Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * IPv6 fragment reassembly | |
1ab1457c | 3 | * Linux INET6 implementation |
1da177e4 LT |
4 | * |
5 | * Authors: | |
1ab1457c | 6 | * Pedro Roque <roque@di.fc.ul.pt> |
1da177e4 | 7 | * |
1da177e4 LT |
8 | * Based on: net/ipv4/ip_fragment.c |
9 | * | |
10 | * This program is free software; you can redistribute it and/or | |
11 | * modify it under the terms of the GNU General Public License | |
12 | * as published by the Free Software Foundation; either version | |
13 | * 2 of the License, or (at your option) any later version. | |
14 | */ | |
15 | ||
1ab1457c YH |
16 | /* |
17 | * Fixes: | |
1da177e4 LT |
18 | * Andi Kleen Make it work with multiple hosts. |
19 | * More RFC compliance. | |
20 | * | |
21 | * Horst von Brand Add missing #include <linux/string.h> | |
22 | * Alexey Kuznetsov SMP races, threading, cleanup. | |
23 | * Patrick McHardy LRU queue of frag heads for evictor. | |
24 | * Mitsuru KANDA @USAGI Register inet6_protocol{}. | |
25 | * David Stevens and | |
26 | * YOSHIFUJI,H. @USAGI Always remove fragment header to | |
27 | * calculate ICV correctly. | |
28 | */ | |
1da177e4 LT |
29 | #include <linux/errno.h> |
30 | #include <linux/types.h> | |
31 | #include <linux/string.h> | |
32 | #include <linux/socket.h> | |
33 | #include <linux/sockios.h> | |
34 | #include <linux/jiffies.h> | |
35 | #include <linux/net.h> | |
36 | #include <linux/list.h> | |
37 | #include <linux/netdevice.h> | |
38 | #include <linux/in6.h> | |
39 | #include <linux/ipv6.h> | |
40 | #include <linux/icmpv6.h> | |
41 | #include <linux/random.h> | |
42 | #include <linux/jhash.h> | |
f61944ef | 43 | #include <linux/skbuff.h> |
1da177e4 LT |
44 | |
45 | #include <net/sock.h> | |
46 | #include <net/snmp.h> | |
47 | ||
48 | #include <net/ipv6.h> | |
a11d206d | 49 | #include <net/ip6_route.h> |
1da177e4 LT |
50 | #include <net/protocol.h> |
51 | #include <net/transp_v6.h> | |
52 | #include <net/rawv6.h> | |
53 | #include <net/ndisc.h> | |
54 | #include <net/addrconf.h> | |
5ab11c98 | 55 | #include <net/inet_frag.h> |
1da177e4 | 56 | |
1da177e4 LT |
57 | struct ip6frag_skb_cb |
58 | { | |
59 | struct inet6_skb_parm h; | |
60 | int offset; | |
61 | }; | |
62 | ||
63 | #define FRAG6_CB(skb) ((struct ip6frag_skb_cb*)((skb)->cb)) | |
64 | ||
65 | ||
66 | /* | |
67 | * Equivalent of ipv4 struct ipq | |
68 | */ | |
69 | ||
70 | struct frag_queue | |
71 | { | |
5ab11c98 | 72 | struct inet_frag_queue q; |
1da177e4 | 73 | |
e69a4adc | 74 | __be32 id; /* fragment id */ |
1da177e4 LT |
75 | struct in6_addr saddr; |
76 | struct in6_addr daddr; | |
77 | ||
1da177e4 | 78 | int iif; |
1da177e4 | 79 | unsigned int csum; |
1da177e4 | 80 | __u16 nhoffset; |
1da177e4 LT |
81 | }; |
82 | ||
7eb95156 | 83 | static struct inet_frags ip6_frags; |
1da177e4 | 84 | |
e5a2bb84 | 85 | int ip6_frag_nqueues(struct net *net) |
7eb95156 | 86 | { |
e5a2bb84 | 87 | return net->ipv6.frags.nqueues; |
7eb95156 | 88 | } |
1da177e4 | 89 | |
6ddc0822 | 90 | int ip6_frag_mem(struct net *net) |
7eb95156 | 91 | { |
6ddc0822 | 92 | return atomic_read(&net->ipv6.frags.mem); |
7eb95156 | 93 | } |
1da177e4 | 94 | |
f61944ef HX |
95 | static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev, |
96 | struct net_device *dev); | |
97 | ||
f6596f9d ZB |
98 | /* |
99 | * callers should be careful not to use the hash value outside the ipfrag_lock | |
100 | * as doing so could race with ipfrag_hash_rnd being recalculated. | |
101 | */ | |
93c8b90f IJ |
102 | unsigned int inet6_hash_frag(__be32 id, const struct in6_addr *saddr, |
103 | const struct in6_addr *daddr, u32 rnd) | |
1da177e4 LT |
104 | { |
105 | u32 a, b, c; | |
106 | ||
e69a4adc AV |
107 | a = (__force u32)saddr->s6_addr32[0]; |
108 | b = (__force u32)saddr->s6_addr32[1]; | |
109 | c = (__force u32)saddr->s6_addr32[2]; | |
1da177e4 LT |
110 | |
