Linux Audio

Check our new training course

In-person Linux kernel drivers training

Jun 16-20, 2025
Register
Loading...
v3.1
   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
  22#include <linux/module.h>
  23#include <linux/capability.h>
  24#include <linux/errno.h>
  25#include <linux/types.h>
  26#include <linux/socket.h>
  27#include <linux/in.h>
  28#include <linux/kernel.h>
  29#include <linux/timer.h>
  30#include <linux/string.h>
  31#include <linux/sockios.h>
  32#include <linux/net.h>
  33#include <linux/fcntl.h>
  34#include <linux/mm.h>
  35#include <linux/interrupt.h>
  36#include <linux/proc_fs.h>
  37#include <linux/stat.h>
  38#include <linux/init.h>
  39#include <linux/slab.h>
  40
  41#include <linux/inet.h>
  42#include <linux/netdevice.h>
  43#include <linux/icmpv6.h>
  44#include <linux/netfilter_ipv6.h>
  45
  46#include <net/ip.h>
  47#include <net/ipv6.h>
  48#include <net/udp.h>
  49#include <net/udplite.h>
  50#include <net/tcp.h>
  51#include <net/ipip.h>
  52#include <net/protocol.h>
  53#include <net/inet_common.h>
  54#include <net/route.h>
  55#include <net/transp_v6.h>
  56#include <net/ip6_route.h>
  57#include <net/addrconf.h>
 
  58#ifdef CONFIG_IPV6_TUNNEL
  59#include <net/ip6_tunnel.h>
  60#endif
  61
  62#include <asm/uaccess.h>
  63#include <asm/system.h>
  64#include <linux/mroute6.h>
  65
  66MODULE_AUTHOR("Cast of dozens");
  67MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  68MODULE_LICENSE("GPL");
  69
  70/* The inetsw6 table contains everything that inet6_create needs to
  71 * build a new socket.
  72 */
  73static struct list_head inetsw6[SOCK_MAX];
  74static DEFINE_SPINLOCK(inetsw6_lock);
  75
  76struct ipv6_params ipv6_defaults = {
  77	.disable_ipv6 = 0,
  78	.autoconf = 1,
  79};
  80
  81static int disable_ipv6_mod = 0;
  82
  83module_param_named(disable, disable_ipv6_mod, int, 0444);
  84MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  85
  86module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  87MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  88
  89module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  90MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  91
  92static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  93{
  94	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  95
  96	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  97}
  98
  99static int inet6_create(struct net *net, struct socket *sock, int protocol,
 100			int kern)
 101{
 102	struct inet_sock *inet;
 103	struct ipv6_pinfo *np;
 104	struct sock *sk;
 105	struct inet_protosw *answer;
 106	struct proto *answer_prot;
 107	unsigned char answer_flags;
 108	char answer_no_check;
 109	int try_loading_module = 0;
 110	int err;
 111
 112	if (sock->type != SOCK_RAW &&
 113	    sock->type != SOCK_DGRAM &&
 114	    !inet_ehash_secret)
 115		build_ehash_secret();
 116
 117	/* Look for the requested type/protocol pair. */
 118lookup_protocol:
 119	err = -ESOCKTNOSUPPORT;
 120	rcu_read_lock();
 121	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 122
 123		err = 0;
 124		/* Check the non-wild match. */
 125		if (protocol == answer->protocol) {
 126			if (protocol != IPPROTO_IP)
 127				break;
 128		} else {
 129			/* Check for the two wild cases. */
 130			if (IPPROTO_IP == protocol) {
 131				protocol = answer->protocol;
 132				break;
 133			}
 134			if (IPPROTO_IP == answer->protocol)
 135				break;
 136		}
 137		err = -EPROTONOSUPPORT;
 138	}
 139
 140	if (err) {
 141		if (try_loading_module < 2) {
 142			rcu_read_unlock();
 143			/*
 144			 * Be more specific, e.g. net-pf-10-proto-132-type-1
 145			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 146			 */
 147			if (++try_loading_module == 1)
 148				request_module("net-pf-%d-proto-%d-type-%d",
 149						PF_INET6, protocol, sock->type);
 150			/*
 151			 * Fall back to generic, e.g. net-pf-10-proto-132
 152			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 153			 */
 154			else
 155				request_module("net-pf-%d-proto-%d",
 156						PF_INET6, protocol);
 157			goto lookup_protocol;
 158		} else
 159			goto out_rcu_unlock;
 160	}
 161
 162	err = -EPERM;
 163	if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
 
 164		goto out_rcu_unlock;
 165
 166	sock->ops = answer->ops;
 167	answer_prot = answer->prot;
 168	answer_no_check = answer->no_check;
 169	answer_flags = answer->flags;
 170	rcu_read_unlock();
 171
 172	WARN_ON(answer_prot->slab == NULL);
 173
 174	err = -ENOBUFS;
 175	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 176	if (sk == NULL)
 177		goto out;
 178
 179	sock_init_data(sock, sk);
 180
 181	err = 0;
 182	sk->sk_no_check = answer_no_check;
 183	if (INET_PROTOSW_REUSE & answer_flags)
 184		sk->sk_reuse = 1;
 185
 186	inet = inet_sk(sk);
 187	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 188
 189	if (SOCK_RAW == sock->type) {
 190		inet->inet_num = protocol;
 191		if (IPPROTO_RAW == protocol)
 192			inet->hdrincl = 1;
 193	}
 194
 195	sk->sk_destruct		= inet_sock_destruct;
 196	sk->sk_family		= PF_INET6;
 197	sk->sk_protocol		= protocol;
 198
 199	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
 200
 201	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 202	np->hop_limit	= -1;
 203	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
 204	np->mc_loop	= 1;
 205	np->pmtudisc	= IPV6_PMTUDISC_WANT;
 206	np->ipv6only	= net->ipv6.sysctl.bindv6only;
 
