xref: /linux/net/ipv6/icmp.c (revision b77e0ce62d63a761ffb7f7245a215a49f5921c2f)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	Internet Control Message Protocol (ICMPv6)
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *
9  *	Based on net/ipv4/icmp.c
10  *
11  *	RFC 1885
12  */
13 
14 /*
15  *	Changes:
16  *
17  *	Andi Kleen		:	exception handling
18  *	Andi Kleen			add rate limits. never reply to a icmp.
19  *					add more length checks and other fixes.
20  *	yoshfuji		:	ensure to sent parameter problem for
21  *					fragments.
22  *	YOSHIFUJI Hideaki @USAGI:	added sysctl for icmp rate limit.
23  *	Randy Dunlap and
24  *	YOSHIFUJI Hideaki @USAGI:	Per-interface statistics support
25  *	Kazunori MIYAZAWA @USAGI:       change output process to use ip6_append_data
26  */
27 
28 #define pr_fmt(fmt) "IPv6: " fmt
29 
30 #include <linux/module.h>
31 #include <linux/errno.h>
32 #include <linux/types.h>
33 #include <linux/socket.h>
34 #include <linux/in.h>
35 #include <linux/kernel.h>
36 #include <linux/sockios.h>
37 #include <linux/net.h>
38 #include <linux/skbuff.h>
39 #include <linux/init.h>
40 #include <linux/netfilter.h>
41 #include <linux/slab.h>
42 
43 #ifdef CONFIG_SYSCTL
44 #include <linux/sysctl.h>
45 #endif
46 
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <linux/icmpv6.h>
50 
51 #include <net/ip.h>
52 #include <net/sock.h>
53 
54 #include <net/ipv6.h>
55 #include <net/ip6_checksum.h>
56 #include <net/ping.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 #include <net/dsfield.h>
67 #include <net/l3mdev.h>
68 
69 #include <linux/uaccess.h>
70 
71 /*
72  *	The ICMP socket(s). This is the most convenient way to flow control
73  *	our ICMP output as well as maintain a clean interface throughout
74  *	all layers. All Socketless IP sends will soon be gone.
75  *
76  *	On SMP we have one ICMP socket per-cpu.
77  */
78 static struct sock *icmpv6_sk(struct net *net)
79 {
80 	return this_cpu_read(*net->ipv6.icmp_sk);
81 }
82 
83 static int icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
84 		       u8 type, u8 code, int offset, __be32 info)
85 {
86 	/* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */
87 	struct icmp6hdr *icmp6 = (struct icmp6hdr *) (skb->data + offset);
88 	struct net *net = dev_net(skb->dev);
89 
90 	if (type == ICMPV6_PKT_TOOBIG)
91 		ip6_update_pmtu(skb, net, info, skb->dev->ifindex, 0, sock_net_uid(net, NULL));
92 	else if (type == NDISC_REDIRECT)
93 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
94 			     sock_net_uid(net, NULL));
95 
96 	if (!(type & ICMPV6_INFOMSG_MASK))
97 		if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST)
98 			ping_err(skb, offset, ntohl(info));
99 
100 	return 0;
101 }
102 
103 static int icmpv6_rcv(struct sk_buff *skb);
104 
105 static const struct inet6_protocol icmpv6_protocol = {
106 	.handler	=	icmpv6_rcv,
107 	.err_handler	=	icmpv6_err,
108 	.flags		=	INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
109 };
110 
111 /* Called with BH disabled */
112 static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
113 {
114 	struct sock *sk;
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 		return NULL;
123 	}
124 	return sk;
125 }
126 
127 static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
128 {
129 	spin_unlock(&sk->sk_lock.slock);
130 }
131 
132 /*
133  * Figure out, may we reply to this packet with icmp error.
134  *
135  * We do not reply, if:
136  *	- it was icmp error message.
137  *	- it is truncated, so that it is known, that protocol is ICMPV6
138  *	  (i.e. in the middle of some exthdr)
139  *
140  *	--ANK (980726)
141  */
142 
143 static bool is_ineligible(const struct sk_buff *skb)
144 {
145 	int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
146 	int len = skb->len - ptr;
147 	__u8 nexthdr = ipv6_hdr(skb)->nexthdr;
148 	__be16 frag_off;
149 
150 	if (len < 0)
151 		return true;
152 
153 	ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off);
154 	if (ptr < 0)
155 		return false;
156 	if (nexthdr == IPPROTO_ICMPV6) {
157 		u8 _type, *tp;
158 		tp = skb_header_pointer(skb,
159 			ptr+offsetof(struct icmp6hdr, icmp6_type),
160 			sizeof(_type), &_type);
161 
162 		/* Based on RFC 8200, Section 4.5 Fragment Header, return
163 		 * false if this is a fragment packet with no icmp header info.