111 | a += JHASH_GOLDEN_RATIO; | |
112 | b += JHASH_GOLDEN_RATIO; | |
93c8b90f | 113 | c += rnd; |
1da177e4 LT |
114 | __jhash_mix(a, b, c); |
115 | ||
e69a4adc AV |
116 | a += (__force u32)saddr->s6_addr32[3]; |
117 | b += (__force u32)daddr->s6_addr32[0]; | |
118 | c += (__force u32)daddr->s6_addr32[1]; | |
1da177e4 LT |
119 | __jhash_mix(a, b, c); |
120 | ||
e69a4adc AV |
121 | a += (__force u32)daddr->s6_addr32[2]; |
122 | b += (__force u32)daddr->s6_addr32[3]; | |
123 | c += (__force u32)id; | |
1da177e4 LT |
124 | __jhash_mix(a, b, c); |
125 | ||
7eb95156 | 126 | return c & (INETFRAGS_HASHSZ - 1); |
1da177e4 | 127 | } |
93c8b90f | 128 | EXPORT_SYMBOL_GPL(inet6_hash_frag); |
1da177e4 | 129 | |
321a3a99 | 130 | static unsigned int ip6_hashfn(struct inet_frag_queue *q) |
1da177e4 | 131 | { |
321a3a99 | 132 | struct frag_queue *fq; |
1da177e4 | 133 | |
321a3a99 | 134 | fq = container_of(q, struct frag_queue, q); |
93c8b90f | 135 | return inet6_hash_frag(fq->id, &fq->saddr, &fq->daddr, ip6_frags.rnd); |
1da177e4 LT |
136 | } |
137 | ||
abd6523d PE |
138 | int ip6_frag_match(struct inet_frag_queue *q, void *a) |
139 | { | |
140 | struct frag_queue *fq; | |
141 | struct ip6_create_arg *arg = a; | |
142 | ||
143 | fq = container_of(q, struct frag_queue, q); | |
144 | return (fq->id == arg->id && | |
145 | ipv6_addr_equal(&fq->saddr, arg->src) && | |
146 | ipv6_addr_equal(&fq->daddr, arg->dst)); | |
147 | } | |
148 | EXPORT_SYMBOL(ip6_frag_match); | |
149 | ||
1da177e4 | 150 | /* Memory Tracking Functions. */ |
6ddc0822 PE |
151 | static inline void frag_kfree_skb(struct netns_frags *nf, |
152 | struct sk_buff *skb, int *work) | |
1da177e4 LT |
153 | { |
154 | if (work) | |
155 | *work -= skb->truesize; | |
6ddc0822 | 156 | atomic_sub(skb->truesize, &nf->mem); |
1da177e4 LT |
157 | kfree_skb(skb); |
158 | } | |
159 | ||
c6fda282 | 160 | void ip6_frag_init(struct inet_frag_queue *q, void *a) |
1da177e4 | 161 | { |
c6fda282 PE |
162 | struct frag_queue *fq = container_of(q, struct frag_queue, q); |
163 | struct ip6_create_arg *arg = a; | |
164 | ||
165 | fq->id = arg->id; | |
166 | ipv6_addr_copy(&fq->saddr, arg->src); | |
167 | ipv6_addr_copy(&fq->daddr, arg->dst); | |
1da177e4 | 168 | } |
c6fda282 | 169 | EXPORT_SYMBOL(ip6_frag_init); |
1da177e4 | 170 | |
1da177e4 LT |
171 | /* Destruction primitives. */ |
172 | ||
4b6cb5d8 | 173 | static __inline__ void fq_put(struct frag_queue *fq) |
1da177e4 | 174 | { |
762cc408 | 175 | inet_frag_put(&fq->q, &ip6_frags); |
1da177e4 LT |
176 | } |
177 | ||
178 | /* Kill fq entry. It is not destroyed immediately, | |
179 | * because caller (and someone more) holds reference count. | |
180 | */ | |
181 | static __inline__ void fq_kill(struct frag_queue *fq) | |
182 | { | |
277e650d | 183 | inet_frag_kill(&fq->q, &ip6_frags); |
1da177e4 LT |
184 | } |
185 | ||
6ddc0822 | 186 | static void ip6_evictor(struct net *net, struct inet6_dev *idev) |
1da177e4 | 187 | { |
8e7999c4 PE |
188 | int evicted; |
189 | ||
6ddc0822 | 190 | evicted = inet_frag_evictor(&net->ipv6.frags, &ip6_frags); |
8e7999c4 | 191 | if (evicted) |
821d5777 | 192 | IP6_ADD_STATS_BH(net, idev, IPSTATS_MIB_REASMFAILS, evicted); |
1da177e4 LT |
193 | } |
194 | ||
195 | static void ip6_frag_expire(unsigned long data) | |
196 | { | |
e521db9d | 197 | struct frag_queue *fq; |
a11d206d | 198 | struct net_device *dev = NULL; |
4ac2ccd0 | 199 | struct net *net; |
1da177e4 | 200 | |
e521db9d PE |
201 | fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q); |
202 | ||
5ab11c98 | 203 | spin_lock(&fq->q.lock); |
1da177e4 | 204 | |
bc578a54 | 205 | if (fq->q.last_in & INET_FRAG_COMPLETE) |
1da177e4 LT |
206 | goto out; |
207 | ||
208 | fq_kill(fq); | |
209 | ||
4ac2ccd0 DL |
210 | net = container_of(fq->q.net, struct net, ipv6.frags); |
211 | dev = dev_get_by_index(net, fq->iif); | |
a11d206d YH |
