]>
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 | 19 | #include <linux/skbuff.h> |
d7fe0f24 | 20 | #include <linux/mm.h> |
39b46fc6 PM |
21 | #include <linux/in.h> |
22 | #include <linux/ip.h> | |
23 | #include <linux/ipv6.h> | |
1da177e4 | 24 | |
39b46fc6 | 25 | #include <linux/netfilter/x_tables.h> |
1da177e4 | 26 | #include <linux/netfilter_ipv4/ip_tables.h> |
39b46fc6 PM |
27 | #include <linux/netfilter_ipv6/ip6_tables.h> |
28 | #include <linux/netfilter/xt_hashlimit.h> | |
14cc3e2b | 29 | #include <linux/mutex.h> |
1da177e4 LT |
30 | |
31 | MODULE_LICENSE("GPL"); | |
32 | MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); | |
33 | MODULE_DESCRIPTION("iptables match for limiting per hash-bucket"); | |
39b46fc6 PM |
34 | MODULE_ALIAS("ipt_hashlimit"); |
35 | MODULE_ALIAS("ip6t_hashlimit"); | |
1da177e4 LT |
36 | |
37 | /* need to declare this at the top */ | |
39b46fc6 PM |
38 | static struct proc_dir_entry *hashlimit_procdir4; |
39 | static struct proc_dir_entry *hashlimit_procdir6; | |
da7071d7 | 40 | static const struct file_operations dl_file_ops; |
1da177e4 LT |
41 | |
42 | /* hash table crap */ | |
1da177e4 | 43 | struct dsthash_dst { |
39b46fc6 PM |
44 | union { |
45 | struct { | |
46 | __be32 src; | |
47 | __be32 dst; | |
48 | } ip; | |
49 | struct { | |
50 | __be32 src[4]; | |
51 | __be32 dst[4]; | |
52 | } ip6; | |
53 | } addr; | |
6a19d614 AV |
54 | __be16 src_port; |
55 | __be16 dst_port; | |
1da177e4 LT |
56 | }; |
57 | ||
58 | struct dsthash_ent { | |
59 | /* static / read-only parts in the beginning */ | |
60 | struct hlist_node node; | |
61 | struct dsthash_dst dst; | |
62 | ||
63 | /* modified structure members in the end */ | |
64 | unsigned long expires; /* precalculated expiry time */ | |
65 | struct { | |
66 | unsigned long prev; /* last modification */ | |
67 | u_int32_t credit; | |
68 | u_int32_t credit_cap, cost; | |
69 | } rateinfo; | |
70 | }; | |
71 | ||
39b46fc6 | 72 | struct xt_hashlimit_htable { |
1da177e4 LT |
73 | struct hlist_node node; /* global list of all htables */ |
74 | atomic_t use; | |
39b46fc6 | 75 | int family; |
1da177e4 LT |
76 | |
77 | struct hashlimit_cfg cfg; /* config */ | |
78 | ||
79 | /* used internally */ | |
80 | spinlock_t lock; /* lock for list_head */ | |
81 | u_int32_t rnd; /* random seed for hash */ | |
bf0857ea | 82 | int rnd_initialized; |
39b46fc6 | 83 | unsigned int count; /* number entries in table */ |
1da177e4 | 84 | struct timer_list timer; /* timer for gc */ |
1da177e4 LT |
85 | |
86 | /* seq_file stuff */ | |
87 | struct proc_dir_entry *pde; | |
88 | ||
89 | struct hlist_head hash[0]; /* hashtable itself */ | |
90 | }; | |
91 | ||
e45b1be8 | 92 | static DEFINE_SPINLOCK(hashlimit_lock); /* protects htables list */ |
14cc3e2b | 93 | static DEFINE_MUTEX(hlimit_mutex); /* additional checkentry protection */ |
1da177e4 | 94 | static HLIST_HEAD(hashlimit_htables); |
e18b890b | 95 | static struct kmem_cache *hashlimit_cachep __read_mostly; |
1da177e4 | 96 | |
1d93a9cb | 97 | static inline bool dst_cmp(const struct dsthash_ent *ent, |
a47362a2 | 98 | const struct dsthash_dst *b) |
1da177e4 | 99 | { |
39b46fc6 | 100 | return !memcmp(&ent->dst, b, sizeof(ent->dst)); |
1da177e4 LT |
101 | } |
102 | ||
39b46fc6 PM |
103 | static u_int32_t |
104 | hash_dst(const struct xt_hashlimit_htable *ht, const struct dsthash_dst *dst) | |
1da177e4 | 105 | { |
39b46fc6 | 106 | return jhash(dst, sizeof(*dst), ht->rnd) % ht->cfg.size; |
1da177e4 LT |
107 | } |
108 | ||
39b46fc6 | 109 | static struct dsthash_ent * |
a47362a2 JE |
110 | dsthash_find(const struct xt_hashlimit_htable *ht, |
111 | const struct dsthash_dst *dst) | |
1da177e4 LT |
112 | { |
