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
ipv6_mod_enabled(void)97 bool ipv6_mod_enabled(void)
98 {
99 return disable_ipv6_mod == 0;
100 }
101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
102
inet6_sk_generic(struct sock * sk)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
inet6_create(struct net * net,struct socket * sock,int protocol,int kern)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 (READ_ONCE(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
__inet6_bind(struct sock * sk,struct sockaddr * uaddr,int addr_len,u32 flags)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 */
inet6_bind(struct socket * sock,struct sockaddr * uaddr,int addr_len)439 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
440 {
441 struct sock *sk = sock->sk;
442 const struct proto *prot;
443 int err = 0;
444
445 /* If the socket has its own bind function then use it. */
446 prot = READ_ONCE(sk->sk_prot);
447 if (prot->bind)
448 return prot->bind(sk, uaddr, addr_len);
449
450 if (addr_len < SIN6_LEN_RFC2133)
451 return -EINVAL;
452
453 /* BPF prog is run before any checks are done so that if the prog
454 * changes context in a wrong way it will be caught.
455 */
456 err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
457 if (err)
458 return err;
459
460 return __inet6_bind(sk, uaddr, addr_len, BIND_WITH_LOCK);
461 }
462 EXPORT_SYMBOL(inet6_bind);
463
inet6_release(struct socket * sock)464 int inet6_release(struct socket *sock)
465 {
466 struct sock *sk = sock->sk;
467
468 if (!sk)
469 return -EINVAL;
470
471 /* Free mc lists */
472 ipv6_sock_mc_close(sk);
473
474 /* Free ac lists */
475 ipv6_sock_ac_close(sk);
476
477 return inet_release(sock);
478 }
479 EXPORT_SYMBOL(inet6_release);
480
inet6_destroy_sock(struct sock * sk)481 void inet6_destroy_sock(struct sock *sk)
482 {
483 struct ipv6_pinfo *np = inet6_sk(sk);
484 struct sk_buff *skb;
485 struct ipv6_txoptions *opt;
486
487 /* Release rx options */
488
489 skb = xchg(&np->pktoptions, NULL);
490 kfree_skb(skb);
491
492 skb = xchg(&np->rxpmtu, NULL);
493 kfree_skb(skb);
494
495 /* Free flowlabels */
496 fl6_free_socklist(sk);
497
498 /* Free tx options */
499
500 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
501 if (opt) {
502 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
503 txopt_put(opt);
504 }
505 }
506 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
507
508 /*
509 * This does both peername and sockname.
510 */
inet6_getname(struct socket * sock,struct sockaddr * uaddr,int peer)511 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
512 int peer)
513 {
514 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
515 struct sock *sk = sock->sk;
516 struct inet_sock *inet = inet_sk(sk);
517 struct ipv6_pinfo *np = inet6_sk(sk);
518
519 sin->sin6_family = AF_INET6;
520 sin->sin6_flowinfo = 0;
521 sin->sin6_scope_id = 0;
522 if (peer) {
523 if (!inet->inet_dport)
524 return -ENOTCONN;
525 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
526 peer == 1)
527 return -ENOTCONN;
528 sin->sin6_port = inet->inet_dport;
529 sin->sin6_addr = sk->sk_v6_daddr;
530 if (np->sndflow)
531 sin->sin6_flowinfo = np->flow_label;
532 } else {
533 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
534 sin->sin6_addr = np->saddr;
535 else
536 sin->sin6_addr = sk->sk_v6_rcv_saddr;
537 sin->sin6_port = inet->inet_sport;
538 }
539 if (cgroup_bpf_enabled)
540 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin,
541 peer ? BPF_CGROUP_INET6_GETPEERNAME :
542 BPF_CGROUP_INET6_GETSOCKNAME,
543 NULL);
544 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
545 sk->sk_bound_dev_if);
