2 * Copyright (c) 2006 Patrick McHardy <kaber@trash.net>
3 * Copyright © CC Computer Consultants GmbH, 2007 - 2008
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
9 * This is a replacement of the old ipt_recent module, which carried the
10 * following copyright notice:
12 * Author: Stephen Frost <sfrost@snowman.net>
13 * Copyright 2002-2003, Stephen Frost, 2.5.x port by laforge@netfilter.org
15 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
16 #include <linux/init.h>
18 #include <linux/ipv6.h>
19 #include <linux/module.h>
20 #include <linux/moduleparam.h>
21 #include <linux/proc_fs.h>
22 #include <linux/seq_file.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/list.h>
26 #include <linux/random.h>
27 #include <linux/jhash.h>
28 #include <linux/bitops.h>
29 #include <linux/skbuff.h>
30 #include <linux/inet.h>
31 #include <linux/slab.h>
32 #include <linux/vmalloc.h>
33 #include <net/net_namespace.h>
34 #include <net/netns/generic.h>
36 #include <linux/netfilter/x_tables.h>
37 #include <linux/netfilter/xt_recent.h>
39 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
40 MODULE_AUTHOR("Jan Engelhardt <jengelh@medozas.de>");
41 MODULE_DESCRIPTION("Xtables: \"recently-seen\" host matching");
42 MODULE_LICENSE("GPL");
43 MODULE_ALIAS("ipt_recent");
44 MODULE_ALIAS("ip6t_recent");
46 static unsigned int ip_list_tot __read_mostly = 100;
47 static unsigned int ip_list_hash_size __read_mostly;
48 static unsigned int ip_list_perms __read_mostly = 0644;
49 static unsigned int ip_list_uid __read_mostly;
50 static unsigned int ip_list_gid __read_mostly;
51 module_param(ip_list_tot, uint, 0400);
52 module_param(ip_list_hash_size, uint, 0400);
53 module_param(ip_list_perms, uint, 0400);
54 module_param(ip_list_uid, uint, S_IRUGO | S_IWUSR);
55 module_param(ip_list_gid, uint, S_IRUGO | S_IWUSR);
56 MODULE_PARM_DESC(ip_list_tot, "number of IPs to remember per list");
57 MODULE_PARM_DESC(ip_list_hash_size, "size of hash table used to look up IPs");
58 MODULE_PARM_DESC(ip_list_perms, "permissions on /proc/net/xt_recent/* files");
59 MODULE_PARM_DESC(ip_list_uid, "default owner of /proc/net/xt_recent/* files");
60 MODULE_PARM_DESC(ip_list_gid, "default owning group of /proc/net/xt_recent/* files");
62 /* retained for backwards compatibility */
63 static unsigned int ip_pkt_list_tot __read_mostly;
64 module_param(ip_pkt_list_tot, uint, 0400);
65 MODULE_PARM_DESC(ip_pkt_list_tot, "number of packets per IP address to remember (max. 255)");
67 #define XT_RECENT_MAX_NSTAMPS 256
70 struct list_head list;
71 struct list_head lru_list;
72 union nf_inet_addr addr;
77 unsigned long stamps[0];
81 struct list_head list;
82 char name[XT_RECENT_NAME_LEN];
83 union nf_inet_addr mask;
87 struct list_head lru_list;
88 struct list_head iphash[0];
92 struct list_head tables;
94 struct proc_dir_entry *xt_recent;
98 static int recent_net_id __read_mostly;
100 static inline struct recent_net *recent_pernet(struct net *net)
102 return net_generic(net, recent_net_id);
105 static DEFINE_SPINLOCK(recent_lock);
106 static DEFINE_MUTEX(recent_mutex);
108 #ifdef CONFIG_PROC_FS
109 static const struct file_operations recent_old_fops, recent_mt_fops;
112 static u_int32_t hash_rnd __read_mostly;
113 static bool hash_rnd_inited __read_mostly;
115 static inline unsigned int recent_entry_hash4(const union nf_inet_addr *addr)
117 return jhash_1word((__force u32)addr->ip, hash_rnd) &
118 (ip_list_hash_size - 1);
121 static inline unsigned int recent_entry_hash6(const union nf_inet_addr *addr)
123 return jhash2((u32 *)addr->ip6, ARRAY_SIZE(addr->ip6), hash_rnd) &
124 (ip_list_hash_size - 1);
127 static struct recent_entry *
128 recent_entry_lookup(const struct recent_table *table,
129 const union nf_inet_addr *addrp, u_int16_t family,
132 struct recent_entry *e;
135 if (family == NFPROTO_IPV4)
136 h = recent_entry_hash4(addrp);
138 h = recent_entry_hash6(addrp);
140 list_for_each_entry(e, &table->iphash[h], list)
141 if (e->family == family &&
142 memcmp(&e->addr, addrp, sizeof(e->addr)) == 0 &&
143 (ttl == e->ttl || ttl == 0 || e->ttl == 0))
148 static void recent_entry_remove(struct recent_table *t, struct recent_entry *e)
151 list_del(&e->lru_list);
157 * Drop entries with timestamps older then 'time'.
