~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/netfilter/xt_hashlimit.c

Version: ~ [ linux-5.6 ] ~ [ linux-5.5.13 ] ~ [ linux-5.4.28 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.113 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.174 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.217 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.217 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.82 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

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

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp