]>
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 | 5 | * (C) 2003-2004 by Harald Welte <laforge@netfilter.org> |
f229f6ce | 6 | * (C) 2006-2012 Patrick McHardy <kaber@trash.net> |
09e410de | 7 | * Copyright © CC Computer Consultants GmbH, 2007 - 2008 |
1da177e4 LT |
8 | * |
9 | * Development of this code was funded by Astaro AG, http://www.astaro.com/ | |
1da177e4 | 10 | */ |
8bee4bad | 11 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
1da177e4 | 12 | #include <linux/module.h> |
1da177e4 LT |
13 | #include <linux/spinlock.h> |
14 | #include <linux/random.h> | |
15 | #include <linux/jhash.h> | |
16 | #include <linux/slab.h> | |
17 | #include <linux/vmalloc.h> | |
1da177e4 LT |
18 | #include <linux/proc_fs.h> |
19 | #include <linux/seq_file.h> | |
20 | #include <linux/list.h> | |
39b46fc6 | 21 | #include <linux/skbuff.h> |
d7fe0f24 | 22 | #include <linux/mm.h> |
39b46fc6 PM |
23 | #include <linux/in.h> |
24 | #include <linux/ip.h> | |
c0cd1156 | 25 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
39b46fc6 | 26 | #include <linux/ipv6.h> |
193b23c5 | 27 | #include <net/ipv6.h> |
7b21e09d ED |
28 | #endif |
29 | ||
457c4cbc | 30 | #include <net/net_namespace.h> |
e89fc3f1 | 31 | #include <net/netns/generic.h> |
1da177e4 | 32 | |
39b46fc6 | 33 | #include <linux/netfilter/x_tables.h> |
1da177e4 | 34 | #include <linux/netfilter_ipv4/ip_tables.h> |
39b46fc6 PM |
35 | #include <linux/netfilter_ipv6/ip6_tables.h> |
36 | #include <linux/netfilter/xt_hashlimit.h> | |
14cc3e2b | 37 | #include <linux/mutex.h> |
1da177e4 LT |
38 | |
39 | MODULE_LICENSE("GPL"); | |
40 | MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); | |
408ffaa4 | 41 | MODULE_AUTHOR("Jan Engelhardt <jengelh@medozas.de>"); |
2ae15b64 | 42 | MODULE_DESCRIPTION("Xtables: per hash-bucket rate-limit match"); |
39b46fc6 PM |
43 | MODULE_ALIAS("ipt_hashlimit"); |
44 | MODULE_ALIAS("ip6t_hashlimit"); | |
1da177e4 | 45 | |
e89fc3f1 AD |
46 | struct hashlimit_net { |
47 | struct hlist_head htables; | |
48 | struct proc_dir_entry *ipt_hashlimit; | |
49 | struct proc_dir_entry *ip6t_hashlimit; | |
50 | }; | |
51 | ||
c7d03a00 | 52 | static unsigned int hashlimit_net_id; |
e89fc3f1 AD |
53 | static inline struct hashlimit_net *hashlimit_pernet(struct net *net) |
54 | { | |
55 | return net_generic(net, hashlimit_net_id); | |
56 | } | |
57 | ||
1da177e4 | 58 | /* need to declare this at the top */ |
0dc60a45 | 59 | static const struct file_operations dl_file_ops_v1; |
11d5f157 | 60 | static const struct file_operations dl_file_ops; |
1da177e4 LT |
61 | |
62 | /* hash table crap */ | |
1da177e4 | 63 | struct dsthash_dst { |
39b46fc6 PM |
64 | union { |
65 | struct { | |
66 | __be32 src; | |
67 | __be32 dst; | |
68 | } ip; | |
c0cd1156 | 69 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
39b46fc6 PM |
70 | struct { |
71 | __be32 src[4]; | |
72 | __be32 dst[4]; | |
73 | } ip6; | |
7b21e09d | 74 | #endif |
09e410de | 75 | }; |
6a19d614 AV |
76 | __be16 src_port; |
77 | __be16 dst_port; | |
1da177e4 LT |
78 | }; |
79 | ||
80 | struct dsthash_ent { | |
81 | /* static / read-only parts in the beginning */ | |
82 | struct hlist_node node; | |
83 | struct dsthash_dst dst; | |
84 | ||
85 | /* modified structure members in the end */ | |
02e4eb75 | 86 | spinlock_t lock; |
1da177e4 LT |
87 | unsigned long expires; /* precalculated expiry time */ |
88 | struct { | |
89 | unsigned long prev; /* last modification */ | |
11d5f157 VP |
90 | u_int64_t credit; |
91 | u_int64_t credit_cap, cost; | |
1da177e4 | 92 | } rateinfo; |
02e4eb75 | 93 | struct rcu_head rcu; |
1da177e4 LT |
94 | }; |
95 | ||
39b46fc6 | 96 | struct xt_hashlimit_htable { |
1da177e4 | 97 | struct hlist_node node; /* global list of all htables */ |
2eff25c1 | 98 | int use; |
76108cea | 99 | u_int8_t family; |
89bc7a0f | 100 | bool rnd_initialized; |
1da177e4 | 101 | |
11d5f157 | 102 | struct hashlimit_cfg2 cfg; /* config */ |
1da177e4 LT |
103 | |
104 | /* used internally */ | |
105 | spinlock_t lock; /* lock for list_head */ | |
106 | u_int32_t rnd; /* random seed for hash */ | |
39b46fc6 | 107 | unsigned int count; /* number entries in table */ |
7bd8490e | 108 | struct delayed_work gc_work; |
1da177e4 LT |
109 | |
110 | /* seq_file stuff */ | |
111 | struct proc_dir_entry *pde; | |
14b872f0 | 112 | const char *name; |
e89fc3f1 | 113 | struct net *net; |
1da177e4 LT |
114 | |
115 | struct hlist_head hash[0]; /* hashtable itself */ | |
116 | }; | |
117 | ||
11d5f157 VP |
118 | static int |
119 | cfg_copy(struct hashlimit_cfg2 *to, void *from, int revision) | |
120 | { | |
121 | if (revision == 1) { | |
122 | struct hashlimit_cfg1 *cfg = (struct hashlimit_cfg1 *)from; | |
123 | ||
124 | to->mode = cfg->mode; | |
125 | to->avg = cfg->avg; | |
126 | to->burst = cfg->burst; | |
127 | to->size = cfg->size; | |
128 | to->max = cfg->max; | |
129 | to->gc_interval = cfg->gc_interval; | |
130 | to->expire = cfg->expire; | |
131 | to->srcmask = cfg->srcmask; | |
132 | to->dstmask = cfg->dstmask; | |
133 | } else if (revision == 2) { | |
134 | memcpy(to, from, sizeof(struct hashlimit_cfg2)); | |
135 | } else { | |
136 | return -EINVAL; | |
137 | } | |
138 | ||
139 | return 0; | |
140 | } | |
141 | ||
2eff25c1 | 142 | static DEFINE_MUTEX(hashlimit_mutex); /* protects htables list */ |
e18b890b | 143 | static struct kmem_cache *hashlimit_cachep __read_mostly; |
1da177e4 | 144 | |
1d93a9cb | 145 | static inline bool dst_cmp(const struct dsthash_ent *ent, |
a47362a2 | 146 | const struct dsthash_dst *b) |
1da177e4 | 147 | { |
39b46fc6 | 148 | return !memcmp(&ent->dst, b, sizeof(ent->dst)); |
