xref: /openbmc/linux/net/ipv6/icmp.c (revision afc98d90)
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 #define pr_fmt(fmt) "IPv6: " fmt
33 
34 #include <linux/module.h>
35 #include <linux/errno.h>
36 #include <linux/types.h>
37 #include <linux/socket.h>
38 #include <linux/in.h>
39 #include <linux/kernel.h>
40 #include <linux/sockios.h>
41 #include <linux/net.h>
42 #include <linux/skbuff.h>
43 #include <linux/init.h>
44 #include <linux/netfilter.h>
45 #include <linux/slab.h>
46 
47 #ifdef CONFIG_SYSCTL
48 #include <linux/sysctl.h>
49 #endif
50 
51 #include <linux/inet.h>
52 #include <linux/netdevice.h>
53 #include <linux/icmpv6.h>
54 
55 #include <net/ip.h>
56 #include <net/sock.h>
57 
58 #include <net/ipv6.h>
59 #include <net/ip6_checksum.h>
60 #include <net/ping.h>
61 #include <net/protocol.h>
62 #include <net/raw.h>
63 #include <net/rawv6.h>
64 #include <net/transp_v6.h>
65 #include <net/ip6_route.h>
66 #include <net/addrconf.h>
67 #include <net/icmp.h>
68 #include <net/xfrm.h>
69 #include <net/inet_common.h>
70 #include <net/dsfield.h>
71 
72 #include <asm/uaccess.h>
73 
74 /*
75  *	The ICMP socket(s). This is the most convenient way to flow control
76  *	our ICMP output as well as maintain a clean interface throughout
77  *	all layers. All Socketless IP sends will soon be gone.
78  *
79  *	On SMP we have one ICMP socket per-cpu.
80  */
81 static inline struct sock *icmpv6_sk(struct net *net)
82 {
83 	return net->ipv6.icmp_sk[smp_processor_id()];
84 }
85 
86 static void icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
87 		       u8 type, u8 code, int offset, __be32 info)
88 {
89 	/* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */
90 	struct icmp6hdr *icmp6 = (struct icmp6hdr *) (skb->data + offset);
91 	struct net *net = dev_net(skb->dev);
92 
93 	if (type == ICMPV6_PKT_TOOBIG)
94 		ip6_update_pmtu(skb, net, info, 0, 0);
95 	else if (type == NDISC_REDIRECT)
96 		ip6_redirect(skb, net, skb->dev->ifindex, 0);
97 
98 	if (!(type & ICMPV6_INFOMSG_MASK))
99 		if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST)
100 			ping_err(skb, offset, info);
101 }
102 
103 static int icmpv6_rcv(struct sk_buff *skb);
104 
105 static const struct inet6_protocol icmpv6_protocol = {
106 	.handler	=	icmpv6_rcv,
107 	.err_handler	=	icmpv6_err,
108 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
109 };
110 
111 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
112 {
113 	struct sock *sk;
114 
115 	local_bh_disable();
116 
117 	sk = icmpv6_sk(net);
118 	if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
119 		/* This can happen if the output path (f.e. SIT or
120 		 * ip6ip6 tunnel) signals dst_link_failure() for an
121 		 * outgoing ICMP6 packet.
122 		 */
123 		local_bh_enable();
124 		return NULL;
125 	}
126 	return sk;
127 }
128 
129 static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
130 {
131 	spin_unlock_bh(&sk->sk_lock.slock);
132 }
133 
134 /*
135  * Figure out, may we reply to this packet with icmp error.
136  *
137  * We do not reply, if:
138  *	- it was icmp error message.
