Linux Audio

Check our new training course

Loading...
v6.8
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	Internet Control Message Protocol (ICMPv6)
   4 *	Linux INET6 implementation
   5 *
   6 *	Authors:
   7 *	Pedro Roque		<roque@di.fc.ul.pt>
   8 *
   9 *	Based on net/ipv4/icmp.c
  10 *
  11 *	RFC 1885
 
 
 
 
 
  12 */
  13
  14/*
  15 *	Changes:
  16 *
  17 *	Andi Kleen		:	exception handling
  18 *	Andi Kleen			add rate limits. never reply to a icmp.
  19 *					add more length checks and other fixes.
  20 *	yoshfuji		:	ensure to sent parameter problem for
  21 *					fragments.
  22 *	YOSHIFUJI Hideaki @USAGI:	added sysctl for icmp rate limit.
  23 *	Randy Dunlap and
  24 *	YOSHIFUJI Hideaki @USAGI:	Per-interface statistics support
  25 *	Kazunori MIYAZAWA @USAGI:       change output process to use ip6_append_data
  26 */
  27
  28#define pr_fmt(fmt) "IPv6: " fmt
  29
  30#include <linux/module.h>
  31#include <linux/errno.h>
  32#include <linux/types.h>
  33#include <linux/socket.h>
  34#include <linux/in.h>
  35#include <linux/kernel.h>
  36#include <linux/sockios.h>
  37#include <linux/net.h>
  38#include <linux/skbuff.h>
  39#include <linux/init.h>
  40#include <linux/netfilter.h>
  41#include <linux/slab.h>
  42
  43#ifdef CONFIG_SYSCTL
  44#include <linux/sysctl.h>
  45#endif
  46
  47#include <linux/inet.h>
  48#include <linux/netdevice.h>
  49#include <linux/icmpv6.h>
  50
  51#include <net/ip.h>
  52#include <net/sock.h>
  53
  54#include <net/ipv6.h>
  55#include <net/ip6_checksum.h>
  56#include <net/ping.h>
  57#include <net/protocol.h>
  58#include <net/raw.h>
  59#include <net/rawv6.h>
  60#include <net/seg6.h>
  61#include <net/transp_v6.h>
  62#include <net/ip6_route.h>
  63#include <net/addrconf.h>
  64#include <net/icmp.h>
  65#include <net/xfrm.h>
  66#include <net/inet_common.h>
  67#include <net/dsfield.h>
  68#include <net/l3mdev.h>
  69
  70#include <linux/uaccess.h>
 
  71
  72static DEFINE_PER_CPU(struct sock *, ipv6_icmp_sk);
  73
  74static int icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
  75		       u8 type, u8 code, int offset, __be32 info)
 
 
 
 
  76{
  77	/* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */
  78	struct icmp6hdr *icmp6 = (struct icmp6hdr *) (skb->data + offset);
  79	struct net *net = dev_net(skb->dev);
  80
  81	if (type == ICMPV6_PKT_TOOBIG)
  82		ip6_update_pmtu(skb, net, info, skb->dev->ifindex, 0, sock_net_uid(net, NULL));
  83	else if (type == NDISC_REDIRECT)
  84		ip6_redirect(skb, net, skb->dev->ifindex, 0,
  85			     sock_net_uid(net, NULL));
  86
  87	if (!(type & ICMPV6_INFOMSG_MASK))
  88		if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST)
  89			ping_err(skb, offset, ntohl(info));
  90
  91	return 0;
  92}
  93
  94static int icmpv6_rcv(struct sk_buff *skb);
  95
  96static const struct inet6_protocol icmpv6_protocol = {
  97	.handler	=	icmpv6_rcv,
  98	.err_handler	=	icmpv6_err,
  99	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
 100};
 101
 102/* Called with BH disabled */
 103static struct sock *icmpv6_xmit_lock(struct net *net)
 104{
 105	struct sock *sk;
 106
 107	sk = this_cpu_read(ipv6_icmp_sk);
 
 
 108	if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
 109		/* This can happen if the output path (f.e. SIT or
 110		 * ip6ip6 tunnel) signals dst_link_failure() for an
 111		 * outgoing ICMP6 packet.
 112		 */
 
 113		return NULL;
 114	}
 115	sock_net_set(sk, net);
 116	return sk;
 117}
 118
 119static void icmpv6_xmit_unlock(struct sock *sk)
 120{
 121	sock_net_set(sk, &init_net);
 122	spin_unlock(&sk->sk_lock.slock);
 
 
 
 
 
 
 
 
 123}
 124
 125/*
 126 * Figure out, may we reply to this packet with icmp error.
 127 *
 128 * We do not reply, if:
 129 *	- it was icmp error message.
 130 *	- it is truncated, so that it is known, that protocol is ICMPV6
 131 *	  (i.e. in the middle of some exthdr)
 132 *
 133 *	--ANK (980726)
 134 */
 135
 136static bool is_ineligible(const struct sk_buff *skb)
 137{
 138	int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
 139	int len = skb->len - ptr;
 140	__u8 nexthdr = ipv6_hdr(skb)->nexthdr;
 141	__be16 frag_off;
 142
 143	if (len < 0)
 144		return true;
 145
 146	ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off);
 147	if (ptr < 0)
 148		return false;
 149	if (nexthdr == IPPROTO_ICMPV6) {
 150		u8 _type, *tp;
 151		tp = skb_header_pointer(skb,
 152			ptr+offsetof(struct icmp6hdr, icmp6_type),
 153			sizeof(_type), &_type);
 154
 155		/* Based on RFC 8200, Section 4.5 Fragment Header, return
 156		 * false if this is a fragment packet with no icmp header info.
 157		 */
 158		if (!tp && frag_off != 0)
 159			return false;
 160		else if (!tp || !(*tp & ICMPV6_INFOMSG_MASK))
 161			return true;
 162	}
 163	return false;
 164}
 165
 166static bool icmpv6_mask_allow(struct net *net, int type)
 167{
 168	if (type > ICMPV6_MSG_MAX)
 169		return true;
 170
 171	/* Limit if icmp type is set in ratemask. */
 172	if (!test_bit(type, net->ipv6.sysctl.icmpv6_ratemask))
 173		return true;
 174
 175	return false;
 176}
 177
 178static bool icmpv6_global_allow(struct net *net, int type)
 179{
 180	if (icmpv6_mask_allow(net, type))
 181		return true;
 182
 183	if (icmp_global_allow())
 184		return true;
 185
 186	__ICMP_INC_STATS(net, ICMP_MIB_RATELIMITGLOBAL);
 187	return false;
 188}
 189
 190/*
 191 * Check the ICMP output rate limit
 192 */
 193static bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
 194			       struct flowi6 *fl6)
 195{
 196	struct net *net = sock_net(sk);
 197	struct dst_entry *dst;
 
 198	bool res = false;
 199
 200	if (icmpv6_mask_allow(net, type))
 
 
 
 
 
 201		return true;
 202
 203	/*
 204	 * Look up the output route.
 205	 * XXX: perhaps the expire for routing entries cloned by
 206	 * this lookup should be more aggressive (not longer than timeout).
 207	 */
 208	dst = ip6_route_output(net, sk, fl6);
 209	if (dst->error) {
 210		IP6_INC_STATS(net, ip6_dst_idev(dst),
 211			      IPSTATS_MIB_OUTNOROUTES);
 212	} else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
 213		res = true;
 214	} else {
 215		struct rt6_info *rt = (struct rt6_info *)dst;
 216		int tmo = net->ipv6.sysctl.icmpv6_time;
 217		struct inet_peer *peer;
 218
 219		/* Give more bandwidth to wider prefixes. */
 220		if (rt->rt6i_dst.plen < 128)
 221			tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
 222
 223		peer = inet_getpeer_v6(net->ipv6.peers, &fl6->daddr, 1);
 224		res = inet_peer_xrlim_allow(peer, tmo);
 225		if (peer)
 226			inet_putpeer(peer);
 227	}
 228	if (!res)
 229		__ICMP6_INC_STATS(net, ip6_dst_idev(dst),
 230				  ICMP6_MIB_RATELIMITHOST);
 231	dst_release(dst);
 232	return res;
 233}
 234
 235static bool icmpv6_rt_has_prefsrc(struct sock *sk, u8 type,
 236				  struct flowi6 *fl6)
 237{
 238	struct net *net = sock_net(sk);
 239	struct dst_entry *dst;
 240	bool res = false;
 241
 242	dst = ip6_route_output(net, sk, fl6);
 243	if (!dst->error) {
 244		struct rt6_info *rt = (struct rt6_info *)dst;
 245		struct in6_addr prefsrc;
 246
 247		rt6_get_prefsrc(rt, &prefsrc);
 248		res = !ipv6_addr_any(&prefsrc);
 249	}
 250	dst_release(dst);
 251	return res;
 252}
 253
 254/*
 255 *	an inline helper for the "simple" if statement below
 256 *	checks if parameter problem report is caused by an
 257 *	unrecognized IPv6 option that has the Option Type
 258 *	highest-order two bits set to 10
 259 */
 260
 261static bool opt_unrec(struct sk_buff *skb, __u32 offset)
 262{
 263	u8 _optval, *op;
 264
 265	offset += skb_network_offset(skb);
 266	op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
 267	if (!op)
 268		return true;
 269	return (*op & 0xC0) == 0x80;
 270}
 271
 272void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
 273				struct icmp6hdr *thdr, int len)
 274{
 275	struct sk_buff *skb;
 276	struct icmp6hdr *icmp6h;
 
