Linux 5.1.15
[linux/fpc-iii.git] / net / ipv6 / af_inet6.c
blob2f45d2a3e3a33e7e43a3384d1a8fe7a683580919
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.
21 #define pr_fmt(fmt) "IPv6: " fmt
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/socket.h>
28 #include <linux/in.h>
29 #include <linux/kernel.h>
30 #include <linux/timer.h>
31 #include <linux/string.h>
32 #include <linux/sockios.h>
33 #include <linux/net.h>
34 #include <linux/fcntl.h>
35 #include <linux/mm.h>
36 #include <linux/interrupt.h>
37 #include <linux/proc_fs.h>
38 #include <linux/stat.h>
39 #include <linux/init.h>
40 #include <linux/slab.h>
42 #include <linux/inet.h>
43 #include <linux/netdevice.h>
44 #include <linux/icmpv6.h>
45 #include <linux/netfilter_ipv6.h>
47 #include <net/ip.h>
48 #include <net/ipv6.h>
49 #include <net/udp.h>
50 #include <net/udplite.h>
51 #include <net/tcp.h>
52 #include <net/ping.h>
53 #include <net/protocol.h>
54 #include <net/inet_common.h>
55 #include <net/route.h>
56 #include <net/transp_v6.h>
57 #include <net/ip6_route.h>
58 #include <net/addrconf.h>
59 #include <net/ndisc.h>
60 #ifdef CONFIG_IPV6_TUNNEL
61 #include <net/ip6_tunnel.h>
62 #endif
63 #include <net/calipso.h>
64 #include <net/seg6.h>
66 #include <linux/uaccess.h>
67 #include <linux/mroute6.h>
69 #include "ip6_offload.h"
71 MODULE_AUTHOR("Cast of dozens");
72 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
73 MODULE_LICENSE("GPL");
75 /* The inetsw6 table contains everything that inet6_create needs to
76 * build a new socket.
78 static struct list_head inetsw6[SOCK_MAX];
79 static DEFINE_SPINLOCK(inetsw6_lock);
81 struct ipv6_params ipv6_defaults = {
82 .disable_ipv6 = 0,
83 .autoconf = 1,
86 static int disable_ipv6_mod;
88 module_param_named(disable, disable_ipv6_mod, int, 0444);
89 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
91 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
92 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
94 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
95 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
97 bool ipv6_mod_enabled(void)
99 return disable_ipv6_mod == 0;
101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
103 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
105 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
107 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
110 static int inet6_create(struct net *net, struct socket *sock, int protocol,
111 int kern)
113 struct inet_sock *inet;
114 struct ipv6_pinfo *np;
115 struct sock *sk;
116 struct inet_protosw *answer;
117 struct proto *answer_prot;
118 unsigned char answer_flags;
119 int try_loading_module = 0;
120 int err;
122 if (protocol < 0 || protocol >= IPPROTO_MAX)
123 return -EINVAL;
125 /* Look for the requested type/protocol pair. */
126 lookup_protocol:
127 err = -ESOCKTNOSUPPORT;
128 rcu_read_lock();
129 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
131 err = 0;
132 /* Check the non-wild match. */
133 if (protocol == answer->protocol) {
134 if (protocol != IPPROTO_IP)
135 break;
136 } else {
137 /* Check for the two wild cases. */
138 if (IPPROTO_IP == protocol) {
139 protocol = answer->protocol;
140 break;
142 if (IPPROTO_IP == answer->protocol)
143 break;
145 err = -EPROTONOSUPPORT;
148 if (err) {
149 if (try_loading_module < 2) {
150 rcu_read_unlock();
152 * Be more specific, e.g. net-pf-10-proto-132-type-1
153 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
155 if (++try_loading_module == 1)
156 request_module("net-pf-%d-proto-%d-type-%d",
157 PF_INET6, protocol, sock->type);
159 * Fall back to generic, e.g. net-pf-10-proto-132
160 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
162 else
163 request_module("net-pf-%d-proto-%d",
164 PF_INET6, protocol);
165 goto lookup_protocol;
166 } else
167 goto out_rcu_unlock;
170 err = -EPERM;
171 if (sock->type == SOCK_RAW && !kern &&
172 !ns_capable(net->user_ns, CAP_NET_RAW))
173 goto out_rcu_unlock;
175 sock->ops = answer->ops;
176 answer_prot = answer->prot;
177 answer_flags = answer->flags;
178 rcu_read_unlock();
180 WARN_ON(!answer_prot->slab);
182 err = -ENOBUFS;
183 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
184 if (!sk)
185 goto out;
187 sock_init_data(sock, sk);
189 err = 0;
190 if (INET_PROTOSW_REUSE & answer_flags)
191 sk->sk_reuse = SK_CAN_REUSE;
193 inet = inet_sk(sk);
194 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
196 if (SOCK_RAW == sock->type) {
197 inet->inet_num = protocol;
198 if (IPPROTO_RAW == protocol)
199 inet->hdrincl = 1;
202 sk->sk_destruct = inet_sock_destruct;
203 sk->sk_family = PF_INET6;
204 sk->sk_protocol = protocol;
206 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
208 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
209 np->hop_limit = -1;
210 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
211 np->mc_loop = 1;
212 np->mc_all = 1;
213 np->pmtudisc = IPV6_PMTUDISC_WANT;
214 np->repflow = net->ipv6.sysctl.flowlabel_reflect;
215 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
217 /* Init the ipv4 part of the socket since we can have sockets
218 * using v6 API for ipv4.
