inet: frag: remove lru list
[linux-2.6-block.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 */
5a3da1fe
HFS
29
30#define pr_fmt(fmt) "IPv6: " fmt
31
1da177e4
LT
32#include <linux/errno.h>
33#include <linux/types.h>
34#include <linux/string.h>
35#include <linux/socket.h>
36#include <linux/sockios.h>
37#include <linux/jiffies.h>
38#include <linux/net.h>
39#include <linux/list.h>
40#include <linux/netdevice.h>
41#include <linux/in6.h>
42#include <linux/ipv6.h>
43#include <linux/icmpv6.h>
44#include <linux/random.h>
45#include <linux/jhash.h>
f61944ef 46#include <linux/skbuff.h>
5a0e3ad6 47#include <linux/slab.h>
bc3b2d7f 48#include <linux/export.h>
1da177e4
LT
49
50#include <net/sock.h>
51#include <net/snmp.h>
52
53#include <net/ipv6.h>
a11d206d 54#include <net/ip6_route.h>
1da177e4
LT
55#include <net/protocol.h>
56#include <net/transp_v6.h>
57#include <net/rawv6.h>
58#include <net/ndisc.h>
59#include <net/addrconf.h>
5ab11c98 60#include <net/inet_frag.h>
eec2e618 61#include <net/inet_ecn.h>
1da177e4 62
1da177e4
LT
63struct ip6frag_skb_cb
64{
65 struct inet6_skb_parm h;
66 int offset;
67};
68
69#define FRAG6_CB(skb) ((struct ip6frag_skb_cb*)((skb)->cb))
70
eec2e618
HFS
71static inline u8 ip6_frag_ecn(const struct ipv6hdr *ipv6h)
72{
73 return 1 << (ipv6_get_dsfield(ipv6h) & INET_ECN_MASK);
74}
1da177e4 75
7eb95156 76static struct inet_frags ip6_frags;
1da177e4 77
f61944ef
HX
78static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
79 struct net_device *dev);
80
f6596f9d
ZB
81/*
82 * callers should be careful not to use the hash value outside the ipfrag_lock
83 * as doing so could race with ipfrag_hash_rnd being recalculated.
84 */
b1190570
HFS
85static unsigned int inet6_hash_frag(__be32 id, const struct in6_addr *saddr,
86 const struct in6_addr *daddr)
1da177e4 87{
b1190570 88 net_get_random_once(&ip6_frags.rnd, sizeof(ip6_frags.rnd));
fb3cfe6e
FW
89 return jhash_3words(ipv6_addr_hash(saddr), ipv6_addr_hash(daddr),
90 (__force u32)id, ip6_frags.rnd);
1da177e4
LT
91}
92
36c77782 93static unsigned int ip6_hashfn(const struct inet_frag_queue *q)
1da177e4 94{
36c77782 95 const struct frag_queue *fq;
1da177e4 96
321a3a99 97 fq = container_of(q, struct frag_queue, q);
b1190570 98 return inet6_hash_frag(fq->id, &fq->saddr, &fq->daddr);
1da177e4
LT
99}
100
36c77782 101bool ip6_frag_match(const struct inet_frag_queue *q, const void *a)
abd6523d 102{
36c77782
FW
103 const struct frag_queue *fq;
104 const struct ip6_create_arg *arg = a;
abd6523d
PE
105
106 fq = container_of(q, struct frag_queue, q);
cbc264ca
ED
107 return fq->id == arg->id &&
108 fq->user == arg->user &&
109 ipv6_addr_equal(&fq->saddr, arg->src) &&
110 ipv6_addr_equal(&fq->daddr, arg->dst);
abd6523d
PE
111}
112EXPORT_SYMBOL(ip6_frag_match);
113
36c77782 114void ip6_frag_init(struct inet_frag_queue *q, const void *a)
1da177e4 115{
c6fda282 116 struct frag_queue *fq = container_of(q, struct frag_queue, q);
36c77782 117 const struct ip6_create_arg *arg = a;
c6fda282
PE
118
119 fq->id = arg->id;
0b5ccb2e 120 fq->user = arg->user;
4e3fd7a0
AD
121 fq->saddr = *arg->src;
