]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * IPv6 fragment reassembly | |
1ab1457c | 3 | * Linux INET6 implementation |
1da177e4 LT |
4 | * |
5 | * Authors: | |
1ab1457c | 6 | * Pedro Roque <[email protected]> |
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 |
63 | struct 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 |
71 | static inline u8 ip6_frag_ecn(const struct ipv6hdr *ipv6h) |
72 | { | |
73 | return 1 << (ipv6_get_dsfield(ipv6h) & INET_ECN_MASK); | |
74 | } | |
1da177e4 | 75 | |
7eb95156 | 76 | static struct inet_frags ip6_frags; |
1da177e4 | 77 | |
f61944ef HX |
78 | static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev, |
79 | struct net_device *dev); | |
80 | ||
f6596f9d ZB |
81 | /* |
82 | * callers should be careful not to use the hash value outside the ipfrag_lock | |
83 | * as doing so could race with ipfrag_hash_rnd being recalculated. | |
84 | */ | |
93c8b90f IJ |
85 | unsigned int inet6_hash_frag(__be32 id, const struct in6_addr *saddr, |
86 | const struct in6_addr *daddr, u32 rnd) | |
1da177e4 | 87 | { |
82a39eb6 JK |
88 | u32 c; |
89 | ||
8064b3cf ED |
90 | c = jhash_3words(ipv6_addr_hash(saddr), ipv6_addr_hash(daddr), |
91 | (__force u32)id, rnd); | |
1da177e4 | 92 | |
7eb95156 | 93 | return c & (INETFRAGS_HASHSZ - 1); |
1da177e4 | 94 | } |
93c8b90f | 95 | EXPORT_SYMBOL_GPL(inet6_hash_frag); |
1da177e4 | 96 | |
321a3a99 | 97 | static unsigned int ip6_hashfn(struct inet_frag_queue *q) |
1da177e4 | 98 | { |
321a3a99 | 99 | struct frag_queue *fq; |
1da177e4 | 100 | |
321a3a99 | 101 | fq = container_of(q, struct frag_queue, q); |
93c8b90f | 102 | return inet6_hash_frag(fq->id, &fq->saddr, &fq->daddr, ip6_frags.rnd); |
1da177e4 LT |
103 | } |
104 | ||
cbc264ca | 105 | bool ip6_frag_match(struct inet_frag_queue *q, void *a) |
abd6523d PE |
106 | { |
107 | struct frag_queue *fq; | |
108 | struct ip6_create_arg *arg = a; | |
109 | ||
110 | fq = container_of(q, struct frag_queue, q); | |
cbc264ca ED |
111 | return fq->id == arg->id && |
112 | fq->user == arg->user && | |
113 | ipv6_addr_equal(&fq->saddr, arg->src) && | |
114 | ipv6_addr_equal(&fq->daddr, arg->dst); | |
abd6523d PE |
115 | } |
116 | EXPORT_SYMBOL(ip6_frag_match); | |
117 | ||
c6fda282 | 118 | void ip6_frag_init(struct inet_frag_queue *q, void *a) |
1da177e4 | 119 | { |
c6fda282 PE |
120 | struct frag_queue *fq = container_of(q, struct frag_queue, q); |
121 | struct ip6_create_arg *arg = a; | |
122 | ||
123 | fq->id = arg->id; | |
0b5ccb2e | 124 | fq->user = arg->user; |
4e3fd7a0 AD |
125 | fq->saddr = *arg->src; |
126 | fq->daddr = *arg->dst; | |
eec2e618 | 127 | fq->ecn = arg->ecn; |
1da177e4 | 128 | } |
c6fda282 | 129 | EXPORT_SYMBOL(ip6_frag_init); |
1da177e4 | 130 | |
b836c99f AW |
131 | void ip6_expire_frag_queue(struct net *net, struct frag_queue *fq, |
132 | struct inet_frags *frags) | |
1da177e4 | 133 | { |
a11d206d | 134 | struct net_device *dev = NULL; |
e521db9d | 135 | |
5ab11c98 | 136 | spin_lock(&fq->q.lock); |
1da177e4 | 137 | |
bc578a54 | 138 | if (fq->q.last_in & INET_FRAG_COMPLETE) |
1da177e4 LT |
139 | goto out; |
140 | ||
b836c99f | 141 | inet_frag_kill(&fq->q, frags); |
1da177e4 | 142 | |
69df9d59 ED |
143 | rcu_read_lock(); |
144 | dev = dev_get_by_index_rcu(net, fq->iif); | |
a11d206d | 145 | if (!dev) |
69df9d59 | 146 | goto out_rcu_unlock; |
a11d206d | 147 | |
483a47d2 DL |
148 | IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMTIMEOUT); |
