nf_conntrack_netlink.c revision cb1cb5c47457ff2b604dac2da44cab4d39d11459
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) 2005-2008 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/types.h>
23#include <linux/timer.h>
24#include <linux/skbuff.h>
25#include <linux/errno.h>
26#include <linux/netlink.h>
27#include <linux/spinlock.h>
28#include <linux/interrupt.h>
29#include <linux/notifier.h>
30
31#include <linux/netfilter.h>
32#include <net/netlink.h>
33#include <net/netfilter/nf_conntrack.h>
34#include <net/netfilter/nf_conntrack_core.h>
35#include <net/netfilter/nf_conntrack_expect.h>
36#include <net/netfilter/nf_conntrack_helper.h>
37#include <net/netfilter/nf_conntrack_l3proto.h>
38#include <net/netfilter/nf_conntrack_l4proto.h>
39#include <net/netfilter/nf_conntrack_tuple.h>
40#include <net/netfilter/nf_conntrack_acct.h>
41#ifdef CONFIG_NF_NAT_NEEDED
42#include <net/netfilter/nf_nat_core.h>
43#include <net/netfilter/nf_nat_protocol.h>
44#endif
45
46#include <linux/netfilter/nfnetlink.h>
47#include <linux/netfilter/nfnetlink_conntrack.h>
48
49MODULE_LICENSE("GPL");
50
51static char __initdata version[] = "0.93";
52
53static inline int
54ctnetlink_dump_tuples_proto(struct sk_buff *skb,
55			    const struct nf_conntrack_tuple *tuple,
56			    struct nf_conntrack_l4proto *l4proto)
57{
58	int ret = 0;
59	struct nlattr *nest_parms;
60
61	nest_parms = nla_nest_start(skb, CTA_TUPLE_PROTO | NLA_F_NESTED);
62	if (!nest_parms)
63		goto nla_put_failure;
64	NLA_PUT_U8(skb, CTA_PROTO_NUM, tuple->dst.protonum);
65
66	if (likely(l4proto->tuple_to_nlattr))
67		ret = l4proto->tuple_to_nlattr(skb, tuple);
68
69	nla_nest_end(skb, nest_parms);
70
71	return ret;
72
73nla_put_failure:
74	return -1;
75}
76
77static inline int
78ctnetlink_dump_tuples_ip(struct sk_buff *skb,
79			 const struct nf_conntrack_tuple *tuple,
80			 struct nf_conntrack_l3proto *l3proto)
81{
82	int ret = 0;
83	struct nlattr *nest_parms;
84
85	nest_parms = nla_nest_start(skb, CTA_TUPLE_IP | NLA_F_NESTED);
86	if (!nest_parms)
87		goto nla_put_failure;
88
89	if (likely(l3proto->tuple_to_nlattr))
90		ret = l3proto->tuple_to_nlattr(skb, tuple);
91
92	nla_nest_end(skb, nest_parms);
93
94	return ret;
95
96nla_put_failure:
97	return -1;
98}
99
100static int
101ctnetlink_dump_tuples(struct sk_buff *skb,
102		      const struct nf_conntrack_tuple *tuple)
103{
104	int ret;
105	struct nf_conntrack_l3proto *l3proto;
106	struct nf_conntrack_l4proto *l4proto;
107
108	l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
109	ret = ctnetlink_dump_tuples_ip(skb, tuple, l3proto);
110	nf_ct_l3proto_put(l3proto);
111
112	if (unlikely(ret < 0))
113		return ret;
114
115	l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
116	ret = ctnetlink_dump_tuples_proto(skb, tuple, l4proto);
117	nf_ct_l4proto_put(l4proto);
118
119	return ret;
120}
121
122static inline int
123ctnetlink_dump_status(struct sk_buff *skb, const struct nf_conn *ct)
124{
125	NLA_PUT_BE32(skb, CTA_STATUS, htonl(ct->status));
126	return 0;
127
128nla_put_failure:
129	return -1;
130}
131
132static inline int
133ctnetlink_dump_timeout(struct sk_buff *skb, const struct nf_conn *ct)
134{
135	long timeout = (ct->timeout.expires - jiffies) / HZ;
136
137	if (timeout < 0)
138		timeout = 0;
139
140	NLA_PUT_BE32(skb, CTA_TIMEOUT, htonl(timeout));
141	return 0;
142
143nla_put_failure:
144	return -1;
145}
146
147static inline int
148ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct nf_conn *ct)
149{
150	struct nf_conntrack_l4proto *l4proto;
151	struct nlattr *nest_proto;
152	int ret;
153
154	l4proto = nf_ct_l4proto_find_get(nf_ct_l3num(ct), nf_ct_protonum(ct));
155	if (!l4proto->to_nlattr) {
156		nf_ct_l4proto_put(l4proto);
157		return 0;
158	}
159
160	nest_proto = nla_nest_start(skb, CTA_PROTOINFO | NLA_F_NESTED);
161	if (!nest_proto)
162		goto nla_put_failure;
163
164	ret = l4proto->to_nlattr(skb, nest_proto, ct);
165
166	nf_ct_l4proto_put(l4proto);
167
168	nla_nest_end(skb, nest_proto);
169
170	return ret;
171
172nla_put_failure:
173	nf_ct_l4proto_put(l4proto);
174	return -1;
175}
176
177static inline int
178ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct nf_conn *ct)
179{
180	struct nlattr *nest_helper;
181	const struct nf_conn_help *help = nfct_help(ct);
182	struct nf_conntrack_helper *helper;
183
184	if (!help)
185		return 0;
186
187	rcu_read_lock();
188	helper = rcu_dereference(help->helper);
189	if (!helper)
190		goto out;
191
192	nest_helper = nla_nest_start(skb, CTA_HELP | NLA_F_NESTED);
193	if (!nest_helper)
194		goto nla_put_failure;
195	NLA_PUT_STRING(skb, CTA_HELP_NAME, helper->name);
196
197	if (helper->to_nlattr)
198		helper->to_nlattr(skb, ct);
199
200	nla_nest_end(skb, nest_helper);
201out:
202	rcu_read_unlock();
203	return 0;
204
205nla_put_failure:
206	rcu_read_unlock();
207	return -1;
208}
209
210static int
211ctnetlink_dump_counters(struct sk_buff *skb, const struct nf_conn *ct,
212			enum ip_conntrack_dir dir)
213{
214	enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
215	struct nlattr *nest_count;
216	const struct nf_conn_counter *acct;
217
218	acct = nf_conn_acct_find(ct);
219	if (!acct)
220		return 0;
221
222	nest_count = nla_nest_start(skb, type | NLA_F_NESTED);
223	if (!nest_count)
224		goto nla_put_failure;
225
226	NLA_PUT_BE64(skb, CTA_COUNTERS_PACKETS,
227		     cpu_to_be64(acct[dir].packets));
228	NLA_PUT_BE64(skb, CTA_COUNTERS_BYTES,
229		     cpu_to_be64(acct[dir].bytes));
230
231	nla_nest_end(skb, nest_count);
232
233	return 0;
234
235nla_put_failure:
236	return -1;
237}
238
239#ifdef CONFIG_NF_CONNTRACK_MARK
240static inline int
241ctnetlink_dump_mark(struct sk_buff *skb, const struct nf_conn *ct)
242{
243	NLA_PUT_BE32(skb, CTA_MARK, htonl(ct->mark));
244	return 0;
245
246nla_put_failure:
247	return -1;
248}
249#else
250#define ctnetlink_dump_mark(a, b) (0)
251#endif
252
253#ifdef CONFIG_NF_CONNTRACK_SECMARK
254static inline int
255ctnetlink_dump_secmark(struct sk_buff *skb, const struct nf_conn *ct)
256{
257	NLA_PUT_BE32(skb, CTA_SECMARK, htonl(ct->secmark));
258	return 0;
259
260nla_put_failure:
261	return -1;
262}
263#else
264#define ctnetlink_dump_secmark(a, b) (0)
265#endif
266
267#define master_tuple(ct) &(ct->master->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
268
