]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Internet Control Message Protocol (ICMPv6) | |
3 | * Linux INET6 implementation | |
4 | * | |
5 | * Authors: | |
6 | * Pedro Roque <roque@di.fc.ul.pt> | |
7 | * | |
8 | * $Id: icmp.c,v 1.38 2002/02/08 03:57:19 davem Exp $ | |
9 | * | |
10 | * Based on net/ipv4/icmp.c | |
11 | * | |
12 | * RFC 1885 | |
13 | * | |
14 | * This program is free software; you can redistribute it and/or | |
15 | * modify it under the terms of the GNU General Public License | |
16 | * as published by the Free Software Foundation; either version | |
17 | * 2 of the License, or (at your option) any later version. | |
18 | */ | |
19 | ||
20 | /* | |
21 | * Changes: | |
22 | * | |
23 | * Andi Kleen : exception handling | |
24 | * Andi Kleen add rate limits. never reply to a icmp. | |
25 | * add more length checks and other fixes. | |
26 | * yoshfuji : ensure to sent parameter problem for | |
27 | * fragments. | |
28 | * YOSHIFUJI Hideaki @USAGI: added sysctl for icmp rate limit. | |
29 | * Randy Dunlap and | |
30 | * YOSHIFUJI Hideaki @USAGI: Per-interface statistics support | |
31 | * Kazunori MIYAZAWA @USAGI: change output process to use ip6_append_data | |
32 | */ | |
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> | |
1da177e4 LT |
40 | #include <linux/sockios.h> |
41 | #include <linux/net.h> | |
42 | #include <linux/skbuff.h> | |
43 | #include <linux/init.h> | |
763ecff1 | 44 | #include <linux/netfilter.h> |
1da177e4 LT |
45 | |
46 | #ifdef CONFIG_SYSCTL | |
47 | #include <linux/sysctl.h> | |
48 | #endif | |
49 | ||
50 | #include <linux/inet.h> | |
51 | #include <linux/netdevice.h> | |
52 | #include <linux/icmpv6.h> | |
53 | ||
54 | #include <net/ip.h> | |
55 | #include <net/sock.h> | |
56 | ||
57 | #include <net/ipv6.h> | |
58 | #include <net/ip6_checksum.h> | |
59 | #include <net/protocol.h> | |
60 | #include <net/raw.h> | |
61 | #include <net/rawv6.h> | |
62 | #include <net/transp_v6.h> | |
63 | #include <net/ip6_route.h> | |
64 | #include <net/addrconf.h> | |
65 | #include <net/icmp.h> | |
8b7817f3 | 66 | #include <net/xfrm.h> |
1da177e4 LT |
67 | |
68 | #include <asm/uaccess.h> | |
69 | #include <asm/system.h> | |
70 | ||
ba89966c | 71 | DEFINE_SNMP_STAT(struct icmpv6_mib, icmpv6_statistics) __read_mostly; |
7159039a | 72 | EXPORT_SYMBOL(icmpv6_statistics); |
14878f75 DS |
73 | DEFINE_SNMP_STAT(struct icmpv6msg_mib, icmpv6msg_statistics) __read_mostly; |
74 | EXPORT_SYMBOL(icmpv6msg_statistics); | |
1da177e4 LT |
75 | |
76 | /* | |
77 | * The ICMP socket(s). This is the most convenient way to flow control | |
78 | * our ICMP output as well as maintain a clean interface throughout | |
79 | * all layers. All Socketless IP sends will soon be gone. | |
80 | * | |
81 | * On SMP we have one ICMP socket per-cpu. | |
82 | */ | |
98c6d1b2 DL |
83 | static inline struct sock *icmpv6_sk(struct net *net) |
84 | { | |
85 | return net->ipv6.icmp_sk[smp_processor_id()]; | |
86 | } | |
1da177e4 | 87 | |
e5bbef20 | 88 | static int icmpv6_rcv(struct sk_buff *skb); |
1da177e4 LT |
89 | |
90 | static struct inet6_protocol icmpv6_protocol = { | |
91 | .handler = icmpv6_rcv, | |
8b7817f3 | 92 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, |
1da177e4 LT |
93 | }; |
94 | ||
405666db | 95 | static __inline__ int icmpv6_xmit_lock(struct sock *sk) |
1da177e4 LT |
96 | { |
97 | local_bh_disable(); | |
98 | ||
405666db | 99 | if (unlikely(!spin_trylock(&sk->sk_lock.slock))) { |
1da177e4 LT |
100 | /* This can happen if the output path (f.e. SIT or |
101 | * ip6ip6 tunnel) signals dst_link_failure() for an | |
102 | * outgoing ICMP6 packet. | |
103 | */ | |
104 | local_bh_enable(); | |
105 | return 1; | |
106 | } | |
107 | return 0; | |
108 | } | |
109 | ||
405666db | 110 | static __inline__ void icmpv6_xmit_unlock(struct sock *sk) |
1da177e4 | 111 | { |
405666db | 112 | spin_unlock_bh(&sk->sk_lock.slock); |
1da177e4 LT |
113 | } |
114 | ||
1ab1457c | 115 | /* |
1da177e4 LT |
