Loading...
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#include <linux/slab.h>
44
45#ifdef CONFIG_SYSCTL
46#include <linux/sysctl.h>
47#endif
48
49#include <linux/inet.h>
50#include <linux/netdevice.h>
51#include <linux/icmpv6.h>
52
53#include <net/ip.h>
54#include <net/sock.h>
55
56#include <net/ipv6.h>
57#include <net/ip6_checksum.h>
58#include <net/protocol.h>
59#include <net/raw.h>
60#include <net/rawv6.h>
61#include <net/transp_v6.h>
62#include <net/ip6_route.h>
63#include <net/addrconf.h>
64#include <net/icmp.h>
65#include <net/xfrm.h>
66#include <net/inet_common.h>
67
68#include <asm/uaccess.h>
69#include <asm/system.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 */
78static inline struct sock *icmpv6_sk(struct net *net)
79{
80 return net->ipv6.icmp_sk[smp_processor_id()];
81}
82
83static int icmpv6_rcv(struct sk_buff *skb);
84
85static const struct inet6_protocol icmpv6_protocol = {
86 .handler = icmpv6_rcv,
87 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
88};
89
90static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
91{
92 struct sock *sk;
93
94 local_bh_disable();
95
96 sk = icmpv6_sk(net);
97 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
98 /* This can happen if the output path (f.e. SIT or
99 * ip6ip6 tunnel) signals dst_link_failure() for an
100 * outgoing ICMP6 packet.
101 */
102 local_bh_enable();
103 return NULL;
104 }
105 return sk;
106}
107
108static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
109{
110 spin_unlock_bh(&sk->sk_lock.slock);
111}
112
113/*
114 * Slightly more convenient version of icmpv6_send.
115 */
116void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
117{
118 icmpv6_send(skb, ICMPV6_PARAMPROB, code, pos);
119 kfree_skb(skb);
120}
121
122/*
123 * Figure out, may we reply to this packet with icmp error.
124 *
125 * We do not reply, if:
126 * - it was icmp error message.
127 * - it is truncated, so that it is known, that protocol is ICMPV6
128 * (i.e. in the middle of some exthdr)
129 *
130 * --ANK (980726)
131 */
132
133static int is_ineligible(struct sk_buff *skb)
134{
135 int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
136 int len = skb->len - ptr;
137 __u8 nexthdr = ipv6_hdr(skb)->nexthdr;
138
139 if (len < 0)
140 return 1;
141
142 ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr);
143 if (ptr < 0)
144 return 0;
145 if (nexthdr == IPPROTO_ICMPV6) {
146 u8 _type, *tp;
147 tp = skb_header_pointer(skb,
148 ptr+offsetof(struct icmp6hdr, icmp6_type),
149 sizeof(_type), &_type);
150 if (tp == NULL ||
151 !(*tp & ICMPV6_INFOMSG_MASK))
152 return 1;
153 }
154 return 0;
155}
156
157/*
158 * Check the ICMP output rate limit
159 */
160static inline bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
161 struct flowi6 *fl6)
162{
163 struct dst_entry *dst;
164 struct net *net = sock_net(sk);
165 bool res = false;
166
167 /* Informational messages are not limited. */
168 if (type & ICMPV6_INFOMSG_MASK)
169 return true;
170
171 /* Do not limit pmtu discovery, it would break it. */
172 if (type == ICMPV6_PKT_TOOBIG)
173 return true;
174
175 /*
176 * Look up the output route.
177 * XXX: perhaps the expire for routing entries cloned by
178 * this lookup should be more aggressive (not longer than timeout).
179 */
180 dst = ip6_route_output(net, sk, fl6);
181 if (dst->error) {
182 IP6_INC_STATS(net, ip6_dst_idev(dst),
183 IPSTATS_MIB_OUTNOROUTES);
184 } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
185 res = true;
186 } else {
187 struct rt6_info *rt = (struct rt6_info *)dst;
188 int tmo = net->ipv6.sysctl.icmpv6_time;
189
190 /* Give more bandwidth to wider prefixes. */
191 if (rt->rt6i_dst.plen < 128)
192 tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
193
194 if (!rt->rt6i_peer)
195 rt6_bind_peer(rt, 1);
196 res = inet_peer_xrlim_allow(rt->rt6i_peer, tmo);
197 }
198 dst_release(dst);
199 return res;
200}
201
202/*
203 * an inline helper for the "simple" if statement below
204 * checks if parameter problem report is caused by an
205 * unrecognized IPv6 option that has the Option Type
206 * highest-order two bits set to 10
207 */
208
209static __inline__ int opt_unrec(struct sk_buff *skb, __u32 offset)
210{
211 u8 _optval, *op;
212
213 offset += skb_network_offset(skb);
214 op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
215 if (op == NULL)
216 return 1;
217 return (*op & 0xC0) == 0x80;
218}
219
220static int icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, struct icmp6hdr *thdr, int len)
221{
222 struct sk_buff *skb;
223 struct icmp6hdr *icmp6h;
224 int err = 0;
225
226 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
227 goto out;
228
229 icmp6h = icmp6_hdr(skb);
230 memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
231 icmp6h->icmp6_cksum = 0;
232
233 if (skb_queue_len(&sk->sk_write_queue) == 1) {
234 skb->csum = csum_partial(icmp6h,
235 sizeof(struct icmp6hdr), skb->csum);
236 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
237 &fl6->daddr,
238 len, fl6->flowi6_proto,
239 skb->csum);
240 } else {
241 __wsum tmp_csum = 0;
242
243 skb_queue_walk(&sk->sk_write_queue, skb) {
244 tmp_csum = csum_add(tmp_csum, skb->csum);
245 }
246
247 tmp_csum = csum_partial(icmp6h,
248 sizeof(struct icmp6hdr), tmp_csum);
249 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
250 &fl6->daddr,
251 len, fl6->flowi6_proto,
252 tmp_csum);
253 }
254 ip6_push_pending_frames(sk);
255out:
256 return err;
257}
258
259struct icmpv6_msg {
260 struct sk_buff *skb;
261 int offset;
262 uint8_t type;
263};
264
265static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
266{
267 struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
268 struct sk_buff *org_skb = msg->skb;
269 __wsum csum = 0;
270
271 csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
272 to, len, csum);
273 skb->csum = csum_block_add(skb->csum, csum, odd);
274 if (!(msg->type & ICMPV6_INFOMSG_MASK))
275 nf_ct_attach(skb, org_skb);
276 return 0;
277}
278
279#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
280static void mip6_addr_swap(struct sk_buff *skb)
281{
282 struct ipv6hdr *iph = ipv6_hdr(skb);
283 struct inet6_skb_parm *opt = IP6CB(skb);
284 struct ipv6_destopt_hao *hao;
285 struct in6_addr tmp;
286 int off;
287
288 if (opt->dsthao) {
289 off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
290 if (likely(off >= 0)) {
291 hao = (struct ipv6_destopt_hao *)
292 (skb_network_header(skb) + off);
293 ipv6_addr_copy(&tmp, &iph->saddr);
294 ipv6_addr_copy(&iph->saddr, &hao->addr);
295 ipv6_addr_copy(&hao->addr, &tmp);
296 }
297 }
298}
299#else
300static inline void mip6_addr_swap(struct sk_buff *skb) {}
301#endif
302
303static struct dst_entry *icmpv6_route_lookup(struct net *net, struct sk_buff *skb,
304 struct sock *sk, struct flowi6 *fl6)
305{
306 struct dst_entry *dst, *dst2;
307 struct flowi6 fl2;
308 int err;
309
310 err = ip6_dst_lookup(sk, &dst, fl6);
311 if (err)
312 return ERR_PTR(err);
313
314 /*
315 * We won't send icmp if the destination is known
316 * anycast.
