1// SPDX-License-Identifier: GPL-2.0
2#include <linux/types.h>
3#include <linux/netfilter.h>
4#include <linux/slab.h>
5#include <linux/module.h>
6#include <linux/skbuff.h>
7#include <linux/proc_fs.h>
8#include <linux/seq_file.h>
9#include <linux/percpu.h>
10#include <linux/netdevice.h>
11#include <linux/security.h>
12#include <net/net_namespace.h>
13#ifdef CONFIG_SYSCTL
14#include <linux/sysctl.h>
15#endif
16
17#include <net/netfilter/nf_log.h>
18#include <net/netfilter/nf_conntrack.h>
19#include <net/netfilter/nf_conntrack_core.h>
20#include <net/netfilter/nf_conntrack_l4proto.h>
21#include <net/netfilter/nf_conntrack_expect.h>
22#include <net/netfilter/nf_conntrack_helper.h>
23#include <net/netfilter/nf_conntrack_acct.h>
24#include <net/netfilter/nf_conntrack_zones.h>
25#include <net/netfilter/nf_conntrack_timestamp.h>
26#include <linux/rculist_nulls.h>
27
28static bool enable_hooks __read_mostly;
29MODULE_PARM_DESC(enable_hooks, "Always enable conntrack hooks");
30module_param(enable_hooks, bool, 0000);
31
32unsigned int nf_conntrack_net_id __read_mostly;
33
34#ifdef CONFIG_NF_CONNTRACK_PROCFS
35void
36print_tuple(struct seq_file *s, const struct nf_conntrack_tuple *tuple,
37 const struct nf_conntrack_l4proto *l4proto)
38{
39 switch (tuple->src.l3num) {
40 case NFPROTO_IPV4:
41 seq_printf(s, "src=%pI4 dst=%pI4 ",
42 &tuple->src.u3.ip, &tuple->dst.u3.ip);
43 break;
44 case NFPROTO_IPV6:
45 seq_printf(s, "src=%pI6 dst=%pI6 ",
46 tuple->src.u3.ip6, tuple->dst.u3.ip6);
47 break;
48 default:
49 break;
50 }
51
52 switch (l4proto->l4proto) {
53 case IPPROTO_ICMP:
54 seq_printf(s, "type=%u code=%u id=%u ",
55 tuple->dst.u.icmp.type,
56 tuple->dst.u.icmp.code,
57 ntohs(tuple->src.u.icmp.id));
58 break;
59 case IPPROTO_TCP:
60 seq_printf(s, "sport=%hu dport=%hu ",
61 ntohs(tuple->src.u.tcp.port),
62 ntohs(tuple->dst.u.tcp.port));
63 break;
64 case IPPROTO_UDPLITE:
65 case IPPROTO_UDP:
66 seq_printf(s, "sport=%hu dport=%hu ",
67 ntohs(tuple->src.u.udp.port),
68 ntohs(tuple->dst.u.udp.port));
69
70 break;
71 case IPPROTO_SCTP:
72 seq_printf(s, "sport=%hu dport=%hu ",
73 ntohs(tuple->src.u.sctp.port),
74 ntohs(tuple->dst.u.sctp.port));
75 break;
76 case IPPROTO_ICMPV6:
77 seq_printf(s, "type=%u code=%u id=%u ",
78 tuple->dst.u.icmp.type,
79 tuple->dst.u.icmp.code,
80 ntohs(tuple->src.u.icmp.id));
81 break;
82 case IPPROTO_GRE:
83 seq_printf(s, "srckey=0x%x dstkey=0x%x ",
84 ntohs(tuple->src.u.gre.key),
85 ntohs(tuple->dst.u.gre.key));
86 break;
87 default:
88 break;
89 }
90}
91EXPORT_SYMBOL_GPL(print_tuple);
92
93struct ct_iter_state {
94 struct seq_net_private p;
95 struct hlist_nulls_head *hash;
96 unsigned int htable_size;
97 unsigned int skip_elems;
98 unsigned int bucket;
99 u_int64_t time_now;
100};
101
102static struct nf_conntrack_tuple_hash *ct_get_next(const struct net *net,
103 struct ct_iter_state *st)
104{
105 struct nf_conntrack_tuple_hash *h;
106 struct hlist_nulls_node *n;
107 unsigned int i;
108
109 for (i = st->bucket; i < st->htable_size; i++) {
110 unsigned int skip = 0;
111
112restart:
113 hlist_nulls_for_each_entry_rcu(h, n, &st->hash[i], hnnode) {
114 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
115 struct hlist_nulls_node *tmp = n;
116
117 if (!net_eq(net, nf_ct_net(ct)))
118 continue;
119
120 if (++skip <= st->skip_elems)
121 continue;
122
123 /* h should be returned, skip to nulls marker. */
124 while (!is_a_nulls(tmp))
125 tmp = rcu_dereference(hlist_nulls_next_rcu(tmp));
126
127 /* check if h is still linked to hash[i] */
128 if (get_nulls_value(tmp) != i) {
129 skip = 0;
130 goto restart;
131 }
132
133 st->skip_elems = skip;
134 st->bucket = i;
135 return h;
136 }
137
138 skip = 0;
139 if (get_nulls_value(n) != i)
140 goto restart;
141
142 st->skip_elems = 0;
143 }
144
145 st->bucket = i;
146 return NULL;
147}
148
149static void *ct_seq_start(struct seq_file *seq, loff_t *pos)
150 __acquires(RCU)
151{
152 struct ct_iter_state *st = seq->private;
153 struct net *net = seq_file_net(seq);
154
155 st->time_now = ktime_get_real_ns();
156 rcu_read_lock();
157
158 nf_conntrack_get_ht(&st->hash, &st->htable_size);
159
160 if (*pos == 0) {
161 st->skip_elems = 0;
162 st->bucket = 0;
163 } else if (st->skip_elems) {
164 /* resume from last dumped entry */
165 st->skip_elems--;
166 }
167
168 return ct_get_next(net, st);
169}
170
