1 // SPDX-License-Identifier: GPL-2.0-only
3 * x_tables core - Backend for {ip,ip6,arp}_tables
5 * Copyright (C) 2006-2006 Harald Welte <laforge@netfilter.org>
6 * Copyright (C) 2006-2012 Patrick McHardy <kaber@trash.net>
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>
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>
23 #include <linux/slab.h>
24 #include <linux/audit.h>
25 #include <linux/user_namespace.h>
26 #include <net/net_namespace.h>
28 #include <linux/netfilter/x_tables.h>
29 #include <linux/netfilter_arp.h>
30 #include <linux/netfilter_ipv4/ip_tables.h>
31 #include <linux/netfilter_ipv6/ip6_tables.h>
32 #include <linux/netfilter_arp/arp_tables.h>
34 MODULE_LICENSE("GPL");
35 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
36 MODULE_DESCRIPTION("{ip,ip6,arp,eb}_tables backend module");
38 #define XT_PCPU_BLOCK_SIZE 4096
39 #define XT_MAX_TABLE_SIZE (512 * 1024 * 1024)
42 unsigned int offset; /* offset in kernel */
43 int delta; /* delta in 32bit user land */
48 struct list_head match;
49 struct list_head target;
51 struct mutex compat_mutex;
52 struct compat_delta *compat_tab;
53 unsigned int number; /* number of slots in compat_tab[] */
54 unsigned int cur; /* number of used slots in compat_tab[] */
58 static struct xt_af *xt;
60 static const char *const xt_prefix[NFPROTO_NUMPROTO] = {
61 [NFPROTO_UNSPEC] = "x",
62 [NFPROTO_IPV4] = "ip",
63 [NFPROTO_ARP] = "arp",
64 [NFPROTO_BRIDGE] = "eb",
65 [NFPROTO_IPV6] = "ip6",
68 /* Registration hooks for targets. */
69 int xt_register_target(struct xt_target *target)
71 u_int8_t af = target->family;
73 mutex_lock(&xt[af].mutex);
74 list_add(&target->list, &xt[af].target);
75 mutex_unlock(&xt[af].mutex);
78 EXPORT_SYMBOL(xt_register_target);
81 xt_unregister_target(struct xt_target *target)
83 u_int8_t af = target->family;
85 mutex_lock(&xt[af].mutex);
86 list_del(&target->list);
87 mutex_unlock(&xt[af].mutex);
89 EXPORT_SYMBOL(xt_unregister_target);
92 xt_register_targets(struct xt_target *target, unsigned int n)
97 for (i = 0; i < n; i++) {
98 err = xt_register_target(&target[i]);
106 xt_unregister_targets(target, i);
109 EXPORT_SYMBOL(xt_register_targets);
112 xt_unregister_targets(struct xt_target *target, unsigned int n)
115 xt_unregister_target(&target[n]);
117 EXPORT_SYMBOL(xt_unregister_targets);
119 int xt_register_match(struct xt_match *match)
121 u_int8_t af = match->family;
123 mutex_lock(&xt[af].mutex);
124 list_add(&match->list, &xt[af].match);
125 mutex_unlock(&xt[af].mutex);
128 EXPORT_SYMBOL(xt_register_match);
131 xt_unregister_match(struct xt_match *match)
133 u_int8_t af = match->family;
135 mutex_lock(&xt[af].mutex);
136 list_del(&match->list);
137 mutex_unlock(&xt[af].mutex);
139 EXPORT_SYMBOL(xt_unregister_match);
142 xt_register_matches(struct xt_match *match, unsigned int n)
147 for (i = 0; i < n; i++) {
148 err = xt_register_match(&match[i]);
156 xt_unregister_matches(match, i);
159 EXPORT_SYMBOL(xt_register_matches);
162 xt_unregister_matches(struct xt_match *match, unsigned int n)
165 xt_unregister_match(&match[n]);
167 EXPORT_SYMBOL(xt_unregister_matches);
171 * These are weird, but module loading must not be done with mutex
172 * held (since they will register), and we have to have a single
176 /* Find match, grabs ref. Returns ERR_PTR() on error. */
177 struct xt_match *xt_find_match(u8 af, const char *name, u8 revision)
182 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN)
183 return ERR_PTR(-EINVAL);
185 mutex_lock(&xt[af].mutex);
186 list_for_each_entry(m, &xt[af].match, list) {
187 if (strcmp(m->name, name) == 0) {
188 if (m->revision == revision) {
189 if (try_module_get(m->me)) {
190 mutex_unlock(&xt[af].mutex);
194 err = -EPROTOTYPE; /* Found something. */
197 mutex_unlock(&xt[af].mutex);
199 if (af != NFPROTO_UNSPEC)
200 /* Try searching again in the family-independent list */
201 return xt_find_match(NFPROTO_UNSPEC, name, revision);
205 EXPORT_SYMBOL(xt_find_match);
208 xt_request_find_match(uint8_t nfproto, const char *name, uint8_t revision)
210 struct xt_match *match;
212 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN)
213 return ERR_PTR(-EINVAL);
215 match = xt_find_match(nfproto, name, revision);
217 request_module("%st_%s", xt_prefix[nfproto], name);
218 match = xt_find_match(nfproto, name, revision);
223 EXPORT_SYMBOL_GPL(xt_request_find_match);
225 /* Find target, grabs ref. Returns ERR_PTR() on error. */
226 static struct xt_target *xt_find_target(u8 af, const char *name, u8 revision)
231 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN)
232 return ERR_PTR(-EINVAL);
234 mutex_lock(&xt[af].mutex);
235 list_for_each_entry(t, &xt[af].target, list) {
236 if (strcmp(t->name, name) == 0) {
237 if (t->revision == revision) {
238 if (try_module_get(t->me)) {
239 mutex_unlock(&xt[af].mutex);
243 err = -EPROTOTYPE; /* Found something. */
246 mutex_unlock(&xt[af].mutex);
248 if (af != NFPROTO_UNSPEC)
249 /* Try searching again in the family-independent list */
250 return xt_find_target(NFPROTO_UNSPEC, name, revision);
255 struct xt_target *xt_request_find_target(u8 af, const char *name, u8 revision)
257 struct xt_target *target;
259 if (strnlen(name, XT_EXTENSION_MAXNAMELEN) == XT_EXTENSION_MAXNAMELEN)
260 return ERR_PTR(-EINVAL);
262 target = xt_find_target(af, name, revision);
263 if (IS_ERR(target)) {
264 request_module("%st_%s", xt_prefix[af], name);
265 target = xt_find_target(af, name, revision);
270 EXPORT_SYMBOL_GPL(xt_request_find_target);
273 static int xt_obj_to_user(u16 __user *psize, u16 size,
274 void __user *pname, const char *name,
275 u8 __user *prev, u8 rev)
