Merge git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending
[GitHub/mt8127/android_kernel_alcatel_ttab.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>
5a0e3ad6 44#include <linux/slab.h>
bc3b2d7f 45#include <linux/export.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 */
0b5ccb2e 77 u32 user;
1da177e4
LT
78 struct in6_addr saddr;
79 struct in6_addr daddr;
80
1da177e4 81 int iif;
1da177e4 82 unsigned int csum;
1da177e4 83 __u16 nhoffset;
1da177e4
LT
84};
85
7eb95156 86static struct inet_frags ip6_frags;
1da177e4 87
e5a2bb84 88int ip6_frag_nqueues(struct net *net)
7eb95156 89{
e5a2bb84 90 return net->ipv6.frags.nqueues;
7eb95156 91}
1da177e4 92
6ddc0822 93int ip6_frag_mem(struct net *net)
7eb95156 94{
6ddc0822 95 return atomic_read(&net->ipv6.frags.mem);
7eb95156 96}
1da177e4 97
f61944ef
HX
98static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
99 struct net_device *dev);
100
f6596f9d
ZB
101/*
102 * callers should be careful not to use the hash value outside the ipfrag_lock
103 * as doing so could race with ipfrag_hash_rnd being recalculated.
104 */
93c8b90f
IJ
105unsigned int inet6_hash_frag(__be32 id, const struct in6_addr *saddr,
106 const struct in6_addr *daddr, u32 rnd)
1da177e4 107{
82a39eb6
JK
108 u32 c;
109
110 c = jhash_3words((__force u32)saddr->s6_addr32[0],
111 (__force u32)saddr->s6_addr32[1],
112 (__force u32)saddr->s6_addr32[2],
113 rnd);
114
115 c = jhash_3words((__force u32)saddr->s6_addr32[3],
116 (__force u32)daddr->s6_addr32[0],
117 (__force u32)daddr->s6_addr32[1],
118 c);
119
120 c = jhash_3words((__force u32)daddr->s6_addr32[2],
121 (__force u32)daddr->s6_addr32[3],
122 (__force u32)id,
123 c);
1da177e4 124
7eb95156 125 return c & (INETFRAGS_HASHSZ - 1);
1da177e4 126}
93c8b90f 127EXPORT_SYMBOL_GPL(inet6_hash_frag);
1da177e4 128
321a3a99 129static unsigned int ip6_hashfn(struct inet_frag_queue *q)
1da177e4 130{
321a3a99 131 struct frag_queue *fq;
1da177e4 132
321a3a99 133 fq = container_of(q, struct frag_queue, q);
93c8b90f 134 return inet6_hash_frag(fq->id, &fq->saddr, &fq->daddr, ip6_frags.rnd);
1da177e4
LT
135}
136
cbc264ca 137bool ip6_frag_match(struct inet_frag_queue *q, void *a)
abd6523d
PE
138{
139 struct frag_queue *fq;
140 struct ip6_create_arg *arg = a;
141
142 fq = container_of(q, struct frag_queue, q);
cbc264ca
ED
143 return fq->id == arg->id &&
144 fq->user == arg->user &&
145 ipv6_addr_equal(&fq->saddr, arg->src) &&
146 ipv6_addr_equal(&fq->daddr, arg->dst);
abd6523d
PE
147}
148EXPORT_SYMBOL(ip6_frag_match);
149
c6fda282 150void ip6_frag_init(struct inet_frag_queue *q, void *a)
1da177e4 151{
c6fda282
PE
152 struct frag_queue *fq = container_of(q, struct frag_queue, q);
153 struct ip6_create_arg *arg = a;
154
155 fq->id = arg->id;
0b5ccb2e 156 fq->user = arg->user;
4e3fd7a0
AD
157 fq->saddr = *arg->src;
158 fq->daddr = *arg->dst;
1da177e4 159}
c6fda282 160EXPORT_SYMBOL(ip6_frag_init);
1da177e4 161
1da177e4
LT
162/* Destruction primitives. */
163
4b6cb5d8 164static __inline__ void fq_put(struct frag_queue *fq)
1da177e4 165{
762cc408 166 inet_frag_put(&fq->q, &ip6_frags);
1da177e4
LT
167}
168
169/* Kill fq entry. It is not destroyed immediately,
170 * because caller (and someone more) holds reference count.
