Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * The IP fragmentation functionality. | |
e905a9ed | 7 | * |
1da177e4 LT |
8 | * Version: $Id: ip_fragment.c,v 1.59 2002/01/12 07:54:56 davem Exp $ |
9 | * | |
10 | * Authors: Fred N. van Kempen <waltje@uWalt.NL.Mugnet.ORG> | |
11 | * Alan Cox <Alan.Cox@linux.org> | |
12 | * | |
13 | * Fixes: | |
14 | * Alan Cox : Split from ip.c , see ip_input.c for history. | |
15 | * David S. Miller : Begin massive cleanup... | |
16 | * Andi Kleen : Add sysctls. | |
17 | * xxxx : Overlapfrag bug. | |
18 | * Ultima : ip_expire() kernel panic. | |
19 | * Bill Hawes : Frag accounting and evictor fixes. | |
20 | * John McDonald : 0 length frag bug. | |
21 | * Alexey Kuznetsov: SMP races, threading, cleanup. | |
22 | * Patrick McHardy : LRU queue of frag heads for evictor. | |
23 | */ | |
24 | ||
89cee8b1 | 25 | #include <linux/compiler.h> |
1da177e4 LT |
26 | #include <linux/module.h> |
27 | #include <linux/types.h> | |
28 | #include <linux/mm.h> | |
29 | #include <linux/jiffies.h> | |
30 | #include <linux/skbuff.h> | |
31 | #include <linux/list.h> | |
32 | #include <linux/ip.h> | |
33 | #include <linux/icmp.h> | |
34 | #include <linux/netdevice.h> | |
35 | #include <linux/jhash.h> | |
36 | #include <linux/random.h> | |
37 | #include <net/sock.h> | |
38 | #include <net/ip.h> | |
39 | #include <net/icmp.h> | |
40 | #include <net/checksum.h> | |
89cee8b1 | 41 | #include <net/inetpeer.h> |
5ab11c98 | 42 | #include <net/inet_frag.h> |
1da177e4 LT |
43 | #include <linux/tcp.h> |
44 | #include <linux/udp.h> | |
45 | #include <linux/inet.h> | |
46 | #include <linux/netfilter_ipv4.h> | |
47 | ||
48 | /* NOTE. Logic of IP defragmentation is parallel to corresponding IPv6 | |
49 | * code now. If you change something here, _PLEASE_ update ipv6/reassembly.c | |
50 | * as well. Or notify me, at least. --ANK | |
51 | */ | |
52 | ||
8d8354d2 | 53 | static int sysctl_ipfrag_max_dist __read_mostly = 64; |
89cee8b1 | 54 | |
1da177e4 LT |
55 | struct ipfrag_skb_cb |
56 | { | |
57 | struct inet_skb_parm h; | |
58 | int offset; | |
59 | }; | |
60 | ||
61 | #define FRAG_CB(skb) ((struct ipfrag_skb_cb*)((skb)->cb)) | |
62 | ||
63 | /* Describe an entry in the "incomplete datagrams" queue. */ | |
64 | struct ipq { | |
5ab11c98 PE |
65 | struct inet_frag_queue q; |
66 | ||
1da177e4 | 67 | u32 user; |
18277770 AV |
68 | __be32 saddr; |
69 | __be32 daddr; | |
70 | __be16 id; | |
1da177e4 | 71 | u8 protocol; |
89cee8b1 HX |
72 | int iif; |
73 | unsigned int rid; | |
74 | struct inet_peer *peer; | |
1da177e4 LT |
75 | }; |
76 | ||
8d8354d2 | 77 | static struct inet_frags_ctl ip4_frags_ctl __read_mostly = { |
04128f23 PE |
78 | /* |
79 | * Fragment cache limits. We will commit 256K at one time. Should we | |
80 | * cross that limit we will prune down to 192K. This should cope with | |
81 | * even the most extreme cases without allowing an attacker to | |
82 | * measurably harm machine performance. | |
83 | */ | |
84 | .high_thresh = 256 * 1024, | |
85 | .low_thresh = 192 * 1024, | |
86 | ||
