Commit | Line | Data |
---|---|---|
09c434b8 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
09e410de JE |
2 | /* |
3 | * xt_hashlimit - Netfilter module to limit the number of packets per time | |
3ad2f3fb | 4 | * separately for each hashbucket (sourceip/sourceport/dstip/dstport) |
1da177e4 | 5 | * |
09e410de | 6 | * (C) 2003-2004 by Harald Welte <laforge@netfilter.org> |
f229f6ce | 7 | * (C) 2006-2012 Patrick McHardy <kaber@trash.net> |
09e410de | 8 | * Copyright © CC Computer Consultants GmbH, 2007 - 2008 |
1da177e4 LT |
9 | * |
10 | * Development of this code was funded by Astaro AG, http://www.astaro.com/ | |
1da177e4 | 11 | */ |
8bee4bad | 12 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
1da177e4 | 13 | #include <linux/module.h> |
1da177e4 LT |
14 | #include <linux/spinlock.h> |
15 | #include <linux/random.h> | |
16 | #include <linux/jhash.h> | |
17 | #include <linux/slab.h> | |
18 | #include <linux/vmalloc.h> | |
1da177e4 LT |
19 | #include <linux/proc_fs.h> |
20 | #include <linux/seq_file.h> | |
21 | #include <linux/list.h> | |
39b46fc6 | 22 | #include <linux/skbuff.h> |
d7fe0f24 | 23 | #include <linux/mm.h> |
39b46fc6 PM |
24 | #include <linux/in.h> |
25 | #include <linux/ip.h> | |
c0cd1156 | 26 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
39b46fc6 | 27 | #include <linux/ipv6.h> |
193b23c5 | 28 | #include <net/ipv6.h> |
7b21e09d ED |
29 | #endif |
30 | ||
457c4cbc | 31 | #include <net/net_namespace.h> |
e89fc3f1 | 32 | #include <net/netns/generic.h> |
1da177e4 | 33 | |
39b46fc6 | 34 | #include <linux/netfilter/x_tables.h> |
1da177e4 | 35 | #include <linux/netfilter_ipv4/ip_tables.h> |
39b46fc6 PM |
36 | #include <linux/netfilter_ipv6/ip6_tables.h> |
37 | #include <linux/netfilter/xt_hashlimit.h> | |
14cc3e2b | 38 | #include <linux/mutex.h> |
90c4ae4e | 39 | #include <linux/kernel.h> |
1da177e4 LT |
40 | |
41 | MODULE_LICENSE("GPL"); | |
42 | MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); | |
408ffaa4 | 43 | MODULE_AUTHOR("Jan Engelhardt <jengelh@medozas.de>"); |
2ae15b64 | 44 | MODULE_DESCRIPTION("Xtables: per hash-bucket rate-limit match"); |
39b46fc6 PM |
45 | MODULE_ALIAS("ipt_hashlimit"); |
46 | MODULE_ALIAS("ip6t_hashlimit"); | |
1da177e4 | 47 | |
e89fc3f1 AD |
48 | struct hashlimit_net { |
49 | struct hlist_head htables; | |
50 | struct proc_dir_entry *ipt_hashlimit; | |
51 | struct proc_dir_entry *ip6t_hashlimit; | |
52 | }; | |
53 | ||
c7d03a00 | 54 | static unsigned int hashlimit_net_id; |
e89fc3f1 AD |
55 | static inline struct hashlimit_net *hashlimit_pernet(struct net *net) |
56 | { | |
57 | return net_generic(net, hashlimit_net_id); | |
58 | } | |
59 | ||
1da177e4 | 60 | /* need to declare this at the top */ |
fd24de89 CH |
61 | static const struct seq_operations dl_seq_ops_v2; |
62 | static const struct seq_operations dl_seq_ops_v1; | |
63 | static const struct seq_operations dl_seq_ops; | |
1da177e4 LT |
64 | |
65 | /* hash table crap */ | |
1da177e4 | 66 | struct dsthash_dst { |
39b46fc6 PM |
67 | union { |
68 | struct { | |
69 | __be32 src; | |
70 | __be32 dst; | |
71 | } ip; | |
c0cd1156 | 72 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
39b46fc6 PM |
73 | struct { |
74 | __be32 src[4]; | |
75 | __be32 dst[4]; | |
76 | } ip6; | |
7b21e09d | 77 | #endif |
09e410de | 78 | }; |
6a19d614 AV |
79 | __be16 src_port; |
80 | __be16 dst_port; | |
1da177e4 LT |
81 | }; |
82 | ||
83 | struct dsthash_ent { | |
84 | /* static / read-only parts in the beginning */ | |
85 | struct hlist_node node; | |
86 | struct dsthash_dst dst; | |
87 | ||
88 | /* modified structure members in the end */ | |
02e4eb75 | 89 | spinlock_t lock; |
1da177e4 LT |
90 | unsigned long expires; /* precalculated expiry time */ |
91 | struct { | |
92 | unsigned long prev; /* last modification */ | |
bea74641 VP |
93 | union { |
94 | struct { | |
95 | u_int64_t credit; | |
96 | u_int64_t credit_cap; | |
97 | u_int64_t cost; | |
98 | }; | |
99 | struct { | |
100 | u_int32_t interval, prev_window; | |
101 | u_int64_t current_rate; | |
102 | u_int64_t rate; | |
103 | int64_t burst; | |
104 | }; | |
105 | }; | |
1da177e4 | 106 | } rateinfo; |
02e4eb75 | 107 | struct rcu_head rcu; |
1da177e4 LT |
108 | }; |
109 | ||
39b46fc6 | 110 | struct xt_hashlimit_htable { |
1da177e4 | 111 | struct hlist_node node; /* global list of all htables */ |
2eff25c1 | 112 | int use; |
76108cea | 113 | u_int8_t family; |
89bc7a0f | 114 | bool rnd_initialized; |
1da177e4 | 115 | |
bea74641 | 116 | struct hashlimit_cfg3 cfg; /* config */ |
1da177e4 LT |
117 | |
118 | /* used internally */ | |
119 | spinlock_t lock; /* lock for list_head */ | |
120 | u_int32_t rnd; /* random seed for hash */ | |
39b46fc6 | 121 | unsigned int count; /* number entries in table */ |
7bd8490e | 122 | struct delayed_work gc_work; |
1da177e4 LT |
123 | |
124 | /* seq_file stuff */ | |
125 | struct proc_dir_entry *pde; | |
14b872f0 | 126 | const char *name; |
e89fc3f1 | 127 | struct net *net; |
1da177e4 LT |
128 | |
129 | struct hlist_head hash[0]; /* hashtable itself */ | |
130 | }; | |
131 | ||
11d5f157 | 132 | static int |
bea74641 | 133 | cfg_copy(struct hashlimit_cfg3 *to, const void *from, int revision) |
11d5f157 VP |
134 | { |
135 | if (revision == 1) { | |
bea74641 | 136 | struct hashlimit_cfg1 *cfg = (struct hashlimit_cfg1 *)from; |
11d5f157 VP |
137 | |
138 | to->mode = cfg->mode; | |
139 | to->avg = cfg->avg; | |
140 | to->burst = cfg->burst; | |
141 | to->size = cfg->size; | |
142 | to->max = cfg->max; | |
143 | to->gc_interval = cfg->gc_interval; | |
144 | to->expire = cfg->expire; | |
145 | to->srcmask = cfg->srcmask; | |
146 | to->dstmask = cfg->dstmask; | |
147 | } else if (revision == 2) { | |
bea74641 VP |
148 | struct hashlimit_cfg2 *cfg = (struct hashlimit_cfg2 *)from; |
149 | ||
150 | to->mode = cfg->mode; | |
151 | to->avg = cfg->avg; | |
152 | to->burst = cfg->burst; | |
153 | to->size = cfg->size; | |
154 | to->max = cfg->max; | |
155 | to->gc_interval = cfg->gc_interval; | |
156 | to->expire = cfg->expire; | |
157 | to->srcmask = cfg->srcmask; | |
158 | to->dstmask = cfg->dstmask; | |
159 | } else if (revision == 3) { | |
160 | memcpy(to, from, sizeof(struct hashlimit_cfg3)); | |
11d5f157 VP |
161 | } else { |
162 | return -EINVAL; | |
163 | } | |
164 | ||
165 | return 0; | |
166 | } | |
167 | ||
2eff25c1 | 168 | static DEFINE_MUTEX(hashlimit_mutex); /* protects htables list */ |
e18b890b | 169 | static struct kmem_cache *hashlimit_cachep __read_mostly; |
1da177e4 | 170 | |
