]> bbs.cooldavid.org Git - net-next-2.6.git/blame - net/ipv6/reassembly.c
IPv6: reassembly: replace magic number with macro definitions
[net-next-2.6.git] / net / ipv6 / reassembly.c
CommitLineData
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
57struct 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
70struct frag_queue
71{
5ab11c98 72 struct inet_frag_queue q;
1da177e4 73
e69a4adc 74 __be32 id; /* fragment id */
0b5ccb2e 75 u32 user;
1da177e4
LT
76 struct in6_addr saddr;
77 struct in6_addr daddr;
78
1da177e4 79 int iif;
1da177e4 80 unsigned int csum;
1da177e4 81 __u16 nhoffset;
1da177e4
LT
82};
83
7eb95156 84static struct inet_frags ip6_frags;
1da177e4 85
e5a2bb84 86int ip6_frag_nqueues(struct net *net)
7eb95156 87{
e5a2bb84 88 return net->ipv6.frags.nqueues;
7eb95156 89}
1da177e4 90
6ddc0822 91int ip6_frag_mem(struct net *net)
7eb95156 92{
6ddc0822 93 return atomic_read(&net->ipv6.frags.mem);
7eb95156 94}
1da177e4 95
f61944ef
HX
96static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
97 struct net_device *dev);
98
f6596f9d
ZB
99/*
100 * callers should be careful not to use the hash value outside the ipfrag_lock
101 * as doing so could race with ipfrag_hash_rnd being recalculated.
102 */
93c8b90f
IJ
103unsigned int inet6_hash_frag(__be32 id, const struct in6_addr *saddr,
104 const struct in6_addr *daddr, u32 rnd)
1da177e4
LT
105{
106 u32 a, b, c;
107
e69a4adc
AV
108 a = (__force u32)saddr->s6_addr32[0];
109 b = (__force u32)saddr->s6_addr32[1];
110 c = (__force u32)saddr->s6_addr32[2];
1da177e4
LT
111
112 a += JHASH_GOLDEN_RATIO;
113 b += JHASH_GOLDEN_RATIO;
93c8b90f 114 c += rnd;
1da177e4
LT
115 __jhash_mix(a, b, c);
116
e69a4adc
AV
117 a += (__force u32)saddr->s6_addr32[3];
118 b += (__force u32)daddr->s6_addr32[0];
119 c += (__force u32)daddr->s6_addr32[1];
1da177e4
LT
120 __jhash_mix(a, b, c);
121
e69a4adc
AV
122 a += (__force u32)daddr->s6_addr32[2];
123 b += (__force u32)daddr->s6_addr32[3];
124 c += (__force u32)id;
1da177e4
LT
125 __jhash_mix(a, b, c);
126
7eb95156 127 return c & (INETFRAGS_HASHSZ - 1);
1da177e4 128}
93c8b90f 129EXPORT_SYMBOL_GPL(inet6_hash_frag);
1da177e4 130
321a3a99 131static unsigned int ip6_hashfn(struct inet_frag_queue *q)
1da177e4 132{
321a3a99 133 struct frag_queue *fq;
1da177e4 134
321a3a99 135 fq = container_of(q, struct frag_queue, q);
93c8b90f 136 return inet6_hash_frag(fq->id, &fq->saddr, &fq->daddr, ip6_frags.rnd);
1da177e4
LT
137}
138
abd6523d
PE
139int ip6_frag_match(struct inet_frag_queue *q, void *a)
140{
141 struct frag_queue *fq;
142 struct ip6_create_arg *arg = a;
143
144 fq = container_of(q, struct frag_queue, q);
0b5ccb2e 145 return (fq->id == arg->id && fq->user == arg->user &&
abd6523d
PE
146 ipv6_addr_equal(&fq->saddr, arg->src) &&
147 ipv6_addr_equal(&fq->daddr, arg->dst));
148}
149EXPORT_SYMBOL(ip6_frag_match);
150
1da177e4 151/* Memory Tracking Functions. */
6ddc0822
PE
152static inline void frag_kfree_skb(struct netns_frags *nf,
153 struct sk_buff *skb, int *work)
1da177e4
LT
154{
155 if (work)
156 *work -= skb->truesize;
6ddc0822 157 atomic_sub(skb->truesize, &nf->mem);
1da177e4
LT
158 kfree_skb(skb);
159}
160
c6fda282 161void ip6_frag_init(struct inet_frag_queue *q, void *a)
1da177e4 162{
c6fda282
PE
163 struct frag_queue *fq = container_of(q, struct frag_queue, q);
164 struct ip6_create_arg *arg = a;
165
166 fq->id = arg->id;
0b5ccb2e 167 fq->user = arg->user;
c6fda282
PE
168 ipv6_addr_copy(&fq->saddr, arg->src);
169 ipv6_addr_copy(&fq->daddr, arg->dst);
1da177e4 170}
c6fda282 171EXPORT_SYMBOL(ip6_frag_init);
1da177e4 172
1da177e4
LT
173/* Destruction primitives. */
174
4b6cb5d8 175static __inline__ void fq_put(struct frag_queue *fq)
1da177e4 176{
762cc408 177 inet_frag_put(&fq->q, &ip6_frags);
1da177e4
LT
178}
179
180/* Kill fq entry. It is not destroyed immediately,
181 * because caller (and someone more) holds reference count.
