?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
[linux-2.6/x86.git] / net / ipv6 / af_inet6.c
blobfbf533cc9dcefdfb5b41f001224192518f855855
1 /*
2 * PF_INET6 socket protocol family
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Adapted from linux/net/ipv4/af_inet.c
9 *
10 * Fixes:
11 * piggy, Karl Knutson : Socket protocol table
12 * Hideaki YOSHIFUJI : sin6_scope_id support
13 * Arnaldo Melo : check proc_net_create return, cleanups
14 *
15 * This program is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU General Public License
17 * as published by the Free Software Foundation; either version
18 * 2 of the License, or (at your option) any later version.
19 */
20
21
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/in.h>
28 #include <linux/kernel.h>
29 #include <linux/timer.h>
30 #include <linux/string.h>
31 #include <linux/sockios.h>
32 #include <linux/net.h>
33 #include <linux/fcntl.h>
34 #include <linux/mm.h>
35 #include <linux/interrupt.h>
36 #include <linux/proc_fs.h>
37 #include <linux/stat.h>
38 #include <linux/init.h>
39
40 #include <linux/inet.h>
41 #include <linux/netdevice.h>
42 #include <linux/icmpv6.h>
43 #include <linux/netfilter_ipv6.h>
44
45 #include <net/ip.h>
46 #include <net/ipv6.h>
47 #include <net/udp.h>
48 #include <net/udplite.h>
49 #include <net/tcp.h>
50 #include <net/ipip.h>
51 #include <net/protocol.h>
52 #include <net/inet_common.h>
53 #include <net/route.h>
54 #include <net/transp_v6.h>
55 #include <net/ip6_route.h>
56 #include <net/addrconf.h>
57 #ifdef CONFIG_IPV6_TUNNEL
58 #include <net/ip6_tunnel.h>
59 #endif
60
61 #include <asm/uaccess.h>
62 #include <asm/system.h>
63 #include <linux/mroute6.h>
64
65 MODULE_AUTHOR("Cast of dozens");
66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
67 MODULE_LICENSE("GPL");
68
69 /* The inetsw6 table contains everything that inet6_create needs to
70 * build a new socket.
71 */
72 static struct list_head inetsw6[SOCK_MAX];
73 static DEFINE_SPINLOCK(inetsw6_lock);
74
75 static int disable_ipv6 = 0;
76 module_param_named(disable, disable_ipv6, int, 0);
77 MODULE_PARM_DESC(disable, "Disable IPv6 such that it is non-functional");
78
79 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
80 {
81 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
82
83 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
84 }
85
86 static int inet6_create(struct net *net, struct socket *sock, int protocol)
87 {
88 struct inet_sock *inet;
89 struct ipv6_pinfo *np;
90 struct sock *sk;
91 struct inet_protosw *answer;
92 struct proto *answer_prot;
93 unsigned char answer_flags;
94 char answer_no_check;
95 int try_loading_module = 0;
96 int err;
97
98 if (sock->type != SOCK_RAW &&
99 sock->type != SOCK_DGRAM &&
100 !inet_ehash_secret)
101 build_ehash_secret();
102
103 /* Look for the requested type/protocol pair. */
104 lookup_protocol:
105 err = -ESOCKTNOSUPPORT;
106 rcu_read_lock();
107 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
108
109 err = 0;
110 /* Check the non-wild match. */
111 if (protocol == answer->protocol) {
112 if (protocol != IPPROTO_IP)
113 break;
114 } else {
115 /* Check for the two wild cases. */
116 if (IPPROTO_IP == protocol) {
117 protocol = answer->protocol;
118 break;
119 }
120 if (IPPROTO_IP == answer->protocol)
121 break;
122 }
123 err = -EPROTONOSUPPORT;
124 }
125
126 if (err) {
127 if (try_loading_module < 2) {
128 rcu_read_unlock();
129 /*
130 * Be more specific, e.g. net-pf-10-proto-132-type-1
131 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
132 */
133 if (++try_loading_module == 1)
134 request_module("net-pf-%d-proto-%d-type-%d",
135 PF_INET6, protocol, sock->type);
136 /*
137 * Fall back to generic, e.g. net-pf-10-proto-132
138 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
139 */
140 else
141 request_module("net-pf-%d-proto-%d",
142 PF_INET6, protocol);
143 goto lookup_protocol;
144 } else
145 goto out_rcu_unlock;
146 }
147
148 err = -EPERM;
149 if (answer->capability > 0 && !capable(answer->capability))
150 goto out_rcu_unlock;
151
152 sock->ops = answer->ops;
153 answer_prot = answer->prot;
154 answer_no_check = answer->no_check;
155 answer_flags = answer->flags;
156 rcu_read_unlock();
157
158 WARN_ON(answer_prot->slab == NULL);
159
160 err = -ENOBUFS;
161 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
162 if (sk == NULL)
163 goto out;
164
165 sock_init_data(sock, sk);
166
167 err = 0;
168 sk->sk_no_check = answer_no_check;
169 if (INET_PROTOSW_REUSE & answer_flags)
170 sk->sk_reuse = 1;
171
172 inet = inet_sk(sk);
173 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
174
175 if (SOCK_RAW == sock->type) {
176 inet->num = protocol;
177 if (IPPROTO_RAW == protocol)
178 inet->hdrincl = 1;
179 }
180
181 sk->sk_destruct = inet_sock_destruct;
182 sk->sk_family = PF_INET6;
183 sk->sk_protocol = protocol;
184
185 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
186
187 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
188 np->hop_limit = -1;
189 np->mcast_hops = -1;
190 np->mc_loop = 1;
191 np->pmtudisc = IPV6_PMTUDISC_WANT;
192 np->ipv6only = net->ipv6.sysctl.bindv6only;
193
194 /* Init the ipv4 part of the socket since we can have sockets
195 * using v6 API for ipv4.
