]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - net/ipv6/tcp_ipv6.c
[IPV6] TCPMD5: Check return value of tcp_alloc_md5sig_pool().
[mirror_ubuntu-jammy-kernel.git] / net / ipv6 / tcp_ipv6.c
CommitLineData
1da177e4
LT
1/*
2 * TCP over IPv6
1ab1457c 3 * Linux INET6 implementation
1da177e4
LT
4 *
5 * Authors:
1ab1457c 6 * Pedro Roque <roque@di.fc.ul.pt>
1da177e4
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>
18134bed 62#include <net/netdma.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);
1ab1457c 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,
1ab1457c
YH
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{
0660e03f
ACM
118 return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
119 ipv6_hdr(skb)->saddr.s6_addr32,
aa8223c7
ACM
120 tcp_hdr(skb)->dest,
121 tcp_hdr(skb)->source);
1da177e4
LT
122}
123
1ab1457c 124static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
1da177e4
LT
125 int addr_len)
126{
127 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
1ab1457c 128 struct inet_sock *inet = inet_sk(sk);
d83d8461 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
1ab1457c 138 if (addr_len < SIN6_LEN_RFC2133)
1da177e4
LT
139 return -EINVAL;
140
1ab1457c 141 if (usin->sin6_family != AF_INET6)
1da177e4
LT
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 /*
1ab1457c
YH
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;
1da177e4
LT
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
14e50e57
DM
268 if ((err = __xfrm_lookup(&dst, &fl, sk, 1)) < 0) {
269 if (err == -EREMOTE)
270 err = ip6_dst_blackhole(sk, &dst, &fl);
271 if (err < 0)
272 goto failure;
273 }
1da177e4
LT
274
275 if (saddr == NULL) {
276 saddr = &fl.fl6_src;
277 ipv6_addr_copy(&np->rcv_saddr, saddr);
278 }
279
280 /* set the source address */
281 ipv6_addr_copy(&np->saddr, saddr);
282 inet->rcv_saddr = LOOPBACK4_IPV6;
283
f83ef8c0 284 sk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 285 __ip6_dst_store(sk, dst, NULL, NULL);
1da177e4 286
d83d8461 287 icsk->icsk_ext_hdr_len = 0;
1da177e4 288 if (np->opt)
d83d8461
ACM
289 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
290 np->opt->opt_nflen);
1da177e4
LT
291
292 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
293
294 inet->dport = usin->sin6_port;
295
296 tcp_set_state(sk, TCP_SYN_SENT);
d8313f5c 297 err = inet6_hash_connect(&tcp_death_row, sk);
1da177e4
LT
298 if (err)
299 goto late_failure;
300
301 if (!tp->write_seq)
302 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
303 np->daddr.s6_addr32,
304 inet->sport,
305 inet->dport);
306
307 err = tcp_connect(sk);
308 if (err)
309 goto late_failure;
310
311 return 0;
312
313late_failure:
314 tcp_set_state(sk, TCP_CLOSE);
315 __sk_dst_reset(sk);
316failure:
317 inet->dport = 0;
318 sk->sk_route_caps = 0;
319 return err;
320}
321
322static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
04ce6909 323 int type, int code, int offset, __be32 info)
1da177e4
LT
324{
325 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
505cbfc5 326 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
1da177e4
LT
327 struct ipv6_pinfo *np;
328 struct sock *sk;
329 int err;
1ab1457c 330 struct tcp_sock *tp;
1da177e4
LT
331 __u32 seq;
332
505cbfc5
ACM
333 sk = inet6_lookup(&tcp_hashinfo, &hdr->daddr, th->dest, &hdr->saddr,
334 th->source, skb->dev->ifindex);
1da177e4
LT
335
336 if (sk == NULL) {
337 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
338 return;
339 }
340
341 if (sk->sk_state == TCP_TIME_WAIT) {
9469c7b4 342 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
343 return;
344 }
345
346 bh_lock_sock(sk);
347 if (sock_owned_by_user(sk))
348 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
349
350 if (sk->sk_state == TCP_CLOSE)
351 goto out;
352
353 tp = tcp_sk(sk);
1ab1457c 354 seq = ntohl(th->seq);
1da177e4
LT
355 if (sk->sk_state != TCP_LISTEN &&
356 !between(seq, tp->snd_una, tp->snd_nxt)) {
357 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
358 goto out;
359 }
360
361 np = inet6_sk(sk);
362
363 if (type == ICMPV6_PKT_TOOBIG) {
364 struct dst_entry *dst = NULL;
365
366 if (sock_owned_by_user(sk))
367 goto out;
368 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
369 goto out;
370
371 /* icmp should have updated the destination cache entry */
372 dst = __sk_dst_check(sk, np->dst_cookie);
373
374 if (dst == NULL) {
375 struct inet_sock *inet = inet_sk(sk);
376 struct flowi fl;
377
378 /* BUGGG_FUTURE: Again, it is not clear how
379 to handle rthdr case. Ignore this complexity
380 for now.
381 */
382 memset(&fl, 0, sizeof(fl));
383 fl.proto = IPPROTO_TCP;
384 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
385 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
386 fl.oif = sk->sk_bound_dev_if;
387 fl.fl_ip_dport = inet->dport;
388 fl.fl_ip_sport = inet->sport;
beb8d13b 389 security_skb_classify_flow(skb, &fl);
1da177e4
LT
390
391 if ((err = ip6_dst_lookup(sk, &dst, &fl))) {
392 sk->sk_err_soft = -err;
393 goto out;
394 }
395
396 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
397 sk->sk_err_soft = -err;
398 goto out;
399 }
400
401 } else
402 dst_hold(dst);
403
d83d8461 404 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
1da177e4
LT
405 tcp_sync_mss(sk, dst_mtu(dst));
406 tcp_simple_retransmit(sk);
407 } /* else let the usual retransmit timer handle it */
408 dst_release(dst);
409 goto out;
410 }
411
412 icmpv6_err_convert(type, code, &err);
413
60236fdd 414 /* Might be for an request_sock */
1da177e4 415 switch (sk->sk_state) {
60236fdd 416 struct request_sock *req, **prev;
1da177e4
LT
417 case TCP_LISTEN:
418 if (sock_owned_by_user(sk))
419 goto out;
420
8129765a
ACM
421 req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
422 &hdr->saddr, inet6_iif(skb));
1da177e4
LT
423 if (!req)
424 goto out;
425
426 /* ICMPs are not backlogged, hence we cannot get
427 * an established socket here.
428 */
429 BUG_TRAP(req->sk == NULL);
430
2e6599cb 431 if (seq != tcp_rsk(req)->snt_isn) {
1da177e4
LT
432 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
433 goto out;
434 }
435
463c84b9 436 inet_csk_reqsk_queue_drop(sk, req, prev);
1da177e4
LT
437 goto out;
438
439 case TCP_SYN_SENT:
440 case TCP_SYN_RECV: /* Cannot happen.
