audit: complex interfield comparison helper
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / af_inet6.c
blob273f48d1df2e974b5722e9133c3dd04ffd39473e
1 /*
2 * PF_INET6 socket protocol family
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Adapted from linux/net/ipv4/af_inet.c
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
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.
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 #include <linux/slab.h>
41 #include <linux/inet.h>
42 #include <linux/netdevice.h>
43 #include <linux/icmpv6.h>
44 #include <linux/netfilter_ipv6.h>
46 #include <net/ip.h>
47 #include <net/ipv6.h>
48 #include <net/udp.h>
49 #include <net/udplite.h>
50 #include <net/tcp.h>
51 #include <net/ipip.h>
52 #include <net/protocol.h>
53 #include <net/inet_common.h>
54 #include <net/route.h>
55 #include <net/transp_v6.h>
56 #include <net/ip6_route.h>
57 #include <net/addrconf.h>
58 #ifdef CONFIG_IPV6_TUNNEL
59 #include <net/ip6_tunnel.h>
60 #endif
62 #include <asm/uaccess.h>
63 #include <asm/system.h>
64 #include <linux/mroute6.h>
66 MODULE_AUTHOR("Cast of dozens");
67 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
68 MODULE_LICENSE("GPL");
70 /* The inetsw6 table contains everything that inet6_create needs to
71 * build a new socket.
73 static struct list_head inetsw6[SOCK_MAX];
74 static DEFINE_SPINLOCK(inetsw6_lock);
76 struct ipv6_params ipv6_defaults = {
77 .disable_ipv6 = 0,
78 .autoconf = 1,
81 static int disable_ipv6_mod = 0;
83 module_param_named(disable, disable_ipv6_mod, int, 0444);
84 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
86 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
87 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
89 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
90 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
92 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
94 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
96 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
99 static int inet6_create(struct net *net, struct socket *sock, int protocol,
100 int kern)
102 struct inet_sock *inet;
103 struct ipv6_pinfo *np;
104 struct sock *sk;
105 struct inet_protosw *answer;
106 struct proto *answer_prot;
107 unsigned char answer_flags;
108 char answer_no_check;
109 int try_loading_module = 0;
110 int err;
112 if (sock->type != SOCK_RAW &&
113 sock->type != SOCK_DGRAM &&
114 !inet_ehash_secret)
115 build_ehash_secret();
117 /* Look for the requested type/protocol pair. */
118 lookup_protocol:
119 err = -ESOCKTNOSUPPORT;
120 rcu_read_lock();
121 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
123 err = 0;
124 /* Check the non-wild match. */
125 if (protocol == answer->protocol) {
126 if (protocol != IPPROTO_IP)
127 break;
128 } else {
129 /* Check for the two wild cases. */
130 if (IPPROTO_IP == protocol) {
131 protocol = answer->protocol;
132 break;
134 if (IPPROTO_IP == answer->protocol)
135 break;
137 err = -EPROTONOSUPPORT;
140 if (err) {
141 if (try_loading_module < 2) {
142 rcu_read_unlock();
144 * Be more specific, e.g. net-pf-10-proto-132-type-1
145 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
147 if (++try_loading_module == 1)
148 request_module("net-pf-%d-proto-%d-type-%d",
149 PF_INET6, protocol, sock->type);
151 * Fall back to generic, e.g. net-pf-10-proto-132
152 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
154 else
155 request_module("net-pf-%d-proto-%d",
156 PF_INET6, protocol);
157 goto lookup_protocol;
158 } else
159 goto out_rcu_unlock;
162 err = -EPERM;
163 if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
164 goto out_rcu_unlock;
166 sock->ops = answer->ops;
167 answer_prot = answer->prot;
168 answer_no_check = answer->no_check;
169 answer_flags = answer->flags;
170 rcu_read_unlock();
172 WARN_ON(answer_prot->slab == NULL);
174 err = -ENOBUFS;
175 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
176 if (sk == NULL)
177 goto out;
179 sock_init_data(sock, sk);
181 err = 0;
182 sk->sk_no_check = answer_no_check;
183 if (INET_PROTOSW_REUSE & answer_flags)
184 sk->sk_reuse = 1;
186 inet = inet_sk(sk);
187 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
189 if (SOCK_RAW == sock->type) {
190 inet->inet_num = protocol;
191 if (IPPROTO_RAW == protocol)
192 inet->hdrincl = 1;
195 sk->sk_destruct = inet_sock_destruct;
196 sk->sk_family = PF_INET6;
197 sk->sk_protocol = protocol;
199 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
201 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
202 np->hop_limit = -1;
203 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
204 np->mc_loop = 1;
205 np->pmtudisc = IPV6_PMTUDISC_WANT;
206 np->ipv6only = net->ipv6.sysctl.bindv6only;
208 /* Init the ipv4 part of the socket since we can have sockets
209 * using v6 API for ipv4.
