KVM: Split mmio completion into a function
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / af_inet6.c
blobafcc7099f96de4c3eb991a4006936b5c7e95b978
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;
275 addr_type = ipv6_addr_type(&addr->sin6_addr);
276 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
277 return -EINVAL;
279 snum = ntohs(addr->sin6_port);
280 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
281 return -EACCES;
283 lock_sock(sk);
285 /* Check these errors (active socket, double bind). */
286 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
287 err = -EINVAL;
288 goto out;
291 /* Check if the address belongs to the host. */
292 if (addr_type == IPV6_ADDR_MAPPED) {
293 int chk_addr_ret;
295 /* Binding to v4-mapped address on a v6-only socket
296 * makes no sense
298 if (np->ipv6only) {
299 err = -EINVAL;
300 goto out;
303 /* Reproduce AF_INET checks to make the bindings consistent */
304 v4addr = addr->sin6_addr.s6_addr32[3];
305 chk_addr_ret = inet_addr_type(net, v4addr);
306 if (!sysctl_ip_nonlocal_bind &&
307 !(inet->freebind || inet->transparent) &&
308 v4addr != htonl(INADDR_ANY) &&
309 chk_addr_ret != RTN_LOCAL &&
310 chk_addr_ret != RTN_MULTICAST &&
311 chk_addr_ret != RTN_BROADCAST) {
312 err = -EADDRNOTAVAIL;
313 goto out;
315 } else {
316 if (addr_type != IPV6_ADDR_ANY) {
317 struct net_device *dev = NULL;
319 rcu_read_lock();
320 if (addr_type & IPV6_ADDR_LINKLOCAL) {
321 if (addr_len >= sizeof(struct sockaddr_in6) &&
322 addr->sin6_scope_id) {
323 /* Override any existing binding, if another one
324 * is supplied by user.
326 sk->sk_bound_dev_if = addr->sin6_scope_id;
329 /* Binding to link-local address requires an interface */
330 if (!sk->sk_bound_dev_if) {
331 err = -EINVAL;
332 goto out_unlock;
334 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
335 if (!dev) {
336 err = -ENODEV;
337 goto out_unlock;
341 /* ipv4 addr of the socket is invalid. Only the
342 * unspecified and mapped address have a v4 equivalent.
344 v4addr = LOOPBACK4_IPV6;
345 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
346 if (!inet->transparent &&
347 !ipv6_chk_addr(net, &addr->sin6_addr,
348 dev, 0)) {
349 err = -EADDRNOTAVAIL;
350 goto out_unlock;
353 rcu_read_unlock();
357 inet->inet_rcv_saddr = v4addr;
358 inet->inet_saddr = v4addr;
360 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
362 if (!(addr_type & IPV6_ADDR_MULTICAST))
363 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
365 /* Make sure we are allowed to bind here. */
366 if (sk->sk_prot->get_port(sk, snum)) {
367 inet_reset_saddr(sk);
368 err = -EADDRINUSE;
369 goto out;
372 if (addr_type != IPV6_ADDR_ANY) {
373 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
374 if (addr_type != IPV6_ADDR_MAPPED)
375 np->ipv6only = 1;
377 if (snum)
378 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
379 inet->inet_sport = htons(inet->inet_num);
380 inet->inet_dport = 0;
381 inet->inet_daddr = 0;
382 out:
383 release_sock(sk);
384 return err;
385 out_unlock:
386 rcu_read_unlock();
387 goto out;
390 EXPORT_SYMBOL(inet6_bind);
392 int inet6_release(struct socket *sock)
394 struct sock *sk = sock->sk;
396 if (sk == NULL)
397 return -EINVAL;
399 /* Free mc lists */
400 ipv6_sock_mc_close(sk);
402 /* Free ac lists */
403 ipv6_sock_ac_close(sk);
405 return inet_release(sock);
408 EXPORT_SYMBOL(inet6_release);
410 void inet6_destroy_sock(struct sock *sk)
412 struct ipv6_pinfo *np = inet6_sk(sk);
413 struct sk_buff *skb;
414 struct ipv6_txoptions *opt;
416 /* Release rx options */
418 if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
419 kfree_skb(skb);
421 if ((skb = xchg(&np->rxpmtu, NULL)) != NULL)
422 kfree_skb(skb);
424 /* Free flowlabels */
425 fl6_free_socklist(sk);
427 /* Free tx options */
429 if ((opt = xchg(&np->opt, NULL)) != NULL)
430 sock_kfree_s(sk, opt, opt->tot_len);
433 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
436 * This does both peername and sockname.
