Loading...
1/*
2 * x_tables core - Backend for {ip,ip6,arp}_tables
3 *
4 * Copyright (C) 2006-2006 Harald Welte <laforge@netfilter.org>
5 * Copyright (C) 2006-2012 Patrick McHardy <kaber@trash.net>
6 *
7 * Based on existing ip_tables code which is
8 * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
9 * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
10 *
11 * This program is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License version 2 as
13 * published by the Free Software Foundation.
14 *
15 */
16#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17#include <linux/kernel.h>
18#include <linux/module.h>
19#include <linux/socket.h>
20#include <linux/net.h>
21#include <linux/proc_fs.h>
22#include <linux/seq_file.h>
23#include <linux/string.h>
24#include <linux/vmalloc.h>
25#include <linux/mutex.h>
26#include <linux/mm.h>
27#include <linux/slab.h>
28#include <linux/audit.h>
29#include <linux/user_namespace.h>
30#include <net/net_namespace.h>
31
32#include <linux/netfilter/x_tables.h>
33#include <linux/netfilter_arp.h>
34#include <linux/netfilter_ipv4/ip_tables.h>
35#include <linux/netfilter_ipv6/ip6_tables.h>
36#include <linux/netfilter_arp/arp_tables.h>
37
38MODULE_LICENSE("GPL");
39MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
40MODULE_DESCRIPTION("{ip,ip6,arp,eb}_tables backend module");
41
42#define SMP_ALIGN(x) (((x) + SMP_CACHE_BYTES-1) & ~(SMP_CACHE_BYTES-1))
43#define XT_PCPU_BLOCK_SIZE 4096
44
45struct compat_delta {
46 unsigned int offset; /* offset in kernel */
47 int delta; /* delta in 32bit user land */
48};
49
50struct xt_af {
51 struct mutex mutex;
52 struct list_head match;
53 struct list_head target;
54#ifdef CONFIG_COMPAT
55 struct mutex compat_mutex;
56 struct compat_delta *compat_tab;
57 unsigned int number; /* number of slots in compat_tab[] */
58 unsigned int cur; /* number of used slots in compat_tab[] */
59#endif
60};
61
62static struct xt_af *xt;
63
64static const char *const xt_prefix[NFPROTO_NUMPROTO] = {
65 [NFPROTO_UNSPEC] = "x",
66 [NFPROTO_IPV4] = "ip",
67 [NFPROTO_ARP] = "arp",
68 [NFPROTO_BRIDGE] = "eb",
69 [NFPROTO_IPV6] = "ip6",
70};
71
72/* Registration hooks for targets. */
73int xt_register_target(struct xt_target *target)
74{
75 u_int8_t af = target->family;
76
77 mutex_lock(&xt[af].mutex);
78 list_add(&target->list, &xt[af].target);
79 mutex_unlock(&xt[af].mutex);
80 return 0;
81}
82EXPORT_SYMBOL(xt_register_target);
83
84void
85xt_unregister_target(struct xt_target *target)
86{
87 u_int8_t af = target->family;
88
89 mutex_lock(&xt[af].mutex);
90 list_del(&target->list);
91 mutex_unlock(&xt[af].mutex);
92}
93EXPORT_SYMBOL(xt_unregister_target);
94
95int
96xt_register_targets(struct xt_target *target, unsigned int n)
97{
98 unsigned int i;
99 int err = 0;
100
101 for (i = 0; i < n; i++) {
102 err = xt_register_target(&target[i]);
103 if (err)
104 goto err;
105 }
106 return err;
107
108err:
109 if (i > 0)
110 xt_unregister_targets(target, i);
111 return err;
112}
113EXPORT_SYMBOL(xt_register_targets);
114
115void
116xt_unregister_targets(struct xt_target *target, unsigned int n)
117{
118 while (n-- > 0)
119 xt_unregister_target(&target[n]);
120}
121EXPORT_SYMBOL(xt_unregister_targets);
122
123int xt_register_match(struct xt_match *match)
124{
125 u_int8_t af = match->family;
126
127 mutex_lock(&xt[af].mutex);
128 list_add(&match->list, &xt[af].match);
129 mutex_unlock(&xt[af].mutex);
130 return 0;
131}
132EXPORT_SYMBOL(xt_register_match);
133
134void
135xt_unregister_match(struct xt_match *match)
136{
137 u_int8_t af = match->family;
138
139 mutex_lock(&xt[af].mutex);
140 list_del(&match->list);
141 mutex_unlock(&xt[af].mutex);
142}
143EXPORT_SYMBOL(xt_unregister_match);
144
145int
146xt_register_matches(struct xt_match *match, unsigned int n)
147{
148 unsigned int i;
149 int err = 0;
150
151 for (i = 0; i < n; i++) {
152 err = xt_register_match(&match[i]);
153 if (err)
154 goto err;
155 }
156 return err;
157
158err:
159 if (i > 0)
160 xt_unregister_matches(match, i);
161 return err;
162}
163EXPORT_SYMBOL(xt_register_matches);
164
165void
166xt_unregister_matches(struct xt_match *match, unsigned int n)
167{
168 while (n-- > 0)
169 xt_unregister_match(&match[n]);
170}
171EXPORT_SYMBOL(xt_unregister_matches);
172
173
174/*
175 * These are weird, but module loading must not be done with mutex
176 * held (since they will register), and we have to have a single
177 * function to use.
178 */
179
180/* Find match, grabs ref. Returns ERR_PTR() on error. */
181struct xt_match *xt_find_match(u8 af, const char *name, u8 revision)
182{
183 struct xt_match *m;
184 int err = -ENOENT;
185
186 mutex_lock(&xt[af].mutex);
187 list_for_each_entry(m, &xt[af].match, list) {
188 if (strcmp(m->name, name) == 0) {
189 if (m->revision == revision) {
190 if (try_module_get(m->me)) {
191 mutex_unlock(&xt[af].mutex);
192 return m;
193 }
194 } else
195 err = -EPROTOTYPE; /* Found something. */
196 }
197 }
198 mutex_unlock(&xt[af].mutex);
199
200 if (af != NFPROTO_UNSPEC)
201 /* Try searching again in the family-independent list */
202 return xt_find_match(NFPROTO_UNSPEC, name, revision);
203
204 return ERR_PTR(err);
205}
206EXPORT_SYMBOL(xt_find_match);
207
208struct xt_match *
209xt_request_find_match(uint8_t nfproto, const char *name, uint8_t revision)
210{
211 struct xt_match *match;
212
213 match = xt_find_match(nfproto, name, revision);
214 if (IS_ERR(match)) {
215 request_module("%st_%s", xt_prefix[nfproto], name);
216 match = xt_find_match(nfproto, name, revision);
217 }
218
219 return match;
220}
221EXPORT_SYMBOL_GPL(xt_request_find_match);
222
223/* Find target, grabs ref. Returns ERR_PTR() on error. */
224struct xt_target *xt_find_target(u8 af, const char *name, u8 revision)
225{
226 struct xt_target *t;
227 int err = -ENOENT;
228
229 mutex_lock(&xt[af].mutex);
230 list_for_each_entry(t, &xt[af].target, list) {
231 if (strcmp(t->name, name) == 0) {
232 if (t->revision == revision) {
233 if (try_module_get(t->me)) {
234 mutex_unlock(&xt[af].mutex);
235 return t;
236 }
237 } else
238 err = -EPROTOTYPE; /* Found something. */
239 }
240 }
241 mutex_unlock(&xt[af].mutex);
242
243 if (af != NFPROTO_UNSPEC)
244 /* Try searching again in the family-independent list */
245 return xt_find_target(NFPROTO_UNSPEC, name, revision);
246
247 return ERR_PTR(err);
248}
249EXPORT_SYMBOL(xt_find_target);
250
251struct xt_target *xt_request_find_target(u8 af, const char *name, u8 revision)
252{
253 struct xt_target *target;
254
255 target = xt_find_target(af, name, revision);
256 if (IS_ERR(target)) {
257 request_module("%st_%s", xt_prefix[af], name);
258 target = xt_find_target(af, name, revision);
259 }
260
261 return target;
262}
263EXPORT_SYMBOL_GPL(xt_request_find_target);
264
265static int match_revfn(u8 af, const char *name, u8 revision, int *bestp)
266{
267 const struct xt_match *m;
268 int have_rev = 0;
269
270 list_for_each_entry(m, &xt[af].match, list) {
271 if (strcmp(m->name, name) == 0) {
272 if (m->revision > *bestp)
273 *bestp = m->revision;
274 if (m->revision == revision)
275 have_rev = 1;
276 }
277 }
278
279 if (af != NFPROTO_UNSPEC && !have_rev)
280 return match_revfn(NFPROTO_UNSPEC, name, revision, bestp);
281
282 return have_rev;
283}
284
285static int target_revfn(u8 af, const char *name, u8 revision, int *bestp)
286{
287 const struct xt_target *t;
288 int have_rev = 0;
289
290 list_for_each_entry(t, &xt[af].target, list) {
291 if (strcmp(t->name, name) == 0) {
292 if (t->revision > *bestp)
293 *bestp = t->revision;
294 if (t->revision == revision)
295 have_rev = 1;
296 }
297 }
298
299 if (af != NFPROTO_UNSPEC && !have_rev)
300 return target_revfn(NFPROTO_UNSPEC, name, revision, bestp);
301
302 return have_rev;
303}
304
305/* Returns true or false (if no such extension at all) */
306int xt_find_revision(u8 af, const char *name, u8 revision, int target,
307 int *err)
308{
309 int have_rev, best = -1;
310
311 mutex_lock(&xt[af].mutex);
312 if (target == 1)
313 have_rev = target_revfn(af, name, revision, &best);
314 else
315 have_rev = match_revfn(af, name, revision, &best);
316 mutex_unlock(&xt[af].mutex);
317
318 /* Nothing at all? Return 0 to try loading module. */
319 if (best == -1) {
320 *err = -ENOENT;
321 return 0;
322 }
323
324 *err = best;
325 if (!have_rev)
326 *err = -EPROTONOSUPPORT;
327 return 1;
328}
329EXPORT_SYMBOL_GPL(xt_find_revision);
330
331static char *
332textify_hooks(char *buf, size_t size, unsigned int mask, uint8_t nfproto)
333{
334 static const char *const inetbr_names[] = {
335 "PREROUTING", "INPUT", "FORWARD",
336 "OUTPUT", "POSTROUTING", "BROUTING",
337 };
338 static const char *const arp_names[] = {
339 "INPUT", "FORWARD", "OUTPUT",
340 };
341 const char *const *names;
342 unsigned int i, max;
343 char *p = buf;
344 bool np = false;
345 int res;
346
347 names = (nfproto == NFPROTO_ARP) ? arp_names : inetbr_names;
348 max = (nfproto == NFPROTO_ARP) ? ARRAY_SIZE(arp_names) :
349 ARRAY_SIZE(inetbr_names);
350 *p = '\0';
351 for (i = 0; i < max; ++i) {
352 if (!(mask & (1 << i)))
353 continue;
354 res = snprintf(p, size, "%s%s", np ? "/" : "", names[i]);
355 if (res > 0) {
356 size -= res;
357 p += res;
358 }
359 np = true;
360 }
361
362 return buf;
363}
364
365int xt_check_match(struct xt_mtchk_param *par,
366 unsigned int size, u_int8_t proto, bool inv_proto)
367{
368 int ret;
369
370 if (XT_ALIGN(par->match->matchsize) != size &&
371 par->match->matchsize != -1) {
372 /*
373 * ebt_among is exempt from centralized matchsize checking
374 * because it uses a dynamic-size data set.
375 */
376 pr_err("%s_tables: %s.%u match: invalid size "
377 "%u (kernel) != (user) %u\n",
378 xt_prefix[par->family], par->match->name,
379 par->match->revision,
380 XT_ALIGN(par->match->matchsize), size);
381 return -EINVAL;
382 }
383 if (par->match->table != NULL &&
384 strcmp(par->match->table, par->table) != 0) {
385 pr_err("%s_tables: %s match: only valid in %s table, not %s\n",
386 xt_prefix[par->family], par->match->name,
387 par->match->table, par->table);
388 return -EINVAL;
389 }
390 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) {
391 char used[64], allow[64];
392
393 pr_err("%s_tables: %s match: used from hooks %s, but only "
394 "valid from %s\n",
395 xt_prefix[par->family], par->match->name,
396 textify_hooks(used, sizeof(used), par->hook_mask,
397 par->family),
398 textify_hooks(allow, sizeof(allow), par->match->hooks,
399 par->family));
400 return -EINVAL;
401 }
402 if (par->match->proto && (par->match->proto != proto || inv_proto)) {
403 pr_err("%s_tables: %s match: only valid for protocol %u\n",
404 xt_prefix[par->family], par->match->name,
405 par->match->proto);
406 return -EINVAL;
407 }
408 if (par->match->checkentry != NULL) {
409 ret = par->match->checkentry(par);
410 if (ret < 0)
411 return ret;
412 else if (ret > 0)
413 /* Flag up potential errors. */
414 return -EIO;
415 }
416 return 0;
417}
418EXPORT_SYMBOL_GPL(xt_check_match);
419
420/** xt_check_entry_match - check that matches end before start of target
421 *
422 * @match: beginning of xt_entry_match
423 * @target: beginning of this rules target (alleged end of matches)
424 * @alignment: alignment requirement of match structures
425 *
426 * Validates that all matches add up to the beginning of the target,
427 * and that each match covers at least the base structure size.
428 *
429 * Return: 0 on success, negative errno on failure.
