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> | |
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> | |
66 | ||
67 | #include <asm/uaccess.h> | |
68 | #include <asm/system.h> | |
69 | ||
ba89966c | 70 | DEFINE_SNMP_STAT(struct icmpv6_mib, icmpv6_statistics) __read_mostly; |
1da177e4 LT |
71 | |
72 | /* | |
73 | * The ICMP socket(s). This is the most convenient way to flow control | |
74 | * our ICMP output as well as maintain a clean interface throughout | |
75 | * all layers. All Socketless IP sends will soon be gone. | |
76 | * | |
77 | * On SMP we have one ICMP socket per-cpu. | |
78 | */ | |
79 | static DEFINE_PER_CPU(struct socket *, __icmpv6_socket) = NULL; | |
80 | #define icmpv6_socket __get_cpu_var(__icmpv6_socket) | |
81 | ||
82 | static int icmpv6_rcv(struct sk_buff **pskb, unsigned int *nhoffp); | |
83 | ||
84 | static struct inet6_protocol icmpv6_protocol = { | |
85 | .handler = icmpv6_rcv, | |
86 | .flags = INET6_PROTO_FINAL, | |
87 | }; | |
88 | ||
89 | static __inline__ int icmpv6_xmit_lock(void) | |
90 | { | |
91 | local_bh_disable(); | |
92 | ||
93 | if (unlikely(!spin_trylock(&icmpv6_socket->sk->sk_lock.slock))) { | |
94 | /* This can happen if the output path (f.e. SIT or | |
95 | * ip6ip6 tunnel) signals dst_link_failure() for an | |
96 | * outgoing ICMP6 packet. | |
97 | */ | |
98 | local_bh_enable(); | |
99 | return 1; | |
100 | } | |
101 | return 0; | |
102 | } | |
103 | ||
104 | static __inline__ void icmpv6_xmit_unlock(void) | |
105 | { | |
106 | spin_unlock_bh(&icmpv6_socket->sk->sk_lock.slock); | |
107 | } | |
108 | ||
109 | /* | |
110 | * Slightly more convenient version of icmpv6_send. | |
111 | */ | |
112 | void icmpv6_param_prob(struct sk_buff *skb, int code, int pos) | |
113 | { | |
114 | icmpv6_send(skb, ICMPV6_PARAMPROB, code, pos, skb->dev); | |
115 | kfree_skb(skb); | |
116 | } | |
117 | ||
118 | /* | |
119 | * Figure out, may we reply to this packet with icmp error. | |
120 | * | |
121 | * We do not reply, if: | |
122 | * - it was icmp error message. | |
123 | * - it is truncated, so that it is known, that protocol is ICMPV6 | |
124 | * (i.e. in the middle of some exthdr) | |
125 | * | |
126 | * --ANK (980726) | |
127 | */ | |
128 | ||
129 | static int is_ineligible(struct sk_buff *skb) | |
130 | { | |
131 | int ptr = (u8*)(skb->nh.ipv6h+1) - skb->data; | |
132 | int len = skb->len - ptr; | |
133 | __u8 nexthdr = skb->nh.ipv6h->nexthdr; | |
134 | ||
135 | if (len < 0) | |
136 | return 1; | |
137 | ||
0d3d077c | 138 | ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr); |
1da177e4 LT |
139 | if (ptr < 0) |
140 | return 0; | |
141 | if (nexthdr == IPPROTO_ICMPV6) { | |
142 | u8 _type, *tp; | |
143 | tp = skb_header_pointer(skb, | |
144 | ptr+offsetof(struct icmp6hdr, icmp6_type), | |
145 | sizeof(_type), &_type); | |
146 | if (tp == NULL || | |
147 | !(*tp & ICMPV6_INFOMSG_MASK)) | |
148 | return 1; | |
149 | } | |
150 | return 0; | |
151 | } | |
152 | ||
153 | static int sysctl_icmpv6_time = 1*HZ; | |
154 | ||
155 | /* | |