159 static void recent_entry_reap(struct recent_table *t, unsigned long time,
160 struct recent_entry *working, bool update)
162 struct recent_entry *e;
165 * The head of the LRU list is always the oldest entry.
167 e = list_entry(t->lru_list.next, struct recent_entry, lru_list);
170 * Do not reap the entry which are going to be updated.
172 if (e == working && update)
176 * The last time stamp is the most recent.
178 if (time_after(time, e->stamps[e->index-1]))
179 recent_entry_remove(t, e);
182 static struct recent_entry *
183 recent_entry_init(struct recent_table *t, const union nf_inet_addr *addr,
184 u_int16_t family, u_int8_t ttl)
186 struct recent_entry *e;
187 unsigned int nstamps_max = t->nstamps_max_mask;
189 if (t->entries >= ip_list_tot) {
190 e = list_entry(t->lru_list.next, struct recent_entry, lru_list);
191 recent_entry_remove(t, e);
195 e = kmalloc(sizeof(*e) + sizeof(e->stamps[0]) * nstamps_max,
199 memcpy(&e->addr, addr, sizeof(e->addr));
201 e->stamps[0] = jiffies;
205 if (family == NFPROTO_IPV4)
206 list_add_tail(&e->list, &t->iphash[recent_entry_hash4(addr)]);
208 list_add_tail(&e->list, &t->iphash[recent_entry_hash6(addr)]);
209 list_add_tail(&e->lru_list, &t->lru_list);
214 static void recent_entry_update(struct recent_table *t, struct recent_entry *e)
216 e->index &= t->nstamps_max_mask;
217 e->stamps[e->index++] = jiffies;
218 if (e->index > e->nstamps)
219 e->nstamps = e->index;
220 list_move_tail(&e->lru_list, &t->lru_list);
223 static struct recent_table *recent_table_lookup(struct recent_net *recent_net,
226 struct recent_table *t;
228 list_for_each_entry(t, &recent_net->tables, list)
229 if (!strcmp(t->name, name))
234 static void recent_table_flush(struct recent_table *t)
236 struct recent_entry *e, *next;
239 for (i = 0; i < ip_list_hash_size; i++)
240 list_for_each_entry_safe(e, next, &t->iphash[i], list)
241 recent_entry_remove(t, e);
245 recent_mt(const struct sk_buff *skb, struct xt_action_param *par)
247 struct net *net = par->net;
248 struct recent_net *recent_net = recent_pernet(net);
249 const struct xt_recent_mtinfo_v1 *info = par->matchinfo;
250 struct recent_table *t;
251 struct recent_entry *e;
252 union nf_inet_addr addr = {}, addr_mask;
254 bool ret = info->invert;
256 if (par->family == NFPROTO_IPV4) {
257 const struct iphdr *iph = ip_hdr(skb);
259 if (info->side == XT_RECENT_DEST)
260 addr.ip = iph->daddr;
262 addr.ip = iph->saddr;
266 const struct ipv6hdr *iph = ipv6_hdr(skb);
268 if (info->side == XT_RECENT_DEST)
269 memcpy(&addr.in6, &iph->daddr, sizeof(addr.in6));
271 memcpy(&addr.in6, &iph->saddr, sizeof(addr.in6));
273 ttl = iph->hop_limit;
276 /* use TTL as seen before forwarding */
277 if (par->out != NULL && skb->sk == NULL)
280 spin_lock_bh(&recent_lock);
281 t = recent_table_lookup(recent_net, info->name);
283 nf_inet_addr_mask(&addr, &addr_mask, &t->mask);
285 e = recent_entry_lookup(t, &addr_mask, par->family,
286 (info->check_set & XT_RECENT_TTL) ? ttl : 0);
288 if (!(info->check_set & XT_RECENT_SET))
290 e = recent_entry_init(t, &addr_mask, par->family, ttl);
297 if (info->check_set & XT_RECENT_SET)
299 else if (info->check_set & XT_RECENT_REMOVE) {
300 recent_entry_remove(t, e);
302 } else if (info->check_set & (XT_RECENT_CHECK | XT_RECENT_UPDATE)) {
303 unsigned long time = jiffies - info->seconds * HZ;
304 unsigned int i, hits = 0;
