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> | |
7b21e09d | 23 | #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE) |
39b46fc6 | 24 | #include <linux/ipv6.h> |
193b23c5 | 25 | #include <net/ipv6.h> |
7b21e09d ED |
26 | #endif |
27 | ||
457c4cbc | 28 | #include <net/net_namespace.h> |
1da177e4 | 29 | |
39b46fc6 | 30 | #include <linux/netfilter/x_tables.h> |
1da177e4 | 31 | #include <linux/netfilter_ipv4/ip_tables.h> |
39b46fc6 PM |
32 | #include <linux/netfilter_ipv6/ip6_tables.h> |
33 | #include <linux/netfilter/xt_hashlimit.h> | |
14cc3e2b | 34 | #include <linux/mutex.h> |
1da177e4 LT |
35 | |
36 | MODULE_LICENSE("GPL"); | |
37 | MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); | |
2ae15b64 | 38 | MODULE_DESCRIPTION("Xtables: per hash-bucket rate-limit match"); |
39b46fc6 PM |
39 | MODULE_ALIAS("ipt_hashlimit"); |
40 | MODULE_ALIAS("ip6t_hashlimit"); | |
1da177e4 LT |
41 | |
42 | /* need to declare this at the top */ | |
39b46fc6 PM |
43 | static struct proc_dir_entry *hashlimit_procdir4; |
44 | static struct proc_dir_entry *hashlimit_procdir6; | |
da7071d7 | 45 | static const struct file_operations dl_file_ops; |
1da177e4 LT |
46 | |
47 | /* hash table crap */ | |
1da177e4 | 48 | struct dsthash_dst { |
39b46fc6 PM |
49 | union { |
50 | struct { | |
51 | __be32 src; | |
52 | __be32 dst; | |
53 | } ip; | |
7b21e09d | 54 | #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE) |
39b46fc6 PM |
55 | struct { |
56 | __be32 src[4]; | |
57 | __be32 dst[4]; | |
58 | } ip6; | |
7b21e09d | 59 | #endif |
39b46fc6 | 60 | } addr; |
6a19d614 AV |
61 | __be16 src_port; |
62 | __be16 dst_port; | |
1da177e4 LT |
63 | }; |
64 | ||
65 | struct dsthash_ent { | |
66 | /* static / read-only parts in the beginning */ | |
67 | struct hlist_node node; | |
68 | struct dsthash_dst dst; | |
69 | ||
70 | /* modified structure members in the end */ | |
71 | unsigned long expires; /* precalculated expiry time */ | |
72 | struct { | |
73 | unsigned long prev; /* last modification */ | |
74 | u_int32_t credit; | |
75 | u_int32_t credit_cap, cost; | |
76 | } rateinfo; | |
77 | }; | |
78 | ||
39b46fc6 | 79 | struct xt_hashlimit_htable { |
1da177e4 LT |
80 | struct hlist_node node; /* global list of all htables */ |
81 | atomic_t use; | |
39b46fc6 | 82 | int family; |
1da177e4 LT |
83 | |
84 | struct hashlimit_cfg cfg; /* config */ | |
85 | ||
86 | /* used internally */ | |
87 | spinlock_t lock; /* lock for list_head */ | |
88 | u_int32_t rnd; /* random seed for hash */ | |
bf0857ea | 89 | int rnd_initialized; |
39b46fc6 | 90 | unsigned int count; /* number entries in table */ |
1da177e4 | 91 | struct timer_list timer; /* timer for gc */ |
1da177e4 LT |
92 | |
93 | /* seq_file stuff */ | |
94 | struct proc_dir_entry *pde; | |
95 | ||
96 | struct hlist_head hash[0]; /* hashtable itself */ | |
97 | }; | |
98 | ||
e45b1be8 | 99 | static DEFINE_SPINLOCK(hashlimit_lock); /* protects htables list */ |
14cc3e2b | 100 | static DEFINE_MUTEX(hlimit_mutex); /* additional checkentry protection */ |
1da177e4 | 101 | static HLIST_HEAD(hashlimit_htables); |
e18b890b | 102 | static struct kmem_cache *hashlimit_cachep __read_mostly; |
1da177e4 | 103 | |
1d93a9cb | 104 | static inline bool dst_cmp(const struct dsthash_ent *ent, |
a47362a2 | 105 | const struct dsthash_dst *b) |
1da177e4 | 106 | { |
39b46fc6 | 107 | return !memcmp(&ent->dst, b, sizeof(ent->dst)); |
1da177e4 LT |
108 | } |
109 | ||
39b46fc6 PM |
110 | static u_int32_t |
111 | hash_dst(const struct xt_hashlimit_htable *ht, const struct dsthash_dst *dst) | |
1da177e4 | 112 | { |
e2f82ac3 ED |
113 | u_int32_t hash = jhash2((const u32 *)dst, |
114 | sizeof(*dst)/sizeof(u32), | |
115 | ht->rnd); | |
116 | /* | |
117 | * Instead of returning hash % ht->cfg.size (implying a divide) | |
118 | * we return the high 32 bits of the (hash * ht->cfg.size) that will | |
