]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - net/netfilter/nf_conntrack_netlink.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'x86-microcode-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[mirror_ubuntu-jammy-kernel.git] / net / netfilter / nf_conntrack_netlink.c
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 MODULE_LICENSE("GPL");
58
59 static int ctnetlink_dump_tuples_proto(struct sk_buff *skb,
60 const struct nf_conntrack_tuple *tuple,
61 const struct nf_conntrack_l4proto *l4proto)
62 {
63 int ret = 0;
64 struct nlattr *nest_parms;
65
66 nest_parms = nla_nest_start(skb, CTA_TUPLE_PROTO);
67 if (!nest_parms)
68 goto nla_put_failure;
69 if (nla_put_u8(skb, CTA_PROTO_NUM, tuple->dst.protonum))
70 goto nla_put_failure;
71
72 if (likely(l4proto->tuple_to_nlattr))
73 ret = l4proto->tuple_to_nlattr(skb, tuple);
74
75 nla_nest_end(skb, nest_parms);
76
77 return ret;
78
79 nla_put_failure:
80 return -1;
81 }
82
83 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
84 const struct nf_conntrack_tuple *tuple)
85 {
86 if (nla_put_in_addr(skb, CTA_IP_V4_SRC, tuple->src.u3.ip) ||
87 nla_put_in_addr(skb, CTA_IP_V4_DST, tuple->dst.u3.ip))
88 return -EMSGSIZE;
89 return 0;
90 }
91
92 static int ipv6_tuple_to_nlattr(struct sk_buff *skb,
93 const struct nf_conntrack_tuple *tuple)
94 {
95 if (nla_put_in6_addr(skb, CTA_IP_V6_SRC, &tuple->src.u3.in6) ||
96 nla_put_in6_addr(skb, CTA_IP_V6_DST, &tuple->dst.u3.in6))
97 return -EMSGSIZE;
98 return 0;
99 }
100
101 static int ctnetlink_dump_tuples_ip(struct sk_buff *skb,
102 const struct nf_conntrack_tuple *tuple)
103 {
104 int ret = 0;
105 struct nlattr *nest_parms;
106
107 nest_parms = nla_nest_start(skb, CTA_TUPLE_IP);
108 if (!nest_parms)
109 goto nla_put_failure;
110
111 switch (tuple->src.l3num) {
112 case NFPROTO_IPV4:
113 ret = ipv4_tuple_to_nlattr(skb, tuple);
114 break;
115 case NFPROTO_IPV6:
116 ret = ipv6_tuple_to_nlattr(skb, tuple);
117 break;
118 }
119
120 nla_nest_end(skb, nest_parms);
121
122 return ret;
123
124 nla_put_failure:
125 return -1;
126 }
127
128 static int ctnetlink_dump_tuples(struct sk_buff *skb,
129 const struct nf_conntrack_tuple *tuple)
130 {
131 const struct nf_conntrack_l4proto *l4proto;
132 int ret;
133
134 rcu_read_lock();
135 ret = ctnetlink_dump_tuples_ip(skb, tuple);
136
137 if (ret >= 0) {
138 l4proto = nf_ct_l4proto_find(tuple->dst.protonum);
139 ret = ctnetlink_dump_tuples_proto(skb, tuple, l4proto);
140 }
141 rcu_read_unlock();
142 return ret;
143 }
144
145 static int ctnetlink_dump_zone_id(struct sk_buff *skb, int attrtype,
146 const struct nf_conntrack_zone *zone, int dir)
147 {
148 if (zone->id == NF_CT_DEFAULT_ZONE_ID || zone->dir != dir)
149 return 0;
150 if (nla_put_be16(skb, attrtype, htons(zone->id)))
151 goto nla_put_failure;
152 return 0;
153
154 nla_put_failure:
155 return -1;
156 }
157
158 static int ctnetlink_dump_status(struct sk_buff *skb, const struct nf_conn *ct)
159 {
160 if (nla_put_be32(skb, CTA_STATUS, htonl(ct->status)))
161 goto nla_put_failure;
162 return 0;
163
164 nla_put_failure:
165 return -1;
166 }
167
168 static int ctnetlink_dump_timeout(struct sk_buff *skb, const struct nf_conn *ct)
169 {
170 long timeout = nf_ct_expires(ct) / HZ;
171
172 if (nla_put_be32(skb, CTA_TIMEOUT, htonl(timeout)))
173 goto nla_put_failure;
174 return 0;
175
176 nla_put_failure:
177 return -1;
178 }
179
180 static int ctnetlink_dump_protoinfo(struct sk_buff *skb, struct nf_conn *ct)
181 {
182 const struct nf_conntrack_l4proto *l4proto;
183 struct nlattr *nest_proto;
184 int ret;
185
186 l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct));
187 if (!l4proto->to_nlattr)
188 return 0;
189
190 nest_proto = nla_nest_start(skb, CTA_PROTOINFO);
191 if (!nest_proto)
192 goto nla_put_failure;
193
194 ret = l4proto->to_nlattr(skb, nest_proto, ct);
195
196 nla_nest_end(skb, nest_proto);
197
198 return ret;
199
200 nla_put_failure:
201 return -1;
202 }
203
204 static int ctnetlink_dump_helpinfo(struct sk_buff *skb,
205 const struct nf_conn *ct)
206 {
207 struct nlattr *nest_helper;
208 const struct nf_conn_help *help = nfct_help(ct);
209 struct nf_conntrack_helper *helper;
210
211 if (!help)
212 return 0;
213
214 helper = rcu_dereference(help->helper);
215 if (!helper)
216 goto out;
217
218 nest_helper = nla_nest_start(skb, CTA_HELP);
219 if (!nest_helper)
220 goto nla_put_failure;
221 if (nla_put_string(skb, CTA_HELP_NAME, helper->name))
222 goto nla_put_failure;
223
224 if (helper->to_nlattr)
225 helper->to_nlattr(skb, ct);
226
227 nla_nest_end(skb, nest_helper);
228 out:
229 return 0;
230
231 nla_put_failure:
232 return -1;
233 }
234
235 static int
236 dump_counters(struct sk_buff *skb, struct nf_conn_acct *acct,
237 enum ip_conntrack_dir dir, int type)
238 {
239 enum ctattr_type attr = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
240 struct nf_conn_counter *counter = acct->counter;
241 struct nlattr *nest_count;
242 u64 pkts, bytes;
243
244 if (type == IPCTNL_MSG_CT_GET_CTRZERO) {
245 pkts = atomic64_xchg(&counter[dir].packets, 0);
246 bytes = atomic64_xchg(&counter[dir].bytes, 0);
247 } else {
248 pkts = atomic64_read(&counter[dir].packets);
249 bytes = atomic64_read(&counter[dir].bytes);
250 }
251
252 nest_count = nla_nest_start(skb, attr);
253 if (!nest_count)
254 goto nla_put_failure;
255
256 if (nla_put_be64(skb, CTA_COUNTERS_PACKETS, cpu_to_be64(pkts),
257 CTA_COUNTERS_PAD) ||
258 nla_put_be64(skb, CTA_COUNTERS_BYTES, cpu_to_be64(bytes),
259 CTA_COUNTERS_PAD))
260 goto nla_put_failure;
261
262 nla_nest_end(skb, nest_count);
263
264 return 0;
265
266 nla_put_failure:
267 return -1;
268 }
269
270 static int
271 ctnetlink_dump_acct(struct sk_buff *skb, const struct nf_conn *ct, int type)
272 {
273 struct nf_conn_acct *acct = nf_conn_acct_find(ct);
274
275 if (!acct)
276 return 0;
277
278 if (dump_counters(skb, acct, IP_CT_DIR_ORIGINAL, type) < 0)
279 return -1;
280 if (dump_counters(skb, acct, IP_CT_DIR_REPLY, type) < 0)
281 return -1;
282
283 return 0;
284 }
285
286 static int
287 ctnetlink_dump_timestamp(struct sk_buff *skb, const struct nf_conn *ct)
288 {
289 struct nlattr *nest_count;
290 const struct nf_conn_tstamp *tstamp;
291
292 tstamp = nf_conn_tstamp_find(ct);
293 if (!tstamp)
294 return 0;
295
296 nest_count = nla_nest_start(skb, CTA_TIMESTAMP);
297 if (!nest_count)
298 goto nla_put_failure;
299
300 if (nla_put_be64(skb, CTA_TIMESTAMP_START, cpu_to_be64(tstamp->start),
301 CTA_TIMESTAMP_PAD) ||
302 (tstamp->stop != 0 && nla_put_be64(skb, CTA_TIMESTAMP_STOP,
303 cpu_to_be64(tstamp->stop),
304 CTA_TIMESTAMP_PAD)))
305 goto nla_put_failure;
306 nla_nest_end(skb, nest_count);
307
308 return 0;
309
310 nla_put_failure:
311 return -1;
312 }
313
314 #ifdef CONFIG_NF_CONNTRACK_MARK
315 static int ctnetlink_dump_mark(struct sk_buff *skb, const struct nf_conn *ct)
316 {
317 if (nla_put_be32(skb, CTA_MARK, htonl(ct->mark)))
318 goto nla_put_failure;
319 return 0;
320
321 nla_put_failure:
322 return -1;
323 }
324 #else
325 #define ctnetlink_dump_mark(a, b) (0)
326 #endif
327
328 #ifdef CONFIG_NF_CONNTRACK_SECMARK
329 static int ctnetlink_dump_secctx(struct sk_buff *skb, const struct nf_conn *ct)
330 {
331 struct nlattr *nest_secctx;
332 int len, ret;
333 char *secctx;
334
335 ret = security_secid_to_secctx(ct->secmark, &secctx, &len);
336 if (ret)
337 return 0;
338
339 ret = -1;
340 nest_secctx = nla_nest_start(skb, CTA_SECCTX);
341 if (!nest_secctx)
342 goto nla_put_failure;
343
344 if (nla_put_string(skb, CTA_SECCTX_NAME, secctx))
345 goto nla_put_failure;
346 nla_nest_end(skb, nest_secctx);
347
348 ret = 0;
349 nla_put_failure:
350 security_release_secctx(secctx, len);
351 return ret;
352 }
353 #else
354 #define ctnetlink_dump_secctx(a, b) (0)
355 #endif
356
357 #ifdef CONFIG_NF_CONNTRACK_LABELS
358 static inline int ctnetlink_label_size(const struct nf_conn *ct)
359 {
360 struct nf_conn_labels *labels = nf_ct_labels_find(ct);
361
362 if (!labels)
363 return 0;
364 return nla_total_size(sizeof(labels->bits));
365 }
366
367 static int
368 ctnetlink_dump_labels(struct sk_buff *skb, const struct nf_conn *ct)
369 {
370 struct nf_conn_labels *labels = nf_ct_labels_find(ct);
371 unsigned int i;
372
373 if (!labels)
374 return 0;
375
376 i = 0;
377 do {
378 if (labels->bits[i] != 0)
379 return nla_put(skb, CTA_LABELS, sizeof(labels->bits),
380 labels->bits);
381 i++;
382 } while (i < ARRAY_SIZE(labels->bits));
383
384 return 0;
385 }
386 #else
387 #define ctnetlink_dump_labels(a, b) (0)
388 #define ctnetlink_label_size(a) (0)
389 #endif
390
391 #define master_tuple(ct) &(ct->master->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
392
393 static int ctnetlink_dump_master(struct sk_buff *skb, const struct nf_conn *ct)
394 {
395 struct nlattr *nest_parms;
396
397 if (!(ct->status & IPS_EXPECTED))
398 return 0;
399
400 nest_parms = nla_nest_start(skb, CTA_TUPLE_MASTER);
401 if (!nest_parms)
402 goto nla_put_failure;
403 if (ctnetlink_dump_tuples(skb, master_tuple(ct)) < 0)
404 goto nla_put_failure;
405 nla_nest_end(skb, nest_parms);
406
407 return 0;
408
409 nla_put_failure:
410 return -1;
411 }
412
413 static int
414 dump_ct_seq_adj(struct sk_buff *skb, const struct nf_ct_seqadj *seq, int type)
415 {
416 struct nlattr *nest_parms;
417
418 nest_parms = nla_nest_start(skb, type);
419 if (!nest_parms)
420 goto nla_put_failure;
421
422 if (nla_put_be32(skb, CTA_SEQADJ_CORRECTION_POS,
423 htonl(seq->correction_pos)) ||
424 nla_put_be32(skb, CTA_SEQADJ_OFFSET_BEFORE,
425 htonl(seq->offset_before)) ||
426 nla_put_be32(skb, CTA_SEQADJ_OFFSET_AFTER,
427 htonl(seq->offset_after)))
428 goto nla_put_failure;
429
430 nla_nest_end(skb, nest_parms);
431
432 return 0;
433
434 nla_put_failure:
435 return -1;
436 }
437
438 static int ctnetlink_dump_ct_seq_adj(struct sk_buff *skb, struct nf_conn *ct)
439 {
440 struct nf_conn_seqadj *seqadj = nfct_seqadj(ct);
441 struct nf_ct_seqadj *seq;
442
443 if (!(ct->status & IPS_SEQ_ADJUST) || !seqadj)
444 return 0;
445
446 spin_lock_bh(&ct->lock);
447 seq = &seqadj->seq[IP_CT_DIR_ORIGINAL];
448 if (dump_ct_seq_adj(skb, seq, CTA_SEQ_ADJ_ORIG) == -1)
449 goto err;
450
451 seq = &seqadj->seq[IP_CT_DIR_REPLY];
452 if (dump_ct_seq_adj(skb, seq, CTA_SEQ_ADJ_REPLY) == -1)
453 goto err;
454
455 spin_unlock_bh(&ct->lock);
456 return 0;
457 err:
458 spin_unlock_bh(&ct->lock);
459 return -1;
460 }
461
462 static int ctnetlink_dump_ct_synproxy(struct sk_buff *skb, struct nf_conn *ct)
463 {
464 struct nf_conn_synproxy *synproxy = nfct_synproxy(ct);
465 struct nlattr *nest_parms;
466
467 if (!synproxy)
468 return 0;
469
470 nest_parms = nla_nest_start(skb, CTA_SYNPROXY);
471 if (!nest_parms)
472 goto nla_put_failure;
473
474 if (nla_put_be32(skb, CTA_SYNPROXY_ISN, htonl(synproxy->isn)) ||
475 nla_put_be32(skb, CTA_SYNPROXY_ITS, htonl(synproxy->its)) ||
476 nla_put_be32(skb, CTA_SYNPROXY_TSOFF, htonl(synproxy->tsoff)))
477 goto nla_put_failure;
478
479 nla_nest_end(skb, nest_parms);
480
481 return 0;
482
483 nla_put_failure:
484 return -1;
485 }
486
487 static int ctnetlink_dump_id(struct sk_buff *skb, const struct nf_conn *ct)
488 {
489 __be32 id = (__force __be32)nf_ct_get_id(ct);
490
491 if (nla_put_be32(skb, CTA_ID, id))
492 goto nla_put_failure;
493 return 0;
494
495 nla_put_failure:
496 return -1;
497 }
498
499 static int ctnetlink_dump_use(struct sk_buff *skb, const struct nf_conn *ct)
500 {
501 if (nla_put_be32(skb, CTA_USE, htonl(atomic_read(&ct->ct_general.use))))
502 goto nla_put_failure;
503 return 0;
504
505 nla_put_failure:
506 return -1;
507 }
508
509 /* all these functions access ct->ext. Caller must either hold a reference
510 * on ct or prevent its deletion by holding either the bucket spinlock or
511 * pcpu dying list lock.