277 if (put_user(size, psize))
279 if (copy_to_user(pname, name, strlen(name) + 1))
281 if (put_user(rev, prev))
287 #define XT_OBJ_TO_USER(U, K, TYPE, C_SIZE) \
288 xt_obj_to_user(&U->u.TYPE##_size, C_SIZE ? : K->u.TYPE##_size, \
289 U->u.user.name, K->u.kernel.TYPE->name, \
290 &U->u.user.revision, K->u.kernel.TYPE->revision)
292 int xt_data_to_user(void __user *dst, const void *src,
293 int usersize, int size, int aligned_size)
295 usersize = usersize ? : size;
296 if (copy_to_user(dst, src, usersize))
298 if (usersize != aligned_size &&
299 clear_user(dst + usersize, aligned_size - usersize))
304 EXPORT_SYMBOL_GPL(xt_data_to_user);
306 #define XT_DATA_TO_USER(U, K, TYPE) \
307 xt_data_to_user(U->data, K->data, \
308 K->u.kernel.TYPE->usersize, \
309 K->u.kernel.TYPE->TYPE##size, \
310 XT_ALIGN(K->u.kernel.TYPE->TYPE##size))
312 int xt_match_to_user(const struct xt_entry_match *m,
313 struct xt_entry_match __user *u)
315 return XT_OBJ_TO_USER(u, m, match, 0) ||
316 XT_DATA_TO_USER(u, m, match);
318 EXPORT_SYMBOL_GPL(xt_match_to_user);
320 int xt_target_to_user(const struct xt_entry_target *t,
321 struct xt_entry_target __user *u)
323 return XT_OBJ_TO_USER(u, t, target, 0) ||
324 XT_DATA_TO_USER(u, t, target);
326 EXPORT_SYMBOL_GPL(xt_target_to_user);
328 static int match_revfn(u8 af, const char *name, u8 revision, int *bestp)
330 const struct xt_match *m;
333 list_for_each_entry(m, &xt[af].match, list) {
334 if (strcmp(m->name, name) == 0) {
335 if (m->revision > *bestp)
336 *bestp = m->revision;
337 if (m->revision == revision)
342 if (af != NFPROTO_UNSPEC && !have_rev)
343 return match_revfn(NFPROTO_UNSPEC, name, revision, bestp);
348 static int target_revfn(u8 af, const char *name, u8 revision, int *bestp)
350 const struct xt_target *t;
353 list_for_each_entry(t, &xt[af].target, list) {
354 if (strcmp(t->name, name) == 0) {
355 if (t->revision > *bestp)
356 *bestp = t->revision;
357 if (t->revision == revision)
362 if (af != NFPROTO_UNSPEC && !have_rev)
363 return target_revfn(NFPROTO_UNSPEC, name, revision, bestp);
368 /* Returns true or false (if no such extension at all) */
369 int xt_find_revision(u8 af, const char *name, u8 revision, int target,
372 int have_rev, best = -1;
374 mutex_lock(&xt[af].mutex);
376 have_rev = target_revfn(af, name, revision, &best);
378 have_rev = match_revfn(af, name, revision, &best);
379 mutex_unlock(&xt[af].mutex);
381 /* Nothing at all? Return 0 to try loading module. */
389 *err = -EPROTONOSUPPORT;
392 EXPORT_SYMBOL_GPL(xt_find_revision);
395 textify_hooks(char *buf, size_t size, unsigned int mask, uint8_t nfproto)
397 static const char *const inetbr_names[] = {
398 "PREROUTING", "INPUT", "FORWARD",
399 "OUTPUT", "POSTROUTING", "BROUTING",
401 static const char *const arp_names[] = {
402 "INPUT", "FORWARD", "OUTPUT",
404 const char *const *names;
410 names = (nfproto == NFPROTO_ARP) ? arp_names : inetbr_names;
411 max = (nfproto == NFPROTO_ARP) ? ARRAY_SIZE(arp_names) :
412 ARRAY_SIZE(inetbr_names);
414 for (i = 0; i < max; ++i) {
415 if (!(mask & (1 << i)))
417 res = snprintf(p, size, "%s%s", np ? "/" : "", names[i]);
429 * xt_check_proc_name - check that name is suitable for /proc file creation
431 * @name: file name candidate
432 * @size: length of buffer
434 * some x_tables modules wish to create a file in /proc.
435 * This function makes sure that the name is suitable for this
436 * purpose, it checks that name is NUL terminated and isn't a 'special'
439 * returns negative number on error or 0 if name is useable.
441 int xt_check_proc_name(const char *name, unsigned int size)
446 if (strnlen(name, size) == size)
447 return -ENAMETOOLONG;
449 if (strcmp(name, ".") == 0 ||
450 strcmp(name, "..") == 0 ||
456 EXPORT_SYMBOL(xt_check_proc_name);
458 int xt_check_match(struct xt_mtchk_param *par,
459 unsigned int size, u16 proto, bool inv_proto)
463 if (XT_ALIGN(par->match->matchsize) != size &&
464 par->match->matchsize != -1) {
466 * ebt_among is exempt from centralized matchsize checking
467 * because it uses a dynamic-size data set.
469 pr_err_ratelimited("%s_tables: %s.%u match: invalid size %u (kernel) != (user) %u\n",
470 xt_prefix[par->family], par->match->name,
471 par->match->revision,
472 XT_ALIGN(par->match->matchsize), size);
475 if (par->match->table != NULL &&
476 strcmp(par->match->table, par->table) != 0) {
477 pr_info_ratelimited("%s_tables: %s match: only valid in %s table, not %s\n",
478 xt_prefix[par->family], par->match->name,
479 par->match->table, par->table);
482 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) {
483 char used[64], allow[64];
485 pr_info_ratelimited("%s_tables: %s match: used from hooks %s, but only valid from %s\n",
486 xt_prefix[par->family], par->match->name,
487 textify_hooks(used, sizeof(used),
488 par->hook_mask, par->family),
489 textify_hooks(allow, sizeof(allow),
494 if (par->match->proto && (par->match->proto != proto || inv_proto)) {
495 pr_info_ratelimited("%s_tables: %s match: only valid for protocol %u\n",
496 xt_prefix[par->family], par->match->name,
500 if (par->match->checkentry != NULL) {
501 ret = par->match->checkentry(par);
505 /* Flag up potential errors. */
510 EXPORT_SYMBOL_GPL(xt_check_match);
512 /** xt_check_entry_match - check that matches end before start of target
514 * @match: beginning of xt_entry_match
515 * @target: beginning of this rules target (alleged end of matches)
516 * @alignment: alignment requirement of match structures
518 * Validates that all matches add up to the beginning of the target,
519 * and that each match covers at least the base structure size.
