[SCTP]: #if 0 sctp_update_copy_cksum()
[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
LT
7 *
8 * $Id: tcp_ipv6.c,v 1.144 2002/02/01 22:01:04 davem Exp $
9 *
1ab1457c 10 * Based on:
1da177e4
LT
11 * linux/net/ipv4/tcp.c
12 * linux/net/ipv4/tcp_input.c
13 * linux/net/ipv4/tcp_output.c
14 *
15 * Fixes:
16 * Hideaki YOSHIFUJI : sin6_scope_id support
17 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
18 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
19 * a single port at the same time.
20 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/tcp6 to seq_file.
21 *
22 * This program is free software; you can redistribute it and/or
23 * modify it under the terms of the GNU General Public License
24 * as published by the Free Software Foundation; either version
25 * 2 of the License, or (at your option) any later version.
26 */
27
28#include <linux/module.h>
1da177e4
LT
29#include <linux/errno.h>
30#include <linux/types.h>
31#include <linux/socket.h>
32#include <linux/sockios.h>
33#include <linux/net.h>
34#include <linux/jiffies.h>
35#include <linux/in.h>
36#include <linux/in6.h>
37#include <linux/netdevice.h>
38#include <linux/init.h>
39#include <linux/jhash.h>
40#include <linux/ipsec.h>
41#include <linux/times.h>
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>
1da177e4
LT
62
63#include <asm/uaccess.h>
64
65#include <linux/proc_fs.h>
66#include <linux/seq_file.h>
67
cfb6eeb4
YH
68#include <linux/crypto.h>
69#include <linux/scatterlist.h>
70
ae0f7d5f
DW
71/* Socket used for sending RSTs and ACKs */
72static struct socket *tcp6_socket;
73
cfb6eeb4 74static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb);
60236fdd 75static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req);
1ab1457c 76static void tcp_v6_send_check(struct sock *sk, int len,
1da177e4
LT
77 struct sk_buff *skb);
78
79static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
1da177e4 80
8292a17a
ACM
81static struct inet_connection_sock_af_ops ipv6_mapped;
82static struct inet_connection_sock_af_ops ipv6_specific;
a928630a 83#ifdef CONFIG_TCP_MD5SIG
cfb6eeb4
YH
84static struct tcp_sock_af_ops tcp_sock_ipv6_specific;
85static struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
a928630a 86#endif
1da177e4 87
1da177e4
LT
88static int tcp_v6_get_port(struct sock *sk, unsigned short snum)
89{
971af18b
ACM
90 return inet_csk_get_port(&tcp_hashinfo, sk, snum,
91 inet6_csk_bind_conflict);
1da177e4
LT
92}
93
1da177e4
LT
94static void tcp_v6_hash(struct sock *sk)
95{
96 if (sk->sk_state != TCP_CLOSE) {
8292a17a 97 if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
1da177e4
LT
98 tcp_prot.hash(sk);
99 return;
100 }
101 local_bh_disable();
90b19d31 102 __inet6_hash(&tcp_hashinfo, sk);
1da177e4
LT
103 local_bh_enable();
104 }
105}
106
868c86bc 107static __inline__ __sum16 tcp_v6_check(struct tcphdr *th, int len,
1ab1457c
YH
108 struct in6_addr *saddr,
109 struct in6_addr *daddr,
868c86bc 110 __wsum base)
1da177e4
LT
111{
112 return csum_ipv6_magic(saddr, daddr, len, IPPROTO_TCP, base);
113}
114
a94f723d 115static __u32 tcp_v6_init_sequence(struct sk_buff *skb)
1da177e4 116{
0660e03f
ACM
117 return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
118 ipv6_hdr(skb)->saddr.s6_addr32,
aa8223c7
ACM
119 tcp_hdr(skb)->dest,
120 tcp_hdr(skb)->source);
1da177e4
LT
121}
122
1ab1457c 123static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
1da177e4
LT
124 int addr_len)
125{
126 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
1ab1457c 127 struct inet_sock *inet = inet_sk(sk);
d83d8461 128 struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4
LT
129 struct ipv6_pinfo *np = inet6_sk(sk);
130 struct tcp_sock *tp = tcp_sk(sk);
131 struct in6_addr *saddr = NULL, *final_p = NULL, final;
132 struct flowi fl;
133 struct dst_entry *dst;
134 int addr_type;
135 int err;
136
1ab1457c 137 if (addr_len < SIN6_LEN_RFC2133)
1da177e4
LT
138 return -EINVAL;
139
1ab1457c 140 if (usin->sin6_family != AF_INET6)
1da177e4
LT
141 return(-EAFNOSUPPORT);
142
143 memset(&fl, 0, sizeof(fl));
144
145 if (np->sndflow) {
146 fl.fl6_flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
147 IP6_ECN_flow_init(fl.fl6_flowlabel);
148 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
149 struct ip6_flowlabel *flowlabel;
150 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
151 if (flowlabel == NULL)
152 return -EINVAL;
153 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
154 fl6_sock_release(flowlabel);
155 }
156 }
157
158 /*
1ab1457c
YH
159 * connect() to INADDR_ANY means loopback (BSD'ism).
160 */
161
162 if(ipv6_addr_any(&usin->sin6_addr))
163 usin->sin6_addr.s6_addr[15] = 0x1;
1da177e4
LT
164
165 addr_type = ipv6_addr_type(&usin->sin6_addr);
166
167 if(addr_type & IPV6_ADDR_MULTICAST)
168 return -ENETUNREACH;
169
170 if (addr_type&IPV6_ADDR_LINKLOCAL) {
171 if (addr_len >= sizeof(struct sockaddr_in6) &&
172 usin->sin6_scope_id) {
173 /* If interface is set while binding, indices
174 * must coincide.
175 */
176 if (sk->sk_bound_dev_if &&
177 sk->sk_bound_dev_if != usin->sin6_scope_id)
178 return -EINVAL;
179
180 sk->sk_bound_dev_if = usin->sin6_scope_id;
181 }
182
183 /* Connect to link-local address requires an interface */
184 if (!sk->sk_bound_dev_if)
185 return -EINVAL;
186 }
187
188 if (tp->rx_opt.ts_recent_stamp &&
189 !ipv6_addr_equal(&np->daddr, &usin->sin6_addr)) {
190 tp->rx_opt.ts_recent = 0;
191 tp->rx_opt.ts_recent_stamp = 0;
192 tp->write_seq = 0;
193 }
194
195 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
196 np->flow_label = fl.fl6_flowlabel;
197
198 /*
199 * TCP over IPv4
200 */
201
202 if (addr_type == IPV6_ADDR_MAPPED) {
d83d8461 203 u32 exthdrlen = icsk->icsk_ext_hdr_len;
1da177e4
LT
204 struct sockaddr_in sin;
205
206 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
207
208 if (__ipv6_only_sock(sk))
209 return -ENETUNREACH;
210
211 sin.sin_family = AF_INET;
212 sin.sin_port = usin->sin6_port;
213 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
214
d83d8461 215 icsk->icsk_af_ops = &ipv6_mapped;
1da177e4 216 sk->sk_backlog_rcv = tcp_v4_do_rcv;
cfb6eeb4
YH
217#ifdef CONFIG_TCP_MD5SIG
218 tp->af_specific = &tcp_sock_ipv6_mapped_specific;
219#endif
1da177e4
LT
220
221 err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
222
223 if (err) {
d83d8461
ACM
224 icsk->icsk_ext_hdr_len = exthdrlen;
225 icsk->icsk_af_ops = &ipv6_specific;
1da177e4 226 sk->sk_backlog_rcv = tcp_v6_do_rcv;
cfb6eeb4
YH
227#ifdef CONFIG_TCP_MD5SIG
228 tp->af_specific = &tcp_sock_ipv6_specific;
229#endif
1da177e4
LT
230 goto failure;
231 } else {
232 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
233 inet->saddr);
234 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
235 inet->rcv_saddr);
236 }
237
238 return err;
239 }
240
241 if (!ipv6_addr_any(&np->rcv_saddr))
242 saddr = &np->rcv_saddr;
243
244 fl.proto = IPPROTO_TCP;
245 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
246 ipv6_addr_copy(&fl.fl6_src,
247 (saddr ? saddr : &np->saddr));
248 fl.oif = sk->sk_bound_dev_if;
249 fl.fl_ip_dport = usin->sin6_port;
250 fl.fl_ip_sport = inet->sport;
251
252 if (np->opt && np->opt->srcrt) {
253 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
254 ipv6_addr_copy(&final, &fl.fl6_dst);
255 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
256 final_p = &final;
257 }
258
beb8d13b
VY
259 security_sk_classify_flow(sk, &fl);
260
1da177e4
LT
261 err = ip6_dst_lookup(sk, &dst, &fl);
262 if (err)
263 goto failure;
264 if (final_p)
265 ipv6_addr_copy(&fl.fl6_dst, final_p);
266
14e50e57
DM
267 if ((err = __xfrm_lookup(&dst, &fl, sk, 1)) < 0) {
268 if (err == -EREMOTE)
269 err = ip6_dst_blackhole(sk, &dst, &fl);
270 if (err < 0)
271 goto failure;
272 }
1da177e4
LT
273
274 if (saddr == NULL) {
275 saddr = &fl.fl6_src;
276 ipv6_addr_copy(&np->rcv_saddr, saddr);
277 }
278
279 /* set the source address */
280 ipv6_addr_copy(&np->saddr, saddr);
281 inet->rcv_saddr = LOOPBACK4_IPV6;
282
f83ef8c0 283 sk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 284 __ip6_dst_store(sk, dst, NULL, NULL);
1da177e4 285
d83d8461 286 icsk->icsk_ext_hdr_len = 0;
1da177e4 287 if (np->opt)
d83d8461
ACM
288 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
289 np->opt->opt_nflen);
1da177e4
LT
290
291 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
292
293 inet->dport = usin->sin6_port;
294
295 tcp_set_state(sk, TCP_SYN_SENT);
d8313f5c 296 err = inet6_hash_connect(&tcp_death_row, sk);
1da177e4
LT
297 if (err)
298 goto late_failure;
299
300 if (!tp->write_seq)
301 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
302 np->daddr.s6_addr32,
303 inet->sport,
304 inet->dport);
305
306 err = tcp_connect(sk);
307 if (err)
308 goto late_failure;
309
310 return 0;
311
312late_failure:
313 tcp_set_state(sk, TCP_CLOSE);
314 __sk_dst_reset(sk);
315failure:
316 inet->dport = 0;
317 sk->sk_route_caps = 0;
318 return err;
319}
320
321static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
04ce6909 322 int type, int code, int offset, __be32 info)
1da177e4
LT
323{
324 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
505cbfc5 325 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
1da177e4
LT
326 struct ipv6_pinfo *np;
327 struct sock *sk;
328 int err;
1ab1457c 329 struct tcp_sock *tp;
1da177e4
LT
330 __u32 seq;
331
505cbfc5
ACM
332 sk = inet6_lookup(&tcp_hashinfo, &hdr->daddr, th->dest, &hdr->saddr,
333 th->source, skb->dev->ifindex);
1da177e4
LT
334
335 if (sk == NULL) {
336 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
337 return;
338 }
339
340 if (sk->sk_state == TCP_TIME_WAIT) {
9469c7b4 341 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
342 return;
343 }
344
345 bh_lock_sock(sk);
346 if (sock_owned_by_user(sk))
347 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
348
349 if (sk->sk_state == TCP_CLOSE)
350 goto out;
351
352 tp = tcp_sk(sk);
1ab1457c 353 seq = ntohl(th->seq);
1da177e4
LT
354 if (sk->sk_state != TCP_LISTEN &&
355 !between(seq, tp->snd_una, tp->snd_nxt)) {
356 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
357 goto out;
358 }
359
360 np = inet6_sk(sk);
361
362 if (type == ICMPV6_PKT_TOOBIG) {
363 struct dst_entry *dst = NULL;
364
365 if (sock_owned_by_user(sk))
366 goto out;
367 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
368 goto out;
369
370 /* icmp should have updated the destination cache entry */
371 dst = __sk_dst_check(sk, np->dst_cookie);
372
373 if (dst == NULL) {
374 struct inet_sock *inet = inet_sk(sk);
375 struct flowi fl;
376
377 /* BUGGG_FUTURE: Again, it is not clear how
378 to handle rthdr case. Ignore this complexity
379 for now.