1da177e4 LT |
149 | } |
150 | ||
39b46fc6 PM |
151 | static u_int32_t |
152 | hash_dst(const struct xt_hashlimit_htable *ht, const struct dsthash_dst *dst) | |
1da177e4 | 153 | { |
e2f82ac3 ED |
154 | u_int32_t hash = jhash2((const u32 *)dst, |
155 | sizeof(*dst)/sizeof(u32), | |
156 | ht->rnd); | |
157 | /* | |
158 | * Instead of returning hash % ht->cfg.size (implying a divide) | |
159 | * we return the high 32 bits of the (hash * ht->cfg.size) that will | |
160 | * give results between [0 and cfg.size-1] and same hash distribution, | |
161 | * but using a multiply, less expensive than a divide | |
162 | */ | |
8fc54f68 | 163 | return reciprocal_scale(hash, ht->cfg.size); |
1da177e4 LT |
164 | } |
165 | ||
39b46fc6 | 166 | static struct dsthash_ent * |
a47362a2 JE |
167 | dsthash_find(const struct xt_hashlimit_htable *ht, |
168 | const struct dsthash_dst *dst) | |
1da177e4 LT |
169 | { |
170 | struct dsthash_ent *ent; | |
1da177e4 LT |
171 | u_int32_t hash = hash_dst(ht, dst); |
172 | ||
39b46fc6 | 173 | if (!hlist_empty(&ht->hash[hash])) { |
b67bfe0d | 174 | hlist_for_each_entry_rcu(ent, &ht->hash[hash], node) |
02e4eb75 ED |
175 | if (dst_cmp(ent, dst)) { |
176 | spin_lock(&ent->lock); | |
1da177e4 | 177 | return ent; |
02e4eb75 | 178 | } |
39b46fc6 | 179 | } |
1da177e4 LT |
180 | return NULL; |
181 | } | |
182 | ||
183 | /* allocate dsthash_ent, initialize dst, put in htable and lock it */ | |
184 | static struct dsthash_ent * | |
a47362a2 | 185 | dsthash_alloc_init(struct xt_hashlimit_htable *ht, |
09181842 | 186 | const struct dsthash_dst *dst, bool *race) |
1da177e4 LT |
187 | { |
188 | struct dsthash_ent *ent; | |
189 | ||
02e4eb75 | 190 | spin_lock(&ht->lock); |
09181842 PNA |
191 | |
192 | /* Two or more packets may race to create the same entry in the | |
193 | * hashtable, double check if this packet lost race. | |
194 | */ | |
195 | ent = dsthash_find(ht, dst); | |
196 | if (ent != NULL) { | |
197 | spin_unlock(&ht->lock); | |
198 | *race = true; | |
199 | return ent; | |
200 | } | |
201 | ||
1da177e4 LT |
202 | /* initialize hash with random val at the time we allocate |
203 | * the first hashtable entry */ | |
02e4eb75 | 204 | if (unlikely(!ht->rnd_initialized)) { |
af07d241 | 205 | get_random_bytes(&ht->rnd, sizeof(ht->rnd)); |
89bc7a0f | 206 | ht->rnd_initialized = true; |
bf0857ea | 207 | } |
1da177e4 | 208 | |
39b46fc6 | 209 | if (ht->cfg.max && ht->count >= ht->cfg.max) { |
1da177e4 | 210 | /* FIXME: do something. question is what.. */ |
e87cc472 | 211 | net_err_ratelimited("max count of %u reached\n", ht->cfg.max); |
02e4eb75 ED |
212 | ent = NULL; |
213 | } else | |
214 | ent = kmem_cache_alloc(hashlimit_cachep, GFP_ATOMIC); | |
0a9ee813 | 215 | if (ent) { |
02e4eb75 ED |
216 | memcpy(&ent->dst, dst, sizeof(ent->dst)); |
217 | spin_lock_init(&ent->lock); | |
1da177e4 | 218 | |
02e4eb75 ED |
219 | spin_lock(&ent->lock); |
220 | hlist_add_head_rcu(&ent->node, &ht->hash[hash_dst(ht, dst)]); | |
221 | ht->count++; | |
222 | } | |
223 | spin_unlock(&ht->lock); | |
1da177e4 LT |
224 | return ent; |
225 | } | |
226 | ||
02e4eb75 ED |
227 | static void dsthash_free_rcu(struct rcu_head *head) |
228 | { | |
229 | struct dsthash_ent *ent = container_of(head, struct dsthash_ent, rcu); | |
230 | ||
231 | kmem_cache_free(hashlimit_cachep, ent); | |
232 | } | |
233 | ||
39b46fc6 PM |
234 | static inline void |
235 | dsthash_free(struct xt_hashlimit_htable *ht, struct dsthash_ent *ent) | |
1da177e4 | 236 | { |
02e4eb75 ED |
237 | hlist_del_rcu(&ent->node); |
238 | call_rcu_bh(&ent->rcu, dsthash_free_rcu); | |
39b46fc6 | 239 | ht->count--; |
1da177e4 | 240 | } |
7bd8490e | 241 | static void htable_gc(struct work_struct *work); |
1da177e4 | 242 | |
11d5f157 VP |
243 | static int htable_create(struct net *net, struct hashlimit_cfg2 *cfg, |
244 | const char *name, u_int8_t family, | |
245 | struct xt_hashlimit_htable **out_hinfo, | |
246 | int revision) | |
09e410de | 247 | { |
e89fc3f1 | 248 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); |
09e410de | 249 | struct xt_hashlimit_htable *hinfo; |
11d5f157 VP |
250 | unsigned int size, i; |
251 | int ret; | |
09e410de | 252 | |
11d5f157 VP |
253 | if (cfg->size) { |
254 | size = cfg->size; | |
09e410de | 255 | } else { |
4481374c | 256 | size = (totalram_pages << PAGE_SHIFT) / 16384 / |
09e410de | 257 | sizeof(struct list_head); |
4481374c | 258 | if (totalram_pages > 1024 * 1024 * 1024 / PAGE_SIZE) |
09e410de JE |
259 | size = 8192; |
260 | if (size < 16) | |
261 | size = 16; | |
262 | } | |
263 | /* FIXME: don't use vmalloc() here or anywhere else -HW */ | |
264 | hinfo = vmalloc(sizeof(struct xt_hashlimit_htable) + | |
265 | sizeof(struct list_head) * size); | |
85bc3f38 | 266 | if (hinfo == NULL) |
4a5a5c73 | 267 | return -ENOMEM; |
11d5f157 | 268 | *out_hinfo = hinfo; |
09e410de JE |
269 | |
270 | /* copy match config into hashtable config */ | |
11d5f157 VP |
271 | ret = cfg_copy(&hinfo->cfg, (void *)cfg, 2); |
272 | ||
273 | if (ret) | |
274 | return ret; | |
275 | ||
09e410de JE |
276 | hinfo->cfg.size = size; |
277 | if (hinfo->cfg.max == 0) | |
278 | hinfo->cfg.max = 8 * hinfo->cfg.size; | |
279 | else if (hinfo->cfg.max < hinfo->cfg.size) | |
280 | hinfo->cfg.max = hinfo->cfg.size; | |
281 | ||
282 | for (i = 0; i < hinfo->cfg.size; i++) | |
283 | INIT_HLIST_HEAD(&hinfo->hash[i]); | |
284 | ||
2eff25c1 | 285 | hinfo->use = 1; |
09e410de JE |
286 | hinfo->count = 0; |
287 | hinfo->family = family; | |
89bc7a0f | 288 | hinfo->rnd_initialized = false; |
11d5f157 | 289 | hinfo->name = kstrdup(name, GFP_KERNEL); |
14b872f0 AV |
290 | if (!hinfo->name) { |
