]> bbs.cooldavid.org Git - net-next-2.6.git/blame - net/netfilter/xt_hashlimit.c
Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
[net-next-2.6.git] / net / netfilter / xt_hashlimit.c
CommitLineData
09e410de
JE
1/*
2 * xt_hashlimit - Netfilter module to limit the number of packets per time
3ad2f3fb 3 * separately for each hashbucket (sourceip/sourceport/dstip/dstport)
1da177e4 4 *
09e410de
JE
5 * (C) 2003-2004 by Harald Welte <laforge@netfilter.org>
6 * Copyright © CC Computer Consultants GmbH, 2007 - 2008
1da177e4
LT
7 *
8 * Development of this code was funded by Astaro AG, http://www.astaro.com/
1da177e4 9 */
8bee4bad 10#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
1da177e4 11#include <linux/module.h>
1da177e4
LT
12#include <linux/spinlock.h>
13#include <linux/random.h>
14#include <linux/jhash.h>
15#include <linux/slab.h>
16#include <linux/vmalloc.h>
1da177e4
LT
17#include <linux/proc_fs.h>
18#include <linux/seq_file.h>
19#include <linux/list.h>
39b46fc6 20#include <linux/skbuff.h>
d7fe0f24 21#include <linux/mm.h>
39b46fc6
PM
22#include <linux/in.h>
23#include <linux/ip.h>
7b21e09d 24#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
39b46fc6 25#include <linux/ipv6.h>
193b23c5 26#include <net/ipv6.h>
7b21e09d
ED
27#endif
28
457c4cbc 29#include <net/net_namespace.h>
e89fc3f1 30#include <net/netns/generic.h>
1da177e4 31
39b46fc6 32#include <linux/netfilter/x_tables.h>
1da177e4 33#include <linux/netfilter_ipv4/ip_tables.h>
39b46fc6
PM
34#include <linux/netfilter_ipv6/ip6_tables.h>
35#include <linux/netfilter/xt_hashlimit.h>
14cc3e2b 36#include <linux/mutex.h>
1da177e4
LT
37
38MODULE_LICENSE("GPL");
39MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
408ffaa4 40MODULE_AUTHOR("Jan Engelhardt <jengelh@medozas.de>");
2ae15b64 41MODULE_DESCRIPTION("Xtables: per hash-bucket rate-limit match");
39b46fc6
PM
42MODULE_ALIAS("ipt_hashlimit");
43MODULE_ALIAS("ip6t_hashlimit");
1da177e4 44
e89fc3f1
AD
45struct hashlimit_net {
46 struct hlist_head htables;
47 struct proc_dir_entry *ipt_hashlimit;
48 struct proc_dir_entry *ip6t_hashlimit;
49};
50
51static int hashlimit_net_id;
52static inline struct hashlimit_net *hashlimit_pernet(struct net *net)
53{
54 return net_generic(net, hashlimit_net_id);
55}
56
1da177e4 57/* need to declare this at the top */
da7071d7 58static const struct file_operations dl_file_ops;
1da177e4
LT
59
60/* hash table crap */
1da177e4 61struct dsthash_dst {
39b46fc6
PM
62 union {
63 struct {
64 __be32 src;
65 __be32 dst;
66 } ip;
7b21e09d 67#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
39b46fc6
PM
68 struct {
69 __be32 src[4];
70 __be32 dst[4];
71 } ip6;
7b21e09d 72#endif
09e410de 73 };
6a19d614
AV
74 __be16 src_port;
75 __be16 dst_port;
1da177e4
LT
76};
77
78struct dsthash_ent {
79 /* static / read-only parts in the beginning */
80 struct hlist_node node;
81 struct dsthash_dst dst;
82
83 /* modified structure members in the end */
02e4eb75 84 spinlock_t lock;
1da177e4
LT
85 unsigned long expires; /* precalculated expiry time */
86 struct {
87 unsigned long prev; /* last modification */
88 u_int32_t credit;
89 u_int32_t credit_cap, cost;
90 } rateinfo;
02e4eb75 91 struct rcu_head rcu;
1da177e4
LT
92};
93
39b46fc6 94struct xt_hashlimit_htable {
1da177e4 95 struct hlist_node node; /* global list of all htables */
2eff25c1 96 int use;
76108cea 97 u_int8_t family;
89bc7a0f 98 bool rnd_initialized;
1da177e4 99
09e410de 100 struct hashlimit_cfg1 cfg; /* config */
1da177e4
LT
