Linux Audio

Check our new training course

Loading...
v4.6
 
   1/*
   2 *	PF_INET6 socket protocol family
   3 *	Linux INET6 implementation
   4 *
   5 *	Authors:
   6 *	Pedro Roque		<roque@di.fc.ul.pt>
   7 *
   8 *	Adapted from linux/net/ipv4/af_inet.c
   9 *
  10 *	Fixes:
  11 *	piggy, Karl Knutson	:	Socket protocol table
  12 *	Hideaki YOSHIFUJI	:	sin6_scope_id support
  13 *	Arnaldo Melo		:	check proc_net_create return, cleanups
  14 *
  15 *	This program is free software; you can redistribute it and/or
  16 *	modify it under the terms of the GNU General Public License
  17 *	as published by the Free Software Foundation; either version
  18 *	2 of the License, or (at your option) any later version.
  19 */
  20
  21#define pr_fmt(fmt) "IPv6: " fmt
  22
  23#include <linux/module.h>
  24#include <linux/capability.h>
  25#include <linux/errno.h>
  26#include <linux/types.h>
  27#include <linux/socket.h>
  28#include <linux/in.h>
  29#include <linux/kernel.h>
  30#include <linux/timer.h>
  31#include <linux/string.h>
  32#include <linux/sockios.h>
  33#include <linux/net.h>
  34#include <linux/fcntl.h>
  35#include <linux/mm.h>
  36#include <linux/interrupt.h>
  37#include <linux/proc_fs.h>
  38#include <linux/stat.h>
  39#include <linux/init.h>
  40#include <linux/slab.h>
  41
  42#include <linux/inet.h>
  43#include <linux/netdevice.h>
  44#include <linux/icmpv6.h>
  45#include <linux/netfilter_ipv6.h>
  46
  47#include <net/ip.h>
  48#include <net/ipv6.h>
  49#include <net/udp.h>
  50#include <net/udplite.h>
  51#include <net/tcp.h>
  52#include <net/ping.h>
  53#include <net/protocol.h>
  54#include <net/inet_common.h>
  55#include <net/route.h>
  56#include <net/transp_v6.h>
  57#include <net/ip6_route.h>
  58#include <net/addrconf.h>
 
  59#include <net/ndisc.h>
  60#ifdef CONFIG_IPV6_TUNNEL
  61#include <net/ip6_tunnel.h>
  62#endif
 
 
 
 
 
 
 
  63
  64#include <asm/uaccess.h>
  65#include <linux/mroute6.h>
  66
 
 
  67MODULE_AUTHOR("Cast of dozens");
  68MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  69MODULE_LICENSE("GPL");
  70
  71/* The inetsw6 table contains everything that inet6_create needs to
  72 * build a new socket.
  73 */
  74static struct list_head inetsw6[SOCK_MAX];
  75static DEFINE_SPINLOCK(inetsw6_lock);
  76
  77struct ipv6_params ipv6_defaults = {
  78	.disable_ipv6 = 0,
  79	.autoconf = 1,
  80};
  81
  82static int disable_ipv6_mod;
  83
  84module_param_named(disable, disable_ipv6_mod, int, 0444);
  85MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  86
  87module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  88MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  89
  90module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  91MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  92
 
 
 
 
 
 
  93static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  94{
  95	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  96
  97	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  98}
  99
 
 
 
 
 
 
 
 100static int inet6_create(struct net *net, struct socket *sock, int protocol,
 101			int kern)
 102{
 103	struct inet_sock *inet;
 104	struct ipv6_pinfo *np;
 105	struct sock *sk;
 106	struct inet_protosw *answer;
 107	struct proto *answer_prot;
 108	unsigned char answer_flags;
 109	int try_loading_module = 0;
 110	int err;
 111
 112	if (protocol < 0 || protocol >= IPPROTO_MAX)
 113		return -EINVAL;
 114
 115	/* Look for the requested type/protocol pair. */
 116lookup_protocol:
 117	err = -ESOCKTNOSUPPORT;
 118	rcu_read_lock();
 119	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 120
 121		err = 0;
 122		/* Check the non-wild match. */
 123		if (protocol == answer->protocol) {
 124			if (protocol != IPPROTO_IP)
 125				break;
 126		} else {
 127			/* Check for the two wild cases. */
 128			if (IPPROTO_IP == protocol) {
 129				protocol = answer->protocol;
 130				break;
 131			}
 132			if (IPPROTO_IP == answer->protocol)
 133				break;
 134		}
 135		err = -EPROTONOSUPPORT;
 136	}
 137
 138	if (err) {
 139		if (try_loading_module < 2) {
 140			rcu_read_unlock();
 141			/*
 142			 * Be more specific, e.g. net-pf-10-proto-132-type-1
 143			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 144			 */
 145			if (++try_loading_module == 1)
 146				request_module("net-pf-%d-proto-%d-type-%d",
 147						PF_INET6, protocol, sock->type);
 148			/*
 149			 * Fall back to generic, e.g. net-pf-10-proto-132
 150			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 151			 */
 152			else
 153				request_module("net-pf-%d-proto-%d",
 154						PF_INET6, protocol);
 155			goto lookup_protocol;
 156		} else
 157			goto out_rcu_unlock;
 158	}
 159
 160	err = -EPERM;
 161	if (sock->type == SOCK_RAW && !kern &&
 162	    !ns_capable(net->user_ns, CAP_NET_RAW))
 163		goto out_rcu_unlock;
 164
 165	sock->ops = answer->ops;
 166	answer_prot = answer->prot;
 167	answer_flags = answer->flags;
 168	rcu_read_unlock();
 169
 170	WARN_ON(!answer_prot->slab);
 171
 172	err = -ENOBUFS;
 173	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
 174	if (!sk)
 175		goto out;
 176
 177	sock_init_data(sock, sk);
 178
 179	err = 0;
 180	if (INET_PROTOSW_REUSE & answer_flags)
 181		sk->sk_reuse = SK_CAN_REUSE;
 182
 183	inet = inet_sk(sk);
 184	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 185
 186	if (SOCK_RAW == sock->type) {
 187		inet->inet_num = protocol;
 188		if (IPPROTO_RAW == protocol)
 189			inet->hdrincl = 1;
 190	}
 191
 192	sk->sk_destruct		= inet_sock_destruct;
 193	sk->sk_family		= PF_INET6;
 194	sk->sk_protocol		= protocol;
 195
 196	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
 197
 198	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 199	np->hop_limit	= -1;
 200	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
 201	np->mc_loop	= 1;
 
 202	np->pmtudisc	= IPV6_PMTUDISC_WANT;
 203	np->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
 204	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
 
 205
 206	/* Init the ipv4 part of the socket since we can have sockets
 207	 * using v6 API for ipv4.
 208	 */
 209	inet->uc_ttl	= -1;
 210
 211	inet->mc_loop	= 1;
 212	inet->mc_ttl	= 1;
 213	inet->mc_index	= 0;
 214	inet->mc_list	= NULL;
 215	inet->rcv_tos	= 0;
 216
 217	if (net->ipv4.sysctl_ip_no_pmtu_disc)
 218		inet->pmtudisc = IP_PMTUDISC_DONT;
 219	else
 220		inet->pmtudisc = IP_PMTUDISC_WANT;
 221	/*
 222	 * Increment only the relevant sk_prot->socks debug field, this changes
 223	 * the previous behaviour of incrementing both the equivalent to
 224	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 225	 *
 226	 * This allows better debug granularity as we'll know exactly how many
 227	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 228	 * transport protocol socks. -acme
 229	 */
 230	sk_refcnt_debug_inc(sk);
 231
 232	if (inet->inet_num) {
 233		/* It assumes that any protocol which allows
 234		 * the user to assign a number at socket
 235		 * creation time automatically shares.
 236		 */
 237		inet->inet_sport = htons(inet->inet_num);
 238		err = sk->sk_prot->hash(sk);
 239		if (err) {
 240			sk_common_release(sk);
 241			goto out;
 242		}
 243	}
 244	if (sk->sk_prot->init) {
 245		err = sk->sk_prot->init(sk);
 246		if (err) {
 247			sk_common_release(sk);
 248			goto out;
 249		}
 250	}
 
 
 
 
 
 
 
 
 251out:
 252	return err;
 253out_rcu_unlock:
 254	rcu_read_unlock();
 255	goto out;
 256}
 257
 258
 259/* bind for INET6 API */
 260int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 261{
 262	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 263	struct sock *sk = sock->sk;
 264	struct inet_sock *inet = inet_sk(sk);
 265	struct ipv6_pinfo *np = inet6_sk(sk);
 266	struct net *net = sock_net(sk);
 267	__be32 v4addr = 0;
 268	unsigned short snum;
 