139  *	- it is truncated, so that it is known, that protocol is ICMPV6
140  *	  (i.e. in the middle of some exthdr)
141  *
142  *	--ANK (980726)
143  */
144 
145 static bool is_ineligible(const struct sk_buff *skb)
146 {
147 	int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
148 	int len = skb->len - ptr;
149 	__u8 nexthdr = ipv6_hdr(skb)->nexthdr;
150 	__be16 frag_off;
151 
152 	if (len < 0)
153 		return true;
154 
155 	ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off);
156 	if (ptr < 0)
157 		return false;
158 	if (nexthdr == IPPROTO_ICMPV6) {
159 		u8 _type, *tp;
160 		tp = skb_header_pointer(skb,
161 			ptr+offsetof(struct icmp6hdr, icmp6_type),
162 			sizeof(_type), &_type);
163 		if (tp == NULL ||
164 		    !(*tp & ICMPV6_INFOMSG_MASK))
165 			return true;
166 	}
167 	return false;
168 }
169 
170 /*
171  * Check the ICMP output rate limit
172  */
173 static inline bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
174 				      struct flowi6 *fl6)
175 {
176 	struct dst_entry *dst;
177 	struct net *net = sock_net(sk);
178 	bool res = false;
179 
180 	/* Informational messages are not limited. */
181 	if (type & ICMPV6_INFOMSG_MASK)
182 		return true;
183 
184 	/* Do not limit pmtu discovery, it would break it. */
185 	if (type == ICMPV6_PKT_TOOBIG)
186 		return true;
187 
188 	/*
189 	 * Look up the output route.
190 	 * XXX: perhaps the expire for routing entries cloned by
191 	 * this lookup should be more aggressive (not longer than timeout).
192 	 */
193 	dst = ip6_route_output(net, sk, fl6);
194 	if (dst->error) {
195 		IP6_INC_STATS(net, ip6_dst_idev(dst),
196 			      IPSTATS_MIB_OUTNOROUTES);
197 	} else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
198 		res = true;
199 	} else {
200 		struct rt6_info *rt = (struct rt6_info *)dst;
201 		int tmo = net->ipv6.sysctl.icmpv6_time;
202 		struct inet_peer *peer;
203 
204 		/* Give more bandwidth to wider prefixes. */
205 		if (rt->rt6i_dst.plen < 128)
206 			tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
207 
208 		peer = inet_getpeer_v6(net->ipv6.peers, &rt->rt6i_dst.addr, 1);
209 		res = inet_peer_xrlim_allow(peer, tmo);
210 		if (peer)
211 			inet_putpeer(peer);
212 	}
213 	dst_release(dst);
214 	return res;
215 }
216 
217 /*
218  *	an inline helper for the "simple" if statement below
219  *	checks if parameter problem report is caused by an
220  *	unrecognized IPv6 option that has the Option Type
221  *	highest-order two bits set to 10
222  */
223 
224 static bool opt_unrec(struct sk_buff *skb, __u32 offset)
225 {
226 	u8 _optval, *op;
227 
228 	offset += skb_network_offset(skb);
229 	op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
230 	if (op == NULL)
231 		return true;
232 	return (*op & 0xC0) == 0x80;
233 }
234 
235 int icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
236 			       struct icmp6hdr *thdr, int len)
237 {
238 	struct sk_buff *skb;
239 	struct icmp6hdr *icmp6h;
240 	int err = 0;
241 
242 	if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
243 		goto out;
244 
245 	icmp6h = icmp6_hdr(skb);
246 	memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
247 	icmp6h->icmp6_cksum = 0;
248 
249 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
250 		skb->csum = csum_partial(icmp6h,
251 					sizeof(struct icmp6hdr), skb->csum);
252 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
253 						      &fl6->daddr,
254 						      len, fl6->flowi6_proto,
255 						      skb->csum);
256 	} else {
257 		__wsum tmp_csum = 0;
258 
259 		skb_queue_walk(&sk->sk_write_queue, skb) {
260 			tmp_csum = csum_add(tmp_csum, skb->csum);
261 		}
262 
263 		tmp_csum = csum_partial(icmp6h,
264 					sizeof(struct icmp6hdr), tmp_csum);
265 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
