]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - net/ipv6/tcp_ipv6.c
[NETFILTER]: ip_nat_snmp_basic annotations.
[mirror_ubuntu-bionic-kernel.git] / net / ipv6 / tcp_ipv6.c
CommitLineData
1da177e4
LT
1/*
2 * TCP over IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * $Id: tcp_ipv6.c,v 1.144 2002/02/01 22:01:04 davem Exp $
9 *
10 * Based on:
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>
59#include <net/addrconf.h>
60#include <net/snmp.h>
61#include <net/dsfield.h>
6d6ee43e 62#include <net/timewait_sock.h>
1da177e4
LT
63
64#include <asm/uaccess.h>
65
66#include <linux/proc_fs.h>
67#include <linux/seq_file.h>
68
cfb6eeb4
YH
69#include <linux/crypto.h>
70#include <linux/scatterlist.h>
71
ae0f7d5f
DW
72/* Socket used for sending RSTs and ACKs */
73static struct socket *tcp6_socket;
74
cfb6eeb4 75static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb);
60236fdd 76static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req);
8292a17a 77static void tcp_v6_send_check(struct sock *sk, int len,
1da177e4
LT
78 struct sk_buff *skb);
79
80static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
1da177e4 81
8292a17a
ACM
82static struct inet_connection_sock_af_ops ipv6_mapped;
83static struct inet_connection_sock_af_ops ipv6_specific;
a928630a 84#ifdef CONFIG_TCP_MD5SIG
cfb6eeb4
YH
85static struct tcp_sock_af_ops tcp_sock_ipv6_specific;
86static struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
a928630a 87#endif
1da177e4 88
1da177e4
LT
89static int tcp_v6_get_port(struct sock *sk, unsigned short snum)
90{
971af18b
ACM
91 return inet_csk_get_port(&tcp_hashinfo, sk, snum,
92 inet6_csk_bind_conflict);
1da177e4
LT
93}
94
1da177e4
LT
95static void tcp_v6_hash(struct sock *sk)
96{
97 if (sk->sk_state != TCP_CLOSE) {
8292a17a 98 if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
1da177e4
LT
99 tcp_prot.hash(sk);
100 return;
101 }
102 local_bh_disable();
90b19d31 103 __inet6_hash(&tcp_hashinfo, sk);
1da177e4
LT
104 local_bh_enable();
105 }
106}
107
868c86bc 108static __inline__ __sum16 tcp_v6_check(struct tcphdr *th, int len,
1da177e4
LT
109 struct in6_addr *saddr,
110 struct in6_addr *daddr,
868c86bc 111 __wsum base)
1da177e4
LT
112{
113 return csum_ipv6_magic(saddr, daddr, len, IPPROTO_TCP, base);
114}
115
a94f723d 116static __u32 tcp_v6_init_sequence(struct sk_buff *skb)
1da177e4 117{
a94f723d
GR
118 return secure_tcpv6_sequence_number(skb->nh.ipv6h->daddr.s6_addr32,
119 skb->nh.ipv6h->saddr.s6_addr32,
120 skb->h.th->dest,
121 skb->h.th->source);
1da177e4
LT
122}
123
1da177e4
LT
124static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
125 int addr_len)
126{
127 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
d83d8461
ACM
128 struct inet_sock *inet = inet_sk(sk);
129 struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4
LT
130 struct ipv6_pinfo *np = inet6_sk(sk);
131 struct tcp_sock *tp = tcp_sk(sk);
132 struct in6_addr *saddr = NULL, *final_p = NULL, final;
133 struct flowi fl;
134 struct dst_entry *dst;
135 int addr_type;
136 int err;
137
138 if (addr_len < SIN6_LEN_RFC2133)
139 return -EINVAL;
140
141 if (usin->sin6_family != AF_INET6)
142 return(-EAFNOSUPPORT);
143
144 memset(&fl, 0, sizeof(fl));
145
146 if (np->sndflow) {
147 fl.fl6_flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
148 IP6_ECN_flow_init(fl.fl6_flowlabel);
149 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
150 struct ip6_flowlabel *flowlabel;
151 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
152 if (flowlabel == NULL)
153 return -EINVAL;
154 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
155 fl6_sock_release(flowlabel);
156 }
157 }
158
159 /*
160 * connect() to INADDR_ANY means loopback (BSD'ism).
161 */
162
163 if(ipv6_addr_any(&usin->sin6_addr))
164 usin->sin6_addr.s6_addr[15] = 0x1;
165
166 addr_type = ipv6_addr_type(&usin->sin6_addr);
167
168 if(addr_type & IPV6_ADDR_MULTICAST)
169 return -ENETUNREACH;
170
171 if (addr_type&IPV6_ADDR_LINKLOCAL) {
172 if (addr_len >= sizeof(struct sockaddr_in6) &&
173 usin->sin6_scope_id) {
174 /* If interface is set while binding, indices
175 * must coincide.
176 */
177 if (sk->sk_bound_dev_if &&
178 sk->sk_bound_dev_if != usin->sin6_scope_id)
179 return -EINVAL;
180
181 sk->sk_bound_dev_if = usin->sin6_scope_id;
182 }
183
184 /* Connect to link-local address requires an interface */
185 if (!sk->sk_bound_dev_if)
186 return -EINVAL;
187 }
188
189 if (tp->rx_opt.ts_recent_stamp &&
190 !ipv6_addr_equal(&np->daddr, &usin->sin6_addr)) {
191 tp->rx_opt.ts_recent = 0;
192 tp->rx_opt.ts_recent_stamp = 0;
193 tp->write_seq = 0;
194 }
195
196 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
197 np->flow_label = fl.fl6_flowlabel;
198
199 /*
200 * TCP over IPv4
201 */
202
203 if (addr_type == IPV6_ADDR_MAPPED) {
d83d8461 204 u32 exthdrlen = icsk->icsk_ext_hdr_len;
1da177e4
LT
205 struct sockaddr_in sin;
206
207 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
208
209 if (__ipv6_only_sock(sk))
210 return -ENETUNREACH;
211
212 sin.sin_family = AF_INET;
213 sin.sin_port = usin->sin6_port;
214 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
215
d83d8461 216 icsk->icsk_af_ops = &ipv6_mapped;
1da177e4 217 sk->sk_backlog_rcv = tcp_v4_do_rcv;
cfb6eeb4
YH
218#ifdef CONFIG_TCP_MD5SIG
219 tp->af_specific = &tcp_sock_ipv6_mapped_specific;
220#endif
1da177e4
LT
221
222 err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
223
224 if (err) {
d83d8461
ACM
225 icsk->icsk_ext_hdr_len = exthdrlen;
226 icsk->icsk_af_ops = &ipv6_specific;
1da177e4 227 sk->sk_backlog_rcv = tcp_v6_do_rcv;
cfb6eeb4
YH
228#ifdef CONFIG_TCP_MD5SIG
229 tp->af_specific = &tcp_sock_ipv6_specific;
230#endif
1da177e4
LT
231 goto failure;
232 } else {
233 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
234 inet->saddr);
235 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
236 inet->rcv_saddr);
237 }
238
239 return err;
240 }
241
242 if (!ipv6_addr_any(&np->rcv_saddr))
243 saddr = &np->rcv_saddr;
244
245 fl.proto = IPPROTO_TCP;
246 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
247 ipv6_addr_copy(&fl.fl6_src,
248 (saddr ? saddr : &np->saddr));
249 fl.oif = sk->sk_bound_dev_if;
250 fl.fl_ip_dport = usin->sin6_port;
251 fl.fl_ip_sport = inet->sport;
252
253 if (np->opt && np->opt->srcrt) {
254 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
255 ipv6_addr_copy(&final, &fl.fl6_dst);
256 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
257 final_p = &final;
258 }
259
beb8d13b
VY
260 security_sk_classify_flow(sk, &fl);
261
1da177e4
LT
262 err = ip6_dst_lookup(sk, &dst, &fl);
263 if (err)
264 goto failure;
265 if (final_p)
266 ipv6_addr_copy(&fl.fl6_dst, final_p);
267
e104411b 268 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
1da177e4 269 goto failure;
1da177e4
LT
270
271 if (saddr == NULL) {
272 saddr = &fl.fl6_src;
273 ipv6_addr_copy(&np->rcv_saddr, saddr);
274 }
275
276 /* set the source address */
277 ipv6_addr_copy(&np->saddr, saddr);
278 inet->rcv_saddr = LOOPBACK4_IPV6;
279
f83ef8c0 280 sk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 281 __ip6_dst_store(sk, dst, NULL, NULL);
1da177e4 282
d83d8461 283 icsk->icsk_ext_hdr_len = 0;
1da177e4 284 if (np->opt)
d83d8461
ACM
285 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
286 np->opt->opt_nflen);
1da177e4
LT
287
288 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
289
290 inet->dport = usin->sin6_port;
291
292 tcp_set_state(sk, TCP_SYN_SENT);
d8313f5c 293 err = inet6_hash_connect(&tcp_death_row, sk);
1da177e4
LT
294 if (err)
295 goto late_failure;
296
297 if (!tp->write_seq)
298 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
299 np->daddr.s6_addr32,
300 inet->sport,
301 inet->dport);
302
303 err = tcp_connect(sk);
304 if (err)
305 goto late_failure;
306
307 return 0;
308
309late_failure:
310 tcp_set_state(sk, TCP_CLOSE);
311 __sk_dst_reset(sk);
312failure:
313 inet->dport = 0;
314 sk->sk_route_caps = 0;
315 return err;
316}
317
318static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
04ce6909 319 int type, int code, int offset, __be32 info)
1da177e4
LT
320{
321 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
505cbfc5 322 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
1da177e4
LT
323 struct ipv6_pinfo *np;
324 struct sock *sk;
325 int err;
326 struct tcp_sock *tp;
327 __u32 seq;
328
505cbfc5
ACM
329 sk = inet6_lookup(&tcp_hashinfo, &hdr->daddr, th->dest, &hdr->saddr,
330 th->source, skb->dev->ifindex);
1da177e4
LT
331
332 if (sk == NULL) {
333 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
334 return;
335 }
336
337 if (sk->sk_state == TCP_TIME_WAIT) {
9469c7b4 338 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
339 return;
340 }
341
342 bh_lock_sock(sk);
343 if (sock_owned_by_user(sk))
344 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
345
346 if (sk->sk_state == TCP_CLOSE)
347 goto out;
348
349 tp = tcp_sk(sk);
350 seq = ntohl(th->seq);
351 if (sk->sk_state != TCP_LISTEN &&
352 !between(seq, tp->snd_una, tp->snd_nxt)) {
353 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
354 goto out;
355 }
356
357 np = inet6_sk(sk);
358
359 if (type == ICMPV6_PKT_TOOBIG) {
360 struct dst_entry *dst = NULL;
361
362 if (sock_owned_by_user(sk))
363 goto out;
364 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
365 goto out;
366
367 /* icmp should have updated the destination cache entry */
368 dst = __sk_dst_check(sk, np->dst_cookie);
369
370 if (dst == NULL) {
371 struct inet_sock *inet = inet_sk(sk);
372 struct flowi fl;
373
374 /* BUGGG_FUTURE: Again, it is not clear how
375 to handle rthdr case. Ignore this complexity
376 for now.
