Loading...
1/*
2 * PF_INET6 socket protocol family
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Adapted from linux/net/ipv4/af_inet.c
9 *
10 * Fixes:
11 * piggy, Karl Knutson : Socket protocol table
12 * Hideaki YOSHIFUJI : sin6_scope_id support
13 * Arnaldo Melo : check proc_net_create return, cleanups
14 *
15 * This program is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU General Public License
17 * as published by the Free Software Foundation; either version
18 * 2 of the License, or (at your option) any later version.
19 */
20
21
22#include <linux/module.h>
23#include <linux/capability.h>
24#include <linux/errno.h>
25#include <linux/types.h>
26#include <linux/socket.h>
27#include <linux/in.h>
28#include <linux/kernel.h>
29#include <linux/timer.h>
30#include <linux/string.h>
31#include <linux/sockios.h>
32#include <linux/net.h>
33#include <linux/fcntl.h>
34#include <linux/mm.h>
35#include <linux/interrupt.h>
36#include <linux/proc_fs.h>
37#include <linux/stat.h>
38#include <linux/init.h>
39#include <linux/slab.h>
40
41#include <linux/inet.h>
42#include <linux/netdevice.h>
43#include <linux/icmpv6.h>
44#include <linux/netfilter_ipv6.h>
45
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
61
62#include <asm/uaccess.h>
63#include <asm/system.h>
64#include <linux/mroute6.h>
65
66MODULE_AUTHOR("Cast of dozens");
67MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
68MODULE_LICENSE("GPL");
69
70/* The inetsw6 table contains everything that inet6_create needs to
71 * build a new socket.
72 */
73static struct list_head inetsw6[SOCK_MAX];
74static DEFINE_SPINLOCK(inetsw6_lock);
75
76struct ipv6_params ipv6_defaults = {
77 .disable_ipv6 = 0,
78 .autoconf = 1,
79};
80
81static int disable_ipv6_mod = 0;
82
83module_param_named(disable, disable_ipv6_mod, int, 0444);
84MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
85
86module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
87MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
88
89module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
90MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
91
92static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
93{
94 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
95
96 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
97}
98
99static int inet6_create(struct net *net, struct socket *sock, int protocol,
100 int kern)
101{
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;
111
112 if (sock->type != SOCK_RAW &&
113 sock->type != SOCK_DGRAM &&
114 !inet_ehash_secret)
115 build_ehash_secret();
116
117 /* Look for the requested type/protocol pair. */
118lookup_protocol:
119 err = -ESOCKTNOSUPPORT;
120 rcu_read_lock();
121 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
122
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;
133 }
134 if (IPPROTO_IP == answer->protocol)
135 break;
136 }
137 err = -EPROTONOSUPPORT;
138 }
139
140 if (err) {
141 if (try_loading_module < 2) {
142 rcu_read_unlock();
143 /*
144 * Be more specific, e.g. net-pf-10-proto-132-type-1
145 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
146 */
147 if (++try_loading_module == 1)
148 request_module("net-pf-%d-proto-%d-type-%d",
149 PF_INET6, protocol, sock->type);
150 /*
151 * Fall back to generic, e.g. net-pf-10-proto-132
152 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
153 */
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;
160 }
161
162 err = -EPERM;
163 if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
164 goto out_rcu_unlock;
165
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();
171
172 WARN_ON(answer_prot->slab == NULL);
173
174 err = -ENOBUFS;
175 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
176 if (sk == NULL)
177 goto out;
178
179 sock_init_data(sock, sk);
180
181 err = 0;
182 sk->sk_no_check = answer_no_check;
183 if (INET_PROTOSW_REUSE & answer_flags)
184 sk->sk_reuse = 1;
185
186 inet = inet_sk(sk);
187 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
188
189 if (SOCK_RAW == sock->type) {
190 inet->inet_num = protocol;
191 if (IPPROTO_RAW == protocol)
192 inet->hdrincl = 1;
193 }
194
195 sk->sk_destruct = inet_sock_destruct;
196 sk->sk_family = PF_INET6;
197 sk->sk_protocol = protocol;
198
199 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
200
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;
207
208 /* Init the ipv4 part of the socket since we can have sockets
209 * using v6 API for ipv4.
210 */
211 inet->uc_ttl = -1;
212
213 inet->mc_loop = 1;
214 inet->mc_ttl = 1;
215 inet->mc_index = 0;
216 inet->mc_list = NULL;
217
218 if (ipv4_config.no_pmtu_disc)
219 inet->pmtudisc = IP_PMTUDISC_DONT;
220 else
221 inet->pmtudisc = IP_PMTUDISC_WANT;
222 /*
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.
226 *
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
230 */
231 sk_refcnt_debug_inc(sk);
232
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.
237 */
238 inet->inet_sport = htons(inet->inet_num);
239 sk->sk_prot->hash(sk);
240 }
241 if (sk->sk_prot->init) {
242 err = sk->sk_prot->init(sk);
243 if (err) {
244 sk_common_release(sk);
245 goto out;
246 }
247 }
248out:
249 return err;
250out_rcu_unlock:
251 rcu_read_unlock();
252 goto out;
253}
254
255
256/* bind for INET6 API */
257int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
258{
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;
268
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);
272
273 if (addr_len < SIN6_LEN_RFC2133)
274 return -EINVAL;
275
276 if (addr->sin6_family != AF_INET6)
277 return -EAFNOSUPPORT;
278
279 addr_type = ipv6_addr_type(&addr->sin6_addr);
280 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
281 return -EINVAL;
282
283 snum = ntohs(addr->sin6_port);
284 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
285 return -EACCES;
286
287 lock_sock(sk);
288
289 /* Check these errors (active socket, double bind). */
290 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
291 err = -EINVAL;
292 goto out;
293 }
294
295 /* Check if the address belongs to the host. */
296 if (addr_type == IPV6_ADDR_MAPPED) {
297 int chk_addr_ret;
298
299 /* Binding to v4-mapped address on a v6-only socket
300 * makes no sense
301 */
302 if (np->ipv6only) {
303 err = -EINVAL;
304 goto out;
305 }
306
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;
318 }
319 } else {
320 if (addr_type != IPV6_ADDR_ANY) {
321 struct net_device *dev = NULL;
322
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.
329 */
330 sk->sk_bound_dev_if = addr->sin6_scope_id;
331 }
332
333 /* Binding to link-local address requires an interface */
334 if (!sk->sk_bound_dev_if) {
335 err = -EINVAL;
336 goto out_unlock;
337 }
338 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
339 if (!dev) {
340 err = -ENODEV;
341 goto out_unlock;
342 }
343 }
344
345 /* ipv4 addr of the socket is invalid. Only the
346 * unspecified and mapped address have a v4 equivalent.
347 */
348 v4addr = LOOPBACK4_IPV6;
349 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
350 if (!inet->transparent &&
351 !ipv6_chk_addr(net, &addr->sin6_addr,
352 dev, 0)) {
353 err = -EADDRNOTAVAIL;
354 goto out_unlock;
355 }
356 }
357 rcu_read_unlock();
358 }
359 }
360
361 inet->inet_rcv_saddr = v4addr;
362 inet->inet_saddr = v4addr;
363
364 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
365
366 if (!(addr_type & IPV6_ADDR_MULTICAST))
367 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
368
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;
374 }
375
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;
380 }
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;
386out:
387 release_sock(sk);
388 return err;
389out_unlock:
390 rcu_read_unlock();
391 goto out;
392}
393
394EXPORT_SYMBOL(inet6_bind);
395
396int inet6_release(struct socket *sock)
397{
398 struct sock *sk = sock->sk;
399
400 if (sk == NULL)
401 return -EINVAL;
402
403 /* Free mc lists */
404 ipv6_sock_mc_close(sk);
405
406 /* Free ac lists */
407 ipv6_sock_ac_close(sk);
408
409 return inet_release(sock);
410}
411
412EXPORT_SYMBOL(inet6_release);
413
414void inet6_destroy_sock(struct sock *sk)
415{
416 struct ipv6_pinfo *np = inet6_sk(sk);
417 struct sk_buff *skb;
418 struct ipv6_txoptions *opt;
419
420 /* Release rx options */
421
422 if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
423 kfree_skb(skb);
424
425 if ((skb = xchg(&np->rxpmtu, NULL)) != NULL)
426 kfree_skb(skb);
427
428 /* Free flowlabels */
429 fl6_free_socklist(sk);
430
431 /* Free tx options */
432
433 if ((opt = xchg(&np->opt, NULL)) != NULL)
434 sock_kfree_s(sk, opt, opt->tot_len);
435}
436
437EXPORT_SYMBOL_GPL(inet6_destroy_sock);
438
439/*
440 * This does both peername and sockname.