439 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
440 int *uaddr_len, int peer)
442 struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
443 struct sock *sk = sock->sk;
444 struct inet_sock *inet = inet_sk(sk);
445 struct ipv6_pinfo *np = inet6_sk(sk);
447 sin->sin6_family = AF_INET6;
448 sin->sin6_flowinfo = 0;
449 sin->sin6_scope_id = 0;
450 if (peer) {
451 if (!inet->inet_dport)
452 return -ENOTCONN;
453 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
454 peer == 1)
455 return -ENOTCONN;
456 sin->sin6_port = inet->inet_dport;
457 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
458 if (np->sndflow)
459 sin->sin6_flowinfo = np->flow_label;
460 } else {
461 if (ipv6_addr_any(&np->rcv_saddr))
462 ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
463 else
464 ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
466 sin->sin6_port = inet->inet_sport;
468 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
469 sin->sin6_scope_id = sk->sk_bound_dev_if;
470 *uaddr_len = sizeof(*sin);
471 return 0;
474 EXPORT_SYMBOL(inet6_getname);
476 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
478 struct sock *sk = sock->sk;
479 struct net *net = sock_net(sk);
481 switch(cmd)
483 case SIOCGSTAMP:
484 return sock_get_timestamp(sk, (struct timeval __user *)arg);
486 case SIOCGSTAMPNS:
487 return sock_get_timestampns(sk, (struct timespec __user *)arg);
489 case SIOCADDRT:
490 case SIOCDELRT:
492 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
494 case SIOCSIFADDR:
495 return addrconf_add_ifaddr(net, (void __user *) arg);
496 case SIOCDIFADDR:
497 return addrconf_del_ifaddr(net, (void __user *) arg);
498 case SIOCSIFDSTADDR:
499 return addrconf_set_dstaddr(net, (void __user *) arg);
500 default:
501 if (!sk->sk_prot->ioctl)
502 return -ENOIOCTLCMD;
503 return sk->sk_prot->ioctl(sk, cmd, arg);
505 /*NOTREACHED*/
506 return 0;
509 EXPORT_SYMBOL(inet6_ioctl);
511 const struct proto_ops inet6_stream_ops = {
512 .family = PF_INET6,
513 .owner = THIS_MODULE,
514 .release = inet6_release,
515 .bind = inet6_bind,
516 .connect = inet_stream_connect, /* ok */
517 .socketpair = sock_no_socketpair, /* a do nothing */
518 .accept = inet_accept, /* ok */
519 .getname = inet6_getname,
520 .poll = tcp_poll, /* ok */
521 .ioctl = inet6_ioctl, /* must change */
522 .listen = inet_listen, /* ok */
523 .shutdown = inet_shutdown, /* ok */
524 .setsockopt = sock_common_setsockopt, /* ok */
525 .getsockopt = sock_common_getsockopt, /* ok */
526 .sendmsg = inet_sendmsg, /* ok */
527 .recvmsg = inet_recvmsg, /* ok */
528 .mmap = sock_no_mmap,
529 .sendpage = inet_sendpage,
530 .splice_read = tcp_splice_read,
531 #ifdef CONFIG_COMPAT
532 .compat_setsockopt = compat_sock_common_setsockopt,
533 .compat_getsockopt = compat_sock_common_getsockopt,
534 #endif
537 const struct proto_ops inet6_dgram_ops = {
538 .family = PF_INET6,
539 .owner = THIS_MODULE,
540 .release = inet6_release,
541 .bind = inet6_bind,
542 .connect = inet_dgram_connect, /* ok */
543 .socketpair = sock_no_socketpair, /* a do nothing */
544 .accept = sock_no_accept, /* a do nothing */
545 .getname = inet6_getname,
546 .poll = udp_poll, /* ok */
547 .ioctl = inet6_ioctl, /* must change */
548 .listen = sock_no_listen, /* ok */
549 .shutdown = inet_shutdown, /* ok */
