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 | ||
5a3da1fe HFS |
17 | #define pr_fmt(fmt) "IPv6-nf: " fmt |
18 | ||
9fb9cbb1 YK |
19 | #include <linux/errno.h> |
20 | #include <linux/types.h> | |
21 | #include <linux/string.h> | |
22 | #include <linux/socket.h> | |
23 | #include <linux/sockios.h> | |
24 | #include <linux/jiffies.h> | |
25 | #include <linux/net.h> | |
26 | #include <linux/list.h> | |
27 | #include <linux/netdevice.h> | |
28 | #include <linux/in6.h> | |
29 | #include <linux/ipv6.h> | |
30 | #include <linux/icmpv6.h> | |
31 | #include <linux/random.h> | |
5a0e3ad6 | 32 | #include <linux/slab.h> |
9fb9cbb1 YK |
33 | |
34 | #include <net/sock.h> | |
35 | #include <net/snmp.h> | |
5ab11c98 | 36 | #include <net/inet_frag.h> |
9fb9cbb1 YK |
37 | |
38 | #include <net/ipv6.h> | |
39 | #include <net/protocol.h> | |
40 | #include <net/transp_v6.h> | |
41 | #include <net/rawv6.h> | |
42 | #include <net/ndisc.h> | |
43 | #include <net/addrconf.h> | |
b8dd6a22 | 44 | #include <net/inet_ecn.h> |
99fa5f53 | 45 | #include <net/netfilter/ipv6/nf_conntrack_ipv6.h> |
9fb9cbb1 YK |
46 | #include <linux/sysctl.h> |
47 | #include <linux/netfilter.h> | |
48 | #include <linux/netfilter_ipv6.h> | |
49 | #include <linux/kernel.h> | |
50 | #include <linux/module.h> | |
bced94ed | 51 | #include <net/netfilter/ipv6/nf_defrag_ipv6.h> |
9fb9cbb1 | 52 | |
9fb9cbb1 | 53 | |
9fb9cbb1 YK |
54 | struct nf_ct_frag6_skb_cb |
55 | { | |
56 | struct inet6_skb_parm h; | |
57 | int offset; | |
58 | struct sk_buff *orig; | |
59 | }; | |
60 | ||
61 | #define NFCT_FRAG6_CB(skb) ((struct nf_ct_frag6_skb_cb*)((skb)->cb)) | |
62 | ||
7eb95156 | 63 | static struct inet_frags nf_frags; |
9fb9cbb1 | 64 | |
8d8354d2 | 65 | #ifdef CONFIG_SYSCTL |
be9e9163 | 66 | static struct ctl_table nf_ct_frag6_sysctl_table[] = { |
8d8354d2 PE |
67 | { |
68 | .procname = "nf_conntrack_frag6_timeout", | |
c038a767 | 69 | .data = &init_net.nf_frag.frags.timeout, |
8d8354d2 PE |
70 | .maxlen = sizeof(unsigned int), |
71 | .mode = 0644, | |
6d9f239a | 72 | .proc_handler = proc_dointvec_jiffies, |
8d8354d2 PE |
73 | }, |
74 | { | |
8d8354d2 | 75 | .procname = "nf_conntrack_frag6_low_thresh", |
c038a767 | 76 | .data = &init_net.nf_frag.frags.low_thresh, |
8d8354d2 PE |
77 | .maxlen = sizeof(unsigned int), |
78 | .mode = 0644, | |
6d9f239a | 79 | .proc_handler = proc_dointvec, |
8d8354d2 PE |
80 | }, |
81 | { | |
8d8354d2 | 82 | .procname = "nf_conntrack_frag6_high_thresh", |
c038a767 | 83 | .data = &init_net.nf_frag.frags.high_thresh, |
8d8354d2 PE |
84 | .maxlen = sizeof(unsigned int), |
85 | .mode = 0644, | |
6d9f239a | 86 | .proc_handler = proc_dointvec, |
8d8354d2 | 87 | }, |
f8572d8f | 88 | { } |
8d8354d2 | 89 | }; |
e97c3e27 | 90 | |
f1df1374 | 91 | static int nf_ct_frag6_sysctl_register(struct net *net) |
c038a767 AW |
92 | { |
93 | struct ctl_table *table; | |
94 | struct ctl_table_header *hdr; | |
95 | ||
96 | table = nf_ct_frag6_sysctl_table; | |
97 | if (!net_eq(net, &init_net)) { | |
98 | table = kmemdup(table, sizeof(nf_ct_frag6_sysctl_table), | |
99 | GFP_KERNEL); | |
100 | if (table == NULL) | |
101 | goto err_alloc; | |
