xref: /openbmc/linux/net/ipv6/icmp.c (revision feac8c8b)
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.
406 	 */
407 	if (unlikely(iif == LOOPBACK_IFINDEX)) {
408 		const struct rt6_info *rt6 = skb_rt6_info(skb);
409 
410 		if (rt6)
411 			iif = rt6->rt6i_idev->dev->ifindex;
412 	}
413 
414 	return iif;
415 }
416 
417 /*
418  *	Send an ICMP message in response to a packet in error
419  */
420 static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info,
421 		       const struct in6_addr *force_saddr)
422 {
423 	struct net *net = dev_net(skb->dev);
424 	struct inet6_dev *idev = NULL;
425 	struct ipv6hdr *hdr = ipv6_hdr(skb);
426 	struct sock *sk;
427 	struct ipv6_pinfo *np;
428 	const struct in6_addr *saddr = NULL;
429 	struct dst_entry *dst;
430 	struct icmp6hdr tmp_hdr;
431 	struct flowi6 fl6;
432 	struct icmpv6_msg msg;
433 	struct sockcm_cookie sockc_unused = {0};
434 	struct ipcm6_cookie ipc6;
435 	int iif = 0;
436 	int addr_type = 0;
437 	int len;
438 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
439 
440 	if ((u8 *)hdr < skb->head ||
441 	    (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb))
442 		return;
443 
444 	/*
445 	 *	Make sure we respect the rules
446 	 *	i.e. RFC 1885 2.4(e)
447 	 *	Rule (e.1) is enforced by not using icmp6_send
448 	 *	in any code that processes icmp errors.
449 	 */
450 	addr_type = ipv6_addr_type(&hdr->daddr);
451 
452 	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) ||
453 	    ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr))
454 		saddr = &hdr->daddr;
455 
456 	/*
457 	 *	Dest addr check
458 	 */
459 
460 	if (addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST) {
461 		if (type != ICMPV6_PKT_TOOBIG &&
462 		    !(type == ICMPV6_PARAMPROB &&
463 		      code == ICMPV6_UNK_OPTION &&
464 		      (opt_unrec(skb, info))))
465 			return;
466 
467 		saddr = NULL;
468 	}
469 
470 	addr_type = ipv6_addr_type(&hdr->saddr);
471 
472 	/*
473 	 *	Source addr check
474 	 */
475 
476 	if (__ipv6_addr_needs_scope_id(addr_type)) {
477 		iif = icmp6_iif(skb);
478 	} else {
479 		dst = skb_dst(skb);
480 		iif = l3mdev_master_ifindex(dst ? dst->dev : skb->dev);
481 	}
482 
483 	/*
484 	 *	Must not send error if the source does not uniquely
485 	 *	identify a single node (RFC2463 Section 2.4).
486 	 *	We check unspecified / multicast addresses here,
487 	 *	and anycast addresses will be checked later.
488 	 */
489 	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
490 		net_dbg_ratelimited("icmp6_send: addr_any/mcast source [%pI6c > %pI6c]\n",
491 				    &hdr->saddr, &hdr->daddr);
492 		return;
493 	}
494 
495 	/*
496 	 *	Never answer to a ICMP packet.