164 		 */
165 		if (!tp && frag_off != 0)
166 			return false;
167 		else if (!tp || !(*tp & ICMPV6_INFOMSG_MASK))
168 			return true;
169 	}
170 	return false;
171 }
172 
173 static bool icmpv6_mask_allow(struct net *net, int type)
174 {
175 	if (type > ICMPV6_MSG_MAX)
176 		return true;
177 
178 	/* Limit if icmp type is set in ratemask. */
179 	if (!test_bit(type, net->ipv6.sysctl.icmpv6_ratemask))
180 		return true;
181 
182 	return false;
183 }
184 
185 static bool icmpv6_global_allow(struct net *net, int type)
186 {
187 	if (icmpv6_mask_allow(net, type))
188 		return true;
189 
190 	if (icmp_global_allow())
191 		return true;
192 
193 	return false;
194 }
195 
196 /*
197  * Check the ICMP output rate limit
198  */
199 static bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
200 			       struct flowi6 *fl6)
201 {
202 	struct net *net = sock_net(sk);
203 	struct dst_entry *dst;
204 	bool res = false;
205 
206 	if (icmpv6_mask_allow(net, type))
207 		return true;
208 
209 	/*
210 	 * Look up the output route.
211 	 * XXX: perhaps the expire for routing entries cloned by
212 	 * this lookup should be more aggressive (not longer than timeout).
213 	 */
214 	dst = ip6_route_output(net, sk, fl6);
215 	if (dst->error) {
216 		IP6_INC_STATS(net, ip6_dst_idev(dst),
217 			      IPSTATS_MIB_OUTNOROUTES);
218 	} else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
219 		res = true;
220 	} else {
221 		struct rt6_info *rt = (struct rt6_info *)dst;
222 		int tmo = net->ipv6.sysctl.icmpv6_time;
223 		struct inet_peer *peer;
224 
225 		/* Give more bandwidth to wider prefixes. */
226 		if (rt->rt6i_dst.plen < 128)
227 			tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
228 
229 		peer = inet_getpeer_v6(net->ipv6.peers, &fl6->daddr, 1);
230 		res = inet_peer_xrlim_allow(peer, tmo);
231 		if (peer)
232 			inet_putpeer(peer);
233 	}
234 	dst_release(dst);
235 	return res;
236 }
237 
238 static bool icmpv6_rt_has_prefsrc(struct sock *sk, u8 type,
239 				  struct flowi6 *fl6)
240 {
241 	struct net *net = sock_net(sk);
242 	struct dst_entry *dst;
243 	bool res = false;
244 
245 	dst = ip6_route_output(net, sk, fl6);
246 	if (!dst->error) {
247 		struct rt6_info *rt = (struct rt6_info *)dst;
248 		struct in6_addr prefsrc;
249 
250 		rt6_get_prefsrc(rt, &prefsrc);
251 		res = !ipv6_addr_any(&prefsrc);
252 	}
253 	dst_release(dst);
254 	return res;
255 }
256 
257 /*
258  *	an inline helper for the "simple" if statement below
259  *	checks if parameter problem report is caused by an
260  *	unrecognized IPv6 option that has the Option Type
261  *	highest-order two bits set to 10
262  */
263 
264 static bool opt_unrec(struct sk_buff *skb, __u32 offset)
265 {
266 	u8 _optval, *op;
267 
268 	offset += skb_network_offset(skb);
269 	op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
270 	if (!op)
271 		return true;
272 	return (*op & 0xC0) == 0x80;
273 }
274 
275 void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
276 				struct icmp6hdr *thdr, int len)
277 {
278 	struct sk_buff *skb;
279 	struct icmp6hdr *icmp6h;
280 
281 	skb = skb_peek(&sk->sk_write_queue);
282 	if (!skb)
283 		return;
284 
285 	icmp6h = icmp6_hdr(skb);
286 	memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
287 	icmp6h->icmp6_cksum = 0;
288 
289 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
290 		skb->csum = csum_partial(icmp6h,
291 					sizeof(struct icmp6hdr), skb->csum);
292 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
293 						      &fl6->daddr,
294 						      len, fl6->flowi6_proto,
295 						      skb->csum);
296 	} else {
297 		__wsum tmp_csum = 0;
298 
299 		skb_queue_walk(&sk->sk_write_queue, skb) {
300 			tmp_csum = csum_add(tmp_csum, skb->csum);
301 		}
302 
303 		tmp_csum = csum_partial(icmp6h,
304 					sizeof(struct icmp6hdr), tmp_csum);
305 		icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
306 						      &fl6->daddr,
307 						      len, fl6->flowi6_proto,
308 						      tmp_csum);
309 	}
310 	ip6_push_pending_frames(sk);
311 }
312 
313 struct icmpv6_msg {
314 	struct sk_buff	*skb;
315 	int		offset;
316 	uint8_t		type;
317 };
318 
319 static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
320 {
321 	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
322 	struct sk_buff *org_skb = msg->skb;
323 	__wsum csum;
324 
325 	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
326 				      to, len);
327 	skb->csum = csum_block_add(skb->csum, csum, odd);
328 	if (!(msg->type & ICMPV6_INFOMSG_MASK))
329 		nf_ct_attach(skb, org_skb);
330 	return 0;
331 }
332 
333 #if IS_ENABLED(CONFIG_IPV6_MIP6)
334 static void mip6_addr_swap(struct sk_buff *skb, const struct inet6_skb_parm *opt)
335 {
336 	struct ipv6hdr *iph = ipv6_hdr(skb);
337 	struct ipv6_destopt_hao *hao;
338 	struct in6_addr tmp;
339 	int off;
340 
341 	if (opt->dsthao) {
342 		off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
343 		if (likely(off >= 0)) {
344 			hao = (struct ipv6_destopt_hao *)
345 					(skb_network_header(skb) + off);
346 			tmp = iph->saddr;
347 			iph->saddr = hao->addr;
348 			hao->addr = tmp;
349 		}
350 	}
351 }
352 #else
353 static inline void mip6_addr_swap(struct sk_buff *skb, const struct inet6_skb_parm *opt) {}
354 #endif
355 
356 static struct dst_entry *icmpv6_route_lookup(struct net *net,
357 					     struct sk_buff *skb,
358 					     struct sock *sk,
359 					     struct flowi6 *fl6)
360 {
361 	struct dst_entry *dst, *dst2;
362 	struct flowi6 fl2;
363 	int err;
364 
365 	err = ip6_dst_lookup(net, sk, &dst, fl6);
366 	if (err)
367 		return ERR_PTR(err);
368 
369 	/*
370 	 * We won't send icmp if the destination is known
371 	 * anycast.