102 | ||
894e2ac8 MK |
103 | table[0].data = &net->nf_frag.frags.timeout; |
104 | table[1].data = &net->nf_frag.frags.low_thresh; | |
105 | table[2].data = &net->nf_frag.frags.high_thresh; | |
c038a767 AW |
106 | } |
107 | ||
108 | hdr = register_net_sysctl(net, "net/netfilter", table); | |
109 | if (hdr == NULL) | |
110 | goto err_reg; | |
111 | ||
4b7cc7fc | 112 | net->nf_frag.sysctl.frags_hdr = hdr; |
c038a767 AW |
113 | return 0; |
114 | ||
115 | err_reg: | |
116 | if (!net_eq(net, &init_net)) | |
117 | kfree(table); | |
118 | err_alloc: | |
119 | return -ENOMEM; | |
120 | } | |
121 | ||
122 | static void __net_exit nf_ct_frags6_sysctl_unregister(struct net *net) | |
123 | { | |
124 | struct ctl_table *table; | |
125 | ||
126 | table = net->nf_frag.sysctl.frags_hdr->ctl_table_arg; | |
127 | unregister_net_sysctl_table(net->nf_frag.sysctl.frags_hdr); | |
128 | if (!net_eq(net, &init_net)) | |
129 | kfree(table); | |
130 | } | |
131 | ||
132 | #else | |
f1df1374 | 133 | static int nf_ct_frag6_sysctl_register(struct net *net) |
c038a767 AW |
134 | { |
135 | return 0; | |
136 | } | |
137 | static void __net_exit nf_ct_frags6_sysctl_unregister(struct net *net) | |
138 | { | |
139 | } | |
8d8354d2 PE |
140 | #endif |
141 | ||
b8dd6a22 HFS |
142 | static inline u8 ip6_frag_ecn(const struct ipv6hdr *ipv6h) |
143 | { | |
144 | return 1 << (ipv6_get_dsfield(ipv6h) & INET_ECN_MASK); | |
145 | } | |
146 | ||
321a3a99 | 147 | static unsigned int nf_hashfn(struct inet_frag_queue *q) |
9fb9cbb1 | 148 | { |
b836c99f | 149 | const struct frag_queue *nq; |
9fb9cbb1 | 150 | |
b836c99f | 151 | nq = container_of(q, struct frag_queue, q); |
93c8b90f | 152 | return inet6_hash_frag(nq->id, &nq->saddr, &nq->daddr, nf_frags.rnd); |
9fb9cbb1 YK |
153 | } |
154 | ||
1e4b8287 PE |
155 | static void nf_skb_free(struct sk_buff *skb) |
156 | { | |
157 | if (NFCT_FRAG6_CB(skb)->orig) | |
158 | kfree_skb(NFCT_FRAG6_CB(skb)->orig); | |
159 | } | |
160 | ||
9fb9cbb1 YK |
161 | static void nf_ct_frag6_expire(unsigned long data) |
162 | { | |
b836c99f AW |
163 | struct frag_queue *fq; |
164 | struct net *net; | |
9fb9cbb1 | 165 | |
b836c99f AW |
166 | fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q); |
167 | net = container_of(fq->q.net, struct net, nf_frag.frags); | |
9fb9cbb1 | 168 | |
b836c99f | 169 | ip6_expire_frag_queue(net, fq, &nf_frags); |
9fb9cbb1 YK |
170 | } |
171 | ||
172 | /* Creation primitives. */ | |
b836c99f AW |
173 | static inline struct frag_queue *fq_find(struct net *net, __be32 id, |
174 | u32 user, struct in6_addr *src, | |
7cd9f602 | 175 | struct in6_addr *dst, int iif, u8 ecn) |
9fb9cbb1 | 176 | { |
2588fe1d | 177 | struct inet_frag_queue *q; |
c6fda282 | 178 | struct ip6_create_arg arg; |
abd6523d | 179 | unsigned int hash; |
9fb9cbb1 | 180 | |
c6fda282 | 181 | arg.id = id; |
0b5ccb2e | 182 | arg.user = user; |
c6fda282 PE |
183 | arg.src = src; |
184 | arg.dst = dst; | |
7cd9f602 | 185 | arg.iif = iif; |
b8dd6a22 | 186 | arg.ecn = ecn; |
9a375803 PE |
187 | |
188 | read_lock_bh(&nf_frags.lock); | |
93c8b90f | 189 | hash = inet6_hash_frag(id, src, dst, nf_frags.rnd); |
9fb9cbb1 | 190 | |
