?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
netfilter: nf_conntrack: add support for "conntrack zones"
[linux/fpc-iii.git] / net / netfilter / xt_hashlimit.c
blobe47fb805ffb485e04f3523892892c7cc111f7d43
1 /*
2 * xt_hashlimit - Netfilter module to limit the number of packets per time
3 * seperately for each hashbucket (sourceip/sourceport/dstip/dstport)
4 *
5 * (C) 2003-2004 by Harald Welte <laforge@netfilter.org>
6 * Copyright © CC Computer Consultants GmbH, 2007 - 2008
7 *
8 * Development of this code was funded by Astaro AG, http://www.astaro.com/
9 */
10 #include <linux/module.h>
11 #include <linux/spinlock.h>
12 #include <linux/random.h>
13 #include <linux/jhash.h>
14 #include <linux/slab.h>
15 #include <linux/vmalloc.h>
16 #include <linux/proc_fs.h>
17 #include <linux/seq_file.h>
18 #include <linux/list.h>
19 #include <linux/skbuff.h>
20 #include <linux/mm.h>
21 #include <linux/in.h>
22 #include <linux/ip.h>
23 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
24 #include <linux/ipv6.h>
25 #include <net/ipv6.h>
26 #endif
27
28 #include <net/net_namespace.h>
29 #include <net/netns/generic.h>
30
31 #include <linux/netfilter/x_tables.h>
32 #include <linux/netfilter_ipv4/ip_tables.h>
33 #include <linux/netfilter_ipv6/ip6_tables.h>
34 #include <linux/netfilter/xt_hashlimit.h>
35 #include <linux/mutex.h>
36
37 MODULE_LICENSE("GPL");
38 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
39 MODULE_AUTHOR("Jan Engelhardt <jengelh@computergmbh.de>");
40 MODULE_DESCRIPTION("Xtables: per hash-bucket rate-limit match");
41 MODULE_ALIAS("ipt_hashlimit");
42 MODULE_ALIAS("ip6t_hashlimit");
43
44 struct hashlimit_net {
45 struct hlist_head htables;
46 struct proc_dir_entry *ipt_hashlimit;
47 struct proc_dir_entry *ip6t_hashlimit;
48 };
49
50 static int hashlimit_net_id;
51 static inline struct hashlimit_net *hashlimit_pernet(struct net *net)
52 {
53 return net_generic(net, hashlimit_net_id);
54 }
55
56 /* need to declare this at the top */
57 static const struct file_operations dl_file_ops;
58
59 /* hash table crap */
60 struct dsthash_dst {
61 union {
62 struct {
63 __be32 src;
64 __be32 dst;
65 } ip;
66 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
67 struct {
68 __be32 src[4];
69 __be32 dst[4];
70 } ip6;
71 #endif
72 };
73 __be16 src_port;
74 __be16 dst_port;
75 };
76
77 struct dsthash_ent {
78 /* static / read-only parts in the beginning */
79 struct hlist_node node;
80 struct dsthash_dst dst;
81
82 /* modified structure members in the end */
83 unsigned long expires; /* precalculated expiry time */
84 struct {
85 unsigned long prev; /* last modification */
86 u_int32_t credit;
87 u_int32_t credit_cap, cost;
88 } rateinfo;
89 };
90
91 struct xt_hashlimit_htable {
92 struct hlist_node node; /* global list of all htables */
93 int use;
94 u_int8_t family;
95 bool rnd_initialized;
96
97 struct hashlimit_cfg1 cfg; /* config */
98
99 /* used internally */
100 spinlock_t lock; /* lock for list_head */
101 u_int32_t rnd; /* random seed for hash */
102 unsigned int count; /* number entries in table */
103 struct timer_list timer; /* timer for gc */
104
105 /* seq_file stuff */
106 struct proc_dir_entry *pde;
107 struct net *net;
108
109 struct hlist_head hash[0]; /* hashtable itself */
110 };
111
112 static DEFINE_MUTEX(hashlimit_mutex); /* protects htables list */
113 static struct kmem_cache *hashlimit_cachep __read_mostly;
114
115 static inline bool dst_cmp(const struct dsthash_ent *ent,
116 const struct dsthash_dst *b)
117 {
118 return !memcmp(&ent->dst, b, sizeof(ent->dst));
119 }
120
121 static u_int32_t
122 hash_dst(const struct xt_hashlimit_htable *ht, const struct dsthash_dst *dst)
123 {
124 u_int32_t hash = jhash2((const u32 *)dst,
125 sizeof(*dst)/sizeof(u32),
126 ht->rnd);
127 /*
