Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv6 / tcp_ipv6.c
1 /*
2 * TCP over IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Based on:
9 * linux/net/ipv4/tcp.c
10 * linux/net/ipv4/tcp_input.c
11 * linux/net/ipv4/tcp_output.c
12 *
13 * Fixes:
14 * Hideaki YOSHIFUJI : sin6_scope_id support
15 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
16 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
17 * a single port at the same time.
18 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/tcp6 to seq_file.
19 *
20 * This program is free software; you can redistribute it and/or
21 * modify it under the terms of the GNU General Public License
22 * as published by the Free Software Foundation; either version
23 * 2 of the License, or (at your option) any later version.
24 */
25
26 #include <linux/module.h>
27 #include <linux/errno.h>
28 #include <linux/types.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/net.h>
32 #include <linux/jiffies.h>
33 #include <linux/in.h>
34 #include <linux/in6.h>
35 #include <linux/netdevice.h>
36 #include <linux/init.h>
37 #include <linux/jhash.h>
38 #include <linux/ipsec.h>
39 #include <linux/times.h>
40
41 #include <linux/ipv6.h>
42 #include <linux/icmpv6.h>
43 #include <linux/random.h>
44
45 #include <net/tcp.h>
46 #include <net/ndisc.h>
47 #include <net/inet6_hashtables.h>
48 #include <net/inet6_connection_sock.h>
49 #include <net/ipv6.h>
50 #include <net/transp_v6.h>
51 #include <net/addrconf.h>
52 #include <net/ip6_route.h>
53 #include <net/ip6_checksum.h>
54 #include <net/inet_ecn.h>
55 #include <net/protocol.h>
56 #include <net/xfrm.h>
57 #include <net/snmp.h>
58 #include <net/dsfield.h>
59 #include <net/timewait_sock.h>
60 #include <net/netdma.h>
61 #include <net/inet_common.h>
62
63 #include <asm/uaccess.h>
64
65 #include <linux/proc_fs.h>
66 #include <linux/seq_file.h>
67
68 #include <linux/crypto.h>
69 #include <linux/scatterlist.h>
70
71 static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb);
72 static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req);
73
74 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
75
76 static struct inet_connection_sock_af_ops ipv6_mapped;
77 static struct inet_connection_sock_af_ops ipv6_specific;
78 #ifdef CONFIG_TCP_MD5SIG
79 static struct tcp_sock_af_ops tcp_sock_ipv6_specific;
80 static struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
81 #else
82 static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
83 struct in6_addr *addr)
84 {
85 return NULL;
86 }
87 #endif
88
89 static void tcp_v6_hash(struct sock *sk)
90 {
91 if (sk->sk_state != TCP_CLOSE) {
92 if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
93 tcp_prot.hash(sk);
94 return;
95 }
96 local_bh_disable();
97 __inet6_hash(sk);
98 local_bh_enable();
99 }
100 }
101
102 static __inline__ __sum16 tcp_v6_check(struct tcphdr *th, int len,
103 struct in6_addr *saddr,
104 struct in6_addr *daddr,
105 __wsum base)
106 {
107 return csum_ipv6_magic(saddr, daddr, len, IPPROTO_TCP, base);
108 }
109
110 static __u32 tcp_v6_init_sequence(struct sk_buff *skb)
111 {
112 return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
113 ipv6_hdr(skb)->saddr.s6_addr32,
114 tcp_hdr(skb)->dest,
115 tcp_hdr(skb)->source);
116 }
117
118 static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
119 int addr_len)
120 {
121 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
122 struct inet_sock *inet = inet_sk(sk);
123 struct inet_connection_sock *icsk = inet_csk(sk);
124 struct ipv6_pinfo *np = inet6_sk(sk);
125 struct tcp_sock *tp = tcp_sk(sk);
126 struct in6_addr *saddr = NULL, *final_p = NULL, final;
127 struct flowi fl;
128 struct dst_entry *dst;
129 int addr_type;
130 int err;
131
132 if (addr_len < SIN6_LEN_RFC2133)
133 return -EINVAL;
134
135 if (usin->sin6_family != AF_INET6)
136 return(-EAFNOSUPPORT);
137
138 memset(&fl, 0, sizeof(fl));
139
140 if (np->sndflow) {
141 fl.fl6_flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
142 IP6_ECN_flow_init(fl.fl6_flowlabel);
143 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
144 struct ip6_flowlabel *flowlabel;
145 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
146 if (flowlabel == NULL)
147 return -EINVAL;
148 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
149 fl6_sock_release(flowlabel);
150 }
151 }
152
153 /*
154 * connect() to INADDR_ANY means loopback (BSD'ism).
155 */
156
157 if(ipv6_addr_any(&usin->sin6_addr))
158 usin->sin6_addr.s6_addr[15] = 0x1;
159
160 addr_type = ipv6_addr_type(&usin->sin6_addr);
161
162 if(addr_type & IPV6_ADDR_MULTICAST)
163 return -ENETUNREACH;
164
165 if (addr_type&IPV6_ADDR_LINKLOCAL) {
166 if (addr_len >= sizeof(struct sockaddr_in6) &&
167 usin->sin6_scope_id) {
168 /* If interface is set while binding, indices
169 * must coincide.
170 */
171 if (sk->sk_bound_dev_if &&
172 sk->sk_bound_dev_if != usin->sin6_scope_id)
173 return -EINVAL;
174
175 sk->sk_bound_dev_if = usin->sin6_scope_id;
176 }
177
178 /* Connect to link-local address requires an interface */
179 if (!sk->sk_bound_dev_if)
180 return -EINVAL;
181 }
182
183 if (tp->rx_opt.ts_recent_stamp &&
184 !ipv6_addr_equal(&np->daddr, &usin->sin6_addr)) {
185 tp->rx_opt.ts_recent = 0;
186 tp->rx_opt.ts_recent_stamp = 0;
187 tp->write_seq = 0;
188 }
189
190 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
191 np->flow_label = fl.fl6_flowlabel;
192
193 /*
194 * TCP over IPv4
195 */
196
197 if (addr_type == IPV6_ADDR_MAPPED) {
198 u32 exthdrlen = icsk->icsk_ext_hdr_len;
199 struct sockaddr_in sin;
200
201 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
202
203 if (__ipv6_only_sock(sk))
204 return -ENETUNREACH;
205
206 sin.sin_family = AF_INET;
207 sin.sin_port = usin->sin6_port;
208 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
209
210 icsk->icsk_af_ops = &ipv6_mapped;
211 sk->sk_backlog_rcv = tcp_v4_do_rcv;
212 #ifdef CONFIG_TCP_MD5SIG
213 tp->af_specific = &tcp_sock_ipv6_mapped_specific;
214 #endif
215
216 err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
217
218 if (err) {
219 icsk->icsk_ext_hdr_len = exthdrlen;
220 icsk->icsk_af_ops = &ipv6_specific;
221 sk->sk_backlog_rcv = tcp_v6_do_rcv;
222 #ifdef CONFIG_TCP_MD5SIG
223 tp->af_specific = &tcp_sock_ipv6_specific;
224 #endif
225 goto failure;
226 } else {
227 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
228 inet->saddr);
229 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
230 inet->rcv_saddr);
231 }
232
233 return err;
234 }
235
236 if (!ipv6_addr_any(&np->rcv_saddr))
237 saddr = &np->rcv_saddr;
238
239 fl.proto = IPPROTO_TCP;
240 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
241 ipv6_addr_copy(&fl.fl6_src,
242 (saddr ? saddr : &np->saddr));
243 fl.oif = sk->sk_bound_dev_if;
244 fl.fl_ip_dport = usin->sin6_port;
245 fl.fl_ip_sport = inet->sport;
246
247 if (np->opt && np->opt->srcrt) {
248 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
249 ipv6_addr_copy(&final, &fl.fl6_dst);
250 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
251 final_p = &final;
252 }
253
254 security_sk_classify_flow(sk, &fl);
255
256 err = ip6_dst_lookup(sk, &dst, &fl);
257 if (err)
258 goto failure;
259 if (final_p)
260 ipv6_addr_copy(&fl.fl6_dst, final_p);
261
262 if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 0) {
263 if (err == -EREMOTE)
264 err = ip6_dst_blackhole(sk, &dst, &fl);
265 if (err < 0)
266 goto failure;
267 }
268
269 if (saddr == NULL) {
270 saddr = &fl.fl6_src;
271 ipv6_addr_copy(&np->rcv_saddr, saddr);
272 }
273
274 /* set the source address */
275 ipv6_addr_copy(&np->saddr, saddr);
276 inet->rcv_saddr = LOOPBACK4_IPV6;
277
278 sk->sk_gso_type = SKB_GSO_TCPV6;
279 __ip6_dst_store(sk, dst, NULL, NULL);
280
281 icsk->icsk_ext_hdr_len = 0;
282 if (np->opt)
283 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
284 np->opt->opt_nflen);
285
286 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
287
288 inet->dport = usin->sin6_port;
289
290 tcp_set_state(sk, TCP_SYN_SENT);
291 err = inet6_hash_connect(&tcp_death_row, sk);
292 if (err)
293 goto late_failure;
294
295 if (!tp->write_seq)
296 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
297 np->daddr.s6_addr32,
298 inet->sport,
299 inet->dport);
300
301 err = tcp_connect(sk);
302 if (err)
303 goto late_failure;
304
305 return 0;
306
307 late_failure:
308 tcp_set_state(sk, TCP_CLOSE);
309 __sk_dst_reset(sk);
310 failure:
311 inet->dport = 0;
312 sk->sk_route_caps = 0;
313 return err;
314 }
315
316 static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
317 int type, int code, int offset, __be32 info)
318 {
319 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
320 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
321 struct ipv6_pinfo *np;
322 struct sock *sk;
323 int err;
324 struct tcp_sock *tp;
325 __u32 seq;
326 struct net *net = dev_net(skb->dev);
327
328 sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr,
329 th->dest, &hdr->saddr, th->source, skb->dev->ifindex);
330
331 if (sk == NULL) {
332 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
333 return;
334 }
335
336 if (sk->sk_state == TCP_TIME_WAIT) {
337 inet_twsk_put(inet_twsk(sk));
338 return;
339 }
340
341 bh_lock_sock(sk);
342 if (sock_owned_by_user(sk))
343 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
344
345 if (sk->sk_state == TCP_CLOSE)
346 goto out;
347
348 tp = tcp_sk(sk);
349 seq = ntohl(th->seq);
350 if (sk->sk_state != TCP_LISTEN &&
351 !between(seq, tp->snd_una, tp->snd_nxt)) {
352 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
353 goto out;
354 }
355
356 np = inet6_sk(sk);
357
358 if (type == ICMPV6_PKT_TOOBIG) {
359 struct dst_entry *dst = NULL;
360
361 if (sock_owned_by_user(sk))
362 goto out;
363 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
364 goto out;
365
366 /* icmp should have updated the destination cache entry */
367 dst = __sk_dst_check(sk, np->dst_cookie);
368
369 if (dst == NULL) {
370 struct inet_sock *inet = inet_sk(sk);
371 struct flowi fl;
372
373 /* BUGGG_FUTURE: Again, it is not clear how
374 to handle rthdr case. Ignore this complexity
375 for now.