441 */
442
443int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
444 int *uaddr_len, int peer)
445{
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);
450
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 ipv6_addr_copy(&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 ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
467 else
468 ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
469
470 sin->sin6_port = inet->inet_sport;
471 }
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;
476}
477
478EXPORT_SYMBOL(inet6_getname);
479
480int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
481{
482 struct sock *sk = sock->sk;
483 struct net *net = sock_net(sk);
484
485 switch(cmd)
486 {
487 case SIOCGSTAMP:
488 return sock_get_timestamp(sk, (struct timeval __user *)arg);
489
490 case SIOCGSTAMPNS:
491 return sock_get_timestampns(sk, (struct timespec __user *)arg);
492
493 case SIOCADDRT:
494 case SIOCDELRT:
495
496 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
497
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);
508 }
509 /*NOTREACHED*/
510 return 0;
511}
512
513EXPORT_SYMBOL(inet6_ioctl);
514
515const 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
539};
540
541const 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
564};
565
566static const struct net_proto_family inet6_family_ops = {
567 .family = PF_INET6,
568 .create = inet6_create,
569 .owner = THIS_MODULE,
570};
571
572int inet6_register_protosw(struct inet_protosw *p)
573{
574 struct list_head *lh;
575 struct inet_protosw *answer;
576 struct list_head *last_perm;
577 int protocol = p->protocol;
578 int ret;
579
580 spin_lock_bh(&inetsw6_lock);
581
582 ret = -EINVAL;
583 if (p->type >= SOCK_MAX)
584 goto out_illegal;
585
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);
592
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;
598 }
599
600 answer = NULL;
601 }
602 if (answer)
603 goto out_permanent;
604
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.
610 */
611 list_add_rcu(&p->list, last_perm);
612 ret = 0;
613out:
614 spin_unlock_bh(&inetsw6_lock);
615 return ret;
616
617out_permanent:
618 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
619 protocol);
620 goto out;
621
622out_illegal:
623 printk(KERN_ERR
624 "Ignoring attempt to register invalid socket type %d.\n",
625 p->type);
626 goto out;
627}
628
629EXPORT_SYMBOL(inet6_register_protosw);
630
631void
632inet6_unregister_protosw(struct inet_protosw *p)
633{
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);
642
643 synchronize_net();
644 }
645}
646
647EXPORT_SYMBOL(inet6_unregister_protosw);
648
649int inet6_sk_rebuild_header(struct sock *sk)
650{
651 struct ipv6_pinfo *np = inet6_sk(sk);
652 struct dst_entry *dst;
653
654 dst = __sk_dst_check(sk, np->dst_cookie);
655
656 if (dst == NULL) {
657 struct inet_sock *inet = inet_sk(sk);
658 struct in6_addr *final_p, final;
659 struct flowi6 fl6;
660
661 memset(&fl6, 0, sizeof(fl6));
662 fl6.flowi6_proto = sk->sk_protocol;
663 ipv6_addr_copy(&fl6.daddr, &np->daddr);
664 ipv6_addr_copy(&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));
671
672 final_p = fl6_update_dst(&fl6, np->opt, &final);
673
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);
679 }
680
681 __ip6_dst_store(sk, dst, NULL, NULL);
682 }
683
684 return 0;
685}
686
687EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
688
689int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
690{
691 struct ipv6_pinfo *np = inet6_sk(sk);
692 struct inet6_skb_parm *opt = IP6CB(skb);
693
694 if (np->rxopt.all) {
695 if ((opt->hop && (np->rxopt.bits.hopopts ||
696 np->rxopt.bits.ohopopts)) ||
697 ((IPV6_FLOWINFO_MASK &
698 *(__be32 *)skb_network_header(skb)) &&
699 np->rxopt.bits.rxflow) ||
700 (opt->srcrt && (np->rxopt.bits.srcrt ||
701 np->rxopt.bits.osrcrt)) ||
702 ((opt->dst1 || opt->dst0) &&
703 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
704 return 1;
705 }
706 return 0;
707}
708
709EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
710
711static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
712{
713 const struct inet6_protocol *ops = NULL;
714
715 for (;;) {
716 struct ipv6_opt_hdr *opth;
717 int len;
718
719 if (proto != NEXTHDR_HOP) {
720 ops = rcu_dereference(inet6_protos[proto]);
721
722 if (unlikely(!ops))
723 break;
724
725 if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
726 break;
727 }
728
729 if (unlikely(!pskb_may_pull(skb, 8)))
730 break;
731
732 opth = (void *)skb->data;
733 len = ipv6_optlen(opth);
734
735 if (unlikely(!pskb_may_pull(skb, len)))
736 break;
737
738 proto = opth->nexthdr;
739 __skb_pull(skb, len);
740 }
741
742 return proto;
743}
744
745static int ipv6_gso_send_check(struct sk_buff *skb)
746{
747 const struct ipv6hdr *ipv6h;
748 const struct inet6_protocol *ops;
749 int err = -EINVAL;
750
751 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
752 goto out;
753
754 ipv6h = ipv6_hdr(skb);
755 __skb_pull(skb, sizeof(*ipv6h));
756 err = -EPROTONOSUPPORT;
757
758 rcu_read_lock();
759 ops = rcu_dereference(inet6_protos[
760 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
761
762 if (likely(ops && ops->gso_send_check)) {
763 skb_reset_transport_header(skb);
764 err = ops->gso_send_check(skb);
765 }
766 rcu_read_unlock();
767
768out:
769 return err;
770}
771
772static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, u32 features)
773{
774 struct sk_buff *segs = ERR_PTR(-EINVAL);
775 struct ipv6hdr *ipv6h;
776 const struct inet6_protocol *ops;
777 int proto;
778 struct frag_hdr *fptr;
779 unsigned int unfrag_ip6hlen;
780 u8 *prevhdr;
781 int offset = 0;
782
783 if (!(features & NETIF_F_V6_CSUM))
784 features &= ~NETIF_F_SG;
785
786 if (unlikely(skb_shinfo(skb)->gso_type &
787 ~(SKB_GSO_UDP |
788 SKB_GSO_DODGY |
789 SKB_GSO_TCP_ECN |
790 SKB_GSO_TCPV6 |
791 0)))
792 goto out;
793
794 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
795 goto out;
796
797 ipv6h = ipv6_hdr(skb);
798 __skb_pull(skb, sizeof(*ipv6h));
799 segs = ERR_PTR(-EPROTONOSUPPORT);
800
801 proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
802 rcu_read_lock();
803 ops = rcu_dereference(inet6_protos[proto]);
804 if (likely(ops && ops->gso_segment)) {
805 skb_reset_transport_header(skb);
806 segs = ops->gso_segment(skb, features);
807 }
808 rcu_read_unlock();
809
810 if (IS_ERR(segs))
811 goto out;
812
813 for (skb = segs; skb; skb = skb->next) {
814 ipv6h = ipv6_hdr(skb);
815 ipv6h->payload_len = htons(skb->len - skb->mac_len -
816 sizeof(*ipv6h));
817 if (proto == IPPROTO_UDP) {
818 unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
819 fptr = (struct frag_hdr *)(skb_network_header(skb) +
820 unfrag_ip6hlen);
821 fptr->frag_off = htons(offset);
822 if (skb->next != NULL)
823 fptr->frag_off |= htons(IP6_MF);
824 offset += (ntohs(ipv6h->payload_len) -
825 sizeof(struct frag_hdr));
826 }
827 }
828
829out:
830 return segs;
831}
832
833struct ipv6_gro_cb {
834 struct napi_gro_cb napi;
835 int proto;
836};
837
838#define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
839
840static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
841 struct sk_buff *skb)
842{
843 const struct inet6_protocol *ops;
844 struct sk_buff **pp = NULL;
845 struct sk_buff *p;
846 struct ipv6hdr *iph;
