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#define pr_fmt(fmt) "IPv6: " fmt
22
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>
41
42#include <linux/inet.h>
43#include <linux/netdevice.h>
44#include <linux/icmpv6.h>
45#include <linux/netfilter_ipv6.h>
46
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/ipip.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#ifdef CONFIG_IPV6_TUNNEL
60#include <net/ip6_tunnel.h>
61#endif
62
63#include <asm/uaccess.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;
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 = SK_CAN_REUSE;
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 inet->rcv_tos = 0;
218
219 if (ipv4_config.no_pmtu_disc)
220 inet->pmtudisc = IP_PMTUDISC_DONT;
221 else
222 inet->pmtudisc = IP_PMTUDISC_WANT;
223 /*
224 * Increment only the relevant sk_prot->socks debug field, this changes
225 * the previous behaviour of incrementing both the equivalent to
226 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
227 *
228 * This allows better debug granularity as we'll know exactly how many
229 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
230 * transport protocol socks. -acme
231 */
232 sk_refcnt_debug_inc(sk);
233
234 if (inet->inet_num) {
235 /* It assumes that any protocol which allows
236 * the user to assign a number at socket
237 * creation time automatically shares.
238 */
239 inet->inet_sport = htons(inet->inet_num);
240 sk->sk_prot->hash(sk);
241 }
242 if (sk->sk_prot->init) {
243 err = sk->sk_prot->init(sk);
244 if (err) {
245 sk_common_release(sk);
246 goto out;
247 }
248 }
249out:
250 return err;
251out_rcu_unlock:
252 rcu_read_unlock();
253 goto out;
254}
255
256
257/* bind for INET6 API */
258int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
259{
260 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
261 struct sock *sk = sock->sk;
262 struct inet_sock *inet = inet_sk(sk);
263 struct ipv6_pinfo *np = inet6_sk(sk);
264 struct net *net = sock_net(sk);
265 __be32 v4addr = 0;
266 unsigned short snum;
267 int addr_type = 0;
268 int err = 0;
269
270 /* If the socket has its own bind function then use it. */
271 if (sk->sk_prot->bind)
272 return sk->sk_prot->bind(sk, uaddr, addr_len);
273
274 if (addr_len < SIN6_LEN_RFC2133)
275 return -EINVAL;
276
277 if (addr->sin6_family != AF_INET6)
278 return -EAFNOSUPPORT;
279
280 addr_type = ipv6_addr_type(&addr->sin6_addr);
281 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
282 return -EINVAL;
283
284 snum = ntohs(addr->sin6_port);
285 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
286 return -EACCES;
287
288 lock_sock(sk);
289
290 /* Check these errors (active socket, double bind). */
291 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
292 err = -EINVAL;
293 goto out;
294 }
295
296 /* Check if the address belongs to the host. */
297 if (addr_type == IPV6_ADDR_MAPPED) {
298 int chk_addr_ret;
299
300 /* Binding to v4-mapped address on a v6-only socket
301 * makes no sense
302 */
303 if (np->ipv6only) {
304 err = -EINVAL;
305 goto out;
306 }
307
308 /* Reproduce AF_INET checks to make the bindings consistent */
309 v4addr = addr->sin6_addr.s6_addr32[3];
310 chk_addr_ret = inet_addr_type(net, v4addr);
311 if (!sysctl_ip_nonlocal_bind &&
312 !(inet->freebind || inet->transparent) &&
313 v4addr != htonl(INADDR_ANY) &&
314 chk_addr_ret != RTN_LOCAL &&
315 chk_addr_ret != RTN_MULTICAST &&
316 chk_addr_ret != RTN_BROADCAST) {
317 err = -EADDRNOTAVAIL;
318 goto out;
319 }
320 } else {
321 if (addr_type != IPV6_ADDR_ANY) {
322 struct net_device *dev = NULL;
323
324 rcu_read_lock();
325 if (addr_type & IPV6_ADDR_LINKLOCAL) {
326 if (addr_len >= sizeof(struct sockaddr_in6) &&
327 addr->sin6_scope_id) {
328 /* Override any existing binding, if another one
329 * is supplied by user.
330 */
331 sk->sk_bound_dev_if = addr->sin6_scope_id;
332 }
333
334 /* Binding to link-local address requires an interface */
335 if (!sk->sk_bound_dev_if) {
336 err = -EINVAL;
337 goto out_unlock;
338 }
339 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
340 if (!dev) {
341 err = -ENODEV;
342 goto out_unlock;
343 }
344 }
345
346 /* ipv4 addr of the socket is invalid. Only the
347 * unspecified and mapped address have a v4 equivalent.
348 */
349 v4addr = LOOPBACK4_IPV6;
350 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
351 if (!(inet->freebind || inet->transparent) &&
352 !ipv6_chk_addr(net, &addr->sin6_addr,
353 dev, 0)) {
354 err = -EADDRNOTAVAIL;
355 goto out_unlock;
356 }
357 }
358 rcu_read_unlock();
359 }
360 }
361
362 inet->inet_rcv_saddr = v4addr;
363 inet->inet_saddr = v4addr;
364
365 np->rcv_saddr = addr->sin6_addr;
366
367 if (!(addr_type & IPV6_ADDR_MULTICAST))
368 np->saddr = addr->sin6_addr;
369
370 /* Make sure we are allowed to bind here. */
371 if (sk->sk_prot->get_port(sk, snum)) {
372 inet_reset_saddr(sk);
373 err = -EADDRINUSE;
374 goto out;
375 }
376
377 if (addr_type != IPV6_ADDR_ANY) {
378 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
379 if (addr_type != IPV6_ADDR_MAPPED)
380 np->ipv6only = 1;
381 }
382 if (snum)
383 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
384 inet->inet_sport = htons(inet->inet_num);
385 inet->inet_dport = 0;
386 inet->inet_daddr = 0;
387out:
388 release_sock(sk);
389 return err;
390out_unlock:
391 rcu_read_unlock();
392 goto out;
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}
411EXPORT_SYMBOL(inet6_release);
412
413void inet6_destroy_sock(struct sock *sk)
414{
415 struct ipv6_pinfo *np = inet6_sk(sk);
416 struct sk_buff *skb;
417 struct ipv6_txoptions *opt;
418
419 /* Release rx options */
420
421 skb = xchg(&np->pktoptions, NULL);
422 if (skb != NULL)
423 kfree_skb(skb);
424
425 skb = xchg(&np->rxpmtu, NULL);
426 if (skb != NULL)
427 kfree_skb(skb);
428
429 /* Free flowlabels */
430 fl6_free_socklist(sk);
431
432 /* Free tx options */
433
434 opt = xchg(&np->opt, NULL);
435 if (opt != NULL)
436 sock_kfree_s(sk, opt, opt->tot_len);
437}
438EXPORT_SYMBOL_GPL(inet6_destroy_sock);
439
440/*
441 * This does both peername and sockname.