1ab1457c 441 It can, it SYNs are crossed. --ANK */
1da177e4 442 if (!sock_owned_by_user(sk)) {
1da177e4
LT
443 sk->sk_err = err;
444 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
445
446 tcp_done(sk);
447 } else
448 sk->sk_err_soft = err;
449 goto out;
450 }
451
452 if (!sock_owned_by_user(sk) && np->recverr) {
453 sk->sk_err = err;
454 sk->sk_error_report(sk);
455 } else
456 sk->sk_err_soft = err;
457
458out:
459 bh_unlock_sock(sk);
460 sock_put(sk);
461}
462
463
60236fdd 464static int tcp_v6_send_synack(struct sock *sk, struct request_sock *req,
1da177e4
LT
465 struct dst_entry *dst)
466{
ca304b61 467 struct inet6_request_sock *treq = inet6_rsk(req);
1da177e4
LT
468 struct ipv6_pinfo *np = inet6_sk(sk);
469 struct sk_buff * skb;
470 struct ipv6_txoptions *opt = NULL;
471 struct in6_addr * final_p = NULL, final;
472 struct flowi fl;
473 int err = -1;
474
475 memset(&fl, 0, sizeof(fl));
476 fl.proto = IPPROTO_TCP;
2e6599cb
ACM
477 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
478 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
1da177e4 479 fl.fl6_flowlabel = 0;
2e6599cb
ACM
480 fl.oif = treq->iif;
481 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
1da177e4 482 fl.fl_ip_sport = inet_sk(sk)->sport;
4237c75c 483 security_req_classify_flow(req, &fl);
1da177e4
LT
484
485 if (dst == NULL) {
486 opt = np->opt;
1da177e4
LT
487 if (opt && opt->srcrt) {
488 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
489 ipv6_addr_copy(&final, &fl.fl6_dst);
490 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
491 final_p = &final;
492 }
493
494 err = ip6_dst_lookup(sk, &dst, &fl);
495 if (err)
496 goto done;
497 if (final_p)
498 ipv6_addr_copy(&fl.fl6_dst, final_p);
499 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
500 goto done;
501 }
502
503 skb = tcp_make_synack(sk, dst, req);
504 if (skb) {
aa8223c7 505 struct tcphdr *th = tcp_hdr(skb);
1da177e4
LT
506
507 th->check = tcp_v6_check(th, skb->len,
2e6599cb 508 &treq->loc_addr, &treq->rmt_addr,
1da177e4
LT
509 csum_partial((char *)th, skb->len, skb->csum));
510
2e6599cb 511 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
1da177e4 512 err = ip6_xmit(sk, skb, &fl, opt, 0);
b9df3cb8 513 err = net_xmit_eval(err);
1da177e4
LT
514 }
515
516done:
1ab1457c 517 if (opt && opt != np->opt)
1da177e4 518 sock_kfree_s(sk, opt, opt->tot_len);
78b91042 519 dst_release(dst);
1da177e4
LT
520 return err;
521}
522
60236fdd 523static void tcp_v6_reqsk_destructor(struct request_sock *req)
1da177e4 524{
ca304b61
ACM
525 if (inet6_rsk(req)->pktopts)
526 kfree_skb(inet6_rsk(req)->pktopts);
1da177e4
LT
527}
528
cfb6eeb4
YH
529#ifdef CONFIG_TCP_MD5SIG
530static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
531 struct in6_addr *addr)
532{
533 struct tcp_sock *tp = tcp_sk(sk);
534 int i;
535
536 BUG_ON(tp == NULL);
537
538 if (!tp->md5sig_info || !tp->md5sig_info->entries6)
539 return NULL;
540
541 for (i = 0; i < tp->md5sig_info->entries6; i++) {
542 if (ipv6_addr_cmp(&tp->md5sig_info->keys6[i].addr, addr) == 0)
f8ab18d2 543 return &tp->md5sig_info->keys6[i].base;
cfb6eeb4
YH
544 }
545 return NULL;
546}
547
548static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
549 struct sock *addr_sk)
550{
551 return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr);
552}
553
554static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
555 struct request_sock *req)
556{
557 return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr);
558}
559
560static int tcp_v6_md5_do_add(struct sock *sk, struct in6_addr *peer,
561 char *newkey, u8 newkeylen)
562{
563 /* Add key to the list */
b0a713e9 564 struct tcp_md5sig_key *key;
cfb6eeb4
YH
565 struct tcp_sock *tp = tcp_sk(sk);
566 struct tcp6_md5sig_key *keys;
567
b0a713e9 568 key = tcp_v6_md5_do_lookup(sk, peer);
cfb6eeb4
YH
569 if (key) {
570 /* modify existing entry - just update that one */
b0a713e9
MD
571 kfree(key->key);
572 key->key = newkey;
573 key->keylen = newkeylen;
cfb6eeb4
YH
574 } else {
575 /* reallocate new list if current one is full. */
576 if (!tp->md5sig_info) {
577 tp->md5sig_info = kzalloc(sizeof(*tp->md5sig_info), GFP_ATOMIC);
578 if (!tp->md5sig_info) {
579 kfree(newkey);
580 return -ENOMEM;
581 }
3d7dbeac 582 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
cfb6eeb4 583 }
aacbe8c8
YH
584 if (tcp_alloc_md5sig_pool() == NULL) {
585 kfree(newkey);
586 return -ENOMEM;
587 }
cfb6eeb4
YH
588 if (tp->md5sig_info->alloced6 == tp->md5sig_info->entries6) {
589 keys = kmalloc((sizeof (tp->md5sig_info->keys6[0]) *
590 (tp->md5sig_info->entries6 + 1)), GFP_ATOMIC);
591
592 if (!keys) {
593 tcp_free_md5sig_pool();
594 kfree(newkey);
595 return -ENOMEM;
596 }
597
598 if (tp->md5sig_info->entries6)
599 memmove(keys, tp->md5sig_info->keys6,
600 (sizeof (tp->md5sig_info->keys6[0]) *
601 tp->md5sig_info->entries6));
602
603 kfree(tp->md5sig_info->keys6);
604 tp->md5sig_info->keys6 = keys;
605 tp->md5sig_info->alloced6++;
606 }
607
608 ipv6_addr_copy(&tp->md5sig_info->keys6[tp->md5sig_info->entries6].addr,
609 peer);
f8ab18d2
DM
610 tp->md5sig_info->keys6[tp->md5sig_info->entries6].base.key = newkey;
611 tp->md5sig_info->keys6[tp->md5sig_info->entries6].base.keylen = newkeylen;
cfb6eeb4
YH
612
613 tp->md5sig_info->entries6++;
614 }
615 return 0;
616}
617
618static int tcp_v6_md5_add_func(struct sock *sk, struct sock *addr_sk,
619 u8 *newkey, __u8 newkeylen)
620{
621 return tcp_v6_md5_do_add(sk, &inet6_sk(addr_sk)->daddr,
622 newkey, newkeylen);
623}
624
625static int tcp_v6_md5_do_del(struct sock *sk, struct in6_addr *peer)
626{
627 struct tcp_sock *tp = tcp_sk(sk);
628 int i;
629
630 for (i = 0; i < tp->md5sig_info->entries6; i++) {
631 if (ipv6_addr_cmp(&tp->md5sig_info->keys6[i].addr, peer) == 0) {
632 /* Free the key */
f8ab18d2 633 kfree(tp->md5sig_info->keys6[i].base.key);
cfb6eeb4
YH
634 tp->md5sig_info->entries6--;
635
636 if (tp->md5sig_info->entries6 == 0) {
637 kfree(tp->md5sig_info->keys6);
638 tp->md5sig_info->keys6 = NULL;
ca983cef 639 tp->md5sig_info->alloced6 = 0;
cfb6eeb4
YH
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++)
f8ab18d2 664 kfree(tp->md5sig_info->keys6[i].base.key);
cfb6eeb4
YH
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++)
f8ab18d2 675 kfree(tp->md5sig_info->keys4[i].base.key);
cfb6eeb4
YH
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;
e773e4fa 704 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
cfb6eeb4
YH
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;
3d7dbeac 721 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
cfb6eeb4
YH
722 }
723
af879cc7 724 newkey = kmemdup(cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
cfb6eeb4
YH
725 if (!newkey)
726 return -ENOMEM;
e773e4fa 727 if (ipv6_addr_v4mapped(&sin6->sin6_addr)) {
cfb6eeb4
YH
728 return tcp_v4_md5_do_add(sk, sin6->sin6_addr.s6_addr32[3],
729 newkey, cmd.tcpm_keylen);
730 }
731 return tcp_v6_md5_do_add(sk, &sin6->sin6_addr, newkey, cmd.tcpm_keylen);
732}
733
734static int tcp_v6_do_calc_md5_hash(char *md5_hash, struct tcp_md5sig_key *key,
735 struct in6_addr *saddr,
736 struct in6_addr *daddr,
737 struct tcphdr *th, int protocol,
738 int tcplen)
739{
740 struct scatterlist sg[4];
741 __u16 data_len;
742 int block = 0;
8e5200f5 743 __sum16 cksum;
cfb6eeb4
YH
744 struct tcp_md5sig_pool *hp;
745 struct tcp6_pseudohdr *bp;
746 struct hash_desc *desc;
747 int err;
748 unsigned int nbytes = 0;
749
750 hp = tcp_get_md5sig_pool();
751 if (!hp) {
