]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blob - net/netfilter/ipset/ip_set_hash_gen.h
projects / mirror_ubuntu-hirsute-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
netfilter: ipset: Add element count to all set types header
[mirror_ubuntu-hirsute-kernel.git] / net / netfilter / ipset / ip_set_hash_gen.h
1 /* Copyright (C) 2013 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
2 *
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.
6 */
7
8 #ifndef _IP_SET_HASH_GEN_H
9 #define _IP_SET_HASH_GEN_H
10
11 #include <linux/rcupdate.h>
12 #include <linux/jhash.h>
13 #include <linux/types.h>
14 #include <linux/netfilter/ipset/ip_set_timeout.h>
15
16 #define __ipset_dereference_protected(p, c) rcu_dereference_protected(p, c)
17 #define ipset_dereference_protected(p, set) \
18 __ipset_dereference_protected(p, spin_is_locked(&(set)->lock))
19
20 #define rcu_dereference_bh_nfnl(p) rcu_dereference_bh_check(p, 1)
21
22 /* Hashing which uses arrays to resolve clashing. The hash table is resized
23 * (doubled) when searching becomes too long.
24 * Internally jhash is used with the assumption that the size of the
25 * stored data is a multiple of sizeof(u32).
26 *
27 * Readers and resizing
28 *
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.
33 */
34
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)
39 /* Max muber of elements in the array block when tuned */
40 #define AHASH_MAX_TUNED 64
41
42 /* Max number of elements can be tuned */
43 #ifdef IP_SET_HASH_WITH_MULTI
44 #define AHASH_MAX(h) ((h)->ahash_max)
45
46 static inline u8
47 tune_ahash_max(u8 curr, u32 multi)
48 {
49 u32 n;
50
51 if (multi < curr)
52 return curr;
53
54 n = curr + AHASH_INIT_SIZE;
55 /* Currently, at listing one hash bucket must fit into a message.
56 * Therefore we have a hard limit here.
57 */
58 return n > curr && n <= AHASH_MAX_TUNED ? n : curr;
59 }
60
61 #define TUNE_AHASH_MAX(h, multi) \
62 ((h)->ahash_max = tune_ahash_max((h)->ahash_max, multi))
63 #else
64 #define AHASH_MAX(h) AHASH_MAX_SIZE
65 #define TUNE_AHASH_MAX(h, multi)
66 #endif
67
68 /* A hash bucket */
69 struct hbucket {
70 struct rcu_head rcu; /* for call_rcu_bh */
71 /* Which positions are used in the array */
72 DECLARE_BITMAP(used, AHASH_MAX_TUNED);
73 u8 size; /* size of the array */
74 u8 pos; /* position of the first free entry */
75 unsigned char value[0] /* the array of the values */
76 __aligned(__alignof__(u64));
77 };
78
79 /* The hash table: the table size stored here in order to make resizing easy */
80 struct htable {
81 atomic_t ref; /* References for resizing */
82 atomic_t uref; /* References for dumping */
83 u8 htable_bits; /* size of hash table == 2^htable_bits */
84 struct hbucket __rcu *bucket[0]; /* hashtable buckets */
85 };
86
87 #define hbucket(h, i) ((h)->bucket[i])
88
89 #ifndef IPSET_NET_COUNT
90 #define IPSET_NET_COUNT 1
91 #endif
92
93 /* Book-keeping of the prefixes added to the set */
94 struct net_prefixes {
95 u32 nets[IPSET_NET_COUNT]; /* number of elements for this cidr */
96 u8 cidr[IPSET_NET_COUNT]; /* the cidr value */
97 };
98
99 /* Compute the hash table size */
100 static size_t
101 htable_size(u8 hbits)
102 {
103 size_t hsize;
104
105 /* We must fit both into u32 in jhash and size_t */
106 if (hbits > 31)
107 return 0;
108 hsize = jhash_size(hbits);
109 if ((((size_t)-1) - sizeof(struct htable)) / sizeof(struct hbucket *)
110 < hsize)
111 return 0;
112
113 return hsize * sizeof(struct hbucket *) + sizeof(struct htable);
114 }
115
116 /* Compute htable_bits from the user input parameter hashsize */
117 static u8
118 htable_bits(u32 hashsize)
119 {
120 /* Assume that hashsize == 2^htable_bits */
121 u8 bits = fls(hashsize - 1);
122
123 if (jhash_size(bits) != hashsize)
124 /* Round up to the first 2^n value */
125 bits = fls(hashsize);
126
127 return bits;
128 }
129
130 #ifdef IP_SET_HASH_WITH_NETS
131 #if IPSET_NET_COUNT > 1
132 #define __CIDR(cidr, i) (cidr[i])
133 #else
134 #define __CIDR(cidr, i) (cidr)
135 #endif
136
137 /* cidr + 1 is stored in net_prefixes to support /0 */
138 #define NCIDR_PUT(cidr) ((cidr) + 1)
139 #define NCIDR_GET(cidr) ((cidr) - 1)
140
141 #ifdef IP_SET_HASH_WITH_NETS_PACKED
142 /* When cidr is packed with nomatch, cidr - 1 is stored in the data entry */
143 #define DCIDR_PUT(cidr) ((cidr) - 1)
144 #define DCIDR_GET(cidr, i) (__CIDR(cidr, i) + 1)
145 #else
146 #define DCIDR_PUT(cidr) (cidr)
147 #define DCIDR_GET(cidr, i) __CIDR(cidr, i)
148 #endif
149
150 #define INIT_CIDR(cidr, host_mask) \
151 DCIDR_PUT(((cidr) ? NCIDR_GET(cidr) : host_mask))
152
153 #define SET_HOST_MASK(family) (family == AF_INET ? 32 : 128)
154