376 */
377 memset(&fl, 0, sizeof(fl));
378 fl.proto = IPPROTO_TCP;
379 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
380 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
381 fl.oif = sk->sk_bound_dev_if;
382 fl.fl_ip_dport = inet->dport;
383 fl.fl_ip_sport = inet->sport;
384 security_skb_classify_flow(skb, &fl);
385
386 if ((err = ip6_dst_lookup(sk, &dst, &fl))) {
387 sk->sk_err_soft = -err;
388 goto out;
389 }
390
391 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
392 sk->sk_err_soft = -err;
393 goto out;
394 }
395
396 } else
397 dst_hold(dst);
398
399 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
400 tcp_sync_mss(sk, dst_mtu(dst));
401 tcp_simple_retransmit(sk);
402 } /* else let the usual retransmit timer handle it */
403 dst_release(dst);
404 goto out;
405 }
406
407 icmpv6_err_convert(type, code, &err);
408
409 /* Might be for an request_sock */
410 switch (sk->sk_state) {
411 struct request_sock *req, **prev;
412 case TCP_LISTEN:
413 if (sock_owned_by_user(sk))
414 goto out;
415
416 req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
417 &hdr->saddr, inet6_iif(skb));
418 if (!req)
419 goto out;
420
421 /* ICMPs are not backlogged, hence we cannot get
422 * an established socket here.
423 */
424 WARN_ON(req->sk != NULL);
425
426 if (seq != tcp_rsk(req)->snt_isn) {
427 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
428 goto out;
429 }
430
431 inet_csk_reqsk_queue_drop(sk, req, prev);
432 goto out;
433
434 case TCP_SYN_SENT:
435 case TCP_SYN_RECV: /* Cannot happen.
436 It can, it SYNs are crossed. --ANK */
437 if (!sock_owned_by_user(sk)) {
438 sk->sk_err = err;
439 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
440
441 tcp_done(sk);
442 } else
443 sk->sk_err_soft = err;
444 goto out;
445 }
446
447 if (!sock_owned_by_user(sk) && np->recverr) {
448 sk->sk_err = err;
449 sk->sk_error_report(sk);
450 } else
451 sk->sk_err_soft = err;
452
453 out:
454 bh_unlock_sock(sk);
455 sock_put(sk);
456 }
457
458
459 static int tcp_v6_send_synack(struct sock *sk, struct request_sock *req)
460 {
461 struct inet6_request_sock *treq = inet6_rsk(req);
462 struct ipv6_pinfo *np = inet6_sk(sk);
463 struct sk_buff * skb;
464 struct ipv6_txoptions *opt = NULL;
465 struct in6_addr * final_p = NULL, final;
466 struct flowi fl;
467 struct dst_entry *dst;
468 int err = -1;
469
470 memset(&fl, 0, sizeof(fl));
471 fl.proto = IPPROTO_TCP;
472 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
473 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
474 fl.fl6_flowlabel = 0;
475 fl.oif = treq->iif;
476 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
477 fl.fl_ip_sport = inet_sk(sk)->sport;
478 security_req_classify_flow(req, &fl);
479
480 opt = np->opt;
481 if (opt && opt->srcrt) {
482 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
483 ipv6_addr_copy(&final, &fl.fl6_dst);
484 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
485 final_p = &final;
486 }
487
488 err = ip6_dst_lookup(sk, &dst, &fl);
489 if (err)
490 goto done;
491 if (final_p)
492 ipv6_addr_copy(&fl.fl6_dst, final_p);
493 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
494 goto done;
495
496 skb = tcp_make_synack(sk, dst, req);
497 if (skb) {
498 struct tcphdr *th = tcp_hdr(skb);
499
500 th->check = tcp_v6_check(th, skb->len,
501 &treq->loc_addr, &treq->rmt_addr,
502 csum_partial((char *)th, skb->len, skb->csum));
503
504 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
505 err = ip6_xmit(sk, skb, &fl, opt, 0);
506 err = net_xmit_eval(err);
507 }
508
509 done:
510 if (opt && opt != np->opt)
511 sock_kfree_s(sk, opt, opt->tot_len);
512 dst_release(dst);
513 return err;
514 }
515
516 static inline void syn_flood_warning(struct sk_buff *skb)
517 {
518 #ifdef CONFIG_SYN_COOKIES
519 if (sysctl_tcp_syncookies)
520 printk(KERN_INFO
521 "TCPv6: Possible SYN flooding on port %d. "
522 "Sending cookies.\n", ntohs(tcp_hdr(skb)->dest));
523 else
524 #endif
525 printk(KERN_INFO
526 "TCPv6: Possible SYN flooding on port %d. "
527 "Dropping request.\n", ntohs(tcp_hdr(skb)->dest));
528 }
529
530 static void tcp_v6_reqsk_destructor(struct request_sock *req)
531 {
532 if (inet6_rsk(req)->pktopts)
533 kfree_skb(inet6_rsk(req)->pktopts);
534 }
535
536 #ifdef CONFIG_TCP_MD5SIG
537 static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
538 struct in6_addr *addr)
539 {
540 struct tcp_sock *tp = tcp_sk(sk);
541 int i;
542
543 BUG_ON(tp == NULL);
544
545 if (!tp->md5sig_info || !tp->md5sig_info->entries6)
546 return NULL;
547
548 for (i = 0; i < tp->md5sig_info->entries6; i++) {
549 if (ipv6_addr_equal(&tp->md5sig_info->keys6[i].addr, addr))
550 return &tp->md5sig_info->keys6[i].base;
551 }
552 return NULL;
553 }
554
555 static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
556 struct sock *addr_sk)
557 {
558 return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr);
559 }
560
561 static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
562 struct request_sock *req)
563 {
564 return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr);
565 }
566
567 static int tcp_v6_md5_do_add(struct sock *sk, struct in6_addr *peer,
568 char *newkey, u8 newkeylen)
569 {
570 /* Add key to the list */
571 struct tcp_md5sig_key *key;
572 struct tcp_sock *tp = tcp_sk(sk);
573 struct tcp6_md5sig_key *keys;
574
575 key = tcp_v6_md5_do_lookup(sk, peer);
576 if (key) {
577 /* modify existing entry - just update that one */
578 kfree(key->key);
579 key->key = newkey;
580 key->keylen = newkeylen;
581 } else {
582 /* reallocate new list if current one is full. */
583 if (!tp->md5sig_info) {
584 tp->md5sig_info = kzalloc(sizeof(*tp->md5sig_info), GFP_ATOMIC);
585 if (!tp->md5sig_info) {
586 kfree(newkey);
587 return -ENOMEM;
588 }
589 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
590 }
591 if (tcp_alloc_md5sig_pool() == NULL) {
592 kfree(newkey);
593 return -ENOMEM;
594 }
595 if (tp->md5sig_info->alloced6 == tp->md5sig_info->entries6) {
596 keys = kmalloc((sizeof (tp->md5sig_info->keys6[0]) *
597 (tp->md5sig_info->entries6 + 1)), GFP_ATOMIC);
598
599 if (!keys) {
600 tcp_free_md5sig_pool();
601 kfree(newkey);
602 return -ENOMEM;
603 }
604
605 if (tp->md5sig_info->entries6)
606 memmove(keys, tp->md5sig_info->keys6,
607 (sizeof (tp->md5sig_info->keys6[0]) *
608 tp->md5sig_info->entries6));
609
610 kfree(tp->md5sig_info->keys6);
611 tp->md5sig_info->keys6 = keys;
612 tp->md5sig_info->alloced6++;
613 }
614
615 ipv6_addr_copy(&tp->md5sig_info->keys6[tp->md5sig_info->entries6].addr,
616 peer);
617 tp->md5sig_info->keys6[tp->md5sig_info->entries6].base.key = newkey;
618 tp->md5sig_info->keys6[tp->md5sig_info->entries6].base.keylen = newkeylen;
619
620 tp->md5sig_info->entries6++;
621 }
622 return 0;
623 }
624
625 static int tcp_v6_md5_add_func(struct sock *sk, struct sock *addr_sk,
626 u8 *newkey, __u8 newkeylen)
627 {
628 return tcp_v6_md5_do_add(sk, &inet6_sk(addr_sk)->daddr,
629 newkey, newkeylen);
630 }
631
632 static int tcp_v6_md5_do_del(struct sock *sk, struct in6_addr *peer)
633 {
634 struct tcp_sock *tp = tcp_sk(sk);
635 int i;
636
637 for (i = 0; i < tp->md5sig_info->entries6; i++) {
638 if (ipv6_addr_equal(&tp->md5sig_info->keys6[i].addr, peer)) {
639 /* Free the key */
640 kfree(tp->md5sig_info->keys6[i].base.key);
641 tp->md5sig_info->entries6--;
642
643 if (tp->md5sig_info->entries6 == 0) {
644 kfree(tp->md5sig_info->keys6);
645 tp->md5sig_info->keys6 = NULL;
