xref: /kernel/linux/linux-5.10/net/ipv6/af_inet6.c (revision 8c2ecf20)
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 *	PF_INET6 socket protocol family
4 *	Linux INET6 implementation
5 *
6 *	Authors:
7 *	Pedro Roque		<roque@di.fc.ul.pt>
8 *
9 *	Adapted from linux/net/ipv4/af_inet.c
10 *
11 *	Fixes:
12 *	piggy, Karl Knutson	:	Socket protocol table
13 *	Hideaki YOSHIFUJI	:	sin6_scope_id support
14 *	Arnaldo Melo		:	check proc_net_create return, cleanups
15 */
16
17#define pr_fmt(fmt) "IPv6: " fmt
18
19#include <linux/module.h>
20#include <linux/capability.h>
21#include <linux/errno.h>
22#include <linux/types.h>
23#include <linux/socket.h>
24#include <linux/in.h>
25#include <linux/kernel.h>
26#include <linux/timer.h>
27#include <linux/string.h>
28#include <linux/sockios.h>
29#include <linux/net.h>
30#include <linux/fcntl.h>
31#include <linux/mm.h>
32#include <linux/interrupt.h>
33#include <linux/proc_fs.h>
34#include <linux/stat.h>
35#include <linux/init.h>
36#include <linux/slab.h>
37
38#include <linux/inet.h>
39#include <linux/netdevice.h>
40#include <linux/icmpv6.h>
41#include <linux/netfilter_ipv6.h>
42
43#include <net/ip.h>
44#include <net/ipv6.h>
45#include <net/udp.h>
46#include <net/udplite.h>
47#include <net/tcp.h>
48#include <net/ping.h>
49#include <net/protocol.h>
50#include <net/inet_common.h>
51#include <net/route.h>
52#include <net/transp_v6.h>
53#include <net/ip6_route.h>
54#include <net/addrconf.h>
55#include <net/ipv6_stubs.h>
56#include <net/ndisc.h>
57#ifdef CONFIG_IPV6_TUNNEL
58#include <net/ip6_tunnel.h>
59#endif
60#include <net/calipso.h>
61#include <net/seg6.h>
62#include <net/rpl.h>
63#include <net/compat.h>
64#include <net/xfrm.h>
65
66#include <linux/uaccess.h>
67#include <linux/mroute6.h>
68
69#include "ip6_offload.h"
70
71MODULE_AUTHOR("Cast of dozens");
72MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
73MODULE_LICENSE("GPL");
74
75/* The inetsw6 table contains everything that inet6_create needs to
76 * build a new socket.
77 */
78static struct list_head inetsw6[SOCK_MAX];
79static DEFINE_SPINLOCK(inetsw6_lock);
80
81struct ipv6_params ipv6_defaults = {
82	.disable_ipv6 = 0,
83	.autoconf = 1,
84};
85
86static int disable_ipv6_mod;
87
88module_param_named(disable, disable_ipv6_mod, int, 0444);
89MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
90
91module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
92MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
93
94module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
95MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
96
97bool ipv6_mod_enabled(void)
98{
99	return disable_ipv6_mod == 0;
100}
101EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
102
103static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
104{
105	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
106
107	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
108}
109
110void inet6_sock_destruct(struct sock *sk)
111{
112	inet6_cleanup_sock(sk);
113	inet_sock_destruct(sk);
114}
115EXPORT_SYMBOL_GPL(inet6_sock_destruct);
116
117static 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. */
133lookup_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	}
280out:
281	return err;
282out_rcu_unlock:
283	rcu_read_unlock();
284	goto out;
285}
286
287static 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;
439out:
440	if (flags & BIND_WITH_LOCK)
441		release_sock(sk);
442	return err;
443out_unlock:
444	rcu_read_unlock();
445	goto out;
446}
447
448/* bind for INET6 API */
449int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
450{
451	struct sock *sk = sock->sk;
452	const struct proto *prot;
453	int err = 0;
454
455	/* If the socket has its own bind function then use it. */
456	prot = READ_ONCE(sk->sk_prot);
457	if (prot->bind)
458		return prot->bind(sk, uaddr, addr_len);
459
460	if (addr_len < SIN6_LEN_RFC2133)
461		return -EINVAL;
462
463	/* BPF prog is run before any checks are done so that if the prog
464	 * changes context in a wrong way it will be caught.
