1 /* Copyright (C) 2013 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
3 * This program is free software; you can redistribute it and/or modify
4 * it under the terms of the GNU General Public License version 2 as
5 * published by the Free Software Foundation.
8 #ifndef _IP_SET_HASH_GEN_H
9 #define _IP_SET_HASH_GEN_H
11 #include <linux/rcupdate.h>
12 #include <linux/jhash.h>
13 #include <linux/netfilter/ipset/ip_set_timeout.h>
14 #ifndef rcu_dereference_bh
15 #define rcu_dereference_bh(p) rcu_dereference(p)
18 #define rcu_dereference_bh_nfnl(p) rcu_dereference_bh_check(p, 1)
20 /* Hashing which uses arrays to resolve clashing. The hash table is resized
21 * (doubled) when searching becomes too long.
22 * Internally jhash is used with the assumption that the size of the
23 * stored data is a multiple of sizeof(u32). If storage supports timeout,
24 * the timeout field must be the last one in the data structure - that field
25 * is ignored when computing the hash key.
27 * Readers and resizing
29 * Resizing can be triggered by userspace command only, and those
30 * are serialized by the nfnl mutex. During resizing the set is
31 * read-locked, so the only possible concurrent operations are
32 * the kernel side readers. Those must be protected by proper RCU locking.
35 /* Number of elements to store in an initial array block */
36 #define AHASH_INIT_SIZE 4
37 /* Max number of elements to store in an array block */
38 #define AHASH_MAX_SIZE (3*AHASH_INIT_SIZE)
40 /* Max number of elements can be tuned */
41 #ifdef IP_SET_HASH_WITH_MULTI
42 #define AHASH_MAX(h) ((h)->ahash_max)
45 tune_ahash_max(u8 curr, u32 multi)
52 n = curr + AHASH_INIT_SIZE;
53 /* Currently, at listing one hash bucket must fit into a message.
54 * Therefore we have a hard limit here.
56 return n > curr && n <= 64 ? n : curr;
58 #define TUNE_AHASH_MAX(h, multi) \
59 ((h)->ahash_max = tune_ahash_max((h)->ahash_max, multi))
61 #define AHASH_MAX(h) AHASH_MAX_SIZE
62 #define TUNE_AHASH_MAX(h, multi)
67 void *value; /* the array of the values */
68 u8 size; /* size of the array */
69 u8 pos; /* position of the first free entry */
72 /* The hash table: the table size stored here in order to make resizing easy */
74 u8 htable_bits; /* size of hash table == 2^htable_bits */
75 struct hbucket bucket[0]; /* hashtable buckets */
78 #define hbucket(h, i) (&((h)->bucket[i]))
80 #ifndef IPSET_NET_COUNT
81 #define IPSET_NET_COUNT 1
84 /* Book-keeping of the prefixes added to the set */
86 u32 nets[IPSET_NET_COUNT]; /* number of elements per cidr */
87 u8 cidr[IPSET_NET_COUNT]; /* the different cidr values in the set */
90 /* Compute the hash table size */
96 /* We must fit both into u32 in jhash and size_t */
99 hsize = jhash_size(hbits);
100 if ((((size_t)-1) - sizeof(struct htable))/sizeof(struct hbucket)
104 return hsize * sizeof(struct hbucket) + sizeof(struct htable);
107 /* Compute htable_bits from the user input parameter hashsize */
109 htable_bits(u32 hashsize)
111 /* Assume that hashsize == 2^htable_bits */
112 u8 bits = fls(hashsize - 1);
113 if (jhash_size(bits) != hashsize)
114 /* Round up to the first 2^n value */
115 bits = fls(hashsize);
121 hbucket_elem_add(struct hbucket *n, u8 ahash_max, size_t dsize)
123 if (n->pos >= n->size) {
126 if (n->size >= ahash_max)
127 /* Trigger rehashing */