171 */
172static __inline__ void fq_kill(struct frag_queue *fq)
173{
277e650d 174 inet_frag_kill(&fq->q, &ip6_frags);
1da177e4
LT
175}
176
6ddc0822 177static void ip6_evictor(struct net *net, struct inet6_dev *idev)
1da177e4 178{
8e7999c4
PE
179 int evicted;
180
6ddc0822 181 evicted = inet_frag_evictor(&net->ipv6.frags, &ip6_frags);
8e7999c4 182 if (evicted)
821d5777 183 IP6_ADD_STATS_BH(net, idev, IPSTATS_MIB_REASMFAILS, evicted);
1da177e4
LT
184}
185
186static void ip6_frag_expire(unsigned long data)
187{
e521db9d 188 struct frag_queue *fq;
a11d206d 189 struct net_device *dev = NULL;
4ac2ccd0 190 struct net *net;
1da177e4 191
e521db9d
PE
192 fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q);
193
5ab11c98 194 spin_lock(&fq->q.lock);
1da177e4 195
bc578a54 196 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
197 goto out;
198
199 fq_kill(fq);
200
4ac2ccd0 201 net = container_of(fq->q.net, struct net, ipv6.frags);
69df9d59
ED
202 rcu_read_lock();
203 dev = dev_get_by_index_rcu(net, fq->iif);
a11d206d 204 if (!dev)
69df9d59 205 goto out_rcu_unlock;
a11d206d 206
483a47d2
DL
207 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMTIMEOUT);
208 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
1da177e4 209
78c784c4 210 /* Don't send error if the first segment did not arrive. */
bc578a54 211 if (!(fq->q.last_in & INET_FRAG_FIRST_IN) || !fq->q.fragments)
69df9d59 212 goto out_rcu_unlock;
78c784c4 213
78c784c4
IO
214 /*
215 But use as source device on which LAST ARRIVED
216 segment was received. And do not use fq->dev
217 pointer directly, device might already disappeared.
218 */
5ab11c98 219 fq->q.fragments->dev = dev;
3ffe533c 220 icmpv6_send(fq->q.fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0);
69df9d59
ED
221out_rcu_unlock:
222 rcu_read_unlock();
1da177e4 223out:
5ab11c98 224 spin_unlock(&fq->q.lock);
4b6cb5d8 225 fq_put(fq);
1da177e4
LT
226}
227
abd6523d 228static __inline__ struct frag_queue *
b71d1d42 229fq_find(struct net *net, __be32 id, const struct in6_addr *src, const struct in6_addr *dst)
1da177e4 230{
c6fda282
PE
231 struct inet_frag_queue *q;
232 struct ip6_create_arg arg;
abd6523d 233 unsigned int hash;
1da177e4 234
c6fda282 235 arg.id = id;
0b5ccb2e 236 arg.user = IP6_DEFRAG_LOCAL_DELIVER;
c6fda282
PE
237 arg.src = src;
238 arg.dst = dst;
9a375803
PE
239
240 read_lock(&ip6_frags.lock);
93c8b90f 241 hash = inet6_hash_frag(id, src, dst, ip6_frags.rnd);
1da177e4 242
ac18e750 243 q = inet_frag_find(&net->ipv6.frags, &ip6_frags, &arg, hash);
c6fda282 244 if (q == NULL)
9546377c 245 return NULL;
1da177e4 246
c6fda282 247 return container_of(q, struct frag_queue, q);
1da177e4
LT
248}
249
f61944ef 250static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
1da177e4
LT
251 struct frag_hdr *fhdr, int nhoff)
252{
253 struct sk_buff *prev, *next;
f61944ef 254 struct net_device *dev;
1da177e4 255 int offset, end;
adf30907 256 struct net *net = dev_net(skb_dst(skb)->dev);
1da177e4 257
bc578a54 258 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
259 goto err;
260
261 offset = ntohs(fhdr->frag_off) & ~0x7;
0660e03f
ACM
262 end = offset + (ntohs(ipv6_hdr(skb)->payload_len) -
263 ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1)));
1da177e4
LT
264
265 if ((unsigned int)end > IPV6_MAXPLEN) {
adf30907 266 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
a11d206d 267 IPSTATS_MIB_INHDRERRORS);
d56f90a7
ACM
268 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
269 ((u8 *)&fhdr->frag_off -
270 skb_network_header(skb)));
f61944ef 271 return -1;
1da177e4
LT
272 }
273
d56f90a7
ACM
274 if (skb->ip_summed == CHECKSUM_COMPLETE) {
275 const unsigned char *nh = skb_network_header(skb);
1ab1457c 276 skb->csum = csum_sub(skb->csum,
d56f90a7
ACM
277 csum_partial(nh, (u8 *)(fhdr + 1) - nh,
278 0));
279 }
1da177e4
LT
280
281 /* Is this the final fragment? */
282 if (!(fhdr->frag_off & htons(IP6_MF))) {
283 /* If we already have some bits beyond end
284 * or have different end, the segment is corrupted.