128 * Instead of returning hash % ht->cfg.size (implying a divide)
129 * we return the high 32 bits of the (hash * ht->cfg.size) that will
130 * give results between [0 and cfg.size-1] and same hash distribution,
131 * but using a multiply, less expensive than a divide
132 */
133 return ((u64)hash * ht->cfg.size) >> 32;
134 }
135
136 static struct dsthash_ent *
137 dsthash_find(const struct xt_hashlimit_htable *ht,
138 const struct dsthash_dst *dst)
139 {
140 struct dsthash_ent *ent;
141 struct hlist_node *pos;
142 u_int32_t hash = hash_dst(ht, dst);
143
144 if (!hlist_empty(&ht->hash[hash])) {
145 hlist_for_each_entry(ent, pos, &ht->hash[hash], node)
146 if (dst_cmp(ent, dst))
147 return ent;
148 }
149 return NULL;
150 }
151
152 /* allocate dsthash_ent, initialize dst, put in htable and lock it */
153 static struct dsthash_ent *
154 dsthash_alloc_init(struct xt_hashlimit_htable *ht,
155 const struct dsthash_dst *dst)
156 {
157 struct dsthash_ent *ent;
158
159 /* initialize hash with random val at the time we allocate
160 * the first hashtable entry */
161 if (!ht->rnd_initialized) {
162 get_random_bytes(&ht->rnd, sizeof(ht->rnd));
163 ht->rnd_initialized = true;
164 }
165
166 if (ht->cfg.max && ht->count >= ht->cfg.max) {
167 /* FIXME: do something. question is what.. */
168 if (net_ratelimit())
169 printk(KERN_WARNING
170 "xt_hashlimit: max count of %u reached\n",
171 ht->cfg.max);
172 return NULL;
173 }
174
175 ent = kmem_cache_alloc(hashlimit_cachep, GFP_ATOMIC);
176 if (!ent) {
177 if (net_ratelimit())
178 printk(KERN_ERR
179 "xt_hashlimit: can't allocate dsthash_ent\n");
180 return NULL;
181 }
182 memcpy(&ent->dst, dst, sizeof(ent->dst));
183
184 hlist_add_head(&ent->node, &ht->hash[hash_dst(ht, dst)]);
185 ht->count++;
186 return ent;
187 }
188
189 static inline void
190 dsthash_free(struct xt_hashlimit_htable *ht, struct dsthash_ent *ent)
191 {
192 hlist_del(&ent->node);
193 kmem_cache_free(hashlimit_cachep, ent);
194 ht->count--;
195 }
196 static void htable_gc(unsigned long htlong);
197
198 static int htable_create_v0(struct net *net, struct xt_hashlimit_info *minfo, u_int8_t family)
199 {
200 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
201 struct xt_hashlimit_htable *hinfo;
202 unsigned int size;
203 unsigned int i;
204
205 if (minfo->cfg.size)
206 size = minfo->cfg.size;
207 else {
208 size = ((totalram_pages << PAGE_SHIFT) / 16384) /
209 sizeof(struct list_head);
210 if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE))
211 size = 8192;
212 if (size < 16)
213 size = 16;
214 }
215 /* FIXME: don't use vmalloc() here or anywhere else -HW */
216 hinfo = vmalloc(sizeof(struct xt_hashlimit_htable) +
217 sizeof(struct list_head) * size);
218 if (!hinfo) {
219 printk(KERN_ERR "xt_hashlimit: unable to create hashtable\n");
220 return -1;
221 }
222 minfo->hinfo = hinfo;
223
224 /* copy match config into hashtable config */
225 hinfo->cfg.mode = minfo->cfg.mode;
226 hinfo->cfg.avg = minfo->cfg.avg;
227 hinfo->cfg.burst = minfo->cfg.burst;
228 hinfo->cfg.max = minfo->cfg.max;
229 hinfo->cfg.gc_interval = minfo->cfg.gc_interval;
230 hinfo->cfg.expire = minfo->cfg.expire;
231
232 if (family == NFPROTO_IPV4)
233 hinfo->cfg.srcmask = hinfo->cfg.dstmask = 32;
234 else
235 hinfo->cfg.srcmask = hinfo->cfg.dstmask = 128;
236
237 hinfo->cfg.size = size;
238 if (!hinfo->cfg.max)
239 hinfo->cfg.max = 8 * hinfo->cfg.size;
240 else if (hinfo->cfg.max < hinfo->cfg.size)
241 hinfo->cfg.max = hinfo->cfg.size;
242
243 for (i = 0; i < hinfo->cfg.size; i++)
244 INIT_HLIST_HEAD(&hinfo->hash[i]);
245
246 hinfo->use = 1;
247 hinfo->count = 0;
248 hinfo->family = family;
249 hinfo->rnd_initialized = false;
250 spin_lock_init(&hinfo->lock);
251 hinfo->pde = proc_create_data(minfo->name, 0,
252 (family == NFPROTO_IPV4) ?