430 */
431static int xt_check_entry_match(const char *match, const char *target,
432 const size_t alignment)
433{
434 const struct xt_entry_match *pos;
435 int length = target - match;
436
437 if (length == 0) /* no matches */
438 return 0;
439
440 pos = (struct xt_entry_match *)match;
441 do {
442 if ((unsigned long)pos % alignment)
443 return -EINVAL;
444
445 if (length < (int)sizeof(struct xt_entry_match))
446 return -EINVAL;
447
448 if (pos->u.match_size < sizeof(struct xt_entry_match))
449 return -EINVAL;
450
451 if (pos->u.match_size > length)
452 return -EINVAL;
453
454 length -= pos->u.match_size;
455 pos = ((void *)((char *)(pos) + (pos)->u.match_size));
456 } while (length > 0);
457
458 return 0;
459}
460
461#ifdef CONFIG_COMPAT
462int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta)
463{
464 struct xt_af *xp = &xt[af];
465
466 if (!xp->compat_tab) {
467 if (!xp->number)
468 return -EINVAL;
469 xp->compat_tab = vmalloc(sizeof(struct compat_delta) * xp->number);
470 if (!xp->compat_tab)
471 return -ENOMEM;
472 xp->cur = 0;
473 }
474
475 if (xp->cur >= xp->number)
476 return -EINVAL;
477
478 if (xp->cur)
479 delta += xp->compat_tab[xp->cur - 1].delta;
480 xp->compat_tab[xp->cur].offset = offset;
481 xp->compat_tab[xp->cur].delta = delta;
482 xp->cur++;
483 return 0;
484}
485EXPORT_SYMBOL_GPL(xt_compat_add_offset);
486
487void xt_compat_flush_offsets(u_int8_t af)
488{
489 if (xt[af].compat_tab) {
490 vfree(xt[af].compat_tab);
491 xt[af].compat_tab = NULL;
492 xt[af].number = 0;
493 xt[af].cur = 0;
494 }
495}
496EXPORT_SYMBOL_GPL(xt_compat_flush_offsets);
497
498int xt_compat_calc_jump(u_int8_t af, unsigned int offset)
499{
500 struct compat_delta *tmp = xt[af].compat_tab;
501 int mid, left = 0, right = xt[af].cur - 1;
502
503 while (left <= right) {
504 mid = (left + right) >> 1;
505 if (offset > tmp[mid].offset)
506 left = mid + 1;
507 else if (offset < tmp[mid].offset)
508 right = mid - 1;
509 else
510 return mid ? tmp[mid - 1].delta : 0;
511 }
512 return left ? tmp[left - 1].delta : 0;
513}
514EXPORT_SYMBOL_GPL(xt_compat_calc_jump);
515
516void xt_compat_init_offsets(u_int8_t af, unsigned int number)
517{
518 xt[af].number = number;
519 xt[af].cur = 0;
520}
521EXPORT_SYMBOL(xt_compat_init_offsets);
522
523int xt_compat_match_offset(const struct xt_match *match)
524{
525 u_int16_t csize = match->compatsize ? : match->matchsize;
526 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize);
527}
528EXPORT_SYMBOL_GPL(xt_compat_match_offset);
529
530void xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr,
531 unsigned int *size)
532{
533 const struct xt_match *match = m->u.kernel.match;
534 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m;
535 int pad, off = xt_compat_match_offset(match);
536 u_int16_t msize = cm->u.user.match_size;
537 char name[sizeof(m->u.user.name)];
538
539 m = *dstptr;
540 memcpy(m, cm, sizeof(*cm));
541 if (match->compat_from_user)
542 match->compat_from_user(m->data, cm->data);
543 else
544 memcpy(m->data, cm->data, msize - sizeof(*cm));
545 pad = XT_ALIGN(match->matchsize) - match->matchsize;
546 if (pad > 0)
547 memset(m->data + match->matchsize, 0, pad);
548
549 msize += off;
550 m->u.user.match_size = msize;
551 strlcpy(name, match->name, sizeof(name));
552 module_put(match->me);
553 strncpy(m->u.user.name, name, sizeof(m->u.user.name));
554
555 *size += off;
556 *dstptr += msize;
557}
558EXPORT_SYMBOL_GPL(xt_compat_match_from_user);
559
560int xt_compat_match_to_user(const struct xt_entry_match *m,
561 void __user **dstptr, unsigned int *size)
562{
563 const struct xt_match *match = m->u.kernel.match;
564 struct compat_xt_entry_match __user *cm = *dstptr;
565 int off = xt_compat_match_offset(match);
566 u_int16_t msize = m->u.user.match_size - off;
567
568 if (copy_to_user(cm, m, sizeof(*cm)) ||
569 put_user(msize, &cm->u.user.match_size) ||
570 copy_to_user(cm->u.user.name, m->u.kernel.match->name,
571 strlen(m->u.kernel.match->name) + 1))
572 return -EFAULT;
573
574 if (match->compat_to_user) {
575 if (match->compat_to_user((void __user *)cm->data, m->data))
576 return -EFAULT;
577 } else {
578 if (copy_to_user(cm->data, m->data, msize - sizeof(*cm)))
579 return -EFAULT;
580 }
581
582 *size -= off;
583 *dstptr += msize;
584 return 0;
585}
586EXPORT_SYMBOL_GPL(xt_compat_match_to_user);
587
588/* non-compat version may have padding after verdict */
589struct compat_xt_standard_target {
590 struct compat_xt_entry_target t;
591 compat_uint_t verdict;
592};
593
594int xt_compat_check_entry_offsets(const void *base, const char *elems,
595 unsigned int target_offset,
596 unsigned int next_offset)
597{
598 long size_of_base_struct = elems - (const char *)base;
599 const struct compat_xt_entry_target *t;
600 const char *e = base;
601
602 if (target_offset < size_of_base_struct)
603 return -EINVAL;
604
605 if (target_offset + sizeof(*t) > next_offset)
606 return -EINVAL;
607
608 t = (void *)(e + target_offset);
609 if (t->u.target_size < sizeof(*t))
610 return -EINVAL;
611
612 if (target_offset + t->u.target_size > next_offset)
613 return -EINVAL;
614
615 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
616 COMPAT_XT_ALIGN(target_offset + sizeof(struct compat_xt_standard_target)) != next_offset)
617 return -EINVAL;
618
619 /* compat_xt_entry match has less strict aligment requirements,
620 * otherwise they are identical. In case of padding differences
621 * we need to add compat version of xt_check_entry_match.
622 */
623 BUILD_BUG_ON(sizeof(struct compat_xt_entry_match) != sizeof(struct xt_entry_match));
624
625 return xt_check_entry_match(elems, base + target_offset,
626 __alignof__(struct compat_xt_entry_match));
627}
628EXPORT_SYMBOL(xt_compat_check_entry_offsets);
629#endif /* CONFIG_COMPAT */
630
631/**
632 * xt_check_entry_offsets - validate arp/ip/ip6t_entry
633 *
634 * @base: pointer to arp/ip/ip6t_entry
635 * @elems: pointer to first xt_entry_match, i.e. ip(6)t_entry->elems
636 * @target_offset: the arp/ip/ip6_t->target_offset
637 * @next_offset: the arp/ip/ip6_t->next_offset
638 *
639 * validates that target_offset and next_offset are sane and that all
640 * match sizes (if any) align with the target offset.
641 *
642 * This function does not validate the targets or matches themselves, it
643 * only tests that all the offsets and sizes are correct, that all
644 * match structures are aligned, and that the last structure ends where
645 * the target structure begins.
646 *
647 * Also see xt_compat_check_entry_offsets for CONFIG_COMPAT version.
648 *
649 * The arp/ip/ip6t_entry structure @base must have passed following tests:
650 * - it must point to a valid memory location
651 * - base to base + next_offset must be accessible, i.e. not exceed allocated
652 * length.
653 *
654 * A well-formed entry looks like this:
655 *
656 * ip(6)t_entry match [mtdata] match [mtdata] target [tgdata] ip(6)t_entry
657 * e->elems[]-----' | |
658 * matchsize | |
659 * matchsize | |
660 * | |
661 * target_offset---------------------------------' |
662 * next_offset---------------------------------------------------'
663 *
664 * elems[]: flexible array member at end of ip(6)/arpt_entry struct.
665 * This is where matches (if any) and the target reside.
666 * target_offset: beginning of target.
667 * next_offset: start of the next rule; also: size of this rule.
668 * Since targets have a minimum size, target_offset + minlen <= next_offset.
669 *
670 * Every match stores its size, sum of sizes must not exceed target_offset.
671 *
672 * Return: 0 on success, negative errno on failure.
673 */
674int xt_check_entry_offsets(const void *base,
675 const char *elems,
676 unsigned int target_offset,
677 unsigned int next_offset)
678{
679 long size_of_base_struct = elems - (const char *)base;
680 const struct xt_entry_target *t;
681 const char *e = base;
682
683 /* target start is within the ip/ip6/arpt_entry struct */
684 if (target_offset < size_of_base_struct)
685 return -EINVAL;
686
687 if (target_offset + sizeof(*t) > next_offset)
688 return -EINVAL;
689
690 t = (void *)(e + target_offset);
691 if (t->u.target_size < sizeof(*t))
692 return -EINVAL;
693
694 if (target_offset + t->u.target_size > next_offset)
695 return -EINVAL;
696
697 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
698 XT_ALIGN(target_offset + sizeof(struct xt_standard_target)) != next_offset)
699 return -EINVAL;
700
701 return xt_check_entry_match(elems, base + target_offset,
702 __alignof__(struct xt_entry_match));
703}
704EXPORT_SYMBOL(xt_check_entry_offsets);
705
706/**
707 * xt_alloc_entry_offsets - allocate array to store rule head offsets
708 *
709 * @size: number of entries
710 *
711 * Return: NULL or kmalloc'd or vmalloc'd array
712 */
713unsigned int *xt_alloc_entry_offsets(unsigned int size)
714{
715 unsigned int *off;
716
717 off = kcalloc(size, sizeof(unsigned int), GFP_KERNEL | __GFP_NOWARN);
718
719 if (off)
720 return off;
721
722 if (size < (SIZE_MAX / sizeof(unsigned int)))
723 off = vmalloc(size * sizeof(unsigned int));
724
725 return off;
726}
727EXPORT_SYMBOL(xt_alloc_entry_offsets);
728
729/**
730 * xt_find_jump_offset - check if target is a valid jump offset
731 *
732 * @offsets: array containing all valid rule start offsets of a rule blob
733 * @target: the jump target to search for
734 * @size: entries in @offset
735 */
736bool xt_find_jump_offset(const unsigned int *offsets,
737 unsigned int target, unsigned int size)
738{
739 int m, low = 0, hi = size;
740
741 while (hi > low) {
742 m = (low + hi) / 2u;
743
744 if (offsets[m] > target)
745 hi = m;
746 else if (offsets[m] < target)
747 low = m + 1;
748 else
749 return true;
750 }
751
752 return false;
753}
754EXPORT_SYMBOL(xt_find_jump_offset);
755
756int xt_check_target(struct xt_tgchk_param *par,
757 unsigned int size, u_int8_t proto, bool inv_proto)
758{
759 int ret;
760
761 if (XT_ALIGN(par->target->targetsize) != size) {
762 pr_err("%s_tables: %s.%u target: invalid size "
763 "%u (kernel) != (user) %u\n",
764 xt_prefix[par->family], par->target->name,
765 par->target->revision,
766 XT_ALIGN(par->target->targetsize), size);
767 return -EINVAL;
768 }
769 if (par->target->table != NULL &&
770 strcmp(par->target->table, par->table) != 0) {
771 pr_err("%s_tables: %s target: only valid in %s table, not %s\n",
772 xt_prefix[par->family], par->target->name,
773 par->target->table, par->table);
774 return -EINVAL;
775 }
776 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) {
777 char used[64], allow[64];
778
779 pr_err("%s_tables: %s target: used from hooks %s, but only "
780 "usable from %s\n",
781 xt_prefix[par->family], par->target->name,
782 textify_hooks(used, sizeof(used), par->hook_mask,
783 par->family),
784 textify_hooks(allow, sizeof(allow), par->target->hooks,
785 par->family));
786 return -EINVAL;
787 }
788 if (par->target->proto && (par->target->proto != proto || inv_proto)) {
789 pr_err("%s_tables: %s target: only valid for protocol %u\n",
790 xt_prefix[par->family], par->target->name,
791 par->target->proto);
792 return -EINVAL;
793 }
794 if (par->target->checkentry != NULL) {
795 ret = par->target->checkentry(par);
796 if (ret < 0)
797 return ret;
798 else if (ret > 0)
799 /* Flag up potential errors. */
800 return -EIO;
801 }
802 return 0;
803}
804EXPORT_SYMBOL_GPL(xt_check_target);
805
806/**
807 * xt_copy_counters_from_user - copy counters and metadata from userspace
808 *
809 * @user: src pointer to userspace memory
810 * @len: alleged size of userspace memory
811 * @info: where to store the xt_counters_info metadata
812 * @compat: true if we setsockopt call is done by 32bit task on 64bit kernel
813 *
814 * Copies counter meta data from @user and stores it in @info.
815 *
816 * vmallocs memory to hold the counters, then copies the counter data
817 * from @user to the new memory and returns a pointer to it.
818 *
819 * If @compat is true, @info gets converted automatically to the 64bit
820 * representation.
821 *
822 * The metadata associated with the counters is stored in @info.
823 *
824 * Return: returns pointer that caller has to test via IS_ERR().
825 * If IS_ERR is false, caller has to vfree the pointer.
