1 /* Connection tracking via netlink socket. Allows for user space
2 * protocol helpers and general trouble making from userspace.
3 *
4 * (C) 2001 by Jay Schulist <jschlst@samba.org>
5 * (C) 2002-2006 by Harald Welte <laforge@gnumonks.org>
6 * (C) 2003 by Patrick Mchardy <kaber@trash.net>
7 * (C) 2005-2012 by Pablo Neira Ayuso <pablo@netfilter.org>
8 *
9 * Initial connection tracking via netlink development funded and
10 * generally made possible by Network Robots, Inc. (www.networkrobots.com)
11 *
12 * Further development of this code funded by Astaro AG (http://www.astaro.com)
13 *
14 * This software may be used and distributed according to the terms
15 * of the GNU General Public License, incorporated herein by reference.
16 */
17
18 #include <linux/init.h>
19 #include <linux/module.h>
20 #include <linux/kernel.h>
21 #include <linux/rculist.h>
22 #include <linux/rculist_nulls.h>
23 #include <linux/types.h>
24 #include <linux/timer.h>
25 #include <linux/security.h>
26 #include <linux/skbuff.h>
27 #include <linux/errno.h>
28 #include <linux/netlink.h>
29 #include <linux/spinlock.h>
30 #include <linux/interrupt.h>
31 #include <linux/slab.h>
32 #include <linux/siphash.h>
33
34 #include <linux/netfilter.h>
35 #include <net/netlink.h>
36 #include <net/sock.h>
37 #include <net/netfilter/nf_conntrack.h>
38 #include <net/netfilter/nf_conntrack_core.h>
39 #include <net/netfilter/nf_conntrack_expect.h>
40 #include <net/netfilter/nf_conntrack_helper.h>
41 #include <net/netfilter/nf_conntrack_seqadj.h>
42 #include <net/netfilter/nf_conntrack_l4proto.h>
43 #include <net/netfilter/nf_conntrack_tuple.h>
44 #include <net/netfilter/nf_conntrack_acct.h>
45 #include <net/netfilter/nf_conntrack_zones.h>
46 #include <net/netfilter/nf_conntrack_timestamp.h>
47 #include <net/netfilter/nf_conntrack_labels.h>
48 #include <net/netfilter/nf_conntrack_synproxy.h>
49 #if IS_ENABLED(CONFIG_NF_NAT)
50 #include <net/netfilter/nf_nat.h>
51 #include <net/netfilter/nf_nat_helper.h>
52 #endif
53
54 #include <linux/netfilter/nfnetlink.h>
55 #include <linux/netfilter/nfnetlink_conntrack.h>
56
57 #include "nf_internals.h"
58
59 MODULE_LICENSE("GPL");
60 MODULE_DESCRIPTION("List and change connection tracking table");
61
62 struct ctnetlink_list_dump_ctx {
63 struct nf_conn *last;
64 unsigned int cpu;
65 bool done;
66 };
67
ctnetlink_dump_tuples_proto(struct sk_buff * skb,const struct nf_conntrack_tuple * tuple,const struct nf_conntrack_l4proto * l4proto)68 static int ctnetlink_dump_tuples_proto(struct sk_buff *skb,
69 const struct nf_conntrack_tuple *tuple,
70 const struct nf_conntrack_l4proto *l4proto)
71 {
72 int ret = 0;
73 struct nlattr *nest_parms;
74
75 nest_parms = nla_nest_start(skb, CTA_TUPLE_PROTO);
76 if (!nest_parms)
77 goto nla_put_failure;
78 if (nla_put_u8(skb, CTA_PROTO_NUM, tuple->dst.protonum))
79 goto nla_put_failure;
80
81 if (likely(l4proto->tuple_to_nlattr))
82 ret = l4proto->tuple_to_nlattr(skb, tuple);
83
84 nla_nest_end(skb, nest_parms);
85
86 return ret;
87
88 nla_put_failure:
89 return -1;
90 }
91
ipv4_tuple_to_nlattr(struct sk_buff * skb,const struct nf_conntrack_tuple * tuple)92 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
93 const struct nf_conntrack_tuple *tuple)
94 {
95 if (nla_put_in_addr(skb, CTA_IP_V4_SRC, tuple->src.u3.ip) ||
96 nla_put_in_addr(skb, CTA_IP_V4_DST, tuple->dst.u3.ip))
97 return -EMSGSIZE;
98 return 0;
99 }
100
ipv6_tuple_to_nlattr(struct sk_buff * skb,const struct nf_conntrack_tuple * tuple)101 static int ipv6_tuple_to_nlattr(struct sk_buff *skb,
102 const struct nf_conntrack_tuple *tuple)
103 {
104 if (nla_put_in6_addr(skb, CTA_IP_V6_SRC, &tuple->src.u3.in6) ||
105 nla_put_in6_addr(skb, CTA_IP_V6_DST, &tuple->dst.u3.in6))
106 return -EMSGSIZE;
107 return 0;
108 }
109
ctnetlink_dump_tuples_ip(struct sk_buff * skb,const struct nf_conntrack_tuple * tuple)110 static int ctnetlink_dump_tuples_ip(struct sk_buff *skb,
111 const struct nf_conntrack_tuple *tuple)
112 {
113 int ret = 0;
114 struct nlattr *nest_parms;
115
116 nest_parms = nla_nest_start(skb, CTA_TUPLE_IP);
117 if (!nest_parms)
118 goto nla_put_failure;
119
120 switch (tuple->src.l3num) {
121 case NFPROTO_IPV4:
122 ret = ipv4_tuple_to_nlattr(skb, tuple);
123 break;
124 case NFPROTO_IPV6:
125 ret = ipv6_tuple_to_nlattr(skb, tuple);
126 break;
127 }
128
129 nla_nest_end(skb, nest_parms);
130
131 return ret;
132
133 nla_put_failure:
134 return -1;
135 }
136
ctnetlink_dump_tuples(struct sk_buff * skb,const struct nf_conntrack_tuple * tuple)137 static int ctnetlink_dump_tuples(struct sk_buff *skb,
138 const struct nf_conntrack_tuple *tuple)
139 {
140 const struct nf_conntrack_l4proto *l4proto;
141 int ret;
142
143 rcu_read_lock();
144 ret = ctnetlink_dump_tuples_ip(skb, tuple);
145
146 if (ret >= 0) {
147 l4proto = nf_ct_l4proto_find(tuple->dst.protonum);
148 ret = ctnetlink_dump_tuples_proto(skb, tuple, l4proto);
149 }
150 rcu_read_unlock();
151 return ret;
152 }
153
ctnetlink_dump_zone_id(struct sk_buff * skb,int attrtype,const struct nf_conntrack_zone * zone,int dir)154 static int ctnetlink_dump_zone_id(struct sk_buff *skb, int attrtype,
155 const struct nf_conntrack_zone *zone, int dir)
156 {
157 if (zone->id == NF_CT_DEFAULT_ZONE_ID || zone->dir != dir)
158 return 0;
159 if (nla_put_be16(skb, attrtype, htons(zone->id)))
160 goto nla_put_failure;
161 return 0;
162
163 nla_put_failure:
164 return -1;
165 }
166
ctnetlink_dump_status(struct sk_buff * skb,const struct nf_conn * ct)167 static int ctnetlink_dump_status(struct sk_buff *skb, const struct nf_conn *ct)
168 {
169 if (nla_put_be32(skb, CTA_STATUS, htonl(ct->status)))
170 goto nla_put_failure;
171 return 0;
172
173 nla_put_failure:
174 return -1;
175 }
176
ctnetlink_dump_timeout(struct sk_buff * skb,const struct nf_conn * ct,bool skip_zero)177 static int ctnetlink_dump_timeout(struct sk_buff *skb, const struct nf_conn *ct,
178 bool skip_zero)
179 {
180 long timeout;
181
182 if (nf_ct_is_confirmed(ct))
183 timeout = nf_ct_expires(ct) / HZ;
184 else
185 timeout = ct->timeout / HZ;
186
187 if (skip_zero && timeout == 0)
188 return 0;
189
190 if (nla_put_be32(skb, CTA_TIMEOUT, htonl(timeout)))
191 goto nla_put_failure;
192 return 0;
193
194 nla_put_failure:
195 return -1;
196 }
197
ctnetlink_dump_protoinfo(struct sk_buff * skb,struct nf_conn * ct,bool destroy)198 static int ctnetlink_dump_protoinfo(struct sk_buff *skb, struct nf_conn *ct,
199 bool destroy)
200 {
201 const struct nf_conntrack_l4proto *l4proto;
202 struct nlattr *nest_proto;
203 int ret;
204
205 l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct));
206 if (!l4proto->to_nlattr)
207 return 0;
208
209 nest_proto = nla_nest_start(skb, CTA_PROTOINFO);
210 if (!nest_proto)
211 goto nla_put_failure;
212
213 ret = l4proto->to_nlattr(skb, nest_proto, ct, destroy);
214
215 nla_nest_end(skb, nest_proto);
216
217 return ret;
218
219 nla_put_failure:
220 return -1;
221 }
222
ctnetlink_dump_helpinfo(struct sk_buff * skb,const struct nf_conn * ct)223 static int ctnetlink_dump_helpinfo(struct sk_buff *skb,
224 const struct nf_conn *ct)
225 {
226 struct nlattr *nest_helper;
227 const struct nf_conn_help *help = nfct_help(ct);
228 struct nf_conntrack_helper *helper;
229
230 if (!help)
231 return 0;
232
233 rcu_read_lock();
234 helper = rcu_dereference(help->helper);
235 if (!helper)
236 goto out;
237
238 nest_helper = nla_nest_start(skb, CTA_HELP);
239 if (!nest_helper)
240 goto nla_put_failure;
241 if (nla_put_string(skb, CTA_HELP_NAME, helper->name))
242 goto nla_put_failure;
243
244 if (helper->to_nlattr)
245 helper->to_nlattr(skb, ct);
246
247 nla_nest_end(skb, nest_helper);
248 out:
249 rcu_read_unlock();
250 return 0;
251
252 nla_put_failure:
253 rcu_read_unlock();
254 return -1;
255 }
256
257 static int
dump_counters(struct sk_buff * skb,struct nf_conn_acct * acct,enum ip_conntrack_dir dir,int type)258 dump_counters(struct sk_buff *skb, struct nf_conn_acct *acct,
259 enum ip_conntrack_dir dir, int type)
260 {
261 enum ctattr_type attr = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
262 struct nf_conn_counter *counter = acct->counter;
263 struct nlattr *nest_count;
264 u64 pkts, bytes;
265
266 if (type == IPCTNL_MSG_CT_GET_CTRZERO) {
267 pkts = atomic64_xchg(&counter[dir].packets, 0);
268 bytes = atomic64_xchg(&counter[dir].bytes, 0);
269 } else {
270 pkts = atomic64_read(&counter[dir].packets);
271 bytes = atomic64_read(&counter[dir].bytes);
272 }
273
274 nest_count = nla_nest_start(skb, attr);
275 if (!nest_count)
276 goto nla_put_failure;
277
278 if (nla_put_be64(skb, CTA_COUNTERS_PACKETS, cpu_to_be64(pkts),
279 CTA_COUNTERS_PAD) ||
280 nla_put_be64(skb, CTA_COUNTERS_BYTES, cpu_to_be64(bytes),
281 CTA_COUNTERS_PAD))
282 goto nla_put_failure;
283
284 nla_nest_end(skb, nest_count);
285
286 return 0;
287
288 nla_put_failure:
289 return -1;
290 }
291
292 static int
ctnetlink_dump_acct(struct sk_buff * skb,const struct nf_conn * ct,int type)293 ctnetlink_dump_acct(struct sk_buff *skb, const struct nf_conn *ct, int type)
294 {
295 struct nf_conn_acct *acct = nf_conn_acct_find(ct);
296
297 if (!acct)
298 return 0;
299
300 if (dump_counters(skb, acct, IP_CT_DIR_ORIGINAL, type) < 0)
301 return -1;
302 if (dump_counters(skb, acct, IP_CT_DIR_REPLY, type) < 0)
303 return -1;
304
305 return 0;
306 }
307
308 static int
ctnetlink_dump_timestamp(struct sk_buff * skb,const struct nf_conn * ct)309 ctnetlink_dump_timestamp(struct sk_buff *skb, const struct nf_conn *ct)
310 {
311 struct nlattr *nest_count;
312 const struct nf_conn_tstamp *tstamp;
313
314 tstamp = nf_conn_tstamp_find(ct);
315 if (!tstamp)
316 return 0;
317
318 nest_count = nla_nest_start(skb, CTA_TIMESTAMP);
319 if (!nest_count)
320 goto nla_put_failure;
321
322 if (nla_put_be64(skb, CTA_TIMESTAMP_START, cpu_to_be64(tstamp->start),
323 CTA_TIMESTAMP_PAD) ||
324 (tstamp->stop != 0 && nla_put_be64(skb, CTA_TIMESTAMP_STOP,
325 cpu_to_be64(tstamp->stop),
326 CTA_TIMESTAMP_PAD)))
327 goto nla_put_failure;
328 nla_nest_end(skb, nest_count);
329
330 return 0;
331
332 nla_put_failure:
333 return -1;
334 }
335
336 #ifdef CONFIG_NF_CONNTRACK_MARK
ctnetlink_dump_mark(struct sk_buff * skb,const struct nf_conn * ct,bool dump)337 static int ctnetlink_dump_mark(struct sk_buff *skb, const struct nf_conn *ct,
338 bool dump)
339 {
340 u32 mark = READ_ONCE(ct->mark);
341
342 if (!mark && !dump)
343 return 0;
344
345 if (nla_put_be32(skb, CTA_MARK, htonl(mark)))
346 goto nla_put_failure;
347 return 0;
348
349 nla_put_failure:
350 return -1;
351 }
352 #else
353 #define ctnetlink_dump_mark(a, b, c) (0)
354 #endif
355
356 #ifdef CONFIG_NF_CONNTRACK_SECMARK
ctnetlink_dump_secctx(struct sk_buff * skb,const struct nf_conn * ct)357 static int ctnetlink_dump_secctx(struct sk_buff *skb, const struct nf_conn *ct)
358 {
359 struct nlattr *nest_secctx;
360 int len, ret;
361 char *secctx;
362
363 ret = security_secid_to_secctx(ct->secmark, &secctx, &len);
364 if (ret)
365 return 0;
366
367 ret = -1;
368 nest_secctx = nla_nest_start(skb, CTA_SECCTX);
369 if (!nest_secctx)
370 goto nla_put_failure;
371
372 if (nla_put_string(skb, CTA_SECCTX_NAME, secctx))
373 goto nla_put_failure;
374 nla_nest_end(skb, nest_secctx);
375
376 ret = 0;
377 nla_put_failure:
378 security_release_secctx(secctx, len);
379 return ret;
380 }
381 #else
382 #define ctnetlink_dump_secctx(a, b) (0)
383 #endif
384
385 #ifdef CONFIG_NF_CONNTRACK_EVENTS
ctnetlink_label_size(const struct nf_conn * ct)386 static inline int ctnetlink_label_size(const struct nf_conn *ct)
387 {
388 struct nf_conn_labels *labels = nf_ct_labels_find(ct);
389
390 if (!labels)
391 return 0;
392 return nla_total_size(sizeof(labels->bits));
393 }
394 #endif
395
396 static int
ctnetlink_dump_labels(struct sk_buff * skb,const struct nf_conn * ct)397 ctnetlink_dump_labels(struct sk_buff *skb, const struct nf_conn *ct)
398 {
399 struct nf_conn_labels *labels = nf_ct_labels_find(ct);
400 unsigned int i;
401
402 if (!labels)
403 return 0;
404
405 i = 0;
406 do {
407 if (labels->bits[i] != 0)
408 return nla_put(skb, CTA_LABELS, sizeof(labels->bits),
409 labels->bits);
410 i++;
411 } while (i < ARRAY_SIZE(labels->bits));
412
413 return 0;
414 }
415
416 #define master_tuple(ct) &(ct->master->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
417
ctnetlink_dump_master(struct sk_buff * skb,const struct nf_conn * ct)418 static int ctnetlink_dump_master(struct sk_buff *skb, const struct nf_conn *ct)
419 {
420 struct nlattr *nest_parms;
421
422 if (!(ct->status & IPS_EXPECTED))
423 return 0;
424
425 nest_parms = nla_nest_start(skb, CTA_TUPLE_MASTER);
426 if (!nest_parms)
427 goto nla_put_failure;
428 if (ctnetlink_dump_tuples(skb, master_tuple(ct)) < 0)
429 goto nla_put_failure;
430 nla_nest_end(skb, nest_parms);
431
432 return 0;
433
434 nla_put_failure:
435 return -1;
436 }
437
438 static int
dump_ct_seq_adj(struct sk_buff * skb,const struct nf_ct_seqadj * seq,int type)439 dump_ct_seq_adj(struct sk_buff *skb, const struct nf_ct_seqadj *seq, int type)
440 {
441 struct nlattr *nest_parms;
442
443 nest_parms = nla_nest_start(skb, type);
444 if (!nest_parms)
445 goto nla_put_failure;
446
447 if (nla_put_be32(skb, CTA_SEQADJ_CORRECTION_POS,
448 htonl(seq->correction_pos)) ||
449 nla_put_be32(skb, CTA_SEQADJ_OFFSET_BEFORE,
450 htonl(seq->offset_before)) ||
451 nla_put_be32(skb, CTA_SEQADJ_OFFSET_AFTER,
452 htonl(seq->offset_after)))
453 goto nla_put_failure;
454
455 nla_nest_end(skb, nest_parms);
456
457 return 0;
458
459 nla_put_failure:
460 return -1;
461 }
462
ctnetlink_dump_ct_seq_adj(struct sk_buff * skb,struct nf_conn * ct)463 static int ctnetlink_dump_ct_seq_adj(struct sk_buff *skb, struct nf_conn *ct)
464 {
465 struct nf_conn_seqadj *seqadj = nfct_seqadj(ct);
466 struct nf_ct_seqadj *seq;
467
468 if (!(ct->status & IPS_SEQ_ADJUST) || !seqadj)
469 return 0;
470
471 spin_lock_bh(&ct->lock);
472 seq = &seqadj->seq[IP_CT_DIR_ORIGINAL];
473 if (dump_ct_seq_adj(skb, seq, CTA_SEQ_ADJ_ORIG) == -1)
474 goto err;
475
476 seq = &seqadj->seq[IP_CT_DIR_REPLY];
477 if (dump_ct_seq_adj(skb, seq, CTA_SEQ_ADJ_REPLY) == -1)
478 goto err;
479
480 spin_unlock_bh(&ct->lock);
481 return 0;
482 err:
483 spin_unlock_bh(&ct->lock);
484 return -1;
485 }
486
ctnetlink_dump_ct_synproxy(struct sk_buff * skb,struct nf_conn * ct)487 static int ctnetlink_dump_ct_synproxy(struct sk_buff *skb, struct nf_conn *ct)
488 {
489 struct nf_conn_synproxy *synproxy = nfct_synproxy(ct);
490 struct nlattr *nest_parms;
491
492 if (!synproxy)
493 return 0;
494
495 nest_parms = nla_nest_start(skb, CTA_SYNPROXY);
496 if (!nest_parms)
497 goto nla_put_failure;
498
499 if (nla_put_be32(skb, CTA_SYNPROXY_ISN, htonl(synproxy->isn)) ||
500 nla_put_be32(skb, CTA_SYNPROXY_ITS, htonl(synproxy->its)) ||
501 nla_put_be32(skb, CTA_SYNPROXY_TSOFF, htonl(synproxy->tsoff)))
502 goto nla_put_failure;
503
504 nla_nest_end(skb, nest_parms);
505
506 return 0;
507
508 nla_put_failure:
509 return -1;
510 }
511
ctnetlink_dump_id(struct sk_buff * skb,const struct nf_conn * ct)512 static int ctnetlink_dump_id(struct sk_buff *skb, const struct nf_conn *ct)
513 {
514 __be32 id = (__force __be32)nf_ct_get_id(ct);
515
516 if (nla_put_be32(skb, CTA_ID, id))
517 goto nla_put_failure;
518 return 0;
519
520 nla_put_failure:
521 return -1;
522 }
523
ctnetlink_dump_use(struct sk_buff * skb,const struct nf_conn * ct)524 static int ctnetlink_dump_use(struct sk_buff *skb, const struct nf_conn *ct)
525 {
526 if (nla_put_be32(skb, CTA_USE, htonl(refcount_read(&ct->ct_general.use))))
527 goto nla_put_failure;
528 return 0;
529
530 nla_put_failure:
531 return -1;
532 }
533
534 /* all these functions access ct->ext. Caller must either hold a reference
535 * on ct or prevent its deletion by holding either the bucket spinlock or
536 * pcpu dying list lock.
