]> bbs.cooldavid.org Git - net-next-2.6.git/blame - net/ipv6/reassembly.c
ipv6: In fragmentation code, handle error returned from register_pernet_subsys.
[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
LT
7 *
8 * $Id: reassembly.c,v 1.26 2001/03/07 22:00:57 davem Exp $
9 *
10 * Based on: net/ipv4/ip_fragment.c
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
16 */
17
1ab1457c
YH
18/*
19 * Fixes:
1da177e4
LT
20 * Andi Kleen Make it work with multiple hosts.
21 * More RFC compliance.
22 *
23 * Horst von Brand Add missing #include <linux/string.h>
24 * Alexey Kuznetsov SMP races, threading, cleanup.
25 * Patrick McHardy LRU queue of frag heads for evictor.
26 * Mitsuru KANDA @USAGI Register inet6_protocol{}.
27 * David Stevens and
28 * YOSHIFUJI,H. @USAGI Always remove fragment header to
29 * calculate ICV correctly.
30 */
1da177e4
LT
31#include <linux/errno.h>
32#include <linux/types.h>
33#include <linux/string.h>
34#include <linux/socket.h>
35#include <linux/sockios.h>
36#include <linux/jiffies.h>
37#include <linux/net.h>
38#include <linux/list.h>
39#include <linux/netdevice.h>
40#include <linux/in6.h>
41#include <linux/ipv6.h>
42#include <linux/icmpv6.h>
43#include <linux/random.h>
44#include <linux/jhash.h>
f61944ef 45#include <linux/skbuff.h>
1da177e4
LT
46
47#include <net/sock.h>
48#include <net/snmp.h>
49
50#include <net/ipv6.h>
a11d206d 51#include <net/ip6_route.h>
1da177e4
LT
52#include <net/protocol.h>
53#include <net/transp_v6.h>
54#include <net/rawv6.h>
55#include <net/ndisc.h>
56#include <net/addrconf.h>
5ab11c98 57#include <net/inet_frag.h>
1da177e4 58
1da177e4
LT
59struct ip6frag_skb_cb
60{
61 struct inet6_skb_parm h;
62 int offset;
63};
64
65#define FRAG6_CB(skb) ((struct ip6frag_skb_cb*)((skb)->cb))
66
67
68/*
69 * Equivalent of ipv4 struct ipq
70 */
71
72struct frag_queue
73{
5ab11c98 74 struct inet_frag_queue q;
1da177e4 75
e69a4adc 76 __be32 id; /* fragment id */
1da177e4
LT
77 struct in6_addr saddr;
78 struct in6_addr daddr;
79
1da177e4 80 int iif;
1da177e4 81 unsigned int csum;
1da177e4 82 __u16 nhoffset;
1da177e4
LT
83};
84
7eb95156 85static struct inet_frags ip6_frags;
1da177e4 86
e5a2bb84 87int ip6_frag_nqueues(struct net *net)
7eb95156 88{
e5a2bb84 89 return net->ipv6.frags.nqueues;
7eb95156 90}
1da177e4 91
6ddc0822 92int ip6_frag_mem(struct net *net)
7eb95156 93{
6ddc0822 94 return atomic_read(&net->ipv6.frags.mem);
7eb95156 95}
1da177e4 96
f61944ef
HX
97static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
98 struct net_device *dev);
99
f6596f9d
ZB
100/*
101 * callers should be careful not to use the hash value outside the ipfrag_lock
102 * as doing so could race with ipfrag_hash_rnd being recalculated.
