Commit | Line | Data |
---|---|---|
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> | |
c0cd1156 | 24 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
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 | |
38 | MODULE_LICENSE("GPL"); | |
39 | MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); | |
408ffaa4 | 40 | MODULE_AUTHOR("Jan Engelhardt <jengelh@medozas.de>"); |
2ae15b64 | 41 | MODULE_DESCRIPTION("Xtables: per hash-bucket rate-limit match"); |
39b46fc6 PM |
42 | MODULE_ALIAS("ipt_hashlimit"); |
43 | MODULE_ALIAS("ip6t_hashlimit"); | |
1da177e4 | 44 | |
e89fc3f1 AD |
45 | struct hashlimit_net { |
46 | struct hlist_head htables; | |
47 | struct proc_dir_entry *ipt_hashlimit; | |
48 | struct proc_dir_entry *ip6t_hashlimit; | |
49 | }; | |
50 | ||
51 | static int hashlimit_net_id; | |
52 | static 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 | 58 | static const struct file_operations dl_file_ops; |
1da177e4 LT |
59 | |
60 | /* hash table crap */ | |
1da177e4 | 61 | struct dsthash_dst { |
39b46fc6 PM |
62 | union { |
63 | struct { | |
64 | __be32 src; | |
65 | __be32 dst; | |
66 | } ip; | |
c0cd1156 | 67 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
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 | ||
78 | struct 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 | 94 | struct 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 | 115 | static DEFINE_MUTEX(hashlimit_mutex); /* protects htables list */ |
e18b890b | 116 | static struct kmem_cache *hashlimit_cachep __read_mostly; |
1da177e4 | 117 | |
1d93a9cb | 118 | static 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 |
124 | static u_int32_t |
125 | hash_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 | 139 | static struct dsthash_ent * |
a47362a2 JE |
140 | dsthash_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 */ | |
158 | static struct dsthash_ent * | |
a47362a2 JE |
159 | dsthash_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 |
191 | static 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 |
198 | static inline void |
199 | dsthash_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 | } |
205 | static void htable_gc(unsigned long htlong); | |
206 | ||
e89fc3f1 AD |
207 | static 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 |
268 | static 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 |
274 | static 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 | 280 | static 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 */ | |
300 | static 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 | 311 | static 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 |
327 | static 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 | 345 | static 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. */ | |
392 | static inline u_int32_t | |
393 | user2credits(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 | ||
403 | static 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 |
411 | static inline __be32 maskl(__be32 a, unsigned int l) |
412 | { | |
1b9b70ea | 413 | return l ? htonl(ntohl(a) & ~0 << (32 - l)) : 0; |
09e410de JE |
414 | } |
415 | ||
c0cd1156 | 416 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
09e410de JE |
417 | static 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 | 441 | static int |
a47362a2 JE |
442 | hashlimit_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; |
75f2811c | 448 | __be16 frag_off; |
aca071c1 | 449 | int poff; |
39b46fc6 PM |
450 | |
451 | memset(dst, 0, sizeof(*dst)); | |
452 | ||
453 | switch (hinfo->family) { | |
ee999d8b | 454 | case NFPROTO_IPV4: |
39b46fc6 | 455 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) |
