xref: /openbmc/linux/net/ipv6/icmp.c (revision 367b8112)
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 #include <linux/module.h>
33 #include <linux/errno.h>
34 #include <linux/types.h>
35 #include <linux/socket.h>
36 #include <linux/in.h>
37 #include <linux/kernel.h>
38 #include <linux/sockios.h>
39 #include <linux/net.h>
40 #include <linux/skbuff.h>
41 #include <linux/init.h>
42 #include <linux/netfilter.h>
43 
44 #ifdef CONFIG_SYSCTL
45 #include <linux/sysctl.h>
46 #endif
47 
48 #include <linux/inet.h>
49 #include <linux/netdevice.h>
50 #include <linux/icmpv6.h>
51 
52 #include <net/ip.h>
53 #include <net/sock.h>
54 
55 #include <net/ipv6.h>
56 #include <net/ip6_checksum.h>
57 #include <net/protocol.h>
58 #include <net/raw.h>
59 #include <net/rawv6.h>
60 #include <net/transp_v6.h>
61 #include <net/ip6_route.h>
62 #include <net/addrconf.h>
63 #include <net/icmp.h>
64 #include <net/xfrm.h>
65 #include <net/inet_common.h>
66 
67 #include <asm/uaccess.h>
68 #include <asm/system.h>
69 
70 DEFINE_SNMP_STAT(struct icmpv6_mib, icmpv6_statistics) __read_mostly;
71 EXPORT_SYMBOL(icmpv6_statistics);
72 DEFINE_SNMP_STAT(struct icmpv6msg_mib, icmpv6msg_statistics) __read_mostly;
73 EXPORT_SYMBOL(icmpv6msg_statistics);
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 int icmpv6_rcv(struct sk_buff *skb);
88 
89 static struct inet6_protocol icmpv6_protocol = {
90 	.handler	=	icmpv6_rcv,
91 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
92 };
93 
94 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
95 {
96 	struct sock *sk;
97 
98 	local_bh_disable();
99 
100 	sk = icmpv6_sk(net);
101 	if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
102 		/* This can happen if the output path (f.e. SIT or
103 		 * ip6ip6 tunnel) signals dst_link_failure() for an
104 		 * outgoing ICMP6 packet.
105 		 */
106 		local_bh_enable();
107 		return NULL;
108 	}
109 	return sk;
110 }
111 
112 static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
113 {
114 	spin_unlock_bh(&sk->sk_lock.slock);
115 }
116 
117 /*
118  * Slightly more convenient version of icmpv6_send.
119  */
120 void icmpv6_param_prob(struct sk_buff *skb, int code, int pos)
121 {
122 	icmpv6_send(skb, ICMPV6_PARAMPROB, code, pos, skb->dev);
123 	kfree_skb(skb);
124 }
125 
126 /*
127  * Figure out, may we reply to this packet with icmp error.
128  *
129  * We do not reply, if:
130  *	- it was icmp error message.
131  *	- it is truncated, so that it is known, that protocol is ICMPV6
132  *	  (i.e. in the middle of some exthdr)
133  *
134  *	--ANK (980726)
135  */
136 
137 static int is_ineligible(struct sk_buff *skb)
138 {
139 	int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
140 	int len = skb->len - ptr;
141 	__u8 nexthdr = ipv6_hdr(skb)->nexthdr;
142 
143 	if (len < 0)
144 		return 1;
145 
146 	ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr);
147 	if (ptr < 0)
148 		return 0;
149 	if (nexthdr == IPPROTO_ICMPV6) {
150 		u8 _type, *tp;
151 		tp = skb_header_pointer(skb,
152 			ptr+offsetof(struct icmp6hdr, icmp6_type),
153 			sizeof(_type), &_type);
154 		if (tp == NULL ||
155 		    !(*tp & ICMPV6_INFOMSG_MASK))
156 			return 1;
157 	}
158 	return 0;
159 }
160 
161 /*
162  * Check the ICMP output rate limit
163  */
164 static inline int icmpv6_xrlim_allow(struct sock *sk, int type,
165 				     struct flowi *fl)
166 {
167 	struct dst_entry *dst;
168 	struct net *net = sock_net(sk);
169 	int res = 0;
170 
171 	/* Informational messages are not limited. */
172 	if (type & ICMPV6_INFOMSG_MASK)
173 		return 1;
174 
175 	/* Do not limit pmtu discovery, it would break it. */
176 	if (type == ICMPV6_PKT_TOOBIG)
177 		return 1;
178 
179 	/*
180 	 * Look up the output route.