182 */
183static __inline__ void fq_kill(struct frag_queue *fq)
184{
277e650d 185 inet_frag_kill(&fq->q, &ip6_frags);
1da177e4
LT
186}
187
6ddc0822 188static void ip6_evictor(struct net *net, struct inet6_dev *idev)
1da177e4 189{
8e7999c4
PE
190 int evicted;
191
6ddc0822 192 evicted = inet_frag_evictor(&net->ipv6.frags, &ip6_frags);
8e7999c4 193 if (evicted)
821d5777 194 IP6_ADD_STATS_BH(net, idev, IPSTATS_MIB_REASMFAILS, evicted);
1da177e4
LT
195}
196
197static void ip6_frag_expire(unsigned long data)
198{
e521db9d 199 struct frag_queue *fq;
a11d206d 200 struct net_device *dev = NULL;
4ac2ccd0 201 struct net *net;
1da177e4 202
e521db9d
PE
203 fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q);
204
5ab11c98 205 spin_lock(&fq->q.lock);
1da177e4 206
bc578a54 207 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
208 goto out;
209
210 fq_kill(fq);
211
4ac2ccd0 212 net = container_of(fq->q.net, struct net, ipv6.frags);
69df9d59
ED
213 rcu_read_lock();
214 dev = dev_get_by_index_rcu(net, fq->iif);
a11d206d 215 if (!dev)
69df9d59 216 goto out_rcu_unlock;
a11d206d 217
483a47d2
DL
218 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMTIMEOUT);
219 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
1da177e4 220
78c784c4 221 /* Don't send error if the first segment did not arrive. */
bc578a54 222 if (!(fq->q.last_in & INET_FRAG_FIRST_IN) || !fq->q.fragments)
69df9d59 223 goto out_rcu_unlock;
78c784c4 224
78c784c4
IO
225 /*
226 But use as source device on which LAST ARRIVED
227 segment was received. And do not use fq->dev
228 pointer directly, device might already disappeared.
229 */
5ab11c98
PE
230 fq->q.fragments->dev = dev;
231 icmpv6_send(fq->q.fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0, dev);
69df9d59
ED
232out_rcu_unlock:
233 rcu_read_unlock();
1da177e4 234out:
5ab11c98 235 spin_unlock(&fq->q.lock);
4b6cb5d8 236 fq_put(fq);
1da177e4
LT
237}
238
abd6523d 239static __inline__ struct frag_queue *
ac18e750 240fq_find(struct net *net, __be32 id, struct in6_addr *src, struct in6_addr *dst,
abd6523d 241 struct inet6_dev *idev)
1da177e4 242{
c6fda282
PE
243 struct inet_frag_queue *q;
244 struct ip6_create_arg arg;
abd6523d 245 unsigned int hash;
1da177e4 246
c6fda282 247 arg.id = id;
0b5ccb2e 248 arg.user = IP6_DEFRAG_LOCAL_DELIVER;
c6fda282
PE
249 arg.src = src;
250 arg.dst = dst;
9a375803
PE
251
252 read_lock(&ip6_frags.lock);
93c8b90f 253 hash = inet6_hash_frag(id, src, dst, ip6_frags.rnd);
1da177e4 254
ac18e750 255 q = inet_frag_find(&net->ipv6.frags, &ip6_frags, &arg, hash);
c6fda282
PE
256 if (q == NULL)
257 goto oom;
1da177e4 258
c6fda282 259 return container_of(q, struct frag_queue, q);
1da177e4
LT
260
261oom:
483a47d2 262 IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_REASMFAILS);
1da177e4
LT
263 return NULL;
264}
265
f61944ef 266static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
1da177e4
LT
267 struct frag_hdr *fhdr, int nhoff)
268{
269 struct sk_buff *prev, *next;
f61944ef 270 struct net_device *dev;
1da177e4 271 int offset, end;
adf30907 272 struct net *net = dev_net(skb_dst(skb)->dev);
1da177e4 273
bc578a54 274 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
275 goto err;
276
277 offset = ntohs(fhdr->frag_off) & ~0x7;
0660e03f
ACM