103 */
e69a4adc 104static unsigned int ip6qhashfn(__be32 id, struct in6_addr *saddr,
1da177e4
LT
105 struct in6_addr *daddr)
106{
107 u32 a, b, c;
108
e69a4adc
AV
109 a = (__force u32)saddr->s6_addr32[0];
110 b = (__force u32)saddr->s6_addr32[1];
111 c = (__force u32)saddr->s6_addr32[2];
1da177e4
LT
112
113 a += JHASH_GOLDEN_RATIO;
114 b += JHASH_GOLDEN_RATIO;
7eb95156 115 c += ip6_frags.rnd;
1da177e4
LT
116 __jhash_mix(a, b, c);
117
e69a4adc
AV
118 a += (__force u32)saddr->s6_addr32[3];
119 b += (__force u32)daddr->s6_addr32[0];
120 c += (__force u32)daddr->s6_addr32[1];
1da177e4
LT
121 __jhash_mix(a, b, c);
122
e69a4adc
AV
123 a += (__force u32)daddr->s6_addr32[2];
124 b += (__force u32)daddr->s6_addr32[3];
125 c += (__force u32)id;
1da177e4
LT
126 __jhash_mix(a, b, c);
127
7eb95156 128 return c & (INETFRAGS_HASHSZ - 1);
1da177e4
LT
129}
130
321a3a99 131static unsigned int ip6_hashfn(struct inet_frag_queue *q)
1da177e4 132{
321a3a99 133 struct frag_queue *fq;
1da177e4 134
321a3a99
PE
135 fq = container_of(q, struct frag_queue, q);
136 return ip6qhashfn(fq->id, &fq->saddr, &fq->daddr);
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);
145 return (fq->id == arg->id &&
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;
167 ipv6_addr_copy(&fq->saddr, arg->src);
168 ipv6_addr_copy(&fq->daddr, arg->dst);
1da177e4 169}
c6fda282 170EXPORT_SYMBOL(ip6_frag_init);
1da177e4 171
1da177e4
LT
172/* Destruction primitives. */
173
4b6cb5d8 174static __inline__ void fq_put(struct frag_queue *fq)
1da177e4 175{
762cc408 176 inet_frag_put(&fq->q, &ip6_frags);
1da177e4
LT
177}
178
179/* Kill fq entry. It is not destroyed immediately,
180 * because caller (and someone more) holds reference count.
181 */
182static __inline__ void fq_kill(struct frag_queue *fq)
183{
277e650d 184 inet_frag_kill(&fq->q, &ip6_frags);
1da177e4
LT
185}
186
6ddc0822 187static void ip6_evictor(struct net *net, struct inet6_dev *idev)
1da177e4 188{
8e7999c4
PE
189 int evicted;
190
6ddc0822 191 evicted = inet_frag_evictor(&net->ipv6.frags, &ip6_frags);
8e7999c4
PE
192 if (evicted)
193 IP6_ADD_STATS_BH(idev, IPSTATS_MIB_REASMFAILS, evicted);
1da177e4
LT
194}
195
196static void ip6_frag_expire(unsigned long data)
197{
e521db9d 198 struct frag_queue *fq;
a11d206d 199 struct net_device *dev = NULL;
4ac2ccd0 200 struct net *net;
1da177e4 201
e521db9d
PE
202 fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q);
203
5ab11c98 204 spin_lock(&fq->q.lock);
1da177e4 205
bc578a54 206 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
207 goto out;
208
209 fq_kill(fq);
210
4ac2ccd0
DL
211 net = container_of(fq->q.net, struct net, ipv6.frags);
212 dev = dev_get_by_index(net, fq->iif);
a11d206d
YH
213 if (!dev)
214 goto out;
215
216 rcu_read_lock();
217 IP6_INC_STATS_BH(__in6_dev_get(dev), IPSTATS_MIB_REASMTIMEOUT);
218 IP6_INC_STATS_BH(__in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
219 rcu_read_unlock();
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)
78c784c4
IO
223 goto out;
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);
1da177e4 232out:
a11d206d
YH
233 if (dev)
234 dev_put(dev);
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
PE
247 arg.id = id;
248 arg.src = src;
249 arg.dst = dst;
abd6523d 250 hash = ip6qhashfn(id, src, dst);
1da177e4 251
ac18e750 252 q = inet_frag_find(&net->ipv6.frags, &ip6_frags, &arg, hash);
c6fda282
PE
253 if (q == NULL)
254 goto oom;
1da177e4 255
c6fda282 256 return container_of(q, struct frag_queue, q);
1da177e4
LT
257
258oom:
a11d206d 259 IP6_INC_STATS_BH(idev, IPSTATS_MIB_REASMFAILS);
1da177e4
LT
260 return NULL;
261}
262