465	 */
466	err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
467	if (err)
468		return err;
469
470	return __inet6_bind(sk, uaddr, addr_len, BIND_WITH_LOCK);
471}
472EXPORT_SYMBOL(inet6_bind);
473
474int inet6_release(struct socket *sock)
475{
476	struct sock *sk = sock->sk;
477
478	if (!sk)
479		return -EINVAL;
480
481	/* Free mc lists */
482	ipv6_sock_mc_close(sk);
483
484	/* Free ac lists */
485	ipv6_sock_ac_close(sk);
486
487	return inet_release(sock);
488}
489EXPORT_SYMBOL(inet6_release);
490
491void inet6_destroy_sock(struct sock *sk)
492{
493	struct ipv6_pinfo *np = inet6_sk(sk);
494	struct sk_buff *skb;
495	struct ipv6_txoptions *opt;
496
497	/* Release rx options */
498
499	skb = xchg(&np->pktoptions, NULL);
500	kfree_skb(skb);
501
502	skb = xchg(&np->rxpmtu, NULL);
503	kfree_skb(skb);
504
505	/* Free flowlabels */
506	fl6_free_socklist(sk);
507
508	/* Free tx options */
509
510	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
511	if (opt) {
512		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
513		txopt_put(opt);
514	}
515}
516EXPORT_SYMBOL_GPL(inet6_destroy_sock);
517
518void inet6_cleanup_sock(struct sock *sk)
519{
520	inet6_destroy_sock(sk);
521}
522EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
523
524/*
525 *	This does both peername and sockname.
526 */
527int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
528		  int peer)
529{
530	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
531	struct sock *sk = sock->sk;
532	struct inet_sock *inet = inet_sk(sk);
533	struct ipv6_pinfo *np = inet6_sk(sk);
534
535	sin->sin6_family = AF_INET6;
536	sin->sin6_flowinfo = 0;
537	sin->sin6_scope_id = 0;
538	if (peer) {
539		if (!inet->inet_dport)
540			return -ENOTCONN;
541		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
542		    peer == 1)
543			return -ENOTCONN;
544		sin->sin6_port = inet->inet_dport;
545		sin->sin6_addr = sk->sk_v6_daddr;
546		if (np->sndflow)
547			sin->sin6_flowinfo = np->flow_label;
548	} else {
549		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
550			sin->sin6_addr = np->saddr;
551		else
552			sin->sin6_addr = sk->sk_v6_rcv_saddr;
553		sin->sin6_port = inet->inet_sport;
554	}
555	if (cgroup_bpf_enabled)
556		BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin,
557					    peer ? BPF_CGROUP_INET6_GETPEERNAME :
558						   BPF_CGROUP_INET6_GETSOCKNAME,
559					    NULL);
560	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
561						 sk->sk_bound_dev_if);
562	return sizeof(*sin);
563}
564EXPORT_SYMBOL(inet6_getname);
565
566int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
567{
568	void __user *argp = (void __user *)arg;
569	struct sock *sk = sock->sk;
570	struct net *net = sock_net(sk);
571	const struct proto *prot;
572
573	switch (cmd) {
574	case SIOCADDRT:
575	case SIOCDELRT: {
576		struct in6_rtmsg rtmsg;
577
578		if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
579			return -EFAULT;
580		return ipv6_route_ioctl(net, cmd, &rtmsg);
581	}
582	case SIOCSIFADDR:
583		return addrconf_add_ifaddr(net, argp);
584	case SIOCDIFADDR:
585		return addrconf_del_ifaddr(net, argp);
586	case SIOCSIFDSTADDR:
587		return addrconf_set_dstaddr(net, argp);
588	default:
589		/* IPV6_ADDRFORM can change sk->sk_prot under us. */
590		prot = READ_ONCE(sk->sk_prot);
591		if (!prot->ioctl)
592			return -ENOIOCTLCMD;
593		return prot->ioctl(sk, cmd, arg);
594	}
595	/*NOTREACHED*/
596	return 0;
597}
598EXPORT_SYMBOL(inet6_ioctl);
599
600#ifdef CONFIG_COMPAT
601struct compat_in6_rtmsg {
602	struct in6_addr		rtmsg_dst;
603	struct in6_addr		rtmsg_src;
604	struct in6_addr		rtmsg_gateway;
605	u32			rtmsg_type;
606	u16			rtmsg_dst_len;
607	u16			rtmsg_src_len;
608	u32			rtmsg_metric;
609	u32			rtmsg_info;
610	u32			rtmsg_flags;
611	s32			rtmsg_ifindex;
612};
613
614static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
615		struct compat_in6_rtmsg __user *ur)
616{
617	struct in6_rtmsg rt;
618
619	if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
620			3 * sizeof(struct in6_addr)) ||
621	    get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
622	    get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
623	    get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
624	    get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
625	    get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
626	    get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
627	    get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
628		return -EFAULT;
629
630
631	return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
632}
633
634int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
635{
636	void __user *argp = compat_ptr(arg);
637	struct sock *sk = sock->sk;
638
639	switch (cmd) {
640	case SIOCADDRT:
641	case SIOCDELRT:
642		return inet6_compat_routing_ioctl(sk, cmd, argp);
643	default:
644		return -ENOIOCTLCMD;
645	}
646}
647EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
648#endif /* CONFIG_COMPAT */
649
650INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
651					    size_t));
652int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
653{
654	struct sock *sk = sock->sk;
655	const struct proto *prot;
656
657	if (unlikely(inet_send_prepare(sk)))
658		return -EAGAIN;
659
660	/* IPV6_ADDRFORM can change sk->sk_prot under us. */
661	prot = READ_ONCE(sk->sk_prot);
662	return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
663			       sk, msg, size);
664}
665
666INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
667					    size_t, int, int, int *));
668int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
669		  int flags)
670{
671	struct sock *sk = sock->sk;
672	const struct proto *prot;
673	int addr_len = 0;
674	int err;
675
676	if (likely(!(flags & MSG_ERRQUEUE)))
677		sock_rps_record_flow(sk);
678
679	/* IPV6_ADDRFORM can change sk->sk_prot under us. */
680	prot = READ_ONCE(sk->sk_prot);
681	err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
682			      sk, msg, size, flags & MSG_DONTWAIT,
683			      flags & ~MSG_DONTWAIT, &addr_len);
684	if (err >= 0)
685		msg->msg_namelen = addr_len;
686	return err;
687}
688
689const struct proto_ops inet6_stream_ops = {
690	.family		   = PF_INET6,
691	.owner		   = THIS_MODULE,
692	.release	   = inet6_release,
693	.bind		   = inet6_bind,
694	.connect	   = inet_stream_connect,	/* ok		*/
695	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