537 */
ctnetlink_dump_extinfo(struct sk_buff * skb,struct nf_conn * ct,u32 type)538 static int ctnetlink_dump_extinfo(struct sk_buff *skb,
539 struct nf_conn *ct, u32 type)
540 {
541 if (ctnetlink_dump_acct(skb, ct, type) < 0 ||
542 ctnetlink_dump_timestamp(skb, ct) < 0 ||
543 ctnetlink_dump_helpinfo(skb, ct) < 0 ||
544 ctnetlink_dump_labels(skb, ct) < 0 ||
545 ctnetlink_dump_ct_seq_adj(skb, ct) < 0 ||
546 ctnetlink_dump_ct_synproxy(skb, ct) < 0)
547 return -1;
548
549 return 0;
550 }
551
ctnetlink_dump_info(struct sk_buff * skb,struct nf_conn * ct)552 static int ctnetlink_dump_info(struct sk_buff *skb, struct nf_conn *ct)
553 {
554 if (ctnetlink_dump_status(skb, ct) < 0 ||
555 ctnetlink_dump_mark(skb, ct, true) < 0 ||
556 ctnetlink_dump_secctx(skb, ct) < 0 ||
557 ctnetlink_dump_id(skb, ct) < 0 ||
558 ctnetlink_dump_use(skb, ct) < 0 ||
559 ctnetlink_dump_master(skb, ct) < 0)
560 return -1;
561
562 if (!test_bit(IPS_OFFLOAD_BIT, &ct->status) &&
563 (ctnetlink_dump_timeout(skb, ct, false) < 0 ||
564 ctnetlink_dump_protoinfo(skb, ct, false) < 0))
565 return -1;
566
567 return 0;
568 }
569
570 static int
ctnetlink_fill_info(struct sk_buff * skb,u32 portid,u32 seq,u32 type,struct nf_conn * ct,bool extinfo,unsigned int flags)571 ctnetlink_fill_info(struct sk_buff *skb, u32 portid, u32 seq, u32 type,
572 struct nf_conn *ct, bool extinfo, unsigned int flags)
573 {
574 const struct nf_conntrack_zone *zone;
575 struct nlmsghdr *nlh;
576 struct nlattr *nest_parms;
577 unsigned int event;
578
579 if (portid)
580 flags |= NLM_F_MULTI;
581 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, IPCTNL_MSG_CT_NEW);
582 nlh = nfnl_msg_put(skb, portid, seq, event, flags, nf_ct_l3num(ct),
583 NFNETLINK_V0, 0);
584 if (!nlh)
585 goto nlmsg_failure;
586
587 zone = nf_ct_zone(ct);
588
589 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG);
590 if (!nest_parms)
591 goto nla_put_failure;
592 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
593 goto nla_put_failure;
594 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
595 NF_CT_ZONE_DIR_ORIG) < 0)
596 goto nla_put_failure;
597 nla_nest_end(skb, nest_parms);
598
599 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY);
600 if (!nest_parms)
601 goto nla_put_failure;
602 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0)
603 goto nla_put_failure;
604 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
605 NF_CT_ZONE_DIR_REPL) < 0)
606 goto nla_put_failure;
607 nla_nest_end(skb, nest_parms);
608
609 if (ctnetlink_dump_zone_id(skb, CTA_ZONE, zone,
610 NF_CT_DEFAULT_ZONE_DIR) < 0)
611 goto nla_put_failure;
612
613 if (ctnetlink_dump_info(skb, ct) < 0)
614 goto nla_put_failure;
615 if (extinfo && ctnetlink_dump_extinfo(skb, ct, type) < 0)
616 goto nla_put_failure;
617
618 nlmsg_end(skb, nlh);
619 return skb->len;
620
621 nlmsg_failure:
622 nla_put_failure:
623 nlmsg_cancel(skb, nlh);
624 return -1;
625 }
626
627 static const struct nla_policy cta_ip_nla_policy[CTA_IP_MAX + 1] = {
628 [CTA_IP_V4_SRC] = { .type = NLA_U32 },
629 [CTA_IP_V4_DST] = { .type = NLA_U32 },
630 [CTA_IP_V6_SRC] = { .len = sizeof(__be32) * 4 },
631 [CTA_IP_V6_DST] = { .len = sizeof(__be32) * 4 },
632 };
633
634 #if defined(CONFIG_NETFILTER_NETLINK_GLUE_CT) || defined(CONFIG_NF_CONNTRACK_EVENTS)
ctnetlink_proto_size(const struct nf_conn * ct)635 static size_t ctnetlink_proto_size(const struct nf_conn *ct)
636 {
637 const struct nf_conntrack_l4proto *l4proto;
638 size_t len, len4 = 0;
639
640 len = nla_policy_len(cta_ip_nla_policy, CTA_IP_MAX + 1);
641 len *= 3u; /* ORIG, REPLY, MASTER */
642
643 l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct));
644 len += l4proto->nlattr_size;
645 if (l4proto->nlattr_tuple_size) {
646 len4 = l4proto->nlattr_tuple_size();
647 len4 *= 3u; /* ORIG, REPLY, MASTER */
648 }
649
650 return len + len4;
651 }
652
ctnetlink_acct_size(const struct nf_conn * ct)653 static inline size_t ctnetlink_acct_size(const struct nf_conn *ct)
654 {
655 if (!nf_ct_ext_exist(ct, NF_CT_EXT_ACCT))
656 return 0;
657 return 2 * nla_total_size(0) /* CTA_COUNTERS_ORIG|REPL */
658 + 2 * nla_total_size_64bit(sizeof(uint64_t)) /* CTA_COUNTERS_PACKETS */
659 + 2 * nla_total_size_64bit(sizeof(uint64_t)) /* CTA_COUNTERS_BYTES */
660 ;
661 }
662
ctnetlink_secctx_size(const struct nf_conn * ct)663 static inline int ctnetlink_secctx_size(const struct nf_conn *ct)
664 {
665 #ifdef CONFIG_NF_CONNTRACK_SECMARK
666 int len, ret;
667
668 ret = security_secid_to_secctx(ct->secmark, NULL, &len);
669 if (ret)
670 return 0;
671
672 return nla_total_size(0) /* CTA_SECCTX */
673 + nla_total_size(sizeof(char) * len); /* CTA_SECCTX_NAME */
674 #else
675 return 0;
676 #endif
677 }
678
ctnetlink_timestamp_size(const struct nf_conn * ct)679 static inline size_t ctnetlink_timestamp_size(const struct nf_conn *ct)
680 {
681 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
682 if (!nf_ct_ext_exist(ct, NF_CT_EXT_TSTAMP))
683 return 0;
684 return nla_total_size(0) + 2 * nla_total_size_64bit(sizeof(uint64_t));
685 #else
686 return 0;
687 #endif
688 }
689 #endif
690
691 #ifdef CONFIG_NF_CONNTRACK_EVENTS
ctnetlink_nlmsg_size(const struct nf_conn * ct)692 static size_t ctnetlink_nlmsg_size(const struct nf_conn *ct)
693 {
694 return NLMSG_ALIGN(sizeof(struct nfgenmsg))
695 + 3 * nla_total_size(0) /* CTA_TUPLE_ORIG|REPL|MASTER */
696 + 3 * nla_total_size(0) /* CTA_TUPLE_IP */
697 + 3 * nla_total_size(0) /* CTA_TUPLE_PROTO */
698 + 3 * nla_total_size(sizeof(u_int8_t)) /* CTA_PROTO_NUM */
699 + nla_total_size(sizeof(u_int32_t)) /* CTA_ID */
700 + nla_total_size(sizeof(u_int32_t)) /* CTA_STATUS */
701 + ctnetlink_acct_size(ct)
702 + ctnetlink_timestamp_size(ct)
703 + nla_total_size(sizeof(u_int32_t)) /* CTA_TIMEOUT */
704 + nla_total_size(0) /* CTA_PROTOINFO */
705 + nla_total_size(0) /* CTA_HELP */
706 + nla_total_size(NF_CT_HELPER_NAME_LEN) /* CTA_HELP_NAME */
707 + ctnetlink_secctx_size(ct)
708 #if IS_ENABLED(CONFIG_NF_NAT)
709 + 2 * nla_total_size(0) /* CTA_NAT_SEQ_ADJ_ORIG|REPL */
710 + 6 * nla_total_size(sizeof(u_int32_t)) /* CTA_NAT_SEQ_OFFSET */
711 #endif
712 #ifdef CONFIG_NF_CONNTRACK_MARK
713 + nla_total_size(sizeof(u_int32_t)) /* CTA_MARK */
714 #endif
715 #ifdef CONFIG_NF_CONNTRACK_ZONES
716 + nla_total_size(sizeof(u_int16_t)) /* CTA_ZONE|CTA_TUPLE_ZONE */
717 #endif
718 + ctnetlink_proto_size(ct)
719 + ctnetlink_label_size(ct)
720 ;
721 }
722
723 static int
ctnetlink_conntrack_event(unsigned int events,const struct nf_ct_event * item)724 ctnetlink_conntrack_event(unsigned int events, const struct nf_ct_event *item)
725 {
726 const struct nf_conntrack_zone *zone;
727 struct net *net;
728 struct nlmsghdr *nlh;
729 struct nlattr *nest_parms;
730 struct nf_conn *ct = item->ct;
731 struct sk_buff *skb;
732 unsigned int type;
733 unsigned int flags = 0, group;
734 int err;
735
736 if (events & (1 << IPCT_DESTROY)) {
737 type = IPCTNL_MSG_CT_DELETE;
738 group = NFNLGRP_CONNTRACK_DESTROY;
739 } else if (events & ((1 << IPCT_NEW) | (1 << IPCT_RELATED))) {
740 type = IPCTNL_MSG_CT_NEW;
741 flags = NLM_F_CREATE|NLM_F_EXCL;
742 group = NFNLGRP_CONNTRACK_NEW;
743 } else if (events) {
744 type = IPCTNL_MSG_CT_NEW;
745 group = NFNLGRP_CONNTRACK_UPDATE;
746 } else
747 return 0;
748
749 net = nf_ct_net(ct);
750 if (!item->report && !nfnetlink_has_listeners(net, group))
751 return 0;
752
753 skb = nlmsg_new(ctnetlink_nlmsg_size(ct), GFP_ATOMIC);
754 if (skb == NULL)
755 goto errout;
756
757 type = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, type);
758 nlh = nfnl_msg_put(skb, item->portid, 0, type, flags, nf_ct_l3num(ct),
759 NFNETLINK_V0, 0);
760 if (!nlh)
761 goto nlmsg_failure;
762
763 zone = nf_ct_zone(ct);
764
765 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG);
766 if (!nest_parms)
767 goto nla_put_failure;
768 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
769 goto nla_put_failure;
770 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
771 NF_CT_ZONE_DIR_ORIG) < 0)
772 goto nla_put_failure;
773 nla_nest_end(skb, nest_parms);
774
775 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY);
776 if (!nest_parms)
777 goto nla_put_failure;
778 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0)
779 goto nla_put_failure;
780 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
781 NF_CT_ZONE_DIR_REPL) < 0)
782 goto nla_put_failure;
783 nla_nest_end(skb, nest_parms);
784
785 if (ctnetlink_dump_zone_id(skb, CTA_ZONE, zone,
786 NF_CT_DEFAULT_ZONE_DIR) < 0)
787 goto nla_put_failure;
788
789 if (ctnetlink_dump_id(skb, ct) < 0)
790 goto nla_put_failure;
791
792 if (ctnetlink_dump_status(skb, ct) < 0)
793 goto nla_put_failure;
794
795 if (events & (1 << IPCT_DESTROY)) {
796 if (ctnetlink_dump_timeout(skb, ct, true) < 0)
797 goto nla_put_failure;
798
799 if (ctnetlink_dump_acct(skb, ct, type) < 0 ||
800 ctnetlink_dump_timestamp(skb, ct) < 0 ||
801 ctnetlink_dump_protoinfo(skb, ct, true) < 0)
802 goto nla_put_failure;
803 } else {
804 if (ctnetlink_dump_timeout(skb, ct, false) < 0)
805 goto nla_put_failure;
806
807 if (events & (1 << IPCT_PROTOINFO) &&
808 ctnetlink_dump_protoinfo(skb, ct, false) < 0)
809 goto nla_put_failure;
810
811 if ((events & (1 << IPCT_HELPER) || nfct_help(ct))
812 && ctnetlink_dump_helpinfo(skb, ct) < 0)
813 goto nla_put_failure;
814
815 #ifdef CONFIG_NF_CONNTRACK_SECMARK
816 if ((events & (1 << IPCT_SECMARK) || ct->secmark)
817 && ctnetlink_dump_secctx(skb, ct) < 0)
818 goto nla_put_failure;
819 #endif
820 if (events & (1 << IPCT_LABEL) &&
821 ctnetlink_dump_labels(skb, ct) < 0)
822 goto nla_put_failure;
823
824 if (events & (1 << IPCT_RELATED) &&
825 ctnetlink_dump_master(skb, ct) < 0)
826 goto nla_put_failure;
827
828 if (events & (1 << IPCT_SEQADJ) &&
829 ctnetlink_dump_ct_seq_adj(skb, ct) < 0)
830 goto nla_put_failure;
831
832 if (events & (1 << IPCT_SYNPROXY) &&
833 ctnetlink_dump_ct_synproxy(skb, ct) < 0)
834 goto nla_put_failure;
835 }
836
837 #ifdef CONFIG_NF_CONNTRACK_MARK
838 if (ctnetlink_dump_mark(skb, ct, events & (1 << IPCT_MARK)))
839 goto nla_put_failure;
840 #endif
841 nlmsg_end(skb, nlh);
842 err = nfnetlink_send(skb, net, item->portid, group, item->report,
843 GFP_ATOMIC);
844 if (err == -ENOBUFS || err == -EAGAIN)
845 return -ENOBUFS;
846
847 return 0;
848
849 nla_put_failure:
850 nlmsg_cancel(skb, nlh);
851 nlmsg_failure:
852 kfree_skb(skb);
853 errout:
854 if (nfnetlink_set_err(net, 0, group, -ENOBUFS) > 0)
855 return -ENOBUFS;
856
857 return 0;
858 }
859 #endif /* CONFIG_NF_CONNTRACK_EVENTS */
860
ctnetlink_done(struct netlink_callback * cb)861 static int ctnetlink_done(struct netlink_callback *cb)
862 {
863 kfree(cb->data);
864 return 0;
865 }
866
867 struct ctnetlink_filter_u32 {
868 u32 val;
869 u32 mask;
870 };
871
872 struct ctnetlink_filter {
873 u8 family;
874 bool zone_filter;
875
876 u_int32_t orig_flags;
877 u_int32_t reply_flags;
878
879 struct nf_conntrack_tuple orig;
880 struct nf_conntrack_tuple reply;
881 struct nf_conntrack_zone zone;
882
883 struct ctnetlink_filter_u32 mark;
884 struct ctnetlink_filter_u32 status;
885 };
886
887 static const struct nla_policy cta_filter_nla_policy[CTA_FILTER_MAX + 1] = {
888 [CTA_FILTER_ORIG_FLAGS] = { .type = NLA_U32 },
889 [CTA_FILTER_REPLY_FLAGS] = { .type = NLA_U32 },
890 };
891
ctnetlink_parse_filter(const struct nlattr * attr,struct ctnetlink_filter * filter)892 static int ctnetlink_parse_filter(const struct nlattr *attr,
893 struct ctnetlink_filter *filter)
894 {
895 struct nlattr *tb[CTA_FILTER_MAX + 1];
896 int ret = 0;
897
898 ret = nla_parse_nested(tb, CTA_FILTER_MAX, attr, cta_filter_nla_policy,
899 NULL);
900 if (ret)
901 return ret;
902
903 if (tb[CTA_FILTER_ORIG_FLAGS]) {
904 filter->orig_flags = nla_get_u32(tb[CTA_FILTER_ORIG_FLAGS]);
905 if (filter->orig_flags & ~CTA_FILTER_F_ALL)
906 return -EOPNOTSUPP;
907 }
908
909 if (tb[CTA_FILTER_REPLY_FLAGS]) {
910 filter->reply_flags = nla_get_u32(tb[CTA_FILTER_REPLY_FLAGS]);
911 if (filter->reply_flags & ~CTA_FILTER_F_ALL)
912 return -EOPNOTSUPP;
913 }
914
915 return 0;
916 }
917
918 static int ctnetlink_parse_zone(const struct nlattr *attr,
919 struct nf_conntrack_zone *zone);
920 static int ctnetlink_parse_tuple_filter(const struct nlattr * const cda[],
921 struct nf_conntrack_tuple *tuple,
922 u32 type, u_int8_t l3num,
923 struct nf_conntrack_zone *zone,
924 u_int32_t flags);
925
ctnetlink_filter_parse_mark(struct ctnetlink_filter_u32 * mark,const struct nlattr * const cda[])926 static int ctnetlink_filter_parse_mark(struct ctnetlink_filter_u32 *mark,
927 const struct nlattr * const cda[])
928 {
929 #ifdef CONFIG_NF_CONNTRACK_MARK
930 if (cda[CTA_MARK]) {
931 mark->val = ntohl(nla_get_be32(cda[CTA_MARK]));
932
933 if (cda[CTA_MARK_MASK])
934 mark->mask = ntohl(nla_get_be32(cda[CTA_MARK_MASK]));
935 else
936 mark->mask = 0xffffffff;
937 } else if (cda[CTA_MARK_MASK]) {
938 return -EINVAL;
939 }
940 #endif
941 return 0;
942 }
943
ctnetlink_filter_parse_status(struct ctnetlink_filter_u32 * status,const struct nlattr * const cda[])944 static int ctnetlink_filter_parse_status(struct ctnetlink_filter_u32 *status,
945 const struct nlattr * const cda[])
946 {
947 if (cda[CTA_STATUS]) {
948 status->val = ntohl(nla_get_be32(cda[CTA_STATUS]));
949 if (cda[CTA_STATUS_MASK])
950 status->mask = ntohl(nla_get_be32(cda[CTA_STATUS_MASK]));
951 else
952 status->mask = status->val;
953
954 /* status->val == 0? always true, else always false. */
955 if (status->mask == 0)
956 return -EINVAL;
957 } else if (cda[CTA_STATUS_MASK]) {
958 return -EINVAL;
959 }
960
961 /* CTA_STATUS is NLA_U32, if this fires UAPI needs to be extended */
962 BUILD_BUG_ON(__IPS_MAX_BIT >= 32);
963 return 0;
964 }
965
966 static struct ctnetlink_filter *
ctnetlink_alloc_filter(const struct nlattr * const cda[],u8 family)967 ctnetlink_alloc_filter(const struct nlattr * const cda[], u8 family)
968 {
969 struct ctnetlink_filter *filter;
970 int err;
971
972 #ifndef CONFIG_NF_CONNTRACK_MARK
973 if (cda[CTA_MARK] || cda[CTA_MARK_MASK])
974 return ERR_PTR(-EOPNOTSUPP);
975 #endif
976
977 filter = kzalloc(sizeof(*filter), GFP_KERNEL);
978 if (filter == NULL)
979 return ERR_PTR(-ENOMEM);
980
981 filter->family = family;
982
983 err = ctnetlink_filter_parse_mark(&filter->mark, cda);
984 if (err)
