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#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
72#include <asm/uaccess.h>
73
74/*
75 * The ICMP socket(s). This is the most convenient way to flow control
76 * our ICMP output as well as maintain a clean interface throughout
77 * all layers. All Socketless IP sends will soon be gone.
78 *
79 * On SMP we have one ICMP socket per-cpu.
80 */
81static inline struct sock *icmpv6_sk(struct net *net)
82{
83 return net->ipv6.icmp_sk[smp_processor_id()];
84}
85
86static void icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
87 u8 type, u8 code, int offset, __be32 info)
88{
89 /* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */
90 struct icmp6hdr *icmp6 = (struct icmp6hdr *) (skb->data + offset);
91 struct net *net = dev_net(skb->dev);
92
93 if (type == ICMPV6_PKT_TOOBIG)
94 ip6_update_pmtu(skb, net, info, 0, 0);
95 else if (type == NDISC_REDIRECT)
96 ip6_redirect(skb, net, skb->dev->ifindex, 0);
97
98 if (!(type & ICMPV6_INFOMSG_MASK))
99 if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST)
100 ping_err(skb, offset, info);
101}
102
103static int icmpv6_rcv(struct sk_buff *skb);
104
105static 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
111static __inline__ struct sock *icmpv6_xmit_lock(struct net *net)
112{
113 struct sock *sk;
114
115 local_bh_disable();
116
117 sk = icmpv6_sk(net);
118 if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
119 /* This can happen if the output path (f.e. SIT or
120 * ip6ip6 tunnel) signals dst_link_failure() for an
121 * outgoing ICMP6 packet.
122 */
123 local_bh_enable();
124 return NULL;
125 }
126 return sk;
127}
128
129static __inline__ void icmpv6_xmit_unlock(struct sock *sk)
130{
131 spin_unlock_bh(&sk->sk_lock.slock);
132}
133
134/*
135 * Figure out, may we reply to this packet with icmp error.
136 *
137 * We do not reply, if:
138 * - it was icmp error message.
139 * - it is truncated, so that it is known, that protocol is ICMPV6
140 * (i.e. in the middle of some exthdr)
141 *
142 * --ANK (980726)
143 */
144
145static bool is_ineligible(const struct sk_buff *skb)
146{
147 int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
148 int len = skb->len - ptr;
149 __u8 nexthdr = ipv6_hdr(skb)->nexthdr;
150 __be16 frag_off;
151
152 if (len < 0)
153 return true;
154
155 ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr, &frag_off);
156 if (ptr < 0)
157 return false;
158 if (nexthdr == IPPROTO_ICMPV6) {
159 u8 _type, *tp;
160 tp = skb_header_pointer(skb,
161 ptr+offsetof(struct icmp6hdr, icmp6_type),
162 sizeof(_type), &_type);
163 if (tp == NULL ||
164 !(*tp & ICMPV6_INFOMSG_MASK))
165 return true;
166 }
167 return false;
168}
169
170/*
171 * Check the ICMP output rate limit
172 */
173static inline bool icmpv6_xrlim_allow(struct sock *sk, u8 type,
174 struct flowi6 *fl6)
175{
176 struct dst_entry *dst;
177 struct net *net = sock_net(sk);
178 bool res = false;
179
180 /* Informational messages are not limited. */
181 if (type & ICMPV6_INFOMSG_MASK)
182 return true;
183
184 /* Do not limit pmtu discovery, it would break it. */
185 if (type == ICMPV6_PKT_TOOBIG)
186 return true;
187
188 /*
189 * Look up the output route.
190 * XXX: perhaps the expire for routing entries cloned by
191 * this lookup should be more aggressive (not longer than timeout).
