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> | |
66 | ||
67 | #include <asm/uaccess.h> | |
68 | #include <asm/system.h> | |
69 | ||
ba89966c | 70 | DEFINE_SNMP_STAT(struct icmpv6_mib, icmpv6_statistics) __read_mostly; |
7159039a | 71 | EXPORT_SYMBOL(icmpv6_statistics); |
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 | ||
1ab1457c | 110 | /* |
1da177e4 LT |
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 | { | |
0660e03f | 132 | int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data; |
1da177e4 | 133 | int len = skb->len - ptr; |
0660e03f | 134 | __u8 nexthdr = ipv6_hdr(skb)->nexthdr; |
1da177e4 LT |
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 | 155 | |
1ab1457c YH |
156 | /* |
157 | * Check the ICMP output rate limit | |
1da177e4 LT |
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 | ||
1ab1457c | 173 | /* |
1da177e4 LT |
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 | |
1ab1457c | 201 | * unrecognized IPv6 option that has the Option Type |
1da177e4 LT |
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 | ||
bbe735e4 | 209 | offset += skb_network_offset(skb); |
1da177e4 LT |
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 | { | |
0660e03f | 278 | struct ipv6hdr *iph = ipv6_hdr(skb); |
79383236 MN |
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)) { | |
d56f90a7 ACM |
287 | hao = (struct ipv6_destopt_hao *) |
288 | (skb_network_header(skb) + off); | |
79383236 MN |
289 | ipv6_addr_copy(&tmp, &iph->saddr); |
290 | ipv6_addr_copy(&iph->saddr, &hao->addr); | |
291 | ipv6_addr_copy(&hao->addr, &tmp); | |
292 | } | |
293 | } | |
294 | } | |
295 | #else | |
296 | static inline void mip6_addr_swap(struct sk_buff *skb) {} | |
297 | #endif | |
298 | ||
1da177e4 LT |
299 | /* |
300 | * Send an ICMP message in response to a packet in error | |
301 | */ | |
1ab1457c | 302 | void icmpv6_send(struct sk_buff *skb, int type, int code, __u32 info, |
1da177e4 LT |
303 | struct net_device *dev) |
304 | { | |
305 | struct inet6_dev *idev = NULL; | |
0660e03f | 306 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
84427d53 YH |
307 | struct sock *sk; |
308 | struct ipv6_pinfo *np; | |
1da177e4 LT |
309 | struct in6_addr *saddr = NULL; |
310 | struct dst_entry *dst; | |
311 | struct icmp6hdr tmp_hdr; | |
312 | struct flowi fl; | |
313 | struct icmpv6_msg msg; | |
314 | int iif = 0; | |
315 | int addr_type = 0; | |
316 | int len; | |
41a1f8ea | 317 | int hlimit, tclass; |
1da177e4 LT |
318 | int err = 0; |
319 | ||
320 | if ((u8*)hdr < skb->head || (u8*)(hdr+1) > skb->tail) | |
321 | return; | |
322 | ||
323 | /* | |
1ab1457c | 324 | * Make sure we respect the rules |
1da177e4 LT |
325 | * i.e. RFC 1885 2.4(e) |
326 | * Rule (e.1) is enforced by not using icmpv6_send | |
327 | * in any code that processes icmp errors. | |
328 | */ | |
329 | addr_type = ipv6_addr_type(&hdr->daddr); | |
330 | ||
331 | if (ipv6_chk_addr(&hdr->daddr, skb->dev, 0)) | |
332 | saddr = &hdr->daddr; | |
333 | ||
334 | /* | |
335 | * Dest addr check | |
336 | */ | |
