2 * x_tables core - Backend for {ip,ip6,arp}_tables
4 * Copyright (C) 2006-2006 Harald Welte <laforge@netfilter.org>
5 * Copyright (C) 2006-2012 Patrick McHardy <kaber@trash.net>
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>
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.
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>
27 #include <linux/slab.h>
28 #include <linux/audit.h>
29 #include <net/net_namespace.h>
31 #include <linux/netfilter/x_tables.h>
32 #include <linux/netfilter_arp.h>
33 #include <linux/netfilter_ipv4/ip_tables.h>
34 #include <linux/netfilter_ipv6/ip6_tables.h>
35 #include <linux/netfilter_arp/arp_tables.h>
37 MODULE_LICENSE("GPL");
38 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
39 MODULE_DESCRIPTION("{ip,ip6,arp,eb}_tables backend module");
41 #define XT_PCPU_BLOCK_SIZE 4096
44 unsigned int offset; /* offset in kernel */
45 int delta; /* delta in 32bit user land */
50 struct list_head match;
51 struct list_head target;
53 struct mutex compat_mutex;
54 struct compat_delta *compat_tab;
55 unsigned int number; /* number of slots in compat_tab[] */
56 unsigned int cur; /* number of used slots in compat_tab[] */
60 static struct xt_af *xt;
62 static const char *const xt_prefix[NFPROTO_NUMPROTO] = {
63 [NFPROTO_UNSPEC] = "x",
64 [NFPROTO_IPV4] = "ip",
65 [NFPROTO_ARP] = "arp",
66 [NFPROTO_BRIDGE] = "eb",
67 [NFPROTO_IPV6] = "ip6",
70 /* Registration hooks for targets. */
71 int xt_register_target(struct xt_target *target)
73 u_int8_t af = target->family;
75 mutex_lock(&xt[af].mutex);
76 list_add(&target->list, &xt[af].target);
77 mutex_unlock(&xt[af].mutex);
80 EXPORT_SYMBOL(xt_register_target);
83 xt_unregister_target(struct xt_target *target)
85 u_int8_t af = target->family;
87 mutex_lock(&xt[af].mutex);
88 list_del(&target->list);
89 mutex_unlock(&xt[af].mutex);
91 EXPORT_SYMBOL(xt_unregister_target);
94 xt_register_targets(struct xt_target *target, unsigned int n)
99 for (i = 0; i < n; i++) {
100 err = xt_register_target(&target[i]);
108 xt_unregister_targets(target, i);
111 EXPORT_SYMBOL(xt_register_targets);
114 xt_unregister_targets(struct xt_target *target, unsigned int n)
117 xt_unregister_target(&target[n]);
119 EXPORT_SYMBOL(xt_unregister_targets);
121 int xt_register_match(struct xt_match *match)
123 u_int8_t af = match->family;
125 mutex_lock(&xt[af].mutex);
126 list_add(&match->list, &xt[af].match);
127 mutex_unlock(&xt[af].mutex);
130 EXPORT_SYMBOL(xt_register_match);
133 xt_unregister_match(struct xt_match *match)
135 u_int8_t af = match->family;
137 mutex_lock(&xt[af].mutex);
138 list_del(&match->list);
139 mutex_unlock(&xt[af].mutex);
141 EXPORT_SYMBOL(xt_unregister_match);
144 xt_register_matches(struct xt_match *match, unsigned int n)
149 for (i = 0; i < n; i++) {
150 err = xt_register_match(&match[i]);
158 xt_unregister_matches(match, i);
161 EXPORT_SYMBOL(xt_register_matches);
164 xt_unregister_matches(struct xt_match *match, unsigned int n)
167 xt_unregister_match(&match[n]);
169 EXPORT_SYMBOL(xt_unregister_matches);
173 * These are weird, but module loading must not be done with mutex
174 * held (since they will register), and we have to have a single
178 /* Find match, grabs ref. Returns ERR_PTR() on error. */
179 struct xt_match *xt_find_match(u8 af, const char *name, u8 revision)
184 mutex_lock(&xt[af].mutex);
185 list_for_each_entry(m, &xt[af].match, list) {
186 if (strcmp(m->name, name) == 0) {
187 if (m->revision == revision) {
188 if (try_module_get(m->me)) {
189 mutex_unlock(&xt[af].mutex);
193 err = -EPROTOTYPE; /* Found something. */
196 mutex_unlock(&xt[af].mutex);
198 if (af != NFPROTO_UNSPEC)
199 /* Try searching again in the family-independent list */
200 return xt_find_match(NFPROTO_UNSPEC, name, revision);
204 EXPORT_SYMBOL(xt_find_match);
207 xt_request_find_match(uint8_t nfproto, const char *name, uint8_t revision)
209 struct xt_match *match;
211 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN)
212 return ERR_PTR(-EINVAL);
214 match = xt_find_match(nfproto, name, revision);
216 request_module("%st_%s", xt_prefix[nfproto], name);
217 match = xt_find_match(nfproto, name, revision);
222 EXPORT_SYMBOL_GPL(xt_request_find_match);
224 /* Find target, grabs ref. Returns ERR_PTR() on error. */
225 struct xt_target *xt_find_target(u8 af, const char *name, u8 revision)
230 mutex_lock(&xt[af].mutex);
231 list_for_each_entry(t, &xt[af].target, list) {
232 if (strcmp(t->name, name) == 0) {
233 if (t->revision == revision) {
234 if (try_module_get(t->me)) {
235 mutex_unlock(&xt[af].mutex);
239 err = -EPROTOTYPE; /* Found something. */
242 mutex_unlock(&xt[af].mutex);
244 if (af != NFPROTO_UNSPEC)
245 /* Try searching again in the family-independent list */
246 return xt_find_target(NFPROTO_UNSPEC, name, revision);
250 EXPORT_SYMBOL(xt_find_target);
