netfilter: xt_NFQUEUE: coalesce IPv4 and IPv6 hashing
[linux-3.10.git] / net / netfilter / nf_conntrack_standalone.c
1 /* (C) 1999-2001 Paul `Rusty' Russell
2  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
3  *
4  * This program is free software; you can redistribute it and/or modify
5  * it under the terms of the GNU General Public License version 2 as
6  * published by the Free Software Foundation.
7  */
8
9 #include <linux/types.h>
10 #include <linux/netfilter.h>
11 #include <linux/slab.h>
12 #include <linux/module.h>
13 #include <linux/skbuff.h>
14 #include <linux/proc_fs.h>
15 #include <linux/seq_file.h>
16 #include <linux/percpu.h>
17 #include <linux/netdevice.h>
18 #include <linux/security.h>
19 #include <net/net_namespace.h>
20 #ifdef CONFIG_SYSCTL
21 #include <linux/sysctl.h>
22 #endif
23
24 #include <net/netfilter/nf_conntrack.h>
25 #include <net/netfilter/nf_conntrack_core.h>
26 #include <net/netfilter/nf_conntrack_l3proto.h>
27 #include <net/netfilter/nf_conntrack_l4proto.h>
28 #include <net/netfilter/nf_conntrack_expect.h>
29 #include <net/netfilter/nf_conntrack_helper.h>
30 #include <net/netfilter/nf_conntrack_acct.h>
31 #include <net/netfilter/nf_conntrack_zones.h>
32 #include <net/netfilter/nf_conntrack_timestamp.h>
33 #include <linux/rculist_nulls.h>
34
35 MODULE_LICENSE("GPL");
36
37 #ifdef CONFIG_NF_CONNTRACK_PROCFS
38 int
39 print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
40             const struct nf_conntrack_l3proto *l3proto,
41             const struct nf_conntrack_l4proto *l4proto)
42 {
43         return l3proto->print_tuple(s, tuple) || l4proto->print_tuple(s, tuple);
44 }
45 EXPORT_SYMBOL_GPL(print_tuple);
46
47 struct ct_iter_state {
48         struct seq_net_private p;
49         unsigned int bucket;
50         u_int64_t time_now;
51 };
52
53 static struct hlist_nulls_node *ct_get_first(struct seq_file *seq)
54 {
55         struct net *net = seq_file_net(seq);
56         struct ct_iter_state *st = seq->private;
57         struct hlist_nulls_node *n;
58
59         for (st->bucket = 0;
60              st->bucket < net->ct.htable_size;
61              st->bucket++) {
62                 n = rcu_dereference(hlist_nulls_first_rcu(&net->ct.hash[st->bucket]));
63                 if (!is_a_nulls(n))
64                         return n;
65         }
66         return NULL;
67 }
68
69 static struct hlist_nulls_node *ct_get_next(struct seq_file *seq,
70                                       struct hlist_nulls_node *head)
71 {
72         struct net *net = seq_file_net(seq);
73         struct ct_iter_state *st = seq->private;
74
75         head = rcu_dereference(hlist_nulls_next_rcu(head));
76         while (is_a_nulls(head)) {
77                 if (likely(get_nulls_value(head) == st->bucket)) {
78                         if (++st->bucket >= net->ct.htable_size)
79                                 return NULL;
80                 }
81                 head = rcu_dereference(
82                                 hlist_nulls_first_rcu(
83                                         &net->ct.hash[st->bucket]));
84         }
85         return head;
86 }
87
88 static struct hlist_nulls_node *ct_get_idx(struct seq_file *seq, loff_t pos)
89 {
90         struct hlist_nulls_node *head = ct_get_first(seq);
91
92         if (head)
93                 while (pos && (head = ct_get_next(seq, head)))
94                         pos--;
95         return pos ? NULL : head;
96 }
97
98 static void *ct_seq_start(struct seq_file *seq, loff_t *pos)
99         __acquires(RCU)
100 {
101         struct ct_iter_state *st = seq->private;
102
103         st->time_now = ktime_to_ns(ktime_get_real());
104         rcu_read_lock();
105         return ct_get_idx(seq, *pos);
106 }
107
108 static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
109 {
110         (*pos)++;
111         return ct_get_next(s, v);
112 }
113
114 static void ct_seq_stop(struct seq_file *s, void *v)
115         __releases(RCU)
116 {
117         rcu_read_unlock();
118 }
119
120 #ifdef CONFIG_NF_CONNTRACK_SECMARK
121 static int ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
122 {
123         int ret;
124         u32 len;
125         char *secctx;
126
127         ret = security_secid_to_secctx(ct->secmark, &secctx, &len);
128         if (ret)
129                 return 0;
130
131         ret = seq_printf(s, "secctx=%s ", secctx);
132
133         security_release_secctx(secctx, len);
134         return ret;
135 }
136 #else
137 static inline int ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
138 {
139         return 0;
140 }
141 #endif
142
143 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
144 static int 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                 return seq_printf(s, "delta-time=%llu ",
159                                   (unsigned long long)delta_time);