09e410de JE |
456 | dst->ip.dst = maskl(ip_hdr(skb)->daddr, |
457 | hinfo->cfg.dstmask); | |
39b46fc6 | 458 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) |
09e410de JE |
459 | dst->ip.src = maskl(ip_hdr(skb)->saddr, |
460 | hinfo->cfg.srcmask); | |
39b46fc6 PM |
461 | |
462 | if (!(hinfo->cfg.mode & | |
463 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
464 | return 0; | |
eddc9ec5 | 465 | nexthdr = ip_hdr(skb)->protocol; |
39b46fc6 | 466 | break; |
c0cd1156 | 467 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
ee999d8b | 468 | case NFPROTO_IPV6: |
09e410de JE |
469 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) { |
470 | memcpy(&dst->ip6.dst, &ipv6_hdr(skb)->daddr, | |
471 | sizeof(dst->ip6.dst)); | |
472 | hashlimit_ipv6_mask(dst->ip6.dst, hinfo->cfg.dstmask); | |
473 | } | |
474 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) { | |
475 | memcpy(&dst->ip6.src, &ipv6_hdr(skb)->saddr, | |
476 | sizeof(dst->ip6.src)); | |
477 | hashlimit_ipv6_mask(dst->ip6.src, hinfo->cfg.srcmask); | |
478 | } | |
39b46fc6 PM |
479 | |
480 | if (!(hinfo->cfg.mode & | |
481 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
482 | return 0; | |
193b23c5 | 483 | nexthdr = ipv6_hdr(skb)->nexthdr; |
75f2811c | 484 | protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr, &frag_off); |
193b23c5 | 485 | if ((int)protoff < 0) |
39b46fc6 PM |
486 | return -1; |
487 | break; | |
488 | #endif | |
489 | default: | |
490 | BUG(); | |
491 | return 0; | |
492 | } | |
493 | ||
aca071c1 CG |
494 | poff = proto_ports_offset(nexthdr); |
495 | if (poff >= 0) { | |
496 | ports = skb_header_pointer(skb, protoff + poff, sizeof(_ports), | |
39b46fc6 | 497 | &_ports); |
aca071c1 | 498 | } else { |
39b46fc6 PM |
499 | _ports[0] = _ports[1] = 0; |
500 | ports = _ports; | |
39b46fc6 PM |
501 | } |
502 | if (!ports) | |
503 | return -1; | |
504 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SPT) | |
505 | dst->src_port = ports[0]; | |
506 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DPT) | |
507 | dst->dst_port = ports[1]; | |
508 | return 0; | |
1da177e4 LT |
509 | } |
510 | ||
ccb79bdc | 511 | static bool |
62fc8051 | 512 | hashlimit_mt(const struct sk_buff *skb, struct xt_action_param *par) |
09e410de | 513 | { |
f7108a20 | 514 | const struct xt_hashlimit_mtinfo1 *info = par->matchinfo; |
09e410de JE |
515 | struct xt_hashlimit_htable *hinfo = info->hinfo; |
516 | unsigned long now = jiffies; | |
517 | struct dsthash_ent *dh; | |
518 | struct dsthash_dst dst; | |
519 | ||
f7108a20 | 520 | if (hashlimit_init_dst(hinfo, &dst, skb, par->thoff) < 0) |
09e410de JE |
521 | goto hotdrop; |
522 | ||
02e4eb75 | 523 | rcu_read_lock_bh(); |
09e410de JE |
524 | dh = dsthash_find(hinfo, &dst); |
525 | if (dh == NULL) { | |
526 | dh = dsthash_alloc_init(hinfo, &dst); | |
527 | if (dh == NULL) { | |
02e4eb75 | 528 | rcu_read_unlock_bh(); |
09e410de JE |
529 | goto hotdrop; |
530 | } | |
09e410de JE |
531 | dh->expires = jiffies + msecs_to_jiffies(hinfo->cfg.expire); |
532 | dh->rateinfo.prev = jiffies; | |
533 | dh->rateinfo.credit = user2credits(hinfo->cfg.avg * | |
534 | hinfo->cfg.burst); | |
535 | dh->rateinfo.credit_cap = user2credits(hinfo->cfg.avg * | |
536 | hinfo->cfg.burst); | |
537 | dh->rateinfo.cost = user2credits(hinfo->cfg.avg); | |
538 | } else { | |
539 | /* update expiration timeout */ | |
540 | dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire); | |
541 | rateinfo_recalc(dh, now); | |
542 | } | |
543 | ||
544 | if (dh->rateinfo.credit >= dh->rateinfo.cost) { | |
545 | /* below the limit */ | |
546 | dh->rateinfo.credit -= dh->rateinfo.cost; | |