269static inline int
270ctnetlink_dump_master(struct sk_buff *skb, const struct nf_conn *ct)
271{
272	struct nlattr *nest_parms;
273
274	if (!(ct->status & IPS_EXPECTED))
275		return 0;
276
277	nest_parms = nla_nest_start(skb, CTA_TUPLE_MASTER | NLA_F_NESTED);
278	if (!nest_parms)
279		goto nla_put_failure;
280	if (ctnetlink_dump_tuples(skb, master_tuple(ct)) < 0)
281		goto nla_put_failure;
282	nla_nest_end(skb, nest_parms);
283
284	return 0;
285
286nla_put_failure:
287	return -1;
288}
289
290#ifdef CONFIG_NF_NAT_NEEDED
291static int
292dump_nat_seq_adj(struct sk_buff *skb, const struct nf_nat_seq *natseq, int type)
293{
294	struct nlattr *nest_parms;
295
296	nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
297	if (!nest_parms)
298		goto nla_put_failure;
299
300	NLA_PUT_BE32(skb, CTA_NAT_SEQ_CORRECTION_POS,
301		     htonl(natseq->correction_pos));
302	NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_BEFORE,
303		     htonl(natseq->offset_before));
304	NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_AFTER,
305		     htonl(natseq->offset_after));
306
307	nla_nest_end(skb, nest_parms);
308
309	return 0;
310
311nla_put_failure:
312	return -1;
313}
314
315static inline int
316ctnetlink_dump_nat_seq_adj(struct sk_buff *skb, const struct nf_conn *ct)
317{
318	struct nf_nat_seq *natseq;
319	struct nf_conn_nat *nat = nfct_nat(ct);
320
321	if (!(ct->status & IPS_SEQ_ADJUST) || !nat)
322		return 0;
323
324	natseq = &nat->seq[IP_CT_DIR_ORIGINAL];
325	if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_ORIG) == -1)
326		return -1;
327
328	natseq = &nat->seq[IP_CT_DIR_REPLY];
329	if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_REPLY) == -1)
330		return -1;
331
332	return 0;
333}
334#else
335#define ctnetlink_dump_nat_seq_adj(a, b) (0)
336#endif
337
338static inline int
339ctnetlink_dump_id(struct sk_buff *skb, const struct nf_conn *ct)
340{
341	NLA_PUT_BE32(skb, CTA_ID, htonl((unsigned long)ct));
342	return 0;
343
344nla_put_failure:
345	return -1;
346}
347
348static inline int
349ctnetlink_dump_use(struct sk_buff *skb, const struct nf_conn *ct)
350{
351	NLA_PUT_BE32(skb, CTA_USE, htonl(atomic_read(&ct->ct_general.use)));
352	return 0;
353
354nla_put_failure:
355	return -1;
356}
357
358#define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
359
360static int
361ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
362		    int event, int nowait,
363		    const struct nf_conn *ct)
364{
365	struct nlmsghdr *nlh;
366	struct nfgenmsg *nfmsg;
367	struct nlattr *nest_parms;
368	unsigned char *b = skb_tail_pointer(skb);
369
370	event |= NFNL_SUBSYS_CTNETLINK << 8;
371	nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
372	nfmsg  = NLMSG_DATA(nlh);
373
374	nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
375	nfmsg->nfgen_family = nf_ct_l3num(ct);
376	nfmsg->version      = NFNETLINK_V0;
377	nfmsg->res_id	    = 0;
378
379	nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
380	if (!nest_parms)
381		goto nla_put_failure;
382	if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
383		goto nla_put_failure;
384	nla_nest_end(skb, nest_parms);
385
386	nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
387	if (!nest_parms)
388		goto nla_put_failure;
389	if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
390		goto nla_put_failure;
391	nla_nest_end(skb, nest_parms);
392
393	if (ctnetlink_dump_status(skb, ct) < 0 ||
394	    ctnetlink_dump_timeout(skb, ct) < 0 ||
395	    ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
396	    ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
397	    ctnetlink_dump_protoinfo(skb, ct) < 0 ||
398	    ctnetlink_dump_helpinfo(skb, ct) < 0 ||
399	    ctnetlink_dump_mark(skb, ct) < 0 ||
400	    ctnetlink_dump_secmark(skb, ct) < 0 ||
401	    ctnetlink_dump_id(skb, ct) < 0 ||
402	    ctnetlink_dump_use(skb, ct) < 0 ||
403	    ctnetlink_dump_master(skb, ct) < 0 ||
404	    ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
405		goto nla_put_failure;
406
407	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
408	return skb->len;
409
410nlmsg_failure:
411nla_put_failure:
412	nlmsg_trim(skb, b);
413	return -1;
414}
415
416#ifdef CONFIG_NF_CONNTRACK_EVENTS
417static int ctnetlink_conntrack_event(struct notifier_block *this,
418				     unsigned long events, void *ptr)
419{
420	struct nlmsghdr *nlh;
421	struct nfgenmsg *nfmsg;
422	struct nlattr *nest_parms;
423	struct nf_conn *ct = (struct nf_conn *)ptr;
424	struct sk_buff *skb;
425	unsigned int type;
426	sk_buff_data_t b;
427	unsigned int flags = 0, group;
428
429	/* ignore our fake conntrack entry */
430	if (ct == &nf_conntrack_untracked)
431		return NOTIFY_DONE;
432
433	if (events & IPCT_DESTROY) {
434		type = IPCTNL_MSG_CT_DELETE;
435		group = NFNLGRP_CONNTRACK_DESTROY;
436	} else  if (events & (IPCT_NEW | IPCT_RELATED)) {
437		type = IPCTNL_MSG_CT_NEW;
438		flags = NLM_F_CREATE|NLM_F_EXCL;
439		group = NFNLGRP_CONNTRACK_NEW;
440	} else  if (events & (IPCT_STATUS | IPCT_PROTOINFO)) {
441		type = IPCTNL_MSG_CT_NEW;
442		group = NFNLGRP_CONNTRACK_UPDATE;
443	} else
444		return NOTIFY_DONE;
445
446	if (!nfnetlink_has_listeners(group))
447		return NOTIFY_DONE;
448
449	skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
450	if (!skb)
451		return NOTIFY_DONE;
452
453	b = skb->tail;
454
455	type |= NFNL_SUBSYS_CTNETLINK << 8;
456	nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
457	nfmsg = NLMSG_DATA(nlh);
458
459	nlh->nlmsg_flags    = flags;
460	nfmsg->nfgen_family = nf_ct_l3num(ct);
461	nfmsg->version	= NFNETLINK_V0;
462	nfmsg->res_id	= 0;
463
464	nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
465	if (!nest_parms)
466		goto nla_put_failure;
467	if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
468		goto nla_put_failure;
469	nla_nest_end(skb, nest_parms);
470
471	nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
472	if (!nest_parms)
473		goto nla_put_failure;
474	if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
475		goto nla_put_failure;
476	nla_nest_end(skb, nest_parms);
477
478	if (ctnetlink_dump_id(skb, ct) < 0)
479		goto nla_put_failure;
480
481	if (ctnetlink_dump_status(skb, ct) < 0)
482		goto nla_put_failure;
483
484	if (events & IPCT_DESTROY) {
485		if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
