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 <linux/user_namespace.h>
30 #include <net/net_namespace.h>
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>
38 MODULE_LICENSE("GPL");
39 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
40 MODULE_DESCRIPTION("{ip,ip6,arp,eb}_tables backend module");
42 #define SMP_ALIGN(x) (((x) + SMP_CACHE_BYTES-1) & ~(SMP_CACHE_BYTES-1))
45 unsigned int offset; /* offset in kernel */
46 int delta; /* delta in 32bit user land */
51 struct list_head match;
52 struct list_head target;
54 struct mutex compat_mutex;
55 struct compat_delta *compat_tab;
56 unsigned int number; /* number of slots in compat_tab[] */
57 unsigned int cur; /* number of used slots in compat_tab[] */
61 static struct xt_af *xt;
63 static const char *const xt_prefix[NFPROTO_NUMPROTO] = {
64 [NFPROTO_UNSPEC] = "x",
65 [NFPROTO_IPV4] = "ip",
66 [NFPROTO_ARP] = "arp",
67 [NFPROTO_BRIDGE] = "eb",
68 [NFPROTO_IPV6] = "ip6",
71 /* Registration hooks for targets. */
72 int xt_register_target(struct xt_target *target)
74 u_int8_t af = target->family;
76 mutex_lock(&xt[af].mutex);
77 list_add(&target->list, &xt[af].target);
78 mutex_unlock(&xt[af].mutex);
81 EXPORT_SYMBOL(xt_register_target);
84 xt_unregister_target(struct xt_target *target)
86 u_int8_t af = target->family;
88 mutex_lock(&xt[af].mutex);
89 list_del(&target->list);
90 mutex_unlock(&xt[af].mutex);
92 EXPORT_SYMBOL(xt_unregister_target);
95 xt_register_targets(struct xt_target *target, unsigned int n)
100 for (i = 0; i < n; i++) {
101 err = xt_register_target(&target[i]);
109 xt_unregister_targets(target, i);
112 EXPORT_SYMBOL(xt_register_targets);
115 xt_unregister_targets(struct xt_target *target, unsigned int n)
118 xt_unregister_target(&target[n]);
120 EXPORT_SYMBOL(xt_unregister_targets);
122 int xt_register_match(struct xt_match *match)
124 u_int8_t af = match->family;
126 mutex_lock(&xt[af].mutex);
127 list_add(&match->list, &xt[af].match);
128 mutex_unlock(&xt[af].mutex);
131 EXPORT_SYMBOL(xt_register_match);
134 xt_unregister_match(struct xt_match *match)
136 u_int8_t af = match->family;
138 mutex_lock(&xt[af].mutex);
139 list_del(&match->list);
140 mutex_unlock(&xt[af].mutex);
142 EXPORT_SYMBOL(xt_unregister_match);
145 xt_register_matches(struct xt_match *match, unsigned int n)
150 for (i = 0; i < n; i++) {
151 err = xt_register_match(&match[i]);
159 xt_unregister_matches(match, i);
162 EXPORT_SYMBOL(xt_register_matches);
165 xt_unregister_matches(struct xt_match *match, unsigned int n)
168 xt_unregister_match(&match[n]);
170 EXPORT_SYMBOL(xt_unregister_matches);
174 * These are weird, but module loading must not be done with mutex
175 * held (since they will register), and we have to have a single
179 /* Find match, grabs ref. Returns ERR_PTR() on error. */
180 struct xt_match *xt_find_match(u8 af, const char *name, u8 revision)
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 match = xt_find_match(nfproto, name, revision);
214 request_module("%st_%s", xt_prefix[nfproto], name);
215 match = xt_find_match(nfproto, name, revision);
220 EXPORT_SYMBOL_GPL(xt_request_find_match);
222 /* Find target, grabs ref. Returns ERR_PTR() on error. */
223 struct xt_target *xt_find_target(u8 af, const char *name, u8 revision)
228 mutex_lock(&xt[af].mutex);
229 list_for_each_entry(t, &xt[af].target, list) {
230 if (strcmp(t->name, name) == 0) {
231 if (t->revision == revision) {
232 if (try_module_get(t->me)) {
233 mutex_unlock(&xt[af].mutex);
237 err = -EPROTOTYPE; /* Found something. */
240 mutex_unlock(&xt[af].mutex);
242 if (af != NFPROTO_UNSPEC)
243 /* Try searching again in the family-independent list */
244 return xt_find_target(NFPROTO_UNSPEC, name, revision);
248 EXPORT_SYMBOL(xt_find_target);
250 struct xt_target *xt_request_find_target(u8 af, const char *name, u8 revision)
252 struct xt_target *target;
