]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - net/netfilter/xt_hashlimit.c
netfilter: Remove unnecessary OOM logging messages
[mirror_ubuntu-artful-kernel.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);
0a9ee813 179 if (ent) {
02e4eb75
ED
180 memcpy(&ent->dst, dst, sizeof(ent->dst));
181 spin_lock_init(&ent->lock);
1da177e4 182
02e4eb75
ED
183 spin_lock(&ent->lock);
184 hlist_add_head_rcu(&ent->node, &ht->hash[hash_dst(ht, dst)]);
185 ht->count++;
186 }
187 spin_unlock(&ht->lock);
1da177e4
LT
188 return ent;
189}
190
02e4eb75
ED
191static void dsthash_free_rcu(struct rcu_head *head)
192{
193 struct dsthash_ent *ent = container_of(head, struct dsthash_ent, rcu);
194
195 kmem_cache_free(hashlimit_cachep, ent);
196}
197
39b46fc6
PM
198static inline void
199dsthash_free(struct xt_hashlimit_htable *ht, struct dsthash_ent *ent)
1da177e4 200{
02e4eb75
ED
201 hlist_del_rcu(&ent->node);
202 call_rcu_bh(&ent->rcu, dsthash_free_rcu);
39b46fc6 203 ht->count--;
1da177e4
LT
204}
205static void htable_gc(unsigned long htlong);
206
e89fc3f1
AD
207static int htable_create(struct net *net, struct xt_hashlimit_mtinfo1 *minfo,
208 u_int8_t family)
09e410de 209{
e89fc3f1 210 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
09e410de
JE
211 struct xt_hashlimit_htable *hinfo;
212 unsigned int size;
213 unsigned int i;
214
215 if (minfo->cfg.size) {
216 size = minfo->cfg.size;
217 } else {
4481374c 218 size = (totalram_pages << PAGE_SHIFT) / 16384 /
09e410de 219 sizeof(struct list_head);
4481374c 220 if (totalram_pages > 1024 * 1024 * 1024 / PAGE_SIZE)
09e410de
JE
221 size = 8192;
222 if (size < 16)
223 size = 16;
224 }
225 /* FIXME: don't use vmalloc() here or anywhere else -HW */
226 hinfo = vmalloc(sizeof(struct xt_hashlimit_htable) +
227 sizeof(struct list_head) * size);
85bc3f38 228 if (hinfo == NULL)
4a5a5c73 229 return -ENOMEM;
09e410de
JE
230 minfo->hinfo = hinfo;
231
232 /* copy match config into hashtable config */
233 memcpy(&hinfo->cfg, &minfo->cfg, sizeof(hinfo->cfg));
234 hinfo->cfg.size = size;
235 if (hinfo->cfg.max == 0)
236 hinfo->cfg.max = 8 * hinfo->cfg.size;
237 else if (hinfo->cfg.max < hinfo->cfg.size)
238 hinfo->cfg.max = hinfo->cfg.size;
239
240 for (i = 0; i < hinfo->cfg.size; i++)
241 INIT_HLIST_HEAD(&hinfo->hash[i]);
242
2eff25c1 243 hinfo->use = 1;
09e410de
JE
244 hinfo->count = 0;
245 hinfo->family = family;
89bc7a0f 246 hinfo->rnd_initialized = false;
09e410de
JE
247 spin_lock_init(&hinfo->lock);
248
ee999d8b
JE
249 hinfo->pde = proc_create_data(minfo->name, 0,
250 (family == NFPROTO_IPV4) ?