211 inet->uc_ttl = -1;
213 inet->mc_loop = 1;
214 inet->mc_ttl = 1;
215 inet->mc_index = 0;
216 inet->mc_list = NULL;
218 if (ipv4_config.no_pmtu_disc)
219 inet->pmtudisc = IP_PMTUDISC_DONT;
220 else
221 inet->pmtudisc = IP_PMTUDISC_WANT;
223 * Increment only the relevant sk_prot->socks debug field, this changes
224 * the previous behaviour of incrementing both the equivalent to
225 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
227 * This allows better debug granularity as we'll know exactly how many
228 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
229 * transport protocol socks. -acme
231 sk_refcnt_debug_inc(sk);
233 if (inet->inet_num) {
234 /* It assumes that any protocol which allows
235 * the user to assign a number at socket
236 * creation time automatically shares.
238 inet->inet_sport = htons(inet->inet_num);
239 sk->sk_prot->hash(sk);
241 if (sk->sk_prot->init) {
242 err = sk->sk_prot->init(sk);
243 if (err) {
244 sk_common_release(sk);
245 goto out;
248 out:
249 return err;
250 out_rcu_unlock:
251 rcu_read_unlock();
252 goto out;
256 /* bind for INET6 API */
257 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
259 struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
260 struct sock *sk = sock->sk;
261 struct inet_sock *inet = inet_sk(sk);
262 struct ipv6_pinfo *np = inet6_sk(sk);
263 struct net *net = sock_net(sk);
264 __be32 v4addr = 0;
265 unsigned short snum;
266 int addr_type = 0;
267 int err = 0;
269 /* If the socket has its own bind function then use it. */
270 if (sk->sk_prot->bind)
271 return sk->sk_prot->bind(sk, uaddr, addr_len);
273 if (addr_len < SIN6_LEN_RFC2133)
274 return -EINVAL;
276 if (addr->sin6_family != AF_INET6)
277 return -EAFNOSUPPORT;
279 addr_type = ipv6_addr_type(&addr->sin6_addr);
280 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
281 return -EINVAL;
283 snum = ntohs(addr->sin6_port);
284 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
285 return -EACCES;
287 lock_sock(sk);
289 /* Check these errors (active socket, double bind). */
290 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
291 err = -EINVAL;
292 goto out;
295 /* Check if the address belongs to the host. */
296 if (addr_type == IPV6_ADDR_MAPPED) {
297 int chk_addr_ret;
299 /* Binding to v4-mapped address on a v6-only socket
300 * makes no sense
302 if (np->ipv6only) {
303 err = -EINVAL;
304 goto out;
307 /* Reproduce AF_INET checks to make the bindings consistent */
308 v4addr = addr->sin6_addr.s6_addr32[3];
309 chk_addr_ret = inet_addr_type(net, v4addr);
310 if (!sysctl_ip_nonlocal_bind &&
311 !(inet->freebind || inet->transparent) &&
312 v4addr != htonl(INADDR_ANY) &&
313 chk_addr_ret != RTN_LOCAL &&
314 chk_addr_ret != RTN_MULTICAST &&
315 chk_addr_ret != RTN_BROADCAST) {
316 err = -EADDRNOTAVAIL;
317 goto out;
319 } else {
320 if (addr_type != IPV6_ADDR_ANY) {
321 struct net_device *dev = NULL;
323 rcu_read_lock();
324 if (addr_type & IPV6_ADDR_LINKLOCAL) {
325 if (addr_len >= sizeof(struct sockaddr_in6) &&
326 addr->sin6_scope_id) {
327 /* Override any existing binding, if another one
328 * is supplied by user.
330 sk->sk_bound_dev_if = addr->sin6_scope_id;
333 /* Binding to link-local address requires an interface */
334 if (!sk->sk_bound_dev_if) {
335 err = -EINVAL;
336 goto out_unlock;
338 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
339 if (!dev) {
340 err = -ENODEV;
341 goto out_unlock;
345 /* ipv4 addr of the socket is invalid. Only the
346 * unspecified and mapped address have a v4 equivalent.
