Merge branches 'topic/asoc', 'topic/misc-fixes' and 'topic/hda' into for-linus
[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>
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 */
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 83static struct inet_frags ip6_frags;
1da177e4 84
e5a2bb84 85int ip6_frag_nqueues(struct net *net)
7eb95156 86{
e5a2bb84 87 return net->ipv6.frags.nqueues;
7eb95156 88}
1da177e4 89
6ddc0822 90int ip6_frag_mem(struct net *net)
7eb95156 91{
6ddc0822 92 return atomic_read(&net->ipv6.frags.mem);
7eb95156 93}
1da177e4 94
f61944ef
HX
95static 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
102unsigned 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 128EXPORT_SYMBOL_GPL(inet6_hash_frag);
1da177e4 129
321a3a99 130static 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
138int 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}
148EXPORT_SYMBOL(ip6_frag_match);
149
1da177e4 150/* Memory Tracking Functions. */
6ddc0822
PE
151static 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 160void 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 169EXPORT_SYMBOL(ip6_frag_init);
1da177e4 170
1da177e4
LT
171/* Destruction primitives. */
172
4b6cb5d8 173static __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 */
181static __inline__ void fq_kill(struct frag_queue *fq)
182{
277e650d 183 inet_frag_kill(&fq->q, &ip6_frags);
1da177e4
LT
184}
185
6ddc0822 186static 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
195static 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 231out:
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 238static __inline__ struct frag_queue *
ac18e750 239fq_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
259oom:
483a47d2 260 IP6_INC_STATS_BH(net, idev, IPSTATS_MIB_REASMFAILS);
1da177e4
LT
261 return NULL;
262}
263
f61944ef 264static 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;
483a47d2 270 struct net *net = dev_net(skb->dst->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) {
483a47d2 280 IP6_INC_STATS_BH(net, ip6_dst_idev(skb->dst),
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 */
483a47d2 313 IP6_INC_STATS_BH(net, ip6_dst_idev(skb->dst),
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
436err:
483a47d2 437 IP6_INC_STATS(net, ip6_dst_idev(skb->dst),
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 452static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
1da177e4
LT
453 struct net_device *dev)
454{
5ab11c98 455 struct sk_buff *fp, *head = fq->q.fragments;
1da177e4
LT
456 int payload_len;
457 unsigned int nhoff;
458
459 fq_kill(fq);
460
f61944ef
HX
461 /* Make the one we just received the head. */
462 if (prev) {
463 head = prev->next;
464 fp = skb_clone(head, GFP_ATOMIC);
465
466 if (!fp)
467 goto out_oom;
468
469 fp->next = head->next;
470 prev->next = fp;
471
5ab11c98
PE
472 skb_morph(head, fq->q.fragments);
473 head->next = fq->q.fragments->next;
f61944ef 474
5ab11c98
PE
475 kfree_skb(fq->q.fragments);
476 fq->q.fragments = head;
f61944ef
HX
477 }
478
547b792c
IJ
479 WARN_ON(head == NULL);
480 WARN_ON(FRAG6_CB(head)->offset != 0);
1da177e4
LT
481
482 /* Unfragmented part is taken from the first segment. */
d56f90a7 483 payload_len = ((head->data - skb_network_header(head)) -
5ab11c98 484 sizeof(struct ipv6hdr) + fq->q.len -
d56f90a7 485 sizeof(struct frag_hdr));
1da177e4
LT
486 if (payload_len > IPV6_MAXPLEN)
487 goto out_oversize;
488
489 /* Head of list must not be cloned. */
490 if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC))