550 .setsockopt = sock_common_setsockopt, /* ok */
551 .getsockopt = sock_common_getsockopt, /* ok */
552 .sendmsg = inet_sendmsg, /* ok */
553 .recvmsg = inet_recvmsg, /* ok */
554 .mmap = sock_no_mmap,
555 .sendpage = sock_no_sendpage,
556 #ifdef CONFIG_COMPAT
557 .compat_setsockopt = compat_sock_common_setsockopt,
558 .compat_getsockopt = compat_sock_common_getsockopt,
559 #endif
562 static const struct net_proto_family inet6_family_ops = {
563 .family = PF_INET6,
564 .create = inet6_create,
565 .owner = THIS_MODULE,
568 int inet6_register_protosw(struct inet_protosw *p)
570 struct list_head *lh;
571 struct inet_protosw *answer;
572 struct list_head *last_perm;
573 int protocol = p->protocol;
574 int ret;
576 spin_lock_bh(&inetsw6_lock);
578 ret = -EINVAL;
579 if (p->type >= SOCK_MAX)
580 goto out_illegal;
582 /* If we are trying to override a permanent protocol, bail. */
583 answer = NULL;
584 ret = -EPERM;
585 last_perm = &inetsw6[p->type];
586 list_for_each(lh, &inetsw6[p->type]) {
587 answer = list_entry(lh, struct inet_protosw, list);
589 /* Check only the non-wild match. */
590 if (INET_PROTOSW_PERMANENT & answer->flags) {
591 if (protocol == answer->protocol)
592 break;
593 last_perm = lh;
596 answer = NULL;
598 if (answer)
599 goto out_permanent;
601 /* Add the new entry after the last permanent entry if any, so that
602 * the new entry does not override a permanent entry when matched with
603 * a wild-card protocol. But it is allowed to override any existing
604 * non-permanent entry. This means that when we remove this entry, the
605 * system automatically returns to the old behavior.
607 list_add_rcu(&p->list, last_perm);
608 ret = 0;
609 out:
610 spin_unlock_bh(&inetsw6_lock);
611 return ret;
613 out_permanent:
614 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
615 protocol);
616 goto out;
618 out_illegal:
619 printk(KERN_ERR
620 "Ignoring attempt to register invalid socket type %d.\n",
621 p->type);
622 goto out;
625 EXPORT_SYMBOL(inet6_register_protosw);
627 void
628 inet6_unregister_protosw(struct inet_protosw *p)
630 if (INET_PROTOSW_PERMANENT & p->flags) {
631 printk(KERN_ERR
632 "Attempt to unregister permanent protocol %d.\n",
633 p->protocol);
634 } else {
635 spin_lock_bh(&inetsw6_lock);
636 list_del_rcu(&p->list);
637 spin_unlock_bh(&inetsw6_lock);
639 synchronize_net();
643 EXPORT_SYMBOL(inet6_unregister_protosw);
645 int inet6_sk_rebuild_header(struct sock *sk)
647 struct ipv6_pinfo *np = inet6_sk(sk);
648 struct dst_entry *dst;
650 dst = __sk_dst_check(sk, np->dst_cookie);
652 if (dst == NULL) {
653 struct inet_sock *inet = inet_sk(sk);
654 struct in6_addr *final_p, final;
655 struct flowi6 fl6;
657 memset(&fl6, 0, sizeof(fl6));
658 fl6.flowi6_proto = sk->sk_protocol;
659 ipv6_addr_copy(&fl6.daddr, &np->daddr);
660 ipv6_addr_copy(&fl6.saddr, &np->saddr);
661 fl6.flowlabel = np->flow_label;
662 fl6.flowi6_oif = sk->sk_bound_dev_if;
663 fl6.flowi6_mark = sk->sk_mark;
664 fl6.fl6_dport = inet->inet_dport;
665 fl6.fl6_sport = inet->inet_sport;
666 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
668 final_p = fl6_update_dst(&fl6, np->opt, &final);
670 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
671 if (IS_ERR(dst)) {
672 sk->sk_route_caps = 0;
673 sk->sk_err_soft = -PTR_ERR(dst);
674 return PTR_ERR(dst);