 277
 278	skb = skb_peek(&sk->sk_write_queue);
 279	if (!skb)
 280		return;
 281
 282	icmp6h = icmp6_hdr(skb);
 283	memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
 284	icmp6h->icmp6_cksum = 0;
 285
 286	if (skb_queue_len(&sk->sk_write_queue) == 1) {
 287		skb->csum = csum_partial(icmp6h,
 288					sizeof(struct icmp6hdr), skb->csum);
 289		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
 290						      &fl6->daddr,
 291						      len, fl6->flowi6_proto,
 292						      skb->csum);
 293	} else {
 294		__wsum tmp_csum = 0;
 295
 296		skb_queue_walk(&sk->sk_write_queue, skb) {
 297			tmp_csum = csum_add(tmp_csum, skb->csum);
 298		}
 299
 300		tmp_csum = csum_partial(icmp6h,
 301					sizeof(struct icmp6hdr), tmp_csum);
 302		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
 303						      &fl6->daddr,
 304						      len, fl6->flowi6_proto,
 305						      tmp_csum);
 306	}
 307	ip6_push_pending_frames(sk);
 
 
 308}
 309
 310struct icmpv6_msg {
 311	struct sk_buff	*skb;
 312	int		offset;
 313	uint8_t		type;
 314};
 315
 316static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
 317{
 318	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
 319	struct sk_buff *org_skb = msg->skb;
 320	__wsum csum;
 321
 322	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
 323				      to, len);
 324	skb->csum = csum_block_add(skb->csum, csum, odd);
 325	if (!(msg->type & ICMPV6_INFOMSG_MASK))
 326		nf_ct_attach(skb, org_skb);
 327	return 0;
 328}
 329
 330#if IS_ENABLED(CONFIG_IPV6_MIP6)
 331static void mip6_addr_swap(struct sk_buff *skb, const struct inet6_skb_parm *opt)
 332{
 333	struct ipv6hdr *iph = ipv6_hdr(skb);
 
 334	struct ipv6_destopt_hao *hao;
 
 335	int off;
 336
 337	if (opt->dsthao) {
 338		off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
 339		if (likely(off >= 0)) {
 340			hao = (struct ipv6_destopt_hao *)
 341					(skb_network_header(skb) + off);
 342			swap(iph->saddr, hao->addr);
 
 
 343		}
 344	}
 345}
 346#else
 347static inline void mip6_addr_swap(struct sk_buff *skb, const struct inet6_skb_parm *opt) {}
 348#endif
 349
 350static struct dst_entry *icmpv6_route_lookup(struct net *net,
 351					     struct sk_buff *skb,
 352					     struct sock *sk,
 353					     struct flowi6 *fl6)
 354{
 355	struct dst_entry *dst, *dst2;
 356	struct flowi6 fl2;
 357	int err;
 358
 359	err = ip6_dst_lookup(net, sk, &dst, fl6);
 360	if (err)
 361		return ERR_PTR(err);
 362
 363	/*
 364	 * We won't send icmp if the destination is known
 365	 * anycast unless we need to treat anycast as unicast.
 366	 */
 367	if (!READ_ONCE(net->ipv6.sysctl.icmpv6_error_anycast_as_unicast) &&
 368	    ipv6_anycast_destination(dst, &fl6->daddr)) {
 369		net_dbg_ratelimited("icmp6_send: acast source\n");
 370		dst_release(dst);
 371		return ERR_PTR(-EINVAL);
 372	}
 373
 374	/* No need to clone since we're just using its address. */
 375	dst2 = dst;
 376
 377	dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
 378	if (!IS_ERR(dst)) {
 379		if (dst != dst2)
 380			return dst;
 381	} else {
 382		if (PTR_ERR(dst) == -EPERM)
 383			dst = NULL;
 384		else
 385			return dst;
 386	}
 387
 388	err = xfrm_decode_session_reverse(net, skb, flowi6_to_flowi(&fl2), AF_INET6);
 389	if (err)
 390		goto relookup_failed;
 391
 392	err = ip6_dst_lookup(net, sk, &dst2, &fl2);
 393	if (err)
 394		goto relookup_failed;
 395
 396	dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
 397	if (!IS_ERR(dst2)) {
 398		dst_release(dst);
 399		dst = dst2;
 400	} else {
 401		err = PTR_ERR(dst2);
 402		if (err == -EPERM) {
 403			dst_release(dst);
 404			return dst2;
 405		} else
 406			goto relookup_failed;
 407	}
 408
 409relookup_failed:
 410	if (dst)
 411		return dst;
 412	return ERR_PTR(err);
 413}
 414
 415static struct net_device *icmp6_dev(const struct sk_buff *skb)
 416{
 417	struct net_device *dev = skb->dev;
 418
 419	/* for local traffic to local address, skb dev is the loopback
 420	 * device. Check if there is a dst attached to the skb and if so
 421	 * get the real device index. Same is needed for replies to a link
 422	 * local address on a device enslaved to an L3 master device
 423	 */
 424	if (unlikely(dev->ifindex == LOOPBACK_IFINDEX || netif_is_l3_master(skb->dev))) {
 425		const struct rt6_info *rt6 = skb_rt6_info(skb);
 426
 427		/* The destination could be an external IP in Ext Hdr (SRv6, RPL, etc.),
 428		 * and ip6_null_entry could be set to skb if no route is found.
 429		 */
 430		if (rt6 && rt6->rt6i_idev)
 431			dev = rt6->rt6i_idev->dev;
 432	}
 433
 434	return dev;
 435}
 436
 437static int icmp6_iif(const struct sk_buff *skb)
 438{
 439	return icmp6_dev(skb)->ifindex;
 440}
 441
 442/*
 443 *	Send an ICMP message in response to a packet in error
 444 */
 445void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info,
 446		const struct in6_addr *force_saddr,
 447		const struct inet6_skb_parm *parm)
 448{
 
 449	struct inet6_dev *idev = NULL;
 450	struct ipv6hdr *hdr = ipv6_hdr(skb);
 451	struct sock *sk;
 452	struct net *net;
 453	struct ipv6_pinfo *np;
 454	const struct in6_addr *saddr = NULL;
 455	struct dst_entry *dst;
 456	struct icmp6hdr tmp_hdr;
 457	struct flowi6 fl6;
 458	struct icmpv6_msg msg;
 459	struct ipcm6_cookie ipc6;
 460	int iif = 0;
 461	int addr_type = 0;
 462	int len;
 463	u32 mark;
 
 464
 465	if ((u8 *)hdr < skb->head ||
 466	    (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb))
 467		return;
 468
 469	if (!skb->dev)
 470		return;
 471	net = dev_net(skb->dev);
 472	mark = IP6_REPLY_MARK(net, skb->mark);
 473	/*
 474	 *	Make sure we respect the rules
 475	 *	i.e. RFC 1885 2.4(e)
 476	 *	Rule (e.1) is enforced by not using icmp6_send
 477	 *	in any code that processes icmp errors.
 478	 */
 479	addr_type = ipv6_addr_type(&hdr->daddr);
 480
 481	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) ||
 482	    ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr))
 483		saddr = &hdr->daddr;
 484
 485	/*
 486	 *	Dest addr check
 487	 */
 488
 489	if (addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST) {
 490		if (type != ICMPV6_PKT_TOOBIG &&
 491		    !(type == ICMPV6_PARAMPROB &&
 492		      code == ICMPV6_UNK_OPTION &&
 493		      (opt_unrec(skb, info))))
 494			return;
 495
 496		saddr = NULL;
 497	}
 498
 499	addr_type = ipv6_addr_type(&hdr->saddr);
 500
 501	/*
 502	 *	Source addr check
 503	 */
 504
 505	if (__ipv6_addr_needs_scope_id(addr_type)) {
 506		iif = icmp6_iif(skb);
 507	} else {
 508		/*
 509		 * The source device is used for looking up which routing table
 510		 * to use for sending an ICMP error.
 511		 */
 512		iif = l3mdev_master_ifindex(skb->dev);
 513	}
 514
 515	/*
 516	 *	Must not send error if the source does not uniquely
 517	 *	identify a single node (RFC2463 Section 2.4).
 518	 *	We check unspecified / multicast addresses here,
 519	 *	and anycast addresses will be checked later.
 520	 */
 521	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
 522		net_dbg_ratelimited("icmp6_send: addr_any/mcast source [%pI6c > %pI6c]\n",
 523				    &hdr->saddr, &hdr->daddr);
 524		return;
 525	}
 526
 527	/*
 528	 *	Never answer to a ICMP packet.
 529	 */
 530	if (is_ineligible(skb)) {
 531		net_dbg_ratelimited("icmp6_send: no reply to icmp error [%pI6c > %pI6c]\n",
 532				    &hdr->saddr, &hdr->daddr);
 533		return;
 534	}
 535
 536	/* Needed by both icmp_global_allow and icmpv6_xmit_lock */
 537	local_bh_disable();
 538
 539	/* Check global sysctl_icmp_msgs_per_sec ratelimit */
 540	if (!(skb->dev->flags & IFF_LOOPBACK) && !icmpv6_global_allow(net, type))
 541		goto out_bh_enable;
 542
 543	mip6_addr_swap(skb, parm);
 544
 545	sk = icmpv6_xmit_lock(net);
 546	if (!sk)
 547		goto out_bh_enable;
 548
 549	memset(&fl6, 0, sizeof(fl6));
 550	fl6.flowi6_proto = IPPROTO_ICMPV6;
 551	fl6.daddr = hdr->saddr;
 552	if (force_saddr)
 553		saddr = force_saddr;
 554	if (saddr) {
 555		fl6.saddr = *saddr;
 556	} else if (!icmpv6_rt_has_prefsrc(sk, type, &fl6)) {
 557		/* select a more meaningful saddr from input if */
 558		struct net_device *in_netdev;
 559
 560		in_netdev = dev_get_by_index(net, parm->iif);
 561		if (in_netdev) {
 562			ipv6_dev_get_saddr(net, in_netdev, &fl6.daddr,
 563					   inet6_sk(sk)->srcprefs,
 564					   &fl6.saddr);
 565			dev_put(in_netdev);
 566		}
 567	}
 568	fl6.flowi6_mark = mark;
 569	fl6.flowi6_oif = iif;
 570	fl6.fl6_icmp_type = type;
 571	fl6.fl6_icmp_code = code;
 572	fl6.flowi6_uid = sock_net_uid(net, NULL);
 573	fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, NULL);
 574	security_skb_classify_flow(skb, flowi6_to_flowi_common(&fl6));
 575
 