254 target = xt_find_target(af, name, revision);
255 if (IS_ERR(target)) {
256 request_module("%st_%s", xt_prefix[af], name);
257 target = xt_find_target(af, name, revision);
262 EXPORT_SYMBOL_GPL(xt_request_find_target);
264 static int match_revfn(u8 af, const char *name, u8 revision, int *bestp)
266 const struct xt_match *m;
269 list_for_each_entry(m, &xt[af].match, list) {
270 if (strcmp(m->name, name) == 0) {
271 if (m->revision > *bestp)
272 *bestp = m->revision;
273 if (m->revision == revision)
278 if (af != NFPROTO_UNSPEC && !have_rev)
279 return match_revfn(NFPROTO_UNSPEC, name, revision, bestp);
284 static int target_revfn(u8 af, const char *name, u8 revision, int *bestp)
286 const struct xt_target *t;
289 list_for_each_entry(t, &xt[af].target, list) {
290 if (strcmp(t->name, name) == 0) {
291 if (t->revision > *bestp)
292 *bestp = t->revision;
293 if (t->revision == revision)
298 if (af != NFPROTO_UNSPEC && !have_rev)
299 return target_revfn(NFPROTO_UNSPEC, name, revision, bestp);
304 /* Returns true or false (if no such extension at all) */
305 int xt_find_revision(u8 af, const char *name, u8 revision, int target,
308 int have_rev, best = -1;
310 mutex_lock(&xt[af].mutex);
312 have_rev = target_revfn(af, name, revision, &best);
314 have_rev = match_revfn(af, name, revision, &best);
315 mutex_unlock(&xt[af].mutex);
317 /* Nothing at all? Return 0 to try loading module. */
325 *err = -EPROTONOSUPPORT;
328 EXPORT_SYMBOL_GPL(xt_find_revision);
331 textify_hooks(char *buf, size_t size, unsigned int mask, uint8_t nfproto)
333 static const char *const inetbr_names[] = {
334 "PREROUTING", "INPUT", "FORWARD",
335 "OUTPUT", "POSTROUTING", "BROUTING",
337 static const char *const arp_names[] = {
338 "INPUT", "FORWARD", "OUTPUT",
340 const char *const *names;
346 names = (nfproto == NFPROTO_ARP) ? arp_names : inetbr_names;
347 max = (nfproto == NFPROTO_ARP) ? ARRAY_SIZE(arp_names) :
348 ARRAY_SIZE(inetbr_names);
350 for (i = 0; i < max; ++i) {
351 if (!(mask & (1 << i)))
353 res = snprintf(p, size, "%s%s", np ? "/" : "", names[i]);
364 int xt_check_match(struct xt_mtchk_param *par,
365 unsigned int size, u_int8_t proto, bool inv_proto)
369 if (XT_ALIGN(par->match->matchsize) != size &&
370 par->match->matchsize != -1) {
372 * ebt_among is exempt from centralized matchsize checking
373 * because it uses a dynamic-size data set.
375 pr_err("%s_tables: %s.%u match: invalid size "
376 "%u (kernel) != (user) %u\n",
377 xt_prefix[par->family], par->match->name,
378 par->match->revision,
379 XT_ALIGN(par->match->matchsize), size);
382 if (par->match->table != NULL &&
383 strcmp(par->match->table, par->table) != 0) {
384 pr_err("%s_tables: %s match: only valid in %s table, not %s\n",
385 xt_prefix[par->family], par->match->name,
386 par->match->table, par->table);
389 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) {
390 char used[64], allow[64];
392 pr_err("%s_tables: %s match: used from hooks %s, but only "
394 xt_prefix[par->family], par->match->name,
395 textify_hooks(used, sizeof(used), par->hook_mask,
397 textify_hooks(allow, sizeof(allow), par->match->hooks,
401 if (par->match->proto && (par->match->proto != proto || inv_proto)) {
402 pr_err("%s_tables: %s match: only valid for protocol %u\n",
403 xt_prefix[par->family], par->match->name,
407 if (par->match->checkentry != NULL) {
408 ret = par->match->checkentry(par);
412 /* Flag up potential errors. */
417 EXPORT_SYMBOL_GPL(xt_check_match);
419 /** xt_check_entry_match - check that matches end before start of target
421 * @match: beginning of xt_entry_match
422 * @target: beginning of this rules target (alleged end of matches)
423 * @alignment: alignment requirement of match structures
425 * Validates that all matches add up to the beginning of the target,
426 * and that each match covers at least the base structure size.
428 * Return: 0 on success, negative errno on failure.