253 hashlimit_net->ipt_hashlimit : hashlimit_net->ip6t_hashlimit,
254 &dl_file_ops, hinfo);
255 if (!hinfo->pde) {
256 vfree(hinfo);
257 return -1;
258 }
259 hinfo->net = net;
260
261 setup_timer(&hinfo->timer, htable_gc, (unsigned long )hinfo);
262 hinfo->timer.expires = jiffies + msecs_to_jiffies(hinfo->cfg.gc_interval);
263 add_timer(&hinfo->timer);
264
265 mutex_lock(&hashlimit_mutex);
266 hlist_add_head(&hinfo->node, &hashlimit_net->htables);
267 mutex_unlock(&hashlimit_mutex);
268
269 return 0;
270 }
271
272 static int htable_create(struct net *net, struct xt_hashlimit_mtinfo1 *minfo,
273 u_int8_t family)
274 {
275 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
276 struct xt_hashlimit_htable *hinfo;
277 unsigned int size;
278 unsigned int i;
279
280 if (minfo->cfg.size) {
281 size = minfo->cfg.size;
282 } else {
283 size = (totalram_pages << PAGE_SHIFT) / 16384 /
284 sizeof(struct list_head);
285 if (totalram_pages > 1024 * 1024 * 1024 / PAGE_SIZE)
286 size = 8192;
287 if (size < 16)
288 size = 16;
289 }
290 /* FIXME: don't use vmalloc() here or anywhere else -HW */
291 hinfo = vmalloc(sizeof(struct xt_hashlimit_htable) +
292 sizeof(struct list_head) * size);
293 if (hinfo == NULL) {
294 printk(KERN_ERR "xt_hashlimit: unable to create hashtable\n");
295 return -1;
296 }
297 minfo->hinfo = hinfo;
298
299 /* copy match config into hashtable config */
300 memcpy(&hinfo->cfg, &minfo->cfg, sizeof(hinfo->cfg));
301 hinfo->cfg.size = size;
302 if (hinfo->cfg.max == 0)
303 hinfo->cfg.max = 8 * hinfo->cfg.size;
304 else if (hinfo->cfg.max < hinfo->cfg.size)
305 hinfo->cfg.max = hinfo->cfg.size;
306
307 for (i = 0; i < hinfo->cfg.size; i++)
308 INIT_HLIST_HEAD(&hinfo->hash[i]);
309
310 hinfo->use = 1;
311 hinfo->count = 0;
312 hinfo->family = family;
313 hinfo->rnd_initialized = false;
314 spin_lock_init(&hinfo->lock);
315
316 hinfo->pde = proc_create_data(minfo->name, 0,
317 (family == NFPROTO_IPV4) ?
318 hashlimit_net->ipt_hashlimit : hashlimit_net->ip6t_hashlimit,
319 &dl_file_ops, hinfo);
320 if (hinfo->pde == NULL) {
321 vfree(hinfo);
322 return -1;
323 }
324 hinfo->net = net;
325
326 setup_timer(&hinfo->timer, htable_gc, (unsigned long)hinfo);
327 hinfo->timer.expires = jiffies + msecs_to_jiffies(hinfo->cfg.gc_interval);
328 add_timer(&hinfo->timer);
329
330 mutex_lock(&hashlimit_mutex);
331 hlist_add_head(&hinfo->node, &hashlimit_net->htables);
332 mutex_unlock(&hashlimit_mutex);
333
334 return 0;
335 }
336
337 static bool select_all(const struct xt_hashlimit_htable *ht,
338 const struct dsthash_ent *he)
339 {
340 return 1;
341 }
342
343 static bool select_gc(const struct xt_hashlimit_htable *ht,
344 const struct dsthash_ent *he)
345 {
346 return time_after_eq(jiffies, he->expires);
347 }
348
349 static void htable_selective_cleanup(struct xt_hashlimit_htable *ht,
350 bool (*select)(const struct xt_hashlimit_htable *ht,
351 const struct dsthash_ent *he))
352 {
353 unsigned int i;
354
355 /* lock hash table and iterate over it */
356 spin_lock_bh(&ht->lock);
357 for (i = 0; i < ht->cfg.size; i++) {
358 struct dsthash_ent *dh;
359 struct hlist_node *pos, *n;
360 hlist_for_each_entry_safe(dh, pos, n, &ht->hash[i], node) {
361 if ((*select)(ht, dh))
362 dsthash_free(ht, dh);
363 }
364 }
365 spin_unlock_bh(&ht->lock);
366 }
367
368 /* hash table garbage collector, run by timer */
369 static void htable_gc(unsigned long htlong)
370 {
371 struct xt_hashlimit_htable *ht = (struct xt_hashlimit_htable *)htlong;
372
373 htable_selective_cleanup(ht, select_gc);
374
375 /* re-add the timer accordingly */
376 ht->timer.expires = jiffies + msecs_to_jiffies(ht->cfg.gc_interval);
377 add_timer(&ht->timer);
378 }
379
380 static void htable_destroy(struct xt_hashlimit_htable *hinfo)
381 {
382 struct hashlimit_net *hashlimit_net = hashlimit_pernet(hinfo->net);
383 struct proc_dir_entry *parent;