87 | /* | |
88 | * Important NOTE! Fragment queue must be destroyed before MSL expires. | |
89 | * RFC791 is wrong proposing to prolongate timer each fragment arrival | |
90 | * by TTL. | |
91 | */ | |
92 | .timeout = IP_FRAG_TIME, | |
93 | .secret_interval = 10 * 60 * HZ, | |
94 | }; | |
95 | ||
7eb95156 | 96 | static struct inet_frags ip4_frags; |
1da177e4 | 97 | |
e5a2bb84 | 98 | int ip_frag_nqueues(struct net *net) |
7eb95156 | 99 | { |
e5a2bb84 | 100 | return net->ipv4.frags.nqueues; |
7eb95156 | 101 | } |
1da177e4 | 102 | |
7eb95156 PE |
103 | int ip_frag_mem(void) |
104 | { | |
105 | return atomic_read(&ip4_frags.mem); | |
106 | } | |
1da177e4 | 107 | |
1706d587 HX |
108 | static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev, |
109 | struct net_device *dev); | |
110 | ||
c6fda282 PE |
111 | struct ip4_create_arg { |
112 | struct iphdr *iph; | |
113 | u32 user; | |
114 | }; | |
115 | ||
18277770 | 116 | static unsigned int ipqhashfn(__be16 id, __be32 saddr, __be32 daddr, u8 prot) |
1da177e4 | 117 | { |
18277770 AV |
118 | return jhash_3words((__force u32)id << 16 | prot, |
119 | (__force u32)saddr, (__force u32)daddr, | |
7eb95156 | 120 | ip4_frags.rnd) & (INETFRAGS_HASHSZ - 1); |
1da177e4 LT |
121 | } |
122 | ||
321a3a99 | 123 | static unsigned int ip4_hashfn(struct inet_frag_queue *q) |
1da177e4 | 124 | { |
321a3a99 | 125 | struct ipq *ipq; |
1da177e4 | 126 | |
321a3a99 PE |
127 | ipq = container_of(q, struct ipq, q); |
128 | return ipqhashfn(ipq->id, ipq->saddr, ipq->daddr, ipq->protocol); | |
1da177e4 LT |
129 | } |
130 | ||
abd6523d PE |
131 | static int ip4_frag_match(struct inet_frag_queue *q, void *a) |
132 | { | |
133 | struct ipq *qp; | |
134 | struct ip4_create_arg *arg = a; | |
135 | ||
136 | qp = container_of(q, struct ipq, q); | |
137 | return (qp->id == arg->iph->id && | |
138 | qp->saddr == arg->iph->saddr && | |
139 | qp->daddr == arg->iph->daddr && | |
140 | qp->protocol == arg->iph->protocol && | |
141 | qp->user == arg->user); | |
142 | } | |
143 | ||
1da177e4 LT |
144 | /* Memory Tracking Functions. */ |
145 | static __inline__ void frag_kfree_skb(struct sk_buff *skb, int *work) | |
146 | { | |
147 | if (work) | |
148 | *work -= skb->truesize; | |
7eb95156 | 149 | atomic_sub(skb->truesize, &ip4_frags.mem); |
1da177e4 LT |
150 | kfree_skb(skb); |
151 | } | |
152 | ||
c6fda282 PE |
153 | static void ip4_frag_init(struct inet_frag_queue *q, void *a) |
154 | { | |
155 | struct ipq *qp = container_of(q, struct ipq, q); | |
156 | struct ip4_create_arg *arg = a; | |
157 | ||
158 | qp->protocol = arg->iph->protocol; | |
159 | qp->id = arg->iph->id; | |
160 | qp->saddr = arg->iph->saddr; | |
161 | qp->daddr = arg->iph->daddr; | |
162 | qp->user = arg->user; | |
163 | qp->peer = sysctl_ipfrag_max_dist ? | |
164 | inet_getpeer(arg->iph->saddr, 1) : NULL; | |
165 | } | |
166 | ||
1e4b8287 | 167 | static __inline__ void ip4_frag_free(struct inet_frag_queue *q) |
1da177e4 | 168 | { |
1e4b8287 PE |
169 | struct ipq *qp; |
170 | ||
171 | qp = container_of(q, struct ipq, q); | |
