]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* iptables match extension to limit the number of packets per second |
2 | * seperately for each hashbucket (sourceip/sourceport/dstip/dstport) | |
3 | * | |
4 | * (C) 2003-2004 by Harald Welte <laforge@netfilter.org> | |
5 | * | |
6 | * $Id: ipt_hashlimit.c 3244 2004-10-20 16:24:29Z laforge@netfilter.org $ | |
7 | * | |
8 | * Development of this code was funded by Astaro AG, http://www.astaro.com/ | |
1da177e4 LT |
9 | */ |
10 | #include <linux/module.h> | |
1da177e4 LT |
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> | |
1da177e4 LT |
16 | #include <linux/proc_fs.h> |
17 | #include <linux/seq_file.h> | |
18 | #include <linux/list.h> | |
39b46fc6 PM |
19 | #include <linux/skbuff.h> |
20 | #include <linux/in.h> | |
21 | #include <linux/ip.h> | |
22 | #include <linux/ipv6.h> | |
1da177e4 | 23 | |
39b46fc6 | 24 | #include <linux/netfilter/x_tables.h> |
1da177e4 | 25 | #include <linux/netfilter_ipv4/ip_tables.h> |
39b46fc6 PM |
26 | #include <linux/netfilter_ipv6/ip6_tables.h> |
27 | #include <linux/netfilter/xt_hashlimit.h> | |
14cc3e2b | 28 | #include <linux/mutex.h> |
1da177e4 LT |
29 | |
30 | MODULE_LICENSE("GPL"); | |
31 | MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); | |
32 | MODULE_DESCRIPTION("iptables match for limiting per hash-bucket"); | |
39b46fc6 PM |
33 | MODULE_ALIAS("ipt_hashlimit"); |
34 | MODULE_ALIAS("ip6t_hashlimit"); | |
1da177e4 LT |
35 | |
36 | /* need to declare this at the top */ | |
39b46fc6 PM |
37 | static struct proc_dir_entry *hashlimit_procdir4; |
38 | static struct proc_dir_entry *hashlimit_procdir6; | |
1da177e4 LT |
39 | static struct file_operations dl_file_ops; |
40 | ||
41 | /* hash table crap */ | |
1da177e4 | 42 | struct dsthash_dst { |
39b46fc6 PM |
43 | union { |
44 | struct { | |
45 | __be32 src; | |
46 | __be32 dst; | |
47 | } ip; | |
48 | struct { | |
49 | __be32 src[4]; | |
50 | __be32 dst[4]; | |
51 | } ip6; | |
52 | } addr; | |
6a19d614 AV |
53 | __be16 src_port; |
54 | __be16 dst_port; | |
1da177e4 LT |
55 | }; |
56 | ||
57 | struct dsthash_ent { | |
58 | /* static / read-only parts in the beginning */ | |
59 | struct hlist_node node; | |
60 | struct dsthash_dst dst; | |
61 | ||
62 | /* modified structure members in the end */ | |
63 | unsigned long expires; /* precalculated expiry time */ | |
64 | struct { | |
65 | unsigned long prev; /* last modification */ | |
66 | u_int32_t credit; | |
67 | u_int32_t credit_cap, cost; | |
68 | } rateinfo; | |
69 | }; | |
70 | ||
39b46fc6 | 71 | struct xt_hashlimit_htable { |
1da177e4 LT |
72 | struct hlist_node node; /* global list of all htables */ |
73 | atomic_t use; | |
39b46fc6 | 74 | int family; |
1da177e4 LT |
75 | |
76 | struct hashlimit_cfg cfg; /* config */ | |
77 | ||
78 | /* used internally */ | |
79 | spinlock_t lock; /* lock for list_head */ | |
80 | u_int32_t rnd; /* random seed for hash */ | |
bf0857ea | 81 | int rnd_initialized; |
39b46fc6 | 82 | unsigned int count; /* number entries in table */ |
1da177e4 | 83 | struct timer_list timer; /* timer for gc */ |
1da177e4 LT |
84 | |
85 | /* seq_file stuff */ | |
86 | struct proc_dir_entry *pde; | |
87 | ||
88 | struct hlist_head hash[0]; /* hashtable itself */ | |
89 | }; | |
90 | ||
e45b1be8 | 91 | static DEFINE_SPINLOCK(hashlimit_lock); /* protects htables list */ |
14cc3e2b | 92 | static DEFINE_MUTEX(hlimit_mutex); /* additional checkentry protection */ |
1da177e4 | 93 | static HLIST_HEAD(hashlimit_htables); |
ba89966c | 94 | static kmem_cache_t *hashlimit_cachep __read_mostly; |
1da177e4 LT |
95 | |
96 | static inline int dst_cmp(const struct dsthash_ent *ent, struct dsthash_dst *b) | |
97 | { | |
39b46fc6 | 98 | return !memcmp(&ent->dst, b, sizeof(ent->dst)); |
1da177e4 LT |
99 | } |
100 | ||
39b46fc6 PM |
101 | static u_int32_t |
102 | hash_dst(const struct xt_hashlimit_htable *ht, const struct dsthash_dst *dst) | |
1da177e4 | 103 | { |
39b46fc6 | 104 | return jhash(dst, sizeof(*dst), ht->rnd) % ht->cfg.size; |
1da177e4 LT |
105 | } |
106 | ||
39b46fc6 PM |
107 | static struct dsthash_ent * |
108 | dsthash_find(const struct xt_hashlimit_htable *ht, struct dsthash_dst *dst) | |
1da177e4 LT |
109 | { |
110 | struct dsthash_ent *ent; | |
111 | struct hlist_node *pos; | |
112 | u_int32_t hash = hash_dst(ht, dst); | |
113 | ||
39b46fc6 PM |
114 | if (!hlist_empty(&ht->hash[hash])) { |
115 | hlist_for_each_entry(ent, pos, &ht->hash[hash], node) | |
116 | if (dst_cmp(ent, dst)) | |
1da177e4 | 117 | return ent; |
39b46fc6 | 118 | } |
1da177e4 LT |
119 | return NULL; |
120 | } | |
121 | ||
122 | /* allocate dsthash_ent, initialize dst, put in htable and lock it */ | |
123 | static struct dsthash_ent * | |
39b46fc6 | 124 | dsthash_alloc_init(struct xt_hashlimit_htable *ht, struct dsthash_dst *dst) |
1da177e4 LT |
125 | { |
126 | struct dsthash_ent *ent; | |
127 | ||
128 | /* initialize hash with random val at the time we allocate | |
129 | * the first hashtable entry */ | |
bf0857ea | 130 | if (!ht->rnd_initialized) { |
1da177e4 | 131 | get_random_bytes(&ht->rnd, 4); |
bf0857ea PM |
132 | ht->rnd_initialized = 1; |
133 | } | |
1da177e4 | 134 | |
39b46fc6 | 135 | if (ht->cfg.max && ht->count >= ht->cfg.max) { |
1da177e4 LT |
136 | /* FIXME: do something. question is what.. */ |
137 | if (net_ratelimit()) | |
39b46fc6 PM |
138 | printk(KERN_WARNING |
139 | "xt_hashlimit: max count of %u reached\n", | |
1da177e4 LT |
140 | ht->cfg.max); |
141 | return NULL; | |
142 | } | |
143 | ||
144 | ent = kmem_cache_alloc(hashlimit_cachep, GFP_ATOMIC); | |
145 | if (!ent) { | |
146 | if (net_ratelimit()) | |
39b46fc6 PM |
147 | printk(KERN_ERR |
148 | "xt_hashlimit: can't allocate dsthash_ent\n"); | |
1da177e4 LT |
149 | return NULL; |
150 | } | |
39b46fc6 | 151 | memcpy(&ent->dst, dst, sizeof(ent->dst)); |
1da177e4 LT |
152 | |
153 | hlist_add_head(&ent->node, &ht->hash[hash_dst(ht, dst)]); | |
39b46fc6 | 154 | ht->count++; |
1da177e4 LT |
155 | return ent; |
156 | } | |
157 | ||
39b46fc6 PM |
158 | static inline void |
159 | dsthash_free(struct xt_hashlimit_htable *ht, struct dsthash_ent *ent) | |
1da177e4 LT |
160 | { |
161 | hlist_del(&ent->node); | |
162 | kmem_cache_free(hashlimit_cachep, ent); | |
39b46fc6 | 163 | ht->count--; |
1da177e4 LT |
164 | } |
165 | static void htable_gc(unsigned long htlong); | |
166 | ||
39b46fc6 | 167 | static int htable_create(struct xt_hashlimit_info *minfo, int family) |
1da177e4 | 168 | { |
39b46fc6 | 169 | struct xt_hashlimit_htable *hinfo; |
1da177e4 | 170 | unsigned int size; |
39b46fc6 | 171 | unsigned int i; |
1da177e4 LT |
172 | |
173 | if (minfo->cfg.size) | |
174 | size = minfo->cfg.size; | |
175 | else { | |
39b46fc6 PM |
176 | size = ((num_physpages << PAGE_SHIFT) / 16384) / |
177 | sizeof(struct list_head); | |
1da177e4 LT |
178 | if (num_physpages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
179 | size = 8192; | |
180 | if (size < 16) | |
181 | size = 16; | |
182 | } | |
183 | /* FIXME: don't use vmalloc() here or anywhere else -HW */ | |
39b46fc6 PM |
184 | hinfo = vmalloc(sizeof(struct xt_hashlimit_htable) + |
185 | sizeof(struct list_head) * size); | |
1da177e4 | 186 | if (!hinfo) { |
39b46fc6 | 187 | printk(KERN_ERR "xt_hashlimit: unable to create hashtable\n"); |
1da177e4 LT |
188 | return -1; |
189 | } | |
190 | minfo->hinfo = hinfo; | |
