xref: /openbmc/linux/net/ipv6/icmp.c (revision 78700c0a)
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 #include <net/l3mdev.h>
72 
73 #include <asm/uaccess.h>
74 
75 /*
76  *	The ICMP socket(s). This is the most convenient way to flow control
77  *	our ICMP output as well as maintain a clean interface throughout
78  *	all layers. All Socketless IP sends will soon be gone.
79  *
80  *	On SMP we have one ICMP socket per-cpu.
81  */
82 static inline struct sock *icmpv6_sk(struct net *net)
83 {
84 	return net->ipv6.icmp_sk[smp_processor_id()];
85 }
86 
87 static void icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
88 		       u8 type, u8 code, int offset, __be32 info)
89 {
90 	/* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */
91 	struct icmp6hdr *icmp6 = (struct icmp6hdr *) (skb->data + offset);
92 	struct net *net = dev_net(skb->dev);
93 
94 	if (type == ICMPV6_PKT_TOOBIG)
95 		ip6_update_pmtu(skb, net, info, 0, 0);
96 	else if (type == NDISC_REDIRECT)
97 		ip6_redirect(skb, net, skb->dev->ifindex, 0);
98 
99 	if (!(type & ICMPV6_INFOMSG_MASK))
100 		if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST)
101 			ping_err(skb, offset, info);
102 }
103 
104 static int icmpv6_rcv(struct sk_buff *skb);
105 
106 static const struct inet6_protocol icmpv6_protocol = {
107 	.handler	=	icmpv6_rcv,
108 	.err_handler	=	icmpv6_err,
109 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
110 };
111 
112 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
113 {
114 	struct sock *sk;
115 
116 	local_bh_disable();
117 
118 	sk = icmpv6_sk(net);
119 	if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
120 		/* This can happen if the output path (f.e. SIT or
121 		 * ip6ip6 tunnel) signals dst_link_failure() for an
122 		 * outgoing ICMP6 packet.
123 		 */
124 		local_bh_enable();
125 		return NULL;
126 	}
127 	return sk;
128 }
129 
130 static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
131 {
132 	spin_unlock_bh(&sk->sk_lock.slock);
133 }
134 
135 /*
136  * Figure out, may we reply to this packet with icmp error.
137  *
138  * We do not reply, if:
139  *	- it was icmp error message.
140  *	- it is truncated, so that it is known, that protocol is ICMPV6
141  *	  (i.e. in the middle of some exthdr)
142  *
143  *	--ANK (980726)
144  */
145 
146 static bool is_ineligible(const struct sk_buff *skb)
147 {
148 	int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
149 	int len = skb->len - ptr;
150 	__u8 nexthdr = ipv6_hdr(skb)->nexthdr;
151 	__be16 frag_off;
152 
153 	if (len < 0)
154 		return true;
155 
156 	ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off);
157 	if (ptr < 0)
158 		return false;
159 	if (nexthdr == IPPROTO_ICMPV6) {
160 		u8 _type, *tp;
161 		tp = skb_header_pointer(skb,
162 			ptr+offsetof(struct icmp6hdr, icmp6_type),
163 			sizeof(_type), &_type);
164 		if (!tp || !(*tp & ICMPV6_INFOMSG_MASK))
165 			return true;
166 	}
167 	return false;
168 }
169 
170 /*
171  * Check the ICMP output rate limit
172  */
173 static bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
174 			       struct flowi6 *fl6)
175 {
176 	struct net *net = sock_net(sk);
177 	struct dst_entry *dst;
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 
203 		/* Give more bandwidth to wider prefixes. */
204 		if (rt->rt6i_dst.plen < 128)
205 			tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
206 
207 		if (icmp_global_allow()) {
208 			struct inet_peer *peer;
209 
210 			peer = inet_getpeer_v6(net->ipv6.peers,
