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> | |
40 | #include <linux/sched.h> | |
41 | #include <linux/sockios.h> | |
42 | #include <linux/net.h> | |
43 | #include <linux/skbuff.h> | |
44 | #include <linux/init.h> | |
763ecff1 | 45 | #include <linux/netfilter.h> |
1da177e4 LT |
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/protocol.h> | |
61 | #include <net/raw.h> | |
62 | #include <net/rawv6.h> | |
63 | #include <net/transp_v6.h> | |
64 | #include <net/ip6_route.h> | |
65 | #include <net/addrconf.h> | |
66 | #include <net/icmp.h> | |
67 | ||
68 | #include <asm/uaccess.h> | |
69 | #include <asm/system.h> | |
70 | ||
ba89966c | 71 | DEFINE_SNMP_STAT(struct icmpv6_mib, icmpv6_statistics) __read_mostly; |
1da177e4 LT |
72 | |
73 | /* | |
74 | * The ICMP socket(s). This is the most convenient way to flow control | |
75 | * our ICMP output as well as maintain a clean interface throughout | |
76 | * all layers. All Socketless IP sends will soon be gone. | |
77 | * | |
78 | * On SMP we have one ICMP socket per-cpu. | |
79 | */ | |
80 | static DEFINE_PER_CPU(struct socket *, __icmpv6_socket) = NULL; | |
81 | #define icmpv6_socket __get_cpu_var(__icmpv6_socket) | |
82 | ||
951dbc8a | 83 | static int icmpv6_rcv(struct sk_buff **pskb); |
1da177e4 LT |
84 | |
85 | static struct inet6_protocol icmpv6_protocol = { | |
86 | .handler = icmpv6_rcv, | |
87 | .flags = INET6_PROTO_FINAL, | |
88 | }; | |
89 | ||
90 | static __inline__ int icmpv6_xmit_lock(void) | |
91 | { | |
92 | local_bh_disable(); | |
93 | ||
94 | if (unlikely(!spin_trylock(&icmpv6_socket->sk->sk_lock.slock))) { | |
95 | /* This can happen if the output path (f.e. SIT or | |
96 | * ip6ip6 tunnel) signals dst_link_failure() for an | |
97 | * outgoing ICMP6 packet. | |
98 | */ | |
99 | local_bh_enable(); | |
100 | return 1; | |
101 | } | |
102 | return 0; | |
103 | } | |
104 | ||
105 | static __inline__ void icmpv6_xmit_unlock(void) | |
106 | { | |
107 | spin_unlock_bh(&icmpv6_socket->sk->sk_lock.slock); | |
108 | } | |
109 | ||
110 | /* | |
111 | * Slightly more convenient version of icmpv6_send. | |
112 | */ | |
113 | void icmpv6_param_prob(struct sk_buff *skb, int code, int pos) | |
114 | { | |
115 | icmpv6_send(skb, ICMPV6_PARAMPROB, code, pos, skb->dev); | |
116 | kfree_skb(skb); | |
117 | } | |
118 | ||
119 | /* | |
120 | * Figure out, may we reply to this packet with icmp error. | |
121 | * | |
122 | * We do not reply, if: | |
123 | * - it was icmp error message. | |
124 | * - it is truncated, so that it is known, that protocol is ICMPV6 | |
125 | * (i.e. in the middle of some exthdr) | |
126 | * | |
127 | * --ANK (980726) | |
128 | */ | |
129 | ||
130 | static int is_ineligible(struct sk_buff *skb) | |
131 | { | |
132 | int ptr = (u8*)(skb->nh.ipv6h+1) - skb->data; | |
133 | int len = skb->len - ptr; | |
134 | __u8 nexthdr = skb->nh.ipv6h->nexthdr; | |
135 | ||
136 | if (len < 0) | |
137 | return 1; | |
138 | ||
0d3d077c | 139 | ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr); |
1da177e4 LT |
140 | if (ptr < 0) |
141 | return 0; | |
142 | if (nexthdr == IPPROTO_ICMPV6) { | |
143 | u8 _type, *tp; | |
144 | tp = skb_header_pointer(skb, | |
145 | ptr+offsetof(struct icmp6hdr, icmp6_type), | |
146 | sizeof(_type), &_type); | |
147 | if (tp == NULL || | |
148 | !(*tp & ICMPV6_INFOMSG_MASK)) | |
149 | return 1; | |
150 | } | |
151 | return 0; | |
152 | } | |
153 | ||