119 | * give results between [0 and cfg.size-1] and same hash distribution, | |
120 | * but using a multiply, less expensive than a divide | |
121 | */ | |
122 | return ((u64)hash * ht->cfg.size) >> 32; | |
1da177e4 LT |
123 | } |
124 | ||
39b46fc6 | 125 | static struct dsthash_ent * |
a47362a2 JE |
126 | dsthash_find(const struct xt_hashlimit_htable *ht, |
127 | const struct dsthash_dst *dst) | |
1da177e4 LT |
128 | { |
129 | struct dsthash_ent *ent; | |
130 | struct hlist_node *pos; | |
131 | u_int32_t hash = hash_dst(ht, dst); | |
132 | ||
39b46fc6 PM |
133 | if (!hlist_empty(&ht->hash[hash])) { |
134 | hlist_for_each_entry(ent, pos, &ht->hash[hash], node) | |
135 | if (dst_cmp(ent, dst)) | |
1da177e4 | 136 | return ent; |
39b46fc6 | 137 | } |
1da177e4 LT |
138 | return NULL; |
139 | } | |
140 | ||
141 | /* allocate dsthash_ent, initialize dst, put in htable and lock it */ | |
142 | static struct dsthash_ent * | |
a47362a2 JE |
143 | dsthash_alloc_init(struct xt_hashlimit_htable *ht, |
144 | const struct dsthash_dst *dst) | |
1da177e4 LT |
145 | { |
146 | struct dsthash_ent *ent; | |
147 | ||
148 | /* initialize hash with random val at the time we allocate | |
149 | * the first hashtable entry */ | |
bf0857ea | 150 | if (!ht->rnd_initialized) { |
1da177e4 | 151 | get_random_bytes(&ht->rnd, 4); |
bf0857ea PM |
152 | ht->rnd_initialized = 1; |
153 | } | |
1da177e4 | 154 | |
39b46fc6 | 155 | if (ht->cfg.max && ht->count >= ht->cfg.max) { |
1da177e4 LT |
156 | /* FIXME: do something. question is what.. */ |
157 | if (net_ratelimit()) | |
39b46fc6 PM |
158 | printk(KERN_WARNING |
159 | "xt_hashlimit: max count of %u reached\n", | |
1da177e4 LT |
160 | ht->cfg.max); |
161 | return NULL; | |
162 | } | |
163 | ||
164 | ent = kmem_cache_alloc(hashlimit_cachep, GFP_ATOMIC); | |
165 | if (!ent) { | |
166 | if (net_ratelimit()) | |
39b46fc6 PM |
167 | printk(KERN_ERR |
168 | "xt_hashlimit: can't allocate dsthash_ent\n"); | |
1da177e4 LT |
169 | return NULL; |
170 | } | |
39b46fc6 | 171 | memcpy(&ent->dst, dst, sizeof(ent->dst)); |
1da177e4 LT |
172 | |
173 | hlist_add_head(&ent->node, &ht->hash[hash_dst(ht, dst)]); | |
39b46fc6 | 174 | ht->count++; |
1da177e4 LT |
175 | return ent; |
176 | } | |
177 | ||
39b46fc6 PM |
178 | static inline void |
179 | dsthash_free(struct xt_hashlimit_htable *ht, struct dsthash_ent *ent) | |
1da177e4 LT |
180 | { |
181 | hlist_del(&ent->node); | |
182 | kmem_cache_free(hashlimit_cachep, ent); | |
39b46fc6 | 183 | ht->count--; |
1da177e4 LT |
184 | } |
185 | static void htable_gc(unsigned long htlong); | |
186 | ||
39b46fc6 | 187 | static int htable_create(struct xt_hashlimit_info *minfo, int family) |
1da177e4 | 188 | { |
39b46fc6 | 189 | struct xt_hashlimit_htable *hinfo; |
1da177e4 | 190 | unsigned int size; |
39b46fc6 | 191 | unsigned int i; |
1da177e4 LT |
192 | |
193 | if (minfo->cfg.size) | |
194 | size = minfo->cfg.size; | |
195 | else { | |
39b46fc6 PM |
196 | size = ((num_physpages << PAGE_SHIFT) / 16384) / |
197 | sizeof(struct list_head); | |
1da177e4 LT |
198 | if (num_physpages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
199 | size = 8192; | |
200 | if (size < 16) | |
201 | size = 16; | |
202 | } | |
203 | /* FIXME: don't use vmalloc() here or anywhere else -HW */ | |
39b46fc6 PM |
204 | hinfo = vmalloc(sizeof(struct xt_hashlimit_htable) + |
205 | sizeof(struct list_head) * size); | |
1da177e4 | 206 | if (!hinfo) { |
39b46fc6 | 207 | printk(KERN_ERR "xt_hashlimit: unable to create hashtable\n"); |
1da177e4 LT |
208 | return -1; |
209 | } | |
210 | minfo->hinfo = hinfo; | |
211 | ||
212 | /* copy match config into hashtable config */ | |
213 | memcpy(&hinfo->cfg, &minfo->cfg, sizeof(hinfo->cfg)); | |
214 | hinfo->cfg.size = size; | |
215 | if (!hinfo->cfg.max) | |
216 | hinfo->cfg.max = 8 * hinfo->cfg.size; | |
217 | else if (hinfo->cfg.max < hinfo->cfg.size) | |