172 | if (qp->peer) | |
173 | inet_putpeer(qp->peer); | |
1da177e4 LT |
174 | } |
175 | ||
1da177e4 LT |
176 | |
177 | /* Destruction primitives. */ | |
178 | ||
4b6cb5d8 | 179 | static __inline__ void ipq_put(struct ipq *ipq) |
1da177e4 | 180 | { |
762cc408 | 181 | inet_frag_put(&ipq->q, &ip4_frags); |
1da177e4 LT |
182 | } |
183 | ||
184 | /* Kill ipq entry. It is not destroyed immediately, | |
185 | * because caller (and someone more) holds reference count. | |
186 | */ | |
187 | static void ipq_kill(struct ipq *ipq) | |
188 | { | |
277e650d | 189 | inet_frag_kill(&ipq->q, &ip4_frags); |
1da177e4 LT |
190 | } |
191 | ||
e905a9ed | 192 | /* Memory limiting on fragments. Evictor trashes the oldest |
1da177e4 LT |
193 | * fragment queue until we are back under the threshold. |
194 | */ | |
195 | static void ip_evictor(void) | |
196 | { | |
8e7999c4 PE |
197 | int evicted; |
198 | ||
199 | evicted = inet_frag_evictor(&ip4_frags); | |
200 | if (evicted) | |
201 | IP_ADD_STATS_BH(IPSTATS_MIB_REASMFAILS, evicted); | |
1da177e4 LT |
202 | } |
203 | ||
204 | /* | |
205 | * Oops, a fragment queue timed out. Kill it and send an ICMP reply. | |
206 | */ | |
207 | static void ip_expire(unsigned long arg) | |
208 | { | |
e521db9d PE |
209 | struct ipq *qp; |
210 | ||
211 | qp = container_of((struct inet_frag_queue *) arg, struct ipq, q); | |
1da177e4 | 212 | |
5ab11c98 | 213 | spin_lock(&qp->q.lock); |
1da177e4 | 214 | |
5ab11c98 | 215 | if (qp->q.last_in & COMPLETE) |
1da177e4 LT |
216 | goto out; |
217 | ||
218 | ipq_kill(qp); | |
219 | ||
220 | IP_INC_STATS_BH(IPSTATS_MIB_REASMTIMEOUT); | |
221 | IP_INC_STATS_BH(IPSTATS_MIB_REASMFAILS); | |
222 | ||
5ab11c98 PE |
223 | if ((qp->q.last_in&FIRST_IN) && qp->q.fragments != NULL) { |
224 | struct sk_buff *head = qp->q.fragments; | |
1da177e4 | 225 | /* Send an ICMP "Fragment Reassembly Timeout" message. */ |
881d966b | 226 | if ((head->dev = dev_get_by_index(&init_net, qp->iif)) != NULL) { |
1da177e4 LT |
227 | icmp_send(head, ICMP_TIME_EXCEEDED, ICMP_EXC_FRAGTIME, 0); |
228 | dev_put(head->dev); | |
229 | } | |
230 | } | |
231 | out: | |
5ab11c98 | 232 | spin_unlock(&qp->q.lock); |
4b6cb5d8 | 233 | ipq_put(qp); |
1da177e4 LT |
234 | } |
235 | ||
abd6523d PE |
236 | /* Find the correct entry in the "incomplete datagrams" queue for |
237 | * this IP datagram, and create new one, if nothing is found. | |
238 | */ | |
ac18e750 | 239 | static inline struct ipq *ip_find(struct net *net, struct iphdr *iph, u32 user) |
1da177e4 | 240 | { |
c6fda282 PE |
241 | struct inet_frag_queue *q; |
242 | struct ip4_create_arg arg; | |
abd6523d | 243 | unsigned int hash; |
1da177e4 | 244 | |
c6fda282 PE |
245 | arg.iph = iph; |
246 | arg.user = user; | |
abd6523d | 247 | hash = ipqhashfn(iph->id, iph->saddr, iph->daddr, iph->protocol); |
1da177e4 | 248 | |
ac18e750 | 249 | q = inet_frag_find(&net->ipv4.frags, &ip4_frags, &arg, hash); |
c6fda282 PE |
250 | if (q == NULL) |
251 | goto out_nomem; | |
1da177e4 | 252 | |
c6fda282 | 253 | return container_of(q, struct ipq, q); |