130 tmp = kzalloc((n->size + AHASH_INIT_SIZE) * dsize,
135 memcpy(tmp, n->value, n->size * dsize);
139 n->size += AHASH_INIT_SIZE;
144 #ifdef IP_SET_HASH_WITH_NETS
145 #if IPSET_NET_COUNT > 1
146 #define __CIDR(cidr, i) (cidr[i])
148 #define __CIDR(cidr, i) (cidr)
150 #ifdef IP_SET_HASH_WITH_NETS_PACKED
151 /* When cidr is packed with nomatch, cidr - 1 is stored in the entry */
152 #define CIDR(cidr, i) (__CIDR(cidr, i) + 1)
154 #define CIDR(cidr, i) (__CIDR(cidr, i))
157 #define SET_HOST_MASK(family) (family == AF_INET ? 32 : 128)
159 #ifdef IP_SET_HASH_WITH_MULTI
160 #define NLEN(family) (SET_HOST_MASK(family) + 1)
162 #define NLEN(family) SET_HOST_MASK(family)
166 #define NLEN(family) 0
167 #endif /* IP_SET_HASH_WITH_NETS */
169 #endif /* _IP_SET_HASH_GEN_H */
171 /* Family dependent templates */
174 #undef mtype_data_equal
175 #undef mtype_do_data_match
176 #undef mtype_data_set_flags
177 #undef mtype_data_reset_flags
178 #undef mtype_data_netmask
179 #undef mtype_data_list
180 #undef mtype_data_next
183 #undef mtype_ahash_destroy
184 #undef mtype_ext_cleanup
185 #undef mtype_add_cidr
186 #undef mtype_del_cidr
187 #undef mtype_ahash_memsize
191 #undef mtype_same_set
198 #undef mtype_test_cidrs
207 #undef mtype_data_match
211 #define mtype_data_equal IPSET_TOKEN(MTYPE, _data_equal)
212 #ifdef IP_SET_HASH_WITH_NETS
213 #define mtype_do_data_match IPSET_TOKEN(MTYPE, _do_data_match)
215 #define mtype_do_data_match(d) 1
217 #define mtype_data_set_flags IPSET_TOKEN(MTYPE, _data_set_flags)
218 #define mtype_data_reset_elem IPSET_TOKEN(MTYPE, _data_reset_elem)
219 #define mtype_data_reset_flags IPSET_TOKEN(MTYPE, _data_reset_flags)
220 #define mtype_data_netmask IPSET_TOKEN(MTYPE, _data_netmask)
221 #define mtype_data_list IPSET_TOKEN(MTYPE, _data_list)
222 #define mtype_data_next IPSET_TOKEN(MTYPE, _data_next)
223 #define mtype_elem IPSET_TOKEN(MTYPE, _elem)
224 #define mtype_ahash_destroy IPSET_TOKEN(MTYPE, _ahash_destroy)
225 #define mtype_ext_cleanup IPSET_TOKEN(MTYPE, _ext_cleanup)
226 #define mtype_add_cidr IPSET_TOKEN(MTYPE, _add_cidr)
227 #define mtype_del_cidr IPSET_TOKEN(MTYPE, _del_cidr)
228 #define mtype_ahash_memsize IPSET_TOKEN(MTYPE, _ahash_memsize)
229 #define mtype_flush IPSET_TOKEN(MTYPE, _flush)
230 #define mtype_destroy IPSET_TOKEN(MTYPE, _destroy)
231 #define mtype_gc_init IPSET_TOKEN(MTYPE, _gc_init)
232 #define mtype_same_set IPSET_TOKEN(MTYPE, _same_set)
233 #define mtype_kadt IPSET_TOKEN(MTYPE, _kadt)
234 #define mtype_uadt IPSET_TOKEN(MTYPE, _uadt)
237 #define mtype_elem IPSET_TOKEN(MTYPE, _elem)
238 #define mtype_add IPSET_TOKEN(MTYPE, _add)
239 #define mtype_del IPSET_TOKEN(MTYPE, _del)
240 #define mtype_test_cidrs IPSET_TOKEN(MTYPE, _test_cidrs)
241 #define mtype_test IPSET_TOKEN(MTYPE, _test)
242 #define mtype_expire IPSET_TOKEN(MTYPE, _expire)
243 #define mtype_resize IPSET_TOKEN(MTYPE, _resize)
244 #define mtype_head IPSET_TOKEN(MTYPE, _head)
245 #define mtype_list IPSET_TOKEN(MTYPE, _list)
246 #define mtype_gc IPSET_TOKEN(MTYPE, _gc)
247 #define mtype_variant IPSET_TOKEN(MTYPE, _variant)
248 #define mtype_data_match IPSET_TOKEN(MTYPE, _data_match)
251 #define HKEY_DATALEN sizeof(struct mtype_elem)
254 #define HKEY(data, initval, htable_bits) \
255 (jhash2((u32 *)(data), HKEY_DATALEN/sizeof(u32), initval) \
256 & jhash_mask(htable_bits))
