Revert "ALSA: hda: Flush interrupts on disabling"
[linux/fpc-iii.git] / net / ipv6 / reassembly.c
blob4aed9c45a91ade2eb68cf90f47bcb37d7bd0088b
1 /*
2 * IPv6 fragment reassembly
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Based on: net/ipv4/ip_fragment.c
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.
17 * Fixes:
18 * Andi Kleen Make it work with multiple hosts.
19 * More RFC compliance.
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.
30 #define pr_fmt(fmt) "IPv6: " fmt
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>
46 #include <linux/skbuff.h>
47 #include <linux/slab.h>
48 #include <linux/export.h>
50 #include <net/sock.h>
51 #include <net/snmp.h>
53 #include <net/ipv6.h>
54 #include <net/ip6_route.h>
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>
60 #include <net/ipv6_frag.h>
61 #include <net/inet_ecn.h>
63 static const char ip6_frag_cache_name[] = "ip6-frags";
65 static u8 ip6_frag_ecn(const struct ipv6hdr *ipv6h)
67 return 1 << (ipv6_get_dsfield(ipv6h) & INET_ECN_MASK);
70 static struct inet_frags ip6_frags;
72 static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *skb,
73 struct sk_buff *prev_tail, struct net_device *dev);
75 static void ip6_frag_expire(unsigned long data)
77 struct frag_queue *fq;
78 struct net *net;
80 fq = container_of((struct inet_frag_queue *)data, struct frag_queue, q);
81 net = container_of(fq->q.net, struct net, ipv6.frags);
83 ip6frag_expire_frag_queue(net, fq);
86 static struct frag_queue *
87 fq_find(struct net *net, __be32 id, const struct ipv6hdr *hdr, int iif)
89 struct frag_v6_compare_key key = {
90 .id = id,
91 .saddr = hdr->saddr,
92 .daddr = hdr->daddr,
93 .user = IP6_DEFRAG_LOCAL_DELIVER,
94 .iif = iif,
96 struct inet_frag_queue *q;
98 if (!(ipv6_addr_type(&hdr->daddr) & (IPV6_ADDR_MULTICAST |
99 IPV6_ADDR_LINKLOCAL)))
100 key.iif = 0;
102 q = inet_frag_find(&net->ipv6.frags, &key);
103 if (!q)
104 return NULL;
106 return container_of(q, struct frag_queue, q);
109 static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
110 struct frag_hdr *fhdr, int nhoff,
111 u32 *prob_offset)
113 struct net *net = dev_net(skb_dst(skb)->dev);
114 int offset, end, fragsize;
115 struct sk_buff *prev_tail;
116 struct net_device *dev;
117 int err = -ENOENT;
118 u8 ecn;
120 if (fq->q.flags & INET_FRAG_COMPLETE)
121 goto err;
123 err = -EINVAL;
124 offset = ntohs(fhdr->frag_off) & ~0x7;
125 end = offset + (ntohs(ipv6_hdr(skb)->payload_len) -
126 ((u8 *)(fhdr + 1) - (u8 *)(ipv6_hdr(skb) + 1)));
128 if ((unsigned int)end > IPV6_MAXPLEN) {
129 *prob_offset = (u8 *)&fhdr->frag_off - skb_network_header(skb);
130 /* note that if prob_offset is set, the skb is freed elsewhere,
131 * we do not free it here.
133 return -1;
136 ecn = ip6_frag_ecn(ipv6_hdr(skb));
138 if (skb->ip_summed == CHECKSUM_COMPLETE) {
139 const unsigned char *nh = skb_network_header(skb);
140 skb->csum = csum_sub(skb->csum,
141 csum_partial(nh, (u8 *)(fhdr + 1) - nh,
142 0));
145 /* Is this the final fragment? */
146 if (!(fhdr->frag_off & htons(IP6_MF))) {
147 /* If we already have some bits beyond end
148 * or have different end, the segment is corrupted.
150 if (end < fq->q.len ||
151 ((fq->q.flags & INET_FRAG_LAST_IN) && end != fq->q.len))
152 goto discard_fq;
153 fq->q.flags |= INET_FRAG_LAST_IN;
154 fq->q.len = end;
155 } else {
156 /* Check if the fragment is rounded to 8 bytes.
157 * Required by the RFC.