02e4eb75 ED |
547 | spin_unlock(&dh->lock); |
548 | rcu_read_unlock_bh(); | |
09e410de JE |
549 | return !(info->cfg.mode & XT_HASHLIMIT_INVERT); |
550 | } | |
551 | ||
02e4eb75 ED |
552 | spin_unlock(&dh->lock); |
553 | rcu_read_unlock_bh(); | |
09e410de JE |
554 | /* default match is underlimit - so over the limit, we need to invert */ |
555 | return info->cfg.mode & XT_HASHLIMIT_INVERT; | |
556 | ||
557 | hotdrop: | |
b4ba2611 | 558 | par->hotdrop = true; |
09e410de JE |
559 | return false; |
560 | } | |
561 | ||
b0f38452 | 562 | static int hashlimit_mt_check(const struct xt_mtchk_param *par) |
09e410de | 563 | { |
e89fc3f1 | 564 | struct net *net = par->net; |
9b4fce7a | 565 | struct xt_hashlimit_mtinfo1 *info = par->matchinfo; |
4a5a5c73 | 566 | int ret; |
09e410de JE |
567 | |
568 | /* Check for overflow. */ | |
569 | if (info->cfg.burst == 0 || | |
570 | user2credits(info->cfg.avg * info->cfg.burst) < | |
571 | user2credits(info->cfg.avg)) { | |
8bee4bad JE |
572 | pr_info("overflow, try lower: %u/%u\n", |
573 | info->cfg.avg, info->cfg.burst); | |
4a5a5c73 | 574 | return -ERANGE; |
09e410de JE |
575 | } |
576 | if (info->cfg.gc_interval == 0 || info->cfg.expire == 0) | |
bd414ee6 | 577 | return -EINVAL; |
09e410de | 578 | if (info->name[sizeof(info->name)-1] != '\0') |
bd414ee6 | 579 | return -EINVAL; |
aa5fa318 | 580 | if (par->family == NFPROTO_IPV4) { |
09e410de | 581 | if (info->cfg.srcmask > 32 || info->cfg.dstmask > 32) |
bd414ee6 | 582 | return -EINVAL; |
09e410de JE |
583 | } else { |
584 | if (info->cfg.srcmask > 128 || info->cfg.dstmask > 128) | |
bd414ee6 | 585 | return -EINVAL; |
09e410de JE |
586 | } |
587 | ||
2eff25c1 | 588 | mutex_lock(&hashlimit_mutex); |
aa5fa318 | 589 | info->hinfo = htable_find_get(net, info->name, par->family); |
4a5a5c73 JE |
590 | if (info->hinfo == NULL) { |
591 | ret = htable_create(net, info, par->family); | |
592 | if (ret < 0) { | |
593 | mutex_unlock(&hashlimit_mutex); | |
594 | return ret; | |
595 | } | |
09e410de | 596 | } |
2eff25c1 | 597 | mutex_unlock(&hashlimit_mutex); |
bd414ee6 | 598 | return 0; |
09e410de JE |
599 | } |
600 | ||
6be3d859 | 601 | static void hashlimit_mt_destroy(const struct xt_mtdtor_param *par) |
09e410de | 602 | { |
6be3d859 | 603 | const struct xt_hashlimit_mtinfo1 *info = par->matchinfo; |
09e410de JE |
604 | |
605 | htable_put(info->hinfo); | |
606 | } | |
607 | ||
d3c5ee6d | 608 | static struct xt_match hashlimit_mt_reg[] __read_mostly = { |
09e410de JE |
609 | { |
610 | .name = "hashlimit", | |
611 | .revision = 1, | |
ee999d8b | 612 | .family = NFPROTO_IPV4, |
09e410de JE |
613 | .match = hashlimit_mt, |
614 | .matchsize = sizeof(struct xt_hashlimit_mtinfo1), | |
615 | .checkentry = hashlimit_mt_check, | |
616 | .destroy = hashlimit_mt_destroy, | |
617 | .me = THIS_MODULE, | |
618 | }, | |
c0cd1156 | 619 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
09e410de JE |
620 | { |
621 | .name = "hashlimit", | |
622 | .revision = 1, | |
ee999d8b | 623 | .family = NFPROTO_IPV6, |
09e410de JE |
624 | .match = hashlimit_mt, |
625 | .matchsize = sizeof(struct xt_hashlimit_mtinfo1), | |
626 | .checkentry = hashlimit_mt_check, | |
627 | .destroy = hashlimit_mt_destroy, | |
628 | .me = THIS_MODULE, | |
629 | }, | |
7b21e09d | 630 | #endif |
1da177e4 LT |
631 | }; |
632 | ||
633 | /* PROC stuff */ | |
1da177e4 | 634 | static void *dl_seq_start(struct seq_file *s, loff_t *pos) |
f4f6fb71 | 635 | __acquires(htable->lock) |
1da177e4 | 636 | { |
a1004d8e | 637 | struct xt_hashlimit_htable *htable = s->private; |
1da177e4 LT |
638 | unsigned int *bucket; |
639 | ||
640 | spin_lock_bh(&htable->lock); | |
641 | if (*pos >= htable->cfg.size) | |