181 	 * XXX: perhaps the expire for routing entries cloned by
182 	 * this lookup should be more aggressive (not longer than timeout).
183 	 */
184 	dst = ip6_route_output(net, sk, fl);
185 	if (dst->error) {
186 		IP6_INC_STATS(net, ip6_dst_idev(dst),
187 			      IPSTATS_MIB_OUTNOROUTES);
188 	} else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
189 		res = 1;
190 	} else {
191 		struct rt6_info *rt = (struct rt6_info *)dst;
192 		int tmo = net->ipv6.sysctl.icmpv6_time;
193 
194 		/* Give more bandwidth to wider prefixes. */
195 		if (rt->rt6i_dst.plen < 128)
196 			tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
197 
198 		res = xrlim_allow(dst, tmo);
199 	}
200 	dst_release(dst);
201 	return res;
202 }
203 
204 /*
205  *	an inline helper for the "simple" if statement below
206  *	checks if parameter problem report is caused by an
207  *	unrecognized IPv6 option that has the Option Type
208  *	highest-order two bits set to 10
209  */
210 
211 static __inline__ int opt_unrec(struct sk_buff *skb, __u32 offset)
212 {
213 	u8 _optval, *op;
214 
215 	offset += skb_network_offset(skb);
216 	op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
217 	if (op == NULL)
218 		return 1;
219 	return (*op & 0xC0) == 0x80;
220 }
221 
222 static int icmpv6_push_pending_frames(struct sock *sk, struct flowi *fl, struct icmp6hdr *thdr, int len)
223 {
224 	struct sk_buff *skb;
225 	struct icmp6hdr *icmp6h;
226 	int err = 0;
227 
228 	if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
229 		goto out;
230 
231 	icmp6h = icmp6_hdr(skb);
232 	memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
233 	icmp6h->icmp6_cksum = 0;
234 
235 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
236 		skb->csum = csum_partial((char *)icmp6h,
237 					sizeof(struct icmp6hdr), skb->csum);
238 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src,
239 						      &fl->fl6_dst,
240 						      len, fl->proto,
241 						      skb->csum);
242 	} else {
243 		__wsum tmp_csum = 0;
244 
245 		skb_queue_walk(&sk->sk_write_queue, skb) {
246 			tmp_csum = csum_add(tmp_csum, skb->csum);
247 		}
248 
249 		tmp_csum = csum_partial((char *)icmp6h,
250 					sizeof(struct icmp6hdr), tmp_csum);
251 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src,
252 						      &fl->fl6_dst,
253 						      len, fl->proto,
254 						      tmp_csum);
255 	}
256 	ip6_push_pending_frames(sk);
257 out:
258 	return err;
259 }
260 
261 struct icmpv6_msg {
262 	struct sk_buff	*skb;
263 	int		offset;
264 	uint8_t		type;
265 };
266 
267 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
268 {
269 	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
270 	struct sk_buff *org_skb = msg->skb;
271 	__wsum csum = 0;
272 
273 	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
274 				      to, len, csum);
275 	skb->csum = csum_block_add(skb->csum, csum, odd);
276 	if (!(msg->type & ICMPV6_INFOMSG_MASK))
277 		nf_ct_attach(skb, org_skb);
278 	return 0;
279 }
280 
281 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
282 static void mip6_addr_swap(struct sk_buff *skb)
283 {
284 	struct ipv6hdr *iph = ipv6_hdr(skb);
285 	struct inet6_skb_parm *opt = IP6CB(skb);
286 	struct ipv6_destopt_hao *hao;
287 	struct in6_addr tmp;
288 	int off;
289 
290 	if (opt->dsthao) {
291 		off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
292 		if (likely(off >= 0)) {
293 			hao = (struct ipv6_destopt_hao *)
294 					(skb_network_header(skb) + off);
295 			ipv6_addr_copy(&tmp, &iph->saddr);
296 			ipv6_addr_copy(&iph->saddr, &hao->addr);
297 			ipv6_addr_copy(&hao->addr, &tmp);
298 		}
299 	}
300 }
301 #else
302 static inline void mip6_addr_swap(struct sk_buff *skb) {}
303 #endif
304 
305 /*
306  *	Send an ICMP message in response to a packet in error
307  */
308 void icmpv6_send(struct sk_buff *skb, int type, int code, __u32 info,
309 		 struct net_device *dev)
310 {
311 	struct net *net = dev_net(skb->dev);
312 	struct inet6_dev *idev = NULL;
313 	struct ipv6hdr *hdr = ipv6_hdr(skb);
314 	struct sock *sk;
315 	struct ipv6_pinfo *np;
316 	struct in6_addr *saddr = NULL;
317 	struct dst_entry *dst;
318 	struct dst_entry *dst2;
319 	struct icmp6hdr tmp_hdr;
320 	struct flowi fl;
321 	struct flowi fl2;
322 	struct icmpv6_msg msg;
323 	int iif = 0;
324 	int addr_type = 0;
325 	int len;
326 	int hlimit, tclass;
327 	int err = 0;
328 
329 	if ((u8 *)hdr < skb->head ||
330 	    (skb->network_header + sizeof(*hdr)) > skb->tail)
331 		return;
332 
333 	/*
334 	 *	Make sure we respect the rules
335 	 *	i.e. RFC 1885 2.4(e)
336 	 *	Rule (e.1) is enforced by not using icmpv6_send
337 	 *	in any code that processes icmp errors.
