e127a32f9540fecfaee5de6332ca1be64ce6541f
[linux-2.6.git] / net / ipv6 / af_inet6.c
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 struct ipv6_params ipv6_defaults = {
76         .disable_ipv6 = 0,
77         .autoconf = 1,
78 };
79
80 static int disable_ipv6_mod = 0;
81
82 module_param_named(disable, disable_ipv6_mod, int, 0444);
83 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
84
85 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
86 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
87
88 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
89 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
90
91 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
92 {
93         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
94
95         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
96 }
97
98 static int inet6_create(struct net *net, struct socket *sock, int protocol)
99 {
100         struct inet_sock *inet;
101         struct ipv6_pinfo *np;
102         struct sock *sk;
103         struct inet_protosw *answer;
104         struct proto *answer_prot;
105         unsigned char answer_flags;
106         char answer_no_check;
107         int try_loading_module = 0;
108         int err;
109
110         if (sock->type != SOCK_RAW &&
111             sock->type != SOCK_DGRAM &&
112             !inet_ehash_secret)
113                 build_ehash_secret();
114
115         /* Look for the requested type/protocol pair. */
116 lookup_protocol:
117         err = -ESOCKTNOSUPPORT;
118         rcu_read_lock();
119         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
120
121                 err = 0;
122                 /* Check the non-wild match. */
123                 if (protocol == answer->protocol) {
124                         if (protocol != IPPROTO_IP)
125                                 break;
126                 } else {
127                         /* Check for the two wild cases. */
128                         if (IPPROTO_IP == protocol) {
129                                 protocol = answer->protocol;
130                                 break;
131                         }
132                         if (IPPROTO_IP == answer->protocol)
133                                 break;
134                 }
135                 err = -EPROTONOSUPPORT;
136         }
137
138         if (err) {
139                 if (try_loading_module < 2) {
140                         rcu_read_unlock();
141                         /*
142                          * Be more specific, e.g. net-pf-10-proto-132-type-1
143                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
144                          */
145                         if (++try_loading_module == 1)
146                                 request_module("net-pf-%d-proto-%d-type-%d",
147                                                 PF_INET6, protocol, sock->type);
148                         /*
149                          * Fall back to generic, e.g. net-pf-10-proto-132
150                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
151                          */
152                         else
153                                 request_module("net-pf-%d-proto-%d",
154                                                 PF_INET6, protocol);
155                         goto lookup_protocol;
156                 } else
157                         goto out_rcu_unlock;
158         }
159
160         err = -EPERM;
161         if (answer->capability > 0 && !capable(answer->capability))
162                 goto out_rcu_unlock;
163
164         sock->ops = answer->ops;
165         answer_prot = answer->prot;
166         answer_no_check = answer->no_check;
167         answer_flags = answer->flags;
168         rcu_read_unlock();
169
170         WARN_ON(answer_prot->slab == NULL);
171
172         err = -ENOBUFS;
173         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
174         if (sk == NULL)
175                 goto out;
176
177         sock_init_data(sock, sk);
178
179         err = 0;
180         sk->sk_no_check = answer_no_check;
181         if (INET_PROTOSW_REUSE & answer_flags)
182                 sk->sk_reuse = 1;
183
184         inet = inet_sk(sk);
185         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
186
187         if (SOCK_RAW == sock->type) {
188                 inet->num = protocol;
189                 if (IPPROTO_RAW == protocol)
190                         inet->hdrincl = 1;
191         }
192
193         sk->sk_destruct         = inet_sock_destruct;
194         sk->sk_family           = PF_INET6;
195         sk->sk_protocol         = protocol;
196
197         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
198
199         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
200         np->hop_limit   = -1;
201         np->mcast_hops  = -1;
202         np->mc_loop     = 1;
203         np->pmtudisc    = IPV6_PMTUDISC_WANT;
204         np->ipv6only    = net->ipv6.sysctl.bindv6only;
205
206         /* Init the ipv4 part of the socket since we can have sockets
207          * using v6 API for ipv4.
208          */
209         inet->uc_ttl    = -1;
210
211         inet->mc_loop   = 1;
212         inet->mc_ttl    = 1;
213         inet->mc_index  = 0;
214         inet->mc_list   = NULL;
215
216         if (ipv4_config.no_pmtu_disc)
217                 inet->pmtudisc = IP_PMTUDISC_DONT;
218         else
219                 inet->pmtudisc = IP_PMTUDISC_WANT;
220         /*
221          * Increment only the relevant sk_prot->socks debug field, this changes
222          * the previous behaviour of incrementing both the equivalent to
223          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
224          *
225          * This allows better debug granularity as we'll know exactly how many
226          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
227          * transport protocol socks. -acme
228          */
229         sk_refcnt_debug_inc(sk);
230
231         if (inet->num) {
232                 /* It assumes that any protocol which allows
233                  * the user to assign a number at socket
234                  * creation time automatically shares.