c038a767 | 191 | q = inet_frag_find(&net->nf_frag.frags, &nf_frags, &arg, hash); |
b9c69896 | 192 | local_bh_enable(); |
5a3da1fe HFS |
193 | if (IS_ERR_OR_NULL(q)) { |
194 | inet_frag_maybe_warn_overflow(q, pr_fmt()); | |
195 | return NULL; | |
196 | } | |
b836c99f | 197 | return container_of(q, struct frag_queue, q); |
9fb9cbb1 YK |
198 | } |
199 | ||
9fb9cbb1 | 200 | |
b836c99f | 201 | static int nf_ct_frag6_queue(struct frag_queue *fq, struct sk_buff *skb, |
58c0fb0d | 202 | const struct frag_hdr *fhdr, int nhoff) |
9fb9cbb1 YK |
203 | { |
204 | struct sk_buff *prev, *next; | |
4cdd3408 | 205 | unsigned int payload_len; |
9fb9cbb1 | 206 | int offset, end; |
b8dd6a22 | 207 | u8 ecn; |
9fb9cbb1 | 208 | |
bc578a54 | 209 | if (fq->q.last_in & INET_FRAG_COMPLETE) { |
698f9315 | 210 | pr_debug("Already completed\n"); |
9fb9cbb1 YK |
211 | goto err; |
212 | } | |
213 | ||
4cdd3408 PM |
214 | payload_len = ntohs(ipv6_hdr(skb)->payload_len); |
215 | ||
9fb9cbb1 | 216 | offset = ntohs(fhdr->frag_off) & ~0x7; |
4cdd3408 | 217 | end = offset + (payload_len - |
0660e03f | 218 | ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1))); |
9fb9cbb1 YK |
219 | |
220 | if ((unsigned int)end > IPV6_MAXPLEN) { | |
0d53778e | 221 | pr_debug("offset is too large.\n"); |
1ab1457c | 222 | return -1; |
9fb9cbb1 YK |
223 | } |
224 | ||
b8dd6a22 HFS |
225 | ecn = ip6_frag_ecn(ipv6_hdr(skb)); |
226 | ||
d56f90a7 ACM |
227 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
228 | const unsigned char *nh = skb_network_header(skb); | |
1ab1457c | 229 | skb->csum = csum_sub(skb->csum, |
d56f90a7 | 230 | csum_partial(nh, (u8 *)(fhdr + 1) - nh, |
9fb9cbb1 | 231 | 0)); |
d56f90a7 | 232 | } |
9fb9cbb1 YK |
233 | |
234 | /* Is this the final fragment? */ | |
235 | if (!(fhdr->frag_off & htons(IP6_MF))) { | |
236 | /* If we already have some bits beyond end | |
237 | * or have different end, the segment is corrupted. | |
238 | */ | |
5ab11c98 | 239 | if (end < fq->q.len || |
bc578a54 | 240 | ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len)) { |
0d53778e | 241 | pr_debug("already received last fragment\n"); |
9fb9cbb1 YK |
242 | goto err; |
243 | } | |
bc578a54 | 244 | fq->q.last_in |= INET_FRAG_LAST_IN; |
5ab11c98 | 245 | fq->q.len = end; |
9fb9cbb1 YK |
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 | */ | |
0d53778e | 254 | pr_debug("end of fragment not rounded to 8 bytes.\n"); |
9fb9cbb1 YK |
255 | return -1; |
256 | } | |
5ab11c98 | 257 | if (end > fq->q.len) { |
9fb9cbb1 | 258 | /* Some bits beyond end -> corruption. */ |
bc578a54 | 259 | if (fq->q.last_in & INET_FRAG_LAST_IN) { |
0d53778e | 260 | pr_debug("last packet already reached.\n"); |
9fb9cbb1 YK |
261 | goto err; |
262 | } | |
5ab11c98 | 263 | fq->q.len = end; |
9fb9cbb1 YK |
264 | } |
265 | } | |
266 | ||
267 | if (end == offset) | |
268 | goto err; | |
269 | ||
270 | /* Point into the IP datagram 'data' part. */ | |
271 | if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data)) { | |
0d53778e | 272 | pr_debug("queue: message is too short.\n"); |
9fb9cbb1 YK |
273 | goto err; |
274 | } | |
b38dfee3 | 275 | if (pskb_trim_rcsum(skb, end - offset)) { |