212 | if (!dev) |
213 | goto out; | |
214 | ||
215 | rcu_read_lock(); | |
483a47d2 DL |
216 | IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMTIMEOUT); |
217 | IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS); | |
a11d206d | 218 | rcu_read_unlock(); |
1da177e4 | 219 | |
78c784c4 | 220 | /* Don't send error if the first segment did not arrive. */ |
bc578a54 | 221 | if (!(fq->q.last_in & INET_FRAG_FIRST_IN) || !fq->q.fragments) |
78c784c4 IO |
222 | goto out; |
223 | ||
78c784c4 IO |
224 | /* |
225 | But use as source device on which LAST ARRIVED | |
226 | segment was received. And do not use fq->dev | |
227 | pointer directly, device might already disappeared. | |
228 | */ | |
5ab11c98 PE |
229 | fq->q.fragments->dev = dev; |
230 | icmpv6_send(fq->q.fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0, dev); | |
1da177e4 | 231 | out: |
a11d206d YH |
232 | if (dev) |
233 | dev_put(dev); | |
5ab11c98 | 234 | spin_unlock(&fq->q.lock); |
4b6cb5d8 | 235 | fq_put(fq); |
1da177e4 LT |
236 | } |
237 | ||
abd6523d | 238 | static __inline__ struct frag_queue * |
ac18e750 | 239 | fq_find(struct net *net, __be32 id, struct in6_addr *src, struct in6_addr *dst, |
abd6523d | 240 | struct inet6_dev *idev) |
1da177e4 | 241 | { |
c6fda282 PE |
242 | struct inet_frag_queue *q; |
243 | struct ip6_create_arg arg; | |
abd6523d | 244 | unsigned int hash; |
1da177e4 | 245 | |
c6fda282 PE |
246 | arg.id = id; |
247 | arg.src = src; | |
248 | arg.dst = dst; | |
9a375803 PE |
249 | |
250 | read_lock(&ip6_frags.lock); | |
93c8b90f | 251 | hash = inet6_hash_frag(id, src, dst, ip6_frags.rnd); |
1da177e4 | 252 | |
ac18e750 | 253 | q = inet_frag_find(&net->ipv6.frags, &ip6_frags, &arg, hash); |
c6fda282 PE |
254 | if (q == NULL) |
255 | goto oom; | |
1da177e4 | 256 | |
c6fda282 | 257 | return container_of(q, struct frag_queue, q); |
1da177e4 LT |
258 | |
259 | oom: | |
483a47d2 | 260 | IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_REASMFAILS); |
1da177e4 LT |
261 | return NULL; |
262 | } | |
263 | ||
f61944ef | 264 | static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb, |
1da177e4 LT |
265 | struct frag_hdr *fhdr, int nhoff) |
266 | { | |
267 | struct sk_buff *prev, *next; | |
f61944ef | 268 | struct net_device *dev; |
1da177e4 | 269 | int offset, end; |
adf30907 | 270 | struct net *net = dev_net(skb_dst(skb)->dev); |
1da177e4 | 271 | |
bc578a54 | 272 | if (fq->q.last_in & INET_FRAG_COMPLETE) |
1da177e4 LT |
273 | goto err; |
274 | ||
275 | offset = ntohs(fhdr->frag_off) & ~0x7; | |
0660e03f ACM |
276 | end = offset + (ntohs(ipv6_hdr(skb)->payload_len) - |
277 | ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1))); | |
1da177e4 LT |
278 | |
279 | if ((unsigned int)end > IPV6_MAXPLEN) { | |
adf30907 | 280 | IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 281 | IPSTATS_MIB_INHDRERRORS); |
d56f90a7 ACM |
282 | icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, |
283 | ((u8 *)&fhdr->frag_off - | |
284 | skb_network_header(skb))); | |
f61944ef | 285 | return -1; |
1da177e4 LT |
286 | } |
287 | ||
d56f90a7 ACM |
288 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
289 | const unsigned char *nh = skb_network_header(skb); | |
1ab1457c | 290 | skb->csum = csum_sub(skb->csum, |
d56f90a7 ACM |
291 | csum_partial(nh, (u8 *)(fhdr + 1) - nh, |
292 | 0)); | |
293 | } | |
1da177e4 LT |
294 | |
295 | /* Is this the final fragment? */ | |
296 | if (!(fhdr->frag_off & htons(IP6_MF))) { | |
297 | /* If we already have some bits beyond end | |
298 | * or have different end, the segment is corrupted. | |
299 | */ | |
5ab11c98 | 300 | if (end < fq->q.len || |
bc578a54 | 301 | ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len)) |
1da177e4 | 302 | goto err; |
bc578a54 | 303 | fq->q.last_in |= INET_FRAG_LAST_IN; |
5ab11c98 | 304 | fq->q.len = end; |
1da177e4 LT |
305 | } else { |