235                  */
236                 inet->sport = htons(inet->num);
237                 sk->sk_prot->hash(sk);
238         }
239         if (sk->sk_prot->init) {
240                 err = sk->sk_prot->init(sk);
241                 if (err) {
242                         sk_common_release(sk);
243                         goto out;
244                 }
245         }
246 out:
247         return err;
248 out_rcu_unlock:
249         rcu_read_unlock();
250         goto out;
251 }
252
253
254 /* bind for INET6 API */
255 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
256 {
257         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
258         struct sock *sk = sock->sk;
259         struct inet_sock *inet = inet_sk(sk);
260         struct ipv6_pinfo *np = inet6_sk(sk);
261         struct net *net = sock_net(sk);
262         __be32 v4addr = 0;
263         unsigned short snum;
264         int addr_type = 0;
265         int err = 0;
266
267         /* If the socket has its own bind function then use it. */
268         if (sk->sk_prot->bind)
269                 return sk->sk_prot->bind(sk, uaddr, addr_len);
270
271         if (addr_len < SIN6_LEN_RFC2133)
272                 return -EINVAL;
273         addr_type = ipv6_addr_type(&addr->sin6_addr);
274         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
275                 return -EINVAL;
276
277         snum = ntohs(addr->sin6_port);
278         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
279                 return -EACCES;
280
281         lock_sock(sk);
282
283         /* Check these errors (active socket, double bind). */
284         if (sk->sk_state != TCP_CLOSE || inet->num) {
285                 err = -EINVAL;
286                 goto out;
287         }
288
289         /* Check if the address belongs to the host. */
290         if (addr_type == IPV6_ADDR_MAPPED) {
291                 int chk_addr_ret;
292
293                 /* Binding to v4-mapped address on a v6-only socket
294                  * makes no sense
295                  */
296                 if (np->ipv6only) {
297                         err = -EINVAL;
298                         goto out;
299                 }
300
301                 /* Reproduce AF_INET checks to make the bindings consitant */
302                 v4addr = addr->sin6_addr.s6_addr32[3];
303                 chk_addr_ret = inet_addr_type(net, v4addr);
304                 if (!sysctl_ip_nonlocal_bind &&
305                     !(inet->freebind || inet->transparent) &&
306                     v4addr != htonl(INADDR_ANY) &&
307                     chk_addr_ret != RTN_LOCAL &&
308                     chk_addr_ret != RTN_MULTICAST &&
309                     chk_addr_ret != RTN_BROADCAST) {
310                         err = -EADDRNOTAVAIL;
311                         goto out;
312                 }
313         } else {
314                 if (addr_type != IPV6_ADDR_ANY) {
315                         struct net_device *dev = NULL;
316
317                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
318                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
319                                     addr->sin6_scope_id) {
320                                         /* Override any existing binding, if another one
321                                          * is supplied by user.
322                                          */
323                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
324                                 }
325
326                                 /* Binding to link-local address requires an interface */
327                                 if (!sk->sk_bound_dev_if) {
328                                         err = -EINVAL;
329                                         goto out;
330                                 }
331                                 dev = dev_get_by_index(net, sk->sk_bound_dev_if);
332                                 if (!dev) {
333                                         err = -ENODEV;
334                                         goto out;
335                                 }
336                         }
337
338                         /* ipv4 addr of the socket is invalid.  Only the
339                          * unspecified and mapped address have a v4 equivalent.
340                          */
341                         v4addr = LOOPBACK4_IPV6;
342                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
343                                 if (!ipv6_chk_addr(net, &addr->sin6_addr,
344                                                    dev, 0)) {
345                                         if (dev)
346                                                 dev_put(dev);
347                                         err = -EADDRNOTAVAIL;
348                                         goto out;
349                                 }
350                         }
351                         if (dev)
352                                 dev_put(dev);
353                 }
354         }
355
356         inet->rcv_saddr = v4addr;
357         inet->saddr = v4addr;
358
359         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
360
361         if (!(addr_type & IPV6_ADDR_MULTICAST))
362                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
363
364         /* Make sure we are allowed to bind here. */
365         if (sk->sk_prot->get_port(sk, snum)) {
366                 inet_reset_saddr(sk);
367                 err = -EADDRINUSE;
368                 goto out;
369         }
370
371         if (addr_type != IPV6_ADDR_ANY) {
372                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
373                 if (addr_type != IPV6_ADDR_MAPPED)
374                         np->ipv6only = 1;
375         }
376         if (snum)
377                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
378         inet->sport = htons(inet->num);
379         inet->dport = 0;
380         inet->daddr = 0;
381 out:
382         release_sock(sk);
383         return err;
384 }
385
386 EXPORT_SYMBOL(inet6_bind);
387
388 int inet6_release(struct socket *sock)
389 {
390         struct sock *sk = sock->sk;
391
392         if (sk == NULL)
393                 return -EINVAL;
394
395         /* Free mc lists */
396         ipv6_sock_mc_close(sk);
397
398         /* Free ac lists */
399         ipv6_sock_ac_close(sk);
400
401         return inet_release(sock);
402 }
403
404 EXPORT_SYMBOL(inet6_release);
405
406 void inet6_destroy_sock(struct sock *sk)
407 {
408         struct ipv6_pinfo *np = inet6_sk(sk);
409         struct sk_buff *skb;
410         struct ipv6_txoptions *opt;
411
412         /* Release rx options */
413
414         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
415                 kfree_skb(skb);
416
417         /* Free flowlabels */
418         fl6_free_socklist(sk);
419
420         /* Free tx options */
421
422         if ((opt = xchg(&np->opt, NULL)) != NULL)
423                 sock_kfree_s(sk, opt, opt->tot_len);
424 }
425
426 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
427
428 /*
429  *      This does both peername and sockname.
430  */
431
432 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
433                  int *uaddr_len, int peer)
434 {
435         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
436         struct sock *sk = sock->sk;
437         struct inet_sock *inet = inet_sk(sk);
438         struct ipv6_pinfo *np = inet6_sk(sk);
439
440         sin->sin6_family = AF_INET6;
441         sin->sin6_flowinfo = 0;
442         sin->sin6_scope_id = 0;
443         if (peer) {
444                 if (!inet->dport)
445                         return -ENOTCONN;
446                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
447                     peer == 1)
448                         return -ENOTCONN;
449                 sin->sin6_port = inet->dport;
450                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
451                 if (np->sndflow)
452                         sin->sin6_flowinfo = np->flow_label;
453         } else {
454                 if (ipv6_addr_any(&np->rcv_saddr))
455                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
456                 else
457                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
458
459                 sin->sin6_port = inet->sport;
460         }
461         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
462                 sin->sin6_scope_id = sk->sk_bound_dev_if;
463         *uaddr_len = sizeof(*sin);
464         return(0);
465 }
466
467 EXPORT_SYMBOL(inet6_getname);
468
469 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
470 {
471         struct sock *sk = sock->sk;
472         struct net *net = sock_net(sk);
473
474         switch(cmd)
475         {
476         case SIOCGSTAMP:
477                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
478
479         case SIOCGSTAMPNS:
480                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
481
482         case SIOCADDRT:
483         case SIOCDELRT:
484
485                 return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
486
487         case SIOCSIFADDR:
488                 return addrconf_add_ifaddr(net, (void __user *) arg);
489         case SIOCDIFADDR:
490                 return addrconf_del_ifaddr(net, (void __user *) arg);
491         case SIOCSIFDSTADDR:
492                 return addrconf_set_dstaddr(net, (void __user *) arg);
493         default:
494                 if (!sk->sk_prot->ioctl)
495                         return -ENOIOCTLCMD;
496                 return sk->sk_prot->ioctl(sk, cmd, arg);
497         }
498         /*NOTREACHED*/
499         return(0);
500 }
501
502 EXPORT_SYMBOL(inet6_ioctl);
503
504 const struct proto_ops inet6_stream_ops = {
505         .family            = PF_INET6,
506         .owner             = THIS_MODULE,
507         .release           = inet6_release,
508         .bind              = inet6_bind,
509         .connect           = inet_stream_connect,       /* ok           */
510         .socketpair        = sock_no_socketpair,        /* a do nothing */
511         .accept            = inet_accept,               /* ok           */
512         .getname           = inet6_getname,
513         .poll              = tcp_poll,                  /* ok           */
514         .ioctl             = inet6_ioctl,               /* must change  */
515         .listen            = inet_listen,               /* ok           */
516         .shutdown          = inet_shutdown,             /* ok           */
517         .setsockopt        = sock_common_setsockopt,    /* ok           */
518         .getsockopt        = sock_common_getsockopt,    /* ok           */
519         .sendmsg           = tcp_sendmsg,               /* ok           */
520         .recvmsg           = sock_common_recvmsg,       /* ok           */
521         .mmap              = sock_no_mmap,
522         .sendpage          = tcp_sendpage,
523         .splice_read       = tcp_splice_read,
524 #ifdef CONFIG_COMPAT
525         .compat_setsockopt = compat_sock_common_setsockopt,
526         .compat_getsockopt = compat_sock_common_getsockopt,
527 #endif
528 };
529
530 const struct proto_ops inet6_dgram_ops = {
531         .family            = PF_INET6,
532         .owner             = THIS_MODULE,
533         .release           = inet6_release,
534         .bind              = inet6_bind,
535         .connect           = inet_dgram_connect,        /* ok           */
536         .socketpair        = sock_no_socketpair,        /* a do nothing */
537         .accept            = sock_no_accept,            /* a do nothing */
538         .getname           = inet6_getname,
539         .poll              = udp_poll,                  /* ok           */
540         .ioctl             = inet6_ioctl,               /* must change  */
541         .listen            = sock_no_listen,            /* ok           */
542         .shutdown          = inet_shutdown,             /* ok           */
543         .setsockopt        = sock_common_setsockopt,    /* ok           */
544         .getsockopt        = sock_common_getsockopt,    /* ok           */
545         .sendmsg           = inet_sendmsg,              /* ok           */
546         .recvmsg           = sock_common_recvmsg,       /* ok           */
547         .mmap              = sock_no_mmap,
548         .sendpage          = sock_no_sendpage,
549 #ifdef CONFIG_COMPAT
550         .compat_setsockopt = compat_sock_common_setsockopt,
551         .compat_getsockopt = compat_sock_common_getsockopt,
552 #endif
553 };
554
555 static struct net_proto_family inet6_family_ops = {
556         .family = PF_INET6,
557         .create = inet6_create,
558         .owner  = THIS_MODULE,
559 };
560
561 int inet6_register_protosw(struct inet_protosw *p)
562 {
563         struct list_head *lh;
564         struct inet_protosw *answer;
565         struct list_head *last_perm;
566         int protocol = p->protocol;
567         int ret;
568
569         spin_lock_bh(&inetsw6_lock);
570
571         ret = -EINVAL;
572         if (p->type >= SOCK_MAX)
573                 goto out_illegal;
574
575         /* If we are trying to override a permanent protocol, bail. */
576         answer = NULL;
577         ret = -EPERM;
578         last_perm = &inetsw6[p->type];
579         list_for_each(lh, &inetsw6[p->type]) {
580                 answer = list_entry(lh, struct inet_protosw, list);
581
582                 /* Check only the non-wild match. */
583                 if (INET_PROTOSW_PERMANENT & answer->flags) {
584                         if (protocol == answer->protocol)
585                                 break;
586                         last_perm = lh;
587                 }
588
589                 answer = NULL;
590         }
591         if (answer)
592                 goto out_permanent;
593
594         /* Add the new entry after the last permanent entry if any, so that
595          * the new entry does not override a permanent entry when matched with
596          * a wild-card protocol. But it is allowed to override any existing
597          * non-permanent entry.  This means that when we remove this entry, the
598          * system automatically returns to the old behavior.
599          */
600         list_add_rcu(&p->list, last_perm);
601         ret = 0;
602 out:
603         spin_unlock_bh(&inetsw6_lock);
604         return ret;
605
606 out_permanent:
607         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
608                protocol);
609         goto out;
610
611 out_illegal:
612         printk(KERN_ERR
613                "Ignoring attempt to register invalid socket type %d.\n",
614                p->type);
615         goto out;
616 }
617
618 EXPORT_SYMBOL(inet6_register_protosw);
619
620 void
621 inet6_unregister_protosw(struct inet_protosw *p)
622 {
623         if (INET_PROTOSW_PERMANENT & p->flags) {
624                 printk(KERN_ERR
625                        "Attempt to unregister permanent protocol %d.\n",
626                        p->protocol);
627         } else {
628                 spin_lock_bh(&inetsw6_lock);
629                 list_del_rcu(&p->list);
630                 spin_unlock_bh(&inetsw6_lock);
631
632                 synchronize_net();
633         }
634 }
635
636 EXPORT_SYMBOL(inet6_unregister_protosw);
637
638 int inet6_sk_rebuild_header(struct sock *sk)
639 {
640         int err;
641         struct dst_entry *dst;
642         struct ipv6_pinfo *np = inet6_sk(sk);
643
644         dst = __sk_dst_check(sk, np->dst_cookie);
645
646         if (dst == NULL) {
647                 struct inet_sock *inet = inet_sk(sk);
648                 struct in6_addr *final_p = NULL, final;
649                 struct flowi fl;
650
651                 memset(&fl, 0, sizeof(fl));
652                 fl.proto = sk->sk_protocol;
653                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
654                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
655                 fl.fl6_flowlabel = np->flow_label;
656                 fl.oif = sk->sk_bound_dev_if;
657                 fl.fl_ip_dport = inet->dport;
658                 fl.fl_ip_sport = inet->sport;
659                 security_sk_classify_flow(sk, &fl);
660
661                 if (np->opt && np->opt->srcrt) {
662                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
663                         ipv6_addr_copy(&final, &fl.fl6_dst);
664                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
665                         final_p = &final;
666                 }
667
668                 err = ip6_dst_lookup(sk, &dst, &fl);
669                 if (err) {
670                         sk->sk_route_caps = 0;
671                         return err;
672                 }
673                 if (final_p)
674                         ipv6_addr_copy(&fl.fl6_dst, final_p);
675
676                 if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
677                         sk->sk_err_soft = -err;
678                         return err;
679                 }
680
681                 __ip6_dst_store(sk, dst, NULL, NULL);
