xref: /openbmc/linux/net/ipv6/icmp.c (revision c000c4f1)
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(net, &fl6, skb, NULL);
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,
633 			skb, 0);
634 
635 	if (rt && rt->dst.dev)
636 		skb2->dev = rt->dst.dev;
637 
638 	ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, &temp_saddr);
639 
640 	if (data_len) {
641 		/* RFC 4884 (partial) support :
642 		 * insert 0 padding at the end, before the extensions
643 		 */
644 		__skb_push(skb2, nhs);
645 		skb_reset_network_header(skb2);
646 		memmove(skb2->data, skb2->data + nhs, data_len - nhs);
647 		memset(skb2->data + data_len - nhs, 0, nhs);
648 		/* RFC 4884 4.5 : Length is measured in 64-bit words,
649 		 * and stored in reserved[0]
650 		 */
651 		info = (data_len/8) << 24;
652 	}
653 	if (type == ICMP_TIME_EXCEEDED)
654 		icmp6_send(skb2, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
655 			   info, &temp_saddr);
656 	else
657 		icmp6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH,
658 			   info, &temp_saddr);
659 	if (rt)
660 		ip6_rt_put(rt);
661 
662 	kfree_skb(skb2);
663 
664 	return 0;
665 }
666 EXPORT_SYMBOL(ip6_err_gen_icmpv6_unreach);
667 
668 static void icmpv6_echo_reply(struct sk_buff *skb)
669 {
670 	struct net *net = dev_net(skb->dev);
671 	struct sock *sk;
672 	struct inet6_dev *idev;
673 	struct ipv6_pinfo *np;
674 	const struct in6_addr *saddr = NULL;
675 	struct icmp6hdr *icmph = icmp6_hdr(skb);
676 	struct icmp6hdr tmp_hdr;
677 	struct flowi6 fl6;
678 	struct icmpv6_msg msg;
679 	struct dst_entry *dst;
680 	struct ipcm6_cookie ipc6;
681 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
682 	struct sockcm_cookie sockc_unused = {0};
683 
684 	saddr = &ipv6_hdr(skb)->daddr;
685 
686 	if (!ipv6_unicast_destination(skb) &&
687 	    !(net->ipv6.sysctl.anycast_src_echo_reply &&
688 	      ipv6_anycast_destination(skb_dst(skb), saddr)))
689 		saddr = NULL;
690 
691 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
692 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
693 
694 	memset(&fl6, 0, sizeof(fl6));
695 	fl6.flowi6_proto = IPPROTO_ICMPV6;
696 	fl6.daddr = ipv6_hdr(skb)->saddr;
697 	if (saddr)
698 		fl6.saddr = *saddr;
699 	fl6.flowi6_oif = icmp6_iif(skb);
700 	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
701 	fl6.flowi6_mark = mark;
702 	fl6.flowi6_uid = sock_net_uid(net, NULL);
703 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
704 
705 	local_bh_disable();
706 	sk = icmpv6_xmit_lock(net);
707 	if (!sk)
708 		goto out_bh_enable;
709 	sk->sk_mark = mark;
710 	np = inet6_sk(sk);
711 
712 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
713 		fl6.flowi6_oif = np->mcast_oif;
714 	else if (!fl6.flowi6_oif)
715 		fl6.flowi6_oif = np->ucast_oif;
716 
717 	if (ip6_dst_lookup(net, sk, &dst, &fl6))
718 		goto out;
719 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
720 	if (IS_ERR(dst))
721 		goto out;
722 
723 	idev = __in6_dev_get(skb->dev);
724 
725 	msg.skb = skb;
726 	msg.offset = 0;
727 	msg.type = ICMPV6_ECHO_REPLY;
728 
729 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
730 	ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb));
731 	ipc6.dontfrag = np->dontfrag;
732 	ipc6.opt = NULL;
733 
734 	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