 269	int addr_type = 0;
 270	int err = 0;
 271
 272	/* If the socket has its own bind function then use it. */
 273	if (sk->sk_prot->bind)
 274		return sk->sk_prot->bind(sk, uaddr, addr_len);
 275
 276	if (addr_len < SIN6_LEN_RFC2133)
 277		return -EINVAL;
 278
 279	if (addr->sin6_family != AF_INET6)
 280		return -EAFNOSUPPORT;
 281
 282	addr_type = ipv6_addr_type(&addr->sin6_addr);
 283	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 284		return -EINVAL;
 285
 286	snum = ntohs(addr->sin6_port);
 287	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 
 
 288		return -EACCES;
 289
 290	lock_sock(sk);
 
 291
 292	/* Check these errors (active socket, double bind). */
 293	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 294		err = -EINVAL;
 295		goto out;
 296	}
 297
 298	/* Check if the address belongs to the host. */
 299	if (addr_type == IPV6_ADDR_MAPPED) {
 
 300		int chk_addr_ret;
 301
 302		/* Binding to v4-mapped address on a v6-only socket
 303		 * makes no sense
 304		 */
 305		if (sk->sk_ipv6only) {
 306			err = -EINVAL;
 307			goto out;
 308		}
 309
 
 
 
 
 
 
 
 
 
 310		/* Reproduce AF_INET checks to make the bindings consistent */
 311		v4addr = addr->sin6_addr.s6_addr32[3];
 312		chk_addr_ret = inet_addr_type(net, v4addr);
 313		if (!net->ipv4.sysctl_ip_nonlocal_bind &&
 314		    !(inet->freebind || inet->transparent) &&
 315		    v4addr != htonl(INADDR_ANY) &&
 316		    chk_addr_ret != RTN_LOCAL &&
 317		    chk_addr_ret != RTN_MULTICAST &&
 318		    chk_addr_ret != RTN_BROADCAST) {
 319			err = -EADDRNOTAVAIL;
 320			goto out;
 321		}
 322	} else {
 323		if (addr_type != IPV6_ADDR_ANY) {
 324			struct net_device *dev = NULL;
 325
 326			rcu_read_lock();
 327			if (__ipv6_addr_needs_scope_id(addr_type)) {
 328				if (addr_len >= sizeof(struct sockaddr_in6) &&
 329				    addr->sin6_scope_id) {
 330					/* Override any existing binding, if another one
 331					 * is supplied by user.
 332					 */
 333					sk->sk_bound_dev_if = addr->sin6_scope_id;
 334				}
 335
 336				/* Binding to link-local address requires an interface */
 337				if (!sk->sk_bound_dev_if) {
 338					err = -EINVAL;
 339					goto out_unlock;
 340				}
 
 
 
 341				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 342				if (!dev) {
 343					err = -ENODEV;
 344					goto out_unlock;
 345				}
 346			}
 347
 348			/* ipv4 addr of the socket is invalid.  Only the
 349			 * unspecified and mapped address have a v4 equivalent.
 350			 */
 351			v4addr = LOOPBACK4_IPV6;
 352			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
 353				if (!net->ipv6.sysctl.ip_nonlocal_bind &&
 354				    !(inet->freebind || inet->transparent) &&
 355				    !ipv6_chk_addr(net, &addr->sin6_addr,
 356						   dev, 0)) {
 357					err = -EADDRNOTAVAIL;
 358					goto out_unlock;
 359				}
 360			}
 361			rcu_read_unlock();
 362		}
 363	}
 364
 365	inet->inet_rcv_saddr = v4addr;
 366	inet->inet_saddr = v4addr;
 367
 368	sk->sk_v6_rcv_saddr = addr->sin6_addr;
 369
 370	if (!(addr_type & IPV6_ADDR_MULTICAST))
 371		np->saddr = addr->sin6_addr;
 372
 
 
 
 
 373	/* Make sure we are allowed to bind here. */
 374	if ((snum || !inet->bind_address_no_port) &&
 375	    sk->sk_prot->get_port(sk, snum)) {
 376		inet_reset_saddr(sk);
 377		err = -EADDRINUSE;
 378		goto out;
 
 
 
 
 
 
 
 
 
 
 
 
 
 379	}
 380
 381	if (addr_type != IPV6_ADDR_ANY) {
 382		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 383		if (addr_type != IPV6_ADDR_MAPPED)
 384			sk->sk_ipv6only = 1;
 385	}
 386	if (snum)
 387		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 388	inet->inet_sport = htons(inet->inet_num);
 389	inet->inet_dport = 0;
 390	inet->inet_daddr = 0;
 391out:
 392	release_sock(sk);
 
 393	return err;
 394out_unlock:
 395	rcu_read_unlock();
 396	goto out;
 397}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 398EXPORT_SYMBOL(inet6_bind);
 399
 400int inet6_release(struct socket *sock)
 401{
 402	struct sock *sk = sock->sk;
 403
 404	if (!sk)
 405		return -EINVAL;
 406
 407	/* Free mc lists */
 408	ipv6_sock_mc_close(sk);
 409
 410	/* Free ac lists */
 411	ipv6_sock_ac_close(sk);
 412
 413	return inet_release(sock);
 414}
 415EXPORT_SYMBOL(inet6_release);
 416
 417void inet6_destroy_sock(struct sock *sk)
 418{
 419	struct ipv6_pinfo *np = inet6_sk(sk);
 420	struct sk_buff *skb;
 421	struct ipv6_txoptions *opt;
 422
 423	/* Release rx options */
 424
 425	skb = xchg(&np->pktoptions, NULL);
 426	if (skb)
 427		kfree_skb(skb);
 428
 429	skb = xchg(&np->rxpmtu, NULL);
 430	if (skb)
 431		kfree_skb(skb);
 432
 433	/* Free flowlabels */
 434	fl6_free_socklist(sk);
 435
 436	/* Free tx options */
 437
 438	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 439	if (opt) {
 440		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 441		txopt_put(opt);
 442	}
 443}
 444EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 445
 446/*
 447 *	This does both peername and sockname.
 448 */
 449
 450int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 451		 int *uaddr_len, int peer)
 452{
 453	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 454	struct sock *sk = sock->sk;
 455	struct inet_sock *inet = inet_sk(sk);
 456	struct ipv6_pinfo *np = inet6_sk(sk);
 457
 458	sin->sin6_family = AF_INET6;
 459	sin->sin6_flowinfo = 0;
 460	sin->sin6_scope_id = 0;
 
 461	if (peer) {
 462		if (!inet->inet_dport)
 463			return -ENOTCONN;
 464		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 465		    peer == 1)
 466			return -ENOTCONN;
 
 467		sin->sin6_port = inet->inet_dport;
 468		sin->sin6_addr = sk->sk_v6_daddr;
 469		if (np->sndflow)
 470			sin->sin6_flowinfo = np->flow_label;
 
 
 471	} else {
 472		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 473			sin->sin6_addr = np->saddr;
 474		else
 475			sin->sin6_addr = sk->sk_v6_rcv_saddr;
 476
 477		sin->sin6_port = inet->inet_sport;
 
 
 478	}
 479	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 480						 sk->sk_bound_dev_if);
 481	*uaddr_len = sizeof(*sin);
 482	return 0;
 483}
 484EXPORT_SYMBOL(inet6_getname);
 485
 486int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 487{
 
 488	struct sock *sk = sock->sk;
 489	struct net *net = sock_net(sk);
 
 490
 491	switch (cmd) {
 492	case SIOCGSTAMP:
 493		return sock_get_timestamp(sk, (struct timeval __user *)arg);
 494
 495	case SIOCGSTAMPNS:
 496		return sock_get_timestampns(sk, (struct timespec __user *)arg);
 497
 498	case SIOCADDRT:
 499	case SIOCDELRT:
 500
 501		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 502
 
 
 
 
 503	case SIOCSIFADDR:
 504		return addrconf_add_ifaddr(net, (void __user *) arg);
 505	case SIOCDIFADDR:
 506		return addrconf_del_ifaddr(net, (void __user *) arg);
 507	case SIOCSIFDSTADDR:
 508		return addrconf_set_dstaddr(net, (void __user *) arg);
 509	default:
 510		if (!sk->sk_prot->ioctl)
 
