]>
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); |
14878f75 DS |
72 | DEFINE_SNMP_STAT(struct icmpv6msg_mib, icmpv6msg_statistics) __read_mostly; |
73 | EXPORT_SYMBOL(icmpv6msg_statistics); | |
1da177e4 LT |
74 | |
75 | /* | |
76 | * The ICMP socket(s). This is the most convenient way to flow control | |
77 | * our ICMP output as well as maintain a clean interface throughout | |
78 | * all layers. All Socketless IP sends will soon be gone. | |
79 | * | |
80 | * On SMP we have one ICMP socket per-cpu. | |
81 | */ | |
82 | static DEFINE_PER_CPU(struct socket *, __icmpv6_socket) = NULL; | |
83 | #define icmpv6_socket __get_cpu_var(__icmpv6_socket) | |
84 | ||
e5bbef20 | 85 | static int icmpv6_rcv(struct sk_buff *skb); |
1da177e4 LT |
86 | |
87 | static struct inet6_protocol icmpv6_protocol = { | |
88 | .handler = icmpv6_rcv, | |
89 | .flags = INET6_PROTO_FINAL, | |
90 | }; | |
91 | ||
92 | static __inline__ int icmpv6_xmit_lock(void) | |
93 | { | |
94 | local_bh_disable(); | |
95 | ||
96 | if (unlikely(!spin_trylock(&icmpv6_socket->sk->sk_lock.slock))) { | |
97 | /* This can happen if the output path (f.e. SIT or | |
98 | * ip6ip6 tunnel) signals dst_link_failure() for an | |
99 | * outgoing ICMP6 packet. | |
100 | */ | |
101 | local_bh_enable(); | |
102 | return 1; | |
103 | } | |
104 | return 0; | |
105 | } | |
106 | ||
107 | static __inline__ void icmpv6_xmit_unlock(void) | |
108 | { | |
109 | spin_unlock_bh(&icmpv6_socket->sk->sk_lock.slock); | |
110 | } | |
111 | ||
1ab1457c | 112 | /* |
1da177e4 LT |
113 | * Slightly more convenient version of icmpv6_send. |
114 | */ | |
115 | void icmpv6_param_prob(struct sk_buff *skb, int code, int pos) | |
116 | { | |
117 | icmpv6_send(skb, ICMPV6_PARAMPROB, code, pos, skb->dev); | |
118 | kfree_skb(skb); | |
119 | } | |
120 | ||
121 | /* | |
122 | * Figure out, may we reply to this packet with icmp error. | |
123 | * | |
124 | * We do not reply, if: | |
125 | * - it was icmp error message. | |
126 | * - it is truncated, so that it is known, that protocol is ICMPV6 | |
127 | * (i.e. in the middle of some exthdr) | |
128 | * | |
129 | * --ANK (980726) | |
130 | */ | |
131 | ||
132 | static int is_ineligible(struct sk_buff *skb) | |
133 | { | |
0660e03f | 134 | int ptr = (u8 *)(ipv6_hdr(skb) + 1) - skb->data; |
1da177e4 | 135 | int len = skb->len - ptr; |
0660e03f | 136 | __u8 nexthdr = ipv6_hdr(skb)->nexthdr; |
1da177e4 LT |
137 | |
138 | if (len < 0) | |
139 | return 1; | |
140 | ||
0d3d077c | 141 | ptr = ipv6_skip_exthdr(skb, ptr, &nexthdr); |
1da177e4 LT |
142 | if (ptr < 0) |
143 | return 0; | |
144 | if (nexthdr == IPPROTO_ICMPV6) { | |
145 | u8 _type, *tp; | |
146 | tp = skb_header_pointer(skb, | |
147 | ptr+offsetof(struct icmp6hdr, icmp6_type), | |
148 | sizeof(_type), &_type); | |
149 | if (tp == NULL || | |
150 | !(*tp & ICMPV6_INFOMSG_MASK)) | |
151 | return 1; | |
152 | } | |
153 | return 0; | |
154 | } | |
155 | ||
ab32ea5d | 156 | static int sysctl_icmpv6_time __read_mostly = 1*HZ; |
1da177e4 | 157 | |
1ab1457c YH |
158 | /* |
159 | * Check the ICMP output rate limit | |
1da177e4 LT |
160 | */ |
161 | static inline int icmpv6_xrlim_allow(struct sock *sk, int type, | |
162 | struct flowi *fl) | |
163 | { | |
164 | struct dst_entry *dst; | |
165 | int res = 0; | |
166 | ||
167 | /* Informational messages are not limited. */ | |
168 | if (type & ICMPV6_INFOMSG_MASK) | |
169 | return 1; | |
170 | ||
171 | /* Do not limit pmtu discovery, it would break it. */ | |
172 | if (type == ICMPV6_PKT_TOOBIG) | |
173 | return 1; | |
174 | ||
1ab1457c | 175 | /* |
1da177e4 LT |
176 | * Look up the output route. |
177 | * XXX: perhaps the expire for routing entries cloned by | |
178 | * this lookup should be more aggressive (not longer than timeout). | |
179 | */ | |
180 | dst = ip6_route_output(sk, fl); | |
181 | if (dst->error) { | |
a11d206d YH |
182 | IP6_INC_STATS(ip6_dst_idev(dst), |
183 | IPSTATS_MIB_OUTNOROUTES); | |
1da177e4 LT |
184 | } else if (dst->dev && (dst->dev->flags&IFF_LOOPBACK)) { |
185 | res = 1; | |
186 | } else { | |
187 | struct rt6_info *rt = (struct rt6_info *)dst; | |
188 | int tmo = sysctl_icmpv6_time; | |
189 | ||
190 | /* Give more bandwidth to wider prefixes. */ | |
191 | if (rt->rt6i_dst.plen < 128) | |
192 | tmo >>= ((128 - rt->rt6i_dst.plen)>>5); | |
193 | ||
194 | res = xrlim_allow(dst, tmo); | |
195 | } | |
196 | dst_release(dst); | |
197 | return res; | |
198 | } | |
199 | ||
200 | /* | |
201 | * an inline helper for the "simple" if statement below | |
202 | * checks if parameter problem report is caused by an | |
1ab1457c | 203 | * unrecognized IPv6 option that has the Option Type |
1da177e4 LT |
204 | * highest-order two bits set to 10 |
205 | */ | |
206 | ||
207 | static __inline__ int opt_unrec(struct sk_buff *skb, __u32 offset) | |
208 | { | |
209 | u8 _optval, *op; | |
210 | ||
bbe735e4 | 211 | offset += skb_network_offset(skb); |
1da177e4 LT |
212 | op = skb_header_pointer(skb, offset, sizeof(_optval), &_optval); |
213 | if (op == NULL) | |
214 | return 1; | |
215 | return (*op & 0xC0) == 0x80; | |
216 | } | |
217 | ||
218 | static int icmpv6_push_pending_frames(struct sock *sk, struct flowi *fl, struct icmp6hdr *thdr, int len) | |
219 | { | |
220 | struct sk_buff *skb; | |
221 | struct icmp6hdr *icmp6h; | |
222 | int err = 0; | |
223 | ||
224 | if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) | |
225 | goto out; | |
226 | ||
cc70ab26 | 227 | icmp6h = icmp6_hdr(skb); |
1da177e4 LT |
228 | memcpy(icmp6h, thdr, sizeof(struct icmp6hdr)); |
229 | icmp6h->icmp6_cksum = 0; | |
230 | ||
231 | if (skb_queue_len(&sk->sk_write_queue) == 1) { | |
232 | skb->csum = csum_partial((char *)icmp6h, | |
233 | sizeof(struct icmp6hdr), skb->csum); | |
234 | icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src, | |
235 | &fl->fl6_dst, | |
236 | len, fl->proto, | |
237 | skb->csum); | |
238 | } else { | |
868c86bc | 239 | __wsum tmp_csum = 0; |
1da177e4 LT |
240 | |
241 | skb_queue_walk(&sk->sk_write_queue, skb) { | |
242 | tmp_csum = csum_add(tmp_csum, skb->csum); | |
243 | } | |
244 | ||
245 | tmp_csum = csum_partial((char *)icmp6h, | |
246 | sizeof(struct icmp6hdr), tmp_csum); | |
868c86bc AV |
247 | icmp6h->icmp6_cksum = csum_ipv6_magic(&fl->fl6_src, |
248 | &fl->fl6_dst, | |
249 | len, fl->proto, | |
250 | tmp_csum); | |
1da177e4 | 251 | } |
1da177e4 LT |
252 | ip6_push_pending_frames(sk); |
253 | out: | |
254 | return err; | |
255 | } | |
256 | ||
257 | struct icmpv6_msg { | |
258 | struct sk_buff *skb; | |
259 | int offset; | |
763ecff1 | 260 | uint8_t type; |
1da177e4 LT |
261 | }; |
262 | ||
263 | static int icmpv6_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb) | |
264 | { | |
265 | struct icmpv6_msg *msg = (struct icmpv6_msg *) from; | |
266 | struct sk_buff *org_skb = msg->skb; | |
5f92a738 | 267 | __wsum csum = 0; |
1da177e4 LT |
268 | |
269 | csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset, | |
270 | to, len, csum); | |
271 | skb->csum = csum_block_add(skb->csum, csum, odd); | |
763ecff1 YK |
272 | if (!(msg->type & ICMPV6_INFOMSG_MASK)) |
273 | nf_ct_attach(skb, org_skb); | |
1da177e4 LT |
274 | return 0; |
275 | } | |
276 | ||
59fbb3a6 | 277 | #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) |
79383236 MN |
278 | static void mip6_addr_swap(struct sk_buff *skb) |
279 | { | |
0660e03f | 280 | struct ipv6hdr *iph = ipv6_hdr(skb); |
79383236 MN |
281 | struct inet6_skb_parm *opt = IP6CB(skb); |
282 | struct ipv6_destopt_hao *hao; | |
283 | struct in6_addr tmp; | |
284 | int off; | |
285 | ||
286 | if (opt->dsthao) { | |
287 | off = ipv6_find_tlv(skb, opt->dsthao, IPV6_TLV_HAO); | |
288 | if (likely(off >= 0)) { | |
d56f90a7 ACM |
289 | hao = (struct ipv6_destopt_hao *) |
290 | (skb_network_header(skb) + off); | |
79383236 MN |
291 | ipv6_addr_copy(&tmp, &iph->saddr); |
292 | ipv6_addr_copy(&iph->saddr, &hao->addr); | |
293 | ipv6_addr_copy(&hao->addr, &tmp); | |
294 | } | |
295 | } | |
296 | } | |
297 | #else | |
298 | static inline void mip6_addr_swap(struct sk_buff *skb) {} | |
299 | #endif | |
300 | ||
1da177e4 LT |
301 | /* |
302 | * Send an ICMP message in response to a packet in error | |
303 | */ | |
1ab1457c | 304 | void icmpv6_send(struct sk_buff *skb, int type, int code, __u32 info, |
1da177e4 LT |
305 | struct net_device *dev) |
306 | { | |
307 | struct inet6_dev *idev = NULL; | |
0660e03f | 308 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
84427d53 YH |
309 | struct sock *sk; |
310 | struct ipv6_pinfo *np; | |
1da177e4 LT |
311 | struct in6_addr *saddr = NULL; |
312 | struct dst_entry *dst; | |
313 | struct icmp6hdr tmp_hdr; | |
314 | struct flowi fl; | |
315 | struct icmpv6_msg msg; | |
316 | int iif = 0; | |
317 | int addr_type = 0; | |
318 | int len; | |
41a1f8ea | 319 | int hlimit, tclass; |
1da177e4 LT |
320 | int err = 0; |
321 | ||
27a884dc ACM |
322 | if ((u8 *)hdr < skb->head || |
323 | (skb->network_header + sizeof(*hdr)) > skb->tail) | |
1da177e4 LT |
324 | return; |
325 | ||
326 | /* | |
1ab1457c | 327 | * Make sure we respect the rules |
1da177e4 LT |
328 | * i.e. RFC 1885 2.4(e) |
329 | * Rule (e.1) is enforced by not using icmpv6_send | |
330 | * in any code that processes icmp errors. | |
331 | */ | |
332 | addr_type = ipv6_addr_type(&hdr->daddr); | |
333 | ||
334 | if (ipv6_chk_addr(&hdr->daddr, skb->dev, 0)) | |
335 | saddr = &hdr->daddr; | |
336 | ||
337 | /* | |
338 | * Dest addr check | |
339 | */ | |
340 | ||
341 | if ((addr_type & IPV6_ADDR_MULTICAST || skb->pkt_type != PACKET_HOST)) { | |
342 | if (type != ICMPV6_PKT_TOOBIG && | |
1ab1457c YH |
343 | !(type == ICMPV6_PARAMPROB && |
344 | code == ICMPV6_UNK_OPTION && | |
1da177e4 LT |
