page allocator: inline buffered_rmqueue()
[linux-2.6] / net / netfilter / nf_conntrack_netlink.c
1 /* Connection tracking via netlink socket. Allows for user space
2  * protocol helpers and general trouble making from userspace.
3  *
4  * (C) 2001 by Jay Schulist <jschlst@samba.org>
5  * (C) 2002-2006 by Harald Welte <laforge@gnumonks.org>
6  * (C) 2003 by Patrick Mchardy <kaber@trash.net>
7  * (C) 2005-2008 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/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
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
49 MODULE_LICENSE("GPL");
50
51 static char __initdata version[] = "0.93";
52
53 static inline int
54 ctnetlink_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
73 nla_put_failure:
74         return -1;
75 }
76
77 static inline int
78 ctnetlink_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
96 nla_put_failure:
97         return -1;
98 }
99
100 static int
101 ctnetlink_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(tuple->src.l3num);
109         ret = ctnetlink_dump_tuples_ip(skb, tuple, l3proto);
110
111         if (unlikely(ret < 0))
112                 return ret;
113
114         l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
115         ret = ctnetlink_dump_tuples_proto(skb, tuple, l4proto);
116
117         return ret;
118 }
119
120 static inline int
121 ctnetlink_dump_status(struct sk_buff *skb, const struct nf_conn *ct)
122 {
123         NLA_PUT_BE32(skb, CTA_STATUS, htonl(ct->status));
124         return 0;
125
126 nla_put_failure:
127         return -1;
128 }
129
130 static inline int
131 ctnetlink_dump_timeout(struct sk_buff *skb, const struct nf_conn *ct)
132 {
133         long timeout = (ct->timeout.expires - jiffies) / HZ;
134
135         if (timeout < 0)
136                 timeout = 0;
137
138         NLA_PUT_BE32(skb, CTA_TIMEOUT, htonl(timeout));
139         return 0;
140
141 nla_put_failure:
142         return -1;
143 }
144
145 static inline int
146 ctnetlink_dump_protoinfo(struct sk_buff *skb, struct nf_conn *ct)
147 {
148         struct nf_conntrack_l4proto *l4proto;
149         struct nlattr *nest_proto;
150         int ret;
151
152         l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
153         if (!l4proto->to_nlattr)
154                 return 0;
155
156         nest_proto = nla_nest_start(skb, CTA_PROTOINFO | NLA_F_NESTED);
157         if (!nest_proto)
158                 goto nla_put_failure;
159
160         ret = l4proto->to_nlattr(skb, nest_proto, ct);
161
162         nla_nest_end(skb, nest_proto);
163
164         return ret;
165
166 nla_put_failure:
167         return -1;
168 }
169
170 static inline int
171 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct nf_conn *ct)
172 {
173         struct nlattr *nest_helper;
174         const struct nf_conn_help *help = nfct_help(ct);
175         struct nf_conntrack_helper *helper;
176
177         if (!help)
178                 return 0;
179
180         helper = rcu_dereference(help->helper);
181         if (!helper)
182                 goto out;
183
184         nest_helper = nla_nest_start(skb, CTA_HELP | NLA_F_NESTED);
185         if (!nest_helper)
186                 goto nla_put_failure;
187         NLA_PUT_STRING(skb, CTA_HELP_NAME, helper->name);
188
189         if (helper->to_nlattr)
190                 helper->to_nlattr(skb, ct);
191
192         nla_nest_end(skb, nest_helper);
193 out:
194         return 0;
195
196 nla_put_failure:
197         return -1;
198 }
199
200 static int
201 ctnetlink_dump_counters(struct sk_buff *skb, const struct nf_conn *ct,
202                         enum ip_conntrack_dir dir)
203 {
204         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
205         struct nlattr *nest_count;
206         const struct nf_conn_counter *acct;
207
208         acct = nf_conn_acct_find(ct);
209         if (!acct)
210                 return 0;
211
212         nest_count = nla_nest_start(skb, type | NLA_F_NESTED);
213         if (!nest_count)
214                 goto nla_put_failure;
215
216         NLA_PUT_BE64(skb, CTA_COUNTERS_PACKETS,
217                      cpu_to_be64(acct[dir].packets));
218         NLA_PUT_BE64(skb, CTA_COUNTERS_BYTES,
219                      cpu_to_be64(acct[dir].bytes));
220
221         nla_nest_end(skb, nest_count);
222
223         return 0;
224
225 nla_put_failure:
226         return -1;
227 }
228
229 #ifdef CONFIG_NF_CONNTRACK_MARK
230 static inline int
231 ctnetlink_dump_mark(struct sk_buff *skb, const struct nf_conn *ct)
232 {
233         NLA_PUT_BE32(skb, CTA_MARK, htonl(ct->mark));
234         return 0;
235
236 nla_put_failure:
237         return -1;
238 }
239 #else
240 #define ctnetlink_dump_mark(a, b) (0)
241 #endif
242
243 #ifdef CONFIG_NF_CONNTRACK_SECMARK
244 static inline int
245 ctnetlink_dump_secmark(struct sk_buff *skb, const struct nf_conn *ct)
246 {
247         NLA_PUT_BE32(skb, CTA_SECMARK, htonl(ct->secmark));
248         return 0;
249
250 nla_put_failure:
251         return -1;
252 }
253 #else
254 #define ctnetlink_dump_secmark(a, b) (0)
255 #endif
256
257 #define master_tuple(ct) &(ct->master->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
258
259 static inline int
260 ctnetlink_dump_master(struct sk_buff *skb, const struct nf_conn *ct)
261 {
262         struct nlattr *nest_parms;
263
264         if (!(ct->status & IPS_EXPECTED))
265                 return 0;
266
267         nest_parms = nla_nest_start(skb, CTA_TUPLE_MASTER | NLA_F_NESTED);
268         if (!nest_parms)
269                 goto nla_put_failure;
270         if (ctnetlink_dump_tuples(skb, master_tuple(ct)) < 0)
271                 goto nla_put_failure;
272         nla_nest_end(skb, nest_parms);
273
274         return 0;
275
276 nla_put_failure:
277         return -1;
278 }
279
280 #ifdef CONFIG_NF_NAT_NEEDED
281 static int
282 dump_nat_seq_adj(struct sk_buff *skb, const struct nf_nat_seq *natseq, int type)
283 {
284         struct nlattr *nest_parms;
285
286         nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
287         if (!nest_parms)
288                 goto nla_put_failure;
289
290         NLA_PUT_BE32(skb, CTA_NAT_SEQ_CORRECTION_POS,
291                      htonl(natseq->correction_pos));
292         NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_BEFORE,
293                      htonl(natseq->offset_before));
294         NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_AFTER,
295                      htonl(natseq->offset_after));
296
297         nla_nest_end(skb, nest_parms);
298
299         return 0;
300
301 nla_put_failure:
302         return -1;
303 }
304
305 static inline int
306 ctnetlink_dump_nat_seq_adj(struct sk_buff *skb, const struct nf_conn *ct)
307 {
308         struct nf_nat_seq *natseq;
309         struct nf_conn_nat *nat = nfct_nat(ct);
310
311         if (!(ct->status & IPS_SEQ_ADJUST) || !nat)
312                 return 0;
313
314         natseq = &nat->seq[IP_CT_DIR_ORIGINAL];
315         if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_ORIG) == -1)
316                 return -1;
317
318         natseq = &nat->seq[IP_CT_DIR_REPLY];
319         if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_REPLY) == -1)
320                 return -1;
321
322         return 0;
323 }
324 #else
325 #define ctnetlink_dump_nat_seq_adj(a, b) (0)
326 #endif
327
328 static inline int
329 ctnetlink_dump_id(struct sk_buff *skb, const struct nf_conn *ct)
330 {
331         NLA_PUT_BE32(skb, CTA_ID, htonl((unsigned long)ct));
332         return 0;
333
334 nla_put_failure:
335         return -1;
336 }
337
338 static inline int
339 ctnetlink_dump_use(struct sk_buff *skb, const struct nf_conn *ct)
340 {
341         NLA_PUT_BE32(skb, CTA_USE, htonl(atomic_read(&ct->ct_general.use)));
342         return 0;
343
344 nla_put_failure:
345         return -1;
346 }
347
348 static int
349 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
350                     int event, struct nf_conn *ct)
351 {
352         struct nlmsghdr *nlh;
353         struct nfgenmsg *nfmsg;
354         struct nlattr *nest_parms;
355         unsigned int flags = pid ? NLM_F_MULTI : 0;
356
357         event |= NFNL_SUBSYS_CTNETLINK << 8;
358         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*nfmsg), flags);