546 return sizeof(*sin);
547 }
548 EXPORT_SYMBOL(inet6_getname);
549
inet6_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)550 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
551 {
552 void __user *argp = (void __user *)arg;
553 struct sock *sk = sock->sk;
554 struct net *net = sock_net(sk);
555 const struct proto *prot;
556
557 switch (cmd) {
558 case SIOCADDRT:
559 case SIOCDELRT: {
560 struct in6_rtmsg rtmsg;
561
562 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
563 return -EFAULT;
564 return ipv6_route_ioctl(net, cmd, &rtmsg);
565 }
566 case SIOCSIFADDR:
567 return addrconf_add_ifaddr(net, argp);
568 case SIOCDIFADDR:
569 return addrconf_del_ifaddr(net, argp);
570 case SIOCSIFDSTADDR:
571 return addrconf_set_dstaddr(net, argp);
572 default:
573 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
574 prot = READ_ONCE(sk->sk_prot);
575 if (!prot->ioctl)
576 return -ENOIOCTLCMD;
577 return prot->ioctl(sk, cmd, arg);
578 }
579 /*NOTREACHED*/
580 return 0;
581 }
582 EXPORT_SYMBOL(inet6_ioctl);
583
584 #ifdef CONFIG_COMPAT
585 struct compat_in6_rtmsg {
586 struct in6_addr rtmsg_dst;
587 struct in6_addr rtmsg_src;
588 struct in6_addr rtmsg_gateway;
589 u32 rtmsg_type;
590 u16 rtmsg_dst_len;
591 u16 rtmsg_src_len;
592 u32 rtmsg_metric;
593 u32 rtmsg_info;
594 u32 rtmsg_flags;
595 s32 rtmsg_ifindex;
596 };
597
inet6_compat_routing_ioctl(struct sock * sk,unsigned int cmd,struct compat_in6_rtmsg __user * ur)598 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
599 struct compat_in6_rtmsg __user *ur)
600 {
601 struct in6_rtmsg rt;
602
603 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
604 3 * sizeof(struct in6_addr)) ||
605 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
606 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
607 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
608 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
609 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
610 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
611 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
612 return -EFAULT;
613
614
615 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
616 }
617
inet6_compat_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)618 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
619 {
620 void __user *argp = compat_ptr(arg);
621 struct sock *sk = sock->sk;
622
623 switch (cmd) {
624 case SIOCADDRT:
625 case SIOCDELRT:
626 return inet6_compat_routing_ioctl(sk, cmd, argp);
627 default:
628 return -ENOIOCTLCMD;
629 }
630 }
631 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
632 #endif /* CONFIG_COMPAT */
633
634 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
635 size_t));
inet6_sendmsg(struct socket * sock,struct msghdr * msg,size_t size)636 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
637 {
638 struct sock *sk = sock->sk;
639 const struct proto *prot;
640
641 if (unlikely(inet_send_prepare(sk)))
642 return -EAGAIN;
643
644 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
645 prot = READ_ONCE(sk->sk_prot);
646 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
647 sk, msg, size);
648 }
649
650 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
651 size_t, int, int, int *));
inet6_recvmsg(struct socket * sock,struct msghdr * msg,size_t size,int flags)652 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
653 int flags)
654 {
655 struct sock *sk = sock->sk;
656 const struct proto *prot;
657 int addr_len = 0;
658 int err;
659
660 if (likely(!(flags & MSG_ERRQUEUE)))
661 sock_rps_record_flow(sk);
662
663 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