191 | ||
192 | /* copy match config into hashtable config */ | |
193 | memcpy(&hinfo->cfg, &minfo->cfg, sizeof(hinfo->cfg)); | |
194 | hinfo->cfg.size = size; | |
195 | if (!hinfo->cfg.max) | |
196 | hinfo->cfg.max = 8 * hinfo->cfg.size; | |
197 | else if (hinfo->cfg.max < hinfo->cfg.size) | |
198 | hinfo->cfg.max = hinfo->cfg.size; | |
199 | ||
200 | for (i = 0; i < hinfo->cfg.size; i++) | |
201 | INIT_HLIST_HEAD(&hinfo->hash[i]); | |
202 | ||
1da177e4 | 203 | atomic_set(&hinfo->use, 1); |
39b46fc6 PM |
204 | hinfo->count = 0; |
205 | hinfo->family = family; | |
bf0857ea | 206 | hinfo->rnd_initialized = 0; |
1da177e4 | 207 | spin_lock_init(&hinfo->lock); |
39b46fc6 PM |
208 | hinfo->pde = create_proc_entry(minfo->name, 0, |
209 | family == AF_INET ? hashlimit_procdir4 : | |
210 | hashlimit_procdir6); | |
1da177e4 LT |
211 | if (!hinfo->pde) { |
212 | vfree(hinfo); | |
213 | return -1; | |
214 | } | |
215 | hinfo->pde->proc_fops = &dl_file_ops; | |
216 | hinfo->pde->data = hinfo; | |
217 | ||
218 | init_timer(&hinfo->timer); | |
219 | hinfo->timer.expires = jiffies + msecs_to_jiffies(hinfo->cfg.gc_interval); | |
220 | hinfo->timer.data = (unsigned long )hinfo; | |
221 | hinfo->timer.function = htable_gc; | |
222 | add_timer(&hinfo->timer); | |
223 | ||
e45b1be8 | 224 | spin_lock_bh(&hashlimit_lock); |
1da177e4 | 225 | hlist_add_head(&hinfo->node, &hashlimit_htables); |
e45b1be8 | 226 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
227 | |
228 | return 0; | |
229 | } | |
230 | ||
39b46fc6 | 231 | static int select_all(struct xt_hashlimit_htable *ht, struct dsthash_ent *he) |
1da177e4 LT |
232 | { |
233 | return 1; | |
234 | } | |
235 | ||
39b46fc6 | 236 | static int select_gc(struct xt_hashlimit_htable *ht, struct dsthash_ent *he) |
1da177e4 LT |
237 | { |
238 | return (jiffies >= he->expires); | |
239 | } | |
240 | ||
39b46fc6 PM |
241 | static void htable_selective_cleanup(struct xt_hashlimit_htable *ht, |
242 | int (*select)(struct xt_hashlimit_htable *ht, | |
1da177e4 LT |
243 | struct dsthash_ent *he)) |
244 | { | |
39b46fc6 | 245 | unsigned int i; |
1da177e4 LT |
246 | |
247 | /* lock hash table and iterate over it */ | |
248 | spin_lock_bh(&ht->lock); | |
249 | for (i = 0; i < ht->cfg.size; i++) { | |
250 | struct dsthash_ent *dh; | |
251 | struct hlist_node *pos, *n; | |
252 | hlist_for_each_entry_safe(dh, pos, n, &ht->hash[i], node) { | |
253 | if ((*select)(ht, dh)) | |
39b46fc6 | 254 | dsthash_free(ht, dh); |
1da177e4 LT |
255 | } |
256 | } | |
257 | spin_unlock_bh(&ht->lock); | |
258 | } | |
259 | ||
260 | /* hash table garbage collector, run by timer */ | |
261 | static void htable_gc(unsigned long htlong) | |
262 | { | |
39b46fc6 | 263 | struct xt_hashlimit_htable *ht = (struct xt_hashlimit_htable *)htlong; |
1da177e4 LT |
264 | |
265 | htable_selective_cleanup(ht, select_gc); | |
266 | ||
267 | /* re-add the timer accordingly */ | |
268 | ht->timer.expires = jiffies + msecs_to_jiffies(ht->cfg.gc_interval); | |
269 | add_timer(&ht->timer); | |
270 | } | |
271 | ||
39b46fc6 | 272 | static void htable_destroy(struct xt_hashlimit_htable *hinfo) |
1da177e4 LT |
273 | { |
274 | /* remove timer, if it is pending */ | |
275 | if (timer_pending(&hinfo->timer)) | |
276 | del_timer(&hinfo->timer); | |
277 | ||
278 | /* remove proc entry */ | |
39b46fc6 PM |
279 | remove_proc_entry(hinfo->pde->name, |
280 | hinfo->family == AF_INET ? hashlimit_procdir4 : | |
281 | hashlimit_procdir6); | |
1da177e4 LT |
282 | htable_selective_cleanup(hinfo, select_all); |
283 | vfree(hinfo); | |
284 | } | |
285 | ||
39b46fc6 | 286 | static struct xt_hashlimit_htable *htable_find_get(char *name, int family) |