285 */
5ab11c98 286 if (end < fq->q.len ||
bc578a54 287 ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len))
1da177e4 288 goto err;
bc578a54 289 fq->q.last_in |= INET_FRAG_LAST_IN;
5ab11c98 290 fq->q.len = end;
1da177e4
LT
291 } else {
292 /* Check if the fragment is rounded to 8 bytes.
293 * Required by the RFC.
294 */
295 if (end & 0x7) {
296 /* RFC2460 says always send parameter problem in
297 * this case. -DaveM
298 */
adf30907 299 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
a11d206d 300 IPSTATS_MIB_INHDRERRORS);
1ab1457c 301 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
1da177e4 302 offsetof(struct ipv6hdr, payload_len));
f61944ef 303 return -1;
1da177e4 304 }
5ab11c98 305 if (end > fq->q.len) {
1da177e4 306 /* Some bits beyond end -> corruption. */
bc578a54 307 if (fq->q.last_in & INET_FRAG_LAST_IN)
1da177e4 308 goto err;
5ab11c98 309 fq->q.len = end;
1da177e4
LT
310 }
311 }
312
313 if (end == offset)
314 goto err;
315
316 /* Point into the IP datagram 'data' part. */
317 if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
318 goto err;
1ab1457c 319
42ca89c1
SH
320 if (pskb_trim_rcsum(skb, end - offset))
321 goto err;
1da177e4
LT
322
323 /* Find out which fragments are in front and at the back of us
324 * in the chain of fragments so far. We must know where to put
325 * this fragment, right?
326 */
d6bebca9
CG
327 prev = fq->q.fragments_tail;
328 if (!prev || FRAG6_CB(prev)->offset < offset) {
329 next = NULL;
330 goto found;
331 }
1da177e4 332 prev = NULL;
5ab11c98 333 for(next = fq->q.fragments; next != NULL; next = next->next) {
1da177e4
LT
334 if (FRAG6_CB(next)->offset >= offset)
335 break; /* bingo! */
336 prev = next;
337 }
338
d6bebca9 339found:
5de658f8
ED
340 /* RFC5722, Section 4, amended by Errata ID : 3089
341 * When reassembling an IPv6 datagram, if
70789d70
ND
342 * one or more its constituent fragments is determined to be an
343 * overlapping fragment, the entire datagram (and any constituent
5de658f8 344 * fragments) MUST be silently discarded.
1da177e4 345 */
1da177e4 346
70789d70
ND
347 /* Check for overlap with preceding fragment. */
348 if (prev &&
f4642141 349 (FRAG6_CB(prev)->offset + prev->len) > offset)
70789d70 350 goto discard_fq;
1da177e4 351
70789d70
ND
352 /* Look for overlap with succeeding segment. */
353 if (next && FRAG6_CB(next)->offset < end)
354 goto discard_fq;
1da177e4
LT
355
356 FRAG6_CB(skb)->offset = offset;
357
358 /* Insert this fragment in the chain of fragments. */
359 skb->next = next;
d6bebca9
CG
360 if (!next)
361 fq->q.fragments_tail = skb;
1da177e4
LT
362 if (prev)
363 prev->next = skb;
364 else
5ab11c98 365 fq->q.fragments = skb;
1da177e4 366
f61944ef
HX
367 dev = skb->dev;
368 if (dev) {
369 fq->iif = dev->ifindex;
370 skb->dev = NULL;
371 }
5ab11c98
PE
372 fq->q.stamp = skb->tstamp;
373 fq->q.meat += skb->len;
6ddc0822 374 atomic_add(skb->truesize, &fq->q.net->mem);
1da177e4
LT
375
376 /* The first fragment.