149 | IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS); | |
1da177e4 | 150 | |
78c784c4 | 151 | /* Don't send error if the first segment did not arrive. */ |
bc578a54 | 152 | if (!(fq->q.last_in & INET_FRAG_FIRST_IN) || !fq->q.fragments) |
69df9d59 | 153 | goto out_rcu_unlock; |
78c784c4 | 154 | |
78c784c4 IO |
155 | /* |
156 | But use as source device on which LAST ARRIVED | |
157 | segment was received. And do not use fq->dev | |
158 | pointer directly, device might already disappeared. | |
159 | */ | |
5ab11c98 | 160 | fq->q.fragments->dev = dev; |
3ffe533c | 161 | icmpv6_send(fq->q.fragments, ICMPV6_TIME_EXCEED, ICMPV6_EXC_FRAGTIME, 0); |
69df9d59 ED |
162 | out_rcu_unlock: |
163 | rcu_read_unlock(); | |
1da177e4 | 164 | out: |
5ab11c98 | 165 | spin_unlock(&fq->q.lock); |
b836c99f AW |
166 | inet_frag_put(&fq->q, frags); |
167 | } | |
168 | EXPORT_SYMBOL(ip6_expire_frag_queue); | |
169 | ||
170 | static void ip6_frag_expire(unsigned long data) | |
171 | { | |
172 | struct frag_queue *fq; | |
173 | struct net *net; | |
174 | ||
175 | fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q); | |
176 | net = container_of(fq->q.net, struct net, ipv6.frags); | |
177 | ||
178 | ip6_expire_frag_queue(net, fq, &ip6_frags); | |
1da177e4 LT |
179 | } |
180 | ||
abd6523d | 181 | static __inline__ struct frag_queue * |
eec2e618 HFS |
182 | fq_find(struct net *net, __be32 id, const struct in6_addr *src, |
183 | const struct in6_addr *dst, u8 ecn) | |
1da177e4 | 184 | { |
c6fda282 PE |
185 | struct inet_frag_queue *q; |
186 | struct ip6_create_arg arg; | |
abd6523d | 187 | unsigned int hash; |
1da177e4 | 188 | |
c6fda282 | 189 | arg.id = id; |
0b5ccb2e | 190 | arg.user = IP6_DEFRAG_LOCAL_DELIVER; |
c6fda282 PE |
191 | arg.src = src; |
192 | arg.dst = dst; | |
eec2e618 | 193 | arg.ecn = ecn; |
9a375803 PE |
194 | |
195 | read_lock(&ip6_frags.lock); | |
93c8b90f | 196 | hash = inet6_hash_frag(id, src, dst, ip6_frags.rnd); |
1da177e4 | 197 | |
ac18e750 | 198 | q = inet_frag_find(&net->ipv6.frags, &ip6_frags, &arg, hash); |
5a3da1fe HFS |
199 | if (IS_ERR_OR_NULL(q)) { |
200 | inet_frag_maybe_warn_overflow(q, pr_fmt()); | |
9546377c | 201 | return NULL; |
5a3da1fe | 202 | } |
c6fda282 | 203 | return container_of(q, struct frag_queue, q); |
1da177e4 LT |
204 | } |
205 | ||
f61944ef | 206 | static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb, |
1da177e4 LT |
207 | struct frag_hdr *fhdr, int nhoff) |
208 | { | |
209 | struct sk_buff *prev, *next; | |
f61944ef | 210 | struct net_device *dev; |
1da177e4 | 211 | int offset, end; |
adf30907 | 212 | struct net *net = dev_net(skb_dst(skb)->dev); |
eec2e618 | 213 | u8 ecn; |
1da177e4 | 214 | |
bc578a54 | 215 | if (fq->q.last_in & INET_FRAG_COMPLETE) |
1da177e4 LT |
216 | goto err; |
217 | ||
218 | offset = ntohs(fhdr->frag_off) & ~0x7; | |
0660e03f ACM |
219 | end = offset + (ntohs(ipv6_hdr(skb)->payload_len) - |
220 | ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1))); | |
1da177e4 LT |
221 | |
222 | if ((unsigned int)end > IPV6_MAXPLEN) { | |
adf30907 | 223 | IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 224 | IPSTATS_MIB_INHDRERRORS); |
d56f90a7 ACM |
225 | icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, |
226 | ((u8 *)&fhdr->frag_off - | |
227 | skb_network_header(skb))); | |
f61944ef | 228 | return -1; |
1da177e4 LT |
229 | } |
230 | ||
eec2e618 HFS |