497 	 */
498 	if (is_ineligible(skb)) {
499 		net_dbg_ratelimited("icmp6_send: no reply to icmp error [%pI6c > %pI6c]\n",
500 				    &hdr->saddr, &hdr->daddr);
501 		return;
502 	}
503 
504 	/* Needed by both icmp_global_allow and icmpv6_xmit_lock */
505 	local_bh_disable();
506 
507 	/* Check global sysctl_icmp_msgs_per_sec ratelimit */
508 	if (!(skb->dev->flags&IFF_LOOPBACK) && !icmpv6_global_allow(type))
509 		goto out_bh_enable;
510 
511 	mip6_addr_swap(skb);
512 
513 	memset(&fl6, 0, sizeof(fl6));
514 	fl6.flowi6_proto = IPPROTO_ICMPV6;
515 	fl6.daddr = hdr->saddr;
516 	if (force_saddr)
517 		saddr = force_saddr;
518 	if (saddr)
519 		fl6.saddr = *saddr;
520 	fl6.flowi6_mark = mark;
521 	fl6.flowi6_oif = iif;
522 	fl6.fl6_icmp_type = type;
523 	fl6.fl6_icmp_code = code;
524 	fl6.flowi6_uid = sock_net_uid(net, NULL);
525 	fl6.mp_hash = rt6_multipath_hash(&fl6, skb);
526 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
527 
528 	sk = icmpv6_xmit_lock(net);
529 	if (!sk)
530 		goto out_bh_enable;
531 
532 	sk->sk_mark = mark;
533 	np = inet6_sk(sk);
534 
535 	if (!icmpv6_xrlim_allow(sk, type, &fl6))
536 		goto out;
537 
538 	tmp_hdr.icmp6_type = type;
539 	tmp_hdr.icmp6_code = code;
540 	tmp_hdr.icmp6_cksum = 0;
541 	tmp_hdr.icmp6_pointer = htonl(info);
542 
543 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
544 		fl6.flowi6_oif = np->mcast_oif;
545 	else if (!fl6.flowi6_oif)
546 		fl6.flowi6_oif = np->ucast_oif;
547 
548 	ipc6.tclass = np->tclass;
549 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
550 
551 	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
552 	if (IS_ERR(dst))
553 		goto out;
554 
555 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
556 	ipc6.dontfrag = np->dontfrag;
557 	ipc6.opt = NULL;
558 
559 	msg.skb = skb;
560 	msg.offset = skb_network_offset(skb);
561 	msg.type = type;
562 
563 	len = skb->len - msg.offset;
564 	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(struct icmp6hdr));
565 	if (len < 0) {
566 		net_dbg_ratelimited("icmp: len problem [%pI6c > %pI6c]\n",
567 				    &hdr->saddr, &hdr->daddr);
568 		goto out_dst_release;
569 	}
570 
571 	rcu_read_lock();
572 	idev = __in6_dev_get(skb->dev);
573 
574 	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
575 			    len + sizeof(struct icmp6hdr),
576 			    sizeof(struct icmp6hdr),
577 			    &ipc6, &fl6, (struct rt6_info *)dst,
578 			    MSG_DONTWAIT, &sockc_unused)) {
579 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
580 		ip6_flush_pending_frames(sk);
581 	} else {
582 		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
583 					   len + sizeof(struct icmp6hdr));
584 	}
585 	rcu_read_unlock();
586 out_dst_release:
587 	dst_release(dst);
588 out:
589 	icmpv6_xmit_unlock(sk);
590 out_bh_enable:
591 	local_bh_enable();
592 }
593 
594 /* Slightly more convenient version of icmp6_send.