377 * nhoffset is obtained from the first fragment, of course.
378 */
379 if (offset == 0) {
380 fq->nhoffset = nhoff;
bc578a54 381 fq->q.last_in |= INET_FRAG_FIRST_IN;
1da177e4 382 }
f61944ef 383
bc578a54
JP
384 if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
385 fq->q.meat == fq->q.len)
f61944ef
HX
386 return ip6_frag_reasm(fq, prev, dev);
387
7eb95156 388 write_lock(&ip6_frags.lock);
3140c25c 389 list_move_tail(&fq->q.lru_list, &fq->q.net->lru_list);
7eb95156 390 write_unlock(&ip6_frags.lock);
f61944ef 391 return -1;
1da177e4 392
70789d70
ND
393discard_fq:
394 fq_kill(fq);
1da177e4 395err:
adf30907 396 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
3bd653c8 397 IPSTATS_MIB_REASMFAILS);
1da177e4 398 kfree_skb(skb);
f61944ef 399 return -1;
1da177e4
LT
400}
401
402/*
403 * Check if this packet is complete.
404 * Returns NULL on failure by any reason, and pointer
405 * to current nexthdr field in reassembled frame.
406 *
407 * It is called with locked fq, and caller must check that
408 * queue is eligible for reassembly i.e. it is not COMPLETE,
409 * the last and the first frames arrived and all the bits are here.
410 */
f61944ef 411static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
1da177e4
LT
412 struct net_device *dev)
413{
2bad35b7 414 struct net *net = container_of(fq->q.net, struct net, ipv6.frags);
5ab11c98 415 struct sk_buff *fp, *head = fq->q.fragments;
1da177e4
LT
416 int payload_len;
417 unsigned int nhoff;
ec16439e 418 int sum_truesize;
1da177e4
LT
419
420 fq_kill(fq);
421
f61944ef
HX
422 /* Make the one we just received the head. */
423 if (prev) {
424 head = prev->next;
425 fp = skb_clone(head, GFP_ATOMIC);
426
427 if (!fp)
428 goto out_oom;
429
430 fp->next = head->next;
d6bebca9
CG
431 if (!fp->next)
432 fq->q.fragments_tail = fp;
f61944ef
HX
433 prev->next = fp;
434
5ab11c98
PE
435 skb_morph(head, fq->q.fragments);
436 head->next = fq->q.fragments->next;
f61944ef 437
808db80a 438 consume_skb(fq->q.fragments);
5ab11c98 439 fq->q.fragments = head;
f61944ef
HX
440 }
441
547b792c
IJ
442 WARN_ON(head == NULL);
443 WARN_ON(FRAG6_CB(head)->offset != 0);
1da177e4
LT
444
445 /* Unfragmented part is taken from the first segment. */
d56f90a7 446 payload_len = ((head->data - skb_network_header(head)) -
5ab11c98 447 sizeof(struct ipv6hdr) + fq->q.len -
d56f90a7 448 sizeof(struct frag_hdr));
1da177e4
LT
449 if (payload_len > IPV6_MAXPLEN)
450 goto out_oversize;
451
452 /* Head of list must not be cloned. */
453 if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC))
454 goto out_oom;
455
456 /* If the first fragment is fragmented itself, we split
457 * it to two chunks: the first with data and paged part
458 * and the second, holding only fragments. */
21dc3301 459 if (skb_has_frag_list(head)) {
1da177e4
LT
460 struct sk_buff *clone;
461 int i, plen = 0;
462
463 if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
464 goto out_oom;
465 clone->next = head->next;
466 head->next = clone;
467 skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
4d9092bb 468 skb_frag_list_init(head);
9e903e08
ED
469 for (i = 0; i < skb_shinfo(head)->nr_frags; i++)
470 plen += skb_frag_size(&skb_shinfo(head)->frags[i]);
1da177e4
LT