 
 511			return -ENOIOCTLCMD;
 512		return sk->sk_prot->ioctl(sk, cmd, arg);
 513	}
 514	/*NOTREACHED*/
 515	return 0;
 516}
 517EXPORT_SYMBOL(inet6_ioctl);
 518
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 519const struct proto_ops inet6_stream_ops = {
 520	.family		   = PF_INET6,
 521	.owner		   = THIS_MODULE,
 522	.release	   = inet6_release,
 523	.bind		   = inet6_bind,
 524	.connect	   = inet_stream_connect,	/* ok		*/
 525	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
 526	.accept		   = inet_accept,		/* ok		*/
 527	.getname	   = inet6_getname,
 528	.poll		   = tcp_poll,			/* ok		*/
 529	.ioctl		   = inet6_ioctl,		/* must change  */
 
 530	.listen		   = inet_listen,		/* ok		*/
 531	.shutdown	   = inet_shutdown,		/* ok		*/
 532	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
 533	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
 534	.sendmsg	   = inet_sendmsg,		/* ok		*/
 535	.recvmsg	   = inet_recvmsg,		/* ok		*/
 536	.mmap		   = sock_no_mmap,
 
 
 537	.sendpage	   = inet_sendpage,
 
 
 538	.splice_read	   = tcp_splice_read,
 
 
 
 539#ifdef CONFIG_COMPAT
 540	.compat_setsockopt = compat_sock_common_setsockopt,
 541	.compat_getsockopt = compat_sock_common_getsockopt,
 542#endif
 
 543};
 544
 545const struct proto_ops inet6_dgram_ops = {
 546	.family		   = PF_INET6,
 547	.owner		   = THIS_MODULE,
 548	.release	   = inet6_release,
 549	.bind		   = inet6_bind,
 550	.connect	   = inet_dgram_connect,	/* ok		*/
 551	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
 552	.accept		   = sock_no_accept,		/* a do nothing	*/
 553	.getname	   = inet6_getname,
 554	.poll		   = udp_poll,			/* ok		*/
 555	.ioctl		   = inet6_ioctl,		/* must change  */
 
 556	.listen		   = sock_no_listen,		/* ok		*/
 557	.shutdown	   = inet_shutdown,		/* ok		*/
 558	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
 559	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
 560	.sendmsg	   = inet_sendmsg,		/* ok		*/
 561	.recvmsg	   = inet_recvmsg,		/* ok		*/
 
 562	.mmap		   = sock_no_mmap,
 563	.sendpage	   = sock_no_sendpage,
 
 564#ifdef CONFIG_COMPAT
 565	.compat_setsockopt = compat_sock_common_setsockopt,
 566	.compat_getsockopt = compat_sock_common_getsockopt,
 567#endif
 568};
 569
 570static const struct net_proto_family inet6_family_ops = {
 571	.family = PF_INET6,
 572	.create = inet6_create,
 573	.owner	= THIS_MODULE,
 574};
 575
 576int inet6_register_protosw(struct inet_protosw *p)
 577{
 578	struct list_head *lh;
 579	struct inet_protosw *answer;
 580	struct list_head *last_perm;
 581	int protocol = p->protocol;
 582	int ret;
 583
 584	spin_lock_bh(&inetsw6_lock);
 585
 586	ret = -EINVAL;
 587	if (p->type >= SOCK_MAX)
 588		goto out_illegal;
 589
 590	/* If we are trying to override a permanent protocol, bail. */
 591	answer = NULL;
 592	ret = -EPERM;
 593	last_perm = &inetsw6[p->type];
 594	list_for_each(lh, &inetsw6[p->type]) {
 595		answer = list_entry(lh, struct inet_protosw, list);
 596
 597		/* Check only the non-wild match. */
 598		if (INET_PROTOSW_PERMANENT & answer->flags) {
 599			if (protocol == answer->protocol)
 600				break;
 601			last_perm = lh;
 602		}
 603
 604		answer = NULL;
 605	}
 606	if (answer)
 607		goto out_permanent;
 608
 609	/* Add the new entry after the last permanent entry if any, so that
 610	 * the new entry does not override a permanent entry when matched with
 611	 * a wild-card protocol. But it is allowed to override any existing
 612	 * non-permanent entry.  This means that when we remove this entry, the
 613	 * system automatically returns to the old behavior.
 614	 */
 615	list_add_rcu(&p->list, last_perm);
 616	ret = 0;
 617out:
 618	spin_unlock_bh(&inetsw6_lock);
 619	return ret;
 620
 621out_permanent:
 622	pr_err("Attempt to override permanent protocol %d\n", protocol);
 623	goto out;
 624
 625out_illegal:
 626	pr_err("Ignoring attempt to register invalid socket type %d\n",
 627	       p->type);
 628	goto out;
 629}
 630EXPORT_SYMBOL(inet6_register_protosw);
 631
 632void
 633inet6_unregister_protosw(struct inet_protosw *p)
 634{
 635	if (INET_PROTOSW_PERMANENT & p->flags) {
 636		pr_err("Attempt to unregister permanent protocol %d\n",
 637		       p->protocol);
 638	} else {
 639		spin_lock_bh(&inetsw6_lock);
 640		list_del_rcu(&p->list);
 641		spin_unlock_bh(&inetsw6_lock);
 642
 643		synchronize_net();
 644	}
 645}
 646EXPORT_SYMBOL(inet6_unregister_protosw);
 647
 648int inet6_sk_rebuild_header(struct sock *sk)
 649{
 650	struct ipv6_pinfo *np = inet6_sk(sk);
 651	struct dst_entry *dst;
 652
 653	dst = __sk_dst_check(sk, np->dst_cookie);
 654
 655	if (!dst) {
 656		struct inet_sock *inet = inet_sk(sk);
 657		struct in6_addr *final_p, final;
 658		struct flowi6 fl6;
 659
 660		memset(&fl6, 0, sizeof(fl6));
 661		fl6.flowi6_proto = sk->sk_protocol;
 662		fl6.daddr = sk->sk_v6_daddr;
 663		fl6.saddr = np->saddr;
 664		fl6.flowlabel = np->flow_label;
 665		fl6.flowi6_oif = sk->sk_bound_dev_if;
 666		fl6.flowi6_mark = sk->sk_mark;
 667		fl6.fl6_dport = inet->inet_dport;
 668		fl6.fl6_sport = inet->inet_sport;
 669		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 
 670
 671		rcu_read_lock();
 672		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 673					 &final);
 674		rcu_read_unlock();
 675
 676		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 677		if (IS_ERR(dst)) {
 678			sk->sk_route_caps = 0;
 679			sk->sk_err_soft = -PTR_ERR(dst);
 680			return PTR_ERR(dst);
 681		}
 682
 683		ip6_dst_store(sk, dst, NULL, NULL);
 684	}
 685
 686	return 0;
 687}
 688EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 689
 690bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 691		       const struct inet6_skb_parm *opt)
 692{
 693	const struct ipv6_pinfo *np = inet6_sk(sk);
 694
 695	if (np->rxopt.all) {
 696		if (((opt->flags & IP6SKB_HOPBYHOP) &&
 697		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 698		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 699		     np->rxopt.bits.rxflow) ||
 700		    (opt->srcrt && (np->rxopt.bits.srcrt ||
 701		     np->rxopt.bits.osrcrt)) ||
 702		    ((opt->dst1 || opt->dst0) &&
 703		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 704			return true;
 705	}
 706	return false;
 707}
 708EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 709
 710static struct packet_type ipv6_packet_type __read_mostly = {
 711	.type = cpu_to_be16(ETH_P_IPV6),
 712	.func = ipv6_rcv,
 