338 	 */
339 	addr_type = ipv6_addr_type(&hdr->daddr);
340 
341 	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0))
342 		saddr = &hdr->daddr;
343 
344 	/*
345 	 *	Dest addr check
346 	 */
347 
348 	if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) {
349 		if (type != ICMPV6_PKT_TOOBIG &&
350 		    !(type == ICMPV6_PARAMPROB &&
351 		      code == ICMPV6_UNK_OPTION &&
352 		      (opt_unrec(skb, info))))
353 			return;
354 
355 		saddr = NULL;
356 	}
357 
358 	addr_type = ipv6_addr_type(&hdr->saddr);
359 
360 	/*
361 	 *	Source addr check
362 	 */
363 
364 	if (addr_type & IPV6_ADDR_LINKLOCAL)
365 		iif = skb->dev->ifindex;
366 
367 	/*
368 	 *	Must not send error if the source does not uniquely
369 	 *	identify a single node (RFC2463 Section 2.4).
370 	 *	We check unspecified / multicast addresses here,
371 	 *	and anycast addresses will be checked later.
372 	 */
373 	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
374 		LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: addr_any/mcast source\n");
375 		return;
376 	}
377 
378 	/*
379 	 *	Never answer to a ICMP packet.
380 	 */
381 	if (is_ineligible(skb)) {
382 		LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: no reply to icmp error\n");
383 		return;
384 	}
385 
386 	mip6_addr_swap(skb);
387 
388 	memset(&fl, 0, sizeof(fl));
389 	fl.proto = IPPROTO_ICMPV6;
390 	ipv6_addr_copy(&fl.fl6_dst, &hdr->saddr);
391 	if (saddr)
392 		ipv6_addr_copy(&fl.fl6_src, saddr);
393 	fl.oif = iif;
394 	fl.fl_icmp_type = type;
395 	fl.fl_icmp_code = code;
396 	security_skb_classify_flow(skb, &fl);
397 
398 	sk = icmpv6_xmit_lock(net);
399 	if (sk == NULL)
400 		return;
401 	np = inet6_sk(sk);
402 
403 	if (!icmpv6_xrlim_allow(sk, type, &fl))
404 		goto out;
405 
406 	tmp_hdr.icmp6_type = type;
407 	tmp_hdr.icmp6_code = code;
408 	tmp_hdr.icmp6_cksum = 0;
409 	tmp_hdr.icmp6_pointer = htonl(info);
410 
411 	if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
412 		fl.oif = np->mcast_oif;
413 
414 	err = ip6_dst_lookup(sk, &dst, &fl);
415 	if (err)
416 		goto out;
417 
418 	/*
419 	 * We won't send icmp if the destination is known
420 	 * anycast.
