xref: /linux/net/ipv6/icmp.c (revision f79e4d5f92a129a1159c973735007d4ddc8541f3)
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 <linux/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, sock_net_uid(net, NULL));
96 	else if (type == NDISC_REDIRECT)
97 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
98 			     sock_net_uid(net, NULL));
99 
100 	if (!(type & ICMPV6_INFOMSG_MASK))
101 		if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST)
102 			ping_err(skb, offset, ntohl(info));
103 }
104 
105 static int icmpv6_rcv(struct sk_buff *skb);
106 
107 static const struct inet6_protocol icmpv6_protocol = {
108 	.handler	=	icmpv6_rcv,
109 	.err_handler	=	icmpv6_err,
110 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
111 };
112 
113 /* Called with BH disabled */
114 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
115 {
116 	struct sock *sk;
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 		return NULL;
125 	}
126 	return sk;
127 }
128 
129 static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
130 {
131 	spin_unlock(&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 || !(*tp & ICMPV6_INFOMSG_MASK))
164 			return true;
165 	}
166 	return false;
167 }
168 
169 static bool icmpv6_mask_allow(int type)
170 {
171 	/* Informational messages are not limited. */
172 	if (type & ICMPV6_INFOMSG_MASK)
173 		return true;
174 
175 	/* Do not limit pmtu discovery, it would break it. */
176 	if (type == ICMPV6_PKT_TOOBIG)
177 		return true;
178 
179 	return false;
180 }
181 
182 static bool icmpv6_global_allow(int type)
183 {
184 	if (icmpv6_mask_allow(type))
185 		return true;
186 
187 	if (icmp_global_allow())
188 		return true;
189 
190 	return false;
191 }
192 
193 /*
194  * Check the ICMP output rate limit
195  */
196 static bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
197 			       struct flowi6 *fl6)
198 {
199 	struct net *net = sock_net(sk);
200 	struct dst_entry *dst;
201 	bool res = false;
202 
203 	if (icmpv6_mask_allow(type))
204 		return true;
205 
206 	/*
207 	 * Look up the output route.
208 	 * XXX: perhaps the expire for routing entries cloned by
209 	 * this lookup should be more aggressive (not longer than timeout).
210 	 */
211 	dst = ip6_route_output(net, sk, fl6);
212 	if (dst->error) {
213 		IP6_INC_STATS(net, ip6_dst_idev(dst),
214 			      IPSTATS_MIB_OUTNOROUTES);
215 	} else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
216 		res = true;
217 	} else {
218 		struct rt6_info *rt = (struct rt6_info *)dst;
219 		int tmo = net->ipv6.sysctl.icmpv6_time;
220 		struct inet_peer *peer;
221 
222 		/* Give more bandwidth to wider prefixes. */
223 		if (rt->rt6i_dst.plen < 128)
224 			tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
225 
226 		peer = inet_getpeer_v6(net->ipv6.peers, &fl6->daddr, 1);
227 		res = inet_peer_xrlim_allow(peer, tmo);
228 		if (peer)
229 			inet_putpeer(peer);
230 	}
231 	dst_release(dst);
232 	return res;
233 }
234 
235 /*
236  *	an inline helper for the "simple" if statement below
237  *	checks if parameter problem report is caused by an
