Merge tag 'v3.10.95' into update
[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 */
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 */
93c8b90f
IJ
85unsigned int inet6_hash_frag(__be32 id, const struct in6_addr *saddr,
86 const struct in6_addr *daddr, u32 rnd)
1da177e4 87{
82a39eb6
JK
88 u32 c;
89
8064b3cf
ED
90 c = jhash_3words(ipv6_addr_hash(saddr), ipv6_addr_hash(daddr),
91 (__force u32)id, rnd);
1da177e4 92
7eb95156 93 return c & (INETFRAGS_HASHSZ - 1);
1da177e4 94}
93c8b90f 95EXPORT_SYMBOL_GPL(inet6_hash_frag);
1da177e4 96
321a3a99 97static unsigned int ip6_hashfn(struct inet_frag_queue *q)
1da177e4 98{
321a3a99 99 struct frag_queue *fq;
1da177e4 100
321a3a99 101 fq = container_of(q, struct frag_queue, q);
93c8b90f 102 return inet6_hash_frag(fq->id, &fq->saddr, &fq->daddr, ip6_frags.rnd);
1da177e4
LT
103}
104
cbc264ca 105bool ip6_frag_match(struct inet_frag_queue *q, void *a)
abd6523d
PE
106{
107 struct frag_queue *fq;
108 struct ip6_create_arg *arg = a;
109
110 fq = container_of(q, struct frag_queue, q);
cbc264ca
ED
111 return fq->id == arg->id &&
112 fq->user == arg->user &&
113 ipv6_addr_equal(&fq->saddr, arg->src) &&
7cd9f602
MK
114 ipv6_addr_equal(&fq->daddr, arg->dst) &&
115 (arg->iif == fq->iif ||
116 !(ipv6_addr_type(arg->dst) & (IPV6_ADDR_MULTICAST |
117 IPV6_ADDR_LINKLOCAL)));
abd6523d
PE
118}
119EXPORT_SYMBOL(ip6_frag_match);
120
c6fda282 121void ip6_frag_init(struct inet_frag_queue *q, void *a)
1da177e4 122{
c6fda282
PE
123 struct frag_queue *fq = container_of(q, struct frag_queue, q);
124 struct ip6_create_arg *arg = a;
125
126 fq->id = arg->id;
0b5ccb2e 127 fq->user = arg->user;
4e3fd7a0
AD
128 fq->saddr = *arg->src;
129 fq->daddr = *arg->dst;
eec2e618 130 fq->ecn = arg->ecn;
1da177e4 131}
c6fda282 132EXPORT_SYMBOL(ip6_frag_init);
1da177e4 133
b836c99f
AW
134void ip6_expire_frag_queue(struct net *net, struct frag_queue *fq,
135 struct inet_frags *frags)
1da177e4 136{
a11d206d 137 struct net_device *dev = NULL;
e521db9d 138
5ab11c98 139 spin_lock(&fq->q.lock);
1da177e4 140
bc578a54 141 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
142 goto out;
143
b836c99f 144 inet_frag_kill(&fq->q, frags);
1da177e4 145
69df9d59
ED
146 rcu_read_lock();
147 dev = dev_get_by_index_rcu(net, fq->iif);
a11d206d 148 if (!dev)
69df9d59 149 goto out_rcu_unlock;
a11d206d 150
483a47d2
DL
151 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMTIMEOUT);
152 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
1da177e4 153
78c784c4 154 /* Don't send error if the first segment did not arrive. */
bc578a54 155 if (!(fq->q.last_in & INET_FRAG_FIRST_IN) || !fq->q.fragments)
69df9d59 156 goto out_rcu_unlock;
78c784c4 157
78c784c4
IO
158 /*
159 But use as source device on which LAST ARRIVED
160 segment was received. And do not use fq->dev
161 pointer directly, device might already disappeared.