337 | ||
338 | if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) { | |
339 | if (type != ICMPV6_PKT_TOOBIG && | |
1ab1457c YH |
340 | !(type == ICMPV6_PARAMPROB && |
341 | code == ICMPV6_UNK_OPTION && | |
1da177e4 LT |
342 | (opt_unrec(skb, info)))) |
343 | return; | |
344 | ||
345 | saddr = NULL; | |
346 | } | |
347 | ||
348 | addr_type = ipv6_addr_type(&hdr->saddr); | |
349 | ||
350 | /* | |
351 | * Source addr check | |
352 | */ | |
353 | ||
354 | if (addr_type & IPV6_ADDR_LINKLOCAL) | |
355 | iif = skb->dev->ifindex; | |
356 | ||
357 | /* | |
8de3351e YH |
358 | * Must not send error if the source does not uniquely |
359 | * identify a single node (RFC2463 Section 2.4). | |
360 | * We check unspecified / multicast addresses here, | |
361 | * and anycast addresses will be checked later. | |
1da177e4 LT |
362 | */ |
363 | if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) { | |
64ce2073 | 364 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: addr_any/mcast source\n"); |
1da177e4 LT |
365 | return; |
366 | } | |
367 | ||
1ab1457c | 368 | /* |
1da177e4 LT |
369 | * Never answer to a ICMP packet. |
370 | */ | |
371 | if (is_ineligible(skb)) { | |
64ce2073 | 372 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: no reply to icmp error\n"); |
1da177e4 LT |
373 | return; |
374 | } | |
375 | ||
79383236 MN |
376 | mip6_addr_swap(skb); |
377 | ||
1da177e4 LT |
378 | memset(&fl, 0, sizeof(fl)); |
379 | fl.proto = IPPROTO_ICMPV6; | |
380 | ipv6_addr_copy(&fl.fl6_dst, &hdr->saddr); | |
381 | if (saddr) | |
382 | ipv6_addr_copy(&fl.fl6_src, saddr); | |
383 | fl.oif = iif; | |
384 | fl.fl_icmp_type = type; | |
385 | fl.fl_icmp_code = code; | |
beb8d13b | 386 | security_skb_classify_flow(skb, &fl); |
1da177e4 LT |
387 | |
388 | if (icmpv6_xmit_lock()) | |
389 | return; | |
390 | ||
84427d53 YH |
391 | sk = icmpv6_socket->sk; |
392 | np = inet6_sk(sk); | |
393 | ||
1da177e4 LT |
394 | if (!icmpv6_xrlim_allow(sk, type, &fl)) |
395 | goto out; | |
396 | ||
397 | tmp_hdr.icmp6_type = type; | |
398 | tmp_hdr.icmp6_code = code; | |
399 | tmp_hdr.icmp6_cksum = 0; | |
400 | tmp_hdr.icmp6_pointer = htonl(info); | |
401 | ||
402 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) | |
403 | fl.oif = np->mcast_oif; | |
404 | ||
405 | err = ip6_dst_lookup(sk, &dst, &fl); | |
406 | if (err) | |
407 | goto out; | |
8de3351e YH |
408 | |
409 | /* | |
410 | * We won't send icmp if the destination is known | |
411 | * anycast. | |
412 | */ | |
413 | if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) { | |
414 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: acast source\n"); | |
415 | goto out_dst_release; | |
416 | } | |
417 | ||
1da177e4 | 418 | if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) |
e104411b | 419 | goto out; |
1da177e4 LT |
420 | |
421 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) | |
422 | hlimit = np->mcast_hops; | |
423 | else | |
424 | hlimit = np->hop_limit; | |
425 | if (hlimit < 0) | |
426 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | |
427 | if (hlimit < 0) | |
428 | hlimit = ipv6_get_hoplimit(dst->dev); | |
429 | ||
e012d51c | 430 | tclass = np->tclass; |
41a1f8ea YH |
431 | if (tclass < 0) |
432 | tclass = 0; | |
433 | ||
1da177e4 | 434 | msg.skb = skb; |
bbe735e4 | 435 | msg.offset = skb_network_offset(skb); |
763ecff1 | 436 | msg.type = type; |
1da177e4 LT |
437 | |
438 | len = skb->len - msg.offset; | |
439 | len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr)); | |
440 | if (len < 0) { | |
64ce2073 | 441 | LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n"); |
1da177e4 LT |
442 | goto out_dst_release; |
443 | } | |
444 | ||
445 | idev = in6_dev_get(skb->dev); | |
446 | ||
447 | err = ip6_append_data(sk, icmpv6_getfrag, &msg, | |
448 | len + sizeof(struct icmp6hdr), | |
449 | sizeof(struct icmp6hdr), | |
41a1f8ea | 450 | hlimit, tclass, NULL, &fl, (struct rt6_info*)dst, |
1da177e4 LT |
451 | MSG_DONTWAIT); |
452 | if (err) { | |
453 | ip6_flush_pending_frames(sk); | |
454 | goto out_put; | |
455 | } | |
456 | err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, len + sizeof(struct icmp6hdr)); | |
457 | ||
458 | if (type >= ICMPV6_DEST_UNREACH && type <= ICMPV6_PARAMPROB) | |
459 | ICMP6_INC_STATS_OFFSET_BH(idev, ICMP6_MIB_OUTDESTUNREACHS, type - ICMPV6_DEST_UNREACH); | |
460 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTMSGS); | |
461 | ||
462 | out_put: | |
463 | if (likely(idev != NULL)) | |
464 | in6_dev_put(idev); | |
465 | out_dst_release: | |
466 | dst_release(dst); | |
467 | out: | |
468 | icmpv6_xmit_unlock(); | |
469 | } | |
470 | ||
7159039a YH |
471 | EXPORT_SYMBOL(icmpv6_send); |
472 | ||
1da177e4 LT |
473 | static void icmpv6_echo_reply(struct sk_buff *skb) |
474 | { | |
84427d53 | 475 | struct sock *sk; |
1da177e4 | 476 | struct inet6_dev *idev; |
84427d53 | 477 | struct ipv6_pinfo *np; |
1da177e4 LT |
478 | struct in6_addr *saddr = NULL; |
479 | struct icmp6hdr *icmph = (struct icmp6hdr *) skb->h.raw; | |
480 | struct icmp6hdr tmp_hdr; | |
481 | struct flowi fl; | |
482 | struct icmpv6_msg msg; | |
483 | struct dst_entry *dst; | |
484 | int err = 0; | |
485 | int hlimit; | |
41a1f8ea | 486 | int tclass; |
1da177e4 | 487 | |
0660e03f | 488 | saddr = &ipv6_hdr(skb)->daddr; |
1da177e4 LT |
489 | |
490 | if (!ipv6_unicast_destination(skb)) | |
491 | saddr = NULL; | |
492 | ||
493 | memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr)); | |
494 | tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY; | |
495 | ||
496 | memset(&fl, 0, sizeof(fl)); | |
497 | fl.proto = IPPROTO_ICMPV6; | |
0660e03f | 498 | ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr); |
1da177e4 LT |
499 | if (saddr) |
500 | ipv6_addr_copy(&fl.fl6_src, saddr); | |
501 | fl.oif = skb->dev->ifindex; | |
502 | fl.fl_icmp_type = ICMPV6_ECHO_REPLY; | |
beb8d13b | 503 | security_skb_classify_flow(skb, &fl); |
1da177e4 LT |
504 | |
505 | if (icmpv6_xmit_lock()) | |
506 | return; | |
507 | ||
84427d53 YH |
508 | sk = icmpv6_socket->sk; |
509 | np = inet6_sk(sk); | |
510 | ||
1da177e4 LT |