113 | struct dsthash_ent *ent; | |
114 | struct hlist_node *pos; | |
115 | u_int32_t hash = hash_dst(ht, dst); | |
116 | ||
39b46fc6 PM |
117 | if (!hlist_empty(&ht->hash[hash])) { |
118 | hlist_for_each_entry(ent, pos, &ht->hash[hash], node) | |
119 | if (dst_cmp(ent, dst)) | |
1da177e4 | 120 | return ent; |
39b46fc6 | 121 | } |
1da177e4 LT |
122 | return NULL; |
123 | } | |
124 | ||
125 | /* allocate dsthash_ent, initialize dst, put in htable and lock it */ | |
126 | static struct dsthash_ent * | |
a47362a2 JE |
127 | dsthash_alloc_init(struct xt_hashlimit_htable *ht, |
128 | const struct dsthash_dst *dst) | |
1da177e4 LT |
129 | { |
130 | struct dsthash_ent *ent; | |
131 | ||
132 | /* initialize hash with random val at the time we allocate | |
133 | * the first hashtable entry */ | |
bf0857ea | 134 | if (!ht->rnd_initialized) { |
1da177e4 | 135 | get_random_bytes(&ht->rnd, 4); |
bf0857ea PM |
136 | ht->rnd_initialized = 1; |
137 | } | |
1da177e4 | 138 | |
39b46fc6 | 139 | if (ht->cfg.max && ht->count >= ht->cfg.max) { |
1da177e4 LT |
140 | /* FIXME: do something. question is what.. */ |
141 | if (net_ratelimit()) | |
39b46fc6 PM |
142 | printk(KERN_WARNING |
143 | "xt_hashlimit: max count of %u reached\n", | |
1da177e4 LT |
144 | ht->cfg.max); |
145 | return NULL; | |
146 | } | |
147 | ||
148 | ent = kmem_cache_alloc(hashlimit_cachep, GFP_ATOMIC); | |
149 | if (!ent) { | |
150 | if (net_ratelimit()) | |
39b46fc6 PM |
151 | printk(KERN_ERR |
152 | "xt_hashlimit: can't allocate dsthash_ent\n"); | |
1da177e4 LT |
153 | return NULL; |
154 | } | |
39b46fc6 | 155 | memcpy(&ent->dst, dst, sizeof(ent->dst)); |
1da177e4 LT |
156 | |
157 | hlist_add_head(&ent->node, &ht->hash[hash_dst(ht, dst)]); | |
39b46fc6 | 158 | ht->count++; |
1da177e4 LT |
159 | return ent; |
160 | } | |
161 | ||
39b46fc6 PM |
162 | static inline void |
163 | dsthash_free(struct xt_hashlimit_htable *ht, struct dsthash_ent *ent) | |
1da177e4 LT |
164 | { |
165 | hlist_del(&ent->node); | |
166 | kmem_cache_free(hashlimit_cachep, ent); | |
39b46fc6 | 167 | ht->count--; |
1da177e4 LT |
168 | } |
169 | static void htable_gc(unsigned long htlong); | |
170 | ||
39b46fc6 | 171 | static int htable_create(struct xt_hashlimit_info *minfo, int family) |
1da177e4 | 172 | { |
39b46fc6 | 173 | struct xt_hashlimit_htable *hinfo; |
1da177e4 | 174 | unsigned int size; |
39b46fc6 | 175 | unsigned int i; |
1da177e4 LT |
176 | |
177 | if (minfo->cfg.size) | |
178 | size = minfo->cfg.size; | |
179 | else { | |
39b46fc6 PM |
180 | size = ((num_physpages << PAGE_SHIFT) / 16384) / |
181 | sizeof(struct list_head); | |
1da177e4 LT |
182 | if (num_physpages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
183 | size = 8192; | |
184 | if (size < 16) | |
185 | size = 16; | |
186 | } | |
187 | /* FIXME: don't use vmalloc() here or anywhere else -HW */ | |
39b46fc6 PM |
188 | hinfo = vmalloc(sizeof(struct xt_hashlimit_htable) + |
189 | sizeof(struct list_head) * size); | |
1da177e4 | 190 | if (!hinfo) { |
39b46fc6 | 191 | printk(KERN_ERR "xt_hashlimit: unable to create hashtable\n"); |
1da177e4 LT |
192 | return -1; |
193 | } | |
194 | minfo->hinfo = hinfo; | |
195 | ||
196 | /* copy match config into hashtable config */ | |
197 | memcpy(&hinfo->cfg, &minfo->cfg, sizeof(hinfo->cfg)); | |
198 | hinfo->cfg.size = size; | |
199 | if (!hinfo->cfg.max) | |
200 | hinfo->cfg.max = 8 * hinfo->cfg.size; | |
201 | else if (hinfo->cfg.max < hinfo->cfg.size) | |
202 | hinfo->cfg.max = hinfo->cfg.size; | |
203 | ||
204 | for (i = 0; i < hinfo->cfg.size; i++) | |
205 | INIT_HLIST_HEAD(&hinfo->hash[i]); | |
206 | ||
1da177e4 | 207 | atomic_set(&hinfo->use, 1); |
39b46fc6 PM |
