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 SMP_ALIGN(x) (((x) + SMP_CACHE_BYTES-1) & ~(SMP_CACHE_BYTES-1))
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 match = xt_find_match(nfproto, name, revision);
213 request_module("%st_%s", xt_prefix[nfproto], name);
214 match = xt_find_match(nfproto, name, revision);
219 EXPORT_SYMBOL_GPL(xt_request_find_match);
221 /* Find target, grabs ref. Returns ERR_PTR() on error. */
222 struct xt_target *xt_find_target(u8 af, const char *name, u8 revision)
227 mutex_lock(&xt[af].mutex);
228 list_for_each_entry(t, &xt[af].target, list) {
229 if (strcmp(t->name, name) == 0) {
230 if (t->revision == revision) {
231 if (try_module_get(t->me)) {
232 mutex_unlock(&xt[af].mutex);
236 err = -EPROTOTYPE; /* Found something. */
239 mutex_unlock(&xt[af].mutex);
241 if (af != NFPROTO_UNSPEC)
242 /* Try searching again in the family-independent list */
243 return xt_find_target(NFPROTO_UNSPEC, name, revision);
247 EXPORT_SYMBOL(xt_find_target);
249 struct xt_target *xt_request_find_target(u8 af, const char *name, u8 revision)
251 struct xt_target *target;
253 target = xt_find_target(af, name, revision);
254 if (IS_ERR(target)) {
255 request_module("%st_%s", xt_prefix[af], name);
256 target = xt_find_target(af, name, revision);
261 EXPORT_SYMBOL_GPL(xt_request_find_target);
263 static int match_revfn(u8 af, const char *name, u8 revision, int *bestp)
265 const struct xt_match *m;
268 list_for_each_entry(m, &xt[af].match, list) {
269 if (strcmp(m->name, name) == 0) {
270 if (m->revision > *bestp)
271 *bestp = m->revision;
272 if (m->revision == revision)
277 if (af != NFPROTO_UNSPEC && !have_rev)
278 return match_revfn(NFPROTO_UNSPEC, name, revision, bestp);
283 static int target_revfn(u8 af, const char *name, u8 revision, int *bestp)
285 const struct xt_target *t;
288 list_for_each_entry(t, &xt[af].target, list) {
289 if (strcmp(t->name, name) == 0) {
290 if (t->revision > *bestp)
291 *bestp = t->revision;
292 if (t->revision == revision)
297 if (af != NFPROTO_UNSPEC && !have_rev)
298 return target_revfn(NFPROTO_UNSPEC, name, revision, bestp);
303 /* Returns true or false (if no such extension at all) */
304 int xt_find_revision(u8 af, const char *name, u8 revision, int target,
307 int have_rev, best = -1;
309 mutex_lock(&xt[af].mutex);
311 have_rev = target_revfn(af, name, revision, &best);
313 have_rev = match_revfn(af, name, revision, &best);
314 mutex_unlock(&xt[af].mutex);
316 /* Nothing at all? Return 0 to try loading module. */
324 *err = -EPROTONOSUPPORT;
327 EXPORT_SYMBOL_GPL(xt_find_revision);
330 textify_hooks(char *buf, size_t size, unsigned int mask, uint8_t nfproto)
332 static const char *const inetbr_names[] = {
333 "PREROUTING", "INPUT", "FORWARD",
334 "OUTPUT", "POSTROUTING", "BROUTING",
336 static const char *const arp_names[] = {
337 "INPUT", "FORWARD", "OUTPUT",
339 const char *const *names;
345 names = (nfproto == NFPROTO_ARP) ? arp_names : inetbr_names;
346 max = (nfproto == NFPROTO_ARP) ? ARRAY_SIZE(arp_names) :
347 ARRAY_SIZE(inetbr_names);
349 for (i = 0; i < max; ++i) {
350 if (!(mask & (1 << i)))
352 res = snprintf(p, size, "%s%s", np ? "/" : "", names[i]);
363 int xt_check_match(struct xt_mtchk_param *par,
364 unsigned int size, u_int8_t proto, bool inv_proto)
368 if (XT_ALIGN(par->match->matchsize) != size &&
369 par->match->matchsize != -1) {
371 * ebt_among is exempt from centralized matchsize checking
372 * because it uses a dynamic-size data set.