f61944ef 263static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
1da177e4
LT
264 struct frag_hdr *fhdr, int nhoff)
265{
266 struct sk_buff *prev, *next;
f61944ef 267 struct net_device *dev;
1da177e4
LT
268 int offset, end;
269
bc578a54 270 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
271 goto err;
272
273 offset = ntohs(fhdr->frag_off) & ~0x7;
0660e03f
ACM
274 end = offset + (ntohs(ipv6_hdr(skb)->payload_len) -
275 ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1)));
1da177e4
LT
276
277 if ((unsigned int)end > IPV6_MAXPLEN) {
a11d206d
YH
278 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
279 IPSTATS_MIB_INHDRERRORS);
d56f90a7
ACM
280 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
281 ((u8 *)&fhdr->frag_off -
282 skb_network_header(skb)));
f61944ef 283 return -1;
1da177e4
LT
284 }
285
d56f90a7
ACM
286 if (skb->ip_summed == CHECKSUM_COMPLETE) {
287 const unsigned char *nh = skb_network_header(skb);
1ab1457c 288 skb->csum = csum_sub(skb->csum,
d56f90a7
ACM
289 csum_partial(nh, (u8 *)(fhdr + 1) - nh,
290 0));
291 }
1da177e4
LT
292
293 /* Is this the final fragment? */
294 if (!(fhdr->frag_off & htons(IP6_MF))) {
295 /* If we already have some bits beyond end
296 * or have different end, the segment is corrupted.
297 */
5ab11c98 298 if (end < fq->q.len ||
bc578a54 299 ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len))
1da177e4 300 goto err;
bc578a54 301 fq->q.last_in |= INET_FRAG_LAST_IN;
5ab11c98 302 fq->q.len = end;
1da177e4
LT
303 } else {
304 /* Check if the fragment is rounded to 8 bytes.
305 * Required by the RFC.
306 */
307 if (end & 0x7) {
308 /* RFC2460 says always send parameter problem in
309 * this case. -DaveM
310 */
a11d206d
YH
311 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst),
312 IPSTATS_MIB_INHDRERRORS);
1ab1457c 313 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
1da177e4 314 offsetof(struct ipv6hdr, payload_len));
f61944ef 315 return -1;
1da177e4 316 }
5ab11c98 317 if (end > fq->q.len) {
1da177e4 318 /* Some bits beyond end -> corruption. */
bc578a54 319 if (fq->q.last_in & INET_FRAG_LAST_IN)
1da177e4 320 goto err;
5ab11c98 321 fq->q.len = end;
1da177e4
LT
322 }
323 }
324
325 if (end == offset)
326 goto err;
327
328 /* Point into the IP datagram 'data' part. */
329 if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
330 goto err;
1ab1457c 331
42ca89c1
SH
332 if (pskb_trim_rcsum(skb, end - offset))
333 goto err;
1da177e4
LT
334
335 /* Find out which fragments are in front and at the back of us
336 * in the chain of fragments so far. We must know where to put
337 * this fragment, right?
338 */
339 prev = NULL;
5ab11c98 340 for(next = fq->q.fragments; next != NULL; next = next->next) {
1da177e4
LT
341 if (FRAG6_CB(next)->offset >= offset)
342 break; /* bingo! */
343 prev = next;
344 }
345
346 /* We found where to put this one. Check for overlap with
347 * preceding fragment, and, if needed, align things so that
348 * any overlaps are eliminated.
349 */
350 if (prev) {
351 int i = (FRAG6_CB(prev)->offset + prev->len) - offset;
352
353 if (i > 0) {
354 offset += i;
355 if (end <= offset)
356 goto err;
357 if (!pskb_pull(skb, i))
358 goto err;
359 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
360 skb->ip_summed = CHECKSUM_NONE;
361 }
362 }
363
364 /* Look for overlap with succeeding segments.
365 * If we can merge fragments, do it.
366 */
367 while (next && FRAG6_CB(next)->offset < end) {
368 int i = end - FRAG6_CB(next)->offset; /* overlap is 'i' bytes */
369
370 if (i < next->len) {
371 /* Eat head of the next overlapped fragment
372 * and leave the loop. The next ones cannot overlap.