261 /* The generic hash structure */
263 struct htable __rcu *table; /* the hash table */
264 u32 maxelem; /* max elements in the hash */
265 u32 elements; /* current element (vs timeout) */
266 u32 initval; /* random jhash init value */
267 struct timer_list gc; /* garbage collection when timeout enabled */
268 struct mtype_elem next; /* temporary storage for uadd */
269 #ifdef IP_SET_HASH_WITH_MULTI
270 u8 ahash_max; /* max elements in an array block */
272 #ifdef IP_SET_HASH_WITH_NETMASK
273 u8 netmask; /* netmask value for subnets to store */
275 #ifdef IP_SET_HASH_WITH_RBTREE
276 struct rb_root rbtree;
278 #ifdef IP_SET_HASH_WITH_NETS
279 struct net_prefixes nets[0]; /* book-keeping of prefixes */
284 #ifdef IP_SET_HASH_WITH_NETS
285 /* Network cidr size book keeping when the hash stores different
288 mtype_add_cidr(struct htype *h, u8 cidr, u8 nets_length, u8 n)
292 /* Add in increasing prefix order, so larger cidr first */
293 for (i = 0, j = -1; i < nets_length && h->nets[i].nets[n]; i++) {
296 else if (h->nets[i].cidr[n] < cidr)
298 else if (h->nets[i].cidr[n] == cidr) {
299 h->nets[i].nets[n]++;
305 h->nets[i].cidr[n] = h->nets[i - 1].cidr[n];
306 h->nets[i].nets[n] = h->nets[i - 1].nets[n];
309 h->nets[i].cidr[n] = cidr;
310 h->nets[i].nets[n] = 1;
314 mtype_del_cidr(struct htype *h, u8 cidr, u8 nets_length, u8 n)
316 u8 i, j, net_end = nets_length - 1;
318 for (i = 0; i < nets_length; i++) {
319 if (h->nets[i].cidr[n] != cidr)
321 if (h->nets[i].nets[n] > 1 || i == net_end ||
322 h->nets[i + 1].nets[n] == 0) {
323 h->nets[i].nets[n]--;
326 for (j = i; j < net_end && h->nets[j].nets[n]; j++) {
327 h->nets[j].cidr[n] = h->nets[j + 1].cidr[n];
328 h->nets[j].nets[n] = h->nets[j + 1].nets[n];
330 h->nets[j].nets[n] = 0;
336 /* Calculate the actual memory size of the set data */
338 mtype_ahash_memsize(const struct htype *h, const struct htable *t,
339 u8 nets_length, size_t dsize)
342 size_t memsize = sizeof(*h)
344 #ifdef IP_SET_HASH_WITH_NETS
345 + sizeof(struct net_prefixes) * nets_length
347 + jhash_size(t->htable_bits) * sizeof(struct hbucket);
349 for (i = 0; i < jhash_size(t->htable_bits); i++)
350 memsize += t->bucket[i].size * dsize;
355 /* Get the ith element from the array block n */
356 #define ahash_data(n, i, dsize) \
357 ((struct mtype_elem *)((n)->value + ((i) * (dsize))))
360 mtype_ext_cleanup(struct ip_set *set, struct hbucket *n)
364 for (i = 0; i < n->pos; i++)
365 ip_set_ext_destroy(set, ahash_data(n, i, set->dsize));
368 /* Flush a hash type of set: destroy all elements */
370 mtype_flush(struct ip_set *set)
372 struct htype *h = set->data;
377 t = rcu_dereference_bh_nfnl(h->table);
378 for (i = 0; i < jhash_size(t->htable_bits); i++) {
381 if (set->extensions & IPSET_EXT_DESTROY)
382 mtype_ext_cleanup(set, n);
383 n->size = n->pos = 0;
384 /* FIXME: use slab cache */
388 #ifdef IP_SET_HASH_WITH_NETS
389 memset(h->nets, 0, sizeof(struct net_prefixes) * NLEN(set->family));
394 /* Destroy the hashtable part of the set */
396 mtype_ahash_destroy(struct ip_set *set, struct htable *t, bool ext_destroy)
401 for (i = 0; i < jhash_size(t->htable_bits); i++) {
404 if (set->extensions & IPSET_EXT_DESTROY && ext_destroy)
405 mtype_ext_cleanup(set, n);
406 /* FIXME: use slab cache */
414 /* Destroy a hash type of set */
416 mtype_destroy(struct ip_set *set)