208 | hinfo->count = 0; |
209 | hinfo->family = family; | |
bf0857ea | 210 | hinfo->rnd_initialized = 0; |
1da177e4 | 211 | spin_lock_init(&hinfo->lock); |
39b46fc6 PM |
212 | hinfo->pde = create_proc_entry(minfo->name, 0, |
213 | family == AF_INET ? hashlimit_procdir4 : | |
601e68e1 | 214 | hashlimit_procdir6); |
1da177e4 LT |
215 | if (!hinfo->pde) { |
216 | vfree(hinfo); | |
217 | return -1; | |
218 | } | |
219 | hinfo->pde->proc_fops = &dl_file_ops; | |
220 | hinfo->pde->data = hinfo; | |
221 | ||
e6f689db | 222 | setup_timer(&hinfo->timer, htable_gc, (unsigned long )hinfo); |
1da177e4 | 223 | hinfo->timer.expires = jiffies + msecs_to_jiffies(hinfo->cfg.gc_interval); |
1da177e4 LT |
224 | add_timer(&hinfo->timer); |
225 | ||
e45b1be8 | 226 | spin_lock_bh(&hashlimit_lock); |
1da177e4 | 227 | hlist_add_head(&hinfo->node, &hashlimit_htables); |
e45b1be8 | 228 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
229 | |
230 | return 0; | |
231 | } | |
232 | ||
a47362a2 JE |
233 | static bool select_all(const struct xt_hashlimit_htable *ht, |
234 | const struct dsthash_ent *he) | |
1da177e4 LT |
235 | { |
236 | return 1; | |
237 | } | |
238 | ||
a47362a2 JE |
239 | static bool select_gc(const struct xt_hashlimit_htable *ht, |
240 | const struct dsthash_ent *he) | |
1da177e4 LT |
241 | { |
242 | return (jiffies >= he->expires); | |
243 | } | |
244 | ||
39b46fc6 | 245 | static void htable_selective_cleanup(struct xt_hashlimit_htable *ht, |
a47362a2 JE |
246 | bool (*select)(const struct xt_hashlimit_htable *ht, |
247 | const struct dsthash_ent *he)) | |
1da177e4 | 248 | { |
39b46fc6 | 249 | unsigned int i; |
1da177e4 LT |
250 | |
251 | /* lock hash table and iterate over it */ | |
252 | spin_lock_bh(&ht->lock); | |
253 | for (i = 0; i < ht->cfg.size; i++) { | |
254 | struct dsthash_ent *dh; | |
255 | struct hlist_node *pos, *n; | |
256 | hlist_for_each_entry_safe(dh, pos, n, &ht->hash[i], node) { | |
257 | if ((*select)(ht, dh)) | |
39b46fc6 | 258 | dsthash_free(ht, dh); |
1da177e4 LT |
259 | } |
260 | } | |
261 | spin_unlock_bh(&ht->lock); | |
262 | } | |
263 | ||
264 | /* hash table garbage collector, run by timer */ | |
265 | static void htable_gc(unsigned long htlong) | |
266 | { | |
39b46fc6 | 267 | struct xt_hashlimit_htable *ht = (struct xt_hashlimit_htable *)htlong; |
1da177e4 LT |
268 | |
269 | htable_selective_cleanup(ht, select_gc); | |
270 | ||
271 | /* re-add the timer accordingly */ | |
272 | ht->timer.expires = jiffies + msecs_to_jiffies(ht->cfg.gc_interval); | |
273 | add_timer(&ht->timer); | |
274 | } | |
275 | ||
39b46fc6 | 276 | static void htable_destroy(struct xt_hashlimit_htable *hinfo) |
1da177e4 LT |
277 | { |
278 | /* remove timer, if it is pending */ | |
279 | if (timer_pending(&hinfo->timer)) | |
280 | del_timer(&hinfo->timer); | |
281 | ||
282 | /* remove proc entry */ | |
39b46fc6 PM |
283 | remove_proc_entry(hinfo->pde->name, |
284 | hinfo->family == AF_INET ? hashlimit_procdir4 : | |
601e68e1 | 285 | hashlimit_procdir6); |
1da177e4 LT |
286 | htable_selective_cleanup(hinfo, select_all); |
287 | vfree(hinfo); | |
288 | } | |
289 | ||
a47362a2 JE |
290 | static struct xt_hashlimit_htable *htable_find_get(const char *name, |
291 | int family) | |
1da177e4 | 292 | { |
39b46fc6 | 293 | struct xt_hashlimit_htable *hinfo; |
1da177e4 LT |
294 | struct hlist_node *pos; |
295 | ||
e45b1be8 | 296 | spin_lock_bh(&hashlimit_lock); |
1da177e4 | 297 | hlist_for_each_entry(hinfo, pos, &hashlimit_htables, node) { |
39b46fc6 PM |
298 | if (!strcmp(name, hinfo->pde->name) && |
299 | hinfo->family == family) { | |
1da177e4 | 300 | atomic_inc(&hinfo->use); |
e45b1be8 | 301 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
