[NETFILTER]: nf_conntrack: add tuplehash l3num/protonum accessors
[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 (events & IPCT_DESTROY) {
476                 if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
477                     ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
478                         goto nla_put_failure;
479         } else {
480                 if (ctnetlink_dump_status(skb, ct) < 0)
481                         goto nla_put_failure;
482
483                 if (ctnetlink_dump_timeout(skb, ct) < 0)
484                         goto nla_put_failure;
485
486                 if (events & IPCT_PROTOINFO
487                     && ctnetlink_dump_protoinfo(skb, ct) < 0)
488                         goto nla_put_failure;
489
490                 if ((events & IPCT_HELPER || nfct_help(ct))
491                     && ctnetlink_dump_helpinfo(skb, ct) < 0)
492                         goto nla_put_failure;
493
494 #ifdef CONFIG_NF_CONNTRACK_SECMARK
495                 if ((events & IPCT_SECMARK || ct->secmark)
496                     && ctnetlink_dump_secmark(skb, ct) < 0)
497                         goto nla_put_failure;
498 #endif
499
500                 if (events & IPCT_COUNTER_FILLING &&
501                     (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
502                      ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0))
503                         goto nla_put_failure;
504
505                 if (events & IPCT_RELATED &&
506                     ctnetlink_dump_master(skb, ct) < 0)
507                         goto nla_put_failure;
508
509                 if (events & IPCT_NATSEQADJ &&
510                     ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
511                         goto nla_put_failure;
512         }
513
514 #ifdef CONFIG_NF_CONNTRACK_MARK
515         if ((events & IPCT_MARK || ct->mark)
516             && ctnetlink_dump_mark(skb, ct) < 0)
517                 goto nla_put_failure;
518 #endif
519
520         nlh->nlmsg_len = skb->tail - b;
521         nfnetlink_send(skb, 0, group, 0);
522         return NOTIFY_DONE;
523
524 nlmsg_failure:
525 nla_put_failure:
526         kfree_skb(skb);
527         return NOTIFY_DONE;
528 }
529 #endif /* CONFIG_NF_CONNTRACK_EVENTS */
530
531 static int ctnetlink_done(struct netlink_callback *cb)
532 {
533         if (cb->args[1])
534                 nf_ct_put((struct nf_conn *)cb->args[1]);
535         return 0;
536 }
537
538 static int
539 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
540 {
541         struct nf_conn *ct, *last;
542         struct nf_conntrack_tuple_hash *h;
543         struct hlist_node *n;
544         struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
545         u_int8_t l3proto = nfmsg->nfgen_family;
546
547         rcu_read_lock();
548         last = (struct nf_conn *)cb->args[1];
549         for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
550 restart:
551                 hlist_for_each_entry_rcu(h, n, &nf_conntrack_hash[cb->args[0]],
552                                          hnode) {
553                         if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
554                                 continue;
555                         ct = nf_ct_tuplehash_to_ctrack(h);
556                         /* Dump entries of a given L3 protocol number.
557                          * If it is not specified, ie. l3proto == 0,
558                          * then dump everything. */
559                         if (l3proto && nf_ct_l3num(ct) != l3proto)
560                                 continue;
561                         if (cb->args[1]) {
562                                 if (ct != last)
563                                         continue;
564                                 cb->args[1] = 0;
565                         }
566                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
567                                                 cb->nlh->nlmsg_seq,
568                                                 IPCTNL_MSG_CT_NEW,
569                                                 1, ct) < 0) {
570                                 if (!atomic_inc_not_zero(&ct->ct_general.use))
571                                         continue;
572                                 cb->args[1] = (unsigned long)ct;
573                                 goto out;
574                         }
575 #ifdef CONFIG_NF_CT_ACCT
576                         if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
577                                                 IPCTNL_MSG_CT_GET_CTRZERO)
578                                 memset(&ct->counters, 0, sizeof(ct->counters));
579 #endif
580                 }
581                 if (cb->args[1]) {
582                         cb->args[1] = 0;
583                         goto restart;
584                 }
585         }
586 out:
587         rcu_read_unlock();
588         if (last)
589                 nf_ct_put(last);
590
591         return skb->len;
592 }
593
594 static inline int
595 ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
596 {
597         struct nlattr *tb[CTA_IP_MAX+1];
598         struct nf_conntrack_l3proto *l3proto;
599         int ret = 0;
600
601         nla_parse_nested(tb, CTA_IP_MAX, attr, NULL);
602
603         l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
604
605         if (likely(l3proto->nlattr_to_tuple)) {
606                 ret = nla_validate_nested(attr, CTA_IP_MAX,
607                                           l3proto->nla_policy);
608                 if (ret == 0)
609                         ret = l3proto->nlattr_to_tuple(tb, tuple);
610         }
611
612         nf_ct_l3proto_put(l3proto);
613
614         return ret;
615 }
616
617 static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = {
618         [CTA_PROTO_NUM] = { .type = NLA_U8 },
619 };
620
621 static inline int
622 ctnetlink_parse_tuple_proto(struct nlattr *attr,
623                             struct nf_conntrack_tuple *tuple)
624 {
625         struct nlattr *tb[CTA_PROTO_MAX+1];
626         struct nf_conntrack_l4proto *l4proto;
627         int ret = 0;
628
629         ret = nla_parse_nested(tb, CTA_PROTO_MAX, attr, proto_nla_policy);
630         if (ret < 0)
631                 return ret;
632
633         if (!tb[CTA_PROTO_NUM])
634                 return -EINVAL;
635         tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