418 struct htype *h = set->data;
420 if (set->extensions & IPSET_EXT_TIMEOUT)
421 del_timer_sync(&h->gc);
423 mtype_ahash_destroy(set, rcu_dereference_bh_nfnl(h->table), true);
424 #ifdef IP_SET_HASH_WITH_RBTREE
425 rbtree_destroy(&h->rbtree);
433 mtype_gc_init(struct ip_set *set, void (*gc)(unsigned long ul_set))
435 struct htype *h = set->data;
438 h->gc.data = (unsigned long) set;
440 h->gc.expires = jiffies + IPSET_GC_PERIOD(set->timeout) * HZ;
442 pr_debug("gc initialized, run in every %u\n",
443 IPSET_GC_PERIOD(set->timeout));
447 mtype_same_set(const struct ip_set *a, const struct ip_set *b)
449 const struct htype *x = a->data;
450 const struct htype *y = b->data;
452 /* Resizing changes htable_bits, so we ignore it */
453 return x->maxelem == y->maxelem &&
454 a->timeout == b->timeout &&
455 #ifdef IP_SET_HASH_WITH_NETMASK
456 x->netmask == y->netmask &&
458 a->extensions == b->extensions;
461 /* Delete expired elements from the hashtable */
463 mtype_expire(struct ip_set *set, struct htype *h, u8 nets_length, size_t dsize)
467 struct mtype_elem *data;
470 #ifdef IP_SET_HASH_WITH_NETS
475 t = rcu_dereference_bh(h->table);
476 for (i = 0; i < jhash_size(t->htable_bits); i++) {
478 for (j = 0; j < n->pos; j++) {
479 data = ahash_data(n, j, dsize);
480 if (ip_set_timeout_expired(ext_timeout(data, set))) {
481 pr_debug("expired %u/%u\n", i, j);
482 #ifdef IP_SET_HASH_WITH_NETS
483 for (k = 0; k < IPSET_NET_COUNT; k++)
484 mtype_del_cidr(h, CIDR(data->cidr, k),
487 ip_set_ext_destroy(set, data);
491 ahash_data(n, n->pos - 1, dsize),
497 if (n->pos + AHASH_INIT_SIZE < n->size) {
498 void *tmp = kzalloc((n->size - AHASH_INIT_SIZE)
502 /* Still try to delete expired elements */
504 n->size -= AHASH_INIT_SIZE;
505 memcpy(tmp, n->value, n->size * dsize);
510 rcu_read_unlock_bh();
514 mtype_gc(unsigned long ul_set)
516 struct ip_set *set = (struct ip_set *) ul_set;
517 struct htype *h = set->data;
519 pr_debug("called\n");
520 write_lock_bh(&set->lock);
521 mtype_expire(set, h, NLEN(set->family), set->dsize);
522 write_unlock_bh(&set->lock);
524 h->gc.expires = jiffies + IPSET_GC_PERIOD(set->timeout) * HZ;
528 /* Resize a hash: create a new hash table with doubling the hashsize
529 * and inserting the elements to it. Repeat until we succeed or
530 * fail due to memory pressures. */
532 mtype_resize(struct ip_set *set, bool retried)
534 struct htype *h = set->data;
535 struct htable *t, *orig = rcu_dereference_bh_nfnl(h->table);
536 u8 htable_bits = orig->htable_bits;
537 #ifdef IP_SET_HASH_WITH_NETS
540 struct mtype_elem *data;
541 struct mtype_elem *d;
542 struct hbucket *n, *m;
546 /* Try to cleanup once */
547 if (SET_WITH_TIMEOUT(set) && !retried) {
549 write_lock_bh(&set->lock);
550 mtype_expire(set, set->data, NLEN(set->family), set->dsize);
551 write_unlock_bh(&set->lock);
559 pr_debug("attempt to resize set %s from %u to %u, t %p\n",
560 set->name, orig->htable_bits, htable_bits, orig);
562 /* In case we have plenty of memory :-) */
563 pr_warning("Cannot increase the hashsize of set %s further\n",
565 return -IPSET_ERR_HASH_FULL;
567 t = ip_set_alloc(sizeof(*t)
568 + jhash_size(htable_bits) * sizeof(struct hbucket));
571 t->htable_bits = htable_bits;
573 read_lock_bh(&set->lock);
574 for (i = 0; i < jhash_size(orig->htable_bits); i++) {
575 n = hbucket(orig, i);