302 | return hinfo; |
303 | } | |
304 | } | |
e45b1be8 | 305 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
306 | return NULL; |
307 | } | |
308 | ||
39b46fc6 | 309 | static void htable_put(struct xt_hashlimit_htable *hinfo) |
1da177e4 LT |
310 | { |
311 | if (atomic_dec_and_test(&hinfo->use)) { | |
e45b1be8 | 312 | spin_lock_bh(&hashlimit_lock); |
1da177e4 | 313 | hlist_del(&hinfo->node); |
e45b1be8 | 314 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
315 | htable_destroy(hinfo); |
316 | } | |
317 | } | |
318 | ||
1da177e4 LT |
319 | /* The algorithm used is the Simple Token Bucket Filter (TBF) |
320 | * see net/sched/sch_tbf.c in the linux source tree | |
321 | */ | |
322 | ||
323 | /* Rusty: This is my (non-mathematically-inclined) understanding of | |
324 | this algorithm. The `average rate' in jiffies becomes your initial | |
325 | amount of credit `credit' and the most credit you can ever have | |
326 | `credit_cap'. The `peak rate' becomes the cost of passing the | |
327 | test, `cost'. | |
328 | ||
329 | `prev' tracks the last packet hit: you gain one credit per jiffy. | |
330 | If you get credit balance more than this, the extra credit is | |
331 | discarded. Every time the match passes, you lose `cost' credits; | |
332 | if you don't have that many, the test fails. | |
333 | ||
334 | See Alexey's formal explanation in net/sched/sch_tbf.c. | |
335 | ||
336 | To get the maximum range, we multiply by this factor (ie. you get N | |
337 | credits per jiffy). We want to allow a rate as low as 1 per day | |
338 | (slowest userspace tool allows), which means | |
339 | CREDITS_PER_JIFFY*HZ*60*60*24 < 2^32 ie. | |
340 | */ | |
341 | #define MAX_CPJ (0xFFFFFFFF / (HZ*60*60*24)) | |
342 | ||
343 | /* Repeated shift and or gives us all 1s, final shift and add 1 gives | |
344 | * us the power of 2 below the theoretical max, so GCC simply does a | |
345 | * shift. */ | |
346 | #define _POW2_BELOW2(x) ((x)|((x)>>1)) | |
347 | #define _POW2_BELOW4(x) (_POW2_BELOW2(x)|_POW2_BELOW2((x)>>2)) | |
348 | #define _POW2_BELOW8(x) (_POW2_BELOW4(x)|_POW2_BELOW4((x)>>4)) | |
349 | #define _POW2_BELOW16(x) (_POW2_BELOW8(x)|_POW2_BELOW8((x)>>8)) | |
350 | #define _POW2_BELOW32(x) (_POW2_BELOW16(x)|_POW2_BELOW16((x)>>16)) | |
351 | #define POW2_BELOW32(x) ((_POW2_BELOW32(x)>>1) + 1) | |
352 | ||
353 | #define CREDITS_PER_JIFFY POW2_BELOW32(MAX_CPJ) | |
354 | ||
355 | /* Precision saver. */ | |
356 | static inline u_int32_t | |
357 | user2credits(u_int32_t user) | |
358 | { | |
359 | /* If multiplying would overflow... */ | |
360 | if (user > 0xFFFFFFFF / (HZ*CREDITS_PER_JIFFY)) | |
361 | /* Divide first. */ | |
39b46fc6 | 362 | return (user / XT_HASHLIMIT_SCALE) * HZ * CREDITS_PER_JIFFY; |
1da177e4 | 363 | |
39b46fc6 | 364 | return (user * HZ * CREDITS_PER_JIFFY) / XT_HASHLIMIT_SCALE; |
1da177e4 LT |
365 | } |
366 | ||
367 | static inline void rateinfo_recalc(struct dsthash_ent *dh, unsigned long now) | |
368 | { | |
39b46fc6 | 369 | dh->rateinfo.credit += (now - dh->rateinfo.prev) * CREDITS_PER_JIFFY; |
1da177e4 LT |
370 | if (dh->rateinfo.credit > dh->rateinfo.credit_cap) |
371 | dh->rateinfo.credit = dh->rateinfo.credit_cap; | |
39b46fc6 PM |
372 | dh->rateinfo.prev = now; |
373 | } | |
374 | ||
375 | static int | |
a47362a2 JE |
376 | hashlimit_init_dst(const struct xt_hashlimit_htable *hinfo, |
377 | struct dsthash_dst *dst, | |
39b46fc6 PM |
378 | const struct sk_buff *skb, unsigned int protoff) |
379 | { | |
380 | __be16 _ports[2], *ports; | |
381 | int nexthdr; | |
382 | ||
383 | memset(dst, 0, sizeof(*dst)); | |
384 | ||
385 | switch (hinfo->family) { | |
386 | case AF_INET: | |
