netfilter: ctnetlink: dump conntrack ID in event messages
[linux-2.6.git] / net / netfilter / nf_conntrack_netlink.c
1 /* Connection tracking via netlink socket. Allows for user space
2  * protocol helpers and general trouble making from userspace.
3  *
4  * (C) 2001 by Jay Schulist <jschlst@samba.org>
5  * (C) 2002-2006 by Harald Welte <laforge@gnumonks.org>
6  * (C) 2003 by Patrick Mchardy <kaber@trash.net>
7  * (C) 2005-2007 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/types.h>
22 #include <linux/timer.h>
23 #include <linux/skbuff.h>
24 #include <linux/errno.h>
25 #include <linux/netlink.h>
26 #include <linux/spinlock.h>
27 #include <linux/interrupt.h>
28 #include <linux/notifier.h>
29
30 #include <linux/netfilter.h>
31 #include <net/netlink.h>
32 #include <net/netfilter/nf_conntrack.h>
33 #include <net/netfilter/nf_conntrack_core.h>
34 #include <net/netfilter/nf_conntrack_expect.h>
35 #include <net/netfilter/nf_conntrack_helper.h>
36 #include <net/netfilter/nf_conntrack_l3proto.h>
37 #include <net/netfilter/nf_conntrack_l4proto.h>
38 #include <net/netfilter/nf_conntrack_tuple.h>
39 #ifdef CONFIG_NF_NAT_NEEDED
40 #include <net/netfilter/nf_nat_core.h>
41 #include <net/netfilter/nf_nat_protocol.h>
42 #endif
43
44 #include <linux/netfilter/nfnetlink.h>
45 #include <linux/netfilter/nfnetlink_conntrack.h>
46
47 MODULE_LICENSE("GPL");
48
49 static char __initdata version[] = "0.93";
50
51 static inline int
52 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
53                             const struct nf_conntrack_tuple *tuple,
54                             struct nf_conntrack_l4proto *l4proto)
55 {
56         int ret = 0;
57         struct nlattr *nest_parms;
58
59         nest_parms = nla_nest_start(skb, CTA_TUPLE_PROTO | NLA_F_NESTED);
60         if (!nest_parms)
61                 goto nla_put_failure;
62         NLA_PUT_U8(skb, CTA_PROTO_NUM, tuple->dst.protonum);
63
64         if (likely(l4proto->tuple_to_nlattr))
65                 ret = l4proto->tuple_to_nlattr(skb, tuple);
66
67         nla_nest_end(skb, nest_parms);
68
69         return ret;
70
71 nla_put_failure:
72         return -1;
73 }
74
75 static inline int
76 ctnetlink_dump_tuples_ip(struct sk_buff *skb,
77                          const struct nf_conntrack_tuple *tuple,
78                          struct nf_conntrack_l3proto *l3proto)
79 {
80         int ret = 0;
81         struct nlattr *nest_parms;
82
83         nest_parms = nla_nest_start(skb, CTA_TUPLE_IP | NLA_F_NESTED);
84         if (!nest_parms)
85                 goto nla_put_failure;
86
87         if (likely(l3proto->tuple_to_nlattr))
88                 ret = l3proto->tuple_to_nlattr(skb, tuple);
89
90         nla_nest_end(skb, nest_parms);
91
92         return ret;
93
94 nla_put_failure:
95         return -1;
96 }
97
98 static int
99 ctnetlink_dump_tuples(struct sk_buff *skb,
100                       const struct nf_conntrack_tuple *tuple)
101 {
102         int ret;
103         struct nf_conntrack_l3proto *l3proto;
104         struct nf_conntrack_l4proto *l4proto;
105
106         l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
107         ret = ctnetlink_dump_tuples_ip(skb, tuple, l3proto);
108         nf_ct_l3proto_put(l3proto);
109
110         if (unlikely(ret < 0))
111                 return ret;
112
113         l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
114         ret = ctnetlink_dump_tuples_proto(skb, tuple, l4proto);
115         nf_ct_l4proto_put(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, const 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_get(nf_ct_l3num(ct), nf_ct_protonum(ct));
153         if (!l4proto->to_nlattr) {
154                 nf_ct_l4proto_put(l4proto);
155                 return 0;
156         }
157
158         nest_proto = nla_nest_start(skb, CTA_PROTOINFO | NLA_F_NESTED);
159         if (!nest_proto)
160                 goto nla_put_failure;
161
162         ret = l4proto->to_nlattr(skb, nest_proto, ct);
163
164         nf_ct_l4proto_put(l4proto);
165
166         nla_nest_end(skb, nest_proto);
167
168         return ret;
169
170 nla_put_failure:
171         nf_ct_l4proto_put(l4proto);
172         return -1;
173 }
174
175 static inline int
176 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct nf_conn *ct)
177 {
178         struct nlattr *nest_helper;
179         const struct nf_conn_help *help = nfct_help(ct);
180         struct nf_conntrack_helper *helper;
181
182         if (!help)
183                 return 0;
184
185         rcu_read_lock();
186         helper = rcu_dereference(help->helper);
187         if (!helper)
188                 goto out;
189
190         nest_helper = nla_nest_start(skb, CTA_HELP | NLA_F_NESTED);
191         if (!nest_helper)
192                 goto nla_put_failure;
193         NLA_PUT_STRING(skb, CTA_HELP_NAME, helper->name);
194
195         if (helper->to_nlattr)
196                 helper->to_nlattr(skb, ct);
197
198         nla_nest_end(skb, nest_helper);
199 out:
200         rcu_read_unlock();
201         return 0;
202
203 nla_put_failure:
204         rcu_read_unlock();
205         return -1;
206 }
207
208 #ifdef CONFIG_NF_CT_ACCT
209 static int
210 ctnetlink_dump_counters(struct sk_buff *skb, const struct nf_conn *ct,
211                         enum ip_conntrack_dir dir)
212 {
213         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
214         struct nlattr *nest_count;
215
216         nest_count = nla_nest_start(skb, type | NLA_F_NESTED);
217         if (!nest_count)
218                 goto nla_put_failure;
219
220         NLA_PUT_BE32(skb, CTA_COUNTERS32_PACKETS,
221                      htonl(ct->counters[dir].packets));
222         NLA_PUT_BE32(skb, CTA_COUNTERS32_BYTES,
223                      htonl(ct->counters[dir].bytes));
224
225         nla_nest_end(skb, nest_count);
226
227         return 0;
228
229 nla_put_failure:
230         return -1;
231 }
232 #else
233 #define ctnetlink_dump_counters(a, b, c) (0)
234 #endif
235
236 #ifdef CONFIG_NF_CONNTRACK_MARK
237 static inline int
238 ctnetlink_dump_mark(struct sk_buff *skb, const struct nf_conn *ct)
239 {
240         NLA_PUT_BE32(skb, CTA_MARK, htonl(ct->mark));
241         return 0;
242
243 nla_put_failure:
244         return -1;
245 }
246 #else
247 #define ctnetlink_dump_mark(a, b) (0)
248 #endif
249
250 #ifdef CONFIG_NF_CONNTRACK_SECMARK
251 static inline int
252 ctnetlink_dump_secmark(struct sk_buff *skb, const struct nf_conn *ct)
253 {
254         NLA_PUT_BE32(skb, CTA_SECMARK, htonl(ct->secmark));
255         return 0;
256
257 nla_put_failure:
258         return -1;
259 }
260 #else
261 #define ctnetlink_dump_secmark(a, b) (0)
262 #endif
263
264 #define master_tuple(ct) &(ct->master->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
265
266 static inline int
267 ctnetlink_dump_master(struct sk_buff *skb, const struct nf_conn *ct)
268 {
269         struct nlattr *nest_parms;
270
271         if (!(ct->status & IPS_EXPECTED))
272                 return 0;
273
274         nest_parms = nla_nest_start(skb, CTA_TUPLE_MASTER | NLA_F_NESTED);
275         if (!nest_parms)
276                 goto nla_put_failure;
277         if (ctnetlink_dump_tuples(skb, master_tuple(ct)) < 0)
278                 goto nla_put_failure;
279         nla_nest_end(skb, nest_parms);
280
281         return 0;
282
283 nla_put_failure:
284         return -1;
285 }
286
287 #ifdef CONFIG_NF_NAT_NEEDED
288 static int
289 dump_nat_seq_adj(struct sk_buff *skb, const struct nf_nat_seq *natseq, int type)
290 {
291         struct nlattr *nest_parms;
292
293         nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
294         if (!nest_parms)
295                 goto nla_put_failure;
296
297         NLA_PUT_BE32(skb, CTA_NAT_SEQ_CORRECTION_POS,
298                      htonl(natseq->correction_pos));