682         }
683
684         return 0;
685 }
686
687 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
688
689 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
690 {
691         struct ipv6_pinfo *np = inet6_sk(sk);
692         struct inet6_skb_parm *opt = IP6CB(skb);
693
694         if (np->rxopt.all) {
695                 if ((opt->hop && (np->rxopt.bits.hopopts ||
696                                   np->rxopt.bits.ohopopts)) ||
697                     ((IPV6_FLOWINFO_MASK &
698                       *(__be32 *)skb_network_header(skb)) &&
699                      np->rxopt.bits.rxflow) ||
700                     (opt->srcrt && (np->rxopt.bits.srcrt ||
701                      np->rxopt.bits.osrcrt)) ||
702                     ((opt->dst1 || opt->dst0) &&
703                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
704                         return 1;
705         }
706         return 0;
707 }
708
709 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
710
711 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
712 {
713         const struct inet6_protocol *ops = NULL;
714
715         for (;;) {
716                 struct ipv6_opt_hdr *opth;
717                 int len;
718
719                 if (proto != NEXTHDR_HOP) {
720                         ops = rcu_dereference(inet6_protos[proto]);
721
722                         if (unlikely(!ops))
723                                 break;
724
725                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
726                                 break;
727                 }
728
729                 if (unlikely(!pskb_may_pull(skb, 8)))
730                         break;
731
732                 opth = (void *)skb->data;
733                 len = ipv6_optlen(opth);
734
735                 if (unlikely(!pskb_may_pull(skb, len)))
736                         break;
737
738                 proto = opth->nexthdr;
739                 __skb_pull(skb, len);
740         }
741
742         return proto;
743 }
744
745 static int ipv6_gso_send_check(struct sk_buff *skb)
746 {
747         struct ipv6hdr *ipv6h;
748         const struct inet6_protocol *ops;
749         int err = -EINVAL;
750
751         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
752                 goto out;
753
754         ipv6h = ipv6_hdr(skb);
755         __skb_pull(skb, sizeof(*ipv6h));
756         err = -EPROTONOSUPPORT;
757
758         rcu_read_lock();
759         ops = rcu_dereference(inet6_protos[
760                 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
761
762         if (likely(ops && ops->gso_send_check)) {
763                 skb_reset_transport_header(skb);
764                 err = ops->gso_send_check(skb);
765         }
766         rcu_read_unlock();
767
768 out:
769         return err;
770 }
771
772 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
773 {
774         struct sk_buff *segs = ERR_PTR(-EINVAL);
775         struct ipv6hdr *ipv6h;
776         const struct inet6_protocol *ops;
777         int proto;
778         struct frag_hdr *fptr;
779         unsigned int unfrag_ip6hlen;
780         u8 *prevhdr;
781         int offset = 0;
782
783         if (!(features & NETIF_F_V6_CSUM))
784                 features &= ~NETIF_F_SG;
785
786         if (unlikely(skb_shinfo(skb)->gso_type &
787                      ~(SKB_GSO_UDP |
788                        SKB_GSO_DODGY |
789                        SKB_GSO_TCP_ECN |
790                        SKB_GSO_TCPV6 |
791                        0)))
792                 goto out;
793
794         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
795                 goto out;
796
797         ipv6h = ipv6_hdr(skb);
798         __skb_pull(skb, sizeof(*ipv6h));
799         segs = ERR_PTR(-EPROTONOSUPPORT);
800
801         proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
802         rcu_read_lock();
803         ops = rcu_dereference(inet6_protos[proto]);
804         if (likely(ops && ops->gso_segment)) {
805                 skb_reset_transport_header(skb);
806                 segs = ops->gso_segment(skb, features);
807         }
808         rcu_read_unlock();
809
810         if (unlikely(IS_ERR(segs)))
811                 goto out;
812
813         for (skb = segs; skb; skb = skb->next) {
814                 ipv6h = ipv6_hdr(skb);
815                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
816                                            sizeof(*ipv6h));
817                 if (proto == IPPROTO_UDP) {
818                         unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
819                         fptr = (struct frag_hdr *)(skb_network_header(skb) +
820                                 unfrag_ip6hlen);
821                         fptr->frag_off = htons(offset);
822                         if (skb->next != NULL)
823                                 fptr->frag_off |= htons(IP6_MF);
824                         offset += (ntohs(ipv6h->payload_len) -
825                                    sizeof(struct frag_hdr));