752 printk(KERN_WARNING "%s(): hash pool not found...\n", __FUNCTION__);
753 goto clear_hash_noput;
754 }
755 bp = &hp->md5_blk.ip6;
756 desc = &hp->md5_desc;
757
758 /* 1. TCP pseudo-header (RFC2460) */
759 ipv6_addr_copy(&bp->saddr, saddr);
760 ipv6_addr_copy(&bp->daddr, daddr);
761 bp->len = htonl(tcplen);
762 bp->protocol = htonl(protocol);
763
c7da57a1
DM
764 sg_init_table(sg, 4);
765
cfb6eeb4
YH
766 sg_set_buf(&sg[block++], bp, sizeof(*bp));
767 nbytes += sizeof(*bp);
768
769 /* 2. TCP header, excluding options */
770 cksum = th->check;
771 th->check = 0;
772 sg_set_buf(&sg[block++], th, sizeof(*th));
773 nbytes += sizeof(*th);
774
775 /* 3. TCP segment data (if any) */
776 data_len = tcplen - (th->doff << 2);
777 if (data_len > 0) {
778 u8 *data = (u8 *)th + (th->doff << 2);
779 sg_set_buf(&sg[block++], data, data_len);
780 nbytes += data_len;
781 }
782
783 /* 4. shared key */
784 sg_set_buf(&sg[block++], key->key, key->keylen);
785 nbytes += key->keylen;
786
c46f2334 787 sg_mark_end(&sg[block - 1]);
c7da57a1 788
cfb6eeb4
YH
789 /* Now store the hash into the packet */
790 err = crypto_hash_init(desc);
791 if (err) {
792 printk(KERN_WARNING "%s(): hash_init failed\n", __FUNCTION__);
793 goto clear_hash;
794 }
795 err = crypto_hash_update(desc, sg, nbytes);
796 if (err) {
797 printk(KERN_WARNING "%s(): hash_update failed\n", __FUNCTION__);
798 goto clear_hash;
799 }
800 err = crypto_hash_final(desc, md5_hash);
801 if (err) {
802 printk(KERN_WARNING "%s(): hash_final failed\n", __FUNCTION__);
803 goto clear_hash;
804 }
805
806 /* Reset header, and free up the crypto */
807 tcp_put_md5sig_pool();
808 th->check = cksum;
809out:
810 return 0;
811clear_hash:
812 tcp_put_md5sig_pool();
813clear_hash_noput:
814 memset(md5_hash, 0, 16);
815 goto out;
816}
817
818static int tcp_v6_calc_md5_hash(char *md5_hash, struct tcp_md5sig_key *key,
819 struct sock *sk,
820 struct dst_entry *dst,
821 struct request_sock *req,
822 struct tcphdr *th, int protocol,
823 int tcplen)
824{
825 struct in6_addr *saddr, *daddr;
826
827 if (sk) {
828 saddr = &inet6_sk(sk)->saddr;
829 daddr = &inet6_sk(sk)->daddr;
830 } else {
831 saddr = &inet6_rsk(req)->loc_addr;
832 daddr = &inet6_rsk(req)->rmt_addr;
833 }
834 return tcp_v6_do_calc_md5_hash(md5_hash, key,
835 saddr, daddr,
836 th, protocol, tcplen);
837}
838
839static int tcp_v6_inbound_md5_hash (struct sock *sk, struct sk_buff *skb)
840{
841 __u8 *hash_location = NULL;
842 struct tcp_md5sig_key *hash_expected;
0660e03f 843 struct ipv6hdr *ip6h = ipv6_hdr(skb);
aa8223c7 844 struct tcphdr *th = tcp_hdr(skb);
cfb6eeb4
YH
845 int length = (th->doff << 2) - sizeof (*th);
846 int genhash;
847 u8 *ptr;
848 u8 newhash[16];
849
850 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
851
852 /* If the TCP option is too short, we can short cut */
853 if (length < TCPOLEN_MD5SIG)
854 return hash_expected ? 1 : 0;
855
856 /* parse options */
857 ptr = (u8*)(th + 1);
858 while (length > 0) {
859 int opcode = *ptr++;
860 int opsize;
861
862 switch(opcode) {
863 case TCPOPT_EOL:
864 goto done_opts;
865 case TCPOPT_NOP:
866 length--;
867 continue;
868 default:
869 opsize = *ptr++;
870 if (opsize < 2 || opsize > length)
871 goto done_opts;
872 if (opcode == TCPOPT_MD5SIG) {
873 hash_location = ptr;
874 goto done_opts;
875 }
876 }
877 ptr += opsize - 2;
878 length -= opsize;
879 }
880
881done_opts:
882 /* do we have a hash as expected? */
883 if (!hash_expected) {
884 if (!hash_location)
885 return 0;
886 if (net_ratelimit()) {
887 printk(KERN_INFO "MD5 Hash NOT expected but found "
888 "(" NIP6_FMT ", %u)->"
889 "(" NIP6_FMT ", %u)\n",
890 NIP6(ip6h->saddr), ntohs(th->source),
891 NIP6(ip6h->daddr), ntohs(th->dest));
892 }
893 return 1;
894 }
895
896 if (!hash_location) {
897 if (net_ratelimit()) {
898 printk(KERN_INFO "MD5 Hash expected but NOT found "
899 "(" NIP6_FMT ", %u)->"
900 "(" NIP6_FMT ", %u)\n",
901 NIP6(ip6h->saddr), ntohs(th->source),
902 NIP6(ip6h->daddr), ntohs(th->dest));
903 }
904 return 1;
905 }
906
907 /* check the signature */
908 genhash = tcp_v6_do_calc_md5_hash(newhash,
909 hash_expected,
910 &ip6h->saddr, &ip6h->daddr,
911 th, sk->sk_protocol,
912 skb->len);
913 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
914 if (net_ratelimit()) {
915 printk(KERN_INFO "MD5 Hash %s for "
916 "(" NIP6_FMT ", %u)->"
917 "(" NIP6_FMT ", %u)\n",
918 genhash ? "failed" : "mismatch",
919 NIP6(ip6h->saddr), ntohs(th->source),
920 NIP6(ip6h->daddr), ntohs(th->dest));
921 }
922 return 1;
923 }
924 return 0;
925}
926#endif
927
9ec75fe8 928static struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
1da177e4 929 .family = AF_INET6,
2e6599cb 930 .obj_size = sizeof(struct tcp6_request_sock),
1da177e4 931 .rtx_syn_ack = tcp_v6_send_synack,
60236fdd
ACM
932 .send_ack = tcp_v6_reqsk_send_ack,
933 .destructor = tcp_v6_reqsk_destructor,
1da177e4
LT
934 .send_reset = tcp_v6_send_reset
935};
936
cfb6eeb4 937#ifdef CONFIG_TCP_MD5SIG
b6332e6c 938static struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
cfb6eeb4 939 .md5_lookup = tcp_v6_reqsk_md5_lookup,
cfb6eeb4 940};
b6332e6c 941#endif
cfb6eeb4 942
6d6ee43e
ACM
943static struct timewait_sock_ops tcp6_timewait_sock_ops = {
944 .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
945 .twsk_unique = tcp_twsk_unique,
cfb6eeb4 946 .twsk_destructor= tcp_twsk_destructor,
6d6ee43e
ACM
947};
948
8292a17a 949static void tcp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
1da177e4
LT
950{
951 struct ipv6_pinfo *np = inet6_sk(sk);
aa8223c7 952 struct tcphdr *th = tcp_hdr(skb);
1da177e4 953
84fa7933 954 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1da177e4 955 th->check = ~csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP, 0);
663ead3b 956 skb->csum_start = skb_transport_header(skb) - skb->head;
ff1dcadb 957 skb->csum_offset = offsetof(struct tcphdr, check);
1da177e4 958 } else {
1ab1457c
YH
959 th->check = csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP,
960 csum_partial((char *)th, th->doff<<2,
1da177e4
LT
961 skb->csum));
962 }
963}
964
a430a43d
HX
965static int tcp_v6_gso_send_check(struct sk_buff *skb)
966{
967 struct ipv6hdr *ipv6h;
968 struct tcphdr *th;
969
970 if (!pskb_may_pull(skb, sizeof(*th)))
971 return -EINVAL;
972
0660e03f 973 ipv6h = ipv6_hdr(skb);
aa8223c7 974 th = tcp_hdr(skb);
a430a43d
HX
975
976 th->check = 0;
977 th->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
978 IPPROTO_TCP, 0);
663ead3b 979 skb->csum_start = skb_transport_header(skb) - skb->head;
ff1dcadb 980 skb->csum_offset = offsetof(struct tcphdr, check);
84fa7933 981 skb->ip_summed = CHECKSUM_PARTIAL;
a430a43d
HX
982 return 0;
983}
1da177e4 984
cfb6eeb4 985static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
1da177e4 986{
aa8223c7 987 struct tcphdr *th = tcp_hdr(skb), *t1;
1da177e4
LT
988 struct sk_buff *buff;
989 struct flowi fl;
cfb6eeb4
YH
990 int tot_len = sizeof(*th);
991#ifdef CONFIG_TCP_MD5SIG
992 struct tcp_md5sig_key *key;
993#endif
1da177e4
LT
994
995 if (th->rst)
996 return;
997
998 if (!ipv6_unicast_destination(skb))
1ab1457c 999 return;
1da177e4 1000
cfb6eeb4
YH
1001#ifdef CONFIG_TCP_MD5SIG
1002 if (sk)
0660e03f 1003 key = tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr);
cfb6eeb4
YH
1004 else
1005 key = NULL;
1006
1007 if (key)
1008 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1009#endif
1010
1da177e4
LT
1011 /*
1012 * We need to grab some memory, and put together an RST,
1013 * and then put it into the queue to be sent.