384
385 del_timer_sync(&hinfo->timer);
386
387 if (hinfo->family == NFPROTO_IPV4)
388 parent = hashlimit_net->ipt_hashlimit;
389 else
390 parent = hashlimit_net->ip6t_hashlimit;
391 remove_proc_entry(hinfo->pde->name, parent);
392 htable_selective_cleanup(hinfo, select_all);
393 vfree(hinfo);
394 }
395
396 static struct xt_hashlimit_htable *htable_find_get(struct net *net,
397 const char *name,
398 u_int8_t family)
399 {
400 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
401 struct xt_hashlimit_htable *hinfo;
402 struct hlist_node *pos;
403
404 hlist_for_each_entry(hinfo, pos, &hashlimit_net->htables, node) {
405 if (!strcmp(name, hinfo->pde->name) &&
406 hinfo->family == family) {
407 hinfo->use++;
408 return hinfo;
409 }
410 }
411 return NULL;
412 }
413
414 static void htable_put(struct xt_hashlimit_htable *hinfo)
415 {
416 mutex_lock(&hashlimit_mutex);
417 if (--hinfo->use == 0) {
418 hlist_del(&hinfo->node);
419 htable_destroy(hinfo);
420 }
421 mutex_unlock(&hashlimit_mutex);
422 }
423
424 /* The algorithm used is the Simple Token Bucket Filter (TBF)
425 * see net/sched/sch_tbf.c in the linux source tree
426 */
427
428 /* Rusty: This is my (non-mathematically-inclined) understanding of
429 this algorithm. The `average rate' in jiffies becomes your initial
430 amount of credit `credit' and the most credit you can ever have
431 `credit_cap'. The `peak rate' becomes the cost of passing the
432 test, `cost'.
433
434 `prev' tracks the last packet hit: you gain one credit per jiffy.
435 If you get credit balance more than this, the extra credit is
436 discarded. Every time the match passes, you lose `cost' credits;
437 if you don't have that many, the test fails.
438
439 See Alexey's formal explanation in net/sched/sch_tbf.c.
440
441 To get the maximum range, we multiply by this factor (ie. you get N
442 credits per jiffy). We want to allow a rate as low as 1 per day
443 (slowest userspace tool allows), which means
444 CREDITS_PER_JIFFY*HZ*60*60*24 < 2^32 ie.
445 */
446 #define MAX_CPJ (0xFFFFFFFF / (HZ*60*60*24))
447
448 /* Repeated shift and or gives us all 1s, final shift and add 1 gives
449 * us the power of 2 below the theoretical max, so GCC simply does a
450 * shift. */
451 #define _POW2_BELOW2(x) ((x)|((x)>>1))
452 #define _POW2_BELOW4(x) (_POW2_BELOW2(x)|_POW2_BELOW2((x)>>2))
453 #define _POW2_BELOW8(x) (_POW2_BELOW4(x)|_POW2_BELOW4((x)>>4))
454 #define _POW2_BELOW16(x) (_POW2_BELOW8(x)|_POW2_BELOW8((x)>>8))
455 #define _POW2_BELOW32(x) (_POW2_BELOW16(x)|_POW2_BELOW16((x)>>16))
456 #define POW2_BELOW32(x) ((_POW2_BELOW32(x)>>1) + 1)
457
458 #define CREDITS_PER_JIFFY POW2_BELOW32(MAX_CPJ)
459
460 /* Precision saver. */
461 static inline u_int32_t
462 user2credits(u_int32_t user)
463 {
464 /* If multiplying would overflow... */
465 if (user > 0xFFFFFFFF / (HZ*CREDITS_PER_JIFFY))
466 /* Divide first. */
467 return (user / XT_HASHLIMIT_SCALE) * HZ * CREDITS_PER_JIFFY;
468
469 return (user * HZ * CREDITS_PER_JIFFY) / XT_HASHLIMIT_SCALE;
470 }
471
472 static inline void rateinfo_recalc(struct dsthash_ent *dh, unsigned long now)
473 {
474 dh->rateinfo.credit += (now - dh->rateinfo.prev) * CREDITS_PER_JIFFY;
475 if (dh->rateinfo.credit > dh->rateinfo.credit_cap)
476 dh->rateinfo.credit = dh->rateinfo.credit_cap;
477 dh->rateinfo.prev = now;
478 }
479
480 static inline __be32 maskl(__be32 a, unsigned int l)
481 {
482 return l ? htonl(ntohl(a) & ~0 << (32 - l)) : 0;
483 }
484
485 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
486 static void hashlimit_ipv6_mask(__be32 *i, unsigned int p)
487 {
488 switch (p) {
489 case 0 ... 31:
490 i[0] = maskl(i[0], p);
491 i[1] = i[2] = i[3] = 0;
492 break;
493 case 32 ... 63:
494 i[1] = maskl(i[1], p - 32);
495 i[2] = i[3] = 0;
496 break;
497 case 64 ... 95:
498 i[2] = maskl(i[2], p - 64);
499 i[3] = 0;
500 case 96 ... 127:
501 i[3] = maskl(i[3], p - 96);
502 break;
503 case 128:
504 break;
505 }
506 }
507 #endif
508
509 static int
510 hashlimit_init_dst(const struct xt_hashlimit_htable *hinfo,
511 struct dsthash_dst *dst,
512 const struct sk_buff *skb, unsigned int protoff)
513 {
514 __be16 _ports[2], *ports;
515 u8 nexthdr;
516
517 memset(dst, 0, sizeof(*dst));
518
519 switch (hinfo->family) {
520 case NFPROTO_IPV4:
521 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP)
522 dst->ip.dst = maskl(ip_hdr(skb)->daddr,
523 hinfo->cfg.dstmask);
524 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP)
525 dst->ip.src = maskl(ip_hdr(skb)->saddr,
526 hinfo->cfg.srcmask);
527
528 if (!(hinfo->cfg.mode &
529 (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT)))
530 return 0;
531 nexthdr = ip_hdr(skb)->protocol;
532 break;
533 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
534 case NFPROTO_IPV6:
535 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) {
536 memcpy(&dst->ip6.dst, &ipv6_hdr(skb)->daddr,
537 sizeof(dst->ip6.dst));
538 hashlimit_ipv6_mask(dst->ip6.dst, hinfo->cfg.dstmask);
539 }
540 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) {
541 memcpy(&dst->ip6.src, &ipv6_hdr(skb)->saddr,
542 sizeof(dst->ip6.src));
543 hashlimit_ipv6_mask(dst->ip6.src, hinfo->cfg.srcmask);
544 }
545
546 if (!(hinfo->cfg.mode &
547 (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT)))
548 return 0;
549 nexthdr = ipv6_hdr(skb)->nexthdr;
550 protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr);
551 if ((int)protoff < 0)
552 return -1;
553 break;
554 #endif
555 default:
556 BUG();
557 return 0;
558 }
559
560 switch (nexthdr) {
561 case IPPROTO_TCP:
562 case IPPROTO_UDP:
563 case IPPROTO_UDPLITE:
564 case IPPROTO_SCTP:
565 case IPPROTO_DCCP:
566 ports = skb_header_pointer(skb, protoff, sizeof(_ports),
567 &_ports);
568 break;
569 default:
570 _ports[0] = _ports[1] = 0;
571 ports = _ports;
572 break;
573 }
574 if (!ports)
575 return -1;
576 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SPT)
577 dst->src_port = ports[0];
578 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DPT)
579 dst->dst_port = ports[1];
580 return 0;
581 }
582
583 static bool
584 hashlimit_mt_v0(const struct sk_buff *skb, const struct xt_match_param *par)
585 {
586 const struct xt_hashlimit_info *r = par->matchinfo;
587 struct xt_hashlimit_htable *hinfo = r->hinfo;
588 unsigned long now = jiffies;
589 struct dsthash_ent *dh;
590 struct dsthash_dst dst;
591
592 if (hashlimit_init_dst(hinfo, &dst, skb, par->thoff) < 0)
593 goto hotdrop;
594
595 spin_lock_bh(&hinfo->lock);
596 dh = dsthash_find(hinfo, &dst);
597 if (!dh) {
598 dh = dsthash_alloc_init(hinfo, &dst);
599 if (!dh) {
600 spin_unlock_bh(&hinfo->lock);
601 goto hotdrop;
602 }
603
604 dh->expires = jiffies + msecs_to_jiffies(hinfo->cfg.expire);
605 dh->rateinfo.prev = jiffies;
606 dh->rateinfo.credit = user2credits(hinfo->cfg.avg *
607 hinfo->cfg.burst);
608 dh->rateinfo.credit_cap = user2credits(hinfo->cfg.avg *
609 hinfo->cfg.burst);
610 dh->rateinfo.cost = user2credits(hinfo->cfg.avg);
611 } else {
612 /* update expiration timeout */
613 dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire);
614 rateinfo_recalc(dh, now);
615 }
616
617 if (dh->rateinfo.credit >= dh->rateinfo.cost) {
618 /* We're underlimit. */
619 dh->rateinfo.credit -= dh->rateinfo.cost;
620 spin_unlock_bh(&hinfo->lock);
621 return true;
622 }
623
624 spin_unlock_bh(&hinfo->lock);
625
626 /* default case: we're overlimit, thus don't match */
627 return false;
628
629 hotdrop:
630 *par->hotdrop = true;
631 return false;
632 }
633
634 static bool
635 hashlimit_mt(const struct sk_buff *skb, const struct xt_match_param *par)
636 {
637 const struct xt_hashlimit_mtinfo1 *info = par->matchinfo;
638 struct xt_hashlimit_htable *hinfo = info->hinfo;
639 unsigned long now = jiffies;
640 struct dsthash_ent *dh;