1da177e4 | 287 | { |
39b46fc6 | 288 | struct xt_hashlimit_htable *hinfo; |
1da177e4 LT |
289 | struct hlist_node *pos; |
290 | ||
e45b1be8 | 291 | spin_lock_bh(&hashlimit_lock); |
1da177e4 | 292 | hlist_for_each_entry(hinfo, pos, &hashlimit_htables, node) { |
39b46fc6 PM |
293 | if (!strcmp(name, hinfo->pde->name) && |
294 | hinfo->family == family) { | |
1da177e4 | 295 | atomic_inc(&hinfo->use); |
e45b1be8 | 296 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
297 | return hinfo; |
298 | } | |
299 | } | |
e45b1be8 | 300 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
301 | return NULL; |
302 | } | |
303 | ||
39b46fc6 | 304 | static void htable_put(struct xt_hashlimit_htable *hinfo) |
1da177e4 LT |
305 | { |
306 | if (atomic_dec_and_test(&hinfo->use)) { | |
e45b1be8 | 307 | spin_lock_bh(&hashlimit_lock); |
1da177e4 | 308 | hlist_del(&hinfo->node); |
e45b1be8 | 309 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
310 | htable_destroy(hinfo); |
311 | } | |
312 | } | |
313 | ||
1da177e4 LT |
314 | /* The algorithm used is the Simple Token Bucket Filter (TBF) |
315 | * see net/sched/sch_tbf.c in the linux source tree | |
316 | */ | |
317 | ||
318 | /* Rusty: This is my (non-mathematically-inclined) understanding of | |
319 | this algorithm. The `average rate' in jiffies becomes your initial | |
320 | amount of credit `credit' and the most credit you can ever have | |
321 | `credit_cap'. The `peak rate' becomes the cost of passing the | |
322 | test, `cost'. | |
323 | ||
324 | `prev' tracks the last packet hit: you gain one credit per jiffy. | |
325 | If you get credit balance more than this, the extra credit is | |
326 | discarded. Every time the match passes, you lose `cost' credits; | |
327 | if you don't have that many, the test fails. | |
328 | ||
329 | See Alexey's formal explanation in net/sched/sch_tbf.c. | |
330 | ||
331 | To get the maximum range, we multiply by this factor (ie. you get N | |
332 | credits per jiffy). We want to allow a rate as low as 1 per day | |
333 | (slowest userspace tool allows), which means | |
334 | CREDITS_PER_JIFFY*HZ*60*60*24 < 2^32 ie. | |
335 | */ | |
336 | #define MAX_CPJ (0xFFFFFFFF / (HZ*60*60*24)) | |
337 | ||
338 | /* Repeated shift and or gives us all 1s, final shift and add 1 gives | |
339 | * us the power of 2 below the theoretical max, so GCC simply does a | |
340 | * shift. */ | |
341 | #define _POW2_BELOW2(x) ((x)|((x)>>1)) | |
342 | #define _POW2_BELOW4(x) (_POW2_BELOW2(x)|_POW2_BELOW2((x)>>2)) | |
343 | #define _POW2_BELOW8(x) (_POW2_BELOW4(x)|_POW2_BELOW4((x)>>4)) | |
344 | #define _POW2_BELOW16(x) (_POW2_BELOW8(x)|_POW2_BELOW8((x)>>8)) | |
345 | #define _POW2_BELOW32(x) (_POW2_BELOW16(x)|_POW2_BELOW16((x)>>16)) | |
346 | #define POW2_BELOW32(x) ((_POW2_BELOW32(x)>>1) + 1) | |
347 | ||
348 | #define CREDITS_PER_JIFFY POW2_BELOW32(MAX_CPJ) | |
349 | ||
350 | /* Precision saver. */ | |
351 | static inline u_int32_t | |
352 | user2credits(u_int32_t user) | |
353 | { | |
354 | /* If multiplying would overflow... */ | |
355 | if (user > 0xFFFFFFFF / (HZ*CREDITS_PER_JIFFY)) | |
356 | /* Divide first. */ | |
39b46fc6 | 357 | return (user / XT_HASHLIMIT_SCALE) * HZ * CREDITS_PER_JIFFY; |
1da177e4 | 358 | |
39b46fc6 | 359 | return (user * HZ * CREDITS_PER_JIFFY) / XT_HASHLIMIT_SCALE; |
1da177e4 LT |
360 | } |
361 | ||
362 | static inline void rateinfo_recalc(struct dsthash_ent *dh, unsigned long now) | |
363 | { | |
39b46fc6 | 364 | dh->rateinfo.credit += (now - dh->rateinfo.prev) * CREDITS_PER_JIFFY; |
1da177e4 LT |
365 | if (dh->rateinfo.credit > dh->rateinfo.credit_cap) |
366 | dh->rateinfo.credit = dh->rateinfo.credit_cap; | |
