net: drop capability from protocol definitions
[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 (sock->type == SOCK_RAW && !capable(CAP_NET_RAW))
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->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->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->inet_sport = htons(inet->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->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                         rcu_read_lock();
318                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
319                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
320                                     addr->sin6_scope_id) {
321                                         /* Override any existing binding, if another one
322                                          * is supplied by user.
323                                          */
324                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
325                                 }
326
327                                 /* Binding to link-local address requires an interface */
328                                 if (!sk->sk_bound_dev_if) {
329                                         err = -EINVAL;
330                                         goto out_unlock;
331                                 }
332                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
333                                 if (!dev) {
334                                         err = -ENODEV;
335                                         goto out_unlock;
336                                 }
337                         }
338
339                         /* ipv4 addr of the socket is invalid.  Only the
340                          * unspecified and mapped address have a v4 equivalent.
341                          */
342                         v4addr = LOOPBACK4_IPV6;
343                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
344                                 if (!ipv6_chk_addr(net, &addr->sin6_addr,
345                                                    dev, 0)) {
346                                         err = -EADDRNOTAVAIL;
347                                         goto out_unlock;
348                                 }
349                         }
350                         rcu_read_unlock();
351                 }
352         }
353
354         inet->inet_rcv_saddr = v4addr;
355         inet->inet_saddr = v4addr;
356
357         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
358
359         if (!(addr_type & IPV6_ADDR_MULTICAST))
360                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
361
362         /* Make sure we are allowed to bind here. */
363         if (sk->sk_prot->get_port(sk, snum)) {
364                 inet_reset_saddr(sk);
365                 err = -EADDRINUSE;
366                 goto out;
367         }
368
369         if (addr_type != IPV6_ADDR_ANY) {
370                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
371                 if (addr_type != IPV6_ADDR_MAPPED)
372                         np->ipv6only = 1;
373         }
374         if (snum)
375                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
376         inet->inet_sport = htons(inet->inet_num);
377         inet->inet_dport = 0;
378         inet->inet_daddr = 0;
379 out:
380         release_sock(sk);
381         return err;
382 out_unlock:
383         rcu_read_unlock();
384         goto out;
385 }
386
387 EXPORT_SYMBOL(inet6_bind);
388
389 int inet6_release(struct socket *sock)
390 {
391         struct sock *sk = sock->sk;
392
393         if (sk == NULL)
394                 return -EINVAL;
395
396         /* Free mc lists */
397         ipv6_sock_mc_close(sk);
398
399         /* Free ac lists */
400         ipv6_sock_ac_close(sk);
401
402         return inet_release(sock);
403 }
404
405 EXPORT_SYMBOL(inet6_release);
406
407 void inet6_destroy_sock(struct sock *sk)
408 {
409         struct ipv6_pinfo *np = inet6_sk(sk);
410         struct sk_buff *skb;
411         struct ipv6_txoptions *opt;
412
413         /* Release rx options */
414
415         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
416                 kfree_skb(skb);
417
418         /* Free flowlabels */
419         fl6_free_socklist(sk);
420
421         /* Free tx options */
422
423         if ((opt = xchg(&np->opt, NULL)) != NULL)
424                 sock_kfree_s(sk, opt, opt->tot_len);
425 }
426
427 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
428
429 /*
430  *      This does both peername and sockname.
