GNU Linux-libre 5.10.217-gnu1
[releases.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 void inet6_sock_destruct(struct sock *sk)
111 {
112         inet6_cleanup_sock(sk);
113         inet_sock_destruct(sk);
114 }
115 EXPORT_SYMBOL_GPL(inet6_sock_destruct);
116
117 static int inet6_create(struct net *net, struct socket *sock, int protocol,
118                         int kern)
119 {
120         struct inet_sock *inet;
121         struct ipv6_pinfo *np;
122         struct sock *sk;
123         struct inet_protosw *answer;
124         struct proto *answer_prot;
125         unsigned char answer_flags;
126         int try_loading_module = 0;
127         int err;
128
129         if (protocol < 0 || protocol >= IPPROTO_MAX)
130                 return -EINVAL;
131
132         /* Look for the requested type/protocol pair. */
133 lookup_protocol:
134         err = -ESOCKTNOSUPPORT;
135         rcu_read_lock();
136         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
137
138                 err = 0;
139                 /* Check the non-wild match. */
140                 if (protocol == answer->protocol) {
141                         if (protocol != IPPROTO_IP)
142                                 break;
143                 } else {
144                         /* Check for the two wild cases. */
145                         if (IPPROTO_IP == protocol) {
146                                 protocol = answer->protocol;
147                                 break;
148                         }
149                         if (IPPROTO_IP == answer->protocol)
150                                 break;
151                 }
152                 err = -EPROTONOSUPPORT;
153         }
154
155         if (err) {
156                 if (try_loading_module < 2) {
157                         rcu_read_unlock();
158                         /*
159                          * Be more specific, e.g. net-pf-10-proto-132-type-1
160                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
161                          */
162                         if (++try_loading_module == 1)
163                                 request_module("net-pf-%d-proto-%d-type-%d",
164                                                 PF_INET6, protocol, sock->type);
165                         /*
166                          * Fall back to generic, e.g. net-pf-10-proto-132
167                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
168                          */
169                         else
170                                 request_module("net-pf-%d-proto-%d",
171                                                 PF_INET6, protocol);
172                         goto lookup_protocol;
173                 } else
174                         goto out_rcu_unlock;
175         }
176
177         err = -EPERM;
178         if (sock->type == SOCK_RAW && !kern &&
179             !ns_capable(net->user_ns, CAP_NET_RAW))
180                 goto out_rcu_unlock;
181
182         sock->ops = answer->ops;
183         answer_prot = answer->prot;
184         answer_flags = answer->flags;
185         rcu_read_unlock();
186
187         WARN_ON(!answer_prot->slab);
188
189         err = -ENOBUFS;
190         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
191         if (!sk)
192                 goto out;
193
194         sock_init_data(sock, sk);
195
196         err = 0;
197         if (INET_PROTOSW_REUSE & answer_flags)
198                 sk->sk_reuse = SK_CAN_REUSE;
199
200         if (INET_PROTOSW_ICSK & answer_flags)
201                 inet_init_csk_locks(sk);
202
203         inet = inet_sk(sk);
204         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
205
206         if (SOCK_RAW == sock->type) {
207                 inet->inet_num = protocol;
208                 if (IPPROTO_RAW == protocol)
209                         inet->hdrincl = 1;
210         }
211
212         sk->sk_destruct         = inet6_sock_destruct;
213         sk->sk_family           = PF_INET6;
214         sk->sk_protocol         = protocol;
215
216         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
217
218         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
219         np->hop_limit   = -1;
220         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
221         np->mc_loop     = 1;
222         np->mc_all      = 1;
223         np->pmtudisc    = IPV6_PMTUDISC_WANT;
224         np->repflow     = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
225         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
226
227         /* Init the ipv4 part of the socket since we can have sockets
228          * using v6 API for ipv4.
229          */
230         inet->uc_ttl    = -1;
231
232         inet->mc_loop   = 1;
233         inet->mc_ttl    = 1;
234         inet->mc_index  = 0;
235         inet->mc_list   = NULL;
236         inet->rcv_tos   = 0;
237
238         if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc))
239                 inet->pmtudisc = IP_PMTUDISC_DONT;
240         else
241                 inet->pmtudisc = IP_PMTUDISC_WANT;
242         /*
243          * Increment only the relevant sk_prot->socks debug field, this changes
244          * the previous behaviour of incrementing both the equivalent to
245          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
246          *
247          * This allows better debug granularity as we'll know exactly how many
248          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
249          * transport protocol socks. -acme
250          */
251         sk_refcnt_debug_inc(sk);
252
253         if (inet->inet_num) {
254                 /* It assumes that any protocol which allows
255                  * the user to assign a number at socket
256                  * creation time automatically shares.