348 v4addr = LOOPBACK4_IPV6;
349 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
350 if (!(inet->freebind || inet->transparent) &&
351 !ipv6_chk_addr(net, &addr->sin6_addr,
352 dev, 0)) {
353 err = -EADDRNOTAVAIL;
354 goto out_unlock;
357 rcu_read_unlock();
361 inet->inet_rcv_saddr = v4addr;
362 inet->inet_saddr = v4addr;
364 np->rcv_saddr = addr->sin6_addr;
366 if (!(addr_type & IPV6_ADDR_MULTICAST))
367 np->saddr = addr->sin6_addr;
369 /* Make sure we are allowed to bind here. */
370 if (sk->sk_prot->get_port(sk, snum)) {
371 inet_reset_saddr(sk);
372 err = -EADDRINUSE;
373 goto out;
376 if (addr_type != IPV6_ADDR_ANY) {
377 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
378 if (addr_type != IPV6_ADDR_MAPPED)
379 np->ipv6only = 1;
381 if (snum)
382 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
383 inet->inet_sport = htons(inet->inet_num);
384 inet->inet_dport = 0;
385 inet->inet_daddr = 0;
386 out:
387 release_sock(sk);
388 return err;
389 out_unlock:
390 rcu_read_unlock();
391 goto out;
394 EXPORT_SYMBOL(inet6_bind);
396 int inet6_release(struct socket *sock)
398 struct sock *sk = sock->sk;
400 if (sk == NULL)
401 return -EINVAL;
403 /* Free mc lists */
404 ipv6_sock_mc_close(sk);
406 /* Free ac lists */
407 ipv6_sock_ac_close(sk);
409 return inet_release(sock);
412 EXPORT_SYMBOL(inet6_release);
414 void inet6_destroy_sock(struct sock *sk)
416 struct ipv6_pinfo *np = inet6_sk(sk);
417 struct sk_buff *skb;
418 struct ipv6_txoptions *opt;
420 /* Release rx options */
422 if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
423 kfree_skb(skb);
425 if ((skb = xchg(&np->rxpmtu, NULL)) != NULL)
426 kfree_skb(skb);
428 /* Free flowlabels */
429 fl6_free_socklist(sk);
431 /* Free tx options */
433 if ((opt = xchg(&np->opt, NULL)) != NULL)
434 sock_kfree_s(sk, opt, opt->tot_len);
437 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
440 * This does both peername and sockname.
443 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
444 int *uaddr_len, int peer)
446 struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
447 struct sock *sk = sock->sk;
448 struct inet_sock *inet = inet_sk(sk);
449 struct ipv6_pinfo *np = inet6_sk(sk);
451 sin->sin6_family = AF_INET6;
452 sin->sin6_flowinfo = 0;
453 sin->sin6_scope_id = 0;
454 if (peer) {
455 if (!inet->inet_dport)
456 return -ENOTCONN;
457 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
458 peer == 1)
459 return -ENOTCONN;
460 sin->sin6_port = inet->inet_dport;
461 sin->sin6_addr = np->daddr;
462 if (np->sndflow)
463 sin->sin6_flowinfo = np->flow_label;
464 } else {
465 if (ipv6_addr_any(&np->rcv_saddr))
466 sin->sin6_addr = np->saddr;
467 else
468 sin->sin6_addr = np->rcv_saddr;
470 sin->sin6_port = inet->inet_sport;
472 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
473 sin->sin6_scope_id = sk->sk_bound_dev_if;
474 *uaddr_len = sizeof(*sin);
475 return 0;
478 EXPORT_SYMBOL(inet6_getname);
480 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
482 struct sock *sk = sock->sk;
483 struct net *net = sock_net(sk);
485 switch(cmd)
487 case SIOCGSTAMP:
488 return sock_get_timestamp(sk, (struct timeval __user *)arg);
490 case SIOCGSTAMPNS:
491 return sock_get_timestampns(sk, (struct timespec __user *)arg);
493 case SIOCADDRT:
494 case SIOCDELRT:
496 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
498 case SIOCSIFADDR:
499 return addrconf_add_ifaddr(net, (void __user *) arg);
500 case SIOCDIFADDR:
501 return addrconf_del_ifaddr(net, (void __user *) arg);
502 case SIOCSIFDSTADDR:
503 return addrconf_set_dstaddr(net, (void __user *) arg);
504 default:
505 if (!sk->sk_prot->ioctl)
506 return -ENOIOCTLCMD;
507 return sk->sk_prot->ioctl(sk, cmd, arg);
509 /*NOTREACHED*/
510 return 0;
513 EXPORT_SYMBOL(inet6_ioctl);
515 const struct proto_ops inet6_stream_ops = {
516 .family = PF_INET6,
517 .owner = THIS_MODULE,
518 .release = inet6_release,
519 .bind = inet6_bind,
520 .connect = inet_stream_connect, /* ok */
521 .socketpair = sock_no_socketpair, /* a do nothing */
522 .accept = inet_accept, /* ok */
523 .getname = inet6_getname,
524 .poll = tcp_poll, /* ok */
525 .ioctl = inet6_ioctl, /* must change */
526 .listen = inet_listen, /* ok */
527 .shutdown = inet_shutdown, /* ok */
528 .setsockopt = sock_common_setsockopt, /* ok */
529 .getsockopt = sock_common_getsockopt, /* ok */
530 .sendmsg = inet_sendmsg, /* ok */
531 .recvmsg = inet_recvmsg, /* ok */
532 .mmap = sock_no_mmap,
533 .sendpage = inet_sendpage,
534 .splice_read = tcp_splice_read,
535 #ifdef CONFIG_COMPAT
536 .compat_setsockopt = compat_sock_common_setsockopt,
537 .compat_getsockopt = compat_sock_common_getsockopt,
538 #endif
541 const struct proto_ops inet6_dgram_ops = {
542 .family = PF_INET6,
543 .owner = THIS_MODULE,
544 .release = inet6_release,
545 .bind = inet6_bind,
546 .connect = inet_dgram_connect, /* ok */
547 .socketpair = sock_no_socketpair, /* a do nothing */
548 .accept = sock_no_accept, /* a do nothing */
549 .getname = inet6_getname,
550 .poll = udp_poll, /* ok */
551 .ioctl = inet6_ioctl, /* must change */
552 .listen = sock_no_listen, /* ok */
553 .shutdown = inet_shutdown, /* ok */
554 .setsockopt = sock_common_setsockopt, /* ok */
555 .getsockopt = sock_common_getsockopt, /* ok */
556 .sendmsg = inet_sendmsg, /* ok */
557 .recvmsg = inet_recvmsg, /* ok */
558 .mmap = sock_no_mmap,
559 .sendpage = sock_no_sendpage,
560 #ifdef CONFIG_COMPAT
561 .compat_setsockopt = compat_sock_common_setsockopt,
562 .compat_getsockopt = compat_sock_common_getsockopt,
563 #endif
566 static const struct net_proto_family inet6_family_ops = {
567 .family = PF_INET6,
568 .create = inet6_create,
569 .owner = THIS_MODULE,
572 int inet6_register_protosw(struct inet_protosw *p)
574 struct list_head *lh;
575 struct inet_protosw *answer;
576 struct list_head *last_perm;
577 int protocol = p->protocol;
578 int ret;
580 spin_lock_bh(&inetsw6_lock);
582 ret = -EINVAL;
583 if (p->type >= SOCK_MAX)
584 goto out_illegal;
586 /* If we are trying to override a permanent protocol, bail. */
587 answer = NULL;
588 ret = -EPERM;
589 last_perm = &inetsw6[p->type];
590 list_for_each(lh, &inetsw6[p->type]) {
591 answer = list_entry(lh, struct inet_protosw, list);
593 /* Check only the non-wild match. */
594 if (INET_PROTOSW_PERMANENT & answer->flags) {
595 if (protocol == answer->protocol)
596 break;
597 last_perm = lh;
600 answer = NULL;
602 if (answer)
603 goto out_permanent;
605 /* Add the new entry after the last permanent entry if any, so that
606 * the new entry does not override a permanent entry when matched with
607 * a wild-card protocol. But it is allowed to override any existing
608 * non-permanent entry. This means that when we remove this entry, the
609 * system automatically returns to the old behavior.