373 */
374 if (!pskb_pull(next, i))
375 goto err;
376 FRAG6_CB(next)->offset += i; /* next fragment */
5ab11c98 377 fq->q.meat -= i;
1da177e4
LT
378 if (next->ip_summed != CHECKSUM_UNNECESSARY)
379 next->ip_summed = CHECKSUM_NONE;
380 break;
381 } else {
382 struct sk_buff *free_it = next;
383
384 /* Old fragment is completely overridden with
385 * new one drop it.
386 */
387 next = next->next;
388
389 if (prev)
390 prev->next = next;
391 else
5ab11c98 392 fq->q.fragments = next;
1da177e4 393
5ab11c98 394 fq->q.meat -= free_it->len;
6ddc0822 395 frag_kfree_skb(fq->q.net, free_it, NULL);
1da177e4
LT
396 }
397 }
398
399 FRAG6_CB(skb)->offset = offset;
400
401 /* Insert this fragment in the chain of fragments. */
402 skb->next = next;
403 if (prev)
404 prev->next = skb;
405 else
5ab11c98 406 fq->q.fragments = skb;
1da177e4 407
f61944ef
HX
408 dev = skb->dev;
409 if (dev) {
410 fq->iif = dev->ifindex;
411 skb->dev = NULL;
412 }
5ab11c98
PE
413 fq->q.stamp = skb->tstamp;
414 fq->q.meat += skb->len;
6ddc0822 415 atomic_add(skb->truesize, &fq->q.net->mem);
1da177e4
LT
416
417 /* The first fragment.
418 * nhoffset is obtained from the first fragment, of course.
419 */
420 if (offset == 0) {
421 fq->nhoffset = nhoff;
bc578a54 422 fq->q.last_in |= INET_FRAG_FIRST_IN;
1da177e4 423 }
f61944ef 424
bc578a54
JP
425 if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
426 fq->q.meat == fq->q.len)
f61944ef
HX
427 return ip6_frag_reasm(fq, prev, dev);
428
7eb95156 429 write_lock(&ip6_frags.lock);
3140c25c 430 list_move_tail(&fq->q.lru_list, &fq->q.net->lru_list);
7eb95156 431 write_unlock(&ip6_frags.lock);
f61944ef 432 return -1;
1da177e4
LT
433
434err:
a11d206d 435 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMFAILS);
1da177e4 436 kfree_skb(skb);
f61944ef 437 return -1;
1da177e4
LT
438}
439
440/*
441 * Check if this packet is complete.
442 * Returns NULL on failure by any reason, and pointer
443 * to current nexthdr field in reassembled frame.
444 *
445 * It is called with locked fq, and caller must check that
446 * queue is eligible for reassembly i.e. it is not COMPLETE,
447 * the last and the first frames arrived and all the bits are here.
448 */
f61944ef 449static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
1da177e4
LT
450 struct net_device *dev)
451{
5ab11c98 452 struct sk_buff *fp, *head = fq->q.fragments;
1da177e4
LT
453 int payload_len;
454 unsigned int nhoff;
455
456 fq_kill(fq);
457
f61944ef
HX
458 /* Make the one we just received the head. */
459 if (prev) {
460 head = prev->next;
461 fp = skb_clone(head, GFP_ATOMIC);
462
463 if (!fp)
464 goto out_oom;
465
466 fp->next = head->next;
467 prev->next = fp;
468
5ab11c98
PE
469 skb_morph(head, fq->q.fragments);
470 head->next = fq->q.fragments->next;
f61944ef 471
5ab11c98
PE
472 kfree_skb(fq->q.fragments);
473 fq->q.fragments = head;
f61944ef
HX
474 }
475
1da177e4
LT
476 BUG_TRAP(head != NULL);
477 BUG_TRAP(FRAG6_CB(head)->offset == 0);
478
479 /* Unfragmented part is taken from the first segment. */
d56f90a7 480 payload_len = ((head->data - skb_network_header(head)) -
5ab11c98 481 sizeof(struct ipv6hdr) + fq->q.len -
d56f90a7 482 sizeof(struct frag_hdr));
1da177e4
LT
483 if (payload_len > IPV6_MAXPLEN)
484 goto out_oversize;
485
486 /* Head of list must not be cloned. */
