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>
4 *
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.
8 */
9
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>
21#ifdef CONFIG_SYSCTL
22#include <linux/sysctl.h>
23#endif
24
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>
35
36MODULE_LICENSE("GPL");
37
38#ifdef CONFIG_NF_CONNTRACK_PROCFS
39int
40print_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)
43{
44	return l3proto->print_tuple(s, tuple) || l4proto->print_tuple(s, tuple);
45}
46EXPORT_SYMBOL_GPL(print_tuple);
47
48struct ct_iter_state {
49	struct seq_net_private p;
50	unsigned int bucket;
51	u_int64_t time_now;
52};
53
54static struct hlist_nulls_node *ct_get_first(struct seq_file *seq)
55{
56	struct net *net = seq_file_net(seq);
57	struct ct_iter_state *st = seq->private;
58	struct hlist_nulls_node *n;
59
60	for (st->bucket = 0;
61	     st->bucket < net->ct.htable_size;
62	     st->bucket++) {
63		n = rcu_dereference(hlist_nulls_first_rcu(&net->ct.hash[st->bucket]));
64		if (!is_a_nulls(n))
65			return n;
66	}
67	return NULL;
68}
69
70static struct hlist_nulls_node *ct_get_next(struct seq_file *seq,
71				      struct hlist_nulls_node *head)
72{
73	struct net *net = seq_file_net(seq);
74	struct ct_iter_state *st = seq->private;
75
76	head = rcu_dereference(hlist_nulls_next_rcu(head));
77	while (is_a_nulls(head)) {
78		if (likely(get_nulls_value(head) == st->bucket)) {
79			if (++st->bucket >= net->ct.htable_size)
80				return NULL;
81		}
82		head = rcu_dereference(
83				hlist_nulls_first_rcu(
84					&net->ct.hash[st->bucket]));
85	}
86	return head;
87}
88
89static struct hlist_nulls_node *ct_get_idx(struct seq_file *seq, loff_t pos)
90{
91	struct hlist_nulls_node *head = ct_get_first(seq);
92
93	if (head)
94		while (pos && (head = ct_get_next(seq, head)))
95			pos--;
96	return pos ? NULL : head;
97}
98
99static void *ct_seq_start(struct seq_file *seq, loff_t *pos)
100	__acquires(RCU)
101{
102	struct ct_iter_state *st = seq->private;
103
104	st->time_now = ktime_get_real_ns();
105	rcu_read_lock();
106	return ct_get_idx(seq, *pos);
107}
108
109static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
110{
111	(*pos)++;
112	return ct_get_next(s, v);
113}
114
115static void ct_seq_stop(struct seq_file *s, void *v)
116	__releases(RCU)
117{
118	rcu_read_unlock();
119}
120
121#ifdef CONFIG_NF_CONNTRACK_SECMARK
122static int ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
123{
124	int ret;
125	u32 len;
126	char *secctx;
127
128	ret = security_secid_to_secctx(ct->secmark, &secctx, &len);
129	if (ret)
130		return 0;
131
132	ret = seq_printf(s, "secctx=%s ", secctx);
133
134	security_release_secctx(secctx, len);
135	return ret;
136}
137#else
138static inline int ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
139{
140	return 0;
141}
142#endif
143
144#ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
145static int ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
146{
147	struct ct_iter_state *st = s->private;
148	struct nf_conn_tstamp *tstamp;
149	s64 delta_time;
150
151	tstamp = nf_conn_tstamp_find(ct);
152	if (tstamp) {
153		delta_time = st->time_now - tstamp->start;
154		if (delta_time > 0)
155			delta_time = div_s64(delta_time, NSEC_PER_SEC);
156		else
157			delta_time = 0;
158
159		return seq_printf(s, "delta-time=%llu ",
160				  (unsigned long long)delta_time);
161	}
162	return 0;
163}
164#else
165static inline int
166ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
167{
168	return 0;
169}
170#endif
171
172/* return 0 on success, 1 in case of error */
173static int ct_seq_show(struct seq_file *s, void *v)
174{
175	struct nf_conntrack_tuple_hash *hash = v;
176	struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(hash);