677 __ip6_dst_store(sk, dst, NULL, NULL);
680 return 0;
683 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
685 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
687 struct ipv6_pinfo *np = inet6_sk(sk);
688 struct inet6_skb_parm *opt = IP6CB(skb);
690 if (np->rxopt.all) {
691 if ((opt->hop && (np->rxopt.bits.hopopts ||
692 np->rxopt.bits.ohopopts)) ||
693 ((IPV6_FLOWINFO_MASK &
694 *(__be32 *)skb_network_header(skb)) &&
695 np->rxopt.bits.rxflow) ||
696 (opt->srcrt && (np->rxopt.bits.srcrt ||
697 np->rxopt.bits.osrcrt)) ||
698 ((opt->dst1 || opt->dst0) &&
699 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
700 return 1;
702 return 0;
705 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
707 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
709 const struct inet6_protocol *ops = NULL;
711 for (;;) {
712 struct ipv6_opt_hdr *opth;
713 int len;
715 if (proto != NEXTHDR_HOP) {
716 ops = rcu_dereference(inet6_protos[proto]);
718 if (unlikely(!ops))
719 break;
721 if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
722 break;
725 if (unlikely(!pskb_may_pull(skb, 8)))
726 break;
728 opth = (void *)skb->data;
729 len = ipv6_optlen(opth);
731 if (unlikely(!pskb_may_pull(skb, len)))
732 break;
734 proto = opth->nexthdr;
735 __skb_pull(skb, len);
738 return proto;
741 static int ipv6_gso_send_check(struct sk_buff *skb)
743 struct ipv6hdr *ipv6h;
744 const struct inet6_protocol *ops;
745 int err = -EINVAL;
747 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
748 goto out;
750 ipv6h = ipv6_hdr(skb);
751 __skb_pull(skb, sizeof(*ipv6h));
752 err = -EPROTONOSUPPORT;
754 rcu_read_lock();
755 ops = rcu_dereference(inet6_protos[
756 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
758 if (likely(ops && ops->gso_send_check)) {
759 skb_reset_transport_header(skb);
760 err = ops->gso_send_check(skb);
762 rcu_read_unlock();
764 out:
765 return err;
768 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, u32 features)
770 struct sk_buff *segs = ERR_PTR(-EINVAL);
771 struct ipv6hdr *ipv6h;
772 const struct inet6_protocol *ops;
773 int proto;
774 struct frag_hdr *fptr;
775 unsigned int unfrag_ip6hlen;
776 u8 *prevhdr;
777 int offset = 0;
779 if (!(features & NETIF_F_V6_CSUM))
780 features &= ~NETIF_F_SG;
782 if (unlikely(skb_shinfo(skb)->gso_type &
783 ~(SKB_GSO_UDP |
784 SKB_GSO_DODGY |
785 SKB_GSO_TCP_ECN |
786 SKB_GSO_TCPV6 |
787 0)))
788 goto out;
790 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
791 goto out;
793 ipv6h = ipv6_hdr(skb);
794 __skb_pull(skb, sizeof(*ipv6h));
795 segs = ERR_PTR(-EPROTONOSUPPORT);
797 proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
798 rcu_read_lock();
799 ops = rcu_dereference(inet6_protos[proto]);
800 if (likely(ops && ops->gso_segment)) {
801 skb_reset_transport_header(skb);
802 segs = ops->gso_segment(skb, features);
804 rcu_read_unlock();
806 if (IS_ERR(segs))
807 goto out;
809 for (skb = segs; skb; skb = skb->next) {
810 ipv6h = ipv6_hdr(skb);
811 ipv6h->payload_len = htons(skb->len - skb->mac_len -
812 sizeof(*ipv6h));
813 if (proto == IPPROTO_UDP) {
814 unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
815 fptr = (struct frag_hdr *)(skb_network_header(skb) +
816 unfrag_ip6hlen);
817 fptr->frag_off = htons(offset);
818 if (skb->next != NULL)
819 fptr->frag_off |= htons(IP6_MF);
