xref: /openbmc/linux/net/ipv6/icmp.c (revision d2574c33)
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 *this_cpu_ptr(net->ipv6.icmp_sk);
85 }
86 
87 static int 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, skb->dev->ifindex, 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 	return 0;
105 }
106 
107 static int icmpv6_rcv(struct sk_buff *skb);
108 
109 static const struct inet6_protocol icmpv6_protocol = {
110 	.handler	=	icmpv6_rcv,
111 	.err_handler	=	icmpv6_err,
112 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
113 };
114 
115 /* Called with BH disabled */
116 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
117 {
118 	struct sock *sk;
119 
120 	sk = icmpv6_sk(net);
121 	if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
122 		/* This can happen if the output path (f.e. SIT or
123 		 * ip6ip6 tunnel) signals dst_link_failure() for an
124 		 * outgoing ICMP6 packet.
125 		 */
126 		return NULL;
127 	}
128 	return sk;
129 }
130 
131 static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
132 {
133 	spin_unlock(&sk->sk_lock.slock);
134 }
135 
136 /*
137  * Figure out, may we reply to this packet with icmp error.
138  *
139  * We do not reply, if:
140  *	- it was icmp error message.
141  *	- it is truncated, so that it is known, that protocol is ICMPV6
142  *	  (i.e. in the middle of some exthdr)
143  *
144  *	--ANK (980726)
145  */
146 
147 static bool is_ineligible(const struct sk_buff *skb)
148 {
149 	int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
150 	int len = skb->len - ptr;
151 	__u8 nexthdr = ipv6_hdr(skb)->nexthdr;
152 	__be16 frag_off;
153 
154 	if (len < 0)
155 		return true;
156 
157 	ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off);
158 	if (ptr < 0)
159 		return false;
160 	if (nexthdr == IPPROTO_ICMPV6) {
161 		u8 _type, *tp;
162 		tp = skb_header_pointer(skb,
163 			ptr+offsetof(struct icmp6hdr, icmp6_type),
164 			sizeof(_type), &_type);
165 		if (!tp || !(*tp & ICMPV6_INFOMSG_MASK))
166 			return true;
167 	}
168 	return false;
169 }
170 
171 static bool icmpv6_mask_allow(int type)
172 {
173 	/* Informational messages are not limited. */
174 	if (type & ICMPV6_INFOMSG_MASK)
175 		return true;
176 
177 	/* Do not limit pmtu discovery, it would break it. */
178 	if (type == ICMPV6_PKT_TOOBIG)
179 		return true;
180 
181 	return false;
182 }
183 
184 static bool icmpv6_global_allow(int type)
185 {
186 	if (icmpv6_mask_allow(type))
187 		return true;
188 
189 	if (icmp_global_allow())
190 		return true;
191 
192 	return false;
193 }
194 
195 /*
196  * Check the ICMP output rate limit
197  */
198 static bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
199 			       struct flowi6 *fl6)
200 {
201 	struct net *net = sock_net(sk);
202 	struct dst_entry *dst;
203 	bool res = false;
204 
205 	if (icmpv6_mask_allow(type))
206 		return true;
207 
208 	/*
209 	 * Look up the output route.
210 	 * XXX: perhaps the expire for routing entries cloned by
211 	 * this lookup should be more aggressive (not longer than timeout).
