]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - net/ipv6/tcp_ipv6.c
[PATCH] tcp: Cache inetpeer in timewait socket, and only when necessary.
[mirror_ubuntu-bionic-kernel.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 &&
280 ipv6_addr_equal(&rt->rt6i_dst.addr, &np->daddr)) {
281 struct inet_peer *peer = rt6_get_peer(rt);
282 /*
283 * VJ's idea. We save last timestamp seen from
284 * the destination in peer table, when entering state
285 * TIME-WAIT * and initialize rx_opt.ts_recent from it,
286 * when trying new connection.
287 */
288 if (peer) {
289 inet_peer_refcheck(peer);
290 if ((u32)get_seconds() - peer->tcp_ts_stamp <= TCP_PAWS_MSL) {
291 tp->rx_opt.ts_recent_stamp = peer->tcp_ts_stamp;
292 tp->rx_opt.ts_recent = peer->tcp_ts;
293 }
294 }
295 }
296
d83d8461 297 icsk->icsk_ext_hdr_len = 0;
1da177e4 298 if (np->opt)
d83d8461
ACM
299 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
300 np->opt->opt_nflen);
1da177e4
LT
301
302 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
303
c720c7e8 304 inet->inet_dport = usin->sin6_port;
1da177e4
LT
305
306 tcp_set_state(sk, TCP_SYN_SENT);
d8313f5c 307 err = inet6_hash_connect(&tcp_death_row, sk);
1da177e4
LT
308 if (err)
309 goto late_failure;
310
311 if (!tp->write_seq)
312 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
313 np->daddr.s6_addr32,
c720c7e8
ED
314 inet->inet_sport,
315 inet->inet_dport);
1da177e4
LT
316
317 err = tcp_connect(sk);
318 if (err)
319 goto late_failure;
320
321 return 0;
322
323late_failure:
324 tcp_set_state(sk, TCP_CLOSE);
325 __sk_dst_reset(sk);
326failure:
c720c7e8 327 inet->inet_dport = 0;
1da177e4
LT
328 sk->sk_route_caps = 0;
329 return err;
330}
331
332static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
d5fdd6ba 333 u8 type, u8 code, int offset, __be32 info)
1da177e4 334{
b71d1d42 335 const struct ipv6hdr *hdr = (const struct ipv6hdr*)skb->data;
505cbfc5 336 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
1da177e4
LT
337 struct ipv6_pinfo *np;
338 struct sock *sk;
339 int err;
1ab1457c 340 struct tcp_sock *tp;
1da177e4 341 __u32 seq;
ca12a1a4 342 struct net *net = dev_net(skb->dev);
1da177e4 343
ca12a1a4 344 sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr,
d86e0dac 345 th->dest, &hdr->saddr, th->source, skb->dev->ifindex);
1da177e4
LT
346
347 if (sk == NULL) {
e41b5368
DL
348 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
349 ICMP6_MIB_INERRORS);
1da177e4
LT
350 return;
351 }
352
353 if (sk->sk_state == TCP_TIME_WAIT) {
9469c7b4 354 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
355 return;
356 }
357
358 bh_lock_sock(sk);
359 if (sock_owned_by_user(sk))
de0744af 360 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
1da177e4
LT
361
362 if (sk->sk_state == TCP_CLOSE)
363 goto out;
364
e802af9c
SH
365 if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) {
366 NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
367 goto out;
368 }
369
1da177e4 370 tp = tcp_sk(sk);
1ab1457c 371 seq = ntohl(th->seq);
1da177e4
LT
372 if (sk->sk_state != TCP_LISTEN &&
373 !between(seq, tp->snd_una, tp->snd_nxt)) {
de0744af 374 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
1da177e4
LT
375 goto out;
376 }
377
378 np = inet6_sk(sk);
379
380 if (type == ICMPV6_PKT_TOOBIG) {
68d0c6d3 381 struct dst_entry *dst;
1da177e4
LT
382
383 if (sock_owned_by_user(sk))
384 goto out;
385 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
386 goto out;
387
388 /* icmp should have updated the destination cache entry */
389 dst = __sk_dst_check(sk, np->dst_cookie);
390
391 if (dst == NULL) {
392 struct inet_sock *inet = inet_sk(sk);
4c9483b2 393 struct flowi6 fl6;
1da177e4
LT
394
395 /* BUGGG_FUTURE: Again, it is not clear how
396 to handle rthdr case. Ignore this complexity
397 for now.
398 */
4c9483b2
DM
399 memset(&fl6, 0, sizeof(fl6));
400 fl6.flowi6_proto = IPPROTO_TCP;
4e3fd7a0
AD
401 fl6.daddr = np->daddr;
402 fl6.saddr = np->saddr;
4c9483b2
DM
403 fl6.flowi6_oif = sk->sk_bound_dev_if;
404 fl6.flowi6_mark = sk->sk_mark;
1958b856
DM
405 fl6.fl6_dport = inet->inet_dport;
406 fl6.fl6_sport = inet->inet_sport;
4c9483b2
DM
407 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
408
409 dst = ip6_dst_lookup_flow(sk, &fl6, NULL, false);
68d0c6d3
DM
410 if (IS_ERR(dst)) {
411 sk->sk_err_soft = -PTR_ERR(dst);
1da177e4
LT
412 goto out;
413 }
414
415 } else
416 dst_hold(dst);
417
d83d8461 418 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
1da177e4
LT
419 tcp_sync_mss(sk, dst_mtu(dst));
420 tcp_simple_retransmit(sk);
421 } /* else let the usual retransmit timer handle it */
422 dst_release(dst);
423 goto out;
424 }
425
426 icmpv6_err_convert(type, code, &err);
427
60236fdd 428 /* Might be for an request_sock */
1da177e4 429 switch (sk->sk_state) {
60236fdd 430 struct request_sock *req, **prev;
1da177e4
LT
431 case TCP_LISTEN:
432 if (sock_owned_by_user(sk))
433 goto out;
434
8129765a
ACM
435 req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
436 &hdr->saddr, inet6_iif(skb));
1da177e4
LT
437 if (!req)
438 goto out;
439
440 /* ICMPs are not backlogged, hence we cannot get
441 * an established socket here.
442 */
547b792c 443 WARN_ON(req->sk != NULL);
1da177e4 444
2e6599cb 445 if (seq != tcp_rsk(req)->snt_isn) {
de0744af 446 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
1da177e4
LT
447 goto out;
448 }
449
463c84b9 450 inet_csk_reqsk_queue_drop(sk, req, prev);
1da177e4
LT
451 goto out;
452
453 case TCP_SYN_SENT:
454 case TCP_SYN_RECV: /* Cannot happen.
1ab1457c 455 It can, it SYNs are crossed. --ANK */
1da177e4 456 if (!sock_owned_by_user(sk)) {
1da177e4
LT
457 sk->sk_err = err;
458 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
459
460 tcp_done(sk);
461 } else
462 sk->sk_err_soft = err;
463 goto out;
464 }
465
466 if (!sock_owned_by_user(sk) && np->recverr) {
467 sk->sk_err = err;
468 sk->sk_error_report(sk);
469 } else
470 sk->sk_err_soft = err;
471
472out:
473 bh_unlock_sock(sk);
474 sock_put(sk);
475}
476
477
e6b4d113 478static int tcp_v6_send_synack(struct sock *sk, struct request_sock *req,
fff32699
ED
479 struct request_values *rvp,
480 u16 queue_mapping)
1da177e4 481{
ca304b61 482 struct inet6_request_sock *treq = inet6_rsk(req);
1da177e4
LT
483 struct ipv6_pinfo *np = inet6_sk(sk);
484 struct sk_buff * skb;
485 struct ipv6_txoptions *opt = NULL;
20c59de2 486 struct in6_addr * final_p, final;
4c9483b2 487 struct flowi6 fl6;
fd80eb94 488 struct dst_entry *dst;
68d0c6d3 489 int err;
1da177e4 490
4c9483b2
DM
491 memset(&fl6, 0, sizeof(fl6));
492 fl6.flowi6_proto = IPPROTO_TCP;
4e3fd7a0
AD
493 fl6.daddr = treq->rmt_addr;
494 fl6.saddr = treq->loc_addr;
4c9483b2
DM
495 fl6.flowlabel = 0;
496 fl6.flowi6_oif = treq->iif;
497 fl6.flowi6_mark = sk->sk_mark;
1958b856
DM
498 fl6.fl6_dport = inet_rsk(req)->rmt_port;
499 fl6.fl6_sport = inet_rsk(req)->loc_port;
4c9483b2 500 security_req_classify_flow(req, flowi6_to_flowi(&fl6));
1da177e4 501
fd80eb94 502 opt = np->opt;
4c9483b2 503 final_p = fl6_update_dst(&fl6, opt, &final);
1da177e4 504
4c9483b2 505 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
68d0c6d3
DM
506 if (IS_ERR(dst)) {
507 err = PTR_ERR(dst);
738faca3 508 dst = NULL;
fd80eb94 509 goto done;
68d0c6d3 510 }
e6b4d113 511 skb = tcp_make_synack(sk, dst, req, rvp);
68d0c6d3 512 err = -ENOMEM;
1da177e4 513 if (skb) {
8ad50d96 514 __tcp_v6_send_check(skb, &treq->loc_addr, &treq->rmt_addr);
1da177e4 515
4e3fd7a0 516 fl6.daddr = treq->rmt_addr;
fff32699 517 skb_set_queue_mapping(skb, queue_mapping);
b903d324 518 err = ip6_xmit(sk, skb, &fl6, opt, np->tclass);
b9df3cb8 519 err = net_xmit_eval(err);
1da177e4
LT
520 }
521
522done:
1ab1457c 523 if (opt && opt != np->opt)
1da177e4
LT
524 sock_kfree_s(sk, opt, opt->tot_len);
525 return err;
526}
527
72659ecc
OP
528static int tcp_v6_rtx_synack(struct sock *sk, struct request_sock *req,
529 struct request_values *rvp)
530{
531 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS);
fff32699 532 return tcp_v6_send_synack(sk, req, rvp, 0);
72659ecc
OP
533}
534
60236fdd 535static void tcp_v6_reqsk_destructor(struct request_sock *req)
1da177e4 536{
800d55f1 537 kfree_skb(inet6_rsk(req)->pktopts);
1da177e4
LT
538}
539
cfb6eeb4
YH