317 */
318 if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) {
319 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: acast source\n");
320 dst_release(dst);
321 return ERR_PTR(-EINVAL);
322 }
323
324 /* No need to clone since we're just using its address. */
325 dst2 = dst;
326
327 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
328 if (!IS_ERR(dst)) {
329 if (dst != dst2)
330 return dst;
331 } else {
332 if (PTR_ERR(dst) == -EPERM)
333 dst = NULL;
334 else
335 return dst;
336 }
337
338 err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
339 if (err)
340 goto relookup_failed;
341
342 err = ip6_dst_lookup(sk, &dst2, &fl2);
343 if (err)
344 goto relookup_failed;
345
346 dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
347 if (!IS_ERR(dst2)) {
348 dst_release(dst);
349 dst = dst2;
350 } else {
351 err = PTR_ERR(dst2);
352 if (err == -EPERM) {
353 dst_release(dst);
354 return dst2;
355 } else
356 goto relookup_failed;
357 }
358
359relookup_failed:
360 if (dst)
361 return dst;
362 return ERR_PTR(err);
363}
364
365/*
366 * Send an ICMP message in response to a packet in error
367 */
368void icmpv6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info)
369{
370 struct net *net = dev_net(skb->dev);
371 struct inet6_dev *idev = NULL;
372 struct ipv6hdr *hdr = ipv6_hdr(skb);
373 struct sock *sk;
374 struct ipv6_pinfo *np;
375 const struct in6_addr *saddr = NULL;
376 struct dst_entry *dst;
377 struct icmp6hdr tmp_hdr;
378 struct flowi6 fl6;
379 struct icmpv6_msg msg;
380 int iif = 0;
381 int addr_type = 0;
382 int len;
383 int hlimit;
384 int err = 0;
385
386 if ((u8 *)hdr < skb->head ||
387 (skb->network_header + sizeof(*hdr)) > skb->tail)
388 return;
389
390 /*
391 * Make sure we respect the rules
392 * i.e. RFC 1885 2.4(e)
393 * Rule (e.1) is enforced by not using icmpv6_send
394 * in any code that processes icmp errors.
395 */
396 addr_type = ipv6_addr_type(&hdr->daddr);
397
398 if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0))
399 saddr = &hdr->daddr;
400
401 /*
402 * Dest addr check
403 */
404
405 if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) {
406 if (type != ICMPV6_PKT_TOOBIG &&
407 !(type == ICMPV6_PARAMPROB &&
408 code == ICMPV6_UNK_OPTION &&
409 (opt_unrec(skb, info))))
410 return;
411
412 saddr = NULL;
413 }
414
415 addr_type = ipv6_addr_type(&hdr->saddr);
416
417 /*
418 * Source addr check
419 */
420
421 if (addr_type & IPV6_ADDR_LINKLOCAL)
422 iif = skb->dev->ifindex;
423
424 /*
425 * Must not send error if the source does not uniquely
426 * identify a single node (RFC2463 Section 2.4).
427 * We check unspecified / multicast addresses here,
428 * and anycast addresses will be checked later.
429 */
430 if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
431 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: addr_any/mcast source\n");
432 return;
433 }
434
435 /*
436 * Never answer to a ICMP packet.
437 */
438 if (is_ineligible(skb)) {
439 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: no reply to icmp error\n");
440 return;
441 }
442
443 mip6_addr_swap(skb);
444
445 memset(&fl6, 0, sizeof(fl6));
446 fl6.flowi6_proto = IPPROTO_ICMPV6;
447 ipv6_addr_copy(&fl6.daddr, &hdr->saddr);
448 if (saddr)
449 ipv6_addr_copy(&fl6.saddr, saddr);
450 fl6.flowi6_oif = iif;
451 fl6.fl6_icmp_type = type;
452 fl6.fl6_icmp_code = code;
453 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
454
455 sk = icmpv6_xmit_lock(net);
456 if (sk == NULL)
457 return;
458 np = inet6_sk(sk);
459
460 if (!icmpv6_xrlim_allow(sk, type, &fl6))
461 goto out;
462
463 tmp_hdr.icmp6_type = type;
464 tmp_hdr.icmp6_code = code;
465 tmp_hdr.icmp6_cksum = 0;
466 tmp_hdr.icmp6_pointer = htonl(info);
467
468 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
469 fl6.flowi6_oif = np->mcast_oif;
470
471 dst = icmpv6_route_lookup(net, skb, sk, &fl6);
472 if (IS_ERR(dst))
473 goto out;
474
475 if (ipv6_addr_is_multicast(&fl6.daddr))
476 hlimit = np->mcast_hops;
477 else
478 hlimit = np->hop_limit;
479 if (hlimit < 0)
480 hlimit = ip6_dst_hoplimit(dst);
481
482 msg.skb = skb;
483 msg.offset = skb_network_offset(skb);
484 msg.type = type;
485
486 len = skb->len - msg.offset;
487 len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr));
488 if (len < 0) {
489 LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n");
490 goto out_dst_release;
491 }
492
493 idev = in6_dev_get(skb->dev);
494
495 err = ip6_append_data(sk, icmpv6_getfrag, &msg,
496 len + sizeof(struct icmp6hdr),
497 sizeof(struct icmp6hdr), hlimit,
498 np->tclass, NULL, &fl6, (struct rt6_info*)dst,
499 MSG_DONTWAIT, np->dontfrag);
500 if (err) {
501 ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTERRORS);
502 ip6_flush_pending_frames(sk);
503 goto out_put;
504 }
505 err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr, len + sizeof(struct icmp6hdr));
506
507out_put:
508 if (likely(idev != NULL))
509 in6_dev_put(idev);
510out_dst_release:
511 dst_release(dst);
512out:
513 icmpv6_xmit_unlock(sk);
514}
515
516EXPORT_SYMBOL(icmpv6_send);
517
518static void icmpv6_echo_reply(struct sk_buff *skb)
519{
520 struct net *net = dev_net(skb->dev);
521 struct sock *sk;
522 struct inet6_dev *idev;
523 struct ipv6_pinfo *np;
524 const struct in6_addr *saddr = NULL;
525 struct icmp6hdr *icmph = icmp6_hdr(skb);
526 struct icmp6hdr tmp_hdr;
527 struct flowi6 fl6;
528 struct icmpv6_msg msg;
529 struct dst_entry *dst;
530 int err = 0;
531 int hlimit;
532
533 saddr = &ipv6_hdr(skb)->daddr;
534
535 if (!ipv6_unicast_destination(skb))
536 saddr = NULL;
537
538 memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
539 tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
540
541 memset(&fl6, 0, sizeof(fl6));
542 fl6.flowi6_proto = IPPROTO_ICMPV6;
543 ipv6_addr_copy(&fl6.daddr, &ipv6_hdr(skb)->saddr);
544 if (saddr)
545 ipv6_addr_copy(&fl6.saddr, saddr);
546 fl6.flowi6_oif = skb->dev->ifindex;
547 fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
548 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
549
550 sk = icmpv6_xmit_lock(net);
551 if (sk == NULL)
552 return;
553 np = inet6_sk(sk);
554
555 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
556 fl6.flowi6_oif = np->mcast_oif;
557
558 err = ip6_dst_lookup(sk, &dst, &fl6);
559 if (err)
560 goto out;
561 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
562 if (IS_ERR(dst))
563 goto out;
564
565 if (ipv6_addr_is_multicast(&fl6.daddr))
566 hlimit = np->mcast_hops;
567 else
568 hlimit = np->hop_limit;
569 if (hlimit < 0)
570 hlimit = ip6_dst_hoplimit(dst);
571
572 idev = in6_dev_get(skb->dev);
573
574 msg.skb = skb;
575 msg.offset = 0;
576 msg.type = ICMPV6_ECHO_REPLY;
577
578 err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
579 sizeof(struct icmp6hdr), hlimit, np->tclass, NULL, &fl6,
580 (struct rt6_info*)dst, MSG_DONTWAIT,
581 np->dontfrag);
582
583 if (err) {
584 ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTERRORS);
585 ip6_flush_pending_frames(sk);
586 goto out_put;
587 }
588 err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr, skb->len + sizeof(struct icmp6hdr));
589
590out_put:
591 if (likely(idev != NULL))
592 in6_dev_put(idev);
593 dst_release(dst);
594out:
595 icmpv6_xmit_unlock(sk);
596}
597
598static void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
599{
600 const struct inet6_protocol *ipprot;
601 int inner_offset;
602 int hash;
603 u8 nexthdr;
604
605 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
606 return;
607
608 nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
609 if (ipv6_ext_hdr(nexthdr)) {
610 /* now skip over extension headers */
611 inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr);
612 if (inner_offset<0)
613 return;
614 } else {
615 inner_offset = sizeof(struct ipv6hdr);
616 }
617
618 /* Checkin header including 8 bytes of inner protocol header. */
619 if (!pskb_may_pull(skb, inner_offset+8))
620 return;
621
622 /* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
623 Without this we will not able f.e. to make source routed
624 pmtu discovery.