122 fq->daddr = *arg->dst;
eec2e618 123 fq->ecn = arg->ecn;
1da177e4 124}
c6fda282 125EXPORT_SYMBOL(ip6_frag_init);
1da177e4 126
b836c99f
AW
127void ip6_expire_frag_queue(struct net *net, struct frag_queue *fq,
128 struct inet_frags *frags)
1da177e4 129{
a11d206d 130 struct net_device *dev = NULL;
e521db9d 131
5ab11c98 132 spin_lock(&fq->q.lock);
1da177e4 133
bc578a54 134 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
135 goto out;
136
b836c99f 137 inet_frag_kill(&fq->q, frags);
1da177e4 138
69df9d59
ED
139 rcu_read_lock();
140 dev = dev_get_by_index_rcu(net, fq->iif);
a11d206d 141 if (!dev)
69df9d59 142 goto out_rcu_unlock;
a11d206d 143
b13d3cbf
FW
144 if (!(fq->q.last_in & INET_FRAG_EVICTED))
145 IP6_INC_STATS_BH(net, __in6_dev_get(dev),
146 IPSTATS_MIB_REASMTIMEOUT);
483a47d2 147 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
1da177e4 148
78c784c4 149 /* Don't send error if the first segment did not arrive. */
bc578a54 150 if (!(fq->q.last_in & INET_FRAG_FIRST_IN) || !fq->q.fragments)
69df9d59 151 goto out_rcu_unlock;
78c784c4 152
78c784c4
IO
153 /*
154 But use as source device on which LAST ARRIVED
155 segment was received. And do not use fq->dev
156 pointer directly, device might already disappeared.
157 */
5ab11c98 158 fq->q.fragments->dev = dev;
3ffe533c 159 icmpv6_send(fq->q.fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0);
69df9d59
ED
160out_rcu_unlock:
161 rcu_read_unlock();
1da177e4 162out:
5ab11c98 163 spin_unlock(&fq->q.lock);
b836c99f
AW
164 inet_frag_put(&fq->q, frags);
165}
166EXPORT_SYMBOL(ip6_expire_frag_queue);
167
168static void ip6_frag_expire(unsigned long data)
169{
170 struct frag_queue *fq;
171 struct net *net;
172
173 fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q);
174 net = container_of(fq->q.net, struct net, ipv6.frags);
175
176 ip6_expire_frag_queue(net, fq, &ip6_frags);
1da177e4
LT
177}
178
abd6523d 179static __inline__ struct frag_queue *
eec2e618
HFS
180fq_find(struct net *net, __be32 id, const struct in6_addr *src,
181 const struct in6_addr *dst, u8 ecn)
1da177e4 182{
c6fda282
PE
183 struct inet_frag_queue *q;
184 struct ip6_create_arg arg;
abd6523d 185 unsigned int hash;
1da177e4 186
c6fda282 187 arg.id = id;
0b5ccb2e 188 arg.user = IP6_DEFRAG_LOCAL_DELIVER;
c6fda282
PE
189 arg.src = src;
190 arg.dst = dst;
eec2e618 191 arg.ecn = ecn;
9a375803
PE
192
193 read_lock(&ip6_frags.lock);
b1190570 194 hash = inet6_hash_frag(id, src, dst);
1da177e4 195
ac18e750 196 q = inet_frag_find(&net->ipv6.frags, &ip6_frags, &arg, hash);
5a3da1fe
HFS
197 if (IS_ERR_OR_NULL(q)) {
198 inet_frag_maybe_warn_overflow(q, pr_fmt());
9546377c 199 return NULL;
5a3da1fe 200 }
c6fda282 201 return container_of(q, struct frag_queue, q);
1da177e4
LT
202}
203
f61944ef 204static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
1da177e4
LT
205 struct frag_hdr *fhdr, int nhoff)
206{
207 struct sk_buff *prev, *next;
f61944ef 208 struct net_device *dev;
1da177e4 209 int offset, end;
adf30907 210 struct net *net = dev_net(skb_dst(skb)->dev);
eec2e618 211 u8 ecn;
1da177e4 212