212 	 */
213 	dst = ip6_route_output(net, sk, fl6);
214 	if (dst->error) {
215 		IP6_INC_STATS(net, ip6_dst_idev(dst),
216 			      IPSTATS_MIB_OUTNOROUTES);
217 	} else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
218 		res = true;
219 	} else {
220 		struct rt6_info *rt = (struct rt6_info *)dst;
221 		int tmo = net->ipv6.sysctl.icmpv6_time;
222 		struct inet_peer *peer;
223 
224 		/* Give more bandwidth to wider prefixes. */
225 		if (rt->rt6i_dst.plen < 128)
226 			tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
227 
228 		peer = inet_getpeer_v6(net->ipv6.peers, &fl6->daddr, 1);
229 		res = inet_peer_xrlim_allow(peer, tmo);
230 		if (peer)
231 			inet_putpeer(peer);
232 	}
233 	dst_release(dst);
234 	return res;
235 }
236 
237 /*
238  *	an inline helper for the "simple" if statement below
239  *	checks if parameter problem report is caused by an
240  *	unrecognized IPv6 option that has the Option Type
241  *	highest-order two bits set to 10
242  */
243 
244 static bool opt_unrec(struct sk_buff *skb, __u32 offset)
245 {
246 	u8 _optval, *op;
247 
248 	offset += skb_network_offset(skb);
249 	op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
250 	if (!op)
251 		return true;
252 	return (*op & 0xC0) == 0x80;
253 }
254 
255 void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
256 				struct icmp6hdr *thdr, int len)
257 {
258 	struct sk_buff *skb;
259 	struct icmp6hdr *icmp6h;
260 
261 	skb = skb_peek(&sk->sk_write_queue);
262 	if (!skb)
263 		return;
264 
265 	icmp6h = icmp6_hdr(skb);
266 	memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
267 	icmp6h->icmp6_cksum = 0;
268 
269 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
270 		skb->csum = csum_partial(icmp6h,
271 					sizeof(struct icmp6hdr), skb->csum);
272 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
273 						      &fl6->daddr,
274 						      len, fl6->flowi6_proto,
275 						      skb->csum);
276 	} else {
277 		__wsum tmp_csum = 0;
278 
279 		skb_queue_walk(&sk->sk_write_queue, skb) {
280 			tmp_csum = csum_add(tmp_csum, skb->csum);
281 		}
282 
283 		tmp_csum = csum_partial(icmp6h,
284 					sizeof(struct icmp6hdr), tmp_csum);
285 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
286 						      &fl6->daddr,
287 						      len, fl6->flowi6_proto,
288 						      tmp_csum);
289 	}
290 	ip6_push_pending_frames(sk);
291 }
292 
293 struct icmpv6_msg {
294 	struct sk_buff	*skb;
295 	int		offset;
296 	uint8_t		type;
297 };
298 
299 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
300 {
301 	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
302 	struct sk_buff *org_skb = msg->skb;
303 	__wsum csum = 0;
304 
305 	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
306 				      to, len, csum);
307 	skb->csum = csum_block_add(skb->csum, csum, odd);
308 	if (!(msg->type & ICMPV6_INFOMSG_MASK))
309 		nf_ct_attach(skb, org_skb);
310 	return 0;
311 }
312 
313 #if IS_ENABLED(CONFIG_IPV6_MIP6)
314 static void mip6_addr_swap(struct sk_buff *skb)
315 {
316 	struct ipv6hdr *iph = ipv6_hdr(skb);
317 	struct inet6_skb_parm *opt = IP6CB(skb);
318 	struct ipv6_destopt_hao *hao;
319 	struct in6_addr tmp;
320 	int off;
321 
322 	if (opt->dsthao) {
323 		off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
324 		if (likely(off >= 0)) {
325 			hao = (struct ipv6_destopt_hao *)
326 					(skb_network_header(skb) + off);
327 			tmp = iph->saddr;
328 			iph->saddr = hao->addr;
329 			hao->addr = tmp;
330 		}
331 	}
332 }
333 #else
334 static inline void mip6_addr_swap(struct sk_buff *skb) {}
335 #endif
336 
337 static struct dst_entry *icmpv6_route_lookup(struct net *net,
338 					     struct sk_buff *skb,
339 					     struct sock *sk,
340 					     struct flowi6 *fl6)
341 {
342 	struct dst_entry *dst, *dst2;
343 	struct flowi6 fl2;
344 	int err;
345 
346 	err = ip6_dst_lookup(net, sk, &dst, fl6);
347 	if (err)
348 		return ERR_PTR(err);
349 
350 	/*
351 	 * We won't send icmp if the destination is known
352 	 * anycast.