231 | ecn = ip6_frag_ecn(ipv6_hdr(skb)); |
232 | ||
d56f90a7 ACM |
233 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
234 | const unsigned char *nh = skb_network_header(skb); | |
1ab1457c | 235 | skb->csum = csum_sub(skb->csum, |
d56f90a7 ACM |
236 | csum_partial(nh, (u8 *)(fhdr + 1) - nh, |
237 | 0)); | |
238 | } | |
1da177e4 LT |
239 | |
240 | /* Is this the final fragment? */ | |
241 | if (!(fhdr->frag_off & htons(IP6_MF))) { | |
242 | /* If we already have some bits beyond end | |
243 | * or have different end, the segment is corrupted. | |
244 | */ | |
5ab11c98 | 245 | if (end < fq->q.len || |
bc578a54 | 246 | ((fq->q.last_in & INET_FRAG_LAST_IN) && end != fq->q.len)) |
1da177e4 | 247 | goto err; |
bc578a54 | 248 | fq->q.last_in |= INET_FRAG_LAST_IN; |
5ab11c98 | 249 | fq->q.len = end; |
1da177e4 LT |
250 | } else { |
251 | /* Check if the fragment is rounded to 8 bytes. | |
252 | * Required by the RFC. | |
253 | */ | |
254 | if (end & 0x7) { | |
255 | /* RFC2460 says always send parameter problem in | |
256 | * this case. -DaveM | |
257 | */ | |
adf30907 | 258 | IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), |
a11d206d | 259 | IPSTATS_MIB_INHDRERRORS); |
1ab1457c | 260 | icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, |
1da177e4 | 261 | offsetof(struct ipv6hdr, payload_len)); |
f61944ef | 262 | return -1; |
1da177e4 | 263 | } |
5ab11c98 | 264 | if (end > fq->q.len) { |
1da177e4 | 265 | /* Some bits beyond end -> corruption. */ |
bc578a54 | 266 | if (fq->q.last_in & INET_FRAG_LAST_IN) |
1da177e4 | 267 | goto err; |
5ab11c98 | 268 | fq->q.len = end; |
1da177e4 LT |
269 | } |
270 | } | |
271 | ||
272 | if (end == offset) | |
273 | goto err; | |
274 | ||
275 | /* Point into the IP datagram 'data' part. */ | |
276 | if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data)) | |
277 | goto err; | |
1ab1457c | 278 | |
42ca89c1 SH |
279 | if (pskb_trim_rcsum(skb, end - offset)) |
280 | goto err; | |
1da177e4 LT |
281 | |
282 | /* Find out which fragments are in front and at the back of us | |
283 | * in the chain of fragments so far. We must know where to put | |
284 | * this fragment, right? | |
285 | */ | |
d6bebca9 CG |
286 | prev = fq->q.fragments_tail; |
287 | if (!prev || FRAG6_CB(prev)->offset < offset) { | |
288 | next = NULL; | |
289 | goto found; | |
290 | } | |
1da177e4 | 291 | prev = NULL; |
5ab11c98 | 292 | for(next = fq->q.fragments; next != NULL; next = next->next) { |
1da177e4 LT |
293 | if (FRAG6_CB(next)->offset >= offset) |
294 | break; /* bingo! */ | |
295 | prev = next; | |
296 | } | |
297 | ||
d6bebca9 | 298 | found: |
5de658f8 ED |
299 | /* RFC5722, Section 4, amended by Errata ID : 3089 |
300 | * When reassembling an IPv6 datagram, if | |
70789d70 ND |
301 | * one or more its constituent fragments is determined to be an |
302 | * overlapping fragment, the entire datagram (and any constituent | |
5de658f8 | 303 | * fragments) MUST be silently discarded. |
1da177e4 | 304 | */ |
1da177e4 | 305 | |
70789d70 ND |
306 | /* Check for overlap with preceding fragment. */ |
307 | if (prev && | |
f4642141 | 308 | (FRAG6_CB(prev)->offset + prev->len) > offset) |
70789d70 | 309 | goto discard_fq; |
1da177e4 | 310 | |
70789d70 ND |
311 | /* Look for overlap with succeeding segment. */ |
312 | if (next && FRAG6_CB(next)->offset < end) | |
313 | goto discard_fq; | |
1da177e4 LT |
314 | |
315 | FRAG6_CB(skb)->offset = offset; | |
316 | ||