291 | vfree(hinfo); | |
292 | return -ENOMEM; | |
293 | } | |
09e410de JE |
294 | spin_lock_init(&hinfo->lock); |
295 | ||
11d5f157 | 296 | hinfo->pde = proc_create_data(name, 0, |
ee999d8b | 297 | (family == NFPROTO_IPV4) ? |
e89fc3f1 | 298 | hashlimit_net->ipt_hashlimit : hashlimit_net->ip6t_hashlimit, |
11d5f157 VP |
299 | (revision == 1) ? &dl_file_ops_v1 : &dl_file_ops, |
300 | hinfo); | |
09e410de | 301 | if (hinfo->pde == NULL) { |
14b872f0 | 302 | kfree(hinfo->name); |
09e410de | 303 | vfree(hinfo); |
4a5a5c73 | 304 | return -ENOMEM; |
09e410de | 305 | } |
e89fc3f1 | 306 | hinfo->net = net; |
09e410de | 307 | |
7bd8490e ED |
308 | INIT_DEFERRABLE_WORK(&hinfo->gc_work, htable_gc); |
309 | queue_delayed_work(system_power_efficient_wq, &hinfo->gc_work, | |
310 | msecs_to_jiffies(hinfo->cfg.gc_interval)); | |
09e410de | 311 | |
e89fc3f1 | 312 | hlist_add_head(&hinfo->node, &hashlimit_net->htables); |
09e410de JE |
313 | |
314 | return 0; | |
315 | } | |
316 | ||
a47362a2 JE |
317 | static bool select_all(const struct xt_hashlimit_htable *ht, |
318 | const struct dsthash_ent *he) | |
1da177e4 LT |
319 | { |
320 | return 1; | |
321 | } | |
322 | ||
a47362a2 JE |
323 | static bool select_gc(const struct xt_hashlimit_htable *ht, |
324 | const struct dsthash_ent *he) | |
1da177e4 | 325 | { |
cbebc51f | 326 | return time_after_eq(jiffies, he->expires); |
1da177e4 LT |
327 | } |
328 | ||
39b46fc6 | 329 | static void htable_selective_cleanup(struct xt_hashlimit_htable *ht, |
a47362a2 JE |
330 | bool (*select)(const struct xt_hashlimit_htable *ht, |
331 | const struct dsthash_ent *he)) | |
1da177e4 | 332 | { |
39b46fc6 | 333 | unsigned int i; |
1da177e4 | 334 | |
1da177e4 LT |
335 | for (i = 0; i < ht->cfg.size; i++) { |
336 | struct dsthash_ent *dh; | |
b67bfe0d | 337 | struct hlist_node *n; |
7bd8490e ED |
338 | |
339 | spin_lock_bh(&ht->lock); | |
b67bfe0d | 340 | hlist_for_each_entry_safe(dh, n, &ht->hash[i], node) { |
1da177e4 | 341 | if ((*select)(ht, dh)) |
39b46fc6 | 342 | dsthash_free(ht, dh); |
1da177e4 | 343 | } |
7bd8490e ED |
344 | spin_unlock_bh(&ht->lock); |
345 | cond_resched(); | |
1da177e4 | 346 | } |
1da177e4 LT |
347 | } |
348 | ||
7bd8490e | 349 | static void htable_gc(struct work_struct *work) |
1da177e4 | 350 | { |
7bd8490e ED |
351 | struct xt_hashlimit_htable *ht; |
352 | ||
353 | ht = container_of(work, struct xt_hashlimit_htable, gc_work.work); | |
1da177e4 LT |
354 | |
355 | htable_selective_cleanup(ht, select_gc); | |
356 | ||
7bd8490e ED |
357 | queue_delayed_work(system_power_efficient_wq, |
358 | &ht->gc_work, msecs_to_jiffies(ht->cfg.gc_interval)); | |
1da177e4 LT |
359 | } |
360 | ||
b4ef4ce0 | 361 | static void htable_remove_proc_entry(struct xt_hashlimit_htable *hinfo) |
1da177e4 | 362 | { |
e89fc3f1 AD |
363 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(hinfo->net); |
364 | struct proc_dir_entry *parent; | |
365 | ||
e89fc3f1 AD |
366 | if (hinfo->family == NFPROTO_IPV4) |
367 | parent = hashlimit_net->ipt_hashlimit; | |
368 | else | |
369 | parent = hashlimit_net->ip6t_hashlimit; | |
32263dd1 | 370 | |
b4ef4ce0 | 371 | if (parent != NULL) |
14b872f0 | 372 | remove_proc_entry(hinfo->name, parent); |
b4ef4ce0 | 373 | } |
32263dd1 | 374 | |
b4ef4ce0 SP |
375 | static void htable_destroy(struct xt_hashlimit_htable *hinfo) |
376 | { | |
7bd8490e | 377 | cancel_delayed_work_sync(&hinfo->gc_work); |
b4ef4ce0 | 378 | htable_remove_proc_entry(hinfo); |
1da177e4 | 379 | htable_selective_cleanup(hinfo, select_all); |
14b872f0 | 380 | kfree(hinfo->name); |
1da177e4 LT |
381 | vfree(hinfo); |
382 | } | |
383 | ||
e89fc3f1 AD |
384 | static struct xt_hashlimit_htable *htable_find_get(struct net *net, |
385 | const char *name, | |
76108cea | 386 | u_int8_t family) |
1da177e4 | 387 | { |
e89fc3f1 | 388 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); |
39b46fc6 | 389 | struct xt_hashlimit_htable *hinfo; |
1da177e4 | 390 | |
b67bfe0d | 391 | hlist_for_each_entry(hinfo, &hashlimit_net->htables, node) { |
14b872f0 | 392 | if (!strcmp(name, hinfo->name) && |
39b46fc6 | 393 | hinfo->family == family) { |
2eff25c1 | 394 | hinfo->use++; |
1da177e4 LT |
395 | return hinfo; |
396 | } | |
397 | } | |
1da177e4 LT |
398 | return NULL; |
399 | } | |
400 | ||
39b46fc6 | 401 | static void htable_put(struct xt_hashlimit_htable *hinfo) |
1da177e4 | 402 | { |
2eff25c1 PM |
403 | mutex_lock(&hashlimit_mutex); |
404 | if (--hinfo->use == 0) { | |
1da177e4 | 405 | hlist_del(&hinfo->node); |
1da177e4 LT |
406 | htable_destroy(hinfo); |
407 | } | |
2eff25c1 | 408 | mutex_unlock(&hashlimit_mutex); |
1da177e4 LT |
409 | } |
410 | ||
1da177e4 LT |
411 | /* The algorithm used is the Simple Token Bucket Filter (TBF) |
412 | * see net/sched/sch_tbf.c in the linux source tree | |
413 | */ | |
414 | ||
415 | /* Rusty: This is my (non-mathematically-inclined) understanding of | |
416 | this algorithm. The `average rate' in jiffies becomes your initial | |
417 | amount of credit `credit' and the most credit you can ever have | |
418 | `credit_cap'. The `peak rate' becomes the cost of passing the | |
419 | test, `cost'. | |
420 | ||
421 | `prev' tracks the last packet hit: you gain one credit per jiffy. | |
422 | If you get credit balance more than this, the extra credit is | |
423 | discarded. Every time the match passes, you lose `cost' credits; | |
424 | if you don't have that many, the test fails. | |
425 | ||
426 | See Alexey's formal explanation in net/sched/sch_tbf.c. | |
427 | ||
428 | To get the maximum range, we multiply by this factor (ie. you get N | |