177	const struct nf_conntrack_l3proto *l3proto;
178	const struct nf_conntrack_l4proto *l4proto;
179	int ret = 0;
180
181	NF_CT_ASSERT(ct);
182	if (unlikely(!atomic_inc_not_zero(&ct->ct_general.use)))
183		return 0;
184
185	/* we only want to print DIR_ORIGINAL */
186	if (NF_CT_DIRECTION(hash))
187		goto release;
188
189	l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct));
190	NF_CT_ASSERT(l3proto);
191	l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
192	NF_CT_ASSERT(l4proto);
193
194	ret = -ENOSPC;
195	if (seq_printf(s, "%-8s %u %-8s %u %ld ",
196		       l3proto->name, nf_ct_l3num(ct),
197		       l4proto->name, nf_ct_protonum(ct),
198		       timer_pending(&ct->timeout)
199		       ? (long)(ct->timeout.expires - jiffies)/HZ : 0) != 0)
200		goto release;
201
202	if (l4proto->print_conntrack && l4proto->print_conntrack(s, ct))
203		goto release;
204
205	if (print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
206			l3proto, l4proto))
207		goto release;
208
209	if (seq_print_acct(s, ct, IP_CT_DIR_ORIGINAL))
210		goto release;
211
212	if (!(test_bit(IPS_SEEN_REPLY_BIT, &ct->status)))
213		if (seq_printf(s, "[UNREPLIED] "))
214			goto release;
215
216	if (print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple,
217			l3proto, l4proto))
218		goto release;
219
220	if (seq_print_acct(s, ct, IP_CT_DIR_REPLY))
221		goto release;
222
223	if (test_bit(IPS_ASSURED_BIT, &ct->status))
224		if (seq_printf(s, "[ASSURED] "))
225			goto release;
226
227#if defined(CONFIG_NF_CONNTRACK_MARK)
228	if (seq_printf(s, "mark=%u ", ct->mark))
229		goto release;
230#endif
231
232	if (ct_show_secctx(s, ct))
233		goto release;
234
235#ifdef CONFIG_NF_CONNTRACK_ZONES
236	if (seq_printf(s, "zone=%u ", nf_ct_zone(ct)))
237		goto release;
238#endif
239
240	if (ct_show_delta_time(s, ct))
241		goto release;
242
243	if (seq_printf(s, "use=%u\n", atomic_read(&ct->ct_general.use)))
244		goto release;
245
246	ret = 0;
247release:
248	nf_ct_put(ct);
249	return ret;
250}
251
252static const struct seq_operations ct_seq_ops = {
253	.start = ct_seq_start,
254	.next  = ct_seq_next,
255	.stop  = ct_seq_stop,
256	.show  = ct_seq_show
257};
258
259static int ct_open(struct inode *inode, struct file *file)
260{
261	return seq_open_net(inode, file, &ct_seq_ops,
262			sizeof(struct ct_iter_state));
263}
264
265static const struct file_operations ct_file_ops = {
266	.owner   = THIS_MODULE,
267	.open    = ct_open,
268	.read    = seq_read,
269	.llseek  = seq_lseek,
270	.release = seq_release_net,
271};
272
273static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
274{
275	struct net *net = seq_file_net(seq);
276	int cpu;
277
278	if (*pos == 0)
279		return SEQ_START_TOKEN;
280
281	for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
282		if (!cpu_possible(cpu))
283			continue;
284		*pos = cpu + 1;
285		return per_cpu_ptr(net->ct.stat, cpu);
286	}
287
288	return NULL;
289}
290
291static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
292{
293	struct net *net = seq_file_net(seq);
294	int cpu;
295
296	for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
297		if (!cpu_possible(cpu))
298			continue;
299		*pos = cpu + 1;
300		return per_cpu_ptr(net->ct.stat, cpu);
301	}
302
303	return NULL;
304}
305
306static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
307{
308}
309
310static int ct_cpu_seq_show(struct seq_file *seq, void *v)
311{
312	struct net *net = seq_file_net(seq);
313	unsigned int nr_conntracks = atomic_read(&net->ct.count);
314	const struct ip_conntrack_stat *st = v;
315
316	if (v == SEQ_START_TOKEN) {
317		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");
318		return 0;
319	}
320
