xref: /openbmc/linux/net/ipv6/icmp.c (revision cd2a9e62c8a3c5cae7691982667d79a0edc65283)
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 <asm/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);
96 	else if (type == NDISC_REDIRECT)
97 		ip6_redirect(skb, net, skb->dev->ifindex, 0);
98 
99 	if (!(type & ICMPV6_INFOMSG_MASK))
100 		if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST)
101 			ping_err(skb, offset, info);
102 }
103 
104 static int icmpv6_rcv(struct sk_buff *skb);
105 
106 static const struct inet6_protocol icmpv6_protocol = {
107 	.handler	=	icmpv6_rcv,
108 	.err_handler	=	icmpv6_err,
109 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
110 };
111 
112 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
113 {
114 	struct sock *sk;
115 
116 	local_bh_disable();
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 		local_bh_enable();
125 		return NULL;
126 	}
127 	return sk;
128 }
129 
130 static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
131 {
132 	spin_unlock_bh(&sk->sk_lock.slock);
133 }
134 
135 /*
136  * Figure out, may we reply to this packet with icmp error.
137  *
138  * We do not reply, if:
139  *	- it was icmp error message.
140  *	- it is truncated, so that it is known, that protocol is ICMPV6
141  *	  (i.e. in the middle of some exthdr)
142  *
143  *	--ANK (980726)
144  */
145 
146 static bool is_ineligible(const struct sk_buff *skb)
147 {
148 	int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
149 	int len = skb->len - ptr;
150 	__u8 nexthdr = ipv6_hdr(skb)->nexthdr;
151 	__be16 frag_off;
152 
153 	if (len < 0)
154 		return true;
155 
156 	ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off);
157 	if (ptr < 0)
158 		return false;
159 	if (nexthdr == IPPROTO_ICMPV6) {
160 		u8 _type, *tp;
161 		tp = skb_header_pointer(skb,
162 			ptr+offsetof(struct icmp6hdr, icmp6_type),
163 			sizeof(_type), &_type);
164 		if (!tp || !(*tp & ICMPV6_INFOMSG_MASK))
165 			return true;
166 	}
167 	return false;
168 }
169 
170 /*
171  * Check the ICMP output rate limit
172  */
173 static bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
174 			       struct flowi6 *fl6)
175 {
176 	struct net *net = sock_net(sk);
177 	struct dst_entry *dst;
178 	bool res = false;
179 
180 	/* Informational messages are not limited. */
181 	if (type & ICMPV6_INFOMSG_MASK)
182 		return true;
183 
184 	/* Do not limit pmtu discovery, it would break it. */
185 	if (type == ICMPV6_PKT_TOOBIG)
186 		return true;
187 
188 	/*
189 	 * Look up the output route.
190 	 * XXX: perhaps the expire for routing entries cloned by
191 	 * this lookup should be more aggressive (not longer than timeout).
192 	 */
193 	dst = ip6_route_output(net, sk, fl6);
194 	if (dst->error) {
195 		IP6_INC_STATS(net, ip6_dst_idev(dst),
196 			      IPSTATS_MIB_OUTNOROUTES);
197 	} else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
198 		res = true;
199 	} else {
200 		struct rt6_info *rt = (struct rt6_info *)dst;
201 		int tmo = net->ipv6.sysctl.icmpv6_time;
202 
203 		/* Give more bandwidth to wider prefixes. */
204 		if (rt->rt6i_dst.plen < 128)
205 			tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
206 
207 		if (icmp_global_allow()) {
208 			struct inet_peer *peer;
209 
210 			peer = inet_getpeer_v6(net->ipv6.peers,
211 					       &fl6->daddr, 1);
212 			res = inet_peer_xrlim_allow(peer, tmo);
213 			if (peer)