218 | hinfo->cfg.max = hinfo->cfg.size; | |
219 | ||
220 | for (i = 0; i < hinfo->cfg.size; i++) | |
221 | INIT_HLIST_HEAD(&hinfo->hash[i]); | |
222 | ||
1da177e4 | 223 | atomic_set(&hinfo->use, 1); |
39b46fc6 PM |
224 | hinfo->count = 0; |
225 | hinfo->family = family; | |
bf0857ea | 226 | hinfo->rnd_initialized = 0; |
1da177e4 | 227 | spin_lock_init(&hinfo->lock); |
39b46fc6 PM |
228 | hinfo->pde = create_proc_entry(minfo->name, 0, |
229 | family == AF_INET ? hashlimit_procdir4 : | |
601e68e1 | 230 | hashlimit_procdir6); |
1da177e4 LT |
231 | if (!hinfo->pde) { |
232 | vfree(hinfo); | |
233 | return -1; | |
234 | } | |
235 | hinfo->pde->proc_fops = &dl_file_ops; | |
236 | hinfo->pde->data = hinfo; | |
237 | ||
e6f689db | 238 | setup_timer(&hinfo->timer, htable_gc, (unsigned long )hinfo); |
1da177e4 | 239 | hinfo->timer.expires = jiffies + msecs_to_jiffies(hinfo->cfg.gc_interval); |
1da177e4 LT |
240 | add_timer(&hinfo->timer); |
241 | ||
e45b1be8 | 242 | spin_lock_bh(&hashlimit_lock); |
1da177e4 | 243 | hlist_add_head(&hinfo->node, &hashlimit_htables); |
e45b1be8 | 244 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
245 | |
246 | return 0; | |
247 | } | |
248 | ||
a47362a2 JE |
249 | static bool select_all(const struct xt_hashlimit_htable *ht, |
250 | const struct dsthash_ent *he) | |
1da177e4 LT |
251 | { |
252 | return 1; | |
253 | } | |
254 | ||
a47362a2 JE |
255 | static bool select_gc(const struct xt_hashlimit_htable *ht, |
256 | const struct dsthash_ent *he) | |
1da177e4 | 257 | { |
cbebc51f | 258 | return time_after_eq(jiffies, he->expires); |
1da177e4 LT |
259 | } |
260 | ||
39b46fc6 | 261 | static void htable_selective_cleanup(struct xt_hashlimit_htable *ht, |
a47362a2 JE |
262 | bool (*select)(const struct xt_hashlimit_htable *ht, |
263 | const struct dsthash_ent *he)) | |
1da177e4 | 264 | { |
39b46fc6 | 265 | unsigned int i; |
1da177e4 LT |
266 | |
267 | /* lock hash table and iterate over it */ | |
268 | spin_lock_bh(&ht->lock); | |
269 | for (i = 0; i < ht->cfg.size; i++) { | |
270 | struct dsthash_ent *dh; | |
271 | struct hlist_node *pos, *n; | |
272 | hlist_for_each_entry_safe(dh, pos, n, &ht->hash[i], node) { | |
273 | if ((*select)(ht, dh)) | |
39b46fc6 | 274 | dsthash_free(ht, dh); |
1da177e4 LT |
275 | } |
276 | } | |
277 | spin_unlock_bh(&ht->lock); | |
278 | } | |
279 | ||
280 | /* hash table garbage collector, run by timer */ | |
281 | static void htable_gc(unsigned long htlong) | |
282 | { | |
39b46fc6 | 283 | struct xt_hashlimit_htable *ht = (struct xt_hashlimit_htable *)htlong; |
1da177e4 LT |
284 | |
285 | htable_selective_cleanup(ht, select_gc); | |
286 | ||
287 | /* re-add the timer accordingly */ | |
288 | ht->timer.expires = jiffies + msecs_to_jiffies(ht->cfg.gc_interval); | |
289 | add_timer(&ht->timer); | |
290 | } | |
291 | ||
39b46fc6 | 292 | static void htable_destroy(struct xt_hashlimit_htable *hinfo) |
1da177e4 LT |
293 | { |
294 | /* remove timer, if it is pending */ | |
295 | if (timer_pending(&hinfo->timer)) | |
296 | del_timer(&hinfo->timer); | |
297 | ||
298 | /* remove proc entry */ | |
39b46fc6 PM |
299 | remove_proc_entry(hinfo->pde->name, |
300 | hinfo->family == AF_INET ? hashlimit_procdir4 : | |
601e68e1 | 301 | hashlimit_procdir6); |
1da177e4 LT |
302 | htable_selective_cleanup(hinfo, select_all); |
303 | vfree(hinfo); | |
304 | } | |
305 | ||
a47362a2 JE |
306 | static struct xt_hashlimit_htable *htable_find_get(const char *name, |
307 | int family) | |
1da177e4 | 308 | { |
39b46fc6 | 309 | struct xt_hashlimit_htable *hinfo; |
1da177e4 LT |
310 | struct hlist_node *pos; |
311 | ||
e45b1be8 | 312 | spin_lock_bh(&hashlimit_lock); |
1da177e4 | 313 | hlist_for_each_entry(hinfo, pos, &hashlimit_htables, node) { |
39b46fc6 PM |
314 | if (!strcmp(name, hinfo->pde->name) && |
315 | hinfo->family == family) { | |
1da177e4 | 316 | atomic_inc(&hinfo->use); |
