Linux Audio

Check our new training course

Linux kernel drivers training

May 6-19, 2025
Register
Loading...
v3.15
   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	char answer_no_check;
 110	int try_loading_module = 0;
 111	int err;
 112
 
 
 
 
 
 113	/* Look for the requested type/protocol pair. */
 114lookup_protocol:
 115	err = -ESOCKTNOSUPPORT;
 116	rcu_read_lock();
 117	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 118
 119		err = 0;
 120		/* Check the non-wild match. */
 121		if (protocol == answer->protocol) {
 122			if (protocol != IPPROTO_IP)
 123				break;
 124		} else {
 125			/* Check for the two wild cases. */
 126			if (IPPROTO_IP == protocol) {
 127				protocol = answer->protocol;
 128				break;
 129			}
 130			if (IPPROTO_IP == answer->protocol)
 131				break;
 132		}
 133		err = -EPROTONOSUPPORT;
 134	}
 135
 136	if (err) {
 137		if (try_loading_module < 2) {
 138			rcu_read_unlock();
 139			/*
 140			 * Be more specific, e.g. net-pf-10-proto-132-type-1
 141			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 142			 */
 143			if (++try_loading_module == 1)
 144				request_module("net-pf-%d-proto-%d-type-%d",
 145						PF_INET6, protocol, sock->type);
 146			/*
 147			 * Fall back to generic, e.g. net-pf-10-proto-132
 148			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 149			 */
 150			else
 151				request_module("net-pf-%d-proto-%d",
 152						PF_INET6, protocol);
 153			goto lookup_protocol;
 154		} else
 155			goto out_rcu_unlock;
 156	}
 157
 158	err = -EPERM;
 159	if (sock->type == SOCK_RAW && !kern &&
 160	    !ns_capable(net->user_ns, CAP_NET_RAW))
 161		goto out_rcu_unlock;
 162
 163	sock->ops = answer->ops;
 164	answer_prot = answer->prot;
 165	answer_no_check = answer->no_check;
 166	answer_flags = answer->flags;
 167	rcu_read_unlock();
 168
 169	WARN_ON(answer_prot->slab == NULL);
 170
 171	err = -ENOBUFS;
 172	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 173	if (sk == NULL)
 174		goto out;
 175
 176	sock_init_data(sock, sk);
 177
 178	err = 0;
 179	sk->sk_no_check = answer_no_check;
 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->ipv6only	= net->ipv6.sysctl.bindv6only;
 204
 205	/* Init the ipv4 part of the socket since we can have sockets
 206	 * using v6 API for ipv4.
 207	 */
 208	inet->uc_ttl	= -1;
 209
 210	inet->mc_loop	= 1;
 211	inet->mc_ttl	= 1;
 212	inet->mc_index	= 0;
 213	inet->mc_list	= NULL;
 214	inet->rcv_tos	= 0;
 215
 216	if (net->ipv4.sysctl_ip_no_pmtu_disc)
 217		inet->pmtudisc = IP_PMTUDISC_DONT;
 218	else
 219		inet->pmtudisc = IP_PMTUDISC_WANT;
 220	/*
 221	 * Increment only the relevant sk_prot->socks debug field, this changes
 222	 * the previous behaviour of incrementing both the equivalent to
 223	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 224	 *
 225	 * This allows better debug granularity as we'll know exactly how many
 226	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 227	 * transport protocol socks. -acme
 228	 */
 229	sk_refcnt_debug_inc(sk);
 230
 231	if (inet->inet_num) {
 232		/* It assumes that any protocol which allows
 233		 * the user to assign a number at socket
 234		 * creation time automatically shares.
 235		 */
 236		inet->inet_sport = htons(inet->inet_num);
 237		sk->sk_prot->hash(sk);
 238	}
 239	if (sk->sk_prot->init) {
 240		err = sk->sk_prot->init(sk);
 241		if (err) {
 242			sk_common_release(sk);
 243			goto out;
 244		}
 245	}
 246out:
 247	return err;
 248out_rcu_unlock:
 249	rcu_read_unlock();
 250	goto out;
 251}
 252
 253
 254/* bind for INET6 API */
 255int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 256{
 257	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 258	struct sock *sk = sock->sk;
 259	struct inet_sock *inet = inet_sk(sk);