257                  */
258                 inet->inet_sport = htons(inet->inet_num);
259                 err = sk->sk_prot->hash(sk);
260                 if (err) {
261                         sk_common_release(sk);
262                         goto out;
263                 }
264         }
265         if (sk->sk_prot->init) {
266                 err = sk->sk_prot->init(sk);
267                 if (err) {
268                         sk_common_release(sk);
269                         goto out;
270                 }
271         }
272
273         if (!kern) {
274                 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
275                 if (err) {
276                         sk_common_release(sk);
277                         goto out;
278                 }
279         }
280 out:
281         return err;
282 out_rcu_unlock:
283         rcu_read_unlock();
284         goto out;
285 }
286
287 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
288                         u32 flags)
289 {
290         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
291         struct inet_sock *inet = inet_sk(sk);
292         struct ipv6_pinfo *np = inet6_sk(sk);
293         struct net *net = sock_net(sk);
294         __be32 v4addr = 0;
295         unsigned short snum;
296         bool saved_ipv6only;
297         int addr_type = 0;
298         int err = 0;
299
300         if (addr->sin6_family != AF_INET6)
301                 return -EAFNOSUPPORT;
302
303         addr_type = ipv6_addr_type(&addr->sin6_addr);
304         if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
305                 return -EINVAL;
306
307         snum = ntohs(addr->sin6_port);
308         if (snum && inet_port_requires_bind_service(net, snum) &&
309             !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
310                 return -EACCES;
311
312         if (flags & BIND_WITH_LOCK)
313                 lock_sock(sk);
314
315         /* Check these errors (active socket, double bind). */
316         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
317                 err = -EINVAL;
318                 goto out;
319         }
320
321         /* Check if the address belongs to the host. */
322         if (addr_type == IPV6_ADDR_MAPPED) {
323                 struct net_device *dev = NULL;
324                 int chk_addr_ret;
325
326                 /* Binding to v4-mapped address on a v6-only socket
327                  * makes no sense
328                  */
329                 if (sk->sk_ipv6only) {
330                         err = -EINVAL;
331                         goto out;
332                 }
333
334                 rcu_read_lock();
335                 if (sk->sk_bound_dev_if) {
336                         dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
337                         if (!dev) {
338                                 err = -ENODEV;
339                                 goto out_unlock;
340                         }
341                 }
342
343                 /* Reproduce AF_INET checks to make the bindings consistent */
344                 v4addr = addr->sin6_addr.s6_addr32[3];
345                 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
346                 rcu_read_unlock();
347
348                 if (!inet_can_nonlocal_bind(net, inet) &&
349                     v4addr != htonl(INADDR_ANY) &&
350                     chk_addr_ret != RTN_LOCAL &&
351                     chk_addr_ret != RTN_MULTICAST &&
352                     chk_addr_ret != RTN_BROADCAST) {
353                         err = -EADDRNOTAVAIL;
354                         goto out;
355                 }
356         } else {
357                 if (addr_type != IPV6_ADDR_ANY) {
358                         struct net_device *dev = NULL;
359
360                         rcu_read_lock();
361                         if (__ipv6_addr_needs_scope_id(addr_type)) {
362                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
363                                     addr->sin6_scope_id) {
364                                         /* Override any existing binding, if another one
365                                          * is supplied by user.
366                                          */
367                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
368                                 }
369
370                                 /* Binding to link-local address requires an interface */
371                                 if (!sk->sk_bound_dev_if) {
372                                         err = -EINVAL;
373                                         goto out_unlock;
374                                 }
375                         }
376
377                         if (sk->sk_bound_dev_if) {
378                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
379                                 if (!dev) {
380                                         err = -ENODEV;
381                                         goto out_unlock;
382                                 }
383                         }
384
385                         /* ipv4 addr of the socket is invalid.  Only the
386                          * unspecified and mapped address have a v4 equivalent.