471 clone->len = clone->data_len = head->data_len - plen;
472 head->data_len -= clone->len;
473 head->len -= clone->len;
474 clone->csum = 0;
475 clone->ip_summed = head->ip_summed;
6ddc0822 476 atomic_add(clone->truesize, &fq->q.net->mem);
1da177e4
LT
477 }
478
479 /* We have to remove fragment header from datagram and to relocate
480 * header in order to calculate ICV correctly. */
481 nhoff = fq->nhoffset;
b0e380b1 482 skb_network_header(head)[nhoff] = skb_transport_header(head)[0];
1ab1457c 483 memmove(head->head + sizeof(struct frag_hdr), head->head,
1da177e4 484 (head->data - head->head) - sizeof(struct frag_hdr));
b0e380b1
ACM
485 head->mac_header += sizeof(struct frag_hdr);
486 head->network_header += sizeof(struct frag_hdr);
1da177e4 487
badff6d0 488 skb_reset_transport_header(head);
d56f90a7 489 skb_push(head, head->data - skb_network_header(head));
1da177e4 490
ec16439e
ED
491 sum_truesize = head->truesize;
492 for (fp = head->next; fp;) {
493 bool headstolen;
494 int delta;
495 struct sk_buff *next = fp->next;
496
497 sum_truesize += fp->truesize;
1da177e4
LT
498 if (head->ip_summed != fp->ip_summed)
499 head->ip_summed = CHECKSUM_NONE;
84fa7933 500 else if (head->ip_summed == CHECKSUM_COMPLETE)
1da177e4 501 head->csum = csum_add(head->csum, fp->csum);
ec16439e
ED
502
503 if (skb_try_coalesce(head, fp, &headstolen, &delta)) {
504 kfree_skb_partial(fp, headstolen);
505 } else {
506 if (!skb_shinfo(head)->frag_list)
507 skb_shinfo(head)->frag_list = fp;
508 head->data_len += fp->len;
509 head->len += fp->len;
510 head->truesize += fp->truesize;
511 }
512 fp = next;
1da177e4 513 }
ec16439e 514 atomic_sub(sum_truesize, &fq->q.net->mem);
1da177e4
LT
515
516 head->next = NULL;
517 head->dev = dev;
5ab11c98 518 head->tstamp = fq->q.stamp;
0660e03f 519 ipv6_hdr(head)->payload_len = htons(payload_len);
951dbc8a 520 IP6CB(head)->nhoff = nhoff;
1da177e4 521
1da177e4 522 /* Yes, and fold redundant checksum back. 8) */
84fa7933 523 if (head->ip_summed == CHECKSUM_COMPLETE)
d56f90a7 524 head->csum = csum_partial(skb_network_header(head),
cfe1fc77 525 skb_network_header_len(head),
d56f90a7 526 head->csum);
1da177e4 527
a11d206d 528 rcu_read_lock();
2bad35b7 529 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
a11d206d 530 rcu_read_unlock();
5ab11c98 531 fq->q.fragments = NULL;
d6bebca9 532 fq->q.fragments_tail = NULL;
1da177e4
LT
533 return 1;
534
535out_oversize:
e87cc472 536 net_dbg_ratelimited("ip6_frag_reasm: payload len = %d\n", payload_len);
1da177e4
LT
537 goto out_fail;
538out_oom:
e87cc472 539 net_dbg_ratelimited("ip6_frag_reasm: no memory for reassembly\n");
1da177e4 540out_fail:
a11d206d 541 rcu_read_lock();
2bad35b7 542 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
a11d206d 543 rcu_read_unlock();
1da177e4
LT
544 return -1;
545}
546
e5bbef20 547static int ipv6_frag_rcv(struct sk_buff *skb)
1da177e4 548{
1da177e4
LT
549 struct frag_hdr *fhdr;
550 struct frag_queue *fq;
b71d1d42 551 const struct ipv6hdr *hdr = ipv6_hdr(skb);
adf30907 552 struct net *net = dev_net(skb_dst(skb)->dev);
1da177e4 553
adf30907 554 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMREQDS);
1da177e4
LT
555
556 /* Jumbo payload inhibits frag. header */