e89fc3f1 251 hashlimit_net->ipt_hashlimit : hashlimit_net->ip6t_hashlimit,
ee999d8b 252 &dl_file_ops, hinfo);
09e410de
JE
253 if (hinfo->pde == NULL) {
254 vfree(hinfo);
4a5a5c73 255 return -ENOMEM;
09e410de 256 }
e89fc3f1 257 hinfo->net = net;
09e410de
JE
258
259 setup_timer(&hinfo->timer, htable_gc, (unsigned long)hinfo);
260 hinfo->timer.expires = jiffies + msecs_to_jiffies(hinfo->cfg.gc_interval);
261 add_timer(&hinfo->timer);
262
e89fc3f1 263 hlist_add_head(&hinfo->node, &hashlimit_net->htables);
09e410de
JE
264
265 return 0;
266}
267
a47362a2
JE
268static bool select_all(const struct xt_hashlimit_htable *ht,
269 const struct dsthash_ent *he)
1da177e4
LT
270{
271 return 1;
272}
273
a47362a2
JE
274static bool select_gc(const struct xt_hashlimit_htable *ht,
275 const struct dsthash_ent *he)
1da177e4 276{
cbebc51f 277 return time_after_eq(jiffies, he->expires);
1da177e4
LT
278}
279
39b46fc6 280static void htable_selective_cleanup(struct xt_hashlimit_htable *ht,
a47362a2
JE
281 bool (*select)(const struct xt_hashlimit_htable *ht,
282 const struct dsthash_ent *he))
1da177e4 283{
39b46fc6 284 unsigned int i;
1da177e4
LT
285
286 /* lock hash table and iterate over it */
287 spin_lock_bh(&ht->lock);
288 for (i = 0; i < ht->cfg.size; i++) {
289 struct dsthash_ent *dh;
290 struct hlist_node *pos, *n;
291 hlist_for_each_entry_safe(dh, pos, n, &ht->hash[i], node) {
292 if ((*select)(ht, dh))
39b46fc6 293 dsthash_free(ht, dh);
1da177e4
LT
294 }
295 }
296 spin_unlock_bh(&ht->lock);
297}
298
299/* hash table garbage collector, run by timer */
300static void htable_gc(unsigned long htlong)
301{
39b46fc6 302 struct xt_hashlimit_htable *ht = (struct xt_hashlimit_htable *)htlong;
1da177e4
LT
303
304 htable_selective_cleanup(ht, select_gc);
305
306 /* re-add the timer accordingly */
307 ht->timer.expires = jiffies + msecs_to_jiffies(ht->cfg.gc_interval);
308 add_timer(&ht->timer);
309}
310
39b46fc6 311static void htable_destroy(struct xt_hashlimit_htable *hinfo)
1da177e4 312{
e89fc3f1
AD
313 struct hashlimit_net *hashlimit_net = hashlimit_pernet(hinfo->net);
314 struct proc_dir_entry *parent;
315
967ab999 316 del_timer_sync(&hinfo->timer);
1da177e4 317
e89fc3f1
AD
318 if (hinfo->family == NFPROTO_IPV4)
319 parent = hashlimit_net->ipt_hashlimit;
320 else
321 parent = hashlimit_net->ip6t_hashlimit;
322 remove_proc_entry(hinfo->pde->name, parent);
1da177e4
LT
323 htable_selective_cleanup(hinfo, select_all);
324 vfree(hinfo);
325}
326
e89fc3f1
AD
327static struct xt_hashlimit_htable *htable_find_get(struct net *net,
328 const char *name,
76108cea 329 u_int8_t family)
1da177e4 330{
e89fc3f1 331 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
39b46fc6 332 struct xt_hashlimit_htable *hinfo;
1da177e4
LT
333 struct hlist_node *pos;
334
e89fc3f1 335 hlist_for_each_entry(hinfo, pos, &hashlimit_net->htables, node) {
39b46fc6
PM
336 if (!strcmp(name, hinfo->pde->name) &&
337 hinfo->family == family) {
2eff25c1 338 hinfo->use++;
1da177e4
LT
339 return hinfo;
340 }
341 }
1da177e4
LT
342 return NULL;
343}
344
39b46fc6 345static void htable_put(struct xt_hashlimit_htable *hinfo)
1da177e4 346{
2eff25c1
PM
347 mutex_lock(&hashlimit_mutex);
348 if (--hinfo->use == 0) {
1da177e4 349 hlist_del(&hinfo->node);
1da177e4
LT
350 htable_destroy(hinfo);
351 }
2eff25c1 352 mutex_unlock(&hashlimit_mutex);
1da177e4
LT
353}
354
1da177e4
LT
355/* The algorithm used is the Simple Token Bucket Filter (TBF)
356 * see net/sched/sch_tbf.c in the linux source tree
357 */
358
359/* Rusty: This is my (non-mathematically-inclined) understanding of
360 this algorithm. The `average rate' in jiffies becomes your initial
361 amount of credit `credit' and the most credit you can ever have
362 `credit_cap'. The `peak rate' becomes the cost of passing the
363 test, `cost'.