826                 }
827         }
828
829 out:
830         return segs;
831 }
832
833 struct ipv6_gro_cb {
834         struct napi_gro_cb napi;
835         int proto;
836 };
837
838 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
839
840 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
841                                          struct sk_buff *skb)
842 {
843         const struct inet6_protocol *ops;
844         struct sk_buff **pp = NULL;
845         struct sk_buff *p;
846         struct ipv6hdr *iph;
847         unsigned int nlen;
848         unsigned int hlen;
849         unsigned int off;
850         int flush = 1;
851         int proto;
852         __wsum csum;
853
854         off = skb_gro_offset(skb);
855         hlen = off + sizeof(*iph);
856         iph = skb_gro_header_fast(skb, off);
857         if (skb_gro_header_hard(skb, hlen)) {
858                 iph = skb_gro_header_slow(skb, hlen, off);
859                 if (unlikely(!iph))
860                         goto out;
861         }
862
863         skb_gro_pull(skb, sizeof(*iph));
864         skb_set_transport_header(skb, skb_gro_offset(skb));
865
866         flush += ntohs(iph->payload_len) != skb_gro_len(skb);
867
868         rcu_read_lock();
869         proto = iph->nexthdr;
870         ops = rcu_dereference(inet6_protos[proto]);
871         if (!ops || !ops->gro_receive) {
872                 __pskb_pull(skb, skb_gro_offset(skb));
873                 proto = ipv6_gso_pull_exthdrs(skb, proto);
874                 skb_gro_pull(skb, -skb_transport_offset(skb));
875                 skb_reset_transport_header(skb);
876                 __skb_push(skb, skb_gro_offset(skb));
877
878                 if (!ops || !ops->gro_receive)
879                         goto out_unlock;
880
881                 iph = ipv6_hdr(skb);
882         }
883
884         IPV6_GRO_CB(skb)->proto = proto;
885
886         flush--;
887         nlen = skb_network_header_len(skb);
888
889         for (p = *head; p; p = p->next) {
890                 struct ipv6hdr *iph2;
891
892                 if (!NAPI_GRO_CB(p)->same_flow)
893                         continue;
894
895                 iph2 = ipv6_hdr(p);
896
897                 /* All fields must match except length. */
898                 if (nlen != skb_network_header_len(p) ||
899                     memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
900                     memcmp(&iph->nexthdr, &iph2->nexthdr,
901                            nlen - offsetof(struct ipv6hdr, nexthdr))) {
902                         NAPI_GRO_CB(p)->same_flow = 0;
903                         continue;
904                 }
905
906                 NAPI_GRO_CB(p)->flush |= flush;
907         }
908
909         NAPI_GRO_CB(skb)->flush |= flush;
910
911         csum = skb->csum;
912         skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
913
914         pp = ops->gro_receive(head, skb);
915
916         skb->csum = csum;
917
918 out_unlock:
919         rcu_read_unlock();
920
921 out:
922         NAPI_GRO_CB(skb)->flush |= flush;
923
924         return pp;
925 }
926
927 static int ipv6_gro_complete(struct sk_buff *skb)
928 {
929         const struct inet6_protocol *ops;
930         struct ipv6hdr *iph = ipv6_hdr(skb);
931         int err = -ENOSYS;
932
933         iph->payload_len = htons(skb->len - skb_network_offset(skb) -
934                                  sizeof(*iph));
935
936         rcu_read_lock();
937         ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
938         if (WARN_ON(!ops || !ops->gro_complete))
939                 goto out_unlock;
940
941         err = ops->gro_complete(skb);
942
943 out_unlock:
944         rcu_read_unlock();
945
946         return err;
947 }
948
949 static struct packet_type ipv6_packet_type __read_mostly = {
950         .type = cpu_to_be16(ETH_P_IPV6),
951         .func = ipv6_rcv,
952         .gso_send_check = ipv6_gso_send_check,
953         .gso_segment = ipv6_gso_segment,
954         .gro_receive = ipv6_gro_receive,
955         .gro_complete = ipv6_gro_complete,
956 };
957
958 static int __init ipv6_packet_init(void)
959 {
960         dev_add_pack(&ipv6_packet_type);
961         return 0;
962 }
963
964 static void ipv6_packet_cleanup(void)
965 {
966         dev_remove_pack(&ipv6_packet_type);
967 }
968
969 static int __net_init ipv6_init_mibs(struct net *net)
970 {
971         if (snmp_mib_init((void **)net->mib.udp_stats_in6,
972                           sizeof (struct udp_mib)) < 0)
973                 return -ENOMEM;
974         if (snmp_mib_init((void **)net->mib.udplite_stats_in6,
975                           sizeof (struct udp_mib)) < 0)
976                 goto err_udplite_mib;
977         if (snmp_mib_init((void **)net->mib.ipv6_statistics,
978                           sizeof(struct ipstats_mib)) < 0)
979                 goto err_ip_mib;
980         if (snmp_mib_init((void **)net->mib.icmpv6_statistics,
981                           sizeof(struct icmpv6_mib)) < 0)
982                 goto err_icmp_mib;