430 static int xt_check_entry_match(const char *match, const char *target,
431 const size_t alignment)
433 const struct xt_entry_match *pos;
434 int length = target - match;
436 if (length == 0) /* no matches */
439 pos = (struct xt_entry_match *)match;
441 if ((unsigned long)pos % alignment)
444 if (length < (int)sizeof(struct xt_entry_match))
447 if (pos->u.match_size < sizeof(struct xt_entry_match))
450 if (pos->u.match_size > length)
453 length -= pos->u.match_size;
454 pos = ((void *)((char *)(pos) + (pos)->u.match_size));
455 } while (length > 0);
461 int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta)
463 struct xt_af *xp = &xt[af];
465 if (!xp->compat_tab) {
468 xp->compat_tab = vmalloc(sizeof(struct compat_delta) * xp->number);
474 if (xp->cur >= xp->number)
478 delta += xp->compat_tab[xp->cur - 1].delta;
479 xp->compat_tab[xp->cur].offset = offset;
480 xp->compat_tab[xp->cur].delta = delta;
484 EXPORT_SYMBOL_GPL(xt_compat_add_offset);
486 void xt_compat_flush_offsets(u_int8_t af)
488 if (xt[af].compat_tab) {
489 vfree(xt[af].compat_tab);
490 xt[af].compat_tab = NULL;
495 EXPORT_SYMBOL_GPL(xt_compat_flush_offsets);
497 int xt_compat_calc_jump(u_int8_t af, unsigned int offset)
499 struct compat_delta *tmp = xt[af].compat_tab;
500 int mid, left = 0, right = xt[af].cur - 1;
502 while (left <= right) {
503 mid = (left + right) >> 1;
504 if (offset > tmp[mid].offset)
506 else if (offset < tmp[mid].offset)
509 return mid ? tmp[mid - 1].delta : 0;
511 return left ? tmp[left - 1].delta : 0;
513 EXPORT_SYMBOL_GPL(xt_compat_calc_jump);
515 void xt_compat_init_offsets(u_int8_t af, unsigned int number)
517 xt[af].number = number;
520 EXPORT_SYMBOL(xt_compat_init_offsets);
522 int xt_compat_match_offset(const struct xt_match *match)
524 u_int16_t csize = match->compatsize ? : match->matchsize;
525 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize);
527 EXPORT_SYMBOL_GPL(xt_compat_match_offset);
529 void xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr,
532 const struct xt_match *match = m->u.kernel.match;
533 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m;
534 int pad, off = xt_compat_match_offset(match);
535 u_int16_t msize = cm->u.user.match_size;
536 char name[sizeof(m->u.user.name)];
539 memcpy(m, cm, sizeof(*cm));
540 if (match->compat_from_user)
541 match->compat_from_user(m->data, cm->data);
543 memcpy(m->data, cm->data, msize - sizeof(*cm));
544 pad = XT_ALIGN(match->matchsize) - match->matchsize;
546 memset(m->data + match->matchsize, 0, pad);
549 m->u.user.match_size = msize;
550 strlcpy(name, match->name, sizeof(name));
551 module_put(match->me);
552 strncpy(m->u.user.name, name, sizeof(m->u.user.name));
557 EXPORT_SYMBOL_GPL(xt_compat_match_from_user);
559 int xt_compat_match_to_user(const struct xt_entry_match *m,
560 void __user **dstptr, unsigned int *size)
562 const struct xt_match *match = m->u.kernel.match;
563 struct compat_xt_entry_match __user *cm = *dstptr;
564 int off = xt_compat_match_offset(match);
565 u_int16_t msize = m->u.user.match_size - off;
567 if (copy_to_user(cm, m, sizeof(*cm)) ||
568 put_user(msize, &cm->u.user.match_size) ||
569 copy_to_user(cm->u.user.name, m->u.kernel.match->name,
570 strlen(m->u.kernel.match->name) + 1))
573 if (match->compat_to_user) {
574 if (match->compat_to_user((void __user *)cm->data, m->data))
577 if (copy_to_user(cm->data, m->data, msize - sizeof(*cm)))
585 EXPORT_SYMBOL_GPL(xt_compat_match_to_user);
587 /* non-compat version may have padding after verdict */
588 struct compat_xt_standard_target {
589 struct compat_xt_entry_target t;
590 compat_uint_t verdict;
593 int xt_compat_check_entry_offsets(const void *base, const char *elems,
594 unsigned int target_offset,
595 unsigned int next_offset)
597 long size_of_base_struct = elems - (const char *)base;
598 const struct compat_xt_entry_target *t;
599 const char *e = base;
601 if (target_offset < size_of_base_struct)
604 if (target_offset + sizeof(*t) > next_offset)
607 t = (void *)(e + target_offset);
608 if (t->u.target_size < sizeof(*t))
611 if (target_offset + t->u.target_size > next_offset)
614 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
615 target_offset + sizeof(struct compat_xt_standard_target) != next_offset)
618 /* compat_xt_entry match has less strict aligment requirements,
619 * otherwise they are identical. In case of padding differences
620 * we need to add compat version of xt_check_entry_match.
622 BUILD_BUG_ON(sizeof(struct compat_xt_entry_match) != sizeof(struct xt_entry_match));
624 return xt_check_entry_match(elems, base + target_offset,
625 __alignof__(struct compat_xt_entry_match));
627 EXPORT_SYMBOL(xt_compat_check_entry_offsets);
628 #endif /* CONFIG_COMPAT */
631 * xt_check_entry_offsets - validate arp/ip/ip6t_entry
633 * @base: pointer to arp/ip/ip6t_entry
634 * @elems: pointer to first xt_entry_match, i.e. ip(6)t_entry->elems
635 * @target_offset: the arp/ip/ip6_t->target_offset
636 * @next_offset: the arp/ip/ip6_t->next_offset
638 * validates that target_offset and next_offset are sane and that all
639 * match sizes (if any) align with the target offset.
641 * This function does not validate the targets or matches themselves, it
642 * only tests that all the offsets and sizes are correct, that all
643 * match structures are aligned, and that the last structure ends where
644 * the target structure begins.
646 * Also see xt_compat_check_entry_offsets for CONFIG_COMPAT version.