486		    ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
487			goto nla_put_failure;
488	} else {
489		if (ctnetlink_dump_timeout(skb, ct) < 0)
490			goto nla_put_failure;
491
492		if (events & IPCT_PROTOINFO
493		    && ctnetlink_dump_protoinfo(skb, ct) < 0)
494			goto nla_put_failure;
495
496		if ((events & IPCT_HELPER || nfct_help(ct))
497		    && ctnetlink_dump_helpinfo(skb, ct) < 0)
498			goto nla_put_failure;
499
500#ifdef CONFIG_NF_CONNTRACK_SECMARK
501		if ((events & IPCT_SECMARK || ct->secmark)
502		    && ctnetlink_dump_secmark(skb, ct) < 0)
503			goto nla_put_failure;
504#endif
505
506		if (events & IPCT_RELATED &&
507		    ctnetlink_dump_master(skb, ct) < 0)
508			goto nla_put_failure;
509
510		if (events & IPCT_NATSEQADJ &&
511		    ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
512			goto nla_put_failure;
513	}
514
515#ifdef CONFIG_NF_CONNTRACK_MARK
516	if ((events & IPCT_MARK || ct->mark)
517	    && ctnetlink_dump_mark(skb, ct) < 0)
518		goto nla_put_failure;
519#endif
520
521	nlh->nlmsg_len = skb->tail - b;
522	nfnetlink_send(skb, 0, group, 0);
523	return NOTIFY_DONE;
524
525nlmsg_failure:
526nla_put_failure:
527	kfree_skb(skb);
528	return NOTIFY_DONE;
529}
530#endif /* CONFIG_NF_CONNTRACK_EVENTS */
531
532static int ctnetlink_done(struct netlink_callback *cb)
533{
534	if (cb->args[1])
535		nf_ct_put((struct nf_conn *)cb->args[1]);
536	return 0;
537}
538
539static int
540ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
541{
542	struct nf_conn *ct, *last;
543	struct nf_conntrack_tuple_hash *h;
544	struct hlist_node *n;
545	struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
546	u_int8_t l3proto = nfmsg->nfgen_family;
547
548	rcu_read_lock();
549	last = (struct nf_conn *)cb->args[1];
550	for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
551restart:
552		hlist_for_each_entry_rcu(h, n, &nf_conntrack_hash[cb->args[0]],
553					 hnode) {
554			if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
555				continue;
556			ct = nf_ct_tuplehash_to_ctrack(h);
557			/* Dump entries of a given L3 protocol number.
558			 * If it is not specified, ie. l3proto == 0,
559			 * then dump everything. */
560			if (l3proto && nf_ct_l3num(ct) != l3proto)
561				continue;
562			if (cb->args[1]) {
563				if (ct != last)
564					continue;
565				cb->args[1] = 0;
566			}
567			if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
568						cb->nlh->nlmsg_seq,
569						IPCTNL_MSG_CT_NEW,
570						1, ct) < 0) {
571				if (!atomic_inc_not_zero(&ct->ct_general.use))
572					continue;
573				cb->args[1] = (unsigned long)ct;
574				goto out;
575			}
576
577			if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
578						IPCTNL_MSG_CT_GET_CTRZERO) {
579				struct nf_conn_counter *acct;
580
581				acct = nf_conn_acct_find(ct);
582				if (acct)
583					memset(acct, 0, sizeof(struct nf_conn_counter[IP_CT_DIR_MAX]));
584			}
585		}
586		if (cb->args[1]) {
587			cb->args[1] = 0;
588			goto restart;
589		}
590	}
591out:
592	rcu_read_unlock();
593	if (last)
594		nf_ct_put(last);
595
596	return skb->len;
597}
598
599static inline int
600ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
601{
602	struct nlattr *tb[CTA_IP_MAX+1];
603	struct nf_conntrack_l3proto *l3proto;
604	int ret = 0;
605
606	nla_parse_nested(tb, CTA_IP_MAX, attr, NULL);
607
608	l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
609
610	if (likely(l3proto->nlattr_to_tuple)) {
611		ret = nla_validate_nested(attr, CTA_IP_MAX,
612					  l3proto->nla_policy);
613		if (ret == 0)
614			ret = l3proto->nlattr_to_tuple(tb, tuple);
615	}
616
617	nf_ct_l3proto_put(l3proto);
618
619	return ret;
620}
621
622static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = {
623	[CTA_PROTO_NUM]	= { .type = NLA_U8 },
624};
625
626static inline int
627ctnetlink_parse_tuple_proto(struct nlattr *attr,
628			    struct nf_conntrack_tuple *tuple)
629{
630	struct nlattr *tb[CTA_PROTO_MAX+1];
631	struct nf_conntrack_l4proto *l4proto;
632	int ret = 0;
633
634	ret = nla_parse_nested(tb, CTA_PROTO_MAX, attr, proto_nla_policy);
635	if (ret < 0)
636		return ret;
637
638	if (!tb[CTA_PROTO_NUM])
639		return -EINVAL;
640	tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
641
642	l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
643
644	if (likely(l4proto->nlattr_to_tuple)) {
645		ret = nla_validate_nested(attr, CTA_PROTO_MAX,
646					  l4proto->nla_policy);
647		if (ret == 0)
648			ret = l4proto->nlattr_to_tuple(tb, tuple);
649	}
650
651	nf_ct_l4proto_put(l4proto);
652
653	return ret;
654}
655
656static int
657ctnetlink_parse_tuple(struct nlattr *cda[], struct nf_conntrack_tuple *tuple,
658		      enum ctattr_tuple type, u_int8_t l3num)
659{
660	struct nlattr *tb[CTA_TUPLE_MAX+1];
661	int err;
662
663	memset(tuple, 0, sizeof(*tuple));
664
665	nla_parse_nested(tb, CTA_TUPLE_MAX, cda[type], NULL);
666
667	if (!tb[CTA_TUPLE_IP])
668		return -EINVAL;
669
670	tuple->src.l3num = l3num;
671
672	err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple);
673	if (err < 0)
674		return err;
675
676	if (!tb[CTA_TUPLE_PROTO])
677		return -EINVAL;
678
679	err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple);
680	if (err < 0)
681		return err;
682
683	/* orig and expect tuples get DIR_ORIGINAL */
684	if (type == CTA_TUPLE_REPLY)
685		tuple->dst.dir = IP_CT_DIR_REPLY;
686	else
687		tuple->dst.dir = IP_CT_DIR_ORIGINAL;
688
689	return 0;
690}
691
692#ifdef CONFIG_NF_NAT_NEEDED
693static const struct nla_policy protonat_nla_policy[CTA_PROTONAT_MAX+1] = {
694	[CTA_PROTONAT_PORT_MIN]	= { .type = NLA_U16 },
695	[CTA_PROTONAT_PORT_MAX]	= { .type = NLA_U16 },
696};
697
698static int nfnetlink_parse_nat_proto(struct nlattr *attr,
699				     const struct nf_conn *ct,
700				     struct nf_nat_range *range)
701{
702	struct nlattr *tb[CTA_PROTONAT_MAX+1];
703	const struct nf_nat_protocol *npt;
704	int err;
705
706	err = nla_parse_nested(tb, CTA_PROTONAT_MAX, attr, protonat_nla_policy);
707	if (err < 0)
708		return err;
709
710	npt = nf_nat_proto_find_get(nf_ct_protonum(ct));
711	if (npt->nlattr_to_range)
712		err = npt->nlattr_to_range(tb, range);
713	nf_nat_proto_put(npt);
714	return err;
715}
716
717static const struct nla_policy nat_nla_policy[CTA_NAT_MAX+1] = {