 207
 208	/* Init the ipv4 part of the socket since we can have sockets
 209	 * using v6 API for ipv4.
 210	 */
 211	inet->uc_ttl	= -1;
 212
 213	inet->mc_loop	= 1;
 214	inet->mc_ttl	= 1;
 215	inet->mc_index	= 0;
 216	inet->mc_list	= NULL;
 
 217
 218	if (ipv4_config.no_pmtu_disc)
 219		inet->pmtudisc = IP_PMTUDISC_DONT;
 220	else
 221		inet->pmtudisc = IP_PMTUDISC_WANT;
 222	/*
 223	 * Increment only the relevant sk_prot->socks debug field, this changes
 224	 * the previous behaviour of incrementing both the equivalent to
 225	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 226	 *
 227	 * This allows better debug granularity as we'll know exactly how many
 228	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 229	 * transport protocol socks. -acme
 230	 */
 231	sk_refcnt_debug_inc(sk);
 232
 233	if (inet->inet_num) {
 234		/* It assumes that any protocol which allows
 235		 * the user to assign a number at socket
 236		 * creation time automatically shares.
 237		 */
 238		inet->inet_sport = htons(inet->inet_num);
 239		sk->sk_prot->hash(sk);
 
 
 
 
 240	}
 241	if (sk->sk_prot->init) {
 242		err = sk->sk_prot->init(sk);
 243		if (err) {
 244			sk_common_release(sk);
 245			goto out;
 246		}
 247	}
 248out:
 249	return err;
 250out_rcu_unlock:
 251	rcu_read_unlock();
 252	goto out;
 253}
 254
 255
 256/* bind for INET6 API */
 257int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 258{
 259	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 260	struct sock *sk = sock->sk;
 261	struct inet_sock *inet = inet_sk(sk);
 262	struct ipv6_pinfo *np = inet6_sk(sk);
 263	struct net *net = sock_net(sk);
 264	__be32 v4addr = 0;
 265	unsigned short snum;
 266	int addr_type = 0;
 267	int err = 0;
 268
 269	/* If the socket has its own bind function then use it. */
 270	if (sk->sk_prot->bind)
 271		return sk->sk_prot->bind(sk, uaddr, addr_len);
 272
 273	if (addr_len < SIN6_LEN_RFC2133)
 274		return -EINVAL;
 275
 276	if (addr->sin6_family != AF_INET6)
 277		return -EAFNOSUPPORT;
 278
 279	addr_type = ipv6_addr_type(&addr->sin6_addr);
 280	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 281		return -EINVAL;
 282
 283	snum = ntohs(addr->sin6_port);
 284	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 285		return -EACCES;
 286
 287	lock_sock(sk);
 288
 289	/* Check these errors (active socket, double bind). */
 290	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 291		err = -EINVAL;
 292		goto out;
 293	}
 294
 295	/* Check if the address belongs to the host. */
 296	if (addr_type == IPV6_ADDR_MAPPED) {
 297		int chk_addr_ret;
 298
 299		/* Binding to v4-mapped address on a v6-only socket
 300		 * makes no sense
 301		 */
 302		if (np->ipv6only) {
 303			err = -EINVAL;
 304			goto out;
 305		}
 306
 307		/* Reproduce AF_INET checks to make the bindings consistent */
 308		v4addr = addr->sin6_addr.s6_addr32[3];
 309		chk_addr_ret = inet_addr_type(net, v4addr);
 310		if (!sysctl_ip_nonlocal_bind &&
 311		    !(inet->freebind || inet->transparent) &&
 312		    v4addr != htonl(INADDR_ANY) &&
 313		    chk_addr_ret != RTN_LOCAL &&
 314		    chk_addr_ret != RTN_MULTICAST &&
 315		    chk_addr_ret != RTN_BROADCAST) {
 316			err = -EADDRNOTAVAIL;
 317			goto out;
 318		}
 319	} else {
 320		if (addr_type != IPV6_ADDR_ANY) {
 321			struct net_device *dev = NULL;
 322
 323			rcu_read_lock();
 324			if (addr_type & IPV6_ADDR_LINKLOCAL) {
 325				if (addr_len >= sizeof(struct sockaddr_in6) &&
 326				    addr->sin6_scope_id) {
 327					/* Override any existing binding, if another one
 328					 * is supplied by user.
 329					 */
 330					sk->sk_bound_dev_if = addr->sin6_scope_id;
 331				}
 332
 333				/* Binding to link-local address requires an interface */
 334				if (!sk->sk_bound_dev_if) {
 335					err = -EINVAL;
 336					goto out_unlock;
 337				}
 338				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 339				if (!dev) {
 340					err = -ENODEV;
 341					goto out_unlock;
 342				}
 343			}
 344
 345			/* ipv4 addr of the socket is invalid.  Only the
 346			 * unspecified and mapped address have a v4 equivalent.
 347			 */
 348			v4addr = LOOPBACK4_IPV6;
 349			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
 350				if (!inet->transparent &&
 