0d53778e | 276 | pr_debug("Can't trim\n"); |
b38dfee3 | 277 | goto err; |
9fb9cbb1 YK |
278 | } |
279 | ||
280 | /* Find out which fragments are in front and at the back of us | |
281 | * in the chain of fragments so far. We must know where to put | |
282 | * this fragment, right? | |
283 | */ | |
ea8fbe8f CG |
284 | prev = fq->q.fragments_tail; |
285 | if (!prev || NFCT_FRAG6_CB(prev)->offset < offset) { | |
286 | next = NULL; | |
287 | goto found; | |
288 | } | |
9fb9cbb1 | 289 | prev = NULL; |
5ab11c98 | 290 | for (next = fq->q.fragments; next != NULL; next = next->next) { |
9fb9cbb1 YK |
291 | if (NFCT_FRAG6_CB(next)->offset >= offset) |
292 | break; /* bingo! */ | |
293 | prev = next; | |
294 | } | |
295 | ||
ea8fbe8f | 296 | found: |
1ee89bd0 ND |
297 | /* RFC5722, Section 4: |
298 | * When reassembling an IPv6 datagram, if | |
299 | * one or more its constituent fragments is determined to be an | |
300 | * overlapping fragment, the entire datagram (and any constituent | |
301 | * fragments, including those not yet received) MUST be silently | |
302 | * discarded. | |
9fb9cbb1 | 303 | */ |
9fb9cbb1 | 304 | |
1ee89bd0 ND |
305 | /* Check for overlap with preceding fragment. */ |
306 | if (prev && | |
22e091e5 | 307 | (NFCT_FRAG6_CB(prev)->offset + prev->len) > offset) |
1ee89bd0 | 308 | goto discard_fq; |
9fb9cbb1 | 309 | |
1ee89bd0 ND |
310 | /* Look for overlap with succeeding segment. */ |
311 | if (next && NFCT_FRAG6_CB(next)->offset < end) | |
312 | goto discard_fq; | |
9fb9cbb1 YK |
313 | |
314 | NFCT_FRAG6_CB(skb)->offset = offset; | |
315 | ||
316 | /* Insert this fragment in the chain of fragments. */ | |
317 | skb->next = next; | |
ea8fbe8f CG |
318 | if (!next) |
319 | fq->q.fragments_tail = skb; | |
9fb9cbb1 YK |
320 | if (prev) |
321 | prev->next = skb; | |
322 | else | |
5ab11c98 | 323 | fq->q.fragments = skb; |
9fb9cbb1 | 324 | |
97cf00e9 HX |
325 | if (skb->dev) { |
326 | fq->iif = skb->dev->ifindex; | |
327 | skb->dev = NULL; | |
328 | } | |
5ab11c98 PE |
329 | fq->q.stamp = skb->tstamp; |
330 | fq->q.meat += skb->len; | |
b8dd6a22 | 331 | fq->ecn |= ecn; |
4cdd3408 PM |
332 | if (payload_len > fq->q.max_size) |
333 | fq->q.max_size = payload_len; | |
d433673e | 334 | add_frag_mem_limit(&fq->q, skb->truesize); |
9fb9cbb1 YK |
335 | |
336 | /* The first fragment. | |
337 | * nhoffset is obtained from the first fragment, of course. | |
338 | */ | |
339 | if (offset == 0) { | |
340 | fq->nhoffset = nhoff; | |
bc578a54 | 341 | fq->q.last_in |= INET_FRAG_FIRST_IN; |
9fb9cbb1 | 342 | } |
3ef0eb0d JDB |
343 | |
344 | inet_frag_lru_move(&fq->q); | |
9fb9cbb1 YK |
345 | return 0; |
346 | ||
1ee89bd0 | 347 | discard_fq: |
b836c99f | 348 | inet_frag_kill(&fq->q, &nf_frags); |
9fb9cbb1 YK |
349 | err: |
350 | return -1; | |
351 | } | |
352 | ||
353 | /* | |
354 | * Check if this packet is complete. | |
355 | * Returns NULL on failure by any reason, and pointer | |
356 | * to current nexthdr field in reassembled frame. | |
357 | * | |
358 | * It is called with locked fq, and caller must check that | |
359 | * queue is eligible for reassembly i.e. it is not COMPLETE, | |