521 * Return: 0 on success, negative errno on failure.
523 static int xt_check_entry_match(const char *match, const char *target,
524 const size_t alignment)
526 const struct xt_entry_match *pos;
527 int length = target - match;
529 if (length == 0) /* no matches */
532 pos = (struct xt_entry_match *)match;
534 if ((unsigned long)pos % alignment)
537 if (length < (int)sizeof(struct xt_entry_match))
540 if (pos->u.match_size < sizeof(struct xt_entry_match))
543 if (pos->u.match_size > length)
546 length -= pos->u.match_size;
547 pos = ((void *)((char *)(pos) + (pos)->u.match_size));
548 } while (length > 0);
553 /** xt_check_table_hooks - check hook entry points are sane
555 * @info xt_table_info to check
556 * @valid_hooks - hook entry points that we can enter from
558 * Validates that the hook entry and underflows points are set up.
560 * Return: 0 on success, negative errno on failure.
562 int xt_check_table_hooks(const struct xt_table_info *info, unsigned int valid_hooks)
564 const char *err = "unsorted underflow";
565 unsigned int i, max_uflow, max_entry;
566 bool check_hooks = false;
568 BUILD_BUG_ON(ARRAY_SIZE(info->hook_entry) != ARRAY_SIZE(info->underflow));
573 for (i = 0; i < ARRAY_SIZE(info->hook_entry); i++) {
574 if (!(valid_hooks & (1 << i)))
577 if (info->hook_entry[i] == 0xFFFFFFFF)
579 if (info->underflow[i] == 0xFFFFFFFF)
583 if (max_uflow > info->underflow[i])
586 if (max_uflow == info->underflow[i]) {
587 err = "duplicate underflow";
590 if (max_entry > info->hook_entry[i]) {
591 err = "unsorted entry";
594 if (max_entry == info->hook_entry[i]) {
595 err = "duplicate entry";
599 max_entry = info->hook_entry[i];
600 max_uflow = info->underflow[i];
606 pr_err_ratelimited("%s at hook %d\n", err, i);
609 EXPORT_SYMBOL(xt_check_table_hooks);
611 static bool verdict_ok(int verdict)
617 int v = -verdict - 1;
619 if (verdict == XT_RETURN)
623 case NF_ACCEPT: return true;
624 case NF_DROP: return true;
625 case NF_QUEUE: return true;
636 static bool error_tg_ok(unsigned int usersize, unsigned int kernsize,
637 const char *msg, unsigned int msglen)
639 return usersize == kernsize && strnlen(msg, msglen) < msglen;
643 int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta)
645 struct xt_af *xp = &xt[af];
647 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex));
649 if (WARN_ON(!xp->compat_tab))
652 if (xp->cur >= xp->number)
656 delta += xp->compat_tab[xp->cur - 1].delta;
657 xp->compat_tab[xp->cur].offset = offset;
658 xp->compat_tab[xp->cur].delta = delta;
662 EXPORT_SYMBOL_GPL(xt_compat_add_offset);
664 void xt_compat_flush_offsets(u_int8_t af)
666 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex));
668 if (xt[af].compat_tab) {
669 vfree(xt[af].compat_tab);
670 xt[af].compat_tab = NULL;
675 EXPORT_SYMBOL_GPL(xt_compat_flush_offsets);
677 int xt_compat_calc_jump(u_int8_t af, unsigned int offset)
679 struct compat_delta *tmp = xt[af].compat_tab;
680 int mid, left = 0, right = xt[af].cur - 1;
682 while (left <= right) {
683 mid = (left + right) >> 1;
684 if (offset > tmp[mid].offset)
686 else if (offset < tmp[mid].offset)
689 return mid ? tmp[mid - 1].delta : 0;
691 return left ? tmp[left - 1].delta : 0;
693 EXPORT_SYMBOL_GPL(xt_compat_calc_jump);
695 int xt_compat_init_offsets(u8 af, unsigned int number)
699 WARN_ON(!mutex_is_locked(&xt[af].compat_mutex));
701 if (!number || number > (INT_MAX / sizeof(struct compat_delta)))
704 if (WARN_ON(xt[af].compat_tab))
707 mem = sizeof(struct compat_delta) * number;
708 if (mem > XT_MAX_TABLE_SIZE)
711 xt[af].compat_tab = vmalloc(mem);
712 if (!xt[af].compat_tab)
715 xt[af].number = number;
720 EXPORT_SYMBOL(xt_compat_init_offsets);
722 int xt_compat_match_offset(const struct xt_match *match)
724 u_int16_t csize = match->compatsize ? : match->matchsize;
725 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize);
727 EXPORT_SYMBOL_GPL(xt_compat_match_offset);
729 void xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr,
732 const struct xt_match *match = m->u.kernel.match;
733 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m;
734 int pad, off = xt_compat_match_offset(match);
735 u_int16_t msize = cm->u.user.match_size;
736 char name[sizeof(m->u.user.name)];
739 memcpy(m, cm, sizeof(*cm));
740 if (match->compat_from_user)
741 match->compat_from_user(m->data, cm->data);
743 memcpy(m->data, cm->data, msize - sizeof(*cm));
744 pad = XT_ALIGN(match->matchsize) - match->matchsize;
746 memset(m->data + match->matchsize, 0, pad);
749 m->u.user.match_size = msize;
750 strlcpy(name, match->name, sizeof(name));
751 module_put(match->me);
752 strncpy(m->u.user.name, name, sizeof(m->u.user.name));
757 EXPORT_SYMBOL_GPL(xt_compat_match_from_user);
759 #define COMPAT_XT_DATA_TO_USER(U, K, TYPE, C_SIZE) \
760 xt_data_to_user(U->data, K->data, \
761 K->u.kernel.TYPE->usersize, \
763 COMPAT_XT_ALIGN(C_SIZE))
765 int xt_compat_match_to_user(const struct xt_entry_match *m,
766 void __user **dstptr, unsigned int *size)
768 const struct xt_match *match = m->u.kernel.match;
769 struct compat_xt_entry_match __user *cm = *dstptr;
770 int off = xt_compat_match_offset(match);
771 u_int16_t msize = m->u.user.match_size - off;
773 if (XT_OBJ_TO_USER(cm, m, match, msize))
776 if (match->compat_to_user) {
777 if (match->compat_to_user((void __user *)cm->data, m->data))
780 if (COMPAT_XT_DATA_TO_USER(cm, m, match, msize - sizeof(*cm)))
788 EXPORT_SYMBOL_GPL(xt_compat_match_to_user);
790 /* non-compat version may have padding after verdict */
791 struct compat_xt_standard_target {
792 struct compat_xt_entry_target t;
793 compat_uint_t verdict;
796 struct compat_xt_error_target {
797 struct compat_xt_entry_target t;
798 char errorname[XT_FUNCTION_MAXNAMELEN];
801 int xt_compat_check_entry_offsets(const void *base, const char *elems,
802 unsigned int target_offset,
803 unsigned int next_offset)
805 long size_of_base_struct = elems - (const char *)base;
806 const struct compat_xt_entry_target *t;
807 const char *e = base;
809 if (target_offset < size_of_base_struct)
812 if (target_offset + sizeof(*t) > next_offset)
815 t = (void *)(e + target_offset);
816 if (t->u.target_size < sizeof(*t))
819 if (target_offset + t->u.target_size > next_offset)
822 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) {
823 const struct compat_xt_standard_target *st = (const void *)t;
825 if (COMPAT_XT_ALIGN(target_offset + sizeof(*st)) != next_offset)
828 if (!verdict_ok(st->verdict))
830 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) {
831 const struct compat_xt_error_target *et = (const void *)t;
833 if (!error_tg_ok(t->u.target_size, sizeof(*et),
834 et->errorname, sizeof(et->errorname)))
838 /* compat_xt_entry match has less strict alignment requirements,
839 * otherwise they are identical. In case of padding differences
840 * we need to add compat version of xt_check_entry_match.