306 | /* Check if the fragment is rounded to 8 bytes. | |
307 | * Required by the RFC. | |
308 | */ | |
309 | if (end & 0x7) { | |
310 | /* RFC2460 says always send parameter problem in | |
311 | * this case. -DaveM | |
312 | */ | |
adf30907 | 313 | IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 314 | IPSTATS_MIB_INHDRERRORS); |
1ab1457c | 315 | icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, |
1da177e4 | 316 | offsetof(struct ipv6hdr, payload_len)); |
f61944ef | 317 | return -1; |
1da177e4 | 318 | } |
5ab11c98 | 319 | if (end > fq->q.len) { |
1da177e4 | 320 | /* Some bits beyond end -> corruption. */ |
bc578a54 | 321 | if (fq->q.last_in & INET_FRAG_LAST_IN) |
1da177e4 | 322 | goto err; |
5ab11c98 | 323 | fq->q.len = end; |
1da177e4 LT |
324 | } |
325 | } | |
326 | ||
327 | if (end == offset) | |
328 | goto err; | |
329 | ||
330 | /* Point into the IP datagram 'data' part. */ | |
331 | if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data)) | |
332 | goto err; | |
1ab1457c | 333 | |
42ca89c1 SH |
334 | if (pskb_trim_rcsum(skb, end - offset)) |
335 | goto err; | |
1da177e4 LT |
336 | |
337 | /* Find out which fragments are in front and at the back of us | |
338 | * in the chain of fragments so far. We must know where to put | |
339 | * this fragment, right? | |
340 | */ | |
341 | prev = NULL; | |
5ab11c98 | 342 | for(next = fq->q.fragments; next != NULL; next = next->next) { |
1da177e4 LT |
343 | if (FRAG6_CB(next)->offset >= offset) |
344 | break; /* bingo! */ | |
345 | prev = next; | |
346 | } | |
347 | ||
348 | /* We found where to put this one. Check for overlap with | |
349 | * preceding fragment, and, if needed, align things so that | |
350 | * any overlaps are eliminated. | |
351 | */ | |
352 | if (prev) { | |
353 | int i = (FRAG6_CB(prev)->offset + prev->len) - offset; | |
354 | ||
355 | if (i > 0) { | |
356 | offset += i; | |
357 | if (end <= offset) | |
358 | goto err; | |
359 | if (!pskb_pull(skb, i)) | |
360 | goto err; | |
361 | if (skb->ip_summed != CHECKSUM_UNNECESSARY) | |
362 | skb->ip_summed = CHECKSUM_NONE; | |
363 | } | |
364 | } | |
365 | ||
366 | /* Look for overlap with succeeding segments. | |
367 | * If we can merge fragments, do it. | |
368 | */ | |
369 | while (next && FRAG6_CB(next)->offset < end) { | |
370 | int i = end - FRAG6_CB(next)->offset; /* overlap is 'i' bytes */ | |
371 | ||
372 | if (i < next->len) { | |
373 | /* Eat head of the next overlapped fragment | |
374 | * and leave the loop. The next ones cannot overlap. | |
375 | */ | |
376 | if (!pskb_pull(next, i)) | |
377 | goto err; | |
378 | FRAG6_CB(next)->offset += i; /* next fragment */ | |
5ab11c98 | 379 | fq->q.meat -= i; |
1da177e4 LT |
380 | if (next->ip_summed != CHECKSUM_UNNECESSARY) |
381 | next->ip_summed = CHECKSUM_NONE; | |
382 | break; | |
383 | } else { | |
384 | struct sk_buff *free_it = next; | |
385 | ||
386 | /* Old fragment is completely overridden with | |
387 | * new one drop it. | |
388 | */ | |
389 | next = next->next; | |
390 | ||
391 | if (prev) | |
392 | prev->next = next; | |
393 | else | |
5ab11c98 | 394 | fq->q.fragments = next; |
1da177e4 | 395 | |
5ab11c98 | 396 | fq->q.meat -= free_it->len; |
6ddc0822 | 397 | frag_kfree_skb(fq->q.net, free_it, NULL); |
1da177e4 LT |
398 | } |
399 | } | |
400 | ||
401 | FRAG6_CB(skb)->offset = offset; | |
402 | ||
403 | /* Insert this fragment in the chain of fragments. */ | |
404 | skb->next = next; | |
405 | if (prev) | |
406 | prev->next = skb; | |
407 | else | |
5ab11c98 | 408 | fq->q.fragments = skb; |
1da177e4 | 409 | |
f61944ef HX |
410 | dev = skb->dev; |
411 | if (dev) { | |
412 | fq->iif = dev->ifindex; | |
413 | skb->dev = NULL; | |
414 | } | |
5ab11c98 PE |
415 | fq->q.stamp = skb->tstamp; |
416 | fq->q.meat += skb->len; | |
6ddc0822 | 417 | atomic_add(skb->truesize, &fq->q.net->mem); |
