• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 
71 #include <asm/uaccess.h>
72 
73 /*
74  *	The ICMP socket(s). This is the most convenient way to flow control
75  *	our ICMP output as well as maintain a clean interface throughout
76  *	all layers. All Socketless IP sends will soon be gone.
77  *
78  *	On SMP we have one ICMP socket per-cpu.
79  */
icmpv6_sk(struct net * net)80 static inline struct sock *icmpv6_sk(struct net *net)
81 {
82 	return net->ipv6.icmp_sk[smp_processor_id()];
83 }
84 
icmpv6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)85 static void icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
86 		       u8 type, u8 code, int offset, __be32 info)
87 {
88 	/* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */
89 	struct icmp6hdr *icmp6 = (struct icmp6hdr *) (skb->data + offset);
90 	struct net *net = dev_net(skb->dev);
91 
92 	if (type == ICMPV6_PKT_TOOBIG)
93 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
94 	else if (type == NDISC_REDIRECT)
95 		ip6_redirect(skb, net, 0, 0, sock_net_uid(net, NULL));
96 
97 	if (!(type & ICMPV6_INFOMSG_MASK))
98 		if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST)
99 			ping_err(skb, offset, ntohl(info));
100 }
101 
102 static int icmpv6_rcv(struct sk_buff *skb);
103 
104 static const struct inet6_protocol icmpv6_protocol = {
105 	.handler	=	icmpv6_rcv,
106 	.err_handler	=	icmpv6_err,
107 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
108 };
109 
icmpv6_xmit_lock(struct net * net)110 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
111 {
112 	struct sock *sk;
113 
114 	local_bh_disable();
115 
116 	sk = icmpv6_sk(net);
117 	if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
118 		/* This can happen if the output path (f.e. SIT or
119 		 * ip6ip6 tunnel) signals dst_link_failure() for an
120 		 * outgoing ICMP6 packet.
121 		 */
122 		local_bh_enable();
123 		return NULL;
124 	}
125 	return sk;
126 }
127 
icmpv6_xmit_unlock(struct sock * sk)128 static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
129 {
130 	spin_unlock_bh(&sk->sk_lock.slock);
131 }
132 
133 /*
134  * Figure out, may we reply to this packet with icmp error.
135  *
136  * We do not reply, if:
137  *	- it was icmp error message.
138  *	- it is truncated, so that it is known, that protocol is ICMPV6
139  *	  (i.e. in the middle of some exthdr)
140  *
141  *	--ANK (980726)
142  */
143 
is_ineligible(const struct sk_buff * skb)144 static bool is_ineligible(const struct sk_buff *skb)
145 {
146 	int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
147 	int len = skb->len - ptr;
148 	__u8 nexthdr = ipv6_hdr(skb)->nexthdr;
149 	__be16 frag_off;
150 
151 	if (len < 0)
152 		return true;
153 
154 	ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off);
155 	if (ptr < 0)
156 		return false;
157 	if (nexthdr == IPPROTO_ICMPV6) {
158 		u8 _type, *tp;
159 		tp = skb_header_pointer(skb,
160 			ptr+offsetof(struct icmp6hdr, icmp6_type),
161 			sizeof(_type), &_type);
162 		if (tp == NULL ||
163 		    !(*tp & ICMPV6_INFOMSG_MASK))
164 			return true;
165 	}
166 	return false;
167 }
168 
169 /*
170  * Check the ICMP output rate limit
171  */
icmpv6_xrlim_allow(struct sock * sk,u8 type,struct flowi6 * fl6)172 static inline bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
173 				      struct flowi6 *fl6)
174 {
175 	struct dst_entry *dst;
176 	struct net *net = sock_net(sk);
177 	bool res = false;
178 
179 	/* Informational messages are not limited. */
180 	if (type & ICMPV6_INFOMSG_MASK)
181 		return true;
182 
183 	/* Do not limit pmtu discovery, it would break it. */
184 	if (type == ICMPV6_PKT_TOOBIG)
185 		return true;
186 
187 	/*
188 	 * Look up the output route.