101
102 /* used internally */
103 spinlock_t lock; /* lock for list_head */
104 u_int32_t rnd; /* random seed for hash */
39b46fc6 105 unsigned int count; /* number entries in table */
1da177e4 106 struct timer_list timer; /* timer for gc */
1da177e4
LT
107
108 /* seq_file stuff */
109 struct proc_dir_entry *pde;
e89fc3f1 110 struct net *net;
1da177e4
LT
111
112 struct hlist_head hash[0]; /* hashtable itself */
113};
114
2eff25c1 115static DEFINE_MUTEX(hashlimit_mutex); /* protects htables list */
e18b890b 116static struct kmem_cache *hashlimit_cachep __read_mostly;
1da177e4 117
1d93a9cb 118static inline bool dst_cmp(const struct dsthash_ent *ent,
a47362a2 119 const struct dsthash_dst *b)
1da177e4 120{
39b46fc6 121 return !memcmp(&ent->dst, b, sizeof(ent->dst));
1da177e4
LT
122}
123
39b46fc6
PM
124static u_int32_t
125hash_dst(const struct xt_hashlimit_htable *ht, const struct dsthash_dst *dst)
1da177e4 126{
e2f82ac3
ED
127 u_int32_t hash = jhash2((const u32 *)dst,
128 sizeof(*dst)/sizeof(u32),
129 ht->rnd);
130 /*
131 * Instead of returning hash % ht->cfg.size (implying a divide)
132 * we return the high 32 bits of the (hash * ht->cfg.size) that will
133 * give results between [0 and cfg.size-1] and same hash distribution,
134 * but using a multiply, less expensive than a divide
135 */
136 return ((u64)hash * ht->cfg.size) >> 32;
1da177e4
LT
137}
138
39b46fc6 139static struct dsthash_ent *
a47362a2
JE
140dsthash_find(const struct xt_hashlimit_htable *ht,
141 const struct dsthash_dst *dst)
1da177e4
LT
142{
143 struct dsthash_ent *ent;
144 struct hlist_node *pos;
145 u_int32_t hash = hash_dst(ht, dst);
146
39b46fc6 147 if (!hlist_empty(&ht->hash[hash])) {
02e4eb75
ED
148 hlist_for_each_entry_rcu(ent, pos, &ht->hash[hash], node)
149 if (dst_cmp(ent, dst)) {
150 spin_lock(&ent->lock);
1da177e4 151 return ent;
02e4eb75 152 }
39b46fc6 153 }
1da177e4
LT
154 return NULL;
155}
156
157/* allocate dsthash_ent, initialize dst, put in htable and lock it */
158static struct dsthash_ent *
a47362a2
JE
159dsthash_alloc_init(struct xt_hashlimit_htable *ht,
160 const struct dsthash_dst *dst)
1da177e4
LT
161{
162 struct dsthash_ent *ent;
163
02e4eb75 164 spin_lock(&ht->lock);
1da177e4
LT
165 /* initialize hash with random val at the time we allocate
166 * the first hashtable entry */
02e4eb75 167 if (unlikely(!ht->rnd_initialized)) {
af07d241 168 get_random_bytes(&ht->rnd, sizeof(ht->rnd));
89bc7a0f 169 ht->rnd_initialized = true;
bf0857ea 170 }
1da177e4 171
39b46fc6 172 if (ht->cfg.max && ht->count >= ht->cfg.max) {
1da177e4
LT
173 /* FIXME: do something. question is what.. */
174 if (net_ratelimit())
8bee4bad 175 pr_err("max count of %u reached\n", ht->cfg.max);
02e4eb75
ED
176 ent = NULL;
177 } else
178 ent = kmem_cache_alloc(hashlimit_cachep, GFP_ATOMIC);
1da177e4
LT
179 if (!ent) {
180 if (net_ratelimit())
8bee4bad 181 pr_err("cannot allocate dsthash_ent\n");
02e4eb75
ED
182 } else {
183 memcpy(&ent->dst, dst, sizeof(ent->dst));
184 spin_lock_init(&ent->lock);
1da177e4 185
02e4eb75
ED
186 spin_lock(&ent->lock);
187 hlist_add_head_rcu(&ent->node, &ht->hash[hash_dst(ht, dst)]);
188 ht->count++;
189 }
190 spin_unlock(&ht->lock);
1da177e4
LT
191 return ent;
192}
193
02e4eb75
ED
194static void dsthash_free_rcu(struct rcu_head *head)
195{
196 struct dsthash_ent *ent = container_of(head, struct dsthash_ent, rcu);
197
198 kmem_cache_free(hashlimit_cachep, ent);
199}
200
39b46fc6
PM
201static inline void