214 				inet_putpeer(peer);
215 		}
216 	}
217 	dst_release(dst);
218 	return res;
219 }
220 
221 /*
222  *	an inline helper for the "simple" if statement below
223  *	checks if parameter problem report is caused by an
224  *	unrecognized IPv6 option that has the Option Type
225  *	highest-order two bits set to 10
226  */
227 
228 static bool opt_unrec(struct sk_buff *skb, __u32 offset)
229 {
230 	u8 _optval, *op;
231 
232 	offset += skb_network_offset(skb);
233 	op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
234 	if (!op)
235 		return true;
236 	return (*op & 0xC0) == 0x80;
237 }
238 
239 int icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
240 			       struct icmp6hdr *thdr, int len)
241 {
242 	struct sk_buff *skb;
243 	struct icmp6hdr *icmp6h;
244 	int err = 0;
245 
246 	skb = skb_peek(&sk->sk_write_queue);
247 	if (!skb)
248 		goto out;
249 
250 	icmp6h = icmp6_hdr(skb);
251 	memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
252 	icmp6h->icmp6_cksum = 0;
253 
254 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
255 		skb->csum = csum_partial(icmp6h,
256 					sizeof(struct icmp6hdr), skb->csum);
257 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
258 						      &fl6->daddr,
259 						      len, fl6->flowi6_proto,
260 						      skb->csum);
261 	} else {
262 		__wsum tmp_csum = 0;
263 
264 		skb_queue_walk(&sk->sk_write_queue, skb) {
265 			tmp_csum = csum_add(tmp_csum, skb->csum);
266 		}
267 
268 		tmp_csum = csum_partial(icmp6h,
269 					sizeof(struct icmp6hdr), tmp_csum);
270 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
271 						      &fl6->daddr,
272 						      len, fl6->flowi6_proto,
273 						      tmp_csum);
274 	}
275 	ip6_push_pending_frames(sk);
276 out:
277 	return err;
278 }
279 
280 struct icmpv6_msg {
281 	struct sk_buff	*skb;
282 	int		offset;
283 	uint8_t		type;
284 };
285 
286 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
287 {
288 	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
289 	struct sk_buff *org_skb = msg->skb;
290 	__wsum csum = 0;
291 
292 	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
293 				      to, len, csum);
294 	skb->csum = csum_block_add(skb->csum, csum, odd);
295 	if (!(msg->type & ICMPV6_INFOMSG_MASK))
296 		nf_ct_attach(skb, org_skb);
297 	return 0;
298 }
299 
300 #if IS_ENABLED(CONFIG_IPV6_MIP6)
301 static void mip6_addr_swap(struct sk_buff *skb)
302 {
303 	struct ipv6hdr *iph = ipv6_hdr(skb);
304 	struct inet6_skb_parm *opt = IP6CB(skb);
305 	struct ipv6_destopt_hao *hao;
306 	struct in6_addr tmp;
307 	int off;
308 
309 	if (opt->dsthao) {
310 		off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
311 		if (likely(off >= 0)) {
312 			hao = (struct ipv6_destopt_hao *)
313 					(skb_network_header(skb) + off);
314 			tmp = iph->saddr;
315 			iph->saddr = hao->addr;
316 			hao->addr = tmp;
317 		}
318 	}
319 }
320 #else
321 static inline void mip6_addr_swap(struct sk_buff *skb) {}
322 #endif
323 
324 static struct dst_entry *icmpv6_route_lookup(struct net *net,
325 					     struct sk_buff *skb,
326 					     struct sock *sk,
327 					     struct flowi6 *fl6)
328 {
329 	struct dst_entry *dst, *dst2;
330 	struct flowi6 fl2;
331 	int err;
332 
333 	err = ip6_dst_lookup(net, sk, &dst, fl6);
334 	if (err)
335 		return ERR_PTR(err);
336 
337 	/*
338 	 * We won't send icmp if the destination is known
339 	 * anycast.
