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:
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>
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, 0644);
55 module_param(ip_list_gid, uint, 0644);
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 unsigned 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_mt_fops;
112 static u_int32_t hash_rnd __read_mostly;
114 static inline unsigned int recent_entry_hash4(const union nf_inet_addr *addr)
116 return jhash_1word((__force u32)addr->ip, hash_rnd) &
117 (ip_list_hash_size - 1);
120 static inline unsigned int recent_entry_hash6(const union nf_inet_addr *addr)
122 return jhash2((u32 *)addr->ip6, ARRAY_SIZE(addr->ip6), hash_rnd) &
123 (ip_list_hash_size - 1);
126 static struct recent_entry *
127 recent_entry_lookup(const struct recent_table *table,
128 const union nf_inet_addr *addrp, u_int16_t family,
131 struct recent_entry *e;
134 if (family == NFPROTO_IPV4)
135 h = recent_entry_hash4(addrp);
137 h = recent_entry_hash6(addrp);
139 list_for_each_entry(e, &table->iphash[h], list)
140 if (e->family == family &&
141 memcmp(&e->addr, addrp, sizeof(e->addr)) == 0 &&
142 (ttl == e->ttl || ttl == 0 || e->ttl == 0))
147 static void recent_entry_remove(struct recent_table *t, struct recent_entry *e)
150 list_del(&e->lru_list);
156 * Drop entries with timestamps older then 'time'.
158 static void recent_entry_reap(struct recent_table *t, unsigned long time)
160 struct recent_entry *e;
163 * The head of the LRU list is always the oldest entry.
165 e = list_entry(t->lru_list.next, struct recent_entry, lru_list);
168 * The last time stamp is the most recent.
170 if (time_after(time, e->stamps[e->index-1]))
171 recent_entry_remove(t, e);
174 static struct recent_entry *
175 recent_entry_init(struct recent_table *t, const union nf_inet_addr *addr,
176 u_int16_t family, u_int8_t ttl)
178 struct recent_entry *e;
179 unsigned int nstamps_max = t->nstamps_max_mask;
181 if (t->entries >= ip_list_tot) {
182 e = list_entry(t->lru_list.next, struct recent_entry, lru_list);
183 recent_entry_remove(t, e);
187 e = kmalloc(struct_size(e, stamps, nstamps_max), GFP_ATOMIC);
190 memcpy(&e->addr, addr, sizeof(e->addr));
192 e->stamps[0] = jiffies;
196 if (family == NFPROTO_IPV4)
197 list_add_tail(&e->list, &t->iphash[recent_entry_hash4(addr)]);
199 list_add_tail(&e->list, &t->iphash[recent_entry_hash6(addr)]);
200 list_add_tail(&e->lru_list, &t->lru_list);
205 static void recent_entry_update(struct recent_table *t, struct recent_entry *e)
207 e->index &= t->nstamps_max_mask;
208 e->stamps[e->index++] = jiffies;
209 if (e->index > e->nstamps)
210 e->nstamps = e->index;
211 list_move_tail(&e->lru_list, &t->lru_list);
214 static struct recent_table *recent_table_lookup(struct recent_net *recent_net,
217 struct recent_table *t;
219 list_for_each_entry(t, &recent_net->tables, list)
220 if (!strcmp(t->name, name))
225 static void recent_table_flush(struct recent_table *t)
227 struct recent_entry *e, *next;
230 for (i = 0; i < ip_list_hash_size; i++)
231 list_for_each_entry_safe(e, next, &t->iphash[i], list)
232 recent_entry_remove(t, e);
236 recent_mt(const struct sk_buff *skb, struct xt_action_param *par)
238 struct net *net = xt_net(par);
239 struct recent_net *recent_net = recent_pernet(net);
240 const struct xt_recent_mtinfo_v1 *info = par->matchinfo;
241 struct recent_table *t;
242 struct recent_entry *e;
243 union nf_inet_addr addr = {}, addr_mask;
245 bool ret = info->invert;
247 if (xt_family(par) == NFPROTO_IPV4) {
248 const struct iphdr *iph = ip_hdr(skb);
250 if (info->side == XT_RECENT_DEST)
251 addr.ip = iph->daddr;
253 addr.ip = iph->saddr;