e45b1be8 | 317 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
318 | return hinfo; |
319 | } | |
320 | } | |
e45b1be8 | 321 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
322 | return NULL; |
323 | } | |
324 | ||
39b46fc6 | 325 | static void htable_put(struct xt_hashlimit_htable *hinfo) |
1da177e4 LT |
326 | { |
327 | if (atomic_dec_and_test(&hinfo->use)) { | |
e45b1be8 | 328 | spin_lock_bh(&hashlimit_lock); |
1da177e4 | 329 | hlist_del(&hinfo->node); |
e45b1be8 | 330 | spin_unlock_bh(&hashlimit_lock); |
1da177e4 LT |
331 | htable_destroy(hinfo); |
332 | } | |
333 | } | |
334 | ||
1da177e4 LT |
335 | /* The algorithm used is the Simple Token Bucket Filter (TBF) |
336 | * see net/sched/sch_tbf.c in the linux source tree | |
337 | */ | |
338 | ||
339 | /* Rusty: This is my (non-mathematically-inclined) understanding of | |
340 | this algorithm. The `average rate' in jiffies becomes your initial | |
341 | amount of credit `credit' and the most credit you can ever have | |
342 | `credit_cap'. The `peak rate' becomes the cost of passing the | |
343 | test, `cost'. | |
344 | ||
345 | `prev' tracks the last packet hit: you gain one credit per jiffy. | |
346 | If you get credit balance more than this, the extra credit is | |
347 | discarded. Every time the match passes, you lose `cost' credits; | |
348 | if you don't have that many, the test fails. | |
349 | ||
350 | See Alexey's formal explanation in net/sched/sch_tbf.c. | |
351 | ||
352 | To get the maximum range, we multiply by this factor (ie. you get N | |
353 | credits per jiffy). We want to allow a rate as low as 1 per day | |
354 | (slowest userspace tool allows), which means | |
355 | CREDITS_PER_JIFFY*HZ*60*60*24 < 2^32 ie. | |
356 | */ | |
357 | #define MAX_CPJ (0xFFFFFFFF / (HZ*60*60*24)) | |
358 | ||
359 | /* Repeated shift and or gives us all 1s, final shift and add 1 gives | |
360 | * us the power of 2 below the theoretical max, so GCC simply does a | |
361 | * shift. */ | |
362 | #define _POW2_BELOW2(x) ((x)|((x)>>1)) | |
363 | #define _POW2_BELOW4(x) (_POW2_BELOW2(x)|_POW2_BELOW2((x)>>2)) | |
364 | #define _POW2_BELOW8(x) (_POW2_BELOW4(x)|_POW2_BELOW4((x)>>4)) | |
365 | #define _POW2_BELOW16(x) (_POW2_BELOW8(x)|_POW2_BELOW8((x)>>8)) | |
366 | #define _POW2_BELOW32(x) (_POW2_BELOW16(x)|_POW2_BELOW16((x)>>16)) | |
367 | #define POW2_BELOW32(x) ((_POW2_BELOW32(x)>>1) + 1) | |
368 | ||
369 | #define CREDITS_PER_JIFFY POW2_BELOW32(MAX_CPJ) | |
370 | ||
371 | /* Precision saver. */ | |
372 | static inline u_int32_t | |
373 | user2credits(u_int32_t user) | |
374 | { | |
375 | /* If multiplying would overflow... */ | |
376 | if (user > 0xFFFFFFFF / (HZ*CREDITS_PER_JIFFY)) | |
377 | /* Divide first. */ | |
39b46fc6 | 378 | return (user / XT_HASHLIMIT_SCALE) * HZ * CREDITS_PER_JIFFY; |
1da177e4 | 379 | |
39b46fc6 | 380 | return (user * HZ * CREDITS_PER_JIFFY) / XT_HASHLIMIT_SCALE; |
1da177e4 LT |
381 | } |
382 | ||
383 | static inline void rateinfo_recalc(struct dsthash_ent *dh, unsigned long now) | |
384 | { | |
39b46fc6 | 385 | dh->rateinfo.credit += (now - dh->rateinfo.prev) * CREDITS_PER_JIFFY; |
1da177e4 LT |
386 | if (dh->rateinfo.credit > dh->rateinfo.credit_cap) |
387 | dh->rateinfo.credit = dh->rateinfo.credit_cap; | |
39b46fc6 PM |
388 | dh->rateinfo.prev = now; |
389 | } | |
390 | ||
391 | static int | |
a47362a2 JE |
392 | hashlimit_init_dst(const struct xt_hashlimit_htable *hinfo, |
393 | struct dsthash_dst *dst, | |
39b46fc6 PM |
394 | const struct sk_buff *skb, unsigned int protoff) |
395 | { | |
396 | __be16 _ports[2], *ports; | |
193b23c5 | 397 | u8 nexthdr; |
39b46fc6 PM |
398 | |
399 | memset(dst, 0, sizeof(*dst)); | |
400 | ||
401 | switch (hinfo->family) { | |
402 | case AF_INET: | |
403 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) | |
eddc9ec5 | 404 | dst->addr.ip.dst = ip_hdr(skb)->daddr; |