116 | * Slightly more convenient version of icmpv6_send. |
117 | */ | |
118 | void icmpv6_param_prob(struct sk_buff *skb, int code, int pos) | |
119 | { | |
120 | icmpv6_send(skb, ICMPV6_PARAMPROB, code, pos, skb->dev); | |
121 | kfree_skb(skb); | |
122 | } | |
123 | ||
124 | /* | |
125 | * Figure out, may we reply to this packet with icmp error. | |
126 | * | |
127 | * We do not reply, if: | |
128 | * - it was icmp error message. | |
129 | * - it is truncated, so that it is known, that protocol is ICMPV6 | |
130 | * (i.e. in the middle of some exthdr) | |
131 | * | |
132 | * --ANK (980726) | |
133 | */ | |
134 | ||
135 | static int is_ineligible(struct sk_buff *skb) | |
136 | { | |
0660e03f | 137 | int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data; |
1da177e4 | 138 | int len = skb->len - ptr; |
0660e03f | 139 | __u8 nexthdr = ipv6_hdr(skb)->nexthdr; |
1da177e4 LT |
140 | |
141 | if (len < 0) | |
142 | return 1; | |
143 | ||
0d3d077c | 144 | ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr); |
1da177e4 LT |
145 | if (ptr < 0) |
146 | return 0; | |
147 | if (nexthdr == IPPROTO_ICMPV6) { | |
148 | u8 _type, *tp; | |
149 | tp = skb_header_pointer(skb, | |
150 | ptr+offsetof(struct icmp6hdr, icmp6_type), | |
151 | sizeof(_type), &_type); | |
152 | if (tp == NULL || | |
153 | !(*tp & ICMPV6_INFOMSG_MASK)) | |
154 | return 1; | |
155 | } | |
156 | return 0; | |
157 | } | |
158 | ||
1ab1457c YH |
159 | /* |
160 | * Check the ICMP output rate limit | |
1da177e4 LT |
161 | */ |
162 | static inline int icmpv6_xrlim_allow(struct sock *sk, int type, | |
163 | struct flowi *fl) | |
164 | { | |
165 | struct dst_entry *dst; | |
9a43b709 | 166 | struct net *net = sk->sk_net; |
1da177e4 LT |
167 | int res = 0; |
168 | ||
169 | /* Informational messages are not limited. */ | |
170 | if (type & ICMPV6_INFOMSG_MASK) | |
171 | return 1; | |
172 | ||
173 | /* Do not limit pmtu discovery, it would break it. */ | |
174 | if (type == ICMPV6_PKT_TOOBIG) | |
175 | return 1; | |
176 | ||
1ab1457c | 177 | /* |
1da177e4 LT |
178 | * Look up the output route. |
179 | * XXX: perhaps the expire for routing entries cloned by | |
180 | * this lookup should be more aggressive (not longer than timeout). | |
181 | */ | |
9a43b709 | 182 | dst = ip6_route_output(net, sk, fl); |
1da177e4 | 183 | if (dst->error) { |
a11d206d YH |
184 | IP6_INC_STATS(ip6_dst_idev(dst), |
185 | IPSTATS_MIB_OUTNOROUTES); | |
1da177e4 LT |
186 | } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) { |
187 | res = 1; | |
188 | } else { | |
189 | struct rt6_info *rt = (struct rt6_info *)dst; | |
9a43b709 | 190 | int tmo = net->ipv6.sysctl.icmpv6_time; |
1da177e4 LT |
191 | |
192 | /* Give more bandwidth to wider prefixes. */ | |
193 | if (rt->rt6i_dst.plen < 128) | |
194 | tmo >>= ((128 - rt->rt6i_dst.plen)>>5); | |
195 | ||
196 | res = xrlim_allow(dst, 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 | |
1ab1457c | 205 | * unrecognized IPv6 option that has the Option Type |
1da177e4 LT |
206 | * highest-order two bits set to 10 |
207 | */ | |
208 | ||
209 | static __inline__ int opt_unrec(struct sk_buff *skb, __u32 offset) | |
210 | { | |
211 | u8 _optval, *op; | |
212 | ||
bbe735e4 | 213 | offset += skb_network_offset(skb); |
1da177e4 LT |
214 | op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval); |
215 | if (op == NULL) | |
216 | return 1; | |
217 | return (*op & 0xC0) == 0x80; | |
218 | } | |
219 | ||
220 | static int icmpv6_push_pending_frames(struct sock *sk, struct flowi *fl, 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 | ||
cc70ab26 | 229 | icmp6h = icmp6_hdr(skb); |
1da177e4 LT |
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((char *)icmp6h, | |
235 | sizeof(struct icmp6hdr), skb->csum); | |
236 | icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src, | |
237 | &fl->fl6_dst, | |
238 | len, fl->proto, | |
239 | skb->csum); | |
240 | } else { | |
868c86bc | 241 | __wsum tmp_csum = 0; |
