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