 351				    !ipv6_chk_addr(net, &addr->sin6_addr,
 352						   dev, 0)) {
 353					err = -EADDRNOTAVAIL;
 354					goto out_unlock;
 355				}
 356			}
 357			rcu_read_unlock();
 358		}
 359	}
 360
 361	inet->inet_rcv_saddr = v4addr;
 362	inet->inet_saddr = v4addr;
 363
 364	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 365
 366	if (!(addr_type & IPV6_ADDR_MULTICAST))
 367		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 368
 369	/* Make sure we are allowed to bind here. */
 370	if (sk->sk_prot->get_port(sk, snum)) {
 
 371		inet_reset_saddr(sk);
 372		err = -EADDRINUSE;
 373		goto out;
 374	}
 375
 376	if (addr_type != IPV6_ADDR_ANY) {
 377		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 378		if (addr_type != IPV6_ADDR_MAPPED)
 379			np->ipv6only = 1;
 380	}
 381	if (snum)
 382		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 383	inet->inet_sport = htons(inet->inet_num);
 384	inet->inet_dport = 0;
 385	inet->inet_daddr = 0;
 386out:
 387	release_sock(sk);
 388	return err;
 389out_unlock:
 390	rcu_read_unlock();
 391	goto out;
 392}
 393
 394EXPORT_SYMBOL(inet6_bind);
 395
 396int inet6_release(struct socket *sock)
 397{
 398	struct sock *sk = sock->sk;
 399
 400	if (sk == NULL)
 401		return -EINVAL;
 402
 403	/* Free mc lists */
 404	ipv6_sock_mc_close(sk);
 405
 406	/* Free ac lists */
 407	ipv6_sock_ac_close(sk);
 408
 409	return inet_release(sock);
 410}
 411
 412EXPORT_SYMBOL(inet6_release);
 413
 414void inet6_destroy_sock(struct sock *sk)
 415{
 416	struct ipv6_pinfo *np = inet6_sk(sk);
 417	struct sk_buff *skb;
 418	struct ipv6_txoptions *opt;
 419
 420	/* Release rx options */
 421
 422	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 
 423		kfree_skb(skb);
 424
 425	if ((skb = xchg(&np->rxpmtu, NULL)) != NULL)
 
 426		kfree_skb(skb);
 427
 428	/* Free flowlabels */
 429	fl6_free_socklist(sk);
 430
 431	/* Free tx options */
 432
 433	if ((opt = xchg(&np->opt, NULL)) != NULL)
 434		sock_kfree_s(sk, opt, opt->tot_len);
 
 
 