220 inet->uc_ttl = -1;
222 inet->mc_loop = 1;
223 inet->mc_ttl = 1;
224 inet->mc_index = 0;
225 inet->mc_list = NULL;
226 inet->rcv_tos = 0;
228 if (net->ipv4.sysctl_ip_no_pmtu_disc)
229 inet->pmtudisc = IP_PMTUDISC_DONT;
230 else
231 inet->pmtudisc = IP_PMTUDISC_WANT;
233 * Increment only the relevant sk_prot->socks debug field, this changes
234 * the previous behaviour of incrementing both the equivalent to
235 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
237 * This allows better debug granularity as we'll know exactly how many
238 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
239 * transport protocol socks. -acme
241 sk_refcnt_debug_inc(sk);
243 if (inet->inet_num) {
244 /* It assumes that any protocol which allows
245 * the user to assign a number at socket
246 * creation time automatically shares.
248 inet->inet_sport = htons(inet->inet_num);
249 err = sk->sk_prot->hash(sk);
250 if (err) {
251 sk_common_release(sk);
252 goto out;
255 if (sk->sk_prot->init) {
256 err = sk->sk_prot->init(sk);
257 if (err) {
258 sk_common_release(sk);
259 goto out;
263 if (!kern) {
264 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
265 if (err) {
266 sk_common_release(sk);
267 goto out;
270 out:
271 return err;
272 out_rcu_unlock:
273 rcu_read_unlock();
274 goto out;
277 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
278 bool force_bind_address_no_port, bool with_lock)
280 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
281 struct inet_sock *inet = inet_sk(sk);
282 struct ipv6_pinfo *np = inet6_sk(sk);
283 struct net *net = sock_net(sk);
284 __be32 v4addr = 0;
285 unsigned short snum;
286 bool saved_ipv6only;
287 int addr_type = 0;
288 int err = 0;
290 if (addr->sin6_family != AF_INET6)
291 return -EAFNOSUPPORT;
293 addr_type = ipv6_addr_type(&addr->sin6_addr);
294 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
295 return -EINVAL;
297 snum = ntohs(addr->sin6_port);
298 if (snum && snum < inet_prot_sock(net) &&
299 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
300 return -EACCES;
302 if (with_lock)
303 lock_sock(sk);
305 /* Check these errors (active socket, double bind). */
306 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
307 err = -EINVAL;
308 goto out;
311 /* Check if the address belongs to the host. */
312 if (addr_type == IPV6_ADDR_MAPPED) {
313 struct net_device *dev = NULL;
314 int chk_addr_ret;
316 /* Binding to v4-mapped address on a v6-only socket
317 * makes no sense
319 if (sk->sk_ipv6only) {
320 err = -EINVAL;
321 goto out;
324 rcu_read_lock();
325 if (sk->sk_bound_dev_if) {
326 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
327 if (!dev) {
328 err = -ENODEV;
329 goto out_unlock;
333 /* Reproduce AF_INET checks to make the bindings consistent */
334 v4addr = addr->sin6_addr.s6_addr32[3];
335 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
336 rcu_read_unlock();
338 if (!inet_can_nonlocal_bind(net, inet) &&
339 v4addr != htonl(INADDR_ANY) &&
340 chk_addr_ret != RTN_LOCAL &&
341 chk_addr_ret != RTN_MULTICAST &&
342 chk_addr_ret != RTN_BROADCAST) {
343 err = -EADDRNOTAVAIL;
344 goto out;
346 } else {
347 if (addr_type != IPV6_ADDR_ANY) {
348 struct net_device *dev = NULL;
350 rcu_read_lock();
351 if (__ipv6_addr_needs_scope_id(addr_type)) {
352 if (addr_len >= sizeof(struct sockaddr_in6) &&
353 addr->sin6_scope_id) {
354 /* Override any existing binding, if another one
355 * is supplied by user.
