 if (tcp_death_row.sysctl_tw_recycle &&
279 !tp->rx_opt.ts_recent_stamp &&
81166dd6
DM
280 ipv6_addr_equal(&rt->rt6i_dst.addr, &np->daddr))
281 tcp_fetch_timewait_stamp(sk, dst);
493f377d 282
d83d8461 283 icsk->icsk_ext_hdr_len = 0;
1da177e4 284 if (np->opt)
d83d8461
ACM
285 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
286 np->opt->opt_nflen);
1da177e4
LT
287
288 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
289
c720c7e8 290 inet->inet_dport = usin->sin6_port;
1da177e4
LT
291
292 tcp_set_state(sk, TCP_SYN_SENT);
d8313f5c 293 err = inet6_hash_connect(&tcp_death_row, sk);
1da177e4
LT
294 if (err)
295 goto late_failure;
296
297 if (!tp->write_seq)
298 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
299 np->daddr.s6_addr32,
c720c7e8
ED
300 inet->inet_sport,
301 inet->inet_dport);
1da177e4
LT
302
303 err = tcp_connect(sk);
304 if (err)
305 goto late_failure;
306
307 return 0;
308
309late_failure:
310 tcp_set_state(sk, TCP_CLOSE);
311 __sk_dst_reset(sk);
312failure:
c720c7e8 313 inet->inet_dport = 0;
1da177e4
LT
314 sk->sk_route_caps = 0;
315 return err;
316}
317
318static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 319 u8 type, u8 code, int offset, __be32 info)
1da177e4 320{
b71d1d42 321 const struct ipv6hdr *hdr = (const struct ipv6hdr*)skb->data;
505cbfc5 322 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
1da177e4
LT
323 struct ipv6_pinfo *np;
324 struct sock *sk;
325 int err;
1ab1457c 326 struct tcp_sock *tp;
1da177e4 327 __u32 seq;
ca12a1a4 328 struct net *net = dev_net(skb->dev);
1da177e4 329
ca12a1a4 330 sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr,
d86e0dac 331 th->dest, &hdr->saddr, th->source, skb->dev->ifindex);
1da177e4
LT
332
333 if (sk == NULL) {
e41b5368
DL
334 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
335 ICMP6_MIB_INERRORS);
1da177e4
LT
336 return;
337 }
338
339 if (sk->sk_state == TCP_TIME_WAIT) {
9469c7b4 340 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
341 return;
342 }
343
344 bh_lock_sock(sk);
345 if (sock_owned_by_user(sk))
de0744af 346 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
1da177e4
LT
347
348 if (sk->sk_state == TCP_CLOSE)
349 goto out;
350
e802af9c
SH
351 if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) {
352 NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
353 goto out;
354 }
355
1da177e4 356 tp = tcp_sk(sk);
1ab1457c 357 seq = ntohl(th->seq);
1da177e4
LT
358 if (sk->sk_state != TCP_LISTEN &&
359 !between(seq, tp->snd_una, tp->snd_nxt)) {
de0744af 360 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
1da177e4
LT
361 goto out;
362 }
363
364 np = inet6_sk(sk);
365
ec18d9a2
DM
366 if (type == NDISC_REDIRECT) {
367 struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
368
1ed5c48f 369 if (dst)
6700c270 370 dst->ops->redirect(dst, sk, skb);
ec18d9a2
DM
371 }
372
1da177e4 373 if (type == ICMPV6_PKT_TOOBIG) {
68d0c6d3 374 struct dst_entry *dst;
1da177e4
LT
375
376 if (sock_owned_by_user(sk))
377 goto out;
378 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
379 goto out;
380
35ad9b9c
DM
381 dst = inet6_csk_update_pmtu(sk, ntohl(info));
382 if (!dst)
383 goto out;
81aded24 384
d83d8461 385 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
1da177e4
LT
386 tcp_sync_mss(sk, dst_mtu(dst));
387 tcp_simple_retransmit(sk);
35ad9b9c 388 }
1da177e4
LT
389 goto out;
390 }
391
392 icmpv6_err_convert(type, code, &err);
393
60236fdd 394 /* Might be for an request_sock */
1da177e4 395 switch (sk->sk_state) {
60236fdd 396 struct request_sock *req, **prev;
1da177e4
LT
397 case TCP_LISTEN:
398 if (sock_owned_by_user(sk))
399 goto out;
400
8129765a
ACM
401 req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
402 &hdr->saddr, inet6_iif(skb));
1da177e4
LT
403 if (!req)
404 goto out;
405
406 /* ICMPs are not backlogged, hence we cannot get
407 * an established socket here.
408 */
547b792c 409 WARN_ON(req->sk != NULL);
1da177e4 410
2e6599cb 411 if (seq != tcp_rsk(req)->snt_isn) {
de0744af 412 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
1da177e4
LT
413 goto out;
414 }
415
463c84b9 416 inet_csk_reqsk_queue_drop(sk, req, prev);
1da177e4
LT
417 goto out;
418
419 case TCP_SYN_SENT:
420 case TCP_SYN_RECV: /* Cannot happen.
1ab1457c 421 It can, it SYNs are crossed. --ANK */
1da177e4 422 if (!sock_owned_by_user(sk)) {
1da177e4
LT
423 sk->sk_err = err;
424 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
425
426 tcp_done(sk);
427 } else
428 sk->sk_err_soft = err;
429 goto out;
430 }
431
432 if (!sock_owned_by_user(sk) && np->recverr) {
433 sk->sk_err = err;
434 sk->sk_error_report(sk);
435 } else
436 sk->sk_err_soft = err;
437
438out:
439 bh_unlock_sock(sk);
440 sock_put(sk);
441}
442
443
9f10d3f6
NC
444static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst,
445 struct flowi6 *fl6,
3840a06e 446 struct request_sock *req,
fff32699
ED
447 struct request_values *rvp,
448 u16 queue_mapping)
1da177e4 449{
ca304b61 450 struct inet6_request_sock *treq = inet6_rsk(req);
1da177e4
LT
451 struct ipv6_pinfo *np = inet6_sk(sk);
452 struct sk_buff * skb;
9494218f 453 int err = -ENOMEM;
1da177e4 454
9f10d3f6
NC
455 /* First, grab a route. */
456 if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL)
fd80eb94 457 goto done;
9494218f 458
e6b4d113 459 skb = tcp_make_synack(sk, dst, req, rvp);
9494218f 460
1da177e4 461 if (skb) {
8ad50d96 462 __tcp_v6_send_check(skb, &treq->loc_addr, &treq->rmt_addr);
1da177e4 463
9f10d3f6 464 fl6->daddr = treq->rmt_addr;
fff32699 465 skb_set_queue_mapping(skb, queue_mapping);
43264e0b 466 err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass);
b9df3cb8 467 err = net_xmit_eval(err);
1da177e4
LT
468 }
469
470done:
1da177e4
LT
471 return err;
472}
473
72659ecc
OP
474static int tcp_v6_rtx_synack(struct sock *sk, struct request_sock *req,
475 struct request_values *rvp)
476{
9f10d3f6
NC
477 struct flowi6 fl6;
478
72659ecc 479 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS);
9f10d3f6 480 return tcp_v6_send_synack(sk, NULL, &fl6, req, rvp, 0);
72659ecc
OP
481}
482
60236fdd 483static void tcp_v6_reqsk_destructor(struct request_sock *req)
1da177e4 484{
800d55f1 485 kfree_skb(inet6_rsk(req)->pktopts);
1da177e4
LT
486}
487
cfb6eeb4
YH
488#ifdef CONFIG_TCP_MD5SIG
489static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
b71d1d42 490 const struct in6_addr *addr)
cfb6eeb4 491{
a915da9b 492 return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
cfb6eeb4
YH
493}
494
495static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
496 struct sock *addr_sk)
497{
498 return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr);
499}
500
501static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
502 struct request_sock *req)
503{
504 return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr);
505}
506
cfb6eeb4
YH
507static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval,
508 int optlen)
509{
510 struct tcp_md5sig cmd;
511 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
cfb6eeb4
YH
512
513 if (optlen < sizeof(cmd))
514 return -EINVAL;
515
516 if (copy_from_user(&cmd, optval, sizeof(cmd)))
517 return -EFAULT;
518
519 if (sin6->sin6_family != AF_INET6)
520 return -EINVAL;
521
522 if (!cmd.tcpm_keylen) {
e773e4fa 523 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
a915da9b
ED
524 return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
525 AF_INET);
526 return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
527 AF_INET6);
cfb6eeb4
YH
528 }
529
530 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