540#ifdef CONFIG_TCP_MD5SIG
541static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
b71d1d42 542 const struct in6_addr *addr)
cfb6eeb4 543{
a915da9b 544 return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
cfb6eeb4
YH
545}
546
547static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
548 struct sock *addr_sk)
549{
550 return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr);
551}
552
553static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
554 struct request_sock *req)
555{
556 return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr);
557}
558
cfb6eeb4
YH
559static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval,
560 int optlen)
561{
562 struct tcp_md5sig cmd;
563 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
cfb6eeb4
YH
564
565 if (optlen < sizeof(cmd))
566 return -EINVAL;
567
568 if (copy_from_user(&cmd, optval, sizeof(cmd)))
569 return -EFAULT;
570
571 if (sin6->sin6_family != AF_INET6)
572 return -EINVAL;
573
574 if (!cmd.tcpm_keylen) {
e773e4fa 575 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
a915da9b
ED
576 return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
577 AF_INET);
578 return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
579 AF_INET6);
cfb6eeb4
YH
580 }
581
582 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
583 return -EINVAL;
584
a915da9b
ED
585 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
586 return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
587 AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
cfb6eeb4 588
a915da9b
ED
589 return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
590 AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
cfb6eeb4
YH
591}
592
49a72dfb 593static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
b71d1d42
ED
594 const struct in6_addr *daddr,
595 const struct in6_addr *saddr, int nbytes)
cfb6eeb4 596{
cfb6eeb4 597 struct tcp6_pseudohdr *bp;
49a72dfb 598 struct scatterlist sg;
8d26d76d 599
cfb6eeb4 600 bp = &hp->md5_blk.ip6;
cfb6eeb4 601 /* 1. TCP pseudo-header (RFC2460) */
4e3fd7a0
AD
602 bp->saddr = *saddr;
603 bp->daddr = *daddr;
49a72dfb 604 bp->protocol = cpu_to_be32(IPPROTO_TCP);
00b1304c 605 bp->len = cpu_to_be32(nbytes);
cfb6eeb4 606
49a72dfb
AL
607 sg_init_one(&sg, bp, sizeof(*bp));
608 return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
609}
c7da57a1 610
49a72dfb 611static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
b71d1d42 612 const struct in6_addr *daddr, struct in6_addr *saddr,
318cf7aa 613 const struct tcphdr *th)
49a72dfb
AL
614{
615 struct tcp_md5sig_pool *hp;
616 struct hash_desc *desc;
617
618 hp = tcp_get_md5sig_pool();
619 if (!hp)
620 goto clear_hash_noput;
621 desc = &hp->md5_desc;
622
623 if (crypto_hash_init(desc))
624 goto clear_hash;
625 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2))
626 goto clear_hash;
627 if (tcp_md5_hash_header(hp, th))
628 goto clear_hash;
629 if (tcp_md5_hash_key(hp, key))
630 goto clear_hash;
631 if (crypto_hash_final(desc, md5_hash))
cfb6eeb4 632 goto clear_hash;
cfb6eeb4 633
cfb6eeb4 634 tcp_put_md5sig_pool();
cfb6eeb4 635 return 0;
49a72dfb 636
cfb6eeb4
YH
637clear_hash:
638 tcp_put_md5sig_pool();
639clear_hash_noput:
640 memset(md5_hash, 0, 16);
49a72dfb 641 return 1;
cfb6eeb4
YH
642}
643
49a72dfb 644static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
318cf7aa
ED
645 const struct sock *sk,
646 const struct request_sock *req,
647 const struct sk_buff *skb)
cfb6eeb4 648{
b71d1d42 649 const struct in6_addr *saddr, *daddr;
49a72dfb
AL
650 struct tcp_md5sig_pool *hp;
651 struct hash_desc *desc;
318cf7aa 652 const struct tcphdr *th = tcp_hdr(skb);
cfb6eeb4
YH
653
654 if (sk) {
655 saddr = &inet6_sk(sk)->saddr;
656 daddr = &inet6_sk(sk)->daddr;
49a72dfb 657 } else if (req) {
cfb6eeb4
YH
658 saddr = &inet6_rsk(req)->loc_addr;
659 daddr = &inet6_rsk(req)->rmt_addr;
49a72dfb 660 } else {
b71d1d42 661 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
49a72dfb
AL
662 saddr = &ip6h->saddr;
663 daddr = &ip6h->daddr;
cfb6eeb4 664 }
49a72dfb
AL
665
666 hp = tcp_get_md5sig_pool();
667 if (!hp)
668 goto clear_hash_noput;
669 desc = &hp->md5_desc;
670
671 if (crypto_hash_init(desc))
672 goto clear_hash;
673
674 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len))
675 goto clear_hash;
676 if (tcp_md5_hash_header(hp, th))
677 goto clear_hash;
678 if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
679 goto clear_hash;
680 if (tcp_md5_hash_key(hp, key))
681 goto clear_hash;
682 if (crypto_hash_final(desc, md5_hash))
683 goto clear_hash;
684
685 tcp_put_md5sig_pool();
686 return 0;
687
688clear_hash:
689 tcp_put_md5sig_pool();
690clear_hash_noput:
691 memset(md5_hash, 0, 16);
692 return 1;
cfb6eeb4
YH
693}
694
318cf7aa 695static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
cfb6eeb4 696{
cf533ea5 697 const __u8 *hash_location = NULL;
cfb6eeb4 698 struct tcp_md5sig_key *hash_expected;
b71d1d42 699 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
318cf7aa 700 const struct tcphdr *th = tcp_hdr(skb);
cfb6eeb4 701 int genhash;
cfb6eeb4
YH
702 u8 newhash[16];
703
704 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
7d5d5525 705 hash_location = tcp_parse_md5sig_option(th);
cfb6eeb4 706
785957d3
DM
707 /* We've parsed the options - do we have a hash? */
708 if (!hash_expected && !hash_location)
709 return 0;
710
711 if (hash_expected && !hash_location) {
712 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
cfb6eeb4
YH
713 return 1;
714 }
715
785957d3
DM
716 if (!hash_expected && hash_location) {
717 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
cfb6eeb4
YH
718 return 1;
719 }
720
721 /* check the signature */
49a72dfb
AL
722 genhash = tcp_v6_md5_hash_skb(newhash,
723 hash_expected,
724 NULL, NULL, skb);
725
cfb6eeb4 726 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
e87cc472
JP
727 net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n",
728 genhash ? "failed" : "mismatch",
729 &ip6h->saddr, ntohs(th->source),
730 &ip6h->daddr, ntohs(th->dest));
cfb6eeb4
YH
731 return 1;
732 }
733 return 0;
734}
735#endif
736
c6aefafb 737struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
1da177e4 738 .family = AF_INET6,
2e6599cb 739 .obj_size = sizeof(struct tcp6_request_sock),
72659ecc 740 .rtx_syn_ack = tcp_v6_rtx_synack,
60236fdd
ACM
741 .send_ack = tcp_v6_reqsk_send_ack,
742 .destructor = tcp_v6_reqsk_destructor,
72659ecc
OP
743 .send_reset = tcp_v6_send_reset,
744 .syn_ack_timeout = tcp_syn_ack_timeout,
1da177e4
LT
745};
746
cfb6eeb4 747#ifdef CONFIG_TCP_MD5SIG
b2e4b3de 748static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
cfb6eeb4 749 .md5_lookup = tcp_v6_reqsk_md5_lookup,
e3afe7b7 750 .calc_md5_hash = tcp_v6_md5_hash_skb,
cfb6eeb4 751};
b6332e6c 752#endif
cfb6eeb4 753
8ad50d96 754static void __tcp_v6_send_check(struct sk_buff *skb,
b71d1d42 755 const struct in6_addr *saddr, const struct in6_addr *daddr)
1da177e4 756{
aa8223c7 757 struct tcphdr *th = tcp_hdr(skb);
1da177e4 758
84fa7933 759 if (skb->ip_summed == CHECKSUM_PARTIAL) {
8ad50d96 760 th->check = ~tcp_v6_check(skb->len, saddr, daddr, 0);
663ead3b 761 skb->csum_start = skb_transport_header(skb) - skb->head;
ff1dcadb 762 skb->csum_offset = offsetof(struct tcphdr, check);
1da177e4 763 } else {
8ad50d96
HX
764 th->check = tcp_v6_check(skb->len, saddr, daddr,
765 csum_partial(th, th->doff << 2,
766 skb->csum));
1da177e4
LT
767 }
768}
769
bb296246 770static void tcp_v6_send_check(struct sock *sk, struct sk_buff *skb)
8ad50d96
HX
771{
772 struct ipv6_pinfo *np = inet6_sk(sk);
773
774 __tcp_v6_send_check(skb, &np->saddr, &np->daddr);
775}
776
a430a43d
HX
777static int tcp_v6_gso_send_check(struct sk_buff *skb)
778{
b71d1d42 779 const struct ipv6hdr *ipv6h;
a430a43d
HX
780 struct tcphdr *th;
781
782 if (!pskb_may_pull(skb, sizeof(*th)))
783 return -EINVAL;
784
0660e03f 785 ipv6h = ipv6_hdr(skb);
aa8223c7 786 th = tcp_hdr(skb);
a430a43d
HX
787
788 th->check = 0;
84fa7933 789 skb->ip_summed = CHECKSUM_PARTIAL;
8ad50d96 790 __tcp_v6_send_check(skb, &ipv6h->saddr, &ipv6h->daddr);
a430a43d
HX
791 return 0;
792}
1da177e4 793
36990673
HX
794static struct sk_buff **tcp6_gro_receive(struct sk_buff **head,
795 struct sk_buff *skb)