202dsthash_free(struct xt_hashlimit_htable *ht, struct dsthash_ent *ent)
1da177e4 203{
02e4eb75
ED
204 hlist_del_rcu(&ent->node);
205 call_rcu_bh(&ent->rcu, dsthash_free_rcu);
39b46fc6 206 ht->count--;
1da177e4
LT
207}
208static void htable_gc(unsigned long htlong);
209
e89fc3f1
AD
210static int htable_create(struct net *net, struct xt_hashlimit_mtinfo1 *minfo,
211 u_int8_t family)
09e410de 212{
e89fc3f1 213 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
09e410de
JE
214 struct xt_hashlimit_htable *hinfo;
215 unsigned int size;
216 unsigned int i;
217
218 if (minfo->cfg.size) {
219 size = minfo->cfg.size;
220 } else {
4481374c 221 size = (totalram_pages << PAGE_SHIFT) / 16384 /
09e410de 222 sizeof(struct list_head);
4481374c 223 if (totalram_pages > 1024 * 1024 * 1024 / PAGE_SIZE)
09e410de
JE
224 size = 8192;
225 if (size < 16)
226 size = 16;
227 }
228 /* FIXME: don't use vmalloc() here or anywhere else -HW */
229 hinfo = vmalloc(sizeof(struct xt_hashlimit_htable) +
230 sizeof(struct list_head) * size);
85bc3f38 231 if (hinfo == NULL)
4a5a5c73 232 return -ENOMEM;
09e410de
JE
233 minfo->hinfo = hinfo;
234
235 /* copy match config into hashtable config */
236 memcpy(&hinfo->cfg, &minfo->cfg, sizeof(hinfo->cfg));
237 hinfo->cfg.size = size;
238 if (hinfo->cfg.max == 0)
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
2eff25c1 246 hinfo->use = 1;
09e410de
JE
247 hinfo->count = 0;
248 hinfo->family = family;
89bc7a0f 249 hinfo->rnd_initialized = false;
09e410de
JE
250 spin_lock_init(&hinfo->lock);
251
ee999d8b
JE
252 hinfo->pde = proc_create_data(minfo->name, 0,
253 (family == NFPROTO_IPV4) ?
e89fc3f1 254 hashlimit_net->ipt_hashlimit : hashlimit_net->ip6t_hashlimit,
ee999d8b 255 &dl_file_ops, hinfo);
09e410de
JE
256 if (hinfo->pde == NULL) {
257 vfree(hinfo);
4a5a5c73 258 return -ENOMEM;
09e410de 259 }
e89fc3f1 260 hinfo->net = net;
09e410de
JE
261
262 setup_timer(&hinfo->timer, htable_gc, (unsigned long)hinfo);
263 hinfo->timer.expires = jiffies + msecs_to_jiffies(hinfo->cfg.gc_interval);
264 add_timer(&hinfo->timer);
265
e89fc3f1 266 hlist_add_head(&hinfo->node, &hashlimit_net->htables);
09e410de
JE
267
268 return 0;
269}
270
a47362a2
JE
271static bool select_all(const struct xt_hashlimit_htable *ht,
272 const struct dsthash_ent *he)
1da177e4
LT
273{
274 return 1;
275}
276
a47362a2
JE
277static bool select_gc(const struct xt_hashlimit_htable *ht,
278 const struct dsthash_ent *he)
1da177e4 279{
cbebc51f 280 return time_after_eq(jiffies, he->expires);
1da177e4
LT
281}
282
39b46fc6 283static void htable_selective_cleanup(struct xt_hashlimit_htable *ht,
a47362a2
JE
284 bool (*select)(const struct xt_hashlimit_htable *ht,
285 const struct dsthash_ent *he))
1da177e4 286{
39b46fc6 287 unsigned int i;
1da177e4
LT
288
289 /* lock hash table and iterate over it */
290 spin_lock_bh(&ht->lock);
291 for (i = 0; i < ht->cfg.size; i++) {
292 struct dsthash_ent *dh;
293 struct hlist_node *pos, *n;
294 hlist_for_each_entry_safe(dh, pos, n, &ht->hash[i], node) {
295 if ((*select)(ht, dh))
39b46fc6 296 dsthash_free(ht, dh);
1da177e4
LT
297 }
298 }
299 spin_unlock_bh(&ht->lock);
300}
301
302/* hash table garbage collector, run by timer */
303static void htable_gc(unsigned long htlong)
304{
39b46fc6 305 struct xt_hashlimit_htable *ht = (struct xt_hashlimit_htable *)htlong;
1da177e4
LT
306
307 htable_selective_cleanup(ht, select_gc);
308
309 /* re-add the timer accordingly */
310 ht->timer.expires = jiffies + msecs_to_jiffies(ht->cfg.gc_interval);