196 */
197 inet->uc_ttl = -1;
198
199 inet->mc_loop = 1;
200 inet->mc_ttl = 1;
201 inet->mc_index = 0;
202 inet->mc_list = NULL;
203
204 if (ipv4_config.no_pmtu_disc)
205 inet->pmtudisc = IP_PMTUDISC_DONT;
206 else
207 inet->pmtudisc = IP_PMTUDISC_WANT;
208 /*
209 * Increment only the relevant sk_prot->socks debug field, this changes
210 * the previous behaviour of incrementing both the equivalent to
211 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
212 *
213 * This allows better debug granularity as we'll know exactly how many
214 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
215 * transport protocol socks. -acme
216 */
217 sk_refcnt_debug_inc(sk);
218
219 if (inet->num) {
220 /* It assumes that any protocol which allows
221 * the user to assign a number at socket
222 * creation time automatically shares.
223 */
224 inet->sport = htons(inet->num);
225 sk->sk_prot->hash(sk);
226 }
227 if (sk->sk_prot->init) {
228 err = sk->sk_prot->init(sk);
229 if (err) {
230 sk_common_release(sk);
231 goto out;
232 }
233 }
234 out:
235 return err;
236 out_rcu_unlock:
237 rcu_read_unlock();
238 goto out;
239 }
240
241
242 /* bind for INET6 API */
243 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
244 {
245 struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
246 struct sock *sk = sock->sk;
247 struct inet_sock *inet = inet_sk(sk);
248 struct ipv6_pinfo *np = inet6_sk(sk);
249 struct net *net = sock_net(sk);
250 __be32 v4addr = 0;
251 unsigned short snum;
252 int addr_type = 0;
253 int err = 0;
254
255 /* If the socket has its own bind function then use it. */
256 if (sk->sk_prot->bind)
257 return sk->sk_prot->bind(sk, uaddr, addr_len);
258
259 if (addr_len < SIN6_LEN_RFC2133)
260 return -EINVAL;
261 addr_type = ipv6_addr_type(&addr->sin6_addr);
262 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
263 return -EINVAL;
264
265 snum = ntohs(addr->sin6_port);
266 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
267 return -EACCES;
268
269 lock_sock(sk);
270
271 /* Check these errors (active socket, double bind). */
272 if (sk->sk_state != TCP_CLOSE || inet->num) {
273 err = -EINVAL;
274 goto out;
275 }
276
277 /* Check if the address belongs to the host. */
278 if (addr_type == IPV6_ADDR_MAPPED) {
279 v4addr = addr->sin6_addr.s6_addr32[3];
280 if (inet_addr_type(net, v4addr) != RTN_LOCAL) {
281 err = -EADDRNOTAVAIL;
282 goto out;
283 }
284 } else {
285 if (addr_type != IPV6_ADDR_ANY) {
286 struct net_device *dev = NULL;
287
288 if (addr_type & IPV6_ADDR_LINKLOCAL) {
289 if (addr_len >= sizeof(struct sockaddr_in6) &&
290 addr->sin6_scope_id) {
291 /* Override any existing binding, if another one
292 * is supplied by user.
293 */
294 sk->sk_bound_dev_if = addr->sin6_scope_id;
295 }
296
297 /* Binding to link-local address requires an interface */
298 if (!sk->sk_bound_dev_if) {
299 err = -EINVAL;
300 goto out;
301 }
302 dev = dev_get_by_index(net, sk->sk_bound_dev_if);
303 if (!dev) {
304 err = -ENODEV;
305 goto out;
306 }
307 }
308
309 /* ipv4 addr of the socket is invalid. Only the
310 * unspecified and mapped address have a v4 equivalent.