431  */
432
433 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
434                  int *uaddr_len, int peer)
435 {
436         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
437         struct sock *sk = sock->sk;
438         struct inet_sock *inet = inet_sk(sk);
439         struct ipv6_pinfo *np = inet6_sk(sk);
440
441         sin->sin6_family = AF_INET6;
442         sin->sin6_flowinfo = 0;
443         sin->sin6_scope_id = 0;
444         if (peer) {
445                 if (!inet->inet_dport)
446                         return -ENOTCONN;
447                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
448                     peer == 1)
449                         return -ENOTCONN;
450                 sin->sin6_port = inet->inet_dport;
451                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
452                 if (np->sndflow)
453                         sin->sin6_flowinfo = np->flow_label;
454         } else {
455                 if (ipv6_addr_any(&np->rcv_saddr))
456                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
457                 else
458                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
459
460                 sin->sin6_port = inet->inet_sport;
461         }
462         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
463                 sin->sin6_scope_id = sk->sk_bound_dev_if;
464         *uaddr_len = sizeof(*sin);
465         return(0);
466 }
467
468 EXPORT_SYMBOL(inet6_getname);
469
470 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
471 {
472         struct sock *sk = sock->sk;
473         struct net *net = sock_net(sk);
474
475         switch(cmd)
476         {
477         case SIOCGSTAMP:
478                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
479
480         case SIOCGSTAMPNS:
481                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
482
483         case SIOCADDRT:
484         case SIOCDELRT:
485
486                 return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
487
488         case SIOCSIFADDR:
489                 return addrconf_add_ifaddr(net, (void __user *) arg);
490         case SIOCDIFADDR:
491                 return addrconf_del_ifaddr(net, (void __user *) arg);
492         case SIOCSIFDSTADDR:
493                 return addrconf_set_dstaddr(net, (void __user *) arg);
494         default:
495                 if (!sk->sk_prot->ioctl)
496                         return -ENOIOCTLCMD;
497                 return sk->sk_prot->ioctl(sk, cmd, arg);
498         }
499         /*NOTREACHED*/
500         return(0);
501 }
502
503 EXPORT_SYMBOL(inet6_ioctl);
504
505 const struct proto_ops inet6_stream_ops = {
506         .family            = PF_INET6,
507         .owner             = THIS_MODULE,
508         .release           = inet6_release,
509         .bind              = inet6_bind,
510         .connect           = inet_stream_connect,       /* ok           */
511         .socketpair        = sock_no_socketpair,        /* a do nothing */
512         .accept            = inet_accept,               /* ok           */
513         .getname           = inet6_getname,
514         .poll              = tcp_poll,                  /* ok           */
515         .ioctl             = inet6_ioctl,               /* must change  */
516         .listen            = inet_listen,               /* ok           */
517         .shutdown          = inet_shutdown,             /* ok           */
518         .setsockopt        = sock_common_setsockopt,    /* ok           */
519         .getsockopt        = sock_common_getsockopt,    /* ok           */
520         .sendmsg           = tcp_sendmsg,               /* ok           */
521         .recvmsg           = sock_common_recvmsg,       /* ok           */
522         .mmap              = sock_no_mmap,
523         .sendpage          = tcp_sendpage,
524         .splice_read       = tcp_splice_read,
525 #ifdef CONFIG_COMPAT
526         .compat_setsockopt = compat_sock_common_setsockopt,
527         .compat_getsockopt = compat_sock_common_getsockopt,
528 #endif
529 };
530
531 const struct proto_ops inet6_dgram_ops = {
532         .family            = PF_INET6,
533         .owner             = THIS_MODULE,
534         .release           = inet6_release,
535         .bind              = inet6_bind,
536         .connect           = inet_dgram_connect,        /* ok           */
537         .socketpair        = sock_no_socketpair,        /* a do nothing */
538         .accept            = sock_no_accept,            /* a do nothing */
539         .getname           = inet6_getname,
540         .poll              = udp_poll,                  /* ok           */
541         .ioctl             = inet6_ioctl,               /* must change  */