641 struct dsthash_dst dst;
642
643 if (hashlimit_init_dst(hinfo, &dst, skb, par->thoff) < 0)
644 goto hotdrop;
645
646 spin_lock_bh(&hinfo->lock);
647 dh = dsthash_find(hinfo, &dst);
648 if (dh == NULL) {
649 dh = dsthash_alloc_init(hinfo, &dst);
650 if (dh == NULL) {
651 spin_unlock_bh(&hinfo->lock);
652 goto hotdrop;
653 }
654
655 dh->expires = jiffies + msecs_to_jiffies(hinfo->cfg.expire);
656 dh->rateinfo.prev = jiffies;
657 dh->rateinfo.credit = user2credits(hinfo->cfg.avg *
658 hinfo->cfg.burst);
659 dh->rateinfo.credit_cap = user2credits(hinfo->cfg.avg *
660 hinfo->cfg.burst);
661 dh->rateinfo.cost = user2credits(hinfo->cfg.avg);
662 } else {
663 /* update expiration timeout */
664 dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire);
665 rateinfo_recalc(dh, now);
666 }
667
668 if (dh->rateinfo.credit >= dh->rateinfo.cost) {
669 /* below the limit */
670 dh->rateinfo.credit -= dh->rateinfo.cost;
671 spin_unlock_bh(&hinfo->lock);
672 return !(info->cfg.mode & XT_HASHLIMIT_INVERT);
673 }
674
675 spin_unlock_bh(&hinfo->lock);
676 /* default match is underlimit - so over the limit, we need to invert */
677 return info->cfg.mode & XT_HASHLIMIT_INVERT;
678
679 hotdrop:
680 *par->hotdrop = true;
681 return false;
682 }
683
684 static bool hashlimit_mt_check_v0(const struct xt_mtchk_param *par)
685 {
686 struct net *net = par->net;
687 struct xt_hashlimit_info *r = par->matchinfo;
688
689 /* Check for overflow. */
690 if (r->cfg.burst == 0 ||
691 user2credits(r->cfg.avg * r->cfg.burst) < user2credits(r->cfg.avg)) {
692 printk(KERN_ERR "xt_hashlimit: overflow, try lower: %u/%u\n",
693 r->cfg.avg, r->cfg.burst);
694 return false;
695 }
696 if (r->cfg.mode == 0 ||
697 r->cfg.mode > (XT_HASHLIMIT_HASH_DPT |
698 XT_HASHLIMIT_HASH_DIP |
699 XT_HASHLIMIT_HASH_SIP |
700 XT_HASHLIMIT_HASH_SPT))
701 return false;
702 if (!r->cfg.gc_interval)
703 return false;
704 if (!r->cfg.expire)
705 return false;
706 if (r->name[sizeof(r->name) - 1] != '\0')
707 return false;
708
709 mutex_lock(&hashlimit_mutex);
710 r->hinfo = htable_find_get(net, r->name, par->match->family);
711 if (!r->hinfo && htable_create_v0(net, r, par->match->family) != 0) {
712 mutex_unlock(&hashlimit_mutex);
713 return false;
714 }
715 mutex_unlock(&hashlimit_mutex);
716
717 return true;
718 }
719
720 static bool hashlimit_mt_check(const struct xt_mtchk_param *par)
721 {
722 struct net *net = par->net;
723 struct xt_hashlimit_mtinfo1 *info = par->matchinfo;
724
725 /* Check for overflow. */
726 if (info->cfg.burst == 0 ||
727 user2credits(info->cfg.avg * info->cfg.burst) <
728 user2credits(info->cfg.avg)) {
729 printk(KERN_ERR "xt_hashlimit: overflow, try lower: %u/%u\n",
730 info->cfg.avg, info->cfg.burst);
731 return false;
732 }
733 if (info->cfg.gc_interval == 0 || info->cfg.expire == 0)
734 return false;
735 if (info->name[sizeof(info->name)-1] != '\0')
736 return false;
737 if (par->match->family == NFPROTO_IPV4) {
738 if (info->cfg.srcmask > 32 || info->cfg.dstmask > 32)
739 return false;
740 } else {
741 if (info->cfg.srcmask > 128 || info->cfg.dstmask > 128)
742 return false;
743 }
744
745 mutex_lock(&hashlimit_mutex);
746 info->hinfo = htable_find_get(net, info->name, par->match->family);
747 if (!info->hinfo && htable_create(net, info, par->match->family) != 0) {
748 mutex_unlock(&hashlimit_mutex);
749 return false;
750 }
751 mutex_unlock(&hashlimit_mutex);
752 return true;
753 }
754
755 static void
756 hashlimit_mt_destroy_v0(const struct xt_mtdtor_param *par)
757 {
758 const struct xt_hashlimit_info *r = par->matchinfo;
759
760 htable_put(r->hinfo);
761 }
762
763 static void hashlimit_mt_destroy(const struct xt_mtdtor_param *par)
764 {
765 const struct xt_hashlimit_mtinfo1 *info = par->matchinfo;
766
767 htable_put(info->hinfo);
768 }
769
770 #ifdef CONFIG_COMPAT
771 struct compat_xt_hashlimit_info {