311 */
312 v4addr = LOOPBACK4_IPV6;
313 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
314 if (!ipv6_chk_addr(net, &addr->sin6_addr,
315 dev, 0)) {
316 if (dev)
317 dev_put(dev);
318 err = -EADDRNOTAVAIL;
319 goto out;
320 }
321 }
322 if (dev)
323 dev_put(dev);
324 }
325 }
326
327 inet->rcv_saddr = v4addr;
328 inet->saddr = v4addr;
329
330 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
331
332 if (!(addr_type & IPV6_ADDR_MULTICAST))
333 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
334
335 /* Make sure we are allowed to bind here. */
336 if (sk->sk_prot->get_port(sk, snum)) {
337 inet_reset_saddr(sk);
338 err = -EADDRINUSE;
339 goto out;
340 }
341
342 if (addr_type != IPV6_ADDR_ANY)
343 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
344 if (snum)
345 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
346 inet->sport = htons(inet->num);
347 inet->dport = 0;
348 inet->daddr = 0;
349 out:
350 release_sock(sk);
351 return err;
352 }
353
354 EXPORT_SYMBOL(inet6_bind);
355
356 int inet6_release(struct socket *sock)
357 {
358 struct sock *sk = sock->sk;
359
360 if (sk == NULL)
361 return -EINVAL;
362
363 /* Free mc lists */
364 ipv6_sock_mc_close(sk);
365
366 /* Free ac lists */
367 ipv6_sock_ac_close(sk);
368
369 return inet_release(sock);
370 }
371
372 EXPORT_SYMBOL(inet6_release);
373
374 void inet6_destroy_sock(struct sock *sk)
375 {
376 struct ipv6_pinfo *np = inet6_sk(sk);
377 struct sk_buff *skb;
378 struct ipv6_txoptions *opt;
379
380 /* Release rx options */
381
382 if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
383 kfree_skb(skb);
384
385 /* Free flowlabels */
386 fl6_free_socklist(sk);
387
388 /* Free tx options */
389
390 if ((opt = xchg(&np->opt, NULL)) != NULL)
391 sock_kfree_s(sk, opt, opt->tot_len);
392 }
393
394 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
395
396 /*
397 * This does both peername and sockname.
398 */
399
400 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
401 int *uaddr_len, int peer)
402 {
403 struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
404 struct sock *sk = sock->sk;
405 struct inet_sock *inet = inet_sk(sk);
406 struct ipv6_pinfo *np = inet6_sk(sk);
407
408 sin->sin6_family = AF_INET6;
409 sin->sin6_flowinfo = 0;
410 sin->sin6_scope_id = 0;
411 if (peer) {
412 if (!inet->dport)
413 return -ENOTCONN;
414 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
415 peer == 1)
416 return -ENOTCONN;
417 sin->sin6_port = inet->dport;
418 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
419 if (np->sndflow)
420 sin->sin6_flowinfo = np->flow_label;
421 } else {
422 if (ipv6_addr_any(&np->rcv_saddr))
423 ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
424 else
425 ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
426
427 sin->sin6_port = inet->sport;
428 }
429 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
430 sin->sin6_scope_id = sk->sk_bound_dev_if;
431 *uaddr_len = sizeof(*sin);
432 return(0);
433 }
434
435 EXPORT_SYMBOL(inet6_getname);
436
437 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
438 {
439 struct sock *sk = sock->sk;
440 struct net *net = sock_net(sk);
441
442 switch(cmd)
443 {
444 case SIOCGSTAMP:
445 return sock_get_timestamp(sk, (struct timeval __user *)arg);
446
447 case SIOCGSTAMPNS:
448 return sock_get_timestampns(sk, (struct timespec __user *)arg);
449
450 case SIOCADDRT:
451 case SIOCDELRT:
452
453 return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
454
455 case SIOCSIFADDR:
456 return addrconf_add_ifaddr(net, (void __user *) arg);
457 case SIOCDIFADDR:
458 return addrconf_del_ifaddr(net, (void __user *) arg);
459 case SIOCSIFDSTADDR:
460 return addrconf_set_dstaddr(net, (void __user *) arg);
461 default:
462 if (!sk->sk_prot->ioctl)
463 return -ENOIOCTLCMD;
464 return sk->sk_prot->ioctl(sk, cmd, arg);
465 }
466 /*NOTREACHED*/
467 return(0);
468 }
469
470 EXPORT_SYMBOL(inet6_ioctl);
471
472 const struct proto_ops inet6_stream_ops = {