156 | * Check the ICMP output rate limit | |
157 | */ | |
158 | static inline int icmpv6_xrlim_allow(struct sock *sk, int type, | |
159 | struct flowi *fl) | |
160 | { | |
161 | struct dst_entry *dst; | |
162 | int res = 0; | |
163 | ||
164 | /* Informational messages are not limited. */ | |
165 | if (type & ICMPV6_INFOMSG_MASK) | |
166 | return 1; | |
167 | ||
168 | /* Do not limit pmtu discovery, it would break it. */ | |
169 | if (type == ICMPV6_PKT_TOOBIG) | |
170 | return 1; | |
171 | ||
172 | /* | |
173 | * Look up the output route. | |
174 | * XXX: perhaps the expire for routing entries cloned by | |
175 | * this lookup should be more aggressive (not longer than timeout). | |
176 | */ | |
177 | dst = ip6_route_output(sk, fl); | |
178 | if (dst->error) { | |
179 | IP6_INC_STATS(IPSTATS_MIB_OUTNOROUTES); | |
180 | } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) { | |
181 | res = 1; | |
182 | } else { | |
183 | struct rt6_info *rt = (struct rt6_info *)dst; | |
184 | int tmo = sysctl_icmpv6_time; | |
185 | ||
186 | /* Give more bandwidth to wider prefixes. */ | |
187 | if (rt->rt6i_dst.plen < 128) | |
188 | tmo >>= ((128 - rt->rt6i_dst.plen)>>5); | |
189 | ||
190 | res = xrlim_allow(dst, tmo); | |
191 | } | |
192 | dst_release(dst); | |
193 | return res; | |
194 | } | |
195 | ||
196 | /* | |
197 | * an inline helper for the "simple" if statement below | |
198 | * checks if parameter problem report is caused by an | |
199 | * unrecognized IPv6 option that has the Option Type | |
200 | * highest-order two bits set to 10 | |
201 | */ | |
202 | ||
203 | static __inline__ int opt_unrec(struct sk_buff *skb, __u32 offset) | |
204 | { | |
205 | u8 _optval, *op; | |
206 | ||
207 | offset += skb->nh.raw - skb->data; | |
208 | op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval); | |
209 | if (op == NULL) | |
210 | return 1; | |
211 | return (*op & 0xC0) == 0x80; | |
212 | } | |
213 | ||
214 | static int icmpv6_push_pending_frames(struct sock *sk, struct flowi *fl, struct icmp6hdr *thdr, int len) | |
215 | { | |
216 | struct sk_buff *skb; | |
217 | struct icmp6hdr *icmp6h; | |
218 | int err = 0; | |
219 | ||
220 | if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) | |
221 | goto out; | |
222 | ||
223 | icmp6h = (struct icmp6hdr*) skb->h.raw; | |
224 | memcpy(icmp6h, thdr, sizeof(struct icmp6hdr)); | |
225 | icmp6h->icmp6_cksum = 0; | |
226 | ||
227 | if (skb_queue_len(&sk->sk_write_queue) == 1) { | |
228 | skb->csum = csum_partial((char *)icmp6h, | |
229 | sizeof(struct icmp6hdr), skb->csum); | |
230 | icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src, | |
231 | &fl->fl6_dst, | |
232 | len, fl->proto, | |
233 | skb->csum); | |
234 | } else { | |
235 | u32 tmp_csum = 0; | |
236 | ||
237 | skb_queue_walk(&sk->sk_write_queue, skb) { | |
238 | tmp_csum = csum_add(tmp_csum, skb->csum); | |
239 | } | |
240 | ||
241 | tmp_csum = csum_partial((char *)icmp6h, | |
242 | sizeof(struct icmp6hdr), tmp_csum); | |
243 | tmp_csum = csum_ipv6_magic(&fl->fl6_src, | |
244 | &fl->fl6_dst, | |