278 end = offset + (ntohs(ipv6_hdr(skb)->payload_len) -
279 ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1)));
1da177e4
LT
280
281 if ((unsigned int)end > IPV6_MAXPLEN) {
adf30907 282 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
a11d206d 283 IPSTATS_MIB_INHDRERRORS);
d56f90a7
ACM
284 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
285 ((u8 *)&fhdr->frag_off -
286 skb_network_header(skb)));
f61944ef 287 return -1;
1da177e4
LT
288 }
289
d56f90a7
ACM
290 if (skb->ip_summed == CHECKSUM_COMPLETE) {
291 const unsigned char *nh = skb_network_header(skb);
1ab1457c 292 skb->csum = csum_sub(skb->csum,
d56f90a7
ACM
293 csum_partial(nh, (u8 *)(fhdr + 1) - nh,
294 0));
295 }
1da177e4
LT
296
297 /* Is this the final fragment? */
298 if (!(fhdr->frag_off & htons(IP6_MF))) {
299 /* If we already have some bits beyond end
300 * or have different end, the segment is corrupted.
301 */
5ab11c98 302 if (end < fq->q.len ||
bc578a54 303 ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len))
1da177e4 304 goto err;
bc578a54 305 fq->q.last_in |= INET_FRAG_LAST_IN;
5ab11c98 306 fq->q.len = end;
1da177e4
LT
307 } else {
308 /* Check if the fragment is rounded to 8 bytes.
309 * Required by the RFC.
310 */
311 if (end & 0x7) {
312 /* RFC2460 says always send parameter problem in
313 * this case. -DaveM
314 */
adf30907 315 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
a11d206d 316 IPSTATS_MIB_INHDRERRORS);
1ab1457c 317 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
1da177e4 318 offsetof(struct ipv6hdr, payload_len));
f61944ef 319 return -1;
1da177e4 320 }
5ab11c98 321 if (end > fq->q.len) {
1da177e4 322 /* Some bits beyond end -> corruption. */
bc578a54 323 if (fq->q.last_in & INET_FRAG_LAST_IN)
1da177e4 324 goto err;
5ab11c98 325 fq->q.len = end;
1da177e4
LT
326 }
327 }
328
329 if (end == offset)
330 goto err;
331
332 /* Point into the IP datagram 'data' part. */
333 if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
334 goto err;
1ab1457c 335
42ca89c1
SH
336 if (pskb_trim_rcsum(skb, end - offset))
337 goto err;
1da177e4
LT
338
339 /* Find out which fragments are in front and at the back of us
340 * in the chain of fragments so far. We must know where to put
341 * this fragment, right?
342 */
343 prev = NULL;
5ab11c98 344 for(next = fq->q.fragments; next != NULL; next = next->next) {
1da177e4
LT
345 if (FRAG6_CB(next)->offset >= offset)
346 break; /* bingo! */
347 prev = next;
348 }
349
350 /* We found where to put this one. Check for overlap with
351 * preceding fragment, and, if needed, align things so that
352 * any overlaps are eliminated.
353 */
354 if (prev) {
355 int i = (FRAG6_CB(prev)->offset + prev->len) - offset;
356
357 if (i > 0) {
358 offset += i;
359 if (end <= offset)
360 goto err;
361 if (!pskb_pull(skb, i))
362 goto err;
363 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
364 skb->ip_summed = CHECKSUM_NONE;
365 }
366 }
367
368 /* Look for overlap with succeeding segments.
369 * If we can merge fragments, do it.
370 */
371 while (next && FRAG6_CB(next)->offset < end) {
372 int i = end - FRAG6_CB(next)->offset; /* overlap is 'i' bytes */
373
374 if (i < next->len) {
375 /* Eat head of the next overlapped fragment
376 * and leave the loop. The next ones cannot overlap.
