Merge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kern
[sfrench/cifs-2.6.git] / net / ipv6 / af_inet6.c
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
71 MODULE_AUTHOR("Cast of dozens");
72 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
73 MODULE_LICENSE("GPL");
74
75 /* The inetsw6 table contains everything that inet6_create needs to
76  * build a new socket.
77  */
78 static struct list_head inetsw6[SOCK_MAX];
79 static DEFINE_SPINLOCK(inetsw6_lock);
80
81 struct ipv6_params ipv6_defaults = {
82         .disable_ipv6 = 0,
83         .autoconf = 1,
84 };
85
86 static int disable_ipv6_mod;
87
88 module_param_named(disable, disable_ipv6_mod, int, 0444);
89 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
90
91 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
92 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
93
94 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
95 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
96
97 bool ipv6_mod_enabled(void)
98 {
99         return disable_ipv6_mod == 0;
100 }
101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
102
103 static __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
110 static 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. */
126 lookup_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         }
270 out:
271         return err;
272 out_rcu_unlock:
273         rcu_read_unlock();
274         goto out;
275 }
276
277 static 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;
429 out:
430         if (flags & BIND_WITH_LOCK)
431                 release_sock(sk);
432         return err;
433 out_unlock:
434         rcu_read_unlock();
435         goto out;
436 }
437
438 /* bind for INET6 API */
439 int 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_LOCK(sk, uaddr);
455         if (err)
456                 return err;
457
458         return __inet6_bind(sk, uaddr, addr_len, BIND_WITH_LOCK);
459 }
460 EXPORT_SYMBOL(inet6_bind);
461
462 int 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 }
477 EXPORT_SYMBOL(inet6_release);
478
479 void 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 }
504 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
505
506 /*
507  *      This does both peername and sockname.
508  */
509 int 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 }
546 EXPORT_SYMBOL(inet6_getname);
547
548 int 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 }
577 EXPORT_SYMBOL(inet6_ioctl);
578
579 #ifdef CONFIG_COMPAT
580 struct 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
593 static 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
613 int 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 }
626 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
627 #endif /* CONFIG_COMPAT */
628
629 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
630                                             size_t));
631 int 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
642 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
643                                             size_t, int, int, int *));
644 int 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
662 const struct proto_ops inet6_stream_ops = {
663         .family            = PF_INET6,
664         .flags             = PROTO_CMSG_DATA_ONLY,
665         .owner             = THIS_MODULE,
666         .release           = inet6_release,
667         .bind              = inet6_bind,
668         .connect           = inet_stream_connect,       /* ok           */
669         .socketpair        = sock_no_socketpair,        /* a do nothing */
670         .accept            = inet_accept,               /* ok           */
671         .getname           = inet6_getname,
672         .poll              = tcp_poll,                  /* ok           */
673         .ioctl             = inet6_ioctl,               /* must change  */
674         .gettstamp         = sock_gettstamp,
675         .listen            = inet_listen,               /* ok           */