377 */
378 memset(&fl, 0, sizeof(fl));
379 fl.proto = IPPROTO_TCP;
380 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
381 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
382 fl.oif = sk->sk_bound_dev_if;
383 fl.fl_ip_dport = inet->dport;
384 fl.fl_ip_sport = inet->sport;
beb8d13b 385 security_skb_classify_flow(skb, &fl);
1da177e4
LT
386
387 if ((err = ip6_dst_lookup(sk, &dst, &fl))) {
388 sk->sk_err_soft = -err;
389 goto out;
390 }
391
392 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
393 sk->sk_err_soft = -err;
394 goto out;
395 }
396
397 } else
398 dst_hold(dst);
399
d83d8461 400 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
1da177e4
LT
401 tcp_sync_mss(sk, dst_mtu(dst));
402 tcp_simple_retransmit(sk);
403 } /* else let the usual retransmit timer handle it */
404 dst_release(dst);
405 goto out;
406 }
407
408 icmpv6_err_convert(type, code, &err);
409
60236fdd 410 /* Might be for an request_sock */
1da177e4 411 switch (sk->sk_state) {
60236fdd 412 struct request_sock *req, **prev;
1da177e4
LT
413 case TCP_LISTEN:
414 if (sock_owned_by_user(sk))
415 goto out;
416
8129765a
ACM
417 req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
418 &hdr->saddr, inet6_iif(skb));
1da177e4
LT
419 if (!req)
420 goto out;
421
422 /* ICMPs are not backlogged, hence we cannot get
423 * an established socket here.
424 */
425 BUG_TRAP(req->sk == NULL);
426
2e6599cb 427 if (seq != tcp_rsk(req)->snt_isn) {
1da177e4
LT
428 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
429 goto out;
430 }
431
463c84b9 432 inet_csk_reqsk_queue_drop(sk, req, prev);
1da177e4
LT
433 goto out;
434
435 case TCP_SYN_SENT:
436 case TCP_SYN_RECV: /* Cannot happen.
437 It can, it SYNs are crossed. --ANK */
438 if (!sock_owned_by_user(sk)) {
1da177e4
LT
439 sk->sk_err = err;
440 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
441
442 tcp_done(sk);
443 } else
444 sk->sk_err_soft = err;
445 goto out;
446 }
447
448 if (!sock_owned_by_user(sk) && np->recverr) {
449 sk->sk_err = err;
450 sk->sk_error_report(sk);
451 } else
452 sk->sk_err_soft = err;
453
454out:
455 bh_unlock_sock(sk);
456 sock_put(sk);
457}
458
459
60236fdd 460static int tcp_v6_send_synack(struct sock *sk, struct request_sock *req,
1da177e4
LT
461 struct dst_entry *dst)
462{
ca304b61 463 struct inet6_request_sock *treq = inet6_rsk(req);
1da177e4
LT
464 struct ipv6_pinfo *np = inet6_sk(sk);
465 struct sk_buff * skb;
466 struct ipv6_txoptions *opt = NULL;
467 struct in6_addr * final_p = NULL, final;
468 struct flowi fl;
469 int err = -1;
470
471 memset(&fl, 0, sizeof(fl));
472 fl.proto = IPPROTO_TCP;
2e6599cb
ACM
473 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
474 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
1da177e4 475 fl.fl6_flowlabel = 0;
2e6599cb
ACM
476 fl.oif = treq->iif;
477 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
1da177e4 478 fl.fl_ip_sport = inet_sk(sk)->sport;
4237c75c 479 security_req_classify_flow(req, &fl);
1da177e4
LT
480
481 if (dst == NULL) {
482 opt = np->opt;
483 if (opt == NULL &&
333fad53 484 np->rxopt.bits.osrcrt == 2 &&
2e6599cb
ACM
485 treq->pktopts) {
486 struct sk_buff *pktopts = treq->pktopts;
1da177e4
LT
487 struct inet6_skb_parm *rxopt = IP6CB(pktopts);
488 if (rxopt->srcrt)
489 opt = ipv6_invert_rthdr(sk, (struct ipv6_rt_hdr*)(pktopts->nh.raw + rxopt->srcrt));
490 }
491
492 if (opt && opt->srcrt) {
493 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
494 ipv6_addr_copy(&final, &fl.fl6_dst);
495 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
496 final_p = &final;
497 }
498
499 err = ip6_dst_lookup(sk, &dst, &fl);
500 if (err)
501 goto done;
502 if (final_p)
503 ipv6_addr_copy(&fl.fl6_dst, final_p);
504 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
505 goto done;
506 }
507
508 skb = tcp_make_synack(sk, dst, req);
509 if (skb) {
510 struct tcphdr *th = skb->h.th;
511
512 th->check = tcp_v6_check(th, skb->len,
2e6599cb 513 &treq->loc_addr, &treq->rmt_addr,
1da177e4
LT
514 csum_partial((char *)th, skb->len, skb->csum));
515
2e6599cb 516 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
1da177e4 517 err = ip6_xmit(sk, skb, &fl, opt, 0);
b9df3cb8 518 err = net_xmit_eval(err);
1da177e4
LT
519 }
520
521done:
1da177e4
LT
522 if (opt && opt != np->opt)
523 sock_kfree_s(sk, opt, opt->tot_len);
78b91042 524 dst_release(dst);
1da177e4
LT
525 return err;
526}
527
60236fdd 528static void tcp_v6_reqsk_destructor(struct request_sock *req)
1da177e4 529{
ca304b61
ACM
530 if (inet6_rsk(req)->pktopts)
531 kfree_skb(inet6_rsk(req)->pktopts);
1da177e4
LT
532}
533
cfb6eeb4
YH
534#ifdef CONFIG_TCP_MD5SIG
535static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
536 struct in6_addr *addr)
537{
538 struct tcp_sock *tp = tcp_sk(sk);
539 int i;
540
541 BUG_ON(tp == NULL);
542
543 if (!tp->md5sig_info || !tp->md5sig_info->entries6)
544 return NULL;
545
546 for (i = 0; i < tp->md5sig_info->entries6; i++) {
547 if (ipv6_addr_cmp(&tp->md5sig_info->keys6[i].addr, addr) == 0)
548 return (struct tcp_md5sig_key *)&tp->md5sig_info->keys6[i];
549 }
550 return NULL;
551}
552
553static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
554 struct sock *addr_sk)
555{
556 return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr);
557}
558
559static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
560 struct request_sock *req)
561{
562 return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr);
563}
564
565static int tcp_v6_md5_do_add(struct sock *sk, struct in6_addr *peer,
566 char *newkey, u8 newkeylen)
567{
568 /* Add key to the list */
569 struct tcp6_md5sig_key *key;
570 struct tcp_sock *tp = tcp_sk(sk);
571 struct tcp6_md5sig_key *keys;
572
573 key = (struct tcp6_md5sig_key*) tcp_v6_md5_do_lookup(sk, peer);
574 if (key) {
575 /* modify existing entry - just update that one */
576 kfree(key->key);
577 key->key = newkey;
578 key->keylen = newkeylen;
579 } else {
580 /* reallocate new list if current one is full. */
581 if (!tp->md5sig_info) {
582 tp->md5sig_info = kzalloc(sizeof(*tp->md5sig_info), GFP_ATOMIC);
583 if (!tp->md5sig_info) {
584 kfree(newkey);
585 return -ENOMEM;
586 }
587 }
588 tcp_alloc_md5sig_pool();
589 if (tp->md5sig_info->alloced6 == tp->md5sig_info->entries6) {
590 keys = kmalloc((sizeof (tp->md5sig_info->keys6[0]) *
591 (tp->md5sig_info->entries6 + 1)), GFP_ATOMIC);
592
593 if (!keys) {
594 tcp_free_md5sig_pool();
595 kfree(newkey);
596 return -ENOMEM;
597 }
598
599 if (tp->md5sig_info->entries6)
600 memmove(keys, tp->md5sig_info->keys6,
601 (sizeof (tp->md5sig_info->keys6[0]) *
602 tp->md5sig_info->entries6));
603
604 kfree(tp->md5sig_info->keys6);
605 tp->md5sig_info->keys6 = keys;
606 tp->md5sig_info->alloced6++;
607 }
608
609 ipv6_addr_copy(&tp->md5sig_info->keys6[tp->md5sig_info->entries6].addr,
610 peer);
611 tp->md5sig_info->keys6[tp->md5sig_info->entries6].key = newkey;
612 tp->md5sig_info->keys6[tp->md5sig_info->entries6].keylen = newkeylen;
613
614 tp->md5sig_info->entries6++;
615 }
616 return 0;
617}
618
619static int tcp_v6_md5_add_func(struct sock *sk, struct sock *addr_sk,
620 u8 *newkey, __u8 newkeylen)
621{
622 return tcp_v6_md5_do_add(sk, &inet6_sk(addr_sk)->daddr,
623 newkey, newkeylen);
624}
625
626static int tcp_v6_md5_do_del(struct sock *sk, struct in6_addr *peer)
627{
628 struct tcp_sock *tp = tcp_sk(sk);
629 int i;
630
631 for (i = 0; i < tp->md5sig_info->entries6; i++) {
632 if (ipv6_addr_cmp(&tp->md5sig_info->keys6[i].addr, peer) == 0) {
633 /* Free the key */
634 kfree(tp->md5sig_info->keys6[i].key);
635 tp->md5sig_info->entries6--;
636
637 if (tp->md5sig_info->entries6 == 0) {
638 kfree(tp->md5sig_info->keys6);
639 tp->md5sig_info->keys6 = NULL;
640
641 tcp_free_md5sig_pool();
642
643 return 0;
644 } else {
645 /* shrink the database */
646 if (tp->md5sig_info->entries6 != i)
647 memmove(&tp->md5sig_info->keys6[i],
648 &tp->md5sig_info->keys6[i+1],
649 (tp->md5sig_info->entries6 - i)
650 * sizeof (tp->md5sig_info->keys6[0]));
651 }
652 }
653 }
654 return -ENOENT;
655}
656
657static void tcp_v6_clear_md5_list (struct sock *sk)
658{
659 struct tcp_sock *tp = tcp_sk(sk);
660 int i;
661
662 if (tp->md5sig_info->entries6) {