595  */
596 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
597 {
598 	icmp6_send(skb, ICMPV6_PARAMPROB, code, pos, NULL);
599 	kfree_skb(skb);
600 }
601 
602 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
603  * if sufficient data bytes are available
604  * @nhs is the size of the tunnel header(s) :
605  *  Either an IPv4 header for SIT encap
606  *         an IPv4 header + GRE header for GRE encap
607  */
608 int ip6_err_gen_icmpv6_unreach(struct sk_buff *skb, int nhs, int type,
609 			       unsigned int data_len)
610 {
611 	struct in6_addr temp_saddr;
612 	struct rt6_info *rt;
613 	struct sk_buff *skb2;
614 	u32 info = 0;
615 
616 	if (!pskb_may_pull(skb, nhs + sizeof(struct ipv6hdr) + 8))
617 		return 1;
618 
619 	/* RFC 4884 (partial) support for ICMP extensions */
620 	if (data_len < 128 || (data_len & 7) || skb->len < data_len)
621 		data_len = 0;
622 
623 	skb2 = data_len ? skb_copy(skb, GFP_ATOMIC) : skb_clone(skb, GFP_ATOMIC);
624 
625 	if (!skb2)
626 		return 1;
627 
628 	skb_dst_drop(skb2);
629 	skb_pull(skb2, nhs);
630 	skb_reset_network_header(skb2);
631 
632 	rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
633 
634 	if (rt && rt->dst.dev)
635 		skb2->dev = rt->dst.dev;
636 
637 	ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, &temp_saddr);
638 
639 	if (data_len) {
640 		/* RFC 4884 (partial) support :
641 		 * insert 0 padding at the end, before the extensions
642 		 */
643 		__skb_push(skb2, nhs);
644 		skb_reset_network_header(skb2);
645 		memmove(skb2->data, skb2->data + nhs, data_len - nhs);
646 		memset(skb2->data + data_len - nhs, 0, nhs);
647 		/* RFC 4884 4.5 : Length is measured in 64-bit words,
648 		 * and stored in reserved[0]
649 		 */
650 		info = (data_len/8) << 24;
651 	}
652 	if (type == ICMP_TIME_EXCEEDED)
653 		icmp6_send(skb2, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
654 			   info, &temp_saddr);
655 	else
656 		icmp6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH,
657 			   info, &temp_saddr);
658 	if (rt)
659 		ip6_rt_put(rt);
660 
661 	kfree_skb(skb2);
662 
663 	return 0;
664 }
665 EXPORT_SYMBOL(ip6_err_gen_icmpv6_unreach);
666 
667 static void icmpv6_echo_reply(struct sk_buff *skb)
668 {
669 	struct net *net = dev_net(skb->dev);
670 	struct sock *sk;
671 	struct inet6_dev *idev;
672 	struct ipv6_pinfo *np;
673 	const struct in6_addr *saddr = NULL;
674 	struct icmp6hdr *icmph = icmp6_hdr(skb);
675 	struct icmp6hdr tmp_hdr;
676 	struct flowi6 fl6;
677 	struct icmpv6_msg msg;
678 	struct dst_entry *dst;
679 	struct ipcm6_cookie ipc6;
680 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
681 	struct sockcm_cookie sockc_unused = {0};
682 
683 	saddr = &ipv6_hdr(skb)->daddr;
684 
685 	if (!ipv6_unicast_destination(skb) &&
686 	    !(net->ipv6.sysctl.anycast_src_echo_reply &&
687 	      ipv6_anycast_destination(skb_dst(skb), saddr)))
688 		saddr = NULL;
689 
690 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
691 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
692 
693 	memset(&fl6, 0, sizeof(fl6));
694 	fl6.flowi6_proto = IPPROTO_ICMPV6;
695 	fl6.daddr = ipv6_hdr(skb)->saddr;
696 	if (saddr)
697 		fl6.saddr = *saddr;
698 	fl6.flowi6_oif = icmp6_iif(skb);
699 	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
700 	fl6.flowi6_mark = mark;
701 	fl6.flowi6_uid = sock_net_uid(net, NULL);
702 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
703 
704 	local_bh_disable();
705 	sk = icmpv6_xmit_lock(net);
706 	if (!sk)
707 		goto out_bh_enable;
708 	sk->sk_mark = mark;
709 	np = inet6_sk(sk);
710 
711 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
712 		fl6.flowi6_oif = np->mcast_oif;
713 	else if (!fl6.flowi6_oif)
714 		fl6.flowi6_oif = np->ucast_oif;
715 
716 	if (ip6_dst_lookup(net, sk, &dst, &fl6))
717 		goto out;
718 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
719 	if (IS_ERR(dst))
720 		goto out;
721 
722 	idev = __in6_dev_get(skb->dev);