696	.accept		   = inet_accept,		/* ok		*/
697	.getname	   = inet6_getname,
698	.poll		   = tcp_poll,			/* ok		*/
699	.ioctl		   = inet6_ioctl,		/* must change  */
700	.gettstamp	   = sock_gettstamp,
701	.listen		   = inet_listen,		/* ok		*/
702	.shutdown	   = inet_shutdown,		/* ok		*/
703	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
704	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
705	.sendmsg	   = inet6_sendmsg,		/* retpoline's sake */
706	.recvmsg	   = inet6_recvmsg,		/* retpoline's sake */
707#ifdef CONFIG_MMU
708	.mmap		   = tcp_mmap,
709#endif
710	.sendpage	   = inet_sendpage,
711	.sendmsg_locked    = tcp_sendmsg_locked,
712	.sendpage_locked   = tcp_sendpage_locked,
713	.splice_read	   = tcp_splice_read,
714	.read_sock	   = tcp_read_sock,
715	.peek_len	   = tcp_peek_len,
716#ifdef CONFIG_COMPAT
717	.compat_ioctl	   = inet6_compat_ioctl,
718#endif
719	.set_rcvlowat	   = tcp_set_rcvlowat,
720};
721
722const struct proto_ops inet6_dgram_ops = {
723	.family		   = PF_INET6,
724	.owner		   = THIS_MODULE,
725	.release	   = inet6_release,
726	.bind		   = inet6_bind,
727	.connect	   = inet_dgram_connect,	/* ok		*/
728	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
729	.accept		   = sock_no_accept,		/* a do nothing	*/
730	.getname	   = inet6_getname,
731	.poll		   = udp_poll,			/* ok		*/
732	.ioctl		   = inet6_ioctl,		/* must change  */
733	.gettstamp	   = sock_gettstamp,
734	.listen		   = sock_no_listen,		/* ok		*/
735	.shutdown	   = inet_shutdown,		/* ok		*/
736	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
737	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
738	.sendmsg	   = inet6_sendmsg,		/* retpoline's sake */
739	.recvmsg	   = inet6_recvmsg,		/* retpoline's sake */
740	.mmap		   = sock_no_mmap,
741	.sendpage	   = sock_no_sendpage,
742	.set_peek_off	   = sk_set_peek_off,
743#ifdef CONFIG_COMPAT
744	.compat_ioctl	   = inet6_compat_ioctl,
745#endif
746};
747
748static const struct net_proto_family inet6_family_ops = {
749	.family = PF_INET6,
750	.create = inet6_create,
751	.owner	= THIS_MODULE,
752};
753
754int inet6_register_protosw(struct inet_protosw *p)
755{
756	struct list_head *lh;
757	struct inet_protosw *answer;
758	struct list_head *last_perm;
759	int protocol = p->protocol;
760	int ret;
761
762	spin_lock_bh(&inetsw6_lock);
763
764	ret = -EINVAL;
765	if (p->type >= SOCK_MAX)
766		goto out_illegal;
767
768	/* If we are trying to override a permanent protocol, bail. */
769	answer = NULL;
770	ret = -EPERM;
771	last_perm = &inetsw6[p->type];
772	list_for_each(lh, &inetsw6[p->type]) {
773		answer = list_entry(lh, struct inet_protosw, list);
774
775		/* Check only the non-wild match. */
776		if (INET_PROTOSW_PERMANENT & answer->flags) {
777			if (protocol == answer->protocol)
778				break;
779			last_perm = lh;
780		}
781
782		answer = NULL;
783	}
784	if (answer)
785		goto out_permanent;
786
787	/* Add the new entry after the last permanent entry if any, so that
788	 * the new entry does not override a permanent entry when matched with
789	 * a wild-card protocol. But it is allowed to override any existing
790	 * non-permanent entry.  This means that when we remove this entry, the
791	 * system automatically returns to the old behavior.