299         NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_BEFORE,
300                      htonl(natseq->offset_before));
301         NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_AFTER,
302                      htonl(natseq->offset_after));
303
304         nla_nest_end(skb, nest_parms);
305
306         return 0;
307
308 nla_put_failure:
309         return -1;
310 }
311
312 static inline int
313 ctnetlink_dump_nat_seq_adj(struct sk_buff *skb, const struct nf_conn *ct)
314 {
315         struct nf_nat_seq *natseq;
316         struct nf_conn_nat *nat = nfct_nat(ct);
317
318         if (!(ct->status & IPS_SEQ_ADJUST) || !nat)
319                 return 0;
320
321         natseq = &nat->seq[IP_CT_DIR_ORIGINAL];
322         if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_ORIG) == -1)
323                 return -1;
324
325         natseq = &nat->seq[IP_CT_DIR_REPLY];
326         if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_REPLY) == -1)
327                 return -1;
328
329         return 0;
330 }
331 #else
332 #define ctnetlink_dump_nat_seq_adj(a, b) (0)
333 #endif
334
335 static inline int
336 ctnetlink_dump_id(struct sk_buff *skb, const struct nf_conn *ct)
337 {
338         NLA_PUT_BE32(skb, CTA_ID, htonl((unsigned long)ct));
339         return 0;
340
341 nla_put_failure:
342         return -1;
343 }
344
345 static inline int
346 ctnetlink_dump_use(struct sk_buff *skb, const struct nf_conn *ct)
347 {
348         NLA_PUT_BE32(skb, CTA_USE, htonl(atomic_read(&ct->ct_general.use)));
349         return 0;
350
351 nla_put_failure:
352         return -1;
353 }
354
355 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
356
357 static int
358 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
359                     int event, int nowait,
360                     const struct nf_conn *ct)
361 {
362         struct nlmsghdr *nlh;
363         struct nfgenmsg *nfmsg;
364         struct nlattr *nest_parms;
365         unsigned char *b = skb_tail_pointer(skb);
366
367         event |= NFNL_SUBSYS_CTNETLINK << 8;
368         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
369         nfmsg  = NLMSG_DATA(nlh);
370
371         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
372         nfmsg->nfgen_family = nf_ct_l3num(ct);
373         nfmsg->version      = NFNETLINK_V0;
374         nfmsg->res_id       = 0;
375
376         nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
377         if (!nest_parms)
378                 goto nla_put_failure;
379         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
380                 goto nla_put_failure;
381         nla_nest_end(skb, nest_parms);
382
383         nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
384         if (!nest_parms)
385                 goto nla_put_failure;
386         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
387                 goto nla_put_failure;
388         nla_nest_end(skb, nest_parms);
389
390         if (ctnetlink_dump_status(skb, ct) < 0 ||
391             ctnetlink_dump_timeout(skb, ct) < 0 ||
392             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
393             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
394             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
395             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
396             ctnetlink_dump_mark(skb, ct) < 0 ||
397             ctnetlink_dump_secmark(skb, ct) < 0 ||
398             ctnetlink_dump_id(skb, ct) < 0 ||
399             ctnetlink_dump_use(skb, ct) < 0 ||
400             ctnetlink_dump_master(skb, ct) < 0 ||
401             ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
402                 goto nla_put_failure;
403
404         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
405         return skb->len;
406
407 nlmsg_failure:
408 nla_put_failure:
409         nlmsg_trim(skb, b);
410         return -1;
411 }
412
413 #ifdef CONFIG_NF_CONNTRACK_EVENTS
414 static int ctnetlink_conntrack_event(struct notifier_block *this,
415                                      unsigned long events, void *ptr)
416 {
417         struct nlmsghdr *nlh;
418         struct nfgenmsg *nfmsg;
419         struct nlattr *nest_parms;
420         struct nf_conn *ct = (struct nf_conn *)ptr;
421         struct sk_buff *skb;
422         unsigned int type;
423         sk_buff_data_t b;
424         unsigned int flags = 0, group;
425
426         /* ignore our fake conntrack entry */
427         if (ct == &nf_conntrack_untracked)
428                 return NOTIFY_DONE;
429
430         if (events & IPCT_DESTROY) {
431                 type = IPCTNL_MSG_CT_DELETE;
432                 group = NFNLGRP_CONNTRACK_DESTROY;
433         } else  if (events & (IPCT_NEW | IPCT_RELATED)) {
434                 type = IPCTNL_MSG_CT_NEW;
435                 flags = NLM_F_CREATE|NLM_F_EXCL;
436                 group = NFNLGRP_CONNTRACK_NEW;
437         } else  if (events & (IPCT_STATUS | IPCT_PROTOINFO)) {
438                 type = IPCTNL_MSG_CT_NEW;
439                 group = NFNLGRP_CONNTRACK_UPDATE;
440         } else
441                 return NOTIFY_DONE;
442
443         if (!nfnetlink_has_listeners(group))
444                 return NOTIFY_DONE;
445
446         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
447         if (!skb)
448                 return NOTIFY_DONE;
449
450         b = skb->tail;
451
452         type |= NFNL_SUBSYS_CTNETLINK << 8;
453         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
454         nfmsg = NLMSG_DATA(nlh);
455
456         nlh->nlmsg_flags    = flags;
457         nfmsg->nfgen_family = nf_ct_l3num(ct);
458         nfmsg->version  = NFNETLINK_V0;
459         nfmsg->res_id   = 0;
460
461         nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
462         if (!nest_parms)
463                 goto nla_put_failure;
464         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
465                 goto nla_put_failure;
466         nla_nest_end(skb, nest_parms);
467
468         nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
469         if (!nest_parms)
470                 goto nla_put_failure;
471         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
472                 goto nla_put_failure;
473         nla_nest_end(skb, nest_parms);
474
475         if (ctnetlink_dump_id(skb, ct) < 0)
476                 goto nla_put_failure;
477
478         if (events & IPCT_DESTROY) {
479                 if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
480                     ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
481                         goto nla_put_failure;
482         } else {
483                 if (ctnetlink_dump_status(skb, ct) < 0)
484                         goto nla_put_failure;
485
486                 if (ctnetlink_dump_timeout(skb, ct) < 0)
487                         goto nla_put_failure;
488
489                 if (events & IPCT_PROTOINFO
490                     && ctnetlink_dump_protoinfo(skb, ct) < 0)
491                         goto nla_put_failure;
492
493                 if ((events & IPCT_HELPER || nfct_help(ct))
494                     && ctnetlink_dump_helpinfo(skb, ct) < 0)
495                         goto nla_put_failure;
496
497 #ifdef CONFIG_NF_CONNTRACK_SECMARK
498                 if ((events & IPCT_SECMARK || ct->secmark)
499                     && ctnetlink_dump_secmark(skb, ct) < 0)
500                         goto nla_put_failure;
501 #endif
502
503                 if (events & IPCT_COUNTER_FILLING &&
504                     (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
505                      ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0))
506                         goto nla_put_failure;
507
508                 if (events & IPCT_RELATED &&