 
 
 576	np = inet6_sk(sk);
 577
 578	if (!icmpv6_xrlim_allow(sk, type, &fl6))
 579		goto out;
 580
 581	tmp_hdr.icmp6_type = type;
 582	tmp_hdr.icmp6_code = code;
 583	tmp_hdr.icmp6_cksum = 0;
 584	tmp_hdr.icmp6_pointer = htonl(info);
 585
 586	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
 587		fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
 588	else if (!fl6.flowi6_oif)
 589		fl6.flowi6_oif = READ_ONCE(np->ucast_oif);
 590
 591	ipcm6_init_sk(&ipc6, sk);
 592	ipc6.sockc.mark = mark;
 593	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
 594
 595	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
 596	if (IS_ERR(dst))
 597		goto out;
 598
 599	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
 
 
 
 
 
 600
 601	msg.skb = skb;
 602	msg.offset = skb_network_offset(skb);
 603	msg.type = type;
 604
 605	len = skb->len - msg.offset;
 606	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(struct icmp6hdr));
 607	if (len < 0) {
 608		net_dbg_ratelimited("icmp: len problem [%pI6c > %pI6c]\n",
 609				    &hdr->saddr, &hdr->daddr);
 610		goto out_dst_release;
 611	}
 612
 613	rcu_read_lock();
 614	idev = __in6_dev_get(skb->dev);
 615
 616	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
 617			    len + sizeof(struct icmp6hdr),
 618			    sizeof(struct icmp6hdr),
 619			    &ipc6, &fl6, (struct rt6_info *)dst,
 620			    MSG_DONTWAIT)) {
 621		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
 
 622		ip6_flush_pending_frames(sk);
 623	} else {
 624		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
 625					   len + sizeof(struct icmp6hdr));
 626	}
 627	rcu_read_unlock();
 
 
 
 
 628out_dst_release:
 629	dst_release(dst);
 630out:
 631	icmpv6_xmit_unlock(sk);
 632out_bh_enable:
 633	local_bh_enable();
 634}
 635EXPORT_SYMBOL(icmp6_send);
 636
 637/* Slightly more convenient version of icmp6_send with drop reasons.
 638 */
 639void icmpv6_param_prob_reason(struct sk_buff *skb, u8 code, int pos,
 640			      enum skb_drop_reason reason)
 641{
 642	icmp6_send(skb, ICMPV6_PARAMPROB, code, pos, NULL, IP6CB(skb));
 643	kfree_skb_reason(skb, reason);
 644}
 645
 646/* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
 647 * if sufficient data bytes are available
 648 * @nhs is the size of the tunnel header(s) :
 649 *  Either an IPv4 header for SIT encap
 650 *         an IPv4 header + GRE header for GRE encap
 651 */
 652int ip6_err_gen_icmpv6_unreach(struct sk_buff *skb, int nhs, int type,
 653			       unsigned int data_len)
 654{
 655	struct in6_addr temp_saddr;
 656	struct rt6_info *rt;
 657	struct sk_buff *skb2;
 658	u32 info = 0;
 659
 660	if (!pskb_may_pull(skb, nhs + sizeof(struct ipv6hdr) + 8))
 661		return 1;
 662
 663	/* RFC 4884 (partial) support for ICMP extensions */
 664	if (data_len < 128 || (data_len & 7) || skb->len < data_len)
 665		data_len = 0;
 666
 667	skb2 = data_len ? skb_copy(skb, GFP_ATOMIC) : skb_clone(skb, GFP_ATOMIC);
 668
 669	if (!skb2)
 670		return 1;
 671
 672	skb_dst_drop(skb2);
 673	skb_pull(skb2, nhs);
 674	skb_reset_network_header(skb2);
 675
 676	rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0,
 677			skb, 0);
 678
 679	if (rt && rt->dst.dev)
 680		skb2->dev = rt->dst.dev;
 681
 682	ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, &temp_saddr);
 683
 684	if (data_len) {
 685		/* RFC 4884 (partial) support :
 686		 * insert 0 padding at the end, before the extensions
 687		 */
 688		__skb_push(skb2, nhs);
 689		skb_reset_network_header(skb2);
 690		memmove(skb2->data, skb2->data + nhs, data_len - nhs);
 691		memset(skb2->data + data_len - nhs, 0, nhs);
 692		/* RFC 4884 4.5 : Length is measured in 64-bit words,
 693		 * and stored in reserved[0]
 694		 */
 695		info = (data_len/8) << 24;
 696	}
 697	if (type == ICMP_TIME_EXCEEDED)
 698		icmp6_send(skb2, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
 699			   info, &temp_saddr, IP6CB(skb2));
 700	else
 701		icmp6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH,
 702			   info, &temp_saddr, IP6CB(skb2));
 703	if (rt)
 704		ip6_rt_put(rt);
 705
 706	kfree_skb(skb2);
 707
 708	return 0;
 709}
 710EXPORT_SYMBOL(ip6_err_gen_icmpv6_unreach);
 711
 712static enum skb_drop_reason icmpv6_echo_reply(struct sk_buff *skb)
 713{
 714	struct net *net = dev_net(skb->dev);
 715	struct sock *sk;
 716	struct inet6_dev *idev;
 717	struct ipv6_pinfo *np;
 718	const struct in6_addr *saddr = NULL;
 719	struct icmp6hdr *icmph = icmp6_hdr(skb);
 720	struct icmp6hdr tmp_hdr;
 721	struct flowi6 fl6;
 722	struct icmpv6_msg msg;
 723	struct dst_entry *dst;
 724	struct ipcm6_cookie ipc6;
 725	u32 mark = IP6_REPLY_MARK(net, skb->mark);
 726	SKB_DR(reason);
 727	bool acast;
 728	u8 type;
 729
 730	if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) &&
 731	    net->ipv6.sysctl.icmpv6_echo_ignore_multicast)
 732		return reason;
 733
 734	saddr = &ipv6_hdr(skb)->daddr;
 735
 736	acast = ipv6_anycast_destination(skb_dst(skb), saddr);
 737	if (acast && net->ipv6.sysctl.icmpv6_echo_ignore_anycast)
 738		return reason;
 739
 740	if (!ipv6_unicast_destination(skb) &&
 741	    !(net->ipv6.sysctl.anycast_src_echo_reply && acast))
 742		saddr = NULL;
 743
 744	if (icmph->icmp6_type == ICMPV6_EXT_ECHO_REQUEST)
 745		type = ICMPV6_EXT_ECHO_REPLY;
 746	else
 747		type = ICMPV6_ECHO_REPLY;
 748
 749	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
 750	tmp_hdr.icmp6_type = type;
 751
 752	memset(&fl6, 0, sizeof(fl6));
 753	if (net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ICMPV6_ECHO_REPLIES)
 754		fl6.flowlabel = ip6_flowlabel(ipv6_hdr(skb));
 755
 756	fl6.flowi6_proto = IPPROTO_ICMPV6;
 757	fl6.daddr = ipv6_hdr(skb)->saddr;
 758	if (saddr)
 759		fl6.saddr = *saddr;
 760	fl6.flowi6_oif = icmp6_iif(skb);
 761	fl6.fl6_icmp_type = type;
 762	fl6.flowi6_mark = mark;
 763	fl6.flowi6_uid = sock_net_uid(net, NULL);
 764	security_skb_classify_flow(skb, flowi6_to_flowi_common(&fl6));
 765
 766	local_bh_disable();
 767	sk = icmpv6_xmit_lock(net);
 768	if (!sk)
 769		goto out_bh_enable;
 770	np = inet6_sk(sk);
 771
 772	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
 773		fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
 774	else if (!fl6.flowi6_oif)
 775		fl6.flowi6_oif = READ_ONCE(np->ucast_oif);
 776
 777	if (ip6_dst_lookup(net, sk, &dst, &fl6))
 