491 goto out_oom;
492
493 /* If the first fragment is fragmented itself, we split
494 * it to two chunks: the first with data and paged part
495 * and the second, holding only fragments. */
496 if (skb_shinfo(head)->frag_list) {
497 struct sk_buff *clone;
498 int i, plen = 0;
499
500 if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
501 goto out_oom;
502 clone->next = head->next;
503 head->next = clone;
504 skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
505 skb_shinfo(head)->frag_list = NULL;
506 for (i=0; i<skb_shinfo(head)->nr_frags; i++)
507 plen += skb_shinfo(head)->frags[i].size;
508 clone->len = clone->data_len = head->data_len - plen;
509 head->data_len -= clone->len;
510 head->len -= clone->len;
511 clone->csum = 0;
512 clone->ip_summed = head->ip_summed;
6ddc0822 513 atomic_add(clone->truesize, &fq->q.net->mem);
1da177e4
LT
514 }
515
516 /* We have to remove fragment header from datagram and to relocate
517 * header in order to calculate ICV correctly. */
518 nhoff = fq->nhoffset;
b0e380b1 519 skb_network_header(head)[nhoff] = skb_transport_header(head)[0];
1ab1457c 520 memmove(head->head + sizeof(struct frag_hdr), head->head,
1da177e4 521 (head->data - head->head) - sizeof(struct frag_hdr));
b0e380b1
ACM
522 head->mac_header += sizeof(struct frag_hdr);
523 head->network_header += sizeof(struct frag_hdr);
1da177e4
LT
524
525 skb_shinfo(head)->frag_list = head->next;
badff6d0 526 skb_reset_transport_header(head);
d56f90a7 527 skb_push(head, head->data - skb_network_header(head));
6ddc0822 528 atomic_sub(head->truesize, &fq->q.net->mem);
1da177e4
LT
529
530 for (fp=head->next; fp; fp = fp->next) {
531 head->data_len += fp->len;
532 head->len += fp->len;
533 if (head->ip_summed != fp->ip_summed)
534 head->ip_summed = CHECKSUM_NONE;
84fa7933 535 else if (head->ip_summed == CHECKSUM_COMPLETE)
1da177e4
LT
536 head->csum = csum_add(head->csum, fp->csum);
537 head->truesize += fp->truesize;
6ddc0822 538 atomic_sub(fp->truesize, &fq->q.net->mem);
1da177e4
LT
539 }
540
541 head->next = NULL;
542 head->dev = dev;
5ab11c98 543 head->tstamp = fq->q.stamp;
0660e03f 544 ipv6_hdr(head)->payload_len = htons(payload_len);
951dbc8a 545 IP6CB(head)->nhoff = nhoff;
1da177e4 546
1da177e4 547 /* Yes, and fold redundant checksum back. 8) */
84fa7933 548 if (head->ip_summed == CHECKSUM_COMPLETE)
d56f90a7 549 head->csum = csum_partial(skb_network_header(head),
cfe1fc77 550 skb_network_header_len(head),
d56f90a7 551 head->csum);
1da177e4 552
a11d206d 553 rcu_read_lock();
483a47d2
DL
554 IP6_INC_STATS_BH(dev_net(dev),
555 __in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
a11d206d 556 rcu_read_unlock();
5ab11c98 557 fq->q.fragments = NULL;
1da177e4
LT
558 return 1;
559
560out_oversize:
561 if (net_ratelimit())
562 printk(KERN_DEBUG "ip6_frag_reasm: payload len = %d\n", payload_len);
563 goto out_fail;
564out_oom:
565 if (net_ratelimit())
566 printk(KERN_DEBUG "ip6_frag_reasm: no memory for reassembly\n");
567out_fail:
a11d206d 568 rcu_read_lock();
483a47d2
DL
569 IP6_INC_STATS_BH(dev_net(dev),
570 __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
a11d206d 571 rcu_read_unlock();
1da177e4
LT
572 return -1;
573}
574
e5bbef20 575static int ipv6_frag_rcv(struct sk_buff *skb)
1da177e4 576{
1da177e4
LT
577 struct frag_hdr *fhdr;
578 struct frag_queue *fq;
0660e03f 579 struct ipv6hdr *hdr = ipv6_hdr(skb);
3bd653c8 580 struct net *net = dev_net(skb->dst->dev);
1da177e4 581
483a47d2 582 IP6_INC_STATS_BH(net, ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMREQDS);