364
365 `prev' tracks the last packet hit: you gain one credit per jiffy.
366 If you get credit balance more than this, the extra credit is
367 discarded. Every time the match passes, you lose `cost' credits;
368 if you don't have that many, the test fails.
369
370 See Alexey's formal explanation in net/sched/sch_tbf.c.
371
372 To get the maximum range, we multiply by this factor (ie. you get N
373 credits per jiffy). We want to allow a rate as low as 1 per day
374 (slowest userspace tool allows), which means
375 CREDITS_PER_JIFFY*HZ*60*60*24 < 2^32 ie.
376*/
377#define MAX_CPJ (0xFFFFFFFF / (HZ*60*60*24))
378
379/* Repeated shift and or gives us all 1s, final shift and add 1 gives
380 * us the power of 2 below the theoretical max, so GCC simply does a
381 * shift. */
382#define _POW2_BELOW2(x) ((x)|((x)>>1))
383#define _POW2_BELOW4(x) (_POW2_BELOW2(x)|_POW2_BELOW2((x)>>2))
384#define _POW2_BELOW8(x) (_POW2_BELOW4(x)|_POW2_BELOW4((x)>>4))
385#define _POW2_BELOW16(x) (_POW2_BELOW8(x)|_POW2_BELOW8((x)>>8))
386#define _POW2_BELOW32(x) (_POW2_BELOW16(x)|_POW2_BELOW16((x)>>16))
387#define POW2_BELOW32(x) ((_POW2_BELOW32(x)>>1) + 1)
388
389#define CREDITS_PER_JIFFY POW2_BELOW32(MAX_CPJ)
390
391/* Precision saver. */
392static inline u_int32_t
393user2credits(u_int32_t user)
394{
395 /* If multiplying would overflow... */
396 if (user > 0xFFFFFFFF / (HZ*CREDITS_PER_JIFFY))
397 /* Divide first. */
39b46fc6 398 return (user / XT_HASHLIMIT_SCALE) * HZ * CREDITS_PER_JIFFY;
1da177e4 399
39b46fc6 400 return (user * HZ * CREDITS_PER_JIFFY) / XT_HASHLIMIT_SCALE;
1da177e4
LT
401}
402
403static inline void rateinfo_recalc(struct dsthash_ent *dh, unsigned long now)
404{
39b46fc6 405 dh->rateinfo.credit += (now - dh->rateinfo.prev) * CREDITS_PER_JIFFY;
1da177e4
LT
406 if (dh->rateinfo.credit > dh->rateinfo.credit_cap)
407 dh->rateinfo.credit = dh->rateinfo.credit_cap;
39b46fc6
PM
408 dh->rateinfo.prev = now;
409}
410
09e410de
JE
411static inline __be32 maskl(__be32 a, unsigned int l)
412{
1b9b70ea 413 return l ? htonl(ntohl(a) & ~0 << (32 - l)) : 0;
09e410de
JE
414}
415
3ed5df44 416#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
09e410de
JE
417static void hashlimit_ipv6_mask(__be32 *i, unsigned int p)
418{
419 switch (p) {
1b9b70ea 420 case 0 ... 31:
09e410de
JE
421 i[0] = maskl(i[0], p);
422 i[1] = i[2] = i[3] = 0;
423 break;
1b9b70ea 424 case 32 ... 63:
09e410de
JE
425 i[1] = maskl(i[1], p - 32);
426 i[2] = i[3] = 0;
427 break;
1b9b70ea 428 case 64 ... 95:
09e410de
JE
429 i[2] = maskl(i[2], p - 64);
430 i[3] = 0;
8f599229 431 break;
1b9b70ea 432 case 96 ... 127:
09e410de
JE
433 i[3] = maskl(i[3], p - 96);
434 break;
435 case 128:
436 break;
437 }
438}
3ed5df44 439#endif
09e410de 440
39b46fc6 441static int
a47362a2
JE
442hashlimit_init_dst(const struct xt_hashlimit_htable *hinfo,
443 struct dsthash_dst *dst,
39b46fc6
PM
444 const struct sk_buff *skb, unsigned int protoff)
445{
446 __be16 _ports[2], *ports;
193b23c5 447 u8 nexthdr;
aca071c1 448 int poff;
39b46fc6
PM
449
450 memset(dst, 0, sizeof(*dst));
451
452 switch (hinfo->family) {