625 Corresponding argument (opt) to notifiers is already added.
626 --ANK (980726)
627 */
628
629 hash = nexthdr & (MAX_INET_PROTOS - 1);
630
631 rcu_read_lock();
632 ipprot = rcu_dereference(inet6_protos[hash]);
633 if (ipprot && ipprot->err_handler)
634 ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
635 rcu_read_unlock();
636
637 raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
638}
639
640/*
641 * Handle icmp messages
642 */
643
644static int icmpv6_rcv(struct sk_buff *skb)
645{
646 struct net_device *dev = skb->dev;
647 struct inet6_dev *idev = __in6_dev_get(dev);
648 const struct in6_addr *saddr, *daddr;
649 const struct ipv6hdr *orig_hdr;
650 struct icmp6hdr *hdr;
651 u8 type;
652
653 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
654 struct sec_path *sp = skb_sec_path(skb);
655 int nh;
656
657 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
658 XFRM_STATE_ICMP))
659 goto drop_no_count;
660
661 if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(*orig_hdr)))
662 goto drop_no_count;
663
664 nh = skb_network_offset(skb);
665 skb_set_network_header(skb, sizeof(*hdr));
666
667 if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
668 goto drop_no_count;
669
670 skb_set_network_header(skb, nh);
671 }
672
673 ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INMSGS);
674
675 saddr = &ipv6_hdr(skb)->saddr;
676 daddr = &ipv6_hdr(skb)->daddr;
677
678 /* Perform checksum. */
679 switch (skb->ip_summed) {
680 case CHECKSUM_COMPLETE:
681 if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6,
682 skb->csum))
683 break;
684 /* fall through */
685 case CHECKSUM_NONE:
686 skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len,
687 IPPROTO_ICMPV6, 0));
688 if (__skb_checksum_complete(skb)) {
689 LIMIT_NETDEBUG(KERN_DEBUG "ICMPv6 checksum failed [%pI6 > %pI6]\n",
690 saddr, daddr);
691 goto discard_it;
692 }
693 }
694
695 if (!pskb_pull(skb, sizeof(*hdr)))
696 goto discard_it;
697
698 hdr = icmp6_hdr(skb);
699
700 type = hdr->icmp6_type;
701
702 ICMP6MSGIN_INC_STATS_BH(dev_net(dev), idev, type);
703
704 switch (type) {
705 case ICMPV6_ECHO_REQUEST:
706 icmpv6_echo_reply(skb);
707 break;
708
709 case ICMPV6_ECHO_REPLY:
710 /* we couldn't care less */
711 break;
712
713 case ICMPV6_PKT_TOOBIG:
714 /* BUGGG_FUTURE: if packet contains rthdr, we cannot update
715 standard destination cache. Seems, only "advanced"
716 destination cache will allow to solve this problem
717 --ANK (980726)
718 */
719 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
720 goto discard_it;
721 hdr = icmp6_hdr(skb);
722 orig_hdr = (struct ipv6hdr *) (hdr + 1);
723 rt6_pmtu_discovery(&orig_hdr->daddr, &orig_hdr->saddr, dev,
724 ntohl(hdr->icmp6_mtu));
725
726 /*
727 * Drop through to notify
728 */
729
730 case ICMPV6_DEST_UNREACH:
731 case ICMPV6_TIME_EXCEED:
732 case ICMPV6_PARAMPROB:
733 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
734 break;
735
736 case NDISC_ROUTER_SOLICITATION:
737 case NDISC_ROUTER_ADVERTISEMENT:
738 case NDISC_NEIGHBOUR_SOLICITATION:
739 case NDISC_NEIGHBOUR_ADVERTISEMENT:
740 case NDISC_REDIRECT:
741 ndisc_rcv(skb);
742 break;
743
744 case ICMPV6_MGM_QUERY:
745 igmp6_event_query(skb);
746 break;
747
748 case ICMPV6_MGM_REPORT:
749 igmp6_event_report(skb);
750 break;
751
752 case ICMPV6_MGM_REDUCTION:
753 case ICMPV6_NI_QUERY:
754 case ICMPV6_NI_REPLY:
755 case ICMPV6_MLD2_REPORT:
756 case ICMPV6_DHAAD_REQUEST:
757 case ICMPV6_DHAAD_REPLY:
758 case ICMPV6_MOBILE_PREFIX_SOL:
759 case ICMPV6_MOBILE_PREFIX_ADV:
760 break;
761
762 default:
763 LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n");
764
765 /* informational */
766 if (type & ICMPV6_INFOMSG_MASK)
767 break;
768
769 /*
770 * error of unknown type.
771 * must pass to upper level
772 */
773
774 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
775 }
776
777 kfree_skb(skb);
778 return 0;
779
780discard_it:
781 ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INERRORS);
782drop_no_count:
783 kfree_skb(skb);
784 return 0;
785}
786
787void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
788 u8 type,
789 const struct in6_addr *saddr,
790 const struct in6_addr *daddr,
791 int oif)
792{
793 memset(fl6, 0, sizeof(*fl6));
794 ipv6_addr_copy(&fl6->saddr, saddr);
795 ipv6_addr_copy(&fl6->daddr, daddr);
796 fl6->flowi6_proto = IPPROTO_ICMPV6;
797 fl6->fl6_icmp_type = type;
798 fl6->fl6_icmp_code = 0;
799 fl6->flowi6_oif = oif;
800 security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
801}
802
803/*
804 * Special lock-class for __icmpv6_sk:
805 */
806static struct lock_class_key icmpv6_socket_sk_dst_lock_key;
807
808static int __net_init icmpv6_sk_init(struct net *net)
809{
810 struct sock *sk;
811 int err, i, j;
812
813 net->ipv6.icmp_sk =
814 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
815 if (net->ipv6.icmp_sk == NULL)
816 return -ENOMEM;
817
818 for_each_possible_cpu(i) {
819 err = inet_ctl_sock_create(&sk, PF_INET6,
820 SOCK_RAW, IPPROTO_ICMPV6, net);
821 if (err < 0) {
822 printk(KERN_ERR
823 "Failed to initialize the ICMP6 control socket "
824 "(err %d).\n",
825 err);
826 goto fail;
827 }
828
829 net->ipv6.icmp_sk[i] = sk;
830
831 /*
832 * Split off their lock-class, because sk->sk_dst_lock
833 * gets used from softirqs, which is safe for
834 * __icmpv6_sk (because those never get directly used
835 * via userspace syscalls), but unsafe for normal sockets.