189 	 * XXX: perhaps the expire for routing entries cloned by
190 	 * this lookup should be more aggressive (not longer than timeout).
191 	 */
192 	dst = ip6_route_output(net, sk, fl6);
193 	if (dst->error) {
194 		IP6_INC_STATS(net, ip6_dst_idev(dst),
195 			      IPSTATS_MIB_OUTNOROUTES);
196 	} else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
197 		res = true;
198 	} else {
199 		struct rt6_info *rt = (struct rt6_info *)dst;
200 		int tmo = net->ipv6.sysctl.icmpv6_time;
201 		struct inet_peer *peer;
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 		peer = inet_getpeer_v6(net->ipv6.peers, &rt->rt6i_dst.addr, 1);
208 		res = inet_peer_xrlim_allow(peer, tmo);
209 		if (peer)
210 			inet_putpeer(peer);
211 	}
212 	dst_release(dst);
213 	return res;
214 }
215 
216 /*
217  *	an inline helper for the "simple" if statement below
218  *	checks if parameter problem report is caused by an
219  *	unrecognized IPv6 option that has the Option Type
220  *	highest-order two bits set to 10
221  */
222 
opt_unrec(struct sk_buff * skb,__u32 offset)223 static bool opt_unrec(struct sk_buff *skb, __u32 offset)
224 {
225 	u8 _optval, *op;
226 
227 	offset += skb_network_offset(skb);
228 	op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
229 	if (op == NULL)
230 		return true;
231 	return (*op & 0xC0) == 0x80;
232 }
233 
icmpv6_push_pending_frames(struct sock * sk,struct flowi6 * fl6,struct icmp6hdr * thdr,int len)234 int icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
235 			       struct icmp6hdr *thdr, int len)
236 {
237 	struct sk_buff *skb;
238 	struct icmp6hdr *icmp6h;
239 	int err = 0;
240 
241 	if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
242 		goto out;
243 
244 	icmp6h = icmp6_hdr(skb);
245 	memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
246 	icmp6h->icmp6_cksum = 0;
247 
248 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
249 		skb->csum = csum_partial(icmp6h,
250 					sizeof(struct icmp6hdr), skb->csum);
251 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
252 						      &fl6->daddr,
253 						      len, fl6->flowi6_proto,
254 						      skb->csum);
255 	} else {
256 		__wsum tmp_csum = 0;
257 
258 		skb_queue_walk(&sk->sk_write_queue, skb) {
259 			tmp_csum = csum_add(tmp_csum, skb->csum);
260 		}
261 
262 		tmp_csum = csum_partial(icmp6h,
263 					sizeof(struct icmp6hdr), tmp_csum);
264 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
265 						      &fl6->daddr,
266 						      len, fl6->flowi6_proto,
267 						      tmp_csum);
268 	}
269 	ip6_push_pending_frames(sk);
270 out:
271 	return err;
272 }
273 
274 struct icmpv6_msg {
275 	struct sk_buff	*skb;
276 	int		offset;
277 	uint8_t		type;
278 };
279 
icmpv6_getfrag(void * from,char * to,int offset,int len,int odd,struct sk_buff * skb)280 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
281 {
282 	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
283 	struct sk_buff *org_skb = msg->skb;
284 	__wsum csum = 0;
285 
286 	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
287 				      to, len, csum);
288 	skb->csum = csum_block_add(skb->csum, csum, odd);
289 	if (!(msg->type & ICMPV6_INFOMSG_MASK))
290 		nf_ct_attach(skb, org_skb);
291 	return 0;
292 }
293 
294 #if IS_ENABLED(CONFIG_IPV6_MIP6)
mip6_addr_swap(struct sk_buff * skb)295 static void mip6_addr_swap(struct sk_buff *skb)
296 {
297 	struct ipv6hdr *iph = ipv6_hdr(skb);
298 	struct inet6_skb_parm *opt = IP6CB(skb);
299 	struct ipv6_destopt_hao *hao;
300 	struct in6_addr tmp;
301 	int off;
302 
303 	if (opt->dsthao) {
304 		off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
305 		if (likely(off >= 0)) {
306 			hao = (struct ipv6_destopt_hao *)
307 					(skb_network_header(skb) + off);
308 			tmp = iph->saddr;
309 			iph->saddr = hao->addr;
310 			hao->addr = tmp;
311 		}
312 	}
313 }
314 #else
mip6_addr_swap(struct sk_buff * skb)315 static inline void mip6_addr_swap(struct sk_buff *skb) {}
316 #endif
317 
icmpv6_route_lookup(struct net * net,struct sk_buff * skb,struct sock * sk,struct flowi6 * fl6)318 struct dst_entry *icmpv6_route_lookup(struct net *net, struct sk_buff *skb,
319 				      struct sock *sk, struct flowi6 *fl6)
320 {
321 	struct dst_entry *dst, *dst2;
322 	struct flowi6 fl2;
323 	int err;
324 
325 	err = ip6_dst_lookup(sk, &dst, fl6);
326 	if (err)
327 		return ERR_PTR(err);
328 
329 	/*
330 	 * We won't send icmp if the destination is known
331 	 * anycast.