245 | len, fl->proto, tmp_csum); | |
246 | icmp6h->icmp6_cksum = tmp_csum; | |
247 | } | |
248 | if (icmp6h->icmp6_cksum == 0) | |
249 | icmp6h->icmp6_cksum = -1; | |
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; | |
258 | }; | |
259 | ||
260 | static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb) | |
261 | { | |
262 | struct icmpv6_msg *msg = (struct icmpv6_msg *) from; | |
263 | struct sk_buff *org_skb = msg->skb; | |
264 | __u32 csum = 0; | |
265 | ||
266 | csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset, | |
267 | to, len, csum); | |
268 | skb->csum = csum_block_add(skb->csum, csum, odd); | |
269 | return 0; | |
270 | } | |
271 | ||
272 | /* | |
273 | * Send an ICMP message in response to a packet in error | |
274 | */ | |
275 | void icmpv6_send(struct sk_buff *skb, int type, int code, __u32 info, | |
276 | struct net_device *dev) | |
277 | { | |
278 | struct inet6_dev *idev = NULL; | |
279 | struct ipv6hdr *hdr = skb->nh.ipv6h; | |
84427d53 YH |
280 | struct sock *sk; |
281 | struct ipv6_pinfo *np; | |
1da177e4 LT |
282 | struct in6_addr *saddr = NULL; |
283 | struct dst_entry *dst; | |
284 | struct icmp6hdr tmp_hdr; | |
285 | struct flowi fl; | |
286 | struct icmpv6_msg msg; | |
287 | int iif = 0; | |
288 | int addr_type = 0; | |
289 | int len; | |
41a1f8ea | 290 | int hlimit, tclass; |
1da177e4 LT |
291 | int err = 0; |
292 | ||
293 | if ((u8*)hdr < skb->head || (u8*)(hdr+1) > skb->tail) | |
294 | return; | |
295 | ||
296 | /* | |
297 | * Make sure we respect the rules | |
298 | * i.e. RFC 1885 2.4(e) | |
299 | * Rule (e.1) is enforced by not using icmpv6_send | |
300 | * in any code that processes icmp errors. | |
301 | */ | |
302 | addr_type = ipv6_addr_type(&hdr->daddr); | |
303 | ||
304 | if (ipv6_chk_addr(&hdr->daddr, skb->dev, 0)) | |
305 | saddr = &hdr->daddr; | |
306 | ||
307 | /* | |
308 | * Dest addr check | |
309 | */ | |
310 | ||
311 | if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) { | |
312 | if (type != ICMPV6_PKT_TOOBIG && | |
313 | !(type == ICMPV6_PARAMPROB && | |
314 | code == ICMPV6_UNK_OPTION && | |
315 | (opt_unrec(skb, info)))) | |
316 | return; | |
317 | ||
318 | saddr = NULL; | |
319 | } | |
320 | ||
321 | addr_type = ipv6_addr_type(&hdr->saddr); | |
322 | ||
323 | /* | |
324 | * Source addr check | |
325 | */ | |
326 | ||
327 | if (addr_type & IPV6_ADDR_LINKLOCAL) | |
328 | iif = skb->dev->ifindex; | |
329 | ||
330 | /* | |
331 | * Must not send if we know that source is Anycast also. | |
332 | * for now we don't know that. | |
333 | */ | |
334 | if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) { | |
64ce2073 | 335 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: addr_any/mcast source\n"); |
1da177e4 LT |
336 | return; |
337 | } | |
338 | ||
339 | /* | |
340 | * Never answer to a ICMP packet. | |
341 | */ | |
342 | if (is_ineligible(skb)) { | |
64ce2073 | 343 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: no reply to icmp error\n"); |
1da177e4 LT |
344 | return; |
345 | } | |
346 | ||
347 | memset(&fl, 0, sizeof(fl)); | |