360 | * the last and the first frames arrived and all the bits are here. | |
361 | */ | |
362 | static struct sk_buff * | |
b836c99f | 363 | nf_ct_frag6_reasm(struct frag_queue *fq, struct net_device *dev) |
9fb9cbb1 | 364 | { |
5ab11c98 | 365 | struct sk_buff *fp, *op, *head = fq->q.fragments; |
9fb9cbb1 | 366 | int payload_len; |
b8dd6a22 | 367 | u8 ecn; |
9fb9cbb1 | 368 | |
b836c99f | 369 | inet_frag_kill(&fq->q, &nf_frags); |
9fb9cbb1 | 370 | |
547b792c IJ |
371 | WARN_ON(head == NULL); |
372 | WARN_ON(NFCT_FRAG6_CB(head)->offset != 0); | |
9fb9cbb1 | 373 | |
b8dd6a22 HFS |
374 | ecn = ip_frag_ecn_table[fq->ecn]; |
375 | if (unlikely(ecn == 0xff)) | |
376 | goto out_fail; | |
377 | ||
9fb9cbb1 | 378 | /* Unfragmented part is taken from the first segment. */ |
d56f90a7 | 379 | payload_len = ((head->data - skb_network_header(head)) - |
5ab11c98 | 380 | sizeof(struct ipv6hdr) + fq->q.len - |
d56f90a7 | 381 | sizeof(struct frag_hdr)); |
9fb9cbb1 | 382 | if (payload_len > IPV6_MAXPLEN) { |
0d53778e | 383 | pr_debug("payload len is too large.\n"); |
9fb9cbb1 YK |
384 | goto out_oversize; |
385 | } | |
386 | ||
387 | /* Head of list must not be cloned. */ | |
14bbd6a5 | 388 | if (skb_unclone(head, GFP_ATOMIC)) { |
0d53778e | 389 | pr_debug("skb is cloned but can't expand head"); |
9fb9cbb1 YK |
390 | goto out_oom; |
391 | } | |
392 | ||
393 | /* If the first fragment is fragmented itself, we split | |
394 | * it to two chunks: the first with data and paged part | |
395 | * and the second, holding only fragments. */ | |
21dc3301 | 396 | if (skb_has_frag_list(head)) { |
9fb9cbb1 YK |
397 | struct sk_buff *clone; |
398 | int i, plen = 0; | |
399 | ||
0a9ee813 JP |
400 | clone = alloc_skb(0, GFP_ATOMIC); |
401 | if (clone == NULL) | |
9fb9cbb1 | 402 | goto out_oom; |
0a9ee813 | 403 | |
9fb9cbb1 YK |
404 | clone->next = head->next; |
405 | head->next = clone; | |
406 | skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list; | |
343a9972 | 407 | skb_frag_list_init(head); |
9e903e08 ED |
408 | for (i = 0; i < skb_shinfo(head)->nr_frags; i++) |
409 | plen += skb_frag_size(&skb_shinfo(head)->frags[i]); | |
9fb9cbb1 YK |
410 | clone->len = clone->data_len = head->data_len - plen; |
411 | head->data_len -= clone->len; | |
412 | head->len -= clone->len; | |
413 | clone->csum = 0; | |
414 | clone->ip_summed = head->ip_summed; | |
415 | ||
416 | NFCT_FRAG6_CB(clone)->orig = NULL; | |
d433673e | 417 | add_frag_mem_limit(&fq->q, clone->truesize); |
9fb9cbb1 YK |
418 | } |
419 | ||
420 | /* We have to remove fragment header from datagram and to relocate | |
421 | * header in order to calculate ICV correctly. */ | |
bff9b61c | 422 | skb_network_header(head)[fq->nhoffset] = skb_transport_header(head)[0]; |
1ab1457c | 423 | memmove(head->head + sizeof(struct frag_hdr), head->head, |
9fb9cbb1 | 424 | (head->data - head->head) - sizeof(struct frag_hdr)); |
b0e380b1 ACM |
425 | head->mac_header += sizeof(struct frag_hdr); |
426 | head->network_header += sizeof(struct frag_hdr); | |
9fb9cbb1 YK |
427 | |
428 | skb_shinfo(head)->frag_list = head->next; | |
badff6d0 | 429 | skb_reset_transport_header(head); |