664 prot = READ_ONCE(sk->sk_prot);
665 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
666 sk, msg, size, flags & MSG_DONTWAIT,
667 flags & ~MSG_DONTWAIT, &addr_len);
668 if (err >= 0)
669 msg->msg_namelen = addr_len;
670 return err;
671 }
672
673 const struct proto_ops inet6_stream_ops = {
674 .family = PF_INET6,
675 .owner = THIS_MODULE,
676 .release = inet6_release,
677 .bind = inet6_bind,
678 .connect = inet_stream_connect, /* ok */
679 .socketpair = sock_no_socketpair, /* a do nothing */
680 .accept = inet_accept, /* ok */
681 .getname = inet6_getname,
682 .poll = tcp_poll, /* ok */
683 .ioctl = inet6_ioctl, /* must change */
684 .gettstamp = sock_gettstamp,
685 .listen = inet_listen, /* ok */
686 .shutdown = inet_shutdown, /* ok */
687 .setsockopt = sock_common_setsockopt, /* ok */
688 .getsockopt = sock_common_getsockopt, /* ok */
689 .sendmsg = inet6_sendmsg, /* retpoline's sake */
690 .recvmsg = inet6_recvmsg, /* retpoline's sake */
691 #ifdef CONFIG_MMU
692 .mmap = tcp_mmap,
693 #endif
694 .sendpage = inet_sendpage,
695 .sendmsg_locked = tcp_sendmsg_locked,
696 .sendpage_locked = tcp_sendpage_locked,
697 .splice_read = tcp_splice_read,
698 .read_sock = tcp_read_sock,
699 .peek_len = tcp_peek_len,
700 #ifdef CONFIG_COMPAT
701 .compat_ioctl = inet6_compat_ioctl,
702 #endif
703 .set_rcvlowat = tcp_set_rcvlowat,
704 };
705
706 const struct proto_ops inet6_dgram_ops = {
707 .family = PF_INET6,
708 .owner = THIS_MODULE,
709 .release = inet6_release,
710 .bind = inet6_bind,
711 .connect = inet_dgram_connect, /* ok */
712 .socketpair = sock_no_socketpair, /* a do nothing */
713 .accept = sock_no_accept, /* a do nothing */
714 .getname = inet6_getname,
715 .poll = udp_poll, /* ok */
716 .ioctl = inet6_ioctl, /* must change */
717 .gettstamp = sock_gettstamp,
718 .listen = sock_no_listen, /* ok */
719 .shutdown = inet_shutdown, /* ok */
720 .setsockopt = sock_common_setsockopt, /* ok */
721 .getsockopt = sock_common_getsockopt, /* ok */
722 .sendmsg = inet6_sendmsg, /* retpoline's sake */
723 .recvmsg = inet6_recvmsg, /* retpoline's sake */
724 .mmap = sock_no_mmap,
725 .sendpage = sock_no_sendpage,
726 .set_peek_off = sk_set_peek_off,
727 #ifdef CONFIG_COMPAT
728 .compat_ioctl = inet6_compat_ioctl,
729 #endif
730 };
731
732 static const struct net_proto_family inet6_family_ops = {
733 .family = PF_INET6,
734 .create = inet6_create,
735 .owner = THIS_MODULE,
736 };
737
inet6_register_protosw(struct inet_protosw * p)738 int inet6_register_protosw(struct inet_protosw *p)
739 {
740 struct list_head *lh;
741 struct inet_protosw *answer;
742 struct list_head *last_perm;
743 int protocol = p->protocol;
744 int ret;
745
746 spin_lock_bh(&inetsw6_lock);
747
748 ret = -EINVAL;
749 if (p->type >= SOCK_MAX)
750 goto out_illegal;
751
752 /* If we are trying to override a permanent protocol, bail. */
753 answer = NULL;
754 ret = -EPERM;
755 last_perm = &inetsw6[p->type];
756 list_for_each(lh, &inetsw6[p->type]) {
757 answer = list_entry(lh, struct inet_protosw, list);
758
759 /* Check only the non-wild match. */
760 if (INET_PROTOSW_PERMANENT & answer->flags) {
761 if (protocol == answer->protocol)
762 break;
763 last_perm = lh;
764 }
765
766 answer = NULL;
767 }
768 if (answer)
769 goto out_permanent;
770
771 /* Add the new entry after the last permanent entry if any, so that
772 * the new entry does not override a permanent entry when matched with
773 * a wild-card protocol. But it is allowed to override any existing
774 * non-permanent entry. This means that when we remove this entry, the
775 * system automatically returns to the old behavior.