663 for (i = 0; i < tp->md5sig_info->entries6; i++)
664 kfree(tp->md5sig_info->keys6[i].key);
665 tp->md5sig_info->entries6 = 0;
666 tcp_free_md5sig_pool();
667 }
668
669 kfree(tp->md5sig_info->keys6);
670 tp->md5sig_info->keys6 = NULL;
671 tp->md5sig_info->alloced6 = 0;
672
673 if (tp->md5sig_info->entries4) {
674 for (i = 0; i < tp->md5sig_info->entries4; i++)
675 kfree(tp->md5sig_info->keys4[i].key);
676 tp->md5sig_info->entries4 = 0;
677 tcp_free_md5sig_pool();
678 }
679
680 kfree(tp->md5sig_info->keys4);
681 tp->md5sig_info->keys4 = NULL;
682 tp->md5sig_info->alloced4 = 0;
683}
684
685static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval,
686 int optlen)
687{
688 struct tcp_md5sig cmd;
689 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
690 u8 *newkey;
691
692 if (optlen < sizeof(cmd))
693 return -EINVAL;
694
695 if (copy_from_user(&cmd, optval, sizeof(cmd)))
696 return -EFAULT;
697
698 if (sin6->sin6_family != AF_INET6)
699 return -EINVAL;
700
701 if (!cmd.tcpm_keylen) {
702 if (!tcp_sk(sk)->md5sig_info)
703 return -ENOENT;
704 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_MAPPED)
705 return tcp_v4_md5_do_del(sk, sin6->sin6_addr.s6_addr32[3]);
706 return tcp_v6_md5_do_del(sk, &sin6->sin6_addr);
707 }
708
709 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
710 return -EINVAL;
711
712 if (!tcp_sk(sk)->md5sig_info) {
713 struct tcp_sock *tp = tcp_sk(sk);
714 struct tcp_md5sig_info *p;
715
716 p = kzalloc(sizeof(struct tcp_md5sig_info), GFP_KERNEL);
717 if (!p)
718 return -ENOMEM;
719
720 tp->md5sig_info = p;
721 }
722
af879cc7 723 newkey = kmemdup(cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
cfb6eeb4
YH
724 if (!newkey)
725 return -ENOMEM;
cfb6eeb4
YH
726 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_MAPPED) {
727 return tcp_v4_md5_do_add(sk, sin6->sin6_addr.s6_addr32[3],
728 newkey, cmd.tcpm_keylen);
729 }
730 return tcp_v6_md5_do_add(sk, &sin6->sin6_addr, newkey, cmd.tcpm_keylen);
731}
732
733static int tcp_v6_do_calc_md5_hash(char *md5_hash, struct tcp_md5sig_key *key,
734 struct in6_addr *saddr,
735 struct in6_addr *daddr,
736 struct tcphdr *th, int protocol,
737 int tcplen)
738{
739 struct scatterlist sg[4];
740 __u16 data_len;
741 int block = 0;
8e5200f5 742 __sum16 cksum;
cfb6eeb4
YH
743 struct tcp_md5sig_pool *hp;
744 struct tcp6_pseudohdr *bp;
745 struct hash_desc *desc;
746 int err;
747 unsigned int nbytes = 0;
748
749 hp = tcp_get_md5sig_pool();
750 if (!hp) {
751 printk(KERN_WARNING "%s(): hash pool not found...\n", __FUNCTION__);
752 goto clear_hash_noput;
753 }
754 bp = &hp->md5_blk.ip6;
755 desc = &hp->md5_desc;
756
757 /* 1. TCP pseudo-header (RFC2460) */
758 ipv6_addr_copy(&bp->saddr, saddr);
759 ipv6_addr_copy(&bp->daddr, daddr);
760 bp->len = htonl(tcplen);
761 bp->protocol = htonl(protocol);
762
763 sg_set_buf(&sg[block++], bp, sizeof(*bp));
764 nbytes += sizeof(*bp);
765
766 /* 2. TCP header, excluding options */
767 cksum = th->check;
768 th->check = 0;
769 sg_set_buf(&sg[block++], th, sizeof(*th));
770 nbytes += sizeof(*th);
771
772 /* 3. TCP segment data (if any) */
773 data_len = tcplen - (th->doff << 2);
774 if (data_len > 0) {
775 u8 *data = (u8 *)th + (th->doff << 2);
776 sg_set_buf(&sg[block++], data, data_len);
777 nbytes += data_len;
778 }
779
780 /* 4. shared key */
781 sg_set_buf(&sg[block++], key->key, key->keylen);
782 nbytes += key->keylen;
783
784 /* Now store the hash into the packet */
785 err = crypto_hash_init(desc);
786 if (err) {
787 printk(KERN_WARNING "%s(): hash_init failed\n", __FUNCTION__);
788 goto clear_hash;
789 }
790 err = crypto_hash_update(desc, sg, nbytes);
791 if (err) {
792 printk(KERN_WARNING "%s(): hash_update failed\n", __FUNCTION__);
793 goto clear_hash;
794 }
795 err = crypto_hash_final(desc, md5_hash);
796 if (err) {
797 printk(KERN_WARNING "%s(): hash_final failed\n", __FUNCTION__);
798 goto clear_hash;
799 }
800
801 /* Reset header, and free up the crypto */
802 tcp_put_md5sig_pool();
803 th->check = cksum;
804out:
805 return 0;
806clear_hash:
807 tcp_put_md5sig_pool();
808clear_hash_noput:
809 memset(md5_hash, 0, 16);
810 goto out;
811}
812
813static int tcp_v6_calc_md5_hash(char *md5_hash, struct tcp_md5sig_key *key,
814 struct sock *sk,
815 struct dst_entry *dst,
816 struct request_sock *req,
817 struct tcphdr *th, int protocol,
818 int tcplen)
819{
820 struct in6_addr *saddr, *daddr;
821
822 if (sk) {
823 saddr = &inet6_sk(sk)->saddr;
824 daddr = &inet6_sk(sk)->daddr;
825 } else {
826 saddr = &inet6_rsk(req)->loc_addr;
827 daddr = &inet6_rsk(req)->rmt_addr;
828 }
829 return tcp_v6_do_calc_md5_hash(md5_hash, key,
830 saddr, daddr,
831 th, protocol, tcplen);
832}
833
834static int tcp_v6_inbound_md5_hash (struct sock *sk, struct sk_buff *skb)
835{
836 __u8 *hash_location = NULL;
837 struct tcp_md5sig_key *hash_expected;
838 struct ipv6hdr *ip6h = skb->nh.ipv6h;
839 struct tcphdr *th = skb->h.th;
840 int length = (th->doff << 2) - sizeof (*th);
841 int genhash;
842 u8 *ptr;
843 u8 newhash[16];
844
845 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
846
847 /* If the TCP option is too short, we can short cut */
848 if (length < TCPOLEN_MD5SIG)
849 return hash_expected ? 1 : 0;
850
851 /* parse options */
852 ptr = (u8*)(th + 1);
853 while (length > 0) {
854 int opcode = *ptr++;
855 int opsize;
856
857 switch(opcode) {
858 case TCPOPT_EOL:
859 goto done_opts;
860 case TCPOPT_NOP:
861 length--;
862 continue;
863 default:
864 opsize = *ptr++;
865 if (opsize < 2 || opsize > length)
866 goto done_opts;
867 if (opcode == TCPOPT_MD5SIG) {
868 hash_location = ptr;
869 goto done_opts;
870 }
871 }
872 ptr += opsize - 2;
873 length -= opsize;
874 }
875
876done_opts:
877 /* do we have a hash as expected? */
878 if (!hash_expected) {
879 if (!hash_location)
880 return 0;
881 if (net_ratelimit()) {
882 printk(KERN_INFO "MD5 Hash NOT expected but found "
883 "(" NIP6_FMT ", %u)->"
884 "(" NIP6_FMT ", %u)\n",
885 NIP6(ip6h->saddr), ntohs(th->source),
886 NIP6(ip6h->daddr), ntohs(th->dest));
887 }
888 return 1;
889 }
890
891 if (!hash_location) {
892 if (net_ratelimit()) {
893 printk(KERN_INFO "MD5 Hash expected but NOT found "
894 "(" NIP6_FMT ", %u)->"
895 "(" NIP6_FMT ", %u)\n",
896 NIP6(ip6h->saddr), ntohs(th->source),
897 NIP6(ip6h->daddr), ntohs(th->dest));
898 }
899 return 1;
900 }
901
902 /* check the signature */
903 genhash = tcp_v6_do_calc_md5_hash(newhash,
904 hash_expected,
905 &ip6h->saddr, &ip6h->daddr,
906 th, sk->sk_protocol,
907 skb->len);
908 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
909 if (net_ratelimit()) {
910 printk(KERN_INFO "MD5 Hash %s for "
911 "(" NIP6_FMT ", %u)->"
912 "(" NIP6_FMT ", %u)\n",
913 genhash ? "failed" : "mismatch",
914 NIP6(ip6h->saddr), ntohs(th->source),
915 NIP6(ip6h->daddr), ntohs(th->dest));
916 }
917 return 1;
918 }
919 return 0;
920}
921#endif
922
9ec75fe8 923static struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
1da177e4 924 .family = AF_INET6,
2e6599cb 925 .obj_size = sizeof(struct tcp6_request_sock),
1da177e4 926 .rtx_syn_ack = tcp_v6_send_synack,
60236fdd
ACM
927 .send_ack = tcp_v6_reqsk_send_ack,
928 .destructor = tcp_v6_reqsk_destructor,
1da177e4
LT
929 .send_reset = tcp_v6_send_reset
930};
931
cfb6eeb4
YH
932struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
933#ifdef CONFIG_TCP_MD5SIG
934 .md5_lookup = tcp_v6_reqsk_md5_lookup,
935#endif
936};
937
6d6ee43e
ACM
938static struct timewait_sock_ops tcp6_timewait_sock_ops = {
939 .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
940 .twsk_unique = tcp_twsk_unique,
cfb6eeb4 941 .twsk_destructor= tcp_twsk_destructor,
6d6ee43e
ACM
942};
943
8292a17a 944static void tcp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
1da177e4
LT
945{
946 struct ipv6_pinfo *np = inet6_sk(sk);
8292a17a 947 struct tcphdr *th = skb->h.th;
1da177e4 948
84fa7933 949 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1da177e4
LT
950 th->check = ~csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP, 0);
951 skb->csum = offsetof(struct tcphdr, check);
952 } else {
953 th->check = csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP,