357 sk->sk_bound_dev_if = addr->sin6_scope_id;
360 /* Binding to link-local address requires an interface */
361 if (!sk->sk_bound_dev_if) {
362 err = -EINVAL;
363 goto out_unlock;
367 if (sk->sk_bound_dev_if) {
368 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
369 if (!dev) {
370 err = -ENODEV;
371 goto out_unlock;
375 /* ipv4 addr of the socket is invalid. Only the
376 * unspecified and mapped address have a v4 equivalent.
378 v4addr = LOOPBACK4_IPV6;
379 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
380 if (!ipv6_can_nonlocal_bind(net, inet) &&
381 !ipv6_chk_addr(net, &addr->sin6_addr,
382 dev, 0)) {
383 err = -EADDRNOTAVAIL;
384 goto out_unlock;
387 rcu_read_unlock();
391 inet->inet_rcv_saddr = v4addr;
392 inet->inet_saddr = v4addr;
394 sk->sk_v6_rcv_saddr = addr->sin6_addr;
396 if (!(addr_type & IPV6_ADDR_MULTICAST))
397 np->saddr = addr->sin6_addr;
399 saved_ipv6only = sk->sk_ipv6only;
400 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
401 sk->sk_ipv6only = 1;
403 /* Make sure we are allowed to bind here. */
404 if (snum || !(inet->bind_address_no_port ||
405 force_bind_address_no_port)) {
406 if (sk->sk_prot->get_port(sk, snum)) {
407 sk->sk_ipv6only = saved_ipv6only;
408 inet_reset_saddr(sk);
409 err = -EADDRINUSE;
410 goto out;
412 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
413 if (err) {
414 sk->sk_ipv6only = saved_ipv6only;
415 inet_reset_saddr(sk);
416 goto out;
420 if (addr_type != IPV6_ADDR_ANY)
421 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
422 if (snum)
423 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
424 inet->inet_sport = htons(inet->inet_num);
425 inet->inet_dport = 0;
426 inet->inet_daddr = 0;
427 out:
428 if (with_lock)
429 release_sock(sk);
430 return err;
431 out_unlock:
432 rcu_read_unlock();
433 goto out;
436 /* bind for INET6 API */
437 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
439 struct sock *sk = sock->sk;
440 int err = 0;
442 /* If the socket has its own bind function then use it. */
443 if (sk->sk_prot->bind)
444 return sk->sk_prot->bind(sk, uaddr, addr_len);
446 if (addr_len < SIN6_LEN_RFC2133)
447 return -EINVAL;
449 /* BPF prog is run before any checks are done so that if the prog
450 * changes context in a wrong way it will be caught.
452 err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
453 if (err)
454 return err;
456 return __inet6_bind(sk, uaddr, addr_len, false, true);
458 EXPORT_SYMBOL(inet6_bind);
460 int inet6_release(struct socket *sock)
462 struct sock *sk = sock->sk;
464 if (!sk)
465 return -EINVAL;
467 /* Free mc lists */
468 ipv6_sock_mc_close(sk);
470 /* Free ac lists */
471 ipv6_sock_ac_close(sk);
473 return inet_release(sock);
475 EXPORT_SYMBOL(inet6_release);
477 void inet6_destroy_sock(struct sock *sk)
479 struct ipv6_pinfo *np = inet6_sk(sk);
480 struct sk_buff *skb;
481 struct ipv6_txoptions *opt;
483 /* Release rx options */
485 skb = xchg(&np->pktoptions, NULL);
486 kfree_skb(skb);
488 skb = xchg(&np->rxpmtu, NULL);
489 kfree_skb(skb);
491 /* Free flowlabels */
492 fl6_free_socklist(sk);
494 /* Free tx options */
496 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
497 if (opt) {
498 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
499 txopt_put(opt);
502 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
505 * This does both peername and sockname.