531 return -EINVAL;
532
a915da9b
ED
533 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
534 return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
535 AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
cfb6eeb4 536
a915da9b
ED
537 return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
538 AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
cfb6eeb4
YH
539}
540
49a72dfb 541static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
b71d1d42
ED
542 const struct in6_addr *daddr,
543 const struct in6_addr *saddr, int nbytes)
cfb6eeb4 544{
cfb6eeb4 545 struct tcp6_pseudohdr *bp;
49a72dfb 546 struct scatterlist sg;
8d26d76d 547
cfb6eeb4 548 bp = &hp->md5_blk.ip6;
cfb6eeb4 549 /* 1. TCP pseudo-header (RFC2460) */
4e3fd7a0
AD
550 bp->saddr = *saddr;
551 bp->daddr = *daddr;
49a72dfb 552 bp->protocol = cpu_to_be32(IPPROTO_TCP);
00b1304c 553 bp->len = cpu_to_be32(nbytes);
cfb6eeb4 554
49a72dfb
AL
555 sg_init_one(&sg, bp, sizeof(*bp));
556 return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
557}
c7da57a1 558
49a72dfb 559static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
b71d1d42 560 const struct in6_addr *daddr, struct in6_addr *saddr,
318cf7aa 561 const struct tcphdr *th)
49a72dfb
AL
562{
563 struct tcp_md5sig_pool *hp;
564 struct hash_desc *desc;
565
566 hp = tcp_get_md5sig_pool();
567 if (!hp)
568 goto clear_hash_noput;
569 desc = &hp->md5_desc;
570
571 if (crypto_hash_init(desc))
572 goto clear_hash;
573 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2))
574 goto clear_hash;
575 if (tcp_md5_hash_header(hp, th))
576 goto clear_hash;
577 if (tcp_md5_hash_key(hp, key))
578 goto clear_hash;
579 if (crypto_hash_final(desc, md5_hash))
cfb6eeb4 580 goto clear_hash;
cfb6eeb4 581
cfb6eeb4 582 tcp_put_md5sig_pool();
cfb6eeb4 583 return 0;
49a72dfb 584
cfb6eeb4
YH
585clear_hash:
586 tcp_put_md5sig_pool();
587clear_hash_noput:
588 memset(md5_hash, 0, 16);
49a72dfb 589 return 1;
cfb6eeb4
YH
590}
591
49a72dfb 592static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
318cf7aa
ED
593 const struct sock *sk,
594 const struct request_sock *req,
595 const struct sk_buff *skb)
cfb6eeb4 596{
b71d1d42 597 const struct in6_addr *saddr, *daddr;
49a72dfb
AL
598 struct tcp_md5sig_pool *hp;
599 struct hash_desc *desc;
318cf7aa 600 const struct tcphdr *th = tcp_hdr(skb);
cfb6eeb4
YH
601
602 if (sk) {
603 saddr = &inet6_sk(sk)->saddr;
604 daddr = &inet6_sk(sk)->daddr;
49a72dfb 605 } else if (req) {
cfb6eeb4
YH
606 saddr = &inet6_rsk(req)->loc_addr;
607 daddr = &inet6_rsk(req)->rmt_addr;
49a72dfb 608 } else {
b71d1d42 609 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
49a72dfb
AL
610 saddr = &ip6h->saddr;
611 daddr = &ip6h->daddr;
cfb6eeb4 612 }
49a72dfb
AL
613
614 hp = tcp_get_md5sig_pool();
615 if (!hp)
616 goto clear_hash_noput;
617 desc = &hp->md5_desc;
618
619 if (crypto_hash_init(desc))
620 goto clear_hash;
621
622 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len))
623 goto clear_hash;
624 if (tcp_md5_hash_header(hp, th))
625 goto clear_hash;
626 if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
627 goto clear_hash;
628 if (tcp_md5_hash_key(hp, key))
629 goto clear_hash;
630 if (crypto_hash_final(desc, md5_hash))
631 goto clear_hash;
632
633 tcp_put_md5sig_pool();
634 return 0;
635
636clear_hash:
637 tcp_put_md5sig_pool();
638clear_hash_noput:
639 memset(md5_hash, 0, 16);
640 return 1;
cfb6eeb4
YH
641}
642
318cf7aa 643static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
cfb6eeb4 644{
cf533ea5 645 const __u8 *hash_location = NULL;
cfb6eeb4 646 struct tcp_md5sig_key *hash_expected;
b71d1d42 647 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
318cf7aa 648 const struct tcphdr *th = tcp_hdr(skb);
cfb6eeb4 649 int genhash;
cfb6eeb4
YH
650 u8 newhash[16];
651
652 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
7d5d5525 653 hash_location = tcp_parse_md5sig_option(th);
cfb6eeb4 654
785957d3
DM
655 /* We've parsed the options - do we have a hash? */
656 if (!hash_expected && !hash_location)
657 return 0;
658
659 if (hash_expected && !hash_location) {
660 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
cfb6eeb4
YH
661 return 1;
662 }
663
785957d3
DM
664 if (!hash_expected && hash_location) {
665 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
cfb6eeb4
YH
666 return 1;
667 }
668
669 /* check the signature */
49a72dfb
AL
670 genhash = tcp_v6_md5_hash_skb(newhash,
671 hash_expected,
672 NULL, NULL, skb);
673
cfb6eeb4 674 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
e87cc472
JP
675 net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n",
676 genhash ? "failed" : "mismatch",
677 &ip6h->saddr, ntohs(th->source),
678 &ip6h->daddr, ntohs(th->dest));
cfb6eeb4
YH
679 return 1;
680 }
681 return 0;
682}
683#endif
684
c6aefafb 685struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
1da177e4 686 .family = AF_INET6,
2e6599cb 687 .obj_size = sizeof(struct tcp6_request_sock),
72659ecc 688 .rtx_syn_ack = tcp_v6_rtx_synack,
60236fdd
ACM
689 .send_ack = tcp_v6_reqsk_send_ack,
690 .destructor = tcp_v6_reqsk_destructor,
72659ecc
OP
691 .send_reset = tcp_v6_send_reset,
692 .syn_ack_timeout = tcp_syn_ack_timeout,
1da177e4
LT
693};
694
cfb6eeb4 695#ifdef CONFIG_TCP_MD5SIG
b2e4b3de 696static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
cfb6eeb4 697 .md5_lookup = tcp_v6_reqsk_md5_lookup,
e3afe7b7 698 .calc_md5_hash = tcp_v6_md5_hash_skb,
cfb6eeb4 699};
b6332e6c 700#endif
cfb6eeb4 701
8ad50d96 702static void __tcp_v6_send_check(struct sk_buff *skb,
b71d1d42 703 const struct in6_addr *saddr, const struct in6_addr *daddr)
1da177e4 704{
aa8223c7 705 struct tcphdr *th = tcp_hdr(skb);
1da177e4 706
84fa7933 707 if (skb->ip_summed == CHECKSUM_PARTIAL) {
8ad50d96 708 th->check = ~tcp_v6_check(skb->len, saddr, daddr, 0);
663ead3b 709 skb->csum_start = skb_transport_header(skb) - skb->head;
ff1dcadb 710 skb->csum_offset = offsetof(struct tcphdr, check);
1da177e4 711 } else {
8ad50d96
HX
712 th->check = tcp_v6_check(skb->len, saddr, daddr,
713 csum_partial(th, th->doff << 2,
714 skb->csum));
1da177e4
LT
715 }
716}
717
bb296246 718static void tcp_v6_send_check(struct sock *sk, struct sk_buff *skb)
8ad50d96
HX
719{
720 struct ipv6_pinfo *np = inet6_sk(sk);
721
722 __tcp_v6_send_check(skb, &np->saddr, &np->daddr);
723}
724
a430a43d
HX
725static int tcp_v6_gso_send_check(struct sk_buff *skb)
726{
b71d1d42 727 const struct ipv6hdr *ipv6h;
a430a43d
HX
728 struct tcphdr *th;
729
730 if (!pskb_may_pull(skb, sizeof(*th)))
731 return -EINVAL;
732
0660e03f 733 ipv6h = ipv6_hdr(skb);
aa8223c7 734 th = tcp_hdr(skb);
a430a43d
HX
735
736 th->check = 0;
84fa7933 737 skb->ip_summed = CHECKSUM_PARTIAL;
8ad50d96 738 __tcp_v6_send_check(skb, &ipv6h->saddr, &ipv6h->daddr);
a430a43d
HX
739 return 0;
740}
1da177e4 741
36990673
HX
742static struct sk_buff **tcp6_gro_receive(struct sk_buff **head,
743 struct sk_buff *skb)
684f2176 744{
b71d1d42 745 const struct ipv6hdr *iph = skb_gro_network_header(skb);
684f2176
HX
746
747 switch (skb->ip_summed) {
748 case CHECKSUM_COMPLETE:
86911732 749 if (!tcp_v6_check(skb_gro_len(skb), &iph->saddr, &iph->daddr,
684f2176
HX
750 skb->csum)) {
751 skb->ip_summed = CHECKSUM_UNNECESSARY;
752 break;
753 }
754
755 /* fall through */
756 case CHECKSUM_NONE:
757 NAPI_GRO_CB(skb)->flush = 1;
758 return NULL;
759 }
760
761 return tcp_gro_receive(head, skb);
762}
684f2176 763
36990673 764static int tcp6_gro_complete(struct sk_buff *skb)
684f2176 765{
b71d1d42 766 const struct ipv6hdr *iph = ipv6_hdr(skb);