954 csum_partial((char *)th, th->doff<<2,
955 skb->csum));
956 }
957}
958
a430a43d
HX
959static int tcp_v6_gso_send_check(struct sk_buff *skb)
960{
961 struct ipv6hdr *ipv6h;
962 struct tcphdr *th;
963
964 if (!pskb_may_pull(skb, sizeof(*th)))
965 return -EINVAL;
966
967 ipv6h = skb->nh.ipv6h;
968 th = skb->h.th;
969
970 th->check = 0;
971 th->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
972 IPPROTO_TCP, 0);
973 skb->csum = offsetof(struct tcphdr, check);
84fa7933 974 skb->ip_summed = CHECKSUM_PARTIAL;
a430a43d
HX
975 return 0;
976}
1da177e4 977
cfb6eeb4 978static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
1da177e4
LT
979{
980 struct tcphdr *th = skb->h.th, *t1;
981 struct sk_buff *buff;
982 struct flowi fl;
cfb6eeb4
YH
983 int tot_len = sizeof(*th);
984#ifdef CONFIG_TCP_MD5SIG
985 struct tcp_md5sig_key *key;
986#endif
1da177e4
LT
987
988 if (th->rst)
989 return;
990
991 if (!ipv6_unicast_destination(skb))
992 return;
993
cfb6eeb4
YH
994#ifdef CONFIG_TCP_MD5SIG
995 if (sk)
996 key = tcp_v6_md5_do_lookup(sk, &skb->nh.ipv6h->daddr);
997 else
998 key = NULL;
999
1000 if (key)
1001 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1002#endif
1003
1da177e4
LT
1004 /*
1005 * We need to grab some memory, and put together an RST,
1006 * and then put it into the queue to be sent.
1007 */
1008
cfb6eeb4 1009 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1da177e4
LT
1010 GFP_ATOMIC);
1011 if (buff == NULL)
1012 return;
1013
cfb6eeb4 1014 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1da177e4 1015
cfb6eeb4 1016 t1 = (struct tcphdr *) skb_push(buff, tot_len);
1da177e4
LT
1017
1018 /* Swap the send and the receive. */
1019 memset(t1, 0, sizeof(*t1));
1020 t1->dest = th->source;
1021 t1->source = th->dest;
cfb6eeb4 1022 t1->doff = tot_len / 4;
1da177e4
LT
1023 t1->rst = 1;
1024
1025 if(th->ack) {
1026 t1->seq = th->ack_seq;
1027 } else {
1028 t1->ack = 1;
1029 t1->ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin
1030 + skb->len - (th->doff<<2));
1031 }
1032
cfb6eeb4
YH
1033#ifdef CONFIG_TCP_MD5SIG
1034 if (key) {
8e5200f5 1035 __be32 *opt = (__be32*)(t1 + 1);
cfb6eeb4
YH
1036 opt[0] = htonl((TCPOPT_NOP << 24) |
1037 (TCPOPT_NOP << 16) |
1038 (TCPOPT_MD5SIG << 8) |
1039 TCPOLEN_MD5SIG);
1040 tcp_v6_do_calc_md5_hash((__u8*)&opt[1],
1041 key,
1042 &skb->nh.ipv6h->daddr,
1043 &skb->nh.ipv6h->saddr,
1044 t1, IPPROTO_TCP,
1045 tot_len);
1046 }
1047#endif
1048
1da177e4
LT
1049 buff->csum = csum_partial((char *)t1, sizeof(*t1), 0);
1050
1051 memset(&fl, 0, sizeof(fl));
1052 ipv6_addr_copy(&fl.fl6_dst, &skb->nh.ipv6h->saddr);
1053 ipv6_addr_copy(&fl.fl6_src, &skb->nh.ipv6h->daddr);
1054
1055 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1056 sizeof(*t1), IPPROTO_TCP,
1057 buff->csum);
1058
1059 fl.proto = IPPROTO_TCP;
505cbfc5 1060 fl.oif = inet6_iif(skb);
1da177e4
LT
1061 fl.fl_ip_dport = t1->dest;
1062 fl.fl_ip_sport = t1->source;
beb8d13b 1063 security_skb_classify_flow(skb, &fl);
1da177e4
LT
1064
1065 /* sk = NULL, but it is safe for now. RST socket required. */
1066 if (!ip6_dst_lookup(NULL, &buff->dst, &fl)) {
1067
ecc51b6d 1068 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
ae0f7d5f 1069 ip6_xmit(tcp6_socket->sk, buff, &fl, NULL, 0);
ecc51b6d
ACM
1070 TCP_INC_STATS_BH(TCP_MIB_OUTSEGS);
1071 TCP_INC_STATS_BH(TCP_MIB_OUTRSTS);
1da177e4 1072 return;
ecc51b6d 1073 }
1da177e4
LT
1074 }
1075
1076 kfree_skb(buff);
1077}
1078
cfb6eeb4
YH
1079static void tcp_v6_send_ack(struct tcp_timewait_sock *tw,
1080 struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts)
1da177e4
LT
1081{
1082 struct tcphdr *th = skb->h.th, *t1;
1083 struct sk_buff *buff;
1084 struct flowi fl;
1085 int tot_len = sizeof(struct tcphdr);
e69a4adc 1086 __be32 *topt;
cfb6eeb4
YH
1087#ifdef CONFIG_TCP_MD5SIG
1088 struct tcp_md5sig_key *key;
1089 struct tcp_md5sig_key tw_key;
1090#endif
1091
1092#ifdef CONFIG_TCP_MD5SIG
1093 if (!tw && skb->sk) {
1094 key = tcp_v6_md5_do_lookup(skb->sk, &skb->nh.ipv6h->daddr);
1095 } else if (tw && tw->tw_md5_keylen) {
1096 tw_key.key = tw->tw_md5_key;
1097 tw_key.keylen = tw->tw_md5_keylen;
1098 key = &tw_key;
1099 } else {
1100 key = NULL;
1101 }
1102#endif
1da177e4
LT
1103
1104 if (ts)
4244f8a9 1105 tot_len += TCPOLEN_TSTAMP_ALIGNED;
cfb6eeb4
YH
1106#ifdef CONFIG_TCP_MD5SIG
1107 if (key)
1108 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1109#endif
1da177e4
LT
1110
1111 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1112 GFP_ATOMIC);
1113 if (buff == NULL)
1114 return;
1115
1116 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1117
1118 t1 = (struct tcphdr *) skb_push(buff,tot_len);
1119
1120 /* Swap the send and the receive. */
1121 memset(t1, 0, sizeof(*t1));
1122 t1->dest = th->source;
1123 t1->source = th->dest;
1124 t1->doff = tot_len/4;
1125 t1->seq = htonl(seq);
1126 t1->ack_seq = htonl(ack);
1127 t1->ack = 1;
1128 t1->window = htons(win);
cfb6eeb4 1129
e69a4adc 1130 topt = (__be32 *)(t1 + 1);
1da177e4
LT
1131
1132 if (ts) {
cfb6eeb4
YH
1133 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1134 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
1135 *topt++ = htonl(tcp_time_stamp);
1136 *topt = htonl(ts);
1da177e4
LT
1137 }
1138
cfb6eeb4
YH
1139#ifdef CONFIG_TCP_MD5SIG
1140 if (key) {
1141 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1142 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
1143 tcp_v6_do_calc_md5_hash((__u8 *)topt,
1144 key,
1145 &skb->nh.ipv6h->daddr,
1146 &skb->nh.ipv6h->saddr,
1147 t1, IPPROTO_TCP,
1148 tot_len);
1149 }
1150#endif
1151
1da177e4
LT
1152 buff->csum = csum_partial((char *)t1, tot_len, 0);
1153
1154 memset(&fl, 0, sizeof(fl));
1155 ipv6_addr_copy(&fl.fl6_dst, &skb->nh.ipv6h->saddr);
1156 ipv6_addr_copy(&fl.fl6_src, &skb->nh.ipv6h->daddr);
1157
1158 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1159 tot_len, IPPROTO_TCP,
1160 buff->csum);
1161
1162 fl.proto = IPPROTO_TCP;
505cbfc5 1163 fl.oif = inet6_iif(skb);
1da177e4
LT
1164 fl.fl_ip_dport = t1->dest;
1165 fl.fl_ip_sport = t1->source;
beb8d13b 1166 security_skb_classify_flow(skb, &fl);
1da177e4
LT
1167
1168 if (!ip6_dst_lookup(NULL, &buff->dst, &fl)) {
ecc51b6d 1169 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
ae0f7d5f 1170 ip6_xmit(tcp6_socket->sk, buff, &fl, NULL, 0);
ecc51b6d 1171 TCP_INC_STATS_BH(TCP_MIB_OUTSEGS);
1da177e4 1172 return;
ecc51b6d 1173 }
1da177e4
LT
1174 }
1175
1176 kfree_skb(buff);
1177}
1178
1179static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
1180{
8feaf0c0 1181 struct inet_timewait_sock *tw = inet_twsk(sk);
cfb6eeb4 1182 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
1da177e4 1183
cfb6eeb4 1184 tcp_v6_send_ack(tcptw, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
8feaf0c0
ACM
1185 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
1186 tcptw->tw_ts_recent);
1da177e4 1187
8feaf0c0 1188 inet_twsk_put(tw);
1da177e4
LT
1189}
1190
60236fdd 1191static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req)
1da177e4 1192{
cfb6eeb4 1193 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
1194}
1195
1196
1197static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
1198{
60236fdd 1199 struct request_sock *req, **prev;
505cbfc5 1200 const struct tcphdr *th = skb->h.th;
1da177e4
LT
1201 struct sock *nsk;
1202
1203 /* Find possible connection requests. */
8129765a
ACM
1204 req = inet6_csk_search_req(sk, &prev, th->source,
1205 &skb->nh.ipv6h->saddr,
1206 &skb->nh.ipv6h->daddr, inet6_iif(skb));
1da177e4
LT
1207 if (req)
1208 return tcp_check_req(sk, skb, req, prev);
1209
505cbfc5
ACM
1210 nsk = __inet6_lookup_established(&tcp_hashinfo, &skb->nh.ipv6h->saddr,
1211 th->source, &skb->nh.ipv6h->daddr,
1212 ntohs(th->dest), inet6_iif(skb));
1da177e4
LT
1213
1214 if (nsk) {
1215 if (nsk->sk_state != TCP_TIME_WAIT) {
1216 bh_lock_sock(nsk);
1217 return nsk;
1218 }
9469c7b4 1219 inet_twsk_put(inet_twsk(nsk));
1da177e4
LT
1220 return NULL;
1221 }
1222
1223#if 0 /*def CONFIG_SYN_COOKIES*/
1224 if (!th->rst && !th->syn && th->ack)
1225 sk = cookie_v6_check(sk, skb, &(IPCB(skb)->opt));
1226#endif
1227 return sk;
1228}
1229
1da177e4
LT
1230/* FIXME: this is substantially similar to the ipv4 code.