847 unsigned int nlen;
848 unsigned int hlen;
849 unsigned int off;
850 int flush = 1;
851 int proto;
852 __wsum csum;
853
854 off = skb_gro_offset(skb);
855 hlen = off + sizeof(*iph);
856 iph = skb_gro_header_fast(skb, off);
857 if (skb_gro_header_hard(skb, hlen)) {
858 iph = skb_gro_header_slow(skb, hlen, off);
859 if (unlikely(!iph))
860 goto out;
861 }
862
863 skb_gro_pull(skb, sizeof(*iph));
864 skb_set_transport_header(skb, skb_gro_offset(skb));
865
866 flush += ntohs(iph->payload_len) != skb_gro_len(skb);
867
868 rcu_read_lock();
869 proto = iph->nexthdr;
870 ops = rcu_dereference(inet6_protos[proto]);
871 if (!ops || !ops->gro_receive) {
872 __pskb_pull(skb, skb_gro_offset(skb));
873 proto = ipv6_gso_pull_exthdrs(skb, proto);
874 skb_gro_pull(skb, -skb_transport_offset(skb));
875 skb_reset_transport_header(skb);
876 __skb_push(skb, skb_gro_offset(skb));
877
878 ops = rcu_dereference(inet6_protos[proto]);
879 if (!ops || !ops->gro_receive)
880 goto out_unlock;
881
882 iph = ipv6_hdr(skb);
883 }
884
885 IPV6_GRO_CB(skb)->proto = proto;
886
887 flush--;
888 nlen = skb_network_header_len(skb);
889
890 for (p = *head; p; p = p->next) {
891 struct ipv6hdr *iph2;
892
893 if (!NAPI_GRO_CB(p)->same_flow)
894 continue;
895
896 iph2 = ipv6_hdr(p);
897
898 /* All fields must match except length. */
899 if (nlen != skb_network_header_len(p) ||
900 memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
901 memcmp(&iph->nexthdr, &iph2->nexthdr,
902 nlen - offsetof(struct ipv6hdr, nexthdr))) {
903 NAPI_GRO_CB(p)->same_flow = 0;
904 continue;
905 }
906
907 NAPI_GRO_CB(p)->flush |= flush;
908 }
909
910 NAPI_GRO_CB(skb)->flush |= flush;
911
912 csum = skb->csum;
913 skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
914
915 pp = ops->gro_receive(head, skb);
916
917 skb->csum = csum;
918
919out_unlock:
920 rcu_read_unlock();
921
922out:
923 NAPI_GRO_CB(skb)->flush |= flush;
924
925 return pp;
926}
927
928static int ipv6_gro_complete(struct sk_buff *skb)
929{
930 const struct inet6_protocol *ops;
931 struct ipv6hdr *iph = ipv6_hdr(skb);
932 int err = -ENOSYS;
933
934 iph->payload_len = htons(skb->len - skb_network_offset(skb) -
935 sizeof(*iph));
936
937 rcu_read_lock();
938 ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
939 if (WARN_ON(!ops || !ops->gro_complete))
940 goto out_unlock;
941
942 err = ops->gro_complete(skb);
943
944out_unlock:
945 rcu_read_unlock();
946
947 return err;
948}
949
950static struct packet_type ipv6_packet_type __read_mostly = {
951 .type = cpu_to_be16(ETH_P_IPV6),
952 .func = ipv6_rcv,
953 .gso_send_check = ipv6_gso_send_check,
954 .gso_segment = ipv6_gso_segment,
955 .gro_receive = ipv6_gro_receive,
956 .gro_complete = ipv6_gro_complete,
957};
958
959static int __init ipv6_packet_init(void)
960{
961 dev_add_pack(&ipv6_packet_type);
962 return 0;
963}
964
965static void ipv6_packet_cleanup(void)
966{
967 dev_remove_pack(&ipv6_packet_type);
968}
969
970static int __net_init ipv6_init_mibs(struct net *net)
971{
972 if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
973 sizeof(struct udp_mib),
974 __alignof__(struct udp_mib)) < 0)
975 return -ENOMEM;
976 if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
977 sizeof(struct udp_mib),
978 __alignof__(struct udp_mib)) < 0)
979 goto err_udplite_mib;
980 if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
981 sizeof(struct ipstats_mib),
982 __alignof__(struct ipstats_mib)) < 0)
983 goto err_ip_mib;
984 if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
985 sizeof(struct icmpv6_mib),
986 __alignof__(struct icmpv6_mib)) < 0)
987 goto err_icmp_mib;
988 if (snmp_mib_init((void __percpu **)net->mib.icmpv6msg_statistics,
989 sizeof(struct icmpv6msg_mib),
990 __alignof__(struct icmpv6msg_mib)) < 0)
991 goto err_icmpmsg_mib;
992 return 0;
993
994err_icmpmsg_mib:
995 snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
996err_icmp_mib:
997 snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
998err_ip_mib:
999 snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1000err_udplite_mib:
1001 snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1002 return -ENOMEM;
1003}
1004
1005static void ipv6_cleanup_mibs(struct net *net)
1006{
1007 snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1008 snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1009 snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1010 snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1011 snmp_mib_free((void __percpu **)net->mib.icmpv6msg_statistics);
1012}
1013
1014static int __net_init inet6_net_init(struct net *net)
1015{
1016 int err = 0;
1017
1018 net->ipv6.sysctl.bindv6only = 0;
1019 net->ipv6.sysctl.icmpv6_time = 1*HZ;
1020
1021 err = ipv6_init_mibs(net);
1022 if (err)
1023 return err;
1024#ifdef CONFIG_PROC_FS
1025 err = udp6_proc_init(net);
1026 if (err)
1027 goto out;
1028 err = tcp6_proc_init(net);
1029 if (err)
1030 goto proc_tcp6_fail;
1031 err = ac6_proc_init(net);
1032 if (err)
1033 goto proc_ac6_fail;
1034#endif
1035 return err;
1036
1037#ifdef CONFIG_PROC_FS
1038proc_ac6_fail:
1039 tcp6_proc_exit(net);
1040proc_tcp6_fail:
1041 udp6_proc_exit(net);
1042out:
1043 ipv6_cleanup_mibs(net);
1044 return err;
1045#endif
1046}
1047
1048static void __net_exit inet6_net_exit(struct net *net)
1049{
1050#ifdef CONFIG_PROC_FS
1051 udp6_proc_exit(net);
1052 tcp6_proc_exit(net);
1053 ac6_proc_exit(net);
1054#endif
1055 ipv6_cleanup_mibs(net);
1056}
1057
1058static struct pernet_operations inet6_net_ops = {
1059 .init = inet6_net_init,
1060 .exit = inet6_net_exit,
1061};
1062
1063static int __init inet6_init(void)
1064{
1065 struct sk_buff *dummy_skb;
1066 struct list_head *r;
1067 int err = 0;
1068
1069 BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1070
1071 /* Register the socket-side information for inet6_create. */
1072 for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1073 INIT_LIST_HEAD(r);
1074
1075 if (disable_ipv6_mod) {
1076 printk(KERN_INFO
1077 "IPv6: Loaded, but administratively disabled, "
1078 "reboot required to enable\n");
1079 goto out;
1080 }
1081
1082 err = proto_register(&tcpv6_prot, 1);
1083 if (err)
1084 goto out;
1085
1086 err = proto_register(&udpv6_prot, 1);
1087 if (err)
1088 goto out_unregister_tcp_proto;
1089
1090 err = proto_register(&udplitev6_prot, 1);
1091 if (err)
1092 goto out_unregister_udp_proto;
1093
1094 err = proto_register(&rawv6_prot, 1);
1095 if (err)
1096 goto out_unregister_udplite_proto;
1097
1098
1099 /* We MUST register RAW sockets before we create the ICMP6,
1100 * IGMP6, or NDISC control sockets.
1101 */
1102 err = rawv6_init();
1103 if (err)
1104 goto out_unregister_raw_proto;
1105
1106 /* Register the family here so that the init calls below will
1107 * be able to create sockets. (?? is this dangerous ??)
1108 */
1109 err = sock_register(&inet6_family_ops);
1110 if (err)
1111 goto out_sock_register_fail;
1112
1113#ifdef CONFIG_SYSCTL
1114 err = ipv6_static_sysctl_register();
1115 if (err)
1116 goto static_sysctl_fail;
1117#endif
1118 /*
1119 * ipngwg API draft makes clear that the correct semantics
1120 * for TCP and UDP is to consider one TCP and UDP instance
1121 * in a host available by both INET and INET6 APIs and
1122 * able to communicate via both network protocols.