171static void *ct_seq_next(struct seq_file *s, void *v, loff_t *pos)
172{
173 struct ct_iter_state *st = s->private;
174 struct net *net = seq_file_net(s);
175
176 (*pos)++;
177 return ct_get_next(net, st);
178}
179
180static void ct_seq_stop(struct seq_file *s, void *v)
181 __releases(RCU)
182{
183 rcu_read_unlock();
184}
185
186#ifdef CONFIG_NF_CONNTRACK_SECMARK
187static void ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
188{
189 struct lsm_context ctx;
190 int ret;
191
192 ret = security_secid_to_secctx(ct->secmark, &ctx);
193 if (ret < 0)
194 return;
195
196 seq_printf(s, "secctx=%s ", ctx.context);
197
198 security_release_secctx(&ctx);
199}
200#else
201static inline void ct_show_secctx(struct seq_file *s, const struct nf_conn *ct)
202{
203}
204#endif
205
206#ifdef CONFIG_NF_CONNTRACK_ZONES
207static void ct_show_zone(struct seq_file *s, const struct nf_conn *ct,
208 int dir)
209{
210 const struct nf_conntrack_zone *zone = nf_ct_zone(ct);
211
212 if (zone->dir != dir)
213 return;
214 switch (zone->dir) {
215 case NF_CT_DEFAULT_ZONE_DIR:
216 seq_printf(s, "zone=%u ", zone->id);
217 break;
218 case NF_CT_ZONE_DIR_ORIG:
219 seq_printf(s, "zone-orig=%u ", zone->id);
220 break;
221 case NF_CT_ZONE_DIR_REPL:
222 seq_printf(s, "zone-reply=%u ", zone->id);
223 break;
224 default:
225 break;
226 }
227}
228#else
229static inline void ct_show_zone(struct seq_file *s, const struct nf_conn *ct,
230 int dir)
231{
232}
233#endif
234
235#ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
236static void ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
237{
238 struct ct_iter_state *st = s->private;
239 struct nf_conn_tstamp *tstamp;
240 s64 delta_time;
241
242 tstamp = nf_conn_tstamp_find(ct);
243 if (tstamp) {
244 delta_time = st->time_now - tstamp->start;
245 if (delta_time > 0)
246 delta_time = div_s64(delta_time, NSEC_PER_SEC);
247 else
248 delta_time = 0;
249
250 seq_printf(s, "delta-time=%llu ",
251 (unsigned long long)delta_time);
252 }
253 return;
254}
255#else
256static inline void
257ct_show_delta_time(struct seq_file *s, const struct nf_conn *ct)
258{
259}
260#endif
261
262static const char* l3proto_name(u16 proto)
263{
264 switch (proto) {
265 case AF_INET: return "ipv4";
266 case AF_INET6: return "ipv6";
267 }
268
269 return "unknown";
270}
271
272static const char* l4proto_name(u16 proto)
273{
274 switch (proto) {
275 case IPPROTO_ICMP: return "icmp";
276 case IPPROTO_TCP: return "tcp";
277 case IPPROTO_UDP: return "udp";
278 case IPPROTO_GRE: return "gre";
279 case IPPROTO_SCTP: return "sctp";
280 case IPPROTO_UDPLITE: return "udplite";
281 case IPPROTO_ICMPV6: return "icmpv6";
282 }
283
284 return "unknown";
285}
286
287static void
288seq_print_acct(struct seq_file *s, const struct nf_conn *ct, int dir)
289{
290 struct nf_conn_acct *acct;
291 struct nf_conn_counter *counter;
292
293 acct = nf_conn_acct_find(ct);
294 if (!acct)
295 return;
296
297 counter = acct->counter;
298 seq_printf(s, "packets=%llu bytes=%llu ",
299 (unsigned long long)atomic64_read(&counter[dir].packets),
300 (unsigned long long)atomic64_read(&counter[dir].bytes));
301}
302
303/* return 0 on success, 1 in case of error */
304static int ct_seq_show(struct seq_file *s, void *v)
305{
306 struct nf_conntrack_tuple_hash *hash = v;
307 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(hash);
308 const struct nf_conntrack_l4proto *l4proto;
309 struct net *net = seq_file_net(s);
310 int ret = 0;
311
312 WARN_ON(!ct);
313 if (unlikely(!refcount_inc_not_zero(&ct->ct_general.use)))
314 return 0;
315
316 /* load ->status after refcount increase */
317 smp_acquire__after_ctrl_dep();
318
319 if (nf_ct_should_gc(ct)) {
320 struct ct_iter_state *st = s->private;
321
322 st->skip_elems--;
323 nf_ct_kill(ct);
324 goto release;
325 }
326
327 /* we only want to print DIR_ORIGINAL */
328 if (NF_CT_DIRECTION(hash))
329 goto release;
330
331 if (!net_eq(nf_ct_net(ct), net))
332 goto release;
333
334 l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct));
335
336 ret = -ENOSPC;
337 seq_printf(s, "%-8s %u %-8s %u ",
338 l3proto_name(nf_ct_l3num(ct)), nf_ct_l3num(ct),
339 l4proto_name(l4proto->l4proto), nf_ct_protonum(ct));
340
341 if (!test_bit(IPS_OFFLOAD_BIT, &ct->status))
342 seq_printf(s, "%ld ", nf_ct_expires(ct) / HZ);
343
344 if (l4proto->print_conntrack)
345 l4proto->print_conntrack(s, ct);