ee999d8b 453 case NFPROTO_IPV4:
39b46fc6 454 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP)
09e410de
JE
455 dst->ip.dst = maskl(ip_hdr(skb)->daddr,
456 hinfo->cfg.dstmask);
39b46fc6 457 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP)
09e410de
JE
458 dst->ip.src = maskl(ip_hdr(skb)->saddr,
459 hinfo->cfg.srcmask);
39b46fc6
PM
460
461 if (!(hinfo->cfg.mode &
462 (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT)))
463 return 0;
eddc9ec5 464 nexthdr = ip_hdr(skb)->protocol;
39b46fc6 465 break;
02dba025 466#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
ee999d8b 467 case NFPROTO_IPV6:
09e410de
JE
468 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) {
469 memcpy(&dst->ip6.dst, &ipv6_hdr(skb)->daddr,
470 sizeof(dst->ip6.dst));
471 hashlimit_ipv6_mask(dst->ip6.dst, hinfo->cfg.dstmask);
472 }
473 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) {
474 memcpy(&dst->ip6.src, &ipv6_hdr(skb)->saddr,
475 sizeof(dst->ip6.src));
476 hashlimit_ipv6_mask(dst->ip6.src, hinfo->cfg.srcmask);
477 }
39b46fc6
PM
478
479 if (!(hinfo->cfg.mode &
480 (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT)))
481 return 0;
193b23c5
PM
482 nexthdr = ipv6_hdr(skb)->nexthdr;
483 protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr);
484 if ((int)protoff < 0)
39b46fc6
PM
485 return -1;
486 break;
487#endif
488 default:
489 BUG();
490 return 0;
491 }
492
aca071c1
CG
493 poff = proto_ports_offset(nexthdr);
494 if (poff >= 0) {
495 ports = skb_header_pointer(skb, protoff + poff, sizeof(_ports),
39b46fc6 496 &_ports);
aca071c1 497 } else {
39b46fc6
PM
498 _ports[0] = _ports[1] = 0;
499 ports = _ports;
39b46fc6
PM
500 }
501 if (!ports)
502 return -1;
503 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SPT)
504 dst->src_port = ports[0];
505 if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DPT)
506 dst->dst_port = ports[1];
507 return 0;
1da177e4
LT
508}
509
ccb79bdc 510static bool
62fc8051 511hashlimit_mt(const struct sk_buff *skb, struct xt_action_param *par)
09e410de 512{
f7108a20 513 const struct xt_hashlimit_mtinfo1 *info = par->matchinfo;
09e410de
JE
514 struct xt_hashlimit_htable *hinfo = info->hinfo;
515 unsigned long now = jiffies;
516 struct dsthash_ent *dh;
517 struct dsthash_dst dst;
518
f7108a20 519 if (hashlimit_init_dst(hinfo, &dst, skb, par->thoff) < 0)
09e410de
JE
520 goto hotdrop;
521
02e4eb75 522 rcu_read_lock_bh();
09e410de
JE
523 dh = dsthash_find(hinfo, &dst);
524 if (dh == NULL) {
525 dh = dsthash_alloc_init(hinfo, &dst);
526 if (dh == NULL) {
02e4eb75 527 rcu_read_unlock_bh();
09e410de
JE
528 goto hotdrop;
529 }
09e410de
JE
530 dh->expires = jiffies + msecs_to_jiffies(hinfo->cfg.expire);
531 dh->rateinfo.prev = jiffies;
532 dh->rateinfo.credit = user2credits(hinfo->cfg.avg *
533 hinfo->cfg.burst);
534 dh->rateinfo.credit_cap = user2credits(hinfo->cfg.avg *
535 hinfo->cfg.burst);
536 dh->rateinfo.cost = user2credits(hinfo->cfg.avg);
537 } else {
538 /* update expiration timeout */
539 dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire);
540 rateinfo_recalc(dh, now);
541 }
542