1123 */
1124
1125 err = register_pernet_subsys(&inet6_net_ops);
1126 if (err)
1127 goto register_pernet_fail;
1128 err = icmpv6_init();
1129 if (err)
1130 goto icmp_fail;
1131 err = ip6_mr_init();
1132 if (err)
1133 goto ipmr_fail;
1134 err = ndisc_init();
1135 if (err)
1136 goto ndisc_fail;
1137 err = igmp6_init();
1138 if (err)
1139 goto igmp_fail;
1140 err = ipv6_netfilter_init();
1141 if (err)
1142 goto netfilter_fail;
1143 /* Create /proc/foo6 entries. */
1144#ifdef CONFIG_PROC_FS
1145 err = -ENOMEM;
1146 if (raw6_proc_init())
1147 goto proc_raw6_fail;
1148 if (udplite6_proc_init())
1149 goto proc_udplite6_fail;
1150 if (ipv6_misc_proc_init())
1151 goto proc_misc6_fail;
1152 if (if6_proc_init())
1153 goto proc_if6_fail;
1154#endif
1155 err = ip6_route_init();
1156 if (err)
1157 goto ip6_route_fail;
1158 err = ip6_flowlabel_init();
1159 if (err)
1160 goto ip6_flowlabel_fail;
1161 err = addrconf_init();
1162 if (err)
1163 goto addrconf_fail;
1164
1165 /* Init v6 extension headers. */
1166 err = ipv6_exthdrs_init();
1167 if (err)
1168 goto ipv6_exthdrs_fail;
1169
1170 err = ipv6_frag_init();
1171 if (err)
1172 goto ipv6_frag_fail;
1173
1174 /* Init v6 transport protocols. */
1175 err = udpv6_init();
1176 if (err)
1177 goto udpv6_fail;
1178
1179 err = udplitev6_init();
1180 if (err)
1181 goto udplitev6_fail;
1182
1183 err = tcpv6_init();
1184 if (err)
1185 goto tcpv6_fail;
1186
1187 err = ipv6_packet_init();
1188 if (err)
1189 goto ipv6_packet_fail;
1190
1191#ifdef CONFIG_SYSCTL
1192 err = ipv6_sysctl_register();
1193 if (err)
1194 goto sysctl_fail;
1195#endif
1196out:
1197 return err;
1198
1199#ifdef CONFIG_SYSCTL
1200sysctl_fail:
1201 ipv6_packet_cleanup();
1202#endif
1203ipv6_packet_fail:
1204 tcpv6_exit();
1205tcpv6_fail:
1206 udplitev6_exit();
1207udplitev6_fail:
1208 udpv6_exit();
1209udpv6_fail:
1210 ipv6_frag_exit();
1211ipv6_frag_fail:
1212 ipv6_exthdrs_exit();
1213ipv6_exthdrs_fail:
1214 addrconf_cleanup();
1215addrconf_fail:
1216 ip6_flowlabel_cleanup();
1217ip6_flowlabel_fail:
1218 ip6_route_cleanup();
1219ip6_route_fail:
1220#ifdef CONFIG_PROC_FS
1221 if6_proc_exit();
1222proc_if6_fail:
1223 ipv6_misc_proc_exit();
1224proc_misc6_fail:
1225 udplite6_proc_exit();
1226proc_udplite6_fail:
1227 raw6_proc_exit();
1228proc_raw6_fail:
1229#endif
1230 ipv6_netfilter_fini();
1231netfilter_fail:
1232 igmp6_cleanup();
1233igmp_fail:
1234 ndisc_cleanup();
1235ndisc_fail:
1236 ip6_mr_cleanup();
1237ipmr_fail:
1238 icmpv6_cleanup();
1239icmp_fail:
1240 unregister_pernet_subsys(&inet6_net_ops);
1241register_pernet_fail:
1242#ifdef CONFIG_SYSCTL
1243 ipv6_static_sysctl_unregister();
1244static_sysctl_fail:
1245#endif
1246 sock_unregister(PF_INET6);
1247 rtnl_unregister_all(PF_INET6);
1248out_sock_register_fail:
1249 rawv6_exit();
1250out_unregister_raw_proto:
1251 proto_unregister(&rawv6_prot);
1252out_unregister_udplite_proto:
1253 proto_unregister(&udplitev6_prot);
1254out_unregister_udp_proto:
1255 proto_unregister(&udpv6_prot);
1256out_unregister_tcp_proto:
1257 proto_unregister(&tcpv6_prot);
1258 goto out;
1259}
1260module_init(inet6_init);
1261
1262static void __exit inet6_exit(void)
1263{
1264 if (disable_ipv6_mod)
1265 return;
1266
1267 /* First of all disallow new sockets creation. */
1268 sock_unregister(PF_INET6);
1269 /* Disallow any further netlink messages */
1270 rtnl_unregister_all(PF_INET6);
1271
1272#ifdef CONFIG_SYSCTL
1273 ipv6_sysctl_unregister();
1274#endif
1275 udpv6_exit();
1276 udplitev6_exit();
1277 tcpv6_exit();
1278
1279 /* Cleanup code parts. */
1280 ipv6_packet_cleanup();
1281 ipv6_frag_exit();
1282 ipv6_exthdrs_exit();
1283 addrconf_cleanup();
1284 ip6_flowlabel_cleanup();
1285 ip6_route_cleanup();
1286#ifdef CONFIG_PROC_FS
1287
1288 /* Cleanup code parts. */
1289 if6_proc_exit();
1290 ipv6_misc_proc_exit();
1291 udplite6_proc_exit();
1292 raw6_proc_exit();
1293#endif
1294 ipv6_netfilter_fini();
1295 igmp6_cleanup();
1296 ndisc_cleanup();
1297 ip6_mr_cleanup();
1298 icmpv6_cleanup();
1299 rawv6_exit();
1300
1301 unregister_pernet_subsys(&inet6_net_ops);
1302#ifdef CONFIG_SYSCTL
1303 ipv6_static_sysctl_unregister();
1304#endif
1305 proto_unregister(&rawv6_prot);
1306 proto_unregister(&udplitev6_prot);
1307 proto_unregister(&udpv6_prot);
1308 proto_unregister(&tcpv6_prot);
1309
1310 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1311}
1312module_exit(inet6_exit);
1313
1314MODULE_ALIAS_NETPROTO(PF_INET6);
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * PF_INET6 socket protocol family
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 *
9 * Adapted from linux/net/ipv4/af_inet.c
10 *
11 * Fixes:
12 * piggy, Karl Knutson : Socket protocol table
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * Arnaldo Melo : check proc_net_create return, cleanups
15 */
16
17#define pr_fmt(fmt) "IPv6: " fmt
18
19#include <linux/module.h>
20#include <linux/capability.h>
21#include <linux/errno.h>
22#include <linux/types.h>
23#include <linux/socket.h>
24#include <linux/in.h>
25#include <linux/kernel.h>
26#include <linux/timer.h>
27#include <linux/string.h>
28#include <linux/sockios.h>
29#include <linux/net.h>
30#include <linux/fcntl.h>
31#include <linux/mm.h>
32#include <linux/interrupt.h>
33#include <linux/proc_fs.h>
34#include <linux/stat.h>
35#include <linux/init.h>
36#include <linux/slab.h>
37
38#include <linux/inet.h>
39#include <linux/netdevice.h>
40#include <linux/icmpv6.h>
41#include <linux/netfilter_ipv6.h>
42
43#include <net/ip.h>
44#include <net/ipv6.h>
45#include <net/udp.h>
46#include <net/udplite.h>
47#include <net/tcp.h>
48#include <net/ping.h>
49#include <net/protocol.h>
50#include <net/inet_common.h>
51#include <net/route.h>
52#include <net/transp_v6.h>
53#include <net/ip6_route.h>
54#include <net/addrconf.h>
55#include <net/ipv6_stubs.h>
56#include <net/ndisc.h>
57#ifdef CONFIG_IPV6_TUNNEL
58#include <net/ip6_tunnel.h>
59#endif
60#include <net/calipso.h>
61#include <net/seg6.h>
62#include <net/rpl.h>
63#include <net/compat.h>
64#include <net/xfrm.h>
65
66#include <linux/uaccess.h>
67#include <linux/mroute6.h>
68
69#include "ip6_offload.h"
70
71MODULE_AUTHOR("Cast of dozens");
72MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
73MODULE_LICENSE("GPL");
74
75/* The inetsw6 table contains everything that inet6_create needs to
76 * build a new socket.