676         .shutdown          = inet_shutdown,             /* ok           */
677         .setsockopt        = sock_common_setsockopt,    /* ok           */
678         .getsockopt        = sock_common_getsockopt,    /* ok           */
679         .sendmsg           = inet6_sendmsg,             /* retpoline's sake */
680         .recvmsg           = inet6_recvmsg,             /* retpoline's sake */
681 #ifdef CONFIG_MMU
682         .mmap              = tcp_mmap,
683 #endif
684         .sendpage          = inet_sendpage,
685         .sendmsg_locked    = tcp_sendmsg_locked,
686         .sendpage_locked   = tcp_sendpage_locked,
687         .splice_read       = tcp_splice_read,
688         .read_sock         = tcp_read_sock,
689         .peek_len          = tcp_peek_len,
690 #ifdef CONFIG_COMPAT
691         .compat_ioctl      = inet6_compat_ioctl,
692 #endif
693         .set_rcvlowat      = tcp_set_rcvlowat,
694 };
695
696 const struct proto_ops inet6_dgram_ops = {
697         .family            = PF_INET6,
698         .owner             = THIS_MODULE,
699         .release           = inet6_release,
700         .bind              = inet6_bind,
701         .connect           = inet_dgram_connect,        /* ok           */
702         .socketpair        = sock_no_socketpair,        /* a do nothing */
703         .accept            = sock_no_accept,            /* a do nothing */
704         .getname           = inet6_getname,
705         .poll              = udp_poll,                  /* ok           */
706         .ioctl             = inet6_ioctl,               /* must change  */
707         .gettstamp         = sock_gettstamp,
708         .listen            = sock_no_listen,            /* ok           */
709         .shutdown          = inet_shutdown,             /* ok           */
710         .setsockopt        = sock_common_setsockopt,    /* ok           */
711         .getsockopt        = sock_common_getsockopt,    /* ok           */
712         .sendmsg           = inet6_sendmsg,             /* retpoline's sake */
713         .recvmsg           = inet6_recvmsg,             /* retpoline's sake */
714         .mmap              = sock_no_mmap,
715         .sendpage          = sock_no_sendpage,
716         .set_peek_off      = sk_set_peek_off,
717 #ifdef CONFIG_COMPAT
718         .compat_ioctl      = inet6_compat_ioctl,
719 #endif
720 };
721
722 static const struct net_proto_family inet6_family_ops = {
723         .family = PF_INET6,
724         .create = inet6_create,
725         .owner  = THIS_MODULE,
726 };
727
728 int inet6_register_protosw(struct inet_protosw *p)
729 {
730         struct list_head *lh;
731         struct inet_protosw *answer;
732         struct list_head *last_perm;
733         int protocol = p->protocol;
734         int ret;
735
736         spin_lock_bh(&inetsw6_lock);
737
738         ret = -EINVAL;
739         if (p->type >= SOCK_MAX)
740                 goto out_illegal;
741
742         /* If we are trying to override a permanent protocol, bail. */
743         answer = NULL;
744         ret = -EPERM;
745         last_perm = &inetsw6[p->type];
746         list_for_each(lh, &inetsw6[p->type]) {
747                 answer = list_entry(lh, struct inet_protosw, list);
748
749                 /* Check only the non-wild match. */
750                 if (INET_PROTOSW_PERMANENT & answer->flags) {
751                         if (protocol == answer->protocol)
752                                 break;
753                         last_perm = lh;
754                 }
755
756                 answer = NULL;
757         }
758         if (answer)
759                 goto out_permanent;
760
761         /* Add the new entry after the last permanent entry if any, so that
762          * the new entry does not override a permanent entry when matched with
763          * a wild-card protocol. But it is allowed to override any existing
764          * non-permanent entry.  This means that when we remove this entry, the
765          * system automatically returns to the old behavior.