1d93a9cb | 171 | static inline bool dst_cmp(const struct dsthash_ent *ent, |
a47362a2 | 172 | const struct dsthash_dst *b) |
1da177e4 | 173 | { |
39b46fc6 | 174 | return !memcmp(&ent->dst, b, sizeof(ent->dst)); |
1da177e4 LT |
175 | } |
176 | ||
39b46fc6 PM |
177 | static u_int32_t |
178 | hash_dst(const struct xt_hashlimit_htable *ht, const struct dsthash_dst *dst) | |
1da177e4 | 179 | { |
e2f82ac3 ED |
180 | u_int32_t hash = jhash2((const u32 *)dst, |
181 | sizeof(*dst)/sizeof(u32), | |
182 | ht->rnd); | |
183 | /* | |
184 | * Instead of returning hash % ht->cfg.size (implying a divide) | |
185 | * we return the high 32 bits of the (hash * ht->cfg.size) that will | |
186 | * give results between [0 and cfg.size-1] and same hash distribution, | |
187 | * but using a multiply, less expensive than a divide | |
188 | */ | |
8fc54f68 | 189 | return reciprocal_scale(hash, ht->cfg.size); |
1da177e4 LT |
190 | } |
191 | ||
39b46fc6 | 192 | static struct dsthash_ent * |
a47362a2 JE |
193 | dsthash_find(const struct xt_hashlimit_htable *ht, |
194 | const struct dsthash_dst *dst) | |
1da177e4 LT |
195 | { |
196 | struct dsthash_ent *ent; | |
1da177e4 LT |
197 | u_int32_t hash = hash_dst(ht, dst); |
198 | ||
39b46fc6 | 199 | if (!hlist_empty(&ht->hash[hash])) { |
b67bfe0d | 200 | hlist_for_each_entry_rcu(ent, &ht->hash[hash], node) |
02e4eb75 ED |
201 | if (dst_cmp(ent, dst)) { |
202 | spin_lock(&ent->lock); | |
1da177e4 | 203 | return ent; |
02e4eb75 | 204 | } |
39b46fc6 | 205 | } |
1da177e4 LT |
206 | return NULL; |
207 | } | |
208 | ||
209 | /* allocate dsthash_ent, initialize dst, put in htable and lock it */ | |
210 | static struct dsthash_ent * | |
a47362a2 | 211 | dsthash_alloc_init(struct xt_hashlimit_htable *ht, |
09181842 | 212 | const struct dsthash_dst *dst, bool *race) |
1da177e4 LT |
213 | { |
214 | struct dsthash_ent *ent; | |
215 | ||
02e4eb75 | 216 | spin_lock(&ht->lock); |
09181842 PNA |
217 | |
218 | /* Two or more packets may race to create the same entry in the | |
219 | * hashtable, double check if this packet lost race. | |
220 | */ | |
221 | ent = dsthash_find(ht, dst); | |
222 | if (ent != NULL) { | |
223 | spin_unlock(&ht->lock); | |
224 | *race = true; | |
225 | return ent; | |
226 | } | |
227 | ||
1da177e4 LT |
228 | /* initialize hash with random val at the time we allocate |
229 | * the first hashtable entry */ | |
02e4eb75 | 230 | if (unlikely(!ht->rnd_initialized)) { |
af07d241 | 231 | get_random_bytes(&ht->rnd, sizeof(ht->rnd)); |
89bc7a0f | 232 | ht->rnd_initialized = true; |
bf0857ea | 233 | } |
1da177e4 | 234 | |
39b46fc6 | 235 | if (ht->cfg.max && ht->count >= ht->cfg.max) { |
1da177e4 | 236 | /* FIXME: do something. question is what.. */ |
e87cc472 | 237 | net_err_ratelimited("max count of %u reached\n", ht->cfg.max); |
02e4eb75 ED |
238 | ent = NULL; |
239 | } else | |
240 | ent = kmem_cache_alloc(hashlimit_cachep, GFP_ATOMIC); | |
0a9ee813 | 241 | if (ent) { |
02e4eb75 ED |
242 | memcpy(&ent->dst, dst, sizeof(ent->dst)); |
243 | spin_lock_init(&ent->lock); | |
1da177e4 | 244 | |
02e4eb75 ED |
245 | spin_lock(&ent->lock); |
246 | hlist_add_head_rcu(&ent->node, &ht->hash[hash_dst(ht, dst)]); | |
247 | ht->count++; | |
248 | } | |
249 | spin_unlock(&ht->lock); | |
1da177e4 LT |
250 | return ent; |
251 | } | |
252 | ||
02e4eb75 ED |
253 | static void dsthash_free_rcu(struct rcu_head *head) |
254 | { | |
255 | struct dsthash_ent *ent = container_of(head, struct dsthash_ent, rcu); | |
256 | ||
257 | kmem_cache_free(hashlimit_cachep, ent); | |
258 | } | |
259 | ||
39b46fc6 PM |
260 | static inline void |
261 | dsthash_free(struct xt_hashlimit_htable *ht, struct dsthash_ent *ent) | |
1da177e4 | 262 | { |
02e4eb75 | 263 | hlist_del_rcu(&ent->node); |
c8d1da40 | 264 | call_rcu(&ent->rcu, dsthash_free_rcu); |
39b46fc6 | 265 | ht->count--; |
1da177e4 | 266 | } |
7bd8490e | 267 | static void htable_gc(struct work_struct *work); |
1da177e4 | 268 | |
bea74641 | 269 | static int htable_create(struct net *net, struct hashlimit_cfg3 *cfg, |
11d5f157 VP |
270 | const char *name, u_int8_t family, |
271 | struct xt_hashlimit_htable **out_hinfo, | |
272 | int revision) | |
09e410de | 273 | { |
e89fc3f1 | 274 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); |
09e410de | 275 | struct xt_hashlimit_htable *hinfo; |
fd24de89 | 276 | const struct seq_operations *ops; |
11d5f157 | 277 | unsigned int size, i; |
ca79b0c2 | 278 | unsigned long nr_pages = totalram_pages(); |
11d5f157 | 279 | int ret; |
09e410de | 280 | |
11d5f157 VP |
281 | if (cfg->size) { |
282 | size = cfg->size; | |
09e410de | 283 | } else { |
3d6357de | 284 | size = (nr_pages << PAGE_SHIFT) / 16384 / |
05d0eae7 | 285 | sizeof(struct hlist_head); |
3d6357de | 286 | if (nr_pages > 1024 * 1024 * 1024 / PAGE_SIZE) |
09e410de JE |
287 | size = 8192; |
288 | if (size < 16) | |
289 | size = 16; | |
290 | } | |
291 | /* FIXME: don't use vmalloc() here or anywhere else -HW */ | |
eabb4782 | 292 | hinfo = vmalloc(struct_size(hinfo, hash, size)); |
85bc3f38 | 293 | if (hinfo == NULL) |
4a5a5c73 | 294 | return -ENOMEM; |
11d5f157 | 295 | *out_hinfo = hinfo; |
09e410de JE |
296 | |
297 | /* copy match config into hashtable config */ | |
bea74641 | 298 | ret = cfg_copy(&hinfo->cfg, (void *)cfg, 3); |
b4e955e9 TY |
299 | if (ret) { |
300 | vfree(hinfo); | |
11d5f157 | 301 | return ret; |
b4e955e9 | 302 | } |
11d5f157 | 303 | |
09e410de JE |
304 | hinfo->cfg.size = size; |
305 | if (hinfo->cfg.max == 0) | |
306 | hinfo->cfg.max = 8 * hinfo->cfg.size; | |
307 | else if (hinfo->cfg.max < hinfo->cfg.size) | |
308 | hinfo->cfg.max = hinfo->cfg.size; | |
309 | ||
310 | for (i = 0; i < hinfo->cfg.size; i++) | |
311 | INIT_HLIST_HEAD(&hinfo->hash[i]); | |
312 | ||
2eff25c1 | 313 | hinfo->use = 1; |
09e410de JE |
314 | hinfo->count = 0; |
315 | hinfo->family = family; | |
89bc7a0f | 316 | hinfo->rnd_initialized = false; |
11d5f157 | 317 | hinfo->name = kstrdup(name, GFP_KERNEL); |
14b872f0 AV |
318 | if (!hinfo->name) { |
319 | vfree(hinfo); | |
320 | return -ENOMEM; | |
321 | } | |
09e410de JE |
322 | spin_lock_init(&hinfo->lock); |
323 | ||
bea74641 VP |
324 | switch (revision) { |
325 | case 1: | |
fd24de89 | 326 | ops = &dl_seq_ops_v1; |
bea74641 VP |
327 | break; |
328 | case 2: | |
fd24de89 | 329 | ops = &dl_seq_ops_v2; |
bea74641 VP |
330 | break; |
331 | default: | |
fd24de89 | 332 | ops = &dl_seq_ops; |
bea74641 VP |
333 | } |
334 | ||
fd24de89 | 335 | hinfo->pde = proc_create_seq_data(name, 0, |
ee999d8b | 336 | (family == NFPROTO_IPV4) ? |