718	[CTA_NAT_MINIP]		= { .type = NLA_U32 },
719	[CTA_NAT_MAXIP]		= { .type = NLA_U32 },
720};
721
722static inline int
723nfnetlink_parse_nat(struct nlattr *nat,
724		    const struct nf_conn *ct, struct nf_nat_range *range)
725{
726	struct nlattr *tb[CTA_NAT_MAX+1];
727	int err;
728
729	memset(range, 0, sizeof(*range));
730
731	err = nla_parse_nested(tb, CTA_NAT_MAX, nat, nat_nla_policy);
732	if (err < 0)
733		return err;
734
735	if (tb[CTA_NAT_MINIP])
736		range->min_ip = nla_get_be32(tb[CTA_NAT_MINIP]);
737
738	if (!tb[CTA_NAT_MAXIP])
739		range->max_ip = range->min_ip;
740	else
741		range->max_ip = nla_get_be32(tb[CTA_NAT_MAXIP]);
742
743	if (range->min_ip)
744		range->flags |= IP_NAT_RANGE_MAP_IPS;
745
746	if (!tb[CTA_NAT_PROTO])
747		return 0;
748
749	err = nfnetlink_parse_nat_proto(tb[CTA_NAT_PROTO], ct, range);
750	if (err < 0)
751		return err;
752
753	return 0;
754}
755#endif
756
757static inline int
758ctnetlink_parse_help(struct nlattr *attr, char **helper_name)
759{
760	struct nlattr *tb[CTA_HELP_MAX+1];
761
762	nla_parse_nested(tb, CTA_HELP_MAX, attr, NULL);
763
764	if (!tb[CTA_HELP_NAME])
765		return -EINVAL;
766
767	*helper_name = nla_data(tb[CTA_HELP_NAME]);
768
769	return 0;
770}
771
772static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
773	[CTA_STATUS] 		= { .type = NLA_U32 },
774	[CTA_TIMEOUT] 		= { .type = NLA_U32 },
775	[CTA_MARK]		= { .type = NLA_U32 },
776	[CTA_USE]		= { .type = NLA_U32 },
777	[CTA_ID]		= { .type = NLA_U32 },
778};
779
780static int
781ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
782			struct nlmsghdr *nlh, struct nlattr *cda[])
783{
784	struct nf_conntrack_tuple_hash *h;
785	struct nf_conntrack_tuple tuple;
786	struct nf_conn *ct;
787	struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
788	u_int8_t u3 = nfmsg->nfgen_family;
789	int err = 0;
790
791	if (cda[CTA_TUPLE_ORIG])
792		err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
793	else if (cda[CTA_TUPLE_REPLY])
794		err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
795	else {
796		/* Flush the whole table */
797		nf_conntrack_flush();
798		return 0;
799	}
800
801	if (err < 0)
802		return err;
803
804	h = nf_conntrack_find_get(&tuple);
805	if (!h)
806		return -ENOENT;
807
808	ct = nf_ct_tuplehash_to_ctrack(h);
809
810	if (cda[CTA_ID]) {
811		u_int32_t id = ntohl(nla_get_be32(cda[CTA_ID]));
812		if (id != (u32)(unsigned long)ct) {
813			nf_ct_put(ct);
814			return -ENOENT;
815		}
816	}
817
818	nf_ct_kill(ct);
819	nf_ct_put(ct);
820
821	return 0;
822}
823
824static int
825ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
826			struct nlmsghdr *nlh, struct nlattr *cda[])
827{
828	struct nf_conntrack_tuple_hash *h;
829	struct nf_conntrack_tuple tuple;
830	struct nf_conn *ct;
831	struct sk_buff *skb2 = NULL;
832	struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
833	u_int8_t u3 = nfmsg->nfgen_family;
834	int err = 0;
835
836	if (nlh->nlmsg_flags & NLM_F_DUMP)
837		return netlink_dump_start(ctnl, skb, nlh, ctnetlink_dump_table,
838					  ctnetlink_done);
839
840	if (cda[CTA_TUPLE_ORIG])
841		err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
842	else if (cda[CTA_TUPLE_REPLY])
843		err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
844	else
845		return -EINVAL;
846
847	if (err < 0)
848		return err;
849
850	h = nf_conntrack_find_get(&tuple);
851	if (!h)
852		return -ENOENT;
853
854	ct = nf_ct_tuplehash_to_ctrack(h);
855
856	err = -ENOMEM;
857	skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
858	if (!skb2) {
859		nf_ct_put(ct);
860		return -ENOMEM;
861	}
862
863	err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
864				  IPCTNL_MSG_CT_NEW, 1, ct);
865	nf_ct_put(ct);
866	if (err <= 0)
867		goto free;
868
869	err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
870	if (err < 0)
871		goto out;
872
873	return 0;
874
875free:
876	kfree_skb(skb2);
877out:
878	return err;
879}
880
881static int
882ctnetlink_change_status(struct nf_conn *ct, struct nlattr *cda[])
883{
884	unsigned long d;
885	unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
886	d = ct->status ^ status;
887
888	if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
889		/* unchangeable */
890		return -EBUSY;
891
892	if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
893		/* SEEN_REPLY bit can only be set */
894		return -EBUSY;
895
896	if (d & IPS_ASSURED && !(status & IPS_ASSURED))
897		/* ASSURED bit can only be set */
898		return -EBUSY;
899
900	if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
901#ifndef CONFIG_NF_NAT_NEEDED
902		return -EOPNOTSUPP;
903#else
904		struct nf_nat_range range;
905
906		if (cda[CTA_NAT_DST]) {
907			if (nfnetlink_parse_nat(cda[CTA_NAT_DST], ct,
908						&range) < 0)
909				return -EINVAL;
910			if (nf_nat_initialized(ct, IP_NAT_MANIP_DST))
911				return -EEXIST;
912			nf_nat_setup_info(ct, &range, IP_NAT_MANIP_DST);
913		}
914		if (cda[CTA_NAT_SRC]) {
915			if (nfnetlink_parse_nat(cda[CTA_NAT_SRC], ct,
916						&range) < 0)
917				return -EINVAL;
918			if (nf_nat_initialized(ct, IP_NAT_MANIP_SRC))
919				return -EEXIST;
920			nf_nat_setup_info(ct, &range, IP_NAT_MANIP_SRC);
921		}
922#endif
923	}
924
925	/* Be careful here, modifying NAT bits can screw up things,
926	 * so don't let users modify them directly if they don't pass
927	 * nf_nat_range. */
928	ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
929	return 0;
930}
931
932
933static inline int
934ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
935{
936	struct nf_conntrack_helper *helper;
937	struct nf_conn_help *help = nfct_help(ct);
938	char *helpname;
939	int err;
940
941	/* don't change helper of sibling connections */
942	if (ct->master)
943		return -EBUSY;
944
945	err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
946	if (err < 0)
947		return err;
948
949	if (!strcmp(helpname, "")) {
950		if (help && help->helper) {
951			/* we had a helper before ... */
952			nf_ct_remove_expectations(ct);
953			rcu_assign_pointer(help->helper, NULL);
954		}
955
956		return 0;
957	}
958
959	helper = __nf_conntrack_helper_find_byname(helpname);
960	if (helper == NULL)
961		return -EOPNOTSUPP;
962
963	if (help) {
964		if (help->helper == helper)
965			return 0;
966		if (help->helper)
967			return -EBUSY;