646 tp->md5sig_info->alloced6 = 0;
647 } else {
648 /* shrink the database */
649 if (tp->md5sig_info->entries6 != i)
650 memmove(&tp->md5sig_info->keys6[i],
651 &tp->md5sig_info->keys6[i+1],
652 (tp->md5sig_info->entries6 - i)
653 * sizeof (tp->md5sig_info->keys6[0]));
654 }
655 tcp_free_md5sig_pool();
656 return 0;
657 }
658 }
659 return -ENOENT;
660 }
661
662 static void tcp_v6_clear_md5_list (struct sock *sk)
663 {
664 struct tcp_sock *tp = tcp_sk(sk);
665 int i;
666
667 if (tp->md5sig_info->entries6) {
668 for (i = 0; i < tp->md5sig_info->entries6; i++)
669 kfree(tp->md5sig_info->keys6[i].base.key);
670 tp->md5sig_info->entries6 = 0;
671 tcp_free_md5sig_pool();
672 }
673
674 kfree(tp->md5sig_info->keys6);
675 tp->md5sig_info->keys6 = NULL;
676 tp->md5sig_info->alloced6 = 0;
677
678 if (tp->md5sig_info->entries4) {
679 for (i = 0; i < tp->md5sig_info->entries4; i++)
680 kfree(tp->md5sig_info->keys4[i].base.key);
681 tp->md5sig_info->entries4 = 0;
682 tcp_free_md5sig_pool();
683 }
684
685 kfree(tp->md5sig_info->keys4);
686 tp->md5sig_info->keys4 = NULL;
687 tp->md5sig_info->alloced4 = 0;
688 }
689
690 static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval,
691 int optlen)
692 {
693 struct tcp_md5sig cmd;
694 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
695 u8 *newkey;
696
697 if (optlen < sizeof(cmd))
698 return -EINVAL;
699
700 if (copy_from_user(&cmd, optval, sizeof(cmd)))
701 return -EFAULT;
702
703 if (sin6->sin6_family != AF_INET6)
704 return -EINVAL;
705
706 if (!cmd.tcpm_keylen) {
707 if (!tcp_sk(sk)->md5sig_info)
708 return -ENOENT;
709 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
710 return tcp_v4_md5_do_del(sk, sin6->sin6_addr.s6_addr32[3]);
711 return tcp_v6_md5_do_del(sk, &sin6->sin6_addr);
712 }
713
714 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
715 return -EINVAL;
716
717 if (!tcp_sk(sk)->md5sig_info) {
718 struct tcp_sock *tp = tcp_sk(sk);
719 struct tcp_md5sig_info *p;
720
721 p = kzalloc(sizeof(struct tcp_md5sig_info), GFP_KERNEL);
722 if (!p)
723 return -ENOMEM;
724
725 tp->md5sig_info = p;
726 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
727 }
728
729 newkey = kmemdup(cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
730 if (!newkey)
731 return -ENOMEM;
732 if (ipv6_addr_v4mapped(&sin6->sin6_addr)) {
733 return tcp_v4_md5_do_add(sk, sin6->sin6_addr.s6_addr32[3],
734 newkey, cmd.tcpm_keylen);
735 }
736 return tcp_v6_md5_do_add(sk, &sin6->sin6_addr, newkey, cmd.tcpm_keylen);
737 }
738
739 static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
740 struct in6_addr *daddr,
741 struct in6_addr *saddr, int nbytes)
742 {
743 struct tcp6_pseudohdr *bp;
744 struct scatterlist sg;
745
746 bp = &hp->md5_blk.ip6;
747 /* 1. TCP pseudo-header (RFC2460) */
748 ipv6_addr_copy(&bp->saddr, saddr);
749 ipv6_addr_copy(&bp->daddr, daddr);
750 bp->protocol = cpu_to_be32(IPPROTO_TCP);
751 bp->len = cpu_to_be32(nbytes);
752
753 sg_init_one(&sg, bp, sizeof(*bp));
754 return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
755 }
756
757 static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
758 struct in6_addr *daddr, struct in6_addr *saddr,
759 struct tcphdr *th)
760 {
761 struct tcp_md5sig_pool *hp;
762 struct hash_desc *desc;
763
764 hp = tcp_get_md5sig_pool();
765 if (!hp)
766 goto clear_hash_noput;
767 desc = &hp->md5_desc;
768
769 if (crypto_hash_init(desc))
770 goto clear_hash;
771 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2))
772 goto clear_hash;
773 if (tcp_md5_hash_header(hp, th))
774 goto clear_hash;
775 if (tcp_md5_hash_key(hp, key))
776 goto clear_hash;
777 if (crypto_hash_final(desc, md5_hash))
778 goto clear_hash;
779
780 tcp_put_md5sig_pool();
781 return 0;
782
783 clear_hash:
784 tcp_put_md5sig_pool();
785 clear_hash_noput:
786 memset(md5_hash, 0, 16);
787 return 1;
788 }
789
790 static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
791 struct sock *sk, struct request_sock *req,
792 struct sk_buff *skb)
793 {
794 struct in6_addr *saddr, *daddr;
795 struct tcp_md5sig_pool *hp;
796 struct hash_desc *desc;
797 struct tcphdr *th = tcp_hdr(skb);
798
799 if (sk) {
800 saddr = &inet6_sk(sk)->saddr;
801 daddr = &inet6_sk(sk)->daddr;
802 } else if (req) {
803 saddr = &inet6_rsk(req)->loc_addr;
804 daddr = &inet6_rsk(req)->rmt_addr;
805 } else {
806 struct ipv6hdr *ip6h = ipv6_hdr(skb);
807 saddr = &ip6h->saddr;
808 daddr = &ip6h->daddr;
809 }
810
811 hp = tcp_get_md5sig_pool();
812 if (!hp)
813 goto clear_hash_noput;
814 desc = &hp->md5_desc;
815
816 if (crypto_hash_init(desc))
817 goto clear_hash;
818
819 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len))
820 goto clear_hash;
821 if (tcp_md5_hash_header(hp, th))
822 goto clear_hash;
823 if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
824 goto clear_hash;
825 if (tcp_md5_hash_key(hp, key))
826 goto clear_hash;
827 if (crypto_hash_final(desc, md5_hash))
828 goto clear_hash;
829
830 tcp_put_md5sig_pool();
831 return 0;
832
833 clear_hash:
834 tcp_put_md5sig_pool();
835 clear_hash_noput:
836 memset(md5_hash, 0, 16);
837 return 1;
838 }
839
840 static int tcp_v6_inbound_md5_hash (struct sock *sk, struct sk_buff *skb)
841 {
842 __u8 *hash_location = NULL;
843 struct tcp_md5sig_key *hash_expected;
844 struct ipv6hdr *ip6h = ipv6_hdr(skb);
845 struct tcphdr *th = tcp_hdr(skb);
846 int genhash;
847 u8 newhash[16];
848
849 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
850 hash_location = tcp_parse_md5sig_option(th);
851
852 /* We've parsed the options - do we have a hash? */
853 if (!hash_expected && !hash_location)
854 return 0;
855
856 if (hash_expected && !hash_location) {
857 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
858 return 1;
859 }
860
861 if (!hash_expected && hash_location) {
862 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
863 return 1;
864 }
865
866 /* check the signature */
867 genhash = tcp_v6_md5_hash_skb(newhash,
868 hash_expected,
869 NULL, NULL, skb);
870
871 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
872 if (net_ratelimit()) {
873 printk(KERN_INFO "MD5 Hash %s for "
874 "(" NIP6_FMT ", %u)->"
875 "(" NIP6_FMT ", %u)\n",
876 genhash ? "failed" : "mismatch",
877 NIP6(ip6h->saddr), ntohs(th->source),
878 NIP6(ip6h->daddr), ntohs(th->dest));
879 }
880 return 1;
881 }
882 return 0;
883 }
884 #endif
885
886 struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
887 .family = AF_INET6,
888 .obj_size = sizeof(struct tcp6_request_sock),
889 .rtx_syn_ack = tcp_v6_send_synack,
890 .send_ack = tcp_v6_reqsk_send_ack,
891 .destructor = tcp_v6_reqsk_destructor,
892 .send_reset = tcp_v6_send_reset
893 };
894
895 #ifdef CONFIG_TCP_MD5SIG
896 static struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
897 .md5_lookup = tcp_v6_reqsk_md5_lookup,
898 };
899 #endif
900
901 static struct timewait_sock_ops tcp6_timewait_sock_ops = {
902 .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
903 .twsk_unique = tcp_twsk_unique,
904 .twsk_destructor= tcp_twsk_destructor,
905 };
906
907 static void tcp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
908 {
909 struct ipv6_pinfo *np = inet6_sk(sk);
910 struct tcphdr *th = tcp_hdr(skb);
911
912 if (skb->ip_summed == CHECKSUM_PARTIAL) {
913 th->check = ~csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP, 0);
914 skb->csum_start = skb_transport_header(skb) - skb->head;
915 skb->csum_offset = offsetof(struct tcphdr, check);
