Commit | Line | Data |
---|---|---|
9fb9cbb1 YK |
1 | /* |
2 | * IPv6 fragment reassembly for connection tracking | |
3 | * | |
4 | * Copyright (C)2004 USAGI/WIDE Project | |
5 | * | |
6 | * Author: | |
7 | * Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp> | |
8 | * | |
9 | * Based on: net/ipv6/reassembly.c | |
10 | * | |
11 | * This program is free software; you can redistribute it and/or | |
12 | * modify it under the terms of the GNU General Public License | |
13 | * as published by the Free Software Foundation; either version | |
14 | * 2 of the License, or (at your option) any later version. | |
15 | */ | |
16 | ||
9fb9cbb1 YK |
17 | #include <linux/errno.h> |
18 | #include <linux/types.h> | |
19 | #include <linux/string.h> | |
20 | #include <linux/socket.h> | |
21 | #include <linux/sockios.h> | |
22 | #include <linux/jiffies.h> | |
23 | #include <linux/net.h> | |
24 | #include <linux/list.h> | |
25 | #include <linux/netdevice.h> | |
26 | #include <linux/in6.h> | |
27 | #include <linux/ipv6.h> | |
28 | #include <linux/icmpv6.h> | |
29 | #include <linux/random.h> | |
9fb9cbb1 YK |
30 | |
31 | #include <net/sock.h> | |
32 | #include <net/snmp.h> | |
5ab11c98 | 33 | #include <net/inet_frag.h> |
9fb9cbb1 YK |
34 | |
35 | #include <net/ipv6.h> | |
36 | #include <net/protocol.h> | |
37 | #include <net/transp_v6.h> | |
38 | #include <net/rawv6.h> | |
39 | #include <net/ndisc.h> | |
40 | #include <net/addrconf.h> | |
99fa5f53 | 41 | #include <net/netfilter/ipv6/nf_conntrack_ipv6.h> |
9fb9cbb1 YK |
42 | #include <linux/sysctl.h> |
43 | #include <linux/netfilter.h> | |
44 | #include <linux/netfilter_ipv6.h> | |
45 | #include <linux/kernel.h> | |
46 | #include <linux/module.h> | |
47 | ||
9fb9cbb1 YK |
48 | #define NF_CT_FRAG6_HIGH_THRESH 262144 /* == 256*1024 */ |
49 | #define NF_CT_FRAG6_LOW_THRESH 196608 /* == 192*1024 */ | |
50 | #define NF_CT_FRAG6_TIMEOUT IPV6_FRAG_TIMEOUT | |
51 | ||
9fb9cbb1 YK |
52 | struct nf_ct_frag6_skb_cb |
53 | { | |
54 | struct inet6_skb_parm h; | |
55 | int offset; | |
56 | struct sk_buff *orig; | |
57 | }; | |
58 | ||
59 | #define NFCT_FRAG6_CB(skb) ((struct nf_ct_frag6_skb_cb*)((skb)->cb)) | |
60 | ||
61 | struct nf_ct_frag6_queue | |
62 | { | |
5ab11c98 | 63 | struct inet_frag_queue q; |
9fb9cbb1 | 64 | |
bff9a89b | 65 | __be32 id; /* fragment id */ |
9fb9cbb1 YK |
66 | struct in6_addr saddr; |
67 | struct in6_addr daddr; | |
68 | ||
9fb9cbb1 | 69 | unsigned int csum; |
9fb9cbb1 | 70 | __u16 nhoffset; |
9fb9cbb1 YK |
71 | }; |
72 | ||
7eb95156 | 73 | static struct inet_frags nf_frags; |
ac18e750 | 74 | static struct netns_frags nf_init_frags; |
9fb9cbb1 | 75 | |
8d8354d2 PE |
76 | #ifdef CONFIG_SYSCTL |
77 | struct ctl_table nf_ct_ipv6_sysctl_table[] = { | |
78 | { | |
79 | .procname = "nf_conntrack_frag6_timeout", | |
b2fd5321 | 80 | .data = &nf_init_frags.timeout, |
8d8354d2 PE |
81 | .maxlen = sizeof(unsigned int), |
82 | .mode = 0644, | |
6d9f239a | 83 | .proc_handler = proc_dointvec_jiffies, |
8d8354d2 PE |
84 | }, |
85 | { | |
86 | .ctl_name = NET_NF_CONNTRACK_FRAG6_LOW_THRESH, | |
87 | .procname = "nf_conntrack_frag6_low_thresh", | |
e31e0bdc | 88 | .data = &nf_init_frags.low_thresh, |