648 * The arp/ip/ip6t_entry structure @base must have passed following tests:
649 * - it must point to a valid memory location
650 * - base to base + next_offset must be accessible, i.e. not exceed allocated
653 * A well-formed entry looks like this:
655 * ip(6)t_entry match [mtdata] match [mtdata] target [tgdata] ip(6)t_entry
656 * e->elems[]-----' | |
660 * target_offset---------------------------------' |
661 * next_offset---------------------------------------------------'
663 * elems[]: flexible array member at end of ip(6)/arpt_entry struct.
664 * This is where matches (if any) and the target reside.
665 * target_offset: beginning of target.
666 * next_offset: start of the next rule; also: size of this rule.
667 * Since targets have a minimum size, target_offset + minlen <= next_offset.
669 * Every match stores its size, sum of sizes must not exceed target_offset.
671 * Return: 0 on success, negative errno on failure.
673 int xt_check_entry_offsets(const void *base,
675 unsigned int target_offset,
676 unsigned int next_offset)
678 long size_of_base_struct = elems - (const char *)base;
679 const struct xt_entry_target *t;
680 const char *e = base;
682 /* target start is within the ip/ip6/arpt_entry struct */
683 if (target_offset < size_of_base_struct)
686 if (target_offset + sizeof(*t) > next_offset)
689 t = (void *)(e + target_offset);
690 if (t->u.target_size < sizeof(*t))
693 if (target_offset + t->u.target_size > next_offset)
696 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
697 target_offset + sizeof(struct xt_standard_target) != next_offset)
700 return xt_check_entry_match(elems, base + target_offset,
701 __alignof__(struct xt_entry_match));
703 EXPORT_SYMBOL(xt_check_entry_offsets);
705 int xt_check_target(struct xt_tgchk_param *par,
706 unsigned int size, u_int8_t proto, bool inv_proto)
710 if (XT_ALIGN(par->target->targetsize) != size) {
711 pr_err("%s_tables: %s.%u target: invalid size "
712 "%u (kernel) != (user) %u\n",
713 xt_prefix[par->family], par->target->name,
714 par->target->revision,
715 XT_ALIGN(par->target->targetsize), size);
718 if (par->target->table != NULL &&
719 strcmp(par->target->table, par->table) != 0) {
720 pr_err("%s_tables: %s target: only valid in %s table, not %s\n",
721 xt_prefix[par->family], par->target->name,
722 par->target->table, par->table);
725 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) {
726 char used[64], allow[64];
728 pr_err("%s_tables: %s target: used from hooks %s, but only "
730 xt_prefix[par->family], par->target->name,
731 textify_hooks(used, sizeof(used), par->hook_mask,
733 textify_hooks(allow, sizeof(allow), par->target->hooks,
737 if (par->target->proto && (par->target->proto != proto || inv_proto)) {
738 pr_err("%s_tables: %s target: only valid for protocol %u\n",
739 xt_prefix[par->family], par->target->name,
743 if (par->target->checkentry != NULL) {
744 ret = par->target->checkentry(par);
748 /* Flag up potential errors. */
753 EXPORT_SYMBOL_GPL(xt_check_target);
756 * xt_copy_counters_from_user - copy counters and metadata from userspace
758 * @user: src pointer to userspace memory
759 * @len: alleged size of userspace memory
760 * @info: where to store the xt_counters_info metadata
761 * @compat: true if we setsockopt call is done by 32bit task on 64bit kernel
763 * Copies counter meta data from @user and stores it in @info.
765 * vmallocs memory to hold the counters, then copies the counter data
766 * from @user to the new memory and returns a pointer to it.
768 * If @compat is true, @info gets converted automatically to the 64bit
771 * The metadata associated with the counters is stored in @info.
773 * Return: returns pointer that caller has to test via IS_ERR().
774 * If IS_ERR is false, caller has to vfree the pointer.
776 void *xt_copy_counters_from_user(const void __user *user, unsigned int len,
777 struct xt_counters_info *info, bool compat)
784 /* structures only differ in size due to alignment */
785 struct compat_xt_counters_info compat_tmp;
787 if (len <= sizeof(compat_tmp))
788 return ERR_PTR(-EINVAL);
790 len -= sizeof(compat_tmp);
791 if (copy_from_user(&compat_tmp, user, sizeof(compat_tmp)) != 0)
792 return ERR_PTR(-EFAULT);
794 strlcpy(info->name, compat_tmp.name, sizeof(info->name));
795 info->num_counters = compat_tmp.num_counters;
796 user += sizeof(compat_tmp);
800 if (len <= sizeof(*info))
801 return ERR_PTR(-EINVAL);
803 len -= sizeof(*info);
804 if (copy_from_user(info, user, sizeof(*info)) != 0)
805 return ERR_PTR(-EFAULT);
807 info->name[sizeof(info->name) - 1] = '\0';
808 user += sizeof(*info);
811 size = sizeof(struct xt_counters);
812 size *= info->num_counters;
814 if (size != (u64)len)
815 return ERR_PTR(-EINVAL);