509                     ctnetlink_dump_master(skb, ct) < 0)
510                         goto nla_put_failure;
511
512                 if (events & IPCT_NATSEQADJ &&
513                     ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
514                         goto nla_put_failure;
515         }
516
517 #ifdef CONFIG_NF_CONNTRACK_MARK
518         if ((events & IPCT_MARK || ct->mark)
519             && ctnetlink_dump_mark(skb, ct) < 0)
520                 goto nla_put_failure;
521 #endif
522
523         nlh->nlmsg_len = skb->tail - b;
524         nfnetlink_send(skb, 0, group, 0);
525         return NOTIFY_DONE;
526
527 nlmsg_failure:
528 nla_put_failure:
529         kfree_skb(skb);
530         return NOTIFY_DONE;
531 }
532 #endif /* CONFIG_NF_CONNTRACK_EVENTS */
533
534 static int ctnetlink_done(struct netlink_callback *cb)
535 {
536         if (cb->args[1])
537                 nf_ct_put((struct nf_conn *)cb->args[1]);
538         return 0;
539 }
540
541 static int
542 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
543 {
544         struct nf_conn *ct, *last;
545         struct nf_conntrack_tuple_hash *h;
546         struct hlist_node *n;
547         struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
548         u_int8_t l3proto = nfmsg->nfgen_family;
549
550         rcu_read_lock();
551         last = (struct nf_conn *)cb->args[1];
552         for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
553 restart:
554                 hlist_for_each_entry_rcu(h, n, &nf_conntrack_hash[cb->args[0]],
555                                          hnode) {
556                         if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
557                                 continue;
558                         ct = nf_ct_tuplehash_to_ctrack(h);
559                         /* Dump entries of a given L3 protocol number.
560                          * If it is not specified, ie. l3proto == 0,
561                          * then dump everything. */
562                         if (l3proto && nf_ct_l3num(ct) != l3proto)
563                                 continue;
564                         if (cb->args[1]) {
565                                 if (ct != last)
566                                         continue;
567                                 cb->args[1] = 0;
568                         }
569                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
570                                                 cb->nlh->nlmsg_seq,
571                                                 IPCTNL_MSG_CT_NEW,
572                                                 1, ct) < 0) {
573                                 if (!atomic_inc_not_zero(&ct->ct_general.use))
574                                         continue;
575                                 cb->args[1] = (unsigned long)ct;
576                                 goto out;
577                         }
578 #ifdef CONFIG_NF_CT_ACCT
579                         if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
580                                                 IPCTNL_MSG_CT_GET_CTRZERO)
581                                 memset(&ct->counters, 0, sizeof(ct->counters));
582 #endif
583                 }
584                 if (cb->args[1]) {
585                         cb->args[1] = 0;
586                         goto restart;
587                 }
588         }
589 out:
590         rcu_read_unlock();
591         if (last)
592                 nf_ct_put(last);
593
594         return skb->len;
595 }
596
597 static inline int
598 ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
599 {
600         struct nlattr *tb[CTA_IP_MAX+1];
601         struct nf_conntrack_l3proto *l3proto;
602         int ret = 0;
603
604         nla_parse_nested(tb, CTA_IP_MAX, attr, NULL);
605
606         l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
607
608         if (likely(l3proto->nlattr_to_tuple)) {
609                 ret = nla_validate_nested(attr, CTA_IP_MAX,
610                                           l3proto->nla_policy);
611                 if (ret == 0)
612                         ret = l3proto->nlattr_to_tuple(tb, tuple);
613         }
614
615         nf_ct_l3proto_put(l3proto);
616
617         return ret;
618 }
619
620 static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = {
621         [CTA_PROTO_NUM] = { .type = NLA_U8 },
622 };
623
624 static inline int
625 ctnetlink_parse_tuple_proto(struct nlattr *attr,
626                             struct nf_conntrack_tuple *tuple)
627 {
628         struct nlattr *tb[CTA_PROTO_MAX+1];
629         struct nf_conntrack_l4proto *l4proto;
630         int ret = 0;
631
632         ret = nla_parse_nested(tb, CTA_PROTO_MAX, attr, proto_nla_policy);
633         if (ret < 0)
634                 return ret;
635
636         if (!tb[CTA_PROTO_NUM])
637                 return -EINVAL;
638         tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
639
640         l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
641
642         if (likely(l4proto->nlattr_to_tuple)) {
643                 ret = nla_validate_nested(attr, CTA_PROTO_MAX,
644                                           l4proto->nla_policy);
645                 if (ret == 0)
646                         ret = l4proto->nlattr_to_tuple(tb, tuple);
647         }
648
649         nf_ct_l4proto_put(l4proto);
650
651         return ret;
652 }
653
654 static int
655 ctnetlink_parse_tuple(struct nlattr *cda[], struct nf_conntrack_tuple *tuple,
656                       enum ctattr_tuple type, u_int8_t l3num)
657 {
658         struct nlattr *tb[CTA_TUPLE_MAX+1];
659         int err;
660
661         memset(tuple, 0, sizeof(*tuple));
662
663         nla_parse_nested(tb, CTA_TUPLE_MAX, cda[type], NULL);
664
665         if (!tb[CTA_TUPLE_IP])
666                 return -EINVAL;
667
668         tuple->src.l3num = l3num;
669
670         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple);
671         if (err < 0)
672                 return err;
673
674         if (!tb[CTA_TUPLE_PROTO])
675                 return -EINVAL;
676
677         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple);
678         if (err < 0)
679                 return err;
680
681         /* orig and expect tuples get DIR_ORIGINAL */
682         if (type == CTA_TUPLE_REPLY)
683                 tuple->dst.dir = IP_CT_DIR_REPLY;
684         else
685                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
686
687         return 0;
688 }
689
690 #ifdef CONFIG_NF_NAT_NEEDED
691 static const struct nla_policy protonat_nla_policy[CTA_PROTONAT_MAX+1] = {
692         [CTA_PROTONAT_PORT_MIN] = { .type = NLA_U16 },
693         [CTA_PROTONAT_PORT_MAX] = { .type = NLA_U16 },
694 };
695
696 static int nfnetlink_parse_nat_proto(struct nlattr *attr,
697                                      const struct nf_conn *ct,
698                                      struct nf_nat_range *range)
699 {
700         struct nlattr *tb[CTA_PROTONAT_MAX+1];
701         const struct nf_nat_protocol *npt;
702         int err;
703
704         err = nla_parse_nested(tb, CTA_PROTONAT_MAX, attr, protonat_nla_policy);
705         if (err < 0)
706                 return err;
707
708         npt = nf_nat_proto_find_get(nf_ct_protonum(ct));
709         if (npt->nlattr_to_range)
710                 err = npt->nlattr_to_range(tb, range);
711         nf_nat_proto_put(npt);
712         return err;
713 }
714
715 static const struct nla_policy nat_nla_policy[CTA_NAT_MAX+1] = {
716         [CTA_NAT_MINIP]         = { .type = NLA_U32 },
717         [CTA_NAT_MAXIP]         = { .type = NLA_U32 },
718 };
719
720 static inline int
721 nfnetlink_parse_nat(struct nlattr *nat,
722                     const struct nf_conn *ct, struct nf_nat_range *range)
723 {
724         struct nlattr *tb[CTA_NAT_MAX+1];