372 	 */
373 	if (ipv6_anycast_destination(dst, &fl6->daddr)) {
374 		net_dbg_ratelimited("icmp6_send: acast source\n");
375 		dst_release(dst);
376 		return ERR_PTR(-EINVAL);
377 	}
378 
379 	/* No need to clone since we're just using its address. */
380 	dst2 = dst;
381 
382 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
383 	if (!IS_ERR(dst)) {
384 		if (dst != dst2)
385 			return dst;
386 	} else {
387 		if (PTR_ERR(dst) == -EPERM)
388 			dst = NULL;
389 		else
390 			return dst;
391 	}
392 
393 	err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
394 	if (err)
395 		goto relookup_failed;
396 
397 	err = ip6_dst_lookup(net, sk, &dst2, &fl2);
398 	if (err)
399 		goto relookup_failed;
400 
401 	dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
402 	if (!IS_ERR(dst2)) {
403 		dst_release(dst);
404 		dst = dst2;
405 	} else {
406 		err = PTR_ERR(dst2);
407 		if (err == -EPERM) {
408 			dst_release(dst);
409 			return dst2;
410 		} else
411 			goto relookup_failed;
412 	}
413 
414 relookup_failed:
415 	if (dst)
416 		return dst;
417 	return ERR_PTR(err);
418 }
419 
420 static struct net_device *icmp6_dev(const struct sk_buff *skb)
421 {
422 	struct net_device *dev = skb->dev;
423 
424 	/* for local traffic to local address, skb dev is the loopback
425 	 * device. Check if there is a dst attached to the skb and if so
426 	 * get the real device index. Same is needed for replies to a link
427 	 * local address on a device enslaved to an L3 master device
428 	 */
429 	if (unlikely(dev->ifindex == LOOPBACK_IFINDEX || netif_is_l3_master(skb->dev))) {
430 		const struct rt6_info *rt6 = skb_rt6_info(skb);
431 
432 		if (rt6)
433 			dev = rt6->rt6i_idev->dev;
434 	}
435 
436 	return dev;
437 }
438 
439 static int icmp6_iif(const struct sk_buff *skb)
440 {
441 	return icmp6_dev(skb)->ifindex;
442 }
443 
444 /*
445  *	Send an ICMP message in response to a packet in error
446  */
447 void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info,
448 		const struct in6_addr *force_saddr,
449 		const struct inet6_skb_parm *parm)
450 {
451 	struct inet6_dev *idev = NULL;
452 	struct ipv6hdr *hdr = ipv6_hdr(skb);
453 	struct sock *sk;
454 	struct net *net;
455 	struct ipv6_pinfo *np;
456 	const struct in6_addr *saddr = NULL;
457 	struct dst_entry *dst;
458 	struct icmp6hdr tmp_hdr;
459 	struct flowi6 fl6;
460 	struct icmpv6_msg msg;
461 	struct ipcm6_cookie ipc6;
462 	int iif = 0;
463 	int addr_type = 0;
464 	int len;
465 	u32 mark;
466 
467 	if ((u8 *)hdr < skb->head ||
468 	    (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb))
469 		return;
470 
471 	if (!skb->dev)
472 		return;
473 	net = dev_net(skb->dev);
474 	mark = IP6_REPLY_MARK(net, skb->mark);
475 	/*
476 	 *	Make sure we respect the rules
477 	 *	i.e. RFC 1885 2.4(e)
478 	 *	Rule (e.1) is enforced by not using icmp6_send
479 	 *	in any code that processes icmp errors.