819 return ERR_PTR(-ENOMEM);
821 if (copy_from_user(mem, user, len) == 0)
825 return ERR_PTR(-EFAULT);
827 EXPORT_SYMBOL_GPL(xt_copy_counters_from_user);
830 int xt_compat_target_offset(const struct xt_target *target)
832 u_int16_t csize = target->compatsize ? : target->targetsize;
833 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize);
835 EXPORT_SYMBOL_GPL(xt_compat_target_offset);
837 void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr,
840 const struct xt_target *target = t->u.kernel.target;
841 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t;
842 int pad, off = xt_compat_target_offset(target);
843 u_int16_t tsize = ct->u.user.target_size;
844 char name[sizeof(t->u.user.name)];
847 memcpy(t, ct, sizeof(*ct));
848 if (target->compat_from_user)
849 target->compat_from_user(t->data, ct->data);
851 memcpy(t->data, ct->data, tsize - sizeof(*ct));
852 pad = XT_ALIGN(target->targetsize) - target->targetsize;
854 memset(t->data + target->targetsize, 0, pad);
857 t->u.user.target_size = tsize;
858 strlcpy(name, target->name, sizeof(name));
859 module_put(target->me);
860 strncpy(t->u.user.name, name, sizeof(t->u.user.name));
865 EXPORT_SYMBOL_GPL(xt_compat_target_from_user);
867 int xt_compat_target_to_user(const struct xt_entry_target *t,
868 void __user **dstptr, unsigned int *size)
870 const struct xt_target *target = t->u.kernel.target;
871 struct compat_xt_entry_target __user *ct = *dstptr;
872 int off = xt_compat_target_offset(target);
873 u_int16_t tsize = t->u.user.target_size - off;
875 if (copy_to_user(ct, t, sizeof(*ct)) ||
876 put_user(tsize, &ct->u.user.target_size) ||
877 copy_to_user(ct->u.user.name, t->u.kernel.target->name,
878 strlen(t->u.kernel.target->name) + 1))
881 if (target->compat_to_user) {
882 if (target->compat_to_user((void __user *)ct->data, t->data))
885 if (copy_to_user(ct->data, t->data, tsize - sizeof(*ct)))
893 EXPORT_SYMBOL_GPL(xt_compat_target_to_user);
896 struct xt_table_info *xt_alloc_table_info(unsigned int size)
898 struct xt_table_info *info = NULL;
899 size_t sz = sizeof(*info) + size;
901 if (sz < sizeof(*info))
904 /* Pedantry: prevent them from hitting BUG() in vmalloc.c --RR */
905 if ((SMP_ALIGN(size) >> PAGE_SHIFT) + 2 > totalram_pages)
908 if (sz <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
909 info = kmalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY);
915 memset(info, 0, sizeof(*info));
919 EXPORT_SYMBOL(xt_alloc_table_info);
921 void xt_free_table_info(struct xt_table_info *info)
925 if (info->jumpstack != NULL) {
926 for_each_possible_cpu(cpu)
927 kvfree(info->jumpstack[cpu]);
928 kvfree(info->jumpstack);
933 EXPORT_SYMBOL(xt_free_table_info);
935 /* Find table by name, grabs mutex & ref. Returns ERR_PTR() on error. */
936 struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af,
939 struct xt_table *t, *found = NULL;
941 mutex_lock(&xt[af].mutex);
942 list_for_each_entry(t, &net->xt.tables[af], list)
943 if (strcmp(t->name, name) == 0 && try_module_get(t->me))
946 if (net == &init_net)
949 /* Table doesn't exist in this netns, re-try init */
950 list_for_each_entry(t, &init_net.xt.tables[af], list) {
951 if (strcmp(t->name, name))
953 if (!try_module_get(t->me))
956 mutex_unlock(&xt[af].mutex);
957 if (t->table_init(net) != 0) {
964 mutex_lock(&xt[af].mutex);
971 /* and once again: */
972 list_for_each_entry(t, &net->xt.tables[af], list)
973 if (strcmp(t->name, name) == 0)
976 module_put(found->me);
978 mutex_unlock(&xt[af].mutex);
981 EXPORT_SYMBOL_GPL(xt_find_table_lock);
983 void xt_table_unlock(struct xt_table *table)
985 mutex_unlock(&xt[table->af].mutex);
987 EXPORT_SYMBOL_GPL(xt_table_unlock);
990 void xt_compat_lock(u_int8_t af)
992 mutex_lock(&xt[af].compat_mutex);
994 EXPORT_SYMBOL_GPL(xt_compat_lock);
996 void xt_compat_unlock(u_int8_t af)
998 mutex_unlock(&xt[af].compat_mutex);
1000 EXPORT_SYMBOL_GPL(xt_compat_unlock);
1003 DEFINE_PER_CPU(seqcount_t, xt_recseq);
1004 EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq);
1006 struct static_key xt_tee_enabled __read_mostly;
1007 EXPORT_SYMBOL_GPL(xt_tee_enabled);
1009 static int xt_jumpstack_alloc(struct xt_table_info *i)
1014 size = sizeof(void **) * nr_cpu_ids;
1015 if (size > PAGE_SIZE)
1016 i->jumpstack = vzalloc(size);
1018 i->jumpstack = kzalloc(size, GFP_KERNEL);
1019 if (i->jumpstack == NULL)
1022 /* ruleset without jumps -- no stack needed */
1023 if (i->stacksize == 0)
1026 /* Jumpstack needs to be able to record two full callchains, one
1027 * from the first rule set traversal, plus one table reentrancy
1028 * via -j TEE without clobbering the callchain that brought us to
1031 * This is done by allocating two jumpstacks per cpu, on reentry
1032 * the upper half of the stack is used.
1034 * see the jumpstack setup in ipt_do_table() for more details.
1036 size = sizeof(void *) * i->stacksize * 2u;
1037 for_each_possible_cpu(cpu) {
1038 if (size > PAGE_SIZE)
1039 i->jumpstack[cpu] = vmalloc_node(size,
1042 i->jumpstack[cpu] = kmalloc_node(size,
1043 GFP_KERNEL, cpu_to_node(cpu));
1044 if (i->jumpstack[cpu] == NULL)
1046 * Freeing will be done later on by the callers. The
1047 * chain is: xt_replace_table -> __do_replace ->
1048 * do_replace -> xt_free_table_info.
1056 struct xt_table_info *
1057 xt_replace_table(struct xt_table *table,
1058 unsigned int num_counters,
1059 struct xt_table_info *newinfo,
1062 struct xt_table_info *private;
1065 ret = xt_jumpstack_alloc(newinfo);
1071 /* Do the substitution. */
1073 private = table->private;
1075 /* Check inside lock: is the old number correct? */
1076 if (num_counters != private->number) {
1077 pr_debug("num_counters != table->private->number (%u/%u)\n",
1078 num_counters, private->number);
1084 newinfo->initial_entries = private->initial_entries;
1086 * Ensure contents of newinfo are visible before assigning to
1090 table->private = newinfo;
1093 * Even though table entries have now been swapped, other CPU's
1094 * may still be using the old entries. This is okay, because
1095 * resynchronization happens because of the locking done
1096 * during the get_counters() routine.
1101 if (audit_enabled) {
1102 struct audit_buffer *ab;
1104 ab = audit_log_start(current->audit_context, GFP_KERNEL,
1105 AUDIT_NETFILTER_CFG);
1107 audit_log_format(ab, "table=%s family=%u entries=%u",
1108 table->name, table->af,
1117 EXPORT_SYMBOL_GPL(xt_replace_table);
1119 struct xt_table *xt_register_table(struct net *net,
1120 const struct xt_table *input_table,
1121 struct xt_table_info *bootstrap,
1122 struct xt_table_info *newinfo)
1125 struct xt_table_info *private;
1126 struct xt_table *t, *table;
1128 /* Don't add one object to multiple lists. */
1129 table = kmemdup(input_table, sizeof(struct xt_table), GFP_KERNEL);
1135 mutex_lock(&xt[table->af].mutex);
1136 /* Don't autoload: we'd eat our tail... */
1137 list_for_each_entry(t, &net->xt.tables[table->af], list) {
1138 if (strcmp(t->name, table->name) == 0) {
1144 /* Simplifies replace_table code. */
1145 table->private = bootstrap;
1147 if (!xt_replace_table(table, 0, newinfo, &ret))
1150 private = table->private;
1151 pr_debug("table->private->number = %u\n", private->number);
1153 /* save number of initial entries */
1154 private->initial_entries = private->number;
1156 list_add(&table->list, &net->xt.tables[table->af]);
1157 mutex_unlock(&xt[table->af].mutex);
1161 mutex_unlock(&xt[table->af].mutex);
1164 return ERR_PTR(ret);
1166 EXPORT_SYMBOL_GPL(xt_register_table);
1168 void *xt_unregister_table(struct xt_table *table)
1170 struct xt_table_info *private;
1172 mutex_lock(&xt[table->af].mutex);
1173 private = table->private;
1174 list_del(&table->list);
1175 mutex_unlock(&xt[table->af].mutex);
1180 EXPORT_SYMBOL_GPL(xt_unregister_table);
1182 #ifdef CONFIG_PROC_FS
1183 struct xt_names_priv {
1184 struct seq_net_private p;
1187 static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos)
1189 struct xt_names_priv *priv = seq->private;
1190 struct net *net = seq_file_net(seq);
1191 u_int8_t af = priv->af;
1193 mutex_lock(&xt[af].mutex);
1194 return seq_list_start(&net->xt.tables[af], *pos);
1197 static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1199 struct xt_names_priv *priv = seq->private;
1200 struct net *net = seq_file_net(seq);
1201 u_int8_t af = priv->af;
1203 return seq_list_next(v, &net->xt.tables[af], pos);
1206 static void xt_table_seq_stop(struct seq_file *seq, void *v)
1208 struct xt_names_priv *priv = seq->private;
1209 u_int8_t af = priv->af;
1211 mutex_unlock(&xt[af].mutex);
1214 static int xt_table_seq_show(struct seq_file *seq, void *v)
1216 struct xt_table *table = list_entry(v, struct xt_table, list);
1219 seq_printf(seq, "%s\n", table->name);
1223 static const struct seq_operations xt_table_seq_ops = {
1224 .start = xt_table_seq_start,
1225 .next = xt_table_seq_next,
1226 .stop = xt_table_seq_stop,
1227 .show = xt_table_seq_show,
1230 static int xt_table_open(struct inode *inode, struct file *file)
1233 struct xt_names_priv *priv;
1235 ret = seq_open_net(inode, file, &xt_table_seq_ops,
1236 sizeof(struct xt_names_priv));
1238 priv = ((struct seq_file *)file->private_data)->private;
1239 priv->af = (unsigned long)PDE_DATA(inode);
1244 static const struct file_operations xt_table_ops = {
1245 .owner = THIS_MODULE,
1246 .open = xt_table_open,
1248 .llseek = seq_lseek,
1249 .release = seq_release_net,
1253 * Traverse state for ip{,6}_{tables,matches} for helping crossing
1254 * the multi-AF mutexes.