 713};
 714
 715static int __init ipv6_packet_init(void)
 716{
 717	dev_add_pack(&ipv6_packet_type);
 718	return 0;
 719}
 720
 721static void ipv6_packet_cleanup(void)
 722{
 723	dev_remove_pack(&ipv6_packet_type);
 724}
 725
 726static int __net_init ipv6_init_mibs(struct net *net)
 727{
 728	int i;
 729
 730	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 731	if (!net->mib.udp_stats_in6)
 732		return -ENOMEM;
 733	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 734	if (!net->mib.udplite_stats_in6)
 735		goto err_udplite_mib;
 736	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 737	if (!net->mib.ipv6_statistics)
 738		goto err_ip_mib;
 739
 740	for_each_possible_cpu(i) {
 741		struct ipstats_mib *af_inet6_stats;
 742		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 743		u64_stats_init(&af_inet6_stats->syncp);
 744	}
 745
 746
 747	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 748	if (!net->mib.icmpv6_statistics)
 749		goto err_icmp_mib;
 750	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 751						GFP_KERNEL);
 752	if (!net->mib.icmpv6msg_statistics)
 753		goto err_icmpmsg_mib;
 754	return 0;
 755
 756err_icmpmsg_mib:
 757	free_percpu(net->mib.icmpv6_statistics);
 758err_icmp_mib:
 759	free_percpu(net->mib.ipv6_statistics);
 760err_ip_mib:
 761	free_percpu(net->mib.udplite_stats_in6);
 762err_udplite_mib:
 763	free_percpu(net->mib.udp_stats_in6);
 764	return -ENOMEM;
 765}
 766
 767static void ipv6_cleanup_mibs(struct net *net)
 768{
 769	free_percpu(net->mib.udp_stats_in6);
 770	free_percpu(net->mib.udplite_stats_in6);
 771	free_percpu(net->mib.ipv6_statistics);
 772	free_percpu(net->mib.icmpv6_statistics);
 773	kfree(net->mib.icmpv6msg_statistics);
 774}
 775
 776static int __net_init inet6_net_init(struct net *net)
 777{
 778	int err = 0;
 779
 780	net->ipv6.sysctl.bindv6only = 0;
 781	net->ipv6.sysctl.icmpv6_time = 1*HZ;
 
 
 
 
 
 
 
 
 
 
 
 
 782	net->ipv6.sysctl.flowlabel_consistency = 1;
 783	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
 784	net->ipv6.sysctl.idgen_retries = 3;
 785	net->ipv6.sysctl.idgen_delay = 1 * HZ;
 786	net->ipv6.sysctl.flowlabel_state_ranges = 0;
 
 
 
 
 
 787	atomic_set(&net->ipv6.fib6_sernum, 1);
 788
 
 
 
 789	err = ipv6_init_mibs(net);
 790	if (err)
 791		return err;
 792#ifdef CONFIG_PROC_FS
 793	err = udp6_proc_init(net);
 794	if (err)
 795		goto out;
 796	err = tcp6_proc_init(net);
 797	if (err)
 798		goto proc_tcp6_fail;
 799	err = ac6_proc_init(net);
 800	if (err)
 801		goto proc_ac6_fail;
 802#endif
 803	return err;
 804
 805#ifdef CONFIG_PROC_FS
 806proc_ac6_fail:
 807	tcp6_proc_exit(net);
 808proc_tcp6_fail:
 809	udp6_proc_exit(net);
 810out:
 811	ipv6_cleanup_mibs(net);
 812	return err;
 813#endif
 814}
 815
 816static void __net_exit inet6_net_exit(struct net *net)
 817{
 818#ifdef CONFIG_PROC_FS
 819	udp6_proc_exit(net);
 820	tcp6_proc_exit(net);
 821	ac6_proc_exit(net);
 822#endif
 823	ipv6_cleanup_mibs(net);
 824}
 825
 826static struct pernet_operations inet6_net_ops = {
 827	.init = inet6_net_init,
 828	.exit = inet6_net_exit,
 829};
 830
 
 
 
 
 
 
 831static const struct ipv6_stub ipv6_stub_impl = {
 832	.ipv6_sock_mc_join = ipv6_sock_mc_join,
 833	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 834	.ipv6_dst_lookup = ip6_dst_lookup,
 
 
 
 
 
 
 
 
 
 
 
 
 835	.udpv6_encap_enable = udpv6_encap_enable,
 836	.ndisc_send_na = ndisc_send_na,
 
 
 
 
 
 837	.nd_tbl	= &nd_tbl,
 
 
 
 
 
 
 
 
 
 838};
 839
 840static int __init inet6_init(void)
 841{
 842	struct list_head *r;
 843	int err = 0;
 844
 845	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 846
 847	/* Register the socket-side information for inet6_create.  */
 848	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 849		INIT_LIST_HEAD(r);
 850
 
 
 851	if (disable_ipv6_mod) {
 852		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 853		goto out;
 854	}
 855
 856	err = proto_register(&tcpv6_prot, 1);
 857	if (err)
 858		goto out;
 859
 860	err = proto_register(&udpv6_prot, 1);
 861	if (err)
 862		goto out_unregister_tcp_proto;
 863
 864	err = proto_register(&udplitev6_prot, 1);
 865	if (err)
 866		goto out_unregister_udp_proto;
 867
 868	err = proto_register(&rawv6_prot, 1);
 869	if (err)
 870		goto out_unregister_udplite_proto;
 871
 872	err = proto_register(&pingv6_prot, 1);
 873	if (err)
 874		goto out_unregister_ping_proto;
 875
 876	/* We MUST register RAW sockets before we create the ICMP6,
 877	 * IGMP6, or NDISC control sockets.
 878	 */
 879	err = rawv6_init();
 880	if (err)
 881		goto out_unregister_raw_proto;
 882
 883	/* Register the family here so that the init calls below will
 884	 * be able to create sockets. (?? is this dangerous ??)
 885	 */
 886	err = sock_register(&inet6_family_ops);
 887	if (err)
 888		goto out_sock_register_fail;
 889
 890	/*
 891	 *	ipngwg API draft makes clear that the correct semantics
 892	 *	for TCP and UDP is to consider one TCP and UDP instance
 893	 *	in a host available by both INET and INET6 APIs and
 894	 *	able to communicate via both network protocols.
 895	 */
 896
 897	err = register_pernet_subsys(&inet6_net_ops);
 898	if (err)
 899		goto register_pernet_fail;
 900	err = icmpv6_init();
 901	if (err)
 902		goto icmp_fail;
 903	err = ip6_mr_init();
 904	if (err)
 905		goto ipmr_fail;
 
 
 
 906	err = ndisc_init();
 907	if (err)
 908		goto ndisc_fail;
 909	err = igmp6_init();
 910	if (err)
 911		goto igmp_fail;
 912
 913	ipv6_stub = &ipv6_stub_impl;
 914
 915	err = ipv6_netfilter_init();
 916	if (err)
 917		goto netfilter_fail;
 918	/* Create /proc/foo6 entries. */
 919#ifdef CONFIG_PROC_FS
 920	err = -ENOMEM;
 921	if (raw6_proc_init())
 922		goto proc_raw6_fail;
 923	if (udplite6_proc_init())
 924		goto proc_udplite6_fail;
 925	if (ipv6_misc_proc_init())
 926		goto proc_misc6_fail;
 927	if (if6_proc_init())
 928		goto proc_if6_fail;
 929#endif
 930	err = ip6_route_init();
 931	if (err)
 932		goto ip6_route_fail;
 933	err = ndisc_late_init();
 934	if (err)
 935		goto ndisc_late_fail;
 936	err = ip6_flowlabel_init();
 937	if (err)
 938		goto ip6_flowlabel_fail;
 
 
 
 939	err = addrconf_init();
 940	if (err)
 941		goto addrconf_fail;
 942
 943	/* Init v6 extension headers. */
 944	err = ipv6_exthdrs_init();
 945	if (err)
 946		goto ipv6_exthdrs_fail;
 947
 948	err = ipv6_frag_init();
 949	if (err)
 950		goto ipv6_frag_fail;
 951
 952	/* Init v6 transport protocols. */
 953	err = udpv6_init();
 954	if (err)
 955		goto udpv6_fail;
 956
 957	err = udplitev6_init();
 958	if (err)
 959		goto udplitev6_fail;
 960
 
 
 
 
 961	err = tcpv6_init();
 962	if (err)
 963		goto tcpv6_fail;
 964
 965	err = ipv6_packet_init();
 966	if (err)
 967		goto ipv6_packet_fail;
 968
 969	err = pingv6_init();
 970	if (err)
 971		goto pingv6_fail;
 972
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 973#ifdef CONFIG_SYSCTL
 974	err = ipv6_sysctl_register();
 975	if (err)
 976		goto sysctl_fail;
 977#endif
 
 
 
 
 
 978out:
 979	return err;
 980
 981#ifdef CONFIG_SYSCTL
 982sysctl_fail:
 983	pingv6_exit();
 984#endif
 
 
 
 
 
 
 
 
 
 
 985pingv6_fail:
 986	ipv6_packet_cleanup();
 987ipv6_packet_fail:
 988	tcpv6_exit();
 989tcpv6_fail:
 
 
 990	udplitev6_exit();
 991udplitev6_fail:
 992	udpv6_exit();
 993udpv6_fail:
 994	ipv6_frag_exit();
 995ipv6_frag_fail:
 996	ipv6_exthdrs_exit();
 997ipv6_exthdrs_fail:
 998	addrconf_cleanup();
 999addrconf_fail:
 