159 if (end & 0x7) {
160 /* RFC2460 says always send parameter problem in
161 * this case. -DaveM
163 *prob_offset = offsetof(struct ipv6hdr, payload_len);
164 return -1;
166 if (end > fq->q.len) {
167 /* Some bits beyond end -> corruption. */
168 if (fq->q.flags & INET_FRAG_LAST_IN)
169 goto discard_fq;
170 fq->q.len = end;
174 if (end == offset)
175 goto discard_fq;
177 err = -ENOMEM;
178 /* Point into the IP datagram 'data' part. */
179 if (!pskb_pull(skb, (u8 *) (fhdr + 1) - skb->data))
180 goto discard_fq;
182 err = pskb_trim_rcsum(skb, end - offset);
183 if (err)
184 goto discard_fq;
186 /* Note : skb->rbnode and skb->dev share the same location. */
187 dev = skb->dev;
188 /* Makes sure compiler wont do silly aliasing games */
189 barrier();
191 prev_tail = fq->q.fragments_tail;
192 err = inet_frag_queue_insert(&fq->q, skb, offset, end);
193 if (err)
194 goto insert_error;
196 if (dev)
197 fq->iif = dev->ifindex;
199 fq->q.stamp = skb->tstamp;
200 fq->q.meat += skb->len;
201 fq->ecn |= ecn;
202 add_frag_mem_limit(fq->q.net, skb->truesize);
204 fragsize = -skb_network_offset(skb) + skb->len;
205 if (fragsize > fq->q.max_size)
206 fq->q.max_size = fragsize;
208 /* The first fragment.
209 * nhoffset is obtained from the first fragment, of course.
211 if (offset == 0) {
212 fq->nhoffset = nhoff;
213 fq->q.flags |= INET_FRAG_FIRST_IN;
216 if (fq->q.flags == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
217 fq->q.meat == fq->q.len) {
218 unsigned long orefdst = skb->_skb_refdst;
220 skb->_skb_refdst = 0UL;
221 err = ip6_frag_reasm(fq, skb, prev_tail, dev);
222 skb->_skb_refdst = orefdst;
223 return err;
226 skb_dst_drop(skb);
227 return -EINPROGRESS;
229 insert_error:
230 if (err == IPFRAG_DUP) {
231 kfree_skb(skb);
232 return -EINVAL;
234 err = -EINVAL;
235 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
236 IPSTATS_MIB_REASM_OVERLAPS);
237 discard_fq:
238 inet_frag_kill(&fq->q);
239 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
240 IPSTATS_MIB_REASMFAILS);
241 err:
242 kfree_skb(skb);
243 return err;
247 * Check if this packet is complete.
249 * It is called with locked fq, and caller must check that
250 * queue is eligible for reassembly i.e. it is not COMPLETE,
251 * the last and the first frames arrived and all the bits are here.
253 static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *skb,
254 struct sk_buff *prev_tail, struct net_device *dev)
256 struct net *net = container_of(fq->q.net, struct net, ipv6.frags);
257 unsigned int nhoff;
258 void *reasm_data;
259 int payload_len;
260 u8 ecn;
262 inet_frag_kill(&fq->q);
264 ecn = ip_frag_ecn_table[fq->ecn];
265 if (unlikely(ecn == 0xff))
266 goto out_fail;
268 reasm_data = inet_frag_reasm_prepare(&fq->q, skb, prev_tail);
269 if (!reasm_data)
270 goto out_oom;
272 payload_len = ((skb->data - skb_network_header(skb)) -
273 sizeof(struct ipv6hdr) + fq->q.len -
274 sizeof(struct frag_hdr));
275 if (payload_len > IPV6_MAXPLEN)
276 goto out_oversize;
278 /* We have to remove fragment header from datagram and to relocate
279 * header in order to calculate ICV correctly. */
280 nhoff = fq->nhoffset;
281 skb_network_header(skb)[nhoff] = skb_transport_header(skb)[0];
282 memmove(skb->head + sizeof(struct frag_hdr), skb->head,
283 (skb->data - skb->head) - sizeof(struct frag_hdr));
284 if (skb_mac_header_was_set(skb))
285 skb->mac_header += sizeof(struct frag_hdr);
286 skb->network_header += sizeof(struct frag_hdr);
288 skb_reset_transport_header(skb);
290 inet_frag_reasm_finish(&fq->q, skb, reasm_data);
292 skb->dev = dev;
293 ipv6_hdr(skb)->payload_len = htons(payload_len);
294 ipv6_change_dsfield(ipv6_hdr(skb), 0xff, ecn);
295 IP6CB(skb)->nhoff = nhoff;
296 IP6CB(skb)->flags |= IP6SKB_FRAGMENTED;
297 IP6CB(skb)->frag_max_size = fq->q.max_size;