1014 */
1015
cfb6eeb4 1016 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1da177e4 1017 GFP_ATOMIC);
1ab1457c
YH
1018 if (buff == NULL)
1019 return;
1da177e4 1020
cfb6eeb4 1021 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1da177e4 1022
cfb6eeb4 1023 t1 = (struct tcphdr *) skb_push(buff, tot_len);
1da177e4
LT
1024
1025 /* Swap the send and the receive. */
1026 memset(t1, 0, sizeof(*t1));
1027 t1->dest = th->source;
1028 t1->source = th->dest;
cfb6eeb4 1029 t1->doff = tot_len / 4;
1da177e4 1030 t1->rst = 1;
1ab1457c 1031
1da177e4 1032 if(th->ack) {
1ab1457c 1033 t1->seq = th->ack_seq;
1da177e4
LT
1034 } else {
1035 t1->ack = 1;
1036 t1->ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin
1037 + skb->len - (th->doff<<2));
1038 }
1039
cfb6eeb4
YH
1040#ifdef CONFIG_TCP_MD5SIG
1041 if (key) {
8e5200f5 1042 __be32 *opt = (__be32*)(t1 + 1);
cfb6eeb4
YH
1043 opt[0] = htonl((TCPOPT_NOP << 24) |
1044 (TCPOPT_NOP << 16) |
1045 (TCPOPT_MD5SIG << 8) |
1046 TCPOLEN_MD5SIG);
0660e03f
ACM
1047 tcp_v6_do_calc_md5_hash((__u8 *)&opt[1], key,
1048 &ipv6_hdr(skb)->daddr,
1049 &ipv6_hdr(skb)->saddr,
1050 t1, IPPROTO_TCP, tot_len);
cfb6eeb4
YH
1051 }
1052#endif
1053
1da177e4
LT
1054 buff->csum = csum_partial((char *)t1, sizeof(*t1), 0);
1055
1056 memset(&fl, 0, sizeof(fl));
0660e03f
ACM
1057 ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
1058 ipv6_addr_copy(&fl.fl6_src, &ipv6_hdr(skb)->daddr);
1da177e4
LT
1059
1060 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1061 sizeof(*t1), IPPROTO_TCP,
1062 buff->csum);
1063
1064 fl.proto = IPPROTO_TCP;
505cbfc5 1065 fl.oif = inet6_iif(skb);
1da177e4
LT
1066 fl.fl_ip_dport = t1->dest;
1067 fl.fl_ip_sport = t1->source;
beb8d13b 1068 security_skb_classify_flow(skb, &fl);
1da177e4
LT
1069
1070 /* sk = NULL, but it is safe for now. RST socket required. */
1071 if (!ip6_dst_lookup(NULL, &buff->dst, &fl)) {
1072
ecc51b6d 1073 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
ae0f7d5f 1074 ip6_xmit(tcp6_socket->sk, buff, &fl, NULL, 0);
ecc51b6d
ACM
1075 TCP_INC_STATS_BH(TCP_MIB_OUTSEGS);
1076 TCP_INC_STATS_BH(TCP_MIB_OUTRSTS);
1da177e4 1077 return;
ecc51b6d 1078 }
1da177e4
LT
1079 }
1080
1081 kfree_skb(buff);
1082}
1083
cfb6eeb4
YH
1084static void tcp_v6_send_ack(struct tcp_timewait_sock *tw,
1085 struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts)
1da177e4 1086{
aa8223c7 1087 struct tcphdr *th = tcp_hdr(skb), *t1;
1da177e4
LT
1088 struct sk_buff *buff;
1089 struct flowi fl;
1090 int tot_len = sizeof(struct tcphdr);
e69a4adc 1091 __be32 *topt;
cfb6eeb4
YH
1092#ifdef CONFIG_TCP_MD5SIG
1093 struct tcp_md5sig_key *key;
1094 struct tcp_md5sig_key tw_key;
1095#endif
1096
1097#ifdef CONFIG_TCP_MD5SIG
1098 if (!tw && skb->sk) {
0660e03f 1099 key = tcp_v6_md5_do_lookup(skb->sk, &ipv6_hdr(skb)->daddr);
cfb6eeb4
YH
1100 } else if (tw && tw->tw_md5_keylen) {
1101 tw_key.key = tw->tw_md5_key;
1102 tw_key.keylen = tw->tw_md5_keylen;
1103 key = &tw_key;
1104 } else {
1105 key = NULL;
1106 }
1107#endif
1da177e4
LT
1108
1109 if (ts)
4244f8a9 1110 tot_len += TCPOLEN_TSTAMP_ALIGNED;
cfb6eeb4
YH
1111#ifdef CONFIG_TCP_MD5SIG
1112 if (key)
1113 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1114#endif
1da177e4
LT
1115
1116 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1117 GFP_ATOMIC);
1118 if (buff == NULL)
1119 return;
1120
1121 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1122
1123 t1 = (struct tcphdr *) skb_push(buff,tot_len);
1124
1125 /* Swap the send and the receive. */
1126 memset(t1, 0, sizeof(*t1));
1127 t1->dest = th->source;
1128 t1->source = th->dest;
1129 t1->doff = tot_len/4;
1130 t1->seq = htonl(seq);
1131 t1->ack_seq = htonl(ack);
1132 t1->ack = 1;
1133 t1->window = htons(win);
cfb6eeb4 1134
e69a4adc 1135 topt = (__be32 *)(t1 + 1);
1ab1457c 1136
1da177e4 1137 if (ts) {
cfb6eeb4
YH
1138 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1139 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
1140 *topt++ = htonl(tcp_time_stamp);
1141 *topt = htonl(ts);
1da177e4
LT
1142 }
1143
cfb6eeb4
YH
1144#ifdef CONFIG_TCP_MD5SIG
1145 if (key) {
1146 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1147 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
0660e03f
ACM
1148 tcp_v6_do_calc_md5_hash((__u8 *)topt, key,
1149 &ipv6_hdr(skb)->daddr,
1150 &ipv6_hdr(skb)->saddr,
1151 t1, IPPROTO_TCP, tot_len);
cfb6eeb4
YH
1152 }
1153#endif
1154
1da177e4
LT
1155 buff->csum = csum_partial((char *)t1, tot_len, 0);
1156
1157 memset(&fl, 0, sizeof(fl));
0660e03f
ACM
1158 ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
1159 ipv6_addr_copy(&fl.fl6_src, &ipv6_hdr(skb)->daddr);
1da177e4
LT
1160
1161 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1162 tot_len, IPPROTO_TCP,
1163 buff->csum);
1164
1165 fl.proto = IPPROTO_TCP;
505cbfc5 1166 fl.oif = inet6_iif(skb);
1da177e4
LT
1167 fl.fl_ip_dport = t1->dest;
1168 fl.fl_ip_sport = t1->source;
beb8d13b 1169 security_skb_classify_flow(skb, &fl);
1da177e4
LT
1170
1171 if (!ip6_dst_lookup(NULL, &buff->dst, &fl)) {
ecc51b6d 1172 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
ae0f7d5f 1173 ip6_xmit(tcp6_socket->sk, buff, &fl, NULL, 0);
ecc51b6d 1174 TCP_INC_STATS_BH(TCP_MIB_OUTSEGS);
1da177e4 1175 return;
ecc51b6d 1176 }
1da177e4
LT
1177 }
1178
1179 kfree_skb(buff);
1180}
1181
1182static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
1183{
8feaf0c0 1184 struct inet_timewait_sock *tw = inet_twsk(sk);
cfb6eeb4 1185 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
1da177e4 1186
cfb6eeb4 1187 tcp_v6_send_ack(tcptw, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
8feaf0c0
ACM
1188 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
1189 tcptw->tw_ts_recent);
1da177e4 1190
8feaf0c0 1191 inet_twsk_put(tw);
1da177e4
LT
1192}
1193
60236fdd 1194static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req)
1da177e4 1195{
cfb6eeb4 1196 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
1197}
1198
1199
1200static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
1201{
60236fdd 1202 struct request_sock *req, **prev;
aa8223c7 1203 const struct tcphdr *th = tcp_hdr(skb);
1da177e4
LT
1204 struct sock *nsk;
1205
1206 /* Find possible connection requests. */
8129765a 1207 req = inet6_csk_search_req(sk, &prev, th->source,
0660e03f
ACM
1208 &ipv6_hdr(skb)->saddr,
1209 &ipv6_hdr(skb)->daddr, inet6_iif(skb));
1da177e4
LT
1210 if (req)
1211 return tcp_check_req(sk, skb, req, prev);
1212
0660e03f
ACM
1213 nsk = __inet6_lookup_established(&tcp_hashinfo, &ipv6_hdr(skb)->saddr,
1214 th->source, &ipv6_hdr(skb)->daddr,
505cbfc5 1215 ntohs(th->dest), inet6_iif(skb));
1da177e4
LT
1216
1217 if (nsk) {
1218 if (nsk->sk_state != TCP_TIME_WAIT) {
1219 bh_lock_sock(nsk);
1220 return nsk;
1221 }
9469c7b4 1222 inet_twsk_put(inet_twsk(nsk));
1da177e4
LT
1223 return NULL;
1224 }
1225
1226#if 0 /*def CONFIG_SYN_COOKIES*/
1227 if (!th->rst && !th->syn && th->ack)
1228 sk = cookie_v6_check(sk, skb, &(IPCB(skb)->opt));
1229#endif
1230 return sk;
1231}
1232
1da177e4
LT
1233/* FIXME: this is substantially similar to the ipv4 code.