266 						      &fl6->daddr,
267 						      len, fl6->flowi6_proto,
268 						      tmp_csum);
269 	}
270 	ip6_push_pending_frames(sk);
271 out:
272 	return err;
273 }
274 
275 struct icmpv6_msg {
276 	struct sk_buff	*skb;
277 	int		offset;
278 	uint8_t		type;
279 };
280 
281 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
282 {
283 	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
284 	struct sk_buff *org_skb = msg->skb;
285 	__wsum csum = 0;
286 
287 	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
288 				      to, len, csum);
289 	skb->csum = csum_block_add(skb->csum, csum, odd);
290 	if (!(msg->type & ICMPV6_INFOMSG_MASK))
291 		nf_ct_attach(skb, org_skb);
292 	return 0;
293 }
294 
295 #if IS_ENABLED(CONFIG_IPV6_MIP6)
296 static void mip6_addr_swap(struct sk_buff *skb)
297 {
298 	struct ipv6hdr *iph = ipv6_hdr(skb);
299 	struct inet6_skb_parm *opt = IP6CB(skb);
300 	struct ipv6_destopt_hao *hao;
301 	struct in6_addr tmp;
302 	int off;
303 
304 	if (opt->dsthao) {
305 		off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
306 		if (likely(off >= 0)) {
307 			hao = (struct ipv6_destopt_hao *)
308 					(skb_network_header(skb) + off);
309 			tmp = iph->saddr;
310 			iph->saddr = hao->addr;
311 			hao->addr = tmp;
312 		}
313 	}
314 }
315 #else
316 static inline void mip6_addr_swap(struct sk_buff *skb) {}
317 #endif
318 
319 static struct dst_entry *icmpv6_route_lookup(struct net *net,
320 					     struct sk_buff *skb,
321 					     struct sock *sk,
322 					     struct flowi6 *fl6)
323 {
324 	struct dst_entry *dst, *dst2;
325 	struct flowi6 fl2;
326 	int err;
327 
328 	err = ip6_dst_lookup(sk, &dst, fl6);
329 	if (err)
330 		return ERR_PTR(err);
331 
332 	/*
333 	 * We won't send icmp if the destination is known
334 	 * anycast.
335 	 */
336 	if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) {
337 		LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: acast source\n");
338 		dst_release(dst);
339 		return ERR_PTR(-EINVAL);
340 	}
341 
342 	/* No need to clone since we're just using its address. */
343 	dst2 = dst;
344 
345 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
346 	if (!IS_ERR(dst)) {
347 		if (dst != dst2)
348 			return dst;
349 	} else {
350 		if (PTR_ERR(dst) == -EPERM)
351 			dst = NULL;
352 		else
353 			return dst;
354 	}
355 
356 	err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
357 	if (err)
358 		goto relookup_failed;
359 
360 	err = ip6_dst_lookup(sk, &dst2, &fl2);
361 	if (err)
362 		goto relookup_failed;
363 
364 	dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
365 	if (!IS_ERR(dst2)) {
366 		dst_release(dst);
367 		dst = dst2;
368 	} else {
369 		err = PTR_ERR(dst2);
370 		if (err == -EPERM) {
371 			dst_release(dst);
372 			return dst2;
373 		} else
374 			goto relookup_failed;
375 	}
376 
377 relookup_failed:
378 	if (dst)
379 		return dst;
380 	return ERR_PTR(err);
381 }
382 
383 /*
384  *	Send an ICMP message in response to a packet in error
385  */
386 static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info)
387 {
388 	struct net *net = dev_net(skb->dev);
389 	struct inet6_dev *idev = NULL;
390 	struct ipv6hdr *hdr = ipv6_hdr(skb);
391 	struct sock *sk;
392 	struct ipv6_pinfo *np;
393 	const struct in6_addr *saddr = NULL;
394 	struct dst_entry *dst;
395 	struct icmp6hdr tmp_hdr;
396 	struct flowi6 fl6;
397 	struct icmpv6_msg msg;
398 	int iif = 0;
399 	int addr_type = 0;
400 	int len;
401 	int hlimit;
402 	int err = 0;
403 
404 	if ((u8 *)hdr < skb->head ||
405 	    (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb))
406 		return;
407 
408 	/*
409 	 *	Make sure we respect the rules
410 	 *	i.e. RFC 1885 2.4(e)
411 	 *	Rule (e.1) is enforced by not using icmp6_send
412 	 *	in any code that processes icmp errors.