 260	struct ipv6_pinfo *np = inet6_sk(sk);
 261	struct net *net = sock_net(sk);
 262	__be32 v4addr = 0;
 263	unsigned short snum;
 264	int addr_type = 0;
 265	int err = 0;
 266
 267	/* If the socket has its own bind function then use it. */
 268	if (sk->sk_prot->bind)
 269		return sk->sk_prot->bind(sk, uaddr, addr_len);
 270
 271	if (addr_len < SIN6_LEN_RFC2133)
 272		return -EINVAL;
 273
 274	if (addr->sin6_family != AF_INET6)
 275		return -EAFNOSUPPORT;
 276
 277	addr_type = ipv6_addr_type(&addr->sin6_addr);
 278	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 279		return -EINVAL;
 280
 281	snum = ntohs(addr->sin6_port);
 282	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 283		return -EACCES;
 284
 285	lock_sock(sk);
 286
 287	/* Check these errors (active socket, double bind). */
 288	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 289		err = -EINVAL;
 290		goto out;
 291	}
 292
 293	/* Check if the address belongs to the host. */
 294	if (addr_type == IPV6_ADDR_MAPPED) {
 295		int chk_addr_ret;
 296
 297		/* Binding to v4-mapped address on a v6-only socket
 298		 * makes no sense
 299		 */
 300		if (np->ipv6only) {
 301			err = -EINVAL;
 302			goto out;
 303		}
 304
 305		/* Reproduce AF_INET checks to make the bindings consistent */
 306		v4addr = addr->sin6_addr.s6_addr32[3];
 307		chk_addr_ret = inet_addr_type(net, v4addr);
 308		if (!sysctl_ip_nonlocal_bind &&
 309		    !(inet->freebind || inet->transparent) &&
 310		    v4addr != htonl(INADDR_ANY) &&
 311		    chk_addr_ret != RTN_LOCAL &&
 312		    chk_addr_ret != RTN_MULTICAST &&
 313		    chk_addr_ret != RTN_BROADCAST) {
 314			err = -EADDRNOTAVAIL;
 315			goto out;
 316		}
 317	} else {
 318		if (addr_type != IPV6_ADDR_ANY) {
 319			struct net_device *dev = NULL;
 320
 321			rcu_read_lock();
 322			if (__ipv6_addr_needs_scope_id(addr_type)) {
 323				if (addr_len >= sizeof(struct sockaddr_in6) &&
 324				    addr->sin6_scope_id) {
 325					/* Override any existing binding, if another one
 326					 * is supplied by user.
 327					 */
 328					sk->sk_bound_dev_if = addr->sin6_scope_id;
 329				}
 330
 331				/* Binding to link-local address requires an interface */
 332				if (!sk->sk_bound_dev_if) {
 333					err = -EINVAL;
 334					goto out_unlock;
 335				}
 336				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 337				if (!dev) {
 338					err = -ENODEV;
 339					goto out_unlock;
 340				}
 341			}
 342
 343			/* ipv4 addr of the socket is invalid.  Only the
 344			 * unspecified and mapped address have a v4 equivalent.
 345			 */
 346			v4addr = LOOPBACK4_IPV6;
 347			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
 348				if (!(inet->freebind || inet->transparent) &&
 349				    !ipv6_chk_addr(net, &addr->sin6_addr,
 350						   dev, 0)) {
 351					err = -EADDRNOTAVAIL;
 352					goto out_unlock;
 353				}
 354			}
 355			rcu_read_unlock();
 356		}
 357	}
 358
 359	inet->inet_rcv_saddr = v4addr;
 360	inet->inet_saddr = v4addr;
 361
 362	sk->sk_v6_rcv_saddr = addr->sin6_addr;
 363
 364	if (!(addr_type & IPV6_ADDR_MULTICAST))
 365		np->saddr = addr->sin6_addr;
 366
 367	/* Make sure we are allowed to bind here. */
 368	if (sk->sk_prot->get_port(sk, snum)) {
 369		inet_reset_saddr(sk);
 370		err = -EADDRINUSE;
 371		goto out;
 372	}
 373
 374	if (addr_type != IPV6_ADDR_ANY) {
 375		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 376		if (addr_type != IPV6_ADDR_MAPPED)
 377			np->ipv6only = 1;
 378	}
 379	if (snum)
 380		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 381	inet->inet_sport = htons(inet->inet_num);
 382	inet->inet_dport = 0;
 383	inet->inet_daddr = 0;
 384out:
 385	release_sock(sk);
 386	return err;
 387out_unlock:
 388	rcu_read_unlock();
 389	goto out;
 390}
 
 391EXPORT_SYMBOL(inet6_bind);
 392
 393int inet6_release(struct socket *sock)
 394{
 395	struct sock *sk = sock->sk;
 396
 397	if (sk == NULL)
 398		return -EINVAL;
 399
 400	/* Free mc lists */
 401	ipv6_sock_mc_close(sk);
 402
 403	/* Free ac lists */
 404	ipv6_sock_ac_close(sk);
 405
 406	return inet_release(sock);
 407}
 
 408EXPORT_SYMBOL(inet6_release);
 409
 410void inet6_destroy_sock(struct sock *sk)
 411{
 412	struct ipv6_pinfo *np = inet6_sk(sk);
 413	struct sk_buff *skb;
 414	struct ipv6_txoptions *opt;
 415
 416	/* Release rx options */
 417
 418	skb = xchg(&np->pktoptions, NULL);
 419	if (skb != NULL)
 420		kfree_skb(skb);
 421
 422	skb = xchg(&np->rxpmtu, NULL);
 423	if (skb != NULL)
 424		kfree_skb(skb);
 425
 426	/* Free flowlabels */
 427	fl6_free_socklist(sk);
 428
 429	/* Free tx options */
 430
 431	opt = xchg(&np->opt, NULL);
 432	if (opt != NULL)
 433		sock_kfree_s(sk, opt, opt->tot_len);
 434}
 
 435EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 436
 437/*
 438 *	This does both peername and sockname.
 439 */
 440
 441int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 442		 int *uaddr_len, int peer)
 443{
 444	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 445	struct sock *sk = sock->sk;
 446	struct inet_sock *inet = inet_sk(sk);
 447	struct ipv6_pinfo *np = inet6_sk(sk);
 448
 449	sin->sin6_family = AF_INET6;
 450	sin->sin6_flowinfo = 0;
 451	sin->sin6_scope_id = 0;
 452	if (peer) {
 453		if (!inet->inet_dport)
 454			return -ENOTCONN;
 455		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 456		    peer == 1)
 457			return -ENOTCONN;
 458		sin->sin6_port = inet->inet_dport;
 459		sin->sin6_addr = sk->sk_v6_daddr;
 460		if (np->sndflow)
 461			sin->sin6_flowinfo = np->flow_label;
 462	} else {
 463		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 464			sin->sin6_addr = np->saddr;
 465		else
 466			sin->sin6_addr = sk->sk_v6_rcv_saddr;
 467
 468		sin->sin6_port = inet->inet_sport;
 469	}
 470	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 471						 sk->sk_bound_dev_if);
 472	*uaddr_len = sizeof(*sin);
 473	return 0;
 474}
 
 475EXPORT_SYMBOL(inet6_getname);
 476
 477int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 478{
 479	struct sock *sk = sock->sk;
 480	struct net *net = sock_net(sk);
 481
 482	switch (cmd) {
 
 483	case SIOCGSTAMP:
 484		return sock_get_timestamp(sk, (struct timeval __user *)arg);
 485
 486	case SIOCGSTAMPNS:
 487		return sock_get_timestampns(sk, (struct timespec __user *)arg);
 488
 489	case SIOCADDRT:
 490	case SIOCDELRT:
 491
 492		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 493
 494	case SIOCSIFADDR:
 495		return addrconf_add_ifaddr(net, (void __user *) arg);
 496	case SIOCDIFADDR:
 497		return addrconf_del_ifaddr(net, (void __user *) arg);
 498	case SIOCSIFDSTADDR:
 499		return addrconf_set_dstaddr(net, (void __user *) arg);
 500	default:
 501		if (!sk->sk_prot->ioctl)
 502			return -ENOIOCTLCMD;
 503		return sk->sk_prot->ioctl(sk, cmd, arg);
 504	}
 505	/*NOTREACHED*/
 506	return 0;
 507}
 
 508EXPORT_SYMBOL(inet6_ioctl);
 509
 510const struct proto_ops inet6_stream_ops = {
 511	.family		   = PF_INET6,
 512	.owner		   = THIS_MODULE,
 513	.release	   = inet6_release,
 514	.bind		   = inet6_bind,
 515	.connect	   = inet_stream_connect,	/* ok		*/
 516	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
 517	.accept		   = inet_accept,		/* ok		*/
 518	.getname	   = inet6_getname,
 519	.poll		   = tcp_poll,			/* ok		*/
 520	.ioctl		   = inet6_ioctl,		/* must change  */
 521	.listen		   = inet_listen,		/* ok		*/
 522	.shutdown	   = inet_shutdown,		/* ok		*/
 523	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
 524	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
 525	.sendmsg	   = inet_sendmsg,		/* ok		*/
 526	.recvmsg	   = inet_recvmsg,		/* ok		*/
 527	.mmap		   = sock_no_mmap,
 528	.sendpage	   = inet_sendpage,
 529	.splice_read	   = tcp_splice_read,
 530#ifdef CONFIG_COMPAT
 531	.compat_setsockopt = compat_sock_common_setsockopt,
 532	.compat_getsockopt = compat_sock_common_getsockopt,
 533#endif
 534};
 535
 536const struct proto_ops inet6_dgram_ops = {
 537	.family		   = PF_INET6,
 538	.owner		   = THIS_MODULE,
 539	.release	   = inet6_release,
 540	.bind		   = inet6_bind,
 541	.connect	   = inet_dgram_connect,	/* ok		*/
 542	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
 543	.accept		   = sock_no_accept,		/* a do nothing	*/
 544	.getname	   = inet6_getname,
 545	.poll		   = udp_poll,			/* ok		*/
 546	.ioctl		   = inet6_ioctl,		/* must change  */
 547	.listen		   = sock_no_listen,		/* ok		*/
 548	.shutdown	   = inet_shutdown,		/* ok		*/
 549	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
 550	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
 551	.sendmsg	   = inet_sendmsg,		/* ok		*/
 552	.recvmsg	   = inet_recvmsg,		/* ok		*/
 553	.mmap		   = sock_no_mmap,
 554	.sendpage	   = sock_no_sendpage,
 555#ifdef CONFIG_COMPAT
 556	.compat_setsockopt = compat_sock_common_setsockopt,
 557	.compat_getsockopt = compat_sock_common_getsockopt,
 558#endif
 559};
 560
 561static const struct net_proto_family inet6_family_ops = {
 562	.family = PF_INET6,
 563	.create = inet6_create,
 564	.owner	= THIS_MODULE,
 565};
 566
 567int inet6_register_protosw(struct inet_protosw *p)
 568{
 569	struct list_head *lh;
 570	struct inet_protosw *answer;
 571	struct list_head *last_perm;
 572	int protocol = p->protocol;
 573	int ret;
 574
 575	spin_lock_bh(&inetsw6_lock);
 576
 577	ret = -EINVAL;
 578	if (p->type >= SOCK_MAX)
 579		goto out_illegal;
 580
 581	/* If we are trying to override a permanent protocol, bail. */
 582	answer = NULL;
 583	ret = -EPERM;
 584	last_perm = &inetsw6[p->type];
 585	list_for_each(lh, &inetsw6[p->type]) {
 586		answer = list_entry(lh, struct inet_protosw, list);
 587
 588		/* Check only the non-wild match. */
 589		if (INET_PROTOSW_PERMANENT & answer->flags) {
 590			if (protocol == answer->protocol)
 591				break;
 592			last_perm = lh;
 593		}
 594
 595		answer = NULL;
 596	}
 597	if (answer)
 598		goto out_permanent;
 599
 600	/* Add the new entry after the last permanent entry if any, so that
 601	 * the new entry does not override a permanent entry when matched with
 602	 * a wild-card protocol. But it is allowed to override any existing
 603	 * non-permanent entry.  This means that when we remove this entry, the
 604	 * system automatically returns to the old behavior.
 605	 */
 606	list_add_rcu(&p->list, last_perm);
 607	ret = 0;
 608out:
 609	spin_unlock_bh(&inetsw6_lock);
 610	return ret;
 611
 612out_permanent:
 613	pr_err("Attempt to override permanent protocol %d\n", protocol);
 