252 struct xt_target *xt_request_find_target(u8 af, const char *name, u8 revision)
254 struct xt_target *target;
256 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN)
257 return ERR_PTR(-EINVAL);
259 target = xt_find_target(af, name, revision);
260 if (IS_ERR(target)) {
261 request_module("%st_%s", xt_prefix[af], name);
262 target = xt_find_target(af, name, revision);
267 EXPORT_SYMBOL_GPL(xt_request_find_target);
269 static int match_revfn(u8 af, const char *name, u8 revision, int *bestp)
271 const struct xt_match *m;
274 mutex_lock(&xt[af].mutex);
275 list_for_each_entry(m, &xt[af].match, list) {
276 if (strcmp(m->name, name) == 0) {
277 if (m->revision > *bestp)
278 *bestp = m->revision;
279 if (m->revision == revision)
283 mutex_unlock(&xt[af].mutex);
285 if (af != NFPROTO_UNSPEC && !have_rev)
286 return match_revfn(NFPROTO_UNSPEC, name, revision, bestp);
291 static int target_revfn(u8 af, const char *name, u8 revision, int *bestp)
293 const struct xt_target *t;
296 mutex_lock(&xt[af].mutex);
297 list_for_each_entry(t, &xt[af].target, list) {
298 if (strcmp(t->name, name) == 0) {
299 if (t->revision > *bestp)
300 *bestp = t->revision;
301 if (t->revision == revision)
305 mutex_unlock(&xt[af].mutex);
307 if (af != NFPROTO_UNSPEC && !have_rev)
308 return target_revfn(NFPROTO_UNSPEC, name, revision, bestp);
313 /* Returns true or false (if no such extension at all) */
314 int xt_find_revision(u8 af, const char *name, u8 revision, int target,
317 int have_rev, best = -1;
320 have_rev = target_revfn(af, name, revision, &best);
322 have_rev = match_revfn(af, name, revision, &best);
324 /* Nothing at all? Return 0 to try loading module. */
332 *err = -EPROTONOSUPPORT;
335 EXPORT_SYMBOL_GPL(xt_find_revision);
338 textify_hooks(char *buf, size_t size, unsigned int mask, uint8_t nfproto)
340 static const char *const inetbr_names[] = {
341 "PREROUTING", "INPUT", "FORWARD",
342 "OUTPUT", "POSTROUTING", "BROUTING",
344 static const char *const arp_names[] = {
345 "INPUT", "FORWARD", "OUTPUT",
347 const char *const *names;
353 names = (nfproto == NFPROTO_ARP) ? arp_names : inetbr_names;
354 max = (nfproto == NFPROTO_ARP) ? ARRAY_SIZE(arp_names) :
355 ARRAY_SIZE(inetbr_names);
357 for (i = 0; i < max; ++i) {
358 if (!(mask & (1 << i)))
360 res = snprintf(p, size, "%s%s", np ? "/" : "", names[i]);
372 * xt_check_proc_name - check that name is suitable for /proc file creation
374 * @name: file name candidate
375 * @size: length of buffer
377 * some x_tables modules wish to create a file in /proc.
378 * This function makes sure that the name is suitable for this
379 * purpose, it checks that name is NUL terminated and isn't a 'special'
382 * returns negative number on error or 0 if name is useable.
384 int xt_check_proc_name(const char *name, unsigned int size)
389 if (strnlen(name, size) == size)
390 return -ENAMETOOLONG;
392 if (strcmp(name, ".") == 0 ||
393 strcmp(name, "..") == 0 ||
399 EXPORT_SYMBOL(xt_check_proc_name);
401 int xt_check_match(struct xt_mtchk_param *par,
402 unsigned int size, u_int8_t proto, bool inv_proto)
406 if (XT_ALIGN(par->match->matchsize) != size &&
407 par->match->matchsize != -1) {
409 * ebt_among is exempt from centralized matchsize checking
410 * because it uses a dynamic-size data set.
412 pr_err("%s_tables: %s.%u match: invalid size "
413 "%u (kernel) != (user) %u\n",
414 xt_prefix[par->family], par->match->name,
415 par->match->revision,
416 XT_ALIGN(par->match->matchsize), size);
419 if (par->match->table != NULL &&
420 strcmp(par->match->table, par->table) != 0) {
421 pr_err("%s_tables: %s match: only valid in %s table, not %s\n",
422 xt_prefix[par->family], par->match->name,
423 par->match->table, par->table);
426 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) {
427 char used[64], allow[64];
429 pr_err("%s_tables: %s match: used from hooks %s, but only "
431 xt_prefix[par->family], par->match->name,
432 textify_hooks(used, sizeof(used), par->hook_mask,
434 textify_hooks(allow, sizeof(allow), par->match->hooks,
438 if (par->match->proto && (par->match->proto != proto || inv_proto)) {
439 pr_err("%s_tables: %s match: only valid for protocol %u\n",
440 xt_prefix[par->family], par->match->name,
444 if (par->match->checkentry != NULL) {
445 ret = par->match->checkentry(par);
449 /* Flag up potential errors. */
454 EXPORT_SYMBOL_GPL(xt_check_match);
456 /** xt_check_entry_match - check that matches end before start of target
458 * @match: beginning of xt_entry_match
459 * @target: beginning of this rules target (alleged end of matches)
460 * @alignment: alignment requirement of match structures
462 * Validates that all matches add up to the beginning of the target,
463 * and that each match covers at least the base structure size.
465 * Return: 0 on success, negative errno on failure.