508 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
509 int peer)
511 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
512 struct sock *sk = sock->sk;
513 struct inet_sock *inet = inet_sk(sk);
514 struct ipv6_pinfo *np = inet6_sk(sk);
516 sin->sin6_family = AF_INET6;
517 sin->sin6_flowinfo = 0;
518 sin->sin6_scope_id = 0;
519 if (peer) {
520 if (!inet->inet_dport)
521 return -ENOTCONN;
522 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
523 peer == 1)
524 return -ENOTCONN;
525 sin->sin6_port = inet->inet_dport;
526 sin->sin6_addr = sk->sk_v6_daddr;
527 if (np->sndflow)
528 sin->sin6_flowinfo = np->flow_label;
529 } else {
530 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
531 sin->sin6_addr = np->saddr;
532 else
533 sin->sin6_addr = sk->sk_v6_rcv_saddr;
535 sin->sin6_port = inet->inet_sport;
537 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
538 sk->sk_bound_dev_if);
539 return sizeof(*sin);
541 EXPORT_SYMBOL(inet6_getname);
543 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
545 struct sock *sk = sock->sk;
546 struct net *net = sock_net(sk);
548 switch (cmd) {
549 case SIOCGSTAMP:
550 return sock_get_timestamp(sk, (struct timeval __user *)arg);
552 case SIOCGSTAMPNS:
553 return sock_get_timestampns(sk, (struct timespec __user *)arg);
555 case SIOCADDRT:
556 case SIOCDELRT:
558 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
560 case SIOCSIFADDR:
561 return addrconf_add_ifaddr(net, (void __user *) arg);
562 case SIOCDIFADDR:
563 return addrconf_del_ifaddr(net, (void __user *) arg);
564 case SIOCSIFDSTADDR:
565 return addrconf_set_dstaddr(net, (void __user *) arg);
566 default:
567 if (!sk->sk_prot->ioctl)
568 return -ENOIOCTLCMD;
569 return sk->sk_prot->ioctl(sk, cmd, arg);
571 /*NOTREACHED*/
572 return 0;
574 EXPORT_SYMBOL(inet6_ioctl);
576 const struct proto_ops inet6_stream_ops = {
577 .family = PF_INET6,
578 .owner = THIS_MODULE,
579 .release = inet6_release,
580 .bind = inet6_bind,
581 .connect = inet_stream_connect, /* ok */
582 .socketpair = sock_no_socketpair, /* a do nothing */
583 .accept = inet_accept, /* ok */
584 .getname = inet6_getname,
585 .poll = tcp_poll, /* ok */
586 .ioctl = inet6_ioctl, /* must change */
587 .listen = inet_listen, /* ok */
588 .shutdown = inet_shutdown, /* ok */
589 .setsockopt = sock_common_setsockopt, /* ok */
590 .getsockopt = sock_common_getsockopt, /* ok */
591 .sendmsg = inet_sendmsg, /* ok */
592 .recvmsg = inet_recvmsg, /* ok */
593 #ifdef CONFIG_MMU
594 .mmap = tcp_mmap,
595 #endif
596 .sendpage = inet_sendpage,
597 .sendmsg_locked = tcp_sendmsg_locked,
598 .sendpage_locked = tcp_sendpage_locked,
599 .splice_read = tcp_splice_read,
600 .read_sock = tcp_read_sock,
601 .peek_len = tcp_peek_len,
602 #ifdef CONFIG_COMPAT
603 .compat_setsockopt = compat_sock_common_setsockopt,
604 .compat_getsockopt = compat_sock_common_getsockopt,
605 #endif
606 .set_rcvlowat = tcp_set_rcvlowat,
609 const struct proto_ops inet6_dgram_ops = {
610 .family = PF_INET6,
611 .owner = THIS_MODULE,
612 .release = inet6_release,
613 .bind = inet6_bind,
614 .connect = inet_dgram_connect, /* ok */
615 .socketpair = sock_no_socketpair, /* a do nothing */
616 .accept = sock_no_accept, /* a do nothing */
617 .getname = inet6_getname,
618 .poll = udp_poll, /* ok */
619 .ioctl = inet6_ioctl, /* must change */
620 .listen = sock_no_listen, /* ok */
621 .shutdown = inet_shutdown, /* ok */