346
347 print_tuple(s, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
348 l4proto);
349
350 ct_show_zone(s, ct, NF_CT_ZONE_DIR_ORIG);
351
352 if (seq_has_overflowed(s))
353 goto release;
354
355 seq_print_acct(s, ct, IP_CT_DIR_ORIGINAL);
356
357 if (!(test_bit(IPS_SEEN_REPLY_BIT, &ct->status)))
358 seq_puts(s, "[UNREPLIED] ");
359
360 print_tuple(s, &ct->tuplehash[IP_CT_DIR_REPLY].tuple, l4proto);
361
362 ct_show_zone(s, ct, NF_CT_ZONE_DIR_REPL);
363
364 seq_print_acct(s, ct, IP_CT_DIR_REPLY);
365
366 if (test_bit(IPS_HW_OFFLOAD_BIT, &ct->status))
367 seq_puts(s, "[HW_OFFLOAD] ");
368 else if (test_bit(IPS_OFFLOAD_BIT, &ct->status))
369 seq_puts(s, "[OFFLOAD] ");
370 else if (test_bit(IPS_ASSURED_BIT, &ct->status))
371 seq_puts(s, "[ASSURED] ");
372
373 if (seq_has_overflowed(s))
374 goto release;
375
376#if defined(CONFIG_NF_CONNTRACK_MARK)
377 seq_printf(s, "mark=%u ", READ_ONCE(ct->mark));
378#endif
379
380 ct_show_secctx(s, ct);
381 ct_show_zone(s, ct, NF_CT_DEFAULT_ZONE_DIR);
382 ct_show_delta_time(s, ct);
383
384 seq_printf(s, "use=%u\n", refcount_read(&ct->ct_general.use));
385
386 if (seq_has_overflowed(s))
387 goto release;
388
389 ret = 0;
390release:
391 nf_ct_put(ct);
392 return ret;
393}
394
395static const struct seq_operations ct_seq_ops = {
396 .start = ct_seq_start,
397 .next = ct_seq_next,
398 .stop = ct_seq_stop,
399 .show = ct_seq_show
400};
401
402static void *ct_cpu_seq_start(struct seq_file *seq, loff_t *pos)
403{
404 struct net *net = seq_file_net(seq);
405 int cpu;
406
407 if (*pos == 0)
408 return SEQ_START_TOKEN;
409
410 for (cpu = *pos-1; cpu < nr_cpu_ids; ++cpu) {
411 if (!cpu_possible(cpu))
412 continue;
413 *pos = cpu + 1;
414 return per_cpu_ptr(net->ct.stat, cpu);
415 }
416
417 return NULL;
418}
419
420static void *ct_cpu_seq_next(struct seq_file *seq, void *v, loff_t *pos)
421{
422 struct net *net = seq_file_net(seq);
423 int cpu;
424
425 for (cpu = *pos; cpu < nr_cpu_ids; ++cpu) {
426 if (!cpu_possible(cpu))
427 continue;
428 *pos = cpu + 1;
429 return per_cpu_ptr(net->ct.stat, cpu);
430 }
431 (*pos)++;
432 return NULL;
433}
434
435static void ct_cpu_seq_stop(struct seq_file *seq, void *v)
436{
437}
438
439static int ct_cpu_seq_show(struct seq_file *seq, void *v)
440{
441 struct net *net = seq_file_net(seq);
442 const struct ip_conntrack_stat *st = v;
443 unsigned int nr_conntracks;
444
445 if (v == SEQ_START_TOKEN) {
446 seq_puts(seq, "entries clashres found new invalid ignore delete chainlength insert insert_failed drop early_drop icmp_error expect_new expect_create expect_delete search_restart\n");
447 return 0;
448 }
449
450 nr_conntracks = nf_conntrack_count(net);
451
452 seq_printf(seq, "%08x %08x %08x %08x %08x %08x %08x %08x "
453 "%08x %08x %08x %08x %08x %08x %08x %08x %08x\n",
454 nr_conntracks,
455 st->clash_resolve,
456 st->found,
457 0,
458 st->invalid,
459 0,
460 0,
461 st->chaintoolong,
462 st->insert,
463 st->insert_failed,
464 st->drop,
465 st->early_drop,
466 st->error,
467
468 st->expect_new,
469 st->expect_create,
470 st->expect_delete,
471 st->search_restart
472 );
473 return 0;
474}
475
476static const struct seq_operations ct_cpu_seq_ops = {
477 .start = ct_cpu_seq_start,
478 .next = ct_cpu_seq_next,
479 .stop = ct_cpu_seq_stop,
480 .show = ct_cpu_seq_show,
481};
482
483static int nf_conntrack_standalone_init_proc(struct net *net)
484{
485 struct proc_dir_entry *pde;
486 kuid_t root_uid;
487 kgid_t root_gid;
488
489 pde = proc_create_net("nf_conntrack", 0440, net->proc_net, &ct_seq_ops,
490 sizeof(struct ct_iter_state));
491 if (!pde)
492 goto out_nf_conntrack;
493
494 root_uid = make_kuid(net->user_ns, 0);
495 root_gid = make_kgid(net->user_ns, 0);
496 if (uid_valid(root_uid) && gid_valid(root_gid))
497 proc_set_user(pde, root_uid, root_gid);
498
499 pde = proc_create_net("nf_conntrack", 0444, net->proc_net_stat,
500 &ct_cpu_seq_ops, sizeof(struct seq_net_private));
501 if (!pde)
502 goto out_stat_nf_conntrack;
503 return 0;
504
505out_stat_nf_conntrack:
506 remove_proc_entry("nf_conntrack", net->proc_net);
507out_nf_conntrack:
508 return -ENOMEM;
509}
510
511static void nf_conntrack_standalone_fini_proc(struct net *net)
512{
513 remove_proc_entry("nf_conntrack", net->proc_net_stat);
514 remove_proc_entry("nf_conntrack", net->proc_net);
515}
516#else
517static int nf_conntrack_standalone_init_proc(struct net *net)