345 | (opt_unrec(skb, info)))) |
346 | return; | |
347 | ||
348 | saddr = NULL; | |
349 | } | |
350 | ||
351 | addr_type = ipv6_addr_type(&hdr->saddr); | |
352 | ||
353 | /* | |
354 | * Source addr check | |
355 | */ | |
356 | ||
357 | if (addr_type & IPV6_ADDR_LINKLOCAL) | |
358 | iif = skb->dev->ifindex; | |
359 | ||
360 | /* | |
8de3351e YH |
361 | * Must not send error if the source does not uniquely |
362 | * identify a single node (RFC2463 Section 2.4). | |
363 | * We check unspecified / multicast addresses here, | |
364 | * and anycast addresses will be checked later. | |
1da177e4 LT |
365 | */ |
366 | if ((addr_type == IPV6_ADDR_ANY) || (addr_type & IPV6_ADDR_MULTICAST)) { | |
64ce2073 | 367 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: addr_any/mcast source\n"); |
1da177e4 LT |
368 | return; |
369 | } | |
370 | ||
1ab1457c | 371 | /* |
1da177e4 LT |
372 | * Never answer to a ICMP packet. |
373 | */ | |
374 | if (is_ineligible(skb)) { | |
64ce2073 | 375 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: no reply to icmp error\n"); |
1da177e4 LT |
376 | return; |
377 | } | |
378 | ||
79383236 MN |
379 | mip6_addr_swap(skb); |
380 | ||
1da177e4 LT |
381 | memset(&fl, 0, sizeof(fl)); |
382 | fl.proto = IPPROTO_ICMPV6; | |
383 | ipv6_addr_copy(&fl.fl6_dst, &hdr->saddr); | |
384 | if (saddr) | |
385 | ipv6_addr_copy(&fl.fl6_src, saddr); | |
386 | fl.oif = iif; | |
387 | fl.fl_icmp_type = type; | |
388 | fl.fl_icmp_code = code; | |
beb8d13b | 389 | security_skb_classify_flow(skb, &fl); |
1da177e4 LT |
390 | |
391 | if (icmpv6_xmit_lock()) | |
392 | return; | |
393 | ||
84427d53 YH |
394 | sk = icmpv6_socket->sk; |
395 | np = inet6_sk(sk); | |
396 | ||
1da177e4 LT |
397 | if (!icmpv6_xrlim_allow(sk, type, &fl)) |
398 | goto out; | |
399 | ||
400 | tmp_hdr.icmp6_type = type; | |
401 | tmp_hdr.icmp6_code = code; | |
402 | tmp_hdr.icmp6_cksum = 0; | |
403 | tmp_hdr.icmp6_pointer = htonl(info); | |
404 | ||
405 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) | |
406 | fl.oif = np->mcast_oif; | |
407 | ||
408 | err = ip6_dst_lookup(sk, &dst, &fl); | |
409 | if (err) | |
410 | goto out; | |
8de3351e YH |
411 | |
412 | /* | |
413 | * We won't send icmp if the destination is known | |
414 | * anycast. | |
415 | */ | |
416 | if (((struct rt6_info *)dst)->rt6i_flags & RTF_ANYCAST) { | |
417 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6_send: acast source\n"); | |
418 | goto out_dst_release; | |
419 | } | |
420 | ||
1da177e4 | 421 | if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) |
e104411b | 422 | goto out; |
1da177e4 LT |
423 | |
424 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) | |
425 | hlimit = np->mcast_hops; | |
426 | else | |
427 | hlimit = np->hop_limit; | |
428 | if (hlimit < 0) | |
429 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | |
430 | if (hlimit < 0) | |
431 | hlimit = ipv6_get_hoplimit(dst->dev); | |
432 | ||
e012d51c | 433 | tclass = np->tclass; |
41a1f8ea YH |
434 | if (tclass < 0) |
435 | tclass = 0; | |
436 | ||
1da177e4 | 437 | msg.skb = skb; |
bbe735e4 | 438 | msg.offset = skb_network_offset(skb); |
763ecff1 | 439 | msg.type = type; |
1da177e4 LT |
440 | |
441 | len = skb->len - msg.offset; | |