776 */
777 list_add_rcu(&p->list, last_perm);
778 ret = 0;
779 out:
780 spin_unlock_bh(&inetsw6_lock);
781 return ret;
782
783 out_permanent:
784 pr_err("Attempt to override permanent protocol %d\n", protocol);
785 goto out;
786
787 out_illegal:
788 pr_err("Ignoring attempt to register invalid socket type %d\n",
789 p->type);
790 goto out;
791 }
792 EXPORT_SYMBOL(inet6_register_protosw);
793
794 void
inet6_unregister_protosw(struct inet_protosw * p)795 inet6_unregister_protosw(struct inet_protosw *p)
796 {
797 if (INET_PROTOSW_PERMANENT & p->flags) {
798 pr_err("Attempt to unregister permanent protocol %d\n",
799 p->protocol);
800 } else {
801 spin_lock_bh(&inetsw6_lock);
802 list_del_rcu(&p->list);
803 spin_unlock_bh(&inetsw6_lock);
804
805 synchronize_net();
806 }
807 }
808 EXPORT_SYMBOL(inet6_unregister_protosw);
809
inet6_sk_rebuild_header(struct sock * sk)810 int inet6_sk_rebuild_header(struct sock *sk)
811 {
812 struct ipv6_pinfo *np = inet6_sk(sk);
813 struct dst_entry *dst;
814
815 dst = __sk_dst_check(sk, np->dst_cookie);
816
817 if (!dst) {
818 struct inet_sock *inet = inet_sk(sk);
819 struct in6_addr *final_p, final;
820 struct flowi6 fl6;
821
822 memset(&fl6, 0, sizeof(fl6));
823 fl6.flowi6_proto = sk->sk_protocol;
824 fl6.daddr = sk->sk_v6_daddr;
825 fl6.saddr = np->saddr;
826 fl6.flowlabel = np->flow_label;
827 fl6.flowi6_oif = sk->sk_bound_dev_if;
828 fl6.flowi6_mark = sk->sk_mark;
829 fl6.fl6_dport = inet->inet_dport;
830 fl6.fl6_sport = inet->inet_sport;
831 fl6.flowi6_uid = sk->sk_uid;
832 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
833
834 rcu_read_lock();
835 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
836 &final);
837 rcu_read_unlock();
838
839 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
840 if (IS_ERR(dst)) {
841 sk->sk_route_caps = 0;
842 sk->sk_err_soft = -PTR_ERR(dst);
843 return PTR_ERR(dst);
844 }
845
846 ip6_dst_store(sk, dst, NULL, NULL);
847 }
848
849 return 0;
850 }
851 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
852
ipv6_opt_accepted(const struct sock * sk,const struct sk_buff * skb,const struct inet6_skb_parm * opt)853 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
854 const struct inet6_skb_parm *opt)
855 {
856 const struct ipv6_pinfo *np = inet6_sk(sk);
857
858 if (np->rxopt.all) {
859 if (((opt->flags & IP6SKB_HOPBYHOP) &&
860 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
861 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
862 np->rxopt.bits.rxflow) ||
863 (opt->srcrt && (np->rxopt.bits.srcrt ||
864 np->rxopt.bits.osrcrt)) ||
865 ((opt->dst1 || opt->dst0) &&
866 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
867 return true;
868 }
869 return false;
870 }
871 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
872
873 static struct packet_type ipv6_packet_type __read_mostly = {
874 .type = cpu_to_be16(ETH_P_IPV6),
875 .func = ipv6_rcv,
876 .list_func = ipv6_list_rcv,
877 };
878
ipv6_packet_init(void)879 static int __init ipv6_packet_init(void)
880 {
881 dev_add_pack(&ipv6_packet_type);
882 return 0;
883 }
884
ipv6_packet_cleanup(void)885 static void ipv6_packet_cleanup(void)
886 {
887 dev_remove_pack(&ipv6_packet_type);
888 }
889
ipv6_init_mibs(struct net * net)890 static int __net_init ipv6_init_mibs(struct net *net)
891 {
892 int i;
893
894 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
895 if (!net->mib.udp_stats_in6)
896 return -ENOMEM;
897 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
898 if (!net->mib.udplite_stats_in6)
899 goto err_udplite_mib;