380 */
381 memset(&fl, 0, sizeof(fl));
382 fl.proto = IPPROTO_TCP;
383 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
384 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
385 fl.oif = sk->sk_bound_dev_if;
386 fl.fl_ip_dport = inet->dport;
387 fl.fl_ip_sport = inet->sport;
beb8d13b 388 security_skb_classify_flow(skb, &fl);
1da177e4
LT
389
390 if ((err = ip6_dst_lookup(sk, &dst, &fl))) {
391 sk->sk_err_soft = -err;
392 goto out;
393 }
394
395 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
396 sk->sk_err_soft = -err;
397 goto out;
398 }
399
400 } else
401 dst_hold(dst);
402
d83d8461 403 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
1da177e4
LT
404 tcp_sync_mss(sk, dst_mtu(dst));
405 tcp_simple_retransmit(sk);
406 } /* else let the usual retransmit timer handle it */
407 dst_release(dst);
408 goto out;
409 }
410
411 icmpv6_err_convert(type, code, &err);
412
60236fdd 413 /* Might be for an request_sock */
1da177e4 414 switch (sk->sk_state) {
60236fdd 415 struct request_sock *req, **prev;
1da177e4
LT
416 case TCP_LISTEN:
417 if (sock_owned_by_user(sk))
418 goto out;
419
8129765a
ACM
420 req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
421 &hdr->saddr, inet6_iif(skb));
1da177e4
LT
422 if (!req)
423 goto out;
424
425 /* ICMPs are not backlogged, hence we cannot get
426 * an established socket here.
427 */
428 BUG_TRAP(req->sk == NULL);
429
2e6599cb 430 if (seq != tcp_rsk(req)->snt_isn) {
1da177e4
LT
431 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
432 goto out;
433 }
434
463c84b9 435 inet_csk_reqsk_queue_drop(sk, req, prev);
1da177e4
LT
436 goto out;
437
438 case TCP_SYN_SENT:
439 case TCP_SYN_RECV: /* Cannot happen.
1ab1457c 440 It can, it SYNs are crossed. --ANK */
1da177e4 441 if (!sock_owned_by_user(sk)) {
1da177e4
LT
442 sk->sk_err = err;
443 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
444
445 tcp_done(sk);
446 } else
447 sk->sk_err_soft = err;
448 goto out;
449 }
450
451 if (!sock_owned_by_user(sk) && np->recverr) {
452 sk->sk_err = err;
453 sk->sk_error_report(sk);
454 } else
455 sk->sk_err_soft = err;
456
457out:
458 bh_unlock_sock(sk);
459 sock_put(sk);
460}
461
462
60236fdd 463static int tcp_v6_send_synack(struct sock *sk, struct request_sock *req,
1da177e4
LT
464 struct dst_entry *dst)
465{
ca304b61 466 struct inet6_request_sock *treq = inet6_rsk(req);
1da177e4
LT
467 struct ipv6_pinfo *np = inet6_sk(sk);
468 struct sk_buff * skb;
469 struct ipv6_txoptions *opt = NULL;
470 struct in6_addr * final_p = NULL, final;
471 struct flowi fl;
472 int err = -1;
473
474 memset(&fl, 0, sizeof(fl));
475 fl.proto = IPPROTO_TCP;
2e6599cb
ACM
476 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
477 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
1da177e4 478 fl.fl6_flowlabel = 0;
2e6599cb
ACM
479 fl.oif = treq->iif;
480 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
1da177e4 481 fl.fl_ip_sport = inet_sk(sk)->sport;
4237c75c 482 security_req_classify_flow(req, &fl);
1da177e4
LT
483
484 if (dst == NULL) {
485 opt = np->opt;
1da177e4
LT
486 if (opt && opt->srcrt) {
487 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
488 ipv6_addr_copy(&final, &fl.fl6_dst);
489 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
490 final_p = &final;
491 }
492
493 err = ip6_dst_lookup(sk, &dst, &fl);
494 if (err)
495 goto done;
496 if (final_p)
497 ipv6_addr_copy(&fl.fl6_dst, final_p);
498 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
499 goto done;
500 }
501
502 skb = tcp_make_synack(sk, dst, req);
503 if (skb) {
aa8223c7 504 struct tcphdr *th = tcp_hdr(skb);
1da177e4
LT
505
506 th->check = tcp_v6_check(th, skb->len,
2e6599cb 507 &treq->loc_addr, &treq->rmt_addr,
1da177e4
LT
508 csum_partial((char *)th, skb->len, skb->csum));
509
2e6599cb 510 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
1da177e4 511 err = ip6_xmit(sk, skb, &fl, opt, 0);
b9df3cb8 512 err = net_xmit_eval(err);
1da177e4
LT
513 }
514
515done:
1ab1457c 516 if (opt && opt != np->opt)
1da177e4 517 sock_kfree_s(sk, opt, opt->tot_len);
78b91042 518 dst_release(dst);
1da177e4
LT
519 return err;
520}
521
60236fdd 522static void tcp_v6_reqsk_destructor(struct request_sock *req)
1da177e4 523{
ca304b61
ACM
524 if (inet6_rsk(req)->pktopts)
525 kfree_skb(inet6_rsk(req)->pktopts);
1da177e4
LT
526}
527
cfb6eeb4
YH
528#ifdef CONFIG_TCP_MD5SIG
529static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
530 struct in6_addr *addr)
531{
532 struct tcp_sock *tp = tcp_sk(sk);
533 int i;
534
535 BUG_ON(tp == NULL);
536
537 if (!tp->md5sig_info || !tp->md5sig_info->entries6)
538 return NULL;
539
540 for (i = 0; i < tp->md5sig_info->entries6; i++) {
541 if (ipv6_addr_cmp(&tp->md5sig_info->keys6[i].addr, addr) == 0)
f8ab18d2 542 return &tp->md5sig_info->keys6[i].base;
cfb6eeb4
YH
543 }
544 return NULL;
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
559static int tcp_v6_md5_do_add(struct sock *sk, struct in6_addr *peer,
560 char *newkey, u8 newkeylen)
561{
562 /* Add key to the list */
563 struct tcp6_md5sig_key *key;
564 struct tcp_sock *tp = tcp_sk(sk);
565 struct tcp6_md5sig_key *keys;
566
567 key = (struct tcp6_md5sig_key*) tcp_v6_md5_do_lookup(sk, peer);
568 if (key) {
569 /* modify existing entry - just update that one */
f8ab18d2
DM
570 kfree(key->base.key);
571 key->base.key = newkey;
572 key->base.keylen = newkeylen;
cfb6eeb4
YH
573 } else {
574 /* reallocate new list if current one is full. */
575 if (!tp->md5sig_info) {
576 tp->md5sig_info = kzalloc(sizeof(*tp->md5sig_info), GFP_ATOMIC);
577 if (!tp->md5sig_info) {
578 kfree(newkey);
579 return -ENOMEM;
580 }
3d7dbeac 581 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
cfb6eeb4
YH
582 }
583 tcp_alloc_md5sig_pool();
584 if (tp->md5sig_info->alloced6 == tp->md5sig_info->entries6) {
585 keys = kmalloc((sizeof (tp->md5sig_info->keys6[0]) *
586 (tp->md5sig_info->entries6 + 1)), GFP_ATOMIC);
587
588 if (!keys) {
589 tcp_free_md5sig_pool();
590 kfree(newkey);
591 return -ENOMEM;
592 }
593
594 if (tp->md5sig_info->entries6)
595 memmove(keys, tp->md5sig_info->keys6,
596 (sizeof (tp->md5sig_info->keys6[0]) *
597 tp->md5sig_info->entries6));
598
599 kfree(tp->md5sig_info->keys6);
600 tp->md5sig_info->keys6 = keys;
601 tp->md5sig_info->alloced6++;
602 }
603
604 ipv6_addr_copy(&tp->md5sig_info->keys6[tp->md5sig_info->entries6].addr,
605 peer);
f8ab18d2
DM
606 tp->md5sig_info->keys6[tp->md5sig_info->entries6].base.key = newkey;
607 tp->md5sig_info->keys6[tp->md5sig_info->entries6].base.keylen = newkeylen;
cfb6eeb4
YH
608
609 tp->md5sig_info->entries6++;
610 }
611 return 0;
612}
613
614static int tcp_v6_md5_add_func(struct sock *sk, struct sock *addr_sk,
615 u8 *newkey, __u8 newkeylen)
616{
617 return tcp_v6_md5_do_add(sk, &inet6_sk(addr_sk)->daddr,
618 newkey, newkeylen);
619}
620
621static int tcp_v6_md5_do_del(struct sock *sk, struct in6_addr *peer)
622{
623 struct tcp_sock *tp = tcp_sk(sk);
624 int i;
625
626 for (i = 0; i < tp->md5sig_info->entries6; i++) {
627 if (ipv6_addr_cmp(&tp->md5sig_info->keys6[i].addr, peer) == 0) {
628 /* Free the key */
f8ab18d2 629 kfree(tp->md5sig_info->keys6[i].base.key);
cfb6eeb4
YH
630 tp->md5sig_info->entries6--;
631
632 if (tp->md5sig_info->entries6 == 0) {
633 kfree(tp->md5sig_info->keys6);
634 tp->md5sig_info->keys6 = NULL;
ca983cef 635 tp->md5sig_info->alloced6 = 0;
cfb6eeb4
YH
636
637 tcp_free_md5sig_pool();
638
639 return 0;
640 } else {
641 /* shrink the database */
642 if (tp->md5sig_info->entries6 != i)