77 */
78static struct list_head inetsw6[SOCK_MAX];
79static DEFINE_SPINLOCK(inetsw6_lock);
80
81struct ipv6_params ipv6_defaults = {
82 .disable_ipv6 = 0,
83 .autoconf = 1,
84};
85
86static int disable_ipv6_mod;
87
88module_param_named(disable, disable_ipv6_mod, int, 0444);
89MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
90
91module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
92MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
93
94module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
95MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
96
97bool ipv6_mod_enabled(void)
98{
99 return disable_ipv6_mod == 0;
100}
101EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
102
103static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
104{
105 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
106
107 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
108}
109
110static int inet6_create(struct net *net, struct socket *sock, int protocol,
111 int kern)
112{
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;
121
122 if (protocol < 0 || protocol >= IPPROTO_MAX)
123 return -EINVAL;
124
125 /* Look for the requested type/protocol pair. */
126lookup_protocol:
127 err = -ESOCKTNOSUPPORT;
128 rcu_read_lock();
129 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
130
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;
141 }
142 if (IPPROTO_IP == answer->protocol)
143 break;
144 }
145 err = -EPROTONOSUPPORT;
146 }
147
148 if (err) {
149 if (try_loading_module < 2) {
150 rcu_read_unlock();
151 /*
152 * Be more specific, e.g. net-pf-10-proto-132-type-1
153 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
154 */
155 if (++try_loading_module == 1)
156 request_module("net-pf-%d-proto-%d-type-%d",
157 PF_INET6, protocol, sock->type);
158 /*
159 * Fall back to generic, e.g. net-pf-10-proto-132
160 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
161 */
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;
168 }
169
170 err = -EPERM;
171 if (sock->type == SOCK_RAW && !kern &&
172 !ns_capable(net->user_ns, CAP_NET_RAW))
173 goto out_rcu_unlock;
174
175 sock->ops = answer->ops;
176 answer_prot = answer->prot;
177 answer_flags = answer->flags;
178 rcu_read_unlock();
179
180 WARN_ON(!answer_prot->slab);
181
182 err = -ENOBUFS;
183 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
184 if (!sk)
185 goto out;
186
187 sock_init_data(sock, sk);
188
189 err = 0;
190 if (INET_PROTOSW_REUSE & answer_flags)
191 sk->sk_reuse = SK_CAN_REUSE;
192
193 inet = inet_sk(sk);
194 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
195
196 if (SOCK_RAW == sock->type) {
197 inet->inet_num = protocol;
198 if (IPPROTO_RAW == protocol)
199 inet->hdrincl = 1;
200 }
201
202 sk->sk_destruct = inet_sock_destruct;
203 sk->sk_family = PF_INET6;
204 sk->sk_protocol = protocol;
205
206 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
207
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 & FLOWLABEL_REFLECT_ESTABLISHED;
215 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
216
217 /* Init the ipv4 part of the socket since we can have sockets
218 * using v6 API for ipv4.
219 */
220 inet->uc_ttl = -1;
221
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;
227
228 if (net->ipv4.sysctl_ip_no_pmtu_disc)
229 inet->pmtudisc = IP_PMTUDISC_DONT;
230 else
231 inet->pmtudisc = IP_PMTUDISC_WANT;
232 /*
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.
236 *
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
240 */
241 sk_refcnt_debug_inc(sk);
242
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.
247 */
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;
253 }
254 }
255 if (sk->sk_prot->init) {
256 err = sk->sk_prot->init(sk);
257 if (err) {
258 sk_common_release(sk);
259 goto out;
260 }
261 }
262
263 if (!kern) {
264 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
265 if (err) {
266 sk_common_release(sk);
267 goto out;
268 }
269 }
270out:
271 return err;
272out_rcu_unlock:
273 rcu_read_unlock();
274 goto out;
275}
276
277static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
278 u32 flags)
279{
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;
289
290 if (addr->sin6_family != AF_INET6)
291 return -EAFNOSUPPORT;
292
293 addr_type = ipv6_addr_type(&addr->sin6_addr);
294 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
295 return -EINVAL;
296
297 snum = ntohs(addr->sin6_port);
298 if (snum && inet_port_requires_bind_service(net, snum) &&
299 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
300 return -EACCES;
301
302 if (flags & BIND_WITH_LOCK)
303 lock_sock(sk);
304
305 /* Check these errors (active socket, double bind). */
306 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
307 err = -EINVAL;
308 goto out;
309 }
310
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;
315
316 /* Binding to v4-mapped address on a v6-only socket
317 * makes no sense
318 */
319 if (sk->sk_ipv6only) {
320 err = -EINVAL;
321 goto out;
322 }
323
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;
330 }
331 }
332
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();
337
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;
345 }
346 } else {
347 if (addr_type != IPV6_ADDR_ANY) {
348 struct net_device *dev = NULL;
349
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.
356 */
357 sk->sk_bound_dev_if = addr->sin6_scope_id;
358 }
359
360 /* Binding to link-local address requires an interface */
361 if (!sk->sk_bound_dev_if) {
362 err = -EINVAL;
363 goto out_unlock;
364 }
365 }
366
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;
372 }
373 }
374
375 /* ipv4 addr of the socket is invalid. Only the
376 * unspecified and mapped address have a v4 equivalent.
377 */
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;
385 }
386 }
387 rcu_read_unlock();
388 }
389 }
390
391 inet->inet_rcv_saddr = v4addr;
392 inet->inet_saddr = v4addr;
393
394 sk->sk_v6_rcv_saddr = addr->sin6_addr;
395
396 if (!(addr_type & IPV6_ADDR_MULTICAST))
397 np->saddr = addr->sin6_addr;
398
399 saved_ipv6only = sk->sk_ipv6only;
400 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
401 sk->sk_ipv6only = 1;
402
403 /* Make sure we are allowed to bind here. */
404 if (snum || !(inet->bind_address_no_port ||
405 (flags & BIND_FORCE_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;
411 }
412 if (!(flags & BIND_FROM_BPF)) {
413 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
414 if (err) {
415 sk->sk_ipv6only = saved_ipv6only;
416 inet_reset_saddr(sk);
417 goto out;
418 }
419 }
420 }
421
422 if (addr_type != IPV6_ADDR_ANY)
423 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
424 if (snum)
425 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
426 inet->inet_sport = htons(inet->inet_num);
427 inet->inet_dport = 0;
428 inet->inet_daddr = 0;
429out:
430 if (flags & BIND_WITH_LOCK)
431 release_sock(sk);
432 return err;
433out_unlock:
434 rcu_read_unlock();
435 goto out;
436}
437
438/* bind for INET6 API */
439int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
440{
441 struct sock *sk = sock->sk;
442 int err = 0;
443
444 /* If the socket has its own bind function then use it. */
445 if (sk->sk_prot->bind)
446 return sk->sk_prot->bind(sk, uaddr, addr_len);
447
448 if (addr_len < SIN6_LEN_RFC2133)
449 return -EINVAL;
450
451 /* BPF prog is run before any checks are done so that if the prog
452 * changes context in a wrong way it will be caught.
453 */
454 err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
455 if (err)
456 return err;
457
458 return __inet6_bind(sk, uaddr, addr_len, BIND_WITH_LOCK);
459}
460EXPORT_SYMBOL(inet6_bind);
461
462int inet6_release(struct socket *sock)
463{
464 struct sock *sk = sock->sk;
465
466 if (!sk)
467 return -EINVAL;
468
469 /* Free mc lists */
470 ipv6_sock_mc_close(sk);
471
472 /* Free ac lists */
473 ipv6_sock_ac_close(sk);
474
475 return inet_release(sock);
476}
477EXPORT_SYMBOL(inet6_release);
478
479void inet6_destroy_sock(struct sock *sk)
480{
481 struct ipv6_pinfo *np = inet6_sk(sk);
482 struct sk_buff *skb;
483 struct ipv6_txoptions *opt;
484
485 /* Release rx options */
486
487 skb = xchg(&np->pktoptions, NULL);
488 kfree_skb(skb);
489
490 skb = xchg(&np->rxpmtu, NULL);
491 kfree_skb(skb);
492
493 /* Free flowlabels */
494 fl6_free_socklist(sk);
495
496 /* Free tx options */
497
498 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
499 if (opt) {
500 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
501 txopt_put(opt);
502 }
503}
504EXPORT_SYMBOL_GPL(inet6_destroy_sock);
505
506/*
507 * This does both peername and sockname.