ab32ea5d | 154 | static int sysctl_icmpv6_time __read_mostly = 1*HZ; |
1da177e4 LT |
155 | |
156 | /* | |
157 | * Check the ICMP output rate limit | |
158 | */ | |
159 | static inline int icmpv6_xrlim_allow(struct sock *sk, int type, | |
160 | struct flowi *fl) | |
161 | { | |
162 | struct dst_entry *dst; | |
163 | int res = 0; | |
164 | ||
165 | /* Informational messages are not limited. */ | |
166 | if (type & ICMPV6_INFOMSG_MASK) | |
167 | return 1; | |
168 | ||
169 | /* Do not limit pmtu discovery, it would break it. */ | |
170 | if (type == ICMPV6_PKT_TOOBIG) | |
171 | return 1; | |
172 | ||
173 | /* | |
174 | * Look up the output route. | |
175 | * XXX: perhaps the expire for routing entries cloned by | |
176 | * this lookup should be more aggressive (not longer than timeout). | |
177 | */ | |
178 | dst = ip6_route_output(sk, fl); | |
179 | if (dst->error) { | |
a11d206d YH |
180 | IP6_INC_STATS(ip6_dst_idev(dst), |
181 | IPSTATS_MIB_OUTNOROUTES); | |
1da177e4 LT |
182 | } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) { |
183 | res = 1; | |
184 | } else { | |
185 | struct rt6_info *rt = (struct rt6_info *)dst; | |
186 | int tmo = sysctl_icmpv6_time; | |
187 | ||
188 | /* Give more bandwidth to wider prefixes. */ | |
189 | if (rt->rt6i_dst.plen < 128) | |
190 | tmo >>= ((128 - rt->rt6i_dst.plen)>>5); | |
191 | ||
192 | res = xrlim_allow(dst, tmo); | |
193 | } | |
194 | dst_release(dst); | |
195 | return res; | |
196 | } | |
197 | ||
198 | /* | |
199 | * an inline helper for the "simple" if statement below | |
200 | * checks if parameter problem report is caused by an | |
201 | * unrecognized IPv6 option that has the Option Type | |
202 | * highest-order two bits set to 10 | |
203 | */ | |
204 | ||
205 | static __inline__ int opt_unrec(struct sk_buff *skb, __u32 offset) | |
206 | { | |
207 | u8 _optval, *op; | |
208 | ||
209 | offset += skb->nh.raw - skb->data; | |
210 | op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval); | |
211 | if (op == NULL) | |
212 | return 1; | |
213 | return (*op & 0xC0) == 0x80; | |
214 | } | |
215 | ||
216 | static int icmpv6_push_pending_frames(struct sock *sk, struct flowi *fl, struct icmp6hdr *thdr, int len) | |
217 | { | |
218 | struct sk_buff *skb; | |
219 | struct icmp6hdr *icmp6h; | |
220 | int err = 0; | |
221 | ||
222 | if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) | |
223 | goto out; | |
224 | ||
225 | icmp6h = (struct icmp6hdr*) skb->h.raw; | |
226 | memcpy(icmp6h, thdr, sizeof(struct icmp6hdr)); | |
227 | icmp6h->icmp6_cksum = 0; | |
228 | ||
229 | if (skb_queue_len(&sk->sk_write_queue) == 1) { | |
230 | skb->csum = csum_partial((char *)icmp6h, | |
231 | sizeof(struct icmp6hdr), skb->csum); | |
232 | icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src, | |
233 | &fl->fl6_dst, | |
234 | len, fl->proto, | |
235 | skb->csum); | |
236 | } else { | |
868c86bc | 237 | __wsum tmp_csum = 0; |
1da177e4 LT |
238 | |
239 | skb_queue_walk(&sk->sk_write_queue, skb) { | |
240 | tmp_csum = csum_add(tmp_csum, skb->csum); | |
241 | } | |
242 | ||
243 | tmp_csum = csum_partial((char *)icmp6h, | |
244 | sizeof(struct icmp6hdr), tmp_csum); | |
868c86bc AV |
245 | icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src, |
246 | &fl->fl6_dst, | |
247 | len, fl->proto, | |
248 | tmp_csum); | |
1da177e4 | 249 | } |
1da177e4 LT |
250 | ip6_push_pending_frames(sk); |
251 | out: | |
252 | return err; | |
253 | } | |
254 | ||