684f2176
HX
767 struct tcphdr *th = tcp_hdr(skb);
768
769 th->check = ~tcp_v6_check(skb->len - skb_transport_offset(skb),
770 &iph->saddr, &iph->daddr, 0);
771 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
772
773 return tcp_gro_complete(skb);
774}
684f2176 775
626e264d 776static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win,
b903d324 777 u32 ts, struct tcp_md5sig_key *key, int rst, u8 tclass)
1da177e4 778{
cf533ea5
ED
779 const struct tcphdr *th = tcp_hdr(skb);
780 struct tcphdr *t1;
1da177e4 781 struct sk_buff *buff;
4c9483b2 782 struct flowi6 fl6;
adf30907 783 struct net *net = dev_net(skb_dst(skb)->dev);
e5047992 784 struct sock *ctl_sk = net->ipv6.tcp_sk;
77c676da 785 unsigned int tot_len = sizeof(struct tcphdr);
adf30907 786 struct dst_entry *dst;
81ada62d 787 __be32 *topt;
1da177e4 788
626e264d
IJ
789 if (ts)
790 tot_len += TCPOLEN_TSTAMP_ALIGNED;
cfb6eeb4 791#ifdef CONFIG_TCP_MD5SIG
cfb6eeb4
YH
792 if (key)
793 tot_len += TCPOLEN_MD5SIG_ALIGNED;
794#endif
795
cfb6eeb4 796 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1da177e4 797 GFP_ATOMIC);
1ab1457c
YH
798 if (buff == NULL)
799 return;
1da177e4 800
cfb6eeb4 801 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1da177e4 802
cfb6eeb4 803 t1 = (struct tcphdr *) skb_push(buff, tot_len);
6651ffc8 804 skb_reset_transport_header(buff);
1da177e4
LT
805
806 /* Swap the send and the receive. */
807 memset(t1, 0, sizeof(*t1));
808 t1->dest = th->source;
809 t1->source = th->dest;
cfb6eeb4 810 t1->doff = tot_len / 4;
626e264d
IJ
811 t1->seq = htonl(seq);
812 t1->ack_seq = htonl(ack);
813 t1->ack = !rst || !th->ack;
814 t1->rst = rst;
815 t1->window = htons(win);
1da177e4 816
81ada62d
IJ
817 topt = (__be32 *)(t1 + 1);
818
626e264d
IJ
819 if (ts) {
820 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
821 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
822 *topt++ = htonl(tcp_time_stamp);
823 *topt++ = htonl(ts);
824 }
825
cfb6eeb4
YH
826#ifdef CONFIG_TCP_MD5SIG
827 if (key) {
81ada62d
IJ
828 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
829 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
830 tcp_v6_md5_hash_hdr((__u8 *)topt, key,
78e645cb
IJ
831 &ipv6_hdr(skb)->saddr,
832 &ipv6_hdr(skb)->daddr, t1);
cfb6eeb4
YH
833 }
834#endif
835
4c9483b2 836 memset(&fl6, 0, sizeof(fl6));
4e3fd7a0
AD
837 fl6.daddr = ipv6_hdr(skb)->saddr;
838 fl6.saddr = ipv6_hdr(skb)->daddr;
1da177e4 839
e5700aff
DM
840 buff->ip_summed = CHECKSUM_PARTIAL;
841 buff->csum = 0;
842
4c9483b2 843 __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
1da177e4 844
4c9483b2
DM
845 fl6.flowi6_proto = IPPROTO_TCP;
846 fl6.flowi6_oif = inet6_iif(skb);
1958b856
DM
847 fl6.fl6_dport = t1->dest;
848 fl6.fl6_sport = t1->source;
4c9483b2 849 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
1da177e4 850
c20121ae
DL
851 /* Pass a socket to ip6_dst_lookup either it is for RST
852 * Underlying function will use this to retrieve the network
853 * namespace
854 */
4c9483b2 855 dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL, false);
68d0c6d3
DM
856 if (!IS_ERR(dst)) {
857 skb_dst_set(buff, dst);
b903d324 858 ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass);
68d0c6d3
DM
859 TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
860 if (rst)
861 TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS);
862 return;
1da177e4
LT
863 }
864
865 kfree_skb(buff);
866}
867
626e264d 868static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
1da177e4 869{
cf533ea5 870 const struct tcphdr *th = tcp_hdr(skb);
626e264d 871 u32 seq = 0, ack_seq = 0;
fa3e5b4e 872 struct tcp_md5sig_key *key = NULL;
658ddaaf
SL
873#ifdef CONFIG_TCP_MD5SIG
874 const __u8 *hash_location = NULL;
875 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
876 unsigned char newhash[16];
877 int genhash;
878 struct sock *sk1 = NULL;
879#endif
1da177e4 880
626e264d 881 if (th->rst)
1da177e4
LT
882 return;
883
626e264d
IJ
884 if (!ipv6_unicast_destination(skb))
885 return;
1da177e4 886
cfb6eeb4 887#ifdef CONFIG_TCP_MD5SIG
658ddaaf
SL
888 hash_location = tcp_parse_md5sig_option(th);
889 if (!sk && hash_location) {
890 /*
891 * active side is lost. Try to find listening socket through
892 * source port, and then find md5 key through listening socket.
893 * we are not loose security here:
894 * Incoming packet is checked with md5 hash with finding key,
895 * no RST generated if md5 hash doesn't match.
896 */
897 sk1 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev),
898 &tcp_hashinfo, &ipv6h->daddr,
899 ntohs(th->source), inet6_iif(skb));
900 if (!sk1)
901 return;
902
903 rcu_read_lock();
904 key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr);
905 if (!key)
906 goto release_sk1;
907
908 genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, NULL, skb);
909 if (genhash || memcmp(hash_location, newhash, 16) != 0)
910 goto release_sk1;
911 } else {
912 key = sk ? tcp_v6_md5_do_lookup(sk, &ipv6h->saddr) : NULL;
913 }
cfb6eeb4
YH
914#endif
915
626e264d
IJ
916 if (th->ack)
917 seq = ntohl(th->ack_seq);
918 else
919 ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len -
920 (th->doff << 2);
1da177e4 921
b903d324 922 tcp_v6_send_response(skb, seq, ack_seq, 0, 0, key, 1, 0);
658ddaaf
SL
923
924#ifdef CONFIG_TCP_MD5SIG
925release_sk1:
926 if (sk1) {
927 rcu_read_unlock();
928 sock_put(sk1);
929 }
930#endif
626e264d 931}
1da177e4 932
626e264d 933static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts,
b903d324 934 struct tcp_md5sig_key *key, u8 tclass)
626e264d 935{
b903d324 936 tcp_v6_send_response(skb, seq, ack, win, ts, key, 0, tclass);
1da177e4
LT
937}
938
939static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
940{
8feaf0c0 941 struct inet_timewait_sock *tw = inet_twsk(sk);
cfb6eeb4 942 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
1da177e4 943
9501f972 944 tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
8feaf0c0 945 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
b903d324
ED
946 tcptw->tw_ts_recent, tcp_twsk_md5_key(tcptw),
947 tw->tw_tclass);
1da177e4 948
8feaf0c0 949 inet_twsk_put(tw);
1da177e4
LT
950}
951
6edafaaf
GJ
952static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
953 struct request_sock *req)
1da177e4 954{
9501f972 955 tcp_v6_send_ack(skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, req->ts_recent,
b903d324 956 tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr), 0);
1da177e4
LT
957}
958
959
960static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
961{
60236fdd 962 struct request_sock *req, **prev;
aa8223c7 963 const struct tcphdr *th = tcp_hdr(skb);
1da177e4
LT
964 struct sock *nsk;
965
966 /* Find possible connection requests. */
8129765a 967 req = inet6_csk_search_req(sk, &prev, th->source,
0660e03f
ACM
968 &ipv6_hdr(skb)->saddr,
969 &ipv6_hdr(skb)->daddr, inet6_iif(skb));
1da177e4
LT
970 if (req)
971 return tcp_check_req(sk, skb, req, prev);
972
3b1e0a65 973 nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
d86e0dac
PE
974 &ipv6_hdr(skb)->saddr, th->source,
975 &ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb));
1da177e4
LT
976
977 if (nsk) {
978 if (nsk->sk_state != TCP_TIME_WAIT) {
979 bh_lock_sock(nsk);
980 return nsk;
981 }
9469c7b4 982 inet_twsk_put(inet_twsk(nsk));
1da177e4
LT
983 return NULL;
984 }
985
c6aefafb 986#ifdef CONFIG_SYN_COOKIES
af9b4738 987 if (!th->syn)
c6aefafb 988 sk = cookie_v6_check(sk, skb);
1da177e4
LT
989#endif
990 return sk;
991}
992
1da177e4
LT
993/* FIXME: this is substantially similar to the ipv4 code.