 
1000	ip6_flowlabel_cleanup();
1001ip6_flowlabel_fail:
1002	ndisc_late_cleanup();
1003ndisc_late_fail:
1004	ip6_route_cleanup();
1005ip6_route_fail:
1006#ifdef CONFIG_PROC_FS
1007	if6_proc_exit();
1008proc_if6_fail:
1009	ipv6_misc_proc_exit();
1010proc_misc6_fail:
1011	udplite6_proc_exit();
1012proc_udplite6_fail:
1013	raw6_proc_exit();
1014proc_raw6_fail:
1015#endif
1016	ipv6_netfilter_fini();
1017netfilter_fail:
1018	igmp6_cleanup();
1019igmp_fail:
1020	ndisc_cleanup();
1021ndisc_fail:
1022	ip6_mr_cleanup();
1023ipmr_fail:
1024	icmpv6_cleanup();
1025icmp_fail:
 
 
1026	unregister_pernet_subsys(&inet6_net_ops);
1027register_pernet_fail:
1028	sock_unregister(PF_INET6);
1029	rtnl_unregister_all(PF_INET6);
1030out_sock_register_fail:
1031	rawv6_exit();
1032out_unregister_ping_proto:
1033	proto_unregister(&pingv6_prot);
1034out_unregister_raw_proto:
1035	proto_unregister(&rawv6_prot);
1036out_unregister_udplite_proto:
1037	proto_unregister(&udplitev6_prot);
1038out_unregister_udp_proto:
1039	proto_unregister(&udpv6_prot);
1040out_unregister_tcp_proto:
1041	proto_unregister(&tcpv6_prot);
1042	goto out;
1043}
1044module_init(inet6_init);
1045
1046MODULE_ALIAS_NETPROTO(PF_INET6);
v6.2
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	PF_INET6 socket protocol family
   4 *	Linux INET6 implementation
   5 *
   6 *	Authors:
   7 *	Pedro Roque		<roque@di.fc.ul.pt>
   8 *
   9 *	Adapted from linux/net/ipv4/af_inet.c
  10 *
  11 *	Fixes:
  12 *	piggy, Karl Knutson	:	Socket protocol table
  13 *	Hideaki YOSHIFUJI	:	sin6_scope_id support
  14 *	Arnaldo Melo		:	check proc_net_create return, cleanups
 
 
 
 
 
  15 */
  16
  17#define pr_fmt(fmt) "IPv6: " fmt
  18
  19#include <linux/module.h>
  20#include <linux/capability.h>
  21#include <linux/errno.h>
  22#include <linux/types.h>
  23#include <linux/socket.h>
  24#include <linux/in.h>
  25#include <linux/kernel.h>
  26#include <linux/timer.h>
  27#include <linux/string.h>
  28#include <linux/sockios.h>
  29#include <linux/net.h>
  30#include <linux/fcntl.h>
  31#include <linux/mm.h>
  32#include <linux/interrupt.h>
  33#include <linux/proc_fs.h>
  34#include <linux/stat.h>
  35#include <linux/init.h>
  36#include <linux/slab.h>
  37
  38#include <linux/inet.h>
  39#include <linux/netdevice.h>
  40#include <linux/icmpv6.h>
  41#include <linux/netfilter_ipv6.h>
  42
  43#include <net/ip.h>
  44#include <net/ipv6.h>
  45#include <net/udp.h>
  46#include <net/udplite.h>
  47#include <net/tcp.h>
  48#include <net/ping.h>
  49#include <net/protocol.h>
  50#include <net/inet_common.h>
  51#include <net/route.h>
  52#include <net/transp_v6.h>
  53#include <net/ip6_route.h>
  54#include <net/addrconf.h>
  55#include <net/ipv6_stubs.h>
  56#include <net/ndisc.h>
  57#ifdef CONFIG_IPV6_TUNNEL
  58#include <net/ip6_tunnel.h>
  59#endif
  60#include <net/calipso.h>
  61#include <net/seg6.h>
  62#include <net/rpl.h>
  63#include <net/compat.h>
  64#include <net/xfrm.h>
  65#include <net/ioam6.h>
  66#include <net/rawv6.h>
  67
  68#include <linux/uaccess.h>
  69#include <linux/mroute6.h>
  70
  71#include "ip6_offload.h"
  72
  73MODULE_AUTHOR("Cast of dozens");
  74MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  75MODULE_LICENSE("GPL");
  76
  77/* The inetsw6 table contains everything that inet6_create needs to
  78 * build a new socket.
  79 */
  80static struct list_head inetsw6[SOCK_MAX];
  81static DEFINE_SPINLOCK(inetsw6_lock);
  82
  83struct ipv6_params ipv6_defaults = {
  84	.disable_ipv6 = 0,
  85	.autoconf = 1,
  86};
  87
  88static int disable_ipv6_mod;
  89
  90module_param_named(disable, disable_ipv6_mod, int, 0444);
  91MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  92
  93module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  94MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  95
  96module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  97MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  98
  99bool ipv6_mod_enabled(void)
 100{
 101	return disable_ipv6_mod == 0;
 102}
 103EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
 104
 105static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
 106{
 107	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
 108
 109	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
 110}
 111
 112void inet6_sock_destruct(struct sock *sk)
 113{
 114	inet6_cleanup_sock(sk);
 115	inet_sock_destruct(sk);
 116}
 117EXPORT_SYMBOL_GPL(inet6_sock_destruct);
 118
 119static int inet6_create(struct net *net, struct socket *sock, int protocol,
 120			int kern)
 121{
 122	struct inet_sock *inet;
 123	struct ipv6_pinfo *np;
 124	struct sock *sk;
 125	struct inet_protosw *answer;
 126	struct proto *answer_prot;
 127	unsigned char answer_flags;
 128	int try_loading_module = 0;
 129	int err;
 130
 131	if (protocol < 0 || protocol >= IPPROTO_MAX)
 132		return -EINVAL;
 133
 134	/* Look for the requested type/protocol pair. */
 135lookup_protocol:
 136	err = -ESOCKTNOSUPPORT;
 137	rcu_read_lock();
 138	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 139
 140		err = 0;
 141		/* Check the non-wild match. */
 142		if (protocol == answer->protocol) {
 143			if (protocol != IPPROTO_IP)
 144				break;
 145		} else {
 146			/* Check for the two wild cases. */
 147			if (IPPROTO_IP == protocol) {
 148				protocol = answer->protocol;
 149				break;
 150			}
 151			if (IPPROTO_IP == answer->protocol)
 152				break;
 153		}
 154		err = -EPROTONOSUPPORT;
 155	}
 156
 157	if (err) {
 158		if (try_loading_module < 2) {
 159			rcu_read_unlock();
 160			/*
 161			 * Be more specific, e.g. net-pf-10-proto-132-type-1
 162			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 163			 */
 164			if (++try_loading_module == 1)
 165				request_module("net-pf-%d-proto-%d-type-%d",
 166						PF_INET6, protocol, sock->type);
 167			/*
 168			 * Fall back to generic, e.g. net-pf-10-proto-132
 169			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 170			 */
 171			else
 172				request_module("net-pf-%d-proto-%d",
 173						PF_INET6, protocol);
 174			goto lookup_protocol;
 175		} else
 176			goto out_rcu_unlock;
 177	}
 178
 179	err = -EPERM;
 180	if (sock->type == SOCK_RAW && !kern &&
 181	    !ns_capable(net->user_ns, CAP_NET_RAW))
 182		goto out_rcu_unlock;
 183
 184	sock->ops = answer->ops;
 185	answer_prot = answer->prot;
 186	answer_flags = answer->flags;
 187	rcu_read_unlock();
 188
 189	WARN_ON(!answer_prot->slab);
 190
 191	err = -ENOBUFS;
 192	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
 193	if (!sk)
 194		goto out;
 195
 196	sock_init_data(sock, sk);
 197
 198	err = 0;
 199	if (INET_PROTOSW_REUSE & answer_flags)
 200		sk->sk_reuse = SK_CAN_REUSE;
 201
 202	inet = inet_sk(sk);
 203	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 204
 205	if (SOCK_RAW == sock->type) {
 206		inet->inet_num = protocol;
 207		if (IPPROTO_RAW == protocol)
 208			inet->hdrincl = 1;
 209	}
 210
 211	sk->sk_destruct		= inet6_sock_destruct;
 212	sk->sk_family		= PF_INET6;
 213	sk->sk_protocol		= protocol;
 214
 215	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
 216
 217	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 218	np->hop_limit	= -1;
 219	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
 220	np->mc_loop	= 1;
 221	np->mc_all	= 1;
 222	np->pmtudisc	= IPV6_PMTUDISC_WANT;
 223	np->repflow	= net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
 224	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
 225	sk->sk_txrehash = READ_ONCE(net->core.sysctl_txrehash);
 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	RCU_INIT_POINTER(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 (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
 309	    snum && inet_port_requires_bind_service(net, snum) &&
 310	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 311		return -EACCES;
 312
 313	if (flags & BIND_WITH_LOCK)
 314		lock_sock(sk);
 315
 316	/* Check these errors (active socket, double bind). */
 317	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 318		err = -EINVAL;
 319		goto out;
 320	}
 321
 322	/* Check if the address belongs to the host. */
 323	if (addr_type == IPV6_ADDR_MAPPED) {
 324		struct net_device *dev = NULL;
 325		int chk_addr_ret;
 326
 327		/* Binding to v4-mapped address on a v6-only socket
 328		 * makes no sense
 329		 */
 330		if (ipv6_only_sock(sk)) {
 331			err = -EINVAL;
 332			goto out;
 333		}
 334
 335		rcu_read_lock();
 336		if (sk->sk_bound_dev_if) {
 337			dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 338			if (!dev) {
 339				err = -ENODEV;
 340				goto out_unlock;
 341			}
 342		}
 343
 344		/* Reproduce AF_INET checks to make the bindings consistent */
 345		v4addr = addr->sin6_addr.s6_addr32[3];
 346		chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
 347		rcu_read_unlock();
 348
 349		if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
 350						 chk_addr_ret)) {
 