766          */
767         list_add_rcu(&p->list, last_perm);
768         ret = 0;
769 out:
770         spin_unlock_bh(&inetsw6_lock);
771         return ret;
772
773 out_permanent:
774         pr_err("Attempt to override permanent protocol %d\n", protocol);
775         goto out;
776
777 out_illegal:
778         pr_err("Ignoring attempt to register invalid socket type %d\n",
779                p->type);
780         goto out;
781 }
782 EXPORT_SYMBOL(inet6_register_protosw);
783
784 void
785 inet6_unregister_protosw(struct inet_protosw *p)
786 {
787         if (INET_PROTOSW_PERMANENT & p->flags) {
788                 pr_err("Attempt to unregister permanent protocol %d\n",
789                        p->protocol);
790         } else {
791                 spin_lock_bh(&inetsw6_lock);
792                 list_del_rcu(&p->list);
793                 spin_unlock_bh(&inetsw6_lock);
794
795                 synchronize_net();
796         }
797 }
798 EXPORT_SYMBOL(inet6_unregister_protosw);
799
800 int inet6_sk_rebuild_header(struct sock *sk)
801 {
802         struct ipv6_pinfo *np = inet6_sk(sk);
803         struct dst_entry *dst;
804
805         dst = __sk_dst_check(sk, np->dst_cookie);
806
807         if (!dst) {
808                 struct inet_sock *inet = inet_sk(sk);
809                 struct in6_addr *final_p, final;
810                 struct flowi6 fl6;
811
812                 memset(&fl6, 0, sizeof(fl6));
813                 fl6.flowi6_proto = sk->sk_protocol;
814                 fl6.daddr = sk->sk_v6_daddr;
815                 fl6.saddr = np->saddr;
816                 fl6.flowlabel = np->flow_label;
817                 fl6.flowi6_oif = sk->sk_bound_dev_if;
818                 fl6.flowi6_mark = sk->sk_mark;
819                 fl6.fl6_dport = inet->inet_dport;
820                 fl6.fl6_sport = inet->inet_sport;
821                 fl6.flowi6_uid = sk->sk_uid;
822                 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
823
824                 rcu_read_lock();
825                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
826                                          &final);
827                 rcu_read_unlock();
828
829                 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
830                 if (IS_ERR(dst)) {
831                         sk->sk_route_caps = 0;
832                         sk->sk_err_soft = -PTR_ERR(dst);
833                         return PTR_ERR(dst);
834                 }
835
836                 ip6_dst_store(sk, dst, NULL, NULL);
837         }
838
839         return 0;
840 }
841 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
842
843 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
844                        const struct inet6_skb_parm *opt)
845 {
846         const struct ipv6_pinfo *np = inet6_sk(sk);
847
848         if (np->rxopt.all) {
849                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
850                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
851                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
852                      np->rxopt.bits.rxflow) ||
853                     (opt->srcrt && (np->rxopt.bits.srcrt ||
854                      np->rxopt.bits.osrcrt)) ||
855                     ((opt->dst1 || opt->dst0) &&
856                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
857                         return true;
858         }
859         return false;
860 }
861 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
862
863 static struct packet_type ipv6_packet_type __read_mostly = {
864         .type = cpu_to_be16(ETH_P_IPV6),
865         .func = ipv6_rcv,
866         .list_func = ipv6_list_rcv,
867 };
868
869 static int __init ipv6_packet_init(void)
870 {
871         dev_add_pack(&ipv6_packet_type);
872         return 0;
873 }
874
875 static void ipv6_packet_cleanup(void)
876 {
877         dev_remove_pack(&ipv6_packet_type);
878 }
879
880 static int __net_init ipv6_init_mibs(struct net *net)
881 {
882         int i;
883
884         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
885         if (!net->mib.udp_stats_in6)
886                 return -ENOMEM;
887         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
888         if (!net->mib.udplite_stats_in6)
889                 goto err_udplite_mib;
890         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
891         if (!net->mib.ipv6_statistics)
892                 goto err_ip_mib;
893
894         for_each_possible_cpu(i) {
895                 struct ipstats_mib *af_inet6_stats;
896                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
897                 u64_stats_init(&af_inet6_stats->syncp);
898         }
899
900
901         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
902         if (!net->mib.icmpv6_statistics)
903                 goto err_icmp_mib;
904         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
905                                                 GFP_KERNEL);
906         if (!net->mib.icmpv6msg_statistics)
907                 goto err_icmpmsg_mib;
908         return 0;
909
910 err_icmpmsg_mib:
911         free_percpu(net->mib.icmpv6_statistics);
912 err_icmp_mib:
913         free_percpu(net->mib.ipv6_statistics);
914 err_ip_mib:
915         free_percpu(net->mib.udplite_stats_in6);
916 err_udplite_mib:
917         free_percpu(net->mib.udp_stats_in6);
918         return -ENOMEM;
919 }
920
921 static void ipv6_cleanup_mibs(struct net *net)
922 {
923         free_percpu(net->mib.udp_stats_in6);
924         free_percpu(net->mib.udplite_stats_in6);
925         free_percpu(net->mib.ipv6_statistics);
926         free_percpu(net->mib.icmpv6_statistics);
927         kfree(net->mib.icmpv6msg_statistics);
928 }
929
930 static int __net_init inet6_net_init(struct net *net)
931 {
932         int err = 0;
933
934         net->ipv6.sysctl.bindv6only = 0;
935         net->ipv6.sysctl.icmpv6_time = 1*HZ;
936         net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
937         net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
938         net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
939
940         /* By default, rate limit error messages.