332 	 */
333 	if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) {
334 		LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: acast source\n");
335 		dst_release(dst);
336 		return ERR_PTR(-EINVAL);
337 	}
338 
339 	/* No need to clone since we're just using its address. */
340 	dst2 = dst;
341 
342 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
343 	if (!IS_ERR(dst)) {
344 		if (dst != dst2)
345 			return dst;
346 	} else {
347 		if (PTR_ERR(dst) == -EPERM)
348 			dst = NULL;
349 		else
350 			return dst;
351 	}
352 
353 	err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
354 	if (err)
355 		goto relookup_failed;
356 
357 	err = ip6_dst_lookup(sk, &dst2, &fl2);
358 	if (err)
359 		goto relookup_failed;
360 
361 	dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
362 	if (!IS_ERR(dst2)) {
363 		dst_release(dst);
364 		dst = dst2;
365 	} else {
366 		err = PTR_ERR(dst2);
367 		if (err == -EPERM) {
368 			dst_release(dst);
369 			return dst2;
370 		} else
371 			goto relookup_failed;
372 	}
373 
374 relookup_failed:
375 	if (dst)
376 		return dst;
377 	return ERR_PTR(err);
378 }
379 
380 /*
381  *	Send an ICMP message in response to a packet in error
382  */
icmp6_send(struct sk_buff * skb,u8 type,u8 code,__u32 info)383 static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info)
384 {
385 	struct net *net = dev_net(skb->dev);
386 	struct inet6_dev *idev = NULL;
387 	struct ipv6hdr *hdr = ipv6_hdr(skb);
388 	struct sock *sk;
389 	struct ipv6_pinfo *np;
390 	const struct in6_addr *saddr = NULL;
391 	struct dst_entry *dst;
392 	struct icmp6hdr tmp_hdr;
393 	struct flowi6 fl6;
394 	struct icmpv6_msg msg;
395 	int iif = 0;
396 	int addr_type = 0;
397 	int len;
398 	int hlimit;
399 	int err = 0;
400 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
401 
402 	if ((u8 *)hdr < skb->head ||
403 	    (skb->network_header + sizeof(*hdr)) > skb->tail)
404 		return;
405 
406 	/*
407 	 *	Make sure we respect the rules
408 	 *	i.e. RFC 1885 2.4(e)
409 	 *	Rule (e.1) is enforced by not using icmp6_send
410 	 *	in any code that processes icmp errors.
411 	 */
412 	addr_type = ipv6_addr_type(&hdr->daddr);
413 
414 	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0))
415 		saddr = &hdr->daddr;
416 
417 	/*
418 	 *	Dest addr check
419 	 */
420 
421 	if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) {
422 		if (type != ICMPV6_PKT_TOOBIG &&
423 		    !(type == ICMPV6_PARAMPROB &&
424 		      code == ICMPV6_UNK_OPTION &&
425 		      (opt_unrec(skb, info))))
426 			return;
427 
428 		saddr = NULL;
429 	}
430 
431 	addr_type = ipv6_addr_type(&hdr->saddr);
432 
433 	/*
434 	 *	Source addr check
435 	 */
436 
437 	if (__ipv6_addr_needs_scope_id(addr_type))
438 		iif = skb->dev->ifindex;
439 
440 	/*
441 	 *	Must not send error if the source does not uniquely
442 	 *	identify a single node (RFC2463 Section 2.4).