 435}
 436
 437EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 438
 439/*
 440 *	This does both peername and sockname.
 441 */
 442
 443int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 444		 int *uaddr_len, int peer)
 445{
 446	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 447	struct sock *sk = sock->sk;
 448	struct inet_sock *inet = inet_sk(sk);
 449	struct ipv6_pinfo *np = inet6_sk(sk);
 450
 451	sin->sin6_family = AF_INET6;
 452	sin->sin6_flowinfo = 0;
 453	sin->sin6_scope_id = 0;
 454	if (peer) {
 455		if (!inet->inet_dport)
 456			return -ENOTCONN;
 457		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 458		    peer == 1)
 459			return -ENOTCONN;
 460		sin->sin6_port = inet->inet_dport;
 461		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 462		if (np->sndflow)
 463			sin->sin6_flowinfo = np->flow_label;
 464	} else {
 465		if (ipv6_addr_any(&np->rcv_saddr))
 466			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 467		else
 468			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 469
 470		sin->sin6_port = inet->inet_sport;
 471	}
 472	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 473		sin->sin6_scope_id = sk->sk_bound_dev_if;
 474	*uaddr_len = sizeof(*sin);
 475	return 0;
 476}
 477
 478EXPORT_SYMBOL(inet6_getname);
 479
 480int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 481{
 482	struct sock *sk = sock->sk;
 483	struct net *net = sock_net(sk);
 484
 485	switch(cmd)
 486	{
 487	case SIOCGSTAMP:
 488		return sock_get_timestamp(sk, (struct timeval __user *)arg);
 489
 490	case SIOCGSTAMPNS:
 491		return sock_get_timestampns(sk, (struct timespec __user *)arg);
 492
 493	case SIOCADDRT:
 494	case SIOCDELRT:
 495
 496		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 497
 498	case SIOCSIFADDR:
 499		return addrconf_add_ifaddr(net, (void __user *) arg);
 500	case SIOCDIFADDR:
 501		return addrconf_del_ifaddr(net, (void __user *) arg);
 502	case SIOCSIFDSTADDR:
 503		return addrconf_set_dstaddr(net, (void __user *) arg);
 504	default:
 505		if (!sk->sk_prot->ioctl)
 506			return -ENOIOCTLCMD;
 507		return sk->sk_prot->ioctl(sk, cmd, arg);
 508	}
 509	/*NOTREACHED*/
 510	return 0;
 511}
 512
 513EXPORT_SYMBOL(inet6_ioctl);
 514
 515const struct proto_ops inet6_stream_ops = {
 516	.family		   = PF_INET6,
 517	.owner		   = THIS_MODULE,
 518	.release	   = inet6_release,
 519	.bind		   = inet6_bind,
 520	.connect	   = inet_stream_connect,	/* ok		*/
 521	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
 522	.accept		   = inet_accept,		/* ok		*/
 523	.getname	   = inet6_getname,
 524	.poll		   = tcp_poll,			/* ok		*/
 525	.ioctl		   = inet6_ioctl,		/* must change  */
 526	.listen		   = inet_listen,		/* ok		*/
 527	.shutdown	   = inet_shutdown,		/* ok		*/
 528	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
 529	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
 530	.sendmsg	   = inet_sendmsg,		/* ok		*/
 531	.recvmsg	   = inet_recvmsg,		/* ok		*/
 532	.mmap		   = sock_no_mmap,
 533	.sendpage	   = inet_sendpage,
 534	.splice_read	   = tcp_splice_read,
 535#ifdef CONFIG_COMPAT
 536	.compat_setsockopt = compat_sock_common_setsockopt,
 537	.compat_getsockopt = compat_sock_common_getsockopt,
 538#endif
 539};
 540
 541const struct proto_ops inet6_dgram_ops = {
 542	.family		   = PF_INET6,
 543	.owner		   = THIS_MODULE,
 544	.release	   = inet6_release,
 545	.bind		   = inet6_bind,
 546	.connect	   = inet_dgram_connect,	/* ok		*/
 547	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
 548	.accept		   = sock_no_accept,		/* a do nothing	*/
 549	.getname	   = inet6_getname,
 550	.poll		   = udp_poll,			/* ok		*/
 551	.ioctl		   = inet6_ioctl,		/* must change  */
 552	.listen		   = sock_no_listen,		/* ok		*/
 553	.shutdown	   = inet_shutdown,		/* ok		*/
 554	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
 555	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
 556	.sendmsg	   = inet_sendmsg,		/* ok		*/
 557	.recvmsg	   = inet_recvmsg,		/* ok		*/
 558	.mmap		   = sock_no_mmap,
 559	.sendpage	   = sock_no_sendpage,
 560#ifdef CONFIG_COMPAT
 561	.compat_setsockopt = compat_sock_common_setsockopt,
 562	.compat_getsockopt = compat_sock_common_getsockopt,
 563#endif
 564};
 565
 566static const struct net_proto_family inet6_family_ops = {
 567	.family = PF_INET6,
 568	.create = inet6_create,
 569	.owner	= THIS_MODULE,
 570};
 571
 572int inet6_register_protosw(struct inet_protosw *p)
 573{
 574	struct list_head *lh;
 575	struct inet_protosw *answer;
 576	struct list_head *last_perm;
 577	int protocol = p->protocol;
 578	int ret;
 579
 580	spin_lock_bh(&inetsw6_lock);
 581
 582	ret = -EINVAL;
 583	if (p->type >= SOCK_MAX)
 584		goto out_illegal;
 585
 586	/* If we are trying to override a permanent protocol, bail. */
 587	answer = NULL;
 588	ret = -EPERM;
 589	last_perm = &inetsw6[p->type];
 590	list_for_each(lh, &inetsw6[p->type]) {
 591		answer = list_entry(lh, struct inet_protosw, list);
 592
 593		/* Check only the non-wild match. */
 594		if (INET_PROTOSW_PERMANENT & answer->flags) {
 595			if (protocol == answer->protocol)
 596				break;
 597			last_perm = lh;
 598		}
 599
 600		answer = NULL;
 601	}
 602	if (answer)
 603		goto out_permanent;
 604
 605	/* Add the new entry after the last permanent entry if any, so that
 606	 * the new entry does not override a permanent entry when matched with
 607	 * a wild-card protocol. But it is allowed to override any existing
 608	 * non-permanent entry.  This means that when we remove this entry, the
 609	 * system automatically returns to the old behavior.
 610	 */
 611	list_add_rcu(&p->list, last_perm);
 612	ret = 0;
 613out:
 614	spin_unlock_bh(&inetsw6_lock);
 615	return ret;
 616
 617out_permanent:
 618	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 619	       protocol);
 620	goto out;
 621
 622out_illegal:
 623	printk(KERN_ERR
 624	       "Ignoring attempt to register invalid socket type %d.\n",
 625	       p->type);
 626	goto out;
 627}
 628
 629EXPORT_SYMBOL(inet6_register_protosw);
 630
 631void
 632inet6_unregister_protosw(struct inet_protosw *p)
 633{
 634	if (INET_PROTOSW_PERMANENT & p->flags) {
 635		printk(KERN_ERR
 636		       "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}
 646
 647EXPORT_SYMBOL(inet6_unregister_protosw);
 648
 649int inet6_sk_rebuild_header(struct sock *sk)
 650{
 651	struct ipv6_pinfo *np = inet6_sk(sk);
 652	struct dst_entry *dst;
 653
 654	dst = __sk_dst_check(sk, np->dst_cookie);
 655
 656	if (dst == NULL) {
 657		struct inet_sock *inet = inet_sk(sk);
 658		struct in6_addr *final_p, final;
 659		struct flowi6 fl6;
 660
 661		memset(&fl6, 0, sizeof(fl6));
 662		fl6.flowi6_proto = sk->sk_protocol;
 663		ipv6_addr_copy(&fl6.daddr, &np->daddr);
 664		ipv6_addr_copy(&fl6.saddr, &np->saddr);
 665		fl6.flowlabel = np->flow_label;
 666		fl6.flowi6_oif = sk->sk_bound_dev_if;
 667		fl6.flowi6_mark = sk->sk_mark;
 668		fl6.fl6_dport = inet->inet_dport;
 669		fl6.fl6_sport = inet->inet_sport;
 670		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 671
 672		final_p = fl6_update_dst(&fl6, np->opt, &final);
 