826 */
827void *xt_copy_counters_from_user(const void __user *user, unsigned int len,
828 struct xt_counters_info *info, bool compat)
829{
830 void *mem;
831 u64 size;
832
833#ifdef CONFIG_COMPAT
834 if (compat) {
835 /* structures only differ in size due to alignment */
836 struct compat_xt_counters_info compat_tmp;
837
838 if (len <= sizeof(compat_tmp))
839 return ERR_PTR(-EINVAL);
840
841 len -= sizeof(compat_tmp);
842 if (copy_from_user(&compat_tmp, user, sizeof(compat_tmp)) != 0)
843 return ERR_PTR(-EFAULT);
844
845 strlcpy(info->name, compat_tmp.name, sizeof(info->name));
846 info->num_counters = compat_tmp.num_counters;
847 user += sizeof(compat_tmp);
848 } else
849#endif
850 {
851 if (len <= sizeof(*info))
852 return ERR_PTR(-EINVAL);
853
854 len -= sizeof(*info);
855 if (copy_from_user(info, user, sizeof(*info)) != 0)
856 return ERR_PTR(-EFAULT);
857
858 info->name[sizeof(info->name) - 1] = '\0';
859 user += sizeof(*info);
860 }
861
862 size = sizeof(struct xt_counters);
863 size *= info->num_counters;
864
865 if (size != (u64)len)
866 return ERR_PTR(-EINVAL);
867
868 mem = vmalloc(len);
869 if (!mem)
870 return ERR_PTR(-ENOMEM);
871
872 if (copy_from_user(mem, user, len) == 0)
873 return mem;
874
875 vfree(mem);
876 return ERR_PTR(-EFAULT);
877}
878EXPORT_SYMBOL_GPL(xt_copy_counters_from_user);
879
880#ifdef CONFIG_COMPAT
881int xt_compat_target_offset(const struct xt_target *target)
882{
883 u_int16_t csize = target->compatsize ? : target->targetsize;
884 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize);
885}
886EXPORT_SYMBOL_GPL(xt_compat_target_offset);
887
888void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr,
889 unsigned int *size)
890{
891 const struct xt_target *target = t->u.kernel.target;
892 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t;
893 int pad, off = xt_compat_target_offset(target);
894 u_int16_t tsize = ct->u.user.target_size;
895 char name[sizeof(t->u.user.name)];
896
897 t = *dstptr;
898 memcpy(t, ct, sizeof(*ct));
899 if (target->compat_from_user)
900 target->compat_from_user(t->data, ct->data);
901 else
902 memcpy(t->data, ct->data, tsize - sizeof(*ct));
903 pad = XT_ALIGN(target->targetsize) - target->targetsize;
904 if (pad > 0)
905 memset(t->data + target->targetsize, 0, pad);
906
907 tsize += off;
908 t->u.user.target_size = tsize;
909 strlcpy(name, target->name, sizeof(name));
910 module_put(target->me);
911 strncpy(t->u.user.name, name, sizeof(t->u.user.name));
912
913 *size += off;
914 *dstptr += tsize;
915}
916EXPORT_SYMBOL_GPL(xt_compat_target_from_user);
917
918int xt_compat_target_to_user(const struct xt_entry_target *t,
919 void __user **dstptr, unsigned int *size)
920{
921 const struct xt_target *target = t->u.kernel.target;
922 struct compat_xt_entry_target __user *ct = *dstptr;
923 int off = xt_compat_target_offset(target);
924 u_int16_t tsize = t->u.user.target_size - off;
925
926 if (copy_to_user(ct, t, sizeof(*ct)) ||
927 put_user(tsize, &ct->u.user.target_size) ||
928 copy_to_user(ct->u.user.name, t->u.kernel.target->name,
929 strlen(t->u.kernel.target->name) + 1))
930 return -EFAULT;
931
932 if (target->compat_to_user) {
933 if (target->compat_to_user((void __user *)ct->data, t->data))
934 return -EFAULT;
935 } else {
936 if (copy_to_user(ct->data, t->data, tsize - sizeof(*ct)))
937 return -EFAULT;
938 }
939
940 *size -= off;
941 *dstptr += tsize;
942 return 0;
943}
944EXPORT_SYMBOL_GPL(xt_compat_target_to_user);
945#endif
946
947struct xt_table_info *xt_alloc_table_info(unsigned int size)
948{
949 struct xt_table_info *info = NULL;
950 size_t sz = sizeof(*info) + size;
951
952 if (sz < sizeof(*info))
953 return NULL;
954
955 /* Pedantry: prevent them from hitting BUG() in vmalloc.c --RR */
956 if ((SMP_ALIGN(size) >> PAGE_SHIFT) + 2 > totalram_pages)
957 return NULL;
958
959 if (sz <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
960 info = kmalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY);
961 if (!info) {
962 info = __vmalloc(sz, GFP_KERNEL | __GFP_NOWARN |
963 __GFP_NORETRY | __GFP_HIGHMEM,
964 PAGE_KERNEL);
965 if (!info)
966 return NULL;
967 }
968 memset(info, 0, sizeof(*info));
969 info->size = size;
970 return info;
971}
972EXPORT_SYMBOL(xt_alloc_table_info);
973
974void xt_free_table_info(struct xt_table_info *info)
975{
976 int cpu;
977
978 if (info->jumpstack != NULL) {
979 for_each_possible_cpu(cpu)
980 kvfree(info->jumpstack[cpu]);
981 kvfree(info->jumpstack);
982 }
983
984 kvfree(info);
985}
986EXPORT_SYMBOL(xt_free_table_info);
987
988/* Find table by name, grabs mutex & ref. Returns NULL on error. */
989struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af,
990 const char *name)
991{
992 struct xt_table *t, *found = NULL;
993
994 mutex_lock(&xt[af].mutex);
995 list_for_each_entry(t, &net->xt.tables[af], list)
996 if (strcmp(t->name, name) == 0 && try_module_get(t->me))
997 return t;
998
999 if (net == &init_net)
1000 goto out;
1001
1002 /* Table doesn't exist in this netns, re-try init */
1003 list_for_each_entry(t, &init_net.xt.tables[af], list) {
1004 if (strcmp(t->name, name))
1005 continue;
1006 if (!try_module_get(t->me))
1007 return NULL;
1008
1009 mutex_unlock(&xt[af].mutex);
1010 if (t->table_init(net) != 0) {
1011 module_put(t->me);
1012 return NULL;
1013 }
1014
1015 found = t;
1016
1017 mutex_lock(&xt[af].mutex);
1018 break;
1019 }
1020
1021 if (!found)
1022 goto out;
1023
1024 /* and once again: */
1025 list_for_each_entry(t, &net->xt.tables[af], list)
1026 if (strcmp(t->name, name) == 0)
1027 return t;
1028
1029 module_put(found->me);
1030 out:
1031 mutex_unlock(&xt[af].mutex);
1032 return NULL;
1033}
1034EXPORT_SYMBOL_GPL(xt_find_table_lock);
1035
1036void xt_table_unlock(struct xt_table *table)
1037{
1038 mutex_unlock(&xt[table->af].mutex);
1039}
1040EXPORT_SYMBOL_GPL(xt_table_unlock);
1041
1042#ifdef CONFIG_COMPAT
1043void xt_compat_lock(u_int8_t af)
1044{
1045 mutex_lock(&xt[af].compat_mutex);
1046}
1047EXPORT_SYMBOL_GPL(xt_compat_lock);
1048
1049void xt_compat_unlock(u_int8_t af)
1050{
1051 mutex_unlock(&xt[af].compat_mutex);
1052}
1053EXPORT_SYMBOL_GPL(xt_compat_unlock);
1054#endif
1055
1056DEFINE_PER_CPU(seqcount_t, xt_recseq);
1057EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq);
1058
1059struct static_key xt_tee_enabled __read_mostly;
1060EXPORT_SYMBOL_GPL(xt_tee_enabled);
1061
1062static int xt_jumpstack_alloc(struct xt_table_info *i)
1063{
1064 unsigned int size;
1065 int cpu;
1066
1067 size = sizeof(void **) * nr_cpu_ids;
1068 if (size > PAGE_SIZE)
1069 i->jumpstack = vzalloc(size);
1070 else
1071 i->jumpstack = kzalloc(size, GFP_KERNEL);
1072 if (i->jumpstack == NULL)
1073 return -ENOMEM;
1074
1075 /* ruleset without jumps -- no stack needed */
1076 if (i->stacksize == 0)
1077 return 0;
1078
1079 /* Jumpstack needs to be able to record two full callchains, one
1080 * from the first rule set traversal, plus one table reentrancy
1081 * via -j TEE without clobbering the callchain that brought us to
1082 * TEE target.
1083 *
1084 * This is done by allocating two jumpstacks per cpu, on reentry
1085 * the upper half of the stack is used.
1086 *
1087 * see the jumpstack setup in ipt_do_table() for more details.
1088 */
1089 size = sizeof(void *) * i->stacksize * 2u;
1090 for_each_possible_cpu(cpu) {
1091 if (size > PAGE_SIZE)
1092 i->jumpstack[cpu] = vmalloc_node(size,
1093 cpu_to_node(cpu));
1094 else
1095 i->jumpstack[cpu] = kmalloc_node(size,
1096 GFP_KERNEL, cpu_to_node(cpu));
1097 if (i->jumpstack[cpu] == NULL)
1098 /*
1099 * Freeing will be done later on by the callers. The
1100 * chain is: xt_replace_table -> __do_replace ->
1101 * do_replace -> xt_free_table_info.
1102 */
1103 return -ENOMEM;
1104 }
1105
1106 return 0;
1107}
1108
1109struct xt_table_info *
1110xt_replace_table(struct xt_table *table,
1111 unsigned int num_counters,
1112 struct xt_table_info *newinfo,
1113 int *error)
1114{
1115 struct xt_table_info *private;
1116 int ret;
1117
1118 ret = xt_jumpstack_alloc(newinfo);
1119 if (ret < 0) {
1120 *error = ret;
1121 return NULL;
1122 }
1123
1124 /* Do the substitution. */
1125 local_bh_disable();
1126 private = table->private;
1127
1128 /* Check inside lock: is the old number correct? */
1129 if (num_counters != private->number) {
1130 pr_debug("num_counters != table->private->number (%u/%u)\n",
1131 num_counters, private->number);
1132 local_bh_enable();
1133 *error = -EAGAIN;
1134 return NULL;
1135 }
1136
1137 newinfo->initial_entries = private->initial_entries;
1138 /*
1139 * Ensure contents of newinfo are visible before assigning to
1140 * private.
1141 */
1142 smp_wmb();
1143 table->private = newinfo;
1144
1145 /*
1146 * Even though table entries have now been swapped, other CPU's
1147 * may still be using the old entries. This is okay, because
1148 * resynchronization happens because of the locking done
1149 * during the get_counters() routine.
1150 */
1151 local_bh_enable();
1152
1153#ifdef CONFIG_AUDIT
1154 if (audit_enabled) {
1155 struct audit_buffer *ab;
1156
1157 ab = audit_log_start(current->audit_context, GFP_KERNEL,
1158 AUDIT_NETFILTER_CFG);
1159 if (ab) {
1160 audit_log_format(ab, "table=%s family=%u entries=%u",
1161 table->name, table->af,
1162 private->number);
1163 audit_log_end(ab);
1164 }
1165 }
1166#endif
1167
1168 return private;
1169}
1170EXPORT_SYMBOL_GPL(xt_replace_table);
1171
1172struct xt_table *xt_register_table(struct net *net,
1173 const struct xt_table *input_table,
1174 struct xt_table_info *bootstrap,
1175 struct xt_table_info *newinfo)
1176{
1177 int ret;
1178 struct xt_table_info *private;
1179 struct xt_table *t, *table;
1180
1181 /* Don't add one object to multiple lists. */
1182 table = kmemdup(input_table, sizeof(struct xt_table), GFP_KERNEL);
1183 if (!table) {
1184 ret = -ENOMEM;
1185 goto out;
1186 }
1187
1188 mutex_lock(&xt[table->af].mutex);
1189 /* Don't autoload: we'd eat our tail... */
1190 list_for_each_entry(t, &net->xt.tables[table->af], list) {
1191 if (strcmp(t->name, table->name) == 0) {
1192 ret = -EEXIST;
1193 goto unlock;
1194 }
1195 }
1196
1197 /* Simplifies replace_table code. */
1198 table->private = bootstrap;
1199
1200 if (!xt_replace_table(table, 0, newinfo, &ret))
1201 goto unlock;
1202
1203 private = table->private;
1204 pr_debug("table->private->number = %u\n", private->number);
1205
1206 /* save number of initial entries */
1207 private->initial_entries = private->number;
1208
1209 list_add(&table->list, &net->xt.tables[table->af]);
1210 mutex_unlock(&xt[table->af].mutex);
1211 return table;
1212
1213unlock:
1214 mutex_unlock(&xt[table->af].mutex);
1215 kfree(table);
1216out:
1217 return ERR_PTR(ret);
1218}
1219EXPORT_SYMBOL_GPL(xt_register_table);
1220
1221void *xt_unregister_table(struct xt_table *table)
1222{
1223 struct xt_table_info *private;
1224
1225 mutex_lock(&xt[table->af].mutex);
1226 private = table->private;
1227 list_del(&table->list);
1228 mutex_unlock(&xt[table->af].mutex);
1229 kfree(table);
1230
1231 return private;
1232}
1233EXPORT_SYMBOL_GPL(xt_unregister_table);
1234
1235#ifdef CONFIG_PROC_FS
1236struct xt_names_priv {
1237 struct seq_net_private p;
1238 u_int8_t af;
1239};
1240static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos)
1241{
1242 struct xt_names_priv *priv = seq->private;
1243 struct net *net = seq_file_net(seq);
1244 u_int8_t af = priv->af;
1245
1246 mutex_lock(&xt[af].mutex);
1247 return seq_list_start(&net->xt.tables[af], *pos);
1248}
1249
1250static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1251{
1252 struct xt_names_priv *priv = seq->private;
1253 struct net *net = seq_file_net(seq);
1254 u_int8_t af = priv->af;
1255
1256 return seq_list_next(v, &net->xt.tables[af], pos);
1257}
1258
1259static void xt_table_seq_stop(struct seq_file *seq, void *v)
1260{
1261 struct xt_names_priv *priv = seq->private;
1262 u_int8_t af = priv->af;
1263
1264 mutex_unlock(&xt[af].mutex);
1265}
1266
1267static int xt_table_seq_show(struct seq_file *seq, void *v)
1268{
1269 struct xt_table *table = list_entry(v, struct xt_table, list);
1270
1271 if (*table->name)
1272 seq_printf(seq, "%s\n", table->name);
1273 return 0;
1274}
1275
1276static const struct seq_operations xt_table_seq_ops = {
1277 .start = xt_table_seq_start,
1278 .next = xt_table_seq_next,
1279 .stop = xt_table_seq_stop,
1280 .show = xt_table_seq_show,
1281};
1282
1283static int xt_table_open(struct inode *inode, struct file *file)
1284{
1285 int ret;
1286 struct xt_names_priv *priv;
1287
1288 ret = seq_open_net(inode, file, &xt_table_seq_ops,
1289 sizeof(struct xt_names_priv));
1290 if (!ret) {
1291 priv = ((struct seq_file *)file->private_data)->private;
1292 priv->af = (unsigned long)PDE_DATA(inode);
1293 }
1294 return ret;
1295}
1296
1297static const struct file_operations xt_table_ops = {
1298 .owner = THIS_MODULE,
1299 .open = xt_table_open,
1300 .read = seq_read,
1301 .llseek = seq_lseek,
1302 .release = seq_release_net,
1303};
1304
1305/*
1306 * Traverse state for ip{,6}_{tables,matches} for helping crossing
1307 * the multi-AF mutexes.
1308 */
1309struct nf_mttg_trav {
1310 struct list_head *head, *curr;
1311 uint8_t class, nfproto;
1312};
1313
1314enum {
1315 MTTG_TRAV_INIT,
1316 MTTG_TRAV_NFP_UNSPEC,
1317 MTTG_TRAV_NFP_SPEC,
1318 MTTG_TRAV_DONE,
1319};
1320
1321static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos,
1322 bool is_target)
1323{
1324 static const uint8_t next_class[] = {
1325 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC,
1326 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE,
1327 };
1328 struct nf_mttg_trav *trav = seq->private;
1329
1330 switch (trav->class) {
1331 case MTTG_TRAV_INIT:
1332 trav->class = MTTG_TRAV_NFP_UNSPEC;
1333 mutex_lock(&xt[NFPROTO_UNSPEC].mutex);
1334 trav->head = trav->curr = is_target ?
1335 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match;
1336 break;
1337 case MTTG_TRAV_NFP_UNSPEC:
1338 trav->curr = trav->curr->next;
1339 if (trav->curr != trav->head)
1340 break;
1341 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1342 mutex_lock(&xt[trav->nfproto].mutex);
1343 trav->head = trav->curr = is_target ?
1344 &xt[trav->nfproto].target : &xt[trav->nfproto].match;
1345 trav->class = next_class[trav->class];
1346 break;
1347 case MTTG_TRAV_NFP_SPEC:
1348 trav->curr = trav->curr->next;
1349 if (trav->curr != trav->head)
1350 break;
1351 /* fallthru, _stop will unlock */
1352 default:
1353 return NULL;
1354 }
1355
1356 if (ppos != NULL)
1357 ++*ppos;
1358 return trav;
1359}
1360
1361static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos,
1362 bool is_target)
1363{
1364 struct nf_mttg_trav *trav = seq->private;
1365 unsigned int j;
1366
1367 trav->class = MTTG_TRAV_INIT;
1368 for (j = 0; j < *pos; ++j)
1369 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL)
1370 return NULL;
1371 return trav;
1372}
1373
1374static void xt_mttg_seq_stop(struct seq_file *seq, void *v)
1375{
1376 struct nf_mttg_trav *trav = seq->private;
1377
1378 switch (trav->class) {
1379 case MTTG_TRAV_NFP_UNSPEC:
1380 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1381 break;
1382 case MTTG_TRAV_NFP_SPEC:
1383 mutex_unlock(&xt[trav->nfproto].mutex);
1384 break;
1385 }
1386}
1387
1388static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos)
1389{
1390 return xt_mttg_seq_start(seq, pos, false);
1391}
1392
1393static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1394{
1395 return xt_mttg_seq_next(seq, v, ppos, false);
1396}
1397
1398static int xt_match_seq_show(struct seq_file *seq, void *v)
1399{
1400 const struct nf_mttg_trav *trav = seq->private;
1401 const struct xt_match *match;
1402
1403 switch (trav->class) {
1404 case MTTG_TRAV_NFP_UNSPEC:
1405 case MTTG_TRAV_NFP_SPEC:
1406 if (trav->curr == trav->head)
1407 return 0;
1408 match = list_entry(trav->curr, struct xt_match, list);
1409 if (*match->name)
1410 seq_printf(seq, "%s\n", match->name);
1411 }
1412 return 0;
1413}
1414
1415static const struct seq_operations xt_match_seq_ops = {
1416 .start = xt_match_seq_start,
1417 .next = xt_match_seq_next,
1418 .stop = xt_mttg_seq_stop,
1419 .show = xt_match_seq_show,
1420};
1421
1422static int xt_match_open(struct inode *inode, struct file *file)
1423{
1424 struct nf_mttg_trav *trav;
1425 trav = __seq_open_private(file, &xt_match_seq_ops, sizeof(*trav));
1426 if (!trav)
1427 return -ENOMEM;
1428
1429 trav->nfproto = (unsigned long)PDE_DATA(inode);
1430 return 0;
1431}
1432
1433static const struct file_operations xt_match_ops = {
1434 .owner = THIS_MODULE,
1435 .open = xt_match_open,
1436 .read = seq_read,
1437 .llseek = seq_lseek,
1438 .release = seq_release_private,
1439};
1440
1441static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos)
1442{
1443 return xt_mttg_seq_start(seq, pos, true);
1444}
1445
1446static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1447{
1448 return xt_mttg_seq_next(seq, v, ppos, true);
1449}
1450
1451static int xt_target_seq_show(struct seq_file *seq, void *v)
1452{
1453 const struct nf_mttg_trav *trav = seq->private;
1454 const struct xt_target *target;
1455
1456 switch (trav->class) {
1457 case MTTG_TRAV_NFP_UNSPEC:
1458 case MTTG_TRAV_NFP_SPEC:
1459 if (trav->curr == trav->head)
1460 return 0;
1461 target = list_entry(trav->curr, struct xt_target, list);
1462 if (*target->name)
1463 seq_printf(seq, "%s\n", target->name);
1464 }
1465 return 0;
1466}
1467
1468static const struct seq_operations xt_target_seq_ops = {
1469 .start = xt_target_seq_start,
1470 .next = xt_target_seq_next,
1471 .stop = xt_mttg_seq_stop,
1472 .show = xt_target_seq_show,
1473};
1474
1475static int xt_target_open(struct inode *inode, struct file *file)
1476{
1477 struct nf_mttg_trav *trav;
1478 trav = __seq_open_private(file, &xt_target_seq_ops, sizeof(*trav));
1479 if (!trav)
1480 return -ENOMEM;
1481
1482 trav->nfproto = (unsigned long)PDE_DATA(inode);
1483 return 0;
1484}
1485
1486static const struct file_operations xt_target_ops = {
1487 .owner = THIS_MODULE,
1488 .open = xt_target_open,
1489 .read = seq_read,
1490 .llseek = seq_lseek,
1491 .release = seq_release_private,
1492};
1493
1494#define FORMAT_TABLES "_tables_names"
1495#define FORMAT_MATCHES "_tables_matches"
1496#define FORMAT_TARGETS "_tables_targets"
1497
1498#endif /* CONFIG_PROC_FS */
1499
1500/**
1501 * xt_hook_ops_alloc - set up hooks for a new table
1502 * @table: table with metadata needed to set up hooks
1503 * @fn: Hook function
1504 *
1505 * This function will create the nf_hook_ops that the x_table needs
1506 * to hand to xt_hook_link_net().