39b46fc6 | 405 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) |
eddc9ec5 | 406 | dst->addr.ip.src = ip_hdr(skb)->saddr; |
39b46fc6 PM |
407 | |
408 | if (!(hinfo->cfg.mode & | |
409 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
410 | return 0; | |
eddc9ec5 | 411 | nexthdr = ip_hdr(skb)->protocol; |
39b46fc6 | 412 | break; |
02dba025 | 413 | #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE) |
39b46fc6 PM |
414 | case AF_INET6: |
415 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) | |
0660e03f | 416 | memcpy(&dst->addr.ip6.dst, &ipv6_hdr(skb)->daddr, |
39b46fc6 PM |
417 | sizeof(dst->addr.ip6.dst)); |
418 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) | |
0660e03f | 419 | memcpy(&dst->addr.ip6.src, &ipv6_hdr(skb)->saddr, |
39b46fc6 PM |
420 | sizeof(dst->addr.ip6.src)); |
421 | ||
422 | if (!(hinfo->cfg.mode & | |
423 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
424 | return 0; | |
193b23c5 PM |
425 | nexthdr = ipv6_hdr(skb)->nexthdr; |
426 | protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr); | |
427 | if ((int)protoff < 0) | |
39b46fc6 PM |
428 | return -1; |
429 | break; | |
430 | #endif | |
431 | default: | |
432 | BUG(); | |
433 | return 0; | |
434 | } | |
435 | ||
436 | switch (nexthdr) { | |
437 | case IPPROTO_TCP: | |
438 | case IPPROTO_UDP: | |
a8d0f952 | 439 | case IPPROTO_UDPLITE: |
39b46fc6 PM |
440 | case IPPROTO_SCTP: |
441 | case IPPROTO_DCCP: | |
442 | ports = skb_header_pointer(skb, protoff, sizeof(_ports), | |
443 | &_ports); | |
444 | break; | |
445 | default: | |
446 | _ports[0] = _ports[1] = 0; | |
447 | ports = _ports; | |
448 | break; | |
449 | } | |
450 | if (!ports) | |
451 | return -1; | |
452 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SPT) | |
453 | dst->src_port = ports[0]; | |
454 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DPT) | |
455 | dst->dst_port = ports[1]; | |
456 | return 0; | |
1da177e4 LT |
457 | } |
458 | ||
1d93a9cb | 459 | static bool |
d3c5ee6d JE |
460 | hashlimit_mt(const struct sk_buff *skb, const struct net_device *in, |
461 | const struct net_device *out, const struct xt_match *match, | |
462 | const void *matchinfo, int offset, unsigned int protoff, | |
463 | bool *hotdrop) | |
1da177e4 | 464 | { |
a47362a2 JE |
465 | const struct xt_hashlimit_info *r = |
466 | ((const struct xt_hashlimit_info *)matchinfo)->u.master; | |
39b46fc6 | 467 | struct xt_hashlimit_htable *hinfo = r->hinfo; |
1da177e4 LT |
468 | unsigned long now = jiffies; |
469 | struct dsthash_ent *dh; | |
470 | struct dsthash_dst dst; | |
471 | ||
39b46fc6 PM |
472 | if (hashlimit_init_dst(hinfo, &dst, skb, protoff) < 0) |
473 | goto hotdrop; | |
1da177e4 LT |
474 | |
475 | spin_lock_bh(&hinfo->lock); | |
39b46fc6 | 476 | dh = dsthash_find(hinfo, &dst); |
1da177e4 | 477 | if (!dh) { |
39b46fc6 | 478 | dh = dsthash_alloc_init(hinfo, &dst); |
1da177e4 | 479 | if (!dh) { |
1da177e4 | 480 | spin_unlock_bh(&hinfo->lock); |
39b46fc6 | 481 | goto hotdrop; |
1da177e4 LT |
482 | } |
483 | ||
484 | dh->expires = jiffies + msecs_to_jiffies(hinfo->cfg.expire); | |
1da177e4 | 485 | dh->rateinfo.prev = jiffies; |
39b46fc6 PM |
486 | dh->rateinfo.credit = user2credits(hinfo->cfg.avg * |
487 | hinfo->cfg.burst); | |
488 | dh->rateinfo.credit_cap = user2credits(hinfo->cfg.avg * | |
489 | hinfo->cfg.burst); | |
1da177e4 | 490 | dh->rateinfo.cost = user2credits(hinfo->cfg.avg); |
1c7628bd PM |
491 | } else { |
492 | /* update expiration timeout */ | |
493 | dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire); | |
494 | rateinfo_recalc(dh, now); | |
1da177e4 LT |
495 | } |
496 | ||
1da177e4 LT |
497 | if (dh->rateinfo.credit >= dh->rateinfo.cost) { |
498 | /* We're underlimit. */ | |
499 | dh->rateinfo.credit -= dh->rateinfo.cost; | |
500 | spin_unlock_bh(&hinfo->lock); | |
1d93a9cb | 501 | return true; |
1da177e4 LT |