bc578a54 213 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
214 goto err;
215
216 offset = ntohs(fhdr->frag_off) & ~0x7;
0660e03f
ACM
217 end = offset + (ntohs(ipv6_hdr(skb)->payload_len) -
218 ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1)));
1da177e4
LT
219
220 if ((unsigned int)end > IPV6_MAXPLEN) {
adf30907 221 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
a11d206d 222 IPSTATS_MIB_INHDRERRORS);
d56f90a7
ACM
223 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
224 ((u8 *)&fhdr->frag_off -
225 skb_network_header(skb)));
f61944ef 226 return -1;
1da177e4
LT
227 }
228
eec2e618
HFS
229 ecn = ip6_frag_ecn(ipv6_hdr(skb));
230
d56f90a7
ACM
231 if (skb->ip_summed == CHECKSUM_COMPLETE) {
232 const unsigned char *nh = skb_network_header(skb);
1ab1457c 233 skb->csum = csum_sub(skb->csum,
d56f90a7
ACM
234 csum_partial(nh, (u8 *)(fhdr + 1) - nh,
235 0));
236 }
1da177e4
LT
237
238 /* Is this the final fragment? */
239 if (!(fhdr->frag_off & htons(IP6_MF))) {
240 /* If we already have some bits beyond end
241 * or have different end, the segment is corrupted.
242 */
5ab11c98 243 if (end < fq->q.len ||
bc578a54 244 ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len))
1da177e4 245 goto err;
bc578a54 246 fq->q.last_in |= INET_FRAG_LAST_IN;
5ab11c98 247 fq->q.len = end;
1da177e4
LT
248 } else {
249 /* Check if the fragment is rounded to 8 bytes.
250 * Required by the RFC.
251 */
252 if (end & 0x7) {
253 /* RFC2460 says always send parameter problem in
254 * this case. -DaveM
255 */
adf30907 256 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
a11d206d 257 IPSTATS_MIB_INHDRERRORS);
1ab1457c 258 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
1da177e4 259 offsetof(struct ipv6hdr, payload_len));
f61944ef 260 return -1;
1da177e4 261 }
5ab11c98 262 if (end > fq->q.len) {
1da177e4 263 /* Some bits beyond end -> corruption. */
bc578a54 264 if (fq->q.last_in & INET_FRAG_LAST_IN)
1da177e4 265 goto err;
5ab11c98 266 fq->q.len = end;
1da177e4
LT
267 }
268 }
269
270 if (end == offset)
271 goto err;
272
273 /* Point into the IP datagram 'data' part. */
274 if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
275 goto err;
1ab1457c 276
42ca89c1
SH
277 if (pskb_trim_rcsum(skb, end - offset))
278 goto err;
1da177e4
LT
279
280 /* Find out which fragments are in front and at the back of us
281 * in the chain of fragments so far. We must know where to put
282 * this fragment, right?
283 */
d6bebca9
CG
284 prev = fq->q.fragments_tail;
285 if (!prev || FRAG6_CB(prev)->offset < offset) {
286 next = NULL;
287 goto found;
288 }
1da177e4 289 prev = NULL;
5ab11c98 290 for(next = fq->q.fragments; next != NULL; next = next->next) {
1da177e4
LT
291 if (FRAG6_CB(next)->offset >= offset)
292 break; /* bingo! */
293 prev = next;
294 }
295
d6bebca9 296found:
5de658f8
ED
297 /* RFC5722, Section 4, amended by Errata ID : 3089
298 * When reassembling an IPv6 datagram, if
70789d70
ND
299 * one or more its constituent fragments is determined to be an
300 * overlapping fragment, the entire datagram (and any constituent
5de658f8 301 * fragments) MUST be silently discarded.