353 	 */
354 	if (ipv6_anycast_destination(dst, &fl6->daddr)) {
355 		net_dbg_ratelimited("icmp6_send: acast source\n");
356 		dst_release(dst);
357 		return ERR_PTR(-EINVAL);
358 	}
359 
360 	/* No need to clone since we're just using its address. */
361 	dst2 = dst;
362 
363 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
364 	if (!IS_ERR(dst)) {
365 		if (dst != dst2)
366 			return dst;
367 	} else {
368 		if (PTR_ERR(dst) == -EPERM)
369 			dst = NULL;
370 		else
371 			return dst;
372 	}
373 
374 	err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
375 	if (err)
376 		goto relookup_failed;
377 
378 	err = ip6_dst_lookup(net, sk, &dst2, &fl2);
379 	if (err)
380 		goto relookup_failed;
381 
382 	dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
383 	if (!IS_ERR(dst2)) {
384 		dst_release(dst);
385 		dst = dst2;
386 	} else {
387 		err = PTR_ERR(dst2);
388 		if (err == -EPERM) {
389 			dst_release(dst);
390 			return dst2;
391 		} else
392 			goto relookup_failed;
393 	}
394 
395 relookup_failed:
396 	if (dst)
397 		return dst;
398 	return ERR_PTR(err);
399 }
400 
401 static int icmp6_iif(const struct sk_buff *skb)
402 {
403 	int iif = skb->dev->ifindex;
404 
405 	/* for local traffic to local address, skb dev is the loopback
406 	 * device. Check if there is a dst attached to the skb and if so
407 	 * get the real device index. Same is needed for replies to a link
408 	 * local address on a device enslaved to an L3 master device
409 	 */
410 	if (unlikely(iif == LOOPBACK_IFINDEX || netif_is_l3_master(skb->dev))) {
411 		const struct rt6_info *rt6 = skb_rt6_info(skb);
412 
413 		if (rt6)
414 			iif = rt6->rt6i_idev->dev->ifindex;
415 	}
416 
417 	return iif;
418 }
419 
420 /*
421  *	Send an ICMP message in response to a packet in error
422  */
423 static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info,
424 		       const struct in6_addr *force_saddr)
425 {
426 	struct inet6_dev *idev = NULL;
427 	struct ipv6hdr *hdr = ipv6_hdr(skb);
428 	struct sock *sk;
429 	struct net *net;
430 	struct ipv6_pinfo *np;
431 	const struct in6_addr *saddr = NULL;
432 	struct dst_entry *dst;
433 	struct icmp6hdr tmp_hdr;
434 	struct flowi6 fl6;
435 	struct icmpv6_msg msg;
436 	struct ipcm6_cookie ipc6;
437 	int iif = 0;
438 	int addr_type = 0;
439 	int len;
440 	u32 mark;
441 
442 	if ((u8 *)hdr < skb->head ||
443 	    (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb))
444 		return;
445 
446 	if (!skb->dev)
447 		return;
448 	net = dev_net(skb->dev);
449 	mark = IP6_REPLY_MARK(net, skb->mark);
450 	/*
451 	 *	Make sure we respect the rules
452 	 *	i.e. RFC 1885 2.4(e)
453 	 *	Rule (e.1) is enforced by not using icmp6_send
454 	 *	in any code that processes icmp errors.
455 	 */
456 	addr_type = ipv6_addr_type(&hdr->daddr);
457 
458 	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) ||
459 	    ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr))
460 		saddr = &hdr->daddr;
461 
462 	/*
463 	 *	Dest addr check
464 	 */
465 
466 	if (addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST) {
467 		if (type != ICMPV6_PKT_TOOBIG &&
468 		    !(type == ICMPV6_PARAMPROB &&
469 		      code == ICMPV6_UNK_OPTION &&
470 		      (opt_unrec(skb, info))))
471 			return;
472 
473 		saddr = NULL;
474 	}
475 
476 	addr_type = ipv6_addr_type(&hdr->saddr);
477 
478 	/*
479 	 *	Source addr check
480 	 */
481 
482 	if (__ipv6_addr_needs_scope_id(addr_type)) {
483 		iif = icmp6_iif(skb);
484 	} else {
485 		dst = skb_dst(skb);
486 		iif = l3mdev_master_ifindex(dst ? dst->dev : skb->dev);
487 	}
488 
489 	/*
490 	 *	Must not send error if the source does not uniquely
491 	 *	identify a single node (RFC2463 Section 2.4).