836 */
837 lockdep_set_class(&sk->sk_dst_lock,
838 &icmpv6_socket_sk_dst_lock_key);
839
840 /* Enough space for 2 64K ICMP packets, including
841 * sk_buff struct overhead.
842 */
843 sk->sk_sndbuf =
844 (2 * ((64 * 1024) + sizeof(struct sk_buff)));
845 }
846 return 0;
847
848 fail:
849 for (j = 0; j < i; j++)
850 inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
851 kfree(net->ipv6.icmp_sk);
852 return err;
853}
854
855static void __net_exit icmpv6_sk_exit(struct net *net)
856{
857 int i;
858
859 for_each_possible_cpu(i) {
860 inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
861 }
862 kfree(net->ipv6.icmp_sk);
863}
864
865static struct pernet_operations icmpv6_sk_ops = {
866 .init = icmpv6_sk_init,
867 .exit = icmpv6_sk_exit,
868};
869
870int __init icmpv6_init(void)
871{
872 int err;
873
874 err = register_pernet_subsys(&icmpv6_sk_ops);
875 if (err < 0)
876 return err;
877
878 err = -EAGAIN;
879 if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
880 goto fail;
881 return 0;
882
883fail:
884 printk(KERN_ERR "Failed to register ICMP6 protocol\n");
885 unregister_pernet_subsys(&icmpv6_sk_ops);
886 return err;
887}
888
889void icmpv6_cleanup(void)
890{
891 unregister_pernet_subsys(&icmpv6_sk_ops);
892 inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
893}
894
895
896static const struct icmp6_err {
897 int err;
898 int fatal;
899} tab_unreach[] = {
900 { /* NOROUTE */
901 .err = ENETUNREACH,
902 .fatal = 0,
903 },
904 { /* ADM_PROHIBITED */
905 .err = EACCES,
906 .fatal = 1,
907 },
908 { /* Was NOT_NEIGHBOUR, now reserved */
909 .err = EHOSTUNREACH,
910 .fatal = 0,
911 },
912 { /* ADDR_UNREACH */
913 .err = EHOSTUNREACH,
914 .fatal = 0,
915 },
916 { /* PORT_UNREACH */
917 .err = ECONNREFUSED,
918 .fatal = 1,
919 },
920};
921
922int icmpv6_err_convert(u8 type, u8 code, int *err)
923{
924 int fatal = 0;
925
926 *err = EPROTO;
927
928 switch (type) {
929 case ICMPV6_DEST_UNREACH:
930 fatal = 1;
931 if (code <= ICMPV6_PORT_UNREACH) {
932 *err = tab_unreach[code].err;
933 fatal = tab_unreach[code].fatal;
934 }
935 break;
936
937 case ICMPV6_PKT_TOOBIG:
938 *err = EMSGSIZE;
939 break;
940
941 case ICMPV6_PARAMPROB:
942 *err = EPROTO;
943 fatal = 1;
944 break;
945
946 case ICMPV6_TIME_EXCEED:
947 *err = EHOSTUNREACH;
948 break;
949 }
950
951 return fatal;
952}
953
954EXPORT_SYMBOL(icmpv6_err_convert);
955
956#ifdef CONFIG_SYSCTL
957ctl_table ipv6_icmp_table_template[] = {
958 {
959 .procname = "ratelimit",
960 .data = &init_net.ipv6.sysctl.icmpv6_time,
961 .maxlen = sizeof(int),
962 .mode = 0644,
963 .proc_handler = proc_dointvec_ms_jiffies,
964 },
965 { },
966};
967
968struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
969{
970 struct ctl_table *table;
971
972 table = kmemdup(ipv6_icmp_table_template,
973 sizeof(ipv6_icmp_table_template),
974 GFP_KERNEL);
975
976 if (table)
977 table[0].data = &net->ipv6.sysctl.icmpv6_time;
978
979 return table;
980}
981#endif
982
1/*
2 * Internet Control Message Protocol (ICMPv6)
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Based on net/ipv4/icmp.c
9 *
10 * RFC 1885
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
16 */
17
18/*
19 * Changes:
20 *
21 * Andi Kleen : exception handling
22 * Andi Kleen add rate limits. never reply to a icmp.
23 * add more length checks and other fixes.
24 * yoshfuji : ensure to sent parameter problem for
25 * fragments.
26 * YOSHIFUJI Hideaki @USAGI: added sysctl for icmp rate limit.
27 * Randy Dunlap and
28 * YOSHIFUJI Hideaki @USAGI: Per-interface statistics support
29 * Kazunori MIYAZAWA @USAGI: change output process to use ip6_append_data
30 */
31
32#define pr_fmt(fmt) "IPv6: " fmt
33
34#include <linux/module.h>
35#include <linux/errno.h>
36#include <linux/types.h>
37#include <linux/socket.h>
38#include <linux/in.h>
39#include <linux/kernel.h>
40#include <linux/sockios.h>
41#include <linux/net.h>
42#include <linux/skbuff.h>
43#include <linux/init.h>
44#include <linux/netfilter.h>
45#include <linux/slab.h>
46
47#ifdef CONFIG_SYSCTL
48#include <linux/sysctl.h>
49#endif
50
51#include <linux/inet.h>
52#include <linux/netdevice.h>
53#include <linux/icmpv6.h>
54
55#include <net/ip.h>
56#include <net/sock.h>
57
58#include <net/ipv6.h>
59#include <net/ip6_checksum.h>
60#include <net/ping.h>
61#include <net/protocol.h>
62#include <net/raw.h>
63#include <net/rawv6.h>
64#include <net/transp_v6.h>
65#include <net/ip6_route.h>
66#include <net/addrconf.h>
67#include <net/icmp.h>
68#include <net/xfrm.h>
69#include <net/inet_common.h>
70#include <net/dsfield.h>
71#include <net/l3mdev.h>
72
73#include <linux/uaccess.h>
74
75/*
76 * The ICMP socket(s). This is the most convenient way to flow control
77 * our ICMP output as well as maintain a clean interface throughout
78 * all layers. All Socketless IP sends will soon be gone.
79 *
80 * On SMP we have one ICMP socket per-cpu.