916 } else {
917 th->check = csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP,
918 csum_partial((char *)th, th->doff<<2,
919 skb->csum));
920 }
921 }
922
923 static int tcp_v6_gso_send_check(struct sk_buff *skb)
924 {
925 struct ipv6hdr *ipv6h;
926 struct tcphdr *th;
927
928 if (!pskb_may_pull(skb, sizeof(*th)))
929 return -EINVAL;
930
931 ipv6h = ipv6_hdr(skb);
932 th = tcp_hdr(skb);
933
934 th->check = 0;
935 th->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
936 IPPROTO_TCP, 0);
937 skb->csum_start = skb_transport_header(skb) - skb->head;
938 skb->csum_offset = offsetof(struct tcphdr, check);
939 skb->ip_summed = CHECKSUM_PARTIAL;
940 return 0;
941 }
942
943 static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
944 {
945 struct tcphdr *th = tcp_hdr(skb), *t1;
946 struct sk_buff *buff;
947 struct flowi fl;
948 struct net *net = dev_net(skb->dst->dev);
949 struct sock *ctl_sk = net->ipv6.tcp_sk;
950 unsigned int tot_len = sizeof(*th);
951 #ifdef CONFIG_TCP_MD5SIG
952 struct tcp_md5sig_key *key;
953 #endif
954
955 if (th->rst)
956 return;
957
958 if (!ipv6_unicast_destination(skb))
959 return;
960
961 #ifdef CONFIG_TCP_MD5SIG
962 if (sk)
963 key = tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr);
964 else
965 key = NULL;
966
967 if (key)
968 tot_len += TCPOLEN_MD5SIG_ALIGNED;
969 #endif
970
971 /*
972 * We need to grab some memory, and put together an RST,
973 * and then put it into the queue to be sent.
974 */
975
976 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
977 GFP_ATOMIC);
978 if (buff == NULL)
979 return;
980
981 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
982
983 t1 = (struct tcphdr *) skb_push(buff, tot_len);
984
985 /* Swap the send and the receive. */
986 memset(t1, 0, sizeof(*t1));
987 t1->dest = th->source;
988 t1->source = th->dest;
989 t1->doff = tot_len / 4;
990 t1->rst = 1;
991
992 if(th->ack) {
993 t1->seq = th->ack_seq;
994 } else {
995 t1->ack = 1;
996 t1->ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin
997 + skb->len - (th->doff<<2));
998 }
999
1000 #ifdef CONFIG_TCP_MD5SIG
1001 if (key) {
1002 __be32 *opt = (__be32*)(t1 + 1);
1003 opt[0] = htonl((TCPOPT_NOP << 24) |
1004 (TCPOPT_NOP << 16) |
1005 (TCPOPT_MD5SIG << 8) |
1006 TCPOLEN_MD5SIG);
1007 tcp_v6_md5_hash_hdr((__u8 *)&opt[1], key,
1008 &ipv6_hdr(skb)->daddr,
1009 &ipv6_hdr(skb)->saddr, t1);
1010 }
1011 #endif
1012
1013 buff->csum = csum_partial((char *)t1, sizeof(*t1), 0);
1014
1015 memset(&fl, 0, sizeof(fl));
1016 ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
1017 ipv6_addr_copy(&fl.fl6_src, &ipv6_hdr(skb)->daddr);
1018
1019 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1020 sizeof(*t1), IPPROTO_TCP,
1021 buff->csum);
1022
1023 fl.proto = IPPROTO_TCP;
1024 fl.oif = inet6_iif(skb);
1025 fl.fl_ip_dport = t1->dest;
1026 fl.fl_ip_sport = t1->source;
1027 security_skb_classify_flow(skb, &fl);
1028
1029 /* Pass a socket to ip6_dst_lookup either it is for RST
1030 * Underlying function will use this to retrieve the network
1031 * namespace
1032 */
1033 if (!ip6_dst_lookup(ctl_sk, &buff->dst, &fl)) {
1034
1035 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
1036 ip6_xmit(ctl_sk, buff, &fl, NULL, 0);
1037 TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
1038 TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS);
1039 return;
1040 }
1041 }
1042
1043 kfree_skb(buff);
1044 }
1045
1046 static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts,
1047 struct tcp_md5sig_key *key)
1048 {
1049 struct tcphdr *th = tcp_hdr(skb), *t1;
1050 struct sk_buff *buff;
1051 struct flowi fl;
1052 struct net *net = dev_net(skb->dev);
1053 struct sock *ctl_sk = net->ipv6.tcp_sk;
1054 unsigned int tot_len = sizeof(struct tcphdr);
1055 __be32 *topt;
1056
1057 if (ts)
1058 tot_len += TCPOLEN_TSTAMP_ALIGNED;
1059 #ifdef CONFIG_TCP_MD5SIG
1060 if (key)
1061 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1062 #endif
1063
1064 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1065 GFP_ATOMIC);
1066 if (buff == NULL)
1067 return;
1068
1069 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1070
1071 t1 = (struct tcphdr *) skb_push(buff,tot_len);
1072
1073 /* Swap the send and the receive. */
1074 memset(t1, 0, sizeof(*t1));
1075 t1->dest = th->source;
1076 t1->source = th->dest;
1077 t1->doff = tot_len/4;
1078 t1->seq = htonl(seq);
1079 t1->ack_seq = htonl(ack);
1080 t1->ack = 1;
1081 t1->window = htons(win);
1082
1083 topt = (__be32 *)(t1 + 1);
1084
1085 if (ts) {
1086 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1087 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
1088 *topt++ = htonl(tcp_time_stamp);
1089 *topt = htonl(ts);
1090 }
1091
1092 #ifdef CONFIG_TCP_MD5SIG
1093 if (key) {
1094 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1095 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
1096 tcp_v6_md5_hash_hdr((__u8 *)topt, key,
1097 &ipv6_hdr(skb)->saddr,
1098 &ipv6_hdr(skb)->daddr, t1);
1099 }
1100 #endif
1101
1102 buff->csum = csum_partial((char *)t1, tot_len, 0);
1103
1104 memset(&fl, 0, sizeof(fl));
1105 ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
1106 ipv6_addr_copy(&fl.fl6_src, &ipv6_hdr(skb)->daddr);
1107
1108 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1109 tot_len, IPPROTO_TCP,
1110 buff->csum);
1111
1112 fl.proto = IPPROTO_TCP;
1113 fl.oif = inet6_iif(skb);
1114 fl.fl_ip_dport = t1->dest;
1115 fl.fl_ip_sport = t1->source;
1116 security_skb_classify_flow(skb, &fl);
1117
1118 if (!ip6_dst_lookup(ctl_sk, &buff->dst, &fl)) {
1119 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
1120 ip6_xmit(ctl_sk, buff, &fl, NULL, 0);
1121 TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
1122 return;
1123 }
1124 }
1125
1126 kfree_skb(buff);
1127 }
1128
1129 static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
1130 {
1131 struct inet_timewait_sock *tw = inet_twsk(sk);
1132 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
1133
1134 tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
1135 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
1136 tcptw->tw_ts_recent, tcp_twsk_md5_key(tcptw));
1137
1138 inet_twsk_put(tw);
1139 }
1140
1141 static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req)
1142 {
1143 tcp_v6_send_ack(skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, req->ts_recent,
1144 tcp_v6_md5_do_lookup(skb->sk, &ipv6_hdr(skb)->daddr));
1145 }
1146
1147
1148 static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
1149 {
1150 struct request_sock *req, **prev;
1151 const struct tcphdr *th = tcp_hdr(skb);
1152 struct sock *nsk;
1153
1154 /* Find possible connection requests. */
1155 req = inet6_csk_search_req(sk, &prev, th->source,
1156 &ipv6_hdr(skb)->saddr,
1157 &ipv6_hdr(skb)->daddr, inet6_iif(skb));
1158 if (req)
1159 return tcp_check_req(sk, skb, req, prev);
1160
1161 nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
1162 &ipv6_hdr(skb)->saddr, th->source,
1163 &ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb));
1164
1165 if (nsk) {
1166 if (nsk->sk_state != TCP_TIME_WAIT) {
1167 bh_lock_sock(nsk);
1168 return nsk;
1169 }
1170 inet_twsk_put(inet_twsk(nsk));
1171 return NULL;
1172 }
1173
1174 #ifdef CONFIG_SYN_COOKIES
1175 if (!th->rst && !th->syn && th->ack)
1176 sk = cookie_v6_check(sk, skb);
1177 #endif
1178 return sk;
1179 }
1180
1181 /* FIXME: this is substantially similar to the ipv4 code.