900 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
901 if (!net->mib.ipv6_statistics)
902 goto err_ip_mib;
903
904 for_each_possible_cpu(i) {
905 struct ipstats_mib *af_inet6_stats;
906 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
907 u64_stats_init(&af_inet6_stats->syncp);
908 }
909
910
911 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
912 if (!net->mib.icmpv6_statistics)
913 goto err_icmp_mib;
914 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
915 GFP_KERNEL);
916 if (!net->mib.icmpv6msg_statistics)
917 goto err_icmpmsg_mib;
918 return 0;
919
920 err_icmpmsg_mib:
921 free_percpu(net->mib.icmpv6_statistics);
922 err_icmp_mib:
923 free_percpu(net->mib.ipv6_statistics);
924 err_ip_mib:
925 free_percpu(net->mib.udplite_stats_in6);
926 err_udplite_mib:
927 free_percpu(net->mib.udp_stats_in6);
928 return -ENOMEM;
929 }
930
ipv6_cleanup_mibs(struct net * net)931 static void ipv6_cleanup_mibs(struct net *net)
932 {
933 free_percpu(net->mib.udp_stats_in6);
934 free_percpu(net->mib.udplite_stats_in6);
935 free_percpu(net->mib.ipv6_statistics);
936 free_percpu(net->mib.icmpv6_statistics);
937 kfree(net->mib.icmpv6msg_statistics);
938 }
939
inet6_net_init(struct net * net)940 static int __net_init inet6_net_init(struct net *net)
941 {
942 int err = 0;
943
944 net->ipv6.sysctl.bindv6only = 0;
945 net->ipv6.sysctl.icmpv6_time = 1*HZ;
946 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
947 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
948 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
949
950 /* By default, rate limit error messages.
951 * Except for pmtu discovery, it would break it.
952 * proc_do_large_bitmap needs pointer to the bitmap.
953 */
954 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
955 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
956 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
957
958 net->ipv6.sysctl.flowlabel_consistency = 1;
959 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
960 net->ipv6.sysctl.idgen_retries = 3;
961 net->ipv6.sysctl.idgen_delay = 1 * HZ;
962 net->ipv6.sysctl.flowlabel_state_ranges = 0;
963 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
964 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
965 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
966 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
967 atomic_set(&net->ipv6.fib6_sernum, 1);
968
969 err = ipv6_init_mibs(net);
970 if (err)
971 return err;
972 #ifdef CONFIG_PROC_FS
973 err = udp6_proc_init(net);
974 if (err)
975 goto out;
976 err = tcp6_proc_init(net);
977 if (err)
978 goto proc_tcp6_fail;
979 err = ac6_proc_init(net);
980 if (err)
981 goto proc_ac6_fail;
982 #endif
983 return err;
984
985 #ifdef CONFIG_PROC_FS
986 proc_ac6_fail:
987 tcp6_proc_exit(net);
988 proc_tcp6_fail:
989 udp6_proc_exit(net);
990 out:
991 ipv6_cleanup_mibs(net);
992 return err;
993 #endif
994 }
995
inet6_net_exit(struct net * net)996 static void __net_exit inet6_net_exit(struct net *net)
997 {
998 #ifdef CONFIG_PROC_FS
999 udp6_proc_exit(net);
1000 tcp6_proc_exit(net);
1001 ac6_proc_exit(net);
1002 #endif
1003 ipv6_cleanup_mibs(net);
1004 }
1005
1006 static struct pernet_operations inet6_net_ops = {
1007 .init = inet6_net_init,
1008 .exit = inet6_net_exit,
1009 };
1010
ipv6_route_input(struct sk_buff * skb)1011 static int ipv6_route_input(struct sk_buff *skb)
1012 {
1013 ip6_route_input(skb);
1014 return skb_dst(skb)->error;
1015 }
1016
1017 static const struct ipv6_stub ipv6_stub_impl = {