467 static int xt_check_entry_match(const char *match, const char *target,
468 const size_t alignment)
470 const struct xt_entry_match *pos;
471 int length = target - match;
473 if (length == 0) /* no matches */
476 pos = (struct xt_entry_match *)match;
478 if ((unsigned long)pos % alignment)
481 if (length < (int)sizeof(struct xt_entry_match))
484 if (pos->u.match_size < sizeof(struct xt_entry_match))
487 if (pos->u.match_size > length)
490 length -= pos->u.match_size;
491 pos = ((void *)((char *)(pos) + (pos)->u.match_size));
492 } while (length > 0);
498 int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta)
500 struct xt_af *xp = &xt[af];
502 if (!xp->compat_tab) {
505 xp->compat_tab = vmalloc(sizeof(struct compat_delta) * xp->number);
511 if (xp->cur >= xp->number)
515 delta += xp->compat_tab[xp->cur - 1].delta;
516 xp->compat_tab[xp->cur].offset = offset;
517 xp->compat_tab[xp->cur].delta = delta;
521 EXPORT_SYMBOL_GPL(xt_compat_add_offset);
523 void xt_compat_flush_offsets(u_int8_t af)
525 if (xt[af].compat_tab) {
526 vfree(xt[af].compat_tab);
527 xt[af].compat_tab = NULL;
532 EXPORT_SYMBOL_GPL(xt_compat_flush_offsets);
534 int xt_compat_calc_jump(u_int8_t af, unsigned int offset)
536 struct compat_delta *tmp = xt[af].compat_tab;
537 int mid, left = 0, right = xt[af].cur - 1;
539 while (left <= right) {
540 mid = (left + right) >> 1;
541 if (offset > tmp[mid].offset)
543 else if (offset < tmp[mid].offset)
546 return mid ? tmp[mid - 1].delta : 0;
548 return left ? tmp[left - 1].delta : 0;
550 EXPORT_SYMBOL_GPL(xt_compat_calc_jump);
552 void xt_compat_init_offsets(u_int8_t af, unsigned int number)
554 xt[af].number = number;
557 EXPORT_SYMBOL(xt_compat_init_offsets);
559 int xt_compat_match_offset(const struct xt_match *match)
561 u_int16_t csize = match->compatsize ? : match->matchsize;
562 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize);
564 EXPORT_SYMBOL_GPL(xt_compat_match_offset);
566 void xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr,
569 const struct xt_match *match = m->u.kernel.match;
570 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m;
571 int off = xt_compat_match_offset(match);
572 u_int16_t msize = cm->u.user.match_size;
573 char name[sizeof(m->u.user.name)];
576 memcpy(m, cm, sizeof(*cm));
577 if (match->compat_from_user)
578 match->compat_from_user(m->data, cm->data);
580 memcpy(m->data, cm->data, msize - sizeof(*cm));
583 m->u.user.match_size = msize;
584 strlcpy(name, match->name, sizeof(name));
585 module_put(match->me);
586 strncpy(m->u.user.name, name, sizeof(m->u.user.name));
591 EXPORT_SYMBOL_GPL(xt_compat_match_from_user);
593 int xt_compat_match_to_user(const struct xt_entry_match *m,
594 void __user **dstptr, unsigned int *size)
596 const struct xt_match *match = m->u.kernel.match;
597 struct compat_xt_entry_match __user *cm = *dstptr;
598 int off = xt_compat_match_offset(match);
599 u_int16_t msize = m->u.user.match_size - off;
601 if (copy_to_user(cm, m, sizeof(*cm)) ||
602 put_user(msize, &cm->u.user.match_size) ||
603 copy_to_user(cm->u.user.name, m->u.kernel.match->name,
604 strlen(m->u.kernel.match->name) + 1))
607 if (match->compat_to_user) {
608 if (match->compat_to_user((void __user *)cm->data, m->data))
611 if (copy_to_user(cm->data, m->data, msize - sizeof(*cm)))
619 EXPORT_SYMBOL_GPL(xt_compat_match_to_user);
621 /* non-compat version may have padding after verdict */
622 struct compat_xt_standard_target {
623 struct compat_xt_entry_target t;
624 compat_uint_t verdict;
627 int xt_compat_check_entry_offsets(const void *base, const char *elems,
628 unsigned int target_offset,
629 unsigned int next_offset)
631 long size_of_base_struct = elems - (const char *)base;
632 const struct compat_xt_entry_target *t;
633 const char *e = base;
635 if (target_offset < size_of_base_struct)
638 if (target_offset + sizeof(*t) > next_offset)
641 t = (void *)(e + target_offset);
642 if (t->u.target_size < sizeof(*t))
645 if (target_offset + t->u.target_size > next_offset)
648 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
649 COMPAT_XT_ALIGN(target_offset + sizeof(struct compat_xt_standard_target)) != next_offset)
652 /* compat_xt_entry match has less strict aligment requirements,
653 * otherwise they are identical. In case of padding differences
654 * we need to add compat version of xt_check_entry_match.
656 BUILD_BUG_ON(sizeof(struct compat_xt_entry_match) != sizeof(struct xt_entry_match));
658 return xt_check_entry_match(elems, base + target_offset,
659 __alignof__(struct compat_xt_entry_match));
661 EXPORT_SYMBOL(xt_compat_check_entry_offsets);
662 #endif /* CONFIG_COMPAT */
665 * xt_check_entry_offsets - validate arp/ip/ip6t_entry
667 * @base: pointer to arp/ip/ip6t_entry
668 * @elems: pointer to first xt_entry_match, i.e. ip(6)t_entry->elems
669 * @target_offset: the arp/ip/ip6_t->target_offset
670 * @next_offset: the arp/ip/ip6_t->next_offset
672 * validates that target_offset and next_offset are sane and that all
673 * match sizes (if any) align with the target offset.