442 */
443
444int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
445 int *uaddr_len, int peer)
446{
447 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
448 struct sock *sk = sock->sk;
449 struct inet_sock *inet = inet_sk(sk);
450 struct ipv6_pinfo *np = inet6_sk(sk);
451
452 sin->sin6_family = AF_INET6;
453 sin->sin6_flowinfo = 0;
454 sin->sin6_scope_id = 0;
455 if (peer) {
456 if (!inet->inet_dport)
457 return -ENOTCONN;
458 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
459 peer == 1)
460 return -ENOTCONN;
461 sin->sin6_port = inet->inet_dport;
462 sin->sin6_addr = np->daddr;
463 if (np->sndflow)
464 sin->sin6_flowinfo = np->flow_label;
465 } else {
466 if (ipv6_addr_any(&np->rcv_saddr))
467 sin->sin6_addr = np->saddr;
468 else
469 sin->sin6_addr = np->rcv_saddr;
470
471 sin->sin6_port = inet->inet_sport;
472 }
473 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
474 sin->sin6_scope_id = sk->sk_bound_dev_if;
475 *uaddr_len = sizeof(*sin);
476 return 0;
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 case SIOCGSTAMP:
487 return sock_get_timestamp(sk, (struct timeval __user *)arg);
488
489 case SIOCGSTAMPNS:
490 return sock_get_timestampns(sk, (struct timespec __user *)arg);
491
492 case SIOCADDRT:
493 case SIOCDELRT:
494
495 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
496
497 case SIOCSIFADDR:
498 return addrconf_add_ifaddr(net, (void __user *) arg);
499 case SIOCDIFADDR:
500 return addrconf_del_ifaddr(net, (void __user *) arg);
501 case SIOCSIFDSTADDR:
502 return addrconf_set_dstaddr(net, (void __user *) arg);
503 default:
504 if (!sk->sk_prot->ioctl)
505 return -ENOIOCTLCMD;
506 return sk->sk_prot->ioctl(sk, cmd, arg);
507 }
508 /*NOTREACHED*/
509 return 0;
510}
511EXPORT_SYMBOL(inet6_ioctl);
512
513const struct proto_ops inet6_stream_ops = {
514 .family = PF_INET6,
515 .owner = THIS_MODULE,
516 .release = inet6_release,
517 .bind = inet6_bind,
518 .connect = inet_stream_connect, /* ok */
519 .socketpair = sock_no_socketpair, /* a do nothing */
520 .accept = inet_accept, /* ok */
521 .getname = inet6_getname,
522 .poll = tcp_poll, /* ok */
523 .ioctl = inet6_ioctl, /* must change */
524 .listen = inet_listen, /* ok */
525 .shutdown = inet_shutdown, /* ok */
526 .setsockopt = sock_common_setsockopt, /* ok */
527 .getsockopt = sock_common_getsockopt, /* ok */
528 .sendmsg = inet_sendmsg, /* ok */
529 .recvmsg = inet_recvmsg, /* ok */
530 .mmap = sock_no_mmap,
531 .sendpage = inet_sendpage,
532 .splice_read = tcp_splice_read,
533#ifdef CONFIG_COMPAT
534 .compat_setsockopt = compat_sock_common_setsockopt,
535 .compat_getsockopt = compat_sock_common_getsockopt,
536#endif
537};
538
539const struct proto_ops inet6_dgram_ops = {
540 .family = PF_INET6,
541 .owner = THIS_MODULE,
542 .release = inet6_release,
543 .bind = inet6_bind,
544 .connect = inet_dgram_connect, /* ok */
545 .socketpair = sock_no_socketpair, /* a do nothing */
546 .accept = sock_no_accept, /* a do nothing */
547 .getname = inet6_getname,
548 .poll = udp_poll, /* ok */
549 .ioctl = inet6_ioctl, /* must change */
550 .listen = sock_no_listen, /* ok */
551 .shutdown = inet_shutdown, /* ok */
552 .setsockopt = sock_common_setsockopt, /* ok */
553 .getsockopt = sock_common_getsockopt, /* ok */
554 .sendmsg = inet_sendmsg, /* ok */
555 .recvmsg = inet_recvmsg, /* ok */
556 .mmap = sock_no_mmap,
557 .sendpage = sock_no_sendpage,
558#ifdef CONFIG_COMPAT
559 .compat_setsockopt = compat_sock_common_setsockopt,
560 .compat_getsockopt = compat_sock_common_getsockopt,
561#endif
562};
563
564static const struct net_proto_family inet6_family_ops = {
565 .family = PF_INET6,
566 .create = inet6_create,
567 .owner = THIS_MODULE,
568};
569
570int inet6_register_protosw(struct inet_protosw *p)
571{
572 struct list_head *lh;
573 struct inet_protosw *answer;
574 struct list_head *last_perm;
575 int protocol = p->protocol;
576 int ret;
577
578 spin_lock_bh(&inetsw6_lock);
579
580 ret = -EINVAL;
581 if (p->type >= SOCK_MAX)
582 goto out_illegal;
583
584 /* If we are trying to override a permanent protocol, bail. */
585 answer = NULL;
586 ret = -EPERM;
587 last_perm = &inetsw6[p->type];
588 list_for_each(lh, &inetsw6[p->type]) {
589 answer = list_entry(lh, struct inet_protosw, list);
590
591 /* Check only the non-wild match. */
592 if (INET_PROTOSW_PERMANENT & answer->flags) {
593 if (protocol == answer->protocol)
594 break;
595 last_perm = lh;
596 }
597
598 answer = NULL;
599 }
600 if (answer)
601 goto out_permanent;
602
603 /* Add the new entry after the last permanent entry if any, so that
604 * the new entry does not override a permanent entry when matched with
605 * a wild-card protocol. But it is allowed to override any existing
606 * non-permanent entry. This means that when we remove this entry, the
607 * system automatically returns to the old behavior.
608 */
609 list_add_rcu(&p->list, last_perm);
610 ret = 0;
611out:
612 spin_unlock_bh(&inetsw6_lock);
613 return ret;
614
615out_permanent:
616 pr_err("Attempt to override permanent protocol %d\n", protocol);
617 goto out;
618
619out_illegal:
620 pr_err("Ignoring attempt to register invalid socket type %d\n",
621 p->type);
622 goto out;
623}
624EXPORT_SYMBOL(inet6_register_protosw);
625
626void
627inet6_unregister_protosw(struct inet_protosw *p)
628{
629 if (INET_PROTOSW_PERMANENT & p->flags) {
630 pr_err("Attempt to unregister permanent protocol %d\n",
631 p->protocol);
632 } else {
633 spin_lock_bh(&inetsw6_lock);
634 list_del_rcu(&p->list);
635 spin_unlock_bh(&inetsw6_lock);
636
637 synchronize_net();
638 }
639}
640EXPORT_SYMBOL(inet6_unregister_protosw);
641
642int inet6_sk_rebuild_header(struct sock *sk)
643{
644 struct ipv6_pinfo *np = inet6_sk(sk);
645 struct dst_entry *dst;
646
647 dst = __sk_dst_check(sk, np->dst_cookie);
648
649 if (dst == NULL) {
650 struct inet_sock *inet = inet_sk(sk);
651 struct in6_addr *final_p, final;
652 struct flowi6 fl6;
653
654 memset(&fl6, 0, sizeof(fl6));
655 fl6.flowi6_proto = sk->sk_protocol;
656 fl6.daddr = np->daddr;
657 fl6.saddr = np->saddr;
658 fl6.flowlabel = np->flow_label;
659 fl6.flowi6_oif = sk->sk_bound_dev_if;
660 fl6.flowi6_mark = sk->sk_mark;
661 fl6.fl6_dport = inet->inet_dport;
662 fl6.fl6_sport = inet->inet_sport;
663 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
664
665 final_p = fl6_update_dst(&fl6, np->opt, &final);
666
667 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
668 if (IS_ERR(dst)) {
669 sk->sk_route_caps = 0;
670 sk->sk_err_soft = -PTR_ERR(dst);
671 return PTR_ERR(dst);
672 }
673
674 __ip6_dst_store(sk, dst, NULL, NULL);
675 }
676
677 return 0;
678}
679EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
680
681bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
682{
683 const struct ipv6_pinfo *np = inet6_sk(sk);
684 const struct inet6_skb_parm *opt = IP6CB(skb);
685
686 if (np->rxopt.all) {
687 if ((opt->hop && (np->rxopt.bits.hopopts ||
688 np->rxopt.bits.ohopopts)) ||
689 ((IPV6_FLOWINFO_MASK &
690 *(__be32 *)skb_network_header(skb)) &&
691 np->rxopt.bits.rxflow) ||
692 (opt->srcrt && (np->rxopt.bits.srcrt ||
693 np->rxopt.bits.osrcrt)) ||
694 ((opt->dst1 || opt->dst0) &&
695 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
696 return true;
697 }
698 return false;
699}
700EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
701
702static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
703{
704 const struct inet6_protocol *ops = NULL;
705
706 for (;;) {
707 struct ipv6_opt_hdr *opth;
708 int len;
709
710 if (proto != NEXTHDR_HOP) {
711 ops = rcu_dereference(inet6_protos[proto]);
712
713 if (unlikely(!ops))
714 break;
715
716 if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
717 break;
718 }
719
720 if (unlikely(!pskb_may_pull(skb, 8)))
721 break;
722
723 opth = (void *)skb->data;
724 len = ipv6_optlen(opth);
725
726 if (unlikely(!pskb_may_pull(skb, len)))
727 break;
728
729 proto = opth->nexthdr;
730 __skb_pull(skb, len);
731 }
732
733 return proto;
734}
735
736static int ipv6_gso_send_check(struct sk_buff *skb)
737{
738 const struct ipv6hdr *ipv6h;
739 const struct inet6_protocol *ops;
740 int err = -EINVAL;
741
742 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
743 goto out;
744
745 ipv6h = ipv6_hdr(skb);
746 __skb_pull(skb, sizeof(*ipv6h));
747 err = -EPROTONOSUPPORT;
748
749 rcu_read_lock();
750 ops = rcu_dereference(inet6_protos[
751 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
752
753 if (likely(ops && ops->gso_send_check)) {
754 skb_reset_transport_header(skb);