255 | struct icmpv6_msg { | |
256 | struct sk_buff *skb; | |
257 | int offset; | |
763ecff1 | 258 | uint8_t type; |
1da177e4 LT |
259 | }; |
260 | ||
261 | static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb) | |
262 | { | |
263 | struct icmpv6_msg *msg = (struct icmpv6_msg *) from; | |
264 | struct sk_buff *org_skb = msg->skb; | |
5f92a738 | 265 | __wsum csum = 0; |
1da177e4 LT |
266 | |
267 | csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset, | |
268 | to, len, csum); | |
269 | skb->csum = csum_block_add(skb->csum, csum, odd); | |
763ecff1 YK |
270 | if (!(msg->type & ICMPV6_INFOMSG_MASK)) |
271 | nf_ct_attach(skb, org_skb); | |
1da177e4 LT |
272 | return 0; |
273 | } | |
274 | ||
79383236 MN |
275 | #ifdef CONFIG_IPV6_MIP6 |
276 | static void mip6_addr_swap(struct sk_buff *skb) | |
277 | { | |
278 | struct ipv6hdr *iph = skb->nh.ipv6h; | |
279 | struct inet6_skb_parm *opt = IP6CB(skb); | |
280 | struct ipv6_destopt_hao *hao; | |
281 | struct in6_addr tmp; | |
282 | int off; | |
283 | ||
284 | if (opt->dsthao) { | |
285 | off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO); | |
286 | if (likely(off >= 0)) { | |
287 | hao = (struct ipv6_destopt_hao *)(skb->nh.raw + off); | |
288 | ipv6_addr_copy(&tmp, &iph->saddr); | |
289 | ipv6_addr_copy(&iph->saddr, &hao->addr); | |
290 | ipv6_addr_copy(&hao->addr, &tmp); | |
291 | } | |
292 | } | |
293 | } | |
294 | #else | |
295 | static inline void mip6_addr_swap(struct sk_buff *skb) {} | |
296 | #endif | |
297 | ||
1da177e4 LT |
298 | /* |
299 | * Send an ICMP message in response to a packet in error | |
300 | */ | |
301 | void icmpv6_send(struct sk_buff *skb, int type, int code, __u32 info, | |
302 | struct net_device *dev) | |
303 | { | |
304 | struct inet6_dev *idev = NULL; | |
305 | struct ipv6hdr *hdr = skb->nh.ipv6h; | |
84427d53 YH |
306 | struct sock *sk; |
307 | struct ipv6_pinfo *np; | |
1da177e4 LT |
308 | struct in6_addr *saddr = NULL; |
309 | struct dst_entry *dst; | |
310 | struct icmp6hdr tmp_hdr; | |
311 | struct flowi fl; | |
312 | struct icmpv6_msg msg; | |
313 | int iif = 0; | |
314 | int addr_type = 0; | |
315 | int len; | |
41a1f8ea | 316 | int hlimit, tclass; |
1da177e4 LT |
317 | int err = 0; |
318 | ||
319 | if ((u8*)hdr < skb->head || (u8*)(hdr+1) > skb->tail) | |
320 | return; | |
321 | ||
322 | /* | |
323 | * Make sure we respect the rules | |
324 | * i.e. RFC 1885 2.4(e) | |
325 | * Rule (e.1) is enforced by not using icmpv6_send | |
326 | * in any code that processes icmp errors. | |
327 | */ | |
328 | addr_type = ipv6_addr_type(&hdr->daddr); | |
329 | ||
330 | if (ipv6_chk_addr(&hdr->daddr, skb->dev, 0)) | |
331 | saddr = &hdr->daddr; | |
332 | ||
333 | /* | |
334 | * Dest addr check | |
335 | */ | |
336 | ||
337 | if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) { | |
338 | if (type != ICMPV6_PKT_TOOBIG && | |
339 | !(type == ICMPV6_PARAMPROB && | |
340 | code == ICMPV6_UNK_OPTION && | |
341 | (opt_unrec(skb, info)))) | |
342 | return; | |
343 | ||
344 | saddr = NULL; | |
345 | } | |
346 | ||
347 | addr_type = ipv6_addr_type(&hdr->saddr); | |
348 | ||
349 | /* | |
350 | * Source addr check | |
351 | */ | |
352 | ||
353 | if (addr_type & IPV6_ADDR_LINKLOCAL) | |
354 | iif = skb->dev->ifindex; | |
355 | ||
356 | /* | |
8de3351e YH |
357 | * Must not send error if the source does not uniquely |
358 | * identify a single node (RFC2463 Section 2.4). | |
