x86, irq: Remove IRQ_DISABLED check in process context IRQ move
[linux-2.6/mini2440.git] / net / netfilter / nf_conntrack_netlink.c
blobc6439c77953c49cf705ac07fe758764d1cc82a9d
1 /* Connection tracking via netlink socket. Allows for user space
2 * protocol helpers and general trouble making from userspace.
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) 2005-2008 by Pablo Neira Ayuso <pablo@netfilter.org>
9 * Initial connection tracking via netlink development funded and
10 * generally made possible by Network Robots, Inc. (www.networkrobots.com)
12 * Further development of this code funded by Astaro AG (http://www.astaro.com)
14 * This software may be used and distributed according to the terms
15 * of the GNU General Public License, incorporated herein by reference.
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/skbuff.h>
26 #include <linux/errno.h>
27 #include <linux/netlink.h>
28 #include <linux/spinlock.h>
29 #include <linux/interrupt.h>
30 #include <linux/notifier.h>
32 #include <linux/netfilter.h>
33 #include <net/netlink.h>
34 #include <net/netfilter/nf_conntrack.h>
35 #include <net/netfilter/nf_conntrack_core.h>
36 #include <net/netfilter/nf_conntrack_expect.h>
37 #include <net/netfilter/nf_conntrack_helper.h>
38 #include <net/netfilter/nf_conntrack_l3proto.h>
39 #include <net/netfilter/nf_conntrack_l4proto.h>
40 #include <net/netfilter/nf_conntrack_tuple.h>
41 #include <net/netfilter/nf_conntrack_acct.h>
42 #ifdef CONFIG_NF_NAT_NEEDED
43 #include <net/netfilter/nf_nat_core.h>
44 #include <net/netfilter/nf_nat_protocol.h>
45 #endif
47 #include <linux/netfilter/nfnetlink.h>
48 #include <linux/netfilter/nfnetlink_conntrack.h>
50 MODULE_LICENSE("GPL");
52 static char __initdata version[] = "0.93";
54 static inline int
55 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
56 const struct nf_conntrack_tuple *tuple,
57 struct nf_conntrack_l4proto *l4proto)
59 int ret = 0;
60 struct nlattr *nest_parms;
62 nest_parms = nla_nest_start(skb, CTA_TUPLE_PROTO | NLA_F_NESTED);
63 if (!nest_parms)
64 goto nla_put_failure;
65 NLA_PUT_U8(skb, CTA_PROTO_NUM, tuple->dst.protonum);
67 if (likely(l4proto->tuple_to_nlattr))
68 ret = l4proto->tuple_to_nlattr(skb, tuple);
70 nla_nest_end(skb, nest_parms);
72 return ret;
74 nla_put_failure:
75 return -1;
78 static inline int
79 ctnetlink_dump_tuples_ip(struct sk_buff *skb,
80 const struct nf_conntrack_tuple *tuple,
81 struct nf_conntrack_l3proto *l3proto)
83 int ret = 0;
84 struct nlattr *nest_parms;
86 nest_parms = nla_nest_start(skb, CTA_TUPLE_IP | NLA_F_NESTED);
87 if (!nest_parms)
88 goto nla_put_failure;
90 if (likely(l3proto->tuple_to_nlattr))
91 ret = l3proto->tuple_to_nlattr(skb, tuple);
93 nla_nest_end(skb, nest_parms);
95 return ret;
97 nla_put_failure:
98 return -1;
101 static int
102 ctnetlink_dump_tuples(struct sk_buff *skb,
103 const struct nf_conntrack_tuple *tuple)
105 int ret;
106 struct nf_conntrack_l3proto *l3proto;
107 struct nf_conntrack_l4proto *l4proto;
109 l3proto = __nf_ct_l3proto_find(tuple->src.l3num);
110 ret = ctnetlink_dump_tuples_ip(skb, tuple, l3proto);
112 if (unlikely(ret < 0))
113 return ret;
115 l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
116 ret = ctnetlink_dump_tuples_proto(skb, tuple, l4proto);
118 return ret;
121 static inline int
122 ctnetlink_dump_status(struct sk_buff *skb, const struct nf_conn *ct)
124 NLA_PUT_BE32(skb, CTA_STATUS, htonl(ct->status));
125 return 0;
127 nla_put_failure:
128 return -1;
131 static inline int
132 ctnetlink_dump_timeout(struct sk_buff *skb, const struct nf_conn *ct)
134 long timeout = (ct->timeout.expires - jiffies) / HZ;
136 if (timeout < 0)
137 timeout = 0;
139 NLA_PUT_BE32(skb, CTA_TIMEOUT, htonl(timeout));
140 return 0;
142 nla_put_failure:
143 return -1;
146 static inline int
147 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct nf_conn *ct)
149 struct nf_conntrack_l4proto *l4proto;
150 struct nlattr *nest_proto;
151 int ret;
153 l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
154 if (!l4proto->to_nlattr)
155 return 0;
157 nest_proto = nla_nest_start(skb, CTA_PROTOINFO | NLA_F_NESTED);
158 if (!nest_proto)
159 goto nla_put_failure;
161 ret = l4proto->to_nlattr(skb, nest_proto, ct);
163 nla_nest_end(skb, nest_proto);
165 return ret;
167 nla_put_failure:
168 return -1;
171 static inline int
172 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct nf_conn *ct)
174 struct nlattr *nest_helper;
175 const struct nf_conn_help *help = nfct_help(ct);
176 struct nf_conntrack_helper *helper;
178 if (!help)
179 return 0;
181 helper = rcu_dereference(help->helper);
182 if (!helper)
183 goto out;
185 nest_helper = nla_nest_start(skb, CTA_HELP | NLA_F_NESTED);
186 if (!nest_helper)
187 goto nla_put_failure;
188 NLA_PUT_STRING(skb, CTA_HELP_NAME, helper->name);
190 if (helper->to_nlattr)
191 helper->to_nlattr(skb, ct);
193 nla_nest_end(skb, nest_helper);
194 out:
195 return 0;
197 nla_put_failure:
198 return -1;
201 static int
202 ctnetlink_dump_counters(struct sk_buff *skb, const struct nf_conn *ct,
203 enum ip_conntrack_dir dir)
205 enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
206 struct nlattr *nest_count;
207 const struct nf_conn_counter *acct;
209 acct = nf_conn_acct_find(ct);
210 if (!acct)
211 return 0;
213 nest_count = nla_nest_start(skb, type | NLA_F_NESTED);
214 if (!nest_count)
215 goto nla_put_failure;
217 NLA_PUT_BE64(skb, CTA_COUNTERS_PACKETS,
218 cpu_to_be64(acct[dir].packets));
219 NLA_PUT_BE64(skb, CTA_COUNTERS_BYTES,
220 cpu_to_be64(acct[dir].bytes));
222 nla_nest_end(skb, nest_count);
224 return 0;