1da177e4 LT |
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((char *)icmp6h, | |
248 | sizeof(struct icmp6hdr), tmp_csum); | |
868c86bc AV |
249 | icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src, |
250 | &fl->fl6_dst, | |
251 | len, fl->proto, | |
252 | tmp_csum); | |
1da177e4 | 253 | } |
1da177e4 LT |
254 | ip6_push_pending_frames(sk); |
255 | out: | |
256 | return err; | |
257 | } | |
258 | ||
259 | struct icmpv6_msg { | |
260 | struct sk_buff *skb; | |
261 | int offset; | |
763ecff1 | 262 | uint8_t type; |
1da177e4 LT |
263 | }; |
264 | ||
265 | static 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; | |
5f92a738 | 269 | __wsum csum = 0; |
1da177e4 LT |
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); | |
763ecff1 YK |
274 | if (!(msg->type & ICMPV6_INFOMSG_MASK)) |
275 | nf_ct_attach(skb, org_skb); | |
1da177e4 LT |
276 | return 0; |
277 | } | |
278 | ||
59fbb3a6 | 279 | #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) |
79383236 MN |
280 | static void mip6_addr_swap(struct sk_buff *skb) |
281 | { | |
0660e03f | 282 | struct ipv6hdr *iph = ipv6_hdr(skb); |
79383236 MN |
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)) { | |
d56f90a7 ACM |
291 | hao = (struct ipv6_destopt_hao *) |
292 | (skb_network_header(skb) + off); | |
79383236 MN |
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 | |
300 | static inline void mip6_addr_swap(struct sk_buff *skb) {} | |
301 | #endif | |
302 | ||
1da177e4 LT |
303 | /* |
304 | * Send an ICMP message in response to a packet in error | |
305 | */ | |
1ab1457c | 306 | void icmpv6_send(struct sk_buff *skb, int type, int code, __u32 info, |
1da177e4 LT |
307 | struct net_device *dev) |
308 | { | |
9a43b709 | 309 | struct net *net = skb->dev->nd_net; |
1da177e4 | 310 | struct inet6_dev *idev = NULL; |
0660e03f | 311 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
84427d53 YH |
312 | struct sock *sk; |
313 | struct ipv6_pinfo *np; | |
1da177e4 LT |
314 | struct in6_addr *saddr = NULL; |
315 | struct dst_entry *dst; | |
8b7817f3 | 316 | struct dst_entry *dst2; |
1da177e4 LT |
317 | struct icmp6hdr tmp_hdr; |
318 | struct flowi fl; | |
8b7817f3 | 319 | struct flowi fl2; |
1da177e4 LT |
320 | struct icmpv6_msg msg; |
321 | int iif = 0; | |
322 | int addr_type = 0; | |
323 | int len; | |
41a1f8ea | 324 | int hlimit, tclass; |
1da177e4 LT |
325 | int err = 0; |
326 | ||
27a884dc ACM |
327 | if ((u8 *)hdr < skb->head || |
328 | (skb->network_header + sizeof(*hdr)) > skb->tail) | |
1da177e4 LT |
329 | return; |
330 | ||
331 | /* | |
1ab1457c | 332 | * Make sure we respect the rules |
1da177e4 LT |
333 | * i.e. RFC 1885 2.4(e) |
334 | * Rule (e.1) is enforced by not using icmpv6_send | |
335 | * in any code that processes icmp errors. | |
336 | */ | |
337 | addr_type = ipv6_addr_type(&hdr->daddr); | |
338 | ||
9a43b709 | 339 | if (ipv6_chk_addr(net, &hdr->daddr, skb->dev, 0)) |
1da177e4 LT |
340 | saddr = &hdr->daddr; |
341 | ||
342 | /* | |
343 | * Dest addr check | |
344 | */ | |
345 | ||
346 | if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) { | |
347 | if (type != ICMPV6_PKT_TOOBIG && | |
1ab1457c YH |
348 | !(type == ICMPV6_PARAMPROB && |
349 | code == ICMPV6_UNK_OPTION && | |
1da177e4 LT |
350 | (opt_unrec(skb, info)))) |
351 | return; | |
352 | ||
353 | saddr = NULL; | |
354 | } | |
355 | ||
356 | addr_type = ipv6_addr_type(&hdr->saddr); | |
357 | ||
358 | /* | |
359 | * Source addr check | |
360 | */ | |
361 | ||
362 | if (addr_type & IPV6_ADDR_LINKLOCAL) | |
363 | iif = skb->dev->ifindex; | |
364 | ||
365 | /* | |
8de3351e YH |
366 | * Must not send error if the source does not uniquely |
367 | * identify a single node (RFC2463 Section 2.4). | |
368 | * We check unspecified / multicast addresses here, | |
369 | * and anycast addresses will be checked later. | |
1da177e4 LT |
370 | */ |
371 | if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) { | |