985 goto err_filter;
986
987 err = ctnetlink_filter_parse_status(&filter->status, cda);
988 if (err)
989 goto err_filter;
990
991 if (cda[CTA_ZONE]) {
992 err = ctnetlink_parse_zone(cda[CTA_ZONE], &filter->zone);
993 if (err < 0)
994 goto err_filter;
995 filter->zone_filter = true;
996 }
997
998 if (!cda[CTA_FILTER])
999 return filter;
1000
1001 err = ctnetlink_parse_filter(cda[CTA_FILTER], filter);
1002 if (err < 0)
1003 goto err_filter;
1004
1005 if (filter->orig_flags) {
1006 if (!cda[CTA_TUPLE_ORIG]) {
1007 err = -EINVAL;
1008 goto err_filter;
1009 }
1010
1011 err = ctnetlink_parse_tuple_filter(cda, &filter->orig,
1012 CTA_TUPLE_ORIG,
1013 filter->family,
1014 &filter->zone,
1015 filter->orig_flags);
1016 if (err < 0)
1017 goto err_filter;
1018 }
1019
1020 if (filter->reply_flags) {
1021 if (!cda[CTA_TUPLE_REPLY]) {
1022 err = -EINVAL;
1023 goto err_filter;
1024 }
1025
1026 err = ctnetlink_parse_tuple_filter(cda, &filter->reply,
1027 CTA_TUPLE_REPLY,
1028 filter->family,
1029 &filter->zone,
1030 filter->reply_flags);
1031 if (err < 0)
1032 goto err_filter;
1033 }
1034
1035 return filter;
1036
1037 err_filter:
1038 kfree(filter);
1039
1040 return ERR_PTR(err);
1041 }
1042
ctnetlink_needs_filter(u8 family,const struct nlattr * const * cda)1043 static bool ctnetlink_needs_filter(u8 family, const struct nlattr * const *cda)
1044 {
1045 return family || cda[CTA_MARK] || cda[CTA_FILTER] || cda[CTA_STATUS] || cda[CTA_ZONE];
1046 }
1047
ctnetlink_start(struct netlink_callback * cb)1048 static int ctnetlink_start(struct netlink_callback *cb)
1049 {
1050 const struct nlattr * const *cda = cb->data;
1051 struct ctnetlink_filter *filter = NULL;
1052 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
1053 u8 family = nfmsg->nfgen_family;
1054
1055 if (ctnetlink_needs_filter(family, cda)) {
1056 filter = ctnetlink_alloc_filter(cda, family);
1057 if (IS_ERR(filter))
1058 return PTR_ERR(filter);
1059 }
1060
1061 cb->data = filter;
1062 return 0;
1063 }
1064
ctnetlink_filter_match_tuple(struct nf_conntrack_tuple * filter_tuple,struct nf_conntrack_tuple * ct_tuple,u_int32_t flags,int family)1065 static int ctnetlink_filter_match_tuple(struct nf_conntrack_tuple *filter_tuple,
1066 struct nf_conntrack_tuple *ct_tuple,
1067 u_int32_t flags, int family)
1068 {
1069 switch (family) {
1070 case NFPROTO_IPV4:
1071 if ((flags & CTA_FILTER_FLAG(CTA_IP_SRC)) &&
1072 filter_tuple->src.u3.ip != ct_tuple->src.u3.ip)
1073 return 0;
1074
1075 if ((flags & CTA_FILTER_FLAG(CTA_IP_DST)) &&
1076 filter_tuple->dst.u3.ip != ct_tuple->dst.u3.ip)
1077 return 0;
1078 break;
1079 case NFPROTO_IPV6:
1080 if ((flags & CTA_FILTER_FLAG(CTA_IP_SRC)) &&
1081 !ipv6_addr_cmp(&filter_tuple->src.u3.in6,
1082 &ct_tuple->src.u3.in6))
1083 return 0;
1084
1085 if ((flags & CTA_FILTER_FLAG(CTA_IP_DST)) &&
1086 !ipv6_addr_cmp(&filter_tuple->dst.u3.in6,
1087 &ct_tuple->dst.u3.in6))
1088 return 0;
1089 break;
1090 }
1091
1092 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_NUM)) &&
1093 filter_tuple->dst.protonum != ct_tuple->dst.protonum)
1094 return 0;
1095
1096 switch (ct_tuple->dst.protonum) {
1097 case IPPROTO_TCP:
1098 case IPPROTO_UDP:
1099 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_SRC_PORT)) &&
1100 filter_tuple->src.u.tcp.port != ct_tuple->src.u.tcp.port)
1101 return 0;
1102
1103 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_DST_PORT)) &&
1104 filter_tuple->dst.u.tcp.port != ct_tuple->dst.u.tcp.port)
1105 return 0;
1106 break;
1107 case IPPROTO_ICMP:
1108 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_TYPE)) &&
1109 filter_tuple->dst.u.icmp.type != ct_tuple->dst.u.icmp.type)
1110 return 0;
1111 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_CODE)) &&
1112 filter_tuple->dst.u.icmp.code != ct_tuple->dst.u.icmp.code)
1113 return 0;
1114 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_ID)) &&
1115 filter_tuple->src.u.icmp.id != ct_tuple->src.u.icmp.id)
1116 return 0;
1117 break;
1118 case IPPROTO_ICMPV6:
1119 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_TYPE)) &&
1120 filter_tuple->dst.u.icmp.type != ct_tuple->dst.u.icmp.type)
1121 return 0;
1122 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_CODE)) &&
1123 filter_tuple->dst.u.icmp.code != ct_tuple->dst.u.icmp.code)
1124 return 0;
1125 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_ID)) &&
1126 filter_tuple->src.u.icmp.id != ct_tuple->src.u.icmp.id)
1127 return 0;
1128 break;
1129 }
1130
1131 return 1;
1132 }
1133
ctnetlink_filter_match(struct nf_conn * ct,void * data)1134 static int ctnetlink_filter_match(struct nf_conn *ct, void *data)
1135 {
1136 struct ctnetlink_filter *filter = data;
1137 struct nf_conntrack_tuple *tuple;
1138 u32 status;
1139
1140 if (filter == NULL)
1141 goto out;
1142
1143 /* Match entries of a given L3 protocol number.
1144 * If it is not specified, ie. l3proto == 0,
1145 * then match everything.
1146 */
1147 if (filter->family && nf_ct_l3num(ct) != filter->family)
1148 goto ignore_entry;
1149
1150 if (filter->zone_filter &&
1151 !nf_ct_zone_equal_any(ct, &filter->zone))
1152 goto ignore_entry;
1153
1154 if (filter->orig_flags) {
1155 tuple = nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL);
1156 if (!ctnetlink_filter_match_tuple(&filter->orig, tuple,
1157 filter->orig_flags,
1158 filter->family))
1159 goto ignore_entry;
1160 }
1161
1162 if (filter->reply_flags) {
1163 tuple = nf_ct_tuple(ct, IP_CT_DIR_REPLY);
1164 if (!ctnetlink_filter_match_tuple(&filter->reply, tuple,
1165 filter->reply_flags,
1166 filter->family))
1167 goto ignore_entry;
1168 }
1169
1170 #ifdef CONFIG_NF_CONNTRACK_MARK
1171 if ((READ_ONCE(ct->mark) & filter->mark.mask) != filter->mark.val)
1172 goto ignore_entry;
1173 #endif
1174 status = (u32)READ_ONCE(ct->status);
1175 if ((status & filter->status.mask) != filter->status.val)
1176 goto ignore_entry;
1177
1178 out:
1179 return 1;
1180
1181 ignore_entry:
1182 return 0;
1183 }
1184
ctnetlink_get_id(const struct nf_conn * ct)1185 static unsigned long ctnetlink_get_id(const struct nf_conn *ct)
1186 {
1187 unsigned long id = nf_ct_get_id(ct);
1188
1189 return id ? id : 1;
1190 }
1191
1192 static int
ctnetlink_dump_table(struct sk_buff * skb,struct netlink_callback * cb)1193 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1194 {
1195 unsigned int flags = cb->data ? NLM_F_DUMP_FILTERED : 0;
1196 struct net *net = sock_net(skb->sk);
1197 unsigned long last_id = cb->args[1];
1198 struct nf_conntrack_tuple_hash *h;
1199 struct hlist_nulls_node *n;
1200 struct nf_conn *nf_ct_evict[8];
1201 struct nf_conn *ct;
1202 int res, i;
1203 spinlock_t *lockp;
1204
1205 i = 0;
1206
1207 local_bh_disable();
1208 for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
1209 restart:
1210 while (i) {
1211 i--;
1212 if (nf_ct_should_gc(nf_ct_evict[i]))
1213 nf_ct_kill(nf_ct_evict[i]);
1214 nf_ct_put(nf_ct_evict[i]);
1215 }
1216
1217 lockp = &nf_conntrack_locks[cb->args[0] % CONNTRACK_LOCKS];
1218 nf_conntrack_lock(lockp);
1219 if (cb->args[0] >= nf_conntrack_htable_size) {
1220 spin_unlock(lockp);
1221 goto out;
1222 }
1223 hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[cb->args[0]],
1224 hnnode) {
1225 ct = nf_ct_tuplehash_to_ctrack(h);
1226 if (nf_ct_is_expired(ct)) {
1227 /* need to defer nf_ct_kill() until lock is released */
1228 if (i < ARRAY_SIZE(nf_ct_evict) &&
1229 refcount_inc_not_zero(&ct->ct_general.use))
1230 nf_ct_evict[i++] = ct;
1231 continue;
1232 }
1233
1234 if (!net_eq(net, nf_ct_net(ct)))
1235 continue;
1236
1237 if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
1238 continue;
1239
1240 if (cb->args[1]) {
1241 if (ctnetlink_get_id(ct) != last_id)
1242 continue;
1243 cb->args[1] = 0;
1244 }
1245 if (!ctnetlink_filter_match(ct, cb->data))
1246 continue;
1247
1248 res =
1249 ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).portid,
1250 cb->nlh->nlmsg_seq,
1251 NFNL_MSG_TYPE(cb->nlh->nlmsg_type),
1252 ct, true, flags);
1253 if (res < 0) {
1254 cb->args[1] = ctnetlink_get_id(ct);
1255 spin_unlock(lockp);
1256 goto out;
1257 }
1258 }
1259 spin_unlock(lockp);
1260 if (cb->args[1]) {
1261 cb->args[1] = 0;
1262 goto restart;
1263 }
1264 }
1265 out:
1266 local_bh_enable();
1267 if (last_id) {
1268 /* nf ct hash resize happened, now clear the leftover. */
1269 if (cb->args[1] == last_id)
1270 cb->args[1] = 0;
1271 }
1272
1273 while (i) {
1274 i--;
1275 if (nf_ct_should_gc(nf_ct_evict[i]))
1276 nf_ct_kill(nf_ct_evict[i]);
1277 nf_ct_put(nf_ct_evict[i]);
1278 }
1279
1280 return skb->len;
1281 }
1282
ipv4_nlattr_to_tuple(struct nlattr * tb[],struct nf_conntrack_tuple * t,u_int32_t flags)1283 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
1284 struct nf_conntrack_tuple *t,
1285 u_int32_t flags)
1286 {
1287 if (flags & CTA_FILTER_FLAG(CTA_IP_SRC)) {
1288 if (!tb[CTA_IP_V4_SRC])
1289 return -EINVAL;
1290
1291 t->src.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_SRC]);
1292 }
1293
1294 if (flags & CTA_FILTER_FLAG(CTA_IP_DST)) {
1295 if (!tb[CTA_IP_V4_DST])
1296 return -EINVAL;
1297
1298 t->dst.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_DST]);
1299 }
1300
1301 return 0;
1302 }
1303
ipv6_nlattr_to_tuple(struct nlattr * tb[],struct nf_conntrack_tuple * t,u_int32_t flags)1304 static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
1305 struct nf_conntrack_tuple *t,
1306 u_int32_t flags)
1307 {
1308 if (flags & CTA_FILTER_FLAG(CTA_IP_SRC)) {
1309 if (!tb[CTA_IP_V6_SRC])
1310 return -EINVAL;
1311
1312 t->src.u3.in6 = nla_get_in6_addr(tb[CTA_IP_V6_SRC]);
1313 }
1314
1315 if (flags & CTA_FILTER_FLAG(CTA_IP_DST)) {
1316 if (!tb[CTA_IP_V6_DST])
1317 return -EINVAL;
1318
1319 t->dst.u3.in6 = nla_get_in6_addr(tb[CTA_IP_V6_DST]);
1320 }
1321
1322 return 0;
1323 }
1324
ctnetlink_parse_tuple_ip(struct nlattr * attr,struct nf_conntrack_tuple * tuple,u_int32_t flags)1325 static int ctnetlink_parse_tuple_ip(struct nlattr *attr,
1326 struct nf_conntrack_tuple *tuple,
1327 u_int32_t flags)
1328 {
1329 struct nlattr *tb[CTA_IP_MAX+1];
1330 int ret = 0;
1331
1332 ret = nla_parse_nested_deprecated(tb, CTA_IP_MAX, attr,
1333 cta_ip_nla_policy, NULL);
1334 if (ret < 0)
1335 return ret;
1336
1337 switch (tuple->src.l3num) {
1338 case NFPROTO_IPV4:
1339 ret = ipv4_nlattr_to_tuple(tb, tuple, flags);
1340 break;
1341 case NFPROTO_IPV6:
1342 ret = ipv6_nlattr_to_tuple(tb, tuple, flags);
1343 break;
1344 }
1345
1346 return ret;
1347 }
1348
1349 static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = {
1350 [CTA_PROTO_NUM] = { .type = NLA_U8 },
1351 };
1352
ctnetlink_parse_tuple_proto(struct nlattr * attr,struct nf_conntrack_tuple * tuple,u_int32_t flags)1353 static int ctnetlink_parse_tuple_proto(struct nlattr *attr,
1354 struct nf_conntrack_tuple *tuple,
1355 u_int32_t flags)
1356 {
1357 const struct nf_conntrack_l4proto *l4proto;
1358 struct nlattr *tb[CTA_PROTO_MAX+1];
1359 int ret = 0;
1360
1361 ret = nla_parse_nested_deprecated(tb, CTA_PROTO_MAX, attr,
1362 proto_nla_policy, NULL);
1363 if (ret < 0)
1364 return ret;
1365
1366 if (!(flags & CTA_FILTER_FLAG(CTA_PROTO_NUM)))
1367 return 0;
1368
1369 if (!tb[CTA_PROTO_NUM])
1370 return -EINVAL;
1371
1372 tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
1373
1374 rcu_read_lock();
1375 l4proto = nf_ct_l4proto_find(tuple->dst.protonum);
1376
1377 if (likely(l4proto->nlattr_to_tuple)) {
1378 ret = nla_validate_nested_deprecated(attr, CTA_PROTO_MAX,
1379 l4proto->nla_policy,
1380 NULL);
1381 if (ret == 0)
1382 ret = l4proto->nlattr_to_tuple(tb, tuple, flags);
1383 }
1384
1385 rcu_read_unlock();
1386
1387 return ret;
1388 }
1389
1390 static int
ctnetlink_parse_zone(const struct nlattr * attr,struct nf_conntrack_zone * zone)1391 ctnetlink_parse_zone(const struct nlattr *attr,
1392 struct nf_conntrack_zone *zone)
1393 {
1394 nf_ct_zone_init(zone, NF_CT_DEFAULT_ZONE_ID,
1395 NF_CT_DEFAULT_ZONE_DIR, 0);
1396 #ifdef CONFIG_NF_CONNTRACK_ZONES
1397 if (attr)
1398 zone->id = ntohs(nla_get_be16(attr));
1399 #else
1400 if (attr)
1401 return -EOPNOTSUPP;
1402 #endif
1403 return 0;
1404 }
1405
1406 static int
ctnetlink_parse_tuple_zone(struct nlattr * attr,enum ctattr_type type,struct nf_conntrack_zone * zone)1407 ctnetlink_parse_tuple_zone(struct nlattr *attr, enum ctattr_type type,
1408 struct nf_conntrack_zone *zone)
1409 {
1410 int ret;
1411
1412 if (zone->id != NF_CT_DEFAULT_ZONE_ID)
1413 return -EINVAL;
1414
1415 ret = ctnetlink_parse_zone(attr, zone);
1416 if (ret < 0)
1417 return ret;
1418
1419 if (type == CTA_TUPLE_REPLY)
1420 zone->dir = NF_CT_ZONE_DIR_REPL;
1421 else
1422 zone->dir = NF_CT_ZONE_DIR_ORIG;
1423
1424 return 0;
1425 }
1426
1427 static const struct nla_policy tuple_nla_policy[CTA_TUPLE_MAX+1] = {
1428 [CTA_TUPLE_IP] = { .type = NLA_NESTED },
1429 [CTA_TUPLE_PROTO] = { .type = NLA_NESTED },
1430 [CTA_TUPLE_ZONE] = { .type = NLA_U16 },
1431 };
1432
1433 #define CTA_FILTER_F_ALL_CTA_PROTO \
1434 (CTA_FILTER_F_CTA_PROTO_SRC_PORT | \
1435 CTA_FILTER_F_CTA_PROTO_DST_PORT | \
1436 CTA_FILTER_F_CTA_PROTO_ICMP_TYPE | \
1437 CTA_FILTER_F_CTA_PROTO_ICMP_CODE | \
1438 CTA_FILTER_F_CTA_PROTO_ICMP_ID | \
1439 CTA_FILTER_F_CTA_PROTO_ICMPV6_TYPE | \
1440 CTA_FILTER_F_CTA_PROTO_ICMPV6_CODE | \
1441 CTA_FILTER_F_CTA_PROTO_ICMPV6_ID)
1442
1443 static int
ctnetlink_parse_tuple_filter(const struct nlattr * const cda[],struct nf_conntrack_tuple * tuple,u32 type,u_int8_t l3num,struct nf_conntrack_zone * zone,u_int32_t flags)1444 ctnetlink_parse_tuple_filter(const struct nlattr * const cda[],
1445 struct nf_conntrack_tuple *tuple, u32 type,
1446 u_int8_t l3num, struct nf_conntrack_zone *zone,
1447 u_int32_t flags)
1448 {
1449 struct nlattr *tb[CTA_TUPLE_MAX+1];
1450 int err;
1451
1452 memset(tuple, 0, sizeof(*tuple));
1453
1454 err = nla_parse_nested_deprecated(tb, CTA_TUPLE_MAX, cda[type],
1455 tuple_nla_policy, NULL);
1456 if (err < 0)
1457 return err;
1458
1459 if (l3num != NFPROTO_IPV4 && l3num != NFPROTO_IPV6)
1460 return -EOPNOTSUPP;
1461 tuple->src.l3num = l3num;
1462
1463 if (flags & CTA_FILTER_FLAG(CTA_IP_DST) ||
1464 flags & CTA_FILTER_FLAG(CTA_IP_SRC)) {
1465 if (!tb[CTA_TUPLE_IP])
1466 return -EINVAL;
1467
1468 err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple, flags);
1469 if (err < 0)
1470 return err;
1471 }
1472