1231 * Can some kind of merge be done? -- erics
1232 */
1233static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
1234{
ca304b61 1235 struct inet6_request_sock *treq;
1da177e4
LT
1236 struct ipv6_pinfo *np = inet6_sk(sk);
1237 struct tcp_options_received tmp_opt;
1238 struct tcp_sock *tp = tcp_sk(sk);
60236fdd 1239 struct request_sock *req = NULL;
1da177e4
LT
1240 __u32 isn = TCP_SKB_CB(skb)->when;
1241
1242 if (skb->protocol == htons(ETH_P_IP))
1243 return tcp_v4_conn_request(sk, skb);
1244
1245 if (!ipv6_unicast_destination(skb))
1246 goto drop;
1247
1248 /*
1249 * There are no SYN attacks on IPv6, yet...
1250 */
463c84b9 1251 if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
1da177e4
LT
1252 if (net_ratelimit())
1253 printk(KERN_INFO "TCPv6: dropping request, synflood is possible\n");
1254 goto drop;
1255 }
1256
463c84b9 1257 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
1da177e4
LT
1258 goto drop;
1259
ca304b61 1260 req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
1da177e4
LT
1261 if (req == NULL)
1262 goto drop;
1263
cfb6eeb4
YH
1264#ifdef CONFIG_TCP_MD5SIG
1265 tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
1266#endif
1267
1da177e4
LT
1268 tcp_clear_options(&tmp_opt);
1269 tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
1270 tmp_opt.user_mss = tp->rx_opt.user_mss;
1271
1272 tcp_parse_options(skb, &tmp_opt, 0);
1273
1274 tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
1275 tcp_openreq_init(req, &tmp_opt, skb);
1276
ca304b61 1277 treq = inet6_rsk(req);
2e6599cb
ACM
1278 ipv6_addr_copy(&treq->rmt_addr, &skb->nh.ipv6h->saddr);
1279 ipv6_addr_copy(&treq->loc_addr, &skb->nh.ipv6h->daddr);
1da177e4 1280 TCP_ECN_create_request(req, skb->h.th);
2e6599cb 1281 treq->pktopts = NULL;
1da177e4 1282 if (ipv6_opt_accepted(sk, skb) ||
333fad53
YH
1283 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
1284 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
1da177e4 1285 atomic_inc(&skb->users);
2e6599cb 1286 treq->pktopts = skb;
1da177e4 1287 }
2e6599cb 1288 treq->iif = sk->sk_bound_dev_if;
1da177e4
LT
1289
1290 /* So that link locals have meaning */
1291 if (!sk->sk_bound_dev_if &&
2e6599cb 1292 ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL)
505cbfc5 1293 treq->iif = inet6_iif(skb);
1da177e4
LT
1294
1295 if (isn == 0)
a94f723d 1296 isn = tcp_v6_init_sequence(skb);
1da177e4 1297
2e6599cb 1298 tcp_rsk(req)->snt_isn = isn;
1da177e4 1299
4237c75c
VY
1300 security_inet_conn_request(sk, skb, req);
1301
1da177e4
LT
1302 if (tcp_v6_send_synack(sk, req, NULL))
1303 goto drop;
1304
8129765a 1305 inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
1da177e4
LT
1306 return 0;
1307
1308drop:
1309 if (req)
60236fdd 1310 reqsk_free(req);
1da177e4 1311
1da177e4
LT
1312 return 0; /* don't send reset */
1313}
1314
1315static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
60236fdd 1316 struct request_sock *req,
1da177e4
LT
1317 struct dst_entry *dst)
1318{
ca304b61 1319 struct inet6_request_sock *treq = inet6_rsk(req);
1da177e4
LT
1320 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
1321 struct tcp6_sock *newtcp6sk;
1322 struct inet_sock *newinet;
1323 struct tcp_sock *newtp;
1324 struct sock *newsk;
1325 struct ipv6_txoptions *opt;
cfb6eeb4
YH
1326#ifdef CONFIG_TCP_MD5SIG
1327 struct tcp_md5sig_key *key;
1328#endif
1da177e4
LT
1329
1330 if (skb->protocol == htons(ETH_P_IP)) {
1331 /*
1332 * v6 mapped
1333 */
1334
1335 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1336
1337 if (newsk == NULL)
1338 return NULL;
1339
1340 newtcp6sk = (struct tcp6_sock *)newsk;
1341 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1342
1343 newinet = inet_sk(newsk);
1344 newnp = inet6_sk(newsk);
1345 newtp = tcp_sk(newsk);
1346
1347 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1348
1349 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
1350 newinet->daddr);
1351
1352 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
1353 newinet->saddr);
1354
1355 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
1356
8292a17a 1357 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1da177e4 1358 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
cfb6eeb4
YH
1359#ifdef CONFIG_TCP_MD5SIG
1360 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1361#endif
1362
1da177e4
LT
1363 newnp->pktoptions = NULL;
1364 newnp->opt = NULL;
505cbfc5 1365 newnp->mcast_oif = inet6_iif(skb);
1da177e4
LT
1366 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
1367
e6848976
ACM
1368 /*
1369 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1370 * here, tcp_create_openreq_child now does this for us, see the comment in
1371 * that function for the gory details. -acme
1da177e4 1372 */
1da177e4
LT
1373
1374 /* It is tricky place. Until this moment IPv4 tcp
8292a17a 1375 worked with IPv6 icsk.icsk_af_ops.
1da177e4
LT
1376 Sync it now.
1377 */
d83d8461 1378 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1da177e4
LT
1379
1380 return newsk;
1381 }
1382
1383 opt = np->opt;
1384
1385 if (sk_acceptq_is_full(sk))
1386 goto out_overflow;
1387
333fad53 1388 if (np->rxopt.bits.osrcrt == 2 &&
2e6599cb
ACM
1389 opt == NULL && treq->pktopts) {
1390 struct inet6_skb_parm *rxopt = IP6CB(treq->pktopts);
1da177e4 1391 if (rxopt->srcrt)
2e6599cb 1392 opt = ipv6_invert_rthdr(sk, (struct ipv6_rt_hdr *)(treq->pktopts->nh.raw + rxopt->srcrt));
1da177e4
LT
1393 }
1394
1395 if (dst == NULL) {
1396 struct in6_addr *final_p = NULL, final;
1397 struct flowi fl;
1398
1399 memset(&fl, 0, sizeof(fl));
1400 fl.proto = IPPROTO_TCP;
2e6599cb 1401 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
1da177e4
LT
1402 if (opt && opt->srcrt) {
1403 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
1404 ipv6_addr_copy(&final, &fl.fl6_dst);
1405 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
1406 final_p = &final;
1407 }
2e6599cb 1408 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
1da177e4 1409 fl.oif = sk->sk_bound_dev_if;
2e6599cb 1410 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
1da177e4 1411 fl.fl_ip_sport = inet_sk(sk)->sport;
4237c75c 1412 security_req_classify_flow(req, &fl);
1da177e4
LT
1413
1414 if (ip6_dst_lookup(sk, &dst, &fl))
1415 goto out;
1416
1417 if (final_p)
1418 ipv6_addr_copy(&fl.fl6_dst, final_p);
1419
1420 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
1421 goto out;
1422 }
1423
1424 newsk = tcp_create_openreq_child(sk, req, skb);
1425 if (newsk == NULL)
1426 goto out;
1427
e6848976
ACM
1428 /*
1429 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1430 * count here, tcp_create_openreq_child now does this for us, see the
1431 * comment in that function for the gory details. -acme
1432 */
1da177e4 1433
59eed279 1434 newsk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 1435 __ip6_dst_store(newsk, dst, NULL, NULL);
1da177e4
LT
1436
1437 newtcp6sk = (struct tcp6_sock *)newsk;
1438 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1439
1440 newtp = tcp_sk(newsk);
1441 newinet = inet_sk(newsk);
1442 newnp = inet6_sk(newsk);
1443
1444 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1445
2e6599cb
ACM
1446 ipv6_addr_copy(&newnp->daddr, &treq->rmt_addr);
1447 ipv6_addr_copy(&newnp->saddr, &treq->loc_addr);
1448 ipv6_addr_copy(&newnp->rcv_saddr, &treq->loc_addr);
1449 newsk->sk_bound_dev_if = treq->iif;
1da177e4
LT
1450
1451 /* Now IPv6 options...