543 if (dh->rateinfo.credit >= dh->rateinfo.cost) {
544 /* below the limit */
545 dh->rateinfo.credit -= dh->rateinfo.cost;
02e4eb75
ED
546 spin_unlock(&dh->lock);
547 rcu_read_unlock_bh();
09e410de
JE
548 return !(info->cfg.mode & XT_HASHLIMIT_INVERT);
549 }
550
02e4eb75
ED
551 spin_unlock(&dh->lock);
552 rcu_read_unlock_bh();
09e410de
JE
553 /* default match is underlimit - so over the limit, we need to invert */
554 return info->cfg.mode & XT_HASHLIMIT_INVERT;
555
556 hotdrop:
b4ba2611 557 par->hotdrop = true;
09e410de
JE
558 return false;
559}
560
b0f38452 561static int hashlimit_mt_check(const struct xt_mtchk_param *par)
09e410de 562{
e89fc3f1 563 struct net *net = par->net;
9b4fce7a 564 struct xt_hashlimit_mtinfo1 *info = par->matchinfo;
4a5a5c73 565 int ret;
09e410de
JE
566
567 /* Check for overflow. */
568 if (info->cfg.burst == 0 ||
569 user2credits(info->cfg.avg * info->cfg.burst) <
570 user2credits(info->cfg.avg)) {
8bee4bad
JE
571 pr_info("overflow, try lower: %u/%u\n",
572 info->cfg.avg, info->cfg.burst);
4a5a5c73 573 return -ERANGE;
09e410de
JE
574 }
575 if (info->cfg.gc_interval == 0 || info->cfg.expire == 0)
bd414ee6 576 return -EINVAL;
09e410de 577 if (info->name[sizeof(info->name)-1] != '\0')
bd414ee6 578 return -EINVAL;
aa5fa318 579 if (par->family == NFPROTO_IPV4) {
09e410de 580 if (info->cfg.srcmask > 32 || info->cfg.dstmask > 32)
bd414ee6 581 return -EINVAL;
09e410de
JE
582 } else {
583 if (info->cfg.srcmask > 128 || info->cfg.dstmask > 128)
bd414ee6 584 return -EINVAL;
09e410de
JE
585 }
586
2eff25c1 587 mutex_lock(&hashlimit_mutex);
aa5fa318 588 info->hinfo = htable_find_get(net, info->name, par->family);
4a5a5c73
JE
589 if (info->hinfo == NULL) {
590 ret = htable_create(net, info, par->family);
591 if (ret < 0) {
592 mutex_unlock(&hashlimit_mutex);
593 return ret;
594 }
09e410de 595 }
2eff25c1 596 mutex_unlock(&hashlimit_mutex);
bd414ee6 597 return 0;
09e410de
JE
598}
599
6be3d859 600static void hashlimit_mt_destroy(const struct xt_mtdtor_param *par)
09e410de 601{
6be3d859 602 const struct xt_hashlimit_mtinfo1 *info = par->matchinfo;
09e410de
JE
603
604 htable_put(info->hinfo);
605}
606
d3c5ee6d 607static struct xt_match hashlimit_mt_reg[] __read_mostly = {
09e410de
JE
608 {
609 .name = "hashlimit",
610 .revision = 1,
ee999d8b 611 .family = NFPROTO_IPV4,
09e410de
JE
612 .match = hashlimit_mt,
613 .matchsize = sizeof(struct xt_hashlimit_mtinfo1),
614 .checkentry = hashlimit_mt_check,
615 .destroy = hashlimit_mt_destroy,
616 .me = THIS_MODULE,
617 },
7b21e09d 618#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
09e410de
JE
619 {
620 .name = "hashlimit",
621 .revision = 1,
ee999d8b 622 .family = NFPROTO_IPV6,
09e410de
JE
623 .match = hashlimit_mt,
624 .matchsize = sizeof(struct xt_hashlimit_mtinfo1),
625 .checkentry = hashlimit_mt_check,
626 .destroy = hashlimit_mt_destroy,
627 .me = THIS_MODULE,
628 },
7b21e09d 629#endif
1da177e4
LT
630};
631
632/* PROC stuff */
1da177e4 633static void *dl_seq_start(struct seq_file *s, loff_t *pos)
f4f6fb71 634 __acquires(htable->lock)
1da177e4 635{
a1004d8e 636 struct xt_hashlimit_htable *htable = s->private;