1256 struct nf_mttg_trav {
1257 struct list_head *head, *curr;
1258 uint8_t class, nfproto;
1263 MTTG_TRAV_NFP_UNSPEC,
1268 static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos,
1271 static const uint8_t next_class[] = {
1272 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC,
1273 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE,
1275 struct nf_mttg_trav *trav = seq->private;
1277 switch (trav->class) {
1278 case MTTG_TRAV_INIT:
1279 trav->class = MTTG_TRAV_NFP_UNSPEC;
1280 mutex_lock(&xt[NFPROTO_UNSPEC].mutex);
1281 trav->head = trav->curr = is_target ?
1282 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match;
1284 case MTTG_TRAV_NFP_UNSPEC:
1285 trav->curr = trav->curr->next;
1286 if (trav->curr != trav->head)
1288 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1289 mutex_lock(&xt[trav->nfproto].mutex);
1290 trav->head = trav->curr = is_target ?
1291 &xt[trav->nfproto].target : &xt[trav->nfproto].match;
1292 trav->class = next_class[trav->class];
1294 case MTTG_TRAV_NFP_SPEC:
1295 trav->curr = trav->curr->next;
1296 if (trav->curr != trav->head)
1298 /* fallthru, _stop will unlock */
1308 static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos,
1311 struct nf_mttg_trav *trav = seq->private;
1314 trav->class = MTTG_TRAV_INIT;
1315 for (j = 0; j < *pos; ++j)
1316 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL)
1321 static void xt_mttg_seq_stop(struct seq_file *seq, void *v)
1323 struct nf_mttg_trav *trav = seq->private;
1325 switch (trav->class) {
1326 case MTTG_TRAV_NFP_UNSPEC:
1327 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1329 case MTTG_TRAV_NFP_SPEC:
1330 mutex_unlock(&xt[trav->nfproto].mutex);
1335 static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos)
1337 return xt_mttg_seq_start(seq, pos, false);
1340 static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1342 return xt_mttg_seq_next(seq, v, ppos, false);
1345 static int xt_match_seq_show(struct seq_file *seq, void *v)
1347 const struct nf_mttg_trav *trav = seq->private;
1348 const struct xt_match *match;
1350 switch (trav->class) {
1351 case MTTG_TRAV_NFP_UNSPEC:
1352 case MTTG_TRAV_NFP_SPEC:
1353 if (trav->curr == trav->head)
1355 match = list_entry(trav->curr, struct xt_match, list);
1357 seq_printf(seq, "%s\n", match->name);
1362 static const struct seq_operations xt_match_seq_ops = {
1363 .start = xt_match_seq_start,
1364 .next = xt_match_seq_next,
1365 .stop = xt_mttg_seq_stop,
1366 .show = xt_match_seq_show,
1369 static int xt_match_open(struct inode *inode, struct file *file)
1371 struct nf_mttg_trav *trav;
1372 trav = __seq_open_private(file, &xt_match_seq_ops, sizeof(*trav));
1376 trav->nfproto = (unsigned long)PDE_DATA(inode);
1380 static const struct file_operations xt_match_ops = {
1381 .owner = THIS_MODULE,
1382 .open = xt_match_open,
1384 .llseek = seq_lseek,
1385 .release = seq_release_private,
1388 static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos)
1390 return xt_mttg_seq_start(seq, pos, true);
1393 static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1395 return xt_mttg_seq_next(seq, v, ppos, true);
1398 static int xt_target_seq_show(struct seq_file *seq, void *v)
1400 const struct nf_mttg_trav *trav = seq->private;
1401 const struct xt_target *target;
1403 switch (trav->class) {
1404 case MTTG_TRAV_NFP_UNSPEC:
1405 case MTTG_TRAV_NFP_SPEC:
1406 if (trav->curr == trav->head)
1408 target = list_entry(trav->curr, struct xt_target, list);
1410 seq_printf(seq, "%s\n", target->name);
1415 static const struct seq_operations xt_target_seq_ops = {
1416 .start = xt_target_seq_start,
1417 .next = xt_target_seq_next,
1418 .stop = xt_mttg_seq_stop,
1419 .show = xt_target_seq_show,
1422 static int xt_target_open(struct inode *inode, struct file *file)
1424 struct nf_mttg_trav *trav;
1425 trav = __seq_open_private(file, &xt_target_seq_ops, sizeof(*trav));
1429 trav->nfproto = (unsigned long)PDE_DATA(inode);
1433 static const struct file_operations xt_target_ops = {
1434 .owner = THIS_MODULE,
1435 .open = xt_target_open,
1437 .llseek = seq_lseek,
1438 .release = seq_release_private,
1441 #define FORMAT_TABLES "_tables_names"
1442 #define FORMAT_MATCHES "_tables_matches"
1443 #define FORMAT_TARGETS "_tables_targets"
1445 #endif /* CONFIG_PROC_FS */
1448 * xt_hook_ops_alloc - set up hooks for a new table
1449 * @table: table with metadata needed to set up hooks
1450 * @fn: Hook function
1452 * This function will create the nf_hook_ops that the x_table needs
1453 * to hand to xt_hook_link_net().