192 */
193 dst = ip6_route_output(net, sk, fl6);
194 if (dst->error) {
195 IP6_INC_STATS(net, ip6_dst_idev(dst),
196 IPSTATS_MIB_OUTNOROUTES);
197 } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) {
198 res = true;
199 } else {
200 struct rt6_info *rt = (struct rt6_info *)dst;
201 int tmo = net->ipv6.sysctl.icmpv6_time;
202 struct inet_peer *peer;
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 peer = inet_getpeer_v6(net->ipv6.peers, &rt->rt6i_dst.addr, 1);
209 res = inet_peer_xrlim_allow(peer, tmo);
210 if (peer)
211 inet_putpeer(peer);
212 }
213 dst_release(dst);
214 return res;
215}
216
217/*
218 * an inline helper for the "simple" if statement below
219 * checks if parameter problem report is caused by an
220 * unrecognized IPv6 option that has the Option Type
221 * highest-order two bits set to 10
222 */
223
224static bool opt_unrec(struct sk_buff *skb, __u32 offset)
225{
226 u8 _optval, *op;
227
228 offset += skb_network_offset(skb);
229 op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval);
230 if (op == NULL)
231 return true;
232 return (*op & 0xC0) == 0x80;
233}
234
235int icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
236 struct icmp6hdr *thdr, int len)
237{
238 struct sk_buff *skb;
239 struct icmp6hdr *icmp6h;
240 int err = 0;
241
242 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
243 goto out;
244
245 icmp6h = icmp6_hdr(skb);
246 memcpy(icmp6h, thdr, sizeof(struct icmp6hdr));
247 icmp6h->icmp6_cksum = 0;
248
249 if (skb_queue_len(&sk->sk_write_queue) == 1) {
250 skb->csum = csum_partial(icmp6h,
251 sizeof(struct icmp6hdr), skb->csum);
252 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
253 &fl6->daddr,
254 len, fl6->flowi6_proto,
255 skb->csum);
256 } else {
257 __wsum tmp_csum = 0;
258
259 skb_queue_walk(&sk->sk_write_queue, skb) {
260 tmp_csum = csum_add(tmp_csum, skb->csum);
261 }
262
263 tmp_csum = csum_partial(icmp6h,
264 sizeof(struct icmp6hdr), tmp_csum);
265 icmp6h->icmp6_cksum = csum_ipv6_magic(&fl6->saddr,
266 &fl6->daddr,
267 len, fl6->flowi6_proto,
268 tmp_csum);
269 }
270 ip6_push_pending_frames(sk);
271out:
272 return err;
273}
274
275struct icmpv6_msg {
276 struct sk_buff *skb;
277 int offset;
278 uint8_t type;
279};
280
281static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
282{
283 struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
284 struct sk_buff *org_skb = msg->skb;
285 __wsum csum = 0;
286
287 csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
288 to, len, csum);
289 skb->csum = csum_block_add(skb->csum, csum, odd);
290 if (!(msg->type & ICMPV6_INFOMSG_MASK))
291 nf_ct_attach(skb, org_skb);
292 return 0;
293}
294
295#if IS_ENABLED(CONFIG_IPV6_MIP6)
296static void mip6_addr_swap(struct sk_buff *skb)
297{
298 struct ipv6hdr *iph = ipv6_hdr(skb);
299 struct inet6_skb_parm *opt = IP6CB(skb);
300 struct ipv6_destopt_hao *hao;
301 struct in6_addr tmp;
302 int off;
303
304 if (opt->dsthao) {
305 off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO);
306 if (likely(off >= 0)) {
307 hao = (struct ipv6_destopt_hao *)
308 (skb_network_header(skb) + off);
309 tmp = iph->saddr;
310 iph->saddr = hao->addr;
311 hao->addr = tmp;
312 }
313 }
314}
315#else
316static inline void mip6_addr_swap(struct sk_buff *skb) {}
317#endif
318
319static struct dst_entry *icmpv6_route_lookup(struct net *net,
320 struct sk_buff *skb,
321 struct sock *sk,
322 struct flowi6 *fl6)
323{
324 struct dst_entry *dst, *dst2;
325 struct flowi6 fl2;
326 int err;
327
328 err = ip6_dst_lookup(sk, &dst, fl6);
329 if (err)
330 return ERR_PTR(err);
331
332 /*
333 * We won't send icmp if the destination is known
334 * anycast.