377 */
378 if (!pskb_pull(next, i))
379 goto err;
380 FRAG6_CB(next)->offset += i; /* next fragment */
5ab11c98 381 fq->q.meat -= i;
1da177e4
LT
382 if (next->ip_summed != CHECKSUM_UNNECESSARY)
383 next->ip_summed = CHECKSUM_NONE;
384 break;
385 } else {
386 struct sk_buff *free_it = next;
387
388 /* Old fragment is completely overridden with
389 * new one drop it.
390 */
391 next = next->next;
392
393 if (prev)
394 prev->next = next;
395 else
5ab11c98 396 fq->q.fragments = next;
1da177e4 397
5ab11c98 398 fq->q.meat -= free_it->len;
6ddc0822 399 frag_kfree_skb(fq->q.net, free_it, NULL);
1da177e4
LT
400 }
401 }
402
403 FRAG6_CB(skb)->offset = offset;
404
405 /* Insert this fragment in the chain of fragments. */
406 skb->next = next;
407 if (prev)
408 prev->next = skb;
409 else
5ab11c98 410 fq->q.fragments = skb;
1da177e4 411
f61944ef
HX
412 dev = skb->dev;
413 if (dev) {
414 fq->iif = dev->ifindex;
415 skb->dev = NULL;
416 }
5ab11c98
PE
417 fq->q.stamp = skb->tstamp;
418 fq->q.meat += skb->len;
6ddc0822 419 atomic_add(skb->truesize, &fq->q.net->mem);
1da177e4
LT
420
421 /* The first fragment.
422 * nhoffset is obtained from the first fragment, of course.
423 */
424 if (offset == 0) {
425 fq->nhoffset = nhoff;
bc578a54 426 fq->q.last_in |= INET_FRAG_FIRST_IN;
1da177e4 427 }
f61944ef 428
bc578a54
JP
429 if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
430 fq->q.meat == fq->q.len)
f61944ef
HX
431 return ip6_frag_reasm(fq, prev, dev);
432
7eb95156 433 write_lock(&ip6_frags.lock);
3140c25c 434 list_move_tail(&fq->q.lru_list, &fq->q.net->lru_list);
7eb95156 435 write_unlock(&ip6_frags.lock);
f61944ef 436 return -1;
1da177e4
LT
437
438err:
adf30907 439 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
3bd653c8 440 IPSTATS_MIB_REASMFAILS);
1da177e4 441 kfree_skb(skb);
f61944ef 442 return -1;
1da177e4
LT
443}
444
445/*
446 * Check if this packet is complete.
447 * Returns NULL on failure by any reason, and pointer
448 * to current nexthdr field in reassembled frame.
449 *
450 * It is called with locked fq, and caller must check that
451 * queue is eligible for reassembly i.e. it is not COMPLETE,
452 * the last and the first frames arrived and all the bits are here.
453 */
f61944ef 454static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
1da177e4
LT
455 struct net_device *dev)
456{
2bad35b7 457 struct net *net = container_of(fq->q.net, struct net, ipv6.frags);
5ab11c98 458 struct sk_buff *fp, *head = fq->q.fragments;
1da177e4
LT
459 int payload_len;
460 unsigned int nhoff;
461
462 fq_kill(fq);
463
f61944ef
HX
464 /* Make the one we just received the head. */
465 if (prev) {
466 head = prev->next;
467 fp = skb_clone(head, GFP_ATOMIC);
468
469 if (!fp)
470 goto out_oom;
471
472 fp->next = head->next;
473 prev->next = fp;
474
5ab11c98
PE
475 skb_morph(head, fq->q.fragments);
476 head->next = fq->q.fragments->next;
f61944ef 477
5ab11c98
PE
478 kfree_skb(fq->q.fragments);
479 fq->q.fragments = head;
f61944ef
HX
480 }
481
547b792c
IJ
482 WARN_ON(head == NULL);
483 WARN_ON(FRAG6_CB(head)->offset != 0);
1da177e4
LT
484
485 /* Unfragmented part is taken from the first segment. */
d56f90a7 486 payload_len = ((head->data - skb_network_header(head)) -
5ab11c98 487 sizeof(struct ipv6hdr) + fq->q.len -
d56f90a7 488 sizeof(struct frag_hdr));
1da177e4
LT
489 if (payload_len > IPV6_MAXPLEN)
490 goto out_oversize;
491
492 /* Head of list must not be cloned. */