160         }
161         return 0;
162 }
163 #else
164 static inline int
165 ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
166 {
167         return 0;
168 }
169 #endif
170
171 /* return 0 on success, 1 in case of error */
172 static int ct_seq_show(struct seq_file *s, void *v)
173 {
174         struct nf_conntrack_tuple_hash *hash = v;
175         struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(hash);
176         const struct nf_conntrack_l3proto *l3proto;
177         const struct nf_conntrack_l4proto *l4proto;
178         int ret = 0;
179
180         NF_CT_ASSERT(ct);
181         if (unlikely(!atomic_inc_not_zero(&ct->ct_general.use)))
182                 return 0;
183
184         /* we only want to print DIR_ORIGINAL */
185         if (NF_CT_DIRECTION(hash))
186                 goto release;
187
188         l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct));
189         NF_CT_ASSERT(l3proto);
190         l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
191         NF_CT_ASSERT(l4proto);
192
193         ret = -ENOSPC;
194         if (seq_printf(s, "%-8s %u %-8s %u %ld ",
195                        l3proto->name, nf_ct_l3num(ct),
196                        l4proto->name, nf_ct_protonum(ct),
197                        timer_pending(&ct->timeout)
198                        ? (long)(ct->timeout.expires - jiffies)/HZ : 0) != 0)
199                 goto release;
200
201         if (l4proto->print_conntrack && l4proto->print_conntrack(s, ct))
202                 goto release;
203
204         if (print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
205                         l3proto, l4proto))
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                 if (seq_printf(s, "[UNREPLIED] "))
213                         goto release;
214
215         if (print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple,
216                         l3proto, l4proto))
217                 goto release;
218
219         if (seq_print_acct(s, ct, IP_CT_DIR_REPLY))
220                 goto release;
221
222         if (test_bit(IPS_ASSURED_BIT, &ct->status))
223                 if (seq_printf(s, "[ASSURED] "))
224                         goto release;
225
226 #if defined(CONFIG_NF_CONNTRACK_MARK)
227         if (seq_printf(s, "mark=%u ", ct->mark))
228                 goto release;
229 #endif
230
231         if (ct_show_secctx(s, ct))
232                 goto release;
233
234 #ifdef CONFIG_NF_CONNTRACK_ZONES
235         if (seq_printf(s, "zone=%u ", nf_ct_zone(ct)))
236                 goto release;
237 #endif
238
239         if (ct_show_delta_time(s, ct))
240                 goto release;
241
242         if (seq_printf(s, "use=%u\n", atomic_read(&ct->ct_general.use)))
243                 goto release;
244
245         ret = 0;
246 release:
247         nf_ct_put(ct);
248         return ret;
249 }
250
251 static const struct seq_operations ct_seq_ops = {
252         .start = ct_seq_start,
253         .next  = ct_seq_next,
254         .stop  = ct_seq_stop,
255         .show  = ct_seq_show
256 };
257
258 static int ct_open(struct inode *inode, struct file *file)
259 {
260         return seq_open_net(inode, file, &ct_seq_ops,
261                         sizeof(struct ct_iter_state));
262 }
263
264 static const struct file_operations ct_file_ops = {
265         .owner   = THIS_MODULE,
266         .open    = ct_open,
267         .read    = seq_read,
268         .llseek  = seq_lseek,
269         .release = seq_release_net,
270 };
271
272 static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
273 {
274         struct net *net = seq_file_net(seq);
275         int cpu;
276
277         if (*pos == 0)
278                 return SEQ_START_TOKEN;
279
280         for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
281                 if (!cpu_possible(cpu))
282                         continue;
283                 *pos = cpu + 1;
284                 return per_cpu_ptr(net->ct.stat, cpu);
285         }
286
287         return NULL;
288 }
289
290 static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
291 {
292         struct net *net = seq_file_net(seq);
293         int cpu;
294
295         for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
296                 if (!cpu_possible(cpu))
297                         continue;
298                 *pos = cpu + 1;
299                 return per_cpu_ptr(net->ct.stat, cpu);
300         }
301
302         return NULL;
303 }
304
305 static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
306 {
307 }
308
309 static int ct_cpu_seq_show(struct seq_file *seq, void *v)
310 {
311         struct net *net = seq_file_net(seq);
312         unsigned int nr_conntracks = atomic_read(&net->ct.count);
313         const struct ip_conntrack_stat *st = v;
314
315         if (v == SEQ_START_TOKEN) {