820 offset += (ntohs(ipv6h->payload_len) -
821 sizeof(struct frag_hdr));
825 out:
826 return segs;
829 struct ipv6_gro_cb {
830 struct napi_gro_cb napi;
831 int proto;
834 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
836 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
837 struct sk_buff *skb)
839 const struct inet6_protocol *ops;
840 struct sk_buff **pp = NULL;
841 struct sk_buff *p;
842 struct ipv6hdr *iph;
843 unsigned int nlen;
844 unsigned int hlen;
845 unsigned int off;
846 int flush = 1;
847 int proto;
848 __wsum csum;
850 off = skb_gro_offset(skb);
851 hlen = off + sizeof(*iph);
852 iph = skb_gro_header_fast(skb, off);
853 if (skb_gro_header_hard(skb, hlen)) {
854 iph = skb_gro_header_slow(skb, hlen, off);
855 if (unlikely(!iph))
856 goto out;
859 skb_gro_pull(skb, sizeof(*iph));
860 skb_set_transport_header(skb, skb_gro_offset(skb));
862 flush += ntohs(iph->payload_len) != skb_gro_len(skb);
864 rcu_read_lock();
865 proto = iph->nexthdr;
866 ops = rcu_dereference(inet6_protos[proto]);
867 if (!ops || !ops->gro_receive) {
868 __pskb_pull(skb, skb_gro_offset(skb));
869 proto = ipv6_gso_pull_exthdrs(skb, proto);
870 skb_gro_pull(skb, -skb_transport_offset(skb));
871 skb_reset_transport_header(skb);
872 __skb_push(skb, skb_gro_offset(skb));
874 if (!ops || !ops->gro_receive)
875 goto out_unlock;
877 iph = ipv6_hdr(skb);
880 IPV6_GRO_CB(skb)->proto = proto;
882 flush--;
883 nlen = skb_network_header_len(skb);
885 for (p = *head; p; p = p->next) {
886 struct ipv6hdr *iph2;
888 if (!NAPI_GRO_CB(p)->same_flow)
889 continue;
891 iph2 = ipv6_hdr(p);
893 /* All fields must match except length. */
894 if (nlen != skb_network_header_len(p) ||
895 memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
896 memcmp(&iph->nexthdr, &iph2->nexthdr,
897 nlen - offsetof(struct ipv6hdr, nexthdr))) {
898 NAPI_GRO_CB(p)->same_flow = 0;
899 continue;
902 NAPI_GRO_CB(p)->flush |= flush;
905 NAPI_GRO_CB(skb)->flush |= flush;
907 csum = skb->csum;
908 skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
910 pp = ops->gro_receive(head, skb);
912 skb->csum = csum;
914 out_unlock:
915 rcu_read_unlock();
917 out:
918 NAPI_GRO_CB(skb)->flush |= flush;
920 return pp;
923 static int ipv6_gro_complete(struct sk_buff *skb)
925 const struct inet6_protocol *ops;
926 struct ipv6hdr *iph = ipv6_hdr(skb);
927 int err = -ENOSYS;
929 iph->payload_len = htons(skb->len - skb_network_offset(skb) -
930 sizeof(*iph));
932 rcu_read_lock();
933 ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
934 if (WARN_ON(!ops || !ops->gro_complete))
935 goto out_unlock;
937 err = ops->gro_complete(skb);
939 out_unlock:
940 rcu_read_unlock();
942 return err;
945 static struct packet_type ipv6_packet_type __read_mostly = {
946 .type = cpu_to_be16(ETH_P_IPV6),
947 .func = ipv6_rcv,
948 .gso_send_check = ipv6_gso_send_check,
949 .gso_segment = ipv6_gso_segment,
950 .gro_receive = ipv6_gro_receive,
951 .gro_complete = ipv6_gro_complete,
954 static int __init ipv6_packet_init(void)
956 dev_add_pack(&ipv6_packet_type);
957 return 0;
960 static void ipv6_packet_cleanup(void)
962 dev_remove_pack(&ipv6_packet_type);
965 static int __net_init ipv6_init_mibs(struct net *net)
967 if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
968 sizeof(struct udp_mib),
969 __alignof__(struct udp_mib)) < 0)