622 .setsockopt = sock_common_setsockopt, /* ok */
623 .getsockopt = sock_common_getsockopt, /* ok */
624 .sendmsg = inet_sendmsg, /* ok */
625 .recvmsg = inet_recvmsg, /* ok */
626 .mmap = sock_no_mmap,
627 .sendpage = sock_no_sendpage,
628 .set_peek_off = sk_set_peek_off,
629 #ifdef CONFIG_COMPAT
630 .compat_setsockopt = compat_sock_common_setsockopt,
631 .compat_getsockopt = compat_sock_common_getsockopt,
632 #endif
635 static const struct net_proto_family inet6_family_ops = {
636 .family = PF_INET6,
637 .create = inet6_create,
638 .owner = THIS_MODULE,
641 int inet6_register_protosw(struct inet_protosw *p)
643 struct list_head *lh;
644 struct inet_protosw *answer;
645 struct list_head *last_perm;
646 int protocol = p->protocol;
647 int ret;
649 spin_lock_bh(&inetsw6_lock);
651 ret = -EINVAL;
652 if (p->type >= SOCK_MAX)
653 goto out_illegal;
655 /* If we are trying to override a permanent protocol, bail. */
656 answer = NULL;
657 ret = -EPERM;
658 last_perm = &inetsw6[p->type];
659 list_for_each(lh, &inetsw6[p->type]) {
660 answer = list_entry(lh, struct inet_protosw, list);
662 /* Check only the non-wild match. */
663 if (INET_PROTOSW_PERMANENT & answer->flags) {
664 if (protocol == answer->protocol)
665 break;
666 last_perm = lh;
669 answer = NULL;
671 if (answer)
672 goto out_permanent;
674 /* Add the new entry after the last permanent entry if any, so that
675 * the new entry does not override a permanent entry when matched with
676 * a wild-card protocol. But it is allowed to override any existing
677 * non-permanent entry. This means that when we remove this entry, the
678 * system automatically returns to the old behavior.
680 list_add_rcu(&p->list, last_perm);
681 ret = 0;
682 out:
683 spin_unlock_bh(&inetsw6_lock);
684 return ret;
686 out_permanent:
687 pr_err("Attempt to override permanent protocol %d\n", protocol);
688 goto out;
690 out_illegal:
691 pr_err("Ignoring attempt to register invalid socket type %d\n",
692 p->type);
693 goto out;
695 EXPORT_SYMBOL(inet6_register_protosw);
697 void
698 inet6_unregister_protosw(struct inet_protosw *p)
700 if (INET_PROTOSW_PERMANENT & p->flags) {
701 pr_err("Attempt to unregister permanent protocol %d\n",
702 p->protocol);
703 } else {
704 spin_lock_bh(&inetsw6_lock);
705 list_del_rcu(&p->list);
706 spin_unlock_bh(&inetsw6_lock);
708 synchronize_net();
711 EXPORT_SYMBOL(inet6_unregister_protosw);
713 int inet6_sk_rebuild_header(struct sock *sk)
715 struct ipv6_pinfo *np = inet6_sk(sk);
716 struct dst_entry *dst;
718 dst = __sk_dst_check(sk, np->dst_cookie);
720 if (!dst) {
721 struct inet_sock *inet = inet_sk(sk);
722 struct in6_addr *final_p, final;
723 struct flowi6 fl6;
725 memset(&fl6, 0, sizeof(fl6));
726 fl6.flowi6_proto = sk->sk_protocol;
727 fl6.daddr = sk->sk_v6_daddr;
728 fl6.saddr = np->saddr;
729 fl6.flowlabel = np->flow_label;
730 fl6.flowi6_oif = sk->sk_bound_dev_if;
731 fl6.flowi6_mark = sk->sk_mark;
732 fl6.fl6_dport = inet->inet_dport;
733 fl6.fl6_sport = inet->inet_sport;
734 fl6.flowi6_uid = sk->sk_uid;
735 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
737 rcu_read_lock();
738 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
739 &final);
740 rcu_read_unlock();
742 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
743 if (IS_ERR(dst)) {
744 sk->sk_route_caps = 0;
745 sk->sk_err_soft = -PTR_ERR(dst);
746 return PTR_ERR(dst);
749 ip6_dst_store(sk, dst, NULL, NULL);