735 			    skb->len + sizeof(struct icmp6hdr),
736 			    sizeof(struct icmp6hdr), &ipc6, &fl6,
737 			    (struct rt6_info *)dst, MSG_DONTWAIT,
738 			    &sockc_unused)) {
739 		__ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
740 		ip6_flush_pending_frames(sk);
741 	} else {
742 		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
743 					   skb->len + sizeof(struct icmp6hdr));
744 	}
745 	dst_release(dst);
746 out:
747 	icmpv6_xmit_unlock(sk);
748 out_bh_enable:
749 	local_bh_enable();
750 }
751 
752 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
753 {
754 	const struct inet6_protocol *ipprot;
755 	int inner_offset;
756 	__be16 frag_off;
757 	u8 nexthdr;
758 	struct net *net = dev_net(skb->dev);
759 
760 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
761 		goto out;
762 
763 	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
764 	if (ipv6_ext_hdr(nexthdr)) {
765 		/* now skip over extension headers */
766 		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
767 						&nexthdr, &frag_off);
768 		if (inner_offset < 0)
769 			goto out;
770 	} else {
771 		inner_offset = sizeof(struct ipv6hdr);
772 	}
773 
774 	/* Checkin header including 8 bytes of inner protocol header. */
775 	if (!pskb_may_pull(skb, inner_offset+8))
776 		goto out;
777 
778 	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
779 	   Without this we will not able f.e. to make source routed
780 	   pmtu discovery.
781 	   Corresponding argument (opt) to notifiers is already added.
782 	   --ANK (980726)
783 	 */
784 
785 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
786 	if (ipprot && ipprot->err_handler)
787 		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
788 
789 	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
790 	return;
791 
792 out:
793 	__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
794 }
795 
796 /*
797  *	Handle icmp messages
798  */
799 
800 static int icmpv6_rcv(struct sk_buff *skb)
801 {
802 	struct net_device *dev = skb->dev;
803 	struct inet6_dev *idev = __in6_dev_get(dev);
804 	const struct in6_addr *saddr, *daddr;
805 	struct icmp6hdr *hdr;
806 	u8 type;
807 	bool success = false;
808 
809 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
810 		struct sec_path *sp = skb_sec_path(skb);
811 		int nh;
812 
813 		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
814 				 XFRM_STATE_ICMP))
815 			goto drop_no_count;
816 
817 		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
818 			goto drop_no_count;
819 
820 		nh = skb_network_offset(skb);
821 		skb_set_network_header(skb, sizeof(*hdr));
822 
823 		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
824 			goto drop_no_count;
825 
826 		skb_set_network_header(skb, nh);
827 	}
828 
829 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS);
830 
831 	saddr = &ipv6_hdr(skb)->saddr;
832 	daddr = &ipv6_hdr(skb)->daddr;
833 
834 	if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) {
835 		net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n",
836 				    saddr, daddr);
837 		goto csum_error;
838 	}
839 
840 	if (!pskb_pull(skb, sizeof(*hdr)))
841 		goto discard_it;
842 
843 	hdr = icmp6_hdr(skb);
844 
845 	type = hdr->icmp6_type;
846 
847 	ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type);
848 
849 	switch (type) {
850 	case ICMPV6_ECHO_REQUEST:
851 		icmpv6_echo_reply(skb);
852 		break;
853 
854 	case ICMPV6_ECHO_REPLY:
855 		success = ping_rcv(skb);
856 		break;
857 
858 	case ICMPV6_PKT_TOOBIG:
859 		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
860 		   standard destination cache. Seems, only "advanced"
861 		   destination cache will allow to solve this problem
862 		   --ANK (980726)
863 		 */
864 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
865 			goto discard_it;
866 		hdr = icmp6_hdr(skb);
867 
868 		/* to notify */
869 		/* fall through */
870 	case ICMPV6_DEST_UNREACH:
871 	case ICMPV6_TIME_EXCEED:
872 	case ICMPV6_PARAMPROB:
873 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
874 		break;
875 
876 	case NDISC_ROUTER_SOLICITATION:
877 	case NDISC_ROUTER_ADVERTISEMENT:
878 	case NDISC_NEIGHBOUR_SOLICITATION:
879 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
880 	case NDISC_REDIRECT:
881 		ndisc_rcv(skb);
882 		break;
883 
884 	case ICMPV6_MGM_QUERY:
885 		igmp6_event_query(skb);
886 		break;
887 
888 	case ICMPV6_MGM_REPORT:
889 		igmp6_event_report(skb);
890 		break;
891 
892 	case ICMPV6_MGM_REDUCTION:
893 	case ICMPV6_NI_QUERY:
894 	case ICMPV6_NI_REPLY:
895 	case ICMPV6_MLD2_REPORT:
896 	case ICMPV6_DHAAD_REQUEST:
897 	case ICMPV6_DHAAD_REPLY:
898 	case ICMPV6_MOBILE_PREFIX_SOL:
899 	case ICMPV6_MOBILE_PREFIX_ADV:
900 		break;
901 
902 	default:
903 		/* informational */
904 		if (type & ICMPV6_INFOMSG_MASK)
905 			break;
906 
907 		net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n",
908 				    saddr, daddr);
909 
910 		/*
911 		 * error of unknown type.
912 		 * must pass to upper level
913 		 */
914 
915 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
916 	}
917 
918 	/* until the v6 path can be better sorted assume failure and
919 	 * preserve the status quo behaviour for the rest of the paths to here
920 	 */
921 	if (success)
922 		consume_skb(skb);
923 	else
924 		kfree_skb(skb);
925 
926 	return 0;
927 
928 csum_error:
929 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
930 discard_it:
931 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS);
932 drop_no_count:
933 	kfree_skb(skb);
934 	return 0;
935 }
936 
937 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
938 		      u8 type,
939 		      const struct in6_addr *saddr,
940 		      const struct in6_addr *daddr,
941 		      int oif)
942 {
943 	memset(fl6, 0, sizeof(*fl6));
944 	fl6->saddr = *saddr;
945 	fl6->daddr = *daddr;
946 	fl6->flowi6_proto	= IPPROTO_ICMPV6;
947 	fl6->fl6_icmp_type	= type;
948 	fl6->fl6_icmp_code	= 0;
949 	fl6->flowi6_oif		= oif;
950 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
951 }
952 
953 static int __net_init icmpv6_sk_init(struct net *net)
954 {
955 	struct sock *sk;
956 	int err, i, j;
957 
958 	net->ipv6.icmp_sk =
959 		kcalloc(nr_cpu_ids, sizeof(struct sock *), GFP_KERNEL);
960 	if (!net->ipv6.icmp_sk)
961 		return -ENOMEM;
962 
963 	for_each_possible_cpu(i) {
964 		err = inet_ctl_sock_create(&sk, PF_INET6,
965 					   SOCK_RAW, IPPROTO_ICMPV6, net);
966 		if (err < 0) {
967 			pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
968 			       err);
969 			goto fail;
970 		}
971 
972 		net->ipv6.icmp_sk[i] = sk;
973 
974 		/* Enough space for 2 64K ICMP packets, including
975 		 * sk_buff struct overhead.