1da177e4 LT |
254 | |
255 | out_nomem: | |
64ce2073 | 256 | LIMIT_NETDEBUG(KERN_ERR "ip_frag_create: no memory left !\n"); |
1da177e4 LT |
257 | return NULL; |
258 | } | |
259 | ||
89cee8b1 HX |
260 | /* Is the fragment too far ahead to be part of ipq? */ |
261 | static inline int ip_frag_too_far(struct ipq *qp) | |
262 | { | |
263 | struct inet_peer *peer = qp->peer; | |
264 | unsigned int max = sysctl_ipfrag_max_dist; | |
265 | unsigned int start, end; | |
266 | ||
267 | int rc; | |
268 | ||
269 | if (!peer || !max) | |
270 | return 0; | |
271 | ||
272 | start = qp->rid; | |
273 | end = atomic_inc_return(&peer->rid); | |
274 | qp->rid = end; | |
275 | ||
5ab11c98 | 276 | rc = qp->q.fragments && (end - start) > max; |
89cee8b1 HX |
277 | |
278 | if (rc) { | |
279 | IP_INC_STATS_BH(IPSTATS_MIB_REASMFAILS); | |
280 | } | |
281 | ||
282 | return rc; | |
283 | } | |
284 | ||
285 | static int ip_frag_reinit(struct ipq *qp) | |
286 | { | |
287 | struct sk_buff *fp; | |
288 | ||
04128f23 | 289 | if (!mod_timer(&qp->q.timer, jiffies + ip4_frags_ctl.timeout)) { |
5ab11c98 | 290 | atomic_inc(&qp->q.refcnt); |
89cee8b1 HX |
291 | return -ETIMEDOUT; |
292 | } | |
293 | ||
5ab11c98 | 294 | fp = qp->q.fragments; |
89cee8b1 HX |
295 | do { |
296 | struct sk_buff *xp = fp->next; | |
297 | frag_kfree_skb(fp, NULL); | |
298 | fp = xp; | |
299 | } while (fp); | |
300 | ||
5ab11c98 PE |
301 | qp->q.last_in = 0; |
302 | qp->q.len = 0; | |
303 | qp->q.meat = 0; | |
304 | qp->q.fragments = NULL; | |
89cee8b1 HX |
305 | qp->iif = 0; |
306 | ||
307 | return 0; | |
308 | } | |
309 | ||
1da177e4 | 310 | /* Add new segment to existing queue. */ |
1706d587 | 311 | static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb) |
1da177e4 LT |
312 | { |
313 | struct sk_buff *prev, *next; | |
1706d587 | 314 | struct net_device *dev; |
1da177e4 LT |
315 | int flags, offset; |
316 | int ihl, end; | |
1706d587 | 317 | int err = -ENOENT; |
1da177e4 | 318 | |
5ab11c98 | 319 | if (qp->q.last_in & COMPLETE) |
1da177e4 LT |
320 | goto err; |
321 | ||
89cee8b1 | 322 | if (!(IPCB(skb)->flags & IPSKB_FRAG_COMPLETE) && |
1706d587 HX |
323 | unlikely(ip_frag_too_far(qp)) && |
324 | unlikely(err = ip_frag_reinit(qp))) { | |
89cee8b1 HX |
325 | ipq_kill(qp); |
326 | goto err; | |
327 | } | |
328 | ||
eddc9ec5 | 329 | offset = ntohs(ip_hdr(skb)->frag_off); |
1da177e4 LT |
330 | flags = offset & ~IP_OFFSET; |
331 | offset &= IP_OFFSET; | |
332 | offset <<= 3; /* offset is in 8-byte chunks */ | |
c9bdd4b5 | 333 | ihl = ip_hdrlen(skb); |
1da177e4 LT |
334 | |
335 | /* Determine the position of this fragment. */ | |
e905a9ed | 336 | end = offset + skb->len - ihl; |
1706d587 | 337 | err = -EINVAL; |
1da177e4 LT |
338 | |
339 | /* Is this the final fragment? */ | |
340 | if ((flags & IP_MF) == 0) { | |
341 | /* If we already have some bits beyond end | |
342 | * or have different end, the segment is corrrupted. | |
343 | */ | |
5ab11c98 PE |
344 | if (end < qp->q.len || |
345 | ((qp->q.last_in & LAST_IN) && end != qp->q.len)) | |
1da177e4 | 346 | goto err; |