684f2176 796{
b71d1d42 797 const struct ipv6hdr *iph = skb_gro_network_header(skb);
684f2176
HX
798
799 switch (skb->ip_summed) {
800 case CHECKSUM_COMPLETE:
86911732 801 if (!tcp_v6_check(skb_gro_len(skb), &iph->saddr, &iph->daddr,
684f2176
HX
802 skb->csum)) {
803 skb->ip_summed = CHECKSUM_UNNECESSARY;
804 break;
805 }
806
807 /* fall through */
808 case CHECKSUM_NONE:
809 NAPI_GRO_CB(skb)->flush = 1;
810 return NULL;
811 }
812
813 return tcp_gro_receive(head, skb);
814}
684f2176 815
36990673 816static int tcp6_gro_complete(struct sk_buff *skb)
684f2176 817{
b71d1d42 818 const struct ipv6hdr *iph = ipv6_hdr(skb);
684f2176
HX
819 struct tcphdr *th = tcp_hdr(skb);
820
821 th->check = ~tcp_v6_check(skb->len - skb_transport_offset(skb),
822 &iph->saddr, &iph->daddr, 0);
823 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
824
825 return tcp_gro_complete(skb);
826}
684f2176 827
626e264d 828static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win,
b903d324 829 u32 ts, struct tcp_md5sig_key *key, int rst, u8 tclass)
1da177e4 830{
cf533ea5
ED
831 const struct tcphdr *th = tcp_hdr(skb);
832 struct tcphdr *t1;
1da177e4 833 struct sk_buff *buff;
4c9483b2 834 struct flowi6 fl6;
adf30907 835 struct net *net = dev_net(skb_dst(skb)->dev);
e5047992 836 struct sock *ctl_sk = net->ipv6.tcp_sk;
77c676da 837 unsigned int tot_len = sizeof(struct tcphdr);
adf30907 838 struct dst_entry *dst;
81ada62d 839 __be32 *topt;
1da177e4 840
626e264d
IJ
841 if (ts)
842 tot_len += TCPOLEN_TSTAMP_ALIGNED;
cfb6eeb4 843#ifdef CONFIG_TCP_MD5SIG
cfb6eeb4
YH
844 if (key)
845 tot_len += TCPOLEN_MD5SIG_ALIGNED;
846#endif
847
cfb6eeb4 848 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1da177e4 849 GFP_ATOMIC);
1ab1457c
YH
850 if (buff == NULL)
851 return;
1da177e4 852
cfb6eeb4 853 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1da177e4 854
cfb6eeb4 855 t1 = (struct tcphdr *) skb_push(buff, tot_len);
6651ffc8 856 skb_reset_transport_header(buff);
1da177e4
LT
857
858 /* Swap the send and the receive. */
859 memset(t1, 0, sizeof(*t1));
860 t1->dest = th->source;
861 t1->source = th->dest;
cfb6eeb4 862 t1->doff = tot_len / 4;
626e264d
IJ
863 t1->seq = htonl(seq);
864 t1->ack_seq = htonl(ack);
865 t1->ack = !rst || !th->ack;
866 t1->rst = rst;
867 t1->window = htons(win);
1da177e4 868
81ada62d
IJ
869 topt = (__be32 *)(t1 + 1);
870
626e264d
IJ
871 if (ts) {
872 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
873 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
874 *topt++ = htonl(tcp_time_stamp);
875 *topt++ = htonl(ts);
876 }
877
cfb6eeb4
YH
878#ifdef CONFIG_TCP_MD5SIG
879 if (key) {
81ada62d
IJ
880 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
881 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
882 tcp_v6_md5_hash_hdr((__u8 *)topt, key,
78e645cb
IJ
883 &ipv6_hdr(skb)->saddr,
884 &ipv6_hdr(skb)->daddr, t1);
cfb6eeb4
YH
885 }
886#endif
887
4c9483b2 888 memset(&fl6, 0, sizeof(fl6));
4e3fd7a0
AD
889 fl6.daddr = ipv6_hdr(skb)->saddr;
890 fl6.saddr = ipv6_hdr(skb)->daddr;
1da177e4 891
e5700aff
DM
892 buff->ip_summed = CHECKSUM_PARTIAL;
893 buff->csum = 0;
894
4c9483b2 895 __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
1da177e4 896
4c9483b2
DM
897 fl6.flowi6_proto = IPPROTO_TCP;
898 fl6.flowi6_oif = inet6_iif(skb);
1958b856
DM
899 fl6.fl6_dport = t1->dest;
900 fl6.fl6_sport = t1->source;
4c9483b2 901 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
1da177e4 902
c20121ae
DL
903 /* Pass a socket to ip6_dst_lookup either it is for RST
904 * Underlying function will use this to retrieve the network
905 * namespace
906 */
4c9483b2 907 dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL, false);
68d0c6d3
DM
908 if (!IS_ERR(dst)) {
909 skb_dst_set(buff, dst);
b903d324 910 ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass);
68d0c6d3
DM
911 TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
912 if (rst)
913 TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS);
914 return;
1da177e4
LT
915 }
916
917 kfree_skb(buff);
918}
919
626e264d 920static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
1da177e4 921{
cf533ea5 922 const struct tcphdr *th = tcp_hdr(skb);
626e264d 923 u32 seq = 0, ack_seq = 0;
fa3e5b4e 924 struct tcp_md5sig_key *key = NULL;
658ddaaf
SL
925#ifdef CONFIG_TCP_MD5SIG
926 const __u8 *hash_location = NULL;
927 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
928 unsigned char newhash[16];
929 int genhash;
930 struct sock *sk1 = NULL;
931#endif
1da177e4 932
626e264d 933 if (th->rst)
1da177e4
LT
934 return;
935
626e264d
IJ
936 if (!ipv6_unicast_destination(skb))
937 return;
1da177e4 938
cfb6eeb4 939#ifdef CONFIG_TCP_MD5SIG
658ddaaf
SL
940 hash_location = tcp_parse_md5sig_option(th);
941 if (!sk && hash_location) {
942 /*
943 * active side is lost. Try to find listening socket through
944 * source port, and then find md5 key through listening socket.
945 * we are not loose security here:
946 * Incoming packet is checked with md5 hash with finding key,
947 * no RST generated if md5 hash doesn't match.
948 */
949 sk1 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev),
950 &tcp_hashinfo, &ipv6h->daddr,
951 ntohs(th->source), inet6_iif(skb));
952 if (!sk1)
953 return;
954
955 rcu_read_lock();
956 key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr);
957 if (!key)
958 goto release_sk1;
959
960 genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, NULL, skb);
961 if (genhash || memcmp(hash_location, newhash, 16) != 0)
962 goto release_sk1;
963 } else {
964 key = sk ? tcp_v6_md5_do_lookup(sk, &ipv6h->saddr) : NULL;
965 }
cfb6eeb4
YH
966#endif
967
626e264d
IJ
968 if (th->ack)
969 seq = ntohl(th->ack_seq);
970 else
971 ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len -
972 (th->doff << 2);
1da177e4 973
b903d324 974 tcp_v6_send_response(skb, seq, ack_seq, 0, 0, key, 1, 0);
658ddaaf
SL
975
976#ifdef CONFIG_TCP_MD5SIG
977release_sk1:
978 if (sk1) {
979 rcu_read_unlock();
980 sock_put(sk1);
981 }
982#endif
626e264d 983}
1da177e4 984
626e264d 985static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts,
b903d324 986 struct tcp_md5sig_key *key, u8 tclass)
626e264d 987{
b903d324 988 tcp_v6_send_response(skb, seq, ack, win, ts, key, 0, tclass);
1da177e4
LT
989}
990
991static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
992{
8feaf0c0 993 struct inet_timewait_sock *tw = inet_twsk(sk);
cfb6eeb4 994 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
1da177e4 995
9501f972 996 tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
8feaf0c0 997 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
b903d324
ED
998 tcptw->tw_ts_recent, tcp_twsk_md5_key(tcptw),
999 tw->tw_tclass);
1da177e4 1000
8feaf0c0 1001 inet_twsk_put(tw);
1da177e4
LT
1002}
1003
6edafaaf
GJ
1004static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
1005 struct request_sock *req)
1da177e4 1006{
9501f972 1007 tcp_v6_send_ack(skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, req->ts_recent,
b903d324 1008 tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr), 0);
1da177e4
LT
1009}
1010
1011
1012static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
1013{
60236fdd 1014 struct request_sock *req, **prev;
aa8223c7 1015 const struct tcphdr *th = tcp_hdr(skb);
1da177e4
LT
1016 struct sock *nsk;
1017
1018 /* Find possible connection requests. */
8129765a 1019 req = inet6_csk_search_req(sk, &prev, th->source,
0660e03f
ACM
1020 &ipv6_hdr(skb)->saddr,
1021 &ipv6_hdr(skb)->daddr, inet6_iif(skb));
1da177e4
LT
1022 if (req)
1023 return tcp_check_req(sk, skb, req, prev);
1024
3b1e0a65 1025 nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
d86e0dac
PE
1026 &ipv6_hdr(skb)->saddr, th->source,
1027 &ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb));
1da177e4
LT
1028
1029 if (nsk) {
1030 if (nsk->sk_state != TCP_TIME_WAIT) {
1031 bh_lock_sock(nsk);
1032 return nsk;
1033 }
9469c7b4 1034 inet_twsk_put(inet_twsk(nsk));
1da177e4
LT
1035 return NULL;
1036 }
1037
c6aefafb 1038#ifdef CONFIG_SYN_COOKIES
af9b4738 1039 if (!th->syn)
c6aefafb 1040 sk = cookie_v6_check(sk, skb);
1da177e4
LT
1041#endif
1042 return sk;
1043}
1044
1da177e4
LT
1045/* FIXME: this is substantially similar to the ipv4 code.