1da177e4 302 */
1da177e4 303
70789d70
ND
304 /* Check for overlap with preceding fragment. */
305 if (prev &&
f4642141 306 (FRAG6_CB(prev)->offset + prev->len) > offset)
70789d70 307 goto discard_fq;
1da177e4 308
70789d70
ND
309 /* Look for overlap with succeeding segment. */
310 if (next && FRAG6_CB(next)->offset < end)
311 goto discard_fq;
1da177e4
LT
312
313 FRAG6_CB(skb)->offset = offset;
314
315 /* Insert this fragment in the chain of fragments. */
316 skb->next = next;
d6bebca9
CG
317 if (!next)
318 fq->q.fragments_tail = skb;
1da177e4
LT
319 if (prev)
320 prev->next = skb;
321 else
5ab11c98 322 fq->q.fragments = skb;
1da177e4 323
f61944ef
HX
324 dev = skb->dev;
325 if (dev) {
326 fq->iif = dev->ifindex;
327 skb->dev = NULL;
328 }
5ab11c98
PE
329 fq->q.stamp = skb->tstamp;
330 fq->q.meat += skb->len;
eec2e618 331 fq->ecn |= ecn;
d433673e 332 add_frag_mem_limit(&fq->q, skb->truesize);
1da177e4
LT
333
334 /* The first fragment.
335 * nhoffset is obtained from the first fragment, of course.
336 */
337 if (offset == 0) {
338 fq->nhoffset = nhoff;
bc578a54 339 fq->q.last_in |= INET_FRAG_FIRST_IN;
1da177e4 340 }
f61944ef 341
bc578a54 342 if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
97599dc7
ED
343 fq->q.meat == fq->q.len) {
344 int res;
345 unsigned long orefdst = skb->_skb_refdst;
346
347 skb->_skb_refdst = 0UL;
348 res = ip6_frag_reasm(fq, prev, dev);
349 skb->_skb_refdst = orefdst;
350 return res;
351 }
f61944ef 352
97599dc7 353 skb_dst_drop(skb);
f61944ef 354 return -1;
1da177e4 355
70789d70 356discard_fq:
b836c99f 357 inet_frag_kill(&fq->q, &ip6_frags);
1da177e4 358err:
adf30907 359 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
3bd653c8 360 IPSTATS_MIB_REASMFAILS);
1da177e4 361 kfree_skb(skb);
f61944ef 362 return -1;
1da177e4
LT
363}
364
365/*
366 * Check if this packet is complete.
367 * Returns NULL on failure by any reason, and pointer
368 * to current nexthdr field in reassembled frame.
369 *
370 * It is called with locked fq, and caller must check that
371 * queue is eligible for reassembly i.e. it is not COMPLETE,
372 * the last and the first frames arrived and all the bits are here.