792	 */
793	list_add_rcu(&p->list, last_perm);
794	ret = 0;
795out:
796	spin_unlock_bh(&inetsw6_lock);
797	return ret;
798
799out_permanent:
800	pr_err("Attempt to override permanent protocol %d\n", protocol);
801	goto out;
802
803out_illegal:
804	pr_err("Ignoring attempt to register invalid socket type %d\n",
805	       p->type);
806	goto out;
807}
808EXPORT_SYMBOL(inet6_register_protosw);
809
810void
811inet6_unregister_protosw(struct inet_protosw *p)
812{
813	if (INET_PROTOSW_PERMANENT & p->flags) {
814		pr_err("Attempt to unregister permanent protocol %d\n",
815		       p->protocol);
816	} else {
817		spin_lock_bh(&inetsw6_lock);
818		list_del_rcu(&p->list);
819		spin_unlock_bh(&inetsw6_lock);
820
821		synchronize_net();
822	}
823}
824EXPORT_SYMBOL(inet6_unregister_protosw);
825
826int inet6_sk_rebuild_header(struct sock *sk)
827{
828	struct ipv6_pinfo *np = inet6_sk(sk);
829	struct dst_entry *dst;
830
831	dst = __sk_dst_check(sk, np->dst_cookie);
832
833	if (!dst) {
834		struct inet_sock *inet = inet_sk(sk);
835		struct in6_addr *final_p, final;
836		struct flowi6 fl6;
837
838		memset(&fl6, 0, sizeof(fl6));
839		fl6.flowi6_proto = sk->sk_protocol;
840		fl6.daddr = sk->sk_v6_daddr;
841		fl6.saddr = np->saddr;
842		fl6.flowlabel = np->flow_label;
843		fl6.flowi6_oif = sk->sk_bound_dev_if;
844		fl6.flowi6_mark = sk->sk_mark;
845		fl6.fl6_dport = inet->inet_dport;
846		fl6.fl6_sport = inet->inet_sport;
847		fl6.flowi6_uid = sk->sk_uid;
848		security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
849
850		rcu_read_lock();
851		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
852					 &final);
853		rcu_read_unlock();
854
855		dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
856		if (IS_ERR(dst)) {
857			sk->sk_route_caps = 0;
858			sk->sk_err_soft = -PTR_ERR(dst);
859			return PTR_ERR(dst);
860		}
861
862		ip6_dst_store(sk, dst, NULL, NULL);
863	}
864
865	return 0;
866}
867EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
868
869bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
870		       const struct inet6_skb_parm *opt)
871{
872	const struct ipv6_pinfo *np = inet6_sk(sk);
873
874	if (np->rxopt.all) {
875		if (((opt->flags & IP6SKB_HOPBYHOP) &&
876		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
877		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
878		     np->rxopt.bits.rxflow) ||
879		    (opt->srcrt && (np->rxopt.bits.srcrt ||
880		     np->rxopt.bits.osrcrt)) ||
881		    ((opt->dst1 || opt->dst0) &&
882		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
883			return true;
884	}
885	return false;
886}
887EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
888
889static struct packet_type ipv6_packet_type __read_mostly = {
890	.type = cpu_to_be16(ETH_P_IPV6),
891	.func = ipv6_rcv,
892	.list_func = ipv6_list_rcv,
893};
894
895static int __init ipv6_packet_init(void)
896{
897	dev_add_pack(&ipv6_packet_type);
898	return 0;
899}
900
901static void ipv6_packet_cleanup(void)
902{
903	dev_remove_pack(&ipv6_packet_type);
904}
905
906static int __net_init ipv6_init_mibs(struct net *net)
907{
908	int i;
909
910	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
911	if (!net->mib.udp_stats_in6)
912		return -ENOMEM;
913	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
914	if (!net->mib.udplite_stats_in6)
915		goto err_udplite_mib;
916	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
917	if (!net->mib.ipv6_statistics)
918		goto err_ip_mib;
919
920	for_each_possible_cpu(i) {
921		struct ipstats_mib *af_inet6_stats;
922		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
923		u64_stats_init(&af_inet6_stats->syncp);
924	}
925
926
927	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
928	if (!net->mib.icmpv6_statistics)
929		goto err_icmp_mib;
930	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
931						GFP_KERNEL);
932	if (!net->mib.icmpv6msg_statistics)
933		goto err_icmpmsg_mib;
934	return 0;
935
936err_icmpmsg_mib:
937	free_percpu(net->mib.icmpv6_statistics);
938err_icmp_mib:
939	free_percpu(net->mib.ipv6_statistics);
940err_ip_mib:
941	free_percpu(net->mib.udplite_stats_in6);
942err_udplite_mib:
943	free_percpu(net->mib.udp_stats_in6);
944	return -ENOMEM;
945}
946
947static void ipv6_cleanup_mibs(struct net *net)
948{
949	free_percpu(net->mib.udp_stats_in6);
950	free_percpu(net->mib.udplite_stats_in6);
951	free_percpu(net->mib.ipv6_statistics);
952	free_percpu(net->mib.icmpv6_statistics);
953	kfree(net->mib.icmpv6msg_statistics);
954}
955
956static int __net_init inet6_net_init(struct net *net)
957{
958	int err = 0;
959
960	net->ipv6.sysctl.bindv6only = 0;
961	net->ipv6.sysctl.icmpv6_time = 1*HZ;
962	net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
963	net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
964	net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
965
966	/* By default, rate limit error messages.