387 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) | |
eddc9ec5 | 388 | dst->addr.ip.dst = ip_hdr(skb)->daddr; |
39b46fc6 | 389 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) |
eddc9ec5 | 390 | dst->addr.ip.src = ip_hdr(skb)->saddr; |
39b46fc6 PM |
391 | |
392 | if (!(hinfo->cfg.mode & | |
393 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
394 | return 0; | |
eddc9ec5 | 395 | nexthdr = ip_hdr(skb)->protocol; |
39b46fc6 | 396 | break; |
02dba025 | 397 | #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE) |
39b46fc6 PM |
398 | case AF_INET6: |
399 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) | |
0660e03f | 400 | memcpy(&dst->addr.ip6.dst, &ipv6_hdr(skb)->daddr, |
39b46fc6 PM |
401 | sizeof(dst->addr.ip6.dst)); |
402 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) | |
0660e03f | 403 | memcpy(&dst->addr.ip6.src, &ipv6_hdr(skb)->saddr, |
39b46fc6 PM |
404 | sizeof(dst->addr.ip6.src)); |
405 | ||
406 | if (!(hinfo->cfg.mode & | |
407 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
408 | return 0; | |
409 | nexthdr = ipv6_find_hdr(skb, &protoff, -1, NULL); | |
410 | if (nexthdr < 0) | |
411 | return -1; | |
412 | break; | |
413 | #endif | |
414 | default: | |
415 | BUG(); | |
416 | return 0; | |
417 | } | |
418 | ||
419 | switch (nexthdr) { | |
420 | case IPPROTO_TCP: | |
421 | case IPPROTO_UDP: | |
a8d0f952 | 422 | case IPPROTO_UDPLITE: |
39b46fc6 PM |
423 | case IPPROTO_SCTP: |
424 | case IPPROTO_DCCP: | |
425 | ports = skb_header_pointer(skb, protoff, sizeof(_ports), | |
426 | &_ports); | |
427 | break; | |
428 | default: | |
429 | _ports[0] = _ports[1] = 0; | |
430 | ports = _ports; | |
431 | break; | |
432 | } | |
433 | if (!ports) | |
434 | return -1; | |
435 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SPT) | |
436 | dst->src_port = ports[0]; | |
437 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DPT) | |
438 | dst->dst_port = ports[1]; | |
439 | return 0; | |
1da177e4 LT |
440 | } |
441 | ||
1d93a9cb | 442 | static bool |
1da177e4 LT |
443 | hashlimit_match(const struct sk_buff *skb, |
444 | const struct net_device *in, | |
445 | const struct net_device *out, | |
c4986734 | 446 | const struct xt_match *match, |
1da177e4 LT |
447 | const void *matchinfo, |
448 | int offset, | |
2e4e6a17 | 449 | unsigned int protoff, |
cff533ac | 450 | bool *hotdrop) |
1da177e4 | 451 | { |
a47362a2 JE |
452 | const struct xt_hashlimit_info *r = |
453 | ((const struct xt_hashlimit_info *)matchinfo)->u.master; | |
39b46fc6 | 454 | struct xt_hashlimit_htable *hinfo = r->hinfo; |
1da177e4 LT |
455 | unsigned long now = jiffies; |
456 | struct dsthash_ent *dh; | |
457 | struct dsthash_dst dst; | |
458 | ||
39b46fc6 PM |
459 | if (hashlimit_init_dst(hinfo, &dst, skb, protoff) < 0) |
460 | goto hotdrop; | |
1da177e4 LT |
461 | |
462 | spin_lock_bh(&hinfo->lock); | |
39b46fc6 | 463 | dh = dsthash_find(hinfo, &dst); |
1da177e4 | 464 | if (!dh) { |
39b46fc6 | 465 | dh = dsthash_alloc_init(hinfo, &dst); |
1da177e4 | 466 | if (!dh) { |
1da177e4 | 467 | spin_unlock_bh(&hinfo->lock); |
39b46fc6 | 468 | goto hotdrop; |
1da177e4 LT |
469 | } |
470 | ||
471 | dh->expires = jiffies + msecs_to_jiffies(hinfo->cfg.expire); | |
1da177e4 | 472 | dh->rateinfo.prev = jiffies; |
39b46fc6 PM |
473 | dh->rateinfo.credit = user2credits(hinfo->cfg.avg * |
474 | hinfo->cfg.burst); | |
475 | dh->rateinfo.credit_cap = user2credits(hinfo->cfg.avg * | |
476 | hinfo->cfg.burst); | |
1da177e4 | 477 | dh->rateinfo.cost = user2credits(hinfo->cfg.avg); |
1c7628bd PM |
478 | } else { |
479 | /* update expiration timeout */ | |
480 | dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire); | |