512 */
513 static int ctnetlink_dump_extinfo(struct sk_buff *skb,
514 struct nf_conn *ct, u32 type)
515 {
516 if (ctnetlink_dump_acct(skb, ct, type) < 0 ||
517 ctnetlink_dump_timestamp(skb, ct) < 0 ||
518 ctnetlink_dump_helpinfo(skb, ct) < 0 ||
519 ctnetlink_dump_labels(skb, ct) < 0 ||
520 ctnetlink_dump_ct_seq_adj(skb, ct) < 0 ||
521 ctnetlink_dump_ct_synproxy(skb, ct) < 0)
522 return -1;
523
524 return 0;
525 }
526
527 static int ctnetlink_dump_info(struct sk_buff *skb, struct nf_conn *ct)
528 {
529 if (ctnetlink_dump_status(skb, ct) < 0 ||
530 ctnetlink_dump_mark(skb, ct) < 0 ||
531 ctnetlink_dump_secctx(skb, ct) < 0 ||
532 ctnetlink_dump_id(skb, ct) < 0 ||
533 ctnetlink_dump_use(skb, ct) < 0 ||
534 ctnetlink_dump_master(skb, ct) < 0)
535 return -1;
536
537 if (!test_bit(IPS_OFFLOAD_BIT, &ct->status) &&
538 (ctnetlink_dump_timeout(skb, ct) < 0 ||
539 ctnetlink_dump_protoinfo(skb, ct) < 0))
540 return -1;
541
542 return 0;
543 }
544
545 static int
546 ctnetlink_fill_info(struct sk_buff *skb, u32 portid, u32 seq, u32 type,
547 struct nf_conn *ct, bool extinfo)
548 {
549 const struct nf_conntrack_zone *zone;
550 struct nlmsghdr *nlh;
551 struct nfgenmsg *nfmsg;
552 struct nlattr *nest_parms;
553 unsigned int flags = portid ? NLM_F_MULTI : 0, event;
554
555 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, IPCTNL_MSG_CT_NEW);
556 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*nfmsg), flags);
557 if (nlh == NULL)
558 goto nlmsg_failure;
559
560 nfmsg = nlmsg_data(nlh);
561 nfmsg->nfgen_family = nf_ct_l3num(ct);
562 nfmsg->version = NFNETLINK_V0;
563 nfmsg->res_id = 0;
564
565 zone = nf_ct_zone(ct);
566
567 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG);
568 if (!nest_parms)
569 goto nla_put_failure;
570 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
571 goto nla_put_failure;
572 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
573 NF_CT_ZONE_DIR_ORIG) < 0)
574 goto nla_put_failure;
575 nla_nest_end(skb, nest_parms);
576
577 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY);
578 if (!nest_parms)
579 goto nla_put_failure;
580 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0)
581 goto nla_put_failure;
582 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
583 NF_CT_ZONE_DIR_REPL) < 0)
584 goto nla_put_failure;
585 nla_nest_end(skb, nest_parms);
586
587 if (ctnetlink_dump_zone_id(skb, CTA_ZONE, zone,
588 NF_CT_DEFAULT_ZONE_DIR) < 0)
589 goto nla_put_failure;
590
591 if (ctnetlink_dump_info(skb, ct) < 0)
592 goto nla_put_failure;
593 if (extinfo && ctnetlink_dump_extinfo(skb, ct, type) < 0)
594 goto nla_put_failure;
595
596 nlmsg_end(skb, nlh);
597 return skb->len;
598
599 nlmsg_failure:
600 nla_put_failure:
601 nlmsg_cancel(skb, nlh);
602 return -1;
603 }
604
605 static const struct nla_policy cta_ip_nla_policy[CTA_IP_MAX + 1] = {
606 [CTA_IP_V4_SRC] = { .type = NLA_U32 },
607 [CTA_IP_V4_DST] = { .type = NLA_U32 },
608 [CTA_IP_V6_SRC] = { .len = sizeof(__be32) * 4 },
609 [CTA_IP_V6_DST] = { .len = sizeof(__be32) * 4 },
610 };
611
612 #if defined(CONFIG_NETFILTER_NETLINK_GLUE_CT) || defined(CONFIG_NF_CONNTRACK_EVENTS)
613 static size_t ctnetlink_proto_size(const struct nf_conn *ct)
614 {
615 const struct nf_conntrack_l4proto *l4proto;
616 size_t len, len4 = 0;
617
618 len = nla_policy_len(cta_ip_nla_policy, CTA_IP_MAX + 1);
619 len *= 3u; /* ORIG, REPLY, MASTER */
620
621 l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct));
622 len += l4proto->nlattr_size;
623 if (l4proto->nlattr_tuple_size) {
624 len4 = l4proto->nlattr_tuple_size();
625 len4 *= 3u; /* ORIG, REPLY, MASTER */
626 }
627
628 return len + len4;
629 }
630 #endif
631
632 static inline size_t ctnetlink_acct_size(const struct nf_conn *ct)
633 {
634 if (!nf_ct_ext_exist(ct, NF_CT_EXT_ACCT))
635 return 0;
636 return 2 * nla_total_size(0) /* CTA_COUNTERS_ORIG|REPL */
637 + 2 * nla_total_size_64bit(sizeof(uint64_t)) /* CTA_COUNTERS_PACKETS */
638 + 2 * nla_total_size_64bit(sizeof(uint64_t)) /* CTA_COUNTERS_BYTES */
639 ;
640 }
641
642 static inline int ctnetlink_secctx_size(const struct nf_conn *ct)
643 {
644 #ifdef CONFIG_NF_CONNTRACK_SECMARK
645 int len, ret;
646
647 ret = security_secid_to_secctx(ct->secmark, NULL, &len);
648 if (ret)
649 return 0;
650
651 return nla_total_size(0) /* CTA_SECCTX */
652 + nla_total_size(sizeof(char) * len); /* CTA_SECCTX_NAME */
653 #else
654 return 0;
655 #endif
656 }
657
658 static inline size_t ctnetlink_timestamp_size(const struct nf_conn *ct)
659 {
660 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP
661 if (!nf_ct_ext_exist(ct, NF_CT_EXT_TSTAMP))
662 return 0;
663 return nla_total_size(0) + 2 * nla_total_size_64bit(sizeof(uint64_t));
664 #else
665 return 0;
666 #endif
667 }
668
669 #ifdef CONFIG_NF_CONNTRACK_EVENTS
670 static size_t ctnetlink_nlmsg_size(const struct nf_conn *ct)
671 {
672 return NLMSG_ALIGN(sizeof(struct nfgenmsg))
673 + 3 * nla_total_size(0) /* CTA_TUPLE_ORIG|REPL|MASTER */
674 + 3 * nla_total_size(0) /* CTA_TUPLE_IP */
675 + 3 * nla_total_size(0) /* CTA_TUPLE_PROTO */
676 + 3 * nla_total_size(sizeof(u_int8_t)) /* CTA_PROTO_NUM */
677 + nla_total_size(sizeof(u_int32_t)) /* CTA_ID */
678 + nla_total_size(sizeof(u_int32_t)) /* CTA_STATUS */
679 + ctnetlink_acct_size(ct)
680 + ctnetlink_timestamp_size(ct)
681 + nla_total_size(sizeof(u_int32_t)) /* CTA_TIMEOUT */
682 + nla_total_size(0) /* CTA_PROTOINFO */
683 + nla_total_size(0) /* CTA_HELP */
684 + nla_total_size(NF_CT_HELPER_NAME_LEN) /* CTA_HELP_NAME */
685 + ctnetlink_secctx_size(ct)
686 #if IS_ENABLED(CONFIG_NF_NAT)
687 + 2 * nla_total_size(0) /* CTA_NAT_SEQ_ADJ_ORIG|REPL */
688 + 6 * nla_total_size(sizeof(u_int32_t)) /* CTA_NAT_SEQ_OFFSET */
689 #endif
690 #ifdef CONFIG_NF_CONNTRACK_MARK
691 + nla_total_size(sizeof(u_int32_t)) /* CTA_MARK */
692 #endif
693 #ifdef CONFIG_NF_CONNTRACK_ZONES
694 + nla_total_size(sizeof(u_int16_t)) /* CTA_ZONE|CTA_TUPLE_ZONE */
695 #endif
696 + ctnetlink_proto_size(ct)
697 + ctnetlink_label_size(ct)
698 ;
699 }
700
701 static int
702 ctnetlink_conntrack_event(unsigned int events, struct nf_ct_event *item)
703 {
704 const struct nf_conntrack_zone *zone;
705 struct net *net;
706 struct nlmsghdr *nlh;
707 struct nfgenmsg *nfmsg;
708 struct nlattr *nest_parms;
709 struct nf_conn *ct = item->ct;
710 struct sk_buff *skb;
711 unsigned int type;
712 unsigned int flags = 0, group;
713 int err;
714
715 if (events & (1 << IPCT_DESTROY)) {
716 type = IPCTNL_MSG_CT_DELETE;
717 group = NFNLGRP_CONNTRACK_DESTROY;
718 } else if (events & ((1 << IPCT_NEW) | (1 << IPCT_RELATED))) {
719 type = IPCTNL_MSG_CT_NEW;
720 flags = NLM_F_CREATE|NLM_F_EXCL;
721 group = NFNLGRP_CONNTRACK_NEW;
722 } else if (events) {
723 type = IPCTNL_MSG_CT_NEW;
724 group = NFNLGRP_CONNTRACK_UPDATE;
725 } else
726 return 0;
727
728 net = nf_ct_net(ct);
729 if (!item->report && !nfnetlink_has_listeners(net, group))
730 return 0;
731
732 skb = nlmsg_new(ctnetlink_nlmsg_size(ct), GFP_ATOMIC);
733 if (skb == NULL)
734 goto errout;
735
736 type = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, type);
737 nlh = nlmsg_put(skb, item->portid, 0, type, sizeof(*nfmsg), flags);
738 if (nlh == NULL)
739 goto nlmsg_failure;
740
741 nfmsg = nlmsg_data(nlh);
742 nfmsg->nfgen_family = nf_ct_l3num(ct);
743 nfmsg->version = NFNETLINK_V0;
744 nfmsg->res_id = 0;
745
746 zone = nf_ct_zone(ct);
747
748 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG);
749 if (!nest_parms)
750 goto nla_put_failure;
751 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
752 goto nla_put_failure;
753 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
754 NF_CT_ZONE_DIR_ORIG) < 0)
755 goto nla_put_failure;
756 nla_nest_end(skb, nest_parms);
757
758 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY);
759 if (!nest_parms)
760 goto nla_put_failure;
761 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0)
762 goto nla_put_failure;
763 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
764 NF_CT_ZONE_DIR_REPL) < 0)
765 goto nla_put_failure;
766 nla_nest_end(skb, nest_parms);
767
768 if (ctnetlink_dump_zone_id(skb, CTA_ZONE, zone,
769 NF_CT_DEFAULT_ZONE_DIR) < 0)
770 goto nla_put_failure;
771
772 if (ctnetlink_dump_id(skb, ct) < 0)
773 goto nla_put_failure;
774
775 if (ctnetlink_dump_status(skb, ct) < 0)
776 goto nla_put_failure;
777
778 if (events & (1 << IPCT_DESTROY)) {
779 if (ctnetlink_dump_acct(skb, ct, type) < 0 ||
780 ctnetlink_dump_timestamp(skb, ct) < 0)
781 goto nla_put_failure;
782 } else {
783 if (ctnetlink_dump_timeout(skb, ct) < 0)
784 goto nla_put_failure;
785
786 if (events & (1 << IPCT_PROTOINFO)
787 && ctnetlink_dump_protoinfo(skb, ct) < 0)
788 goto nla_put_failure;
789
790 if ((events & (1 << IPCT_HELPER) || nfct_help(ct))
791 && ctnetlink_dump_helpinfo(skb, ct) < 0)
792 goto nla_put_failure;
793
794 #ifdef CONFIG_NF_CONNTRACK_SECMARK
795 if ((events & (1 << IPCT_SECMARK) || ct->secmark)
796 && ctnetlink_dump_secctx(skb, ct) < 0)
797 goto nla_put_failure;
798 #endif
799 if (events & (1 << IPCT_LABEL) &&
800 ctnetlink_dump_labels(skb, ct) < 0)
801 goto nla_put_failure;
802
803 if (events & (1 << IPCT_RELATED) &&
804 ctnetlink_dump_master(skb, ct) < 0)
805 goto nla_put_failure;
806
807 if (events & (1 << IPCT_SEQADJ) &&
808 ctnetlink_dump_ct_seq_adj(skb, ct) < 0)
809 goto nla_put_failure;
810
811 if (events & (1 << IPCT_SYNPROXY) &&
812 ctnetlink_dump_ct_synproxy(skb, ct) < 0)
813 goto nla_put_failure;
814 }
815
816 #ifdef CONFIG_NF_CONNTRACK_MARK
817 if ((events & (1 << IPCT_MARK) || ct->mark)
818 && ctnetlink_dump_mark(skb, ct) < 0)
819 goto nla_put_failure;
820 #endif
821 nlmsg_end(skb, nlh);
822 err = nfnetlink_send(skb, net, item->portid, group, item->report,
823 GFP_ATOMIC);
824 if (err == -ENOBUFS || err == -EAGAIN)
825 return -ENOBUFS;
826
827 return 0;
828
829 nla_put_failure:
830 nlmsg_cancel(skb, nlh);
831 nlmsg_failure:
832 kfree_skb(skb);
833 errout:
834 if (nfnetlink_set_err(net, 0, group, -ENOBUFS) > 0)
835 return -ENOBUFS;
836
837 return 0;
838 }
839 #endif /* CONFIG_NF_CONNTRACK_EVENTS */
840
841 static int ctnetlink_done(struct netlink_callback *cb)
842 {
843 if (cb->args[1])
844 nf_ct_put((struct nf_conn *)cb->args[1]);
845 kfree(cb->data);
846 return 0;
847 }
848
849 struct ctnetlink_filter {
850 u8 family;
851 struct {
852 u_int32_t val;
853 u_int32_t mask;
854 } mark;
855 };
856
857 static struct ctnetlink_filter *
858 ctnetlink_alloc_filter(const struct nlattr * const cda[], u8 family)
859 {
860 struct ctnetlink_filter *filter;
861
862 #ifndef CONFIG_NF_CONNTRACK_MARK
863 if (cda[CTA_MARK] && cda[CTA_MARK_MASK])
864 return ERR_PTR(-EOPNOTSUPP);
865 #endif
866
867 filter = kzalloc(sizeof(*filter), GFP_KERNEL);
868 if (filter == NULL)
869 return ERR_PTR(-ENOMEM);
870
871 filter->family = family;
872
873 #ifdef CONFIG_NF_CONNTRACK_MARK
874 if (cda[CTA_MARK] && cda[CTA_MARK_MASK]) {
875 filter->mark.val = ntohl(nla_get_be32(cda[CTA_MARK]));
876 filter->mark.mask = ntohl(nla_get_be32(cda[CTA_MARK_MASK]));
877 }
878 #endif
879 return filter;
880 }
881
882 static int ctnetlink_start(struct netlink_callback *cb)
883 {
884 const struct nlattr * const *cda = cb->data;
885 struct ctnetlink_filter *filter = NULL;
886 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
887 u8 family = nfmsg->nfgen_family;
888
889 if (family || (cda[CTA_MARK] && cda[CTA_MARK_MASK])) {
890 filter = ctnetlink_alloc_filter(cda, family);
891 if (IS_ERR(filter))
892 return PTR_ERR(filter);
893 }
894
895 cb->data = filter;
896 return 0;
897 }
898
899 static int ctnetlink_filter_match(struct nf_conn *ct, void *data)
900 {
901 struct ctnetlink_filter *filter = data;
902
903 if (filter == NULL)
904 goto out;
905
906 /* Match entries of a given L3 protocol number.
907 * If it is not specified, ie. l3proto == 0,
908 * then match everything.