 
 
 673
 674		dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
 675		if (IS_ERR(dst)) {
 676			sk->sk_route_caps = 0;
 677			sk->sk_err_soft = -PTR_ERR(dst);
 678			return PTR_ERR(dst);
 679		}
 680
 681		__ip6_dst_store(sk, dst, NULL, NULL);
 682	}
 683
 684	return 0;
 685}
 686
 687EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 688
 689int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 
 690{
 691	struct ipv6_pinfo *np = inet6_sk(sk);
 692	struct inet6_skb_parm *opt = IP6CB(skb);
 693
 694	if (np->rxopt.all) {
 695		if ((opt->hop && (np->rxopt.bits.hopopts ||
 696				  np->rxopt.bits.ohopopts)) ||
 697		    ((IPV6_FLOWINFO_MASK &
 698		      *(__be32 *)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 1;
 705	}
 706	return 0;
 707}
 708
 709EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 710
 711static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
 712{
 713	const struct inet6_protocol *ops = NULL;
 714
 715	for (;;) {
 716		struct ipv6_opt_hdr *opth;
 717		int len;
 718
 719		if (proto != NEXTHDR_HOP) {
 720			ops = rcu_dereference(inet6_protos[proto]);
 721
 722			if (unlikely(!ops))
 723				break;
 724
 725			if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
 726				break;
 727		}
 728
 729		if (unlikely(!pskb_may_pull(skb, 8)))
 730			break;
 731
 732		opth = (void *)skb->data;
 733		len = ipv6_optlen(opth);
 734
 735		if (unlikely(!pskb_may_pull(skb, len)))
 736			break;
 737
 738		proto = opth->nexthdr;
 739		__skb_pull(skb, len);
 740	}
 741
 742	return proto;
 743}
 744
 745static int ipv6_gso_send_check(struct sk_buff *skb)
 746{
 747	const struct ipv6hdr *ipv6h;
 748	const struct inet6_protocol *ops;
 749	int err = -EINVAL;
 750
 751	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 752		goto out;
 753
 754	ipv6h = ipv6_hdr(skb);
 755	__skb_pull(skb, sizeof(*ipv6h));
 756	err = -EPROTONOSUPPORT;
 757
 758	rcu_read_lock();
 759	ops = rcu_dereference(inet6_protos[
 760		ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
 761
 762	if (likely(ops && ops->gso_send_check)) {
 763		skb_reset_transport_header(skb);
 764		err = ops->gso_send_check(skb);
 765	}
 766	rcu_read_unlock();
 767
 768out:
 769	return err;
 770}
 771
 772static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, u32 features)
 773{
 774	struct sk_buff *segs = ERR_PTR(-EINVAL);
 775	struct ipv6hdr *ipv6h;
 776	const struct inet6_protocol *ops;
 777	int proto;
 778	struct frag_hdr *fptr;
 779	unsigned int unfrag_ip6hlen;
 780	u8 *prevhdr;
 781	int offset = 0;
 782
 783	if (!(features & NETIF_F_V6_CSUM))
 784		features &= ~NETIF_F_SG;
 785
 786	if (unlikely(skb_shinfo(skb)->gso_type &
 787		     ~(SKB_GSO_UDP |
 788		       SKB_GSO_DODGY |
 789		       SKB_GSO_TCP_ECN |
 790		       SKB_GSO_TCPV6 |
 791		       0)))
 792		goto out;
 793
 794	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 795		goto out;
 796
 797	ipv6h = ipv6_hdr(skb);
 798	__skb_pull(skb, sizeof(*ipv6h));
 799	segs = ERR_PTR(-EPROTONOSUPPORT);
 800
 801	proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
 802	rcu_read_lock();
 803	ops = rcu_dereference(inet6_protos[proto]);
 804	if (likely(ops && ops->gso_segment)) {
 805		skb_reset_transport_header(skb);
 806		segs = ops->gso_segment(skb, features);
 807	}
 808	rcu_read_unlock();
 809
 810	if (IS_ERR(segs))
 811		goto out;
 812
 813	for (skb = segs; skb; skb = skb->next) {
 814		ipv6h = ipv6_hdr(skb);
 815		ipv6h->payload_len = htons(skb->len - skb->mac_len -
 816					   sizeof(*ipv6h));
 817		if (proto == IPPROTO_UDP) {
 818			unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
 819			fptr = (struct frag_hdr *)(skb_network_header(skb) +
 820				unfrag_ip6hlen);
 821			fptr->frag_off = htons(offset);
 822			if (skb->next != NULL)
 823				fptr->frag_off |= htons(IP6_MF);