487 if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC))
488 goto out_oom;
489
490 /* If the first fragment is fragmented itself, we split
491 * it to two chunks: the first with data and paged part
492 * and the second, holding only fragments. */
493 if (skb_shinfo(head)->frag_list) {
494 struct sk_buff *clone;
495 int i, plen = 0;
496
497 if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
498 goto out_oom;
499 clone->next = head->next;
500 head->next = clone;
501 skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
502 skb_shinfo(head)->frag_list = NULL;
503 for (i=0; i<skb_shinfo(head)->nr_frags; i++)
504 plen += skb_shinfo(head)->frags[i].size;
505 clone->len = clone->data_len = head->data_len - plen;
506 head->data_len -= clone->len;
507 head->len -= clone->len;
508 clone->csum = 0;
509 clone->ip_summed = head->ip_summed;
6ddc0822 510 atomic_add(clone->truesize, &fq->q.net->mem);
1da177e4
LT
511 }
512
513 /* We have to remove fragment header from datagram and to relocate
514 * header in order to calculate ICV correctly. */
515 nhoff = fq->nhoffset;
b0e380b1 516 skb_network_header(head)[nhoff] = skb_transport_header(head)[0];
1ab1457c 517 memmove(head->head + sizeof(struct frag_hdr), head->head,
1da177e4 518 (head->data - head->head) - sizeof(struct frag_hdr));
b0e380b1
ACM
519 head->mac_header += sizeof(struct frag_hdr);
520 head->network_header += sizeof(struct frag_hdr);
1da177e4
LT
521
522 skb_shinfo(head)->frag_list = head->next;
badff6d0 523 skb_reset_transport_header(head);
d56f90a7 524 skb_push(head, head->data - skb_network_header(head));
6ddc0822 525 atomic_sub(head->truesize, &fq->q.net->mem);
1da177e4
LT
526
527 for (fp=head->next; fp; fp = fp->next) {
528 head->data_len += fp->len;
529 head->len += fp->len;
530 if (head->ip_summed != fp->ip_summed)
531 head->ip_summed = CHECKSUM_NONE;
84fa7933 532 else if (head->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
533 head->csum = csum_add(head->csum, fp->csum);
534 head->truesize += fp->truesize;
6ddc0822 535 atomic_sub(fp->truesize, &fq->q.net->mem);
1da177e4
LT
536 }
537
538 head->next = NULL;
539 head->dev = dev;
5ab11c98 540 head->tstamp = fq->q.stamp;
0660e03f 541 ipv6_hdr(head)->payload_len = htons(payload_len);
951dbc8a 542 IP6CB(head)->nhoff = nhoff;
1da177e4 543
1da177e4 544 /* Yes, and fold redundant checksum back. 8) */
84fa7933 545 if (head->ip_summed == CHECKSUM_COMPLETE)
d56f90a7 546 head->csum = csum_partial(skb_network_header(head),
cfe1fc77 547 skb_network_header_len(head),
d56f90a7 548 head->csum);
1da177e4 549
a11d206d
YH
550 rcu_read_lock();
551 IP6_INC_STATS_BH(__in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
552 rcu_read_unlock();
5ab11c98 553 fq->q.fragments = NULL;
1da177e4
LT
554 return 1;
555
556out_oversize:
557 if (net_ratelimit())
558 printk(KERN_DEBUG "ip6_frag_reasm: payload len = %d\n", payload_len);
559 goto out_fail;
560out_oom:
561 if (net_ratelimit())
562 printk(KERN_DEBUG "ip6_frag_reasm: no memory for reassembly\n");
563out_fail:
a11d206d
YH
564 rcu_read_lock();
565 IP6_INC_STATS_BH(__in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
566 rcu_read_unlock();
1da177e4
LT
567 return -1;
568}
569
e5bbef20 570static int ipv6_frag_rcv(struct sk_buff *skb)
1da177e4 571{
1da177e4
LT