443 	 *	We check unspecified / multicast addresses here,
444 	 *	and anycast addresses will be checked later.
445 	 */
446 	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
447 		LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: addr_any/mcast source\n");
448 		return;
449 	}
450 
451 	/*
452 	 *	Never answer to a ICMP packet.
453 	 */
454 	if (is_ineligible(skb)) {
455 		LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: no reply to icmp error\n");
456 		return;
457 	}
458 
459 	mip6_addr_swap(skb);
460 
461 	memset(&fl6, 0, sizeof(fl6));
462 	fl6.flowi6_proto = IPPROTO_ICMPV6;
463 	fl6.daddr = hdr->saddr;
464 	if (saddr)
465 		fl6.saddr = *saddr;
466 	fl6.flowi6_mark = mark;
467 	fl6.flowi6_oif = iif;
468 	fl6.fl6_icmp_type = type;
469 	fl6.fl6_icmp_code = code;
470 	fl6.flowi6_uid = sock_net_uid(net, NULL);
471 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
472 
473 	sk = icmpv6_xmit_lock(net);
474 	if (sk == NULL)
475 		return;
476 	sk->sk_mark = mark;
477 	np = inet6_sk(sk);
478 
479 	if (!icmpv6_xrlim_allow(sk, type, &fl6))
480 		goto out;
481 
482 	tmp_hdr.icmp6_type = type;
483 	tmp_hdr.icmp6_code = code;
484 	tmp_hdr.icmp6_cksum = 0;
485 	tmp_hdr.icmp6_pointer = htonl(info);
486 
487 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
488 		fl6.flowi6_oif = np->mcast_oif;
489 	else if (!fl6.flowi6_oif)
490 		fl6.flowi6_oif = np->ucast_oif;
491 
492 	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
493 	if (IS_ERR(dst))
494 		goto out;
495 
496 	if (ipv6_addr_is_multicast(&fl6.daddr))
497 		hlimit = np->mcast_hops;
498 	else
499 		hlimit = np->hop_limit;
500 	if (hlimit < 0)
501 		hlimit = ip6_dst_hoplimit(dst);
502 
503 	msg.skb = skb;
504 	msg.offset = skb_network_offset(skb);
505 	msg.type = type;
506 
507 	len = skb->len - msg.offset;
508 	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr));
509 	if (len < 0) {
510 		LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n");
511 		goto out_dst_release;
512 	}
513 
514 	rcu_read_lock();
515 	idev = __in6_dev_get(skb->dev);
516 
517 	err = ip6_append_data(sk, icmpv6_getfrag, &msg,
518 			      len + sizeof(struct icmp6hdr),
519 			      sizeof(struct icmp6hdr), hlimit,
520 			      np->tclass, NULL, &fl6, (struct rt6_info *)dst,
521 			      MSG_DONTWAIT, np->dontfrag);
522 	if (err) {
523 		ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTERRORS);
524 		ip6_flush_pending_frames(sk);
525 	} else {
526 		err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
527 						 len + sizeof(struct icmp6hdr));
528 	}
529 	rcu_read_unlock();
530 out_dst_release:
531 	dst_release(dst);
532 out:
533 	icmpv6_xmit_unlock(sk);
534 }
535 
536 /* Slightly more convenient version of icmp6_send.