359         if (nlh == NULL)
360                 goto nlmsg_failure;
361
362         nfmsg = nlmsg_data(nlh);
363         nfmsg->nfgen_family = nf_ct_l3num(ct);
364         nfmsg->version      = NFNETLINK_V0;
365         nfmsg->res_id       = 0;
366
367         nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
368         if (!nest_parms)
369                 goto nla_put_failure;
370         if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
371                 goto nla_put_failure;
372         nla_nest_end(skb, nest_parms);
373
374         nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
375         if (!nest_parms)
376                 goto nla_put_failure;
377         if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0)
378                 goto nla_put_failure;
379         nla_nest_end(skb, nest_parms);
380
381         if (ctnetlink_dump_status(skb, ct) < 0 ||
382             ctnetlink_dump_timeout(skb, ct) < 0 ||
383             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
384             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
385             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
386             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
387             ctnetlink_dump_mark(skb, ct) < 0 ||
388             ctnetlink_dump_secmark(skb, ct) < 0 ||
389             ctnetlink_dump_id(skb, ct) < 0 ||
390             ctnetlink_dump_use(skb, ct) < 0 ||
391             ctnetlink_dump_master(skb, ct) < 0 ||
392             ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
393                 goto nla_put_failure;
394
395         nlmsg_end(skb, nlh);
396         return skb->len;
397
398 nlmsg_failure:
399 nla_put_failure:
400         nlmsg_cancel(skb, nlh);
401         return -1;
402 }
403
404 #ifdef CONFIG_NF_CONNTRACK_EVENTS
405 static inline size_t
406 ctnetlink_proto_size(const struct nf_conn *ct)
407 {
408         struct nf_conntrack_l3proto *l3proto;
409         struct nf_conntrack_l4proto *l4proto;
410         size_t len = 0;
411
412         rcu_read_lock();
413         l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct));
414         len += l3proto->nla_size;
415
416         l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
417         len += l4proto->nla_size;
418         rcu_read_unlock();
419
420         return len;
421 }
422
423 static inline size_t
424 ctnetlink_nlmsg_size(const struct nf_conn *ct)
425 {
426         return NLMSG_ALIGN(sizeof(struct nfgenmsg))
427                + 3 * nla_total_size(0) /* CTA_TUPLE_ORIG|REPL|MASTER */
428                + 3 * nla_total_size(0) /* CTA_TUPLE_IP */
429                + 3 * nla_total_size(0) /* CTA_TUPLE_PROTO */
430                + 3 * nla_total_size(sizeof(u_int8_t)) /* CTA_PROTO_NUM */
431                + nla_total_size(sizeof(u_int32_t)) /* CTA_ID */
432                + nla_total_size(sizeof(u_int32_t)) /* CTA_STATUS */
433 #ifdef CONFIG_NF_CT_ACCT
434                + 2 * nla_total_size(0) /* CTA_COUNTERS_ORIG|REPL */
435                + 2 * nla_total_size(sizeof(uint64_t)) /* CTA_COUNTERS_PACKETS */
436                + 2 * nla_total_size(sizeof(uint64_t)) /* CTA_COUNTERS_BYTES */
437 #endif
438                + nla_total_size(sizeof(u_int32_t)) /* CTA_TIMEOUT */
439                + nla_total_size(0) /* CTA_PROTOINFO */
440                + nla_total_size(0) /* CTA_HELP */
441                + nla_total_size(NF_CT_HELPER_NAME_LEN) /* CTA_HELP_NAME */
442 #ifdef CONFIG_NF_CONNTRACK_SECMARK
443                + nla_total_size(sizeof(u_int32_t)) /* CTA_SECMARK */
444 #endif
445 #ifdef CONFIG_NF_NAT_NEEDED
446                + 2 * nla_total_size(0) /* CTA_NAT_SEQ_ADJ_ORIG|REPL */
447                + 6 * nla_total_size(sizeof(u_int32_t)) /* CTA_NAT_SEQ_OFFSET */
448 #endif
449 #ifdef CONFIG_NF_CONNTRACK_MARK
450                + nla_total_size(sizeof(u_int32_t)) /* CTA_MARK */
451 #endif
452                + ctnetlink_proto_size(ct)
453                ;
454 }
455
456 static int
457 ctnetlink_conntrack_event(unsigned int events, struct nf_ct_event *item)
458 {
459         struct nlmsghdr *nlh;
460         struct nfgenmsg *nfmsg;
461         struct nlattr *nest_parms;
462         struct nf_conn *ct = item->ct;
463         struct sk_buff *skb;
464         unsigned int type;
465         unsigned int flags = 0, group;
466         int err;
467
468         /* ignore our fake conntrack entry */
469         if (ct == &nf_conntrack_untracked)
470                 return 0;
471
472         if (events & (1 << IPCT_DESTROY)) {
473                 type = IPCTNL_MSG_CT_DELETE;
474                 group = NFNLGRP_CONNTRACK_DESTROY;
475         } else  if (events & ((1 << IPCT_NEW) | (1 << IPCT_RELATED))) {
476                 type = IPCTNL_MSG_CT_NEW;
477                 flags = NLM_F_CREATE|NLM_F_EXCL;
478                 group = NFNLGRP_CONNTRACK_NEW;
479         } else  if (events) {
480                 type = IPCTNL_MSG_CT_NEW;
481                 group = NFNLGRP_CONNTRACK_UPDATE;
482         } else
483                 return 0;
484
485         if (!item->report && !nfnetlink_has_listeners(group))
486                 return 0;
487
488         skb = nlmsg_new(ctnetlink_nlmsg_size(ct), GFP_ATOMIC);
489         if (skb == NULL)
490                 goto errout;
491
492         type |= NFNL_SUBSYS_CTNETLINK << 8;
493         nlh = nlmsg_put(skb, item->pid, 0, type, sizeof(*nfmsg), flags);
494         if (nlh == NULL)
495                 goto nlmsg_failure;
496
497         nfmsg = nlmsg_data(nlh);
498         nfmsg->nfgen_family = nf_ct_l3num(ct);
499         nfmsg->version  = NFNETLINK_V0;
500         nfmsg->res_id   = 0;
501
502         rcu_read_lock();
503         nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
504         if (!nest_parms)
505                 goto nla_put_failure;
506         if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
507                 goto nla_put_failure;
508         nla_nest_end(skb, nest_parms);
509
510         nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
511         if (!nest_parms)
512                 goto nla_put_failure;
513         if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0)
514                 goto nla_put_failure;
515         nla_nest_end(skb, nest_parms);
516
517         if (ctnetlink_dump_id(skb, ct) < 0)
518                 goto nla_put_failure;
519
520         if (ctnetlink_dump_status(skb, ct) < 0)
521                 goto nla_put_failure;
522
523         if (events & (1 << IPCT_DESTROY)) {
524                 if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
525                     ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
526                         goto nla_put_failure;
527         } else {
528                 if (ctnetlink_dump_timeout(skb, ct) < 0)
529                         goto nla_put_failure;
530
531                 if (events & (1 << IPCT_PROTOINFO)
532                     && ctnetlink_dump_protoinfo(skb, ct) < 0)
533                         goto nla_put_failure;
534
535                 if ((events & (1 << IPCT_HELPER) || nfct_help(ct))
536                     && ctnetlink_dump_helpinfo(skb, ct) < 0)
537                         goto nla_put_failure;
538
539 #ifdef CONFIG_NF_CONNTRACK_SECMARK
540                 if ((events & (1 << IPCT_SECMARK) || ct->secmark)
541                     && ctnetlink_dump_secmark(skb, ct) < 0)
542                         goto nla_put_failure;
543 #endif
544
545                 if (events & (1 << IPCT_RELATED) &&
546                     ctnetlink_dump_master(skb, ct) < 0)
547                         goto nla_put_failure;
548
549                 if (events & (1 << IPCT_NATSEQADJ) &&
550                     ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
551                         goto nla_put_failure;
552         }
553
554 #ifdef CONFIG_NF_CONNTRACK_MARK
555         if ((events & (1 << IPCT_MARK) || ct->mark)
556             && ctnetlink_dump_mark(skb, ct) < 0)
557                 goto nla_put_failure;
558 #endif
559         rcu_read_unlock();
560
561         nlmsg_end(skb, nlh);
562         err = nfnetlink_send(skb, item->pid, group, item->report, GFP_ATOMIC);
563         if (err == -ENOBUFS || err == -EAGAIN)