643 memmove(&tp->md5sig_info->keys6[i],
644 &tp->md5sig_info->keys6[i+1],
645 (tp->md5sig_info->entries6 - i)
646 * sizeof (tp->md5sig_info->keys6[0]));
647 }
648 }
649 }
650 return -ENOENT;
651}
652
653static void tcp_v6_clear_md5_list (struct sock *sk)
654{
655 struct tcp_sock *tp = tcp_sk(sk);
656 int i;
657
658 if (tp->md5sig_info->entries6) {
659 for (i = 0; i < tp->md5sig_info->entries6; i++)
f8ab18d2 660 kfree(tp->md5sig_info->keys6[i].base.key);
cfb6eeb4
YH
661 tp->md5sig_info->entries6 = 0;
662 tcp_free_md5sig_pool();
663 }
664
665 kfree(tp->md5sig_info->keys6);
666 tp->md5sig_info->keys6 = NULL;
667 tp->md5sig_info->alloced6 = 0;
668
669 if (tp->md5sig_info->entries4) {
670 for (i = 0; i < tp->md5sig_info->entries4; i++)
f8ab18d2 671 kfree(tp->md5sig_info->keys4[i].base.key);
cfb6eeb4
YH
672 tp->md5sig_info->entries4 = 0;
673 tcp_free_md5sig_pool();
674 }
675
676 kfree(tp->md5sig_info->keys4);
677 tp->md5sig_info->keys4 = NULL;
678 tp->md5sig_info->alloced4 = 0;
679}
680
681static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval,
682 int optlen)
683{
684 struct tcp_md5sig cmd;
685 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
686 u8 *newkey;
687
688 if (optlen < sizeof(cmd))
689 return -EINVAL;
690
691 if (copy_from_user(&cmd, optval, sizeof(cmd)))
692 return -EFAULT;
693
694 if (sin6->sin6_family != AF_INET6)
695 return -EINVAL;
696
697 if (!cmd.tcpm_keylen) {
698 if (!tcp_sk(sk)->md5sig_info)
699 return -ENOENT;
e773e4fa 700 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
cfb6eeb4
YH
701 return tcp_v4_md5_do_del(sk, sin6->sin6_addr.s6_addr32[3]);
702 return tcp_v6_md5_do_del(sk, &sin6->sin6_addr);
703 }
704
705 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
706 return -EINVAL;
707
708 if (!tcp_sk(sk)->md5sig_info) {
709 struct tcp_sock *tp = tcp_sk(sk);
710 struct tcp_md5sig_info *p;
711
712 p = kzalloc(sizeof(struct tcp_md5sig_info), GFP_KERNEL);
713 if (!p)
714 return -ENOMEM;
715
716 tp->md5sig_info = p;
3d7dbeac 717 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
cfb6eeb4
YH
718 }
719
af879cc7 720 newkey = kmemdup(cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
cfb6eeb4
YH
721 if (!newkey)
722 return -ENOMEM;
e773e4fa 723 if (ipv6_addr_v4mapped(&sin6->sin6_addr)) {
cfb6eeb4
YH
724 return tcp_v4_md5_do_add(sk, sin6->sin6_addr.s6_addr32[3],
725 newkey, cmd.tcpm_keylen);
726 }
727 return tcp_v6_md5_do_add(sk, &sin6->sin6_addr, newkey, cmd.tcpm_keylen);
728}
729
730static int tcp_v6_do_calc_md5_hash(char *md5_hash, struct tcp_md5sig_key *key,
731 struct in6_addr *saddr,
732 struct in6_addr *daddr,
733 struct tcphdr *th, int protocol,
734 int tcplen)
735{
736 struct scatterlist sg[4];
737 __u16 data_len;
738 int block = 0;
8e5200f5 739 __sum16 cksum;
cfb6eeb4
YH
740 struct tcp_md5sig_pool *hp;
741 struct tcp6_pseudohdr *bp;
742 struct hash_desc *desc;
743 int err;
744 unsigned int nbytes = 0;
745
746 hp = tcp_get_md5sig_pool();
747 if (!hp) {
748 printk(KERN_WARNING "%s(): hash pool not found...\n", __FUNCTION__);
749 goto clear_hash_noput;
750 }
751 bp = &hp->md5_blk.ip6;
752 desc = &hp->md5_desc;
753
754 /* 1. TCP pseudo-header (RFC2460) */
755 ipv6_addr_copy(&bp->saddr, saddr);
756 ipv6_addr_copy(&bp->daddr, daddr);
757 bp->len = htonl(tcplen);
758 bp->protocol = htonl(protocol);
759
c7da57a1
DM
760 sg_init_table(sg, 4);
761
cfb6eeb4
YH
762 sg_set_buf(&sg[block++], bp, sizeof(*bp));
763 nbytes += sizeof(*bp);
764
765 /* 2. TCP header, excluding options */
766 cksum = th->check;
767 th->check = 0;
768 sg_set_buf(&sg[block++], th, sizeof(*th));
769 nbytes += sizeof(*th);
770
771 /* 3. TCP segment data (if any) */
772 data_len = tcplen - (th->doff << 2);
773 if (data_len > 0) {
774 u8 *data = (u8 *)th + (th->doff << 2);
775 sg_set_buf(&sg[block++], data, data_len);
776 nbytes += data_len;
777 }
778
779 /* 4. shared key */
780 sg_set_buf(&sg[block++], key->key, key->keylen);
781 nbytes += key->keylen;
782
c7da57a1
DM
783 sg_mark_end(sg, block);
784
cfb6eeb4
YH
785 /* Now store the hash into the packet */
786 err = crypto_hash_init(desc);
787 if (err) {
788 printk(KERN_WARNING "%s(): hash_init failed\n", __FUNCTION__);
789 goto clear_hash;
790 }
791 err = crypto_hash_update(desc, sg, nbytes);
792 if (err) {
793 printk(KERN_WARNING "%s(): hash_update failed\n", __FUNCTION__);
794 goto clear_hash;
795 }
796 err = crypto_hash_final(desc, md5_hash);
797 if (err) {
798 printk(KERN_WARNING "%s(): hash_final failed\n", __FUNCTION__);
799 goto clear_hash;
800 }
801
802 /* Reset header, and free up the crypto */
803 tcp_put_md5sig_pool();
804 th->check = cksum;
805out:
806 return 0;
807clear_hash:
808 tcp_put_md5sig_pool();
809clear_hash_noput:
810 memset(md5_hash, 0, 16);
811 goto out;
812}
813
814static int tcp_v6_calc_md5_hash(char *md5_hash, struct tcp_md5sig_key *key,
815 struct sock *sk,
816 struct dst_entry *dst,
817 struct request_sock *req,
818 struct tcphdr *th, int protocol,
819 int tcplen)
820{
821 struct in6_addr *saddr, *daddr;
822
823 if (sk) {
824 saddr = &inet6_sk(sk)->saddr;
825 daddr = &inet6_sk(sk)->daddr;
826 } else {
827 saddr = &inet6_rsk(req)->loc_addr;
828 daddr = &inet6_rsk(req)->rmt_addr;
829 }
830 return tcp_v6_do_calc_md5_hash(md5_hash, key,
831 saddr, daddr,
832 th, protocol, tcplen);
833}
834
835static int tcp_v6_inbound_md5_hash (struct sock *sk, struct sk_buff *skb)
836{
837 __u8 *hash_location = NULL;
838 struct tcp_md5sig_key *hash_expected;
0660e03f 839 struct ipv6hdr *ip6h = ipv6_hdr(skb);
aa8223c7 840 struct tcphdr *th = tcp_hdr(skb);
cfb6eeb4
YH
841 int length = (th->doff << 2) - sizeof (*th);
842 int genhash;
843 u8 *ptr;
844 u8 newhash[16];
845
846 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
847
848 /* If the TCP option is too short, we can short cut */
849 if (length < TCPOLEN_MD5SIG)
850 return hash_expected ? 1 : 0;
851
852 /* parse options */
853 ptr = (u8*)(th + 1);
854 while (length > 0) {
855 int opcode = *ptr++;
856 int opsize;
857
858 switch(opcode) {
859 case TCPOPT_EOL:
860 goto done_opts;
861 case TCPOPT_NOP:
862 length--;
863 continue;
864 default:
865 opsize = *ptr++;
866 if (opsize < 2 || opsize > length)
867 goto done_opts;
868 if (opcode == TCPOPT_MD5SIG) {
869 hash_location = ptr;
870 goto done_opts;
871 }
872 }
873 ptr += opsize - 2;
874 length -= opsize;
875 }
876
877done_opts:
878 /* do we have a hash as expected? */
879 if (!hash_expected) {
880 if (!hash_location)
881 return 0;
882 if (net_ratelimit()) {
883 printk(KERN_INFO "MD5 Hash NOT expected but found "
884 "(" NIP6_FMT ", %u)->"
885 "(" NIP6_FMT ", %u)\n",
886 NIP6(ip6h->saddr), ntohs(th->source),
887 NIP6(ip6h->daddr), ntohs(th->dest));
888 }
889 return 1;
890 }
891
892 if (!hash_location) {
893 if (net_ratelimit()) {
894 printk(KERN_INFO "MD5 Hash expected but NOT found "
895 "(" NIP6_FMT ", %u)->"
896 "(" NIP6_FMT ", %u)\n",
897 NIP6(ip6h->saddr), ntohs(th->source),
898 NIP6(ip6h->daddr), ntohs(th->dest));
899 }
900 return 1;
901 }
902
903 /* check the signature */
904 genhash = tcp_v6_do_calc_md5_hash(newhash,
905 hash_expected,
906 &ip6h->saddr, &ip6h->daddr,
907 th, sk->sk_protocol,
908 skb->len);
909 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
910 if (net_ratelimit()) {
911 printk(KERN_INFO "MD5 Hash %s for "
912 "(" NIP6_FMT ", %u)->"
913 "(" NIP6_FMT ", %u)\n",
914 genhash ? "failed" : "mismatch",
915 NIP6(ip6h->saddr), ntohs(th->source),
916 NIP6(ip6h->daddr), ntohs(th->dest));
917 }
918 return 1;
919 }
920 return 0;
921}
922#endif
923