909 */
910 if (filter->family && nf_ct_l3num(ct) != filter->family)
911 goto ignore_entry;
912
913 #ifdef CONFIG_NF_CONNTRACK_MARK
914 if ((ct->mark & filter->mark.mask) != filter->mark.val)
915 goto ignore_entry;
916 #endif
917
918 out:
919 return 1;
920
921 ignore_entry:
922 return 0;
923 }
924
925 static int
926 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
927 {
928 struct net *net = sock_net(skb->sk);
929 struct nf_conn *ct, *last;
930 struct nf_conntrack_tuple_hash *h;
931 struct hlist_nulls_node *n;
932 struct nf_conn *nf_ct_evict[8];
933 int res, i;
934 spinlock_t *lockp;
935
936 last = (struct nf_conn *)cb->args[1];
937 i = 0;
938
939 local_bh_disable();
940 for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
941 restart:
942 while (i) {
943 i--;
944 if (nf_ct_should_gc(nf_ct_evict[i]))
945 nf_ct_kill(nf_ct_evict[i]);
946 nf_ct_put(nf_ct_evict[i]);
947 }
948
949 lockp = &nf_conntrack_locks[cb->args[0] % CONNTRACK_LOCKS];
950 nf_conntrack_lock(lockp);
951 if (cb->args[0] >= nf_conntrack_htable_size) {
952 spin_unlock(lockp);
953 goto out;
954 }
955 hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[cb->args[0]],
956 hnnode) {
957 if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
958 continue;
959 ct = nf_ct_tuplehash_to_ctrack(h);
960 if (nf_ct_is_expired(ct)) {
961 if (i < ARRAY_SIZE(nf_ct_evict) &&
962 atomic_inc_not_zero(&ct->ct_general.use))
963 nf_ct_evict[i++] = ct;
964 continue;
965 }
966
967 if (!net_eq(net, nf_ct_net(ct)))
968 continue;
969
970 if (cb->args[1]) {
971 if (ct != last)
972 continue;
973 cb->args[1] = 0;
974 }
975 if (!ctnetlink_filter_match(ct, cb->data))
976 continue;
977
978 res =
979 ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).portid,
980 cb->nlh->nlmsg_seq,
981 NFNL_MSG_TYPE(cb->nlh->nlmsg_type),
982 ct, true);
983 if (res < 0) {
984 nf_conntrack_get(&ct->ct_general);
985 cb->args[1] = (unsigned long)ct;
986 spin_unlock(lockp);
987 goto out;
988 }
989 }
990 spin_unlock(lockp);
991 if (cb->args[1]) {
992 cb->args[1] = 0;
993 goto restart;
994 }
995 }
996 out:
997 local_bh_enable();
998 if (last) {
999 /* nf ct hash resize happened, now clear the leftover. */
1000 if ((struct nf_conn *)cb->args[1] == last)
1001 cb->args[1] = 0;
1002
1003 nf_ct_put(last);
1004 }
1005
1006 while (i) {
1007 i--;
1008 if (nf_ct_should_gc(nf_ct_evict[i]))
1009 nf_ct_kill(nf_ct_evict[i]);
1010 nf_ct_put(nf_ct_evict[i]);
1011 }
1012
1013 return skb->len;
1014 }
1015
1016 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
1017 struct nf_conntrack_tuple *t)
1018 {
1019 if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
1020 return -EINVAL;
1021
1022 t->src.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_SRC]);
1023 t->dst.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_DST]);
1024
1025 return 0;
1026 }
1027
1028 static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
1029 struct nf_conntrack_tuple *t)
1030 {
1031 if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST])
1032 return -EINVAL;
1033
1034 t->src.u3.in6 = nla_get_in6_addr(tb[CTA_IP_V6_SRC]);
1035 t->dst.u3.in6 = nla_get_in6_addr(tb[CTA_IP_V6_DST]);
1036
1037 return 0;
1038 }
1039
1040 static int ctnetlink_parse_tuple_ip(struct nlattr *attr,
1041 struct nf_conntrack_tuple *tuple)
1042 {
1043 struct nlattr *tb[CTA_IP_MAX+1];
1044 int ret = 0;
1045
1046 ret = nla_parse_nested_deprecated(tb, CTA_IP_MAX, attr, NULL, NULL);
1047 if (ret < 0)
1048 return ret;
1049
1050 ret = nla_validate_nested_deprecated(attr, CTA_IP_MAX,
1051 cta_ip_nla_policy, NULL);
1052 if (ret)
1053 return ret;
1054
1055 switch (tuple->src.l3num) {
1056 case NFPROTO_IPV4:
1057 ret = ipv4_nlattr_to_tuple(tb, tuple);
1058 break;
1059 case NFPROTO_IPV6:
1060 ret = ipv6_nlattr_to_tuple(tb, tuple);
1061 break;
1062 }
1063
1064 return ret;
1065 }
1066
1067 static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = {
1068 [CTA_PROTO_NUM] = { .type = NLA_U8 },
1069 };
1070
1071 static int ctnetlink_parse_tuple_proto(struct nlattr *attr,
1072 struct nf_conntrack_tuple *tuple)
1073 {
1074 const struct nf_conntrack_l4proto *l4proto;
1075 struct nlattr *tb[CTA_PROTO_MAX+1];
1076 int ret = 0;
1077
1078 ret = nla_parse_nested_deprecated(tb, CTA_PROTO_MAX, attr,
1079 proto_nla_policy, NULL);
1080 if (ret < 0)
1081 return ret;
1082
1083 if (!tb[CTA_PROTO_NUM])
1084 return -EINVAL;
1085 tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
1086
1087 rcu_read_lock();
1088 l4proto = nf_ct_l4proto_find(tuple->dst.protonum);
1089
1090 if (likely(l4proto->nlattr_to_tuple)) {
1091 ret = nla_validate_nested_deprecated(attr, CTA_PROTO_MAX,
1092 l4proto->nla_policy,
1093 NULL);
1094 if (ret == 0)
1095 ret = l4proto->nlattr_to_tuple(tb, tuple);
1096 }
1097
1098 rcu_read_unlock();
1099
1100 return ret;
1101 }
1102
1103 static int
1104 ctnetlink_parse_zone(const struct nlattr *attr,
1105 struct nf_conntrack_zone *zone)
1106 {
1107 nf_ct_zone_init(zone, NF_CT_DEFAULT_ZONE_ID,
1108 NF_CT_DEFAULT_ZONE_DIR, 0);
1109 #ifdef CONFIG_NF_CONNTRACK_ZONES
1110 if (attr)
1111 zone->id = ntohs(nla_get_be16(attr));
1112 #else
1113 if (attr)
1114 return -EOPNOTSUPP;
1115 #endif
1116 return 0;
1117 }
1118
1119 static int
1120 ctnetlink_parse_tuple_zone(struct nlattr *attr, enum ctattr_type type,
1121 struct nf_conntrack_zone *zone)
1122 {
1123 int ret;
1124
1125 if (zone->id != NF_CT_DEFAULT_ZONE_ID)
1126 return -EINVAL;
1127
1128 ret = ctnetlink_parse_zone(attr, zone);
1129 if (ret < 0)
1130 return ret;
1131
1132 if (type == CTA_TUPLE_REPLY)
1133 zone->dir = NF_CT_ZONE_DIR_REPL;
1134 else
1135 zone->dir = NF_CT_ZONE_DIR_ORIG;
1136
1137 return 0;
1138 }
1139
1140 static const struct nla_policy tuple_nla_policy[CTA_TUPLE_MAX+1] = {
1141 [CTA_TUPLE_IP] = { .type = NLA_NESTED },
1142 [CTA_TUPLE_PROTO] = { .type = NLA_NESTED },
1143 [CTA_TUPLE_ZONE] = { .type = NLA_U16 },
1144 };
1145
1146 static int
1147 ctnetlink_parse_tuple(const struct nlattr * const cda[],
1148 struct nf_conntrack_tuple *tuple, u32 type,
1149 u_int8_t l3num, struct nf_conntrack_zone *zone)
1150 {
1151 struct nlattr *tb[CTA_TUPLE_MAX+1];
1152 int err;
1153
1154 memset(tuple, 0, sizeof(*tuple));
1155
1156 err = nla_parse_nested_deprecated(tb, CTA_TUPLE_MAX, cda[type],
1157 tuple_nla_policy, NULL);
1158 if (err < 0)
1159 return err;
1160
1161 if (!tb[CTA_TUPLE_IP])
1162 return -EINVAL;
1163
1164 tuple->src.l3num = l3num;
1165
1166 err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple);
1167 if (err < 0)
1168 return err;
1169
1170 if (!tb[CTA_TUPLE_PROTO])
1171 return -EINVAL;
1172
1173 err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple);
1174 if (err < 0)
1175 return err;
1176
1177 if (tb[CTA_TUPLE_ZONE]) {
1178 if (!zone)
1179 return -EINVAL;
1180
1181 err = ctnetlink_parse_tuple_zone(tb[CTA_TUPLE_ZONE],
1182 type, zone);
1183 if (err < 0)
1184 return err;
1185 }
1186
1187 /* orig and expect tuples get DIR_ORIGINAL */
1188 if (type == CTA_TUPLE_REPLY)
1189 tuple->dst.dir = IP_CT_DIR_REPLY;
1190 else
1191 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
1192
1193 return 0;
1194 }
1195
1196 static const struct nla_policy help_nla_policy[CTA_HELP_MAX+1] = {
1197 [CTA_HELP_NAME] = { .type = NLA_NUL_STRING,
1198 .len = NF_CT_HELPER_NAME_LEN - 1 },
1199 };
1200
1201 static int ctnetlink_parse_help(const struct nlattr *attr, char **helper_name,
1202 struct nlattr **helpinfo)
1203 {
1204 int err;
1205 struct nlattr *tb[CTA_HELP_MAX+1];
1206
1207 err = nla_parse_nested_deprecated(tb, CTA_HELP_MAX, attr,
1208 help_nla_policy, NULL);
1209 if (err < 0)
1210 return err;
1211
1212 if (!tb[CTA_HELP_NAME])
1213 return -EINVAL;
1214
1215 *helper_name = nla_data(tb[CTA_HELP_NAME]);
1216
1217 if (tb[CTA_HELP_INFO])
1218 *helpinfo = tb[CTA_HELP_INFO];
1219
1220 return 0;
1221 }
1222
1223 static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
1224 [CTA_TUPLE_ORIG] = { .type = NLA_NESTED },
1225 [CTA_TUPLE_REPLY] = { .type = NLA_NESTED },
1226 [CTA_STATUS] = { .type = NLA_U32 },
1227 [CTA_PROTOINFO] = { .type = NLA_NESTED },
1228 [CTA_HELP] = { .type = NLA_NESTED },
1229 [CTA_NAT_SRC] = { .type = NLA_NESTED },
1230 [CTA_TIMEOUT] = { .type = NLA_U32 },
1231 [CTA_MARK] = { .type = NLA_U32 },
1232 [CTA_ID] = { .type = NLA_U32 },
1233 [CTA_NAT_DST] = { .type = NLA_NESTED },
1234 [CTA_TUPLE_MASTER] = { .type = NLA_NESTED },
1235 [CTA_NAT_SEQ_ADJ_ORIG] = { .type = NLA_NESTED },
1236 [CTA_NAT_SEQ_ADJ_REPLY] = { .type = NLA_NESTED },
1237 [CTA_ZONE] = { .type = NLA_U16 },
1238 [CTA_MARK_MASK] = { .type = NLA_U32 },
1239 [CTA_LABELS] = { .type = NLA_BINARY,
1240 .len = NF_CT_LABELS_MAX_SIZE },
1241 [CTA_LABELS_MASK] = { .type = NLA_BINARY,
1242 .len = NF_CT_LABELS_MAX_SIZE },
1243 };
1244
1245 static int ctnetlink_flush_iterate(struct nf_conn *ct, void *data)
1246 {
1247 if (test_bit(IPS_OFFLOAD_BIT, &ct->status))
1248 return 0;
1249
1250 return ctnetlink_filter_match(ct, data);
1251 }
1252
1253 static int ctnetlink_flush_conntrack(struct net *net,
1254 const struct nlattr * const cda[],
1255 u32 portid, int report, u8 family)
1256 {
1257 struct ctnetlink_filter *filter = NULL;
1258
1259 if (family || (cda[CTA_MARK] && cda[CTA_MARK_MASK])) {
1260 filter = ctnetlink_alloc_filter(cda, family);
1261 if (IS_ERR(filter))
1262 return PTR_ERR(filter);
1263 }
1264
1265 nf_ct_iterate_cleanup_net(net, ctnetlink_flush_iterate, filter,
1266 portid, report);
1267 kfree(filter);
1268
1269 return 0;
1270 }
1271
1272 static int ctnetlink_del_conntrack(struct net *net, struct sock *ctnl,
1273 struct sk_buff *skb,
1274 const struct nlmsghdr *nlh,
1275 const struct nlattr * const cda[],
1276 struct netlink_ext_ack *extack)
1277 {
1278 struct nf_conntrack_tuple_hash *h;
1279 struct nf_conntrack_tuple tuple;
1280 struct nf_conn *ct;
1281 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1282 struct nf_conntrack_zone zone;
1283 int err;
1284
1285 err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone);
1286 if (err < 0)
1287 return err;
1288
1289 if (cda[CTA_TUPLE_ORIG])
1290 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG,
1291 nfmsg->nfgen_family, &zone);
1292 else if (cda[CTA_TUPLE_REPLY])
1293 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY,
1294 nfmsg->nfgen_family, &zone);
1295 else {
1296 u_int8_t u3 = nfmsg->version ? nfmsg->nfgen_family : AF_UNSPEC;
1297
1298 return ctnetlink_flush_conntrack(net, cda,
1299 NETLINK_CB(skb).portid,
1300 nlmsg_report(nlh), u3);
1301 }
1302
1303 if (err < 0)
1304 return err;
1305
1306 h = nf_conntrack_find_get(net, &zone, &tuple);
1307 if (!h)
1308 return -ENOENT;
1309
1310 ct = nf_ct_tuplehash_to_ctrack(h);
1311
1312 if (test_bit(IPS_OFFLOAD_BIT, &ct->status)) {
1313 nf_ct_put(ct);
1314 return -EBUSY;
1315 }
1316
1317 if (cda[CTA_ID]) {
1318 __be32 id = nla_get_be32(cda[CTA_ID]);
1319
1320 if (id != (__force __be32)nf_ct_get_id(ct)) {
1321 nf_ct_put(ct);
1322 return -ENOENT;
1323 }
1324 }
1325
1326 nf_ct_delete(ct, NETLINK_CB(skb).portid, nlmsg_report(nlh));
1327 nf_ct_put(ct);
1328
1329 return 0;
1330 }
1331
1332 static int ctnetlink_get_conntrack(struct net *net, struct sock *ctnl,
1333 struct sk_buff *skb,
1334 const struct nlmsghdr *nlh,
1335 const struct nlattr * const cda[],
1336 struct netlink_ext_ack *extack)
1337 {
1338 struct nf_conntrack_tuple_hash *h;
1339 struct nf_conntrack_tuple tuple;
1340 struct nf_conn *ct;
1341 struct sk_buff *skb2 = NULL;
1342 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1343 u_int8_t u3 = nfmsg->nfgen_family;
1344 struct nf_conntrack_zone zone;
1345 int err;
1346
1347 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1348 struct netlink_dump_control c = {
1349 .start = ctnetlink_start,
1350 .dump = ctnetlink_dump_table,
1351 .done = ctnetlink_done,
1352 .data = (void *)cda,
1353 };
1354
1355 return netlink_dump_start(ctnl, skb, nlh, &c);
1356 }
1357
1358 err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone);
1359 if (err < 0)
1360 return err;
1361
1362 if (cda[CTA_TUPLE_ORIG])
1363 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG,
1364 u3, &zone);
1365 else if (cda[CTA_TUPLE_REPLY])
1366 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY,
1367 u3, &zone);
1368 else
1369 return -EINVAL;
1370
1371 if (err < 0)
1372 return err;
1373
1374 h = nf_conntrack_find_get(net, &zone, &tuple);
1375 if (!h)
1376 return -ENOENT;
1377
1378 ct = nf_ct_tuplehash_to_ctrack(h);
1379
1380 err = -ENOMEM;
1381 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1382 if (skb2 == NULL) {
1383 nf_ct_put(ct);
1384 return -ENOMEM;
1385 }
1386
1387 err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).portid, nlh->nlmsg_seq,
1388 NFNL_MSG_TYPE(nlh->nlmsg_type), ct, true);
1389 nf_ct_put(ct);
1390 if (err <= 0)
1391 goto free;
1392
1393 err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
1394 if (err < 0)
1395 goto out;
1396
1397 return 0;
1398
1399 free:
1400 kfree_skb(skb2);
1401 out:
1402 /* this avoids a loop in nfnetlink. */
1403 return err == -EAGAIN ? -ENOBUFS : err;
1404 }
1405
1406 static int ctnetlink_done_list(struct netlink_callback *cb)
1407 {
1408 if (cb->args[1])
1409 nf_ct_put((struct nf_conn *)cb->args[1]);
1410 return 0;
1411 }
1412
1413 static int
1414 ctnetlink_dump_list(struct sk_buff *skb, struct netlink_callback *cb, bool dying)
1415 {
1416 struct nf_conn *ct, *last;
1417 struct nf_conntrack_tuple_hash *h;
1418 struct hlist_nulls_node *n;
1419 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
1420 u_int8_t l3proto = nfmsg->nfgen_family;
1421 int res;
1422 int cpu;
1423 struct hlist_nulls_head *list;
1424 struct net *net = sock_net(skb->sk);
1425
1426 if (cb->args[2])
1427 return 0;
1428
1429 last = (struct nf_conn *)cb->args[1];
1430
1431 for (cpu = cb->args[0]; cpu < nr_cpu_ids; cpu++) {
1432 struct ct_pcpu *pcpu;
1433
1434 if (!cpu_possible(cpu))
1435 continue;
1436
1437 pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu);
1438 spin_lock_bh(&pcpu->lock);
1439 list = dying ? &pcpu->dying : &pcpu->unconfirmed;
1440 restart:
1441 hlist_nulls_for_each_entry(h, n, list, hnnode) {
1442 ct = nf_ct_tuplehash_to_ctrack(h);
1443 if (l3proto && nf_ct_l3num(ct) != l3proto)
1444 continue;
1445 if (cb->args[1]) {
1446 if (ct != last)
1447 continue;
1448 cb->args[1] = 0;
1449 }
1450
1451 /* We can't dump extension info for the unconfirmed
1452 * list because unconfirmed conntracks can have
1453 * ct->ext reallocated (and thus freed).
1454 *
1455 * In the dying list case ct->ext can't be free'd
1456 * until after we drop pcpu->lock.