994 * Can some kind of merge be done? -- erics
995 */
996static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
997{
4957faad 998 struct tcp_extend_values tmp_ext;
e6b4d113 999 struct tcp_options_received tmp_opt;
cf533ea5 1000 const u8 *hash_location;
e6b4d113 1001 struct request_sock *req;
ca304b61 1002 struct inet6_request_sock *treq;
1da177e4 1003 struct ipv6_pinfo *np = inet6_sk(sk);
1da177e4 1004 struct tcp_sock *tp = tcp_sk(sk);
e6b4d113 1005 __u32 isn = TCP_SKB_CB(skb)->when;
493f377d 1006 struct dst_entry *dst = NULL;
3840a06e 1007 struct flowi6 fl6;
a2a385d6 1008 bool want_cookie = false;
1da177e4
LT
1009
1010 if (skb->protocol == htons(ETH_P_IP))
1011 return tcp_v4_conn_request(sk, skb);
1012
1013 if (!ipv6_unicast_destination(skb))
1ab1457c 1014 goto drop;
1da177e4 1015
463c84b9 1016 if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
946cedcc
ED
1017 want_cookie = tcp_syn_flood_action(sk, skb, "TCPv6");
1018 if (!want_cookie)
1019 goto drop;
1da177e4
LT
1020 }
1021
463c84b9 1022 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
1da177e4
LT
1023 goto drop;
1024
ca304b61 1025 req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
1da177e4
LT
1026 if (req == NULL)
1027 goto drop;
1028
cfb6eeb4
YH
1029#ifdef CONFIG_TCP_MD5SIG
1030 tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
1031#endif
1032
1da177e4
LT
1033 tcp_clear_options(&tmp_opt);
1034 tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
1035 tmp_opt.user_mss = tp->rx_opt.user_mss;
2100c8d2 1036 tcp_parse_options(skb, &tmp_opt, &hash_location, 0, NULL);
4957faad
WAS
1037
1038 if (tmp_opt.cookie_plus > 0 &&
1039 tmp_opt.saw_tstamp &&
1040 !tp->rx_opt.cookie_out_never &&
1041 (sysctl_tcp_cookie_size > 0 ||
1042 (tp->cookie_values != NULL &&
1043 tp->cookie_values->cookie_desired > 0))) {
1044 u8 *c;
1045 u32 *d;
1046 u32 *mess = &tmp_ext.cookie_bakery[COOKIE_DIGEST_WORDS];
1047 int l = tmp_opt.cookie_plus - TCPOLEN_COOKIE_BASE;
1048
1049 if (tcp_cookie_generator(&tmp_ext.cookie_bakery[0]) != 0)
1050 goto drop_and_free;
1051
1052 /* Secret recipe starts with IP addresses */
0eae88f3 1053 d = (__force u32 *)&ipv6_hdr(skb)->daddr.s6_addr32[0];
4957faad
WAS
1054 *mess++ ^= *d++;
1055 *mess++ ^= *d++;
1056 *mess++ ^= *d++;
1057 *mess++ ^= *d++;
0eae88f3 1058 d = (__force u32 *)&ipv6_hdr(skb)->saddr.s6_addr32[0];
4957faad
WAS
1059 *mess++ ^= *d++;
1060 *mess++ ^= *d++;
1061 *mess++ ^= *d++;
1062 *mess++ ^= *d++;
1063
1064 /* plus variable length Initiator Cookie */
1065 c = (u8 *)mess;
1066 while (l-- > 0)
1067 *c++ ^= *hash_location++;
1da177e4 1068
a2a385d6 1069 want_cookie = false; /* not our kind of cookie */
4957faad
WAS
1070 tmp_ext.cookie_out_never = 0; /* false */
1071 tmp_ext.cookie_plus = tmp_opt.cookie_plus;
1072 } else if (!tp->rx_opt.cookie_in_always) {
1073 /* redundant indications, but ensure initialization. */
1074 tmp_ext.cookie_out_never = 1; /* true */
1075 tmp_ext.cookie_plus = 0;
1076 } else {
1077 goto drop_and_free;
1078 }
1079 tmp_ext.cookie_in_always = tp->rx_opt.cookie_in_always;
1da177e4 1080
4dfc2817 1081 if (want_cookie && !tmp_opt.saw_tstamp)
c6aefafb 1082 tcp_clear_options(&tmp_opt);
c6aefafb 1083
1da177e4
LT
1084 tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
1085 tcp_openreq_init(req, &tmp_opt, skb);
1086
ca304b61 1087 treq = inet6_rsk(req);
4e3fd7a0
AD
1088 treq->rmt_addr = ipv6_hdr(skb)->saddr;
1089 treq->loc_addr = ipv6_hdr(skb)->daddr;
172d69e6 1090 if (!want_cookie || tmp_opt.tstamp_ok)
bd14b1b2 1091 TCP_ECN_create_request(req, skb);
c6aefafb 1092
4d0fe50c
ED
1093 treq->iif = sk->sk_bound_dev_if;
1094
1095 /* So that link locals have meaning */
1096 if (!sk->sk_bound_dev_if &&
1097 ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL)
1098 treq->iif = inet6_iif(skb);
1099
2bbdf389 1100 if (!isn) {
c6aefafb
GG
1101 if (ipv6_opt_accepted(sk, skb) ||
1102 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
1103 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
1104 atomic_inc(&skb->users);
1105 treq->pktopts = skb;
1106 }
493f377d
DM
1107
1108 if (want_cookie) {
2bbdf389
FW
1109 isn = cookie_v6_init_sequence(sk, skb, &req->mss);
1110 req->cookie_ts = tmp_opt.tstamp_ok;
493f377d
DM
1111 goto have_isn;
1112 }
1113
1114 /* VJ's idea. We save last timestamp seen
1115 * from the destination in peer table, when entering
1116 * state TIME-WAIT, and check against it before
1117 * accepting new connection request.
1118 *
1119 * If "isn" is not zero, this request hit alive
1120 * timewait bucket, so that all the necessary checks
1121 * are made in the function processing timewait state.
1122 */
1123 if (tmp_opt.saw_tstamp &&
1124 tcp_death_row.sysctl_tw_recycle &&
81166dd6
DM
1125 (dst = inet6_csk_route_req(sk, &fl6, req)) != NULL) {
1126 if (!tcp_peer_is_proven(req, dst, true)) {
493f377d
DM
1127 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED);
1128 goto drop_and_release;
1129 }
1130 }
1131 /* Kill the following clause, if you dislike this way. */
1132 else if (!sysctl_tcp_syncookies &&
1133 (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) <
1134 (sysctl_max_syn_backlog >> 2)) &&
81166dd6 1135 !tcp_peer_is_proven(req, dst, false)) {
493f377d
DM
1136 /* Without syncookies last quarter of
1137 * backlog is filled with destinations,
1138 * proven to be alive.
1139 * It means that we continue to communicate
1140 * to destinations, already remembered
1141 * to the moment of synflood.
1142 */
1143 LIMIT_NETDEBUG(KERN_DEBUG "TCP: drop open request from %pI6/%u\n",
1144 &treq->rmt_addr, ntohs(tcp_hdr(skb)->source));
1145 goto drop_and_release;
2bbdf389 1146 }
493f377d
DM
1147
1148 isn = tcp_v6_init_sequence(skb);
c6aefafb 1149 }
493f377d 1150have_isn:
2e6599cb 1151 tcp_rsk(req)->snt_isn = isn;
9ad7c049 1152 tcp_rsk(req)->snt_synack = tcp_time_stamp;
1da177e4 1153
437c5b53
NC
1154 if (security_inet_conn_request(sk, skb, req))
1155 goto drop_and_release;
4237c75c 1156
9f10d3f6 1157 if (tcp_v6_send_synack(sk, dst, &fl6, req,
fff32699
ED
1158 (struct request_values *)&tmp_ext,
1159 skb_get_queue_mapping(skb)) ||
4957faad 1160 want_cookie)
e6b4d113 1161 goto drop_and_free;
1da177e4 1162
e6b4d113
WAS
1163 inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
1164 return 0;
1da177e4 1165
493f377d
DM
1166drop_and_release:
1167 dst_release(dst);
e6b4d113
WAS
1168drop_and_free:
1169 reqsk_free(req);
1da177e4 1170drop:
1da177e4
LT
1171 return 0; /* don't send reset */
1172}
1173
1174static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
60236fdd 1175 struct request_sock *req,
1da177e4
LT
1176 struct dst_entry *dst)
1177{
78d15e82 1178 struct inet6_request_sock *treq;
1da177e4
LT
1179 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
1180 struct tcp6_sock *newtcp6sk;
1181 struct inet_sock *newinet;
1182 struct tcp_sock *newtp;
1183 struct sock *newsk;
cfb6eeb4
YH
1184#ifdef CONFIG_TCP_MD5SIG
1185 struct tcp_md5sig_key *key;
1186#endif
3840a06e 1187 struct flowi6 fl6;
1da177e4
LT
1188
1189 if (skb->protocol == htons(ETH_P_IP)) {
1190 /*
1191 * v6 mapped
1192 */
1193
1194 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1195
1ab1457c 1196 if (newsk == NULL)
1da177e4
LT
1197 return NULL;
1198
1199 newtcp6sk = (struct tcp6_sock *)newsk;
1200 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1201
1202 newinet = inet_sk(newsk);
1203 newnp = inet6_sk(newsk);
1204 newtp = tcp_sk(newsk);
1205
1206 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1207
c720c7e8 1208 ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr);
1da177e4 1209
c720c7e8 1210 ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr);
1da177e4 1211
4e3fd7a0 1212 newnp->rcv_saddr = newnp->saddr;
1da177e4 1213
8292a17a 1214 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1da177e4 1215 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
cfb6eeb4
YH
1216#ifdef CONFIG_TCP_MD5SIG
1217 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1218#endif
1219
676a1184
YZ
1220 newnp->ipv6_ac_list = NULL;
1221 newnp->ipv6_fl_list = NULL;
1da177e4
LT
1222 newnp->pktoptions = NULL;
1223 newnp->opt = NULL;
505cbfc5 1224 newnp->mcast_oif = inet6_iif(skb);
0660e03f 1225 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
4c507d28 1226 newnp->rcv_tclass = ipv6_tclass(ipv6_hdr(skb));
1da177e4 1227
e6848976
ACM
1228 /*
1229 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1230 * here, tcp_create_openreq_child now does this for us, see the comment in
1231 * that function for the gory details. -acme
1da177e4 1232 */
1da177e4
LT
1233
1234 /* It is tricky place. Until this moment IPv4 tcp
8292a17a 1235 worked with IPv6 icsk.icsk_af_ops.