e89fc3f1 | 337 | hashlimit_net->ipt_hashlimit : hashlimit_net->ip6t_hashlimit, |
fd24de89 | 338 | ops, hinfo); |
09e410de | 339 | if (hinfo->pde == NULL) { |
14b872f0 | 340 | kfree(hinfo->name); |
09e410de | 341 | vfree(hinfo); |
4a5a5c73 | 342 | return -ENOMEM; |
09e410de | 343 | } |
e89fc3f1 | 344 | hinfo->net = net; |
09e410de | 345 | |
7bd8490e ED |
346 | INIT_DEFERRABLE_WORK(&hinfo->gc_work, htable_gc); |
347 | queue_delayed_work(system_power_efficient_wq, &hinfo->gc_work, | |
348 | msecs_to_jiffies(hinfo->cfg.gc_interval)); | |
09e410de | 349 | |
e89fc3f1 | 350 | hlist_add_head(&hinfo->node, &hashlimit_net->htables); |
09e410de JE |
351 | |
352 | return 0; | |
353 | } | |
354 | ||
a47362a2 JE |
355 | static bool select_all(const struct xt_hashlimit_htable *ht, |
356 | const struct dsthash_ent *he) | |
1da177e4 | 357 | { |
d384e65f | 358 | return true; |
1da177e4 LT |
359 | } |
360 | ||
a47362a2 JE |
361 | static bool select_gc(const struct xt_hashlimit_htable *ht, |
362 | const struct dsthash_ent *he) | |
1da177e4 | 363 | { |
cbebc51f | 364 | return time_after_eq(jiffies, he->expires); |
1da177e4 LT |
365 | } |
366 | ||
39b46fc6 | 367 | static void htable_selective_cleanup(struct xt_hashlimit_htable *ht, |
a47362a2 JE |
368 | bool (*select)(const struct xt_hashlimit_htable *ht, |
369 | const struct dsthash_ent *he)) | |
1da177e4 | 370 | { |
39b46fc6 | 371 | unsigned int i; |
1da177e4 | 372 | |
1da177e4 LT |
373 | for (i = 0; i < ht->cfg.size; i++) { |
374 | struct dsthash_ent *dh; | |
b67bfe0d | 375 | struct hlist_node *n; |
7bd8490e ED |
376 | |
377 | spin_lock_bh(&ht->lock); | |
b67bfe0d | 378 | hlist_for_each_entry_safe(dh, n, &ht->hash[i], node) { |
1da177e4 | 379 | if ((*select)(ht, dh)) |
39b46fc6 | 380 | dsthash_free(ht, dh); |
1da177e4 | 381 | } |
7bd8490e ED |
382 | spin_unlock_bh(&ht->lock); |
383 | cond_resched(); | |
1da177e4 | 384 | } |
1da177e4 LT |
385 | } |
386 | ||
7bd8490e | 387 | static void htable_gc(struct work_struct *work) |
1da177e4 | 388 | { |
7bd8490e ED |
389 | struct xt_hashlimit_htable *ht; |
390 | ||
391 | ht = container_of(work, struct xt_hashlimit_htable, gc_work.work); | |
1da177e4 LT |
392 | |
393 | htable_selective_cleanup(ht, select_gc); | |
394 | ||
7bd8490e ED |
395 | queue_delayed_work(system_power_efficient_wq, |
396 | &ht->gc_work, msecs_to_jiffies(ht->cfg.gc_interval)); | |
1da177e4 LT |
397 | } |
398 | ||
b4ef4ce0 | 399 | static void htable_remove_proc_entry(struct xt_hashlimit_htable *hinfo) |
1da177e4 | 400 | { |
e89fc3f1 AD |
401 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(hinfo->net); |
402 | struct proc_dir_entry *parent; | |
403 | ||
e89fc3f1 AD |
404 | if (hinfo->family == NFPROTO_IPV4) |
405 | parent = hashlimit_net->ipt_hashlimit; | |
406 | else | |
407 | parent = hashlimit_net->ip6t_hashlimit; | |
32263dd1 | 408 | |
b4ef4ce0 | 409 | if (parent != NULL) |
14b872f0 | 410 | remove_proc_entry(hinfo->name, parent); |
b4ef4ce0 | 411 | } |
32263dd1 | 412 | |
b4ef4ce0 SP |
413 | static void htable_destroy(struct xt_hashlimit_htable *hinfo) |
414 | { | |
7bd8490e | 415 | cancel_delayed_work_sync(&hinfo->gc_work); |
b4ef4ce0 | 416 | htable_remove_proc_entry(hinfo); |
1da177e4 | 417 | htable_selective_cleanup(hinfo, select_all); |
14b872f0 | 418 | kfree(hinfo->name); |
1da177e4 LT |
419 | vfree(hinfo); |
420 | } | |
421 | ||
e89fc3f1 AD |
422 | static struct xt_hashlimit_htable *htable_find_get(struct net *net, |
423 | const char *name, | |
76108cea | 424 | u_int8_t family) |
1da177e4 | 425 | { |
e89fc3f1 | 426 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); |
39b46fc6 | 427 | struct xt_hashlimit_htable *hinfo; |
1da177e4 | 428 | |
b67bfe0d | 429 | hlist_for_each_entry(hinfo, &hashlimit_net->htables, node) { |
14b872f0 | 430 | if (!strcmp(name, hinfo->name) && |
39b46fc6 | 431 | hinfo->family == family) { |
2eff25c1 | 432 | hinfo->use++; |
1da177e4 LT |
433 | return hinfo; |
434 | } | |
435 | } | |
1da177e4 LT |
436 | return NULL; |
437 | } | |
438 | ||
39b46fc6 | 439 | static void htable_put(struct xt_hashlimit_htable *hinfo) |
1da177e4 | 440 | { |
2eff25c1 PM |
441 | mutex_lock(&hashlimit_mutex); |
442 | if (--hinfo->use == 0) { | |
1da177e4 | 443 | hlist_del(&hinfo->node); |
1da177e4 LT |
444 | htable_destroy(hinfo); |
445 | } | |
2eff25c1 | 446 | mutex_unlock(&hashlimit_mutex); |
1da177e4 LT |
447 | } |
448 | ||
1da177e4 LT |
449 | /* The algorithm used is the Simple Token Bucket Filter (TBF) |
450 | * see net/sched/sch_tbf.c in the linux source tree | |
451 | */ | |
452 | ||
453 | /* Rusty: This is my (non-mathematically-inclined) understanding of | |
454 | this algorithm. The `average rate' in jiffies becomes your initial | |
455 | amount of credit `credit' and the most credit you can ever have | |
456 | `credit_cap'. The `peak rate' becomes the cost of passing the | |
457 | test, `cost'. | |
458 | ||
459 | `prev' tracks the last packet hit: you gain one credit per jiffy. | |
460 | If you get credit balance more than this, the extra credit is | |
461 | discarded. Every time the match passes, you lose `cost' credits; | |
462 | if you don't have that many, the test fails. | |
463 | ||
464 | See Alexey's formal explanation in net/sched/sch_tbf.c. | |
465 | ||
466 | To get the maximum range, we multiply by this factor (ie. you get N | |
467 | credits per jiffy). We want to allow a rate as low as 1 per day | |
468 | (slowest userspace tool allows), which means | |
469 | CREDITS_PER_JIFFY*HZ*60*60*24 < 2^32 ie. | |
470 | */ | |
0dc60a45 | 471 | #define MAX_CPJ_v1 (0xFFFFFFFF / (HZ*60*60*24)) |
1b203c13 | 472 | #define MAX_CPJ (0xFFFFFFFFFFFFFFFFULL / (HZ*60*60*24)) |
1da177e4 LT |
473 | |
474 | /* Repeated shift and or gives us all 1s, final shift and add 1 gives | |
475 | * us the power of 2 below the theoretical max, so GCC simply does a | |
476 | * shift. */ | |
477 | #define _POW2_BELOW2(x) ((x)|((x)>>1)) | |
478 | #define _POW2_BELOW4(x) (_POW2_BELOW2(x)|_POW2_BELOW2((x)>>2)) | |
479 | #define _POW2_BELOW8(x) (_POW2_BELOW4(x)|_POW2_BELOW4((x)>>4)) | |
480 | #define _POW2_BELOW16(x) (_POW2_BELOW8(x)|_POW2_BELOW8((x)>>8)) | |
481 | #define _POW2_BELOW32(x) (_POW2_BELOW16(x)|_POW2_BELOW16((x)>>16)) | |
11d5f157 | 482 | #define _POW2_BELOW64(x) (_POW2_BELOW32(x)|_POW2_BELOW32((x)>>32)) |
1da177e4 | 483 | #define POW2_BELOW32(x) ((_POW2_BELOW32(x)>>1) + 1) |
11d5f157 | 484 | #define POW2_BELOW64(x) ((_POW2_BELOW64(x)>>1) + 1) |
1da177e4 | 485 | |
11d5f157 | 486 | #define CREDITS_PER_JIFFY POW2_BELOW64(MAX_CPJ) |