373 */
f61944ef 374static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
1da177e4
LT
375 struct net_device *dev)
376{
2bad35b7 377 struct net *net = container_of(fq->q.net, struct net, ipv6.frags);
5ab11c98 378 struct sk_buff *fp, *head = fq->q.fragments;
1da177e4
LT
379 int payload_len;
380 unsigned int nhoff;
ec16439e 381 int sum_truesize;
eec2e618 382 u8 ecn;
1da177e4 383
b836c99f 384 inet_frag_kill(&fq->q, &ip6_frags);
1da177e4 385
eec2e618
HFS
386 ecn = ip_frag_ecn_table[fq->ecn];
387 if (unlikely(ecn == 0xff))
388 goto out_fail;
389
f61944ef
HX
390 /* Make the one we just received the head. */
391 if (prev) {
392 head = prev->next;
393 fp = skb_clone(head, GFP_ATOMIC);
394
395 if (!fp)
396 goto out_oom;
397
398 fp->next = head->next;
d6bebca9
CG
399 if (!fp->next)
400 fq->q.fragments_tail = fp;
f61944ef
HX
401 prev->next = fp;
402
5ab11c98
PE
403 skb_morph(head, fq->q.fragments);
404 head->next = fq->q.fragments->next;
f61944ef 405
808db80a 406 consume_skb(fq->q.fragments);
5ab11c98 407 fq->q.fragments = head;
f61944ef
HX
408 }
409
547b792c
IJ
410 WARN_ON(head == NULL);
411 WARN_ON(FRAG6_CB(head)->offset != 0);
1da177e4
LT
412
413 /* Unfragmented part is taken from the first segment. */
d56f90a7 414 payload_len = ((head->data - skb_network_header(head)) -
5ab11c98 415 sizeof(struct ipv6hdr) + fq->q.len -
d56f90a7 416 sizeof(struct frag_hdr));
1da177e4
LT
417 if (payload_len > IPV6_MAXPLEN)
418 goto out_oversize;
419
420 /* Head of list must not be cloned. */
14bbd6a5 421 if (skb_unclone(head, GFP_ATOMIC))
1da177e4
LT
422 goto out_oom;
423
424 /* If the first fragment is fragmented itself, we split
425 * it to two chunks: the first with data and paged part
426 * and the second, holding only fragments. */
21dc3301 427 if (skb_has_frag_list(head)) {
1da177e4
LT
428 struct sk_buff *clone;
429 int i, plen = 0;
430
431 if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
432 goto out_oom;
433 clone->next = head->next;
434 head->next = clone;
435 skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
4d9092bb 436 skb_frag_list_init(head);
9e903e08
ED
437 for (i = 0; i < skb_shinfo(head)->nr_frags; i++)
438 plen += skb_frag_size(&skb_shinfo(head)->frags[i]);
1da177e4
LT
439 clone->len = clone->data_len = head->data_len - plen;
440 head->data_len -= clone->len;
441 head->len -= clone->len;
442 clone->csum = 0;
443 clone->ip_summed = head->ip_summed;
d433673e 444 add_frag_mem_limit(&fq->q, clone->truesize);
1da177e4
LT
445 }
446
447 /* We have to remove fragment header from datagram and to relocate
448 * header in order to calculate ICV correctly. */
449 nhoff = fq->nhoffset;
b0e380b1 450 skb_network_header(head)[nhoff] = skb_transport_header(head)[0];
1ab1457c 451 memmove(head->head + sizeof(struct frag_hdr), head->head,
1da177e4 452 (head->data - head->head) - sizeof(struct frag_hdr));
b0e380b1
ACM
453 head->mac_header += sizeof(struct frag_hdr);
454 head->network_header += sizeof(struct frag_hdr);
1da177e4 455
badff6d0 456 skb_reset_transport_header(head);
d56f90a7 457 skb_push(head, head->data - skb_network_header(head));
1da177e4 458
ec16439e
ED
459 sum_truesize = head->truesize;
460 for (fp = head->next; fp;) {