492 	 *	We check unspecified / multicast addresses here,
493 	 *	and anycast addresses will be checked later.
494 	 */
495 	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
496 		net_dbg_ratelimited("icmp6_send: addr_any/mcast source [%pI6c > %pI6c]\n",
497 				    &hdr->saddr, &hdr->daddr);
498 		return;
499 	}
500 
501 	/*
502 	 *	Never answer to a ICMP packet.
503 	 */
504 	if (is_ineligible(skb)) {
505 		net_dbg_ratelimited("icmp6_send: no reply to icmp error [%pI6c > %pI6c]\n",
506 				    &hdr->saddr, &hdr->daddr);
507 		return;
508 	}
509 
510 	/* Needed by both icmp_global_allow and icmpv6_xmit_lock */
511 	local_bh_disable();
512 
513 	/* Check global sysctl_icmp_msgs_per_sec ratelimit */
514 	if (!(skb->dev->flags&IFF_LOOPBACK) && !icmpv6_global_allow(type))
515 		goto out_bh_enable;
516 
517 	mip6_addr_swap(skb);
518 
519 	memset(&fl6, 0, sizeof(fl6));
520 	fl6.flowi6_proto = IPPROTO_ICMPV6;
521 	fl6.daddr = hdr->saddr;
522 	if (force_saddr)
523 		saddr = force_saddr;
524 	if (saddr)
525 		fl6.saddr = *saddr;
526 	fl6.flowi6_mark = mark;
527 	fl6.flowi6_oif = iif;
528 	fl6.fl6_icmp_type = type;
529 	fl6.fl6_icmp_code = code;
530 	fl6.flowi6_uid = sock_net_uid(net, NULL);
531 	fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, NULL);
532 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
533 
534 	sk = icmpv6_xmit_lock(net);
535 	if (!sk)
536 		goto out_bh_enable;
537 
538 	sk->sk_mark = mark;
539 	np = inet6_sk(sk);
540 
541 	if (!icmpv6_xrlim_allow(sk, type, &fl6))
542 		goto out;
543 
544 	tmp_hdr.icmp6_type = type;
545 	tmp_hdr.icmp6_code = code;
546 	tmp_hdr.icmp6_cksum = 0;
547 	tmp_hdr.icmp6_pointer = htonl(info);
548 
549 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
550 		fl6.flowi6_oif = np->mcast_oif;
551 	else if (!fl6.flowi6_oif)
552 		fl6.flowi6_oif = np->ucast_oif;
553 
554 	ipcm6_init_sk(&ipc6, np);
555 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
556 
557 	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
558 	if (IS_ERR(dst))
559 		goto out;
560 
561 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
562 
563 	msg.skb = skb;
564 	msg.offset = skb_network_offset(skb);
565 	msg.type = type;
566 
567 	len = skb->len - msg.offset;
568 	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(struct icmp6hdr));
569 	if (len < 0) {
570 		net_dbg_ratelimited("icmp: len problem [%pI6c > %pI6c]\n",
571 				    &hdr->saddr, &hdr->daddr);
572 		goto out_dst_release;
573 	}
574 
575 	rcu_read_lock();
576 	idev = __in6_dev_get(skb->dev);
577 
578 	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
579 			    len + sizeof(struct icmp6hdr),
580 			    sizeof(struct icmp6hdr),
581 			    &ipc6, &fl6, (struct rt6_info *)dst,
582 			    MSG_DONTWAIT)) {
583 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
584 		ip6_flush_pending_frames(sk);
585 	} else {
586 		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
587 					   len + sizeof(struct icmp6hdr));
588 	}
589 	rcu_read_unlock();
590 out_dst_release:
591 	dst_release(dst);
592 out:
593 	icmpv6_xmit_unlock(sk);
594 out_bh_enable:
595 	local_bh_enable();
596 }
597 
598 /* Slightly more convenient version of icmp6_send.