81 */
82static inline struct sock *icmpv6_sk(struct net *net)
83{
84 return net->ipv6.icmp_sk[smp_processor_id()];
85}
86
87static void icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
88 u8 type, u8 code, int offset, __be32 info)
89{
90 /* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */
91 struct icmp6hdr *icmp6 = (struct icmp6hdr *) (skb->data + offset);
92 struct net *net = dev_net(skb->dev);
93
94 if (type == ICMPV6_PKT_TOOBIG)
95 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
96 else if (type == NDISC_REDIRECT)
97 ip6_redirect(skb, net, skb->dev->ifindex, 0,
98 sock_net_uid(net, NULL));
99
100 if (!(type & ICMPV6_INFOMSG_MASK))
101 if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST)
102 ping_err(skb, offset, ntohl(info));
103}
104
105static int icmpv6_rcv(struct sk_buff *skb);
106
107static const struct inet6_protocol icmpv6_protocol = {
108 .handler = icmpv6_rcv,
109 .err_handler = icmpv6_err,
110 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
111};
112
113static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
114{
115 struct sock *sk;
116
117 local_bh_disable();
118
119 sk = icmpv6_sk(net);
120 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
121 /* This can happen if the output path (f.e. SIT or
122 * ip6ip6 tunnel) signals dst_link_failure() for an
123 * outgoing ICMP6 packet.
124 */
125 local_bh_enable();
126 return NULL;
127 }
128 return sk;
129}
130
131static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
132{
133 spin_unlock_bh(&sk->sk_lock.slock);
134}
135
136/*
137 * Figure out, may we reply to this packet with icmp error.
138 *
139 * We do not reply, if:
140 * - it was icmp error message.
141 * - it is truncated, so that it is known, that protocol is ICMPV6
142 * (i.e. in the middle of some exthdr)
143 *
144 * --ANK (980726)
145 */
146
147static bool is_ineligible(const struct sk_buff *skb)
148{
149 int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
150 int len = skb->len - ptr;
151 __u8 nexthdr = ipv6_hdr(skb)->nexthdr;
152 __be16 frag_off;
153
154 if (len < 0)
155 return true;
156
157 ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off);
158 if (ptr < 0)
159 return false;
160 if (nexthdr == IPPROTO_ICMPV6) {
161 u8 _type, *tp;
162 tp = skb_header_pointer(skb,
163 ptr+offsetof(struct icmp6hdr, icmp6_type),
164 sizeof(_type), &_type);
165 if (!tp || !(*tp & ICMPV6_INFOMSG_MASK))
166 return true;
167 }
168 return false;
169}
170
171/*
172 * Check the ICMP output rate limit
173 */
174static bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
175 struct flowi6 *fl6)
176{
177 struct net *net = sock_net(sk);
178 struct dst_entry *dst;
179 bool res = false;
180
181 /* Informational messages are not limited. */
182 if (type & ICMPV6_INFOMSG_MASK)
183 return true;
184
185 /* Do not limit pmtu discovery, it would break it. */
186 if (type == ICMPV6_PKT_TOOBIG)
187 return true;
188
189 /*
190 * Look up the output route.
191 * XXX: perhaps the expire for routing entries cloned by
192 * this lookup should be more aggressive (not longer than timeout).
193 */
194 dst = ip6_route_output(net, sk, fl6);
195 if (dst->error) {
196 IP6_INC_STATS(net, ip6_dst_idev(dst),
197 IPSTATS_MIB_OUTNOROUTES);
198 } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
199 res = true;
200 } else {
201 struct rt6_info *rt = (struct rt6_info *)dst;
202 int tmo = net->ipv6.sysctl.icmpv6_time;
203
204 /* Give more bandwidth to wider prefixes. */
205 if (rt->rt6i_dst.plen < 128)
206 tmo >>= ((128 - rt->rt6i_dst.plen)>>5);
207
208 if (icmp_global_allow()) {
209 struct inet_peer *peer;
210
211 peer = inet_getpeer_v6(net->ipv6.peers,
212 &fl6->daddr, 1);
213 res = inet_peer_xrlim_allow(peer, tmo);
214 if (peer)
215 inet_putpeer(peer);
216 }
217 }
218 dst_release(dst);
219 return res;
220}
221
222/*
223 * an inline helper for the "simple" if statement below
224 * checks if parameter problem report is caused by an
225 * unrecognized IPv6 option that has the Option Type
226 * highest-order two bits set to 10
227 */
228
229static bool opt_unrec(struct sk_buff *skb, __u32 offset)
230{
231 u8 _optval, *op;
232
233 offset += skb_network_offset(skb);
234 op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
235 if (!op)
236 return true;
237 return (*op & 0xC0) == 0x80;
238}
239
240int icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
241 struct icmp6hdr *thdr, int len)
242{
243 struct sk_buff *skb;
244 struct icmp6hdr *icmp6h;
245 int err = 0;
246
247 skb = skb_peek(&sk->sk_write_queue);
248 if (!skb)
249 goto out;
250
251 icmp6h = icmp6_hdr(skb);
252 memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
253 icmp6h->icmp6_cksum = 0;
254
255 if (skb_queue_len(&sk->sk_write_queue) == 1) {
256 skb->csum = csum_partial(icmp6h,
257 sizeof(struct icmp6hdr), skb->csum);
258 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
259 &fl6->daddr,
260 len, fl6->flowi6_proto,
261 skb->csum);
262 } else {
263 __wsum tmp_csum = 0;
264
265 skb_queue_walk(&sk->sk_write_queue, skb) {
266 tmp_csum = csum_add(tmp_csum, skb->csum);
267 }
268
269 tmp_csum = csum_partial(icmp6h,
270 sizeof(struct icmp6hdr), tmp_csum);
271 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
272 &fl6->daddr,
273 len, fl6->flowi6_proto,
274 tmp_csum);
275 }
276 ip6_push_pending_frames(sk);
277out:
278 return err;
279}
280
281struct icmpv6_msg {
282 struct sk_buff *skb;
283 int offset;
284 uint8_t type;
285};
286
287static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
288{
289 struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
290 struct sk_buff *org_skb = msg->skb;
291 __wsum csum = 0;
292
293 csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
294 to, len, csum);
295 skb->csum = csum_block_add(skb->csum, csum, odd);
296 if (!(msg->type & ICMPV6_INFOMSG_MASK))
297 nf_ct_attach(skb, org_skb);
298 return 0;
299}
300
301#if IS_ENABLED(CONFIG_IPV6_MIP6)
302static void mip6_addr_swap(struct sk_buff *skb)
303{
304 struct ipv6hdr *iph = ipv6_hdr(skb);
305 struct inet6_skb_parm *opt = IP6CB(skb);
306 struct ipv6_destopt_hao *hao;
307 struct in6_addr tmp;
308 int off;
309
310 if (opt->dsthao) {
311 off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
312 if (likely(off >= 0)) {
313 hao = (struct ipv6_destopt_hao *)
314 (skb_network_header(skb) + off);
315 tmp = iph->saddr;
316 iph->saddr = hao->addr;
317 hao->addr = tmp;
318 }
319 }
320}
321#else
322static inline void mip6_addr_swap(struct sk_buff *skb) {}
323#endif
324
325static struct dst_entry *icmpv6_route_lookup(struct net *net,
326 struct sk_buff *skb,
327 struct sock *sk,
328 struct flowi6 *fl6)
329{
330 struct dst_entry *dst, *dst2;
331 struct flowi6 fl2;
332 int err;
333
334 err = ip6_dst_lookup(net, sk, &dst, fl6);
335 if (err)
336 return ERR_PTR(err);
337
338 /*
339 * We won't send icmp if the destination is known
340 * anycast.