316                 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");
317                 return 0;
318         }
319
320         seq_printf(seq, "%08x  %08x %08x %08x %08x %08x %08x %08x "
321                         "%08x %08x %08x %08x %08x  %08x %08x %08x %08x\n",
322                    nr_conntracks,
323                    st->searched,
324                    st->found,
325                    st->new,
326                    st->invalid,
327                    st->ignore,
328                    st->delete,
329                    st->delete_list,
330                    st->insert,
331                    st->insert_failed,
332                    st->drop,
333                    st->early_drop,
334                    st->error,
335
336                    st->expect_new,
337                    st->expect_create,
338                    st->expect_delete,
339                    st->search_restart
340                 );
341         return 0;
342 }
343
344 static const struct seq_operations ct_cpu_seq_ops = {
345         .start  = ct_cpu_seq_start,
346         .next   = ct_cpu_seq_next,
347         .stop   = ct_cpu_seq_stop,
348         .show   = ct_cpu_seq_show,
349 };
350
351 static int ct_cpu_seq_open(struct inode *inode, struct file *file)
352 {
353         return seq_open_net(inode, file, &ct_cpu_seq_ops,
354                             sizeof(struct seq_net_private));
355 }
356
357 static const struct file_operations ct_cpu_seq_fops = {
358         .owner   = THIS_MODULE,
359         .open    = ct_cpu_seq_open,
360         .read    = seq_read,
361         .llseek  = seq_lseek,
362         .release = seq_release_net,
363 };
364
365 static int nf_conntrack_standalone_init_proc(struct net *net)
366 {
367         struct proc_dir_entry *pde;
368
369         pde = proc_create("nf_conntrack", 0440, net->proc_net, &ct_file_ops);
370         if (!pde)
371                 goto out_nf_conntrack;
372
373         pde = proc_create("nf_conntrack", S_IRUGO, net->proc_net_stat,
374                           &ct_cpu_seq_fops);
375         if (!pde)
376                 goto out_stat_nf_conntrack;
377         return 0;
378
379 out_stat_nf_conntrack:
380         remove_proc_entry("nf_conntrack", net->proc_net);
381 out_nf_conntrack:
382         return -ENOMEM;
383 }
384
385 static void nf_conntrack_standalone_fini_proc(struct net *net)
386 {
387         remove_proc_entry("nf_conntrack", net->proc_net_stat);
388         remove_proc_entry("nf_conntrack", net->proc_net);
389 }
390 #else
391 static int nf_conntrack_standalone_init_proc(struct net *net)
392 {
393         return 0;
394 }
395
396 static void nf_conntrack_standalone_fini_proc(struct net *net)
397 {
398 }
399 #endif /* CONFIG_NF_CONNTRACK_PROCFS */
400
401 /* Sysctl support */
402
403 #ifdef CONFIG_SYSCTL
404 /* Log invalid packets of a given protocol */
405 static int log_invalid_proto_min = 0;
406 static int log_invalid_proto_max = 255;
407
408 static struct ctl_table_header *nf_ct_netfilter_header;
409
410 static ctl_table nf_ct_sysctl_table[] = {
411         {
412                 .procname       = "nf_conntrack_max",
413                 .data           = &nf_conntrack_max,
414                 .maxlen         = sizeof(int),
415                 .mode           = 0644,
416                 .proc_handler   = proc_dointvec,
417         },
418         {
419                 .procname       = "nf_conntrack_count",
420                 .data           = &init_net.ct.count,
421                 .maxlen         = sizeof(int),
422                 .mode           = 0444,
423                 .proc_handler   = proc_dointvec,
424         },
425         {
426                 .procname       = "nf_conntrack_buckets",
427                 .data           = &init_net.ct.htable_size,
428                 .maxlen         = sizeof(unsigned int),
429                 .mode           = 0444,
430                 .proc_handler   = proc_dointvec,
431         },
432         {
433                 .procname       = "nf_conntrack_checksum",
434                 .data           = &init_net.ct.sysctl_checksum,
435                 .maxlen         = sizeof(unsigned int),
436                 .mode           = 0644,
437                 .proc_handler   = proc_dointvec,
438         },
439         {
440                 .procname       = "nf_conntrack_log_invalid",
441                 .data           = &init_net.ct.sysctl_log_invalid,
442                 .maxlen         = sizeof(unsigned int),
443                 .mode           = 0644,
444                 .proc_handler   = proc_dointvec_minmax,
445                 .extra1         = &log_invalid_proto_min,
446                 .extra2         = &log_invalid_proto_max,
447         },
448         {
449                 .procname       = "nf_conntrack_expect_max",
450                 .data           = &nf_ct_expect_max,