473 .family = PF_INET6,
474 .owner = THIS_MODULE,
475 .release = inet6_release,
476 .bind = inet6_bind,
477 .connect = inet_stream_connect, /* ok */
478 .socketpair = sock_no_socketpair, /* a do nothing */
479 .accept = inet_accept, /* ok */
480 .getname = inet6_getname,
481 .poll = tcp_poll, /* ok */
482 .ioctl = inet6_ioctl, /* must change */
483 .listen = inet_listen, /* ok */
484 .shutdown = inet_shutdown, /* ok */
485 .setsockopt = sock_common_setsockopt, /* ok */
486 .getsockopt = sock_common_getsockopt, /* ok */
487 .sendmsg = tcp_sendmsg, /* ok */
488 .recvmsg = sock_common_recvmsg, /* ok */
489 .mmap = sock_no_mmap,
490 .sendpage = tcp_sendpage,
491 .splice_read = tcp_splice_read,
492 #ifdef CONFIG_COMPAT
493 .compat_setsockopt = compat_sock_common_setsockopt,
494 .compat_getsockopt = compat_sock_common_getsockopt,
495 #endif
496 };
497
498 const struct proto_ops inet6_dgram_ops = {
499 .family = PF_INET6,
500 .owner = THIS_MODULE,
501 .release = inet6_release,
502 .bind = inet6_bind,
503 .connect = inet_dgram_connect, /* ok */
504 .socketpair = sock_no_socketpair, /* a do nothing */
505 .accept = sock_no_accept, /* a do nothing */
506 .getname = inet6_getname,
507 .poll = udp_poll, /* ok */
508 .ioctl = inet6_ioctl, /* must change */
509 .listen = sock_no_listen, /* ok */
510 .shutdown = inet_shutdown, /* ok */
511 .setsockopt = sock_common_setsockopt, /* ok */
512 .getsockopt = sock_common_getsockopt, /* ok */
513 .sendmsg = inet_sendmsg, /* ok */
514 .recvmsg = sock_common_recvmsg, /* ok */
515 .mmap = sock_no_mmap,
516 .sendpage = sock_no_sendpage,
517 #ifdef CONFIG_COMPAT
518 .compat_setsockopt = compat_sock_common_setsockopt,
519 .compat_getsockopt = compat_sock_common_getsockopt,
520 #endif
521 };
522
523 static struct net_proto_family inet6_family_ops = {
524 .family = PF_INET6,
525 .create = inet6_create,
526 .owner = THIS_MODULE,
527 };
528
529 int inet6_register_protosw(struct inet_protosw *p)
530 {
531 struct list_head *lh;
532 struct inet_protosw *answer;
533 struct list_head *last_perm;
534 int protocol = p->protocol;
535 int ret;
536
537 spin_lock_bh(&inetsw6_lock);
538
539 ret = -EINVAL;
540 if (p->type >= SOCK_MAX)
541 goto out_illegal;
542
543 /* If we are trying to override a permanent protocol, bail. */
544 answer = NULL;
545 ret = -EPERM;
546 last_perm = &inetsw6[p->type];
547 list_for_each(lh, &inetsw6[p->type]) {
548 answer = list_entry(lh, struct inet_protosw, list);
549
550 /* Check only the non-wild match. */
551 if (INET_PROTOSW_PERMANENT & answer->flags) {
552 if (protocol == answer->protocol)
553 break;
554 last_perm = lh;
555 }
556
557 answer = NULL;
558 }
559 if (answer)
560 goto out_permanent;
561
562 /* Add the new entry after the last permanent entry if any, so that
563 * the new entry does not override a permanent entry when matched with
564 * a wild-card protocol. But it is allowed to override any existing
565 * non-permanent entry. This means that when we remove this entry, the
566 * system automatically returns to the old behavior.
567 */
568 list_add_rcu(&p->list, last_perm);
569 ret = 0;
570 out:
571 spin_unlock_bh(&inetsw6_lock);
572 return ret;
573
574 out_permanent:
575 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
576 protocol);
577 goto out;
578
579 out_illegal:
580 printk(KERN_ERR
581 "Ignoring attempt to register invalid socket type %d.\n",
582 p->type);
583 goto out;
584 }
585
586 EXPORT_SYMBOL(inet6_register_protosw);
587
588 void
589 inet6_unregister_protosw(struct inet_protosw *p)
590 {
591 if (INET_PROTOSW_PERMANENT & p->flags) {
592 printk(KERN_ERR
593 "Attempt to unregister permanent protocol %d.\n",
594 p->protocol);
595 } else {
596 spin_lock_bh(&inetsw6_lock);
597 list_del_rcu(&p->list);
598 spin_unlock_bh(&inetsw6_lock);
599
600 synchronize_net();
601 }
602 }
603
604 EXPORT_SYMBOL(inet6_unregister_protosw);
605
606 int inet6_sk_rebuild_header(struct sock *sk)
607 {
608 int err;