611 list_add_rcu(&p->list, last_perm);
612 ret = 0;
613 out:
614 spin_unlock_bh(&inetsw6_lock);
615 return ret;
617 out_permanent:
618 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
619 protocol);
620 goto out;
622 out_illegal:
623 printk(KERN_ERR
624 "Ignoring attempt to register invalid socket type %d.\n",
625 p->type);
626 goto out;
629 EXPORT_SYMBOL(inet6_register_protosw);
631 void
632 inet6_unregister_protosw(struct inet_protosw *p)
634 if (INET_PROTOSW_PERMANENT & p->flags) {
635 printk(KERN_ERR
636 "Attempt to unregister permanent protocol %d.\n",
637 p->protocol);
638 } else {
639 spin_lock_bh(&inetsw6_lock);
640 list_del_rcu(&p->list);
641 spin_unlock_bh(&inetsw6_lock);
643 synchronize_net();
647 EXPORT_SYMBOL(inet6_unregister_protosw);
649 int inet6_sk_rebuild_header(struct sock *sk)
651 struct ipv6_pinfo *np = inet6_sk(sk);
652 struct dst_entry *dst;
654 dst = __sk_dst_check(sk, np->dst_cookie);
656 if (dst == NULL) {
657 struct inet_sock *inet = inet_sk(sk);
658 struct in6_addr *final_p, final;
659 struct flowi6 fl6;
661 memset(&fl6, 0, sizeof(fl6));
662 fl6.flowi6_proto = sk->sk_protocol;
663 fl6.daddr = np->daddr;
664 fl6.saddr = np->saddr;
665 fl6.flowlabel = np->flow_label;
666 fl6.flowi6_oif = sk->sk_bound_dev_if;
667 fl6.flowi6_mark = sk->sk_mark;
668 fl6.fl6_dport = inet->inet_dport;
669 fl6.fl6_sport = inet->inet_sport;
670 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
672 final_p = fl6_update_dst(&fl6, np->opt, &final);
674 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
675 if (IS_ERR(dst)) {
676 sk->sk_route_caps = 0;
677 sk->sk_err_soft = -PTR_ERR(dst);
678 return PTR_ERR(dst);
681 __ip6_dst_store(sk, dst, NULL, NULL);
684 return 0;
687 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
689 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
691 struct ipv6_pinfo *np = inet6_sk(sk);
692 struct inet6_skb_parm *opt = IP6CB(skb);
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;
706 return 0;
709 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
711 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
713 const struct inet6_protocol *ops = NULL;
715 for (;;) {
716 struct ipv6_opt_hdr *opth;
717 int len;
719 if (proto != NEXTHDR_HOP) {
720 ops = rcu_dereference(inet6_protos[proto]);
722 if (unlikely(!ops))
723 break;
725 if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
726 break;
729 if (unlikely(!pskb_may_pull(skb, 8)))
730 break;
732 opth = (void *)skb->data;
733 len = ipv6_optlen(opth);
735 if (unlikely(!pskb_may_pull(skb, len)))
736 break;
738 proto = opth->nexthdr;
739 __skb_pull(skb, len);
742 return proto;
745 static int ipv6_gso_send_check(struct sk_buff *skb)
747 const struct ipv6hdr *ipv6h;
748 const struct inet6_protocol *ops;
749 int err = -EINVAL;
751 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
752 goto out;
754 ipv6h = ipv6_hdr(skb);
755 __skb_pull(skb, sizeof(*ipv6h));
756 err = -EPROTONOSUPPORT;
758 rcu_read_lock();
759 ops = rcu_dereference(inet6_protos[
760 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
762 if (likely(ops && ops->gso_send_check)) {
763 skb_reset_transport_header(skb);
764 err = ops->gso_send_check(skb);
766 rcu_read_unlock();
768 out:
769 return err;
772 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb,
773 netdev_features_t features)
775 struct sk_buff *segs = ERR_PTR(-EINVAL);
776 struct ipv6hdr *ipv6h;
777 const struct inet6_protocol *ops;
778 int proto;
779 struct frag_hdr *fptr;
780 unsigned int unfrag_ip6hlen;
781 u8 *prevhdr;
782 int offset = 0;
784 if (!(features & NETIF_F_V6_CSUM))
785 features &= ~NETIF_F_SG;
787 if (unlikely(skb_shinfo(skb)->gso_type &
788 ~(SKB_GSO_UDP |
789 SKB_GSO_DODGY |
790 SKB_GSO_TCP_ECN |
791 SKB_GSO_TCPV6 |
792 0)))
793 goto out;
795 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
796 goto out;
798 ipv6h = ipv6_hdr(skb);
799 __skb_pull(skb, sizeof(*ipv6h));
800 segs = ERR_PTR(-EPROTONOSUPPORT);