564                 return -ENOBUFS;
565
566         return 0;
567
568 nla_put_failure:
569         rcu_read_unlock();
570         nlmsg_cancel(skb, nlh);
571 nlmsg_failure:
572         kfree_skb(skb);
573 errout:
574         nfnetlink_set_err(0, group, -ENOBUFS);
575         return 0;
576 }
577 #endif /* CONFIG_NF_CONNTRACK_EVENTS */
578
579 static int ctnetlink_done(struct netlink_callback *cb)
580 {
581         if (cb->args[1])
582                 nf_ct_put((struct nf_conn *)cb->args[1]);
583         return 0;
584 }
585
586 static int
587 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
588 {
589         struct nf_conn *ct, *last;
590         struct nf_conntrack_tuple_hash *h;
591         struct hlist_nulls_node *n;
592         struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
593         u_int8_t l3proto = nfmsg->nfgen_family;
594
595         rcu_read_lock();
596         last = (struct nf_conn *)cb->args[1];
597         for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
598 restart:
599                 hlist_nulls_for_each_entry_rcu(h, n, &init_net.ct.hash[cb->args[0]],
600                                          hnnode) {
601                         if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
602                                 continue;
603                         ct = nf_ct_tuplehash_to_ctrack(h);
604                         if (!atomic_inc_not_zero(&ct->ct_general.use))
605                                 continue;
606                         /* Dump entries of a given L3 protocol number.
607                          * If it is not specified, ie. l3proto == 0,
608                          * then dump everything. */
609                         if (l3proto && nf_ct_l3num(ct) != l3proto)
610                                 goto releasect;
611                         if (cb->args[1]) {
612                                 if (ct != last)
613                                         goto releasect;
614                                 cb->args[1] = 0;
615                         }
616                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
617                                                 cb->nlh->nlmsg_seq,
618                                                 IPCTNL_MSG_CT_NEW, ct) < 0) {
619                                 cb->args[1] = (unsigned long)ct;
620                                 goto out;
621                         }
622
623                         if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
624                                                 IPCTNL_MSG_CT_GET_CTRZERO) {
625                                 struct nf_conn_counter *acct;
626
627                                 acct = nf_conn_acct_find(ct);
628                                 if (acct)
629                                         memset(acct, 0, sizeof(struct nf_conn_counter[IP_CT_DIR_MAX]));
630                         }
631 releasect:
632                 nf_ct_put(ct);
633                 }
634                 if (cb->args[1]) {
635                         cb->args[1] = 0;
636                         goto restart;
637                 }
638         }
639 out:
640         rcu_read_unlock();
641         if (last)
642                 nf_ct_put(last);
643
644         return skb->len;
645 }
646
647 static inline int
648 ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
649 {
650         struct nlattr *tb[CTA_IP_MAX+1];
651         struct nf_conntrack_l3proto *l3proto;
652         int ret = 0;
653
654         nla_parse_nested(tb, CTA_IP_MAX, attr, NULL);
655
656         rcu_read_lock();
657         l3proto = __nf_ct_l3proto_find(tuple->src.l3num);
658
659         if (likely(l3proto->nlattr_to_tuple)) {
660                 ret = nla_validate_nested(attr, CTA_IP_MAX,
661                                           l3proto->nla_policy);
662                 if (ret == 0)
663                         ret = l3proto->nlattr_to_tuple(tb, tuple);
664         }
665
666         rcu_read_unlock();
667
668         return ret;
669 }
670
671 static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = {
672         [CTA_PROTO_NUM] = { .type = NLA_U8 },
673 };
674
675 static inline int
676 ctnetlink_parse_tuple_proto(struct nlattr *attr,
677                             struct nf_conntrack_tuple *tuple)
678 {
679         struct nlattr *tb[CTA_PROTO_MAX+1];
680         struct nf_conntrack_l4proto *l4proto;
681         int ret = 0;
682
683         ret = nla_parse_nested(tb, CTA_PROTO_MAX, attr, proto_nla_policy);
684         if (ret < 0)
685                 return ret;
686
687         if (!tb[CTA_PROTO_NUM])
688                 return -EINVAL;
689         tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
690
691         rcu_read_lock();
692         l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
693
694         if (likely(l4proto->nlattr_to_tuple)) {
695                 ret = nla_validate_nested(attr, CTA_PROTO_MAX,
696                                           l4proto->nla_policy);
697                 if (ret == 0)
698                         ret = l4proto->nlattr_to_tuple(tb, tuple);
699         }
700
701         rcu_read_unlock();
702
703         return ret;
704 }
705
706 static int
707 ctnetlink_parse_tuple(struct nlattr *cda[], struct nf_conntrack_tuple *tuple,
708                       enum ctattr_tuple type, u_int8_t l3num)
709 {
710         struct nlattr *tb[CTA_TUPLE_MAX+1];
711         int err;
712
713         memset(tuple, 0, sizeof(*tuple));
714
715         nla_parse_nested(tb, CTA_TUPLE_MAX, cda[type], NULL);
716
717         if (!tb[CTA_TUPLE_IP])
718                 return -EINVAL;
719
720         tuple->src.l3num = l3num;
721
722         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple);
723         if (err < 0)
724                 return err;
725
726         if (!tb[CTA_TUPLE_PROTO])
727                 return -EINVAL;
728
729         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple);
730         if (err < 0)
731                 return err;
732
733         /* orig and expect tuples get DIR_ORIGINAL */
734         if (type == CTA_TUPLE_REPLY)
735                 tuple->dst.dir = IP_CT_DIR_REPLY;
736         else
737                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
738
739         return 0;
740 }
741
742 static inline int
743 ctnetlink_parse_help(struct nlattr *attr, char **helper_name)
744 {
745         struct nlattr *tb[CTA_HELP_MAX+1];
746
747         nla_parse_nested(tb, CTA_HELP_MAX, attr, NULL);
748
749         if (!tb[CTA_HELP_NAME])
750                 return -EINVAL;
751
752         *helper_name = nla_data(tb[CTA_HELP_NAME]);
753
754         return 0;
755 }
756
757 static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
758         [CTA_STATUS]            = { .type = NLA_U32 },
759         [CTA_TIMEOUT]           = { .type = NLA_U32 },
760         [CTA_MARK]              = { .type = NLA_U32 },
761         [CTA_USE]               = { .type = NLA_U32 },
762         [CTA_ID]                = { .type = NLA_U32 },
763 };
764
765 static int
766 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
767                         struct nlmsghdr *nlh, struct nlattr *cda[])
768 {
769         struct nf_conntrack_tuple_hash *h;
770         struct nf_conntrack_tuple tuple;
771         struct nf_conn *ct;
772         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
773         u_int8_t u3 = nfmsg->nfgen_family;
774         int err = 0;
775
776         if (cda[CTA_TUPLE_ORIG])
777                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
778         else if (cda[CTA_TUPLE_REPLY])
779                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
780         else {
781                 /* Flush the whole table */
782                 nf_conntrack_flush_report(&init_net,
783                                          NETLINK_CB(skb).pid,
784                                          nlmsg_report(nlh));
785                 return 0;
786         }
787
788         if (err < 0)
789                 return err;
790
791         h = nf_conntrack_find_get(&init_net, &tuple);
792         if (!h)
793                 return -ENOENT;
794
795         ct = nf_ct_tuplehash_to_ctrack(h);
796
797         if (cda[CTA_ID]) {
