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