387                          */
388                         v4addr = LOOPBACK4_IPV6;
389                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
390                                 if (!ipv6_can_nonlocal_bind(net, inet) &&
391                                     !ipv6_chk_addr(net, &addr->sin6_addr,
392                                                    dev, 0)) {
393                                         err = -EADDRNOTAVAIL;
394                                         goto out_unlock;
395                                 }
396                         }
397                         rcu_read_unlock();
398                 }
399         }
400
401         inet->inet_rcv_saddr = v4addr;
402         inet->inet_saddr = v4addr;
403
404         sk->sk_v6_rcv_saddr = addr->sin6_addr;
405
406         if (!(addr_type & IPV6_ADDR_MULTICAST))
407                 np->saddr = addr->sin6_addr;
408
409         saved_ipv6only = sk->sk_ipv6only;
410         if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
411                 sk->sk_ipv6only = 1;
412
413         /* Make sure we are allowed to bind here. */
414         if (snum || !(inet->bind_address_no_port ||
415                       (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
416                 if (sk->sk_prot->get_port(sk, snum)) {
417                         sk->sk_ipv6only = saved_ipv6only;
418                         inet_reset_saddr(sk);
419                         err = -EADDRINUSE;
420                         goto out;
421                 }
422                 if (!(flags & BIND_FROM_BPF)) {
423                         err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
424                         if (err) {
425                                 sk->sk_ipv6only = saved_ipv6only;
426                                 inet_reset_saddr(sk);
427                                 goto out;
428                         }
429                 }
430         }
431
432         if (addr_type != IPV6_ADDR_ANY)
433                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
434         if (snum)
435                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
436         inet->inet_sport = htons(inet->inet_num);
437         inet->inet_dport = 0;
438         inet->inet_daddr = 0;
439 out:
440         if (flags & BIND_WITH_LOCK)
441                 release_sock(sk);
442         return err;
443 out_unlock:
444         rcu_read_unlock();
445         goto out;
446 }
447
448 /* bind for INET6 API */
449 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
450 {
451         struct sock *sk = sock->sk;
452         int err = 0;
453
454         /* If the socket has its own bind function then use it. */
455         if (sk->sk_prot->bind)
456                 return sk->sk_prot->bind(sk, uaddr, addr_len);
457
458         if (addr_len < SIN6_LEN_RFC2133)
459                 return -EINVAL;
460
461         /* BPF prog is run before any checks are done so that if the prog
462          * changes context in a wrong way it will be caught.
463          */
464         err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
465         if (err)
466                 return err;
467
468         return __inet6_bind(sk, uaddr, addr_len, BIND_WITH_LOCK);
469 }
470 EXPORT_SYMBOL(inet6_bind);
471
472 int inet6_release(struct socket *sock)
473 {
474         struct sock *sk = sock->sk;
475
476         if (!sk)
477                 return -EINVAL;
478
479         /* Free mc lists */
480         ipv6_sock_mc_close(sk);
481
482         /* Free ac lists */
483         ipv6_sock_ac_close(sk);
484
485         return inet_release(sock);
486 }
487 EXPORT_SYMBOL(inet6_release);
488
489 void inet6_destroy_sock(struct sock *sk)
490 {
491         struct ipv6_pinfo *np = inet6_sk(sk);
492         struct sk_buff *skb;
493         struct ipv6_txoptions *opt;
494
495         /* Release rx options */
496
497         skb = xchg(&np->pktoptions, NULL);
498         kfree_skb(skb);
499
500         skb = xchg(&np->rxpmtu, NULL);
501         kfree_skb(skb);
502
503         /* Free flowlabels */
504         fl6_free_socklist(sk);
505
506         /* Free tx options */
507
508         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
509         if (opt) {
510                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
511                 txopt_put(opt);
512         }
513 }
514 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
515
516 void inet6_cleanup_sock(struct sock *sk)
517 {
518         inet6_destroy_sock(sk);
519 }
520 EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
521
522 /*
523  *      This does both peername and sockname.
524  */
525 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
526                   int peer)
527 {
528         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
529         struct sock *sk = sock->sk;
530         struct inet_sock *inet = inet_sk(sk);
531         struct ipv6_pinfo *np = inet6_sk(sk);
532
533         sin->sin6_family = AF_INET6;
534         sin->sin6_flowinfo = 0;
535         sin->sin6_scope_id = 0;
536         if (peer) {
537                 if (!inet->inet_dport)
538                         return -ENOTCONN;
539                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
540                     peer == 1)
541                         return -ENOTCONN;
542                 sin->sin6_port = inet->inet_dport;