802 proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
803 rcu_read_lock();
804 ops = rcu_dereference(inet6_protos[proto]);
805 if (likely(ops && ops->gso_segment)) {
806 skb_reset_transport_header(skb);
807 segs = ops->gso_segment(skb, features);
809 rcu_read_unlock();
811 if (IS_ERR(segs))
812 goto out;
814 for (skb = segs; skb; skb = skb->next) {
815 ipv6h = ipv6_hdr(skb);
816 ipv6h->payload_len = htons(skb->len - skb->mac_len -
817 sizeof(*ipv6h));
818 if (proto == IPPROTO_UDP) {
819 unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
820 fptr = (struct frag_hdr *)(skb_network_header(skb) +
821 unfrag_ip6hlen);
822 fptr->frag_off = htons(offset);
823 if (skb->next != NULL)
824 fptr->frag_off |= htons(IP6_MF);
825 offset += (ntohs(ipv6h->payload_len) -
826 sizeof(struct frag_hdr));
830 out:
831 return segs;
834 struct ipv6_gro_cb {
835 struct napi_gro_cb napi;
836 int proto;
839 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
841 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
842 struct sk_buff *skb)
844 const struct inet6_protocol *ops;
845 struct sk_buff **pp = NULL;
846 struct sk_buff *p;
847 struct ipv6hdr *iph;
848 unsigned int nlen;
849 unsigned int hlen;
850 unsigned int off;
851 int flush = 1;
852 int proto;
853 __wsum csum;
855 off = skb_gro_offset(skb);
856 hlen = off + sizeof(*iph);
857 iph = skb_gro_header_fast(skb, off);
858 if (skb_gro_header_hard(skb, hlen)) {
859 iph = skb_gro_header_slow(skb, hlen, off);
860 if (unlikely(!iph))
861 goto out;
864 skb_gro_pull(skb, sizeof(*iph));
865 skb_set_transport_header(skb, skb_gro_offset(skb));
867 flush += ntohs(iph->payload_len) != skb_gro_len(skb);
869 rcu_read_lock();
870 proto = iph->nexthdr;
871 ops = rcu_dereference(inet6_protos[proto]);
872 if (!ops || !ops->gro_receive) {
873 __pskb_pull(skb, skb_gro_offset(skb));
874 proto = ipv6_gso_pull_exthdrs(skb, proto);
875 skb_gro_pull(skb, -skb_transport_offset(skb));
876 skb_reset_transport_header(skb);
877 __skb_push(skb, skb_gro_offset(skb));
879 ops = rcu_dereference(inet6_protos[proto]);
880 if (!ops || !ops->gro_receive)
881 goto out_unlock;
883 iph = ipv6_hdr(skb);
886 IPV6_GRO_CB(skb)->proto = proto;
888 flush--;
889 nlen = skb_network_header_len(skb);
891 for (p = *head; p; p = p->next) {
892 struct ipv6hdr *iph2;
894 if (!NAPI_GRO_CB(p)->same_flow)
895 continue;
897 iph2 = ipv6_hdr(p);
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;
908 NAPI_GRO_CB(p)->flush |= flush;
911 NAPI_GRO_CB(skb)->flush |= flush;
913 csum = skb->csum;
914 skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
916 pp = ops->gro_receive(head, skb);
918 skb->csum = csum;
920 out_unlock:
921 rcu_read_unlock();
923 out:
924 NAPI_GRO_CB(skb)->flush |= flush;
926 return pp;
929 static int ipv6_gro_complete(struct sk_buff *skb)
931 const struct inet6_protocol *ops;
932 struct ipv6hdr *iph = ipv6_hdr(skb);
933 int err = -ENOSYS;
935 iph->payload_len = htons(skb->len - skb_network_offset(skb) -
936 sizeof(*iph));
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;
943 err = ops->gro_complete(skb);
945 out_unlock:
946 rcu_read_unlock();
948 return err;
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,
960 static int __init ipv6_packet_init(void)
962 dev_add_pack(&ipv6_packet_type);
963 return 0;
966 static void ipv6_packet_cleanup(void)
968 dev_remove_pack(&ipv6_packet_type);
971 static int __net_init ipv6_init_mibs(struct net *net)
973 if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
974 sizeof(struct udp_mib),
975 __alignof__(struct udp_mib)) < 0)
976 return -ENOMEM;
977 if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
978 sizeof(struct udp_mib),
979 __alignof__(struct udp_mib)) < 0)
980 goto err_udplite_mib;
981 if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
982 sizeof(struct ipstats_mib),
983 __alignof__(struct ipstats_mib)) < 0)
984 goto err_ip_mib;
985 if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
986 sizeof(struct icmpv6_mib),