1473 if (flags & CTA_FILTER_FLAG(CTA_PROTO_NUM)) {
1474 if (!tb[CTA_TUPLE_PROTO])
1475 return -EINVAL;
1476
1477 err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple, flags);
1478 if (err < 0)
1479 return err;
1480 } else if (flags & CTA_FILTER_FLAG(ALL_CTA_PROTO)) {
1481 /* Can't manage proto flags without a protonum */
1482 return -EINVAL;
1483 }
1484
1485 if ((flags & CTA_FILTER_FLAG(CTA_TUPLE_ZONE)) && tb[CTA_TUPLE_ZONE]) {
1486 if (!zone)
1487 return -EINVAL;
1488
1489 err = ctnetlink_parse_tuple_zone(tb[CTA_TUPLE_ZONE],
1490 type, zone);
1491 if (err < 0)
1492 return err;
1493 }
1494
1495 /* orig and expect tuples get DIR_ORIGINAL */
1496 if (type == CTA_TUPLE_REPLY)
1497 tuple->dst.dir = IP_CT_DIR_REPLY;
1498 else
1499 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
1500
1501 return 0;
1502 }
1503
1504 static int
ctnetlink_parse_tuple(const struct nlattr * const cda[],struct nf_conntrack_tuple * tuple,u32 type,u_int8_t l3num,struct nf_conntrack_zone * zone)1505 ctnetlink_parse_tuple(const struct nlattr * const cda[],
1506 struct nf_conntrack_tuple *tuple, u32 type,
1507 u_int8_t l3num, struct nf_conntrack_zone *zone)
1508 {
1509 return ctnetlink_parse_tuple_filter(cda, tuple, type, l3num, zone,
1510 CTA_FILTER_FLAG(ALL));
1511 }
1512
1513 static const struct nla_policy help_nla_policy[CTA_HELP_MAX+1] = {
1514 [CTA_HELP_NAME] = { .type = NLA_NUL_STRING,
1515 .len = NF_CT_HELPER_NAME_LEN - 1 },
1516 };
1517
ctnetlink_parse_help(const struct nlattr * attr,char ** helper_name,struct nlattr ** helpinfo)1518 static int ctnetlink_parse_help(const struct nlattr *attr, char **helper_name,
1519 struct nlattr **helpinfo)
1520 {
1521 int err;
1522 struct nlattr *tb[CTA_HELP_MAX+1];
1523
1524 err = nla_parse_nested_deprecated(tb, CTA_HELP_MAX, attr,
1525 help_nla_policy, NULL);
1526 if (err < 0)
1527 return err;
1528
1529 if (!tb[CTA_HELP_NAME])
1530 return -EINVAL;
1531
1532 *helper_name = nla_data(tb[CTA_HELP_NAME]);
1533
1534 if (tb[CTA_HELP_INFO])
1535 *helpinfo = tb[CTA_HELP_INFO];
1536
1537 return 0;
1538 }
1539
1540 static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
1541 [CTA_TUPLE_ORIG] = { .type = NLA_NESTED },
1542 [CTA_TUPLE_REPLY] = { .type = NLA_NESTED },
1543 [CTA_STATUS] = { .type = NLA_U32 },
1544 [CTA_PROTOINFO] = { .type = NLA_NESTED },
1545 [CTA_HELP] = { .type = NLA_NESTED },
1546 [CTA_NAT_SRC] = { .type = NLA_NESTED },
1547 [CTA_TIMEOUT] = { .type = NLA_U32 },
1548 [CTA_MARK] = { .type = NLA_U32 },
1549 [CTA_ID] = { .type = NLA_U32 },
1550 [CTA_NAT_DST] = { .type = NLA_NESTED },
1551 [CTA_TUPLE_MASTER] = { .type = NLA_NESTED },
1552 [CTA_NAT_SEQ_ADJ_ORIG] = { .type = NLA_NESTED },
1553 [CTA_NAT_SEQ_ADJ_REPLY] = { .type = NLA_NESTED },
1554 [CTA_ZONE] = { .type = NLA_U16 },
1555 [CTA_MARK_MASK] = { .type = NLA_U32 },
1556 [CTA_LABELS] = { .type = NLA_BINARY,
1557 .len = NF_CT_LABELS_MAX_SIZE },
1558 [CTA_LABELS_MASK] = { .type = NLA_BINARY,
1559 .len = NF_CT_LABELS_MAX_SIZE },
1560 [CTA_FILTER] = { .type = NLA_NESTED },
1561 [CTA_STATUS_MASK] = { .type = NLA_U32 },
1562 };
1563
ctnetlink_flush_iterate(struct nf_conn * ct,void * data)1564 static int ctnetlink_flush_iterate(struct nf_conn *ct, void *data)
1565 {
1566 return ctnetlink_filter_match(ct, data);
1567 }
1568
ctnetlink_flush_conntrack(struct net * net,const struct nlattr * const cda[],u32 portid,int report,u8 family)1569 static int ctnetlink_flush_conntrack(struct net *net,
1570 const struct nlattr * const cda[],
1571 u32 portid, int report, u8 family)
1572 {
1573 struct ctnetlink_filter *filter = NULL;
1574 struct nf_ct_iter_data iter = {
1575 .net = net,
1576 .portid = portid,
1577 .report = report,
1578 };
1579
1580 if (ctnetlink_needs_filter(family, cda)) {
1581 filter = ctnetlink_alloc_filter(cda, family);
1582 if (IS_ERR(filter))
1583 return PTR_ERR(filter);
1584
1585 iter.data = filter;
1586 }
1587
1588 nf_ct_iterate_cleanup_net(ctnetlink_flush_iterate, &iter);
1589 kfree(filter);
1590
1591 return 0;
1592 }
1593
ctnetlink_del_conntrack(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])1594 static int ctnetlink_del_conntrack(struct sk_buff *skb,
1595 const struct nfnl_info *info,
1596 const struct nlattr * const cda[])
1597 {
1598 u8 family = info->nfmsg->nfgen_family;
1599 struct nf_conntrack_tuple_hash *h;
1600 struct nf_conntrack_tuple tuple;
1601 struct nf_conntrack_zone zone;
1602 struct nf_conn *ct;
1603 int err;
1604
1605 err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone);
1606 if (err < 0)
1607 return err;
1608
1609 if (cda[CTA_TUPLE_ORIG] && !cda[CTA_FILTER])
1610 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG,
1611 family, &zone);
1612 else if (cda[CTA_TUPLE_REPLY] && !cda[CTA_FILTER])
1613 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY,
1614 family, &zone);
1615 else {
1616 u8 u3 = info->nfmsg->version || cda[CTA_FILTER] ? family : AF_UNSPEC;
1617
1618 return ctnetlink_flush_conntrack(info->net, cda,
1619 NETLINK_CB(skb).portid,
1620 nlmsg_report(info->nlh), u3);
1621 }
1622
1623 if (err < 0)
1624 return err;
1625
1626 h = nf_conntrack_find_get(info->net, &zone, &tuple);
1627 if (!h)
1628 return -ENOENT;
1629
1630 ct = nf_ct_tuplehash_to_ctrack(h);
1631
1632 if (cda[CTA_ID]) {
1633 __be32 id = nla_get_be32(cda[CTA_ID]);
1634
1635 if (id != (__force __be32)nf_ct_get_id(ct)) {
1636 nf_ct_put(ct);
1637 return -ENOENT;
1638 }
1639 }
1640
1641 nf_ct_delete(ct, NETLINK_CB(skb).portid, nlmsg_report(info->nlh));
1642 nf_ct_put(ct);
1643
1644 return 0;
1645 }
1646
ctnetlink_get_conntrack(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])1647 static int ctnetlink_get_conntrack(struct sk_buff *skb,
1648 const struct nfnl_info *info,
1649 const struct nlattr * const cda[])
1650 {
1651 u_int8_t u3 = info->nfmsg->nfgen_family;
1652 struct nf_conntrack_tuple_hash *h;
1653 struct nf_conntrack_tuple tuple;
1654 struct nf_conntrack_zone zone;
1655 struct sk_buff *skb2;
1656 struct nf_conn *ct;
1657 int err;
1658
1659 if (info->nlh->nlmsg_flags & NLM_F_DUMP) {
1660 struct netlink_dump_control c = {
1661 .start = ctnetlink_start,
1662 .dump = ctnetlink_dump_table,
1663 .done = ctnetlink_done,
1664 .data = (void *)cda,
1665 };
1666
1667 return netlink_dump_start(info->sk, skb, info->nlh, &c);
1668 }
1669
1670 err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone);
1671 if (err < 0)
1672 return err;
1673
1674 if (cda[CTA_TUPLE_ORIG])
1675 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG,
1676 u3, &zone);
1677 else if (cda[CTA_TUPLE_REPLY])
1678 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY,
1679 u3, &zone);
1680 else
1681 return -EINVAL;
1682
1683 if (err < 0)
1684 return err;
1685
1686 h = nf_conntrack_find_get(info->net, &zone, &tuple);
1687 if (!h)
1688 return -ENOENT;
1689
1690 ct = nf_ct_tuplehash_to_ctrack(h);
1691
1692 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1693 if (!skb2) {
1694 nf_ct_put(ct);
1695 return -ENOMEM;
1696 }
1697
1698 err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).portid,
1699 info->nlh->nlmsg_seq,
1700 NFNL_MSG_TYPE(info->nlh->nlmsg_type), ct,
1701 true, 0);
1702 nf_ct_put(ct);
1703 if (err <= 0) {
1704 kfree_skb(skb2);
1705 return -ENOMEM;
1706 }
1707
1708 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
1709 }
1710
ctnetlink_done_list(struct netlink_callback * cb)1711 static int ctnetlink_done_list(struct netlink_callback *cb)
1712 {
1713 struct ctnetlink_list_dump_ctx *ctx = (void *)cb->ctx;
1714
1715 if (ctx->last)
1716 nf_ct_put(ctx->last);
1717
1718 return 0;
1719 }
1720
1721 #ifdef CONFIG_NF_CONNTRACK_EVENTS
ctnetlink_dump_one_entry(struct sk_buff * skb,struct netlink_callback * cb,struct nf_conn * ct,bool dying)1722 static int ctnetlink_dump_one_entry(struct sk_buff *skb,
1723 struct netlink_callback *cb,
1724 struct nf_conn *ct,
1725 bool dying)
1726 {
1727 struct ctnetlink_list_dump_ctx *ctx = (void *)cb->ctx;
1728 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
1729 u8 l3proto = nfmsg->nfgen_family;
1730 int res;
1731
1732 if (l3proto && nf_ct_l3num(ct) != l3proto)
1733 return 0;
1734
1735 if (ctx->last) {
1736 if (ct != ctx->last)
1737 return 0;
1738
1739 ctx->last = NULL;
1740 }
1741
1742 /* We can't dump extension info for the unconfirmed
1743 * list because unconfirmed conntracks can have
1744 * ct->ext reallocated (and thus freed).
1745 *
1746 * In the dying list case ct->ext can't be free'd
1747 * until after we drop pcpu->lock.
1748 */
1749 res = ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).portid,
1750 cb->nlh->nlmsg_seq,
1751 NFNL_MSG_TYPE(cb->nlh->nlmsg_type),
1752 ct, dying, 0);
1753 if (res < 0) {
1754 if (!refcount_inc_not_zero(&ct->ct_general.use))
1755 return 0;
1756
1757 ctx->last = ct;
1758 }
1759
1760 return res;
1761 }
1762 #endif
1763
1764 static int
ctnetlink_dump_unconfirmed(struct sk_buff * skb,struct netlink_callback * cb)1765 ctnetlink_dump_unconfirmed(struct sk_buff *skb, struct netlink_callback *cb)
1766 {
1767 return 0;
1768 }
1769
1770 static int
ctnetlink_dump_dying(struct sk_buff * skb,struct netlink_callback * cb)1771 ctnetlink_dump_dying(struct sk_buff *skb, struct netlink_callback *cb)
1772 {
1773 struct ctnetlink_list_dump_ctx *ctx = (void *)cb->ctx;
1774 struct nf_conn *last = ctx->last;
1775 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1776 const struct net *net = sock_net(skb->sk);
1777 struct nf_conntrack_net_ecache *ecache_net;
1778 struct nf_conntrack_tuple_hash *h;
1779 struct hlist_nulls_node *n;
1780 #endif
1781
1782 if (ctx->done)
1783 return 0;
1784
1785 ctx->last = NULL;
1786
1787 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1788 ecache_net = nf_conn_pernet_ecache(net);
1789 spin_lock_bh(&ecache_net->dying_lock);
1790
1791 hlist_nulls_for_each_entry(h, n, &ecache_net->dying_list, hnnode) {
1792 struct nf_conn *ct;
1793 int res;
1794
1795 ct = nf_ct_tuplehash_to_ctrack(h);
1796 if (last && last != ct)
1797 continue;
1798
1799 res = ctnetlink_dump_one_entry(skb, cb, ct, true);
1800 if (res < 0) {
1801 spin_unlock_bh(&ecache_net->dying_lock);
1802 nf_ct_put(last);
1803 return skb->len;
1804 }
1805
1806 nf_ct_put(last);
1807 last = NULL;
1808 }
1809
1810 spin_unlock_bh(&ecache_net->dying_lock);
1811 #endif
1812 ctx->done = true;
1813 nf_ct_put(last);
1814
1815 return skb->len;
1816 }
1817
ctnetlink_get_ct_dying(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])1818 static int ctnetlink_get_ct_dying(struct sk_buff *skb,
1819 const struct nfnl_info *info,
1820 const struct nlattr * const cda[])
1821 {
1822 if (info->nlh->nlmsg_flags & NLM_F_DUMP) {
1823 struct netlink_dump_control c = {
1824 .dump = ctnetlink_dump_dying,
1825 .done = ctnetlink_done_list,
1826 };
1827 return netlink_dump_start(info->sk, skb, info->nlh, &c);
1828 }
1829
1830 return -EOPNOTSUPP;
1831 }
1832
ctnetlink_get_ct_unconfirmed(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])1833 static int ctnetlink_get_ct_unconfirmed(struct sk_buff *skb,
1834 const struct nfnl_info *info,
1835 const struct nlattr * const cda[])
1836 {
1837 if (info->nlh->nlmsg_flags & NLM_F_DUMP) {
1838 struct netlink_dump_control c = {
1839 .dump = ctnetlink_dump_unconfirmed,
1840 .done = ctnetlink_done_list,
1841 };
1842 return netlink_dump_start(info->sk, skb, info->nlh, &c);
1843 }
1844
1845 return -EOPNOTSUPP;
1846 }
1847
1848 #if IS_ENABLED(CONFIG_NF_NAT)
1849 static int
ctnetlink_parse_nat_setup(struct nf_conn * ct,enum nf_nat_manip_type manip,const struct nlattr * attr)1850 ctnetlink_parse_nat_setup(struct nf_conn *ct,
1851 enum nf_nat_manip_type manip,
1852 const struct nlattr *attr)
1853 __must_hold(RCU)
1854 {
1855 const struct nf_nat_hook *nat_hook;
1856 int err;
1857
1858 nat_hook = rcu_dereference(nf_nat_hook);
1859 if (!nat_hook) {
1860 #ifdef CONFIG_MODULES
1861 rcu_read_unlock();
1862 nfnl_unlock(NFNL_SUBSYS_CTNETLINK);
1863 if (request_module("nf-nat") < 0) {
1864 nfnl_lock(NFNL_SUBSYS_CTNETLINK);
1865 rcu_read_lock();
1866 return -EOPNOTSUPP;
1867 }
1868 nfnl_lock(NFNL_SUBSYS_CTNETLINK);
1869 rcu_read_lock();
1870 nat_hook = rcu_dereference(nf_nat_hook);
1871 if (nat_hook)
1872 return -EAGAIN;
1873 #endif
1874 return -EOPNOTSUPP;
1875 }
1876
1877 err = nat_hook->parse_nat_setup(ct, manip, attr);
1878 if (err == -EAGAIN) {
1879 #ifdef CONFIG_MODULES
1880 rcu_read_unlock();
1881 nfnl_unlock(NFNL_SUBSYS_CTNETLINK);
1882 if (request_module("nf-nat-%u", nf_ct_l3num(ct)) < 0) {
1883 nfnl_lock(NFNL_SUBSYS_CTNETLINK);
1884 rcu_read_lock();
1885 return -EOPNOTSUPP;
1886 }
1887 nfnl_lock(NFNL_SUBSYS_CTNETLINK);
1888 rcu_read_lock();
1889 #else
1890 err = -EOPNOTSUPP;
1891 #endif
1892 }
1893 return err;
1894 }
1895 #endif
1896
1897 static int
ctnetlink_change_status(struct nf_conn * ct,const struct nlattr * const cda[])1898 ctnetlink_change_status(struct nf_conn *ct, const struct nlattr * const cda[])
1899 {
1900 return nf_ct_change_status_common(ct, ntohl(nla_get_be32(cda[CTA_STATUS])));
1901 }
1902
1903 static int
ctnetlink_setup_nat(struct nf_conn * ct,const struct nlattr * const cda[])1904 ctnetlink_setup_nat(struct nf_conn *ct, const struct nlattr * const cda[])
1905 {
1906 #if IS_ENABLED(CONFIG_NF_NAT)
1907 int ret;
1908
1909 if (!cda[CTA_NAT_DST] && !cda[CTA_NAT_SRC])
1910 return 0;
1911
1912 ret = ctnetlink_parse_nat_setup(ct, NF_NAT_MANIP_DST,
1913 cda[CTA_NAT_DST]);
1914 if (ret < 0)
1915 return ret;
1916
1917 return ctnetlink_parse_nat_setup(ct, NF_NAT_MANIP_SRC,
1918 cda[CTA_NAT_SRC]);
1919 #else
1920 if (!cda[CTA_NAT_DST] && !cda[CTA_NAT_SRC])
1921 return 0;
1922 return -EOPNOTSUPP;
1923 #endif
1924 }
1925
ctnetlink_change_helper(struct nf_conn * ct,const struct nlattr * const cda[])1926 static int ctnetlink_change_helper(struct nf_conn *ct,
1927 const struct nlattr * const cda[])
1928 {
1929 struct nf_conntrack_helper *helper;
1930 struct nf_conn_help *help = nfct_help(ct);
1931 char *helpname = NULL;
1932 struct nlattr *helpinfo = NULL;
1933 int err;
1934
1935 err = ctnetlink_parse_help(cda[CTA_HELP], &helpname, &helpinfo);
1936 if (err < 0)
1937 return err;
1938
1939 /* don't change helper of sibling connections */
1940 if (ct->master) {
1941 /* If we try to change the helper to the same thing twice,
1942 * treat the second attempt as a no-op instead of returning
1943 * an error.