413 	 */
414 	addr_type = ipv6_addr_type(&hdr->daddr);
415 
416 	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) ||
417 	    ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr))
418 		saddr = &hdr->daddr;
419 
420 	/*
421 	 *	Dest addr check
422 	 */
423 
424 	if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) {
425 		if (type != ICMPV6_PKT_TOOBIG &&
426 		    !(type == ICMPV6_PARAMPROB &&
427 		      code == ICMPV6_UNK_OPTION &&
428 		      (opt_unrec(skb, info))))
429 			return;
430 
431 		saddr = NULL;
432 	}
433 
434 	addr_type = ipv6_addr_type(&hdr->saddr);
435 
436 	/*
437 	 *	Source addr check
438 	 */
439 
440 	if (__ipv6_addr_needs_scope_id(addr_type))
441 		iif = skb->dev->ifindex;
442 
443 	/*
444 	 *	Must not send error if the source does not uniquely
445 	 *	identify a single node (RFC2463 Section 2.4).
446 	 *	We check unspecified / multicast addresses here,
447 	 *	and anycast addresses will be checked later.
448 	 */
449 	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
450 		LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: addr_any/mcast source\n");
451 		return;
452 	}
453 
454 	/*
455 	 *	Never answer to a ICMP packet.
456 	 */
457 	if (is_ineligible(skb)) {
458 		LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: no reply to icmp error\n");
459 		return;
460 	}
461 
462 	mip6_addr_swap(skb);
463 
464 	memset(&fl6, 0, sizeof(fl6));
465 	fl6.flowi6_proto = IPPROTO_ICMPV6;
466 	fl6.daddr = hdr->saddr;
467 	if (saddr)
468 		fl6.saddr = *saddr;
469 	fl6.flowi6_oif = iif;
470 	fl6.fl6_icmp_type = type;
471 	fl6.fl6_icmp_code = code;
472 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
473 
474 	sk = icmpv6_xmit_lock(net);
475 	if (sk == NULL)
476 		return;
477 	np = inet6_sk(sk);
478 
479 	if (!icmpv6_xrlim_allow(sk, type, &fl6))
480 		goto out;
481 
482 	tmp_hdr.icmp6_type = type;
483 	tmp_hdr.icmp6_code = code;
484 	tmp_hdr.icmp6_cksum = 0;
485 	tmp_hdr.icmp6_pointer = htonl(info);
486 
487 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
488 		fl6.flowi6_oif = np->mcast_oif;
489 	else if (!fl6.flowi6_oif)
490 		fl6.flowi6_oif = np->ucast_oif;
491 
492 	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
493 	if (IS_ERR(dst))
494 		goto out;
495 
496 	if (ipv6_addr_is_multicast(&fl6.daddr))
497 		hlimit = np->mcast_hops;
498 	else
499 		hlimit = np->hop_limit;
500 	if (hlimit < 0)
501 		hlimit = ip6_dst_hoplimit(dst);
502 
503 	msg.skb = skb;
504 	msg.offset = skb_network_offset(skb);
505 	msg.type = type;
506 
507 	len = skb->len - msg.offset;
508 	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr));
509 	if (len < 0) {
510 		LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n");
511 		goto out_dst_release;
512 	}
513 
514 	rcu_read_lock();
515 	idev = __in6_dev_get(skb->dev);
516 
517 	err = ip6_append_data(sk, icmpv6_getfrag, &msg,
518 			      len + sizeof(struct icmp6hdr),
519 			      sizeof(struct icmp6hdr), hlimit,
520 			      np->tclass, NULL, &fl6, (struct rt6_info *)dst,
521 			      MSG_DONTWAIT, np->dontfrag);
522 	if (err) {
523 		ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTERRORS);
524 		ip6_flush_pending_frames(sk);
525 	} else {
526 		err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
527 						 len + sizeof(struct icmp6hdr));
528 	}
529 	rcu_read_unlock();
530 out_dst_release:
531 	dst_release(dst);
532 out:
533 	icmpv6_xmit_unlock(sk);
534 }
535 
536 /* Slightly more convenient version of icmp6_send.