348 | fl.proto = IPPROTO_ICMPV6; | |
349 | ipv6_addr_copy(&fl.fl6_dst, &hdr->saddr); | |
350 | if (saddr) | |
351 | ipv6_addr_copy(&fl.fl6_src, saddr); | |
352 | fl.oif = iif; | |
353 | fl.fl_icmp_type = type; | |
354 | fl.fl_icmp_code = code; | |
355 | ||
356 | if (icmpv6_xmit_lock()) | |
357 | return; | |
358 | ||
84427d53 YH |
359 | sk = icmpv6_socket->sk; |
360 | np = inet6_sk(sk); | |
361 | ||
1da177e4 LT |
362 | if (!icmpv6_xrlim_allow(sk, type, &fl)) |
363 | goto out; | |
364 | ||
365 | tmp_hdr.icmp6_type = type; | |
366 | tmp_hdr.icmp6_code = code; | |
367 | tmp_hdr.icmp6_cksum = 0; | |
368 | tmp_hdr.icmp6_pointer = htonl(info); | |
369 | ||
370 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) | |
371 | fl.oif = np->mcast_oif; | |
372 | ||
373 | err = ip6_dst_lookup(sk, &dst, &fl); | |
374 | if (err) | |
375 | goto out; | |
376 | if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) | |
e104411b | 377 | goto out; |
1da177e4 LT |
378 | |
379 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) | |
380 | hlimit = np->mcast_hops; | |
381 | else | |
382 | hlimit = np->hop_limit; | |
383 | if (hlimit < 0) | |
384 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | |
385 | if (hlimit < 0) | |
386 | hlimit = ipv6_get_hoplimit(dst->dev); | |
387 | ||
41a1f8ea YH |
388 | tclass = np->cork.tclass; |
389 | if (tclass < 0) | |
390 | tclass = 0; | |
391 | ||
1da177e4 LT |
392 | msg.skb = skb; |
393 | msg.offset = skb->nh.raw - skb->data; | |
394 | ||
395 | len = skb->len - msg.offset; | |
396 | len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr)); | |
397 | if (len < 0) { | |
64ce2073 | 398 | LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n"); |
1da177e4 LT |
399 | goto out_dst_release; |
400 | } | |
401 | ||
402 | idev = in6_dev_get(skb->dev); | |
403 | ||
404 | err = ip6_append_data(sk, icmpv6_getfrag, &msg, | |
405 | len + sizeof(struct icmp6hdr), | |
406 | sizeof(struct icmp6hdr), | |
41a1f8ea | 407 | hlimit, tclass, NULL, &fl, (struct rt6_info*)dst, |
1da177e4 LT |
408 | MSG_DONTWAIT); |
409 | if (err) { | |
410 | ip6_flush_pending_frames(sk); | |
411 | goto out_put; | |
412 | } | |
413 | err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, len + sizeof(struct icmp6hdr)); | |
414 | ||
415 | if (type >= ICMPV6_DEST_UNREACH && type <= ICMPV6_PARAMPROB) | |
416 | ICMP6_INC_STATS_OFFSET_BH(idev, ICMP6_MIB_OUTDESTUNREACHS, type - ICMPV6_DEST_UNREACH); | |
417 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTMSGS); | |
418 | ||
419 | out_put: | |
420 | if (likely(idev != NULL)) | |
421 | in6_dev_put(idev); | |
422 | out_dst_release: | |
423 | dst_release(dst); | |
424 | out: | |
425 | icmpv6_xmit_unlock(); | |
426 | } | |
427 | ||
428 | static void icmpv6_echo_reply(struct sk_buff *skb) | |
429 | { | |
84427d53 | 430 | struct sock *sk; |
1da177e4 | 431 | struct inet6_dev *idev; |
84427d53 | 432 | struct ipv6_pinfo *np; |
1da177e4 LT |
433 | struct in6_addr *saddr = NULL; |
434 | struct icmp6hdr *icmph = (struct icmp6hdr *) skb->h.raw; | |