675 * This function does not validate the targets or matches themselves, it
676 * only tests that all the offsets and sizes are correct, that all
677 * match structures are aligned, and that the last structure ends where
678 * the target structure begins.
680 * Also see xt_compat_check_entry_offsets for CONFIG_COMPAT version.
682 * The arp/ip/ip6t_entry structure @base must have passed following tests:
683 * - it must point to a valid memory location
684 * - base to base + next_offset must be accessible, i.e. not exceed allocated
687 * A well-formed entry looks like this:
689 * ip(6)t_entry match [mtdata] match [mtdata] target [tgdata] ip(6)t_entry
690 * e->elems[]-----' | |
694 * target_offset---------------------------------' |
695 * next_offset---------------------------------------------------'
697 * elems[]: flexible array member at end of ip(6)/arpt_entry struct.
698 * This is where matches (if any) and the target reside.
699 * target_offset: beginning of target.
700 * next_offset: start of the next rule; also: size of this rule.
701 * Since targets have a minimum size, target_offset + minlen <= next_offset.
703 * Every match stores its size, sum of sizes must not exceed target_offset.
705 * Return: 0 on success, negative errno on failure.
707 int xt_check_entry_offsets(const void *base,
709 unsigned int target_offset,
710 unsigned int next_offset)
712 long size_of_base_struct = elems - (const char *)base;
713 const struct xt_entry_target *t;
714 const char *e = base;
716 /* target start is within the ip/ip6/arpt_entry struct */
717 if (target_offset < size_of_base_struct)
720 if (target_offset + sizeof(*t) > next_offset)
723 t = (void *)(e + target_offset);
724 if (t->u.target_size < sizeof(*t))
727 if (target_offset + t->u.target_size > next_offset)
730 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
731 XT_ALIGN(target_offset + sizeof(struct xt_standard_target)) != next_offset)
734 return xt_check_entry_match(elems, base + target_offset,
735 __alignof__(struct xt_entry_match));
737 EXPORT_SYMBOL(xt_check_entry_offsets);
740 * xt_alloc_entry_offsets - allocate array to store rule head offsets
742 * @size: number of entries
744 * Return: NULL or kmalloc'd or vmalloc'd array
746 unsigned int *xt_alloc_entry_offsets(unsigned int size)
750 off = kcalloc(size, sizeof(unsigned int), GFP_KERNEL | __GFP_NOWARN);
755 if (size < (SIZE_MAX / sizeof(unsigned int)))
756 off = vmalloc(size * sizeof(unsigned int));
760 EXPORT_SYMBOL(xt_alloc_entry_offsets);
763 * xt_find_jump_offset - check if target is a valid jump offset
765 * @offsets: array containing all valid rule start offsets of a rule blob
766 * @target: the jump target to search for
767 * @size: entries in @offset
769 bool xt_find_jump_offset(const unsigned int *offsets,
770 unsigned int target, unsigned int size)
772 int m, low = 0, hi = size;
777 if (offsets[m] > target)
779 else if (offsets[m] < target)
787 EXPORT_SYMBOL(xt_find_jump_offset);
789 int xt_check_target(struct xt_tgchk_param *par,
790 unsigned int size, u_int8_t proto, bool inv_proto)
794 if (XT_ALIGN(par->target->targetsize) != size) {
795 pr_err("%s_tables: %s.%u target: invalid size "
796 "%u (kernel) != (user) %u\n",
797 xt_prefix[par->family], par->target->name,
798 par->target->revision,
799 XT_ALIGN(par->target->targetsize), size);
802 if (par->target->table != NULL &&
803 strcmp(par->target->table, par->table) != 0) {
804 pr_err("%s_tables: %s target: only valid in %s table, not %s\n",
805 xt_prefix[par->family], par->target->name,
806 par->target->table, par->table);
809 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) {
810 char used[64], allow[64];
812 pr_err("%s_tables: %s target: used from hooks %s, but only "
814 xt_prefix[par->family], par->target->name,
815 textify_hooks(used, sizeof(used), par->hook_mask,
817 textify_hooks(allow, sizeof(allow), par->target->hooks,
821 if (par->target->proto && (par->target->proto != proto || inv_proto)) {
822 pr_err("%s_tables: %s target: only valid for protocol %u\n",
823 xt_prefix[par->family], par->target->name,
827 if (par->target->checkentry != NULL) {
828 ret = par->target->checkentry(par);
832 /* Flag up potential errors. */
837 EXPORT_SYMBOL_GPL(xt_check_target);
840 * xt_copy_counters_from_user - copy counters and metadata from userspace
842 * @user: src pointer to userspace memory
843 * @len: alleged size of userspace memory
844 * @info: where to store the xt_counters_info metadata
845 * @compat: true if we setsockopt call is done by 32bit task on 64bit kernel
847 * Copies counter meta data from @user and stores it in @info.
849 * vmallocs memory to hold the counters, then copies the counter data
850 * from @user to the new memory and returns a pointer to it.
852 * If @compat is true, @info gets converted automatically to the 64bit
855 * The metadata associated with the counters is stored in @info.
857 * Return: returns pointer that caller has to test via IS_ERR().
858 * If IS_ERR is false, caller has to vfree the pointer.