 614	goto out;
 615
 616out_illegal:
 617	pr_err("Ignoring attempt to register invalid socket type %d\n",
 
 618	       p->type);
 619	goto out;
 620}
 
 621EXPORT_SYMBOL(inet6_register_protosw);
 622
 623void
 624inet6_unregister_protosw(struct inet_protosw *p)
 625{
 626	if (INET_PROTOSW_PERMANENT & p->flags) {
 627		pr_err("Attempt to unregister permanent protocol %d\n",
 
 628		       p->protocol);
 629	} else {
 630		spin_lock_bh(&inetsw6_lock);
 631		list_del_rcu(&p->list);
 632		spin_unlock_bh(&inetsw6_lock);
 633
 634		synchronize_net();
 635	}
 636}
 
 637EXPORT_SYMBOL(inet6_unregister_protosw);
 638
 639int inet6_sk_rebuild_header(struct sock *sk)
 640{
 641	struct ipv6_pinfo *np = inet6_sk(sk);
 642	struct dst_entry *dst;
 643
 644	dst = __sk_dst_check(sk, np->dst_cookie);
 645
 646	if (dst == NULL) {
 647		struct inet_sock *inet = inet_sk(sk);
 648		struct in6_addr *final_p, final;
 649		struct flowi6 fl6;
 650
 651		memset(&fl6, 0, sizeof(fl6));
 652		fl6.flowi6_proto = sk->sk_protocol;
 653		fl6.daddr = sk->sk_v6_daddr;
 654		fl6.saddr = np->saddr;
 655		fl6.flowlabel = np->flow_label;
 656		fl6.flowi6_oif = sk->sk_bound_dev_if;
 657		fl6.flowi6_mark = sk->sk_mark;
 658		fl6.fl6_dport = inet->inet_dport;
 659		fl6.fl6_sport = inet->inet_sport;
 660		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 661
 662		final_p = fl6_update_dst(&fl6, np->opt, &final);
 663
 664		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 665		if (IS_ERR(dst)) {
 666			sk->sk_route_caps = 0;
 667			sk->sk_err_soft = -PTR_ERR(dst);
 668			return PTR_ERR(dst);
 669		}
 670
 671		__ip6_dst_store(sk, dst, NULL, NULL);
 672	}
 673
 674	return 0;
 675}
 