1234 * Can some kind of merge be done? -- erics
1235 */
1236static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
1237{
ca304b61 1238 struct inet6_request_sock *treq;
1da177e4
LT
1239 struct ipv6_pinfo *np = inet6_sk(sk);
1240 struct tcp_options_received tmp_opt;
1241 struct tcp_sock *tp = tcp_sk(sk);
60236fdd 1242 struct request_sock *req = NULL;
1da177e4
LT
1243 __u32 isn = TCP_SKB_CB(skb)->when;
1244
1245 if (skb->protocol == htons(ETH_P_IP))
1246 return tcp_v4_conn_request(sk, skb);
1247
1248 if (!ipv6_unicast_destination(skb))
1ab1457c 1249 goto drop;
1da177e4
LT
1250
1251 /*
1ab1457c 1252 * There are no SYN attacks on IPv6, yet...
1da177e4 1253 */
463c84b9 1254 if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
1da177e4
LT
1255 if (net_ratelimit())
1256 printk(KERN_INFO "TCPv6: dropping request, synflood is possible\n");
1ab1457c 1257 goto drop;
1da177e4
LT
1258 }
1259
463c84b9 1260 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
1da177e4
LT
1261 goto drop;
1262
ca304b61 1263 req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
1da177e4
LT
1264 if (req == NULL)
1265 goto drop;
1266
cfb6eeb4
YH
1267#ifdef CONFIG_TCP_MD5SIG
1268 tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
1269#endif
1270
1da177e4
LT
1271 tcp_clear_options(&tmp_opt);
1272 tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
1273 tmp_opt.user_mss = tp->rx_opt.user_mss;
1274
1275 tcp_parse_options(skb, &tmp_opt, 0);
1276
1277 tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
1278 tcp_openreq_init(req, &tmp_opt, skb);
1279
ca304b61 1280 treq = inet6_rsk(req);
0660e03f
ACM
1281 ipv6_addr_copy(&treq->rmt_addr, &ipv6_hdr(skb)->saddr);
1282 ipv6_addr_copy(&treq->loc_addr, &ipv6_hdr(skb)->daddr);
aa8223c7 1283 TCP_ECN_create_request(req, tcp_hdr(skb));
2e6599cb 1284 treq->pktopts = NULL;
1da177e4 1285 if (ipv6_opt_accepted(sk, skb) ||
333fad53
YH
1286 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
1287 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
1da177e4 1288 atomic_inc(&skb->users);
2e6599cb 1289 treq->pktopts = skb;
1da177e4 1290 }
2e6599cb 1291 treq->iif = sk->sk_bound_dev_if;
1da177e4
LT
1292
1293 /* So that link locals have meaning */
1294 if (!sk->sk_bound_dev_if &&
2e6599cb 1295 ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL)
505cbfc5 1296 treq->iif = inet6_iif(skb);
1da177e4 1297
1ab1457c 1298 if (isn == 0)
a94f723d 1299 isn = tcp_v6_init_sequence(skb);
1da177e4 1300
2e6599cb 1301 tcp_rsk(req)->snt_isn = isn;
1da177e4 1302
4237c75c
VY
1303 security_inet_conn_request(sk, skb, req);
1304
1da177e4
LT
1305 if (tcp_v6_send_synack(sk, req, NULL))
1306 goto drop;
1307
8129765a 1308 inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
1da177e4
LT
1309 return 0;
1310
1311drop:
1312 if (req)
60236fdd 1313 reqsk_free(req);
1da177e4 1314
1da177e4
LT
1315 return 0; /* don't send reset */
1316}
1317
1318static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
60236fdd 1319 struct request_sock *req,
1da177e4
LT
1320 struct dst_entry *dst)
1321{
ca304b61 1322 struct inet6_request_sock *treq = inet6_rsk(req);
1da177e4
LT
1323 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
1324 struct tcp6_sock *newtcp6sk;
1325 struct inet_sock *newinet;
1326 struct tcp_sock *newtp;
1327 struct sock *newsk;
1328 struct ipv6_txoptions *opt;
cfb6eeb4
YH
1329#ifdef CONFIG_TCP_MD5SIG
1330 struct tcp_md5sig_key *key;
1331#endif
1da177e4
LT
1332
1333 if (skb->protocol == htons(ETH_P_IP)) {
1334 /*
1335 * v6 mapped
1336 */
1337
1338 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1339
1ab1457c 1340 if (newsk == NULL)
1da177e4
LT
1341 return NULL;
1342
1343 newtcp6sk = (struct tcp6_sock *)newsk;
1344 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1345
1346 newinet = inet_sk(newsk);
1347 newnp = inet6_sk(newsk);
1348 newtp = tcp_sk(newsk);
1349
1350 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1351
1352 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
1353 newinet->daddr);
1354
1355 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
1356 newinet->saddr);
1357
1358 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
1359
8292a17a 1360 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1da177e4 1361 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
cfb6eeb4
YH
1362#ifdef CONFIG_TCP_MD5SIG
1363 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1364#endif
1365
1da177e4
LT
1366 newnp->pktoptions = NULL;
1367 newnp->opt = NULL;
505cbfc5 1368 newnp->mcast_oif = inet6_iif(skb);
0660e03f 1369 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1da177e4 1370
e6848976
ACM
1371 /*
1372 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1373 * here, tcp_create_openreq_child now does this for us, see the comment in
1374 * that function for the gory details. -acme
1da177e4 1375 */
1da177e4
LT
1376
1377 /* It is tricky place. Until this moment IPv4 tcp
8292a17a 1378 worked with IPv6 icsk.icsk_af_ops.
1da177e4
LT
1379 Sync it now.
1380 */
d83d8461 1381 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1da177e4
LT
1382
1383 return newsk;
1384 }
1385
1386 opt = np->opt;
1387
1388 if (sk_acceptq_is_full(sk))
1389 goto out_overflow;
1390
1da177e4
LT
1391 if (dst == NULL) {
1392 struct in6_addr *final_p = NULL, final;
1393 struct flowi fl;
1394
1395 memset(&fl, 0, sizeof(fl));
1396 fl.proto = IPPROTO_TCP;
2e6599cb 1397 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
1da177e4
LT
1398 if (opt && opt->srcrt) {
1399 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
1400 ipv6_addr_copy(&final, &fl.fl6_dst);
1401 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
1402 final_p = &final;
1403 }
2e6599cb 1404 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
1da177e4 1405 fl.oif = sk->sk_bound_dev_if;
2e6599cb 1406 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
1da177e4 1407 fl.fl_ip_sport = inet_sk(sk)->sport;
4237c75c 1408 security_req_classify_flow(req, &fl);
1da177e4
LT
1409
1410 if (ip6_dst_lookup(sk, &dst, &fl))
1411 goto out;
1412
1413 if (final_p)
1414 ipv6_addr_copy(&fl.fl6_dst, final_p);
1415
1416 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
1417 goto out;
1ab1457c 1418 }
1da177e4
LT
1419
1420 newsk = tcp_create_openreq_child(sk, req, skb);
1421 if (newsk == NULL)
1422 goto out;
1423
e6848976
ACM
1424 /*
1425 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1426 * count here, tcp_create_openreq_child now does this for us, see the
1427 * comment in that function for the gory details. -acme
1428 */
1da177e4 1429
59eed279 1430 newsk->sk_gso_type = SKB_GSO_TCPV6;
8e1ef0a9 1431 __ip6_dst_store(newsk, dst, NULL, NULL);
1da177e4
LT
1432
1433 newtcp6sk = (struct tcp6_sock *)newsk;
1434 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1435
1436 newtp = tcp_sk(newsk);
1437 newinet = inet_sk(newsk);
1438 newnp = inet6_sk(newsk);
1439
1440 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1441
2e6599cb
ACM
1442 ipv6_addr_copy(&newnp->daddr, &treq->rmt_addr);
1443 ipv6_addr_copy(&newnp->saddr, &treq->loc_addr);
1444 ipv6_addr_copy(&newnp->rcv_saddr, &treq->loc_addr);
1445 newsk->sk_bound_dev_if = treq->iif;
1da177e4 1446
1ab1457c 1447 /* Now IPv6 options...
1da177e4
LT
1448
1449 First: no IPv4 options.
1450 */
1451 newinet->opt = NULL;
d35690be 1452 newnp->ipv6_fl_list = NULL;
1da177e4
LT
1453
1454 /* Clone RX bits */
1455 newnp->rxopt.all = np->rxopt.all;
1456
1457 /* Clone pktoptions received with SYN */
1458 newnp->pktoptions = NULL;
2e6599cb
ACM
1459 if (treq->pktopts != NULL) {
1460 newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC);
1461 kfree_skb(treq->pktopts);
1462 treq->pktopts = NULL;
1da177e4
LT
1463 if (newnp->pktoptions)
1464 skb_set_owner_r(newnp->pktoptions, newsk);
1465 }
1466 newnp->opt = NULL;
505cbfc5 1467 newnp->mcast_oif = inet6_iif(skb);
0660e03f 1468 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1da177e4
LT
1469
1470 /* Clone native IPv6 options from listening socket (if any)
1471
1472 Yes, keeping reference count would be much more clever,
1473 but we make one more one thing there: reattach optmem
1474 to newsk.
