]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/blame - net/ipv6/reassembly.c
inet: frag: move evictor calls into frag_find function
[mirror_ubuntu-eoan-kernel.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
483a47d2
DL
144 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMTIMEOUT);
145 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
1da177e4 146
78c784c4 147 /* Don't send error if the first segment did not arrive. */
bc578a54 148 if (!(fq->q.last_in & INET_FRAG_FIRST_IN) || !fq->q.fragments)
69df9d59 149 goto out_rcu_unlock;
78c784c4 150
78c784c4
IO
151 /*
152 But use as source device on which LAST ARRIVED
153 segment was received. And do not use fq->dev
154 pointer directly, device might already disappeared.
155 */
5ab11c98 156 fq->q.fragments->dev = dev;
3ffe533c 157 icmpv6_send(fq->q.fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0);
69df9d59
ED
158out_rcu_unlock:
159 rcu_read_unlock();
1da177e4 160out:
5ab11c98 161 spin_unlock(&fq->q.lock);
b836c99f
AW
162 inet_frag_put(&fq->q, frags);
163}
164EXPORT_SYMBOL(ip6_expire_frag_queue);
165
166static void ip6_frag_expire(unsigned long data)
167{
168 struct frag_queue *fq;
169 struct net *net;
170
171 fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q);
172 net = container_of(fq->q.net, struct net, ipv6.frags);
173
174 ip6_expire_frag_queue(net, fq, &ip6_frags);
1da177e4
LT
175}
176
abd6523d 177static __inline__ struct frag_queue *
eec2e618
HFS
178fq_find(struct net *net, __be32 id, const struct in6_addr *src,
179 const struct in6_addr *dst, u8 ecn)
1da177e4 180{
c6fda282
PE
181 struct inet_frag_queue *q;
182 struct ip6_create_arg arg;
abd6523d 183 unsigned int hash;
1da177e4 184
c6fda282 185 arg.id = id;
0b5ccb2e 186 arg.user = IP6_DEFRAG_LOCAL_DELIVER;
c6fda282
PE
187 arg.src = src;
188 arg.dst = dst;
eec2e618 189 arg.ecn = ecn;
9a375803
PE
190
191 read_lock(&ip6_frags.lock);
b1190570 192 hash = inet6_hash_frag(id, src, dst);
1da177e4 193
ac18e750 194 q = inet_frag_find(&net->ipv6.frags, &ip6_frags, &arg, hash);
5a3da1fe
HFS
195 if (IS_ERR_OR_NULL(q)) {
196 inet_frag_maybe_warn_overflow(q, pr_fmt());
9546377c 197 return NULL;
5a3da1fe 198 }
c6fda282 199 return container_of(q, struct frag_queue, q);
1da177e4
LT
200}
201
f61944ef 202static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
1da177e4
LT
203 struct frag_hdr *fhdr, int nhoff)
204{
205 struct sk_buff *prev, *next;
f61944ef 206 struct net_device *dev;
1da177e4 207 int offset, end;
adf30907 208 struct net *net = dev_net(skb_dst(skb)->dev);
eec2e618 209 u8 ecn;
1da177e4 210
bc578a54 211 if (fq->q.last_in & INET_FRAG_COMPLETE)
1da177e4
LT
212 goto err;
213
214 offset = ntohs(fhdr->frag_off) & ~0x7;
0660e03f
ACM
215 end = offset + (ntohs(ipv6_hdr(skb)->payload_len) -
216 ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1)));
1da177e4
LT
217
218 if ((unsigned int)end > IPV6_MAXPLEN) {
adf30907 219 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
a11d206d 220 IPSTATS_MIB_INHDRERRORS);
d56f90a7
ACM
221 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
222 ((u8 *)&fhdr->frag_off -
223 skb_network_header(skb)));
f61944ef 224 return -1;
1da177e4
LT
225 }
226
eec2e618
HFS
227 ecn = ip6_frag_ecn(ipv6_hdr(skb));
228
d56f90a7
ACM
229 if (skb->ip_summed == CHECKSUM_COMPLETE) {
230 const unsigned char *nh = skb_network_header(skb);
1ab1457c 231 skb->csum = csum_sub(skb->csum,
d56f90a7
ACM
232 csum_partial(nh, (u8 *)(fhdr + 1) - nh,
233 0));
234 }
1da177e4
LT
235
236 /* Is this the final fragment? */
237 if (!(fhdr->frag_off & htons(IP6_MF))) {
238 /* If we already have some bits beyond end
239 * or have different end, the segment is corrupted.