537  */
icmpv6_param_prob(struct sk_buff * skb,u8 code,int pos)538 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
539 {
540 	icmp6_send(skb, ICMPV6_PARAMPROB, code, pos);
541 	kfree_skb(skb);
542 }
543 
icmpv6_echo_reply(struct sk_buff * skb)544 static void icmpv6_echo_reply(struct sk_buff *skb)
545 {
546 	struct net *net = dev_net(skb->dev);
547 	struct sock *sk;
548 	struct inet6_dev *idev;
549 	struct ipv6_pinfo *np;
550 	const struct in6_addr *saddr = NULL;
551 	struct icmp6hdr *icmph = icmp6_hdr(skb);
552 	struct icmp6hdr tmp_hdr;
553 	struct flowi6 fl6;
554 	struct icmpv6_msg msg;
555 	struct dst_entry *dst;
556 	int err = 0;
557 	int hlimit;
558 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
559 
560 	saddr = &ipv6_hdr(skb)->daddr;
561 
562 	if (!ipv6_unicast_destination(skb))
563 		saddr = NULL;
564 
565 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
566 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
567 
568 	memset(&fl6, 0, sizeof(fl6));
569 	fl6.flowi6_proto = IPPROTO_ICMPV6;
570 	fl6.daddr = ipv6_hdr(skb)->saddr;
571 	if (saddr)
572 		fl6.saddr = *saddr;
573 	fl6.flowi6_oif = skb->dev->ifindex;
574 	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
575 	fl6.flowi6_mark = mark;
576 	fl6.flowi6_uid = sock_net_uid(net, NULL);
577 	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
578 
579 	sk = icmpv6_xmit_lock(net);
580 	if (sk == NULL)
581 		return;
582 	sk->sk_mark = mark;
583 	np = inet6_sk(sk);
584 
585 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
586 		fl6.flowi6_oif = np->mcast_oif;
587 	else if (!fl6.flowi6_oif)
588 		fl6.flowi6_oif = np->ucast_oif;
589 
590 	err = ip6_dst_lookup(sk, &dst, &fl6);
591 	if (err)
592 		goto out;
593 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
594 	if (IS_ERR(dst))
595 		goto out;
596 
597 	if (ipv6_addr_is_multicast(&fl6.daddr))
598 		hlimit = np->mcast_hops;
599 	else
600 		hlimit = np->hop_limit;
601 	if (hlimit < 0)
602 		hlimit = ip6_dst_hoplimit(dst);
603 
604 	idev = __in6_dev_get(skb->dev);
605 
606 	msg.skb = skb;
607 	msg.offset = 0;
608 	msg.type = ICMPV6_ECHO_REPLY;
609 
610 	err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
611 				sizeof(struct icmp6hdr), hlimit, np->tclass, NULL, &fl6,
612 				(struct rt6_info *)dst, MSG_DONTWAIT,
613 				np->dontfrag);
614 
615 	if (err) {
616 		ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTERRORS);
617 		ip6_flush_pending_frames(sk);
618 	} else {
619 		err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
620 						 skb->len + sizeof(struct icmp6hdr));
621 	}
622 	dst_release(dst);
623 out:
624 	icmpv6_xmit_unlock(sk);
625 }
626 
icmpv6_notify(struct sk_buff * skb,u8 type,u8 code,__be32 info)627 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
628 {
629 	const struct inet6_protocol *ipprot;
630 	int inner_offset;
631 	__be16 frag_off;
632 	u8 nexthdr;
633 
634 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
635 		return;
636 
637 	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
638 	if (ipv6_ext_hdr(nexthdr)) {
639 		/* now skip over extension headers */
640 		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
641 						&nexthdr, &frag_off);
642 		if (inner_offset<0)
643 			return;
644 	} else {
645 		inner_offset = sizeof(struct ipv6hdr);
646 	}
647 
648 	/* Checkin header including 8 bytes of inner protocol header. */
649 	if (!pskb_may_pull(skb, inner_offset+8))
650 		return;
651 
652 	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
653 	   Without this we will not able f.e. to make source routed
654 	   pmtu discovery.