967	 * Except for pmtu discovery, it would break it.
968	 * proc_do_large_bitmap needs pointer to the bitmap.
969	 */
970	bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
971	bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
972	net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
973
974	net->ipv6.sysctl.flowlabel_consistency = 1;
975	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
976	net->ipv6.sysctl.idgen_retries = 3;
977	net->ipv6.sysctl.idgen_delay = 1 * HZ;
978	net->ipv6.sysctl.flowlabel_state_ranges = 0;
979	net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
980	net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
981	net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
982	net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
983	atomic_set(&net->ipv6.fib6_sernum, 1);
984
985	err = ipv6_init_mibs(net);
986	if (err)
987		return err;
988#ifdef CONFIG_PROC_FS
989	err = udp6_proc_init(net);
990	if (err)
991		goto out;
992	err = tcp6_proc_init(net);
993	if (err)
994		goto proc_tcp6_fail;
995	err = ac6_proc_init(net);
996	if (err)
997		goto proc_ac6_fail;
998#endif
999	return err;
1000
1001#ifdef CONFIG_PROC_FS
1002proc_ac6_fail:
1003	tcp6_proc_exit(net);
1004proc_tcp6_fail:
1005	udp6_proc_exit(net);
1006out:
1007	ipv6_cleanup_mibs(net);
1008	return err;
1009#endif
1010}
1011
1012static void __net_exit inet6_net_exit(struct net *net)
1013{
1014#ifdef CONFIG_PROC_FS
1015	udp6_proc_exit(net);
1016	tcp6_proc_exit(net);
1017	ac6_proc_exit(net);
1018#endif
1019	ipv6_cleanup_mibs(net);
1020}
1021
1022static struct pernet_operations inet6_net_ops = {
1023	.init = inet6_net_init,
1024	.exit = inet6_net_exit,
1025};
1026
1027static int ipv6_route_input(struct sk_buff *skb)
1028{
1029	ip6_route_input(skb);
1030	return skb_dst(skb)->error;
1031}
1032
1033static const struct ipv6_stub ipv6_stub_impl = {
1034	.ipv6_sock_mc_join = ipv6_sock_mc_join,
1035	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1036	.ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1037	.ipv6_route_input  = ipv6_route_input,
1038	.fib6_get_table	   = fib6_get_table,
1039	.fib6_table_lookup = fib6_table_lookup,
1040	.fib6_lookup       = fib6_lookup,
1041	.fib6_select_path  = fib6_select_path,
1042	.ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1043	.fib6_nh_init	   = fib6_nh_init,
1044	.fib6_nh_release   = fib6_nh_release,
1045	.fib6_nh_release_dsts = fib6_nh_release_dsts,
1046	.fib6_update_sernum = fib6_update_sernum_stub,
1047	.fib6_rt_update	   = fib6_rt_update,
1048	.ip6_del_rt	   = ip6_del_rt,
1049	.udpv6_encap_enable = udpv6_encap_enable,
1050	.ndisc_send_na = ndisc_send_na,
1051#if IS_ENABLED(CONFIG_XFRM)
1052	.xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1053	.xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1054	.xfrm6_rcv_encap = xfrm6_rcv_encap,
1055#endif
1056	.nd_tbl	= &nd_tbl,
1057	.ipv6_fragment = ip6_fragment,
1058};
1059