543                 sin->sin6_addr = sk->sk_v6_daddr;
544                 if (np->sndflow)
545                         sin->sin6_flowinfo = np->flow_label;
546         } else {
547                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
548                         sin->sin6_addr = np->saddr;
549                 else
550                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
551                 sin->sin6_port = inet->inet_sport;
552         }
553         if (cgroup_bpf_enabled)
554                 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin,
555                                             peer ? BPF_CGROUP_INET6_GETPEERNAME :
556                                                    BPF_CGROUP_INET6_GETSOCKNAME,
557                                             NULL);
558         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
559                                                  sk->sk_bound_dev_if);
560         return sizeof(*sin);
561 }
562 EXPORT_SYMBOL(inet6_getname);
563
564 int 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
570         switch (cmd) {
571         case SIOCADDRT:
572         case SIOCDELRT: {
573                 struct in6_rtmsg rtmsg;
574
575                 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
576                         return -EFAULT;
577                 return ipv6_route_ioctl(net, cmd, &rtmsg);
578         }
579         case SIOCSIFADDR:
580                 return addrconf_add_ifaddr(net, argp);
581         case SIOCDIFADDR:
582                 return addrconf_del_ifaddr(net, argp);
583         case SIOCSIFDSTADDR:
584                 return addrconf_set_dstaddr(net, argp);
585         default:
586                 if (!sk->sk_prot->ioctl)
587                         return -ENOIOCTLCMD;
588                 return sk->sk_prot->ioctl(sk, cmd, arg);
589         }
590         /*NOTREACHED*/
591         return 0;
592 }
593 EXPORT_SYMBOL(inet6_ioctl);
594
595 #ifdef CONFIG_COMPAT
596 struct compat_in6_rtmsg {
597         struct in6_addr         rtmsg_dst;
598         struct in6_addr         rtmsg_src;
599         struct in6_addr         rtmsg_gateway;
600         u32                     rtmsg_type;
601         u16                     rtmsg_dst_len;
602         u16                     rtmsg_src_len;
603         u32                     rtmsg_metric;
604         u32                     rtmsg_info;
605         u32                     rtmsg_flags;
606         s32                     rtmsg_ifindex;
607 };
608
609 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
610                 struct compat_in6_rtmsg __user *ur)
611 {
612         struct in6_rtmsg rt;
613
614         if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
615                         3 * sizeof(struct in6_addr)) ||
616             get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
617             get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
618             get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
619             get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
620             get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
621             get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
622             get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
623                 return -EFAULT;
624
625
626         return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
627 }
628
629 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
630 {
631         void __user *argp = compat_ptr(arg);
632         struct sock *sk = sock->sk;
633
634         switch (cmd) {
635         case SIOCADDRT:
636         case SIOCDELRT:
637                 return inet6_compat_routing_ioctl(sk, cmd, argp);
638         default:
639                 return -ENOIOCTLCMD;
640         }
641 }
642 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
643 #endif /* CONFIG_COMPAT */
644
645 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
646                                             size_t));
647 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
648 {
649         struct sock *sk = sock->sk;
650
651         if (unlikely(inet_send_prepare(sk)))
652                 return -EAGAIN;
653
654         return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
655                                sk, msg, size);
656 }
657
658 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
659                                             size_t, int, int, int *));
660 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
661                   int flags)
662 {
663         struct sock *sk = sock->sk;
664         int addr_len = 0;
665         int err;
666
667         if (likely(!(flags & MSG_ERRQUEUE)))
668                 sock_rps_record_flow(sk);
669
670         err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
671                               sk, msg, size, flags & MSG_DONTWAIT,
672                               flags & ~MSG_DONTWAIT, &addr_len);
673         if (err >= 0)
674                 msg->msg_namelen = addr_len;
675         return err;
676 }
677
678 const struct proto_ops inet6_stream_ops = {
679         .family            = PF_INET6,
680         .owner             = THIS_MODULE,
681         .release           = inet6_release,
682         .bind              = inet6_bind,
683         .connect           = inet_stream_connect,       /* ok           */