842 BUILD_BUG_ON(sizeof(struct compat_xt_entry_match) != sizeof(struct xt_entry_match));
844 return xt_check_entry_match(elems, base + target_offset,
845 __alignof__(struct compat_xt_entry_match));
847 EXPORT_SYMBOL(xt_compat_check_entry_offsets);
848 #endif /* CONFIG_COMPAT */
851 * xt_check_entry_offsets - validate arp/ip/ip6t_entry
853 * @base: pointer to arp/ip/ip6t_entry
854 * @elems: pointer to first xt_entry_match, i.e. ip(6)t_entry->elems
855 * @target_offset: the arp/ip/ip6_t->target_offset
856 * @next_offset: the arp/ip/ip6_t->next_offset
858 * validates that target_offset and next_offset are sane and that all
859 * match sizes (if any) align with the target offset.
861 * This function does not validate the targets or matches themselves, it
862 * only tests that all the offsets and sizes are correct, that all
863 * match structures are aligned, and that the last structure ends where
864 * the target structure begins.
866 * Also see xt_compat_check_entry_offsets for CONFIG_COMPAT version.
868 * The arp/ip/ip6t_entry structure @base must have passed following tests:
869 * - it must point to a valid memory location
870 * - base to base + next_offset must be accessible, i.e. not exceed allocated
873 * A well-formed entry looks like this:
875 * ip(6)t_entry match [mtdata] match [mtdata] target [tgdata] ip(6)t_entry
876 * e->elems[]-----' | |
880 * target_offset---------------------------------' |
881 * next_offset---------------------------------------------------'
883 * elems[]: flexible array member at end of ip(6)/arpt_entry struct.
884 * This is where matches (if any) and the target reside.
885 * target_offset: beginning of target.
886 * next_offset: start of the next rule; also: size of this rule.
887 * Since targets have a minimum size, target_offset + minlen <= next_offset.
889 * Every match stores its size, sum of sizes must not exceed target_offset.
891 * Return: 0 on success, negative errno on failure.
893 int xt_check_entry_offsets(const void *base,
895 unsigned int target_offset,
896 unsigned int next_offset)
898 long size_of_base_struct = elems - (const char *)base;
899 const struct xt_entry_target *t;
900 const char *e = base;
902 /* target start is within the ip/ip6/arpt_entry struct */
903 if (target_offset < size_of_base_struct)
906 if (target_offset + sizeof(*t) > next_offset)
909 t = (void *)(e + target_offset);
910 if (t->u.target_size < sizeof(*t))
913 if (target_offset + t->u.target_size > next_offset)
916 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0) {
917 const struct xt_standard_target *st = (const void *)t;
919 if (XT_ALIGN(target_offset + sizeof(*st)) != next_offset)
922 if (!verdict_ok(st->verdict))
924 } else if (strcmp(t->u.user.name, XT_ERROR_TARGET) == 0) {
925 const struct xt_error_target *et = (const void *)t;
927 if (!error_tg_ok(t->u.target_size, sizeof(*et),
928 et->errorname, sizeof(et->errorname)))
932 return xt_check_entry_match(elems, base + target_offset,
933 __alignof__(struct xt_entry_match));
935 EXPORT_SYMBOL(xt_check_entry_offsets);
938 * xt_alloc_entry_offsets - allocate array to store rule head offsets
940 * @size: number of entries
942 * Return: NULL or zeroed kmalloc'd or vmalloc'd array
944 unsigned int *xt_alloc_entry_offsets(unsigned int size)
946 if (size > XT_MAX_TABLE_SIZE / sizeof(unsigned int))
949 return kvcalloc(size, sizeof(unsigned int), GFP_KERNEL);
952 EXPORT_SYMBOL(xt_alloc_entry_offsets);
955 * xt_find_jump_offset - check if target is a valid jump offset
957 * @offsets: array containing all valid rule start offsets of a rule blob
958 * @target: the jump target to search for
959 * @size: entries in @offset
961 bool xt_find_jump_offset(const unsigned int *offsets,
962 unsigned int target, unsigned int size)
964 int m, low = 0, hi = size;
969 if (offsets[m] > target)
971 else if (offsets[m] < target)
979 EXPORT_SYMBOL(xt_find_jump_offset);
981 int xt_check_target(struct xt_tgchk_param *par,
982 unsigned int size, u16 proto, bool inv_proto)
986 if (XT_ALIGN(par->target->targetsize) != size) {
987 pr_err_ratelimited("%s_tables: %s.%u target: invalid size %u (kernel) != (user) %u\n",
988 xt_prefix[par->family], par->target->name,
989 par->target->revision,
990 XT_ALIGN(par->target->targetsize), size);
993 if (par->target->table != NULL &&
994 strcmp(par->target->table, par->table) != 0) {
995 pr_info_ratelimited("%s_tables: %s target: only valid in %s table, not %s\n",
996 xt_prefix[par->family], par->target->name,
997 par->target->table, par->table);
1000 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) {
1001 char used[64], allow[64];
1003 pr_info_ratelimited("%s_tables: %s target: used from hooks %s, but only usable from %s\n",
1004 xt_prefix[par->family], par->target->name,
1005 textify_hooks(used, sizeof(used),
1006 par->hook_mask, par->family),
1007 textify_hooks(allow, sizeof(allow),
1012 if (par->target->proto && (par->target->proto != proto || inv_proto)) {
1013 pr_info_ratelimited("%s_tables: %s target: only valid for protocol %u\n",
1014 xt_prefix[par->family], par->target->name,
1015 par->target->proto);
1018 if (par->target->checkentry != NULL) {
1019 ret = par->target->checkentry(par);
1023 /* Flag up potential errors. */
1028 EXPORT_SYMBOL_GPL(xt_check_target);
1031 * xt_copy_counters - copy counters and metadata from a sockptr_t
1034 * @len: alleged size of userspace memory
1035 * @info: where to store the xt_counters_info metadata
1037 * Copies counter meta data from @user and stores it in @info.
1039 * vmallocs memory to hold the counters, then copies the counter data
1040 * from @user to the new memory and returns a pointer to it.
1042 * If called from a compat syscall, @info gets converted automatically to the
1043 * 64bit representation.
1045 * The metadata associated with the counters is stored in @info.
1047 * Return: returns pointer that caller has to test via IS_ERR().
1048 * If IS_ERR is false, caller has to vfree the pointer.
1050 void *xt_copy_counters(sockptr_t arg, unsigned int len,
1051 struct xt_counters_info *info)
1057 #ifdef CONFIG_COMPAT
1058 if (in_compat_syscall()) {
1059 /* structures only differ in size due to alignment */
1060 struct compat_xt_counters_info compat_tmp;
1062 if (len <= sizeof(compat_tmp))
1063 return ERR_PTR(-EINVAL);
1065 len -= sizeof(compat_tmp);
1066 if (copy_from_sockptr(&compat_tmp, arg, sizeof(compat_tmp)) != 0)
1067 return ERR_PTR(-EFAULT);
1069 memcpy(info->name, compat_tmp.name, sizeof(info->name) - 1);
1070 info->num_counters = compat_tmp.num_counters;
1071 offset = sizeof(compat_tmp);
1075 if (len <= sizeof(*info))
1076 return ERR_PTR(-EINVAL);
1078 len -= sizeof(*info);
1079 if (copy_from_sockptr(info, arg, sizeof(*info)) != 0)
1080 return ERR_PTR(-EFAULT);
1082 offset = sizeof(*info);
1084 info->name[sizeof(info->name) - 1] = '\0';
1086 size = sizeof(struct xt_counters);
1087 size *= info->num_counters;
1089 if (size != (u64)len)
1090 return ERR_PTR(-EINVAL);
1094 return ERR_PTR(-ENOMEM);
1096 if (copy_from_sockptr_offset(mem, arg, offset, len) == 0)
1100 return ERR_PTR(-EFAULT);
1102 EXPORT_SYMBOL_GPL(xt_copy_counters);
1104 #ifdef CONFIG_COMPAT
1105 int xt_compat_target_offset(const struct xt_target *target)
1107 u_int16_t csize = target->compatsize ? : target->targetsize;
1108 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize);
1110 EXPORT_SYMBOL_GPL(xt_compat_target_offset);
1112 void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr,
1115 const struct xt_target *target = t->u.kernel.target;
1116 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t;
1117 int pad, off = xt_compat_target_offset(target);
1118 u_int16_t tsize = ct->u.user.target_size;
1119 char name[sizeof(t->u.user.name)];
1122 memcpy(t, ct, sizeof(*ct));
1123 if (target->compat_from_user)
1124 target->compat_from_user(t->data, ct->data);
1126 memcpy(t->data, ct->data, tsize - sizeof(*ct));
1127 pad = XT_ALIGN(target->targetsize) - target->targetsize;
1129 memset(t->data + target->targetsize, 0, pad);
1132 t->u.user.target_size = tsize;
1133 strlcpy(name, target->name, sizeof(name));
1134 module_put(target->me);
1135 strncpy(t->u.user.name, name, sizeof(t->u.user.name));
1140 EXPORT_SYMBOL_GPL(xt_compat_target_from_user);
1142 int xt_compat_target_to_user(const struct xt_entry_target *t,
1143 void __user **dstptr, unsigned int *size)
1145 const struct xt_target *target = t->u.kernel.target;
1146 struct compat_xt_entry_target __user *ct = *dstptr;
1147 int off = xt_compat_target_offset(target);
1148 u_int16_t tsize = t->u.user.target_size - off;
1150 if (XT_OBJ_TO_USER(ct, t, target, tsize))
1153 if (target->compat_to_user) {
1154 if (target->compat_to_user((void __user *)ct->data, t->data))
1157 if (COMPAT_XT_DATA_TO_USER(ct, t, target, tsize - sizeof(*ct)))
1165 EXPORT_SYMBOL_GPL(xt_compat_target_to_user);
1168 struct xt_table_info *xt_alloc_table_info(unsigned int size)
1170 struct xt_table_info *info = NULL;
1171 size_t sz = sizeof(*info) + size;
1173 if (sz < sizeof(*info) || sz >= XT_MAX_TABLE_SIZE)
1176 info = kvmalloc(sz, GFP_KERNEL_ACCOUNT);
1180 memset(info, 0, sizeof(*info));
1184 EXPORT_SYMBOL(xt_alloc_table_info);
1186 void xt_free_table_info(struct xt_table_info *info)
1190 if (info->jumpstack != NULL) {
1191 for_each_possible_cpu(cpu)
1192 kvfree(info->jumpstack[cpu]);
1193 kvfree(info->jumpstack);
1198 EXPORT_SYMBOL(xt_free_table_info);
1200 /* Find table by name, grabs mutex & ref. Returns ERR_PTR on error. */
1201 struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af,
1204 struct xt_table *t, *found = NULL;
1206 mutex_lock(&xt[af].mutex);
1207 list_for_each_entry(t, &net->xt.tables[af], list)
1208 if (strcmp(t->name, name) == 0 && try_module_get(t->me))
1211 if (net == &init_net)
1214 /* Table doesn't exist in this netns, re-try init */
1215 list_for_each_entry(t, &init_net.xt.tables[af], list) {
1218 if (strcmp(t->name, name))
1220 if (!try_module_get(t->me))
1222 mutex_unlock(&xt[af].mutex);
1223 err = t->table_init(net);
1226 return ERR_PTR(err);
1231 mutex_lock(&xt[af].mutex);
1238 /* and once again: */
1239 list_for_each_entry(t, &net->xt.tables[af], list)
1240 if (strcmp(t->name, name) == 0)
1243 module_put(found->me);
1245 mutex_unlock(&xt[af].mutex);
1246 return ERR_PTR(-ENOENT);
1248 EXPORT_SYMBOL_GPL(xt_find_table_lock);
1250 struct xt_table *xt_request_find_table_lock(struct net *net, u_int8_t af,
1253 struct xt_table *t = xt_find_table_lock(net, af, name);
1255 #ifdef CONFIG_MODULES
1257 int err = request_module("%stable_%s", xt_prefix[af], name);
1259 return ERR_PTR(err);
1260 t = xt_find_table_lock(net, af, name);
1266 EXPORT_SYMBOL_GPL(xt_request_find_table_lock);
1268 void xt_table_unlock(struct xt_table *table)
1270 mutex_unlock(&xt[table->af].mutex);
1272 EXPORT_SYMBOL_GPL(xt_table_unlock);
1274 #ifdef CONFIG_COMPAT
1275 void xt_compat_lock(u_int8_t af)
1277 mutex_lock(&xt[af].compat_mutex);
1279 EXPORT_SYMBOL_GPL(xt_compat_lock);
1281 void xt_compat_unlock(u_int8_t af)
1283 mutex_unlock(&xt[af].compat_mutex);
1285 EXPORT_SYMBOL_GPL(xt_compat_unlock);
1288 DEFINE_PER_CPU(seqcount_t, xt_recseq);
1289 EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq);
1291 struct static_key xt_tee_enabled __read_mostly;
1292 EXPORT_SYMBOL_GPL(xt_tee_enabled);
1294 static int xt_jumpstack_alloc(struct xt_table_info *i)
1299 size = sizeof(void **) * nr_cpu_ids;
1300 if (size > PAGE_SIZE)
1301 i->jumpstack = kvzalloc(size, GFP_KERNEL);
1303 i->jumpstack = kzalloc(size, GFP_KERNEL);
1304 if (i->jumpstack == NULL)
1307 /* ruleset without jumps -- no stack needed */
1308 if (i->stacksize == 0)
1311 /* Jumpstack needs to be able to record two full callchains, one
1312 * from the first rule set traversal, plus one table reentrancy
1313 * via -j TEE without clobbering the callchain that brought us to
1316 * This is done by allocating two jumpstacks per cpu, on reentry
1317 * the upper half of the stack is used.
1319 * see the jumpstack setup in ipt_do_table() for more details.
1321 size = sizeof(void *) * i->stacksize * 2u;
1322 for_each_possible_cpu(cpu) {
1323 i->jumpstack[cpu] = kvmalloc_node(size, GFP_KERNEL,
1325 if (i->jumpstack[cpu] == NULL)
1327 * Freeing will be done later on by the callers. The
1328 * chain is: xt_replace_table -> __do_replace ->
1329 * do_replace -> xt_free_table_info.
1337 struct xt_counters *xt_counters_alloc(unsigned int counters)
1339 struct xt_counters *mem;
1341 if (counters == 0 || counters > INT_MAX / sizeof(*mem))
1344 counters *= sizeof(*mem);
1345 if (counters > XT_MAX_TABLE_SIZE)
1348 return vzalloc(counters);
1350 EXPORT_SYMBOL(xt_counters_alloc);
1352 struct xt_table_info *
1353 xt_replace_table(struct xt_table *table,
1354 unsigned int num_counters,
1355 struct xt_table_info *newinfo,
1358 struct xt_table_info *private;
1362 ret = xt_jumpstack_alloc(newinfo);
1368 /* Do the substitution. */
1370 private = table->private;
1372 /* Check inside lock: is the old number correct? */
1373 if (num_counters != private->number) {
1374 pr_debug("num_counters != table->private->number (%u/%u)\n",
1375 num_counters, private->number);
1381 newinfo->initial_entries = private->initial_entries;
1383 * Ensure contents of newinfo are visible before assigning to
1387 table->private = newinfo;
1389 /* make sure all cpus see new ->private value */
1393 * Even though table entries have now been swapped, other CPU's
1394 * may still be using the old entries...
1398 /* ... so wait for even xt_recseq on all cpus */
1399 for_each_possible_cpu(cpu) {
1400 seqcount_t *s = &per_cpu(xt_recseq, cpu);
1401 u32 seq = raw_read_seqcount(s);
1407 } while (seq == raw_read_seqcount(s));
1411 audit_log_nfcfg(table->name, table->af, private->number,
1412 !private->number ? AUDIT_XT_OP_REGISTER :
1413 AUDIT_XT_OP_REPLACE,
1417 EXPORT_SYMBOL_GPL(xt_replace_table);
1419 struct xt_table *xt_register_table(struct net *net,
1420 const struct xt_table *input_table,
1421 struct xt_table_info *bootstrap,
1422 struct xt_table_info *newinfo)
1425 struct xt_table_info *private;
1426 struct xt_table *t, *table;
1428 /* Don't add one object to multiple lists. */
1429 table = kmemdup(input_table, sizeof(struct xt_table), GFP_KERNEL);
1435 mutex_lock(&xt[table->af].mutex);
1436 /* Don't autoload: we'd eat our tail... */
1437 list_for_each_entry(t, &net->xt.tables[table->af], list) {
1438 if (strcmp(t->name, table->name) == 0) {
1444 /* Simplifies replace_table code. */
1445 table->private = bootstrap;
1447 if (!xt_replace_table(table, 0, newinfo, &ret))
1450 private = table->private;
1451 pr_debug("table->private->number = %u\n", private->number);
1453 /* save number of initial entries */
1454 private->initial_entries = private->number;
1456 list_add(&table->list, &net->xt.tables[table->af]);
1457 mutex_unlock(&xt[table->af].mutex);
1461 mutex_unlock(&xt[table->af].mutex);
1464 return ERR_PTR(ret);
1466 EXPORT_SYMBOL_GPL(xt_register_table);
1468 void *xt_unregister_table(struct xt_table *table)
1470 struct xt_table_info *private;
1472 mutex_lock(&xt[table->af].mutex);
1473 private = table->private;
1474 list_del(&table->list);
1475 mutex_unlock(&xt[table->af].mutex);
1476 audit_log_nfcfg(table->name, table->af, private->number,
1477 AUDIT_XT_OP_UNREGISTER, GFP_KERNEL);
1482 EXPORT_SYMBOL_GPL(xt_unregister_table);
1484 #ifdef CONFIG_PROC_FS
1485 static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos)
1487 struct net *net = seq_file_net(seq);
1488 u_int8_t af = (unsigned long)PDE_DATA(file_inode(seq->file));
1490 mutex_lock(&xt[af].mutex);
1491 return seq_list_start(&net->xt.tables[af], *pos);
1494 static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1496 struct net *net = seq_file_net(seq);
1497 u_int8_t af = (unsigned long)PDE_DATA(file_inode(seq->file));
1499 return seq_list_next(v, &net->xt.tables[af], pos);
1502 static void xt_table_seq_stop(struct seq_file *seq, void *v)
1504 u_int8_t af = (unsigned long)PDE_DATA(file_inode(seq->file));
1506 mutex_unlock(&xt[af].mutex);
1509 static int xt_table_seq_show(struct seq_file *seq, void *v)
1511 struct xt_table *table = list_entry(v, struct xt_table, list);
1514 seq_printf(seq, "%s\n", table->name);
1518 static const struct seq_operations xt_table_seq_ops = {
1519 .start = xt_table_seq_start,
1520 .next = xt_table_seq_next,
1521 .stop = xt_table_seq_stop,
1522 .show = xt_table_seq_show,
1526 * Traverse state for ip{,6}_{tables,matches} for helping crossing
1527 * the multi-AF mutexes.