421 	 */
422 	if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) {
423 		LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: acast source\n");
424 		goto out_dst_release;
425 	}
426 
427 	/* No need to clone since we're just using its address. */
428 	dst2 = dst;
429 
430 	err = xfrm_lookup(&dst, &fl, sk, 0);
431 	switch (err) {
432 	case 0:
433 		if (dst != dst2)
434 			goto route_done;
435 		break;
436 	case -EPERM:
437 		dst = NULL;
438 		break;
439 	default:
440 		goto out;
441 	}
442 
443 	if (xfrm_decode_session_reverse(skb, &fl2, AF_INET6))
444 		goto relookup_failed;
445 
446 	if (ip6_dst_lookup(sk, &dst2, &fl))
447 		goto relookup_failed;
448 
449 	err = xfrm_lookup(&dst2, &fl, sk, XFRM_LOOKUP_ICMP);
450 	switch (err) {
451 	case 0:
452 		dst_release(dst);
453 		dst = dst2;
454 		break;
455 	case -EPERM:
456 		goto out_dst_release;
457 	default:
458 relookup_failed:
459 		if (!dst)
460 			goto out;
461 		break;
462 	}
463 
464 route_done:
465 	if (ipv6_addr_is_multicast(&fl.fl6_dst))
466 		hlimit = np->mcast_hops;
467 	else
468 		hlimit = np->hop_limit;
469 	if (hlimit < 0)
470 		hlimit = ip6_dst_hoplimit(dst);
471 
472 	tclass = np->tclass;
473 	if (tclass < 0)
474 		tclass = 0;
475 
476 	msg.skb = skb;
477 	msg.offset = skb_network_offset(skb);
478 	msg.type = type;
479 
480 	len = skb->len - msg.offset;
481 	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr));
482 	if (len < 0) {
483 		LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n");
484 		goto out_dst_release;
485 	}
486 
487 	idev = in6_dev_get(skb->dev);
488 
489 	err = ip6_append_data(sk, icmpv6_getfrag, &msg,
490 			      len + sizeof(struct icmp6hdr),
491 			      sizeof(struct icmp6hdr),
492 			      hlimit, tclass, NULL, &fl, (struct rt6_info*)dst,
493 			      MSG_DONTWAIT);
494 	if (err) {
495 		ip6_flush_pending_frames(sk);
496 		goto out_put;
497 	}
498 	err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, len + sizeof(struct icmp6hdr));
499 
500 out_put:
501 	if (likely(idev != NULL))
502 		in6_dev_put(idev);
503 out_dst_release:
504 	dst_release(dst);
505 out:
506 	icmpv6_xmit_unlock(sk);
507 }
508 
509 EXPORT_SYMBOL(icmpv6_send);
510 
511 static void icmpv6_echo_reply(struct sk_buff *skb)
512 {
513 	struct net *net = dev_net(skb->dev);
514 	struct sock *sk;
515 	struct inet6_dev *idev;
516 	struct ipv6_pinfo *np;
517 	struct in6_addr *saddr = NULL;
518 	struct icmp6hdr *icmph = icmp6_hdr(skb);
519 	struct icmp6hdr tmp_hdr;
520 	struct flowi fl;
521 	struct icmpv6_msg msg;
522 	struct dst_entry *dst;
523 	int err = 0;
524 	int hlimit;
525 	int tclass;
526 
527 	saddr = &ipv6_hdr(skb)->daddr;
528 
529 	if (!ipv6_unicast_destination(skb))
530 		saddr = NULL;
531 
532 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
533 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
534 
535 	memset(&fl, 0, sizeof(fl));
536 	fl.proto = IPPROTO_ICMPV6;
537 	ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
538 	if (saddr)
539 		ipv6_addr_copy(&fl.fl6_src, saddr);
540 	fl.oif = skb->dev->ifindex;
541 	fl.fl_icmp_type = ICMPV6_ECHO_REPLY;
542 	security_skb_classify_flow(skb, &fl);
543 
544 	sk = icmpv6_xmit_lock(net);
545 	if (sk == NULL)
546 		return;
547 	np = inet6_sk(sk);
548 
549 	if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
550 		fl.oif = np->mcast_oif;
551 
552 	err = ip6_dst_lookup(sk, &dst, &fl);
553 	if (err)
554 		goto out;
555 	if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
556 		goto out;
557 
558 	if (ipv6_addr_is_multicast(&fl.fl6_dst))
559 		hlimit = np->mcast_hops;
560 	else
561 		hlimit = np->hop_limit;
562 	if (hlimit < 0)
563 		hlimit = ip6_dst_hoplimit(dst);