461 bool headstolen;
462 int delta;
463 struct sk_buff *next = fp->next;
464
465 sum_truesize += fp->truesize;
1da177e4
LT
466 if (head->ip_summed != fp->ip_summed)
467 head->ip_summed = CHECKSUM_NONE;
84fa7933 468 else if (head->ip_summed == CHECKSUM_COMPLETE)
1da177e4 469 head->csum = csum_add(head->csum, fp->csum);
ec16439e
ED
470
471 if (skb_try_coalesce(head, fp, &headstolen, &delta)) {
472 kfree_skb_partial(fp, headstolen);
473 } else {
474 if (!skb_shinfo(head)->frag_list)
475 skb_shinfo(head)->frag_list = fp;
476 head->data_len += fp->len;
477 head->len += fp->len;
478 head->truesize += fp->truesize;
479 }
480 fp = next;
1da177e4 481 }
d433673e 482 sub_frag_mem_limit(&fq->q, sum_truesize);
1da177e4
LT
483
484 head->next = NULL;
485 head->dev = dev;
5ab11c98 486 head->tstamp = fq->q.stamp;
0660e03f 487 ipv6_hdr(head)->payload_len = htons(payload_len);
eec2e618 488 ipv6_change_dsfield(ipv6_hdr(head), 0xff, ecn);
951dbc8a 489 IP6CB(head)->nhoff = nhoff;
f46078cf 490 IP6CB(head)->flags |= IP6SKB_FRAGMENTED;
1da177e4 491
1da177e4 492 /* Yes, and fold redundant checksum back. 8) */
84fa7933 493 if (head->ip_summed == CHECKSUM_COMPLETE)
d56f90a7 494 head->csum = csum_partial(skb_network_header(head),
cfe1fc77 495 skb_network_header_len(head),
d56f90a7 496 head->csum);
1da177e4 497
a11d206d 498 rcu_read_lock();
2bad35b7 499 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
a11d206d 500 rcu_read_unlock();
5ab11c98 501 fq->q.fragments = NULL;
d6bebca9 502 fq->q.fragments_tail = NULL;
1da177e4
LT
503 return 1;
504
505out_oversize:
e87cc472 506 net_dbg_ratelimited("ip6_frag_reasm: payload len = %d\n", payload_len);
1da177e4
LT
507 goto out_fail;
508out_oom:
e87cc472 509 net_dbg_ratelimited("ip6_frag_reasm: no memory for reassembly\n");
1da177e4 510out_fail:
a11d206d 511 rcu_read_lock();
2bad35b7 512 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
a11d206d 513 rcu_read_unlock();
1da177e4
LT
514 return -1;
515}
516
e5bbef20 517static int ipv6_frag_rcv(struct sk_buff *skb)
1da177e4 518{
1da177e4
LT
519 struct frag_hdr *fhdr;
520 struct frag_queue *fq;
b71d1d42 521 const struct ipv6hdr *hdr = ipv6_hdr(skb);
adf30907 522 struct net *net = dev_net(skb_dst(skb)->dev);
1da177e4 523
f46078cf
HFS
524 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED)
525 goto fail_hdr;
526
adf30907 527 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMREQDS);
1da177e4
LT
528
529 /* Jumbo payload inhibits frag. header */
98b3377c
DL
530 if (hdr->payload_len==0)
531 goto fail_hdr;
532
ea2ae17d 533 if (!pskb_may_pull(skb, (skb_transport_offset(skb) +
98b3377c
DL
534 sizeof(struct frag_hdr))))
535 goto fail_hdr;
1da177e4 536
0660e03f 537 hdr = ipv6_hdr(skb);
9c70220b 538 fhdr = (struct frag_hdr *)skb_transport_header(skb);
1da177e4
LT
539
540 if (!(fhdr->frag_off & htons(0xFFF9))) {
541 /* It is not a fragmented frame */
b0e380b1 542 skb->transport_header += sizeof(struct frag_hdr);
483a47d2 543 IP6_INC_STATS_BH(net,
adf30907 544 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMOKS);
1da177e4 545
d56f90a7 546 IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
f46078cf 547 IP6CB(skb)->flags |= IP6SKB_FRAGMENTED;