968		/* need to zero data of old helper */
969		memset(&help->help, 0, sizeof(help->help));
970	} else {
971		help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
972		if (help == NULL)
973			return -ENOMEM;
974	}
975
976	rcu_assign_pointer(help->helper, helper);
977
978	return 0;
979}
980
981static inline int
982ctnetlink_change_timeout(struct nf_conn *ct, struct nlattr *cda[])
983{
984	u_int32_t timeout = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
985
986	if (!del_timer(&ct->timeout))
987		return -ETIME;
988
989	ct->timeout.expires = jiffies + timeout * HZ;
990	add_timer(&ct->timeout);
991
992	return 0;
993}
994
995static inline int
996ctnetlink_change_protoinfo(struct nf_conn *ct, struct nlattr *cda[])
997{
998	struct nlattr *tb[CTA_PROTOINFO_MAX+1], *attr = cda[CTA_PROTOINFO];
999	struct nf_conntrack_l4proto *l4proto;
1000	int err = 0;
1001
1002	nla_parse_nested(tb, CTA_PROTOINFO_MAX, attr, NULL);
1003
1004	l4proto = nf_ct_l4proto_find_get(nf_ct_l3num(ct), nf_ct_protonum(ct));
1005	if (l4proto->from_nlattr)
1006		err = l4proto->from_nlattr(tb, ct);
1007	nf_ct_l4proto_put(l4proto);
1008
1009	return err;
1010}
1011
1012#ifdef CONFIG_NF_NAT_NEEDED
1013static inline int
1014change_nat_seq_adj(struct nf_nat_seq *natseq, struct nlattr *attr)
1015{
1016	struct nlattr *cda[CTA_NAT_SEQ_MAX+1];
1017
1018	nla_parse_nested(cda, CTA_NAT_SEQ_MAX, attr, NULL);
1019
1020	if (!cda[CTA_NAT_SEQ_CORRECTION_POS])
1021		return -EINVAL;
1022
1023	natseq->correction_pos =
1024		ntohl(nla_get_be32(cda[CTA_NAT_SEQ_CORRECTION_POS]));
1025
1026	if (!cda[CTA_NAT_SEQ_OFFSET_BEFORE])
1027		return -EINVAL;
1028
1029	natseq->offset_before =
1030		ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_BEFORE]));
1031
1032	if (!cda[CTA_NAT_SEQ_OFFSET_AFTER])
1033		return -EINVAL;
1034
1035	natseq->offset_after =
1036		ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_AFTER]));
1037
1038	return 0;
1039}
1040
1041static int
1042ctnetlink_change_nat_seq_adj(struct nf_conn *ct, struct nlattr *cda[])
1043{
1044	int ret = 0;
1045	struct nf_conn_nat *nat = nfct_nat(ct);
1046
1047	if (!nat)
1048		return 0;
1049
1050	if (cda[CTA_NAT_SEQ_ADJ_ORIG]) {
1051		ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_ORIGINAL],
1052					 cda[CTA_NAT_SEQ_ADJ_ORIG]);
1053		if (ret < 0)
1054			return ret;
1055
1056		ct->status |= IPS_SEQ_ADJUST;
1057	}
1058
1059	if (cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1060		ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_REPLY],
1061					 cda[CTA_NAT_SEQ_ADJ_REPLY]);
1062		if (ret < 0)
1063			return ret;
1064
1065		ct->status |= IPS_SEQ_ADJUST;
1066	}
1067
1068	return 0;
1069}
1070#endif
1071
1072static int
1073ctnetlink_change_conntrack(struct nf_conn *ct, struct nlattr *cda[])
1074{
1075	int err;
1076
1077	if (cda[CTA_HELP]) {
1078		err = ctnetlink_change_helper(ct, cda);
1079		if (err < 0)
1080			return err;
1081	}
1082
1083	if (cda[CTA_TIMEOUT]) {
1084		err = ctnetlink_change_timeout(ct, cda);
1085		if (err < 0)
1086			return err;
1087	}
1088
1089	if (cda[CTA_STATUS]) {
1090		err = ctnetlink_change_status(ct, cda);
1091		if (err < 0)
1092			return err;
1093	}
1094
1095	if (cda[CTA_PROTOINFO]) {
1096		err = ctnetlink_change_protoinfo(ct, cda);
1097		if (err < 0)
1098			return err;
1099	}
1100
1101#if defined(CONFIG_NF_CONNTRACK_MARK)
1102	if (cda[CTA_MARK])
1103		ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1104#endif
1105
1106#ifdef CONFIG_NF_NAT_NEEDED
1107	if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1108		err = ctnetlink_change_nat_seq_adj(ct, cda);
1109		if (err < 0)
1110			return err;
1111	}
1112#endif
1113
1114	return 0;
1115}
1116
1117static int
1118ctnetlink_create_conntrack(struct nlattr *cda[],
1119			   struct nf_conntrack_tuple *otuple,
1120			   struct nf_conntrack_tuple *rtuple,
1121			   struct nf_conn *master_ct)
1122{
1123	struct nf_conn *ct;
1124	int err = -EINVAL;
1125	struct nf_conn_help *help;
1126	struct nf_conntrack_helper *helper;
1127
1128	ct = nf_conntrack_alloc(otuple, rtuple, GFP_KERNEL);
1129	if (ct == NULL || IS_ERR(ct))
1130		return -ENOMEM;
1131
1132	if (!cda[CTA_TIMEOUT])
1133		goto err;
1134	ct->timeout.expires = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
1135
1136	ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1137	ct->status |= IPS_CONFIRMED;
1138
1139	rcu_read_lock();
1140	helper = __nf_ct_helper_find(rtuple);
1141	if (helper) {
1142		help = nf_ct_helper_ext_add(ct, GFP_ATOMIC);
1143		if (help == NULL) {
1144			rcu_read_unlock();
1145			err = -ENOMEM;
1146			goto err;
1147		}
1148		/* not in hash table yet so not strictly necessary */
1149		rcu_assign_pointer(help->helper, helper);
1150	}
1151
1152	if (cda[CTA_STATUS]) {
1153		err = ctnetlink_change_status(ct, cda);
1154		if (err < 0) {
1155			rcu_read_unlock();
1156			goto err;
1157		}
1158	}
1159
1160	if (cda[CTA_PROTOINFO]) {
1161		err = ctnetlink_change_protoinfo(ct, cda);
1162		if (err < 0) {
1163			rcu_read_unlock();
1164			goto err;
1165		}
1166	}
1167
1168	nf_ct_acct_ext_add(ct, GFP_KERNEL);
1169
1170#if defined(CONFIG_NF_CONNTRACK_MARK)
1171	if (cda[CTA_MARK])
1172		ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1173#endif
1174
1175	/* setup master conntrack: this is a confirmed expectation */
1176	if (master_ct) {
1177		__set_bit(IPS_EXPECTED_BIT, &ct->status);
1178		ct->master = master_ct;
1179	}
1180
1181	add_timer(&ct->timeout);
1182	nf_conntrack_hash_insert(ct);
1183	rcu_read_unlock();
1184
1185	return 0;
1186
1187err:
1188	nf_conntrack_free(ct);
1189	return err;
1190}
1191
1192static int
1193ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1194			struct nlmsghdr *nlh, struct nlattr *cda[])
1195{
1196	struct nf_conntrack_tuple otuple, rtuple;
1197	struct nf_conntrack_tuple_hash *h = NULL;
1198	struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1199	u_int8_t u3 = nfmsg->nfgen_family;
1200	int err = 0;
1201
1202	if (cda[CTA_TUPLE_ORIG]) {
1203		err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG, u3);
1204		if (err < 0)
1205			return err;
1206	}
1207
1208	if (cda[CTA_TUPLE_REPLY]) {
1209		err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY, u3);
1210		if (err < 0)
1211			return err;
1212	}
1213
1214	spin_lock_bh(&nf_conntrack_lock);
1215	if (cda[CTA_TUPLE_ORIG])