599  */
600 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
601 {
602 	icmp6_send(skb, ICMPV6_PARAMPROB, code, pos, NULL);
603 	kfree_skb(skb);
604 }
605 
606 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
607  * if sufficient data bytes are available
608  * @nhs is the size of the tunnel header(s) :
609  *  Either an IPv4 header for SIT encap
610  *         an IPv4 header + GRE header for GRE encap
611  */
612 int ip6_err_gen_icmpv6_unreach(struct sk_buff *skb, int nhs, int type,
613 			       unsigned int data_len)
614 {
615 	struct in6_addr temp_saddr;
616 	struct rt6_info *rt;
617 	struct sk_buff *skb2;
618 	u32 info = 0;
619 
620 	if (!pskb_may_pull(skb, nhs + sizeof(struct ipv6hdr) + 8))
621 		return 1;
622 
623 	/* RFC 4884 (partial) support for ICMP extensions */
624 	if (data_len < 128 || (data_len & 7) || skb->len < data_len)
625 		data_len = 0;
626 
627 	skb2 = data_len ? skb_copy(skb, GFP_ATOMIC) : skb_clone(skb, GFP_ATOMIC);
628 
629 	if (!skb2)
630 		return 1;
631 
632 	skb_dst_drop(skb2);
633 	skb_pull(skb2, nhs);
634 	skb_reset_network_header(skb2);
635 
636 	rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0,
637 			skb, 0);
638 
639 	if (rt && rt->dst.dev)
640 		skb2->dev = rt->dst.dev;
641 
642 	ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, &temp_saddr);
643 
644 	if (data_len) {
645 		/* RFC 4884 (partial) support :
646 		 * insert 0 padding at the end, before the extensions
647 		 */
648 		__skb_push(skb2, nhs);
649 		skb_reset_network_header(skb2);
650 		memmove(skb2->data, skb2->data + nhs, data_len - nhs);
651 		memset(skb2->data + data_len - nhs, 0, nhs);
652 		/* RFC 4884 4.5 : Length is measured in 64-bit words,
653 		 * and stored in reserved[0]
654 		 */
655 		info = (data_len/8) << 24;
656 	}
657 	if (type == ICMP_TIME_EXCEEDED)
658 		icmp6_send(skb2, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
659 			   info, &temp_saddr);
660 	else
661 		icmp6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH,
662 			   info, &temp_saddr);
663 	if (rt)
664 		ip6_rt_put(rt);
665 
666 	kfree_skb(skb2);
667 
668 	return 0;
669 }
670 EXPORT_SYMBOL(ip6_err_gen_icmpv6_unreach);
671 
672 static void icmpv6_echo_reply(struct sk_buff *skb)
673 {
674 	struct net *net = dev_net(skb->dev);
675 	struct sock *sk;
676 	struct inet6_dev *idev;
677 	struct ipv6_pinfo *np;
678 	const struct in6_addr *saddr = NULL;
679 	struct icmp6hdr *icmph = icmp6_hdr(skb);
680 	struct icmp6hdr tmp_hdr;
681 	struct flowi6 fl6;
682 	struct icmpv6_msg msg;
683 	struct dst_entry *dst;
684 	struct ipcm6_cookie ipc6;
685 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
686 
687 	saddr = &ipv6_hdr(skb)->daddr;
688 
689 	if (!ipv6_unicast_destination(skb) &&
690 	    !(net->ipv6.sysctl.anycast_src_echo_reply &&
691 	      ipv6_anycast_destination(skb_dst(skb), saddr)))
692 		saddr = NULL;
693 
694 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
695 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
696 
697 	memset(&fl6, 0, sizeof(fl6));
698 	fl6.flowi6_proto = IPPROTO_ICMPV6;
699 	fl6.daddr = ipv6_hdr(skb)->saddr;
700 	if (saddr)
701 		fl6.saddr = *saddr;
702 	fl6.flowi6_oif = icmp6_iif(skb);
703 	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
704 	fl6.flowi6_mark = mark;
705 	fl6.flowi6_uid = sock_net_uid(net, NULL);
706 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
707 
708 	local_bh_disable();
709 	sk = icmpv6_xmit_lock(net);
710 	if (!sk)
711 		goto out_bh_enable;
712 	sk->sk_mark = mark;
713 	np = inet6_sk(sk);
714 
715 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
716 		fl6.flowi6_oif = np->mcast_oif;
717 	else if (!fl6.flowi6_oif)
718 		fl6.flowi6_oif = np->ucast_oif;
719 