0dc60a45 | 487 | #define CREDITS_PER_JIFFY_v1 POW2_BELOW32(MAX_CPJ_v1) |
1da177e4 | 488 | |
0197dee7 FW |
489 | /* in byte mode, the lowest possible rate is one packet/second. |
490 | * credit_cap is used as a counter that tells us how many times we can | |
491 | * refill the "credits available" counter when it becomes empty. | |
492 | */ | |
493 | #define MAX_CPJ_BYTES (0xFFFFFFFF / HZ) | |
494 | #define CREDITS_PER_JIFFY_BYTES POW2_BELOW32(MAX_CPJ_BYTES) | |
495 | ||
496 | static u32 xt_hashlimit_len_to_chunks(u32 len) | |
497 | { | |
498 | return (len >> XT_HASHLIMIT_BYTE_SHIFT) + 1; | |
499 | } | |
500 | ||
1da177e4 | 501 | /* Precision saver. */ |
11d5f157 | 502 | static u64 user2credits(u64 user, int revision) |
1da177e4 | 503 | { |
ad5b5576 AB |
504 | u64 scale = (revision == 1) ? |
505 | XT_HASHLIMIT_SCALE : XT_HASHLIMIT_SCALE_v2; | |
506 | u64 cpj = (revision == 1) ? | |
507 | CREDITS_PER_JIFFY_v1 : CREDITS_PER_JIFFY; | |
1da177e4 | 508 | |
ad5b5576 AB |
509 | /* Avoid overflow: divide the constant operands first */ |
510 | if (scale >= HZ * cpj) | |
511 | return div64_u64(user, div64_u64(scale, HZ * cpj)); | |
512 | ||
513 | return user * div64_u64(HZ * cpj, scale); | |
1da177e4 LT |
514 | } |
515 | ||
0197dee7 | 516 | static u32 user2credits_byte(u32 user) |
1da177e4 | 517 | { |
0197dee7 FW |
518 | u64 us = user; |
519 | us *= HZ * CREDITS_PER_JIFFY_BYTES; | |
520 | return (u32) (us >> 32); | |
521 | } | |
522 | ||
bea74641 VP |
523 | static u64 user2rate(u64 user) |
524 | { | |
525 | if (user != 0) { | |
526 | return div64_u64(XT_HASHLIMIT_SCALE_v2, user); | |
527 | } else { | |
b2606644 FW |
528 | pr_info_ratelimited("invalid rate from userspace: %llu\n", |
529 | user); | |
bea74641 VP |
530 | return 0; |
531 | } | |
532 | } | |
533 | ||
90c4ae4e | 534 | static u64 user2rate_bytes(u32 user) |
bea74641 VP |
535 | { |
536 | u64 r; | |
537 | ||
90c4ae4e | 538 | r = user ? U32_MAX / user : U32_MAX; |
c47d36b3 | 539 | return (r - 1) << XT_HASHLIMIT_BYTE_SHIFT; |
bea74641 VP |
540 | } |
541 | ||
11d5f157 VP |
542 | static void rateinfo_recalc(struct dsthash_ent *dh, unsigned long now, |
543 | u32 mode, int revision) | |
0197dee7 FW |
544 | { |
545 | unsigned long delta = now - dh->rateinfo.prev; | |
11d5f157 | 546 | u64 cap, cpj; |
0197dee7 FW |
547 | |
548 | if (delta == 0) | |
549 | return; | |
550 | ||
bea74641 VP |
551 | if (revision >= 3 && mode & XT_HASHLIMIT_RATE_MATCH) { |
552 | u64 interval = dh->rateinfo.interval * HZ; | |
553 | ||
554 | if (delta < interval) | |
555 | return; | |
556 | ||
557 | dh->rateinfo.prev = now; | |
558 | dh->rateinfo.prev_window = | |
559 | ((dh->rateinfo.current_rate * interval) > | |
560 | (delta * dh->rateinfo.rate)); | |
561 | dh->rateinfo.current_rate = 0; | |
562 | ||
563 | return; | |
564 | } | |
565 | ||
39b46fc6 | 566 | dh->rateinfo.prev = now; |
0197dee7 FW |
567 | |
568 | if (mode & XT_HASHLIMIT_BYTES) { | |
11d5f157 | 569 | u64 tmp = dh->rateinfo.credit; |
0197dee7 FW |
570 | dh->rateinfo.credit += CREDITS_PER_JIFFY_BYTES * delta; |
571 | cap = CREDITS_PER_JIFFY_BYTES * HZ; | |
572 | if (tmp >= dh->rateinfo.credit) {/* overflow */ | |
573 | dh->rateinfo.credit = cap; | |
574 | return; | |
575 | } | |
576 | } else { | |
11d5f157 VP |
577 | cpj = (revision == 1) ? |
578 | CREDITS_PER_JIFFY_v1 : CREDITS_PER_JIFFY; | |
579 | dh->rateinfo.credit += delta * cpj; | |
0197dee7 FW |
580 | cap = dh->rateinfo.credit_cap; |
581 | } | |
582 | if (dh->rateinfo.credit > cap) | |
583 | dh->rateinfo.credit = cap; | |
39b46fc6 PM |
584 | } |
585 | ||
817e076f | 586 | static void rateinfo_init(struct dsthash_ent *dh, |
11d5f157 | 587 | struct xt_hashlimit_htable *hinfo, int revision) |
817e076f FW |
588 | { |
589 | dh->rateinfo.prev = jiffies; | |
bea74641 VP |
590 | if (revision >= 3 && hinfo->cfg.mode & XT_HASHLIMIT_RATE_MATCH) { |
591 | dh->rateinfo.prev_window = 0; | |
592 | dh->rateinfo.current_rate = 0; | |
593 | if (hinfo->cfg.mode & XT_HASHLIMIT_BYTES) { | |
90c4ae4e VP |
594 | dh->rateinfo.rate = |
595 | user2rate_bytes((u32)hinfo->cfg.avg); | |
bea74641 VP |
596 | if (hinfo->cfg.burst) |
597 | dh->rateinfo.burst = | |
598 | hinfo->cfg.burst * dh->rateinfo.rate; | |
599 | else | |
600 | dh->rateinfo.burst = dh->rateinfo.rate; | |
601 | } else { | |
602 | dh->rateinfo.rate = user2rate(hinfo->cfg.avg); | |
603 | dh->rateinfo.burst = | |
604 | hinfo->cfg.burst + dh->rateinfo.rate; | |
605 | } | |
606 | dh->rateinfo.interval = hinfo->cfg.interval; | |
607 | } else if (hinfo->cfg.mode & XT_HASHLIMIT_BYTES) { | |
0197dee7 FW |
608 | dh->rateinfo.credit = CREDITS_PER_JIFFY_BYTES * HZ; |
609 | dh->rateinfo.cost = user2credits_byte(hinfo->cfg.avg); | |
610 | dh->rateinfo.credit_cap = hinfo->cfg.burst; | |
611 | } else { | |
612 | dh->rateinfo.credit = user2credits(hinfo->cfg.avg * | |
11d5f157 VP |
613 | hinfo->cfg.burst, revision); |
614 | dh->rateinfo.cost = user2credits(hinfo->cfg.avg, revision); | |
0197dee7 FW |
615 | dh->rateinfo.credit_cap = dh->rateinfo.credit; |
616 | } | |
817e076f FW |
617 | } |
618 | ||
09e410de JE |
619 | static inline __be32 maskl(__be32 a, unsigned int l) |
620 | { | |
1b9b70ea | 621 | return l ? htonl(ntohl(a) & ~0 << (32 - l)) : 0; |
09e410de JE |
622 | } |
623 | ||
c0cd1156 | 624 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
09e410de JE |
625 | static void hashlimit_ipv6_mask(__be32 *i, unsigned int p) |
626 | { | |
627 | switch (p) { | |
1b9b70ea | 628 | case 0 ... 31: |
09e410de JE |
629 | i[0] = maskl(i[0], p); |
630 | i[1] = i[2] = i[3] = 0; | |
631 | break; | |
1b9b70ea | 632 | case 32 ... 63: |
09e410de JE |
633 | i[1] = maskl(i[1], p - 32); |
634 | i[2] = i[3] = 0; | |
635 | break; | |
1b9b70ea | 636 | case 64 ... 95: |
09e410de JE |
637 | i[2] = maskl(i[2], p - 64); |
638 | i[3] = 0; | |
8f599229 | 639 | break; |
1b9b70ea | 640 | case 96 ... 127: |
09e410de JE |
641 | i[3] = maskl(i[3], p - 96); |
642 | break; | |
643 | case 128: | |
644 | break; | |
645 | } | |
646 | } | |
3ed5df44 | 647 | #endif |
09e410de | 648 | |
39b46fc6 | 649 | static int |
a47362a2 JE |
650 | hashlimit_init_dst(const struct xt_hashlimit_htable *hinfo, |
651 | struct dsthash_dst *dst, | |
39b46fc6 PM |
652 | const struct sk_buff *skb, unsigned int protoff) |
653 | { | |
654 | __be16 _ports[2], *ports; | |
193b23c5 | 655 | u8 nexthdr; |
aca071c1 | 656 | int poff; |
39b46fc6 PM |
657 | |
658 | memset(dst, 0, sizeof(*dst)); | |
659 | ||
660 | switch (hinfo->family) { | |
ee999d8b | 661 | case NFPROTO_IPV4: |
39b46fc6 | 662 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) |