317 | /* Insert this fragment in the chain of fragments. */ | |
318 | skb->next = next; | |
d6bebca9 CG |
319 | if (!next) |
320 | fq->q.fragments_tail = skb; | |
1da177e4 LT |
321 | if (prev) |
322 | prev->next = skb; | |
323 | else | |
5ab11c98 | 324 | fq->q.fragments = skb; |
1da177e4 | 325 | |
f61944ef HX |
326 | dev = skb->dev; |
327 | if (dev) { | |
328 | fq->iif = dev->ifindex; | |
329 | skb->dev = NULL; | |
330 | } | |
5ab11c98 PE |
331 | fq->q.stamp = skb->tstamp; |
332 | fq->q.meat += skb->len; | |
eec2e618 | 333 | fq->ecn |= ecn; |
d433673e | 334 | add_frag_mem_limit(&fq->q, skb->truesize); |
1da177e4 LT |
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; |
1da177e4 | 342 | } |
f61944ef | 343 | |
bc578a54 | 344 | if (fq->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) && |
97599dc7 ED |
345 | fq->q.meat == fq->q.len) { |
346 | int res; | |
347 | unsigned long orefdst = skb->_skb_refdst; | |
348 | ||
349 | skb->_skb_refdst = 0UL; | |
350 | res = ip6_frag_reasm(fq, prev, dev); | |
351 | skb->_skb_refdst = orefdst; | |
352 | return res; | |
353 | } | |
f61944ef | 354 | |
97599dc7 | 355 | skb_dst_drop(skb); |
3ef0eb0d | 356 | inet_frag_lru_move(&fq->q); |
f61944ef | 357 | return -1; |
1da177e4 | 358 | |
70789d70 | 359 | discard_fq: |
b836c99f | 360 | inet_frag_kill(&fq->q, &ip6_frags); |
1da177e4 | 361 | err: |
adf30907 | 362 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), |
3bd653c8 | 363 | IPSTATS_MIB_REASMFAILS); |
1da177e4 | 364 | kfree_skb(skb); |
f61944ef | 365 | return -1; |
1da177e4 LT |
366 | } |
367 | ||
368 | /* | |
369 | * Check if this packet is complete. | |
370 | * Returns NULL on failure by any reason, and pointer | |
371 | * to current nexthdr field in reassembled frame. | |
372 | * | |
373 | * It is called with locked fq, and caller must check that | |
374 | * queue is eligible for reassembly i.e. it is not COMPLETE, | |
375 | * the last and the first frames arrived and all the bits are here. | |
376 | */ | |
f61944ef | 377 | static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev, |
1da177e4 LT |
378 | struct net_device *dev) |
379 | { | |
2bad35b7 | 380 | struct net *net = container_of(fq->q.net, struct net, ipv6.frags); |
5ab11c98 | 381 | struct sk_buff *fp, *head = fq->q.fragments; |
1da177e4 LT |
382 | int payload_len; |
383 | unsigned int nhoff; | |
ec16439e | 384 | int sum_truesize; |
eec2e618 | 385 | u8 ecn; |
1da177e4 | 386 | |
b836c99f | 387 | inet_frag_kill(&fq->q, &ip6_frags); |
1da177e4 | 388 | |
eec2e618 HFS |
389 | ecn = ip_frag_ecn_table[fq->ecn]; |
390 | if (unlikely(ecn == 0xff)) | |
391 | goto out_fail; | |
392 | ||
f61944ef HX |
393 | /* Make the one we just received the head. */ |
394 | if (prev) { | |
395 | head = prev->next; | |
396 | fp = skb_clone(head, GFP_ATOMIC); | |
397 | ||
398 | if (!fp) | |
399 | goto out_oom; | |
400 | ||
401 | fp->next = head->next; | |
d6bebca9 CG |
402 | if (!fp->next) |
403 | fq->q.fragments_tail = fp; | |
f61944ef HX |
404 | prev->next = fp; |
405 | ||
5ab11c98 PE |
406 | skb_morph(head, fq->q.fragments); |
407 | head->next = fq->q.fragments->next; | |
f61944ef | 408 | |
808db80a | 409 | consume_skb(fq->q.fragments); |
5ab11c98 | 410 | fq->q.fragments = head; |
f61944ef HX |
411 | } |
412 | ||
547b792c IJ |
413 | WARN_ON(head == NULL); |
414 | WARN_ON(FRAG6_CB(head)->offset != 0); | |
1da177e4 LT |
415 | |
416 | /* Unfragmented part is taken from the first segment. */ | |