 676EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 677
 678bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
 679{
 680	const struct ipv6_pinfo *np = inet6_sk(sk);
 681	const struct inet6_skb_parm *opt = IP6CB(skb);
 682
 683	if (np->rxopt.all) {
 684		if ((opt->hop && (np->rxopt.bits.hopopts ||
 685				  np->rxopt.bits.ohopopts)) ||
 686		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 
 687		     np->rxopt.bits.rxflow) ||
 688		    (opt->srcrt && (np->rxopt.bits.srcrt ||
 689		     np->rxopt.bits.osrcrt)) ||
 690		    ((opt->dst1 || opt->dst0) &&
 691		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 692			return true;
 693	}
 694	return false;
 695}
 
 696EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 697
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 698static struct packet_type ipv6_packet_type __read_mostly = {
 699	.type = cpu_to_be16(ETH_P_IPV6),
 700	.func = ipv6_rcv,
 
 
 
 
 701};
 702
 703static int __init ipv6_packet_init(void)
 704{
 705	dev_add_pack(&ipv6_packet_type);
 706	return 0;
 707}
 708
 709static void ipv6_packet_cleanup(void)
 710{
 711	dev_remove_pack(&ipv6_packet_type);
 712}
 713
 714static int __net_init ipv6_init_mibs(struct net *net)
 715{
 716	int i;
 717
 718	if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
 719			  sizeof(struct udp_mib),
 720			  __alignof__(struct udp_mib)) < 0)
 721		return -ENOMEM;
 722	if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
 723			  sizeof(struct udp_mib),
 724			  __alignof__(struct udp_mib)) < 0)
 725		goto err_udplite_mib;
 726	if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
 727			  sizeof(struct ipstats_mib),
 728			  __alignof__(struct ipstats_mib)) < 0)
 729		goto err_ip_mib;
 730
 731	for_each_possible_cpu(i) {
 732		struct ipstats_mib *af_inet6_stats;
 733		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics[0], i);
 734		u64_stats_init(&af_inet6_stats->syncp);
 735#if SNMP_ARRAY_SZ == 2
 736		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics[1], i);
 737		u64_stats_init(&af_inet6_stats->syncp);
 738#endif
 739	}
 740
 741
 742	if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
 743			  sizeof(struct icmpv6_mib),
 744			  __alignof__(struct icmpv6_mib)) < 0)
 745		goto err_icmp_mib;
 746	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 747						GFP_KERNEL);
 748	if (!net->mib.icmpv6msg_statistics)
 749		goto err_icmpmsg_mib;
 750	return 0;
 751
 752err_icmpmsg_mib:
 753	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 754err_icmp_mib:
 755	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 756err_ip_mib:
 757	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 758err_udplite_mib:
 759	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 760	return -ENOMEM;
 761}
 762
 763static void ipv6_cleanup_mibs(struct net *net)
 764{
 765	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 766	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 767	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 768	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 769	kfree(net->mib.icmpv6msg_statistics);
 770}
 771
 772static int __net_init inet6_net_init(struct net *net)
 773{
 774	int err = 0;
 775
 776	net->ipv6.sysctl.bindv6only = 0;
 777	net->ipv6.sysctl.icmpv6_time = 1*HZ;
 778	net->ipv6.sysctl.flowlabel_consistency = 1;
 779	atomic_set(&net->ipv6.rt_genid, 0);
 780
 781	err = ipv6_init_mibs(net);
 782	if (err)
 783		return err;
 784#ifdef CONFIG_PROC_FS
 785	err = udp6_proc_init(net);
 786	if (err)
 787		goto out;
 788	err = tcp6_proc_init(net);
 789	if (err)
 790		goto proc_tcp6_fail;
 791	err = ac6_proc_init(net);
 792	if (err)
 793		goto proc_ac6_fail;
 794#endif
 795	return err;
 796
 797#ifdef CONFIG_PROC_FS
 798proc_ac6_fail:
 799	tcp6_proc_exit(net);
 800proc_tcp6_fail:
 801	udp6_proc_exit(net);
 802out:
 803	ipv6_cleanup_mibs(net);
 804	return err;
 805#endif
 806}
 807
 808static void __net_exit inet6_net_exit(struct net *net)
 809{
 810#ifdef CONFIG_PROC_FS
 811	udp6_proc_exit(net);
 812	tcp6_proc_exit(net);
 813	ac6_proc_exit(net);
 814#endif
 815	ipv6_cleanup_mibs(net);
 816}
 817
 818static struct pernet_operations inet6_net_ops = {
 819	.init = inet6_net_init,
 820	.exit = inet6_net_exit,
 821};
 822
 823static const struct ipv6_stub ipv6_stub_impl = {
 824	.ipv6_sock_mc_join = ipv6_sock_mc_join,
 825	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 826	.ipv6_dst_lookup = ip6_dst_lookup,
 827	.udpv6_encap_enable = udpv6_encap_enable,
 828	.ndisc_send_na = ndisc_send_na,
 829	.nd_tbl	= &nd_tbl,
 830};
 831
 832static int __init inet6_init(void)
 833{
 