162 */
5ab11c98 163 fq->q.fragments->dev = dev;
3ffe533c 164 icmpv6_send(fq->q.fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0);
69df9d59
ED
165out_rcu_unlock:
166 rcu_read_unlock();
1da177e4 167out:
5ab11c98 168 spin_unlock(&fq->q.lock);
b836c99f
AW
169 inet_frag_put(&fq->q, frags);
170}
171EXPORT_SYMBOL(ip6_expire_frag_queue);
172
173static void ip6_frag_expire(unsigned long data)
174{
175 struct frag_queue *fq;
176 struct net *net;
177
178 fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q);
179 net = container_of(fq->q.net, struct net, ipv6.frags);
180
181 ip6_expire_frag_queue(net, fq, &ip6_frags);
1da177e4
LT
182}
183
abd6523d 184static __inline__ struct frag_queue *
eec2e618 185fq_find(struct net *net, __be32 id, const struct in6_addr *src,
7cd9f602 186 const struct in6_addr *dst, int iif, u8 ecn)
1da177e4 187{
c6fda282
PE
188 struct inet_frag_queue *q;
189 struct ip6_create_arg arg;
abd6523d 190 unsigned int hash;
1da177e4 191
c6fda282 192 arg.id = id;
0b5ccb2e 193 arg.user = IP6_DEFRAG_LOCAL_DELIVER;
c6fda282
PE
194 arg.src = src;
195 arg.dst = dst;
7cd9f602 196 arg.iif = iif;
eec2e618 197 arg.ecn = ecn;
9a375803
PE
198
199 read_lock(&ip6_frags.lock);
93c8b90f 200 hash = inet6_hash_frag(id, src, dst, ip6_frags.rnd);
1da177e4 201
ac18e750 202 q = inet_frag_find(&net->ipv6.frags, &ip6_frags, &arg, hash);
5a3da1fe
HFS
203 if (IS_ERR_OR_NULL(q)) {
204 inet_frag_maybe_warn_overflow(q, pr_fmt());
9546377c 205 return NULL;
5a3da1fe 206 }
c6fda282 207 return container_of(q, struct frag_queue, q);
1da177e4
LT
208}
209
f61944ef 210static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
1da177e4
LT
211 struct frag_hdr *fhdr, int nhoff)
212{
213 struct sk_buff *prev, *next;
f61944ef 214 struct net_device *dev;
1da177e4 215 int offset, end;
adf30907 216 struct net *net = dev_net(skb_dst(skb)->dev);
eec2e618 217 u8 ecn;
1da177e4 218
bc578a54 219 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
220 goto err;
221
222 offset = ntohs(fhdr->frag_off) & ~0x7;
0660e03f
ACM
223 end = offset + (ntohs(ipv6_hdr(skb)->payload_len) -
224 ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1)));
1da177e4
LT
225
226 if ((unsigned int)end > IPV6_MAXPLEN) {
adf30907 227 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
a11d206d 228 IPSTATS_MIB_INHDRERRORS);
d56f90a7
ACM
229 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
230 ((u8 *)&fhdr->frag_off -
231 skb_network_header(skb)));
f61944ef 232 return -1;
1da177e4
LT
233 }
234
eec2e618
HFS
235 ecn = ip6_frag_ecn(ipv6_hdr(skb));
236
d56f90a7
ACM
237 if (skb->ip_summed == CHECKSUM_COMPLETE) {
238 const unsigned char *nh = skb_network_header(skb);
1ab1457c 239 skb->csum = csum_sub(skb->csum,
d56f90a7
ACM
240 csum_partial(nh, (u8 *)(fhdr + 1) - nh,
241 0));
242 }
1da177e4
LT
243
244 /* Is this the final fragment? */
245 if (!(fhdr->frag_off & htons(IP6_MF))) {
246 /* If we already have some bits beyond end
247 * or have different end, the segment is corrupted.
248 */
5ab11c98 249 if (end < fq->q.len ||
bc578a54 250 ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len))
1da177e4 251 goto err;
bc578a54 252 fq->q.last_in |= INET_FRAG_LAST_IN;
5ab11c98 253 fq->q.len = end;
1da177e4
LT
254 } else {
255 /* Check if the fragment is rounded to 8 bytes.
256 * Required by the RFC.