860 void *xt_copy_counters_from_user(const void __user *user, unsigned int len,
861 struct xt_counters_info *info, bool compat)
868 /* structures only differ in size due to alignment */
869 struct compat_xt_counters_info compat_tmp;
871 if (len <= sizeof(compat_tmp))
872 return ERR_PTR(-EINVAL);
874 len -= sizeof(compat_tmp);
875 if (copy_from_user(&compat_tmp, user, sizeof(compat_tmp)) != 0)
876 return ERR_PTR(-EFAULT);
878 memcpy(info->name, compat_tmp.name, sizeof(info->name) - 1);
879 info->num_counters = compat_tmp.num_counters;
880 user += sizeof(compat_tmp);
884 if (len <= sizeof(*info))
885 return ERR_PTR(-EINVAL);
887 len -= sizeof(*info);
888 if (copy_from_user(info, user, sizeof(*info)) != 0)
889 return ERR_PTR(-EFAULT);
891 user += sizeof(*info);
893 info->name[sizeof(info->name) - 1] = '\0';
895 size = sizeof(struct xt_counters);
896 size *= info->num_counters;
898 if (size != (u64)len)
899 return ERR_PTR(-EINVAL);
903 return ERR_PTR(-ENOMEM);
905 if (copy_from_user(mem, user, len) == 0)
909 return ERR_PTR(-EFAULT);
911 EXPORT_SYMBOL_GPL(xt_copy_counters_from_user);
914 int xt_compat_target_offset(const struct xt_target *target)
916 u_int16_t csize = target->compatsize ? : target->targetsize;
917 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize);
919 EXPORT_SYMBOL_GPL(xt_compat_target_offset);
921 void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr,
924 const struct xt_target *target = t->u.kernel.target;
925 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t;
926 int off = xt_compat_target_offset(target);
927 u_int16_t tsize = ct->u.user.target_size;
928 char name[sizeof(t->u.user.name)];
931 memcpy(t, ct, sizeof(*ct));
932 if (target->compat_from_user)
933 target->compat_from_user(t->data, ct->data);
935 memcpy(t->data, ct->data, tsize - sizeof(*ct));
938 t->u.user.target_size = tsize;
939 strlcpy(name, target->name, sizeof(name));
940 module_put(target->me);
941 strncpy(t->u.user.name, name, sizeof(t->u.user.name));
946 EXPORT_SYMBOL_GPL(xt_compat_target_from_user);
948 int xt_compat_target_to_user(const struct xt_entry_target *t,
949 void __user **dstptr, unsigned int *size)
951 const struct xt_target *target = t->u.kernel.target;
952 struct compat_xt_entry_target __user *ct = *dstptr;
953 int off = xt_compat_target_offset(target);
954 u_int16_t tsize = t->u.user.target_size - off;
956 if (copy_to_user(ct, t, sizeof(*ct)) ||
957 put_user(tsize, &ct->u.user.target_size) ||
958 copy_to_user(ct->u.user.name, t->u.kernel.target->name,
959 strlen(t->u.kernel.target->name) + 1))
962 if (target->compat_to_user) {
963 if (target->compat_to_user((void __user *)ct->data, t->data))
966 if (copy_to_user(ct->data, t->data, tsize - sizeof(*ct)))
974 EXPORT_SYMBOL_GPL(xt_compat_target_to_user);
977 struct xt_table_info *xt_alloc_table_info(unsigned int size)
979 struct xt_table_info *info = NULL;
980 size_t sz = sizeof(*info) + size;
982 if (sz < sizeof(*info))
985 if (sz < sizeof(*info))
988 /* Pedantry: prevent them from hitting BUG() in vmalloc.c --RR */
989 if ((size >> PAGE_SHIFT) + 2 > totalram_pages)
992 if (sz <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
993 info = kmalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY);
999 memset(info, 0, sizeof(*info));
1003 EXPORT_SYMBOL(xt_alloc_table_info);
1005 void xt_free_table_info(struct xt_table_info *info)
1009 if (info->jumpstack != NULL) {
1010 for_each_possible_cpu(cpu)
1011 kvfree(info->jumpstack[cpu]);
1012 kvfree(info->jumpstack);
1017 EXPORT_SYMBOL(xt_free_table_info);
1019 /* Find table by name, grabs mutex & ref. Returns ERR_PTR() on error. */
1020 struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af,
1025 mutex_lock(&xt[af].mutex);
1026 list_for_each_entry(t, &net->xt.tables[af], list)
1027 if (strcmp(t->name, name) == 0 && try_module_get(t->me))
1029 mutex_unlock(&xt[af].mutex);
1032 EXPORT_SYMBOL_GPL(xt_find_table_lock);
1034 void xt_table_unlock(struct xt_table *table)
1036 mutex_unlock(&xt[table->af].mutex);
1038 EXPORT_SYMBOL_GPL(xt_table_unlock);
1040 #ifdef CONFIG_COMPAT
1041 void xt_compat_lock(u_int8_t af)
1043 mutex_lock(&xt[af].compat_mutex);
1045 EXPORT_SYMBOL_GPL(xt_compat_lock);
1047 void xt_compat_unlock(u_int8_t af)
1049 mutex_unlock(&xt[af].compat_mutex);
1051 EXPORT_SYMBOL_GPL(xt_compat_unlock);
1054 DEFINE_PER_CPU(seqcount_t, xt_recseq);
1055 EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq);
1057 struct static_key xt_tee_enabled __read_mostly;
1058 EXPORT_SYMBOL_GPL(xt_tee_enabled);
1060 static int xt_jumpstack_alloc(struct xt_table_info *i)
1065 size = sizeof(void **) * nr_cpu_ids;
1066 if (size > PAGE_SIZE)
1067 i->jumpstack = vzalloc(size);
1069 i->jumpstack = kzalloc(size, GFP_KERNEL);
1070 if (i->jumpstack == NULL)
1073 /* ruleset without jumps -- no stack needed */
1074 if (i->stacksize == 0)
1077 /* Jumpstack needs to be able to record two full callchains, one
1078 * from the first rule set traversal, plus one table reentrancy
1079 * via -j TEE without clobbering the callchain that brought us to
1082 * This is done by allocating two jumpstacks per cpu, on reentry
1083 * the upper half of the stack is used.
1085 * see the jumpstack setup in ipt_do_table() for more details.
1087 size = sizeof(void *) * i->stacksize * 2u;
1088 for_each_possible_cpu(cpu) {
1089 if (size > PAGE_SIZE)
1090 i->jumpstack[cpu] = vmalloc_node(size,
1093 i->jumpstack[cpu] = kmalloc_node(size,
1094 GFP_KERNEL, cpu_to_node(cpu));
1095 if (i->jumpstack[cpu] == NULL)
1097 * Freeing will be done later on by the callers. The
1098 * chain is: xt_replace_table -> __do_replace ->
1099 * do_replace -> xt_free_table_info.
1107 struct xt_table_info *
1108 xt_replace_table(struct xt_table *table,
1109 unsigned int num_counters,
1110 struct xt_table_info *newinfo,
1113 struct xt_table_info *private;
1116 ret = xt_jumpstack_alloc(newinfo);
1122 /* Do the substitution. */
1124 private = table->private;
1126 /* Check inside lock: is the old number correct? */
1127 if (num_counters != private->number) {
1128 pr_debug("num_counters != table->private->number (%u/%u)\n",
1129 num_counters, private->number);
1135 newinfo->initial_entries = private->initial_entries;
1137 * Ensure contents of newinfo are visible before assigning to
1141 table->private = newinfo;
1143 /* make sure all cpus see new ->private value */
1147 * Even though table entries have now been swapped, other CPU's
1148 * may still be using the old entries. This is okay, because
1149 * resynchronization happens because of the locking done
1150 * during the get_counters() routine.
1155 if (audit_enabled) {
1156 struct audit_buffer *ab;
1158 ab = audit_log_start(current->audit_context, GFP_KERNEL,
1159 AUDIT_NETFILTER_CFG);
1161 audit_log_format(ab, "table=%s family=%u entries=%u",
1162 table->name, table->af,
1171 EXPORT_SYMBOL_GPL(xt_replace_table);
1173 struct xt_table *xt_register_table(struct net *net,
1174 const struct xt_table *input_table,
1175 struct xt_table_info *bootstrap,
1176 struct xt_table_info *newinfo)
1179 struct xt_table_info *private;
1180 struct xt_table *t, *table;
1182 /* Don't add one object to multiple lists. */
1183 table = kmemdup(input_table, sizeof(struct xt_table), GFP_KERNEL);
1189 mutex_lock(&xt[table->af].mutex);
1190 /* Don't autoload: we'd eat our tail... */
1191 list_for_each_entry(t, &net->xt.tables[table->af], list) {
1192 if (strcmp(t->name, table->name) == 0) {
1198 /* Simplifies replace_table code. */
1199 table->private = bootstrap;
1201 if (!xt_replace_table(table, 0, newinfo, &ret))
1204 private = table->private;
1205 pr_debug("table->private->number = %u\n", private->number);
1207 /* save number of initial entries */
1208 private->initial_entries = private->number;
1210 list_add(&table->list, &net->xt.tables[table->af]);
1211 mutex_unlock(&xt[table->af].mutex);
1215 mutex_unlock(&xt[table->af].mutex);
1218 return ERR_PTR(ret);
1220 EXPORT_SYMBOL_GPL(xt_register_table);
1222 void *xt_unregister_table(struct xt_table *table)
1224 struct xt_table_info *private;
1226 mutex_lock(&xt[table->af].mutex);
1227 private = table->private;
1228 list_del(&table->list);
1229 mutex_unlock(&xt[table->af].mutex);
1234 EXPORT_SYMBOL_GPL(xt_unregister_table);
1236 #ifdef CONFIG_PROC_FS
1237 struct xt_names_priv {
1238 struct seq_net_private p;
1241 static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos)
1243 struct xt_names_priv *priv = seq->private;
1244 struct net *net = seq_file_net(seq);
1245 u_int8_t af = priv->af;
1247 mutex_lock(&xt[af].mutex);
1248 return seq_list_start(&net->xt.tables[af], *pos);
1251 static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1253 struct xt_names_priv *priv = seq->private;
1254 struct net *net = seq_file_net(seq);
1255 u_int8_t af = priv->af;
1257 return seq_list_next(v, &net->xt.tables[af], pos);
1260 static void xt_table_seq_stop(struct seq_file *seq, void *v)
1262 struct xt_names_priv *priv = seq->private;
1263 u_int8_t af = priv->af;
1265 mutex_unlock(&xt[af].mutex);
1268 static int xt_table_seq_show(struct seq_file *seq, void *v)
1270 struct xt_table *table = list_entry(v, struct xt_table, list);
1273 seq_printf(seq, "%s\n", table->name);
1277 static const struct seq_operations xt_table_seq_ops = {
1278 .start = xt_table_seq_start,
1279 .next = xt_table_seq_next,
1280 .stop = xt_table_seq_stop,
1281 .show = xt_table_seq_show,
1284 static int xt_table_open(struct inode *inode, struct file *file)
1287 struct xt_names_priv *priv;
1289 ret = seq_open_net(inode, file, &xt_table_seq_ops,
1290 sizeof(struct xt_names_priv));
1292 priv = ((struct seq_file *)file->private_data)->private;
1293 priv->af = (unsigned long)PDE_DATA(inode);
1298 static const struct file_operations xt_table_ops = {
1299 .owner = THIS_MODULE,
1300 .open = xt_table_open,
1302 .llseek = seq_lseek,
1303 .release = seq_release_net,
1307 * Traverse state for ip{,6}_{tables,matches} for helping crossing
1308 * the multi-AF mutexes.
1310 struct nf_mttg_trav {
1311 struct list_head *head, *curr;
1312 uint8_t class, nfproto;
1317 MTTG_TRAV_NFP_UNSPEC,
1322 static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos,
1325 static const uint8_t next_class[] = {
1326 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC,
1327 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE,
1329 struct nf_mttg_trav *trav = seq->private;
1331 switch (trav->class) {
1332 case MTTG_TRAV_INIT:
1333 trav->class = MTTG_TRAV_NFP_UNSPEC;
1334 mutex_lock(&xt[NFPROTO_UNSPEC].mutex);
1335 trav->head = trav->curr = is_target ?
1336 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match;
1338 case MTTG_TRAV_NFP_UNSPEC:
1339 trav->curr = trav->curr->next;
1340 if (trav->curr != trav->head)
1342 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1343 mutex_lock(&xt[trav->nfproto].mutex);
1344 trav->head = trav->curr = is_target ?
1345 &xt[trav->nfproto].target : &xt[trav->nfproto].match;
1346 trav->class = next_class[trav->class];
1348 case MTTG_TRAV_NFP_SPEC:
1349 trav->curr = trav->curr->next;
1350 if (trav->curr != trav->head)
1352 /* fallthru, _stop will unlock */
1362 static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos,
1365 struct nf_mttg_trav *trav = seq->private;
1368 trav->class = MTTG_TRAV_INIT;
1369 for (j = 0; j < *pos; ++j)
1370 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL)
1375 static void xt_mttg_seq_stop(struct seq_file *seq, void *v)
1377 struct nf_mttg_trav *trav = seq->private;
1379 switch (trav->class) {
1380 case MTTG_TRAV_NFP_UNSPEC:
1381 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1383 case MTTG_TRAV_NFP_SPEC:
1384 mutex_unlock(&xt[trav->nfproto].mutex);
1389 static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos)
1391 return xt_mttg_seq_start(seq, pos, false);
1394 static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1396 return xt_mttg_seq_next(seq, v, ppos, false);
1399 static int xt_match_seq_show(struct seq_file *seq, void *v)
1401 const struct nf_mttg_trav *trav = seq->private;
1402 const struct xt_match *match;
1404 switch (trav->class) {
1405 case MTTG_TRAV_NFP_UNSPEC:
1406 case MTTG_TRAV_NFP_SPEC:
1407 if (trav->curr == trav->head)
1409 match = list_entry(trav->curr, struct xt_match, list);
1411 seq_printf(seq, "%s\n", match->name);
1416 static const struct seq_operations xt_match_seq_ops = {
1417 .start = xt_match_seq_start,
1418 .next = xt_match_seq_next,
1419 .stop = xt_mttg_seq_stop,
1420 .show = xt_match_seq_show,
1423 static int xt_match_open(struct inode *inode, struct file *file)
1425 struct nf_mttg_trav *trav;
1426 trav = __seq_open_private(file, &xt_match_seq_ops, sizeof(*trav));
1430 trav->nfproto = (unsigned long)PDE_DATA(inode);
1434 static const struct file_operations xt_match_ops = {
1435 .owner = THIS_MODULE,
1436 .open = xt_match_open,
1438 .llseek = seq_lseek,
1439 .release = seq_release_private,
1442 static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos)
1444 return xt_mttg_seq_start(seq, pos, true);
1447 static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1449 return xt_mttg_seq_next(seq, v, ppos, true);
1452 static int xt_target_seq_show(struct seq_file *seq, void *v)
1454 const struct nf_mttg_trav *trav = seq->private;
1455 const struct xt_target *target;
1457 switch (trav->class) {
1458 case MTTG_TRAV_NFP_UNSPEC:
1459 case MTTG_TRAV_NFP_SPEC:
1460 if (trav->curr == trav->head)
1462 target = list_entry(trav->curr, struct xt_target, list);
1464 seq_printf(seq, "%s\n", target->name);
1469 static const struct seq_operations xt_target_seq_ops = {
1470 .start = xt_target_seq_start,
1471 .next = xt_target_seq_next,
1472 .stop = xt_mttg_seq_stop,
1473 .show = xt_target_seq_show,
1476 static int xt_target_open(struct inode *inode, struct file *file)
1478 struct nf_mttg_trav *trav;
1479 trav = __seq_open_private(file, &xt_target_seq_ops, sizeof(*trav));
1483 trav->nfproto = (unsigned long)PDE_DATA(inode);
1487 static const struct file_operations xt_target_ops = {
1488 .owner = THIS_MODULE,
1489 .open = xt_target_open,
1491 .llseek = seq_lseek,
1492 .release = seq_release_private,
1495 #define FORMAT_TABLES "_tables_names"
1496 #define FORMAT_MATCHES "_tables_matches"
1497 #define FORMAT_TARGETS "_tables_targets"
1499 #endif /* CONFIG_PROC_FS */
1502 * xt_hook_link - set up hooks for a new table
1503 * @table: table with metadata needed to set up hooks
1504 * @fn: Hook function
1506 * This function will take care of creating and registering the necessary
1507 * Netfilter hooks for XT tables.