1457 */
1458 res = ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).portid,
1459 cb->nlh->nlmsg_seq,
1460 NFNL_MSG_TYPE(cb->nlh->nlmsg_type),
1461 ct, dying ? true : false);
1462 if (res < 0) {
1463 if (!atomic_inc_not_zero(&ct->ct_general.use))
1464 continue;
1465 cb->args[0] = cpu;
1466 cb->args[1] = (unsigned long)ct;
1467 spin_unlock_bh(&pcpu->lock);
1468 goto out;
1469 }
1470 }
1471 if (cb->args[1]) {
1472 cb->args[1] = 0;
1473 goto restart;
1474 }
1475 spin_unlock_bh(&pcpu->lock);
1476 }
1477 cb->args[2] = 1;
1478 out:
1479 if (last)
1480 nf_ct_put(last);
1481
1482 return skb->len;
1483 }
1484
1485 static int
1486 ctnetlink_dump_dying(struct sk_buff *skb, struct netlink_callback *cb)
1487 {
1488 return ctnetlink_dump_list(skb, cb, true);
1489 }
1490
1491 static int ctnetlink_get_ct_dying(struct net *net, struct sock *ctnl,
1492 struct sk_buff *skb,
1493 const struct nlmsghdr *nlh,
1494 const struct nlattr * const cda[],
1495 struct netlink_ext_ack *extack)
1496 {
1497 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1498 struct netlink_dump_control c = {
1499 .dump = ctnetlink_dump_dying,
1500 .done = ctnetlink_done_list,
1501 };
1502 return netlink_dump_start(ctnl, skb, nlh, &c);
1503 }
1504
1505 return -EOPNOTSUPP;
1506 }
1507
1508 static int
1509 ctnetlink_dump_unconfirmed(struct sk_buff *skb, struct netlink_callback *cb)
1510 {
1511 return ctnetlink_dump_list(skb, cb, false);
1512 }
1513
1514 static int ctnetlink_get_ct_unconfirmed(struct net *net, struct sock *ctnl,
1515 struct sk_buff *skb,
1516 const struct nlmsghdr *nlh,
1517 const struct nlattr * const cda[],
1518 struct netlink_ext_ack *extack)
1519 {
1520 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1521 struct netlink_dump_control c = {
1522 .dump = ctnetlink_dump_unconfirmed,
1523 .done = ctnetlink_done_list,
1524 };
1525 return netlink_dump_start(ctnl, skb, nlh, &c);
1526 }
1527
1528 return -EOPNOTSUPP;
1529 }
1530
1531 #if IS_ENABLED(CONFIG_NF_NAT)
1532 static int
1533 ctnetlink_parse_nat_setup(struct nf_conn *ct,
1534 enum nf_nat_manip_type manip,
1535 const struct nlattr *attr)
1536 {
1537 struct nf_nat_hook *nat_hook;
1538 int err;
1539
1540 nat_hook = rcu_dereference(nf_nat_hook);
1541 if (!nat_hook) {
1542 #ifdef CONFIG_MODULES
1543 rcu_read_unlock();
1544 nfnl_unlock(NFNL_SUBSYS_CTNETLINK);
1545 if (request_module("nf-nat") < 0) {
1546 nfnl_lock(NFNL_SUBSYS_CTNETLINK);
1547 rcu_read_lock();
1548 return -EOPNOTSUPP;
1549 }
1550 nfnl_lock(NFNL_SUBSYS_CTNETLINK);
1551 rcu_read_lock();
1552 nat_hook = rcu_dereference(nf_nat_hook);
1553 if (nat_hook)
1554 return -EAGAIN;
1555 #endif
1556 return -EOPNOTSUPP;
1557 }
1558
1559 err = nat_hook->parse_nat_setup(ct, manip, attr);
1560 if (err == -EAGAIN) {
1561 #ifdef CONFIG_MODULES
1562 rcu_read_unlock();
1563 nfnl_unlock(NFNL_SUBSYS_CTNETLINK);
1564 if (request_module("nf-nat-%u", nf_ct_l3num(ct)) < 0) {
1565 nfnl_lock(NFNL_SUBSYS_CTNETLINK);
1566 rcu_read_lock();
1567 return -EOPNOTSUPP;
1568 }
1569 nfnl_lock(NFNL_SUBSYS_CTNETLINK);
1570 rcu_read_lock();
1571 #else
1572 err = -EOPNOTSUPP;
1573 #endif
1574 }
1575 return err;
1576 }
1577 #endif
1578
1579 static void
1580 __ctnetlink_change_status(struct nf_conn *ct, unsigned long on,
1581 unsigned long off)
1582 {
1583 unsigned int bit;
1584
1585 /* Ignore these unchangable bits */
1586 on &= ~IPS_UNCHANGEABLE_MASK;
1587 off &= ~IPS_UNCHANGEABLE_MASK;
1588
1589 for (bit = 0; bit < __IPS_MAX_BIT; bit++) {
1590 if (on & (1 << bit))
1591 set_bit(bit, &ct->status);
1592 else if (off & (1 << bit))
1593 clear_bit(bit, &ct->status);
1594 }
1595 }
1596
1597 static int
1598 ctnetlink_change_status(struct nf_conn *ct, const struct nlattr * const cda[])
1599 {
1600 unsigned long d;
1601 unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
1602 d = ct->status ^ status;
1603
1604 if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
1605 /* unchangeable */
1606 return -EBUSY;
1607
1608 if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
1609 /* SEEN_REPLY bit can only be set */
1610 return -EBUSY;
1611
1612 if (d & IPS_ASSURED && !(status & IPS_ASSURED))
1613 /* ASSURED bit can only be set */
1614 return -EBUSY;
1615
1616 __ctnetlink_change_status(ct, status, 0);
1617 return 0;
1618 }
1619
1620 static int
1621 ctnetlink_setup_nat(struct nf_conn *ct, const struct nlattr * const cda[])
1622 {
1623 #if IS_ENABLED(CONFIG_NF_NAT)
1624 int ret;
1625
1626 if (!cda[CTA_NAT_DST] && !cda[CTA_NAT_SRC])
1627 return 0;
1628
1629 ret = ctnetlink_parse_nat_setup(ct, NF_NAT_MANIP_DST,
1630 cda[CTA_NAT_DST]);
1631 if (ret < 0)
1632 return ret;
1633
1634 return ctnetlink_parse_nat_setup(ct, NF_NAT_MANIP_SRC,
1635 cda[CTA_NAT_SRC]);
1636 #else
1637 if (!cda[CTA_NAT_DST] && !cda[CTA_NAT_SRC])
1638 return 0;
1639 return -EOPNOTSUPP;
1640 #endif
1641 }
1642
1643 static int ctnetlink_change_helper(struct nf_conn *ct,
1644 const struct nlattr * const cda[])
1645 {
1646 struct nf_conntrack_helper *helper;
1647 struct nf_conn_help *help = nfct_help(ct);
1648 char *helpname = NULL;
1649 struct nlattr *helpinfo = NULL;
1650 int err;
1651
1652 err = ctnetlink_parse_help(cda[CTA_HELP], &helpname, &helpinfo);
1653 if (err < 0)
1654 return err;
1655
1656 /* don't change helper of sibling connections */
1657 if (ct->master) {
1658 /* If we try to change the helper to the same thing twice,
1659 * treat the second attempt as a no-op instead of returning
1660 * an error.
1661 */
1662 err = -EBUSY;
1663 if (help) {
1664 rcu_read_lock();
1665 helper = rcu_dereference(help->helper);
1666 if (helper && !strcmp(helper->name, helpname))
1667 err = 0;
1668 rcu_read_unlock();
1669 }
1670
1671 return err;
1672 }
1673
1674 if (!strcmp(helpname, "")) {
1675 if (help && help->helper) {
1676 /* we had a helper before ... */
1677 nf_ct_remove_expectations(ct);
1678 RCU_INIT_POINTER(help->helper, NULL);
1679 }
1680
1681 return 0;
1682 }
1683
1684 rcu_read_lock();
1685 helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct),
1686 nf_ct_protonum(ct));
1687 if (helper == NULL) {
1688 rcu_read_unlock();
1689 return -EOPNOTSUPP;
1690 }
1691
1692 if (help) {
1693 if (help->helper == helper) {
1694 /* update private helper data if allowed. */
1695 if (helper->from_nlattr)
1696 helper->from_nlattr(helpinfo, ct);
1697 err = 0;
1698 } else
1699 err = -EBUSY;
1700 } else {
1701 /* we cannot set a helper for an existing conntrack */
1702 err = -EOPNOTSUPP;
1703 }
1704
1705 rcu_read_unlock();
1706 return err;
1707 }
1708
1709 static int ctnetlink_change_timeout(struct nf_conn *ct,
1710 const struct nlattr * const cda[])
1711 {
1712 u64 timeout = (u64)ntohl(nla_get_be32(cda[CTA_TIMEOUT])) * HZ;
1713
1714 if (timeout > INT_MAX)
1715 timeout = INT_MAX;
1716 ct->timeout = nfct_time_stamp + (u32)timeout;
1717
1718 if (test_bit(IPS_DYING_BIT, &ct->status))
1719 return -ETIME;
1720
1721 return 0;
1722 }
1723
1724 #if defined(CONFIG_NF_CONNTRACK_MARK)
1725 static void ctnetlink_change_mark(struct nf_conn *ct,
1726 const struct nlattr * const cda[])
1727 {
1728 u32 mark, newmark, mask = 0;
1729
1730 if (cda[CTA_MARK_MASK])
1731 mask = ~ntohl(nla_get_be32(cda[CTA_MARK_MASK]));
1732
1733 mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1734 newmark = (ct->mark & mask) ^ mark;
1735 if (newmark != ct->mark)
1736 ct->mark = newmark;
1737 }
1738 #endif
1739
1740 static const struct nla_policy protoinfo_policy[CTA_PROTOINFO_MAX+1] = {
1741 [CTA_PROTOINFO_TCP] = { .type = NLA_NESTED },
1742 [CTA_PROTOINFO_DCCP] = { .type = NLA_NESTED },
1743 [CTA_PROTOINFO_SCTP] = { .type = NLA_NESTED },
1744 };
1745
1746 static int ctnetlink_change_protoinfo(struct nf_conn *ct,
1747 const struct nlattr * const cda[])
1748 {
1749 const struct nlattr *attr = cda[CTA_PROTOINFO];
1750 const struct nf_conntrack_l4proto *l4proto;
1751 struct nlattr *tb[CTA_PROTOINFO_MAX+1];
1752 int err = 0;
1753
1754 err = nla_parse_nested_deprecated(tb, CTA_PROTOINFO_MAX, attr,
1755 protoinfo_policy, NULL);
1756 if (err < 0)
1757 return err;
1758
1759 l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct));
1760 if (l4proto->from_nlattr)
1761 err = l4proto->from_nlattr(tb, ct);
1762
1763 return err;
1764 }
1765
1766 static const struct nla_policy seqadj_policy[CTA_SEQADJ_MAX+1] = {
1767 [CTA_SEQADJ_CORRECTION_POS] = { .type = NLA_U32 },
1768 [CTA_SEQADJ_OFFSET_BEFORE] = { .type = NLA_U32 },
1769 [CTA_SEQADJ_OFFSET_AFTER] = { .type = NLA_U32 },
1770 };
1771
1772 static int change_seq_adj(struct nf_ct_seqadj *seq,
1773 const struct nlattr * const attr)
1774 {
1775 int err;
1776 struct nlattr *cda[CTA_SEQADJ_MAX+1];
1777
1778 err = nla_parse_nested_deprecated(cda, CTA_SEQADJ_MAX, attr,
1779 seqadj_policy, NULL);
1780 if (err < 0)
1781 return err;
1782
1783 if (!cda[CTA_SEQADJ_CORRECTION_POS])
1784 return -EINVAL;
1785
1786 seq->correction_pos =
1787 ntohl(nla_get_be32(cda[CTA_SEQADJ_CORRECTION_POS]));
1788
1789 if (!cda[CTA_SEQADJ_OFFSET_BEFORE])
1790 return -EINVAL;
1791
1792 seq->offset_before =
1793 ntohl(nla_get_be32(cda[CTA_SEQADJ_OFFSET_BEFORE]));
1794
1795 if (!cda[CTA_SEQADJ_OFFSET_AFTER])
1796 return -EINVAL;
1797
1798 seq->offset_after =
1799 ntohl(nla_get_be32(cda[CTA_SEQADJ_OFFSET_AFTER]));
1800
1801 return 0;
1802 }
1803
1804 static int
1805 ctnetlink_change_seq_adj(struct nf_conn *ct,
1806 const struct nlattr * const cda[])
1807 {
1808 struct nf_conn_seqadj *seqadj = nfct_seqadj(ct);
1809 int ret = 0;
1810
1811 if (!seqadj)
1812 return 0;
1813
1814 spin_lock_bh(&ct->lock);
1815 if (cda[CTA_SEQ_ADJ_ORIG]) {
1816 ret = change_seq_adj(&seqadj->seq[IP_CT_DIR_ORIGINAL],
1817 cda[CTA_SEQ_ADJ_ORIG]);
1818 if (ret < 0)
1819 goto err;
1820
1821 set_bit(IPS_SEQ_ADJUST_BIT, &ct->status);
1822 }
1823
1824 if (cda[CTA_SEQ_ADJ_REPLY]) {
1825 ret = change_seq_adj(&seqadj->seq[IP_CT_DIR_REPLY],
1826 cda[CTA_SEQ_ADJ_REPLY]);
1827 if (ret < 0)
1828 goto err;
1829
1830 set_bit(IPS_SEQ_ADJUST_BIT, &ct->status);
1831 }
1832
1833 spin_unlock_bh(&ct->lock);
1834 return 0;
1835 err:
1836 spin_unlock_bh(&ct->lock);
1837 return ret;
1838 }
1839
1840 static const struct nla_policy synproxy_policy[CTA_SYNPROXY_MAX + 1] = {
1841 [CTA_SYNPROXY_ISN] = { .type = NLA_U32 },
1842 [CTA_SYNPROXY_ITS] = { .type = NLA_U32 },
1843 [CTA_SYNPROXY_TSOFF] = { .type = NLA_U32 },
1844 };
1845
1846 static int ctnetlink_change_synproxy(struct nf_conn *ct,
1847 const struct nlattr * const cda[])
1848 {
1849 struct nf_conn_synproxy *synproxy = nfct_synproxy(ct);
1850 struct nlattr *tb[CTA_SYNPROXY_MAX + 1];
1851 int err;
1852
1853 if (!synproxy)
1854 return 0;
1855
1856 err = nla_parse_nested_deprecated(tb, CTA_SYNPROXY_MAX,
1857 cda[CTA_SYNPROXY], synproxy_policy,
1858 NULL);
1859 if (err < 0)
1860 return err;
1861
1862 if (!tb[CTA_SYNPROXY_ISN] ||
1863 !tb[CTA_SYNPROXY_ITS] ||
1864 !tb[CTA_SYNPROXY_TSOFF])
1865 return -EINVAL;
1866
1867 synproxy->isn = ntohl(nla_get_be32(tb[CTA_SYNPROXY_ISN]));
1868 synproxy->its = ntohl(nla_get_be32(tb[CTA_SYNPROXY_ITS]));
1869 synproxy->tsoff = ntohl(nla_get_be32(tb[CTA_SYNPROXY_TSOFF]));
1870
1871 return 0;
1872 }
1873
1874 static int
1875 ctnetlink_attach_labels(struct nf_conn *ct, const struct nlattr * const cda[])
1876 {
1877 #ifdef CONFIG_NF_CONNTRACK_LABELS
1878 size_t len = nla_len(cda[CTA_LABELS]);
1879 const void *mask = cda[CTA_LABELS_MASK];
1880
1881 if (len & (sizeof(u32)-1)) /* must be multiple of u32 */
1882 return -EINVAL;
1883
1884 if (mask) {
1885 if (nla_len(cda[CTA_LABELS_MASK]) == 0 ||
1886 nla_len(cda[CTA_LABELS_MASK]) != len)
1887 return -EINVAL;
1888 mask = nla_data(cda[CTA_LABELS_MASK]);
1889 }
1890
1891 len /= sizeof(u32);
1892
1893 return nf_connlabels_replace(ct, nla_data(cda[CTA_LABELS]), mask, len);
1894 #else
1895 return -EOPNOTSUPP;
1896 #endif
1897 }
1898
1899 static int
1900 ctnetlink_change_conntrack(struct nf_conn *ct,
1901 const struct nlattr * const cda[])
1902 {
1903 int err;
1904
1905 /* only allow NAT changes and master assignation for new conntracks */
1906 if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST] || cda[CTA_TUPLE_MASTER])
1907 return -EOPNOTSUPP;
1908
1909 if (cda[CTA_HELP]) {