983         if (snmp_mib_init((void **)net->mib.icmpv6msg_statistics,
984                           sizeof(struct icmpv6msg_mib)) < 0)
985                 goto err_icmpmsg_mib;
986         return 0;
987
988 err_icmpmsg_mib:
989         snmp_mib_free((void **)net->mib.icmpv6_statistics);
990 err_icmp_mib:
991         snmp_mib_free((void **)net->mib.ipv6_statistics);
992 err_ip_mib:
993         snmp_mib_free((void **)net->mib.udplite_stats_in6);
994 err_udplite_mib:
995         snmp_mib_free((void **)net->mib.udp_stats_in6);
996         return -ENOMEM;
997 }
998
999 static void __net_exit ipv6_cleanup_mibs(struct net *net)
1000 {
1001         snmp_mib_free((void **)net->mib.udp_stats_in6);
1002         snmp_mib_free((void **)net->mib.udplite_stats_in6);
1003         snmp_mib_free((void **)net->mib.ipv6_statistics);
1004         snmp_mib_free((void **)net->mib.icmpv6_statistics);
1005         snmp_mib_free((void **)net->mib.icmpv6msg_statistics);
1006 }
1007
1008 static int __net_init inet6_net_init(struct net *net)
1009 {
1010         int err = 0;
1011
1012         net->ipv6.sysctl.bindv6only = 0;
1013         net->ipv6.sysctl.icmpv6_time = 1*HZ;
1014
1015         err = ipv6_init_mibs(net);
1016         if (err)
1017                 return err;
1018 #ifdef CONFIG_PROC_FS
1019         err = udp6_proc_init(net);
1020         if (err)
1021                 goto out;
1022         err = tcp6_proc_init(net);
1023         if (err)
1024                 goto proc_tcp6_fail;
1025         err = ac6_proc_init(net);
1026         if (err)
1027                 goto proc_ac6_fail;
1028 #endif
1029         return err;
1030
1031 #ifdef CONFIG_PROC_FS
1032 proc_ac6_fail:
1033         tcp6_proc_exit(net);
1034 proc_tcp6_fail:
1035         udp6_proc_exit(net);
1036 out:
1037         ipv6_cleanup_mibs(net);
1038         return err;
1039 #endif
1040 }
1041
1042 static void inet6_net_exit(struct net *net)
1043 {
1044 #ifdef CONFIG_PROC_FS
1045         udp6_proc_exit(net);
1046         tcp6_proc_exit(net);
1047         ac6_proc_exit(net);
1048 #endif
1049         ipv6_cleanup_mibs(net);
1050 }
1051
1052 static struct pernet_operations inet6_net_ops = {
1053         .init = inet6_net_init,
1054         .exit = inet6_net_exit,
1055 };
1056
1057 static int __init inet6_init(void)
1058 {
1059         struct sk_buff *dummy_skb;
1060         struct list_head *r;
1061         int err = 0;
1062
1063         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1064
1065         /* Register the socket-side information for inet6_create.  */
1066         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1067                 INIT_LIST_HEAD(r);
1068
1069         if (disable_ipv6_mod) {
1070                 printk(KERN_INFO
1071                        "IPv6: Loaded, but administratively disabled, "
1072                        "reboot required to enable\n");
1073                 goto out;
1074         }
1075
1076         err = proto_register(&tcpv6_prot, 1);
1077         if (err)
1078                 goto out;
1079
1080         err = proto_register(&udpv6_prot, 1);
1081         if (err)
1082                 goto out_unregister_tcp_proto;
1083
1084         err = proto_register(&udplitev6_prot, 1);
1085         if (err)
1086                 goto out_unregister_udp_proto;
1087
1088         err = proto_register(&rawv6_prot, 1);
1089         if (err)
1090                 goto out_unregister_udplite_proto;
1091
1092
1093         /* We MUST register RAW sockets before we create the ICMP6,
1094          * IGMP6, or NDISC control sockets.
1095          */
1096         err = rawv6_init();
1097         if (err)
1098                 goto out_unregister_raw_proto;
1099
1100         /* Register the family here so that the init calls below will
1101          * be able to create sockets. (?? is this dangerous ??)
1102          */
1103         err = sock_register(&inet6_family_ops);
1104         if (err)
1105                 goto out_sock_register_fail;
1106
1107 #ifdef CONFIG_SYSCTL
1108         err = ipv6_static_sysctl_register();
1109         if (err)
1110                 goto static_sysctl_fail;
1111 #endif
1112         /*
1113          *      ipngwg API draft makes clear that the correct semantics
1114          *      for TCP and UDP is to consider one TCP and UDP instance
1115          *      in a host availiable by both INET and INET6 APIs and
1116          *      able to communicate via both network protocols.
1117          */
1118
1119         err = register_pernet_subsys(&inet6_net_ops);
1120         if (err)
1121                 goto register_pernet_fail;
1122         err = icmpv6_init();
1123         if (err)
1124                 goto icmp_fail;
1125         err = ip6_mr_init();
1126         if (err)
1127                 goto ipmr_fail;
1128         err = ndisc_init();
1129         if (err)
1130                 goto ndisc_fail;
1131         err = igmp6_init();
1132         if (err)
1133                 goto igmp_fail;
1134         err = ipv6_netfilter_init();
1135         if (err)