720 	if (ip6_dst_lookup(net, sk, &dst, &fl6))
721 		goto out;
722 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
723 	if (IS_ERR(dst))
724 		goto out;
725 
726 	idev = __in6_dev_get(skb->dev);
727 
728 	msg.skb = skb;
729 	msg.offset = 0;
730 	msg.type = ICMPV6_ECHO_REPLY;
731 
732 	ipcm6_init_sk(&ipc6, np);
733 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
734 	ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb));
735 
736 	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
737 			    skb->len + sizeof(struct icmp6hdr),
738 			    sizeof(struct icmp6hdr), &ipc6, &fl6,
739 			    (struct rt6_info *)dst, MSG_DONTWAIT)) {
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 *net = dev_net(skb->dev);
804 	struct net_device *dev = skb->dev;
805 	struct inet6_dev *idev = __in6_dev_get(dev);
806 	const struct in6_addr *saddr, *daddr;
807 	struct icmp6hdr *hdr;
808 	u8 type;
809 	bool success = false;
810 
811 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
812 		struct sec_path *sp = skb_sec_path(skb);
813 		int nh;
814 
815 		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
816 				 XFRM_STATE_ICMP))
817 			goto drop_no_count;
818 
819 		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
820 			goto drop_no_count;
821 
822 		nh = skb_network_offset(skb);
823 		skb_set_network_header(skb, sizeof(*hdr));
824 
825 		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
826 			goto drop_no_count;
827 
828 		skb_set_network_header(skb, nh);
829 	}
830 
831 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS);
832 
833 	saddr = &ipv6_hdr(skb)->saddr;
834 	daddr = &ipv6_hdr(skb)->daddr;
835 
836 	if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) {
837 		net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n",
838 				    saddr, daddr);
839 		goto csum_error;
840 	}
841 
842 	if (!pskb_pull(skb, sizeof(*hdr)))
843 		goto discard_it;
844 
845 	hdr = icmp6_hdr(skb);
846 
847 	type = hdr->icmp6_type;
848 
849 	ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type);
850 
851 	switch (type) {
852 	case ICMPV6_ECHO_REQUEST:
853 		if (!net->ipv6.sysctl.icmpv6_echo_ignore_all)
854 			icmpv6_echo_reply(skb);
855 		break;
856 
857 	case ICMPV6_ECHO_REPLY:
858 		success = ping_rcv(skb);
859 		break;
860 
861 	case ICMPV6_PKT_TOOBIG:
862 		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
863 		   standard destination cache. Seems, only "advanced"
864 		   destination cache will allow to solve this problem
865 		   --ANK (980726)
866 		 */
867 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
868 			goto discard_it;
869 		hdr = icmp6_hdr(skb);
870 
871 		/* to notify */
872 		/* fall through */
873 	case ICMPV6_DEST_UNREACH:
874 	case ICMPV6_TIME_EXCEED:
875 	case ICMPV6_PARAMPROB:
876 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
877 		break;
878 
879 	case NDISC_ROUTER_SOLICITATION:
880 	case NDISC_ROUTER_ADVERTISEMENT:
881 	case NDISC_NEIGHBOUR_SOLICITATION:
882 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
883 	case NDISC_REDIRECT:
884 		ndisc_rcv(skb);
885 		break;
886 
887 	case ICMPV6_MGM_QUERY:
888 		igmp6_event_query(skb);
889 		break;
890 
891 	case ICMPV6_MGM_REPORT:
892 		igmp6_event_report(skb);
893 		break;
894 
895 	case ICMPV6_MGM_REDUCTION:
896 	case ICMPV6_NI_QUERY:
897 	case ICMPV6_NI_REPLY:
898 	case ICMPV6_MLD2_REPORT:
899 	case ICMPV6_DHAAD_REQUEST:
900 	case ICMPV6_DHAAD_REPLY:
901 	case ICMPV6_MOBILE_PREFIX_SOL:
902 	case ICMPV6_MOBILE_PREFIX_ADV:
903 		break;
904 
905 	default:
906 		/* informational */
907 		if (type & ICMPV6_INFOMSG_MASK)
908 			break;
909 
910 		net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n",
911 				    saddr, daddr);
912 
913 		/*
914 		 * error of unknown type.