311 add_timer(&ht->timer);
312}
313
39b46fc6 314static void htable_destroy(struct xt_hashlimit_htable *hinfo)
1da177e4 315{
e89fc3f1
AD
316 struct hashlimit_net *hashlimit_net = hashlimit_pernet(hinfo->net);
317 struct proc_dir_entry *parent;
318
967ab999 319 del_timer_sync(&hinfo->timer);
1da177e4 320
e89fc3f1
AD
321 if (hinfo->family == NFPROTO_IPV4)
322 parent = hashlimit_net->ipt_hashlimit;
323 else
324 parent = hashlimit_net->ip6t_hashlimit;
325 remove_proc_entry(hinfo->pde->name, parent);
1da177e4
LT
326 htable_selective_cleanup(hinfo, select_all);
327 vfree(hinfo);
328}
329
e89fc3f1
AD
330static struct xt_hashlimit_htable *htable_find_get(struct net *net,
331 const char *name,
76108cea 332 u_int8_t family)
1da177e4 333{
e89fc3f1 334 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
39b46fc6 335 struct xt_hashlimit_htable *hinfo;
1da177e4
LT
336 struct hlist_node *pos;
337
e89fc3f1 338 hlist_for_each_entry(hinfo, pos, &hashlimit_net->htables, node) {
39b46fc6
PM
339 if (!strcmp(name, hinfo->pde->name) &&
340 hinfo->family == family) {
2eff25c1 341 hinfo->use++;
1da177e4
LT
342 return hinfo;
343 }
344 }
1da177e4
LT
345 return NULL;
346}
347
39b46fc6 348static void htable_put(struct xt_hashlimit_htable *hinfo)
1da177e4 349{
2eff25c1
PM
350 mutex_lock(&hashlimit_mutex);
351 if (--hinfo->use == 0) {
1da177e4 352 hlist_del(&hinfo->node);
1da177e4
LT
353 htable_destroy(hinfo);
354 }
2eff25c1 355 mutex_unlock(&hashlimit_mutex);
1da177e4
LT
356}
357
1da177e4
LT
358/* The algorithm used is the Simple Token Bucket Filter (TBF)
359 * see net/sched/sch_tbf.c in the linux source tree
360 */
361
362/* Rusty: This is my (non-mathematically-inclined) understanding of
363 this algorithm. The `average rate' in jiffies becomes your initial
364 amount of credit `credit' and the most credit you can ever have
365 `credit_cap'. The `peak rate' becomes the cost of passing the
366 test, `cost'.
367
368 `prev' tracks the last packet hit: you gain one credit per jiffy.
369 If you get credit balance more than this, the extra credit is
370 discarded. Every time the match passes, you lose `cost' credits;
371 if you don't have that many, the test fails.
372
373 See Alexey's formal explanation in net/sched/sch_tbf.c.
374
375 To get the maximum range, we multiply by this factor (ie. you get N
376 credits per jiffy). We want to allow a rate as low as 1 per day
377 (slowest userspace tool allows), which means
378 CREDITS_PER_JIFFY*HZ*60*60*24 < 2^32 ie.
379*/
380#define MAX_CPJ (0xFFFFFFFF / (HZ*60*60*24))
381
382/* Repeated shift and or gives us all 1s, final shift and add 1 gives
383 * us the power of 2 below the theoretical max, so GCC simply does a
384 * shift. */
385#define _POW2_BELOW2(x) ((x)|((x)>>1))
386#define _POW2_BELOW4(x) (_POW2_BELOW2(x)|_POW2_BELOW2((x)>>2))
387#define _POW2_BELOW8(x) (_POW2_BELOW4(x)|_POW2_BELOW4((x)>>4))
388#define _POW2_BELOW16(x) (_POW2_BELOW8(x)|_POW2_BELOW8((x)>>8))
389#define _POW2_BELOW32(x) (_POW2_BELOW16(x)|_POW2_BELOW16((x)>>16))
390#define POW2_BELOW32(x) ((_POW2_BELOW32(x)>>1) + 1)
391
392#define CREDITS_PER_JIFFY POW2_BELOW32(MAX_CPJ)
393
394/* Precision saver. */
395static inline u_int32_t
396user2credits(u_int32_t user)
397{
398 /* If multiplying would overflow... */
399 if (user > 0xFFFFFFFF / (HZ*CREDITS_PER_JIFFY))
400 /* Divide first. */
39b46fc6 401 return (user / XT_HASHLIMIT_SCALE) * HZ * CREDITS_PER_JIFFY;
1da177e4 402
39b46fc6 403 return (user * HZ * CREDITS_PER_JIFFY) / XT_HASHLIMIT_SCALE;