636
637         l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
638
639         if (likely(l4proto->nlattr_to_tuple)) {
640                 ret = nla_validate_nested(attr, CTA_PROTO_MAX,
641                                           l4proto->nla_policy);
642                 if (ret == 0)
643                         ret = l4proto->nlattr_to_tuple(tb, tuple);
644         }
645
646         nf_ct_l4proto_put(l4proto);
647
648         return ret;
649 }
650
651 static int
652 ctnetlink_parse_tuple(struct nlattr *cda[], struct nf_conntrack_tuple *tuple,
653                       enum ctattr_tuple type, u_int8_t l3num)
654 {
655         struct nlattr *tb[CTA_TUPLE_MAX+1];
656         int err;
657
658         memset(tuple, 0, sizeof(*tuple));
659
660         nla_parse_nested(tb, CTA_TUPLE_MAX, cda[type], NULL);
661
662         if (!tb[CTA_TUPLE_IP])
663                 return -EINVAL;
664
665         tuple->src.l3num = l3num;
666
667         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple);
668         if (err < 0)
669                 return err;
670
671         if (!tb[CTA_TUPLE_PROTO])
672                 return -EINVAL;
673
674         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple);
675         if (err < 0)
676                 return err;
677
678         /* orig and expect tuples get DIR_ORIGINAL */
679         if (type == CTA_TUPLE_REPLY)
680                 tuple->dst.dir = IP_CT_DIR_REPLY;
681         else
682                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
683
684         return 0;
685 }
686
687 #ifdef CONFIG_NF_NAT_NEEDED
688 static const struct nla_policy protonat_nla_policy[CTA_PROTONAT_MAX+1] = {
689         [CTA_PROTONAT_PORT_MIN] = { .type = NLA_U16 },
690         [CTA_PROTONAT_PORT_MAX] = { .type = NLA_U16 },
691 };
692
693 static int nfnetlink_parse_nat_proto(struct nlattr *attr,
694                                      const struct nf_conn *ct,
695                                      struct nf_nat_range *range)
696 {
697         struct nlattr *tb[CTA_PROTONAT_MAX+1];
698         const struct nf_nat_protocol *npt;
699         int err;
700
701         err = nla_parse_nested(tb, CTA_PROTONAT_MAX, attr, protonat_nla_policy);
702         if (err < 0)
703                 return err;
704
705         npt = nf_nat_proto_find_get(nf_ct_protonum(ct));
706         if (npt->nlattr_to_range)
707                 err = npt->nlattr_to_range(tb, range);
708         nf_nat_proto_put(npt);
709         return err;
710 }
711
712 static const struct nla_policy nat_nla_policy[CTA_NAT_MAX+1] = {
713         [CTA_NAT_MINIP]         = { .type = NLA_U32 },
714         [CTA_NAT_MAXIP]         = { .type = NLA_U32 },
715 };
716
717 static inline int
718 nfnetlink_parse_nat(struct nlattr *nat,
719                     const struct nf_conn *ct, struct nf_nat_range *range)
720 {
721         struct nlattr *tb[CTA_NAT_MAX+1];
722         int err;
723
724         memset(range, 0, sizeof(*range));
725
726         err = nla_parse_nested(tb, CTA_NAT_MAX, nat, nat_nla_policy);
727         if (err < 0)
728                 return err;
729
730         if (tb[CTA_NAT_MINIP])
731                 range->min_ip = nla_get_be32(tb[CTA_NAT_MINIP]);
732
733         if (!tb[CTA_NAT_MAXIP])
734                 range->max_ip = range->min_ip;
735         else
736                 range->max_ip = nla_get_be32(tb[CTA_NAT_MAXIP]);
737
738         if (range->min_ip)
739                 range->flags |= IP_NAT_RANGE_MAP_IPS;
740
741         if (!tb[CTA_NAT_PROTO])
742                 return 0;
743
744         err = nfnetlink_parse_nat_proto(tb[CTA_NAT_PROTO], ct, range);
745         if (err < 0)
746                 return err;
747
748         return 0;
749 }
750 #endif
751
752 static inline int
753 ctnetlink_parse_help(struct nlattr *attr, char **helper_name)
754 {
755         struct nlattr *tb[CTA_HELP_MAX+1];
756
757         nla_parse_nested(tb, CTA_HELP_MAX, attr, NULL);
758
759         if (!tb[CTA_HELP_NAME])
760                 return -EINVAL;
761
762         *helper_name = nla_data(tb[CTA_HELP_NAME]);
763
764         return 0;
765 }
766
767 static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
768         [CTA_STATUS]            = { .type = NLA_U32 },
769         [CTA_TIMEOUT]           = { .type = NLA_U32 },
770         [CTA_MARK]              = { .type = NLA_U32 },
771         [CTA_USE]               = { .type = NLA_U32 },
772         [CTA_ID]                = { .type = NLA_U32 },
773 };
774
775 static int
776 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
777                         struct nlmsghdr *nlh, struct nlattr *cda[])
778 {
779         struct nf_conntrack_tuple_hash *h;
780         struct nf_conntrack_tuple tuple;
781         struct nf_conn *ct;
782         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
783         u_int8_t u3 = nfmsg->nfgen_family;
784         int err = 0;
785
786         if (cda[CTA_TUPLE_ORIG])
787                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
788         else if (cda[CTA_TUPLE_REPLY])
789                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
790         else {
791                 /* Flush the whole table */
792                 nf_conntrack_flush();
793                 return 0;
794         }
795
796         if (err < 0)
797                 return err;
798
799         h = nf_conntrack_find_get(&tuple);
800         if (!h)
801                 return -ENOENT;
802
803         ct = nf_ct_tuplehash_to_ctrack(h);
804
805         if (cda[CTA_ID]) {
806                 u_int32_t id = ntohl(nla_get_be32(cda[CTA_ID]));
807                 if (id != (u32)(unsigned long)ct) {
808                         nf_ct_put(ct);
809                         return -ENOENT;
810                 }
811         }
812         if (del_timer(&ct->timeout))
813                 ct->timeout.function((unsigned long)ct);
814
815         nf_ct_put(ct);
816
817         return 0;
818 }
819
820 static int
821 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
822                         struct nlmsghdr *nlh, struct nlattr *cda[])
823 {
824         struct nf_conntrack_tuple_hash *h;
825         struct nf_conntrack_tuple tuple;
826         struct nf_conn *ct;
827         struct sk_buff *skb2 = NULL;
828         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
829         u_int8_t u3 = nfmsg->nfgen_family;