1910 err = ctnetlink_change_helper(ct, cda);
1911 if (err < 0)
1912 return err;
1913 }
1914
1915 if (cda[CTA_TIMEOUT]) {
1916 err = ctnetlink_change_timeout(ct, cda);
1917 if (err < 0)
1918 return err;
1919 }
1920
1921 if (cda[CTA_STATUS]) {
1922 err = ctnetlink_change_status(ct, cda);
1923 if (err < 0)
1924 return err;
1925 }
1926
1927 if (cda[CTA_PROTOINFO]) {
1928 err = ctnetlink_change_protoinfo(ct, cda);
1929 if (err < 0)
1930 return err;
1931 }
1932
1933 #if defined(CONFIG_NF_CONNTRACK_MARK)
1934 if (cda[CTA_MARK])
1935 ctnetlink_change_mark(ct, cda);
1936 #endif
1937
1938 if (cda[CTA_SEQ_ADJ_ORIG] || cda[CTA_SEQ_ADJ_REPLY]) {
1939 err = ctnetlink_change_seq_adj(ct, cda);
1940 if (err < 0)
1941 return err;
1942 }
1943
1944 if (cda[CTA_SYNPROXY]) {
1945 err = ctnetlink_change_synproxy(ct, cda);
1946 if (err < 0)
1947 return err;
1948 }
1949
1950 if (cda[CTA_LABELS]) {
1951 err = ctnetlink_attach_labels(ct, cda);
1952 if (err < 0)
1953 return err;
1954 }
1955
1956 return 0;
1957 }
1958
1959 static struct nf_conn *
1960 ctnetlink_create_conntrack(struct net *net,
1961 const struct nf_conntrack_zone *zone,
1962 const struct nlattr * const cda[],
1963 struct nf_conntrack_tuple *otuple,
1964 struct nf_conntrack_tuple *rtuple,
1965 u8 u3)
1966 {
1967 struct nf_conn *ct;
1968 int err = -EINVAL;
1969 struct nf_conntrack_helper *helper;
1970 struct nf_conn_tstamp *tstamp;
1971 u64 timeout;
1972
1973 ct = nf_conntrack_alloc(net, zone, otuple, rtuple, GFP_ATOMIC);
1974 if (IS_ERR(ct))
1975 return ERR_PTR(-ENOMEM);
1976
1977 if (!cda[CTA_TIMEOUT])
1978 goto err1;
1979
1980 timeout = (u64)ntohl(nla_get_be32(cda[CTA_TIMEOUT])) * HZ;
1981 if (timeout > INT_MAX)
1982 timeout = INT_MAX;
1983 ct->timeout = (u32)timeout + nfct_time_stamp;
1984
1985 rcu_read_lock();
1986 if (cda[CTA_HELP]) {
1987 char *helpname = NULL;
1988 struct nlattr *helpinfo = NULL;
1989
1990 err = ctnetlink_parse_help(cda[CTA_HELP], &helpname, &helpinfo);
1991 if (err < 0)
1992 goto err2;
1993
1994 helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct),
1995 nf_ct_protonum(ct));
1996 if (helper == NULL) {
1997 rcu_read_unlock();
1998 #ifdef CONFIG_MODULES
1999 if (request_module("nfct-helper-%s", helpname) < 0) {
2000 err = -EOPNOTSUPP;
2001 goto err1;
2002 }
2003
2004 rcu_read_lock();
2005 helper = __nf_conntrack_helper_find(helpname,
2006 nf_ct_l3num(ct),
2007 nf_ct_protonum(ct));
2008 if (helper) {
2009 err = -EAGAIN;
2010 goto err2;
2011 }
2012 rcu_read_unlock();
2013 #endif
2014 err = -EOPNOTSUPP;
2015 goto err1;
2016 } else {
2017 struct nf_conn_help *help;
2018
2019 help = nf_ct_helper_ext_add(ct, GFP_ATOMIC);
2020 if (help == NULL) {
2021 err = -ENOMEM;
2022 goto err2;
2023 }
2024 /* set private helper data if allowed. */
2025 if (helper->from_nlattr)
2026 helper->from_nlattr(helpinfo, ct);
2027
2028 /* not in hash table yet so not strictly necessary */
2029 RCU_INIT_POINTER(help->helper, helper);
2030 }
2031 } else {
2032 /* try an implicit helper assignation */
2033 err = __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC);
2034 if (err < 0)
2035 goto err2;
2036 }
2037
2038 err = ctnetlink_setup_nat(ct, cda);
2039 if (err < 0)
2040 goto err2;
2041
2042 nf_ct_acct_ext_add(ct, GFP_ATOMIC);
2043 nf_ct_tstamp_ext_add(ct, GFP_ATOMIC);
2044 nf_ct_ecache_ext_add(ct, 0, 0, GFP_ATOMIC);
2045 nf_ct_labels_ext_add(ct);
2046 nfct_seqadj_ext_add(ct);
2047 nfct_synproxy_ext_add(ct);
2048
2049 /* we must add conntrack extensions before confirmation. */
2050 ct->status |= IPS_CONFIRMED;
2051
2052 if (cda[CTA_STATUS]) {
2053 err = ctnetlink_change_status(ct, cda);
2054 if (err < 0)
2055 goto err2;
2056 }
2057
2058 if (cda[CTA_SEQ_ADJ_ORIG] || cda[CTA_SEQ_ADJ_REPLY]) {
2059 err = ctnetlink_change_seq_adj(ct, cda);
2060 if (err < 0)
2061 goto err2;
2062 }
2063
2064 memset(&ct->proto, 0, sizeof(ct->proto));
2065 if (cda[CTA_PROTOINFO]) {
2066 err = ctnetlink_change_protoinfo(ct, cda);
2067 if (err < 0)
2068 goto err2;
2069 }
2070
2071 if (cda[CTA_SYNPROXY]) {
2072 err = ctnetlink_change_synproxy(ct, cda);
2073 if (err < 0)
2074 goto err2;
2075 }
2076
2077 #if defined(CONFIG_NF_CONNTRACK_MARK)
2078 if (cda[CTA_MARK])
2079 ctnetlink_change_mark(ct, cda);
2080 #endif
2081
2082 /* setup master conntrack: this is a confirmed expectation */
2083 if (cda[CTA_TUPLE_MASTER]) {
2084 struct nf_conntrack_tuple master;
2085 struct nf_conntrack_tuple_hash *master_h;
2086 struct nf_conn *master_ct;
2087
2088 err = ctnetlink_parse_tuple(cda, &master, CTA_TUPLE_MASTER,
2089 u3, NULL);
2090 if (err < 0)
2091 goto err2;
2092
2093 master_h = nf_conntrack_find_get(net, zone, &master);
2094 if (master_h == NULL) {
2095 err = -ENOENT;
2096 goto err2;
2097 }
2098 master_ct = nf_ct_tuplehash_to_ctrack(master_h);
2099 __set_bit(IPS_EXPECTED_BIT, &ct->status);
2100 ct->master = master_ct;
2101 }
2102 tstamp = nf_conn_tstamp_find(ct);
2103 if (tstamp)
2104 tstamp->start = ktime_get_real_ns();
2105
2106 err = nf_conntrack_hash_check_insert(ct);
2107 if (err < 0)
2108 goto err2;
2109
2110 rcu_read_unlock();
2111
2112 return ct;
2113
2114 err2:
2115 rcu_read_unlock();
2116 err1:
2117 nf_conntrack_free(ct);
2118 return ERR_PTR(err);
2119 }
2120
2121 static int ctnetlink_new_conntrack(struct net *net, struct sock *ctnl,
2122 struct sk_buff *skb,
2123 const struct nlmsghdr *nlh,
2124 const struct nlattr * const cda[],
2125 struct netlink_ext_ack *extack)
2126 {
2127 struct nf_conntrack_tuple otuple, rtuple;
2128 struct nf_conntrack_tuple_hash *h = NULL;
2129 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
2130 struct nf_conn *ct;
2131 u_int8_t u3 = nfmsg->nfgen_family;
2132 struct nf_conntrack_zone zone;
2133 int err;
2134
2135 err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone);
2136 if (err < 0)
2137 return err;
2138
2139 if (cda[CTA_TUPLE_ORIG]) {
2140 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG,
2141 u3, &zone);
2142 if (err < 0)
2143 return err;
2144 }
2145
2146 if (cda[CTA_TUPLE_REPLY]) {
2147 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY,
2148 u3, &zone);
2149 if (err < 0)
2150 return err;
2151 }
2152
2153 if (cda[CTA_TUPLE_ORIG])
2154 h = nf_conntrack_find_get(net, &zone, &otuple);
2155 else if (cda[CTA_TUPLE_REPLY])
2156 h = nf_conntrack_find_get(net, &zone, &rtuple);
2157
2158 if (h == NULL) {
2159 err = -ENOENT;
2160 if (nlh->nlmsg_flags & NLM_F_CREATE) {
2161 enum ip_conntrack_events events;
2162
2163 if (!cda[CTA_TUPLE_ORIG] || !cda[CTA_TUPLE_REPLY])
2164 return -EINVAL;
2165 if (otuple.dst.protonum != rtuple.dst.protonum)
2166 return -EINVAL;
2167
2168 ct = ctnetlink_create_conntrack(net, &zone, cda, &otuple,
2169 &rtuple, u3);
2170 if (IS_ERR(ct))
2171 return PTR_ERR(ct);
2172
2173 err = 0;
2174 if (test_bit(IPS_EXPECTED_BIT, &ct->status))
2175 events = 1 << IPCT_RELATED;
2176 else
2177 events = 1 << IPCT_NEW;
2178
2179 if (cda[CTA_LABELS] &&
2180 ctnetlink_attach_labels(ct, cda) == 0)
2181 events |= (1 << IPCT_LABEL);
2182
2183 nf_conntrack_eventmask_report((1 << IPCT_REPLY) |
2184 (1 << IPCT_ASSURED) |
2185 (1 << IPCT_HELPER) |
2186 (1 << IPCT_PROTOINFO) |
2187 (1 << IPCT_SEQADJ) |
2188 (1 << IPCT_MARK) |
2189 (1 << IPCT_SYNPROXY) |
2190 events,
2191 ct, NETLINK_CB(skb).portid,
2192 nlmsg_report(nlh));
2193 nf_ct_put(ct);
2194 }
2195
2196 return err;
2197 }
2198 /* implicit 'else' */
2199
2200 err = -EEXIST;
2201 ct = nf_ct_tuplehash_to_ctrack(h);
2202 if (!(nlh->nlmsg_flags & NLM_F_EXCL)) {
2203 err = ctnetlink_change_conntrack(ct, cda);
2204 if (err == 0) {
2205 nf_conntrack_eventmask_report((1 << IPCT_REPLY) |
2206 (1 << IPCT_ASSURED) |
2207 (1 << IPCT_HELPER) |
2208 (1 << IPCT_LABEL) |
2209 (1 << IPCT_PROTOINFO) |
2210 (1 << IPCT_SEQADJ) |
2211 (1 << IPCT_MARK) |
2212 (1 << IPCT_SYNPROXY),
2213 ct, NETLINK_CB(skb).portid,
2214 nlmsg_report(nlh));
2215 }
2216 }
2217
2218 nf_ct_put(ct);
2219 return err;
2220 }
2221
2222 static int
2223 ctnetlink_ct_stat_cpu_fill_info(struct sk_buff *skb, u32 portid, u32 seq,
2224 __u16 cpu, const struct ip_conntrack_stat *st)
2225 {
2226 struct nlmsghdr *nlh;
2227 struct nfgenmsg *nfmsg;
2228 unsigned int flags = portid ? NLM_F_MULTI : 0, event;
2229
2230 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK,
2231 IPCTNL_MSG_CT_GET_STATS_CPU);
2232 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*nfmsg), flags);
2233 if (nlh == NULL)
2234 goto nlmsg_failure;
2235
2236 nfmsg = nlmsg_data(nlh);
2237 nfmsg->nfgen_family = AF_UNSPEC;
2238 nfmsg->version = NFNETLINK_V0;
2239 nfmsg->res_id = htons(cpu);
2240
2241 if (nla_put_be32(skb, CTA_STATS_FOUND, htonl(st->found)) ||
2242 nla_put_be32(skb, CTA_STATS_INVALID, htonl(st->invalid)) ||
2243 nla_put_be32(skb, CTA_STATS_IGNORE, htonl(st->ignore)) ||
2244 nla_put_be32(skb, CTA_STATS_INSERT, htonl(st->insert)) ||
2245 nla_put_be32(skb, CTA_STATS_INSERT_FAILED,
2246 htonl(st->insert_failed)) ||
2247 nla_put_be32(skb, CTA_STATS_DROP, htonl(st->drop)) ||
2248 nla_put_be32(skb, CTA_STATS_EARLY_DROP, htonl(st->early_drop)) ||
2249 nla_put_be32(skb, CTA_STATS_ERROR, htonl(st->error)) ||
2250 nla_put_be32(skb, CTA_STATS_SEARCH_RESTART,
2251 htonl(st->search_restart)))
2252 goto nla_put_failure;
2253
2254 nlmsg_end(skb, nlh);
2255 return skb->len;
2256
2257 nla_put_failure:
2258 nlmsg_failure:
2259 nlmsg_cancel(skb, nlh);
2260 return -1;
2261 }
2262
2263 static int
2264 ctnetlink_ct_stat_cpu_dump(struct sk_buff *skb, struct netlink_callback *cb)
2265 {
2266 int cpu;
2267 struct net *net = sock_net(skb->sk);
2268
2269 if (cb->args[0] == nr_cpu_ids)
2270 return 0;
2271
2272 for (cpu = cb->args[0]; cpu < nr_cpu_ids; cpu++) {
2273 const struct ip_conntrack_stat *st;
2274
2275 if (!cpu_possible(cpu))
2276 continue;
2277
2278 st = per_cpu_ptr(net->ct.stat, cpu);
2279 if (ctnetlink_ct_stat_cpu_fill_info(skb,
2280 NETLINK_CB(cb->skb).portid,
2281 cb->nlh->nlmsg_seq,
2282 cpu, st) < 0)
2283 break;
2284 }
2285 cb->args[0] = cpu;
2286
2287 return skb->len;
2288 }
2289
2290 static int ctnetlink_stat_ct_cpu(struct net *net, struct sock *ctnl,
2291 struct sk_buff *skb,
2292 const struct nlmsghdr *nlh,
2293 const struct nlattr * const cda[],
2294 struct netlink_ext_ack *extack)
2295 {
2296 if (nlh->nlmsg_flags & NLM_F_DUMP) {
2297 struct netlink_dump_control c = {
2298 .dump = ctnetlink_ct_stat_cpu_dump,
2299 };
2300 return netlink_dump_start(ctnl, skb, nlh, &c);
2301 }
2302
2303 return 0;
2304 }
2305
2306 static int
2307 ctnetlink_stat_ct_fill_info(struct sk_buff *skb, u32 portid, u32 seq, u32 type,
2308 struct net *net)
2309 {
2310 struct nlmsghdr *nlh;
2311 struct nfgenmsg *nfmsg;
2312 unsigned int flags = portid ? NLM_F_MULTI : 0, event;
2313 unsigned int nr_conntracks = atomic_read(&net->ct.count);
2314
2315 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, IPCTNL_MSG_CT_GET_STATS);
2316 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*nfmsg), flags);
2317 if (nlh == NULL)
2318 goto nlmsg_failure;
2319
2320 nfmsg = nlmsg_data(nlh);
2321 nfmsg->nfgen_family = AF_UNSPEC;
2322 nfmsg->version = NFNETLINK_V0;
2323 nfmsg->res_id = 0;
2324
2325 if (nla_put_be32(skb, CTA_STATS_GLOBAL_ENTRIES, htonl(nr_conntracks)))
2326 goto nla_put_failure;
2327
2328 if (nla_put_be32(skb, CTA_STATS_GLOBAL_MAX_ENTRIES, htonl(nf_conntrack_max)))
2329 goto nla_put_failure;
2330
2331 nlmsg_end(skb, nlh);
2332 return skb->len;
2333
2334 nla_put_failure:
2335 nlmsg_failure:
2336 nlmsg_cancel(skb, nlh);
2337 return -1;
2338 }
2339
2340 static int ctnetlink_stat_ct(struct net *net, struct sock *ctnl,
2341 struct sk_buff *skb, const struct nlmsghdr *nlh,
2342 const struct nlattr * const cda[],
2343 struct netlink_ext_ack *extack)
2344 {
2345 struct sk_buff *skb2;
2346 int err;
2347
2348 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
2349 if (skb2 == NULL)
2350 return -ENOMEM;
2351
2352 err = ctnetlink_stat_ct_fill_info(skb2, NETLINK_CB(skb).portid,
2353 nlh->nlmsg_seq,
2354 NFNL_MSG_TYPE(nlh->nlmsg_type),
2355 sock_net(skb->sk));
2356 if (err <= 0)
2357 goto free;
2358
2359 err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
2360 if (err < 0)
2361 goto out;
2362
2363 return 0;
2364
2365 free:
2366 kfree_skb(skb2);
2367 out:
2368 /* this avoids a loop in nfnetlink. */
2369 return err == -EAGAIN ? -ENOBUFS : err;
2370 }
2371
2372 static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = {
2373 [CTA_EXPECT_MASTER] = { .type = NLA_NESTED },
2374 [CTA_EXPECT_TUPLE] = { .type = NLA_NESTED },
2375 [CTA_EXPECT_MASK] = { .type = NLA_NESTED },
2376 [CTA_EXPECT_TIMEOUT] = { .type = NLA_U32 },
2377 [CTA_EXPECT_ID] = { .type = NLA_U32 },
2378 [CTA_EXPECT_HELP_NAME] = { .type = NLA_NUL_STRING,
2379 .len = NF_CT_HELPER_NAME_LEN - 1 },
2380 [CTA_EXPECT_ZONE] = { .type = NLA_U16 },
2381 [CTA_EXPECT_FLAGS] = { .type = NLA_U32 },
2382 [CTA_EXPECT_CLASS] = { .type = NLA_U32 },
2383 [CTA_EXPECT_NAT] = { .type = NLA_NESTED },
2384 [CTA_EXPECT_FN] = { .type = NLA_NUL_STRING },
2385 };
2386
2387 static struct nf_conntrack_expect *
2388 ctnetlink_alloc_expect(const struct nlattr *const cda[], struct nf_conn *ct,
2389 struct nf_conntrack_helper *helper,
2390 struct nf_conntrack_tuple *tuple,
2391 struct nf_conntrack_tuple *mask);
2392
2393 #ifdef CONFIG_NETFILTER_NETLINK_GLUE_CT
2394 static size_t
2395 ctnetlink_glue_build_size(const struct nf_conn *ct)
2396 {
2397 return 3 * nla_total_size(0) /* CTA_TUPLE_ORIG|REPL|MASTER */
2398 + 3 * nla_total_size(0) /* CTA_TUPLE_IP */
2399 + 3 * nla_total_size(0) /* CTA_TUPLE_PROTO */
2400 + 3 * nla_total_size(sizeof(u_int8_t)) /* CTA_PROTO_NUM */
2401 + nla_total_size(sizeof(u_int32_t)) /* CTA_ID */
2402 + nla_total_size(sizeof(u_int32_t)) /* CTA_STATUS */
2403 + nla_total_size(sizeof(u_int32_t)) /* CTA_TIMEOUT */
2404 + nla_total_size(0) /* CTA_PROTOINFO */
2405 + nla_total_size(0) /* CTA_HELP */
2406 + nla_total_size(NF_CT_HELPER_NAME_LEN) /* CTA_HELP_NAME */
2407 + ctnetlink_secctx_size(ct)
2408 #if IS_ENABLED(CONFIG_NF_NAT)
2409 + 2 * nla_total_size(0) /* CTA_NAT_SEQ_ADJ_ORIG|REPL */
2410 + 6 * nla_total_size(sizeof(u_int32_t)) /* CTA_NAT_SEQ_OFFSET */
2411 #endif
2412 #ifdef CONFIG_NF_CONNTRACK_MARK
2413 + nla_total_size(sizeof(u_int32_t)) /* CTA_MARK */
2414 #endif
2415 #ifdef CONFIG_NF_CONNTRACK_ZONES
2416 + nla_total_size(sizeof(u_int16_t)) /* CTA_ZONE|CTA_TUPLE_ZONE */
2417 #endif
2418 + ctnetlink_proto_size(ct)
2419 ;
2420 }
2421
2422 static struct nf_conn *ctnetlink_glue_get_ct(const struct sk_buff *skb,
2423 enum ip_conntrack_info *ctinfo)
2424 {
2425 return nf_ct_get(skb, ctinfo);
2426 }
2427
2428 static int __ctnetlink_glue_build(struct sk_buff *skb, struct nf_conn *ct)
2429 {
2430 const struct nf_conntrack_zone *zone;
2431 struct nlattr *nest_parms;
2432
2433 zone = nf_ct_zone(ct);
2434
2435 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG);
2436 if (!nest_parms)
2437 goto nla_put_failure;
2438 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
2439 goto nla_put_failure;
2440 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
2441 NF_CT_ZONE_DIR_ORIG) < 0)
2442 goto nla_put_failure;
2443 nla_nest_end(skb, nest_parms);
2444
2445 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY);
2446 if (!nest_parms)
2447 goto nla_put_failure;
2448 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0)
2449 goto nla_put_failure;
2450 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone,
2451 NF_CT_ZONE_DIR_REPL) < 0)
2452 goto nla_put_failure;
2453 nla_nest_end(skb, nest_parms);
2454
2455 if (ctnetlink_dump_zone_id(skb, CTA_ZONE, zone,
2456 NF_CT_DEFAULT_ZONE_DIR) < 0)
2457 goto nla_put_failure;
2458
2459 if (ctnetlink_dump_id(skb, ct) < 0)
2460 goto nla_put_failure;
2461
2462 if (ctnetlink_dump_status(skb, ct) < 0)
2463 goto nla_put_failure;
2464
2465 if (ctnetlink_dump_timeout(skb, ct) < 0)
2466 goto nla_put_failure;
2467
2468 if (ctnetlink_dump_protoinfo(skb, ct) < 0)
2469 goto nla_put_failure;
2470
2471 if (ctnetlink_dump_helpinfo(skb, ct) < 0)
2472 goto nla_put_failure;
2473
2474 #ifdef CONFIG_NF_CONNTRACK_SECMARK
2475 if (ct->secmark && ctnetlink_dump_secctx(skb, ct) < 0)
2476 goto nla_put_failure;
2477 #endif
2478 if (ct->master && ctnetlink_dump_master(skb, ct) < 0)
2479 goto nla_put_failure;
2480
2481 if ((ct->status & IPS_SEQ_ADJUST) &&
2482 ctnetlink_dump_ct_seq_adj(skb, ct) < 0)
2483 goto nla_put_failure;
2484
2485 if (ctnetlink_dump_ct_synproxy(skb, ct) < 0)
2486 goto nla_put_failure;
2487
2488 #ifdef CONFIG_NF_CONNTRACK_MARK
2489 if (ct->mark && ctnetlink_dump_mark(skb, ct) < 0)
2490 goto nla_put_failure;
2491 #endif
2492 if (ctnetlink_dump_labels(skb, ct) < 0)
2493 goto nla_put_failure;
2494 return 0;
2495
2496 nla_put_failure:
2497 return -ENOSPC;
2498 }
2499
2500 static int
2501 ctnetlink_glue_build(struct sk_buff *skb, struct nf_conn *ct,
2502 enum ip_conntrack_info ctinfo,
2503 u_int16_t ct_attr, u_int16_t ct_info_attr)
2504 {
2505 struct nlattr *nest_parms;
2506
2507 nest_parms = nla_nest_start(skb, ct_attr);
2508 if (!nest_parms)
2509 goto nla_put_failure;
2510
2511 if (__ctnetlink_glue_build(skb, ct) < 0)
2512 goto nla_put_failure;
2513
2514 nla_nest_end(skb, nest_parms);
2515
2516 if (nla_put_be32(skb, ct_info_attr, htonl(ctinfo)))
2517 goto nla_put_failure;
2518
2519 return 0;
2520
2521 nla_put_failure:
2522 return -ENOSPC;
2523 }
2524
2525 static int
2526 ctnetlink_update_status(struct nf_conn *ct, const struct nlattr * const cda[])
2527 {
2528 unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
2529 unsigned long d = ct->status ^ status;
2530
2531 if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
2532 /* SEEN_REPLY bit can only be set */
2533 return -EBUSY;
2534
2535 if (d & IPS_ASSURED && !(status & IPS_ASSURED))
2536 /* ASSURED bit can only be set */
2537 return -EBUSY;
2538
2539 /* This check is less strict than ctnetlink_change_status()
2540 * because callers often flip IPS_EXPECTED bits when sending
2541 * an NFQA_CT attribute to the kernel. So ignore the
2542 * unchangeable bits but do not error out. Also user programs
2543 * are allowed to clear the bits that they are allowed to change.
2544 */
2545 __ctnetlink_change_status(ct, status, ~status);
2546 return 0;
2547 }
2548
2549 static int
2550 ctnetlink_glue_parse_ct(const struct nlattr *cda[], struct nf_conn *ct)
2551 {
2552 int err;
2553
2554 if (cda[CTA_TIMEOUT]) {
2555 err = ctnetlink_change_timeout(ct, cda);
2556 if (err < 0)
2557 return err;
2558 }
2559 if (cda[CTA_STATUS]) {
2560 err = ctnetlink_update_status(ct, cda);
2561 if (err < 0)
2562 return err;
2563 }
2564 if (cda[CTA_HELP]) {
2565 err = ctnetlink_change_helper(ct, cda);
2566 if (err < 0)
2567 return err;
2568 }
2569 if (cda[CTA_LABELS]) {
2570 err = ctnetlink_attach_labels(ct, cda);
2571 if (err < 0)
2572 return err;
2573 }
2574 #if defined(CONFIG_NF_CONNTRACK_MARK)
2575 if (cda[CTA_MARK]) {
2576 ctnetlink_change_mark(ct, cda);
2577 }
2578 #endif
2579 return 0;
2580 }
2581
2582 static int
2583 ctnetlink_glue_parse(const struct nlattr *attr, struct nf_conn *ct)
2584 {
2585 struct nlattr *cda[CTA_MAX+1];
2586 int ret;
2587
2588 ret = nla_parse_nested_deprecated(cda, CTA_MAX, attr, ct_nla_policy,
2589 NULL);
2590 if (ret < 0)
2591 return ret;
2592
2593 return ctnetlink_glue_parse_ct((const struct nlattr **)cda, ct);
2594 }
2595
2596 static int ctnetlink_glue_exp_parse(const struct nlattr * const *cda,
2597 const struct nf_conn *ct,
2598 struct nf_conntrack_tuple *tuple,
2599 struct nf_conntrack_tuple *mask)
2600 {
2601 int err;
2602
2603 err = ctnetlink_parse_tuple(cda, tuple, CTA_EXPECT_TUPLE,
2604 nf_ct_l3num(ct), NULL);
2605 if (err < 0)
2606 return err;
2607
2608 return ctnetlink_parse_tuple(cda, mask, CTA_EXPECT_MASK,
2609 nf_ct_l3num(ct), NULL);
2610 }
2611
2612 static int
2613 ctnetlink_glue_attach_expect(const struct nlattr *attr, struct nf_conn *ct,
2614 u32 portid, u32 report)
2615 {
2616 struct nlattr *cda[CTA_EXPECT_MAX+1];
2617 struct nf_conntrack_tuple tuple, mask;
2618 struct nf_conntrack_helper *helper = NULL;
2619 struct nf_conntrack_expect *exp;
2620 int err;
2621
2622 err = nla_parse_nested_deprecated(cda, CTA_EXPECT_MAX, attr,
2623 exp_nla_policy, NULL);
2624 if (err < 0)
2625 return err;
2626
2627 err = ctnetlink_glue_exp_parse((const struct nlattr * const *)cda,
2628 ct, &tuple, &mask);
2629 if (err < 0)
2630 return err;
2631
2632 if (cda[CTA_EXPECT_HELP_NAME]) {
2633 const char *helpname = nla_data(cda[CTA_EXPECT_HELP_NAME]);
2634
2635 helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct),
2636 nf_ct_protonum(ct));
2637 if (helper == NULL)
2638 return -EOPNOTSUPP;
2639 }
2640
2641 exp = ctnetlink_alloc_expect((const struct nlattr * const *)cda, ct,
2642 helper, &tuple, &mask);
2643 if (IS_ERR(exp))
2644 return PTR_ERR(exp);
2645
2646 err = nf_ct_expect_related_report(exp, portid, report, 0);
2647 nf_ct_expect_put(exp);
2648 return err;
2649 }
2650
2651 static void ctnetlink_glue_seqadj(struct sk_buff *skb, struct nf_conn *ct,
2652 enum ip_conntrack_info ctinfo, int diff)
2653 {
2654 if (!(ct->status & IPS_NAT_MASK))
2655 return;
2656
2657 nf_ct_tcp_seqadj_set(skb, ct, ctinfo, diff);
2658 }
2659
2660 static struct nfnl_ct_hook ctnetlink_glue_hook = {
2661 .get_ct = ctnetlink_glue_get_ct,
2662 .build_size = ctnetlink_glue_build_size,
2663 .build = ctnetlink_glue_build,
2664 .parse = ctnetlink_glue_parse,
2665 .attach_expect = ctnetlink_glue_attach_expect,
2666 .seq_adjust = ctnetlink_glue_seqadj,
2667 };
2668 #endif /* CONFIG_NETFILTER_NETLINK_GLUE_CT */
2669
2670 /***********************************************************************
2671 * EXPECT
2672 ***********************************************************************/
2673
2674 static int ctnetlink_exp_dump_tuple(struct sk_buff *skb,
2675 const struct nf_conntrack_tuple *tuple,
2676 u32 type)
2677 {
2678 struct nlattr *nest_parms;
2679
2680 nest_parms = nla_nest_start(skb, type);
2681 if (!nest_parms)
2682 goto nla_put_failure;
2683 if (ctnetlink_dump_tuples(skb, tuple) < 0)
2684 goto nla_put_failure;
2685 nla_nest_end(skb, nest_parms);
2686
2687 return 0;
2688
2689 nla_put_failure:
2690 return -1;
2691 }
2692
2693 static int ctnetlink_exp_dump_mask(struct sk_buff *skb,
2694 const struct nf_conntrack_tuple *tuple,
2695 const struct nf_conntrack_tuple_mask *mask)
2696 {
2697 const struct nf_conntrack_l4proto *l4proto;
2698 struct nf_conntrack_tuple m;
2699 struct nlattr *nest_parms;
2700 int ret;
2701
2702 memset(&m, 0xFF, sizeof(m));
2703 memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3));
2704 m.src.u.all = mask->src.u.all;
2705 m.dst.protonum = tuple->dst.protonum;
2706
2707 nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK);
2708 if (!nest_parms)
2709 goto nla_put_failure;
2710
2711 rcu_read_lock();
2712 ret = ctnetlink_dump_tuples_ip(skb, &m);
2713 if (ret >= 0) {
2714 l4proto = nf_ct_l4proto_find(tuple->dst.protonum);
2715 ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto);
2716 }
2717 rcu_read_unlock();
2718
2719 if (unlikely(ret < 0))
2720 goto nla_put_failure;
2721
2722 nla_nest_end(skb, nest_parms);
2723
2724 return 0;
2725
2726 nla_put_failure:
2727 return -1;
2728 }
2729
2730 static const union nf_inet_addr any_addr;
2731
2732 static __be32 nf_expect_get_id(const struct nf_conntrack_expect *exp)
2733 {
2734 static __read_mostly siphash_key_t exp_id_seed;
2735 unsigned long a, b, c, d;
2736
2737 net_get_random_once(&exp_id_seed, sizeof(exp_id_seed));
2738
2739 a = (unsigned long)exp;
2740 b = (unsigned long)exp->helper;
2741 c = (unsigned long)exp->master;
2742 d = (unsigned long)siphash(&exp->tuple, sizeof(exp->tuple), &exp_id_seed);
2743
2744 #ifdef CONFIG_64BIT
2745 return (__force __be32)siphash_4u64((u64)a, (u64)b, (u64)c, (u64)d, &exp_id_seed);
2746 #else
2747 return (__force __be32)siphash_4u32((u32)a, (u32)b, (u32)c, (u32)d, &exp_id_seed);
2748 #endif
2749 }
2750
2751 static int
2752 ctnetlink_exp_dump_expect(struct sk_buff *skb,
2753 const struct nf_conntrack_expect *exp)
2754 {
2755 struct nf_conn *master = exp->master;
2756 long timeout = ((long)exp->timeout.expires - (long)jiffies) / HZ;
2757 struct nf_conn_help *help;
2758 #if IS_ENABLED(CONFIG_NF_NAT)
2759 struct nlattr *nest_parms;
2760 struct nf_conntrack_tuple nat_tuple = {};
2761 #endif
2762 struct nf_ct_helper_expectfn *expfn;
2763
2764 if (timeout < 0)
2765 timeout = 0;