240 */
5ab11c98 241 if (end < fq->q.len ||
bc578a54 242 ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len))
1da177e4 243 goto err;
bc578a54 244 fq->q.last_in |= INET_FRAG_LAST_IN;
5ab11c98 245 fq->q.len = end;
1da177e4
LT
246 } else {
247 /* Check if the fragment is rounded to 8 bytes.
248 * Required by the RFC.
249 */
250 if (end & 0x7) {
251 /* RFC2460 says always send parameter problem in
252 * this case. -DaveM
253 */
adf30907 254 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
a11d206d 255 IPSTATS_MIB_INHDRERRORS);
1ab1457c 256 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
1da177e4 257 offsetof(struct ipv6hdr, payload_len));
f61944ef 258 return -1;
1da177e4 259 }
5ab11c98 260 if (end > fq->q.len) {
1da177e4 261 /* Some bits beyond end -> corruption. */
bc578a54 262 if (fq->q.last_in & INET_FRAG_LAST_IN)
1da177e4 263 goto err;
5ab11c98 264 fq->q.len = end;
1da177e4
LT
265 }
266 }
267
268 if (end == offset)
269 goto err;
270
271 /* Point into the IP datagram 'data' part. */
272 if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
273 goto err;
1ab1457c 274
42ca89c1
SH
275 if (pskb_trim_rcsum(skb, end - offset))
276 goto err;
1da177e4
LT
277
278 /* Find out which fragments are in front and at the back of us
279 * in the chain of fragments so far. We must know where to put
280 * this fragment, right?
281 */
d6bebca9
CG
282 prev = fq->q.fragments_tail;
283 if (!prev || FRAG6_CB(prev)->offset < offset) {
284 next = NULL;
285 goto found;
286 }
1da177e4 287 prev = NULL;
5ab11c98 288 for(next = fq->q.fragments; next != NULL; next = next->next) {
1da177e4
LT
289 if (FRAG6_CB(next)->offset >= offset)
290 break; /* bingo! */
291 prev = next;
292 }
293
d6bebca9 294found:
5de658f8
ED
295 /* RFC5722, Section 4, amended by Errata ID : 3089
296 * When reassembling an IPv6 datagram, if
70789d70
ND
297 * one or more its constituent fragments is determined to be an
298 * overlapping fragment, the entire datagram (and any constituent
5de658f8 299 * fragments) MUST be silently discarded.
1da177e4 300 */
1da177e4 301
70789d70
ND
302 /* Check for overlap with preceding fragment. */
303 if (prev &&
f4642141 304 (FRAG6_CB(prev)->offset + prev->len) > offset)
70789d70 305 goto discard_fq;
1da177e4 306
70789d70
ND
307 /* Look for overlap with succeeding segment. */
308 if (next && FRAG6_CB(next)->offset < end)
309 goto discard_fq;
1da177e4
LT
310
311 FRAG6_CB(skb)->offset = offset;
312
313 /* Insert this fragment in the chain of fragments. */
314 skb->next = next;
d6bebca9
CG
315 if (!next)
316 fq->q.fragments_tail = skb;
1da177e4
LT
317 if (prev)
318 prev->next = skb;
319 else
5ab11c98 320 fq->q.fragments = skb;
1da177e4 321
f61944ef
HX
322 dev = skb->dev;
323 if (dev) {
324 fq->iif = dev->ifindex;
325 skb->dev = NULL;
326 }
5ab11c98
PE
327 fq->q.stamp = skb->tstamp;
328 fq->q.meat += skb->len;
eec2e618 329 fq->ecn |= ecn;
d433673e 330 add_frag_mem_limit(&fq->q, skb->truesize);
1da177e4
LT
331
332 /* The first fragment.
333 * nhoffset is obtained from the first fragment, of course.