830         int err = 0;
831
832         if (nlh->nlmsg_flags & NLM_F_DUMP) {
833 #ifndef CONFIG_NF_CT_ACCT
834                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) == IPCTNL_MSG_CT_GET_CTRZERO)
835                         return -ENOTSUPP;
836 #endif
837                 return netlink_dump_start(ctnl, skb, nlh, ctnetlink_dump_table,
838                                           ctnetlink_done);
839         }
840
841         if (cda[CTA_TUPLE_ORIG])
842                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
843         else if (cda[CTA_TUPLE_REPLY])
844                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
845         else
846                 return -EINVAL;
847
848         if (err < 0)
849                 return err;
850
851         h = nf_conntrack_find_get(&tuple);
852         if (!h)
853                 return -ENOENT;
854
855         ct = nf_ct_tuplehash_to_ctrack(h);
856
857         err = -ENOMEM;
858         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
859         if (!skb2) {
860                 nf_ct_put(ct);
861                 return -ENOMEM;
862         }
863
864         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
865                                   IPCTNL_MSG_CT_NEW, 1, ct);
866         nf_ct_put(ct);
867         if (err <= 0)
868                 goto free;
869
870         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
871         if (err < 0)
872                 goto out;
873
874         return 0;
875
876 free:
877         kfree_skb(skb2);
878 out:
879         return err;
880 }
881
882 static int
883 ctnetlink_change_status(struct nf_conn *ct, struct nlattr *cda[])
884 {
885         unsigned long d;
886         unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
887         d = ct->status ^ status;
888
889         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
890                 /* unchangeable */
891                 return -EINVAL;
892
893         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
894                 /* SEEN_REPLY bit can only be set */
895                 return -EINVAL;
896
897
898         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
899                 /* ASSURED bit can only be set */
900                 return -EINVAL;
901
902         if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
903 #ifndef CONFIG_NF_NAT_NEEDED
904                 return -EINVAL;
905 #else
906                 struct nf_nat_range range;
907
908                 if (cda[CTA_NAT_DST]) {
909                         if (nfnetlink_parse_nat(cda[CTA_NAT_DST], ct,
910                                                 &range) < 0)
911                                 return -EINVAL;
912                         if (nf_nat_initialized(ct, IP_NAT_MANIP_DST))
913                                 return -EEXIST;
914                         nf_nat_setup_info(ct, &range, IP_NAT_MANIP_DST);
915                 }
916                 if (cda[CTA_NAT_SRC]) {
917                         if (nfnetlink_parse_nat(cda[CTA_NAT_SRC], ct,
918                                                 &range) < 0)
919                                 return -EINVAL;
920                         if (nf_nat_initialized(ct, IP_NAT_MANIP_SRC))
921                                 return -EEXIST;
922                         nf_nat_setup_info(ct, &range, IP_NAT_MANIP_SRC);
923                 }
924 #endif
925         }
926
927         /* Be careful here, modifying NAT bits can screw up things,
928          * so don't let users modify them directly if they don't pass
929          * nf_nat_range. */
930         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
931         return 0;
932 }
933
934
935 static inline int
936 ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
937 {
938         struct nf_conntrack_helper *helper;
939         struct nf_conn_help *help = nfct_help(ct);
940         char *helpname;
941         int err;
942
943         /* don't change helper of sibling connections */
944         if (ct->master)
945                 return -EINVAL;
946
947         err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
948         if (err < 0)
949                 return err;
950
951         if (!strcmp(helpname, "")) {
952                 if (help && help->helper) {
953                         /* we had a helper before ... */
954                         nf_ct_remove_expectations(ct);
955                         rcu_assign_pointer(help->helper, NULL);
956                 }
957
958                 return 0;
959         }
960
961         helper = __nf_conntrack_helper_find_byname(helpname);
962         if (helper == NULL)
963                 return -EINVAL;
964
965         if (help) {
966                 if (help->helper == helper)
967                         return 0;
968                 if (help->helper)
969                         return -EBUSY;
970                 /* need to zero data of old helper */
971                 memset(&help->help, 0, sizeof(help->help));
972         } else {
973                 help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
974                 if (help == NULL)
975                         return -ENOMEM;
976         }
977
978         rcu_assign_pointer(help->helper, helper);
979
980         return 0;
981 }
982
983 static inline int
984 ctnetlink_change_timeout(struct nf_conn *ct, struct nlattr *cda[])
985 {
986         u_int32_t timeout = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
987
988         if (!del_timer(&ct->timeout))
989                 return -ETIME;
990
991         ct->timeout.expires = jiffies + timeout * HZ;
992         add_timer(&ct->timeout);
993
994         return 0;
995 }
996
997 static inline int
998 ctnetlink_change_protoinfo(struct nf_conn *ct, struct nlattr *cda[])
999 {
1000         struct nlattr *tb[CTA_PROTOINFO_MAX+1], *attr = cda[CTA_PROTOINFO];
1001         struct nf_conntrack_l4proto *l4proto;
1002         int err = 0;
1003
1004         nla_parse_nested(tb, CTA_PROTOINFO_MAX, attr, NULL);
1005
1006         l4proto = nf_ct_l4proto_find_get(nf_ct_l3num(ct), nf_ct_protonum(ct));
1007         if (l4proto->from_nlattr)
1008                 err = l4proto->from_nlattr(tb, ct);
1009         nf_ct_l4proto_put(l4proto);
1010
1011         return err;
1012 }
1013
1014 #ifdef CONFIG_NF_NAT_NEEDED
1015 static inline int
1016 change_nat_seq_adj(struct nf_nat_seq *natseq, struct nlattr *attr)