481 | rateinfo_recalc(dh, now); | |
1da177e4 LT |
482 | } |
483 | ||
1da177e4 LT |
484 | if (dh->rateinfo.credit >= dh->rateinfo.cost) { |
485 | /* We're underlimit. */ | |
486 | dh->rateinfo.credit -= dh->rateinfo.cost; | |
487 | spin_unlock_bh(&hinfo->lock); | |
1d93a9cb | 488 | return true; |
1da177e4 LT |
489 | } |
490 | ||
601e68e1 | 491 | spin_unlock_bh(&hinfo->lock); |
1da177e4 LT |
492 | |
493 | /* default case: we're overlimit, thus don't match */ | |
1d93a9cb | 494 | return false; |
39b46fc6 PM |
495 | |
496 | hotdrop: | |
cff533ac | 497 | *hotdrop = true; |
1d93a9cb | 498 | return false; |
1da177e4 LT |
499 | } |
500 | ||
ccb79bdc | 501 | static bool |
1da177e4 | 502 | hashlimit_checkentry(const char *tablename, |
2e4e6a17 | 503 | const void *inf, |
c4986734 | 504 | const struct xt_match *match, |
1da177e4 | 505 | void *matchinfo, |
1da177e4 LT |
506 | unsigned int hook_mask) |
507 | { | |
39b46fc6 | 508 | struct xt_hashlimit_info *r = matchinfo; |
1da177e4 | 509 | |
1da177e4 | 510 | /* Check for overflow. */ |
39b46fc6 PM |
511 | if (r->cfg.burst == 0 || |
512 | user2credits(r->cfg.avg * r->cfg.burst) < user2credits(r->cfg.avg)) { | |
513 | printk(KERN_ERR "xt_hashlimit: overflow, try lower: %u/%u\n", | |
1da177e4 | 514 | r->cfg.avg, r->cfg.burst); |
ccb79bdc | 515 | return false; |
1da177e4 | 516 | } |
39b46fc6 PM |
517 | if (r->cfg.mode == 0 || |
518 | r->cfg.mode > (XT_HASHLIMIT_HASH_DPT | | |
519 | XT_HASHLIMIT_HASH_DIP | | |
520 | XT_HASHLIMIT_HASH_SIP | | |
521 | XT_HASHLIMIT_HASH_SPT)) | |
ccb79bdc | 522 | return false; |
1da177e4 | 523 | if (!r->cfg.gc_interval) |
ccb79bdc | 524 | return false; |
1da177e4 | 525 | if (!r->cfg.expire) |
ccb79bdc | 526 | return false; |
3ab72088 | 527 | if (r->name[sizeof(r->name) - 1] != '\0') |
ccb79bdc | 528 | return false; |
3ab72088 | 529 | |
1da177e4 | 530 | /* This is the best we've got: We cannot release and re-grab lock, |
39b46fc6 PM |
531 | * since checkentry() is called before x_tables.c grabs xt_mutex. |
532 | * We also cannot grab the hashtable spinlock, since htable_create will | |
1da177e4 LT |
533 | * call vmalloc, and that can sleep. And we cannot just re-search |
534 | * the list of htable's in htable_create(), since then we would | |
535 | * create duplicate proc files. -HW */ | |
14cc3e2b | 536 | mutex_lock(&hlimit_mutex); |
39b46fc6 PM |
537 | r->hinfo = htable_find_get(r->name, match->family); |
538 | if (!r->hinfo && htable_create(r, match->family) != 0) { | |
14cc3e2b | 539 | mutex_unlock(&hlimit_mutex); |
ccb79bdc | 540 | return false; |
1da177e4 | 541 | } |
14cc3e2b | 542 | mutex_unlock(&hlimit_mutex); |
1da177e4 LT |
543 | |
544 | /* Ugly hack: For SMP, we only want to use one set */ | |
545 | r->u.master = r; | |
ccb79bdc | 546 | return true; |
1da177e4 LT |
547 | } |
548 | ||
549 | static void | |
efa74165 | 550 | hashlimit_destroy(const struct xt_match *match, void *matchinfo) |
1da177e4 | 551 | { |
a47362a2 | 552 | const struct xt_hashlimit_info *r = matchinfo; |
1da177e4 LT |
553 | |
554 | htable_put(r->hinfo); | |
555 | } | |
556 | ||
127f15dd | 557 | #ifdef CONFIG_COMPAT |
39b46fc6 | 558 | struct compat_xt_hashlimit_info { |
127f15dd PM |
559 | char name[IFNAMSIZ]; |
560 | struct hashlimit_cfg cfg; | |
561 | compat_uptr_t hinfo; | |
562 | compat_uptr_t master; | |
563 | }; | |
564 | ||
565 | static void compat_from_user(void *dst, void *src) | |
566 | { | |
39b46fc6 | 567 | int off = offsetof(struct compat_xt_hashlimit_info, hinfo); |
127f15dd PM |
568 | |
569 | memcpy(dst, src, off); | |
39b46fc6 | 570 | memset(dst + off, 0, sizeof(struct compat_xt_hashlimit_info) - off); |
127f15dd PM |
571 | } |
572 | ||