359 | * We check unspecified / multicast addresses here, | |
360 | * and anycast addresses will be checked later. | |
1da177e4 LT |
361 | */ |
362 | if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) { | |
64ce2073 | 363 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: addr_any/mcast source\n"); |
1da177e4 LT |
364 | return; |
365 | } | |
366 | ||
367 | /* | |
368 | * Never answer to a ICMP packet. | |
369 | */ | |
370 | if (is_ineligible(skb)) { | |
64ce2073 | 371 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: no reply to icmp error\n"); |
1da177e4 LT |
372 | return; |
373 | } | |
374 | ||
79383236 MN |
375 | mip6_addr_swap(skb); |
376 | ||
1da177e4 LT |
377 | memset(&fl, 0, sizeof(fl)); |
378 | fl.proto = IPPROTO_ICMPV6; | |
379 | ipv6_addr_copy(&fl.fl6_dst, &hdr->saddr); | |
380 | if (saddr) | |
381 | ipv6_addr_copy(&fl.fl6_src, saddr); | |
382 | fl.oif = iif; | |
383 | fl.fl_icmp_type = type; | |
384 | fl.fl_icmp_code = code; | |
beb8d13b | 385 | security_skb_classify_flow(skb, &fl); |
1da177e4 LT |
386 | |
387 | if (icmpv6_xmit_lock()) | |
388 | return; | |
389 | ||
84427d53 YH |
390 | sk = icmpv6_socket->sk; |
391 | np = inet6_sk(sk); | |
392 | ||
1da177e4 LT |
393 | if (!icmpv6_xrlim_allow(sk, type, &fl)) |
394 | goto out; | |
395 | ||
396 | tmp_hdr.icmp6_type = type; | |
397 | tmp_hdr.icmp6_code = code; | |
398 | tmp_hdr.icmp6_cksum = 0; | |
399 | tmp_hdr.icmp6_pointer = htonl(info); | |
400 | ||
401 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) | |
402 | fl.oif = np->mcast_oif; | |
403 | ||
404 | err = ip6_dst_lookup(sk, &dst, &fl); | |
405 | if (err) | |
406 | goto out; | |
8de3351e YH |
407 | |
408 | /* | |
409 | * We won't send icmp if the destination is known | |
410 | * anycast. | |
411 | */ | |
412 | if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) { | |
413 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: acast source\n"); | |
414 | goto out_dst_release; | |
415 | } | |
416 | ||
1da177e4 | 417 | if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) |
e104411b | 418 | goto out; |
1da177e4 LT |
419 | |
420 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) | |
421 | hlimit = np->mcast_hops; | |
422 | else | |
423 | hlimit = np->hop_limit; | |
424 | if (hlimit < 0) | |
425 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | |
426 | if (hlimit < 0) | |
427 | hlimit = ipv6_get_hoplimit(dst->dev); | |
428 | ||
e012d51c | 429 | tclass = np->tclass; |
41a1f8ea YH |
430 | if (tclass < 0) |
431 | tclass = 0; | |
432 | ||
1da177e4 LT |
433 | msg.skb = skb; |
434 | msg.offset = skb->nh.raw - skb->data; | |
763ecff1 | 435 | msg.type = type; |
1da177e4 LT |
436 | |
437 | len = skb->len - msg.offset; | |
438 | len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr)); | |
439 | if (len < 0) { | |
64ce2073 | 440 | LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n"); |
1da177e4 LT |
441 | goto out_dst_release; |
442 | } | |
443 | ||
444 | idev = in6_dev_get(skb->dev); | |
445 | ||
446 | err = ip6_append_data(sk, icmpv6_getfrag, &msg, | |
447 | len + sizeof(struct icmp6hdr), | |
448 | sizeof(struct icmp6hdr), | |
41a1f8ea | 449 | hlimit, tclass, NULL, &fl, (struct rt6_info*)dst, |
1da177e4 LT |
450 | MSG_DONTWAIT); |
451 | if (err) { | |
452 | ip6_flush_pending_frames(sk); | |
453 | goto out_put; | |
454 | } | |
455 | err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, len + sizeof(struct icmp6hdr)); | |
456 | ||