725         int err;
726
727         memset(range, 0, sizeof(*range));
728
729         err = nla_parse_nested(tb, CTA_NAT_MAX, nat, nat_nla_policy);
730         if (err < 0)
731                 return err;
732
733         if (tb[CTA_NAT_MINIP])
734                 range->min_ip = nla_get_be32(tb[CTA_NAT_MINIP]);
735
736         if (!tb[CTA_NAT_MAXIP])
737                 range->max_ip = range->min_ip;
738         else
739                 range->max_ip = nla_get_be32(tb[CTA_NAT_MAXIP]);
740
741         if (range->min_ip)
742                 range->flags |= IP_NAT_RANGE_MAP_IPS;
743
744         if (!tb[CTA_NAT_PROTO])
745                 return 0;
746
747         err = nfnetlink_parse_nat_proto(tb[CTA_NAT_PROTO], ct, range);
748         if (err < 0)
749                 return err;
750
751         return 0;
752 }
753 #endif
754
755 static inline int
756 ctnetlink_parse_help(struct nlattr *attr, char **helper_name)
757 {
758         struct nlattr *tb[CTA_HELP_MAX+1];
759
760         nla_parse_nested(tb, CTA_HELP_MAX, attr, NULL);
761
762         if (!tb[CTA_HELP_NAME])
763                 return -EINVAL;
764
765         *helper_name = nla_data(tb[CTA_HELP_NAME]);
766
767         return 0;
768 }
769
770 static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
771         [CTA_STATUS]            = { .type = NLA_U32 },
772         [CTA_TIMEOUT]           = { .type = NLA_U32 },
773         [CTA_MARK]              = { .type = NLA_U32 },
774         [CTA_USE]               = { .type = NLA_U32 },
775         [CTA_ID]                = { .type = NLA_U32 },
776 };
777
778 static int
779 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
780                         struct nlmsghdr *nlh, struct nlattr *cda[])
781 {
782         struct nf_conntrack_tuple_hash *h;
783         struct nf_conntrack_tuple tuple;
784         struct nf_conn *ct;
785         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
786         u_int8_t u3 = nfmsg->nfgen_family;
787         int err = 0;
788
789         if (cda[CTA_TUPLE_ORIG])
790                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
791         else if (cda[CTA_TUPLE_REPLY])
792                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
793         else {
794                 /* Flush the whole table */
795                 nf_conntrack_flush();
796                 return 0;
797         }
798
799         if (err < 0)
800                 return err;
801
802         h = nf_conntrack_find_get(&tuple);
803         if (!h)
804                 return -ENOENT;
805
806         ct = nf_ct_tuplehash_to_ctrack(h);
807
808         if (cda[CTA_ID]) {
809                 u_int32_t id = ntohl(nla_get_be32(cda[CTA_ID]));
810                 if (id != (u32)(unsigned long)ct) {
811                         nf_ct_put(ct);
812                         return -ENOENT;
813                 }
814         }
815         if (del_timer(&ct->timeout))
816                 ct->timeout.function((unsigned long)ct);
817
818         nf_ct_put(ct);
819
820         return 0;
821 }
822
823 static int
824 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
825                         struct nlmsghdr *nlh, struct nlattr *cda[])
826 {
827         struct nf_conntrack_tuple_hash *h;
828         struct nf_conntrack_tuple tuple;
829         struct nf_conn *ct;
830         struct sk_buff *skb2 = NULL;
831         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
832         u_int8_t u3 = nfmsg->nfgen_family;
833         int err = 0;
834
835         if (nlh->nlmsg_flags & NLM_F_DUMP) {
836 #ifndef CONFIG_NF_CT_ACCT
837                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) == IPCTNL_MSG_CT_GET_CTRZERO)
838                         return -ENOTSUPP;
839 #endif
840                 return netlink_dump_start(ctnl, skb, nlh, ctnetlink_dump_table,
841                                           ctnetlink_done);
842         }
843
844         if (cda[CTA_TUPLE_ORIG])
845                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
846         else if (cda[CTA_TUPLE_REPLY])
847                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
848         else
849                 return -EINVAL;
850
851         if (err < 0)
852                 return err;
853
854         h = nf_conntrack_find_get(&tuple);
855         if (!h)
856                 return -ENOENT;
857
858         ct = nf_ct_tuplehash_to_ctrack(h);
859
860         err = -ENOMEM;
861         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
862         if (!skb2) {
863                 nf_ct_put(ct);
864                 return -ENOMEM;
865         }
866
867         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
868                                   IPCTNL_MSG_CT_NEW, 1, ct);
869         nf_ct_put(ct);
870         if (err <= 0)
871                 goto free;
872
873         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
874         if (err < 0)
875                 goto out;
876
877         return 0;
878
879 free:
880         kfree_skb(skb2);
881 out:
882         return err;
883 }
884
885 static int
886 ctnetlink_change_status(struct nf_conn *ct, struct nlattr *cda[])
887 {
888         unsigned long d;
889         unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
890         d = ct->status ^ status;
891
892         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
893                 /* unchangeable */
894                 return -EINVAL;
895
896         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
897                 /* SEEN_REPLY bit can only be set */
898                 return -EINVAL;
899
900
901         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
902                 /* ASSURED bit can only be set */
903                 return -EINVAL;
904
905         if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
906 #ifndef CONFIG_NF_NAT_NEEDED
907                 return -EINVAL;
908 #else
909                 struct nf_nat_range range;
910
911                 if (cda[CTA_NAT_DST]) {
912                         if (nfnetlink_parse_nat(cda[CTA_NAT_DST], ct,
913                                                 &range) < 0)
914                                 return -EINVAL;
915                         if (nf_nat_initialized(ct, IP_NAT_MANIP_DST))
916                                 return -EEXIST;
917                         nf_nat_setup_info(ct, &range, IP_NAT_MANIP_DST);
918                 }
919                 if (cda[CTA_NAT_SRC]) {
920                         if (nfnetlink_parse_nat(cda[CTA_NAT_SRC], ct,
921                                                 &range) < 0)
922                                 return -EINVAL;
923                         if (nf_nat_initialized(ct, IP_NAT_MANIP_SRC))
924                                 return -EEXIST;
925                         nf_nat_setup_info(ct, &range, IP_NAT_MANIP_SRC);
926                 }
927 #endif
928         }
929
930         /* Be careful here, modifying NAT bits can screw up things,
931          * so don't let users modify them directly if they don't pass
932          * nf_nat_range. */
933         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
934         return 0;
935 }
936
937
938 static inline int
939 ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
940 {
941         struct nf_conntrack_helper *helper;
942         struct nf_conn_help *help = nfct_help(ct);
943         char *helpname;
944         int err;
945
946         /* don't change helper of sibling connections */
947         if (ct->master)
948                 return -EINVAL;
949
950         err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
951         if (err < 0)
952                 return err;
953
954         if (!strcmp(helpname, "")) {
955                 if (help && help->helper) {