508 */
509int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
510 int peer)
511{
512 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
513 struct sock *sk = sock->sk;
514 struct inet_sock *inet = inet_sk(sk);
515 struct ipv6_pinfo *np = inet6_sk(sk);
516
517 sin->sin6_family = AF_INET6;
518 sin->sin6_flowinfo = 0;
519 sin->sin6_scope_id = 0;
520 if (peer) {
521 if (!inet->inet_dport)
522 return -ENOTCONN;
523 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
524 peer == 1)
525 return -ENOTCONN;
526 sin->sin6_port = inet->inet_dport;
527 sin->sin6_addr = sk->sk_v6_daddr;
528 if (np->sndflow)
529 sin->sin6_flowinfo = np->flow_label;
530 } else {
531 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
532 sin->sin6_addr = np->saddr;
533 else
534 sin->sin6_addr = sk->sk_v6_rcv_saddr;
535 sin->sin6_port = inet->inet_sport;
536 }
537 if (cgroup_bpf_enabled)
538 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin,
539 peer ? BPF_CGROUP_INET6_GETPEERNAME :
540 BPF_CGROUP_INET6_GETSOCKNAME,
541 NULL);
542 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
543 sk->sk_bound_dev_if);
544 return sizeof(*sin);
545}
546EXPORT_SYMBOL(inet6_getname);
547
548int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
549{
550 void __user *argp = (void __user *)arg;
551 struct sock *sk = sock->sk;
552 struct net *net = sock_net(sk);
553
554 switch (cmd) {
555 case SIOCADDRT:
556 case SIOCDELRT: {
557 struct in6_rtmsg rtmsg;
558
559 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
560 return -EFAULT;
561 return ipv6_route_ioctl(net, cmd, &rtmsg);
562 }
563 case SIOCSIFADDR:
564 return addrconf_add_ifaddr(net, argp);
565 case SIOCDIFADDR:
566 return addrconf_del_ifaddr(net, argp);
567 case SIOCSIFDSTADDR:
568 return addrconf_set_dstaddr(net, argp);
569 default:
570 if (!sk->sk_prot->ioctl)
571 return -ENOIOCTLCMD;
572 return sk->sk_prot->ioctl(sk, cmd, arg);
573 }
574 /*NOTREACHED*/
575 return 0;
576}
577EXPORT_SYMBOL(inet6_ioctl);
578
579#ifdef CONFIG_COMPAT
580struct compat_in6_rtmsg {
581 struct in6_addr rtmsg_dst;
582 struct in6_addr rtmsg_src;
583 struct in6_addr rtmsg_gateway;
584 u32 rtmsg_type;
585 u16 rtmsg_dst_len;
586 u16 rtmsg_src_len;
587 u32 rtmsg_metric;
588 u32 rtmsg_info;
589 u32 rtmsg_flags;
590 s32 rtmsg_ifindex;
591};
592
593static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
594 struct compat_in6_rtmsg __user *ur)
595{
596 struct in6_rtmsg rt;
597
598 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
599 3 * sizeof(struct in6_addr)) ||
600 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
601 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
602 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
603 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
604 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
605 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
606 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
607 return -EFAULT;
608
609
610 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
611}
612
613int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
614{
615 void __user *argp = compat_ptr(arg);
616 struct sock *sk = sock->sk;
617
618 switch (cmd) {
619 case SIOCADDRT:
620 case SIOCDELRT:
621 return inet6_compat_routing_ioctl(sk, cmd, argp);
622 default:
623 return -ENOIOCTLCMD;
624 }
625}
626EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
627#endif /* CONFIG_COMPAT */
628
629INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
630 size_t));
631int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
632{
633 struct sock *sk = sock->sk;
634
635 if (unlikely(inet_send_prepare(sk)))
636 return -EAGAIN;
637
638 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
639 sk, msg, size);
640}
641
642INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
643 size_t, int, int, int *));
644int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
645 int flags)
646{
647 struct sock *sk = sock->sk;
648 int addr_len = 0;
649 int err;
650
651 if (likely(!(flags & MSG_ERRQUEUE)))
652 sock_rps_record_flow(sk);
653
654 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
655 sk, msg, size, flags & MSG_DONTWAIT,
656 flags & ~MSG_DONTWAIT, &addr_len);
657 if (err >= 0)
658 msg->msg_namelen = addr_len;
659 return err;
660}
661
662const struct proto_ops inet6_stream_ops = {
663 .family = PF_INET6,
664 .owner = THIS_MODULE,
665 .release = inet6_release,
666 .bind = inet6_bind,
667 .connect = inet_stream_connect, /* ok */
668 .socketpair = sock_no_socketpair, /* a do nothing */
669 .accept = inet_accept, /* ok */
670 .getname = inet6_getname,
671 .poll = tcp_poll, /* ok */
672 .ioctl = inet6_ioctl, /* must change */
673 .gettstamp = sock_gettstamp,
674 .listen = inet_listen, /* ok */
675 .shutdown = inet_shutdown, /* ok */
676 .setsockopt = sock_common_setsockopt, /* ok */
677 .getsockopt = sock_common_getsockopt, /* ok */
678 .sendmsg = inet6_sendmsg, /* retpoline's sake */
679 .recvmsg = inet6_recvmsg, /* retpoline's sake */
680#ifdef CONFIG_MMU
681 .mmap = tcp_mmap,
682#endif
683 .sendpage = inet_sendpage,
684 .sendmsg_locked = tcp_sendmsg_locked,
685 .sendpage_locked = tcp_sendpage_locked,
686 .splice_read = tcp_splice_read,
687 .read_sock = tcp_read_sock,
688 .peek_len = tcp_peek_len,
689#ifdef CONFIG_COMPAT
690 .compat_ioctl = inet6_compat_ioctl,
691#endif
692 .set_rcvlowat = tcp_set_rcvlowat,
693};
694
695const struct proto_ops inet6_dgram_ops = {
696 .family = PF_INET6,
697 .owner = THIS_MODULE,
698 .release = inet6_release,
699 .bind = inet6_bind,
700 .connect = inet_dgram_connect, /* ok */
701 .socketpair = sock_no_socketpair, /* a do nothing */
702 .accept = sock_no_accept, /* a do nothing */
703 .getname = inet6_getname,
704 .poll = udp_poll, /* ok */
705 .ioctl = inet6_ioctl, /* must change */
706 .gettstamp = sock_gettstamp,
707 .listen = sock_no_listen, /* ok */
708 .shutdown = inet_shutdown, /* ok */
709 .setsockopt = sock_common_setsockopt, /* ok */
710 .getsockopt = sock_common_getsockopt, /* ok */
711 .sendmsg = inet6_sendmsg, /* retpoline's sake */
712 .recvmsg = inet6_recvmsg, /* retpoline's sake */
713 .mmap = sock_no_mmap,
714 .sendpage = sock_no_sendpage,
715 .set_peek_off = sk_set_peek_off,
716#ifdef CONFIG_COMPAT
717 .compat_ioctl = inet6_compat_ioctl,
718#endif
719};
720
721static const struct net_proto_family inet6_family_ops = {
722 .family = PF_INET6,
723 .create = inet6_create,
724 .owner = THIS_MODULE,
725};
726
727int inet6_register_protosw(struct inet_protosw *p)
728{
729 struct list_head *lh;
730 struct inet_protosw *answer;
731 struct list_head *last_perm;
732 int protocol = p->protocol;
733 int ret;
734
735 spin_lock_bh(&inetsw6_lock);
736
737 ret = -EINVAL;
738 if (p->type >= SOCK_MAX)
739 goto out_illegal;
740
741 /* If we are trying to override a permanent protocol, bail. */
742 answer = NULL;
743 ret = -EPERM;
744 last_perm = &inetsw6[p->type];
745 list_for_each(lh, &inetsw6[p->type]) {
746 answer = list_entry(lh, struct inet_protosw, list);
747
748 /* Check only the non-wild match. */
749 if (INET_PROTOSW_PERMANENT & answer->flags) {
750 if (protocol == answer->protocol)
751 break;
752 last_perm = lh;
753 }
754
755 answer = NULL;
756 }
757 if (answer)
758 goto out_permanent;
759
760 /* Add the new entry after the last permanent entry if any, so that
761 * the new entry does not override a permanent entry when matched with
762 * a wild-card protocol. But it is allowed to override any existing
763 * non-permanent entry. This means that when we remove this entry, the
764 * system automatically returns to the old behavior.