537  */
538 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
539 {
540 	icmp6_send(skb, ICMPV6_PARAMPROB, code, pos);
541 	kfree_skb(skb);
542 }
543 
544 static void icmpv6_echo_reply(struct sk_buff *skb)
545 {
546 	struct net *net = dev_net(skb->dev);
547 	struct sock *sk;
548 	struct inet6_dev *idev;
549 	struct ipv6_pinfo *np;
550 	const struct in6_addr *saddr = NULL;
551 	struct icmp6hdr *icmph = icmp6_hdr(skb);
552 	struct icmp6hdr tmp_hdr;
553 	struct flowi6 fl6;
554 	struct icmpv6_msg msg;
555 	struct dst_entry *dst;
556 	int err = 0;
557 	int hlimit;
558 	u8 tclass;
559 
560 	saddr = &ipv6_hdr(skb)->daddr;
561 
562 	if (!ipv6_unicast_destination(skb) &&
563 	    !(net->ipv6.sysctl.anycast_src_echo_reply &&
564 	      ipv6_anycast_destination(skb)))
565 		saddr = NULL;
566 
567 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
568 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
569 
570 	memset(&fl6, 0, sizeof(fl6));
571 	fl6.flowi6_proto = IPPROTO_ICMPV6;
572 	fl6.daddr = ipv6_hdr(skb)->saddr;
573 	if (saddr)
574 		fl6.saddr = *saddr;
575 	fl6.flowi6_oif = skb->dev->ifindex;
576 	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
577 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
578 
579 	sk = icmpv6_xmit_lock(net);
580 	if (sk == NULL)
581 		return;
582 	np = inet6_sk(sk);
583 
584 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
585 		fl6.flowi6_oif = np->mcast_oif;
586 	else if (!fl6.flowi6_oif)
587 		fl6.flowi6_oif = np->ucast_oif;
588 
589 	err = ip6_dst_lookup(sk, &dst, &fl6);
590 	if (err)
591 		goto out;
592 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
593 	if (IS_ERR(dst))
594 		goto out;
595 
596 	if (ipv6_addr_is_multicast(&fl6.daddr))
597 		hlimit = np->mcast_hops;
598 	else
599 		hlimit = np->hop_limit;
600 	if (hlimit < 0)
601 		hlimit = ip6_dst_hoplimit(dst);
602 
603 	idev = __in6_dev_get(skb->dev);
604 
605 	msg.skb = skb;
606 	msg.offset = 0;
607 	msg.type = ICMPV6_ECHO_REPLY;
608 
609 	tclass = ipv6_get_dsfield(ipv6_hdr(skb));
610 	err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
611 				sizeof(struct icmp6hdr), hlimit, tclass, NULL, &fl6,
612 				(struct rt6_info *)dst, MSG_DONTWAIT,
613 				np->dontfrag);
614 
615 	if (err) {
616 		ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTERRORS);
617 		ip6_flush_pending_frames(sk);
618 	} else {
619 		err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
620 						 skb->len + sizeof(struct icmp6hdr));
621 	}
622 	dst_release(dst);
623 out:
624 	icmpv6_xmit_unlock(sk);
625 }
626 
627 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
628 {
629 	const struct inet6_protocol *ipprot;
630 	int inner_offset;
631 	__be16 frag_off;
632 	u8 nexthdr;
633 
634 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
635 		return;
636 
637 	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
638 	if (ipv6_ext_hdr(nexthdr)) {
639 		/* now skip over extension headers */
640 		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
641 						&nexthdr, &frag_off);
642 		if (inner_offset<0)
643 			return;
644 	} else {
645 		inner_offset = sizeof(struct ipv6hdr);
646 	}
647 
648 	/* Checkin header including 8 bytes of inner protocol header. */
649 	if (!pskb_may_pull(skb, inner_offset+8))
650 		return;
651 
652 	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
653 	   Without this we will not able f.e. to make source routed
654 	   pmtu discovery.