1046 * Can some kind of merge be done? -- erics
1047 */
1048static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
1049{
4957faad 1050 struct tcp_extend_values tmp_ext;
e6b4d113 1051 struct tcp_options_received tmp_opt;
cf533ea5 1052 const u8 *hash_location;
e6b4d113 1053 struct request_sock *req;
ca304b61 1054 struct inet6_request_sock *treq;
1da177e4 1055 struct ipv6_pinfo *np = inet6_sk(sk);
1da177e4 1056 struct tcp_sock *tp = tcp_sk(sk);
e6b4d113 1057 __u32 isn = TCP_SKB_CB(skb)->when;
493f377d 1058 struct dst_entry *dst = NULL;
a2a385d6 1059 bool want_cookie = false;
1da177e4
LT
1060
1061 if (skb->protocol == htons(ETH_P_IP))
1062 return tcp_v4_conn_request(sk, skb);
1063
1064 if (!ipv6_unicast_destination(skb))
1ab1457c 1065 goto drop;
1da177e4 1066
463c84b9 1067 if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
946cedcc
ED
1068 want_cookie = tcp_syn_flood_action(sk, skb, "TCPv6");
1069 if (!want_cookie)
1070 goto drop;
1da177e4
LT
1071 }
1072
463c84b9 1073 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
1da177e4
LT
1074 goto drop;
1075
ca304b61 1076 req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
1da177e4
LT
1077 if (req == NULL)
1078 goto drop;
1079
cfb6eeb4
YH
1080#ifdef CONFIG_TCP_MD5SIG
1081 tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
1082#endif
1083
1da177e4
LT
1084 tcp_clear_options(&tmp_opt);
1085 tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
1086 tmp_opt.user_mss = tp->rx_opt.user_mss;
bb5b7c11 1087 tcp_parse_options(skb, &tmp_opt, &hash_location, 0);
4957faad
WAS
1088
1089 if (tmp_opt.cookie_plus > 0 &&
1090 tmp_opt.saw_tstamp &&
1091 !tp->rx_opt.cookie_out_never &&
1092 (sysctl_tcp_cookie_size > 0 ||
1093 (tp->cookie_values != NULL &&
1094 tp->cookie_values->cookie_desired > 0))) {
1095 u8 *c;
1096 u32 *d;
1097 u32 *mess = &tmp_ext.cookie_bakery[COOKIE_DIGEST_WORDS];
1098 int l = tmp_opt.cookie_plus - TCPOLEN_COOKIE_BASE;
1099
1100 if (tcp_cookie_generator(&tmp_ext.cookie_bakery[0]) != 0)
1101 goto drop_and_free;
1102
1103 /* Secret recipe starts with IP addresses */
0eae88f3 1104 d = (__force u32 *)&ipv6_hdr(skb)->daddr.s6_addr32[0];
4957faad
WAS
1105 *mess++ ^= *d++;
1106 *mess++ ^= *d++;
1107 *mess++ ^= *d++;
1108 *mess++ ^= *d++;
0eae88f3 1109 d = (__force u32 *)&ipv6_hdr(skb)->saddr.s6_addr32[0];
4957faad
WAS
1110 *mess++ ^= *d++;
1111 *mess++ ^= *d++;
1112 *mess++ ^= *d++;
1113 *mess++ ^= *d++;
1114
1115 /* plus variable length Initiator Cookie */
1116 c = (u8 *)mess;
1117 while (l-- > 0)
1118 *c++ ^= *hash_location++;
1da177e4 1119
a2a385d6 1120 want_cookie = false; /* not our kind of cookie */
4957faad
WAS
1121 tmp_ext.cookie_out_never = 0; /* false */
1122 tmp_ext.cookie_plus = tmp_opt.cookie_plus;
1123 } else if (!tp->rx_opt.cookie_in_always) {
1124 /* redundant indications, but ensure initialization. */
1125 tmp_ext.cookie_out_never = 1; /* true */
1126 tmp_ext.cookie_plus = 0;
1127 } else {
1128 goto drop_and_free;
1129 }
1130 tmp_ext.cookie_in_always = tp->rx_opt.cookie_in_always;
1da177e4 1131
4dfc2817 1132 if (want_cookie && !tmp_opt.saw_tstamp)
c6aefafb 1133 tcp_clear_options(&tmp_opt);
c6aefafb 1134
1da177e4
LT
1135 tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
1136 tcp_openreq_init(req, &tmp_opt, skb);
1137
ca304b61 1138 treq = inet6_rsk(req);
4e3fd7a0
AD
1139 treq->rmt_addr = ipv6_hdr(skb)->saddr;
1140 treq->loc_addr = ipv6_hdr(skb)->daddr;
172d69e6 1141 if (!want_cookie || tmp_opt.tstamp_ok)
bd14b1b2 1142 TCP_ECN_create_request(req, skb);
c6aefafb 1143
4d0fe50c
ED
1144 treq->iif = sk->sk_bound_dev_if;
1145
1146 /* So that link locals have meaning */
1147 if (!sk->sk_bound_dev_if &&
1148 ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL)
1149 treq->iif = inet6_iif(skb);
1150
2bbdf389 1151 if (!isn) {
493f377d
DM
1152 struct inet_peer *peer = NULL;
1153
c6aefafb
GG
1154 if (ipv6_opt_accepted(sk, skb) ||
1155 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
1156 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
1157 atomic_inc(&skb->users);
1158 treq->pktopts = skb;
1159 }
493f377d
DM
1160
1161 if (want_cookie) {
2bbdf389
FW
1162 isn = cookie_v6_init_sequence(sk, skb, &req->mss);
1163 req->cookie_ts = tmp_opt.tstamp_ok;
493f377d
DM
1164 goto have_isn;
1165 }
1166
1167 /* VJ's idea. We save last timestamp seen
1168 * from the destination in peer table, when entering
1169 * state TIME-WAIT, and check against it before
1170 * accepting new connection request.
1171 *
1172 * If "isn" is not zero, this request hit alive
1173 * timewait bucket, so that all the necessary checks
1174 * are made in the function processing timewait state.
1175 */
1176 if (tmp_opt.saw_tstamp &&
1177 tcp_death_row.sysctl_tw_recycle &&
1178 (dst = inet6_csk_route_req(sk, req)) != NULL &&
1179 (peer = rt6_get_peer((struct rt6_info *)dst)) != NULL &&
7a71ed89 1180 ipv6_addr_equal((struct in6_addr *)peer->daddr.addr.a6,
493f377d
DM
1181 &treq->rmt_addr)) {
1182 inet_peer_refcheck(peer);
1183 if ((u32)get_seconds() - peer->tcp_ts_stamp < TCP_PAWS_MSL &&
1184 (s32)(peer->tcp_ts - req->ts_recent) >
1185 TCP_PAWS_WINDOW) {
1186 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED);
1187 goto drop_and_release;
1188 }
1189 }
1190 /* Kill the following clause, if you dislike this way. */
1191 else if (!sysctl_tcp_syncookies &&
1192 (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) <
1193 (sysctl_max_syn_backlog >> 2)) &&
1194 (!peer || !peer->tcp_ts_stamp) &&
1195 (!dst || !dst_metric(dst, RTAX_RTT))) {
1196 /* Without syncookies last quarter of
1197 * backlog is filled with destinations,
1198 * proven to be alive.
1199 * It means that we continue to communicate
1200 * to destinations, already remembered
1201 * to the moment of synflood.