1136                 goto netfilter_fail;
1137         /* Create /proc/foo6 entries. */
1138 #ifdef CONFIG_PROC_FS
1139         err = -ENOMEM;
1140         if (raw6_proc_init())
1141                 goto proc_raw6_fail;
1142         if (udplite6_proc_init())
1143                 goto proc_udplite6_fail;
1144         if (ipv6_misc_proc_init())
1145                 goto proc_misc6_fail;
1146         if (if6_proc_init())
1147                 goto proc_if6_fail;
1148 #endif
1149         err = ip6_route_init();
1150         if (err)
1151                 goto ip6_route_fail;
1152         err = ip6_flowlabel_init();
1153         if (err)
1154                 goto ip6_flowlabel_fail;
1155         err = addrconf_init();
1156         if (err)
1157                 goto addrconf_fail;
1158
1159         /* Init v6 extension headers. */
1160         err = ipv6_exthdrs_init();
1161         if (err)
1162                 goto ipv6_exthdrs_fail;
1163
1164         err = ipv6_frag_init();
1165         if (err)
1166                 goto ipv6_frag_fail;
1167
1168         /* Init v6 transport protocols. */
1169         err = udpv6_init();
1170         if (err)
1171                 goto udpv6_fail;
1172
1173         err = udplitev6_init();
1174         if (err)
1175                 goto udplitev6_fail;
1176
1177         err = tcpv6_init();
1178         if (err)
1179                 goto tcpv6_fail;
1180
1181         err = ipv6_packet_init();
1182         if (err)
1183                 goto ipv6_packet_fail;
1184
1185 #ifdef CONFIG_SYSCTL
1186         err = ipv6_sysctl_register();
1187         if (err)
1188                 goto sysctl_fail;
1189 #endif
1190 out:
1191         return err;
1192
1193 #ifdef CONFIG_SYSCTL
1194 sysctl_fail:
1195         ipv6_packet_cleanup();
1196 #endif
1197 ipv6_packet_fail:
1198         tcpv6_exit();
1199 tcpv6_fail:
1200         udplitev6_exit();
1201 udplitev6_fail:
1202         udpv6_exit();
1203 udpv6_fail:
1204         ipv6_frag_exit();
1205 ipv6_frag_fail:
1206         ipv6_exthdrs_exit();
1207 ipv6_exthdrs_fail:
1208         addrconf_cleanup();
1209 addrconf_fail:
1210         ip6_flowlabel_cleanup();
1211 ip6_flowlabel_fail:
1212         ip6_route_cleanup();
1213 ip6_route_fail:
1214 #ifdef CONFIG_PROC_FS
1215         if6_proc_exit();
1216 proc_if6_fail:
1217         ipv6_misc_proc_exit();
1218 proc_misc6_fail:
1219         udplite6_proc_exit();
1220 proc_udplite6_fail:
1221         raw6_proc_exit();
1222 proc_raw6_fail:
1223 #endif
1224         ipv6_netfilter_fini();
1225 netfilter_fail:
1226         igmp6_cleanup();
1227 igmp_fail:
1228         ndisc_cleanup();
1229 ndisc_fail:
1230         ip6_mr_cleanup();
1231 ipmr_fail:
1232         icmpv6_cleanup();
1233 icmp_fail:
1234         unregister_pernet_subsys(&inet6_net_ops);
1235 register_pernet_fail:
1236 #ifdef CONFIG_SYSCTL
1237         ipv6_static_sysctl_unregister();
1238 static_sysctl_fail:
1239 #endif
1240         sock_unregister(PF_INET6);
1241         rtnl_unregister_all(PF_INET6);
1242 out_sock_register_fail:
1243         rawv6_exit();
1244 out_unregister_raw_proto:
1245         proto_unregister(&rawv6_prot);
1246 out_unregister_udplite_proto:
1247         proto_unregister(&udplitev6_prot);
1248 out_unregister_udp_proto:
1249         proto_unregister(&udpv6_prot);
1250 out_unregister_tcp_proto:
1251         proto_unregister(&tcpv6_prot);
1252         goto out;
1253 }
1254 module_init(inet6_init);
1255
1256 static void __exit inet6_exit(void)
1257 {
1258         if (disable_ipv6_mod)
1259                 return;
1260
1261         /* First of all disallow new sockets creation. */
1262         sock_unregister(PF_INET6);
1263         /* Disallow any further netlink messages */
1264         rtnl_unregister_all(PF_INET6);
1265
1266 #ifdef CONFIG_SYSCTL
1267         ipv6_sysctl_unregister();
1268 #endif
1269         udpv6_exit();
1270         udplitev6_exit();
1271         tcpv6_exit();
1272
1273         /* Cleanup code parts. */
1274         ipv6_packet_cleanup();
1275         ipv6_frag_exit();
1276         ipv6_exthdrs_exit();
1277         addrconf_cleanup();
1278         ip6_flowlabel_cleanup();
1279         ip6_route_cleanup();
1280 #ifdef CONFIG_PROC_FS
1281
1282         /* Cleanup code parts. */
1283         if6_proc_exit();
1284         ipv6_misc_proc_exit();
1285         udplite6_proc_exit();
1286         raw6_proc_exit();
1287 #endif
1288         ipv6_netfilter_fini();
1289         igmp6_cleanup();
1290         ndisc_cleanup();
1291         ip6_mr_cleanup();
1292         icmpv6_cleanup();
1293         rawv6_exit();
1294
1295         unregister_pernet_subsys(&inet6_net_ops);
1296 #ifdef CONFIG_SYSCTL
1297         ipv6_static_sysctl_unregister();
1298 #endif
1299         proto_unregister(&rawv6_prot);
1300         proto_unregister(&udplitev6_prot);
1301         proto_unregister(&udpv6_prot);
1302         proto_unregister(&tcpv6_prot);
1303
1304         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1305 }
1306 module_exit(inet6_exit);
1307
1308 MODULE_ALIAS_NETPROTO(PF_INET6);