 824			offset += (ntohs(ipv6h->payload_len) -
 825				   sizeof(struct frag_hdr));
 826		}
 827	}
 828
 829out:
 830	return segs;
 831}
 832
 833struct ipv6_gro_cb {
 834	struct napi_gro_cb napi;
 835	int proto;
 836};
 837
 838#define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
 839
 840static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
 841					 struct sk_buff *skb)
 842{
 843	const struct inet6_protocol *ops;
 844	struct sk_buff **pp = NULL;
 845	struct sk_buff *p;
 846	struct ipv6hdr *iph;
 847	unsigned int nlen;
 848	unsigned int hlen;
 849	unsigned int off;
 850	int flush = 1;
 851	int proto;
 852	__wsum csum;
 853
 854	off = skb_gro_offset(skb);
 855	hlen = off + sizeof(*iph);
 856	iph = skb_gro_header_fast(skb, off);
 857	if (skb_gro_header_hard(skb, hlen)) {
 858		iph = skb_gro_header_slow(skb, hlen, off);
 859		if (unlikely(!iph))
 860			goto out;
 861	}
 862
 863	skb_gro_pull(skb, sizeof(*iph));
 864	skb_set_transport_header(skb, skb_gro_offset(skb));
 865
 866	flush += ntohs(iph->payload_len) != skb_gro_len(skb);
 867
 868	rcu_read_lock();
 869	proto = iph->nexthdr;
 870	ops = rcu_dereference(inet6_protos[proto]);
 871	if (!ops || !ops->gro_receive) {
 872		__pskb_pull(skb, skb_gro_offset(skb));
 873		proto = ipv6_gso_pull_exthdrs(skb, proto);
 874		skb_gro_pull(skb, -skb_transport_offset(skb));
 875		skb_reset_transport_header(skb);
 876		__skb_push(skb, skb_gro_offset(skb));
 877
 878		ops = rcu_dereference(inet6_protos[proto]);
 879		if (!ops || !ops->gro_receive)
 880			goto out_unlock;
 881
 882		iph = ipv6_hdr(skb);
 883	}
 884
 885	IPV6_GRO_CB(skb)->proto = proto;
 886
 887	flush--;
 888	nlen = skb_network_header_len(skb);
 889
 890	for (p = *head; p; p = p->next) {
 891		struct ipv6hdr *iph2;
 892
 893		if (!NAPI_GRO_CB(p)->same_flow)
 894			continue;
 895
 896		iph2 = ipv6_hdr(p);
 897
 898		/* All fields must match except length. */
 899		if (nlen != skb_network_header_len(p) ||
 900		    memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
 901		    memcmp(&iph->nexthdr, &iph2->nexthdr,
 902			   nlen - offsetof(struct ipv6hdr, nexthdr))) {
 903			NAPI_GRO_CB(p)->same_flow = 0;
 904			continue;
 905		}
 906
 907		NAPI_GRO_CB(p)->flush |= flush;
 908	}
 909
 910	NAPI_GRO_CB(skb)->flush |= flush;
 911
 912	csum = skb->csum;
 913	skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
 914
 915	pp = ops->gro_receive(head, skb);
 916
 917	skb->csum = csum;
 918
 919out_unlock:
 920	rcu_read_unlock();
 921
 922out:
 923	NAPI_GRO_CB(skb)->flush |= flush;
 924
 925	return pp;
 926}
 927
 928static int ipv6_gro_complete(struct sk_buff *skb)
 929{
 930	const struct inet6_protocol *ops;
 931	struct ipv6hdr *iph = ipv6_hdr(skb);
 932	int err = -ENOSYS;
 933
 934	iph->payload_len = htons(skb->len - skb_network_offset(skb) -
 935				 sizeof(*iph));
 936
 937	rcu_read_lock();
 938	ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
 939	if (WARN_ON(!ops || !ops->gro_complete))
 940		goto out_unlock;
 941
 942	err = ops->gro_complete(skb);
 943
 944out_unlock:
 945	rcu_read_unlock();
 946
 947	return err;
 948}
 949
 950static struct packet_type ipv6_packet_type __read_mostly = {
 951	.type = cpu_to_be16(ETH_P_IPV6),
 952	.func = ipv6_rcv,
 953	.gso_send_check = ipv6_gso_send_check,
 954	.gso_segment = ipv6_gso_segment,
 955	.gro_receive = ipv6_gro_receive,
 956	.gro_complete = ipv6_gro_complete,
 957};
 958
 959static int __init ipv6_packet_init(void)
 960{
 961	dev_add_pack(&ipv6_packet_type);
 962	return 0;
 963}
 964
 965static void ipv6_packet_cleanup(void)
 966{
 967	dev_remove_pack(&ipv6_packet_type);
 968}
 969
 970static int __net_init ipv6_init_mibs(struct net *net)
 971{
 972	if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
 973			  sizeof(struct udp_mib),
 974			  __alignof__(struct udp_mib)) < 0)
 
