1 /* (C) 1999-2001 Paul `Rusty' Russell
2 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
3 * (C) 2005-2012 Patrick McHardy <kaber@trash.net>
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.
10 #include <linux/types.h>
11 #include <linux/netfilter.h>
12 #include <linux/slab.h>
13 #include <linux/module.h>
14 #include <linux/skbuff.h>
15 #include <linux/proc_fs.h>
16 #include <linux/seq_file.h>
17 #include <linux/percpu.h>
18 #include <linux/netdevice.h>
19 #include <linux/security.h>
20 #include <net/net_namespace.h>
22 #include <linux/sysctl.h>
25 #include <net/netfilter/nf_conntrack.h>
26 #include <net/netfilter/nf_conntrack_core.h>
27 #include <net/netfilter/nf_conntrack_l3proto.h>
28 #include <net/netfilter/nf_conntrack_l4proto.h>
29 #include <net/netfilter/nf_conntrack_expect.h>
30 #include <net/netfilter/nf_conntrack_helper.h>
31 #include <net/netfilter/nf_conntrack_acct.h>
32 #include <net/netfilter/nf_conntrack_zones.h>
33 #include <net/netfilter/nf_conntrack_timestamp.h>
34 #include <linux/rculist_nulls.h>
36 MODULE_LICENSE("GPL");
38 #ifdef CONFIG_NF_CONNTRACK_PROCFS
40 print_tuple(struct seq_file
*s
, const struct nf_conntrack_tuple
*tuple
,
41 const struct nf_conntrack_l3proto
*l3proto
,
42 const struct nf_conntrack_l4proto
*l4proto
)
44 l3proto
->print_tuple(s
, tuple
);
45 l4proto
->print_tuple(s
, tuple
);
47 EXPORT_SYMBOL_GPL(print_tuple
);
49 struct ct_iter_state
{
50 struct seq_net_private p
;
51 struct hlist_nulls_head
*hash
;
52 unsigned int htable_size
;
57 static struct hlist_nulls_node
*ct_get_first(struct seq_file
*seq
)
59 struct ct_iter_state
*st
= seq
->private;
60 struct hlist_nulls_node
*n
;
63 st
->bucket
< st
->htable_size
;
66 hlist_nulls_first_rcu(&st
->hash
[st
->bucket
]));
73 static struct hlist_nulls_node
*ct_get_next(struct seq_file
*seq
,
74 struct hlist_nulls_node
*head
)
76 struct ct_iter_state
*st
= seq
->private;
78 head
= rcu_dereference(hlist_nulls_next_rcu(head
));
79 while (is_a_nulls(head
)) {
80 if (likely(get_nulls_value(head
) == st
->bucket
)) {
81 if (++st
->bucket
>= st
->htable_size
)
84 head
= rcu_dereference(
85 hlist_nulls_first_rcu(&st
->hash
[st
->bucket
]));
90 static struct hlist_nulls_node
*ct_get_idx(struct seq_file
*seq
, loff_t pos
)
92 struct hlist_nulls_node
*head
= ct_get_first(seq
);
95 while (pos
&& (head
= ct_get_next(seq
, head
)))
97 return pos
? NULL
: head
;
100 static void *ct_seq_start(struct seq_file
*seq
, loff_t
*pos
)
103 struct ct_iter_state
*st
= seq
->private;
105 st
->time_now
= ktime_get_real_ns();
108 nf_conntrack_get_ht(&st
->hash
, &st
->htable_size
);
109 return ct_get_idx(seq
, *pos
);
112 static void *ct_seq_next(struct seq_file
*s
, void *v
, loff_t
*pos
)
115 return ct_get_next(s
, v
);
118 static void ct_seq_stop(struct seq_file
*s
, void *v
)
124 #ifdef CONFIG_NF_CONNTRACK_SECMARK
125 static void ct_show_secctx(struct seq_file
*s
, const struct nf_conn
*ct
)
131 ret
= security_secid_to_secctx(ct
->secmark
, &secctx
, &len
);
135 seq_printf(s
, "secctx=%s ", secctx
);
137 security_release_secctx(secctx
, len
);
140 static inline void ct_show_secctx(struct seq_file
*s
, const struct nf_conn
*ct
)
145 #ifdef CONFIG_NF_CONNTRACK_ZONES
146 static void ct_show_zone(struct seq_file
*s
, const struct nf_conn
*ct
,
149 const struct nf_conntrack_zone
*zone
= nf_ct_zone(ct
);
151 if (zone
->dir
!= dir
)
154 case NF_CT_DEFAULT_ZONE_DIR
:
155 seq_printf(s
, "zone=%u ", zone
->id
);
157 case NF_CT_ZONE_DIR_ORIG