1182 * Can some kind of merge be done? -- erics
1183 */
1184 static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
1185 {
1186 struct inet6_request_sock *treq;
1187 struct ipv6_pinfo *np = inet6_sk(sk);
1188 struct tcp_options_received tmp_opt;
1189 struct tcp_sock *tp = tcp_sk(sk);
1190 struct request_sock *req = NULL;
1191 __u32 isn = TCP_SKB_CB(skb)->when;
1192 #ifdef CONFIG_SYN_COOKIES
1193 int want_cookie = 0;
1194 #else
1195 #define want_cookie 0
1196 #endif
1197
1198 if (skb->protocol == htons(ETH_P_IP))
1199 return tcp_v4_conn_request(sk, skb);
1200
1201 if (!ipv6_unicast_destination(skb))
1202 goto drop;
1203
1204 if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
1205 if (net_ratelimit())
1206 syn_flood_warning(skb);
1207 #ifdef CONFIG_SYN_COOKIES
1208 if (sysctl_tcp_syncookies)
1209 want_cookie = 1;
1210 else
1211 #endif
1212 goto drop;
1213 }
1214
1215 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
1216 goto drop;
1217
1218 req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
1219 if (req == NULL)
1220 goto drop;
1221
1222 #ifdef CONFIG_TCP_MD5SIG
1223 tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
1224 #endif
1225
1226 tcp_clear_options(&tmp_opt);
1227 tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
1228 tmp_opt.user_mss = tp->rx_opt.user_mss;
1229
1230 tcp_parse_options(skb, &tmp_opt, 0);
1231
1232 if (want_cookie && !tmp_opt.saw_tstamp)
1233 tcp_clear_options(&tmp_opt);
1234
1235 tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
1236 tcp_openreq_init(req, &tmp_opt, skb);
1237
1238 treq = inet6_rsk(req);
1239 ipv6_addr_copy(&treq->rmt_addr, &ipv6_hdr(skb)->saddr);
1240 ipv6_addr_copy(&treq->loc_addr, &ipv6_hdr(skb)->daddr);
1241 if (!want_cookie)
1242 TCP_ECN_create_request(req, tcp_hdr(skb));
1243
1244 if (want_cookie) {
1245 isn = cookie_v6_init_sequence(sk, skb, &req->mss);
1246 req->cookie_ts = tmp_opt.tstamp_ok;
1247 } else if (!isn) {
1248 if (ipv6_opt_accepted(sk, skb) ||
1249 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
1250 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
1251 atomic_inc(&skb->users);
1252 treq->pktopts = skb;
1253 }
1254 treq->iif = sk->sk_bound_dev_if;
1255
1256 /* So that link locals have meaning */
1257 if (!sk->sk_bound_dev_if &&
1258 ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL)
1259 treq->iif = inet6_iif(skb);
1260
1261 isn = tcp_v6_init_sequence(skb);
1262 }
1263
1264 tcp_rsk(req)->snt_isn = isn;
1265
1266 security_inet_conn_request(sk, skb, req);
1267
1268 if (tcp_v6_send_synack(sk, req))
1269 goto drop;
1270
1271 if (!want_cookie) {
1272 inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
1273 return 0;
1274 }
1275
1276 drop:
1277 if (req)
1278 reqsk_free(req);
1279
1280 return 0; /* don't send reset */
1281 }
1282
1283 static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
1284 struct request_sock *req,
1285 struct dst_entry *dst)
1286 {
1287 struct inet6_request_sock *treq = inet6_rsk(req);
1288 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
1289 struct tcp6_sock *newtcp6sk;
1290 struct inet_sock *newinet;
1291 struct tcp_sock *newtp;
1292 struct sock *newsk;
1293 struct ipv6_txoptions *opt;
1294 #ifdef CONFIG_TCP_MD5SIG
1295 struct tcp_md5sig_key *key;
1296 #endif
1297
1298 if (skb->protocol == htons(ETH_P_IP)) {
1299 /*
1300 * v6 mapped
1301 */
1302
1303 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1304
1305 if (newsk == NULL)
1306 return NULL;
1307
1308 newtcp6sk = (struct tcp6_sock *)newsk;
1309 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1310
1311 newinet = inet_sk(newsk);
1312 newnp = inet6_sk(newsk);
1313 newtp = tcp_sk(newsk);
1314
1315 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1316
1317 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
1318 newinet->daddr);
1319
1320 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
1321 newinet->saddr);
1322
1323 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
1324
1325 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1326 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1327 #ifdef CONFIG_TCP_MD5SIG
1328 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1329 #endif
1330
1331 newnp->pktoptions = NULL;
1332 newnp->opt = NULL;
1333 newnp->mcast_oif = inet6_iif(skb);
1334 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1335
1336 /*
1337 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1338 * here, tcp_create_openreq_child now does this for us, see the comment in
1339 * that function for the gory details. -acme
1340 */
1341
1342 /* It is tricky place. Until this moment IPv4 tcp
1343 worked with IPv6 icsk.icsk_af_ops.
1344 Sync it now.
1345 */
1346 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1347
1348 return newsk;
1349 }
1350
1351 opt = np->opt;
1352
1353 if (sk_acceptq_is_full(sk))
1354 goto out_overflow;
1355
1356 if (dst == NULL) {
1357 struct in6_addr *final_p = NULL, final;
1358 struct flowi fl;
1359
1360 memset(&fl, 0, sizeof(fl));
1361 fl.proto = IPPROTO_TCP;
1362 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
1363 if (opt && opt->srcrt) {
1364 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
1365 ipv6_addr_copy(&final, &fl.fl6_dst);
1366 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
1367 final_p = &final;
1368 }
1369 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
1370 fl.oif = sk->sk_bound_dev_if;
1371 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
1372 fl.fl_ip_sport = inet_sk(sk)->sport;
1373 security_req_classify_flow(req, &fl);
1374
1375 if (ip6_dst_lookup(sk, &dst, &fl))
1376 goto out;
1377
1378 if (final_p)
1379 ipv6_addr_copy(&fl.fl6_dst, final_p);
1380
1381 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
1382 goto out;
1383 }
1384
1385 newsk = tcp_create_openreq_child(sk, req, skb);
1386 if (newsk == NULL)
1387 goto out;
1388
1389 /*
1390 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1391 * count here, tcp_create_openreq_child now does this for us, see the
1392 * comment in that function for the gory details. -acme
1393 */
1394
1395 newsk->sk_gso_type = SKB_GSO_TCPV6;
1396 __ip6_dst_store(newsk, dst, NULL, NULL);
1397
1398 newtcp6sk = (struct tcp6_sock *)newsk;
1399 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1400
1401 newtp = tcp_sk(newsk);
1402 newinet = inet_sk(newsk);
1403 newnp = inet6_sk(newsk);
1404
1405 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1406
1407 ipv6_addr_copy(&newnp->daddr, &treq->rmt_addr);
1408 ipv6_addr_copy(&newnp->saddr, &treq->loc_addr);
1409 ipv6_addr_copy(&newnp->rcv_saddr, &treq->loc_addr);
1410 newsk->sk_bound_dev_if = treq->iif;
1411
1412 /* Now IPv6 options...