340 	 */
341 	if (ipv6_anycast_destination(dst, &fl6->daddr)) {
342 		net_dbg_ratelimited("icmp6_send: acast source\n");
343 		dst_release(dst);
344 		return ERR_PTR(-EINVAL);
345 	}
346 
347 	/* No need to clone since we're just using its address. */
348 	dst2 = dst;
349 
350 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
351 	if (!IS_ERR(dst)) {
352 		if (dst != dst2)
353 			return dst;
354 	} else {
355 		if (PTR_ERR(dst) == -EPERM)
356 			dst = NULL;
357 		else
358 			return dst;
359 	}
360 
361 	err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
362 	if (err)
363 		goto relookup_failed;
364 
365 	err = ip6_dst_lookup(net, sk, &dst2, &fl2);
366 	if (err)
367 		goto relookup_failed;
368 
369 	dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
370 	if (!IS_ERR(dst2)) {
371 		dst_release(dst);
372 		dst = dst2;
373 	} else {
374 		err = PTR_ERR(dst2);
375 		if (err == -EPERM) {
376 			dst_release(dst);
377 			return dst2;
378 		} else
379 			goto relookup_failed;
380 	}
381 
382 relookup_failed:
383 	if (dst)
384 		return dst;
385 	return ERR_PTR(err);
386 }
387 
388 /*
389  *	Send an ICMP message in response to a packet in error
390  */
391 static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info)
392 {
393 	struct net *net = dev_net(skb->dev);
394 	struct inet6_dev *idev = NULL;
395 	struct ipv6hdr *hdr = ipv6_hdr(skb);
396 	struct sock *sk;
397 	struct ipv6_pinfo *np;
398 	const struct in6_addr *saddr = NULL;
399 	struct dst_entry *dst;
400 	struct icmp6hdr tmp_hdr;
401 	struct flowi6 fl6;
402 	struct icmpv6_msg msg;
403 	struct sockcm_cookie sockc_unused = {0};
404 	struct ipcm6_cookie ipc6;
405 	int iif = 0;
406 	int addr_type = 0;
407 	int len;
408 	int err = 0;
409 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
410 
411 	if ((u8 *)hdr < skb->head ||
412 	    (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb))
413 		return;
414 
415 	/*
416 	 *	Make sure we respect the rules
417 	 *	i.e. RFC 1885 2.4(e)
418 	 *	Rule (e.1) is enforced by not using icmp6_send
419 	 *	in any code that processes icmp errors.
420 	 */
421 	addr_type = ipv6_addr_type(&hdr->daddr);
422 
423 	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) ||
424 	    ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr))
425 		saddr = &hdr->daddr;
426 
427 	/*
428 	 *	Dest addr check
429 	 */
430 
431 	if (addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST) {
432 		if (type != ICMPV6_PKT_TOOBIG &&
433 		    !(type == ICMPV6_PARAMPROB &&
434 		      code == ICMPV6_UNK_OPTION &&
435 		      (opt_unrec(skb, info))))
436 			return;
437 
438 		saddr = NULL;
439 	}
440 
441 	addr_type = ipv6_addr_type(&hdr->saddr);
442 
443 	/*
444 	 *	Source addr check
445 	 */
446 
447 	if (__ipv6_addr_needs_scope_id(addr_type))
448 		iif = skb->dev->ifindex;
449 	else
450 		iif = l3mdev_master_ifindex(skb->dev);
451 
452 	/*
453 	 *	Must not send error if the source does not uniquely
454 	 *	identify a single node (RFC2463 Section 2.4).
455 	 *	We check unspecified / multicast addresses here,
456 	 *	and anycast addresses will be checked later.
457 	 */
458 	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
459 		net_dbg_ratelimited("icmp6_send: addr_any/mcast source [%pI6c > %pI6c]\n",
460 				    &hdr->saddr, &hdr->daddr);
461 		return;
462 	}
463 
464 	/*
465 	 *	Never answer to a ICMP packet.