755 err = ops->gso_send_check(skb);
756 }
757 rcu_read_unlock();
758
759out:
760 return err;
761}
762
763static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb,
764 netdev_features_t features)
765{
766 struct sk_buff *segs = ERR_PTR(-EINVAL);
767 struct ipv6hdr *ipv6h;
768 const struct inet6_protocol *ops;
769 int proto;
770 struct frag_hdr *fptr;
771 unsigned int unfrag_ip6hlen;
772 u8 *prevhdr;
773 int offset = 0;
774
775 if (!(features & NETIF_F_V6_CSUM))
776 features &= ~NETIF_F_SG;
777
778 if (unlikely(skb_shinfo(skb)->gso_type &
779 ~(SKB_GSO_UDP |
780 SKB_GSO_DODGY |
781 SKB_GSO_TCP_ECN |
782 SKB_GSO_TCPV6 |
783 0)))
784 goto out;
785
786 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
787 goto out;
788
789 ipv6h = ipv6_hdr(skb);
790 __skb_pull(skb, sizeof(*ipv6h));
791 segs = ERR_PTR(-EPROTONOSUPPORT);
792
793 proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
794 rcu_read_lock();
795 ops = rcu_dereference(inet6_protos[proto]);
796 if (likely(ops && ops->gso_segment)) {
797 skb_reset_transport_header(skb);
798 segs = ops->gso_segment(skb, features);
799 }
800 rcu_read_unlock();
801
802 if (IS_ERR(segs))
803 goto out;
804
805 for (skb = segs; skb; skb = skb->next) {
806 ipv6h = ipv6_hdr(skb);
807 ipv6h->payload_len = htons(skb->len - skb->mac_len -
808 sizeof(*ipv6h));
809 if (proto == IPPROTO_UDP) {
810 unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
811 fptr = (struct frag_hdr *)(skb_network_header(skb) +
812 unfrag_ip6hlen);
813 fptr->frag_off = htons(offset);
814 if (skb->next != NULL)
815 fptr->frag_off |= htons(IP6_MF);
816 offset += (ntohs(ipv6h->payload_len) -
817 sizeof(struct frag_hdr));
818 }
819 }
820
821out:
822 return segs;
823}
824
825struct ipv6_gro_cb {
826 struct napi_gro_cb napi;
827 int proto;
828};
829
830#define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
831
832static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
833 struct sk_buff *skb)
834{
835 const struct inet6_protocol *ops;
836 struct sk_buff **pp = NULL;
837 struct sk_buff *p;
838 struct ipv6hdr *iph;
839 unsigned int nlen;
840 unsigned int hlen;
841 unsigned int off;
842 int flush = 1;
843 int proto;
844 __wsum csum;
845
846 off = skb_gro_offset(skb);
847 hlen = off + sizeof(*iph);
848 iph = skb_gro_header_fast(skb, off);
849 if (skb_gro_header_hard(skb, hlen)) {
850 iph = skb_gro_header_slow(skb, hlen, off);
851 if (unlikely(!iph))
852 goto out;
853 }
854
855 skb_gro_pull(skb, sizeof(*iph));
856 skb_set_transport_header(skb, skb_gro_offset(skb));
857
858 flush += ntohs(iph->payload_len) != skb_gro_len(skb);
859
860 rcu_read_lock();
861 proto = iph->nexthdr;
862 ops = rcu_dereference(inet6_protos[proto]);
863 if (!ops || !ops->gro_receive) {
864 __pskb_pull(skb, skb_gro_offset(skb));
865 proto = ipv6_gso_pull_exthdrs(skb, proto);
866 skb_gro_pull(skb, -skb_transport_offset(skb));
867 skb_reset_transport_header(skb);
868 __skb_push(skb, skb_gro_offset(skb));
869
870 ops = rcu_dereference(inet6_protos[proto]);
871 if (!ops || !ops->gro_receive)
872 goto out_unlock;
873
874 iph = ipv6_hdr(skb);
875 }
876
877 IPV6_GRO_CB(skb)->proto = proto;
878
879 flush--;
880 nlen = skb_network_header_len(skb);
881
882 for (p = *head; p; p = p->next) {
883 struct ipv6hdr *iph2;
884
885 if (!NAPI_GRO_CB(p)->same_flow)
886 continue;
887
888 iph2 = ipv6_hdr(p);
889
890 /* All fields must match except length. */
891 if (nlen != skb_network_header_len(p) ||
892 memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
893 memcmp(&iph->nexthdr, &iph2->nexthdr,
894 nlen - offsetof(struct ipv6hdr, nexthdr))) {
895 NAPI_GRO_CB(p)->same_flow = 0;
896 continue;
897 }
898
899 NAPI_GRO_CB(p)->flush |= flush;
900 }
901
902 NAPI_GRO_CB(skb)->flush |= flush;
903
904 csum = skb->csum;
905 skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
906
907 pp = ops->gro_receive(head, skb);
908
909 skb->csum = csum;
910
911out_unlock:
912 rcu_read_unlock();
913
914out:
915 NAPI_GRO_CB(skb)->flush |= flush;
916
917 return pp;
918}
919
920static int ipv6_gro_complete(struct sk_buff *skb)
921{
922 const struct inet6_protocol *ops;
923 struct ipv6hdr *iph = ipv6_hdr(skb);
924 int err = -ENOSYS;
925
926 iph->payload_len = htons(skb->len - skb_network_offset(skb) -
927 sizeof(*iph));
928
929 rcu_read_lock();
930 ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
931 if (WARN_ON(!ops || !ops->gro_complete))
932 goto out_unlock;
933
934 err = ops->gro_complete(skb);
935
936out_unlock:
937 rcu_read_unlock();
938
939 return err;
940}
941
942static struct packet_type ipv6_packet_type __read_mostly = {
943 .type = cpu_to_be16(ETH_P_IPV6),
944 .func = ipv6_rcv,
945 .gso_send_check = ipv6_gso_send_check,
946 .gso_segment = ipv6_gso_segment,
947 .gro_receive = ipv6_gro_receive,
948 .gro_complete = ipv6_gro_complete,
949};
950
951static int __init ipv6_packet_init(void)
952{
953 dev_add_pack(&ipv6_packet_type);
954 return 0;
955}
956
957static void ipv6_packet_cleanup(void)
958{
959 dev_remove_pack(&ipv6_packet_type);
960}
961
962static int __net_init ipv6_init_mibs(struct net *net)
963{
964 if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
965 sizeof(struct udp_mib),
966 __alignof__(struct udp_mib)) < 0)
967 return -ENOMEM;
968 if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
969 sizeof(struct udp_mib),
970 __alignof__(struct udp_mib)) < 0)
971 goto err_udplite_mib;
972 if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
973 sizeof(struct ipstats_mib),
974 __alignof__(struct ipstats_mib)) < 0)
975 goto err_ip_mib;
976 if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
977 sizeof(struct icmpv6_mib),
978 __alignof__(struct icmpv6_mib)) < 0)
979 goto err_icmp_mib;
980 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
981 GFP_KERNEL);
982 if (!net->mib.icmpv6msg_statistics)
983 goto err_icmpmsg_mib;
984 return 0;
985
986err_icmpmsg_mib:
987 snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
988err_icmp_mib:
989 snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
990err_ip_mib:
991 snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
992err_udplite_mib:
993 snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
994 return -ENOMEM;
995}
996
997static void ipv6_cleanup_mibs(struct net *net)
998{
999 snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1000 snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1001 snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1002 snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1003 kfree(net->mib.icmpv6msg_statistics);
1004}
1005
1006static int __net_init inet6_net_init(struct net *net)
1007{
1008 int err = 0;
1009
1010 net->ipv6.sysctl.bindv6only = 0;
1011 net->ipv6.sysctl.icmpv6_time = 1*HZ;
1012
1013 err = ipv6_init_mibs(net);
1014 if (err)
1015 return err;
1016#ifdef CONFIG_PROC_FS
1017 err = udp6_proc_init(net);
1018 if (err)
1019 goto out;
1020 err = tcp6_proc_init(net);
1021 if (err)
1022 goto proc_tcp6_fail;
1023 err = ac6_proc_init(net);
1024 if (err)
1025 goto proc_ac6_fail;
1026#endif
1027 return err;
1028
1029#ifdef CONFIG_PROC_FS
1030proc_ac6_fail:
1031 tcp6_proc_exit(net);
1032proc_tcp6_fail:
1033 udp6_proc_exit(net);
1034out:
1035 ipv6_cleanup_mibs(net);
1036 return err;
1037#endif
1038}
1039
1040static void __net_exit inet6_net_exit(struct net *net)
1041{
1042#ifdef CONFIG_PROC_FS
1043 udp6_proc_exit(net);
1044 tcp6_proc_exit(net);
1045 ac6_proc_exit(net);
1046#endif
1047 ipv6_cleanup_mibs(net);
1048}
1049
1050static struct pernet_operations inet6_net_ops = {
1051 .init = inet6_net_init,
1052 .exit = inet6_net_exit,
1053};
1054
1055static int __init inet6_init(void)
1056{
1057 struct sk_buff *dummy_skb;
1058 struct list_head *r;
1059 int err = 0;
1060
1061 BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1062
1063 /* Register the socket-side information for inet6_create. */
1064 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1065 INIT_LIST_HEAD(r);
1066
1067 if (disable_ipv6_mod) {
1068 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1069 goto out;
1070 }
1071
1072 err = proto_register(&tcpv6_prot, 1);
1073 if (err)
1074 goto out;
1075
1076 err = proto_register(&udpv6_prot, 1);
1077 if (err)
1078 goto out_unregister_tcp_proto;
1079
1080 err = proto_register(&udplitev6_prot, 1);
1081 if (err)
1082 goto out_unregister_udp_proto;
1083
1084 err = proto_register(&rawv6_prot, 1);
1085 if (err)
1086 goto out_unregister_udplite_proto;
1087
1088
1089 /* We MUST register RAW sockets before we create the ICMP6,
1090 * IGMP6, or NDISC control sockets.