1da177e4
LT
404}
405
406static inline void rateinfo_recalc(struct dsthash_ent *dh, unsigned long now)
407{
39b46fc6 408 dh->rateinfo.credit += (now - dh->rateinfo.prev) * CREDITS_PER_JIFFY;
1da177e4
LT
409 if (dh->rateinfo.credit > dh->rateinfo.credit_cap)
410 dh->rateinfo.credit = dh->rateinfo.credit_cap;
39b46fc6
PM
411 dh->rateinfo.prev = now;
412}
413
09e410de
JE
414static inline __be32 maskl(__be32 a, unsigned int l)
415{
1b9b70ea 416 return l ? htonl(ntohl(a) & ~0 << (32 - l)) : 0;
09e410de
JE
417}
418
3ed5df44 419#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
09e410de
JE
420static void hashlimit_ipv6_mask(__be32 *i, unsigned int p)
421{
422 switch (p) {
1b9b70ea 423 case 0 ... 31:
09e410de
JE
424 i[0] = maskl(i[0], p);
425 i[1] = i[2] = i[3] = 0;
426 break;
1b9b70ea 427 case 32 ... 63:
09e410de
JE
428 i[1] = maskl(i[1], p - 32);
429 i[2] = i[3] = 0;
430 break;
1b9b70ea 431 case 64 ... 95:
09e410de
JE
432 i[2] = maskl(i[2], p - 64);
433 i[3] = 0;
8f599229 434 break;
1b9b70ea 435 case 96 ... 127:
09e410de
JE
436 i[3] = maskl(i[3], p - 96);
437 break;
438 case 128:
439 break;
440 }
441}
3ed5df44 442#endif
09e410de 443
39b46fc6 444static int
a47362a2
JE
445hashlimit_init_dst(const struct xt_hashlimit_htable *hinfo,
446 struct dsthash_dst *dst,
39b46fc6
PM
447 const struct sk_buff *skb, unsigned int protoff)
448{
449 __be16 _ports[2], *ports;
193b23c5 450 u8 nexthdr;
aca071c1 451 int poff;
39b46fc6
PM
452
453 memset(dst, 0, sizeof(*dst));
454
455 switch (hinfo->family) {
ee999d8b 456 case NFPROTO_IPV4:
39b46fc6 457 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP)
09e410de
JE
458 dst->ip.dst = maskl(ip_hdr(skb)->daddr,
459 hinfo->cfg.dstmask);
39b46fc6 460 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP)
09e410de
JE
461 dst->ip.src = maskl(ip_hdr(skb)->saddr,
462 hinfo->cfg.srcmask);
39b46fc6
PM
463
464 if (!(hinfo->cfg.mode &
465 (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT)))
466 return 0;
eddc9ec5 467 nexthdr = ip_hdr(skb)->protocol;
39b46fc6 468 break;
02dba025 469#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
ee999d8b 470 case NFPROTO_IPV6:
09e410de
JE
471 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) {
472 memcpy(&dst->ip6.dst, &ipv6_hdr(skb)->daddr,
473 sizeof(dst->ip6.dst));
474 hashlimit_ipv6_mask(dst->ip6.dst, hinfo->cfg.dstmask);
475 }
476 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) {
477 memcpy(&dst->ip6.src, &ipv6_hdr(skb)->saddr,
478 sizeof(dst->ip6.src));
479 hashlimit_ipv6_mask(dst->ip6.src, hinfo->cfg.srcmask);
480 }
39b46fc6
PM
481
482 if (!(hinfo->cfg.mode &
483 (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT)))
484 return 0;
193b23c5
PM
485 nexthdr = ipv6_hdr(skb)->nexthdr;
486 protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr);
487 if ((int)protoff < 0)
39b46fc6
PM
488 return -1;
489 break;
490#endif
491 default:
492 BUG();
493 return 0;
494 }
495
aca071c1
CG
496 poff = proto_ports_offset(nexthdr);
497 if (poff >= 0) {
498 ports = skb_header_pointer(skb, protoff + poff, sizeof(_ports),
39b46fc6 499 &_ports);
aca071c1 500 } else {
39b46fc6
PM
501 _ports[0] = _ports[1] = 0;
502 ports = _ports;
39b46fc6
PM
503 }
504 if (!ports)
505 return -1;
506 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SPT)
507 dst->src_port = ports[0];
508 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DPT)
509 dst->dst_port = ports[1];
510 return 0;
1da177e4
LT
511}
512
ccb79bdc 513static bool