518{
519 return 0;
520}
521
522static void nf_conntrack_standalone_fini_proc(struct net *net)
523{
524}
525#endif /* CONFIG_NF_CONNTRACK_PROCFS */
526
527u32 nf_conntrack_count(const struct net *net)
528{
529 const struct nf_conntrack_net *cnet = nf_ct_pernet(net);
530
531 return atomic_read(v: &cnet->count);
532}
533EXPORT_SYMBOL_GPL(nf_conntrack_count);
534
535/* Sysctl support */
536
537#ifdef CONFIG_SYSCTL
538/* size the user *wants to set */
539static unsigned int nf_conntrack_htable_size_user __read_mostly;
540
541static int
542nf_conntrack_hash_sysctl(const struct ctl_table *table, int write,
543 void *buffer, size_t *lenp, loff_t *ppos)
544{
545 int ret;
546
547 /* module_param hashsize could have changed value */
548 nf_conntrack_htable_size_user = nf_conntrack_htable_size;
549
550 ret = proc_dointvec(table, write, buffer, lenp, ppos);
551 if (ret < 0 || !write)
552 return ret;
553
554 /* update ret, we might not be able to satisfy request */
555 ret = nf_conntrack_hash_resize(hashsize: nf_conntrack_htable_size_user);
556
557 /* update it to the actual value used by conntrack */
558 nf_conntrack_htable_size_user = nf_conntrack_htable_size;
559 return ret;
560}
561
562static int
563nf_conntrack_log_invalid_sysctl(const struct ctl_table *table, int write,
564 void *buffer, size_t *lenp, loff_t *ppos)
565{
566 int ret, i;
567
568 ret = proc_dou8vec_minmax(table, write, buffer, lenp, ppos);
569 if (ret < 0 || !write)
570 return ret;
571
572 if (*(u8 *)table->data == 0)
573 return 0;
574
575 /* Load nf_log_syslog only if no logger is currently registered */
576 for (i = 0; i < NFPROTO_NUMPROTO; i++) {
577 if (nf_log_is_registered(pf: i))
578 return 0;
579 }
580 request_module("%s", "nf_log_syslog");
581
582 return 0;
583}
584
585static struct ctl_table_header *nf_ct_netfilter_header;
586
587enum nf_ct_sysctl_index {
588 NF_SYSCTL_CT_MAX,
589 NF_SYSCTL_CT_COUNT,
590 NF_SYSCTL_CT_BUCKETS,
591 NF_SYSCTL_CT_CHECKSUM,
592 NF_SYSCTL_CT_LOG_INVALID,
593 NF_SYSCTL_CT_EXPECT_MAX,
594 NF_SYSCTL_CT_ACCT,
595#ifdef CONFIG_NF_CONNTRACK_EVENTS
596 NF_SYSCTL_CT_EVENTS,
597#endif
598#ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
599 NF_SYSCTL_CT_TIMESTAMP,
600#endif
601 NF_SYSCTL_CT_PROTO_TIMEOUT_GENERIC,
602 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_SYN_SENT,
603 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_SYN_RECV,
604 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_ESTABLISHED,
605 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_FIN_WAIT,
606 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_CLOSE_WAIT,
607 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_LAST_ACK,
608 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_TIME_WAIT,
609 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_CLOSE,
610 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_RETRANS,
611 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_UNACK,
612#if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
613 NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_OFFLOAD,
614#endif
615 NF_SYSCTL_CT_PROTO_TCP_LOOSE,
616 NF_SYSCTL_CT_PROTO_TCP_LIBERAL,
617 NF_SYSCTL_CT_PROTO_TCP_IGNORE_INVALID_RST,
618 NF_SYSCTL_CT_PROTO_TCP_MAX_RETRANS,
619 NF_SYSCTL_CT_PROTO_TIMEOUT_UDP,
620 NF_SYSCTL_CT_PROTO_TIMEOUT_UDP_STREAM,
621#if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
622 NF_SYSCTL_CT_PROTO_TIMEOUT_UDP_OFFLOAD,
623#endif
624 NF_SYSCTL_CT_PROTO_TIMEOUT_ICMP,
625 NF_SYSCTL_CT_PROTO_TIMEOUT_ICMPV6,
626#ifdef CONFIG_NF_CT_PROTO_SCTP
627 NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_CLOSED,
628 NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_COOKIE_WAIT,
629 NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_COOKIE_ECHOED,
630 NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_ESTABLISHED,
631 NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_SHUTDOWN_SENT,
632 NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_SHUTDOWN_RECD,
633 NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_SHUTDOWN_ACK_SENT,
634 NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_HEARTBEAT_SENT,
635#endif
636#ifdef CONFIG_NF_CT_PROTO_GRE
637 NF_SYSCTL_CT_PROTO_TIMEOUT_GRE,
638 NF_SYSCTL_CT_PROTO_TIMEOUT_GRE_STREAM,
639#endif
640
641 NF_SYSCTL_CT_LAST_SYSCTL,
642};
643
644static struct ctl_table nf_ct_sysctl_table[] = {