226 nla_put_failure:
227 return -1;
230 #ifdef CONFIG_NF_CONNTRACK_MARK
231 static inline int
232 ctnetlink_dump_mark(struct sk_buff *skb, const struct nf_conn *ct)
234 NLA_PUT_BE32(skb, CTA_MARK, htonl(ct->mark));
235 return 0;
237 nla_put_failure:
238 return -1;
240 #else
241 #define ctnetlink_dump_mark(a, b) (0)
242 #endif
244 #ifdef CONFIG_NF_CONNTRACK_SECMARK
245 static inline int
246 ctnetlink_dump_secmark(struct sk_buff *skb, const struct nf_conn *ct)
248 NLA_PUT_BE32(skb, CTA_SECMARK, htonl(ct->secmark));
249 return 0;
251 nla_put_failure:
252 return -1;
254 #else
255 #define ctnetlink_dump_secmark(a, b) (0)
256 #endif
258 #define master_tuple(ct) &(ct->master->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
260 static inline int
261 ctnetlink_dump_master(struct sk_buff *skb, const struct nf_conn *ct)
263 struct nlattr *nest_parms;
265 if (!(ct->status & IPS_EXPECTED))
266 return 0;
268 nest_parms = nla_nest_start(skb, CTA_TUPLE_MASTER | NLA_F_NESTED);
269 if (!nest_parms)
270 goto nla_put_failure;
271 if (ctnetlink_dump_tuples(skb, master_tuple(ct)) < 0)
272 goto nla_put_failure;
273 nla_nest_end(skb, nest_parms);
275 return 0;
277 nla_put_failure:
278 return -1;
281 #ifdef CONFIG_NF_NAT_NEEDED
282 static int
283 dump_nat_seq_adj(struct sk_buff *skb, const struct nf_nat_seq *natseq, int type)
285 struct nlattr *nest_parms;
287 nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
288 if (!nest_parms)
289 goto nla_put_failure;
291 NLA_PUT_BE32(skb, CTA_NAT_SEQ_CORRECTION_POS,
292 htonl(natseq->correction_pos));
293 NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_BEFORE,
294 htonl(natseq->offset_before));
295 NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_AFTER,
296 htonl(natseq->offset_after));
298 nla_nest_end(skb, nest_parms);
300 return 0;
302 nla_put_failure:
303 return -1;
306 static inline int
307 ctnetlink_dump_nat_seq_adj(struct sk_buff *skb, const struct nf_conn *ct)
309 struct nf_nat_seq *natseq;
310 struct nf_conn_nat *nat = nfct_nat(ct);
312 if (!(ct->status & IPS_SEQ_ADJUST) || !nat)
313 return 0;
315 natseq = &nat->seq[IP_CT_DIR_ORIGINAL];
316 if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_ORIG) == -1)
317 return -1;
319 natseq = &nat->seq[IP_CT_DIR_REPLY];
320 if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_REPLY) == -1)
321 return -1;
323 return 0;
325 #else
326 #define ctnetlink_dump_nat_seq_adj(a, b) (0)
327 #endif
329 static inline int
330 ctnetlink_dump_id(struct sk_buff *skb, const struct nf_conn *ct)
332 NLA_PUT_BE32(skb, CTA_ID, htonl((unsigned long)ct));
333 return 0;
335 nla_put_failure:
336 return -1;
339 static inline int
340 ctnetlink_dump_use(struct sk_buff *skb, const struct nf_conn *ct)
342 NLA_PUT_BE32(skb, CTA_USE, htonl(atomic_read(&ct->ct_general.use)));
343 return 0;
345 nla_put_failure:
346 return -1;
349 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
351 static int
352 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
353 int event, int nowait,
354 const struct nf_conn *ct)
356 struct nlmsghdr *nlh;
357 struct nfgenmsg *nfmsg;
358 struct nlattr *nest_parms;
359 unsigned char *b = skb_tail_pointer(skb);
361 event |= NFNL_SUBSYS_CTNETLINK << 8;
362 nlh = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
363 nfmsg = NLMSG_DATA(nlh);
365 nlh->nlmsg_flags = (nowait && pid) ? NLM_F_MULTI : 0;
366 nfmsg->nfgen_family = nf_ct_l3num(ct);
367 nfmsg->version = NFNETLINK_V0;
368 nfmsg->res_id = 0;
370 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
371 if (!nest_parms)
372 goto nla_put_failure;
373 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
374 goto nla_put_failure;
375 nla_nest_end(skb, nest_parms);
377 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
378 if (!nest_parms)
379 goto nla_put_failure;
380 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
381 goto nla_put_failure;
382 nla_nest_end(skb, nest_parms);
384 if (ctnetlink_dump_status(skb, ct) < 0 ||
385 ctnetlink_dump_timeout(skb, ct) < 0 ||
386 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
387 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
388 ctnetlink_dump_protoinfo(skb, ct) < 0 ||
389 ctnetlink_dump_helpinfo(skb, ct) < 0 ||
390 ctnetlink_dump_mark(skb, ct) < 0 ||
391 ctnetlink_dump_secmark(skb, ct) < 0 ||
392 ctnetlink_dump_id(skb, ct) < 0 ||
393 ctnetlink_dump_use(skb, ct) < 0 ||
394 ctnetlink_dump_master(skb, ct) < 0 ||
395 ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
396 goto nla_put_failure;
398 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
399 return skb->len;
401 nlmsg_failure:
402 nla_put_failure:
403 nlmsg_trim(skb, b);
404 return -1;
407 #ifdef CONFIG_NF_CONNTRACK_EVENTS
409 * The general structure of a ctnetlink event is
411 * CTA_TUPLE_ORIG
412 * <l3/l4-proto-attributes>
413 * CTA_TUPLE_REPLY
414 * <l3/l4-proto-attributes>
415 * CTA_ID
416 * ...
417 * CTA_PROTOINFO
418 * <l4-proto-attributes>
419 * CTA_TUPLE_MASTER
420 * <l3/l4-proto-attributes>
422 * Therefore the formular is
424 * size = sizeof(headers) + sizeof(generic_nlas) + 3 * sizeof(tuple_nlas)
425 * + sizeof(protoinfo_nlas)
427 static struct sk_buff *
428 ctnetlink_alloc_skb(const struct nf_conntrack_tuple *tuple, gfp_t gfp)
430 struct nf_conntrack_l3proto *l3proto;
431 struct nf_conntrack_l4proto *l4proto;
432 int len;
434 #define NLA_TYPE_SIZE(type) nla_total_size(sizeof(type))
436 /* proto independant part */
437 len = NLMSG_SPACE(sizeof(struct nfgenmsg))
438 + 3 * nla_total_size(0) /* CTA_TUPLE_ORIG|REPL|MASTER */
439 + 3 * nla_total_size(0) /* CTA_TUPLE_IP */
440 + 3 * nla_total_size(0) /* CTA_TUPLE_PROTO */