435 | struct icmp6hdr tmp_hdr; | |
436 | struct flowi fl; | |
437 | struct icmpv6_msg msg; | |
438 | struct dst_entry *dst; | |
439 | int err = 0; | |
440 | int hlimit; | |
41a1f8ea | 441 | int tclass; |
1da177e4 LT |
442 | |
443 | saddr = &skb->nh.ipv6h->daddr; | |
444 | ||
445 | if (!ipv6_unicast_destination(skb)) | |
446 | saddr = NULL; | |
447 | ||
448 | memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr)); | |
449 | tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY; | |
450 | ||
451 | memset(&fl, 0, sizeof(fl)); | |
452 | fl.proto = IPPROTO_ICMPV6; | |
453 | ipv6_addr_copy(&fl.fl6_dst, &skb->nh.ipv6h->saddr); | |
454 | if (saddr) | |
455 | ipv6_addr_copy(&fl.fl6_src, saddr); | |
456 | fl.oif = skb->dev->ifindex; | |
457 | fl.fl_icmp_type = ICMPV6_ECHO_REPLY; | |
458 | ||
459 | if (icmpv6_xmit_lock()) | |
460 | return; | |
461 | ||
84427d53 YH |
462 | sk = icmpv6_socket->sk; |
463 | np = inet6_sk(sk); | |
464 | ||
1da177e4 LT |
465 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) |
466 | fl.oif = np->mcast_oif; | |
467 | ||
468 | err = ip6_dst_lookup(sk, &dst, &fl); | |
469 | if (err) | |
470 | goto out; | |
471 | if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) | |
e104411b | 472 | goto out; |
1da177e4 LT |
473 | |
474 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) | |
475 | hlimit = np->mcast_hops; | |
476 | else | |
477 | hlimit = np->hop_limit; | |
478 | if (hlimit < 0) | |
479 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | |
480 | if (hlimit < 0) | |
481 | hlimit = ipv6_get_hoplimit(dst->dev); | |
482 | ||
41a1f8ea YH |
483 | tclass = np->cork.tclass; |
484 | if (tclass < 0) | |
485 | tclass = 0; | |
486 | ||
1da177e4 LT |
487 | idev = in6_dev_get(skb->dev); |
488 | ||
489 | msg.skb = skb; | |
490 | msg.offset = 0; | |
491 | ||
492 | err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr), | |
41a1f8ea | 493 | sizeof(struct icmp6hdr), hlimit, tclass, NULL, &fl, |
1da177e4 LT |
494 | (struct rt6_info*)dst, MSG_DONTWAIT); |
495 | ||
496 | if (err) { | |
497 | ip6_flush_pending_frames(sk); | |
498 | goto out_put; | |
499 | } | |
500 | err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, skb->len + sizeof(struct icmp6hdr)); | |
501 | ||
502 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTECHOREPLIES); | |
503 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTMSGS); | |
504 | ||
505 | out_put: | |
506 | if (likely(idev != NULL)) | |
507 | in6_dev_put(idev); | |
1da177e4 LT |
508 | dst_release(dst); |
509 | out: | |
510 | icmpv6_xmit_unlock(); | |
511 | } | |
512 | ||
513 | static void icmpv6_notify(struct sk_buff *skb, int type, int code, u32 info) | |
514 | { | |
515 | struct in6_addr *saddr, *daddr; | |
516 | struct inet6_protocol *ipprot; | |
517 | struct sock *sk; | |
518 | int inner_offset; | |
519 | int hash; | |
520 | u8 nexthdr; | |
521 | ||
522 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
523 | return; | |
524 | ||
525 | nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr; | |
526 | if (ipv6_ext_hdr(nexthdr)) { | |