d56f90a7 | 430 | skb_push(head, head->data - skb_network_header(head)); |
9fb9cbb1 YK |
431 | |
432 | for (fp=head->next; fp; fp = fp->next) { | |
433 | head->data_len += fp->len; | |
434 | head->len += fp->len; | |
435 | if (head->ip_summed != fp->ip_summed) | |
436 | head->ip_summed = CHECKSUM_NONE; | |
84fa7933 | 437 | else if (head->ip_summed == CHECKSUM_COMPLETE) |
9fb9cbb1 YK |
438 | head->csum = csum_add(head->csum, fp->csum); |
439 | head->truesize += fp->truesize; | |
9fb9cbb1 | 440 | } |
d433673e | 441 | sub_frag_mem_limit(&fq->q, head->truesize); |
9fb9cbb1 | 442 | |
4cdd3408 | 443 | head->local_df = 1; |
9fb9cbb1 YK |
444 | head->next = NULL; |
445 | head->dev = dev; | |
5ab11c98 | 446 | head->tstamp = fq->q.stamp; |
0660e03f | 447 | ipv6_hdr(head)->payload_len = htons(payload_len); |
b8dd6a22 | 448 | ipv6_change_dsfield(ipv6_hdr(head), 0xff, ecn); |
4cdd3408 | 449 | IP6CB(head)->frag_max_size = sizeof(struct ipv6hdr) + fq->q.max_size; |
9fb9cbb1 YK |
450 | |
451 | /* Yes, and fold redundant checksum back. 8) */ | |
84fa7933 | 452 | if (head->ip_summed == CHECKSUM_COMPLETE) |
d56f90a7 | 453 | head->csum = csum_partial(skb_network_header(head), |
cfe1fc77 | 454 | skb_network_header_len(head), |
d56f90a7 | 455 | head->csum); |
9fb9cbb1 | 456 | |
5ab11c98 | 457 | fq->q.fragments = NULL; |
ea8fbe8f | 458 | fq->q.fragments_tail = NULL; |
9fb9cbb1 YK |
459 | |
460 | /* all original skbs are linked into the NFCT_FRAG6_CB(head).orig */ | |
461 | fp = skb_shinfo(head)->frag_list; | |
9e2dcf72 | 462 | if (fp && NFCT_FRAG6_CB(fp)->orig == NULL) |
9fb9cbb1 YK |
463 | /* at above code, head skb is divided into two skbs. */ |
464 | fp = fp->next; | |
465 | ||
466 | op = NFCT_FRAG6_CB(head)->orig; | |
467 | for (; fp; fp = fp->next) { | |
468 | struct sk_buff *orig = NFCT_FRAG6_CB(fp)->orig; | |
469 | ||
470 | op->next = orig; | |
471 | op = orig; | |
472 | NFCT_FRAG6_CB(fp)->orig = NULL; | |
473 | } | |
474 | ||
475 | return head; | |
476 | ||
477 | out_oversize: | |
e87cc472 JP |
478 | net_dbg_ratelimited("nf_ct_frag6_reasm: payload len = %d\n", |
479 | payload_len); | |
9fb9cbb1 YK |
480 | goto out_fail; |
481 | out_oom: | |
e87cc472 | 482 | net_dbg_ratelimited("nf_ct_frag6_reasm: no memory for reassembly\n"); |
9fb9cbb1 YK |
483 | out_fail: |
484 | return NULL; | |
485 | } | |
486 | ||
487 | /* | |
488 | * find the header just before Fragment Header. | |
489 | * | |
490 | * if success return 0 and set ... | |
491 | * (*prevhdrp): the value of "Next Header Field" in the header | |
492 | * just before Fragment Header. | |
493 | * (*prevhoff): the offset of "Next Header Field" in the header | |
494 | * just before Fragment Header. | |
495 | * (*fhoff) : the offset of Fragment Header. | |
496 | * | |
497 | * Based on ipv6_skip_hdr() in net/ipv6/exthdr.c | |
498 | * | |
499 | */ | |
500 | static int | |
501 | find_prev_fhdr(struct sk_buff *skb, u8 *prevhdrp, int *prevhoff, int *fhoff) | |
502 | { | |
0660e03f | 503 | u8 nexthdr = ipv6_hdr(skb)->nexthdr; |
6b88dd96 ACM |
504 | const int netoff = skb_network_offset(skb); |
505 | u8 prev_nhoff = netoff + offsetof(struct ipv6hdr, nexthdr); | |