1413
1414 First: no IPv4 options.
1415 */
1416 newinet->opt = NULL;
1417 newnp->ipv6_fl_list = NULL;
1418
1419 /* Clone RX bits */
1420 newnp->rxopt.all = np->rxopt.all;
1421
1422 /* Clone pktoptions received with SYN */
1423 newnp->pktoptions = NULL;
1424 if (treq->pktopts != NULL) {
1425 newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC);
1426 kfree_skb(treq->pktopts);
1427 treq->pktopts = NULL;
1428 if (newnp->pktoptions)
1429 skb_set_owner_r(newnp->pktoptions, newsk);
1430 }
1431 newnp->opt = NULL;
1432 newnp->mcast_oif = inet6_iif(skb);
1433 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1434
1435 /* Clone native IPv6 options from listening socket (if any)
1436
1437 Yes, keeping reference count would be much more clever,
1438 but we make one more one thing there: reattach optmem
1439 to newsk.
1440 */
1441 if (opt) {
1442 newnp->opt = ipv6_dup_options(newsk, opt);
1443 if (opt != np->opt)
1444 sock_kfree_s(sk, opt, opt->tot_len);
1445 }
1446
1447 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1448 if (newnp->opt)
1449 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1450 newnp->opt->opt_flen);
1451
1452 tcp_mtup_init(newsk);
1453 tcp_sync_mss(newsk, dst_mtu(dst));
1454 newtp->advmss = dst_metric(dst, RTAX_ADVMSS);
1455 tcp_initialize_rcv_mss(newsk);
1456
1457 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
1458
1459 #ifdef CONFIG_TCP_MD5SIG
1460 /* Copy over the MD5 key from the original socket */
1461 if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) {
1462 /* We're using one, so create a matching key
1463 * on the newsk structure. If we fail to get
1464 * memory, then we end up not copying the key
1465 * across. Shucks.
1466 */
1467 char *newkey = kmemdup(key->key, key->keylen, GFP_ATOMIC);
1468 if (newkey != NULL)
1469 tcp_v6_md5_do_add(newsk, &inet6_sk(sk)->daddr,
1470 newkey, key->keylen);
1471 }
1472 #endif
1473
1474 __inet6_hash(newsk);
1475 __inet_inherit_port(sk, newsk);
1476
1477 return newsk;
1478
1479 out_overflow:
1480 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1481 out:
1482 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
1483 if (opt && opt != np->opt)
1484 sock_kfree_s(sk, opt, opt->tot_len);
1485 dst_release(dst);
1486 return NULL;
1487 }
1488
1489 static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
1490 {
1491 if (skb->ip_summed == CHECKSUM_COMPLETE) {
1492 if (!tcp_v6_check(tcp_hdr(skb), skb->len, &ipv6_hdr(skb)->saddr,
1493 &ipv6_hdr(skb)->daddr, skb->csum)) {
1494 skb->ip_summed = CHECKSUM_UNNECESSARY;
1495 return 0;
1496 }
1497 }
1498
1499 skb->csum = ~csum_unfold(tcp_v6_check(tcp_hdr(skb), skb->len,
1500 &ipv6_hdr(skb)->saddr,
1501 &ipv6_hdr(skb)->daddr, 0));
1502
1503 if (skb->len <= 76) {
1504 return __skb_checksum_complete(skb);
1505 }
1506 return 0;
1507 }
1508
1509 /* The socket must have it's spinlock held when we get
1510 * here.
1511 *
1512 * We have a potential double-lock case here, so even when
1513 * doing backlog processing we use the BH locking scheme.
1514 * This is because we cannot sleep with the original spinlock
1515 * held.
1516 */
1517 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1518 {
1519 struct ipv6_pinfo *np = inet6_sk(sk);
1520 struct tcp_sock *tp;
1521 struct sk_buff *opt_skb = NULL;
1522
1523 /* Imagine: socket is IPv6. IPv4 packet arrives,
1524 goes to IPv4 receive handler and backlogged.
1525 From backlog it always goes here. Kerboom...
1526 Fortunately, tcp_rcv_established and rcv_established
1527 handle them correctly, but it is not case with
1528 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1529 */
1530
1531 if (skb->protocol == htons(ETH_P_IP))
1532 return tcp_v4_do_rcv(sk, skb);
1533
1534 #ifdef CONFIG_TCP_MD5SIG
1535 if (tcp_v6_inbound_md5_hash (sk, skb))
1536 goto discard;
1537 #endif
1538
1539 if (sk_filter(sk, skb))
1540 goto discard;
1541
1542 /*
1543 * socket locking is here for SMP purposes as backlog rcv
1544 * is currently called with bh processing disabled.
1545 */
1546
1547 /* Do Stevens' IPV6_PKTOPTIONS.
1548
1549 Yes, guys, it is the only place in our code, where we
1550 may make it not affecting IPv4.
1551 The rest of code is protocol independent,
1552 and I do not like idea to uglify IPv4.
1553
1554 Actually, all the idea behind IPV6_PKTOPTIONS
1555 looks not very well thought. For now we latch
1556 options, received in the last packet, enqueued
1557 by tcp. Feel free to propose better solution.
1558 --ANK (980728)
1559 */
1560 if (np->rxopt.all)
1561 opt_skb = skb_clone(skb, GFP_ATOMIC);
1562
1563 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1564 TCP_CHECK_TIMER(sk);
1565 if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len))
1566 goto reset;
1567 TCP_CHECK_TIMER(sk);
1568 if (opt_skb)
1569 goto ipv6_pktoptions;
1570 return 0;
1571 }
1572
1573 if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
1574 goto csum_err;
1575
1576 if (sk->sk_state == TCP_LISTEN) {
1577 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1578 if (!nsk)
1579 goto discard;
1580
1581 /*
1582 * Queue it on the new socket if the new socket is active,
1583 * otherwise we just shortcircuit this and continue with
1584 * the new socket..
1585 */
1586 if(nsk != sk) {
1587 if (tcp_child_process(sk, nsk, skb))
1588 goto reset;
1589 if (opt_skb)
1590 __kfree_skb(opt_skb);
1591 return 0;
1592 }
1593 }
1594
1595 TCP_CHECK_TIMER(sk);
1596 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
1597 goto reset;
1598 TCP_CHECK_TIMER(sk);
1599 if (opt_skb)
1600 goto ipv6_pktoptions;
1601 return 0;
1602
1603 reset:
1604 tcp_v6_send_reset(sk, skb);
1605 discard:
1606 if (opt_skb)
1607 __kfree_skb(opt_skb);
1608 kfree_skb(skb);
1609 return 0;
1610 csum_err:
1611 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
1612 goto discard;
1613
1614
1615 ipv6_pktoptions:
1616 /* Do you ask, what is it?
1617
1618 1. skb was enqueued by tcp.
1619 2. skb is added to tail of read queue, rather than out of order.
1620 3. socket is not in passive state.
1621 4. Finally, it really contains options, which user wants to receive.
1622 */
1623 tp = tcp_sk(sk);
1624 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1625 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
1626 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
1627 np->mcast_oif = inet6_iif(opt_skb);
1628 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1629 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1630 if (ipv6_opt_accepted(sk, opt_skb)) {
1631 skb_set_owner_r(opt_skb, sk);
1632 opt_skb = xchg(&np->pktoptions, opt_skb);
1633 } else {
1634 __kfree_skb(opt_skb);
1635 opt_skb = xchg(&np->pktoptions, NULL);
1636 }
1637 }
1638
1639 if (opt_skb)
1640 kfree_skb(opt_skb);
1641 return 0;
1642 }
1643
1644 static int tcp_v6_rcv(struct sk_buff *skb)
1645 {
1646 struct tcphdr *th;
1647 struct sock *sk;
1648 int ret;
1649 struct net *net = dev_net(skb->dev);
1650
1651 if (skb->pkt_type != PACKET_HOST)
1652 goto discard_it;
1653
1654 /*
1655 * Count it even if it's bad.