1da177e4
LT
637 unsigned int *bucket;
638
639 spin_lock_bh(&htable->lock);
640 if (*pos >= htable->cfg.size)
641 return NULL;
642
643 bucket = kmalloc(sizeof(unsigned int), GFP_ATOMIC);
644 if (!bucket)
645 return ERR_PTR(-ENOMEM);
646
647 *bucket = *pos;
648 return bucket;
649}
650
651static void *dl_seq_next(struct seq_file *s, void *v, loff_t *pos)
652{
a1004d8e 653 struct xt_hashlimit_htable *htable = s->private;
1da177e4
LT
654 unsigned int *bucket = (unsigned int *)v;
655
656 *pos = ++(*bucket);
657 if (*pos >= htable->cfg.size) {
658 kfree(v);
659 return NULL;
660 }
661 return bucket;
662}
663
664static void dl_seq_stop(struct seq_file *s, void *v)
f4f6fb71 665 __releases(htable->lock)
1da177e4 666{
a1004d8e 667 struct xt_hashlimit_htable *htable = s->private;
1da177e4
LT
668 unsigned int *bucket = (unsigned int *)v;
669
55e0d7cf
ED
670 if (!IS_ERR(bucket))
671 kfree(bucket);
1da177e4
LT
672 spin_unlock_bh(&htable->lock);
673}
674
76108cea 675static int dl_seq_real_show(struct dsthash_ent *ent, u_int8_t family,
39b46fc6 676 struct seq_file *s)
1da177e4 677{
02e4eb75
ED
678 int res;
679
680 spin_lock(&ent->lock);
1da177e4
LT
681 /* recalculate to show accurate numbers */
682 rateinfo_recalc(ent, jiffies);
683
39b46fc6 684 switch (family) {
ee999d8b 685 case NFPROTO_IPV4:
02e4eb75 686 res = seq_printf(s, "%ld %pI4:%u->%pI4:%u %u %u %u\n",
39b46fc6 687 (long)(ent->expires - jiffies)/HZ,
14d5e834 688 &ent->dst.ip.src,
39b46fc6 689 ntohs(ent->dst.src_port),
14d5e834 690 &ent->dst.ip.dst,
39b46fc6
PM
691 ntohs(ent->dst.dst_port),
692 ent->rateinfo.credit, ent->rateinfo.credit_cap,
693 ent->rateinfo.cost);
02e4eb75 694 break;
7b21e09d 695#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
ee999d8b 696 case NFPROTO_IPV6:
02e4eb75 697 res = seq_printf(s, "%ld %pI6:%u->%pI6:%u %u %u %u\n",
39b46fc6 698 (long)(ent->expires - jiffies)/HZ,
38ff4fa4 699 &ent->dst.ip6.src,
39b46fc6 700 ntohs(ent->dst.src_port),
38ff4fa4 701 &ent->dst.ip6.dst,
39b46fc6
PM
702 ntohs(ent->dst.dst_port),
703 ent->rateinfo.credit, ent->rateinfo.credit_cap,
704 ent->rateinfo.cost);
02e4eb75 705 break;
7b21e09d 706#endif
39b46fc6
PM
707 default:
708 BUG();
02e4eb75 709 res = 0;
39b46fc6 710 }
02e4eb75
ED
711 spin_unlock(&ent->lock);
712 return res;
1da177e4
LT
713}
714
715static int dl_seq_show(struct seq_file *s, void *v)
716{
a1004d8e 717 struct xt_hashlimit_htable *htable = s->private;
1da177e4
LT
718 unsigned int *bucket = (unsigned int *)v;
719 struct dsthash_ent *ent;
720 struct hlist_node *pos;
721
39b46fc6
PM
722 if (!hlist_empty(&htable->hash[*bucket])) {
723 hlist_for_each_entry(ent, pos, &htable->hash[*bucket], node)
724 if (dl_seq_real_show(ent, htable->family, s))
683a04ce 725 return -1;
39b46fc6 726 }
1da177e4
LT
727 return 0;
728}
729
56b3d975 730static const struct seq_operations dl_seq_ops = {
1da177e4
LT
731 .start = dl_seq_start,
732 .next = dl_seq_next,
733 .stop = dl_seq_stop,
734 .show = dl_seq_show
735};
736
737static int dl_proc_open(struct inode *inode, struct file *file)
738{
739 int ret = seq_open(file, &dl_seq_ops);
740