1018 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1019 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1020 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1021 .ipv6_route_input = ipv6_route_input,
1022 .fib6_get_table = fib6_get_table,
1023 .fib6_table_lookup = fib6_table_lookup,
1024 .fib6_lookup = fib6_lookup,
1025 .fib6_select_path = fib6_select_path,
1026 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1027 .fib6_nh_init = fib6_nh_init,
1028 .fib6_nh_release = fib6_nh_release,
1029 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1030 .fib6_update_sernum = fib6_update_sernum_stub,
1031 .fib6_rt_update = fib6_rt_update,
1032 .ip6_del_rt = ip6_del_rt,
1033 .udpv6_encap_enable = udpv6_encap_enable,
1034 .ndisc_send_na = ndisc_send_na,
1035 #if IS_ENABLED(CONFIG_XFRM)
1036 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1037 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1038 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1039 #endif
1040 .nd_tbl = &nd_tbl,
1041 .ipv6_fragment = ip6_fragment,
1042 };
1043
1044 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1045 .inet6_bind = __inet6_bind,
1046 .udp6_lib_lookup = __udp6_lib_lookup,
1047 };
1048
inet6_init(void)1049 static int __init inet6_init(void)
1050 {
1051 struct list_head *r;
1052 int err = 0;
1053
1054 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1055
1056 /* Register the socket-side information for inet6_create. */
1057 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1058 INIT_LIST_HEAD(r);
1059
1060 if (disable_ipv6_mod) {
1061 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1062 goto out;
1063 }
1064
1065 err = proto_register(&tcpv6_prot, 1);
1066 if (err)
1067 goto out;
1068
1069 err = proto_register(&udpv6_prot, 1);
1070 if (err)
1071 goto out_unregister_tcp_proto;
1072
1073 err = proto_register(&udplitev6_prot, 1);
1074 if (err)
1075 goto out_unregister_udp_proto;
1076
1077 err = proto_register(&rawv6_prot, 1);
1078 if (err)
1079 goto out_unregister_udplite_proto;
1080
1081 err = proto_register(&pingv6_prot, 1);
1082 if (err)
1083 goto out_unregister_raw_proto;
1084
1085 /* We MUST register RAW sockets before we create the ICMP6,
1086 * IGMP6, or NDISC control sockets.
1087 */
1088 err = rawv6_init();
1089 if (err)
1090 goto out_unregister_ping_proto;
1091
1092 /* Register the family here so that the init calls below will
1093 * be able to create sockets. (?? is this dangerous ??)
1094 */
1095 err = sock_register(&inet6_family_ops);
1096 if (err)
1097 goto out_sock_register_fail;
1098
1099 /*
1100 * ipngwg API draft makes clear that the correct semantics
1101 * for TCP and UDP is to consider one TCP and UDP instance
1102 * in a host available by both INET and INET6 APIs and
1103 * able to communicate via both network protocols.
1104 */
1105
1106 err = register_pernet_subsys(&inet6_net_ops);
1107 if (err)
1108 goto register_pernet_fail;
1109 err = ip6_mr_init();
1110 if (err)
1111 goto ipmr_fail;
1112 err = icmpv6_init();
1113 if (err)
1114 goto icmp_fail;
1115 err = ndisc_init();
1116 if (err)
1117 goto ndisc_fail;
1118 err = igmp6_init();
1119 if (err)
1120 goto igmp_fail;
1121
1122 err = ipv6_netfilter_init();
1123 if (err)
1124 goto netfilter_fail;
1125 /* Create /proc/foo6 entries. */
1126 #ifdef CONFIG_PROC_FS
1127 err = -ENOMEM;
1128 if (raw6_proc_init())
1129 goto proc_raw6_fail;
1130 if (udplite6_proc_init())
1131 goto proc_udplite6_fail;
1132 if (ipv6_misc_proc_init())
1133 goto proc_misc6_fail;
1134 if (if6_proc_init())