238  *	unrecognized IPv6 option that has the Option Type
239  *	highest-order two bits set to 10
240  */
241 
242 static bool opt_unrec(struct sk_buff *skb, __u32 offset)
243 {
244 	u8 _optval, *op;
245 
246 	offset += skb_network_offset(skb);
247 	op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
248 	if (!op)
249 		return true;
250 	return (*op & 0xC0) == 0x80;
251 }
252 
253 void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
254 				struct icmp6hdr *thdr, int len)
255 {
256 	struct sk_buff *skb;
257 	struct icmp6hdr *icmp6h;
258 
259 	skb = skb_peek(&sk->sk_write_queue);
260 	if (!skb)
261 		return;
262 
263 	icmp6h = icmp6_hdr(skb);
264 	memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
265 	icmp6h->icmp6_cksum = 0;
266 
267 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
268 		skb->csum = csum_partial(icmp6h,
269 					sizeof(struct icmp6hdr), skb->csum);
270 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
271 						      &fl6->daddr,
272 						      len, fl6->flowi6_proto,
273 						      skb->csum);
274 	} else {
275 		__wsum tmp_csum = 0;
276 
277 		skb_queue_walk(&sk->sk_write_queue, skb) {
278 			tmp_csum = csum_add(tmp_csum, skb->csum);
279 		}
280 
281 		tmp_csum = csum_partial(icmp6h,
282 					sizeof(struct icmp6hdr), tmp_csum);
283 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
284 						      &fl6->daddr,
285 						      len, fl6->flowi6_proto,
286 						      tmp_csum);
287 	}
288 	ip6_push_pending_frames(sk);
289 }
290 
291 struct icmpv6_msg {
292 	struct sk_buff	*skb;
293 	int		offset;
294 	uint8_t		type;
295 };
296 
297 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
298 {
299 	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
300 	struct sk_buff *org_skb = msg->skb;
301 	__wsum csum = 0;
302 
303 	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
304 				      to, len, csum);
305 	skb->csum = csum_block_add(skb->csum, csum, odd);
306 	if (!(msg->type & ICMPV6_INFOMSG_MASK))
307 		nf_ct_attach(skb, org_skb);
308 	return 0;
309 }
310 
311 #if IS_ENABLED(CONFIG_IPV6_MIP6)
312 static void mip6_addr_swap(struct sk_buff *skb)
313 {
314 	struct ipv6hdr *iph = ipv6_hdr(skb);
315 	struct inet6_skb_parm *opt = IP6CB(skb);
316 	struct ipv6_destopt_hao *hao;
317 	struct in6_addr tmp;
318 	int off;
319 
320 	if (opt->dsthao) {
321 		off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
322 		if (likely(off >= 0)) {
323 			hao = (struct ipv6_destopt_hao *)
324 					(skb_network_header(skb) + off);
325 			tmp = iph->saddr;
326 			iph->saddr = hao->addr;
327 			hao->addr = tmp;
328 		}
329 	}
330 }
331 #else
332 static inline void mip6_addr_swap(struct sk_buff *skb) {}
333 #endif
334 
335 static struct dst_entry *icmpv6_route_lookup(struct net *net,
336 					     struct sk_buff *skb,
337 					     struct sock *sk,
338 					     struct flowi6 *fl6)
339 {
340 	struct dst_entry *dst, *dst2;
341 	struct flowi6 fl2;
342 	int err;
343 
344 	err = ip6_dst_lookup(net, sk, &dst, fl6);
345 	if (err)
346 		return ERR_PTR(err);
347 
348 	/*
349 	 * We won't send icmp if the destination is known
350 	 * anycast.