1202 */
1203 LIMIT_NETDEBUG(KERN_DEBUG "TCP: drop open request from %pI6/%u\n",
1204 &treq->rmt_addr, ntohs(tcp_hdr(skb)->source));
1205 goto drop_and_release;
2bbdf389 1206 }
493f377d
DM
1207
1208 isn = tcp_v6_init_sequence(skb);
c6aefafb 1209 }
493f377d 1210have_isn:
2e6599cb 1211 tcp_rsk(req)->snt_isn = isn;
9ad7c049 1212 tcp_rsk(req)->snt_synack = tcp_time_stamp;
1da177e4 1213
4237c75c
VY
1214 security_inet_conn_request(sk, skb, req);
1215
4957faad 1216 if (tcp_v6_send_synack(sk, req,
fff32699
ED
1217 (struct request_values *)&tmp_ext,
1218 skb_get_queue_mapping(skb)) ||
4957faad 1219 want_cookie)
e6b4d113 1220 goto drop_and_free;
1da177e4 1221
e6b4d113
WAS
1222 inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
1223 return 0;
1da177e4 1224
493f377d
DM
1225drop_and_release:
1226 dst_release(dst);
e6b4d113
WAS
1227drop_and_free:
1228 reqsk_free(req);
1da177e4 1229drop:
1da177e4
LT
1230 return 0; /* don't send reset */
1231}
1232
1233static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
60236fdd 1234 struct request_sock *req,
1da177e4
LT
1235 struct dst_entry *dst)
1236{
78d15e82 1237 struct inet6_request_sock *treq;
1da177e4
LT
1238 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
1239 struct tcp6_sock *newtcp6sk;
1240 struct inet_sock *newinet;
1241 struct tcp_sock *newtp;
1242 struct sock *newsk;
1243 struct ipv6_txoptions *opt;
cfb6eeb4
YH
1244#ifdef CONFIG_TCP_MD5SIG
1245 struct tcp_md5sig_key *key;
1246#endif
1da177e4
LT
1247
1248 if (skb->protocol == htons(ETH_P_IP)) {
1249 /*
1250 * v6 mapped
1251 */
1252
1253 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1254
1ab1457c 1255 if (newsk == NULL)
1da177e4
LT
1256 return NULL;
1257
1258 newtcp6sk = (struct tcp6_sock *)newsk;
1259 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1260
1261 newinet = inet_sk(newsk);
1262 newnp = inet6_sk(newsk);
1263 newtp = tcp_sk(newsk);
1264
1265 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1266
c720c7e8 1267 ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr);
1da177e4 1268
c720c7e8 1269 ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr);
1da177e4 1270
4e3fd7a0 1271 newnp->rcv_saddr = newnp->saddr;
1da177e4 1272
8292a17a 1273 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1da177e4 1274 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
cfb6eeb4
YH
1275#ifdef CONFIG_TCP_MD5SIG
1276 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1277#endif
1278
676a1184
YZ
1279 newnp->ipv6_ac_list = NULL;
1280 newnp->ipv6_fl_list = NULL;
1da177e4
LT
1281 newnp->pktoptions = NULL;
1282 newnp->opt = NULL;
505cbfc5 1283 newnp->mcast_oif = inet6_iif(skb);
0660e03f 1284 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
4c507d28 1285 newnp->rcv_tclass = ipv6_tclass(ipv6_hdr(skb));
1da177e4 1286
e6848976
ACM
1287 /*
1288 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1289 * here, tcp_create_openreq_child now does this for us, see the comment in
1290 * that function for the gory details. -acme
1da177e4 1291 */
1da177e4
LT
1292
1293 /* It is tricky place. Until this moment IPv4 tcp
8292a17a 1294 worked with IPv6 icsk.icsk_af_ops.
1da177e4
LT
1295 Sync it now.
1296 */
d83d8461 1297 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1da177e4
LT
1298
1299 return newsk;
1300 }
1301
78d15e82 1302 treq = inet6_rsk(req);
1da177e4
LT
1303 opt = np->opt;
1304
1305 if (sk_acceptq_is_full(sk))
1306 goto out_overflow;
1307
493f377d
DM
1308 if (!dst) {
1309 dst = inet6_csk_route_req(sk, req);
1310 if (!dst)
1da177e4 1311 goto out;
1ab1457c 1312 }
1da177e4
LT
1313
1314 newsk = tcp_create_openreq_child(sk, req, skb);
1315 if (newsk == NULL)
093d2823 1316 goto out_nonewsk;
1da177e4 1317
e6848976
ACM
1318 /*
1319 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1320 * count here, tcp_create_openreq_child now does this for us, see the
1321 * comment in that function for the gory details. -acme
1322 */
1da177e4 1323
59eed279 1324 newsk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 1325 __ip6_dst_store(newsk, dst, NULL, NULL);
1da177e4
LT
1326
1327 newtcp6sk = (struct tcp6_sock *)newsk;
1328 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1329
1330 newtp = tcp_sk(newsk);
1331 newinet = inet_sk(newsk);
1332 newnp = inet6_sk(newsk);
1333
1334 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1335
4e3fd7a0
AD
1336 newnp->daddr = treq->rmt_addr;
1337 newnp->saddr = treq->loc_addr;
1338 newnp->rcv_saddr = treq->loc_addr;
2e6599cb 1339 newsk->sk_bound_dev_if = treq->iif;
1da177e4 1340
1ab1457c 1341 /* Now IPv6 options...
1da177e4
LT
1342
1343 First: no IPv4 options.
1344 */
f6d8bd05 1345 newinet->inet_opt = NULL;
676a1184 1346 newnp->ipv6_ac_list = NULL;
d35690be 1347 newnp->ipv6_fl_list = NULL;
1da177e4
LT
1348
1349 /* Clone RX bits */
1350 newnp->rxopt.all = np->rxopt.all;
1351
1352 /* Clone pktoptions received with SYN */
1353 newnp->pktoptions = NULL;
2e6599cb
ACM
1354 if (treq->pktopts != NULL) {
1355 newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC);
ab185d7b 1356 consume_skb(treq->pktopts);
2e6599cb 1357 treq->pktopts = NULL;
1da177e4
LT
1358 if (newnp->pktoptions)
1359 skb_set_owner_r(newnp->pktoptions, newsk);
1360 }
1361 newnp->opt = NULL;
505cbfc5 1362 newnp->mcast_oif = inet6_iif(skb);
0660e03f 1363 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
4c507d28 1364 newnp->rcv_tclass = ipv6_tclass(ipv6_hdr(skb));
1da177e4
LT
1365
1366 /* Clone native IPv6 options from listening socket (if any)
1367
1368 Yes, keeping reference count would be much more clever,
1369 but we make one more one thing there: reattach optmem
1370 to newsk.
1371 */
1372 if (opt) {
1373 newnp->opt = ipv6_dup_options(newsk, opt);
1374 if (opt != np->opt)
1375 sock_kfree_s(sk, opt, opt->tot_len);
1376 }
1377
d83d8461 1378 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1da177e4 1379 if (newnp->opt)
d83d8461
ACM
1380 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1381 newnp->opt->opt_flen);
1da177e4 1382
5d424d5a 1383 tcp_mtup_init(newsk);
1da177e4 1384 tcp_sync_mss(newsk, dst_mtu(dst));
0dbaee3b 1385 newtp->advmss = dst_metric_advmss(dst);
d135c522
NC
1386 if (tcp_sk(sk)->rx_opt.user_mss &&
1387 tcp_sk(sk)->rx_opt.user_mss < newtp->advmss)
1388 newtp->advmss = tcp_sk(sk)->rx_opt.user_mss;
1389
1da177e4 1390 tcp_initialize_rcv_mss(newsk);
9ad7c049
JC
1391 if (tcp_rsk(req)->snt_synack)
1392 tcp_valid_rtt_meas(newsk,
1393 tcp_time_stamp - tcp_rsk(req)->snt_synack);
1394 newtp->total_retrans = req->retrans;
1da177e4 1395
c720c7e8
ED
1396 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
1397 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
1da177e4 1398
cfb6eeb4
YH
1399#ifdef CONFIG_TCP_MD5SIG
1400 /* Copy over the MD5 key from the original socket */
1401 if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) {
1402 /* We're using one, so create a matching key
1403 * on the newsk structure. If we fail to get
1404 * memory, then we end up not copying the key
1405 * across. Shucks.
1406 */
a915da9b
ED
1407 tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newnp->daddr,
1408 AF_INET6, key->key, key->keylen, GFP_ATOMIC);
cfb6eeb4
YH
1409 }
1410#endif
1411
093d2823
BS
1412 if (__inet_inherit_port(sk, newsk) < 0) {
1413 sock_put(newsk);
1414 goto out;
1415 }
9327f705 1416 __inet6_hash(newsk, NULL);
1da177e4
LT
1417
1418 return newsk;
1419
1420out_overflow:
de0744af 1421 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
093d2823 1422out_nonewsk:
1da177e4
LT
1423 if (opt && opt != np->opt)
1424 sock_kfree_s(sk, opt, opt->tot_len);
1425 dst_release(dst);
093d2823
BS
1426out:
1427 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
1da177e4
LT
1428 return NULL;
1429}
1430
b51655b9 1431static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
1da177e4 1432{
84fa7933 1433 if (skb->ip_summed == CHECKSUM_COMPLETE) {
684f2176 1434 if (!tcp_v6_check(skb->len, &ipv6_hdr(skb)->saddr,
0660e03f 1435 &ipv6_hdr(skb)->daddr, skb->csum)) {
fb286bb2 1436 skb->ip_summed = CHECKSUM_UNNECESSARY;
1da177e4 1437 return 0;
fb286bb2 1438 }
1da177e4 1439 }
fb286bb2 1440
684f2176 1441 skb->csum = ~csum_unfold(tcp_v6_check(skb->len,
0660e03f
ACM
1442 &ipv6_hdr(skb)->saddr,
1443 &ipv6_hdr(skb)->daddr, 0));
fb286bb2 1444
1da177e4 1445 if (skb->len <= 76) {
fb286bb2 1446 return __skb_checksum_complete(skb);
1da177e4
LT
1447 }
1448 return 0;
1449}
1450
1451/* The socket must have it's spinlock held when we get
1452 * here.