:
158 seq_printf(s
, "zone-orig=%u ", zone
->id
);
160 case NF_CT_ZONE_DIR_REPL
:
161 seq_printf(s
, "zone-reply=%u ", zone
->id
);
168 static inline void ct_show_zone(struct seq_file
*s
, const struct nf_conn
*ct
,
174 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
175 static void ct_show_delta_time(struct seq_file
*s
, const struct nf_conn
*ct
)
177 struct ct_iter_state
*st
= s
->private;
178 struct nf_conn_tstamp
*tstamp
;
181 tstamp
= nf_conn_tstamp_find(ct
);
183 delta_time
= st
->time_now
- tstamp
->start
;
185 delta_time
= div_s64(delta_time
, NSEC_PER_SEC
);
189 seq_printf(s
, "delta-time=%llu ",
190 (unsigned long long)delta_time
);
196 ct_show_delta_time(struct seq_file
*s
, const struct nf_conn
*ct
)
201 /* return 0 on success, 1 in case of error */
202 static int ct_seq_show(struct seq_file
*s
, void *v
)
204 struct nf_conntrack_tuple_hash
*hash
= v
;
205 struct nf_conn
*ct
= nf_ct_tuplehash_to_ctrack(hash
);
206 const struct nf_conntrack_l3proto
*l3proto
;
207 const struct nf_conntrack_l4proto
*l4proto
;
211 if (unlikely(!atomic_inc_not_zero(&ct
->ct_general
.use
)))
214 /* we only want to print DIR_ORIGINAL */
215 if (NF_CT_DIRECTION(hash
))
218 l3proto
= __nf_ct_l3proto_find(nf_ct_l3num(ct
));
219 NF_CT_ASSERT(l3proto
);
220 l4proto
= __nf_ct_l4proto_find(nf_ct_l3num(ct
), nf_ct_protonum(ct
));
221 NF_CT_ASSERT(l4proto
);
224 seq_printf(s
, "%-8s %u %-8s %u %ld ",
225 l3proto
->name
, nf_ct_l3num(ct
),
226 l4proto
->name
, nf_ct_protonum(ct
),
227 timer_pending(&ct
->timeout
)
228 ? (long)(ct
->timeout
.expires
- jiffies
)/HZ
: 0);
230 if (l4proto
->print_conntrack
)
231 l4proto
->print_conntrack(s
, ct
);
233 print_tuple(s
, &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
,
236 ct_show_zone(s
, ct
, NF_CT_ZONE_DIR_ORIG
);
238 if (seq_has_overflowed(s
))
241 if (seq_print_acct(s
, ct
, IP_CT_DIR_ORIGINAL
))
244 if (!(test_bit(IPS_SEEN_REPLY_BIT
, &ct
->status
)))
245 seq_printf(s
, "[UNREPLIED] ");
247 print_tuple(s
, &ct
->tuplehash
[IP_CT_DIR_REPLY
].tuple
,
250 ct_show_zone(s
, ct
, NF_CT_ZONE_DIR_REPL
);
252 if (seq_print_acct(s
, ct
, IP_CT_DIR_REPLY
))
255 if (test_bit(IPS_ASSURED_BIT
, &ct
->status
))
256 seq_printf(s
, "[ASSURED] ");
258 if (seq_has_overflowed(s
))
261 #if defined(CONFIG_NF_CONNTRACK_MARK)
262 seq_printf(s
, "mark=%u ", ct
->mark
);
265 ct_show_secctx(s
, ct
);
266 ct_show_zone(s
, ct
, NF_CT_DEFAULT_ZONE_DIR
);
267 ct_show_delta_time(s
, ct
);
269 seq_printf(s
, "use=%u\n", atomic_read(&ct
->ct_general
.use
));
271 if (seq_has_overflowed(s
))
280 static const struct seq_operations ct_seq_ops
= {
281 .start
= ct_seq_start
,
287 static int ct_open(struct inode
*inode
, struct file
*file
)
289 return seq_open_net(inode
, file
, &ct_seq_ops
,
290 sizeof(struct ct_iter_state
));
293 static const struct file_operations ct_file_ops
= {
294 .owner
= THIS_MODULE
,
298 .release
= seq_release_net
,
301 static void *ct_cpu_seq_start(struct seq_file
*seq
, loff_t
*pos
)
303 struct net
*net
= seq_file_net(seq
);
307 return SEQ_START_TOKEN
;
309 for (cpu
= *pos
-1; cpu
< nr_cpu_ids
; ++cpu
) {
310 if (!cpu_possible(cpu
))
313 return per_cpu_ptr(net
->ct
.stat
, cpu
);
319 static void *ct_cpu_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
321 struct net
*net
= seq_file_net(seq
);
324 for (cpu
= *pos
; cpu
< nr_cpu_ids
; ++cpu
) {
325 if (!cpu_possible(cpu
))
328 return per_cpu_ptr(net
->ct
.stat
, cpu
);
334 static void ct_cpu_seq_stop(struct seq_file