1da177e4 LT |
418 | |
419 | /* The first fragment. | |
420 | * nhoffset is obtained from the first fragment, of course. | |
421 | */ | |
422 | if (offset == 0) { | |
423 | fq->nhoffset = nhoff; | |
bc578a54 | 424 | fq->q.last_in |= INET_FRAG_FIRST_IN; |
1da177e4 | 425 | } |
f61944ef | 426 | |
bc578a54 JP |
427 | if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) && |
428 | fq->q.meat == fq->q.len) | |
f61944ef HX |
429 | return ip6_frag_reasm(fq, prev, dev); |
430 | ||
7eb95156 | 431 | write_lock(&ip6_frags.lock); |
3140c25c | 432 | list_move_tail(&fq->q.lru_list, &fq->q.net->lru_list); |
7eb95156 | 433 | write_unlock(&ip6_frags.lock); |
f61944ef | 434 | return -1; |
1da177e4 LT |
435 | |
436 | err: | |
adf30907 | 437 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
3bd653c8 | 438 | IPSTATS_MIB_REASMFAILS); |
1da177e4 | 439 | kfree_skb(skb); |
f61944ef | 440 | return -1; |
1da177e4 LT |
441 | } |
442 | ||
443 | /* | |
444 | * Check if this packet is complete. | |
445 | * Returns NULL on failure by any reason, and pointer | |
446 | * to current nexthdr field in reassembled frame. | |
447 | * | |
448 | * It is called with locked fq, and caller must check that | |
449 | * queue is eligible for reassembly i.e. it is not COMPLETE, | |
450 | * the last and the first frames arrived and all the bits are here. | |
451 | */ | |
f61944ef | 452 | static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev, |
1da177e4 LT |
453 | struct net_device *dev) |
454 | { | |
2bad35b7 | 455 | struct net *net = container_of(fq->q.net, struct net, ipv6.frags); |
5ab11c98 | 456 | struct sk_buff *fp, *head = fq->q.fragments; |
1da177e4 LT |
457 | int payload_len; |
458 | unsigned int nhoff; | |
459 | ||
460 | fq_kill(fq); | |
461 | ||
f61944ef HX |
462 | /* Make the one we just received the head. */ |
463 | if (prev) { | |
464 | head = prev->next; | |
465 | fp = skb_clone(head, GFP_ATOMIC); | |
466 | ||
467 | if (!fp) | |
468 | goto out_oom; | |
469 | ||
470 | fp->next = head->next; | |
471 | prev->next = fp; | |
472 | ||
5ab11c98 PE |
473 | skb_morph(head, fq->q.fragments); |
474 | head->next = fq->q.fragments->next; | |
f61944ef | 475 | |
5ab11c98 PE |
476 | kfree_skb(fq->q.fragments); |
477 | fq->q.fragments = head; | |
f61944ef HX |
478 | } |
479 | ||
547b792c IJ |
480 | WARN_ON(head == NULL); |
481 | WARN_ON(FRAG6_CB(head)->offset != 0); | |
1da177e4 LT |
482 | |
483 | /* Unfragmented part is taken from the first segment. */ | |
d56f90a7 | 484 | payload_len = ((head->data - skb_network_header(head)) - |
5ab11c98 | 485 | sizeof(struct ipv6hdr) + fq->q.len - |
d56f90a7 | 486 | sizeof(struct frag_hdr)); |
1da177e4 LT |
487 | if (payload_len > IPV6_MAXPLEN) |
488 | goto out_oversize; | |
489 | ||
490 | /* Head of list must not be cloned. */ | |
491 | if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC)) | |
492 | goto out_oom; | |
493 | ||
494 | /* If the first fragment is fragmented itself, we split | |
495 | * it to two chunks: the first with data and paged part | |
496 | * and the second, holding only fragments. */ | |
4d9092bb | 497 | if (skb_has_frags(head)) { |
1da177e4 LT |
498 | struct sk_buff *clone; |
499 | int i, plen = 0; | |
500 | ||
501 | if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL) | |
502 | goto out_oom; | |
503 | clone->next = head->next; | |
504 | head->next = clone; | |
505 | skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list; | |
4d9092bb | 506 | skb_frag_list_init(head); |
1da177e4 LT |
507 | for (i=0; i<skb_shinfo(head)->nr_frags; i++) |
508 | plen += skb_shinfo(head)->frags[i].size; | |
509 | clone->len = clone->data_len = head->data_len - plen; | |
510 | head->data_len -= clone->len; | |
511 | head->len -= clone->len; | |
512 | clone->csum = 0; | |
513 | clone->ip_summed = head->ip_summed; | |
6ddc0822 | 514 | atomic_add(clone->truesize, &fq->q.net->mem); |