609 struct dst_entry *dst;
610 struct ipv6_pinfo *np = inet6_sk(sk);
611
612 dst = __sk_dst_check(sk, np->dst_cookie);
613
614 if (dst == NULL) {
615 struct inet_sock *inet = inet_sk(sk);
616 struct in6_addr *final_p = NULL, final;
617 struct flowi fl;
618
619 memset(&fl, 0, sizeof(fl));
620 fl.proto = sk->sk_protocol;
621 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
622 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
623 fl.fl6_flowlabel = np->flow_label;
624 fl.oif = sk->sk_bound_dev_if;
625 fl.fl_ip_dport = inet->dport;
626 fl.fl_ip_sport = inet->sport;
627 security_sk_classify_flow(sk, &fl);
628
629 if (np->opt && np->opt->srcrt) {
630 struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
631 ipv6_addr_copy(&final, &fl.fl6_dst);
632 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
633 final_p = &final;
634 }
635
636 err = ip6_dst_lookup(sk, &dst, &fl);
637 if (err) {
638 sk->sk_route_caps = 0;
639 return err;
640 }
641 if (final_p)
642 ipv6_addr_copy(&fl.fl6_dst, final_p);
643
644 if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
645 sk->sk_err_soft = -err;
646 return err;
647 }
648
649 __ip6_dst_store(sk, dst, NULL, NULL);
650 }
651
652 return 0;
653 }
654
655 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
656
657 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
658 {
659 struct ipv6_pinfo *np = inet6_sk(sk);
660 struct inet6_skb_parm *opt = IP6CB(skb);
661
662 if (np->rxopt.all) {
663 if ((opt->hop && (np->rxopt.bits.hopopts ||
664 np->rxopt.bits.ohopopts)) ||
665 ((IPV6_FLOWINFO_MASK &
666 *(__be32 *)skb_network_header(skb)) &&
667 np->rxopt.bits.rxflow) ||
668 (opt->srcrt && (np->rxopt.bits.srcrt ||
669 np->rxopt.bits.osrcrt)) ||
670 ((opt->dst1 || opt->dst0) &&
671 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
672 return 1;
673 }
674 return 0;
675 }
676
677 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
678
679 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
680 {
681 struct inet6_protocol *ops = NULL;
682
683 for (;;) {
684 struct ipv6_opt_hdr *opth;
685 int len;
686
687 if (proto != NEXTHDR_HOP) {
688 ops = rcu_dereference(inet6_protos[proto]);
689
690 if (unlikely(!ops))
691 break;
692
693 if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
694 break;
695 }
696
697 if (unlikely(!pskb_may_pull(skb, 8)))
698 break;
699
700 opth = (void *)skb->data;
701 len = ipv6_optlen(opth);
702
703 if (unlikely(!pskb_may_pull(skb, len)))
704 break;
705
706 proto = opth->nexthdr;
707 __skb_pull(skb, len);
708 }
709
710 return proto;
711 }
712
713 static int ipv6_gso_send_check(struct sk_buff *skb)
714 {
715 struct ipv6hdr *ipv6h;
716 struct inet6_protocol *ops;
717 int err = -EINVAL;
718
719 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
720 goto out;
721
722 ipv6h = ipv6_hdr(skb);
723 __skb_pull(skb, sizeof(*ipv6h));
724 err = -EPROTONOSUPPORT;
725
726 rcu_read_lock();
727 ops = rcu_dereference(inet6_protos[
728 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
729
730 if (likely(ops && ops->gso_send_check)) {
731 skb_reset_transport_header(skb);
732 err = ops->gso_send_check(skb);
733 }
734 rcu_read_unlock();
735
736 out:
737 return err;
738 }
739
740 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
741 {
742 struct sk_buff *segs = ERR_PTR(-EINVAL);
743 struct ipv6hdr *ipv6h;
744 struct inet6_protocol *ops;
745
746 if (!(features & NETIF_F_V6_CSUM))
747 features &= ~NETIF_F_SG;
748
749 if (unlikely(skb_shinfo(skb)->gso_type &
750 ~(SKB_GSO_UDP |
751 SKB_GSO_DODGY |
752 SKB_GSO_TCP_ECN |
753 SKB_GSO_TCPV6 |
754 0)))
755 goto out;
756
757 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
758 goto out;
759
760 ipv6h = ipv6_hdr(skb);
761 __skb_pull(skb, sizeof(*ipv6h));
762 segs = ERR_PTR(-EPROTONOSUPPORT);
763
764 rcu_read_lock();
765 ops = rcu_dereference(inet6_protos[
766 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
767
768 if (likely(ops && ops->gso_segment)) {
769 skb_reset_transport_header(skb);