98b3377c
DL
557 if (hdr->payload_len==0)
558 goto fail_hdr;
559
ea2ae17d 560 if (!pskb_may_pull(skb, (skb_transport_offset(skb) +
98b3377c
DL
561 sizeof(struct frag_hdr))))
562 goto fail_hdr;
1da177e4 563
0660e03f 564 hdr = ipv6_hdr(skb);
9c70220b 565 fhdr = (struct frag_hdr *)skb_transport_header(skb);
1da177e4
LT
566
567 if (!(fhdr->frag_off & htons(0xFFF9))) {
568 /* It is not a fragmented frame */
b0e380b1 569 skb->transport_header += sizeof(struct frag_hdr);
483a47d2 570 IP6_INC_STATS_BH(net,
adf30907 571 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMOKS);
1da177e4 572
d56f90a7 573 IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
1da177e4
LT
574 return 1;
575 }
576
e31e0bdc 577 if (atomic_read(&net->ipv6.frags.mem) > net->ipv6.frags.high_thresh)
adf30907 578 ip6_evictor(net, ip6_dst_idev(skb_dst(skb)));
1da177e4 579
9546377c
SW
580 fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr);
581 if (fq != NULL) {
f61944ef 582 int ret;
1da177e4 583
5ab11c98 584 spin_lock(&fq->q.lock);
1da177e4 585
f61944ef 586 ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);
1da177e4 587
5ab11c98 588 spin_unlock(&fq->q.lock);
4b6cb5d8 589 fq_put(fq);
1da177e4
LT
590 return ret;
591 }
592
adf30907 593 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMFAILS);
1da177e4
LT
594 kfree_skb(skb);
595 return -1;
98b3377c
DL
596
597fail_hdr:
adf30907 598 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS);
98b3377c
DL
599 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb));
600 return -1;
1da177e4
LT
601}
602
41135cc8 603static const struct inet6_protocol frag_protocol =
1da177e4
LT
604{
605 .handler = ipv6_frag_rcv,
606 .flags = INET6_PROTO_NOPOLICY,
607};
608
8d8354d2 609#ifdef CONFIG_SYSCTL
0a64b4b8 610static struct ctl_table ip6_frags_ns_ctl_table[] = {
8d8354d2 611 {
8d8354d2 612 .procname = "ip6frag_high_thresh",
e31e0bdc 613 .data = &init_net.ipv6.frags.high_thresh,
8d8354d2
PE
614 .maxlen = sizeof(int),
615 .mode = 0644,
6d9f239a 616 .proc_handler = proc_dointvec
8d8354d2
PE
617 },
618 {
8d8354d2 619 .procname = "ip6frag_low_thresh",
e31e0bdc 620 .data = &init_net.ipv6.frags.low_thresh,
8d8354d2
PE
621 .maxlen = sizeof(int),
622 .mode = 0644,
6d9f239a 623 .proc_handler = proc_dointvec
8d8354d2
PE
624 },
625 {
8d8354d2 626 .procname = "ip6frag_time",
b2fd5321 627 .data = &init_net.ipv6.frags.timeout,
8d8354d2
PE
628 .maxlen = sizeof(int),
629 .mode = 0644,
6d9f239a 630 .proc_handler = proc_dointvec_jiffies,
8d8354d2 631 },
7d291ebb
PE
632 { }
633};
634
635static struct ctl_table ip6_frags_ctl_table[] = {
8d8354d2 636 {
8d8354d2 637 .procname = "ip6frag_secret_interval",
3b4bc4a2 638 .data = &ip6_frags.secret_interval,
8d8354d2
PE
639 .maxlen = sizeof(int),
640 .mode = 0644,
6d9f239a 641 .proc_handler = proc_dointvec_jiffies,
8d8354d2
PE
642 },
643 { }
644};
645
2c8c1e72 646static int __net_init ip6_frags_ns_sysctl_register(struct net *net)
8d8354d2 647{
e4a2d5c2 648 struct ctl_table *table;
8d8354d2
PE
649 struct ctl_table_header *hdr;
650
0a64b4b8 651 table = ip6_frags_ns_ctl_table;
09ad9bc7 652 if (!net_eq(net, &init_net)) {
0a64b4b8 653 table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL);
e4a2d5c2
PE
654 if (table == NULL)