480 	 */
481 	addr_type = ipv6_addr_type(&hdr->daddr);
482 
483 	if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) ||
484 	    ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr))
485 		saddr = &hdr->daddr;
486 
487 	/*
488 	 *	Dest addr check
489 	 */
490 
491 	if (addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST) {
492 		if (type != ICMPV6_PKT_TOOBIG &&
493 		    !(type == ICMPV6_PARAMPROB &&
494 		      code == ICMPV6_UNK_OPTION &&
495 		      (opt_unrec(skb, info))))
496 			return;
497 
498 		saddr = NULL;
499 	}
500 
501 	addr_type = ipv6_addr_type(&hdr->saddr);
502 
503 	/*
504 	 *	Source addr check
505 	 */
506 
507 	if (__ipv6_addr_needs_scope_id(addr_type)) {
508 		iif = icmp6_iif(skb);
509 	} else {
510 		/*
511 		 * The source device is used for looking up which routing table
512 		 * to use for sending an ICMP error.
513 		 */
514 		iif = l3mdev_master_ifindex(skb->dev);
515 	}
516 
517 	/*
518 	 *	Must not send error if the source does not uniquely
519 	 *	identify a single node (RFC2463 Section 2.4).
520 	 *	We check unspecified / multicast addresses here,
521 	 *	and anycast addresses will be checked later.
522 	 */
523 	if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
524 		net_dbg_ratelimited("icmp6_send: addr_any/mcast source [%pI6c > %pI6c]\n",
525 				    &hdr->saddr, &hdr->daddr);
526 		return;
527 	}
528 
529 	/*
530 	 *	Never answer to a ICMP packet.
531 	 */
532 	if (is_ineligible(skb)) {
533 		net_dbg_ratelimited("icmp6_send: no reply to icmp error [%pI6c > %pI6c]\n",
534 				    &hdr->saddr, &hdr->daddr);
535 		return;
536 	}
537 
538 	/* Needed by both icmp_global_allow and icmpv6_xmit_lock */
539 	local_bh_disable();
540 
541 	/* Check global sysctl_icmp_msgs_per_sec ratelimit */
542 	if (!(skb->dev->flags & IFF_LOOPBACK) && !icmpv6_global_allow(net, type))
543 		goto out_bh_enable;
544 
545 	mip6_addr_swap(skb, parm);
546 
547 	sk = icmpv6_xmit_lock(net);
548 	if (!sk)
549 		goto out_bh_enable;
550 
551 	memset(&fl6, 0, sizeof(fl6));
552 	fl6.flowi6_proto = IPPROTO_ICMPV6;
553 	fl6.daddr = hdr->saddr;
554 	if (force_saddr)
555 		saddr = force_saddr;
556 	if (saddr) {
557 		fl6.saddr = *saddr;
558 	} else if (!icmpv6_rt_has_prefsrc(sk, type, &fl6)) {
559 		/* select a more meaningful saddr from input if */
560 		struct net_device *in_netdev;
561 
562 		in_netdev = dev_get_by_index(net, parm->iif);
563 		if (in_netdev) {
564 			ipv6_dev_get_saddr(net, in_netdev, &fl6.daddr,
565 					   inet6_sk(sk)->srcprefs,
566 					   &fl6.saddr);
567 			dev_put(in_netdev);
568 		}
569 	}
570 	fl6.flowi6_mark = mark;
571 	fl6.flowi6_oif = iif;
572 	fl6.fl6_icmp_type = type;
573 	fl6.fl6_icmp_code = code;
574 	fl6.flowi6_uid = sock_net_uid(net, NULL);
575 	fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, NULL);
576 	security_skb_classify_flow(skb, flowi6_to_flowi_common(&fl6));
577 
578 	np = inet6_sk(sk);
579 
580 	if (!icmpv6_xrlim_allow(sk, type, &fl6))
581 		goto out;
582 
583 	tmp_hdr.icmp6_type = type;
584 	tmp_hdr.icmp6_code = code;
585 	tmp_hdr.icmp6_cksum = 0;
586 	tmp_hdr.icmp6_pointer = htonl(info);
587 
588 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
589 		fl6.flowi6_oif = np->mcast_oif;
590 	else if (!fl6.flowi6_oif)
591 		fl6.flowi6_oif = np->ucast_oif;
592 
593 	ipcm6_init_sk(&ipc6, np);
594 	ipc6.sockc.mark = mark;
595 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
596 
597 	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
598 	if (IS_ERR(dst))
599 		goto out;
600 
601 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
602 
603 	msg.skb = skb;
604 	msg.offset = skb_network_offset(skb);
605 	msg.type = type;
606 
607 	len = skb->len - msg.offset;
608 	len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(struct icmp6hdr));
609 	if (len < 0) {
610 		net_dbg_ratelimited("icmp: len problem [%pI6c > %pI6c]\n",
611 				    &hdr->saddr, &hdr->daddr);
612 		goto out_dst_release;
613 	}
614 
615 	rcu_read_lock();
616 	idev = __in6_dev_get(skb->dev);
617 
618 	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
619 			    len + sizeof(struct icmp6hdr),
620 			    sizeof(struct icmp6hdr),
621 			    &ipc6, &fl6, (struct rt6_info *)dst,
622 			    MSG_DONTWAIT)) {
623 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
624 		ip6_flush_pending_frames(sk);
625 	} else {
626 		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
627 					   len + sizeof(struct icmp6hdr));
628 	}
629 	rcu_read_unlock();
630 out_dst_release:
631 	dst_release(dst);
632 out:
633 	icmpv6_xmit_unlock(sk);
634 out_bh_enable:
635 	local_bh_enable();
636 }
637 EXPORT_SYMBOL(icmp6_send);
638 
639 /* Slightly more convenient version of icmp6_send.