335 */
336 if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) {
337 LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: acast source\n");
338 dst_release(dst);
339 return ERR_PTR(-EINVAL);
340 }
341
342 /* No need to clone since we're just using its address. */
343 dst2 = dst;
344
345 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), sk, 0);
346 if (!IS_ERR(dst)) {
347 if (dst != dst2)
348 return dst;
349 } else {
350 if (PTR_ERR(dst) == -EPERM)
351 dst = NULL;
352 else
353 return dst;
354 }
355
356 err = xfrm_decode_session_reverse(skb, flowi6_to_flowi(&fl2), AF_INET6);
357 if (err)
358 goto relookup_failed;
359
360 err = ip6_dst_lookup(sk, &dst2, &fl2);
361 if (err)
362 goto relookup_failed;
363
364 dst2 = xfrm_lookup(net, dst2, flowi6_to_flowi(&fl2), sk, XFRM_LOOKUP_ICMP);
365 if (!IS_ERR(dst2)) {
366 dst_release(dst);
367 dst = dst2;
368 } else {
369 err = PTR_ERR(dst2);
370 if (err == -EPERM) {
371 dst_release(dst);
372 return dst2;
373 } else
374 goto relookup_failed;
375 }
376
377relookup_failed:
378 if (dst)
379 return dst;
380 return ERR_PTR(err);
381}
382
383/*
384 * Send an ICMP message in response to a packet in error
385 */
386static void icmp6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info)
387{
388 struct net *net = dev_net(skb->dev);
389 struct inet6_dev *idev = NULL;
390 struct ipv6hdr *hdr = ipv6_hdr(skb);
391 struct sock *sk;
392 struct ipv6_pinfo *np;
393 const struct in6_addr *saddr = NULL;
394 struct dst_entry *dst;
395 struct icmp6hdr tmp_hdr;
396 struct flowi6 fl6;
397 struct icmpv6_msg msg;
398 int iif = 0;
399 int addr_type = 0;
400 int len;
401 int hlimit;
402 int err = 0;
403
404 if ((u8 *)hdr < skb->head ||
405 (skb_network_header(skb) + sizeof(*hdr)) > skb_tail_pointer(skb))
406 return;
407
408 /*
409 * Make sure we respect the rules
410 * i.e. RFC 1885 2.4(e)
411 * Rule (e.1) is enforced by not using icmp6_send
412 * in any code that processes icmp errors.
413 */
414 addr_type = ipv6_addr_type(&hdr->daddr);
415
416 if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0) ||
417 ipv6_chk_acast_addr_src(net, skb->dev, &hdr->daddr))
418 saddr = &hdr->daddr;
419
420 /*
421 * Dest addr check
422 */
423
424 if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) {
425 if (type != ICMPV6_PKT_TOOBIG &&
426 !(type == ICMPV6_PARAMPROB &&
427 code == ICMPV6_UNK_OPTION &&
428 (opt_unrec(skb, info))))
429 return;
430
431 saddr = NULL;
432 }
433
434 addr_type = ipv6_addr_type(&hdr->saddr);
435
436 /*
437 * Source addr check
438 */
439
440 if (__ipv6_addr_needs_scope_id(addr_type))
441 iif = skb->dev->ifindex;
442
443 /*
444 * Must not send error if the source does not uniquely
445 * identify a single node (RFC2463 Section 2.4).
446 * We check unspecified / multicast addresses here,
447 * and anycast addresses will be checked later.
448 */
449 if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) {
450 LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: addr_any/mcast source\n");
451 return;
452 }
453
454 /*
455 * Never answer to a ICMP packet.
456 */
457 if (is_ineligible(skb)) {
458 LIMIT_NETDEBUG(KERN_DEBUG "icmp6_send: no reply to icmp error\n");
459 return;
460 }
461
462 mip6_addr_swap(skb);
463
464 memset(&fl6, 0, sizeof(fl6));
465 fl6.flowi6_proto = IPPROTO_ICMPV6;
466 fl6.daddr = hdr->saddr;
467 if (saddr)
468 fl6.saddr = *saddr;
469 fl6.flowi6_oif = iif;
470 fl6.fl6_icmp_type = type;
471 fl6.fl6_icmp_code = code;
472 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
473
474 sk = icmpv6_xmit_lock(net);
475 if (sk == NULL)
476 return;
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(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();
530out_dst_release:
531 dst_release(dst);
532out:
533 icmpv6_xmit_unlock(sk);
534}
535
536/* Slightly more convenient version of icmp6_send.