 
 351			err = -EADDRNOTAVAIL;
 352			goto out;
 353		}
 354	} else {
 355		if (addr_type != IPV6_ADDR_ANY) {
 356			struct net_device *dev = NULL;
 357
 358			rcu_read_lock();
 359			if (__ipv6_addr_needs_scope_id(addr_type)) {
 360				if (addr_len >= sizeof(struct sockaddr_in6) &&
 361				    addr->sin6_scope_id) {
 362					/* Override any existing binding, if another one
 363					 * is supplied by user.
 364					 */
 365					sk->sk_bound_dev_if = addr->sin6_scope_id;
 366				}
 367
 368				/* Binding to link-local address requires an interface */
 369				if (!sk->sk_bound_dev_if) {
 370					err = -EINVAL;
 371					goto out_unlock;
 372				}
 373			}
 374
 375			if (sk->sk_bound_dev_if) {
 376				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 377				if (!dev) {
 378					err = -ENODEV;
 379					goto out_unlock;
 380				}
 381			}
 382
 383			/* ipv4 addr of the socket is invalid.  Only the
 384			 * unspecified and mapped address have a v4 equivalent.
 385			 */
 386			v4addr = LOOPBACK4_IPV6;
 387			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
 388				if (!ipv6_can_nonlocal_bind(net, inet) &&
 
 389				    !ipv6_chk_addr(net, &addr->sin6_addr,
 390						   dev, 0)) {
 391					err = -EADDRNOTAVAIL;
 392					goto out_unlock;
 393				}
 394			}
 395			rcu_read_unlock();
 396		}
 397	}
 398
 399	inet->inet_rcv_saddr = v4addr;
 400	inet->inet_saddr = v4addr;
 401
 402	sk->sk_v6_rcv_saddr = addr->sin6_addr;
 403
 404	if (!(addr_type & IPV6_ADDR_MULTICAST))
 405		np->saddr = addr->sin6_addr;
 406
 407	saved_ipv6only = sk->sk_ipv6only;
 408	if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
 409		sk->sk_ipv6only = 1;
 410
 411	/* Make sure we are allowed to bind here. */
 412	if (snum || !(inet->bind_address_no_port ||
 413		      (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
 414		err = sk->sk_prot->get_port(sk, snum);
 415		if (err) {
 416			sk->sk_ipv6only = saved_ipv6only;
 417			inet_reset_saddr(sk);
 418			goto out;
 419		}
 420		if (!(flags & BIND_FROM_BPF)) {
 421			err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
 422			if (err) {
 423				sk->sk_ipv6only = saved_ipv6only;
 424				inet_reset_saddr(sk);
 425				if (sk->sk_prot->put_port)
 426					sk->sk_prot->put_port(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	u32 flags = BIND_WITH_LOCK;
 453	const struct proto *prot;
 454	int err = 0;
 455
 456	/* IPV6_ADDRFORM can change sk->sk_prot under us. */
 457	prot = READ_ONCE(sk->sk_prot);
 458	/* If the socket has its own bind function then use it. */
 459	if (prot->bind)
 460		return prot->bind(sk, uaddr, addr_len);
 461
 462	if (addr_len < SIN6_LEN_RFC2133)
 463		return -EINVAL;
 464
 465	/* BPF prog is run before any checks are done so that if the prog
 466	 * changes context in a wrong way it will be caught.
 467	 */
 468	err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr,
 469						 CGROUP_INET6_BIND, &flags);
 470	if (err)
 471		return err;
 472
 473	return __inet6_bind(sk, uaddr, addr_len, flags);
 474}
 475EXPORT_SYMBOL(inet6_bind);
 476
 477int inet6_release(struct socket *sock)
 478{
 479	struct sock *sk = sock->sk;
 480
 481	if (!sk)
 482		return -EINVAL;
 483
 484	/* Free mc lists */
 485	ipv6_sock_mc_close(sk);
 486
 487	/* Free ac lists */
 488	ipv6_sock_ac_close(sk);
 489
 490	return inet_release(sock);
 491}
 492EXPORT_SYMBOL(inet6_release);
 493
 494void inet6_cleanup_sock(struct sock *sk)
 495{
 496	struct ipv6_pinfo *np = inet6_sk(sk);
 497	struct sk_buff *skb;
 498	struct ipv6_txoptions *opt;
 499
 500	/* Release rx options */
 501
 502	skb = xchg(&np->pktoptions, NULL);
 503	kfree_skb(skb);
 
 504
 505	skb = xchg(&np->rxpmtu, NULL);
 506	kfree_skb(skb);
 
 507
 508	/* Free flowlabels */
 509	fl6_free_socklist(sk);
 510
 511	/* Free tx options */
 512
 513	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 514	if (opt) {
 515		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 516		txopt_put(opt);
 517	}
 518}
 519EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
 520
 521/*
 522 *	This does both peername and sockname.
 523 */
 
 524int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 525		  int peer)
 526{
 527	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 528	struct sock *sk = sock->sk;
 529	struct inet_sock *inet = inet_sk(sk);
 530	struct ipv6_pinfo *np = inet6_sk(sk);
 531
 532	sin->sin6_family = AF_INET6;
 533	sin->sin6_flowinfo = 0;
 534	sin->sin6_scope_id = 0;
 535	lock_sock(sk);
 536	if (peer) {
 537		if (!inet->inet_dport ||
 538		    (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 539		    peer == 1)) {
 540			release_sock(sk);
 541			return -ENOTCONN;
 542		}
 543		sin->sin6_port = inet->inet_dport;
 544		sin->sin6_addr = sk->sk_v6_daddr;
 545		if (np->sndflow)
 546			sin->sin6_flowinfo = np->flow_label;
 547		BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
 548				       CGROUP_INET6_GETPEERNAME);
 549	} else {
 550		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 551			sin->sin6_addr = np->saddr;
 552		else
 553			sin->sin6_addr = sk->sk_v6_rcv_saddr;
 
 554		sin->sin6_port = inet->inet_sport;
 555		BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
 556				       CGROUP_INET6_GETSOCKNAME);
 557	}
 558	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 559						 sk->sk_bound_dev_if);
 560	release_sock(sk);
 561	return sizeof(*sin);
 562}
 563EXPORT_SYMBOL(inet6_getname);
 564
 565int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 566{
 567	void __user *argp = (void __user *)arg;
 568	struct sock *sk = sock->sk;
 569	struct net *net = sock_net(sk);
 570	const struct proto *prot;
 571
 572	switch (cmd) {
 
 
 
 
 
 
 573	case SIOCADDRT:
 574	case SIOCDELRT: {
 575		struct in6_rtmsg rtmsg;
 