1da177e4 LT |
515 | } |
516 | ||
517 | /* We have to remove fragment header from datagram and to relocate | |
518 | * header in order to calculate ICV correctly. */ | |
519 | nhoff = fq->nhoffset; | |
b0e380b1 | 520 | skb_network_header(head)[nhoff] = skb_transport_header(head)[0]; |
1ab1457c | 521 | memmove(head->head + sizeof(struct frag_hdr), head->head, |
1da177e4 | 522 | (head->data - head->head) - sizeof(struct frag_hdr)); |
b0e380b1 ACM |
523 | head->mac_header += sizeof(struct frag_hdr); |
524 | head->network_header += sizeof(struct frag_hdr); | |
1da177e4 LT |
525 | |
526 | skb_shinfo(head)->frag_list = head->next; | |
badff6d0 | 527 | skb_reset_transport_header(head); |
d56f90a7 | 528 | skb_push(head, head->data - skb_network_header(head)); |
6ddc0822 | 529 | atomic_sub(head->truesize, &fq->q.net->mem); |
1da177e4 LT |
530 | |
531 | for (fp=head->next; fp; fp = fp->next) { | |
532 | head->data_len += fp->len; | |
533 | head->len += fp->len; | |
534 | if (head->ip_summed != fp->ip_summed) | |
535 | head->ip_summed = CHECKSUM_NONE; | |
84fa7933 | 536 | else if (head->ip_summed == CHECKSUM_COMPLETE) |
1da177e4 LT |
537 | head->csum = csum_add(head->csum, fp->csum); |
538 | head->truesize += fp->truesize; | |
6ddc0822 | 539 | atomic_sub(fp->truesize, &fq->q.net->mem); |
1da177e4 LT |
540 | } |
541 | ||
542 | head->next = NULL; | |
543 | head->dev = dev; | |
5ab11c98 | 544 | head->tstamp = fq->q.stamp; |
0660e03f | 545 | ipv6_hdr(head)->payload_len = htons(payload_len); |
951dbc8a | 546 | IP6CB(head)->nhoff = nhoff; |
1da177e4 | 547 | |
1da177e4 | 548 | /* Yes, and fold redundant checksum back. 8) */ |
84fa7933 | 549 | if (head->ip_summed == CHECKSUM_COMPLETE) |
d56f90a7 | 550 | head->csum = csum_partial(skb_network_header(head), |
cfe1fc77 | 551 | skb_network_header_len(head), |
d56f90a7 | 552 | head->csum); |
1da177e4 | 553 | |
a11d206d | 554 | rcu_read_lock(); |
2bad35b7 | 555 | IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS); |
a11d206d | 556 | rcu_read_unlock(); |
5ab11c98 | 557 | fq->q.fragments = NULL; |
1da177e4 LT |
558 | return 1; |
559 | ||
560 | out_oversize: | |
561 | if (net_ratelimit()) | |
562 | printk(KERN_DEBUG "ip6_frag_reasm: payload len = %d\n", payload_len); | |
563 | goto out_fail; | |
564 | out_oom: | |
565 | if (net_ratelimit()) | |
566 | printk(KERN_DEBUG "ip6_frag_reasm: no memory for reassembly\n"); | |
567 | out_fail: | |
a11d206d | 568 | rcu_read_lock(); |
2bad35b7 | 569 | IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS); |
a11d206d | 570 | rcu_read_unlock(); |
1da177e4 LT |
571 | return -1; |
572 | } | |
573 | ||
e5bbef20 | 574 | static int ipv6_frag_rcv(struct sk_buff *skb) |
1da177e4 | 575 | { |
1da177e4 LT |
576 | struct frag_hdr *fhdr; |
577 | struct frag_queue *fq; | |
0660e03f | 578 | struct ipv6hdr *hdr = ipv6_hdr(skb); |
adf30907 | 579 | struct net *net = dev_net(skb_dst(skb)->dev); |
1da177e4 | 580 | |
adf30907 | 581 | IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMREQDS); |
1da177e4 LT |
582 | |
583 | /* Jumbo payload inhibits frag. header */ | |
98b3377c DL |
584 | if (hdr->payload_len==0) |
585 | goto fail_hdr; | |
586 | ||
ea2ae17d | 587 | if (!pskb_may_pull(skb, (skb_transport_offset(skb) + |
98b3377c DL |
588 | sizeof(struct frag_hdr)))) |
589 | goto fail_hdr; | |
1da177e4 | 590 | |
0660e03f | 591 | hdr = ipv6_hdr(skb); |
9c70220b | 592 | fhdr = (struct frag_hdr *)skb_transport_header(skb); |
1da177e4 LT |
593 | |
594 | if (!(fhdr->frag_off & htons(0xFFF9))) { | |
595 | /* It is not a fragmented frame */ | |
b0e380b1 | 596 | skb->transport_header += sizeof(struct frag_hdr); |
483a47d2 | 597 | IP6_INC_STATS_BH(net, |
adf30907 | 598 | ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMOKS); |
1da177e4 | 599 | |
d56f90a7 | 600 | IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb); |
1da177e4 LT |