642 | return NULL; | |
643 | ||
644 | bucket = kmalloc(sizeof(unsigned int), GFP_ATOMIC); | |
645 | if (!bucket) | |
646 | return ERR_PTR(-ENOMEM); | |
647 | ||
648 | *bucket = *pos; | |
649 | return bucket; | |
650 | } | |
651 | ||
652 | static void *dl_seq_next(struct seq_file *s, void *v, loff_t *pos) | |
653 | { | |
a1004d8e | 654 | struct xt_hashlimit_htable *htable = s->private; |
1da177e4 LT |
655 | unsigned int *bucket = (unsigned int *)v; |
656 | ||
657 | *pos = ++(*bucket); | |
658 | if (*pos >= htable->cfg.size) { | |
659 | kfree(v); | |
660 | return NULL; | |
661 | } | |
662 | return bucket; | |
663 | } | |
664 | ||
665 | static void dl_seq_stop(struct seq_file *s, void *v) | |
f4f6fb71 | 666 | __releases(htable->lock) |
1da177e4 | 667 | { |
a1004d8e | 668 | struct xt_hashlimit_htable *htable = s->private; |
1da177e4 LT |
669 | unsigned int *bucket = (unsigned int *)v; |
670 | ||
55e0d7cf ED |
671 | if (!IS_ERR(bucket)) |
672 | kfree(bucket); | |
1da177e4 LT |
673 | spin_unlock_bh(&htable->lock); |
674 | } | |
675 | ||
76108cea | 676 | static int dl_seq_real_show(struct dsthash_ent *ent, u_int8_t family, |
39b46fc6 | 677 | struct seq_file *s) |
1da177e4 | 678 | { |
02e4eb75 ED |
679 | int res; |
680 | ||
681 | spin_lock(&ent->lock); | |
1da177e4 LT |
682 | /* recalculate to show accurate numbers */ |
683 | rateinfo_recalc(ent, jiffies); | |
684 | ||
39b46fc6 | 685 | switch (family) { |
ee999d8b | 686 | case NFPROTO_IPV4: |
02e4eb75 | 687 | res = seq_printf(s, "%ld %pI4:%u->%pI4:%u %u %u %u\n", |
39b46fc6 | 688 | (long)(ent->expires - jiffies)/HZ, |
14d5e834 | 689 | &ent->dst.ip.src, |
39b46fc6 | 690 | ntohs(ent->dst.src_port), |
14d5e834 | 691 | &ent->dst.ip.dst, |
39b46fc6 PM |
692 | ntohs(ent->dst.dst_port), |
693 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
694 | ent->rateinfo.cost); | |
02e4eb75 | 695 | break; |
c0cd1156 | 696 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
ee999d8b | 697 | case NFPROTO_IPV6: |
02e4eb75 | 698 | res = seq_printf(s, "%ld %pI6:%u->%pI6:%u %u %u %u\n", |
39b46fc6 | 699 | (long)(ent->expires - jiffies)/HZ, |
38ff4fa4 | 700 | &ent->dst.ip6.src, |
39b46fc6 | 701 | ntohs(ent->dst.src_port), |
38ff4fa4 | 702 | &ent->dst.ip6.dst, |
39b46fc6 PM |
703 | ntohs(ent->dst.dst_port), |
704 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
705 | ent->rateinfo.cost); | |
02e4eb75 | 706 | break; |
7b21e09d | 707 | #endif |
39b46fc6 PM |
708 | default: |
709 | BUG(); | |
02e4eb75 | 710 | res = 0; |
39b46fc6 | 711 | } |
02e4eb75 ED |
712 | spin_unlock(&ent->lock); |
713 | return res; | |
1da177e4 LT |
714 | } |
715 | ||
716 | static int dl_seq_show(struct seq_file *s, void *v) | |
717 | { | |
a1004d8e | 718 | struct xt_hashlimit_htable *htable = s->private; |
1da177e4 LT |
719 | unsigned int *bucket = (unsigned int *)v; |
720 | struct dsthash_ent *ent; | |
721 | struct hlist_node *pos; | |
722 | ||
39b46fc6 PM |
723 | if (!hlist_empty(&htable->hash[*bucket])) { |
724 | hlist_for_each_entry(ent, pos, &htable->hash[*bucket], node) | |
725 | if (dl_seq_real_show(ent, htable->family, s)) | |
683a04ce | 726 | return -1; |
39b46fc6 | 727 | } |
1da177e4 LT |
728 | return 0; |
729 | } | |
730 | ||
56b3d975 | 731 | static const struct seq_operations dl_seq_ops = { |
1da177e4 LT |
732 | .start = dl_seq_start, |
733 | .next = dl_seq_next, | |
734 | .stop = dl_seq_stop, | |
735 | .show = dl_seq_show | |
736 | }; | |
737 | ||
738 | static int dl_proc_open(struct inode *inode, struct file *file) | |
739 | { | |
740 | int ret = seq_open(file, &dl_seq_ops); | |
741 | ||
742 | if (!ret) { | |
743 | struct seq_file *sf = file->private_data; | |