576 for (j = 0; j < n->pos; j++) {
577 data = ahash_data(n, j, set->dsize);
578 #ifdef IP_SET_HASH_WITH_NETS
580 mtype_data_reset_flags(data, &flags);
582 m = hbucket(t, HKEY(data, h->initval, htable_bits));
583 ret = hbucket_elem_add(m, AHASH_MAX(h), set->dsize);
585 #ifdef IP_SET_HASH_WITH_NETS
586 mtype_data_reset_flags(data, &flags);
588 read_unlock_bh(&set->lock);
589 mtype_ahash_destroy(set, t, false);
594 d = ahash_data(m, m->pos++, set->dsize);
595 memcpy(d, data, set->dsize);
596 #ifdef IP_SET_HASH_WITH_NETS
597 mtype_data_reset_flags(d, &flags);
602 rcu_assign_pointer(h->table, t);
603 read_unlock_bh(&set->lock);
605 /* Give time to other readers of the set */
606 synchronize_rcu_bh();
608 pr_debug("set %s resized from %u (%p) to %u (%p)\n", set->name,
609 orig->htable_bits, orig, t->htable_bits, t);
610 mtype_ahash_destroy(set, orig, false);
615 /* Add an element to a hash and update the internal counters when succeeded,
616 * otherwise report the proper error code. */
618 mtype_add(struct ip_set *set, void *value, const struct ip_set_ext *ext,
619 struct ip_set_ext *mext, u32 flags)
621 struct htype *h = set->data;
623 const struct mtype_elem *d = value;
624 struct mtype_elem *data;
627 int j = AHASH_MAX(h) + 1;
628 bool flag_exist = flags & IPSET_FLAG_EXIST;
631 if (SET_WITH_TIMEOUT(set) && h->elements >= h->maxelem)
632 /* FIXME: when set is full, we slow down here */
633 mtype_expire(set, h, NLEN(set->family), set->dsize);
635 if (h->elements >= h->maxelem) {
637 pr_warning("Set %s is full, maxelem %u reached\n",
638 set->name, h->maxelem);
639 return -IPSET_ERR_HASH_FULL;
643 t = rcu_dereference_bh(h->table);
644 key = HKEY(value, h->initval, t->htable_bits);
646 for (i = 0; i < n->pos; i++) {
647 data = ahash_data(n, i, set->dsize);
648 if (mtype_data_equal(data, d, &multi)) {
650 (SET_WITH_TIMEOUT(set) &&
651 ip_set_timeout_expired(ext_timeout(data, set)))) {
652 /* Just the extensions could be overwritten */
656 ret = -IPSET_ERR_EXIST;
660 /* Reuse first timed out entry */
661 if (SET_WITH_TIMEOUT(set) &&
662 ip_set_timeout_expired(ext_timeout(data, set)) &&
663 j != AHASH_MAX(h) + 1)
667 if (j != AHASH_MAX(h) + 1) {
668 /* Fill out reused slot */
669 data = ahash_data(n, j, set->dsize);
670 #ifdef IP_SET_HASH_WITH_NETS
671 for (i = 0; i < IPSET_NET_COUNT; i++) {
672 mtype_del_cidr(h, CIDR(data->cidr, i),
673 NLEN(set->family), i);
674 mtype_add_cidr(h, CIDR(d->cidr, i),
675 NLEN(set->family), i);
678 ip_set_ext_destroy(set, data);
680 /* Use/create a new slot */
681 TUNE_AHASH_MAX(h, multi);
682 ret = hbucket_elem_add(n, AHASH_MAX(h), set->dsize);
685 mtype_data_next(&h->next, d);
688 data = ahash_data(n, n->pos++, set->dsize);
689 #ifdef IP_SET_HASH_WITH_NETS
690 for (i = 0; i < IPSET_NET_COUNT; i++)
691 mtype_add_cidr(h, CIDR(d->cidr, i), NLEN(set->family),
696 memcpy(data, d, sizeof(struct mtype_elem));
697 #ifdef IP_SET_HASH_WITH_NETS
698 mtype_data_set_flags(data, flags);
700 if (SET_WITH_TIMEOUT(set))
701 ip_set_timeout_set(ext_timeout(data, set), ext->timeout);
702 if (SET_WITH_COUNTER(set))
703 ip_set_init_counter(ext_counter(data, set), ext);
704 if (SET_WITH_COMMENT(set))
705 ip_set_init_comment(ext_comment(data, set), ext);
708 rcu_read_unlock_bh();
712 /* Delete an element from the hash: swap it with the last element
713 * and free up space if possible.