441 + 3 * NLA_TYPE_SIZE(u_int8_t) /* CTA_PROTO_NUM */
442 + NLA_TYPE_SIZE(u_int32_t) /* CTA_ID */
443 + NLA_TYPE_SIZE(u_int32_t) /* CTA_STATUS */
444 #ifdef CONFIG_NF_CT_ACCT
445 + 2 * nla_total_size(0) /* CTA_COUNTERS_ORIG|REPL */
446 + 2 * NLA_TYPE_SIZE(uint64_t) /* CTA_COUNTERS_PACKETS */
447 + 2 * NLA_TYPE_SIZE(uint64_t) /* CTA_COUNTERS_BYTES */
448 #endif
449 + NLA_TYPE_SIZE(u_int32_t) /* CTA_TIMEOUT */
450 + nla_total_size(0) /* CTA_PROTOINFO */
451 + nla_total_size(0) /* CTA_HELP */
452 + nla_total_size(NF_CT_HELPER_NAME_LEN) /* CTA_HELP_NAME */
453 #ifdef CONFIG_NF_CONNTRACK_SECMARK
454 + NLA_TYPE_SIZE(u_int32_t) /* CTA_SECMARK */
455 #endif
456 #ifdef CONFIG_NF_NAT_NEEDED
457 + 2 * nla_total_size(0) /* CTA_NAT_SEQ_ADJ_ORIG|REPL */
458 + 2 * NLA_TYPE_SIZE(u_int32_t) /* CTA_NAT_SEQ_CORRECTION_POS */
459 + 2 * NLA_TYPE_SIZE(u_int32_t) /* CTA_NAT_SEQ_CORRECTION_BEFORE */
460 + 2 * NLA_TYPE_SIZE(u_int32_t) /* CTA_NAT_SEQ_CORRECTION_AFTER */
461 #endif
462 #ifdef CONFIG_NF_CONNTRACK_MARK
463 + NLA_TYPE_SIZE(u_int32_t) /* CTA_MARK */
464 #endif
467 #undef NLA_TYPE_SIZE
469 rcu_read_lock();
470 l3proto = __nf_ct_l3proto_find(tuple->src.l3num);
471 len += l3proto->nla_size;
473 l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
474 len += l4proto->nla_size;
475 rcu_read_unlock();
477 return alloc_skb(len, gfp);
480 static int ctnetlink_conntrack_event(struct notifier_block *this,
481 unsigned long events, void *ptr)
483 struct nlmsghdr *nlh;
484 struct nfgenmsg *nfmsg;
485 struct nlattr *nest_parms;
486 struct nf_ct_event *item = (struct nf_ct_event *)ptr;
487 struct nf_conn *ct = item->ct;
488 struct sk_buff *skb;
489 unsigned int type;
490 sk_buff_data_t b;
491 unsigned int flags = 0, group;
493 /* ignore our fake conntrack entry */
494 if (ct == &nf_conntrack_untracked)
495 return NOTIFY_DONE;
497 if (events & IPCT_DESTROY) {
498 type = IPCTNL_MSG_CT_DELETE;
499 group = NFNLGRP_CONNTRACK_DESTROY;
500 } else if (events & (IPCT_NEW | IPCT_RELATED)) {
501 type = IPCTNL_MSG_CT_NEW;
502 flags = NLM_F_CREATE|NLM_F_EXCL;
503 group = NFNLGRP_CONNTRACK_NEW;
504 } else if (events & (IPCT_STATUS | IPCT_PROTOINFO)) {
505 type = IPCTNL_MSG_CT_NEW;
506 group = NFNLGRP_CONNTRACK_UPDATE;
507 } else
508 return NOTIFY_DONE;
510 if (!item->report && !nfnetlink_has_listeners(group))
511 return NOTIFY_DONE;
513 skb = ctnetlink_alloc_skb(tuple(ct, IP_CT_DIR_ORIGINAL), GFP_ATOMIC);
514 if (!skb)
515 return NOTIFY_DONE;
517 b = skb->tail;
519 type |= NFNL_SUBSYS_CTNETLINK << 8;
520 nlh = NLMSG_PUT(skb, item->pid, 0, type, sizeof(struct nfgenmsg));
521 nfmsg = NLMSG_DATA(nlh);
523 nlh->nlmsg_flags = flags;
524 nfmsg->nfgen_family = nf_ct_l3num(ct);
525 nfmsg->version = NFNETLINK_V0;
526 nfmsg->res_id = 0;
528 rcu_read_lock();
529 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
530 if (!nest_parms)
531 goto nla_put_failure;
532 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
533 goto nla_put_failure;
534 nla_nest_end(skb, nest_parms);
536 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
537 if (!nest_parms)
538 goto nla_put_failure;
539 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
540 goto nla_put_failure;
541 nla_nest_end(skb, nest_parms);
543 if (ctnetlink_dump_id(skb, ct) < 0)
544 goto nla_put_failure;
546 if (ctnetlink_dump_status(skb, ct) < 0)
547 goto nla_put_failure;
549 if (events & IPCT_DESTROY) {
550 if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
551 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
552 goto nla_put_failure;
553 } else {
554 if (ctnetlink_dump_timeout(skb, ct) < 0)
555 goto nla_put_failure;
557 if (events & IPCT_PROTOINFO
558 && ctnetlink_dump_protoinfo(skb, ct) < 0)
559 goto nla_put_failure;
561 if ((events & IPCT_HELPER || nfct_help(ct))
562 && ctnetlink_dump_helpinfo(skb, ct) < 0)
563 goto nla_put_failure;
565 #ifdef CONFIG_NF_CONNTRACK_SECMARK
566 if ((events & IPCT_SECMARK || ct->secmark)
567 && ctnetlink_dump_secmark(skb, ct) < 0)
568 goto nla_put_failure;
569 #endif
571 if (events & IPCT_RELATED &&
572 ctnetlink_dump_master(skb, ct) < 0)
573 goto nla_put_failure;
575 if (events & IPCT_NATSEQADJ &&
576 ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
577 goto nla_put_failure;
580 #ifdef CONFIG_NF_CONNTRACK_MARK
581 if ((events & IPCT_MARK || ct->mark)
582 && ctnetlink_dump_mark(skb, ct) < 0)
583 goto nla_put_failure;
584 #endif
585 rcu_read_unlock();
587 nlh->nlmsg_len = skb->tail - b;
588 nfnetlink_send(skb, item->pid, group, item->report);
589 return NOTIFY_DONE;
591 nla_put_failure:
592 rcu_read_unlock();
593 nlmsg_failure:
594 nfnetlink_set_err(0, group, -ENOBUFS);
595 kfree_skb(skb);
596 return NOTIFY_DONE;
598 #endif /* CONFIG_NF_CONNTRACK_EVENTS */
600 static int ctnetlink_done(struct netlink_callback *cb)
602 if (cb->args[1])
603 nf_ct_put((struct nf_conn *)cb->args[1]);
604 return 0;
607 static int
608 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
610 struct nf_conn *ct, *last;
611 struct nf_conntrack_tuple_hash *h;
612 struct hlist_nulls_node *n;
613 struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
614 u_int8_t l3proto = nfmsg->nfgen_family;
616 rcu_read_lock();
617 last = (struct nf_conn *)cb->args[1];
618 for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
619 restart:
620 hlist_nulls_for_each_entry_rcu(h, n, &init_net.ct.hash[cb->args[0]],
621 hnnode) {
622 if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
623 continue;
624 ct = nf_ct_tuplehash_to_ctrack(h);
625 if (!atomic_inc_not_zero(&ct->ct_general.use))
626 continue;
627 /* Dump entries of a given L3 protocol number.