1da177e4
LT
583
584 /* Jumbo payload inhibits frag. header */
98b3377c
DL
585 if (hdr->payload_len==0)
586 goto fail_hdr;
587
ea2ae17d 588 if (!pskb_may_pull(skb, (skb_transport_offset(skb) +
98b3377c
DL
589 sizeof(struct frag_hdr))))
590 goto fail_hdr;
1da177e4 591
0660e03f 592 hdr = ipv6_hdr(skb);
9c70220b 593 fhdr = (struct frag_hdr *)skb_transport_header(skb);
1da177e4
LT
594
595 if (!(fhdr->frag_off & htons(0xFFF9))) {
596 /* It is not a fragmented frame */
b0e380b1 597 skb->transport_header += sizeof(struct frag_hdr);
483a47d2
DL
598 IP6_INC_STATS_BH(net,
599 ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMOKS);
1da177e4 600
d56f90a7 601 IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
1da177e4
LT
602 return 1;
603 }
604
e31e0bdc 605 if (atomic_read(&net->ipv6.frags.mem) > net->ipv6.frags.high_thresh)
6ddc0822 606 ip6_evictor(net, ip6_dst_idev(skb->dst));
1da177e4 607
ac18e750 608 if ((fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr,
a11d206d 609 ip6_dst_idev(skb->dst))) != NULL) {
f61944ef 610 int ret;
1da177e4 611
5ab11c98 612 spin_lock(&fq->q.lock);
1da177e4 613
f61944ef 614 ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);
1da177e4 615
5ab11c98 616 spin_unlock(&fq->q.lock);
4b6cb5d8 617 fq_put(fq);
1da177e4
LT
618 return ret;
619 }
620
483a47d2 621 IP6_INC_STATS_BH(net, ip6_dst_idev(skb->dst), IPSTATS_MIB_REASMFAILS);
1da177e4
LT
622 kfree_skb(skb);
623 return -1;
98b3377c
DL
624
625fail_hdr:
3bd653c8 626 IP6_INC_STATS(net, ip6_dst_idev(skb->dst), IPSTATS_MIB_INHDRERRORS);
98b3377c
DL
627 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb));
628 return -1;
1da177e4
LT
629}
630
631static struct inet6_protocol frag_protocol =
632{
633 .handler = ipv6_frag_rcv,
634 .flags = INET6_PROTO_NOPOLICY,
635};
636
8d8354d2 637#ifdef CONFIG_SYSCTL
0a64b4b8 638static struct ctl_table ip6_frags_ns_ctl_table[] = {
8d8354d2
PE
639 {
640 .ctl_name = NET_IPV6_IP6FRAG_HIGH_THRESH,
641 .procname = "ip6frag_high_thresh",
e31e0bdc 642 .data = &init_net.ipv6.frags.high_thresh,
8d8354d2
PE
643 .maxlen = sizeof(int),
644 .mode = 0644,
645 .proc_handler = &proc_dointvec
646 },
647 {
648 .ctl_name = NET_IPV6_IP6FRAG_LOW_THRESH,
649 .procname = "ip6frag_low_thresh",
e31e0bdc 650 .data = &init_net.ipv6.frags.low_thresh,
8d8354d2
PE
651 .maxlen = sizeof(int),
652 .mode = 0644,
653 .proc_handler = &proc_dointvec
654 },
655 {
656 .ctl_name = NET_IPV6_IP6FRAG_TIME,
657 .procname = "ip6frag_time",
b2fd5321 658 .data = &init_net.ipv6.frags.timeout,
8d8354d2
PE
659 .maxlen = sizeof(int),
660 .mode = 0644,
661 .proc_handler = &proc_dointvec_jiffies,
662 .strategy = &sysctl_jiffies,
663 },
7d291ebb
PE
664 { }
665};
666
667static struct ctl_table ip6_frags_ctl_table[] = {
8d8354d2
PE
668 {
669 .ctl_name = NET_IPV6_IP6FRAG_SECRET_INTERVAL,
670 .procname = "ip6frag_secret_interval",
3b4bc4a2 671 .data = &ip6_frags.secret_interval,
8d8354d2
PE
672 .maxlen = sizeof(int),
673 .mode = 0644,
674 .proc_handler = &proc_dointvec_jiffies,
675 .strategy = &sysctl_jiffies
676 },
677 { }
678};
679
0a64b4b8 680static int ip6_frags_ns_sysctl_register(struct net *net)
8d8354d2 681{
e4a2d5c2 682 struct ctl_table *table;
8d8354d2
PE
683 struct ctl_table_header *hdr;
684
0a64b4b8 685 table = ip6_frags_ns_ctl_table;
e4a2d5c2 686 if (net != &init_net) {
0a64b4b8 687 table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL);
e4a2d5c2
PE
688 if (table == NULL)
689 goto err_alloc;
690
e31e0bdc
PE
691 table[0].data = &net->ipv6.frags.high_thresh;