502 | } |
503 | ||
601e68e1 | 504 | spin_unlock_bh(&hinfo->lock); |
1da177e4 LT |
505 | |
506 | /* default case: we're overlimit, thus don't match */ | |
1d93a9cb | 507 | return false; |
39b46fc6 PM |
508 | |
509 | hotdrop: | |
cff533ac | 510 | *hotdrop = true; |
1d93a9cb | 511 | return false; |
1da177e4 LT |
512 | } |
513 | ||
ccb79bdc | 514 | static bool |
d3c5ee6d JE |
515 | hashlimit_mt_check(const char *tablename, const void *inf, |
516 | const struct xt_match *match, void *matchinfo, | |
517 | unsigned int hook_mask) | |
1da177e4 | 518 | { |
39b46fc6 | 519 | struct xt_hashlimit_info *r = matchinfo; |
1da177e4 | 520 | |
1da177e4 | 521 | /* Check for overflow. */ |
39b46fc6 PM |
522 | if (r->cfg.burst == 0 || |
523 | user2credits(r->cfg.avg * r->cfg.burst) < user2credits(r->cfg.avg)) { | |
524 | printk(KERN_ERR "xt_hashlimit: overflow, try lower: %u/%u\n", | |
1da177e4 | 525 | r->cfg.avg, r->cfg.burst); |
ccb79bdc | 526 | return false; |
1da177e4 | 527 | } |
39b46fc6 PM |
528 | if (r->cfg.mode == 0 || |
529 | r->cfg.mode > (XT_HASHLIMIT_HASH_DPT | | |
530 | XT_HASHLIMIT_HASH_DIP | | |
531 | XT_HASHLIMIT_HASH_SIP | | |
532 | XT_HASHLIMIT_HASH_SPT)) | |
ccb79bdc | 533 | return false; |
1da177e4 | 534 | if (!r->cfg.gc_interval) |
ccb79bdc | 535 | return false; |
1da177e4 | 536 | if (!r->cfg.expire) |
ccb79bdc | 537 | return false; |
3ab72088 | 538 | if (r->name[sizeof(r->name) - 1] != '\0') |
ccb79bdc | 539 | return false; |
3ab72088 | 540 | |
1da177e4 | 541 | /* This is the best we've got: We cannot release and re-grab lock, |
39b46fc6 PM |
542 | * since checkentry() is called before x_tables.c grabs xt_mutex. |
543 | * We also cannot grab the hashtable spinlock, since htable_create will | |
1da177e4 LT |
544 | * call vmalloc, and that can sleep. And we cannot just re-search |
545 | * the list of htable's in htable_create(), since then we would | |
546 | * create duplicate proc files. -HW */ | |
14cc3e2b | 547 | mutex_lock(&hlimit_mutex); |
39b46fc6 PM |
548 | r->hinfo = htable_find_get(r->name, match->family); |
549 | if (!r->hinfo && htable_create(r, match->family) != 0) { | |
14cc3e2b | 550 | mutex_unlock(&hlimit_mutex); |
ccb79bdc | 551 | return false; |
1da177e4 | 552 | } |
14cc3e2b | 553 | mutex_unlock(&hlimit_mutex); |
1da177e4 LT |
554 | |
555 | /* Ugly hack: For SMP, we only want to use one set */ | |
556 | r->u.master = r; | |
ccb79bdc | 557 | return true; |
1da177e4 LT |
558 | } |
559 | ||
560 | static void | |
d3c5ee6d | 561 | hashlimit_mt_destroy(const struct xt_match *match, void *matchinfo) |
1da177e4 | 562 | { |
a47362a2 | 563 | const struct xt_hashlimit_info *r = matchinfo; |
1da177e4 LT |
564 | |
565 | htable_put(r->hinfo); | |
566 | } | |
567 | ||
127f15dd | 568 | #ifdef CONFIG_COMPAT |
39b46fc6 | 569 | struct compat_xt_hashlimit_info { |
127f15dd PM |
570 | char name[IFNAMSIZ]; |
571 | struct hashlimit_cfg cfg; | |
572 | compat_uptr_t hinfo; | |
573 | compat_uptr_t master; | |
574 | }; | |
575 | ||
d3c5ee6d | 576 | static void hashlimit_mt_compat_from_user(void *dst, void *src) |
127f15dd | 577 | { |
39b46fc6 | 578 | int off = offsetof(struct compat_xt_hashlimit_info, hinfo); |
127f15dd PM |
579 | |
580 | memcpy(dst, src, off); | |
39b46fc6 | 581 | memset(dst + off, 0, sizeof(struct compat_xt_hashlimit_info) - off); |
127f15dd PM |
582 | } |
583 | ||
d3c5ee6d | 584 | static int hashlimit_mt_compat_to_user(void __user *dst, void *src) |
127f15dd | 585 | { |
39b46fc6 | 586 | int off = offsetof(struct compat_xt_hashlimit_info, hinfo); |
127f15dd PM |
587 | |
588 | return copy_to_user(dst, src, off) ? -EFAULT : 0; | |
589 | } | |
590 | #endif | |
591 | ||
d3c5ee6d | 592 | static struct xt_match hashlimit_mt_reg[] __read_mostly = { |
39b46fc6 PM |
593 | { |
594 | .name = "hashlimit", | |
595 | .family = AF_INET, | |
d3c5ee6d | 596 | .match = hashlimit_mt, |
39b46fc6 PM |