 778		goto out;
 779	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
 780	if (IS_ERR(dst))
 781		goto out;
 782
 783	/* Check the ratelimit */
 784	if ((!(skb->dev->flags & IFF_LOOPBACK) && !icmpv6_global_allow(net, ICMPV6_ECHO_REPLY)) ||
 785	    !icmpv6_xrlim_allow(sk, ICMPV6_ECHO_REPLY, &fl6))
 786		goto out_dst_release;
 
 
 787
 788	idev = __in6_dev_get(skb->dev);
 789
 790	msg.skb = skb;
 791	msg.offset = 0;
 792	msg.type = type;
 793
 794	ipcm6_init_sk(&ipc6, sk);
 795	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
 796	ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb));
 797	ipc6.sockc.mark = mark;
 798
 799	if (icmph->icmp6_type == ICMPV6_EXT_ECHO_REQUEST)
 800		if (!icmp_build_probe(skb, (struct icmphdr *)&tmp_hdr))
 801			goto out_dst_release;
 802
 803	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
 804			    skb->len + sizeof(struct icmp6hdr),
 805			    sizeof(struct icmp6hdr), &ipc6, &fl6,
 806			    (struct rt6_info *)dst, MSG_DONTWAIT)) {
 807		__ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
 808		ip6_flush_pending_frames(sk);
 809	} else {
 810		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
 811					   skb->len + sizeof(struct icmp6hdr));
 812		reason = SKB_CONSUMED;
 813	}
 814out_dst_release:
 
 
 
 
 815	dst_release(dst);
 816out:
 817	icmpv6_xmit_unlock(sk);
 818out_bh_enable:
 819	local_bh_enable();
 820	return reason;
 821}
 822
 823enum skb_drop_reason icmpv6_notify(struct sk_buff *skb, u8 type,
 824				   u8 code, __be32 info)
 825{
 826	struct inet6_skb_parm *opt = IP6CB(skb);
 827	struct net *net = dev_net(skb->dev);
 828	const struct inet6_protocol *ipprot;
 829	enum skb_drop_reason reason;
 830	int inner_offset;
 831	__be16 frag_off;
 832	u8 nexthdr;
 833
 834	reason = pskb_may_pull_reason(skb, sizeof(struct ipv6hdr));
 835	if (reason != SKB_NOT_DROPPED_YET)
 836		goto out;
 837
 838	seg6_icmp_srh(skb, opt);
 839
 840	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
 841	if (ipv6_ext_hdr(nexthdr)) {
 842		/* now skip over extension headers */
 843		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
 844						&nexthdr, &frag_off);
 845		if (inner_offset < 0) {
 846			SKB_DR_SET(reason, IPV6_BAD_EXTHDR);
 847			goto out;
 848		}
 849	} else {
 850		inner_offset = sizeof(struct ipv6hdr);
 851	}
 852
 853	/* Checkin header including 8 bytes of inner protocol header. */
 854	reason = pskb_may_pull_reason(skb, inner_offset + 8);
 855	if (reason != SKB_NOT_DROPPED_YET)
 856		goto out;
 857
 858	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
 859	   Without this we will not able f.e. to make source routed
 860	   pmtu discovery.
 861	   Corresponding argument (opt) to notifiers is already added.
 862	   --ANK (980726)
 863	 */
 864
 865	ipprot = rcu_dereference(inet6_protos[nexthdr]);
 
 
 
 866	if (ipprot && ipprot->err_handler)
 867		ipprot->err_handler(skb, opt, type, code, inner_offset, info);
 
 868
 869	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
 870	return SKB_CONSUMED;
 871
 872out:
 873	__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
 874	return reason;
 875}
 876
 877/*
 878 *	Handle icmp messages
 879 */
 880
 881static int icmpv6_rcv(struct sk_buff *skb)
 882{
 883	enum skb_drop_reason reason = SKB_DROP_REASON_NOT_SPECIFIED;
 884	struct net *net = dev_net(skb->dev);
 885	struct net_device *dev = icmp6_dev(skb);
 886	struct inet6_dev *idev = __in6_dev_get(dev);
 887	const struct in6_addr *saddr, *daddr;
 
 888	struct icmp6hdr *hdr;
 889	u8 type;
 890
 891	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 892		struct sec_path *sp = skb_sec_path(skb);
 893		int nh;
 894
 895		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
 896				 XFRM_STATE_ICMP)) {
 897			reason = SKB_DROP_REASON_XFRM_POLICY;
 898			goto drop_no_count;
 899		}
 900
 901		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
 902			goto drop_no_count;
 903
 904		nh = skb_network_offset(skb);
 905		skb_set_network_header(skb, sizeof(*hdr));
 906
 907		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN,
 908						skb)) {
 909			reason = SKB_DROP_REASON_XFRM_POLICY;
 910			goto drop_no_count;
 911		}
 912
 913		skb_set_network_header(skb, nh);
 914	}
 915
 916	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS);
 917
 918	saddr = &ipv6_hdr(skb)->saddr;
 919	daddr = &ipv6_hdr(skb)->daddr;
 920
 921	if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) {
 922		net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n",
 923				    saddr, daddr);
 924		goto csum_error;
 
 
 
 
 
 
 
 
 
 
 
 925	}
 926
 927	if (!pskb_pull(skb, sizeof(*hdr)))
 928		goto discard_it;
 929
 930	hdr = icmp6_hdr(skb);
 931
 932	type = hdr->icmp6_type;
 933
 934	ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type);
 935
 936	switch (type) {
 937	case ICMPV6_ECHO_REQUEST:
 938		if (!net->ipv6.sysctl.icmpv6_echo_ignore_all)
 939			reason = icmpv6_echo_reply(skb);
 940		break;
 941	case ICMPV6_EXT_ECHO_REQUEST:
 942		if (!net->ipv6.sysctl.icmpv6_echo_ignore_all &&
 943		    READ_ONCE(net->ipv4.sysctl_icmp_echo_enable_probe))
 944			reason = icmpv6_echo_reply(skb);
 945		break;
 946
 947	case ICMPV6_ECHO_REPLY:
 948		reason = ping_rcv(skb);
 949		break;
 950
 951	case ICMPV6_EXT_ECHO_REPLY:
 952		reason = ping_rcv(skb);
 953		break;
 954
 955	case ICMPV6_PKT_TOOBIG:
 956		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
 957		   standard destination cache. Seems, only "advanced"
 958		   destination cache will allow to solve this problem
 959		   --ANK (980726)
 960		 */
 961		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 962			goto discard_it;
 963		hdr = icmp6_hdr(skb);
 
 
 
 
 
 
 
 964
 965		/* to notify */
 966		fallthrough;
 967	case ICMPV6_DEST_UNREACH:
 968	case ICMPV6_TIME_EXCEED:
 969	case ICMPV6_PARAMPROB:
 970		reason = icmpv6_notify(skb, type, hdr->icmp6_code,
 971				       hdr->icmp6_mtu);
 972		break;
 973
 974	case NDISC_ROUTER_SOLICITATION:
 975	case NDISC_ROUTER_ADVERTISEMENT:
 976	case NDISC_NEIGHBOUR_SOLICITATION:
 977	case NDISC_NEIGHBOUR_ADVERTISEMENT:
 978	case NDISC_REDIRECT:
 979		reason = ndisc_rcv(skb);
 980		break;
 981
 982	case ICMPV6_MGM_QUERY:
 983		igmp6_event_query(skb);
 984		return 0;
 985
 986	case ICMPV6_MGM_REPORT:
 987		igmp6_event_report(skb);
 988		return 0;
 989
 990	case ICMPV6_MGM_REDUCTION:
 991	case ICMPV6_NI_QUERY:
 992	case ICMPV6_NI_REPLY:
 993	case ICMPV6_MLD2_REPORT:
 994	case ICMPV6_DHAAD_REQUEST:
 995	case ICMPV6_DHAAD_REPLY:
 996	case ICMPV6_MOBILE_PREFIX_SOL:
 997	case ICMPV6_MOBILE_PREFIX_ADV:
 998		break;
 999
1000	default:
 
 
1001		/* informational */
1002		if (type & ICMPV6_INFOMSG_MASK)
1003			break;
1004
1005		net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n",
1006				    saddr, daddr);
1007
1008		/*
1009		 * error of unknown type.
1010		 * must pass to upper level
1011		 */
1012
1013		reason = icmpv6_notify(skb, type, hdr->icmp6_code,
1014				       hdr->icmp6_mtu);
1015	}
1016
1017	/* until the v6 path can be better sorted assume failure and
1018	 * preserve the status quo behaviour for the rest of the paths to here
1019	 */
1020	if (reason)
1021		kfree_skb_reason(skb, reason);
1022	else
1023		consume_skb(skb);
1024
1025	return 0;
1026
1027csum_error:
1028	reason = SKB_DROP_REASON_ICMP_CSUM;
1029	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
1030discard_it:
1031	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS);
1032drop_no_count:
1033	kfree_skb_reason(skb, reason);
1034	return 0;
1035}
1036
1037void icmpv6_flow_init(const struct sock *sk, struct flowi6 *fl6, u8 type,
 
1038		      const struct in6_addr *saddr,
1039		      const struct in6_addr *daddr, int oif)
 