1507 */
1508struct nf_hook_ops *
1509xt_hook_ops_alloc(const struct xt_table *table, nf_hookfn *fn)
1510{
1511 unsigned int hook_mask = table->valid_hooks;
1512 uint8_t i, num_hooks = hweight32(hook_mask);
1513 uint8_t hooknum;
1514 struct nf_hook_ops *ops;
1515
1516 if (!num_hooks)
1517 return ERR_PTR(-EINVAL);
1518
1519 ops = kcalloc(num_hooks, sizeof(*ops), GFP_KERNEL);
1520 if (ops == NULL)
1521 return ERR_PTR(-ENOMEM);
1522
1523 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0;
1524 hook_mask >>= 1, ++hooknum) {
1525 if (!(hook_mask & 1))
1526 continue;
1527 ops[i].hook = fn;
1528 ops[i].pf = table->af;
1529 ops[i].hooknum = hooknum;
1530 ops[i].priority = table->priority;
1531 ++i;
1532 }
1533
1534 return ops;
1535}
1536EXPORT_SYMBOL_GPL(xt_hook_ops_alloc);
1537
1538int xt_proto_init(struct net *net, u_int8_t af)
1539{
1540#ifdef CONFIG_PROC_FS
1541 char buf[XT_FUNCTION_MAXNAMELEN];
1542 struct proc_dir_entry *proc;
1543 kuid_t root_uid;
1544 kgid_t root_gid;
1545#endif
1546
1547 if (af >= ARRAY_SIZE(xt_prefix))
1548 return -EINVAL;
1549
1550
1551#ifdef CONFIG_PROC_FS
1552 root_uid = make_kuid(net->user_ns, 0);
1553 root_gid = make_kgid(net->user_ns, 0);
1554
1555 strlcpy(buf, xt_prefix[af], sizeof(buf));
1556 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1557 proc = proc_create_data(buf, 0440, net->proc_net, &xt_table_ops,
1558 (void *)(unsigned long)af);
1559 if (!proc)
1560 goto out;
1561 if (uid_valid(root_uid) && gid_valid(root_gid))
1562 proc_set_user(proc, root_uid, root_gid);
1563
1564 strlcpy(buf, xt_prefix[af], sizeof(buf));
1565 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1566 proc = proc_create_data(buf, 0440, net->proc_net, &xt_match_ops,
1567 (void *)(unsigned long)af);
1568 if (!proc)
1569 goto out_remove_tables;
1570 if (uid_valid(root_uid) && gid_valid(root_gid))
1571 proc_set_user(proc, root_uid, root_gid);
1572
1573 strlcpy(buf, xt_prefix[af], sizeof(buf));
1574 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1575 proc = proc_create_data(buf, 0440, net->proc_net, &xt_target_ops,
1576 (void *)(unsigned long)af);
1577 if (!proc)
1578 goto out_remove_matches;
1579 if (uid_valid(root_uid) && gid_valid(root_gid))
1580 proc_set_user(proc, root_uid, root_gid);
1581#endif
1582
1583 return 0;
1584
1585#ifdef CONFIG_PROC_FS
1586out_remove_matches:
1587 strlcpy(buf, xt_prefix[af], sizeof(buf));
1588 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1589 remove_proc_entry(buf, net->proc_net);
1590
1591out_remove_tables:
1592 strlcpy(buf, xt_prefix[af], sizeof(buf));
1593 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1594 remove_proc_entry(buf, net->proc_net);
1595out:
1596 return -1;
1597#endif
1598}
1599EXPORT_SYMBOL_GPL(xt_proto_init);
1600
1601void xt_proto_fini(struct net *net, u_int8_t af)
1602{
1603#ifdef CONFIG_PROC_FS
1604 char buf[XT_FUNCTION_MAXNAMELEN];
1605
1606 strlcpy(buf, xt_prefix[af], sizeof(buf));
1607 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1608 remove_proc_entry(buf, net->proc_net);
1609
1610 strlcpy(buf, xt_prefix[af], sizeof(buf));
1611 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1612 remove_proc_entry(buf, net->proc_net);
1613
1614 strlcpy(buf, xt_prefix[af], sizeof(buf));
1615 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1616 remove_proc_entry(buf, net->proc_net);
1617#endif /*CONFIG_PROC_FS*/
1618}
1619EXPORT_SYMBOL_GPL(xt_proto_fini);
1620
1621/**
1622 * xt_percpu_counter_alloc - allocate x_tables rule counter
1623 *
1624 * @state: pointer to xt_percpu allocation state
1625 * @counter: pointer to counter struct inside the ip(6)/arpt_entry struct
1626 *
1627 * On SMP, the packet counter [ ip(6)t_entry->counters.pcnt ] will then
1628 * contain the address of the real (percpu) counter.
1629 *
1630 * Rule evaluation needs to use xt_get_this_cpu_counter() helper
1631 * to fetch the real percpu counter.
1632 *
1633 * To speed up allocation and improve data locality, a 4kb block is
1634 * allocated.
1635 *
1636 * xt_percpu_counter_alloc_state contains the base address of the
1637 * allocated page and the current sub-offset.
1638 *
1639 * returns false on error.
1640 */
1641bool xt_percpu_counter_alloc(struct xt_percpu_counter_alloc_state *state,
1642 struct xt_counters *counter)
1643{
1644 BUILD_BUG_ON(XT_PCPU_BLOCK_SIZE < (sizeof(*counter) * 2));
1645
1646 if (nr_cpu_ids <= 1)
1647 return true;
1648
1649 if (!state->mem) {
1650 state->mem = __alloc_percpu(XT_PCPU_BLOCK_SIZE,
1651 XT_PCPU_BLOCK_SIZE);
1652 if (!state->mem)
1653 return false;
1654 }
1655 counter->pcnt = (__force unsigned long)(state->mem + state->off);
1656 state->off += sizeof(*counter);
1657 if (state->off > (XT_PCPU_BLOCK_SIZE - sizeof(*counter))) {
1658 state->mem = NULL;
1659 state->off = 0;
1660 }
1661 return true;
1662}
1663EXPORT_SYMBOL_GPL(xt_percpu_counter_alloc);
1664
1665void xt_percpu_counter_free(struct xt_counters *counters)
1666{
1667 unsigned long pcnt = counters->pcnt;
1668
1669 if (nr_cpu_ids > 1 && (pcnt & (XT_PCPU_BLOCK_SIZE - 1)) == 0)
1670 free_percpu((void __percpu *)pcnt);
1671}
1672EXPORT_SYMBOL_GPL(xt_percpu_counter_free);
1673
1674static int __net_init xt_net_init(struct net *net)
1675{
1676 int i;
1677
1678 for (i = 0; i < NFPROTO_NUMPROTO; i++)
1679 INIT_LIST_HEAD(&net->xt.tables[i]);
1680 return 0;
1681}
1682
1683static struct pernet_operations xt_net_ops = {
1684 .init = xt_net_init,
1685};
1686
1687static int __init xt_init(void)
1688{
1689 unsigned int i;
1690 int rv;
1691
1692 for_each_possible_cpu(i) {
1693 seqcount_init(&per_cpu(xt_recseq, i));
1694 }
1695
1696 xt = kmalloc(sizeof(struct xt_af) * NFPROTO_NUMPROTO, GFP_KERNEL);
1697 if (!xt)
1698 return -ENOMEM;
1699
1700 for (i = 0; i < NFPROTO_NUMPROTO; i++) {
1701 mutex_init(&xt[i].mutex);
1702#ifdef CONFIG_COMPAT
1703 mutex_init(&xt[i].compat_mutex);
1704 xt[i].compat_tab = NULL;
1705#endif
1706 INIT_LIST_HEAD(&xt[i].target);
1707 INIT_LIST_HEAD(&xt[i].match);
1708 }
1709 rv = register_pernet_subsys(&xt_net_ops);
1710 if (rv < 0)
1711 kfree(xt);
1712 return rv;
1713}
1714
1715static void __exit xt_fini(void)
1716{
1717 unregister_pernet_subsys(&xt_net_ops);
1718 kfree(xt);
1719}
1720
1721module_init(xt_init);
1722module_exit(xt_fini);
1723
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * x_tables core - Backend for {ip,ip6,arp}_tables
4 *
5 * Copyright (C) 2006-2006 Harald Welte <laforge@netfilter.org>
6 * Copyright (C) 2006-2012 Patrick McHardy <kaber@trash.net>
7 *
8 * Based on existing ip_tables code which is
9 * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
10 * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
11 */
12#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13#include <linux/kernel.h>
14#include <linux/module.h>
15#include <linux/socket.h>
16#include <linux/net.h>
17#include <linux/proc_fs.h>
18#include <linux/seq_file.h>
19#include <linux/string.h>
20#include <linux/vmalloc.h>
21#include <linux/mutex.h>
22#include <linux/mm.h>
23#include <linux/slab.h>
24#include <linux/audit.h>
25#include <linux/user_namespace.h>
26#include <net/net_namespace.h>
27#include <net/netns/generic.h>
28
29#include <linux/netfilter/x_tables.h>
30#include <linux/netfilter_arp.h>
31#include <linux/netfilter_ipv4/ip_tables.h>
32#include <linux/netfilter_ipv6/ip6_tables.h>
33#include <linux/netfilter_arp/arp_tables.h>
34
35MODULE_LICENSE("GPL");
36MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
37MODULE_DESCRIPTION("{ip,ip6,arp,eb}_tables backend module");
38
39#define XT_PCPU_BLOCK_SIZE 4096
40#define XT_MAX_TABLE_SIZE (512 * 1024 * 1024)
41
42struct xt_template {
43 struct list_head list;
44
45 /* called when table is needed in the given netns */
46 int (*table_init)(struct net *net);
47
48 struct module *me;
49
50 /* A unique name... */
51 char name[XT_TABLE_MAXNAMELEN];
52};
53
54static struct list_head xt_templates[NFPROTO_NUMPROTO];
55
56struct xt_pernet {
57 struct list_head tables[NFPROTO_NUMPROTO];
58};
59
60struct compat_delta {
61 unsigned int offset; /* offset in kernel */
62 int delta; /* delta in 32bit user land */
63};
64
65struct xt_af {
66 struct mutex mutex;
67 struct list_head match;
68 struct list_head target;
69#ifdef CONFIG_NETFILTER_XTABLES_COMPAT
70 struct mutex compat_mutex;
71 struct compat_delta *compat_tab;
72 unsigned int number; /* number of slots in compat_tab[] */
73 unsigned int cur; /* number of used slots in compat_tab[] */
74#endif
75};
76
77static unsigned int xt_pernet_id __read_mostly;
78static struct xt_af *xt __read_mostly;
79
80static const char *const xt_prefix[NFPROTO_NUMPROTO] = {
81 [NFPROTO_UNSPEC] = "x",
82 [NFPROTO_IPV4] = "ip",
83 [NFPROTO_ARP] = "arp",
84 [NFPROTO_BRIDGE] = "eb",
85 [NFPROTO_IPV6] = "ip6",
86};
87
88/* Registration hooks for targets. */
89int xt_register_target(struct xt_target *target)
90{
91 u_int8_t af = target->family;
92
93 mutex_lock(&xt[af].mutex);
94 list_add(&target->list, &xt[af].target);
95 mutex_unlock(&xt[af].mutex);
96 return 0;
97}
98EXPORT_SYMBOL(xt_register_target);
99
100void
101xt_unregister_target(struct xt_target *target)
102{
103 u_int8_t af = target->family;
104
105 mutex_lock(&xt[af].mutex);
106 list_del(&target->list);
107 mutex_unlock(&xt[af].mutex);
108}
109EXPORT_SYMBOL(xt_unregister_target);
110
111int
112xt_register_targets(struct xt_target *target, unsigned int n)
113{
114 unsigned int i;
115 int err = 0;
116
117 for (i = 0; i < n; i++) {
118 err = xt_register_target(&target[i]);
119 if (err)
120 goto err;
121 }
122 return err;
123
124err:
125 if (i > 0)
126 xt_unregister_targets(target, i);
127 return err;
128}
129EXPORT_SYMBOL(xt_register_targets);
130
131void
132xt_unregister_targets(struct xt_target *target, unsigned int n)
133{
134 while (n-- > 0)
135 xt_unregister_target(&target[n]);
136}
137EXPORT_SYMBOL(xt_unregister_targets);
138
139int xt_register_match(struct xt_match *match)
140{
141 u_int8_t af = match->family;
142
143 mutex_lock(&xt[af].mutex);
144 list_add(&match->list, &xt[af].match);
145 mutex_unlock(&xt[af].mutex);
146 return 0;
147}
148EXPORT_SYMBOL(xt_register_match);
149
150void
151xt_unregister_match(struct xt_match *match)
152{
153 u_int8_t af = match->family;
154
155 mutex_lock(&xt[af].mutex);
156 list_del(&match->list);
157 mutex_unlock(&xt[af].mutex);
158}
159EXPORT_SYMBOL(xt_unregister_match);
160
161int
162xt_register_matches(struct xt_match *match, unsigned int n)
163{
164 unsigned int i;
165 int err = 0;
166
167 for (i = 0; i < n; i++) {
168 err = xt_register_match(&match[i]);
169 if (err)
170 goto err;
171 }
172 return err;
173
174err:
175 if (i > 0)
176 xt_unregister_matches(match, i);
177 return err;
178}
179EXPORT_SYMBOL(xt_register_matches);
180
181void
182xt_unregister_matches(struct xt_match *match, unsigned int n)
183{
184 while (n-- > 0)
185 xt_unregister_match(&match[n]);
186}
187EXPORT_SYMBOL(xt_unregister_matches);
188
189
190/*
191 * These are weird, but module loading must not be done with mutex
192 * held (since they will register), and we have to have a single
193 * function to use.