628 * If it is not specified, ie. l3proto == 0,
629 * then dump everything. */
630 if (l3proto && nf_ct_l3num(ct) != l3proto)
631 goto releasect;
632 if (cb->args[1]) {
633 if (ct != last)
634 goto releasect;
635 cb->args[1] = 0;
637 if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
638 cb->nlh->nlmsg_seq,
639 IPCTNL_MSG_CT_NEW,
640 1, ct) < 0) {
641 cb->args[1] = (unsigned long)ct;
642 goto out;
645 if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
646 IPCTNL_MSG_CT_GET_CTRZERO) {
647 struct nf_conn_counter *acct;
649 acct = nf_conn_acct_find(ct);
650 if (acct)
651 memset(acct, 0, sizeof(struct nf_conn_counter[IP_CT_DIR_MAX]));
653 releasect:
654 nf_ct_put(ct);
656 if (cb->args[1]) {
657 cb->args[1] = 0;
658 goto restart;
661 out:
662 rcu_read_unlock();
663 if (last)
664 nf_ct_put(last);
666 return skb->len;
669 static inline int
670 ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
672 struct nlattr *tb[CTA_IP_MAX+1];
673 struct nf_conntrack_l3proto *l3proto;
674 int ret = 0;
676 nla_parse_nested(tb, CTA_IP_MAX, attr, NULL);
678 rcu_read_lock();
679 l3proto = __nf_ct_l3proto_find(tuple->src.l3num);
681 if (likely(l3proto->nlattr_to_tuple)) {
682 ret = nla_validate_nested(attr, CTA_IP_MAX,
683 l3proto->nla_policy);
684 if (ret == 0)
685 ret = l3proto->nlattr_to_tuple(tb, tuple);
688 rcu_read_unlock();
690 return ret;
693 static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = {
694 [CTA_PROTO_NUM] = { .type = NLA_U8 },
697 static inline int
698 ctnetlink_parse_tuple_proto(struct nlattr *attr,
699 struct nf_conntrack_tuple *tuple)
701 struct nlattr *tb[CTA_PROTO_MAX+1];
702 struct nf_conntrack_l4proto *l4proto;
703 int ret = 0;
705 ret = nla_parse_nested(tb, CTA_PROTO_MAX, attr, proto_nla_policy);
706 if (ret < 0)
707 return ret;
709 if (!tb[CTA_PROTO_NUM])
710 return -EINVAL;
711 tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
713 rcu_read_lock();
714 l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
716 if (likely(l4proto->nlattr_to_tuple)) {
717 ret = nla_validate_nested(attr, CTA_PROTO_MAX,
718 l4proto->nla_policy);
719 if (ret == 0)
720 ret = l4proto->nlattr_to_tuple(tb, tuple);
723 rcu_read_unlock();
725 return ret;
728 static int
729 ctnetlink_parse_tuple(struct nlattr *cda[], struct nf_conntrack_tuple *tuple,
730 enum ctattr_tuple type, u_int8_t l3num)
732 struct nlattr *tb[CTA_TUPLE_MAX+1];
733 int err;
735 memset(tuple, 0, sizeof(*tuple));
737 nla_parse_nested(tb, CTA_TUPLE_MAX, cda[type], NULL);
739 if (!tb[CTA_TUPLE_IP])
740 return -EINVAL;
742 tuple->src.l3num = l3num;
744 err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple);
745 if (err < 0)
746 return err;
748 if (!tb[CTA_TUPLE_PROTO])
749 return -EINVAL;
751 err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple);
752 if (err < 0)
753 return err;
755 /* orig and expect tuples get DIR_ORIGINAL */
756 if (type == CTA_TUPLE_REPLY)
757 tuple->dst.dir = IP_CT_DIR_REPLY;
758 else
759 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
761 return 0;
764 static inline int
765 ctnetlink_parse_help(struct nlattr *attr, char **helper_name)
767 struct nlattr *tb[CTA_HELP_MAX+1];
769 nla_parse_nested(tb, CTA_HELP_MAX, attr, NULL);
771 if (!tb[CTA_HELP_NAME])
772 return -EINVAL;
774 *helper_name = nla_data(tb[CTA_HELP_NAME]);
776 return 0;
779 static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
780 [CTA_STATUS] = { .type = NLA_U32 },
781 [CTA_TIMEOUT] = { .type = NLA_U32 },
782 [CTA_MARK] = { .type = NLA_U32 },
783 [CTA_USE] = { .type = NLA_U32 },
784 [CTA_ID] = { .type = NLA_U32 },
787 static int
788 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
789 struct nlmsghdr *nlh, struct nlattr *cda[])
791 struct nf_conntrack_tuple_hash *h;
792 struct nf_conntrack_tuple tuple;
793 struct nf_conn *ct;
794 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
795 u_int8_t u3 = nfmsg->nfgen_family;
796 int err = 0;
798 if (cda[CTA_TUPLE_ORIG])
799 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
800 else if (cda[CTA_TUPLE_REPLY])
801 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
802 else {
803 /* Flush the whole table */
804 nf_conntrack_flush(&init_net,
805 NETLINK_CB(skb).pid,
806 nlmsg_report(nlh));
807 return 0;
810 if (err < 0)
811 return err;
813 h = nf_conntrack_find_get(&init_net, &tuple);
814 if (!h)
815 return -ENOENT;
817 ct = nf_ct_tuplehash_to_ctrack(h);
819 if (cda[CTA_ID]) {
820 u_int32_t id = ntohl(nla_get_be32(cda[CTA_ID]));
821 if (id != (u32)(unsigned long)ct) {
822 nf_ct_put(ct);
823 return -ENOENT;
827 nf_conntrack_event_report(IPCT_DESTROY,
829 NETLINK_CB(skb).pid,
830 nlmsg_report(nlh));
832 /* death_by_timeout would report the event again */
833 set_bit(IPS_DYING_BIT, &ct->status);
835 nf_ct_kill(ct);
836 nf_ct_put(ct);
838 return 0;
841 static int
842 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
843 struct nlmsghdr *nlh, struct nlattr *cda[])
845 struct nf_conntrack_tuple_hash *h;
846 struct nf_conntrack_tuple tuple;
847 struct nf_conn *ct;
848 struct sk_buff *skb2 = NULL;
849 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
850 u_int8_t u3 = nfmsg->nfgen_family;
851 int err = 0;
853 if (nlh->nlmsg_flags & NLM_F_DUMP)
854 return netlink_dump_start(ctnl, skb, nlh, ctnetlink_dump_table,
855 ctnetlink_done);
857 if (cda[CTA_TUPLE_ORIG])
858 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
859 else if (cda[CTA_TUPLE_REPLY])
860 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
861 else
862 return -EINVAL;
864 if (err < 0)
865 return err;
867 h = nf_conntrack_find_get(&init_net, &tuple);
868 if (!h)
869 return -ENOENT;
871 ct = nf_ct_tuplehash_to_ctrack(h);
873 err = -ENOMEM;
874 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
875 if (!skb2) {
876 nf_ct_put(ct);
877 return -ENOMEM;
880 rcu_read_lock();
881 err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
882 IPCTNL_MSG_CT_NEW, 1, ct);
883 rcu_read_unlock();
884 nf_ct_put(ct);
885 if (err <= 0)
886 goto free;
888 err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
889 if (err < 0)
890 goto out;
892 return 0;
894 free:
895 kfree_skb(skb2);
896 out:
897 return err;
900 #ifdef CONFIG_NF_NAT_NEEDED
901 static int
902 ctnetlink_parse_nat_setup(struct nf_conn *ct,
903 enum nf_nat_manip_type manip,
904 struct nlattr *attr)
906 typeof(nfnetlink_parse_nat_setup_hook) parse_nat_setup;
908 parse_nat_setup = rcu_dereference(nfnetlink_parse_nat_setup_hook);
909 if (!parse_nat_setup) {
910 #ifdef CONFIG_MODULES
911 rcu_read_unlock();
912 spin_unlock_bh(&nf_conntrack_lock);
913 nfnl_unlock();
914 if (request_module("nf-nat-ipv4") < 0) {
915 nfnl_lock();
916 spin_lock_bh(&nf_conntrack_lock);
917 rcu_read_lock();
918 return -EOPNOTSUPP;
920 nfnl_lock();
921 spin_lock_bh(&nf_conntrack_lock);
922 rcu_read_lock();