351 	 */
352 	if (ipv6_anycast_destination(dst, &fl6->daddr)) {
353 		net_dbg_ratelimited("icmp6_send: acast source\n");
354 		dst_release(dst);
355 		return ERR_PTR(-EINVAL);
356 	}
357 
358 	/* No need to clone since we're just using its address. */
359 	dst2 = dst;
360 
361 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
362 	if (!IS_ERR(dst)) {
363 		if (dst != dst2)
364 			return dst;
365 	} else {
366 		if (PTR_ERR(dst) == -EPERM)
367 			dst = NULL;
368 		else
369 			return dst;
370 	}
371 
372 	err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
373 	if (err)
374 		goto relookup_failed;
375 
376 	err = ip6_dst_lookup(net, sk, &dst2, &fl2);
377 	if (err)
378 		goto relookup_failed;
379 
380 	dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
381 	if (!IS_ERR(dst2)) {
382 		dst_release(dst);
383 		dst = dst2;
384 	} else {
385 		err = PTR_ERR(dst2);
386 		if (err == -EPERM) {
387 			dst_release(dst);
388 			return dst2;
389 		} else
390 			goto relookup_failed;
391 	}
392 
393 relookup_failed:
394 	if (dst)
395 		return dst;
396 	return ERR_PTR(err);
397 }
398 
399 static int icmp6_iif(const struct sk_buff *skb)
400 {
401 	int iif = skb->dev->ifindex;
402 
403 	/* for local traffic to local address, skb dev is the loopback
404 	 * device. Check if there is a dst attached to the skb and if so
405 	 * get the real device index. Same is needed for replies to a link
406 	 * local address on a device enslaved to an L3 master device
407 	 */
408 	if (unlikely(iif == LOOPBACK_IFINDEX || netif_is_l3_master(skb->dev))) {
409 		const struct rt6_info *rt6 = skb_rt6_info(skb);
410 
411 		if (rt6)
412 			iif = rt6->rt6i_idev->dev->ifindex;
413 	}
414 
415 	return iif;
416 }
417 
418 /*
419  *	Send an ICMP message in response to a packet in error
420  */
421 static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info,
422 		       const struct in6_addr *force_saddr)
423 {
424 	struct net *net = dev_net(skb->dev);
425 	struct inet6_dev *idev = NULL;
426 	struct ipv6hdr *hdr = ipv6_hdr(skb);
427 	struct sock *sk;
428 	struct ipv6_pinfo *np;
429 	const struct in6_addr *saddr = NULL;
430 	struct dst_entry *dst;
431 	struct icmp6hdr tmp_hdr;
432 	struct flowi6 fl6;
433 	struct icmpv6_msg msg;
434 	struct sockcm_cookie sockc_unused = {0};
435 	struct ipcm6_cookie ipc6;
436 	int iif = 0;
437 	int addr_type = 0;
438 	int len;
439 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
440 
441 	if ((u8 *)hdr < skb->head ||
442 	    (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb))
443 		return;
444 
445 	/*
446 	 *	Make sure we respect the rules
447 	 *	i.e. RFC 1885 2.4(e)
448 	 *	Rule (e.1) is enforced by not using icmp6_send
449 	 *	in any code that processes icmp errors.
450 	 */
451 	addr_type = ipv6_addr_type(&hdr->daddr);
452 
453 	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) ||
454 	    ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr))
455 		saddr = &hdr->daddr;
456 
457 	/*
458 	 *	Dest addr check
459 	 */
460 
461 	if (addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST) {
462 		if (type != ICMPV6_PKT_TOOBIG &&
463 		    !(type == ICMPV6_PARAMPROB &&
464 		      code == ICMPV6_UNK_OPTION &&
465 		      (opt_unrec(skb, info))))
466 			return;
467 
468 		saddr = NULL;
469 	}
470 
471 	addr_type = ipv6_addr_type(&hdr->saddr);
472 
473 	/*
474 	 *	Source addr check
475 	 */
476 
477 	if (__ipv6_addr_needs_scope_id(addr_type)) {
478 		iif = icmp6_iif(skb);
479 	} else {
480 		dst = skb_dst(skb);
481 		iif = l3mdev_master_ifindex(dst ? dst->dev : skb->dev);
482 	}
483 
484 	/*
485 	 *	Must not send error if the source does not uniquely
486 	 *	identify a single node (RFC2463 Section 2.4).
487 	 *	We check unspecified / multicast addresses here,
488 	 *	and anycast addresses will be checked later.
489 	 */
490 	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
491 		net_dbg_ratelimited("icmp6_send: addr_any/mcast source [%pI6c > %pI6c]\n",
492 				    &hdr->saddr, &hdr->daddr);
493 		return;
494 	}
495 
496 	/*
497 	 *	Never answer to a ICMP packet.