1040{
1041	memset(fl6, 0, sizeof(*fl6));
1042	fl6->saddr = *saddr;
1043	fl6->daddr = *daddr;
1044	fl6->flowi6_proto	= IPPROTO_ICMPV6;
1045	fl6->fl6_icmp_type	= type;
1046	fl6->fl6_icmp_code	= 0;
1047	fl6->flowi6_oif		= oif;
1048	security_sk_classify_flow(sk, flowi6_to_flowi_common(fl6));
1049}
1050
1051int __init icmpv6_init(void)
 
 
 
 
 
1052{
1053	struct sock *sk;
1054	int err, i;
 
 
 
 
 
1055
1056	for_each_possible_cpu(i) {
1057		err = inet_ctl_sock_create(&sk, PF_INET6,
1058					   SOCK_RAW, IPPROTO_ICMPV6, &init_net);
1059		if (err < 0) {
1060			pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
 
 
1061			       err);
1062			return err;
1063		}
1064
1065		per_cpu(ipv6_icmp_sk, i) = sk;
 
 
 
 
 
 
 
 
 
1066
1067		/* Enough space for 2 64K ICMP packets, including
1068		 * sk_buff struct overhead.
1069		 */
1070		sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
 
1071	}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1072
1073	err = -EAGAIN;
1074	if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
1075		goto fail;
1076
1077	err = inet6_register_icmp_sender(icmp6_send);
1078	if (err)
1079		goto sender_reg_err;
1080	return 0;
1081
1082sender_reg_err:
1083	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
1084fail:
1085	pr_err("Failed to register ICMP6 protocol\n");
 
1086	return err;
1087}
1088
1089void icmpv6_cleanup(void)
1090{
1091	inet6_unregister_icmp_sender(icmp6_send);
1092	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
1093}
1094
1095
1096static const struct icmp6_err {
1097	int err;
1098	int fatal;
1099} tab_unreach[] = {
1100	{	/* NOROUTE */
1101		.err	= ENETUNREACH,
1102		.fatal	= 0,
1103	},
1104	{	/* ADM_PROHIBITED */
1105		.err	= EACCES,
1106		.fatal	= 1,
1107	},
1108	{	/* Was NOT_NEIGHBOUR, now reserved */
1109		.err	= EHOSTUNREACH,
1110		.fatal	= 0,
1111	},
1112	{	/* ADDR_UNREACH	*/
1113		.err	= EHOSTUNREACH,
1114		.fatal	= 0,
1115	},
1116	{	/* PORT_UNREACH	*/
1117		.err	= ECONNREFUSED,
1118		.fatal	= 1,
1119	},
1120	{	/* POLICY_FAIL */
1121		.err	= EACCES,
1122		.fatal	= 1,
1123	},
1124	{	/* REJECT_ROUTE	*/
1125		.err	= EACCES,
1126		.fatal	= 1,
1127	},
1128};
1129
1130int icmpv6_err_convert(u8 type, u8 code, int *err)
1131{
1132	int fatal = 0;
1133
1134	*err = EPROTO;
1135
1136	switch (type) {
1137	case ICMPV6_DEST_UNREACH:
1138		fatal = 1;
1139		if (code < ARRAY_SIZE(tab_unreach)) {
1140			*err  = tab_unreach[code].err;
1141			fatal = tab_unreach[code].fatal;
1142		}
1143		break;
1144
1145	case ICMPV6_PKT_TOOBIG:
1146		*err = EMSGSIZE;
1147		break;
1148
1149	case ICMPV6_PARAMPROB:
1150		*err = EPROTO;
1151		fatal = 1;
1152		break;
1153
1154	case ICMPV6_TIME_EXCEED:
1155		*err = EHOSTUNREACH;
1156		break;
1157	}
1158
1159	return fatal;
1160}
 
1161EXPORT_SYMBOL(icmpv6_err_convert);
1162
1163#ifdef CONFIG_SYSCTL
1164static struct ctl_table ipv6_icmp_table_template[] = {
1165	{
1166		.procname	= "ratelimit",
1167		.data		= &init_net.ipv6.sysctl.icmpv6_time,
1168		.maxlen		= sizeof(int),
1169		.mode		= 0644,
1170		.proc_handler	= proc_dointvec_ms_jiffies,
1171	},
1172	{
1173		.procname	= "echo_ignore_all",
1174		.data		= &init_net.ipv6.sysctl.icmpv6_echo_ignore_all,
1175		.maxlen		= sizeof(u8),
1176		.mode		= 0644,
1177		.proc_handler = proc_dou8vec_minmax,
1178	},
1179	{
1180		.procname	= "echo_ignore_multicast",
1181		.data		= &init_net.ipv6.sysctl.icmpv6_echo_ignore_multicast,
1182		.maxlen		= sizeof(u8),
1183		.mode		= 0644,
1184		.proc_handler = proc_dou8vec_minmax,
1185	},
1186	{
1187		.procname	= "echo_ignore_anycast",
1188		.data		= &init_net.ipv6.sysctl.icmpv6_echo_ignore_anycast,
1189		.maxlen		= sizeof(u8),
1190		.mode		= 0644,
1191		.proc_handler = proc_dou8vec_minmax,
1192	},
1193	{
1194		.procname	= "ratemask",
1195		.data		= &init_net.ipv6.sysctl.icmpv6_ratemask_ptr,
1196		.maxlen		= ICMPV6_MSG_MAX + 1,
1197		.mode		= 0644,
1198		.proc_handler = proc_do_large_bitmap,
1199	},
1200	{
1201		.procname	= "error_anycast_as_unicast",
1202		.data		= &init_net.ipv6.sysctl.icmpv6_error_anycast_as_unicast,
1203		.maxlen		= sizeof(u8),
1204		.mode		= 0644,
1205		.proc_handler	= proc_dou8vec_minmax,
1206		.extra1		= SYSCTL_ZERO,
1207		.extra2		= SYSCTL_ONE,
1208	},
1209	{ },
1210};
1211
1212struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1213{
1214	struct ctl_table *table;
1215
1216	table = kmemdup(ipv6_icmp_table_template,
1217			sizeof(ipv6_icmp_table_template),
1218			GFP_KERNEL);
1219
1220	if (table) {
1221		table[0].data = &net->ipv6.sysctl.icmpv6_time;
1222		table[1].data = &net->ipv6.sysctl.icmpv6_echo_ignore_all;
1223		table[2].data = &net->ipv6.sysctl.icmpv6_echo_ignore_multicast;
1224		table[3].data = &net->ipv6.sysctl.icmpv6_echo_ignore_anycast;
1225		table[4].data = &net->ipv6.sysctl.icmpv6_ratemask_ptr;
1226		table[5].data = &net->ipv6.sysctl.icmpv6_error_anycast_as_unicast;
1227	}
1228	return table;
1229}
1230
1231size_t ipv6_icmp_sysctl_table_size(void)
1232{
1233	return ARRAY_SIZE(ipv6_icmp_table_template);
1234}
1235#endif
v3.1
 
  1/*
  2 *	Internet Control Message Protocol (ICMPv6)
  3 *	Linux INET6 implementation
  4 *
  5 *	Authors:
  6 *	Pedro Roque		<roque@di.fc.ul.pt>
  7 *
  8 *	Based on net/ipv4/icmp.c
  9 *
 10 *	RFC 1885
 11 *
 12 *	This program is free software; you can redistribute it and/or
 13 *      modify it under the terms of the GNU General Public License
 14 *      as published by the Free Software Foundation; either version
 15 *      2 of the License, or (at your option) any later version.
 16 */
 17
 18/*
 19 *	Changes:
 20 *
 21 *	Andi Kleen		:	exception handling
 22 *	Andi Kleen			add rate limits. never reply to a icmp.
 23 *					add more length checks and other fixes.
 24 *	yoshfuji		:	ensure to sent parameter problem for
 25 *					fragments.
 26 *	YOSHIFUJI Hideaki @USAGI:	added sysctl for icmp rate limit.
 27 *	Randy Dunlap and
 28 *	YOSHIFUJI Hideaki @USAGI:	Per-interface statistics support
 29 *	Kazunori MIYAZAWA @USAGI:       change output process to use ip6_append_data
 30 */
 31
 
 
 32#include <linux/module.h>
 33#include <linux/errno.h>
 34#include <linux/types.h>
 35#include <linux/socket.h>
 36#include <linux/in.h>
 37#include <linux/kernel.h>
 38#include <linux/sockios.h>
 39#include <linux/net.h>
 40#include <linux/skbuff.h>
 41#include <linux/init.h>
 42#include <linux/netfilter.h>
 43#include <linux/slab.h>
 44
 45#ifdef CONFIG_SYSCTL
 46#include <linux/sysctl.h>
 47#endif
 48
 49#include <linux/inet.h>
 50#include <linux/netdevice.h>
 51#include <linux/icmpv6.h>
 52
 53#include <net/ip.h>
 54#include <net/sock.h>
 55
 56#include <net/ipv6.h>
 57#include <net/ip6_checksum.h>
 
 58#include <net/protocol.h>
 59#include <net/raw.h>
 60#include <net/rawv6.h>
 
 61#include <net/transp_v6.h>
 62#include <net/ip6_route.h>
 63#include <net/addrconf.h>
 64#include <net/icmp.h>
 65#include <net/xfrm.h>
 66#include <net/inet_common.h>
 