527 | /* now skip over extension headers */ | |
0d3d077c | 528 | inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr); |
1da177e4 LT |
529 | if (inner_offset<0) |
530 | return; | |
531 | } else { | |
532 | inner_offset = sizeof(struct ipv6hdr); | |
533 | } | |
534 | ||
535 | /* Checkin header including 8 bytes of inner protocol header. */ | |
536 | if (!pskb_may_pull(skb, inner_offset+8)) | |
537 | return; | |
538 | ||
539 | saddr = &skb->nh.ipv6h->saddr; | |
540 | daddr = &skb->nh.ipv6h->daddr; | |
541 | ||
542 | /* BUGGG_FUTURE: we should try to parse exthdrs in this packet. | |
543 | Without this we will not able f.e. to make source routed | |
544 | pmtu discovery. | |
545 | Corresponding argument (opt) to notifiers is already added. | |
546 | --ANK (980726) | |
547 | */ | |
548 | ||
549 | hash = nexthdr & (MAX_INET_PROTOS - 1); | |
550 | ||
551 | rcu_read_lock(); | |
552 | ipprot = rcu_dereference(inet6_protos[hash]); | |
553 | if (ipprot && ipprot->err_handler) | |
554 | ipprot->err_handler(skb, NULL, type, code, inner_offset, info); | |
555 | rcu_read_unlock(); | |
556 | ||
557 | read_lock(&raw_v6_lock); | |
558 | if ((sk = sk_head(&raw_v6_htable[hash])) != NULL) { | |
0bd1b59b | 559 | while((sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, |
2dac4b96 | 560 | IP6CB(skb)->iif))) { |
1da177e4 LT |
561 | rawv6_err(sk, skb, NULL, type, code, inner_offset, info); |
562 | sk = sk_next(sk); | |
563 | } | |
564 | } | |
565 | read_unlock(&raw_v6_lock); | |
566 | } | |
567 | ||
568 | /* | |
569 | * Handle icmp messages | |
570 | */ | |
571 | ||
572 | static int icmpv6_rcv(struct sk_buff **pskb, unsigned int *nhoffp) | |
573 | { | |
574 | struct sk_buff *skb = *pskb; | |
575 | struct net_device *dev = skb->dev; | |
576 | struct inet6_dev *idev = __in6_dev_get(dev); | |
577 | struct in6_addr *saddr, *daddr; | |
578 | struct ipv6hdr *orig_hdr; | |
579 | struct icmp6hdr *hdr; | |
580 | int type; | |
581 | ||
582 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INMSGS); | |
583 | ||
584 | saddr = &skb->nh.ipv6h->saddr; | |
585 | daddr = &skb->nh.ipv6h->daddr; | |
586 | ||
587 | /* Perform checksum. */ | |
fb286bb2 HX |
588 | switch (skb->ip_summed) { |
589 | case CHECKSUM_HW: | |
590 | if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6, | |
591 | skb->csum)) | |
592 | break; | |
593 | /* fall through */ | |
594 | case CHECKSUM_NONE: | |
595 | skb->csum = ~csum_ipv6_magic(saddr, daddr, skb->len, | |
596 | IPPROTO_ICMPV6, 0); | |
597 | if (__skb_checksum_complete(skb)) { | |
64ce2073 PM |
598 | LIMIT_NETDEBUG(KERN_DEBUG "ICMPv6 checksum failed [%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x > %04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x]\n", |
599 | NIP6(*saddr), NIP6(*daddr)); | |
1da177e4 LT |
600 | goto discard_it; |
601 | } | |
602 | } | |
603 | ||
604 | if (!pskb_pull(skb, sizeof(struct icmp6hdr))) | |
605 | goto discard_it; | |
606 | ||
607 | hdr = (struct icmp6hdr *) skb->h.raw; | |
608 | ||
609 | type = hdr->icmp6_type; | |
610 | ||