 834	struct list_head *r;
 835	int err = 0;
 836
 837	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
 838
 839	/* Register the socket-side information for inet6_create.  */
 840	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 841		INIT_LIST_HEAD(r);
 842
 843	if (disable_ipv6_mod) {
 844		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 
 
 845		goto out;
 846	}
 847
 848	err = proto_register(&tcpv6_prot, 1);
 849	if (err)
 850		goto out;
 851
 852	err = proto_register(&udpv6_prot, 1);
 853	if (err)
 854		goto out_unregister_tcp_proto;
 855
 856	err = proto_register(&udplitev6_prot, 1);
 857	if (err)
 858		goto out_unregister_udp_proto;
 859
 860	err = proto_register(&rawv6_prot, 1);
 861	if (err)
 862		goto out_unregister_udplite_proto;
 863
 864	err = proto_register(&pingv6_prot, 1);
 865	if (err)
 866		goto out_unregister_ping_proto;
 867
 868	/* We MUST register RAW sockets before we create the ICMP6,
 869	 * IGMP6, or NDISC control sockets.
 870	 */
 871	err = rawv6_init();
 872	if (err)
 873		goto out_unregister_raw_proto;
 874
 875	/* Register the family here so that the init calls below will
 876	 * be able to create sockets. (?? is this dangerous ??)
 877	 */
 878	err = sock_register(&inet6_family_ops);
 879	if (err)
 880		goto out_sock_register_fail;
 881
 
 
 
 
 