62fc8051 514hashlimit_mt(const struct sk_buff *skb, struct xt_action_param *par)
09e410de 515{
f7108a20 516 const struct xt_hashlimit_mtinfo1 *info = par->matchinfo;
09e410de
JE
517 struct xt_hashlimit_htable *hinfo = info->hinfo;
518 unsigned long now = jiffies;
519 struct dsthash_ent *dh;
520 struct dsthash_dst dst;
521
f7108a20 522 if (hashlimit_init_dst(hinfo, &dst, skb, par->thoff) < 0)
09e410de
JE
523 goto hotdrop;
524
02e4eb75 525 rcu_read_lock_bh();
09e410de
JE
526 dh = dsthash_find(hinfo, &dst);
527 if (dh == NULL) {
528 dh = dsthash_alloc_init(hinfo, &dst);
529 if (dh == NULL) {
02e4eb75 530 rcu_read_unlock_bh();
09e410de
JE
531 goto hotdrop;
532 }
09e410de
JE
533 dh->expires = jiffies + msecs_to_jiffies(hinfo->cfg.expire);
534 dh->rateinfo.prev = jiffies;
535 dh->rateinfo.credit = user2credits(hinfo->cfg.avg *
536 hinfo->cfg.burst);
537 dh->rateinfo.credit_cap = user2credits(hinfo->cfg.avg *
538 hinfo->cfg.burst);
539 dh->rateinfo.cost = user2credits(hinfo->cfg.avg);
540 } else {
541 /* update expiration timeout */
542 dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire);
543 rateinfo_recalc(dh, now);
544 }
545
546 if (dh->rateinfo.credit >= dh->rateinfo.cost) {
547 /* below the limit */
548 dh->rateinfo.credit -= dh->rateinfo.cost;
02e4eb75
ED
549 spin_unlock(&dh->lock);
550 rcu_read_unlock_bh();
09e410de
JE
551 return !(info->cfg.mode & XT_HASHLIMIT_INVERT);
552 }
553
02e4eb75
ED
554 spin_unlock(&dh->lock);
555 rcu_read_unlock_bh();
09e410de
JE
556 /* default match is underlimit - so over the limit, we need to invert */
557 return info->cfg.mode & XT_HASHLIMIT_INVERT;
558
559 hotdrop:
b4ba2611 560 par->hotdrop = true;
09e410de
JE
561 return false;
562}
563
b0f38452 564static int hashlimit_mt_check(const struct xt_mtchk_param *par)
09e410de 565{
e89fc3f1 566 struct net *net = par->net;
9b4fce7a 567 struct xt_hashlimit_mtinfo1 *info = par->matchinfo;
4a5a5c73 568 int ret;
09e410de
JE
569
570 /* Check for overflow. */
571 if (info->cfg.burst == 0 ||
572 user2credits(info->cfg.avg * info->cfg.burst) <
573 user2credits(info->cfg.avg)) {
8bee4bad
JE
574 pr_info("overflow, try lower: %u/%u\n",
575 info->cfg.avg, info->cfg.burst);
4a5a5c73 576 return -ERANGE;
09e410de
JE
577 }
578 if (info->cfg.gc_interval == 0 || info->cfg.expire == 0)
bd414ee6 579 return -EINVAL;
09e410de 580 if (info->name[sizeof(info->name)-1] != '\0')
bd414ee6 581 return -EINVAL;
aa5fa318 582 if (par->family == NFPROTO_IPV4) {
09e410de 583 if (info->cfg.srcmask > 32 || info->cfg.dstmask > 32)
bd414ee6 584 return -EINVAL;
09e410de
JE
585 } else {
586 if (info->cfg.srcmask > 128 || info->cfg.dstmask > 128)
bd414ee6 587 return -EINVAL;
09e410de
JE
588 }
589
2eff25c1 590 mutex_lock(&hashlimit_mutex);
aa5fa318 591 info->hinfo = htable_find_get(net, info->name, par->family);
4a5a5c73
JE
592 if (info->hinfo == NULL) {
593 ret = htable_create(net, info, par->family);
594 if (ret < 0) {
595 mutex_unlock(&hashlimit_mutex);
596 return ret;
597 }
09e410de 598 }
2eff25c1 599 mutex_unlock(&hashlimit_mutex);
bd414ee6 600 return 0;
09e410de
JE
601}
602
6be3d859 603static void hashlimit_mt_destroy(const struct xt_mtdtor_param *par)
09e410de 604{
6be3d859 605 const struct xt_hashlimit_mtinfo1 *info = par->matchinfo;
09e410de
JE
606
607 htable_put(info->hinfo);
608}
609
d3c5ee6d 610static struct xt_match hashlimit_mt_reg[] __read_mostly = {
09e410de
JE
611 {
612 .name = "hashlimit",
613 .revision = 1,
ee999d8b 614 .family = NFPROTO_IPV4,
09e410de
JE
615 .match = hashlimit_mt,