 
 67
 68#include <asm/uaccess.h>
 69#include <asm/system.h>
 70
 71/*
 72 *	The ICMP socket(s). This is the most convenient way to flow control
 73 *	our ICMP output as well as maintain a clean interface throughout
 74 *	all layers. All Socketless IP sends will soon be gone.
 75 *
 76 *	On SMP we have one ICMP socket per-cpu.
 77 */
 78static inline struct sock *icmpv6_sk(struct net *net)
 79{
 80	return net->ipv6.icmp_sk[smp_processor_id()];
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 81}
 82
 83static int icmpv6_rcv(struct sk_buff *skb);
 84
 85static const struct inet6_protocol icmpv6_protocol = {
 86	.handler	=	icmpv6_rcv,
 
 87	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
 88};
 89
 90static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
 
 91{
 92	struct sock *sk;
 93
 94	local_bh_disable();
 95
 96	sk = icmpv6_sk(net);
 97	if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
 98		/* This can happen if the output path (f.e. SIT or
 99		 * ip6ip6 tunnel) signals dst_link_failure() for an
100		 * outgoing ICMP6 packet.
101		 */
102		local_bh_enable();
103		return NULL;
104	}
 
105	return sk;
106}
107
108static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
109{
110	spin_unlock_bh(&sk->sk_lock.slock);
111}
112
113/*
114 * Slightly more convenient version of icmpv6_send.
115 */
116void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
117{
118	icmpv6_send(skb, ICMPV6_PARAMPROB, code, pos);
119	kfree_skb(skb);
120}
121
122/*
123 * Figure out, may we reply to this packet with icmp error.
124 *
125 * We do not reply, if:
126 *	- it was icmp error message.
127 *	- it is truncated, so that it is known, that protocol is ICMPV6
128 *	  (i.e. in the middle of some exthdr)
129 *
130 *	--ANK (980726)
131 */
132
133static int is_ineligible(struct sk_buff *skb)
134{
135	int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
136	int len = skb->len - ptr;
137	__u8 nexthdr = ipv6_hdr(skb)->nexthdr;
 
138
139	if (len < 0)
140		return 1;
141
142	ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr);
143	if (ptr < 0)
144		return 0;
145	if (nexthdr == IPPROTO_ICMPV6) {
146		u8 _type, *tp;
147		tp = skb_header_pointer(skb,
148			ptr+offsetof(struct icmp6hdr, icmp6_type),
149			sizeof(_type), &_type);
150		if (tp == NULL ||
151		    !(*tp & ICMPV6_INFOMSG_MASK))
152			return 1;
 
 
 
 
 
153	}
154	return 0;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
155}
156
157/*
158 * Check the ICMP output rate limit
159 */
160static inline bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
161				      struct flowi6 *fl6)
162{
 
163	struct dst_entry *dst;
164	struct net *net = sock_net(sk);
165	bool res = false;
166
167	/* Informational messages are not limited. */
168	if (type & ICMPV6_INFOMSG_MASK)
169		return true;
170
171	/* Do not limit pmtu discovery, it would break it. */
172	if (type == ICMPV6_PKT_TOOBIG)
173		return true;
174
175	/*
176	 * Look up the output route.
177	 * XXX: perhaps the expire for routing entries cloned by
178	 * this lookup should be more aggressive (not longer than timeout).
179	 */
180	dst = ip6_route_output(net, sk, fl6);
181	if (dst->error) {
182		IP6_INC_STATS(net, ip6_dst_idev(dst),
183			      IPSTATS_MIB_OUTNOROUTES);
184	} else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
185		res = true;
186	} else {
187		struct rt6_info *rt = (struct rt6_info *)dst;
188		int tmo = net->ipv6.sysctl.icmpv6_time;
 
189
190		/* Give more bandwidth to wider prefixes. */
191		if (rt->rt6i_dst.plen < 128)
192			tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
193
194		if (!rt->rt6i_peer)
195			rt6_bind_peer(rt, 1);
196		res = inet_peer_xrlim_allow(rt->rt6i_peer, tmo);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
197	}
198	dst_release(dst);
199	return res;
200}
201
202/*
203 *	an inline helper for the "simple" if statement below
204 *	checks if parameter problem report is caused by an
205 *	unrecognized IPv6 option that has the Option Type
206 *	highest-order two bits set to 10
207 */
208
209static __inline__ int opt_unrec(struct sk_buff *skb, __u32 offset)
210{
211	u8 _optval, *op;
212
213	offset += skb_network_offset(skb);
214	op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
215	if (op == NULL)
216		return 1;
217	return (*op & 0xC0) == 0x80;
218}
219
220static int icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, struct icmp6hdr *thdr, int len)
 
221{
222	struct sk_buff *skb;
223	struct icmp6hdr *icmp6h;
224	int err = 0;
225
226	if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
227		goto out;
 
228
229	icmp6h = icmp6_hdr(skb);
230	memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
231	icmp6h->icmp6_cksum = 0;
232
233	if (skb_queue_len(&sk->sk_write_queue) == 1) {
234		skb->csum = csum_partial(icmp6h,
235					sizeof(struct icmp6hdr), skb->csum);
236		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
237						      &fl6->daddr,
238						      len, fl6->flowi6_proto,
239						      skb->csum);
240	} else {
241		__wsum tmp_csum = 0;
242
243		skb_queue_walk(&sk->sk_write_queue, skb) {
244			tmp_csum = csum_add(tmp_csum, skb->csum);
245		}
246
247		tmp_csum = csum_partial(icmp6h,
248					sizeof(struct icmp6hdr), tmp_csum);
249		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
250						      &fl6->daddr,
251						      len, fl6->flowi6_proto,
252						      tmp_csum);
253	}
254	ip6_push_pending_frames(sk);
255out:
256	return err;
257}
258
259struct icmpv6_msg {
260	struct sk_buff	*skb;
261	int		offset;
262	uint8_t		type;
263};
264
265static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
266{
267	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
268	struct sk_buff *org_skb = msg->skb;
269	__wsum csum = 0;
270
271	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
272				      to, len, csum);
273	skb->csum = csum_block_add(skb->csum, csum, odd);
274	if (!(msg->type & ICMPV6_INFOMSG_MASK))
275		nf_ct_attach(skb, org_skb);
276	return 0;
277}
278
279#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
280static void mip6_addr_swap(struct sk_buff *skb)
281{
282	struct ipv6hdr *iph = ipv6_hdr(skb);
283	struct inet6_skb_parm *opt = IP6CB(skb);
284	struct ipv6_destopt_hao *hao;
285	struct in6_addr tmp;
286	int off;
287
288	if (opt->dsthao) {
289		off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
290		if (likely(off >= 0)) {
291			hao = (struct ipv6_destopt_hao *)
292					(skb_network_header(skb) + off);
293			ipv6_addr_copy(&tmp, &iph->saddr);
294			ipv6_addr_copy(&iph->saddr, &hao->addr);
295			ipv6_addr_copy(&hao->addr, &tmp);
296		}
297	}
298}
299#else
300static inline void mip6_addr_swap(struct sk_buff *skb) {}
301#endif
302
303static struct dst_entry *icmpv6_route_lookup(struct net *net, struct sk_buff *skb,
304					     struct sock *sk, struct flowi6 *fl6)
 
 
305{
306	struct dst_entry *dst, *dst2;
307	struct flowi6 fl2;
308	int err;
309
310	err = ip6_dst_lookup(sk, &dst, fl6);
311	if (err)
312		return ERR_PTR(err);
313
314	/*
315	 * We won't send icmp if the destination is known
316	 * anycast.
317	 */
318	if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) {
319		LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: acast source\n");
 
320		dst_release(dst);
321		return ERR_PTR(-EINVAL);
322	}
323
324	/* No need to clone since we're just using its address. */
325	dst2 = dst;
326
327	dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
328	if (!IS_ERR(dst)) {
329		if (dst != dst2)
330			return dst;
331	} else {
332		if (PTR_ERR(dst) == -EPERM)
333			dst = NULL;
334		else
335			return dst;
336	}
337
338	err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
339	if (err)
340		goto relookup_failed;
341
342	err = ip6_dst_lookup(sk, &dst2, &fl2);
343	if (err)
344		goto relookup_failed;
345
346	dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
347	if (!IS_ERR(dst2)) {
348		dst_release(dst);
349		dst = dst2;
350	} else {
351		err = PTR_ERR(dst2);
352		if (err == -EPERM) {
353			dst_release(dst);
354			return dst2;
355		} else
356			goto relookup_failed;
357	}
358
359relookup_failed:
360	if (dst)
361		return dst;
362	return ERR_PTR(err);
363}
364
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
365/*
366 *	Send an ICMP message in response to a packet in error
367 */
368void icmpv6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info)
 
 
369{
370	struct net *net = dev_net(skb->dev);
371	struct inet6_dev *idev = NULL;
372	struct ipv6hdr *hdr = ipv6_hdr(skb);
373	struct sock *sk;
 
374	struct ipv6_pinfo *np;
375	const struct in6_addr *saddr = NULL;
376	struct dst_entry *dst;
377	struct icmp6hdr tmp_hdr;
378	struct flowi6 fl6;
379	struct icmpv6_msg msg;
 