 882	/*
 883	 *	ipngwg API draft makes clear that the correct semantics
 884	 *	for TCP and UDP is to consider one TCP and UDP instance
 885	 *	in a host available by both INET and INET6 APIs and
 886	 *	able to communicate via both network protocols.
 887	 */
 888
 889	err = register_pernet_subsys(&inet6_net_ops);
 890	if (err)
 891		goto register_pernet_fail;
 892	err = icmpv6_init();
 893	if (err)
 894		goto icmp_fail;
 895	err = ip6_mr_init();
 896	if (err)
 897		goto ipmr_fail;
 898	err = ndisc_init();
 899	if (err)
 900		goto ndisc_fail;
 901	err = igmp6_init();
 902	if (err)
 903		goto igmp_fail;
 904
 905	ipv6_stub = &ipv6_stub_impl;
 906
 907	err = ipv6_netfilter_init();
 908	if (err)
 909		goto netfilter_fail;
 910	/* Create /proc/foo6 entries. */
 911#ifdef CONFIG_PROC_FS
 912	err = -ENOMEM;
 913	if (raw6_proc_init())
 914		goto proc_raw6_fail;
 915	if (udplite6_proc_init())
 916		goto proc_udplite6_fail;
 917	if (ipv6_misc_proc_init())
 918		goto proc_misc6_fail;
 919	if (if6_proc_init())
 920		goto proc_if6_fail;
 921#endif
 922	err = ip6_route_init();
 923	if (err)
 924		goto ip6_route_fail;
 925	err = ndisc_late_init();
 926	if (err)
 927		goto ndisc_late_fail;
 928	err = ip6_flowlabel_init();
 929	if (err)
 930		goto ip6_flowlabel_fail;
 931	err = addrconf_init();
 932	if (err)
 933		goto addrconf_fail;
 934
 935	/* Init v6 extension headers. */
 936	err = ipv6_exthdrs_init();
 937	if (err)
 938		goto ipv6_exthdrs_fail;
 939
 940	err = ipv6_frag_init();
 941	if (err)
 942		goto ipv6_frag_fail;
 943
 944	/* Init v6 transport protocols. */
 945	err = udpv6_init();
 946	if (err)
 947		goto udpv6_fail;
 948
 949	err = udplitev6_init();
 950	if (err)
 951		goto udplitev6_fail;
 952
 953	err = tcpv6_init();
 954	if (err)
 955		goto tcpv6_fail;
 956
 957	err = ipv6_packet_init();
 958	if (err)
 959		goto ipv6_packet_fail;
 960
 961	err = pingv6_init();
 962	if (err)
 963		goto pingv6_fail;
 964
 965#ifdef CONFIG_SYSCTL
 966	err = ipv6_sysctl_register();
 967	if (err)
 968		goto sysctl_fail;
 969#endif
 970out:
 971	return err;
 972
 973#ifdef CONFIG_SYSCTL
 974sysctl_fail:
 975	pingv6_exit();
 976#endif
 977pingv6_fail:
 978	ipv6_packet_cleanup();
 
 979ipv6_packet_fail:
 980	tcpv6_exit();
 981tcpv6_fail:
 982	udplitev6_exit();
 983udplitev6_fail:
 984	udpv6_exit();
 985udpv6_fail:
 986	ipv6_frag_exit();
 987ipv6_frag_fail:
 988	ipv6_exthdrs_exit();
 989ipv6_exthdrs_fail:
 990	addrconf_cleanup();
 991addrconf_fail:
 992	ip6_flowlabel_cleanup();
 993ip6_flowlabel_fail:
 994	ndisc_late_cleanup();
 995ndisc_late_fail:
 996	ip6_route_cleanup();
 997ip6_route_fail:
 998#ifdef CONFIG_PROC_FS
 999	if6_proc_exit();
1000proc_if6_fail:
1001	ipv6_misc_proc_exit();
1002proc_misc6_fail:
1003	udplite6_proc_exit();
1004proc_udplite6_fail:
1005	raw6_proc_exit();
1006proc_raw6_fail:
1007#endif
1008	ipv6_netfilter_fini();
1009netfilter_fail:
1010	igmp6_cleanup();
1011igmp_fail:
1012	ndisc_cleanup();
1013ndisc_fail:
1014	ip6_mr_cleanup();
1015ipmr_fail:
1016	icmpv6_cleanup();
1017icmp_fail:
1018	unregister_pernet_subsys(&inet6_net_ops);
1019register_pernet_fail:
 
 
 
 
1020	sock_unregister(PF_INET6);
1021	rtnl_unregister_all(PF_INET6);
1022out_sock_register_fail:
1023	rawv6_exit();
1024out_unregister_ping_proto:
1025	proto_unregister(&pingv6_prot);
1026out_unregister_raw_proto:
1027	proto_unregister(&rawv6_prot);
1028out_unregister_udplite_proto:
1029	proto_unregister(&udplitev6_prot);
1030out_unregister_udp_proto:
1031	proto_unregister(&udpv6_prot);
1032out_unregister_tcp_proto:
1033	proto_unregister(&tcpv6_prot);
1034	goto out;
1035}
1036module_init(inet6_init);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1037
1038MODULE_ALIAS_NETPROTO(PF_INET6);
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);