511 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) |
512 | fl.oif = np->mcast_oif; | |
513 | ||
514 | err = ip6_dst_lookup(sk, &dst, &fl); | |
515 | if (err) | |
516 | goto out; | |
517 | if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) | |
e104411b | 518 | goto out; |
1da177e4 LT |
519 | |
520 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) | |
521 | hlimit = np->mcast_hops; | |
522 | else | |
523 | hlimit = np->hop_limit; | |
524 | if (hlimit < 0) | |
525 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | |
526 | if (hlimit < 0) | |
527 | hlimit = ipv6_get_hoplimit(dst->dev); | |
528 | ||
e012d51c | 529 | tclass = np->tclass; |
41a1f8ea YH |
530 | if (tclass < 0) |
531 | tclass = 0; | |
532 | ||
1da177e4 LT |
533 | idev = in6_dev_get(skb->dev); |
534 | ||
535 | msg.skb = skb; | |
536 | msg.offset = 0; | |
763ecff1 | 537 | msg.type = ICMPV6_ECHO_REPLY; |
1da177e4 LT |
538 | |
539 | err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr), | |
41a1f8ea | 540 | sizeof(struct icmp6hdr), hlimit, tclass, NULL, &fl, |
1da177e4 LT |
541 | (struct rt6_info*)dst, MSG_DONTWAIT); |
542 | ||
543 | if (err) { | |
544 | ip6_flush_pending_frames(sk); | |
545 | goto out_put; | |
546 | } | |
547 | err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, skb->len + sizeof(struct icmp6hdr)); | |
548 | ||
1ab1457c YH |
549 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTECHOREPLIES); |
550 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTMSGS); | |
1da177e4 | 551 | |
1ab1457c | 552 | out_put: |
1da177e4 LT |
553 | if (likely(idev != NULL)) |
554 | in6_dev_put(idev); | |
1da177e4 | 555 | dst_release(dst); |
1ab1457c | 556 | out: |
1da177e4 LT |
557 | icmpv6_xmit_unlock(); |
558 | } | |
559 | ||
04ce6909 | 560 | static void icmpv6_notify(struct sk_buff *skb, int type, int code, __be32 info) |
1da177e4 LT |
561 | { |
562 | struct in6_addr *saddr, *daddr; | |
563 | struct inet6_protocol *ipprot; | |
564 | struct sock *sk; | |
565 | int inner_offset; | |
566 | int hash; | |
567 | u8 nexthdr; | |
568 | ||
569 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
570 | return; | |
571 | ||
572 | nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr; | |
573 | if (ipv6_ext_hdr(nexthdr)) { | |
574 | /* now skip over extension headers */ | |
0d3d077c | 575 | inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr); |
1da177e4 LT |
576 | if (inner_offset<0) |
577 | return; | |
578 | } else { | |
579 | inner_offset = sizeof(struct ipv6hdr); | |
580 | } | |
581 | ||
582 | /* Checkin header including 8 bytes of inner protocol header. */ | |
583 | if (!pskb_may_pull(skb, inner_offset+8)) | |
584 | return; | |
585 | ||
0660e03f ACM |
586 | saddr = &ipv6_hdr(skb)->saddr; |
587 | daddr = &ipv6_hdr(skb)->daddr; | |
1da177e4 LT |
588 | |
589 | /* BUGGG_FUTURE: we should try to parse exthdrs in this packet. | |
590 | Without this we will not able f.e. to make source routed | |
591 | pmtu discovery. | |
592 | Corresponding argument (opt) to notifiers is already added. | |
593 | --ANK (980726) | |