640  */
641 void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
642 {
643 	icmp6_send(skb, ICMPV6_PARAMPROB, code, pos, NULL, IP6CB(skb));
644 	kfree_skb(skb);
645 }
646 
647 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
648  * if sufficient data bytes are available
649  * @nhs is the size of the tunnel header(s) :
650  *  Either an IPv4 header for SIT encap
651  *         an IPv4 header + GRE header for GRE encap
652  */
653 int ip6_err_gen_icmpv6_unreach(struct sk_buff *skb, int nhs, int type,
654 			       unsigned int data_len)
655 {
656 	struct in6_addr temp_saddr;
657 	struct rt6_info *rt;
658 	struct sk_buff *skb2;
659 	u32 info = 0;
660 
661 	if (!pskb_may_pull(skb, nhs + sizeof(struct ipv6hdr) + 8))
662 		return 1;
663 
664 	/* RFC 4884 (partial) support for ICMP extensions */
665 	if (data_len < 128 || (data_len & 7) || skb->len < data_len)
666 		data_len = 0;
667 
668 	skb2 = data_len ? skb_copy(skb, GFP_ATOMIC) : skb_clone(skb, GFP_ATOMIC);
669 
670 	if (!skb2)
671 		return 1;
672 
673 	skb_dst_drop(skb2);
674 	skb_pull(skb2, nhs);
675 	skb_reset_network_header(skb2);
676 
677 	rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0,
678 			skb, 0);
679 
680 	if (rt && rt->dst.dev)
681 		skb2->dev = rt->dst.dev;
682 
683 	ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, &temp_saddr);
684 
685 	if (data_len) {
686 		/* RFC 4884 (partial) support :
687 		 * insert 0 padding at the end, before the extensions
688 		 */
689 		__skb_push(skb2, nhs);
690 		skb_reset_network_header(skb2);
691 		memmove(skb2->data, skb2->data + nhs, data_len - nhs);
692 		memset(skb2->data + data_len - nhs, 0, nhs);
693 		/* RFC 4884 4.5 : Length is measured in 64-bit words,
694 		 * and stored in reserved[0]
695 		 */
696 		info = (data_len/8) << 24;
697 	}
698 	if (type == ICMP_TIME_EXCEEDED)
699 		icmp6_send(skb2, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
700 			   info, &temp_saddr, IP6CB(skb2));
701 	else
702 		icmp6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH,
703 			   info, &temp_saddr, IP6CB(skb2));
704 	if (rt)
705 		ip6_rt_put(rt);
706 
707 	kfree_skb(skb2);
708 
709 	return 0;
710 }
711 EXPORT_SYMBOL(ip6_err_gen_icmpv6_unreach);
712 
713 static void icmpv6_echo_reply(struct sk_buff *skb)
714 {
715 	struct net *net = dev_net(skb->dev);
716 	struct sock *sk;
717 	struct inet6_dev *idev;
718 	struct ipv6_pinfo *np;
719 	const struct in6_addr *saddr = NULL;
720 	struct icmp6hdr *icmph = icmp6_hdr(skb);
721 	struct icmp6hdr tmp_hdr;
722 	struct flowi6 fl6;
723 	struct icmpv6_msg msg;
724 	struct dst_entry *dst;
725 	struct ipcm6_cookie ipc6;
726 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
727 	bool acast;
728 
729 	if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) &&
730 	    net->ipv6.sysctl.icmpv6_echo_ignore_multicast)
731 		return;
732 
733 	saddr = &ipv6_hdr(skb)->daddr;
734 
735 	acast = ipv6_anycast_destination(skb_dst(skb), saddr);
736 	if (acast && net->ipv6.sysctl.icmpv6_echo_ignore_anycast)
737 		return;
738 
739 	if (!ipv6_unicast_destination(skb) &&
740 	    !(net->ipv6.sysctl.anycast_src_echo_reply && acast))
741 		saddr = NULL;
742 
743 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
744 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
745 
746 	memset(&fl6, 0, sizeof(fl6));
747 	if (net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ICMPV6_ECHO_REPLIES)
748 		fl6.flowlabel = ip6_flowlabel(ipv6_hdr(skb));
749 
750 	fl6.flowi6_proto = IPPROTO_ICMPV6;
751 	fl6.daddr = ipv6_hdr(skb)->saddr;
752 	if (saddr)
753 		fl6.saddr = *saddr;
754 	fl6.flowi6_oif = icmp6_iif(skb);
755 	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
756 	fl6.flowi6_mark = mark;
757 	fl6.flowi6_uid = sock_net_uid(net, NULL);
758 	security_skb_classify_flow(skb, flowi6_to_flowi_common(&fl6));
759 
760 	local_bh_disable();
761 	sk = icmpv6_xmit_lock(net);
762 	if (!sk)
763 		goto out_bh_enable;
764 	np = inet6_sk(sk);
765 