1529 struct nf_mttg_trav {
1530 struct list_head *head, *curr;
1536 MTTG_TRAV_NFP_UNSPEC,
1541 static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos,
1544 static const uint8_t next_class[] = {
1545 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC,
1546 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE,
1548 uint8_t nfproto = (unsigned long)PDE_DATA(file_inode(seq->file));
1549 struct nf_mttg_trav *trav = seq->private;
1554 switch (trav->class) {
1555 case MTTG_TRAV_INIT:
1556 trav->class = MTTG_TRAV_NFP_UNSPEC;
1557 mutex_lock(&xt[NFPROTO_UNSPEC].mutex);
1558 trav->head = trav->curr = is_target ?
1559 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match;
1561 case MTTG_TRAV_NFP_UNSPEC:
1562 trav->curr = trav->curr->next;
1563 if (trav->curr != trav->head)
1565 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1566 mutex_lock(&xt[nfproto].mutex);
1567 trav->head = trav->curr = is_target ?
1568 &xt[nfproto].target : &xt[nfproto].match;
1569 trav->class = next_class[trav->class];
1571 case MTTG_TRAV_NFP_SPEC:
1572 trav->curr = trav->curr->next;
1573 if (trav->curr != trav->head)
1582 static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos,
1585 struct nf_mttg_trav *trav = seq->private;
1588 trav->class = MTTG_TRAV_INIT;
1589 for (j = 0; j < *pos; ++j)
1590 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL)
1595 static void xt_mttg_seq_stop(struct seq_file *seq, void *v)
1597 uint8_t nfproto = (unsigned long)PDE_DATA(file_inode(seq->file));
1598 struct nf_mttg_trav *trav = seq->private;
1600 switch (trav->class) {
1601 case MTTG_TRAV_NFP_UNSPEC:
1602 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1604 case MTTG_TRAV_NFP_SPEC:
1605 mutex_unlock(&xt[nfproto].mutex);
1610 static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos)
1612 return xt_mttg_seq_start(seq, pos, false);
1615 static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1617 return xt_mttg_seq_next(seq, v, ppos, false);
1620 static int xt_match_seq_show(struct seq_file *seq, void *v)
1622 const struct nf_mttg_trav *trav = seq->private;
1623 const struct xt_match *match;
1625 switch (trav->class) {
1626 case MTTG_TRAV_NFP_UNSPEC:
1627 case MTTG_TRAV_NFP_SPEC:
1628 if (trav->curr == trav->head)
1630 match = list_entry(trav->curr, struct xt_match, list);
1632 seq_printf(seq, "%s\n", match->name);
1637 static const struct seq_operations xt_match_seq_ops = {
1638 .start = xt_match_seq_start,
1639 .next = xt_match_seq_next,
1640 .stop = xt_mttg_seq_stop,
1641 .show = xt_match_seq_show,
1644 static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos)
1646 return xt_mttg_seq_start(seq, pos, true);
1649 static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1651 return xt_mttg_seq_next(seq, v, ppos, true);
1654 static int xt_target_seq_show(struct seq_file *seq, void *v)
1656 const struct nf_mttg_trav *trav = seq->private;
1657 const struct xt_target *target;
1659 switch (trav->class) {
1660 case MTTG_TRAV_NFP_UNSPEC:
1661 case MTTG_TRAV_NFP_SPEC:
1662 if (trav->curr == trav->head)
1664 target = list_entry(trav->curr, struct xt_target, list);
1666 seq_printf(seq, "%s\n", target->name);
1671 static const struct seq_operations xt_target_seq_ops = {
1672 .start = xt_target_seq_start,
1673 .next = xt_target_seq_next,
1674 .stop = xt_mttg_seq_stop,
1675 .show = xt_target_seq_show,
1678 #define FORMAT_TABLES "_tables_names"
1679 #define FORMAT_MATCHES "_tables_matches"
1680 #define FORMAT_TARGETS "_tables_targets"
1682 #endif /* CONFIG_PROC_FS */
1685 * xt_hook_ops_alloc - set up hooks for a new table
1686 * @table: table with metadata needed to set up hooks
1687 * @fn: Hook function
1689 * This function will create the nf_hook_ops that the x_table needs
1690 * to hand to xt_hook_link_net().