655 	   Corresponding argument (opt) to notifiers is already added.
656 	   --ANK (980726)
657 	 */
658 
659 	rcu_read_lock();
660 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
661 	if (ipprot && ipprot->err_handler)
662 		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
663 	rcu_read_unlock();
664 
665 	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
666 }
667 
668 /*
669  *	Handle icmp messages
670  */
671 
672 static int icmpv6_rcv(struct sk_buff *skb)
673 {
674 	struct net_device *dev = skb->dev;
675 	struct inet6_dev *idev = __in6_dev_get(dev);
676 	const struct in6_addr *saddr, *daddr;
677 	struct icmp6hdr *hdr;
678 	u8 type;
679 
680 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
681 		struct sec_path *sp = skb_sec_path(skb);
682 		int nh;
683 
684 		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
685 				 XFRM_STATE_ICMP))
686 			goto drop_no_count;
687 
688 		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
689 			goto drop_no_count;
690 
691 		nh = skb_network_offset(skb);
692 		skb_set_network_header(skb, sizeof(*hdr));
693 
694 		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
695 			goto drop_no_count;
696 
697 		skb_set_network_header(skb, nh);
698 	}
699 
700 	ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INMSGS);
701 
702 	saddr = &ipv6_hdr(skb)->saddr;
703 	daddr = &ipv6_hdr(skb)->daddr;
704 
705 	/* Perform checksum. */
706 	switch (skb->ip_summed) {
707 	case CHECKSUM_COMPLETE:
708 		if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6,
709 				     skb->csum))
710 			break;
711 		/* fall through */
712 	case CHECKSUM_NONE:
713 		skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len,
714 					     IPPROTO_ICMPV6, 0));
715 		if (__skb_checksum_complete(skb)) {
716 			LIMIT_NETDEBUG(KERN_DEBUG
717 				       "ICMPv6 checksum failed [%pI6c > %pI6c]\n",
718 				       saddr, daddr);
719 			goto csum_error;
720 		}
721 	}
722 
723 	if (!pskb_pull(skb, sizeof(*hdr)))
724 		goto discard_it;
725 
726 	hdr = icmp6_hdr(skb);
727 
728 	type = hdr->icmp6_type;
729 
730 	ICMP6MSGIN_INC_STATS_BH(dev_net(dev), idev, type);
731 
732 	switch (type) {
733 	case ICMPV6_ECHO_REQUEST:
734 		icmpv6_echo_reply(skb);
735 		break;
736 
737 	case ICMPV6_ECHO_REPLY:
738 		ping_rcv(skb);
739 		break;
740 
741 	case ICMPV6_PKT_TOOBIG:
742 		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
743 		   standard destination cache. Seems, only "advanced"
744 		   destination cache will allow to solve this problem
745 		   --ANK (980726)
746 		 */
747 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
748 			goto discard_it;
749 		hdr = icmp6_hdr(skb);
750 
751 		/*
752 		 *	Drop through to notify
753 		 */
754 
755 	case ICMPV6_DEST_UNREACH:
756 	case ICMPV6_TIME_EXCEED:
757 	case ICMPV6_PARAMPROB:
758 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
759 		break;
760 
761 	case NDISC_ROUTER_SOLICITATION:
762 	case NDISC_ROUTER_ADVERTISEMENT:
763 	case NDISC_NEIGHBOUR_SOLICITATION:
764 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
765 	case NDISC_REDIRECT:
766 		ndisc_rcv(skb);
767 		break;
768 
769 	case ICMPV6_MGM_QUERY:
770 		igmp6_event_query(skb);
771 		break;
772 
773 	case ICMPV6_MGM_REPORT:
774 		igmp6_event_report(skb);
775 		break;
776 
777 	case ICMPV6_MGM_REDUCTION:
778 	case ICMPV6_NI_QUERY:
779 	case ICMPV6_NI_REPLY:
780 	case ICMPV6_MLD2_REPORT:
781 	case ICMPV6_DHAAD_REQUEST:
782 	case ICMPV6_DHAAD_REPLY:
783 	case ICMPV6_MOBILE_PREFIX_SOL:
784 	case ICMPV6_MOBILE_PREFIX_ADV:
785 		break;
786 
787 	default:
788 		LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n");
789 
790 		/* informational */
791 		if (type & ICMPV6_INFOMSG_MASK)
792 			break;
793 
794 		/*
795 		 * error of unknown type.