970 return -ENOMEM;
971 if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
972 sizeof(struct udp_mib),
973 __alignof__(struct udp_mib)) < 0)
974 goto err_udplite_mib;
975 if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
976 sizeof(struct ipstats_mib),
977 __alignof__(struct ipstats_mib)) < 0)
978 goto err_ip_mib;
979 if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
980 sizeof(struct icmpv6_mib),
981 __alignof__(struct icmpv6_mib)) < 0)
982 goto err_icmp_mib;
983 if (snmp_mib_init((void __percpu **)net->mib.icmpv6msg_statistics,
984 sizeof(struct icmpv6msg_mib),
985 __alignof__(struct icmpv6msg_mib)) < 0)
986 goto err_icmpmsg_mib;
987 return 0;
989 err_icmpmsg_mib:
990 snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
991 err_icmp_mib:
992 snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
993 err_ip_mib:
994 snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
995 err_udplite_mib:
996 snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
997 return -ENOMEM;
1000 static void ipv6_cleanup_mibs(struct net *net)
1002 snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1003 snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1004 snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1005 snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1006 snmp_mib_free((void __percpu **)net->mib.icmpv6msg_statistics);
1009 static int __net_init inet6_net_init(struct net *net)
1011 int err = 0;
1013 net->ipv6.sysctl.bindv6only = 0;
1014 net->ipv6.sysctl.icmpv6_time = 1*HZ;
1016 err = ipv6_init_mibs(net);
1017 if (err)
1018 return err;
1019 #ifdef CONFIG_PROC_FS
1020 err = udp6_proc_init(net);
1021 if (err)
1022 goto out;
1023 err = tcp6_proc_init(net);
1024 if (err)
1025 goto proc_tcp6_fail;
1026 err = ac6_proc_init(net);
1027 if (err)
1028 goto proc_ac6_fail;
1029 #endif
1030 return err;
1032 #ifdef CONFIG_PROC_FS
1033 proc_ac6_fail:
1034 tcp6_proc_exit(net);
1035 proc_tcp6_fail:
1036 udp6_proc_exit(net);
1037 out:
1038 ipv6_cleanup_mibs(net);
1039 return err;
1040 #endif
1043 static void __net_exit inet6_net_exit(struct net *net)
1045 #ifdef CONFIG_PROC_FS
1046 udp6_proc_exit(net);
1047 tcp6_proc_exit(net);
1048 ac6_proc_exit(net);
1049 #endif
1050 ipv6_cleanup_mibs(net);
1053 static struct pernet_operations inet6_net_ops = {
1054 .init = inet6_net_init,
1055 .exit = inet6_net_exit,
1058 static int __init inet6_init(void)
1060 struct sk_buff *dummy_skb;
1061 struct list_head *r;
1062 int err = 0;
1064 BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1066 /* Register the socket-side information for inet6_create. */
1067 for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1068 INIT_LIST_HEAD(r);
1070 if (disable_ipv6_mod) {
1071 printk(KERN_INFO
1072 "IPv6: Loaded, but administratively disabled, "
1073 "reboot required to enable\n");
1074 goto out;
1077 err = proto_register(&tcpv6_prot, 1);
1078 if (err)
1079 goto out;
1081 err = proto_register(&udpv6_prot, 1);
1082 if (err)
1083 goto out_unregister_tcp_proto;
1085 err = proto_register(&udplitev6_prot, 1);
1086 if (err)
1087 goto out_unregister_udp_proto;
1089 err = proto_register(&rawv6_prot, 1);
1090 if (err)
1091 goto out_unregister_udplite_proto;
1094 /* We MUST register RAW sockets before we create the ICMP6,
1095 * IGMP6, or NDISC control sockets.