770 segs = ops->gso_segment(skb, features);
771 }
772 rcu_read_unlock();
773
774 if (unlikely(IS_ERR(segs)))
775 goto out;
776
777 for (skb = segs; skb; skb = skb->next) {
778 ipv6h = ipv6_hdr(skb);
779 ipv6h->payload_len = htons(skb->len - skb->mac_len -
780 sizeof(*ipv6h));
781 }
782
783 out:
784 return segs;
785 }
786
787 struct ipv6_gro_cb {
788 struct napi_gro_cb napi;
789 int proto;
790 };
791
792 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
793
794 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
795 struct sk_buff *skb)
796 {
797 struct inet6_protocol *ops;
798 struct sk_buff **pp = NULL;
799 struct sk_buff *p;
800 struct ipv6hdr *iph;
801 unsigned int nlen;
802 int flush = 1;
803 int proto;
804 __wsum csum;
805
806 iph = skb_gro_header(skb, sizeof(*iph));
807 if (unlikely(!iph))
808 goto out;
809
810 skb_gro_pull(skb, sizeof(*iph));
811 skb_set_transport_header(skb, skb_gro_offset(skb));
812
813 flush += ntohs(iph->payload_len) != skb_gro_len(skb);
814
815 rcu_read_lock();
816 proto = iph->nexthdr;
817 ops = rcu_dereference(inet6_protos[proto]);
818 if (!ops || !ops->gro_receive) {
819 __pskb_pull(skb, skb_gro_offset(skb));
820 proto = ipv6_gso_pull_exthdrs(skb, proto);
821 skb_gro_pull(skb, -skb_transport_offset(skb));
822 skb_reset_transport_header(skb);
823 __skb_push(skb, skb_gro_offset(skb));
824
825 if (!ops || !ops->gro_receive)
826 goto out_unlock;
827
828 iph = ipv6_hdr(skb);
829 }
830
831 IPV6_GRO_CB(skb)->proto = proto;
832
833 flush--;
834 nlen = skb_network_header_len(skb);
835
836 for (p = *head; p; p = p->next) {
837 struct ipv6hdr *iph2;
838
839 if (!NAPI_GRO_CB(p)->same_flow)
840 continue;
841
842 iph2 = ipv6_hdr(p);
843
844 /* All fields must match except length. */
845 if (nlen != skb_network_header_len(p) ||
846 memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
847 memcmp(&iph->nexthdr, &iph2->nexthdr,
848 nlen - offsetof(struct ipv6hdr, nexthdr))) {
849 NAPI_GRO_CB(p)->same_flow = 0;
850 continue;
851 }
852
853 NAPI_GRO_CB(p)->flush |= flush;
854 }
855
856 NAPI_GRO_CB(skb)->flush |= flush;
857
858 csum = skb->csum;
859 skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
860
861 pp = ops->gro_receive(head, skb);
862
863 skb->csum = csum;
864
865 out_unlock:
866 rcu_read_unlock();
867
868 out:
869 NAPI_GRO_CB(skb)->flush |= flush;
870
871 return pp;
872 }
873
874 static int ipv6_gro_complete(struct sk_buff *skb)
875 {
876 struct inet6_protocol *ops;
877 struct ipv6hdr *iph = ipv6_hdr(skb);
878 int err = -ENOSYS;
879
880 iph->payload_len = htons(skb->len - skb_network_offset(skb) -
881 sizeof(*iph));
882
883 rcu_read_lock();
884 ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
885 if (WARN_ON(!ops || !ops->gro_complete))
886 goto out_unlock;
887
888 err = ops->gro_complete(skb);
889
890 out_unlock:
891 rcu_read_unlock();
892
893 return err;
894 }
895
896 static struct packet_type ipv6_packet_type __read_mostly = {
897 .type = cpu_to_be16(ETH_P_IPV6),
898 .func = ipv6_rcv,
899 .gso_send_check = ipv6_gso_send_check,
900 .gso_segment = ipv6_gso_segment,
901 .gro_receive = ipv6_gro_receive,
902 .gro_complete = ipv6_gro_complete,
903 };
904
905 static int __init ipv6_packet_init(void)
906 {
907 dev_add_pack(&ipv6_packet_type);
908 return 0;
909 }
910
911 static void ipv6_packet_cleanup(void)
912 {
913 dev_remove_pack(&ipv6_packet_type);
914 }
915
916 static int __net_init ipv6_init_mibs(struct net *net)
917 {
918 if (snmp_mib_init((void **)net->mib.udp_stats_in6,
919 sizeof (struct udp_mib)) < 0)
920 return -ENOMEM;
921 if (snmp_mib_init((void **)net->mib.udplite_stats_in6,
922 sizeof (struct udp_mib)) < 0)
923 goto err_udplite_mib;
924 if (snmp_mib_init((void **)net->mib.ipv6_statistics,
925 sizeof(struct ipstats_mib)) < 0)
926 goto err_ip_mib;
927 if (snmp_mib_init((void **)net->mib.icmpv6_statistics,
928 sizeof(struct icmpv6_mib)) < 0)
929 goto err_icmp_mib;
930 if (snmp_mib_init((void **)net->mib.icmpv6msg_statistics,