956                         /* we had a helper before ... */
957                         nf_ct_remove_expectations(ct);
958                         rcu_assign_pointer(help->helper, NULL);
959                 }
960
961                 return 0;
962         }
963
964         helper = __nf_conntrack_helper_find_byname(helpname);
965         if (helper == NULL)
966                 return -EINVAL;
967
968         if (help) {
969                 if (help->helper == helper)
970                         return 0;
971                 if (help->helper)
972                         return -EBUSY;
973                 /* need to zero data of old helper */
974                 memset(&help->help, 0, sizeof(help->help));
975         } else {
976                 help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
977                 if (help == NULL)
978                         return -ENOMEM;
979         }
980
981         rcu_assign_pointer(help->helper, helper);
982
983         return 0;
984 }
985
986 static inline int
987 ctnetlink_change_timeout(struct nf_conn *ct, struct nlattr *cda[])
988 {
989         u_int32_t timeout = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
990
991         if (!del_timer(&ct->timeout))
992                 return -ETIME;
993
994         ct->timeout.expires = jiffies + timeout * HZ;
995         add_timer(&ct->timeout);
996
997         return 0;
998 }
999
1000 static inline int
1001 ctnetlink_change_protoinfo(struct nf_conn *ct, struct nlattr *cda[])
1002 {
1003         struct nlattr *tb[CTA_PROTOINFO_MAX+1], *attr = cda[CTA_PROTOINFO];
1004         struct nf_conntrack_l4proto *l4proto;
1005         int err = 0;
1006
1007         nla_parse_nested(tb, CTA_PROTOINFO_MAX, attr, NULL);
1008
1009         l4proto = nf_ct_l4proto_find_get(nf_ct_l3num(ct), nf_ct_protonum(ct));
1010         if (l4proto->from_nlattr)
1011                 err = l4proto->from_nlattr(tb, ct);
1012         nf_ct_l4proto_put(l4proto);
1013
1014         return err;
1015 }
1016
1017 #ifdef CONFIG_NF_NAT_NEEDED
1018 static inline int
1019 change_nat_seq_adj(struct nf_nat_seq *natseq, struct nlattr *attr)
1020 {
1021         struct nlattr *cda[CTA_NAT_SEQ_MAX+1];
1022
1023         nla_parse_nested(cda, CTA_NAT_SEQ_MAX, attr, NULL);
1024
1025         if (!cda[CTA_NAT_SEQ_CORRECTION_POS])
1026                 return -EINVAL;
1027
1028         natseq->correction_pos =
1029                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_CORRECTION_POS]));
1030
1031         if (!cda[CTA_NAT_SEQ_OFFSET_BEFORE])
1032                 return -EINVAL;
1033
1034         natseq->offset_before =
1035                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_BEFORE]));
1036
1037         if (!cda[CTA_NAT_SEQ_OFFSET_AFTER])
1038                 return -EINVAL;
1039
1040         natseq->offset_after =
1041                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_AFTER]));
1042
1043         return 0;
1044 }
1045
1046 static int
1047 ctnetlink_change_nat_seq_adj(struct nf_conn *ct, struct nlattr *cda[])
1048 {
1049         int ret = 0;
1050         struct nf_conn_nat *nat = nfct_nat(ct);
1051
1052         if (!nat)
1053                 return 0;
1054
1055         if (cda[CTA_NAT_SEQ_ADJ_ORIG]) {
1056                 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_ORIGINAL],
1057                                          cda[CTA_NAT_SEQ_ADJ_ORIG]);
1058                 if (ret < 0)
1059                         return ret;
1060
1061                 ct->status |= IPS_SEQ_ADJUST;
1062         }
1063
1064         if (cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1065                 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_REPLY],
1066                                          cda[CTA_NAT_SEQ_ADJ_REPLY]);
1067                 if (ret < 0)
1068                         return ret;
1069
1070                 ct->status |= IPS_SEQ_ADJUST;
1071         }
1072
1073         return 0;
1074 }
1075 #endif
1076
1077 static int
1078 ctnetlink_change_conntrack(struct nf_conn *ct, struct nlattr *cda[])
1079 {
1080         int err;
1081
1082         if (cda[CTA_HELP]) {
1083                 err = ctnetlink_change_helper(ct, cda);
1084                 if (err < 0)
1085                         return err;
1086         }
1087
1088         if (cda[CTA_TIMEOUT]) {
1089                 err = ctnetlink_change_timeout(ct, cda);
1090                 if (err < 0)
1091                         return err;
1092         }
1093
1094         if (cda[CTA_STATUS]) {
1095                 err = ctnetlink_change_status(ct, cda);
1096                 if (err < 0)
1097                         return err;
1098         }
1099
1100         if (cda[CTA_PROTOINFO]) {
1101                 err = ctnetlink_change_protoinfo(ct, cda);
1102                 if (err < 0)
1103                         return err;
1104         }
1105
1106 #if defined(CONFIG_NF_CONNTRACK_MARK)
1107         if (cda[CTA_MARK])
1108                 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1109 #endif
1110
1111 #ifdef CONFIG_NF_NAT_NEEDED
1112         if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1113                 err = ctnetlink_change_nat_seq_adj(ct, cda);
1114                 if (err < 0)
1115                         return err;
1116         }
1117 #endif
1118
1119         return 0;
1120 }
1121
1122 static int
1123 ctnetlink_create_conntrack(struct nlattr *cda[],
1124                            struct nf_conntrack_tuple *otuple,
1125                            struct nf_conntrack_tuple *rtuple,
1126                            struct nf_conn *master_ct)
1127 {
1128         struct nf_conn *ct;
1129         int err = -EINVAL;
1130         struct nf_conn_help *help;
1131         struct nf_conntrack_helper *helper;
1132
1133         ct = nf_conntrack_alloc(otuple, rtuple);
1134         if (ct == NULL || IS_ERR(ct))
1135                 return -ENOMEM;
1136
1137         if (!cda[CTA_TIMEOUT])
1138                 goto err;
1139         ct->timeout.expires = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
1140
1141         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1142         ct->status |= IPS_CONFIRMED;
1143
1144         if (cda[CTA_STATUS]) {
1145                 err = ctnetlink_change_status(ct, cda);
1146                 if (err < 0)
1147                         goto err;
1148         }
1149
1150         if (cda[CTA_PROTOINFO]) {
1151                 err = ctnetlink_change_protoinfo(ct, cda);
1152                 if (err < 0)
1153                         goto err;
1154         }
1155
1156 #if defined(CONFIG_NF_CONNTRACK_MARK)
1157         if (cda[CTA_MARK])
1158                 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1159 #endif
1160
1161         rcu_read_lock();
1162         helper = __nf_ct_helper_find(rtuple);
1163         if (helper) {
1164                 help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
1165                 if (help == NULL) {
1166                         rcu_read_unlock();
1167                         err = -ENOMEM;
1168                         goto err;
1169                 }
1170                 /* not in hash table yet so not strictly necessary */
1171                 rcu_assign_pointer(help->helper, helper);
1172         }
1173
1174         /* setup master conntrack: this is a confirmed expectation */
1175         if (master_ct) {
1176                 __set_bit(IPS_EXPECTED_BIT, &ct->status);
1177                 ct->master = master_ct;
1178         }
1179
1180         add_timer(&ct->timeout);
1181         nf_conntrack_hash_insert(ct);
1182         rcu_read_unlock();
1183
1184         return 0;
1185
1186 err:
1187         nf_conntrack_free(ct);
1188         return err;
1189 }
1190
1191 static int