429 | credits per jiffy). We want to allow a rate as low as 1 per day | |
430 | (slowest userspace tool allows), which means | |
431 | CREDITS_PER_JIFFY*HZ*60*60*24 < 2^32 ie. | |
432 | */ | |
0dc60a45 | 433 | #define MAX_CPJ_v1 (0xFFFFFFFF / (HZ*60*60*24)) |
1b203c13 | 434 | #define MAX_CPJ (0xFFFFFFFFFFFFFFFFULL / (HZ*60*60*24)) |
1da177e4 LT |
435 | |
436 | /* Repeated shift and or gives us all 1s, final shift and add 1 gives | |
437 | * us the power of 2 below the theoretical max, so GCC simply does a | |
438 | * shift. */ | |
439 | #define _POW2_BELOW2(x) ((x)|((x)>>1)) | |
440 | #define _POW2_BELOW4(x) (_POW2_BELOW2(x)|_POW2_BELOW2((x)>>2)) | |
441 | #define _POW2_BELOW8(x) (_POW2_BELOW4(x)|_POW2_BELOW4((x)>>4)) | |
442 | #define _POW2_BELOW16(x) (_POW2_BELOW8(x)|_POW2_BELOW8((x)>>8)) | |
443 | #define _POW2_BELOW32(x) (_POW2_BELOW16(x)|_POW2_BELOW16((x)>>16)) | |
11d5f157 | 444 | #define _POW2_BELOW64(x) (_POW2_BELOW32(x)|_POW2_BELOW32((x)>>32)) |
1da177e4 | 445 | #define POW2_BELOW32(x) ((_POW2_BELOW32(x)>>1) + 1) |
11d5f157 | 446 | #define POW2_BELOW64(x) ((_POW2_BELOW64(x)>>1) + 1) |
1da177e4 | 447 | |
11d5f157 | 448 | #define CREDITS_PER_JIFFY POW2_BELOW64(MAX_CPJ) |
0dc60a45 | 449 | #define CREDITS_PER_JIFFY_v1 POW2_BELOW32(MAX_CPJ_v1) |
1da177e4 | 450 | |
0197dee7 FW |
451 | /* in byte mode, the lowest possible rate is one packet/second. |
452 | * credit_cap is used as a counter that tells us how many times we can | |
453 | * refill the "credits available" counter when it becomes empty. | |
454 | */ | |
455 | #define MAX_CPJ_BYTES (0xFFFFFFFF / HZ) | |
456 | #define CREDITS_PER_JIFFY_BYTES POW2_BELOW32(MAX_CPJ_BYTES) | |
457 | ||
458 | static u32 xt_hashlimit_len_to_chunks(u32 len) | |
459 | { | |
460 | return (len >> XT_HASHLIMIT_BYTE_SHIFT) + 1; | |
461 | } | |
462 | ||
1da177e4 | 463 | /* Precision saver. */ |
11d5f157 | 464 | static u64 user2credits(u64 user, int revision) |
1da177e4 | 465 | { |
ad5b5576 AB |
466 | u64 scale = (revision == 1) ? |
467 | XT_HASHLIMIT_SCALE : XT_HASHLIMIT_SCALE_v2; | |
468 | u64 cpj = (revision == 1) ? | |
469 | CREDITS_PER_JIFFY_v1 : CREDITS_PER_JIFFY; | |
1da177e4 | 470 | |
ad5b5576 AB |
471 | /* Avoid overflow: divide the constant operands first */ |
472 | if (scale >= HZ * cpj) | |
473 | return div64_u64(user, div64_u64(scale, HZ * cpj)); | |
474 | ||
475 | return user * div64_u64(HZ * cpj, scale); | |
1da177e4 LT |
476 | } |
477 | ||
0197dee7 | 478 | static u32 user2credits_byte(u32 user) |
1da177e4 | 479 | { |
0197dee7 FW |
480 | u64 us = user; |
481 | us *= HZ * CREDITS_PER_JIFFY_BYTES; | |
482 | return (u32) (us >> 32); | |
483 | } | |
484 | ||
11d5f157 VP |
485 | static void rateinfo_recalc(struct dsthash_ent *dh, unsigned long now, |
486 | u32 mode, int revision) | |
0197dee7 FW |
487 | { |
488 | unsigned long delta = now - dh->rateinfo.prev; | |
11d5f157 | 489 | u64 cap, cpj; |
0197dee7 FW |
490 | |
491 | if (delta == 0) | |
492 | return; | |
493 | ||
39b46fc6 | 494 | dh->rateinfo.prev = now; |
0197dee7 FW |
495 | |
496 | if (mode & XT_HASHLIMIT_BYTES) { | |
11d5f157 | 497 | u64 tmp = dh->rateinfo.credit; |
0197dee7 FW |
498 | dh->rateinfo.credit += CREDITS_PER_JIFFY_BYTES * delta; |
499 | cap = CREDITS_PER_JIFFY_BYTES * HZ; | |
500 | if (tmp >= dh->rateinfo.credit) {/* overflow */ | |
501 | dh->rateinfo.credit = cap; | |
502 | return; | |
503 | } | |
504 | } else { | |
11d5f157 VP |
505 | cpj = (revision == 1) ? |
506 | CREDITS_PER_JIFFY_v1 : CREDITS_PER_JIFFY; | |
507 | dh->rateinfo.credit += delta * cpj; | |
0197dee7 FW |
508 | cap = dh->rateinfo.credit_cap; |
509 | } | |
510 | if (dh->rateinfo.credit > cap) | |
511 | dh->rateinfo.credit = cap; | |
39b46fc6 PM |
512 | } |
513 | ||
817e076f | 514 | static void rateinfo_init(struct dsthash_ent *dh, |
11d5f157 | 515 | struct xt_hashlimit_htable *hinfo, int revision) |
817e076f FW |
516 | { |
517 | dh->rateinfo.prev = jiffies; | |
0197dee7 FW |
518 | if (hinfo->cfg.mode & XT_HASHLIMIT_BYTES) { |
519 | dh->rateinfo.credit = CREDITS_PER_JIFFY_BYTES * HZ; | |
520 | dh->rateinfo.cost = user2credits_byte(hinfo->cfg.avg); | |
521 | dh->rateinfo.credit_cap = hinfo->cfg.burst; | |
522 | } else { | |
523 | dh->rateinfo.credit = user2credits(hinfo->cfg.avg * | |
11d5f157 VP |
524 | hinfo->cfg.burst, revision); |
525 | dh->rateinfo.cost = user2credits(hinfo->cfg.avg, revision); | |
0197dee7 FW |
526 | dh->rateinfo.credit_cap = dh->rateinfo.credit; |
527 | } | |
817e076f FW |
528 | } |
529 | ||
09e410de JE |
530 | static inline __be32 maskl(__be32 a, unsigned int l) |
531 | { | |
1b9b70ea | 532 | return l ? htonl(ntohl(a) & ~0 << (32 - l)) : 0; |
09e410de JE |
533 | } |
534 | ||
c0cd1156 | 535 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
09e410de JE |
536 | static void hashlimit_ipv6_mask(__be32 *i, unsigned int p) |
537 | { | |
538 | switch (p) { | |
1b9b70ea | 539 | case 0 ... 31: |
09e410de JE |
540 | i[0] = maskl(i[0], p); |
541 | i[1] = i[2] = i[3] = 0; | |
542 | break; | |
1b9b70ea | 543 | case 32 ... 63: |
09e410de JE |
544 | i[1] = maskl(i[1], p - 32); |
545 | i[2] = i[3] = 0; | |
546 | break; | |
1b9b70ea | 547 | case 64 ... 95: |
09e410de JE |
548 | i[2] = maskl(i[2], p - 64); |
549 | i[3] = 0; | |
8f599229 | 550 | break; |
1b9b70ea | 551 | case 96 ... 127: |
09e410de JE |
552 | i[3] = maskl(i[3], p - 96); |
553 | break; | |
554 | case 128: | |
555 | break; | |
556 | } | |
557 | } | |
3ed5df44 | 558 | #endif |
09e410de | 559 | |
39b46fc6 | 560 | static int |
a47362a2 JE |
561 | hashlimit_init_dst(const struct xt_hashlimit_htable *hinfo, |
562 | struct dsthash_dst *dst, | |
39b46fc6 PM |