211 					       &fl6->daddr, 1);
212 			res = inet_peer_xrlim_allow(peer, tmo);
213 			if (peer)
214 				inet_putpeer(peer);
215 		}
216 	}
217 	dst_release(dst);
218 	return res;
219 }
220 
221 /*
222  *	an inline helper for the "simple" if statement below
223  *	checks if parameter problem report is caused by an
224  *	unrecognized IPv6 option that has the Option Type
225  *	highest-order two bits set to 10
226  */
227 
228 static bool opt_unrec(struct sk_buff *skb, __u32 offset)
229 {
230 	u8 _optval, *op;
231 
232 	offset += skb_network_offset(skb);
233 	op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
234 	if (!op)
235 		return true;
236 	return (*op & 0xC0) == 0x80;
237 }
238 
239 int icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
240 			       struct icmp6hdr *thdr, int len)
241 {
242 	struct sk_buff *skb;
243 	struct icmp6hdr *icmp6h;
244 	int err = 0;
245 
246 	skb = skb_peek(&sk->sk_write_queue);
247 	if (!skb)
248 		goto out;
249 
250 	icmp6h = icmp6_hdr(skb);
251 	memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
252 	icmp6h->icmp6_cksum = 0;
253 
254 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
255 		skb->csum = csum_partial(icmp6h,
256 					sizeof(struct icmp6hdr), skb->csum);
257 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
258 						      &fl6->daddr,
259 						      len, fl6->flowi6_proto,
260 						      skb->csum);
261 	} else {
262 		__wsum tmp_csum = 0;
263 
264 		skb_queue_walk(&sk->sk_write_queue, skb) {
265 			tmp_csum = csum_add(tmp_csum, skb->csum);
266 		}
267 
268 		tmp_csum = csum_partial(icmp6h,
269 					sizeof(struct icmp6hdr), tmp_csum);
270 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
271 						      &fl6->daddr,
272 						      len, fl6->flowi6_proto,
273 						      tmp_csum);
274 	}
275 	ip6_push_pending_frames(sk);
276 out:
277 	return err;
278 }
279 
280 struct icmpv6_msg {
281 	struct sk_buff	*skb;
282 	int		offset;
283 	uint8_t		type;
284 };
285 
286 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
287 {
288 	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
289 	struct sk_buff *org_skb = msg->skb;
290 	__wsum csum = 0;
291 
292 	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
293 				      to, len, csum);
294 	skb->csum = csum_block_add(skb->csum, csum, odd);
295 	if (!(msg->type & ICMPV6_INFOMSG_MASK))
296 		nf_ct_attach(skb, org_skb);
297 	return 0;
298 }
299 
300 #if IS_ENABLED(CONFIG_IPV6_MIP6)
301 static void mip6_addr_swap(struct sk_buff *skb)
302 {
303 	struct ipv6hdr *iph = ipv6_hdr(skb);
304 	struct inet6_skb_parm *opt = IP6CB(skb);
305 	struct ipv6_destopt_hao *hao;
306 	struct in6_addr tmp;
307 	int off;
308 
309 	if (opt->dsthao) {
310 		off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
311 		if (likely(off >= 0)) {
312 			hao = (struct ipv6_destopt_hao *)
313 					(skb_network_header(skb) + off);
314 			tmp = iph->saddr;
315 			iph->saddr = hao->addr;
316 			hao->addr = tmp;
317 		}
318 	}
319 }
320 #else
321 static inline void mip6_addr_swap(struct sk_buff *skb) {}
322 #endif
323 
324 static struct dst_entry *icmpv6_route_lookup(struct net *net,
325 					     struct sk_buff *skb,
326 					     struct sock *sk,
327 					     struct flowi6 *fl6)
328 {
329 	struct dst_entry *dst, *dst2;
330 	struct flowi6 fl2;
331 	int err;
332 
333 	err = ip6_dst_lookup(net, sk, &dst, fl6);
334 	if (err)
335 		return ERR_PTR(err);
336 
337 	/*
338 	 * We won't send icmp if the destination is known
339 	 * anycast.