597 | .matchsize = sizeof(struct xt_hashlimit_info), |
598 | #ifdef CONFIG_COMPAT | |
599 | .compatsize = sizeof(struct compat_xt_hashlimit_info), | |
d3c5ee6d JE |
600 | .compat_from_user = hashlimit_mt_compat_from_user, |
601 | .compat_to_user = hashlimit_mt_compat_to_user, | |
39b46fc6 | 602 | #endif |
d3c5ee6d JE |
603 | .checkentry = hashlimit_mt_check, |
604 | .destroy = hashlimit_mt_destroy, | |
39b46fc6 PM |
605 | .me = THIS_MODULE |
606 | }, | |
7b21e09d | 607 | #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE) |
39b46fc6 PM |
608 | { |
609 | .name = "hashlimit", | |
610 | .family = AF_INET6, | |
d3c5ee6d | 611 | .match = hashlimit_mt, |
39b46fc6 | 612 | .matchsize = sizeof(struct xt_hashlimit_info), |
127f15dd | 613 | #ifdef CONFIG_COMPAT |
39b46fc6 | 614 | .compatsize = sizeof(struct compat_xt_hashlimit_info), |
d3c5ee6d JE |
615 | .compat_from_user = hashlimit_mt_compat_from_user, |
616 | .compat_to_user = hashlimit_mt_compat_to_user, | |
127f15dd | 617 | #endif |
d3c5ee6d JE |
618 | .checkentry = hashlimit_mt_check, |
619 | .destroy = hashlimit_mt_destroy, | |
39b46fc6 PM |
620 | .me = THIS_MODULE |
621 | }, | |
7b21e09d | 622 | #endif |
1da177e4 LT |
623 | }; |
624 | ||
625 | /* PROC stuff */ | |
1da177e4 LT |
626 | static void *dl_seq_start(struct seq_file *s, loff_t *pos) |
627 | { | |
628 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 629 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
630 | unsigned int *bucket; |
631 | ||
632 | spin_lock_bh(&htable->lock); | |
633 | if (*pos >= htable->cfg.size) | |
634 | return NULL; | |
635 | ||
636 | bucket = kmalloc(sizeof(unsigned int), GFP_ATOMIC); | |
637 | if (!bucket) | |
638 | return ERR_PTR(-ENOMEM); | |
639 | ||
640 | *bucket = *pos; | |
641 | return bucket; | |
642 | } | |
643 | ||
644 | static void *dl_seq_next(struct seq_file *s, void *v, loff_t *pos) | |
645 | { | |
646 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 647 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
648 | unsigned int *bucket = (unsigned int *)v; |
649 | ||
650 | *pos = ++(*bucket); | |
651 | if (*pos >= htable->cfg.size) { | |
652 | kfree(v); | |
653 | return NULL; | |
654 | } | |
655 | return bucket; | |
656 | } | |
657 | ||
658 | static void dl_seq_stop(struct seq_file *s, void *v) | |
659 | { | |
660 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 661 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
662 | unsigned int *bucket = (unsigned int *)v; |
663 | ||
664 | kfree(bucket); | |
1da177e4 LT |
665 | spin_unlock_bh(&htable->lock); |
666 | } | |
667 | ||
39b46fc6 PM |
668 | static int dl_seq_real_show(struct dsthash_ent *ent, int family, |
669 | struct seq_file *s) | |
1da177e4 LT |
670 | { |
671 | /* recalculate to show accurate numbers */ | |
672 | rateinfo_recalc(ent, jiffies); | |
673 | ||
39b46fc6 PM |
674 | switch (family) { |
675 | case AF_INET: | |
676 | return seq_printf(s, "%ld %u.%u.%u.%u:%u->" | |
677 | "%u.%u.%u.%u:%u %u %u %u\n", | |
678 | (long)(ent->expires - jiffies)/HZ, | |
679 | NIPQUAD(ent->dst.addr.ip.src), | |
680 | ntohs(ent->dst.src_port), | |
681 | NIPQUAD(ent->dst.addr.ip.dst), | |
682 | ntohs(ent->dst.dst_port), | |
683 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
684 | ent->rateinfo.cost); | |
7b21e09d | 685 | #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE) |
39b46fc6 PM |
686 | case AF_INET6: |
687 | return seq_printf(s, "%ld " NIP6_FMT ":%u->" | |
688 | NIP6_FMT ":%u %u %u %u\n", | |
689 | (long)(ent->expires - jiffies)/HZ, | |
690 | NIP6(*(struct in6_addr *)&ent->dst.addr.ip6.src), | |
691 | ntohs(ent->dst.src_port), | |
692 | NIP6(*(struct in6_addr *)&ent->dst.addr.ip6.dst), | |
693 | ntohs(ent->dst.dst_port), | |
694 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
695 | ent->rateinfo.cost); | |
7b21e09d | 696 | #endif |
39b46fc6 PM |
697 | default: |
698 | BUG(); | |