d56f90a7 | 417 | payload_len = ((head->data - skb_network_header(head)) - |
5ab11c98 | 418 | sizeof(struct ipv6hdr) + fq->q.len - |
d56f90a7 | 419 | sizeof(struct frag_hdr)); |
1da177e4 LT |
420 | if (payload_len > IPV6_MAXPLEN) |
421 | goto out_oversize; | |
422 | ||
423 | /* Head of list must not be cloned. */ | |
14bbd6a5 | 424 | if (skb_unclone(head, GFP_ATOMIC)) |
1da177e4 LT |
425 | goto out_oom; |
426 | ||
427 | /* If the first fragment is fragmented itself, we split | |
428 | * it to two chunks: the first with data and paged part | |
429 | * and the second, holding only fragments. */ | |
21dc3301 | 430 | if (skb_has_frag_list(head)) { |
1da177e4 LT |
431 | struct sk_buff *clone; |
432 | int i, plen = 0; | |
433 | ||
434 | if ((clone = alloc_skb(0, GFP_ATOMIC)) == NULL) | |
435 | goto out_oom; | |
436 | clone->next = head->next; | |
437 | head->next = clone; | |
438 | skb_shinfo(clone)->frag_list = skb_shinfo(head)->frag_list; | |
4d9092bb | 439 | skb_frag_list_init(head); |
9e903e08 ED |
440 | for (i = 0; i < skb_shinfo(head)->nr_frags; i++) |
441 | plen += skb_frag_size(&skb_shinfo(head)->frags[i]); | |
1da177e4 LT |
442 | clone->len = clone->data_len = head->data_len - plen; |
443 | head->data_len -= clone->len; | |
444 | head->len -= clone->len; | |
445 | clone->csum = 0; | |
446 | clone->ip_summed = head->ip_summed; | |
d433673e | 447 | add_frag_mem_limit(&fq->q, clone->truesize); |
1da177e4 LT |
448 | } |
449 | ||
450 | /* We have to remove fragment header from datagram and to relocate | |
451 | * header in order to calculate ICV correctly. */ | |
452 | nhoff = fq->nhoffset; | |
b0e380b1 | 453 | skb_network_header(head)[nhoff] = skb_transport_header(head)[0]; |
1ab1457c | 454 | memmove(head->head + sizeof(struct frag_hdr), head->head, |
1da177e4 | 455 | (head->data - head->head) - sizeof(struct frag_hdr)); |
b0e380b1 ACM |
456 | head->mac_header += sizeof(struct frag_hdr); |
457 | head->network_header += sizeof(struct frag_hdr); | |
1da177e4 | 458 | |
badff6d0 | 459 | skb_reset_transport_header(head); |
d56f90a7 | 460 | skb_push(head, head->data - skb_network_header(head)); |
1da177e4 | 461 | |
ec16439e ED |
462 | sum_truesize = head->truesize; |
463 | for (fp = head->next; fp;) { | |
464 | bool headstolen; | |
465 | int delta; | |
466 | struct sk_buff *next = fp->next; | |
467 | ||
468 | sum_truesize += fp->truesize; | |
1da177e4 LT |
469 | if (head->ip_summed != fp->ip_summed) |
470 | head->ip_summed = CHECKSUM_NONE; | |
84fa7933 | 471 | else if (head->ip_summed == CHECKSUM_COMPLETE) |
1da177e4 | 472 | head->csum = csum_add(head->csum, fp->csum); |
ec16439e ED |
473 | |
474 | if (skb_try_coalesce(head, fp, &headstolen, &delta)) { | |
475 | kfree_skb_partial(fp, headstolen); | |
476 | } else { | |
477 | if (!skb_shinfo(head)->frag_list) | |
478 | skb_shinfo(head)->frag_list = fp; | |
479 | head->data_len += fp->len; | |
480 | head->len += fp->len; | |
481 | head->truesize += fp->truesize; | |
482 | } | |
483 | fp = next; | |
1da177e4 | 484 | } |
d433673e | 485 | sub_frag_mem_limit(&fq->q, sum_truesize); |
1da177e4 LT |
486 | |
487 | head->next = NULL; | |
488 | head->dev = dev; | |
5ab11c98 | 489 | head->tstamp = fq->q.stamp; |
0660e03f | 490 | ipv6_hdr(head)->payload_len = htons(payload_len); |
eec2e618 | 491 | ipv6_change_dsfield(ipv6_hdr(head), 0xff, ecn); |
951dbc8a | 492 | IP6CB(head)->nhoff = nhoff; |
1da177e4 | 493 | |