334 */
335 if (offset == 0) {
336 fq->nhoffset = nhoff;
bc578a54 337 fq->q.last_in |= INET_FRAG_FIRST_IN;
1da177e4 338 }
f61944ef 339
bc578a54 340 if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
97599dc7
ED
341 fq->q.meat == fq->q.len) {
342 int res;
343 unsigned long orefdst = skb->_skb_refdst;
344
345 skb->_skb_refdst = 0UL;
346 res = ip6_frag_reasm(fq, prev, dev);
347 skb->_skb_refdst = orefdst;
348 return res;
349 }
f61944ef 350
97599dc7 351 skb_dst_drop(skb);
3ef0eb0d 352 inet_frag_lru_move(&fq->q);
f61944ef 353 return -1;
1da177e4 354
70789d70 355discard_fq:
b836c99f 356 inet_frag_kill(&fq->q, &ip6_frags);
1da177e4 357err:
adf30907 358 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
3bd653c8 359 IPSTATS_MIB_REASMFAILS);
1da177e4 360 kfree_skb(skb);
f61944ef 361 return -1;
1da177e4
LT
362}
363
364/*
365 * Check if this packet is complete.
366 * Returns NULL on failure by any reason, and pointer
367 * to current nexthdr field in reassembled frame.
368 *
369 * It is called with locked fq, and caller must check that
370 * queue is eligible for reassembly i.e. it is not COMPLETE,
371 * the last and the first frames arrived and all the bits are here.
372 */
f61944ef 373static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
1da177e4
LT
374 struct net_device *dev)
375{
2bad35b7 376 struct net *net = container_of(fq->q.net, struct net, ipv6.frags);
5ab11c98 377 struct sk_buff *fp, *head = fq->q.fragments;
1da177e4
LT
378 int payload_len;
379 unsigned int nhoff;
ec16439e 380 int sum_truesize;
eec2e618 381 u8 ecn;
1da177e4 382
b836c99f 383 inet_frag_kill(&fq->q, &ip6_frags);
1da177e4 384
eec2e618
HFS
385 ecn = ip_frag_ecn_table[fq->ecn];
386 if (unlikely(ecn == 0xff))
387 goto out_fail;
388
f61944ef
HX
389 /* Make the one we just received the head. */
390 if (prev) {
391 head = prev->next;
392 fp = skb_clone(head, GFP_ATOMIC);
393
394 if (!fp)
395 goto out_oom;
396
397 fp->next = head->next;
d6bebca9
CG
398 if (!fp->next)
399 fq->q.fragments_tail = fp;
f61944ef
HX
400 prev->next = fp;
401
5ab11c98
PE
402 skb_morph(head, fq->q.fragments);
403 head->next = fq->q.fragments->next;
f61944ef 404
808db80a 405 consume_skb(fq->q.fragments);
5ab11c98 406 fq->q.fragments = head;
f61944ef
HX
407 }
408
547b792c
IJ
409 WARN_ON(head == NULL);
410 WARN_ON(FRAG6_CB(head)->offset != 0);
1da177e4
LT
411
412 /* Unfragmented part is taken from the first segment. */
d56f90a7 413 payload_len = ((head->data - skb_network_header(head)) -
5ab11c98 414 sizeof(struct ipv6hdr) + fq->q.len -
d56f90a7 415 sizeof(struct frag_hdr));
1da177e4
LT
416 if (payload_len > IPV6_MAXPLEN)
417 goto out_oversize;
418
419 /* Head of list must not be cloned. */
14bbd6a5 420 if (skb_unclone(head, GFP_ATOMIC))
1da177e4
LT
421 goto out_oom;
422
423 /* If the first fragment is fragmented itself, we split
424 * it to two chunks: the first with data and paged part
425 * and the second, holding only fragments. */
21dc3301 426 if (skb_has_frag_list(head)) {
1da177e4
LT
427 struct sk_buff *clone;
428 int i, plen = 0;
429
430 if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL)
431 goto out_oom;
432 clone->next = head->next;
433 head->next = clone;
434 skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list;
4d9092bb 435 skb_frag_list_init(head);
9e903e08
ED
436 for (i = 0; i < skb_shinfo(head)->nr_frags; i++)
437 plen += skb_frag_size(&skb_shinfo(head)->frags[i]);
1da177e4
LT
438 clone->len = clone->data_len = head->data_len - plen;