772 char name[IFNAMSIZ];
773 struct hashlimit_cfg cfg;
774 compat_uptr_t hinfo;
775 compat_uptr_t master;
776 };
777
778 static void hashlimit_mt_compat_from_user(void *dst, const void *src)
779 {
780 int off = offsetof(struct compat_xt_hashlimit_info, hinfo);
781
782 memcpy(dst, src, off);
783 memset(dst + off, 0, sizeof(struct compat_xt_hashlimit_info) - off);
784 }
785
786 static int hashlimit_mt_compat_to_user(void __user *dst, const void *src)
787 {
788 int off = offsetof(struct compat_xt_hashlimit_info, hinfo);
789
790 return copy_to_user(dst, src, off) ? -EFAULT : 0;
791 }
792 #endif
793
794 static struct xt_match hashlimit_mt_reg[] __read_mostly = {
795 {
796 .name = "hashlimit",
797 .revision = 0,
798 .family = NFPROTO_IPV4,
799 .match = hashlimit_mt_v0,
800 .matchsize = sizeof(struct xt_hashlimit_info),
801 #ifdef CONFIG_COMPAT
802 .compatsize = sizeof(struct compat_xt_hashlimit_info),
803 .compat_from_user = hashlimit_mt_compat_from_user,
804 .compat_to_user = hashlimit_mt_compat_to_user,
805 #endif
806 .checkentry = hashlimit_mt_check_v0,
807 .destroy = hashlimit_mt_destroy_v0,
808 .me = THIS_MODULE
809 },
810 {
811 .name = "hashlimit",
812 .revision = 1,
813 .family = NFPROTO_IPV4,
814 .match = hashlimit_mt,
815 .matchsize = sizeof(struct xt_hashlimit_mtinfo1),
816 .checkentry = hashlimit_mt_check,
817 .destroy = hashlimit_mt_destroy,
818 .me = THIS_MODULE,
819 },
820 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
821 {
822 .name = "hashlimit",
823 .family = NFPROTO_IPV6,
824 .match = hashlimit_mt_v0,
825 .matchsize = sizeof(struct xt_hashlimit_info),
826 #ifdef CONFIG_COMPAT
827 .compatsize = sizeof(struct compat_xt_hashlimit_info),
828 .compat_from_user = hashlimit_mt_compat_from_user,
829 .compat_to_user = hashlimit_mt_compat_to_user,
830 #endif
831 .checkentry = hashlimit_mt_check_v0,
832 .destroy = hashlimit_mt_destroy_v0,
833 .me = THIS_MODULE
834 },
835 {
836 .name = "hashlimit",
837 .revision = 1,
838 .family = NFPROTO_IPV6,
839 .match = hashlimit_mt,
840 .matchsize = sizeof(struct xt_hashlimit_mtinfo1),
841 .checkentry = hashlimit_mt_check,
842 .destroy = hashlimit_mt_destroy,
843 .me = THIS_MODULE,
844 },
845 #endif
846 };
847
848 /* PROC stuff */
849 static void *dl_seq_start(struct seq_file *s, loff_t *pos)
850 __acquires(htable->lock)
851 {
852 struct xt_hashlimit_htable *htable = s->private;
853 unsigned int *bucket;
854
855 spin_lock_bh(&htable->lock);
856 if (*pos >= htable->cfg.size)
857 return NULL;
858
859 bucket = kmalloc(sizeof(unsigned int), GFP_ATOMIC);
860 if (!bucket)
861 return ERR_PTR(-ENOMEM);
862
863 *bucket = *pos;
864 return bucket;
865 }
866
867 static void *dl_seq_next(struct seq_file *s, void *v, loff_t *pos)
868 {
869 struct xt_hashlimit_htable *htable = s->private;
870 unsigned int *bucket = (unsigned int *)v;
871
872 *pos = ++(*bucket);
873 if (*pos >= htable->cfg.size) {
874 kfree(v);
875 return NULL;
876 }
877 return bucket;
878 }
879
880 static void dl_seq_stop(struct seq_file *s, void *v)
881 __releases(htable->lock)
882 {
883 struct xt_hashlimit_htable *htable = s->private;
884 unsigned int *bucket = (unsigned int *)v;
885
886 kfree(bucket);
887 spin_unlock_bh(&htable->lock);
888 }
889
890 static int dl_seq_real_show(struct dsthash_ent *ent, u_int8_t family,
891 struct seq_file *s)
892 {
893 /* recalculate to show accurate numbers */
894 rateinfo_recalc(ent, jiffies);
895
896 switch (family) {
897 case NFPROTO_IPV4:
898 return seq_printf(s, "%ld %pI4:%u->%pI4:%u %u %u %u\n",
899 (long)(ent->expires - jiffies)/HZ,
900 &ent->dst.ip.src,
901 ntohs(ent->dst.src_port),
902 &ent->dst.ip.dst,
903 ntohs(ent->dst.dst_port),
904 ent->rateinfo.credit, ent->rateinfo.credit_cap,
905 ent->rateinfo.cost);
906 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
907 case NFPROTO_IPV6:
908 return seq_printf(s, "%ld %pI6:%u->%pI6:%u %u %u %u\n",