64ce2073 | 372 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: addr_any/mcast source\n"); |
1da177e4 LT |
373 | return; |
374 | } | |
375 | ||
1ab1457c | 376 | /* |
1da177e4 LT |
377 | * Never answer to a ICMP packet. |
378 | */ | |
379 | if (is_ineligible(skb)) { | |
64ce2073 | 380 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: no reply to icmp error\n"); |
1da177e4 LT |
381 | return; |
382 | } | |
383 | ||
79383236 MN |
384 | mip6_addr_swap(skb); |
385 | ||
1da177e4 LT |
386 | memset(&fl, 0, sizeof(fl)); |
387 | fl.proto = IPPROTO_ICMPV6; | |
388 | ipv6_addr_copy(&fl.fl6_dst, &hdr->saddr); | |
389 | if (saddr) | |
390 | ipv6_addr_copy(&fl.fl6_src, saddr); | |
391 | fl.oif = iif; | |
392 | fl.fl_icmp_type = type; | |
393 | fl.fl_icmp_code = code; | |
beb8d13b | 394 | security_skb_classify_flow(skb, &fl); |
1da177e4 | 395 | |
9a43b709 | 396 | sk = icmpv6_sk(net); |
84427d53 YH |
397 | np = inet6_sk(sk); |
398 | ||
405666db DL |
399 | if (icmpv6_xmit_lock(sk)) |
400 | return; | |
401 | ||
1da177e4 LT |
402 | if (!icmpv6_xrlim_allow(sk, type, &fl)) |
403 | goto out; | |
404 | ||
405 | tmp_hdr.icmp6_type = type; | |
406 | tmp_hdr.icmp6_code = code; | |
407 | tmp_hdr.icmp6_cksum = 0; | |
408 | tmp_hdr.icmp6_pointer = htonl(info); | |
409 | ||
410 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) | |
411 | fl.oif = np->mcast_oif; | |
412 | ||
413 | err = ip6_dst_lookup(sk, &dst, &fl); | |
414 | if (err) | |
415 | goto out; | |
8de3351e YH |
416 | |
417 | /* | |
418 | * We won't send icmp if the destination is known | |
419 | * anycast. | |
420 | */ | |
421 | if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) { | |
422 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: acast source\n"); | |
423 | goto out_dst_release; | |
424 | } | |
425 | ||
8b7817f3 HX |
426 | /* No need to clone since we're just using its address. */ |
427 | dst2 = dst; | |
428 | ||
429 | err = xfrm_lookup(&dst, &fl, sk, 0); | |
430 | switch (err) { | |
431 | case 0: | |
432 | if (dst != dst2) | |
433 | goto route_done; | |
434 | break; | |
435 | case -EPERM: | |
436 | dst = NULL; | |
437 | break; | |
438 | default: | |
439 | goto out; | |
440 | } | |
441 | ||
442 | if (xfrm_decode_session_reverse(skb, &fl2, AF_INET6)) | |
443 | goto out; | |
444 | ||
445 | if (ip6_dst_lookup(sk, &dst2, &fl)) | |
e104411b | 446 | goto out; |
1da177e4 | 447 | |
8b7817f3 HX |
448 | err = xfrm_lookup(&dst2, &fl, sk, XFRM_LOOKUP_ICMP); |
449 | if (err == -ENOENT) { | |
450 | if (!dst) | |
451 | goto out; | |
452 | goto route_done; | |
453 | } | |
454 | ||
455 | dst_release(dst); | |
456 | dst = dst2; | |
457 | ||
458 | if (err) | |
459 | goto out; | |
460 | ||
461 | route_done: | |
1da177e4 LT |
462 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) |
463 | hlimit = np->mcast_hops; | |
464 | else | |
465 | hlimit = np->hop_limit; | |
466 | if (hlimit < 0) | |
6b75d090 | 467 | hlimit = ip6_dst_hoplimit(dst); |
1da177e4 | 468 | |
e012d51c | 469 | tclass = np->tclass; |
41a1f8ea YH |
470 | if (tclass < 0) |
471 | tclass = 0; | |
472 | ||
1da177e4 | 473 | msg.skb = skb; |
bbe735e4 | 474 | msg.offset = skb_network_offset(skb); |
763ecff1 | 475 | msg.type = type; |
1da177e4 LT |
476 | |
477 | len = skb->len - msg.offset; | |
478 | len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr)); | |
479 | if (len < 0) { | |
64ce2073 | 480 | LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n"); |
1da177e4 LT |
481 | goto out_dst_release; |
482 | } | |
483 | ||
484 | idev = in6_dev_get(skb->dev); | |
485 | ||
486 | err = ip6_append_data(sk, icmpv6_getfrag, &msg, | |
487 | len + sizeof(struct icmp6hdr), | |
488 | sizeof(struct icmp6hdr), | |
41a1f8ea | 489 | hlimit, tclass, NULL, &fl, (struct rt6_info*)dst, |
1da177e4 LT |
490 | MSG_DONTWAIT); |
491 | if (err) { | |
492 | ip6_flush_pending_frames(sk); | |
493 | goto out_put; | |
494 | } | |
495 | err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, len + sizeof(struct icmp6hdr)); | |
496 | ||
1da177e4 LT |