340 	 */
341 	if (ipv6_anycast_destination(dst, &fl6->daddr)) {
342 		net_dbg_ratelimited("icmp6_send: acast source\n");
343 		dst_release(dst);
344 		return ERR_PTR(-EINVAL);
345 	}
346 
347 	/* No need to clone since we're just using its address. */
348 	dst2 = dst;
349 
350 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
351 	if (!IS_ERR(dst)) {
352 		if (dst != dst2)
353 			return dst;
354 	} else {
355 		if (PTR_ERR(dst) == -EPERM)
356 			dst = NULL;
357 		else
358 			return dst;
359 	}
360 
361 	err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
362 	if (err)
363 		goto relookup_failed;
364 
365 	err = ip6_dst_lookup(net, sk, &dst2, &fl2);
366 	if (err)
367 		goto relookup_failed;
368 
369 	dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
370 	if (!IS_ERR(dst2)) {
371 		dst_release(dst);
372 		dst = dst2;
373 	} else {
374 		err = PTR_ERR(dst2);
375 		if (err == -EPERM) {
376 			dst_release(dst);
377 			return dst2;
378 		} else
379 			goto relookup_failed;
380 	}
381 
382 relookup_failed:
383 	if (dst)
384 		return dst;
385 	return ERR_PTR(err);
386 }
387 
388 /*
389  *	Send an ICMP message in response to a packet in error
390  */
391 static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info)
392 {
393 	struct net *net = dev_net(skb->dev);
394 	struct inet6_dev *idev = NULL;
395 	struct ipv6hdr *hdr = ipv6_hdr(skb);
396 	struct sock *sk;
397 	struct ipv6_pinfo *np;
398 	const struct in6_addr *saddr = NULL;
399 	struct dst_entry *dst;
400 	struct icmp6hdr tmp_hdr;
401 	struct flowi6 fl6;
402 	struct icmpv6_msg msg;
403 	struct sockcm_cookie sockc_unused = {0};
404 	struct ipcm6_cookie ipc6;
405 	int iif = 0;
406 	int addr_type = 0;
407 	int len;
408 	int err = 0;
409 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
410 
411 	if ((u8 *)hdr < skb->head ||
412 	    (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb))
413 		return;
414 
415 	/*
416 	 *	Make sure we respect the rules
417 	 *	i.e. RFC 1885 2.4(e)
418 	 *	Rule (e.1) is enforced by not using icmp6_send
419 	 *	in any code that processes icmp errors.
420 	 */
421 	addr_type = ipv6_addr_type(&hdr->daddr);
422 
423 	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) ||
424 	    ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr))
425 		saddr = &hdr->daddr;
426 
427 	/*
428 	 *	Dest addr check
429 	 */
430 
431 	if (addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST) {
432 		if (type != ICMPV6_PKT_TOOBIG &&
433 		    !(type == ICMPV6_PARAMPROB &&
434 		      code == ICMPV6_UNK_OPTION &&
435 		      (opt_unrec(skb, info))))
436 			return;
437 
438 		saddr = NULL;
439 	}
440 
441 	addr_type = ipv6_addr_type(&hdr->saddr);
442 
443 	/*
444 	 *	Source addr check
445 	 */
446 
447 	if (__ipv6_addr_needs_scope_id(addr_type))
448 		iif = skb->dev->ifindex;
449 	else
450 		iif = l3mdev_master_ifindex(skb->dev);
451 
452 	/*
453 	 *	Must not send error if the source does not uniquely
454 	 *	identify a single node (RFC2463 Section 2.4).
455 	 *	We check unspecified / multicast addresses here,
456 	 *	and anycast addresses will be checked later.
457 	 */
458 	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
459 		net_dbg_ratelimited("icmp6_send: addr_any/mcast source [%pI6c > %pI6c]\n",
460 				    &hdr->saddr, &hdr->daddr);
461 		return;
462 	}
463 
464 	/*
465 	 *	Never answer to a ICMP packet.