542         .listen            = sock_no_listen,            /* ok           */
543         .shutdown          = inet_shutdown,             /* ok           */
544         .setsockopt        = sock_common_setsockopt,    /* ok           */
545         .getsockopt        = sock_common_getsockopt,    /* ok           */
546         .sendmsg           = inet_sendmsg,              /* ok           */
547         .recvmsg           = sock_common_recvmsg,       /* ok           */
548         .mmap              = sock_no_mmap,
549         .sendpage          = sock_no_sendpage,
550 #ifdef CONFIG_COMPAT
551         .compat_setsockopt = compat_sock_common_setsockopt,
552         .compat_getsockopt = compat_sock_common_getsockopt,
553 #endif
554 };
555
556 static const struct net_proto_family inet6_family_ops = {
557         .family = PF_INET6,
558         .create = inet6_create,
559         .owner  = THIS_MODULE,
560 };
561
562 int inet6_register_protosw(struct inet_protosw *p)
563 {
564         struct list_head *lh;
565         struct inet_protosw *answer;
566         struct list_head *last_perm;
567         int protocol = p->protocol;
568         int ret;
569
570         spin_lock_bh(&inetsw6_lock);
571
572         ret = -EINVAL;
573         if (p->type >= SOCK_MAX)
574                 goto out_illegal;
575
576         /* If we are trying to override a permanent protocol, bail. */
577         answer = NULL;
578         ret = -EPERM;
579         last_perm = &inetsw6[p->type];
580         list_for_each(lh, &inetsw6[p->type]) {
581                 answer = list_entry(lh, struct inet_protosw, list);
582
583                 /* Check only the non-wild match. */
584                 if (INET_PROTOSW_PERMANENT & answer->flags) {
585                         if (protocol == answer->protocol)
586                                 break;
587                         last_perm = lh;
588                 }
589
590                 answer = NULL;
591         }
592         if (answer)
593                 goto out_permanent;
594
595         /* Add the new entry after the last permanent entry if any, so that
596          * the new entry does not override a permanent entry when matched with
597          * a wild-card protocol. But it is allowed to override any existing
598          * non-permanent entry.  This means that when we remove this entry, the
599          * system automatically returns to the old behavior.
600          */
601         list_add_rcu(&p->list, last_perm);
602         ret = 0;
603 out:
604         spin_unlock_bh(&inetsw6_lock);
605         return ret;
606
607 out_permanent:
608         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
609                protocol);
610         goto out;
611
612 out_illegal:
613         printk(KERN_ERR
614                "Ignoring attempt to register invalid socket type %d.\n",
615                p->type);
616         goto out;
617 }
618
619 EXPORT_SYMBOL(inet6_register_protosw);
620
621 void
622 inet6_unregister_protosw(struct inet_protosw *p)
623 {
624         if (INET_PROTOSW_PERMANENT & p->flags) {
625                 printk(KERN_ERR
626                        "Attempt to unregister permanent protocol %d.\n",
627                        p->protocol);
628         } else {
629                 spin_lock_bh(&inetsw6_lock);
630                 list_del_rcu(&p->list);
631                 spin_unlock_bh(&inetsw6_lock);
632
633                 synchronize_net();
634         }
635 }
636
637 EXPORT_SYMBOL(inet6_unregister_protosw);
638
639 int inet6_sk_rebuild_header(struct sock *sk)
640 {
641         int err;
642         struct dst_entry *dst;
643         struct ipv6_pinfo *np = inet6_sk(sk);
644
645         dst = __sk_dst_check(sk, np->dst_cookie);
646
647         if (dst == NULL) {
648                 struct inet_sock *inet = inet_sk(sk);
649                 struct in6_addr *final_p = NULL, final;
650                 struct flowi fl;
651
652                 memset(&fl, 0, sizeof(fl));
653                 fl.proto = sk->sk_protocol;
654                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
655                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
656                 fl.fl6_flowlabel = np->flow_label;
657                 fl.oif = sk->sk_bound_dev_if;
658                 fl.mark = sk->sk_mark;
659                 fl.fl_ip_dport = inet->inet_dport;
660                 fl.fl_ip_sport = inet->inet_sport;
661                 security_sk_classify_flow(sk, &fl);
662
663                 if (np->opt && np->opt->srcrt) {
664                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
665                         ipv6_addr_copy(&final, &fl.fl6_dst);
666                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