299 /* Yes, and fold redundant checksum back. 8) */
300 skb_postpush_rcsum(skb, skb_network_header(skb),
301 skb_network_header_len(skb));
303 rcu_read_lock();
304 __IP6_INC_STATS(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
305 rcu_read_unlock();
306 fq->q.fragments = NULL;
307 fq->q.rb_fragments = RB_ROOT;
308 fq->q.fragments_tail = NULL;
309 fq->q.last_run_head = NULL;
310 return 1;
312 out_oversize:
313 net_dbg_ratelimited("ip6_frag_reasm: payload len = %d\n", payload_len);
314 goto out_fail;
315 out_oom:
316 net_dbg_ratelimited("ip6_frag_reasm: no memory for reassembly\n");
317 out_fail:
318 rcu_read_lock();
319 __IP6_INC_STATS(net, __in6_dev_get(dev), IPSTATS_MIB_REASMFAILS);
320 rcu_read_unlock();
321 inet_frag_kill(&fq->q);
322 return -1;
325 static int ipv6_frag_rcv(struct sk_buff *skb)
327 struct frag_hdr *fhdr;
328 struct frag_queue *fq;
329 const struct ipv6hdr *hdr = ipv6_hdr(skb);
330 struct net *net = dev_net(skb_dst(skb)->dev);
331 int iif;
333 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED)
334 goto fail_hdr;
336 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMREQDS);
338 /* Jumbo payload inhibits frag. header */
339 if (hdr->payload_len == 0)
340 goto fail_hdr;
342 if (!pskb_may_pull(skb, (skb_transport_offset(skb) +
343 sizeof(struct frag_hdr))))
344 goto fail_hdr;
346 hdr = ipv6_hdr(skb);
347 fhdr = (struct frag_hdr *)skb_transport_header(skb);
349 if (!(fhdr->frag_off & htons(0xFFF9))) {
350 /* It is not a fragmented frame */
351 skb->transport_header += sizeof(struct frag_hdr);
352 __IP6_INC_STATS(net,
353 ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMOKS);
355 IP6CB(skb)->nhoff = (u8 *)fhdr - skb_network_header(skb);
356 IP6CB(skb)->flags |= IP6SKB_FRAGMENTED;
357 return 1;
360 iif = skb->dev ? skb->dev->ifindex : 0;
361 fq = fq_find(net, fhdr->identification, hdr, iif);
362 if (fq) {
363 u32 prob_offset = 0;
364 int ret;
366 spin_lock(&fq->q.lock);
368 fq->iif = iif;
369 ret = ip6_frag_queue(fq, skb, fhdr, IP6CB(skb)->nhoff,
370 &prob_offset);
372 spin_unlock(&fq->q.lock);
373 inet_frag_put(&fq->q);
374 if (prob_offset) {
375 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
376 IPSTATS_MIB_INHDRERRORS);
377 /* icmpv6_param_prob() calls kfree_skb(skb) */
378 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, prob_offset);
380 return ret;
383 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_REASMFAILS);
384 kfree_skb(skb);
385 return -1;
387 fail_hdr:
388 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
389 IPSTATS_MIB_INHDRERRORS);
390 icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, skb_network_header_len(skb));
391 return -1;
394 static const struct inet6_protocol frag_protocol = {
395 .handler = ipv6_frag_rcv,
396 .flags = INET6_PROTO_NOPOLICY,
399 #ifdef CONFIG_SYSCTL
401 static struct ctl_table ip6_frags_ns_ctl_table[] = {
403 .procname = "ip6frag_high_thresh",
404 .data = &init_net.ipv6.frags.high_thresh,
405 .maxlen = sizeof(unsigned long),
406 .mode = 0644,
407 .proc_handler = proc_doulongvec_minmax,
408 .extra1 = &init_net.ipv6.frags.low_thresh
411 .procname = "ip6frag_low_thresh",
412 .data = &init_net.ipv6.frags.low_thresh,
413 .maxlen = sizeof(unsigned long),
414 .mode = 0644,
415 .proc_handler = proc_doulongvec_minmax,
416 .extra2 = &init_net.ipv6.frags.high_thresh
419 .procname = "ip6frag_time",
420 .data = &init_net.ipv6.frags.timeout,
421 .maxlen = sizeof(int),
422 .mode = 0644,
423 .proc_handler = proc_dointvec_jiffies,
428 /* secret interval has been deprecated */
429 static int ip6_frags_secret_interval_unused;
430 static struct ctl_table ip6_frags_ctl_table[] = {
432 .procname = "ip6frag_secret_interval",
433 .data = &ip6_frags_secret_interval_unused,
434 .maxlen = sizeof(int),
435 .mode = 0644,
436 .proc_handler = proc_dointvec_jiffies,