1060static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1061	.inet6_bind = __inet6_bind,
1062	.udp6_lib_lookup = __udp6_lib_lookup,
1063};
1064
1065static int __init inet6_init(void)
1066{
1067	struct list_head *r;
1068	int err = 0;
1069
1070	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1071
1072	/* Register the socket-side information for inet6_create.  */
1073	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1074		INIT_LIST_HEAD(r);
1075
1076	if (disable_ipv6_mod) {
1077		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1078		goto out;
1079	}
1080
1081	err = proto_register(&tcpv6_prot, 1);
1082	if (err)
1083		goto out;
1084
1085	err = proto_register(&udpv6_prot, 1);
1086	if (err)
1087		goto out_unregister_tcp_proto;
1088
1089	err = proto_register(&udplitev6_prot, 1);
1090	if (err)
1091		goto out_unregister_udp_proto;
1092
1093	err = proto_register(&rawv6_prot, 1);
1094	if (err)
1095		goto out_unregister_udplite_proto;
1096
1097	err = proto_register(&pingv6_prot, 1);
1098	if (err)
1099		goto out_unregister_raw_proto;
1100
1101	/* We MUST register RAW sockets before we create the ICMP6,
1102	 * IGMP6, or NDISC control sockets.
1103	 */
1104	err = rawv6_init();
1105	if (err)
1106		goto out_unregister_ping_proto;
1107
1108	/* Register the family here so that the init calls below will
1109	 * be able to create sockets. (?? is this dangerous ??)
1110	 */
1111	err = sock_register(&inet6_family_ops);
1112	if (err)
1113		goto out_sock_register_fail;
1114
1115	/*
1116	 *	ipngwg API draft makes clear that the correct semantics
1117	 *	for TCP and UDP is to consider one TCP and UDP instance
1118	 *	in a host available by both INET and INET6 APIs and
1119	 *	able to communicate via both network protocols.
1120	 */
1121
1122	err = register_pernet_subsys(&inet6_net_ops);
1123	if (err)
1124		goto register_pernet_fail;
1125	err = ip6_mr_init();
1126	if (err)
1127		goto ipmr_fail;
1128	err = icmpv6_init();
1129	if (err)
1130		goto icmp_fail;
1131	err = ndisc_init();
1132	if (err)
1133		goto ndisc_fail;
1134	err = igmp6_init();
1135	if (err)
1136		goto igmp_fail;
1137
1138	err = ipv6_netfilter_init();
1139	if (err)
1140		goto netfilter_fail;
1141	/* Create /proc/foo6 entries. */
1142#ifdef CONFIG_PROC_FS
1143	err = -ENOMEM;
1144	if (raw6_proc_init())
1145		goto proc_raw6_fail;
1146	if (udplite6_proc_init())
1147		goto proc_udplite6_fail;
1148	if (ipv6_misc_proc_init())
1149		goto proc_misc6_fail;
1150	if (if6_proc_init())
1151		goto proc_if6_fail;
1152#endif
1153	err = ip6_route_init();
1154	if (err)
1155		goto ip6_route_fail;
1156	err = ndisc_late_init();
1157	if (err)
1158		goto ndisc_late_fail;
1159	err = ip6_flowlabel_init();
1160	if (err)
1161		goto ip6_flowlabel_fail;
1162	err = ipv6_anycast_init();
1163	if (err)
1164		goto ipv6_anycast_fail;
1165	err = addrconf_init();
1166	if (err)
1167		goto addrconf_fail;
1168
1169	/* Init v6 extension headers. */
1170	err = ipv6_exthdrs_init();
1171	if (err)
1172		goto ipv6_exthdrs_fail;
1173
1174	err = ipv6_frag_init();