1091 */
1092 err = rawv6_init();
1093 if (err)
1094 goto out_unregister_raw_proto;
1095
1096 /* Register the family here so that the init calls below will
1097 * be able to create sockets. (?? is this dangerous ??)
1098 */
1099 err = sock_register(&inet6_family_ops);
1100 if (err)
1101 goto out_sock_register_fail;
1102
1103 tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
1104
1105 /*
1106 * ipngwg API draft makes clear that the correct semantics
1107 * for TCP and UDP is to consider one TCP and UDP instance
1108 * in a host available by both INET and INET6 APIs and
1109 * able to communicate via both network protocols.
1110 */
1111
1112 err = register_pernet_subsys(&inet6_net_ops);
1113 if (err)
1114 goto register_pernet_fail;
1115 err = icmpv6_init();
1116 if (err)
1117 goto icmp_fail;
1118 err = ip6_mr_init();
1119 if (err)
1120 goto ipmr_fail;
1121 err = ndisc_init();
1122 if (err)
1123 goto ndisc_fail;
1124 err = igmp6_init();
1125 if (err)
1126 goto igmp_fail;
1127 err = ipv6_netfilter_init();
1128 if (err)
1129 goto netfilter_fail;
1130 /* Create /proc/foo6 entries. */
1131#ifdef CONFIG_PROC_FS
1132 err = -ENOMEM;
1133 if (raw6_proc_init())
1134 goto proc_raw6_fail;
1135 if (udplite6_proc_init())
1136 goto proc_udplite6_fail;
1137 if (ipv6_misc_proc_init())
1138 goto proc_misc6_fail;
1139 if (if6_proc_init())
1140 goto proc_if6_fail;
1141#endif
1142 err = ip6_route_init();
1143 if (err)
1144 goto ip6_route_fail;
1145 err = ip6_flowlabel_init();
1146 if (err)
1147 goto ip6_flowlabel_fail;
1148 err = addrconf_init();
1149 if (err)
1150 goto addrconf_fail;
1151
1152 /* Init v6 extension headers. */
1153 err = ipv6_exthdrs_init();
1154 if (err)
1155 goto ipv6_exthdrs_fail;
1156
1157 err = ipv6_frag_init();
1158 if (err)
1159 goto ipv6_frag_fail;
1160
1161 /* Init v6 transport protocols. */
1162 err = udpv6_init();
1163 if (err)
1164 goto udpv6_fail;
1165
1166 err = udplitev6_init();
1167 if (err)
1168 goto udplitev6_fail;
1169
1170 err = tcpv6_init();
1171 if (err)
1172 goto tcpv6_fail;
1173
1174 err = ipv6_packet_init();
1175 if (err)
1176 goto ipv6_packet_fail;
1177
1178#ifdef CONFIG_SYSCTL
1179 err = ipv6_sysctl_register();
1180 if (err)
1181 goto sysctl_fail;
1182#endif
1183out:
1184 return err;
1185
1186#ifdef CONFIG_SYSCTL
1187sysctl_fail:
1188 ipv6_packet_cleanup();
1189#endif
1190ipv6_packet_fail:
1191 tcpv6_exit();
1192tcpv6_fail:
1193 udplitev6_exit();
1194udplitev6_fail:
1195 udpv6_exit();
1196udpv6_fail:
1197 ipv6_frag_exit();
1198ipv6_frag_fail:
1199 ipv6_exthdrs_exit();
1200ipv6_exthdrs_fail:
1201 addrconf_cleanup();
1202addrconf_fail:
1203 ip6_flowlabel_cleanup();
1204ip6_flowlabel_fail:
1205 ip6_route_cleanup();
1206ip6_route_fail:
1207#ifdef CONFIG_PROC_FS
1208 if6_proc_exit();
1209proc_if6_fail:
1210 ipv6_misc_proc_exit();
1211proc_misc6_fail:
1212 udplite6_proc_exit();
1213proc_udplite6_fail:
1214 raw6_proc_exit();
1215proc_raw6_fail:
1216#endif
1217 ipv6_netfilter_fini();
1218netfilter_fail:
1219 igmp6_cleanup();
1220igmp_fail:
1221 ndisc_cleanup();
1222ndisc_fail:
1223 ip6_mr_cleanup();
1224ipmr_fail:
1225 icmpv6_cleanup();
1226icmp_fail:
1227 unregister_pernet_subsys(&inet6_net_ops);
1228register_pernet_fail:
1229 sock_unregister(PF_INET6);
1230 rtnl_unregister_all(PF_INET6);
1231out_sock_register_fail:
1232 rawv6_exit();
1233out_unregister_raw_proto:
1234 proto_unregister(&rawv6_prot);
1235out_unregister_udplite_proto:
1236 proto_unregister(&udplitev6_prot);
1237out_unregister_udp_proto:
1238 proto_unregister(&udpv6_prot);
1239out_unregister_tcp_proto:
1240 proto_unregister(&tcpv6_prot);
1241 goto out;
1242}
1243module_init(inet6_init);
1244
1245static void __exit inet6_exit(void)
1246{
1247 if (disable_ipv6_mod)
1248 return;
1249
1250 /* First of all disallow new sockets creation. */
1251 sock_unregister(PF_INET6);
1252 /* Disallow any further netlink messages */
1253 rtnl_unregister_all(PF_INET6);
1254
1255 udpv6_exit();
1256 udplitev6_exit();
1257 tcpv6_exit();
1258
1259 /* Cleanup code parts. */
1260 ipv6_packet_cleanup();
1261 ipv6_frag_exit();
1262 ipv6_exthdrs_exit();
1263 addrconf_cleanup();
1264 ip6_flowlabel_cleanup();
1265 ip6_route_cleanup();
1266#ifdef CONFIG_PROC_FS
1267
1268 /* Cleanup code parts. */
1269 if6_proc_exit();
1270 ipv6_misc_proc_exit();
1271 udplite6_proc_exit();
1272 raw6_proc_exit();
1273#endif
1274 ipv6_netfilter_fini();
1275 igmp6_cleanup();
1276 ndisc_cleanup();
1277 ip6_mr_cleanup();
1278 icmpv6_cleanup();
1279 rawv6_exit();
1280
1281 unregister_pernet_subsys(&inet6_net_ops);
1282 proto_unregister(&rawv6_prot);
1283 proto_unregister(&udplitev6_prot);
1284 proto_unregister(&udpv6_prot);
1285 proto_unregister(&tcpv6_prot);
1286
1287 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1288}
1289module_exit(inet6_exit);
1290
1291MODULE_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#include <net/ioam6.h>
66#include <net/rawv6.h>
67
68#include <linux/uaccess.h>
69#include <linux/mroute6.h>
70
71#include "ip6_offload.h"
72
73MODULE_AUTHOR("Cast of dozens");
74MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
75MODULE_LICENSE("GPL");
76
77/* The inetsw6 table contains everything that inet6_create needs to
78 * build a new socket.