1453 *
1454 * We have a potential double-lock case here, so even when
1455 * doing backlog processing we use the BH locking scheme.
1456 * This is because we cannot sleep with the original spinlock
1457 * held.
1458 */
1459static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1460{
1461 struct ipv6_pinfo *np = inet6_sk(sk);
1462 struct tcp_sock *tp;
1463 struct sk_buff *opt_skb = NULL;
1464
1465 /* Imagine: socket is IPv6. IPv4 packet arrives,
1466 goes to IPv4 receive handler and backlogged.
1467 From backlog it always goes here. Kerboom...
1468 Fortunately, tcp_rcv_established and rcv_established
1469 handle them correctly, but it is not case with
1470 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1471 */
1472
1473 if (skb->protocol == htons(ETH_P_IP))
1474 return tcp_v4_do_rcv(sk, skb);
1475
cfb6eeb4
YH
1476#ifdef CONFIG_TCP_MD5SIG
1477 if (tcp_v6_inbound_md5_hash (sk, skb))
1478 goto discard;
1479#endif
1480
fda9ef5d 1481 if (sk_filter(sk, skb))
1da177e4
LT
1482 goto discard;
1483
1484 /*
1485 * socket locking is here for SMP purposes as backlog rcv
1486 * is currently called with bh processing disabled.
1487 */
1488
1489 /* Do Stevens' IPV6_PKTOPTIONS.
1490
1491 Yes, guys, it is the only place in our code, where we
1492 may make it not affecting IPv4.
1493 The rest of code is protocol independent,
1494 and I do not like idea to uglify IPv4.
1495
1496 Actually, all the idea behind IPV6_PKTOPTIONS
1497 looks not very well thought. For now we latch
1498 options, received in the last packet, enqueued
1499 by tcp. Feel free to propose better solution.
1ab1457c 1500 --ANK (980728)
1da177e4
LT
1501 */
1502 if (np->rxopt.all)
1503 opt_skb = skb_clone(skb, GFP_ATOMIC);
1504
1505 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
bdeab991 1506 sock_rps_save_rxhash(sk, skb);
aa8223c7 1507 if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len))
1da177e4 1508 goto reset;
1da177e4
LT
1509 if (opt_skb)
1510 goto ipv6_pktoptions;
1511 return 0;
1512 }
1513
ab6a5bb6 1514 if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
1da177e4
LT
1515 goto csum_err;
1516
1ab1457c 1517 if (sk->sk_state == TCP_LISTEN) {
1da177e4
LT
1518 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1519 if (!nsk)
1520 goto discard;
1521
1522 /*
1523 * Queue it on the new socket if the new socket is active,
1524 * otherwise we just shortcircuit this and continue with
1525 * the new socket..
1526 */
1ab1457c 1527 if(nsk != sk) {
bdeab991 1528 sock_rps_save_rxhash(nsk, skb);
1da177e4
LT
1529 if (tcp_child_process(sk, nsk, skb))
1530 goto reset;
1531 if (opt_skb)
1532 __kfree_skb(opt_skb);
1533 return 0;
1534 }
47482f13 1535 } else
bdeab991 1536 sock_rps_save_rxhash(sk, skb);
1da177e4 1537
aa8223c7 1538 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
1da177e4 1539 goto reset;
1da177e4
LT
1540 if (opt_skb)
1541 goto ipv6_pktoptions;
1542 return 0;
1543
1544reset:
cfb6eeb4 1545 tcp_v6_send_reset(sk, skb);
1da177e4
LT
1546discard:
1547 if (opt_skb)
1548 __kfree_skb(opt_skb);
1549 kfree_skb(skb);
1550 return 0;
1551csum_err:
63231bdd 1552 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
1da177e4
LT
1553 goto discard;
1554
1555
1556ipv6_pktoptions:
1557 /* Do you ask, what is it?
1558
1559 1. skb was enqueued by tcp.
1560 2. skb is added to tail of read queue, rather than out of order.
1561 3. socket is not in passive state.
1562 4. Finally, it really contains options, which user wants to receive.
1563 */
1564 tp = tcp_sk(sk);
1565 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1566 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
333fad53 1567 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
505cbfc5 1568 np->mcast_oif = inet6_iif(opt_skb);
333fad53 1569 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
0660e03f 1570 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
4c507d28
JB
1571 if (np->rxopt.bits.rxtclass)
1572 np->rcv_tclass = ipv6_tclass(ipv6_hdr(skb));
1da177e4
LT
1573 if (ipv6_opt_accepted(sk, opt_skb)) {
1574 skb_set_owner_r(opt_skb, sk);
1575 opt_skb = xchg(&np->pktoptions, opt_skb);
1576 } else {
1577 __kfree_skb(opt_skb);
1578 opt_skb = xchg(&np->pktoptions, NULL);
1579 }
1580 }
1581
800d55f1 1582 kfree_skb(opt_skb);
1da177e4
LT
1583 return 0;
1584}
1585
e5bbef20 1586static int tcp_v6_rcv(struct sk_buff *skb)
1da177e4 1587{
cf533ea5 1588 const struct tcphdr *th;
b71d1d42 1589 const struct ipv6hdr *hdr;
1da177e4
LT
1590 struct sock *sk;
1591 int ret;
a86b1e30 1592 struct net *net = dev_net(skb->dev);
1da177e4
LT
1593
1594 if (skb->pkt_type != PACKET_HOST)
1595 goto discard_it;
1596
1597 /*
1598 * Count it even if it's bad.
1599 */
63231bdd 1600 TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
1da177e4
LT
1601
1602 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1603 goto discard_it;
1604
aa8223c7 1605 th = tcp_hdr(skb);
1da177e4
LT
1606
1607 if (th->doff < sizeof(struct tcphdr)/4)
1608 goto bad_packet;
1609 if (!pskb_may_pull(skb, th->doff*4))
1610 goto discard_it;
1611
60476372 1612 if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
1da177e4
LT
1613 goto bad_packet;
1614
aa8223c7 1615 th = tcp_hdr(skb);
e802af9c 1616 hdr = ipv6_hdr(skb);
1da177e4
LT
1617 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1618 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1619 skb->len - th->doff*4);
1620 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1621 TCP_SKB_CB(skb)->when = 0;
b82d1bb4 1622 TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
1da177e4
LT
1623 TCP_SKB_CB(skb)->sacked = 0;
1624
9a1f27c4 1625 sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest);
1da177e4
LT
1626 if (!sk)
1627 goto no_tcp_socket;
1628
1629process:
1630 if (sk->sk_state == TCP_TIME_WAIT)
1631 goto do_time_wait;
1632
e802af9c
SH
1633 if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) {
1634 NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
1635 goto discard_and_relse;
1636 }
1637
1da177e4
LT
1638 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1639 goto discard_and_relse;
1640
fda9ef5d 1641 if (sk_filter(sk, skb))
1da177e4
LT
1642 goto discard_and_relse;
1643
1644 skb->dev = NULL;
1645
293b9c42 1646 bh_lock_sock_nested(sk);
1da177e4
LT
1647 ret = 0;
1648 if (!sock_owned_by_user(sk)) {
1a2449a8 1649#ifdef CONFIG_NET_DMA
1ab1457c 1650 struct tcp_sock *tp = tcp_sk(sk);
b4caea8a 1651 if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
a2bd1140 1652 tp->ucopy.dma_chan = net_dma_find_channel();
1ab1457c
YH
1653 if (tp->ucopy.dma_chan)
1654 ret = tcp_v6_do_rcv(sk, skb);
1655 else
1a2449a8
CL
1656#endif
1657 {
1658 if (!tcp_prequeue(sk, skb))
1659 ret = tcp_v6_do_rcv(sk, skb);
1660 }
da882c1f
ED
1661 } else if (unlikely(sk_add_backlog(sk, skb,
1662 sk->sk_rcvbuf + sk->sk_sndbuf))) {
6b03a53a 1663 bh_unlock_sock(sk);
6cce09f8 1664 NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
6b03a53a
ZY
1665 goto discard_and_relse;
1666 }
1da177e4
LT
1667 bh_unlock_sock(sk);
1668
1669 sock_put(sk);
1670 return ret ? -1 : 0;
1671
1672no_tcp_socket:
1673 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1674 goto discard_it;
1675
1676 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1677bad_packet:
63231bdd 1678 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1da177e4 1679 } else {
cfb6eeb4 1680 tcp_v6_send_reset(NULL, skb);
1da177e4
LT
1681 }
1682
1683discard_it:
1684
1685 /*
1686 * Discard frame
1687 */
1688
1689 kfree_skb(skb);
1690 return 0;
1691
1692discard_and_relse:
1693 sock_put(sk);
1694 goto discard_it;
1695
1696do_time_wait:
1697 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
9469c7b4 1698 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1699 goto discard_it;
1700 }
1701
1702 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
63231bdd 1703 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
9469c7b4 1704 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1705 goto discard_it;
1706 }
1707
9469c7b4 1708 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1da177e4
LT
1709 case TCP_TW_SYN:
1710 {
1711 struct sock *sk2;
1712
c346dca1 1713 sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
0660e03f 1714 &ipv6_hdr(skb)->daddr,
505cbfc5 1715 ntohs(th->dest), inet6_iif(skb));
1da177e4 1716 if (sk2 != NULL) {
295ff7ed
ACM
1717 struct inet_timewait_sock *tw = inet_twsk(sk);
1718 inet_twsk_deschedule(tw, &tcp_death_row);
1719 inet_twsk_put(tw);
1da177e4
LT
1720 sk = sk2;
1721 goto process;
1722 }
1723 /* Fall through to ACK */
1724 }
1725 case TCP_TW_ACK:
1726 tcp_v6_timewait_ack(sk, skb);
1727 break;
1728 case TCP_TW_RST:
1729 goto no_tcp_socket;
1730 case TCP_TW_SUCCESS:;
1731 }
1732 goto discard_it;
1733}
1734
4670fd81 1735static struct inet_peer *tcp_v6_get_peer(struct sock *sk)
ccb7c410 1736{
db3949c4
DM
1737 struct rt6_info *rt = (struct rt6_info *) __sk_dst_get(sk);
1738 struct ipv6_pinfo *np = inet6_sk(sk);
db3949c4 1739
4670fd81
DM
1740 /* If we don't have a valid cached route, or we're doing IP
1741 * options which make the IPv6 header destination address
1742 * different from our peer's, do not bother with this.