1017 {
1018         struct nlattr *cda[CTA_NAT_SEQ_MAX+1];
1019
1020         nla_parse_nested(cda, CTA_NAT_SEQ_MAX, attr, NULL);
1021
1022         if (!cda[CTA_NAT_SEQ_CORRECTION_POS])
1023                 return -EINVAL;
1024
1025         natseq->correction_pos =
1026                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_CORRECTION_POS]));
1027
1028         if (!cda[CTA_NAT_SEQ_OFFSET_BEFORE])
1029                 return -EINVAL;
1030
1031         natseq->offset_before =
1032                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_BEFORE]));
1033
1034         if (!cda[CTA_NAT_SEQ_OFFSET_AFTER])
1035                 return -EINVAL;
1036
1037         natseq->offset_after =
1038                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_AFTER]));
1039
1040         return 0;
1041 }
1042
1043 static int
1044 ctnetlink_change_nat_seq_adj(struct nf_conn *ct, struct nlattr *cda[])
1045 {
1046         int ret = 0;
1047         struct nf_conn_nat *nat = nfct_nat(ct);
1048
1049         if (!nat)
1050                 return 0;
1051
1052         if (cda[CTA_NAT_SEQ_ADJ_ORIG]) {
1053                 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_ORIGINAL],
1054                                          cda[CTA_NAT_SEQ_ADJ_ORIG]);
1055                 if (ret < 0)
1056                         return ret;
1057
1058                 ct->status |= IPS_SEQ_ADJUST;
1059         }
1060
1061         if (cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1062                 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_REPLY],
1063                                          cda[CTA_NAT_SEQ_ADJ_REPLY]);
1064                 if (ret < 0)
1065                         return ret;
1066
1067                 ct->status |= IPS_SEQ_ADJUST;
1068         }
1069
1070         return 0;
1071 }
1072 #endif
1073
1074 static int
1075 ctnetlink_change_conntrack(struct nf_conn *ct, struct nlattr *cda[])
1076 {
1077         int err;
1078
1079         if (cda[CTA_HELP]) {
1080                 err = ctnetlink_change_helper(ct, cda);
1081                 if (err < 0)
1082                         return err;
1083         }
1084
1085         if (cda[CTA_TIMEOUT]) {
1086                 err = ctnetlink_change_timeout(ct, cda);
1087                 if (err < 0)
1088                         return err;
1089         }
1090
1091         if (cda[CTA_STATUS]) {
1092                 err = ctnetlink_change_status(ct, cda);
1093                 if (err < 0)
1094                         return err;
1095         }
1096
1097         if (cda[CTA_PROTOINFO]) {
1098                 err = ctnetlink_change_protoinfo(ct, cda);
1099                 if (err < 0)
1100                         return err;
1101         }
1102
1103 #if defined(CONFIG_NF_CONNTRACK_MARK)
1104         if (cda[CTA_MARK])
1105                 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1106 #endif
1107
1108 #ifdef CONFIG_NF_NAT_NEEDED
1109         if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1110                 err = ctnetlink_change_nat_seq_adj(ct, cda);
1111                 if (err < 0)
1112                         return err;
1113         }
1114 #endif
1115
1116         return 0;
1117 }
1118
1119 static int
1120 ctnetlink_create_conntrack(struct nlattr *cda[],
1121                            struct nf_conntrack_tuple *otuple,
1122                            struct nf_conntrack_tuple *rtuple,
1123                            struct nf_conn *master_ct)
1124 {
1125         struct nf_conn *ct;
1126         int err = -EINVAL;
1127         struct nf_conn_help *help;
1128         struct nf_conntrack_helper *helper;
1129
1130         ct = nf_conntrack_alloc(otuple, rtuple);
1131         if (ct == NULL || IS_ERR(ct))
1132                 return -ENOMEM;
1133
1134         if (!cda[CTA_TIMEOUT])
1135                 goto err;
1136         ct->timeout.expires = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
1137
1138         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1139         ct->status |= IPS_CONFIRMED;
1140
1141         if (cda[CTA_STATUS]) {
1142                 err = ctnetlink_change_status(ct, cda);
1143                 if (err < 0)
1144                         goto err;
1145         }
1146
1147         if (cda[CTA_PROTOINFO]) {
1148                 err = ctnetlink_change_protoinfo(ct, cda);
1149                 if (err < 0)
1150                         goto err;
1151         }
1152
1153 #if defined(CONFIG_NF_CONNTRACK_MARK)
1154         if (cda[CTA_MARK])
1155                 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1156 #endif
1157
1158         rcu_read_lock();
1159         helper = __nf_ct_helper_find(rtuple);
1160         if (helper) {
1161                 help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
1162                 if (help == NULL) {
1163                         rcu_read_unlock();
1164                         err = -ENOMEM;
1165                         goto err;
1166                 }
1167                 /* not in hash table yet so not strictly necessary */
1168                 rcu_assign_pointer(help->helper, helper);
1169         }
1170
1171         /* setup master conntrack: this is a confirmed expectation */
1172         if (master_ct) {
1173                 __set_bit(IPS_EXPECTED_BIT, &ct->status);
1174                 ct->master = master_ct;
1175         }
1176
1177         add_timer(&ct->timeout);
1178         nf_conntrack_hash_insert(ct);
1179         rcu_read_unlock();
1180
1181         return 0;
1182
1183 err:
1184         nf_conntrack_free(ct);
1185         return err;
1186 }
1187
1188 static int
1189 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1190                         struct nlmsghdr *nlh, struct nlattr *cda[])
1191 {
1192         struct nf_conntrack_tuple otuple, rtuple;
1193         struct nf_conntrack_tuple_hash *h = NULL;
1194         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1195         u_int8_t u3 = nfmsg->nfgen_family;
1196         int err = 0;
1197
1198         if (cda[CTA_TUPLE_ORIG]) {
1199                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG, u3);
1200                 if (err < 0)
1201                         return err;
1202         }
1203
1204         if (cda[CTA_TUPLE_REPLY]) {
1205                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY, u3);