645 [NF_SYSCTL_CT_MAX] = {
646 .procname = "nf_conntrack_max",
647 .data = &nf_conntrack_max,
648 .maxlen = sizeof(int),
649 .mode = 0644,
650 .proc_handler = proc_dointvec_minmax,
651 .extra1 = SYSCTL_ZERO,
652 .extra2 = SYSCTL_INT_MAX,
653 },
654 [NF_SYSCTL_CT_COUNT] = {
655 .procname = "nf_conntrack_count",
656 .maxlen = sizeof(int),
657 .mode = 0444,
658 .proc_handler = proc_dointvec,
659 },
660 [NF_SYSCTL_CT_BUCKETS] = {
661 .procname = "nf_conntrack_buckets",
662 .data = &nf_conntrack_htable_size_user,
663 .maxlen = sizeof(unsigned int),
664 .mode = 0644,
665 .proc_handler = nf_conntrack_hash_sysctl,
666 },
667 [NF_SYSCTL_CT_CHECKSUM] = {
668 .procname = "nf_conntrack_checksum",
669 .data = &init_net.ct.sysctl_checksum,
670 .maxlen = sizeof(u8),
671 .mode = 0644,
672 .proc_handler = proc_dou8vec_minmax,
673 .extra1 = SYSCTL_ZERO,
674 .extra2 = SYSCTL_ONE,
675 },
676 [NF_SYSCTL_CT_LOG_INVALID] = {
677 .procname = "nf_conntrack_log_invalid",
678 .data = &init_net.ct.sysctl_log_invalid,
679 .maxlen = sizeof(u8),
680 .mode = 0644,
681 .proc_handler = nf_conntrack_log_invalid_sysctl,
682 },
683 [NF_SYSCTL_CT_EXPECT_MAX] = {
684 .procname = "nf_conntrack_expect_max",
685 .data = &nf_ct_expect_max,
686 .maxlen = sizeof(int),
687 .mode = 0644,
688 .proc_handler = proc_dointvec_minmax,
689 .extra1 = SYSCTL_ONE,
690 .extra2 = SYSCTL_INT_MAX,
691 },
692 [NF_SYSCTL_CT_ACCT] = {
693 .procname = "nf_conntrack_acct",
694 .data = &init_net.ct.sysctl_acct,
695 .maxlen = sizeof(u8),
696 .mode = 0644,
697 .proc_handler = proc_dou8vec_minmax,
698 .extra1 = SYSCTL_ZERO,
699 .extra2 = SYSCTL_ONE,
700 },
701#ifdef CONFIG_NF_CONNTRACK_EVENTS
702 [NF_SYSCTL_CT_EVENTS] = {
703 .procname = "nf_conntrack_events",
704 .data = &init_net.ct.sysctl_events,
705 .maxlen = sizeof(u8),
706 .mode = 0644,
707 .proc_handler = proc_dou8vec_minmax,
708 .extra1 = SYSCTL_ZERO,
709 .extra2 = SYSCTL_TWO,
710 },
711#endif
712#ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
713 [NF_SYSCTL_CT_TIMESTAMP] = {
714 .procname = "nf_conntrack_timestamp",
715 .data = &init_net.ct.sysctl_tstamp,
716 .maxlen = sizeof(u8),
717 .mode = 0644,
718 .proc_handler = proc_dou8vec_minmax,
719 .extra1 = SYSCTL_ZERO,
720 .extra2 = SYSCTL_ONE,
721 },
722#endif
723 [NF_SYSCTL_CT_PROTO_TIMEOUT_GENERIC] = {
724 .procname = "nf_conntrack_generic_timeout",
725 .maxlen = sizeof(unsigned int),
726 .mode = 0644,
727 .proc_handler = proc_dointvec_jiffies,
728 },
729 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_SYN_SENT] = {
730 .procname = "nf_conntrack_tcp_timeout_syn_sent",
731 .maxlen = sizeof(unsigned int),
732 .mode = 0644,
733 .proc_handler = proc_dointvec_jiffies,
734 },
735 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_SYN_RECV] = {
736 .procname = "nf_conntrack_tcp_timeout_syn_recv",
737 .maxlen = sizeof(unsigned int),
738 .mode = 0644,
739 .proc_handler = proc_dointvec_jiffies,
740 },
741 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_ESTABLISHED] = {
742 .procname = "nf_conntrack_tcp_timeout_established",
743 .maxlen = sizeof(unsigned int),
744 .mode = 0644,
745 .proc_handler = proc_dointvec_jiffies,
746 },
747 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_FIN_WAIT] = {
748 .procname = "nf_conntrack_tcp_timeout_fin_wait",
749 .maxlen = sizeof(unsigned int),
750 .mode = 0644,
751 .proc_handler = proc_dointvec_jiffies,
752 },
753 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_CLOSE_WAIT] = {
754 .procname = "nf_conntrack_tcp_timeout_close_wait",
755 .maxlen = sizeof(unsigned int),
756 .mode = 0644,
757 .proc_handler = proc_dointvec_jiffies,
758 },
759 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_LAST_ACK] = {
760 .procname = "nf_conntrack_tcp_timeout_last_ack",
761 .maxlen = sizeof(unsigned int),
762 .mode = 0644,
763 .proc_handler = proc_dointvec_jiffies,
764 },
765 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_TIME_WAIT] = {
766 .procname = "nf_conntrack_tcp_timeout_time_wait",
767 .maxlen = sizeof(unsigned int),
768 .mode = 0644,
769 .proc_handler = proc_dointvec_jiffies,
770 },
771 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_CLOSE] = {
772 .procname = "nf_conntrack_tcp_timeout_close",
773 .maxlen = sizeof(unsigned int),
774 .mode = 0644,
775 .proc_handler = proc_dointvec_jiffies,
776 },
777 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_RETRANS] = {