466 	 */
467 	if (is_ineligible(skb)) {
468 		net_dbg_ratelimited("icmp6_send: no reply to icmp error [%pI6c > %pI6c]\n",
469 				    &hdr->saddr, &hdr->daddr);
470 		return;
471 	}
472 
473 	mip6_addr_swap(skb);
474 
475 	memset(&fl6, 0, sizeof(fl6));
476 	fl6.flowi6_proto = IPPROTO_ICMPV6;
477 	fl6.daddr = hdr->saddr;
478 	if (saddr)
479 		fl6.saddr = *saddr;
480 	fl6.flowi6_mark = mark;
481 	fl6.flowi6_oif = iif;
482 	fl6.fl6_icmp_type = type;
483 	fl6.fl6_icmp_code = code;
484 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
485 
486 	sk = icmpv6_xmit_lock(net);
487 	if (!sk)
488 		return;
489 	sk->sk_mark = mark;
490 	np = inet6_sk(sk);
491 
492 	if (!icmpv6_xrlim_allow(sk, type, &fl6))
493 		goto out;
494 
495 	tmp_hdr.icmp6_type = type;
496 	tmp_hdr.icmp6_code = code;
497 	tmp_hdr.icmp6_cksum = 0;
498 	tmp_hdr.icmp6_pointer = htonl(info);
499 
500 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
501 		fl6.flowi6_oif = np->mcast_oif;
502 	else if (!fl6.flowi6_oif)
503 		fl6.flowi6_oif = np->ucast_oif;
504 
505 	ipc6.tclass = np->tclass;
506 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
507 
508 	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
509 	if (IS_ERR(dst))
510 		goto out;
511 
512 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
513 	ipc6.dontfrag = np->dontfrag;
514 	ipc6.opt = NULL;
515 
516 	msg.skb = skb;
517 	msg.offset = skb_network_offset(skb);
518 	msg.type = type;
519 
520 	len = skb->len - msg.offset;
521 	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(struct icmp6hdr));
522 	if (len < 0) {
523 		net_dbg_ratelimited("icmp: len problem [%pI6c > %pI6c]\n",
524 				    &hdr->saddr, &hdr->daddr);
525 		goto out_dst_release;
526 	}
527 
528 	rcu_read_lock();
529 	idev = __in6_dev_get(skb->dev);
530 
531 	err = ip6_append_data(sk, icmpv6_getfrag, &msg,
532 			      len + sizeof(struct icmp6hdr),
533 			      sizeof(struct icmp6hdr),
534 			      &ipc6, &fl6, (struct rt6_info *)dst,
535 			      MSG_DONTWAIT, &sockc_unused);
536 	if (err) {
537 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
538 		ip6_flush_pending_frames(sk);
539 	} else {
540 		err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
541 						 len + sizeof(struct icmp6hdr));
542 	}
543 	rcu_read_unlock();
544 out_dst_release:
545 	dst_release(dst);
546 out:
547 	icmpv6_xmit_unlock(sk);
548 }
549 
550 /* Slightly more convenient version of icmp6_send.
551  */
552 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
553 {
554 	icmp6_send(skb, ICMPV6_PARAMPROB, code, pos);
555 	kfree_skb(skb);
556 }
557 
558 static void icmpv6_echo_reply(struct sk_buff *skb)
559 {
560 	struct net *net = dev_net(skb->dev);
561 	struct sock *sk;
562 	struct inet6_dev *idev;
563 	struct ipv6_pinfo *np;
564 	const struct in6_addr *saddr = NULL;
565 	struct icmp6hdr *icmph = icmp6_hdr(skb);
566 	struct icmp6hdr tmp_hdr;
567 	struct flowi6 fl6;
568 	struct icmpv6_msg msg;
569 	struct dst_entry *dst;
570 	struct ipcm6_cookie ipc6;
571 	int err = 0;
572 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
573 	struct sockcm_cookie sockc_unused = {0};
574 
575 	saddr = &ipv6_hdr(skb)->daddr;
576 
577 	if (!ipv6_unicast_destination(skb) &&
578 	    !(net->ipv6.sysctl.anycast_src_echo_reply &&
579 	      ipv6_anycast_destination(skb_dst(skb), saddr)))
580 		saddr = NULL;
581 
582 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
583 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
584 
585 	memset(&fl6, 0, sizeof(fl6));
586 	fl6.flowi6_proto = IPPROTO_ICMPV6;
587 	fl6.daddr = ipv6_hdr(skb)->saddr;
588 	if (saddr)
589 		fl6.saddr = *saddr;
590 	fl6.flowi6_oif = l3mdev_fib_oif(skb->dev);
591 	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