155 #ifdef IP_SET_HASH_WITH_NET0
156 /* cidr from 0 to SET_HOST_MASK() value and c = cidr + 1 */
157 #define NLEN(family) (SET_HOST_MASK(family) + 1)
158 #define CIDR_POS(c) ((c) - 1)
159 #else
160 /* cidr from 1 to SET_HOST_MASK() value and c = cidr + 1 */
161 #define NLEN(family) SET_HOST_MASK(family)
162 #define CIDR_POS(c) ((c) - 2)
163 #endif
164
165 #else
166 #define NLEN(family) 0
167 #endif /* IP_SET_HASH_WITH_NETS */
168
169 #endif /* _IP_SET_HASH_GEN_H */
170
171 /* Family dependent templates */
172
173 #undef ahash_data
174 #undef mtype_data_equal
175 #undef mtype_do_data_match
176 #undef mtype_data_set_flags
177 #undef mtype_data_reset_elem
178 #undef mtype_data_reset_flags
179 #undef mtype_data_netmask
180 #undef mtype_data_list
181 #undef mtype_data_next
182 #undef mtype_elem
183
184 #undef mtype_ahash_destroy
185 #undef mtype_ext_cleanup
186 #undef mtype_add_cidr
187 #undef mtype_del_cidr
188 #undef mtype_ahash_memsize
189 #undef mtype_flush
190 #undef mtype_destroy
191 #undef mtype_same_set
192 #undef mtype_kadt
193 #undef mtype_uadt
194 #undef mtype
195
196 #undef mtype_add
197 #undef mtype_del
198 #undef mtype_test_cidrs
199 #undef mtype_test
200 #undef mtype_uref
201 #undef mtype_expire
202 #undef mtype_resize
203 #undef mtype_head
204 #undef mtype_list
205 #undef mtype_gc
206 #undef mtype_gc_init
207 #undef mtype_variant
208 #undef mtype_data_match
209
210 #undef HKEY
211
212 #define mtype_data_equal IPSET_TOKEN(MTYPE, _data_equal)
213 #ifdef IP_SET_HASH_WITH_NETS
214 #define mtype_do_data_match IPSET_TOKEN(MTYPE, _do_data_match)
215 #else
216 #define mtype_do_data_match(d) 1
217 #endif
218 #define mtype_data_set_flags IPSET_TOKEN(MTYPE, _data_set_flags)
219 #define mtype_data_reset_elem IPSET_TOKEN(MTYPE, _data_reset_elem)
220 #define mtype_data_reset_flags IPSET_TOKEN(MTYPE, _data_reset_flags)
221 #define mtype_data_netmask IPSET_TOKEN(MTYPE, _data_netmask)
222 #define mtype_data_list IPSET_TOKEN(MTYPE, _data_list)
223 #define mtype_data_next IPSET_TOKEN(MTYPE, _data_next)
224 #define mtype_elem IPSET_TOKEN(MTYPE, _elem)
225
226 #define mtype_ahash_destroy IPSET_TOKEN(MTYPE, _ahash_destroy)
227 #define mtype_ext_cleanup IPSET_TOKEN(MTYPE, _ext_cleanup)
228 #define mtype_add_cidr IPSET_TOKEN(MTYPE, _add_cidr)
229 #define mtype_del_cidr IPSET_TOKEN(MTYPE, _del_cidr)
230 #define mtype_ahash_memsize IPSET_TOKEN(MTYPE, _ahash_memsize)
231 #define mtype_flush IPSET_TOKEN(MTYPE, _flush)
232 #define mtype_destroy IPSET_TOKEN(MTYPE, _destroy)
233 #define mtype_same_set IPSET_TOKEN(MTYPE, _same_set)
234 #define mtype_kadt IPSET_TOKEN(MTYPE, _kadt)
235 #define mtype_uadt IPSET_TOKEN(MTYPE, _uadt)
236 #define mtype MTYPE
237
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_uref IPSET_TOKEN(MTYPE, _uref)
243 #define mtype_expire IPSET_TOKEN(MTYPE, _expire)
244 #define mtype_resize IPSET_TOKEN(MTYPE, _resize)
245 #define mtype_head IPSET_TOKEN(MTYPE, _head)
246 #define mtype_list IPSET_TOKEN(MTYPE, _list)
247 #define mtype_gc IPSET_TOKEN(MTYPE, _gc)
248 #define mtype_gc_init IPSET_TOKEN(MTYPE, _gc_init)
249 #define mtype_variant IPSET_TOKEN(MTYPE, _variant)
250 #define mtype_data_match IPSET_TOKEN(MTYPE, _data_match)
251
252 #ifndef MTYPE
253 #error "MTYPE is not defined!"
254 #endif
255
256 #ifndef HOST_MASK
257 #error "HOST_MASK is not defined!"
258 #endif
259
260 #ifndef HKEY_DATALEN
261 #define HKEY_DATALEN sizeof(struct mtype_elem)
262 #endif
263
264 #define HKEY(data, initval, htable_bits) \
265 (jhash2((u32 *)(data), HKEY_DATALEN / sizeof(u32), initval) \
266 & jhash_mask(htable_bits))
267
268 #ifndef htype
269 #ifndef HTYPE
270 #error "HTYPE is not defined!"
271 #endif /* HTYPE */
272 #define htype HTYPE
273
274 /* The generic hash structure */
275 struct htype {
276 struct htable __rcu *table; /* the hash table */
277 u32 maxelem; /* max elements in the hash */
278 u32 initval; /* random jhash init value */
279 #ifdef IP_SET_HASH_WITH_MARKMASK
280 u32 markmask; /* markmask value for mark mask to store */
281 #endif
282 struct timer_list gc; /* garbage collection when timeout enabled */
283 struct mtype_elem next; /* temporary storage for uadd */
284 #ifdef IP_SET_HASH_WITH_MULTI
285 u8 ahash_max; /* max elements in an array block */
286 #endif
287 #ifdef IP_SET_HASH_WITH_NETMASK
288 u8 netmask; /* netmask value for subnets to store */
289 #endif
290 #ifdef IP_SET_HASH_WITH_NETS
291 struct net_prefixes nets[0]; /* book-keeping of prefixes */
292 #endif
293 };
294 #endif /* htype */
295
296 #ifdef IP_SET_HASH_WITH_NETS
297 /* Network cidr size book keeping when the hash stores different
298 * sized networks. cidr == real cidr + 1 to support /0.