616 .matchsize = sizeof(struct xt_hashlimit_mtinfo1),
617 .checkentry = hashlimit_mt_check,
618 .destroy = hashlimit_mt_destroy,
619 .me = THIS_MODULE,
620 },
7b21e09d 621#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
09e410de
JE
622 {
623 .name = "hashlimit",
624 .revision = 1,
ee999d8b 625 .family = NFPROTO_IPV6,
09e410de
JE
626 .match = hashlimit_mt,
627 .matchsize = sizeof(struct xt_hashlimit_mtinfo1),
628 .checkentry = hashlimit_mt_check,
629 .destroy = hashlimit_mt_destroy,
630 .me = THIS_MODULE,
631 },
7b21e09d 632#endif
1da177e4
LT
633};
634
635/* PROC stuff */
1da177e4 636static void *dl_seq_start(struct seq_file *s, loff_t *pos)
f4f6fb71 637 __acquires(htable->lock)
1da177e4 638{
a1004d8e 639 struct xt_hashlimit_htable *htable = s->private;
1da177e4
LT
640 unsigned int *bucket;
641
642 spin_lock_bh(&htable->lock);
643 if (*pos >= htable->cfg.size)
644 return NULL;
645
646 bucket = kmalloc(sizeof(unsigned int), GFP_ATOMIC);
647 if (!bucket)
648 return ERR_PTR(-ENOMEM);
649
650 *bucket = *pos;
651 return bucket;
652}
653
654static void *dl_seq_next(struct seq_file *s, void *v, loff_t *pos)
655{
a1004d8e 656 struct xt_hashlimit_htable *htable = s->private;
1da177e4
LT
657 unsigned int *bucket = (unsigned int *)v;
658
659 *pos = ++(*bucket);
660 if (*pos >= htable->cfg.size) {
661 kfree(v);
662 return NULL;
663 }
664 return bucket;
665}
666
667static void dl_seq_stop(struct seq_file *s, void *v)
f4f6fb71 668 __releases(htable->lock)
1da177e4 669{
a1004d8e 670 struct xt_hashlimit_htable *htable = s->private;
1da177e4
LT
671 unsigned int *bucket = (unsigned int *)v;
672
55e0d7cf
ED
673 if (!IS_ERR(bucket))
674 kfree(bucket);
1da177e4
LT
675 spin_unlock_bh(&htable->lock);
676}
677
76108cea 678static int dl_seq_real_show(struct dsthash_ent *ent, u_int8_t family,
39b46fc6 679 struct seq_file *s)
1da177e4 680{
02e4eb75
ED
681 int res;
682
683 spin_lock(&ent->lock);
1da177e4
LT
684 /* recalculate to show accurate numbers */
685 rateinfo_recalc(ent, jiffies);
686
39b46fc6 687 switch (family) {
ee999d8b 688 case NFPROTO_IPV4:
02e4eb75 689 res = seq_printf(s, "%ld %pI4:%u->%pI4:%u %u %u %u\n",
39b46fc6 690 (long)(ent->expires - jiffies)/HZ,
14d5e834 691 &ent->dst.ip.src,
39b46fc6 692 ntohs(ent->dst.src_port),
14d5e834 693 &ent->dst.ip.dst,
39b46fc6
PM
694 ntohs(ent->dst.dst_port),
695 ent->rateinfo.credit, ent->rateinfo.credit_cap,
696 ent->rateinfo.cost);
02e4eb75 697 break;
7b21e09d 698#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
ee999d8b 699 case NFPROTO_IPV6:
02e4eb75 700 res = seq_printf(s, "%ld %pI6:%u->%pI6:%u %u %u %u\n",
39b46fc6 701 (long)(ent->expires - jiffies)/HZ,
38ff4fa4 702 &ent->dst.ip6.src,
39b46fc6 703 ntohs(ent->dst.src_port),
38ff4fa4 704 &ent->dst.ip6.dst,
39b46fc6
PM
705 ntohs(ent->dst.dst_port),
706 ent->rateinfo.credit, ent->rateinfo.credit_cap,
707 ent->rateinfo.cost);
02e4eb75 708 break;
7b21e09d 709#endif
39b46fc6
PM
710 default:
711 BUG();
02e4eb75 712 res = 0;
39b46fc6 713 }
02e4eb75
ED
714 spin_unlock(&ent->lock);
715 return res;
1da177e4
LT
716}
717
718static int dl_seq_show(struct seq_file *s, void *v)
719{
a1004d8e 720 struct xt_hashlimit_htable *htable = s->private;
1da177e4
LT
721 unsigned int *bucket = (unsigned int *)v;
722 struct dsthash_ent *ent;
723 struct hlist_node *pos;
724
39b46fc6
PM
725 if (!hlist_empty(&htable->hash[*bucket])) {
726 hlist_for_each_entry(ent, pos, &htable->hash[*bucket], node)
727 if (dl_seq_real_show(ent, htable->family, s))