8d8354d2 PE |
89 | .maxlen = sizeof(unsigned int), |
90 | .mode = 0644, | |
6d9f239a | 91 | .proc_handler = proc_dointvec, |
8d8354d2 PE |
92 | }, |
93 | { | |
94 | .ctl_name = NET_NF_CONNTRACK_FRAG6_HIGH_THRESH, | |
95 | .procname = "nf_conntrack_frag6_high_thresh", | |
e31e0bdc | 96 | .data = &nf_init_frags.high_thresh, |
8d8354d2 PE |
97 | .maxlen = sizeof(unsigned int), |
98 | .mode = 0644, | |
6d9f239a | 99 | .proc_handler = proc_dointvec, |
8d8354d2 PE |
100 | }, |
101 | { .ctl_name = 0 } | |
102 | }; | |
103 | #endif | |
104 | ||
321a3a99 | 105 | static unsigned int nf_hashfn(struct inet_frag_queue *q) |
9fb9cbb1 | 106 | { |
58c0fb0d | 107 | const struct nf_ct_frag6_queue *nq; |
9fb9cbb1 | 108 | |
321a3a99 | 109 | nq = container_of(q, struct nf_ct_frag6_queue, q); |
93c8b90f | 110 | return inet6_hash_frag(nq->id, &nq->saddr, &nq->daddr, nf_frags.rnd); |
9fb9cbb1 YK |
111 | } |
112 | ||
1e4b8287 PE |
113 | static void nf_skb_free(struct sk_buff *skb) |
114 | { | |
115 | if (NFCT_FRAG6_CB(skb)->orig) | |
116 | kfree_skb(NFCT_FRAG6_CB(skb)->orig); | |
117 | } | |
118 | ||
9fb9cbb1 | 119 | /* Memory Tracking Functions. */ |
1ba430bc | 120 | static inline void frag_kfree_skb(struct sk_buff *skb, unsigned int *work) |
9fb9cbb1 | 121 | { |
1ba430bc YK |
122 | if (work) |
123 | *work -= skb->truesize; | |
6ddc0822 | 124 | atomic_sub(skb->truesize, &nf_init_frags.mem); |
1e4b8287 | 125 | nf_skb_free(skb); |
9fb9cbb1 YK |
126 | kfree_skb(skb); |
127 | } | |
128 | ||
9fb9cbb1 YK |
129 | /* Destruction primitives. */ |
130 | ||
4b6cb5d8 | 131 | static __inline__ void fq_put(struct nf_ct_frag6_queue *fq) |
9fb9cbb1 | 132 | { |
762cc408 | 133 | inet_frag_put(&fq->q, &nf_frags); |
9fb9cbb1 YK |
134 | } |
135 | ||
136 | /* Kill fq entry. It is not destroyed immediately, | |
137 | * because caller (and someone more) holds reference count. | |
138 | */ | |
139 | static __inline__ void fq_kill(struct nf_ct_frag6_queue *fq) | |
140 | { | |
277e650d | 141 | inet_frag_kill(&fq->q, &nf_frags); |
9fb9cbb1 YK |
142 | } |
143 | ||
144 | static void nf_ct_frag6_evictor(void) | |
145 | { | |
e8e16b70 | 146 | local_bh_disable(); |
6ddc0822 | 147 | inet_frag_evictor(&nf_init_frags, &nf_frags); |
e8e16b70 | 148 | local_bh_enable(); |
9fb9cbb1 YK |
149 | } |
150 | ||
151 | static void nf_ct_frag6_expire(unsigned long data) | |
152 | { | |
e521db9d PE |
153 | struct nf_ct_frag6_queue *fq; |
154 | ||
155 | fq = container_of((struct inet_frag_queue *)data, | |
156 | struct nf_ct_frag6_queue, q); | |
9fb9cbb1 | 157 | |
5ab11c98 | 158 | spin_lock(&fq->q.lock); |
9fb9cbb1 | 159 | |
bc578a54 | 160 | if (fq->q.last_in & INET_FRAG_COMPLETE) |
9fb9cbb1 YK |
161 | goto out; |
162 | ||
163 | fq_kill(fq); | |
164 | ||
165 | out: | |
5ab11c98 | 166 | spin_unlock(&fq->q.lock); |
4b6cb5d8 | 167 | fq_put(fq); |
9fb9cbb1 YK |
168 | } |
169 | ||
170 | /* Creation primitives. */ | |
171 | ||
abd6523d PE |
172 | static __inline__ struct nf_ct_frag6_queue * |
173 | fq_find(__be32 id, struct in6_addr *src, struct in6_addr *dst) | |
9fb9cbb1 | 174 | { |
2588fe1d | 175 | struct inet_frag_queue *q; |
c6fda282 | 176 | struct ip6_create_arg arg; |