374 pr_err("%s_tables: %s.%u match: invalid size "
375 "%u (kernel) != (user) %u\n",
376 xt_prefix[par->family], par->match->name,
377 par->match->revision,
378 XT_ALIGN(par->match->matchsize), size);
381 if (par->match->table != NULL &&
382 strcmp(par->match->table, par->table) != 0) {
383 pr_err("%s_tables: %s match: only valid in %s table, not %s\n",
384 xt_prefix[par->family], par->match->name,
385 par->match->table, par->table);
388 if (par->match->hooks && (par->hook_mask & ~par->match->hooks) != 0) {
389 char used[64], allow[64];
391 pr_err("%s_tables: %s match: used from hooks %s, but only "
393 xt_prefix[par->family], par->match->name,
394 textify_hooks(used, sizeof(used), par->hook_mask,
396 textify_hooks(allow, sizeof(allow), par->match->hooks,
400 if (par->match->proto && (par->match->proto != proto || inv_proto)) {
401 pr_err("%s_tables: %s match: only valid for protocol %u\n",
402 xt_prefix[par->family], par->match->name,
406 if (par->match->checkentry != NULL) {
407 ret = par->match->checkentry(par);
411 /* Flag up potential errors. */
416 EXPORT_SYMBOL_GPL(xt_check_match);
418 /** xt_check_entry_match - check that matches end before start of target
420 * @match: beginning of xt_entry_match
421 * @target: beginning of this rules target (alleged end of matches)
422 * @alignment: alignment requirement of match structures
424 * Validates that all matches add up to the beginning of the target,
425 * and that each match covers at least the base structure size.
427 * Return: 0 on success, negative errno on failure.
429 static int xt_check_entry_match(const char *match, const char *target,
430 const size_t alignment)
432 const struct xt_entry_match *pos;
433 int length = target - match;
435 if (length == 0) /* no matches */
438 pos = (struct xt_entry_match *)match;
440 if ((unsigned long)pos % alignment)
443 if (length < (int)sizeof(struct xt_entry_match))
446 if (pos->u.match_size < sizeof(struct xt_entry_match))
449 if (pos->u.match_size > length)
452 length -= pos->u.match_size;
453 pos = ((void *)((char *)(pos) + (pos)->u.match_size));
454 } while (length > 0);
460 int xt_compat_add_offset(u_int8_t af, unsigned int offset, int delta)
462 struct xt_af *xp = &xt[af];
464 if (!xp->compat_tab) {
467 xp->compat_tab = vmalloc(sizeof(struct compat_delta) * xp->number);
473 if (xp->cur >= xp->number)
477 delta += xp->compat_tab[xp->cur - 1].delta;
478 xp->compat_tab[xp->cur].offset = offset;
479 xp->compat_tab[xp->cur].delta = delta;
483 EXPORT_SYMBOL_GPL(xt_compat_add_offset);
485 void xt_compat_flush_offsets(u_int8_t af)
487 if (xt[af].compat_tab) {
488 vfree(xt[af].compat_tab);
489 xt[af].compat_tab = NULL;
494 EXPORT_SYMBOL_GPL(xt_compat_flush_offsets);
496 int xt_compat_calc_jump(u_int8_t af, unsigned int offset)
498 struct compat_delta *tmp = xt[af].compat_tab;
499 int mid, left = 0, right = xt[af].cur - 1;
501 while (left <= right) {
502 mid = (left + right) >> 1;
503 if (offset > tmp[mid].offset)
505 else if (offset < tmp[mid].offset)
508 return mid ? tmp[mid - 1].delta : 0;
510 return left ? tmp[left - 1].delta : 0;
512 EXPORT_SYMBOL_GPL(xt_compat_calc_jump);
514 void xt_compat_init_offsets(u_int8_t af, unsigned int number)
516 xt[af].number = number;
519 EXPORT_SYMBOL(xt_compat_init_offsets);
521 int xt_compat_match_offset(const struct xt_match *match)
523 u_int16_t csize = match->compatsize ? : match->matchsize;
524 return XT_ALIGN(match->matchsize) - COMPAT_XT_ALIGN(csize);
526 EXPORT_SYMBOL_GPL(xt_compat_match_offset);
528 int xt_compat_match_from_user(struct xt_entry_match *m, void **dstptr,
531 const struct xt_match *match = m->u.kernel.match;
532 struct compat_xt_entry_match *cm = (struct compat_xt_entry_match *)m;
533 int pad, off = xt_compat_match_offset(match);
534 u_int16_t msize = cm->u.user.match_size;
537 memcpy(m, cm, sizeof(*cm));
538 if (match->compat_from_user)
539 match->compat_from_user(m->data, cm->data);