466 	 */
467 	if (is_ineligible(skb)) {
468 		net_dbg_ratelimited("icmp6_send: no reply to icmp error [%pI6c > %pI6c]\n",
469 				    &hdr->saddr, &hdr->daddr);
470 		return;
471 	}
472 
473 	mip6_addr_swap(skb);
474 
475 	memset(&fl6, 0, sizeof(fl6));
476 	fl6.flowi6_proto = IPPROTO_ICMPV6;
477 	fl6.daddr = hdr->saddr;
478 	if (saddr)
479 		fl6.saddr = *saddr;
480 	fl6.flowi6_mark = mark;
481 	fl6.flowi6_oif = iif;
482 	fl6.fl6_icmp_type = type;
483 	fl6.fl6_icmp_code = code;
484 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
485 
486 	sk = icmpv6_xmit_lock(net);
487 	if (!sk)
488 		return;
489 	sk->sk_mark = mark;
490 	np = inet6_sk(sk);
491 
492 	if (!icmpv6_xrlim_allow(sk, type, &fl6))
493 		goto out;
494 
495 	tmp_hdr.icmp6_type = type;
496 	tmp_hdr.icmp6_code = code;
497 	tmp_hdr.icmp6_cksum = 0;
498 	tmp_hdr.icmp6_pointer = htonl(info);
499 
500 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
501 		fl6.flowi6_oif = np->mcast_oif;
502 	else if (!fl6.flowi6_oif)
503 		fl6.flowi6_oif = np->ucast_oif;
504 
505 	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
506 	if (IS_ERR(dst))
507 		goto out;
508 
509 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
510 	ipc6.tclass = np->tclass;
511 	ipc6.dontfrag = np->dontfrag;
512 	ipc6.opt = NULL;
513 
514 	msg.skb = skb;
515 	msg.offset = skb_network_offset(skb);
516 	msg.type = type;
517 
518 	len = skb->len - msg.offset;
519 	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(struct icmp6hdr));
520 	if (len < 0) {
521 		net_dbg_ratelimited("icmp: len problem [%pI6c > %pI6c]\n",
522 				    &hdr->saddr, &hdr->daddr);
523 		goto out_dst_release;
524 	}
525 
526 	rcu_read_lock();
527 	idev = __in6_dev_get(skb->dev);
528 
529 	err = ip6_append_data(sk, icmpv6_getfrag, &msg,
530 			      len + sizeof(struct icmp6hdr),
531 			      sizeof(struct icmp6hdr),
532 			      &ipc6, &fl6, (struct rt6_info *)dst,
533 			      MSG_DONTWAIT, &sockc_unused);
534 	if (err) {
535 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
536 		ip6_flush_pending_frames(sk);
537 	} else {
538 		err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
539 						 len + sizeof(struct icmp6hdr));
540 	}
541 	rcu_read_unlock();
542 out_dst_release:
543 	dst_release(dst);
544 out:
545 	icmpv6_xmit_unlock(sk);
546 }
547 
548 /* Slightly more convenient version of icmp6_send.
549  */
550 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
551 {
552 	icmp6_send(skb, ICMPV6_PARAMPROB, code, pos);
553 	kfree_skb(skb);
554 }
555 
556 static void icmpv6_echo_reply(struct sk_buff *skb)
557 {
558 	struct net *net = dev_net(skb->dev);
559 	struct sock *sk;
560 	struct inet6_dev *idev;
561 	struct ipv6_pinfo *np;
562 	const struct in6_addr *saddr = NULL;
563 	struct icmp6hdr *icmph = icmp6_hdr(skb);
564 	struct icmp6hdr tmp_hdr;
565 	struct flowi6 fl6;
566 	struct icmpv6_msg msg;
567 	struct dst_entry *dst;
568 	struct ipcm6_cookie ipc6;
569 	int err = 0;
570 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
571 	struct sockcm_cookie sockc_unused = {0};
572 
573 	saddr = &ipv6_hdr(skb)->daddr;
574 
575 	if (!ipv6_unicast_destination(skb) &&
576 	    !(net->ipv6.sysctl.anycast_src_echo_reply &&
577 	      ipv6_anycast_destination(skb_dst(skb), saddr)))
578 		saddr = NULL;
579 
580 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
581 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
582 
583 	memset(&fl6, 0, sizeof(fl6));
584 	fl6.flowi6_proto = IPPROTO_ICMPV6;
585 	fl6.daddr = ipv6_hdr(skb)->saddr;
586 	if (saddr)
587 		fl6.saddr = *saddr;
588 	fl6.flowi6_oif = l3mdev_fib_oif(skb->dev);
589 	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
590 	fl6.flowi6_mark = mark;