798                 u_int32_t id = ntohl(nla_get_be32(cda[CTA_ID]));
799                 if (id != (u32)(unsigned long)ct) {
800                         nf_ct_put(ct);
801                         return -ENOENT;
802                 }
803         }
804
805         if (nf_conntrack_event_report(IPCT_DESTROY, ct,
806                                       NETLINK_CB(skb).pid,
807                                       nlmsg_report(nlh)) < 0) {
808                 nf_ct_delete_from_lists(ct);
809                 /* we failed to report the event, try later */
810                 nf_ct_insert_dying_list(ct);
811                 nf_ct_put(ct);
812                 return 0;
813         }
814
815         /* death_by_timeout would report the event again */
816         set_bit(IPS_DYING_BIT, &ct->status);
817
818         nf_ct_kill(ct);
819         nf_ct_put(ct);
820
821         return 0;
822 }
823
824 static int
825 ctnetlink_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(&init_net, &tuple);
851         if (!h)
852                 return -ENOENT;
853
854         ct = nf_ct_tuplehash_to_ctrack(h);
855
856         err = -ENOMEM;
857         skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
858         if (skb2 == NULL) {
859                 nf_ct_put(ct);
860                 return -ENOMEM;
861         }
862
863         rcu_read_lock();
864         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
865                                   IPCTNL_MSG_CT_NEW, ct);
866         rcu_read_unlock();
867         nf_ct_put(ct);
868         if (err <= 0)
869                 goto free;
870
871         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
872         if (err < 0)
873                 goto out;
874
875         return 0;
876
877 free:
878         kfree_skb(skb2);
879 out:
880         return err;
881 }
882
883 #ifdef CONFIG_NF_NAT_NEEDED
884 static int
885 ctnetlink_parse_nat_setup(struct nf_conn *ct,
886                           enum nf_nat_manip_type manip,
887                           struct nlattr *attr)
888 {
889         typeof(nfnetlink_parse_nat_setup_hook) parse_nat_setup;
890
891         parse_nat_setup = rcu_dereference(nfnetlink_parse_nat_setup_hook);
892         if (!parse_nat_setup) {
893 #ifdef CONFIG_MODULES
894                 rcu_read_unlock();
895                 spin_unlock_bh(&nf_conntrack_lock);
896                 nfnl_unlock();
897                 if (request_module("nf-nat-ipv4") < 0) {
898                         nfnl_lock();
899                         spin_lock_bh(&nf_conntrack_lock);
900                         rcu_read_lock();
901                         return -EOPNOTSUPP;
902                 }
903                 nfnl_lock();
904                 spin_lock_bh(&nf_conntrack_lock);
905                 rcu_read_lock();
906                 if (nfnetlink_parse_nat_setup_hook)
907                         return -EAGAIN;
908 #endif
909                 return -EOPNOTSUPP;
910         }
911
912         return parse_nat_setup(ct, manip, attr);
913 }
914 #endif
915
916 static int
917 ctnetlink_change_status(struct nf_conn *ct, struct nlattr *cda[])
918 {
919         unsigned long d;
920         unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
921         d = ct->status ^ status;
922
923         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
924                 /* unchangeable */
925                 return -EBUSY;
926
927         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
928                 /* SEEN_REPLY bit can only be set */
929                 return -EBUSY;
930
931         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
932                 /* ASSURED bit can only be set */
933                 return -EBUSY;
934
935         /* Be careful here, modifying NAT bits can screw up things,
936          * so don't let users modify them directly if they don't pass
937          * nf_nat_range. */
938         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
939         return 0;
940 }
941
942 static int
943 ctnetlink_change_nat(struct nf_conn *ct, struct nlattr *cda[])
944 {
945 #ifdef CONFIG_NF_NAT_NEEDED
946         int ret;
947
948         if (cda[CTA_NAT_DST]) {
949                 ret = ctnetlink_parse_nat_setup(ct,
950                                                 IP_NAT_MANIP_DST,
951                                                 cda[CTA_NAT_DST]);
952                 if (ret < 0)
953                         return ret;
954         }
955         if (cda[CTA_NAT_SRC]) {
956                 ret = ctnetlink_parse_nat_setup(ct,
957                                                 IP_NAT_MANIP_SRC,
958                                                 cda[CTA_NAT_SRC]);
959                 if (ret < 0)
960                         return ret;
961         }
962         return 0;
963 #else
964         return -EOPNOTSUPP;
965 #endif
966 }
967
968 static inline int
969 ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
970 {
971         struct nf_conntrack_helper *helper;
972         struct nf_conn_help *help = nfct_help(ct);
973         char *helpname = NULL;
974         int err;
975
976         /* don't change helper of sibling connections */
977         if (ct->master)
978                 return -EBUSY;
979
980         err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
981         if (err < 0)
982                 return err;
983
984         if (!strcmp(helpname, "")) {
985                 if (help && help->helper) {
986                         /* we had a helper before ... */
987                         nf_ct_remove_expectations(ct);
988                         rcu_assign_pointer(help->helper, NULL);
989                 }
990
991                 return 0;
992         }
993
994         helper = __nf_conntrack_helper_find_byname(helpname);
995         if (helper == NULL) {
996 #ifdef CONFIG_MODULES
997                 spin_unlock_bh(&nf_conntrack_lock);
998
999                 if (request_module("nfct-helper-%s", helpname) < 0) {
1000                         spin_lock_bh(&nf_conntrack_lock);
1001                         return -EOPNOTSUPP;
1002                 }
1003
1004                 spin_lock_bh(&nf_conntrack_lock);
1005                 helper = __nf_conntrack_helper_find_byname(helpname);
1006                 if (helper)
1007                         return -EAGAIN;
1008 #endif
1009                 return -EOPNOTSUPP;
1010         }
1011
1012         if (help) {
1013                 if (help->helper == helper)
1014                         return 0;
1015                 if (help->helper)
1016                         return -EBUSY;
1017                 /* need to zero data of old helper */
1018                 memset(&help->help, 0, sizeof(help->help));
1019         } else {
1020                 help = nf_ct_helper_ext_add(ct, GFP_ATOMIC);
1021                 if (help == NULL)
1022                         return -ENOMEM;
1023         }
1024
1025         rcu_assign_pointer(help->helper, helper);
1026
1027         return 0;
1028 }
1029
1030 static inline int
1031 ctnetlink_change_timeout(struct nf_conn *ct, struct nlattr *cda[])
1032 {
1033         u_int32_t timeout = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
1034
1035         if (!del_timer(&ct->timeout))
1036                 return -ETIME;
1037
1038         ct->timeout.expires = jiffies + timeout * HZ;
1039         add_timer(&ct->timeout);
1040
1041         return 0;
1042 }
1043
1044 static inline int
1045 ctnetlink_change_protoinfo(struct nf_conn *ct, struct nlattr *cda[])
1046 {
1047         struct nlattr *tb[CTA_PROTOINFO_MAX+1], *attr = cda[CTA_PROTOINFO];
1048         struct nf_conntrack_l4proto *l4proto;
1049         int err = 0;
1050
1051         nla_parse_nested(tb, CTA_PROTOINFO_MAX, attr, NULL);
1052
1053         rcu_read_lock();
1054         l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
1055         if (l4proto->from_nlattr)
1056                 err = l4proto->from_nlattr(tb, ct);
1057         rcu_read_unlock();
1058
1059         return err;
1060 }
1061
1062 #ifdef CONFIG_NF_NAT_NEEDED
1063 static inline int
1064 change_nat_seq_adj(struct nf_nat_seq *natseq, struct nlattr *attr)
1065 {
1066         struct nlattr *cda[CTA_NAT_SEQ_MAX+1];
1067
1068         nla_parse_nested(cda, CTA_NAT_SEQ_MAX, attr, NULL);