1475 */
1476 if (opt) {
1477 newnp->opt = ipv6_dup_options(newsk, opt);
1478 if (opt != np->opt)
1479 sock_kfree_s(sk, opt, opt->tot_len);
1480 }
1481
d83d8461 1482 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1da177e4 1483 if (newnp->opt)
d83d8461
ACM
1484 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1485 newnp->opt->opt_flen);
1da177e4 1486
5d424d5a 1487 tcp_mtup_init(newsk);
1da177e4
LT
1488 tcp_sync_mss(newsk, dst_mtu(dst));
1489 newtp->advmss = dst_metric(dst, RTAX_ADVMSS);
1490 tcp_initialize_rcv_mss(newsk);
1491
1492 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
1493
cfb6eeb4
YH
1494#ifdef CONFIG_TCP_MD5SIG
1495 /* Copy over the MD5 key from the original socket */
1496 if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) {
1497 /* We're using one, so create a matching key
1498 * on the newsk structure. If we fail to get
1499 * memory, then we end up not copying the key
1500 * across. Shucks.
1501 */
af879cc7
ACM
1502 char *newkey = kmemdup(key->key, key->keylen, GFP_ATOMIC);
1503 if (newkey != NULL)
cfb6eeb4
YH
1504 tcp_v6_md5_do_add(newsk, &inet6_sk(sk)->daddr,
1505 newkey, key->keylen);
cfb6eeb4
YH
1506 }
1507#endif
1508
90b19d31 1509 __inet6_hash(&tcp_hashinfo, newsk);
2d8c4ce5 1510 inet_inherit_port(&tcp_hashinfo, sk, newsk);
1da177e4
LT
1511
1512 return newsk;
1513
1514out_overflow:
1515 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
1516out:
1517 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
1518 if (opt && opt != np->opt)
1519 sock_kfree_s(sk, opt, opt->tot_len);
1520 dst_release(dst);
1521 return NULL;
1522}
1523
b51655b9 1524static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
1da177e4 1525{
84fa7933 1526 if (skb->ip_summed == CHECKSUM_COMPLETE) {
aa8223c7 1527 if (!tcp_v6_check(tcp_hdr(skb), skb->len, &ipv6_hdr(skb)->saddr,
0660e03f 1528 &ipv6_hdr(skb)->daddr, skb->csum)) {
fb286bb2 1529 skb->ip_summed = CHECKSUM_UNNECESSARY;
1da177e4 1530 return 0;
fb286bb2 1531 }
1da177e4 1532 }
fb286bb2 1533
aa8223c7 1534 skb->csum = ~csum_unfold(tcp_v6_check(tcp_hdr(skb), skb->len,
0660e03f
ACM
1535 &ipv6_hdr(skb)->saddr,
1536 &ipv6_hdr(skb)->daddr, 0));
fb286bb2 1537
1da177e4 1538 if (skb->len <= 76) {
fb286bb2 1539 return __skb_checksum_complete(skb);
1da177e4
LT
1540 }
1541 return 0;
1542}
1543
1544/* The socket must have it's spinlock held when we get
1545 * here.
1546 *
1547 * We have a potential double-lock case here, so even when
1548 * doing backlog processing we use the BH locking scheme.
1549 * This is because we cannot sleep with the original spinlock
1550 * held.
1551 */
1552static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1553{
1554 struct ipv6_pinfo *np = inet6_sk(sk);
1555 struct tcp_sock *tp;
1556 struct sk_buff *opt_skb = NULL;
1557
1558 /* Imagine: socket is IPv6. IPv4 packet arrives,
1559 goes to IPv4 receive handler and backlogged.
1560 From backlog it always goes here. Kerboom...
1561 Fortunately, tcp_rcv_established and rcv_established
1562 handle them correctly, but it is not case with
1563 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1564 */
1565
1566 if (skb->protocol == htons(ETH_P_IP))
1567 return tcp_v4_do_rcv(sk, skb);
1568
cfb6eeb4
YH
1569#ifdef CONFIG_TCP_MD5SIG
1570 if (tcp_v6_inbound_md5_hash (sk, skb))
1571 goto discard;
1572#endif
1573
fda9ef5d 1574 if (sk_filter(sk, skb))
1da177e4
LT
1575 goto discard;
1576
1577 /*
1578 * socket locking is here for SMP purposes as backlog rcv
1579 * is currently called with bh processing disabled.
1580 */
1581
1582 /* Do Stevens' IPV6_PKTOPTIONS.
1583
1584 Yes, guys, it is the only place in our code, where we
1585 may make it not affecting IPv4.
1586 The rest of code is protocol independent,
1587 and I do not like idea to uglify IPv4.
1588
1589 Actually, all the idea behind IPV6_PKTOPTIONS
1590 looks not very well thought. For now we latch
1591 options, received in the last packet, enqueued
1592 by tcp. Feel free to propose better solution.
1ab1457c 1593 --ANK (980728)
1da177e4
LT
1594 */
1595 if (np->rxopt.all)
1596 opt_skb = skb_clone(skb, GFP_ATOMIC);
1597
1598 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1599 TCP_CHECK_TIMER(sk);
aa8223c7 1600 if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len))
1da177e4
LT
1601 goto reset;
1602 TCP_CHECK_TIMER(sk);
1603 if (opt_skb)
1604 goto ipv6_pktoptions;
1605 return 0;
1606 }
1607
ab6a5bb6 1608 if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
1da177e4
LT
1609 goto csum_err;
1610
1ab1457c 1611 if (sk->sk_state == TCP_LISTEN) {
1da177e4
LT
1612 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1613 if (!nsk)
1614 goto discard;
1615
1616 /*
1617 * Queue it on the new socket if the new socket is active,
1618 * otherwise we just shortcircuit this and continue with
1619 * the new socket..
1620 */
1ab1457c 1621 if(nsk != sk) {
1da177e4
LT
1622 if (tcp_child_process(sk, nsk, skb))
1623 goto reset;
1624 if (opt_skb)
1625 __kfree_skb(opt_skb);
1626 return 0;
1627 }
1628 }
1629
1630 TCP_CHECK_TIMER(sk);
aa8223c7 1631 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
1da177e4
LT
1632 goto reset;
1633 TCP_CHECK_TIMER(sk);
1634 if (opt_skb)
1635 goto ipv6_pktoptions;
1636 return 0;
1637
1638reset:
cfb6eeb4 1639 tcp_v6_send_reset(sk, skb);
1da177e4
LT
1640discard:
1641 if (opt_skb)
1642 __kfree_skb(opt_skb);
1643 kfree_skb(skb);
1644 return 0;
1645csum_err:
1646 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1647 goto discard;
1648
1649
1650ipv6_pktoptions:
1651 /* Do you ask, what is it?
1652
1653 1. skb was enqueued by tcp.
1654 2. skb is added to tail of read queue, rather than out of order.
1655 3. socket is not in passive state.
1656 4. Finally, it really contains options, which user wants to receive.
1657 */
1658 tp = tcp_sk(sk);
1659 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1660 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
333fad53 1661 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
505cbfc5 1662 np->mcast_oif = inet6_iif(opt_skb);
333fad53 1663 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
0660e03f 1664 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1da177e4
LT
1665 if (ipv6_opt_accepted(sk, opt_skb)) {
1666 skb_set_owner_r(opt_skb, sk);
1667 opt_skb = xchg(&np->pktoptions, opt_skb);
1668 } else {
1669 __kfree_skb(opt_skb);
1670 opt_skb = xchg(&np->pktoptions, NULL);
1671 }
1672 }
1673
1674 if (opt_skb)
1675 kfree_skb(opt_skb);
1676 return 0;
1677}
1678
e5bbef20 1679static int tcp_v6_rcv(struct sk_buff *skb)
1da177e4 1680{
1ab1457c 1681 struct tcphdr *th;
1da177e4
LT
1682 struct sock *sk;
1683 int ret;
1684
1685 if (skb->pkt_type != PACKET_HOST)
1686 goto discard_it;
1687
1688 /*
1689 * Count it even if it's bad.