9ec75fe8 924static struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
1da177e4 925 .family = AF_INET6,
2e6599cb 926 .obj_size = sizeof(struct tcp6_request_sock),
1da177e4 927 .rtx_syn_ack = tcp_v6_send_synack,
60236fdd
ACM
928 .send_ack = tcp_v6_reqsk_send_ack,
929 .destructor = tcp_v6_reqsk_destructor,
1da177e4
LT
930 .send_reset = tcp_v6_send_reset
931};
932
cfb6eeb4 933#ifdef CONFIG_TCP_MD5SIG
b6332e6c 934static struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
cfb6eeb4 935 .md5_lookup = tcp_v6_reqsk_md5_lookup,
cfb6eeb4 936};
b6332e6c 937#endif
cfb6eeb4 938
6d6ee43e
ACM
939static struct timewait_sock_ops tcp6_timewait_sock_ops = {
940 .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
941 .twsk_unique = tcp_twsk_unique,
cfb6eeb4 942 .twsk_destructor= tcp_twsk_destructor,
6d6ee43e
ACM
943};
944
8292a17a 945static void tcp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
1da177e4
LT
946{
947 struct ipv6_pinfo *np = inet6_sk(sk);
aa8223c7 948 struct tcphdr *th = tcp_hdr(skb);
1da177e4 949
84fa7933 950 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1da177e4 951 th->check = ~csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP, 0);
663ead3b 952 skb->csum_start = skb_transport_header(skb) - skb->head;
ff1dcadb 953 skb->csum_offset = offsetof(struct tcphdr, check);
1da177e4 954 } else {
1ab1457c
YH
955 th->check = csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP,
956 csum_partial((char *)th, th->doff<<2,
1da177e4
LT
957 skb->csum));
958 }
959}
960
a430a43d
HX
961static int tcp_v6_gso_send_check(struct sk_buff *skb)
962{
963 struct ipv6hdr *ipv6h;
964 struct tcphdr *th;
965
966 if (!pskb_may_pull(skb, sizeof(*th)))
967 return -EINVAL;
968
0660e03f 969 ipv6h = ipv6_hdr(skb);
aa8223c7 970 th = tcp_hdr(skb);
a430a43d
HX
971
972 th->check = 0;
973 th->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
974 IPPROTO_TCP, 0);
663ead3b 975 skb->csum_start = skb_transport_header(skb) - skb->head;
ff1dcadb 976 skb->csum_offset = offsetof(struct tcphdr, check);
84fa7933 977 skb->ip_summed = CHECKSUM_PARTIAL;
a430a43d
HX
978 return 0;
979}
1da177e4 980
cfb6eeb4 981static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
1da177e4 982{
aa8223c7 983 struct tcphdr *th = tcp_hdr(skb), *t1;
1da177e4
LT
984 struct sk_buff *buff;
985 struct flowi fl;
cfb6eeb4
YH
986 int tot_len = sizeof(*th);
987#ifdef CONFIG_TCP_MD5SIG
988 struct tcp_md5sig_key *key;
989#endif
1da177e4
LT
990
991 if (th->rst)
992 return;
993
994 if (!ipv6_unicast_destination(skb))
1ab1457c 995 return;
1da177e4 996
cfb6eeb4
YH
997#ifdef CONFIG_TCP_MD5SIG
998 if (sk)
0660e03f 999 key = tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr);
cfb6eeb4
YH
1000 else
1001 key = NULL;
1002
1003 if (key)
1004 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1005#endif
1006
1da177e4
LT
1007 /*
1008 * We need to grab some memory, and put together an RST,
1009 * and then put it into the queue to be sent.
1010 */
1011
cfb6eeb4 1012 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1da177e4 1013 GFP_ATOMIC);
1ab1457c
YH
1014 if (buff == NULL)
1015 return;
1da177e4 1016
cfb6eeb4 1017 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1da177e4 1018
cfb6eeb4 1019 t1 = (struct tcphdr *) skb_push(buff, tot_len);
1da177e4
LT
1020
1021 /* Swap the send and the receive. */
1022 memset(t1, 0, sizeof(*t1));
1023 t1->dest = th->source;
1024 t1->source = th->dest;
cfb6eeb4 1025 t1->doff = tot_len / 4;
1da177e4 1026 t1->rst = 1;
1ab1457c 1027
1da177e4 1028 if(th->ack) {
1ab1457c 1029 t1->seq = th->ack_seq;
1da177e4
LT
1030 } else {
1031 t1->ack = 1;
1032 t1->ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin
1033 + skb->len - (th->doff<<2));
1034 }
1035
cfb6eeb4
YH
1036#ifdef CONFIG_TCP_MD5SIG
1037 if (key) {
8e5200f5 1038 __be32 *opt = (__be32*)(t1 + 1);
cfb6eeb4
YH
1039 opt[0] = htonl((TCPOPT_NOP << 24) |
1040 (TCPOPT_NOP << 16) |
1041 (TCPOPT_MD5SIG << 8) |
1042 TCPOLEN_MD5SIG);
0660e03f
ACM
1043 tcp_v6_do_calc_md5_hash((__u8 *)&opt[1], key,
1044 &ipv6_hdr(skb)->daddr,
1045 &ipv6_hdr(skb)->saddr,
1046 t1, IPPROTO_TCP, tot_len);
cfb6eeb4
YH
1047 }
1048#endif
1049
1da177e4
LT
1050 buff->csum = csum_partial((char *)t1, sizeof(*t1), 0);
1051
1052 memset(&fl, 0, sizeof(fl));
0660e03f
ACM
1053 ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
1054 ipv6_addr_copy(&fl.fl6_src, &ipv6_hdr(skb)->daddr);
1da177e4
LT
1055
1056 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1057 sizeof(*t1), IPPROTO_TCP,
1058 buff->csum);
1059
1060 fl.proto = IPPROTO_TCP;
505cbfc5 1061 fl.oif = inet6_iif(skb);
1da177e4
LT
1062 fl.fl_ip_dport = t1->dest;
1063 fl.fl_ip_sport = t1->source;
beb8d13b 1064 security_skb_classify_flow(skb, &fl);
1da177e4
LT
1065
1066 /* sk = NULL, but it is safe for now. RST socket required. */
1067 if (!ip6_dst_lookup(NULL, &buff->dst, &fl)) {
1068
ecc51b6d 1069 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
ae0f7d5f 1070 ip6_xmit(tcp6_socket->sk, buff, &fl, NULL, 0);
ecc51b6d
ACM
1071 TCP_INC_STATS_BH(TCP_MIB_OUTSEGS);
1072 TCP_INC_STATS_BH(TCP_MIB_OUTRSTS);
1da177e4 1073 return;
ecc51b6d 1074 }
1da177e4
LT
1075 }
1076
1077 kfree_skb(buff);
1078}
1079
cfb6eeb4
YH
1080static void tcp_v6_send_ack(struct tcp_timewait_sock *tw,
1081 struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts)
1da177e4 1082{
aa8223c7 1083 struct tcphdr *th = tcp_hdr(skb), *t1;
1da177e4
LT
1084 struct sk_buff *buff;
1085 struct flowi fl;
1086 int tot_len = sizeof(struct tcphdr);
e69a4adc 1087 __be32 *topt;
cfb6eeb4
YH
1088#ifdef CONFIG_TCP_MD5SIG
1089 struct tcp_md5sig_key *key;
1090 struct tcp_md5sig_key tw_key;
1091#endif
1092
1093#ifdef CONFIG_TCP_MD5SIG
1094 if (!tw && skb->sk) {
0660e03f 1095 key = tcp_v6_md5_do_lookup(skb->sk, &ipv6_hdr(skb)->daddr);
cfb6eeb4
YH
1096 } else if (tw && tw->tw_md5_keylen) {
1097 tw_key.key = tw->tw_md5_key;
1098 tw_key.keylen = tw->tw_md5_keylen;
1099 key = &tw_key;
1100 } else {
1101 key = NULL;
1102 }
1103#endif
1da177e4
LT
1104
1105 if (ts)
4244f8a9 1106 tot_len += TCPOLEN_TSTAMP_ALIGNED;
cfb6eeb4
YH
1107#ifdef CONFIG_TCP_MD5SIG
1108 if (key)
1109 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1110#endif
1da177e4
LT
1111
1112 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1113 GFP_ATOMIC);
1114 if (buff == NULL)
1115 return;
1116
1117 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1118
1119 t1 = (struct tcphdr *) skb_push(buff,tot_len);
1120
1121 /* Swap the send and the receive. */
1122 memset(t1, 0, sizeof(*t1));
1123 t1->dest = th->source;
1124 t1->source = th->dest;
1125 t1->doff = tot_len/4;
1126 t1->seq = htonl(seq);
1127 t1->ack_seq = htonl(ack);
1128 t1->ack = 1;
1129 t1->window = htons(win);
cfb6eeb4 1130
e69a4adc 1131 topt = (__be32 *)(t1 + 1);
1ab1457c 1132
1da177e4 1133 if (ts) {
cfb6eeb4
YH
1134 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1135 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
1136 *topt++ = htonl(tcp_time_stamp);
1137 *topt = htonl(ts);
1da177e4
LT
1138 }
1139
cfb6eeb4
YH
1140#ifdef CONFIG_TCP_MD5SIG
1141 if (key) {
1142 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1143 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
0660e03f
ACM
1144 tcp_v6_do_calc_md5_hash((__u8 *)topt, key,
1145 &ipv6_hdr(skb)->daddr,
1146 &ipv6_hdr(skb)->saddr,
1147 t1, IPPROTO_TCP, tot_len);
cfb6eeb4
YH
1148 }
1149#endif
1150
1da177e4
LT
1151 buff->csum = csum_partial((char *)t1, tot_len, 0);
1152
1153 memset(&fl, 0, sizeof(fl));
0660e03f
ACM
1154 ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