778 .procname = "nf_conntrack_tcp_timeout_max_retrans",
779 .maxlen = sizeof(unsigned int),
780 .mode = 0644,
781 .proc_handler = proc_dointvec_jiffies,
782 },
783 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_UNACK] = {
784 .procname = "nf_conntrack_tcp_timeout_unacknowledged",
785 .maxlen = sizeof(unsigned int),
786 .mode = 0644,
787 .proc_handler = proc_dointvec_jiffies,
788 },
789#if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
790 [NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_OFFLOAD] = {
791 .procname = "nf_flowtable_tcp_timeout",
792 .maxlen = sizeof(unsigned int),
793 .mode = 0644,
794 .proc_handler = proc_dointvec_jiffies,
795 },
796#endif
797 [NF_SYSCTL_CT_PROTO_TCP_LOOSE] = {
798 .procname = "nf_conntrack_tcp_loose",
799 .maxlen = sizeof(u8),
800 .mode = 0644,
801 .proc_handler = proc_dou8vec_minmax,
802 .extra1 = SYSCTL_ZERO,
803 .extra2 = SYSCTL_ONE,
804 },
805 [NF_SYSCTL_CT_PROTO_TCP_LIBERAL] = {
806 .procname = "nf_conntrack_tcp_be_liberal",
807 .maxlen = sizeof(u8),
808 .mode = 0644,
809 .proc_handler = proc_dou8vec_minmax,
810 .extra1 = SYSCTL_ZERO,
811 .extra2 = SYSCTL_ONE,
812 },
813 [NF_SYSCTL_CT_PROTO_TCP_IGNORE_INVALID_RST] = {
814 .procname = "nf_conntrack_tcp_ignore_invalid_rst",
815 .maxlen = sizeof(u8),
816 .mode = 0644,
817 .proc_handler = proc_dou8vec_minmax,
818 .extra1 = SYSCTL_ZERO,
819 .extra2 = SYSCTL_ONE,
820 },
821 [NF_SYSCTL_CT_PROTO_TCP_MAX_RETRANS] = {
822 .procname = "nf_conntrack_tcp_max_retrans",
823 .maxlen = sizeof(u8),
824 .mode = 0644,
825 .proc_handler = proc_dou8vec_minmax,
826 },
827 [NF_SYSCTL_CT_PROTO_TIMEOUT_UDP] = {
828 .procname = "nf_conntrack_udp_timeout",
829 .maxlen = sizeof(unsigned int),
830 .mode = 0644,
831 .proc_handler = proc_dointvec_jiffies,
832 },
833 [NF_SYSCTL_CT_PROTO_TIMEOUT_UDP_STREAM] = {
834 .procname = "nf_conntrack_udp_timeout_stream",
835 .maxlen = sizeof(unsigned int),
836 .mode = 0644,
837 .proc_handler = proc_dointvec_jiffies,
838 },
839#if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
840 [NF_SYSCTL_CT_PROTO_TIMEOUT_UDP_OFFLOAD] = {
841 .procname = "nf_flowtable_udp_timeout",
842 .maxlen = sizeof(unsigned int),
843 .mode = 0644,
844 .proc_handler = proc_dointvec_jiffies,
845 },
846#endif
847 [NF_SYSCTL_CT_PROTO_TIMEOUT_ICMP] = {
848 .procname = "nf_conntrack_icmp_timeout",
849 .maxlen = sizeof(unsigned int),
850 .mode = 0644,
851 .proc_handler = proc_dointvec_jiffies,
852 },
853 [NF_SYSCTL_CT_PROTO_TIMEOUT_ICMPV6] = {
854 .procname = "nf_conntrack_icmpv6_timeout",
855 .maxlen = sizeof(unsigned int),
856 .mode = 0644,
857 .proc_handler = proc_dointvec_jiffies,
858 },
859#ifdef CONFIG_NF_CT_PROTO_SCTP
860 [NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_CLOSED] = {
861 .procname = "nf_conntrack_sctp_timeout_closed",
862 .maxlen = sizeof(unsigned int),
863 .mode = 0644,
864 .proc_handler = proc_dointvec_jiffies,
865 },
866 [NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_COOKIE_WAIT] = {
867 .procname = "nf_conntrack_sctp_timeout_cookie_wait",
868 .maxlen = sizeof(unsigned int),
869 .mode = 0644,
870 .proc_handler = proc_dointvec_jiffies,
871 },
872 [NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_COOKIE_ECHOED] = {
873 .procname = "nf_conntrack_sctp_timeout_cookie_echoed",
874 .maxlen = sizeof(unsigned int),
875 .mode = 0644,
876 .proc_handler = proc_dointvec_jiffies,
877 },
878 [NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_ESTABLISHED] = {
879 .procname = "nf_conntrack_sctp_timeout_established",
880 .maxlen = sizeof(unsigned int),
881 .mode = 0644,
882 .proc_handler = proc_dointvec_jiffies,
883 },
884 [NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_SHUTDOWN_SENT] = {
885 .procname = "nf_conntrack_sctp_timeout_shutdown_sent",
886 .maxlen = sizeof(unsigned int),
887 .mode = 0644,
888 .proc_handler = proc_dointvec_jiffies,
889 },
890 [NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_SHUTDOWN_RECD] = {
891 .procname = "nf_conntrack_sctp_timeout_shutdown_recd",
892 .maxlen = sizeof(unsigned int),
893 .mode = 0644,
894 .proc_handler = proc_dointvec_jiffies,
895 },
896 [NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_SHUTDOWN_ACK_SENT] = {
897 .procname = "nf_conntrack_sctp_timeout_shutdown_ack_sent",
898 .maxlen = sizeof(unsigned int),
899 .mode = 0644,
900 .proc_handler = proc_dointvec_jiffies,