498 	 */
499 	if (is_ineligible(skb)) {
500 		net_dbg_ratelimited("icmp6_send: no reply to icmp error [%pI6c > %pI6c]\n",
501 				    &hdr->saddr, &hdr->daddr);
502 		return;
503 	}
504 
505 	/* Needed by both icmp_global_allow and icmpv6_xmit_lock */
506 	local_bh_disable();
507 
508 	/* Check global sysctl_icmp_msgs_per_sec ratelimit */
509 	if (!(skb->dev->flags&IFF_LOOPBACK) && !icmpv6_global_allow(type))
510 		goto out_bh_enable;
511 
512 	mip6_addr_swap(skb);
513 
514 	memset(&fl6, 0, sizeof(fl6));
515 	fl6.flowi6_proto = IPPROTO_ICMPV6;
516 	fl6.daddr = hdr->saddr;
517 	if (force_saddr)
518 		saddr = force_saddr;
519 	if (saddr)
520 		fl6.saddr = *saddr;
521 	fl6.flowi6_mark = mark;
522 	fl6.flowi6_oif = iif;
523 	fl6.fl6_icmp_type = type;
524 	fl6.fl6_icmp_code = code;
525 	fl6.flowi6_uid = sock_net_uid(net, NULL);
526 	fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, NULL);
527 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
528 
529 	sk = icmpv6_xmit_lock(net);
530 	if (!sk)
531 		goto out_bh_enable;
532 
533 	sk->sk_mark = mark;
534 	np = inet6_sk(sk);
535 
536 	if (!icmpv6_xrlim_allow(sk, type, &fl6))
537 		goto out;
538 
539 	tmp_hdr.icmp6_type = type;
540 	tmp_hdr.icmp6_code = code;
541 	tmp_hdr.icmp6_cksum = 0;
542 	tmp_hdr.icmp6_pointer = htonl(info);
543 
544 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
545 		fl6.flowi6_oif = np->mcast_oif;
546 	else if (!fl6.flowi6_oif)
547 		fl6.flowi6_oif = np->ucast_oif;
548 
549 	ipc6.tclass = np->tclass;
550 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
551 
552 	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
553 	if (IS_ERR(dst))
554 		goto out;
555 
556 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
557 	ipc6.dontfrag = np->dontfrag;
558 	ipc6.opt = NULL;
559 
560 	msg.skb = skb;
561 	msg.offset = skb_network_offset(skb);
562 	msg.type = type;
563 
564 	len = skb->len - msg.offset;
565 	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(struct icmp6hdr));
566 	if (len < 0) {
567 		net_dbg_ratelimited("icmp: len problem [%pI6c > %pI6c]\n",
568 				    &hdr->saddr, &hdr->daddr);
569 		goto out_dst_release;
570 	}
571 
572 	rcu_read_lock();
573 	idev = __in6_dev_get(skb->dev);
574 
575 	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
576 			    len + sizeof(struct icmp6hdr),
577 			    sizeof(struct icmp6hdr),
578 			    &ipc6, &fl6, (struct rt6_info *)dst,
579 			    MSG_DONTWAIT, &sockc_unused)) {
580 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
581 		ip6_flush_pending_frames(sk);
582 	} else {
583 		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
584 					   len + sizeof(struct icmp6hdr));
585 	}
586 	rcu_read_unlock();
587 out_dst_release:
588 	dst_release(dst);
589 out:
590 	icmpv6_xmit_unlock(sk);
591 out_bh_enable:
592 	local_bh_enable();
593 }
594 
595 /* Slightly more convenient version of icmp6_send.