1da177e4 | 494 | /* Yes, and fold redundant checksum back. 8) */ |
84fa7933 | 495 | if (head->ip_summed == CHECKSUM_COMPLETE) |
d56f90a7 | 496 | head->csum = csum_partial(skb_network_header(head), |
cfe1fc77 | 497 | skb_network_header_len(head), |
d56f90a7 | 498 | head->csum); |
1da177e4 | 499 | |
a11d206d | 500 | rcu_read_lock(); |
2bad35b7 | 501 | IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS); |
a11d206d | 502 | rcu_read_unlock(); |
5ab11c98 | 503 | fq->q.fragments = NULL; |
d6bebca9 | 504 | fq->q.fragments_tail = NULL; |
1da177e4 LT |
505 | return 1; |
506 | ||
507 | out_oversize: | |
e87cc472 | 508 | net_dbg_ratelimited("ip6_frag_reasm: payload len = %d\n", payload_len); |
1da177e4 LT |
509 | goto out_fail; |
510 | out_oom: | |
e87cc472 | 511 | net_dbg_ratelimited("ip6_frag_reasm: no memory for reassembly\n"); |
1da177e4 | 512 | out_fail: |
a11d206d | 513 | rcu_read_lock(); |
2bad35b7 | 514 | IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS); |
a11d206d | 515 | rcu_read_unlock(); |
1da177e4 LT |
516 | return -1; |
517 | } | |
518 | ||
e5bbef20 | 519 | static int ipv6_frag_rcv(struct sk_buff *skb) |
1da177e4 | 520 | { |
1da177e4 LT |
521 | struct frag_hdr *fhdr; |
522 | struct frag_queue *fq; | |
b71d1d42 | 523 | const struct ipv6hdr *hdr = ipv6_hdr(skb); |
adf30907 | 524 | struct net *net = dev_net(skb_dst(skb)->dev); |
6b102865 | 525 | int evicted; |
1da177e4 | 526 | |
adf30907 | 527 | IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMREQDS); |
1da177e4 LT |
528 | |
529 | /* Jumbo payload inhibits frag. header */ | |
98b3377c DL |
530 | if (hdr->payload_len==0) |
531 | goto fail_hdr; | |
532 | ||
ea2ae17d | 533 | if (!pskb_may_pull(skb, (skb_transport_offset(skb) + |
98b3377c DL |
534 | sizeof(struct frag_hdr)))) |
535 | goto fail_hdr; | |
1da177e4 | 536 | |
0660e03f | 537 | hdr = ipv6_hdr(skb); |
9c70220b | 538 | fhdr = (struct frag_hdr *)skb_transport_header(skb); |
1da177e4 LT |
539 | |
540 | if (!(fhdr->frag_off & htons(0xFFF9))) { | |
541 | /* It is not a fragmented frame */ | |
b0e380b1 | 542 | skb->transport_header += sizeof(struct frag_hdr); |
483a47d2 | 543 | IP6_INC_STATS_BH(net, |
adf30907 | 544 | ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMOKS); |
1da177e4 | 545 | |
d56f90a7 | 546 | IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb); |
1da177e4 LT |
547 | return 1; |
548 | } | |
549 | ||
6b102865 AW |
550 | evicted = inet_frag_evictor(&net->ipv6.frags, &ip6_frags, false); |
551 | if (evicted) | |
552 | IP6_ADD_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), | |
553 | IPSTATS_MIB_REASMFAILS, evicted); | |
1da177e4 | 554 | |
eec2e618 HFS |
555 | fq = fq_find(net, fhdr->identification, &hdr->saddr, &hdr->daddr, |
556 | ip6_frag_ecn(hdr)); | |
9546377c | 557 | if (fq != NULL) { |
f61944ef | 558 | int ret; |
1da177e4 | 559 | |
5ab11c98 | 560 | spin_lock(&fq->q.lock); |
1da177e4 | 561 | |
f61944ef | 562 | ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff); |
1da177e4 | 563 | |
5ab11c98 | 564 | spin_unlock(&fq->q.lock); |
b836c99f | 565 | inet_frag_put(&fq->q, &ip6_frags); |
1da177e4 LT |
566 | return ret; |
567 | } | |
568 | ||
adf30907 | 569 | IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMFAILS); |
1da177e4 LT |
570 | kfree_skb(skb); |
571 | return -1; | |
98b3377c DL |
572 | |
573 | fail_hdr: | |
adf30907 | 574 | IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS); |
98b3377c DL |