931 sizeof(struct icmpv6msg_mib)) < 0)
932 goto err_icmpmsg_mib;
933 return 0;
934
935 err_icmpmsg_mib:
936 snmp_mib_free((void **)net->mib.icmpv6_statistics);
937 err_icmp_mib:
938 snmp_mib_free((void **)net->mib.ipv6_statistics);
939 err_ip_mib:
940 snmp_mib_free((void **)net->mib.udplite_stats_in6);
941 err_udplite_mib:
942 snmp_mib_free((void **)net->mib.udp_stats_in6);
943 return -ENOMEM;
944 }
945
946 static void __net_exit ipv6_cleanup_mibs(struct net *net)
947 {
948 snmp_mib_free((void **)net->mib.udp_stats_in6);
949 snmp_mib_free((void **)net->mib.udplite_stats_in6);
950 snmp_mib_free((void **)net->mib.ipv6_statistics);
951 snmp_mib_free((void **)net->mib.icmpv6_statistics);
952 snmp_mib_free((void **)net->mib.icmpv6msg_statistics);
953 }
954
955 static int __net_init inet6_net_init(struct net *net)
956 {
957 int err = 0;
958
959 net->ipv6.sysctl.bindv6only = 0;
960 net->ipv6.sysctl.icmpv6_time = 1*HZ;
961
962 err = ipv6_init_mibs(net);
963 if (err)
964 return err;
965 #ifdef CONFIG_PROC_FS
966 err = udp6_proc_init(net);
967 if (err)
968 goto out;
969 err = tcp6_proc_init(net);
970 if (err)
971 goto proc_tcp6_fail;
972 err = ac6_proc_init(net);
973 if (err)
974 goto proc_ac6_fail;
975 #endif
976 return err;
977
978 #ifdef CONFIG_PROC_FS
979 proc_ac6_fail:
980 tcp6_proc_exit(net);
981 proc_tcp6_fail:
982 udp6_proc_exit(net);
983 out:
984 ipv6_cleanup_mibs(net);
985 return err;
986 #endif
987 }
988
989 static void inet6_net_exit(struct net *net)
990 {
991 #ifdef CONFIG_PROC_FS
992 udp6_proc_exit(net);
993 tcp6_proc_exit(net);
994 ac6_proc_exit(net);
995 #endif
996 ipv6_cleanup_mibs(net);
997 }
998
999 static struct pernet_operations inet6_net_ops = {
1000 .init = inet6_net_init,
1001 .exit = inet6_net_exit,
1002 };
1003
1004 static int __init inet6_init(void)
1005 {
1006 struct sk_buff *dummy_skb;
1007 struct list_head *r;
1008 int err = 0;
1009
1010 BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1011
1012 /* Register the socket-side information for inet6_create. */
1013 for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1014 INIT_LIST_HEAD(r);
1015
1016 if (disable_ipv6) {
1017 printk(KERN_INFO
1018 "IPv6: Loaded, but administratively disabled, "
1019 "reboot required to enable\n");
1020 goto out;
1021 }
1022
1023 err = proto_register(&tcpv6_prot, 1);
1024 if (err)
1025 goto out;
1026
1027 err = proto_register(&udpv6_prot, 1);
1028 if (err)
1029 goto out_unregister_tcp_proto;
1030
1031 err = proto_register(&udplitev6_prot, 1);
1032 if (err)
1033 goto out_unregister_udp_proto;
1034
1035 err = proto_register(&rawv6_prot, 1);
1036 if (err)
1037 goto out_unregister_udplite_proto;
1038
1039
1040 /* We MUST register RAW sockets before we create the ICMP6,
1041 * IGMP6, or NDISC control sockets.
1042 */
1043 err = rawv6_init();
1044 if (err)
1045 goto out_unregister_raw_proto;
1046
1047 /* Register the family here so that the init calls below will
1048 * be able to create sockets. (?? is this dangerous ??)
1049 */
1050 err = sock_register(&inet6_family_ops);
1051 if (err)
1052 goto out_sock_register_fail;
1053
1054 #ifdef CONFIG_SYSCTL
1055 err = ipv6_static_sysctl_register();
1056 if (err)
1057 goto static_sysctl_fail;
1058 #endif
1059 /*
1060 * ipngwg API draft makes clear that the correct semantics
1061 * for TCP and UDP is to consider one TCP and UDP instance
1062 * in a host availiable by both INET and INET6 APIs and
1063 * able to communicate via both network protocols.
1064 */
1065
1066 err = register_pernet_subsys(&inet6_net_ops);
1067 if (err)
1068 goto register_pernet_fail;
1069 err = icmpv6_init();
1070 if (err)
1071 goto icmp_fail;
1072 err = ip6_mr_init();
1073 if (err)
1074 goto ipmr_fail;
1075 err = ndisc_init();
1076 if (err)
1077 goto ndisc_fail;
1078 err = igmp6_init();
1079 if (err)
1080 goto igmp_fail;
1081 err = ipv6_netfilter_init();
1082 if (err)
1083 goto netfilter_fail;
1084 /* Create /proc/foo6 entries. */
1085 #ifdef CONFIG_PROC_FS