655 	   Corresponding argument (opt) to notifiers is already added.
656 	   --ANK (980726)
657 	 */
658 
659 	rcu_read_lock();
660 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
661 	if (ipprot && ipprot->err_handler)
662 		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
663 	rcu_read_unlock();
664 
665 	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
666 }
667 
668 /*
669  *	Handle icmp messages
670  */
671 
icmpv6_rcv(struct sk_buff * skb)672 static int icmpv6_rcv(struct sk_buff *skb)
673 {
674 	struct net_device *dev = skb->dev;
675 	struct inet6_dev *idev = __in6_dev_get(dev);
676 	const struct in6_addr *saddr, *daddr;
677 	struct icmp6hdr *hdr;
678 	u8 type;
679 
680 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
681 		struct sec_path *sp = skb_sec_path(skb);
682 		int nh;
683 
684 		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
685 				 XFRM_STATE_ICMP))
686 			goto drop_no_count;
687 
688 		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
689 			goto drop_no_count;
690 
691 		nh = skb_network_offset(skb);
692 		skb_set_network_header(skb, sizeof(*hdr));
693 
694 		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
695 			goto drop_no_count;
696 
697 		skb_set_network_header(skb, nh);
698 	}
699 
700 	ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INMSGS);
701 
702 	saddr = &ipv6_hdr(skb)->saddr;
703 	daddr = &ipv6_hdr(skb)->daddr;
704 
705 	/* Perform checksum. */
706 	switch (skb->ip_summed) {
707 	case CHECKSUM_COMPLETE:
708 		if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6,
709 				     skb->csum))
710 			break;
711 		/* fall through */
712 	case CHECKSUM_NONE:
713 		skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len,
714 					     IPPROTO_ICMPV6, 0));
715 		if (__skb_checksum_complete(skb)) {
716 			LIMIT_NETDEBUG(KERN_DEBUG
717 				       "ICMPv6 checksum failed [%pI6c > %pI6c]\n",
718 				       saddr, daddr);
719 			goto csum_error;
720 		}
721 	}
722 
723 	if (!pskb_pull(skb, sizeof(*hdr)))
724 		goto discard_it;
725 
726 	hdr = icmp6_hdr(skb);
727 
728 	type = hdr->icmp6_type;
729 
730 	ICMP6MSGIN_INC_STATS_BH(dev_net(dev), idev, type);
731 
732 	switch (type) {
733 	case ICMPV6_ECHO_REQUEST:
734 		icmpv6_echo_reply(skb);
735 		break;
736 
737 	case ICMPV6_ECHO_REPLY:
738 		ping_rcv(skb);
739 		break;
740 
741 	case ICMPV6_PKT_TOOBIG:
742 		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
743 		   standard destination cache. Seems, only "advanced"
744 		   destination cache will allow to solve this problem
745 		   --ANK (980726)
746 		 */
747 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
748 			goto discard_it;
749 		hdr = icmp6_hdr(skb);
750 
751 		/*
752 		 *	Drop through to notify
753 		 */
754 
755 	case ICMPV6_DEST_UNREACH:
756 	case ICMPV6_TIME_EXCEED:
757 	case ICMPV6_PARAMPROB:
758 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
759 		break;
760 
761 	case NDISC_ROUTER_SOLICITATION:
762 	case NDISC_ROUTER_ADVERTISEMENT:
763 	case NDISC_NEIGHBOUR_SOLICITATION:
764 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
765 	case NDISC_REDIRECT:
766 		ndisc_rcv(skb);
767 		break;
768 
769 	case ICMPV6_MGM_QUERY:
770 		igmp6_event_query(skb);
771 		break;
772 
773 	case ICMPV6_MGM_REPORT:
774 		igmp6_event_report(skb);
775 		break;
776 
777 	case ICMPV6_MGM_REDUCTION:
778 	case ICMPV6_NI_QUERY:
779 	case ICMPV6_NI_REPLY:
780 	case ICMPV6_MLD2_REPORT:
781 	case ICMPV6_DHAAD_REQUEST:
782 	case ICMPV6_DHAAD_REPLY:
783 	case ICMPV6_MOBILE_PREFIX_SOL:
784 	case ICMPV6_MOBILE_PREFIX_ADV:
785 		break;
786 
787 	default:
788 		LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n");
789 
790 		/* informational */
791 		if (type & ICMPV6_INFOMSG_MASK)
792 			break;
793 
794 		/*
795 		 * error of unknown type.