667                         final_p = &final;
668                 }
669
670                 err = ip6_dst_lookup(sk, &dst, &fl);
671                 if (err) {
672                         sk->sk_route_caps = 0;
673                         return err;
674                 }
675                 if (final_p)
676                         ipv6_addr_copy(&fl.fl6_dst, final_p);
677
678                 if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
679                         sk->sk_err_soft = -err;
680                         return err;
681                 }
682
683                 __ip6_dst_store(sk, dst, NULL, NULL);
684         }
685
686         return 0;
687 }
688
689 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
690
691 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
692 {
693         struct ipv6_pinfo *np = inet6_sk(sk);
694         struct inet6_skb_parm *opt = IP6CB(skb);
695
696         if (np->rxopt.all) {
697                 if ((opt->hop && (np->rxopt.bits.hopopts ||
698                                   np->rxopt.bits.ohopopts)) ||
699                     ((IPV6_FLOWINFO_MASK &
700                       *(__be32 *)skb_network_header(skb)) &&
701                      np->rxopt.bits.rxflow) ||
702                     (opt->srcrt && (np->rxopt.bits.srcrt ||
703                      np->rxopt.bits.osrcrt)) ||
704                     ((opt->dst1 || opt->dst0) &&
705                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
706                         return 1;
707         }
708         return 0;
709 }
710
711 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
712
713 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
714 {
715         const struct inet6_protocol *ops = NULL;
716
717         for (;;) {
718                 struct ipv6_opt_hdr *opth;
719                 int len;
720
721                 if (proto != NEXTHDR_HOP) {
722                         ops = rcu_dereference(inet6_protos[proto]);
723
724                         if (unlikely(!ops))
725                                 break;
726
727                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
728                                 break;
729                 }
730
731                 if (unlikely(!pskb_may_pull(skb, 8)))
732                         break;
733
734                 opth = (void *)skb->data;
735                 len = ipv6_optlen(opth);
736
737                 if (unlikely(!pskb_may_pull(skb, len)))
738                         break;
739
740                 proto = opth->nexthdr;
741                 __skb_pull(skb, len);
742         }
743
744         return proto;
745 }
746
747 static int ipv6_gso_send_check(struct sk_buff *skb)
748 {
749         struct ipv6hdr *ipv6h;
750         const struct inet6_protocol *ops;
751         int err = -EINVAL;
752
753         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
754                 goto out;
755
756         ipv6h = ipv6_hdr(skb);
757         __skb_pull(skb, sizeof(*ipv6h));
758         err = -EPROTONOSUPPORT;
759
760         rcu_read_lock();
761         ops = rcu_dereference(inet6_protos[
762                 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
763
764         if (likely(ops && ops->gso_send_check)) {
765                 skb_reset_transport_header(skb);
766                 err = ops->gso_send_check(skb);
767         }
768         rcu_read_unlock();
769
770 out:
771         return err;
772 }
773
774 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
775 {
776         struct sk_buff *segs = ERR_PTR(-EINVAL);
777         struct ipv6hdr *ipv6h;
778         const struct inet6_protocol *ops;
779         int proto;
780         struct frag_hdr *fptr;
781         unsigned int unfrag_ip6hlen;
782         u8 *prevhdr;
783         int offset = 0;
784
785         if (!(features & NETIF_F_V6_CSUM))
786                 features &= ~NETIF_F_SG;
787
788         if (unlikely(skb_shinfo(skb)->gso_type &
789                      ~(SKB_GSO_UDP |
790                        SKB_GSO_DODGY |
791                        SKB_GSO_TCP_ECN |
792                        SKB_GSO_TCPV6 |
793                        0)))
794                 goto out;
795
796         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
797                 goto out;
798
799         ipv6h = ipv6_hdr(skb);
800         __skb_pull(skb, sizeof(*ipv6h));
801         segs = ERR_PTR(-EPROTONOSUPPORT);
802
803         proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
804         rcu_read_lock();
805         ops = rcu_dereference(inet6_protos[proto]);
806         if (likely(ops && ops->gso_segment)) {
807                 skb_reset_transport_header(skb);
808                 segs = ops->gso_segment(skb, features);
809         }