39b46fc6 PM |
367 | dh->rateinfo.prev = now; |
368 | } | |
369 | ||
370 | static int | |
371 | hashlimit_init_dst(struct xt_hashlimit_htable *hinfo, struct dsthash_dst *dst, | |
372 | const struct sk_buff *skb, unsigned int protoff) | |
373 | { | |
374 | __be16 _ports[2], *ports; | |
375 | int nexthdr; | |
376 | ||
377 | memset(dst, 0, sizeof(*dst)); | |
378 | ||
379 | switch (hinfo->family) { | |
380 | case AF_INET: | |
381 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) | |
382 | dst->addr.ip.dst = skb->nh.iph->daddr; | |
383 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) | |
384 | dst->addr.ip.src = skb->nh.iph->saddr; | |
385 | ||
386 | if (!(hinfo->cfg.mode & | |
387 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
388 | return 0; | |
389 | nexthdr = skb->nh.iph->protocol; | |
390 | break; | |
391 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) | |
392 | case AF_INET6: | |
393 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) | |
394 | memcpy(&dst->addr.ip6.dst, &skb->nh.ipv6h->daddr, | |
395 | sizeof(dst->addr.ip6.dst)); | |
396 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) | |
397 | memcpy(&dst->addr.ip6.src, &skb->nh.ipv6h->saddr, | |
398 | sizeof(dst->addr.ip6.src)); | |
399 | ||
400 | if (!(hinfo->cfg.mode & | |
401 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
402 | return 0; | |
403 | nexthdr = ipv6_find_hdr(skb, &protoff, -1, NULL); | |
404 | if (nexthdr < 0) | |
405 | return -1; | |
406 | break; | |
407 | #endif | |
408 | default: | |
409 | BUG(); | |
410 | return 0; | |
411 | } | |
412 | ||
413 | switch (nexthdr) { | |
414 | case IPPROTO_TCP: | |
415 | case IPPROTO_UDP: | |
416 | case IPPROTO_SCTP: | |
417 | case IPPROTO_DCCP: | |
418 | ports = skb_header_pointer(skb, protoff, sizeof(_ports), | |
419 | &_ports); | |
420 | break; | |
421 | default: | |
422 | _ports[0] = _ports[1] = 0; | |
423 | ports = _ports; | |
424 | break; | |
425 | } | |
426 | if (!ports) | |
427 | return -1; | |
428 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SPT) | |
429 | dst->src_port = ports[0]; | |
430 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DPT) | |
431 | dst->dst_port = ports[1]; | |
432 | return 0; | |
1da177e4 LT |
433 | } |
434 | ||
1da177e4 LT |
435 | static int |
436 | hashlimit_match(const struct sk_buff *skb, | |
437 | const struct net_device *in, | |
438 | const struct net_device *out, | |
c4986734 | 439 | const struct xt_match *match, |
1da177e4 LT |
440 | const void *matchinfo, |
441 | int offset, | |
2e4e6a17 | 442 | unsigned int protoff, |
1da177e4 LT |
443 | int *hotdrop) |
444 | { | |
39b46fc6 PM |
445 | struct xt_hashlimit_info *r = |
446 | ((struct xt_hashlimit_info *)matchinfo)->u.master; | |
447 | struct xt_hashlimit_htable *hinfo = r->hinfo; | |
1da177e4 LT |
448 | unsigned long now = jiffies; |
449 | struct dsthash_ent *dh; | |
450 | struct dsthash_dst dst; | |
451 | ||
39b46fc6 PM |
452 | if (hashlimit_init_dst(hinfo, &dst, skb, protoff) < 0) |
453 | goto hotdrop; | |
1da177e4 LT |
454 | |
455 | spin_lock_bh(&hinfo->lock); | |
39b46fc6 | 456 | dh = dsthash_find(hinfo, &dst); |
1da177e4 | 457 | if (!dh) { |
39b46fc6 | 458 | dh = dsthash_alloc_init(hinfo, &dst); |
1da177e4 | 459 | if (!dh) { |
1da177e4 | 460 | spin_unlock_bh(&hinfo->lock); |
39b46fc6 | 461 | goto hotdrop; |
1da177e4 LT |
462 | } |
463 | ||
464 | dh->expires = jiffies + msecs_to_jiffies(hinfo->cfg.expire); | |
1da177e4 | 465 | dh->rateinfo.prev = jiffies; |
39b46fc6 PM |
466 | dh->rateinfo.credit = user2credits(hinfo->cfg.avg * |
467 | hinfo->cfg.burst); | |
468 | dh->rateinfo.credit_cap = user2credits(hinfo->cfg.avg * | |
469 | hinfo->cfg.burst); | |