987 __alignof__(struct icmpv6_mib)) < 0)
988 goto err_icmp_mib;
989 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
990 GFP_KERNEL);
991 if (!net->mib.icmpv6msg_statistics)
992 goto err_icmpmsg_mib;
993 return 0;
995 err_icmpmsg_mib:
996 snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
997 err_icmp_mib:
998 snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
999 err_ip_mib:
1000 snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1001 err_udplite_mib:
1002 snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1003 return -ENOMEM;
1006 static void ipv6_cleanup_mibs(struct net *net)
1008 snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1009 snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1010 snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1011 snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1012 kfree(net->mib.icmpv6msg_statistics);
1015 static int __net_init inet6_net_init(struct net *net)
1017 int err = 0;
1019 net->ipv6.sysctl.bindv6only = 0;
1020 net->ipv6.sysctl.icmpv6_time = 1*HZ;
1022 err = ipv6_init_mibs(net);
1023 if (err)
1024 return err;
1025 #ifdef CONFIG_PROC_FS
1026 err = udp6_proc_init(net);
1027 if (err)
1028 goto out;
1029 err = tcp6_proc_init(net);
1030 if (err)
1031 goto proc_tcp6_fail;
1032 err = ac6_proc_init(net);
1033 if (err)
1034 goto proc_ac6_fail;
1035 #endif
1036 return err;
1038 #ifdef CONFIG_PROC_FS
1039 proc_ac6_fail:
1040 tcp6_proc_exit(net);
1041 proc_tcp6_fail:
1042 udp6_proc_exit(net);
1043 out:
1044 ipv6_cleanup_mibs(net);
1045 return err;
1046 #endif
1049 static void __net_exit inet6_net_exit(struct net *net)
1051 #ifdef CONFIG_PROC_FS
1052 udp6_proc_exit(net);
1053 tcp6_proc_exit(net);
1054 ac6_proc_exit(net);
1055 #endif
1056 ipv6_cleanup_mibs(net);
1059 static struct pernet_operations inet6_net_ops = {
1060 .init = inet6_net_init,
1061 .exit = inet6_net_exit,
1064 static int __init inet6_init(void)
1066 struct sk_buff *dummy_skb;
1067 struct list_head *r;
1068 int err = 0;
1070 BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1072 /* Register the socket-side information for inet6_create. */
1073 for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1074 INIT_LIST_HEAD(r);
1076 if (disable_ipv6_mod) {
1077 printk(KERN_INFO
1078 "IPv6: Loaded, but administratively disabled, "
1079 "reboot required to enable\n");
1080 goto out;
1083 err = proto_register(&tcpv6_prot, 1);
1084 if (err)
1085 goto out;
1087 err = proto_register(&udpv6_prot, 1);
1088 if (err)
1089 goto out_unregister_tcp_proto;
1091 err = proto_register(&udplitev6_prot, 1);
1092 if (err)
1093 goto out_unregister_udp_proto;
1095 err = proto_register(&rawv6_prot, 1);
1096 if (err)
1097 goto out_unregister_udplite_proto;
1100 /* We MUST register RAW sockets before we create the ICMP6,
1101 * IGMP6, or NDISC control sockets.
1103 err = rawv6_init();
1104 if (err)
1105 goto out_unregister_raw_proto;
1107 /* Register the family here so that the init calls below will
1108 * be able to create sockets. (?? is this dangerous ??)
1110 err = sock_register(&inet6_family_ops);
1111 if (err)
1112 goto out_sock_register_fail;
1114 #ifdef CONFIG_SYSCTL
1115 err = ipv6_static_sysctl_register();
1116 if (err)
1117 goto static_sysctl_fail;
1118 #endif
1119 tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
1122 * ipngwg API draft makes clear that the correct semantics
1123 * for TCP and UDP is to consider one TCP and UDP instance
1124 * in a host available by both INET and INET6 APIs and
1125 * able to communicate via both network protocols.
1128 err = register_pernet_subsys(&inet6_net_ops);
1129 if (err)
1130 goto register_pernet_fail;
1131 err = icmpv6_init();
1132 if (err)
1133 goto icmp_fail;
1134 err = ip6_mr_init();
1135 if (err)
1136 goto ipmr_fail;
1137 err = ndisc_init();
1138 if (err)
1139 goto ndisc_fail;
1140 err = igmp6_init();
1141 if (err)
1142 goto igmp_fail;
1143 err = ipv6_netfilter_init();
1144 if (err)
1145 goto netfilter_fail;