1452
1453 First: no IPv4 options.
1454 */
1455 newinet->opt = NULL;
1456
1457 /* Clone RX bits */
1458 newnp->rxopt.all = np->rxopt.all;
1459
1460 /* Clone pktoptions received with SYN */
1461 newnp->pktoptions = NULL;
2e6599cb
ACM
1462 if (treq->pktopts != NULL) {
1463 newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC);
1464 kfree_skb(treq->pktopts);
1465 treq->pktopts = NULL;
1da177e4
LT
1466 if (newnp->pktoptions)
1467 skb_set_owner_r(newnp->pktoptions, newsk);
1468 }
1469 newnp->opt = NULL;
505cbfc5 1470 newnp->mcast_oif = inet6_iif(skb);
1da177e4
LT
1471 newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
1472
1473 /* Clone native IPv6 options from listening socket (if any)
1474
1475 Yes, keeping reference count would be much more clever,
1476 but we make one more one thing there: reattach optmem
1477 to newsk.
1478 */
1479 if (opt) {
1480 newnp->opt = ipv6_dup_options(newsk, opt);
1481 if (opt != np->opt)
1482 sock_kfree_s(sk, opt, opt->tot_len);
1483 }
1484
d83d8461 1485 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1da177e4 1486 if (newnp->opt)
d83d8461
ACM
1487 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1488 newnp->opt->opt_flen);
1da177e4 1489
5d424d5a 1490 tcp_mtup_init(newsk);
1da177e4
LT
1491 tcp_sync_mss(newsk, dst_mtu(dst));
1492 newtp->advmss = dst_metric(dst, RTAX_ADVMSS);
1493 tcp_initialize_rcv_mss(newsk);
1494
1495 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
1496
cfb6eeb4
YH
1497#ifdef CONFIG_TCP_MD5SIG
1498 /* Copy over the MD5 key from the original socket */
1499 if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) {
1500 /* We're using one, so create a matching key
1501 * on the newsk structure. If we fail to get
1502 * memory, then we end up not copying the key
1503 * across. Shucks.
1504 */
af879cc7
ACM
1505 char *newkey = kmemdup(key->key, key->keylen, GFP_ATOMIC);
1506 if (newkey != NULL)
cfb6eeb4
YH
1507 tcp_v6_md5_do_add(newsk, &inet6_sk(sk)->daddr,
1508 newkey, key->keylen);
cfb6eeb4
YH
1509 }
1510#endif
1511
90b19d31 1512 __inet6_hash(&tcp_hashinfo, newsk);
2d8c4ce5 1513 inet_inherit_port(&tcp_hashinfo, sk, newsk);
1da177e4
LT
1514
1515 return newsk;
1516
1517out_overflow:
1518 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
1519out:
1520 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
1521 if (opt && opt != np->opt)
1522 sock_kfree_s(sk, opt, opt->tot_len);
1523 dst_release(dst);
1524 return NULL;
1525}
1526
b51655b9 1527static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
1da177e4 1528{
84fa7933 1529 if (skb->ip_summed == CHECKSUM_COMPLETE) {
1da177e4 1530 if (!tcp_v6_check(skb->h.th,skb->len,&skb->nh.ipv6h->saddr,
fb286bb2
HX
1531 &skb->nh.ipv6h->daddr,skb->csum)) {
1532 skb->ip_summed = CHECKSUM_UNNECESSARY;
1da177e4 1533 return 0;
fb286bb2 1534 }
1da177e4 1535 }
fb286bb2 1536
868c86bc
AV
1537 skb->csum = ~csum_unfold(tcp_v6_check(skb->h.th,skb->len,&skb->nh.ipv6h->saddr,
1538 &skb->nh.ipv6h->daddr, 0));
fb286bb2 1539
1da177e4 1540 if (skb->len <= 76) {
fb286bb2 1541 return __skb_checksum_complete(skb);
1da177e4
LT
1542 }
1543 return 0;
1544}
1545
1546/* The socket must have it's spinlock held when we get
1547 * here.
1548 *
1549 * We have a potential double-lock case here, so even when
1550 * doing backlog processing we use the BH locking scheme.
1551 * This is because we cannot sleep with the original spinlock
1552 * held.
1553 */
1554static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1555{
1556 struct ipv6_pinfo *np = inet6_sk(sk);
1557 struct tcp_sock *tp;
1558 struct sk_buff *opt_skb = NULL;
1559
1560 /* Imagine: socket is IPv6. IPv4 packet arrives,
1561 goes to IPv4 receive handler and backlogged.
1562 From backlog it always goes here. Kerboom...
1563 Fortunately, tcp_rcv_established and rcv_established
1564 handle them correctly, but it is not case with
1565 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1566 */
1567
1568 if (skb->protocol == htons(ETH_P_IP))
1569 return tcp_v4_do_rcv(sk, skb);
1570
cfb6eeb4
YH
1571#ifdef CONFIG_TCP_MD5SIG
1572 if (tcp_v6_inbound_md5_hash (sk, skb))
1573 goto discard;
1574#endif
1575
fda9ef5d 1576 if (sk_filter(sk, skb))
1da177e4
LT
1577 goto discard;
1578
1579 /*
1580 * socket locking is here for SMP purposes as backlog rcv
1581 * is currently called with bh processing disabled.
1582 */
1583
1584 /* Do Stevens' IPV6_PKTOPTIONS.
1585
1586 Yes, guys, it is the only place in our code, where we
1587 may make it not affecting IPv4.
1588 The rest of code is protocol independent,
1589 and I do not like idea to uglify IPv4.
1590
1591 Actually, all the idea behind IPV6_PKTOPTIONS
1592 looks not very well thought. For now we latch
1593 options, received in the last packet, enqueued
1594 by tcp. Feel free to propose better solution.
1595 --ANK (980728)
1596 */
1597 if (np->rxopt.all)
1598 opt_skb = skb_clone(skb, GFP_ATOMIC);
1599
1600 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1601 TCP_CHECK_TIMER(sk);
1602 if (tcp_rcv_established(sk, skb, skb->h.th, skb->len))
1603 goto reset;
1604 TCP_CHECK_TIMER(sk);
1605 if (opt_skb)
1606 goto ipv6_pktoptions;
1607 return 0;
1608 }
1609
1610 if (skb->len < (skb->h.th->doff<<2) || tcp_checksum_complete(skb))
1611 goto csum_err;
1612
1613 if (sk->sk_state == TCP_LISTEN) {
1614 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1615 if (!nsk)
1616 goto discard;
1617
1618 /*
1619 * Queue it on the new socket if the new socket is active,
1620 * otherwise we just shortcircuit this and continue with
1621 * the new socket..
1622 */
1623 if(nsk != sk) {
1624 if (tcp_child_process(sk, nsk, skb))
1625 goto reset;
1626 if (opt_skb)
1627 __kfree_skb(opt_skb);
1628 return 0;
1629 }
1630 }
1631
1632 TCP_CHECK_TIMER(sk);
1633 if (tcp_rcv_state_process(sk, skb, skb->h.th, skb->len))
1634 goto reset;
1635 TCP_CHECK_TIMER(sk);
1636 if (opt_skb)
1637 goto ipv6_pktoptions;
1638 return 0;
1639
1640reset:
cfb6eeb4 1641 tcp_v6_send_reset(sk, skb);
1da177e4
LT
1642discard:
1643 if (opt_skb)
1644 __kfree_skb(opt_skb);
1645 kfree_skb(skb);
1646 return 0;
1647csum_err:
1648 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1649 goto discard;
1650
1651
1652ipv6_pktoptions:
1653 /* Do you ask, what is it?
1654
1655 1. skb was enqueued by tcp.
1656 2. skb is added to tail of read queue, rather than out of order.
1657 3. socket is not in passive state.
1658 4. Finally, it really contains options, which user wants to receive.
1659 */
1660 tp = tcp_sk(sk);
1661 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1662 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
333fad53 1663 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
505cbfc5 1664 np->mcast_oif = inet6_iif(opt_skb);
333fad53 1665 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1da177e4
LT
1666 np->mcast_hops = opt_skb->nh.ipv6h->hop_limit;
1667 if (ipv6_opt_accepted(sk, opt_skb)) {
1668 skb_set_owner_r(opt_skb, sk);
1669 opt_skb = xchg(&np->pktoptions, opt_skb);
1670 } else {
1671 __kfree_skb(opt_skb);
1672 opt_skb = xchg(&np->pktoptions, NULL);
1673 }
1674 }
1675
1676 if (opt_skb)
1677 kfree_skb(opt_skb);
1678 return 0;
1679}
1680
951dbc8a 1681static int tcp_v6_rcv(struct sk_buff **pskb)
1da177e4
LT
1682{
1683 struct sk_buff *skb = *pskb;
1684 struct tcphdr *th;
1685 struct sock *sk;
1686 int ret;
1687
1688 if (skb->pkt_type != PACKET_HOST)
1689 goto discard_it;
1690
1691 /*
1692 * Count it even if it's bad.