766 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
767 		fl6.flowi6_oif = np->mcast_oif;
768 	else if (!fl6.flowi6_oif)
769 		fl6.flowi6_oif = np->ucast_oif;
770 
771 	if (ip6_dst_lookup(net, sk, &dst, &fl6))
772 		goto out;
773 	dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
774 	if (IS_ERR(dst))
775 		goto out;
776 
777 	/* Check the ratelimit */
778 	if ((!(skb->dev->flags & IFF_LOOPBACK) && !icmpv6_global_allow(net, ICMPV6_ECHO_REPLY)) ||
779 	    !icmpv6_xrlim_allow(sk, ICMPV6_ECHO_REPLY, &fl6))
780 		goto out_dst_release;
781 
782 	idev = __in6_dev_get(skb->dev);
783 
784 	msg.skb = skb;
785 	msg.offset = 0;
786 	msg.type = ICMPV6_ECHO_REPLY;
787 
788 	ipcm6_init_sk(&ipc6, np);
789 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
790 	ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb));
791 	ipc6.sockc.mark = mark;
792 
793 	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
794 			    skb->len + sizeof(struct icmp6hdr),
795 			    sizeof(struct icmp6hdr), &ipc6, &fl6,
796 			    (struct rt6_info *)dst, MSG_DONTWAIT)) {
797 		__ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
798 		ip6_flush_pending_frames(sk);
799 	} else {
800 		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
801 					   skb->len + sizeof(struct icmp6hdr));
802 	}
803 out_dst_release:
804 	dst_release(dst);
805 out:
806 	icmpv6_xmit_unlock(sk);
807 out_bh_enable:
808 	local_bh_enable();
809 }
810 
811 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
812 {
813 	const struct inet6_protocol *ipprot;
814 	int inner_offset;
815 	__be16 frag_off;
816 	u8 nexthdr;
817 	struct net *net = dev_net(skb->dev);
818 
819 	if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
820 		goto out;
821 
822 	nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
823 	if (ipv6_ext_hdr(nexthdr)) {
824 		/* now skip over extension headers */
825 		inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
826 						&nexthdr, &frag_off);
827 		if (inner_offset < 0)
828 			goto out;
829 	} else {
830 		inner_offset = sizeof(struct ipv6hdr);
831 	}
832 
833 	/* Checkin header including 8 bytes of inner protocol header. */
834 	if (!pskb_may_pull(skb, inner_offset+8))
835 		goto out;
836 
837 	/* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
838 	   Without this we will not able f.e. to make source routed
839 	   pmtu discovery.
840 	   Corresponding argument (opt) to notifiers is already added.
841 	   --ANK (980726)
842 	 */
843 
844 	ipprot = rcu_dereference(inet6_protos[nexthdr]);
845 	if (ipprot && ipprot->err_handler)
846 		ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
847 
848 	raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
849 	return;
850 
851 out:
852 	__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
853 }
854 
855 /*
856  *	Handle icmp messages
857  */
858 
859 static int icmpv6_rcv(struct sk_buff *skb)
860 {
861 	struct net *net = dev_net(skb->dev);
862 	struct net_device *dev = icmp6_dev(skb);
863 	struct inet6_dev *idev = __in6_dev_get(dev);
864 	const struct in6_addr *saddr, *daddr;
865 	struct icmp6hdr *hdr;
866 	u8 type;
867 	bool success = false;
868 
869 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
870 		struct sec_path *sp = skb_sec_path(skb);
871 		int nh;
872 
873 		if (!(sp && sp->xvec[sp->len - 1]->props.flags &
874 				 XFRM_STATE_ICMP))
875 			goto drop_no_count;
876 
877 		if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
878 			goto drop_no_count;
879 
880 		nh = skb_network_offset(skb);
881 		skb_set_network_header(skb, sizeof(*hdr));
882 
883 		if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
884 			goto drop_no_count;
885 
886 		skb_set_network_header(skb, nh);
887 	}
888 
889 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS);
890 
891 	saddr = &ipv6_hdr(skb)->saddr;
892 	daddr = &ipv6_hdr(skb)->daddr;
893 
894 	if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) {
895 		net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n",