457 | if (type >= ICMPV6_DEST_UNREACH && type <= ICMPV6_PARAMPROB) | |
458 | ICMP6_INC_STATS_OFFSET_BH(idev, ICMP6_MIB_OUTDESTUNREACHS, type - ICMPV6_DEST_UNREACH); | |
459 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTMSGS); | |
460 | ||
461 | out_put: | |
462 | if (likely(idev != NULL)) | |
463 | in6_dev_put(idev); | |
464 | out_dst_release: | |
465 | dst_release(dst); | |
466 | out: | |
467 | icmpv6_xmit_unlock(); | |
468 | } | |
469 | ||
470 | static void icmpv6_echo_reply(struct sk_buff *skb) | |
471 | { | |
84427d53 | 472 | struct sock *sk; |
1da177e4 | 473 | struct inet6_dev *idev; |
84427d53 | 474 | struct ipv6_pinfo *np; |
1da177e4 LT |
475 | struct in6_addr *saddr = NULL; |
476 | struct icmp6hdr *icmph = (struct icmp6hdr *) skb->h.raw; | |
477 | struct icmp6hdr tmp_hdr; | |
478 | struct flowi fl; | |
479 | struct icmpv6_msg msg; | |
480 | struct dst_entry *dst; | |
481 | int err = 0; | |
482 | int hlimit; | |
41a1f8ea | 483 | int tclass; |
1da177e4 LT |
484 | |
485 | saddr = &skb->nh.ipv6h->daddr; | |
486 | ||
487 | if (!ipv6_unicast_destination(skb)) | |
488 | saddr = NULL; | |
489 | ||
490 | memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr)); | |
491 | tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY; | |
492 | ||
493 | memset(&fl, 0, sizeof(fl)); | |
494 | fl.proto = IPPROTO_ICMPV6; | |
495 | ipv6_addr_copy(&fl.fl6_dst, &skb->nh.ipv6h->saddr); | |
496 | if (saddr) | |
497 | ipv6_addr_copy(&fl.fl6_src, saddr); | |
498 | fl.oif = skb->dev->ifindex; | |
499 | fl.fl_icmp_type = ICMPV6_ECHO_REPLY; | |
beb8d13b | 500 | security_skb_classify_flow(skb, &fl); |
1da177e4 LT |
501 | |
502 | if (icmpv6_xmit_lock()) | |
503 | return; | |
504 | ||
84427d53 YH |
505 | sk = icmpv6_socket->sk; |
506 | np = inet6_sk(sk); | |
507 | ||
1da177e4 LT |
508 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) |
509 | fl.oif = np->mcast_oif; | |
510 | ||
511 | err = ip6_dst_lookup(sk, &dst, &fl); | |
512 | if (err) | |
513 | goto out; | |
514 | if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) | |
e104411b | 515 | goto out; |
1da177e4 LT |
516 | |
517 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) | |
518 | hlimit = np->mcast_hops; | |
519 | else | |
520 | hlimit = np->hop_limit; | |
521 | if (hlimit < 0) | |
522 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | |
523 | if (hlimit < 0) | |
524 | hlimit = ipv6_get_hoplimit(dst->dev); | |
525 | ||
e012d51c | 526 | tclass = np->tclass; |
41a1f8ea YH |
527 | if (tclass < 0) |
528 | tclass = 0; | |
529 | ||
1da177e4 LT |
530 | idev = in6_dev_get(skb->dev); |
531 | ||
532 | msg.skb = skb; | |
533 | msg.offset = 0; | |
763ecff1 | 534 | msg.type = ICMPV6_ECHO_REPLY; |
1da177e4 LT |
535 | |
536 | err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr), | |
41a1f8ea | 537 | sizeof(struct icmp6hdr), hlimit, tclass, NULL, &fl, |
1da177e4 LT |
538 | (struct rt6_info*)dst, MSG_DONTWAIT); |
539 | ||
540 | if (err) { | |
541 | ip6_flush_pending_frames(sk); | |
542 | goto out_put; | |
543 | } | |
544 | err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, skb->len + sizeof(struct icmp6hdr)); | |
545 | ||
546 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTECHOREPLIES); | |
547 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTMSGS); | |
548 | ||
549 | out_put: | |
550 | if (likely(idev != NULL)) | |
551 | in6_dev_put(idev); | |
1da177e4 LT |
552 | dst_release(dst); |
553 | out: | |