564 
565 	tclass = np->tclass;
566 	if (tclass < 0)
567 		tclass = 0;
568 
569 	idev = in6_dev_get(skb->dev);
570 
571 	msg.skb = skb;
572 	msg.offset = 0;
573 	msg.type = ICMPV6_ECHO_REPLY;
574 
575 	err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
576 				sizeof(struct icmp6hdr), hlimit, tclass, NULL, &fl,
577 				(struct rt6_info*)dst, MSG_DONTWAIT);
578 
579 	if (err) {
580 		ip6_flush_pending_frames(sk);
581 		goto out_put;
582 	}
583 	err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, skb->len + sizeof(struct icmp6hdr));
584 
585 out_put:
586 	if (likely(idev != NULL))
587 		in6_dev_put(idev);
588 	dst_release(dst);
589 out:
590 	icmpv6_xmit_unlock(sk);
591 }
592 
593 static void icmpv6_notify(struct sk_buff *skb, int type, int code, __be32 info)
594 {
595 	struct inet6_protocol *ipprot;
596 	int inner_offset;
597 	int hash;
598 	u8 nexthdr;
599 
600 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
601 		return;
602 
603 	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
604 	if (ipv6_ext_hdr(nexthdr)) {
605 		/* now skip over extension headers */
606 		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr);
607 		if (inner_offset<0)
608 			return;
609 	} else {
610 		inner_offset = sizeof(struct ipv6hdr);
611 	}
612 
613 	/* Checkin header including 8 bytes of inner protocol header. */
614 	if (!pskb_may_pull(skb, inner_offset+8))
615 		return;
616 
617 	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
618 	   Without this we will not able f.e. to make source routed
619 	   pmtu discovery.
620 	   Corresponding argument (opt) to notifiers is already added.
621 	   --ANK (980726)
622 	 */
623 
624 	hash = nexthdr & (MAX_INET_PROTOS - 1);
625 
626 	rcu_read_lock();
627 	ipprot = rcu_dereference(inet6_protos[hash]);
628 	if (ipprot && ipprot->err_handler)
629 		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
630 	rcu_read_unlock();
631 
632 	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
633 }
634 
635 /*
636  *	Handle icmp messages
637  */
638 
639 static int icmpv6_rcv(struct sk_buff *skb)
640 {
641 	struct net_device *dev = skb->dev;
642 	struct inet6_dev *idev = __in6_dev_get(dev);
643 	struct in6_addr *saddr, *daddr;
644 	struct ipv6hdr *orig_hdr;
645 	struct icmp6hdr *hdr;
646 	int type;
647 
648 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
649 		int nh;
650 
651 		if (!(skb->sp && skb->sp->xvec[skb->sp->len - 1]->props.flags &
652 				 XFRM_STATE_ICMP))
653 			goto drop_no_count;
654 
655 		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(*orig_hdr)))
656 			goto drop_no_count;
657 
658 		nh = skb_network_offset(skb);
659 		skb_set_network_header(skb, sizeof(*hdr));
660 
661 		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
662 			goto drop_no_count;
663 
664 		skb_set_network_header(skb, nh);
665 	}
666 
667 	ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INMSGS);
668 
669 	saddr = &ipv6_hdr(skb)->saddr;
670 	daddr = &ipv6_hdr(skb)->daddr;
671 
672 	/* Perform checksum. */
673 	switch (skb->ip_summed) {
674 	case CHECKSUM_COMPLETE:
675 		if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6,
676 				     skb->csum))
677 			break;
678 		/* fall through */
679 	case CHECKSUM_NONE:
680 		skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len,
681 					     IPPROTO_ICMPV6, 0));
682 		if (__skb_checksum_complete(skb)) {
683 			LIMIT_NETDEBUG(KERN_DEBUG "ICMPv6 checksum failed [" NIP6_FMT " > " NIP6_FMT "]\n",
684 				       NIP6(*saddr), NIP6(*daddr));
685 			goto discard_it;
686 		}
687 	}
688 
689 	if (!pskb_pull(skb, sizeof(*hdr)))
690 		goto discard_it;
691 
692 	hdr = icmp6_hdr(skb);