1175	if (err)
1176		goto ipv6_frag_fail;
1177
1178	/* Init v6 transport protocols. */
1179	err = udpv6_init();
1180	if (err)
1181		goto udpv6_fail;
1182
1183	err = udplitev6_init();
1184	if (err)
1185		goto udplitev6_fail;
1186
1187	err = udpv6_offload_init();
1188	if (err)
1189		goto udpv6_offload_fail;
1190
1191	err = tcpv6_init();
1192	if (err)
1193		goto tcpv6_fail;
1194
1195	err = ipv6_packet_init();
1196	if (err)
1197		goto ipv6_packet_fail;
1198
1199	err = pingv6_init();
1200	if (err)
1201		goto pingv6_fail;
1202
1203	err = calipso_init();
1204	if (err)
1205		goto calipso_fail;
1206
1207	err = seg6_init();
1208	if (err)
1209		goto seg6_fail;
1210
1211	err = rpl_init();
1212	if (err)
1213		goto rpl_fail;
1214
1215	err = igmp6_late_init();
1216	if (err)
1217		goto igmp6_late_err;
1218
1219#ifdef CONFIG_SYSCTL
1220	err = ipv6_sysctl_register();
1221	if (err)
1222		goto sysctl_fail;
1223#endif
1224
1225	/* ensure that ipv6 stubs are visible only after ipv6 is ready */
1226	wmb();
1227	ipv6_stub = &ipv6_stub_impl;
1228	ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1229out:
1230	return err;
1231
1232#ifdef CONFIG_SYSCTL
1233sysctl_fail:
1234	igmp6_late_cleanup();
1235#endif
1236igmp6_late_err:
1237	rpl_exit();
1238rpl_fail:
1239	seg6_exit();
1240seg6_fail:
1241	calipso_exit();
1242calipso_fail:
1243	pingv6_exit();
1244pingv6_fail:
1245	ipv6_packet_cleanup();
1246ipv6_packet_fail:
1247	tcpv6_exit();
1248tcpv6_fail:
1249	udpv6_offload_exit();
1250udpv6_offload_fail:
1251	udplitev6_exit();
1252udplitev6_fail:
1253	udpv6_exit();
1254udpv6_fail:
1255	ipv6_frag_exit();
1256ipv6_frag_fail:
1257	ipv6_exthdrs_exit();
1258ipv6_exthdrs_fail:
1259	addrconf_cleanup();
1260addrconf_fail:
1261	ipv6_anycast_cleanup();
1262ipv6_anycast_fail:
1263	ip6_flowlabel_cleanup();
1264ip6_flowlabel_fail:
1265	ndisc_late_cleanup();
1266ndisc_late_fail:
1267	ip6_route_cleanup();
1268ip6_route_fail:
1269#ifdef CONFIG_PROC_FS
1270	if6_proc_exit();
1271proc_if6_fail:
1272	ipv6_misc_proc_exit();
1273proc_misc6_fail:
1274	udplite6_proc_exit();
1275proc_udplite6_fail:
1276	raw6_proc_exit();
1277proc_raw6_fail:
1278#endif
1279	ipv6_netfilter_fini();
1280netfilter_fail:
1281	igmp6_cleanup();
1282igmp_fail:
1283	ndisc_cleanup();
1284ndisc_fail:
1285	icmpv6_cleanup();
1286icmp_fail:
1287	ip6_mr_cleanup();
1288ipmr_fail:
1289	unregister_pernet_subsys(&inet6_net_ops);
1290register_pernet_fail:
1291	sock_unregister(PF_INET6);
1292	rtnl_unregister_all(PF_INET6);
1293out_sock_register_fail:
1294	rawv6_exit();
1295out_unregister_ping_proto:
1296	proto_unregister(&pingv6_prot);
1297out_unregister_raw_proto:
1298	proto_unregister(&rawv6_prot);
1299out_unregister_udplite_proto:
1300	proto_unregister(&udplitev6_prot);
1301out_unregister_udp_proto:
1302	proto_unregister(&udpv6_prot);
1303out_unregister_tcp_proto:
1304	proto_unregister(&tcpv6_prot);
1305	goto out;
1306}
1307module_init(inet6_init);
1308
1309MODULE_ALIAS_NETPROTO(PF_INET6);
1310