796 		 * must pass to upper level
797 		 */
798 
799 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
800 	}
801 
802 	kfree_skb(skb);
803 	return 0;
804 
805 csum_error:
806 	ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
807 discard_it:
808 	ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INERRORS);
809 drop_no_count:
810 	kfree_skb(skb);
811 	return 0;
812 }
813 
icmpv6_flow_init(struct sock * sk,struct flowi6 * fl6,u8 type,const struct in6_addr * saddr,const struct in6_addr * daddr,int oif)814 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
815 		      u8 type,
816 		      const struct in6_addr *saddr,
817 		      const struct in6_addr *daddr,
818 		      int oif)
819 {
820 	memset(fl6, 0, sizeof(*fl6));
821 	fl6->saddr = *saddr;
822 	fl6->daddr = *daddr;
823 	fl6->flowi6_proto 	= IPPROTO_ICMPV6;
824 	fl6->fl6_icmp_type	= type;
825 	fl6->fl6_icmp_code	= 0;
826 	fl6->flowi6_oif		= oif;
827 	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
828 }
829 
830 /*
831  * Special lock-class for __icmpv6_sk:
832  */
833 static struct lock_class_key icmpv6_socket_sk_dst_lock_key;
834 
icmpv6_sk_init(struct net * net)835 static int __net_init icmpv6_sk_init(struct net *net)
836 {
837 	struct sock *sk;
838 	int err, i, j;
839 
840 	net->ipv6.icmp_sk =
841 		kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
842 	if (net->ipv6.icmp_sk == NULL)
843 		return -ENOMEM;
844 
845 	for_each_possible_cpu(i) {
846 		err = inet_ctl_sock_create(&sk, PF_INET6,
847 					   SOCK_RAW, IPPROTO_ICMPV6, net);
848 		if (err < 0) {
849 			pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
850 			       err);
851 			goto fail;
852 		}
853 
854 		net->ipv6.icmp_sk[i] = sk;
855 
856 		/*
857 		 * Split off their lock-class, because sk->sk_dst_lock
858 		 * gets used from softirqs, which is safe for
859 		 * __icmpv6_sk (because those never get directly used
860 		 * via userspace syscalls), but unsafe for normal sockets.
861 		 */
862 		lockdep_set_class(&sk->sk_dst_lock,
863 				  &icmpv6_socket_sk_dst_lock_key);
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 
icmpv6_sk_exit(struct net * net)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 
icmpv6_init(void)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 
icmpv6_cleanup(void)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 };
952 
icmpv6_err_convert(u8 type,u8 code,int * err)953 int icmpv6_err_convert(u8 type, u8 code, int *err)
954 {
955 	int fatal = 0;
956 
957 	*err = EPROTO;
958 
959 	switch (type) {
960 	case ICMPV6_DEST_UNREACH:
961 		fatal = 1;
962 		if (code <= ICMPV6_PORT_UNREACH) {
963 			*err  = tab_unreach[code].err;
964 			fatal = tab_unreach[code].fatal;
965 		}
966 		break;
967 
968 	case ICMPV6_PKT_TOOBIG:
969 		*err = EMSGSIZE;
970 		break;
971 
972 	case ICMPV6_PARAMPROB:
973 		*err = EPROTO;
974 		fatal = 1;
975 		break;
976 
977 	case ICMPV6_TIME_EXCEED:
978 		*err = EHOSTUNREACH;
979 		break;
980 	}
981 
982 	return fatal;
983 }
984 EXPORT_SYMBOL(icmpv6_err_convert);
985 
986 #ifdef CONFIG_SYSCTL
987 ctl_table ipv6_icmp_table_template[] = {
988 	{
989 		.procname	= "ratelimit",
990 		.data		= &init_net.ipv6.sysctl.icmpv6_time,
991 		.maxlen		= sizeof(int),
992 		.mode		= 0644,
993 		.proc_handler	= proc_dointvec_ms_jiffies,
994 	},
995 	{ },
996 };
997 
ipv6_icmp_sysctl_init(struct net * net)998 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
999 {
1000 	struct ctl_table *table;
1001 
1002 	table = kmemdup(ipv6_icmp_table_template,
1003 			sizeof(ipv6_icmp_table_template),
1004 			GFP_KERNEL);
1005 
1006 	if (table)
1007 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
1008 
1009 	return table;
1010 }
1011 #endif
1012 
1013