1da177e4
LT
1236 Sync it now.
1237 */
d83d8461 1238 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1da177e4
LT
1239
1240 return newsk;
1241 }
1242
78d15e82 1243 treq = inet6_rsk(req);
1da177e4
LT
1244
1245 if (sk_acceptq_is_full(sk))
1246 goto out_overflow;
1247
493f377d 1248 if (!dst) {
3840a06e 1249 dst = inet6_csk_route_req(sk, &fl6, req);
493f377d 1250 if (!dst)
1da177e4 1251 goto out;
1ab1457c 1252 }
1da177e4
LT
1253
1254 newsk = tcp_create_openreq_child(sk, req, skb);
1255 if (newsk == NULL)
093d2823 1256 goto out_nonewsk;
1da177e4 1257
e6848976
ACM
1258 /*
1259 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1260 * count here, tcp_create_openreq_child now does this for us, see the
1261 * comment in that function for the gory details. -acme
1262 */
1da177e4 1263
59eed279 1264 newsk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 1265 __ip6_dst_store(newsk, dst, NULL, NULL);
1da177e4
LT
1266
1267 newtcp6sk = (struct tcp6_sock *)newsk;
1268 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1269
1270 newtp = tcp_sk(newsk);
1271 newinet = inet_sk(newsk);
1272 newnp = inet6_sk(newsk);
1273
1274 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1275
4e3fd7a0
AD
1276 newnp->daddr = treq->rmt_addr;
1277 newnp->saddr = treq->loc_addr;
1278 newnp->rcv_saddr = treq->loc_addr;
2e6599cb 1279 newsk->sk_bound_dev_if = treq->iif;
1da177e4 1280
1ab1457c 1281 /* Now IPv6 options...
1da177e4
LT
1282
1283 First: no IPv4 options.
1284 */
f6d8bd05 1285 newinet->inet_opt = NULL;
676a1184 1286 newnp->ipv6_ac_list = NULL;
d35690be 1287 newnp->ipv6_fl_list = NULL;
1da177e4
LT
1288
1289 /* Clone RX bits */
1290 newnp->rxopt.all = np->rxopt.all;
1291
1292 /* Clone pktoptions received with SYN */
1293 newnp->pktoptions = NULL;
2e6599cb
ACM
1294 if (treq->pktopts != NULL) {
1295 newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC);
ab185d7b 1296 consume_skb(treq->pktopts);
2e6599cb 1297 treq->pktopts = NULL;
1da177e4
LT
1298 if (newnp->pktoptions)
1299 skb_set_owner_r(newnp->pktoptions, newsk);
1300 }
1301 newnp->opt = NULL;
505cbfc5 1302 newnp->mcast_oif = inet6_iif(skb);
0660e03f 1303 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
4c507d28 1304 newnp->rcv_tclass = ipv6_tclass(ipv6_hdr(skb));
1da177e4
LT
1305
1306 /* Clone native IPv6 options from listening socket (if any)
1307
1308 Yes, keeping reference count would be much more clever,
1309 but we make one more one thing there: reattach optmem
1310 to newsk.
1311 */
43264e0b
RL
1312 if (np->opt)
1313 newnp->opt = ipv6_dup_options(newsk, np->opt);
1da177e4 1314
d83d8461 1315 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1da177e4 1316 if (newnp->opt)
d83d8461
ACM
1317 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1318 newnp->opt->opt_flen);
1da177e4 1319
5d424d5a 1320 tcp_mtup_init(newsk);
1da177e4 1321 tcp_sync_mss(newsk, dst_mtu(dst));
0dbaee3b 1322 newtp->advmss = dst_metric_advmss(dst);
d135c522
NC
1323 if (tcp_sk(sk)->rx_opt.user_mss &&
1324 tcp_sk(sk)->rx_opt.user_mss < newtp->advmss)
1325 newtp->advmss = tcp_sk(sk)->rx_opt.user_mss;
1326
1da177e4 1327 tcp_initialize_rcv_mss(newsk);
9ad7c049
JC
1328 if (tcp_rsk(req)->snt_synack)
1329 tcp_valid_rtt_meas(newsk,
1330 tcp_time_stamp - tcp_rsk(req)->snt_synack);
1331 newtp->total_retrans = req->retrans;
1da177e4 1332
c720c7e8
ED
1333 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
1334 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
1da177e4 1335
cfb6eeb4
YH
1336#ifdef CONFIG_TCP_MD5SIG
1337 /* Copy over the MD5 key from the original socket */
1338 if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) {
1339 /* We're using one, so create a matching key
1340 * on the newsk structure. If we fail to get
1341 * memory, then we end up not copying the key
1342 * across. Shucks.
1343 */
a915da9b
ED
1344 tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newnp->daddr,
1345 AF_INET6, key->key, key->keylen, GFP_ATOMIC);
cfb6eeb4
YH
1346 }
1347#endif
1348
093d2823
BS
1349 if (__inet_inherit_port(sk, newsk) < 0) {
1350 sock_put(newsk);
1351 goto out;
1352 }
9327f705 1353 __inet6_hash(newsk, NULL);
1da177e4
LT
1354
1355 return newsk;
1356
1357out_overflow:
de0744af 1358 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
093d2823 1359out_nonewsk:
1da177e4 1360 dst_release(dst);
093d2823
BS
1361out:
1362 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
1da177e4
LT
1363 return NULL;
1364}
1365
b51655b9 1366static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
1da177e4 1367{
84fa7933 1368 if (skb->ip_summed == CHECKSUM_COMPLETE) {
684f2176 1369 if (!tcp_v6_check(skb->len, &ipv6_hdr(skb)->saddr,
0660e03f 1370 &ipv6_hdr(skb)->daddr, skb->csum)) {
fb286bb2 1371 skb->ip_summed = CHECKSUM_UNNECESSARY;
1da177e4 1372 return 0;
fb286bb2 1373 }
1da177e4 1374 }
fb286bb2 1375
684f2176 1376 skb->csum = ~csum_unfold(tcp_v6_check(skb->len,
0660e03f
ACM
1377 &ipv6_hdr(skb)->saddr,
1378 &ipv6_hdr(skb)->daddr, 0));
fb286bb2 1379
1da177e4 1380 if (skb->len <= 76) {
fb286bb2 1381 return __skb_checksum_complete(skb);
1da177e4
LT
1382 }
1383 return 0;
1384}
1385
1386/* The socket must have it's spinlock held when we get
1387 * here.
1388 *
1389 * We have a potential double-lock case here, so even when
1390 * doing backlog processing we use the BH locking scheme.
1391 * This is because we cannot sleep with the original spinlock
1392 * held.
1393 */
1394static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1395{
1396 struct ipv6_pinfo *np = inet6_sk(sk);
1397 struct tcp_sock *tp;
1398 struct sk_buff *opt_skb = NULL;
1399
1400 /* Imagine: socket is IPv6. IPv4 packet arrives,
1401 goes to IPv4 receive handler and backlogged.