741 if (!ret) {
742 struct seq_file *sf = file->private_data;
a1004d8e 743 sf->private = PDE(inode)->data;
1da177e4
LT
744 }
745 return ret;
746}
747
da7071d7 748static const struct file_operations dl_file_ops = {
1da177e4
LT
749 .owner = THIS_MODULE,
750 .open = dl_proc_open,
751 .read = seq_read,
752 .llseek = seq_lseek,
753 .release = seq_release
754};
755
e89fc3f1
AD
756static int __net_init hashlimit_proc_net_init(struct net *net)
757{
758 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
759
760 hashlimit_net->ipt_hashlimit = proc_mkdir("ipt_hashlimit", net->proc_net);
761 if (!hashlimit_net->ipt_hashlimit)
762 return -ENOMEM;
763#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
764 hashlimit_net->ip6t_hashlimit = proc_mkdir("ip6t_hashlimit", net->proc_net);
765 if (!hashlimit_net->ip6t_hashlimit) {
766 proc_net_remove(net, "ipt_hashlimit");
767 return -ENOMEM;
768 }
769#endif
770 return 0;
771}
772
773static void __net_exit hashlimit_proc_net_exit(struct net *net)
774{
775 proc_net_remove(net, "ipt_hashlimit");
776#if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE)
777 proc_net_remove(net, "ip6t_hashlimit");
778#endif
779}
780
781static int __net_init hashlimit_net_init(struct net *net)
782{
783 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
784
785 INIT_HLIST_HEAD(&hashlimit_net->htables);
786 return hashlimit_proc_net_init(net);
787}
788
789static void __net_exit hashlimit_net_exit(struct net *net)
790{
791 struct hashlimit_net *hashlimit_net = hashlimit_pernet(net);
792
793 BUG_ON(!hlist_empty(&hashlimit_net->htables));
794 hashlimit_proc_net_exit(net);
795}
796
797static struct pernet_operations hashlimit_net_ops = {
798 .init = hashlimit_net_init,
799 .exit = hashlimit_net_exit,
800 .id = &hashlimit_net_id,
801 .size = sizeof(struct hashlimit_net),
802};
803
d3c5ee6d 804static int __init hashlimit_mt_init(void)
1da177e4 805{
39b46fc6 806 int err;
1da177e4 807
e89fc3f1
AD
808 err = register_pernet_subsys(&hashlimit_net_ops);
809 if (err < 0)
810 return err;
d3c5ee6d
JE
811 err = xt_register_matches(hashlimit_mt_reg,
812 ARRAY_SIZE(hashlimit_mt_reg));
39b46fc6
PM
813 if (err < 0)
814 goto err1;
1da177e4 815
39b46fc6
PM
816 err = -ENOMEM;
817 hashlimit_cachep = kmem_cache_create("xt_hashlimit",
818 sizeof(struct dsthash_ent), 0, 0,
20c2df83 819 NULL);
1da177e4 820 if (!hashlimit_cachep) {
8bee4bad 821 pr_warning("unable to create slab cache\n");
39b46fc6 822 goto err2;
1da177e4 823 }
e89fc3f1
AD
824 return 0;
825
39b46fc6 826err2:
d3c5ee6d 827 xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg));
39b46fc6 828err1:
e89fc3f1 829 unregister_pernet_subsys(&hashlimit_net_ops);
39b46fc6 830 return err;
1da177e4 831
1da177e4
LT
832}
833
d3c5ee6d 834static void __exit hashlimit_mt_exit(void)
1da177e4 835{
d3c5ee6d 836 xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg));
e89fc3f1 837 unregister_pernet_subsys(&hashlimit_net_ops);
02e4eb75
ED
838
839 rcu_barrier_bh();
840 kmem_cache_destroy(hashlimit_cachep);
1da177e4
LT
841}
842
d3c5ee6d
JE
843module_init(hashlimit_mt_init);
844module_exit(hashlimit_mt_exit);