2766
2767 if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
2768 goto nla_put_failure;
2769 if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
2770 goto nla_put_failure;
2771 if (ctnetlink_exp_dump_tuple(skb,
2772 &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
2773 CTA_EXPECT_MASTER) < 0)
2774 goto nla_put_failure;
2775
2776 #if IS_ENABLED(CONFIG_NF_NAT)
2777 if (!nf_inet_addr_cmp(&exp->saved_addr, &any_addr) ||
2778 exp->saved_proto.all) {
2779 nest_parms = nla_nest_start(skb, CTA_EXPECT_NAT);
2780 if (!nest_parms)
2781 goto nla_put_failure;
2782
2783 if (nla_put_be32(skb, CTA_EXPECT_NAT_DIR, htonl(exp->dir)))
2784 goto nla_put_failure;
2785
2786 nat_tuple.src.l3num = nf_ct_l3num(master);
2787 nat_tuple.src.u3 = exp->saved_addr;
2788 nat_tuple.dst.protonum = nf_ct_protonum(master);
2789 nat_tuple.src.u = exp->saved_proto;
2790
2791 if (ctnetlink_exp_dump_tuple(skb, &nat_tuple,
2792 CTA_EXPECT_NAT_TUPLE) < 0)
2793 goto nla_put_failure;
2794 nla_nest_end(skb, nest_parms);
2795 }
2796 #endif
2797 if (nla_put_be32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout)) ||
2798 nla_put_be32(skb, CTA_EXPECT_ID, nf_expect_get_id(exp)) ||
2799 nla_put_be32(skb, CTA_EXPECT_FLAGS, htonl(exp->flags)) ||
2800 nla_put_be32(skb, CTA_EXPECT_CLASS, htonl(exp->class)))
2801 goto nla_put_failure;
2802 help = nfct_help(master);
2803 if (help) {
2804 struct nf_conntrack_helper *helper;
2805
2806 helper = rcu_dereference(help->helper);
2807 if (helper &&
2808 nla_put_string(skb, CTA_EXPECT_HELP_NAME, helper->name))
2809 goto nla_put_failure;
2810 }
2811 expfn = nf_ct_helper_expectfn_find_by_symbol(exp->expectfn);
2812 if (expfn != NULL &&
2813 nla_put_string(skb, CTA_EXPECT_FN, expfn->name))
2814 goto nla_put_failure;
2815
2816 return 0;
2817
2818 nla_put_failure:
2819 return -1;
2820 }
2821
2822 static int
2823 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 portid, u32 seq,
2824 int event, const struct nf_conntrack_expect *exp)
2825 {
2826 struct nlmsghdr *nlh;
2827 struct nfgenmsg *nfmsg;
2828 unsigned int flags = portid ? NLM_F_MULTI : 0;
2829
2830 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK_EXP, event);
2831 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*nfmsg), flags);
2832 if (nlh == NULL)
2833 goto nlmsg_failure;
2834
2835 nfmsg = nlmsg_data(nlh);
2836 nfmsg->nfgen_family = exp->tuple.src.l3num;
2837 nfmsg->version = NFNETLINK_V0;
2838 nfmsg->res_id = 0;
2839
2840 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
2841 goto nla_put_failure;
2842
2843 nlmsg_end(skb, nlh);
2844 return skb->len;
2845
2846 nlmsg_failure:
2847 nla_put_failure:
2848 nlmsg_cancel(skb, nlh);
2849 return -1;
2850 }
2851
2852 #ifdef CONFIG_NF_CONNTRACK_EVENTS
2853 static int
2854 ctnetlink_expect_event(unsigned int events, struct nf_exp_event *item)
2855 {
2856 struct nf_conntrack_expect *exp = item->exp;
2857 struct net *net = nf_ct_exp_net(exp);
2858 struct nlmsghdr *nlh;
2859 struct nfgenmsg *nfmsg;
2860 struct sk_buff *skb;
2861 unsigned int type, group;
2862 int flags = 0;
2863
2864 if (events & (1 << IPEXP_DESTROY)) {
2865 type = IPCTNL_MSG_EXP_DELETE;
2866 group = NFNLGRP_CONNTRACK_EXP_DESTROY;
2867 } else if (events & (1 << IPEXP_NEW)) {
2868 type = IPCTNL_MSG_EXP_NEW;
2869 flags = NLM_F_CREATE|NLM_F_EXCL;
2870 group = NFNLGRP_CONNTRACK_EXP_NEW;
2871 } else
2872 return 0;
2873
2874 if (!item->report && !nfnetlink_has_listeners(net, group))
2875 return 0;
2876
2877 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
2878 if (skb == NULL)
2879 goto errout;
2880
2881 type = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK_EXP, type);
2882 nlh = nlmsg_put(skb, item->portid, 0, type, sizeof(*nfmsg), flags);
2883 if (nlh == NULL)
2884 goto nlmsg_failure;
2885
2886 nfmsg = nlmsg_data(nlh);
2887 nfmsg->nfgen_family = exp->tuple.src.l3num;
2888 nfmsg->version = NFNETLINK_V0;
2889 nfmsg->res_id = 0;
2890
2891 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
2892 goto nla_put_failure;
2893
2894 nlmsg_end(skb, nlh);
2895 nfnetlink_send(skb, net, item->portid, group, item->report, GFP_ATOMIC);
2896 return 0;
2897
2898 nla_put_failure:
2899 nlmsg_cancel(skb, nlh);
2900 nlmsg_failure:
2901 kfree_skb(skb);
2902 errout:
2903 nfnetlink_set_err(net, 0, 0, -ENOBUFS);
2904 return 0;
2905 }
2906 #endif
2907 static int ctnetlink_exp_done(struct netlink_callback *cb)
2908 {
2909 if (cb->args[1])
2910 nf_ct_expect_put((struct nf_conntrack_expect *)cb->args[1]);
2911 return 0;
2912 }
2913
2914 static int
2915 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
2916 {
2917 struct net *net = sock_net(skb->sk);
2918 struct nf_conntrack_expect *exp, *last;
2919 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
2920 u_int8_t l3proto = nfmsg->nfgen_family;
2921
2922 rcu_read_lock();
2923 last = (struct nf_conntrack_expect *)cb->args[1];
2924 for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) {
2925 restart:
2926 hlist_for_each_entry_rcu(exp, &nf_ct_expect_hash[cb->args[0]],
2927 hnode) {
2928 if (l3proto && exp->tuple.src.l3num != l3proto)
2929 continue;
2930
2931 if (!net_eq(nf_ct_net(exp->master), net))
2932 continue;
2933
2934 if (cb->args[1]) {
2935 if (exp != last)
2936 continue;
2937 cb->args[1] = 0;
2938 }
2939 if (ctnetlink_exp_fill_info(skb,
2940 NETLINK_CB(cb->skb).portid,
2941 cb->nlh->nlmsg_seq,
2942 IPCTNL_MSG_EXP_NEW,
2943 exp) < 0) {
2944 if (!refcount_inc_not_zero(&exp->use))
2945 continue;
2946 cb->args[1] = (unsigned long)exp;
2947 goto out;
2948 }
2949 }
2950 if (cb->args[1]) {
2951 cb->args[1] = 0;
2952 goto restart;
2953 }
2954 }
2955 out:
2956 rcu_read_unlock();
2957 if (last)
2958 nf_ct_expect_put(last);
2959
2960 return skb->len;
2961 }
2962
2963 static int
2964 ctnetlink_exp_ct_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
2965 {
2966 struct nf_conntrack_expect *exp, *last;
2967 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
2968 struct nf_conn *ct = cb->data;
2969 struct nf_conn_help *help = nfct_help(ct);
2970 u_int8_t l3proto = nfmsg->nfgen_family;
2971
2972 if (cb->args[0])
2973 return 0;
2974
2975 rcu_read_lock();
2976 last = (struct nf_conntrack_expect *)cb->args[1];
2977 restart:
2978 hlist_for_each_entry_rcu(exp, &help->expectations, lnode) {
2979 if (l3proto && exp->tuple.src.l3num != l3proto)
2980 continue;
2981 if (cb->args[1]) {
2982 if (exp != last)
2983 continue;
2984 cb->args[1] = 0;
2985 }
2986 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).portid,
2987 cb->nlh->nlmsg_seq,
2988 IPCTNL_MSG_EXP_NEW,
2989 exp) < 0) {
2990 if (!refcount_inc_not_zero(&exp->use))
2991 continue;
2992 cb->args[1] = (unsigned long)exp;
2993 goto out;
2994 }
2995 }
2996 if (cb->args[1]) {
2997 cb->args[1] = 0;
2998 goto restart;
2999 }
3000 cb->args[0] = 1;
3001 out:
3002 rcu_read_unlock();
3003 if (last)
3004 nf_ct_expect_put(last);
3005
3006 return skb->len;
3007 }
3008
3009 static int ctnetlink_dump_exp_ct(struct net *net, struct sock *ctnl,
3010 struct sk_buff *skb,
3011 const struct nlmsghdr *nlh,
3012 const struct nlattr * const cda[],
3013 struct netlink_ext_ack *extack)
3014 {
3015 int err;
3016 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
3017 u_int8_t u3 = nfmsg->nfgen_family;
3018 struct nf_conntrack_tuple tuple;
3019 struct nf_conntrack_tuple_hash *h;
3020 struct nf_conn *ct;
3021 struct nf_conntrack_zone zone;
3022 struct netlink_dump_control c = {
3023 .dump = ctnetlink_exp_ct_dump_table,
3024 .done = ctnetlink_exp_done,
3025 };
3026
3027 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER,
3028 u3, NULL);
3029 if (err < 0)
3030 return err;
3031
3032 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
3033 if (err < 0)
3034 return err;
3035
3036 h = nf_conntrack_find_get(net, &zone, &tuple);
3037 if (!h)
3038 return -ENOENT;
3039
3040 ct = nf_ct_tuplehash_to_ctrack(h);
3041 /* No expectation linked to this connection tracking. */
3042 if (!nfct_help(ct)) {
3043 nf_ct_put(ct);
3044 return 0;
3045 }
3046
3047 c.data = ct;
3048
3049 err = netlink_dump_start(ctnl, skb, nlh, &c);
3050 nf_ct_put(ct);
3051
3052 return err;
3053 }
3054
3055 static int ctnetlink_get_expect(struct net *net, struct sock *ctnl,
3056 struct sk_buff *skb, const struct nlmsghdr *nlh,
3057 const struct nlattr * const cda[],
3058 struct netlink_ext_ack *extack)
3059 {
3060 struct nf_conntrack_tuple tuple;
3061 struct nf_conntrack_expect *exp;
3062 struct sk_buff *skb2;
3063 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
3064 u_int8_t u3 = nfmsg->nfgen_family;
3065 struct nf_conntrack_zone zone;
3066 int err;
3067
3068 if (nlh->nlmsg_flags & NLM_F_DUMP) {
3069 if (cda[CTA_EXPECT_MASTER])
3070 return ctnetlink_dump_exp_ct(net, ctnl, skb, nlh, cda,
3071 extack);
3072 else {
3073 struct netlink_dump_control c = {
3074 .dump = ctnetlink_exp_dump_table,
3075 .done = ctnetlink_exp_done,
3076 };
3077 return netlink_dump_start(ctnl, skb, nlh, &c);
3078 }
3079 }
3080
3081 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
3082 if (err < 0)
3083 return err;
3084
3085 if (cda[CTA_EXPECT_TUPLE])
3086 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE,
3087 u3, NULL);
3088 else if (cda[CTA_EXPECT_MASTER])
3089 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER,
3090 u3, NULL);
3091 else
3092 return -EINVAL;
3093
3094 if (err < 0)
3095 return err;
3096
3097 exp = nf_ct_expect_find_get(net, &zone, &tuple);
3098 if (!exp)
3099 return -ENOENT;
3100
3101 if (cda[CTA_EXPECT_ID]) {
3102 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
3103
3104 if (id != nf_expect_get_id(exp)) {
3105 nf_ct_expect_put(exp);
3106 return -ENOENT;
3107 }
3108 }
3109
3110 err = -ENOMEM;
3111 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3112 if (skb2 == NULL) {
3113 nf_ct_expect_put(exp);
3114 goto out;
3115 }
3116
3117 rcu_read_lock();
3118 err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).portid,
3119 nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW, exp);
3120 rcu_read_unlock();
3121 nf_ct_expect_put(exp);
3122 if (err <= 0)
3123 goto free;
3124
3125 err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).portid, MSG_DONTWAIT);
3126 if (err < 0)
3127 goto out;
3128
3129 return 0;
3130
3131 free:
3132 kfree_skb(skb2);
3133 out:
3134 /* this avoids a loop in nfnetlink. */
3135 return err == -EAGAIN ? -ENOBUFS : err;
3136 }
3137
3138 static bool expect_iter_name(struct nf_conntrack_expect *exp, void *data)
3139 {
3140 const struct nf_conn_help *m_help;
3141 const char *name = data;
3142
3143 m_help = nfct_help(exp->master);
3144
3145 return strcmp(m_help->helper->name, name) == 0;
3146 }
3147
3148 static bool expect_iter_all(struct nf_conntrack_expect *exp, void *data)
3149 {
3150 return true;
3151 }
3152
3153 static int ctnetlink_del_expect(struct net *net, struct sock *ctnl,
3154 struct sk_buff *skb, const struct nlmsghdr *nlh,
3155 const struct nlattr * const cda[],
3156 struct netlink_ext_ack *extack)
3157 {
3158 struct nf_conntrack_expect *exp;
3159 struct nf_conntrack_tuple tuple;
3160 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
3161 u_int8_t u3 = nfmsg->nfgen_family;
3162 struct nf_conntrack_zone zone;
3163 int err;
3164
3165 if (cda[CTA_EXPECT_TUPLE]) {
3166 /* delete a single expect by tuple */
3167 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
3168 if (err < 0)
3169 return err;
3170
3171 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE,
3172 u3, NULL);
3173 if (err < 0)
3174 return err;
3175
3176 /* bump usage count to 2 */
3177 exp = nf_ct_expect_find_get(net, &zone, &tuple);
3178 if (!exp)
3179 return -ENOENT;
3180
3181 if (cda[CTA_EXPECT_ID]) {
3182 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
3183 if (ntohl(id) != (u32)(unsigned long)exp) {
3184 nf_ct_expect_put(exp);
3185 return -ENOENT;
3186 }
3187 }
3188
3189 /* after list removal, usage count == 1 */
3190 spin_lock_bh(&nf_conntrack_expect_lock);
3191 if (del_timer(&exp->timeout)) {
3192 nf_ct_unlink_expect_report(exp, NETLINK_CB(skb).portid,
3193 nlmsg_report(nlh));
3194 nf_ct_expect_put(exp);
3195 }
3196 spin_unlock_bh(&nf_conntrack_expect_lock);
3197 /* have to put what we 'get' above.
3198 * after this line usage count == 0 */
3199 nf_ct_expect_put(exp);
3200 } else if (cda[CTA_EXPECT_HELP_NAME]) {
3201 char *name = nla_data(cda[CTA_EXPECT_HELP_NAME]);
3202
3203 nf_ct_expect_iterate_net(net, expect_iter_name, name,
3204 NETLINK_CB(skb).portid,
3205 nlmsg_report(nlh));
3206 } else {
3207 /* This basically means we have to flush everything*/
3208 nf_ct_expect_iterate_net(net, expect_iter_all, NULL,
3209 NETLINK_CB(skb).portid,
3210 nlmsg_report(nlh));
3211 }
3212
3213 return 0;
3214 }
3215 static int
3216 ctnetlink_change_expect(struct nf_conntrack_expect *x,
3217 const struct nlattr * const cda[])
3218 {
3219 if (cda[CTA_EXPECT_TIMEOUT]) {
3220 if (!del_timer(&x->timeout))
3221 return -ETIME;
3222
3223 x->timeout.expires = jiffies +
3224 ntohl(nla_get_be32(cda[CTA_EXPECT_TIMEOUT])) * HZ;
3225 add_timer(&x->timeout);
3226 }
3227 return 0;
3228 }
3229
3230 static const struct nla_policy exp_nat_nla_policy[CTA_EXPECT_NAT_MAX+1] = {
3231 [CTA_EXPECT_NAT_DIR] = { .type = NLA_U32 },
3232 [CTA_EXPECT_NAT_TUPLE] = { .type = NLA_NESTED },
3233 };
3234
3235 static int
3236 ctnetlink_parse_expect_nat(const struct nlattr *attr,
3237 struct nf_conntrack_expect *exp,
3238 u_int8_t u3)
3239 {
3240 #if IS_ENABLED(CONFIG_NF_NAT)
3241 struct nlattr *tb[CTA_EXPECT_NAT_MAX+1];
3242 struct nf_conntrack_tuple nat_tuple = {};
3243 int err;
3244
3245 err = nla_parse_nested_deprecated(tb, CTA_EXPECT_NAT_MAX, attr,
3246 exp_nat_nla_policy, NULL);
3247 if (err < 0)
3248 return err;
3249
3250 if (!tb[CTA_EXPECT_NAT_DIR] || !tb[CTA_EXPECT_NAT_TUPLE])
3251 return -EINVAL;
3252
3253 err = ctnetlink_parse_tuple((const struct nlattr * const *)tb,
3254 &nat_tuple, CTA_EXPECT_NAT_TUPLE,
3255 u3, NULL);
3256 if (err < 0)
3257 return err;
3258
3259 exp->saved_addr = nat_tuple.src.u3;
3260 exp->saved_proto = nat_tuple.src.u;
3261 exp->dir = ntohl(nla_get_be32(tb[CTA_EXPECT_NAT_DIR]));
3262
3263 return 0;
3264 #else
3265 return -EOPNOTSUPP;
3266 #endif
3267 }
3268
3269 static struct nf_conntrack_expect *
3270 ctnetlink_alloc_expect(const struct nlattr * const cda[], struct nf_conn *ct,
3271 struct nf_conntrack_helper *helper,
3272 struct nf_conntrack_tuple *tuple,
3273 struct nf_conntrack_tuple *mask)
3274 {
3275 u_int32_t class = 0;
3276 struct nf_conntrack_expect *exp;
3277 struct nf_conn_help *help;
3278 int err;
3279
3280 help = nfct_help(ct);
3281 if (!help)
3282 return ERR_PTR(-EOPNOTSUPP);
3283
3284 if (cda[CTA_EXPECT_CLASS] && helper) {
3285 class = ntohl(nla_get_be32(cda[CTA_EXPECT_CLASS]));
3286 if (class > helper->expect_class_max)
3287 return ERR_PTR(-EINVAL);
3288 }
3289 exp = nf_ct_expect_alloc(ct);
3290 if (!exp)
3291 return ERR_PTR(-ENOMEM);
3292
3293 if (cda[CTA_EXPECT_FLAGS]) {
3294 exp->flags = ntohl(nla_get_be32(cda[CTA_EXPECT_FLAGS]));
3295 exp->flags &= ~NF_CT_EXPECT_USERSPACE;
3296 } else {
3297 exp->flags = 0;
3298 }
3299 if (cda[CTA_EXPECT_FN]) {
3300 const char *name = nla_data(cda[CTA_EXPECT_FN]);
3301 struct nf_ct_helper_expectfn *expfn;
3302
3303 expfn = nf_ct_helper_expectfn_find_by_name(name);
3304 if (expfn == NULL) {
3305 err = -EINVAL;
3306 goto err_out;
3307 }
3308 exp->expectfn = expfn->expectfn;
3309 } else
3310 exp->expectfn = NULL;
3311
3312 exp->class = class;
3313 exp->master = ct;
3314 exp->helper = helper;
3315 exp->tuple = *tuple;
3316 exp->mask.src.u3 = mask->src.u3;
3317 exp->mask.src.u.all = mask->src.u.all;
3318
3319 if (cda[CTA_EXPECT_NAT]) {
3320 err = ctnetlink_parse_expect_nat(cda[CTA_EXPECT_NAT],
3321 exp, nf_ct_l3num(ct));
3322 if (err < 0)
3323 goto err_out;
3324 }
3325 return exp;
3326 err_out:
3327 nf_ct_expect_put(exp);
3328 return ERR_PTR(err);
3329 }
3330
3331 static int
3332 ctnetlink_create_expect(struct net *net,
3333 const struct nf_conntrack_zone *zone,
3334 const struct nlattr * const cda[],
3335 u_int8_t u3, u32 portid, int report)
3336 {
3337 struct nf_conntrack_tuple tuple, mask, master_tuple;
3338 struct nf_conntrack_tuple_hash *h = NULL;
3339 struct nf_conntrack_helper *helper = NULL;
3340 struct nf_conntrack_expect *exp;
3341 struct nf_conn *ct;
3342 int err;
3343
3344 /* caller guarantees that those three CTA_EXPECT_* exist */
3345 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE,
3346 u3, NULL);
3347 if (err < 0)
3348 return err;
3349 err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK,
3350 u3, NULL);
3351 if (err < 0)
3352 return err;
3353 err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER,
3354 u3, NULL);
3355 if (err < 0)
3356 return err;
3357
3358 /* Look for master conntrack of this expectation */
3359 h = nf_conntrack_find_get(net, zone, &master_tuple);
3360 if (!h)
3361 return -ENOENT;
3362 ct = nf_ct_tuplehash_to_ctrack(h);
3363
3364 rcu_read_lock();
3365 if (cda[CTA_EXPECT_HELP_NAME]) {
3366 const char *helpname = nla_data(cda[CTA_EXPECT_HELP_NAME]);
3367
3368 helper = __nf_conntrack_helper_find(helpname, u3,
3369 nf_ct_protonum(ct));
3370 if (helper == NULL) {
3371 rcu_read_unlock();
3372 #ifdef CONFIG_MODULES
3373 if (request_module("nfct-helper-%s", helpname) < 0) {
3374 err = -EOPNOTSUPP;
3375 goto err_ct;
3376 }
3377 rcu_read_lock();
3378 helper = __nf_conntrack_helper_find(helpname, u3,
3379 nf_ct_protonum(ct));
3380 if (helper) {
3381 err = -EAGAIN;
3382 goto err_rcu;
3383 }
3384 rcu_read_unlock();
3385 #endif
3386 err = -EOPNOTSUPP;
3387 goto err_ct;
3388 }
3389 }
3390
3391 exp = ctnetlink_alloc_expect(cda, ct, helper, &tuple, &mask);
3392 if (IS_ERR(exp)) {
3393 err = PTR_ERR(exp);
3394 goto err_rcu;
3395 }
3396
3397 err = nf_ct_expect_related_report(exp, portid, report, 0);
3398 nf_ct_expect_put(exp);
3399 err_rcu:
3400 rcu_read_unlock();
3401 err_ct:
3402 nf_ct_put(ct);
3403 return err;
3404 }
3405
3406 static int ctnetlink_new_expect(struct net *net, struct sock *ctnl,
3407 struct sk_buff *skb, const struct nlmsghdr *nlh,
3408 const struct nlattr * const cda[],
3409 struct netlink_ext_ack *extack)
3410 {
3411 struct nf_conntrack_tuple tuple;
3412 struct nf_conntrack_expect *exp;
3413 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
3414 u_int8_t u3 = nfmsg->nfgen_family;
3415 struct nf_conntrack_zone zone;
3416 int err;
3417
3418 if (!cda[CTA_EXPECT_TUPLE]
3419 || !cda[CTA_EXPECT_MASK]
3420 || !cda[CTA_EXPECT_MASTER])
3421 return -EINVAL;
3422
3423 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
3424 if (err < 0)
3425 return err;
3426
3427 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE,
3428 u3, NULL);
3429 if (err < 0)
3430 return err;
3431
3432 spin_lock_bh(&nf_conntrack_expect_lock);
3433 exp = __nf_ct_expect_find(net, &zone, &tuple);
3434 if (!exp) {
3435 spin_unlock_bh(&nf_conntrack_expect_lock);
3436 err = -ENOENT;
3437 if (nlh->nlmsg_flags & NLM_F_CREATE) {
3438 err = ctnetlink_create_expect(net, &zone, cda, u3,
3439 NETLINK_CB(skb).portid,
3440 nlmsg_report(nlh));
3441 }
3442 return err;
3443 }
3444
3445 err = -EEXIST;
3446 if (!(nlh->nlmsg_flags & NLM_F_EXCL))
3447 err = ctnetlink_change_expect(exp, cda);
3448 spin_unlock_bh(&nf_conntrack_expect_lock);
3449
3450 return err;
3451 }
3452
3453 static int
3454 ctnetlink_exp_stat_fill_info(struct sk_buff *skb, u32 portid, u32 seq, int cpu,
3455 const struct ip_conntrack_stat *st)
3456 {
3457 struct nlmsghdr *nlh;
3458 struct nfgenmsg *nfmsg;
3459 unsigned int flags = portid ? NLM_F_MULTI : 0, event;
3460
3461 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK,
3462 IPCTNL_MSG_EXP_GET_STATS_CPU);
3463 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*nfmsg), flags);
3464 if (nlh == NULL)
3465 goto nlmsg_failure;
3466
3467 nfmsg = nlmsg_data(nlh);
3468 nfmsg->nfgen_family = AF_UNSPEC;
3469 nfmsg->version = NFNETLINK_V0;
3470 nfmsg->res_id = htons(cpu);
3471
3472 if (nla_put_be32(skb, CTA_STATS_EXP_NEW, htonl(st->expect_new)) ||
3473 nla_put_be32(skb, CTA_STATS_EXP_CREATE, htonl(st->expect_create)) ||
3474 nla_put_be32(skb, CTA_STATS_EXP_DELETE, htonl(st->expect_delete)))
3475 goto nla_put_failure;
3476
3477 nlmsg_end(skb, nlh);
3478 return skb->len;
3479
3480 nla_put_failure:
3481 nlmsg_failure:
3482 nlmsg_cancel(skb, nlh);
3483 return -1;
3484 }
3485
3486 static int
3487 ctnetlink_exp_stat_cpu_dump(struct sk_buff *skb, struct netlink_callback *cb)
3488 {
3489 int cpu;
3490 struct net *net = sock_net(skb->sk);
3491
3492 if (cb->args[0] == nr_cpu_ids)
3493 return 0;
3494
3495 for (cpu = cb->args[0]; cpu < nr_cpu_ids; cpu++) {
3496 const struct ip_conntrack_stat *st;
3497
3498 if (!cpu_possible(cpu))
3499 continue;
3500
3501 st = per_cpu_ptr(net->ct.stat, cpu);
3502 if (ctnetlink_exp_stat_fill_info(skb, NETLINK_CB(cb->skb).portid,
3503 cb->nlh->nlmsg_seq,
3504 cpu, st) < 0)
3505 break;
3506 }
3507 cb->args[0] = cpu;
3508
3509 return skb->len;
3510 }
3511
3512 static int ctnetlink_stat_exp_cpu(struct net *net, struct sock *ctnl,
3513 struct sk_buff *skb,
3514 const struct nlmsghdr *nlh,
3515 const struct nlattr * const cda[],
3516 struct netlink_ext_ack *extack)
3517 {
3518 if (nlh->nlmsg_flags & NLM_F_DUMP) {
3519 struct netlink_dump_control c = {
3520 .dump = ctnetlink_exp_stat_cpu_dump,
3521 };
3522 return netlink_dump_start(ctnl, skb, nlh, &c);
3523 }
3524
3525 return 0;
3526 }
3527
3528 #ifdef CONFIG_NF_CONNTRACK_EVENTS
3529 static struct nf_ct_event_notifier ctnl_notifier = {
3530 .fcn = ctnetlink_conntrack_event,
3531 };
3532
3533 static struct nf_exp_event_notifier ctnl_notifier_exp = {
3534 .fcn = ctnetlink_expect_event,
3535 };
3536 #endif
3537
3538 static const struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
3539 [IPCTNL_MSG_CT_NEW] = { .call = ctnetlink_new_conntrack,
3540 .attr_count = CTA_MAX,
3541 .policy = ct_nla_policy },
3542 [IPCTNL_MSG_CT_GET] = { .call = ctnetlink_get_conntrack,
3543 .attr_count = CTA_MAX,
3544 .policy = ct_nla_policy },
3545 [IPCTNL_MSG_CT_DELETE] = { .call = ctnetlink_del_conntrack,
3546 .attr_count = CTA_MAX,
3547 .policy = ct_nla_policy },
3548 [IPCTNL_MSG_CT_GET_CTRZERO] = { .call = ctnetlink_get_conntrack,
3549 .attr_count = CTA_MAX,
3550 .policy = ct_nla_policy },
3551 [IPCTNL_MSG_CT_GET_STATS_CPU] = { .call = ctnetlink_stat_ct_cpu },
3552 [IPCTNL_MSG_CT_GET_STATS] = { .call = ctnetlink_stat_ct },
3553 [IPCTNL_MSG_CT_GET_DYING] = { .call = ctnetlink_get_ct_dying },
3554 [IPCTNL_MSG_CT_GET_UNCONFIRMED] = { .call = ctnetlink_get_ct_unconfirmed },
3555 };
3556
3557 static const struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
3558 [IPCTNL_MSG_EXP_GET] = { .call = ctnetlink_get_expect,
3559 .attr_count = CTA_EXPECT_MAX,
3560 .policy = exp_nla_policy },
3561 [IPCTNL_MSG_EXP_NEW] = { .call = ctnetlink_new_expect,
3562 .attr_count = CTA_EXPECT_MAX,
3563 .policy = exp_nla_policy },
3564 [IPCTNL_MSG_EXP_DELETE] = { .call = ctnetlink_del_expect,
3565 .attr_count = CTA_EXPECT_MAX,
3566 .policy = exp_nla_policy },
3567 [IPCTNL_MSG_EXP_GET_STATS_CPU] = { .call = ctnetlink_stat_exp_cpu },
3568 };
3569
3570 static const struct nfnetlink_subsystem ctnl_subsys = {
3571 .name = "conntrack",
3572 .subsys_id = NFNL_SUBSYS_CTNETLINK,
3573 .cb_count = IPCTNL_MSG_MAX,
3574 .cb = ctnl_cb,
3575 };
3576
3577 static const struct nfnetlink_subsystem ctnl_exp_subsys = {
3578 .name = "conntrack_expect",
3579 .subsys_id = NFNL_SUBSYS_CTNETLINK_EXP,
3580 .cb_count = IPCTNL_MSG_EXP_MAX,
3581 .cb = ctnl_exp_cb,
3582 };
3583
3584 MODULE_ALIAS("ip_conntrack_netlink");
3585 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
3586 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
3587
3588 static int __net_init ctnetlink_net_init(struct net *net)
3589 {
3590 #ifdef CONFIG_NF_CONNTRACK_EVENTS
3591 int ret;
3592
3593 ret = nf_conntrack_register_notifier(net, &ctnl_notifier);
3594 if (ret < 0) {
3595 pr_err("ctnetlink_init: cannot register notifier.\n");
3596 goto err_out;
3597 }
3598
3599 ret = nf_ct_expect_register_notifier(net, &ctnl_notifier_exp);
3600 if (ret < 0) {
3601 pr_err("ctnetlink_init: cannot expect register notifier.\n");
3602 goto err_unreg_notifier;
3603 }
3604 #endif
3605 return 0;
3606
3607 #ifdef CONFIG_NF_CONNTRACK_EVENTS
3608 err_unreg_notifier:
3609 nf_conntrack_unregister_notifier(net, &ctnl_notifier);
3610 err_out:
3611 return ret;
3612 #endif
3613 }
3614
3615 static void ctnetlink_net_exit(struct net *net)
3616 {
3617 #ifdef CONFIG_NF_CONNTRACK_EVENTS
3618 nf_ct_expect_unregister_notifier(net, &ctnl_notifier_exp);
3619 nf_conntrack_unregister_notifier(net, &ctnl_notifier);
3620 #endif
3621 }
3622
3623 static void __net_exit ctnetlink_net_exit_batch(struct list_head *net_exit_list)
3624 {
3625 struct net *net;
3626
3627 list_for_each_entry(net, net_exit_list, exit_list)
3628 ctnetlink_net_exit(net);
3629
3630 /* wait for other cpus until they are done with ctnl_notifiers */
3631 synchronize_rcu();
3632 }
3633
3634 static struct pernet_operations ctnetlink_net_ops = {
3635 .init = ctnetlink_net_init,
3636 .exit_batch = ctnetlink_net_exit_batch,
3637 };
3638
3639 static int __init ctnetlink_init(void)
3640 {
3641 int ret;
3642
3643 ret = nfnetlink_subsys_register(&ctnl_subsys);
3644 if (ret < 0) {
3645 pr_err("ctnetlink_init: cannot register with nfnetlink.\n");
3646 goto err_out;
3647 }
3648
3649 ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
3650 if (ret < 0) {
3651 pr_err("ctnetlink_init: cannot register exp with nfnetlink.\n");
3652 goto err_unreg_subsys;
3653 }
3654
3655 ret = register_pernet_subsys(&ctnetlink_net_ops);
3656 if (ret < 0) {
3657 pr_err("ctnetlink_init: cannot register pernet operations\n");
3658 goto err_unreg_exp_subsys;
3659 }
3660 #ifdef CONFIG_NETFILTER_NETLINK_GLUE_CT
3661 /* setup interaction between nf_queue and nf_conntrack_netlink. */
3662 RCU_INIT_POINTER(nfnl_ct_hook, &ctnetlink_glue_hook);
3663 #endif
3664 return 0;
3665
3666 err_unreg_exp_subsys:
3667 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
3668 err_unreg_subsys:
3669 nfnetlink_subsys_unregister(&ctnl_subsys);
3670 err_out:
3671 return ret;
3672 }
3673
3674 static void __exit ctnetlink_exit(void)
3675 {
3676 unregister_pernet_subsys(&ctnetlink_net_ops);
3677 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
3678 nfnetlink_subsys_unregister(&ctnl_subsys);
3679 #ifdef CONFIG_NETFILTER_NETLINK_GLUE_CT
3680 RCU_INIT_POINTER(nfnl_ct_hook, NULL);
3681 #endif
3682 synchronize_rcu();
3683 }
3684
3685 module_init(ctnetlink_init);
3686 module_exit(ctnetlink_exit);
Ubuntu Jammy Linux kernel mirror