1155 ipv6_addr_copy(&fl.fl6_src, &ipv6_hdr(skb)->daddr);
1da177e4
LT
1156
1157 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1158 tot_len, IPPROTO_TCP,
1159 buff->csum);
1160
1161 fl.proto = IPPROTO_TCP;
505cbfc5 1162 fl.oif = inet6_iif(skb);
1da177e4
LT
1163 fl.fl_ip_dport = t1->dest;
1164 fl.fl_ip_sport = t1->source;
beb8d13b 1165 security_skb_classify_flow(skb, &fl);
1da177e4
LT
1166
1167 if (!ip6_dst_lookup(NULL, &buff->dst, &fl)) {
ecc51b6d 1168 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
ae0f7d5f 1169 ip6_xmit(tcp6_socket->sk, buff, &fl, NULL, 0);
ecc51b6d 1170 TCP_INC_STATS_BH(TCP_MIB_OUTSEGS);
1da177e4 1171 return;
ecc51b6d 1172 }
1da177e4
LT
1173 }
1174
1175 kfree_skb(buff);
1176}
1177
1178static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
1179{
8feaf0c0 1180 struct inet_timewait_sock *tw = inet_twsk(sk);
cfb6eeb4 1181 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
1da177e4 1182
cfb6eeb4 1183 tcp_v6_send_ack(tcptw, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
8feaf0c0
ACM
1184 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
1185 tcptw->tw_ts_recent);
1da177e4 1186
8feaf0c0 1187 inet_twsk_put(tw);
1da177e4
LT
1188}
1189
60236fdd 1190static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req)
1da177e4 1191{
cfb6eeb4 1192 tcp_v6_send_ack(NULL, skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, req->ts_recent);
1da177e4
LT
1193}
1194
1195
1196static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
1197{
60236fdd 1198 struct request_sock *req, **prev;
aa8223c7 1199 const struct tcphdr *th = tcp_hdr(skb);
1da177e4
LT
1200 struct sock *nsk;
1201
1202 /* Find possible connection requests. */
8129765a 1203 req = inet6_csk_search_req(sk, &prev, th->source,
0660e03f
ACM
1204 &ipv6_hdr(skb)->saddr,
1205 &ipv6_hdr(skb)->daddr, inet6_iif(skb));
1da177e4
LT
1206 if (req)
1207 return tcp_check_req(sk, skb, req, prev);
1208
0660e03f
ACM
1209 nsk = __inet6_lookup_established(&tcp_hashinfo, &ipv6_hdr(skb)->saddr,
1210 th->source, &ipv6_hdr(skb)->daddr,
505cbfc5 1211 ntohs(th->dest), inet6_iif(skb));
1da177e4
LT
1212
1213 if (nsk) {
1214 if (nsk->sk_state != TCP_TIME_WAIT) {
1215 bh_lock_sock(nsk);
1216 return nsk;
1217 }
9469c7b4 1218 inet_twsk_put(inet_twsk(nsk));
1da177e4
LT
1219 return NULL;
1220 }
1221
1222#if 0 /*def CONFIG_SYN_COOKIES*/
1223 if (!th->rst && !th->syn && th->ack)
1224 sk = cookie_v6_check(sk, skb, &(IPCB(skb)->opt));
1225#endif
1226 return sk;
1227}
1228
1da177e4
LT
1229/* FIXME: this is substantially similar to the ipv4 code.
1230 * Can some kind of merge be done? -- erics
1231 */
1232static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
1233{
ca304b61 1234 struct inet6_request_sock *treq;
1da177e4
LT
1235 struct ipv6_pinfo *np = inet6_sk(sk);
1236 struct tcp_options_received tmp_opt;
1237 struct tcp_sock *tp = tcp_sk(sk);
60236fdd 1238 struct request_sock *req = NULL;
1da177e4
LT
1239 __u32 isn = TCP_SKB_CB(skb)->when;
1240
1241 if (skb->protocol == htons(ETH_P_IP))
1242 return tcp_v4_conn_request(sk, skb);
1243
1244 if (!ipv6_unicast_destination(skb))
1ab1457c 1245 goto drop;
1da177e4
LT
1246
1247 /*
1ab1457c 1248 * There are no SYN attacks on IPv6, yet...
1da177e4 1249 */
463c84b9 1250 if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
1da177e4
LT
1251 if (net_ratelimit())
1252 printk(KERN_INFO "TCPv6: dropping request, synflood is possible\n");
1ab1457c 1253 goto drop;
1da177e4
LT
1254 }
1255
463c84b9 1256 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
1da177e4
LT
1257 goto drop;
1258
ca304b61 1259 req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
1da177e4
LT
1260 if (req == NULL)
1261 goto drop;
1262
cfb6eeb4
YH
1263#ifdef CONFIG_TCP_MD5SIG
1264 tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
1265#endif
1266
1da177e4
LT
1267 tcp_clear_options(&tmp_opt);
1268 tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
1269 tmp_opt.user_mss = tp->rx_opt.user_mss;
1270
1271 tcp_parse_options(skb, &tmp_opt, 0);
1272
1273 tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
1274 tcp_openreq_init(req, &tmp_opt, skb);
1275
ca304b61 1276 treq = inet6_rsk(req);
0660e03f
ACM
1277 ipv6_addr_copy(&treq->rmt_addr, &ipv6_hdr(skb)->saddr);
1278 ipv6_addr_copy(&treq->loc_addr, &ipv6_hdr(skb)->daddr);
aa8223c7 1279 TCP_ECN_create_request(req, tcp_hdr(skb));
2e6599cb 1280 treq->pktopts = NULL;
1da177e4 1281 if (ipv6_opt_accepted(sk, skb) ||
333fad53
YH
1282 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
1283 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
1da177e4 1284 atomic_inc(&skb->users);
2e6599cb 1285 treq->pktopts = skb;
1da177e4 1286 }
2e6599cb 1287 treq->iif = sk->sk_bound_dev_if;
1da177e4
LT
1288
1289 /* So that link locals have meaning */
1290 if (!sk->sk_bound_dev_if &&
2e6599cb 1291 ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL)
505cbfc5 1292 treq->iif = inet6_iif(skb);
1da177e4 1293
1ab1457c 1294 if (isn == 0)
a94f723d 1295 isn = tcp_v6_init_sequence(skb);
1da177e4 1296
2e6599cb 1297 tcp_rsk(req)->snt_isn = isn;
1da177e4 1298
4237c75c
VY
1299 security_inet_conn_request(sk, skb, req);
1300
1da177e4
LT
1301 if (tcp_v6_send_synack(sk, req, NULL))
1302 goto drop;
1303
8129765a 1304 inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
1da177e4
LT
1305 return 0;
1306
1307drop:
1308 if (req)
60236fdd 1309 reqsk_free(req);
1da177e4 1310
1da177e4
LT
1311 return 0; /* don't send reset */
1312}
1313
1314static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
60236fdd 1315 struct request_sock *req,
1da177e4
LT
1316 struct dst_entry *dst)
1317{
ca304b61 1318 struct inet6_request_sock *treq = inet6_rsk(req);
1da177e4
LT
1319 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
1320 struct tcp6_sock *newtcp6sk;
1321 struct inet_sock *newinet;
1322 struct tcp_sock *newtp;
1323 struct sock *newsk;
1324 struct ipv6_txoptions *opt;
cfb6eeb4
YH
1325#ifdef CONFIG_TCP_MD5SIG
1326 struct tcp_md5sig_key *key;
1327#endif
1da177e4
LT
1328
1329 if (skb->protocol == htons(ETH_P_IP)) {
1330 /*
1331 * v6 mapped
1332 */
1333
1334 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1335
1ab1457c 1336 if (newsk == NULL)
1da177e4
LT
1337 return NULL;
1338
1339 newtcp6sk = (struct tcp6_sock *)newsk;
1340 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1341
1342 newinet = inet_sk(newsk);
1343 newnp = inet6_sk(newsk);
1344 newtp = tcp_sk(newsk);
1345
1346 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1347
1348 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
1349 newinet->daddr);
1350
1351 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
1352 newinet->saddr);
1353
1354 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
1355
8292a17a 1356 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1da177e4 1357 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
cfb6eeb4
YH
1358#ifdef CONFIG_TCP_MD5SIG
1359 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1360#endif
1361
1da177e4
LT
1362 newnp->pktoptions = NULL;
1363 newnp->opt = NULL;
505cbfc5 1364 newnp->mcast_oif = inet6_iif(skb);
0660e03f 1365 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1da177e4 1366
e6848976
ACM
1367 /*
1368 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1369 * here, tcp_create_openreq_child now does this for us, see the comment in
1370 * that function for the gory details. -acme
1da177e4 1371 */
1da177e4
LT
1372
1373 /* It is tricky place. Until this moment IPv4 tcp
8292a17a 1374 worked with IPv6 icsk.icsk_af_ops.