1944 */
1945 err = -EBUSY;
1946 if (help) {
1947 rcu_read_lock();
1948 helper = rcu_dereference(help->helper);
1949 if (helper && !strcmp(helper->name, helpname))
1950 err = 0;
1951 rcu_read_unlock();
1952 }
1953
1954 return err;
1955 }
1956
1957 if (!strcmp(helpname, "")) {
1958 if (help && help->helper) {
1959 /* we had a helper before ... */
1960 nf_ct_remove_expectations(ct);
1961 RCU_INIT_POINTER(help->helper, NULL);
1962 }
1963
1964 return 0;
1965 }
1966
1967 rcu_read_lock();
1968 helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct),
1969 nf_ct_protonum(ct));
1970 if (helper == NULL) {
1971 rcu_read_unlock();
1972 return -EOPNOTSUPP;
1973 }
1974
1975 if (help) {
1976 if (rcu_access_pointer(help->helper) == helper) {
1977 /* update private helper data if allowed. */
1978 if (helper->from_nlattr)
1979 helper->from_nlattr(helpinfo, ct);
1980 err = 0;
1981 } else
1982 err = -EBUSY;
1983 } else {
1984 /* we cannot set a helper for an existing conntrack */
1985 err = -EOPNOTSUPP;
1986 }
1987
1988 rcu_read_unlock();
1989 return err;
1990 }
1991
ctnetlink_change_timeout(struct nf_conn * ct,const struct nlattr * const cda[])1992 static int ctnetlink_change_timeout(struct nf_conn *ct,
1993 const struct nlattr * const cda[])
1994 {
1995 return __nf_ct_change_timeout(ct, (u64)ntohl(nla_get_be32(cda[CTA_TIMEOUT])) * HZ);
1996 }
1997
1998 #if defined(CONFIG_NF_CONNTRACK_MARK)
ctnetlink_change_mark(struct nf_conn * ct,const struct nlattr * const cda[])1999 static void ctnetlink_change_mark(struct nf_conn *ct,
2000 const struct nlattr * const cda[])
2001 {
2002 u32 mark, newmark, mask = 0;
2003
2004 if (cda[CTA_MARK_MASK])
2005 mask = ~ntohl(nla_get_be32(cda[CTA_MARK_MASK]));
2006
2007 mark = ntohl(nla_get_be32(cda[CTA_MARK]));
2008 newmark = (READ_ONCE(ct->mark) & mask) ^ mark;
2009 if (newmark != READ_ONCE(ct->mark))
2010 WRITE_ONCE(ct->mark, newmark);
2011 }
2012 #endif
2013
2014 static const struct nla_policy protoinfo_policy[CTA_PROTOINFO_MAX+1] = {
2015 [CTA_PROTOINFO_TCP] = { .type = NLA_NESTED },
2016 [CTA_PROTOINFO_DCCP] = { .type = NLA_NESTED },
2017 [CTA_PROTOINFO_SCTP] = { .type = NLA_NESTED },
2018 };
2019
ctnetlink_change_protoinfo(struct nf_conn * ct,const struct nlattr * const cda[])2020 static int ctnetlink_change_protoinfo(struct nf_conn *ct,
2021 const struct nlattr * const cda[])
2022 {
2023 const struct nlattr *attr = cda[CTA_PROTOINFO];
2024 const struct nf_conntrack_l4proto *l4proto;
2025 struct nlattr *tb[CTA_PROTOINFO_MAX+1];
2026 int err = 0;
2027
2028 err = nla_parse_nested_deprecated(tb, CTA_PROTOINFO_MAX, attr,
2029 protoinfo_policy, NULL);
2030 if (err < 0)
2031 return err;
2032
2033 l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct));
2034 if (l4proto->from_nlattr)
2035 err = l4proto->from_nlattr(tb, ct);
2036
2037 return err;
2038 }
2039
2040 static const struct nla_policy seqadj_policy[CTA_SEQADJ_MAX+1] = {
2041 [CTA_SEQADJ_CORRECTION_POS] = { .type = NLA_U32 },
2042 [CTA_SEQADJ_OFFSET_BEFORE] = { .type = NLA_U32 },
2043 [CTA_SEQADJ_OFFSET_AFTER] = { .type = NLA_U32 },
2044 };
2045
change_seq_adj(struct nf_ct_seqadj * seq,const struct nlattr * const attr)2046 static int change_seq_adj(struct nf_ct_seqadj *seq,
2047 const struct nlattr * const attr)
2048 {
2049 int err;
2050 struct nlattr *cda[CTA_SEQADJ_MAX+1];
2051
2052 err = nla_parse_nested_deprecated(cda, CTA_SEQADJ_MAX, attr,
2053 seqadj_policy, NULL);
2054 if (err < 0)
2055 return err;
2056
2057 if (!cda[CTA_SEQADJ_CORRECTION_POS])
2058 return -EINVAL;
2059
2060 seq->correction_pos =
2061 ntohl(nla_get_be32(cda[CTA_SEQADJ_CORRECTION_POS]));
2062
2063 if (!cda[CTA_SEQADJ_OFFSET_BEFORE])
2064 return -EINVAL;
2065
2066 seq->offset_before =
2067 ntohl(nla_get_be32(cda[CTA_SEQADJ_OFFSET_BEFORE]));
2068
2069 if (!cda[CTA_SEQADJ_OFFSET_AFTER])
2070 return -EINVAL;
2071
2072 seq->offset_after =
2073 ntohl(nla_get_be32(cda[CTA_SEQADJ_OFFSET_AFTER]));
2074
2075 return 0;
2076 }
2077
2078 static int
ctnetlink_change_seq_adj(struct nf_conn * ct,const struct nlattr * const cda[])2079 ctnetlink_change_seq_adj(struct nf_conn *ct,
2080 const struct nlattr * const cda[])
2081 {
2082 struct nf_conn_seqadj *seqadj = nfct_seqadj(ct);
2083 int ret = 0;
2084
2085 if (!seqadj)
2086 return 0;
2087
2088 spin_lock_bh(&ct->lock);
2089 if (cda[CTA_SEQ_ADJ_ORIG]) {
2090 ret = change_seq_adj(&seqadj->seq[IP_CT_DIR_ORIGINAL],
2091 cda[CTA_SEQ_ADJ_ORIG]);
2092 if (ret < 0)
2093 goto err;
2094
2095 set_bit(IPS_SEQ_ADJUST_BIT, &ct->status);
2096 }
2097
2098 if (cda[CTA_SEQ_ADJ_REPLY]) {
2099 ret = change_seq_adj(&seqadj->seq[IP_CT_DIR_REPLY],
2100 cda[CTA_SEQ_ADJ_REPLY]);
2101 if (ret < 0)
2102 goto err;
2103
2104 set_bit(IPS_SEQ_ADJUST_BIT, &ct->status);
2105 }
2106
2107 spin_unlock_bh(&ct->lock);
2108 return 0;
2109 err:
2110 spin_unlock_bh(&ct->lock);
2111 return ret;
2112 }
2113
2114 static const struct nla_policy synproxy_policy[CTA_SYNPROXY_MAX + 1] = {
2115 [CTA_SYNPROXY_ISN] = { .type = NLA_U32 },
2116 [CTA_SYNPROXY_ITS] = { .type = NLA_U32 },
2117 [CTA_SYNPROXY_TSOFF] = { .type = NLA_U32 },
2118 };
2119
ctnetlink_change_synproxy(struct nf_conn * ct,const struct nlattr * const cda[])2120 static int ctnetlink_change_synproxy(struct nf_conn *ct,
2121 const struct nlattr * const cda[])
2122 {
2123 struct nf_conn_synproxy *synproxy = nfct_synproxy(ct);
2124 struct nlattr *tb[CTA_SYNPROXY_MAX + 1];
2125 int err;
2126
2127 if (!synproxy)
2128 return 0;
2129
2130 err = nla_parse_nested_deprecated(tb, CTA_SYNPROXY_MAX,
2131 cda[CTA_SYNPROXY], synproxy_policy,
2132 NULL);
2133 if (err < 0)
2134 return err;
2135
2136 if (!tb[CTA_SYNPROXY_ISN] ||
2137 !tb[CTA_SYNPROXY_ITS] ||
2138 !tb[CTA_SYNPROXY_TSOFF])
2139 return -EINVAL;
2140
2141 synproxy->isn = ntohl(nla_get_be32(tb[CTA_SYNPROXY_ISN]));
2142 synproxy->its = ntohl(nla_get_be32(tb[CTA_SYNPROXY_ITS]));
2143 synproxy->tsoff = ntohl(nla_get_be32(tb[CTA_SYNPROXY_TSOFF]));
2144
2145 return 0;
2146 }
2147
2148 static int
ctnetlink_attach_labels(struct nf_conn * ct,const struct nlattr * const cda[])2149 ctnetlink_attach_labels(struct nf_conn *ct, const struct nlattr * const cda[])
2150 {
2151 #ifdef CONFIG_NF_CONNTRACK_LABELS
2152 size_t len = nla_len(cda[CTA_LABELS]);
2153 const void *mask = cda[CTA_LABELS_MASK];
2154
2155 if (len & (sizeof(u32)-1)) /* must be multiple of u32 */
2156 return -EINVAL;
2157
2158 if (mask) {
2159 if (nla_len(cda[CTA_LABELS_MASK]) == 0 ||
2160 nla_len(cda[CTA_LABELS_MASK]) != len)
2161 return -EINVAL;
2162 mask = nla_data(cda[CTA_LABELS_MASK]);
2163 }
2164
2165 len /= sizeof(u32);
2166
2167 return nf_connlabels_replace(ct, nla_data(cda[CTA_LABELS]), mask, len);
2168 #else
2169 return -EOPNOTSUPP;
2170 #endif
2171 }
2172
2173 static int
ctnetlink_change_conntrack(struct nf_conn * ct,const struct nlattr * const cda[])2174 ctnetlink_change_conntrack(struct nf_conn *ct,
2175 const struct nlattr * const cda[])
2176 {
2177 int err;
2178
2179 /* only allow NAT changes and master assignation for new conntracks */
2180 if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST] || cda[CTA_TUPLE_MASTER])
2181 return -EOPNOTSUPP;
2182
2183 if (cda[CTA_HELP]) {
2184 err = ctnetlink_change_helper(ct, cda);
2185 if (err < 0)
2186 return err;
2187 }
2188
2189 if (cda[CTA_TIMEOUT]) {
2190 err = ctnetlink_change_timeout(ct, cda);
2191 if (err < 0)
2192 return err;
2193 }
2194
2195 if (cda[CTA_STATUS]) {
2196 err = ctnetlink_change_status(ct, cda);
2197 if (err < 0)
2198 return err;
2199 }
2200
2201 if (cda[CTA_PROTOINFO]) {
2202 err = ctnetlink_change_protoinfo(ct, cda);
2203 if (err < 0)
2204 return err;
2205 }
2206
2207 #if defined(CONFIG_NF_CONNTRACK_MARK)
2208 if (cda[CTA_MARK])
2209 ctnetlink_change_mark(ct, cda);
2210 #endif
2211
2212 if (cda[CTA_SEQ_ADJ_ORIG] || cda[CTA_SEQ_ADJ_REPLY]) {
2213 err = ctnetlink_change_seq_adj(ct, cda);
2214 if (err < 0)
2215 return err;
2216 }
2217
2218 if (cda[CTA_SYNPROXY]) {
2219 err = ctnetlink_change_synproxy(ct, cda);
2220 if (err < 0)
2221 return err;
2222 }
2223
2224 if (cda[CTA_LABELS]) {
2225 err = ctnetlink_attach_labels(ct, cda);
2226 if (err < 0)
2227 return err;
2228 }
2229
2230 return 0;
2231 }
2232
2233 static struct nf_conn *
ctnetlink_create_conntrack(struct net * net,const struct nf_conntrack_zone * zone,const struct nlattr * const cda[],struct nf_conntrack_tuple * otuple,struct nf_conntrack_tuple * rtuple,u8 u3)2234 ctnetlink_create_conntrack(struct net *net,
2235 const struct nf_conntrack_zone *zone,
2236 const struct nlattr * const cda[],
2237 struct nf_conntrack_tuple *otuple,
2238 struct nf_conntrack_tuple *rtuple,
2239 u8 u3)
2240 {
2241 struct nf_conn *ct;
2242 int err = -EINVAL;
2243 struct nf_conntrack_helper *helper;
2244 struct nf_conn_tstamp *tstamp;
2245 u64 timeout;
2246
2247 ct = nf_conntrack_alloc(net, zone, otuple, rtuple, GFP_ATOMIC);
2248 if (IS_ERR(ct))
2249 return ERR_PTR(-ENOMEM);
2250
2251 if (!cda[CTA_TIMEOUT])
2252 goto err1;
2253
2254 rcu_read_lock();
2255 if (cda[CTA_HELP]) {
2256 char *helpname = NULL;
2257 struct nlattr *helpinfo = NULL;
2258
2259 err = ctnetlink_parse_help(cda[CTA_HELP], &helpname, &helpinfo);
2260 if (err < 0)
2261 goto err2;
2262
2263 helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct),
2264 nf_ct_protonum(ct));
2265 if (helper == NULL) {
2266 rcu_read_unlock();
2267 #ifdef CONFIG_MODULES
2268 if (request_module("nfct-helper-%s", helpname) < 0) {
2269 err = -EOPNOTSUPP;
2270 goto err1;
2271 }
2272
2273 rcu_read_lock();
2274 helper = __nf_conntrack_helper_find(helpname,
2275 nf_ct_l3num(ct),
2276 nf_ct_protonum(ct));
2277 if (helper) {
2278 err = -EAGAIN;
2279 goto err2;
2280 }
2281 rcu_read_unlock();
2282 #endif
2283 err = -EOPNOTSUPP;
2284 goto err1;
2285 } else {
2286 struct nf_conn_help *help;
2287
2288 help = nf_ct_helper_ext_add(ct, GFP_ATOMIC);
2289 if (help == NULL) {
2290 err = -ENOMEM;
2291 goto err2;
2292 }
2293 /* set private helper data if allowed. */
2294 if (helper->from_nlattr)
2295 helper->from_nlattr(helpinfo, ct);
2296
2297 /* disable helper auto-assignment for this entry */
2298 ct->status |= IPS_HELPER;
2299 RCU_INIT_POINTER(help->helper, helper);
2300 }
2301 }
2302
2303 err = ctnetlink_setup_nat(ct, cda);
2304 if (err < 0)
2305 goto err2;
2306
2307 nf_ct_acct_ext_add(ct, GFP_ATOMIC);
2308 nf_ct_tstamp_ext_add(ct, GFP_ATOMIC);
2309 nf_ct_ecache_ext_add(ct, 0, 0, GFP_ATOMIC);
2310 nf_ct_labels_ext_add(ct);
2311 nfct_seqadj_ext_add(ct);
2312 nfct_synproxy_ext_add(ct);
2313
2314 /* we must add conntrack extensions before confirmation. */
2315 ct->status |= IPS_CONFIRMED;
2316
2317 timeout = (u64)ntohl(nla_get_be32(cda[CTA_TIMEOUT])) * HZ;
2318 __nf_ct_set_timeout(ct, timeout);
2319
2320 if (cda[CTA_STATUS]) {
2321 err = ctnetlink_change_status(ct, cda);
2322 if (err < 0)
2323 goto err2;
2324 }
2325
2326 if (cda[CTA_SEQ_ADJ_ORIG] || cda[CTA_SEQ_ADJ_REPLY]) {
2327 err = ctnetlink_change_seq_adj(ct, cda);
2328 if (err < 0)
2329 goto err2;
2330 }
2331
2332 memset(&ct->proto, 0, sizeof(ct->proto));
2333 if (cda[CTA_PROTOINFO]) {
2334 err = ctnetlink_change_protoinfo(ct, cda);
2335 if (err < 0)
2336 goto err2;
2337 }
2338
2339 if (cda[CTA_SYNPROXY]) {
2340 err = ctnetlink_change_synproxy(ct, cda);
2341 if (err < 0)
2342 goto err2;
2343 }
2344
2345 #if defined(CONFIG_NF_CONNTRACK_MARK)
2346 if (cda[CTA_MARK])
2347 ctnetlink_change_mark(ct, cda);
2348 #endif
2349
2350 /* setup master conntrack: this is a confirmed expectation */
2351 if (cda[CTA_TUPLE_MASTER]) {
2352 struct nf_conntrack_tuple master;
2353 struct nf_conntrack_tuple_hash *master_h;
2354 struct nf_conn *master_ct;
2355
2356 err = ctnetlink_parse_tuple(cda, &master, CTA_TUPLE_MASTER,
2357 u3, NULL);
2358 if (err < 0)
2359 goto err2;
2360
2361 master_h = nf_conntrack_find_get(net, zone, &master);
2362 if (master_h == NULL) {
2363 err = -ENOENT;
2364 goto err2;
2365 }
2366 master_ct = nf_ct_tuplehash_to_ctrack(master_h);
2367 __set_bit(IPS_EXPECTED_BIT, &ct->status);
2368 ct->master = master_ct;
2369 }
2370 tstamp = nf_conn_tstamp_find(ct);
2371 if (tstamp)
2372 tstamp->start = ktime_get_real_ns();
2373
2374 err = nf_conntrack_hash_check_insert(ct);
2375 if (err < 0)
2376 goto err3;
2377
2378 rcu_read_unlock();
2379
2380 return ct;
2381
2382 err3:
2383 if (ct->master)
2384 nf_ct_put(ct->master);
2385 err2:
2386 rcu_read_unlock();
2387 err1:
2388 nf_conntrack_free(ct);
2389 return ERR_PTR(err);
2390 }
2391
ctnetlink_new_conntrack(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])2392 static int ctnetlink_new_conntrack(struct sk_buff *skb,
2393 const struct nfnl_info *info,
2394 const struct nlattr * const cda[])
2395 {
2396 struct nf_conntrack_tuple otuple, rtuple;
2397 struct nf_conntrack_tuple_hash *h = NULL;
2398 u_int8_t u3 = info->nfmsg->nfgen_family;