596  */
597 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
598 {
599 	icmp6_send(skb, ICMPV6_PARAMPROB, code, pos, NULL);
600 	kfree_skb(skb);
601 }
602 
603 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
604  * if sufficient data bytes are available
605  * @nhs is the size of the tunnel header(s) :
606  *  Either an IPv4 header for SIT encap
607  *         an IPv4 header + GRE header for GRE encap
608  */
609 int ip6_err_gen_icmpv6_unreach(struct sk_buff *skb, int nhs, int type,
610 			       unsigned int data_len)
611 {
612 	struct in6_addr temp_saddr;
613 	struct rt6_info *rt;
614 	struct sk_buff *skb2;
615 	u32 info = 0;
616 
617 	if (!pskb_may_pull(skb, nhs + sizeof(struct ipv6hdr) + 8))
618 		return 1;
619 
620 	/* RFC 4884 (partial) support for ICMP extensions */
621 	if (data_len < 128 || (data_len & 7) || skb->len < data_len)
622 		data_len = 0;
623 
624 	skb2 = data_len ? skb_copy(skb, GFP_ATOMIC) : skb_clone(skb, GFP_ATOMIC);
625 
626 	if (!skb2)
627 		return 1;
628 
629 	skb_dst_drop(skb2);
630 	skb_pull(skb2, nhs);
631 	skb_reset_network_header(skb2);
632 
633 	rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0,
634 			skb, 0);
635 
636 	if (rt && rt->dst.dev)
637 		skb2->dev = rt->dst.dev;
638 
639 	ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, &temp_saddr);
640 
641 	if (data_len) {
642 		/* RFC 4884 (partial) support :
643 		 * insert 0 padding at the end, before the extensions
644 		 */
645 		__skb_push(skb2, nhs);
646 		skb_reset_network_header(skb2);
647 		memmove(skb2->data, skb2->data + nhs, data_len - nhs);
648 		memset(skb2->data + data_len - nhs, 0, nhs);
649 		/* RFC 4884 4.5 : Length is measured in 64-bit words,
650 		 * and stored in reserved[0]
651 		 */
652 		info = (data_len/8) << 24;
653 	}
654 	if (type == ICMP_TIME_EXCEEDED)
655 		icmp6_send(skb2, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
656 			   info, &temp_saddr);
657 	else
658 		icmp6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH,
659 			   info, &temp_saddr);
660 	if (rt)
661 		ip6_rt_put(rt);
662 
663 	kfree_skb(skb2);
664 
665 	return 0;
666 }
667 EXPORT_SYMBOL(ip6_err_gen_icmpv6_unreach);
668 
669 static void icmpv6_echo_reply(struct sk_buff *skb)
670 {
671 	struct net *net = dev_net(skb->dev);
672 	struct sock *sk;
673 	struct inet6_dev *idev;
674 	struct ipv6_pinfo *np;
675 	const struct in6_addr *saddr = NULL;
676 	struct icmp6hdr *icmph = icmp6_hdr(skb);
677 	struct icmp6hdr tmp_hdr;
678 	struct flowi6 fl6;
679 	struct icmpv6_msg msg;
680 	struct dst_entry *dst;
681 	struct ipcm6_cookie ipc6;
682 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
683 	struct sockcm_cookie sockc_unused = {0};
684 
685 	saddr = &ipv6_hdr(skb)->daddr;
686 
687 	if (!ipv6_unicast_destination(skb) &&
688 	    !(net->ipv6.sysctl.anycast_src_echo_reply &&
689 	      ipv6_anycast_destination(skb_dst(skb), saddr)))
690 		saddr = NULL;
691 
692 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
693 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
694 
695 	memset(&fl6, 0, sizeof(fl6));
696 	fl6.flowi6_proto = IPPROTO_ICMPV6;
697 	fl6.daddr = ipv6_hdr(skb)->saddr;
698 	if (saddr)
699 		fl6.saddr = *saddr;
700 	fl6.flowi6_oif = icmp6_iif(skb);
701 	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
702 	fl6.flowi6_mark = mark;
703 	fl6.flowi6_uid = sock_net_uid(net, NULL);
704 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
705 
706 	local_bh_disable();
707 	sk = icmpv6_xmit_lock(net);
708 	if (!sk)
709 		goto out_bh_enable;
710 	sk->sk_mark = mark;
711 	np = inet6_sk(sk);
712 
713 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