1216		h = __nf_conntrack_find(&otuple);
1217	else if (cda[CTA_TUPLE_REPLY])
1218		h = __nf_conntrack_find(&rtuple);
1219
1220	if (h == NULL) {
1221		struct nf_conntrack_tuple master;
1222		struct nf_conntrack_tuple_hash *master_h = NULL;
1223		struct nf_conn *master_ct = NULL;
1224
1225		if (cda[CTA_TUPLE_MASTER]) {
1226			err = ctnetlink_parse_tuple(cda,
1227						    &master,
1228						    CTA_TUPLE_MASTER,
1229						    u3);
1230			if (err < 0)
1231				goto out_unlock;
1232
1233			master_h = __nf_conntrack_find(&master);
1234			if (master_h == NULL) {
1235				err = -ENOENT;
1236				goto out_unlock;
1237			}
1238			master_ct = nf_ct_tuplehash_to_ctrack(master_h);
1239			atomic_inc(&master_ct->ct_general.use);
1240		}
1241
1242		spin_unlock_bh(&nf_conntrack_lock);
1243		err = -ENOENT;
1244		if (nlh->nlmsg_flags & NLM_F_CREATE)
1245			err = ctnetlink_create_conntrack(cda,
1246							 &otuple,
1247							 &rtuple,
1248							 master_ct);
1249		if (err < 0 && master_ct)
1250			nf_ct_put(master_ct);
1251
1252		return err;
1253	}
1254	/* implicit 'else' */
1255
1256	/* We manipulate the conntrack inside the global conntrack table lock,
1257	 * so there's no need to increase the refcount */
1258	err = -EEXIST;
1259	if (!(nlh->nlmsg_flags & NLM_F_EXCL)) {
1260		/* we only allow nat config for new conntracks */
1261		if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
1262			err = -EOPNOTSUPP;
1263			goto out_unlock;
1264		}
1265		/* can't link an existing conntrack to a master */
1266		if (cda[CTA_TUPLE_MASTER]) {
1267			err = -EOPNOTSUPP;
1268			goto out_unlock;
1269		}
1270		err = ctnetlink_change_conntrack(nf_ct_tuplehash_to_ctrack(h),
1271						 cda);
1272	}
1273
1274out_unlock:
1275	spin_unlock_bh(&nf_conntrack_lock);
1276	return err;
1277}
1278
1279/***********************************************************************
1280 * EXPECT
1281 ***********************************************************************/
1282
1283static inline int
1284ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1285			 const struct nf_conntrack_tuple *tuple,
1286			 enum ctattr_expect type)
1287{
1288	struct nlattr *nest_parms;
1289
1290	nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
1291	if (!nest_parms)
1292		goto nla_put_failure;
1293	if (ctnetlink_dump_tuples(skb, tuple) < 0)
1294		goto nla_put_failure;
1295	nla_nest_end(skb, nest_parms);
1296
1297	return 0;
1298
1299nla_put_failure:
1300	return -1;
1301}
1302
1303static inline int
1304ctnetlink_exp_dump_mask(struct sk_buff *skb,
1305			const struct nf_conntrack_tuple *tuple,
1306			const struct nf_conntrack_tuple_mask *mask)
1307{
1308	int ret;
1309	struct nf_conntrack_l3proto *l3proto;
1310	struct nf_conntrack_l4proto *l4proto;
1311	struct nf_conntrack_tuple m;
1312	struct nlattr *nest_parms;
1313
1314	memset(&m, 0xFF, sizeof(m));
1315	m.src.u.all = mask->src.u.all;
1316	memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3));
1317
1318	nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK | NLA_F_NESTED);
1319	if (!nest_parms)
1320		goto nla_put_failure;
1321
1322	l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
1323	ret = ctnetlink_dump_tuples_ip(skb, &m, l3proto);
1324	nf_ct_l3proto_put(l3proto);
1325
1326	if (unlikely(ret < 0))
1327		goto nla_put_failure;
1328
1329	l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
1330	ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto);
1331	nf_ct_l4proto_put(l4proto);
1332	if (unlikely(ret < 0))
1333		goto nla_put_failure;
1334
1335	nla_nest_end(skb, nest_parms);
1336
1337	return 0;
1338
1339nla_put_failure:
1340	return -1;
1341}
1342
1343static int
1344ctnetlink_exp_dump_expect(struct sk_buff *skb,
1345			  const struct nf_conntrack_expect *exp)
1346{
1347	struct nf_conn *master = exp->master;
1348	long timeout = (exp->timeout.expires - jiffies) / HZ;
1349
1350	if (timeout < 0)
1351		timeout = 0;
1352
1353	if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1354		goto nla_put_failure;
1355	if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1356		goto nla_put_failure;
1357	if (ctnetlink_exp_dump_tuple(skb,
1358				 &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1359				 CTA_EXPECT_MASTER) < 0)
1360		goto nla_put_failure;
1361
1362	NLA_PUT_BE32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout));
1363	NLA_PUT_BE32(skb, CTA_EXPECT_ID, htonl((unsigned long)exp));
1364
1365	return 0;
1366
1367nla_put_failure:
1368	return -1;
1369}
1370
1371static int
1372ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1373		    int event,
1374		    int nowait,
1375		    const struct nf_conntrack_expect *exp)
1376{
1377	struct nlmsghdr *nlh;
1378	struct nfgenmsg *nfmsg;
1379	unsigned char *b = skb_tail_pointer(skb);
1380
1381	event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1382	nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1383	nfmsg  = NLMSG_DATA(nlh);
1384
1385	nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1386	nfmsg->nfgen_family = exp->tuple.src.l3num;
1387	nfmsg->version	    = NFNETLINK_V0;
1388	nfmsg->res_id	    = 0;
1389
1390	if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1391		goto nla_put_failure;
1392
1393	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
1394	return skb->len;
1395
1396nlmsg_failure:
1397nla_put_failure:
1398	nlmsg_trim(skb, b);
1399	return -1;
1400}
1401
1402#ifdef CONFIG_NF_CONNTRACK_EVENTS
1403static int ctnetlink_expect_event(struct notifier_block *this,
1404				  unsigned long events, void *ptr)
1405{
1406	struct nlmsghdr *nlh;
1407	struct nfgenmsg *nfmsg;
1408	struct nf_conntrack_expect *exp = (struct nf_conntrack_expect *)ptr;
1409	struct sk_buff *skb;
1410	unsigned int type;
1411	sk_buff_data_t b;
1412	int flags = 0;
1413
1414	if (events & IPEXP_NEW) {
1415		type = IPCTNL_MSG_EXP_NEW;
1416		flags = NLM_F_CREATE|NLM_F_EXCL;
1417	} else
1418		return NOTIFY_DONE;
1419
1420	if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1421		return NOTIFY_DONE;
1422
1423	skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1424	if (!skb)
1425		return NOTIFY_DONE;
1426
1427	b = skb->tail;
1428
1429	type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1430	nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1431	nfmsg = NLMSG_DATA(nlh);