1402 From backlog it always goes here. Kerboom...
1403 Fortunately, tcp_rcv_established and rcv_established
1404 handle them correctly, but it is not case with
1405 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1406 */
1407
1408 if (skb->protocol == htons(ETH_P_IP))
1409 return tcp_v4_do_rcv(sk, skb);
1410
cfb6eeb4
YH
1411#ifdef CONFIG_TCP_MD5SIG
1412 if (tcp_v6_inbound_md5_hash (sk, skb))
1413 goto discard;
1414#endif
1415
fda9ef5d 1416 if (sk_filter(sk, skb))
1da177e4
LT
1417 goto discard;
1418
1419 /*
1420 * socket locking is here for SMP purposes as backlog rcv
1421 * is currently called with bh processing disabled.
1422 */
1423
1424 /* Do Stevens' IPV6_PKTOPTIONS.
1425
1426 Yes, guys, it is the only place in our code, where we
1427 may make it not affecting IPv4.
1428 The rest of code is protocol independent,
1429 and I do not like idea to uglify IPv4.
1430
1431 Actually, all the idea behind IPV6_PKTOPTIONS
1432 looks not very well thought. For now we latch
1433 options, received in the last packet, enqueued
1434 by tcp. Feel free to propose better solution.
1ab1457c 1435 --ANK (980728)
1da177e4
LT
1436 */
1437 if (np->rxopt.all)
1438 opt_skb = skb_clone(skb, GFP_ATOMIC);
1439
1440 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
bdeab991 1441 sock_rps_save_rxhash(sk, skb);
aa8223c7 1442 if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len))
1da177e4 1443 goto reset;
1da177e4
LT
1444 if (opt_skb)
1445 goto ipv6_pktoptions;
1446 return 0;
1447 }
1448
ab6a5bb6 1449 if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
1da177e4
LT
1450 goto csum_err;
1451
1ab1457c 1452 if (sk->sk_state == TCP_LISTEN) {
1da177e4
LT
1453 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1454 if (!nsk)
1455 goto discard;
1456
1457 /*
1458 * Queue it on the new socket if the new socket is active,
1459 * otherwise we just shortcircuit this and continue with
1460 * the new socket..
1461 */
1ab1457c 1462 if(nsk != sk) {
bdeab991 1463 sock_rps_save_rxhash(nsk, skb);
1da177e4
LT
1464 if (tcp_child_process(sk, nsk, skb))
1465 goto reset;
1466 if (opt_skb)
1467 __kfree_skb(opt_skb);
1468 return 0;
1469 }
47482f13 1470 } else
bdeab991 1471 sock_rps_save_rxhash(sk, skb);
1da177e4 1472
aa8223c7 1473 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
1da177e4 1474 goto reset;
1da177e4
LT
1475 if (opt_skb)
1476 goto ipv6_pktoptions;
1477 return 0;
1478
1479reset:
cfb6eeb4 1480 tcp_v6_send_reset(sk, skb);
1da177e4
LT
1481discard:
1482 if (opt_skb)
1483 __kfree_skb(opt_skb);
1484 kfree_skb(skb);
1485 return 0;
1486csum_err:
63231bdd 1487 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
1da177e4
LT
1488 goto discard;
1489
1490
1491ipv6_pktoptions:
1492 /* Do you ask, what is it?
1493
1494 1. skb was enqueued by tcp.
1495 2. skb is added to tail of read queue, rather than out of order.
1496 3. socket is not in passive state.
1497 4. Finally, it really contains options, which user wants to receive.
1498 */
1499 tp = tcp_sk(sk);
1500 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1501 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
333fad53 1502 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
505cbfc5 1503 np->mcast_oif = inet6_iif(opt_skb);
333fad53 1504 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
0660e03f 1505 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
4c507d28
JB
1506 if (np->rxopt.bits.rxtclass)
1507 np->rcv_tclass = ipv6_tclass(ipv6_hdr(skb));
1da177e4
LT
1508 if (ipv6_opt_accepted(sk, opt_skb)) {
1509 skb_set_owner_r(opt_skb, sk);
1510 opt_skb = xchg(&np->pktoptions, opt_skb);
1511 } else {
1512 __kfree_skb(opt_skb);
1513 opt_skb = xchg(&np->pktoptions, NULL);
1514 }
1515 }
1516
800d55f1 1517 kfree_skb(opt_skb);
1da177e4
LT
1518 return 0;
1519}
1520
e5bbef20 1521static int tcp_v6_rcv(struct sk_buff *skb)
1da177e4 1522{
cf533ea5 1523 const struct tcphdr *th;
b71d1d42 1524 const struct ipv6hdr *hdr;
1da177e4
LT
1525 struct sock *sk;
1526 int ret;
a86b1e30 1527 struct net *net = dev_net(skb->dev);
1da177e4
LT
1528
1529 if (skb->pkt_type != PACKET_HOST)
1530 goto discard_it;
1531
1532 /*
1533 * Count it even if it's bad.
1534 */
63231bdd 1535 TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
1da177e4
LT
1536
1537 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1538 goto discard_it;
1539
aa8223c7 1540 th = tcp_hdr(skb);
1da177e4
LT
1541
1542 if (th->doff < sizeof(struct tcphdr)/4)
1543 goto bad_packet;
1544 if (!pskb_may_pull(skb, th->doff*4))
1545 goto discard_it;
1546
60476372 1547 if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
1da177e4
LT
1548 goto bad_packet;
1549
aa8223c7 1550 th = tcp_hdr(skb);
e802af9c 1551 hdr = ipv6_hdr(skb);
1da177e4
LT
1552 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1553 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1554 skb->len - th->doff*4);
1555 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1556 TCP_SKB_CB(skb)->when = 0;
b82d1bb4 1557 TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
1da177e4
LT
1558 TCP_SKB_CB(skb)->sacked = 0;
1559
9a1f27c4 1560 sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest);
1da177e4
LT
1561 if (!sk)
1562 goto no_tcp_socket;
1563
1564process:
1565 if (sk->sk_state == TCP_TIME_WAIT)
1566 goto do_time_wait;
1567
e802af9c
SH
1568 if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) {
1569 NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
1570 goto discard_and_relse;
1571 }
1572
1da177e4
LT
1573 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1574 goto discard_and_relse;
1575
fda9ef5d 1576 if (sk_filter(sk, skb))
1da177e4
LT
1577 goto discard_and_relse;
1578
1579 skb->dev = NULL;
1580
293b9c42 1581 bh_lock_sock_nested(sk);
1da177e4
LT
1582 ret = 0;
1583 if (!sock_owned_by_user(sk)) {
1a2449a8 1584#ifdef CONFIG_NET_DMA
1ab1457c 1585 struct tcp_sock *tp = tcp_sk(sk);
b4caea8a 1586 if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
a2bd1140 1587 tp->ucopy.dma_chan = net_dma_find_channel();
1ab1457c
YH
1588 if (tp->ucopy.dma_chan)
1589 ret = tcp_v6_do_rcv(sk, skb);
1590 else
1a2449a8
CL
1591#endif
1592 {
1593 if (!tcp_prequeue(sk, skb))
1594 ret = tcp_v6_do_rcv(sk, skb);
1595 }
da882c1f
ED
1596 } else if (unlikely(sk_add_backlog(sk, skb,
1597 sk->sk_rcvbuf + sk->sk_sndbuf))) {
6b03a53a 1598 bh_unlock_sock(sk);
6cce09f8 1599 NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
6b03a53a
ZY
1600 goto discard_and_relse;
1601 }
1da177e4
LT
1602 bh_unlock_sock(sk);
1603
1604 sock_put(sk);
1605 return ret ? -1 : 0;
1606
1607no_tcp_socket:
1608 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1609 goto discard_it;
1610
1611 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1612bad_packet:
63231bdd 1613 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1da177e4 1614 } else {
cfb6eeb4 1615 tcp_v6_send_reset(NULL, skb);
1da177e4
LT
1616 }
1617
1618discard_it:
1619
1620 /*
1621 * Discard frame
1622 */
1623
1624 kfree_skb(skb);
1625 return 0;
1626
1627discard_and_relse:
1628 sock_put(sk);
1629 goto discard_it;
1630
1631do_time_wait:
1632 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
9469c7b4 1633 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1634 goto discard_it;
1635 }