1692 struct nf_hook_ops *
1693 xt_hook_ops_alloc(const struct xt_table *table, nf_hookfn *fn)
1695 unsigned int hook_mask = table->valid_hooks;
1696 uint8_t i, num_hooks = hweight32(hook_mask);
1698 struct nf_hook_ops *ops;
1701 return ERR_PTR(-EINVAL);
1703 ops = kcalloc(num_hooks, sizeof(*ops), GFP_KERNEL);
1705 return ERR_PTR(-ENOMEM);
1707 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0;
1708 hook_mask >>= 1, ++hooknum) {
1709 if (!(hook_mask & 1))
1712 ops[i].pf = table->af;
1713 ops[i].hooknum = hooknum;
1714 ops[i].priority = table->priority;
1720 EXPORT_SYMBOL_GPL(xt_hook_ops_alloc);
1722 int xt_proto_init(struct net *net, u_int8_t af)
1724 #ifdef CONFIG_PROC_FS
1725 char buf[XT_FUNCTION_MAXNAMELEN];
1726 struct proc_dir_entry *proc;
1731 if (af >= ARRAY_SIZE(xt_prefix))
1735 #ifdef CONFIG_PROC_FS
1736 root_uid = make_kuid(net->user_ns, 0);
1737 root_gid = make_kgid(net->user_ns, 0);
1739 strlcpy(buf, xt_prefix[af], sizeof(buf));
1740 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1741 proc = proc_create_net_data(buf, 0440, net->proc_net, &xt_table_seq_ops,
1742 sizeof(struct seq_net_private),
1743 (void *)(unsigned long)af);
1746 if (uid_valid(root_uid) && gid_valid(root_gid))
1747 proc_set_user(proc, root_uid, root_gid);
1749 strlcpy(buf, xt_prefix[af], sizeof(buf));
1750 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1751 proc = proc_create_seq_private(buf, 0440, net->proc_net,
1752 &xt_match_seq_ops, sizeof(struct nf_mttg_trav),
1753 (void *)(unsigned long)af);
1755 goto out_remove_tables;
1756 if (uid_valid(root_uid) && gid_valid(root_gid))
1757 proc_set_user(proc, root_uid, root_gid);
1759 strlcpy(buf, xt_prefix[af], sizeof(buf));
1760 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1761 proc = proc_create_seq_private(buf, 0440, net->proc_net,
1762 &xt_target_seq_ops, sizeof(struct nf_mttg_trav),
1763 (void *)(unsigned long)af);
1765 goto out_remove_matches;
1766 if (uid_valid(root_uid) && gid_valid(root_gid))
1767 proc_set_user(proc, root_uid, root_gid);
1772 #ifdef CONFIG_PROC_FS
1774 strlcpy(buf, xt_prefix[af], sizeof(buf));
1775 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1776 remove_proc_entry(buf, net->proc_net);
1779 strlcpy(buf, xt_prefix[af], sizeof(buf));
1780 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1781 remove_proc_entry(buf, net->proc_net);
1786 EXPORT_SYMBOL_GPL(xt_proto_init);
1788 void xt_proto_fini(struct net *net, u_int8_t af)
1790 #ifdef CONFIG_PROC_FS
1791 char buf[XT_FUNCTION_MAXNAMELEN];
1793 strlcpy(buf, xt_prefix[af], sizeof(buf));
1794 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1795 remove_proc_entry(buf, net->proc_net);
1797 strlcpy(buf, xt_prefix[af], sizeof(buf));
1798 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1799 remove_proc_entry(buf, net->proc_net);
1801 strlcpy(buf, xt_prefix[af], sizeof(buf));
1802 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1803 remove_proc_entry(buf, net->proc_net);
1804 #endif /*CONFIG_PROC_FS*/
1806 EXPORT_SYMBOL_GPL(xt_proto_fini);
1809 * xt_percpu_counter_alloc - allocate x_tables rule counter
1811 * @state: pointer to xt_percpu allocation state
1812 * @counter: pointer to counter struct inside the ip(6)/arpt_entry struct
1814 * On SMP, the packet counter [ ip(6)t_entry->counters.pcnt ] will then
1815 * contain the address of the real (percpu) counter.
1817 * Rule evaluation needs to use xt_get_this_cpu_counter() helper
1818 * to fetch the real percpu counter.
1820 * To speed up allocation and improve data locality, a 4kb block is
1821 * allocated. Freeing any counter may free an entire block, so all
1822 * counters allocated using the same state must be freed at the same
1825 * xt_percpu_counter_alloc_state contains the base address of the
1826 * allocated page and the current sub-offset.
1828 * returns false on error.
1830 bool xt_percpu_counter_alloc(struct xt_percpu_counter_alloc_state *state,
1831 struct xt_counters *counter)
1833 BUILD_BUG_ON(XT_PCPU_BLOCK_SIZE < (sizeof(*counter) * 2));
1835 if (nr_cpu_ids <= 1)
1839 state->mem = __alloc_percpu(XT_PCPU_BLOCK_SIZE,
1840 XT_PCPU_BLOCK_SIZE);
1844 counter->pcnt = (__force unsigned long)(state->mem + state->off);
1845 state->off += sizeof(*counter);
1846 if (state->off > (XT_PCPU_BLOCK_SIZE - sizeof(*counter))) {
1852 EXPORT_SYMBOL_GPL(xt_percpu_counter_alloc);
1854 void xt_percpu_counter_free(struct xt_counters *counters)
1856 unsigned long pcnt = counters->pcnt;
1858 if (nr_cpu_ids > 1 && (pcnt & (XT_PCPU_BLOCK_SIZE - 1)) == 0)
1859 free_percpu((void __percpu *)pcnt);
1861 EXPORT_SYMBOL_GPL(xt_percpu_counter_free);
1863 static int __net_init xt_net_init(struct net *net)
1867 for (i = 0; i < NFPROTO_NUMPROTO; i++)
1868 INIT_LIST_HEAD(&net->xt.tables[i]);
1872 static void __net_exit xt_net_exit(struct net *net)
1876 for (i = 0; i < NFPROTO_NUMPROTO; i++)
1877 WARN_ON_ONCE(!list_empty(&net->xt.tables[i]));
1880 static struct pernet_operations xt_net_ops = {
1881 .init = xt_net_init,
1882 .exit = xt_net_exit,
1885 static int __init xt_init(void)
1890 for_each_possible_cpu(i) {
1891 seqcount_init(&per_cpu(xt_recseq, i));
1894 xt = kcalloc(NFPROTO_NUMPROTO, sizeof(struct xt_af), GFP_KERNEL);
1898 for (i = 0; i < NFPROTO_NUMPROTO; i++) {
1899 mutex_init(&xt[i].mutex);
1900 #ifdef CONFIG_COMPAT
1901 mutex_init(&xt[i].compat_mutex);
1902 xt[i].compat_tab = NULL;
1904 INIT_LIST_HEAD(&xt[i].target);
1905 INIT_LIST_HEAD(&xt[i].match);
1907 rv = register_pernet_subsys(&xt_net_ops);
1913 static void __exit xt_fini(void)
1915 unregister_pernet_subsys(&xt_net_ops);
1919 module_init(xt_init);
1920 module_exit(xt_fini);