716 mtype_del(struct ip_set *set, void *value, const struct ip_set_ext *ext,
717 struct ip_set_ext *mext, u32 flags)
719 struct htype *h = set->data;
721 const struct mtype_elem *d = value;
722 struct mtype_elem *data;
724 int i, ret = -IPSET_ERR_EXIST;
725 #ifdef IP_SET_HASH_WITH_NETS
731 t = rcu_dereference_bh(h->table);
732 key = HKEY(value, h->initval, t->htable_bits);
734 for (i = 0; i < n->pos; i++) {
735 data = ahash_data(n, i, set->dsize);
736 if (!mtype_data_equal(data, d, &multi))
738 if (SET_WITH_TIMEOUT(set) &&
739 ip_set_timeout_expired(ext_timeout(data, set)))
743 memcpy(data, ahash_data(n, n->pos - 1, set->dsize),
748 #ifdef IP_SET_HASH_WITH_NETS
749 for (j = 0; j < IPSET_NET_COUNT; j++)
750 mtype_del_cidr(h, CIDR(d->cidr, j), NLEN(set->family),
753 ip_set_ext_destroy(set, data);
754 if (n->pos + AHASH_INIT_SIZE < n->size) {
755 void *tmp = kzalloc((n->size - AHASH_INIT_SIZE)
762 n->size -= AHASH_INIT_SIZE;
763 memcpy(tmp, n->value, n->size * set->dsize);
772 rcu_read_unlock_bh();
777 mtype_data_match(struct mtype_elem *data, const struct ip_set_ext *ext,
778 struct ip_set_ext *mext, struct ip_set *set, u32 flags)
780 if (SET_WITH_COUNTER(set))
781 ip_set_update_counter(ext_counter(data, set),
783 return mtype_do_data_match(data);
786 #ifdef IP_SET_HASH_WITH_NETS
787 /* Special test function which takes into account the different network
788 * sizes added to the set */
790 mtype_test_cidrs(struct ip_set *set, struct mtype_elem *d,
791 const struct ip_set_ext *ext,
792 struct ip_set_ext *mext, u32 flags)
794 struct htype *h = set->data;
795 struct htable *t = rcu_dereference_bh(h->table);
797 struct mtype_elem *data;
798 #if IPSET_NET_COUNT == 2
799 struct mtype_elem orig = *d;
805 u8 nets_length = NLEN(set->family);
807 pr_debug("test by nets\n");
808 for (; j < nets_length && h->nets[j].nets[0] && !multi; j++) {
809 #if IPSET_NET_COUNT == 2
810 mtype_data_reset_elem(d, &orig);
811 mtype_data_netmask(d, h->nets[j].cidr[0], false);
812 for (k = 0; k < nets_length && h->nets[k].nets[1] && !multi;
814 mtype_data_netmask(d, h->nets[k].cidr[1], true);
816 mtype_data_netmask(d, h->nets[j].cidr[0]);
818 key = HKEY(d, h->initval, t->htable_bits);
820 for (i = 0; i < n->pos; i++) {
821 data = ahash_data(n, i, set->dsize);
822 if (!mtype_data_equal(data, d, &multi))
824 if (SET_WITH_TIMEOUT(set)) {
825 if (!ip_set_timeout_expired(
826 ext_timeout(data, set)))
827 return mtype_data_match(data, ext,
830 #ifdef IP_SET_HASH_WITH_MULTI
834 return mtype_data_match(data, ext,
837 #if IPSET_NET_COUNT == 2
845 /* Test whether the element is added to the set */
847 mtype_test(struct ip_set *set, void *value, const struct ip_set_ext *ext,
848 struct ip_set_ext *mext, u32 flags)
850 struct htype *h = set->data;
852 struct mtype_elem *d = value;
854 struct mtype_elem *data;
859 t = rcu_dereference_bh(h->table);
860 #ifdef IP_SET_HASH_WITH_NETS
861 /* If we test an IP address and not a network address,
862 * try all possible network sizes */
863 for (i = 0; i < IPSET_NET_COUNT; i++)
864 if (CIDR(d->cidr, i) != SET_HOST_MASK(set->family))
866 if (i == IPSET_NET_COUNT) {
867 ret = mtype_test_cidrs(set, d, ext, mext, flags);
872 key = HKEY(d, h->initval, t->htable_bits);
874 for (i = 0; i < n->pos; i++) {
875 data = ahash_data(n, i, set->dsize);
876 if (mtype_data_equal(data, d, &multi) &&
877 !(SET_WITH_TIMEOUT(set) &&
878 ip_set_timeout_expired(ext_timeout(data, set)))) {
879 ret = mtype_data_match(data, ext, mext, set, flags);
884 rcu_read_unlock_bh();
888 /* Reply a HEADER request: fill out the header part of the set */
890 mtype_head(struct ip_set *set, struct sk_buff *skb)
892 const struct htype *h = set->data;
893 const struct htable *t;
894 struct nlattr *nested;
897 t = rcu_dereference_bh_nfnl(h->table);
898 memsize = mtype_ahash_memsize(h, t, NLEN(set->family), set->dsize);
900 nested = ipset_nest_start(skb, IPSET_ATTR_DATA);
902 goto nla_put_failure;
903 if (nla_put_net32(skb, IPSET_ATTR_HASHSIZE,
904 htonl(jhash_size(t->htable_bits))) ||
905 nla_put_net32(skb, IPSET_ATTR_MAXELEM, htonl(h->maxelem)))
906 goto nla_put_failure;
907 #ifdef IP_SET_HASH_WITH_NETMASK
908 if (h->netmask != HOST_MASK &&
909 nla_put_u8(skb, IPSET_ATTR_NETMASK, h->netmask))
910 goto nla_put_failure;
912 if (nla_put_net32(skb, IPSET_ATTR_REFERENCES, htonl(set->ref - 1)) ||
913 nla_put_net32(skb, IPSET_ATTR_MEMSIZE, htonl(memsize)))
914 goto nla_put_failure;
915 if (unlikely(ip_set_put_flags(skb, set)))
916 goto nla_put_failure;
917 ipset_nest_end(skb, nested);
924 /* Reply a LIST/SAVE request: dump the elements of the specified set */
926 mtype_list(const struct ip_set *set,
927 struct sk_buff *skb, struct netlink_callback *cb)
929 const struct htype *h = set->data;
930 const struct htable *t = rcu_dereference_bh_nfnl(h->table);
931 struct nlattr *atd, *nested;
932 const struct hbucket *n;
933 const struct mtype_elem *e;
934 u32 first = cb->args[2];
935 /* We assume that one hash bucket fills into one page */
939 atd = ipset_nest_start(skb, IPSET_ATTR_ADT);
942 pr_debug("list hash set %s\n", set->name);
943 for (; cb->args[2] < jhash_size(t->htable_bits); cb->args[2]++) {
944 incomplete = skb_tail_pointer(skb);
945 n = hbucket(t, cb->args[2]);
946 pr_debug("cb->args[2]: %lu, t %p n %p\n", cb->args[2], t, n);
947 for (i = 0; i < n->pos; i++) {
948 e = ahash_data(n, i, set->dsize);
949 if (SET_WITH_TIMEOUT(set) &&
950 ip_set_timeout_expired(ext_timeout(e, set)))
952 pr_debug("list hash %lu hbucket %p i %u, data %p\n",
953 cb->args[2], n, i, e);
954 nested = ipset_nest_start(skb, IPSET_ATTR_DATA);
956 if (cb->args[2] == first) {
957 nla_nest_cancel(skb, atd);
960 goto nla_put_failure;
962 if (mtype_data_list(skb, e))
963 goto nla_put_failure;
964 if (ip_set_put_extensions(skb, set, e, true))
965 goto nla_put_failure;
966 ipset_nest_end(skb, nested);
969 ipset_nest_end(skb, atd);
970 /* Set listing finished */
976 nlmsg_trim(skb, incomplete);
977 if (unlikely(first == cb->args[2])) {
978 pr_warning("Can't list set %s: one bucket does not fit into "
979 "a message. Please report it!\n", set->name);
983 ipset_nest_end(skb, atd);
988 IPSET_TOKEN(MTYPE, _kadt)(struct ip_set *set, const struct sk_buff *skb,
989 const struct xt_action_param *par,
990 enum ipset_adt adt, struct ip_set_adt_opt *opt);