655 goto err_alloc;
656
e31e0bdc
PE
657 table[0].data = &net->ipv6.frags.high_thresh;
658 table[1].data = &net->ipv6.frags.low_thresh;
b2fd5321 659 table[2].data = &net->ipv6.frags.timeout;
e4a2d5c2
PE
660 }
661
ec8f23ce 662 hdr = register_net_sysctl(net, "net/ipv6", table);
e4a2d5c2
PE
663 if (hdr == NULL)
664 goto err_reg;
665
666 net->ipv6.sysctl.frags_hdr = hdr;
667 return 0;
668
669err_reg:
09ad9bc7 670 if (!net_eq(net, &init_net))
e4a2d5c2
PE
671 kfree(table);
672err_alloc:
673 return -ENOMEM;
674}
675
2c8c1e72 676static void __net_exit ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
677{
678 struct ctl_table *table;
679
680 table = net->ipv6.sysctl.frags_hdr->ctl_table_arg;
681 unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr);
3705e11a
YH
682 if (!net_eq(net, &init_net))
683 kfree(table);
8d8354d2 684}
7d291ebb
PE
685
686static struct ctl_table_header *ip6_ctl_header;
687
688static int ip6_frags_sysctl_register(void)
689{
43444757 690 ip6_ctl_header = register_net_sysctl(&init_net, "net/ipv6",
7d291ebb
PE
691 ip6_frags_ctl_table);
692 return ip6_ctl_header == NULL ? -ENOMEM : 0;
693}
694
695static void ip6_frags_sysctl_unregister(void)
696{
697 unregister_net_sysctl_table(ip6_ctl_header);
698}
8d8354d2 699#else
0a64b4b8 700static inline int ip6_frags_ns_sysctl_register(struct net *net)
e71e0349 701{
8d8354d2
PE
702 return 0;
703}
e4a2d5c2 704
0a64b4b8 705static inline void ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
706{
707}
7d291ebb
PE
708
709static inline int ip6_frags_sysctl_register(void)
710{
711 return 0;
712}
713
714static inline void ip6_frags_sysctl_unregister(void)
715{
716}
8d8354d2 717#endif
7d460db9 718
2c8c1e72 719static int __net_init ipv6_frags_init_net(struct net *net)
8d8354d2 720{
7c070aa9
SW
721 net->ipv6.frags.high_thresh = IPV6_FRAG_HIGH_THRESH;
722 net->ipv6.frags.low_thresh = IPV6_FRAG_LOW_THRESH;
b2fd5321 723 net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT;
8d8354d2 724
e5a2bb84
PE
725 inet_frags_init_net(&net->ipv6.frags);
726
0a64b4b8 727 return ip6_frags_ns_sysctl_register(net);
e71e0349
DL
728}
729
2c8c1e72 730static void __net_exit ipv6_frags_exit_net(struct net *net)
81566e83 731{
0a64b4b8 732 ip6_frags_ns_sysctl_unregister(net);
81566e83
PE
733 inet_frags_exit_net(&net->ipv6.frags, &ip6_frags);
734}
735
736static struct pernet_operations ip6_frags_ops = {
737 .init = ipv6_frags_init_net,
738 .exit = ipv6_frags_exit_net,
739};
740
853cbbaa 741int __init ipv6_frag_init(void)
1da177e4 742{
853cbbaa 743 int ret;
1da177e4 744
853cbbaa
DL
745 ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT);
746 if (ret)
747 goto out;
e71e0349 748
7d291ebb
PE
749 ret = ip6_frags_sysctl_register();
750 if (ret)
751 goto err_sysctl;
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:
7d291ebb
PE
770 ip6_frags_sysctl_unregister();
771err_sysctl:
0002c630
PE
772 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
773 goto out;
853cbbaa
DL
774}
775
776void ipv6_frag_exit(void)
777{
778 inet_frags_fini(&ip6_frags);
7d291ebb 779 ip6_frags_sysctl_unregister();
81566e83 780 unregister_pernet_subsys(&ip6_frags_ops);
853cbbaa 781 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
1da177e4 782}