765 */
766 list_add_rcu(&p->list, last_perm);
767 ret = 0;
768out:
769 spin_unlock_bh(&inetsw6_lock);
770 return ret;
771
772out_permanent:
773 pr_err("Attempt to override permanent protocol %d\n", protocol);
774 goto out;
775
776out_illegal:
777 pr_err("Ignoring attempt to register invalid socket type %d\n",
778 p->type);
779 goto out;
780}
781EXPORT_SYMBOL(inet6_register_protosw);
782
783void
784inet6_unregister_protosw(struct inet_protosw *p)
785{
786 if (INET_PROTOSW_PERMANENT & p->flags) {
787 pr_err("Attempt to unregister permanent protocol %d\n",
788 p->protocol);
789 } else {
790 spin_lock_bh(&inetsw6_lock);
791 list_del_rcu(&p->list);
792 spin_unlock_bh(&inetsw6_lock);
793
794 synchronize_net();
795 }
796}
797EXPORT_SYMBOL(inet6_unregister_protosw);
798
799int inet6_sk_rebuild_header(struct sock *sk)
800{
801 struct ipv6_pinfo *np = inet6_sk(sk);
802 struct dst_entry *dst;
803
804 dst = __sk_dst_check(sk, np->dst_cookie);
805
806 if (!dst) {
807 struct inet_sock *inet = inet_sk(sk);
808 struct in6_addr *final_p, final;
809 struct flowi6 fl6;
810
811 memset(&fl6, 0, sizeof(fl6));
812 fl6.flowi6_proto = sk->sk_protocol;
813 fl6.daddr = sk->sk_v6_daddr;
814 fl6.saddr = np->saddr;
815 fl6.flowlabel = np->flow_label;
816 fl6.flowi6_oif = sk->sk_bound_dev_if;
817 fl6.flowi6_mark = sk->sk_mark;
818 fl6.fl6_dport = inet->inet_dport;
819 fl6.fl6_sport = inet->inet_sport;
820 fl6.flowi6_uid = sk->sk_uid;
821 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
822
823 rcu_read_lock();
824 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
825 &final);
826 rcu_read_unlock();
827
828 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
829 if (IS_ERR(dst)) {
830 sk->sk_route_caps = 0;
831 sk->sk_err_soft = -PTR_ERR(dst);
832 return PTR_ERR(dst);
833 }
834
835 ip6_dst_store(sk, dst, NULL, NULL);
836 }
837
838 return 0;
839}
840EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
841
842bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
843 const struct inet6_skb_parm *opt)
844{
845 const struct ipv6_pinfo *np = inet6_sk(sk);
846
847 if (np->rxopt.all) {
848 if (((opt->flags & IP6SKB_HOPBYHOP) &&
849 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
850 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
851 np->rxopt.bits.rxflow) ||
852 (opt->srcrt && (np->rxopt.bits.srcrt ||
853 np->rxopt.bits.osrcrt)) ||
854 ((opt->dst1 || opt->dst0) &&
855 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
856 return true;
857 }
858 return false;
859}
860EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
861
862static struct packet_type ipv6_packet_type __read_mostly = {
863 .type = cpu_to_be16(ETH_P_IPV6),
864 .func = ipv6_rcv,
865 .list_func = ipv6_list_rcv,
866};
867
868static int __init ipv6_packet_init(void)
869{
870 dev_add_pack(&ipv6_packet_type);
871 return 0;
872}
873
874static void ipv6_packet_cleanup(void)
875{
876 dev_remove_pack(&ipv6_packet_type);
877}
878
879static int __net_init ipv6_init_mibs(struct net *net)
880{
881 int i;
882
883 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
884 if (!net->mib.udp_stats_in6)
885 return -ENOMEM;
886 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
887 if (!net->mib.udplite_stats_in6)
888 goto err_udplite_mib;
889 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
890 if (!net->mib.ipv6_statistics)
891 goto err_ip_mib;
892
893 for_each_possible_cpu(i) {
894 struct ipstats_mib *af_inet6_stats;
895 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
896 u64_stats_init(&af_inet6_stats->syncp);
897 }
898
899
900 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
901 if (!net->mib.icmpv6_statistics)
902 goto err_icmp_mib;
903 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
904 GFP_KERNEL);
905 if (!net->mib.icmpv6msg_statistics)
906 goto err_icmpmsg_mib;
907 return 0;
908
909err_icmpmsg_mib:
910 free_percpu(net->mib.icmpv6_statistics);
911err_icmp_mib:
912 free_percpu(net->mib.ipv6_statistics);
913err_ip_mib:
914 free_percpu(net->mib.udplite_stats_in6);
915err_udplite_mib:
916 free_percpu(net->mib.udp_stats_in6);
917 return -ENOMEM;
918}
919
920static void ipv6_cleanup_mibs(struct net *net)
921{
922 free_percpu(net->mib.udp_stats_in6);
923 free_percpu(net->mib.udplite_stats_in6);
924 free_percpu(net->mib.ipv6_statistics);
925 free_percpu(net->mib.icmpv6_statistics);
926 kfree(net->mib.icmpv6msg_statistics);
927}
928
929static int __net_init inet6_net_init(struct net *net)
930{
931 int err = 0;
932
933 net->ipv6.sysctl.bindv6only = 0;
934 net->ipv6.sysctl.icmpv6_time = 1*HZ;
935 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
936 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
937 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
938
939 /* By default, rate limit error messages.
940 * Except for pmtu discovery, it would break it.
941 * proc_do_large_bitmap needs pointer to the bitmap.
942 */
943 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
944 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
945 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
946
947 net->ipv6.sysctl.flowlabel_consistency = 1;
948 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
949 net->ipv6.sysctl.idgen_retries = 3;
950 net->ipv6.sysctl.idgen_delay = 1 * HZ;
951 net->ipv6.sysctl.flowlabel_state_ranges = 0;
952 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
953 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
954 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
955 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
956 atomic_set(&net->ipv6.fib6_sernum, 1);
957
958 err = ipv6_init_mibs(net);
959 if (err)
960 return err;
961#ifdef CONFIG_PROC_FS
962 err = udp6_proc_init(net);
963 if (err)
964 goto out;
965 err = tcp6_proc_init(net);
966 if (err)
967 goto proc_tcp6_fail;
968 err = ac6_proc_init(net);
969 if (err)
970 goto proc_ac6_fail;
971#endif
972 return err;
973
974#ifdef CONFIG_PROC_FS
975proc_ac6_fail:
976 tcp6_proc_exit(net);
977proc_tcp6_fail:
978 udp6_proc_exit(net);
979out:
980 ipv6_cleanup_mibs(net);
981 return err;
982#endif
983}
984
985static void __net_exit inet6_net_exit(struct net *net)
986{
987#ifdef CONFIG_PROC_FS
988 udp6_proc_exit(net);
989 tcp6_proc_exit(net);
990 ac6_proc_exit(net);
991#endif
992 ipv6_cleanup_mibs(net);
993}
994
995static struct pernet_operations inet6_net_ops = {
996 .init = inet6_net_init,
997 .exit = inet6_net_exit,
998};
999
1000static int ipv6_route_input(struct sk_buff *skb)
1001{
1002 ip6_route_input(skb);
1003 return skb_dst(skb)->error;
1004}
1005
1006static const struct ipv6_stub ipv6_stub_impl = {
1007 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1008 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1009 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1010 .ipv6_route_input = ipv6_route_input,
1011 .fib6_get_table = fib6_get_table,
1012 .fib6_table_lookup = fib6_table_lookup,
1013 .fib6_lookup = fib6_lookup,
1014 .fib6_select_path = fib6_select_path,
1015 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1016 .fib6_nh_init = fib6_nh_init,
1017 .fib6_nh_release = fib6_nh_release,
1018 .fib6_update_sernum = fib6_update_sernum_stub,
1019 .fib6_rt_update = fib6_rt_update,
1020 .ip6_del_rt = ip6_del_rt,
1021 .udpv6_encap_enable = udpv6_encap_enable,
1022 .ndisc_send_na = ndisc_send_na,
1023#if IS_ENABLED(CONFIG_XFRM)
1024 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1025 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1026 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1027#endif
1028 .nd_tbl = &nd_tbl,
1029};
1030
1031static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1032 .inet6_bind = __inet6_bind,
1033 .udp6_lib_lookup = __udp6_lib_lookup,
1034};
1035
1036static int __init inet6_init(void)
1037{
1038 struct list_head *r;
1039 int err = 0;
1040
1041 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1042
1043 /* Register the socket-side information for inet6_create. */
1044 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1045 INIT_LIST_HEAD(r);
1046
1047 if (disable_ipv6_mod) {
1048 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1049 goto out;
1050 }
1051
1052 err = proto_register(&tcpv6_prot, 1);
1053 if (err)
1054 goto out;
1055
1056 err = proto_register(&udpv6_prot, 1);
1057 if (err)
1058 goto out_unregister_tcp_proto;
1059
1060 err = proto_register(&udplitev6_prot, 1);
1061 if (err)
1062 goto out_unregister_udp_proto;
1063
1064 err = proto_register(&rawv6_prot, 1);
1065 if (err)
1066 goto out_unregister_udplite_proto;
1067
1068 err = proto_register(&pingv6_prot, 1);
1069 if (err)
1070 goto out_unregister_raw_proto;
1071
1072 /* We MUST register RAW sockets before we create the ICMP6,
1073 * IGMP6, or NDISC control sockets.