611 | if (type >= ICMPV6_DEST_UNREACH && type <= ICMPV6_PARAMPROB) | |
612 | ICMP6_INC_STATS_OFFSET_BH(idev, ICMP6_MIB_INDESTUNREACHS, type - ICMPV6_DEST_UNREACH); | |
613 | else if (type >= ICMPV6_ECHO_REQUEST && type <= NDISC_REDIRECT) | |
614 | ICMP6_INC_STATS_OFFSET_BH(idev, ICMP6_MIB_INECHOS, type - ICMPV6_ECHO_REQUEST); | |
615 | ||
616 | switch (type) { | |
617 | case ICMPV6_ECHO_REQUEST: | |
618 | icmpv6_echo_reply(skb); | |
619 | break; | |
620 | ||
621 | case ICMPV6_ECHO_REPLY: | |
622 | /* we couldn't care less */ | |
623 | break; | |
624 | ||
625 | case ICMPV6_PKT_TOOBIG: | |
626 | /* BUGGG_FUTURE: if packet contains rthdr, we cannot update | |
627 | standard destination cache. Seems, only "advanced" | |
628 | destination cache will allow to solve this problem | |
629 | --ANK (980726) | |
630 | */ | |
631 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
632 | goto discard_it; | |
633 | hdr = (struct icmp6hdr *) skb->h.raw; | |
634 | orig_hdr = (struct ipv6hdr *) (hdr + 1); | |
635 | rt6_pmtu_discovery(&orig_hdr->daddr, &orig_hdr->saddr, dev, | |
636 | ntohl(hdr->icmp6_mtu)); | |
637 | ||
638 | /* | |
639 | * Drop through to notify | |
640 | */ | |
641 | ||
642 | case ICMPV6_DEST_UNREACH: | |
643 | case ICMPV6_TIME_EXCEED: | |
644 | case ICMPV6_PARAMPROB: | |
645 | icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); | |
646 | break; | |
647 | ||
648 | case NDISC_ROUTER_SOLICITATION: | |
649 | case NDISC_ROUTER_ADVERTISEMENT: | |
650 | case NDISC_NEIGHBOUR_SOLICITATION: | |
651 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
652 | case NDISC_REDIRECT: | |
653 | ndisc_rcv(skb); | |
654 | break; | |
655 | ||
656 | case ICMPV6_MGM_QUERY: | |
657 | igmp6_event_query(skb); | |
658 | break; | |
659 | ||
660 | case ICMPV6_MGM_REPORT: | |
661 | igmp6_event_report(skb); | |
662 | break; | |
663 | ||
664 | case ICMPV6_MGM_REDUCTION: | |
665 | case ICMPV6_NI_QUERY: | |
666 | case ICMPV6_NI_REPLY: | |
667 | case ICMPV6_MLD2_REPORT: | |
668 | case ICMPV6_DHAAD_REQUEST: | |
669 | case ICMPV6_DHAAD_REPLY: | |
670 | case ICMPV6_MOBILE_PREFIX_SOL: | |
671 | case ICMPV6_MOBILE_PREFIX_ADV: | |
672 | break; | |
673 | ||
674 | default: | |
64ce2073 | 675 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n"); |
1da177e4 LT |
676 | |
677 | /* informational */ | |
678 | if (type & ICMPV6_INFOMSG_MASK) | |
679 | break; | |
680 | ||
681 | /* | |
682 | * error of unknown type. | |
683 | * must pass to upper level | |
684 | */ | |
685 | ||
686 | icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); | |
687 | }; | |
688 | kfree_skb(skb); | |
689 | return 0; | |
690 | ||
691 | discard_it: | |
692 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INERRORS); | |
693 | kfree_skb(skb); | |
694 | return 0; | |
695 | } | |
696 | ||
697 | int __init icmpv6_init(struct net_proto_family *ops) | |
698 | { | |
699 | struct sock *sk; | |
700 | int err, i, j; | |
701 | ||
670c02c2 | 702 | for_each_cpu(i) { |
1da177e4 LT |
703 | err = sock_create_kern(PF_INET6, SOCK_RAW, IPPROTO_ICMPV6, |
704 | &per_cpu(__icmpv6_socket, i)); | |