941          * Except for pmtu discovery, it would break it.
942          * proc_do_large_bitmap needs pointer to the bitmap.
943          */
944         bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
945         bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
946         net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
947
948         net->ipv6.sysctl.flowlabel_consistency = 1;
949         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
950         net->ipv6.sysctl.idgen_retries = 3;
951         net->ipv6.sysctl.idgen_delay = 1 * HZ;
952         net->ipv6.sysctl.flowlabel_state_ranges = 0;
953         net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
954         net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
955         net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
956         net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
957         net->ipv6.sysctl.fib_notify_on_flag_change = 0;
958         atomic_set(&net->ipv6.fib6_sernum, 1);
959
960         err = ipv6_init_mibs(net);
961         if (err)
962                 return err;
963 #ifdef CONFIG_PROC_FS
964         err = udp6_proc_init(net);
965         if (err)
966                 goto out;
967         err = tcp6_proc_init(net);
968         if (err)
969                 goto proc_tcp6_fail;
970         err = ac6_proc_init(net);
971         if (err)
972                 goto proc_ac6_fail;
973 #endif
974         return err;
975
976 #ifdef CONFIG_PROC_FS
977 proc_ac6_fail:
978         tcp6_proc_exit(net);
979 proc_tcp6_fail:
980         udp6_proc_exit(net);
981 out:
982         ipv6_cleanup_mibs(net);
983         return err;
984 #endif
985 }
986
987 static void __net_exit inet6_net_exit(struct net *net)
988 {
989 #ifdef CONFIG_PROC_FS
990         udp6_proc_exit(net);
991         tcp6_proc_exit(net);
992         ac6_proc_exit(net);
993 #endif
994         ipv6_cleanup_mibs(net);
995 }
996
997 static struct pernet_operations inet6_net_ops = {
998         .init = inet6_net_init,
999         .exit = inet6_net_exit,
1000 };
1001
1002 static int ipv6_route_input(struct sk_buff *skb)
1003 {
1004         ip6_route_input(skb);
1005         return skb_dst(skb)->error;
1006 }
1007
1008 static const struct ipv6_stub ipv6_stub_impl = {
1009         .ipv6_sock_mc_join = ipv6_sock_mc_join,
1010         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1011         .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1012         .ipv6_route_input  = ipv6_route_input,
1013         .fib6_get_table    = fib6_get_table,
1014         .fib6_table_lookup = fib6_table_lookup,
1015         .fib6_lookup       = fib6_lookup,
1016         .fib6_select_path  = fib6_select_path,
1017         .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1018         .fib6_nh_init      = fib6_nh_init,
1019         .fib6_nh_release   = fib6_nh_release,
1020         .fib6_update_sernum = fib6_update_sernum_stub,
1021         .fib6_rt_update    = fib6_rt_update,
1022         .ip6_del_rt        = ip6_del_rt,
1023         .udpv6_encap_enable = udpv6_encap_enable,
1024         .ndisc_send_na = ndisc_send_na,
1025 #if IS_ENABLED(CONFIG_XFRM)
1026         .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1027         .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1028         .xfrm6_rcv_encap = xfrm6_rcv_encap,
1029 #endif
1030         .nd_tbl = &nd_tbl,
1031         .ipv6_fragment = ip6_fragment,
1032 };
1033
1034 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1035         .inet6_bind = __inet6_bind,
1036         .udp6_lib_lookup = __udp6_lib_lookup,
1037 };
1038
1039 static int __init inet6_init(void)
1040 {
1041         struct list_head *r;
1042         int err = 0;
1043
1044         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1045
1046         /* Register the socket-side information for inet6_create.  */
1047         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1048                 INIT_LIST_HEAD(r);
1049
1050         if (disable_ipv6_mod) {
1051                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1052                 goto out;
1053         }
1054
1055         err = proto_register(&tcpv6_prot, 1);
1056         if (err)
1057                 goto out;
1058
1059         err = proto_register(&udpv6_prot, 1);
1060         if (err)
1061                 goto out_unregister_tcp_proto;
1062
1063         err = proto_register(&udplitev6_prot, 1);
1064         if (err)
1065                 goto out_unregister_udp_proto;
1066
1067         err = proto_register(&rawv6_prot, 1);
1068         if (err)
1069                 goto out_unregister_udplite_proto;
1070
1071         err = proto_register(&pingv6_prot, 1);
1072         if (err)
1073                 goto out_unregister_raw_proto;
1074
1075         /* We MUST register RAW sockets before we create the ICMP6,
1076          * IGMP6, or NDISC control sockets.