684         .socketpair        = sock_no_socketpair,        /* a do nothing */
685         .accept            = inet_accept,               /* ok           */
686         .getname           = inet6_getname,
687         .poll              = tcp_poll,                  /* ok           */
688         .ioctl             = inet6_ioctl,               /* must change  */
689         .gettstamp         = sock_gettstamp,
690         .listen            = inet_listen,               /* ok           */
691         .shutdown          = inet_shutdown,             /* ok           */
692         .setsockopt        = sock_common_setsockopt,    /* ok           */
693         .getsockopt        = sock_common_getsockopt,    /* ok           */
694         .sendmsg           = inet6_sendmsg,             /* retpoline's sake */
695         .recvmsg           = inet6_recvmsg,             /* retpoline's sake */
696 #ifdef CONFIG_MMU
697         .mmap              = tcp_mmap,
698 #endif
699         .sendpage          = inet_sendpage,
700         .sendmsg_locked    = tcp_sendmsg_locked,
701         .sendpage_locked   = tcp_sendpage_locked,
702         .splice_read       = tcp_splice_read,
703         .read_sock         = tcp_read_sock,
704         .peek_len          = tcp_peek_len,
705 #ifdef CONFIG_COMPAT
706         .compat_ioctl      = inet6_compat_ioctl,
707 #endif
708         .set_rcvlowat      = tcp_set_rcvlowat,
709 };
710
711 const struct proto_ops inet6_dgram_ops = {
712         .family            = PF_INET6,
713         .owner             = THIS_MODULE,
714         .release           = inet6_release,
715         .bind              = inet6_bind,
716         .connect           = inet_dgram_connect,        /* ok           */
717         .socketpair        = sock_no_socketpair,        /* a do nothing */
718         .accept            = sock_no_accept,            /* a do nothing */
719         .getname           = inet6_getname,
720         .poll              = udp_poll,                  /* ok           */
721         .ioctl             = inet6_ioctl,               /* must change  */
722         .gettstamp         = sock_gettstamp,
723         .listen            = sock_no_listen,            /* ok           */
724         .shutdown          = inet_shutdown,             /* ok           */
725         .setsockopt        = sock_common_setsockopt,    /* ok           */
726         .getsockopt        = sock_common_getsockopt,    /* ok           */
727         .sendmsg           = inet6_sendmsg,             /* retpoline's sake */
728         .recvmsg           = inet6_recvmsg,             /* retpoline's sake */
729         .mmap              = sock_no_mmap,
730         .sendpage          = sock_no_sendpage,
731         .set_peek_off      = sk_set_peek_off,
732 #ifdef CONFIG_COMPAT
733         .compat_ioctl      = inet6_compat_ioctl,
734 #endif
735 };
736
737 static const struct net_proto_family inet6_family_ops = {
738         .family = PF_INET6,
739         .create = inet6_create,
740         .owner  = THIS_MODULE,
741 };
742
743 int inet6_register_protosw(struct inet_protosw *p)
744 {
745         struct list_head *lh;
746         struct inet_protosw *answer;
747         struct list_head *last_perm;
748         int protocol = p->protocol;
749         int ret;
750
751         spin_lock_bh(&inetsw6_lock);
752
753         ret = -EINVAL;
754         if (p->type >= SOCK_MAX)
755                 goto out_illegal;
756
757         /* If we are trying to override a permanent protocol, bail. */
758         answer = NULL;
759         ret = -EPERM;
760         last_perm = &inetsw6[p->type];
761         list_for_each(lh, &inetsw6[p->type]) {
762                 answer = list_entry(lh, struct inet_protosw, list);
763
764                 /* Check only the non-wild match. */
765                 if (INET_PROTOSW_PERMANENT & answer->flags) {
766                         if (protocol == answer->protocol)
767                                 break;
768                         last_perm = lh;
769                 }
770
771                 answer = NULL;
772         }
773         if (answer)
774                 goto out_permanent;
775
776         /* Add the new entry after the last permanent entry if any, so that
777          * the new entry does not override a permanent entry when matched with
778          * a wild-card protocol. But it is allowed to override any existing
779          * non-permanent entry.  This means that when we remove this entry, the
780          * system automatically returns to the old behavior.
781          */
782         list_add_rcu(&p->list, last_perm);
783         ret = 0;
784 out:
785         spin_unlock_bh(&inetsw6_lock);
786         return ret;
787
788 out_permanent:
789         pr_err("Attempt to override permanent protocol %d\n", protocol);
790         goto out;
791
792 out_illegal:
793         pr_err("Ignoring attempt to register invalid socket type %d\n",
794                p->type);
795         goto out;
796 }
797 EXPORT_SYMBOL(inet6_register_protosw);
798
799 void
800 inet6_unregister_protosw(struct inet_protosw *p)
801 {
802         if (INET_PROTOSW_PERMANENT & p->flags) {
803                 pr_err("Attempt to unregister permanent protocol %d\n",
804                        p->protocol);
805         } else {
806                 spin_lock_bh(&inetsw6_lock);
807                 list_del_rcu(&p->list);
808                 spin_unlock_bh(&inetsw6_lock);
809
810                 synchronize_net();