257 */
258 if (end & 0x7) {
259 /* RFC2460 says always send parameter problem in
260 * this case. -DaveM
261 */
adf30907 262 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
a11d206d 263 IPSTATS_MIB_INHDRERRORS);
1ab1457c 264 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
1da177e4 265 offsetof(struct ipv6hdr, payload_len));
f61944ef 266 return -1;
1da177e4 267 }
5ab11c98 268 if (end > fq->q.len) {
1da177e4 269 /* Some bits beyond end -> corruption. */
bc578a54 270 if (fq->q.last_in & INET_FRAG_LAST_IN)
1da177e4 271 goto err;
5ab11c98 272 fq->q.len = end;
1da177e4
LT
273 }
274 }
275
276 if (end == offset)
277 goto err;
278
279 /* Point into the IP datagram 'data' part. */
280 if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
281 goto err;
1ab1457c 282
42ca89c1
SH
283 if (pskb_trim_rcsum(skb, end - offset))
284 goto err;
1da177e4
LT
285
286 /* Find out which fragments are in front and at the back of us
287 * in the chain of fragments so far. We must know where to put
288 * this fragment, right?
289 */
d6bebca9
CG
290 prev = fq->q.fragments_tail;
291 if (!prev || FRAG6_CB(prev)->offset < offset) {
292 next = NULL;
293 goto found;
294 }
1da177e4 295 prev = NULL;
5ab11c98 296 for(next = fq->q.fragments; next != NULL; next = next->next) {
1da177e4
LT
297 if (FRAG6_CB(next)->offset >= offset)
298 break; /* bingo! */
299 prev = next;
300 }
301
d6bebca9 302found:
5de658f8
ED
303 /* RFC5722, Section 4, amended by Errata ID : 3089
304 * When reassembling an IPv6 datagram, if
70789d70
ND
305 * one or more its constituent fragments is determined to be an
306 * overlapping fragment, the entire datagram (and any constituent
5de658f8 307 * fragments) MUST be silently discarded.
1da177e4 308 */
1da177e4 309
70789d70
ND
310 /* Check for overlap with preceding fragment. */
311 if (prev &&
f4642141 312 (FRAG6_CB(prev)->offset + prev->len) > offset)
70789d70 313 goto discard_fq;
1da177e4 314
70789d70
ND
315 /* Look for overlap with succeeding segment. */
316 if (next && FRAG6_CB(next)->offset < end)
317 goto discard_fq;
1da177e4
LT
318
319 FRAG6_CB(skb)->offset = offset;
320
321 /* Insert this fragment in the chain of fragments. */
322 skb->next = next;
d6bebca9
CG
323 if (!next)
324 fq->q.fragments_tail = skb;
1da177e4
LT
325 if (prev)
326 prev->next = skb;
327 else
5ab11c98 328 fq->q.fragments = skb;
1da177e4 329
f61944ef
HX
330 dev = skb->dev;
331 if (dev) {
332 fq->iif = dev->ifindex;
333 skb->dev = NULL;
334 }
5ab11c98
PE
335 fq->q.stamp = skb->tstamp;
336 fq->q.meat += skb->len;
eec2e618 337 fq->ecn |= ecn;
d433673e 338 add_frag_mem_limit(&fq->q, skb->truesize);
1da177e4
LT
339
340 /* The first fragment.
341 * nhoffset is obtained from the first fragment, of course.
342 */
343 if (offset == 0) {
344 fq->nhoffset = nhoff;
bc578a54 345 fq->q.last_in |= INET_FRAG_FIRST_IN;
1da177e4 346 }
f61944ef 347
bc578a54 348 if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
97599dc7
ED
349 fq->q.meat == fq->q.len) {
350 int res;
351 unsigned long orefdst = skb->_skb_refdst;
352
353 skb->_skb_refdst = 0UL;
354 res = ip6_frag_reasm(fq, prev, dev);
355 skb->_skb_refdst = orefdst;
356 return res;
357 }
f61944ef 358
97599dc7 359 skb_dst_drop(skb);
3ef0eb0d 360 inet_frag_lru_move(&fq->q);
f61944ef 361 return -1;
1da177e4 362
70789d70 363discard_fq:
b836c99f 364 inet_frag_kill(&fq->q, &ip6_frags);
1da177e4 365err:
adf30907 366 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
3bd653c8 367 IPSTATS_MIB_REASMFAILS);
1da177e4 368 kfree_skb(skb);
f61944ef 369 return -1;
1da177e4
LT
370}
371
372/*
373 * Check if this packet is complete.
374 * Returns NULL on failure by any reason, and pointer
375 * to current nexthdr field in reassembled frame.