abd6523d | 177 | unsigned int hash; |
9fb9cbb1 | 178 | |
c6fda282 PE |
179 | arg.id = id; |
180 | arg.src = src; | |
181 | arg.dst = dst; | |
9a375803 PE |
182 | |
183 | read_lock_bh(&nf_frags.lock); | |
93c8b90f | 184 | hash = inet6_hash_frag(id, src, dst, nf_frags.rnd); |
9fb9cbb1 | 185 | |
ac18e750 | 186 | q = inet_frag_find(&nf_init_frags, &nf_frags, &arg, hash); |
b9c69896 | 187 | local_bh_enable(); |
c6fda282 | 188 | if (q == NULL) |
9fb9cbb1 | 189 | goto oom; |
9fb9cbb1 | 190 | |
c6fda282 | 191 | return container_of(q, struct nf_ct_frag6_queue, q); |
9fb9cbb1 YK |
192 | |
193 | oom: | |
c6fda282 | 194 | pr_debug("Can't alloc new queue\n"); |
9fb9cbb1 YK |
195 | return NULL; |
196 | } | |
197 | ||
9fb9cbb1 | 198 | |
1ab1457c | 199 | static int nf_ct_frag6_queue(struct nf_ct_frag6_queue *fq, struct sk_buff *skb, |
58c0fb0d | 200 | const struct frag_hdr *fhdr, int nhoff) |
9fb9cbb1 YK |
201 | { |
202 | struct sk_buff *prev, *next; | |
203 | int offset, end; | |
204 | ||
bc578a54 | 205 | if (fq->q.last_in & INET_FRAG_COMPLETE) { |
0d53778e | 206 | pr_debug("Allready completed\n"); |
9fb9cbb1 YK |
207 | goto err; |
208 | } | |
209 | ||
210 | offset = ntohs(fhdr->frag_off) & ~0x7; | |
0660e03f ACM |
211 | end = offset + (ntohs(ipv6_hdr(skb)->payload_len) - |
212 | ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1))); | |
9fb9cbb1 YK |
213 | |
214 | if ((unsigned int)end > IPV6_MAXPLEN) { | |
0d53778e | 215 | pr_debug("offset is too large.\n"); |
1ab1457c | 216 | return -1; |
9fb9cbb1 YK |
217 | } |
218 | ||
d56f90a7 ACM |
219 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
220 | const unsigned char *nh = skb_network_header(skb); | |
1ab1457c | 221 | skb->csum = csum_sub(skb->csum, |
d56f90a7 | 222 | csum_partial(nh, (u8 *)(fhdr + 1) - nh, |
9fb9cbb1 | 223 | 0)); |
d56f90a7 | 224 | } |
9fb9cbb1 YK |
225 | |
226 | /* Is this the final fragment? */ | |
227 | if (!(fhdr->frag_off & htons(IP6_MF))) { | |
228 | /* If we already have some bits beyond end | |
229 | * or have different end, the segment is corrupted. | |
230 | */ | |
5ab11c98 | 231 | if (end < fq->q.len || |
bc578a54 | 232 | ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len)) { |
0d53778e | 233 | pr_debug("already received last fragment\n"); |
9fb9cbb1 YK |
234 | goto err; |
235 | } | |
bc578a54 | 236 | fq->q.last_in |= INET_FRAG_LAST_IN; |
5ab11c98 | 237 | fq->q.len = end; |
9fb9cbb1 YK |
238 | } else { |
239 | /* Check if the fragment is rounded to 8 bytes. | |
240 | * Required by the RFC. | |
241 | */ | |
242 | if (end & 0x7) { | |
243 | /* RFC2460 says always send parameter problem in | |
244 | * this case. -DaveM | |
245 | */ | |
0d53778e | 246 | pr_debug("end of fragment not rounded to 8 bytes.\n"); |
9fb9cbb1 YK |
247 | return -1; |
248 | } | |
5ab11c98 | 249 | if (end > fq->q.len) { |
9fb9cbb1 | 250 | /* Some bits beyond end -> corruption. */ |
bc578a54 | 251 | if (fq->q.last_in & INET_FRAG_LAST_IN) { |
0d53778e | 252 | pr_debug("last packet already reached.\n"); |
9fb9cbb1 YK |
253 | goto err; |
254 | } | |
5ab11c98 | 255 | fq->q.len = end; |
9fb9cbb1 YK |
256 | } |
257 | } | |
258 | ||
259 | if (end == offset) | |