09e410de JE |
663 | dst->ip.dst = maskl(ip_hdr(skb)->daddr, |
664 | hinfo->cfg.dstmask); | |
39b46fc6 | 665 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) |
09e410de JE |
666 | dst->ip.src = maskl(ip_hdr(skb)->saddr, |
667 | hinfo->cfg.srcmask); | |
39b46fc6 PM |
668 | |
669 | if (!(hinfo->cfg.mode & | |
670 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
671 | return 0; | |
eddc9ec5 | 672 | nexthdr = ip_hdr(skb)->protocol; |
39b46fc6 | 673 | break; |
c0cd1156 | 674 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
ee999d8b | 675 | case NFPROTO_IPV6: |
412662d2 SR |
676 | { |
677 | __be16 frag_off; | |
678 | ||
09e410de JE |
679 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) { |
680 | memcpy(&dst->ip6.dst, &ipv6_hdr(skb)->daddr, | |
681 | sizeof(dst->ip6.dst)); | |
682 | hashlimit_ipv6_mask(dst->ip6.dst, hinfo->cfg.dstmask); | |
683 | } | |
684 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) { | |
685 | memcpy(&dst->ip6.src, &ipv6_hdr(skb)->saddr, | |
686 | sizeof(dst->ip6.src)); | |
687 | hashlimit_ipv6_mask(dst->ip6.src, hinfo->cfg.srcmask); | |
688 | } | |
39b46fc6 PM |
689 | |
690 | if (!(hinfo->cfg.mode & | |
691 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
692 | return 0; | |
193b23c5 | 693 | nexthdr = ipv6_hdr(skb)->nexthdr; |
75f2811c | 694 | protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr, &frag_off); |
193b23c5 | 695 | if ((int)protoff < 0) |
39b46fc6 PM |
696 | return -1; |
697 | break; | |
412662d2 | 698 | } |
39b46fc6 PM |
699 | #endif |
700 | default: | |
701 | BUG(); | |
702 | return 0; | |
703 | } | |
704 | ||
aca071c1 CG |
705 | poff = proto_ports_offset(nexthdr); |
706 | if (poff >= 0) { | |
707 | ports = skb_header_pointer(skb, protoff + poff, sizeof(_ports), | |
39b46fc6 | 708 | &_ports); |
aca071c1 | 709 | } else { |
39b46fc6 PM |
710 | _ports[0] = _ports[1] = 0; |
711 | ports = _ports; | |
39b46fc6 PM |
712 | } |
713 | if (!ports) | |
714 | return -1; | |
715 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SPT) | |
716 | dst->src_port = ports[0]; | |
717 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DPT) | |
718 | dst->dst_port = ports[1]; | |
719 | return 0; | |
1da177e4 LT |
720 | } |
721 | ||
0197dee7 FW |
722 | static u32 hashlimit_byte_cost(unsigned int len, struct dsthash_ent *dh) |
723 | { | |
724 | u64 tmp = xt_hashlimit_len_to_chunks(len); | |
725 | tmp = tmp * dh->rateinfo.cost; | |
726 | ||
727 | if (unlikely(tmp > CREDITS_PER_JIFFY_BYTES * HZ)) | |
728 | tmp = CREDITS_PER_JIFFY_BYTES * HZ; | |
729 | ||
730 | if (dh->rateinfo.credit < tmp && dh->rateinfo.credit_cap) { | |
731 | dh->rateinfo.credit_cap--; | |
732 | dh->rateinfo.credit = CREDITS_PER_JIFFY_BYTES * HZ; | |
733 | } | |
734 | return (u32) tmp; | |
735 | } | |
736 | ||
ccb79bdc | 737 | static bool |
11d5f157 VP |
738 | hashlimit_mt_common(const struct sk_buff *skb, struct xt_action_param *par, |
739 | struct xt_hashlimit_htable *hinfo, | |
bea74641 | 740 | const struct hashlimit_cfg3 *cfg, int revision) |
09e410de | 741 | { |
09e410de JE |
742 | unsigned long now = jiffies; |
743 | struct dsthash_ent *dh; | |
744 | struct dsthash_dst dst; | |
09181842 | 745 | bool race = false; |
11d5f157 | 746 | u64 cost; |
09e410de | 747 | |
f7108a20 | 748 | if (hashlimit_init_dst(hinfo, &dst, skb, par->thoff) < 0) |
09e410de JE |
749 | goto hotdrop; |
750 | ||
0b35f603 | 751 | local_bh_disable(); |
09e410de JE |
752 | dh = dsthash_find(hinfo, &dst); |
753 | if (dh == NULL) { | |
09181842 | 754 | dh = dsthash_alloc_init(hinfo, &dst, &race); |
09e410de | 755 | if (dh == NULL) { |
0b35f603 | 756 | local_bh_enable(); |
09e410de | 757 | goto hotdrop; |
09181842 PNA |
758 | } else if (race) { |
759 | /* Already got an entry, update expiration timeout */ | |
760 | dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire); | |
11d5f157 | 761 | rateinfo_recalc(dh, now, hinfo->cfg.mode, revision); |
09181842 PNA |
762 | } else { |
763 | dh->expires = jiffies + msecs_to_jiffies(hinfo->cfg.expire); | |
11d5f157 | 764 | rateinfo_init(dh, hinfo, revision); |
09e410de | 765 | } |
09e410de JE |
766 | } else { |
767 | /* update expiration timeout */ | |
768 | dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire); | |
11d5f157 | 769 | rateinfo_recalc(dh, now, hinfo->cfg.mode, revision); |
09e410de JE |
770 | } |
771 | ||
bea74641 VP |
772 | if (cfg->mode & XT_HASHLIMIT_RATE_MATCH) { |
773 | cost = (cfg->mode & XT_HASHLIMIT_BYTES) ? skb->len : 1; | |
774 | dh->rateinfo.current_rate += cost; | |
775 | ||
776 | if (!dh->rateinfo.prev_window && | |
777 | (dh->rateinfo.current_rate <= dh->rateinfo.burst)) { | |
778 | spin_unlock(&dh->lock); | |
de526f40 | 779 | local_bh_enable(); |
bea74641 VP |
780 | return !(cfg->mode & XT_HASHLIMIT_INVERT); |
781 | } else { | |
782 | goto overlimit; | |
783 | } | |
784 | } | |
785 | ||
11d5f157 | 786 | if (cfg->mode & XT_HASHLIMIT_BYTES) |
0197dee7 FW |
787 | cost = hashlimit_byte_cost(skb->len, dh); |
788 | else | |
789 | cost = dh->rateinfo.cost; | |
790 | ||
791 | if (dh->rateinfo.credit >= cost) { | |
09e410de | 792 | /* below the limit */ |
0197dee7 | 793 | dh->rateinfo.credit -= cost; |
02e4eb75 | 794 | spin_unlock(&dh->lock); |
0b35f603 | 795 | local_bh_enable(); |
11d5f157 | 796 | return !(cfg->mode & XT_HASHLIMIT_INVERT); |
09e410de JE |
797 | } |
798 | ||
bea74641 | 799 | overlimit: |
02e4eb75 | 800 | spin_unlock(&dh->lock); |
0b35f603 | 801 | local_bh_enable(); |
09e410de | 802 | /* default match is underlimit - so over the limit, we need to invert */ |
11d5f157 | 803 | return cfg->mode & XT_HASHLIMIT_INVERT; |
09e410de JE |
804 | |
805 | hotdrop: | |
b4ba2611 | 806 | par->hotdrop = true; |
09e410de JE |
807 | return false; |
808 | } | |
809 | ||
11d5f157 VP |
810 | static bool |
811 | hashlimit_mt_v1(const struct sk_buff *skb, struct xt_action_param *par) | |
812 | { | |
813 | const struct xt_hashlimit_mtinfo1 *info = par->matchinfo; | |
814 | struct xt_hashlimit_htable *hinfo = info->hinfo; | |
bea74641 | 815 | struct hashlimit_cfg3 cfg = {}; |
11d5f157 VP |
816 | int ret; |
817 | ||
818 | ret = cfg_copy(&cfg, (void *)&info->cfg, 1); | |
11d5f157 VP |
819 | if (ret) |
820 | return ret; | |
821 | ||
822 | return hashlimit_mt_common(skb, par, hinfo, &cfg, 1); | |
823 | } | |
824 | ||
825 | static bool | |
bea74641 | 826 | hashlimit_mt_v2(const struct sk_buff *skb, struct xt_action_param *par) |
11d5f157 VP |
827 | { |
828 | const struct xt_hashlimit_mtinfo2 *info = par->matchinfo; | |