194 */
195
196/* Find match, grabs ref. Returns ERR_PTR() on error. */
197struct xt_match *xt_find_match(u8 af, const char *name, u8 revision)
198{
199 struct xt_match *m;
200 int err = -ENOENT;
201
202 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN)
203 return ERR_PTR(-EINVAL);
204
205 mutex_lock(&xt[af].mutex);
206 list_for_each_entry(m, &xt[af].match, list) {
207 if (strcmp(m->name, name) == 0) {
208 if (m->revision == revision) {
209 if (try_module_get(m->me)) {
210 mutex_unlock(&xt[af].mutex);
211 return m;
212 }
213 } else
214 err = -EPROTOTYPE; /* Found something. */
215 }
216 }
217 mutex_unlock(&xt[af].mutex);
218
219 if (af != NFPROTO_UNSPEC)
220 /* Try searching again in the family-independent list */
221 return xt_find_match(NFPROTO_UNSPEC, name, revision);
222
223 return ERR_PTR(err);
224}
225EXPORT_SYMBOL(xt_find_match);
226
227struct xt_match *
228xt_request_find_match(uint8_t nfproto, const char *name, uint8_t revision)
229{
230 struct xt_match *match;
231
232 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN)
233 return ERR_PTR(-EINVAL);
234
235 match = xt_find_match(nfproto, name, revision);
236 if (IS_ERR(match)) {
237 request_module("%st_%s", xt_prefix[nfproto], name);
238 match = xt_find_match(nfproto, name, revision);
239 }
240
241 return match;
242}
243EXPORT_SYMBOL_GPL(xt_request_find_match);
244
245/* Find target, grabs ref. Returns ERR_PTR() on error. */
246static struct xt_target *xt_find_target(u8 af, const char *name, u8 revision)
247{
248 struct xt_target *t;
249 int err = -ENOENT;
250
251 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN)
252 return ERR_PTR(-EINVAL);
253
254 mutex_lock(&xt[af].mutex);
255 list_for_each_entry(t, &xt[af].target, list) {
256 if (strcmp(t->name, name) == 0) {
257 if (t->revision == revision) {
258 if (try_module_get(t->me)) {
259 mutex_unlock(&xt[af].mutex);
260 return t;
261 }
262 } else
263 err = -EPROTOTYPE; /* Found something. */
264 }
265 }
266 mutex_unlock(&xt[af].mutex);
267
268 if (af != NFPROTO_UNSPEC)
269 /* Try searching again in the family-independent list */
270 return xt_find_target(NFPROTO_UNSPEC, name, revision);
271
272 return ERR_PTR(err);
273}
274
275struct xt_target *xt_request_find_target(u8 af, const char *name, u8 revision)
276{
277 struct xt_target *target;
278
279 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN)
280 return ERR_PTR(-EINVAL);
281
282 target = xt_find_target(af, name, revision);
283 if (IS_ERR(target)) {
284 request_module("%st_%s", xt_prefix[af], name);
285 target = xt_find_target(af, name, revision);
286 }
287
288 return target;
289}
290EXPORT_SYMBOL_GPL(xt_request_find_target);
291
292
293static int xt_obj_to_user(u16 __user *psize, u16 size,
294 void __user *pname, const char *name,
295 u8 __user *prev, u8 rev)
296{
297 if (put_user(size, psize))
298 return -EFAULT;
299 if (copy_to_user(pname, name, strlen(name) + 1))
300 return -EFAULT;
301 if (put_user(rev, prev))
302 return -EFAULT;
303
304 return 0;
305}
306
307#define XT_OBJ_TO_USER(U, K, TYPE, C_SIZE) \
308 xt_obj_to_user(&U->u.TYPE##_size, C_SIZE ? : K->u.TYPE##_size, \
309 U->u.user.name, K->u.kernel.TYPE->name, \
310 &U->u.user.revision, K->u.kernel.TYPE->revision)
311
312int xt_data_to_user(void __user *dst, const void *src,
313 int usersize, int size, int aligned_size)
314{
315 usersize = usersize ? : size;
316 if (copy_to_user(dst, src, usersize))
317 return -EFAULT;
318 if (usersize != aligned_size &&
319 clear_user(dst + usersize, aligned_size - usersize))
320 return -EFAULT;
321
322 return 0;
323}
324EXPORT_SYMBOL_GPL(xt_data_to_user);
325
326#define XT_DATA_TO_USER(U, K, TYPE) \
327 xt_data_to_user(U->data, K->data, \
328 K->u.kernel.TYPE->usersize, \
329 K->u.kernel.TYPE->TYPE##size, \
330 XT_ALIGN(K->u.kernel.TYPE->TYPE##size))
331
332int xt_match_to_user(const struct xt_entry_match *m,
333 struct xt_entry_match __user *u)
334{
335 return XT_OBJ_TO_USER(u, m, match, 0) ||
336 XT_DATA_TO_USER(u, m, match);
337}
338EXPORT_SYMBOL_GPL(xt_match_to_user);
339
340int xt_target_to_user(const struct xt_entry_target *t,
341 struct xt_entry_target __user *u)
342{
343 return XT_OBJ_TO_USER(u, t, target, 0) ||
344 XT_DATA_TO_USER(u, t, target);
345}
346EXPORT_SYMBOL_GPL(xt_target_to_user);
347
348static int match_revfn(u8 af, const char *name, u8 revision, int *bestp)
349{
350 const struct xt_match *m;
351 int have_rev = 0;
352
353 mutex_lock(&xt[af].mutex);
354 list_for_each_entry(m, &xt[af].match, list) {
355 if (strcmp(m->name, name) == 0) {
356 if (m->revision > *bestp)
357 *bestp = m->revision;
358 if (m->revision == revision)
359 have_rev = 1;
360 }
361 }
362 mutex_unlock(&xt[af].mutex);
363
364 if (af != NFPROTO_UNSPEC && !have_rev)
365 return match_revfn(NFPROTO_UNSPEC, name, revision, bestp);
366
367 return have_rev;
368}
369
370static int target_revfn(u8 af, const char *name, u8 revision, int *bestp)
371{
372 const struct xt_target *t;
373 int have_rev = 0;
374
375 mutex_lock(&xt[af].mutex);
376 list_for_each_entry(t, &xt[af].target, list) {
377 if (strcmp(t->name, name) == 0) {
378 if (t->revision > *bestp)
379 *bestp = t->revision;
380 if (t->revision == revision)
381 have_rev = 1;
382 }
383 }
384 mutex_unlock(&xt[af].mutex);
385
386 if (af != NFPROTO_UNSPEC && !have_rev)
387 return target_revfn(NFPROTO_UNSPEC, name, revision, bestp);
388
389 return have_rev;
390}
391
392/* Returns true or false (if no such extension at all) */
393int xt_find_revision(u8 af, const char *name, u8 revision, int target,
394 int *err)
395{
396 int have_rev, best = -1;
397
398 if (target == 1)
399 have_rev = target_revfn(af, name, revision, &best);
400 else
401 have_rev = match_revfn(af, name, revision, &best);
402
403 /* Nothing at all? Return 0 to try loading module. */
404 if (best == -1) {
405 *err = -ENOENT;
406 return 0;
407 }
408
409 *err = best;
410 if (!have_rev)
411 *err = -EPROTONOSUPPORT;
412 return 1;
413}
414EXPORT_SYMBOL_GPL(xt_find_revision);
415
416static char *
417textify_hooks(char *buf, size_t size, unsigned int mask, uint8_t nfproto)
418{
419 static const char *const inetbr_names[] = {
420 "PREROUTING", "INPUT", "FORWARD",
421 "OUTPUT", "POSTROUTING", "BROUTING",
422 };
423 static const char *const arp_names[] = {
424 "INPUT", "FORWARD", "OUTPUT",
425 };
426 const char *const *names;
427 unsigned int i, max;
428 char *p = buf;
429 bool np = false;
430 int res;
431
432 names = (nfproto == NFPROTO_ARP) ? arp_names : inetbr_names;
433 max = (nfproto == NFPROTO_ARP) ? ARRAY_SIZE(arp_names) :
434 ARRAY_SIZE(inetbr_names);
435 *p = '\0';
436 for (i = 0; i < max; ++i) {
437 if (!(mask & (1 << i)))
438 continue;
439 res = snprintf(p, size, "%s%s", np ? "/" : "", names[i]);
440 if (res > 0) {
441 size -= res;
442 p += res;
443 }
444 np = true;
445 }
446
447 return buf;
448}
449
450/**
451 * xt_check_proc_name - check that name is suitable for /proc file creation
452 *
453 * @name: file name candidate
454 * @size: length of buffer
455 *
456 * some x_tables modules wish to create a file in /proc.
457 * This function makes sure that the name is suitable for this
458 * purpose, it checks that name is NUL terminated and isn't a 'special'
459 * name, like "..".
460 *
461 * returns negative number on error or 0 if name is useable.
462 */
463int xt_check_proc_name(const char *name, unsigned int size)
464{
465 if (name[0] == '\0')
466 return -EINVAL;
467
468 if (strnlen(name, size) == size)
469 return -ENAMETOOLONG;
470
471 if (strcmp(name, ".") == 0 ||
472 strcmp(name, "..") == 0 ||
473 strchr(name, '/'))
474 return -EINVAL;
475
476 return 0;
477}
478EXPORT_SYMBOL(xt_check_proc_name);
479
480int xt_check_match(struct xt_mtchk_param *par,
481 unsigned int size, u16 proto, bool inv_proto)
482{
483 int ret;
484
485 if (XT_ALIGN(par->match->matchsize) != size &&
486 par->match->matchsize != -1) {
487 /*
488 * ebt_among is exempt from centralized matchsize checking
489 * because it uses a dynamic-size data set.
490 */
491 pr_err_ratelimited("%s_tables: %s.%u match: invalid size %u (kernel) != (user) %u\n",
492 xt_prefix[par->family], par->match->name,
493 par->match->revision,
494 XT_ALIGN(par->match->matchsize), size);
495 return -EINVAL;
496 }
497 if (par->match->table != NULL &&
498 strcmp(par->match->table, par->table) != 0) {
499 pr_info_ratelimited("%s_tables: %s match: only valid in %s table, not %s\n",
500 xt_prefix[par->family], par->match->name,
501 par->match->table, par->table);
502 return -EINVAL;
503 }
504 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) {
505 char used[64], allow[64];
506
507 pr_info_ratelimited("%s_tables: %s match: used from hooks %s, but only valid from %s\n",
508 xt_prefix[par->family], par->match->name,
509 textify_hooks(used, sizeof(used),
510 par->hook_mask, par->family),
511 textify_hooks(allow, sizeof(allow),
512 par->match->hooks,
513 par->family));
514 return -EINVAL;
515 }
516 if (par->match->proto && (par->match->proto != proto || inv_proto)) {
517 pr_info_ratelimited("%s_tables: %s match: only valid for protocol %u\n",
518 xt_prefix[par->family], par->match->name,
519 par->match->proto);
520 return -EINVAL;
521 }
522 if (par->match->checkentry != NULL) {
523 ret = par->match->checkentry(par);
524 if (ret < 0)
525 return ret;
526 else if (ret > 0)
527 /* Flag up potential errors. */
528 return -EIO;
529 }
530 return 0;
531}
532EXPORT_SYMBOL_GPL(xt_check_match);
533
534/** xt_check_entry_match - check that matches end before start of target
535 *
536 * @match: beginning of xt_entry_match
537 * @target: beginning of this rules target (alleged end of matches)
538 * @alignment: alignment requirement of match structures
539 *
540 * Validates that all matches add up to the beginning of the target,
541 * and that each match covers at least the base structure size.
542 *
543 * Return: 0 on success, negative errno on failure.
544 */
545static int xt_check_entry_match(const char *match, const char *target,
546 const size_t alignment)
547{
548 const struct xt_entry_match *pos;
549 int length = target - match;
550
551 if (length == 0) /* no matches */
552 return 0;
553
554 pos = (struct xt_entry_match *)match;
555 do {
556 if ((unsigned long)pos % alignment)
557 return -EINVAL;
558
559 if (length < (int)sizeof(struct xt_entry_match))
560 return -EINVAL;
561
562 if (pos->u.match_size < sizeof(struct xt_entry_match))
563 return -EINVAL;
564
565 if (pos->u.match_size > length)
566 return -EINVAL;
567
568 length -= pos->u.match_size;
569 pos = ((void *)((char *)(pos) + (pos)->u.match_size));
570 } while (length > 0);
571
572 return 0;
573}
574
575/** xt_check_table_hooks - check hook entry points are sane
576 *
577 * @info xt_table_info to check
578 * @valid_hooks - hook entry points that we can enter from
579 *
580 * Validates that the hook entry and underflows points are set up.
581 *
582 * Return: 0 on success, negative errno on failure.
583 */
584int xt_check_table_hooks(const struct xt_table_info *info, unsigned int valid_hooks)
585{
586 const char *err = "unsorted underflow";
587 unsigned int i, max_uflow, max_entry;
588 bool check_hooks = false;
589
590 BUILD_BUG_ON(ARRAY_SIZE(info->hook_entry) != ARRAY_SIZE(info->underflow));
591
592 max_entry = 0;
593 max_uflow = 0;
594
595 for (i = 0; i < ARRAY_SIZE(info->hook_entry); i++) {
596 if (!(valid_hooks & (1 << i)))
597 continue;
598
599 if (info->hook_entry[i] == 0xFFFFFFFF)
600 return -EINVAL;
601 if (info->underflow[i] == 0xFFFFFFFF)
602 return -EINVAL;
603
604 if (check_hooks) {
605 if (max_uflow > info->underflow[i])
606 goto error;
607
608 if (max_uflow == info->underflow[i]) {
609 err = "duplicate underflow";
610 goto error;
611 }
612 if (max_entry > info->hook_entry[i]) {
613 err = "unsorted entry";
614 goto error;
615 }
616 if (max_entry == info->hook_entry[i]) {
617 err = "duplicate entry";
618 goto error;
619 }
620 }
621 max_entry = info->hook_entry[i];
622 max_uflow = info->underflow[i];
623 check_hooks = true;
624 }
625
626 return 0;
627error:
628 pr_err_ratelimited("%s at hook %d\n", err, i);
629 return -EINVAL;
630}
631EXPORT_SYMBOL(xt_check_table_hooks);
632
633static bool verdict_ok(int verdict)
634{
635 if (verdict > 0)
636 return true;
637
638 if (verdict < 0) {
639 int v = -verdict - 1;
640
641 if (verdict == XT_RETURN)
642 return true;
643
644 switch (v) {
645 case NF_ACCEPT: return true;
646 case NF_DROP: return true;
647 case NF_QUEUE: return true;
648 default:
649 break;
650 }
651
652 return false;
653 }
654
655 return false;
656}
657
658static bool error_tg_ok(unsigned int usersize, unsigned int kernsize,
659 const char *msg, unsigned int msglen)
660{
661 return usersize == kernsize && strnlen(msg, msglen) < msglen;
662}
663
664#ifdef CONFIG_NETFILTER_XTABLES_COMPAT
665int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta)
666{
667 struct xt_af *xp = &xt[af];
668
669 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex));
670
671 if (WARN_ON(!xp->compat_tab))
672 return -ENOMEM;
673
674 if (xp->cur >= xp->number)
675 return -EINVAL;
676
677 if (xp->cur)
678 delta += xp->compat_tab[xp->cur - 1].delta;
679 xp->compat_tab[xp->cur].offset = offset;
680 xp->compat_tab[xp->cur].delta = delta;
681 xp->cur++;
682 return 0;
683}
684EXPORT_SYMBOL_GPL(xt_compat_add_offset);
685
686void xt_compat_flush_offsets(u_int8_t af)
687{
688 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex));
689
690 if (xt[af].compat_tab) {
691 vfree(xt[af].compat_tab);
692 xt[af].compat_tab = NULL;
693 xt[af].number = 0;
694 xt[af].cur = 0;
695 }
696}
697EXPORT_SYMBOL_GPL(xt_compat_flush_offsets);
698
699int xt_compat_calc_jump(u_int8_t af, unsigned int offset)
700{
701 struct compat_delta *tmp = xt[af].compat_tab;
702 int mid, left = 0, right = xt[af].cur - 1;
703
704 while (left <= right) {
705 mid = (left + right) >> 1;
706 if (offset > tmp[mid].offset)
707 left = mid + 1;
708 else if (offset < tmp[mid].offset)
709 right = mid - 1;
710 else
711 return mid ? tmp[mid - 1].delta : 0;
712 }
713 return left ? tmp[left - 1].delta : 0;
714}
715EXPORT_SYMBOL_GPL(xt_compat_calc_jump);
716
717int xt_compat_init_offsets(u8 af, unsigned int number)
718{
719 size_t mem;
720
721 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex));
722
723 if (!number || number > (INT_MAX / sizeof(struct compat_delta)))
724 return -EINVAL;
725
726 if (WARN_ON(xt[af].compat_tab))
727 return -EINVAL;
728
729 mem = sizeof(struct compat_delta) * number;
730 if (mem > XT_MAX_TABLE_SIZE)
731 return -ENOMEM;
732
733 xt[af].compat_tab = vmalloc(mem);
734 if (!xt[af].compat_tab)
735 return -ENOMEM;
736
737 xt[af].number = number;
738 xt[af].cur = 0;
739
740 return 0;
741}
742EXPORT_SYMBOL(xt_compat_init_offsets);
743
744int xt_compat_match_offset(const struct xt_match *match)
745{
746 u_int16_t csize = match->compatsize ? : match->matchsize;
747 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize);
748}
749EXPORT_SYMBOL_GPL(xt_compat_match_offset);
750
751void xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr,
752 unsigned int *size)
753{
754 const struct xt_match *match = m->u.kernel.match;
755 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m;
756 int off = xt_compat_match_offset(match);
757 u_int16_t msize = cm->u.user.match_size;
758 char name[sizeof(m->u.user.name)];
759
760 m = *dstptr;
761 memcpy(m, cm, sizeof(*cm));
762 if (match->compat_from_user)
763 match->compat_from_user(m->data, cm->data);
764 else
765 memcpy(m->data, cm->data, msize - sizeof(*cm));
766
767 msize += off;
768 m->u.user.match_size = msize;
769 strscpy(name, match->name, sizeof(name));
770 module_put(match->me);
771 strscpy_pad(m->u.user.name, name, sizeof(m->u.user.name));
772
773 *size += off;
774 *dstptr += msize;
775}
776EXPORT_SYMBOL_GPL(xt_compat_match_from_user);
777
778#define COMPAT_XT_DATA_TO_USER(U, K, TYPE, C_SIZE) \
779 xt_data_to_user(U->data, K->data, \
780 K->u.kernel.TYPE->usersize, \
781 C_SIZE, \
782 COMPAT_XT_ALIGN(C_SIZE))
783
784int xt_compat_match_to_user(const struct xt_entry_match *m,
785 void __user **dstptr, unsigned int *size)
786{
787 const struct xt_match *match = m->u.kernel.match;
788 struct compat_xt_entry_match __user *cm = *dstptr;
789 int off = xt_compat_match_offset(match);
790 u_int16_t msize = m->u.user.match_size - off;
791
792 if (XT_OBJ_TO_USER(cm, m, match, msize))
793 return -EFAULT;
794
795 if (match->compat_to_user) {
796 if (match->compat_to_user((void __user *)cm->data, m->data))
797 return -EFAULT;
798 } else {
799 if (COMPAT_XT_DATA_TO_USER(cm, m, match, msize - sizeof(*cm)))
800 return -EFAULT;
801 }
802
803 *size -= off;
804 *dstptr += msize;
805 return 0;
806}
807EXPORT_SYMBOL_GPL(xt_compat_match_to_user);
808
809/* non-compat version may have padding after verdict */
810struct compat_xt_standard_target {
811 struct compat_xt_entry_target t;
812 compat_uint_t verdict;
813};
814
815struct compat_xt_error_target {
816 struct compat_xt_entry_target t;
817 char errorname[XT_FUNCTION_MAXNAMELEN];
818};
819
820int xt_compat_check_entry_offsets(const void *base, const char *elems,
821 unsigned int target_offset,
822 unsigned int next_offset)
823{
824 long size_of_base_struct = elems - (const char *)base;
825 const struct compat_xt_entry_target *t;
826 const char *e = base;
827
828 if (target_offset < size_of_base_struct)
829 return -EINVAL;
830
831 if (target_offset + sizeof(*t) > next_offset)
832 return -EINVAL;
833
834 t = (void *)(e + target_offset);
835 if (t->u.target_size < sizeof(*t))
836 return -EINVAL;
837
838 if (target_offset + t->u.target_size > next_offset)
839 return -EINVAL;
840
841 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) {
842 const struct compat_xt_standard_target *st = (const void *)t;
843
844 if (COMPAT_XT_ALIGN(target_offset + sizeof(*st)) != next_offset)
845 return -EINVAL;
846
847 if (!verdict_ok(st->verdict))
848 return -EINVAL;
849 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) {
850 const struct compat_xt_error_target *et = (const void *)t;
851
852 if (!error_tg_ok(t->u.target_size, sizeof(*et),
853 et->errorname, sizeof(et->errorname)))
854 return -EINVAL;
855 }
856
857 /* compat_xt_entry match has less strict alignment requirements,
858 * otherwise they are identical. In case of padding differences
859 * we need to add compat version of xt_check_entry_match.
860 */
861 BUILD_BUG_ON(sizeof(struct compat_xt_entry_match) != sizeof(struct xt_entry_match));
862
863 return xt_check_entry_match(elems, base + target_offset,
864 __alignof__(struct compat_xt_entry_match));
865}
866EXPORT_SYMBOL(xt_compat_check_entry_offsets);
867#endif /* CONFIG_NETFILTER_XTABLES_COMPAT */
868
869/**
870 * xt_check_entry_offsets - validate arp/ip/ip6t_entry
871 *
872 * @base: pointer to arp/ip/ip6t_entry
873 * @elems: pointer to first xt_entry_match, i.e. ip(6)t_entry->elems
874 * @target_offset: the arp/ip/ip6_t->target_offset
875 * @next_offset: the arp/ip/ip6_t->next_offset
876 *
877 * validates that target_offset and next_offset are sane and that all
878 * match sizes (if any) align with the target offset.
879 *
880 * This function does not validate the targets or matches themselves, it
881 * only tests that all the offsets and sizes are correct, that all
882 * match structures are aligned, and that the last structure ends where
883 * the target structure begins.
884 *
885 * Also see xt_compat_check_entry_offsets for CONFIG_NETFILTER_XTABLES_COMPAT version.
886 *
887 * The arp/ip/ip6t_entry structure @base must have passed following tests:
888 * - it must point to a valid memory location
889 * - base to base + next_offset must be accessible, i.e. not exceed allocated
890 * length.
891 *
892 * A well-formed entry looks like this:
893 *
894 * ip(6)t_entry match [mtdata] match [mtdata] target [tgdata] ip(6)t_entry
895 * e->elems[]-----' | |
896 * matchsize | |
897 * matchsize | |
898 * | |
899 * target_offset---------------------------------' |
900 * next_offset---------------------------------------------------'
901 *
902 * elems[]: flexible array member at end of ip(6)/arpt_entry struct.
903 * This is where matches (if any) and the target reside.
904 * target_offset: beginning of target.
905 * next_offset: start of the next rule; also: size of this rule.
906 * Since targets have a minimum size, target_offset + minlen <= next_offset.
907 *
908 * Every match stores its size, sum of sizes must not exceed target_offset.
909 *
910 * Return: 0 on success, negative errno on failure.
911 */
912int xt_check_entry_offsets(const void *base,
913 const char *elems,
914 unsigned int target_offset,
915 unsigned int next_offset)
916{
917 long size_of_base_struct = elems - (const char *)base;
918 const struct xt_entry_target *t;
919 const char *e = base;
920
921 /* target start is within the ip/ip6/arpt_entry struct */
922 if (target_offset < size_of_base_struct)
923 return -EINVAL;
924
925 if (target_offset + sizeof(*t) > next_offset)
926 return -EINVAL;
927
928 t = (void *)(e + target_offset);
929 if (t->u.target_size < sizeof(*t))
930 return -EINVAL;
931
932 if (target_offset + t->u.target_size > next_offset)
933 return -EINVAL;
934
935 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) {
936 const struct xt_standard_target *st = (const void *)t;
937
938 if (XT_ALIGN(target_offset + sizeof(*st)) != next_offset)
939 return -EINVAL;
940
941 if (!verdict_ok(st->verdict))
942 return -EINVAL;
943 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) {
944 const struct xt_error_target *et = (const void *)t;
945
946 if (!error_tg_ok(t->u.target_size, sizeof(*et),
947 et->errorname, sizeof(et->errorname)))
948 return -EINVAL;
949 }
950
951 return xt_check_entry_match(elems, base + target_offset,
952 __alignof__(struct xt_entry_match));
953}
954EXPORT_SYMBOL(xt_check_entry_offsets);
955
956/**
957 * xt_alloc_entry_offsets - allocate array to store rule head offsets
958 *
959 * @size: number of entries
960 *
961 * Return: NULL or zeroed kmalloc'd or vmalloc'd array
962 */
963unsigned int *xt_alloc_entry_offsets(unsigned int size)
964{
965 if (size > XT_MAX_TABLE_SIZE / sizeof(unsigned int))
966 return NULL;
967
968 return kvcalloc(size, sizeof(unsigned int), GFP_KERNEL);
969
970}
971EXPORT_SYMBOL(xt_alloc_entry_offsets);
972
973/**
974 * xt_find_jump_offset - check if target is a valid jump offset
975 *
976 * @offsets: array containing all valid rule start offsets of a rule blob
977 * @target: the jump target to search for
978 * @size: entries in @offset
979 */
980bool xt_find_jump_offset(const unsigned int *offsets,
981 unsigned int target, unsigned int size)
982{
983 int m, low = 0, hi = size;
984
985 while (hi > low) {
986 m = (low + hi) / 2u;
987
988 if (offsets[m] > target)
989 hi = m;
990 else if (offsets[m] < target)
991 low = m + 1;
992 else
993 return true;
994 }
995
996 return false;
997}
998EXPORT_SYMBOL(xt_find_jump_offset);
999
1000int xt_check_target(struct xt_tgchk_param *par,
1001 unsigned int size, u16 proto, bool inv_proto)
1002{
1003 int ret;
1004
1005 if (XT_ALIGN(par->target->targetsize) != size) {
1006 pr_err_ratelimited("%s_tables: %s.%u target: invalid size %u (kernel) != (user) %u\n",
1007 xt_prefix[par->family], par->target->name,
1008 par->target->revision,
1009 XT_ALIGN(par->target->targetsize), size);
1010 return -EINVAL;
1011 }
1012 if (par->target->table != NULL &&
1013 strcmp(par->target->table, par->table) != 0) {
1014 pr_info_ratelimited("%s_tables: %s target: only valid in %s table, not %s\n",
1015 xt_prefix[par->family], par->target->name,
1016 par->target->table, par->table);
1017 return -EINVAL;
1018 }
1019 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) {
1020 char used[64], allow[64];
1021
1022 pr_info_ratelimited("%s_tables: %s target: used from hooks %s, but only usable from %s\n",
1023 xt_prefix[par->family], par->target->name,
1024 textify_hooks(used, sizeof(used),
1025 par->hook_mask, par->family),
1026 textify_hooks(allow, sizeof(allow),
1027 par->target->hooks,
1028 par->family));
1029 return -EINVAL;
1030 }
1031 if (par->target->proto && (par->target->proto != proto || inv_proto)) {
1032 pr_info_ratelimited("%s_tables: %s target: only valid for protocol %u\n",
1033 xt_prefix[par->family], par->target->name,
1034 par->target->proto);
1035 return -EINVAL;
1036 }
1037 if (par->target->checkentry != NULL) {
1038 ret = par->target->checkentry(par);
1039 if (ret < 0)
1040 return ret;
1041 else if (ret > 0)
1042 /* Flag up potential errors. */
1043 return -EIO;
1044 }
1045 return 0;
1046}
1047EXPORT_SYMBOL_GPL(xt_check_target);
1048
1049/**
1050 * xt_copy_counters - copy counters and metadata from a sockptr_t
1051 *
1052 * @arg: src sockptr
1053 * @len: alleged size of userspace memory
1054 * @info: where to store the xt_counters_info metadata
1055 *
1056 * Copies counter meta data from @user and stores it in @info.
1057 *
1058 * vmallocs memory to hold the counters, then copies the counter data
1059 * from @user to the new memory and returns a pointer to it.
1060 *
1061 * If called from a compat syscall, @info gets converted automatically to the
1062 * 64bit representation.
1063 *
1064 * The metadata associated with the counters is stored in @info.
1065 *
1066 * Return: returns pointer that caller has to test via IS_ERR().
1067 * If IS_ERR is false, caller has to vfree the pointer.
1068 */
1069void *xt_copy_counters(sockptr_t arg, unsigned int len,
1070 struct xt_counters_info *info)
1071{
1072 size_t offset;
1073 void *mem;
1074 u64 size;
1075
1076#ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1077 if (in_compat_syscall()) {
1078 /* structures only differ in size due to alignment */
1079 struct compat_xt_counters_info compat_tmp;
1080
1081 if (len <= sizeof(compat_tmp))
1082 return ERR_PTR(-EINVAL);
1083
1084 len -= sizeof(compat_tmp);
1085 if (copy_from_sockptr(&compat_tmp, arg, sizeof(compat_tmp)) != 0)
1086 return ERR_PTR(-EFAULT);
1087
1088 memcpy(info->name, compat_tmp.name, sizeof(info->name) - 1);
1089 info->num_counters = compat_tmp.num_counters;
1090 offset = sizeof(compat_tmp);
1091 } else
1092#endif
1093 {
1094 if (len <= sizeof(*info))
1095 return ERR_PTR(-EINVAL);
1096
1097 len -= sizeof(*info);
1098 if (copy_from_sockptr(info, arg, sizeof(*info)) != 0)
1099 return ERR_PTR(-EFAULT);
1100
1101 offset = sizeof(*info);
1102 }
1103 info->name[sizeof(info->name) - 1] = '\0';
1104
1105 size = sizeof(struct xt_counters);
1106 size *= info->num_counters;
1107
1108 if (size != (u64)len)
1109 return ERR_PTR(-EINVAL);
1110
1111 mem = vmalloc(len);
1112 if (!mem)
1113 return ERR_PTR(-ENOMEM);
1114
1115 if (copy_from_sockptr_offset(mem, arg, offset, len) == 0)
1116 return mem;
1117
1118 vfree(mem);
1119 return ERR_PTR(-EFAULT);
1120}
1121EXPORT_SYMBOL_GPL(xt_copy_counters);
1122
1123#ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1124int xt_compat_target_offset(const struct xt_target *target)
1125{
1126 u_int16_t csize = target->compatsize ? : target->targetsize;
1127 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize);
1128}
1129EXPORT_SYMBOL_GPL(xt_compat_target_offset);
1130
1131void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr,
1132 unsigned int *size)
1133{
1134 const struct xt_target *target = t->u.kernel.target;
1135 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t;
1136 int off = xt_compat_target_offset(target);
1137 u_int16_t tsize = ct->u.user.target_size;
1138 char name[sizeof(t->u.user.name)];
1139
1140 t = *dstptr;
1141 memcpy(t, ct, sizeof(*ct));
1142 if (target->compat_from_user)
1143 target->compat_from_user(t->data, ct->data);
1144 else
1145 unsafe_memcpy(t->data, ct->data, tsize - sizeof(*ct),
1146 /* UAPI 0-sized destination */);
1147
1148 tsize += off;
1149 t->u.user.target_size = tsize;
1150 strscpy(name, target->name, sizeof(name));
1151 module_put(target->me);
1152 strscpy_pad(t->u.user.name, name, sizeof(t->u.user.name));
1153
1154 *size += off;
1155 *dstptr += tsize;
1156}
1157EXPORT_SYMBOL_GPL(xt_compat_target_from_user);
1158
1159int xt_compat_target_to_user(const struct xt_entry_target *t,
1160 void __user **dstptr, unsigned int *size)
1161{
1162 const struct xt_target *target = t->u.kernel.target;
1163 struct compat_xt_entry_target __user *ct = *dstptr;
1164 int off = xt_compat_target_offset(target);
1165 u_int16_t tsize = t->u.user.target_size - off;
1166
1167 if (XT_OBJ_TO_USER(ct, t, target, tsize))
1168 return -EFAULT;
1169
1170 if (target->compat_to_user) {
1171 if (target->compat_to_user((void __user *)ct->data, t->data))
1172 return -EFAULT;
1173 } else {
1174 if (COMPAT_XT_DATA_TO_USER(ct, t, target, tsize - sizeof(*ct)))
1175 return -EFAULT;
1176 }
1177
1178 *size -= off;
1179 *dstptr += tsize;
1180 return 0;
1181}
1182EXPORT_SYMBOL_GPL(xt_compat_target_to_user);
1183#endif
1184
1185struct xt_table_info *xt_alloc_table_info(unsigned int size)
1186{
1187 struct xt_table_info *info = NULL;
1188 size_t sz = sizeof(*info) + size;
1189
1190 if (sz < sizeof(*info) || sz >= XT_MAX_TABLE_SIZE)
1191 return NULL;
1192
1193 info = kvmalloc(sz, GFP_KERNEL_ACCOUNT);
1194 if (!info)
1195 return NULL;
1196
1197 memset(info, 0, sizeof(*info));
1198 info->size = size;
1199 return info;
1200}
1201EXPORT_SYMBOL(xt_alloc_table_info);
1202
1203void xt_free_table_info(struct xt_table_info *info)
1204{
1205 int cpu;
1206
1207 if (info->jumpstack != NULL) {
1208 for_each_possible_cpu(cpu)
1209 kvfree(info->jumpstack[cpu]);
1210 kvfree(info->jumpstack);
1211 }
1212
1213 kvfree(info);
1214}
1215EXPORT_SYMBOL(xt_free_table_info);
1216
1217struct xt_table *xt_find_table(struct net *net, u8 af, const char *name)
1218{
1219 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id);
1220 struct xt_table *t;
1221
1222 mutex_lock(&xt[af].mutex);
1223 list_for_each_entry(t, &xt_net->tables[af], list) {
1224 if (strcmp(t->name, name) == 0) {
1225 mutex_unlock(&xt[af].mutex);
1226 return t;
1227 }
1228 }
1229 mutex_unlock(&xt[af].mutex);
1230 return NULL;
1231}
1232EXPORT_SYMBOL(xt_find_table);
1233
1234/* Find table by name, grabs mutex & ref. Returns ERR_PTR on error. */
1235struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af,
1236 const char *name)
1237{
1238 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id);
1239 struct module *owner = NULL;
1240 struct xt_template *tmpl;
1241 struct xt_table *t;
1242
1243 mutex_lock(&xt[af].mutex);
1244 list_for_each_entry(t, &xt_net->tables[af], list)
1245 if (strcmp(t->name, name) == 0 && try_module_get(t->me))
1246 return t;
1247
1248 /* Table doesn't exist in this netns, check larval list */
1249 list_for_each_entry(tmpl, &xt_templates[af], list) {
1250 int err;
1251
1252 if (strcmp(tmpl->name, name))
1253 continue;
1254 if (!try_module_get(tmpl->me))
1255 goto out;
1256
1257 owner = tmpl->me;
1258
1259 mutex_unlock(&xt[af].mutex);
1260 err = tmpl->table_init(net);
1261 if (err < 0) {
1262 module_put(owner);
1263 return ERR_PTR(err);
1264 }
1265
1266 mutex_lock(&xt[af].mutex);
1267 break;
1268 }
1269
1270 /* and once again: */
1271 list_for_each_entry(t, &xt_net->tables[af], list)
1272 if (strcmp(t->name, name) == 0 && owner == t->me)
1273 return t;
1274
1275 module_put(owner);
1276 out:
1277 mutex_unlock(&xt[af].mutex);
1278 return ERR_PTR(-ENOENT);
1279}
1280EXPORT_SYMBOL_GPL(xt_find_table_lock);
1281
1282struct xt_table *xt_request_find_table_lock(struct net *net, u_int8_t af,
1283 const char *name)
1284{
1285 struct xt_table *t = xt_find_table_lock(net, af, name);
1286
1287#ifdef CONFIG_MODULES
1288 if (IS_ERR(t)) {
1289 int err = request_module("%stable_%s", xt_prefix[af], name);
1290 if (err < 0)
1291 return ERR_PTR(err);
1292 t = xt_find_table_lock(net, af, name);
1293 }
1294#endif
1295
1296 return t;
1297}
1298EXPORT_SYMBOL_GPL(xt_request_find_table_lock);
1299
1300void xt_table_unlock(struct xt_table *table)
1301{
1302 mutex_unlock(&xt[table->af].mutex);
1303}
1304EXPORT_SYMBOL_GPL(xt_table_unlock);
1305
1306#ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1307void xt_compat_lock(u_int8_t af)
1308{
1309 mutex_lock(&xt[af].compat_mutex);
1310}
1311EXPORT_SYMBOL_GPL(xt_compat_lock);
1312
1313void xt_compat_unlock(u_int8_t af)
1314{
1315 mutex_unlock(&xt[af].compat_mutex);
1316}
1317EXPORT_SYMBOL_GPL(xt_compat_unlock);
1318#endif
1319
1320DEFINE_PER_CPU(seqcount_t, xt_recseq);
1321EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq);
1322
1323struct static_key xt_tee_enabled __read_mostly;
1324EXPORT_SYMBOL_GPL(xt_tee_enabled);
1325
1326static int xt_jumpstack_alloc(struct xt_table_info *i)
1327{
1328 unsigned int size;
1329 int cpu;
1330
1331 size = sizeof(void **) * nr_cpu_ids;
1332 if (size > PAGE_SIZE)
1333 i->jumpstack = kvzalloc(size, GFP_KERNEL);
1334 else
1335 i->jumpstack = kzalloc(size, GFP_KERNEL);
1336 if (i->jumpstack == NULL)
1337 return -ENOMEM;
1338
1339 /* ruleset without jumps -- no stack needed */
1340 if (i->stacksize == 0)
1341 return 0;
1342
1343 /* Jumpstack needs to be able to record two full callchains, one
1344 * from the first rule set traversal, plus one table reentrancy
1345 * via -j TEE without clobbering the callchain that brought us to
1346 * TEE target.
1347 *
1348 * This is done by allocating two jumpstacks per cpu, on reentry
1349 * the upper half of the stack is used.
1350 *
1351 * see the jumpstack setup in ipt_do_table() for more details.
1352 */
1353 size = sizeof(void *) * i->stacksize * 2u;
1354 for_each_possible_cpu(cpu) {
1355 i->jumpstack[cpu] = kvmalloc_node(size, GFP_KERNEL,
1356 cpu_to_node(cpu));
1357 if (i->jumpstack[cpu] == NULL)
1358 /*
1359 * Freeing will be done later on by the callers. The
1360 * chain is: xt_replace_table -> __do_replace ->
1361 * do_replace -> xt_free_table_info.
1362 */
1363 return -ENOMEM;
1364 }
1365
1366 return 0;
1367}
1368
1369struct xt_counters *xt_counters_alloc(unsigned int counters)
1370{
1371 struct xt_counters *mem;
1372
1373 if (counters == 0 || counters > INT_MAX / sizeof(*mem))
1374 return NULL;
1375
1376 counters *= sizeof(*mem);
1377 if (counters > XT_MAX_TABLE_SIZE)
1378 return NULL;
1379
1380 return vzalloc(counters);
1381}
1382EXPORT_SYMBOL(xt_counters_alloc);
1383
1384struct xt_table_info *
1385xt_replace_table(struct xt_table *table,
1386 unsigned int num_counters,
1387 struct xt_table_info *newinfo,
1388 int *error)
1389{
1390 struct xt_table_info *private;
1391 unsigned int cpu;
1392 int ret;
1393
1394 ret = xt_jumpstack_alloc(newinfo);
1395 if (ret < 0) {
1396 *error = ret;
1397 return NULL;
1398 }
1399
1400 /* Do the substitution. */
1401 local_bh_disable();
1402 private = table->private;
1403
1404 /* Check inside lock: is the old number correct? */
1405 if (num_counters != private->number) {
1406 pr_debug("num_counters != table->private->number (%u/%u)\n",
1407 num_counters, private->number);
1408 local_bh_enable();
1409 *error = -EAGAIN;
1410 return NULL;
1411 }
1412
1413 newinfo->initial_entries = private->initial_entries;
1414 /*
1415 * Ensure contents of newinfo are visible before assigning to
1416 * private.
1417 */
1418 smp_wmb();
1419 table->private = newinfo;
1420
1421 /* make sure all cpus see new ->private value */
1422 smp_mb();
1423
1424 /*
1425 * Even though table entries have now been swapped, other CPU's
1426 * may still be using the old entries...
1427 */
1428 local_bh_enable();
1429
1430 /* ... so wait for even xt_recseq on all cpus */
1431 for_each_possible_cpu(cpu) {
1432 seqcount_t *s = &per_cpu(xt_recseq, cpu);
1433 u32 seq = raw_read_seqcount(s);
1434
1435 if (seq & 1) {
1436 do {
1437 cond_resched();
1438 cpu_relax();
1439 } while (seq == raw_read_seqcount(s));
1440 }
1441 }
1442
1443 audit_log_nfcfg(table->name, table->af, private->number,
1444 !private->number ? AUDIT_XT_OP_REGISTER :
1445 AUDIT_XT_OP_REPLACE,
1446 GFP_KERNEL);
1447 return private;
1448}
1449EXPORT_SYMBOL_GPL(xt_replace_table);
1450
1451struct xt_table *xt_register_table(struct net *net,
1452 const struct xt_table *input_table,
1453 struct xt_table_info *bootstrap,
1454 struct xt_table_info *newinfo)
1455{
1456 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id);
1457 struct xt_table_info *private;
1458 struct xt_table *t, *table;
1459 int ret;
1460
1461 /* Don't add one object to multiple lists. */
1462 table = kmemdup(input_table, sizeof(struct xt_table), GFP_KERNEL);
1463 if (!table) {
1464 ret = -ENOMEM;
1465 goto out;
1466 }
1467
1468 mutex_lock(&xt[table->af].mutex);
1469 /* Don't autoload: we'd eat our tail... */
1470 list_for_each_entry(t, &xt_net->tables[table->af], list) {
1471 if (strcmp(t->name, table->name) == 0) {
1472 ret = -EEXIST;
1473 goto unlock;
1474 }
1475 }
1476
1477 /* Simplifies replace_table code. */
1478 table->private = bootstrap;
1479
1480 if (!xt_replace_table(table, 0, newinfo, &ret))
1481 goto unlock;
1482
1483 private = table->private;
1484 pr_debug("table->private->number = %u\n", private->number);
1485
1486 /* save number of initial entries */
1487 private->initial_entries = private->number;
1488
1489 list_add(&table->list, &xt_net->tables[table->af]);
1490 mutex_unlock(&xt[table->af].mutex);
1491 return table;
1492
1493unlock:
1494 mutex_unlock(&xt[table->af].mutex);
1495 kfree(table);
1496out:
1497 return ERR_PTR(ret);
1498}
1499EXPORT_SYMBOL_GPL(xt_register_table);
1500
1501void *xt_unregister_table(struct xt_table *table)
1502{
1503 struct xt_table_info *private;
1504
1505 mutex_lock(&xt[table->af].mutex);
1506 private = table->private;
1507 list_del(&table->list);
1508 mutex_unlock(&xt[table->af].mutex);
1509 audit_log_nfcfg(table->name, table->af, private->number,
1510 AUDIT_XT_OP_UNREGISTER, GFP_KERNEL);
1511 kfree(table->ops);
1512 kfree(table);
1513
1514 return private;
1515}
1516EXPORT_SYMBOL_GPL(xt_unregister_table);
1517
1518#ifdef CONFIG_PROC_FS
1519static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos)
1520{
1521 u8 af = (unsigned long)pde_data(file_inode(seq->file));
1522 struct net *net = seq_file_net(seq);
1523 struct xt_pernet *xt_net;
1524
1525 xt_net = net_generic(net, xt_pernet_id);
1526
1527 mutex_lock(&xt[af].mutex);
1528 return seq_list_start(&xt_net->tables[af], *pos);
1529}
1530
1531static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1532{
1533 u8 af = (unsigned long)pde_data(file_inode(seq->file));
1534 struct net *net = seq_file_net(seq);
1535 struct xt_pernet *xt_net;
1536
1537 xt_net = net_generic(net, xt_pernet_id);
1538
1539 return seq_list_next(v, &xt_net->tables[af], pos);
1540}
1541
1542static void xt_table_seq_stop(struct seq_file *seq, void *v)
1543{
1544 u_int8_t af = (unsigned long)pde_data(file_inode(seq->file));
1545
1546 mutex_unlock(&xt[af].mutex);
1547}
1548
1549static int xt_table_seq_show(struct seq_file *seq, void *v)
1550{
1551 struct xt_table *table = list_entry(v, struct xt_table, list);
1552
1553 if (*table->name)
1554 seq_printf(seq, "%s\n", table->name);
1555 return 0;
1556}
1557
1558static const struct seq_operations xt_table_seq_ops = {
1559 .start = xt_table_seq_start,
1560 .next = xt_table_seq_next,
1561 .stop = xt_table_seq_stop,
1562 .show = xt_table_seq_show,
1563};
1564
1565/*
1566 * Traverse state for ip{,6}_{tables,matches} for helping crossing
1567 * the multi-AF mutexes.
1568 */
1569struct nf_mttg_trav {
1570 struct list_head *head, *curr;
1571 uint8_t class;
1572};
1573
1574enum {
1575 MTTG_TRAV_INIT,
1576 MTTG_TRAV_NFP_UNSPEC,
1577 MTTG_TRAV_NFP_SPEC,
1578 MTTG_TRAV_DONE,
1579};
1580
1581static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos,
1582 bool is_target)
1583{
1584 static const uint8_t next_class[] = {
1585 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC,
1586 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE,
1587 };
1588 uint8_t nfproto = (unsigned long)pde_data(file_inode(seq->file));
1589 struct nf_mttg_trav *trav = seq->private;
1590
1591 if (ppos != NULL)
1592 ++(*ppos);
1593
1594 switch (trav->class) {
1595 case MTTG_TRAV_INIT:
1596 trav->class = MTTG_TRAV_NFP_UNSPEC;
1597 mutex_lock(&xt[NFPROTO_UNSPEC].mutex);
1598 trav->head = trav->curr = is_target ?
1599 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match;
1600 break;
1601 case MTTG_TRAV_NFP_UNSPEC:
1602 trav->curr = trav->curr->next;
1603 if (trav->curr != trav->head)
1604 break;
1605 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1606 mutex_lock(&xt[nfproto].mutex);
1607 trav->head = trav->curr = is_target ?
1608 &xt[nfproto].target : &xt[nfproto].match;
1609 trav->class = next_class[trav->class];
1610 break;
1611 case MTTG_TRAV_NFP_SPEC:
1612 trav->curr = trav->curr->next;
1613 if (trav->curr != trav->head)
1614 break;
1615 fallthrough;
1616 default:
1617 return NULL;
1618 }
1619 return trav;
1620}
1621
1622static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos,
1623 bool is_target)
1624{
1625 struct nf_mttg_trav *trav = seq->private;
1626 unsigned int j;
1627
1628 trav->class = MTTG_TRAV_INIT;
1629 for (j = 0; j < *pos; ++j)
1630 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL)
1631 return NULL;
1632 return trav;
1633}
1634
1635static void xt_mttg_seq_stop(struct seq_file *seq, void *v)
1636{
1637 uint8_t nfproto = (unsigned long)pde_data(file_inode(seq->file));
1638 struct nf_mttg_trav *trav = seq->private;
1639
1640 switch (trav->class) {
1641 case MTTG_TRAV_NFP_UNSPEC:
1642 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1643 break;
1644 case MTTG_TRAV_NFP_SPEC:
1645 mutex_unlock(&xt[nfproto].mutex);
1646 break;
1647 }
1648}
1649
1650static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos)
1651{
1652 return xt_mttg_seq_start(seq, pos, false);
1653}
1654
1655static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1656{
1657 return xt_mttg_seq_next(seq, v, ppos, false);
1658}
1659
1660static int xt_match_seq_show(struct seq_file *seq, void *v)
1661{
1662 const struct nf_mttg_trav *trav = seq->private;
1663 const struct xt_match *match;
1664
1665 switch (trav->class) {
1666 case MTTG_TRAV_NFP_UNSPEC:
1667 case MTTG_TRAV_NFP_SPEC:
1668 if (trav->curr == trav->head)
1669 return 0;
1670 match = list_entry(trav->curr, struct xt_match, list);
1671 if (*match->name)
1672 seq_printf(seq, "%s\n", match->name);
1673 }
1674 return 0;
1675}
1676
1677static const struct seq_operations xt_match_seq_ops = {
1678 .start = xt_match_seq_start,
1679 .next = xt_match_seq_next,
1680 .stop = xt_mttg_seq_stop,
1681 .show = xt_match_seq_show,
1682};
1683
1684static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos)
1685{
1686 return xt_mttg_seq_start(seq, pos, true);
1687}
1688
1689static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1690{
1691 return xt_mttg_seq_next(seq, v, ppos, true);
1692}
1693
1694static int xt_target_seq_show(struct seq_file *seq, void *v)
1695{
1696 const struct nf_mttg_trav *trav = seq->private;
1697 const struct xt_target *target;
1698
1699 switch (trav->class) {
1700 case MTTG_TRAV_NFP_UNSPEC:
1701 case MTTG_TRAV_NFP_SPEC:
1702 if (trav->curr == trav->head)
1703 return 0;
1704 target = list_entry(trav->curr, struct xt_target, list);
1705 if (*target->name)
1706 seq_printf(seq, "%s\n", target->name);
1707 }
1708 return 0;
1709}
1710
1711static const struct seq_operations xt_target_seq_ops = {
1712 .start = xt_target_seq_start,
1713 .next = xt_target_seq_next,
1714 .stop = xt_mttg_seq_stop,
1715 .show = xt_target_seq_show,
1716};
1717
1718#define FORMAT_TABLES "_tables_names"
1719#define FORMAT_MATCHES "_tables_matches"
1720#define FORMAT_TARGETS "_tables_targets"
1721
1722#endif /* CONFIG_PROC_FS */
1723
1724/**
1725 * xt_hook_ops_alloc - set up hooks for a new table
1726 * @table: table with metadata needed to set up hooks
1727 * @fn: Hook function
1728 *
1729 * This function will create the nf_hook_ops that the x_table needs
1730 * to hand to xt_hook_link_net().
1731 */
1732struct nf_hook_ops *
1733xt_hook_ops_alloc(const struct xt_table *table, nf_hookfn *fn)
1734{
1735 unsigned int hook_mask = table->valid_hooks;
1736 uint8_t i, num_hooks = hweight32(hook_mask);
1737 uint8_t hooknum;
1738 struct nf_hook_ops *ops;
1739
1740 if (!num_hooks)
1741 return ERR_PTR(-EINVAL);
1742
1743 ops = kcalloc(num_hooks, sizeof(*ops), GFP_KERNEL);
1744 if (ops == NULL)
1745 return ERR_PTR(-ENOMEM);
1746
1747 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0;
1748 hook_mask >>= 1, ++hooknum) {
1749 if (!(hook_mask & 1))
1750 continue;
1751 ops[i].hook = fn;
1752 ops[i].pf = table->af;
1753 ops[i].hooknum = hooknum;
1754 ops[i].priority = table->priority;
1755 ++i;
1756 }
1757
1758 return ops;
1759}
1760EXPORT_SYMBOL_GPL(xt_hook_ops_alloc);
1761
1762int xt_register_template(const struct xt_table *table,
1763 int (*table_init)(struct net *net))
1764{
1765 int ret = -EEXIST, af = table->af;
1766 struct xt_template *t;
1767
1768 mutex_lock(&xt[af].mutex);
1769
1770 list_for_each_entry(t, &xt_templates[af], list) {
1771 if (WARN_ON_ONCE(strcmp(table->name, t->name) == 0))
1772 goto out_unlock;
1773 }
1774
1775 ret = -ENOMEM;
1776 t = kzalloc(sizeof(*t), GFP_KERNEL);
1777 if (!t)
1778 goto out_unlock;
1779
1780 BUILD_BUG_ON(sizeof(t->name) != sizeof(table->name));
1781
1782 strscpy(t->name, table->name, sizeof(t->name));
1783 t->table_init = table_init;
1784 t->me = table->me;
1785 list_add(&t->list, &xt_templates[af]);
1786 ret = 0;
1787out_unlock:
1788 mutex_unlock(&xt[af].mutex);
1789 return ret;
1790}
1791EXPORT_SYMBOL_GPL(xt_register_template);
1792
1793void xt_unregister_template(const struct xt_table *table)
1794{
1795 struct xt_template *t;
1796 int af = table->af;
1797
1798 mutex_lock(&xt[af].mutex);
1799 list_for_each_entry(t, &xt_templates[af], list) {
1800 if (strcmp(table->name, t->name))
1801 continue;
1802
1803 list_del(&t->list);
1804 mutex_unlock(&xt[af].mutex);
1805 kfree(t);
1806 return;
1807 }
1808
1809 mutex_unlock(&xt[af].mutex);
1810 WARN_ON_ONCE(1);
1811}
1812EXPORT_SYMBOL_GPL(xt_unregister_template);
1813
1814int xt_proto_init(struct net *net, u_int8_t af)
1815{
1816#ifdef CONFIG_PROC_FS
1817 char buf[XT_FUNCTION_MAXNAMELEN];
1818 struct proc_dir_entry *proc;
1819 kuid_t root_uid;
1820 kgid_t root_gid;
1821#endif
1822
1823 if (af >= ARRAY_SIZE(xt_prefix))
1824 return -EINVAL;
1825
1826
1827#ifdef CONFIG_PROC_FS
1828 root_uid = make_kuid(net->user_ns, 0);
1829 root_gid = make_kgid(net->user_ns, 0);
1830
1831 strscpy(buf, xt_prefix[af], sizeof(buf));
1832 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1833 proc = proc_create_net_data(buf, 0440, net->proc_net, &xt_table_seq_ops,
1834 sizeof(struct seq_net_private),
1835 (void *)(unsigned long)af);
1836 if (!proc)
1837 goto out;
1838 if (uid_valid(root_uid) && gid_valid(root_gid))
1839 proc_set_user(proc, root_uid, root_gid);
1840
1841 strscpy(buf, xt_prefix[af], sizeof(buf));
1842 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1843 proc = proc_create_seq_private(buf, 0440, net->proc_net,
1844 &xt_match_seq_ops, sizeof(struct nf_mttg_trav),
1845 (void *)(unsigned long)af);
1846 if (!proc)
1847 goto out_remove_tables;
1848 if (uid_valid(root_uid) && gid_valid(root_gid))
1849 proc_set_user(proc, root_uid, root_gid);
1850
1851 strscpy(buf, xt_prefix[af], sizeof(buf));
1852 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1853 proc = proc_create_seq_private(buf, 0440, net->proc_net,
1854 &xt_target_seq_ops, sizeof(struct nf_mttg_trav),
1855 (void *)(unsigned long)af);
1856 if (!proc)
1857 goto out_remove_matches;
1858 if (uid_valid(root_uid) && gid_valid(root_gid))
1859 proc_set_user(proc, root_uid, root_gid);
1860#endif
1861
1862 return 0;
1863
1864#ifdef CONFIG_PROC_FS
1865out_remove_matches:
1866 strscpy(buf, xt_prefix[af], sizeof(buf));
1867 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1868 remove_proc_entry(buf, net->proc_net);
1869
1870out_remove_tables:
1871 strscpy(buf, xt_prefix[af], sizeof(buf));
1872 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1873 remove_proc_entry(buf, net->proc_net);
1874out:
1875 return -1;
1876#endif
1877}
1878EXPORT_SYMBOL_GPL(xt_proto_init);
1879
1880void xt_proto_fini(struct net *net, u_int8_t af)
1881{
1882#ifdef CONFIG_PROC_FS
1883 char buf[XT_FUNCTION_MAXNAMELEN];
1884
1885 strscpy(buf, xt_prefix[af], sizeof(buf));
1886 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1887 remove_proc_entry(buf, net->proc_net);
1888
1889 strscpy(buf, xt_prefix[af], sizeof(buf));
1890 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1891 remove_proc_entry(buf, net->proc_net);
1892
1893 strscpy(buf, xt_prefix[af], sizeof(buf));
1894 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1895 remove_proc_entry(buf, net->proc_net);
1896#endif /*CONFIG_PROC_FS*/
1897}
1898EXPORT_SYMBOL_GPL(xt_proto_fini);
1899
1900/**
1901 * xt_percpu_counter_alloc - allocate x_tables rule counter
1902 *
1903 * @state: pointer to xt_percpu allocation state
1904 * @counter: pointer to counter struct inside the ip(6)/arpt_entry struct
1905 *
1906 * On SMP, the packet counter [ ip(6)t_entry->counters.pcnt ] will then
1907 * contain the address of the real (percpu) counter.
1908 *
1909 * Rule evaluation needs to use xt_get_this_cpu_counter() helper
1910 * to fetch the real percpu counter.
1911 *
1912 * To speed up allocation and improve data locality, a 4kb block is
1913 * allocated. Freeing any counter may free an entire block, so all
1914 * counters allocated using the same state must be freed at the same
1915 * time.
1916 *
1917 * xt_percpu_counter_alloc_state contains the base address of the
1918 * allocated page and the current sub-offset.
1919 *
1920 * returns false on error.
1921 */
1922bool xt_percpu_counter_alloc(struct xt_percpu_counter_alloc_state *state,
1923 struct xt_counters *counter)
1924{
1925 BUILD_BUG_ON(XT_PCPU_BLOCK_SIZE < (sizeof(*counter) * 2));
1926
1927 if (nr_cpu_ids <= 1)
1928 return true;
1929
1930 if (!state->mem) {
1931 state->mem = __alloc_percpu(XT_PCPU_BLOCK_SIZE,
1932 XT_PCPU_BLOCK_SIZE);
1933 if (!state->mem)
1934 return false;
1935 }
1936 counter->pcnt = (__force unsigned long)(state->mem + state->off);
1937 state->off += sizeof(*counter);
1938 if (state->off > (XT_PCPU_BLOCK_SIZE - sizeof(*counter))) {
1939 state->mem = NULL;
1940 state->off = 0;
1941 }
1942 return true;
1943}
1944EXPORT_SYMBOL_GPL(xt_percpu_counter_alloc);
1945
1946void xt_percpu_counter_free(struct xt_counters *counters)
1947{
1948 unsigned long pcnt = counters->pcnt;
1949
1950 if (nr_cpu_ids > 1 && (pcnt & (XT_PCPU_BLOCK_SIZE - 1)) == 0)
1951 free_percpu((void __percpu *)pcnt);
1952}
1953EXPORT_SYMBOL_GPL(xt_percpu_counter_free);
1954
1955static int __net_init xt_net_init(struct net *net)
1956{
1957 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id);
1958 int i;
1959
1960 for (i = 0; i < NFPROTO_NUMPROTO; i++)
1961 INIT_LIST_HEAD(&xt_net->tables[i]);
1962 return 0;
1963}
1964
1965static void __net_exit xt_net_exit(struct net *net)
1966{
1967 struct xt_pernet *xt_net = net_generic(net, xt_pernet_id);
1968 int i;
1969
1970 for (i = 0; i < NFPROTO_NUMPROTO; i++)
1971 WARN_ON_ONCE(!list_empty(&xt_net->tables[i]));
1972}
1973
1974static struct pernet_operations xt_net_ops = {
1975 .init = xt_net_init,
1976 .exit = xt_net_exit,
1977 .id = &xt_pernet_id,
1978 .size = sizeof(struct xt_pernet),
1979};
1980
1981static int __init xt_init(void)
1982{
1983 unsigned int i;
1984 int rv;
1985
1986 for_each_possible_cpu(i) {
1987 seqcount_init(&per_cpu(xt_recseq, i));
1988 }
1989
1990 xt = kcalloc(NFPROTO_NUMPROTO, sizeof(struct xt_af), GFP_KERNEL);
1991 if (!xt)
1992 return -ENOMEM;
1993
1994 for (i = 0; i < NFPROTO_NUMPROTO; i++) {
1995 mutex_init(&xt[i].mutex);
1996#ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1997 mutex_init(&xt[i].compat_mutex);
1998 xt[i].compat_tab = NULL;
1999#endif
2000 INIT_LIST_HEAD(&xt[i].target);
2001 INIT_LIST_HEAD(&xt[i].match);
2002 INIT_LIST_HEAD(&xt_templates[i]);
2003 }
2004 rv = register_pernet_subsys(&xt_net_ops);
2005 if (rv < 0)
2006 kfree(xt);
2007 return rv;
2008}
2009
2010static void __exit xt_fini(void)
2011{
2012 unregister_pernet_subsys(&xt_net_ops);
2013 kfree(xt);
2014}
2015
2016module_init(xt_init);
2017module_exit(xt_fini);