321	seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
322			"%08x %08x %08x %08x %08x  %08x %08x %08x %08x\n",
323		   nr_conntracks,
324		   st->searched,
325		   st->found,
326		   st->new,
327		   st->invalid,
328		   st->ignore,
329		   st->delete,
330		   st->delete_list,
331		   st->insert,
332		   st->insert_failed,
333		   st->drop,
334		   st->early_drop,
335		   st->error,
336
337		   st->expect_new,
338		   st->expect_create,
339		   st->expect_delete,
340		   st->search_restart
341		);
342	return 0;
343}
344
345static const struct seq_operations ct_cpu_seq_ops = {
346	.start	= ct_cpu_seq_start,
347	.next	= ct_cpu_seq_next,
348	.stop	= ct_cpu_seq_stop,
349	.show	= ct_cpu_seq_show,
350};
351
352static int ct_cpu_seq_open(struct inode *inode, struct file *file)
353{
354	return seq_open_net(inode, file, &ct_cpu_seq_ops,
355			    sizeof(struct seq_net_private));
356}
357
358static const struct file_operations ct_cpu_seq_fops = {
359	.owner	 = THIS_MODULE,
360	.open	 = ct_cpu_seq_open,
361	.read	 = seq_read,
362	.llseek	 = seq_lseek,
363	.release = seq_release_net,
364};
365
366static int nf_conntrack_standalone_init_proc(struct net *net)
367{
368	struct proc_dir_entry *pde;
369
370	pde = proc_create("nf_conntrack", 0440, net->proc_net, &ct_file_ops);
371	if (!pde)
372		goto out_nf_conntrack;
373
374	pde = proc_create("nf_conntrack", S_IRUGO, net->proc_net_stat,
375			  &ct_cpu_seq_fops);
376	if (!pde)
377		goto out_stat_nf_conntrack;
378	return 0;
379
380out_stat_nf_conntrack:
381	remove_proc_entry("nf_conntrack", net->proc_net);
382out_nf_conntrack:
383	return -ENOMEM;
384}
385
386static void nf_conntrack_standalone_fini_proc(struct net *net)
387{
388	remove_proc_entry("nf_conntrack", net->proc_net_stat);
389	remove_proc_entry("nf_conntrack", net->proc_net);
390}
391#else
392static int nf_conntrack_standalone_init_proc(struct net *net)
393{
394	return 0;
395}
396
397static void nf_conntrack_standalone_fini_proc(struct net *net)
398{
399}
400#endif /* CONFIG_NF_CONNTRACK_PROCFS */
401
402/* Sysctl support */
403
404#ifdef CONFIG_SYSCTL
405/* Log invalid packets of a given protocol */
406static int log_invalid_proto_min = 0;
407static int log_invalid_proto_max = 255;
408
409static struct ctl_table_header *nf_ct_netfilter_header;
410
411static struct ctl_table nf_ct_sysctl_table[] = {
412	{
413		.procname	= "nf_conntrack_max",
414		.data		= &nf_conntrack_max,
415		.maxlen		= sizeof(int),
416		.mode		= 0644,
417		.proc_handler	= proc_dointvec,
418	},
419	{
420		.procname	= "nf_conntrack_count",
421		.data		= &init_net.ct.count,
422		.maxlen		= sizeof(int),
423		.mode		= 0444,
424		.proc_handler	= proc_dointvec,
425	},
426	{
427		.procname       = "nf_conntrack_buckets",
428		.data           = &init_net.ct.htable_size,
429		.maxlen         = sizeof(unsigned int),
430		.mode           = 0444,
431		.proc_handler   = proc_dointvec,
432	},
433	{
434		.procname	= "nf_conntrack_checksum",
435		.data		= &init_net.ct.sysctl_checksum,
436		.maxlen		= sizeof(unsigned int),
437		.mode		= 0644,
438		.proc_handler	= proc_dointvec,
439	},
440	{
441		.procname	= "nf_conntrack_log_invalid",
442		.data		= &init_net.ct.sysctl_log_invalid,
443		.maxlen		= sizeof(unsigned int),
444		.mode		= 0644,
445		.proc_handler	= proc_dointvec_minmax,
446		.extra1		= &log_invalid_proto_min,
447		.extra2		= &log_invalid_proto_max,
448	},
449	{
450		.procname	= "nf_conntrack_expect_max",
451		.data		= &nf_ct_expect_max,
452		.maxlen		= sizeof(int),
453		.mode		= 0644,
454		.proc_handler	= proc_dointvec,
455	},
456	{ }
457};
458
459#define NET_NF_CONNTRACK_MAX 2089
460
461static struct ctl_table nf_ct_netfilter_table[] = {
462	{
463		.procname	= "nf_conntrack_max",
464		.data		= &nf_conntrack_max,