591 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
592 
593 	sk = icmpv6_xmit_lock(net);
594 	if (!sk)
595 		return;
596 	sk->sk_mark = mark;
597 	np = inet6_sk(sk);
598 
599 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
600 		fl6.flowi6_oif = np->mcast_oif;
601 	else if (!fl6.flowi6_oif)
602 		fl6.flowi6_oif = np->ucast_oif;
603 
604 	err = ip6_dst_lookup(net, sk, &dst, &fl6);
605 	if (err)
606 		goto out;
607 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
608 	if (IS_ERR(dst))
609 		goto out;
610 
611 	idev = __in6_dev_get(skb->dev);
612 
613 	msg.skb = skb;
614 	msg.offset = 0;
615 	msg.type = ICMPV6_ECHO_REPLY;
616 
617 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
618 	ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb));
619 	ipc6.dontfrag = np->dontfrag;
620 	ipc6.opt = NULL;
621 
622 	err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
623 				sizeof(struct icmp6hdr), &ipc6, &fl6,
624 				(struct rt6_info *)dst, MSG_DONTWAIT,
625 				&sockc_unused);
626 
627 	if (err) {
628 		__ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
629 		ip6_flush_pending_frames(sk);
630 	} else {
631 		err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
632 						 skb->len + sizeof(struct icmp6hdr));
633 	}
634 	dst_release(dst);
635 out:
636 	icmpv6_xmit_unlock(sk);
637 }
638 
639 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
640 {
641 	const struct inet6_protocol *ipprot;
642 	int inner_offset;
643 	__be16 frag_off;
644 	u8 nexthdr;
645 	struct net *net = dev_net(skb->dev);
646 
647 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
648 		goto out;
649 
650 	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
651 	if (ipv6_ext_hdr(nexthdr)) {
652 		/* now skip over extension headers */
653 		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
654 						&nexthdr, &frag_off);
655 		if (inner_offset < 0)
656 			goto out;
657 	} else {
658 		inner_offset = sizeof(struct ipv6hdr);
659 	}
660 
661 	/* Checkin header including 8 bytes of inner protocol header. */
662 	if (!pskb_may_pull(skb, inner_offset+8))
663 		goto out;
664 
665 	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
666 	   Without this we will not able f.e. to make source routed
667 	   pmtu discovery.
668 	   Corresponding argument (opt) to notifiers is already added.
669 	   --ANK (980726)
670 	 */
671 
672 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
673 	if (ipprot && ipprot->err_handler)
674 		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
675 
676 	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
677 	return;
678 
679 out:
680 	__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
681 }
682 
683 /*
684  *	Handle icmp messages
685  */
686 
687 static int icmpv6_rcv(struct sk_buff *skb)
688 {
689 	struct net_device *dev = skb->dev;
690 	struct inet6_dev *idev = __in6_dev_get(dev);
691 	const struct in6_addr *saddr, *daddr;
692 	struct icmp6hdr *hdr;
693 	u8 type;
694 	bool success = false;
695 
696 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
697 		struct sec_path *sp = skb_sec_path(skb);
698 		int nh;
699 
700 		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
701 				 XFRM_STATE_ICMP))
702 			goto drop_no_count;
703 
704 		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
705 			goto drop_no_count;
706 
707 		nh = skb_network_offset(skb);
708 		skb_set_network_header(skb, sizeof(*hdr));
709 
710 		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
711 			goto drop_no_count;
712 
713 		skb_set_network_header(skb, nh);
714 	}
715 
716 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS);
717 
718 	saddr = &ipv6_hdr(skb)->saddr;
719 	daddr = &ipv6_hdr(skb)->daddr;
720 
721 	if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) {