439 head->data_len -= clone->len;
440 head->len -= clone->len;
441 clone->csum = 0;
442 clone->ip_summed = head->ip_summed;
d433673e 443 add_frag_mem_limit(&fq->q, clone->truesize);
1da177e4
LT
444 }
445
446 /* We have to remove fragment header from datagram and to relocate
447 * header in order to calculate ICV correctly. */
448 nhoff = fq->nhoffset;
b0e380b1 449 skb_network_header(head)[nhoff] = skb_transport_header(head)[0];
1ab1457c 450 memmove(head->head + sizeof(struct frag_hdr), head->head,
1da177e4 451 (head->data - head->head) - sizeof(struct frag_hdr));
b0e380b1
ACM
452 head->mac_header += sizeof(struct frag_hdr);
453 head->network_header += sizeof(struct frag_hdr);
1da177e4 454
badff6d0 455 skb_reset_transport_header(head);
d56f90a7 456 skb_push(head, head->data - skb_network_header(head));
1da177e4 457
ec16439e
ED
458 sum_truesize = head->truesize;
459 for (fp = head->next; fp;) {
460 bool headstolen;
461 int delta;
462 struct sk_buff *next = fp->next;
463
464 sum_truesize += fp->truesize;
1da177e4
LT
465 if (head->ip_summed != fp->ip_summed)
466 head->ip_summed = CHECKSUM_NONE;
84fa7933 467 else if (head->ip_summed == CHECKSUM_COMPLETE)
1da177e4 468 head->csum = csum_add(head->csum, fp->csum);
ec16439e
ED
469
470 if (skb_try_coalesce(head, fp, &headstolen, &delta)) {
471 kfree_skb_partial(fp, headstolen);
472 } else {
473 if (!skb_shinfo(head)->frag_list)
474 skb_shinfo(head)->frag_list = fp;
475 head->data_len += fp->len;
476 head->len += fp->len;
477 head->truesize += fp->truesize;
478 }
479 fp = next;
1da177e4 480 }
d433673e 481 sub_frag_mem_limit(&fq->q, sum_truesize);
1da177e4
LT
482
483 head->next = NULL;
484 head->dev = dev;
5ab11c98 485 head->tstamp = fq->q.stamp;
0660e03f 486 ipv6_hdr(head)->payload_len = htons(payload_len);
eec2e618 487 ipv6_change_dsfield(ipv6_hdr(head), 0xff, ecn);
951dbc8a 488 IP6CB(head)->nhoff = nhoff;
f46078cf 489 IP6CB(head)->flags |= IP6SKB_FRAGMENTED;
1da177e4 490
1da177e4 491 /* Yes, and fold redundant checksum back. 8) */
84fa7933 492 if (head->ip_summed == CHECKSUM_COMPLETE)
d56f90a7 493 head->csum = csum_partial(skb_network_header(head),
cfe1fc77 494 skb_network_header_len(head),
d56f90a7 495 head->csum);
1da177e4 496
a11d206d 497 rcu_read_lock();
2bad35b7 498 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
a11d206d 499 rcu_read_unlock();
5ab11c98 500 fq->q.fragments = NULL;
d6bebca9 501 fq->q.fragments_tail = NULL;
1da177e4
LT
502 return 1;
503
504out_oversize:
e87cc472 505 net_dbg_ratelimited("ip6_frag_reasm: payload len = %d\n", payload_len);
1da177e4
LT
506 goto out_fail;
507out_oom:
e87cc472 508 net_dbg_ratelimited("ip6_frag_reasm: no memory for reassembly\n");
1da177e4 509out_fail:
a11d206d 510 rcu_read_lock();
2bad35b7 511 IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
a11d206d 512 rcu_read_unlock();
1da177e4
LT
513 return -1;
514}
515
e5bbef20 516static int ipv6_frag_rcv(struct sk_buff *skb)
1da177e4 517{
1da177e4
LT
518 struct frag_hdr *fhdr;
519 struct frag_queue *fq;
b71d1d42 520 const struct ipv6hdr *hdr = ipv6_hdr(skb);
adf30907 521 struct net *net = dev_net(skb_dst(skb)->dev);
1da177e4 522
f46078cf
HFS
523 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED)
524 goto fail_hdr;
525
adf30907 526 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMREQDS);
1da177e4
LT
527
528 /* Jumbo payload inhibits frag. header */
98b3377c
DL
529 if (hdr->payload_len==0)
530 goto fail_hdr;
531
ea2ae17d 532 if (!pskb_may_pull(skb, (skb_transport_offset(skb) +