1074 */
1075 err = rawv6_init();
1076 if (err)
1077 goto out_unregister_ping_proto;
1078
1079 /* Register the family here so that the init calls below will
1080 * be able to create sockets. (?? is this dangerous ??)
1081 */
1082 err = sock_register(&inet6_family_ops);
1083 if (err)
1084 goto out_sock_register_fail;
1085
1086 /*
1087 * ipngwg API draft makes clear that the correct semantics
1088 * for TCP and UDP is to consider one TCP and UDP instance
1089 * in a host available by both INET and INET6 APIs and
1090 * able to communicate via both network protocols.
1091 */
1092
1093 err = register_pernet_subsys(&inet6_net_ops);
1094 if (err)
1095 goto register_pernet_fail;
1096 err = ip6_mr_init();
1097 if (err)
1098 goto ipmr_fail;
1099 err = icmpv6_init();
1100 if (err)
1101 goto icmp_fail;
1102 err = ndisc_init();
1103 if (err)
1104 goto ndisc_fail;
1105 err = igmp6_init();
1106 if (err)
1107 goto igmp_fail;
1108
1109 err = ipv6_netfilter_init();
1110 if (err)
1111 goto netfilter_fail;
1112 /* Create /proc/foo6 entries. */
1113#ifdef CONFIG_PROC_FS
1114 err = -ENOMEM;
1115 if (raw6_proc_init())
1116 goto proc_raw6_fail;
1117 if (udplite6_proc_init())
1118 goto proc_udplite6_fail;
1119 if (ipv6_misc_proc_init())
1120 goto proc_misc6_fail;
1121 if (if6_proc_init())
1122 goto proc_if6_fail;
1123#endif
1124 err = ip6_route_init();
1125 if (err)
1126 goto ip6_route_fail;
1127 err = ndisc_late_init();
1128 if (err)
1129 goto ndisc_late_fail;
1130 err = ip6_flowlabel_init();
1131 if (err)
1132 goto ip6_flowlabel_fail;
1133 err = ipv6_anycast_init();
1134 if (err)
1135 goto ipv6_anycast_fail;
1136 err = addrconf_init();
1137 if (err)
1138 goto addrconf_fail;
1139
1140 /* Init v6 extension headers. */
1141 err = ipv6_exthdrs_init();
1142 if (err)
1143 goto ipv6_exthdrs_fail;
1144
1145 err = ipv6_frag_init();
1146 if (err)
1147 goto ipv6_frag_fail;
1148
1149 /* Init v6 transport protocols. */
1150 err = udpv6_init();
1151 if (err)
1152 goto udpv6_fail;
1153
1154 err = udplitev6_init();
1155 if (err)
1156 goto udplitev6_fail;
1157
1158 err = udpv6_offload_init();
1159 if (err)
1160 goto udpv6_offload_fail;
1161
1162 err = tcpv6_init();
1163 if (err)
1164 goto tcpv6_fail;
1165
1166 err = ipv6_packet_init();
1167 if (err)
1168 goto ipv6_packet_fail;
1169
1170 err = pingv6_init();
1171 if (err)
1172 goto pingv6_fail;
1173
1174 err = calipso_init();
1175 if (err)
1176 goto calipso_fail;
1177
1178 err = seg6_init();
1179 if (err)
1180 goto seg6_fail;
1181
1182 err = rpl_init();
1183 if (err)
1184 goto rpl_fail;
1185
1186 err = igmp6_late_init();
1187 if (err)
1188 goto igmp6_late_err;
1189
1190#ifdef CONFIG_SYSCTL
1191 err = ipv6_sysctl_register();
1192 if (err)
1193 goto sysctl_fail;
1194#endif
1195
1196 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1197 wmb();
1198 ipv6_stub = &ipv6_stub_impl;
1199 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1200out:
1201 return err;
1202
1203#ifdef CONFIG_SYSCTL
1204sysctl_fail:
1205 igmp6_late_cleanup();
1206#endif
1207igmp6_late_err:
1208 rpl_exit();
1209rpl_fail:
1210 seg6_exit();
1211seg6_fail:
1212 calipso_exit();
1213calipso_fail:
1214 pingv6_exit();
1215pingv6_fail:
1216 ipv6_packet_cleanup();
1217ipv6_packet_fail:
1218 tcpv6_exit();
1219tcpv6_fail:
1220 udpv6_offload_exit();
1221udpv6_offload_fail:
1222 udplitev6_exit();
1223udplitev6_fail:
1224 udpv6_exit();
1225udpv6_fail:
1226 ipv6_frag_exit();
1227ipv6_frag_fail:
1228 ipv6_exthdrs_exit();
1229ipv6_exthdrs_fail:
1230 addrconf_cleanup();
1231addrconf_fail:
1232 ipv6_anycast_cleanup();
1233ipv6_anycast_fail:
1234 ip6_flowlabel_cleanup();
1235ip6_flowlabel_fail:
1236 ndisc_late_cleanup();
1237ndisc_late_fail:
1238 ip6_route_cleanup();
1239ip6_route_fail:
1240#ifdef CONFIG_PROC_FS
1241 if6_proc_exit();
1242proc_if6_fail:
1243 ipv6_misc_proc_exit();
1244proc_misc6_fail:
1245 udplite6_proc_exit();
1246proc_udplite6_fail:
1247 raw6_proc_exit();
1248proc_raw6_fail:
1249#endif
1250 ipv6_netfilter_fini();
1251netfilter_fail:
1252 igmp6_cleanup();
1253igmp_fail:
1254 ndisc_cleanup();
1255ndisc_fail:
1256 icmpv6_cleanup();
1257icmp_fail:
1258 ip6_mr_cleanup();
1259ipmr_fail:
1260 unregister_pernet_subsys(&inet6_net_ops);
1261register_pernet_fail:
1262 sock_unregister(PF_INET6);
1263 rtnl_unregister_all(PF_INET6);
1264out_sock_register_fail:
1265 rawv6_exit();
1266out_unregister_ping_proto:
1267 proto_unregister(&pingv6_prot);
1268out_unregister_raw_proto:
1269 proto_unregister(&rawv6_prot);
1270out_unregister_udplite_proto:
1271 proto_unregister(&udplitev6_prot);
1272out_unregister_udp_proto:
1273 proto_unregister(&udpv6_prot);
1274out_unregister_tcp_proto:
1275 proto_unregister(&tcpv6_prot);
1276 goto out;
1277}
1278module_init(inet6_init);
1279
1280MODULE_ALIAS_NETPROTO(PF_INET6);