1069
1070         if (!cda[CTA_NAT_SEQ_CORRECTION_POS])
1071                 return -EINVAL;
1072
1073         natseq->correction_pos =
1074                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_CORRECTION_POS]));
1075
1076         if (!cda[CTA_NAT_SEQ_OFFSET_BEFORE])
1077                 return -EINVAL;
1078
1079         natseq->offset_before =
1080                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_BEFORE]));
1081
1082         if (!cda[CTA_NAT_SEQ_OFFSET_AFTER])
1083                 return -EINVAL;
1084
1085         natseq->offset_after =
1086                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_AFTER]));
1087
1088         return 0;
1089 }
1090
1091 static int
1092 ctnetlink_change_nat_seq_adj(struct nf_conn *ct, struct nlattr *cda[])
1093 {
1094         int ret = 0;
1095         struct nf_conn_nat *nat = nfct_nat(ct);
1096
1097         if (!nat)
1098                 return 0;
1099
1100         if (cda[CTA_NAT_SEQ_ADJ_ORIG]) {
1101                 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_ORIGINAL],
1102                                          cda[CTA_NAT_SEQ_ADJ_ORIG]);
1103                 if (ret < 0)
1104                         return ret;
1105
1106                 ct->status |= IPS_SEQ_ADJUST;
1107         }
1108
1109         if (cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1110                 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_REPLY],
1111                                          cda[CTA_NAT_SEQ_ADJ_REPLY]);
1112                 if (ret < 0)
1113                         return ret;
1114
1115                 ct->status |= IPS_SEQ_ADJUST;
1116         }
1117
1118         return 0;
1119 }
1120 #endif
1121
1122 static int
1123 ctnetlink_change_conntrack(struct nf_conn *ct, struct nlattr *cda[])
1124 {
1125         int err;
1126
1127         /* only allow NAT changes and master assignation for new conntracks */
1128         if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST] || cda[CTA_TUPLE_MASTER])
1129                 return -EOPNOTSUPP;
1130
1131         if (cda[CTA_HELP]) {
1132                 err = ctnetlink_change_helper(ct, cda);
1133                 if (err < 0)
1134                         return err;
1135         }
1136
1137         if (cda[CTA_TIMEOUT]) {
1138                 err = ctnetlink_change_timeout(ct, cda);
1139                 if (err < 0)
1140                         return err;
1141         }
1142
1143         if (cda[CTA_STATUS]) {
1144                 err = ctnetlink_change_status(ct, cda);
1145                 if (err < 0)
1146                         return err;
1147         }
1148
1149         if (cda[CTA_PROTOINFO]) {
1150                 err = ctnetlink_change_protoinfo(ct, cda);
1151                 if (err < 0)
1152                         return err;
1153         }
1154
1155 #if defined(CONFIG_NF_CONNTRACK_MARK)
1156         if (cda[CTA_MARK])
1157                 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1158 #endif
1159
1160 #ifdef CONFIG_NF_NAT_NEEDED
1161         if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1162                 err = ctnetlink_change_nat_seq_adj(ct, cda);
1163                 if (err < 0)
1164                         return err;
1165         }
1166 #endif
1167
1168         return 0;
1169 }
1170
1171 static struct nf_conn *
1172 ctnetlink_create_conntrack(struct nlattr *cda[],
1173                            struct nf_conntrack_tuple *otuple,
1174                            struct nf_conntrack_tuple *rtuple,
1175                            u8 u3)
1176 {
1177         struct nf_conn *ct;
1178         int err = -EINVAL;
1179         struct nf_conntrack_helper *helper;
1180
1181         ct = nf_conntrack_alloc(&init_net, otuple, rtuple, GFP_ATOMIC);
1182         if (IS_ERR(ct))
1183                 return ERR_PTR(-ENOMEM);
1184
1185         if (!cda[CTA_TIMEOUT])
1186                 goto err1;
1187         ct->timeout.expires = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
1188
1189         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1190         ct->status |= IPS_CONFIRMED;
1191
1192         rcu_read_lock();
1193         if (cda[CTA_HELP]) {
1194                 char *helpname = NULL;
1195  
1196                 err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
1197                 if (err < 0)
1198                         goto err2;
1199
1200                 helper = __nf_conntrack_helper_find_byname(helpname);
1201                 if (helper == NULL) {
1202                         rcu_read_unlock();
1203 #ifdef CONFIG_MODULES
1204                         if (request_module("nfct-helper-%s", helpname) < 0) {
1205                                 err = -EOPNOTSUPP;
1206                                 goto err1;
1207                         }
1208
1209                         rcu_read_lock();
1210                         helper = __nf_conntrack_helper_find_byname(helpname);
1211                         if (helper) {
1212                                 err = -EAGAIN;
1213                                 goto err2;
1214                         }
1215                         rcu_read_unlock();
1216 #endif
1217                         err = -EOPNOTSUPP;
1218                         goto err1;
1219                 } else {
1220                         struct nf_conn_help *help;
1221
1222                         help = nf_ct_helper_ext_add(ct, GFP_ATOMIC);
1223                         if (help == NULL) {
1224                                 err = -ENOMEM;
1225                                 goto err2;
1226                         }
1227
1228                         /* not in hash table yet so not strictly necessary */
1229                         rcu_assign_pointer(help->helper, helper);
1230                 }
1231         } else {
1232                 /* try an implicit helper assignation */
1233                 err = __nf_ct_try_assign_helper(ct, GFP_ATOMIC);
1234                 if (err < 0)
1235                         goto err2;
1236         }
1237
1238         if (cda[CTA_STATUS]) {
1239                 err = ctnetlink_change_status(ct, cda);
1240                 if (err < 0)
1241                         goto err2;
1242         }
1243
1244         if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
1245                 err = ctnetlink_change_nat(ct, cda);
1246                 if (err < 0)
1247                         goto err2;
1248         }
1249
1250 #ifdef CONFIG_NF_NAT_NEEDED
1251         if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1252                 err = ctnetlink_change_nat_seq_adj(ct, cda);
1253                 if (err < 0)
1254                         goto err2;
1255         }
1256 #endif
1257
1258         if (cda[CTA_PROTOINFO]) {
1259                 err = ctnetlink_change_protoinfo(ct, cda);
1260                 if (err < 0)
1261                         goto err2;
1262         }
1263
1264         nf_ct_acct_ext_add(ct, GFP_ATOMIC);
1265         nf_ct_ecache_ext_add(ct, GFP_ATOMIC);
1266
1267 #if defined(CONFIG_NF_CONNTRACK_MARK)
1268         if (cda[CTA_MARK])
1269                 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1270 #endif
1271
1272         /* setup master conntrack: this is a confirmed expectation */
1273         if (cda[CTA_TUPLE_MASTER]) {
1274                 struct nf_conntrack_tuple master;
1275                 struct nf_conntrack_tuple_hash *master_h;
1276                 struct nf_conn *master_ct;
1277
1278                 err = ctnetlink_parse_tuple(cda, &master, CTA_TUPLE_MASTER, u3);
1279                 if (err < 0)
1280                         goto err2;
1281
1282                 master_h = nf_conntrack_find_get(&init_net, &master);
1283                 if (master_h == NULL) {
1284                         err = -ENOENT;
1285                         goto err2;
1286                 }
1287                 master_ct = nf_ct_tuplehash_to_ctrack(master_h);
1288                 __set_bit(IPS_EXPECTED_BIT, &ct->status);
1289                 ct->master = master_ct;
1290         }
1291
1292         add_timer(&ct->timeout);
1293         nf_conntrack_hash_insert(ct);
1294         rcu_read_unlock();
1295
1296         return ct;
1297