79 */
80static struct list_head inetsw6[SOCK_MAX];
81static DEFINE_SPINLOCK(inetsw6_lock);
82
83struct ipv6_params ipv6_defaults = {
84 .disable_ipv6 = 0,
85 .autoconf = 1,
86};
87
88static int disable_ipv6_mod;
89
90module_param_named(disable, disable_ipv6_mod, int, 0444);
91MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
92
93module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
94MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
95
96module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
97MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
98
99bool ipv6_mod_enabled(void)
100{
101 return disable_ipv6_mod == 0;
102}
103EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
104
105static struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
106{
107 const int offset = sk->sk_prot->ipv6_pinfo_offset;
108
109 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
110}
111
112void inet6_sock_destruct(struct sock *sk)
113{
114 inet6_cleanup_sock(sk);
115 inet_sock_destruct(sk);
116}
117EXPORT_SYMBOL_GPL(inet6_sock_destruct);
118
119static int inet6_create(struct net *net, struct socket *sock, int protocol,
120 int kern)
121{
122 struct inet_sock *inet;
123 struct ipv6_pinfo *np;
124 struct sock *sk;
125 struct inet_protosw *answer;
126 struct proto *answer_prot;
127 unsigned char answer_flags;
128 int try_loading_module = 0;
129 int err;
130
131 if (protocol < 0 || protocol >= IPPROTO_MAX)
132 return -EINVAL;
133
134 /* Look for the requested type/protocol pair. */
135lookup_protocol:
136 err = -ESOCKTNOSUPPORT;
137 rcu_read_lock();
138 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
139
140 err = 0;
141 /* Check the non-wild match. */
142 if (protocol == answer->protocol) {
143 if (protocol != IPPROTO_IP)
144 break;
145 } else {
146 /* Check for the two wild cases. */
147 if (IPPROTO_IP == protocol) {
148 protocol = answer->protocol;
149 break;
150 }
151 if (IPPROTO_IP == answer->protocol)
152 break;
153 }
154 err = -EPROTONOSUPPORT;
155 }
156
157 if (err) {
158 if (try_loading_module < 2) {
159 rcu_read_unlock();
160 /*
161 * Be more specific, e.g. net-pf-10-proto-132-type-1
162 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
163 */
164 if (++try_loading_module == 1)
165 request_module("net-pf-%d-proto-%d-type-%d",
166 PF_INET6, protocol, sock->type);
167 /*
168 * Fall back to generic, e.g. net-pf-10-proto-132
169 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
170 */
171 else
172 request_module("net-pf-%d-proto-%d",
173 PF_INET6, protocol);
174 goto lookup_protocol;
175 } else
176 goto out_rcu_unlock;
177 }
178
179 err = -EPERM;
180 if (sock->type == SOCK_RAW && !kern &&
181 !ns_capable(net->user_ns, CAP_NET_RAW))
182 goto out_rcu_unlock;
183
184 sock->ops = answer->ops;
185 answer_prot = answer->prot;
186 answer_flags = answer->flags;
187 rcu_read_unlock();
188
189 WARN_ON(!answer_prot->slab);
190
191 err = -ENOBUFS;
192 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
193 if (!sk)
194 goto out;
195
196 sock_init_data(sock, sk);
197
198 err = 0;
199 if (INET_PROTOSW_REUSE & answer_flags)
200 sk->sk_reuse = SK_CAN_REUSE;
201
202 if (INET_PROTOSW_ICSK & answer_flags)
203 inet_init_csk_locks(sk);
204
205 inet = inet_sk(sk);
206 inet_assign_bit(IS_ICSK, sk, INET_PROTOSW_ICSK & answer_flags);
207
208 if (SOCK_RAW == sock->type) {
209 inet->inet_num = protocol;
210 if (IPPROTO_RAW == protocol)
211 inet_set_bit(HDRINCL, sk);
212 }
213
214 sk->sk_destruct = inet6_sock_destruct;
215 sk->sk_family = PF_INET6;
216 sk->sk_protocol = protocol;
217
218 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
219
220 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
221 np->hop_limit = -1;
222 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
223 inet6_set_bit(MC6_LOOP, sk);
224 inet6_set_bit(MC6_ALL, sk);
225 np->pmtudisc = IPV6_PMTUDISC_WANT;
226 inet6_assign_bit(REPFLOW, sk, net->ipv6.sysctl.flowlabel_reflect &
227 FLOWLABEL_REFLECT_ESTABLISHED);
228 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
229 sk->sk_txrehash = READ_ONCE(net->core.sysctl_txrehash);
230
231 /* Init the ipv4 part of the socket since we can have sockets
232 * using v6 API for ipv4.
233 */
234 inet->uc_ttl = -1;
235
236 inet_set_bit(MC_LOOP, sk);
237 inet->mc_ttl = 1;
238 inet->mc_index = 0;
239 RCU_INIT_POINTER(inet->mc_list, NULL);
240 inet->rcv_tos = 0;
241
242 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc))
243 inet->pmtudisc = IP_PMTUDISC_DONT;
244 else
245 inet->pmtudisc = IP_PMTUDISC_WANT;
246
247 if (inet->inet_num) {
248 /* It assumes that any protocol which allows
249 * the user to assign a number at socket
250 * creation time automatically shares.
251 */
252 inet->inet_sport = htons(inet->inet_num);
253 err = sk->sk_prot->hash(sk);
254 if (err) {
255 sk_common_release(sk);
256 goto out;
257 }
258 }
259 if (sk->sk_prot->init) {
260 err = sk->sk_prot->init(sk);
261 if (err) {
262 sk_common_release(sk);
263 goto out;
264 }
265 }
266
267 if (!kern) {
268 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
269 if (err) {
270 sk_common_release(sk);
271 goto out;
272 }
273 }
274out:
275 return err;
276out_rcu_unlock:
277 rcu_read_unlock();
278 goto out;
279}
280
281static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
282 u32 flags)
283{
284 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
285 struct inet_sock *inet = inet_sk(sk);
286 struct ipv6_pinfo *np = inet6_sk(sk);
287 struct net *net = sock_net(sk);
288 __be32 v4addr = 0;
289 unsigned short snum;
290 bool saved_ipv6only;
291 int addr_type = 0;
292 int err = 0;
293
294 if (addr->sin6_family != AF_INET6)
295 return -EAFNOSUPPORT;
296
297 addr_type = ipv6_addr_type(&addr->sin6_addr);
298 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
299 return -EINVAL;
300
301 snum = ntohs(addr->sin6_port);
302 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
303 snum && inet_port_requires_bind_service(net, snum) &&
304 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
305 return -EACCES;
306
307 if (flags & BIND_WITH_LOCK)
308 lock_sock(sk);
309
310 /* Check these errors (active socket, double bind). */
311 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
312 err = -EINVAL;
313 goto out;
314 }
315
316 /* Check if the address belongs to the host. */
317 if (addr_type == IPV6_ADDR_MAPPED) {
318 struct net_device *dev = NULL;
319 int chk_addr_ret;
320
321 /* Binding to v4-mapped address on a v6-only socket
322 * makes no sense
323 */
324 if (ipv6_only_sock(sk)) {
325 err = -EINVAL;
326 goto out;
327 }
328
329 rcu_read_lock();
330 if (sk->sk_bound_dev_if) {
331 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
332 if (!dev) {
333 err = -ENODEV;
334 goto out_unlock;
335 }
336 }
337
338 /* Reproduce AF_INET checks to make the bindings consistent */
339 v4addr = addr->sin6_addr.s6_addr32[3];
340 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
341 rcu_read_unlock();
342
343 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
344 chk_addr_ret)) {
345 err = -EADDRNOTAVAIL;
346 goto out;
347 }
348 } else {
349 if (addr_type != IPV6_ADDR_ANY) {
350 struct net_device *dev = NULL;
351
352 rcu_read_lock();
353 if (__ipv6_addr_needs_scope_id(addr_type)) {
354 if (addr_len >= sizeof(struct sockaddr_in6) &&
355 addr->sin6_scope_id) {
356 /* Override any existing binding, if another one
357 * is supplied by user.
358 */
359 sk->sk_bound_dev_if = addr->sin6_scope_id;
360 }
361
362 /* Binding to link-local address requires an interface */
363 if (!sk->sk_bound_dev_if) {
364 err = -EINVAL;
365 goto out_unlock;
366 }
367 }
368
369 if (sk->sk_bound_dev_if) {
370 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
371 if (!dev) {
372 err = -ENODEV;
373 goto out_unlock;
374 }
375 }
376
377 /* ipv4 addr of the socket is invalid. Only the
378 * unspecified and mapped address have a v4 equivalent.
379 */
380 v4addr = LOOPBACK4_IPV6;
381 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
382 if (!ipv6_can_nonlocal_bind(net, inet) &&
383 !ipv6_chk_addr(net, &addr->sin6_addr,
384 dev, 0)) {
385 err = -EADDRNOTAVAIL;
386 goto out_unlock;
387 }
388 }
389 rcu_read_unlock();
390 }
391 }
392
393 inet->inet_rcv_saddr = v4addr;
394 inet->inet_saddr = v4addr;
395
396 sk->sk_v6_rcv_saddr = addr->sin6_addr;
397
398 if (!(addr_type & IPV6_ADDR_MULTICAST))
399 np->saddr = addr->sin6_addr;
400
401 saved_ipv6only = sk->sk_ipv6only;
402 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
403 sk->sk_ipv6only = 1;
404
405 /* Make sure we are allowed to bind here. */
406 if (snum || !(inet_test_bit(BIND_ADDRESS_NO_PORT, sk) ||
407 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
408 err = sk->sk_prot->get_port(sk, snum);
409 if (err) {
410 sk->sk_ipv6only = saved_ipv6only;
411 inet_reset_saddr(sk);
412 goto out;
413 }
414 if (!(flags & BIND_FROM_BPF)) {
415 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
416 if (err) {
417 sk->sk_ipv6only = saved_ipv6only;
418 inet_reset_saddr(sk);
419 if (sk->sk_prot->put_port)
420 sk->sk_prot->put_port(sk);
421 goto out;
422 }
423 }
424 }
425
426 if (addr_type != IPV6_ADDR_ANY)
427 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
428 if (snum)
429 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
430 inet->inet_sport = htons(inet->inet_num);
431 inet->inet_dport = 0;
432 inet->inet_daddr = 0;
433out:
434 if (flags & BIND_WITH_LOCK)
435 release_sock(sk);
436 return err;
437out_unlock:
438 rcu_read_unlock();
439 goto out;
440}
441
442int inet6_bind_sk(struct sock *sk, struct sockaddr *uaddr, int addr_len)
443{
444 u32 flags = BIND_WITH_LOCK;
445 const struct proto *prot;
446 int err = 0;
447
448 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
449 prot = READ_ONCE(sk->sk_prot);
450 /* If the socket has its own bind function then use it. */
451 if (prot->bind)
452 return prot->bind(sk, uaddr, addr_len);
453
454 if (addr_len < SIN6_LEN_RFC2133)
455 return -EINVAL;
456
457 /* BPF prog is run before any checks are done so that if the prog
458 * changes context in a wrong way it will be caught.