554 | icmpv6_xmit_unlock(); | |
555 | } | |
556 | ||
04ce6909 | 557 | static void icmpv6_notify(struct sk_buff *skb, int type, int code, __be32 info) |
1da177e4 LT |
558 | { |
559 | struct in6_addr *saddr, *daddr; | |
560 | struct inet6_protocol *ipprot; | |
561 | struct sock *sk; | |
562 | int inner_offset; | |
563 | int hash; | |
564 | u8 nexthdr; | |
565 | ||
566 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
567 | return; | |
568 | ||
569 | nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr; | |
570 | if (ipv6_ext_hdr(nexthdr)) { | |
571 | /* now skip over extension headers */ | |
0d3d077c | 572 | inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr); |
1da177e4 LT |
573 | if (inner_offset<0) |
574 | return; | |
575 | } else { | |
576 | inner_offset = sizeof(struct ipv6hdr); | |
577 | } | |
578 | ||
579 | /* Checkin header including 8 bytes of inner protocol header. */ | |
580 | if (!pskb_may_pull(skb, inner_offset+8)) | |
581 | return; | |
582 | ||
583 | saddr = &skb->nh.ipv6h->saddr; | |
584 | daddr = &skb->nh.ipv6h->daddr; | |
585 | ||
586 | /* BUGGG_FUTURE: we should try to parse exthdrs in this packet. | |
587 | Without this we will not able f.e. to make source routed | |
588 | pmtu discovery. | |
589 | Corresponding argument (opt) to notifiers is already added. | |
590 | --ANK (980726) | |
591 | */ | |
592 | ||
593 | hash = nexthdr & (MAX_INET_PROTOS - 1); | |
594 | ||
595 | rcu_read_lock(); | |
596 | ipprot = rcu_dereference(inet6_protos[hash]); | |
597 | if (ipprot && ipprot->err_handler) | |
598 | ipprot->err_handler(skb, NULL, type, code, inner_offset, info); | |
599 | rcu_read_unlock(); | |
600 | ||
601 | read_lock(&raw_v6_lock); | |
602 | if ((sk = sk_head(&raw_v6_htable[hash])) != NULL) { | |
0bd1b59b | 603 | while((sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, |
2dac4b96 | 604 | IP6CB(skb)->iif))) { |
1da177e4 LT |
605 | rawv6_err(sk, skb, NULL, type, code, inner_offset, info); |
606 | sk = sk_next(sk); | |
607 | } | |
608 | } | |
609 | read_unlock(&raw_v6_lock); | |
610 | } | |
611 | ||
612 | /* | |
613 | * Handle icmp messages | |
614 | */ | |
615 | ||
951dbc8a | 616 | static int icmpv6_rcv(struct sk_buff **pskb) |
1da177e4 LT |
617 | { |
618 | struct sk_buff *skb = *pskb; | |
619 | struct net_device *dev = skb->dev; | |
620 | struct inet6_dev *idev = __in6_dev_get(dev); | |
621 | struct in6_addr *saddr, *daddr; | |
622 | struct ipv6hdr *orig_hdr; | |
623 | struct icmp6hdr *hdr; | |
624 | int type; | |
625 | ||
626 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INMSGS); | |
627 | ||
628 | saddr = &skb->nh.ipv6h->saddr; | |
629 | daddr = &skb->nh.ipv6h->daddr; | |
630 | ||
631 | /* Perform checksum. */ | |
fb286bb2 | 632 | switch (skb->ip_summed) { |
84fa7933 | 633 | case CHECKSUM_COMPLETE: |
fb286bb2 HX |
634 | if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6, |
635 | skb->csum)) | |
636 | break; | |
637 | /* fall through */ | |
638 | case CHECKSUM_NONE: | |
868c86bc AV |
639 | skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len, |
640 | IPPROTO_ICMPV6, 0)); | |
fb286bb2 | 641 | if (__skb_checksum_complete(skb)) { |
46b86a2d | 642 | LIMIT_NETDEBUG(KERN_DEBUG "ICMPv6 checksum failed [" NIP6_FMT " > " NIP6_FMT "]\n", |
64ce2073 | 643 | NIP6(*saddr), NIP6(*daddr)); |
1da177e4 LT |
644 | goto discard_it; |
645 | } | |
646 | } | |
647 | ||
648 | if (!pskb_pull(skb, sizeof(struct icmp6hdr))) | |