714 		fl6.flowi6_oif = np->mcast_oif;
715 	else if (!fl6.flowi6_oif)
716 		fl6.flowi6_oif = np->ucast_oif;
717 
718 	if (ip6_dst_lookup(net, sk, &dst, &fl6))
719 		goto out;
720 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
721 	if (IS_ERR(dst))
722 		goto out;
723 
724 	idev = __in6_dev_get(skb->dev);
725 
726 	msg.skb = skb;
727 	msg.offset = 0;
728 	msg.type = ICMPV6_ECHO_REPLY;
729 
730 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
731 	ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb));
732 	ipc6.dontfrag = np->dontfrag;
733 	ipc6.opt = NULL;
734 
735 	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
736 			    skb->len + sizeof(struct icmp6hdr),
737 			    sizeof(struct icmp6hdr), &ipc6, &fl6,
738 			    (struct rt6_info *)dst, MSG_DONTWAIT,
739 			    &sockc_unused)) {
740 		__ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
741 		ip6_flush_pending_frames(sk);
742 	} else {
743 		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
744 					   skb->len + sizeof(struct icmp6hdr));
745 	}
746 	dst_release(dst);
747 out:
748 	icmpv6_xmit_unlock(sk);
749 out_bh_enable:
750 	local_bh_enable();
751 }
752 
753 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
754 {
755 	const struct inet6_protocol *ipprot;
756 	int inner_offset;
757 	__be16 frag_off;
758 	u8 nexthdr;
759 	struct net *net = dev_net(skb->dev);
760 
761 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
762 		goto out;
763 
764 	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
765 	if (ipv6_ext_hdr(nexthdr)) {
766 		/* now skip over extension headers */
767 		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
768 						&nexthdr, &frag_off);
769 		if (inner_offset < 0)
770 			goto out;
771 	} else {
772 		inner_offset = sizeof(struct ipv6hdr);
773 	}
774 
775 	/* Checkin header including 8 bytes of inner protocol header. */
776 	if (!pskb_may_pull(skb, inner_offset+8))
777 		goto out;
778 
779 	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
780 	   Without this we will not able f.e. to make source routed
781 	   pmtu discovery.
782 	   Corresponding argument (opt) to notifiers is already added.
783 	   --ANK (980726)
784 	 */
785 
786 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
787 	if (ipprot && ipprot->err_handler)
788 		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
789 
790 	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
791 	return;
792 
793 out:
794 	__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
795 }
796 
797 /*
798  *	Handle icmp messages
799  */
800 
801 static int icmpv6_rcv(struct sk_buff *skb)
802 {
803 	struct net_device *dev = skb->dev;
804 	struct inet6_dev *idev = __in6_dev_get(dev);
805 	const struct in6_addr *saddr, *daddr;
806 	struct icmp6hdr *hdr;
807 	u8 type;
808 	bool success = false;
809 
810 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
811 		struct sec_path *sp = skb_sec_path(skb);
812 		int nh;
813 
814 		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
815 				 XFRM_STATE_ICMP))
816 			goto drop_no_count;
817 
818 		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
819 			goto drop_no_count;
820 
821 		nh = skb_network_offset(skb);
822 		skb_set_network_header(skb, sizeof(*hdr));
823 
824 		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
825 			goto drop_no_count;
826 
827 		skb_set_network_header(skb, nh);
828 	}
829 
830 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS);
831 
832 	saddr = &ipv6_hdr(skb)->saddr;
833 	daddr = &ipv6_hdr(skb)->daddr;
834 
835 	if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) {
836 		net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n",