722 		net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n",
723 				    saddr, daddr);
724 		goto csum_error;
725 	}
726 
727 	if (!pskb_pull(skb, sizeof(*hdr)))
728 		goto discard_it;
729 
730 	hdr = icmp6_hdr(skb);
731 
732 	type = hdr->icmp6_type;
733 
734 	ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type);
735 
736 	switch (type) {
737 	case ICMPV6_ECHO_REQUEST:
738 		icmpv6_echo_reply(skb);
739 		break;
740 
741 	case ICMPV6_ECHO_REPLY:
742 		success = ping_rcv(skb);
743 		break;
744 
745 	case ICMPV6_PKT_TOOBIG:
746 		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
747 		   standard destination cache. Seems, only "advanced"
748 		   destination cache will allow to solve this problem
749 		   --ANK (980726)
750 		 */
751 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
752 			goto discard_it;
753 		hdr = icmp6_hdr(skb);
754 
755 		/*
756 		 *	Drop through to notify
757 		 */
758 
759 	case ICMPV6_DEST_UNREACH:
760 	case ICMPV6_TIME_EXCEED:
761 	case ICMPV6_PARAMPROB:
762 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
763 		break;
764 
765 	case NDISC_ROUTER_SOLICITATION:
766 	case NDISC_ROUTER_ADVERTISEMENT:
767 	case NDISC_NEIGHBOUR_SOLICITATION:
768 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
769 	case NDISC_REDIRECT:
770 		ndisc_rcv(skb);
771 		break;
772 
773 	case ICMPV6_MGM_QUERY:
774 		igmp6_event_query(skb);
775 		break;
776 
777 	case ICMPV6_MGM_REPORT:
778 		igmp6_event_report(skb);
779 		break;
780 
781 	case ICMPV6_MGM_REDUCTION:
782 	case ICMPV6_NI_QUERY:
783 	case ICMPV6_NI_REPLY:
784 	case ICMPV6_MLD2_REPORT:
785 	case ICMPV6_DHAAD_REQUEST:
786 	case ICMPV6_DHAAD_REPLY:
787 	case ICMPV6_MOBILE_PREFIX_SOL:
788 	case ICMPV6_MOBILE_PREFIX_ADV:
789 		break;
790 
791 	default:
792 		/* informational */
793 		if (type & ICMPV6_INFOMSG_MASK)
794 			break;
795 
796 		net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n",
797 				    saddr, daddr);
798 
799 		/*
800 		 * error of unknown type.
801 		 * must pass to upper level
802 		 */
803 
804 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
805 	}
806 
807 	/* until the v6 path can be better sorted assume failure and
808 	 * preserve the status quo behaviour for the rest of the paths to here
809 	 */
810 	if (success)
811 		consume_skb(skb);
812 	else
813 		kfree_skb(skb);
814 
815 	return 0;
816 
817 csum_error:
818 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
819 discard_it:
820 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS);
821 drop_no_count:
822 	kfree_skb(skb);
823 	return 0;
824 }
825 
826 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
827 		      u8 type,
828 		      const struct in6_addr *saddr,
829 		      const struct in6_addr *daddr,
830 		      int oif)
831 {
832 	memset(fl6, 0, sizeof(*fl6));
833 	fl6->saddr = *saddr;
834 	fl6->daddr = *daddr;
835 	fl6->flowi6_proto	= IPPROTO_ICMPV6;
836 	fl6->fl6_icmp_type	= type;
837 	fl6->fl6_icmp_code	= 0;
838 	fl6->flowi6_oif		= oif;
839 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
840 }
841 
842 static int __net_init icmpv6_sk_init(struct net *net)
843 {
844 	struct sock *sk;
845 	int err, i, j;
846 
847 	net->ipv6.icmp_sk =
848 		kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
849 	if (!net->ipv6.icmp_sk)
850 		return -ENOMEM;
851 
852 	for_each_possible_cpu(i) {
853 		err = inet_ctl_sock_create(&sk, PF_INET6,
854 					   SOCK_RAW, IPPROTO_ICMPV6, net);
855 		if (err < 0) {
856 			pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
857 			       err);
858 			goto fail;
859 		}
860 
861 		net->ipv6.icmp_sk[i] = sk;
862 
863 		/* Enough space for 2 64K ICMP packets, including
864 		 * sk_buff struct overhead.