923 if (nfnetlink_parse_nat_setup_hook)
924 return -EAGAIN;
925 #endif
926 return -EOPNOTSUPP;
929 return parse_nat_setup(ct, manip, attr);
931 #endif
933 static int
934 ctnetlink_change_status(struct nf_conn *ct, struct nlattr *cda[])
936 unsigned long d;
937 unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
938 d = ct->status ^ status;
940 if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
941 /* unchangeable */
942 return -EBUSY;
944 if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
945 /* SEEN_REPLY bit can only be set */
946 return -EBUSY;
948 if (d & IPS_ASSURED && !(status & IPS_ASSURED))
949 /* ASSURED bit can only be set */
950 return -EBUSY;
952 /* Be careful here, modifying NAT bits can screw up things,
953 * so don't let users modify them directly if they don't pass
954 * nf_nat_range. */
955 ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
956 return 0;
959 static int
960 ctnetlink_change_nat(struct nf_conn *ct, struct nlattr *cda[])
962 #ifdef CONFIG_NF_NAT_NEEDED
963 int ret;
965 if (cda[CTA_NAT_DST]) {
966 ret = ctnetlink_parse_nat_setup(ct,
967 IP_NAT_MANIP_DST,
968 cda[CTA_NAT_DST]);
969 if (ret < 0)
970 return ret;
972 if (cda[CTA_NAT_SRC]) {
973 ret = ctnetlink_parse_nat_setup(ct,
974 IP_NAT_MANIP_SRC,
975 cda[CTA_NAT_SRC]);
976 if (ret < 0)
977 return ret;
979 return 0;
980 #else
981 return -EOPNOTSUPP;
982 #endif
985 static inline int
986 ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
988 struct nf_conntrack_helper *helper;
989 struct nf_conn_help *help = nfct_help(ct);
990 char *helpname;
991 int err;
993 /* don't change helper of sibling connections */
994 if (ct->master)
995 return -EBUSY;
997 err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
998 if (err < 0)
999 return err;
1001 if (!strcmp(helpname, "")) {
1002 if (help && help->helper) {
1003 /* we had a helper before ... */
1004 nf_ct_remove_expectations(ct);
1005 rcu_assign_pointer(help->helper, NULL);
1008 return 0;
1011 helper = __nf_conntrack_helper_find_byname(helpname);
1012 if (helper == NULL) {
1013 #ifdef CONFIG_MODULES
1014 spin_unlock_bh(&nf_conntrack_lock);
1016 if (request_module("nfct-helper-%s", helpname) < 0) {
1017 spin_lock_bh(&nf_conntrack_lock);
1018 return -EOPNOTSUPP;
1021 spin_lock_bh(&nf_conntrack_lock);
1022 helper = __nf_conntrack_helper_find_byname(helpname);
1023 if (helper)
1024 return -EAGAIN;
1025 #endif
1026 return -EOPNOTSUPP;
1029 if (help) {
1030 if (help->helper == helper)
1031 return 0;
1032 if (help->helper)
1033 return -EBUSY;
1034 /* need to zero data of old helper */
1035 memset(&help->help, 0, sizeof(help->help));
1036 } else {
1037 help = nf_ct_helper_ext_add(ct, GFP_ATOMIC);
1038 if (help == NULL)
1039 return -ENOMEM;
1042 rcu_assign_pointer(help->helper, helper);
1044 return 0;
1047 static inline int
1048 ctnetlink_change_timeout(struct nf_conn *ct, struct nlattr *cda[])
1050 u_int32_t timeout = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
1052 if (!del_timer(&ct->timeout))
1053 return -ETIME;
1055 ct->timeout.expires = jiffies + timeout * HZ;
1056 add_timer(&ct->timeout);
1058 return 0;
1061 static inline int
1062 ctnetlink_change_protoinfo(struct nf_conn *ct, struct nlattr *cda[])
1064 struct nlattr *tb[CTA_PROTOINFO_MAX+1], *attr = cda[CTA_PROTOINFO];
1065 struct nf_conntrack_l4proto *l4proto;
1066 int err = 0;
1068 nla_parse_nested(tb, CTA_PROTOINFO_MAX, attr, NULL);
1070 rcu_read_lock();
1071 l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
1072 if (l4proto->from_nlattr)
1073 err = l4proto->from_nlattr(tb, ct);
1074 rcu_read_unlock();
1076 return err;
1079 #ifdef CONFIG_NF_NAT_NEEDED
1080 static inline int
1081 change_nat_seq_adj(struct nf_nat_seq *natseq, struct nlattr *attr)
1083 struct nlattr *cda[CTA_NAT_SEQ_MAX+1];
1085 nla_parse_nested(cda, CTA_NAT_SEQ_MAX, attr, NULL);
1087 if (!cda[CTA_NAT_SEQ_CORRECTION_POS])
1088 return -EINVAL;
1090 natseq->correction_pos =
1091 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_CORRECTION_POS]));
1093 if (!cda[CTA_NAT_SEQ_OFFSET_BEFORE])
1094 return -EINVAL;
1096 natseq->offset_before =
1097 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_BEFORE]));
1099 if (!cda[CTA_NAT_SEQ_OFFSET_AFTER])
1100 return -EINVAL;
1102 natseq->offset_after =
1103 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_AFTER]));
1105 return 0;
1108 static int
1109 ctnetlink_change_nat_seq_adj(struct nf_conn *ct, struct nlattr *cda[])
1111 int ret = 0;
1112 struct nf_conn_nat *nat = nfct_nat(ct);
1114 if (!nat)
1115 return 0;
1117 if (cda[CTA_NAT_SEQ_ADJ_ORIG]) {
1118 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_ORIGINAL],
1119 cda[CTA_NAT_SEQ_ADJ_ORIG]);
1120 if (ret < 0)
1121 return ret;
1123 ct->status |= IPS_SEQ_ADJUST;
1126 if (cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1127 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_REPLY],
1128 cda[CTA_NAT_SEQ_ADJ_REPLY]);
1129 if (ret < 0)
1130 return ret;
1132 ct->status |= IPS_SEQ_ADJUST;
1135 return 0;
1137 #endif
1139 static int
1140 ctnetlink_change_conntrack(struct nf_conn *ct, struct nlattr *cda[])
1142 int err;
1144 /* only allow NAT changes and master assignation for new conntracks */
1145 if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST] || cda[CTA_TUPLE_MASTER])
1146 return -EOPNOTSUPP;
1148 if (cda[CTA_HELP]) {
1149 err = ctnetlink_change_helper(ct, cda);
1150 if (err < 0)
1151 return err;
1154 if (cda[CTA_TIMEOUT]) {
1155 err = ctnetlink_change_timeout(ct, cda);
1156 if (err < 0)
1157 return err;
1160 if (cda[CTA_STATUS]) {
1161 err = ctnetlink_change_status(ct, cda);
1162 if (err < 0)
1163 return err;
1166 if (cda[CTA_PROTOINFO]) {
1167 err = ctnetlink_change_protoinfo(ct, cda);
1168 if (err < 0)
1169 return err;
1172 #if defined(CONFIG_NF_CONNTRACK_MARK)
1173 if (cda[CTA_MARK])
1174 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1175 #endif
1177 #ifdef CONFIG_NF_NAT_NEEDED
1178 if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1179 err = ctnetlink_change_nat_seq_adj(ct, cda);
1180 if (err < 0)
1181 return err;
1183 #endif
1185 return 0;
1188 static inline void
1189 ctnetlink_event_report(struct nf_conn *ct, u32 pid, int report)
1191 unsigned int events = 0;
1193 if (test_bit(IPS_EXPECTED_BIT, &ct->status))
1194 events |= IPCT_RELATED;
1195 else
1196 events |= IPCT_NEW;
1198 nf_conntrack_event_report(IPCT_STATUS |
1199 IPCT_HELPER |
1200 IPCT_REFRESH |
1201 IPCT_PROTOINFO |
1202 IPCT_NATSEQADJ |
1203 IPCT_MARK |
1204 events,
1206 pid,
1207 report);
1210 static struct nf_conn *
1211 ctnetlink_create_conntrack(struct nlattr *cda[],
1212 struct nf_conntrack_tuple *otuple,
1213 struct nf_conntrack_tuple *rtuple,
1214 u8 u3)
1216 struct nf_conn *ct;
1217 int err = -EINVAL;
1218 struct nf_conntrack_helper *helper;