a1004d8e | 744 | sf->private = PDE(inode)->data; |
1da177e4 LT |
745 | } |
746 | return ret; | |
747 | } | |
748 | ||
da7071d7 | 749 | static const struct file_operations dl_file_ops = { |
1da177e4 LT |
750 | .owner = THIS_MODULE, |
751 | .open = dl_proc_open, | |
752 | .read = seq_read, | |
753 | .llseek = seq_lseek, | |
754 | .release = seq_release | |
755 | }; | |
756 | ||
e89fc3f1 AD |
757 | static int __net_init hashlimit_proc_net_init(struct net *net) |
758 | { | |
759 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); | |
760 | ||
761 | hashlimit_net->ipt_hashlimit = proc_mkdir("ipt_hashlimit", net->proc_net); | |
762 | if (!hashlimit_net->ipt_hashlimit) | |
763 | return -ENOMEM; | |
c0cd1156 | 764 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
e89fc3f1 AD |
765 | hashlimit_net->ip6t_hashlimit = proc_mkdir("ip6t_hashlimit", net->proc_net); |
766 | if (!hashlimit_net->ip6t_hashlimit) { | |
767 | proc_net_remove(net, "ipt_hashlimit"); | |
768 | return -ENOMEM; | |
769 | } | |
770 | #endif | |
771 | return 0; | |
772 | } | |
773 | ||
774 | static void __net_exit hashlimit_proc_net_exit(struct net *net) | |
775 | { | |
776 | proc_net_remove(net, "ipt_hashlimit"); | |
c0cd1156 | 777 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
e89fc3f1 AD |
778 | proc_net_remove(net, "ip6t_hashlimit"); |
779 | #endif | |
780 | } | |
781 | ||
782 | static int __net_init hashlimit_net_init(struct net *net) | |
783 | { | |
784 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); | |
785 | ||
786 | INIT_HLIST_HEAD(&hashlimit_net->htables); | |
787 | return hashlimit_proc_net_init(net); | |
788 | } | |
789 | ||
790 | static void __net_exit hashlimit_net_exit(struct net *net) | |
791 | { | |
792 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); | |
793 | ||
794 | BUG_ON(!hlist_empty(&hashlimit_net->htables)); | |
795 | hashlimit_proc_net_exit(net); | |
796 | } | |
797 | ||
798 | static struct pernet_operations hashlimit_net_ops = { | |
799 | .init = hashlimit_net_init, | |
800 | .exit = hashlimit_net_exit, | |
801 | .id = &hashlimit_net_id, | |
802 | .size = sizeof(struct hashlimit_net), | |
803 | }; | |
804 | ||
d3c5ee6d | 805 | static int __init hashlimit_mt_init(void) |
1da177e4 | 806 | { |
39b46fc6 | 807 | int err; |
1da177e4 | 808 | |
e89fc3f1 AD |
809 | err = register_pernet_subsys(&hashlimit_net_ops); |
810 | if (err < 0) | |
811 | return err; | |
d3c5ee6d JE |
812 | err = xt_register_matches(hashlimit_mt_reg, |
813 | ARRAY_SIZE(hashlimit_mt_reg)); | |
39b46fc6 PM |
814 | if (err < 0) |
815 | goto err1; | |
1da177e4 | 816 | |
39b46fc6 PM |
817 | err = -ENOMEM; |
818 | hashlimit_cachep = kmem_cache_create("xt_hashlimit", | |
819 | sizeof(struct dsthash_ent), 0, 0, | |
20c2df83 | 820 | NULL); |
1da177e4 | 821 | if (!hashlimit_cachep) { |
8bee4bad | 822 | pr_warning("unable to create slab cache\n"); |
39b46fc6 | 823 | goto err2; |
1da177e4 | 824 | } |
e89fc3f1 AD |
825 | return 0; |
826 | ||
39b46fc6 | 827 | err2: |
d3c5ee6d | 828 | xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg)); |
39b46fc6 | 829 | err1: |
e89fc3f1 | 830 | unregister_pernet_subsys(&hashlimit_net_ops); |
39b46fc6 | 831 | return err; |
1da177e4 | 832 | |
1da177e4 LT |
833 | } |
834 | ||
d3c5ee6d | 835 | static void __exit hashlimit_mt_exit(void) |
1da177e4 | 836 | { |
d3c5ee6d | 837 | xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg)); |
e89fc3f1 | 838 | unregister_pernet_subsys(&hashlimit_net_ops); |
02e4eb75 ED |
839 | |
840 | rcu_barrier_bh(); | |
841 | kmem_cache_destroy(hashlimit_cachep); | |
1da177e4 LT |
842 | } |
843 | ||
d3c5ee6d JE |
844 | module_init(hashlimit_mt_init); |
845 | module_exit(hashlimit_mt_exit); |