442 | len = min_t(unsigned int, len, IPV6_MIN_MTU - sizeof(struct ipv6hdr) -sizeof(struct icmp6hdr)); | |
443 | if (len < 0) { | |
64ce2073 | 444 | LIMIT_NETDEBUG(KERN_DEBUG "icmp: len problem\n"); |
1da177e4 LT |
445 | goto out_dst_release; |
446 | } | |
447 | ||
448 | idev = in6_dev_get(skb->dev); | |
449 | ||
450 | err = ip6_append_data(sk, icmpv6_getfrag, &msg, | |
451 | len + sizeof(struct icmp6hdr), | |
452 | sizeof(struct icmp6hdr), | |
41a1f8ea | 453 | hlimit, tclass, NULL, &fl, (struct rt6_info*)dst, |
1da177e4 LT |
454 | MSG_DONTWAIT); |
455 | if (err) { | |
456 | ip6_flush_pending_frames(sk); | |
457 | goto out_put; | |
458 | } | |
459 | err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, len + sizeof(struct icmp6hdr)); | |
460 | ||
1da177e4 LT |
461 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTMSGS); |
462 | ||
463 | out_put: | |
464 | if (likely(idev != NULL)) | |
465 | in6_dev_put(idev); | |
466 | out_dst_release: | |
467 | dst_release(dst); | |
468 | out: | |
469 | icmpv6_xmit_unlock(); | |
470 | } | |
471 | ||
7159039a YH |
472 | EXPORT_SYMBOL(icmpv6_send); |
473 | ||
1da177e4 LT |
474 | static void icmpv6_echo_reply(struct sk_buff *skb) |
475 | { | |
84427d53 | 476 | struct sock *sk; |
1da177e4 | 477 | struct inet6_dev *idev; |
84427d53 | 478 | struct ipv6_pinfo *np; |
1da177e4 | 479 | struct in6_addr *saddr = NULL; |
cc70ab26 | 480 | struct icmp6hdr *icmph = icmp6_hdr(skb); |
1da177e4 LT |
481 | struct icmp6hdr tmp_hdr; |
482 | struct flowi fl; | |
483 | struct icmpv6_msg msg; | |
484 | struct dst_entry *dst; | |
485 | int err = 0; | |
486 | int hlimit; | |
41a1f8ea | 487 | int tclass; |
1da177e4 | 488 | |
0660e03f | 489 | saddr = &ipv6_hdr(skb)->daddr; |
1da177e4 LT |
490 | |
491 | if (!ipv6_unicast_destination(skb)) | |
492 | saddr = NULL; | |
493 | ||
494 | memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr)); | |
495 | tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY; | |
496 | ||
497 | memset(&fl, 0, sizeof(fl)); | |
498 | fl.proto = IPPROTO_ICMPV6; | |
0660e03f | 499 | ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr); |
1da177e4 LT |
500 | if (saddr) |
501 | ipv6_addr_copy(&fl.fl6_src, saddr); | |
502 | fl.oif = skb->dev->ifindex; | |
503 | fl.fl_icmp_type = ICMPV6_ECHO_REPLY; | |
beb8d13b | 504 | security_skb_classify_flow(skb, &fl); |
1da177e4 LT |
505 | |
506 | if (icmpv6_xmit_lock()) | |
507 | return; | |
508 | ||
84427d53 YH |
509 | sk = icmpv6_socket->sk; |
510 | np = inet6_sk(sk); | |
511 | ||
1da177e4 LT |
512 | if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) |
513 | fl.oif = np->mcast_oif; | |
514 | ||
515 | err = ip6_dst_lookup(sk, &dst, &fl); | |
516 | if (err) | |
517 | goto out; | |
518 | if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) | |
e104411b | 519 | goto out; |
1da177e4 LT |
520 | |
521 | if (ipv6_addr_is_multicast(&fl.fl6_dst)) | |
522 | hlimit = np->mcast_hops; | |
523 | else | |
524 | hlimit = np->hop_limit; | |
525 | if (hlimit < 0) | |
526 | hlimit = dst_metric(dst, RTAX_HOPLIMIT); | |
527 | if (hlimit < 0) | |
528 | hlimit = ipv6_get_hoplimit(dst->dev); | |
529 | ||
e012d51c | 530 | tclass = np->tclass; |
41a1f8ea YH |
531 | if (tclass < 0) |
532 | tclass = 0; | |
533 | ||
1da177e4 LT |