506 | int start = netoff + sizeof(struct ipv6hdr); | |
9fb9cbb1 YK |
507 | int len = skb->len - start; |
508 | u8 prevhdr = NEXTHDR_IPV6; | |
509 | ||
1ab1457c YH |
510 | while (nexthdr != NEXTHDR_FRAGMENT) { |
511 | struct ipv6_opt_hdr hdr; | |
512 | int hdrlen; | |
9fb9cbb1 YK |
513 | |
514 | if (!ipv6_ext_hdr(nexthdr)) { | |
515 | return -1; | |
516 | } | |
1ab1457c | 517 | if (nexthdr == NEXTHDR_NONE) { |
0d53778e | 518 | pr_debug("next header is none\n"); |
9fb9cbb1 YK |
519 | return -1; |
520 | } | |
d1238d53 CP |
521 | if (len < (int)sizeof(struct ipv6_opt_hdr)) { |
522 | pr_debug("too short\n"); | |
523 | return -1; | |
524 | } | |
1ab1457c YH |
525 | if (skb_copy_bits(skb, start, &hdr, sizeof(hdr))) |
526 | BUG(); | |
527 | if (nexthdr == NEXTHDR_AUTH) | |
528 | hdrlen = (hdr.hdrlen+2)<<2; | |
529 | else | |
530 | hdrlen = ipv6_optlen(&hdr); | |
9fb9cbb1 YK |
531 | |
532 | prevhdr = nexthdr; | |
533 | prev_nhoff = start; | |
534 | ||
1ab1457c YH |
535 | nexthdr = hdr.nexthdr; |
536 | len -= hdrlen; | |
537 | start += hdrlen; | |
538 | } | |
9fb9cbb1 YK |
539 | |
540 | if (len < 0) | |
541 | return -1; | |
542 | ||
543 | *prevhdrp = prevhdr; | |
544 | *prevhoff = prev_nhoff; | |
545 | *fhoff = start; | |
546 | ||
547 | return 0; | |
548 | } | |
549 | ||
0b5ccb2e | 550 | struct sk_buff *nf_ct_frag6_gather(struct sk_buff *skb, u32 user) |
9fb9cbb1 | 551 | { |
1ab1457c | 552 | struct sk_buff *clone; |
9fb9cbb1 | 553 | struct net_device *dev = skb->dev; |
c038a767 AW |
554 | struct net *net = skb_dst(skb) ? dev_net(skb_dst(skb)->dev) |
555 | : dev_net(skb->dev); | |
9fb9cbb1 | 556 | struct frag_hdr *fhdr; |
b836c99f | 557 | struct frag_queue *fq; |
9fb9cbb1 YK |
558 | struct ipv6hdr *hdr; |
559 | int fhoff, nhoff; | |
560 | u8 prevhdr; | |
561 | struct sk_buff *ret_skb = NULL; | |
562 | ||
563 | /* Jumbo payload inhibits frag. header */ | |
0660e03f | 564 | if (ipv6_hdr(skb)->payload_len == 0) { |
0d53778e | 565 | pr_debug("payload len = 0\n"); |
9fb9cbb1 YK |
566 | return skb; |
567 | } | |
568 | ||
569 | if (find_prev_fhdr(skb, &prevhdr, &nhoff, &fhoff) < 0) | |
570 | return skb; | |
571 | ||
572 | clone = skb_clone(skb, GFP_ATOMIC); | |
573 | if (clone == NULL) { | |
0d53778e | 574 | pr_debug("Can't clone skb\n"); |
9fb9cbb1 YK |
575 | return skb; |
576 | } | |
577 | ||
578 | NFCT_FRAG6_CB(clone)->orig = skb; | |
579 | ||
580 | if (!pskb_may_pull(clone, fhoff + sizeof(*fhdr))) { | |
0d53778e | 581 | pr_debug("message is too short.\n"); |
9fb9cbb1 YK |
582 | goto ret_orig; |
583 | } | |
584 | ||
967b05f6 | 585 | skb_set_transport_header(clone, fhoff); |
0660e03f | 586 | hdr = ipv6_hdr(clone); |
bff9b61c | 587 | fhdr = (struct frag_hdr *)skb_transport_header(clone); |
9fb9cbb1 | 588 | |
6b102865 AW |
589 | local_bh_disable(); |
590 | inet_frag_evictor(&net->nf_frag.frags, &nf_frags, false); | |
591 | local_bh_enable(); | |
9fb9cbb1 | 592 | |
b8dd6a22 | 593 | fq = fq_find(net, fhdr->identification, user, &hdr->saddr, &hdr->daddr, |
7cd9f602 | 594 | skb->dev ? skb->dev->ifindex : 0, ip6_frag_ecn(hdr)); |
9fb9cbb1 | 595 | if (fq == NULL) { |