573 | static int compat_to_user(void __user *dst, void *src) | |
574 | { | |
39b46fc6 | 575 | int off = offsetof(struct compat_xt_hashlimit_info, hinfo); |
127f15dd PM |
576 | |
577 | return copy_to_user(dst, src, off) ? -EFAULT : 0; | |
578 | } | |
579 | #endif | |
580 | ||
39b46fc6 PM |
581 | static struct xt_match xt_hashlimit[] = { |
582 | { | |
583 | .name = "hashlimit", | |
584 | .family = AF_INET, | |
585 | .match = hashlimit_match, | |
586 | .matchsize = sizeof(struct xt_hashlimit_info), | |
587 | #ifdef CONFIG_COMPAT | |
588 | .compatsize = sizeof(struct compat_xt_hashlimit_info), | |
589 | .compat_from_user = compat_from_user, | |
590 | .compat_to_user = compat_to_user, | |
591 | #endif | |
592 | .checkentry = hashlimit_checkentry, | |
593 | .destroy = hashlimit_destroy, | |
594 | .me = THIS_MODULE | |
595 | }, | |
596 | { | |
597 | .name = "hashlimit", | |
598 | .family = AF_INET6, | |
599 | .match = hashlimit_match, | |
600 | .matchsize = sizeof(struct xt_hashlimit_info), | |
127f15dd | 601 | #ifdef CONFIG_COMPAT |
39b46fc6 PM |
602 | .compatsize = sizeof(struct compat_xt_hashlimit_info), |
603 | .compat_from_user = compat_from_user, | |
604 | .compat_to_user = compat_to_user, | |
127f15dd | 605 | #endif |
39b46fc6 PM |
606 | .checkentry = hashlimit_checkentry, |
607 | .destroy = hashlimit_destroy, | |
608 | .me = THIS_MODULE | |
609 | }, | |
1da177e4 LT |
610 | }; |
611 | ||
612 | /* PROC stuff */ | |
1da177e4 LT |
613 | static void *dl_seq_start(struct seq_file *s, loff_t *pos) |
614 | { | |
615 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 616 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
617 | unsigned int *bucket; |
618 | ||
619 | spin_lock_bh(&htable->lock); | |
620 | if (*pos >= htable->cfg.size) | |
621 | return NULL; | |
622 | ||
623 | bucket = kmalloc(sizeof(unsigned int), GFP_ATOMIC); | |
624 | if (!bucket) | |
625 | return ERR_PTR(-ENOMEM); | |
626 | ||
627 | *bucket = *pos; | |
628 | return bucket; | |
629 | } | |
630 | ||
631 | static void *dl_seq_next(struct seq_file *s, void *v, loff_t *pos) | |
632 | { | |
633 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 634 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
635 | unsigned int *bucket = (unsigned int *)v; |
636 | ||
637 | *pos = ++(*bucket); | |
638 | if (*pos >= htable->cfg.size) { | |
639 | kfree(v); | |
640 | return NULL; | |
641 | } | |
642 | return bucket; | |
643 | } | |
644 | ||
645 | static void dl_seq_stop(struct seq_file *s, void *v) | |
646 | { | |
647 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 648 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
649 | unsigned int *bucket = (unsigned int *)v; |
650 | ||
651 | kfree(bucket); | |
1da177e4 LT |
652 | spin_unlock_bh(&htable->lock); |
653 | } | |
654 | ||
39b46fc6 PM |
655 | static int dl_seq_real_show(struct dsthash_ent *ent, int family, |
656 | struct seq_file *s) | |
1da177e4 LT |
657 | { |
658 | /* recalculate to show accurate numbers */ | |
659 | rateinfo_recalc(ent, jiffies); | |
660 | ||
39b46fc6 PM |
661 | switch (family) { |
662 | case AF_INET: | |
663 | return seq_printf(s, "%ld %u.%u.%u.%u:%u->" | |
664 | "%u.%u.%u.%u:%u %u %u %u\n", | |
665 | (long)(ent->expires - jiffies)/HZ, | |
666 | NIPQUAD(ent->dst.addr.ip.src), | |
667 | ntohs(ent->dst.src_port), | |
668 | NIPQUAD(ent->dst.addr.ip.dst), | |
669 | ntohs(ent->dst.dst_port), | |
670 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
671 | ent->rateinfo.cost); | |
672 | case AF_INET6: | |
673 | return seq_printf(s, "%ld " NIP6_FMT ":%u->" | |
674 | NIP6_FMT ":%u %u %u %u\n", | |
675 | (long)(ent->expires - jiffies)/HZ, | |
676 | NIP6(*(struct in6_addr *)&ent->dst.addr.ip6.src), | |