1da177e4 | 470 | dh->rateinfo.cost = user2credits(hinfo->cfg.avg); |
1c7628bd PM |
471 | } else { |
472 | /* update expiration timeout */ | |
473 | dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire); | |
474 | rateinfo_recalc(dh, now); | |
1da177e4 LT |
475 | } |
476 | ||
1da177e4 LT |
477 | if (dh->rateinfo.credit >= dh->rateinfo.cost) { |
478 | /* We're underlimit. */ | |
479 | dh->rateinfo.credit -= dh->rateinfo.cost; | |
480 | spin_unlock_bh(&hinfo->lock); | |
481 | return 1; | |
482 | } | |
483 | ||
484 | spin_unlock_bh(&hinfo->lock); | |
485 | ||
486 | /* default case: we're overlimit, thus don't match */ | |
487 | return 0; | |
39b46fc6 PM |
488 | |
489 | hotdrop: | |
490 | *hotdrop = 1; | |
491 | return 0; | |
1da177e4 LT |
492 | } |
493 | ||
494 | static int | |
495 | hashlimit_checkentry(const char *tablename, | |
2e4e6a17 | 496 | const void *inf, |
c4986734 | 497 | const struct xt_match *match, |
1da177e4 | 498 | void *matchinfo, |
1da177e4 LT |
499 | unsigned int hook_mask) |
500 | { | |
39b46fc6 | 501 | struct xt_hashlimit_info *r = matchinfo; |
1da177e4 | 502 | |
1da177e4 | 503 | /* Check for overflow. */ |
39b46fc6 PM |
504 | if (r->cfg.burst == 0 || |
505 | user2credits(r->cfg.avg * r->cfg.burst) < user2credits(r->cfg.avg)) { | |
506 | printk(KERN_ERR "xt_hashlimit: overflow, try lower: %u/%u\n", | |
1da177e4 LT |
507 | r->cfg.avg, r->cfg.burst); |
508 | return 0; | |
509 | } | |
39b46fc6 PM |
510 | if (r->cfg.mode == 0 || |
511 | r->cfg.mode > (XT_HASHLIMIT_HASH_DPT | | |
512 | XT_HASHLIMIT_HASH_DIP | | |
513 | XT_HASHLIMIT_HASH_SIP | | |
514 | XT_HASHLIMIT_HASH_SPT)) | |
1da177e4 | 515 | return 0; |
1da177e4 LT |
516 | if (!r->cfg.gc_interval) |
517 | return 0; | |
1da177e4 LT |
518 | if (!r->cfg.expire) |
519 | return 0; | |
3ab72088 PM |
520 | if (r->name[sizeof(r->name) - 1] != '\0') |
521 | return 0; | |
522 | ||
1da177e4 | 523 | /* This is the best we've got: We cannot release and re-grab lock, |
39b46fc6 PM |
524 | * since checkentry() is called before x_tables.c grabs xt_mutex. |
525 | * We also cannot grab the hashtable spinlock, since htable_create will | |
1da177e4 LT |
526 | * call vmalloc, and that can sleep. And we cannot just re-search |
527 | * the list of htable's in htable_create(), since then we would | |
528 | * create duplicate proc files. -HW */ | |
14cc3e2b | 529 | mutex_lock(&hlimit_mutex); |
39b46fc6 PM |
530 | r->hinfo = htable_find_get(r->name, match->family); |
531 | if (!r->hinfo && htable_create(r, match->family) != 0) { | |
14cc3e2b | 532 | mutex_unlock(&hlimit_mutex); |
1da177e4 LT |
533 | return 0; |
534 | } | |
14cc3e2b | 535 | mutex_unlock(&hlimit_mutex); |
1da177e4 LT |
536 | |
537 | /* Ugly hack: For SMP, we only want to use one set */ | |
538 | r->u.master = r; | |
1da177e4 LT |
539 | return 1; |
540 | } | |
541 | ||
542 | static void | |
efa74165 | 543 | hashlimit_destroy(const struct xt_match *match, void *matchinfo) |
1da177e4 | 544 | { |
39b46fc6 | 545 | struct xt_hashlimit_info *r = matchinfo; |
1da177e4 LT |
546 | |
547 | htable_put(r->hinfo); | |
548 | } | |
549 | ||
127f15dd | 550 | #ifdef CONFIG_COMPAT |
39b46fc6 | 551 | struct compat_xt_hashlimit_info { |
127f15dd PM |
552 | char name[IFNAMSIZ]; |
553 | struct hashlimit_cfg cfg; | |
554 | compat_uptr_t hinfo; | |
555 | compat_uptr_t master; | |
556 | }; | |
557 | ||
558 | static void compat_from_user(void *dst, void *src) | |
559 | { | |
39b46fc6 | 560 | int off = offsetof(struct compat_xt_hashlimit_info, hinfo); |
127f15dd PM |
561 | |
562 | memcpy(dst, src, off); | |
39b46fc6 | 563 | memset(dst + off, 0, sizeof(struct compat_xt_hashlimit_info) - off); |
127f15dd PM |
564 | } |