829 | struct xt_hashlimit_htable *hinfo = info->hinfo; | |
bea74641 VP |
830 | struct hashlimit_cfg3 cfg = {}; |
831 | int ret; | |
832 | ||
833 | ret = cfg_copy(&cfg, (void *)&info->cfg, 2); | |
bea74641 VP |
834 | if (ret) |
835 | return ret; | |
836 | ||
837 | return hashlimit_mt_common(skb, par, hinfo, &cfg, 2); | |
838 | } | |
839 | ||
840 | static bool | |
841 | hashlimit_mt(const struct sk_buff *skb, struct xt_action_param *par) | |
842 | { | |
843 | const struct xt_hashlimit_mtinfo3 *info = par->matchinfo; | |
844 | struct xt_hashlimit_htable *hinfo = info->hinfo; | |
11d5f157 | 845 | |
bea74641 | 846 | return hashlimit_mt_common(skb, par, hinfo, &info->cfg, 3); |
11d5f157 VP |
847 | } |
848 | ||
849 | static int hashlimit_mt_check_common(const struct xt_mtchk_param *par, | |
850 | struct xt_hashlimit_htable **hinfo, | |
bea74641 | 851 | struct hashlimit_cfg3 *cfg, |
11d5f157 | 852 | const char *name, int revision) |
09e410de | 853 | { |
e89fc3f1 | 854 | struct net *net = par->net; |
4a5a5c73 | 855 | int ret; |
09e410de | 856 | |
11d5f157 | 857 | if (cfg->gc_interval == 0 || cfg->expire == 0) |
bd414ee6 | 858 | return -EINVAL; |
aa5fa318 | 859 | if (par->family == NFPROTO_IPV4) { |
11d5f157 | 860 | if (cfg->srcmask > 32 || cfg->dstmask > 32) |
bd414ee6 | 861 | return -EINVAL; |
09e410de | 862 | } else { |
11d5f157 | 863 | if (cfg->srcmask > 128 || cfg->dstmask > 128) |
bd414ee6 | 864 | return -EINVAL; |
09e410de JE |
865 | } |
866 | ||
11d5f157 | 867 | if (cfg->mode & ~XT_HASHLIMIT_ALL) { |
b2606644 FW |
868 | pr_info_ratelimited("Unknown mode mask %X, kernel too old?\n", |
869 | cfg->mode); | |
0197dee7 FW |
870 | return -EINVAL; |
871 | } | |
872 | ||
873 | /* Check for overflow. */ | |
bea74641 | 874 | if (revision >= 3 && cfg->mode & XT_HASHLIMIT_RATE_MATCH) { |
90c4ae4e | 875 | if (cfg->avg == 0 || cfg->avg > U32_MAX) { |
b2606644 | 876 | pr_info_ratelimited("invalid rate\n"); |
bea74641 VP |
877 | return -ERANGE; |
878 | } | |
879 | ||
880 | if (cfg->interval == 0) { | |
b2606644 | 881 | pr_info_ratelimited("invalid interval\n"); |
bea74641 VP |
882 | return -EINVAL; |
883 | } | |
884 | } else if (cfg->mode & XT_HASHLIMIT_BYTES) { | |
11d5f157 | 885 | if (user2credits_byte(cfg->avg) == 0) { |
b2606644 FW |
886 | pr_info_ratelimited("overflow, rate too high: %llu\n", |
887 | cfg->avg); | |
0197dee7 FW |
888 | return -EINVAL; |
889 | } | |
11d5f157 | 890 | } else if (cfg->burst == 0 || |
b2606644 FW |
891 | user2credits(cfg->avg * cfg->burst, revision) < |
892 | user2credits(cfg->avg, revision)) { | |
893 | pr_info_ratelimited("overflow, try lower: %llu/%llu\n", | |
894 | cfg->avg, cfg->burst); | |
895 | return -ERANGE; | |
0197dee7 FW |
896 | } |
897 | ||
2eff25c1 | 898 | mutex_lock(&hashlimit_mutex); |
11d5f157 VP |
899 | *hinfo = htable_find_get(net, name, par->family); |
900 | if (*hinfo == NULL) { | |
901 | ret = htable_create(net, cfg, name, par->family, | |
902 | hinfo, revision); | |
4a5a5c73 JE |
903 | if (ret < 0) { |
904 | mutex_unlock(&hashlimit_mutex); | |
905 | return ret; | |
906 | } | |
09e410de | 907 | } |
2eff25c1 | 908 | mutex_unlock(&hashlimit_mutex); |
11d5f157 | 909 | |
bd414ee6 | 910 | return 0; |
09e410de JE |
911 | } |
912 | ||
11d5f157 VP |
913 | static int hashlimit_mt_check_v1(const struct xt_mtchk_param *par) |
914 | { | |
915 | struct xt_hashlimit_mtinfo1 *info = par->matchinfo; | |
bea74641 | 916 | struct hashlimit_cfg3 cfg = {}; |
11d5f157 VP |
917 | int ret; |
918 | ||
b1d0a5d0 FW |
919 | ret = xt_check_proc_name(info->name, sizeof(info->name)); |
920 | if (ret) | |
921 | return ret; | |
11d5f157 VP |
922 | |
923 | ret = cfg_copy(&cfg, (void *)&info->cfg, 1); | |
11d5f157 VP |
924 | if (ret) |
925 | return ret; | |
926 | ||
927 | return hashlimit_mt_check_common(par, &info->hinfo, | |
928 | &cfg, info->name, 1); | |
929 | } | |
930 | ||
bea74641 | 931 | static int hashlimit_mt_check_v2(const struct xt_mtchk_param *par) |
11d5f157 VP |
932 | { |
933 | struct xt_hashlimit_mtinfo2 *info = par->matchinfo; | |
bea74641 VP |
934 | struct hashlimit_cfg3 cfg = {}; |
935 | int ret; | |
936 | ||
b1d0a5d0 FW |
937 | ret = xt_check_proc_name(info->name, sizeof(info->name)); |
938 | if (ret) | |
939 | return ret; | |
bea74641 VP |
940 | |
941 | ret = cfg_copy(&cfg, (void *)&info->cfg, 2); | |
bea74641 VP |
942 | if (ret) |
943 | return ret; | |
944 | ||
945 | return hashlimit_mt_check_common(par, &info->hinfo, | |
946 | &cfg, info->name, 2); | |
947 | } | |
948 | ||
949 | static int hashlimit_mt_check(const struct xt_mtchk_param *par) | |
950 | { | |
951 | struct xt_hashlimit_mtinfo3 *info = par->matchinfo; | |
b1d0a5d0 | 952 | int ret; |
11d5f157 | 953 | |
b1d0a5d0 FW |
954 | ret = xt_check_proc_name(info->name, sizeof(info->name)); |
955 | if (ret) | |
956 | return ret; | |
11d5f157 VP |
957 | |
958 | return hashlimit_mt_check_common(par, &info->hinfo, &info->cfg, | |
bea74641 VP |
959 | info->name, 3); |
960 | } | |
961 | ||
962 | static void hashlimit_mt_destroy_v2(const struct xt_mtdtor_param *par) | |
963 | { | |
964 | const struct xt_hashlimit_mtinfo2 *info = par->matchinfo; | |
965 | ||
966 | htable_put(info->hinfo); | |
11d5f157 VP |
967 | } |
968 | ||
0dc60a45 | 969 | static void hashlimit_mt_destroy_v1(const struct xt_mtdtor_param *par) |
09e410de | 970 | { |
6be3d859 | 971 | const struct xt_hashlimit_mtinfo1 *info = par->matchinfo; |
09e410de JE |
972 | |
973 | htable_put(info->hinfo); | |
974 | } | |
975 | ||
11d5f157 VP |
976 | static void hashlimit_mt_destroy(const struct xt_mtdtor_param *par) |
977 | { | |
bea74641 | 978 | const struct xt_hashlimit_mtinfo3 *info = par->matchinfo; |
11d5f157 VP |
979 | |
980 | htable_put(info->hinfo); | |
981 | } | |
982 | ||
d3c5ee6d | 983 | static struct xt_match hashlimit_mt_reg[] __read_mostly = { |
09e410de JE |
984 | { |
985 | .name = "hashlimit", | |
986 | .revision = 1, | |
ee999d8b | 987 | .family = NFPROTO_IPV4, |
0dc60a45 | 988 | .match = hashlimit_mt_v1, |
09e410de | 989 | .matchsize = sizeof(struct xt_hashlimit_mtinfo1), |
ec231890 | 990 | .usersize = offsetof(struct xt_hashlimit_mtinfo1, hinfo), |
0dc60a45 VP |
991 | .checkentry = hashlimit_mt_check_v1, |
992 | .destroy = hashlimit_mt_destroy_v1, | |
09e410de JE |
993 | .me = THIS_MODULE, |
994 | }, | |
11d5f157 VP |
995 | { |
996 | .name = "hashlimit", | |
997 | .revision = 2, | |
998 | .family = NFPROTO_IPV4, | |
bea74641 | 999 | .match = hashlimit_mt_v2, |
11d5f157 | 1000 | .matchsize = sizeof(struct xt_hashlimit_mtinfo2), |
ec231890 | 1001 | .usersize = offsetof(struct xt_hashlimit_mtinfo2, hinfo), |