1097 err = rawv6_init();
1098 if (err)
1099 goto out_unregister_raw_proto;
1101 /* Register the family here so that the init calls below will
1102 * be able to create sockets. (?? is this dangerous ??)
1104 err = sock_register(&inet6_family_ops);
1105 if (err)
1106 goto out_sock_register_fail;
1108 #ifdef CONFIG_SYSCTL
1109 err = ipv6_static_sysctl_register();
1110 if (err)
1111 goto static_sysctl_fail;
1112 #endif
1114 * ipngwg API draft makes clear that the correct semantics
1115 * for TCP and UDP is to consider one TCP and UDP instance
1116 * in a host available by both INET and INET6 APIs and
1117 * able to communicate via both network protocols.
1120 err = register_pernet_subsys(&inet6_net_ops);
1121 if (err)
1122 goto register_pernet_fail;
1123 err = icmpv6_init();
1124 if (err)
1125 goto icmp_fail;
1126 err = ip6_mr_init();
1127 if (err)
1128 goto ipmr_fail;
1129 err = ndisc_init();
1130 if (err)
1131 goto ndisc_fail;
1132 err = igmp6_init();
1133 if (err)
1134 goto igmp_fail;
1135 err = ipv6_netfilter_init();
1136 if (err)
1137 goto netfilter_fail;
1138 /* Create /proc/foo6 entries. */
1139 #ifdef CONFIG_PROC_FS
1140 err = -ENOMEM;
1141 if (raw6_proc_init())
1142 goto proc_raw6_fail;
1143 if (udplite6_proc_init())
1144 goto proc_udplite6_fail;
1145 if (ipv6_misc_proc_init())
1146 goto proc_misc6_fail;
1147 if (if6_proc_init())
1148 goto proc_if6_fail;
1149 #endif
1150 err = ip6_route_init();
1151 if (err)
1152 goto ip6_route_fail;
1153 err = ip6_flowlabel_init();
1154 if (err)
1155 goto ip6_flowlabel_fail;
1156 err = addrconf_init();
1157 if (err)
1158 goto addrconf_fail;
1160 /* Init v6 extension headers. */
1161 err = ipv6_exthdrs_init();
1162 if (err)
1163 goto ipv6_exthdrs_fail;
1165 err = ipv6_frag_init();
1166 if (err)
1167 goto ipv6_frag_fail;
1169 /* Init v6 transport protocols. */
1170 err = udpv6_init();
1171 if (err)
1172 goto udpv6_fail;
1174 err = udplitev6_init();
1175 if (err)
1176 goto udplitev6_fail;
1178 err = tcpv6_init();
1179 if (err)
1180 goto tcpv6_fail;
1182 err = ipv6_packet_init();
1183 if (err)
1184 goto ipv6_packet_fail;
1186 #ifdef CONFIG_SYSCTL
1187 err = ipv6_sysctl_register();
1188 if (err)
1189 goto sysctl_fail;
1190 #endif
1191 out:
1192 return err;
1194 #ifdef CONFIG_SYSCTL
1195 sysctl_fail:
1196 ipv6_packet_cleanup();
1197 #endif
1198 ipv6_packet_fail:
1199 tcpv6_exit();
1200 tcpv6_fail:
1201 udplitev6_exit();
1202 udplitev6_fail:
1203 udpv6_exit();
1204 udpv6_fail:
1205 ipv6_frag_exit();
1206 ipv6_frag_fail:
1207 ipv6_exthdrs_exit();
1208 ipv6_exthdrs_fail:
1209 addrconf_cleanup();
1210 addrconf_fail:
1211 ip6_flowlabel_cleanup();
1212 ip6_flowlabel_fail:
1213 ip6_route_cleanup();
1214 ip6_route_fail:
1215 #ifdef CONFIG_PROC_FS
1216 if6_proc_exit();
1217 proc_if6_fail:
1218 ipv6_misc_proc_exit();
1219 proc_misc6_fail:
1220 udplite6_proc_exit();
1221 proc_udplite6_fail:
1222 raw6_proc_exit();
1223 proc_raw6_fail:
1224 #endif
1225 ipv6_netfilter_fini();
1226 netfilter_fail:
1227 igmp6_cleanup();
1228 igmp_fail:
1229 ndisc_cleanup();
1230 ndisc_fail:
1231 ip6_mr_cleanup();
1232 ipmr_fail:
1233 icmpv6_cleanup();
1234 icmp_fail:
1235 unregister_pernet_subsys(&inet6_net_ops);
1236 register_pernet_fail:
1237 #ifdef CONFIG_SYSCTL
1238 ipv6_static_sysctl_unregister();
1239 static_sysctl_fail:
1240 #endif
1241 sock_unregister(PF_INET6);
1242 rtnl_unregister_all(PF_INET6);
1243 out_sock_register_fail:
1244 rawv6_exit();
1245 out_unregister_raw_proto:
1246 proto_unregister(&rawv6_prot);
1247 out_unregister_udplite_proto:
1248 proto_unregister(&udplitev6_prot);
1249 out_unregister_udp_proto:
1250 proto_unregister(&udpv6_prot);
1251 out_unregister_tcp_proto:
1252 proto_unregister(&tcpv6_prot);
1253 goto out;
1255 module_init(inet6_init);
1257 static void __exit inet6_exit(void)
1259 if (disable_ipv6_mod)
1260 return;
1262 /* First of all disallow new sockets creation. */
1263 sock_unregister(PF_INET6);
1264 /* Disallow any further netlink messages */
1265 rtnl_unregister_all(PF_INET6);
1267 #ifdef CONFIG_SYSCTL
1268 ipv6_sysctl_unregister();
1269 #endif
1270 udpv6_exit();
1271 udplitev6_exit();
1272 tcpv6_exit();
1274 /* Cleanup code parts. */
1275 ipv6_packet_cleanup();
1276 ipv6_frag_exit();
1277 ipv6_exthdrs_exit();
1278 addrconf_cleanup();
1279 ip6_flowlabel_cleanup();
1280 ip6_route_cleanup();
1281 #ifdef CONFIG_PROC_FS
1283 /* Cleanup code parts. */
1284 if6_proc_exit();
1285 ipv6_misc_proc_exit();
1286 udplite6_proc_exit();
1287 raw6_proc_exit();
1288 #endif
1289 ipv6_netfilter_fini();
1290 igmp6_cleanup();
1291 ndisc_cleanup();
1292 ip6_mr_cleanup();
1293 icmpv6_cleanup();
1294 rawv6_exit();
1296 unregister_pernet_subsys(&inet6_net_ops);
1297 #ifdef CONFIG_SYSCTL
1298 ipv6_static_sysctl_unregister();
1299 #endif
1300 proto_unregister(&rawv6_prot);
1301 proto_unregister(&udplitev6_prot);
1302 proto_unregister(&udpv6_prot);
1303 proto_unregister(&tcpv6_prot);
1305 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1307 module_exit(inet6_exit);
1309 MODULE_ALIAS_NETPROTO(PF_INET6);