752 return 0;
754 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
756 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
757 const struct inet6_skb_parm *opt)
759 const struct ipv6_pinfo *np = inet6_sk(sk);
761 if (np->rxopt.all) {
762 if (((opt->flags & IP6SKB_HOPBYHOP) &&
763 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
764 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
765 np->rxopt.bits.rxflow) ||
766 (opt->srcrt && (np->rxopt.bits.srcrt ||
767 np->rxopt.bits.osrcrt)) ||
768 ((opt->dst1 || opt->dst0) &&
769 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
770 return true;
772 return false;
774 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
776 static struct packet_type ipv6_packet_type __read_mostly = {
777 .type = cpu_to_be16(ETH_P_IPV6),
778 .func = ipv6_rcv,
779 .list_func = ipv6_list_rcv,
782 static int __init ipv6_packet_init(void)
784 dev_add_pack(&ipv6_packet_type);
785 return 0;
788 static void ipv6_packet_cleanup(void)
790 dev_remove_pack(&ipv6_packet_type);
793 static int __net_init ipv6_init_mibs(struct net *net)
795 int i;
797 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
798 if (!net->mib.udp_stats_in6)
799 return -ENOMEM;
800 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
801 if (!net->mib.udplite_stats_in6)
802 goto err_udplite_mib;
803 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
804 if (!net->mib.ipv6_statistics)
805 goto err_ip_mib;
807 for_each_possible_cpu(i) {
808 struct ipstats_mib *af_inet6_stats;
809 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
810 u64_stats_init(&af_inet6_stats->syncp);
814 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
815 if (!net->mib.icmpv6_statistics)
816 goto err_icmp_mib;
817 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
818 GFP_KERNEL);
819 if (!net->mib.icmpv6msg_statistics)
820 goto err_icmpmsg_mib;
821 return 0;
823 err_icmpmsg_mib:
824 free_percpu(net->mib.icmpv6_statistics);
825 err_icmp_mib:
826 free_percpu(net->mib.ipv6_statistics);
827 err_ip_mib:
828 free_percpu(net->mib.udplite_stats_in6);
829 err_udplite_mib:
830 free_percpu(net->mib.udp_stats_in6);
831 return -ENOMEM;
834 static void ipv6_cleanup_mibs(struct net *net)
836 free_percpu(net->mib.udp_stats_in6);
837 free_percpu(net->mib.udplite_stats_in6);
838 free_percpu(net->mib.ipv6_statistics);
839 free_percpu(net->mib.icmpv6_statistics);
840 kfree(net->mib.icmpv6msg_statistics);
843 static int __net_init inet6_net_init(struct net *net)
845 int err = 0;
847 net->ipv6.sysctl.bindv6only = 0;
848 net->ipv6.sysctl.icmpv6_time = 1*HZ;
849 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
850 net->ipv6.sysctl.flowlabel_consistency = 1;
851 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
852 net->ipv6.sysctl.idgen_retries = 3;
853 net->ipv6.sysctl.idgen_delay = 1 * HZ;
854 net->ipv6.sysctl.flowlabel_state_ranges = 0;
855 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
856 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
857 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
858 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
859 atomic_set(&net->ipv6.fib6_sernum, 1);
861 err = ipv6_init_mibs(net);
862 if (err)
863 return err;
864 #ifdef CONFIG_PROC_FS
865 err = udp6_proc_init(net);
866 if (err)
867 goto out;
868 err = tcp6_proc_init(net);
869 if (err)
870 goto proc_tcp6_fail;
871 err = ac6_proc_init(net);
872 if (err)