901 },
902 [NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_HEARTBEAT_SENT] = {
903 .procname = "nf_conntrack_sctp_timeout_heartbeat_sent",
904 .maxlen = sizeof(unsigned int),
905 .mode = 0644,
906 .proc_handler = proc_dointvec_jiffies,
907 },
908#endif
909#ifdef CONFIG_NF_CT_PROTO_GRE
910 [NF_SYSCTL_CT_PROTO_TIMEOUT_GRE] = {
911 .procname = "nf_conntrack_gre_timeout",
912 .maxlen = sizeof(unsigned int),
913 .mode = 0644,
914 .proc_handler = proc_dointvec_jiffies,
915 },
916 [NF_SYSCTL_CT_PROTO_TIMEOUT_GRE_STREAM] = {
917 .procname = "nf_conntrack_gre_timeout_stream",
918 .maxlen = sizeof(unsigned int),
919 .mode = 0644,
920 .proc_handler = proc_dointvec_jiffies,
921 },
922#endif
923};
924
925static struct ctl_table nf_ct_netfilter_table[] = {
926 {
927 .procname = "nf_conntrack_max",
928 .data = &nf_conntrack_max,
929 .maxlen = sizeof(int),
930 .mode = 0644,
931 .proc_handler = proc_dointvec_minmax,
932 .extra1 = SYSCTL_ZERO,
933 .extra2 = SYSCTL_INT_MAX,
934 },
935};
936
937static void nf_conntrack_standalone_init_tcp_sysctl(struct net *net,
938 struct ctl_table *table)
939{
940 struct nf_tcp_net *tn = nf_tcp_pernet(net);
941
942#define XASSIGN(XNAME, tn) \
943 table[NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_ ## XNAME].data = \
944 &(tn)->timeouts[TCP_CONNTRACK_ ## XNAME]
945
946 XASSIGN(SYN_SENT, tn);
947 XASSIGN(SYN_RECV, tn);
948 XASSIGN(ESTABLISHED, tn);
949 XASSIGN(FIN_WAIT, tn);
950 XASSIGN(CLOSE_WAIT, tn);
951 XASSIGN(LAST_ACK, tn);
952 XASSIGN(TIME_WAIT, tn);
953 XASSIGN(CLOSE, tn);
954 XASSIGN(RETRANS, tn);
955 XASSIGN(UNACK, tn);
956#undef XASSIGN
957#define XASSIGN(XNAME, rval) \
958 table[NF_SYSCTL_CT_PROTO_TCP_ ## XNAME].data = (rval)
959
960 XASSIGN(LOOSE, &tn->tcp_loose);
961 XASSIGN(LIBERAL, &tn->tcp_be_liberal);
962 XASSIGN(MAX_RETRANS, &tn->tcp_max_retrans);
963 XASSIGN(IGNORE_INVALID_RST, &tn->tcp_ignore_invalid_rst);
964#undef XASSIGN
965
966#if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
967 table[NF_SYSCTL_CT_PROTO_TIMEOUT_TCP_OFFLOAD].data = &tn->offload_timeout;
968#endif
969
970}
971
972static void nf_conntrack_standalone_init_sctp_sysctl(struct net *net,
973 struct ctl_table *table)
974{
975#ifdef CONFIG_NF_CT_PROTO_SCTP
976 struct nf_sctp_net *sn = nf_sctp_pernet(net);
977
978#define XASSIGN(XNAME, sn) \
979 table[NF_SYSCTL_CT_PROTO_TIMEOUT_SCTP_ ## XNAME].data = \
980 &(sn)->timeouts[SCTP_CONNTRACK_ ## XNAME]
981
982 XASSIGN(CLOSED, sn);
983 XASSIGN(COOKIE_WAIT, sn);
984 XASSIGN(COOKIE_ECHOED, sn);
985 XASSIGN(ESTABLISHED, sn);
986 XASSIGN(SHUTDOWN_SENT, sn);
987 XASSIGN(SHUTDOWN_RECD, sn);
988 XASSIGN(SHUTDOWN_ACK_SENT, sn);
989 XASSIGN(HEARTBEAT_SENT, sn);
990#undef XASSIGN
991#endif
992}
993
994static void nf_conntrack_standalone_init_gre_sysctl(struct net *net,
995 struct ctl_table *table)
996{
997#ifdef CONFIG_NF_CT_PROTO_GRE
998 struct nf_gre_net *gn = nf_gre_pernet(net);
999
1000 table[NF_SYSCTL_CT_PROTO_TIMEOUT_GRE].data = &gn->timeouts[GRE_CT_UNREPLIED];
1001 table[NF_SYSCTL_CT_PROTO_TIMEOUT_GRE_STREAM].data = &gn->timeouts[GRE_CT_REPLIED];
1002#endif
1003}
1004
1005static int nf_conntrack_standalone_init_sysctl(struct net *net)
1006{
1007 struct nf_conntrack_net *cnet = nf_ct_pernet(net);
1008 struct nf_udp_net *un = nf_udp_pernet(net);
1009 struct ctl_table *table;
1010
1011 BUILD_BUG_ON(ARRAY_SIZE(nf_ct_sysctl_table) != NF_SYSCTL_CT_LAST_SYSCTL);
1012
1013 table = kmemdup(nf_ct_sysctl_table, sizeof(nf_ct_sysctl_table),
1014 GFP_KERNEL);
1015 if (!table)
1016 return -ENOMEM;
1017
1018 table[NF_SYSCTL_CT_COUNT].data = &cnet->count;
1019 table[NF_SYSCTL_CT_CHECKSUM].data = &net->ct.sysctl_checksum;
1020 table[NF_SYSCTL_CT_LOG_INVALID].data = &net->ct.sysctl_log_invalid;
1021 table[NF_SYSCTL_CT_ACCT].data = &net->ct.sysctl_acct;
1022#ifdef CONFIG_NF_CONNTRACK_EVENTS
1023 table[NF_SYSCTL_CT_EVENTS].data = &net->ct.sysctl_events;
1024#endif
1025#ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
1026 table[NF_SYSCTL_CT_TIMESTAMP].data = &net->ct.sysctl_tstamp;
1027#endif
1028 table[NF_SYSCTL_CT_PROTO_TIMEOUT_GENERIC].data = &nf_generic_pernet(net)->timeout;
1029 table[NF_SYSCTL_CT_PROTO_TIMEOUT_ICMP].data = &nf_icmp_pernet(net)->timeout;
1030 table[NF_SYSCTL_CT_PROTO_TIMEOUT_ICMPV6].data = &nf_icmpv6_pernet(net)->timeout;
1031 table[NF_SYSCTL_CT_PROTO_TIMEOUT_UDP].data = &un->timeouts[UDP_CT_UNREPLIED];