810         rcu_read_unlock();
811
812         if (unlikely(IS_ERR(segs)))
813                 goto out;
814
815         for (skb = segs; skb; skb = skb->next) {
816                 ipv6h = ipv6_hdr(skb);
817                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
818                                            sizeof(*ipv6h));
819                 if (proto == IPPROTO_UDP) {
820                         unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
821                         fptr = (struct frag_hdr *)(skb_network_header(skb) +
822                                 unfrag_ip6hlen);
823                         fptr->frag_off = htons(offset);
824                         if (skb->next != NULL)
825                                 fptr->frag_off |= htons(IP6_MF);
826                         offset += (ntohs(ipv6h->payload_len) -
827                                    sizeof(struct frag_hdr));
828                 }
829         }
830
831 out:
832         return segs;
833 }
834
835 struct ipv6_gro_cb {
836         struct napi_gro_cb napi;
837         int proto;
838 };
839
840 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
841
842 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
843                                          struct sk_buff *skb)
844 {
845         const struct inet6_protocol *ops;
846         struct sk_buff **pp = NULL;
847         struct sk_buff *p;
848         struct ipv6hdr *iph;
849         unsigned int nlen;
850         unsigned int hlen;
851         unsigned int off;
852         int flush = 1;
853         int proto;
854         __wsum csum;
855
856         off = skb_gro_offset(skb);
857         hlen = off + sizeof(*iph);
858         iph = skb_gro_header_fast(skb, off);
859         if (skb_gro_header_hard(skb, hlen)) {
860                 iph = skb_gro_header_slow(skb, hlen, off);
861                 if (unlikely(!iph))
862                         goto out;
863         }
864
865         skb_gro_pull(skb, sizeof(*iph));
866         skb_set_transport_header(skb, skb_gro_offset(skb));
867
868         flush += ntohs(iph->payload_len) != skb_gro_len(skb);
869
870         rcu_read_lock();
871         proto = iph->nexthdr;
872         ops = rcu_dereference(inet6_protos[proto]);
873         if (!ops || !ops->gro_receive) {
874                 __pskb_pull(skb, skb_gro_offset(skb));
875                 proto = ipv6_gso_pull_exthdrs(skb, proto);
876                 skb_gro_pull(skb, -skb_transport_offset(skb));
877                 skb_reset_transport_header(skb);
878                 __skb_push(skb, skb_gro_offset(skb));
879
880                 if (!ops || !ops->gro_receive)
881                         goto out_unlock;
882
883                 iph = ipv6_hdr(skb);
884         }
885
886         IPV6_GRO_CB(skb)->proto = proto;
887
888         flush--;
889         nlen = skb_network_header_len(skb);
890
891         for (p = *head; p; p = p->next) {
892                 struct ipv6hdr *iph2;
893
894                 if (!NAPI_GRO_CB(p)->same_flow)
895                         continue;
896
897                 iph2 = ipv6_hdr(p);
898
899                 /* All fields must match except length. */
900                 if (nlen != skb_network_header_len(p) ||
901                     memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
902                     memcmp(&iph->nexthdr, &iph2->nexthdr,
903                            nlen - offsetof(struct ipv6hdr, nexthdr))) {
904                         NAPI_GRO_CB(p)->same_flow = 0;
905                         continue;
906                 }
907
908                 NAPI_GRO_CB(p)->flush |= flush;
909         }
910
911         NAPI_GRO_CB(skb)->flush |= flush;
912
913         csum = skb->csum;
914         skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
915
916         pp = ops->gro_receive(head, skb);
917
918         skb->csum = csum;
919
920 out_unlock:
921         rcu_read_unlock();
922
923 out:
924         NAPI_GRO_CB(skb)->flush |= flush;
925
926         return pp;
927 }
928
929 static int ipv6_gro_complete(struct sk_buff *skb)
930 {
931         const struct inet6_protocol *ops;
932         struct ipv6hdr *iph = ipv6_hdr(skb);
933         int err = -ENOSYS;
934
935         iph->payload_len = htons(skb->len - skb_network_offset(skb) -
936                                  sizeof(*iph));
937
938         rcu_read_lock();
939         ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
940         if (WARN_ON(!ops || !ops->gro_complete))
941                 goto out_unlock;
942
943         err = ops->gro_complete(skb);
944
945 out_unlock:
946         rcu_read_unlock();
947
948         return err;
949 }
950
951 static struct packet_type ipv6_packet_type __read_mostly = {
952         .type = cpu_to_be16(ETH_P_IPV6),