376 *
377 * It is called with locked fq, and caller must check that
378 * queue is eligible for reassembly i.e. it is not COMPLETE,
379 * the last and the first frames arrived and all the bits are here.
380 */
f61944ef 381static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
1da177e4
LT
382 struct net_device *dev)
383{
2bad35b7 384 struct net *net = container_of(fq->q.net, struct net, ipv6.frags);
5ab11c98 385 struct sk_buff *fp, *head = fq->q.fragments;
1da177e4
LT
386 int payload_len;
387 unsigned int nhoff;
ec16439e 388 int sum_truesize;
eec2e618 389 u8 ecn;
1da177e4 390
b836c99f 391 inet_frag_kill(&fq->q, &ip6_frags);
1da177e4 392
eec2e618
HFS
393 ecn = ip_frag_ecn_table[fq->ecn];
394 if (unlikely(ecn == 0xff))
395 goto out_fail;
396
f61944ef
HX
397 /* Make the one we just received the head. */
398 if (prev) {
399 head = prev->next;
400 fp = skb_clone(head, GFP_ATOMIC);
401
402 if (!fp)
403 goto out_oom;
404
405 fp->next = head->next;
d6bebca9
CG
406 if (!fp->next)
407 fq->q.fragments_tail = fp;
f61944ef
HX
408 prev->next = fp;
409
5ab11c98
PE
410 skb_morph(head, fq->q.fragments);
411 head->next = fq->q.fragments->next;
f61944ef 412
808db80a 413 consume_skb(fq->q.fragments);
5ab11c98 414 fq->q.fragments = head;
f61944ef
HX
415 }
416
547b792c
IJ
417 WARN_ON(head == NULL);
418 WARN_ON(FRAG6_CB(head)->offset != 0);
1da177e4
LT
419
420 /* Unfragmented part is taken from the first segment. */
d56f90a7 421 payload_len = ((head->data - skb_network_header(head)) -
5ab11c98 422 sizeof(struct ipv6hdr) + fq->q.len -
d56f90a7 423 sizeof(struct frag_hdr));
1da177e4
LT
424 if (payload_len > IPV6_MAXPLEN)
425 goto out_oversize;
426
427 /* Head of list must not be cloned. */
14bbd6a5 428 if (skb_unclone(head, GFP_ATOMIC))
1da177e4
LT
429 goto out_oom;
430
431 /* If the first fragment is fragmented itself, we split
432 * it to two chunks: the first with data and paged part
433 * and the second, holding only fragments. */
21dc3301 434 if (skb_has_frag_list(head)) {
1da177e4
LT
435 struct sk_buff *clone;
436 int i, plen = 0;
437
438 if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
439 goto out_oom;
440 clone->next = head->next;
441 head->next = clone;
442 skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
4d9092bb 443 skb_frag_list_init(head);
9e903e08
ED
444 for (i = 0; i < skb_shinfo(head)->nr_frags; i++)
445 plen += skb_frag_size(&skb_shinfo(head)->frags[i]);
1da177e4
LT
446 clone->len = clone->data_len = head->data_len - plen;
447 head->data_len -= clone->len;
448 head->len -= clone->len;
449 clone->csum = 0;
450 clone->ip_summed = head->ip_summed;
d433673e 451 add_frag_mem_limit(&fq->q, clone->truesize);
1da177e4
LT
452 }
453
454 /* We have to remove fragment header from datagram and to relocate
455 * header in order to calculate ICV correctly. */
456 nhoff = fq->nhoffset;
b0e380b1 457 skb_network_header(head)[nhoff] = skb_transport_header(head)[0];
1ab1457c 458 memmove(head->head + sizeof(struct frag_hdr), head->head,
1da177e4 459 (head->data - head->head) - sizeof(struct frag_hdr));
b0e380b1
ACM
460 head->mac_header += sizeof(struct frag_hdr);
461 head->network_header += sizeof(struct frag_hdr);
1da177e4 462
badff6d0 463 skb_reset_transport_header(head);
d56f90a7 464 skb_push(head, head->data - skb_network_header(head));