260 | goto err; | |
261 | ||
262 | /* Point into the IP datagram 'data' part. */ | |
263 | if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data)) { | |
0d53778e | 264 | pr_debug("queue: message is too short.\n"); |
9fb9cbb1 YK |
265 | goto err; |
266 | } | |
b38dfee3 | 267 | if (pskb_trim_rcsum(skb, end - offset)) { |
0d53778e | 268 | pr_debug("Can't trim\n"); |
b38dfee3 | 269 | goto err; |
9fb9cbb1 YK |
270 | } |
271 | ||
272 | /* Find out which fragments are in front and at the back of us | |
273 | * in the chain of fragments so far. We must know where to put | |
274 | * this fragment, right? | |
275 | */ | |
276 | prev = NULL; | |
5ab11c98 | 277 | for (next = fq->q.fragments; next != NULL; next = next->next) { |
9fb9cbb1 YK |
278 | if (NFCT_FRAG6_CB(next)->offset >= offset) |
279 | break; /* bingo! */ | |
280 | prev = next; | |
281 | } | |
282 | ||
283 | /* We found where to put this one. Check for overlap with | |
284 | * preceding fragment, and, if needed, align things so that | |
285 | * any overlaps are eliminated. | |
286 | */ | |
287 | if (prev) { | |
288 | int i = (NFCT_FRAG6_CB(prev)->offset + prev->len) - offset; | |
289 | ||
290 | if (i > 0) { | |
291 | offset += i; | |
292 | if (end <= offset) { | |
0d53778e | 293 | pr_debug("overlap\n"); |
9fb9cbb1 YK |
294 | goto err; |
295 | } | |
296 | if (!pskb_pull(skb, i)) { | |
0d53778e | 297 | pr_debug("Can't pull\n"); |
9fb9cbb1 YK |
298 | goto err; |
299 | } | |
300 | if (skb->ip_summed != CHECKSUM_UNNECESSARY) | |
301 | skb->ip_summed = CHECKSUM_NONE; | |
302 | } | |
303 | } | |
304 | ||
305 | /* Look for overlap with succeeding segments. | |
306 | * If we can merge fragments, do it. | |
307 | */ | |
308 | while (next && NFCT_FRAG6_CB(next)->offset < end) { | |
309 | /* overlap is 'i' bytes */ | |
310 | int i = end - NFCT_FRAG6_CB(next)->offset; | |
311 | ||
312 | if (i < next->len) { | |
313 | /* Eat head of the next overlapped fragment | |
314 | * and leave the loop. The next ones cannot overlap. | |
315 | */ | |
0d53778e | 316 | pr_debug("Eat head of the overlapped parts.: %d", i); |
9fb9cbb1 YK |
317 | if (!pskb_pull(next, i)) |
318 | goto err; | |
319 | ||
320 | /* next fragment */ | |
321 | NFCT_FRAG6_CB(next)->offset += i; | |
5ab11c98 | 322 | fq->q.meat -= i; |
9fb9cbb1 YK |
323 | if (next->ip_summed != CHECKSUM_UNNECESSARY) |
324 | next->ip_summed = CHECKSUM_NONE; | |
325 | break; | |
326 | } else { | |
327 | struct sk_buff *free_it = next; | |
328 | ||
329 | /* Old fragmnet is completely overridden with | |
330 | * new one drop it. | |
331 | */ | |
332 | next = next->next; | |
333 | ||
334 | if (prev) | |
335 | prev->next = next; | |
336 | else | |
5ab11c98 | 337 | fq->q.fragments = next; |
9fb9cbb1 | 338 | |
5ab11c98 | 339 | fq->q.meat -= free_it->len; |
1ba430bc | 340 | frag_kfree_skb(free_it, NULL); |
9fb9cbb1 YK |
341 | } |
342 | } | |
343 | ||
344 | NFCT_FRAG6_CB(skb)->offset = offset; | |
345 | ||
346 | /* Insert this fragment in the chain of fragments. */ | |
347 | skb->next = next; | |
348 | if (prev) | |
349 | prev->next = skb; | |
350 | else | |
5ab11c98 | 351 | fq->q.fragments = skb; |
9fb9cbb1 YK |
352 | |
353 | skb->dev = NULL; | |