1da177e4
LT
548 return 1;
549 }
550
eec2e618
HFS
551 fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr,
552 ip6_frag_ecn(hdr));
9546377c 553 if (fq != NULL) {
f61944ef 554 int ret;
1da177e4 555
5ab11c98 556 spin_lock(&fq->q.lock);
1da177e4 557
f61944ef 558 ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);
1da177e4 559
5ab11c98 560 spin_unlock(&fq->q.lock);
b836c99f 561 inet_frag_put(&fq->q, &ip6_frags);
1da177e4
LT
562 return ret;
563 }
564
adf30907 565 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMFAILS);
1da177e4
LT
566 kfree_skb(skb);
567 return -1;
98b3377c
DL
568
569fail_hdr:
adf30907 570 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS);
98b3377c
DL
571 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb));
572 return -1;
1da177e4
LT
573}
574
41135cc8 575static const struct inet6_protocol frag_protocol =
1da177e4
LT
576{
577 .handler = ipv6_frag_rcv,
578 .flags = INET6_PROTO_NOPOLICY,
579};
580
8d8354d2 581#ifdef CONFIG_SYSCTL
0a64b4b8 582static struct ctl_table ip6_frags_ns_ctl_table[] = {
8d8354d2 583 {
8d8354d2 584 .procname = "ip6frag_high_thresh",
e31e0bdc 585 .data = &init_net.ipv6.frags.high_thresh,
8d8354d2
PE
586 .maxlen = sizeof(int),
587 .mode = 0644,
6d9f239a 588 .proc_handler = proc_dointvec
8d8354d2
PE
589 },
590 {
8d8354d2 591 .procname = "ip6frag_low_thresh",
e31e0bdc 592 .data = &init_net.ipv6.frags.low_thresh,
8d8354d2
PE
593 .maxlen = sizeof(int),
594 .mode = 0644,
6d9f239a 595 .proc_handler = proc_dointvec
8d8354d2
PE
596 },
597 {
8d8354d2 598 .procname = "ip6frag_time",
b2fd5321 599 .data = &init_net.ipv6.frags.timeout,
8d8354d2
PE
600 .maxlen = sizeof(int),
601 .mode = 0644,
6d9f239a 602 .proc_handler = proc_dointvec_jiffies,
8d8354d2 603 },
7d291ebb
PE
604 { }
605};
606
607static struct ctl_table ip6_frags_ctl_table[] = {
8d8354d2 608 {
8d8354d2 609 .procname = "ip6frag_secret_interval",
3b4bc4a2 610 .data = &ip6_frags.secret_interval,
8d8354d2
PE
611 .maxlen = sizeof(int),
612 .mode = 0644,
6d9f239a 613 .proc_handler = proc_dointvec_jiffies,
8d8354d2
PE
614 },
615 { }
616};
617
2c8c1e72 618static int __net_init ip6_frags_ns_sysctl_register(struct net *net)
8d8354d2 619{
e4a2d5c2 620 struct ctl_table *table;
8d8354d2
PE
621 struct ctl_table_header *hdr;
622
0a64b4b8 623 table = ip6_frags_ns_ctl_table;
09ad9bc7 624 if (!net_eq(net, &init_net)) {
0a64b4b8 625 table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL);
e4a2d5c2
PE
626 if (table == NULL)
627 goto err_alloc;
628
e31e0bdc
PE
629 table[0].data = &net->ipv6.frags.high_thresh;
630 table[1].data = &net->ipv6.frags.low_thresh;
b2fd5321 631 table[2].data = &net->ipv6.frags.timeout;
464dc801
EB
632
633 /* Don't export sysctls to unprivileged users */
634 if (net->user_ns != &init_user_ns)
635 table[0].procname = NULL;
e4a2d5c2
PE
636 }
637
ec8f23ce 638 hdr = register_net_sysctl(net, "net/ipv6", table);
e4a2d5c2
PE
639 if (hdr == NULL)
640 goto err_reg;
641
642 net->ipv6.sysctl.frags_hdr = hdr;
643 return 0;
644
645err_reg:
09ad9bc7 646 if (!net_eq(net, &init_net))
e4a2d5c2
PE
647 kfree(table);
648err_alloc:
649 return -ENOMEM;
650}
651
2c8c1e72 652static void __net_exit ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