98b3377c
DL
533 sizeof(struct frag_hdr))))
534 goto fail_hdr;
1da177e4 535
0660e03f 536 hdr = ipv6_hdr(skb);
9c70220b 537 fhdr = (struct frag_hdr *)skb_transport_header(skb);
1da177e4
LT
538
539 if (!(fhdr->frag_off & htons(0xFFF9))) {
540 /* It is not a fragmented frame */
b0e380b1 541 skb->transport_header += sizeof(struct frag_hdr);
483a47d2 542 IP6_INC_STATS_BH(net,
adf30907 543 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMOKS);
1da177e4 544
d56f90a7 545 IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
f46078cf 546 IP6CB(skb)->flags |= IP6SKB_FRAGMENTED;
1da177e4
LT
547 return 1;
548 }
549
eec2e618
HFS
550 fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr,
551 ip6_frag_ecn(hdr));
9546377c 552 if (fq != NULL) {
f61944ef 553 int ret;
1da177e4 554
5ab11c98 555 spin_lock(&fq->q.lock);
1da177e4 556
f61944ef 557 ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff);
1da177e4 558
5ab11c98 559 spin_unlock(&fq->q.lock);
b836c99f 560 inet_frag_put(&fq->q, &ip6_frags);
1da177e4
LT
561 return ret;
562 }
563
adf30907 564 IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMFAILS);
1da177e4
LT
565 kfree_skb(skb);
566 return -1;
98b3377c
DL
567
568fail_hdr:
adf30907 569 IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS);
98b3377c
DL
570 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb));
571 return -1;
1da177e4
LT
572}
573
41135cc8 574static const struct inet6_protocol frag_protocol =
1da177e4
LT
575{
576 .handler = ipv6_frag_rcv,
577 .flags = INET6_PROTO_NOPOLICY,
578};
579
8d8354d2 580#ifdef CONFIG_SYSCTL
0a64b4b8 581static struct ctl_table ip6_frags_ns_ctl_table[] = {
8d8354d2 582 {
8d8354d2 583 .procname = "ip6frag_high_thresh",
e31e0bdc 584 .data = &init_net.ipv6.frags.high_thresh,
8d8354d2
PE
585 .maxlen = sizeof(int),
586 .mode = 0644,
6d9f239a 587 .proc_handler = proc_dointvec
8d8354d2
PE
588 },
589 {
8d8354d2 590 .procname = "ip6frag_low_thresh",
e31e0bdc 591 .data = &init_net.ipv6.frags.low_thresh,
8d8354d2
PE
592 .maxlen = sizeof(int),
593 .mode = 0644,
6d9f239a 594 .proc_handler = proc_dointvec
8d8354d2
PE
595 },
596 {
8d8354d2 597 .procname = "ip6frag_time",
b2fd5321 598 .data = &init_net.ipv6.frags.timeout,
8d8354d2
PE
599 .maxlen = sizeof(int),
600 .mode = 0644,
6d9f239a 601 .proc_handler = proc_dointvec_jiffies,
8d8354d2 602 },
7d291ebb
PE
603 { }
604};
605
606static struct ctl_table ip6_frags_ctl_table[] = {
8d8354d2 607 {
8d8354d2 608 .procname = "ip6frag_secret_interval",
3b4bc4a2 609 .data = &ip6_frags.secret_interval,
8d8354d2
PE
610 .maxlen = sizeof(int),
611 .mode = 0644,
6d9f239a 612 .proc_handler = proc_dointvec_jiffies,
8d8354d2
PE
613 },
614 { }
615};
616
2c8c1e72 617static int __net_init ip6_frags_ns_sysctl_register(struct net *net)
8d8354d2 618{
e4a2d5c2 619 struct ctl_table *table;
8d8354d2
PE
620 struct ctl_table_header *hdr;
621
0a64b4b8 622 table = ip6_frags_ns_ctl_table;
09ad9bc7 623 if (!net_eq(net, &init_net)) {
0a64b4b8 624 table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL);
e4a2d5c2
PE
625 if (table == NULL)
626 goto err_alloc;
627
e31e0bdc
PE
628 table[0].data = &net->ipv6.frags.high_thresh;
629 table[1].data = &net->ipv6.frags.low_thresh;
b2fd5321 630 table[2].data = &net->ipv6.frags.timeout;
464dc801
EB
631
632 /* Don't export sysctls to unprivileged users */
633 if (net->user_ns != &init_user_ns)
634 table[0].procname = NULL;
e4a2d5c2
PE
635 }
636
ec8f23ce 637 hdr = register_net_sysctl(net, "net/ipv6", table);
e4a2d5c2
PE