299 */
300 static void
301 mtype_add_cidr(struct htype *h, u8 cidr, u8 nets_length, u8 n)
302 {
303 int i, j;
304
305 /* Add in increasing prefix order, so larger cidr first */
306 for (i = 0, j = -1; i < nets_length && h->nets[i].cidr[n]; i++) {
307 if (j != -1) {
308 continue;
309 } else if (h->nets[i].cidr[n] < cidr) {
310 j = i;
311 } else if (h->nets[i].cidr[n] == cidr) {
312 h->nets[CIDR_POS(cidr)].nets[n]++;
313 return;
314 }
315 }
316 if (j != -1) {
317 for (; i > j; i--)
318 h->nets[i].cidr[n] = h->nets[i - 1].cidr[n];
319 }
320 h->nets[i].cidr[n] = cidr;
321 h->nets[CIDR_POS(cidr)].nets[n] = 1;
322 }
323
324 static void
325 mtype_del_cidr(struct htype *h, u8 cidr, u8 nets_length, u8 n)
326 {
327 u8 i, j, net_end = nets_length - 1;
328
329 for (i = 0; i < nets_length; i++) {
330 if (h->nets[i].cidr[n] != cidr)
331 continue;
332 h->nets[CIDR_POS(cidr)].nets[n]--;
333 if (h->nets[CIDR_POS(cidr)].nets[n] > 0)
334 return;
335 for (j = i; j < net_end && h->nets[j].cidr[n]; j++)
336 h->nets[j].cidr[n] = h->nets[j + 1].cidr[n];
337 h->nets[j].cidr[n] = 0;
338 return;
339 }
340 }
341 #endif
342
343 /* Calculate the actual memory size of the set data */
344 static size_t
345 mtype_ahash_memsize(const struct htype *h, const struct htable *t,
346 u8 nets_length, size_t dsize)
347 {
348 u32 i;
349 struct hbucket *n;
350 size_t memsize = sizeof(*h) + sizeof(*t);
351
352 #ifdef IP_SET_HASH_WITH_NETS
353 memsize += sizeof(struct net_prefixes) * nets_length;
354 #endif
355 for (i = 0; i < jhash_size(t->htable_bits); i++) {
356 n = rcu_dereference_bh(hbucket(t, i));
357 if (!n)
358 continue;
359 memsize += sizeof(struct hbucket) + n->size * dsize;
360 }
361
362 return memsize;
363 }
364
365 /* Get the ith element from the array block n */
366 #define ahash_data(n, i, dsize) \
367 ((struct mtype_elem *)((n)->value + ((i) * (dsize))))
368
369 static void
370 mtype_ext_cleanup(struct ip_set *set, struct hbucket *n)
371 {
372 int i;
373
374 for (i = 0; i < n->pos; i++)
375 if (test_bit(i, n->used))
376 ip_set_ext_destroy(set, ahash_data(n, i, set->dsize));
377 }
378
379 /* Flush a hash type of set: destroy all elements */
380 static void
381 mtype_flush(struct ip_set *set)
382 {
383 struct htype *h = set->data;
384 struct htable *t;
385 struct hbucket *n;
386 u32 i;
387
388 t = ipset_dereference_protected(h->table, set);
389 for (i = 0; i < jhash_size(t->htable_bits); i++) {
390 n = __ipset_dereference_protected(hbucket(t, i), 1);
391 if (!n)
392 continue;
393 if (set->extensions & IPSET_EXT_DESTROY)
394 mtype_ext_cleanup(set, n);
395 /* FIXME: use slab cache */
396 rcu_assign_pointer(hbucket(t, i), NULL);
397 kfree_rcu(n, rcu);
398 }
399 #ifdef IP_SET_HASH_WITH_NETS
400 memset(h->nets, 0, sizeof(struct net_prefixes) * NLEN(set->family));
401 #endif
402 set->elements = 0;
403 }
404
405 /* Destroy the hashtable part of the set */
406 static void
407 mtype_ahash_destroy(struct ip_set *set, struct htable *t, bool ext_destroy)
408 {
409 struct hbucket *n;
410 u32 i;
411
412 for (i = 0; i < jhash_size(t->htable_bits); i++) {
413 n = __ipset_dereference_protected(hbucket(t, i), 1);
414 if (!n)
415 continue;
416 if (set->extensions & IPSET_EXT_DESTROY && ext_destroy)
417 mtype_ext_cleanup(set, n);
418 /* FIXME: use slab cache */
419 kfree(n);
420 }
421
422 ip_set_free(t);
423 }
424
425 /* Destroy a hash type of set */
426 static void
427 mtype_destroy(struct ip_set *set)
428 {
429 struct htype *h = set->data;
430
431 if (SET_WITH_TIMEOUT(set))
432 del_timer_sync(&h->gc);
433
434 mtype_ahash_destroy(set,
435 __ipset_dereference_protected(h->table, 1), true);
436 kfree(h);
437
438 set->data = NULL;
439 }
440
441 static void
442 mtype_gc_init(struct ip_set *set, void (*gc)(unsigned long ul_set))
443 {
444 struct htype *h = set->data;
445
446 init_timer(&h->gc);
447 h->gc.data = (unsigned long)set;
448 h->gc.function = gc;
449 h->gc.expires = jiffies + IPSET_GC_PERIOD(set->timeout) * HZ;
450 add_timer(&h->gc);
451 pr_debug("gc initialized, run in every %u\n",
452 IPSET_GC_PERIOD(set->timeout));
453 }
454
455 static bool
456 mtype_same_set(const struct ip_set *a, const struct ip_set *b)
457 {
458 const struct htype *x = a->data;
459 const struct htype *y = b->data;
460
461 /* Resizing changes htable_bits, so we ignore it */
462 return x->maxelem == y->maxelem &&
463 a->timeout == b->timeout &&
464 #ifdef IP_SET_HASH_WITH_NETMASK
465 x->netmask == y->netmask &&
466 #endif
467 #ifdef IP_SET_HASH_WITH_MARKMASK
468 x->markmask == y->markmask &&
469 #endif
470 a->extensions == b->extensions;
471 }
472
473 /* Delete expired elements from the hashtable */
474 static void
475 mtype_expire(struct ip_set *set, struct htype *h, u8 nets_length, size_t dsize)
476 {
477 struct htable *t;
478 struct hbucket *n, *tmp;
479 struct mtype_elem *data;
480 u32 i, j, d;
481 #ifdef IP_SET_HASH_WITH_NETS
482 u8 k;
483 #endif
484
485 t = ipset_dereference_protected(h->table, set);
486 for (i = 0; i < jhash_size(t->htable_bits); i++) {
487 n = __ipset_dereference_protected(hbucket(t, i), 1);
488 if (!n)
489 continue;
490 for (j = 0, d = 0; j < n->pos; j++) {
491 if (!test_bit(j, n->used)) {
492 d++;
493 continue;
494 }
495 data = ahash_data(n, j, dsize);
496 if (ip_set_timeout_expired(ext_timeout(data, set))) {
497 pr_debug("expired %u/%u\n", i, j);
498 clear_bit(j, n->used);
499 smp_mb__after_atomic();
500 #ifdef IP_SET_HASH_WITH_NETS
501 for (k = 0; k < IPSET_NET_COUNT; k++)
502 mtype_del_cidr(h,
503 NCIDR_PUT(DCIDR_GET(data->cidr,
504 k)),
505 nets_length, k);
506 #endif
507 ip_set_ext_destroy(set, data);
508 set->elements--;
509 d++;
510 }
511 }
512 if (d >= AHASH_INIT_SIZE) {
513 if (d >= n->size) {
514 rcu_assign_pointer(hbucket(t, i), NULL);
515 kfree_rcu(n, rcu);
516 continue;
517 }
518 tmp = kzalloc(sizeof(*tmp) +
519 (n->size - AHASH_INIT_SIZE) * dsize,
520 GFP_ATOMIC);
521 if (!tmp)
522 /* Still try to delete expired elements */
523 continue;
524 tmp->size = n->size - AHASH_INIT_SIZE;
525 for (j = 0, d = 0; j < n->pos; j++) {
526 if (!test_bit(j, n->used))
527 continue;
528 data = ahash_data(n, j, dsize);
529 memcpy(tmp->value + d * dsize, data, dsize);
530 set_bit(d, tmp->used);
531 d++;
532 }
533 tmp->pos = d;
534 rcu_assign_pointer(hbucket(t, i), tmp);
535 kfree_rcu(n, rcu);
536 }
537 }
538 }
539
540 static void
541 mtype_gc(unsigned long ul_set)
542 {
543 struct ip_set *set = (struct ip_set *)ul_set;
544 struct htype *h = set->data;
545
546 pr_debug("called\n");
547 spin_lock_bh(&set->lock);
548 mtype_expire(set, h, NLEN(set->family), set->dsize);
549 spin_unlock_bh(&set->lock);
550
551 h->gc.expires = jiffies + IPSET_GC_PERIOD(set->timeout) * HZ;
552 add_timer(&h->gc);
553 }
554
555 /* Resize a hash: create a new hash table with doubling the hashsize
556 * and inserting the elements to it. Repeat until we succeed or
557 * fail due to memory pressures.