563 | const struct sk_buff *skb, unsigned int protoff) |
564 | { | |
565 | __be16 _ports[2], *ports; | |
193b23c5 | 566 | u8 nexthdr; |
aca071c1 | 567 | int poff; |
39b46fc6 PM |
568 | |
569 | memset(dst, 0, sizeof(*dst)); | |
570 | ||
571 | switch (hinfo->family) { | |
ee999d8b | 572 | case NFPROTO_IPV4: |
39b46fc6 | 573 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) |
09e410de JE |
574 | dst->ip.dst = maskl(ip_hdr(skb)->daddr, |
575 | hinfo->cfg.dstmask); | |
39b46fc6 | 576 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) |
09e410de JE |
577 | dst->ip.src = maskl(ip_hdr(skb)->saddr, |
578 | hinfo->cfg.srcmask); | |
39b46fc6 PM |
579 | |
580 | if (!(hinfo->cfg.mode & | |
581 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
582 | return 0; | |
eddc9ec5 | 583 | nexthdr = ip_hdr(skb)->protocol; |
39b46fc6 | 584 | break; |
c0cd1156 | 585 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
ee999d8b | 586 | case NFPROTO_IPV6: |
412662d2 SR |
587 | { |
588 | __be16 frag_off; | |
589 | ||
09e410de JE |
590 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DIP) { |
591 | memcpy(&dst->ip6.dst, &ipv6_hdr(skb)->daddr, | |
592 | sizeof(dst->ip6.dst)); | |
593 | hashlimit_ipv6_mask(dst->ip6.dst, hinfo->cfg.dstmask); | |
594 | } | |
595 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SIP) { | |
596 | memcpy(&dst->ip6.src, &ipv6_hdr(skb)->saddr, | |
597 | sizeof(dst->ip6.src)); | |
598 | hashlimit_ipv6_mask(dst->ip6.src, hinfo->cfg.srcmask); | |
599 | } | |
39b46fc6 PM |
600 | |
601 | if (!(hinfo->cfg.mode & | |
602 | (XT_HASHLIMIT_HASH_DPT | XT_HASHLIMIT_HASH_SPT))) | |
603 | return 0; | |
193b23c5 | 604 | nexthdr = ipv6_hdr(skb)->nexthdr; |
75f2811c | 605 | protoff = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr, &frag_off); |
193b23c5 | 606 | if ((int)protoff < 0) |
39b46fc6 PM |
607 | return -1; |
608 | break; | |
412662d2 | 609 | } |
39b46fc6 PM |
610 | #endif |
611 | default: | |
612 | BUG(); | |
613 | return 0; | |
614 | } | |
615 | ||
aca071c1 CG |
616 | poff = proto_ports_offset(nexthdr); |
617 | if (poff >= 0) { | |
618 | ports = skb_header_pointer(skb, protoff + poff, sizeof(_ports), | |
39b46fc6 | 619 | &_ports); |
aca071c1 | 620 | } else { |
39b46fc6 PM |
621 | _ports[0] = _ports[1] = 0; |
622 | ports = _ports; | |
39b46fc6 PM |
623 | } |
624 | if (!ports) | |
625 | return -1; | |
626 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_SPT) | |
627 | dst->src_port = ports[0]; | |
628 | if (hinfo->cfg.mode & XT_HASHLIMIT_HASH_DPT) | |
629 | dst->dst_port = ports[1]; | |
630 | return 0; | |
1da177e4 LT |
631 | } |
632 | ||
0197dee7 FW |
633 | static u32 hashlimit_byte_cost(unsigned int len, struct dsthash_ent *dh) |
634 | { | |
635 | u64 tmp = xt_hashlimit_len_to_chunks(len); | |
636 | tmp = tmp * dh->rateinfo.cost; | |
637 | ||
638 | if (unlikely(tmp > CREDITS_PER_JIFFY_BYTES * HZ)) | |
639 | tmp = CREDITS_PER_JIFFY_BYTES * HZ; | |
640 | ||
641 | if (dh->rateinfo.credit < tmp && dh->rateinfo.credit_cap) { | |
642 | dh->rateinfo.credit_cap--; | |
643 | dh->rateinfo.credit = CREDITS_PER_JIFFY_BYTES * HZ; | |
644 | } | |
645 | return (u32) tmp; | |
646 | } | |
647 | ||
ccb79bdc | 648 | static bool |
11d5f157 VP |
649 | hashlimit_mt_common(const struct sk_buff *skb, struct xt_action_param *par, |
650 | struct xt_hashlimit_htable *hinfo, | |
651 | const struct hashlimit_cfg2 *cfg, int revision) | |
09e410de | 652 | { |
09e410de JE |
653 | unsigned long now = jiffies; |
654 | struct dsthash_ent *dh; | |
655 | struct dsthash_dst dst; | |
09181842 | 656 | bool race = false; |
11d5f157 | 657 | u64 cost; |
09e410de | 658 | |
f7108a20 | 659 | if (hashlimit_init_dst(hinfo, &dst, skb, par->thoff) < 0) |
09e410de JE |
660 | goto hotdrop; |
661 | ||
02e4eb75 | 662 | rcu_read_lock_bh(); |
09e410de JE |
663 | dh = dsthash_find(hinfo, &dst); |
664 | if (dh == NULL) { | |
09181842 | 665 | dh = dsthash_alloc_init(hinfo, &dst, &race); |
09e410de | 666 | if (dh == NULL) { |
02e4eb75 | 667 | rcu_read_unlock_bh(); |
09e410de | 668 | goto hotdrop; |
09181842 PNA |
669 | } else if (race) { |
670 | /* Already got an entry, update expiration timeout */ | |
671 | dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire); | |
11d5f157 | 672 | rateinfo_recalc(dh, now, hinfo->cfg.mode, revision); |
09181842 PNA |
673 | } else { |
674 | dh->expires = jiffies + msecs_to_jiffies(hinfo->cfg.expire); | |
11d5f157 | 675 | rateinfo_init(dh, hinfo, revision); |
09e410de | 676 | } |
09e410de JE |
677 | } else { |
678 | /* update expiration timeout */ | |
679 | dh->expires = now + msecs_to_jiffies(hinfo->cfg.expire); | |
11d5f157 | 680 | rateinfo_recalc(dh, now, hinfo->cfg.mode, revision); |
09e410de JE |
681 | } |
682 | ||
11d5f157 | 683 | if (cfg->mode & XT_HASHLIMIT_BYTES) |
0197dee7 FW |
684 | cost = hashlimit_byte_cost(skb->len, dh); |
685 | else | |
686 | cost = dh->rateinfo.cost; | |
687 | ||
688 | if (dh->rateinfo.credit >= cost) { | |
09e410de | 689 | /* below the limit */ |
0197dee7 | 690 | dh->rateinfo.credit -= cost; |
02e4eb75 ED |
691 | spin_unlock(&dh->lock); |
692 | rcu_read_unlock_bh(); | |
11d5f157 | 693 | return !(cfg->mode & XT_HASHLIMIT_INVERT); |
09e410de JE |
694 | } |
695 | ||
02e4eb75 ED |
696 | spin_unlock(&dh->lock); |
697 | rcu_read_unlock_bh(); | |
09e410de | 698 | /* default match is underlimit - so over the limit, we need to invert */ |
11d5f157 | 699 | return cfg->mode & XT_HASHLIMIT_INVERT; |
09e410de JE |
700 | |
701 | hotdrop: | |
b4ba2611 | 702 | par->hotdrop = true; |
09e410de JE |
703 | return false; |
704 | } | |
705 | ||
11d5f157 VP |
706 | static bool |
707 | hashlimit_mt_v1(const struct sk_buff *skb, struct xt_action_param *par) | |