306 for (i = 0; i < e->nstamps; i++) {
307 if (info->seconds && time_after(time, e->stamps[i]))
309 if (!info->hit_count || ++hits >= info->hit_count) {
315 /* info->seconds must be non-zero */
316 if (info->check_set & XT_RECENT_REAP)
317 recent_entry_reap(t, time, e,
318 info->check_set & XT_RECENT_UPDATE && ret);
321 if (info->check_set & XT_RECENT_SET ||
322 (info->check_set & XT_RECENT_UPDATE && ret)) {
323 recent_entry_update(t, e);
327 spin_unlock_bh(&recent_lock);
331 static void recent_table_free(void *addr)
336 static int recent_mt_check(const struct xt_mtchk_param *par,
337 const struct xt_recent_mtinfo_v1 *info)
339 struct recent_net *recent_net = recent_pernet(par->net);
340 struct recent_table *t;
341 #ifdef CONFIG_PROC_FS
342 struct proc_dir_entry *pde;
346 unsigned int nstamp_mask;
351 if (unlikely(!hash_rnd_inited)) {
352 get_random_bytes(&hash_rnd, sizeof(hash_rnd));
353 hash_rnd_inited = true;
355 if (info->check_set & ~XT_RECENT_VALID_FLAGS) {
356 pr_info("Unsupported user space flags (%08x)\n",
360 if (hweight8(info->check_set &
361 (XT_RECENT_SET | XT_RECENT_REMOVE |
362 XT_RECENT_CHECK | XT_RECENT_UPDATE)) != 1)
364 if ((info->check_set & (XT_RECENT_SET | XT_RECENT_REMOVE)) &&
365 (info->seconds || info->hit_count ||
366 (info->check_set & XT_RECENT_MODIFIERS)))
368 if ((info->check_set & XT_RECENT_REAP) && !info->seconds)
370 if (info->hit_count >= XT_RECENT_MAX_NSTAMPS) {
371 pr_info("hitcount (%u) is larger than allowed maximum (%u)\n",
372 info->hit_count, XT_RECENT_MAX_NSTAMPS - 1);
375 ret = xt_check_proc_name(info->name, sizeof(info->name));
379 if (ip_pkt_list_tot && info->hit_count < ip_pkt_list_tot)
380 nstamp_mask = roundup_pow_of_two(ip_pkt_list_tot) - 1;
381 else if (info->hit_count)
382 nstamp_mask = roundup_pow_of_two(info->hit_count) - 1;
384 nstamp_mask = 32 - 1;
386 mutex_lock(&recent_mutex);
387 t = recent_table_lookup(recent_net, info->name);
389 if (nstamp_mask > t->nstamps_max_mask) {
390 spin_lock_bh(&recent_lock);
391 recent_table_flush(t);
392 t->nstamps_max_mask = nstamp_mask;
393 spin_unlock_bh(&recent_lock);
401 sz = sizeof(*t) + sizeof(t->iphash[0]) * ip_list_hash_size;
403 t = kzalloc(sz, GFP_KERNEL);
411 t->nstamps_max_mask = nstamp_mask;
413 memcpy(&t->mask, &info->mask, sizeof(t->mask));
414 strcpy(t->name, info->name);
415 INIT_LIST_HEAD(&t->lru_list);
416 for (i = 0; i < ip_list_hash_size; i++)
417 INIT_LIST_HEAD(&t->iphash[i]);
418 #ifdef CONFIG_PROC_FS
419 uid = make_kuid(&init_user_ns, ip_list_uid);
420 gid = make_kgid(&init_user_ns, ip_list_gid);
421 if (!uid_valid(uid) || !gid_valid(gid)) {
422 recent_table_free(t);
426 pde = proc_create_data(t->name, ip_list_perms, recent_net->xt_recent,
429 recent_table_free(t);
433 proc_set_user(pde, uid, gid);
435 spin_lock_bh(&recent_lock);
436 list_add_tail(&t->list, &recent_net->tables);
437 spin_unlock_bh(&recent_lock);
440 mutex_unlock(&recent_mutex);
444 static int recent_mt_check_v0(const struct xt_mtchk_param *par)
446 const struct xt_recent_mtinfo_v0 *info_v0 = par->matchinfo;
447 struct xt_recent_mtinfo_v1 info_v1;
449 /* Copy revision 0 structure to revision 1 */
450 memcpy(&info_v1, info_v0, sizeof(struct xt_recent_mtinfo));
451 /* Set default mask to ensure backward compatible behaviour */
452 memset(info_v1.mask.all, 0xFF, sizeof(info_v1.mask.all));
454 return recent_mt_check(par, &info_v1);