1206                 if (err < 0)
1207                         return err;
1208         }
1209
1210         spin_lock_bh(&nf_conntrack_lock);
1211         if (cda[CTA_TUPLE_ORIG])
1212                 h = __nf_conntrack_find(&otuple);
1213         else if (cda[CTA_TUPLE_REPLY])
1214                 h = __nf_conntrack_find(&rtuple);
1215
1216         if (h == NULL) {
1217                 struct nf_conntrack_tuple master;
1218                 struct nf_conntrack_tuple_hash *master_h = NULL;
1219                 struct nf_conn *master_ct = NULL;
1220
1221                 if (cda[CTA_TUPLE_MASTER]) {
1222                         err = ctnetlink_parse_tuple(cda,
1223                                                     &master,
1224                                                     CTA_TUPLE_MASTER,
1225                                                     u3);
1226                         if (err < 0)
1227                                 goto out_unlock;
1228
1229                         master_h = __nf_conntrack_find(&master);
1230                         if (master_h == NULL) {
1231                                 err = -ENOENT;
1232                                 goto out_unlock;
1233                         }
1234                         master_ct = nf_ct_tuplehash_to_ctrack(master_h);
1235                         atomic_inc(&master_ct->ct_general.use);
1236                 }
1237
1238                 spin_unlock_bh(&nf_conntrack_lock);
1239                 err = -ENOENT;
1240                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1241                         err = ctnetlink_create_conntrack(cda,
1242                                                          &otuple,
1243                                                          &rtuple,
1244                                                          master_ct);
1245                 if (err < 0 && master_ct)
1246                         nf_ct_put(master_ct);
1247
1248                 return err;
1249         }
1250         /* implicit 'else' */
1251
1252         /* We manipulate the conntrack inside the global conntrack table lock,
1253          * so there's no need to increase the refcount */
1254         err = -EEXIST;
1255         if (!(nlh->nlmsg_flags & NLM_F_EXCL)) {
1256                 /* we only allow nat config for new conntracks */
1257                 if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
1258                         err = -EINVAL;
1259                         goto out_unlock;
1260                 }
1261                 /* can't link an existing conntrack to a master */
1262                 if (cda[CTA_TUPLE_MASTER]) {
1263                         err = -EINVAL;
1264                         goto out_unlock;
1265                 }
1266                 err = ctnetlink_change_conntrack(nf_ct_tuplehash_to_ctrack(h),
1267                                                  cda);
1268         }
1269
1270 out_unlock:
1271         spin_unlock_bh(&nf_conntrack_lock);
1272         return err;
1273 }
1274
1275 /***********************************************************************
1276  * EXPECT
1277  ***********************************************************************/
1278
1279 static inline int
1280 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1281                          const struct nf_conntrack_tuple *tuple,
1282                          enum ctattr_expect type)
1283 {
1284         struct nlattr *nest_parms;
1285
1286         nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
1287         if (!nest_parms)
1288                 goto nla_put_failure;
1289         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1290                 goto nla_put_failure;
1291         nla_nest_end(skb, nest_parms);
1292
1293         return 0;
1294
1295 nla_put_failure:
1296         return -1;
1297 }
1298
1299 static inline int
1300 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1301                         const struct nf_conntrack_tuple *tuple,
1302                         const struct nf_conntrack_tuple_mask *mask)
1303 {
1304         int ret;
1305         struct nf_conntrack_l3proto *l3proto;
1306         struct nf_conntrack_l4proto *l4proto;
1307         struct nf_conntrack_tuple m;
1308         struct nlattr *nest_parms;
1309
1310         memset(&m, 0xFF, sizeof(m));
1311         m.src.u.all = mask->src.u.all;
1312         memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3));
1313
1314         nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK | NLA_F_NESTED);
1315         if (!nest_parms)
1316                 goto nla_put_failure;
1317
1318         l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
1319         ret = ctnetlink_dump_tuples_ip(skb, &m, l3proto);
1320         nf_ct_l3proto_put(l3proto);
1321
1322         if (unlikely(ret < 0))
1323                 goto nla_put_failure;
1324
1325         l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
1326         ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto);
1327         nf_ct_l4proto_put(l4proto);
1328         if (unlikely(ret < 0))
1329                 goto nla_put_failure;
1330
1331         nla_nest_end(skb, nest_parms);
1332
1333         return 0;
1334
1335 nla_put_failure:
1336         return -1;
1337 }
1338
1339 static int
1340 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1341                           const struct nf_conntrack_expect *exp)
1342 {
1343         struct nf_conn *master = exp->master;
1344         long timeout = (exp->timeout.expires - jiffies) / HZ;
1345
1346         if (timeout < 0)
1347                 timeout = 0;
1348
1349         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1350                 goto nla_put_failure;
1351         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1352                 goto nla_put_failure;
1353         if (ctnetlink_exp_dump_tuple(skb,
1354                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1355                                  CTA_EXPECT_MASTER) < 0)
1356                 goto nla_put_failure;
1357
1358         NLA_PUT_BE32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout));
1359         NLA_PUT_BE32(skb, CTA_EXPECT_ID, htonl((unsigned long)exp));
1360
1361         return 0;
1362
1363 nla_put_failure:
1364         return -1;
1365 }
1366
1367 static int
1368 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1369                     int event,
1370                     int nowait,
1371                     const struct nf_conntrack_expect *exp)
1372 {
1373         struct nlmsghdr *nlh;
1374         struct nfgenmsg *nfmsg;
1375         unsigned char *b = skb_tail_pointer(skb);
1376
1377         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1378         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1379         nfmsg  = NLMSG_DATA(nlh);
1380
1381         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1382         nfmsg->nfgen_family = exp->tuple.src.l3num;
1383         nfmsg->version      = NFNETLINK_V0;
1384         nfmsg->res_id       = 0;
1385
1386         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1387                 goto nla_put_failure;
1388
1389         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
1390         return skb->len;
1391
1392 nlmsg_failure:
1393 nla_put_failure:
1394         nlmsg_trim(skb, b);
1395         return -1;
1396 }
1397
1398 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1399 static int ctnetlink_expect_event(struct notifier_block *this,
1400                                   unsigned long events, void *ptr)
1401 {
1402         struct nlmsghdr *nlh;
1403         struct nfgenmsg *nfmsg;
1404         struct nf_conntrack_expect *exp = (struct nf_conntrack_expect *)ptr;
1405         struct sk_buff *skb;
1406         unsigned int type;
1407         sk_buff_data_t b;
1408         int flags = 0;
1409
1410         if (events & IPEXP_NEW) {
1411                 type = IPCTNL_MSG_EXP_NEW;
1412                 flags = NLM_F_CREATE|NLM_F_EXCL;
1413         } else
1414                 return NOTIFY_DONE;
1415
1416         if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1417                 return NOTIFY_DONE;
1418
1419         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1420         if (!skb)
1421                 return NOTIFY_DONE;
1422
1423         b = skb->tail;
1424
1425         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1426         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1427         nfmsg = NLMSG_DATA(nlh);
1428
1429         nlh->nlmsg_flags    = flags;
1430         nfmsg->nfgen_family = exp->tuple.src.l3num;
1431         nfmsg->version      = NFNETLINK_V0;
1432         nfmsg->res_id       = 0;
1433
1434         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1435                 goto nla_put_failure;
1436
1437         nlh->nlmsg_len = skb->tail - b;
1438         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1439         return NOTIFY_DONE;
1440
1441 nlmsg_failure:
1442 nla_put_failure:
1443         kfree_skb(skb);
1444         return NOTIFY_DONE;
1445 }
1446 #endif
1447 static int ctnetlink_exp_done(struct netlink_callback *cb)
1448 {
1449         if (cb->args[1])
1450                 nf_ct_expect_put((struct nf_conntrack_expect *)cb->args[1]);
1451         return 0;
1452 }
1453
1454 static int
1455 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1456 {
1457         struct nf_conntrack_expect *exp, *last;
1458         struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
1459         struct hlist_node *n;
1460         u_int8_t l3proto = nfmsg->nfgen_family;
1461
1462         rcu_read_lock();
1463         last = (struct nf_conntrack_expect *)cb->args[1];
1464         for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) {
1465 restart:
1466                 hlist_for_each_entry(exp, n, &nf_ct_expect_hash[cb->args[0]],
1467                                      hnode) {
1468                         if (l3proto && exp->tuple.src.l3num != l3proto)
1469                                 continue;
1470                         if (cb->args[1]) {
1471                                 if (exp != last)
1472                                         continue;
1473                                 cb->args[1] = 0;
1474                         }
1475                         if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1476                                                     cb->nlh->nlmsg_seq,
1477                                                     IPCTNL_MSG_EXP_NEW,
1478                                                     1, exp) < 0) {
1479                                 if (!atomic_inc_not_zero(&exp->use))
1480                                         continue;
1481                                 cb->args[1] = (unsigned long)exp;
1482                                 goto out;
1483                         }
1484                 }
1485                 if (cb->args[1]) {
1486                         cb->args[1] = 0;
1487                         goto restart;
1488                 }
1489         }
1490 out:
1491         rcu_read_unlock();
1492         if (last)
1493                 nf_ct_expect_put(last);
1494
1495         return skb->len;
1496 }
1497
1498 static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = {
1499         [CTA_EXPECT_TIMEOUT]    = { .type = NLA_U32 },
1500         [CTA_EXPECT_ID]         = { .type = NLA_U32 },
1501 };
1502
1503 static int
1504 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1505                      struct nlmsghdr *nlh, struct nlattr *cda[])
1506 {
1507         struct nf_conntrack_tuple tuple;
1508         struct nf_conntrack_expect *exp;
1509         struct sk_buff *skb2;
1510         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1511         u_int8_t u3 = nfmsg->nfgen_family;
1512         int err = 0;
1513
1514         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1515                 return netlink_dump_start(ctnl, skb, nlh,
1516                                           ctnetlink_exp_dump_table,
1517                                           ctnetlink_exp_done);
1518         }
1519
1520         if (cda[CTA_EXPECT_MASTER])
1521                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER, u3);
1522         else
1523                 return -EINVAL;
1524
1525         if (err < 0)
1526                 return err;
1527
1528         exp = nf_ct_expect_find_get(&tuple);
1529         if (!exp)
1530                 return -ENOENT;
1531
1532         if (cda[CTA_EXPECT_ID]) {
1533                 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1534                 if (ntohl(id) != (u32)(unsigned long)exp) {
1535                         nf_ct_expect_put(exp);
1536                         return -ENOENT;
1537                 }