2399 struct nf_conntrack_zone zone;
2400 struct nf_conn *ct;
2401 int err;
2402
2403 err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone);
2404 if (err < 0)
2405 return err;
2406
2407 if (cda[CTA_TUPLE_ORIG]) {
2408 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG,
2409 u3, &zone);
2410 if (err < 0)
2411 return err;
2412 }
2413
2414 if (cda[CTA_TUPLE_REPLY]) {
2415 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY,
2416 u3, &zone);
2417 if (err < 0)
2418 return err;
2419 }
2420
2421 if (cda[CTA_TUPLE_ORIG])
2422 h = nf_conntrack_find_get(info->net, &zone, &otuple);
2423 else if (cda[CTA_TUPLE_REPLY])
2424 h = nf_conntrack_find_get(info->net, &zone, &rtuple);
2425
2426 if (h == NULL) {
2427 err = -ENOENT;
2428 if (info->nlh->nlmsg_flags & NLM_F_CREATE) {
2429 enum ip_conntrack_events events;
2430
2431 if (!cda[CTA_TUPLE_ORIG] || !cda[CTA_TUPLE_REPLY])
2432 return -EINVAL;
2433 if (otuple.dst.protonum != rtuple.dst.protonum)
2434 return -EINVAL;
2435
2436 ct = ctnetlink_create_conntrack(info->net, &zone, cda,
2437 &otuple, &rtuple, u3);
2438 if (IS_ERR(ct))
2439 return PTR_ERR(ct);
2440
2441 err = 0;
2442 if (test_bit(IPS_EXPECTED_BIT, &ct->status))
2443 events = 1 << IPCT_RELATED;
2444 else
2445 events = 1 << IPCT_NEW;
2446
2447 if (cda[CTA_LABELS] &&
2448 ctnetlink_attach_labels(ct, cda) == 0)
2449 events |= (1 << IPCT_LABEL);
2450
2451 nf_conntrack_eventmask_report((1 << IPCT_REPLY) |
2452 (1 << IPCT_ASSURED) |
2453 (1 << IPCT_HELPER) |
2454 (1 << IPCT_PROTOINFO) |
2455 (1 << IPCT_SEQADJ) |
2456 (1 << IPCT_MARK) |
2457 (1 << IPCT_SYNPROXY) |
2458 events,
2459 ct, NETLINK_CB(skb).portid,
2460 nlmsg_report(info->nlh));
2461 nf_ct_put(ct);
2462 }
2463
2464 return err;
2465 }
2466 /* implicit 'else' */
2467
2468 err = -EEXIST;
2469 ct = nf_ct_tuplehash_to_ctrack(h);
2470 if (!(info->nlh->nlmsg_flags & NLM_F_EXCL)) {
2471 err = ctnetlink_change_conntrack(ct, cda);
2472 if (err == 0) {
2473 nf_conntrack_eventmask_report((1 << IPCT_REPLY) |
2474 (1 << IPCT_ASSURED) |
2475 (1 << IPCT_HELPER) |
2476 (1 << IPCT_LABEL) |
2477 (1 << IPCT_PROTOINFO) |
2478 (1 << IPCT_SEQADJ) |
2479 (1 << IPCT_MARK) |
2480 (1 << IPCT_SYNPROXY),
2481 ct, NETLINK_CB(skb).portid,
2482 nlmsg_report(info->nlh));
2483 }
2484 }
2485
2486 nf_ct_put(ct);
2487 return err;
2488 }
2489
2490 static int
ctnetlink_ct_stat_cpu_fill_info(struct sk_buff * skb,u32 portid,u32 seq,__u16 cpu,const struct ip_conntrack_stat * st)2491 ctnetlink_ct_stat_cpu_fill_info(struct sk_buff *skb, u32 portid, u32 seq,
2492 __u16 cpu, const struct ip_conntrack_stat *st)
2493 {
2494 struct nlmsghdr *nlh;
2495 unsigned int flags = portid ? NLM_F_MULTI : 0, event;
2496
2497 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK,
2498 IPCTNL_MSG_CT_GET_STATS_CPU);
2499 nlh = nfnl_msg_put(skb, portid, seq, event, flags, AF_UNSPEC,
2500 NFNETLINK_V0, htons(cpu));
2501 if (!nlh)
2502 goto nlmsg_failure;
2503
2504 if (nla_put_be32(skb, CTA_STATS_FOUND, htonl(st->found)) ||
2505 nla_put_be32(skb, CTA_STATS_INVALID, htonl(st->invalid)) ||
2506 nla_put_be32(skb, CTA_STATS_INSERT, htonl(st->insert)) ||
2507 nla_put_be32(skb, CTA_STATS_INSERT_FAILED,
2508 htonl(st->insert_failed)) ||
2509 nla_put_be32(skb, CTA_STATS_DROP, htonl(st->drop)) ||
2510 nla_put_be32(skb, CTA_STATS_EARLY_DROP, htonl(st->early_drop)) ||
2511 nla_put_be32(skb, CTA_STATS_ERROR, htonl(st->error)) ||
2512 nla_put_be32(skb, CTA_STATS_SEARCH_RESTART,
2513 htonl(st->search_restart)) ||
2514 nla_put_be32(skb, CTA_STATS_CLASH_RESOLVE,
2515 htonl(st->clash_resolve)) ||
2516 nla_put_be32(skb, CTA_STATS_CHAIN_TOOLONG,
2517 htonl(st->chaintoolong)))
2518 goto nla_put_failure;
2519
2520 nlmsg_end(skb, nlh);
2521 return skb->len;
2522
2523 nla_put_failure:
2524 nlmsg_failure:
2525 nlmsg_cancel(skb, nlh);
2526 return -1;
2527 }
2528
2529 static int
ctnetlink_ct_stat_cpu_dump(struct sk_buff * skb,struct netlink_callback * cb)2530 ctnetlink_ct_stat_cpu_dump(struct sk_buff *skb, struct netlink_callback *cb)
2531 {
2532 int cpu;
2533 struct net *net = sock_net(skb->sk);
2534
2535 if (cb->args[0] == nr_cpu_ids)
2536 return 0;
2537
2538 for (cpu = cb->args[0]; cpu < nr_cpu_ids; cpu++) {
2539 const struct ip_conntrack_stat *st;
2540
2541 if (!cpu_possible(cpu))
2542 continue;
2543
2544 st = per_cpu_ptr(net->ct.stat, cpu);
2545 if (ctnetlink_ct_stat_cpu_fill_info(skb,
2546 NETLINK_CB(cb->skb).portid,
2547 cb->nlh->nlmsg_seq,
2548 cpu, st) < 0)
2549 break;
2550 }
2551 cb->args[0] = cpu;
2552
2553 return skb->len;
2554 }
2555
ctnetlink_stat_ct_cpu(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])2556 static int ctnetlink_stat_ct_cpu(struct sk_buff *skb,
2557 const struct nfnl_info *info,
2558 const struct nlattr * const cda[])
2559 {
2560 if (info->nlh->nlmsg_flags & NLM_F_DUMP) {
2561 struct netlink_dump_control c = {
2562 .dump = ctnetlink_ct_stat_cpu_dump,
2563 };
2564 return netlink_dump_start(info->sk, skb, info->nlh, &c);
2565 }
2566
2567 return 0;
2568 }
2569
2570 static int
ctnetlink_stat_ct_fill_info(struct sk_buff * skb,u32 portid,u32 seq,u32 type,struct net * net)2571 ctnetlink_stat_ct_fill_info(struct sk_buff *skb, u32 portid, u32 seq, u32 type,
2572 struct net *net)
2573 {
2574 unsigned int flags = portid ? NLM_F_MULTI : 0, event;
2575 unsigned int nr_conntracks;
2576 struct nlmsghdr *nlh;
2577
2578 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, IPCTNL_MSG_CT_GET_STATS);
2579 nlh = nfnl_msg_put(skb, portid, seq, event, flags, AF_UNSPEC,
2580 NFNETLINK_V0, 0);
2581 if (!nlh)
2582 goto nlmsg_failure;
2583
2584 nr_conntracks = nf_conntrack_count(net);
2585 if (nla_put_be32(skb, CTA_STATS_GLOBAL_ENTRIES, htonl(nr_conntracks)))
2586 goto nla_put_failure;
2587
2588 if (nla_put_be32(skb, CTA_STATS_GLOBAL_MAX_ENTRIES, htonl(nf_conntrack_max)))
2589 goto nla_put_failure;
2590
2591 nlmsg_end(skb, nlh);
2592 return skb->len;
2593
2594 nla_put_failure:
2595 nlmsg_failure:
2596 nlmsg_cancel(skb, nlh);
2597 return -1;
2598 }
2599
ctnetlink_stat_ct(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])2600 static int ctnetlink_stat_ct(struct sk_buff *skb, const struct nfnl_info *info,
2601 const struct nlattr * const cda[])
2602 {
2603 struct sk_buff *skb2;
2604 int err;
2605
2606 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2607 if (skb2 == NULL)
2608 return -ENOMEM;
2609
2610 err = ctnetlink_stat_ct_fill_info(skb2, NETLINK_CB(skb).portid,
2611 info->nlh->nlmsg_seq,
2612 NFNL_MSG_TYPE(info->nlh->nlmsg_type),
2613 sock_net(skb->sk));
2614 if (err <= 0) {
2615 kfree_skb(skb2);
2616 return -ENOMEM;
2617 }
2618
2619 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
2620 }
2621
2622 static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = {
2623 [CTA_EXPECT_MASTER] = { .type = NLA_NESTED },
2624 [CTA_EXPECT_TUPLE] = { .type = NLA_NESTED },
2625 [CTA_EXPECT_MASK] = { .type = NLA_NESTED },
2626 [CTA_EXPECT_TIMEOUT] = { .type = NLA_U32 },
2627 [CTA_EXPECT_ID] = { .type = NLA_U32 },
2628 [CTA_EXPECT_HELP_NAME] = { .type = NLA_NUL_STRING,
2629 .len = NF_CT_HELPER_NAME_LEN - 1 },
2630 [CTA_EXPECT_ZONE] = { .type = NLA_U16 },
2631 [CTA_EXPECT_FLAGS] = { .type = NLA_U32 },
2632 [CTA_EXPECT_CLASS] = { .type = NLA_U32 },
2633 [CTA_EXPECT_NAT] = { .type = NLA_NESTED },
2634 [CTA_EXPECT_FN] = { .type = NLA_NUL_STRING },
2635 };
2636
2637 static struct nf_conntrack_expect *
2638 ctnetlink_alloc_expect(const struct nlattr *const cda[], struct nf_conn *ct,
2639 struct nf_conntrack_helper *helper,
2640 struct nf_conntrack_tuple *tuple,
2641 struct nf_conntrack_tuple *mask);
2642
2643 #ifdef CONFIG_NETFILTER_NETLINK_GLUE_CT
2644 static size_t
ctnetlink_glue_build_size(const struct nf_conn * ct)2645 ctnetlink_glue_build_size(const struct nf_conn *ct)
2646 {
2647 return 3 * nla_total_size(0) /* CTA_TUPLE_ORIG|REPL|MASTER */
2648 + 3 * nla_total_size(0) /* CTA_TUPLE_IP */
2649 + 3 * nla_total_size(0) /* CTA_TUPLE_PROTO */
2650 + 3 * nla_total_size(sizeof(u_int8_t)) /* CTA_PROTO_NUM */
2651 + nla_total_size(sizeof(u_int32_t)) /* CTA_ID */
2652 + nla_total_size(sizeof(u_int32_t)) /* CTA_STATUS */
2653 + nla_total_size(sizeof(u_int32_t)) /* CTA_TIMEOUT */
2654 + nla_total_size(0) /* CTA_PROTOINFO */
2655 + nla_total_size(0) /* CTA_HELP */
2656 + nla_total_size(NF_CT_HELPER_NAME_LEN) /* CTA_HELP_NAME */
2657 + ctnetlink_secctx_size(ct)
2658 + ctnetlink_acct_size(ct)
2659 + ctnetlink_timestamp_size(ct)
2660 #if IS_ENABLED(CONFIG_NF_NAT)
2661 + 2 * nla_total_size(0) /* CTA_NAT_SEQ_ADJ_ORIG|REPL */
2662 + 6 * nla_total_size(sizeof(u_int32_t)) /* CTA_NAT_SEQ_OFFSET */
2663 #endif
2664 #ifdef CONFIG_NF_CONNTRACK_MARK
2665 + nla_total_size(sizeof(u_int32_t)) /* CTA_MARK */
2666 #endif
2667 #ifdef CONFIG_NF_CONNTRACK_ZONES
2668 + nla_total_size(sizeof(u_int16_t)) /* CTA_ZONE|CTA_TUPLE_ZONE */
2669 #endif
2670 + ctnetlink_proto_size(ct)
2671 ;
2672 }
2673
__ctnetlink_glue_build(struct sk_buff * skb,struct nf_conn * ct)2674 static int __ctnetlink_glue_build(struct sk_buff *skb, struct nf_conn *ct)
2675 {
2676 const struct nf_conntrack_zone *zone;
2677 struct nlattr *nest_parms;
2678
2679 zone = nf_ct_zone(ct);
2680
2681 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG);
2682 if (!nest_parms)
2683 goto nla_put_failure;
2684 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
2685 goto nla_put_failure;
2686 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
2687 NF_CT_ZONE_DIR_ORIG) < 0)
2688 goto nla_put_failure;
2689 nla_nest_end(skb, nest_parms);
2690
2691 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY);
2692 if (!nest_parms)
2693 goto nla_put_failure;
2694 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0)
2695 goto nla_put_failure;
2696 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
2697 NF_CT_ZONE_DIR_REPL) < 0)
2698 goto nla_put_failure;
2699 nla_nest_end(skb, nest_parms);
2700
2701 if (ctnetlink_dump_zone_id(skb, CTA_ZONE, zone,
2702 NF_CT_DEFAULT_ZONE_DIR) < 0)
2703 goto nla_put_failure;
2704
2705 if (ctnetlink_dump_id(skb, ct) < 0)
2706 goto nla_put_failure;
2707
2708 if (ctnetlink_dump_status(skb, ct) < 0)
2709 goto nla_put_failure;
2710
2711 if (ctnetlink_dump_timeout(skb, ct, false) < 0)
2712 goto nla_put_failure;
2713
2714 if (ctnetlink_dump_protoinfo(skb, ct, false) < 0)
2715 goto nla_put_failure;
2716
2717 if (ctnetlink_dump_acct(skb, ct, IPCTNL_MSG_CT_GET) < 0 ||
2718 ctnetlink_dump_timestamp(skb, ct) < 0)
2719 goto nla_put_failure;
2720
2721 if (ctnetlink_dump_helpinfo(skb, ct) < 0)
2722 goto nla_put_failure;
2723
2724 #ifdef CONFIG_NF_CONNTRACK_SECMARK
2725 if (ct->secmark && ctnetlink_dump_secctx(skb, ct) < 0)
2726 goto nla_put_failure;
2727 #endif
2728 if (ct->master && ctnetlink_dump_master(skb, ct) < 0)
2729 goto nla_put_failure;
2730
2731 if ((ct->status & IPS_SEQ_ADJUST) &&
2732 ctnetlink_dump_ct_seq_adj(skb, ct) < 0)
2733 goto nla_put_failure;
2734
2735 if (ctnetlink_dump_ct_synproxy(skb, ct) < 0)
2736 goto nla_put_failure;
2737
2738 #ifdef CONFIG_NF_CONNTRACK_MARK
2739 if (ctnetlink_dump_mark(skb, ct, true) < 0)
2740 goto nla_put_failure;
2741 #endif
2742 if (ctnetlink_dump_labels(skb, ct) < 0)
2743 goto nla_put_failure;
2744 return 0;
2745
2746 nla_put_failure:
2747 return -ENOSPC;
2748 }
2749
2750 static int
ctnetlink_glue_build(struct sk_buff * skb,struct nf_conn * ct,enum ip_conntrack_info ctinfo,u_int16_t ct_attr,u_int16_t ct_info_attr)2751 ctnetlink_glue_build(struct sk_buff *skb, struct nf_conn *ct,
2752 enum ip_conntrack_info ctinfo,
2753 u_int16_t ct_attr, u_int16_t ct_info_attr)
2754 {
2755 struct nlattr *nest_parms;
2756
2757 nest_parms = nla_nest_start(skb, ct_attr);
2758 if (!nest_parms)
2759 goto nla_put_failure;
2760
2761 if (__ctnetlink_glue_build(skb, ct) < 0)
2762 goto nla_put_failure;
2763
2764 nla_nest_end(skb, nest_parms);
2765
2766 if (nla_put_be32(skb, ct_info_attr, htonl(ctinfo)))
2767 goto nla_put_failure;
2768
2769 return 0;
2770
2771 nla_put_failure:
2772 return -ENOSPC;
2773 }
2774
2775 static int
ctnetlink_update_status(struct nf_conn * ct,const struct nlattr * const cda[])2776 ctnetlink_update_status(struct nf_conn *ct, const struct nlattr * const cda[])
2777 {
2778 unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
2779 unsigned long d = ct->status ^ status;
2780
2781 if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
2782 /* SEEN_REPLY bit can only be set */
2783 return -EBUSY;
2784
2785 if (d & IPS_ASSURED && !(status & IPS_ASSURED))
2786 /* ASSURED bit can only be set */
2787 return -EBUSY;
2788
2789 /* This check is less strict than ctnetlink_change_status()
2790 * because callers often flip IPS_EXPECTED bits when sending
2791 * an NFQA_CT attribute to the kernel. So ignore the
2792 * unchangeable bits but do not error out. Also user programs
2793 * are allowed to clear the bits that they are allowed to change.
2794 */
2795 __nf_ct_change_status(ct, status, ~status);
2796 return 0;
2797 }
2798
2799 static int
ctnetlink_glue_parse_ct(const struct nlattr * cda[],struct nf_conn * ct)2800 ctnetlink_glue_parse_ct(const struct nlattr *cda[], struct nf_conn *ct)
2801 {
2802 int err;
2803
2804 if (cda[CTA_TIMEOUT]) {
2805 err = ctnetlink_change_timeout(ct, cda);
2806 if (err < 0)
2807 return err;
2808 }
2809 if (cda[CTA_STATUS]) {
2810 err = ctnetlink_update_status(ct, cda);
2811 if (err < 0)
2812 return err;
2813 }
2814 if (cda[CTA_HELP]) {
2815 err = ctnetlink_change_helper(ct, cda);
2816 if (err < 0)
2817 return err;
2818 }
2819 if (cda[CTA_LABELS]) {
2820 err = ctnetlink_attach_labels(ct, cda);
2821 if (err < 0)
2822 return err;
2823 }
2824 #if defined(CONFIG_NF_CONNTRACK_MARK)
2825 if (cda[CTA_MARK]) {
2826 ctnetlink_change_mark(ct, cda);
2827 }
2828 #endif
2829 return 0;
2830 }
2831
2832 static int
ctnetlink_glue_parse(const struct nlattr * attr,struct nf_conn * ct)2833 ctnetlink_glue_parse(const struct nlattr *attr, struct nf_conn *ct)
2834 {
2835 struct nlattr *cda[CTA_MAX+1];
2836 int ret;
2837
2838 ret = nla_parse_nested_deprecated(cda, CTA_MAX, attr, ct_nla_policy,
2839 NULL);
2840 if (ret < 0)
2841 return ret;
2842
2843 return ctnetlink_glue_parse_ct((const struct nlattr **)cda, ct);
2844 }
2845
ctnetlink_glue_exp_parse(const struct nlattr * const * cda,const struct nf_conn * ct,struct nf_conntrack_tuple * tuple,struct nf_conntrack_tuple * mask)2846 static int ctnetlink_glue_exp_parse(const struct nlattr * const *cda,
2847 const struct nf_conn *ct,
2848 struct nf_conntrack_tuple *tuple,
2849 struct nf_conntrack_tuple *mask)
2850 {
2851 int err;
2852
2853 err = ctnetlink_parse_tuple(cda, tuple, CTA_EXPECT_TUPLE,
2854 nf_ct_l3num(ct), NULL);
2855 if (err < 0)
2856 return err;
2857
2858 return ctnetlink_parse_tuple(cda, mask, CTA_EXPECT_MASK,
2859 nf_ct_l3num(ct), NULL);
2860 }
2861
2862 static int
ctnetlink_glue_attach_expect(const struct nlattr * attr,struct nf_conn * ct,u32 portid,u32 report)2863 ctnetlink_glue_attach_expect(const struct nlattr *attr, struct nf_conn *ct,
2864 u32 portid, u32 report)
2865 {
2866 struct nlattr *cda[CTA_EXPECT_MAX+1];
2867 struct nf_conntrack_tuple tuple, mask;
2868 struct nf_conntrack_helper *helper = NULL;
2869 struct nf_conntrack_expect *exp;
2870 int err;
2871
2872 err = nla_parse_nested_deprecated(cda, CTA_EXPECT_MAX, attr,
2873 exp_nla_policy, NULL);
2874 if (err < 0)
2875 return err;
2876
2877 err = ctnetlink_glue_exp_parse((const struct nlattr * const *)cda,
2878 ct, &tuple, &mask);
2879 if (err < 0)
2880 return err;
2881
2882 if (cda[CTA_EXPECT_HELP_NAME]) {
2883 const char *helpname = nla_data(cda[CTA_EXPECT_HELP_NAME]);
2884
2885 helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct),
2886 nf_ct_protonum(ct));
2887 if (helper == NULL)
2888 return -EOPNOTSUPP;
2889 }
2890
2891 exp = ctnetlink_alloc_expect((const struct nlattr * const *)cda, ct,
2892 helper, &tuple, &mask);
2893 if (IS_ERR(exp))
2894 return PTR_ERR(exp);
2895
2896 err = nf_ct_expect_related_report(exp, portid, report, 0);
2897 nf_ct_expect_put(exp);
2898 return err;
2899 }
2900
ctnetlink_glue_seqadj(struct sk_buff * skb,struct nf_conn * ct,enum ip_conntrack_info ctinfo,int diff)2901 static void ctnetlink_glue_seqadj(struct sk_buff *skb, struct nf_conn *ct,
2902 enum ip_conntrack_info ctinfo, int diff)
2903 {
2904 if (!(ct->status & IPS_NAT_MASK))
2905 return;
2906
2907 nf_ct_tcp_seqadj_set(skb, ct, ctinfo, diff);
2908 }
2909
2910 static const struct nfnl_ct_hook ctnetlink_glue_hook = {
2911 .build_size = ctnetlink_glue_build_size,
2912 .build = ctnetlink_glue_build,
2913 .parse = ctnetlink_glue_parse,
2914 .attach_expect = ctnetlink_glue_attach_expect,
2915 .seq_adjust = ctnetlink_glue_seqadj,
2916 };
2917 #endif /* CONFIG_NETFILTER_NETLINK_GLUE_CT */
2918
2919 /***********************************************************************
2920 * EXPECT
2921 ***********************************************************************/
2922
ctnetlink_exp_dump_tuple(struct sk_buff * skb,const struct nf_conntrack_tuple * tuple,u32 type)2923 static int ctnetlink_exp_dump_tuple(struct sk_buff *skb,
2924 const struct nf_conntrack_tuple *tuple,
2925 u32 type)
2926 {
2927 struct nlattr *nest_parms;
2928
2929 nest_parms = nla_nest_start(skb, type);
2930 if (!nest_parms)
2931 goto nla_put_failure;
2932 if (ctnetlink_dump_tuples(skb, tuple) < 0)
2933 goto nla_put_failure;
2934 nla_nest_end(skb, nest_parms);
2935
2936 return 0;
2937
2938 nla_put_failure:
2939 return -1;
2940 }
2941
ctnetlink_exp_dump_mask(struct sk_buff * skb,const struct nf_conntrack_tuple * tuple,const struct nf_conntrack_tuple_mask * mask)2942 static int ctnetlink_exp_dump_mask(struct sk_buff *skb,
2943 const struct nf_conntrack_tuple *tuple,
2944 const struct nf_conntrack_tuple_mask *mask)
2945 {
2946 const struct nf_conntrack_l4proto *l4proto;
2947 struct nf_conntrack_tuple m;
2948 struct nlattr *nest_parms;
2949 int ret;
2950
2951 memset(&m, 0xFF, sizeof(m));
2952 memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3));
2953 m.src.u.all = mask->src.u.all;
2954 m.src.l3num = tuple->src.l3num;
2955 m.dst.protonum = tuple->dst.protonum;
2956
2957 nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK);
2958 if (!nest_parms)
2959 goto nla_put_failure;
2960
2961 rcu_read_lock();
2962 ret = ctnetlink_dump_tuples_ip(skb, &m);
2963 if (ret >= 0) {
2964 l4proto = nf_ct_l4proto_find(tuple->dst.protonum);
2965 ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto);
2966 }
2967 rcu_read_unlock();
2968
2969 if (unlikely(ret < 0))
2970 goto nla_put_failure;
2971
2972 nla_nest_end(skb, nest_parms);
2973
2974 return 0;
2975
2976 nla_put_failure:
2977 return -1;
2978 }
2979
2980 #if IS_ENABLED(CONFIG_NF_NAT)
2981 static const union nf_inet_addr any_addr;
2982 #endif
2983
nf_expect_get_id(const struct nf_conntrack_expect * exp)2984 static __be32 nf_expect_get_id(const struct nf_conntrack_expect *exp)
2985 {
2986 static siphash_aligned_key_t exp_id_seed;
2987 unsigned long a, b, c, d;
2988
2989 net_get_random_once(&exp_id_seed, sizeof(exp_id_seed));
2990
2991 a = (unsigned long)exp;
2992 b = (unsigned long)exp->helper;
2993 c = (unsigned long)exp->master;
2994 d = (unsigned long)siphash(&exp->tuple, sizeof(exp->tuple), &exp_id_seed);
2995
2996 #ifdef CONFIG_64BIT
2997 return (__force __be32)siphash_4u64((u64)a, (u64)b, (u64)c, (u64)d, &exp_id_seed);
2998 #else
2999 return (__force __be32)siphash_4u32((u32)a, (u32)b, (u32)c, (u32)d, &exp_id_seed);
3000 #endif
3001 }
3002
3003 static int
ctnetlink_exp_dump_expect(struct sk_buff * skb,const struct nf_conntrack_expect * exp)3004 ctnetlink_exp_dump_expect(struct sk_buff *skb,
3005 const struct nf_conntrack_expect *exp)
3006 {
3007 struct nf_conn *master = exp->master;
3008 long timeout = ((long)exp->timeout.expires - (long)jiffies) / HZ;
3009 struct nf_conn_help *help;
3010 #if IS_ENABLED(CONFIG_NF_NAT)
3011 struct nlattr *nest_parms;
3012 struct nf_conntrack_tuple nat_tuple = {};
3013 #endif
3014 struct nf_ct_helper_expectfn *expfn;
3015
3016 if (timeout < 0)
3017 timeout = 0;
3018
3019 if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
3020 goto nla_put_failure;
3021 if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
3022 goto nla_put_failure;
3023 if (ctnetlink_exp_dump_tuple(skb,
3024 &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
3025 CTA_EXPECT_MASTER) < 0)
3026 goto nla_put_failure;
3027
3028 #if IS_ENABLED(CONFIG_NF_NAT)
3029 if (!nf_inet_addr_cmp(&exp->saved_addr, &any_addr) ||
3030 exp->saved_proto.all) {
3031 nest_parms = nla_nest_start(skb, CTA_EXPECT_NAT);
3032 if (!nest_parms)
3033 goto nla_put_failure;
3034
3035 if (nla_put_be32(skb, CTA_EXPECT_NAT_DIR, htonl(exp->dir)))
3036 goto nla_put_failure;
3037
3038 nat_tuple.src.l3num = nf_ct_l3num(master);
3039 nat_tuple.src.u3 = exp->saved_addr;
3040 nat_tuple.dst.protonum = nf_ct_protonum(master);
3041 nat_tuple.src.u = exp->saved_proto;
3042
3043 if (ctnetlink_exp_dump_tuple(skb, &nat_tuple,
3044 CTA_EXPECT_NAT_TUPLE) < 0)
3045 goto nla_put_failure;
3046 nla_nest_end(skb, nest_parms);
3047 }
3048 #endif
3049 if (nla_put_be32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout)) ||
3050 nla_put_be32(skb, CTA_EXPECT_ID, nf_expect_get_id(exp)) ||
3051 nla_put_be32(skb, CTA_EXPECT_FLAGS, htonl(exp->flags)) ||
3052 nla_put_be32(skb, CTA_EXPECT_CLASS, htonl(exp->class)))
3053 goto nla_put_failure;
3054 help = nfct_help(master);
3055 if (help) {
3056 struct nf_conntrack_helper *helper;
3057
3058 helper = rcu_dereference(help->helper);
3059 if (helper &&
3060 nla_put_string(skb, CTA_EXPECT_HELP_NAME, helper->name))
3061 goto nla_put_failure;
3062 }
3063 expfn = nf_ct_helper_expectfn_find_by_symbol(exp->expectfn);
3064 if (expfn != NULL &&
3065 nla_put_string(skb, CTA_EXPECT_FN, expfn->name))
3066 goto nla_put_failure;
3067
3068 return 0;
3069
3070 nla_put_failure:
3071 return -1;
3072 }
3073
3074 static int
ctnetlink_exp_fill_info(struct sk_buff * skb,u32 portid,u32 seq,int event,const struct nf_conntrack_expect * exp)3075 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 portid, u32 seq,
3076 int event, const struct nf_conntrack_expect *exp)
3077 {
3078 struct nlmsghdr *nlh;
3079 unsigned int flags = portid ? NLM_F_MULTI : 0;
3080
3081 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK_EXP, event);
3082 nlh = nfnl_msg_put(skb, portid, seq, event, flags,
3083 exp->tuple.src.l3num, NFNETLINK_V0, 0);
3084 if (!nlh)
3085 goto nlmsg_failure;
3086
3087 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
3088 goto nla_put_failure;
3089
3090 nlmsg_end(skb, nlh);
3091 return skb->len;
3092
3093 nlmsg_failure:
3094 nla_put_failure:
3095 nlmsg_cancel(skb, nlh);
3096 return -1;
3097 }
3098
3099 #ifdef CONFIG_NF_CONNTRACK_EVENTS
3100 static int
ctnetlink_expect_event(unsigned int events,const struct nf_exp_event * item)3101 ctnetlink_expect_event(unsigned int events, const struct nf_exp_event *item)
3102 {
3103 struct nf_conntrack_expect *exp = item->exp;
3104 struct net *net = nf_ct_exp_net(exp);
3105 struct nlmsghdr *nlh;
3106 struct sk_buff *skb;
3107 unsigned int type, group;
3108 int flags = 0;
3109
3110 if (events & (1 << IPEXP_DESTROY)) {
3111 type = IPCTNL_MSG_EXP_DELETE;
3112 group = NFNLGRP_CONNTRACK_EXP_DESTROY;
3113 } else if (events & (1 << IPEXP_NEW)) {
3114 type = IPCTNL_MSG_EXP_NEW;
3115 flags = NLM_F_CREATE|NLM_F_EXCL;
3116 group = NFNLGRP_CONNTRACK_EXP_NEW;
3117 } else
3118 return 0;
3119
3120 if (!item->report && !nfnetlink_has_listeners(net, group))
3121 return 0;
3122
3123 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
3124 if (skb == NULL)
3125 goto errout;
3126
3127 type = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK_EXP, type);
3128 nlh = nfnl_msg_put(skb, item->portid, 0, type, flags,
3129 exp->tuple.src.l3num, NFNETLINK_V0, 0);
3130 if (!nlh)
3131 goto nlmsg_failure;
3132
3133 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
3134 goto nla_put_failure;
3135
3136 nlmsg_end(skb, nlh);
3137 nfnetlink_send(skb, net, item->portid, group, item->report, GFP_ATOMIC);
3138 return 0;
3139
3140 nla_put_failure:
3141 nlmsg_cancel(skb, nlh);
3142 nlmsg_failure:
3143 kfree_skb(skb);
3144 errout:
3145 nfnetlink_set_err(net, 0, 0, -ENOBUFS);
3146 return 0;
3147 }
3148 #endif
ctnetlink_exp_done(struct netlink_callback * cb)3149 static int ctnetlink_exp_done(struct netlink_callback *cb)
3150 {
3151 if (cb->args[1])
3152 nf_ct_expect_put((struct nf_conntrack_expect *)cb->args[1]);
3153 return 0;
3154 }
3155
3156 static int
ctnetlink_exp_dump_table(struct sk_buff * skb,struct netlink_callback * cb)3157 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
3158 {
3159 struct net *net = sock_net(skb->sk);
3160 struct nf_conntrack_expect *exp, *last;
3161 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
3162 u_int8_t l3proto = nfmsg->nfgen_family;
3163
3164 rcu_read_lock();
3165 last = (struct nf_conntrack_expect *)cb->args[1];
3166 for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) {
3167 restart:
3168 hlist_for_each_entry_rcu(exp, &nf_ct_expect_hash[cb->args[0]],
3169 hnode) {
3170 if (l3proto && exp->tuple.src.l3num != l3proto)
3171 continue;
3172
3173 if (!net_eq(nf_ct_net(exp->master), net))
3174 continue;
3175
3176 if (cb->args[1]) {
3177 if (exp != last)
3178 continue;
3179 cb->args[1] = 0;
3180 }
3181 if (ctnetlink_exp_fill_info(skb,
3182 NETLINK_CB(cb->skb).portid,
3183 cb->nlh->nlmsg_seq,
3184 IPCTNL_MSG_EXP_NEW,
3185 exp) < 0) {
3186 if (!refcount_inc_not_zero(&exp->use))
3187 continue;
3188 cb->args[1] = (unsigned long)exp;
3189 goto out;
3190 }
3191 }
3192 if (cb->args[1]) {
3193 cb->args[1] = 0;
3194 goto restart;
3195 }
3196 }
3197 out:
3198 rcu_read_unlock();
3199 if (last)
3200 nf_ct_expect_put(last);
3201
3202 return skb->len;
3203 }
3204
3205 static int
ctnetlink_exp_ct_dump_table(struct sk_buff * skb,struct netlink_callback * cb)3206 ctnetlink_exp_ct_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
3207 {
3208 struct nf_conntrack_expect *exp, *last;
3209 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
3210 struct nf_conn *ct = cb->data;
3211 struct nf_conn_help *help = nfct_help(ct);
3212 u_int8_t l3proto = nfmsg->nfgen_family;
3213
3214 if (cb->args[0])
3215 return 0;
3216
3217 rcu_read_lock();
3218 last = (struct nf_conntrack_expect *)cb->args[1];
3219 restart:
3220 hlist_for_each_entry_rcu(exp, &help->expectations, lnode) {
3221 if (l3proto && exp->tuple.src.l3num != l3proto)
3222 continue;
3223 if (cb->args[1]) {
3224 if (exp != last)
3225 continue;
3226 cb->args[1] = 0;
3227 }
3228 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).portid,
3229 cb->nlh->nlmsg_seq,
3230 IPCTNL_MSG_EXP_NEW,
3231 exp) < 0) {
3232 if (!refcount_inc_not_zero(&exp->use))
3233 continue;
3234 cb->args[1] = (unsigned long)exp;
3235 goto out;
3236 }
3237 }
3238 if (cb->args[1]) {
3239 cb->args[1] = 0;
3240 goto restart;
3241 }
3242 cb->args[0] = 1;
3243 out:
3244 rcu_read_unlock();
3245 if (last)
3246 nf_ct_expect_put(last);
3247
3248 return skb->len;
3249 }
3250
ctnetlink_dump_exp_ct(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const cda[],struct netlink_ext_ack * extack)3251 static int ctnetlink_dump_exp_ct(struct net *net, struct sock *ctnl,
3252 struct sk_buff *skb,
3253 const struct nlmsghdr *nlh,
3254 const struct nlattr * const cda[],
3255 struct netlink_ext_ack *extack)
3256 {
3257 int err;
3258 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
3259 u_int8_t u3 = nfmsg->nfgen_family;
3260 struct nf_conntrack_tuple tuple;
3261 struct nf_conntrack_tuple_hash *h;
3262 struct nf_conn *ct;
3263 struct nf_conntrack_zone zone;
3264 struct netlink_dump_control c = {
3265 .dump = ctnetlink_exp_ct_dump_table,
3266 .done = ctnetlink_exp_done,
3267 };
3268
3269 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER,
3270 u3, NULL);
3271 if (err < 0)
3272 return err;
3273
3274 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
3275 if (err < 0)
3276 return err;
3277
3278 h = nf_conntrack_find_get(net, &zone, &tuple);
3279 if (!h)
3280 return -ENOENT;
3281
3282 ct = nf_ct_tuplehash_to_ctrack(h);
3283 /* No expectation linked to this connection tracking. */
3284 if (!nfct_help(ct)) {
3285 nf_ct_put(ct);
3286 return 0;
3287 }
3288
3289 c.data = ct;
3290
3291 err = netlink_dump_start(ctnl, skb, nlh, &c);
3292 nf_ct_put(ct);
3293
3294 return err;
3295 }
3296
ctnetlink_get_expect(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])3297 static int ctnetlink_get_expect(struct sk_buff *skb,
3298 const struct nfnl_info *info,
3299 const struct nlattr * const cda[])
3300 {
3301 u_int8_t u3 = info->nfmsg->nfgen_family;
3302 struct nf_conntrack_tuple tuple;
3303 struct nf_conntrack_expect *exp;
3304 struct nf_conntrack_zone zone;
3305 struct sk_buff *skb2;
3306 int err;
3307
3308 if (info->nlh->nlmsg_flags & NLM_F_DUMP) {
3309 if (cda[CTA_EXPECT_MASTER])
3310 return ctnetlink_dump_exp_ct(info->net, info->sk, skb,
3311 info->nlh, cda,
3312 info->extack);
3313 else {
3314 struct netlink_dump_control c = {
3315 .dump = ctnetlink_exp_dump_table,
3316 .done = ctnetlink_exp_done,
3317 };
3318 return netlink_dump_start(info->sk, skb, info->nlh, &c);
3319 }
3320 }
3321
3322 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
3323 if (err < 0)
3324 return err;
3325
3326 if (cda[CTA_EXPECT_TUPLE])
3327 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE,
3328 u3, NULL);
3329 else if (cda[CTA_EXPECT_MASTER])
3330 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER,
3331 u3, NULL);
3332 else
3333 return -EINVAL;
3334
3335 if (err < 0)
3336 return err;
3337
3338 exp = nf_ct_expect_find_get(info->net, &zone, &tuple);
3339 if (!exp)
3340 return -ENOENT;
3341
3342 if (cda[CTA_EXPECT_ID]) {
3343 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
3344
3345 if (id != nf_expect_get_id(exp)) {
3346 nf_ct_expect_put(exp);
3347 return -ENOENT;
3348 }
3349 }
3350
3351 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3352 if (!skb2) {
3353 nf_ct_expect_put(exp);
3354 return -ENOMEM;
3355 }
3356
3357 rcu_read_lock();
3358 err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).portid,
3359 info->nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
3360 exp);
3361 rcu_read_unlock();
3362 nf_ct_expect_put(exp);
3363 if (err <= 0) {
3364 kfree_skb(skb2);
3365 return -ENOMEM;
3366 }
3367
3368 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid);
3369 }
3370
expect_iter_name(struct nf_conntrack_expect * exp,void * data)3371 static bool expect_iter_name(struct nf_conntrack_expect *exp, void *data)
3372 {
3373 struct nf_conntrack_helper *helper;
3374 const struct nf_conn_help *m_help;
3375 const char *name = data;
3376
3377 m_help = nfct_help(exp->master);
3378
3379 helper = rcu_dereference(m_help->helper);
3380 if (!helper)
3381 return false;
3382
3383 return strcmp(helper->name, name) == 0;
3384 }
3385
expect_iter_all(struct nf_conntrack_expect * exp,void * data)3386 static bool expect_iter_all(struct nf_conntrack_expect *exp, void *data)
3387 {
3388 return true;
3389 }
3390
ctnetlink_del_expect(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])3391 static int ctnetlink_del_expect(struct sk_buff *skb,
3392 const struct nfnl_info *info,
3393 const struct nlattr * const cda[])
3394 {
3395 u_int8_t u3 = info->nfmsg->nfgen_family;
3396 struct nf_conntrack_expect *exp;
3397 struct nf_conntrack_tuple tuple;
3398 struct nf_conntrack_zone zone;
3399 int err;
3400
3401 if (cda[CTA_EXPECT_TUPLE]) {
3402 /* delete a single expect by tuple */
3403 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
3404 if (err < 0)
3405 return err;
3406
3407 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE,
3408 u3, NULL);
3409 if (err < 0)
3410 return err;
3411
3412 /* bump usage count to 2 */
3413 exp = nf_ct_expect_find_get(info->net, &zone, &tuple);
3414 if (!exp)
3415 return -ENOENT;
3416
3417 if (cda[CTA_EXPECT_ID]) {
3418 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
3419
3420 if (id != nf_expect_get_id(exp)) {
3421 nf_ct_expect_put(exp);
3422 return -ENOENT;
3423 }
3424 }
3425
3426 /* after list removal, usage count == 1 */
3427 spin_lock_bh(&nf_conntrack_expect_lock);
3428 if (del_timer(&exp->timeout)) {
3429 nf_ct_unlink_expect_report(exp, NETLINK_CB(skb).portid,
3430 nlmsg_report(info->nlh));
3431 nf_ct_expect_put(exp);
3432 }
3433 spin_unlock_bh(&nf_conntrack_expect_lock);
3434 /* have to put what we 'get' above.