1656 */
1657 TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
1658
1659 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1660 goto discard_it;
1661
1662 th = tcp_hdr(skb);
1663
1664 if (th->doff < sizeof(struct tcphdr)/4)
1665 goto bad_packet;
1666 if (!pskb_may_pull(skb, th->doff*4))
1667 goto discard_it;
1668
1669 if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
1670 goto bad_packet;
1671
1672 th = tcp_hdr(skb);
1673 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1674 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1675 skb->len - th->doff*4);
1676 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1677 TCP_SKB_CB(skb)->when = 0;
1678 TCP_SKB_CB(skb)->flags = ipv6_get_dsfield(ipv6_hdr(skb));
1679 TCP_SKB_CB(skb)->sacked = 0;
1680
1681 sk = __inet6_lookup(net, &tcp_hashinfo,
1682 &ipv6_hdr(skb)->saddr, th->source,
1683 &ipv6_hdr(skb)->daddr, ntohs(th->dest),
1684 inet6_iif(skb));
1685
1686 if (!sk)
1687 goto no_tcp_socket;
1688
1689 process:
1690 if (sk->sk_state == TCP_TIME_WAIT)
1691 goto do_time_wait;
1692
1693 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1694 goto discard_and_relse;
1695
1696 if (sk_filter(sk, skb))
1697 goto discard_and_relse;
1698
1699 skb->dev = NULL;
1700
1701 bh_lock_sock_nested(sk);
1702 ret = 0;
1703 if (!sock_owned_by_user(sk)) {
1704 #ifdef CONFIG_NET_DMA
1705 struct tcp_sock *tp = tcp_sk(sk);
1706 if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
1707 tp->ucopy.dma_chan = get_softnet_dma();
1708 if (tp->ucopy.dma_chan)
1709 ret = tcp_v6_do_rcv(sk, skb);
1710 else
1711 #endif
1712 {
1713 if (!tcp_prequeue(sk, skb))
1714 ret = tcp_v6_do_rcv(sk, skb);
1715 }
1716 } else
1717 sk_add_backlog(sk, skb);
1718 bh_unlock_sock(sk);
1719
1720 sock_put(sk);
1721 return ret ? -1 : 0;
1722
1723 no_tcp_socket:
1724 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1725 goto discard_it;
1726
1727 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1728 bad_packet:
1729 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1730 } else {
1731 tcp_v6_send_reset(NULL, skb);
1732 }
1733
1734 discard_it:
1735
1736 /*
1737 * Discard frame
1738 */
1739
1740 kfree_skb(skb);
1741 return 0;
1742
1743 discard_and_relse:
1744 sock_put(sk);
1745 goto discard_it;
1746
1747 do_time_wait:
1748 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1749 inet_twsk_put(inet_twsk(sk));
1750 goto discard_it;
1751 }
1752
1753 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1754 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1755 inet_twsk_put(inet_twsk(sk));
1756 goto discard_it;
1757 }
1758
1759 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1760 case TCP_TW_SYN:
1761 {
1762 struct sock *sk2;
1763
1764 sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1765 &ipv6_hdr(skb)->daddr,
1766 ntohs(th->dest), inet6_iif(skb));
1767 if (sk2 != NULL) {
1768 struct inet_timewait_sock *tw = inet_twsk(sk);
1769 inet_twsk_deschedule(tw, &tcp_death_row);
1770 inet_twsk_put(tw);
1771 sk = sk2;
1772 goto process;
1773 }
1774 /* Fall through to ACK */
1775 }
1776 case TCP_TW_ACK:
1777 tcp_v6_timewait_ack(sk, skb);
1778 break;
1779 case TCP_TW_RST:
1780 goto no_tcp_socket;
1781 case TCP_TW_SUCCESS:;
1782 }
1783 goto discard_it;
1784 }
1785
1786 static int tcp_v6_remember_stamp(struct sock *sk)
1787 {
1788 /* Alas, not yet... */
1789 return 0;
1790 }
1791
1792 static struct inet_connection_sock_af_ops ipv6_specific = {
1793 .queue_xmit = inet6_csk_xmit,
1794 .send_check = tcp_v6_send_check,
1795 .rebuild_header = inet6_sk_rebuild_header,
1796 .conn_request = tcp_v6_conn_request,
1797 .syn_recv_sock = tcp_v6_syn_recv_sock,
1798 .remember_stamp = tcp_v6_remember_stamp,
1799 .net_header_len = sizeof(struct ipv6hdr),
1800 .setsockopt = ipv6_setsockopt,
1801 .getsockopt = ipv6_getsockopt,
1802 .addr2sockaddr = inet6_csk_addr2sockaddr,
1803 .sockaddr_len = sizeof(struct sockaddr_in6),
1804 .bind_conflict = inet6_csk_bind_conflict,
1805 #ifdef CONFIG_COMPAT
1806 .compat_setsockopt = compat_ipv6_setsockopt,
1807 .compat_getsockopt = compat_ipv6_getsockopt,
1808 #endif
1809 };
1810
1811 #ifdef CONFIG_TCP_MD5SIG
1812 static struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1813 .md5_lookup = tcp_v6_md5_lookup,
1814 .calc_md5_hash = tcp_v6_md5_hash_skb,
1815 .md5_add = tcp_v6_md5_add_func,
1816 .md5_parse = tcp_v6_parse_md5_keys,
1817 };
1818 #endif
1819
1820 /*
1821 * TCP over IPv4 via INET6 API
1822 */
1823
1824 static struct inet_connection_sock_af_ops ipv6_mapped = {
1825 .queue_xmit = ip_queue_xmit,
1826 .send_check = tcp_v4_send_check,
1827 .rebuild_header = inet_sk_rebuild_header,
1828 .conn_request = tcp_v6_conn_request,
1829 .syn_recv_sock = tcp_v6_syn_recv_sock,
1830 .remember_stamp = tcp_v4_remember_stamp,
1831 .net_header_len = sizeof(struct iphdr),
1832 .setsockopt = ipv6_setsockopt,
1833 .getsockopt = ipv6_getsockopt,
1834 .addr2sockaddr = inet6_csk_addr2sockaddr,
1835 .sockaddr_len = sizeof(struct sockaddr_in6),
1836 .bind_conflict = inet6_csk_bind_conflict,
1837 #ifdef CONFIG_COMPAT
1838 .compat_setsockopt = compat_ipv6_setsockopt,
1839 .compat_getsockopt = compat_ipv6_getsockopt,
1840 #endif
1841 };
1842
1843 #ifdef CONFIG_TCP_MD5SIG
1844 static struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1845 .md5_lookup = tcp_v4_md5_lookup,
1846 .calc_md5_hash = tcp_v4_md5_hash_skb,
1847 .md5_add = tcp_v6_md5_add_func,
1848 .md5_parse = tcp_v6_parse_md5_keys,
1849 };
1850 #endif
1851
1852 /* NOTE: A lot of things set to zero explicitly by call to
1853 * sk_alloc() so need not be done here.
1854 */
1855 static int tcp_v6_init_sock(struct sock *sk)
1856 {
1857 struct inet_connection_sock *icsk = inet_csk(sk);
1858 struct tcp_sock *tp = tcp_sk(sk);
1859
1860 skb_queue_head_init(&tp->out_of_order_queue);
1861 tcp_init_xmit_timers(sk);
1862 tcp_prequeue_init(tp);
1863
1864 icsk->icsk_rto = TCP_TIMEOUT_INIT;
1865 tp->mdev = TCP_TIMEOUT_INIT;
1866
1867 /* So many TCP implementations out there (incorrectly) count the
1868 * initial SYN frame in their delayed-ACK and congestion control
1869 * algorithms that we must have the following bandaid to talk
1870 * efficiently to them. -DaveM
1871 */
1872 tp->snd_cwnd = 2;
1873
1874 /* See draft-stevens-tcpca-spec-01 for discussion of the
1875 * initialization of these values.