bea74641 VP |
1002 | .checkentry = hashlimit_mt_check_v2, |
1003 | .destroy = hashlimit_mt_destroy_v2, | |
1004 | .me = THIS_MODULE, | |
1005 | }, | |
1006 | { | |
1007 | .name = "hashlimit", | |
1008 | .revision = 3, | |
1009 | .family = NFPROTO_IPV4, | |
1010 | .match = hashlimit_mt, | |
1011 | .matchsize = sizeof(struct xt_hashlimit_mtinfo3), | |
1012 | .usersize = offsetof(struct xt_hashlimit_mtinfo3, hinfo), | |
11d5f157 VP |
1013 | .checkentry = hashlimit_mt_check, |
1014 | .destroy = hashlimit_mt_destroy, | |
1015 | .me = THIS_MODULE, | |
1016 | }, | |
c0cd1156 | 1017 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
09e410de JE |
1018 | { |
1019 | .name = "hashlimit", | |
1020 | .revision = 1, | |
ee999d8b | 1021 | .family = NFPROTO_IPV6, |
0dc60a45 | 1022 | .match = hashlimit_mt_v1, |
09e410de | 1023 | .matchsize = sizeof(struct xt_hashlimit_mtinfo1), |
ec231890 | 1024 | .usersize = offsetof(struct xt_hashlimit_mtinfo1, hinfo), |
0dc60a45 VP |
1025 | .checkentry = hashlimit_mt_check_v1, |
1026 | .destroy = hashlimit_mt_destroy_v1, | |
09e410de JE |
1027 | .me = THIS_MODULE, |
1028 | }, | |
11d5f157 VP |
1029 | { |
1030 | .name = "hashlimit", | |
1031 | .revision = 2, | |
1032 | .family = NFPROTO_IPV6, | |
bea74641 | 1033 | .match = hashlimit_mt_v2, |
11d5f157 | 1034 | .matchsize = sizeof(struct xt_hashlimit_mtinfo2), |
ec231890 | 1035 | .usersize = offsetof(struct xt_hashlimit_mtinfo2, hinfo), |
bea74641 VP |
1036 | .checkentry = hashlimit_mt_check_v2, |
1037 | .destroy = hashlimit_mt_destroy_v2, | |
1038 | .me = THIS_MODULE, | |
1039 | }, | |
1040 | { | |
1041 | .name = "hashlimit", | |
1042 | .revision = 3, | |
1043 | .family = NFPROTO_IPV6, | |
1044 | .match = hashlimit_mt, | |
1045 | .matchsize = sizeof(struct xt_hashlimit_mtinfo3), | |
1046 | .usersize = offsetof(struct xt_hashlimit_mtinfo3, hinfo), | |
11d5f157 VP |
1047 | .checkentry = hashlimit_mt_check, |
1048 | .destroy = hashlimit_mt_destroy, | |
1049 | .me = THIS_MODULE, | |
1050 | }, | |
7b21e09d | 1051 | #endif |
1da177e4 LT |
1052 | }; |
1053 | ||
1054 | /* PROC stuff */ | |
1da177e4 | 1055 | static void *dl_seq_start(struct seq_file *s, loff_t *pos) |
f4f6fb71 | 1056 | __acquires(htable->lock) |
1da177e4 | 1057 | { |
1286df26 | 1058 | struct xt_hashlimit_htable *htable = PDE_DATA(file_inode(s->file)); |
1da177e4 LT |
1059 | unsigned int *bucket; |
1060 | ||
1061 | spin_lock_bh(&htable->lock); | |
1062 | if (*pos >= htable->cfg.size) | |
1063 | return NULL; | |
1064 | ||
1065 | bucket = kmalloc(sizeof(unsigned int), GFP_ATOMIC); | |
1066 | if (!bucket) | |
1067 | return ERR_PTR(-ENOMEM); | |
1068 | ||
1069 | *bucket = *pos; | |
1070 | return bucket; | |
1071 | } | |
1072 | ||
1073 | static void *dl_seq_next(struct seq_file *s, void *v, loff_t *pos) | |
1074 | { | |
1286df26 | 1075 | struct xt_hashlimit_htable *htable = PDE_DATA(file_inode(s->file)); |
68ad546a | 1076 | unsigned int *bucket = v; |
1da177e4 LT |
1077 | |
1078 | *pos = ++(*bucket); | |
1079 | if (*pos >= htable->cfg.size) { | |
1080 | kfree(v); | |
1081 | return NULL; | |
1082 | } | |
1083 | return bucket; | |
1084 | } | |
1085 | ||
1086 | static void dl_seq_stop(struct seq_file *s, void *v) | |
f4f6fb71 | 1087 | __releases(htable->lock) |
1da177e4 | 1088 | { |
1286df26 | 1089 | struct xt_hashlimit_htable *htable = PDE_DATA(file_inode(s->file)); |
68ad546a | 1090 | unsigned int *bucket = v; |
1da177e4 | 1091 | |
55e0d7cf ED |
1092 | if (!IS_ERR(bucket)) |
1093 | kfree(bucket); | |
1da177e4 LT |
1094 | spin_unlock_bh(&htable->lock); |
1095 | } | |
1096 | ||
11d5f157 VP |
1097 | static void dl_seq_print(struct dsthash_ent *ent, u_int8_t family, |
1098 | struct seq_file *s) | |
1da177e4 | 1099 | { |
39b46fc6 | 1100 | switch (family) { |
ee999d8b | 1101 | case NFPROTO_IPV4: |
11d5f157 | 1102 | seq_printf(s, "%ld %pI4:%u->%pI4:%u %llu %llu %llu\n", |
e71456ae SRRH |
1103 | (long)(ent->expires - jiffies)/HZ, |
1104 | &ent->dst.ip.src, | |
1105 | ntohs(ent->dst.src_port), | |
1106 | &ent->dst.ip.dst, | |
1107 | ntohs(ent->dst.dst_port), | |
1108 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
1109 | ent->rateinfo.cost); | |
02e4eb75 | 1110 | break; |
c0cd1156 | 1111 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
ee999d8b | 1112 | case NFPROTO_IPV6: |
11d5f157 | 1113 | seq_printf(s, "%ld %pI6:%u->%pI6:%u %llu %llu %llu\n", |
e71456ae SRRH |
1114 | (long)(ent->expires - jiffies)/HZ, |
1115 | &ent->dst.ip6.src, | |
1116 | ntohs(ent->dst.src_port), | |
1117 | &ent->dst.ip6.dst, | |
1118 | ntohs(ent->dst.dst_port), | |
1119 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
1120 | ent->rateinfo.cost); | |
02e4eb75 | 1121 | break; |
7b21e09d | 1122 | #endif |
39b46fc6 PM |
1123 | default: |
1124 | BUG(); | |
39b46fc6 | 1125 | } |
11d5f157 VP |
1126 | } |
1127 | ||
bea74641 VP |
1128 | static int dl_seq_real_show_v2(struct dsthash_ent *ent, u_int8_t family, |
1129 | struct seq_file *s) | |
1130 | { | |
1286df26 | 1131 | struct xt_hashlimit_htable *ht = PDE_DATA(file_inode(s->file)); |
bea74641 VP |
1132 | |
1133 | spin_lock(&ent->lock); | |
1134 | /* recalculate to show accurate numbers */ | |
1135 | rateinfo_recalc(ent, jiffies, ht->cfg.mode, 2); | |
1136 | ||
1137 | dl_seq_print(ent, family, s); | |
1138 | ||
1139 | spin_unlock(&ent->lock); | |
1140 | return seq_has_overflowed(s); | |
1141 | } | |
1142 | ||
11d5f157 VP |
1143 | static int dl_seq_real_show_v1(struct dsthash_ent *ent, u_int8_t family, |
1144 | struct seq_file *s) | |
1145 | { | |
1286df26 | 1146 | struct xt_hashlimit_htable *ht = PDE_DATA(file_inode(s->file)); |
11d5f157 VP |
1147 | |
1148 | spin_lock(&ent->lock); | |
1149 | /* recalculate to show accurate numbers */ | |
1150 | rateinfo_recalc(ent, jiffies, ht->cfg.mode, 1); | |
1151 | ||
1152 | dl_seq_print(ent, family, s); | |
1153 | ||
1154 | spin_unlock(&ent->lock); | |
1155 | return seq_has_overflowed(s); | |
1156 | } | |
1157 | ||
1158 | static int dl_seq_real_show(struct dsthash_ent *ent, u_int8_t family, | |
1159 | struct seq_file *s) | |
1160 | { | |
1286df26 | 1161 | struct xt_hashlimit_htable *ht = PDE_DATA(file_inode(s->file)); |
11d5f157 VP |
1162 | |
1163 | spin_lock(&ent->lock); | |
1164 | /* recalculate to show accurate numbers */ | |
bea74641 | 1165 | rateinfo_recalc(ent, jiffies, ht->cfg.mode, 3); |
11d5f157 VP |
1166 | |
1167 | dl_seq_print(ent, family, s); | |
1168 | ||
02e4eb75 | 1169 | spin_unlock(&ent->lock); |
e71456ae | 1170 | return seq_has_overflowed(s); |
1da177e4 LT |
1171 | } |