1693 */
1694 TCP_INC_STATS_BH(TCP_MIB_INSEGS);
1695
1696 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1697 goto discard_it;
1698
1699 th = skb->h.th;
1700
1701 if (th->doff < sizeof(struct tcphdr)/4)
1702 goto bad_packet;
1703 if (!pskb_may_pull(skb, th->doff*4))
1704 goto discard_it;
1705
1706 if ((skb->ip_summed != CHECKSUM_UNNECESSARY &&
fb286bb2 1707 tcp_v6_checksum_init(skb)))
1da177e4
LT
1708 goto bad_packet;
1709
1710 th = skb->h.th;
1711 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1712 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1713 skb->len - th->doff*4);
1714 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1715 TCP_SKB_CB(skb)->when = 0;
1716 TCP_SKB_CB(skb)->flags = ipv6_get_dsfield(skb->nh.ipv6h);
1717 TCP_SKB_CB(skb)->sacked = 0;
1718
505cbfc5
ACM
1719 sk = __inet6_lookup(&tcp_hashinfo, &skb->nh.ipv6h->saddr, th->source,
1720 &skb->nh.ipv6h->daddr, ntohs(th->dest),
1721 inet6_iif(skb));
1da177e4
LT
1722
1723 if (!sk)
1724 goto no_tcp_socket;
1725
1726process:
1727 if (sk->sk_state == TCP_TIME_WAIT)
1728 goto do_time_wait;
1729
1730 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1731 goto discard_and_relse;
1732
fda9ef5d 1733 if (sk_filter(sk, skb))
1da177e4
LT
1734 goto discard_and_relse;
1735
1736 skb->dev = NULL;
1737
293b9c42 1738 bh_lock_sock_nested(sk);
1da177e4
LT
1739 ret = 0;
1740 if (!sock_owned_by_user(sk)) {
1a2449a8
CL
1741#ifdef CONFIG_NET_DMA
1742 struct tcp_sock *tp = tcp_sk(sk);
1743 if (tp->ucopy.dma_chan)
1744 ret = tcp_v6_do_rcv(sk, skb);
1745 else
1746#endif
1747 {
1748 if (!tcp_prequeue(sk, skb))
1749 ret = tcp_v6_do_rcv(sk, skb);
1750 }
1da177e4
LT
1751 } else
1752 sk_add_backlog(sk, skb);
1753 bh_unlock_sock(sk);
1754
1755 sock_put(sk);
1756 return ret ? -1 : 0;
1757
1758no_tcp_socket:
1759 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1760 goto discard_it;
1761
1762 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1763bad_packet:
1764 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1765 } else {
cfb6eeb4 1766 tcp_v6_send_reset(NULL, skb);
1da177e4
LT
1767 }
1768
1769discard_it:
1770
1771 /*
1772 * Discard frame
1773 */
1774
1775 kfree_skb(skb);
1776 return 0;
1777
1778discard_and_relse:
1779 sock_put(sk);
1780 goto discard_it;
1781
1782do_time_wait:
1783 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
9469c7b4 1784 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1785 goto discard_it;
1786 }
1787
1788 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1789 TCP_INC_STATS_BH(TCP_MIB_INERRS);
9469c7b4 1790 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1791 goto discard_it;
1792 }
1793
9469c7b4 1794 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1da177e4
LT
1795 case TCP_TW_SYN:
1796 {
1797 struct sock *sk2;
1798
505cbfc5
ACM
1799 sk2 = inet6_lookup_listener(&tcp_hashinfo,
1800 &skb->nh.ipv6h->daddr,
1801 ntohs(th->dest), inet6_iif(skb));
1da177e4 1802 if (sk2 != NULL) {
295ff7ed
ACM
1803 struct inet_timewait_sock *tw = inet_twsk(sk);
1804 inet_twsk_deschedule(tw, &tcp_death_row);
1805 inet_twsk_put(tw);
1da177e4
LT
1806 sk = sk2;
1807 goto process;
1808 }
1809 /* Fall through to ACK */
1810 }
1811 case TCP_TW_ACK:
1812 tcp_v6_timewait_ack(sk, skb);
1813 break;
1814 case TCP_TW_RST:
1815 goto no_tcp_socket;
1816 case TCP_TW_SUCCESS:;
1817 }
1818 goto discard_it;
1819}
1820
1da177e4
LT
1821static int tcp_v6_remember_stamp(struct sock *sk)
1822{
1823 /* Alas, not yet... */
1824 return 0;
1825}
1826
8292a17a 1827static struct inet_connection_sock_af_ops ipv6_specific = {
543d9cfe
ACM
1828 .queue_xmit = inet6_csk_xmit,
1829 .send_check = tcp_v6_send_check,
1830 .rebuild_header = inet6_sk_rebuild_header,
1831 .conn_request = tcp_v6_conn_request,
1832 .syn_recv_sock = tcp_v6_syn_recv_sock,
1833 .remember_stamp = tcp_v6_remember_stamp,
1834 .net_header_len = sizeof(struct ipv6hdr),
1835 .setsockopt = ipv6_setsockopt,
1836 .getsockopt = ipv6_getsockopt,
1837 .addr2sockaddr = inet6_csk_addr2sockaddr,
1838 .sockaddr_len = sizeof(struct sockaddr_in6),
3fdadf7d 1839#ifdef CONFIG_COMPAT
543d9cfe
ACM
1840 .compat_setsockopt = compat_ipv6_setsockopt,
1841 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1842#endif
1da177e4
LT
1843};
1844
cfb6eeb4 1845#ifdef CONFIG_TCP_MD5SIG
a928630a 1846static struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
cfb6eeb4
YH
1847 .md5_lookup = tcp_v6_md5_lookup,
1848 .calc_md5_hash = tcp_v6_calc_md5_hash,
1849 .md5_add = tcp_v6_md5_add_func,
1850 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1851};
a928630a 1852#endif
cfb6eeb4 1853
1da177e4
LT
1854/*
1855 * TCP over IPv4 via INET6 API
1856 */
1857
8292a17a 1858static struct inet_connection_sock_af_ops ipv6_mapped = {
543d9cfe
ACM
1859 .queue_xmit = ip_queue_xmit,
1860 .send_check = tcp_v4_send_check,
1861 .rebuild_header = inet_sk_rebuild_header,
1862 .conn_request = tcp_v6_conn_request,
1863 .syn_recv_sock = tcp_v6_syn_recv_sock,
1864 .remember_stamp = tcp_v4_remember_stamp,
1865 .net_header_len = sizeof(struct iphdr),
1866 .setsockopt = ipv6_setsockopt,
1867 .getsockopt = ipv6_getsockopt,
1868 .addr2sockaddr = inet6_csk_addr2sockaddr,
1869 .sockaddr_len = sizeof(struct sockaddr_in6),
3fdadf7d 1870#ifdef CONFIG_COMPAT
543d9cfe
ACM
1871 .compat_setsockopt = compat_ipv6_setsockopt,
1872 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1873#endif
1da177e4
LT
1874};
1875
cfb6eeb4 1876#ifdef CONFIG_TCP_MD5SIG
a928630a 1877static struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
cfb6eeb4
YH
1878 .md5_lookup = tcp_v4_md5_lookup,
1879 .calc_md5_hash = tcp_v4_calc_md5_hash,
1880 .md5_add = tcp_v6_md5_add_func,
1881 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1882};
a928630a 1883#endif
cfb6eeb4 1884
1da177e4
LT
1885/* NOTE: A lot of things set to zero explicitly by call to
1886 * sk_alloc() so need not be done here.
1887 */
1888static int tcp_v6_init_sock(struct sock *sk)
1889{
6687e988 1890 struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4
LT
1891 struct tcp_sock *tp = tcp_sk(sk);
1892
1893 skb_queue_head_init(&tp->out_of_order_queue);
1894 tcp_init_xmit_timers(sk);
1895 tcp_prequeue_init(tp);
1896
6687e988 1897 icsk->icsk_rto = TCP_TIMEOUT_INIT;
1da177e4
LT
1898 tp->mdev = TCP_TIMEOUT_INIT;
1899
1900 /* So many TCP implementations out there (incorrectly) count the
1901 * initial SYN frame in their delayed-ACK and congestion control
1902 * algorithms that we must have the following bandaid to talk
1903 * efficiently to them. -DaveM
1904 */
1905 tp->snd_cwnd = 2;
1906
1907 /* See draft-stevens-tcpca-spec-01 for discussion of the
1908 * initialization of these values.