1636
1637 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
63231bdd 1638 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
9469c7b4 1639 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1640 goto discard_it;
1641 }
1642
9469c7b4 1643 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1da177e4
LT
1644 case TCP_TW_SYN:
1645 {
1646 struct sock *sk2;
1647
c346dca1 1648 sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
0660e03f 1649 &ipv6_hdr(skb)->daddr,
505cbfc5 1650 ntohs(th->dest), inet6_iif(skb));
1da177e4 1651 if (sk2 != NULL) {
295ff7ed
ACM
1652 struct inet_timewait_sock *tw = inet_twsk(sk);
1653 inet_twsk_deschedule(tw, &tcp_death_row);
1654 inet_twsk_put(tw);
1da177e4
LT
1655 sk = sk2;
1656 goto process;
1657 }
1658 /* Fall through to ACK */
1659 }
1660 case TCP_TW_ACK:
1661 tcp_v6_timewait_ack(sk, skb);
1662 break;
1663 case TCP_TW_RST:
1664 goto no_tcp_socket;
1665 case TCP_TW_SUCCESS:;
1666 }
1667 goto discard_it;
1668}
1669
ccb7c410
DM
1670static struct timewait_sock_ops tcp6_timewait_sock_ops = {
1671 .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
1672 .twsk_unique = tcp_twsk_unique,
1673 .twsk_destructor= tcp_twsk_destructor,
ccb7c410
DM
1674};
1675
3b401a81 1676static const struct inet_connection_sock_af_ops ipv6_specific = {
543d9cfe
ACM
1677 .queue_xmit = inet6_csk_xmit,
1678 .send_check = tcp_v6_send_check,
1679 .rebuild_header = inet6_sk_rebuild_header,
1680 .conn_request = tcp_v6_conn_request,
1681 .syn_recv_sock = tcp_v6_syn_recv_sock,
543d9cfe 1682 .net_header_len = sizeof(struct ipv6hdr),
67469601 1683 .net_frag_header_len = sizeof(struct frag_hdr),
543d9cfe
ACM
1684 .setsockopt = ipv6_setsockopt,
1685 .getsockopt = ipv6_getsockopt,
1686 .addr2sockaddr = inet6_csk_addr2sockaddr,
1687 .sockaddr_len = sizeof(struct sockaddr_in6),
ab1e0a13 1688 .bind_conflict = inet6_csk_bind_conflict,
3fdadf7d 1689#ifdef CONFIG_COMPAT
543d9cfe
ACM
1690 .compat_setsockopt = compat_ipv6_setsockopt,
1691 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1692#endif
1da177e4
LT
1693};
1694
cfb6eeb4 1695#ifdef CONFIG_TCP_MD5SIG
b2e4b3de 1696static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
cfb6eeb4 1697 .md5_lookup = tcp_v6_md5_lookup,
49a72dfb 1698 .calc_md5_hash = tcp_v6_md5_hash_skb,
cfb6eeb4 1699 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1700};
a928630a 1701#endif
cfb6eeb4 1702
1da177e4
LT
1703/*
1704 * TCP over IPv4 via INET6 API
1705 */
1706
3b401a81 1707static const struct inet_connection_sock_af_ops ipv6_mapped = {
543d9cfe
ACM
1708 .queue_xmit = ip_queue_xmit,
1709 .send_check = tcp_v4_send_check,
1710 .rebuild_header = inet_sk_rebuild_header,
1711 .conn_request = tcp_v6_conn_request,
1712 .syn_recv_sock = tcp_v6_syn_recv_sock,
543d9cfe
ACM
1713 .net_header_len = sizeof(struct iphdr),
1714 .setsockopt = ipv6_setsockopt,
1715 .getsockopt = ipv6_getsockopt,
1716 .addr2sockaddr = inet6_csk_addr2sockaddr,
1717 .sockaddr_len = sizeof(struct sockaddr_in6),
ab1e0a13 1718 .bind_conflict = inet6_csk_bind_conflict,
3fdadf7d 1719#ifdef CONFIG_COMPAT
543d9cfe
ACM
1720 .compat_setsockopt = compat_ipv6_setsockopt,
1721 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1722#endif
1da177e4
LT
1723};
1724
cfb6eeb4 1725#ifdef CONFIG_TCP_MD5SIG
b2e4b3de 1726static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
cfb6eeb4 1727 .md5_lookup = tcp_v4_md5_lookup,
49a72dfb 1728 .calc_md5_hash = tcp_v4_md5_hash_skb,
cfb6eeb4 1729 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1730};
a928630a 1731#endif
cfb6eeb4 1732
1da177e4
LT
1733/* NOTE: A lot of things set to zero explicitly by call to
1734 * sk_alloc() so need not be done here.
1735 */
1736static int tcp_v6_init_sock(struct sock *sk)
1737{
6687e988 1738 struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4 1739
900f65d3 1740 tcp_init_sock(sk);
1da177e4 1741
8292a17a 1742 icsk->icsk_af_ops = &ipv6_specific;
1da177e4 1743
cfb6eeb4 1744#ifdef CONFIG_TCP_MD5SIG
ac807fa8 1745 tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
cfb6eeb4
YH
1746#endif
1747
1da177e4
LT
1748 return 0;
1749}
1750
7d06b2e0 1751static void tcp_v6_destroy_sock(struct sock *sk)
1da177e4 1752{
1da177e4 1753 tcp_v4_destroy_sock(sk);
7d06b2e0 1754 inet6_destroy_sock(sk);
1da177e4
LT
1755}
1756
952a10be 1757#ifdef CONFIG_PROC_FS
1da177e4 1758/* Proc filesystem TCPv6 sock list dumping. */
1ab1457c 1759static void get_openreq6(struct seq_file *seq,
cf533ea5 1760 const struct sock *sk, struct request_sock *req, int i, int uid)
1da177e4 1761{
1da177e4 1762 int ttd = req->expires - jiffies;
b71d1d42
ED
1763 const struct in6_addr *src = &inet6_rsk(req)->loc_addr;
1764 const struct in6_addr *dest = &inet6_rsk(req)->rmt_addr;
1da177e4
LT
1765
1766 if (ttd < 0)
1767 ttd = 0;
1768
1da177e4
LT
1769 seq_printf(seq,
1770 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
71338aa7 1771 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
1da177e4
LT
1772 i,
1773 src->s6_addr32[0], src->s6_addr32[1],
1774 src->s6_addr32[2], src->s6_addr32[3],
fd507037 1775 ntohs(inet_rsk(req)->loc_port),
1da177e4
LT
1776 dest->s6_addr32[0], dest->s6_addr32[1],
1777 dest->s6_addr32[2], dest->s6_addr32[3],
2e6599cb 1778 ntohs(inet_rsk(req)->rmt_port),
1da177e4
LT
1779 TCP_SYN_RECV,
1780 0,0, /* could print option size, but that is af dependent. */
1ab1457c
YH
1781 1, /* timers active (only the expire timer) */
1782 jiffies_to_clock_t(ttd),
1da177e4
LT
1783 req->retrans,
1784 uid,
1ab1457c 1785 0, /* non standard timer */
1da177e4
LT
1786 0, /* open_requests have no inode */
1787 0, req);
1788}
1789
1790static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1791{
b71d1d42 1792 const struct in6_addr *dest, *src;
1da177e4
LT
1793 __u16 destp, srcp;
1794 int timer_active;
1795 unsigned long timer_expires;
cf533ea5
ED
1796 const struct inet_sock *inet = inet_sk(sp);
1797 const struct tcp_sock *tp = tcp_sk(sp);
463c84b9 1798 const struct inet_connection_sock *icsk = inet_csk(sp);
cf533ea5 1799 const struct ipv6_pinfo *np = inet6_sk(sp);
1da177e4
LT
1800
1801 dest = &np->daddr;
1802 src = &np->rcv_saddr;
c720c7e8
ED
1803 destp = ntohs(inet->inet_dport);
1804 srcp = ntohs(inet->inet_sport);
463c84b9
ACM
1805
1806 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
1da177e4 1807 timer_active = 1;
463c84b9
ACM
1808 timer_expires = icsk->icsk_timeout;
1809 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
1da177e4 1810 timer_active = 4;
463c84b9 1811 timer_expires = icsk->icsk_timeout;
1da177e4
LT
1812 } else if (timer_pending(&sp->sk_timer)) {
1813 timer_active = 2;
1814 timer_expires = sp->sk_timer.expires;
1815 } else {
1816 timer_active = 0;
1817 timer_expires = jiffies;
1818 }
1819
1820 seq_printf(seq,
1821 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
71338aa7 1822 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %lu %lu %u %u %d\n",
1da177e4
LT
1823 i,
1824 src->s6_addr32[0], src->s6_addr32[1],
1825 src->s6_addr32[2], src->s6_addr32[3], srcp,
1826 dest->s6_addr32[0], dest->s6_addr32[1],
1827 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1ab1457c 1828 sp->sk_state,
47da8ee6
SS
1829 tp->write_seq-tp->snd_una,
1830 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
1da177e4
LT