915 		 * must pass to upper level
916 		 */
917 
918 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
919 	}
920 
921 	/* until the v6 path can be better sorted assume failure and
922 	 * preserve the status quo behaviour for the rest of the paths to here
923 	 */
924 	if (success)
925 		consume_skb(skb);
926 	else
927 		kfree_skb(skb);
928 
929 	return 0;
930 
931 csum_error:
932 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
933 discard_it:
934 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS);
935 drop_no_count:
936 	kfree_skb(skb);
937 	return 0;
938 }
939 
940 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
941 		      u8 type,
942 		      const struct in6_addr *saddr,
943 		      const struct in6_addr *daddr,
944 		      int oif)
945 {
946 	memset(fl6, 0, sizeof(*fl6));
947 	fl6->saddr = *saddr;
948 	fl6->daddr = *daddr;
949 	fl6->flowi6_proto	= IPPROTO_ICMPV6;
950 	fl6->fl6_icmp_type	= type;
951 	fl6->fl6_icmp_code	= 0;
952 	fl6->flowi6_oif		= oif;
953 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
954 }
955 
956 static void __net_exit icmpv6_sk_exit(struct net *net)
957 {
958 	int i;
959 
960 	for_each_possible_cpu(i)
961 		inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv6.icmp_sk, i));
962 	free_percpu(net->ipv6.icmp_sk);
963 }
964 
965 static int __net_init icmpv6_sk_init(struct net *net)
966 {
967 	struct sock *sk;
968 	int err, i;
969 
970 	net->ipv6.icmp_sk = alloc_percpu(struct sock *);
971 	if (!net->ipv6.icmp_sk)
972 		return -ENOMEM;
973 
974 	for_each_possible_cpu(i) {
975 		err = inet_ctl_sock_create(&sk, PF_INET6,
976 					   SOCK_RAW, IPPROTO_ICMPV6, net);
977 		if (err < 0) {
978 			pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
979 			       err);
980 			goto fail;
981 		}
982 
983 		*per_cpu_ptr(net->ipv6.icmp_sk, i) = sk;
984 
985 		/* Enough space for 2 64K ICMP packets, including
986 		 * sk_buff struct overhead.