341 */
342 if (ipv6_anycast_destination(dst, &fl6->daddr)) {
343 net_dbg_ratelimited("icmp6_send: acast source\n");
344 dst_release(dst);
345 return ERR_PTR(-EINVAL);
346 }
347
348 /* No need to clone since we're just using its address. */
349 dst2 = dst;
350
351 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
352 if (!IS_ERR(dst)) {
353 if (dst != dst2)
354 return dst;
355 } else {
356 if (PTR_ERR(dst) == -EPERM)
357 dst = NULL;
358 else
359 return dst;
360 }
361
362 err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
363 if (err)
364 goto relookup_failed;
365
366 err = ip6_dst_lookup(net, sk, &dst2, &fl2);
367 if (err)
368 goto relookup_failed;
369
370 dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
371 if (!IS_ERR(dst2)) {
372 dst_release(dst);
373 dst = dst2;
374 } else {
375 err = PTR_ERR(dst2);
376 if (err == -EPERM) {
377 dst_release(dst);
378 return dst2;
379 } else
380 goto relookup_failed;
381 }
382
383relookup_failed:
384 if (dst)
385 return dst;
386 return ERR_PTR(err);
387}
388
389/*
390 * Send an ICMP message in response to a packet in error
391 */
392static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info,
393 const struct in6_addr *force_saddr)
394{
395 struct net *net = dev_net(skb->dev);
396 struct inet6_dev *idev = NULL;
397 struct ipv6hdr *hdr = ipv6_hdr(skb);
398 struct sock *sk;
399 struct ipv6_pinfo *np;
400 const struct in6_addr *saddr = NULL;
401 struct dst_entry *dst;
402 struct icmp6hdr tmp_hdr;
403 struct flowi6 fl6;
404 struct icmpv6_msg msg;
405 struct sockcm_cookie sockc_unused = {0};
406 struct ipcm6_cookie ipc6;
407 int iif = 0;
408 int addr_type = 0;
409 int len;
410 int err = 0;
411 u32 mark = IP6_REPLY_MARK(net, skb->mark);
412
413 if ((u8 *)hdr < skb->head ||
414 (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb))
415 return;
416
417 /*
418 * Make sure we respect the rules
419 * i.e. RFC 1885 2.4(e)
420 * Rule (e.1) is enforced by not using icmp6_send
421 * in any code that processes icmp errors.
422 */
423 addr_type = ipv6_addr_type(&hdr->daddr);
424
425 if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) ||
426 ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr))
427 saddr = &hdr->daddr;
428
429 /*
430 * Dest addr check
431 */
432
433 if (addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST) {
434 if (type != ICMPV6_PKT_TOOBIG &&
435 !(type == ICMPV6_PARAMPROB &&
436 code == ICMPV6_UNK_OPTION &&
437 (opt_unrec(skb, info))))
438 return;
439
440 saddr = NULL;
441 }
442
443 addr_type = ipv6_addr_type(&hdr->saddr);
444
445 /*
446 * Source addr check
447 */
448
449 if (__ipv6_addr_needs_scope_id(addr_type))
450 iif = skb->dev->ifindex;
451 else {
452 dst = skb_dst(skb);
453 iif = l3mdev_master_ifindex(dst ? dst->dev : skb->dev);
454 }
455
456 /*
457 * Must not send error if the source does not uniquely
458 * identify a single node (RFC2463 Section 2.4).
459 * We check unspecified / multicast addresses here,
460 * and anycast addresses will be checked later.
461 */
462 if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
463 net_dbg_ratelimited("icmp6_send: addr_any/mcast source [%pI6c > %pI6c]\n",
464 &hdr->saddr, &hdr->daddr);
465 return;
466 }
467
468 /*
469 * Never answer to a ICMP packet.
470 */
471 if (is_ineligible(skb)) {
472 net_dbg_ratelimited("icmp6_send: no reply to icmp error [%pI6c > %pI6c]\n",
473 &hdr->saddr, &hdr->daddr);
474 return;
475 }
476
477 mip6_addr_swap(skb);
478
479 memset(&fl6, 0, sizeof(fl6));
480 fl6.flowi6_proto = IPPROTO_ICMPV6;
481 fl6.daddr = hdr->saddr;
482 if (force_saddr)
483 saddr = force_saddr;
484 if (saddr)
485 fl6.saddr = *saddr;
486 fl6.flowi6_mark = mark;
487 fl6.flowi6_oif = iif;
488 fl6.fl6_icmp_type = type;
489 fl6.fl6_icmp_code = code;
490 fl6.flowi6_uid = sock_net_uid(net, NULL);
491 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
492
493 sk = icmpv6_xmit_lock(net);
494 if (!sk)
495 return;
496 sk->sk_mark = mark;
497 np = inet6_sk(sk);
498
499 if (!icmpv6_xrlim_allow(sk, type, &fl6))
500 goto out;
501
502 tmp_hdr.icmp6_type = type;
503 tmp_hdr.icmp6_code = code;
504 tmp_hdr.icmp6_cksum = 0;
505 tmp_hdr.icmp6_pointer = htonl(info);
506
507 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
508 fl6.flowi6_oif = np->mcast_oif;
509 else if (!fl6.flowi6_oif)
510 fl6.flowi6_oif = np->ucast_oif;
511
512 ipc6.tclass = np->tclass;
513 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
514
515 dst = icmpv6_route_lookup(net, skb, sk, &fl6);
516 if (IS_ERR(dst))
517 goto out;
518
519 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
520 ipc6.dontfrag = np->dontfrag;
521 ipc6.opt = NULL;
522
523 msg.skb = skb;
524 msg.offset = skb_network_offset(skb);
525 msg.type = type;
526
527 len = skb->len - msg.offset;
528 len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(struct icmp6hdr));
529 if (len < 0) {
530 net_dbg_ratelimited("icmp: len problem [%pI6c > %pI6c]\n",
531 &hdr->saddr, &hdr->daddr);
532 goto out_dst_release;
533 }
534
535 rcu_read_lock();
536 idev = __in6_dev_get(skb->dev);
537
538 err = ip6_append_data(sk, icmpv6_getfrag, &msg,
539 len + sizeof(struct icmp6hdr),
540 sizeof(struct icmp6hdr),
541 &ipc6, &fl6, (struct rt6_info *)dst,
542 MSG_DONTWAIT, &sockc_unused);
543 if (err) {
544 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
545 ip6_flush_pending_frames(sk);
546 } else {
547 err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
548 len + sizeof(struct icmp6hdr));
549 }
550 rcu_read_unlock();
551out_dst_release:
552 dst_release(dst);
553out:
554 icmpv6_xmit_unlock(sk);
555}
556
557/* Slightly more convenient version of icmp6_send.