1220 ct = nf_conntrack_alloc(&init_net, otuple, rtuple, GFP_ATOMIC);
1221 if (IS_ERR(ct))
1222 return ERR_PTR(-ENOMEM);
1224 if (!cda[CTA_TIMEOUT])
1225 goto err1;
1226 ct->timeout.expires = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
1228 ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1229 ct->status |= IPS_CONFIRMED;
1231 rcu_read_lock();
1232 if (cda[CTA_HELP]) {
1233 char *helpname;
1235 err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
1236 if (err < 0)
1237 goto err2;
1239 helper = __nf_conntrack_helper_find_byname(helpname);
1240 if (helper == NULL) {
1241 rcu_read_unlock();
1242 #ifdef CONFIG_MODULES
1243 if (request_module("nfct-helper-%s", helpname) < 0) {
1244 err = -EOPNOTSUPP;
1245 goto err1;
1248 rcu_read_lock();
1249 helper = __nf_conntrack_helper_find_byname(helpname);
1250 if (helper) {
1251 err = -EAGAIN;
1252 goto err2;
1254 rcu_read_unlock();
1255 #endif
1256 err = -EOPNOTSUPP;
1257 goto err1;
1258 } else {
1259 struct nf_conn_help *help;
1261 help = nf_ct_helper_ext_add(ct, GFP_ATOMIC);
1262 if (help == NULL) {
1263 err = -ENOMEM;
1264 goto err2;
1267 /* not in hash table yet so not strictly necessary */
1268 rcu_assign_pointer(help->helper, helper);
1270 } else {
1271 /* try an implicit helper assignation */
1272 err = __nf_ct_try_assign_helper(ct, GFP_ATOMIC);
1273 if (err < 0)
1274 goto err2;
1277 if (cda[CTA_STATUS]) {
1278 err = ctnetlink_change_status(ct, cda);
1279 if (err < 0)
1280 goto err2;
1283 if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
1284 err = ctnetlink_change_nat(ct, cda);
1285 if (err < 0)
1286 goto err2;
1289 #ifdef CONFIG_NF_NAT_NEEDED
1290 if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1291 err = ctnetlink_change_nat_seq_adj(ct, cda);
1292 if (err < 0)
1293 goto err2;
1295 #endif
1297 if (cda[CTA_PROTOINFO]) {
1298 err = ctnetlink_change_protoinfo(ct, cda);
1299 if (err < 0)
1300 goto err2;
1303 nf_ct_acct_ext_add(ct, GFP_ATOMIC);
1305 #if defined(CONFIG_NF_CONNTRACK_MARK)
1306 if (cda[CTA_MARK])
1307 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1308 #endif
1310 /* setup master conntrack: this is a confirmed expectation */
1311 if (cda[CTA_TUPLE_MASTER]) {
1312 struct nf_conntrack_tuple master;
1313 struct nf_conntrack_tuple_hash *master_h;
1314 struct nf_conn *master_ct;
1316 err = ctnetlink_parse_tuple(cda, &master, CTA_TUPLE_MASTER, u3);
1317 if (err < 0)
1318 goto err2;
1320 master_h = nf_conntrack_find_get(&init_net, &master);
1321 if (master_h == NULL) {
1322 err = -ENOENT;
1323 goto err2;
1325 master_ct = nf_ct_tuplehash_to_ctrack(master_h);
1326 __set_bit(IPS_EXPECTED_BIT, &ct->status);
1327 ct->master = master_ct;
1330 add_timer(&ct->timeout);
1331 nf_conntrack_hash_insert(ct);
1332 rcu_read_unlock();
1334 return ct;
1336 err2:
1337 rcu_read_unlock();
1338 err1:
1339 nf_conntrack_free(ct);
1340 return ERR_PTR(err);
1343 static int
1344 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1345 struct nlmsghdr *nlh, struct nlattr *cda[])
1347 struct nf_conntrack_tuple otuple, rtuple;
1348 struct nf_conntrack_tuple_hash *h = NULL;
1349 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1350 u_int8_t u3 = nfmsg->nfgen_family;
1351 int err = 0;
1353 if (cda[CTA_TUPLE_ORIG]) {
1354 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG, u3);
1355 if (err < 0)
1356 return err;
1359 if (cda[CTA_TUPLE_REPLY]) {
1360 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY, u3);
1361 if (err < 0)
1362 return err;
1365 spin_lock_bh(&nf_conntrack_lock);
1366 if (cda[CTA_TUPLE_ORIG])
1367 h = __nf_conntrack_find(&init_net, &otuple);
1368 else if (cda[CTA_TUPLE_REPLY])
1369 h = __nf_conntrack_find(&init_net, &rtuple);
1371 if (h == NULL) {
1372 err = -ENOENT;
1373 if (nlh->nlmsg_flags & NLM_F_CREATE) {
1374 struct nf_conn *ct;
1376 ct = ctnetlink_create_conntrack(cda, &otuple,
1377 &rtuple, u3);
1378 if (IS_ERR(ct)) {
1379 err = PTR_ERR(ct);
1380 goto out_unlock;
1382 err = 0;
1383 nf_conntrack_get(&ct->ct_general);
1384 spin_unlock_bh(&nf_conntrack_lock);
1385 ctnetlink_event_report(ct,
1386 NETLINK_CB(skb).pid,
1387 nlmsg_report(nlh));
1388 nf_ct_put(ct);
1389 } else
1390 spin_unlock_bh(&nf_conntrack_lock);
1392 return err;
1394 /* implicit 'else' */
1396 /* We manipulate the conntrack inside the global conntrack table lock,
1397 * so there's no need to increase the refcount */
1398 err = -EEXIST;
1399 if (!(nlh->nlmsg_flags & NLM_F_EXCL)) {
1400 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
1402 err = ctnetlink_change_conntrack(ct, cda);
1403 if (err == 0) {
1404 nf_conntrack_get(&ct->ct_general);
1405 spin_unlock_bh(&nf_conntrack_lock);
1406 ctnetlink_event_report(ct,
1407 NETLINK_CB(skb).pid,
1408 nlmsg_report(nlh));
1409 nf_ct_put(ct);
1410 } else
1411 spin_unlock_bh(&nf_conntrack_lock);
1413 return err;
1416 out_unlock:
1417 spin_unlock_bh(&nf_conntrack_lock);
1418 return err;
1421 /***********************************************************************
1422 * EXPECT
1423 ***********************************************************************/
1425 static inline int
1426 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1427 const struct nf_conntrack_tuple *tuple,
1428 enum ctattr_expect type)
1430 struct nlattr *nest_parms;
1432 nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
1433 if (!nest_parms)
1434 goto nla_put_failure;
1435 if (ctnetlink_dump_tuples(skb, tuple) < 0)
1436 goto nla_put_failure;
1437 nla_nest_end(skb, nest_parms);
1439 return 0;
1441 nla_put_failure:
1442 return -1;
1445 static inline int
1446 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1447 const struct nf_conntrack_tuple *tuple,
1448 const struct nf_conntrack_tuple_mask *mask)
1450 int ret;
1451 struct nf_conntrack_l3proto *l3proto;
1452 struct nf_conntrack_l4proto *l4proto;
1453 struct nf_conntrack_tuple m;
1454 struct nlattr *nest_parms;
1456 memset(&m, 0xFF, sizeof(m));
1457 m.src.u.all = mask->src.u.all;
1458 memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3));
1460 nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK | NLA_F_NESTED);
1461 if (!nest_parms)
1462 goto nla_put_failure;
1464 l3proto = __nf_ct_l3proto_find(tuple->src.l3num);
1465 ret = ctnetlink_dump_tuples_ip(skb, &m, l3proto);
1467 if (unlikely(ret < 0))
1468 goto nla_put_failure;
1470 l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
1471 ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto);
1472 if (unlikely(ret < 0))
1473 goto nla_put_failure;
1475 nla_nest_end(skb, nest_parms);
1477 return 0;
1479 nla_put_failure:
1480 return -1;
1483 static int
1484 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1485 const struct nf_conntrack_expect *exp)
1487 struct nf_conn *master = exp->master;
1488 long timeout = (exp->timeout.expires - jiffies) / HZ;
1490 if (timeout < 0)