451                 .maxlen         = sizeof(int),
452                 .mode           = 0644,
453                 .proc_handler   = proc_dointvec,
454         },
455         { }
456 };
457
458 #define NET_NF_CONNTRACK_MAX 2089
459
460 static ctl_table nf_ct_netfilter_table[] = {
461         {
462                 .procname       = "nf_conntrack_max",
463                 .data           = &nf_conntrack_max,
464                 .maxlen         = sizeof(int),
465                 .mode           = 0644,
466                 .proc_handler   = proc_dointvec,
467         },
468         { }
469 };
470
471 static int nf_conntrack_standalone_init_sysctl(struct net *net)
472 {
473         struct ctl_table *table;
474
475         table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table),
476                         GFP_KERNEL);
477         if (!table)
478                 goto out_kmemdup;
479
480         table[1].data = &net->ct.count;
481         table[2].data = &net->ct.htable_size;
482         table[3].data = &net->ct.sysctl_checksum;
483         table[4].data = &net->ct.sysctl_log_invalid;
484
485         /* Don't export sysctls to unprivileged users */
486         if (net->user_ns != &init_user_ns)
487                 table[0].procname = NULL;
488
489         net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table);
490         if (!net->ct.sysctl_header)
491                 goto out_unregister_netfilter;
492
493         return 0;
494
495 out_unregister_netfilter:
496         kfree(table);
497 out_kmemdup:
498         return -ENOMEM;
499 }
500
501 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
502 {
503         struct ctl_table *table;
504
505         table = net->ct.sysctl_header->ctl_table_arg;
506         unregister_net_sysctl_table(net->ct.sysctl_header);
507         kfree(table);
508 }
509 #else
510 static int nf_conntrack_standalone_init_sysctl(struct net *net)
511 {
512         return 0;
513 }
514
515 static void nf_conntrack_standalone_fini_sysctl(struct net *net)
516 {
517 }
518 #endif /* CONFIG_SYSCTL */
519
520 static int nf_conntrack_pernet_init(struct net *net)
521 {
522         int ret;
523
524         ret = nf_conntrack_init_net(net);
525         if (ret < 0)
526                 goto out_init;
527
528         ret = nf_conntrack_standalone_init_proc(net);
529         if (ret < 0)
530                 goto out_proc;
531
532         net->ct.sysctl_checksum = 1;
533         net->ct.sysctl_log_invalid = 0;
534         ret = nf_conntrack_standalone_init_sysctl(net);
535         if (ret < 0)
536                 goto out_sysctl;
537
538         return 0;
539
540 out_sysctl:
541         nf_conntrack_standalone_fini_proc(net);
542 out_proc:
543         nf_conntrack_cleanup_net(net);
544 out_init:
545         return ret;
546 }
547
548 static void nf_conntrack_pernet_exit(struct list_head *net_exit_list)
549 {
550         struct net *net;
551
552         list_for_each_entry(net, net_exit_list, exit_list) {
553                 nf_conntrack_standalone_fini_sysctl(net);
554                 nf_conntrack_standalone_fini_proc(net);
555         }
556         nf_conntrack_cleanup_net_list(net_exit_list);
557 }
558
559 static struct pernet_operations nf_conntrack_net_ops = {
560         .init           = nf_conntrack_pernet_init,
561         .exit_batch     = nf_conntrack_pernet_exit,
562 };
563
564 static int __init nf_conntrack_standalone_init(void)
565 {
566         int ret = nf_conntrack_init_start();
567         if (ret < 0)
568                 goto out_start;
569
570 #ifdef CONFIG_SYSCTL
571         nf_ct_netfilter_header =
572                 register_net_sysctl(&init_net, "net", nf_ct_netfilter_table);
573         if (!nf_ct_netfilter_header) {
574                 pr_err("nf_conntrack: can't register to sysctl.\n");
575                 goto out_sysctl;
576         }
577 #endif
578
579         ret = register_pernet_subsys(&nf_conntrack_net_ops);
580         if (ret < 0)
581                 goto out_pernet;
582
583         nf_conntrack_init_end();
584         return 0;
585
586 out_pernet:
587 #ifdef CONFIG_SYSCTL
588         unregister_net_sysctl_table(nf_ct_netfilter_header);
589 out_sysctl:
590 #endif
591         nf_conntrack_cleanup_end();
592 out_start:
593         return ret;
594 }
595
596 static void __exit nf_conntrack_standalone_fini(void)
597 {
598         nf_conntrack_cleanup_start();
599         unregister_pernet_subsys(&nf_conntrack_net_ops);
600 #ifdef CONFIG_SYSCTL
601         unregister_net_sysctl_table(nf_ct_netfilter_header);
602 #endif
603         nf_conntrack_cleanup_end();
604 }
605
606 module_init(nf_conntrack_standalone_init);
607 module_exit(nf_conntrack_standalone_fini);
608
609 /* Some modules need us, but don't depend directly on any symbol.
610    They should call this. */
611 void need_conntrack(void)
612 {
613 }
614 EXPORT_SYMBOL_GPL(need_conntrack);