457 static int recent_mt_check_v1(const struct xt_mtchk_param *par)
459 return recent_mt_check(par, par->matchinfo);
462 static void recent_mt_destroy(const struct xt_mtdtor_param *par)
464 struct recent_net *recent_net = recent_pernet(par->net);
465 const struct xt_recent_mtinfo_v1 *info = par->matchinfo;
466 struct recent_table *t;
468 mutex_lock(&recent_mutex);
469 t = recent_table_lookup(recent_net, info->name);
470 if (--t->refcnt == 0) {
471 spin_lock_bh(&recent_lock);
473 spin_unlock_bh(&recent_lock);
474 #ifdef CONFIG_PROC_FS
475 if (recent_net->xt_recent != NULL)
476 remove_proc_entry(t->name, recent_net->xt_recent);
478 recent_table_flush(t);
479 recent_table_free(t);
481 mutex_unlock(&recent_mutex);
484 #ifdef CONFIG_PROC_FS
485 struct recent_iter_state {
486 const struct recent_table *table;
490 static void *recent_seq_start(struct seq_file *seq, loff_t *pos)
491 __acquires(recent_lock)
493 struct recent_iter_state *st = seq->private;
494 const struct recent_table *t = st->table;
495 struct recent_entry *e;
498 spin_lock_bh(&recent_lock);
500 for (st->bucket = 0; st->bucket < ip_list_hash_size; st->bucket++)
501 list_for_each_entry(e, &t->iphash[st->bucket], list)
507 static void *recent_seq_next(struct seq_file *seq, void *v, loff_t *pos)
509 struct recent_iter_state *st = seq->private;
510 const struct recent_table *t = st->table;
511 const struct recent_entry *e = v;
512 const struct list_head *head = e->list.next;
514 while (head == &t->iphash[st->bucket]) {
515 if (++st->bucket >= ip_list_hash_size)
517 head = t->iphash[st->bucket].next;
520 return list_entry(head, struct recent_entry, list);
523 static void recent_seq_stop(struct seq_file *s, void *v)
524 __releases(recent_lock)
526 spin_unlock_bh(&recent_lock);
529 static int recent_seq_show(struct seq_file *seq, void *v)
531 const struct recent_entry *e = v;
532 struct recent_iter_state *st = seq->private;
533 const struct recent_table *t = st->table;
536 i = (e->index - 1) & t->nstamps_max_mask;
538 if (e->family == NFPROTO_IPV4)
539 seq_printf(seq, "src=%pI4 ttl: %u last_seen: %lu oldest_pkt: %u",
540 &e->addr.ip, e->ttl, e->stamps[i], e->index);
542 seq_printf(seq, "src=%pI6 ttl: %u last_seen: %lu oldest_pkt: %u",
543 &e->addr.in6, e->ttl, e->stamps[i], e->index);
544 for (i = 0; i < e->nstamps; i++)
545 seq_printf(seq, "%s %lu", i ? "," : "", e->stamps[i]);
546 seq_printf(seq, "\n");
550 static const struct seq_operations recent_seq_ops = {
551 .start = recent_seq_start,
552 .next = recent_seq_next,
553 .stop = recent_seq_stop,
554 .show = recent_seq_show,
557 static int recent_seq_open(struct inode *inode, struct file *file)
559 struct recent_iter_state *st;
561 st = __seq_open_private(file, &recent_seq_ops, sizeof(*st));
565 st->table = PDE_DATA(inode);
570 recent_mt_proc_write(struct file *file, const char __user *input,
571 size_t size, loff_t *loff)
573 struct recent_table *t = PDE_DATA(file_inode(file));
574 struct recent_entry *e;
575 char buf[sizeof("+b335:1d35:1e55:dead:c0de:1715:5afe:c0de")];
577 union nf_inet_addr addr = {};
583 if (size > sizeof(buf))
585 if (copy_from_user(buf, input, size) != 0)
588 /* Strict protocol! */
592 case '/': /* flush table */
593 spin_lock_bh(&recent_lock);
594 recent_table_flush(t);
595 spin_unlock_bh(&recent_lock);
597 case '-': /* remove address */
600 case '+': /* add address */
604 pr_info("Need \"+ip\", \"-ip\" or \"/\"\n");