649 | goto discard_it; | |
650 | ||
651 | hdr = (struct icmp6hdr *) skb->h.raw; | |
652 | ||
653 | type = hdr->icmp6_type; | |
654 | ||
655 | if (type >= ICMPV6_DEST_UNREACH && type <= ICMPV6_PARAMPROB) | |
656 | ICMP6_INC_STATS_OFFSET_BH(idev, ICMP6_MIB_INDESTUNREACHS, type - ICMPV6_DEST_UNREACH); | |
657 | else if (type >= ICMPV6_ECHO_REQUEST && type <= NDISC_REDIRECT) | |
658 | ICMP6_INC_STATS_OFFSET_BH(idev, ICMP6_MIB_INECHOS, type - ICMPV6_ECHO_REQUEST); | |
659 | ||
660 | switch (type) { | |
661 | case ICMPV6_ECHO_REQUEST: | |
662 | icmpv6_echo_reply(skb); | |
663 | break; | |
664 | ||
665 | case ICMPV6_ECHO_REPLY: | |
666 | /* we couldn't care less */ | |
667 | break; | |
668 | ||
669 | case ICMPV6_PKT_TOOBIG: | |
670 | /* BUGGG_FUTURE: if packet contains rthdr, we cannot update | |
671 | standard destination cache. Seems, only "advanced" | |
672 | destination cache will allow to solve this problem | |
673 | --ANK (980726) | |
674 | */ | |
675 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
676 | goto discard_it; | |
677 | hdr = (struct icmp6hdr *) skb->h.raw; | |
678 | orig_hdr = (struct ipv6hdr *) (hdr + 1); | |
679 | rt6_pmtu_discovery(&orig_hdr->daddr, &orig_hdr->saddr, dev, | |
680 | ntohl(hdr->icmp6_mtu)); | |
681 | ||
682 | /* | |
683 | * Drop through to notify | |
684 | */ | |
685 | ||
686 | case ICMPV6_DEST_UNREACH: | |
687 | case ICMPV6_TIME_EXCEED: | |
688 | case ICMPV6_PARAMPROB: | |
689 | icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); | |
690 | break; | |
691 | ||
692 | case NDISC_ROUTER_SOLICITATION: | |
693 | case NDISC_ROUTER_ADVERTISEMENT: | |
694 | case NDISC_NEIGHBOUR_SOLICITATION: | |
695 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
696 | case NDISC_REDIRECT: | |
697 | ndisc_rcv(skb); | |
698 | break; | |
699 | ||
700 | case ICMPV6_MGM_QUERY: | |
701 | igmp6_event_query(skb); | |
702 | break; | |
703 | ||
704 | case ICMPV6_MGM_REPORT: | |
705 | igmp6_event_report(skb); | |
706 | break; | |
707 | ||
708 | case ICMPV6_MGM_REDUCTION: | |
709 | case ICMPV6_NI_QUERY: | |
710 | case ICMPV6_NI_REPLY: | |
711 | case ICMPV6_MLD2_REPORT: | |
712 | case ICMPV6_DHAAD_REQUEST: | |
713 | case ICMPV6_DHAAD_REPLY: | |
714 | case ICMPV6_MOBILE_PREFIX_SOL: | |
715 | case ICMPV6_MOBILE_PREFIX_ADV: | |
716 | break; | |
717 | ||
718 | default: | |
64ce2073 | 719 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n"); |
1da177e4 LT |
720 | |
721 | /* informational */ | |
722 | if (type & ICMPV6_INFOMSG_MASK) | |
723 | break; | |
724 | ||
725 | /* | |
726 | * error of unknown type. | |
727 | * must pass to upper level | |
728 | */ | |
729 | ||
730 | icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); | |
731 | }; | |
732 | kfree_skb(skb); | |
733 | return 0; | |
734 | ||
735 | discard_it: | |
736 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INERRORS); | |
737 | kfree_skb(skb); | |
738 | return 0; | |
739 | } | |
740 | ||
640c41c7 IM |
741 | /* |
742 | * Special lock-class for __icmpv6_socket: | |
743 | */ | |
744 | static struct lock_class_key icmpv6_socket_sk_dst_lock_key; | |
745 | ||
1da177e4 LT |
746 | int __init icmpv6_init(struct net_proto_family *ops) |
747 | { | |
748 | struct sock *sk; | |
749 | int err, i, j; | |
750 | ||
6f912042 | 751 | for_each_possible_cpu(i) { |
1da177e4 LT |
752 | err = sock_create_kern(PF_INET6, SOCK_RAW, IPPROTO_ICMPV6, |
753 | &per_cpu(__icmpv6_socket, i)); | |
754 | if (err < 0) { | |