497 | out_put: |
498 | if (likely(idev != NULL)) | |
499 | in6_dev_put(idev); | |
500 | out_dst_release: | |
501 | dst_release(dst); | |
502 | out: | |
405666db | 503 | icmpv6_xmit_unlock(sk); |
1da177e4 LT |
504 | } |
505 | ||
7159039a YH |
506 | EXPORT_SYMBOL(icmpv6_send); |
507 | ||
1da177e4 LT |
508 | static void icmpv6_echo_reply(struct sk_buff *skb) |
509 | { | |
9a43b709 | 510 | struct net *net = skb->dev->nd_net; |
84427d53 | 511 | struct sock *sk; |
1da177e4 | 512 | struct inet6_dev *idev; |
84427d53 | 513 | struct ipv6_pinfo *np; |
1da177e4 | 514 | struct in6_addr *saddr = NULL; |
cc70ab26 | 515 | struct icmp6hdr *icmph = icmp6_hdr(skb); |
1da177e4 LT |
516 | struct icmp6hdr tmp_hdr; |
517 | struct flowi fl; | |
518 | struct icmpv6_msg msg; | |
519 | struct dst_entry *dst; | |
520 | int err = 0; | |
521 | int hlimit; | |
41a1f8ea | 522 | int tclass; |
1da177e4 | 523 | |
0660e03f | 524 | saddr = &ipv6_hdr(skb)->daddr; |
1da177e4 LT |
525 | |
526 | if (!ipv6_unicast_destination(skb)) | |
527 | saddr = NULL; | |
528 | ||
529 | memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr)); | |
530 | tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY; | |
531 | ||
532 | memset(&fl, 0, sizeof(fl)); | |
533 | fl.proto = IPPROTO_ICMPV6; | |
0660e03f | 534 | ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr); |
1da177e4 LT |
535 | if (saddr) |
536 | ipv6_addr_copy(&fl.fl6_src, saddr); | |
537 | fl.oif = skb->dev->ifindex; | |
538 | fl.fl_icmp_type = ICMPV6_ECHO_REPLY; | |
beb8d13b | 539 | security_skb_classify_flow(skb, &fl); |
1da177e4 | 540 | |
9a43b709 | 541 | sk = icmpv6_sk(net); |
84427d53 YH |
542 | np = inet6_sk(sk); |
543 | ||
405666db DL |
544 | if (icmpv6_xmit_lock(sk)) |
545 | return; | |
546 | ||
1da177e4 LT |
547 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) |
548 | fl.oif = np->mcast_oif; | |
549 | ||
550 | err = ip6_dst_lookup(sk, &dst, &fl); | |
551 | if (err) | |
552 | goto out; | |
553 | if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) | |
e104411b | 554 | goto out; |
1da177e4 LT |
555 | |
556 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) | |
557 | hlimit = np->mcast_hops; | |
558 | else | |
559 | hlimit = np->hop_limit; | |
560 | if (hlimit < 0) | |
6b75d090 | 561 | hlimit = ip6_dst_hoplimit(dst); |
1da177e4 | 562 | |
e012d51c | 563 | tclass = np->tclass; |
41a1f8ea YH |
564 | if (tclass < 0) |
565 | tclass = 0; | |
566 | ||
1da177e4 LT |
567 | idev = in6_dev_get(skb->dev); |
568 | ||
569 | msg.skb = skb; | |
570 | msg.offset = 0; | |
763ecff1 | 571 | msg.type = ICMPV6_ECHO_REPLY; |
1da177e4 LT |
572 | |
573 | err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr), | |
41a1f8ea | 574 | sizeof(struct icmp6hdr), hlimit, tclass, NULL, &fl, |
1da177e4 LT |
575 | (struct rt6_info*)dst, MSG_DONTWAIT); |
576 | ||
577 | if (err) { | |
578 | ip6_flush_pending_frames(sk); | |
579 | goto out_put; | |
580 | } | |
581 | err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, skb->len + sizeof(struct icmp6hdr)); | |
582 | ||
1ab1457c | 583 | out_put: |
1da177e4 LT |
584 | if (likely(idev != NULL)) |
585 | in6_dev_put(idev); | |
1da177e4 | 586 | dst_release(dst); |
1ab1457c | 587 | out: |
405666db | 588 | icmpv6_xmit_unlock(sk); |
1da177e4 LT |
589 | } |
590 | ||
04ce6909 | 591 | static void icmpv6_notify(struct sk_buff *skb, int type, int code, __be32 info) |
1da177e4 | 592 | { |
1da177e4 | 593 | struct inet6_protocol *ipprot; |
1da177e4 LT |
594 | int inner_offset; |
595 | int hash; | |
596 | u8 nexthdr; | |
597 | ||
598 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
599 | return; | |
600 | ||
601 | nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr; | |
602 | if (ipv6_ext_hdr(nexthdr)) { | |
603 | /* now skip over extension headers */ | |
0d3d077c | 604 | inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr); |
1da177e4 LT |
605 | if (inner_offset<0) |
606 | return; | |
607 | } else { | |
608 | inner_offset = sizeof(struct ipv6hdr); | |