380	int iif = 0;
381	int addr_type = 0;
382	int len;
383	int hlimit;
384	int err = 0;
385
386	if ((u8 *)hdr < skb->head ||
387	    (skb->network_header + sizeof(*hdr)) > skb->tail)
388		return;
389
 
 
 
 
390	/*
391	 *	Make sure we respect the rules
392	 *	i.e. RFC 1885 2.4(e)
393	 *	Rule (e.1) is enforced by not using icmpv6_send
394	 *	in any code that processes icmp errors.
395	 */
396	addr_type = ipv6_addr_type(&hdr->daddr);
397
398	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0))
 
399		saddr = &hdr->daddr;
400
401	/*
402	 *	Dest addr check
403	 */
404
405	if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) {
406		if (type != ICMPV6_PKT_TOOBIG &&
407		    !(type == ICMPV6_PARAMPROB &&
408		      code == ICMPV6_UNK_OPTION &&
409		      (opt_unrec(skb, info))))
410			return;
411
412		saddr = NULL;
413	}
414
415	addr_type = ipv6_addr_type(&hdr->saddr);
416
417	/*
418	 *	Source addr check
419	 */
420
421	if (addr_type & IPV6_ADDR_LINKLOCAL)
422		iif = skb->dev->ifindex;
 
 
 
 
 
 
 
423
424	/*
425	 *	Must not send error if the source does not uniquely
426	 *	identify a single node (RFC2463 Section 2.4).
427	 *	We check unspecified / multicast addresses here,
428	 *	and anycast addresses will be checked later.
429	 */
430	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
431		LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: addr_any/mcast source\n");
 
432		return;
433	}
434
435	/*
436	 *	Never answer to a ICMP packet.
437	 */
438	if (is_ineligible(skb)) {
439		LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: no reply to icmp error\n");
 
440		return;
441	}
442
443	mip6_addr_swap(skb);
 
 
 
 
 
 
 
 
 
 
 
444
445	memset(&fl6, 0, sizeof(fl6));
446	fl6.flowi6_proto = IPPROTO_ICMPV6;
447	ipv6_addr_copy(&fl6.daddr, &hdr->saddr);
448	if (saddr)
449		ipv6_addr_copy(&fl6.saddr, saddr);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
450	fl6.flowi6_oif = iif;
451	fl6.fl6_icmp_type = type;
452	fl6.fl6_icmp_code = code;
453	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
 
 
454
455	sk = icmpv6_xmit_lock(net);
456	if (sk == NULL)
457		return;
458	np = inet6_sk(sk);
459
460	if (!icmpv6_xrlim_allow(sk, type, &fl6))
461		goto out;
462
463	tmp_hdr.icmp6_type = type;
464	tmp_hdr.icmp6_code = code;
465	tmp_hdr.icmp6_cksum = 0;
466	tmp_hdr.icmp6_pointer = htonl(info);
467
468	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
469		fl6.flowi6_oif = np->mcast_oif;
 
 
 
 
 
 
470
471	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
472	if (IS_ERR(dst))
473		goto out;
474
475	if (ipv6_addr_is_multicast(&fl6.daddr))
476		hlimit = np->mcast_hops;
477	else
478		hlimit = np->hop_limit;
479	if (hlimit < 0)
480		hlimit = ip6_dst_hoplimit(dst);
481
482	msg.skb = skb;
483	msg.offset = skb_network_offset(skb);
484	msg.type = type;
485
486	len = skb->len - msg.offset;
487	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr));
488	if (len < 0) {
489		LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n");
 
490		goto out_dst_release;
491	}
492
493	idev = in6_dev_get(skb->dev);
 
494
495	err = ip6_append_data(sk, icmpv6_getfrag, &msg,
496			      len + sizeof(struct icmp6hdr),
497			      sizeof(struct icmp6hdr), hlimit,
498			      np->tclass, NULL, &fl6, (struct rt6_info*)dst,
499			      MSG_DONTWAIT, np->dontfrag);
500	if (err) {
501		ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTERRORS);
502		ip6_flush_pending_frames(sk);
503		goto out_put;
 
 
504	}
505	err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr, len + sizeof(struct icmp6hdr));
506
507out_put:
508	if (likely(idev != NULL))
509		in6_dev_put(idev);
510out_dst_release:
511	dst_release(dst);
512out:
513	icmpv6_xmit_unlock(sk);
 
 
514}
 
515
516EXPORT_SYMBOL(icmpv6_send);
 
 
 
 
 
 
 
517
518static void icmpv6_echo_reply(struct sk_buff *skb)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
519{
520	struct net *net = dev_net(skb->dev);
521	struct sock *sk;
522	struct inet6_dev *idev;
523	struct ipv6_pinfo *np;
524	const struct in6_addr *saddr = NULL;
525	struct icmp6hdr *icmph = icmp6_hdr(skb);
526	struct icmp6hdr tmp_hdr;
527	struct flowi6 fl6;
528	struct icmpv6_msg msg;
529	struct dst_entry *dst;
530	int err = 0;
531	int hlimit;
 
 
 
 
 
 
 
532
533	saddr = &ipv6_hdr(skb)->daddr;
534
535	if (!ipv6_unicast_destination(skb))
 
 
 
 
 
536		saddr = NULL;
537
 
 
 
 
 
538	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
539	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
540
541	memset(&fl6, 0, sizeof(fl6));
 
 
 
542	fl6.flowi6_proto = IPPROTO_ICMPV6;
543	ipv6_addr_copy(&fl6.daddr, &ipv6_hdr(skb)->saddr);
544	if (saddr)
545		ipv6_addr_copy(&fl6.saddr, saddr);
546	fl6.flowi6_oif = skb->dev->ifindex;
547	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
548	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
 
 
549
 
550	sk = icmpv6_xmit_lock(net);
551	if (sk == NULL)
552		return;
553	np = inet6_sk(sk);
554
555	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
556		fl6.flowi6_oif = np->mcast_oif;
 
 
557
558	err = ip6_dst_lookup(sk, &dst, &fl6);
559	if (err)
560		goto out;
561	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
562	if (IS_ERR(dst))
563		goto out;
564
565	if (ipv6_addr_is_multicast(&fl6.daddr))
566		hlimit = np->mcast_hops;
567	else
568		hlimit = np->hop_limit;
569	if (hlimit < 0)
570		hlimit = ip6_dst_hoplimit(dst);
571
572	idev = in6_dev_get(skb->dev);
573
574	msg.skb = skb;
575	msg.offset = 0;
576	msg.type = ICMPV6_ECHO_REPLY;
577
578	err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
579				sizeof(struct icmp6hdr), hlimit, np->tclass, NULL, &fl6,
580				(struct rt6_info*)dst, MSG_DONTWAIT,
581				np->dontfrag);
582
583	if (err) {
584		ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTERRORS);
 
 
 
 
 
 
 
585		ip6_flush_pending_frames(sk);
586		goto out_put;
 
 
 
587	}
588	err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr, skb->len + sizeof(struct icmp6hdr));
589
590out_put:
591	if (likely(idev != NULL))
592		in6_dev_put(idev);
593	dst_release(dst);
594out:
595	icmpv6_xmit_unlock(sk);
 
 
 
596}
597
598static void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
 
599{
 
 
600	const struct inet6_protocol *ipprot;
 
601	int inner_offset;
602	int hash;
603	u8 nexthdr;
604
605	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
606		return;
 
 
 
607
608	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
609	if (ipv6_ext_hdr(nexthdr)) {
610		/* now skip over extension headers */
611		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr);
612		if (inner_offset<0)
613			return;
 
 
 
614	} else {
615		inner_offset = sizeof(struct ipv6hdr);
616	}
617
618	/* Checkin header including 8 bytes of inner protocol header. */
619	if (!pskb_may_pull(skb, inner_offset+8))
620		return;
 
621
622	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
623	   Without this we will not able f.e. to make source routed
624	   pmtu discovery.
625	   Corresponding argument (opt) to notifiers is already added.
626	   --ANK (980726)
627	 */
628
629	hash = nexthdr & (MAX_INET_PROTOS - 1);
630
631	rcu_read_lock();
632	ipprot = rcu_dereference(inet6_protos[hash]);
633	if (ipprot && ipprot->err_handler)
634		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
635	rcu_read_unlock();
636
637	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
 
 
 
 
 
638}
639
640/*
641 *	Handle icmp messages
642 */
643
644static int icmpv6_rcv(struct sk_buff *skb)
645{
646	struct net_device *dev = skb->dev;
 
 
647	struct inet6_dev *idev = __in6_dev_get(dev);
648	const struct in6_addr *saddr, *daddr;
649	const struct ipv6hdr *orig_hdr;
650	struct icmp6hdr *hdr;
651	u8 type;
652
653	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
654		struct sec_path *sp = skb_sec_path(skb);
655		int nh;
656
657		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
658				 XFRM_STATE_ICMP))
 
659			goto drop_no_count;
 
660
661		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(*orig_hdr)))
662			goto drop_no_count;
663
664		nh = skb_network_offset(skb);
665		skb_set_network_header(skb, sizeof(*hdr));
666
667		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
 
 
668			goto drop_no_count;
 