5ab11c98 PE |
347 | qp->q.last_in |= LAST_IN; |
348 | qp->q.len = end; | |
1da177e4 LT |
349 | } else { |
350 | if (end&7) { | |
351 | end &= ~7; | |
352 | if (skb->ip_summed != CHECKSUM_UNNECESSARY) | |
353 | skb->ip_summed = CHECKSUM_NONE; | |
354 | } | |
5ab11c98 | 355 | if (end > qp->q.len) { |
1da177e4 | 356 | /* Some bits beyond end -> corruption. */ |
5ab11c98 | 357 | if (qp->q.last_in & LAST_IN) |
1da177e4 | 358 | goto err; |
5ab11c98 | 359 | qp->q.len = end; |
1da177e4 LT |
360 | } |
361 | } | |
362 | if (end == offset) | |
363 | goto err; | |
364 | ||
1706d587 | 365 | err = -ENOMEM; |
1da177e4 LT |
366 | if (pskb_pull(skb, ihl) == NULL) |
367 | goto err; | |
1706d587 HX |
368 | |
369 | err = pskb_trim_rcsum(skb, end - offset); | |
370 | if (err) | |
1da177e4 LT |
371 | goto err; |
372 | ||
373 | /* Find out which fragments are in front and at the back of us | |
374 | * in the chain of fragments so far. We must know where to put | |
375 | * this fragment, right? | |
376 | */ | |
377 | prev = NULL; | |
5ab11c98 | 378 | for (next = qp->q.fragments; next != NULL; next = next->next) { |
1da177e4 LT |
379 | if (FRAG_CB(next)->offset >= offset) |
380 | break; /* bingo! */ | |
381 | prev = next; | |
382 | } | |
383 | ||
384 | /* We found where to put this one. Check for overlap with | |
385 | * preceding fragment, and, if needed, align things so that | |
386 | * any overlaps are eliminated. | |
387 | */ | |
388 | if (prev) { | |
389 | int i = (FRAG_CB(prev)->offset + prev->len) - offset; | |
390 | ||
391 | if (i > 0) { | |
392 | offset += i; | |
1706d587 | 393 | err = -EINVAL; |
1da177e4 LT |
394 | if (end <= offset) |
395 | goto err; | |
1706d587 | 396 | err = -ENOMEM; |
1da177e4 LT |
397 | if (!pskb_pull(skb, i)) |
398 | goto err; | |
399 | if (skb->ip_summed != CHECKSUM_UNNECESSARY) | |
400 | skb->ip_summed = CHECKSUM_NONE; | |
401 | } | |
402 | } | |
403 | ||
1706d587 HX |
404 | err = -ENOMEM; |
405 | ||
1da177e4 LT |
406 | while (next && FRAG_CB(next)->offset < end) { |
407 | int i = end - FRAG_CB(next)->offset; /* overlap is 'i' bytes */ | |
408 | ||
409 | if (i < next->len) { | |
410 | /* Eat head of the next overlapped fragment | |
411 | * and leave the loop. The next ones cannot overlap. | |
412 | */ | |
413 | if (!pskb_pull(next, i)) | |
414 | goto err; | |
415 | FRAG_CB(next)->offset += i; | |
5ab11c98 | 416 | qp->q.meat -= i; |
1da177e4 LT |
417 | if (next->ip_summed != CHECKSUM_UNNECESSARY) |
418 | next->ip_summed = CHECKSUM_NONE; | |
419 | break; | |
420 | } else { | |
421 | struct sk_buff *free_it = next; | |
422 | ||
47c6bf77 | 423 | /* Old fragment is completely overridden with |
1da177e4 LT |
424 | * new one drop it. |
425 | */ | |
426 | next = next->next; | |
427 | ||
428 | if (prev) | |
429 | prev->next = next; | |
430 | else | |
5ab11c98 | 431 | qp->q.fragments = next; |
1da177e4 | 432 | |
5ab11c98 | 433 | qp->q.meat -= free_it->len; |
1da177e4 LT |
434 | frag_kfree_skb(free_it, NULL); |
435 | } | |
436 | } | |
437 | ||
438 | FRAG_CB(skb)->offset = offset; | |
439 | ||