594 | */ | |
595 | ||
596 | hash = nexthdr & (MAX_INET_PROTOS - 1); | |
597 | ||
598 | rcu_read_lock(); | |
599 | ipprot = rcu_dereference(inet6_protos[hash]); | |
600 | if (ipprot && ipprot->err_handler) | |
601 | ipprot->err_handler(skb, NULL, type, code, inner_offset, info); | |
602 | rcu_read_unlock(); | |
603 | ||
604 | read_lock(&raw_v6_lock); | |
605 | if ((sk = sk_head(&raw_v6_htable[hash])) != NULL) { | |
0bd1b59b | 606 | while((sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, |
2dac4b96 | 607 | IP6CB(skb)->iif))) { |
1da177e4 LT |
608 | rawv6_err(sk, skb, NULL, type, code, inner_offset, info); |
609 | sk = sk_next(sk); | |
610 | } | |
611 | } | |
612 | read_unlock(&raw_v6_lock); | |
613 | } | |
1ab1457c | 614 | |
1da177e4 LT |
615 | /* |
616 | * Handle icmp messages | |
617 | */ | |
618 | ||
951dbc8a | 619 | static int icmpv6_rcv(struct sk_buff **pskb) |
1da177e4 LT |
620 | { |
621 | struct sk_buff *skb = *pskb; | |
622 | struct net_device *dev = skb->dev; | |
623 | struct inet6_dev *idev = __in6_dev_get(dev); | |
624 | struct in6_addr *saddr, *daddr; | |
625 | struct ipv6hdr *orig_hdr; | |
626 | struct icmp6hdr *hdr; | |
627 | int type; | |
628 | ||
629 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INMSGS); | |
630 | ||
0660e03f ACM |
631 | saddr = &ipv6_hdr(skb)->saddr; |
632 | daddr = &ipv6_hdr(skb)->daddr; | |
1da177e4 LT |
633 | |
634 | /* Perform checksum. */ | |
fb286bb2 | 635 | switch (skb->ip_summed) { |
84fa7933 | 636 | case CHECKSUM_COMPLETE: |
fb286bb2 HX |
637 | if (!csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_ICMPV6, |
638 | skb->csum)) | |
639 | break; | |
640 | /* fall through */ | |
641 | case CHECKSUM_NONE: | |
868c86bc AV |
642 | skb->csum = ~csum_unfold(csum_ipv6_magic(saddr, daddr, skb->len, |
643 | IPPROTO_ICMPV6, 0)); | |
fb286bb2 | 644 | if (__skb_checksum_complete(skb)) { |
46b86a2d | 645 | LIMIT_NETDEBUG(KERN_DEBUG "ICMPv6 checksum failed [" NIP6_FMT " > " NIP6_FMT "]\n", |
64ce2073 | 646 | NIP6(*saddr), NIP6(*daddr)); |
1da177e4 LT |
647 | goto discard_it; |
648 | } | |
649 | } | |
650 | ||
651 | if (!pskb_pull(skb, sizeof(struct icmp6hdr))) | |
652 | goto discard_it; | |
653 | ||
654 | hdr = (struct icmp6hdr *) skb->h.raw; | |
655 | ||
656 | type = hdr->icmp6_type; | |
657 | ||
658 | if (type >= ICMPV6_DEST_UNREACH && type <= ICMPV6_PARAMPROB) | |
659 | ICMP6_INC_STATS_OFFSET_BH(idev, ICMP6_MIB_INDESTUNREACHS, type - ICMPV6_DEST_UNREACH); | |
660 | else if (type >= ICMPV6_ECHO_REQUEST && type <= NDISC_REDIRECT) | |
661 | ICMP6_INC_STATS_OFFSET_BH(idev, ICMP6_MIB_INECHOS, type - ICMPV6_ECHO_REQUEST); | |
662 | ||
663 | switch (type) { | |
664 | case ICMPV6_ECHO_REQUEST: | |
665 | icmpv6_echo_reply(skb); | |
666 | break; | |
667 | ||
668 | case ICMPV6_ECHO_REPLY: | |
669 | /* we couldn't care less */ | |
670 | break; | |
671 | ||
672 | case ICMPV6_PKT_TOOBIG: | |
673 | /* BUGGG_FUTURE: if packet contains rthdr, we cannot update | |
674 | standard destination cache. Seems, only "advanced" | |
675 | destination cache will allow to solve this problem | |