534 | idev = in6_dev_get(skb->dev); |
535 | ||
536 | msg.skb = skb; | |
537 | msg.offset = 0; | |
763ecff1 | 538 | msg.type = ICMPV6_ECHO_REPLY; |
1da177e4 LT |
539 | |
540 | err = ip6_append_data(sk, icmpv6_getfrag, &msg, skb->len + sizeof(struct icmp6hdr), | |
41a1f8ea | 541 | sizeof(struct icmp6hdr), hlimit, tclass, NULL, &fl, |
1da177e4 LT |
542 | (struct rt6_info*)dst, MSG_DONTWAIT); |
543 | ||
544 | if (err) { | |
545 | ip6_flush_pending_frames(sk); | |
546 | goto out_put; | |
547 | } | |
548 | err = icmpv6_push_pending_frames(sk, &fl, &tmp_hdr, skb->len + sizeof(struct icmp6hdr)); | |
549 | ||
1ab1457c | 550 | out_put: |
1da177e4 LT |
551 | if (likely(idev != NULL)) |
552 | in6_dev_put(idev); | |
1da177e4 | 553 | dst_release(dst); |
1ab1457c | 554 | out: |
1da177e4 LT |
555 | icmpv6_xmit_unlock(); |
556 | } | |
557 | ||
04ce6909 | 558 | static void icmpv6_notify(struct sk_buff *skb, int type, int code, __be32 info) |
1da177e4 LT |
559 | { |
560 | struct in6_addr *saddr, *daddr; | |
561 | struct inet6_protocol *ipprot; | |
562 | struct sock *sk; | |
563 | int inner_offset; | |
564 | int hash; | |
565 | u8 nexthdr; | |
566 | ||
567 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
568 | return; | |
569 | ||
570 | nexthdr = ((struct ipv6hdr *)skb->data)->nexthdr; | |
571 | if (ipv6_ext_hdr(nexthdr)) { | |
572 | /* now skip over extension headers */ | |
0d3d077c | 573 | inner_offset = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr); |
1da177e4 LT |
574 | if (inner_offset<0) |
575 | return; | |
576 | } else { | |
577 | inner_offset = sizeof(struct ipv6hdr); | |
578 | } | |
579 | ||
580 | /* Checkin header including 8 bytes of inner protocol header. */ | |
581 | if (!pskb_may_pull(skb, inner_offset+8)) | |
582 | return; | |
583 | ||
0660e03f ACM |
584 | saddr = &ipv6_hdr(skb)->saddr; |
585 | daddr = &ipv6_hdr(skb)->daddr; | |
1da177e4 LT |
586 | |
587 | /* BUGGG_FUTURE: we should try to parse exthdrs in this packet. | |
588 | Without this we will not able f.e. to make source routed | |
589 | pmtu discovery. | |
590 | Corresponding argument (opt) to notifiers is already added. | |
591 | --ANK (980726) | |
592 | */ | |
593 | ||
594 | hash = nexthdr & (MAX_INET_PROTOS - 1); | |
595 | ||
596 | rcu_read_lock(); | |
597 | ipprot = rcu_dereference(inet6_protos[hash]); | |
598 | if (ipprot && ipprot->err_handler) | |
599 | ipprot->err_handler(skb, NULL, type, code, inner_offset, info); | |
600 | rcu_read_unlock(); | |
601 | ||
602 | read_lock(&raw_v6_lock); | |
603 | if ((sk = sk_head(&raw_v6_htable[hash])) != NULL) { | |
f13ec93f | 604 | while ((sk = __raw_v6_lookup(sk, nexthdr, saddr, daddr, |
2dac4b96 | 605 | IP6CB(skb)->iif))) { |
1da177e4 LT |
606 | rawv6_err(sk, skb, NULL, type, code, inner_offset, info); |
607 | sk = sk_next(sk); | |
608 | } | |
609 | } | |
610 | read_unlock(&raw_v6_lock); | |
611 | } | |
1ab1457c | 612 | |
1da177e4 LT |
613 | /* |
614 | * Handle icmp messages | |
615 | */ | |
616 | ||
e5bbef20 | 617 | static int icmpv6_rcv(struct sk_buff *skb) |
1da177e4 | 618 | { |
1da177e4 LT |
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 | ||
0660e03f ACM |
628 | saddr = &ipv6_hdr(skb)->saddr; |
629 | daddr = &ipv6_hdr(skb)->daddr; | |
1da177e4 LT |