440 | /* Insert this fragment in the chain of fragments. */ | |
441 | skb->next = next; | |
442 | if (prev) | |
443 | prev->next = skb; | |
444 | else | |
5ab11c98 | 445 | qp->q.fragments = skb; |
1da177e4 | 446 | |
1706d587 HX |
447 | dev = skb->dev; |
448 | if (dev) { | |
449 | qp->iif = dev->ifindex; | |
450 | skb->dev = NULL; | |
451 | } | |
5ab11c98 PE |
452 | qp->q.stamp = skb->tstamp; |
453 | qp->q.meat += skb->len; | |
7eb95156 | 454 | atomic_add(skb->truesize, &ip4_frags.mem); |
1da177e4 | 455 | if (offset == 0) |
5ab11c98 | 456 | qp->q.last_in |= FIRST_IN; |
1da177e4 | 457 | |
5ab11c98 | 458 | if (qp->q.last_in == (FIRST_IN | LAST_IN) && qp->q.meat == qp->q.len) |
1706d587 HX |
459 | return ip_frag_reasm(qp, prev, dev); |
460 | ||
7eb95156 PE |
461 | write_lock(&ip4_frags.lock); |
462 | list_move_tail(&qp->q.lru_list, &ip4_frags.lru_list); | |
463 | write_unlock(&ip4_frags.lock); | |
1706d587 | 464 | return -EINPROGRESS; |
1da177e4 LT |
465 | |
466 | err: | |
467 | kfree_skb(skb); | |
1706d587 | 468 | return err; |
1da177e4 LT |
469 | } |
470 | ||
471 | ||
472 | /* Build a new IP datagram from all its fragments. */ | |
473 | ||
1706d587 HX |
474 | static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev, |
475 | struct net_device *dev) | |
1da177e4 LT |
476 | { |
477 | struct iphdr *iph; | |
5ab11c98 | 478 | struct sk_buff *fp, *head = qp->q.fragments; |
1da177e4 LT |
479 | int len; |
480 | int ihlen; | |
1706d587 | 481 | int err; |
1da177e4 LT |
482 | |
483 | ipq_kill(qp); | |
484 | ||
1706d587 HX |
485 | /* Make the one we just received the head. */ |
486 | if (prev) { | |
487 | head = prev->next; | |
488 | fp = skb_clone(head, GFP_ATOMIC); | |
1706d587 HX |
489 | if (!fp) |
490 | goto out_nomem; | |
491 | ||
492 | fp->next = head->next; | |
493 | prev->next = fp; | |
494 | ||
5ab11c98 PE |
495 | skb_morph(head, qp->q.fragments); |
496 | head->next = qp->q.fragments->next; | |
1706d587 | 497 | |
5ab11c98 PE |
498 | kfree_skb(qp->q.fragments); |
499 | qp->q.fragments = head; | |
1706d587 HX |
500 | } |
501 | ||
1da177e4 LT |
502 | BUG_TRAP(head != NULL); |
503 | BUG_TRAP(FRAG_CB(head)->offset == 0); | |
504 | ||
505 | /* Allocate a new buffer for the datagram. */ | |
c9bdd4b5 | 506 | ihlen = ip_hdrlen(head); |
5ab11c98 | 507 | len = ihlen + qp->q.len; |
1da177e4 | 508 | |
1706d587 | 509 | err = -E2BIG; |
132adf54 | 510 | if (len > 65535) |
1da177e4 LT |
511 | goto out_oversize; |
512 | ||
513 | /* Head of list must not be cloned. */ | |
514 | if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC)) | |
515 | goto out_nomem; | |
516 | ||
517 | /* If the first fragment is fragmented itself, we split | |
518 | * it to two chunks: the first with data and paged part | |
519 | * and the second, holding only fragments. */ | |
520 | if (skb_shinfo(head)->frag_list) { | |
521 | struct sk_buff *clone; | |
522 | int i, plen = 0; | |
523 | ||
524 | if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL) | |
525 | goto out_nomem; | |
526 | clone->next = head->next; | |
527 | head->next = clone; | |