1da177e4 465
ec16439e
ED
466 sum_truesize = head->truesize;
467 for (fp = head->next; fp;) {
468 bool headstolen;
469 int delta;
470 struct sk_buff *next = fp->next;
471
472 sum_truesize += fp->truesize;
1da177e4
LT
473 if (head->ip_summed != fp->ip_summed)
474 head->ip_summed = CHECKSUM_NONE;
84fa7933 475 else if (head->ip_summed == CHECKSUM_COMPLETE)
1da177e4 476 head->csum = csum_add(head->csum, fp->csum);
ec16439e
ED
477
478 if (skb_try_coalesce(head, fp, &headstolen, &delta)) {
479 kfree_skb_partial(fp, headstolen);
480 } else {
481 if (!skb_shinfo(head)->frag_list)
482 skb_shinfo(head)->frag_list = fp;
483 head->data_len += fp->len;
484 head->len += fp->len;
485 head->truesize += fp->truesize;
486 }
487 fp = next;
1da177e4 488 }
d433673e 489 sub_frag_mem_limit(&fq->q, sum_truesize);
1da177e4
LT
490
491 head->next = NULL;
492 head->dev = dev;
5ab11c98 493 head->tstamp = fq->q.stamp;
0660e03f 494 ipv6_hdr(head)->payload_len = htons(payload_len);
eec2e618 495 ipv6_change_dsfield(ipv6_hdr(head), 0xff, ecn);
951dbc8a 496 IP6CB(head)->nhoff = nhoff;
a829a288 497 IP6CB(head)->flags |= IP6SKB_FRAGMENTED;
1da177e4 498
1da177e4 499 /* Yes, and fold redundant checksum back. 8) */
84fa7933 500 if (head->ip_summed == CHECKSUM_COMPLETE)
d56f90a7 501 head->csum = csum_partial(skb_network_header(head),
cfe1fc77 502 skb_network_header_len(head),
d56f90a7 503 head->csum);
1da177e4 504
a11d206d 505 rcu_read_lock();
2bad35b7 506 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
a11d206d 507 rcu_read_unlock();
5ab11c98 508 fq->q.fragments = NULL;
d6bebca9 509 fq->q.fragments_tail = NULL;
1da177e4
LT
510 return 1;
511
512out_oversize:
e87cc472 513 net_dbg_ratelimited("ip6_frag_reasm: payload len = %d\n", payload_len);
1da177e4
LT
514 goto out_fail;
515out_oom:
e87cc472 516 net_dbg_ratelimited("ip6_frag_reasm: no memory for reassembly\n");
1da177e4 517out_fail:
a11d206d 518 rcu_read_lock();
2bad35b7 519 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
a11d206d 520 rcu_read_unlock();
1da177e4
LT
521 return -1;
522}
523
e5bbef20 524static int ipv6_frag_rcv(struct sk_buff *skb)
1da177e4 525{
1da177e4
LT
526 struct frag_hdr *fhdr;
527 struct frag_queue *fq;
b71d1d42 528 const struct ipv6hdr *hdr = ipv6_hdr(skb);
adf30907 529 struct net *net = dev_net(skb_dst(skb)->dev);
6b102865 530 int evicted;
1da177e4 531
a829a288
HFS
532 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED)
533 goto fail_hdr;
534
adf30907 535 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMREQDS);
1da177e4
LT
536
537 /* Jumbo payload inhibits frag. header */
98b3377c
DL
538 if (hdr->payload_len==0)
539 goto fail_hdr;
540
ea2ae17d 541 if (!pskb_may_pull(skb, (skb_transport_offset(skb) +
98b3377c
DL
542 sizeof(struct frag_hdr))))
543 goto fail_hdr;
1da177e4 544
0660e03f 545 hdr = ipv6_hdr(skb);
9c70220b 546 fhdr = (struct frag_hdr *)skb_transport_header(skb);
1da177e4
LT
547
548 if (!(fhdr->frag_off & htons(0xFFF9))) {
549 /* It is not a fragmented frame */
b0e380b1 550 skb->transport_header += sizeof(struct frag_hdr);
483a47d2 551 IP6_INC_STATS_BH(net,
adf30907 552 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMOKS);
1da177e4 553
d56f90a7 554 IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
a829a288 555 IP6CB(skb)->flags |= IP6SKB_FRAGMENTED;
1da177e4
LT