708 | { | |
709 | const struct xt_hashlimit_mtinfo1 *info = par->matchinfo; | |
710 | struct xt_hashlimit_htable *hinfo = info->hinfo; | |
711 | struct hashlimit_cfg2 cfg = {}; | |
712 | int ret; | |
713 | ||
714 | ret = cfg_copy(&cfg, (void *)&info->cfg, 1); | |
715 | ||
716 | if (ret) | |
717 | return ret; | |
718 | ||
719 | return hashlimit_mt_common(skb, par, hinfo, &cfg, 1); | |
720 | } | |
721 | ||
722 | static bool | |
723 | hashlimit_mt(const struct sk_buff *skb, struct xt_action_param *par) | |
724 | { | |
725 | const struct xt_hashlimit_mtinfo2 *info = par->matchinfo; | |
726 | struct xt_hashlimit_htable *hinfo = info->hinfo; | |
727 | ||
728 | return hashlimit_mt_common(skb, par, hinfo, &info->cfg, 2); | |
729 | } | |
730 | ||
731 | static int hashlimit_mt_check_common(const struct xt_mtchk_param *par, | |
732 | struct xt_hashlimit_htable **hinfo, | |
733 | struct hashlimit_cfg2 *cfg, | |
734 | const char *name, int revision) | |
09e410de | 735 | { |
e89fc3f1 | 736 | struct net *net = par->net; |
4a5a5c73 | 737 | int ret; |
09e410de | 738 | |
11d5f157 | 739 | if (cfg->gc_interval == 0 || cfg->expire == 0) |
bd414ee6 | 740 | return -EINVAL; |
aa5fa318 | 741 | if (par->family == NFPROTO_IPV4) { |
11d5f157 | 742 | if (cfg->srcmask > 32 || cfg->dstmask > 32) |
bd414ee6 | 743 | return -EINVAL; |
09e410de | 744 | } else { |
11d5f157 | 745 | if (cfg->srcmask > 128 || cfg->dstmask > 128) |
bd414ee6 | 746 | return -EINVAL; |
09e410de JE |
747 | } |
748 | ||
11d5f157 | 749 | if (cfg->mode & ~XT_HASHLIMIT_ALL) { |
0197dee7 | 750 | pr_info("Unknown mode mask %X, kernel too old?\n", |
11d5f157 | 751 | cfg->mode); |
0197dee7 FW |
752 | return -EINVAL; |
753 | } | |
754 | ||
755 | /* Check for overflow. */ | |
11d5f157 VP |
756 | if (cfg->mode & XT_HASHLIMIT_BYTES) { |
757 | if (user2credits_byte(cfg->avg) == 0) { | |
758 | pr_info("overflow, rate too high: %llu\n", cfg->avg); | |
0197dee7 FW |
759 | return -EINVAL; |
760 | } | |
11d5f157 VP |
761 | } else if (cfg->burst == 0 || |
762 | user2credits(cfg->avg * cfg->burst, revision) < | |
763 | user2credits(cfg->avg, revision)) { | |
764 | pr_info("overflow, try lower: %llu/%llu\n", | |
765 | cfg->avg, cfg->burst); | |
0197dee7 FW |
766 | return -ERANGE; |
767 | } | |
768 | ||
2eff25c1 | 769 | mutex_lock(&hashlimit_mutex); |
11d5f157 VP |
770 | *hinfo = htable_find_get(net, name, par->family); |
771 | if (*hinfo == NULL) { | |
772 | ret = htable_create(net, cfg, name, par->family, | |
773 | hinfo, revision); | |
4a5a5c73 JE |
774 | if (ret < 0) { |
775 | mutex_unlock(&hashlimit_mutex); | |
776 | return ret; | |
777 | } | |
09e410de | 778 | } |
2eff25c1 | 779 | mutex_unlock(&hashlimit_mutex); |
11d5f157 | 780 | |
bd414ee6 | 781 | return 0; |
09e410de JE |
782 | } |
783 | ||
11d5f157 VP |
784 | static int hashlimit_mt_check_v1(const struct xt_mtchk_param *par) |
785 | { | |
786 | struct xt_hashlimit_mtinfo1 *info = par->matchinfo; | |
787 | struct hashlimit_cfg2 cfg = {}; | |
788 | int ret; | |
789 | ||
790 | if (info->name[sizeof(info->name) - 1] != '\0') | |
791 | return -EINVAL; | |
792 | ||
793 | ret = cfg_copy(&cfg, (void *)&info->cfg, 1); | |
794 | ||
795 | if (ret) | |
796 | return ret; | |
797 | ||
798 | return hashlimit_mt_check_common(par, &info->hinfo, | |
799 | &cfg, info->name, 1); | |
800 | } | |
801 | ||
802 | static int hashlimit_mt_check(const struct xt_mtchk_param *par) | |
803 | { | |
804 | struct xt_hashlimit_mtinfo2 *info = par->matchinfo; | |
805 | ||
806 | if (info->name[sizeof(info->name) - 1] != '\0') | |
807 | return -EINVAL; | |
808 | ||
809 | return hashlimit_mt_check_common(par, &info->hinfo, &info->cfg, | |
810 | info->name, 2); | |
811 | } | |
812 | ||
0dc60a45 | 813 | static void hashlimit_mt_destroy_v1(const struct xt_mtdtor_param *par) |
09e410de | 814 | { |
6be3d859 | 815 | const struct xt_hashlimit_mtinfo1 *info = par->matchinfo; |
09e410de JE |
816 | |
817 | htable_put(info->hinfo); | |
818 | } | |
819 | ||
11d5f157 VP |
820 | static void hashlimit_mt_destroy(const struct xt_mtdtor_param *par) |
821 | { | |
822 | const struct xt_hashlimit_mtinfo2 *info = par->matchinfo; | |
823 | ||
824 | htable_put(info->hinfo); | |
825 | } | |
826 | ||
d3c5ee6d | 827 | static struct xt_match hashlimit_mt_reg[] __read_mostly = { |
09e410de JE |
828 | { |
829 | .name = "hashlimit", | |
830 | .revision = 1, | |
ee999d8b | 831 | .family = NFPROTO_IPV4, |
0dc60a45 | 832 | .match = hashlimit_mt_v1, |
09e410de | 833 | .matchsize = sizeof(struct xt_hashlimit_mtinfo1), |
ec231890 | 834 | .usersize = offsetof(struct xt_hashlimit_mtinfo1, hinfo), |
0dc60a45 VP |
835 | .checkentry = hashlimit_mt_check_v1, |
836 | .destroy = hashlimit_mt_destroy_v1, | |
09e410de JE |
837 | .me = THIS_MODULE, |
838 | }, | |
11d5f157 VP |
839 | { |
840 | .name = "hashlimit", | |
841 | .revision = 2, | |
842 | .family = NFPROTO_IPV4, | |
843 | .match = hashlimit_mt, | |
844 | .matchsize = sizeof(struct xt_hashlimit_mtinfo2), | |
ec231890 | 845 | .usersize = offsetof(struct xt_hashlimit_mtinfo2, hinfo), |
11d5f157 VP |
846 | .checkentry = hashlimit_mt_check, |
847 | .destroy = hashlimit_mt_destroy, | |
848 | .me = THIS_MODULE, | |
849 | }, | |
c0cd1156 | 850 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
09e410de JE |
851 | { |
852 | .name = "hashlimit", | |
853 | .revision = 1, | |
ee999d8b | 854 | .family = NFPROTO_IPV6, |
0dc60a45 | 855 | .match = hashlimit_mt_v1, |
09e410de | 856 | .matchsize = sizeof(struct xt_hashlimit_mtinfo1), |
ec231890 | 857 | .usersize = offsetof(struct xt_hashlimit_mtinfo1, hinfo), |
0dc60a45 VP |