705 | if (err < 0) { | |
706 | printk(KERN_ERR | |
707 | "Failed to initialize the ICMP6 control socket " | |
708 | "(err %d).\n", | |
709 | err); | |
710 | goto fail; | |
711 | } | |
712 | ||
713 | sk = per_cpu(__icmpv6_socket, i)->sk; | |
714 | sk->sk_allocation = GFP_ATOMIC; | |
715 | ||
716 | /* Enough space for 2 64K ICMP packets, including | |
717 | * sk_buff struct overhead. | |
718 | */ | |
719 | sk->sk_sndbuf = | |
720 | (2 * ((64 * 1024) + sizeof(struct sk_buff))); | |
721 | ||
722 | sk->sk_prot->unhash(sk); | |
723 | } | |
724 | ||
725 | ||
726 | if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0) { | |
727 | printk(KERN_ERR "Failed to register ICMP6 protocol\n"); | |
728 | err = -EAGAIN; | |
729 | goto fail; | |
730 | } | |
731 | ||
732 | return 0; | |
733 | ||
734 | fail: | |
735 | for (j = 0; j < i; j++) { | |
736 | if (!cpu_possible(j)) | |
737 | continue; | |
738 | sock_release(per_cpu(__icmpv6_socket, j)); | |
739 | } | |
740 | ||
741 | return err; | |
742 | } | |
743 | ||
744 | void icmpv6_cleanup(void) | |
745 | { | |
746 | int i; | |
747 | ||
670c02c2 | 748 | for_each_cpu(i) { |
1da177e4 LT |
749 | sock_release(per_cpu(__icmpv6_socket, i)); |
750 | } | |
751 | inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); | |
752 | } | |
753 | ||
754 | static struct icmp6_err { | |
755 | int err; | |
756 | int fatal; | |
757 | } tab_unreach[] = { | |
758 | { /* NOROUTE */ | |
759 | .err = ENETUNREACH, | |
760 | .fatal = 0, | |
761 | }, | |
762 | { /* ADM_PROHIBITED */ | |
763 | .err = EACCES, | |
764 | .fatal = 1, | |
765 | }, | |
766 | { /* Was NOT_NEIGHBOUR, now reserved */ | |
767 | .err = EHOSTUNREACH, | |
768 | .fatal = 0, | |
769 | }, | |
770 | { /* ADDR_UNREACH */ | |
771 | .err = EHOSTUNREACH, | |
772 | .fatal = 0, | |
773 | }, | |
774 | { /* PORT_UNREACH */ | |
775 | .err = ECONNREFUSED, | |
776 | .fatal = 1, | |
777 | }, | |
778 | }; | |
779 | ||
780 | int icmpv6_err_convert(int type, int code, int *err) | |
781 | { | |
782 | int fatal = 0; | |
783 | ||
784 | *err = EPROTO; | |
785 | ||
786 | switch (type) { | |
787 | case ICMPV6_DEST_UNREACH: | |
788 | fatal = 1; | |
789 | if (code <= ICMPV6_PORT_UNREACH) { | |
790 | *err = tab_unreach[code].err; | |
791 | fatal = tab_unreach[code].fatal; | |
792 | } | |
793 | break; | |
794 | ||
795 | case ICMPV6_PKT_TOOBIG: | |
796 | *err = EMSGSIZE; | |
797 | break; | |
798 | ||
799 | case ICMPV6_PARAMPROB: | |
800 | *err = EPROTO; | |
801 | fatal = 1; | |
802 | break; | |
803 | ||
804 | case ICMPV6_TIME_EXCEED: | |
805 | *err = EHOSTUNREACH; | |
806 | break; | |
807 | }; | |
808 | ||
809 | return fatal; | |
810 | } | |
811 | ||
812 | #ifdef CONFIG_SYSCTL | |
813 | ctl_table ipv6_icmp_table[] = { | |
814 | { | |
815 | .ctl_name = NET_IPV6_ICMP_RATELIMIT, | |
816 | .procname = "ratelimit", | |
817 | .data = &sysctl_icmpv6_time, | |
818 | .maxlen = sizeof(int), | |
819 | .mode = 0644, | |
820 | .proc_handler = &proc_dointvec | |
821 | }, | |
822 | { .ctl_name = 0 }, | |
823 | }; | |
824 | #endif | |
825 |