459 */
460 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, &addr_len,
461 CGROUP_INET6_BIND, &flags);
462 if (err)
463 return err;
464
465 return __inet6_bind(sk, uaddr, addr_len, flags);
466}
467
468/* bind for INET6 API */
469int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
470{
471 return inet6_bind_sk(sock->sk, uaddr, addr_len);
472}
473EXPORT_SYMBOL(inet6_bind);
474
475int inet6_release(struct socket *sock)
476{
477 struct sock *sk = sock->sk;
478
479 if (!sk)
480 return -EINVAL;
481
482 /* Free mc lists */
483 ipv6_sock_mc_close(sk);
484
485 /* Free ac lists */
486 ipv6_sock_ac_close(sk);
487
488 return inet_release(sock);
489}
490EXPORT_SYMBOL(inet6_release);
491
492void inet6_cleanup_sock(struct sock *sk)
493{
494 struct ipv6_pinfo *np = inet6_sk(sk);
495 struct sk_buff *skb;
496 struct ipv6_txoptions *opt;
497
498 /* Release rx options */
499
500 skb = xchg(&np->pktoptions, NULL);
501 kfree_skb(skb);
502
503 skb = xchg(&np->rxpmtu, NULL);
504 kfree_skb(skb);
505
506 /* Free flowlabels */
507 fl6_free_socklist(sk);
508
509 /* Free tx options */
510
511 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
512 if (opt) {
513 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
514 txopt_put(opt);
515 }
516}
517EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
518
519/*
520 * This does both peername and sockname.
521 */
522int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
523 int peer)
524{
525 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
526 int sin_addr_len = sizeof(*sin);
527 struct sock *sk = sock->sk;
528 struct inet_sock *inet = inet_sk(sk);
529 struct ipv6_pinfo *np = inet6_sk(sk);
530
531 sin->sin6_family = AF_INET6;
532 sin->sin6_flowinfo = 0;
533 sin->sin6_scope_id = 0;
534 lock_sock(sk);
535 if (peer) {
536 if (!inet->inet_dport ||
537 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
538 peer == 1)) {
539 release_sock(sk);
540 return -ENOTCONN;
541 }
542 sin->sin6_port = inet->inet_dport;
543 sin->sin6_addr = sk->sk_v6_daddr;
544 if (inet6_test_bit(SNDFLOW, sk))
545 sin->sin6_flowinfo = np->flow_label;
546 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, &sin_addr_len,
547 CGROUP_INET6_GETPEERNAME);
548 } else {
549 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
550 sin->sin6_addr = np->saddr;
551 else
552 sin->sin6_addr = sk->sk_v6_rcv_saddr;
553 sin->sin6_port = inet->inet_sport;
554 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, &sin_addr_len,
555 CGROUP_INET6_GETSOCKNAME);
556 }
557 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
558 sk->sk_bound_dev_if);
559 release_sock(sk);
560 return sin_addr_len;
561}
562EXPORT_SYMBOL(inet6_getname);
563
564int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
565{
566 void __user *argp = (void __user *)arg;
567 struct sock *sk = sock->sk;
568 struct net *net = sock_net(sk);
569 const struct proto *prot;
570
571 switch (cmd) {
572 case SIOCADDRT:
573 case SIOCDELRT: {
574 struct in6_rtmsg rtmsg;
575
576 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
577 return -EFAULT;
578 return ipv6_route_ioctl(net, cmd, &rtmsg);
579 }
580 case SIOCSIFADDR:
581 return addrconf_add_ifaddr(net, argp);
582 case SIOCDIFADDR:
583 return addrconf_del_ifaddr(net, argp);
584 case SIOCSIFDSTADDR:
585 return addrconf_set_dstaddr(net, argp);
586 default:
587 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
588 prot = READ_ONCE(sk->sk_prot);
589 if (!prot->ioctl)
590 return -ENOIOCTLCMD;
591 return sk_ioctl(sk, cmd, (void __user *)arg);
592 }
593 /*NOTREACHED*/
594 return 0;
595}
596EXPORT_SYMBOL(inet6_ioctl);
597
598#ifdef CONFIG_COMPAT
599struct compat_in6_rtmsg {
600 struct in6_addr rtmsg_dst;
601 struct in6_addr rtmsg_src;
602 struct in6_addr rtmsg_gateway;
603 u32 rtmsg_type;
604 u16 rtmsg_dst_len;
605 u16 rtmsg_src_len;
606 u32 rtmsg_metric;
607 u32 rtmsg_info;
608 u32 rtmsg_flags;
609 s32 rtmsg_ifindex;
610};
611
612static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
613 struct compat_in6_rtmsg __user *ur)
614{
615 struct in6_rtmsg rt;
616
617 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
618 3 * sizeof(struct in6_addr)) ||
619 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
620 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
621 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
622 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
623 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
624 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
625 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
626 return -EFAULT;
627
628
629 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
630}
631
632int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
633{
634 void __user *argp = compat_ptr(arg);
635 struct sock *sk = sock->sk;
636
637 switch (cmd) {
638 case SIOCADDRT:
639 case SIOCDELRT:
640 return inet6_compat_routing_ioctl(sk, cmd, argp);
641 default:
642 return -ENOIOCTLCMD;
643 }
644}
645EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
646#endif /* CONFIG_COMPAT */
647
648INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
649 size_t));
650int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
651{
652 struct sock *sk = sock->sk;
653 const struct proto *prot;
654
655 if (unlikely(inet_send_prepare(sk)))
656 return -EAGAIN;
657
658 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
659 prot = READ_ONCE(sk->sk_prot);
660 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
661 sk, msg, size);
662}
663
664INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
665 size_t, int, int *));
666int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
667 int flags)
668{
669 struct sock *sk = sock->sk;
670 const struct proto *prot;
671 int addr_len = 0;
672 int err;
673
674 if (likely(!(flags & MSG_ERRQUEUE)))
675 sock_rps_record_flow(sk);
676
677 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
678 prot = READ_ONCE(sk->sk_prot);
679 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
680 sk, msg, size, flags, &addr_len);
681 if (err >= 0)
682 msg->msg_namelen = addr_len;
683 return err;
684}
685
686const struct proto_ops inet6_stream_ops = {
687 .family = PF_INET6,
688 .owner = THIS_MODULE,
689 .release = inet6_release,
690 .bind = inet6_bind,
691 .connect = inet_stream_connect, /* ok */
692 .socketpair = sock_no_socketpair, /* a do nothing */
693 .accept = inet_accept, /* ok */
694 .getname = inet6_getname,
695 .poll = tcp_poll, /* ok */
696 .ioctl = inet6_ioctl, /* must change */
697 .gettstamp = sock_gettstamp,
698 .listen = inet_listen, /* ok */
699 .shutdown = inet_shutdown, /* ok */
700 .setsockopt = sock_common_setsockopt, /* ok */
701 .getsockopt = sock_common_getsockopt, /* ok */
702 .sendmsg = inet6_sendmsg, /* retpoline's sake */
703 .recvmsg = inet6_recvmsg, /* retpoline's sake */
704#ifdef CONFIG_MMU
705 .mmap = tcp_mmap,
706#endif
707 .splice_eof = inet_splice_eof,
708 .sendmsg_locked = tcp_sendmsg_locked,
709 .splice_read = tcp_splice_read,
710 .read_sock = tcp_read_sock,
711 .read_skb = tcp_read_skb,
712 .peek_len = tcp_peek_len,
713#ifdef CONFIG_COMPAT
714 .compat_ioctl = inet6_compat_ioctl,
715#endif
716 .set_rcvlowat = tcp_set_rcvlowat,
717};
718
719const struct proto_ops inet6_dgram_ops = {
720 .family = PF_INET6,
721 .owner = THIS_MODULE,
722 .release = inet6_release,
723 .bind = inet6_bind,
724 .connect = inet_dgram_connect, /* ok */
725 .socketpair = sock_no_socketpair, /* a do nothing */
726 .accept = sock_no_accept, /* a do nothing */
727 .getname = inet6_getname,
728 .poll = udp_poll, /* ok */
729 .ioctl = inet6_ioctl, /* must change */
730 .gettstamp = sock_gettstamp,
731 .listen = sock_no_listen, /* ok */
732 .shutdown = inet_shutdown, /* ok */
733 .setsockopt = sock_common_setsockopt, /* ok */
734 .getsockopt = sock_common_getsockopt, /* ok */
735 .sendmsg = inet6_sendmsg, /* retpoline's sake */
736 .recvmsg = inet6_recvmsg, /* retpoline's sake */
737 .read_skb = udp_read_skb,
738 .mmap = sock_no_mmap,
739 .set_peek_off = sk_set_peek_off,
740#ifdef CONFIG_COMPAT
741 .compat_ioctl = inet6_compat_ioctl,
742#endif
743};
744
745static const struct net_proto_family inet6_family_ops = {
746 .family = PF_INET6,
747 .create = inet6_create,
748 .owner = THIS_MODULE,
749};
750
751int inet6_register_protosw(struct inet_protosw *p)
752{
753 struct list_head *lh;
754 struct inet_protosw *answer;
755 struct list_head *last_perm;
756 int protocol = p->protocol;
757 int ret;
758
759 spin_lock_bh(&inetsw6_lock);
760
761 ret = -EINVAL;
762 if (p->type >= SOCK_MAX)
763 goto out_illegal;
764
765 /* If we are trying to override a permanent protocol, bail. */
766 answer = NULL;
767 ret = -EPERM;
768 last_perm = &inetsw6[p->type];
769 list_for_each(lh, &inetsw6[p->type]) {
770 answer = list_entry(lh, struct inet_protosw, list);
771
772 /* Check only the non-wild match. */
773 if (INET_PROTOSW_PERMANENT & answer->flags) {
774 if (protocol == answer->protocol)
775 break;
776 last_perm = lh;
777 }
778
779 answer = NULL;
780 }
781 if (answer)
782 goto out_permanent;
783
784 /* Add the new entry after the last permanent entry if any, so that
785 * the new entry does not override a permanent entry when matched with
786 * a wild-card protocol. But it is allowed to override any existing
787 * non-permanent entry. This means that when we remove this entry, the
788 * system automatically returns to the old behavior.