811         }
812 }
813 EXPORT_SYMBOL(inet6_unregister_protosw);
814
815 int inet6_sk_rebuild_header(struct sock *sk)
816 {
817         struct ipv6_pinfo *np = inet6_sk(sk);
818         struct dst_entry *dst;
819
820         dst = __sk_dst_check(sk, np->dst_cookie);
821
822         if (!dst) {
823                 struct inet_sock *inet = inet_sk(sk);
824                 struct in6_addr *final_p, final;
825                 struct flowi6 fl6;
826
827                 memset(&fl6, 0, sizeof(fl6));
828                 fl6.flowi6_proto = sk->sk_protocol;
829                 fl6.daddr = sk->sk_v6_daddr;
830                 fl6.saddr = np->saddr;
831                 fl6.flowlabel = np->flow_label;
832                 fl6.flowi6_oif = sk->sk_bound_dev_if;
833                 fl6.flowi6_mark = sk->sk_mark;
834                 fl6.fl6_dport = inet->inet_dport;
835                 fl6.fl6_sport = inet->inet_sport;
836                 fl6.flowi6_uid = sk->sk_uid;
837                 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
838
839                 rcu_read_lock();
840                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
841                                          &final);
842                 rcu_read_unlock();
843
844                 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
845                 if (IS_ERR(dst)) {
846                         sk->sk_route_caps = 0;
847                         sk->sk_err_soft = -PTR_ERR(dst);
848                         return PTR_ERR(dst);
849                 }
850
851                 ip6_dst_store(sk, dst, NULL, NULL);
852         }
853
854         return 0;
855 }
856 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
857
858 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
859                        const struct inet6_skb_parm *opt)
860 {
861         const struct ipv6_pinfo *np = inet6_sk(sk);
862
863         if (np->rxopt.all) {
864                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
865                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
866                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
867                      np->rxopt.bits.rxflow) ||
868                     (opt->srcrt && (np->rxopt.bits.srcrt ||
869                      np->rxopt.bits.osrcrt)) ||
870                     ((opt->dst1 || opt->dst0) &&
871                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
872                         return true;
873         }
874         return false;
875 }
876 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
877
878 static struct packet_type ipv6_packet_type __read_mostly = {
879         .type = cpu_to_be16(ETH_P_IPV6),
880         .func = ipv6_rcv,
881         .list_func = ipv6_list_rcv,
882 };
883
884 static int __init ipv6_packet_init(void)
885 {
886         dev_add_pack(&ipv6_packet_type);
887         return 0;
888 }
889
890 static void ipv6_packet_cleanup(void)
891 {
892         dev_remove_pack(&ipv6_packet_type);
893 }
894
895 static int __net_init ipv6_init_mibs(struct net *net)
896 {
897         int i;
898
899         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
900         if (!net->mib.udp_stats_in6)
901                 return -ENOMEM;
902         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
903         if (!net->mib.udplite_stats_in6)
904                 goto err_udplite_mib;
905         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
906         if (!net->mib.ipv6_statistics)
907                 goto err_ip_mib;
908
909         for_each_possible_cpu(i) {
910                 struct ipstats_mib *af_inet6_stats;
911                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
912                 u64_stats_init(&af_inet6_stats->syncp);
913         }
914
915
916         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
917         if (!net->mib.icmpv6_statistics)
918                 goto err_icmp_mib;
919         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
920                                                 GFP_KERNEL);
921         if (!net->mib.icmpv6msg_statistics)
922                 goto err_icmpmsg_mib;
923         return 0;
924
925 err_icmpmsg_mib:
926         free_percpu(net->mib.icmpv6_statistics);
927 err_icmp_mib:
928         free_percpu(net->mib.ipv6_statistics);
929 err_ip_mib:
930         free_percpu(net->mib.udplite_stats_in6);
931 err_udplite_mib:
932         free_percpu(net->mib.udp_stats_in6);
933         return -ENOMEM;
934 }
935
936 static void ipv6_cleanup_mibs(struct net *net)
937 {
938         free_percpu(net->mib.udp_stats_in6);
939         free_percpu(net->mib.udplite_stats_in6);
940         free_percpu(net->mib.ipv6_statistics);
941         free_percpu(net->mib.icmpv6_statistics);
942         kfree(net->mib.icmpv6msg_statistics);
943 }
944
945 static int __net_init inet6_net_init(struct net *net)
946 {
947         int err = 0;
948
949         net->ipv6.sysctl.bindv6only = 0;
950         net->ipv6.sysctl.icmpv6_time = 1*HZ;
951         net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
952         net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
953         net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
954
955         /* By default, rate limit error messages.
956          * Except for pmtu discovery, it would break it.
957          * proc_do_large_bitmap needs pointer to the bitmap.