528 | skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list; | |
529 | skb_shinfo(head)->frag_list = NULL; | |
530 | for (i=0; i<skb_shinfo(head)->nr_frags; i++) | |
531 | plen += skb_shinfo(head)->frags[i].size; | |
532 | clone->len = clone->data_len = head->data_len - plen; | |
533 | head->data_len -= clone->len; | |
534 | head->len -= clone->len; | |
535 | clone->csum = 0; | |
536 | clone->ip_summed = head->ip_summed; | |
7eb95156 | 537 | atomic_add(clone->truesize, &ip4_frags.mem); |
1da177e4 LT |
538 | } |
539 | ||
540 | skb_shinfo(head)->frag_list = head->next; | |
d56f90a7 | 541 | skb_push(head, head->data - skb_network_header(head)); |
7eb95156 | 542 | atomic_sub(head->truesize, &ip4_frags.mem); |
1da177e4 LT |
543 | |
544 | for (fp=head->next; fp; fp = fp->next) { | |
545 | head->data_len += fp->len; | |
546 | head->len += fp->len; | |
547 | if (head->ip_summed != fp->ip_summed) | |
548 | head->ip_summed = CHECKSUM_NONE; | |
84fa7933 | 549 | else if (head->ip_summed == CHECKSUM_COMPLETE) |
1da177e4 LT |
550 | head->csum = csum_add(head->csum, fp->csum); |
551 | head->truesize += fp->truesize; | |
7eb95156 | 552 | atomic_sub(fp->truesize, &ip4_frags.mem); |
1da177e4 LT |
553 | } |
554 | ||
555 | head->next = NULL; | |
556 | head->dev = dev; | |
5ab11c98 | 557 | head->tstamp = qp->q.stamp; |
1da177e4 | 558 | |
eddc9ec5 | 559 | iph = ip_hdr(head); |
1da177e4 LT |
560 | iph->frag_off = 0; |
561 | iph->tot_len = htons(len); | |
562 | IP_INC_STATS_BH(IPSTATS_MIB_REASMOKS); | |
5ab11c98 | 563 | qp->q.fragments = NULL; |
1706d587 | 564 | return 0; |
1da177e4 LT |
565 | |
566 | out_nomem: | |
e905a9ed | 567 | LIMIT_NETDEBUG(KERN_ERR "IP: queue_glue: no memory for gluing " |
64ce2073 | 568 | "queue %p\n", qp); |
45542479 | 569 | err = -ENOMEM; |
1da177e4 LT |
570 | goto out_fail; |
571 | out_oversize: | |
572 | if (net_ratelimit()) | |
573 | printk(KERN_INFO | |
574 | "Oversized IP packet from %d.%d.%d.%d.\n", | |
575 | NIPQUAD(qp->saddr)); | |
576 | out_fail: | |
577 | IP_INC_STATS_BH(IPSTATS_MIB_REASMFAILS); | |
1706d587 | 578 | return err; |
1da177e4 LT |
579 | } |
580 | ||
581 | /* Process an incoming IP datagram fragment. */ | |
776c729e | 582 | int ip_defrag(struct sk_buff *skb, u32 user) |
1da177e4 | 583 | { |
1da177e4 | 584 | struct ipq *qp; |
ac18e750 | 585 | struct net *net; |
e905a9ed | 586 | |
1da177e4 LT |
587 | IP_INC_STATS_BH(IPSTATS_MIB_REASMREQDS); |
588 | ||
ac18e750 | 589 | net = skb->dev->nd_net; |
1da177e4 | 590 | /* Start by cleaning up the memory. */ |
04128f23 | 591 | if (atomic_read(&ip4_frags.mem) > ip4_frags_ctl.high_thresh) |
1da177e4 LT |
592 | ip_evictor(); |
593 | ||
1da177e4 | 594 | /* Lookup (or create) queue header */ |
ac18e750 | 595 | if ((qp = ip_find(net, ip_hdr(skb), user)) != NULL) { |
1706d587 | 596 | int ret; |
1da177e4 | 597 | |
5ab11c98 | 598 | spin_lock(&qp->q.lock); |
1da177e4 | 599 | |
1706d587 | 600 | ret = ip_frag_queue(qp, skb); |
1da177e4 | 601 | |
5ab11c98 | 602 | spin_unlock(&qp->q.lock); |
4b6cb5d8 | 603 | ipq_put(qp); |