1690 */
1691 TCP_INC_STATS_BH(TCP_MIB_INSEGS);
1692
1693 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1694 goto discard_it;
1695
aa8223c7 1696 th = tcp_hdr(skb);
1da177e4
LT
1697
1698 if (th->doff < sizeof(struct tcphdr)/4)
1699 goto bad_packet;
1700 if (!pskb_may_pull(skb, th->doff*4))
1701 goto discard_it;
1702
60476372 1703 if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
1da177e4
LT
1704 goto bad_packet;
1705
aa8223c7 1706 th = tcp_hdr(skb);
1da177e4
LT
1707 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1708 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1709 skb->len - th->doff*4);
1710 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1711 TCP_SKB_CB(skb)->when = 0;
0660e03f 1712 TCP_SKB_CB(skb)->flags = ipv6_get_dsfield(ipv6_hdr(skb));
1da177e4
LT
1713 TCP_SKB_CB(skb)->sacked = 0;
1714
0660e03f
ACM
1715 sk = __inet6_lookup(&tcp_hashinfo, &ipv6_hdr(skb)->saddr, th->source,
1716 &ipv6_hdr(skb)->daddr, ntohs(th->dest),
505cbfc5 1717 inet6_iif(skb));
1da177e4
LT
1718
1719 if (!sk)
1720 goto no_tcp_socket;
1721
1722process:
1723 if (sk->sk_state == TCP_TIME_WAIT)
1724 goto do_time_wait;
1725
1726 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1727 goto discard_and_relse;
1728
fda9ef5d 1729 if (sk_filter(sk, skb))
1da177e4
LT
1730 goto discard_and_relse;
1731
1732 skb->dev = NULL;
1733
293b9c42 1734 bh_lock_sock_nested(sk);
1da177e4
LT
1735 ret = 0;
1736 if (!sock_owned_by_user(sk)) {
1a2449a8 1737#ifdef CONFIG_NET_DMA
1ab1457c 1738 struct tcp_sock *tp = tcp_sk(sk);
b4caea8a
DM
1739 if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
1740 tp->ucopy.dma_chan = get_softnet_dma();
1ab1457c
YH
1741 if (tp->ucopy.dma_chan)
1742 ret = tcp_v6_do_rcv(sk, skb);
1743 else
1a2449a8
CL
1744#endif
1745 {
1746 if (!tcp_prequeue(sk, skb))
1747 ret = tcp_v6_do_rcv(sk, skb);
1748 }
1da177e4
LT
1749 } else
1750 sk_add_backlog(sk, skb);
1751 bh_unlock_sock(sk);
1752
1753 sock_put(sk);
1754 return ret ? -1 : 0;
1755
1756no_tcp_socket:
1757 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1758 goto discard_it;
1759
1760 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1761bad_packet:
1762 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1763 } else {
cfb6eeb4 1764 tcp_v6_send_reset(NULL, skb);
1da177e4
LT
1765 }
1766
1767discard_it:
1768
1769 /*
1770 * Discard frame
1771 */
1772
1773 kfree_skb(skb);
1774 return 0;
1775
1776discard_and_relse:
1777 sock_put(sk);
1778 goto discard_it;
1779
1780do_time_wait:
1781 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
9469c7b4 1782 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1783 goto discard_it;
1784 }
1785
1786 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1787 TCP_INC_STATS_BH(TCP_MIB_INERRS);
9469c7b4 1788 inet_twsk_put(inet_twsk(sk));
1da177e4
LT
1789 goto discard_it;
1790 }
1791
9469c7b4 1792 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1da177e4
LT
1793 case TCP_TW_SYN:
1794 {
1795 struct sock *sk2;
1796
505cbfc5 1797 sk2 = inet6_lookup_listener(&tcp_hashinfo,
0660e03f 1798 &ipv6_hdr(skb)->daddr,
505cbfc5 1799 ntohs(th->dest), inet6_iif(skb));
1da177e4 1800 if (sk2 != NULL) {
295ff7ed
ACM
1801 struct inet_timewait_sock *tw = inet_twsk(sk);
1802 inet_twsk_deschedule(tw, &tcp_death_row);
1803 inet_twsk_put(tw);
1da177e4
LT
1804 sk = sk2;
1805 goto process;
1806 }
1807 /* Fall through to ACK */
1808 }
1809 case TCP_TW_ACK:
1810 tcp_v6_timewait_ack(sk, skb);
1811 break;
1812 case TCP_TW_RST:
1813 goto no_tcp_socket;
1814 case TCP_TW_SUCCESS:;
1815 }
1816 goto discard_it;
1817}
1818
1da177e4
LT
1819static int tcp_v6_remember_stamp(struct sock *sk)
1820{
1821 /* Alas, not yet... */
1822 return 0;
1823}
1824
8292a17a 1825static struct inet_connection_sock_af_ops ipv6_specific = {
543d9cfe
ACM
1826 .queue_xmit = inet6_csk_xmit,
1827 .send_check = tcp_v6_send_check,
1828 .rebuild_header = inet6_sk_rebuild_header,
1829 .conn_request = tcp_v6_conn_request,
1830 .syn_recv_sock = tcp_v6_syn_recv_sock,
1831 .remember_stamp = tcp_v6_remember_stamp,
1832 .net_header_len = sizeof(struct ipv6hdr),
1833 .setsockopt = ipv6_setsockopt,
1834 .getsockopt = ipv6_getsockopt,
1835 .addr2sockaddr = inet6_csk_addr2sockaddr,
1836 .sockaddr_len = sizeof(struct sockaddr_in6),
3fdadf7d 1837#ifdef CONFIG_COMPAT
543d9cfe
ACM
1838 .compat_setsockopt = compat_ipv6_setsockopt,
1839 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1840#endif
1da177e4
LT
1841};
1842
cfb6eeb4 1843#ifdef CONFIG_TCP_MD5SIG
a928630a 1844static struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
cfb6eeb4
YH
1845 .md5_lookup = tcp_v6_md5_lookup,
1846 .calc_md5_hash = tcp_v6_calc_md5_hash,
1847 .md5_add = tcp_v6_md5_add_func,
1848 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1849};
a928630a 1850#endif
cfb6eeb4 1851
1da177e4
LT
1852/*
1853 * TCP over IPv4 via INET6 API
1854 */
1855
8292a17a 1856static struct inet_connection_sock_af_ops ipv6_mapped = {
543d9cfe
ACM
1857 .queue_xmit = ip_queue_xmit,
1858 .send_check = tcp_v4_send_check,
1859 .rebuild_header = inet_sk_rebuild_header,
1860 .conn_request = tcp_v6_conn_request,
1861 .syn_recv_sock = tcp_v6_syn_recv_sock,
1862 .remember_stamp = tcp_v4_remember_stamp,
1863 .net_header_len = sizeof(struct iphdr),
1864 .setsockopt = ipv6_setsockopt,
1865 .getsockopt = ipv6_getsockopt,
1866 .addr2sockaddr = inet6_csk_addr2sockaddr,
1867 .sockaddr_len = sizeof(struct sockaddr_in6),
3fdadf7d 1868#ifdef CONFIG_COMPAT
543d9cfe
ACM
1869 .compat_setsockopt = compat_ipv6_setsockopt,
1870 .compat_getsockopt = compat_ipv6_getsockopt,
3fdadf7d 1871#endif
1da177e4
LT
1872};
1873
cfb6eeb4 1874#ifdef CONFIG_TCP_MD5SIG
a928630a 1875static struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
cfb6eeb4
YH
1876 .md5_lookup = tcp_v4_md5_lookup,
1877 .calc_md5_hash = tcp_v4_calc_md5_hash,
1878 .md5_add = tcp_v6_md5_add_func,
1879 .md5_parse = tcp_v6_parse_md5_keys,
cfb6eeb4 1880};
a928630a 1881#endif
cfb6eeb4 1882
1da177e4
LT
1883/* NOTE: A lot of things set to zero explicitly by call to
1884 * sk_alloc() so need not be done here.
1885 */
1886static int tcp_v6_init_sock(struct sock *sk)
1887{
6687e988 1888 struct inet_connection_sock *icsk = inet_csk(sk);
1da177e4
LT
1889 struct tcp_sock *tp = tcp_sk(sk);
1890
1891 skb_queue_head_init(&tp->out_of_order_queue);
1892 tcp_init_xmit_timers(sk);
1893 tcp_prequeue_init(tp);
1894
6687e988 1895 icsk->icsk_rto = TCP_TIMEOUT_INIT;
1da177e4
LT
1896 tp->mdev = TCP_TIMEOUT_INIT;
1897
1898 /* So many TCP implementations out there (incorrectly) count the
1899 * initial SYN frame in their delayed-ACK and congestion control
1900 * algorithms that we must have the following bandaid to talk
1901 * efficiently to them. -DaveM
1902 */
1903 tp->snd_cwnd = 2;
1904
1905 /* See draft-stevens-tcpca-spec-01 for discussion of the
1906 * initialization of these values.