1298 err2:
1299         rcu_read_unlock();
1300 err1:
1301         nf_conntrack_free(ct);
1302         return ERR_PTR(err);
1303 }
1304
1305 static int
1306 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1307                         struct nlmsghdr *nlh, struct nlattr *cda[])
1308 {
1309         struct nf_conntrack_tuple otuple, rtuple;
1310         struct nf_conntrack_tuple_hash *h = NULL;
1311         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1312         u_int8_t u3 = nfmsg->nfgen_family;
1313         int err = 0;
1314
1315         if (cda[CTA_TUPLE_ORIG]) {
1316                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG, u3);
1317                 if (err < 0)
1318                         return err;
1319         }
1320
1321         if (cda[CTA_TUPLE_REPLY]) {
1322                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY, u3);
1323                 if (err < 0)
1324                         return err;
1325         }
1326
1327         spin_lock_bh(&nf_conntrack_lock);
1328         if (cda[CTA_TUPLE_ORIG])
1329                 h = __nf_conntrack_find(&init_net, &otuple);
1330         else if (cda[CTA_TUPLE_REPLY])
1331                 h = __nf_conntrack_find(&init_net, &rtuple);
1332
1333         if (h == NULL) {
1334                 err = -ENOENT;
1335                 if (nlh->nlmsg_flags & NLM_F_CREATE) {
1336                         struct nf_conn *ct;
1337                         enum ip_conntrack_events events;
1338
1339                         ct = ctnetlink_create_conntrack(cda, &otuple,
1340                                                         &rtuple, u3);
1341                         if (IS_ERR(ct)) {
1342                                 err = PTR_ERR(ct);
1343                                 goto out_unlock;
1344                         }
1345                         err = 0;
1346                         nf_conntrack_get(&ct->ct_general);
1347                         spin_unlock_bh(&nf_conntrack_lock);
1348                         if (test_bit(IPS_EXPECTED_BIT, &ct->status))
1349                                 events = IPCT_RELATED;
1350                         else
1351                                 events = IPCT_NEW;
1352
1353                         nf_conntrack_eventmask_report((1 << IPCT_STATUS) |
1354                                                       (1 << IPCT_HELPER) |
1355                                                       (1 << IPCT_PROTOINFO) |
1356                                                       (1 << IPCT_NATSEQADJ) |
1357                                                       (1 << IPCT_MARK) | events,
1358                                                       ct, NETLINK_CB(skb).pid,
1359                                                       nlmsg_report(nlh));
1360                         nf_ct_put(ct);
1361                 } else
1362                         spin_unlock_bh(&nf_conntrack_lock);
1363
1364                 return err;
1365         }
1366         /* implicit 'else' */
1367
1368         /* We manipulate the conntrack inside the global conntrack table lock,
1369          * so there's no need to increase the refcount */
1370         err = -EEXIST;
1371         if (!(nlh->nlmsg_flags & NLM_F_EXCL)) {
1372                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
1373
1374                 err = ctnetlink_change_conntrack(ct, cda);
1375                 if (err == 0) {
1376                         nf_conntrack_get(&ct->ct_general);
1377                         spin_unlock_bh(&nf_conntrack_lock);
1378                         nf_conntrack_eventmask_report((1 << IPCT_STATUS) |
1379                                                       (1 << IPCT_HELPER) |
1380                                                       (1 << IPCT_PROTOINFO) |
1381                                                       (1 << IPCT_NATSEQADJ) |
1382                                                       (1 << IPCT_MARK),
1383                                                       ct, NETLINK_CB(skb).pid,
1384                                                       nlmsg_report(nlh));
1385                         nf_ct_put(ct);
1386                 } else
1387                         spin_unlock_bh(&nf_conntrack_lock);
1388
1389                 return err;
1390         }
1391
1392 out_unlock:
1393         spin_unlock_bh(&nf_conntrack_lock);
1394         return err;
1395 }
1396
1397 /***********************************************************************
1398  * EXPECT
1399  ***********************************************************************/
1400
1401 static inline int
1402 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1403                          const struct nf_conntrack_tuple *tuple,
1404                          enum ctattr_expect type)
1405 {
1406         struct nlattr *nest_parms;
1407
1408         nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
1409         if (!nest_parms)
1410                 goto nla_put_failure;
1411         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1412                 goto nla_put_failure;
1413         nla_nest_end(skb, nest_parms);
1414
1415         return 0;
1416
1417 nla_put_failure:
1418         return -1;
1419 }
1420
1421 static inline int
1422 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1423                         const struct nf_conntrack_tuple *tuple,
1424                         const struct nf_conntrack_tuple_mask *mask)
1425 {
1426         int ret;
1427         struct nf_conntrack_l3proto *l3proto;
1428         struct nf_conntrack_l4proto *l4proto;
1429         struct nf_conntrack_tuple m;
1430         struct nlattr *nest_parms;
1431
1432         memset(&m, 0xFF, sizeof(m));
1433         m.src.u.all = mask->src.u.all;
1434         memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3));
1435
1436         nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK | NLA_F_NESTED);
1437         if (!nest_parms)
1438                 goto nla_put_failure;
1439
1440         l3proto = __nf_ct_l3proto_find(tuple->src.l3num);
1441         ret = ctnetlink_dump_tuples_ip(skb, &m, l3proto);
1442
1443         if (unlikely(ret < 0))
1444                 goto nla_put_failure;
1445
1446         l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
1447         ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto);
1448         if (unlikely(ret < 0))
1449                 goto nla_put_failure;
1450
1451         nla_nest_end(skb, nest_parms);
1452
1453         return 0;
1454
1455 nla_put_failure:
1456         return -1;
1457 }
1458
1459 static int
1460 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1461                           const struct nf_conntrack_expect *exp)
1462 {
1463         struct nf_conn *master = exp->master;
1464         long timeout = (exp->timeout.expires - jiffies) / HZ;
1465
1466         if (timeout < 0)
1467                 timeout = 0;
1468
1469         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1470                 goto nla_put_failure;
1471         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1472                 goto nla_put_failure;
1473         if (ctnetlink_exp_dump_tuple(skb,
1474                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1475                                  CTA_EXPECT_MASTER) < 0)
1476                 goto nla_put_failure;
1477
1478         NLA_PUT_BE32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout));
1479         NLA_PUT_BE32(skb, CTA_EXPECT_ID, htonl((unsigned long)exp));
1480
1481         return 0;
1482
1483 nla_put_failure:
1484         return -1;
1485 }
1486
1487 static int
1488 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1489                         int event, const struct nf_conntrack_expect *exp)
1490 {
1491         struct nlmsghdr *nlh;
1492         struct nfgenmsg *nfmsg;
1493         unsigned int flags = pid ? NLM_F_MULTI : 0;
1494
1495         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1496         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*nfmsg), flags);
1497         if (nlh == NULL)
1498                 goto nlmsg_failure;
1499
1500         nfmsg = nlmsg_data(nlh);
1501         nfmsg->nfgen_family = exp->tuple.src.l3num;