610 if (strnchr(c, size, ':') != NULL) {
611 family = NFPROTO_IPV6;
612 succ = in6_pton(c, size, (void *)&addr, '\n', NULL);
614 family = NFPROTO_IPV4;
615 succ = in4_pton(c, size, (void *)&addr, '\n', NULL);
619 pr_info("illegal address written to procfs\n");
623 spin_lock_bh(&recent_lock);
624 e = recent_entry_lookup(t, &addr, family, 0);
627 recent_entry_init(t, &addr, family, 0);
630 recent_entry_update(t, e);
632 recent_entry_remove(t, e);
634 spin_unlock_bh(&recent_lock);
635 /* Note we removed one above */
640 static const struct file_operations recent_mt_fops = {
641 .open = recent_seq_open,
643 .write = recent_mt_proc_write,
644 .release = seq_release_private,
645 .owner = THIS_MODULE,
649 static int __net_init recent_proc_net_init(struct net *net)
651 struct recent_net *recent_net = recent_pernet(net);
653 recent_net->xt_recent = proc_mkdir("xt_recent", net->proc_net);
654 if (!recent_net->xt_recent)
659 static void __net_exit recent_proc_net_exit(struct net *net)
661 struct recent_net *recent_net = recent_pernet(net);
662 struct recent_table *t;
664 /* recent_net_exit() is called before recent_mt_destroy(). Make sure
665 * that the parent xt_recent proc entry is is empty before trying to
668 spin_lock_bh(&recent_lock);
669 list_for_each_entry(t, &recent_net->tables, list)
670 remove_proc_entry(t->name, recent_net->xt_recent);
672 recent_net->xt_recent = NULL;
673 spin_unlock_bh(&recent_lock);
675 remove_proc_entry("xt_recent", net->proc_net);
678 static inline int recent_proc_net_init(struct net *net)
683 static inline void recent_proc_net_exit(struct net *net)
686 #endif /* CONFIG_PROC_FS */
688 static int __net_init recent_net_init(struct net *net)
690 struct recent_net *recent_net = recent_pernet(net);
692 INIT_LIST_HEAD(&recent_net->tables);
693 return recent_proc_net_init(net);
696 static void __net_exit recent_net_exit(struct net *net)
698 recent_proc_net_exit(net);
701 static struct pernet_operations recent_net_ops = {
702 .init = recent_net_init,
703 .exit = recent_net_exit,
704 .id = &recent_net_id,
705 .size = sizeof(struct recent_net),
708 static struct xt_match recent_mt_reg[] __read_mostly = {
712 .family = NFPROTO_IPV4,
714 .matchsize = sizeof(struct xt_recent_mtinfo),
715 .checkentry = recent_mt_check_v0,
716 .destroy = recent_mt_destroy,
722 .family = NFPROTO_IPV6,
724 .matchsize = sizeof(struct xt_recent_mtinfo),
725 .checkentry = recent_mt_check_v0,
726 .destroy = recent_mt_destroy,
732 .family = NFPROTO_IPV4,
734 .matchsize = sizeof(struct xt_recent_mtinfo_v1),
735 .checkentry = recent_mt_check_v1,
736 .destroy = recent_mt_destroy,
742 .family = NFPROTO_IPV6,
744 .matchsize = sizeof(struct xt_recent_mtinfo_v1),
745 .checkentry = recent_mt_check_v1,
746 .destroy = recent_mt_destroy,
751 static int __init recent_mt_init(void)
755 BUILD_BUG_ON_NOT_POWER_OF_2(XT_RECENT_MAX_NSTAMPS);
757 if (!ip_list_tot || ip_pkt_list_tot >= XT_RECENT_MAX_NSTAMPS)
759 ip_list_hash_size = 1 << fls(ip_list_tot);
761 err = register_pernet_subsys(&recent_net_ops);
764 err = xt_register_matches(recent_mt_reg, ARRAY_SIZE(recent_mt_reg));
766 unregister_pernet_subsys(&recent_net_ops);
770 static void __exit recent_mt_exit(void)
772 xt_unregister_matches(recent_mt_reg, ARRAY_SIZE(recent_mt_reg));
773 unregister_pernet_subsys(&recent_net_ops);
776 module_init(recent_mt_init);
777 module_exit(recent_mt_exit);