575 | icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb)); |
576 | return -1; | |
1da177e4 LT |
577 | } |
578 | ||
41135cc8 | 579 | static const struct inet6_protocol frag_protocol = |
1da177e4 LT |
580 | { |
581 | .handler = ipv6_frag_rcv, | |
582 | .flags = INET6_PROTO_NOPOLICY, | |
583 | }; | |
584 | ||
8d8354d2 | 585 | #ifdef CONFIG_SYSCTL |
0a64b4b8 | 586 | static struct ctl_table ip6_frags_ns_ctl_table[] = { |
8d8354d2 | 587 | { |
8d8354d2 | 588 | .procname = "ip6frag_high_thresh", |
e31e0bdc | 589 | .data = &init_net.ipv6.frags.high_thresh, |
8d8354d2 PE |
590 | .maxlen = sizeof(int), |
591 | .mode = 0644, | |
6d9f239a | 592 | .proc_handler = proc_dointvec |
8d8354d2 PE |
593 | }, |
594 | { | |
8d8354d2 | 595 | .procname = "ip6frag_low_thresh", |
e31e0bdc | 596 | .data = &init_net.ipv6.frags.low_thresh, |
8d8354d2 PE |
597 | .maxlen = sizeof(int), |
598 | .mode = 0644, | |
6d9f239a | 599 | .proc_handler = proc_dointvec |
8d8354d2 PE |
600 | }, |
601 | { | |
8d8354d2 | 602 | .procname = "ip6frag_time", |
b2fd5321 | 603 | .data = &init_net.ipv6.frags.timeout, |
8d8354d2 PE |
604 | .maxlen = sizeof(int), |
605 | .mode = 0644, | |
6d9f239a | 606 | .proc_handler = proc_dointvec_jiffies, |
8d8354d2 | 607 | }, |
7d291ebb PE |
608 | { } |
609 | }; | |
610 | ||
611 | static struct ctl_table ip6_frags_ctl_table[] = { | |
8d8354d2 | 612 | { |
8d8354d2 | 613 | .procname = "ip6frag_secret_interval", |
3b4bc4a2 | 614 | .data = &ip6_frags.secret_interval, |
8d8354d2 PE |
615 | .maxlen = sizeof(int), |
616 | .mode = 0644, | |
6d9f239a | 617 | .proc_handler = proc_dointvec_jiffies, |
8d8354d2 PE |
618 | }, |
619 | { } | |
620 | }; | |
621 | ||
2c8c1e72 | 622 | static int __net_init ip6_frags_ns_sysctl_register(struct net *net) |
8d8354d2 | 623 | { |
e4a2d5c2 | 624 | struct ctl_table *table; |
8d8354d2 PE |
625 | struct ctl_table_header *hdr; |
626 | ||
0a64b4b8 | 627 | table = ip6_frags_ns_ctl_table; |
09ad9bc7 | 628 | if (!net_eq(net, &init_net)) { |
0a64b4b8 | 629 | table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL); |
e4a2d5c2 PE |
630 | if (table == NULL) |
631 | goto err_alloc; | |
632 | ||
e31e0bdc PE |
633 | table[0].data = &net->ipv6.frags.high_thresh; |
634 | table[1].data = &net->ipv6.frags.low_thresh; | |
b2fd5321 | 635 | table[2].data = &net->ipv6.frags.timeout; |
464dc801 EB |
636 | |
637 | /* Don't export sysctls to unprivileged users */ | |
638 | if (net->user_ns != &init_user_ns) | |
639 | table[0].procname = NULL; | |
e4a2d5c2 PE |
640 | } |
641 | ||
ec8f23ce | 642 | hdr = register_net_sysctl(net, "net/ipv6", table); |
e4a2d5c2 PE |
643 | if (hdr == NULL) |
644 | goto err_reg; | |
645 | ||
646 | net->ipv6.sysctl.frags_hdr = hdr; | |
647 | return 0; | |
648 | ||
649 | err_reg: | |
09ad9bc7 | 650 | if (!net_eq(net, &init_net)) |
e4a2d5c2 PE |
651 | kfree(table); |
652 | err_alloc: | |
653 | return -ENOMEM; | |
654 | } | |
655 | ||
2c8c1e72 | 656 | static void __net_exit ip6_frags_ns_sysctl_unregister(struct net *net) |
e4a2d5c2 PE |
657 | { |
658 | struct ctl_table *table; | |
659 | ||
660 | table = net->ipv6.sysctl.frags_hdr->ctl_table_arg; | |
661 | unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr); | |
3705e11a YH |
662 | if (!net_eq(net, &init_net)) |
663 | kfree(table); | |
8d8354d2 | 664 | } |
7d291ebb PE |
665 | |
666 | static struct ctl_table_header *ip6_ctl_header; | |
667 | ||
668 | static int ip6_frags_sysctl_register(void) | |