556 return 1;
557 }
558
6b102865
AW
559 evicted = inet_frag_evictor(&net->ipv6.frags, &ip6_frags, false);
560 if (evicted)
561 IP6_ADD_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
562 IPSTATS_MIB_REASMFAILS, evicted);
1da177e4 563
eec2e618 564 fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr,
7cd9f602 565 skb->dev ? skb->dev->ifindex : 0, ip6_frag_ecn(hdr));
9546377c 566 if (fq != NULL) {
f61944ef 567 int ret;
1da177e4 568
5ab11c98 569 spin_lock(&fq->q.lock);
1da177e4 570
f61944ef 571 ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);
1da177e4 572
5ab11c98 573 spin_unlock(&fq->q.lock);
b836c99f 574 inet_frag_put(&fq->q, &ip6_frags);
1da177e4
LT
575 return ret;
576 }
577
adf30907 578 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMFAILS);
1da177e4
LT
579 kfree_skb(skb);
580 return -1;
98b3377c
DL
581
582fail_hdr:
adf30907 583 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS);
98b3377c
DL
584 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb));
585 return -1;
1da177e4
LT
586}
587
41135cc8 588static const struct inet6_protocol frag_protocol =
1da177e4
LT
589{
590 .handler = ipv6_frag_rcv,
591 .flags = INET6_PROTO_NOPOLICY,
592};
593
8d8354d2 594#ifdef CONFIG_SYSCTL
0a64b4b8 595static struct ctl_table ip6_frags_ns_ctl_table[] = {
8d8354d2 596 {
8d8354d2 597 .procname = "ip6frag_high_thresh",
e31e0bdc 598 .data = &init_net.ipv6.frags.high_thresh,
8d8354d2
PE
599 .maxlen = sizeof(int),
600 .mode = 0644,
6d9f239a 601 .proc_handler = proc_dointvec
8d8354d2
PE
602 },
603 {
8d8354d2 604 .procname = "ip6frag_low_thresh",
e31e0bdc 605 .data = &init_net.ipv6.frags.low_thresh,
8d8354d2
PE
606 .maxlen = sizeof(int),
607 .mode = 0644,
6d9f239a 608 .proc_handler = proc_dointvec
8d8354d2
PE
609 },
610 {
8d8354d2 611 .procname = "ip6frag_time",
b2fd5321 612 .data = &init_net.ipv6.frags.timeout,
8d8354d2
PE
613 .maxlen = sizeof(int),
614 .mode = 0644,
6d9f239a 615 .proc_handler = proc_dointvec_jiffies,
8d8354d2 616 },
7d291ebb
PE
617 { }
618};
619
620static struct ctl_table ip6_frags_ctl_table[] = {
8d8354d2 621 {
8d8354d2 622 .procname = "ip6frag_secret_interval",
3b4bc4a2 623 .data = &ip6_frags.secret_interval,
8d8354d2
PE
624 .maxlen = sizeof(int),
625 .mode = 0644,
6d9f239a 626 .proc_handler = proc_dointvec_jiffies,
8d8354d2
PE
627 },
628 { }
629};
630
2c8c1e72 631static int __net_init ip6_frags_ns_sysctl_register(struct net *net)
8d8354d2 632{
e4a2d5c2 633 struct ctl_table *table;
8d8354d2
PE
634 struct ctl_table_header *hdr;
635
0a64b4b8 636 table = ip6_frags_ns_ctl_table;
09ad9bc7 637 if (!net_eq(net, &init_net)) {
0a64b4b8 638 table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL);
e4a2d5c2
PE
639 if (table == NULL)
640 goto err_alloc;
641
e31e0bdc
PE
642 table[0].data = &net->ipv6.frags.high_thresh;
643 table[1].data = &net->ipv6.frags.low_thresh;
b2fd5321 644 table[2].data = &net->ipv6.frags.timeout;
464dc801
EB
645
646 /* Don't export sysctls to unprivileged users */
647 if (net->user_ns != &init_user_ns)
648 table[0].procname = NULL;
e4a2d5c2
PE
649 }
650
ec8f23ce 651 hdr = register_net_sysctl(net, "net/ipv6", table);
e4a2d5c2
PE
652 if (hdr == NULL)
653 goto err_reg;
654
655 net->ipv6.sysctl.frags_hdr = hdr;
656 return 0;
657
658err_reg:
09ad9bc7 659 if (!net_eq(net, &init_net))
e4a2d5c2
PE
660 kfree(table);