592 	fl6.flowi6_mark = mark;
593 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
594 
595 	sk = icmpv6_xmit_lock(net);
596 	if (!sk)
597 		return;
598 	sk->sk_mark = mark;
599 	np = inet6_sk(sk);
600 
601 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
602 		fl6.flowi6_oif = np->mcast_oif;
603 	else if (!fl6.flowi6_oif)
604 		fl6.flowi6_oif = np->ucast_oif;
605 
606 	err = ip6_dst_lookup(net, sk, &dst, &fl6);
607 	if (err)
608 		goto out;
609 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
610 	if (IS_ERR(dst))
611 		goto out;
612 
613 	idev = __in6_dev_get(skb->dev);
614 
615 	msg.skb = skb;
616 	msg.offset = 0;
617 	msg.type = ICMPV6_ECHO_REPLY;
618 
619 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
620 	ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb));
621 	ipc6.dontfrag = np->dontfrag;
622 	ipc6.opt = NULL;
623 
624 	err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
625 				sizeof(struct icmp6hdr), &ipc6, &fl6,
626 				(struct rt6_info *)dst, MSG_DONTWAIT,
627 				&sockc_unused);
628 
629 	if (err) {
630 		__ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
631 		ip6_flush_pending_frames(sk);
632 	} else {
633 		err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
634 						 skb->len + sizeof(struct icmp6hdr));
635 	}
636 	dst_release(dst);
637 out:
638 	icmpv6_xmit_unlock(sk);
639 }
640 
641 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
642 {
643 	const struct inet6_protocol *ipprot;
644 	int inner_offset;
645 	__be16 frag_off;
646 	u8 nexthdr;
647 	struct net *net = dev_net(skb->dev);
648 
649 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
650 		goto out;
651 
652 	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
653 	if (ipv6_ext_hdr(nexthdr)) {
654 		/* now skip over extension headers */
655 		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
656 						&nexthdr, &frag_off);
657 		if (inner_offset < 0)
658 			goto out;
659 	} else {
660 		inner_offset = sizeof(struct ipv6hdr);
661 	}
662 
663 	/* Checkin header including 8 bytes of inner protocol header. */
664 	if (!pskb_may_pull(skb, inner_offset+8))
665 		goto out;
666 
667 	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
668 	   Without this we will not able f.e. to make source routed
669 	   pmtu discovery.
670 	   Corresponding argument (opt) to notifiers is already added.
671 	   --ANK (980726)
672 	 */
673 
674 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
675 	if (ipprot && ipprot->err_handler)
676 		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
677 
678 	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
679 	return;
680 
681 out:
682 	__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
683 }
684 
685 /*
686  *	Handle icmp messages
687  */
688 
689 static int icmpv6_rcv(struct sk_buff *skb)
690 {
691 	struct net_device *dev = skb->dev;
692 	struct inet6_dev *idev = __in6_dev_get(dev);
693 	const struct in6_addr *saddr, *daddr;
694 	struct icmp6hdr *hdr;
695 	u8 type;
696 	bool success = false;
697 
698 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
699 		struct sec_path *sp = skb_sec_path(skb);
700 		int nh;
701 
702 		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
703 				 XFRM_STATE_ICMP))
704 			goto drop_no_count;
705 
706 		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
707 			goto drop_no_count;
708 
709 		nh = skb_network_offset(skb);
710 		skb_set_network_header(skb, sizeof(*hdr));
711 
712 		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
713 			goto drop_no_count;
714 
715 		skb_set_network_header(skb, nh);
716 	}
717 
718 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS);
719 
720 	saddr = &ipv6_hdr(skb)->saddr;
721 	daddr = &ipv6_hdr(skb)->daddr;
722 