541 memcpy(m->data, cm->data, msize - sizeof(*cm));
542 pad = XT_ALIGN(match->matchsize) - match->matchsize;
544 memset(m->data + match->matchsize, 0, pad);
547 m->u.user.match_size = msize;
553 EXPORT_SYMBOL_GPL(xt_compat_match_from_user);
555 int xt_compat_match_to_user(const struct xt_entry_match *m,
556 void __user **dstptr, unsigned int *size)
558 const struct xt_match *match = m->u.kernel.match;
559 struct compat_xt_entry_match __user *cm = *dstptr;
560 int off = xt_compat_match_offset(match);
561 u_int16_t msize = m->u.user.match_size - off;
563 if (copy_to_user(cm, m, sizeof(*cm)) ||
564 put_user(msize, &cm->u.user.match_size) ||
565 copy_to_user(cm->u.user.name, m->u.kernel.match->name,
566 strlen(m->u.kernel.match->name) + 1))
569 if (match->compat_to_user) {
570 if (match->compat_to_user((void __user *)cm->data, m->data))
573 if (copy_to_user(cm->data, m->data, msize - sizeof(*cm)))
581 EXPORT_SYMBOL_GPL(xt_compat_match_to_user);
583 /* non-compat version may have padding after verdict */
584 struct compat_xt_standard_target {
585 struct compat_xt_entry_target t;
586 compat_uint_t verdict;
589 int xt_compat_check_entry_offsets(const void *base, const char *elems,
590 unsigned int target_offset,
591 unsigned int next_offset)
593 long size_of_base_struct = elems - (const char *)base;
594 const struct compat_xt_entry_target *t;
595 const char *e = base;
597 if (target_offset < size_of_base_struct)
600 if (target_offset + sizeof(*t) > next_offset)
603 t = (void *)(e + target_offset);
604 if (t->u.target_size < sizeof(*t))
607 if (target_offset + t->u.target_size > next_offset)
610 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
611 target_offset + sizeof(struct compat_xt_standard_target) != next_offset)
614 /* compat_xt_entry match has less strict aligment requirements,
615 * otherwise they are identical. In case of padding differences
616 * we need to add compat version of xt_check_entry_match.
618 BUILD_BUG_ON(sizeof(struct compat_xt_entry_match) != sizeof(struct xt_entry_match));
620 return xt_check_entry_match(elems, base + target_offset,
621 __alignof__(struct compat_xt_entry_match));
623 EXPORT_SYMBOL(xt_compat_check_entry_offsets);
624 #endif /* CONFIG_COMPAT */
627 * xt_check_entry_offsets - validate arp/ip/ip6t_entry
629 * @base: pointer to arp/ip/ip6t_entry
630 * @elems: pointer to first xt_entry_match, i.e. ip(6)t_entry->elems
631 * @target_offset: the arp/ip/ip6_t->target_offset
632 * @next_offset: the arp/ip/ip6_t->next_offset
634 * validates that target_offset and next_offset are sane and that all
635 * match sizes (if any) align with the target offset.
637 * This function does not validate the targets or matches themselves, it
638 * only tests that all the offsets and sizes are correct, that all
639 * match structures are aligned, and that the last structure ends where
640 * the target structure begins.
642 * Also see xt_compat_check_entry_offsets for CONFIG_COMPAT version.
644 * The arp/ip/ip6t_entry structure @base must have passed following tests:
645 * - it must point to a valid memory location
646 * - base to base + next_offset must be accessible, i.e. not exceed allocated
649 * A well-formed entry looks like this:
651 * ip(6)t_entry match [mtdata] match [mtdata] target [tgdata] ip(6)t_entry
652 * e->elems[]-----' | |
656 * target_offset---------------------------------' |
657 * next_offset---------------------------------------------------'
659 * elems[]: flexible array member at end of ip(6)/arpt_entry struct.
660 * This is where matches (if any) and the target reside.
661 * target_offset: beginning of target.
662 * next_offset: start of the next rule; also: size of this rule.
663 * Since targets have a minimum size, target_offset + minlen <= next_offset.
665 * Every match stores its size, sum of sizes must not exceed target_offset.
667 * Return: 0 on success, negative errno on failure.
669 int xt_check_entry_offsets(const void *base,
671 unsigned int target_offset,
672 unsigned int next_offset)
674 long size_of_base_struct = elems - (const char *)base;
675 const struct xt_entry_target *t;
676 const char *e = base;
678 /* target start is within the ip/ip6/arpt_entry struct */
679 if (target_offset < size_of_base_struct)
682 if (target_offset + sizeof(*t) > next_offset)
685 t = (void *)(e + target_offset);
686 if (t->u.target_size < sizeof(*t))
689 if (target_offset + t->u.target_size > next_offset)
692 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
693 target_offset + sizeof(struct xt_standard_target) != next_offset)
696 return xt_check_entry_match(elems, base + target_offset,
697 __alignof__(struct xt_entry_match));
699 EXPORT_SYMBOL(xt_check_entry_offsets);
701 int xt_check_target(struct xt_tgchk_param *par,
702 unsigned int size, u_int8_t proto, bool inv_proto)
706 if (XT_ALIGN(par->target->targetsize) != size) {
707 pr_err("%s_tables: %s.%u target: invalid size "
708 "%u (kernel) != (user) %u\n",
709 xt_prefix[par->family], par->target->name,
710 par->target->revision,
711 XT_ALIGN(par->target->targetsize), size);
714 if (par->target->table != NULL &&
715 strcmp(par->target->table, par->table) != 0) {
716 pr_err("%s_tables: %s target: only valid in %s table, not %s\n",
717 xt_prefix[par->family], par->target->name,
718 par->target->table, par->table);
721 if (par->target->hooks && (par->hook_mask & ~par->target->hooks) != 0) {
722 char used[64], allow[64];
724 pr_err("%s_tables: %s target: used from hooks %s, but only "
726 xt_prefix[par->family], par->target->name,
727 textify_hooks(used, sizeof(used), par->hook_mask,
729 textify_hooks(allow, sizeof(allow), par->target->hooks,
733 if (par->target->proto && (par->target->proto != proto || inv_proto)) {
734 pr_err("%s_tables: %s target: only valid for protocol %u\n",
735 xt_prefix[par->family], par->target->name,
739 if (par->target->checkentry != NULL) {
740 ret = par->target->checkentry(par);
744 /* Flag up potential errors. */
749 EXPORT_SYMBOL_GPL(xt_check_target);
752 int xt_compat_target_offset(const struct xt_target *target)
754 u_int16_t csize = target->compatsize ? : target->targetsize;
755 return XT_ALIGN(target->targetsize) - COMPAT_XT_ALIGN(csize);
757 EXPORT_SYMBOL_GPL(xt_compat_target_offset);
759 void xt_compat_target_from_user(struct xt_entry_target *t, void **dstptr,
762 const struct xt_target *target = t->u.kernel.target;
763 struct compat_xt_entry_target *ct = (struct compat_xt_entry_target *)t;
764 int pad, off = xt_compat_target_offset(target);
765 u_int16_t tsize = ct->u.user.target_size;
768 memcpy(t, ct, sizeof(*ct));
769 if (target->compat_from_user)
770 target->compat_from_user(t->data, ct->data);
772 memcpy(t->data, ct->data, tsize - sizeof(*ct));
773 pad = XT_ALIGN(target->targetsize) - target->targetsize;
775 memset(t->data + target->targetsize, 0, pad);
778 t->u.user.target_size = tsize;
783 EXPORT_SYMBOL_GPL(xt_compat_target_from_user);
785 int xt_compat_target_to_user(const struct xt_entry_target *t,
786 void __user **dstptr, unsigned int *size)
788 const struct xt_target *target = t->u.kernel.target;
789 struct compat_xt_entry_target __user *ct = *dstptr;
790 int off = xt_compat_target_offset(target);
791 u_int16_t tsize = t->u.user.target_size - off;
793 if (copy_to_user(ct, t, sizeof(*ct)) ||
794 put_user(tsize, &ct->u.user.target_size) ||
795 copy_to_user(ct->u.user.name, t->u.kernel.target->name,
796 strlen(t->u.kernel.target->name) + 1))
799 if (target->compat_to_user) {
800 if (target->compat_to_user((void __user *)ct->data, t->data))
803 if (copy_to_user(ct->data, t->data, tsize - sizeof(*ct)))
811 EXPORT_SYMBOL_GPL(xt_compat_target_to_user);
814 struct xt_table_info *xt_alloc_table_info(unsigned int size)
816 struct xt_table_info *info = NULL;
817 size_t sz = sizeof(*info) + size;
819 /* Pedantry: prevent them from hitting BUG() in vmalloc.c --RR */
820 if ((SMP_ALIGN(size) >> PAGE_SHIFT) + 2 > totalram_pages)
823 if (sz <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER))
824 info = kmalloc(sz, GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY);
830 memset(info, 0, sizeof(*info));
834 EXPORT_SYMBOL(xt_alloc_table_info);
836 void xt_free_table_info(struct xt_table_info *info)
840 if (info->jumpstack != NULL) {
841 for_each_possible_cpu(cpu)
842 kvfree(info->jumpstack[cpu]);
843 kvfree(info->jumpstack);
848 EXPORT_SYMBOL(xt_free_table_info);
850 /* Find table by name, grabs mutex & ref. Returns ERR_PTR() on error. */
851 struct xt_table *xt_find_table_lock(struct net *net, u_int8_t af,
856 mutex_lock(&xt[af].mutex);
857 list_for_each_entry(t, &net->xt.tables[af], list)
858 if (strcmp(t->name, name) == 0 && try_module_get(t->me))
860 mutex_unlock(&xt[af].mutex);
863 EXPORT_SYMBOL_GPL(xt_find_table_lock);
865 void xt_table_unlock(struct xt_table *table)
867 mutex_unlock(&xt[table->af].mutex);
869 EXPORT_SYMBOL_GPL(xt_table_unlock);
872 void xt_compat_lock(u_int8_t af)
874 mutex_lock(&xt[af].compat_mutex);
876 EXPORT_SYMBOL_GPL(xt_compat_lock);
878 void xt_compat_unlock(u_int8_t af)
880 mutex_unlock(&xt[af].compat_mutex);
882 EXPORT_SYMBOL_GPL(xt_compat_unlock);
885 DEFINE_PER_CPU(seqcount_t, xt_recseq);
886 EXPORT_PER_CPU_SYMBOL_GPL(xt_recseq);
888 struct static_key xt_tee_enabled __read_mostly;
889 EXPORT_SYMBOL_GPL(xt_tee_enabled);
891 static int xt_jumpstack_alloc(struct xt_table_info *i)
896 size = sizeof(void **) * nr_cpu_ids;
897 if (size > PAGE_SIZE)
898 i->jumpstack = vzalloc(size);
900 i->jumpstack = kzalloc(size, GFP_KERNEL);
901 if (i->jumpstack == NULL)
904 /* ruleset without jumps -- no stack needed */
905 if (i->stacksize == 0)
908 /* Jumpstack needs to be able to record two full callchains, one
909 * from the first rule set traversal, plus one table reentrancy
910 * via -j TEE without clobbering the callchain that brought us to
913 * This is done by allocating two jumpstacks per cpu, on reentry
914 * the upper half of the stack is used.
916 * see the jumpstack setup in ipt_do_table() for more details.
918 size = sizeof(void *) * i->stacksize * 2u;
919 for_each_possible_cpu(cpu) {
920 if (size > PAGE_SIZE)
921 i->jumpstack[cpu] = vmalloc_node(size,
924 i->jumpstack[cpu] = kmalloc_node(size,
925 GFP_KERNEL, cpu_to_node(cpu));
926 if (i->jumpstack[cpu] == NULL)
928 * Freeing will be done later on by the callers. The
929 * chain is: xt_replace_table -> __do_replace ->
930 * do_replace -> xt_free_table_info.
938 struct xt_table_info *
939 xt_replace_table(struct xt_table *table,
940 unsigned int num_counters,
941 struct xt_table_info *newinfo,
944 struct xt_table_info *private;
947 ret = xt_jumpstack_alloc(newinfo);
953 /* Do the substitution. */
955 private = table->private;
957 /* Check inside lock: is the old number correct? */
958 if (num_counters != private->number) {
959 pr_debug("num_counters != table->private->number (%u/%u)\n",
960 num_counters, private->number);
966 newinfo->initial_entries = private->initial_entries;
968 * Ensure contents of newinfo are visible before assigning to
972 table->private = newinfo;
975 * Even though table entries have now been swapped, other CPU's
976 * may still be using the old entries. This is okay, because
977 * resynchronization happens because of the locking done
978 * during the get_counters() routine.
984 struct audit_buffer *ab;
986 ab = audit_log_start(current->audit_context, GFP_KERNEL,
987 AUDIT_NETFILTER_CFG);
989 audit_log_format(ab, "table=%s family=%u entries=%u",
990 table->name, table->af,
999 EXPORT_SYMBOL_GPL(xt_replace_table);
1001 struct xt_table *xt_register_table(struct net *net,
1002 const struct xt_table *input_table,
1003 struct xt_table_info *bootstrap,
1004 struct xt_table_info *newinfo)
1007 struct xt_table_info *private;
1008 struct xt_table *t, *table;
1010 /* Don't add one object to multiple lists. */
1011 table = kmemdup(input_table, sizeof(struct xt_table), GFP_KERNEL);
1017 mutex_lock(&xt[table->af].mutex);
1018 /* Don't autoload: we'd eat our tail... */
1019 list_for_each_entry(t, &net->xt.tables[table->af], list) {
1020 if (strcmp(t->name, table->name) == 0) {
1026 /* Simplifies replace_table code. */
1027 table->private = bootstrap;
1029 if (!xt_replace_table(table, 0, newinfo, &ret))
1032 private = table->private;
1033 pr_debug("table->private->number = %u\n", private->number);
1035 /* save number of initial entries */
1036 private->initial_entries = private->number;
1038 list_add(&table->list, &net->xt.tables[table->af]);
1039 mutex_unlock(&xt[table->af].mutex);
1043 mutex_unlock(&xt[table->af].mutex);
1046 return ERR_PTR(ret);
1048 EXPORT_SYMBOL_GPL(xt_register_table);
1050 void *xt_unregister_table(struct xt_table *table)
1052 struct xt_table_info *private;
1054 mutex_lock(&xt[table->af].mutex);
1055 private = table->private;
1056 list_del(&table->list);
1057 mutex_unlock(&xt[table->af].mutex);
1062 EXPORT_SYMBOL_GPL(xt_unregister_table);
1064 #ifdef CONFIG_PROC_FS
1065 struct xt_names_priv {
1066 struct seq_net_private p;
1069 static void *xt_table_seq_start(struct seq_file *seq, loff_t *pos)
1071 struct xt_names_priv *priv = seq->private;
1072 struct net *net = seq_file_net(seq);
1073 u_int8_t af = priv->af;
1075 mutex_lock(&xt[af].mutex);
1076 return seq_list_start(&net->xt.tables[af], *pos);
1079 static void *xt_table_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1081 struct xt_names_priv *priv = seq->private;
1082 struct net *net = seq_file_net(seq);
1083 u_int8_t af = priv->af;
1085 return seq_list_next(v, &net->xt.tables[af], pos);
1088 static void xt_table_seq_stop(struct seq_file *seq, void *v)
1090 struct xt_names_priv *priv = seq->private;
1091 u_int8_t af = priv->af;
1093 mutex_unlock(&xt[af].mutex);
1096 static int xt_table_seq_show(struct seq_file *seq, void *v)
1098 struct xt_table *table = list_entry(v, struct xt_table, list);
1101 seq_printf(seq, "%s\n", table->name);
1105 static const struct seq_operations xt_table_seq_ops = {
1106 .start = xt_table_seq_start,
1107 .next = xt_table_seq_next,
1108 .stop = xt_table_seq_stop,
1109 .show = xt_table_seq_show,
1112 static int xt_table_open(struct inode *inode, struct file *file)
1115 struct xt_names_priv *priv;
1117 ret = seq_open_net(inode, file, &xt_table_seq_ops,
1118 sizeof(struct xt_names_priv));
1120 priv = ((struct seq_file *)file->private_data)->private;
1121 priv->af = (unsigned long)PDE_DATA(inode);
1126 static const struct file_operations xt_table_ops = {
1127 .owner = THIS_MODULE,
1128 .open = xt_table_open,
1130 .llseek = seq_lseek,
1131 .release = seq_release_net,
1135 * Traverse state for ip{,6}_{tables,matches} for helping crossing
1136 * the multi-AF mutexes.
1138 struct nf_mttg_trav {
1139 struct list_head *head, *curr;
1140 uint8_t class, nfproto;
1145 MTTG_TRAV_NFP_UNSPEC,
1150 static void *xt_mttg_seq_next(struct seq_file *seq, void *v, loff_t *ppos,
1153 static const uint8_t next_class[] = {
1154 [MTTG_TRAV_NFP_UNSPEC] = MTTG_TRAV_NFP_SPEC,
1155 [MTTG_TRAV_NFP_SPEC] = MTTG_TRAV_DONE,
1157 struct nf_mttg_trav *trav = seq->private;
1159 switch (trav->class) {
1160 case MTTG_TRAV_INIT:
1161 trav->class = MTTG_TRAV_NFP_UNSPEC;
1162 mutex_lock(&xt[NFPROTO_UNSPEC].mutex);
1163 trav->head = trav->curr = is_target ?
1164 &xt[NFPROTO_UNSPEC].target : &xt[NFPROTO_UNSPEC].match;
1166 case MTTG_TRAV_NFP_UNSPEC:
1167 trav->curr = trav->curr->next;
1168 if (trav->curr != trav->head)
1170 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1171 mutex_lock(&xt[trav->nfproto].mutex);
1172 trav->head = trav->curr = is_target ?
1173 &xt[trav->nfproto].target : &xt[trav->nfproto].match;
1174 trav->class = next_class[trav->class];
1176 case MTTG_TRAV_NFP_SPEC:
1177 trav->curr = trav->curr->next;
1178 if (trav->curr != trav->head)
1180 /* fallthru, _stop will unlock */
1190 static void *xt_mttg_seq_start(struct seq_file *seq, loff_t *pos,
1193 struct nf_mttg_trav *trav = seq->private;
1196 trav->class = MTTG_TRAV_INIT;
1197 for (j = 0; j < *pos; ++j)
1198 if (xt_mttg_seq_next(seq, NULL, NULL, is_target) == NULL)
1203 static void xt_mttg_seq_stop(struct seq_file *seq, void *v)
1205 struct nf_mttg_trav *trav = seq->private;
1207 switch (trav->class) {
1208 case MTTG_TRAV_NFP_UNSPEC:
1209 mutex_unlock(&xt[NFPROTO_UNSPEC].mutex);
1211 case MTTG_TRAV_NFP_SPEC:
1212 mutex_unlock(&xt[trav->nfproto].mutex);
1217 static void *xt_match_seq_start(struct seq_file *seq, loff_t *pos)
1219 return xt_mttg_seq_start(seq, pos, false);
1222 static void *xt_match_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1224 return xt_mttg_seq_next(seq, v, ppos, false);
1227 static int xt_match_seq_show(struct seq_file *seq, void *v)
1229 const struct nf_mttg_trav *trav = seq->private;
1230 const struct xt_match *match;
1232 switch (trav->class) {
1233 case MTTG_TRAV_NFP_UNSPEC:
1234 case MTTG_TRAV_NFP_SPEC:
1235 if (trav->curr == trav->head)
1237 match = list_entry(trav->curr, struct xt_match, list);
1239 seq_printf(seq, "%s\n", match->name);
1244 static const struct seq_operations xt_match_seq_ops = {
1245 .start = xt_match_seq_start,
1246 .next = xt_match_seq_next,
1247 .stop = xt_mttg_seq_stop,
1248 .show = xt_match_seq_show,
1251 static int xt_match_open(struct inode *inode, struct file *file)
1253 struct nf_mttg_trav *trav;
1254 trav = __seq_open_private(file, &xt_match_seq_ops, sizeof(*trav));
1258 trav->nfproto = (unsigned long)PDE_DATA(inode);
1262 static const struct file_operations xt_match_ops = {
1263 .owner = THIS_MODULE,
1264 .open = xt_match_open,
1266 .llseek = seq_lseek,
1267 .release = seq_release_private,
1270 static void *xt_target_seq_start(struct seq_file *seq, loff_t *pos)
1272 return xt_mttg_seq_start(seq, pos, true);
1275 static void *xt_target_seq_next(struct seq_file *seq, void *v, loff_t *ppos)
1277 return xt_mttg_seq_next(seq, v, ppos, true);
1280 static int xt_target_seq_show(struct seq_file *seq, void *v)
1282 const struct nf_mttg_trav *trav = seq->private;
1283 const struct xt_target *target;
1285 switch (trav->class) {
1286 case MTTG_TRAV_NFP_UNSPEC:
1287 case MTTG_TRAV_NFP_SPEC:
1288 if (trav->curr == trav->head)
1290 target = list_entry(trav->curr, struct xt_target, list);
1292 seq_printf(seq, "%s\n", target->name);
1297 static const struct seq_operations xt_target_seq_ops = {
1298 .start = xt_target_seq_start,
1299 .next = xt_target_seq_next,
1300 .stop = xt_mttg_seq_stop,
1301 .show = xt_target_seq_show,
1304 static int xt_target_open(struct inode *inode, struct file *file)
1306 struct nf_mttg_trav *trav;
1307 trav = __seq_open_private(file, &xt_target_seq_ops, sizeof(*trav));
1311 trav->nfproto = (unsigned long)PDE_DATA(inode);
1315 static const struct file_operations xt_target_ops = {
1316 .owner = THIS_MODULE,
1317 .open = xt_target_open,
1319 .llseek = seq_lseek,
1320 .release = seq_release_private,
1323 #define FORMAT_TABLES "_tables_names"
1324 #define FORMAT_MATCHES "_tables_matches"
1325 #define FORMAT_TARGETS "_tables_targets"
1327 #endif /* CONFIG_PROC_FS */
1330 * xt_hook_link - set up hooks for a new table
1331 * @table: table with metadata needed to set up hooks
1332 * @fn: Hook function
1334 * This function will take care of creating and registering the necessary
1335 * Netfilter hooks for XT tables.
1337 struct nf_hook_ops *xt_hook_link(const struct xt_table *table, nf_hookfn *fn)
1339 unsigned int hook_mask = table->valid_hooks;
1340 uint8_t i, num_hooks = hweight32(hook_mask);
1342 struct nf_hook_ops *ops;
1345 ops = kmalloc(sizeof(*ops) * num_hooks, GFP_KERNEL);
1347 return ERR_PTR(-ENOMEM);
1349 for (i = 0, hooknum = 0; i < num_hooks && hook_mask != 0;
1350 hook_mask >>= 1, ++hooknum) {
1351 if (!(hook_mask & 1))
1354 ops[i].pf = table->af;
1355 ops[i].hooknum = hooknum;
1356 ops[i].priority = table->priority;
1360 ret = nf_register_hooks(ops, num_hooks);
1363 return ERR_PTR(ret);
1368 EXPORT_SYMBOL_GPL(xt_hook_link);
1371 * xt_hook_unlink - remove hooks for a table
1372 * @ops: nf_hook_ops array as returned by nf_hook_link
1373 * @hook_mask: the very same mask that was passed to nf_hook_link
1375 void xt_hook_unlink(const struct xt_table *table, struct nf_hook_ops *ops)
1377 nf_unregister_hooks(ops, hweight32(table->valid_hooks));
1380 EXPORT_SYMBOL_GPL(xt_hook_unlink);
1382 int xt_proto_init(struct net *net, u_int8_t af)
1384 #ifdef CONFIG_PROC_FS
1385 char buf[XT_FUNCTION_MAXNAMELEN];
1386 struct proc_dir_entry *proc;
1389 if (af >= ARRAY_SIZE(xt_prefix))
1393 #ifdef CONFIG_PROC_FS
1394 strlcpy(buf, xt_prefix[af], sizeof(buf));
1395 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1396 proc = proc_create_data(buf, 0440, net->proc_net, &xt_table_ops,
1397 (void *)(unsigned long)af);
1401 strlcpy(buf, xt_prefix[af], sizeof(buf));
1402 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1403 proc = proc_create_data(buf, 0440, net->proc_net, &xt_match_ops,
1404 (void *)(unsigned long)af);
1406 goto out_remove_tables;
1408 strlcpy(buf, xt_prefix[af], sizeof(buf));
1409 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1410 proc = proc_create_data(buf, 0440, net->proc_net, &xt_target_ops,
1411 (void *)(unsigned long)af);
1413 goto out_remove_matches;
1418 #ifdef CONFIG_PROC_FS
1420 strlcpy(buf, xt_prefix[af], sizeof(buf));
1421 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1422 remove_proc_entry(buf, net->proc_net);
1425 strlcpy(buf, xt_prefix[af], sizeof(buf));
1426 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1427 remove_proc_entry(buf, net->proc_net);
1432 EXPORT_SYMBOL_GPL(xt_proto_init);
1434 void xt_proto_fini(struct net *net, u_int8_t af)
1436 #ifdef CONFIG_PROC_FS
1437 char buf[XT_FUNCTION_MAXNAMELEN];
1439 strlcpy(buf, xt_prefix[af], sizeof(buf));
1440 strlcat(buf, FORMAT_TABLES, sizeof(buf));
1441 remove_proc_entry(buf, net->proc_net);
1443 strlcpy(buf, xt_prefix[af], sizeof(buf));
1444 strlcat(buf, FORMAT_TARGETS, sizeof(buf));
1445 remove_proc_entry(buf, net->proc_net);
1447 strlcpy(buf, xt_prefix[af], sizeof(buf));
1448 strlcat(buf, FORMAT_MATCHES, sizeof(buf));
1449 remove_proc_entry(buf, net->proc_net);
1450 #endif /*CONFIG_PROC_FS*/
1452 EXPORT_SYMBOL_GPL(xt_proto_fini);
1454 static int __net_init xt_net_init(struct net *net)
1458 for (i = 0; i < NFPROTO_NUMPROTO; i++)
1459 INIT_LIST_HEAD(&net->xt.tables[i]);
1463 static struct pernet_operations xt_net_ops = {
1464 .init = xt_net_init,
1467 static int __init xt_init(void)
1472 for_each_possible_cpu(i) {
1473 seqcount_init(&per_cpu(xt_recseq, i));
1476 xt = kmalloc(sizeof(struct xt_af) * NFPROTO_NUMPROTO, GFP_KERNEL);
1480 for (i = 0; i < NFPROTO_NUMPROTO; i++) {
1481 mutex_init(&xt[i].mutex);
1482 #ifdef CONFIG_COMPAT
1483 mutex_init(&xt[i].compat_mutex);
1484 xt[i].compat_tab = NULL;
1486 INIT_LIST_HEAD(&xt[i].target);
1487 INIT_LIST_HEAD(&xt[i].match);
1489 rv = register_pernet_subsys(&xt_net_ops);
1495 static void __exit xt_fini(void)
1497 unregister_pernet_subsys(&xt_net_ops);
1501 module_init(xt_init);
1502 module_exit(xt_fini);