1831 timer_active,
1832 jiffies_to_clock_t(timer_expires - jiffies),
463c84b9 1833 icsk->icsk_retransmits,
1da177e4 1834 sock_i_uid(sp),
6687e988 1835 icsk->icsk_probes_out,
1da177e4
LT
1836 sock_i_ino(sp),
1837 atomic_read(&sp->sk_refcnt), sp,
7be87351
SH
1838 jiffies_to_clock_t(icsk->icsk_rto),
1839 jiffies_to_clock_t(icsk->icsk_ack.ato),
463c84b9 1840 (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
0b6a05c1
IJ
1841 tp->snd_cwnd,
1842 tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh
1da177e4
LT
1843 );
1844}
1845
1ab1457c 1846static void get_timewait6_sock(struct seq_file *seq,
8feaf0c0 1847 struct inet_timewait_sock *tw, int i)
1da177e4 1848{
b71d1d42 1849 const struct in6_addr *dest, *src;
1da177e4 1850 __u16 destp, srcp;
cf533ea5 1851 const struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
1da177e4
LT
1852 int ttd = tw->tw_ttd - jiffies;
1853
1854 if (ttd < 0)
1855 ttd = 0;
1856
0fa1a53e
ACM
1857 dest = &tw6->tw_v6_daddr;
1858 src = &tw6->tw_v6_rcv_saddr;
1da177e4
LT
1859 destp = ntohs(tw->tw_dport);
1860 srcp = ntohs(tw->tw_sport);
1861
1862 seq_printf(seq,
1863 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
71338aa7 1864 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
1da177e4
LT
1865 i,
1866 src->s6_addr32[0], src->s6_addr32[1],
1867 src->s6_addr32[2], src->s6_addr32[3], srcp,
1868 dest->s6_addr32[0], dest->s6_addr32[1],
1869 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1870 tw->tw_substate, 0, 0,
1871 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
1872 atomic_read(&tw->tw_refcnt), tw);
1873}
1874
1da177e4
LT
1875static int tcp6_seq_show(struct seq_file *seq, void *v)
1876{
1877 struct tcp_iter_state *st;
1878
1879 if (v == SEQ_START_TOKEN) {
1880 seq_puts(seq,
1881 " sl "
1882 "local_address "
1883 "remote_address "
1884 "st tx_queue rx_queue tr tm->when retrnsmt"
1885 " uid timeout inode\n");
1886 goto out;
1887 }
1888 st = seq->private;
1889
1890 switch (st->state) {
1891 case TCP_SEQ_STATE_LISTENING:
1892 case TCP_SEQ_STATE_ESTABLISHED:
1893 get_tcp6_sock(seq, v, st->num);
1894 break;
1895 case TCP_SEQ_STATE_OPENREQ:
1896 get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
1897 break;
1898 case TCP_SEQ_STATE_TIME_WAIT:
1899 get_timewait6_sock(seq, v, st->num);
1900 break;
1901 }
1902out:
1903 return 0;
1904}
1905
73cb88ec
AV
1906static const struct file_operations tcp6_afinfo_seq_fops = {
1907 .owner = THIS_MODULE,
1908 .open = tcp_seq_open,
1909 .read = seq_read,
1910 .llseek = seq_lseek,
1911 .release = seq_release_net
1912};
1913
1da177e4 1914static struct tcp_seq_afinfo tcp6_seq_afinfo = {
1da177e4
LT
1915 .name = "tcp6",
1916 .family = AF_INET6,
73cb88ec 1917 .seq_fops = &tcp6_afinfo_seq_fops,
9427c4b3
DL
1918 .seq_ops = {
1919 .show = tcp6_seq_show,
1920 },
1da177e4
LT
1921};
1922
2c8c1e72 1923int __net_init tcp6_proc_init(struct net *net)
1da177e4 1924{
6f8b13bc 1925 return tcp_proc_register(net, &tcp6_seq_afinfo);
1da177e4
LT
1926}
1927
6f8b13bc 1928void tcp6_proc_exit(struct net *net)
1da177e4 1929{
6f8b13bc 1930 tcp_proc_unregister(net, &tcp6_seq_afinfo);
1da177e4
LT
1931}
1932#endif
1933
1934struct proto tcpv6_prot = {
1935 .name = "TCPv6",
1936 .owner = THIS_MODULE,
1937 .close = tcp_close,
1938 .connect = tcp_v6_connect,
1939 .disconnect = tcp_disconnect,
463c84b9 1940 .accept = inet_csk_accept,
1da177e4
LT
1941 .ioctl = tcp_ioctl,
1942 .init = tcp_v6_init_sock,
1943 .destroy = tcp_v6_destroy_sock,
1944 .shutdown = tcp_shutdown,
1945 .setsockopt = tcp_setsockopt,
1946 .getsockopt = tcp_getsockopt,
1da177e4 1947 .recvmsg = tcp_recvmsg,
7ba42910
CG
1948 .sendmsg = tcp_sendmsg,
1949 .sendpage = tcp_sendpage,
1da177e4 1950 .backlog_rcv = tcp_v6_do_rcv,
46d3ceab 1951 .release_cb = tcp_release_cb,
1da177e4 1952 .hash = tcp_v6_hash,
ab1e0a13
ACM
1953 .unhash = inet_unhash,
1954 .get_port = inet_csk_get_port,
1da177e4
LT
1955 .enter_memory_pressure = tcp_enter_memory_pressure,
1956 .sockets_allocated = &tcp_sockets_allocated,
1957 .memory_allocated = &tcp_memory_allocated,
1958 .memory_pressure = &tcp_memory_pressure,
0a5578cf 1959 .orphan_count = &tcp_orphan_count,
1da177e4
LT
1960 .sysctl_wmem = sysctl_tcp_wmem,
1961 .sysctl_rmem = sysctl_tcp_rmem,
1962 .max_header = MAX_TCP_HEADER,
1963 .obj_size = sizeof(struct tcp6_sock),
3ab5aee7 1964 .slab_flags = SLAB_DESTROY_BY_RCU,
6d6ee43e 1965 .twsk_prot = &tcp6_timewait_sock_ops,
60236fdd 1966 .rsk_prot = &tcp6_request_sock_ops,
39d8cda7 1967 .h.hashinfo = &tcp_hashinfo,
7ba42910 1968 .no_autobind = true,
543d9cfe
ACM
1969#ifdef CONFIG_COMPAT
1970 .compat_setsockopt = compat_tcp_setsockopt,
1971 .compat_getsockopt = compat_tcp_getsockopt,
1972#endif
d1a4c0b3
GC
1973#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
1974 .proto_cgroup = tcp_proto_cgroup,
1975#endif
1da177e4
LT
1976};
1977
41135cc8 1978static const struct inet6_protocol tcpv6_protocol = {
1da177e4
LT
1979 .handler = tcp_v6_rcv,
1980 .err_handler = tcp_v6_err,
a430a43d 1981 .gso_send_check = tcp_v6_gso_send_check,
adcfc7d0 1982 .gso_segment = tcp_tso_segment,
684f2176
HX
1983 .gro_receive = tcp6_gro_receive,
1984 .gro_complete = tcp6_gro_complete,
1da177e4
LT
1985 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1986};
1987
1da177e4
LT
1988static struct inet_protosw tcpv6_protosw = {
1989 .type = SOCK_STREAM,
1990 .protocol = IPPROTO_TCP,
1991 .prot = &tcpv6_prot,
1992 .ops = &inet6_stream_ops,
1da177e4 1993 .no_check = 0,
d83d8461
ACM
1994 .flags = INET_PROTOSW_PERMANENT |
1995 INET_PROTOSW_ICSK,
1da177e4
LT
1996};
1997
2c8c1e72 1998static int __net_init tcpv6_net_init(struct net *net)
93ec926b 1999{
5677242f
DL
2000 return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
2001 SOCK_RAW, IPPROTO_TCP, net);
93ec926b
DL
2002}
2003
2c8c1e72 2004static void __net_exit tcpv6_net_exit(struct net *net)
93ec926b 2005{
5677242f 2006 inet_ctl_sock_destroy(net->ipv6.tcp_sk);
b099ce26
EB
2007}
2008
2c8c1e72 2009static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
b099ce26
EB
2010{
2011 inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6);
93ec926b
DL
2012}
2013
2014static struct pernet_operations tcpv6_net_ops = {
b099ce26
EB
2015 .init = tcpv6_net_init,
2016 .exit = tcpv6_net_exit,
2017 .exit_batch = tcpv6_net_exit_batch,
93ec926b
DL
2018};
2019
7f4e4868 2020int __init tcpv6_init(void)
1da177e4 2021{
7f4e4868
DL
2022 int ret;
2023
2024 ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
2025 if (ret)
2026 goto out;
2027
1da177e4 2028 /* register inet6 protocol */
7f4e4868
DL
2029 ret = inet6_register_protosw(&tcpv6_protosw);
2030 if (ret)
2031 goto out_tcpv6_protocol;
2032
93ec926b 2033 ret = register_pernet_subsys(&tcpv6_net_ops);
7f4e4868
DL
2034 if (ret)
2035 goto out_tcpv6_protosw;
2036out:
2037 return ret;
ae0f7d5f 2038
7f4e4868
DL
2039out_tcpv6_protocol:
2040 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2041out_tcpv6_protosw:
2042 inet6_unregister_protosw(&tcpv6_protosw);
2043 goto out;
2044}
2045
09f7709f 2046void tcpv6_exit(void)
7f4e4868 2047{
93ec926b 2048 unregister_pernet_subsys(&tcpv6_net_ops);
7f4e4868
DL
2049 inet6_unregister_protosw(&tcpv6_protosw);
2050 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
1da177e4 2051}