1876 */
1877 tp->snd_ssthresh = 0x7fffffff;
1878 tp->snd_cwnd_clamp = ~0;
1879 tp->mss_cache = 536;
1880
1881 tp->reordering = sysctl_tcp_reordering;
1882
1883 sk->sk_state = TCP_CLOSE;
1884
1885 icsk->icsk_af_ops = &ipv6_specific;
1886 icsk->icsk_ca_ops = &tcp_init_congestion_ops;
1887 icsk->icsk_sync_mss = tcp_sync_mss;
1888 sk->sk_write_space = sk_stream_write_space;
1889 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
1890
1891 #ifdef CONFIG_TCP_MD5SIG
1892 tp->af_specific = &tcp_sock_ipv6_specific;
1893 #endif
1894
1895 sk->sk_sndbuf = sysctl_tcp_wmem[1];
1896 sk->sk_rcvbuf = sysctl_tcp_rmem[1];
1897
1898 atomic_inc(&tcp_sockets_allocated);
1899
1900 return 0;
1901 }
1902
1903 static void tcp_v6_destroy_sock(struct sock *sk)
1904 {
1905 #ifdef CONFIG_TCP_MD5SIG
1906 /* Clean up the MD5 key list */
1907 if (tcp_sk(sk)->md5sig_info)
1908 tcp_v6_clear_md5_list(sk);
1909 #endif
1910 tcp_v4_destroy_sock(sk);
1911 inet6_destroy_sock(sk);
1912 }
1913
1914 #ifdef CONFIG_PROC_FS
1915 /* Proc filesystem TCPv6 sock list dumping. */
1916 static void get_openreq6(struct seq_file *seq,
1917 struct sock *sk, struct request_sock *req, int i, int uid)
1918 {
1919 int ttd = req->expires - jiffies;
1920 struct in6_addr *src = &inet6_rsk(req)->loc_addr;
1921 struct in6_addr *dest = &inet6_rsk(req)->rmt_addr;
1922
1923 if (ttd < 0)
1924 ttd = 0;
1925
1926 seq_printf(seq,
1927 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1928 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
1929 i,
1930 src->s6_addr32[0], src->s6_addr32[1],
1931 src->s6_addr32[2], src->s6_addr32[3],
1932 ntohs(inet_sk(sk)->sport),
1933 dest->s6_addr32[0], dest->s6_addr32[1],
1934 dest->s6_addr32[2], dest->s6_addr32[3],
1935 ntohs(inet_rsk(req)->rmt_port),
1936 TCP_SYN_RECV,
1937 0,0, /* could print option size, but that is af dependent. */
1938 1, /* timers active (only the expire timer) */
1939 jiffies_to_clock_t(ttd),
1940 req->retrans,
1941 uid,
1942 0, /* non standard timer */
1943 0, /* open_requests have no inode */
1944 0, req);
1945 }
1946
1947 static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1948 {
1949 struct in6_addr *dest, *src;
1950 __u16 destp, srcp;
1951 int timer_active;
1952 unsigned long timer_expires;
1953 struct inet_sock *inet = inet_sk(sp);
1954 struct tcp_sock *tp = tcp_sk(sp);
1955 const struct inet_connection_sock *icsk = inet_csk(sp);
1956 struct ipv6_pinfo *np = inet6_sk(sp);
1957
1958 dest = &np->daddr;
1959 src = &np->rcv_saddr;
1960 destp = ntohs(inet->dport);
1961 srcp = ntohs(inet->sport);
1962
1963 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
1964 timer_active = 1;
1965 timer_expires = icsk->icsk_timeout;
1966 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
1967 timer_active = 4;
1968 timer_expires = icsk->icsk_timeout;
1969 } else if (timer_pending(&sp->sk_timer)) {
1970 timer_active = 2;
1971 timer_expires = sp->sk_timer.expires;
1972 } else {
1973 timer_active = 0;
1974 timer_expires = jiffies;
1975 }
1976
1977 seq_printf(seq,
1978 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1979 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %lu %lu %u %u %d\n",
1980 i,
1981 src->s6_addr32[0], src->s6_addr32[1],
1982 src->s6_addr32[2], src->s6_addr32[3], srcp,
1983 dest->s6_addr32[0], dest->s6_addr32[1],
1984 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1985 sp->sk_state,
1986 tp->write_seq-tp->snd_una,
1987 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
1988 timer_active,
1989 jiffies_to_clock_t(timer_expires - jiffies),
1990 icsk->icsk_retransmits,
1991 sock_i_uid(sp),
1992 icsk->icsk_probes_out,
1993 sock_i_ino(sp),
1994 atomic_read(&sp->sk_refcnt), sp,
1995 jiffies_to_clock_t(icsk->icsk_rto),
1996 jiffies_to_clock_t(icsk->icsk_ack.ato),
1997 (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
1998 tp->snd_cwnd, tp->snd_ssthresh>=0xFFFF?-1:tp->snd_ssthresh
1999 );
2000 }
2001
2002 static void get_timewait6_sock(struct seq_file *seq,
2003 struct inet_timewait_sock *tw, int i)
2004 {
2005 struct in6_addr *dest, *src;
2006 __u16 destp, srcp;
2007 struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
2008 int ttd = tw->tw_ttd - jiffies;
2009
2010 if (ttd < 0)
2011 ttd = 0;
2012
2013 dest = &tw6->tw_v6_daddr;
2014 src = &tw6->tw_v6_rcv_saddr;
2015 destp = ntohs(tw->tw_dport);
2016 srcp = ntohs(tw->tw_sport);
2017
2018 seq_printf(seq,
2019 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2020 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
2021 i,
2022 src->s6_addr32[0], src->s6_addr32[1],
2023 src->s6_addr32[2], src->s6_addr32[3], srcp,
2024 dest->s6_addr32[0], dest->s6_addr32[1],
2025 dest->s6_addr32[2], dest->s6_addr32[3], destp,
2026 tw->tw_substate, 0, 0,
2027 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
2028 atomic_read(&tw->tw_refcnt), tw);
2029 }
2030
2031 static int tcp6_seq_show(struct seq_file *seq, void *v)
2032 {
2033 struct tcp_iter_state *st;
2034
2035 if (v == SEQ_START_TOKEN) {
2036 seq_puts(seq,
2037 " sl "
2038 "local_address "
2039 "remote_address "
2040 "st tx_queue rx_queue tr tm->when retrnsmt"
2041 " uid timeout inode\n");
2042 goto out;
2043 }
2044 st = seq->private;
2045
2046 switch (st->state) {
2047 case TCP_SEQ_STATE_LISTENING:
2048 case TCP_SEQ_STATE_ESTABLISHED:
2049 get_tcp6_sock(seq, v, st->num);
2050 break;
2051 case TCP_SEQ_STATE_OPENREQ:
2052 get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
2053 break;
2054 case TCP_SEQ_STATE_TIME_WAIT:
2055 get_timewait6_sock(seq, v, st->num);
2056 break;
2057 }
2058 out:
2059 return 0;
2060 }
2061
2062 static struct tcp_seq_afinfo tcp6_seq_afinfo = {
2063 .name = "tcp6",
2064 .family = AF_INET6,
2065 .seq_fops = {
2066 .owner = THIS_MODULE,
2067 },
2068 .seq_ops = {
2069 .show = tcp6_seq_show,
2070 },
2071 };
2072
2073 int tcp6_proc_init(struct net *net)
2074 {
2075 return tcp_proc_register(net, &tcp6_seq_afinfo);
2076 }
2077
2078 void tcp6_proc_exit(struct net *net)
2079 {
2080 tcp_proc_unregister(net, &tcp6_seq_afinfo);
2081 }
2082 #endif
2083
2084 struct proto tcpv6_prot = {
2085 .name = "TCPv6",
2086 .owner = THIS_MODULE,
2087 .close = tcp_close,
2088 .connect = tcp_v6_connect,
2089 .disconnect = tcp_disconnect,
2090 .accept = inet_csk_accept,
2091 .ioctl = tcp_ioctl,
2092 .init = tcp_v6_init_sock,
2093 .destroy = tcp_v6_destroy_sock,
2094 .shutdown = tcp_shutdown,
2095 .setsockopt = tcp_setsockopt,
2096 .getsockopt = tcp_getsockopt,
2097 .recvmsg = tcp_recvmsg,
2098 .backlog_rcv = tcp_v6_do_rcv,
2099 .hash = tcp_v6_hash,
2100 .unhash = inet_unhash,
2101 .get_port = inet_csk_get_port,
2102 .enter_memory_pressure = tcp_enter_memory_pressure,
2103 .sockets_allocated = &tcp_sockets_allocated,
2104 .memory_allocated = &tcp_memory_allocated,
2105 .memory_pressure = &tcp_memory_pressure,
2106 .orphan_count = &tcp_orphan_count,
2107 .sysctl_mem = sysctl_tcp_mem,
2108 .sysctl_wmem = sysctl_tcp_wmem,
2109 .sysctl_rmem = sysctl_tcp_rmem,
2110 .max_header = MAX_TCP_HEADER,
2111 .obj_size = sizeof(struct tcp6_sock),
2112 .twsk_prot = &tcp6_timewait_sock_ops,
2113 .rsk_prot = &tcp6_request_sock_ops,
2114 .h.hashinfo = &tcp_hashinfo,
2115 #ifdef CONFIG_COMPAT
2116 .compat_setsockopt = compat_tcp_setsockopt,
2117 .compat_getsockopt = compat_tcp_getsockopt,
2118 #endif
2119 };
2120
2121 static struct inet6_protocol tcpv6_protocol = {
2122 .handler = tcp_v6_rcv,
2123 .err_handler = tcp_v6_err,
2124 .gso_send_check = tcp_v6_gso_send_check,
2125 .gso_segment = tcp_tso_segment,
2126 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
2127 };
2128
2129 static struct inet_protosw tcpv6_protosw = {
2130 .type = SOCK_STREAM,
2131 .protocol = IPPROTO_TCP,
2132 .prot = &tcpv6_prot,
2133 .ops = &inet6_stream_ops,
2134 .capability = -1,
2135 .no_check = 0,
2136 .flags = INET_PROTOSW_PERMANENT |
2137 INET_PROTOSW_ICSK,
2138 };
2139
2140 static int tcpv6_net_init(struct net *net)
2141 {
2142 return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
2143 SOCK_RAW, IPPROTO_TCP, net);
2144 }
2145
2146 static void tcpv6_net_exit(struct net *net)
2147 {
2148 inet_ctl_sock_destroy(net->ipv6.tcp_sk);
2149 }
2150
2151 static struct pernet_operations tcpv6_net_ops = {
2152 .init = tcpv6_net_init,
2153 .exit = tcpv6_net_exit,
2154 };
2155
2156 int __init tcpv6_init(void)
2157 {
2158 int ret;
2159
2160 ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
2161 if (ret)
2162 goto out;
2163
2164 /* register inet6 protocol */
2165 ret = inet6_register_protosw(&tcpv6_protosw);
2166 if (ret)
2167 goto out_tcpv6_protocol;
2168
2169 ret = register_pernet_subsys(&tcpv6_net_ops);
2170 if (ret)
2171 goto out_tcpv6_protosw;
2172 out:
2173 return ret;
2174
2175 out_tcpv6_protocol:
2176 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2177 out_tcpv6_protosw:
2178 inet6_unregister_protosw(&tcpv6_protosw);
2179 goto out;
2180 }
2181
2182 void tcpv6_exit(void)
2183 {
2184 unregister_pernet_subsys(&tcpv6_net_ops);
2185 inet6_unregister_protosw(&tcpv6_protosw);
2186 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2187 }