1432
1433	nlh->nlmsg_flags    = flags;
1434	nfmsg->nfgen_family = exp->tuple.src.l3num;
1435	nfmsg->version	    = NFNETLINK_V0;
1436	nfmsg->res_id	    = 0;
1437
1438	if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1439		goto nla_put_failure;
1440
1441	nlh->nlmsg_len = skb->tail - b;
1442	nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1443	return NOTIFY_DONE;
1444
1445nlmsg_failure:
1446nla_put_failure:
1447	kfree_skb(skb);
1448	return NOTIFY_DONE;
1449}
1450#endif
1451static int ctnetlink_exp_done(struct netlink_callback *cb)
1452{
1453	if (cb->args[1])
1454		nf_ct_expect_put((struct nf_conntrack_expect *)cb->args[1]);
1455	return 0;
1456}
1457
1458static int
1459ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1460{
1461	struct nf_conntrack_expect *exp, *last;
1462	struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
1463	struct hlist_node *n;
1464	u_int8_t l3proto = nfmsg->nfgen_family;
1465
1466	rcu_read_lock();
1467	last = (struct nf_conntrack_expect *)cb->args[1];
1468	for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) {
1469restart:
1470		hlist_for_each_entry(exp, n, &nf_ct_expect_hash[cb->args[0]],
1471				     hnode) {
1472			if (l3proto && exp->tuple.src.l3num != l3proto)
1473				continue;
1474			if (cb->args[1]) {
1475				if (exp != last)
1476					continue;
1477				cb->args[1] = 0;
1478			}
1479			if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1480						    cb->nlh->nlmsg_seq,
1481						    IPCTNL_MSG_EXP_NEW,
1482						    1, exp) < 0) {
1483				if (!atomic_inc_not_zero(&exp->use))
1484					continue;
1485				cb->args[1] = (unsigned long)exp;
1486				goto out;
1487			}
1488		}
1489		if (cb->args[1]) {
1490			cb->args[1] = 0;
1491			goto restart;
1492		}
1493	}
1494out:
1495	rcu_read_unlock();
1496	if (last)
1497		nf_ct_expect_put(last);
1498
1499	return skb->len;
1500}
1501
1502static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = {
1503	[CTA_EXPECT_TIMEOUT]	= { .type = NLA_U32 },
1504	[CTA_EXPECT_ID]		= { .type = NLA_U32 },
1505};
1506
1507static int
1508ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1509		     struct nlmsghdr *nlh, struct nlattr *cda[])
1510{
1511	struct nf_conntrack_tuple tuple;
1512	struct nf_conntrack_expect *exp;
1513	struct sk_buff *skb2;
1514	struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1515	u_int8_t u3 = nfmsg->nfgen_family;
1516	int err = 0;
1517
1518	if (nlh->nlmsg_flags & NLM_F_DUMP) {
1519		return netlink_dump_start(ctnl, skb, nlh,
1520					  ctnetlink_exp_dump_table,
1521					  ctnetlink_exp_done);
1522	}
1523
1524	if (cda[CTA_EXPECT_MASTER])
1525		err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER, u3);
1526	else
1527		return -EINVAL;
1528
1529	if (err < 0)
1530		return err;
1531
1532	exp = nf_ct_expect_find_get(&tuple);
1533	if (!exp)
1534		return -ENOENT;
1535
1536	if (cda[CTA_EXPECT_ID]) {
1537		__be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1538		if (ntohl(id) != (u32)(unsigned long)exp) {
1539			nf_ct_expect_put(exp);
1540			return -ENOENT;
1541		}
1542	}
1543
1544	err = -ENOMEM;
1545	skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1546	if (!skb2)
1547		goto out;
1548
1549	err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1550				      nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1551				      1, exp);
1552	if (err <= 0)
1553		goto free;
1554
1555	nf_ct_expect_put(exp);
1556
1557	return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1558
1559free:
1560	kfree_skb(skb2);
1561out:
1562	nf_ct_expect_put(exp);
1563	return err;
1564}
1565
1566static int
1567ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1568		     struct nlmsghdr *nlh, struct nlattr *cda[])
1569{
1570	struct nf_conntrack_expect *exp;
1571	struct nf_conntrack_tuple tuple;
1572	struct nf_conntrack_helper *h;
1573	struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1574	struct hlist_node *n, *next;
1575	u_int8_t u3 = nfmsg->nfgen_family;
1576	unsigned int i;
1577	int err;
1578
1579	if (cda[CTA_EXPECT_TUPLE]) {
1580		/* delete a single expect by tuple */
1581		err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1582		if (err < 0)
1583			return err;
1584
1585		/* bump usage count to 2 */
1586		exp = nf_ct_expect_find_get(&tuple);
1587		if (!exp)
1588			return -ENOENT;
1589
1590		if (cda[CTA_EXPECT_ID]) {
1591			__be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1592			if (ntohl(id) != (u32)(unsigned long)exp) {
1593				nf_ct_expect_put(exp);
1594				return -ENOENT;
1595			}
1596		}
1597
1598		/* after list removal, usage count == 1 */
1599		nf_ct_unexpect_related(exp);
1600		/* have to put what we 'get' above.
1601		 * after this line usage count == 0 */
1602		nf_ct_expect_put(exp);
1603	} else if (cda[CTA_EXPECT_HELP_NAME]) {
1604		char *name = nla_data(cda[CTA_EXPECT_HELP_NAME]);
1605		struct nf_conn_help *m_help;
1606
1607		/* delete all expectations for this helper */
1608		spin_lock_bh(&nf_conntrack_lock);
1609		h = __nf_conntrack_helper_find_byname(name);
1610		if (!h) {
1611			spin_unlock_bh(&nf_conntrack_lock);
1612			return -EOPNOTSUPP;
1613		}
1614		for (i = 0; i < nf_ct_expect_hsize; i++) {
1615			hlist_for_each_entry_safe(exp, n, next,
1616						  &nf_ct_expect_hash[i],
1617						  hnode) {
1618				m_help = nfct_help(exp->master);
1619				if (m_help->helper == h
1620				    && del_timer(&exp->timeout)) {
1621					nf_ct_unlink_expect(exp);
1622					nf_ct_expect_put(exp);
1623				}
1624			}
1625		}
1626		spin_unlock_bh(&nf_conntrack_lock);
1627	} else {
1628		/* This basically means we have to flush everything*/
1629		spin_lock_bh(&nf_conntrack_lock);
1630		for (i = 0; i < nf_ct_expect_hsize; i++) {
1631			hlist_for_each_entry_safe(exp, n, next,
1632						  &nf_ct_expect_hash[i],
1633						  hnode) {
1634				if (del_timer(&exp->timeout)) {
1635					nf_ct_unlink_expect(exp);
1636					nf_ct_expect_put(exp);
1637				}
1638			}
1639		}
1640		spin_unlock_bh(&nf_conntrack_lock);
1641	}
1642
1643	return 0;
1644}
1645static int
1646ctnetlink_change_expect(struct nf_conntrack_expect *x, struct nlattr *cda[])
1647{
1648	return -EOPNOTSUPP;
1649}
1650
1651static int
1652ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3)
1653{
1654	struct nf_conntrack_tuple tuple, mask, master_tuple;