*seq
, void *v
)
338 static int ct_cpu_seq_show(struct seq_file
*seq
, void *v
)
340 struct net
*net
= seq_file_net(seq
);
341 unsigned int nr_conntracks
= atomic_read(&net
->ct
.count
);
342 const struct ip_conntrack_stat
*st
= v
;
344 if (v
== SEQ_START_TOKEN
) {
345 seq_printf(seq
, "entries searched found new invalid ignore delete delete_list insert insert_failed drop early_drop icmp_error expect_new expect_create expect_delete search_restart\n");
349 seq_printf(seq
, "%08x %08x %08x %08x %08x %08x %08x %08x "
350 "%08x %08x %08x %08x %08x %08x %08x %08x %08x\n",
373 static const struct seq_operations ct_cpu_seq_ops
= {
374 .start
= ct_cpu_seq_start
,
375 .next
= ct_cpu_seq_next
,
376 .stop
= ct_cpu_seq_stop
,
377 .show
= ct_cpu_seq_show
,
380 static int ct_cpu_seq_open(struct inode
*inode
, struct file
*file
)
382 return seq_open_net(inode
, file
, &ct_cpu_seq_ops
,
383 sizeof(struct seq_net_private
));
386 static const struct file_operations ct_cpu_seq_fops
= {
387 .owner
= THIS_MODULE
,
388 .open
= ct_cpu_seq_open
,
391 .release
= seq_release_net
,
394 static int nf_conntrack_standalone_init_proc(struct net
*net
)
396 struct proc_dir_entry
*pde
;
400 pde
= proc_create("nf_conntrack", 0440, net
->proc_net
, &ct_file_ops
);
402 goto out_nf_conntrack
;
404 root_uid
= make_kuid(net
->user_ns
, 0);
405 root_gid
= make_kgid(net
->user_ns
, 0);
406 if (uid_valid(root_uid
) && gid_valid(root_gid
))
407 proc_set_user(pde
, root_uid
, root_gid
);
409 pde
= proc_create("nf_conntrack", S_IRUGO
, net
->proc_net_stat
,
412 goto out_stat_nf_conntrack
;
415 out_stat_nf_conntrack
:
416 remove_proc_entry("nf_conntrack", net
->proc_net
);
421 static void nf_conntrack_standalone_fini_proc(struct net
*net
)
423 remove_proc_entry("nf_conntrack", net
->proc_net_stat
);
424 remove_proc_entry("nf_conntrack", net
->proc_net
);
427 static int nf_conntrack_standalone_init_proc(struct net
*net
)
432 static void nf_conntrack_standalone_fini_proc(struct net
*net
)
435 #endif /* CONFIG_NF_CONNTRACK_PROCFS */
440 /* Log invalid packets of a given protocol */
441 static int log_invalid_proto_min __read_mostly
;
442 static int log_invalid_proto_max __read_mostly
= 255;
444 /* size the user *wants to set */
445 static unsigned int nf_conntrack_htable_size_user __read_mostly
;
448 nf_conntrack_hash_sysctl(struct ctl_table
*table
, int write
,
449 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
453 ret
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
454 if (ret
< 0 || !write
)
457 /* update ret, we might not be able to satisfy request */
458 ret
= nf_conntrack_hash_resize(nf_conntrack_htable_size_user
);
460 /* update it to the actual value used by conntrack */
461 nf_conntrack_htable_size_user
= nf_conntrack_htable_size
;
465 static struct ctl_table_header
*nf_ct_netfilter_header
;
467 static struct ctl_table nf_ct_sysctl_table
[] = {
469 .procname
= "nf_conntrack_max",
470 .data
= &nf_conntrack_max
,
471 .maxlen
= sizeof(int),
473 .proc_handler
= proc_dointvec
,
476 .procname
= "nf_conntrack_count",
477 .data
= &init_net
.ct
.count
,
478 .maxlen
= sizeof(int),
480 .proc_handler
= proc_dointvec
,
483 .procname
= "nf_conntrack_buckets",
484 .data
= &nf_conntrack_htable_size_user
,
485 .maxlen
= sizeof(unsigned int),
487 .proc_handler
= nf_conntrack_hash_sysctl
,
490 .procname
= "nf_conntrack_checksum",
491 .data
= &init_net
.ct
.sysctl_checksum
,
492 .maxlen
= sizeof(unsigned int),
494 .proc_handler