776c729e | 604 | return ret; |
1da177e4 LT |
605 | } |
606 | ||
607 | IP_INC_STATS_BH(IPSTATS_MIB_REASMFAILS); | |
608 | kfree_skb(skb); | |
776c729e | 609 | return -ENOMEM; |
1da177e4 LT |
610 | } |
611 | ||
8d8354d2 PE |
612 | #ifdef CONFIG_SYSCTL |
613 | static int zero; | |
614 | ||
615 | static struct ctl_table ip4_frags_ctl_table[] = { | |
616 | { | |
617 | .ctl_name = NET_IPV4_IPFRAG_HIGH_THRESH, | |
618 | .procname = "ipfrag_high_thresh", | |
619 | .data = &ip4_frags_ctl.high_thresh, | |
620 | .maxlen = sizeof(int), | |
621 | .mode = 0644, | |
622 | .proc_handler = &proc_dointvec | |
623 | }, | |
624 | { | |
625 | .ctl_name = NET_IPV4_IPFRAG_LOW_THRESH, | |
626 | .procname = "ipfrag_low_thresh", | |
627 | .data = &ip4_frags_ctl.low_thresh, | |
628 | .maxlen = sizeof(int), | |
629 | .mode = 0644, | |
630 | .proc_handler = &proc_dointvec | |
631 | }, | |
632 | { | |
633 | .ctl_name = NET_IPV4_IPFRAG_TIME, | |
634 | .procname = "ipfrag_time", | |
635 | .data = &ip4_frags_ctl.timeout, | |
636 | .maxlen = sizeof(int), | |
637 | .mode = 0644, | |
638 | .proc_handler = &proc_dointvec_jiffies, | |
639 | .strategy = &sysctl_jiffies | |
640 | }, | |
641 | { | |
642 | .ctl_name = NET_IPV4_IPFRAG_SECRET_INTERVAL, | |
643 | .procname = "ipfrag_secret_interval", | |
644 | .data = &ip4_frags_ctl.secret_interval, | |
645 | .maxlen = sizeof(int), | |
646 | .mode = 0644, | |
647 | .proc_handler = &proc_dointvec_jiffies, | |
648 | .strategy = &sysctl_jiffies | |
649 | }, | |
650 | { | |
651 | .procname = "ipfrag_max_dist", | |
652 | .data = &sysctl_ipfrag_max_dist, | |
653 | .maxlen = sizeof(int), | |
654 | .mode = 0644, | |
655 | .proc_handler = &proc_dointvec_minmax, | |
656 | .extra1 = &zero | |
657 | }, | |
658 | { } | |
659 | }; | |
660 | ||
661 | static int ip4_frags_ctl_register(struct net *net) | |
662 | { | |
663 | struct ctl_table_header *hdr; | |
664 | ||
665 | hdr = register_net_sysctl_table(net, net_ipv4_ctl_path, | |
666 | ip4_frags_ctl_table); | |
667 | return hdr == NULL ? -ENOMEM : 0; | |
668 | } | |
669 | #else | |
670 | static inline int ip4_frags_ctl_register(struct net *net) | |
671 | { | |
672 | return 0; | |
673 | } | |
674 | #endif | |
675 | ||
676 | static int ipv4_frags_init_net(struct net *net) | |
677 | { | |
e5a2bb84 PE |
678 | inet_frags_init_net(&net->ipv4.frags); |
679 | ||
8d8354d2 PE |
680 | return ip4_frags_ctl_register(net); |
681 | } | |
682 | ||
b7aa0bf7 | 683 | void __init ipfrag_init(void) |
1da177e4 | 684 | { |
8d8354d2 | 685 | ipv4_frags_init_net(&init_net); |
04128f23 | 686 | ip4_frags.ctl = &ip4_frags_ctl; |
321a3a99 | 687 | ip4_frags.hashfn = ip4_hashfn; |
c6fda282 | 688 | ip4_frags.constructor = ip4_frag_init; |
1e4b8287 PE |
689 | ip4_frags.destructor = ip4_frag_free; |
690 | ip4_frags.skb_free = NULL; | |
691 | ip4_frags.qsize = sizeof(struct ipq); | |
abd6523d | 692 | ip4_frags.match = ip4_frag_match; |
e521db9d | 693 | ip4_frags.frag_expire = ip_expire; |
7eb95156 | 694 | inet_frags_init(&ip4_frags); |
1da177e4 LT |
695 | } |
696 | ||
697 | EXPORT_SYMBOL(ip_defrag); |