683a04ce 728 return -1;
39b46fc6 729 }
1da177e4
LT
730 return 0;
731}
732
56b3d975 733static const struct seq_operations dl_seq_ops = {
1da177e4
LT
734 .start = dl_seq_start,
735 .next = dl_seq_next,
736 .stop = dl_seq_stop,
737 .show = dl_seq_show
738};
739
740static int dl_proc_open(struct inode *inode, struct file *file)
741{
742 int ret = seq_open(file, &dl_seq_ops);
743
744 if (!ret) {
745 struct seq_file *sf = file->private_data;
a1004d8e 746 sf->private = PDE(inode)->data;
1da177e4
LT
747 }
748 return ret;
749}
750
da7071d7 751static const struct file_operations dl_file_ops = {
1da177e4
LT
752 .owner = THIS_MODULE,
753 .open = dl_proc_open,
754 .read = seq_read,
755 .llseek = seq_lseek,
756 .release = seq_release
757};
758
e89fc3f1
AD
759static int __net_init hashlimit_proc_net_init(struct net *net)
760{
761 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
762
763 hashlimit_net->ipt_hashlimit = proc_mkdir("ipt_hashlimit", net->proc_net);
764 if (!hashlimit_net->ipt_hashlimit)
765 return -ENOMEM;
766#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
767 hashlimit_net->ip6t_hashlimit = proc_mkdir("ip6t_hashlimit", net->proc_net);
768 if (!hashlimit_net->ip6t_hashlimit) {
769 proc_net_remove(net, "ipt_hashlimit");
770 return -ENOMEM;
771 }
772#endif
773 return 0;
774}
775
776static void __net_exit hashlimit_proc_net_exit(struct net *net)
777{
778 proc_net_remove(net, "ipt_hashlimit");
779#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
780 proc_net_remove(net, "ip6t_hashlimit");
781#endif
782}
783
784static int __net_init hashlimit_net_init(struct net *net)
785{
786 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
787
788 INIT_HLIST_HEAD(&hashlimit_net->htables);
789 return hashlimit_proc_net_init(net);
790}
791
792static void __net_exit hashlimit_net_exit(struct net *net)
793{
794 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
795
796 BUG_ON(!hlist_empty(&hashlimit_net->htables));
797 hashlimit_proc_net_exit(net);
798}
799
800static struct pernet_operations hashlimit_net_ops = {
801 .init = hashlimit_net_init,
802 .exit = hashlimit_net_exit,
803 .id = &hashlimit_net_id,
804 .size = sizeof(struct hashlimit_net),
805};
806
d3c5ee6d 807static int __init hashlimit_mt_init(void)
1da177e4 808{
39b46fc6 809 int err;
1da177e4 810
e89fc3f1
AD
811 err = register_pernet_subsys(&hashlimit_net_ops);
812 if (err < 0)
813 return err;
d3c5ee6d
JE
814 err = xt_register_matches(hashlimit_mt_reg,
815 ARRAY_SIZE(hashlimit_mt_reg));
39b46fc6
PM
816 if (err < 0)
817 goto err1;
1da177e4 818
39b46fc6
PM
819 err = -ENOMEM;
820 hashlimit_cachep = kmem_cache_create("xt_hashlimit",
821 sizeof(struct dsthash_ent), 0, 0,
20c2df83 822 NULL);
1da177e4 823 if (!hashlimit_cachep) {
8bee4bad 824 pr_warning("unable to create slab cache\n");
39b46fc6 825 goto err2;
1da177e4 826 }
e89fc3f1
AD
827 return 0;
828
39b46fc6 829err2:
d3c5ee6d 830 xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg));
39b46fc6 831err1:
e89fc3f1 832 unregister_pernet_subsys(&hashlimit_net_ops);
39b46fc6 833 return err;
1da177e4 834
1da177e4
LT
835}
836
d3c5ee6d 837static void __exit hashlimit_mt_exit(void)
1da177e4 838{
d3c5ee6d 839 xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg));
e89fc3f1 840 unregister_pernet_subsys(&hashlimit_net_ops);
02e4eb75
ED
841
842 rcu_barrier_bh();
843 kmem_cache_destroy(hashlimit_cachep);
1da177e4
LT
844}
845
d3c5ee6d
JE
846module_init(hashlimit_mt_init);
847module_exit(hashlimit_mt_exit);