1502         nfmsg->version      = NFNETLINK_V0;
1503         nfmsg->res_id       = 0;
1504
1505         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1506                 goto nla_put_failure;
1507
1508         nlmsg_end(skb, nlh);
1509         return skb->len;
1510
1511 nlmsg_failure:
1512 nla_put_failure:
1513         nlmsg_cancel(skb, nlh);
1514         return -1;
1515 }
1516
1517 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1518 static int
1519 ctnetlink_expect_event(unsigned int events, struct nf_exp_event *item)
1520 {
1521         struct nlmsghdr *nlh;
1522         struct nfgenmsg *nfmsg;
1523         struct nf_conntrack_expect *exp = item->exp;
1524         struct sk_buff *skb;
1525         unsigned int type;
1526         int flags = 0;
1527
1528         if (events & (1 << IPEXP_NEW)) {
1529                 type = IPCTNL_MSG_EXP_NEW;
1530                 flags = NLM_F_CREATE|NLM_F_EXCL;
1531         } else
1532                 return 0;
1533
1534         if (!item->report &&
1535             !nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1536                 return 0;
1537
1538         skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1539         if (skb == NULL)
1540                 goto errout;
1541
1542         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1543         nlh = nlmsg_put(skb, item->pid, 0, type, sizeof(*nfmsg), flags);
1544         if (nlh == NULL)
1545                 goto nlmsg_failure;
1546
1547         nfmsg = nlmsg_data(nlh);
1548         nfmsg->nfgen_family = exp->tuple.src.l3num;
1549         nfmsg->version      = NFNETLINK_V0;
1550         nfmsg->res_id       = 0;
1551
1552         rcu_read_lock();
1553         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1554                 goto nla_put_failure;
1555         rcu_read_unlock();
1556
1557         nlmsg_end(skb, nlh);
1558         nfnetlink_send(skb, item->pid, NFNLGRP_CONNTRACK_EXP_NEW,
1559                        item->report, GFP_ATOMIC);
1560         return 0;
1561
1562 nla_put_failure:
1563         rcu_read_unlock();
1564         nlmsg_cancel(skb, nlh);
1565 nlmsg_failure:
1566         kfree_skb(skb);
1567 errout:
1568         nfnetlink_set_err(0, 0, -ENOBUFS);
1569         return 0;
1570 }
1571 #endif
1572 static int ctnetlink_exp_done(struct netlink_callback *cb)
1573 {
1574         if (cb->args[1])
1575                 nf_ct_expect_put((struct nf_conntrack_expect *)cb->args[1]);
1576         return 0;
1577 }
1578
1579 static int
1580 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1581 {
1582         struct net *net = &init_net;
1583         struct nf_conntrack_expect *exp, *last;
1584         struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
1585         struct hlist_node *n;
1586         u_int8_t l3proto = nfmsg->nfgen_family;
1587
1588         rcu_read_lock();
1589         last = (struct nf_conntrack_expect *)cb->args[1];
1590         for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) {
1591 restart:
1592                 hlist_for_each_entry(exp, n, &net->ct.expect_hash[cb->args[0]],
1593                                      hnode) {
1594                         if (l3proto && exp->tuple.src.l3num != l3proto)
1595                                 continue;
1596                         if (cb->args[1]) {
1597                                 if (exp != last)
1598                                         continue;
1599                                 cb->args[1] = 0;
1600                         }
1601                         if (ctnetlink_exp_fill_info(skb,
1602                                                     NETLINK_CB(cb->skb).pid,
1603                                                     cb->nlh->nlmsg_seq,
1604                                                     IPCTNL_MSG_EXP_NEW,
1605                                                     exp) < 0) {
1606                                 if (!atomic_inc_not_zero(&exp->use))
1607                                         continue;
1608                                 cb->args[1] = (unsigned long)exp;
1609                                 goto out;
1610                         }
1611                 }
1612                 if (cb->args[1]) {
1613                         cb->args[1] = 0;
1614                         goto restart;
1615                 }
1616         }
1617 out:
1618         rcu_read_unlock();
1619         if (last)
1620                 nf_ct_expect_put(last);
1621
1622         return skb->len;
1623 }
1624
1625 static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = {
1626         [CTA_EXPECT_TIMEOUT]    = { .type = NLA_U32 },
1627         [CTA_EXPECT_ID]         = { .type = NLA_U32 },
1628 };
1629
1630 static int
1631 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1632                      struct nlmsghdr *nlh, struct nlattr *cda[])
1633 {
1634         struct nf_conntrack_tuple tuple;
1635         struct nf_conntrack_expect *exp;
1636         struct sk_buff *skb2;
1637         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1638         u_int8_t u3 = nfmsg->nfgen_family;
1639         int err = 0;
1640
1641         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1642                 return netlink_dump_start(ctnl, skb, nlh,
1643                                           ctnetlink_exp_dump_table,
1644                                           ctnetlink_exp_done);
1645         }
1646
1647         if (cda[CTA_EXPECT_MASTER])
1648                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER, u3);
1649         else
1650                 return -EINVAL;
1651
1652         if (err < 0)
1653                 return err;
1654
1655         exp = nf_ct_expect_find_get(&init_net, &tuple);
1656         if (!exp)
1657                 return -ENOENT;
1658
1659         if (cda[CTA_EXPECT_ID]) {
1660                 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1661                 if (ntohl(id) != (u32)(unsigned long)exp) {
1662                         nf_ct_expect_put(exp);
1663                         return -ENOENT;
1664                 }
1665         }
1666
1667         err = -ENOMEM;
1668         skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1669         if (skb2 == NULL)
1670                 goto out;
1671
1672         rcu_read_lock();
1673         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1674                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW, exp);
1675         rcu_read_unlock();
1676         if (err <= 0)
1677                 goto free;
1678
1679         nf_ct_expect_put(exp);
1680
1681         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1682
1683 free:
1684         kfree_skb(skb2);
1685 out:
1686         nf_ct_expect_put(exp);
1687         return err;
1688 }
1689
1690 static int
1691 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1692                      struct nlmsghdr *nlh, struct nlattr *cda[])
1693 {
1694         struct nf_conntrack_expect *exp;
1695         struct nf_conntrack_tuple tuple;
1696         struct nf_conntrack_helper *h;
1697         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1698         struct hlist_node *n, *next;
1699         u_int8_t u3 = nfmsg->nfgen_family;
1700         unsigned int i;
1701         int err;
1702
1703         if (cda[CTA_EXPECT_TUPLE]) {
1704                 /* delete a single expect by tuple */
1705                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1706                 if (err < 0)
1707                         return err;
1708
1709                 /* bump usage count to 2 */
1710                 exp = nf_ct_expect_find_get(&init_net, &tuple);
1711                 if (!exp)
1712                         return -ENOENT;
1713
1714                 if (cda[CTA_EXPECT_ID]) {
1715                         __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1716                         if (ntohl(id) != (u32)(unsigned long)exp) {
1717                                 nf_ct_expect_put(exp);
1718                                 return -ENOENT;
1719                         }
1720                 }
1721
1722                 /* after list removal, usage count == 1 */
1723                 nf_ct_unexpect_related(exp);
1724                 /* have to put what we 'get' above.