558 */
559void icmpv6_param_prob(struct sk_buff *skb, u8 code, int pos)
560{
561 icmp6_send(skb, ICMPV6_PARAMPROB, code, pos, NULL);
562 kfree_skb(skb);
563}
564
565/* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
566 * if sufficient data bytes are available
567 * @nhs is the size of the tunnel header(s) :
568 * Either an IPv4 header for SIT encap
569 * an IPv4 header + GRE header for GRE encap
570 */
571int ip6_err_gen_icmpv6_unreach(struct sk_buff *skb, int nhs, int type,
572 unsigned int data_len)
573{
574 struct in6_addr temp_saddr;
575 struct rt6_info *rt;
576 struct sk_buff *skb2;
577 u32 info = 0;
578
579 if (!pskb_may_pull(skb, nhs + sizeof(struct ipv6hdr) + 8))
580 return 1;
581
582 /* RFC 4884 (partial) support for ICMP extensions */
583 if (data_len < 128 || (data_len & 7) || skb->len < data_len)
584 data_len = 0;
585
586 skb2 = data_len ? skb_copy(skb, GFP_ATOMIC) : skb_clone(skb, GFP_ATOMIC);
587
588 if (!skb2)
589 return 1;
590
591 skb_dst_drop(skb2);
592 skb_pull(skb2, nhs);
593 skb_reset_network_header(skb2);
594
595 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
596
597 if (rt && rt->dst.dev)
598 skb2->dev = rt->dst.dev;
599
600 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, &temp_saddr);
601
602 if (data_len) {
603 /* RFC 4884 (partial) support :
604 * insert 0 padding at the end, before the extensions
605 */
606 __skb_push(skb2, nhs);
607 skb_reset_network_header(skb2);
608 memmove(skb2->data, skb2->data + nhs, data_len - nhs);
609 memset(skb2->data + data_len - nhs, 0, nhs);
610 /* RFC 4884 4.5 : Length is measured in 64-bit words,
611 * and stored in reserved[0]
612 */
613 info = (data_len/8) << 24;
614 }
615 if (type == ICMP_TIME_EXCEEDED)
616 icmp6_send(skb2, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
617 info, &temp_saddr);
618 else
619 icmp6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH,
620 info, &temp_saddr);
621 if (rt)
622 ip6_rt_put(rt);
623
624 kfree_skb(skb2);
625
626 return 0;
627}
628EXPORT_SYMBOL(ip6_err_gen_icmpv6_unreach);
629
630static void icmpv6_echo_reply(struct sk_buff *skb)
631{
632 struct net *net = dev_net(skb->dev);
633 struct sock *sk;
634 struct inet6_dev *idev;
635 struct ipv6_pinfo *np;
636 const struct in6_addr *saddr = NULL;
637 struct icmp6hdr *icmph = icmp6_hdr(skb);
638 struct icmp6hdr tmp_hdr;
639 struct flowi6 fl6;
640 struct icmpv6_msg msg;
641 struct dst_entry *dst;
642 struct ipcm6_cookie ipc6;
643 int err = 0;
644 u32 mark = IP6_REPLY_MARK(net, skb->mark);
645 struct sockcm_cookie sockc_unused = {0};
646
647 saddr = &ipv6_hdr(skb)->daddr;
648
649 if (!ipv6_unicast_destination(skb) &&
650 !(net->ipv6.sysctl.anycast_src_echo_reply &&
651 ipv6_anycast_destination(skb_dst(skb), saddr)))
652 saddr = NULL;
653
654 memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
655 tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
656
657 memset(&fl6, 0, sizeof(fl6));
658 fl6.flowi6_proto = IPPROTO_ICMPV6;
659 fl6.daddr = ipv6_hdr(skb)->saddr;
660 if (saddr)
661 fl6.saddr = *saddr;
662 fl6.flowi6_oif = skb->dev->ifindex;
663 fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
664 fl6.flowi6_mark = mark;
665 fl6.flowi6_uid = sock_net_uid(net, NULL);
666 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
667
668 sk = icmpv6_xmit_lock(net);
669 if (!sk)
670 return;
671 sk->sk_mark = mark;
672 np = inet6_sk(sk);
673
674 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
675 fl6.flowi6_oif = np->mcast_oif;
676 else if (!fl6.flowi6_oif)
677 fl6.flowi6_oif = np->ucast_oif;
678
679 err = ip6_dst_lookup(net, sk, &dst, &fl6);
680 if (err)
681 goto out;
682 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
683 if (IS_ERR(dst))
684 goto out;
685
686 idev = __in6_dev_get(skb->dev);
687
688 msg.skb = skb;
689 msg.offset = 0;
690 msg.type = ICMPV6_ECHO_REPLY;
691
692 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
693 ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb));
694 ipc6.dontfrag = np->dontfrag;
695 ipc6.opt = NULL;
696
697 err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
698 sizeof(struct icmp6hdr), &ipc6, &fl6,
699 (struct rt6_info *)dst, MSG_DONTWAIT,
700 &sockc_unused);
701
702 if (err) {
703 __ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTERRORS);
704 ip6_flush_pending_frames(sk);
705 } else {
706 err = icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
707 skb->len + sizeof(struct icmp6hdr));
708 }
709 dst_release(dst);
710out:
711 icmpv6_xmit_unlock(sk);
712}
713
714void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info)
715{
716 const struct inet6_protocol *ipprot;
717 int inner_offset;
718 __be16 frag_off;
719 u8 nexthdr;
720 struct net *net = dev_net(skb->dev);
721
722 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
723 goto out;
724
725 nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr;
726 if (ipv6_ext_hdr(nexthdr)) {
727 /* now skip over extension headers */
728 inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
729 &nexthdr, &frag_off);
730 if (inner_offset < 0)
731 goto out;
732 } else {
733 inner_offset = sizeof(struct ipv6hdr);
734 }
735
736 /* Checkin header including 8 bytes of inner protocol header. */
737 if (!pskb_may_pull(skb, inner_offset+8))
738 goto out;
739
740 /* BUGGG_FUTURE: we should try to parse exthdrs in this packet.
741 Without this we will not able f.e. to make source routed
742 pmtu discovery.
743 Corresponding argument (opt) to notifiers is already added.
744 --ANK (980726)
745 */
746
747 ipprot = rcu_dereference(inet6_protos[nexthdr]);
748 if (ipprot && ipprot->err_handler)
749 ipprot->err_handler(skb, NULL, type, code, inner_offset, info);
750
751 raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info);
752 return;
753
754out:
755 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
756}
757
758/*
759 * Handle icmp messages
760 */
761
762static int icmpv6_rcv(struct sk_buff *skb)
763{
764 struct net_device *dev = skb->dev;
765 struct inet6_dev *idev = __in6_dev_get(dev);
766 const struct in6_addr *saddr, *daddr;
767 struct icmp6hdr *hdr;
768 u8 type;
769 bool success = false;
770
771 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
772 struct sec_path *sp = skb_sec_path(skb);
773 int nh;
774
775 if (!(sp && sp->xvec[sp->len - 1]->props.flags &
776 XFRM_STATE_ICMP))
777 goto drop_no_count;
778
779 if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(struct ipv6hdr)))
780 goto drop_no_count;
781
782 nh = skb_network_offset(skb);
783 skb_set_network_header(skb, sizeof(*hdr));
784
785 if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb))
786 goto drop_no_count;
787
788 skb_set_network_header(skb, nh);
789 }
790
791 __ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INMSGS);
792
793 saddr = &ipv6_hdr(skb)->saddr;
794 daddr = &ipv6_hdr(skb)->daddr;
795
796 if (skb_checksum_validate(skb, IPPROTO_ICMPV6, ip6_compute_pseudo)) {
797 net_dbg_ratelimited("ICMPv6 checksum failed [%pI6c > %pI6c]\n",
798 saddr, daddr);
799 goto csum_error;
800 }
801
802 if (!pskb_pull(skb, sizeof(*hdr)))
803 goto discard_it;
804
805 hdr = icmp6_hdr(skb);
806
807 type = hdr->icmp6_type;
808
809 ICMP6MSGIN_INC_STATS(dev_net(dev), idev, type);
810
811 switch (type) {
812 case ICMPV6_ECHO_REQUEST:
813 icmpv6_echo_reply(skb);
814 break;
815
816 case ICMPV6_ECHO_REPLY:
817 success = ping_rcv(skb);
818 break;
819
820 case ICMPV6_PKT_TOOBIG:
821 /* BUGGG_FUTURE: if packet contains rthdr, we cannot update
822 standard destination cache. Seems, only "advanced"
823 destination cache will allow to solve this problem
824 --ANK (980726)
825 */
826 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
827 goto discard_it;
828 hdr = icmp6_hdr(skb);
829
830 /*
831 * Drop through to notify
832 */
833
834 case ICMPV6_DEST_UNREACH:
835 case ICMPV6_TIME_EXCEED:
836 case ICMPV6_PARAMPROB:
837 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
838 break;
839
840 case NDISC_ROUTER_SOLICITATION:
841 case NDISC_ROUTER_ADVERTISEMENT:
842 case NDISC_NEIGHBOUR_SOLICITATION:
843 case NDISC_NEIGHBOUR_ADVERTISEMENT:
844 case NDISC_REDIRECT:
845 ndisc_rcv(skb);
846 break;
847
848 case ICMPV6_MGM_QUERY:
849 igmp6_event_query(skb);
850 break;
851
852 case ICMPV6_MGM_REPORT:
853 igmp6_event_report(skb);
854 break;
855
856 case ICMPV6_MGM_REDUCTION:
857 case ICMPV6_NI_QUERY:
858 case ICMPV6_NI_REPLY:
859 case ICMPV6_MLD2_REPORT:
860 case ICMPV6_DHAAD_REQUEST:
861 case ICMPV6_DHAAD_REPLY:
862 case ICMPV6_MOBILE_PREFIX_SOL:
863 case ICMPV6_MOBILE_PREFIX_ADV:
864 break;
865
866 default:
867 /* informational */
868 if (type & ICMPV6_INFOMSG_MASK)
869 break;
870
871 net_dbg_ratelimited("icmpv6: msg of unknown type [%pI6c > %pI6c]\n",
872 saddr, daddr);
873
874 /*
875 * error of unknown type.