1077          */
1078         err = rawv6_init();
1079         if (err)
1080                 goto out_unregister_ping_proto;
1081
1082         /* Register the family here so that the init calls below will
1083          * be able to create sockets. (?? is this dangerous ??)
1084          */
1085         err = sock_register(&inet6_family_ops);
1086         if (err)
1087                 goto out_sock_register_fail;
1088
1089         /*
1090          *      ipngwg API draft makes clear that the correct semantics
1091          *      for TCP and UDP is to consider one TCP and UDP instance
1092          *      in a host available by both INET and INET6 APIs and
1093          *      able to communicate via both network protocols.
1094          */
1095
1096         err = register_pernet_subsys(&inet6_net_ops);
1097         if (err)
1098                 goto register_pernet_fail;
1099         err = ip6_mr_init();
1100         if (err)
1101                 goto ipmr_fail;
1102         err = icmpv6_init();
1103         if (err)
1104                 goto icmp_fail;
1105         err = ndisc_init();
1106         if (err)
1107                 goto ndisc_fail;
1108         err = igmp6_init();
1109         if (err)
1110                 goto igmp_fail;
1111
1112         err = ipv6_netfilter_init();
1113         if (err)
1114                 goto netfilter_fail;
1115         /* Create /proc/foo6 entries. */
1116 #ifdef CONFIG_PROC_FS
1117         err = -ENOMEM;
1118         if (raw6_proc_init())
1119                 goto proc_raw6_fail;
1120         if (udplite6_proc_init())
1121                 goto proc_udplite6_fail;
1122         if (ipv6_misc_proc_init())
1123                 goto proc_misc6_fail;
1124         if (if6_proc_init())
1125                 goto proc_if6_fail;
1126 #endif
1127         err = ip6_route_init();
1128         if (err)
1129                 goto ip6_route_fail;
1130         err = ndisc_late_init();
1131         if (err)
1132                 goto ndisc_late_fail;
1133         err = ip6_flowlabel_init();
1134         if (err)
1135                 goto ip6_flowlabel_fail;
1136         err = ipv6_anycast_init();
1137         if (err)
1138                 goto ipv6_anycast_fail;
1139         err = addrconf_init();
1140         if (err)
1141                 goto addrconf_fail;
1142
1143         /* Init v6 extension headers. */
1144         err = ipv6_exthdrs_init();
1145         if (err)
1146                 goto ipv6_exthdrs_fail;
1147
1148         err = ipv6_frag_init();
1149         if (err)
1150                 goto ipv6_frag_fail;
1151
1152         /* Init v6 transport protocols. */
1153         err = udpv6_init();
1154         if (err)
1155                 goto udpv6_fail;
1156
1157         err = udplitev6_init();
1158         if (err)
1159                 goto udplitev6_fail;
1160
1161         err = udpv6_offload_init();
1162         if (err)
1163                 goto udpv6_offload_fail;
1164
1165         err = tcpv6_init();
1166         if (err)
1167                 goto tcpv6_fail;
1168
1169         err = ipv6_packet_init();