653{
654 struct ctl_table *table;
655
656 table = net->ipv6.sysctl.frags_hdr->ctl_table_arg;
657 unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr);
3705e11a
YH
658 if (!net_eq(net, &init_net))
659 kfree(table);
8d8354d2 660}
7d291ebb
PE
661
662static struct ctl_table_header *ip6_ctl_header;
663
664static int ip6_frags_sysctl_register(void)
665{
43444757 666 ip6_ctl_header = register_net_sysctl(&init_net, "net/ipv6",
7d291ebb
PE
667 ip6_frags_ctl_table);
668 return ip6_ctl_header == NULL ? -ENOMEM : 0;
669}
670
671static void ip6_frags_sysctl_unregister(void)
672{
673 unregister_net_sysctl_table(ip6_ctl_header);
674}
8d8354d2 675#else
0a64b4b8 676static inline int ip6_frags_ns_sysctl_register(struct net *net)
e71e0349 677{
8d8354d2
PE
678 return 0;
679}
e4a2d5c2 680
0a64b4b8 681static inline void ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
682{
683}
7d291ebb
PE
684
685static inline int ip6_frags_sysctl_register(void)
686{
687 return 0;
688}
689
690static inline void ip6_frags_sysctl_unregister(void)
691{
692}
8d8354d2 693#endif
7d460db9 694
2c8c1e72 695static int __net_init ipv6_frags_init_net(struct net *net)
8d8354d2 696{
7c070aa9
SW
697 net->ipv6.frags.high_thresh = IPV6_FRAG_HIGH_THRESH;
698 net->ipv6.frags.low_thresh = IPV6_FRAG_LOW_THRESH;
b2fd5321 699 net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT;
8d8354d2 700
e5a2bb84
PE
701 inet_frags_init_net(&net->ipv6.frags);
702
0a64b4b8 703 return ip6_frags_ns_sysctl_register(net);
e71e0349
DL
704}
705
2c8c1e72 706static void __net_exit ipv6_frags_exit_net(struct net *net)
81566e83 707{
0a64b4b8 708 ip6_frags_ns_sysctl_unregister(net);
81566e83
PE
709 inet_frags_exit_net(&net->ipv6.frags, &ip6_frags);
710}
711
712static struct pernet_operations ip6_frags_ops = {
713 .init = ipv6_frags_init_net,
714 .exit = ipv6_frags_exit_net,
715};
716
853cbbaa 717int __init ipv6_frag_init(void)
1da177e4 718{
853cbbaa 719 int ret;
1da177e4 720
853cbbaa
DL
721 ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT);
722 if (ret)
723 goto out;
e71e0349 724
7d291ebb
PE
725 ret = ip6_frags_sysctl_register();
726 if (ret)
727 goto err_sysctl;
728
0002c630
PE
729 ret = register_pernet_subsys(&ip6_frags_ops);
730 if (ret)
731 goto err_pernet;
8d8354d2 732
321a3a99 733 ip6_frags.hashfn = ip6_hashfn;
c6fda282 734 ip6_frags.constructor = ip6_frag_init;
c9547709 735 ip6_frags.destructor = NULL;
1e4b8287
PE
736 ip6_frags.skb_free = NULL;
737 ip6_frags.qsize = sizeof(struct frag_queue);
abd6523d 738 ip6_frags.match = ip6_frag_match;
e521db9d 739 ip6_frags.frag_expire = ip6_frag_expire;
3b4bc4a2 740 ip6_frags.secret_interval = 10 * 60 * HZ;
7eb95156 741 inet_frags_init(&ip6_frags);
853cbbaa
DL
742out:
743 return ret;
0002c630
PE
744
745err_pernet:
7d291ebb
PE
746 ip6_frags_sysctl_unregister();
747err_sysctl:
0002c630
PE
748 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
749 goto out;
853cbbaa
DL
750}
751
752void ipv6_frag_exit(void)
753{
754 inet_frags_fini(&ip6_frags);
7d291ebb 755 ip6_frags_sysctl_unregister();
81566e83 756 unregister_pernet_subsys(&ip6_frags_ops);
853cbbaa 757 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
1da177e4 758}