565 | ||
566 | static int compat_to_user(void __user *dst, void *src) | |
567 | { | |
39b46fc6 | 568 | int off = offsetof(struct compat_xt_hashlimit_info, hinfo); |
127f15dd PM |
569 | |
570 | return copy_to_user(dst, src, off) ? -EFAULT : 0; | |
571 | } | |
572 | #endif | |
573 | ||
39b46fc6 PM |
574 | static struct xt_match xt_hashlimit[] = { |
575 | { | |
576 | .name = "hashlimit", | |
577 | .family = AF_INET, | |
578 | .match = hashlimit_match, | |
579 | .matchsize = sizeof(struct xt_hashlimit_info), | |
580 | #ifdef CONFIG_COMPAT | |
581 | .compatsize = sizeof(struct compat_xt_hashlimit_info), | |
582 | .compat_from_user = compat_from_user, | |
583 | .compat_to_user = compat_to_user, | |
584 | #endif | |
585 | .checkentry = hashlimit_checkentry, | |
586 | .destroy = hashlimit_destroy, | |
587 | .me = THIS_MODULE | |
588 | }, | |
589 | { | |
590 | .name = "hashlimit", | |
591 | .family = AF_INET6, | |
592 | .match = hashlimit_match, | |
593 | .matchsize = sizeof(struct xt_hashlimit_info), | |
127f15dd | 594 | #ifdef CONFIG_COMPAT |
39b46fc6 PM |
595 | .compatsize = sizeof(struct compat_xt_hashlimit_info), |
596 | .compat_from_user = compat_from_user, | |
597 | .compat_to_user = compat_to_user, | |
127f15dd | 598 | #endif |
39b46fc6 PM |
599 | .checkentry = hashlimit_checkentry, |
600 | .destroy = hashlimit_destroy, | |
601 | .me = THIS_MODULE | |
602 | }, | |
1da177e4 LT |
603 | }; |
604 | ||
605 | /* PROC stuff */ | |
1da177e4 LT |
606 | static void *dl_seq_start(struct seq_file *s, loff_t *pos) |
607 | { | |
608 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 609 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
610 | unsigned int *bucket; |
611 | ||
612 | spin_lock_bh(&htable->lock); | |
613 | if (*pos >= htable->cfg.size) | |
614 | return NULL; | |
615 | ||
616 | bucket = kmalloc(sizeof(unsigned int), GFP_ATOMIC); | |
617 | if (!bucket) | |
618 | return ERR_PTR(-ENOMEM); | |
619 | ||
620 | *bucket = *pos; | |
621 | return bucket; | |
622 | } | |
623 | ||
624 | static void *dl_seq_next(struct seq_file *s, void *v, loff_t *pos) | |
625 | { | |
626 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 627 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
628 | unsigned int *bucket = (unsigned int *)v; |
629 | ||
630 | *pos = ++(*bucket); | |
631 | if (*pos >= htable->cfg.size) { | |
632 | kfree(v); | |
633 | return NULL; | |
634 | } | |
635 | return bucket; | |
636 | } | |
637 | ||
638 | static void dl_seq_stop(struct seq_file *s, void *v) | |
639 | { | |
640 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 641 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
642 | unsigned int *bucket = (unsigned int *)v; |
643 | ||
644 | kfree(bucket); | |
1da177e4 LT |
645 | spin_unlock_bh(&htable->lock); |
646 | } | |
647 | ||
39b46fc6 PM |
648 | static int dl_seq_real_show(struct dsthash_ent *ent, int family, |
649 | struct seq_file *s) | |
1da177e4 LT |
650 | { |
651 | /* recalculate to show accurate numbers */ | |
652 | rateinfo_recalc(ent, jiffies); | |
653 | ||
39b46fc6 PM |
654 | switch (family) { |
655 | case AF_INET: | |
656 | return seq_printf(s, "%ld %u.%u.%u.%u:%u->" | |
657 | "%u.%u.%u.%u:%u %u %u %u\n", | |
658 | (long)(ent->expires - jiffies)/HZ, | |
659 | NIPQUAD(ent->dst.addr.ip.src), | |
660 | ntohs(ent->dst.src_port), | |
661 | NIPQUAD(ent->dst.addr.ip.dst), | |
662 | ntohs(ent->dst.dst_port), | |
663 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
664 | ent->rateinfo.cost); | |
665 | case AF_INET6: | |
666 | return seq_printf(s, "%ld " NIP6_FMT ":%u->" | |
667 | NIP6_FMT ":%u %u %u %u\n", | |
668 | (long)(ent->expires - jiffies)/HZ, | |