1170         if (err)
1171                 goto ipv6_packet_fail;
1172
1173         err = pingv6_init();
1174         if (err)
1175                 goto pingv6_fail;
1176
1177         err = calipso_init();
1178         if (err)
1179                 goto calipso_fail;
1180
1181         err = seg6_init();
1182         if (err)
1183                 goto seg6_fail;
1184
1185         err = rpl_init();
1186         if (err)
1187                 goto rpl_fail;
1188
1189         err = igmp6_late_init();
1190         if (err)
1191                 goto igmp6_late_err;
1192
1193 #ifdef CONFIG_SYSCTL
1194         err = ipv6_sysctl_register();
1195         if (err)
1196                 goto sysctl_fail;
1197 #endif
1198
1199         /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1200         wmb();
1201         ipv6_stub = &ipv6_stub_impl;
1202         ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1203 out:
1204         return err;
1205
1206 #ifdef CONFIG_SYSCTL
1207 sysctl_fail:
1208         igmp6_late_cleanup();
1209 #endif
1210 igmp6_late_err:
1211         rpl_exit();
1212 rpl_fail:
1213         seg6_exit();
1214 seg6_fail:
1215         calipso_exit();
1216 calipso_fail:
1217         pingv6_exit();
1218 pingv6_fail:
1219         ipv6_packet_cleanup();
1220 ipv6_packet_fail:
1221         tcpv6_exit();
1222 tcpv6_fail:
1223         udpv6_offload_exit();
1224 udpv6_offload_fail:
1225         udplitev6_exit();
1226 udplitev6_fail:
1227         udpv6_exit();
1228 udpv6_fail:
1229         ipv6_frag_exit();
1230 ipv6_frag_fail:
1231         ipv6_exthdrs_exit();
1232 ipv6_exthdrs_fail:
1233         addrconf_cleanup();
1234 addrconf_fail:
1235         ipv6_anycast_cleanup();
1236 ipv6_anycast_fail:
1237         ip6_flowlabel_cleanup();
1238 ip6_flowlabel_fail:
1239         ndisc_late_cleanup();
1240 ndisc_late_fail:
1241         ip6_route_cleanup();
1242 ip6_route_fail:
1243 #ifdef CONFIG_PROC_FS
1244         if6_proc_exit();
1245 proc_if6_fail:
1246         ipv6_misc_proc_exit();
1247 proc_misc6_fail:
1248         udplite6_proc_exit();
1249 proc_udplite6_fail:
1250         raw6_proc_exit();
1251 proc_raw6_fail:
1252 #endif
1253         ipv6_netfilter_fini();
1254 netfilter_fail:
1255         igmp6_cleanup();
1256 igmp_fail:
1257         ndisc_cleanup();
1258 ndisc_fail:
1259         icmpv6_cleanup();
1260 icmp_fail:
1261         ip6_mr_cleanup();
1262 ipmr_fail:
1263         unregister_pernet_subsys(&inet6_net_ops);
1264 register_pernet_fail:
1265         sock_unregister(PF_INET6);
1266         rtnl_unregister_all(PF_INET6);
1267 out_sock_register_fail:
1268         rawv6_exit();
1269 out_unregister_ping_proto:
1270         proto_unregister(&pingv6_prot);
1271 out_unregister_raw_proto:
1272         proto_unregister(&rawv6_prot);
1273 out_unregister_udplite_proto:
1274         proto_unregister(&udplitev6_prot);
1275 out_unregister_udp_proto:
1276         proto_unregister(&udpv6_prot);
1277 out_unregister_tcp_proto:
1278         proto_unregister(&tcpv6_prot);
1279         goto out;
1280 }
1281 module_init(inet6_init);
1282
1283 MODULE_ALIAS_NETPROTO(PF_INET6);