558 */
559 static int
560 mtype_resize(struct ip_set *set, bool retried)
561 {
562 struct htype *h = set->data;
563 struct htable *t, *orig;
564 u8 htable_bits;
565 size_t dsize = set->dsize;
566 #ifdef IP_SET_HASH_WITH_NETS
567 u8 flags;
568 struct mtype_elem *tmp;
569 #endif
570 struct mtype_elem *data;
571 struct mtype_elem *d;
572 struct hbucket *n, *m;
573 u32 i, j, key;
574 int ret;
575
576 #ifdef IP_SET_HASH_WITH_NETS
577 tmp = kmalloc(dsize, GFP_KERNEL);
578 if (!tmp)
579 return -ENOMEM;
580 #endif
581 rcu_read_lock_bh();
582 orig = rcu_dereference_bh_nfnl(h->table);
583 htable_bits = orig->htable_bits;
584 rcu_read_unlock_bh();
585
586 retry:
587 ret = 0;
588 htable_bits++;
589 if (!htable_bits) {
590 /* In case we have plenty of memory :-) */
591 pr_warn("Cannot increase the hashsize of set %s further\n",
592 set->name);
593 ret = -IPSET_ERR_HASH_FULL;
594 goto out;
595 }
596 t = ip_set_alloc(htable_size(htable_bits));
597 if (!t) {
598 ret = -ENOMEM;
599 goto out;
600 }
601 t->htable_bits = htable_bits;
602
603 spin_lock_bh(&set->lock);
604 orig = __ipset_dereference_protected(h->table, 1);
605 /* There can't be another parallel resizing, but dumping is possible */
606 atomic_set(&orig->ref, 1);
607 atomic_inc(&orig->uref);
608 pr_debug("attempt to resize set %s from %u to %u, t %p\n",
609 set->name, orig->htable_bits, htable_bits, orig);
610 for (i = 0; i < jhash_size(orig->htable_bits); i++) {
611 n = __ipset_dereference_protected(hbucket(orig, i), 1);
612 if (!n)
613 continue;
614 for (j = 0; j < n->pos; j++) {
615 if (!test_bit(j, n->used))
616 continue;
617 data = ahash_data(n, j, dsize);
618 #ifdef IP_SET_HASH_WITH_NETS
619 /* We have readers running parallel with us,
620 * so the live data cannot be modified.
621 */
622 flags = 0;
623 memcpy(tmp, data, dsize);
624 data = tmp;
625 mtype_data_reset_flags(data, &flags);
626 #endif
627 key = HKEY(data, h->initval, htable_bits);
628 m = __ipset_dereference_protected(hbucket(t, key), 1);
629 if (!m) {
630 m = kzalloc(sizeof(*m) +
631 AHASH_INIT_SIZE * dsize,
632 GFP_ATOMIC);
633 if (!m) {
634 ret = -ENOMEM;
635 goto cleanup;
636 }
637 m->size = AHASH_INIT_SIZE;
638 RCU_INIT_POINTER(hbucket(t, key), m);
639 } else if (m->pos >= m->size) {
640 struct hbucket *ht;
641
642 if (m->size >= AHASH_MAX(h)) {
643 ret = -EAGAIN;
644 } else {
645 ht = kzalloc(sizeof(*ht) +
646 (m->size + AHASH_INIT_SIZE)
647 * dsize,
648 GFP_ATOMIC);
649 if (!ht)
650 ret = -ENOMEM;
651 }
652 if (ret < 0)
653 goto cleanup;
654 memcpy(ht, m, sizeof(struct hbucket) +
655 m->size * dsize);
656 ht->size = m->size + AHASH_INIT_SIZE;
657 kfree(m);
658 m = ht;
659 RCU_INIT_POINTER(hbucket(t, key), ht);
660 }
661 d = ahash_data(m, m->pos, dsize);
662 memcpy(d, data, dsize);
663 set_bit(m->pos++, m->used);
664 #ifdef IP_SET_HASH_WITH_NETS
665 mtype_data_reset_flags(d, &flags);
666 #endif
667 }
668 }
669 rcu_assign_pointer(h->table, t);
670
671 spin_unlock_bh(&set->lock);
672
673 /* Give time to other readers of the set */
674 synchronize_rcu_bh();
675
676 pr_debug("set %s resized from %u (%p) to %u (%p)\n", set->name,
677 orig->htable_bits, orig, t->htable_bits, t);
678 /* If there's nobody else dumping the table, destroy it */
679 if (atomic_dec_and_test(&orig->uref)) {
680 pr_debug("Table destroy by resize %p\n", orig);
681 mtype_ahash_destroy(set, orig, false);
682 }
683
684 out:
685 #ifdef IP_SET_HASH_WITH_NETS
686 kfree(tmp);
687 #endif
688 return ret;
689
690 cleanup:
691 atomic_set(&orig->ref, 0);
692 atomic_dec(&orig->uref);
693 spin_unlock_bh(&set->lock);
694 mtype_ahash_destroy(set, t, false);
695 if (ret == -EAGAIN)
696 goto retry;
697 goto out;
698 }
699
700 /* Add an element to a hash and update the internal counters when succeeded,
701 * otherwise report the proper error code.
702 */
703 static int
704 mtype_add(struct ip_set *set, void *value, const struct ip_set_ext *ext,
705 struct ip_set_ext *mext, u32 flags)
706 {
707 struct htype *h = set->data;
708 struct htable *t;
709 const struct mtype_elem *d = value;
710 struct mtype_elem *data;
711 struct hbucket *n, *old = ERR_PTR(-ENOENT);
712 int i, j = -1;
713 bool flag_exist = flags & IPSET_FLAG_EXIST;
714 bool deleted = false, forceadd = false, reuse = false;
715 u32 key, multi = 0;
716
717 if (set->elements >= h->maxelem) {
718 if (SET_WITH_TIMEOUT(set))
719 /* FIXME: when set is full, we slow down here */
720 mtype_expire(set, h, NLEN(set->family), set->dsize);
721 if (set->elements >= h->maxelem && SET_WITH_FORCEADD(set))
722 forceadd = true;
723 }
724
725 t = ipset_dereference_protected(h->table, set);
726 key = HKEY(value, h->initval, t->htable_bits);
727 n = __ipset_dereference_protected(hbucket(t, key), 1);
728 if (!n) {
729 if (forceadd) {
730 if (net_ratelimit())
731 pr_warn("Set %s is full, maxelem %u reached\n",
732 set->name, h->maxelem);
733 return -IPSET_ERR_HASH_FULL;
734 } else if (set->elements >= h->maxelem) {
735 goto set_full;
736 }
737 old = NULL;
738 n = kzalloc(sizeof(*n) + AHASH_INIT_SIZE * set->dsize,
739 GFP_ATOMIC);
740 if (!n)
741 return -ENOMEM;
742 n->size = AHASH_INIT_SIZE;
743 goto copy_elem;
744 }
745 for (i = 0; i < n->pos; i++) {
746 if (!test_bit(i, n->used)) {
747 /* Reuse first deleted entry */
748 if (j == -1) {
749 deleted = reuse = true;
750 j = i;
751 }
752 continue;
753 }
754 data = ahash_data(n, i, set->dsize);
755 if (mtype_data_equal(data, d, &multi)) {
756 if (flag_exist ||
757 (SET_WITH_TIMEOUT(set) &&
758 ip_set_timeout_expired(ext_timeout(data, set)))) {
759 /* Just the extensions could be overwritten */
760 j = i;
761 goto overwrite_extensions;
762 }
763 return -IPSET_ERR_EXIST;
764 }
765 /* Reuse first timed out entry */
766 if (SET_WITH_TIMEOUT(set) &&
767 ip_set_timeout_expired(ext_timeout(data, set)) &&
768 j == -1) {
769 j = i;
770 reuse = true;
771 }
772 }
773 if (reuse || forceadd) {
774 data = ahash_data(n, j, set->dsize);
775 if (!deleted) {
776 #ifdef IP_SET_HASH_WITH_NETS
777 for (i = 0; i < IPSET_NET_COUNT; i++)
778 mtype_del_cidr(h,
779 NCIDR_PUT(DCIDR_GET(data->cidr, i)),
780 NLEN(set->family), i);
781 #endif
782 ip_set_ext_destroy(set, data);
783 set->elements--;
784 }
785 goto copy_data;
786 }
787 if (set->elements >= h->maxelem)
788 goto set_full;
789 /* Create a new slot */
790 if (n->pos >= n->size) {
791 TUNE_AHASH_MAX(h, multi);
792 if (n->size >= AHASH_MAX(h)) {
793 /* Trigger rehashing */
794 mtype_data_next(&h->next, d);
795 return -EAGAIN;
796 }
797 old = n;
798 n = kzalloc(sizeof(*n) +
799 (old->size + AHASH_INIT_SIZE) * set->dsize,
800 GFP_ATOMIC);
801 if (!n)
802 return -ENOMEM;
803 memcpy(n, old, sizeof(struct hbucket) +
804 old->size * set->dsize);
805 n->size = old->size + AHASH_INIT_SIZE;
806 }
807
808 copy_elem:
809 j = n->pos++;
810 data = ahash_data(n, j, set->dsize);
811 copy_data:
812 set->elements++;
813 #ifdef IP_SET_HASH_WITH_NETS
814 for (i = 0; i < IPSET_NET_COUNT; i++)
815 mtype_add_cidr(h, NCIDR_PUT(DCIDR_GET(d->cidr, i)),
816 NLEN(set->family), i);
817 #endif
818 memcpy(data, d, sizeof(struct mtype_elem));
819 overwrite_extensions:
820 #ifdef IP_SET_HASH_WITH_NETS
821 mtype_data_set_flags(data, flags);
822 #endif
823 if (SET_WITH_COUNTER(set))
824 ip_set_init_counter(ext_counter(data, set), ext);
825 if (SET_WITH_COMMENT(set))
826 ip_set_init_comment(ext_comment(data, set), ext);
827 if (SET_WITH_SKBINFO(set))
828 ip_set_init_skbinfo(ext_skbinfo(data, set), ext);
829 /* Must come last for the case when timed out entry is reused */
830 if (SET_WITH_TIMEOUT(set))
831 ip_set_timeout_set(ext_timeout(data, set), ext->timeout);
832 smp_mb__before_atomic();
833 set_bit(j, n->used);
834 if (old != ERR_PTR(-ENOENT)) {
835 rcu_assign_pointer(hbucket(t, key), n);
836 if (old)
837 kfree_rcu(old, rcu);
838 }
839
840 return 0;
841 set_full:
842 if (net_ratelimit())
843 pr_warn("Set %s is full, maxelem %u reached\n",
844 set->name, h->maxelem);
845 return -IPSET_ERR_HASH_FULL;
846 }
847
848 /* Delete an element from the hash and free up space if possible.
