[NETFILTER]: nf_conntrack: use RCU for conntrack hash
[safe/jmp/linux-2.6] / net / netfilter / nf_conntrack_netlink.c
1 /* Connection tracking via netlink socket. Allows for user space
2  * protocol helpers and general trouble making from userspace.
3  *
4  * (C) 2001 by Jay Schulist <jschlst@samba.org>
5  * (C) 2002-2006 by Harald Welte <laforge@gnumonks.org>
6  * (C) 2003 by Patrick Mchardy <kaber@trash.net>
7  * (C) 2005-2007 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 = nf_ct_l4proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num, ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
149         struct nlattr *nest_proto;
150         int ret;
151
152         if (!l4proto->to_nlattr) {
153                 nf_ct_l4proto_put(l4proto);
154                 return 0;
155         }
156
157         nest_proto = nla_nest_start(skb, CTA_PROTOINFO | NLA_F_NESTED);
158         if (!nest_proto)
159                 goto nla_put_failure;
160
161         ret = l4proto->to_nlattr(skb, nest_proto, ct);
162
163         nf_ct_l4proto_put(l4proto);
164
165         nla_nest_end(skb, nest_proto);
166
167         return ret;
168
169 nla_put_failure:
170         nf_ct_l4proto_put(l4proto);
171         return -1;
172 }
173
174 static inline int
175 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct nf_conn *ct)
176 {
177         struct nlattr *nest_helper;
178         const struct nf_conn_help *help = nfct_help(ct);
179         struct nf_conntrack_helper *helper;
180
181         if (!help)
182                 return 0;
183
184         rcu_read_lock();
185         helper = rcu_dereference(help->helper);
186         if (!helper)
187                 goto out;
188
189         nest_helper = nla_nest_start(skb, CTA_HELP | NLA_F_NESTED);
190         if (!nest_helper)
191                 goto nla_put_failure;
192         NLA_PUT_STRING(skb, CTA_HELP_NAME, helper->name);
193
194         if (helper->to_nlattr)
195                 helper->to_nlattr(skb, ct);
196
197         nla_nest_end(skb, nest_helper);
198 out:
199         rcu_read_unlock();
200         return 0;
201
202 nla_put_failure:
203         rcu_read_unlock();
204         return -1;
205 }
206
207 #ifdef CONFIG_NF_CT_ACCT
208 static int
209 ctnetlink_dump_counters(struct sk_buff *skb, const struct nf_conn *ct,
210                         enum ip_conntrack_dir dir)
211 {
212         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
213         struct nlattr *nest_count;
214
215         nest_count = nla_nest_start(skb, type | NLA_F_NESTED);
216         if (!nest_count)
217                 goto nla_put_failure;
218
219         NLA_PUT_BE32(skb, CTA_COUNTERS32_PACKETS,
220                      htonl(ct->counters[dir].packets));
221         NLA_PUT_BE32(skb, CTA_COUNTERS32_BYTES,
222                      htonl(ct->counters[dir].bytes));
223
224         nla_nest_end(skb, nest_count);
225
226         return 0;
227
228 nla_put_failure:
229         return -1;
230 }
231 #else
232 #define ctnetlink_dump_counters(a, b, c) (0)
233 #endif
234
235 #ifdef CONFIG_NF_CONNTRACK_MARK
236 static inline int
237 ctnetlink_dump_mark(struct sk_buff *skb, const struct nf_conn *ct)
238 {
239         NLA_PUT_BE32(skb, CTA_MARK, htonl(ct->mark));
240         return 0;
241
242 nla_put_failure:
243         return -1;
244 }
245 #else
246 #define ctnetlink_dump_mark(a, b) (0)
247 #endif
248
249 #ifdef CONFIG_NF_CONNTRACK_SECMARK
250 static inline int
251 ctnetlink_dump_secmark(struct sk_buff *skb, const struct nf_conn *ct)
252 {
253         NLA_PUT_BE32(skb, CTA_SECMARK, htonl(ct->secmark));
254         return 0;
255
256 nla_put_failure:
257         return -1;
258 }
259 #else
260 #define ctnetlink_dump_secmark(a, b) (0)
261 #endif
262
263 #define master_tuple(ct) &(ct->master->tuplehash[IP_CT_DIR_ORIGINAL].tuple)
264
265 static inline int
266 ctnetlink_dump_master(struct sk_buff *skb, const struct nf_conn *ct)
267 {
268         struct nlattr *nest_parms;
269
270         if (!(ct->status & IPS_EXPECTED))
271                 return 0;