601 | return 1; |
602 | } | |
603 | ||
e31e0bdc | 604 | if (atomic_read(&net->ipv6.frags.mem) > net->ipv6.frags.high_thresh) |
adf30907 | 605 | ip6_evictor(net, ip6_dst_idev(skb_dst(skb))); |
1da177e4 | 606 | |
ac18e750 | 607 | if ((fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr, |
adf30907 | 608 | ip6_dst_idev(skb_dst(skb)))) != NULL) { |
f61944ef | 609 | int ret; |
1da177e4 | 610 | |
5ab11c98 | 611 | spin_lock(&fq->q.lock); |
1da177e4 | 612 | |
f61944ef | 613 | ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff); |
1da177e4 | 614 | |
5ab11c98 | 615 | spin_unlock(&fq->q.lock); |
4b6cb5d8 | 616 | fq_put(fq); |
1da177e4 LT |
617 | return ret; |
618 | } | |
619 | ||
adf30907 | 620 | IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMFAILS); |
1da177e4 LT |
621 | kfree_skb(skb); |
622 | return -1; | |
98b3377c DL |
623 | |
624 | fail_hdr: | |
adf30907 | 625 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS); |
98b3377c DL |
626 | icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb)); |
627 | return -1; | |
1da177e4 LT |
628 | } |
629 | ||
41135cc8 | 630 | static const struct inet6_protocol frag_protocol = |
1da177e4 LT |
631 | { |
632 | .handler = ipv6_frag_rcv, | |
633 | .flags = INET6_PROTO_NOPOLICY, | |
634 | }; | |
635 | ||
8d8354d2 | 636 | #ifdef CONFIG_SYSCTL |
0a64b4b8 | 637 | static struct ctl_table ip6_frags_ns_ctl_table[] = { |
8d8354d2 PE |
638 | { |
639 | .ctl_name = NET_IPV6_IP6FRAG_HIGH_THRESH, | |
640 | .procname = "ip6frag_high_thresh", | |
e31e0bdc | 641 | .data = &init_net.ipv6.frags.high_thresh, |
8d8354d2 PE |
642 | .maxlen = sizeof(int), |
643 | .mode = 0644, | |
6d9f239a | 644 | .proc_handler = proc_dointvec |
8d8354d2 PE |
645 | }, |
646 | { | |
647 | .ctl_name = NET_IPV6_IP6FRAG_LOW_THRESH, | |
648 | .procname = "ip6frag_low_thresh", | |
e31e0bdc | 649 | .data = &init_net.ipv6.frags.low_thresh, |
8d8354d2 PE |
650 | .maxlen = sizeof(int), |
651 | .mode = 0644, | |
6d9f239a | 652 | .proc_handler = proc_dointvec |
8d8354d2 PE |
653 | }, |
654 | { | |
655 | .ctl_name = NET_IPV6_IP6FRAG_TIME, | |
656 | .procname = "ip6frag_time", | |
b2fd5321 | 657 | .data = &init_net.ipv6.frags.timeout, |
8d8354d2 PE |
658 | .maxlen = sizeof(int), |
659 | .mode = 0644, | |
6d9f239a AD |
660 | .proc_handler = proc_dointvec_jiffies, |
661 | .strategy = sysctl_jiffies, | |
8d8354d2 | 662 | }, |
7d291ebb PE |
663 | { } |
664 | }; | |
665 | ||
666 | static struct ctl_table ip6_frags_ctl_table[] = { | |
8d8354d2 PE |
667 | { |
668 | .ctl_name = NET_IPV6_IP6FRAG_SECRET_INTERVAL, | |
669 | .procname = "ip6frag_secret_interval", | |
3b4bc4a2 | 670 | .data = &ip6_frags.secret_interval, |
8d8354d2 PE |
671 | .maxlen = sizeof(int), |
672 | .mode = 0644, | |
6d9f239a AD |
673 | .proc_handler = proc_dointvec_jiffies, |
674 | .strategy = sysctl_jiffies | |
8d8354d2 PE |
675 | }, |
676 | { } | |
677 | }; | |
678 | ||
0a64b4b8 | 679 | static int ip6_frags_ns_sysctl_register(struct net *net) |
8d8354d2 | 680 | { |
e4a2d5c2 | 681 | struct ctl_table *table; |
8d8354d2 PE |
682 | struct ctl_table_header *hdr; |
683 | ||
0a64b4b8 | 684 | table = ip6_frags_ns_ctl_table; |
e4a2d5c2 | 685 | if (net != &init_net) { |
0a64b4b8 | 686 | table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL); |
e4a2d5c2 PE |
687 | if (table == NULL) |
688 | goto err_alloc; | |
689 | ||
e31e0bdc PE |
690 | table[0].data = &net->ipv6.frags.high_thresh; |
691 | table[1].data = &net->ipv6.frags.low_thresh; | |
b2fd5321 | 692 | table[2].data = &net->ipv6.frags.timeout; |
e4a2d5c2 PE |
693 | } |
694 | ||
695 | hdr = register_net_sysctl_table(net, net_ipv6_ctl_path, table); | |
696 | if (hdr == NULL) | |
697 | goto err_reg; | |
698 | ||
699 | net->ipv6.sysctl.frags_hdr = hdr; | |
700 | return 0; | |
701 | ||
702 | err_reg: | |
703 | if (net != &init_net) | |
704 | kfree(table); | |
705 | err_alloc: | |