849 */
850 static int
851 mtype_del(struct ip_set *set, void *value, const struct ip_set_ext *ext,
852 struct ip_set_ext *mext, u32 flags)
853 {
854 struct htype *h = set->data;
855 struct htable *t;
856 const struct mtype_elem *d = value;
857 struct mtype_elem *data;
858 struct hbucket *n;
859 int i, j, k, ret = -IPSET_ERR_EXIST;
860 u32 key, multi = 0;
861 size_t dsize = set->dsize;
862
863 t = ipset_dereference_protected(h->table, set);
864 key = HKEY(value, h->initval, t->htable_bits);
865 n = __ipset_dereference_protected(hbucket(t, key), 1);
866 if (!n)
867 goto out;
868 for (i = 0, k = 0; i < n->pos; i++) {
869 if (!test_bit(i, n->used)) {
870 k++;
871 continue;
872 }
873 data = ahash_data(n, i, dsize);
874 if (!mtype_data_equal(data, d, &multi))
875 continue;
876 if (SET_WITH_TIMEOUT(set) &&
877 ip_set_timeout_expired(ext_timeout(data, set)))
878 goto out;
879
880 ret = 0;
881 clear_bit(i, n->used);
882 smp_mb__after_atomic();
883 if (i + 1 == n->pos)
884 n->pos--;
885 set->elements--;
886 #ifdef IP_SET_HASH_WITH_NETS
887 for (j = 0; j < IPSET_NET_COUNT; j++)
888 mtype_del_cidr(h, NCIDR_PUT(DCIDR_GET(d->cidr, j)),
889 NLEN(set->family), j);
890 #endif
891 ip_set_ext_destroy(set, data);
892
893 for (; i < n->pos; i++) {
894 if (!test_bit(i, n->used))
895 k++;
896 }
897 if (n->pos == 0 && k == 0) {
898 rcu_assign_pointer(hbucket(t, key), NULL);
899 kfree_rcu(n, rcu);
900 } else if (k >= AHASH_INIT_SIZE) {
901 struct hbucket *tmp = kzalloc(sizeof(*tmp) +
902 (n->size - AHASH_INIT_SIZE) * dsize,
903 GFP_ATOMIC);
904 if (!tmp)
905 goto out;
906 tmp->size = n->size - AHASH_INIT_SIZE;
907 for (j = 0, k = 0; j < n->pos; j++) {
908 if (!test_bit(j, n->used))
909 continue;
910 data = ahash_data(n, j, dsize);
911 memcpy(tmp->value + k * dsize, data, dsize);
912 set_bit(j, tmp->used);
913 k++;
914 }
915 tmp->pos = k;
916 rcu_assign_pointer(hbucket(t, key), tmp);
917 kfree_rcu(n, rcu);
918 }
919 goto out;
920 }
921
922 out:
923 return ret;
924 }
925
926 static inline int
927 mtype_data_match(struct mtype_elem *data, const struct ip_set_ext *ext,
928 struct ip_set_ext *mext, struct ip_set *set, u32 flags)
929 {
930 if (SET_WITH_COUNTER(set))
931 ip_set_update_counter(ext_counter(data, set),
932 ext, mext, flags);
933 if (SET_WITH_SKBINFO(set))
934 ip_set_get_skbinfo(ext_skbinfo(data, set),
935 ext, mext, flags);
936 return mtype_do_data_match(data);
937 }
938
939 #ifdef IP_SET_HASH_WITH_NETS
940 /* Special test function which takes into account the different network
941 * sizes added to the set
942 */
943 static int
944 mtype_test_cidrs(struct ip_set *set, struct mtype_elem *d,
945 const struct ip_set_ext *ext,
946 struct ip_set_ext *mext, u32 flags)
947 {
948 struct htype *h = set->data;
949 struct htable *t = rcu_dereference_bh(h->table);
950 struct hbucket *n;
951 struct mtype_elem *data;
952 #if IPSET_NET_COUNT == 2
953 struct mtype_elem orig = *d;
954 int i, j = 0, k;
955 #else
956 int i, j = 0;
957 #endif
958 u32 key, multi = 0;
959 u8 nets_length = NLEN(set->family);
960
961 pr_debug("test by nets\n");
962 for (; j < nets_length && h->nets[j].cidr[0] && !multi; j++) {
963 #if IPSET_NET_COUNT == 2
964 mtype_data_reset_elem(d, &orig);
965 mtype_data_netmask(d, NCIDR_GET(h->nets[j].cidr[0]), false);
966 for (k = 0; k < nets_length && h->nets[k].cidr[1] && !multi;
967 k++) {
968 mtype_data_netmask(d, NCIDR_GET(h->nets[k].cidr[1]),
969 true);
970 #else
971 mtype_data_netmask(d, NCIDR_GET(h->nets[j].cidr[0]));
972 #endif
973 key = HKEY(d, h->initval, t->htable_bits);
974 n = rcu_dereference_bh(hbucket(t, key));
975 if (!n)
976 continue;
977 for (i = 0; i < n->pos; i++) {
978 if (!test_bit(i, n->used))
979 continue;
980 data = ahash_data(n, i, set->dsize);
981 if (!mtype_data_equal(data, d, &multi))
982 continue;
983 if (SET_WITH_TIMEOUT(set)) {
984 if (!ip_set_timeout_expired(
985 ext_timeout(data, set)))
986 return mtype_data_match(data, ext,
987 mext, set,
988 flags);
989 #ifdef IP_SET_HASH_WITH_MULTI
990 multi = 0;
991 #endif
992 } else
993 return mtype_data_match(data, ext,
994 mext, set, flags);
995 }
996 #if IPSET_NET_COUNT == 2
997 }
998 #endif
999 }
1000 return 0;
1001 }
1002 #endif
1003
1004 /* Test whether the element is added to the set */
1005 static int
1006 mtype_test(struct ip_set *set, void *value, const struct ip_set_ext *ext,
1007 struct ip_set_ext *mext, u32 flags)
1008 {
1009 struct htype *h = set->data;
1010 struct htable *t;
1011 struct mtype_elem *d = value;
1012 struct hbucket *n;
1013 struct mtype_elem *data;
1014 int i, ret = 0;
1015 u32 key, multi = 0;
1016
1017 t = rcu_dereference_bh(h->table);
1018 #ifdef IP_SET_HASH_WITH_NETS
1019 /* If we test an IP address and not a network address,
1020 * try all possible network sizes
1021 */
1022 for (i = 0; i < IPSET_NET_COUNT; i++)
1023 if (DCIDR_GET(d->cidr, i) != SET_HOST_MASK(set->family))
1024 break;
1025 if (i == IPSET_NET_COUNT) {
1026 ret = mtype_test_cidrs(set, d, ext, mext, flags);
1027 goto out;
1028 }
1029 #endif
1030
1031 key = HKEY(d, h->initval, t->htable_bits);
1032 n = rcu_dereference_bh(hbucket(t, key));