976 		 */
977 		sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
978 	}
979 	return 0;
980 
981  fail:
982 	for (j = 0; j < i; j++)
983 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
984 	kfree(net->ipv6.icmp_sk);
985 	return err;
986 }
987 
988 static void __net_exit icmpv6_sk_exit(struct net *net)
989 {
990 	int i;
991 
992 	for_each_possible_cpu(i) {
993 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
994 	}
995 	kfree(net->ipv6.icmp_sk);
996 }
997 
998 static struct pernet_operations icmpv6_sk_ops = {
999 	.init = icmpv6_sk_init,
1000 	.exit = icmpv6_sk_exit,
1001 };
1002 
1003 int __init icmpv6_init(void)
1004 {
1005 	int err;
1006 
1007 	err = register_pernet_subsys(&icmpv6_sk_ops);
1008 	if (err < 0)
1009 		return err;
1010 
1011 	err = -EAGAIN;
1012 	if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
1013 		goto fail;
1014 
1015 	err = inet6_register_icmp_sender(icmp6_send);
1016 	if (err)
1017 		goto sender_reg_err;
1018 	return 0;
1019 
1020 sender_reg_err:
1021 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
1022 fail:
1023 	pr_err("Failed to register ICMP6 protocol\n");
1024 	unregister_pernet_subsys(&icmpv6_sk_ops);
1025 	return err;
1026 }
1027 
1028 void icmpv6_cleanup(void)
1029 {
1030 	inet6_unregister_icmp_sender(icmp6_send);
1031 	unregister_pernet_subsys(&icmpv6_sk_ops);
1032 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
1033 }
1034 
1035 
1036 static const struct icmp6_err {
1037 	int err;
1038 	int fatal;
1039 } tab_unreach[] = {
1040 	{	/* NOROUTE */
1041 		.err	= ENETUNREACH,
1042 		.fatal	= 0,
1043 	},
1044 	{	/* ADM_PROHIBITED */
1045 		.err	= EACCES,
1046 		.fatal	= 1,
1047 	},
1048 	{	/* Was NOT_NEIGHBOUR, now reserved */
1049 		.err	= EHOSTUNREACH,
1050 		.fatal	= 0,
1051 	},
1052 	{	/* ADDR_UNREACH	*/
1053 		.err	= EHOSTUNREACH,
1054 		.fatal	= 0,
1055 	},
1056 	{	/* PORT_UNREACH	*/
1057 		.err	= ECONNREFUSED,
1058 		.fatal	= 1,
1059 	},
1060 	{	/* POLICY_FAIL */
1061 		.err	= EACCES,
1062 		.fatal	= 1,
1063 	},
1064 	{	/* REJECT_ROUTE	*/
1065 		.err	= EACCES,
1066 		.fatal	= 1,
1067 	},
1068 };
1069 
1070 int icmpv6_err_convert(u8 type, u8 code, int *err)
1071 {
1072 	int fatal = 0;
1073 
1074 	*err = EPROTO;
1075 
1076 	switch (type) {
1077 	case ICMPV6_DEST_UNREACH:
1078 		fatal = 1;
1079 		if (code < ARRAY_SIZE(tab_unreach)) {
1080 			*err  = tab_unreach[code].err;
1081 			fatal = tab_unreach[code].fatal;
1082 		}
1083 		break;
1084 
1085 	case ICMPV6_PKT_TOOBIG:
1086 		*err = EMSGSIZE;
1087 		break;
1088 
1089 	case ICMPV6_PARAMPROB:
1090 		*err = EPROTO;
1091 		fatal = 1;
1092 		break;
1093 
1094 	case ICMPV6_TIME_EXCEED:
1095 		*err = EHOSTUNREACH;
1096 		break;
1097 	}
1098 
1099 	return fatal;
1100 }
1101 EXPORT_SYMBOL(icmpv6_err_convert);
1102 
1103 #ifdef CONFIG_SYSCTL
1104 static struct ctl_table ipv6_icmp_table_template[] = {
1105 	{
1106 		.procname	= "ratelimit",
1107 		.data		= &init_net.ipv6.sysctl.icmpv6_time,
1108 		.maxlen		= sizeof(int),
1109 		.mode		= 0644,
1110 		.proc_handler	= proc_dointvec_ms_jiffies,
1111 	},
1112 	{ },
1113 };
1114 
1115 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1116 {
1117 	struct ctl_table *table;
1118 
1119 	table = kmemdup(ipv6_icmp_table_template,
1120 			sizeof(ipv6_icmp_table_template),
1121 			GFP_KERNEL);
1122 
1123 	if (table)
1124 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
1125 
1126 	return table;
1127 }
1128 #endif
1129