1192 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1193                         struct nlmsghdr *nlh, struct nlattr *cda[])
1194 {
1195         struct nf_conntrack_tuple otuple, rtuple;
1196         struct nf_conntrack_tuple_hash *h = NULL;
1197         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1198         u_int8_t u3 = nfmsg->nfgen_family;
1199         int err = 0;
1200
1201         if (cda[CTA_TUPLE_ORIG]) {
1202                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG, u3);
1203                 if (err < 0)
1204                         return err;
1205         }
1206
1207         if (cda[CTA_TUPLE_REPLY]) {
1208                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY, u3);
1209                 if (err < 0)
1210                         return err;
1211         }
1212
1213         spin_lock_bh(&nf_conntrack_lock);
1214         if (cda[CTA_TUPLE_ORIG])
1215                 h = __nf_conntrack_find(&otuple);
1216         else if (cda[CTA_TUPLE_REPLY])
1217                 h = __nf_conntrack_find(&rtuple);
1218
1219         if (h == NULL) {
1220                 struct nf_conntrack_tuple master;
1221                 struct nf_conntrack_tuple_hash *master_h = NULL;
1222                 struct nf_conn *master_ct = NULL;
1223
1224                 if (cda[CTA_TUPLE_MASTER]) {
1225                         err = ctnetlink_parse_tuple(cda,
1226                                                     &master,
1227                                                     CTA_TUPLE_MASTER,
1228                                                     u3);
1229                         if (err < 0)
1230                                 goto out_unlock;
1231
1232                         master_h = __nf_conntrack_find(&master);
1233                         if (master_h == NULL) {
1234                                 err = -ENOENT;
1235                                 goto out_unlock;
1236                         }
1237                         master_ct = nf_ct_tuplehash_to_ctrack(master_h);
1238                         atomic_inc(&master_ct->ct_general.use);
1239                 }
1240
1241                 spin_unlock_bh(&nf_conntrack_lock);
1242                 err = -ENOENT;
1243                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1244                         err = ctnetlink_create_conntrack(cda,
1245                                                          &otuple,
1246                                                          &rtuple,
1247                                                          master_ct);
1248                 if (err < 0 && master_ct)
1249                         nf_ct_put(master_ct);
1250
1251                 return err;
1252         }
1253         /* implicit 'else' */
1254
1255         /* We manipulate the conntrack inside the global conntrack table lock,
1256          * so there's no need to increase the refcount */
1257         err = -EEXIST;
1258         if (!(nlh->nlmsg_flags & NLM_F_EXCL)) {
1259                 /* we only allow nat config for new conntracks */
1260                 if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
1261                         err = -EINVAL;
1262                         goto out_unlock;
1263                 }
1264                 /* can't link an existing conntrack to a master */
1265                 if (cda[CTA_TUPLE_MASTER]) {
1266                         err = -EINVAL;
1267                         goto out_unlock;
1268                 }
1269                 err = ctnetlink_change_conntrack(nf_ct_tuplehash_to_ctrack(h),
1270                                                  cda);
1271         }
1272
1273 out_unlock:
1274         spin_unlock_bh(&nf_conntrack_lock);
1275         return err;
1276 }
1277
1278 /***********************************************************************
1279  * EXPECT
1280  ***********************************************************************/
1281
1282 static inline int
1283 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1284                          const struct nf_conntrack_tuple *tuple,
1285                          enum ctattr_expect type)
1286 {
1287         struct nlattr *nest_parms;
1288
1289         nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
1290         if (!nest_parms)
1291                 goto nla_put_failure;
1292         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1293                 goto nla_put_failure;
1294         nla_nest_end(skb, nest_parms);
1295
1296         return 0;
1297
1298 nla_put_failure:
1299         return -1;
1300 }
1301
1302 static inline int
1303 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1304                         const struct nf_conntrack_tuple *tuple,
1305                         const struct nf_conntrack_tuple_mask *mask)
1306 {
1307         int ret;
1308         struct nf_conntrack_l3proto *l3proto;
1309         struct nf_conntrack_l4proto *l4proto;
1310         struct nf_conntrack_tuple m;
1311         struct nlattr *nest_parms;
1312
1313         memset(&m, 0xFF, sizeof(m));
1314         m.src.u.all = mask->src.u.all;
1315         memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3));
1316
1317         nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK | NLA_F_NESTED);
1318         if (!nest_parms)
1319                 goto nla_put_failure;
1320
1321         l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
1322         ret = ctnetlink_dump_tuples_ip(skb, &m, l3proto);
1323         nf_ct_l3proto_put(l3proto);
1324
1325         if (unlikely(ret < 0))
1326                 goto nla_put_failure;
1327
1328         l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
1329         ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto);
1330         nf_ct_l4proto_put(l4proto);
1331         if (unlikely(ret < 0))
1332                 goto nla_put_failure;
1333
1334         nla_nest_end(skb, nest_parms);
1335
1336         return 0;
1337
1338 nla_put_failure:
1339         return -1;
1340 }
1341
1342 static int
1343 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1344                           const struct nf_conntrack_expect *exp)
1345 {
1346         struct nf_conn *master = exp->master;
1347         long timeout = (exp->timeout.expires - jiffies) / HZ;
1348
1349         if (timeout < 0)
1350                 timeout = 0;
1351
1352         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1353                 goto nla_put_failure;
1354         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1355                 goto nla_put_failure;
1356         if (ctnetlink_exp_dump_tuple(skb,
1357                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1358                                  CTA_EXPECT_MASTER) < 0)
1359                 goto nla_put_failure;
1360
1361         NLA_PUT_BE32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout));
1362         NLA_PUT_BE32(skb, CTA_EXPECT_ID, htonl((unsigned long)exp));
1363
1364         return 0;
1365
1366 nla_put_failure:
1367         return -1;
1368 }
1369
1370 static int
1371 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1372                     int event,
1373                     int nowait,
1374                     const struct nf_conntrack_expect *exp)
1375 {
1376         struct nlmsghdr *nlh;
1377         struct nfgenmsg *nfmsg;
1378         unsigned char *b = skb_tail_pointer(skb);
1379
1380         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1381         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1382         nfmsg  = NLMSG_DATA(nlh);
1383
1384         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1385         nfmsg->nfgen_family = exp->tuple.src.l3num;
1386         nfmsg->version      = NFNETLINK_V0;
1387         nfmsg->res_id       = 0;
1388
1389         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1390                 goto nla_put_failure;