572 struct frag_hdr *fhdr;
573 struct frag_queue *fq;
0660e03f 574 struct ipv6hdr *hdr = ipv6_hdr(skb);
ac18e750 575 struct net *net;
1da177e4 576
a11d206d 577 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMREQDS);
1da177e4
LT
578
579 /* Jumbo payload inhibits frag. header */
580 if (hdr->payload_len==0) {
a11d206d 581 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INHDRERRORS);
cfe1fc77
ACM
582 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
583 skb_network_header_len(skb));
1da177e4
LT
584 return -1;
585 }
ea2ae17d
ACM
586 if (!pskb_may_pull(skb, (skb_transport_offset(skb) +
587 sizeof(struct frag_hdr)))) {
a11d206d 588 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INHDRERRORS);
cfe1fc77
ACM
589 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
590 skb_network_header_len(skb));
1da177e4
LT
591 return -1;
592 }
593
0660e03f 594 hdr = ipv6_hdr(skb);
9c70220b 595 fhdr = (struct frag_hdr *)skb_transport_header(skb);
1da177e4
LT
596
597 if (!(fhdr->frag_off & htons(0xFFF9))) {
598 /* It is not a fragmented frame */
b0e380b1 599 skb->transport_header += sizeof(struct frag_hdr);
a11d206d 600 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMOKS);
1da177e4 601
d56f90a7 602 IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
1da177e4
LT
603 return 1;
604 }
605
c346dca1 606 net = dev_net(skb->dev);
e31e0bdc 607 if (atomic_read(&net->ipv6.frags.mem) > net->ipv6.frags.high_thresh)
6ddc0822 608 ip6_evictor(net, ip6_dst_idev(skb->dst));
1da177e4 609
ac18e750 610 if ((fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr,
a11d206d 611 ip6_dst_idev(skb->dst))) != NULL) {
f61944ef 612 int ret;
1da177e4 613
5ab11c98 614 spin_lock(&fq->q.lock);
1da177e4 615
f61944ef 616 ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);
1da177e4 617
5ab11c98 618 spin_unlock(&fq->q.lock);
4b6cb5d8 619 fq_put(fq);
1da177e4
LT
620 return ret;
621 }
622
a11d206d 623 IP6_INC_STATS_BH(ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMFAILS);
1da177e4
LT
624 kfree_skb(skb);
625 return -1;
626}
627
628static struct inet6_protocol frag_protocol =
629{
630 .handler = ipv6_frag_rcv,
631 .flags = INET6_PROTO_NOPOLICY,
632};
633
8d8354d2 634#ifdef CONFIG_SYSCTL
0a64b4b8 635static struct ctl_table ip6_frags_ns_ctl_table[] = {
8d8354d2
PE
636 {
637 .ctl_name = NET_IPV6_IP6FRAG_HIGH_THRESH,
638 .procname = "ip6frag_high_thresh",
e31e0bdc 639 .data = &init_net.ipv6.frags.high_thresh,
8d8354d2
PE
640 .maxlen = sizeof(int),
641 .mode = 0644,
642 .proc_handler = &proc_dointvec
643 },
644 {
645 .ctl_name = NET_IPV6_IP6FRAG_LOW_THRESH,
646 .procname = "ip6frag_low_thresh",
e31e0bdc 647 .data = &init_net.ipv6.frags.low_thresh,
8d8354d2
PE
648 .maxlen = sizeof(int),
649 .mode = 0644,
650 .proc_handler = &proc_dointvec
651 },
652 {
653 .ctl_name = NET_IPV6_IP6FRAG_TIME,
654 .procname = "ip6frag_time",
b2fd5321 655 .data = &init_net.ipv6.frags.timeout,
8d8354d2
PE
656 .maxlen = sizeof(int),
657 .mode = 0644,
658 .proc_handler = &proc_dointvec_jiffies,
659 .strategy = &sysctl_jiffies,
660 },
661 {
662 .ctl_name = NET_IPV6_IP6FRAG_SECRET_INTERVAL,
663 .procname = "ip6frag_secret_interval",
3b4bc4a2 664 .data = &ip6_frags.secret_interval,
8d8354d2
PE
665 .maxlen = sizeof(int),
666 .mode = 0644,
667 .proc_handler = &proc_dointvec_jiffies,