755 | printk(KERN_ERR | |
756 | "Failed to initialize the ICMP6 control socket " | |
757 | "(err %d).\n", | |
758 | err); | |
759 | goto fail; | |
760 | } | |
761 | ||
762 | sk = per_cpu(__icmpv6_socket, i)->sk; | |
763 | sk->sk_allocation = GFP_ATOMIC; | |
640c41c7 IM |
764 | /* |
765 | * Split off their lock-class, because sk->sk_dst_lock | |
766 | * gets used from softirqs, which is safe for | |
767 | * __icmpv6_socket (because those never get directly used | |
768 | * via userspace syscalls), but unsafe for normal sockets. | |
769 | */ | |
770 | lockdep_set_class(&sk->sk_dst_lock, | |
771 | &icmpv6_socket_sk_dst_lock_key); | |
1da177e4 LT |
772 | |
773 | /* Enough space for 2 64K ICMP packets, including | |
774 | * sk_buff struct overhead. | |
775 | */ | |
776 | sk->sk_sndbuf = | |
777 | (2 * ((64 * 1024) + sizeof(struct sk_buff))); | |
778 | ||
779 | sk->sk_prot->unhash(sk); | |
780 | } | |
781 | ||
782 | ||
783 | if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0) { | |
784 | printk(KERN_ERR "Failed to register ICMP6 protocol\n"); | |
785 | err = -EAGAIN; | |
786 | goto fail; | |
787 | } | |
788 | ||
789 | return 0; | |
790 | ||
791 | fail: | |
792 | for (j = 0; j < i; j++) { | |
793 | if (!cpu_possible(j)) | |
794 | continue; | |
795 | sock_release(per_cpu(__icmpv6_socket, j)); | |
796 | } | |
797 | ||
798 | return err; | |
799 | } | |
800 | ||
801 | void icmpv6_cleanup(void) | |
802 | { | |
803 | int i; | |
804 | ||
6f912042 | 805 | for_each_possible_cpu(i) { |
1da177e4 LT |
806 | sock_release(per_cpu(__icmpv6_socket, i)); |
807 | } | |
808 | inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); | |
809 | } | |
810 | ||
9b5b5cff | 811 | static const struct icmp6_err { |
1da177e4 LT |
812 | int err; |
813 | int fatal; | |
814 | } tab_unreach[] = { | |
815 | { /* NOROUTE */ | |
816 | .err = ENETUNREACH, | |
817 | .fatal = 0, | |
818 | }, | |
819 | { /* ADM_PROHIBITED */ | |
820 | .err = EACCES, | |
821 | .fatal = 1, | |
822 | }, | |
823 | { /* Was NOT_NEIGHBOUR, now reserved */ | |
824 | .err = EHOSTUNREACH, | |
825 | .fatal = 0, | |
826 | }, | |
827 | { /* ADDR_UNREACH */ | |
828 | .err = EHOSTUNREACH, | |
829 | .fatal = 0, | |
830 | }, | |
831 | { /* PORT_UNREACH */ | |
832 | .err = ECONNREFUSED, | |
833 | .fatal = 1, | |
834 | }, | |
835 | }; | |
836 | ||
837 | int icmpv6_err_convert(int type, int code, int *err) | |
838 | { | |
839 | int fatal = 0; | |
840 | ||
841 | *err = EPROTO; | |
842 | ||
843 | switch (type) { | |
844 | case ICMPV6_DEST_UNREACH: | |
845 | fatal = 1; | |
846 | if (code <= ICMPV6_PORT_UNREACH) { | |
847 | *err = tab_unreach[code].err; | |
848 | fatal = tab_unreach[code].fatal; | |
849 | } | |
850 | break; | |
851 | ||
852 | case ICMPV6_PKT_TOOBIG: | |
853 | *err = EMSGSIZE; | |
854 | break; | |
855 | ||
856 | case ICMPV6_PARAMPROB: | |
857 | *err = EPROTO; | |
858 | fatal = 1; | |
859 | break; | |
860 | ||
861 | case ICMPV6_TIME_EXCEED: | |
862 | *err = EHOSTUNREACH; | |
863 | break; | |
864 | }; | |
865 | ||
866 | return fatal; | |
867 | } | |
868 | ||
869 | #ifdef CONFIG_SYSCTL | |
870 | ctl_table ipv6_icmp_table[] = { | |
871 | { | |
872 | .ctl_name = NET_IPV6_ICMP_RATELIMIT, | |
873 | .procname = "ratelimit", | |
874 | .data = &sysctl_icmpv6_time, | |
875 | .maxlen = sizeof(int), | |
876 | .mode = 0644, | |
877 | .proc_handler = &proc_dointvec | |
878 | }, | |
879 | { .ctl_name = 0 }, | |
880 | }; | |
881 | #endif | |
882 |