1391
1392         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
1393         return skb->len;
1394
1395 nlmsg_failure:
1396 nla_put_failure:
1397         nlmsg_trim(skb, b);
1398         return -1;
1399 }
1400
1401 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1402 static int ctnetlink_expect_event(struct notifier_block *this,
1403                                   unsigned long events, void *ptr)
1404 {
1405         struct nlmsghdr *nlh;
1406         struct nfgenmsg *nfmsg;
1407         struct nf_conntrack_expect *exp = (struct nf_conntrack_expect *)ptr;
1408         struct sk_buff *skb;
1409         unsigned int type;
1410         sk_buff_data_t b;
1411         int flags = 0;
1412
1413         if (events & IPEXP_NEW) {
1414                 type = IPCTNL_MSG_EXP_NEW;
1415                 flags = NLM_F_CREATE|NLM_F_EXCL;
1416         } else
1417                 return NOTIFY_DONE;
1418
1419         if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1420                 return NOTIFY_DONE;
1421
1422         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1423         if (!skb)
1424                 return NOTIFY_DONE;
1425
1426         b = skb->tail;
1427
1428         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1429         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1430         nfmsg = NLMSG_DATA(nlh);
1431
1432         nlh->nlmsg_flags    = flags;
1433         nfmsg->nfgen_family = exp->tuple.src.l3num;
1434         nfmsg->version      = NFNETLINK_V0;
1435         nfmsg->res_id       = 0;
1436
1437         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1438                 goto nla_put_failure;
1439
1440         nlh->nlmsg_len = skb->tail - b;
1441         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1442         return NOTIFY_DONE;
1443
1444 nlmsg_failure:
1445 nla_put_failure:
1446         kfree_skb(skb);
1447         return NOTIFY_DONE;
1448 }
1449 #endif
1450 static int ctnetlink_exp_done(struct netlink_callback *cb)
1451 {
1452         if (cb->args[1])
1453                 nf_ct_expect_put((struct nf_conntrack_expect *)cb->args[1]);
1454         return 0;
1455 }
1456
1457 static int
1458 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1459 {
1460         struct nf_conntrack_expect *exp, *last;
1461         struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
1462         struct hlist_node *n;
1463         u_int8_t l3proto = nfmsg->nfgen_family;
1464
1465         rcu_read_lock();
1466         last = (struct nf_conntrack_expect *)cb->args[1];
1467         for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) {
1468 restart:
1469                 hlist_for_each_entry(exp, n, &nf_ct_expect_hash[cb->args[0]],
1470                                      hnode) {
1471                         if (l3proto && exp->tuple.src.l3num != l3proto)
1472                                 continue;
1473                         if (cb->args[1]) {
1474                                 if (exp != last)
1475                                         continue;
1476                                 cb->args[1] = 0;
1477                         }
1478                         if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1479                                                     cb->nlh->nlmsg_seq,
1480                                                     IPCTNL_MSG_EXP_NEW,
1481                                                     1, exp) < 0) {
1482                                 if (!atomic_inc_not_zero(&exp->use))
1483                                         continue;
1484                                 cb->args[1] = (unsigned long)exp;
1485                                 goto out;
1486                         }
1487                 }
1488                 if (cb->args[1]) {
1489                         cb->args[1] = 0;
1490                         goto restart;
1491                 }
1492         }
1493 out:
1494         rcu_read_unlock();
1495         if (last)
1496                 nf_ct_expect_put(last);
1497
1498         return skb->len;
1499 }
1500
1501 static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = {
1502         [CTA_EXPECT_TIMEOUT]    = { .type = NLA_U32 },
1503         [CTA_EXPECT_ID]         = { .type = NLA_U32 },
1504 };
1505
1506 static int
1507 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1508                      struct nlmsghdr *nlh, struct nlattr *cda[])
1509 {
1510         struct nf_conntrack_tuple tuple;
1511         struct nf_conntrack_expect *exp;
1512         struct sk_buff *skb2;
1513         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1514         u_int8_t u3 = nfmsg->nfgen_family;
1515         int err = 0;
1516
1517         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1518                 return netlink_dump_start(ctnl, skb, nlh,
1519                                           ctnetlink_exp_dump_table,
1520                                           ctnetlink_exp_done);
1521         }
1522
1523         if (cda[CTA_EXPECT_MASTER])
1524                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER, u3);
1525         else
1526                 return -EINVAL;
1527
1528         if (err < 0)
1529                 return err;
1530
1531         exp = nf_ct_expect_find_get(&tuple);
1532         if (!exp)
1533                 return -ENOENT;
1534
1535         if (cda[CTA_EXPECT_ID]) {
1536                 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1537                 if (ntohl(id) != (u32)(unsigned long)exp) {
1538                         nf_ct_expect_put(exp);
1539                         return -ENOENT;
1540                 }
1541         }
1542
1543         err = -ENOMEM;
1544         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1545         if (!skb2)
1546                 goto out;
1547
1548         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1549                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1550                                       1, exp);
1551         if (err <= 0)
1552                 goto free;
1553
1554         nf_ct_expect_put(exp);
1555
1556         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1557
1558 free:
1559         kfree_skb(skb2);
1560 out:
1561         nf_ct_expect_put(exp);
1562         return err;
1563 }
1564
1565 static int
1566 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1567                      struct nlmsghdr *nlh, struct nlattr *cda[])
1568 {
1569         struct nf_conntrack_expect *exp;
1570         struct nf_conntrack_tuple tuple;
1571         struct nf_conntrack_helper *h;
1572         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1573         struct hlist_node *n, *next;
1574         u_int8_t u3 = nfmsg->nfgen_family;
1575         unsigned int i;
1576         int err;
1577
1578         if (cda[CTA_EXPECT_TUPLE]) {
1579                 /* delete a single expect by tuple */
1580                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1581                 if (err < 0)
1582                         return err;
1583
1584                 /* bump usage count to 2 */
1585                 exp = nf_ct_expect_find_get(&tuple);
1586                 if (!exp)
1587                         return -ENOENT;
1588
1589                 if (cda[CTA_EXPECT_ID]) {
1590                         __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1591                         if (ntohl(id) != (u32)(unsigned long)exp) {
1592                                 nf_ct_expect_put(exp);
1593                                 return -ENOENT;
1594                         }
1595                 }
1596
1597                 /* after list removal, usage count == 1 */
1598                 nf_ct_unexpect_related(exp);
1599                 /* have to put what we 'get' above.