638 if (hdr == NULL)
639 goto err_reg;
640
641 net->ipv6.sysctl.frags_hdr = hdr;
642 return 0;
643
644err_reg:
09ad9bc7 645 if (!net_eq(net, &init_net))
e4a2d5c2
PE
646 kfree(table);
647err_alloc:
648 return -ENOMEM;
649}
650
2c8c1e72 651static void __net_exit ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
652{
653 struct ctl_table *table;
654
655 table = net->ipv6.sysctl.frags_hdr->ctl_table_arg;
656 unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr);
3705e11a
YH
657 if (!net_eq(net, &init_net))
658 kfree(table);
8d8354d2 659}
7d291ebb
PE
660
661static struct ctl_table_header *ip6_ctl_header;
662
663static int ip6_frags_sysctl_register(void)
664{
43444757 665 ip6_ctl_header = register_net_sysctl(&init_net, "net/ipv6",
7d291ebb
PE
666 ip6_frags_ctl_table);
667 return ip6_ctl_header == NULL ? -ENOMEM : 0;
668}
669
670static void ip6_frags_sysctl_unregister(void)
671{
672 unregister_net_sysctl_table(ip6_ctl_header);
673}
8d8354d2 674#else
0a64b4b8 675static inline int ip6_frags_ns_sysctl_register(struct net *net)
e71e0349 676{
8d8354d2
PE
677 return 0;
678}
e4a2d5c2 679
0a64b4b8 680static inline void ip6_frags_ns_sysctl_unregister(struct net *net)
e4a2d5c2
PE
681{
682}
7d291ebb
PE
683
684static inline int ip6_frags_sysctl_register(void)
685{
686 return 0;
687}
688
689static inline void ip6_frags_sysctl_unregister(void)
690{
691}
8d8354d2 692#endif
7d460db9 693
2c8c1e72 694static int __net_init ipv6_frags_init_net(struct net *net)
8d8354d2 695{
7c070aa9
SW
696 net->ipv6.frags.high_thresh = IPV6_FRAG_HIGH_THRESH;
697 net->ipv6.frags.low_thresh = IPV6_FRAG_LOW_THRESH;
b2fd5321 698 net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT;
8d8354d2 699
e5a2bb84
PE
700 inet_frags_init_net(&net->ipv6.frags);
701
0a64b4b8 702 return ip6_frags_ns_sysctl_register(net);
e71e0349
DL
703}
704
2c8c1e72 705static void __net_exit ipv6_frags_exit_net(struct net *net)
81566e83 706{
0a64b4b8 707 ip6_frags_ns_sysctl_unregister(net);
81566e83
PE
708 inet_frags_exit_net(&net->ipv6.frags, &ip6_frags);
709}
710
711static struct pernet_operations ip6_frags_ops = {
712 .init = ipv6_frags_init_net,
713 .exit = ipv6_frags_exit_net,
714};
715
853cbbaa 716int __init ipv6_frag_init(void)
1da177e4 717{
853cbbaa 718 int ret;
1da177e4 719
853cbbaa
DL
720 ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT);
721 if (ret)
722 goto out;
e71e0349 723
7d291ebb
PE
724 ret = ip6_frags_sysctl_register();
725 if (ret)
726 goto err_sysctl;
727
0002c630
PE
728 ret = register_pernet_subsys(&ip6_frags_ops);
729 if (ret)
730 goto err_pernet;
8d8354d2 731
321a3a99 732 ip6_frags.hashfn = ip6_hashfn;
c6fda282 733 ip6_frags.constructor = ip6_frag_init;
c9547709 734 ip6_frags.destructor = NULL;
1e4b8287
PE
735 ip6_frags.skb_free = NULL;
736 ip6_frags.qsize = sizeof(struct frag_queue);
abd6523d 737 ip6_frags.match = ip6_frag_match;
e521db9d 738 ip6_frags.frag_expire = ip6_frag_expire;
3b4bc4a2 739 ip6_frags.secret_interval = 10 * 60 * HZ;
7eb95156 740 inet_frags_init(&ip6_frags);
853cbbaa
DL
741out:
742 return ret;
0002c630
PE
743
744err_pernet:
7d291ebb
PE
745 ip6_frags_sysctl_unregister();
746err_sysctl:
0002c630
PE
747 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
748 goto out;
853cbbaa
DL
749}
750
751void ipv6_frag_exit(void)
752{
753 inet_frags_fini(&ip6_frags);
7d291ebb 754 ip6_frags_sysctl_unregister();
81566e83 755 unregister_pernet_subsys(&ip6_frags_ops);
853cbbaa 756 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
1da177e4 757}