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 | ||
cc70ab26 | 651 | hdr = icmp6_hdr(skb); |
1da177e4 LT |
652 | |
653 | type = hdr->icmp6_type; | |
654 | ||
14878f75 | 655 | ICMP6MSGIN_INC_STATS_BH(idev, type); |
1da177e4 LT |
656 | |
657 | switch (type) { | |
658 | case ICMPV6_ECHO_REQUEST: | |
659 | icmpv6_echo_reply(skb); | |
660 | break; | |
661 | ||
662 | case ICMPV6_ECHO_REPLY: | |
663 | /* we couldn't care less */ | |
664 | break; | |
665 | ||
666 | case ICMPV6_PKT_TOOBIG: | |
667 | /* BUGGG_FUTURE: if packet contains rthdr, we cannot update | |
668 | standard destination cache. Seems, only "advanced" | |
669 | destination cache will allow to solve this problem | |
670 | --ANK (980726) | |
671 | */ | |
672 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
673 | goto discard_it; | |
cc70ab26 | 674 | hdr = icmp6_hdr(skb); |
1da177e4 LT |
675 | orig_hdr = (struct ipv6hdr *) (hdr + 1); |
676 | rt6_pmtu_discovery(&orig_hdr->daddr, &orig_hdr->saddr, dev, | |
677 | ntohl(hdr->icmp6_mtu)); | |
678 | ||
679 | /* | |
680 | * Drop through to notify | |
681 | */ | |
682 | ||
683 | case ICMPV6_DEST_UNREACH: | |
684 | case ICMPV6_TIME_EXCEED: | |
685 | case ICMPV6_PARAMPROB: | |
686 | icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); | |
687 | break; | |
688 | ||
689 | case NDISC_ROUTER_SOLICITATION: | |
690 | case NDISC_ROUTER_ADVERTISEMENT: | |
691 | case NDISC_NEIGHBOUR_SOLICITATION: | |
692 | case NDISC_NEIGHBOUR_ADVERTISEMENT: | |
693 | case NDISC_REDIRECT: | |
694 | ndisc_rcv(skb); | |
695 | break; | |
696 | ||
697 | case ICMPV6_MGM_QUERY: | |
698 | igmp6_event_query(skb); | |
699 | break; | |
700 | ||
701 | case ICMPV6_MGM_REPORT: | |
702 | igmp6_event_report(skb); | |
703 | break; | |
704 | ||
705 | case ICMPV6_MGM_REDUCTION: | |
706 | case ICMPV6_NI_QUERY: | |
707 | case ICMPV6_NI_REPLY: | |
708 | case ICMPV6_MLD2_REPORT: | |
709 | case ICMPV6_DHAAD_REQUEST: | |
710 | case ICMPV6_DHAAD_REPLY: | |
711 | case ICMPV6_MOBILE_PREFIX_SOL: | |
712 | case ICMPV6_MOBILE_PREFIX_ADV: | |
713 | break; | |
714 | ||
715 | default: | |
64ce2073 | 716 | LIMIT_NETDEBUG(KERN_DEBUG "icmpv6: msg of unknown type\n"); |
1da177e4 LT |
717 | |
718 | /* informational */ | |
719 | if (type & ICMPV6_INFOMSG_MASK) | |
720 | break; | |
721 | ||
1ab1457c YH |
722 | /* |
723 | * error of unknown type. | |
724 | * must pass to upper level | |
1da177e4 LT |
725 | */ |
726 | ||
727 | icmpv6_notify(skb, type, hdr->icmp6_code, hdr->icmp6_mtu); | |
3ff50b79 SH |
728 | } |
729 | ||
1da177e4 LT |
730 | kfree_skb(skb); |
731 | return 0; | |
732 | ||
733 | discard_it: | |
734 | ICMP6_INC_STATS_BH(idev, ICMP6_MIB_INERRORS); | |
735 | kfree_skb(skb); | |
736 | return 0; | |
737 | } | |
738 | ||
640c41c7 IM |
739 | /* |
740 | * Special lock-class for __icmpv6_socket: | |
741 | */ | |
742 | static struct lock_class_key icmpv6_socket_sk_dst_lock_key; | |
743 | ||
1da177e4 LT |
744 | int __init icmpv6_init(struct net_proto_family *ops) |
745 | { | |
746 | struct sock *sk; | |
747 | int err, i, j; | |
748 | ||
6f912042 | 749 | for_each_possible_cpu(i) { |
1da177e4 LT |
750 | err = sock_create_kern(PF_INET6, SOCK_RAW, IPPROTO_ICMPV6, |
751 | &per_cpu(__icmpv6_socket, i)); | |
752 | if (err < 0) { | |
753 | printk(KERN_ERR | |
754 | "Failed to initialize the ICMP6 control socket " | |
755 | "(err %d).\n", | |
756 | err); | |
757 | goto fail; | |
758 | } | |
759 | ||
760 | sk = per_cpu(__icmpv6_socket, i)->sk; | |
761 | sk->sk_allocation = GFP_ATOMIC; | |
640c41c7 IM |
762 | /* |
763 | * Split off their lock-class, because sk->sk_dst_lock | |
764 | * gets used from softirqs, which is safe for | |
765 | * __icmpv6_socket (because those never get directly used | |
766 | * via userspace syscalls), but unsafe for normal sockets. | |
767 | */ | |
768 | lockdep_set_class(&sk->sk_dst_lock, | |
769 | &icmpv6_socket_sk_dst_lock_key); | |
1da177e4 LT |
770 | |
771 | /* Enough space for 2 64K ICMP packets, including | |
772 | * sk_buff struct overhead. | |
773 | */ | |
774 | sk->sk_sndbuf = | |
775 | (2 * ((64 * 1024) + sizeof(struct sk_buff))); | |
776 | ||
777 | sk->sk_prot->unhash(sk); | |
778 | } | |
779 | ||
780 | ||
781 | if (inet6_add_protocol(&icmpv6_protocol, IPPROTO_ICMPV6) < 0) { | |
782 | printk(KERN_ERR "Failed to register ICMP6 protocol\n"); | |
783 | err = -EAGAIN; | |
784 | goto fail; | |
785 | } | |
786 | ||
787 | return 0; | |
788 | ||
789 | fail: | |
790 | for (j = 0; j < i; j++) { | |
791 | if (!cpu_possible(j)) | |
792 | continue; | |
793 | sock_release(per_cpu(__icmpv6_socket, j)); | |
794 | } | |
795 | ||
796 | return err; | |
797 | } | |
798 | ||
799 | void icmpv6_cleanup(void) | |
800 | { | |
801 | int i; | |
802 | ||
6f912042 | 803 | for_each_possible_cpu(i) { |
1da177e4 LT |
804 | sock_release(per_cpu(__icmpv6_socket, i)); |
805 | } | |
806 | inet6_del_protocol(&icmpv6_protocol, IPPROTO_ICMPV6); | |
807 | } | |
808 | ||
9b5b5cff | 809 | static const struct icmp6_err { |
1da177e4 LT |
810 | int err; |
811 | int fatal; | |
812 | } tab_unreach[] = { | |
813 | { /* NOROUTE */ | |
814 | .err = ENETUNREACH, | |
815 | .fatal = 0, | |
816 | }, | |
817 | { /* ADM_PROHIBITED */ | |
818 | .err = EACCES, | |
819 | .fatal = 1, | |
820 | }, | |
821 | { /* Was NOT_NEIGHBOUR, now reserved */ | |
822 | .err = EHOSTUNREACH, | |
823 | .fatal = 0, | |
824 | }, | |
825 | { /* ADDR_UNREACH */ | |
826 | .err = EHOSTUNREACH, | |
827 | .fatal = 0, | |
828 | }, | |
829 | { /* PORT_UNREACH */ | |
830 | .err = ECONNREFUSED, | |
831 | .fatal = 1, | |
832 | }, | |
833 | }; | |
834 | ||
835 | int icmpv6_err_convert(int type, int code, int *err) | |
836 | { | |
837 | int fatal = 0; | |
838 | ||
839 | *err = EPROTO; | |
840 | ||
841 | switch (type) { | |
842 | case ICMPV6_DEST_UNREACH: | |
843 | fatal = 1; | |
844 | if (code <= ICMPV6_PORT_UNREACH) { | |
845 | *err = tab_unreach[code].err; | |
846 | fatal = tab_unreach[code].fatal; | |
847 | } | |
848 | break; | |
849 | ||
850 | case ICMPV6_PKT_TOOBIG: | |
851 | *err = EMSGSIZE; | |
852 | break; | |
1ab1457c | 853 | |
1da177e4 LT |
854 | case ICMPV6_PARAMPROB: |
855 | *err = EPROTO; | |
856 | fatal = 1; | |
857 | break; | |
858 | ||
859 | case ICMPV6_TIME_EXCEED: | |
860 | *err = EHOSTUNREACH; | |
861 | break; | |
3ff50b79 | 862 | } |
1da177e4 LT |
863 | |
864 | return fatal; | |
865 | } | |
866 | ||
7159039a YH |
867 | EXPORT_SYMBOL(icmpv6_err_convert); |
868 | ||
1da177e4 LT |
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 |