1538         }
1539
1540         err = -ENOMEM;
1541         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1542         if (!skb2)
1543                 goto out;
1544
1545         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1546                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1547                                       1, exp);
1548         if (err <= 0)
1549                 goto free;
1550
1551         nf_ct_expect_put(exp);
1552
1553         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1554
1555 free:
1556         kfree_skb(skb2);
1557 out:
1558         nf_ct_expect_put(exp);
1559         return err;
1560 }
1561
1562 static int
1563 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1564                      struct nlmsghdr *nlh, struct nlattr *cda[])
1565 {
1566         struct nf_conntrack_expect *exp;
1567         struct nf_conntrack_tuple tuple;
1568         struct nf_conntrack_helper *h;
1569         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1570         struct hlist_node *n, *next;
1571         u_int8_t u3 = nfmsg->nfgen_family;
1572         unsigned int i;
1573         int err;
1574
1575         if (cda[CTA_EXPECT_TUPLE]) {
1576                 /* delete a single expect by tuple */
1577                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1578                 if (err < 0)
1579                         return err;
1580
1581                 /* bump usage count to 2 */
1582                 exp = nf_ct_expect_find_get(&tuple);
1583                 if (!exp)
1584                         return -ENOENT;
1585
1586                 if (cda[CTA_EXPECT_ID]) {
1587                         __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1588                         if (ntohl(id) != (u32)(unsigned long)exp) {
1589                                 nf_ct_expect_put(exp);
1590                                 return -ENOENT;
1591                         }
1592                 }
1593
1594                 /* after list removal, usage count == 1 */
1595                 nf_ct_unexpect_related(exp);
1596                 /* have to put what we 'get' above.
1597                  * after this line usage count == 0 */
1598                 nf_ct_expect_put(exp);
1599         } else if (cda[CTA_EXPECT_HELP_NAME]) {
1600                 char *name = nla_data(cda[CTA_EXPECT_HELP_NAME]);
1601                 struct nf_conn_help *m_help;
1602
1603                 /* delete all expectations for this helper */
1604                 spin_lock_bh(&nf_conntrack_lock);
1605                 h = __nf_conntrack_helper_find_byname(name);
1606                 if (!h) {
1607                         spin_unlock_bh(&nf_conntrack_lock);
1608                         return -EINVAL;
1609                 }
1610                 for (i = 0; i < nf_ct_expect_hsize; i++) {
1611                         hlist_for_each_entry_safe(exp, n, next,
1612                                                   &nf_ct_expect_hash[i],
1613                                                   hnode) {
1614                                 m_help = nfct_help(exp->master);
1615                                 if (m_help->helper == h
1616                                     && del_timer(&exp->timeout)) {
1617                                         nf_ct_unlink_expect(exp);
1618                                         nf_ct_expect_put(exp);
1619                                 }
1620                         }
1621                 }
1622                 spin_unlock_bh(&nf_conntrack_lock);
1623         } else {
1624                 /* This basically means we have to flush everything*/
1625                 spin_lock_bh(&nf_conntrack_lock);
1626                 for (i = 0; i < nf_ct_expect_hsize; i++) {
1627                         hlist_for_each_entry_safe(exp, n, next,
1628                                                   &nf_ct_expect_hash[i],
1629                                                   hnode) {
1630                                 if (del_timer(&exp->timeout)) {
1631                                         nf_ct_unlink_expect(exp);
1632                                         nf_ct_expect_put(exp);
1633                                 }
1634                         }
1635                 }
1636                 spin_unlock_bh(&nf_conntrack_lock);
1637         }
1638
1639         return 0;
1640 }
1641 static int
1642 ctnetlink_change_expect(struct nf_conntrack_expect *x, struct nlattr *cda[])
1643 {
1644         return -EOPNOTSUPP;
1645 }
1646
1647 static int
1648 ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3)
1649 {
1650         struct nf_conntrack_tuple tuple, mask, master_tuple;
1651         struct nf_conntrack_tuple_hash *h = NULL;
1652         struct nf_conntrack_expect *exp;
1653         struct nf_conn *ct;
1654         struct nf_conn_help *help;
1655         int err = 0;
1656
1657         /* caller guarantees that those three CTA_EXPECT_* exist */
1658         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1659         if (err < 0)
1660                 return err;
1661         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK, u3);
1662         if (err < 0)
1663                 return err;
1664         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER, u3);
1665         if (err < 0)
1666                 return err;
1667
1668         /* Look for master conntrack of this expectation */
1669         h = nf_conntrack_find_get(&master_tuple);
1670         if (!h)
1671                 return -ENOENT;
1672         ct = nf_ct_tuplehash_to_ctrack(h);
1673         help = nfct_help(ct);
1674
1675         if (!help || !help->helper) {
1676                 /* such conntrack hasn't got any helper, abort */
1677                 err = -EINVAL;
1678                 goto out;
1679         }
1680
1681         exp = nf_ct_expect_alloc(ct);
1682         if (!exp) {
1683                 err = -ENOMEM;
1684                 goto out;
1685         }
1686
1687         exp->expectfn = NULL;
1688         exp->flags = 0;
1689         exp->master = ct;
1690         exp->helper = NULL;
1691         memcpy(&exp->tuple, &tuple, sizeof(struct nf_conntrack_tuple));
1692         memcpy(&exp->mask.src.u3, &mask.src.u3, sizeof(exp->mask.src.u3));
1693         exp->mask.src.u.all = mask.src.u.all;
1694
1695         err = nf_ct_expect_related(exp);
1696         nf_ct_expect_put(exp);