1455 struct nf_hook_ops *
1456 xt_hook_ops_alloc(const struct xt_table *table, nf_hookfn *fn)
1458 unsigned int hook_mask = table->valid_hooks;
1459 uint8_t i, num_hooks = hweight32(hook_mask);
1461 struct nf_hook_ops *ops;
1463 ops = kmalloc(sizeof(*ops) * num_hooks, GFP_KERNEL);
1465 return ERR_PTR(-ENOMEM);
1467 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0;
1468 hook_mask >>= 1, ++hooknum) {
1469 if (!(hook_mask & 1))
1472 ops[i].pf = table->af;
1473 ops[i].hooknum = hooknum;
1474 ops[i].priority = table->priority;
1480 EXPORT_SYMBOL_GPL(xt_hook_ops_alloc);
1482 int xt_proto_init(struct net *net, u_int8_t af)
1484 #ifdef CONFIG_PROC_FS
1485 char buf[XT_FUNCTION_MAXNAMELEN];
1486 struct proc_dir_entry *proc;
1491 if (af >= ARRAY_SIZE(xt_prefix))
1495 #ifdef CONFIG_PROC_FS
1496 root_uid = make_kuid(net->user_ns, 0);
1497 root_gid = make_kgid(net->user_ns, 0);
1499 strlcpy(buf, xt_prefix[af], sizeof(buf));
1500 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1501 proc = proc_create_data(buf, 0440, net->proc_net, &xt_table_ops,
1502 (void *)(unsigned long)af);
1505 if (uid_valid(root_uid) && gid_valid(root_gid))
1506 proc_set_user(proc, root_uid, root_gid);
1508 strlcpy(buf, xt_prefix[af], sizeof(buf));
1509 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1510 proc = proc_create_data(buf, 0440, net->proc_net, &xt_match_ops,
1511 (void *)(unsigned long)af);
1513 goto out_remove_tables;
1514 if (uid_valid(root_uid) && gid_valid(root_gid))
1515 proc_set_user(proc, root_uid, root_gid);
1517 strlcpy(buf, xt_prefix[af], sizeof(buf));
1518 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1519 proc = proc_create_data(buf, 0440, net->proc_net, &xt_target_ops,
1520 (void *)(unsigned long)af);
1522 goto out_remove_matches;
1523 if (uid_valid(root_uid) && gid_valid(root_gid))
1524 proc_set_user(proc, root_uid, root_gid);
1529 #ifdef CONFIG_PROC_FS
1531 strlcpy(buf, xt_prefix[af], sizeof(buf));
1532 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1533 remove_proc_entry(buf, net->proc_net);
1536 strlcpy(buf, xt_prefix[af], sizeof(buf));
1537 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1538 remove_proc_entry(buf, net->proc_net);
1543 EXPORT_SYMBOL_GPL(xt_proto_init);
1545 void xt_proto_fini(struct net *net, u_int8_t af)
1547 #ifdef CONFIG_PROC_FS
1548 char buf[XT_FUNCTION_MAXNAMELEN];
1550 strlcpy(buf, xt_prefix[af], sizeof(buf));
1551 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1552 remove_proc_entry(buf, net->proc_net);
1554 strlcpy(buf, xt_prefix[af], sizeof(buf));
1555 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1556 remove_proc_entry(buf, net->proc_net);
1558 strlcpy(buf, xt_prefix[af], sizeof(buf));
1559 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1560 remove_proc_entry(buf, net->proc_net);
1561 #endif /*CONFIG_PROC_FS*/
1563 EXPORT_SYMBOL_GPL(xt_proto_fini);
1565 static int __net_init xt_net_init(struct net *net)
1569 for (i = 0; i < NFPROTO_NUMPROTO; i++)
1570 INIT_LIST_HEAD(&net->xt.tables[i]);
1574 static struct pernet_operations xt_net_ops = {
1575 .init = xt_net_init,
1578 static int __init xt_init(void)
1583 for_each_possible_cpu(i) {
1584 seqcount_init(&per_cpu(xt_recseq, i));
1587 xt = kmalloc(sizeof(struct xt_af) * NFPROTO_NUMPROTO, GFP_KERNEL);
1591 for (i = 0; i < NFPROTO_NUMPROTO; i++) {
1592 mutex_init(&xt[i].mutex);
1593 #ifdef CONFIG_COMPAT
1594 mutex_init(&xt[i].compat_mutex);
1595 xt[i].compat_tab = NULL;
1597 INIT_LIST_HEAD(&xt[i].target);
1598 INIT_LIST_HEAD(&xt[i].match);
1600 rv = register_pernet_subsys(&xt_net_ops);
1606 static void __exit xt_fini(void)
1608 unregister_pernet_subsys(&xt_net_ops);
1612 module_init(xt_init);
1613 module_exit(xt_fini);