0d53778e | 596 | pr_debug("Can't find and can't create new queue\n"); |
9fb9cbb1 YK |
597 | goto ret_orig; |
598 | } | |
599 | ||
b9c69896 | 600 | spin_lock_bh(&fq->q.lock); |
9fb9cbb1 YK |
601 | |
602 | if (nf_ct_frag6_queue(fq, clone, fhdr, nhoff) < 0) { | |
b9c69896 | 603 | spin_unlock_bh(&fq->q.lock); |
0d53778e | 604 | pr_debug("Can't insert skb to queue\n"); |
b836c99f | 605 | inet_frag_put(&fq->q, &nf_frags); |
9fb9cbb1 YK |
606 | goto ret_orig; |
607 | } | |
608 | ||
bc578a54 JP |
609 | if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) && |
610 | fq->q.meat == fq->q.len) { | |
9fb9cbb1 YK |
611 | ret_skb = nf_ct_frag6_reasm(fq, dev); |
612 | if (ret_skb == NULL) | |
0d53778e | 613 | pr_debug("Can't reassemble fragmented packets\n"); |
9fb9cbb1 | 614 | } |
b9c69896 | 615 | spin_unlock_bh(&fq->q.lock); |
9fb9cbb1 | 616 | |
b836c99f | 617 | inet_frag_put(&fq->q, &nf_frags); |
9fb9cbb1 YK |
618 | return ret_skb; |
619 | ||
620 | ret_orig: | |
621 | kfree_skb(clone); | |
622 | return skb; | |
623 | } | |
624 | ||
fddd8b50 | 625 | void nf_ct_frag6_consume_orig(struct sk_buff *skb) |
9fb9cbb1 YK |
626 | { |
627 | struct sk_buff *s, *s2; | |
628 | ||
629 | for (s = NFCT_FRAG6_CB(skb)->orig; s;) { | |
9fb9cbb1 | 630 | s2 = s->next; |
f9f02cca | 631 | s->next = NULL; |
fddd8b50 | 632 | consume_skb(s); |
9fb9cbb1 YK |
633 | s = s2; |
634 | } | |
9fb9cbb1 YK |
635 | } |
636 | ||
c038a767 AW |
637 | static int nf_ct_net_init(struct net *net) |
638 | { | |
639 | net->nf_frag.frags.high_thresh = IPV6_FRAG_HIGH_THRESH; | |
640 | net->nf_frag.frags.low_thresh = IPV6_FRAG_LOW_THRESH; | |
641 | net->nf_frag.frags.timeout = IPV6_FRAG_TIMEOUT; | |
642 | inet_frags_init_net(&net->nf_frag.frags); | |
643 | ||
644 | return nf_ct_frag6_sysctl_register(net); | |
645 | } | |
646 | ||
647 | static void nf_ct_net_exit(struct net *net) | |
648 | { | |
649 | nf_ct_frags6_sysctl_unregister(net); | |
650 | inet_frags_exit_net(&net->nf_frag.frags, &nf_frags); | |
651 | } | |
652 | ||
653 | static struct pernet_operations nf_ct_net_ops = { | |
654 | .init = nf_ct_net_init, | |
655 | .exit = nf_ct_net_exit, | |
656 | }; | |
657 | ||
9fb9cbb1 YK |
658 | int nf_ct_frag6_init(void) |
659 | { | |
c038a767 AW |
660 | int ret = 0; |
661 | ||
321a3a99 | 662 | nf_frags.hashfn = nf_hashfn; |
c6fda282 | 663 | nf_frags.constructor = ip6_frag_init; |
c9547709 | 664 | nf_frags.destructor = NULL; |
1e4b8287 | 665 | nf_frags.skb_free = nf_skb_free; |
b836c99f | 666 | nf_frags.qsize = sizeof(struct frag_queue); |
abd6523d | 667 | nf_frags.match = ip6_frag_match; |
e521db9d | 668 | nf_frags.frag_expire = nf_ct_frag6_expire; |
3b4bc4a2 | 669 | nf_frags.secret_interval = 10 * 60 * HZ; |
7eb95156 | 670 | inet_frags_init(&nf_frags); |
9fb9cbb1 | 671 | |
c038a767 AW |
672 | ret = register_pernet_subsys(&nf_ct_net_ops); |
673 | if (ret) | |
e97c3e27 | 674 | inet_frags_fini(&nf_frags); |
e97c3e27 | 675 | |
c038a767 | 676 | return ret; |
9fb9cbb1 YK |
677 | } |
678 | ||
679 | void nf_ct_frag6_cleanup(void) | |
680 | { | |
c038a767 | 681 | unregister_pernet_subsys(&nf_ct_net_ops); |
7eb95156 | 682 | inet_frags_fini(&nf_frags); |
9fb9cbb1 | 683 | } |