257 const struct ipv6hdr *iph = ipv6_hdr(skb);
259 if (info->side == XT_RECENT_DEST)
260 memcpy(&addr.in6, &iph->daddr, sizeof(addr.in6));
262 memcpy(&addr.in6, &iph->saddr, sizeof(addr.in6));
264 ttl = iph->hop_limit;
267 /* use TTL as seen before forwarding */
268 if (xt_out(par) != NULL &&
269 (!skb->sk || !net_eq(net, sock_net(skb->sk))))
272 spin_lock_bh(&recent_lock);
273 t = recent_table_lookup(recent_net, info->name);
275 nf_inet_addr_mask(&addr, &addr_mask, &t->mask);
277 e = recent_entry_lookup(t, &addr_mask, xt_family(par),
278 (info->check_set & XT_RECENT_TTL) ? ttl : 0);
280 if (!(info->check_set & XT_RECENT_SET))
282 e = recent_entry_init(t, &addr_mask, xt_family(par), ttl);
289 if (info->check_set & XT_RECENT_SET)
291 else if (info->check_set & XT_RECENT_REMOVE) {
292 recent_entry_remove(t, e);
294 } else if (info->check_set & (XT_RECENT_CHECK | XT_RECENT_UPDATE)) {
295 unsigned long time = jiffies - info->seconds * HZ;
296 unsigned int i, hits = 0;
298 for (i = 0; i < e->nstamps; i++) {
299 if (info->seconds && time_after(time, e->stamps[i]))
301 if (!info->hit_count || ++hits >= info->hit_count) {
307 /* info->seconds must be non-zero */
308 if (info->check_set & XT_RECENT_REAP)
309 recent_entry_reap(t, time);
312 if (info->check_set & XT_RECENT_SET ||
313 (info->check_set & XT_RECENT_UPDATE && ret)) {
314 recent_entry_update(t, e);
318 spin_unlock_bh(&recent_lock);
322 static void recent_table_free(void *addr)
327 static int recent_mt_check(const struct xt_mtchk_param *par,
328 const struct xt_recent_mtinfo_v1 *info)
330 struct recent_net *recent_net = recent_pernet(par->net);
331 struct recent_table *t;
332 #ifdef CONFIG_PROC_FS
333 struct proc_dir_entry *pde;
337 unsigned int nstamp_mask;
341 net_get_random_once(&hash_rnd, sizeof(hash_rnd));
343 if (info->check_set & ~XT_RECENT_VALID_FLAGS) {
344 pr_info_ratelimited("Unsupported userspace flags (%08x)\n",
348 if (hweight8(info->check_set &
349 (XT_RECENT_SET | XT_RECENT_REMOVE |
350 XT_RECENT_CHECK | XT_RECENT_UPDATE)) != 1)
352 if ((info->check_set & (XT_RECENT_SET | XT_RECENT_REMOVE)) &&
353 (info->seconds || info->hit_count ||
354 (info->check_set & XT_RECENT_MODIFIERS)))
356 if ((info->check_set & XT_RECENT_REAP) && !info->seconds)
358 if (info->hit_count >= XT_RECENT_MAX_NSTAMPS) {
359 pr_info_ratelimited("hitcount (%u) is larger than allowed maximum (%u)\n",
360 info->hit_count, XT_RECENT_MAX_NSTAMPS - 1);
363 ret = xt_check_proc_name(info->name, sizeof(info->name));
367 if (ip_pkt_list_tot && info->hit_count < ip_pkt_list_tot)
368 nstamp_mask = roundup_pow_of_two(ip_pkt_list_tot) - 1;
369 else if (info->hit_count)
370 nstamp_mask = roundup_pow_of_two(info->hit_count) - 1;
372 nstamp_mask = 32 - 1;
374 mutex_lock(&recent_mutex);
375 t = recent_table_lookup(recent_net, info->name);
377 if (nstamp_mask > t->nstamps_max_mask) {
378 spin_lock_bh(&recent_lock);
379 recent_table_flush(t);
380 t->nstamps_max_mask = nstamp_mask;
381 spin_unlock_bh(&recent_lock);
389 t = kvzalloc(struct_size(t, iphash, ip_list_hash_size), GFP_KERNEL);
395 t->nstamps_max_mask = nstamp_mask;
397 memcpy(&t->mask, &info->mask, sizeof(t->mask));
398 strcpy(t->name, info->name);
399 INIT_LIST_HEAD(&t->lru_list);
400 for (i = 0; i < ip_list_hash_size; i++)
401 INIT_LIST_HEAD(&t->iphash[i]);
402 #ifdef CONFIG_PROC_FS
403 uid = make_kuid(&init_user_ns, ip_list_uid);
404 gid = make_kgid(&init_user_ns, ip_list_gid);
405 if (!uid_valid(uid) || !gid_valid(gid)) {
406 recent_table_free(t);
410 pde = proc_create_data(t->name, ip_list_perms, recent_net->xt_recent,
413 recent_table_free(t);
417 proc_set_user(pde, uid, gid);
419 spin_lock_bh(&recent_lock);