953         .func = ipv6_rcv,
954         .gso_send_check = ipv6_gso_send_check,
955         .gso_segment = ipv6_gso_segment,
956         .gro_receive = ipv6_gro_receive,
957         .gro_complete = ipv6_gro_complete,
958 };
959
960 static int __init ipv6_packet_init(void)
961 {
962         dev_add_pack(&ipv6_packet_type);
963         return 0;
964 }
965
966 static void ipv6_packet_cleanup(void)
967 {
968         dev_remove_pack(&ipv6_packet_type);
969 }
970
971 static int __net_init ipv6_init_mibs(struct net *net)
972 {
973         if (snmp_mib_init((void **)net->mib.udp_stats_in6,
974                           sizeof (struct udp_mib)) < 0)
975                 return -ENOMEM;
976         if (snmp_mib_init((void **)net->mib.udplite_stats_in6,
977                           sizeof (struct udp_mib)) < 0)
978                 goto err_udplite_mib;
979         if (snmp_mib_init((void **)net->mib.ipv6_statistics,
980                           sizeof(struct ipstats_mib)) < 0)
981                 goto err_ip_mib;
982         if (snmp_mib_init((void **)net->mib.icmpv6_statistics,
983                           sizeof(struct icmpv6_mib)) < 0)
984                 goto err_icmp_mib;
985         if (snmp_mib_init((void **)net->mib.icmpv6msg_statistics,
986                           sizeof(struct icmpv6msg_mib)) < 0)
987                 goto err_icmpmsg_mib;
988         return 0;
989
990 err_icmpmsg_mib:
991         snmp_mib_free((void **)net->mib.icmpv6_statistics);
992 err_icmp_mib:
993         snmp_mib_free((void **)net->mib.ipv6_statistics);
994 err_ip_mib:
995         snmp_mib_free((void **)net->mib.udplite_stats_in6);
996 err_udplite_mib:
997         snmp_mib_free((void **)net->mib.udp_stats_in6);
998         return -ENOMEM;
999 }
1000
1001 static void __net_exit ipv6_cleanup_mibs(struct net *net)
1002 {
1003         snmp_mib_free((void **)net->mib.udp_stats_in6);
1004         snmp_mib_free((void **)net->mib.udplite_stats_in6);
1005         snmp_mib_free((void **)net->mib.ipv6_statistics);
1006         snmp_mib_free((void **)net->mib.icmpv6_statistics);
1007         snmp_mib_free((void **)net->mib.icmpv6msg_statistics);
1008 }
1009
1010 static int __net_init inet6_net_init(struct net *net)
1011 {
1012         int err = 0;
1013
1014         net->ipv6.sysctl.bindv6only = 0;
1015         net->ipv6.sysctl.icmpv6_time = 1*HZ;
1016
1017         err = ipv6_init_mibs(net);
1018         if (err)
1019                 return err;
1020 #ifdef CONFIG_PROC_FS
1021         err = udp6_proc_init(net);
1022         if (err)
1023                 goto out;
1024         err = tcp6_proc_init(net);
1025         if (err)
1026                 goto proc_tcp6_fail;
1027         err = ac6_proc_init(net);
1028         if (err)
1029                 goto proc_ac6_fail;
1030 #endif
1031         return err;
1032
1033 #ifdef CONFIG_PROC_FS
1034 proc_ac6_fail:
1035         tcp6_proc_exit(net);
1036 proc_tcp6_fail:
1037         udp6_proc_exit(net);
1038 out:
1039         ipv6_cleanup_mibs(net);
1040         return err;
1041 #endif
1042 }
1043
1044 static void inet6_net_exit(struct net *net)
1045 {
1046 #ifdef CONFIG_PROC_FS
1047         udp6_proc_exit(net);
1048         tcp6_proc_exit(net);
1049         ac6_proc_exit(net);
1050 #endif
1051         ipv6_cleanup_mibs(net);
1052 }
1053
1054 static struct pernet_operations inet6_net_ops = {
1055         .init = inet6_net_init,
1056         .exit = inet6_net_exit,
1057 };
1058
1059 static int __init inet6_init(void)
1060 {
1061         struct sk_buff *dummy_skb;
1062         struct list_head *r;
1063         int err = 0;
1064
1065         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1066
1067         /* Register the socket-side information for inet6_create.  */
1068         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1069                 INIT_LIST_HEAD(r);
1070
1071         if (disable_ipv6_mod) {
1072                 printk(KERN_INFO
1073                        "IPv6: Loaded, but administratively disabled, "
1074                        "reboot required to enable\n");
1075                 goto out;
1076         }
1077
1078         err = proto_register(&tcpv6_prot, 1);
1079         if (err)
1080                 goto out;
1081
1082         err = proto_register(&udpv6_prot, 1);
1083         if (err)
1084                 goto out_unregister_tcp_proto;
1085
1086         err = proto_register(&udplitev6_prot, 1);
1087         if (err)
1088                 goto out_unregister_udp_proto;
1089
1090         err = proto_register(&rawv6_prot, 1);
1091         if (err)
1092                 goto out_unregister_udplite_proto;
1093
1094
1095         /* We MUST register RAW sockets before we create the ICMP6,
1096          * IGMP6, or NDISC control sockets.