1909 */
1910 tp->snd_ssthresh = 0x7fffffff;
1911 tp->snd_cwnd_clamp = ~0;
c1b4a7e6 1912 tp->mss_cache = 536;
1da177e4
LT
1913
1914 tp->reordering = sysctl_tcp_reordering;
1915
1916 sk->sk_state = TCP_CLOSE;
1917
8292a17a 1918 icsk->icsk_af_ops = &ipv6_specific;
6687e988 1919 icsk->icsk_ca_ops = &tcp_init_congestion_ops;
d83d8461 1920 icsk->icsk_sync_mss = tcp_sync_mss;
1da177e4
LT
1921 sk->sk_write_space = sk_stream_write_space;
1922 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
1923
cfb6eeb4
YH
1924#ifdef CONFIG_TCP_MD5SIG
1925 tp->af_specific = &tcp_sock_ipv6_specific;
1926#endif
1927
1da177e4
LT
1928 sk->sk_sndbuf = sysctl_tcp_wmem[1];
1929 sk->sk_rcvbuf = sysctl_tcp_rmem[1];
1930
1931 atomic_inc(&tcp_sockets_allocated);
1932
1933 return 0;
1934}
1935
1936static int tcp_v6_destroy_sock(struct sock *sk)
1937{
cfb6eeb4
YH
1938#ifdef CONFIG_TCP_MD5SIG
1939 /* Clean up the MD5 key list */
1940 if (tcp_sk(sk)->md5sig_info)
1941 tcp_v6_clear_md5_list(sk);
1942#endif
1da177e4
LT
1943 tcp_v4_destroy_sock(sk);
1944 return inet6_destroy_sock(sk);
1945}
1946
1947/* Proc filesystem TCPv6 sock list dumping. */
1948static void get_openreq6(struct seq_file *seq,
60236fdd 1949 struct sock *sk, struct request_sock *req, int i, int uid)
1da177e4 1950{
1da177e4 1951 int ttd = req->expires - jiffies;
ca304b61
ACM
1952 struct in6_addr *src = &inet6_rsk(req)->loc_addr;
1953 struct in6_addr *dest = &inet6_rsk(req)->rmt_addr;
1da177e4
LT
1954
1955 if (ttd < 0)
1956 ttd = 0;
1957
1da177e4
LT
1958 seq_printf(seq,
1959 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1960 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
1961 i,
1962 src->s6_addr32[0], src->s6_addr32[1],
1963 src->s6_addr32[2], src->s6_addr32[3],
1964 ntohs(inet_sk(sk)->sport),
1965 dest->s6_addr32[0], dest->s6_addr32[1],
1966 dest->s6_addr32[2], dest->s6_addr32[3],
2e6599cb 1967 ntohs(inet_rsk(req)->rmt_port),
1da177e4
LT
1968 TCP_SYN_RECV,
1969 0,0, /* could print option size, but that is af dependent. */
1970 1, /* timers active (only the expire timer) */
1971 jiffies_to_clock_t(ttd),
1972 req->retrans,
1973 uid,
1974 0, /* non standard timer */
1975 0, /* open_requests have no inode */
1976 0, req);
1977}
1978
1979static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1980{
1981 struct in6_addr *dest, *src;
1982 __u16 destp, srcp;
1983 int timer_active;
1984 unsigned long timer_expires;
1985 struct inet_sock *inet = inet_sk(sp);
1986 struct tcp_sock *tp = tcp_sk(sp);
463c84b9 1987 const struct inet_connection_sock *icsk = inet_csk(sp);
1da177e4
LT
1988 struct ipv6_pinfo *np = inet6_sk(sp);
1989
1990 dest = &np->daddr;
1991 src = &np->rcv_saddr;
1992 destp = ntohs(inet->dport);
1993 srcp = ntohs(inet->sport);
463c84b9
ACM
1994
1995 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
1da177e4 1996 timer_active = 1;
463c84b9
ACM
1997 timer_expires = icsk->icsk_timeout;
1998 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
1da177e4 1999 timer_active = 4;
463c84b9 2000 timer_expires = icsk->icsk_timeout;
1da177e4
LT
2001 } else if (timer_pending(&sp->sk_timer)) {
2002 timer_active = 2;
2003 timer_expires = sp->sk_timer.expires;
2004 } else {
2005 timer_active = 0;
2006 timer_expires = jiffies;
2007 }
2008
2009 seq_printf(seq,
2010 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2011 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %u %u %u %u %d\n",
2012 i,
2013 src->s6_addr32[0], src->s6_addr32[1],
2014 src->s6_addr32[2], src->s6_addr32[3], srcp,
2015 dest->s6_addr32[0], dest->s6_addr32[1],
2016 dest->s6_addr32[2], dest->s6_addr32[3], destp,
2017 sp->sk_state,
47da8ee6
SS
2018 tp->write_seq-tp->snd_una,
2019 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
1da177e4
LT
2020 timer_active,
2021 jiffies_to_clock_t(timer_expires - jiffies),
463c84b9 2022 icsk->icsk_retransmits,
1da177e4 2023 sock_i_uid(sp),
6687e988 2024 icsk->icsk_probes_out,
1da177e4
LT
2025 sock_i_ino(sp),
2026 atomic_read(&sp->sk_refcnt), sp,
463c84b9
ACM
2027 icsk->icsk_rto,
2028 icsk->icsk_ack.ato,
2029 (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
1da177e4
LT
2030 tp->snd_cwnd, tp->snd_ssthresh>=0xFFFF?-1:tp->snd_ssthresh
2031 );
2032}
2033
2034static void get_timewait6_sock(struct seq_file *seq,
8feaf0c0 2035 struct inet_timewait_sock *tw, int i)
1da177e4
LT
2036{
2037 struct in6_addr *dest, *src;
2038 __u16 destp, srcp;
0fa1a53e 2039 struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
1da177e4
LT
2040 int ttd = tw->tw_ttd - jiffies;
2041
2042 if (ttd < 0)
2043 ttd = 0;
2044
0fa1a53e
ACM
2045 dest = &tw6->tw_v6_daddr;
2046 src = &tw6->tw_v6_rcv_saddr;
1da177e4
LT
2047 destp = ntohs(tw->tw_dport);
2048 srcp = ntohs(tw->tw_sport);
2049
2050 seq_printf(seq,
2051 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2052 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
2053 i,
2054 src->s6_addr32[0], src->s6_addr32[1],
2055 src->s6_addr32[2], src->s6_addr32[3], srcp,
2056 dest->s6_addr32[0], dest->s6_addr32[1],
2057 dest->s6_addr32[2], dest->s6_addr32[3], destp,
2058 tw->tw_substate, 0, 0,
2059 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
2060 atomic_read(&tw->tw_refcnt), tw);
2061}
2062
2063#ifdef CONFIG_PROC_FS
2064static int tcp6_seq_show(struct seq_file *seq, void *v)
2065{
2066 struct tcp_iter_state *st;
2067
2068 if (v == SEQ_START_TOKEN) {
2069 seq_puts(seq,
2070 " sl "
2071 "local_address "
2072 "remote_address "
2073 "st tx_queue rx_queue tr tm->when retrnsmt"
2074 " uid timeout inode\n");
2075 goto out;
2076 }
2077 st = seq->private;
2078
2079 switch (st->state) {
2080 case TCP_SEQ_STATE_LISTENING:
2081 case TCP_SEQ_STATE_ESTABLISHED:
2082 get_tcp6_sock(seq, v, st->num);
2083 break;
2084 case TCP_SEQ_STATE_OPENREQ:
2085 get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
2086 break;
2087 case TCP_SEQ_STATE_TIME_WAIT:
2088 get_timewait6_sock(seq, v, st->num);
2089 break;
2090 }
2091out:
2092 return 0;
2093}
2094
2095static struct file_operations tcp6_seq_fops;
2096static struct tcp_seq_afinfo tcp6_seq_afinfo = {
2097 .owner = THIS_MODULE,
2098 .name = "tcp6",
2099 .family = AF_INET6,
2100 .seq_show = tcp6_seq_show,
2101 .seq_fops = &tcp6_seq_fops,
2102};
2103
2104int __init tcp6_proc_init(void)
2105{
2106 return tcp_proc_register(&tcp6_seq_afinfo);
2107}
2108
2109void tcp6_proc_exit(void)
2110{
2111 tcp_proc_unregister(&tcp6_seq_afinfo);
2112}
2113#endif
2114
2115struct proto tcpv6_prot = {
2116 .name = "TCPv6",
2117 .owner = THIS_MODULE,
2118 .close = tcp_close,
2119 .connect = tcp_v6_connect,
2120 .disconnect = tcp_disconnect,
463c84b9 2121 .accept = inet_csk_accept,
1da177e4
LT
2122 .ioctl = tcp_ioctl,
2123 .init = tcp_v6_init_sock,
2124 .destroy = tcp_v6_destroy_sock,
2125 .shutdown = tcp_shutdown,
2126 .setsockopt = tcp_setsockopt,
2127 .getsockopt = tcp_getsockopt,
2128 .sendmsg = tcp_sendmsg,
2129 .recvmsg = tcp_recvmsg,
2130 .backlog_rcv = tcp_v6_do_rcv,
2131 .hash = tcp_v6_hash,
2132 .unhash = tcp_unhash,
2133 .get_port = tcp_v6_get_port,
2134 .enter_memory_pressure = tcp_enter_memory_pressure,
2135 .sockets_allocated = &tcp_sockets_allocated,
2136 .memory_allocated = &tcp_memory_allocated,
2137 .memory_pressure = &tcp_memory_pressure,
0a5578cf 2138 .orphan_count = &tcp_orphan_count,
1da177e4
LT
2139 .sysctl_mem = sysctl_tcp_mem,
2140 .sysctl_wmem = sysctl_tcp_wmem,
2141 .sysctl_rmem = sysctl_tcp_rmem,
2142 .max_header = MAX_TCP_HEADER,
2143 .obj_size = sizeof(struct tcp6_sock),
6d6ee43e 2144 .twsk_prot = &tcp6_timewait_sock_ops,
60236fdd 2145 .rsk_prot = &tcp6_request_sock_ops,
543d9cfe
ACM
2146#ifdef CONFIG_COMPAT
2147 .compat_setsockopt = compat_tcp_setsockopt,
2148 .compat_getsockopt = compat_tcp_getsockopt,
2149#endif
1da177e4
LT
2150};
2151
2152static struct inet6_protocol tcpv6_protocol = {
2153 .handler = tcp_v6_rcv,
2154 .err_handler = tcp_v6_err,
a430a43d 2155 .gso_send_check = tcp_v6_gso_send_check,
adcfc7d0 2156 .gso_segment = tcp_tso_segment,
1da177e4
LT
2157 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
2158};
2159
1da177e4
LT
2160static struct inet_protosw tcpv6_protosw = {
2161 .type = SOCK_STREAM,
2162 .protocol = IPPROTO_TCP,
2163 .prot = &tcpv6_prot,
2164 .ops = &inet6_stream_ops,
2165 .capability = -1,
2166 .no_check = 0,
d83d8461
ACM
2167 .flags = INET_PROTOSW_PERMANENT |
2168 INET_PROTOSW_ICSK,
1da177e4
LT
2169};
2170
2171void __init tcpv6_init(void)
2172{
2173 /* register inet6 protocol */
2174 if (inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP) < 0)
2175 printk(KERN_ERR "tcpv6_init: Could not register protocol\n");
2176 inet6_register_protosw(&tcpv6_protosw);
ae0f7d5f 2177
c4d93909
ACM
2178 if (inet_csk_ctl_sock_create(&tcp6_socket, PF_INET6, SOCK_RAW,
2179 IPPROTO_TCP) < 0)
ae0f7d5f 2180 panic("Failed to create the TCPv6 control socket.\n");
1da177e4 2181}