1743 */
1744 if (!rt || !ipv6_addr_equal(&np->daddr, &rt->rt6i_dst.addr))
1745 return NULL;
1746 return rt6_get_peer_create(rt);
ccb7c410
DM
1747}
1748
ccb7c410
DM
1749static struct timewait_sock_ops tcp6_timewait_sock_ops = {
1750 .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
1751 .twsk_unique = tcp_twsk_unique,
1752 .twsk_destructor= tcp_twsk_destructor,
ccb7c410
DM
1753};
1754
3b401a81 1755static const struct inet_connection_sock_af_ops ipv6_specific = {
543d9cfe
ACM
1756 .queue_xmit = inet6_csk_xmit,
1757 .send_check = tcp_v6_send_check,
1758 .rebuild_header = inet6_sk_rebuild_header,
1759 .conn_request = tcp_v6_conn_request,
1760 .syn_recv_sock = tcp_v6_syn_recv_sock,
3f419d2d 1761 .get_peer = tcp_v6_get_peer,
543d9cfe 1762 .net_header_len = sizeof(struct ipv6hdr),
67469601 1763 .net_frag_header_len = sizeof(struct frag_hdr),
543d9cfe
ACM
1764 .setsockopt = ipv6_setsockopt,
1765 .getsockopt = ipv6_getsockopt,
1766 .addr2sockaddr = inet6_csk_addr2sockaddr,
1767 .sockaddr_len = sizeof(struct sockaddr_in6),
ab1e0a13 1768 .bind_conflict = inet6_csk_bind_conflict,
3fdadf7d 1769#ifdef CONFIG_COMPAT
543d9cfe
ACM
1770 .compat_setsockopt = compat_ipv6_setsockopt,
1771 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1772#endif
1da177e4
LT
1773};
1774
cfb6eeb4 1775#ifdef CONFIG_TCP_MD5SIG
b2e4b3de 1776static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
cfb6eeb4 1777 .md5_lookup = tcp_v6_md5_lookup,
49a72dfb 1778 .calc_md5_hash = tcp_v6_md5_hash_skb,
cfb6eeb4 1779 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1780};
a928630a 1781#endif
cfb6eeb4 1782
1da177e4
LT
1783/*
1784 * TCP over IPv4 via INET6 API
1785 */
1786
3b401a81 1787static const struct inet_connection_sock_af_ops ipv6_mapped = {
543d9cfe
ACM
1788 .queue_xmit = ip_queue_xmit,
1789 .send_check = tcp_v4_send_check,
1790 .rebuild_header = inet_sk_rebuild_header,
1791 .conn_request = tcp_v6_conn_request,
1792 .syn_recv_sock = tcp_v6_syn_recv_sock,
3f419d2d 1793 .get_peer = tcp_v4_get_peer,
543d9cfe
ACM
1794 .net_header_len = sizeof(struct iphdr),
1795 .setsockopt = ipv6_setsockopt,
1796 .getsockopt = ipv6_getsockopt,
1797 .addr2sockaddr = inet6_csk_addr2sockaddr,
1798 .sockaddr_len = sizeof(struct sockaddr_in6),
ab1e0a13 1799 .bind_conflict = inet6_csk_bind_conflict,
3fdadf7d 1800#ifdef CONFIG_COMPAT
543d9cfe
ACM
1801 .compat_setsockopt = compat_ipv6_setsockopt,
1802 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1803#endif
1da177e4
LT
1804};
1805
cfb6eeb4 1806#ifdef CONFIG_TCP_MD5SIG
b2e4b3de 1807static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
cfb6eeb4 1808 .md5_lookup = tcp_v4_md5_lookup,
49a72dfb 1809 .calc_md5_hash = tcp_v4_md5_hash_skb,
cfb6eeb4 1810 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1811};
a928630a 1812#endif
cfb6eeb4 1813
1da177e4
LT
1814/* NOTE: A lot of things set to zero explicitly by call to
1815 * sk_alloc() so need not be done here.
1816 */
1817static int tcp_v6_init_sock(struct sock *sk)
1818{
6687e988 1819 struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4 1820
900f65d3 1821 tcp_init_sock(sk);
1da177e4 1822
8292a17a 1823 icsk->icsk_af_ops = &ipv6_specific;
1da177e4 1824
cfb6eeb4 1825#ifdef CONFIG_TCP_MD5SIG
ac807fa8 1826 tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
cfb6eeb4
YH
1827#endif
1828
1da177e4
LT
1829 return 0;
1830}
1831
7d06b2e0 1832static void tcp_v6_destroy_sock(struct sock *sk)
1da177e4 1833{
1da177e4 1834 tcp_v4_destroy_sock(sk);
7d06b2e0 1835 inet6_destroy_sock(sk);
1da177e4
LT
1836}
1837
952a10be 1838#ifdef CONFIG_PROC_FS
1da177e4 1839/* Proc filesystem TCPv6 sock list dumping. */
1ab1457c 1840static void get_openreq6(struct seq_file *seq,
cf533ea5 1841 const struct sock *sk, struct request_sock *req, int i, int uid)
1da177e4 1842{
1da177e4 1843 int ttd = req->expires - jiffies;
b71d1d42
ED
1844 const struct in6_addr *src = &inet6_rsk(req)->loc_addr;
1845 const struct in6_addr *dest = &inet6_rsk(req)->rmt_addr;
1da177e4
LT
1846
1847 if (ttd < 0)
1848 ttd = 0;
1849
1da177e4
LT
1850 seq_printf(seq,
1851 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
71338aa7 1852 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
1da177e4
LT
1853 i,
1854 src->s6_addr32[0], src->s6_addr32[1],
1855 src->s6_addr32[2], src->s6_addr32[3],
fd507037 1856 ntohs(inet_rsk(req)->loc_port),
1da177e4
LT
1857 dest->s6_addr32[0], dest->s6_addr32[1],
1858 dest->s6_addr32[2], dest->s6_addr32[3],
2e6599cb 1859 ntohs(inet_rsk(req)->rmt_port),
1da177e4
LT
1860 TCP_SYN_RECV,
1861 0,0, /* could print option size, but that is af dependent. */
1ab1457c
YH
1862 1, /* timers active (only the expire timer) */
1863 jiffies_to_clock_t(ttd),
1da177e4
LT
1864 req->retrans,
1865 uid,
1ab1457c 1866 0, /* non standard timer */
1da177e4
LT
1867 0, /* open_requests have no inode */
1868 0, req);
1869}
1870
1871static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1872{
b71d1d42 1873 const struct in6_addr *dest, *src;
1da177e4
LT
1874 __u16 destp, srcp;
1875 int timer_active;
1876 unsigned long timer_expires;
cf533ea5
ED
1877 const struct inet_sock *inet = inet_sk(sp);
1878 const struct tcp_sock *tp = tcp_sk(sp);
463c84b9 1879 const struct inet_connection_sock *icsk = inet_csk(sp);
cf533ea5 1880 const struct ipv6_pinfo *np = inet6_sk(sp);
1da177e4
LT
1881
1882 dest = &np->daddr;
1883 src = &np->rcv_saddr;
c720c7e8
ED
1884 destp = ntohs(inet->inet_dport);
1885 srcp = ntohs(inet->inet_sport);
463c84b9
ACM
1886
1887 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
1da177e4 1888 timer_active = 1;
463c84b9
ACM
1889 timer_expires = icsk->icsk_timeout;
1890 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
1da177e4 1891 timer_active = 4;
463c84b9 1892 timer_expires = icsk->icsk_timeout;
1da177e4
LT
1893 } else if (timer_pending(&sp->sk_timer)) {
1894 timer_active = 2;
1895 timer_expires = sp->sk_timer.expires;
1896 } else {
1897 timer_active = 0;
1898 timer_expires = jiffies;
1899 }
1900
1901 seq_printf(seq,
1902 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
71338aa7 1903 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %lu %lu %u %u %d\n",
1da177e4
LT
1904 i,
1905 src->s6_addr32[0], src->s6_addr32[1],
1906 src->s6_addr32[2], src->s6_addr32[3], srcp,
1907 dest->s6_addr32[0], dest->s6_addr32[1],
1908 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1ab1457c 1909 sp->sk_state,
47da8ee6
SS
1910 tp->write_seq-tp->snd_una,
1911 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
1da177e4
LT
1912 timer_active,
1913 jiffies_to_clock_t(timer_expires - jiffies),
463c84b9 1914 icsk->icsk_retransmits,
1da177e4 1915 sock_i_uid(sp),
6687e988 1916 icsk->icsk_probes_out,
1da177e4
LT
1917 sock_i_ino(sp),
1918 atomic_read(&sp->sk_refcnt), sp,
7be87351
SH
1919 jiffies_to_clock_t(icsk->icsk_rto),
1920 jiffies_to_clock_t(icsk->icsk_ack.ato),