= proc_dointvec
,
497 .procname
= "nf_conntrack_log_invalid",
498 .data
= &init_net
.ct
.sysctl_log_invalid
,
499 .maxlen
= sizeof(unsigned int),
501 .proc_handler
= proc_dointvec_minmax
,
502 .extra1
= &log_invalid_proto_min
,
503 .extra2
= &log_invalid_proto_max
,
506 .procname
= "nf_conntrack_expect_max",
507 .data
= &nf_ct_expect_max
,
508 .maxlen
= sizeof(int),
510 .proc_handler
= proc_dointvec
,
515 static struct ctl_table nf_ct_netfilter_table
[] = {
517 .procname
= "nf_conntrack_max",
518 .data
= &nf_conntrack_max
,
519 .maxlen
= sizeof(int),
521 .proc_handler
= proc_dointvec
,
526 static int nf_conntrack_standalone_init_sysctl(struct net
*net
)
528 struct ctl_table
*table
;
530 table
= kmemdup(nf_ct_sysctl_table
, sizeof(nf_ct_sysctl_table
),
535 table
[1].data
= &net
->ct
.count
;
536 table
[3].data
= &net
->ct
.sysctl_checksum
;
537 table
[4].data
= &net
->ct
.sysctl_log_invalid
;
539 /* Don't export sysctls to unprivileged users */
540 if (net
->user_ns
!= &init_user_ns
)
541 table
[0].procname
= NULL
;
543 if (!net_eq(&init_net
, net
))
544 table
[2].mode
= 0444;
546 net
->ct
.sysctl_header
= register_net_sysctl(net
, "net/netfilter", table
);
547 if (!net
->ct
.sysctl_header
)
548 goto out_unregister_netfilter
;
552 out_unregister_netfilter
:
558 static void nf_conntrack_standalone_fini_sysctl(struct net
*net
)
560 struct ctl_table
*table
;
562 table
= net
->ct
.sysctl_header
->ctl_table_arg
;
563 unregister_net_sysctl_table(net
->ct
.sysctl_header
);
567 static int nf_conntrack_standalone_init_sysctl(struct net
*net
)
572 static void nf_conntrack_standalone_fini_sysctl(struct net
*net
)
575 #endif /* CONFIG_SYSCTL */
577 static int nf_conntrack_pernet_init(struct net
*net
)
581 ret
= nf_conntrack_init_net(net
);
585 ret
= nf_conntrack_standalone_init_proc(net
);
589 net
->ct
.sysctl_checksum
= 1;
590 net
->ct
.sysctl_log_invalid
= 0;
591 ret
= nf_conntrack_standalone_init_sysctl(net
);
598 nf_conntrack_standalone_fini_proc(net
);
600 nf_conntrack_cleanup_net(net
);
605 static void nf_conntrack_pernet_exit(struct list_head
*net_exit_list
)
609 list_for_each_entry(net
, net_exit_list
, exit_list
) {
610 nf_conntrack_standalone_fini_sysctl(net
);
611 nf_conntrack_standalone_fini_proc(net
);
613 nf_conntrack_cleanup_net_list(net_exit_list
);
616 static struct pernet_operations nf_conntrack_net_ops
= {
617 .init
= nf_conntrack_pernet_init
,
618 .exit_batch
= nf_conntrack_pernet_exit
,
621 static int __init
nf_conntrack_standalone_init(void)
623 int ret
= nf_conntrack_init_start();
628 nf_ct_netfilter_header
=
629 register_net_sysctl(&init_net
, "net", nf_ct_netfilter_table
);
630 if (!nf_ct_netfilter_header
) {
631 pr_err("nf_conntrack: can't register to sysctl.\n");
636 nf_conntrack_htable_size_user
= nf_conntrack_htable_size
;
639 ret
= register_pernet_subsys(&nf_conntrack_net_ops
);
643 nf_conntrack_init_end();
648 unregister_net_sysctl_table(nf_ct_netfilter_header
);
651 nf_conntrack_cleanup_end();
656 static void __exit
nf_conntrack_standalone_fini(void)
658 nf_conntrack_cleanup_start();
659 unregister_pernet_subsys(&nf_conntrack_net_ops
);
661 unregister_net_sysctl_table(nf_ct_netfilter_header
);
663 nf_conntrack_cleanup_end();
666 module_init(nf_conntrack_standalone_init
);
667 module_exit(nf_conntrack_standalone_fini
);
669 /* Some modules need us, but don't depend directly on any symbol.
670 They should call this. */
671 void need_conntrack(void)
674 EXPORT_SYMBOL_GPL(need_conntrack
);