609 | } | |
610 | ||
611 | /* Checkin header including 8 bytes of inner protocol header. */ | |
612 | if (!pskb_may_pull(skb, inner_offset+8)) | |
613 | return; | |
614 | ||
1da177e4 LT |
615 | /* BUGGG_FUTURE: we should try to parse exthdrs in this packet. |
616 | Without this we will not able f.e. to make source routed | |
617 | pmtu discovery. | |
618 | Corresponding argument (opt) to notifiers is already added. | |
619 | --ANK (980726) | |
620 | */ | |
621 | ||
622 | hash = nexthdr & (MAX_INET_PROTOS - 1); | |
623 | ||
624 | rcu_read_lock(); | |
625 | ipprot = rcu_dereference(inet6_protos[hash]); | |
626 | if (ipprot && ipprot->err_handler) | |
627 | ipprot->err_handler(skb, NULL, type, code, inner_offset, info); | |
628 | rcu_read_unlock(); | |
629 | ||
69d6da0b | 630 | raw6_icmp_error(skb, nexthdr, type, code, inner_offset, info); |
1da177e4 | 631 | } |
1ab1457c | 632 | |
1da177e4 LT |
633 | /* |
634 | * Handle icmp messages | |
635 | */ | |
636 | ||
e5bbef20 | 637 | static int icmpv6_rcv(struct sk_buff *skb) |
1da177e4 | 638 | { |
1da177e4 LT |
639 | struct net_device *dev = skb->dev; |
640 | struct inet6_dev *idev = __in6_dev_get(dev); | |
641 | struct in6_addr *saddr, *daddr; | |
642 | struct ipv6hdr *orig_hdr; | |
643 | struct icmp6hdr *hdr; | |
644 | int type; | |
645 | ||
aebcf82c | 646 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { |
8b7817f3 HX |
647 | int nh; |
648 | ||
aebcf82c HX |
649 | if (!(skb->sp && skb->sp->xvec[skb->sp->len - 1]->props.flags & |
650 | XFRM_STATE_ICMP)) | |
651 | goto drop_no_count; | |
652 | ||
8b7817f3 HX |
653 | if (!pskb_may_pull(skb, sizeof(*hdr) + sizeof(*orig_hdr))) |
654 | goto drop_no_count; | |
655 | ||
656 | nh = skb_network_offset(skb); | |
657 | skb_set_network_header(skb, sizeof(*hdr)); | |
658 | ||
659 | if (!xfrm6_policy_check_reverse(NULL, XFRM_POLICY_IN, skb)) | |
660 | goto drop_no_count; | |
661 | ||
662 | skb_set_network_header(skb, nh); | |
663 | } | |
664 | ||
1da177e4 LT |
665 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INMSGS); |
666 | ||
0660e03f ACM |
667 | saddr = &ipv6_hdr(skb)->saddr; |
668 | daddr = &ipv6_hdr(skb)->daddr; | |
1da177e4 LT |
669 | |
670 | /* Perform checksum. */ | |
fb286bb2 | 671 | switch (skb->ip_summed) { |
84fa7933 | 672 | case CHECKSUM_COMPLETE: |
fb286bb2 HX |
673 | if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6, |
674 | skb->csum)) | |
675 | break; | |
676 | /* fall through */ | |
677 | case CHECKSUM_NONE: | |
868c86bc AV |
678 | skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len, |
679 | IPPROTO_ICMPV6, 0)); | |
fb286bb2 | 680 | if (__skb_checksum_complete(skb)) { |
46b86a2d | 681 | LIMIT_NETDEBUG(KERN_DEBUG "ICMPv6 checksum failed [" NIP6_FMT " > " NIP6_FMT "]\n", |
64ce2073 | 682 | NIP6(*saddr), NIP6(*daddr)); |
1da177e4 LT |
683 | goto discard_it; |
684 | } | |
685 | } | |
686 | ||
8cf22943 HX |
687 | if (!pskb_pull(skb, sizeof(*hdr))) |
688 | goto discard_it; | |
1da177e4 | 689 | |
cc70ab26 | 690 | hdr = icmp6_hdr(skb); |
1da177e4 LT |
691 | |
692 | type = hdr->icmp6_type; | |
693 | ||
14878f75 | 694 | ICMP6MSGIN_INC_STATS_BH(idev, type); |
1da177e4 LT |
695 | |
696 | switch (type) { | |
697 | case ICMPV6_ECHO_REQUEST: | |
698 | icmpv6_echo_reply(skb); | |
699 | break; | |
700 | ||
701 | case ICMPV6_ECHO_REPLY: | |
702 | /* we couldn't care less */ | |
703 | break; | |
704 | ||
705 | case ICMPV6_PKT_TOOBIG: | |
706 | /* BUGGG_FUTURE: if packet contains rthdr, we cannot update | |
707 | standard destination cache. Seems, only "advanced" | |
708 | destination cache will allow to solve this problem | |
709 | --ANK (980726) | |
710 | */ | |
711 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
712 | goto discard_it; | |
cc70ab26 | 713 | hdr = icmp6_hdr(skb); |
1da177e4 LT |
714 | orig_hdr = (struct ipv6hdr *) (hdr + 1); |
715 | rt6_pmtu_discovery(&orig_hdr->daddr, &orig_hdr->saddr, dev, | |
716 | ntohl(hdr->icmp6_mtu)); | |
717 | ||
718 | /* | |