 576
 577		if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
 578			return -EFAULT;
 579		return ipv6_route_ioctl(net, cmd, &rtmsg);
 580	}
 581	case SIOCSIFADDR:
 582		return addrconf_add_ifaddr(net, argp);
 583	case SIOCDIFADDR:
 584		return addrconf_del_ifaddr(net, argp);
 585	case SIOCSIFDSTADDR:
 586		return addrconf_set_dstaddr(net, argp);
 587	default:
 588		/* IPV6_ADDRFORM can change sk->sk_prot under us. */
 589		prot = READ_ONCE(sk->sk_prot);
 590		if (!prot->ioctl)
 591			return -ENOIOCTLCMD;
 592		return prot->ioctl(sk, cmd, arg);
 593	}
 594	/*NOTREACHED*/
 595	return 0;
 596}
 597EXPORT_SYMBOL(inet6_ioctl);
 598
 599#ifdef CONFIG_COMPAT
 600struct compat_in6_rtmsg {
 601	struct in6_addr		rtmsg_dst;
 602	struct in6_addr		rtmsg_src;
 603	struct in6_addr		rtmsg_gateway;
 604	u32			rtmsg_type;
 605	u16			rtmsg_dst_len;
 606	u16			rtmsg_src_len;
 607	u32			rtmsg_metric;
 608	u32			rtmsg_info;
 609	u32			rtmsg_flags;
 610	s32			rtmsg_ifindex;
 611};
 612
 613static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
 614		struct compat_in6_rtmsg __user *ur)
 615{
 616	struct in6_rtmsg rt;
 617
 618	if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
 619			3 * sizeof(struct in6_addr)) ||
 620	    get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
 621	    get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
 622	    get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
 623	    get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
 624	    get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
 625	    get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
 626	    get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
 627		return -EFAULT;
 628
 629
 630	return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
 631}
 632
 633int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 634{
 635	void __user *argp = compat_ptr(arg);
 636	struct sock *sk = sock->sk;
 637
 638	switch (cmd) {
 639	case SIOCADDRT:
 640	case SIOCDELRT:
 641		return inet6_compat_routing_ioctl(sk, cmd, argp);
 642	default:
 643		return -ENOIOCTLCMD;
 644	}
 645}
 646EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
 647#endif /* CONFIG_COMPAT */
 648
 649INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
 650					    size_t));
 651int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 652{
 653	struct sock *sk = sock->sk;
 654	const struct proto *prot;
 655
 656	if (unlikely(inet_send_prepare(sk)))
 657		return -EAGAIN;
 658
 659	/* IPV6_ADDRFORM can change sk->sk_prot under us. */
 660	prot = READ_ONCE(sk->sk_prot);
 661	return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
 662			       sk, msg, size);
 663}
 664
 665INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
 666					    size_t, int, int *));
 667int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
 668		  int flags)
 669{
 670	struct sock *sk = sock->sk;
 671	const struct proto *prot;
 672	int addr_len = 0;
 673	int err;
 674
 675	if (likely(!(flags & MSG_ERRQUEUE)))
 676		sock_rps_record_flow(sk);
 677
 678	/* IPV6_ADDRFORM can change sk->sk_prot under us. */
 679	prot = READ_ONCE(sk->sk_prot);
 680	err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
 681			      sk, msg, size, flags, &addr_len);
 682	if (err >= 0)
 683		msg->msg_namelen = addr_len;
 684	return err;
 685}
 686
 687const struct proto_ops inet6_stream_ops = {
 688	.family		   = PF_INET6,
 689	.owner		   = THIS_MODULE,
 690	.release	   = inet6_release,
 691	.bind		   = inet6_bind,
 692	.connect	   = inet_stream_connect,	/* ok		*/
 693	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
 694	.accept		   = inet_accept,		/* ok		*/
 695	.getname	   = inet6_getname,
 696	.poll		   = tcp_poll,			/* ok		*/
 697	.ioctl		   = inet6_ioctl,		/* must change  */
 698	.gettstamp	   = sock_gettstamp,
 699	.listen		   = inet_listen,		/* ok		*/
 700	.shutdown	   = inet_shutdown,		/* ok		*/
 701	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
 702	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
 703	.sendmsg	   = inet6_sendmsg,		/* retpoline's sake */
 704	.recvmsg	   = inet6_recvmsg,		/* retpoline's sake */
 705#ifdef CONFIG_MMU
 706	.mmap		   = tcp_mmap,
 707#endif
 708	.sendpage	   = inet_sendpage,
 709	.sendmsg_locked    = tcp_sendmsg_locked,
 710	.sendpage_locked   = tcp_sendpage_locked,
 711	.splice_read	   = tcp_splice_read,
 712	.read_sock	   = tcp_read_sock,
 713	.read_skb	   = tcp_read_skb,
 714	.peek_len	   = tcp_peek_len,
 715#ifdef CONFIG_COMPAT
 716	.compat_ioctl	   = inet6_compat_ioctl,
 
 717#endif
 718	.set_rcvlowat	   = tcp_set_rcvlowat,
 719};
 720
 721const struct proto_ops inet6_dgram_ops = {
 722	.family		   = PF_INET6,
 723	.owner		   = THIS_MODULE,
 724	.release	   = inet6_release,
 725	.bind		   = inet6_bind,
 726	.connect	   = inet_dgram_connect,	/* ok		*/
 727	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
 728	.accept		   = sock_no_accept,		/* a do nothing	*/
 729	.getname	   = inet6_getname,
 730	.poll		   = udp_poll,			/* ok		*/
 731	.ioctl		   = inet6_ioctl,		/* must change  */
 732	.gettstamp	   = sock_gettstamp,
 733	.listen		   = sock_no_listen,		/* ok		*/
 734	.shutdown	   = inet_shutdown,		/* ok		*/
 735	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
 736	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
 737	.sendmsg	   = inet6_sendmsg,		/* retpoline's sake */
 738	.recvmsg	   = inet6_recvmsg,		/* retpoline's sake */
 739	.read_skb	   = udp_read_skb,
 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	net->ipv6.sysctl.fib_notify_on_flag_change = 0;
 984	atomic_set(&net->ipv6.fib6_sernum, 1);
 985
 986	net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
 987	net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
 988
 989	err = ipv6_init_mibs(net);
 990	if (err)
 991		return err;
 992#ifdef CONFIG_PROC_FS
 993	err = udp6_proc_init(net);
 994	if (err)
 995		goto out;
 996	err = tcp6_proc_init(net);
 997	if (err)
 998		goto proc_tcp6_fail;
 999	err = ac6_proc_init(net);
1000	if (err)
1001		goto proc_ac6_fail;
1002#endif
1003	return err;
1004
1005#ifdef CONFIG_PROC_FS
1006proc_ac6_fail:
1007	tcp6_proc_exit(net);
1008proc_tcp6_fail:
1009	udp6_proc_exit(net);
1010out:
1011	ipv6_cleanup_mibs(net);
1012	return err;
1013#endif
1014}
1015
1016static void __net_exit inet6_net_exit(struct net *net)
1017{
1018#ifdef CONFIG_PROC_FS
1019	udp6_proc_exit(net);
1020	tcp6_proc_exit(net);
1021	ac6_proc_exit(net);
1022#endif
1023	ipv6_cleanup_mibs(net);
1024}
1025
1026static struct pernet_operations inet6_net_ops = {
1027	.init = inet6_net_init,
1028	.exit = inet6_net_exit,
1029};
1030
1031static int ipv6_route_input(struct sk_buff *skb)
1032{
1033	ip6_route_input(skb);
1034	return skb_dst(skb)->error;
1035}
1036
1037static const struct ipv6_stub ipv6_stub_impl = {
1038	.ipv6_sock_mc_join = ipv6_sock_mc_join,
1039	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1040	.ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1041	.ipv6_route_input  = ipv6_route_input,
1042	.fib6_get_table	   = fib6_get_table,
1043	.fib6_table_lookup = fib6_table_lookup,
1044	.fib6_lookup       = fib6_lookup,
1045	.fib6_select_path  = fib6_select_path,
1046	.ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1047	.fib6_nh_init	   = fib6_nh_init,
1048	.fib6_nh_release   = fib6_nh_release,
1049	.fib6_nh_release_dsts = fib6_nh_release_dsts,
1050	.fib6_update_sernum = fib6_update_sernum_stub,
1051	.fib6_rt_update	   = fib6_rt_update,
1052	.ip6_del_rt	   = ip6_del_rt,
1053	.udpv6_encap_enable = udpv6_encap_enable,
1054	.ndisc_send_na = ndisc_send_na,
1055#if IS_ENABLED(CONFIG_XFRM)
1056	.xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1057	.xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1058	.xfrm6_rcv_encap = xfrm6_rcv_encap,
1059#endif
1060	.nd_tbl	= &nd_tbl,
1061	.ipv6_fragment = ip6_fragment,
1062	.ipv6_dev_find = ipv6_dev_find,
1063};
1064
1065static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1066	.inet6_bind = __inet6_bind,
1067	.udp6_lib_lookup = __udp6_lib_lookup,
1068	.ipv6_setsockopt = do_ipv6_setsockopt,
1069	.ipv6_getsockopt = do_ipv6_getsockopt,
1070};
1071
1072static int __init inet6_init(void)
1073{
1074	struct list_head *r;
1075	int err = 0;
1076
1077	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1078
1079	/* Register the socket-side information for inet6_create.  */
1080	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1081		INIT_LIST_HEAD(r);
1082
1083	raw_hashinfo_init(&raw_v6_hashinfo);
1084
1085	if (disable_ipv6_mod) {
1086		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1087		goto out;
1088	}
1089
1090	err = proto_register(&tcpv6_prot, 1);
1091	if (err)
1092		goto out;
1093
1094	err = proto_register(&udpv6_prot, 1);
1095	if (err)
1096		goto out_unregister_tcp_proto;
1097
1098	err = proto_register(&udplitev6_prot, 1);
1099	if (err)
1100		goto out_unregister_udp_proto;
1101
1102	err = proto_register(&rawv6_prot, 1);
1103	if (err)
1104		goto out_unregister_udplite_proto;
1105
1106	err = proto_register(&pingv6_prot, 1);
1107	if (err)
1108		goto out_unregister_raw_proto;
1109
1110	/* We MUST register RAW sockets before we create the ICMP6,
1111	 * IGMP6, or NDISC control sockets.
1112	 */
1113	err = rawv6_init();
1114	if (err)
1115		goto out_unregister_ping_proto;
1116
1117	/* Register the family here so that the init calls below will
1118	 * be able to create sockets. (?? is this dangerous ??)
1119	 */
1120	err = sock_register(&inet6_family_ops);
1121	if (err)
1122		goto out_sock_register_fail;
1123
1124	/*
1125	 *	ipngwg API draft makes clear that the correct semantics
1126	 *	for TCP and UDP is to consider one TCP and UDP instance
1127	 *	in a host available by both INET and INET6 APIs and
1128	 *	able to communicate via both network protocols.
1129	 */
1130
1131	err = register_pernet_subsys(&inet6_net_ops);
1132	if (err)
1133		goto register_pernet_fail;
 