873 goto proc_ac6_fail;
874 #endif
875 return err;
877 #ifdef CONFIG_PROC_FS
878 proc_ac6_fail:
879 tcp6_proc_exit(net);
880 proc_tcp6_fail:
881 udp6_proc_exit(net);
882 out:
883 ipv6_cleanup_mibs(net);
884 return err;
885 #endif
888 static void __net_exit inet6_net_exit(struct net *net)
890 #ifdef CONFIG_PROC_FS
891 udp6_proc_exit(net);
892 tcp6_proc_exit(net);
893 ac6_proc_exit(net);
894 #endif
895 ipv6_cleanup_mibs(net);
898 static struct pernet_operations inet6_net_ops = {
899 .init = inet6_net_init,
900 .exit = inet6_net_exit,
903 static int ipv6_route_input(struct sk_buff *skb)
905 ip6_route_input(skb);
906 return skb_dst(skb)->error;
909 static const struct ipv6_stub ipv6_stub_impl = {
910 .ipv6_sock_mc_join = ipv6_sock_mc_join,
911 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
912 .ipv6_dst_lookup = ip6_dst_lookup,
913 .ipv6_route_input = ipv6_route_input,
914 .fib6_get_table = fib6_get_table,
915 .fib6_table_lookup = fib6_table_lookup,
916 .fib6_lookup = fib6_lookup,
917 .fib6_multipath_select = fib6_multipath_select,
918 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
919 .udpv6_encap_enable = udpv6_encap_enable,
920 .ndisc_send_na = ndisc_send_na,
921 .nd_tbl = &nd_tbl,
924 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
925 .inet6_bind = __inet6_bind,
926 .udp6_lib_lookup = __udp6_lib_lookup,
929 static int __init inet6_init(void)
931 struct list_head *r;
932 int err = 0;
934 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
936 /* Register the socket-side information for inet6_create. */
937 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
938 INIT_LIST_HEAD(r);
940 if (disable_ipv6_mod) {
941 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
942 goto out;
945 err = proto_register(&tcpv6_prot, 1);
946 if (err)
947 goto out;
949 err = proto_register(&udpv6_prot, 1);
950 if (err)
951 goto out_unregister_tcp_proto;
953 err = proto_register(&udplitev6_prot, 1);
954 if (err)
955 goto out_unregister_udp_proto;
957 err = proto_register(&rawv6_prot, 1);
958 if (err)
959 goto out_unregister_udplite_proto;
961 err = proto_register(&pingv6_prot, 1);
962 if (err)
963 goto out_unregister_raw_proto;
965 /* We MUST register RAW sockets before we create the ICMP6,
966 * IGMP6, or NDISC control sockets.
968 err = rawv6_init();
969 if (err)
970 goto out_unregister_ping_proto;
972 /* Register the family here so that the init calls below will
973 * be able to create sockets. (?? is this dangerous ??)
975 err = sock_register(&inet6_family_ops);
976 if (err)
977 goto out_sock_register_fail;
980 * ipngwg API draft makes clear that the correct semantics
981 * for TCP and UDP is to consider one TCP and UDP instance
982 * in a host available by both INET and INET6 APIs and
983 * able to communicate via both network protocols.
986 err = register_pernet_subsys(&inet6_net_ops);
987 if (err)
988 goto register_pernet_fail;
989 err = ip6_mr_init();
990 if (err)
991 goto ipmr_fail;
992 err = icmpv6_init();
993 if (err)
994 goto icmp_fail;
995 err = ndisc_init();
996 if (err)
997 goto ndisc_fail;
998 err = igmp6_init();
999 if (err)
1000 goto igmp_fail;
1002 err = ipv6_netfilter_init();
1003 if (err)
1004 goto netfilter_fail;
1005 /* Create /proc/foo6 entries. */
1006 #ifdef CONFIG_PROC_FS
1007 err = -ENOMEM;
1008 if (raw6_proc_init())
1009 goto proc_raw6_fail;
1010 if (udplite6_proc_init())
1011 goto proc_udplite6_fail;
1012 if (ipv6_misc_proc_init())