677 | ntohs(ent->dst.src_port), | |
678 | NIP6(*(struct in6_addr *)&ent->dst.addr.ip6.dst), | |
679 | ntohs(ent->dst.dst_port), | |
680 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
681 | ent->rateinfo.cost); | |
682 | default: | |
683 | BUG(); | |
684 | return 0; | |
685 | } | |
1da177e4 LT |
686 | } |
687 | ||
688 | static int dl_seq_show(struct seq_file *s, void *v) | |
689 | { | |
690 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 691 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
692 | unsigned int *bucket = (unsigned int *)v; |
693 | struct dsthash_ent *ent; | |
694 | struct hlist_node *pos; | |
695 | ||
39b46fc6 PM |
696 | if (!hlist_empty(&htable->hash[*bucket])) { |
697 | hlist_for_each_entry(ent, pos, &htable->hash[*bucket], node) | |
698 | if (dl_seq_real_show(ent, htable->family, s)) | |
1da177e4 | 699 | return 1; |
39b46fc6 | 700 | } |
1da177e4 LT |
701 | return 0; |
702 | } | |
703 | ||
704 | static struct seq_operations dl_seq_ops = { | |
705 | .start = dl_seq_start, | |
706 | .next = dl_seq_next, | |
707 | .stop = dl_seq_stop, | |
708 | .show = dl_seq_show | |
709 | }; | |
710 | ||
711 | static int dl_proc_open(struct inode *inode, struct file *file) | |
712 | { | |
713 | int ret = seq_open(file, &dl_seq_ops); | |
714 | ||
715 | if (!ret) { | |
716 | struct seq_file *sf = file->private_data; | |
717 | sf->private = PDE(inode); | |
718 | } | |
719 | return ret; | |
720 | } | |
721 | ||
da7071d7 | 722 | static const struct file_operations dl_file_ops = { |
1da177e4 LT |
723 | .owner = THIS_MODULE, |
724 | .open = dl_proc_open, | |
725 | .read = seq_read, | |
726 | .llseek = seq_lseek, | |
727 | .release = seq_release | |
728 | }; | |
729 | ||
39b46fc6 | 730 | static int __init xt_hashlimit_init(void) |
1da177e4 | 731 | { |
39b46fc6 | 732 | int err; |
1da177e4 | 733 | |
39b46fc6 PM |
734 | err = xt_register_matches(xt_hashlimit, ARRAY_SIZE(xt_hashlimit)); |
735 | if (err < 0) | |
736 | goto err1; | |
1da177e4 | 737 | |
39b46fc6 PM |
738 | err = -ENOMEM; |
739 | hashlimit_cachep = kmem_cache_create("xt_hashlimit", | |
740 | sizeof(struct dsthash_ent), 0, 0, | |
741 | NULL, NULL); | |
1da177e4 | 742 | if (!hashlimit_cachep) { |
39b46fc6 PM |
743 | printk(KERN_ERR "xt_hashlimit: unable to create slab cache\n"); |
744 | goto err2; | |
1da177e4 | 745 | } |
39b46fc6 PM |
746 | hashlimit_procdir4 = proc_mkdir("ipt_hashlimit", proc_net); |
747 | if (!hashlimit_procdir4) { | |
748 | printk(KERN_ERR "xt_hashlimit: unable to create proc dir " | |
749 | "entry\n"); | |
750 | goto err3; | |
1da177e4 | 751 | } |
39b46fc6 PM |
752 | hashlimit_procdir6 = proc_mkdir("ip6t_hashlimit", proc_net); |
753 | if (!hashlimit_procdir6) { | |
9c2440bb | 754 | printk(KERN_ERR "xt_hashlimit: unable to create proc dir " |
39b46fc6 PM |
755 | "entry\n"); |
756 | goto err4; | |
757 | } | |
758 | return 0; | |
759 | err4: | |
1da177e4 | 760 | remove_proc_entry("ipt_hashlimit", proc_net); |
39b46fc6 | 761 | err3: |
1da177e4 | 762 | kmem_cache_destroy(hashlimit_cachep); |
39b46fc6 PM |
763 | err2: |
764 | xt_unregister_matches(xt_hashlimit, ARRAY_SIZE(xt_hashlimit)); | |
765 | err1: | |
766 | return err; | |
1da177e4 | 767 | |
1da177e4 LT |
768 | } |
769 | ||
39b46fc6 | 770 | static void __exit xt_hashlimit_fini(void) |
1da177e4 | 771 | { |
39b46fc6 PM |
772 | remove_proc_entry("ipt_hashlimit", proc_net); |
773 | remove_proc_entry("ip6t_hashlimit", proc_net); | |
774 | kmem_cache_destroy(hashlimit_cachep); | |
775 | xt_unregister_matches(xt_hashlimit, ARRAY_SIZE(xt_hashlimit)); | |
1da177e4 LT |
776 | } |
777 | ||
39b46fc6 PM |
778 | module_init(xt_hashlimit_init); |
779 | module_exit(xt_hashlimit_fini); |