723 
724 	msg.skb = skb;
725 	msg.offset = 0;
726 	msg.type = ICMPV6_ECHO_REPLY;
727 
728 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
729 	ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb));
730 	ipc6.dontfrag = np->dontfrag;
731 	ipc6.opt = NULL;
732 
733 	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
734 			    skb->len + sizeof(struct icmp6hdr),
735 			    sizeof(struct icmp6hdr), &ipc6, &fl6,
736 			    (struct rt6_info *)dst, MSG_DONTWAIT,
737 			    &sockc_unused)) {
738 		__ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
739 		ip6_flush_pending_frames(sk);
740 	} else {
741 		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
742 					   skb->len + sizeof(struct icmp6hdr));
743 	}
744 	dst_release(dst);
745 out:
746 	icmpv6_xmit_unlock(sk);
747 out_bh_enable:
748 	local_bh_enable();
749 }
750 
751 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
752 {
753 	const struct inet6_protocol *ipprot;
754 	int inner_offset;
755 	__be16 frag_off;
756 	u8 nexthdr;
757 	struct net *net = dev_net(skb->dev);
758 
759 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
760 		goto out;
761 
762 	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
763 	if (ipv6_ext_hdr(nexthdr)) {
764 		/* now skip over extension headers */
765 		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
766 						&nexthdr, &frag_off);
767 		if (inner_offset < 0)
768 			goto out;
769 	} else {
770 		inner_offset = sizeof(struct ipv6hdr);
771 	}
772 
773 	/* Checkin header including 8 bytes of inner protocol header. */
774 	if (!pskb_may_pull(skb, inner_offset+8))
775 		goto out;
776 
777 	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
778 	   Without this we will not able f.e. to make source routed
779 	   pmtu discovery.
780 	   Corresponding argument (opt) to notifiers is already added.
781 	   --ANK (980726)
782 	 */
783 
784 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
785 	if (ipprot && ipprot->err_handler)
786 		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
787 
788 	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
789 	return;
790 
791 out:
792 	__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
793 }
794 
795 /*
796  *	Handle icmp messages
797  */
798 
799 static int icmpv6_rcv(struct sk_buff *skb)
800 {
801 	struct net_device *dev = skb->dev;
802 	struct inet6_dev *idev = __in6_dev_get(dev);
803 	const struct in6_addr *saddr, *daddr;
804 	struct icmp6hdr *hdr;
805 	u8 type;
806 	bool success = false;
807 
808 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
809 		struct sec_path *sp = skb_sec_path(skb);
810 		int nh;
811 
812 		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
813 				 XFRM_STATE_ICMP))
814 			goto drop_no_count;
815 
816 		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
817 			goto drop_no_count;
818 
819 		nh = skb_network_offset(skb);
820 		skb_set_network_header(skb, sizeof(*hdr));
821 
822 		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
823 			goto drop_no_count;
824 
825 		skb_set_network_header(skb, nh);
826 	}
827 
828 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS);
829 
830 	saddr = &ipv6_hdr(skb)->saddr;
831 	daddr = &ipv6_hdr(skb)->daddr;
832 
833 	if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) {
834 		net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n",
835 				    saddr, daddr);
836 		goto csum_error;
837 	}
838 
839 	if (!pskb_pull(skb, sizeof(*hdr)))
840 		goto discard_it;
841 
842 	hdr = icmp6_hdr(skb);
843 
844 	type = hdr->icmp6_type;
845 
846 	ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type);
847 
848 	switch (type) {
849 	case ICMPV6_ECHO_REQUEST:
850 		icmpv6_echo_reply(skb);
851 		break;
852 
853 	case ICMPV6_ECHO_REPLY:
854 		success = ping_rcv(skb);
855 		break;
856 
857 	case ICMPV6_PKT_TOOBIG:
858 		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
859 		   standard destination cache. Seems, only "advanced"
860 		   destination cache will allow to solve this problem
861 		   --ANK (980726)
862 		 */
863 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
864 			goto discard_it;
865 		hdr = icmp6_hdr(skb);
866 
867 		/* to notify */
868 		/* fall through */
869 	case ICMPV6_DEST_UNREACH:
870 	case ICMPV6_TIME_EXCEED:
871 	case ICMPV6_PARAMPROB:
872 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
873 		break;
874 
875 	case NDISC_ROUTER_SOLICITATION:
876 	case NDISC_ROUTER_ADVERTISEMENT:
877 	case NDISC_NEIGHBOUR_SOLICITATION:
878 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
879 	case NDISC_REDIRECT:
880 		ndisc_rcv(skb);
881 		break;
882 
883 	case ICMPV6_MGM_QUERY:
884 		igmp6_event_query(skb);
885 		break;
886 
887 	case ICMPV6_MGM_REPORT:
888 		igmp6_event_report(skb);
889 		break;
890 
891 	case ICMPV6_MGM_REDUCTION:
892 	case ICMPV6_NI_QUERY:
893 	case ICMPV6_NI_REPLY:
894 	case ICMPV6_MLD2_REPORT:
895 	case ICMPV6_DHAAD_REQUEST:
896 	case ICMPV6_DHAAD_REPLY:
897 	case ICMPV6_MOBILE_PREFIX_SOL:
898 	case ICMPV6_MOBILE_PREFIX_ADV:
899 		break;
900 
901 	default:
902 		/* informational */
903 		if (type & ICMPV6_INFOMSG_MASK)
904 			break;
905 
906 		net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n",
907 				    saddr, daddr);
908 
909 		/*
910 		 * error of unknown type.
911 		 * must pass to upper level
912 		 */
913 
914 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
915 	}
916 
917 	/* until the v6 path can be better sorted assume failure and
918 	 * preserve the status quo behaviour for the rest of the paths to here
919 	 */
920 	if (success)
921 		consume_skb(skb);
922 	else
923 		kfree_skb(skb);
924 
925 	return 0;
926 
927 csum_error:
928 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
929 discard_it:
930 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS);
931 drop_no_count:
932 	kfree_skb(skb);
933 	return 0;
934 }
935 
936 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
937 		      u8 type,
938 		      const struct in6_addr *saddr,
939 		      const struct in6_addr *daddr,
940 		      int oif)
941 {
942 	memset(fl6, 0, sizeof(*fl6));
943 	fl6->saddr = *saddr;
944 	fl6->daddr = *daddr;
945 	fl6->flowi6_proto	= IPPROTO_ICMPV6;
946 	fl6->fl6_icmp_type	= type;
947 	fl6->fl6_icmp_code	= 0;
948 	fl6->flowi6_oif		= oif;
949 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
950 }
951 
952 static int __net_init icmpv6_sk_init(struct net *net)
953 {
954 	struct sock *sk;
955 	int err, i, j;
956 
957 	net->ipv6.icmp_sk =
958 		kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
959 	if (!net->ipv6.icmp_sk)
960 		return -ENOMEM;
961 
962 	for_each_possible_cpu(i) {
963 		err = inet_ctl_sock_create(&sk, PF_INET6,
964 					   SOCK_RAW, IPPROTO_ICMPV6, net);
965 		if (err < 0) {
966 			pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
967 			       err);
968 			goto fail;
969 		}
970 
971 		net->ipv6.icmp_sk[i] = sk;
972 
973 		/* Enough space for 2 64K ICMP packets, including
974 		 * sk_buff struct overhead.
975 		 */
976 		sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
977 	}
978 	return 0;
979 
980  fail:
981 	for (j = 0; j < i; j++)
982 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
983 	kfree(net->ipv6.icmp_sk);
984 	return err;
985 }
986 
987 static void __net_exit icmpv6_sk_exit(struct net *net)
988 {
989 	int i;
990 
991 	for_each_possible_cpu(i) {
992 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
993 	}
994 	kfree(net->ipv6.icmp_sk);
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 };
1113 
1114 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1115 {
1116 	struct ctl_table *table;
1117 
1118 	table = kmemdup(ipv6_icmp_table_template,
1119 			sizeof(ipv6_icmp_table_template),
1120 			GFP_KERNEL);
1121 
1122 	if (table)
1123 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
1124 
1125 	return table;
1126 }
1127 #endif
1128