693 
694 	type = hdr->icmp6_type;
695 
696 	ICMP6MSGIN_INC_STATS_BH(dev_net(dev), idev, type);
697 
698 	switch (type) {
699 	case ICMPV6_ECHO_REQUEST:
700 		icmpv6_echo_reply(skb);
701 		break;
702 
703 	case ICMPV6_ECHO_REPLY:
704 		/* we couldn't care less */
705 		break;
706 
707 	case ICMPV6_PKT_TOOBIG:
708 		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
709 		   standard destination cache. Seems, only "advanced"
710 		   destination cache will allow to solve this problem
711 		   --ANK (980726)
712 		 */
713 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
714 			goto discard_it;
715 		hdr = icmp6_hdr(skb);
716 		orig_hdr = (struct ipv6hdr *) (hdr + 1);
717 		rt6_pmtu_discovery(&orig_hdr->daddr, &orig_hdr->saddr, dev,
718 				   ntohl(hdr->icmp6_mtu));
719 
720 		/*
721 		 *	Drop through to notify
722 		 */
723 
724 	case ICMPV6_DEST_UNREACH:
725 	case ICMPV6_TIME_EXCEED:
726 	case ICMPV6_PARAMPROB:
727 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
728 		break;
729 
730 	case NDISC_ROUTER_SOLICITATION:
731 	case NDISC_ROUTER_ADVERTISEMENT:
732 	case NDISC_NEIGHBOUR_SOLICITATION:
733 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
734 	case NDISC_REDIRECT:
735 		ndisc_rcv(skb);
736 		break;
737 
738 	case ICMPV6_MGM_QUERY:
739 		igmp6_event_query(skb);
740 		break;
741 
742 	case ICMPV6_MGM_REPORT:
743 		igmp6_event_report(skb);
744 		break;
745 
746 	case ICMPV6_MGM_REDUCTION:
747 	case ICMPV6_NI_QUERY:
748 	case ICMPV6_NI_REPLY:
749 	case ICMPV6_MLD2_REPORT:
750 	case ICMPV6_DHAAD_REQUEST:
751 	case ICMPV6_DHAAD_REPLY:
752 	case ICMPV6_MOBILE_PREFIX_SOL:
753 	case ICMPV6_MOBILE_PREFIX_ADV:
754 		break;
755 
756 	default:
757 		LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n");
758 
759 		/* informational */
760 		if (type & ICMPV6_INFOMSG_MASK)
761 			break;
762 
763 		/*
764 		 * error of unknown type.
765 		 * must pass to upper level
766 		 */
767 
768 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
769 	}
770 
771 	kfree_skb(skb);
772 	return 0;
773 
774 discard_it:
775 	ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INERRORS);
776 drop_no_count:
777 	kfree_skb(skb);
778 	return 0;
779 }
780 
781 void icmpv6_flow_init(struct sock *sk, struct flowi *fl,
782 		      u8 type,
783 		      const struct in6_addr *saddr,
784 		      const struct in6_addr *daddr,
785 		      int oif)
786 {
787 	memset(fl, 0, sizeof(*fl));
788 	ipv6_addr_copy(&fl->fl6_src, saddr);
789 	ipv6_addr_copy(&fl->fl6_dst, daddr);
790 	fl->proto	 	= IPPROTO_ICMPV6;
791 	fl->fl_icmp_type	= type;
792 	fl->fl_icmp_code	= 0;
793 	fl->oif			= oif;
794 	security_sk_classify_flow(sk, fl);
795 }
796 
797 /*
798  * Special lock-class for __icmpv6_sk:
799  */
800 static struct lock_class_key icmpv6_socket_sk_dst_lock_key;
801 
802 static int __net_init icmpv6_sk_init(struct net *net)
803 {
804 	struct sock *sk;
805 	int err, i, j;
806 
807 	net->ipv6.icmp_sk =
808 		kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
809 	if (net->ipv6.icmp_sk == NULL)
810 		return -ENOMEM;
811 
812 	for_each_possible_cpu(i) {
813 		err = inet_ctl_sock_create(&sk, PF_INET6,
814 					   SOCK_RAW, IPPROTO_ICMPV6, net);
815 		if (err < 0) {
816 			printk(KERN_ERR
817 			       "Failed to initialize the ICMP6 control socket "
818 			       "(err %d).\n",
819 			       err);
820 			goto fail;
821 		}
822 
823 		net->ipv6.icmp_sk[i] = sk;
824 
825 		/*
826 		 * Split off their lock-class, because sk->sk_dst_lock
827 		 * gets used from softirqs, which is safe for
828 		 * __icmpv6_sk (because those never get directly used
829 		 * via userspace syscalls), but unsafe for normal sockets.