1013 goto proc_misc6_fail;
1014 if (if6_proc_init())
1015 goto proc_if6_fail;
1016 #endif
1017 err = ip6_route_init();
1018 if (err)
1019 goto ip6_route_fail;
1020 err = ndisc_late_init();
1021 if (err)
1022 goto ndisc_late_fail;
1023 err = ip6_flowlabel_init();
1024 if (err)
1025 goto ip6_flowlabel_fail;
1026 err = ipv6_anycast_init();
1027 if (err)
1028 goto ipv6_anycast_fail;
1029 err = addrconf_init();
1030 if (err)
1031 goto addrconf_fail;
1033 /* Init v6 extension headers. */
1034 err = ipv6_exthdrs_init();
1035 if (err)
1036 goto ipv6_exthdrs_fail;
1038 err = ipv6_frag_init();
1039 if (err)
1040 goto ipv6_frag_fail;
1042 /* Init v6 transport protocols. */
1043 err = udpv6_init();
1044 if (err)
1045 goto udpv6_fail;
1047 err = udplitev6_init();
1048 if (err)
1049 goto udplitev6_fail;
1051 err = udpv6_offload_init();
1052 if (err)
1053 goto udpv6_offload_fail;
1055 err = tcpv6_init();
1056 if (err)
1057 goto tcpv6_fail;
1059 err = ipv6_packet_init();
1060 if (err)
1061 goto ipv6_packet_fail;
1063 err = pingv6_init();
1064 if (err)
1065 goto pingv6_fail;
1067 err = calipso_init();
1068 if (err)
1069 goto calipso_fail;
1071 err = seg6_init();
1072 if (err)
1073 goto seg6_fail;
1075 err = igmp6_late_init();
1076 if (err)
1077 goto igmp6_late_err;
1079 #ifdef CONFIG_SYSCTL
1080 err = ipv6_sysctl_register();
1081 if (err)
1082 goto sysctl_fail;
1083 #endif
1085 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1086 wmb();
1087 ipv6_stub = &ipv6_stub_impl;
1088 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1089 out:
1090 return err;
1092 #ifdef CONFIG_SYSCTL
1093 sysctl_fail:
1094 igmp6_late_cleanup();
1095 #endif
1096 igmp6_late_err:
1097 seg6_exit();
1098 seg6_fail:
1099 calipso_exit();
1100 calipso_fail:
1101 pingv6_exit();
1102 pingv6_fail:
1103 ipv6_packet_cleanup();
1104 ipv6_packet_fail:
1105 tcpv6_exit();
1106 tcpv6_fail:
1107 udpv6_offload_exit();
1108 udpv6_offload_fail:
1109 udplitev6_exit();
1110 udplitev6_fail:
1111 udpv6_exit();
1112 udpv6_fail:
1113 ipv6_frag_exit();
1114 ipv6_frag_fail:
1115 ipv6_exthdrs_exit();
1116 ipv6_exthdrs_fail:
1117 addrconf_cleanup();
1118 addrconf_fail:
1119 ipv6_anycast_cleanup();
1120 ipv6_anycast_fail:
1121 ip6_flowlabel_cleanup();
1122 ip6_flowlabel_fail:
1123 ndisc_late_cleanup();
1124 ndisc_late_fail:
1125 ip6_route_cleanup();
1126 ip6_route_fail:
1127 #ifdef CONFIG_PROC_FS
1128 if6_proc_exit();
1129 proc_if6_fail:
1130 ipv6_misc_proc_exit();
1131 proc_misc6_fail:
1132 udplite6_proc_exit();
1133 proc_udplite6_fail:
1134 raw6_proc_exit();
1135 proc_raw6_fail:
1136 #endif
1137 ipv6_netfilter_fini();
1138 netfilter_fail:
1139 igmp6_cleanup();
1140 igmp_fail:
1141 ndisc_cleanup();
1142 ndisc_fail:
1143 icmpv6_cleanup();
1144 icmp_fail:
1145 ip6_mr_cleanup();
1146 ipmr_fail:
1147 unregister_pernet_subsys(&inet6_net_ops);
1148 register_pernet_fail:
1149 sock_unregister(PF_INET6);
1150 rtnl_unregister_all(PF_INET6);
1151 out_sock_register_fail:
1152 rawv6_exit();
1153 out_unregister_ping_proto:
1154 proto_unregister(&pingv6_prot);
1155 out_unregister_raw_proto:
1156 proto_unregister(&rawv6_prot);
1157 out_unregister_udplite_proto:
1158 proto_unregister(&udplitev6_prot);
1159 out_unregister_udp_proto:
1160 proto_unregister(&udpv6_prot);
1161 out_unregister_tcp_proto:
1162 proto_unregister(&tcpv6_prot);
1163 goto out;
1165 module_init(inet6_init);
1167 MODULE_ALIAS_NETPROTO(PF_INET6);