837 				    saddr, daddr);
838 		goto csum_error;
839 	}
840 
841 	if (!pskb_pull(skb, sizeof(*hdr)))
842 		goto discard_it;
843 
844 	hdr = icmp6_hdr(skb);
845 
846 	type = hdr->icmp6_type;
847 
848 	ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type);
849 
850 	switch (type) {
851 	case ICMPV6_ECHO_REQUEST:
852 		icmpv6_echo_reply(skb);
853 		break;
854 
855 	case ICMPV6_ECHO_REPLY:
856 		success = ping_rcv(skb);
857 		break;
858 
859 	case ICMPV6_PKT_TOOBIG:
860 		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
861 		   standard destination cache. Seems, only "advanced"
862 		   destination cache will allow to solve this problem
863 		   --ANK (980726)
864 		 */
865 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
866 			goto discard_it;
867 		hdr = icmp6_hdr(skb);
868 
869 		/* to notify */
870 		/* fall through */
871 	case ICMPV6_DEST_UNREACH:
872 	case ICMPV6_TIME_EXCEED:
873 	case ICMPV6_PARAMPROB:
874 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
875 		break;
876 
877 	case NDISC_ROUTER_SOLICITATION:
878 	case NDISC_ROUTER_ADVERTISEMENT:
879 	case NDISC_NEIGHBOUR_SOLICITATION:
880 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
881 	case NDISC_REDIRECT:
882 		ndisc_rcv(skb);
883 		break;
884 
885 	case ICMPV6_MGM_QUERY:
886 		igmp6_event_query(skb);
887 		break;
888 
889 	case ICMPV6_MGM_REPORT:
890 		igmp6_event_report(skb);
891 		break;
892 
893 	case ICMPV6_MGM_REDUCTION:
894 	case ICMPV6_NI_QUERY:
895 	case ICMPV6_NI_REPLY:
896 	case ICMPV6_MLD2_REPORT:
897 	case ICMPV6_DHAAD_REQUEST:
898 	case ICMPV6_DHAAD_REPLY:
899 	case ICMPV6_MOBILE_PREFIX_SOL:
900 	case ICMPV6_MOBILE_PREFIX_ADV:
901 		break;
902 
903 	default:
904 		/* informational */
905 		if (type & ICMPV6_INFOMSG_MASK)
906 			break;
907 
908 		net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n",
909 				    saddr, daddr);
910 
911 		/*
912 		 * error of unknown type.
913 		 * must pass to upper level
914 		 */
915 
916 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
917 	}
918 
919 	/* until the v6 path can be better sorted assume failure and
920 	 * preserve the status quo behaviour for the rest of the paths to here
921 	 */
922 	if (success)
923 		consume_skb(skb);
924 	else
925 		kfree_skb(skb);
926 
927 	return 0;
928 
929 csum_error:
930 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
931 discard_it:
932 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS);
933 drop_no_count:
934 	kfree_skb(skb);
935 	return 0;
936 }
937 
938 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
939 		      u8 type,
940 		      const struct in6_addr *saddr,
941 		      const struct in6_addr *daddr,
942 		      int oif)
943 {
944 	memset(fl6, 0, sizeof(*fl6));
945 	fl6->saddr = *saddr;
946 	fl6->daddr = *daddr;
947 	fl6->flowi6_proto	= IPPROTO_ICMPV6;
948 	fl6->fl6_icmp_type	= type;
949 	fl6->fl6_icmp_code	= 0;
950 	fl6->flowi6_oif		= oif;
951 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
952 }
953 
954 static int __net_init icmpv6_sk_init(struct net *net)
955 {
956 	struct sock *sk;
957 	int err, i, j;
958 
959 	net->ipv6.icmp_sk =
960 		kcalloc(nr_cpu_ids, sizeof(struct sock *), GFP_KERNEL);
961 	if (!net->ipv6.icmp_sk)
962 		return -ENOMEM;
963 
964 	for_each_possible_cpu(i) {
965 		err = inet_ctl_sock_create(&sk, PF_INET6,
966 					   SOCK_RAW, IPPROTO_ICMPV6, net);
967 		if (err < 0) {
968 			pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
969 			       err);
970 			goto fail;
971 		}
972 
973 		net->ipv6.icmp_sk[i] = sk;
974 
975 		/* Enough space for 2 64K ICMP packets, including
976 		 * sk_buff struct overhead.