1da177e4
LT
1375 Sync it now.
1376 */
d83d8461 1377 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1da177e4
LT
1378
1379 return newsk;
1380 }
1381
1382 opt = np->opt;
1383
1384 if (sk_acceptq_is_full(sk))
1385 goto out_overflow;
1386
1da177e4
LT
1387 if (dst == NULL) {
1388 struct in6_addr *final_p = NULL, final;
1389 struct flowi fl;
1390
1391 memset(&fl, 0, sizeof(fl));
1392 fl.proto = IPPROTO_TCP;
2e6599cb 1393 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
1da177e4
LT
1394 if (opt && opt->srcrt) {
1395 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
1396 ipv6_addr_copy(&final, &fl.fl6_dst);
1397 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
1398 final_p = &final;
1399 }
2e6599cb 1400 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
1da177e4 1401 fl.oif = sk->sk_bound_dev_if;
2e6599cb 1402 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
1da177e4 1403 fl.fl_ip_sport = inet_sk(sk)->sport;
4237c75c 1404 security_req_classify_flow(req, &fl);
1da177e4
LT
1405
1406 if (ip6_dst_lookup(sk, &dst, &fl))
1407 goto out;
1408
1409 if (final_p)
1410 ipv6_addr_copy(&fl.fl6_dst, final_p);
1411
1412 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
1413 goto out;
1ab1457c 1414 }
1da177e4
LT
1415
1416 newsk = tcp_create_openreq_child(sk, req, skb);
1417 if (newsk == NULL)
1418 goto out;
1419
e6848976
ACM
1420 /*
1421 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1422 * count here, tcp_create_openreq_child now does this for us, see the
1423 * comment in that function for the gory details. -acme
1424 */
1da177e4 1425
59eed279 1426 newsk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 1427 __ip6_dst_store(newsk, dst, NULL, NULL);
1da177e4
LT
1428
1429 newtcp6sk = (struct tcp6_sock *)newsk;
1430 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1431
1432 newtp = tcp_sk(newsk);
1433 newinet = inet_sk(newsk);
1434 newnp = inet6_sk(newsk);
1435
1436 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1437
2e6599cb
ACM
1438 ipv6_addr_copy(&newnp->daddr, &treq->rmt_addr);
1439 ipv6_addr_copy(&newnp->saddr, &treq->loc_addr);
1440 ipv6_addr_copy(&newnp->rcv_saddr, &treq->loc_addr);
1441 newsk->sk_bound_dev_if = treq->iif;
1da177e4 1442
1ab1457c 1443 /* Now IPv6 options...
1da177e4
LT
1444
1445 First: no IPv4 options.
1446 */
1447 newinet->opt = NULL;
d35690be 1448 newnp->ipv6_fl_list = NULL;
1da177e4
LT
1449
1450 /* Clone RX bits */
1451 newnp->rxopt.all = np->rxopt.all;
1452
1453 /* Clone pktoptions received with SYN */
1454 newnp->pktoptions = NULL;
2e6599cb
ACM
1455 if (treq->pktopts != NULL) {
1456 newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC);
1457 kfree_skb(treq->pktopts);
1458 treq->pktopts = NULL;
1da177e4
LT
1459 if (newnp->pktoptions)
1460 skb_set_owner_r(newnp->pktoptions, newsk);
1461 }
1462 newnp->opt = NULL;
505cbfc5 1463 newnp->mcast_oif = inet6_iif(skb);
0660e03f 1464 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1da177e4
LT
1465
1466 /* Clone native IPv6 options from listening socket (if any)
1467
1468 Yes, keeping reference count would be much more clever,
1469 but we make one more one thing there: reattach optmem
1470 to newsk.
1471 */
1472 if (opt) {
1473 newnp->opt = ipv6_dup_options(newsk, opt);
1474 if (opt != np->opt)
1475 sock_kfree_s(sk, opt, opt->tot_len);
1476 }
1477
d83d8461 1478 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1da177e4 1479 if (newnp->opt)
d83d8461
ACM
1480 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1481 newnp->opt->opt_flen);
1da177e4 1482
5d424d5a 1483 tcp_mtup_init(newsk);
1da177e4
LT
1484 tcp_sync_mss(newsk, dst_mtu(dst));
1485 newtp->advmss = dst_metric(dst, RTAX_ADVMSS);
1486 tcp_initialize_rcv_mss(newsk);
1487
1488 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
1489
cfb6eeb4
YH
1490#ifdef CONFIG_TCP_MD5SIG
1491 /* Copy over the MD5 key from the original socket */
1492 if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) {
1493 /* We're using one, so create a matching key
1494 * on the newsk structure. If we fail to get
1495 * memory, then we end up not copying the key
1496 * across. Shucks.
1497 */
af879cc7
ACM
1498 char *newkey = kmemdup(key->key, key->keylen, GFP_ATOMIC);
1499 if (newkey != NULL)
cfb6eeb4
YH
1500 tcp_v6_md5_do_add(newsk, &inet6_sk(sk)->daddr,
1501 newkey, key->keylen);
cfb6eeb4
YH
1502 }
1503#endif
1504
90b19d31 1505 __inet6_hash(&tcp_hashinfo, newsk);
2d8c4ce5 1506 inet_inherit_port(&tcp_hashinfo, sk, newsk);
1da177e4
LT
1507
1508 return newsk;
1509
1510out_overflow:
1511 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
1512out:
1513 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
1514 if (opt && opt != np->opt)
1515 sock_kfree_s(sk, opt, opt->tot_len);
1516 dst_release(dst);
1517 return NULL;
1518}
1519
b51655b9 1520static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
1da177e4 1521{
84fa7933 1522 if (skb->ip_summed == CHECKSUM_COMPLETE) {
aa8223c7 1523 if (!tcp_v6_check(tcp_hdr(skb), skb->len, &ipv6_hdr(skb)->saddr,
0660e03f 1524 &ipv6_hdr(skb)->daddr, skb->csum)) {
fb286bb2 1525 skb->ip_summed = CHECKSUM_UNNECESSARY;
1da177e4 1526 return 0;
fb286bb2 1527 }
1da177e4 1528 }
fb286bb2 1529
aa8223c7 1530 skb->csum = ~csum_unfold(tcp_v6_check(tcp_hdr(skb), skb->len,
0660e03f
ACM
1531 &ipv6_hdr(skb)->saddr,
1532 &ipv6_hdr(skb)->daddr, 0));
fb286bb2 1533
1da177e4 1534 if (skb->len <= 76) {
fb286bb2 1535 return __skb_checksum_complete(skb);
1da177e4
LT
1536 }
1537 return 0;
1538}
1539
1540/* The socket must have it's spinlock held when we get
1541 * here.
1542 *
1543 * We have a potential double-lock case here, so even when
1544 * doing backlog processing we use the BH locking scheme.
1545 * This is because we cannot sleep with the original spinlock
1546 * held.
1547 */
1548static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1549{
1550 struct ipv6_pinfo *np = inet6_sk(sk);
1551 struct tcp_sock *tp;
1552 struct sk_buff *opt_skb = NULL;
1553
1554 /* Imagine: socket is IPv6. IPv4 packet arrives,
1555 goes to IPv4 receive handler and backlogged.
1556 From backlog it always goes here. Kerboom...
1557 Fortunately, tcp_rcv_established and rcv_established
1558 handle them correctly, but it is not case with
1559 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1560 */
1561
1562 if (skb->protocol == htons(ETH_P_IP))
1563 return tcp_v4_do_rcv(sk, skb);
1564
cfb6eeb4
YH
1565#ifdef CONFIG_TCP_MD5SIG
1566 if (tcp_v6_inbound_md5_hash (sk, skb))
1567 goto discard;
1568#endif
1569
fda9ef5d 1570 if (sk_filter(sk, skb))
1da177e4
LT
1571 goto discard;
1572
1573 /*
1574 * socket locking is here for SMP purposes as backlog rcv
1575 * is currently called with bh processing disabled.
1576 */
1577
1578 /* Do Stevens' IPV6_PKTOPTIONS.
1579
1580 Yes, guys, it is the only place in our code, where we
1581 may make it not affecting IPv4.
1582 The rest of code is protocol independent,
1583 and I do not like idea to uglify IPv4.
1584
1585 Actually, all the idea behind IPV6_PKTOPTIONS
1586 looks not very well thought. For now we latch
1587 options, received in the last packet, enqueued
1588 by tcp. Feel free to propose better solution.
1ab1457c 1589 --ANK (980728)
1da177e4
LT
1590 */
1591 if (np->rxopt.all)
1592 opt_skb = skb_clone(skb, GFP_ATOMIC);
1593
1594 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1595 TCP_CHECK_TIMER(sk);
aa8223c7 1596 if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len))
1da177e4
LT
1597 goto reset;
1598 TCP_CHECK_TIMER(sk);
1599 if (opt_skb)
1600 goto ipv6_pktoptions;
1601 return 0;
1602 }
1603
ab6a5bb6 1604 if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
1da177e4
LT
1605 goto csum_err;
1606
1ab1457c 1607 if (sk->sk_state == TCP_LISTEN) {
1da177e4
LT
1608 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1609 if (!nsk)
1610 goto discard;
1611
1612 /*
1613 * Queue it on the new socket if the new socket is active,
1614 * otherwise we just shortcircuit this and continue with
1615 * the new socket..
1616 */
1ab1457c 1617 if(nsk != sk) {
1da177e4
LT
1618 if (tcp_child_process(sk, nsk, skb))
1619 goto reset;
1620 if (opt_skb)
1621 __kfree_skb(opt_skb);
1622 return 0;
1623 }
1624 }
1625
1626 TCP_CHECK_TIMER(sk);
aa8223c7 1627 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
1da177e4
LT
1628 goto reset;
1629 TCP_CHECK_TIMER(sk);
1630 if (opt_skb)
1631 goto ipv6_pktoptions;
1632 return 0;
1633
1634reset:
cfb6eeb4 1635 tcp_v6_send_reset(sk, skb);
1da177e4
LT
1636discard:
1637 if (opt_skb)
1638 __kfree_skb(opt_skb);
1639 kfree_skb(skb);
1640 return 0;
1641csum_err:
1642 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1643 goto discard;
1644
1645
1646ipv6_pktoptions:
1647 /* Do you ask, what is it?
1648
1649 1. skb was enqueued by tcp.
1650 2. skb is added to tail of read queue, rather than out of order.
1651 3. socket is not in passive state.
1652 4. Finally, it really contains options, which user wants to receive.
1653 */
1654 tp = tcp_sk(sk);
1655 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1656 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
333fad53 1657 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
505cbfc5 1658 np->mcast_oif = inet6_iif(opt_skb);
333fad53 1659 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
0660e03f 1660 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1da177e4
LT
1661 if (ipv6_opt_accepted(sk, opt_skb)) {
1662 skb_set_owner_r(opt_skb, sk);
1663 opt_skb = xchg(&np->pktoptions, opt_skb);
1664 } else {
1665 __kfree_skb(opt_skb);
1666 opt_skb = xchg(&np->pktoptions, NULL);
1667 }
1668 }
1669
1670 if (opt_skb)
1671 kfree_skb(opt_skb);
1672 return 0;
1673}
1674
e5bbef20 1675static int tcp_v6_rcv(struct sk_buff *skb)
1da177e4 1676{
1ab1457c 1677 struct tcphdr *th;
1da177e4
LT
1678 struct sock *sk;
1679 int ret;
1680
1681 if (skb->pkt_type != PACKET_HOST)
1682 goto discard_it;
1683
1684 /*
1685 * Count it even if it's bad.