493 if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC))
494 goto out_oom;
495
496 /* If the first fragment is fragmented itself, we split
497 * it to two chunks: the first with data and paged part
498 * and the second, holding only fragments. */
4d9092bb 499 if (skb_has_frags(head)) {
1da177e4
LT
500 struct sk_buff *clone;
501 int i, plen = 0;
502
503 if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
504 goto out_oom;
505 clone->next = head->next;
506 head->next = clone;
507 skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
4d9092bb 508 skb_frag_list_init(head);
1da177e4
LT
509 for (i=0; i<skb_shinfo(head)->nr_frags; i++)
510 plen += skb_shinfo(head)->frags[i].size;
511 clone->len = clone->data_len = head->data_len - plen;
512 head->data_len -= clone->len;
513 head->len -= clone->len;
514 clone->csum = 0;
515 clone->ip_summed = head->ip_summed;
6ddc0822 516 atomic_add(clone->truesize, &fq->q.net->mem);
1da177e4
LT
517 }
518
519 /* We have to remove fragment header from datagram and to relocate
520 * header in order to calculate ICV correctly. */
521 nhoff = fq->nhoffset;
b0e380b1 522 skb_network_header(head)[nhoff] = skb_transport_header(head)[0];
1ab1457c 523 memmove(head->head + sizeof(struct frag_hdr), head->head,
1da177e4 524 (head->data - head->head) - sizeof(struct frag_hdr));
b0e380b1
ACM
525 head->mac_header += sizeof(struct frag_hdr);
526 head->network_header += sizeof(struct frag_hdr);
1da177e4
LT
527
528 skb_shinfo(head)->frag_list = head->next;
badff6d0 529 skb_reset_transport_header(head);
d56f90a7 530 skb_push(head, head->data - skb_network_header(head));
6ddc0822 531 atomic_sub(head->truesize, &fq->q.net->mem);
1da177e4
LT
532
533 for (fp=head->next; fp; fp = fp->next) {
534 head->data_len += fp->len;
535 head->len += fp->len;
536 if (head->ip_summed != fp->ip_summed)
537 head->ip_summed = CHECKSUM_NONE;
84fa7933 538 else if (head->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
539 head->csum = csum_add(head->csum, fp->csum);
540 head->truesize += fp->truesize;
6ddc0822 541 atomic_sub(fp->truesize, &fq->q.net->mem);
1da177e4
LT
542 }
543
544 head->next = NULL;
545 head->dev = dev;
5ab11c98 546 head->tstamp = fq->q.stamp;
0660e03f 547 ipv6_hdr(head)->payload_len = htons(payload_len);
951dbc8a 548 IP6CB(head)->nhoff = nhoff;
1da177e4 549
1da177e4 550 /* Yes, and fold redundant checksum back. 8) */
84fa7933 551 if (head->ip_summed == CHECKSUM_COMPLETE)
d56f90a7 552 head->csum = csum_partial(skb_network_header(head),
cfe1fc77 553 skb_network_header_len(head),
d56f90a7 554 head->csum);
1da177e4 555
a11d206d 556 rcu_read_lock();
2bad35b7 557 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
a11d206d 558 rcu_read_unlock();
5ab11c98 559 fq->q.fragments = NULL;
1da177e4
LT
560 return 1;
561
562out_oversize:
563 if (net_ratelimit())
564 printk(KERN_DEBUG "ip6_frag_reasm: payload len = %d\n", payload_len);
565 goto out_fail;
566out_oom:
567 if (net_ratelimit())
568 printk(KERN_DEBUG "ip6_frag_reasm: no memory for reassembly\n");
569out_fail:
a11d206d 570 rcu_read_lock();
2bad35b7 571 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
a11d206d 572 rcu_read_unlock();
1da177e4
LT
573 return -1;
574}
575
e5bbef20 576static int ipv6_frag_rcv(struct sk_buff *skb)
1da177e4 577{
1da177e4
LT
578 struct frag_hdr *fhdr;
579 struct frag_queue *fq;
0660e03f 580 struct ipv6hdr *hdr = ipv6_hdr(skb);
adf30907 581 struct net *net = dev_net(skb_dst(skb)->dev);
1da177e4 582