958          */
959         bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
960         bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
961         net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
962
963         net->ipv6.sysctl.flowlabel_consistency = 1;
964         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
965         net->ipv6.sysctl.idgen_retries = 3;
966         net->ipv6.sysctl.idgen_delay = 1 * HZ;
967         net->ipv6.sysctl.flowlabel_state_ranges = 0;
968         net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
969         net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
970         net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
971         net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
972         atomic_set(&net->ipv6.fib6_sernum, 1);
973
974         err = ipv6_init_mibs(net);
975         if (err)
976                 return err;
977 #ifdef CONFIG_PROC_FS
978         err = udp6_proc_init(net);
979         if (err)
980                 goto out;
981         err = tcp6_proc_init(net);
982         if (err)
983                 goto proc_tcp6_fail;
984         err = ac6_proc_init(net);
985         if (err)
986                 goto proc_ac6_fail;
987 #endif
988         return err;
989
990 #ifdef CONFIG_PROC_FS
991 proc_ac6_fail:
992         tcp6_proc_exit(net);
993 proc_tcp6_fail:
994         udp6_proc_exit(net);
995 out:
996         ipv6_cleanup_mibs(net);
997         return err;
998 #endif
999 }
1000
1001 static void __net_exit inet6_net_exit(struct net *net)
1002 {
1003 #ifdef CONFIG_PROC_FS
1004         udp6_proc_exit(net);
1005         tcp6_proc_exit(net);
1006         ac6_proc_exit(net);
1007 #endif
1008         ipv6_cleanup_mibs(net);
1009 }
1010
1011 static struct pernet_operations inet6_net_ops = {
1012         .init = inet6_net_init,
1013         .exit = inet6_net_exit,
1014 };
1015
1016 static int ipv6_route_input(struct sk_buff *skb)
1017 {
1018         ip6_route_input(skb);
1019         return skb_dst(skb)->error;
1020 }
1021
1022 static const struct ipv6_stub ipv6_stub_impl = {
1023         .ipv6_sock_mc_join = ipv6_sock_mc_join,
1024         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1025         .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1026         .ipv6_route_input  = ipv6_route_input,
1027         .fib6_get_table    = fib6_get_table,
1028         .fib6_table_lookup = fib6_table_lookup,
1029         .fib6_lookup       = fib6_lookup,
1030         .fib6_select_path  = fib6_select_path,
1031         .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1032         .fib6_nh_init      = fib6_nh_init,
1033         .fib6_nh_release   = fib6_nh_release,
1034         .fib6_nh_release_dsts = fib6_nh_release_dsts,
1035         .fib6_update_sernum = fib6_update_sernum_stub,
1036         .fib6_rt_update    = fib6_rt_update,
1037         .ip6_del_rt        = ip6_del_rt,
1038         .udpv6_encap_enable = udpv6_encap_enable,
1039         .ndisc_send_na = ndisc_send_na,
1040 #if IS_ENABLED(CONFIG_XFRM)
1041         .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1042         .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1043         .xfrm6_rcv_encap = xfrm6_rcv_encap,
1044 #endif
1045         .nd_tbl = &nd_tbl,
1046         .ipv6_fragment = ip6_fragment,
1047 };
1048
1049 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1050         .inet6_bind = __inet6_bind,
1051         .udp6_lib_lookup = __udp6_lib_lookup,
1052 };
1053
1054 static int __init inet6_init(void)
1055 {
1056         struct list_head *r;
1057         int err = 0;
1058
1059         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1060
1061         /* Register the socket-side information for inet6_create.  */
1062         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1063                 INIT_LIST_HEAD(r);
1064
1065         if (disable_ipv6_mod) {
1066                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1067                 goto out;
1068         }
1069
1070         err = proto_register(&tcpv6_prot, 1);
1071         if (err)
1072                 goto out;
1073
1074         err = proto_register(&udpv6_prot, 1);
1075         if (err)
1076                 goto out_unregister_tcp_proto;
1077
1078         err = proto_register(&udplitev6_prot, 1);
1079         if (err)
1080                 goto out_unregister_udp_proto;
1081
1082         err = proto_register(&rawv6_prot, 1);
1083         if (err)
1084                 goto out_unregister_udplite_proto;
1085
1086         err = proto_register(&pingv6_prot, 1);
1087         if (err)
1088                 goto out_unregister_raw_proto;
1089
1090         /* We MUST register RAW sockets before we create the ICMP6,
1091          * IGMP6, or NDISC control sockets.
1092          */
1093         err = rawv6_init();
1094         if (err)
1095                 goto out_unregister_ping_proto;
1096
1097         /* Register the family here so that the init calls below will
1098          * be able to create sockets. (?? is this dangerous ??)
1099          */
1100         err = sock_register(&inet6_family_ops);
1101         if (err)
1102                 goto out_sock_register_fail;
1103
1104         /*
1105          *      ipngwg API draft makes clear that the correct semantics
1106          *      for TCP and UDP is to consider one TCP and UDP instance
1107          *      in a host available by both INET and INET6 APIs and
1108          *      able to communicate via both network protocols.