1600                  * after this line usage count == 0 */
1601                 nf_ct_expect_put(exp);
1602         } else if (cda[CTA_EXPECT_HELP_NAME]) {
1603                 char *name = nla_data(cda[CTA_EXPECT_HELP_NAME]);
1604                 struct nf_conn_help *m_help;
1605
1606                 /* delete all expectations for this helper */
1607                 spin_lock_bh(&nf_conntrack_lock);
1608                 h = __nf_conntrack_helper_find_byname(name);
1609                 if (!h) {
1610                         spin_unlock_bh(&nf_conntrack_lock);
1611                         return -EINVAL;
1612                 }
1613                 for (i = 0; i < nf_ct_expect_hsize; i++) {
1614                         hlist_for_each_entry_safe(exp, n, next,
1615                                                   &nf_ct_expect_hash[i],
1616                                                   hnode) {
1617                                 m_help = nfct_help(exp->master);
1618                                 if (m_help->helper == h
1619                                     && del_timer(&exp->timeout)) {
1620                                         nf_ct_unlink_expect(exp);
1621                                         nf_ct_expect_put(exp);
1622                                 }
1623                         }
1624                 }
1625                 spin_unlock_bh(&nf_conntrack_lock);
1626         } else {
1627                 /* This basically means we have to flush everything*/
1628                 spin_lock_bh(&nf_conntrack_lock);
1629                 for (i = 0; i < nf_ct_expect_hsize; i++) {
1630                         hlist_for_each_entry_safe(exp, n, next,
1631                                                   &nf_ct_expect_hash[i],
1632                                                   hnode) {
1633                                 if (del_timer(&exp->timeout)) {
1634                                         nf_ct_unlink_expect(exp);
1635                                         nf_ct_expect_put(exp);
1636                                 }
1637                         }
1638                 }
1639                 spin_unlock_bh(&nf_conntrack_lock);
1640         }
1641
1642         return 0;
1643 }
1644 static int
1645 ctnetlink_change_expect(struct nf_conntrack_expect *x, struct nlattr *cda[])
1646 {
1647         return -EOPNOTSUPP;
1648 }
1649
1650 static int
1651 ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3)
1652 {
1653         struct nf_conntrack_tuple tuple, mask, master_tuple;
1654         struct nf_conntrack_tuple_hash *h = NULL;
1655         struct nf_conntrack_expect *exp;
1656         struct nf_conn *ct;
1657         struct nf_conn_help *help;
1658         int err = 0;
1659
1660         /* caller guarantees that those three CTA_EXPECT_* exist */
1661         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1662         if (err < 0)
1663                 return err;
1664         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK, u3);
1665         if (err < 0)
1666                 return err;
1667         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER, u3);
1668         if (err < 0)
1669                 return err;
1670
1671         /* Look for master conntrack of this expectation */
1672         h = nf_conntrack_find_get(&master_tuple);
1673         if (!h)
1674                 return -ENOENT;
1675         ct = nf_ct_tuplehash_to_ctrack(h);
1676         help = nfct_help(ct);
1677
1678         if (!help || !help->helper) {
1679                 /* such conntrack hasn't got any helper, abort */
1680                 err = -EINVAL;
1681                 goto out;
1682         }
1683
1684         exp = nf_ct_expect_alloc(ct);
1685         if (!exp) {
1686                 err = -ENOMEM;
1687                 goto out;
1688         }
1689
1690         exp->expectfn = NULL;
1691         exp->flags = 0;
1692         exp->master = ct;
1693         exp->helper = NULL;
1694         memcpy(&exp->tuple, &tuple, sizeof(struct nf_conntrack_tuple));
1695         memcpy(&exp->mask.src.u3, &mask.src.u3, sizeof(exp->mask.src.u3));
1696         exp->mask.src.u.all = mask.src.u.all;
1697
1698         err = nf_ct_expect_related(exp);
1699         nf_ct_expect_put(exp);
1700
1701 out:
1702         nf_ct_put(nf_ct_tuplehash_to_ctrack(h));
1703         return err;
1704 }
1705
1706 static int
1707 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1708                      struct nlmsghdr *nlh, struct nlattr *cda[])
1709 {
1710         struct nf_conntrack_tuple tuple;
1711         struct nf_conntrack_expect *exp;
1712         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1713         u_int8_t u3 = nfmsg->nfgen_family;
1714         int err = 0;
1715
1716         if (!cda[CTA_EXPECT_TUPLE]
1717             || !cda[CTA_EXPECT_MASK]
1718             || !cda[CTA_EXPECT_MASTER])
1719                 return -EINVAL;
1720
1721         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1722         if (err < 0)
1723                 return err;
1724
1725         spin_lock_bh(&nf_conntrack_lock);
1726         exp = __nf_ct_expect_find(&tuple);
1727
1728         if (!exp) {
1729                 spin_unlock_bh(&nf_conntrack_lock);
1730                 err = -ENOENT;
1731                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1732                         err = ctnetlink_create_expect(cda, u3);
1733                 return err;
1734         }
1735
1736         err = -EEXIST;
1737         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1738                 err = ctnetlink_change_expect(exp, cda);
1739         spin_unlock_bh(&nf_conntrack_lock);
1740
1741         return err;
1742 }
1743
1744 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1745 static struct notifier_block ctnl_notifier = {
1746         .notifier_call  = ctnetlink_conntrack_event,
1747 };
1748
1749 static struct notifier_block ctnl_notifier_exp = {
1750         .notifier_call  = ctnetlink_expect_event,
1751 };
1752 #endif
1753
1754 static const struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1755         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1756                                             .attr_count = CTA_MAX,
1757                                             .policy = ct_nla_policy },
1758         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1759                                             .attr_count = CTA_MAX,
1760                                             .policy = ct_nla_policy },
1761         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1762                                             .attr_count = CTA_MAX,
1763                                             .policy = ct_nla_policy },
1764         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1765                                             .attr_count = CTA_MAX,
1766                                             .policy = ct_nla_policy },
1767 };
1768
1769 static const struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1770         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1771                                             .attr_count = CTA_EXPECT_MAX,
1772                                             .policy = exp_nla_policy },
1773         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1774                                             .attr_count = CTA_EXPECT_MAX,
1775                                             .policy = exp_nla_policy },
1776         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1777                                             .attr_count = CTA_EXPECT_MAX,
1778                                             .policy = exp_nla_policy },
1779 };
1780
1781 static const struct nfnetlink_subsystem ctnl_subsys = {
1782         .name                           = "conntrack",
1783         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1784         .cb_count                       = IPCTNL_MSG_MAX,
1785         .cb                             = ctnl_cb,
1786 };
1787
1788 static const struct nfnetlink_subsystem ctnl_exp_subsys = {
1789         .name                           = "conntrack_expect",
1790         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1791         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1792         .cb                             = ctnl_exp_cb,
1793 };
1794
1795 MODULE_ALIAS("ip_conntrack_netlink");
1796 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1797 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1798
1799 static int __init ctnetlink_init(void)
1800 {
1801         int ret;
1802
1803         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1804         ret = nfnetlink_subsys_register(&ctnl_subsys);
1805         if (ret < 0) {
1806                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1807                 goto err_out;
1808         }
1809
1810         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1811         if (ret < 0) {
1812                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1813                 goto err_unreg_subsys;
1814         }
1815
1816 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1817         ret = nf_conntrack_register_notifier(&ctnl_notifier);
1818         if (ret < 0) {
1819                 printk("ctnetlink_init: cannot register notifier.\n");
1820                 goto err_unreg_exp_subsys;
1821         }
1822
1823         ret = nf_ct_expect_register_notifier(&ctnl_notifier_exp);
1824         if (ret < 0) {
1825                 printk("ctnetlink_init: cannot expect register notifier.\n");
1826                 goto err_unreg_notifier;
1827         }
1828 #endif
1829
1830         return 0;
1831
1832 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1833 err_unreg_notifier:
1834         nf_conntrack_unregister_notifier(&ctnl_notifier);
1835 err_unreg_exp_subsys:
1836         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1837 #endif
1838 err_unreg_subsys:
1839         nfnetlink_subsys_unregister(&ctnl_subsys);
1840 err_out:
1841         return ret;
1842 }
1843
1844 static void __exit ctnetlink_exit(void)
1845 {
1846         printk("ctnetlink: unregistering from nfnetlink.\n");
1847
1848 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1849         nf_ct_expect_unregister_notifier(&ctnl_notifier_exp);
1850         nf_conntrack_unregister_notifier(&ctnl_notifier);
1851 #endif
1852
1853         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1854         nfnetlink_subsys_unregister(&ctnl_subsys);
1855         return;
1856 }
1857
1858 module_init(ctnetlink_init);
1859 module_exit(ctnetlink_exit);