676 | --ANK (980726) | |
677 | */ | |
678 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
679 | goto discard_it; | |
680 | hdr = (struct icmp6hdr *) skb->h.raw; | |
681 | orig_hdr = (struct ipv6hdr *) (hdr + 1); | |
682 | rt6_pmtu_discovery(&orig_hdr->daddr, &orig_hdr->saddr, dev, | |
683 | ntohl(hdr->icmp6_mtu)); | |
684 | ||
685 | /* | |
686 | * Drop through to notify | |
687 | */ | |
688 | ||
689 | case ICMPV6_DEST_UNREACH: | |
690 | case ICMPV6_TIME_EXCEED: | |
691 | case ICMPV6_PARAMPROB: | |
692 | icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); | |
693 | break; | |
694 | ||
695 | case NDISC_ROUTER_SOLICITATION: | |
696 | case NDISC_ROUTER_ADVERTISEMENT: | |
697 | case NDISC_NEIGHBOUR_SOLICITATION: | |
698 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
699 | case NDISC_REDIRECT: | |
700 | ndisc_rcv(skb); | |
701 | break; | |
702 | ||
703 | case ICMPV6_MGM_QUERY: | |
704 | igmp6_event_query(skb); | |
705 | break; | |
706 | ||
707 | case ICMPV6_MGM_REPORT: | |
708 | igmp6_event_report(skb); | |
709 | break; | |
710 | ||
711 | case ICMPV6_MGM_REDUCTION: | |
712 | case ICMPV6_NI_QUERY: | |
713 | case ICMPV6_NI_REPLY: | |
714 | case ICMPV6_MLD2_REPORT: | |
715 | case ICMPV6_DHAAD_REQUEST: | |
716 | case ICMPV6_DHAAD_REPLY: | |
717 | case ICMPV6_MOBILE_PREFIX_SOL: | |
718 | case ICMPV6_MOBILE_PREFIX_ADV: | |
719 | break; | |
720 | ||
721 | default: | |
64ce2073 | 722 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n"); |
1da177e4 LT |
723 | |
724 | /* informational */ | |
725 | if (type & ICMPV6_INFOMSG_MASK) | |
726 | break; | |
727 | ||
1ab1457c YH |
728 | /* |
729 | * error of unknown type. | |
730 | * must pass to upper level | |
1da177e4 LT |
731 | */ |
732 | ||
733 | icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); | |
734 | }; | |
735 | kfree_skb(skb); | |
736 | return 0; | |
737 | ||
738 | discard_it: | |
739 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INERRORS); | |
740 | kfree_skb(skb); | |
741 | return 0; | |
742 | } | |
743 | ||
640c41c7 IM |
744 | /* |
745 | * Special lock-class for __icmpv6_socket: | |
746 | */ | |
747 | static struct lock_class_key icmpv6_socket_sk_dst_lock_key; | |
748 | ||
1da177e4 LT |
749 | int __init icmpv6_init(struct net_proto_family *ops) |
750 | { | |
751 | struct sock *sk; | |
752 | int err, i, j; | |
753 | ||
6f912042 | 754 | for_each_possible_cpu(i) { |
1da177e4 LT |
755 | err = sock_create_kern(PF_INET6, SOCK_RAW, IPPROTO_ICMPV6, |
756 | &per_cpu(__icmpv6_socket, i)); | |
757 | if (err < 0) { | |
758 | printk(KERN_ERR | |
759 | "Failed to initialize the ICMP6 control socket " | |
760 | "(err %d).\n", | |
761 | err); | |
762 | goto fail; | |
763 | } | |
764 | ||
765 | sk = per_cpu(__icmpv6_socket, i)->sk; | |
766 | sk->sk_allocation = GFP_ATOMIC; | |
640c41c7 IM |
767 | /* |
768 | * Split off their lock-class, because sk->sk_dst_lock | |
769 | * gets used from softirqs, which is safe for | |
770 | * __icmpv6_socket (because those never get directly used | |
771 | * via userspace syscalls), but unsafe for normal sockets. | |
772 | */ | |
773 | lockdep_set_class(&sk->sk_dst_lock, | |
774 | &icmpv6_socket_sk_dst_lock_key); | |
1da177e4 LT |
775 | |
776 | /* Enough space for 2 64K ICMP packets, including | |
777 | * sk_buff struct overhead. | |
778 | */ | |
779 | sk->sk_sndbuf = | |
780 | (2 * ((64 * 1024) + sizeof(struct sk_buff))); | |
781 | ||
782 | sk->sk_prot->unhash(sk); | |
783 | } | |
784 | ||
785 | ||
786 | if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0) { | |
787 | printk(KERN_ERR "Failed to register ICMP6 protocol\n"); | |
788 | err = -EAGAIN; | |
789 | goto fail; | |
790 | } | |
791 | ||
792 | return 0; | |
793 | ||
794 | fail: | |
795 | for (j = 0; j < i; j++) { | |
796 | if (!cpu_possible(j)) | |
797 | continue; | |
798 | sock_release(per_cpu(__icmpv6_socket, j)); | |
799 | } | |
800 | ||
801 | return err; | |
802 | } | |
803 | ||
804 | void icmpv6_cleanup(void) | |
805 | { | |
806 | int i; | |
807 | ||
6f912042 | 808 | for_each_possible_cpu(i) { |
1da177e4 LT |
809 | sock_release(per_cpu(__icmpv6_socket, i)); |
810 | } | |
811 | inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); | |
812 | } | |
813 | ||
9b5b5cff | 814 | static const struct icmp6_err { |
1da177e4 LT |
815 | int err; |
816 | int fatal; | |
817 | } tab_unreach[] = { | |
818 | { /* NOROUTE */ | |
819 | .err = ENETUNREACH, | |
820 | .fatal = 0, | |
821 | }, | |
822 | { /* ADM_PROHIBITED */ | |
823 | .err = EACCES, | |
824 | .fatal = 1, | |
825 | }, | |
826 | { /* Was NOT_NEIGHBOUR, now reserved */ | |
827 | .err = EHOSTUNREACH, | |
828 | .fatal = 0, | |
829 | }, | |
830 | { /* ADDR_UNREACH */ | |
831 | .err = EHOSTUNREACH, | |
832 | .fatal = 0, | |
833 | }, | |
834 | { /* PORT_UNREACH */ | |
835 | .err = ECONNREFUSED, | |
836 | .fatal = 1, | |
837 | }, | |
838 | }; | |
839 | ||
840 | int icmpv6_err_convert(int type, int code, int *err) | |
841 | { | |
842 | int fatal = 0; | |
843 | ||
844 | *err = EPROTO; | |
845 | ||
846 | switch (type) { | |
847 | case ICMPV6_DEST_UNREACH: | |
848 | fatal = 1; | |
849 | if (code <= ICMPV6_PORT_UNREACH) { | |
850 | *err = tab_unreach[code].err; | |
851 | fatal = tab_unreach[code].fatal; | |
852 | } | |
853 | break; | |
854 | ||
855 | case ICMPV6_PKT_TOOBIG: | |
856 | *err = EMSGSIZE; | |
857 | break; | |
1ab1457c | 858 | |
1da177e4 LT |
859 | case ICMPV6_PARAMPROB: |
860 | *err = EPROTO; | |
861 | fatal = 1; | |
862 | break; | |
863 | ||
864 | case ICMPV6_TIME_EXCEED: | |
865 | *err = EHOSTUNREACH; | |
866 | break; | |
867 | }; | |
868 | ||
869 | return fatal; | |
870 | } | |
871 | ||
7159039a YH |
872 | EXPORT_SYMBOL(icmpv6_err_convert); |
873 | ||
1da177e4 LT |
874 | #ifdef CONFIG_SYSCTL |
875 | ctl_table ipv6_icmp_table[] = { | |
876 | { | |
877 | .ctl_name = NET_IPV6_ICMP_RATELIMIT, | |
878 | .procname = "ratelimit", | |
879 | .data = &sysctl_icmpv6_time, | |
880 | .maxlen = sizeof(int), | |
881 | .mode = 0644, | |
882 | .proc_handler = &proc_dointvec | |
883 | }, | |
884 | { .ctl_name = 0 }, | |
885 | }; | |
886 | #endif | |
887 |