977 		 */
978 		sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
979 	}
980 	return 0;
981 
982  fail:
983 	for (j = 0; j < i; j++)
984 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
985 	kfree(net->ipv6.icmp_sk);
986 	return err;
987 }
988 
989 static void __net_exit icmpv6_sk_exit(struct net *net)
990 {
991 	int i;
992 
993 	for_each_possible_cpu(i) {
994 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
995 	}
996 	kfree(net->ipv6.icmp_sk);
997 }
998 
999 static struct pernet_operations icmpv6_sk_ops = {
1000 	.init = icmpv6_sk_init,
1001 	.exit = icmpv6_sk_exit,
1002 };
1003 
1004 int __init icmpv6_init(void)
1005 {
1006 	int err;
1007 
1008 	err = register_pernet_subsys(&icmpv6_sk_ops);
1009 	if (err < 0)
1010 		return err;
1011 
1012 	err = -EAGAIN;
1013 	if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
1014 		goto fail;
1015 
1016 	err = inet6_register_icmp_sender(icmp6_send);
1017 	if (err)
1018 		goto sender_reg_err;
1019 	return 0;
1020 
1021 sender_reg_err:
1022 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
1023 fail:
1024 	pr_err("Failed to register ICMP6 protocol\n");
1025 	unregister_pernet_subsys(&icmpv6_sk_ops);
1026 	return err;
1027 }
1028 
1029 void icmpv6_cleanup(void)
1030 {
1031 	inet6_unregister_icmp_sender(icmp6_send);
1032 	unregister_pernet_subsys(&icmpv6_sk_ops);
1033 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
1034 }
1035 
1036 
1037 static const struct icmp6_err {
1038 	int err;
1039 	int fatal;
1040 } tab_unreach[] = {
1041 	{	/* NOROUTE */
1042 		.err	= ENETUNREACH,
1043 		.fatal	= 0,
1044 	},
1045 	{	/* ADM_PROHIBITED */
1046 		.err	= EACCES,
1047 		.fatal	= 1,
1048 	},
1049 	{	/* Was NOT_NEIGHBOUR, now reserved */
1050 		.err	= EHOSTUNREACH,
1051 		.fatal	= 0,
1052 	},
1053 	{	/* ADDR_UNREACH	*/
1054 		.err	= EHOSTUNREACH,
1055 		.fatal	= 0,
1056 	},
1057 	{	/* PORT_UNREACH	*/
1058 		.err	= ECONNREFUSED,
1059 		.fatal	= 1,
1060 	},
1061 	{	/* POLICY_FAIL */
1062 		.err	= EACCES,
1063 		.fatal	= 1,
1064 	},
1065 	{	/* REJECT_ROUTE	*/
1066 		.err	= EACCES,
1067 		.fatal	= 1,
1068 	},
1069 };
1070 
1071 int icmpv6_err_convert(u8 type, u8 code, int *err)
1072 {
1073 	int fatal = 0;
1074 
1075 	*err = EPROTO;
1076 
1077 	switch (type) {
1078 	case ICMPV6_DEST_UNREACH:
1079 		fatal = 1;
1080 		if (code < ARRAY_SIZE(tab_unreach)) {
1081 			*err  = tab_unreach[code].err;
1082 			fatal = tab_unreach[code].fatal;
1083 		}
1084 		break;
1085 
1086 	case ICMPV6_PKT_TOOBIG:
1087 		*err = EMSGSIZE;
1088 		break;
1089 
1090 	case ICMPV6_PARAMPROB:
1091 		*err = EPROTO;
1092 		fatal = 1;
1093 		break;
1094 
1095 	case ICMPV6_TIME_EXCEED:
1096 		*err = EHOSTUNREACH;
1097 		break;
1098 	}
1099 
1100 	return fatal;
1101 }
1102 EXPORT_SYMBOL(icmpv6_err_convert);
1103 
1104 #ifdef CONFIG_SYSCTL
1105 static struct ctl_table ipv6_icmp_table_template[] = {
1106 	{
1107 		.procname	= "ratelimit",
1108 		.data		= &init_net.ipv6.sysctl.icmpv6_time,
1109 		.maxlen		= sizeof(int),
1110 		.mode		= 0644,
1111 		.proc_handler	= proc_dointvec_ms_jiffies,
1112 	},
1113 	{ },
1114 };
1115 
1116 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1117 {
1118 	struct ctl_table *table;
1119 
1120 	table = kmemdup(ipv6_icmp_table_template,
1121 			sizeof(ipv6_icmp_table_template),
1122 			GFP_KERNEL);
1123 
1124 	if (table)
1125 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
1126 
1127 	return table;
1128 }
1129 #endif
1130