1686 */
1687 TCP_INC_STATS_BH(TCP_MIB_INSEGS);
1688
1689 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1690 goto discard_it;
1691
aa8223c7 1692 th = tcp_hdr(skb);
1da177e4
LT
1693
1694 if (th->doff < sizeof(struct tcphdr)/4)
1695 goto bad_packet;
1696 if (!pskb_may_pull(skb, th->doff*4))
1697 goto discard_it;
1698
60476372 1699 if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
1da177e4
LT
1700 goto bad_packet;
1701
aa8223c7 1702 th = tcp_hdr(skb);
1da177e4
LT
1703 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1704 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1705 skb->len - th->doff*4);
1706 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1707 TCP_SKB_CB(skb)->when = 0;
0660e03f 1708 TCP_SKB_CB(skb)->flags = ipv6_get_dsfield(ipv6_hdr(skb));
1da177e4
LT
1709 TCP_SKB_CB(skb)->sacked = 0;
1710
0660e03f
ACM
1711 sk = __inet6_lookup(&tcp_hashinfo, &ipv6_hdr(skb)->saddr, th->source,
1712 &ipv6_hdr(skb)->daddr, ntohs(th->dest),
505cbfc5 1713 inet6_iif(skb));
1da177e4
LT
1714
1715 if (!sk)
1716 goto no_tcp_socket;
1717
1718process:
1719 if (sk->sk_state == TCP_TIME_WAIT)
1720 goto do_time_wait;
1721
1722 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1723 goto discard_and_relse;
1724
fda9ef5d 1725 if (sk_filter(sk, skb))
1da177e4
LT
1726 goto discard_and_relse;
1727
1728 skb->dev = NULL;
1729
293b9c42 1730 bh_lock_sock_nested(sk);
1da177e4
LT
1731 ret = 0;
1732 if (!sock_owned_by_user(sk)) {
1a2449a8 1733#ifdef CONFIG_NET_DMA
1ab1457c
YH
1734 struct tcp_sock *tp = tcp_sk(sk);
1735 if (tp->ucopy.dma_chan)
1736 ret = tcp_v6_do_rcv(sk, skb);
1737 else
1a2449a8
CL
1738#endif
1739 {
1740 if (!tcp_prequeue(sk, skb))
1741 ret = tcp_v6_do_rcv(sk, skb);
1742 }
1da177e4
LT
1743 } else
1744 sk_add_backlog(sk, skb);
1745 bh_unlock_sock(sk);
1746
1747 sock_put(sk);
1748 return ret ? -1 : 0;
1749
1750no_tcp_socket:
1751 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1752 goto discard_it;
1753
1754 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1755bad_packet:
1756 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1757 } else {
cfb6eeb4 1758 tcp_v6_send_reset(NULL, skb);
1da177e4
LT
1759 }
1760
1761discard_it:
1762
1763 /*
1764 * Discard frame
1765 */
1766
1767 kfree_skb(skb);
1768 return 0;
1769
1770discard_and_relse:
1771 sock_put(sk);
1772 goto discard_it;
1773
1774do_time_wait:
1775 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
9469c7b4 1776 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1777 goto discard_it;
1778 }
1779
1780 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1781 TCP_INC_STATS_BH(TCP_MIB_INERRS);
9469c7b4 1782 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1783 goto discard_it;
1784 }
1785
9469c7b4 1786 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1da177e4
LT
1787 case TCP_TW_SYN:
1788 {
1789 struct sock *sk2;
1790
505cbfc5 1791 sk2 = inet6_lookup_listener(&tcp_hashinfo,
0660e03f 1792 &ipv6_hdr(skb)->daddr,
505cbfc5 1793 ntohs(th->dest), inet6_iif(skb));
1da177e4 1794 if (sk2 != NULL) {
295ff7ed
ACM
1795 struct inet_timewait_sock *tw = inet_twsk(sk);
1796 inet_twsk_deschedule(tw, &tcp_death_row);
1797 inet_twsk_put(tw);
1da177e4
LT
1798 sk = sk2;
1799 goto process;
1800 }
1801 /* Fall through to ACK */
1802 }
1803 case TCP_TW_ACK:
1804 tcp_v6_timewait_ack(sk, skb);
1805 break;
1806 case TCP_TW_RST:
1807 goto no_tcp_socket;
1808 case TCP_TW_SUCCESS:;
1809 }
1810 goto discard_it;
1811}
1812
1da177e4
LT
1813static int tcp_v6_remember_stamp(struct sock *sk)
1814{
1815 /* Alas, not yet... */
1816 return 0;
1817}
1818
8292a17a 1819static struct inet_connection_sock_af_ops ipv6_specific = {
543d9cfe
ACM
1820 .queue_xmit = inet6_csk_xmit,
1821 .send_check = tcp_v6_send_check,
1822 .rebuild_header = inet6_sk_rebuild_header,
1823 .conn_request = tcp_v6_conn_request,
1824 .syn_recv_sock = tcp_v6_syn_recv_sock,
1825 .remember_stamp = tcp_v6_remember_stamp,
1826 .net_header_len = sizeof(struct ipv6hdr),
1827 .setsockopt = ipv6_setsockopt,
1828 .getsockopt = ipv6_getsockopt,
1829 .addr2sockaddr = inet6_csk_addr2sockaddr,
1830 .sockaddr_len = sizeof(struct sockaddr_in6),
3fdadf7d 1831#ifdef CONFIG_COMPAT
543d9cfe
ACM
1832 .compat_setsockopt = compat_ipv6_setsockopt,
1833 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1834#endif
1da177e4
LT
1835};
1836
cfb6eeb4 1837#ifdef CONFIG_TCP_MD5SIG
a928630a 1838static struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
cfb6eeb4
YH
1839 .md5_lookup = tcp_v6_md5_lookup,
1840 .calc_md5_hash = tcp_v6_calc_md5_hash,
1841 .md5_add = tcp_v6_md5_add_func,
1842 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1843};
a928630a 1844#endif
cfb6eeb4 1845
1da177e4
LT
1846/*
1847 * TCP over IPv4 via INET6 API
1848 */
1849
8292a17a 1850static struct inet_connection_sock_af_ops ipv6_mapped = {
543d9cfe
ACM
1851 .queue_xmit = ip_queue_xmit,
1852 .send_check = tcp_v4_send_check,
1853 .rebuild_header = inet_sk_rebuild_header,
1854 .conn_request = tcp_v6_conn_request,
1855 .syn_recv_sock = tcp_v6_syn_recv_sock,
1856 .remember_stamp = tcp_v4_remember_stamp,
1857 .net_header_len = sizeof(struct iphdr),
1858 .setsockopt = ipv6_setsockopt,
1859 .getsockopt = ipv6_getsockopt,
1860 .addr2sockaddr = inet6_csk_addr2sockaddr,
1861 .sockaddr_len = sizeof(struct sockaddr_in6),
3fdadf7d 1862#ifdef CONFIG_COMPAT
543d9cfe
ACM
1863 .compat_setsockopt = compat_ipv6_setsockopt,
1864 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1865#endif
1da177e4
LT
1866};
1867
cfb6eeb4 1868#ifdef CONFIG_TCP_MD5SIG
a928630a 1869static struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
cfb6eeb4
YH
1870 .md5_lookup = tcp_v4_md5_lookup,
1871 .calc_md5_hash = tcp_v4_calc_md5_hash,
1872 .md5_add = tcp_v6_md5_add_func,
1873 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1874};
a928630a 1875#endif
cfb6eeb4 1876
1da177e4
LT
1877/* NOTE: A lot of things set to zero explicitly by call to
1878 * sk_alloc() so need not be done here.
1879 */
1880static int tcp_v6_init_sock(struct sock *sk)
1881{
6687e988 1882 struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4
LT
1883 struct tcp_sock *tp = tcp_sk(sk);
1884
1885 skb_queue_head_init(&tp->out_of_order_queue);
1886 tcp_init_xmit_timers(sk);
1887 tcp_prequeue_init(tp);
1888
6687e988 1889 icsk->icsk_rto = TCP_TIMEOUT_INIT;
1da177e4
LT
1890 tp->mdev = TCP_TIMEOUT_INIT;
1891
1892 /* So many TCP implementations out there (incorrectly) count the
1893 * initial SYN frame in their delayed-ACK and congestion control
1894 * algorithms that we must have the following bandaid to talk
1895 * efficiently to them. -DaveM
1896 */
1897 tp->snd_cwnd = 2;
1898
1899 /* See draft-stevens-tcpca-spec-01 for discussion of the
1900 * initialization of these values.