876 * must pass to upper level
877 */
878
879 icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu);
880 }
881
882 /* until the v6 path can be better sorted assume failure and
883 * preserve the status quo behaviour for the rest of the paths to here
884 */
885 if (success)
886 consume_skb(skb);
887 else
888 kfree_skb(skb);
889
890 return 0;
891
892csum_error:
893 __ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
894discard_it:
895 __ICMP6_INC_STATS(dev_net(dev), idev, ICMP6_MIB_INERRORS);
896drop_no_count:
897 kfree_skb(skb);
898 return 0;
899}
900
901void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6,
902 u8 type,
903 const struct in6_addr *saddr,
904 const struct in6_addr *daddr,
905 int oif)
906{
907 memset(fl6, 0, sizeof(*fl6));
908 fl6->saddr = *saddr;
909 fl6->daddr = *daddr;
910 fl6->flowi6_proto = IPPROTO_ICMPV6;
911 fl6->fl6_icmp_type = type;
912 fl6->fl6_icmp_code = 0;
913 fl6->flowi6_oif = oif;
914 security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
915}
916
917static int __net_init icmpv6_sk_init(struct net *net)
918{
919 struct sock *sk;
920 int err, i, j;
921
922 net->ipv6.icmp_sk =
923 kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL);
924 if (!net->ipv6.icmp_sk)
925 return -ENOMEM;
926
927 for_each_possible_cpu(i) {
928 err = inet_ctl_sock_create(&sk, PF_INET6,
929 SOCK_RAW, IPPROTO_ICMPV6, net);
930 if (err < 0) {
931 pr_err("Failed to initialize the ICMP6 control socket (err %d)\n",
932 err);
933 goto fail;
934 }
935
936 net->ipv6.icmp_sk[i] = sk;
937
938 /* Enough space for 2 64K ICMP packets, including
939 * sk_buff struct overhead.
940 */
941 sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
942 }
943 return 0;
944
945 fail:
946 for (j = 0; j < i; j++)
947 inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
948 kfree(net->ipv6.icmp_sk);
949 return err;
950}
951
952static void __net_exit icmpv6_sk_exit(struct net *net)
953{
954 int i;
955
956 for_each_possible_cpu(i) {
957 inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
958 }
959 kfree(net->ipv6.icmp_sk);
960}
961
962static struct pernet_operations icmpv6_sk_ops = {
963 .init = icmpv6_sk_init,
964 .exit = icmpv6_sk_exit,
965};
966
967int __init icmpv6_init(void)
968{
969 int err;
970
971 err = register_pernet_subsys(&icmpv6_sk_ops);
972 if (err < 0)
973 return err;
974
975 err = -EAGAIN;
976 if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0)
977 goto fail;
978
979 err = inet6_register_icmp_sender(icmp6_send);
980 if (err)
981 goto sender_reg_err;
982 return 0;
983
984sender_reg_err:
985 inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
986fail:
987 pr_err("Failed to register ICMP6 protocol\n");
988 unregister_pernet_subsys(&icmpv6_sk_ops);
989 return err;
990}
991
992void icmpv6_cleanup(void)
993{
994 inet6_unregister_icmp_sender(icmp6_send);
995 unregister_pernet_subsys(&icmpv6_sk_ops);
996 inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
997}
998
999
1000static const struct icmp6_err {
1001 int err;
1002 int fatal;
1003} tab_unreach[] = {
1004 { /* NOROUTE */
1005 .err = ENETUNREACH,
1006 .fatal = 0,
1007 },
1008 { /* ADM_PROHIBITED */
1009 .err = EACCES,
1010 .fatal = 1,
1011 },
1012 { /* Was NOT_NEIGHBOUR, now reserved */
1013 .err = EHOSTUNREACH,
1014 .fatal = 0,
1015 },
1016 { /* ADDR_UNREACH */
1017 .err = EHOSTUNREACH,
1018 .fatal = 0,
1019 },
1020 { /* PORT_UNREACH */
1021 .err = ECONNREFUSED,
1022 .fatal = 1,
1023 },
1024 { /* POLICY_FAIL */
1025 .err = EACCES,
1026 .fatal = 1,
1027 },
1028 { /* REJECT_ROUTE */
1029 .err = EACCES,
1030 .fatal = 1,
1031 },
1032};
1033
1034int icmpv6_err_convert(u8 type, u8 code, int *err)
1035{
1036 int fatal = 0;
1037
1038 *err = EPROTO;
1039
1040 switch (type) {
1041 case ICMPV6_DEST_UNREACH:
1042 fatal = 1;
1043 if (code < ARRAY_SIZE(tab_unreach)) {
1044 *err = tab_unreach[code].err;
1045 fatal = tab_unreach[code].fatal;
1046 }
1047 break;
1048
1049 case ICMPV6_PKT_TOOBIG:
1050 *err = EMSGSIZE;
1051 break;
1052
1053 case ICMPV6_PARAMPROB:
1054 *err = EPROTO;
1055 fatal = 1;
1056 break;
1057
1058 case ICMPV6_TIME_EXCEED:
1059 *err = EHOSTUNREACH;
1060 break;
1061 }
1062
1063 return fatal;
1064}
1065EXPORT_SYMBOL(icmpv6_err_convert);
1066
1067#ifdef CONFIG_SYSCTL
1068static struct ctl_table ipv6_icmp_table_template[] = {
1069 {
1070 .procname = "ratelimit",
1071 .data = &init_net.ipv6.sysctl.icmpv6_time,
1072 .maxlen = sizeof(int),
1073 .mode = 0644,
1074 .proc_handler = proc_dointvec_ms_jiffies,
1075 },
1076 { },
1077};
1078
1079struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1080{
1081 struct ctl_table *table;
1082
1083 table = kmemdup(ipv6_icmp_table_template,
1084 sizeof(ipv6_icmp_table_template),
1085 GFP_KERNEL);
1086
1087 if (table)
1088 table[0].data = &net->ipv6.sysctl.icmpv6_time;
1089
1090 return table;
1091}
1092#endif