1491 timeout = 0;
1493 if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1494 goto nla_put_failure;
1495 if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1496 goto nla_put_failure;
1497 if (ctnetlink_exp_dump_tuple(skb,
1498 &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1499 CTA_EXPECT_MASTER) < 0)
1500 goto nla_put_failure;
1502 NLA_PUT_BE32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout));
1503 NLA_PUT_BE32(skb, CTA_EXPECT_ID, htonl((unsigned long)exp));
1505 return 0;
1507 nla_put_failure:
1508 return -1;
1511 static int
1512 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1513 int event,
1514 int nowait,
1515 const struct nf_conntrack_expect *exp)
1517 struct nlmsghdr *nlh;
1518 struct nfgenmsg *nfmsg;
1519 unsigned char *b = skb_tail_pointer(skb);
1521 event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1522 nlh = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1523 nfmsg = NLMSG_DATA(nlh);
1525 nlh->nlmsg_flags = (nowait && pid) ? NLM_F_MULTI : 0;
1526 nfmsg->nfgen_family = exp->tuple.src.l3num;
1527 nfmsg->version = NFNETLINK_V0;
1528 nfmsg->res_id = 0;
1530 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1531 goto nla_put_failure;
1533 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
1534 return skb->len;
1536 nlmsg_failure:
1537 nla_put_failure:
1538 nlmsg_trim(skb, b);
1539 return -1;
1542 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1543 static int ctnetlink_expect_event(struct notifier_block *this,
1544 unsigned long events, void *ptr)
1546 struct nlmsghdr *nlh;
1547 struct nfgenmsg *nfmsg;
1548 struct nf_exp_event *item = (struct nf_exp_event *)ptr;
1549 struct nf_conntrack_expect *exp = item->exp;
1550 struct sk_buff *skb;
1551 unsigned int type;
1552 sk_buff_data_t b;
1553 int flags = 0;
1555 if (events & IPEXP_NEW) {
1556 type = IPCTNL_MSG_EXP_NEW;
1557 flags = NLM_F_CREATE|NLM_F_EXCL;
1558 } else
1559 return NOTIFY_DONE;
1561 if (!item->report &&
1562 !nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1563 return NOTIFY_DONE;
1565 skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1566 if (!skb)
1567 return NOTIFY_DONE;
1569 b = skb->tail;
1571 type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1572 nlh = NLMSG_PUT(skb, item->pid, 0, type, sizeof(struct nfgenmsg));
1573 nfmsg = NLMSG_DATA(nlh);
1575 nlh->nlmsg_flags = flags;
1576 nfmsg->nfgen_family = exp->tuple.src.l3num;
1577 nfmsg->version = NFNETLINK_V0;
1578 nfmsg->res_id = 0;
1580 rcu_read_lock();
1581 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1582 goto nla_put_failure;
1583 rcu_read_unlock();
1585 nlh->nlmsg_len = skb->tail - b;
1586 nfnetlink_send(skb, item->pid, NFNLGRP_CONNTRACK_EXP_NEW, item->report);
1587 return NOTIFY_DONE;
1589 nla_put_failure:
1590 rcu_read_unlock();
1591 nlmsg_failure:
1592 nfnetlink_set_err(0, 0, -ENOBUFS);
1593 kfree_skb(skb);
1594 return NOTIFY_DONE;
1596 #endif
1597 static int ctnetlink_exp_done(struct netlink_callback *cb)
1599 if (cb->args[1])
1600 nf_ct_expect_put((struct nf_conntrack_expect *)cb->args[1]);
1601 return 0;
1604 static int
1605 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1607 struct net *net = &init_net;
1608 struct nf_conntrack_expect *exp, *last;
1609 struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
1610 struct hlist_node *n;
1611 u_int8_t l3proto = nfmsg->nfgen_family;
1613 rcu_read_lock();
1614 last = (struct nf_conntrack_expect *)cb->args[1];
1615 for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) {
1616 restart:
1617 hlist_for_each_entry(exp, n, &net->ct.expect_hash[cb->args[0]],
1618 hnode) {
1619 if (l3proto && exp->tuple.src.l3num != l3proto)
1620 continue;
1621 if (cb->args[1]) {
1622 if (exp != last)
1623 continue;
1624 cb->args[1] = 0;
1626 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1627 cb->nlh->nlmsg_seq,
1628 IPCTNL_MSG_EXP_NEW,
1629 1, exp) < 0) {
1630 if (!atomic_inc_not_zero(&exp->use))
1631 continue;
1632 cb->args[1] = (unsigned long)exp;
1633 goto out;
1636 if (cb->args[1]) {
1637 cb->args[1] = 0;
1638 goto restart;
1641 out:
1642 rcu_read_unlock();
1643 if (last)
1644 nf_ct_expect_put(last);
1646 return skb->len;
1649 static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = {
1650 [CTA_EXPECT_TIMEOUT] = { .type = NLA_U32 },
1651 [CTA_EXPECT_ID] = { .type = NLA_U32 },
1654 static int
1655 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1656 struct nlmsghdr *nlh, struct nlattr *cda[])
1658 struct nf_conntrack_tuple tuple;
1659 struct nf_conntrack_expect *exp;
1660 struct sk_buff *skb2;
1661 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1662 u_int8_t u3 = nfmsg->nfgen_family;
1663 int err = 0;
1665 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1666 return netlink_dump_start(ctnl, skb, nlh,
1667 ctnetlink_exp_dump_table,
1668 ctnetlink_exp_done);
1671 if (cda[CTA_EXPECT_MASTER])
1672 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER, u3);
1673 else
1674 return -EINVAL;
1676 if (err < 0)
1677 return err;
1679 exp = nf_ct_expect_find_get(&init_net, &tuple);
1680 if (!exp)
1681 return -ENOENT;
1683 if (cda[CTA_EXPECT_ID]) {
1684 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1685 if (ntohl(id) != (u32)(unsigned long)exp) {
1686 nf_ct_expect_put(exp);
1687 return -ENOENT;
1691 err = -ENOMEM;
1692 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1693 if (!skb2)
1694 goto out;
1696 rcu_read_lock();
1697 err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1698 nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1699 1, exp);
1700 rcu_read_unlock();
1701 if (err <= 0)
1702 goto free;
1704 nf_ct_expect_put(exp);
1706 return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1708 free:
1709 kfree_skb(skb2);
1710 out:
1711 nf_ct_expect_put(exp);
1712 return err;
1715 static int
1716 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1717 struct nlmsghdr *nlh, struct nlattr *cda[])
1719 struct nf_conntrack_expect *exp;
1720 struct nf_conntrack_tuple tuple;
1721 struct nf_conntrack_helper *h;
1722 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1723 struct hlist_node *n, *next;
1724 u_int8_t u3 = nfmsg->nfgen_family;
1725 unsigned int i;
1726 int err;
1728 if (cda[CTA_EXPECT_TUPLE]) {
1729 /* delete a single expect by tuple */
1730 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1731 if (err < 0)
1732 return err;
1734 /* bump usage count to 2 */
1735 exp = nf_ct_expect_find_get(&init_net, &tuple);
1736 if (!exp)
1737 return -ENOENT;
1739 if (cda[CTA_EXPECT_ID]) {
1740 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1741 if (ntohl(id) != (u32)(unsigned long)exp) {
1742 nf_ct_expect_put(exp);
1743 return -ENOENT;
1747 /* after list removal, usage count == 1 */
1748 nf_ct_unexpect_related(exp);
1749 /* have to put what we 'get' above.
