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