1901 */
1902 tp->snd_ssthresh = 0x7fffffff;
1903 tp->snd_cwnd_clamp = ~0;
c1b4a7e6 1904 tp->mss_cache = 536;
1da177e4
LT
1905
1906 tp->reordering = sysctl_tcp_reordering;
1907
1908 sk->sk_state = TCP_CLOSE;
1909
8292a17a 1910 icsk->icsk_af_ops = &ipv6_specific;
6687e988 1911 icsk->icsk_ca_ops = &tcp_init_congestion_ops;
d83d8461 1912 icsk->icsk_sync_mss = tcp_sync_mss;
1da177e4
LT
1913 sk->sk_write_space = sk_stream_write_space;
1914 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
1915
cfb6eeb4
YH
1916#ifdef CONFIG_TCP_MD5SIG
1917 tp->af_specific = &tcp_sock_ipv6_specific;
1918#endif
1919
1da177e4
LT
1920 sk->sk_sndbuf = sysctl_tcp_wmem[1];
1921 sk->sk_rcvbuf = sysctl_tcp_rmem[1];
1922
1923 atomic_inc(&tcp_sockets_allocated);
1924
1925 return 0;
1926}
1927
1928static int tcp_v6_destroy_sock(struct sock *sk)
1929{
cfb6eeb4
YH
1930#ifdef CONFIG_TCP_MD5SIG
1931 /* Clean up the MD5 key list */
1932 if (tcp_sk(sk)->md5sig_info)
1933 tcp_v6_clear_md5_list(sk);
1934#endif
1da177e4
LT
1935 tcp_v4_destroy_sock(sk);
1936 return inet6_destroy_sock(sk);
1937}
1938
952a10be 1939#ifdef CONFIG_PROC_FS
1da177e4 1940/* Proc filesystem TCPv6 sock list dumping. */
1ab1457c 1941static void get_openreq6(struct seq_file *seq,
60236fdd 1942 struct sock *sk, struct request_sock *req, int i, int uid)
1da177e4 1943{
1da177e4 1944 int ttd = req->expires - jiffies;
ca304b61
ACM
1945 struct in6_addr *src = &inet6_rsk(req)->loc_addr;
1946 struct in6_addr *dest = &inet6_rsk(req)->rmt_addr;
1da177e4
LT
1947
1948 if (ttd < 0)
1949 ttd = 0;
1950
1da177e4
LT
1951 seq_printf(seq,
1952 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1953 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
1954 i,
1955 src->s6_addr32[0], src->s6_addr32[1],
1956 src->s6_addr32[2], src->s6_addr32[3],
1957 ntohs(inet_sk(sk)->sport),
1958 dest->s6_addr32[0], dest->s6_addr32[1],
1959 dest->s6_addr32[2], dest->s6_addr32[3],
2e6599cb 1960 ntohs(inet_rsk(req)->rmt_port),
1da177e4
LT
1961 TCP_SYN_RECV,
1962 0,0, /* could print option size, but that is af dependent. */
1ab1457c
YH
1963 1, /* timers active (only the expire timer) */
1964 jiffies_to_clock_t(ttd),
1da177e4
LT
1965 req->retrans,
1966 uid,
1ab1457c 1967 0, /* non standard timer */
1da177e4
LT
1968 0, /* open_requests have no inode */
1969 0, req);
1970}
1971
1972static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1973{
1974 struct in6_addr *dest, *src;
1975 __u16 destp, srcp;
1976 int timer_active;
1977 unsigned long timer_expires;
1978 struct inet_sock *inet = inet_sk(sp);
1979 struct tcp_sock *tp = tcp_sk(sp);
463c84b9 1980 const struct inet_connection_sock *icsk = inet_csk(sp);
1da177e4
LT
1981 struct ipv6_pinfo *np = inet6_sk(sp);
1982
1983 dest = &np->daddr;
1984 src = &np->rcv_saddr;
1985 destp = ntohs(inet->dport);
1986 srcp = ntohs(inet->sport);
463c84b9
ACM
1987
1988 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
1da177e4 1989 timer_active = 1;
463c84b9
ACM
1990 timer_expires = icsk->icsk_timeout;
1991 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
1da177e4 1992 timer_active = 4;
463c84b9 1993 timer_expires = icsk->icsk_timeout;
1da177e4
LT
1994 } else if (timer_pending(&sp->sk_timer)) {
1995 timer_active = 2;
1996 timer_expires = sp->sk_timer.expires;
1997 } else {
1998 timer_active = 0;
1999 timer_expires = jiffies;
2000 }
2001
2002 seq_printf(seq,
2003 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2004 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %u %u %u %u %d\n",
2005 i,
2006 src->s6_addr32[0], src->s6_addr32[1],
2007 src->s6_addr32[2], src->s6_addr32[3], srcp,
2008 dest->s6_addr32[0], dest->s6_addr32[1],
2009 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1ab1457c 2010 sp->sk_state,
47da8ee6
SS
2011 tp->write_seq-tp->snd_una,
2012 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
1da177e4
LT
2013 timer_active,
2014 jiffies_to_clock_t(timer_expires - jiffies),
463c84b9 2015 icsk->icsk_retransmits,
1da177e4 2016 sock_i_uid(sp),
6687e988 2017 icsk->icsk_probes_out,
1da177e4
LT
2018 sock_i_ino(sp),
2019 atomic_read(&sp->sk_refcnt), sp,
463c84b9
ACM
2020 icsk->icsk_rto,
2021 icsk->icsk_ack.ato,
2022 (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
1da177e4
LT
2023 tp->snd_cwnd, tp->snd_ssthresh>=0xFFFF?-1:tp->snd_ssthresh
2024 );
2025}
2026
1ab1457c 2027static void get_timewait6_sock(struct seq_file *seq,
8feaf0c0 2028 struct inet_timewait_sock *tw, int i)
1da177e4
LT
2029{
2030 struct in6_addr *dest, *src;
2031 __u16 destp, srcp;
0fa1a53e 2032 struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
1da177e4
LT
2033 int ttd = tw->tw_ttd - jiffies;
2034
2035 if (ttd < 0)
2036 ttd = 0;
2037
0fa1a53e
ACM
2038 dest = &tw6->tw_v6_daddr;
2039 src = &tw6->tw_v6_rcv_saddr;
1da177e4
LT
2040 destp = ntohs(tw->tw_dport);
2041 srcp = ntohs(tw->tw_sport);
2042
2043 seq_printf(seq,
2044 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2045 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
2046 i,
2047 src->s6_addr32[0], src->s6_addr32[1],
2048 src->s6_addr32[2], src->s6_addr32[3], srcp,
2049 dest->s6_addr32[0], dest->s6_addr32[1],
2050 dest->s6_addr32[2], dest->s6_addr32[3], destp,
2051 tw->tw_substate, 0, 0,
2052 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
2053 atomic_read(&tw->tw_refcnt), tw);
2054}
2055
1da177e4
LT
2056static int tcp6_seq_show(struct seq_file *seq, void *v)
2057{
2058 struct tcp_iter_state *st;
2059
2060 if (v == SEQ_START_TOKEN) {
2061 seq_puts(seq,
2062 " sl "
2063 "local_address "
2064 "remote_address "
2065 "st tx_queue rx_queue tr tm->when retrnsmt"
2066 " uid timeout inode\n");
2067 goto out;
2068 }
2069 st = seq->private;
2070
2071 switch (st->state) {
2072 case TCP_SEQ_STATE_LISTENING:
2073 case TCP_SEQ_STATE_ESTABLISHED:
2074 get_tcp6_sock(seq, v, st->num);
2075 break;
2076 case TCP_SEQ_STATE_OPENREQ:
2077 get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
2078 break;
2079 case TCP_SEQ_STATE_TIME_WAIT:
2080 get_timewait6_sock(seq, v, st->num);
2081 break;
2082 }
2083out:
2084 return 0;
2085}
2086
2087static struct file_operations tcp6_seq_fops;
2088static struct tcp_seq_afinfo tcp6_seq_afinfo = {
2089 .owner = THIS_MODULE,
2090 .name = "tcp6",
2091 .family = AF_INET6,
2092 .seq_show = tcp6_seq_show,
2093 .seq_fops = &tcp6_seq_fops,
2094};
2095
2096int __init tcp6_proc_init(void)
2097{
2098 return tcp_proc_register(&tcp6_seq_afinfo);
2099}
2100
2101void tcp6_proc_exit(void)
2102{
2103 tcp_proc_unregister(&tcp6_seq_afinfo);
2104}
2105#endif
2106
2107struct proto tcpv6_prot = {
2108 .name = "TCPv6",
2109 .owner = THIS_MODULE,
2110 .close = tcp_close,
2111 .connect = tcp_v6_connect,
2112 .disconnect = tcp_disconnect,
463c84b9 2113 .accept = inet_csk_accept,
1da177e4
LT
2114 .ioctl = tcp_ioctl,
2115 .init = tcp_v6_init_sock,
2116 .destroy = tcp_v6_destroy_sock,
2117 .shutdown = tcp_shutdown,
2118 .setsockopt = tcp_setsockopt,
2119 .getsockopt = tcp_getsockopt,
1da177e4
LT
2120 .recvmsg = tcp_recvmsg,
2121 .backlog_rcv = tcp_v6_do_rcv,
2122 .hash = tcp_v6_hash,
2123 .unhash = tcp_unhash,
2124 .get_port = tcp_v6_get_port,
2125 .enter_memory_pressure = tcp_enter_memory_pressure,
2126 .sockets_allocated = &tcp_sockets_allocated,
2127 .memory_allocated = &tcp_memory_allocated,
2128 .memory_pressure = &tcp_memory_pressure,
0a5578cf 2129 .orphan_count = &tcp_orphan_count,
1da177e4
LT
2130 .sysctl_mem = sysctl_tcp_mem,
2131 .sysctl_wmem = sysctl_tcp_wmem,
2132 .sysctl_rmem = sysctl_tcp_rmem,
2133 .max_header = MAX_TCP_HEADER,
2134 .obj_size = sizeof(struct tcp6_sock),
6d6ee43e 2135 .twsk_prot = &tcp6_timewait_sock_ops,
60236fdd 2136 .rsk_prot = &tcp6_request_sock_ops,
543d9cfe
ACM
2137#ifdef CONFIG_COMPAT
2138 .compat_setsockopt = compat_tcp_setsockopt,
2139 .compat_getsockopt = compat_tcp_getsockopt,
2140#endif
1da177e4
LT
2141};
2142
2143static struct inet6_protocol tcpv6_protocol = {
2144 .handler = tcp_v6_rcv,
2145 .err_handler = tcp_v6_err,
a430a43d 2146 .gso_send_check = tcp_v6_gso_send_check,
adcfc7d0 2147 .gso_segment = tcp_tso_segment,
1da177e4
LT
2148 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
2149};
2150
1da177e4
LT
2151static struct inet_protosw tcpv6_protosw = {
2152 .type = SOCK_STREAM,
2153 .protocol = IPPROTO_TCP,
2154 .prot = &tcpv6_prot,
2155 .ops = &inet6_stream_ops,
2156 .capability = -1,
2157 .no_check = 0,
d83d8461
ACM
2158 .flags = INET_PROTOSW_PERMANENT |
2159 INET_PROTOSW_ICSK,
1da177e4
LT
2160};
2161
2162void __init tcpv6_init(void)
2163{
2164 /* register inet6 protocol */
2165 if (inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP) < 0)
2166 printk(KERN_ERR "tcpv6_init: Could not register protocol\n");
2167 inet6_register_protosw(&tcpv6_protosw);
ae0f7d5f 2168
c4d93909
ACM
2169 if (inet_csk_ctl_sock_create(&tcp6_socket, PF_INET6, SOCK_RAW,
2170 IPPROTO_TCP) < 0)
ae0f7d5f 2171 panic("Failed to create the TCPv6 control socket.\n");
1da177e4 2172}