789 */
790 list_add_rcu(&p->list, last_perm);
791 ret = 0;
792out:
793 spin_unlock_bh(&inetsw6_lock);
794 return ret;
795
796out_permanent:
797 pr_err("Attempt to override permanent protocol %d\n", protocol);
798 goto out;
799
800out_illegal:
801 pr_err("Ignoring attempt to register invalid socket type %d\n",
802 p->type);
803 goto out;
804}
805EXPORT_SYMBOL(inet6_register_protosw);
806
807void
808inet6_unregister_protosw(struct inet_protosw *p)
809{
810 if (INET_PROTOSW_PERMANENT & p->flags) {
811 pr_err("Attempt to unregister permanent protocol %d\n",
812 p->protocol);
813 } else {
814 spin_lock_bh(&inetsw6_lock);
815 list_del_rcu(&p->list);
816 spin_unlock_bh(&inetsw6_lock);
817
818 synchronize_net();
819 }
820}
821EXPORT_SYMBOL(inet6_unregister_protosw);
822
823int inet6_sk_rebuild_header(struct sock *sk)
824{
825 struct ipv6_pinfo *np = inet6_sk(sk);
826 struct dst_entry *dst;
827
828 dst = __sk_dst_check(sk, np->dst_cookie);
829
830 if (!dst) {
831 struct inet_sock *inet = inet_sk(sk);
832 struct in6_addr *final_p, final;
833 struct flowi6 fl6;
834
835 memset(&fl6, 0, sizeof(fl6));
836 fl6.flowi6_proto = sk->sk_protocol;
837 fl6.daddr = sk->sk_v6_daddr;
838 fl6.saddr = np->saddr;
839 fl6.flowlabel = np->flow_label;
840 fl6.flowi6_oif = sk->sk_bound_dev_if;
841 fl6.flowi6_mark = sk->sk_mark;
842 fl6.fl6_dport = inet->inet_dport;
843 fl6.fl6_sport = inet->inet_sport;
844 fl6.flowi6_uid = sk->sk_uid;
845 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
846
847 rcu_read_lock();
848 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
849 &final);
850 rcu_read_unlock();
851
852 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
853 if (IS_ERR(dst)) {
854 sk->sk_route_caps = 0;
855 WRITE_ONCE(sk->sk_err_soft, -PTR_ERR(dst));
856 return PTR_ERR(dst);
857 }
858
859 ip6_dst_store(sk, dst, NULL, NULL);
860 }
861
862 return 0;
863}
864EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
865
866bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
867 const struct inet6_skb_parm *opt)
868{
869 const struct ipv6_pinfo *np = inet6_sk(sk);
870
871 if (np->rxopt.all) {
872 if (((opt->flags & IP6SKB_HOPBYHOP) &&
873 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
874 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
875 np->rxopt.bits.rxflow) ||
876 (opt->srcrt && (np->rxopt.bits.srcrt ||
877 np->rxopt.bits.osrcrt)) ||
878 ((opt->dst1 || opt->dst0) &&
879 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
880 return true;
881 }
882 return false;
883}
884EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
885
886static struct packet_type ipv6_packet_type __read_mostly = {
887 .type = cpu_to_be16(ETH_P_IPV6),
888 .func = ipv6_rcv,
889 .list_func = ipv6_list_rcv,
890};
891
892static int __init ipv6_packet_init(void)
893{
894 dev_add_pack(&ipv6_packet_type);
895 return 0;
896}
897
898static void ipv6_packet_cleanup(void)
899{
900 dev_remove_pack(&ipv6_packet_type);
901}
902
903static int __net_init ipv6_init_mibs(struct net *net)
904{
905 int i;
906
907 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
908 if (!net->mib.udp_stats_in6)
909 return -ENOMEM;
910 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
911 if (!net->mib.udplite_stats_in6)
912 goto err_udplite_mib;
913 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
914 if (!net->mib.ipv6_statistics)
915 goto err_ip_mib;
916
917 for_each_possible_cpu(i) {
918 struct ipstats_mib *af_inet6_stats;
919 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
920 u64_stats_init(&af_inet6_stats->syncp);
921 }
922
923
924 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
925 if (!net->mib.icmpv6_statistics)
926 goto err_icmp_mib;
927 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
928 GFP_KERNEL);
929 if (!net->mib.icmpv6msg_statistics)
930 goto err_icmpmsg_mib;
931 return 0;
932
933err_icmpmsg_mib:
934 free_percpu(net->mib.icmpv6_statistics);
935err_icmp_mib:
936 free_percpu(net->mib.ipv6_statistics);
937err_ip_mib:
938 free_percpu(net->mib.udplite_stats_in6);
939err_udplite_mib:
940 free_percpu(net->mib.udp_stats_in6);
941 return -ENOMEM;
942}
943
944static void ipv6_cleanup_mibs(struct net *net)
945{
946 free_percpu(net->mib.udp_stats_in6);
947 free_percpu(net->mib.udplite_stats_in6);
948 free_percpu(net->mib.ipv6_statistics);
949 free_percpu(net->mib.icmpv6_statistics);
950 kfree(net->mib.icmpv6msg_statistics);
951}
952
953static int __net_init inet6_net_init(struct net *net)
954{
955 int err = 0;
956
957 net->ipv6.sysctl.bindv6only = 0;
958 net->ipv6.sysctl.icmpv6_time = 1*HZ;
959 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
960 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
961 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
962 net->ipv6.sysctl.icmpv6_error_anycast_as_unicast = 0;
963
964 /* By default, rate limit error messages.
965 * Except for pmtu discovery, it would break it.
966 * proc_do_large_bitmap needs pointer to the bitmap.