441 static int __net_init ip6_frags_ns_sysctl_register(struct net *net)
443 struct ctl_table *table;
444 struct ctl_table_header *hdr;
446 table = ip6_frags_ns_ctl_table;
447 if (!net_eq(net, &init_net)) {
448 table = kmemdup(table, sizeof(ip6_frags_ns_ctl_table), GFP_KERNEL);
449 if (!table)
450 goto err_alloc;
452 table[0].data = &net->ipv6.frags.high_thresh;
453 table[0].extra1 = &net->ipv6.frags.low_thresh;
454 table[0].extra2 = &init_net.ipv6.frags.high_thresh;
455 table[1].data = &net->ipv6.frags.low_thresh;
456 table[1].extra2 = &net->ipv6.frags.high_thresh;
457 table[2].data = &net->ipv6.frags.timeout;
460 hdr = register_net_sysctl(net, "net/ipv6", table);
461 if (!hdr)
462 goto err_reg;
464 net->ipv6.sysctl.frags_hdr = hdr;
465 return 0;
467 err_reg:
468 if (!net_eq(net, &init_net))
469 kfree(table);
470 err_alloc:
471 return -ENOMEM;
474 static void __net_exit ip6_frags_ns_sysctl_unregister(struct net *net)
476 struct ctl_table *table;
478 table = net->ipv6.sysctl.frags_hdr->ctl_table_arg;
479 unregister_net_sysctl_table(net->ipv6.sysctl.frags_hdr);
480 if (!net_eq(net, &init_net))
481 kfree(table);
484 static struct ctl_table_header *ip6_ctl_header;
486 static int ip6_frags_sysctl_register(void)
488 ip6_ctl_header = register_net_sysctl(&init_net, "net/ipv6",
489 ip6_frags_ctl_table);
490 return ip6_ctl_header == NULL ? -ENOMEM : 0;
493 static void ip6_frags_sysctl_unregister(void)
495 unregister_net_sysctl_table(ip6_ctl_header);
497 #else
498 static int ip6_frags_ns_sysctl_register(struct net *net)
500 return 0;
503 static void ip6_frags_ns_sysctl_unregister(struct net *net)
507 static int ip6_frags_sysctl_register(void)
509 return 0;
512 static void ip6_frags_sysctl_unregister(void)
515 #endif
517 static int __net_init ipv6_frags_init_net(struct net *net)
519 int res;
521 net->ipv6.frags.high_thresh = IPV6_FRAG_HIGH_THRESH;
522 net->ipv6.frags.low_thresh = IPV6_FRAG_LOW_THRESH;
523 net->ipv6.frags.timeout = IPV6_FRAG_TIMEOUT;
524 net->ipv6.frags.f = &ip6_frags;
526 res = inet_frags_init_net(&net->ipv6.frags);
527 if (res < 0)
528 return res;
530 res = ip6_frags_ns_sysctl_register(net);
531 if (res < 0)
532 inet_frags_exit_net(&net->ipv6.frags);
533 return res;
536 static void __net_exit ipv6_frags_exit_net(struct net *net)
538 ip6_frags_ns_sysctl_unregister(net);
539 inet_frags_exit_net(&net->ipv6.frags);
542 static struct pernet_operations ip6_frags_ops = {
543 .init = ipv6_frags_init_net,
544 .exit = ipv6_frags_exit_net,
547 static const struct rhashtable_params ip6_rhash_params = {
548 .head_offset = offsetof(struct inet_frag_queue, node),
549 .hashfn = ip6frag_key_hashfn,
550 .obj_hashfn = ip6frag_obj_hashfn,
551 .obj_cmpfn = ip6frag_obj_cmpfn,
552 .automatic_shrinking = true,
555 int __init ipv6_frag_init(void)
557 int ret;
559 ip6_frags.constructor = ip6frag_init;
560 ip6_frags.destructor = NULL;
561 ip6_frags.qsize = sizeof(struct frag_queue);
562 ip6_frags.frag_expire = ip6_frag_expire;
563 ip6_frags.frags_cache_name = ip6_frag_cache_name;
564 ip6_frags.rhash_params = ip6_rhash_params;
565 ret = inet_frags_init(&ip6_frags);
566 if (ret)
567 goto out;
569 ret = inet6_add_protocol(&frag_protocol, IPPROTO_FRAGMENT);
570 if (ret)
571 goto err_protocol;
573 ret = ip6_frags_sysctl_register();
574 if (ret)
575 goto err_sysctl;
577 ret = register_pernet_subsys(&ip6_frags_ops);
578 if (ret)
579 goto err_pernet;
581 out:
582 return ret;
584 err_pernet:
585 ip6_frags_sysctl_unregister();
586 err_sysctl:
587 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);
588 err_protocol:
589 inet_frags_fini(&ip6_frags);
590 goto out;
593 void ipv6_frag_exit(void)
595 inet_frags_fini(&ip6_frags);
596 ip6_frags_sysctl_unregister();
597 unregister_pernet_subsys(&ip6_frags_ops);
598 inet6_del_protocol(&frag_protocol, IPPROTO_FRAGMENT);