909 (long)(ent->expires - jiffies)/HZ,
910 &ent->dst.ip6.src,
911 ntohs(ent->dst.src_port),
912 &ent->dst.ip6.dst,
913 ntohs(ent->dst.dst_port),
914 ent->rateinfo.credit, ent->rateinfo.credit_cap,
915 ent->rateinfo.cost);
916 #endif
917 default:
918 BUG();
919 return 0;
920 }
921 }
922
923 static int dl_seq_show(struct seq_file *s, void *v)
924 {
925 struct xt_hashlimit_htable *htable = s->private;
926 unsigned int *bucket = (unsigned int *)v;
927 struct dsthash_ent *ent;
928 struct hlist_node *pos;
929
930 if (!hlist_empty(&htable->hash[*bucket])) {
931 hlist_for_each_entry(ent, pos, &htable->hash[*bucket], node)
932 if (dl_seq_real_show(ent, htable->family, s))
933 return -1;
934 }
935 return 0;
936 }
937
938 static const struct seq_operations dl_seq_ops = {
939 .start = dl_seq_start,
940 .next = dl_seq_next,
941 .stop = dl_seq_stop,
942 .show = dl_seq_show
943 };
944
945 static int dl_proc_open(struct inode *inode, struct file *file)
946 {
947 int ret = seq_open(file, &dl_seq_ops);
948
949 if (!ret) {
950 struct seq_file *sf = file->private_data;
951 sf->private = PDE(inode)->data;
952 }
953 return ret;
954 }
955
956 static const struct file_operations dl_file_ops = {
957 .owner = THIS_MODULE,
958 .open = dl_proc_open,
959 .read = seq_read,
960 .llseek = seq_lseek,
961 .release = seq_release
962 };
963
964 static int __net_init hashlimit_proc_net_init(struct net *net)
965 {
966 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
967
968 hashlimit_net->ipt_hashlimit = proc_mkdir("ipt_hashlimit", net->proc_net);
969 if (!hashlimit_net->ipt_hashlimit)
970 return -ENOMEM;
971 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
972 hashlimit_net->ip6t_hashlimit = proc_mkdir("ip6t_hashlimit", net->proc_net);
973 if (!hashlimit_net->ip6t_hashlimit) {
974 proc_net_remove(net, "ipt_hashlimit");
975 return -ENOMEM;
976 }
977 #endif
978 return 0;
979 }
980
981 static void __net_exit hashlimit_proc_net_exit(struct net *net)
982 {
983 proc_net_remove(net, "ipt_hashlimit");
984 #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
985 proc_net_remove(net, "ip6t_hashlimit");
986 #endif
987 }
988
989 static int __net_init hashlimit_net_init(struct net *net)
990 {
991 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
992
993 INIT_HLIST_HEAD(&hashlimit_net->htables);
994 return hashlimit_proc_net_init(net);
995 }
996
997 static void __net_exit hashlimit_net_exit(struct net *net)
998 {
999 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
1000
1001 BUG_ON(!hlist_empty(&hashlimit_net->htables));
1002 hashlimit_proc_net_exit(net);
1003 }
1004
1005 static struct pernet_operations hashlimit_net_ops = {
1006 .init = hashlimit_net_init,
1007 .exit = hashlimit_net_exit,
1008 .id = &hashlimit_net_id,
1009 .size = sizeof(struct hashlimit_net),
1010 };
1011
1012 static int __init hashlimit_mt_init(void)
1013 {
1014 int err;
1015
1016 err = register_pernet_subsys(&hashlimit_net_ops);
1017 if (err < 0)
1018 return err;
1019 err = xt_register_matches(hashlimit_mt_reg,
1020 ARRAY_SIZE(hashlimit_mt_reg));
1021 if (err < 0)
1022 goto err1;
1023
1024 err = -ENOMEM;
1025 hashlimit_cachep = kmem_cache_create("xt_hashlimit",
1026 sizeof(struct dsthash_ent), 0, 0,
1027 NULL);
1028 if (!hashlimit_cachep) {
1029 printk(KERN_ERR "xt_hashlimit: unable to create slab cache\n");
1030 goto err2;
1031 }
1032 return 0;
1033
1034 err2:
1035 xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg));
1036 err1:
1037 unregister_pernet_subsys(&hashlimit_net_ops);
1038 return err;
1039
1040 }
1041
1042 static void __exit hashlimit_mt_exit(void)
1043 {
1044 kmem_cache_destroy(hashlimit_cachep);
1045 xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg));
1046 unregister_pernet_subsys(&hashlimit_net_ops);
1047 }
1048
1049 module_init(hashlimit_mt_init);
1050 module_exit(hashlimit_mt_exit);
Linux with added FPC-III support