993 IPSET_TOKEN(MTYPE, _uadt)(struct ip_set *set, struct nlattr *tb[],
994 enum ipset_adt adt, u32 *lineno, u32 flags, bool retried);
996 static const struct ip_set_type_variant mtype_variant = {
1000 [IPSET_ADD] = mtype_add,
1001 [IPSET_DEL] = mtype_del,
1002 [IPSET_TEST] = mtype_test,
1004 .destroy = mtype_destroy,
1005 .flush = mtype_flush,
1008 .resize = mtype_resize,
1009 .same_set = mtype_same_set,
1012 #ifdef IP_SET_EMIT_CREATE
1014 IPSET_TOKEN(HTYPE, _create)(struct net *net, struct ip_set *set,
1015 struct nlattr *tb[], u32 flags)
1017 u32 hashsize = IPSET_DEFAULT_HASHSIZE, maxelem = IPSET_DEFAULT_MAXELEM;
1019 #ifdef IP_SET_HASH_WITH_NETMASK
1026 if (!(set->family == NFPROTO_IPV4 || set->family == NFPROTO_IPV6))
1027 return -IPSET_ERR_INVALID_FAMILY;
1028 #ifdef IP_SET_HASH_WITH_NETMASK
1029 netmask = set->family == NFPROTO_IPV4 ? 32 : 128;
1030 pr_debug("Create set %s with family %s\n",
1031 set->name, set->family == NFPROTO_IPV4 ? "inet" : "inet6");
1034 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_HASHSIZE) ||
1035 !ip_set_optattr_netorder(tb, IPSET_ATTR_MAXELEM) ||
1036 !ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
1037 !ip_set_optattr_netorder(tb, IPSET_ATTR_CADT_FLAGS)))
1038 return -IPSET_ERR_PROTOCOL;
1040 if (tb[IPSET_ATTR_HASHSIZE]) {
1041 hashsize = ip_set_get_h32(tb[IPSET_ATTR_HASHSIZE]);
1042 if (hashsize < IPSET_MIMINAL_HASHSIZE)
1043 hashsize = IPSET_MIMINAL_HASHSIZE;
1046 if (tb[IPSET_ATTR_MAXELEM])
1047 maxelem = ip_set_get_h32(tb[IPSET_ATTR_MAXELEM]);
1049 #ifdef IP_SET_HASH_WITH_NETMASK
1050 if (tb[IPSET_ATTR_NETMASK]) {
1051 netmask = nla_get_u8(tb[IPSET_ATTR_NETMASK]);
1053 if ((set->family == NFPROTO_IPV4 && netmask > 32) ||
1054 (set->family == NFPROTO_IPV6 && netmask > 128) ||
1056 return -IPSET_ERR_INVALID_NETMASK;
1061 #ifdef IP_SET_HASH_WITH_NETS
1062 hsize += sizeof(struct net_prefixes) *
1063 (set->family == NFPROTO_IPV4 ? 32 : 128);
1065 h = kzalloc(hsize, GFP_KERNEL);
1069 h->maxelem = maxelem;
1070 #ifdef IP_SET_HASH_WITH_NETMASK
1071 h->netmask = netmask;
1073 get_random_bytes(&h->initval, sizeof(h->initval));
1074 set->timeout = IPSET_NO_TIMEOUT;
1076 hbits = htable_bits(hashsize);
1077 hsize = htable_size(hbits);
1082 t = ip_set_alloc(hsize);
1087 t->htable_bits = hbits;
1088 rcu_assign_pointer(h->table, t);
1091 if (set->family == NFPROTO_IPV4) {
1092 set->variant = &IPSET_TOKEN(HTYPE, 4_variant);
1093 set->dsize = ip_set_elem_len(set, tb,
1094 sizeof(struct IPSET_TOKEN(HTYPE, 4_elem)));
1096 set->variant = &IPSET_TOKEN(HTYPE, 6_variant);
1097 set->dsize = ip_set_elem_len(set, tb,
1098 sizeof(struct IPSET_TOKEN(HTYPE, 6_elem)));
1100 if (tb[IPSET_ATTR_TIMEOUT]) {
1101 set->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
1102 if (set->family == NFPROTO_IPV4)
1103 IPSET_TOKEN(HTYPE, 4_gc_init)(set,
1104 IPSET_TOKEN(HTYPE, 4_gc));
1106 IPSET_TOKEN(HTYPE, 6_gc_init)(set,
1107 IPSET_TOKEN(HTYPE, 6_gc));
1110 pr_debug("create %s hashsize %u (%u) maxelem %u: %p(%p)\n",
1111 set->name, jhash_size(t->htable_bits),
1112 t->htable_bits, h->maxelem, set->data, t);
1116 #endif /* IP_SET_EMIT_CREATE */