465		.maxlen		= sizeof(int),
466		.mode		= 0644,
467		.proc_handler	= proc_dointvec,
468	},
469	{ }
470};
471
472static int nf_conntrack_standalone_init_sysctl(struct net *net)
473{
474	struct ctl_table *table;
475
476	table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table),
477			GFP_KERNEL);
478	if (!table)
479		goto out_kmemdup;
480
481	table[1].data = &net->ct.count;
482	table[2].data = &net->ct.htable_size;
483	table[3].data = &net->ct.sysctl_checksum;
484	table[4].data = &net->ct.sysctl_log_invalid;
485
486	/* Don't export sysctls to unprivileged users */
487	if (net->user_ns != &init_user_ns)
488		table[0].procname = NULL;
489
490	net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table);
491	if (!net->ct.sysctl_header)
492		goto out_unregister_netfilter;
493
494	return 0;
495
496out_unregister_netfilter:
497	kfree(table);
498out_kmemdup:
499	return -ENOMEM;
500}
501
502static void nf_conntrack_standalone_fini_sysctl(struct net *net)
503{
504	struct ctl_table *table;
505
506	table = net->ct.sysctl_header->ctl_table_arg;
507	unregister_net_sysctl_table(net->ct.sysctl_header);
508	kfree(table);
509}
510#else
511static int nf_conntrack_standalone_init_sysctl(struct net *net)
512{
513	return 0;
514}
515
516static void nf_conntrack_standalone_fini_sysctl(struct net *net)
517{
518}
519#endif /* CONFIG_SYSCTL */
520
521static int nf_conntrack_pernet_init(struct net *net)
522{
523	int ret;
524
525	ret = nf_conntrack_init_net(net);
526	if (ret < 0)
527		goto out_init;
528
529	ret = nf_conntrack_standalone_init_proc(net);
530	if (ret < 0)
531		goto out_proc;
532
533	net->ct.sysctl_checksum = 1;
534	net->ct.sysctl_log_invalid = 0;
535	ret = nf_conntrack_standalone_init_sysctl(net);
536	if (ret < 0)
537		goto out_sysctl;
538
539	return 0;
540
541out_sysctl:
542	nf_conntrack_standalone_fini_proc(net);
543out_proc:
544	nf_conntrack_cleanup_net(net);
545out_init:
546	return ret;
547}
548
549static void nf_conntrack_pernet_exit(struct list_head *net_exit_list)
550{
551	struct net *net;
552
553	list_for_each_entry(net, net_exit_list, exit_list) {
554		nf_conntrack_standalone_fini_sysctl(net);
555		nf_conntrack_standalone_fini_proc(net);
556	}
557	nf_conntrack_cleanup_net_list(net_exit_list);
558}
559
560static struct pernet_operations nf_conntrack_net_ops = {
561	.init		= nf_conntrack_pernet_init,
562	.exit_batch	= nf_conntrack_pernet_exit,
563};
564
565static int __init nf_conntrack_standalone_init(void)
566{
567	int ret = nf_conntrack_init_start();
568	if (ret < 0)
569		goto out_start;
570
571#ifdef CONFIG_SYSCTL
572	nf_ct_netfilter_header =
573		register_net_sysctl(&init_net, "net", nf_ct_netfilter_table);
574	if (!nf_ct_netfilter_header) {
575		pr_err("nf_conntrack: can't register to sysctl.\n");
576		ret = -ENOMEM;
577		goto out_sysctl;
578	}
579#endif
580
581	ret = register_pernet_subsys(&nf_conntrack_net_ops);
582	if (ret < 0)
583		goto out_pernet;
584
585	nf_conntrack_init_end();
586	return 0;
587
588out_pernet:
589#ifdef CONFIG_SYSCTL
590	unregister_net_sysctl_table(nf_ct_netfilter_header);
591out_sysctl:
592#endif
593	nf_conntrack_cleanup_end();
594out_start:
595	return ret;
596}
597
598static void __exit nf_conntrack_standalone_fini(void)
599{
600	nf_conntrack_cleanup_start();
601	unregister_pernet_subsys(&nf_conntrack_net_ops);
602#ifdef CONFIG_SYSCTL
603	unregister_net_sysctl_table(nf_ct_netfilter_header);
604#endif
605	nf_conntrack_cleanup_end();
606}
607
608module_init(nf_conntrack_standalone_init);
609module_exit(nf_conntrack_standalone_fini);
610
611/* Some modules need us, but don't depend directly on any symbol.
612   They should call this. */
613void need_conntrack(void)
614{
615}
616EXPORT_SYMBOL_GPL(need_conntrack);
617