1097          */
1098         err = rawv6_init();
1099         if (err)
1100                 goto out_unregister_raw_proto;
1101
1102         /* Register the family here so that the init calls below will
1103          * be able to create sockets. (?? is this dangerous ??)
1104          */
1105         err = sock_register(&inet6_family_ops);
1106         if (err)
1107                 goto out_sock_register_fail;
1108
1109 #ifdef CONFIG_SYSCTL
1110         err = ipv6_static_sysctl_register();
1111         if (err)
1112                 goto static_sysctl_fail;
1113 #endif
1114         /*
1115          *      ipngwg API draft makes clear that the correct semantics
1116          *      for TCP and UDP is to consider one TCP and UDP instance
1117          *      in a host availiable by both INET and INET6 APIs and
1118          *      able to communicate via both network protocols.
1119          */
1120
1121         err = register_pernet_subsys(&inet6_net_ops);
1122         if (err)
1123                 goto register_pernet_fail;
1124         err = icmpv6_init();
1125         if (err)
1126                 goto icmp_fail;
1127         err = ip6_mr_init();
1128         if (err)
1129                 goto ipmr_fail;
1130         err = ndisc_init();
1131         if (err)
1132                 goto ndisc_fail;
1133         err = igmp6_init();
1134         if (err)
1135                 goto igmp_fail;
1136         err = ipv6_netfilter_init();
1137         if (err)
1138                 goto netfilter_fail;
1139         /* Create /proc/foo6 entries. */
1140 #ifdef CONFIG_PROC_FS
1141         err = -ENOMEM;
1142         if (raw6_proc_init())
1143                 goto proc_raw6_fail;
1144         if (udplite6_proc_init())
1145                 goto proc_udplite6_fail;
1146         if (ipv6_misc_proc_init())
1147                 goto proc_misc6_fail;
1148         if (if6_proc_init())
1149                 goto proc_if6_fail;
1150 #endif
1151         err = ip6_route_init();
1152         if (err)
1153                 goto ip6_route_fail;
1154         err = ip6_flowlabel_init();
1155         if (err)
1156                 goto ip6_flowlabel_fail;
1157         err = addrconf_init();
1158         if (err)
1159                 goto addrconf_fail;
1160
1161         /* Init v6 extension headers. */
1162         err = ipv6_exthdrs_init();
1163         if (err)
1164                 goto ipv6_exthdrs_fail;
1165
1166         err = ipv6_frag_init();
1167         if (err)
1168                 goto ipv6_frag_fail;
1169
1170         /* Init v6 transport protocols. */
1171         err = udpv6_init();
1172         if (err)
1173                 goto udpv6_fail;
1174
1175         err = udplitev6_init();
1176         if (err)
1177                 goto udplitev6_fail;
1178
1179         err = tcpv6_init();
1180         if (err)
1181                 goto tcpv6_fail;
1182
1183         err = ipv6_packet_init();
1184         if (err)
1185                 goto ipv6_packet_fail;
1186
1187 #ifdef CONFIG_SYSCTL
1188         err = ipv6_sysctl_register();
1189         if (err)
1190                 goto sysctl_fail;
1191 #endif
1192 out:
1193         return err;
1194
1195 #ifdef CONFIG_SYSCTL
1196 sysctl_fail:
1197         ipv6_packet_cleanup();
1198 #endif