830 		 */
831 		lockdep_set_class(&sk->sk_dst_lock,
832 				  &icmpv6_socket_sk_dst_lock_key);
833 
834 		/* Enough space for 2 64K ICMP packets, including
835 		 * sk_buff struct overhead.
836 		 */
837 		sk->sk_sndbuf =
838 			(2 * ((64 * 1024) + sizeof(struct sk_buff)));
839 	}
840 	return 0;
841 
842  fail:
843 	for (j = 0; j < i; j++)
844 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
845 	kfree(net->ipv6.icmp_sk);
846 	return err;
847 }
848 
849 static void __net_exit icmpv6_sk_exit(struct net *net)
850 {
851 	int i;
852 
853 	for_each_possible_cpu(i) {
854 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
855 	}
856 	kfree(net->ipv6.icmp_sk);
857 }
858 
859 static struct pernet_operations icmpv6_sk_ops = {
860        .init = icmpv6_sk_init,
861        .exit = icmpv6_sk_exit,
862 };
863 
864 int __init icmpv6_init(void)
865 {
866 	int err;
867 
868 	err = register_pernet_subsys(&icmpv6_sk_ops);
869 	if (err < 0)
870 		return err;
871 
872 	err = -EAGAIN;
873 	if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
874 		goto fail;
875 	return 0;
876 
877 fail:
878 	printk(KERN_ERR "Failed to register ICMP6 protocol\n");
879 	unregister_pernet_subsys(&icmpv6_sk_ops);
880 	return err;
881 }
882 
883 void icmpv6_cleanup(void)
884 {
885 	unregister_pernet_subsys(&icmpv6_sk_ops);
886 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
887 }
888 
889 
890 static const struct icmp6_err {
891 	int err;
892 	int fatal;
893 } tab_unreach[] = {
894 	{	/* NOROUTE */
895 		.err	= ENETUNREACH,
896 		.fatal	= 0,
897 	},
898 	{	/* ADM_PROHIBITED */
899 		.err	= EACCES,
900 		.fatal	= 1,
901 	},
902 	{	/* Was NOT_NEIGHBOUR, now reserved */
903 		.err	= EHOSTUNREACH,
904 		.fatal	= 0,
905 	},
906 	{	/* ADDR_UNREACH	*/
907 		.err	= EHOSTUNREACH,
908 		.fatal	= 0,
909 	},
910 	{	/* PORT_UNREACH	*/
911 		.err	= ECONNREFUSED,
912 		.fatal	= 1,
913 	},
914 };
915 
916 int icmpv6_err_convert(int type, int code, int *err)
917 {
918 	int fatal = 0;
919 
920 	*err = EPROTO;
921 
922 	switch (type) {
923 	case ICMPV6_DEST_UNREACH:
924 		fatal = 1;
925 		if (code <= ICMPV6_PORT_UNREACH) {
926 			*err  = tab_unreach[code].err;
927 			fatal = tab_unreach[code].fatal;
928 		}
929 		break;
930 
931 	case ICMPV6_PKT_TOOBIG:
932 		*err = EMSGSIZE;
933 		break;
934 
935 	case ICMPV6_PARAMPROB:
936 		*err = EPROTO;
937 		fatal = 1;
938 		break;
939 
940 	case ICMPV6_TIME_EXCEED:
941 		*err = EHOSTUNREACH;
942 		break;
943 	}
944 
945 	return fatal;
946 }
947 
948 EXPORT_SYMBOL(icmpv6_err_convert);
949 
950 #ifdef CONFIG_SYSCTL
951 ctl_table ipv6_icmp_table_template[] = {
952 	{
953 		.ctl_name	= NET_IPV6_ICMP_RATELIMIT,
954 		.procname	= "ratelimit",
955 		.data		= &init_net.ipv6.sysctl.icmpv6_time,
956 		.maxlen		= sizeof(int),
957 		.mode		= 0644,
958 		.proc_handler	= &proc_dointvec_ms_jiffies,
959 		.strategy	= &sysctl_ms_jiffies
960 	},
961 	{ .ctl_name = 0 },
962 };
963 
964 struct ctl_table *ipv6_icmp_sysctl_init(struct net *net)
965 {
966 	struct ctl_table *table;
967 
968 	table = kmemdup(ipv6_icmp_table_template,
969 			sizeof(ipv6_icmp_table_template),
970 			GFP_KERNEL);
971 
972 	if (table)
973 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
974 
975 	return table;
976 }
977 #endif
978 
979