 975		return -ENOMEM;
 976	if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
 977			  sizeof(struct udp_mib),
 978			  __alignof__(struct udp_mib)) < 0)
 979		goto err_udplite_mib;
 980	if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
 981			  sizeof(struct ipstats_mib),
 982			  __alignof__(struct ipstats_mib)) < 0)
 983		goto err_ip_mib;
 984	if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
 985			  sizeof(struct icmpv6_mib),
 986			  __alignof__(struct icmpv6_mib)) < 0)
 
 
 
 
 
 
 
 987		goto err_icmp_mib;
 988	if (snmp_mib_init((void __percpu **)net->mib.icmpv6msg_statistics,
 989			  sizeof(struct icmpv6msg_mib),
 990			  __alignof__(struct icmpv6msg_mib)) < 0)
 991		goto err_icmpmsg_mib;
 992	return 0;
 993
 994err_icmpmsg_mib:
 995	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 996err_icmp_mib:
 997	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 998err_ip_mib:
 999	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1000err_udplite_mib:
1001	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1002	return -ENOMEM;
1003}
1004
1005static void ipv6_cleanup_mibs(struct net *net)
1006{
1007	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1008	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1009	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1010	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1011	snmp_mib_free((void __percpu **)net->mib.icmpv6msg_statistics);
1012}
1013
1014static int __net_init inet6_net_init(struct net *net)
1015{
1016	int err = 0;
1017
1018	net->ipv6.sysctl.bindv6only = 0;
1019	net->ipv6.sysctl.icmpv6_time = 1*HZ;
 
 
 
 
 
 
1020
1021	err = ipv6_init_mibs(net);
1022	if (err)
1023		return err;
1024#ifdef CONFIG_PROC_FS
1025	err = udp6_proc_init(net);
1026	if (err)
1027		goto out;
1028	err = tcp6_proc_init(net);
1029	if (err)
1030		goto proc_tcp6_fail;
1031	err = ac6_proc_init(net);
1032	if (err)
1033		goto proc_ac6_fail;
1034#endif
1035	return err;
1036
1037#ifdef CONFIG_PROC_FS
1038proc_ac6_fail:
1039	tcp6_proc_exit(net);
1040proc_tcp6_fail:
1041	udp6_proc_exit(net);
1042out:
1043	ipv6_cleanup_mibs(net);
1044	return err;
1045#endif
1046}
1047
1048static void __net_exit inet6_net_exit(struct net *net)
1049{
1050#ifdef CONFIG_PROC_FS
1051	udp6_proc_exit(net);
1052	tcp6_proc_exit(net);
1053	ac6_proc_exit(net);
1054#endif
1055	ipv6_cleanup_mibs(net);
1056}
1057
1058static struct pernet_operations inet6_net_ops = {
1059	.init = inet6_net_init,
1060	.exit = inet6_net_exit,
1061};
1062
 
 
 
 
 
 
 
 
 
1063static int __init inet6_init(void)
1064{
1065	struct sk_buff *dummy_skb;
1066	struct list_head *r;
1067	int err = 0;
1068
1069	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1070
1071	/* Register the socket-side information for inet6_create.  */
1072	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1073		INIT_LIST_HEAD(r);
1074
1075	if (disable_ipv6_mod) {
1076		printk(KERN_INFO
1077		       "IPv6: Loaded, but administratively disabled, "
1078		       "reboot required to enable\n");
1079		goto out;
1080	}
1081
1082	err = proto_register(&tcpv6_prot, 1);
1083	if (err)
1084		goto out;
1085
1086	err = proto_register(&udpv6_prot, 1);
1087	if (err)
1088		goto out_unregister_tcp_proto;
1089
1090	err = proto_register(&udplitev6_prot, 1);
1091	if (err)
1092		goto out_unregister_udp_proto;
1093
1094	err = proto_register(&rawv6_prot, 1);
1095	if (err)
1096		goto out_unregister_udplite_proto;
1097
 
 
 
1098
1099	/* We MUST register RAW sockets before we create the ICMP6,
1100	 * IGMP6, or NDISC control sockets.
1101	 */
1102	err = rawv6_init();
1103	if (err)
1104		goto out_unregister_raw_proto;
1105
1106	/* Register the family here so that the init calls below will
1107	 * be able to create sockets. (?? is this dangerous ??)
1108	 */
1109	err = sock_register(&inet6_family_ops);
1110	if (err)
1111		goto out_sock_register_fail;
1112
1113#ifdef CONFIG_SYSCTL
1114	err = ipv6_static_sysctl_register();
1115	if (err)
1116		goto static_sysctl_fail;
1117#endif
1118	/*
1119	 *	ipngwg API draft makes clear that the correct semantics
1120	 *	for TCP and UDP is to consider one TCP and UDP instance
1121	 *	in a host available by both INET and INET6 APIs and
1122	 *	able to communicate via both network protocols.
1123	 */
1124
1125	err = register_pernet_subsys(&inet6_net_ops);
1126	if (err)
1127		goto register_pernet_fail;
1128	err = icmpv6_init();
1129	if (err)
1130		goto icmp_fail;
1131	err = ip6_mr_init();
1132	if (err)
1133		goto ipmr_fail;
1134	err = ndisc_init();
1135	if (err)
1136		goto ndisc_fail;
1137	err = igmp6_init();
1138	if (err)
1139		goto igmp_fail;
 
 
 
1140	err = ipv6_netfilter_init();
1141	if (err)
1142		goto netfilter_fail;
1143	/* Create /proc/foo6 entries. */
1144#ifdef CONFIG_PROC_FS
1145	err = -ENOMEM;
1146	if (raw6_proc_init())
1147		goto proc_raw6_fail;
1148	if (udplite6_proc_init())
1149		goto proc_udplite6_fail;
1150	if (ipv6_misc_proc_init())
1151		goto proc_misc6_fail;
1152	if (if6_proc_init())
1153		goto proc_if6_fail;
1154#endif
1155	err = ip6_route_init();
1156	if (err)
1157		goto ip6_route_fail;
 