661err_alloc:
662 return -ENOMEM;
663}
664
2c8c1e72 665static void __net_exit ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
666{
667 struct ctl_table *table;
668
669 table = net->ipv6.sysctl.frags_hdr->ctl_table_arg;
670 unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr);
3705e11a
YH
671 if (!net_eq(net, &init_net))
672 kfree(table);
8d8354d2 673}
7d291ebb
PE
674
675static struct ctl_table_header *ip6_ctl_header;
676
677static int ip6_frags_sysctl_register(void)
678{
43444757 679 ip6_ctl_header = register_net_sysctl(&init_net, "net/ipv6",
7d291ebb
PE
680 ip6_frags_ctl_table);
681 return ip6_ctl_header == NULL ? -ENOMEM : 0;
682}
683
684static void ip6_frags_sysctl_unregister(void)
685{
686 unregister_net_sysctl_table(ip6_ctl_header);
687}
8d8354d2 688#else
0a64b4b8 689static inline int ip6_frags_ns_sysctl_register(struct net *net)
e71e0349 690{
8d8354d2
PE
691 return 0;
692}
e4a2d5c2 693
0a64b4b8 694static inline void ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
695{
696}
7d291ebb
PE
697
698static inline int ip6_frags_sysctl_register(void)
699{
700 return 0;
701}
702
703static inline void ip6_frags_sysctl_unregister(void)
704{
705}
8d8354d2 706#endif
7d460db9 707
2c8c1e72 708static int __net_init ipv6_frags_init_net(struct net *net)
8d8354d2 709{
7c070aa9
SW
710 net->ipv6.frags.high_thresh = IPV6_FRAG_HIGH_THRESH;
711 net->ipv6.frags.low_thresh = IPV6_FRAG_LOW_THRESH;
b2fd5321 712 net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT;
8d8354d2 713
e5a2bb84
PE
714 inet_frags_init_net(&net->ipv6.frags);
715
0a64b4b8 716 return ip6_frags_ns_sysctl_register(net);
e71e0349
DL
717}
718
2c8c1e72 719static void __net_exit ipv6_frags_exit_net(struct net *net)
81566e83 720{
0a64b4b8 721 ip6_frags_ns_sysctl_unregister(net);
81566e83
PE
722 inet_frags_exit_net(&net->ipv6.frags, &ip6_frags);
723}
724
725static struct pernet_operations ip6_frags_ops = {
726 .init = ipv6_frags_init_net,
727 .exit = ipv6_frags_exit_net,
728};
729
853cbbaa 730int __init ipv6_frag_init(void)
1da177e4 731{
853cbbaa 732 int ret;
1da177e4 733
853cbbaa
DL
734 ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT);
735 if (ret)
736 goto out;
e71e0349 737
7d291ebb
PE
738 ret = ip6_frags_sysctl_register();
739 if (ret)
740 goto err_sysctl;
741
0002c630
PE
742 ret = register_pernet_subsys(&ip6_frags_ops);
743 if (ret)
744 goto err_pernet;
8d8354d2 745
321a3a99 746 ip6_frags.hashfn = ip6_hashfn;
c6fda282 747 ip6_frags.constructor = ip6_frag_init;
c9547709 748 ip6_frags.destructor = NULL;
1e4b8287
PE
749 ip6_frags.skb_free = NULL;
750 ip6_frags.qsize = sizeof(struct frag_queue);
abd6523d 751 ip6_frags.match = ip6_frag_match;
e521db9d 752 ip6_frags.frag_expire = ip6_frag_expire;
3b4bc4a2 753 ip6_frags.secret_interval = 10 * 60 * HZ;
7eb95156 754 inet_frags_init(&ip6_frags);
853cbbaa
DL
755out:
756 return ret;
0002c630
PE
757
758err_pernet:
7d291ebb
PE
759 ip6_frags_sysctl_unregister();
760err_sysctl:
0002c630
PE
761 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
762 goto out;
853cbbaa
DL
763}
764
765void ipv6_frag_exit(void)
766{
767 inet_frags_fini(&ip6_frags);
7d291ebb 768 ip6_frags_sysctl_unregister();
81566e83 769 unregister_pernet_subsys(&ip6_frags_ops);
853cbbaa 770 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
1da177e4 771}