1146 /* Create /proc/foo6 entries. */
1147 #ifdef CONFIG_PROC_FS
1148 err = -ENOMEM;
1149 if (raw6_proc_init())
1150 goto proc_raw6_fail;
1151 if (udplite6_proc_init())
1152 goto proc_udplite6_fail;
1153 if (ipv6_misc_proc_init())
1154 goto proc_misc6_fail;
1155 if (if6_proc_init())
1156 goto proc_if6_fail;
1157 #endif
1158 err = ip6_route_init();
1159 if (err)
1160 goto ip6_route_fail;
1161 err = ip6_flowlabel_init();
1162 if (err)
1163 goto ip6_flowlabel_fail;
1164 err = addrconf_init();
1165 if (err)
1166 goto addrconf_fail;
1168 /* Init v6 extension headers. */
1169 err = ipv6_exthdrs_init();
1170 if (err)
1171 goto ipv6_exthdrs_fail;
1173 err = ipv6_frag_init();
1174 if (err)
1175 goto ipv6_frag_fail;
1177 /* Init v6 transport protocols. */
1178 err = udpv6_init();
1179 if (err)
1180 goto udpv6_fail;
1182 err = udplitev6_init();
1183 if (err)
1184 goto udplitev6_fail;
1186 err = tcpv6_init();
1187 if (err)
1188 goto tcpv6_fail;
1190 err = ipv6_packet_init();
1191 if (err)
1192 goto ipv6_packet_fail;
1194 #ifdef CONFIG_SYSCTL
1195 err = ipv6_sysctl_register();
1196 if (err)
1197 goto sysctl_fail;
1198 #endif
1199 out:
1200 return err;
1202 #ifdef CONFIG_SYSCTL
1203 sysctl_fail:
1204 ipv6_packet_cleanup();
1205 #endif
1206 ipv6_packet_fail:
1207 tcpv6_exit();
1208 tcpv6_fail:
1209 udplitev6_exit();
1210 udplitev6_fail:
1211 udpv6_exit();
1212 udpv6_fail:
1213 ipv6_frag_exit();
1214 ipv6_frag_fail:
1215 ipv6_exthdrs_exit();
1216 ipv6_exthdrs_fail:
1217 addrconf_cleanup();
1218 addrconf_fail:
1219 ip6_flowlabel_cleanup();
1220 ip6_flowlabel_fail:
1221 ip6_route_cleanup();
1222 ip6_route_fail:
1223 #ifdef CONFIG_PROC_FS
1224 if6_proc_exit();
1225 proc_if6_fail:
1226 ipv6_misc_proc_exit();
1227 proc_misc6_fail:
1228 udplite6_proc_exit();
1229 proc_udplite6_fail:
1230 raw6_proc_exit();
1231 proc_raw6_fail:
1232 #endif
1233 ipv6_netfilter_fini();
1234 netfilter_fail:
1235 igmp6_cleanup();
1236 igmp_fail:
1237 ndisc_cleanup();
1238 ndisc_fail:
1239 ip6_mr_cleanup();
1240 ipmr_fail:
1241 icmpv6_cleanup();
1242 icmp_fail:
1243 unregister_pernet_subsys(&inet6_net_ops);
1244 register_pernet_fail:
1245 #ifdef CONFIG_SYSCTL
1246 ipv6_static_sysctl_unregister();
1247 static_sysctl_fail:
1248 #endif
1249 sock_unregister(PF_INET6);
1250 rtnl_unregister_all(PF_INET6);
1251 out_sock_register_fail:
1252 rawv6_exit();
1253 out_unregister_raw_proto:
1254 proto_unregister(&rawv6_prot);
1255 out_unregister_udplite_proto:
1256 proto_unregister(&udplitev6_prot);
1257 out_unregister_udp_proto:
1258 proto_unregister(&udpv6_prot);
1259 out_unregister_tcp_proto:
1260 proto_unregister(&tcpv6_prot);
1261 goto out;
1263 module_init(inet6_init);
1265 static void __exit inet6_exit(void)
1267 if (disable_ipv6_mod)
1268 return;
1270 /* First of all disallow new sockets creation. */
1271 sock_unregister(PF_INET6);
1272 /* Disallow any further netlink messages */
1273 rtnl_unregister_all(PF_INET6);
1275 #ifdef CONFIG_SYSCTL
1276 ipv6_sysctl_unregister();
1277 #endif
1278 udpv6_exit();
1279 udplitev6_exit();
1280 tcpv6_exit();
1282 /* Cleanup code parts. */
1283 ipv6_packet_cleanup();
1284 ipv6_frag_exit();
1285 ipv6_exthdrs_exit();
1286 addrconf_cleanup();
1287 ip6_flowlabel_cleanup();
1288 ip6_route_cleanup();
1289 #ifdef CONFIG_PROC_FS
1291 /* Cleanup code parts. */
1292 if6_proc_exit();
1293 ipv6_misc_proc_exit();
1294 udplite6_proc_exit();
1295 raw6_proc_exit();
1296 #endif
1297 ipv6_netfilter_fini();
1298 igmp6_cleanup();
1299 ndisc_cleanup();
1300 ip6_mr_cleanup();
1301 icmpv6_cleanup();
1302 rawv6_exit();
1304 unregister_pernet_subsys(&inet6_net_ops);
1305 #ifdef CONFIG_SYSCTL
1306 ipv6_static_sysctl_unregister();
1307 #endif
1308 proto_unregister(&rawv6_prot);
1309 proto_unregister(&udplitev6_prot);
1310 proto_unregister(&udpv6_prot);
1311 proto_unregister(&tcpv6_prot);
1313 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1315 module_exit(inet6_exit);
1317 MODULE_ALIAS_NETPROTO(PF_INET6);