699 | return 0; | |
700 | } | |
1da177e4 LT |
701 | } |
702 | ||
703 | static int dl_seq_show(struct seq_file *s, void *v) | |
704 | { | |
705 | struct proc_dir_entry *pde = s->private; | |
39b46fc6 | 706 | struct xt_hashlimit_htable *htable = pde->data; |
1da177e4 LT |
707 | unsigned int *bucket = (unsigned int *)v; |
708 | struct dsthash_ent *ent; | |
709 | struct hlist_node *pos; | |
710 | ||
39b46fc6 PM |
711 | if (!hlist_empty(&htable->hash[*bucket])) { |
712 | hlist_for_each_entry(ent, pos, &htable->hash[*bucket], node) | |
713 | if (dl_seq_real_show(ent, htable->family, s)) | |
1da177e4 | 714 | return 1; |
39b46fc6 | 715 | } |
1da177e4 LT |
716 | return 0; |
717 | } | |
718 | ||
56b3d975 | 719 | static const struct seq_operations dl_seq_ops = { |
1da177e4 LT |
720 | .start = dl_seq_start, |
721 | .next = dl_seq_next, | |
722 | .stop = dl_seq_stop, | |
723 | .show = dl_seq_show | |
724 | }; | |
725 | ||
726 | static int dl_proc_open(struct inode *inode, struct file *file) | |
727 | { | |
728 | int ret = seq_open(file, &dl_seq_ops); | |
729 | ||
730 | if (!ret) { | |
731 | struct seq_file *sf = file->private_data; | |
732 | sf->private = PDE(inode); | |
733 | } | |
734 | return ret; | |
735 | } | |
736 | ||
da7071d7 | 737 | static const struct file_operations dl_file_ops = { |
1da177e4 LT |
738 | .owner = THIS_MODULE, |
739 | .open = dl_proc_open, | |
740 | .read = seq_read, | |
741 | .llseek = seq_lseek, | |
742 | .release = seq_release | |
743 | }; | |
744 | ||
d3c5ee6d | 745 | static int __init hashlimit_mt_init(void) |
1da177e4 | 746 | { |
39b46fc6 | 747 | int err; |
1da177e4 | 748 | |
d3c5ee6d JE |
749 | err = xt_register_matches(hashlimit_mt_reg, |
750 | ARRAY_SIZE(hashlimit_mt_reg)); | |
39b46fc6 PM |
751 | if (err < 0) |
752 | goto err1; | |
1da177e4 | 753 | |
39b46fc6 PM |
754 | err = -ENOMEM; |
755 | hashlimit_cachep = kmem_cache_create("xt_hashlimit", | |
756 | sizeof(struct dsthash_ent), 0, 0, | |
20c2df83 | 757 | NULL); |
1da177e4 | 758 | if (!hashlimit_cachep) { |
39b46fc6 PM |
759 | printk(KERN_ERR "xt_hashlimit: unable to create slab cache\n"); |
760 | goto err2; | |
1da177e4 | 761 | } |
457c4cbc | 762 | hashlimit_procdir4 = proc_mkdir("ipt_hashlimit", init_net.proc_net); |
39b46fc6 PM |
763 | if (!hashlimit_procdir4) { |
764 | printk(KERN_ERR "xt_hashlimit: unable to create proc dir " | |
765 | "entry\n"); | |
766 | goto err3; | |
1da177e4 | 767 | } |
7b21e09d ED |
768 | err = 0; |
769 | #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE) | |
457c4cbc | 770 | hashlimit_procdir6 = proc_mkdir("ip6t_hashlimit", init_net.proc_net); |
39b46fc6 | 771 | if (!hashlimit_procdir6) { |
9c2440bb | 772 | printk(KERN_ERR "xt_hashlimit: unable to create proc dir " |
39b46fc6 | 773 | "entry\n"); |
7b21e09d | 774 | err = -ENOMEM; |
39b46fc6 | 775 | } |
7b21e09d ED |
776 | #endif |
777 | if (!err) | |
778 | return 0; | |
457c4cbc | 779 | remove_proc_entry("ipt_hashlimit", init_net.proc_net); |
39b46fc6 | 780 | err3: |
1da177e4 | 781 | kmem_cache_destroy(hashlimit_cachep); |
39b46fc6 | 782 | err2: |
d3c5ee6d | 783 | xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg)); |
39b46fc6 PM |
784 | err1: |
785 | return err; | |
1da177e4 | 786 | |
1da177e4 LT |
787 | } |
788 | ||
d3c5ee6d | 789 | static void __exit hashlimit_mt_exit(void) |
1da177e4 | 790 | { |
457c4cbc | 791 | remove_proc_entry("ipt_hashlimit", init_net.proc_net); |
7b21e09d | 792 | #if defined(CONFIG_IP6_NF_IPTABLES) || defined(CONFIG_IP6_NF_IPTABLES_MODULE) |
457c4cbc | 793 | remove_proc_entry("ip6t_hashlimit", init_net.proc_net); |
7b21e09d | 794 | #endif |
39b46fc6 | 795 | kmem_cache_destroy(hashlimit_cachep); |
d3c5ee6d | 796 | xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg)); |
1da177e4 LT |
797 | } |
798 | ||
d3c5ee6d JE |
799 | module_init(hashlimit_mt_init); |
800 | module_exit(hashlimit_mt_exit); |