5ab11c98 PE |
354 | fq->q.stamp = skb->tstamp; |
355 | fq->q.meat += skb->len; | |
6ddc0822 | 356 | atomic_add(skb->truesize, &nf_init_frags.mem); |
9fb9cbb1 YK |
357 | |
358 | /* The first fragment. | |
359 | * nhoffset is obtained from the first fragment, of course. | |
360 | */ | |
361 | if (offset == 0) { | |
362 | fq->nhoffset = nhoff; | |
bc578a54 | 363 | fq->q.last_in |= INET_FRAG_FIRST_IN; |
9fb9cbb1 | 364 | } |
7eb95156 | 365 | write_lock(&nf_frags.lock); |
3140c25c | 366 | list_move_tail(&fq->q.lru_list, &nf_init_frags.lru_list); |
7eb95156 | 367 | write_unlock(&nf_frags.lock); |
9fb9cbb1 YK |
368 | return 0; |
369 | ||
370 | err: | |
371 | return -1; | |
372 | } | |
373 | ||
374 | /* | |
375 | * Check if this packet is complete. | |
376 | * Returns NULL on failure by any reason, and pointer | |
377 | * to current nexthdr field in reassembled frame. | |
378 | * | |
379 | * It is called with locked fq, and caller must check that | |
380 | * queue is eligible for reassembly i.e. it is not COMPLETE, | |
381 | * the last and the first frames arrived and all the bits are here. | |
382 | */ | |
383 | static struct sk_buff * | |
384 | nf_ct_frag6_reasm(struct nf_ct_frag6_queue *fq, struct net_device *dev) | |
385 | { | |
5ab11c98 | 386 | struct sk_buff *fp, *op, *head = fq->q.fragments; |
9fb9cbb1 YK |
387 | int payload_len; |
388 | ||
389 | fq_kill(fq); | |
390 | ||
547b792c IJ |
391 | WARN_ON(head == NULL); |
392 | WARN_ON(NFCT_FRAG6_CB(head)->offset != 0); | |
9fb9cbb1 YK |
393 | |
394 | /* Unfragmented part is taken from the first segment. */ | |
d56f90a7 | 395 | payload_len = ((head->data - skb_network_header(head)) - |
5ab11c98 | 396 | sizeof(struct ipv6hdr) + fq->q.len - |
d56f90a7 | 397 | sizeof(struct frag_hdr)); |
9fb9cbb1 | 398 | if (payload_len > IPV6_MAXPLEN) { |
0d53778e | 399 | pr_debug("payload len is too large.\n"); |
9fb9cbb1 YK |
400 | goto out_oversize; |
401 | } | |
402 | ||
403 | /* Head of list must not be cloned. */ | |
404 | if (skb_cloned(head) && pskb_expand_head(head, 0, 0, GFP_ATOMIC)) { | |
0d53778e | 405 | pr_debug("skb is cloned but can't expand head"); |
9fb9cbb1 YK |
406 | goto out_oom; |
407 | } | |
408 | ||
409 | /* If the first fragment is fragmented itself, we split | |
410 | * it to two chunks: the first with data and paged part | |
411 | * and the second, holding only fragments. */ | |
412 | if (skb_shinfo(head)->frag_list) { | |
413 | struct sk_buff *clone; | |
414 | int i, plen = 0; | |
415 | ||
416 | if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL) { | |
0d53778e | 417 | pr_debug("Can't alloc skb\n"); |
9fb9cbb1 YK |
418 | goto out_oom; |
419 | } | |
420 | clone->next = head->next; | |
421 | head->next = clone; | |
422 | skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list; | |
423 | skb_shinfo(head)->frag_list = NULL; | |
424 | for (i=0; i<skb_shinfo(head)->nr_frags; i++) | |
425 | plen += skb_shinfo(head)->frags[i].size; | |
426 | clone->len = clone->data_len = head->data_len - plen; | |
427 | head->data_len -= clone->len; | |
428 | head->len -= clone->len; | |
429 | clone->csum = 0; | |
430 | clone->ip_summed = head->ip_summed; | |
431 | ||