1655	struct nf_conntrack_tuple_hash *h = NULL;
1656	struct nf_conntrack_expect *exp;
1657	struct nf_conn *ct;
1658	struct nf_conn_help *help;
1659	int err = 0;
1660
1661	/* caller guarantees that those three CTA_EXPECT_* exist */
1662	err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1663	if (err < 0)
1664		return err;
1665	err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK, u3);
1666	if (err < 0)
1667		return err;
1668	err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER, u3);
1669	if (err < 0)
1670		return err;
1671
1672	/* Look for master conntrack of this expectation */
1673	h = nf_conntrack_find_get(&master_tuple);
1674	if (!h)
1675		return -ENOENT;
1676	ct = nf_ct_tuplehash_to_ctrack(h);
1677	help = nfct_help(ct);
1678
1679	if (!help || !help->helper) {
1680		/* such conntrack hasn't got any helper, abort */
1681		err = -EINVAL;
1682		goto out;
1683	}
1684
1685	exp = nf_ct_expect_alloc(ct);
1686	if (!exp) {
1687		err = -ENOMEM;
1688		goto out;
1689	}
1690
1691	exp->expectfn = NULL;
1692	exp->flags = 0;
1693	exp->master = ct;
1694	exp->helper = NULL;
1695	memcpy(&exp->tuple, &tuple, sizeof(struct nf_conntrack_tuple));
1696	memcpy(&exp->mask.src.u3, &mask.src.u3, sizeof(exp->mask.src.u3));
1697	exp->mask.src.u.all = mask.src.u.all;
1698
1699	err = nf_ct_expect_related(exp);
1700	nf_ct_expect_put(exp);
1701
1702out:
1703	nf_ct_put(nf_ct_tuplehash_to_ctrack(h));
1704	return err;
1705}
1706
1707static int
1708ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1709		     struct nlmsghdr *nlh, struct nlattr *cda[])
1710{
1711	struct nf_conntrack_tuple tuple;
1712	struct nf_conntrack_expect *exp;
1713	struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1714	u_int8_t u3 = nfmsg->nfgen_family;
1715	int err = 0;
1716
1717	if (!cda[CTA_EXPECT_TUPLE]
1718	    || !cda[CTA_EXPECT_MASK]
1719	    || !cda[CTA_EXPECT_MASTER])
1720		return -EINVAL;
1721
1722	err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1723	if (err < 0)
1724		return err;
1725
1726	spin_lock_bh(&nf_conntrack_lock);
1727	exp = __nf_ct_expect_find(&tuple);
1728
1729	if (!exp) {
1730		spin_unlock_bh(&nf_conntrack_lock);
1731		err = -ENOENT;
1732		if (nlh->nlmsg_flags & NLM_F_CREATE)
1733			err = ctnetlink_create_expect(cda, u3);
1734		return err;
1735	}
1736
1737	err = -EEXIST;
1738	if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1739		err = ctnetlink_change_expect(exp, cda);
1740	spin_unlock_bh(&nf_conntrack_lock);
1741
1742	return err;
1743}
1744
1745#ifdef CONFIG_NF_CONNTRACK_EVENTS
1746static struct notifier_block ctnl_notifier = {
1747	.notifier_call	= ctnetlink_conntrack_event,
1748};
1749
1750static struct notifier_block ctnl_notifier_exp = {
1751	.notifier_call	= ctnetlink_expect_event,
1752};
1753#endif
1754
1755static const struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1756	[IPCTNL_MSG_CT_NEW]		= { .call = ctnetlink_new_conntrack,
1757					    .attr_count = CTA_MAX,
1758					    .policy = ct_nla_policy },
1759	[IPCTNL_MSG_CT_GET] 		= { .call = ctnetlink_get_conntrack,
1760					    .attr_count = CTA_MAX,
1761					    .policy = ct_nla_policy },
1762	[IPCTNL_MSG_CT_DELETE]  	= { .call = ctnetlink_del_conntrack,
1763					    .attr_count = CTA_MAX,
1764					    .policy = ct_nla_policy },
1765	[IPCTNL_MSG_CT_GET_CTRZERO] 	= { .call = ctnetlink_get_conntrack,
1766					    .attr_count = CTA_MAX,
1767					    .policy = ct_nla_policy },
1768};
1769
1770static const struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1771	[IPCTNL_MSG_EXP_GET]		= { .call = ctnetlink_get_expect,
1772					    .attr_count = CTA_EXPECT_MAX,
1773					    .policy = exp_nla_policy },
1774	[IPCTNL_MSG_EXP_NEW]		= { .call = ctnetlink_new_expect,
1775					    .attr_count = CTA_EXPECT_MAX,
1776					    .policy = exp_nla_policy },
1777	[IPCTNL_MSG_EXP_DELETE]		= { .call = ctnetlink_del_expect,
1778					    .attr_count = CTA_EXPECT_MAX,
1779					    .policy = exp_nla_policy },
1780};
1781
1782static const struct nfnetlink_subsystem ctnl_subsys = {
1783	.name				= "conntrack",
1784	.subsys_id			= NFNL_SUBSYS_CTNETLINK,
1785	.cb_count			= IPCTNL_MSG_MAX,
1786	.cb				= ctnl_cb,
1787};
1788
1789static const struct nfnetlink_subsystem ctnl_exp_subsys = {
1790	.name				= "conntrack_expect",
1791	.subsys_id			= NFNL_SUBSYS_CTNETLINK_EXP,
1792	.cb_count			= IPCTNL_MSG_EXP_MAX,
1793	.cb				= ctnl_exp_cb,
1794};
1795
1796MODULE_ALIAS("ip_conntrack_netlink");
1797MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1798MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1799
1800static int __init ctnetlink_init(void)
1801{
1802	int ret;
1803
1804	printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1805	ret = nfnetlink_subsys_register(&ctnl_subsys);
1806	if (ret < 0) {
1807		printk("ctnetlink_init: cannot register with nfnetlink.\n");
1808		goto err_out;
1809	}
1810
1811	ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1812	if (ret < 0) {
1813		printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1814		goto err_unreg_subsys;
1815	}
1816
1817#ifdef CONFIG_NF_CONNTRACK_EVENTS
1818	ret = nf_conntrack_register_notifier(&ctnl_notifier);
1819	if (ret < 0) {
1820		printk("ctnetlink_init: cannot register notifier.\n");
1821		goto err_unreg_exp_subsys;
1822	}
1823
1824	ret = nf_ct_expect_register_notifier(&ctnl_notifier_exp);
1825	if (ret < 0) {
1826		printk("ctnetlink_init: cannot expect register notifier.\n");
1827		goto err_unreg_notifier;
1828	}
1829#endif
1830
1831	return 0;
1832
1833#ifdef CONFIG_NF_CONNTRACK_EVENTS
1834err_unreg_notifier:
1835	nf_conntrack_unregister_notifier(&ctnl_notifier);
1836err_unreg_exp_subsys:
1837	nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1838#endif
1839err_unreg_subsys:
1840	nfnetlink_subsys_unregister(&ctnl_subsys);
1841err_out:
1842	return ret;
1843}
1844
1845static void __exit ctnetlink_exit(void)
1846{
1847	printk("ctnetlink: unregistering from nfnetlink.\n");
1848
1849#ifdef CONFIG_NF_CONNTRACK_EVENTS
1850	nf_ct_expect_unregister_notifier(&ctnl_notifier_exp);
1851	nf_conntrack_unregister_notifier(&ctnl_notifier);
1852#endif
1853
1854	nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1855	nfnetlink_subsys_unregister(&ctnl_subsys);
1856	return;
1857}
1858
1859module_init(ctnetlink_init);
1860module_exit(ctnetlink_exit);
1861