463c84b9 1921 (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
0b6a05c1
IJ
1922 tp->snd_cwnd,
1923 tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh
1da177e4
LT
1924 );
1925}
1926
1ab1457c 1927static void get_timewait6_sock(struct seq_file *seq,
8feaf0c0 1928 struct inet_timewait_sock *tw, int i)
1da177e4 1929{
b71d1d42 1930 const struct in6_addr *dest, *src;
1da177e4 1931 __u16 destp, srcp;
cf533ea5 1932 const struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
1da177e4
LT
1933 int ttd = tw->tw_ttd - jiffies;
1934
1935 if (ttd < 0)
1936 ttd = 0;
1937
0fa1a53e
ACM
1938 dest = &tw6->tw_v6_daddr;
1939 src = &tw6->tw_v6_rcv_saddr;
1da177e4
LT
1940 destp = ntohs(tw->tw_dport);
1941 srcp = ntohs(tw->tw_sport);
1942
1943 seq_printf(seq,
1944 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
71338aa7 1945 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
1da177e4
LT
1946 i,
1947 src->s6_addr32[0], src->s6_addr32[1],
1948 src->s6_addr32[2], src->s6_addr32[3], srcp,
1949 dest->s6_addr32[0], dest->s6_addr32[1],
1950 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1951 tw->tw_substate, 0, 0,
1952 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
1953 atomic_read(&tw->tw_refcnt), tw);
1954}
1955
1da177e4
LT
1956static int tcp6_seq_show(struct seq_file *seq, void *v)
1957{
1958 struct tcp_iter_state *st;
1959
1960 if (v == SEQ_START_TOKEN) {
1961 seq_puts(seq,
1962 " sl "
1963 "local_address "
1964 "remote_address "
1965 "st tx_queue rx_queue tr tm->when retrnsmt"
1966 " uid timeout inode\n");
1967 goto out;
1968 }
1969 st = seq->private;
1970
1971 switch (st->state) {
1972 case TCP_SEQ_STATE_LISTENING:
1973 case TCP_SEQ_STATE_ESTABLISHED:
1974 get_tcp6_sock(seq, v, st->num);
1975 break;
1976 case TCP_SEQ_STATE_OPENREQ:
1977 get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
1978 break;
1979 case TCP_SEQ_STATE_TIME_WAIT:
1980 get_timewait6_sock(seq, v, st->num);
1981 break;
1982 }
1983out:
1984 return 0;
1985}
1986
73cb88ec
AV
1987static const struct file_operations tcp6_afinfo_seq_fops = {
1988 .owner = THIS_MODULE,
1989 .open = tcp_seq_open,
1990 .read = seq_read,
1991 .llseek = seq_lseek,
1992 .release = seq_release_net
1993};
1994
1da177e4 1995static struct tcp_seq_afinfo tcp6_seq_afinfo = {
1da177e4
LT
1996 .name = "tcp6",
1997 .family = AF_INET6,
73cb88ec 1998 .seq_fops = &tcp6_afinfo_seq_fops,
9427c4b3
DL
1999 .seq_ops = {
2000 .show = tcp6_seq_show,
2001 },
1da177e4
LT
2002};
2003
2c8c1e72 2004int __net_init tcp6_proc_init(struct net *net)
1da177e4 2005{
6f8b13bc 2006 return tcp_proc_register(net, &tcp6_seq_afinfo);
1da177e4
LT
2007}
2008
6f8b13bc 2009void tcp6_proc_exit(struct net *net)
1da177e4 2010{
6f8b13bc 2011 tcp_proc_unregister(net, &tcp6_seq_afinfo);
1da177e4
LT
2012}
2013#endif
2014
2015struct proto tcpv6_prot = {
2016 .name = "TCPv6",
2017 .owner = THIS_MODULE,
2018 .close = tcp_close,
2019 .connect = tcp_v6_connect,
2020 .disconnect = tcp_disconnect,
463c84b9 2021 .accept = inet_csk_accept,
1da177e4
LT
2022 .ioctl = tcp_ioctl,
2023 .init = tcp_v6_init_sock,
2024 .destroy = tcp_v6_destroy_sock,
2025 .shutdown = tcp_shutdown,
2026 .setsockopt = tcp_setsockopt,
2027 .getsockopt = tcp_getsockopt,
1da177e4 2028 .recvmsg = tcp_recvmsg,
7ba42910
CG
2029 .sendmsg = tcp_sendmsg,
2030 .sendpage = tcp_sendpage,
1da177e4
LT
2031 .backlog_rcv = tcp_v6_do_rcv,
2032 .hash = tcp_v6_hash,
ab1e0a13
ACM
2033 .unhash = inet_unhash,
2034 .get_port = inet_csk_get_port,
1da177e4
LT
2035 .enter_memory_pressure = tcp_enter_memory_pressure,
2036 .sockets_allocated = &tcp_sockets_allocated,
2037 .memory_allocated = &tcp_memory_allocated,
2038 .memory_pressure = &tcp_memory_pressure,
0a5578cf 2039 .orphan_count = &tcp_orphan_count,
1da177e4
LT
2040 .sysctl_wmem = sysctl_tcp_wmem,
2041 .sysctl_rmem = sysctl_tcp_rmem,
2042 .max_header = MAX_TCP_HEADER,
2043 .obj_size = sizeof(struct tcp6_sock),
3ab5aee7 2044 .slab_flags = SLAB_DESTROY_BY_RCU,
6d6ee43e 2045 .twsk_prot = &tcp6_timewait_sock_ops,
60236fdd 2046 .rsk_prot = &tcp6_request_sock_ops,
39d8cda7 2047 .h.hashinfo = &tcp_hashinfo,
7ba42910 2048 .no_autobind = true,
543d9cfe
ACM
2049#ifdef CONFIG_COMPAT
2050 .compat_setsockopt = compat_tcp_setsockopt,
2051 .compat_getsockopt = compat_tcp_getsockopt,
2052#endif
d1a4c0b3
GC
2053#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
2054 .proto_cgroup = tcp_proto_cgroup,
2055#endif
1da177e4
LT
2056};
2057
41135cc8 2058static const struct inet6_protocol tcpv6_protocol = {
1da177e4
LT
2059 .handler = tcp_v6_rcv,
2060 .err_handler = tcp_v6_err,
a430a43d 2061 .gso_send_check = tcp_v6_gso_send_check,
adcfc7d0 2062 .gso_segment = tcp_tso_segment,
684f2176
HX
2063 .gro_receive = tcp6_gro_receive,
2064 .gro_complete = tcp6_gro_complete,
1da177e4
LT
2065 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
2066};
2067
1da177e4
LT
2068static struct inet_protosw tcpv6_protosw = {
2069 .type = SOCK_STREAM,
2070 .protocol = IPPROTO_TCP,
2071 .prot = &tcpv6_prot,
2072 .ops = &inet6_stream_ops,
1da177e4 2073 .no_check = 0,
d83d8461
ACM
2074 .flags = INET_PROTOSW_PERMANENT |
2075 INET_PROTOSW_ICSK,
1da177e4
LT
2076};
2077
2c8c1e72 2078static int __net_init tcpv6_net_init(struct net *net)
93ec926b 2079{
5677242f
DL
2080 return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
2081 SOCK_RAW, IPPROTO_TCP, net);
93ec926b
DL
2082}
2083
2c8c1e72 2084static void __net_exit tcpv6_net_exit(struct net *net)
93ec926b 2085{
5677242f 2086 inet_ctl_sock_destroy(net->ipv6.tcp_sk);
b099ce26
EB
2087}
2088
2c8c1e72 2089static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
b099ce26
EB
2090{
2091 inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6);
93ec926b
DL
2092}
2093
2094static struct pernet_operations tcpv6_net_ops = {
b099ce26
EB
2095 .init = tcpv6_net_init,
2096 .exit = tcpv6_net_exit,
2097 .exit_batch = tcpv6_net_exit_batch,
93ec926b
DL
2098};
2099
7f4e4868 2100int __init tcpv6_init(void)
1da177e4 2101{
7f4e4868
DL
2102 int ret;
2103
2104 ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
2105 if (ret)
2106 goto out;
2107
1da177e4 2108 /* register inet6 protocol */
7f4e4868
DL
2109 ret = inet6_register_protosw(&tcpv6_protosw);
2110 if (ret)
2111 goto out_tcpv6_protocol;
2112
93ec926b 2113 ret = register_pernet_subsys(&tcpv6_net_ops);
7f4e4868
DL
2114 if (ret)
2115 goto out_tcpv6_protosw;
2116out:
2117 return ret;
ae0f7d5f 2118
7f4e4868
DL
2119out_tcpv6_protocol:
2120 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2121out_tcpv6_protosw:
2122 inet6_unregister_protosw(&tcpv6_protosw);
2123 goto out;
2124}
2125
09f7709f 2126void tcpv6_exit(void)
7f4e4868 2127{
93ec926b 2128 unregister_pernet_subsys(&tcpv6_net_ops);
7f4e4868
DL
2129 inet6_unregister_protosw(&tcpv6_protosw);
2130 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
1da177e4 2131}