1086 err = -ENOMEM;
1087 if (raw6_proc_init())
1088 goto proc_raw6_fail;
1089 if (udplite6_proc_init())
1090 goto proc_udplite6_fail;
1091 if (ipv6_misc_proc_init())
1092 goto proc_misc6_fail;
1093 if (if6_proc_init())
1094 goto proc_if6_fail;
1095 #endif
1096 err = ip6_route_init();
1097 if (err)
1098 goto ip6_route_fail;
1099 err = ip6_flowlabel_init();
1100 if (err)
1101 goto ip6_flowlabel_fail;
1102 err = addrconf_init();
1103 if (err)
1104 goto addrconf_fail;
1105
1106 /* Init v6 extension headers. */
1107 err = ipv6_exthdrs_init();
1108 if (err)
1109 goto ipv6_exthdrs_fail;
1110
1111 err = ipv6_frag_init();
1112 if (err)
1113 goto ipv6_frag_fail;
1114
1115 /* Init v6 transport protocols. */
1116 err = udpv6_init();
1117 if (err)
1118 goto udpv6_fail;
1119
1120 err = udplitev6_init();
1121 if (err)
1122 goto udplitev6_fail;
1123
1124 err = tcpv6_init();
1125 if (err)
1126 goto tcpv6_fail;
1127
1128 err = ipv6_packet_init();
1129 if (err)
1130 goto ipv6_packet_fail;
1131
1132 #ifdef CONFIG_SYSCTL
1133 err = ipv6_sysctl_register();
1134 if (err)
1135 goto sysctl_fail;
1136 #endif
1137 out:
1138 return err;
1139
1140 #ifdef CONFIG_SYSCTL
1141 sysctl_fail:
1142 ipv6_packet_cleanup();
1143 #endif
1144 ipv6_packet_fail:
1145 tcpv6_exit();
1146 tcpv6_fail:
1147 udplitev6_exit();
1148 udplitev6_fail:
1149 udpv6_exit();
1150 udpv6_fail:
1151 ipv6_frag_exit();
1152 ipv6_frag_fail:
1153 ipv6_exthdrs_exit();
1154 ipv6_exthdrs_fail:
1155 addrconf_cleanup();
1156 addrconf_fail:
1157 ip6_flowlabel_cleanup();
1158 ip6_flowlabel_fail:
1159 ip6_route_cleanup();
1160 ip6_route_fail:
1161 #ifdef CONFIG_PROC_FS
1162 if6_proc_exit();
1163 proc_if6_fail:
1164 ipv6_misc_proc_exit();
1165 proc_misc6_fail:
1166 udplite6_proc_exit();
1167 proc_udplite6_fail:
1168 raw6_proc_exit();
1169 proc_raw6_fail:
1170 #endif
1171 ipv6_netfilter_fini();
1172 netfilter_fail:
1173 igmp6_cleanup();
1174 igmp_fail:
1175 ndisc_cleanup();
1176 ndisc_fail:
1177 ip6_mr_cleanup();
1178 ipmr_fail:
1179 icmpv6_cleanup();
1180 icmp_fail:
1181 unregister_pernet_subsys(&inet6_net_ops);
1182 register_pernet_fail:
1183 #ifdef CONFIG_SYSCTL
1184 ipv6_static_sysctl_unregister();
1185 static_sysctl_fail:
1186 #endif
1187 sock_unregister(PF_INET6);
1188 rtnl_unregister_all(PF_INET6);
1189 out_sock_register_fail:
1190 rawv6_exit();
1191 out_unregister_raw_proto:
1192 proto_unregister(&rawv6_prot);
1193 out_unregister_udplite_proto:
1194 proto_unregister(&udplitev6_prot);
1195 out_unregister_udp_proto:
1196 proto_unregister(&udpv6_prot);
1197 out_unregister_tcp_proto:
1198 proto_unregister(&tcpv6_prot);
1199 goto out;
1200 }
1201 module_init(inet6_init);
1202
1203 static void __exit inet6_exit(void)
1204 {
1205 if (disable_ipv6)
1206 return;
1207
1208 /* First of all disallow new sockets creation. */
1209 sock_unregister(PF_INET6);
1210 /* Disallow any further netlink messages */
1211 rtnl_unregister_all(PF_INET6);
1212
1213 #ifdef CONFIG_SYSCTL
1214 ipv6_sysctl_unregister();
1215 #endif
1216 udpv6_exit();
1217 udplitev6_exit();
1218 tcpv6_exit();
1219
1220 /* Cleanup code parts. */
1221 ipv6_packet_cleanup();
1222 ipv6_frag_exit();
1223 ipv6_exthdrs_exit();
1224 addrconf_cleanup();
1225 ip6_flowlabel_cleanup();
1226 ip6_route_cleanup();
1227 #ifdef CONFIG_PROC_FS
1228
1229 /* Cleanup code parts. */
1230 if6_proc_exit();
1231 ipv6_misc_proc_exit();
1232 udplite6_proc_exit();
1233 raw6_proc_exit();
1234 #endif
1235 ipv6_netfilter_fini();
1236 igmp6_cleanup();
1237 ndisc_cleanup();
1238 ip6_mr_cleanup();
1239 icmpv6_cleanup();
1240 rawv6_exit();
1241
1242 unregister_pernet_subsys(&inet6_net_ops);
1243 #ifdef CONFIG_SYSCTL
1244 ipv6_static_sysctl_unregister();
1245 #endif
1246 proto_unregister(&rawv6_prot);
1247 proto_unregister(&udplitev6_prot);
1248 proto_unregister(&udpv6_prot);
1249 proto_unregister(&tcpv6_prot);
1250 }
1251 module_exit(inet6_exit);
1252
1253 MODULE_ALIAS_NETPROTO(PF_INET6);
x86 "subsystem" repository