420 list_add_tail(&t->list, &recent_net->tables);
421 spin_unlock_bh(&recent_lock);
424 mutex_unlock(&recent_mutex);
428 static int recent_mt_check_v0(const struct xt_mtchk_param *par)
430 const struct xt_recent_mtinfo_v0 *info_v0 = par->matchinfo;
431 struct xt_recent_mtinfo_v1 info_v1;
433 /* Copy revision 0 structure to revision 1 */
434 memcpy(&info_v1, info_v0, sizeof(struct xt_recent_mtinfo));
435 /* Set default mask to ensure backward compatible behaviour */
436 memset(info_v1.mask.all, 0xFF, sizeof(info_v1.mask.all));
438 return recent_mt_check(par, &info_v1);
441 static int recent_mt_check_v1(const struct xt_mtchk_param *par)
443 return recent_mt_check(par, par->matchinfo);
446 static void recent_mt_destroy(const struct xt_mtdtor_param *par)
448 struct recent_net *recent_net = recent_pernet(par->net);
449 const struct xt_recent_mtinfo_v1 *info = par->matchinfo;
450 struct recent_table *t;
452 mutex_lock(&recent_mutex);
453 t = recent_table_lookup(recent_net, info->name);
454 if (--t->refcnt == 0) {
455 spin_lock_bh(&recent_lock);
457 spin_unlock_bh(&recent_lock);
458 #ifdef CONFIG_PROC_FS
459 if (recent_net->xt_recent != NULL)
460 remove_proc_entry(t->name, recent_net->xt_recent);
462 recent_table_flush(t);
463 recent_table_free(t);
465 mutex_unlock(&recent_mutex);
468 #ifdef CONFIG_PROC_FS
469 struct recent_iter_state {
470 const struct recent_table *table;
474 static void *recent_seq_start(struct seq_file *seq, loff_t *pos)
475 __acquires(recent_lock)
477 struct recent_iter_state *st = seq->private;
478 const struct recent_table *t = st->table;
479 struct recent_entry *e;
482 spin_lock_bh(&recent_lock);
484 for (st->bucket = 0; st->bucket < ip_list_hash_size; st->bucket++)
485 list_for_each_entry(e, &t->iphash[st->bucket], list)
491 static void *recent_seq_next(struct seq_file *seq, void *v, loff_t *pos)
493 struct recent_iter_state *st = seq->private;
494 const struct recent_table *t = st->table;
495 const struct recent_entry *e = v;
496 const struct list_head *head = e->list.next;
498 while (head == &t->iphash[st->bucket]) {
499 if (++st->bucket >= ip_list_hash_size)
501 head = t->iphash[st->bucket].next;
504 return list_entry(head, struct recent_entry, list);
507 static void recent_seq_stop(struct seq_file *s, void *v)
508 __releases(recent_lock)
510 spin_unlock_bh(&recent_lock);
513 static int recent_seq_show(struct seq_file *seq, void *v)
515 const struct recent_entry *e = v;
516 struct recent_iter_state *st = seq->private;
517 const struct recent_table *t = st->table;
520 i = (e->index - 1) & t->nstamps_max_mask;
522 if (e->family == NFPROTO_IPV4)
523 seq_printf(seq, "src=%pI4 ttl: %u last_seen: %lu oldest_pkt: %u",
524 &e->addr.ip, e->ttl, e->stamps[i], e->index);
526 seq_printf(seq, "src=%pI6 ttl: %u last_seen: %lu oldest_pkt: %u",
527 &e->addr.in6, e->ttl, e->stamps[i], e->index);
528 for (i = 0; i < e->nstamps; i++)
529 seq_printf(seq, "%s %lu", i ? "," : "", e->stamps[i]);
534 static const struct seq_operations recent_seq_ops = {
535 .start = recent_seq_start,
536 .next = recent_seq_next,
537 .stop = recent_seq_stop,
538 .show = recent_seq_show,
541 static int recent_seq_open(struct inode *inode, struct file *file)
543 struct recent_iter_state *st;
545 st = __seq_open_private(file, &recent_seq_ops, sizeof(*st));
549 st->table = PDE_DATA(inode);
554 recent_mt_proc_write(struct file *file, const char __user *input,
555 size_t size, loff_t *loff)
557 struct recent_table *t = PDE_DATA(file_inode(file));
558 struct recent_entry *e;
559 char buf[sizeof("+b335:1d35:1e55:dead:c0de:1715:5afe:c0de")];
561 union nf_inet_addr addr = {};
567 if (size > sizeof(buf))