1725                  * after this line usage count == 0 */
1726                 nf_ct_expect_put(exp);
1727         } else if (cda[CTA_EXPECT_HELP_NAME]) {
1728                 char *name = nla_data(cda[CTA_EXPECT_HELP_NAME]);
1729                 struct nf_conn_help *m_help;
1730
1731                 /* delete all expectations for this helper */
1732                 spin_lock_bh(&nf_conntrack_lock);
1733                 h = __nf_conntrack_helper_find_byname(name);
1734                 if (!h) {
1735                         spin_unlock_bh(&nf_conntrack_lock);
1736                         return -EOPNOTSUPP;
1737                 }
1738                 for (i = 0; i < nf_ct_expect_hsize; i++) {
1739                         hlist_for_each_entry_safe(exp, n, next,
1740                                                   &init_net.ct.expect_hash[i],
1741                                                   hnode) {
1742                                 m_help = nfct_help(exp->master);
1743                                 if (m_help->helper == h
1744                                     && del_timer(&exp->timeout)) {
1745                                         nf_ct_unlink_expect(exp);
1746                                         nf_ct_expect_put(exp);
1747                                 }
1748                         }
1749                 }
1750                 spin_unlock_bh(&nf_conntrack_lock);
1751         } else {
1752                 /* This basically means we have to flush everything*/
1753                 spin_lock_bh(&nf_conntrack_lock);
1754                 for (i = 0; i < nf_ct_expect_hsize; i++) {
1755                         hlist_for_each_entry_safe(exp, n, next,
1756                                                   &init_net.ct.expect_hash[i],
1757                                                   hnode) {
1758                                 if (del_timer(&exp->timeout)) {
1759                                         nf_ct_unlink_expect(exp);
1760                                         nf_ct_expect_put(exp);
1761                                 }
1762                         }
1763                 }
1764                 spin_unlock_bh(&nf_conntrack_lock);
1765         }
1766
1767         return 0;
1768 }
1769 static int
1770 ctnetlink_change_expect(struct nf_conntrack_expect *x, struct nlattr *cda[])
1771 {
1772         return -EOPNOTSUPP;
1773 }
1774
1775 static int
1776 ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3, u32 pid, int report)
1777 {
1778         struct nf_conntrack_tuple tuple, mask, master_tuple;
1779         struct nf_conntrack_tuple_hash *h = NULL;
1780         struct nf_conntrack_expect *exp;
1781         struct nf_conn *ct;
1782         struct nf_conn_help *help;
1783         int err = 0;
1784
1785         /* caller guarantees that those three CTA_EXPECT_* exist */
1786         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1787         if (err < 0)
1788                 return err;
1789         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK, u3);
1790         if (err < 0)
1791                 return err;
1792         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER, u3);
1793         if (err < 0)
1794                 return err;
1795
1796         /* Look for master conntrack of this expectation */
1797         h = nf_conntrack_find_get(&init_net, &master_tuple);
1798         if (!h)
1799                 return -ENOENT;
1800         ct = nf_ct_tuplehash_to_ctrack(h);
1801         help = nfct_help(ct);
1802
1803         if (!help || !help->helper) {
1804                 /* such conntrack hasn't got any helper, abort */
1805                 err = -EOPNOTSUPP;
1806                 goto out;
1807         }
1808
1809         exp = nf_ct_expect_alloc(ct);
1810         if (!exp) {
1811                 err = -ENOMEM;
1812                 goto out;
1813         }
1814
1815         exp->class = 0;
1816         exp->expectfn = NULL;
1817         exp->flags = 0;
1818         exp->master = ct;
1819         exp->helper = NULL;
1820         memcpy(&exp->tuple, &tuple, sizeof(struct nf_conntrack_tuple));
1821         memcpy(&exp->mask.src.u3, &mask.src.u3, sizeof(exp->mask.src.u3));
1822         exp->mask.src.u.all = mask.src.u.all;
1823
1824         err = nf_ct_expect_related_report(exp, pid, report);
1825         nf_ct_expect_put(exp);
1826
1827 out:
1828         nf_ct_put(nf_ct_tuplehash_to_ctrack(h));
1829         return err;
1830 }
1831
1832 static int
1833 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1834                      struct nlmsghdr *nlh, struct nlattr *cda[])
1835 {
1836         struct nf_conntrack_tuple tuple;
1837         struct nf_conntrack_expect *exp;
1838         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1839         u_int8_t u3 = nfmsg->nfgen_family;
1840         int err = 0;
1841
1842         if (!cda[CTA_EXPECT_TUPLE]
1843             || !cda[CTA_EXPECT_MASK]
1844             || !cda[CTA_EXPECT_MASTER])
1845                 return -EINVAL;
1846
1847         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1848         if (err < 0)
1849                 return err;
1850
1851         spin_lock_bh(&nf_conntrack_lock);
1852         exp = __nf_ct_expect_find(&init_net, &tuple);
1853
1854         if (!exp) {
1855                 spin_unlock_bh(&nf_conntrack_lock);
1856                 err = -ENOENT;
1857                 if (nlh->nlmsg_flags & NLM_F_CREATE) {
1858                         err = ctnetlink_create_expect(cda,
1859                                                       u3,
1860                                                       NETLINK_CB(skb).pid,
1861                                                       nlmsg_report(nlh));
1862                 }
1863                 return err;
1864         }
1865
1866         err = -EEXIST;
1867         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1868                 err = ctnetlink_change_expect(exp, cda);
1869         spin_unlock_bh(&nf_conntrack_lock);
1870
1871         return err;
1872 }
1873
1874 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1875 static struct nf_ct_event_notifier ctnl_notifier = {
1876         .fcn = ctnetlink_conntrack_event,
1877 };
1878
1879 static struct nf_exp_event_notifier ctnl_notifier_exp = {
1880         .fcn = ctnetlink_expect_event,
1881 };
1882 #endif
1883
1884 static const struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1885         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1886                                             .attr_count = CTA_MAX,
1887                                             .policy = ct_nla_policy },
1888         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1889                                             .attr_count = CTA_MAX,
1890                                             .policy = ct_nla_policy },
1891         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1892                                             .attr_count = CTA_MAX,
1893                                             .policy = ct_nla_policy },
1894         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1895                                             .attr_count = CTA_MAX,
1896                                             .policy = ct_nla_policy },
1897 };
1898
1899 static const struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1900         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1901                                             .attr_count = CTA_EXPECT_MAX,
1902                                             .policy = exp_nla_policy },
1903         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1904                                             .attr_count = CTA_EXPECT_MAX,
1905                                             .policy = exp_nla_policy },
1906         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1907                                             .attr_count = CTA_EXPECT_MAX,
1908                                             .policy = exp_nla_policy },
1909 };
1910
1911 static const struct nfnetlink_subsystem ctnl_subsys = {
1912         .name                           = "conntrack",
1913         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1914         .cb_count                       = IPCTNL_MSG_MAX,
1915         .cb                             = ctnl_cb,
1916 };
1917
1918 static const struct nfnetlink_subsystem ctnl_exp_subsys = {
1919         .name                           = "conntrack_expect",
1920         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1921         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1922         .cb                             = ctnl_exp_cb,
1923 };
1924
1925 MODULE_ALIAS("ip_conntrack_netlink");
1926 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1927 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1928
1929 static int __init ctnetlink_init(void)
1930 {
1931         int ret;
1932
1933         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1934         ret = nfnetlink_subsys_register(&ctnl_subsys);
1935         if (ret < 0) {
1936                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1937                 goto err_out;
1938         }
1939
1940         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1941         if (ret < 0) {
1942                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1943                 goto err_unreg_subsys;
1944         }
1945
1946 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1947         ret = nf_conntrack_register_notifier(&ctnl_notifier);
1948         if (ret < 0) {
1949                 printk("ctnetlink_init: cannot register notifier.\n");
1950                 goto err_unreg_exp_subsys;
1951         }
1952
1953         ret = nf_ct_expect_register_notifier(&ctnl_notifier_exp);
1954         if (ret < 0) {
1955                 printk("ctnetlink_init: cannot expect register notifier.\n");
1956                 goto err_unreg_notifier;
1957         }
1958 #endif
1959
1960         return 0;
1961
1962 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1963 err_unreg_notifier:
1964         nf_conntrack_unregister_notifier(&ctnl_notifier);
1965 err_unreg_exp_subsys:
1966         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1967 #endif
1968 err_unreg_subsys:
1969         nfnetlink_subsys_unregister(&ctnl_subsys);
1970 err_out:
1971         return ret;
1972 }
1973
1974 static void __exit ctnetlink_exit(void)
1975 {
1976         printk("ctnetlink: unregistering from nfnetlink.\n");
1977
1978 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1979         nf_ct_expect_unregister_notifier(&ctnl_notifier_exp);
1980         nf_conntrack_unregister_notifier(&ctnl_notifier);
1981 #endif
1982
1983         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1984         nfnetlink_subsys_unregister(&ctnl_subsys);
1985         return;
1986 }
1987
1988 module_init(ctnetlink_init);
1989 module_exit(ctnetlink_exit);