1750 * after this line usage count == 0 */
1751 nf_ct_expect_put(exp);
1752 } else if (cda[CTA_EXPECT_HELP_NAME]) {
1753 char *name = nla_data(cda[CTA_EXPECT_HELP_NAME]);
1754 struct nf_conn_help *m_help;
1756 /* delete all expectations for this helper */
1757 spin_lock_bh(&nf_conntrack_lock);
1758 h = __nf_conntrack_helper_find_byname(name);
1759 if (!h) {
1760 spin_unlock_bh(&nf_conntrack_lock);
1761 return -EOPNOTSUPP;
1763 for (i = 0; i < nf_ct_expect_hsize; i++) {
1764 hlist_for_each_entry_safe(exp, n, next,
1765 &init_net.ct.expect_hash[i],
1766 hnode) {
1767 m_help = nfct_help(exp->master);
1768 if (m_help->helper == h
1769 && del_timer(&exp->timeout)) {
1770 nf_ct_unlink_expect(exp);
1771 nf_ct_expect_put(exp);
1775 spin_unlock_bh(&nf_conntrack_lock);
1776 } else {
1777 /* This basically means we have to flush everything*/
1778 spin_lock_bh(&nf_conntrack_lock);
1779 for (i = 0; i < nf_ct_expect_hsize; i++) {
1780 hlist_for_each_entry_safe(exp, n, next,
1781 &init_net.ct.expect_hash[i],
1782 hnode) {
1783 if (del_timer(&exp->timeout)) {
1784 nf_ct_unlink_expect(exp);
1785 nf_ct_expect_put(exp);
1789 spin_unlock_bh(&nf_conntrack_lock);
1792 return 0;
1794 static int
1795 ctnetlink_change_expect(struct nf_conntrack_expect *x, struct nlattr *cda[])
1797 return -EOPNOTSUPP;
1800 static int
1801 ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3, u32 pid, int report)
1803 struct nf_conntrack_tuple tuple, mask, master_tuple;
1804 struct nf_conntrack_tuple_hash *h = NULL;
1805 struct nf_conntrack_expect *exp;
1806 struct nf_conn *ct;
1807 struct nf_conn_help *help;
1808 int err = 0;
1810 /* caller guarantees that those three CTA_EXPECT_* exist */
1811 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1812 if (err < 0)
1813 return err;
1814 err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK, u3);
1815 if (err < 0)
1816 return err;
1817 err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER, u3);
1818 if (err < 0)
1819 return err;
1821 /* Look for master conntrack of this expectation */
1822 h = nf_conntrack_find_get(&init_net, &master_tuple);
1823 if (!h)
1824 return -ENOENT;
1825 ct = nf_ct_tuplehash_to_ctrack(h);
1826 help = nfct_help(ct);
1828 if (!help || !help->helper) {
1829 /* such conntrack hasn't got any helper, abort */
1830 err = -EOPNOTSUPP;
1831 goto out;
1834 exp = nf_ct_expect_alloc(ct);
1835 if (!exp) {
1836 err = -ENOMEM;
1837 goto out;
1840 exp->class = 0;
1841 exp->expectfn = NULL;
1842 exp->flags = 0;
1843 exp->master = ct;
1844 exp->helper = NULL;
1845 memcpy(&exp->tuple, &tuple, sizeof(struct nf_conntrack_tuple));
1846 memcpy(&exp->mask.src.u3, &mask.src.u3, sizeof(exp->mask.src.u3));
1847 exp->mask.src.u.all = mask.src.u.all;
1849 err = nf_ct_expect_related_report(exp, pid, report);
1850 nf_ct_expect_put(exp);
1852 out:
1853 nf_ct_put(nf_ct_tuplehash_to_ctrack(h));
1854 return err;
1857 static int
1858 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1859 struct nlmsghdr *nlh, struct nlattr *cda[])
1861 struct nf_conntrack_tuple tuple;
1862 struct nf_conntrack_expect *exp;
1863 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1864 u_int8_t u3 = nfmsg->nfgen_family;
1865 int err = 0;
1867 if (!cda[CTA_EXPECT_TUPLE]
1868 || !cda[CTA_EXPECT_MASK]
1869 || !cda[CTA_EXPECT_MASTER])
1870 return -EINVAL;
1872 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1873 if (err < 0)
1874 return err;
1876 spin_lock_bh(&nf_conntrack_lock);
1877 exp = __nf_ct_expect_find(&init_net, &tuple);
1879 if (!exp) {
1880 spin_unlock_bh(&nf_conntrack_lock);
1881 err = -ENOENT;
1882 if (nlh->nlmsg_flags & NLM_F_CREATE) {
1883 err = ctnetlink_create_expect(cda,
1885 NETLINK_CB(skb).pid,
1886 nlmsg_report(nlh));
1888 return err;
1891 err = -EEXIST;
1892 if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1893 err = ctnetlink_change_expect(exp, cda);
1894 spin_unlock_bh(&nf_conntrack_lock);
1896 return err;
1899 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1900 static struct notifier_block ctnl_notifier = {
1901 .notifier_call = ctnetlink_conntrack_event,
1904 static struct notifier_block ctnl_notifier_exp = {
1905 .notifier_call = ctnetlink_expect_event,
1907 #endif
1909 static const struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1910 [IPCTNL_MSG_CT_NEW] = { .call = ctnetlink_new_conntrack,
1911 .attr_count = CTA_MAX,
1912 .policy = ct_nla_policy },
1913 [IPCTNL_MSG_CT_GET] = { .call = ctnetlink_get_conntrack,
1914 .attr_count = CTA_MAX,
1915 .policy = ct_nla_policy },
1916 [IPCTNL_MSG_CT_DELETE] = { .call = ctnetlink_del_conntrack,
1917 .attr_count = CTA_MAX,
1918 .policy = ct_nla_policy },
1919 [IPCTNL_MSG_CT_GET_CTRZERO] = { .call = ctnetlink_get_conntrack,
1920 .attr_count = CTA_MAX,
1921 .policy = ct_nla_policy },
1924 static const struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1925 [IPCTNL_MSG_EXP_GET] = { .call = ctnetlink_get_expect,
1926 .attr_count = CTA_EXPECT_MAX,
1927 .policy = exp_nla_policy },
1928 [IPCTNL_MSG_EXP_NEW] = { .call = ctnetlink_new_expect,
1929 .attr_count = CTA_EXPECT_MAX,
1930 .policy = exp_nla_policy },
1931 [IPCTNL_MSG_EXP_DELETE] = { .call = ctnetlink_del_expect,
1932 .attr_count = CTA_EXPECT_MAX,
1933 .policy = exp_nla_policy },
1936 static const struct nfnetlink_subsystem ctnl_subsys = {
1937 .name = "conntrack",
1938 .subsys_id = NFNL_SUBSYS_CTNETLINK,
1939 .cb_count = IPCTNL_MSG_MAX,
1940 .cb = ctnl_cb,
1943 static const struct nfnetlink_subsystem ctnl_exp_subsys = {
1944 .name = "conntrack_expect",
1945 .subsys_id = NFNL_SUBSYS_CTNETLINK_EXP,
1946 .cb_count = IPCTNL_MSG_EXP_MAX,
1947 .cb = ctnl_exp_cb,
1950 MODULE_ALIAS("ip_conntrack_netlink");
1951 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1952 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1954 static int __init ctnetlink_init(void)
1956 int ret;
1958 printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1959 ret = nfnetlink_subsys_register(&ctnl_subsys);
1960 if (ret < 0) {
1961 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1962 goto err_out;
1965 ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1966 if (ret < 0) {
1967 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1968 goto err_unreg_subsys;
1971 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1972 ret = nf_conntrack_register_notifier(&ctnl_notifier);
1973 if (ret < 0) {
1974 printk("ctnetlink_init: cannot register notifier.\n");
1975 goto err_unreg_exp_subsys;
1978 ret = nf_ct_expect_register_notifier(&ctnl_notifier_exp);
1979 if (ret < 0) {
1980 printk("ctnetlink_init: cannot expect register notifier.\n");
1981 goto err_unreg_notifier;
1983 #endif
1985 return 0;
1987 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1988 err_unreg_notifier:
1989 nf_conntrack_unregister_notifier(&ctnl_notifier);
1990 err_unreg_exp_subsys:
1991 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1992 #endif
1993 err_unreg_subsys:
1994 nfnetlink_subsys_unregister(&ctnl_subsys);
1995 err_out:
1996 return ret;
1999 static void __exit ctnetlink_exit(void)
2001 printk("ctnetlink: unregistering from nfnetlink.\n");
2003 #ifdef CONFIG_NF_CONNTRACK_EVENTS
2004 nf_ct_expect_unregister_notifier(&ctnl_notifier_exp);
2005 nf_conntrack_unregister_notifier(&ctnl_notifier);
2006 #endif
2008 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
2009 nfnetlink_subsys_unregister(&ctnl_subsys);
2010 return;
2013 module_init(ctnetlink_init);
2014 module_exit(ctnetlink_exit);