669
670		skb_set_network_header(skb, nh);
671	}
672
673	ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INMSGS);
674
675	saddr = &ipv6_hdr(skb)->saddr;
676	daddr = &ipv6_hdr(skb)->daddr;
677
678	/* Perform checksum. */
679	switch (skb->ip_summed) {
680	case CHECKSUM_COMPLETE:
681		if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6,
682				     skb->csum))
683			break;
684		/* fall through */
685	case CHECKSUM_NONE:
686		skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len,
687					     IPPROTO_ICMPV6, 0));
688		if (__skb_checksum_complete(skb)) {
689			LIMIT_NETDEBUG(KERN_DEBUG "ICMPv6 checksum failed [%pI6 > %pI6]\n",
690				       saddr, daddr);
691			goto discard_it;
692		}
693	}
694
695	if (!pskb_pull(skb, sizeof(*hdr)))
696		goto discard_it;
697
698	hdr = icmp6_hdr(skb);
699
700	type = hdr->icmp6_type;
701
702	ICMP6MSGIN_INC_STATS_BH(dev_net(dev), idev, type);
703
704	switch (type) {
705	case ICMPV6_ECHO_REQUEST:
706		icmpv6_echo_reply(skb);
 
 
 
 
 
 
707		break;
708
709	case ICMPV6_ECHO_REPLY:
710		/* we couldn't care less */
 
 
 
 
711		break;
712
713	case ICMPV6_PKT_TOOBIG:
714		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
715		   standard destination cache. Seems, only "advanced"
716		   destination cache will allow to solve this problem
717		   --ANK (980726)
718		 */
719		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
720			goto discard_it;
721		hdr = icmp6_hdr(skb);
722		orig_hdr = (struct ipv6hdr *) (hdr + 1);
723		rt6_pmtu_discovery(&orig_hdr->daddr, &orig_hdr->saddr, dev,
724				   ntohl(hdr->icmp6_mtu));
725
726		/*
727		 *	Drop through to notify
728		 */
729
 
 
730	case ICMPV6_DEST_UNREACH:
731	case ICMPV6_TIME_EXCEED:
732	case ICMPV6_PARAMPROB:
733		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
 
734		break;
735
736	case NDISC_ROUTER_SOLICITATION:
737	case NDISC_ROUTER_ADVERTISEMENT:
738	case NDISC_NEIGHBOUR_SOLICITATION:
739	case NDISC_NEIGHBOUR_ADVERTISEMENT:
740	case NDISC_REDIRECT:
741		ndisc_rcv(skb);
742		break;
743
744	case ICMPV6_MGM_QUERY:
745		igmp6_event_query(skb);
746		break;
747
748	case ICMPV6_MGM_REPORT:
749		igmp6_event_report(skb);
750		break;
751
752	case ICMPV6_MGM_REDUCTION:
753	case ICMPV6_NI_QUERY:
754	case ICMPV6_NI_REPLY:
755	case ICMPV6_MLD2_REPORT:
756	case ICMPV6_DHAAD_REQUEST:
757	case ICMPV6_DHAAD_REPLY:
758	case ICMPV6_MOBILE_PREFIX_SOL:
759	case ICMPV6_MOBILE_PREFIX_ADV:
760		break;
761
762	default:
763		LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n");
764
765		/* informational */
766		if (type & ICMPV6_INFOMSG_MASK)
767			break;
768
 
 
 
769		/*
770		 * error of unknown type.
771		 * must pass to upper level
772		 */
773
774		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
 
775	}
776
777	kfree_skb(skb);
 
 
 
 
 
 
 
778	return 0;
779
 
 
 
780discard_it:
781	ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INERRORS);
782drop_no_count:
783	kfree_skb(skb);
784	return 0;
785}
786
787void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
788		      u8 type,
789		      const struct in6_addr *saddr,
790		      const struct in6_addr *daddr,
791		      int oif)
792{
793	memset(fl6, 0, sizeof(*fl6));
794	ipv6_addr_copy(&fl6->saddr, saddr);
795	ipv6_addr_copy(&fl6->daddr, daddr);
796	fl6->flowi6_proto 	= IPPROTO_ICMPV6;
797	fl6->fl6_icmp_type	= type;
798	fl6->fl6_icmp_code	= 0;
799	fl6->flowi6_oif		= oif;
800	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
801}
802
803/*
804 * Special lock-class for __icmpv6_sk:
805 */
806static struct lock_class_key icmpv6_socket_sk_dst_lock_key;
807
808static int __net_init icmpv6_sk_init(struct net *net)
809{
810	struct sock *sk;
811	int err, i, j;
812
813	net->ipv6.icmp_sk =
814		kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
815	if (net->ipv6.icmp_sk == NULL)
816		return -ENOMEM;
817
818	for_each_possible_cpu(i) {
819		err = inet_ctl_sock_create(&sk, PF_INET6,
820					   SOCK_RAW, IPPROTO_ICMPV6, net);
821		if (err < 0) {
822			printk(KERN_ERR
823			       "Failed to initialize the ICMP6 control socket "
824			       "(err %d).\n",
825			       err);
826			goto fail;
827		}
828
829		net->ipv6.icmp_sk[i] = sk;
830
831		/*
832		 * Split off their lock-class, because sk->sk_dst_lock
833		 * gets used from softirqs, which is safe for
834		 * __icmpv6_sk (because those never get directly used
835		 * via userspace syscalls), but unsafe for normal sockets.
836		 */
837		lockdep_set_class(&sk->sk_dst_lock,
838				  &icmpv6_socket_sk_dst_lock_key);
839
840		/* Enough space for 2 64K ICMP packets, including
841		 * sk_buff struct overhead.
842		 */
843		sk->sk_sndbuf =
844			(2 * ((64 * 1024) + sizeof(struct sk_buff)));
845	}
846	return 0;
847
848 fail:
849	for (j = 0; j < i; j++)
850		inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
851	kfree(net->ipv6.icmp_sk);
852	return err;
853}
854
855static void __net_exit icmpv6_sk_exit(struct net *net)
856{
857	int i;
858
859	for_each_possible_cpu(i) {
860		inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
861	}
862	kfree(net->ipv6.icmp_sk);
863}
864
865static struct pernet_operations icmpv6_sk_ops = {
866       .init = icmpv6_sk_init,
867       .exit = icmpv6_sk_exit,
868};
869
870int __init icmpv6_init(void)
871{
872	int err;
873
874	err = register_pernet_subsys(&icmpv6_sk_ops);
875	if (err < 0)
876		return err;
877
878	err = -EAGAIN;
879	if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
880		goto fail;
 
 
 
 
881	return 0;
882
 
 
883fail:
884	printk(KERN_ERR "Failed to register ICMP6 protocol\n");
885	unregister_pernet_subsys(&icmpv6_sk_ops);
886	return err;
887}
888
889void icmpv6_cleanup(void)
890{
891	unregister_pernet_subsys(&icmpv6_sk_ops);
892	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
893}
894
895
896static const struct icmp6_err {
897	int err;
898	int fatal;
899} tab_unreach[] = {
900	{	/* NOROUTE */
901		.err	= ENETUNREACH,
902		.fatal	= 0,
903	},
904	{	/* ADM_PROHIBITED */
905		.err	= EACCES,
906		.fatal	= 1,
907	},
908	{	/* Was NOT_NEIGHBOUR, now reserved */
909		.err	= EHOSTUNREACH,
910		.fatal	= 0,
911	},
912	{	/* ADDR_UNREACH	*/
913		.err	= EHOSTUNREACH,
914		.fatal	= 0,
915	},
916	{	/* PORT_UNREACH	*/
917		.err	= ECONNREFUSED,
918		.fatal	= 1,
919	},
 
 
 
 
 
 
 
 
920};
921
922int icmpv6_err_convert(u8 type, u8 code, int *err)
923{
924	int fatal = 0;
925
926	*err = EPROTO;
927
928	switch (type) {
929	case ICMPV6_DEST_UNREACH:
930		fatal = 1;
931		if (code <= ICMPV6_PORT_UNREACH) {
932			*err  = tab_unreach[code].err;
933			fatal = tab_unreach[code].fatal;
934		}
935		break;
936
937	case ICMPV6_PKT_TOOBIG:
938		*err = EMSGSIZE;
939		break;
940
941	case ICMPV6_PARAMPROB:
942		*err = EPROTO;
943		fatal = 1;
944		break;
945
946	case ICMPV6_TIME_EXCEED:
947		*err = EHOSTUNREACH;
948		break;
949	}
950
951	return fatal;
952}
953
954EXPORT_SYMBOL(icmpv6_err_convert);
955
956#ifdef CONFIG_SYSCTL
957ctl_table ipv6_icmp_table_template[] = {
958	{
959		.procname	= "ratelimit",
960		.data		= &init_net.ipv6.sysctl.icmpv6_time,
961		.maxlen		= sizeof(int),
962		.mode		= 0644,
963		.proc_handler	= proc_dointvec_ms_jiffies,
964	},
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
965	{ },
966};
967
968struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
969{
970	struct ctl_table *table;
971
972	table = kmemdup(ipv6_icmp_table_template,
973			sizeof(ipv6_icmp_table_template),
974			GFP_KERNEL);
975
976	if (table)
977		table[0].data = &net->ipv6.sysctl.icmpv6_time;
 
 
 
 
 
 
 
 
978
979	return table;
 
 
980}
981#endif
982