719 | * Drop through to notify | |
720 | */ | |
721 | ||
722 | case ICMPV6_DEST_UNREACH: | |
723 | case ICMPV6_TIME_EXCEED: | |
724 | case ICMPV6_PARAMPROB: | |
725 | icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); | |
726 | break; | |
727 | ||
728 | case NDISC_ROUTER_SOLICITATION: | |
729 | case NDISC_ROUTER_ADVERTISEMENT: | |
730 | case NDISC_NEIGHBOUR_SOLICITATION: | |
731 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
732 | case NDISC_REDIRECT: | |
733 | ndisc_rcv(skb); | |
734 | break; | |
735 | ||
736 | case ICMPV6_MGM_QUERY: | |
737 | igmp6_event_query(skb); | |
738 | break; | |
739 | ||
740 | case ICMPV6_MGM_REPORT: | |
741 | igmp6_event_report(skb); | |
742 | break; | |
743 | ||
744 | case ICMPV6_MGM_REDUCTION: | |
745 | case ICMPV6_NI_QUERY: | |
746 | case ICMPV6_NI_REPLY: | |
747 | case ICMPV6_MLD2_REPORT: | |
748 | case ICMPV6_DHAAD_REQUEST: | |
749 | case ICMPV6_DHAAD_REPLY: | |
750 | case ICMPV6_MOBILE_PREFIX_SOL: | |
751 | case ICMPV6_MOBILE_PREFIX_ADV: | |
752 | break; | |
753 | ||
754 | default: | |
64ce2073 | 755 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n"); |
1da177e4 LT |
756 | |
757 | /* informational */ | |
758 | if (type & ICMPV6_INFOMSG_MASK) | |
759 | break; | |
760 | ||
1ab1457c YH |
761 | /* |
762 | * error of unknown type. | |
763 | * must pass to upper level | |
1da177e4 LT |
764 | */ |
765 | ||
766 | icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); | |
3ff50b79 SH |
767 | } |
768 | ||
1da177e4 LT |
769 | kfree_skb(skb); |
770 | return 0; | |
771 | ||
772 | discard_it: | |
773 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INERRORS); | |
8b7817f3 | 774 | drop_no_count: |
1da177e4 LT |
775 | kfree_skb(skb); |
776 | return 0; | |
777 | } | |
778 | ||
95e41e93 YH |
779 | void icmpv6_flow_init(struct sock *sk, struct flowi *fl, |
780 | u8 type, | |
781 | const struct in6_addr *saddr, | |
782 | const struct in6_addr *daddr, | |
783 | int oif) | |
784 | { | |
785 | memset(fl, 0, sizeof(*fl)); | |
786 | ipv6_addr_copy(&fl->fl6_src, saddr); | |
787 | ipv6_addr_copy(&fl->fl6_dst, daddr); | |
788 | fl->proto = IPPROTO_ICMPV6; | |
789 | fl->fl_icmp_type = type; | |
790 | fl->fl_icmp_code = 0; | |
791 | fl->oif = oif; | |
792 | security_sk_classify_flow(sk, fl); | |
793 | } | |
794 | ||
640c41c7 | 795 | /* |
b7e729c4 | 796 | * Special lock-class for __icmpv6_sk: |
640c41c7 IM |
797 | */ |
798 | static struct lock_class_key icmpv6_socket_sk_dst_lock_key; | |
799 | ||
98c6d1b2 | 800 | static int __net_init icmpv6_sk_init(struct net *net) |
1da177e4 LT |
801 | { |
802 | struct sock *sk; | |
803 | int err, i, j; | |
804 | ||
98c6d1b2 DL |
805 | net->ipv6.icmp_sk = |
806 | kzalloc(nr_cpu_ids * sizeof(struct sock *), GFP_KERNEL); | |
807 | if (net->ipv6.icmp_sk == NULL) | |
79c91159 DL |
808 | return -ENOMEM; |
809 | ||
6f912042 | 810 | for_each_possible_cpu(i) { |
b7e729c4 | 811 | struct socket *sock; |
1da177e4 | 812 | err = sock_create_kern(PF_INET6, SOCK_RAW, IPPROTO_ICMPV6, |
b7e729c4 | 813 | &sock); |
1da177e4 LT |
814 | if (err < 0) { |
815 | printk(KERN_ERR | |
816 | "Failed to initialize the ICMP6 control socket " | |
817 | "(err %d).\n", | |
818 | err); | |
819 | goto fail; | |
820 | } | |
821 | ||
98c6d1b2 DL |
822 | net->ipv6.icmp_sk[i] = sk = sock->sk; |
823 | sk_change_net(sk, net); | |
5c8cafd6 | 824 | |
1da177e4 | 825 | sk->sk_allocation = GFP_ATOMIC; |
640c41c7 IM |
826 | /* |
827 | * Split off their lock-class, because sk->sk_dst_lock | |
828 | * gets used from softirqs, which is safe for | |
b7e729c4 | 829 | * __icmpv6_sk (because those never get directly used |
640c41c7 IM |
830 | * via userspace syscalls), but unsafe for normal sockets. |
831 | */ | |
832 | lockdep_set_class(&sk->sk_dst_lock, | |
833 | &icmpv6_socket_sk_dst_lock_key); | |
1da177e4 LT |
834 | |
835 | /* Enough space for 2 64K ICMP packets, including | |
836 | * sk_buff struct overhead. | |
837 | */ | |
838 | sk->sk_sndbuf = | |