987 		 */
988 		sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
989 	}
990 	return 0;
991 
992  fail:
993 	icmpv6_sk_exit(net);
994 	return err;
995 }
996 
997 static struct pernet_operations icmpv6_sk_ops = {
998 	.init = icmpv6_sk_init,
999 	.exit = icmpv6_sk_exit,
1000 };
1001 
1002 int __init icmpv6_init(void)
1003 {
1004 	int err;
1005 
1006 	err = register_pernet_subsys(&icmpv6_sk_ops);
1007 	if (err < 0)
1008 		return err;
1009 
1010 	err = -EAGAIN;
1011 	if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
1012 		goto fail;
1013 
1014 	err = inet6_register_icmp_sender(icmp6_send);
1015 	if (err)
1016 		goto sender_reg_err;
1017 	return 0;
1018 
1019 sender_reg_err:
1020 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
1021 fail:
1022 	pr_err("Failed to register ICMP6 protocol\n");
1023 	unregister_pernet_subsys(&icmpv6_sk_ops);
1024 	return err;
1025 }
1026 
1027 void icmpv6_cleanup(void)
1028 {
1029 	inet6_unregister_icmp_sender(icmp6_send);
1030 	unregister_pernet_subsys(&icmpv6_sk_ops);
1031 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
1032 }
1033 
1034 
1035 static const struct icmp6_err {
1036 	int err;
1037 	int fatal;
1038 } tab_unreach[] = {
1039 	{	/* NOROUTE */
1040 		.err	= ENETUNREACH,
1041 		.fatal	= 0,
1042 	},
1043 	{	/* ADM_PROHIBITED */
1044 		.err	= EACCES,
1045 		.fatal	= 1,
1046 	},
1047 	{	/* Was NOT_NEIGHBOUR, now reserved */
1048 		.err	= EHOSTUNREACH,
1049 		.fatal	= 0,
1050 	},
1051 	{	/* ADDR_UNREACH	*/
1052 		.err	= EHOSTUNREACH,
1053 		.fatal	= 0,
1054 	},
1055 	{	/* PORT_UNREACH	*/
1056 		.err	= ECONNREFUSED,
1057 		.fatal	= 1,
1058 	},
1059 	{	/* POLICY_FAIL */
1060 		.err	= EACCES,
1061 		.fatal	= 1,
1062 	},
1063 	{	/* REJECT_ROUTE	*/
1064 		.err	= EACCES,
1065 		.fatal	= 1,
1066 	},
1067 };
1068 
1069 int icmpv6_err_convert(u8 type, u8 code, int *err)
1070 {
1071 	int fatal = 0;
1072 
1073 	*err = EPROTO;
1074 
1075 	switch (type) {
1076 	case ICMPV6_DEST_UNREACH:
1077 		fatal = 1;
1078 		if (code < ARRAY_SIZE(tab_unreach)) {
1079 			*err  = tab_unreach[code].err;
1080 			fatal = tab_unreach[code].fatal;
1081 		}
1082 		break;
1083 
1084 	case ICMPV6_PKT_TOOBIG:
1085 		*err = EMSGSIZE;
1086 		break;
1087 
1088 	case ICMPV6_PARAMPROB:
1089 		*err = EPROTO;
1090 		fatal = 1;
1091 		break;
1092 
1093 	case ICMPV6_TIME_EXCEED:
1094 		*err = EHOSTUNREACH;
1095 		break;
1096 	}
1097 
1098 	return fatal;
1099 }
1100 EXPORT_SYMBOL(icmpv6_err_convert);
1101 
1102 #ifdef CONFIG_SYSCTL
1103 static struct ctl_table ipv6_icmp_table_template[] = {
1104 	{
1105 		.procname	= "ratelimit",
1106 		.data		= &init_net.ipv6.sysctl.icmpv6_time,
1107 		.maxlen		= sizeof(int),
1108 		.mode		= 0644,
1109 		.proc_handler	= proc_dointvec_ms_jiffies,
1110 	},
1111 	{
1112 		.procname	= "echo_ignore_all",
1113 		.data		= &init_net.ipv6.sysctl.icmpv6_echo_ignore_all,
1114 		.maxlen		= sizeof(int),
1115 		.mode		= 0644,
1116 		.proc_handler = proc_dointvec,
1117 	},
1118 	{ },
1119 };
1120 
1121 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1122 {
1123 	struct ctl_table *table;
1124 
1125 	table = kmemdup(ipv6_icmp_table_template,
1126 			sizeof(ipv6_icmp_table_template),
1127 			GFP_KERNEL);
1128 
1129 	if (table) {
1130 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
1131 		table[1].data = &net->ipv6.sysctl.icmpv6_echo_ignore_all;
1132 	}
1133 	return table;
1134 }
1135 #endif
1136