723 	if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) {
724 		net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n",
725 				    saddr, daddr);
726 		goto csum_error;
727 	}
728 
729 	if (!pskb_pull(skb, sizeof(*hdr)))
730 		goto discard_it;
731 
732 	hdr = icmp6_hdr(skb);
733 
734 	type = hdr->icmp6_type;
735 
736 	ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type);
737 
738 	switch (type) {
739 	case ICMPV6_ECHO_REQUEST:
740 		icmpv6_echo_reply(skb);
741 		break;
742 
743 	case ICMPV6_ECHO_REPLY:
744 		success = ping_rcv(skb);
745 		break;
746 
747 	case ICMPV6_PKT_TOOBIG:
748 		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
749 		   standard destination cache. Seems, only "advanced"
750 		   destination cache will allow to solve this problem
751 		   --ANK (980726)
752 		 */
753 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
754 			goto discard_it;
755 		hdr = icmp6_hdr(skb);
756 
757 		/*
758 		 *	Drop through to notify
759 		 */
760 
761 	case ICMPV6_DEST_UNREACH:
762 	case ICMPV6_TIME_EXCEED:
763 	case ICMPV6_PARAMPROB:
764 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
765 		break;
766 
767 	case NDISC_ROUTER_SOLICITATION:
768 	case NDISC_ROUTER_ADVERTISEMENT:
769 	case NDISC_NEIGHBOUR_SOLICITATION:
770 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
771 	case NDISC_REDIRECT:
772 		ndisc_rcv(skb);
773 		break;
774 
775 	case ICMPV6_MGM_QUERY:
776 		igmp6_event_query(skb);
777 		break;
778 
779 	case ICMPV6_MGM_REPORT:
780 		igmp6_event_report(skb);
781 		break;
782 
783 	case ICMPV6_MGM_REDUCTION:
784 	case ICMPV6_NI_QUERY:
785 	case ICMPV6_NI_REPLY:
786 	case ICMPV6_MLD2_REPORT:
787 	case ICMPV6_DHAAD_REQUEST:
788 	case ICMPV6_DHAAD_REPLY:
789 	case ICMPV6_MOBILE_PREFIX_SOL:
790 	case ICMPV6_MOBILE_PREFIX_ADV:
791 		break;
792 
793 	default:
794 		/* informational */
795 		if (type & ICMPV6_INFOMSG_MASK)
796 			break;
797 
798 		net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n",
799 				    saddr, daddr);
800 
801 		/*
802 		 * error of unknown type.
803 		 * must pass to upper level
804 		 */
805 
806 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
807 	}
808 
809 	/* until the v6 path can be better sorted assume failure and
810 	 * preserve the status quo behaviour for the rest of the paths to here
811 	 */
812 	if (success)
813 		consume_skb(skb);
814 	else
815 		kfree_skb(skb);
816 
817 	return 0;
818 
819 csum_error:
820 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
821 discard_it:
822 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS);
823 drop_no_count:
824 	kfree_skb(skb);
825 	return 0;
826 }
827 
828 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
829 		      u8 type,
830 		      const struct in6_addr *saddr,
831 		      const struct in6_addr *daddr,
832 		      int oif)
833 {
834 	memset(fl6, 0, sizeof(*fl6));
835 	fl6->saddr = *saddr;
836 	fl6->daddr = *daddr;
837 	fl6->flowi6_proto	= IPPROTO_ICMPV6;
838 	fl6->fl6_icmp_type	= type;
839 	fl6->fl6_icmp_code	= 0;
840 	fl6->flowi6_oif		= oif;
841 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
842 }
843 
844 static int __net_init icmpv6_sk_init(struct net *net)
845 {
846 	struct sock *sk;
847 	int err, i, j;
848 
849 	net->ipv6.icmp_sk =
850 		kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
851 	if (!net->ipv6.icmp_sk)
852 		return -ENOMEM;
853 
854 	for_each_possible_cpu(i) {
855 		err = inet_ctl_sock_create(&sk, PF_INET6,
856 					   SOCK_RAW, IPPROTO_ICMPV6, net);
857 		if (err < 0) {
858 			pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
859 			       err);
860 			goto fail;
861 		}
862 
863 		net->ipv6.icmp_sk[i] = sk;
864 
865 		/* Enough space for 2 64K ICMP packets, including
866 		 * sk_buff struct overhead.