839 | (2 * ((64 * 1024) + sizeof(struct sk_buff))); | |
840 | ||
841 | sk->sk_prot->unhash(sk); | |
842 | } | |
1da177e4 LT |
843 | return 0; |
844 | ||
845 | fail: | |
5c8cafd6 | 846 | for (j = 0; j < i; j++) |
98c6d1b2 DL |
847 | sk_release_kernel(net->ipv6.icmp_sk[j]); |
848 | kfree(net->ipv6.icmp_sk); | |
1da177e4 LT |
849 | return err; |
850 | } | |
851 | ||
98c6d1b2 | 852 | static void __net_exit icmpv6_sk_exit(struct net *net) |
1da177e4 LT |
853 | { |
854 | int i; | |
855 | ||
6f912042 | 856 | for_each_possible_cpu(i) { |
98c6d1b2 | 857 | sk_release_kernel(net->ipv6.icmp_sk[i]); |
1da177e4 | 858 | } |
98c6d1b2 DL |
859 | kfree(net->ipv6.icmp_sk); |
860 | } | |
861 | ||
8ed7edce | 862 | static struct pernet_operations icmpv6_sk_ops = { |
98c6d1b2 DL |
863 | .init = icmpv6_sk_init, |
864 | .exit = icmpv6_sk_exit, | |
865 | }; | |
866 | ||
867 | int __init icmpv6_init(void) | |
868 | { | |
869 | int err; | |
870 | ||
871 | err = register_pernet_subsys(&icmpv6_sk_ops); | |
872 | if (err < 0) | |
873 | return err; | |
874 | ||
875 | err = -EAGAIN; | |
876 | if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0) | |
877 | goto fail; | |
878 | return 0; | |
879 | ||
880 | fail: | |
881 | printk(KERN_ERR "Failed to register ICMP6 protocol\n"); | |
882 | unregister_pernet_subsys(&icmpv6_sk_ops); | |
883 | return err; | |
884 | } | |
885 | ||
8ed7edce | 886 | void icmpv6_cleanup(void) |
98c6d1b2 DL |
887 | { |
888 | unregister_pernet_subsys(&icmpv6_sk_ops); | |
1da177e4 LT |
889 | inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); |
890 | } | |
891 | ||
98c6d1b2 | 892 | |
9b5b5cff | 893 | static const struct icmp6_err { |
1da177e4 LT |
894 | int err; |
895 | int fatal; | |
896 | } tab_unreach[] = { | |
897 | { /* NOROUTE */ | |
898 | .err = ENETUNREACH, | |
899 | .fatal = 0, | |
900 | }, | |
901 | { /* ADM_PROHIBITED */ | |
902 | .err = EACCES, | |
903 | .fatal = 1, | |
904 | }, | |
905 | { /* Was NOT_NEIGHBOUR, now reserved */ | |
906 | .err = EHOSTUNREACH, | |
907 | .fatal = 0, | |
908 | }, | |
909 | { /* ADDR_UNREACH */ | |
910 | .err = EHOSTUNREACH, | |
911 | .fatal = 0, | |
912 | }, | |
913 | { /* PORT_UNREACH */ | |
914 | .err = ECONNREFUSED, | |
915 | .fatal = 1, | |
916 | }, | |
917 | }; | |
918 | ||
919 | int icmpv6_err_convert(int type, int code, int *err) | |
920 | { | |
921 | int fatal = 0; | |
922 | ||
923 | *err = EPROTO; | |
924 | ||
925 | switch (type) { | |
926 | case ICMPV6_DEST_UNREACH: | |
927 | fatal = 1; | |
928 | if (code <= ICMPV6_PORT_UNREACH) { | |
929 | *err = tab_unreach[code].err; | |
930 | fatal = tab_unreach[code].fatal; | |
931 | } | |
932 | break; | |
933 | ||
934 | case ICMPV6_PKT_TOOBIG: | |
935 | *err = EMSGSIZE; | |
936 | break; | |
1ab1457c | 937 | |
1da177e4 LT |
938 | case ICMPV6_PARAMPROB: |
939 | *err = EPROTO; | |
940 | fatal = 1; | |
941 | break; | |
942 | ||
943 | case ICMPV6_TIME_EXCEED: | |
944 | *err = EHOSTUNREACH; | |
945 | break; | |
3ff50b79 | 946 | } |
1da177e4 LT |
947 | |
948 | return fatal; | |
949 | } | |
950 | ||
7159039a YH |
951 | EXPORT_SYMBOL(icmpv6_err_convert); |
952 | ||
1da177e4 | 953 | #ifdef CONFIG_SYSCTL |
760f2d01 | 954 | ctl_table ipv6_icmp_table_template[] = { |
1da177e4 LT |
955 | { |
956 | .ctl_name = NET_IPV6_ICMP_RATELIMIT, | |
957 | .procname = "ratelimit", | |
41a76906 | 958 | .data = &init_net.ipv6.sysctl.icmpv6_time, |
1da177e4 LT |
959 | .maxlen = sizeof(int), |
960 | .mode = 0644, | |
961 | .proc_handler = &proc_dointvec | |
962 | }, | |
963 | { .ctl_name = 0 }, | |
964 | }; | |
760f2d01 DL |
965 | |
966 | struct ctl_table *ipv6_icmp_sysctl_init(struct net *net) | |
967 | { | |
968 | struct ctl_table *table; | |
969 | ||
970 | table = kmemdup(ipv6_icmp_table_template, | |
971 | sizeof(ipv6_icmp_table_template), | |
972 | GFP_KERNEL); | |
5ee09105 YH |
973 | |
974 | if (table) | |
975 | table[0].data = &net->ipv6.sysctl.icmpv6_time; | |
976 | ||
760f2d01 DL |
977 | return table; |
978 | } | |
1da177e4 LT |
979 | #endif |
980 |