1172 | ||
bea74641 VP |
1173 | static int dl_seq_show_v2(struct seq_file *s, void *v) |
1174 | { | |
1286df26 | 1175 | struct xt_hashlimit_htable *htable = PDE_DATA(file_inode(s->file)); |
bea74641 VP |
1176 | unsigned int *bucket = (unsigned int *)v; |
1177 | struct dsthash_ent *ent; | |
1178 | ||
1179 | if (!hlist_empty(&htable->hash[*bucket])) { | |
1180 | hlist_for_each_entry(ent, &htable->hash[*bucket], node) | |
1181 | if (dl_seq_real_show_v2(ent, htable->family, s)) | |
1182 | return -1; | |
1183 | } | |
1184 | return 0; | |
1185 | } | |
1186 | ||
0dc60a45 | 1187 | static int dl_seq_show_v1(struct seq_file *s, void *v) |
1da177e4 | 1188 | { |
1286df26 | 1189 | struct xt_hashlimit_htable *htable = PDE_DATA(file_inode(s->file)); |
68ad546a | 1190 | unsigned int *bucket = v; |
1da177e4 | 1191 | struct dsthash_ent *ent; |
1da177e4 | 1192 | |
39b46fc6 | 1193 | if (!hlist_empty(&htable->hash[*bucket])) { |
b67bfe0d | 1194 | hlist_for_each_entry(ent, &htable->hash[*bucket], node) |
0dc60a45 | 1195 | if (dl_seq_real_show_v1(ent, htable->family, s)) |
683a04ce | 1196 | return -1; |
39b46fc6 | 1197 | } |
1da177e4 LT |
1198 | return 0; |
1199 | } | |
1200 | ||
11d5f157 VP |
1201 | static int dl_seq_show(struct seq_file *s, void *v) |
1202 | { | |
1286df26 | 1203 | struct xt_hashlimit_htable *htable = PDE_DATA(file_inode(s->file)); |
68ad546a | 1204 | unsigned int *bucket = v; |
11d5f157 VP |
1205 | struct dsthash_ent *ent; |
1206 | ||
1207 | if (!hlist_empty(&htable->hash[*bucket])) { | |
1208 | hlist_for_each_entry(ent, &htable->hash[*bucket], node) | |
1209 | if (dl_seq_real_show(ent, htable->family, s)) | |
1210 | return -1; | |
1211 | } | |
1212 | return 0; | |
1213 | } | |
1214 | ||
0dc60a45 | 1215 | static const struct seq_operations dl_seq_ops_v1 = { |
1da177e4 LT |
1216 | .start = dl_seq_start, |
1217 | .next = dl_seq_next, | |
1218 | .stop = dl_seq_stop, | |
0dc60a45 | 1219 | .show = dl_seq_show_v1 |
1da177e4 LT |
1220 | }; |
1221 | ||
bea74641 VP |
1222 | static const struct seq_operations dl_seq_ops_v2 = { |
1223 | .start = dl_seq_start, | |
1224 | .next = dl_seq_next, | |
1225 | .stop = dl_seq_stop, | |
1226 | .show = dl_seq_show_v2 | |
1227 | }; | |
1228 | ||
11d5f157 VP |
1229 | static const struct seq_operations dl_seq_ops = { |
1230 | .start = dl_seq_start, | |
1231 | .next = dl_seq_next, | |
1232 | .stop = dl_seq_stop, | |
1233 | .show = dl_seq_show | |
1234 | }; | |
1235 | ||
e89fc3f1 AD |
1236 | static int __net_init hashlimit_proc_net_init(struct net *net) |
1237 | { | |
1238 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); | |
1239 | ||
1240 | hashlimit_net->ipt_hashlimit = proc_mkdir("ipt_hashlimit", net->proc_net); | |
1241 | if (!hashlimit_net->ipt_hashlimit) | |
1242 | return -ENOMEM; | |
c0cd1156 | 1243 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
e89fc3f1 AD |
1244 | hashlimit_net->ip6t_hashlimit = proc_mkdir("ip6t_hashlimit", net->proc_net); |
1245 | if (!hashlimit_net->ip6t_hashlimit) { | |
ece31ffd | 1246 | remove_proc_entry("ipt_hashlimit", net->proc_net); |
e89fc3f1 AD |
1247 | return -ENOMEM; |
1248 | } | |
1249 | #endif | |
1250 | return 0; | |
1251 | } | |
1252 | ||
1253 | static void __net_exit hashlimit_proc_net_exit(struct net *net) | |
1254 | { | |
32263dd1 | 1255 | struct xt_hashlimit_htable *hinfo; |
32263dd1 VL |
1256 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); |
1257 | ||
b4ef4ce0 SP |
1258 | /* hashlimit_net_exit() is called before hashlimit_mt_destroy(). |
1259 | * Make sure that the parent ipt_hashlimit and ip6t_hashlimit proc | |
1260 | * entries is empty before trying to remove it. | |
32263dd1 VL |
1261 | */ |
1262 | mutex_lock(&hashlimit_mutex); | |
b67bfe0d | 1263 | hlist_for_each_entry(hinfo, &hashlimit_net->htables, node) |
b4ef4ce0 | 1264 | htable_remove_proc_entry(hinfo); |
32263dd1 VL |
1265 | hashlimit_net->ipt_hashlimit = NULL; |
1266 | hashlimit_net->ip6t_hashlimit = NULL; | |
1267 | mutex_unlock(&hashlimit_mutex); | |
1268 | ||
ece31ffd | 1269 | remove_proc_entry("ipt_hashlimit", net->proc_net); |
c0cd1156 | 1270 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
ece31ffd | 1271 | remove_proc_entry("ip6t_hashlimit", net->proc_net); |
e89fc3f1 AD |
1272 | #endif |
1273 | } | |
1274 | ||
1275 | static int __net_init hashlimit_net_init(struct net *net) | |
1276 | { | |
1277 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); | |
1278 | ||
1279 | INIT_HLIST_HEAD(&hashlimit_net->htables); | |
1280 | return hashlimit_proc_net_init(net); | |
1281 | } | |
1282 | ||
1283 | static void __net_exit hashlimit_net_exit(struct net *net) | |
1284 | { | |
e89fc3f1 AD |
1285 | hashlimit_proc_net_exit(net); |
1286 | } | |
1287 | ||
1288 | static struct pernet_operations hashlimit_net_ops = { | |
1289 | .init = hashlimit_net_init, | |
1290 | .exit = hashlimit_net_exit, | |
1291 | .id = &hashlimit_net_id, | |
1292 | .size = sizeof(struct hashlimit_net), | |
1293 | }; | |
1294 | ||
d3c5ee6d | 1295 | static int __init hashlimit_mt_init(void) |
1da177e4 | 1296 | { |
39b46fc6 | 1297 | int err; |
1da177e4 | 1298 | |
e89fc3f1 AD |
1299 | err = register_pernet_subsys(&hashlimit_net_ops); |
1300 | if (err < 0) | |
1301 | return err; | |
d3c5ee6d JE |
1302 | err = xt_register_matches(hashlimit_mt_reg, |
1303 | ARRAY_SIZE(hashlimit_mt_reg)); | |
39b46fc6 PM |
1304 | if (err < 0) |
1305 | goto err1; | |
1da177e4 | 1306 | |
39b46fc6 PM |
1307 | err = -ENOMEM; |
1308 | hashlimit_cachep = kmem_cache_create("xt_hashlimit", | |
1309 | sizeof(struct dsthash_ent), 0, 0, | |
20c2df83 | 1310 | NULL); |
1da177e4 | 1311 | if (!hashlimit_cachep) { |
b167a37c | 1312 | pr_warn("unable to create slab cache\n"); |
39b46fc6 | 1313 | goto err2; |
1da177e4 | 1314 | } |
e89fc3f1 AD |
1315 | return 0; |
1316 | ||
39b46fc6 | 1317 | err2: |
d3c5ee6d | 1318 | xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg)); |
39b46fc6 | 1319 | err1: |
e89fc3f1 | 1320 | unregister_pernet_subsys(&hashlimit_net_ops); |
39b46fc6 | 1321 | return err; |
1da177e4 | 1322 | |
1da177e4 LT |
1323 | } |
1324 | ||
d3c5ee6d | 1325 | static void __exit hashlimit_mt_exit(void) |
1da177e4 | 1326 | { |
d3c5ee6d | 1327 | xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg)); |
e89fc3f1 | 1328 | unregister_pernet_subsys(&hashlimit_net_ops); |
02e4eb75 | 1329 | |
c8d1da40 | 1330 | rcu_barrier(); |
02e4eb75 | 1331 | kmem_cache_destroy(hashlimit_cachep); |
1da177e4 LT |
1332 | } |
1333 | ||
d3c5ee6d JE |
1334 | module_init(hashlimit_mt_init); |
1335 | module_exit(hashlimit_mt_exit); |