432 | NFCT_FRAG6_CB(clone)->orig = NULL; | |
6ddc0822 | 433 | atomic_add(clone->truesize, &nf_init_frags.mem); |
9fb9cbb1 YK |
434 | } |
435 | ||
436 | /* We have to remove fragment header from datagram and to relocate | |
437 | * header in order to calculate ICV correctly. */ | |
bff9b61c | 438 | skb_network_header(head)[fq->nhoffset] = skb_transport_header(head)[0]; |
1ab1457c | 439 | memmove(head->head + sizeof(struct frag_hdr), head->head, |
9fb9cbb1 | 440 | (head->data - head->head) - sizeof(struct frag_hdr)); |
b0e380b1 ACM |
441 | head->mac_header += sizeof(struct frag_hdr); |
442 | head->network_header += sizeof(struct frag_hdr); | |
9fb9cbb1 YK |
443 | |
444 | skb_shinfo(head)->frag_list = head->next; | |
badff6d0 | 445 | skb_reset_transport_header(head); |
d56f90a7 | 446 | skb_push(head, head->data - skb_network_header(head)); |
6ddc0822 | 447 | atomic_sub(head->truesize, &nf_init_frags.mem); |
9fb9cbb1 YK |
448 | |
449 | for (fp=head->next; fp; fp = fp->next) { | |
450 | head->data_len += fp->len; | |
451 | head->len += fp->len; | |
452 | if (head->ip_summed != fp->ip_summed) | |
453 | head->ip_summed = CHECKSUM_NONE; | |
84fa7933 | 454 | else if (head->ip_summed == CHECKSUM_COMPLETE) |
9fb9cbb1 YK |
455 | head->csum = csum_add(head->csum, fp->csum); |
456 | head->truesize += fp->truesize; | |
6ddc0822 | 457 | atomic_sub(fp->truesize, &nf_init_frags.mem); |
9fb9cbb1 YK |
458 | } |
459 | ||
460 | head->next = NULL; | |
461 | head->dev = dev; | |
5ab11c98 | 462 | head->tstamp = fq->q.stamp; |
0660e03f | 463 | ipv6_hdr(head)->payload_len = htons(payload_len); |
9fb9cbb1 YK |
464 | |
465 | /* Yes, and fold redundant checksum back. 8) */ | |
84fa7933 | 466 | if (head->ip_summed == CHECKSUM_COMPLETE) |
d56f90a7 | 467 | head->csum = csum_partial(skb_network_header(head), |
cfe1fc77 | 468 | skb_network_header_len(head), |
d56f90a7 | 469 | head->csum); |
9fb9cbb1 | 470 | |
5ab11c98 | 471 | fq->q.fragments = NULL; |
9fb9cbb1 YK |
472 | |
473 | /* all original skbs are linked into the NFCT_FRAG6_CB(head).orig */ | |
474 | fp = skb_shinfo(head)->frag_list; | |
475 | if (NFCT_FRAG6_CB(fp)->orig == NULL) | |
476 | /* at above code, head skb is divided into two skbs. */ | |
477 | fp = fp->next; | |
478 | ||
479 | op = NFCT_FRAG6_CB(head)->orig; | |
480 | for (; fp; fp = fp->next) { | |
481 | struct sk_buff *orig = NFCT_FRAG6_CB(fp)->orig; | |
482 | ||
483 | op->next = orig; | |
484 | op = orig; | |
485 | NFCT_FRAG6_CB(fp)->orig = NULL; | |
486 | } | |
487 | ||
488 | return head; | |
489 | ||
490 | out_oversize: | |
491 | if (net_ratelimit()) | |
492 | printk(KERN_DEBUG "nf_ct_frag6_reasm: payload len = %d\n", payload_len); | |
493 | goto out_fail; | |
494 | out_oom: | |
495 | if (net_ratelimit()) | |
496 | printk(KERN_DEBUG "nf_ct_frag6_reasm: no memory for reassembly\n"); | |
497 | out_fail: | |
498 | return NULL; | |
499 | } | |
500 | ||
501 | /* | |
502 | * find the header just before Fragment Header. | |
503 | * | |
504 | * if success return 0 and set ... | |
505 | * (*prevhdrp): the value of "Next Header Field" in the header | |
506 | * just before Fragment Header. | |
507 | * (*prevhoff): the offset of "Next Header Field" in the header | |