adf30907 583 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMREQDS);
1da177e4
LT
584
585 /* Jumbo payload inhibits frag. header */
98b3377c
DL
586 if (hdr->payload_len==0)
587 goto fail_hdr;
588
ea2ae17d 589 if (!pskb_may_pull(skb, (skb_transport_offset(skb) +
98b3377c
DL
590 sizeof(struct frag_hdr))))
591 goto fail_hdr;
1da177e4 592
0660e03f 593 hdr = ipv6_hdr(skb);
9c70220b 594 fhdr = (struct frag_hdr *)skb_transport_header(skb);
1da177e4
LT
595
596 if (!(fhdr->frag_off & htons(0xFFF9))) {
597 /* It is not a fragmented frame */
b0e380b1 598 skb->transport_header += sizeof(struct frag_hdr);
483a47d2 599 IP6_INC_STATS_BH(net,
adf30907 600 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMOKS);
1da177e4 601
d56f90a7 602 IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
1da177e4
LT
603 return 1;
604 }
605
e31e0bdc 606 if (atomic_read(&net->ipv6.frags.mem) > net->ipv6.frags.high_thresh)
adf30907 607 ip6_evictor(net, ip6_dst_idev(skb_dst(skb)));
1da177e4 608
ac18e750 609 if ((fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr,
adf30907 610 ip6_dst_idev(skb_dst(skb)))) != NULL) {
f61944ef 611 int ret;
1da177e4 612
5ab11c98 613 spin_lock(&fq->q.lock);
1da177e4 614
f61944ef 615 ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);
1da177e4 616
5ab11c98 617 spin_unlock(&fq->q.lock);
4b6cb5d8 618 fq_put(fq);
1da177e4
LT
619 return ret;
620 }
621
adf30907 622 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMFAILS);
1da177e4
LT
623 kfree_skb(skb);
624 return -1;
98b3377c
DL
625
626fail_hdr:
adf30907 627 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS);
98b3377c
DL
628 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb));
629 return -1;
1da177e4
LT
630}
631
41135cc8 632static const struct inet6_protocol frag_protocol =
1da177e4
LT
633{
634 .handler = ipv6_frag_rcv,
635 .flags = INET6_PROTO_NOPOLICY,
636};
637
8d8354d2 638#ifdef CONFIG_SYSCTL
0a64b4b8 639static struct ctl_table ip6_frags_ns_ctl_table[] = {
8d8354d2 640 {
8d8354d2 641 .procname = "ip6frag_high_thresh",
e31e0bdc 642 .data = &init_net.ipv6.frags.high_thresh,
8d8354d2
PE
643 .maxlen = sizeof(int),
644 .mode = 0644,
6d9f239a 645 .proc_handler = proc_dointvec
8d8354d2
PE
646 },
647 {
8d8354d2 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 {
8d8354d2 655 .procname = "ip6frag_time",
b2fd5321 656 .data = &init_net.ipv6.frags.timeout,
8d8354d2
PE
657 .maxlen = sizeof(int),
658 .mode = 0644,
6d9f239a 659 .proc_handler = proc_dointvec_jiffies,
8d8354d2 660 },
7d291ebb
PE
661 { }
662};
663
664static struct ctl_table ip6_frags_ctl_table[] = {
8d8354d2 665 {
8d8354d2 666 .procname = "ip6frag_secret_interval",
3b4bc4a2 667 .data = &ip6_frags.secret_interval,
8d8354d2
PE
668 .maxlen = sizeof(int),
669 .mode = 0644,
6d9f239a 670 .proc_handler = proc_dointvec_jiffies,
8d8354d2
PE
671 },
672 { }
673};
674
0a64b4b8 675static int ip6_frags_ns_sysctl_register(struct net *net)
8d8354d2 676{
e4a2d5c2 677 struct ctl_table *table;
8d8354d2
PE
678 struct ctl_table_header *hdr;
679
0a64b4b8 680 table = ip6_frags_ns_ctl_table;
09ad9bc7 681 if (!net_eq(net, &init_net)) {
0a64b4b8 682 table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL);
e4a2d5c2
PE
683 if (table == NULL)
684 goto err_alloc;
685
e31e0bdc
PE
686 table[0].data = &net->ipv6.frags.high_thresh;
687 table[1].data = &net->ipv6.frags.low_thresh;
b2fd5321 688 table[2].data = &net->ipv6.frags.timeout;
e4a2d5c2
PE
689 }
690
691 hdr = register_net_sysctl_table(net, net_ipv6_ctl_path, table);