669 | NIP6(*(struct in6_addr *)&ent->dst.addr.ip6.src), | |
670 | ntohs(ent->dst.src_port), | |
671 | NIP6(*(struct in6_addr *)&ent->dst.addr.ip6.dst), | |
672 | ntohs(ent->dst.dst_port), | |
673 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
674 | ent->rateinfo.cost); | |
675 | default: | |
676 | BUG(); | |
677 | return 0; | |
678 | } | |
1da177e4 LT |
679 | } |
680 | ||
681 | static int dl_seq_show(struct seq_file *s, void *v) | |
682 | { | |
683 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 684 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
685 | unsigned int *bucket = (unsigned int *)v; |
686 | struct dsthash_ent *ent; | |
687 | struct hlist_node *pos; | |
688 | ||
39b46fc6 PM |
689 | if (!hlist_empty(&htable->hash[*bucket])) { |
690 | hlist_for_each_entry(ent, pos, &htable->hash[*bucket], node) | |
691 | if (dl_seq_real_show(ent, htable->family, s)) | |
1da177e4 | 692 | return 1; |
39b46fc6 | 693 | } |
1da177e4 LT |
694 | return 0; |
695 | } | |
696 | ||
697 | static struct seq_operations dl_seq_ops = { | |
698 | .start = dl_seq_start, | |
699 | .next = dl_seq_next, | |
700 | .stop = dl_seq_stop, | |
701 | .show = dl_seq_show | |
702 | }; | |
703 | ||
704 | static int dl_proc_open(struct inode *inode, struct file *file) | |
705 | { | |
706 | int ret = seq_open(file, &dl_seq_ops); | |
707 | ||
708 | if (!ret) { | |
709 | struct seq_file *sf = file->private_data; | |
710 | sf->private = PDE(inode); | |
711 | } | |
712 | return ret; | |
713 | } | |
714 | ||
715 | static struct file_operations dl_file_ops = { | |
716 | .owner = THIS_MODULE, | |
717 | .open = dl_proc_open, | |
718 | .read = seq_read, | |
719 | .llseek = seq_lseek, | |
720 | .release = seq_release | |
721 | }; | |
722 | ||
39b46fc6 | 723 | static int __init xt_hashlimit_init(void) |
1da177e4 | 724 | { |
39b46fc6 | 725 | int err; |
1da177e4 | 726 | |
39b46fc6 PM |
727 | err = xt_register_matches(xt_hashlimit, ARRAY_SIZE(xt_hashlimit)); |
728 | if (err < 0) | |
729 | goto err1; | |
1da177e4 | 730 | |
39b46fc6 PM |
731 | err = -ENOMEM; |
732 | hashlimit_cachep = kmem_cache_create("xt_hashlimit", | |
733 | sizeof(struct dsthash_ent), 0, 0, | |
734 | NULL, NULL); | |
1da177e4 | 735 | if (!hashlimit_cachep) { |
39b46fc6 PM |
736 | printk(KERN_ERR "xt_hashlimit: unable to create slab cache\n"); |
737 | goto err2; | |
1da177e4 | 738 | } |
39b46fc6 PM |
739 | hashlimit_procdir4 = proc_mkdir("ipt_hashlimit", proc_net); |
740 | if (!hashlimit_procdir4) { | |
741 | printk(KERN_ERR "xt_hashlimit: unable to create proc dir " | |
742 | "entry\n"); | |
743 | goto err3; | |
1da177e4 | 744 | } |
39b46fc6 PM |
745 | hashlimit_procdir6 = proc_mkdir("ip6t_hashlimit", proc_net); |
746 | if (!hashlimit_procdir6) { | |
747 | printk(KERN_ERR "xt_hashlimit: tnable to create proc dir " | |
748 | "entry\n"); | |
749 | goto err4; | |
750 | } | |
751 | return 0; | |
752 | err4: | |
1da177e4 | 753 | remove_proc_entry("ipt_hashlimit", proc_net); |
39b46fc6 | 754 | err3: |
1da177e4 | 755 | kmem_cache_destroy(hashlimit_cachep); |
39b46fc6 PM |
756 | err2: |
757 | xt_unregister_matches(xt_hashlimit, ARRAY_SIZE(xt_hashlimit)); | |
758 | err1: | |
759 | return err; | |
1da177e4 | 760 | |
1da177e4 LT |
761 | } |
762 | ||
39b46fc6 | 763 | static void __exit xt_hashlimit_fini(void) |
1da177e4 | 764 | { |
39b46fc6 PM |
765 | remove_proc_entry("ipt_hashlimit", proc_net); |
766 | remove_proc_entry("ip6t_hashlimit", proc_net); | |
767 | kmem_cache_destroy(hashlimit_cachep); | |
768 | xt_unregister_matches(xt_hashlimit, ARRAY_SIZE(xt_hashlimit)); | |
1da177e4 LT |
769 | } |
770 | ||
39b46fc6 PM |
771 | module_init(xt_hashlimit_init); |
772 | module_exit(xt_hashlimit_fini); |