692 table[1].data = &net->ipv6.frags.low_thresh;
b2fd5321 693 table[2].data = &net->ipv6.frags.timeout;
e4a2d5c2
PE
694 }
695
696 hdr = register_net_sysctl_table(net, net_ipv6_ctl_path, table);
697 if (hdr == NULL)
698 goto err_reg;
699
700 net->ipv6.sysctl.frags_hdr = hdr;
701 return 0;
702
703err_reg:
704 if (net != &init_net)
705 kfree(table);
706err_alloc:
707 return -ENOMEM;
708}
709
0a64b4b8 710static void ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
711{
712 struct ctl_table *table;
713
714 table = net->ipv6.sysctl.frags_hdr->ctl_table_arg;
715 unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr);
716 kfree(table);
8d8354d2 717}
7d291ebb
PE
718
719static struct ctl_table_header *ip6_ctl_header;
720
721static int ip6_frags_sysctl_register(void)
722{
723 ip6_ctl_header = register_net_sysctl_rotable(net_ipv6_ctl_path,
724 ip6_frags_ctl_table);
725 return ip6_ctl_header == NULL ? -ENOMEM : 0;
726}
727
728static void ip6_frags_sysctl_unregister(void)
729{
730 unregister_net_sysctl_table(ip6_ctl_header);
731}
8d8354d2 732#else
0a64b4b8 733static inline int ip6_frags_ns_sysctl_register(struct net *net)
e71e0349 734{
8d8354d2
PE
735 return 0;
736}
e4a2d5c2 737
0a64b4b8 738static inline void ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
739{
740}
7d291ebb
PE
741
742static inline int ip6_frags_sysctl_register(void)
743{
744 return 0;
745}
746
747static inline void ip6_frags_sysctl_unregister(void)
748{
749}
8d8354d2 750#endif
7d460db9 751
8d8354d2
PE
752static int ipv6_frags_init_net(struct net *net)
753{
e31e0bdc
PE
754 net->ipv6.frags.high_thresh = 256 * 1024;
755 net->ipv6.frags.low_thresh = 192 * 1024;
b2fd5321 756 net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT;
8d8354d2 757
e5a2bb84
PE
758 inet_frags_init_net(&net->ipv6.frags);
759
0a64b4b8 760 return ip6_frags_ns_sysctl_register(net);
e71e0349
DL
761}
762
81566e83
PE
763static void ipv6_frags_exit_net(struct net *net)
764{
0a64b4b8 765 ip6_frags_ns_sysctl_unregister(net);
81566e83
PE
766 inet_frags_exit_net(&net->ipv6.frags, &ip6_frags);
767}
768
769static struct pernet_operations ip6_frags_ops = {
770 .init = ipv6_frags_init_net,
771 .exit = ipv6_frags_exit_net,
772};
773
853cbbaa 774int __init ipv6_frag_init(void)
1da177e4 775{
853cbbaa 776 int ret;
1da177e4 777
853cbbaa
DL
778 ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT);
779 if (ret)
780 goto out;
e71e0349 781
7d291ebb
PE
782 ret = ip6_frags_sysctl_register();
783 if (ret)
784 goto err_sysctl;
785
0002c630
PE
786 ret = register_pernet_subsys(&ip6_frags_ops);
787 if (ret)
788 goto err_pernet;
8d8354d2 789
321a3a99 790 ip6_frags.hashfn = ip6_hashfn;
c6fda282 791 ip6_frags.constructor = ip6_frag_init;
c9547709 792 ip6_frags.destructor = NULL;
1e4b8287
PE
793 ip6_frags.skb_free = NULL;
794 ip6_frags.qsize = sizeof(struct frag_queue);
abd6523d 795 ip6_frags.match = ip6_frag_match;
e521db9d 796 ip6_frags.frag_expire = ip6_frag_expire;
3b4bc4a2 797 ip6_frags.secret_interval = 10 * 60 * HZ;
7eb95156 798 inet_frags_init(&ip6_frags);
853cbbaa
DL
799out:
800 return ret;
0002c630
PE
801
802err_pernet:
7d291ebb
PE
803 ip6_frags_sysctl_unregister();
804err_sysctl:
0002c630
PE
805 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
806 goto out;
853cbbaa
DL
807}
808
809void ipv6_frag_exit(void)
810{
811 inet_frags_fini(&ip6_frags);
7d291ebb 812 ip6_frags_sysctl_unregister();
81566e83 813 unregister_pernet_subsys(&ip6_frags_ops);
853cbbaa 814 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
1da177e4 815}