692 if (hdr == NULL)
693 goto err_reg;
694
695 net->ipv6.sysctl.frags_hdr = hdr;
696 return 0;
697
698err_reg:
09ad9bc7 699 if (!net_eq(net, &init_net))
e4a2d5c2
PE
700 kfree(table);
701err_alloc:
702 return -ENOMEM;
703}
704
0a64b4b8 705static void ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
706{
707 struct ctl_table *table;
708
709 table = net->ipv6.sysctl.frags_hdr->ctl_table_arg;
710 unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr);
3705e11a
YH
711 if (!net_eq(net, &init_net))
712 kfree(table);
8d8354d2 713}
7d291ebb
PE
714
715static struct ctl_table_header *ip6_ctl_header;
716
717static int ip6_frags_sysctl_register(void)
718{
719 ip6_ctl_header = register_net_sysctl_rotable(net_ipv6_ctl_path,
720 ip6_frags_ctl_table);
721 return ip6_ctl_header == NULL ? -ENOMEM : 0;
722}
723
724static void ip6_frags_sysctl_unregister(void)
725{
726 unregister_net_sysctl_table(ip6_ctl_header);
727}
8d8354d2 728#else
0a64b4b8 729static inline int ip6_frags_ns_sysctl_register(struct net *net)
e71e0349 730{
8d8354d2
PE
731 return 0;
732}
e4a2d5c2 733
0a64b4b8 734static inline void ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
735{
736}
7d291ebb
PE
737
738static inline int ip6_frags_sysctl_register(void)
739{
740 return 0;
741}
742
743static inline void ip6_frags_sysctl_unregister(void)
744{
745}
8d8354d2 746#endif
7d460db9 747
8d8354d2
PE
748static int ipv6_frags_init_net(struct net *net)
749{
7c070aa9
SW
750 net->ipv6.frags.high_thresh = IPV6_FRAG_HIGH_THRESH;
751 net->ipv6.frags.low_thresh = IPV6_FRAG_LOW_THRESH;
b2fd5321 752 net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT;
8d8354d2 753
e5a2bb84
PE
754 inet_frags_init_net(&net->ipv6.frags);
755
0a64b4b8 756 return ip6_frags_ns_sysctl_register(net);
e71e0349
DL
757}
758
81566e83
PE
759static void ipv6_frags_exit_net(struct net *net)
760{
0a64b4b8 761 ip6_frags_ns_sysctl_unregister(net);
81566e83
PE
762 inet_frags_exit_net(&net->ipv6.frags, &ip6_frags);
763}
764
765static struct pernet_operations ip6_frags_ops = {
766 .init = ipv6_frags_init_net,
767 .exit = ipv6_frags_exit_net,
768};
769
853cbbaa 770int __init ipv6_frag_init(void)
1da177e4 771{
853cbbaa 772 int ret;
1da177e4 773
853cbbaa
DL
774 ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT);
775 if (ret)
776 goto out;
e71e0349 777
7d291ebb
PE
778 ret = ip6_frags_sysctl_register();
779 if (ret)
780 goto err_sysctl;
781
0002c630
PE
782 ret = register_pernet_subsys(&ip6_frags_ops);
783 if (ret)
784 goto err_pernet;
8d8354d2 785
321a3a99 786 ip6_frags.hashfn = ip6_hashfn;
c6fda282 787 ip6_frags.constructor = ip6_frag_init;
c9547709 788 ip6_frags.destructor = NULL;
1e4b8287
PE
789 ip6_frags.skb_free = NULL;
790 ip6_frags.qsize = sizeof(struct frag_queue);
abd6523d 791 ip6_frags.match = ip6_frag_match;
e521db9d 792 ip6_frags.frag_expire = ip6_frag_expire;
3b4bc4a2 793 ip6_frags.secret_interval = 10 * 60 * HZ;
7eb95156 794 inet_frags_init(&ip6_frags);
853cbbaa
DL
795out:
796 return ret;
0002c630
PE
797
798err_pernet:
7d291ebb
PE
799 ip6_frags_sysctl_unregister();
800err_sysctl:
0002c630
PE
801 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
802 goto out;
853cbbaa
DL
803}
804
805void ipv6_frag_exit(void)
806{
807 inet_frags_fini(&ip6_frags);
7d291ebb 808 ip6_frags_sysctl_unregister();
81566e83 809 unregister_pernet_subsys(&ip6_frags_ops);
853cbbaa 810 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
1da177e4 811}