537 */
538void 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
544static 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 u8 tclass;
559
560 saddr = &ipv6_hdr(skb)->daddr;
561
562 if (!ipv6_unicast_destination(skb) &&
563 !(net->ipv6.sysctl.anycast_src_echo_reply &&
564 ipv6_anycast_destination(skb)))
565 saddr = NULL;
566
567 memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
568 tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
569
570 memset(&fl6, 0, sizeof(fl6));
571 fl6.flowi6_proto = IPPROTO_ICMPV6;
572 fl6.daddr = ipv6_hdr(skb)->saddr;
573 if (saddr)
574 fl6.saddr = *saddr;
575 fl6.flowi6_oif = skb->dev->ifindex;
576 fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
577 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
578
579 sk = icmpv6_xmit_lock(net);
580 if (sk == NULL)
581 return;
582 np = inet6_sk(sk);
583
584 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
585 fl6.flowi6_oif = np->mcast_oif;
586 else if (!fl6.flowi6_oif)
587 fl6.flowi6_oif = np->ucast_oif;
588
589 err = ip6_dst_lookup(sk, &dst, &fl6);
590 if (err)
591 goto out;
592 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
593 if (IS_ERR(dst))
594 goto out;
595
596 if (ipv6_addr_is_multicast(&fl6.daddr))
597 hlimit = np->mcast_hops;
598 else
599 hlimit = np->hop_limit;
600 if (hlimit < 0)
601 hlimit = ip6_dst_hoplimit(dst);
602
603 idev = __in6_dev_get(skb->dev);
604
605 msg.skb = skb;
606 msg.offset = 0;
607 msg.type = ICMPV6_ECHO_REPLY;
608
609 tclass = ipv6_get_dsfield(ipv6_hdr(skb));
610 err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr),
611 sizeof(struct icmp6hdr), hlimit, 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);
623out:
624 icmpv6_xmit_unlock(sk);
625}
626
627void 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
672static 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
805csum_error:
806 ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_CSUMERRORS);
807discard_it:
808 ICMP6_INC_STATS_BH(dev_net(dev), idev, ICMP6_MIB_INERRORS);
809drop_no_count:
810 kfree_skb(skb);
811 return 0;
812}
813
814void 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 */
833static struct lock_class_key icmpv6_socket_sk_dst_lock_key;
834
835static 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
879static 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
889static struct pernet_operations icmpv6_sk_ops = {
890 .init = icmpv6_sk_init,
891 .exit = icmpv6_sk_exit,
892};
893
894int __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
911sender_reg_err:
912 inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6);
913fail:
914 pr_err("Failed to register ICMP6 protocol\n");
915 unregister_pernet_subsys(&icmpv6_sk_ops);
916 return err;
917}
918
919void 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
927static const struct icmp6_err {
928 int err;
929 int fatal;
930} tab_unreach[] = {
931 { /* NOROUTE */
932 .err = ENETUNREACH,
933 .fatal = 0,
934 },
935 { /* ADM_PROHIBITED */
936 .err = EACCES,
937 .fatal = 1,
938 },
939 { /* Was NOT_NEIGHBOUR, now reserved */
940 .err = EHOSTUNREACH,
941 .fatal = 0,
942 },
943 { /* ADDR_UNREACH */
944 .err = EHOSTUNREACH,
945 .fatal = 0,
946 },
947 { /* PORT_UNREACH */
948 .err = ECONNREFUSED,
949 .fatal = 1,
950 },
951 { /* POLICY_FAIL */
952 .err = EACCES,
953 .fatal = 1,
954 },
955 { /* REJECT_ROUTE */
956 .err = EACCES,
957 .fatal = 1,
958 },
959};
960
961int icmpv6_err_convert(u8 type, u8 code, int *err)
962{
963 int fatal = 0;
964
965 *err = EPROTO;
966
967 switch (type) {
968 case ICMPV6_DEST_UNREACH:
969 fatal = 1;
970 if (code < ARRAY_SIZE(tab_unreach)) {
971 *err = tab_unreach[code].err;
972 fatal = tab_unreach[code].fatal;
973 }
974 break;
975
976 case ICMPV6_PKT_TOOBIG:
977 *err = EMSGSIZE;
978 break;
979
980 case ICMPV6_PARAMPROB:
981 *err = EPROTO;
982 fatal = 1;
983 break;
984
985 case ICMPV6_TIME_EXCEED:
986 *err = EHOSTUNREACH;
987 break;
988 }
989
990 return fatal;
991}
992EXPORT_SYMBOL(icmpv6_err_convert);
993
994#ifdef CONFIG_SYSCTL
995static struct ctl_table ipv6_icmp_table_template[] = {
996 {
997 .procname = "ratelimit",
998 .data = &init_net.ipv6.sysctl.icmpv6_time,
999 .maxlen = sizeof(int),
1000 .mode = 0644,
1001 .proc_handler = proc_dointvec_ms_jiffies,
1002 },
1003 { },
1004};
1005
1006struct ctl_table * __net_init ipv6_icmp_sysctl_init(struct net *net)
1007{
1008 struct ctl_table *table;
1009
1010 table = kmemdup(ipv6_icmp_table_template,
1011 sizeof(ipv6_icmp_table_template),
1012 GFP_KERNEL);
1013
1014 if (table)
1015 table[0].data = &net->ipv6.sysctl.icmpv6_time;
1016
1017 return table;
1018}
1019#endif
1020
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