1199 ipv6_packet_fail:
1200         tcpv6_exit();
1201 tcpv6_fail:
1202         udplitev6_exit();
1203 udplitev6_fail:
1204         udpv6_exit();
1205 udpv6_fail:
1206         ipv6_frag_exit();
1207 ipv6_frag_fail:
1208         ipv6_exthdrs_exit();
1209 ipv6_exthdrs_fail:
1210         addrconf_cleanup();
1211 addrconf_fail:
1212         ip6_flowlabel_cleanup();
1213 ip6_flowlabel_fail:
1214         ip6_route_cleanup();
1215 ip6_route_fail:
1216 #ifdef CONFIG_PROC_FS
1217         if6_proc_exit();
1218 proc_if6_fail:
1219         ipv6_misc_proc_exit();
1220 proc_misc6_fail:
1221         udplite6_proc_exit();
1222 proc_udplite6_fail:
1223         raw6_proc_exit();
1224 proc_raw6_fail:
1225 #endif
1226         ipv6_netfilter_fini();
1227 netfilter_fail:
1228         igmp6_cleanup();
1229 igmp_fail:
1230         ndisc_cleanup();
1231 ndisc_fail:
1232         ip6_mr_cleanup();
1233 ipmr_fail:
1234         icmpv6_cleanup();
1235 icmp_fail:
1236         unregister_pernet_subsys(&inet6_net_ops);
1237 register_pernet_fail:
1238 #ifdef CONFIG_SYSCTL
1239         ipv6_static_sysctl_unregister();
1240 static_sysctl_fail:
1241 #endif
1242         sock_unregister(PF_INET6);
1243         rtnl_unregister_all(PF_INET6);
1244 out_sock_register_fail:
1245         rawv6_exit();
1246 out_unregister_raw_proto:
1247         proto_unregister(&rawv6_prot);
1248 out_unregister_udplite_proto:
1249         proto_unregister(&udplitev6_prot);
1250 out_unregister_udp_proto:
1251         proto_unregister(&udpv6_prot);
1252 out_unregister_tcp_proto:
1253         proto_unregister(&tcpv6_prot);
1254         goto out;
1255 }
1256 module_init(inet6_init);
1257
1258 static void __exit inet6_exit(void)
1259 {
1260         if (disable_ipv6_mod)
1261                 return;
1262
1263         /* First of all disallow new sockets creation. */
1264         sock_unregister(PF_INET6);
1265         /* Disallow any further netlink messages */
1266         rtnl_unregister_all(PF_INET6);
1267
1268 #ifdef CONFIG_SYSCTL
1269         ipv6_sysctl_unregister();
1270 #endif
1271         udpv6_exit();
1272         udplitev6_exit();
1273         tcpv6_exit();
1274
1275         /* Cleanup code parts. */
1276         ipv6_packet_cleanup();
1277         ipv6_frag_exit();
1278         ipv6_exthdrs_exit();
1279         addrconf_cleanup();
1280         ip6_flowlabel_cleanup();
1281         ip6_route_cleanup();
1282 #ifdef CONFIG_PROC_FS
1283
1284         /* Cleanup code parts. */
1285         if6_proc_exit();
1286         ipv6_misc_proc_exit();
1287         udplite6_proc_exit();
1288         raw6_proc_exit();
1289 #endif
1290         ipv6_netfilter_fini();
1291         igmp6_cleanup();
1292         ndisc_cleanup();
1293         ip6_mr_cleanup();
1294         icmpv6_cleanup();
1295         rawv6_exit();
1296
1297         unregister_pernet_subsys(&inet6_net_ops);
1298 #ifdef CONFIG_SYSCTL
1299         ipv6_static_sysctl_unregister();
1300 #endif
1301         proto_unregister(&rawv6_prot);
1302         proto_unregister(&udplitev6_prot);
1303         proto_unregister(&udpv6_prot);
1304         proto_unregister(&tcpv6_prot);
1305
1306         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1307 }
1308 module_exit(inet6_exit);
1309
1310 MODULE_ALIAS_NETPROTO(PF_INET6);