858 | .checkentry = hashlimit_mt_check_v1, |
859 | .destroy = hashlimit_mt_destroy_v1, | |
09e410de JE |
860 | .me = THIS_MODULE, |
861 | }, | |
11d5f157 VP |
862 | { |
863 | .name = "hashlimit", | |
864 | .revision = 2, | |
865 | .family = NFPROTO_IPV6, | |
866 | .match = hashlimit_mt, | |
867 | .matchsize = sizeof(struct xt_hashlimit_mtinfo2), | |
ec231890 | 868 | .usersize = offsetof(struct xt_hashlimit_mtinfo2, hinfo), |
11d5f157 VP |
869 | .checkentry = hashlimit_mt_check, |
870 | .destroy = hashlimit_mt_destroy, | |
871 | .me = THIS_MODULE, | |
872 | }, | |
7b21e09d | 873 | #endif |
1da177e4 LT |
874 | }; |
875 | ||
876 | /* PROC stuff */ | |
1da177e4 | 877 | static void *dl_seq_start(struct seq_file *s, loff_t *pos) |
f4f6fb71 | 878 | __acquires(htable->lock) |
1da177e4 | 879 | { |
a1004d8e | 880 | struct xt_hashlimit_htable *htable = s->private; |
1da177e4 LT |
881 | unsigned int *bucket; |
882 | ||
883 | spin_lock_bh(&htable->lock); | |
884 | if (*pos >= htable->cfg.size) | |
885 | return NULL; | |
886 | ||
887 | bucket = kmalloc(sizeof(unsigned int), GFP_ATOMIC); | |
888 | if (!bucket) | |
889 | return ERR_PTR(-ENOMEM); | |
890 | ||
891 | *bucket = *pos; | |
892 | return bucket; | |
893 | } | |
894 | ||
895 | static void *dl_seq_next(struct seq_file *s, void *v, loff_t *pos) | |
896 | { | |
a1004d8e | 897 | struct xt_hashlimit_htable *htable = s->private; |
1da177e4 LT |
898 | unsigned int *bucket = (unsigned int *)v; |
899 | ||
900 | *pos = ++(*bucket); | |
901 | if (*pos >= htable->cfg.size) { | |
902 | kfree(v); | |
903 | return NULL; | |
904 | } | |
905 | return bucket; | |
906 | } | |
907 | ||
908 | static void dl_seq_stop(struct seq_file *s, void *v) | |
f4f6fb71 | 909 | __releases(htable->lock) |
1da177e4 | 910 | { |
a1004d8e | 911 | struct xt_hashlimit_htable *htable = s->private; |
1da177e4 LT |
912 | unsigned int *bucket = (unsigned int *)v; |
913 | ||
55e0d7cf ED |
914 | if (!IS_ERR(bucket)) |
915 | kfree(bucket); | |
1da177e4 LT |
916 | spin_unlock_bh(&htable->lock); |
917 | } | |
918 | ||
11d5f157 VP |
919 | static void dl_seq_print(struct dsthash_ent *ent, u_int8_t family, |
920 | struct seq_file *s) | |
1da177e4 | 921 | { |
39b46fc6 | 922 | switch (family) { |
ee999d8b | 923 | case NFPROTO_IPV4: |
11d5f157 | 924 | seq_printf(s, "%ld %pI4:%u->%pI4:%u %llu %llu %llu\n", |
e71456ae SRRH |
925 | (long)(ent->expires - jiffies)/HZ, |
926 | &ent->dst.ip.src, | |
927 | ntohs(ent->dst.src_port), | |
928 | &ent->dst.ip.dst, | |
929 | ntohs(ent->dst.dst_port), | |
930 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
931 | ent->rateinfo.cost); | |
02e4eb75 | 932 | break; |
c0cd1156 | 933 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
ee999d8b | 934 | case NFPROTO_IPV6: |
11d5f157 | 935 | seq_printf(s, "%ld %pI6:%u->%pI6:%u %llu %llu %llu\n", |
e71456ae SRRH |
936 | (long)(ent->expires - jiffies)/HZ, |
937 | &ent->dst.ip6.src, | |
938 | ntohs(ent->dst.src_port), | |
939 | &ent->dst.ip6.dst, | |
940 | ntohs(ent->dst.dst_port), | |
941 | ent->rateinfo.credit, ent->rateinfo.credit_cap, | |
942 | ent->rateinfo.cost); | |
02e4eb75 | 943 | break; |
7b21e09d | 944 | #endif |
39b46fc6 PM |
945 | default: |
946 | BUG(); | |
39b46fc6 | 947 | } |
11d5f157 VP |
948 | } |
949 | ||
950 | static int dl_seq_real_show_v1(struct dsthash_ent *ent, u_int8_t family, | |
951 | struct seq_file *s) | |
952 | { | |
953 | const struct xt_hashlimit_htable *ht = s->private; | |
954 | ||
955 | spin_lock(&ent->lock); | |
956 | /* recalculate to show accurate numbers */ | |
957 | rateinfo_recalc(ent, jiffies, ht->cfg.mode, 1); | |
958 | ||
959 | dl_seq_print(ent, family, s); | |
960 | ||
961 | spin_unlock(&ent->lock); | |
962 | return seq_has_overflowed(s); | |
963 | } | |
964 | ||
965 | static int dl_seq_real_show(struct dsthash_ent *ent, u_int8_t family, | |
966 | struct seq_file *s) | |
967 | { | |
968 | const struct xt_hashlimit_htable *ht = s->private; | |
969 | ||
970 | spin_lock(&ent->lock); | |
971 | /* recalculate to show accurate numbers */ | |
972 | rateinfo_recalc(ent, jiffies, ht->cfg.mode, 2); | |
973 | ||
974 | dl_seq_print(ent, family, s); | |
975 | ||
02e4eb75 | 976 | spin_unlock(&ent->lock); |
e71456ae | 977 | return seq_has_overflowed(s); |
1da177e4 LT |
978 | } |
979 | ||
0dc60a45 | 980 | static int dl_seq_show_v1(struct seq_file *s, void *v) |
1da177e4 | 981 | { |
a1004d8e | 982 | struct xt_hashlimit_htable *htable = s->private; |
1da177e4 LT |
983 | unsigned int *bucket = (unsigned int *)v; |
984 | struct dsthash_ent *ent; | |
1da177e4 | 985 | |
39b46fc6 | 986 | if (!hlist_empty(&htable->hash[*bucket])) { |
b67bfe0d | 987 | hlist_for_each_entry(ent, &htable->hash[*bucket], node) |
0dc60a45 | 988 | if (dl_seq_real_show_v1(ent, htable->family, s)) |
683a04ce | 989 | return -1; |
39b46fc6 | 990 | } |
1da177e4 LT |
991 | return 0; |
992 | } | |
993 | ||
11d5f157 VP |
994 | static int dl_seq_show(struct seq_file *s, void *v) |
995 | { | |
996 | struct xt_hashlimit_htable *htable = s->private; | |
997 | unsigned int *bucket = (unsigned int *)v; | |
998 | struct dsthash_ent *ent; | |
999 | ||
1000 | if (!hlist_empty(&htable->hash[*bucket])) { | |
1001 | hlist_for_each_entry(ent, &htable->hash[*bucket], node) | |
1002 | if (dl_seq_real_show(ent, htable->family, s)) | |
1003 | return -1; | |
1004 | } | |
1005 | return 0; | |
1006 | } | |
1007 | ||
0dc60a45 | 1008 | static const struct seq_operations dl_seq_ops_v1 = { |
1da177e4 LT |
1009 | .start = dl_seq_start, |
1010 | .next = dl_seq_next, | |
1011 | .stop = dl_seq_stop, | |
0dc60a45 | 1012 | .show = dl_seq_show_v1 |
1da177e4 LT |
1013 | }; |