796 		 * must pass to upper level
797 		 */
798 
799 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
800 	}
801 
802 	kfree_skb(skb);
803 	return 0;
804 
805 csum_error:
806 	ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
807 discard_it:
808 	ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INERRORS);
809 drop_no_count:
810 	kfree_skb(skb);
811 	return 0;
812 }
813 
814 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
815 		      u8 type,
816 		      const struct in6_addr *saddr,
817 		      const struct in6_addr *daddr,
818 		      int oif)
819 {
820 	memset(fl6, 0, sizeof(*fl6));
821 	fl6->saddr = *saddr;
822 	fl6->daddr = *daddr;
823 	fl6->flowi6_proto 	= IPPROTO_ICMPV6;
824 	fl6->fl6_icmp_type	= type;
825 	fl6->fl6_icmp_code	= 0;
826 	fl6->flowi6_oif		= oif;
827 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
828 }
829 
830 /*
831  * Special lock-class for __icmpv6_sk:
832  */
833 static struct lock_class_key icmpv6_socket_sk_dst_lock_key;
834 
835 static int __net_init icmpv6_sk_init(struct net *net)
836 {
837 	struct sock *sk;
838 	int err, i, j;
839 
840 	net->ipv6.icmp_sk =
841 		kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
842 	if (net->ipv6.icmp_sk == NULL)
843 		return -ENOMEM;
844 
845 	for_each_possible_cpu(i) {
846 		err = inet_ctl_sock_create(&sk, PF_INET6,
847 					   SOCK_RAW, IPPROTO_ICMPV6, net);
848 		if (err < 0) {
849 			pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
850 			       err);
851 			goto fail;
852 		}
853 
854 		net->ipv6.icmp_sk[i] = sk;
855 
856 		/*
857 		 * Split off their lock-class, because sk->sk_dst_lock
858 		 * gets used from softirqs, which is safe for
859 		 * __icmpv6_sk (because those never get directly used
860 		 * via userspace syscalls), but unsafe for normal sockets.
861 		 */
862 		lockdep_set_class(&sk->sk_dst_lock,
863 				  &icmpv6_socket_sk_dst_lock_key);
864 
865 		/* Enough space for 2 64K ICMP packets, including
866 		 * sk_buff struct overhead.