1697
1698 out:
1699         nf_ct_put(nf_ct_tuplehash_to_ctrack(h));
1700         return err;
1701 }
1702
1703 static int
1704 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1705                      struct nlmsghdr *nlh, struct nlattr *cda[])
1706 {
1707         struct nf_conntrack_tuple tuple;
1708         struct nf_conntrack_expect *exp;
1709         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1710         u_int8_t u3 = nfmsg->nfgen_family;
1711         int err = 0;
1712
1713         if (!cda[CTA_EXPECT_TUPLE]
1714             || !cda[CTA_EXPECT_MASK]
1715             || !cda[CTA_EXPECT_MASTER])
1716                 return -EINVAL;
1717
1718         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1719         if (err < 0)
1720                 return err;
1721
1722         spin_lock_bh(&nf_conntrack_lock);
1723         exp = __nf_ct_expect_find(&tuple);
1724
1725         if (!exp) {
1726                 spin_unlock_bh(&nf_conntrack_lock);
1727                 err = -ENOENT;
1728                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1729                         err = ctnetlink_create_expect(cda, u3);
1730                 return err;
1731         }
1732
1733         err = -EEXIST;
1734         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1735                 err = ctnetlink_change_expect(exp, cda);
1736         spin_unlock_bh(&nf_conntrack_lock);
1737
1738         return err;
1739 }
1740
1741 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1742 static struct notifier_block ctnl_notifier = {
1743         .notifier_call  = ctnetlink_conntrack_event,
1744 };
1745
1746 static struct notifier_block ctnl_notifier_exp = {
1747         .notifier_call  = ctnetlink_expect_event,
1748 };
1749 #endif
1750
1751 static const struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1752         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1753                                             .attr_count = CTA_MAX,
1754                                             .policy = ct_nla_policy },
1755         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1756                                             .attr_count = CTA_MAX,
1757                                             .policy = ct_nla_policy },
1758         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1759                                             .attr_count = CTA_MAX,
1760                                             .policy = ct_nla_policy },
1761         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1762                                             .attr_count = CTA_MAX,
1763                                             .policy = ct_nla_policy },
1764 };
1765
1766 static const struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1767         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1768                                             .attr_count = CTA_EXPECT_MAX,
1769                                             .policy = exp_nla_policy },
1770         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1771                                             .attr_count = CTA_EXPECT_MAX,
1772                                             .policy = exp_nla_policy },
1773         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1774                                             .attr_count = CTA_EXPECT_MAX,
1775                                             .policy = exp_nla_policy },
1776 };
1777
1778 static const struct nfnetlink_subsystem ctnl_subsys = {
1779         .name                           = "conntrack",
1780         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1781         .cb_count                       = IPCTNL_MSG_MAX,
1782         .cb                             = ctnl_cb,
1783 };
1784
1785 static const struct nfnetlink_subsystem ctnl_exp_subsys = {
1786         .name                           = "conntrack_expect",
1787         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1788         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1789         .cb                             = ctnl_exp_cb,
1790 };
1791
1792 MODULE_ALIAS("ip_conntrack_netlink");
1793 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1794 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1795
1796 static int __init ctnetlink_init(void)
1797 {
1798         int ret;
1799
1800         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1801         ret = nfnetlink_subsys_register(&ctnl_subsys);
1802         if (ret < 0) {
1803                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1804                 goto err_out;
1805         }
1806
1807         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1808         if (ret < 0) {
1809                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1810                 goto err_unreg_subsys;
1811         }
1812
1813 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1814         ret = nf_conntrack_register_notifier(&ctnl_notifier);
1815         if (ret < 0) {
1816                 printk("ctnetlink_init: cannot register notifier.\n");
1817                 goto err_unreg_exp_subsys;
1818         }
1819
1820         ret = nf_ct_expect_register_notifier(&ctnl_notifier_exp);
1821         if (ret < 0) {
1822                 printk("ctnetlink_init: cannot expect register notifier.\n");
1823                 goto err_unreg_notifier;
1824         }
1825 #endif
1826
1827         return 0;
1828
1829 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1830 err_unreg_notifier:
1831         nf_conntrack_unregister_notifier(&ctnl_notifier);
1832 err_unreg_exp_subsys:
1833         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1834 #endif
1835 err_unreg_subsys:
1836         nfnetlink_subsys_unregister(&ctnl_subsys);
1837 err_out:
1838         return ret;
1839 }
1840
1841 static void __exit ctnetlink_exit(void)
1842 {
1843         printk("ctnetlink: unregistering from nfnetlink.\n");
1844
1845 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1846         nf_ct_expect_unregister_notifier(&ctnl_notifier_exp);
1847         nf_conntrack_unregister_notifier(&ctnl_notifier);
1848 #endif
1849
1850         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1851         nfnetlink_subsys_unregister(&ctnl_subsys);
1852         return;
1853 }
1854
1855 module_init(ctnetlink_init);
1856 module_exit(ctnetlink_exit);