1014 | ||
11d5f157 VP |
1015 | static const struct seq_operations dl_seq_ops = { |
1016 | .start = dl_seq_start, | |
1017 | .next = dl_seq_next, | |
1018 | .stop = dl_seq_stop, | |
1019 | .show = dl_seq_show | |
1020 | }; | |
1021 | ||
0dc60a45 | 1022 | static int dl_proc_open_v1(struct inode *inode, struct file *file) |
1da177e4 | 1023 | { |
0dc60a45 | 1024 | int ret = seq_open(file, &dl_seq_ops_v1); |
1da177e4 LT |
1025 | |
1026 | if (!ret) { | |
1027 | struct seq_file *sf = file->private_data; | |
d9dda78b | 1028 | sf->private = PDE_DATA(inode); |
1da177e4 LT |
1029 | } |
1030 | return ret; | |
1031 | } | |
1032 | ||
11d5f157 VP |
1033 | static int dl_proc_open(struct inode *inode, struct file *file) |
1034 | { | |
1035 | int ret = seq_open(file, &dl_seq_ops); | |
1036 | ||
1037 | if (!ret) { | |
1038 | struct seq_file *sf = file->private_data; | |
1039 | ||
1040 | sf->private = PDE_DATA(inode); | |
1041 | } | |
1042 | return ret; | |
1043 | } | |
1044 | ||
0dc60a45 | 1045 | static const struct file_operations dl_file_ops_v1 = { |
1da177e4 | 1046 | .owner = THIS_MODULE, |
0dc60a45 | 1047 | .open = dl_proc_open_v1, |
1da177e4 LT |
1048 | .read = seq_read, |
1049 | .llseek = seq_lseek, | |
1050 | .release = seq_release | |
1051 | }; | |
1052 | ||
11d5f157 VP |
1053 | static const struct file_operations dl_file_ops = { |
1054 | .owner = THIS_MODULE, | |
1055 | .open = dl_proc_open, | |
1056 | .read = seq_read, | |
1057 | .llseek = seq_lseek, | |
1058 | .release = seq_release | |
1059 | }; | |
1060 | ||
e89fc3f1 AD |
1061 | static int __net_init hashlimit_proc_net_init(struct net *net) |
1062 | { | |
1063 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); | |
1064 | ||
1065 | hashlimit_net->ipt_hashlimit = proc_mkdir("ipt_hashlimit", net->proc_net); | |
1066 | if (!hashlimit_net->ipt_hashlimit) | |
1067 | return -ENOMEM; | |
c0cd1156 | 1068 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
e89fc3f1 AD |
1069 | hashlimit_net->ip6t_hashlimit = proc_mkdir("ip6t_hashlimit", net->proc_net); |
1070 | if (!hashlimit_net->ip6t_hashlimit) { | |
ece31ffd | 1071 | remove_proc_entry("ipt_hashlimit", net->proc_net); |
e89fc3f1 AD |
1072 | return -ENOMEM; |
1073 | } | |
1074 | #endif | |
1075 | return 0; | |
1076 | } | |
1077 | ||
1078 | static void __net_exit hashlimit_proc_net_exit(struct net *net) | |
1079 | { | |
32263dd1 | 1080 | struct xt_hashlimit_htable *hinfo; |
32263dd1 VL |
1081 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); |
1082 | ||
b4ef4ce0 SP |
1083 | /* hashlimit_net_exit() is called before hashlimit_mt_destroy(). |
1084 | * Make sure that the parent ipt_hashlimit and ip6t_hashlimit proc | |
1085 | * entries is empty before trying to remove it. | |
32263dd1 VL |
1086 | */ |
1087 | mutex_lock(&hashlimit_mutex); | |
b67bfe0d | 1088 | hlist_for_each_entry(hinfo, &hashlimit_net->htables, node) |
b4ef4ce0 | 1089 | htable_remove_proc_entry(hinfo); |
32263dd1 VL |
1090 | hashlimit_net->ipt_hashlimit = NULL; |
1091 | hashlimit_net->ip6t_hashlimit = NULL; | |
1092 | mutex_unlock(&hashlimit_mutex); | |
1093 | ||
ece31ffd | 1094 | remove_proc_entry("ipt_hashlimit", net->proc_net); |
c0cd1156 | 1095 | #if IS_ENABLED(CONFIG_IP6_NF_IPTABLES) |
ece31ffd | 1096 | remove_proc_entry("ip6t_hashlimit", net->proc_net); |
e89fc3f1 AD |
1097 | #endif |
1098 | } | |
1099 | ||
1100 | static int __net_init hashlimit_net_init(struct net *net) | |
1101 | { | |
1102 | struct hashlimit_net *hashlimit_net = hashlimit_pernet(net); | |
1103 | ||
1104 | INIT_HLIST_HEAD(&hashlimit_net->htables); | |
1105 | return hashlimit_proc_net_init(net); | |
1106 | } | |
1107 | ||
1108 | static void __net_exit hashlimit_net_exit(struct net *net) | |
1109 | { | |
e89fc3f1 AD |
1110 | hashlimit_proc_net_exit(net); |
1111 | } | |
1112 | ||
1113 | static struct pernet_operations hashlimit_net_ops = { | |
1114 | .init = hashlimit_net_init, | |
1115 | .exit = hashlimit_net_exit, | |
1116 | .id = &hashlimit_net_id, | |
1117 | .size = sizeof(struct hashlimit_net), | |
1118 | }; | |
1119 | ||
d3c5ee6d | 1120 | static int __init hashlimit_mt_init(void) |
1da177e4 | 1121 | { |
39b46fc6 | 1122 | int err; |
1da177e4 | 1123 | |
e89fc3f1 AD |
1124 | err = register_pernet_subsys(&hashlimit_net_ops); |
1125 | if (err < 0) | |
1126 | return err; | |
d3c5ee6d JE |
1127 | err = xt_register_matches(hashlimit_mt_reg, |
1128 | ARRAY_SIZE(hashlimit_mt_reg)); | |
39b46fc6 PM |
1129 | if (err < 0) |
1130 | goto err1; | |
1da177e4 | 1131 | |
39b46fc6 PM |
1132 | err = -ENOMEM; |
1133 | hashlimit_cachep = kmem_cache_create("xt_hashlimit", | |
1134 | sizeof(struct dsthash_ent), 0, 0, | |
20c2df83 | 1135 | NULL); |
1da177e4 | 1136 | if (!hashlimit_cachep) { |
b167a37c | 1137 | pr_warn("unable to create slab cache\n"); |
39b46fc6 | 1138 | goto err2; |
1da177e4 | 1139 | } |
e89fc3f1 AD |
1140 | return 0; |
1141 | ||
39b46fc6 | 1142 | err2: |
d3c5ee6d | 1143 | xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg)); |
39b46fc6 | 1144 | err1: |
e89fc3f1 | 1145 | unregister_pernet_subsys(&hashlimit_net_ops); |
39b46fc6 | 1146 | return err; |
1da177e4 | 1147 | |
1da177e4 LT |
1148 | } |
1149 | ||
d3c5ee6d | 1150 | static void __exit hashlimit_mt_exit(void) |
1da177e4 | 1151 | { |
d3c5ee6d | 1152 | xt_unregister_matches(hashlimit_mt_reg, ARRAY_SIZE(hashlimit_mt_reg)); |
e89fc3f1 | 1153 | unregister_pernet_subsys(&hashlimit_net_ops); |
02e4eb75 ED |
1154 | |
1155 | rcu_barrier_bh(); | |
1156 | kmem_cache_destroy(hashlimit_cachep); | |
1da177e4 LT |
1157 | } |
1158 | ||
d3c5ee6d JE |
1159 | module_init(hashlimit_mt_init); |
1160 | module_exit(hashlimit_mt_exit); |