3435 * after this line usage count == 0 */
3436 nf_ct_expect_put(exp);
3437 } else if (cda[CTA_EXPECT_HELP_NAME]) {
3438 char *name = nla_data(cda[CTA_EXPECT_HELP_NAME]);
3439
3440 nf_ct_expect_iterate_net(info->net, expect_iter_name, name,
3441 NETLINK_CB(skb).portid,
3442 nlmsg_report(info->nlh));
3443 } else {
3444 /* This basically means we have to flush everything*/
3445 nf_ct_expect_iterate_net(info->net, expect_iter_all, NULL,
3446 NETLINK_CB(skb).portid,
3447 nlmsg_report(info->nlh));
3448 }
3449
3450 return 0;
3451 }
3452 static int
ctnetlink_change_expect(struct nf_conntrack_expect * x,const struct nlattr * const cda[])3453 ctnetlink_change_expect(struct nf_conntrack_expect *x,
3454 const struct nlattr * const cda[])
3455 {
3456 if (cda[CTA_EXPECT_TIMEOUT]) {
3457 if (!del_timer(&x->timeout))
3458 return -ETIME;
3459
3460 x->timeout.expires = jiffies +
3461 ntohl(nla_get_be32(cda[CTA_EXPECT_TIMEOUT])) * HZ;
3462 add_timer(&x->timeout);
3463 }
3464 return 0;
3465 }
3466
3467 #if IS_ENABLED(CONFIG_NF_NAT)
3468 static const struct nla_policy exp_nat_nla_policy[CTA_EXPECT_NAT_MAX+1] = {
3469 [CTA_EXPECT_NAT_DIR] = { .type = NLA_U32 },
3470 [CTA_EXPECT_NAT_TUPLE] = { .type = NLA_NESTED },
3471 };
3472 #endif
3473
3474 static int
ctnetlink_parse_expect_nat(const struct nlattr * attr,struct nf_conntrack_expect * exp,u_int8_t u3)3475 ctnetlink_parse_expect_nat(const struct nlattr *attr,
3476 struct nf_conntrack_expect *exp,
3477 u_int8_t u3)
3478 {
3479 #if IS_ENABLED(CONFIG_NF_NAT)
3480 struct nlattr *tb[CTA_EXPECT_NAT_MAX+1];
3481 struct nf_conntrack_tuple nat_tuple = {};
3482 int err;
3483
3484 err = nla_parse_nested_deprecated(tb, CTA_EXPECT_NAT_MAX, attr,
3485 exp_nat_nla_policy, NULL);
3486 if (err < 0)
3487 return err;
3488
3489 if (!tb[CTA_EXPECT_NAT_DIR] || !tb[CTA_EXPECT_NAT_TUPLE])
3490 return -EINVAL;
3491
3492 err = ctnetlink_parse_tuple((const struct nlattr * const *)tb,
3493 &nat_tuple, CTA_EXPECT_NAT_TUPLE,
3494 u3, NULL);
3495 if (err < 0)
3496 return err;
3497
3498 exp->saved_addr = nat_tuple.src.u3;
3499 exp->saved_proto = nat_tuple.src.u;
3500 exp->dir = ntohl(nla_get_be32(tb[CTA_EXPECT_NAT_DIR]));
3501
3502 return 0;
3503 #else
3504 return -EOPNOTSUPP;
3505 #endif
3506 }
3507
3508 static struct nf_conntrack_expect *
ctnetlink_alloc_expect(const struct nlattr * const cda[],struct nf_conn * ct,struct nf_conntrack_helper * helper,struct nf_conntrack_tuple * tuple,struct nf_conntrack_tuple * mask)3509 ctnetlink_alloc_expect(const struct nlattr * const cda[], struct nf_conn *ct,
3510 struct nf_conntrack_helper *helper,
3511 struct nf_conntrack_tuple *tuple,
3512 struct nf_conntrack_tuple *mask)
3513 {
3514 u_int32_t class = 0;
3515 struct nf_conntrack_expect *exp;
3516 struct nf_conn_help *help;
3517 int err;
3518
3519 help = nfct_help(ct);
3520 if (!help)
3521 return ERR_PTR(-EOPNOTSUPP);
3522
3523 if (cda[CTA_EXPECT_CLASS] && helper) {
3524 class = ntohl(nla_get_be32(cda[CTA_EXPECT_CLASS]));
3525 if (class > helper->expect_class_max)
3526 return ERR_PTR(-EINVAL);
3527 }
3528 exp = nf_ct_expect_alloc(ct);
3529 if (!exp)
3530 return ERR_PTR(-ENOMEM);
3531
3532 if (cda[CTA_EXPECT_FLAGS]) {
3533 exp->flags = ntohl(nla_get_be32(cda[CTA_EXPECT_FLAGS]));
3534 exp->flags &= ~NF_CT_EXPECT_USERSPACE;
3535 } else {
3536 exp->flags = 0;
3537 }
3538 if (cda[CTA_EXPECT_FN]) {
3539 const char *name = nla_data(cda[CTA_EXPECT_FN]);
3540 struct nf_ct_helper_expectfn *expfn;
3541
3542 expfn = nf_ct_helper_expectfn_find_by_name(name);
3543 if (expfn == NULL) {
3544 err = -EINVAL;
3545 goto err_out;
3546 }
3547 exp->expectfn = expfn->expectfn;
3548 } else
3549 exp->expectfn = NULL;
3550
3551 exp->class = class;
3552 exp->master = ct;
3553 exp->helper = helper;
3554 exp->tuple = *tuple;
3555 exp->mask.src.u3 = mask->src.u3;
3556 exp->mask.src.u.all = mask->src.u.all;
3557
3558 if (cda[CTA_EXPECT_NAT]) {
3559 err = ctnetlink_parse_expect_nat(cda[CTA_EXPECT_NAT],
3560 exp, nf_ct_l3num(ct));
3561 if (err < 0)
3562 goto err_out;
3563 }
3564 return exp;
3565 err_out:
3566 nf_ct_expect_put(exp);
3567 return ERR_PTR(err);
3568 }
3569
3570 static int
ctnetlink_create_expect(struct net * net,const struct nf_conntrack_zone * zone,const struct nlattr * const cda[],u_int8_t u3,u32 portid,int report)3571 ctnetlink_create_expect(struct net *net,
3572 const struct nf_conntrack_zone *zone,
3573 const struct nlattr * const cda[],
3574 u_int8_t u3, u32 portid, int report)
3575 {
3576 struct nf_conntrack_tuple tuple, mask, master_tuple;
3577 struct nf_conntrack_tuple_hash *h = NULL;
3578 struct nf_conntrack_helper *helper = NULL;
3579 struct nf_conntrack_expect *exp;
3580 struct nf_conn *ct;
3581 int err;
3582
3583 /* caller guarantees that those three CTA_EXPECT_* exist */
3584 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE,
3585 u3, NULL);
3586 if (err < 0)
3587 return err;
3588 err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK,
3589 u3, NULL);
3590 if (err < 0)
3591 return err;
3592 err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER,
3593 u3, NULL);
3594 if (err < 0)
3595 return err;
3596
3597 /* Look for master conntrack of this expectation */
3598 h = nf_conntrack_find_get(net, zone, &master_tuple);
3599 if (!h)
3600 return -ENOENT;
3601 ct = nf_ct_tuplehash_to_ctrack(h);
3602
3603 rcu_read_lock();
3604 if (cda[CTA_EXPECT_HELP_NAME]) {
3605 const char *helpname = nla_data(cda[CTA_EXPECT_HELP_NAME]);
3606
3607 helper = __nf_conntrack_helper_find(helpname, u3,
3608 nf_ct_protonum(ct));
3609 if (helper == NULL) {
3610 rcu_read_unlock();
3611 #ifdef CONFIG_MODULES
3612 if (request_module("nfct-helper-%s", helpname) < 0) {
3613 err = -EOPNOTSUPP;
3614 goto err_ct;
3615 }
3616 rcu_read_lock();
3617 helper = __nf_conntrack_helper_find(helpname, u3,
3618 nf_ct_protonum(ct));
3619 if (helper) {
3620 err = -EAGAIN;
3621 goto err_rcu;
3622 }
3623 rcu_read_unlock();
3624 #endif
3625 err = -EOPNOTSUPP;
3626 goto err_ct;
3627 }
3628 }
3629
3630 exp = ctnetlink_alloc_expect(cda, ct, helper, &tuple, &mask);
3631 if (IS_ERR(exp)) {
3632 err = PTR_ERR(exp);
3633 goto err_rcu;
3634 }
3635
3636 err = nf_ct_expect_related_report(exp, portid, report, 0);
3637 nf_ct_expect_put(exp);
3638 err_rcu:
3639 rcu_read_unlock();
3640 err_ct:
3641 nf_ct_put(ct);
3642 return err;
3643 }
3644
ctnetlink_new_expect(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])3645 static int ctnetlink_new_expect(struct sk_buff *skb,
3646 const struct nfnl_info *info,
3647 const struct nlattr * const cda[])
3648 {
3649 u_int8_t u3 = info->nfmsg->nfgen_family;
3650 struct nf_conntrack_tuple tuple;
3651 struct nf_conntrack_expect *exp;
3652 struct nf_conntrack_zone zone;
3653 int err;
3654
3655 if (!cda[CTA_EXPECT_TUPLE]
3656 || !cda[CTA_EXPECT_MASK]
3657 || !cda[CTA_EXPECT_MASTER])
3658 return -EINVAL;
3659
3660 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
3661 if (err < 0)
3662 return err;
3663
3664 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE,
3665 u3, NULL);
3666 if (err < 0)
3667 return err;
3668
3669 spin_lock_bh(&nf_conntrack_expect_lock);
3670 exp = __nf_ct_expect_find(info->net, &zone, &tuple);
3671 if (!exp) {
3672 spin_unlock_bh(&nf_conntrack_expect_lock);
3673 err = -ENOENT;
3674 if (info->nlh->nlmsg_flags & NLM_F_CREATE) {
3675 err = ctnetlink_create_expect(info->net, &zone, cda, u3,
3676 NETLINK_CB(skb).portid,
3677 nlmsg_report(info->nlh));
3678 }
3679 return err;
3680 }
3681
3682 err = -EEXIST;
3683 if (!(info->nlh->nlmsg_flags & NLM_F_EXCL))
3684 err = ctnetlink_change_expect(exp, cda);
3685 spin_unlock_bh(&nf_conntrack_expect_lock);
3686
3687 return err;
3688 }
3689
3690 static int
ctnetlink_exp_stat_fill_info(struct sk_buff * skb,u32 portid,u32 seq,int cpu,const struct ip_conntrack_stat * st)3691 ctnetlink_exp_stat_fill_info(struct sk_buff *skb, u32 portid, u32 seq, int cpu,
3692 const struct ip_conntrack_stat *st)
3693 {
3694 struct nlmsghdr *nlh;
3695 unsigned int flags = portid ? NLM_F_MULTI : 0, event;
3696
3697 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK,
3698 IPCTNL_MSG_EXP_GET_STATS_CPU);
3699 nlh = nfnl_msg_put(skb, portid, seq, event, flags, AF_UNSPEC,
3700 NFNETLINK_V0, htons(cpu));
3701 if (!nlh)
3702 goto nlmsg_failure;
3703
3704 if (nla_put_be32(skb, CTA_STATS_EXP_NEW, htonl(st->expect_new)) ||
3705 nla_put_be32(skb, CTA_STATS_EXP_CREATE, htonl(st->expect_create)) ||
3706 nla_put_be32(skb, CTA_STATS_EXP_DELETE, htonl(st->expect_delete)))
3707 goto nla_put_failure;
3708
3709 nlmsg_end(skb, nlh);
3710 return skb->len;
3711
3712 nla_put_failure:
3713 nlmsg_failure:
3714 nlmsg_cancel(skb, nlh);
3715 return -1;
3716 }
3717
3718 static int
ctnetlink_exp_stat_cpu_dump(struct sk_buff * skb,struct netlink_callback * cb)3719 ctnetlink_exp_stat_cpu_dump(struct sk_buff *skb, struct netlink_callback *cb)
3720 {
3721 int cpu;
3722 struct net *net = sock_net(skb->sk);
3723
3724 if (cb->args[0] == nr_cpu_ids)
3725 return 0;
3726
3727 for (cpu = cb->args[0]; cpu < nr_cpu_ids; cpu++) {
3728 const struct ip_conntrack_stat *st;
3729
3730 if (!cpu_possible(cpu))
3731 continue;
3732
3733 st = per_cpu_ptr(net->ct.stat, cpu);
3734 if (ctnetlink_exp_stat_fill_info(skb, NETLINK_CB(cb->skb).portid,
3735 cb->nlh->nlmsg_seq,
3736 cpu, st) < 0)
3737 break;
3738 }
3739 cb->args[0] = cpu;
3740
3741 return skb->len;
3742 }
3743
ctnetlink_stat_exp_cpu(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])3744 static int ctnetlink_stat_exp_cpu(struct sk_buff *skb,
3745 const struct nfnl_info *info,
3746 const struct nlattr * const cda[])
3747 {
3748 if (info->nlh->nlmsg_flags & NLM_F_DUMP) {
3749 struct netlink_dump_control c = {
3750 .dump = ctnetlink_exp_stat_cpu_dump,
3751 };
3752 return netlink_dump_start(info->sk, skb, info->nlh, &c);
3753 }
3754
3755 return 0;
3756 }
3757
3758 #ifdef CONFIG_NF_CONNTRACK_EVENTS
3759 static struct nf_ct_event_notifier ctnl_notifier = {
3760 .ct_event = ctnetlink_conntrack_event,
3761 .exp_event = ctnetlink_expect_event,
3762 };
3763 #endif
3764
3765 static const struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
3766 [IPCTNL_MSG_CT_NEW] = {
3767 .call = ctnetlink_new_conntrack,
3768 .type = NFNL_CB_MUTEX,
3769 .attr_count = CTA_MAX,
3770 .policy = ct_nla_policy
3771 },
3772 [IPCTNL_MSG_CT_GET] = {
3773 .call = ctnetlink_get_conntrack,
3774 .type = NFNL_CB_MUTEX,
3775 .attr_count = CTA_MAX,
3776 .policy = ct_nla_policy
3777 },
3778 [IPCTNL_MSG_CT_DELETE] = {
3779 .call = ctnetlink_del_conntrack,
3780 .type = NFNL_CB_MUTEX,
3781 .attr_count = CTA_MAX,
3782 .policy = ct_nla_policy
3783 },
3784 [IPCTNL_MSG_CT_GET_CTRZERO] = {
3785 .call = ctnetlink_get_conntrack,
3786 .type = NFNL_CB_MUTEX,
3787 .attr_count = CTA_MAX,
3788 .policy = ct_nla_policy
3789 },
3790 [IPCTNL_MSG_CT_GET_STATS_CPU] = {
3791 .call = ctnetlink_stat_ct_cpu,
3792 .type = NFNL_CB_MUTEX,
3793 },
3794 [IPCTNL_MSG_CT_GET_STATS] = {
3795 .call = ctnetlink_stat_ct,
3796 .type = NFNL_CB_MUTEX,
3797 },
3798 [IPCTNL_MSG_CT_GET_DYING] = {
3799 .call = ctnetlink_get_ct_dying,
3800 .type = NFNL_CB_MUTEX,
3801 },
3802 [IPCTNL_MSG_CT_GET_UNCONFIRMED] = {
3803 .call = ctnetlink_get_ct_unconfirmed,
3804 .type = NFNL_CB_MUTEX,
3805 },
3806 };
3807
3808 static const struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
3809 [IPCTNL_MSG_EXP_GET] = {
3810 .call = ctnetlink_get_expect,
3811 .type = NFNL_CB_MUTEX,
3812 .attr_count = CTA_EXPECT_MAX,
3813 .policy = exp_nla_policy
3814 },
3815 [IPCTNL_MSG_EXP_NEW] = {
3816 .call = ctnetlink_new_expect,
3817 .type = NFNL_CB_MUTEX,
3818 .attr_count = CTA_EXPECT_MAX,
3819 .policy = exp_nla_policy
3820 },
3821 [IPCTNL_MSG_EXP_DELETE] = {
3822 .call = ctnetlink_del_expect,
3823 .type = NFNL_CB_MUTEX,
3824 .attr_count = CTA_EXPECT_MAX,
3825 .policy = exp_nla_policy
3826 },
3827 [IPCTNL_MSG_EXP_GET_STATS_CPU] = {
3828 .call = ctnetlink_stat_exp_cpu,
3829 .type = NFNL_CB_MUTEX,
3830 },
3831 };
3832
3833 static const struct nfnetlink_subsystem ctnl_subsys = {
3834 .name = "conntrack",
3835 .subsys_id = NFNL_SUBSYS_CTNETLINK,
3836 .cb_count = IPCTNL_MSG_MAX,
3837 .cb = ctnl_cb,
3838 };
3839
3840 static const struct nfnetlink_subsystem ctnl_exp_subsys = {
3841 .name = "conntrack_expect",
3842 .subsys_id = NFNL_SUBSYS_CTNETLINK_EXP,
3843 .cb_count = IPCTNL_MSG_EXP_MAX,
3844 .cb = ctnl_exp_cb,
3845 };
3846
3847 MODULE_ALIAS("ip_conntrack_netlink");
3848 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
3849 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
3850
ctnetlink_net_init(struct net * net)3851 static int __net_init ctnetlink_net_init(struct net *net)
3852 {
3853 #ifdef CONFIG_NF_CONNTRACK_EVENTS
3854 nf_conntrack_register_notifier(net, &ctnl_notifier);
3855 #endif
3856 return 0;
3857 }
3858
ctnetlink_net_pre_exit(struct net * net)3859 static void ctnetlink_net_pre_exit(struct net *net)
3860 {
3861 #ifdef CONFIG_NF_CONNTRACK_EVENTS
3862 nf_conntrack_unregister_notifier(net);
3863 #endif
3864 }
3865
3866 static struct pernet_operations ctnetlink_net_ops = {
3867 .init = ctnetlink_net_init,
3868 .pre_exit = ctnetlink_net_pre_exit,
3869 };
3870
ctnetlink_init(void)3871 static int __init ctnetlink_init(void)
3872 {
3873 int ret;
3874
3875 NL_ASSERT_DUMP_CTX_FITS(struct ctnetlink_list_dump_ctx);
3876
3877 ret = nfnetlink_subsys_register(&ctnl_subsys);
3878 if (ret < 0) {
3879 pr_err("ctnetlink_init: cannot register with nfnetlink.\n");
3880 goto err_out;
3881 }
3882
3883 ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
3884 if (ret < 0) {
3885 pr_err("ctnetlink_init: cannot register exp with nfnetlink.\n");
3886 goto err_unreg_subsys;
3887 }
3888
3889 ret = register_pernet_subsys(&ctnetlink_net_ops);
3890 if (ret < 0) {
3891 pr_err("ctnetlink_init: cannot register pernet operations\n");
3892 goto err_unreg_exp_subsys;
3893 }
3894 #ifdef CONFIG_NETFILTER_NETLINK_GLUE_CT
3895 /* setup interaction between nf_queue and nf_conntrack_netlink. */
3896 RCU_INIT_POINTER(nfnl_ct_hook, &ctnetlink_glue_hook);
3897 #endif
3898 return 0;
3899
3900 err_unreg_exp_subsys:
3901 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
3902 err_unreg_subsys:
3903 nfnetlink_subsys_unregister(&ctnl_subsys);
3904 err_out:
3905 return ret;
3906 }
3907
ctnetlink_exit(void)3908 static void __exit ctnetlink_exit(void)
3909 {
3910 unregister_pernet_subsys(&ctnetlink_net_ops);
3911 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
3912 nfnetlink_subsys_unregister(&ctnl_subsys);
3913 #ifdef CONFIG_NETFILTER_NETLINK_GLUE_CT
3914 RCU_INIT_POINTER(nfnl_ct_hook, NULL);
3915 #endif
3916 synchronize_rcu();
3917 }
3918
3919 module_init(ctnetlink_init);
3920 module_exit(ctnetlink_exit);
3921