1907 */
1908 tp->snd_ssthresh = 0x7fffffff;
1909 tp->snd_cwnd_clamp = ~0;
c1b4a7e6 1910 tp->mss_cache = 536;
1da177e4
LT
1911
1912 tp->reordering = sysctl_tcp_reordering;
1913
1914 sk->sk_state = TCP_CLOSE;
1915
8292a17a 1916 icsk->icsk_af_ops = &ipv6_specific;
6687e988 1917 icsk->icsk_ca_ops = &tcp_init_congestion_ops;
d83d8461 1918 icsk->icsk_sync_mss = tcp_sync_mss;
1da177e4
LT
1919 sk->sk_write_space = sk_stream_write_space;
1920 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
1921
cfb6eeb4
YH
1922#ifdef CONFIG_TCP_MD5SIG
1923 tp->af_specific = &tcp_sock_ipv6_specific;
1924#endif
1925
1da177e4
LT
1926 sk->sk_sndbuf = sysctl_tcp_wmem[1];
1927 sk->sk_rcvbuf = sysctl_tcp_rmem[1];
1928
1929 atomic_inc(&tcp_sockets_allocated);
1930
1931 return 0;
1932}
1933
1934static int tcp_v6_destroy_sock(struct sock *sk)
1935{
cfb6eeb4
YH
1936#ifdef CONFIG_TCP_MD5SIG
1937 /* Clean up the MD5 key list */
1938 if (tcp_sk(sk)->md5sig_info)
1939 tcp_v6_clear_md5_list(sk);
1940#endif
1da177e4
LT
1941 tcp_v4_destroy_sock(sk);
1942 return inet6_destroy_sock(sk);
1943}
1944
952a10be 1945#ifdef CONFIG_PROC_FS
1da177e4 1946/* Proc filesystem TCPv6 sock list dumping. */
1ab1457c 1947static void get_openreq6(struct seq_file *seq,
60236fdd 1948 struct sock *sk, struct request_sock *req, int i, int uid)
1da177e4 1949{
1da177e4 1950 int ttd = req->expires - jiffies;
ca304b61
ACM
1951 struct in6_addr *src = &inet6_rsk(req)->loc_addr;
1952 struct in6_addr *dest = &inet6_rsk(req)->rmt_addr;
1da177e4
LT
1953
1954 if (ttd < 0)
1955 ttd = 0;
1956
1da177e4
LT
1957 seq_printf(seq,
1958 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1959 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
1960 i,
1961 src->s6_addr32[0], src->s6_addr32[1],
1962 src->s6_addr32[2], src->s6_addr32[3],
1963 ntohs(inet_sk(sk)->sport),
1964 dest->s6_addr32[0], dest->s6_addr32[1],
1965 dest->s6_addr32[2], dest->s6_addr32[3],
2e6599cb 1966 ntohs(inet_rsk(req)->rmt_port),
1da177e4
LT
1967 TCP_SYN_RECV,
1968 0,0, /* could print option size, but that is af dependent. */
1ab1457c
YH
1969 1, /* timers active (only the expire timer) */
1970 jiffies_to_clock_t(ttd),
1da177e4
LT
1971 req->retrans,
1972 uid,
1ab1457c 1973 0, /* non standard timer */
1da177e4
LT
1974 0, /* open_requests have no inode */
1975 0, req);
1976}
1977
1978static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1979{
1980 struct in6_addr *dest, *src;
1981 __u16 destp, srcp;
1982 int timer_active;
1983 unsigned long timer_expires;
1984 struct inet_sock *inet = inet_sk(sp);
1985 struct tcp_sock *tp = tcp_sk(sp);
463c84b9 1986 const struct inet_connection_sock *icsk = inet_csk(sp);
1da177e4
LT
1987 struct ipv6_pinfo *np = inet6_sk(sp);
1988
1989 dest = &np->daddr;
1990 src = &np->rcv_saddr;
1991 destp = ntohs(inet->dport);
1992 srcp = ntohs(inet->sport);
463c84b9
ACM
1993
1994 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
1da177e4 1995 timer_active = 1;
463c84b9
ACM
1996 timer_expires = icsk->icsk_timeout;
1997 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
1da177e4 1998 timer_active = 4;
463c84b9 1999 timer_expires = icsk->icsk_timeout;
1da177e4
LT
2000 } else if (timer_pending(&sp->sk_timer)) {
2001 timer_active = 2;
2002 timer_expires = sp->sk_timer.expires;
2003 } else {
2004 timer_active = 0;
2005 timer_expires = jiffies;
2006 }
2007
2008 seq_printf(seq,
2009 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2010 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %u %u %u %u %d\n",
2011 i,
2012 src->s6_addr32[0], src->s6_addr32[1],
2013 src->s6_addr32[2], src->s6_addr32[3], srcp,
2014 dest->s6_addr32[0], dest->s6_addr32[1],
2015 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1ab1457c 2016 sp->sk_state,
47da8ee6
SS
2017 tp->write_seq-tp->snd_una,
2018 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
1da177e4
LT
2019 timer_active,
2020 jiffies_to_clock_t(timer_expires - jiffies),
463c84b9 2021 icsk->icsk_retransmits,
1da177e4 2022 sock_i_uid(sp),
6687e988 2023 icsk->icsk_probes_out,
1da177e4
LT
2024 sock_i_ino(sp),
2025 atomic_read(&sp->sk_refcnt), sp,
463c84b9
ACM
2026 icsk->icsk_rto,
2027 icsk->icsk_ack.ato,
2028 (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
1da177e4
LT
2029 tp->snd_cwnd, tp->snd_ssthresh>=0xFFFF?-1:tp->snd_ssthresh
2030 );
2031}
2032
1ab1457c 2033static void get_timewait6_sock(struct seq_file *seq,
8feaf0c0 2034 struct inet_timewait_sock *tw, int i)
1da177e4
LT
2035{
2036 struct in6_addr *dest, *src;
2037 __u16 destp, srcp;
0fa1a53e 2038 struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
1da177e4
LT
2039 int ttd = tw->tw_ttd - jiffies;
2040
2041 if (ttd < 0)
2042 ttd = 0;
2043
0fa1a53e
ACM
2044 dest = &tw6->tw_v6_daddr;
2045 src = &tw6->tw_v6_rcv_saddr;
1da177e4
LT
2046 destp = ntohs(tw->tw_dport);
2047 srcp = ntohs(tw->tw_sport);
2048
2049 seq_printf(seq,
2050 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2051 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
2052 i,
2053 src->s6_addr32[0], src->s6_addr32[1],
2054 src->s6_addr32[2], src->s6_addr32[3], srcp,
2055 dest->s6_addr32[0], dest->s6_addr32[1],
2056 dest->s6_addr32[2], dest->s6_addr32[3], destp,
2057 tw->tw_substate, 0, 0,
2058 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
2059 atomic_read(&tw->tw_refcnt), tw);
2060}
2061
1da177e4
LT
2062static int tcp6_seq_show(struct seq_file *seq, void *v)
2063{
2064 struct tcp_iter_state *st;
2065
2066 if (v == SEQ_START_TOKEN) {
2067 seq_puts(seq,
2068 " sl "
2069 "local_address "
2070 "remote_address "
2071 "st tx_queue rx_queue tr tm->when retrnsmt"
2072 " uid timeout inode\n");
2073 goto out;
2074 }
2075 st = seq->private;
2076
2077 switch (st->state) {
2078 case TCP_SEQ_STATE_LISTENING:
2079 case TCP_SEQ_STATE_ESTABLISHED:
2080 get_tcp6_sock(seq, v, st->num);
2081 break;
2082 case TCP_SEQ_STATE_OPENREQ:
2083 get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
2084 break;
2085 case TCP_SEQ_STATE_TIME_WAIT:
2086 get_timewait6_sock(seq, v, st->num);
2087 break;
2088 }
2089out:
2090 return 0;
2091}
2092
2093static struct file_operations tcp6_seq_fops;
2094static struct tcp_seq_afinfo tcp6_seq_afinfo = {
2095 .owner = THIS_MODULE,
2096 .name = "tcp6",
2097 .family = AF_INET6,
2098 .seq_show = tcp6_seq_show,
2099 .seq_fops = &tcp6_seq_fops,
2100};
2101
2102int __init tcp6_proc_init(void)
2103{
2104 return tcp_proc_register(&tcp6_seq_afinfo);
2105}
2106
2107void tcp6_proc_exit(void)
2108{
2109 tcp_proc_unregister(&tcp6_seq_afinfo);
2110}
2111#endif
2112
c5a432f1
ED
2113DEFINE_PROTO_INUSE(tcpv6)
2114
1da177e4
LT
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,
1da177e4
LT
2128 .recvmsg = tcp_recvmsg,
2129 .backlog_rcv = tcp_v6_do_rcv,
2130 .hash = tcp_v6_hash,
2131 .unhash = tcp_unhash,
2132 .get_port = tcp_v6_get_port,
2133 .enter_memory_pressure = tcp_enter_memory_pressure,
2134 .sockets_allocated = &tcp_sockets_allocated,
2135 .memory_allocated = &tcp_memory_allocated,
2136 .memory_pressure = &tcp_memory_pressure,
0a5578cf 2137 .orphan_count = &tcp_orphan_count,
1da177e4
LT
2138 .sysctl_mem = sysctl_tcp_mem,
2139 .sysctl_wmem = sysctl_tcp_wmem,
2140 .sysctl_rmem = sysctl_tcp_rmem,
2141 .max_header = MAX_TCP_HEADER,
2142 .obj_size = sizeof(struct tcp6_sock),
6d6ee43e 2143 .twsk_prot = &tcp6_timewait_sock_ops,
60236fdd 2144 .rsk_prot = &tcp6_request_sock_ops,
543d9cfe
ACM
2145#ifdef CONFIG_COMPAT
2146 .compat_setsockopt = compat_tcp_setsockopt,
2147 .compat_getsockopt = compat_tcp_getsockopt,
2148#endif
c5a432f1 2149 REF_PROTO_INUSE(tcpv6)
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}