669 | { | |
43444757 | 670 | ip6_ctl_header = register_net_sysctl(&init_net, "net/ipv6", |
7d291ebb PE |
671 | ip6_frags_ctl_table); |
672 | return ip6_ctl_header == NULL ? -ENOMEM : 0; | |
673 | } | |
674 | ||
675 | static void ip6_frags_sysctl_unregister(void) | |
676 | { | |
677 | unregister_net_sysctl_table(ip6_ctl_header); | |
678 | } | |
8d8354d2 | 679 | #else |
0a64b4b8 | 680 | static inline int ip6_frags_ns_sysctl_register(struct net *net) |
e71e0349 | 681 | { |
8d8354d2 PE |
682 | return 0; |
683 | } | |
e4a2d5c2 | 684 | |
0a64b4b8 | 685 | static inline void ip6_frags_ns_sysctl_unregister(struct net *net) |
e4a2d5c2 PE |
686 | { |
687 | } | |
7d291ebb PE |
688 | |
689 | static inline int ip6_frags_sysctl_register(void) | |
690 | { | |
691 | return 0; | |
692 | } | |
693 | ||
694 | static inline void ip6_frags_sysctl_unregister(void) | |
695 | { | |
696 | } | |
8d8354d2 | 697 | #endif |
7d460db9 | 698 | |
2c8c1e72 | 699 | static int __net_init ipv6_frags_init_net(struct net *net) |
8d8354d2 | 700 | { |
7c070aa9 SW |
701 | net->ipv6.frags.high_thresh = IPV6_FRAG_HIGH_THRESH; |
702 | net->ipv6.frags.low_thresh = IPV6_FRAG_LOW_THRESH; | |
b2fd5321 | 703 | net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT; |
8d8354d2 | 704 | |
e5a2bb84 PE |
705 | inet_frags_init_net(&net->ipv6.frags); |
706 | ||
0a64b4b8 | 707 | return ip6_frags_ns_sysctl_register(net); |
e71e0349 DL |
708 | } |
709 | ||
2c8c1e72 | 710 | static void __net_exit ipv6_frags_exit_net(struct net *net) |
81566e83 | 711 | { |
0a64b4b8 | 712 | ip6_frags_ns_sysctl_unregister(net); |
81566e83 PE |
713 | inet_frags_exit_net(&net->ipv6.frags, &ip6_frags); |
714 | } | |
715 | ||
716 | static struct pernet_operations ip6_frags_ops = { | |
717 | .init = ipv6_frags_init_net, | |
718 | .exit = ipv6_frags_exit_net, | |
719 | }; | |
720 | ||
853cbbaa | 721 | int __init ipv6_frag_init(void) |
1da177e4 | 722 | { |
853cbbaa | 723 | int ret; |
1da177e4 | 724 | |
853cbbaa DL |
725 | ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT); |
726 | if (ret) | |
727 | goto out; | |
e71e0349 | 728 | |
7d291ebb PE |
729 | ret = ip6_frags_sysctl_register(); |
730 | if (ret) | |
731 | goto err_sysctl; | |
732 | ||
0002c630 PE |
733 | ret = register_pernet_subsys(&ip6_frags_ops); |
734 | if (ret) | |
735 | goto err_pernet; | |
8d8354d2 | 736 | |
321a3a99 | 737 | ip6_frags.hashfn = ip6_hashfn; |
c6fda282 | 738 | ip6_frags.constructor = ip6_frag_init; |
c9547709 | 739 | ip6_frags.destructor = NULL; |
1e4b8287 PE |
740 | ip6_frags.skb_free = NULL; |
741 | ip6_frags.qsize = sizeof(struct frag_queue); | |
abd6523d | 742 | ip6_frags.match = ip6_frag_match; |
e521db9d | 743 | ip6_frags.frag_expire = ip6_frag_expire; |
3b4bc4a2 | 744 | ip6_frags.secret_interval = 10 * 60 * HZ; |
7eb95156 | 745 | inet_frags_init(&ip6_frags); |
853cbbaa DL |
746 | out: |
747 | return ret; | |
0002c630 PE |
748 | |
749 | err_pernet: | |
7d291ebb PE |
750 | ip6_frags_sysctl_unregister(); |
751 | err_sysctl: | |
0002c630 PE |
752 | inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT); |
753 | goto out; | |
853cbbaa DL |
754 | } |
755 | ||
756 | void ipv6_frag_exit(void) | |
757 | { | |
758 | inet_frags_fini(&ip6_frags); | |
7d291ebb | 759 | ip6_frags_sysctl_unregister(); |
81566e83 | 760 | unregister_pernet_subsys(&ip6_frags_ops); |
853cbbaa | 761 | inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT); |
1da177e4 | 762 | } |