 
 
1134	err = ip6_mr_init();
1135	if (err)
1136		goto ipmr_fail;
1137	err = icmpv6_init();
1138	if (err)
1139		goto icmp_fail;
1140	err = ndisc_init();
1141	if (err)
1142		goto ndisc_fail;
1143	err = igmp6_init();
1144	if (err)
1145		goto igmp_fail;
1146
 
 
1147	err = ipv6_netfilter_init();
1148	if (err)
1149		goto netfilter_fail;
1150	/* Create /proc/foo6 entries. */
1151#ifdef CONFIG_PROC_FS
1152	err = -ENOMEM;
1153	if (raw6_proc_init())
1154		goto proc_raw6_fail;
1155	if (udplite6_proc_init())
1156		goto proc_udplite6_fail;
1157	if (ipv6_misc_proc_init())
1158		goto proc_misc6_fail;
1159	if (if6_proc_init())
1160		goto proc_if6_fail;
1161#endif
1162	err = ip6_route_init();
1163	if (err)
1164		goto ip6_route_fail;
1165	err = ndisc_late_init();
1166	if (err)
1167		goto ndisc_late_fail;
1168	err = ip6_flowlabel_init();
1169	if (err)
1170		goto ip6_flowlabel_fail;
1171	err = ipv6_anycast_init();
1172	if (err)
1173		goto ipv6_anycast_fail;
1174	err = addrconf_init();
1175	if (err)
1176		goto addrconf_fail;
1177
1178	/* Init v6 extension headers. */
1179	err = ipv6_exthdrs_init();
1180	if (err)
1181		goto ipv6_exthdrs_fail;
1182
1183	err = ipv6_frag_init();
1184	if (err)
1185		goto ipv6_frag_fail;
1186
1187	/* Init v6 transport protocols. */
1188	err = udpv6_init();
1189	if (err)
1190		goto udpv6_fail;
1191
1192	err = udplitev6_init();
1193	if (err)
1194		goto udplitev6_fail;
1195
1196	err = udpv6_offload_init();
1197	if (err)
1198		goto udpv6_offload_fail;
1199
1200	err = tcpv6_init();
1201	if (err)
1202		goto tcpv6_fail;
1203
1204	err = ipv6_packet_init();
1205	if (err)
1206		goto ipv6_packet_fail;
1207
1208	err = pingv6_init();
1209	if (err)
1210		goto pingv6_fail;
1211
1212	err = calipso_init();
1213	if (err)
1214		goto calipso_fail;
1215
1216	err = seg6_init();
1217	if (err)
1218		goto seg6_fail;
1219
1220	err = rpl_init();
1221	if (err)
1222		goto rpl_fail;
1223
1224	err = ioam6_init();
1225	if (err)
1226		goto ioam6_fail;
1227
1228	err = igmp6_late_init();
1229	if (err)
1230		goto igmp6_late_err;
1231
1232#ifdef CONFIG_SYSCTL
1233	err = ipv6_sysctl_register();
1234	if (err)
1235		goto sysctl_fail;
1236#endif
1237
1238	/* ensure that ipv6 stubs are visible only after ipv6 is ready */
1239	wmb();
1240	ipv6_stub = &ipv6_stub_impl;
1241	ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1242out:
1243	return err;
1244
1245#ifdef CONFIG_SYSCTL
1246sysctl_fail:
1247	igmp6_late_cleanup();
1248#endif
1249igmp6_late_err:
1250	ioam6_exit();
1251ioam6_fail:
1252	rpl_exit();
1253rpl_fail:
1254	seg6_exit();
1255seg6_fail:
1256	calipso_exit();
1257calipso_fail:
1258	pingv6_exit();
1259pingv6_fail:
1260	ipv6_packet_cleanup();
1261ipv6_packet_fail:
1262	tcpv6_exit();
1263tcpv6_fail:
1264	udpv6_offload_exit();
1265udpv6_offload_fail:
1266	udplitev6_exit();
1267udplitev6_fail:
1268	udpv6_exit();
1269udpv6_fail:
1270	ipv6_frag_exit();
1271ipv6_frag_fail:
1272	ipv6_exthdrs_exit();
1273ipv6_exthdrs_fail:
1274	addrconf_cleanup();
1275addrconf_fail:
1276	ipv6_anycast_cleanup();
1277ipv6_anycast_fail:
1278	ip6_flowlabel_cleanup();
1279ip6_flowlabel_fail:
1280	ndisc_late_cleanup();
1281ndisc_late_fail:
1282	ip6_route_cleanup();
1283ip6_route_fail:
1284#ifdef CONFIG_PROC_FS
1285	if6_proc_exit();
1286proc_if6_fail:
1287	ipv6_misc_proc_exit();
1288proc_misc6_fail:
1289	udplite6_proc_exit();
1290proc_udplite6_fail:
1291	raw6_proc_exit();
1292proc_raw6_fail:
1293#endif
1294	ipv6_netfilter_fini();
1295netfilter_fail:
1296	igmp6_cleanup();
1297igmp_fail:
1298	ndisc_cleanup();
1299ndisc_fail:
 
 
1300	icmpv6_cleanup();
1301icmp_fail:
1302	ip6_mr_cleanup();
1303ipmr_fail:
1304	unregister_pernet_subsys(&inet6_net_ops);
1305register_pernet_fail:
1306	sock_unregister(PF_INET6);
1307	rtnl_unregister_all(PF_INET6);
1308out_sock_register_fail:
1309	rawv6_exit();
1310out_unregister_ping_proto:
1311	proto_unregister(&pingv6_prot);
1312out_unregister_raw_proto:
1313	proto_unregister(&rawv6_prot);
1314out_unregister_udplite_proto:
1315	proto_unregister(&udplitev6_prot);
1316out_unregister_udp_proto:
1317	proto_unregister(&udpv6_prot);
1318out_unregister_tcp_proto:
1319	proto_unregister(&tcpv6_prot);
1320	goto out;
1321}
1322module_init(inet6_init);
1323
1324MODULE_ALIAS_NETPROTO(PF_INET6);