1032 table[NF_SYSCTL_CT_PROTO_TIMEOUT_UDP_STREAM].data = &un->timeouts[UDP_CT_REPLIED];
1033#if IS_ENABLED(CONFIG_NF_FLOW_TABLE)
1034 table[NF_SYSCTL_CT_PROTO_TIMEOUT_UDP_OFFLOAD].data = &un->offload_timeout;
1035#endif
1036
1037 nf_conntrack_standalone_init_tcp_sysctl(net, table);
1038 nf_conntrack_standalone_init_sctp_sysctl(net, table);
1039 nf_conntrack_standalone_init_gre_sysctl(net, table);
1040
1041 /* Don't allow non-init_net ns to alter global sysctls */
1042 if (!net_eq(net1: &init_net, net2: net)) {
1043 table[NF_SYSCTL_CT_MAX].mode = 0444;
1044 table[NF_SYSCTL_CT_EXPECT_MAX].mode = 0444;
1045 table[NF_SYSCTL_CT_BUCKETS].mode = 0444;
1046 }
1047
1048 cnet->sysctl_header = register_net_sysctl_sz(net, path: "net/netfilter",
1049 table,
1050 ARRAY_SIZE(nf_ct_sysctl_table));
1051 if (!cnet->sysctl_header)
1052 goto out_unregister_netfilter;
1053
1054 return 0;
1055
1056out_unregister_netfilter:
1057 kfree(objp: table);
1058 return -ENOMEM;
1059}
1060
1061static void nf_conntrack_standalone_fini_sysctl(struct net *net)
1062{
1063 struct nf_conntrack_net *cnet = nf_ct_pernet(net);
1064 const struct ctl_table *table;
1065
1066 table = cnet->sysctl_header->ctl_table_arg;
1067 unregister_net_sysctl_table(header: cnet->sysctl_header);
1068 kfree(objp: table);
1069}
1070#else
1071static int nf_conntrack_standalone_init_sysctl(struct net *net)
1072{
1073 return 0;
1074}
1075
1076static void nf_conntrack_standalone_fini_sysctl(struct net *net)
1077{
1078}
1079#endif /* CONFIG_SYSCTL */
1080
1081static void nf_conntrack_fini_net(struct net *net)
1082{
1083 if (enable_hooks)
1084 nf_ct_netns_put(net, nfproto: NFPROTO_INET);
1085
1086 nf_conntrack_standalone_fini_proc(net);
1087 nf_conntrack_standalone_fini_sysctl(net);
1088}
1089
1090static int nf_conntrack_pernet_init(struct net *net)
1091{
1092 int ret;
1093
1094 net->ct.sysctl_checksum = 1;
1095
1096 ret = nf_conntrack_standalone_init_sysctl(net);
1097 if (ret < 0)
1098 return ret;
1099
1100 ret = nf_conntrack_standalone_init_proc(net);
1101 if (ret < 0)
1102 goto out_proc;
1103
1104 ret = nf_conntrack_init_net(net);
1105 if (ret < 0)
1106 goto out_init_net;
1107
1108 if (enable_hooks) {
1109 ret = nf_ct_netns_get(net, nfproto: NFPROTO_INET);
1110 if (ret < 0)
1111 goto out_hooks;
1112 }
1113
1114 return 0;
1115
1116out_hooks:
1117 nf_conntrack_cleanup_net(net);
1118out_init_net:
1119 nf_conntrack_standalone_fini_proc(net);
1120out_proc:
1121 nf_conntrack_standalone_fini_sysctl(net);
1122 return ret;
1123}
1124
1125static void nf_conntrack_pernet_exit(struct list_head *net_exit_list)
1126{
1127 struct net *net;
1128
1129 list_for_each_entry(net, net_exit_list, exit_list)
1130 nf_conntrack_fini_net(net);
1131
1132 nf_conntrack_cleanup_net_list(net_exit_list);
1133}
1134
1135static struct pernet_operations nf_conntrack_net_ops = {
1136 .init = nf_conntrack_pernet_init,
1137 .exit_batch = nf_conntrack_pernet_exit,
1138 .id = &nf_conntrack_net_id,
1139 .size = sizeof(struct nf_conntrack_net),
1140};
1141
1142static int __init nf_conntrack_standalone_init(void)
1143{
1144 int ret = nf_conntrack_init_start();
1145 if (ret < 0)
1146 goto out_start;
1147
1148 BUILD_BUG_ON(NFCT_INFOMASK <= IP_CT_NUMBER);
1149
1150#ifdef CONFIG_SYSCTL
1151 nf_ct_netfilter_header =
1152 register_net_sysctl(&init_net, "net", nf_ct_netfilter_table);
1153 if (!nf_ct_netfilter_header) {
1154 pr_err("nf_conntrack: can't register to sysctl.\n");
1155 ret = -ENOMEM;
1156 goto out_sysctl;
1157 }
1158
1159 nf_conntrack_htable_size_user = nf_conntrack_htable_size;
1160#endif
1161
1162 nf_conntrack_init_end();
1163
1164 ret = register_pernet_subsys(&nf_conntrack_net_ops);
1165 if (ret < 0)
1166 goto out_pernet;
1167
1168 return 0;
1169
1170out_pernet:
1171#ifdef CONFIG_SYSCTL
1172 unregister_net_sysctl_table(header: nf_ct_netfilter_header);
1173out_sysctl:
1174#endif
1175 nf_conntrack_cleanup_end();
1176out_start:
1177 return ret;
1178}
1179
1180static void __exit nf_conntrack_standalone_fini(void)
1181{
1182 nf_conntrack_cleanup_start();
1183 unregister_pernet_subsys(&nf_conntrack_net_ops);
1184#ifdef CONFIG_SYSCTL
1185 unregister_net_sysctl_table(header: nf_ct_netfilter_header);
1186#endif
1187 nf_conntrack_cleanup_end();
1188}
1189
1190module_init(nf_conntrack_standalone_init);
1191module_exit(nf_conntrack_standalone_fini);
1192