668 .strategy = &sysctl_jiffies
669 },
670 { }
671};
672
0a64b4b8 673static int ip6_frags_ns_sysctl_register(struct net *net)
8d8354d2 674{
e4a2d5c2 675 struct ctl_table *table;
8d8354d2
PE
676 struct ctl_table_header *hdr;
677
0a64b4b8 678 table = ip6_frags_ns_ctl_table;
e4a2d5c2 679 if (net != &init_net) {
0a64b4b8 680 table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL);
e4a2d5c2
PE
681 if (table == NULL)
682 goto err_alloc;
683
e31e0bdc
PE
684 table[0].data = &net->ipv6.frags.high_thresh;
685 table[1].data = &net->ipv6.frags.low_thresh;
b2fd5321 686 table[2].data = &net->ipv6.frags.timeout;
e4a2d5c2
PE
687 table[3].mode &= ~0222;
688 }
689
690 hdr = register_net_sysctl_table(net, net_ipv6_ctl_path, table);
691 if (hdr == NULL)
692 goto err_reg;
693
694 net->ipv6.sysctl.frags_hdr = hdr;
695 return 0;
696
697err_reg:
698 if (net != &init_net)
699 kfree(table);
700err_alloc:
701 return -ENOMEM;
702}
703
0a64b4b8 704static void ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
705{
706 struct ctl_table *table;
707
708 table = net->ipv6.sysctl.frags_hdr->ctl_table_arg;
709 unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr);
710 kfree(table);
8d8354d2
PE
711}
712#else
0a64b4b8 713static inline int ip6_frags_ns_sysctl_register(struct net *net)
e71e0349 714{
8d8354d2
PE
715 return 0;
716}
e4a2d5c2 717
0a64b4b8 718static inline void ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
719{
720}
8d8354d2 721#endif
7d460db9 722
8d8354d2
PE
723static int ipv6_frags_init_net(struct net *net)
724{
e31e0bdc
PE
725 net->ipv6.frags.high_thresh = 256 * 1024;
726 net->ipv6.frags.low_thresh = 192 * 1024;
b2fd5321 727 net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT;
8d8354d2 728
e5a2bb84
PE
729 inet_frags_init_net(&net->ipv6.frags);
730
0a64b4b8 731 return ip6_frags_ns_sysctl_register(net);
e71e0349
DL
732}
733
81566e83
PE
734static void ipv6_frags_exit_net(struct net *net)
735{
0a64b4b8 736 ip6_frags_ns_sysctl_unregister(net);
81566e83
PE
737 inet_frags_exit_net(&net->ipv6.frags, &ip6_frags);
738}
739
740static struct pernet_operations ip6_frags_ops = {
741 .init = ipv6_frags_init_net,
742 .exit = ipv6_frags_exit_net,
743};
744
853cbbaa 745int __init ipv6_frag_init(void)
1da177e4 746{
853cbbaa 747 int ret;
1da177e4 748
853cbbaa
DL
749 ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT);
750 if (ret)
751 goto out;
e71e0349 752
0002c630
PE
753 ret = register_pernet_subsys(&ip6_frags_ops);
754 if (ret)
755 goto err_pernet;
8d8354d2 756
321a3a99 757 ip6_frags.hashfn = ip6_hashfn;
c6fda282 758 ip6_frags.constructor = ip6_frag_init;
c9547709 759 ip6_frags.destructor = NULL;
1e4b8287
PE
760 ip6_frags.skb_free = NULL;
761 ip6_frags.qsize = sizeof(struct frag_queue);
abd6523d 762 ip6_frags.match = ip6_frag_match;
e521db9d 763 ip6_frags.frag_expire = ip6_frag_expire;
3b4bc4a2 764 ip6_frags.secret_interval = 10 * 60 * HZ;
7eb95156 765 inet_frags_init(&ip6_frags);
853cbbaa
DL
766out:
767 return ret;
0002c630
PE
768
769err_pernet:
770 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
771 goto out;
853cbbaa
DL
772}
773
774void ipv6_frag_exit(void)
775{
776 inet_frags_fini(&ip6_frags);
81566e83 777 unregister_pernet_subsys(&ip6_frags_ops);
853cbbaa 778 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
1da177e4 779}