967 */
968 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
969 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
970 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
971
972 net->ipv6.sysctl.flowlabel_consistency = 1;
973 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
974 net->ipv6.sysctl.idgen_retries = 3;
975 net->ipv6.sysctl.idgen_delay = 1 * HZ;
976 net->ipv6.sysctl.flowlabel_state_ranges = 0;
977 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
978 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
979 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
980 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
981 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
982 atomic_set(&net->ipv6.fib6_sernum, 1);
983
984 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
985 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
986
987 err = ipv6_init_mibs(net);
988 if (err)
989 return err;
990#ifdef CONFIG_PROC_FS
991 err = udp6_proc_init(net);
992 if (err)
993 goto out;
994 err = tcp6_proc_init(net);
995 if (err)
996 goto proc_tcp6_fail;
997 err = ac6_proc_init(net);
998 if (err)
999 goto proc_ac6_fail;
1000#endif
1001 return err;
1002
1003#ifdef CONFIG_PROC_FS
1004proc_ac6_fail:
1005 tcp6_proc_exit(net);
1006proc_tcp6_fail:
1007 udp6_proc_exit(net);
1008out:
1009 ipv6_cleanup_mibs(net);
1010 return err;
1011#endif
1012}
1013
1014static void __net_exit inet6_net_exit(struct net *net)
1015{
1016#ifdef CONFIG_PROC_FS
1017 udp6_proc_exit(net);
1018 tcp6_proc_exit(net);
1019 ac6_proc_exit(net);
1020#endif
1021 ipv6_cleanup_mibs(net);
1022}
1023
1024static struct pernet_operations inet6_net_ops = {
1025 .init = inet6_net_init,
1026 .exit = inet6_net_exit,
1027};
1028
1029static int ipv6_route_input(struct sk_buff *skb)
1030{
1031 ip6_route_input(skb);
1032 return skb_dst(skb)->error;
1033}
1034
1035static const struct ipv6_stub ipv6_stub_impl = {
1036 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1037 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1038 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1039 .ipv6_route_input = ipv6_route_input,
1040 .fib6_get_table = fib6_get_table,
1041 .fib6_table_lookup = fib6_table_lookup,
1042 .fib6_lookup = fib6_lookup,
1043 .fib6_select_path = fib6_select_path,
1044 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1045 .fib6_nh_init = fib6_nh_init,
1046 .fib6_nh_release = fib6_nh_release,
1047 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1048 .fib6_update_sernum = fib6_update_sernum_stub,
1049 .fib6_rt_update = fib6_rt_update,
1050 .ip6_del_rt = ip6_del_rt,
1051 .udpv6_encap_enable = udpv6_encap_enable,
1052 .ndisc_send_na = ndisc_send_na,
1053#if IS_ENABLED(CONFIG_XFRM)
1054 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1055 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1056 .xfrm6_gro_udp_encap_rcv = xfrm6_gro_udp_encap_rcv,
1057 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1058#endif
1059 .nd_tbl = &nd_tbl,
1060 .ipv6_fragment = ip6_fragment,
1061 .ipv6_dev_find = ipv6_dev_find,
1062};
1063
1064static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1065 .inet6_bind = __inet6_bind,
1066 .udp6_lib_lookup = __udp6_lib_lookup,
1067 .ipv6_setsockopt = do_ipv6_setsockopt,
1068 .ipv6_getsockopt = do_ipv6_getsockopt,
1069 .ipv6_dev_get_saddr = ipv6_dev_get_saddr,
1070};
1071
1072static int __init inet6_init(void)
1073{
1074 struct list_head *r;
1075 int err = 0;
1076
1077 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1078
1079 /* Register the socket-side information for inet6_create. */
1080 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1081 INIT_LIST_HEAD(r);
1082
1083 raw_hashinfo_init(&raw_v6_hashinfo);
1084
1085 if (disable_ipv6_mod) {
1086 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1087 goto out;
1088 }
1089
1090 err = proto_register(&tcpv6_prot, 1);
1091 if (err)
1092 goto out;
1093
1094 err = proto_register(&udpv6_prot, 1);
1095 if (err)
1096 goto out_unregister_tcp_proto;
1097
1098 err = proto_register(&udplitev6_prot, 1);
1099 if (err)
1100 goto out_unregister_udp_proto;
1101
1102 err = proto_register(&rawv6_prot, 1);
1103 if (err)
1104 goto out_unregister_udplite_proto;
1105
1106 err = proto_register(&pingv6_prot, 1);
1107 if (err)
1108 goto out_unregister_raw_proto;
1109
1110 /* We MUST register RAW sockets before we create the ICMP6,
1111 * IGMP6, or NDISC control sockets.
1112 */
1113 err = rawv6_init();
1114 if (err)
1115 goto out_unregister_ping_proto;
1116
1117 /* Register the family here so that the init calls below will
1118 * be able to create sockets. (?? is this dangerous ??)
1119 */
1120 err = sock_register(&inet6_family_ops);
1121 if (err)
1122 goto out_sock_register_fail;
1123
1124 /*
1125 * ipngwg API draft makes clear that the correct semantics
1126 * for TCP and UDP is to consider one TCP and UDP instance
1127 * in a host available by both INET and INET6 APIs and
1128 * able to communicate via both network protocols.
1129 */
1130
1131 err = register_pernet_subsys(&inet6_net_ops);
1132 if (err)
1133 goto register_pernet_fail;
1134 err = ip6_mr_init();
1135 if (err)
1136 goto ipmr_fail;
1137 err = icmpv6_init();
1138 if (err)
1139 goto icmp_fail;
1140 err = ndisc_init();
1141 if (err)
1142 goto ndisc_fail;
1143 err = igmp6_init();
1144 if (err)
1145 goto igmp_fail;
1146
1147 err = ipv6_netfilter_init();
1148 if (err)
1149 goto netfilter_fail;
1150 /* Create /proc/foo6 entries. */
1151#ifdef CONFIG_PROC_FS
1152 err = -ENOMEM;
1153 if (raw6_proc_init())
1154 goto proc_raw6_fail;
1155 if (udplite6_proc_init())
1156 goto proc_udplite6_fail;
1157 if (ipv6_misc_proc_init())
1158 goto proc_misc6_fail;
1159 if (if6_proc_init())
1160 goto proc_if6_fail;
1161#endif
1162 err = ip6_route_init();
1163 if (err)
1164 goto ip6_route_fail;
1165 err = ndisc_late_init();
1166 if (err)
1167 goto ndisc_late_fail;
1168 err = ip6_flowlabel_init();
1169 if (err)
1170 goto ip6_flowlabel_fail;
1171 err = ipv6_anycast_init();
1172 if (err)
1173 goto ipv6_anycast_fail;
1174 err = addrconf_init();
1175 if (err)
1176 goto addrconf_fail;
1177
1178 /* Init v6 extension headers. */
1179 err = ipv6_exthdrs_init();
1180 if (err)
1181 goto ipv6_exthdrs_fail;
1182
1183 err = ipv6_frag_init();
1184 if (err)
1185 goto ipv6_frag_fail;
1186
1187 /* Init v6 transport protocols. */
1188 err = udpv6_init();
1189 if (err)
1190 goto udpv6_fail;
1191
1192 err = udplitev6_init();
1193 if (err)
1194 goto udplitev6_fail;
1195
1196 err = udpv6_offload_init();
1197 if (err)
1198 goto udpv6_offload_fail;
1199
1200 err = tcpv6_init();
1201 if (err)
1202 goto tcpv6_fail;
1203
1204 err = ipv6_packet_init();
1205 if (err)
1206 goto ipv6_packet_fail;
1207
1208 err = pingv6_init();
1209 if (err)
1210 goto pingv6_fail;
1211
1212 err = calipso_init();
1213 if (err)
1214 goto calipso_fail;
1215
1216 err = seg6_init();
1217 if (err)
1218 goto seg6_fail;
1219
1220 err = rpl_init();
1221 if (err)
1222 goto rpl_fail;
1223
1224 err = ioam6_init();
1225 if (err)
1226 goto ioam6_fail;
1227
1228 err = igmp6_late_init();
1229 if (err)
1230 goto igmp6_late_err;
1231
1232#ifdef CONFIG_SYSCTL
1233 err = ipv6_sysctl_register();
1234 if (err)
1235 goto sysctl_fail;
1236#endif
1237
1238 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1239 wmb();
1240 ipv6_stub = &ipv6_stub_impl;
1241 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1242out:
1243 return err;
1244
1245#ifdef CONFIG_SYSCTL
1246sysctl_fail:
1247 igmp6_late_cleanup();
1248#endif
1249igmp6_late_err:
1250 ioam6_exit();
1251ioam6_fail:
1252 rpl_exit();
1253rpl_fail:
1254 seg6_exit();
1255seg6_fail:
1256 calipso_exit();
1257calipso_fail:
1258 pingv6_exit();
1259pingv6_fail:
1260 ipv6_packet_cleanup();
1261ipv6_packet_fail:
1262 tcpv6_exit();
1263tcpv6_fail:
1264 udpv6_offload_exit();
1265udpv6_offload_fail:
1266 udplitev6_exit();
1267udplitev6_fail:
1268 udpv6_exit();
1269udpv6_fail:
1270 ipv6_frag_exit();
1271ipv6_frag_fail:
1272 ipv6_exthdrs_exit();
1273ipv6_exthdrs_fail:
1274 addrconf_cleanup();
1275addrconf_fail:
1276 ipv6_anycast_cleanup();
1277ipv6_anycast_fail:
1278 ip6_flowlabel_cleanup();
1279ip6_flowlabel_fail:
1280 ndisc_late_cleanup();
1281ndisc_late_fail:
1282 ip6_route_cleanup();
1283ip6_route_fail:
1284#ifdef CONFIG_PROC_FS
1285 if6_proc_exit();
1286proc_if6_fail:
1287 ipv6_misc_proc_exit();
1288proc_misc6_fail:
1289 udplite6_proc_exit();
1290proc_udplite6_fail:
1291 raw6_proc_exit();
1292proc_raw6_fail:
1293#endif
1294 ipv6_netfilter_fini();
1295netfilter_fail:
1296 igmp6_cleanup();
1297igmp_fail:
1298 ndisc_cleanup();
1299ndisc_fail:
1300 icmpv6_cleanup();
1301icmp_fail:
1302 ip6_mr_cleanup();
1303ipmr_fail:
1304 unregister_pernet_subsys(&inet6_net_ops);
1305register_pernet_fail:
1306 sock_unregister(PF_INET6);
1307 rtnl_unregister_all(PF_INET6);
1308out_sock_register_fail:
1309 rawv6_exit();
1310out_unregister_ping_proto:
1311 proto_unregister(&pingv6_prot);
1312out_unregister_raw_proto:
1313 proto_unregister(&rawv6_prot);
1314out_unregister_udplite_proto:
1315 proto_unregister(&udplitev6_prot);
1316out_unregister_udp_proto:
1317 proto_unregister(&udpv6_prot);
1318out_unregister_tcp_proto:
1319 proto_unregister(&tcpv6_prot);
1320 goto out;
1321}
1322module_init(inet6_init);
1323
1324MODULE_ALIAS_NETPROTO(PF_INET6);