1033 if (!n) {
1034 ret = 0;
1035 goto out;
1036 }
1037 for (i = 0; i < n->pos; i++) {
1038 if (!test_bit(i, n->used))
1039 continue;
1040 data = ahash_data(n, i, set->dsize);
1041 if (mtype_data_equal(data, d, &multi) &&
1042 !(SET_WITH_TIMEOUT(set) &&
1043 ip_set_timeout_expired(ext_timeout(data, set)))) {
1044 ret = mtype_data_match(data, ext, mext, set, flags);
1045 goto out;
1046 }
1047 }
1048 out:
1049 return ret;
1050 }
1051
1052 /* Reply a HEADER request: fill out the header part of the set */
1053 static int
1054 mtype_head(struct ip_set *set, struct sk_buff *skb)
1055 {
1056 const struct htype *h = set->data;
1057 const struct htable *t;
1058 struct nlattr *nested;
1059 size_t memsize;
1060 u8 htable_bits;
1061
1062 rcu_read_lock_bh();
1063 t = rcu_dereference_bh_nfnl(h->table);
1064 memsize = mtype_ahash_memsize(h, t, NLEN(set->family), set->dsize);
1065 htable_bits = t->htable_bits;
1066 rcu_read_unlock_bh();
1067
1068 nested = ipset_nest_start(skb, IPSET_ATTR_DATA);
1069 if (!nested)
1070 goto nla_put_failure;
1071 if (nla_put_net32(skb, IPSET_ATTR_HASHSIZE,
1072 htonl(jhash_size(htable_bits))) ||
1073 nla_put_net32(skb, IPSET_ATTR_MAXELEM, htonl(h->maxelem)))
1074 goto nla_put_failure;
1075 #ifdef IP_SET_HASH_WITH_NETMASK
1076 if (h->netmask != HOST_MASK &&
1077 nla_put_u8(skb, IPSET_ATTR_NETMASK, h->netmask))
1078 goto nla_put_failure;
1079 #endif
1080 #ifdef IP_SET_HASH_WITH_MARKMASK
1081 if (nla_put_u32(skb, IPSET_ATTR_MARKMASK, h->markmask))
1082 goto nla_put_failure;
1083 #endif
1084 if (nla_put_net32(skb, IPSET_ATTR_REFERENCES, htonl(set->ref)) ||
1085 nla_put_net32(skb, IPSET_ATTR_MEMSIZE, htonl(memsize)) ||
1086 nla_put_net32(skb, IPSET_ATTR_ELEMENTS, htonl(set->elements)))
1087 goto nla_put_failure;
1088 if (unlikely(ip_set_put_flags(skb, set)))
1089 goto nla_put_failure;
1090 ipset_nest_end(skb, nested);
1091
1092 return 0;
1093 nla_put_failure:
1094 return -EMSGSIZE;
1095 }
1096
1097 /* Make possible to run dumping parallel with resizing */
1098 static void
1099 mtype_uref(struct ip_set *set, struct netlink_callback *cb, bool start)
1100 {
1101 struct htype *h = set->data;
1102 struct htable *t;
1103
1104 if (start) {
1105 rcu_read_lock_bh();
1106 t = rcu_dereference_bh_nfnl(h->table);
1107 atomic_inc(&t->uref);
1108 cb->args[IPSET_CB_PRIVATE] = (unsigned long)t;
1109 rcu_read_unlock_bh();
1110 } else if (cb->args[IPSET_CB_PRIVATE]) {
1111 t = (struct htable *)cb->args[IPSET_CB_PRIVATE];
1112 if (atomic_dec_and_test(&t->uref) && atomic_read(&t->ref)) {
1113 /* Resizing didn't destroy the hash table */
1114 pr_debug("Table destroy by dump: %p\n", t);
1115 mtype_ahash_destroy(set, t, false);
1116 }
1117 cb->args[IPSET_CB_PRIVATE] = 0;
1118 }
1119 }
1120
1121 /* Reply a LIST/SAVE request: dump the elements of the specified set */
1122 static int
1123 mtype_list(const struct ip_set *set,
1124 struct sk_buff *skb, struct netlink_callback *cb)
1125 {
1126 const struct htable *t;
1127 struct nlattr *atd, *nested;
1128 const struct hbucket *n;
1129 const struct mtype_elem *e;
1130 u32 first = cb->args[IPSET_CB_ARG0];
1131 /* We assume that one hash bucket fills into one page */
1132 void *incomplete;
1133 int i, ret = 0;
1134
1135 atd = ipset_nest_start(skb, IPSET_ATTR_ADT);
1136 if (!atd)
1137 return -EMSGSIZE;
1138
1139 pr_debug("list hash set %s\n", set->name);
1140 t = (const struct htable *)cb->args[IPSET_CB_PRIVATE];
1141 /* Expire may replace a hbucket with another one */
1142 rcu_read_lock();
1143 for (; cb->args[IPSET_CB_ARG0] < jhash_size(t->htable_bits);
1144 cb->args[IPSET_CB_ARG0]++) {
1145 incomplete = skb_tail_pointer(skb);
1146 n = rcu_dereference(hbucket(t, cb->args[IPSET_CB_ARG0]));
1147 pr_debug("cb->arg bucket: %lu, t %p n %p\n",
1148 cb->args[IPSET_CB_ARG0], t, n);
1149 if (!n)
1150 continue;
1151 for (i = 0; i < n->pos; i++) {
1152 if (!test_bit(i, n->used))
1153 continue;
1154 e = ahash_data(n, i, set->dsize);
1155 if (SET_WITH_TIMEOUT(set) &&
1156 ip_set_timeout_expired(ext_timeout(e, set)))
1157 continue;
1158 pr_debug("list hash %lu hbucket %p i %u, data %p\n",
1159 cb->args[IPSET_CB_ARG0], n, i, e);
1160 nested = ipset_nest_start(skb, IPSET_ATTR_DATA);
1161 if (!nested) {
1162 if (cb->args[IPSET_CB_ARG0] == first) {
1163 nla_nest_cancel(skb, atd);
1164 ret = -EMSGSIZE;
1165 goto out;
1166 }
1167 goto nla_put_failure;
1168 }
1169 if (mtype_data_list(skb, e))
1170 goto nla_put_failure;
1171 if (ip_set_put_extensions(skb, set, e, true))
1172 goto nla_put_failure;
1173 ipset_nest_end(skb, nested);
1174 }
1175 }
1176 ipset_nest_end(skb, atd);
1177 /* Set listing finished */
1178 cb->args[IPSET_CB_ARG0] = 0;
1179
1180 goto out;
1181
1182 nla_put_failure:
1183 nlmsg_trim(skb, incomplete);
1184 if (unlikely(first == cb->args[IPSET_CB_ARG0])) {
1185 pr_warn("Can't list set %s: one bucket does not fit into a message. Please report it!\n",
1186 set->name);
1187 cb->args[IPSET_CB_ARG0] = 0;
1188 ret = -EMSGSIZE;
1189 } else {
1190 ipset_nest_end(skb, atd);
1191 }
1192 out:
1193 rcu_read_unlock();
1194 return ret;
1195 }
1196