706 | return -ENOMEM; | |
707 | } | |
708 | ||
0a64b4b8 | 709 | static void ip6_frags_ns_sysctl_unregister(struct net *net) |
e4a2d5c2 PE |
710 | { |
711 | struct ctl_table *table; | |
712 | ||
713 | table = net->ipv6.sysctl.frags_hdr->ctl_table_arg; | |
714 | unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr); | |
715 | kfree(table); | |
8d8354d2 | 716 | } |
7d291ebb PE |
717 | |
718 | static struct ctl_table_header *ip6_ctl_header; | |
719 | ||
720 | static int ip6_frags_sysctl_register(void) | |
721 | { | |
722 | ip6_ctl_header = register_net_sysctl_rotable(net_ipv6_ctl_path, | |
723 | ip6_frags_ctl_table); | |
724 | return ip6_ctl_header == NULL ? -ENOMEM : 0; | |
725 | } | |
726 | ||
727 | static void ip6_frags_sysctl_unregister(void) | |
728 | { | |
729 | unregister_net_sysctl_table(ip6_ctl_header); | |
730 | } | |
8d8354d2 | 731 | #else |
0a64b4b8 | 732 | static inline int ip6_frags_ns_sysctl_register(struct net *net) |
e71e0349 | 733 | { |
8d8354d2 PE |
734 | return 0; |
735 | } | |
e4a2d5c2 | 736 | |
0a64b4b8 | 737 | static inline void ip6_frags_ns_sysctl_unregister(struct net *net) |
e4a2d5c2 PE |
738 | { |
739 | } | |
7d291ebb PE |
740 | |
741 | static inline int ip6_frags_sysctl_register(void) | |
742 | { | |
743 | return 0; | |
744 | } | |
745 | ||
746 | static inline void ip6_frags_sysctl_unregister(void) | |
747 | { | |
748 | } | |
8d8354d2 | 749 | #endif |
7d460db9 | 750 | |
8d8354d2 PE |
751 | static int ipv6_frags_init_net(struct net *net) |
752 | { | |
e31e0bdc PE |
753 | net->ipv6.frags.high_thresh = 256 * 1024; |
754 | net->ipv6.frags.low_thresh = 192 * 1024; | |
b2fd5321 | 755 | net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT; |
8d8354d2 | 756 | |
e5a2bb84 PE |
757 | inet_frags_init_net(&net->ipv6.frags); |
758 | ||
0a64b4b8 | 759 | return ip6_frags_ns_sysctl_register(net); |
e71e0349 DL |
760 | } |
761 | ||
81566e83 PE |
762 | static void ipv6_frags_exit_net(struct net *net) |
763 | { | |
0a64b4b8 | 764 | ip6_frags_ns_sysctl_unregister(net); |
81566e83 PE |
765 | inet_frags_exit_net(&net->ipv6.frags, &ip6_frags); |
766 | } | |
767 | ||
768 | static struct pernet_operations ip6_frags_ops = { | |
769 | .init = ipv6_frags_init_net, | |
770 | .exit = ipv6_frags_exit_net, | |
771 | }; | |
772 | ||
853cbbaa | 773 | int __init ipv6_frag_init(void) |
1da177e4 | 774 | { |
853cbbaa | 775 | int ret; |
1da177e4 | 776 | |
853cbbaa DL |
777 | ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT); |
778 | if (ret) | |
779 | goto out; | |
e71e0349 | 780 | |
7d291ebb PE |
781 | ret = ip6_frags_sysctl_register(); |
782 | if (ret) | |
783 | goto err_sysctl; | |
784 | ||
0002c630 PE |
785 | ret = register_pernet_subsys(&ip6_frags_ops); |
786 | if (ret) | |
787 | goto err_pernet; | |
8d8354d2 | 788 | |
321a3a99 | 789 | ip6_frags.hashfn = ip6_hashfn; |
c6fda282 | 790 | ip6_frags.constructor = ip6_frag_init; |
c9547709 | 791 | ip6_frags.destructor = NULL; |
1e4b8287 PE |
792 | ip6_frags.skb_free = NULL; |
793 | ip6_frags.qsize = sizeof(struct frag_queue); | |
abd6523d | 794 | ip6_frags.match = ip6_frag_match; |
e521db9d | 795 | ip6_frags.frag_expire = ip6_frag_expire; |
3b4bc4a2 | 796 | ip6_frags.secret_interval = 10 * 60 * HZ; |
7eb95156 | 797 | inet_frags_init(&ip6_frags); |
853cbbaa DL |
798 | out: |
799 | return ret; | |
0002c630 PE |
800 | |
801 | err_pernet: | |
7d291ebb PE |
802 | ip6_frags_sysctl_unregister(); |
803 | err_sysctl: | |
0002c630 PE |
804 | inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT); |
805 | goto out; | |
853cbbaa DL |
806 | } |
807 | ||
808 | void ipv6_frag_exit(void) | |
809 | { | |
810 | inet_frags_fini(&ip6_frags); | |
7d291ebb | 811 | ip6_frags_sysctl_unregister(); |
81566e83 | 812 | unregister_pernet_subsys(&ip6_frags_ops); |
853cbbaa | 813 | inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT); |
1da177e4 | 814 | } |