896 				    saddr, daddr);
897 		goto csum_error;
898 	}
899 
900 	if (!pskb_pull(skb, sizeof(*hdr)))
901 		goto discard_it;
902 
903 	hdr = icmp6_hdr(skb);
904 
905 	type = hdr->icmp6_type;
906 
907 	ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type);
908 
909 	switch (type) {
910 	case ICMPV6_ECHO_REQUEST:
911 		if (!net->ipv6.sysctl.icmpv6_echo_ignore_all)
912 			icmpv6_echo_reply(skb);
913 		break;
914 
915 	case ICMPV6_ECHO_REPLY:
916 		success = ping_rcv(skb);
917 		break;
918 
919 	case ICMPV6_PKT_TOOBIG:
920 		/* BUGGG_FUTURE: if packet contains rthdr, we cannot update
921 		   standard destination cache. Seems, only "advanced"
922 		   destination cache will allow to solve this problem
923 		   --ANK (980726)
924 		 */
925 		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
926 			goto discard_it;
927 		hdr = icmp6_hdr(skb);
928 
929 		/* to notify */
930 		fallthrough;
931 	case ICMPV6_DEST_UNREACH:
932 	case ICMPV6_TIME_EXCEED:
933 	case ICMPV6_PARAMPROB:
934 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
935 		break;
936 
937 	case NDISC_ROUTER_SOLICITATION:
938 	case NDISC_ROUTER_ADVERTISEMENT:
939 	case NDISC_NEIGHBOUR_SOLICITATION:
940 	case NDISC_NEIGHBOUR_ADVERTISEMENT:
941 	case NDISC_REDIRECT:
942 		ndisc_rcv(skb);
943 		break;
944 
945 	case ICMPV6_MGM_QUERY:
946 		igmp6_event_query(skb);
947 		break;
948 
949 	case ICMPV6_MGM_REPORT:
950 		igmp6_event_report(skb);
951 		break;
952 
953 	case ICMPV6_MGM_REDUCTION:
954 	case ICMPV6_NI_QUERY:
955 	case ICMPV6_NI_REPLY:
956 	case ICMPV6_MLD2_REPORT:
957 	case ICMPV6_DHAAD_REQUEST:
958 	case ICMPV6_DHAAD_REPLY:
959 	case ICMPV6_MOBILE_PREFIX_SOL:
960 	case ICMPV6_MOBILE_PREFIX_ADV:
961 		break;
962 
963 	default:
964 		/* informational */
965 		if (type & ICMPV6_INFOMSG_MASK)
966 			break;
967 
968 		net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n",
969 				    saddr, daddr);
970 
971 		/*
972 		 * error of unknown type.
973 		 * must pass to upper level
974 		 */
975 
976 		icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
977 	}
978 
979 	/* until the v6 path can be better sorted assume failure and
980 	 * preserve the status quo behaviour for the rest of the paths to here
981 	 */
982 	if (success)
983 		consume_skb(skb);
984 	else
985 		kfree_skb(skb);
986 
987 	return 0;
988 
989 csum_error:
990 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
991 discard_it:
992 	__ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS);
993 drop_no_count:
994 	kfree_skb(skb);
995 	return 0;
996 }
997 
998 void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
999 		      u8 type,
1000 		      const struct in6_addr *saddr,
1001 		      const struct in6_addr *daddr,
1002 		      int oif)
1003 {
1004 	memset(fl6, 0, sizeof(*fl6));
1005 	fl6->saddr = *saddr;
1006 	fl6->daddr = *daddr;
1007 	fl6->flowi6_proto	= IPPROTO_ICMPV6;
1008 	fl6->fl6_icmp_type	= type;
1009 	fl6->fl6_icmp_code	= 0;
1010 	fl6->flowi6_oif		= oif;
1011 	security_sk_classify_flow(sk, flowi6_to_flowi_common(fl6));
1012 }
1013 
1014 static void __net_exit icmpv6_sk_exit(struct net *net)
1015 {
1016 	int i;
1017 
1018 	for_each_possible_cpu(i)
1019 		inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv6.icmp_sk, i));
1020 	free_percpu(net->ipv6.icmp_sk);
1021 }
1022 
1023 static int __net_init icmpv6_sk_init(struct net *net)
1024 {
1025 	struct sock *sk;
1026 	int err, i;
1027 
1028 	net->ipv6.icmp_sk = alloc_percpu(struct sock *);
1029 	if (!net->ipv6.icmp_sk)
1030 		return -ENOMEM;
1031 
1032 	for_each_possible_cpu(i) {
1033 		err = inet_ctl_sock_create(&sk, PF_INET6,
1034 					   SOCK_RAW, IPPROTO_ICMPV6, net);
1035 		if (err < 0) {
1036 			pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
1037 			       err);
1038 			goto fail;
1039 		}
1040 
1041 		*per_cpu_ptr(net->ipv6.icmp_sk, i) = sk;
1042 
1043 		/* Enough space for 2 64K ICMP packets, including
1044 		 * sk_buff struct overhead.