1109          */
1110
1111         err = register_pernet_subsys(&inet6_net_ops);
1112         if (err)
1113                 goto register_pernet_fail;
1114         err = ip6_mr_init();
1115         if (err)
1116                 goto ipmr_fail;
1117         err = icmpv6_init();
1118         if (err)
1119                 goto icmp_fail;
1120         err = ndisc_init();
1121         if (err)
1122                 goto ndisc_fail;
1123         err = igmp6_init();
1124         if (err)
1125                 goto igmp_fail;
1126
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 = ndisc_late_init();
1146         if (err)
1147                 goto ndisc_late_fail;
1148         err = ip6_flowlabel_init();
1149         if (err)
1150                 goto ip6_flowlabel_fail;
1151         err = ipv6_anycast_init();
1152         if (err)
1153                 goto ipv6_anycast_fail;
1154         err = addrconf_init();
1155         if (err)
1156                 goto addrconf_fail;
1157
1158         /* Init v6 extension headers. */
1159         err = ipv6_exthdrs_init();
1160         if (err)
1161                 goto ipv6_exthdrs_fail;
1162
1163         err = ipv6_frag_init();
1164         if (err)
1165                 goto ipv6_frag_fail;
1166
1167         /* Init v6 transport protocols. */
1168         err = udpv6_init();
1169         if (err)
1170                 goto udpv6_fail;
1171
1172         err = udplitev6_init();
1173         if (err)
1174                 goto udplitev6_fail;
1175
1176         err = udpv6_offload_init();
1177         if (err)
1178                 goto udpv6_offload_fail;
1179
1180         err = tcpv6_init();
1181         if (err)
1182                 goto tcpv6_fail;
1183
1184         err = ipv6_packet_init();
1185         if (err)
1186                 goto ipv6_packet_fail;
1187
1188         err = pingv6_init();
1189         if (err)
1190                 goto pingv6_fail;
1191
1192         err = calipso_init();
1193         if (err)
1194                 goto calipso_fail;
1195
1196         err = seg6_init();
1197         if (err)
1198                 goto seg6_fail;
1199
1200         err = rpl_init();
1201         if (err)
1202                 goto rpl_fail;
1203
1204         err = igmp6_late_init();
1205         if (err)
1206                 goto igmp6_late_err;
1207
1208 #ifdef CONFIG_SYSCTL
1209         err = ipv6_sysctl_register();
1210         if (err)
1211                 goto sysctl_fail;
1212 #endif
1213
1214         /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1215         wmb();
1216         ipv6_stub = &ipv6_stub_impl;
1217         ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1218 out:
1219         return err;
1220
1221 #ifdef CONFIG_SYSCTL
1222 sysctl_fail:
1223         igmp6_late_cleanup();
1224 #endif
1225 igmp6_late_err:
1226         rpl_exit();
1227 rpl_fail:
1228         seg6_exit();
1229 seg6_fail:
1230         calipso_exit();
1231 calipso_fail:
1232         pingv6_exit();
1233 pingv6_fail:
1234         ipv6_packet_cleanup();
1235 ipv6_packet_fail:
1236         tcpv6_exit();
1237 tcpv6_fail:
1238         udpv6_offload_exit();
1239 udpv6_offload_fail:
1240         udplitev6_exit();
1241 udplitev6_fail:
1242         udpv6_exit();
1243 udpv6_fail:
1244         ipv6_frag_exit();
1245 ipv6_frag_fail:
1246         ipv6_exthdrs_exit();
1247 ipv6_exthdrs_fail:
1248         addrconf_cleanup();
1249 addrconf_fail:
1250         ipv6_anycast_cleanup();
1251 ipv6_anycast_fail:
1252         ip6_flowlabel_cleanup();
1253 ip6_flowlabel_fail:
1254         ndisc_late_cleanup();
1255 ndisc_late_fail:
1256         ip6_route_cleanup();
1257 ip6_route_fail:
1258 #ifdef CONFIG_PROC_FS
1259         if6_proc_exit();
1260 proc_if6_fail:
1261         ipv6_misc_proc_exit();
1262 proc_misc6_fail:
1263         udplite6_proc_exit();
1264 proc_udplite6_fail:
1265         raw6_proc_exit();
1266 proc_raw6_fail:
1267 #endif
1268         ipv6_netfilter_fini();
1269 netfilter_fail:
1270         igmp6_cleanup();
1271 igmp_fail:
1272         ndisc_cleanup();
1273 ndisc_fail:
1274         icmpv6_cleanup();
1275 icmp_fail:
1276         ip6_mr_cleanup();
1277 ipmr_fail:
1278         unregister_pernet_subsys(&inet6_net_ops);
1279 register_pernet_fail:
1280         sock_unregister(PF_INET6);
1281         rtnl_unregister_all(PF_INET6);
1282 out_sock_register_fail:
1283         rawv6_exit();
1284 out_unregister_ping_proto:
1285         proto_unregister(&pingv6_prot);
1286 out_unregister_raw_proto:
1287         proto_unregister(&rawv6_prot);
1288 out_unregister_udplite_proto:
1289         proto_unregister(&udplitev6_prot);
1290 out_unregister_udp_proto:
1291         proto_unregister(&udpv6_prot);
1292 out_unregister_tcp_proto:
1293         proto_unregister(&tcpv6_prot);
1294         goto out;
1295 }
1296 module_init(inet6_init);
1297
1298 MODULE_ALIAS_NETPROTO(PF_INET6);