1197 static int
1198 IPSET_TOKEN(MTYPE, _kadt)(struct ip_set *set, const struct sk_buff *skb,
1199 const struct xt_action_param *par,
1200 enum ipset_adt adt, struct ip_set_adt_opt *opt);
1201
1202 static int
1203 IPSET_TOKEN(MTYPE, _uadt)(struct ip_set *set, struct nlattr *tb[],
1204 enum ipset_adt adt, u32 *lineno, u32 flags,
1205 bool retried);
1206
1207 static const struct ip_set_type_variant mtype_variant = {
1208 .kadt = mtype_kadt,
1209 .uadt = mtype_uadt,
1210 .adt = {
1211 [IPSET_ADD] = mtype_add,
1212 [IPSET_DEL] = mtype_del,
1213 [IPSET_TEST] = mtype_test,
1214 },
1215 .destroy = mtype_destroy,
1216 .flush = mtype_flush,
1217 .head = mtype_head,
1218 .list = mtype_list,
1219 .uref = mtype_uref,
1220 .resize = mtype_resize,
1221 .same_set = mtype_same_set,
1222 };
1223
1224 #ifdef IP_SET_EMIT_CREATE
1225 static int
1226 IPSET_TOKEN(HTYPE, _create)(struct net *net, struct ip_set *set,
1227 struct nlattr *tb[], u32 flags)
1228 {
1229 u32 hashsize = IPSET_DEFAULT_HASHSIZE, maxelem = IPSET_DEFAULT_MAXELEM;
1230 #ifdef IP_SET_HASH_WITH_MARKMASK
1231 u32 markmask;
1232 #endif
1233 u8 hbits;
1234 #ifdef IP_SET_HASH_WITH_NETMASK
1235 u8 netmask;
1236 #endif
1237 size_t hsize;
1238 struct htype *h;
1239 struct htable *t;
1240
1241 #ifndef IP_SET_PROTO_UNDEF
1242 if (!(set->family == NFPROTO_IPV4 || set->family == NFPROTO_IPV6))
1243 return -IPSET_ERR_INVALID_FAMILY;
1244 #endif
1245
1246 #ifdef IP_SET_HASH_WITH_MARKMASK
1247 markmask = 0xffffffff;
1248 #endif
1249 #ifdef IP_SET_HASH_WITH_NETMASK
1250 netmask = set->family == NFPROTO_IPV4 ? 32 : 128;
1251 pr_debug("Create set %s with family %s\n",
1252 set->name, set->family == NFPROTO_IPV4 ? "inet" : "inet6");
1253 #endif
1254
1255 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_HASHSIZE) ||
1256 !ip_set_optattr_netorder(tb, IPSET_ATTR_MAXELEM) ||
1257 !ip_set_optattr_netorder(tb, IPSET_ATTR_TIMEOUT) ||
1258 !ip_set_optattr_netorder(tb, IPSET_ATTR_CADT_FLAGS)))
1259 return -IPSET_ERR_PROTOCOL;
1260 #ifdef IP_SET_HASH_WITH_MARKMASK
1261 /* Separated condition in order to avoid directive in argument list */
1262 if (unlikely(!ip_set_optattr_netorder(tb, IPSET_ATTR_MARKMASK)))
1263 return -IPSET_ERR_PROTOCOL;
1264 #endif
1265
1266 if (tb[IPSET_ATTR_HASHSIZE]) {
1267 hashsize = ip_set_get_h32(tb[IPSET_ATTR_HASHSIZE]);
1268 if (hashsize < IPSET_MIMINAL_HASHSIZE)
1269 hashsize = IPSET_MIMINAL_HASHSIZE;
1270 }
1271
1272 if (tb[IPSET_ATTR_MAXELEM])
1273 maxelem = ip_set_get_h32(tb[IPSET_ATTR_MAXELEM]);
1274
1275 #ifdef IP_SET_HASH_WITH_NETMASK
1276 if (tb[IPSET_ATTR_NETMASK]) {
1277 netmask = nla_get_u8(tb[IPSET_ATTR_NETMASK]);
1278
1279 if ((set->family == NFPROTO_IPV4 && netmask > 32) ||
1280 (set->family == NFPROTO_IPV6 && netmask > 128) ||
1281 netmask == 0)
1282 return -IPSET_ERR_INVALID_NETMASK;
1283 }
1284 #endif
1285 #ifdef IP_SET_HASH_WITH_MARKMASK
1286 if (tb[IPSET_ATTR_MARKMASK]) {
1287 markmask = ntohl(nla_get_be32(tb[IPSET_ATTR_MARKMASK]));
1288
1289 if (markmask == 0)
1290 return -IPSET_ERR_INVALID_MARKMASK;
1291 }
1292 #endif
1293
1294 hsize = sizeof(*h);
1295 #ifdef IP_SET_HASH_WITH_NETS
1296 hsize += sizeof(struct net_prefixes) * NLEN(set->family);
1297 #endif
1298 h = kzalloc(hsize, GFP_KERNEL);
1299 if (!h)
1300 return -ENOMEM;
1301
1302 h->maxelem = maxelem;
1303 #ifdef IP_SET_HASH_WITH_NETMASK
1304 h->netmask = netmask;
1305 #endif
1306 #ifdef IP_SET_HASH_WITH_MARKMASK
1307 h->markmask = markmask;
1308 #endif
1309 get_random_bytes(&h->initval, sizeof(h->initval));
1310 set->timeout = IPSET_NO_TIMEOUT;
1311
1312 hbits = htable_bits(hashsize);
1313 hsize = htable_size(hbits);
1314 if (hsize == 0) {
1315 kfree(h);
1316 return -ENOMEM;
1317 }
1318 t = ip_set_alloc(hsize);
1319 if (!t) {
1320 kfree(h);
1321 return -ENOMEM;
1322 }
1323 t->htable_bits = hbits;
1324 rcu_assign_pointer(h->table, t);
1325
1326 set->data = h;
1327 #ifndef IP_SET_PROTO_UNDEF
1328 if (set->family == NFPROTO_IPV4) {
1329 #endif
1330 set->variant = &IPSET_TOKEN(HTYPE, 4_variant);
1331 set->dsize = ip_set_elem_len(set, tb,
1332 sizeof(struct IPSET_TOKEN(HTYPE, 4_elem)),
1333 __alignof__(struct IPSET_TOKEN(HTYPE, 4_elem)));
1334 #ifndef IP_SET_PROTO_UNDEF
1335 } else {
1336 set->variant = &IPSET_TOKEN(HTYPE, 6_variant);
1337 set->dsize = ip_set_elem_len(set, tb,
1338 sizeof(struct IPSET_TOKEN(HTYPE, 6_elem)),
1339 __alignof__(struct IPSET_TOKEN(HTYPE, 6_elem)));
1340 }
1341 #endif
1342 if (tb[IPSET_ATTR_TIMEOUT]) {
1343 set->timeout = ip_set_timeout_uget(tb[IPSET_ATTR_TIMEOUT]);
1344 #ifndef IP_SET_PROTO_UNDEF
1345 if (set->family == NFPROTO_IPV4)
1346 #endif
1347 IPSET_TOKEN(HTYPE, 4_gc_init)(set,
1348 IPSET_TOKEN(HTYPE, 4_gc));
1349 #ifndef IP_SET_PROTO_UNDEF
1350 else
1351 IPSET_TOKEN(HTYPE, 6_gc_init)(set,
1352 IPSET_TOKEN(HTYPE, 6_gc));
1353 #endif
1354 }
1355 pr_debug("create %s hashsize %u (%u) maxelem %u: %p(%p)\n",
1356 set->name, jhash_size(t->htable_bits),
1357 t->htable_bits, h->maxelem, set->data, t);
1358
1359 return 0;
1360 }
1361 #endif /* IP_SET_EMIT_CREATE */
1362
1363 #undef HKEY_DATALEN
Ubuntu Hirsute Linux kernel mirror