569 if (copy_from_user(buf, input, size) != 0)
572 /* Strict protocol! */
576 case '/': /* flush table */
577 spin_lock_bh(&recent_lock);
578 recent_table_flush(t);
579 spin_unlock_bh(&recent_lock);
581 case '-': /* remove address */
584 case '+': /* add address */
588 pr_info_ratelimited("Need \"+ip\", \"-ip\" or \"/\"\n");
594 if (strnchr(c, size, ':') != NULL) {
595 family = NFPROTO_IPV6;
596 succ = in6_pton(c, size, (void *)&addr, '\n', NULL);
598 family = NFPROTO_IPV4;
599 succ = in4_pton(c, size, (void *)&addr, '\n', NULL);
605 spin_lock_bh(&recent_lock);
606 e = recent_entry_lookup(t, &addr, family, 0);
609 recent_entry_init(t, &addr, family, 0);
612 recent_entry_update(t, e);
614 recent_entry_remove(t, e);
616 spin_unlock_bh(&recent_lock);
617 /* Note we removed one above */
622 static const struct file_operations recent_mt_fops = {
623 .open = recent_seq_open,
625 .write = recent_mt_proc_write,
626 .release = seq_release_private,
627 .owner = THIS_MODULE,
631 static int __net_init recent_proc_net_init(struct net *net)
633 struct recent_net *recent_net = recent_pernet(net);
635 recent_net->xt_recent = proc_mkdir("xt_recent", net->proc_net);
636 if (!recent_net->xt_recent)
641 static void __net_exit recent_proc_net_exit(struct net *net)
643 struct recent_net *recent_net = recent_pernet(net);
644 struct recent_table *t;
646 /* recent_net_exit() is called before recent_mt_destroy(). Make sure
647 * that the parent xt_recent proc entry is is empty before trying to
650 spin_lock_bh(&recent_lock);
651 list_for_each_entry(t, &recent_net->tables, list)
652 remove_proc_entry(t->name, recent_net->xt_recent);
654 recent_net->xt_recent = NULL;
655 spin_unlock_bh(&recent_lock);
657 remove_proc_entry("xt_recent", net->proc_net);
660 static inline int recent_proc_net_init(struct net *net)
665 static inline void recent_proc_net_exit(struct net *net)
668 #endif /* CONFIG_PROC_FS */
670 static int __net_init recent_net_init(struct net *net)
672 struct recent_net *recent_net = recent_pernet(net);
674 INIT_LIST_HEAD(&recent_net->tables);
675 return recent_proc_net_init(net);
678 static void __net_exit recent_net_exit(struct net *net)
680 recent_proc_net_exit(net);
683 static struct pernet_operations recent_net_ops = {
684 .init = recent_net_init,
685 .exit = recent_net_exit,
686 .id = &recent_net_id,
687 .size = sizeof(struct recent_net),
690 static struct xt_match recent_mt_reg[] __read_mostly = {
694 .family = NFPROTO_IPV4,
696 .matchsize = sizeof(struct xt_recent_mtinfo),
697 .checkentry = recent_mt_check_v0,
698 .destroy = recent_mt_destroy,
704 .family = NFPROTO_IPV6,
706 .matchsize = sizeof(struct xt_recent_mtinfo),
707 .checkentry = recent_mt_check_v0,
708 .destroy = recent_mt_destroy,
714 .family = NFPROTO_IPV4,
716 .matchsize = sizeof(struct xt_recent_mtinfo_v1),
717 .checkentry = recent_mt_check_v1,
718 .destroy = recent_mt_destroy,
724 .family = NFPROTO_IPV6,
726 .matchsize = sizeof(struct xt_recent_mtinfo_v1),
727 .checkentry = recent_mt_check_v1,
728 .destroy = recent_mt_destroy,
733 static int __init recent_mt_init(void)
737 BUILD_BUG_ON_NOT_POWER_OF_2(XT_RECENT_MAX_NSTAMPS);
739 if (!ip_list_tot || ip_pkt_list_tot >= XT_RECENT_MAX_NSTAMPS)
741 ip_list_hash_size = 1 << fls(ip_list_tot);
743 err = register_pernet_subsys(&recent_net_ops);
746 err = xt_register_matches(recent_mt_reg, ARRAY_SIZE(recent_mt_reg));
748 unregister_pernet_subsys(&recent_net_ops);
752 static void __exit recent_mt_exit(void)
754 xt_unregister_matches(recent_mt_reg, ARRAY_SIZE(recent_mt_reg));
755 unregister_pernet_subsys(&recent_net_ops);
758 module_init(recent_mt_init);
759 module_exit(recent_mt_exit);