508 | * just before Fragment Header. | |
509 | * (*fhoff) : the offset of Fragment Header. | |
510 | * | |
511 | * Based on ipv6_skip_hdr() in net/ipv6/exthdr.c | |
512 | * | |
513 | */ | |
514 | static int | |
515 | find_prev_fhdr(struct sk_buff *skb, u8 *prevhdrp, int *prevhoff, int *fhoff) | |
516 | { | |
0660e03f | 517 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; |
6b88dd96 ACM |
518 | const int netoff = skb_network_offset(skb); |
519 | u8 prev_nhoff = netoff + offsetof(struct ipv6hdr, nexthdr); | |
520 | int start = netoff + sizeof(struct ipv6hdr); | |
9fb9cbb1 YK |
521 | int len = skb->len - start; |
522 | u8 prevhdr = NEXTHDR_IPV6; | |
523 | ||
1ab1457c YH |
524 | while (nexthdr != NEXTHDR_FRAGMENT) { |
525 | struct ipv6_opt_hdr hdr; | |
526 | int hdrlen; | |
9fb9cbb1 YK |
527 | |
528 | if (!ipv6_ext_hdr(nexthdr)) { | |
529 | return -1; | |
530 | } | |
1ab1457c | 531 | if (len < (int)sizeof(struct ipv6_opt_hdr)) { |
0d53778e | 532 | pr_debug("too short\n"); |
9fb9cbb1 YK |
533 | return -1; |
534 | } | |
1ab1457c | 535 | if (nexthdr == NEXTHDR_NONE) { |
0d53778e | 536 | pr_debug("next header is none\n"); |
9fb9cbb1 YK |
537 | return -1; |
538 | } | |
1ab1457c YH |
539 | if (skb_copy_bits(skb, start, &hdr, sizeof(hdr))) |
540 | BUG(); | |
541 | if (nexthdr == NEXTHDR_AUTH) | |
542 | hdrlen = (hdr.hdrlen+2)<<2; | |
543 | else | |
544 | hdrlen = ipv6_optlen(&hdr); | |
9fb9cbb1 YK |
545 | |
546 | prevhdr = nexthdr; | |
547 | prev_nhoff = start; | |
548 | ||
1ab1457c YH |
549 | nexthdr = hdr.nexthdr; |
550 | len -= hdrlen; | |
551 | start += hdrlen; | |
552 | } | |
9fb9cbb1 YK |
553 | |
554 | if (len < 0) | |
555 | return -1; | |
556 | ||
557 | *prevhdrp = prevhdr; | |
558 | *prevhoff = prev_nhoff; | |
559 | *fhoff = start; | |
560 | ||
561 | return 0; | |
562 | } | |
563 | ||
564 | struct sk_buff *nf_ct_frag6_gather(struct sk_buff *skb) | |
565 | { | |
1ab1457c | 566 | struct sk_buff *clone; |
9fb9cbb1 YK |
567 | struct net_device *dev = skb->dev; |
568 | struct frag_hdr *fhdr; | |
569 | struct nf_ct_frag6_queue *fq; | |
570 | struct ipv6hdr *hdr; | |
571 | int fhoff, nhoff; | |
572 | u8 prevhdr; | |
573 | struct sk_buff *ret_skb = NULL; | |
574 | ||
575 | /* Jumbo payload inhibits frag. header */ | |
0660e03f | 576 | if (ipv6_hdr(skb)->payload_len == 0) { |
0d53778e | 577 | pr_debug("payload len = 0\n"); |
9fb9cbb1 YK |
578 | return skb; |
579 | } | |
580 | ||
581 | if (find_prev_fhdr(skb, &prevhdr, &nhoff, &fhoff) < 0) | |
582 | return skb; | |
583 | ||
584 | clone = skb_clone(skb, GFP_ATOMIC); | |
585 | if (clone == NULL) { | |
0d53778e | 586 | pr_debug("Can't clone skb\n"); |
9fb9cbb1 YK |
587 | return skb; |
588 | } | |
589 | ||
590 | NFCT_FRAG6_CB(clone)->orig = skb; | |
591 | ||
592 | if (!pskb_may_pull(clone, fhoff + sizeof(*fhdr))) { | |
0d53778e | 593 | pr_debug("message is too short.\n"); |
9fb9cbb1 YK |
594 | goto ret_orig; |
595 | } | |
596 | ||
967b05f6 | 597 | skb_set_transport_header(clone, fhoff); |
0660e03f | 598 | hdr = ipv6_hdr(clone); |
bff9b61c | 599 | fhdr = (struct frag_hdr *)skb_transport_header(clone); |
9fb9cbb1 YK |
600 | |