 
 
1158	err = ip6_flowlabel_init();
1159	if (err)
1160		goto ip6_flowlabel_fail;
1161	err = addrconf_init();
1162	if (err)
1163		goto addrconf_fail;
1164
1165	/* Init v6 extension headers. */
1166	err = ipv6_exthdrs_init();
1167	if (err)
1168		goto ipv6_exthdrs_fail;
1169
1170	err = ipv6_frag_init();
1171	if (err)
1172		goto ipv6_frag_fail;
1173
1174	/* Init v6 transport protocols. */
1175	err = udpv6_init();
1176	if (err)
1177		goto udpv6_fail;
1178
1179	err = udplitev6_init();
1180	if (err)
1181		goto udplitev6_fail;
1182
1183	err = tcpv6_init();
1184	if (err)
1185		goto tcpv6_fail;
1186
1187	err = ipv6_packet_init();
1188	if (err)
1189		goto ipv6_packet_fail;
1190
 
 
 
 
1191#ifdef CONFIG_SYSCTL
1192	err = ipv6_sysctl_register();
1193	if (err)
1194		goto sysctl_fail;
1195#endif
1196out:
1197	return err;
1198
1199#ifdef CONFIG_SYSCTL
1200sysctl_fail:
1201	ipv6_packet_cleanup();
1202#endif
 
 
1203ipv6_packet_fail:
1204	tcpv6_exit();
1205tcpv6_fail:
1206	udplitev6_exit();
1207udplitev6_fail:
1208	udpv6_exit();
1209udpv6_fail:
1210	ipv6_frag_exit();
1211ipv6_frag_fail:
1212	ipv6_exthdrs_exit();
1213ipv6_exthdrs_fail:
1214	addrconf_cleanup();
1215addrconf_fail:
1216	ip6_flowlabel_cleanup();
1217ip6_flowlabel_fail:
 
 
1218	ip6_route_cleanup();
1219ip6_route_fail:
1220#ifdef CONFIG_PROC_FS
1221	if6_proc_exit();
1222proc_if6_fail:
1223	ipv6_misc_proc_exit();
1224proc_misc6_fail:
1225	udplite6_proc_exit();
1226proc_udplite6_fail:
1227	raw6_proc_exit();
1228proc_raw6_fail:
1229#endif
1230	ipv6_netfilter_fini();
1231netfilter_fail:
1232	igmp6_cleanup();
1233igmp_fail:
1234	ndisc_cleanup();
1235ndisc_fail:
1236	ip6_mr_cleanup();
1237ipmr_fail:
1238	icmpv6_cleanup();
1239icmp_fail:
1240	unregister_pernet_subsys(&inet6_net_ops);
1241register_pernet_fail:
1242#ifdef CONFIG_SYSCTL
1243	ipv6_static_sysctl_unregister();
1244static_sysctl_fail:
1245#endif
1246	sock_unregister(PF_INET6);
1247	rtnl_unregister_all(PF_INET6);
1248out_sock_register_fail:
1249	rawv6_exit();
 
 
1250out_unregister_raw_proto:
1251	proto_unregister(&rawv6_prot);
1252out_unregister_udplite_proto:
1253	proto_unregister(&udplitev6_prot);
1254out_unregister_udp_proto:
1255	proto_unregister(&udpv6_prot);
1256out_unregister_tcp_proto:
1257	proto_unregister(&tcpv6_prot);
1258	goto out;
1259}
1260module_init(inet6_init);
1261
1262static void __exit inet6_exit(void)
1263{
1264	if (disable_ipv6_mod)
1265		return;
1266
1267	/* First of all disallow new sockets creation. */
1268	sock_unregister(PF_INET6);
1269	/* Disallow any further netlink messages */
1270	rtnl_unregister_all(PF_INET6);
1271
1272#ifdef CONFIG_SYSCTL
1273	ipv6_sysctl_unregister();
1274#endif
1275	udpv6_exit();
1276	udplitev6_exit();
1277	tcpv6_exit();
1278
1279	/* Cleanup code parts. */
1280	ipv6_packet_cleanup();
1281	ipv6_frag_exit();
1282	ipv6_exthdrs_exit();
1283	addrconf_cleanup();
1284	ip6_flowlabel_cleanup();
1285	ip6_route_cleanup();
1286#ifdef CONFIG_PROC_FS
1287
1288	/* Cleanup code parts. */
1289	if6_proc_exit();
1290	ipv6_misc_proc_exit();
1291	udplite6_proc_exit();
1292	raw6_proc_exit();
1293#endif
1294	ipv6_netfilter_fini();
1295	igmp6_cleanup();
1296	ndisc_cleanup();
1297	ip6_mr_cleanup();
1298	icmpv6_cleanup();
1299	rawv6_exit();
1300
1301	unregister_pernet_subsys(&inet6_net_ops);
1302#ifdef CONFIG_SYSCTL
1303	ipv6_static_sysctl_unregister();
1304#endif
1305	proto_unregister(&rawv6_prot);
1306	proto_unregister(&udplitev6_prot);
1307	proto_unregister(&udpv6_prot);
1308	proto_unregister(&tcpv6_prot);
1309
1310	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1311}
1312module_exit(inet6_exit);
1313
1314MODULE_ALIAS_NETPROTO(PF_INET6);
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);