867 		 */
868 		sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
869 	}
870 	return 0;
871 
872  fail:
873 	for (j = 0; j < i; j++)
874 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
875 	kfree(net->ipv6.icmp_sk);
876 	return err;
877 }
878 
879 static void __net_exit icmpv6_sk_exit(struct net *net)
880 {
881 	int i;
882 
883 	for_each_possible_cpu(i) {
884 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
885 	}
886 	kfree(net->ipv6.icmp_sk);
887 }
888 
889 static struct pernet_operations icmpv6_sk_ops = {
890        .init = icmpv6_sk_init,
891        .exit = icmpv6_sk_exit,
892 };
893 
894 int __init icmpv6_init(void)
895 {
896 	int err;
897 
898 	err = register_pernet_subsys(&icmpv6_sk_ops);
899 	if (err < 0)
900 		return err;
901 
902 	err = -EAGAIN;
903 	if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
904 		goto fail;
905 
906 	err = inet6_register_icmp_sender(icmp6_send);
907 	if (err)
908 		goto sender_reg_err;
909 	return 0;
910 
911 sender_reg_err:
912 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
913 fail:
914 	pr_err("Failed to register ICMP6 protocol\n");
915 	unregister_pernet_subsys(&icmpv6_sk_ops);
916 	return err;
917 }
918 
919 void icmpv6_cleanup(void)
920 {
921 	inet6_unregister_icmp_sender(icmp6_send);
922 	unregister_pernet_subsys(&icmpv6_sk_ops);
923 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
924 }
925 
926 
927 static const struct icmp6_err {
928 	int err;
929 	int fatal;
930 } tab_unreach[] = {
931 	{	/* NOROUTE */
932 		.err	= ENETUNREACH,
933 		.fatal	= 0,
934 	},
935 	{	/* ADM_PROHIBITED */
936 		.err	= EACCES,
937 		.fatal	= 1,
938 	},
939 	{	/* Was NOT_NEIGHBOUR, now reserved */
940 		.err	= EHOSTUNREACH,
941 		.fatal	= 0,
942 	},
943 	{	/* ADDR_UNREACH	*/
944 		.err	= EHOSTUNREACH,
945 		.fatal	= 0,
946 	},
947 	{	/* PORT_UNREACH	*/
948 		.err	= ECONNREFUSED,
949 		.fatal	= 1,
950 	},
951 	{	/* POLICY_FAIL */
952 		.err	= EACCES,
953 		.fatal	= 1,
954 	},
955 	{	/* REJECT_ROUTE	*/
956 		.err	= EACCES,
957 		.fatal	= 1,
958 	},
959 };
960 
961 int icmpv6_err_convert(u8 type, u8 code, int *err)
962 {
963 	int fatal = 0;
964 
965 	*err = EPROTO;
966 
967 	switch (type) {
968 	case ICMPV6_DEST_UNREACH:
969 		fatal = 1;
970 		if (code < ARRAY_SIZE(tab_unreach)) {
971 			*err  = tab_unreach[code].err;
972 			fatal = tab_unreach[code].fatal;
973 		}
974 		break;
975 
976 	case ICMPV6_PKT_TOOBIG:
977 		*err = EMSGSIZE;
978 		break;
979 
980 	case ICMPV6_PARAMPROB:
981 		*err = EPROTO;
982 		fatal = 1;
983 		break;
984 
985 	case ICMPV6_TIME_EXCEED:
986 		*err = EHOSTUNREACH;
987 		break;
988 	}
989 
990 	return fatal;
991 }
992 EXPORT_SYMBOL(icmpv6_err_convert);
993 
994 #ifdef CONFIG_SYSCTL
995 static struct ctl_table ipv6_icmp_table_template[] = {
996 	{
997 		.procname	= "ratelimit",
998 		.data		= &init_net.ipv6.sysctl.icmpv6_time,
999 		.maxlen		= sizeof(int),
1000 		.mode		= 0644,
1001 		.proc_handler	= proc_dointvec_ms_jiffies,
1002 	},
1003 	{ },
1004 };
1005 
1006 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1007 {
1008 	struct ctl_table *table;
1009 
1010 	table = kmemdup(ipv6_icmp_table_template,
1011 			sizeof(ipv6_icmp_table_template),
1012 			GFP_KERNEL);
1013 
1014 	if (table)
1015 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
1016 
1017 	return table;
1018 }
1019 #endif
1020 
1021