865 		 */
866 		sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
867 	}
868 	return 0;
869 
870  fail:
871 	for (j = 0; j < i; j++)
872 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
873 	kfree(net->ipv6.icmp_sk);
874 	return err;
875 }
876 
877 static void __net_exit icmpv6_sk_exit(struct net *net)
878 {
879 	int i;
880 
881 	for_each_possible_cpu(i) {
882 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
883 	}
884 	kfree(net->ipv6.icmp_sk);
885 }
886 
887 static struct pernet_operations icmpv6_sk_ops = {
888 	.init = icmpv6_sk_init,
889 	.exit = icmpv6_sk_exit,
890 };
891 
892 int __init icmpv6_init(void)
893 {
894 	int err;
895 
896 	err = register_pernet_subsys(&icmpv6_sk_ops);
897 	if (err < 0)
898 		return err;
899 
900 	err = -EAGAIN;
901 	if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
902 		goto fail;
903 
904 	err = inet6_register_icmp_sender(icmp6_send);
905 	if (err)
906 		goto sender_reg_err;
907 	return 0;
908 
909 sender_reg_err:
910 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
911 fail:
912 	pr_err("Failed to register ICMP6 protocol\n");
913 	unregister_pernet_subsys(&icmpv6_sk_ops);
914 	return err;
915 }
916 
917 void icmpv6_cleanup(void)
918 {
919 	inet6_unregister_icmp_sender(icmp6_send);
920 	unregister_pernet_subsys(&icmpv6_sk_ops);
921 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
922 }
923 
924 
925 static const struct icmp6_err {
926 	int err;
927 	int fatal;
928 } tab_unreach[] = {
929 	{	/* NOROUTE */
930 		.err	= ENETUNREACH,
931 		.fatal	= 0,
932 	},
933 	{	/* ADM_PROHIBITED */
934 		.err	= EACCES,
935 		.fatal	= 1,
936 	},
937 	{	/* Was NOT_NEIGHBOUR, now reserved */
938 		.err	= EHOSTUNREACH,
939 		.fatal	= 0,
940 	},
941 	{	/* ADDR_UNREACH	*/
942 		.err	= EHOSTUNREACH,
943 		.fatal	= 0,
944 	},
945 	{	/* PORT_UNREACH	*/
946 		.err	= ECONNREFUSED,
947 		.fatal	= 1,
948 	},
949 	{	/* POLICY_FAIL */
950 		.err	= EACCES,
951 		.fatal	= 1,
952 	},
953 	{	/* REJECT_ROUTE	*/
954 		.err	= EACCES,
955 		.fatal	= 1,
956 	},
957 };
958 
959 int icmpv6_err_convert(u8 type, u8 code, int *err)
960 {
961 	int fatal = 0;
962 
963 	*err = EPROTO;
964 
965 	switch (type) {
966 	case ICMPV6_DEST_UNREACH:
967 		fatal = 1;
968 		if (code < ARRAY_SIZE(tab_unreach)) {
969 			*err  = tab_unreach[code].err;
970 			fatal = tab_unreach[code].fatal;
971 		}
972 		break;
973 
974 	case ICMPV6_PKT_TOOBIG:
975 		*err = EMSGSIZE;
976 		break;
977 
978 	case ICMPV6_PARAMPROB:
979 		*err = EPROTO;
980 		fatal = 1;
981 		break;
982 
983 	case ICMPV6_TIME_EXCEED:
984 		*err = EHOSTUNREACH;
985 		break;
986 	}
987 
988 	return fatal;
989 }
990 EXPORT_SYMBOL(icmpv6_err_convert);
991 
992 #ifdef CONFIG_SYSCTL
993 static struct ctl_table ipv6_icmp_table_template[] = {
994 	{
995 		.procname	= "ratelimit",
996 		.data		= &init_net.ipv6.sysctl.icmpv6_time,
997 		.maxlen		= sizeof(int),
998 		.mode		= 0644,
999 		.proc_handler	= proc_dointvec_ms_jiffies,
1000 	},
1001 	{ },
1002 };
1003 
1004 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1005 {
1006 	struct ctl_table *table;
1007 
1008 	table = kmemdup(ipv6_icmp_table_template,
1009 			sizeof(ipv6_icmp_table_template),
1010 			GFP_KERNEL);
1011 
1012 	if (table)
1013 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
1014 
1015 	return table;
1016 }
1017 #endif
1018