1509 struct nf_hook_ops *xt_hook_link(const struct xt_table *table, nf_hookfn *fn)
1511 unsigned int hook_mask = table->valid_hooks;
1512 uint8_t i, num_hooks = hweight32(hook_mask);
1514 struct nf_hook_ops *ops;
1517 ops = kmalloc(sizeof(*ops) * num_hooks, GFP_KERNEL);
1519 return ERR_PTR(-ENOMEM);
1521 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0;
1522 hook_mask >>= 1, ++hooknum) {
1523 if (!(hook_mask & 1))
1526 ops[i].pf = table->af;
1527 ops[i].hooknum = hooknum;
1528 ops[i].priority = table->priority;
1532 ret = nf_register_hooks(ops, num_hooks);
1535 return ERR_PTR(ret);
1540 EXPORT_SYMBOL_GPL(xt_hook_link);
1543 * xt_hook_unlink - remove hooks for a table
1544 * @ops: nf_hook_ops array as returned by nf_hook_link
1545 * @hook_mask: the very same mask that was passed to nf_hook_link
1547 void xt_hook_unlink(const struct xt_table *table, struct nf_hook_ops *ops)
1549 nf_unregister_hooks(ops, hweight32(table->valid_hooks));
1552 EXPORT_SYMBOL_GPL(xt_hook_unlink);
1554 int xt_proto_init(struct net *net, u_int8_t af)
1556 #ifdef CONFIG_PROC_FS
1557 char buf[XT_FUNCTION_MAXNAMELEN];
1558 struct proc_dir_entry *proc;
1561 if (af >= ARRAY_SIZE(xt_prefix))
1565 #ifdef CONFIG_PROC_FS
1566 strlcpy(buf, xt_prefix[af], sizeof(buf));
1567 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1568 proc = proc_create_data(buf, 0440, net->proc_net, &xt_table_ops,
1569 (void *)(unsigned long)af);
1573 strlcpy(buf, xt_prefix[af], sizeof(buf));
1574 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1575 proc = proc_create_data(buf, 0440, net->proc_net, &xt_match_ops,
1576 (void *)(unsigned long)af);
1578 goto out_remove_tables;
1580 strlcpy(buf, xt_prefix[af], sizeof(buf));
1581 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1582 proc = proc_create_data(buf, 0440, net->proc_net, &xt_target_ops,
1583 (void *)(unsigned long)af);
1585 goto out_remove_matches;
1590 #ifdef CONFIG_PROC_FS
1592 strlcpy(buf, xt_prefix[af], sizeof(buf));
1593 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1594 remove_proc_entry(buf, net->proc_net);
1597 strlcpy(buf, xt_prefix[af], sizeof(buf));
1598 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1599 remove_proc_entry(buf, net->proc_net);
1604 EXPORT_SYMBOL_GPL(xt_proto_init);
1606 void xt_proto_fini(struct net *net, u_int8_t af)
1608 #ifdef CONFIG_PROC_FS
1609 char buf[XT_FUNCTION_MAXNAMELEN];
1611 strlcpy(buf, xt_prefix[af], sizeof(buf));
1612 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1613 remove_proc_entry(buf, net->proc_net);
1615 strlcpy(buf, xt_prefix[af], sizeof(buf));
1616 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1617 remove_proc_entry(buf, net->proc_net);
1619 strlcpy(buf, xt_prefix[af], sizeof(buf));
1620 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1621 remove_proc_entry(buf, net->proc_net);
1622 #endif /*CONFIG_PROC_FS*/
1624 EXPORT_SYMBOL_GPL(xt_proto_fini);
1627 * xt_percpu_counter_alloc - allocate x_tables rule counter
1629 * @state: pointer to xt_percpu allocation state
1630 * @counter: pointer to counter struct inside the ip(6)/arpt_entry struct
1632 * On SMP, the packet counter [ ip(6)t_entry->counters.pcnt ] will then
1633 * contain the address of the real (percpu) counter.
1635 * Rule evaluation needs to use xt_get_this_cpu_counter() helper
1636 * to fetch the real percpu counter.
1638 * To speed up allocation and improve data locality, a 4kb block is
1641 * xt_percpu_counter_alloc_state contains the base address of the
1642 * allocated page and the current sub-offset.
1644 * returns false on error.
1646 bool xt_percpu_counter_alloc(struct xt_percpu_counter_alloc_state *state,
1647 struct xt_counters *counter)
1649 BUILD_BUG_ON(XT_PCPU_BLOCK_SIZE < (sizeof(*counter) * 2));
1651 if (nr_cpu_ids <= 1)
1655 state->mem = __alloc_percpu(XT_PCPU_BLOCK_SIZE,
1656 XT_PCPU_BLOCK_SIZE);
1660 counter->pcnt = (__force unsigned long)(state->mem + state->off);
1661 state->off += sizeof(*counter);
1662 if (state->off > (XT_PCPU_BLOCK_SIZE - sizeof(*counter))) {
1668 EXPORT_SYMBOL_GPL(xt_percpu_counter_alloc);
1670 void xt_percpu_counter_free(struct xt_counters *counters)
1672 unsigned long pcnt = counters->pcnt;
1674 if (nr_cpu_ids > 1 && (pcnt & (XT_PCPU_BLOCK_SIZE - 1)) == 0)
1675 free_percpu((void __percpu *)pcnt);
1677 EXPORT_SYMBOL_GPL(xt_percpu_counter_free);
1679 static int __net_init xt_net_init(struct net *net)
1683 for (i = 0; i < NFPROTO_NUMPROTO; i++)
1684 INIT_LIST_HEAD(&net->xt.tables[i]);
1688 static struct pernet_operations xt_net_ops = {
1689 .init = xt_net_init,
1692 static int __init xt_init(void)
1697 for_each_possible_cpu(i) {
1698 seqcount_init(&per_cpu(xt_recseq, i));
1701 xt = kcalloc(NFPROTO_NUMPROTO, sizeof(struct xt_af), GFP_KERNEL);
1705 for (i = 0; i < NFPROTO_NUMPROTO; i++) {
1706 mutex_init(&xt[i].mutex);
1707 #ifdef CONFIG_COMPAT
1708 mutex_init(&xt[i].compat_mutex);
1709 xt[i].compat_tab = NULL;
1711 INIT_LIST_HEAD(&xt[i].target);
1712 INIT_LIST_HEAD(&xt[i].match);
1714 rv = register_pernet_subsys(&xt_net_ops);
1720 static void __exit xt_fini(void)
1722 unregister_pernet_subsys(&xt_net_ops);
1726 module_init(xt_init);
1727 module_exit(xt_fini);