601 | if (!(fhdr->frag_off & htons(0xFFF9))) { | |
0d53778e | 602 | pr_debug("Invalid fragment offset\n"); |
9fb9cbb1 YK |
603 | /* It is not a fragmented frame */ |
604 | goto ret_orig; | |
605 | } | |
606 | ||
e31e0bdc | 607 | if (atomic_read(&nf_init_frags.mem) > nf_init_frags.high_thresh) |
9fb9cbb1 YK |
608 | nf_ct_frag6_evictor(); |
609 | ||
610 | fq = fq_find(fhdr->identification, &hdr->saddr, &hdr->daddr); | |
611 | if (fq == NULL) { | |
0d53778e | 612 | pr_debug("Can't find and can't create new queue\n"); |
9fb9cbb1 YK |
613 | goto ret_orig; |
614 | } | |
615 | ||
b9c69896 | 616 | spin_lock_bh(&fq->q.lock); |
9fb9cbb1 YK |
617 | |
618 | if (nf_ct_frag6_queue(fq, clone, fhdr, nhoff) < 0) { | |
b9c69896 | 619 | spin_unlock_bh(&fq->q.lock); |
0d53778e | 620 | pr_debug("Can't insert skb to queue\n"); |
4b6cb5d8 | 621 | fq_put(fq); |
9fb9cbb1 YK |
622 | goto ret_orig; |
623 | } | |
624 | ||
bc578a54 JP |
625 | if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) && |
626 | fq->q.meat == fq->q.len) { | |
9fb9cbb1 YK |
627 | ret_skb = nf_ct_frag6_reasm(fq, dev); |
628 | if (ret_skb == NULL) | |
0d53778e | 629 | pr_debug("Can't reassemble fragmented packets\n"); |
9fb9cbb1 | 630 | } |
b9c69896 | 631 | spin_unlock_bh(&fq->q.lock); |
9fb9cbb1 | 632 | |
4b6cb5d8 | 633 | fq_put(fq); |
9fb9cbb1 YK |
634 | return ret_skb; |
635 | ||
636 | ret_orig: | |
637 | kfree_skb(clone); | |
638 | return skb; | |
639 | } | |
640 | ||
641 | void nf_ct_frag6_output(unsigned int hooknum, struct sk_buff *skb, | |
642 | struct net_device *in, struct net_device *out, | |
643 | int (*okfn)(struct sk_buff *)) | |
644 | { | |
645 | struct sk_buff *s, *s2; | |
646 | ||
647 | for (s = NFCT_FRAG6_CB(skb)->orig; s;) { | |
648 | nf_conntrack_put_reasm(s->nfct_reasm); | |
649 | nf_conntrack_get_reasm(skb); | |
650 | s->nfct_reasm = skb; | |
651 | ||
652 | s2 = s->next; | |
f9f02cca PM |
653 | s->next = NULL; |
654 | ||
9fb9cbb1 YK |
655 | NF_HOOK_THRESH(PF_INET6, hooknum, s, in, out, okfn, |
656 | NF_IP6_PRI_CONNTRACK_DEFRAG + 1); | |
657 | s = s2; | |
658 | } | |
659 | nf_conntrack_put_reasm(skb); | |
660 | } | |
661 | ||
9fb9cbb1 YK |
662 | int nf_ct_frag6_init(void) |
663 | { | |
321a3a99 | 664 | nf_frags.hashfn = nf_hashfn; |
c6fda282 | 665 | nf_frags.constructor = ip6_frag_init; |
c9547709 | 666 | nf_frags.destructor = NULL; |
1e4b8287 PE |
667 | nf_frags.skb_free = nf_skb_free; |
668 | nf_frags.qsize = sizeof(struct nf_ct_frag6_queue); | |
abd6523d | 669 | nf_frags.match = ip6_frag_match; |
e521db9d | 670 | nf_frags.frag_expire = nf_ct_frag6_expire; |
3b4bc4a2 | 671 | nf_frags.secret_interval = 10 * 60 * HZ; |
b2fd5321 | 672 | nf_init_frags.timeout = IPV6_FRAG_TIMEOUT; |
e31e0bdc PE |
673 | nf_init_frags.high_thresh = 256 * 1024; |
674 | nf_init_frags.low_thresh = 192 * 1024; | |
e5a2bb84 | 675 | inet_frags_init_net(&nf_init_frags); |
7eb95156 | 676 | inet_frags_init(&nf_frags); |
9fb9cbb1 YK |
677 | |
678 | return 0; | |
679 | } | |
680 | ||
681 | void nf_ct_frag6_cleanup(void) | |
682 | { | |
7eb95156 PE |
683 | inet_frags_fini(&nf_frags); |
684 | ||
e31e0bdc | 685 | nf_init_frags.low_thresh = 0; |
9fb9cbb1 YK |
686 | nf_ct_frag6_evictor(); |
687 | } |