1045 		 */
1046 		sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
1047 	}
1048 	return 0;
1049 
1050  fail:
1051 	icmpv6_sk_exit(net);
1052 	return err;
1053 }
1054 
1055 static struct pernet_operations icmpv6_sk_ops = {
1056 	.init = icmpv6_sk_init,
1057 	.exit = icmpv6_sk_exit,
1058 };
1059 
1060 int __init icmpv6_init(void)
1061 {
1062 	int err;
1063 
1064 	err = register_pernet_subsys(&icmpv6_sk_ops);
1065 	if (err < 0)
1066 		return err;
1067 
1068 	err = -EAGAIN;
1069 	if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
1070 		goto fail;
1071 
1072 	err = inet6_register_icmp_sender(icmp6_send);
1073 	if (err)
1074 		goto sender_reg_err;
1075 	return 0;
1076 
1077 sender_reg_err:
1078 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
1079 fail:
1080 	pr_err("Failed to register ICMP6 protocol\n");
1081 	unregister_pernet_subsys(&icmpv6_sk_ops);
1082 	return err;
1083 }
1084 
1085 void icmpv6_cleanup(void)
1086 {
1087 	inet6_unregister_icmp_sender(icmp6_send);
1088 	unregister_pernet_subsys(&icmpv6_sk_ops);
1089 	inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
1090 }
1091 
1092 
1093 static const struct icmp6_err {
1094 	int err;
1095 	int fatal;
1096 } tab_unreach[] = {
1097 	{	/* NOROUTE */
1098 		.err	= ENETUNREACH,
1099 		.fatal	= 0,
1100 	},
1101 	{	/* ADM_PROHIBITED */
1102 		.err	= EACCES,
1103 		.fatal	= 1,
1104 	},
1105 	{	/* Was NOT_NEIGHBOUR, now reserved */
1106 		.err	= EHOSTUNREACH,
1107 		.fatal	= 0,
1108 	},
1109 	{	/* ADDR_UNREACH	*/
1110 		.err	= EHOSTUNREACH,
1111 		.fatal	= 0,
1112 	},
1113 	{	/* PORT_UNREACH	*/
1114 		.err	= ECONNREFUSED,
1115 		.fatal	= 1,
1116 	},
1117 	{	/* POLICY_FAIL */
1118 		.err	= EACCES,
1119 		.fatal	= 1,
1120 	},
1121 	{	/* REJECT_ROUTE	*/
1122 		.err	= EACCES,
1123 		.fatal	= 1,
1124 	},
1125 };
1126 
1127 int icmpv6_err_convert(u8 type, u8 code, int *err)
1128 {
1129 	int fatal = 0;
1130 
1131 	*err = EPROTO;
1132 
1133 	switch (type) {
1134 	case ICMPV6_DEST_UNREACH:
1135 		fatal = 1;
1136 		if (code < ARRAY_SIZE(tab_unreach)) {
1137 			*err  = tab_unreach[code].err;
1138 			fatal = tab_unreach[code].fatal;
1139 		}
1140 		break;
1141 
1142 	case ICMPV6_PKT_TOOBIG:
1143 		*err = EMSGSIZE;
1144 		break;
1145 
1146 	case ICMPV6_PARAMPROB:
1147 		*err = EPROTO;
1148 		fatal = 1;
1149 		break;
1150 
1151 	case ICMPV6_TIME_EXCEED:
1152 		*err = EHOSTUNREACH;
1153 		break;
1154 	}
1155 
1156 	return fatal;
1157 }
1158 EXPORT_SYMBOL(icmpv6_err_convert);
1159 
1160 #ifdef CONFIG_SYSCTL
1161 static struct ctl_table ipv6_icmp_table_template[] = {
1162 	{
1163 		.procname	= "ratelimit",
1164 		.data		= &init_net.ipv6.sysctl.icmpv6_time,
1165 		.maxlen		= sizeof(int),
1166 		.mode		= 0644,
1167 		.proc_handler	= proc_dointvec_ms_jiffies,
1168 	},
1169 	{
1170 		.procname	= "echo_ignore_all",
1171 		.data		= &init_net.ipv6.sysctl.icmpv6_echo_ignore_all,
1172 		.maxlen		= sizeof(int),
1173 		.mode		= 0644,
1174 		.proc_handler = proc_dointvec,
1175 	},
1176 	{
1177 		.procname	= "echo_ignore_multicast",
1178 		.data		= &init_net.ipv6.sysctl.icmpv6_echo_ignore_multicast,
1179 		.maxlen		= sizeof(int),
1180 		.mode		= 0644,
1181 		.proc_handler = proc_dointvec,
1182 	},
1183 	{
1184 		.procname	= "echo_ignore_anycast",
1185 		.data		= &init_net.ipv6.sysctl.icmpv6_echo_ignore_anycast,
1186 		.maxlen		= sizeof(int),
1187 		.mode		= 0644,
1188 		.proc_handler = proc_dointvec,
1189 	},
1190 	{
1191 		.procname	= "ratemask",
1192 		.data		= &init_net.ipv6.sysctl.icmpv6_ratemask_ptr,
1193 		.maxlen		= ICMPV6_MSG_MAX + 1,
1194 		.mode		= 0644,
1195 		.proc_handler = proc_do_large_bitmap,
1196 	},
1197 	{ },
1198 };
1199 
1200 struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1201 {
1202 	struct ctl_table *table;
1203 
1204 	table = kmemdup(ipv6_icmp_table_template,
1205 			sizeof(ipv6_icmp_table_template),
1206 			GFP_KERNEL);
1207 
1208 	if (table) {
1209 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
1210 		table[1].data = &net->ipv6.sysctl.icmpv6_echo_ignore_all;
1211 		table[2].data = &net->ipv6.sysctl.icmpv6_echo_ignore_multicast;
1212 		table[3].data = &net->ipv6.sysctl.icmpv6_echo_ignore_anycast;
1213 		table[4].data = &net->ipv6.sysctl.icmpv6_ratemask_ptr;
1214 	}
1215 	return table;
1216 }
1217 #endif
1218