867 		 */
868 		sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
869 	}
870 	return 0;
871 
872  fail:
873 	for (j = 0; j < i; j++)
874 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
875 	kfree(net->ipv6.icmp_sk);
876 	return err;
877 }
878 
879 static void __net_exit icmpv6_sk_exit(struct net *net)
880 {
881 	int i;
882 
883 	for_each_possible_cpu(i) {
884 		inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
885 	}
886 	kfree(net->ipv6.icmp_sk);
887 }
888 
889 static struct pernet_operations icmpv6_sk_ops = {
890 	.init = icmpv6_sk_init,
891 	.exit = icmpv6_sk_exit,
892 };
893 
894 int __init icmpv6_init(void)
895 {
896 	int err;
897 
898 	err = register_pernet_subsys(&icmpv6_sk_ops);
899 	if (err < 0)
900 		return err;
901 
902 	err = -EAGAIN;
903 	if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
904 		goto fail;
905 
906 	err = inet6_register_icmp_sender(icmp6_send);
907 	if (err)
908 		goto sender_reg_err;
909 	return 0;
910 
911 sender_reg_err:
912 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
913 fail:
914 	pr_err("Failed to register ICMP6 protocol\n");
915 	unregister_pernet_subsys(&icmpv6_sk_ops);
916 	return err;
917 }
918 
919 void icmpv6_cleanup(void)
920 {
921 	inet6_unregister_icmp_sender(icmp6_send);
922 	unregister_pernet_subsys(&icmpv6_sk_ops);
923 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
924 }
925 
926 
927 static const struct icmp6_err {
928 	int err;
929 	int fatal;
930 } tab_unreach[] = {
931 	{	/* NOROUTE */
932 		.err	= ENETUNREACH,
933 		.fatal	= 0,
934 	},
935 	{	/* ADM_PROHIBITED */
936 		.err	= EACCES,
937 		.fatal	= 1,
938 	},
939 	{	/* Was NOT_NEIGHBOUR, now reserved */
940 		.err	= EHOSTUNREACH,
941 		.fatal	= 0,
942 	},
943 	{	/* ADDR_UNREACH	*/
944 		.err	= EHOSTUNREACH,
945 		.fatal	= 0,
946 	},
947 	{	/* PORT_UNREACH	*/
948 		.err	= ECONNREFUSED,
949 		.fatal	= 1,
950 	},
951 	{	/* POLICY_FAIL */
952 		.err	= EACCES,
953 		.fatal	= 1,
954 	},
955 	{	/* REJECT_ROUTE	*/
956 		.err	= EACCES,
957 		.fatal	= 1,
958 	},
959 };
960 
961 int icmpv6_err_convert(u8 type, u8 code, int *err)
962 {
963 	int fatal = 0;
964 
965 	*err = EPROTO;
966 
967 	switch (type) {
968 	case ICMPV6_DEST_UNREACH:
969 		fatal = 1;
970 		if (code < ARRAY_SIZE(tab_unreach)) {
971 			*err  = tab_unreach[code].err;
972 			fatal = tab_unreach[code].fatal;
973 		}
974 		break;
975 
976 	case ICMPV6_PKT_TOOBIG:
977 		*err = EMSGSIZE;
978 		break;
979 
980 	case ICMPV6_PARAMPROB:
981 		*err = EPROTO;
982 		fatal = 1;
983 		break;
984 
985 	case ICMPV6_TIME_EXCEED:
986 		*err = EHOSTUNREACH;
987 		break;
988 	}
989 
990 	return fatal;
991 }
992 EXPORT_SYMBOL(icmpv6_err_convert);
993 
994 #ifdef CONFIG_SYSCTL
995 static struct ctl_table ipv6_icmp_table_template[] = {
996 	{
997 		.procname	= "ratelimit",
998 		.data		= &init_net.ipv6.sysctl.icmpv6_time,
999 		.maxlen		= sizeof(int),
1000 		.mode		= 0644,
1001 		.proc_handler	= proc_dointvec_ms_jiffies,
1002 	},
1003 	{ },
1004 };
1005 
1006 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1007 {
1008 	struct ctl_table *table;
1009 
1010 	table = kmemdup(ipv6_icmp_table_template,
1011 			sizeof(ipv6_icmp_table_template),
1012 			GFP_KERNEL);
1013 
1014 	if (table)
1015 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
1016 
1017 	return table;
1018 }
1019 #endif
1020