1135 goto proc_if6_fail;
1136 #endif
1137 err = ip6_route_init();
1138 if (err)
1139 goto ip6_route_fail;
1140 err = ndisc_late_init();
1141 if (err)
1142 goto ndisc_late_fail;
1143 err = ip6_flowlabel_init();
1144 if (err)
1145 goto ip6_flowlabel_fail;
1146 err = ipv6_anycast_init();
1147 if (err)
1148 goto ipv6_anycast_fail;
1149 err = addrconf_init();
1150 if (err)
1151 goto addrconf_fail;
1152
1153 /* Init v6 extension headers. */
1154 err = ipv6_exthdrs_init();
1155 if (err)
1156 goto ipv6_exthdrs_fail;
1157
1158 err = ipv6_frag_init();
1159 if (err)
1160 goto ipv6_frag_fail;
1161
1162 /* Init v6 transport protocols. */
1163 err = udpv6_init();
1164 if (err)
1165 goto udpv6_fail;
1166
1167 err = udplitev6_init();
1168 if (err)
1169 goto udplitev6_fail;
1170
1171 err = udpv6_offload_init();
1172 if (err)
1173 goto udpv6_offload_fail;
1174
1175 err = tcpv6_init();
1176 if (err)
1177 goto tcpv6_fail;
1178
1179 err = ipv6_packet_init();
1180 if (err)
1181 goto ipv6_packet_fail;
1182
1183 err = pingv6_init();
1184 if (err)
1185 goto pingv6_fail;
1186
1187 err = calipso_init();
1188 if (err)
1189 goto calipso_fail;
1190
1191 err = seg6_init();
1192 if (err)
1193 goto seg6_fail;
1194
1195 err = rpl_init();
1196 if (err)
1197 goto rpl_fail;
1198
1199 err = igmp6_late_init();
1200 if (err)
1201 goto igmp6_late_err;
1202
1203 #ifdef CONFIG_SYSCTL
1204 err = ipv6_sysctl_register();
1205 if (err)
1206 goto sysctl_fail;
1207 #endif
1208
1209 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1210 wmb();
1211 ipv6_stub = &ipv6_stub_impl;
1212 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1213 out:
1214 return err;
1215
1216 #ifdef CONFIG_SYSCTL
1217 sysctl_fail:
1218 igmp6_late_cleanup();
1219 #endif
1220 igmp6_late_err:
1221 rpl_exit();
1222 rpl_fail:
1223 seg6_exit();
1224 seg6_fail:
1225 calipso_exit();
1226 calipso_fail:
1227 pingv6_exit();
1228 pingv6_fail:
1229 ipv6_packet_cleanup();
1230 ipv6_packet_fail:
1231 tcpv6_exit();
1232 tcpv6_fail:
1233 udpv6_offload_exit();
1234 udpv6_offload_fail:
1235 udplitev6_exit();
1236 udplitev6_fail:
1237 udpv6_exit();
1238 udpv6_fail:
1239 ipv6_frag_exit();
1240 ipv6_frag_fail:
1241 ipv6_exthdrs_exit();
1242 ipv6_exthdrs_fail:
1243 addrconf_cleanup();
1244 addrconf_fail:
1245 ipv6_anycast_cleanup();
1246 ipv6_anycast_fail:
1247 ip6_flowlabel_cleanup();
1248 ip6_flowlabel_fail:
1249 ndisc_late_cleanup();
1250 ndisc_late_fail:
1251 ip6_route_cleanup();
1252 ip6_route_fail:
1253 #ifdef CONFIG_PROC_FS
1254 if6_proc_exit();
1255 proc_if6_fail:
1256 ipv6_misc_proc_exit();
1257 proc_misc6_fail:
1258 udplite6_proc_exit();
1259 proc_udplite6_fail:
1260 raw6_proc_exit();
1261 proc_raw6_fail:
1262 #endif
1263 ipv6_netfilter_fini();
1264 netfilter_fail:
1265 igmp6_cleanup();
1266 igmp_fail:
1267 ndisc_cleanup();
1268 ndisc_fail:
1269 icmpv6_cleanup();
1270 icmp_fail:
1271 ip6_mr_cleanup();
1272 ipmr_fail:
1273 unregister_pernet_subsys(&inet6_net_ops);
1274 register_pernet_fail:
1275 sock_unregister(PF_INET6);
1276 rtnl_unregister_all(PF_INET6);
1277 out_sock_register_fail:
1278 rawv6_exit();
1279 out_unregister_ping_proto:
1280 proto_unregister(&pingv6_prot);
1281 out_unregister_raw_proto:
1282 proto_unregister(&rawv6_prot);
1283 out_unregister_udplite_proto:
1284 proto_unregister(&udplitev6_prot);
1285 out_unregister_udp_proto:
1286 proto_unregister(&udpv6_prot);
1287 out_unregister_tcp_proto:
1288 proto_unregister(&tcpv6_prot);
1289 goto out;
1290 }
1291 module_init(inet6_init);
1292
1293 MODULE_ALIAS_NETPROTO(PF_INET6);
1294