272
273         nest_parms = nla_nest_start(skb, CTA_TUPLE_MASTER | NLA_F_NESTED);
274         if (!nest_parms)
275                 goto nla_put_failure;
276         if (ctnetlink_dump_tuples(skb, master_tuple(ct)) < 0)
277                 goto nla_put_failure;
278         nla_nest_end(skb, nest_parms);
279
280         return 0;
281
282 nla_put_failure:
283         return -1;
284 }
285
286 #ifdef CONFIG_NF_NAT_NEEDED
287 static int
288 dump_nat_seq_adj(struct sk_buff *skb, const struct nf_nat_seq *natseq, int type)
289 {
290         struct nlattr *nest_parms;
291
292         nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
293         if (!nest_parms)
294                 goto nla_put_failure;
295
296         NLA_PUT_BE32(skb, CTA_NAT_SEQ_CORRECTION_POS,
297                      htonl(natseq->correction_pos));
298         NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_BEFORE,
299                      htonl(natseq->offset_before));
300         NLA_PUT_BE32(skb, CTA_NAT_SEQ_OFFSET_AFTER,
301                      htonl(natseq->offset_after));
302
303         nla_nest_end(skb, nest_parms);
304
305         return 0;
306
307 nla_put_failure:
308         return -1;
309 }
310
311 static inline int
312 ctnetlink_dump_nat_seq_adj(struct sk_buff *skb, const struct nf_conn *ct)
313 {
314         struct nf_nat_seq *natseq;
315         struct nf_conn_nat *nat = nfct_nat(ct);
316
317         if (!(ct->status & IPS_SEQ_ADJUST) || !nat)
318                 return 0;
319
320         natseq = &nat->seq[IP_CT_DIR_ORIGINAL];
321         if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_ORIG) == -1)
322                 return -1;
323
324         natseq = &nat->seq[IP_CT_DIR_REPLY];
325         if (dump_nat_seq_adj(skb, natseq, CTA_NAT_SEQ_ADJ_REPLY) == -1)
326                 return -1;
327
328         return 0;
329 }
330 #else
331 #define ctnetlink_dump_nat_seq_adj(a, b) (0)
332 #endif
333
334 static inline int
335 ctnetlink_dump_id(struct sk_buff *skb, const struct nf_conn *ct)
336 {
337         NLA_PUT_BE32(skb, CTA_ID, htonl((unsigned long)ct));
338         return 0;
339
340 nla_put_failure:
341         return -1;
342 }
343
344 static inline int
345 ctnetlink_dump_use(struct sk_buff *skb, const struct nf_conn *ct)
346 {
347         NLA_PUT_BE32(skb, CTA_USE, htonl(atomic_read(&ct->ct_general.use)));
348         return 0;
349
350 nla_put_failure:
351         return -1;
352 }
353
354 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
355
356 static int
357 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
358                     int event, int nowait,
359                     const struct nf_conn *ct)
360 {
361         struct nlmsghdr *nlh;
362         struct nfgenmsg *nfmsg;
363         struct nlattr *nest_parms;
364         unsigned char *b = skb_tail_pointer(skb);
365
366         event |= NFNL_SUBSYS_CTNETLINK << 8;
367         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
368         nfmsg  = NLMSG_DATA(nlh);
369
370         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
371         nfmsg->nfgen_family =
372                 ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
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 = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
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_MARK
495                 if ((events & IPCT_MARK || ct->mark)
496                     && ctnetlink_dump_mark(skb, ct) < 0)
497                         goto nla_put_failure;
498 #endif
499 #ifdef CONFIG_NF_CONNTRACK_SECMARK
500                 if ((events & IPCT_SECMARK || ct->secmark)
501                     && ctnetlink_dump_secmark(skb, ct) < 0)
502                         goto nla_put_failure;
503 #endif
504
505                 if (events & IPCT_COUNTER_FILLING &&
506                     (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
507                      ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0))
508                         goto nla_put_failure;
509
510                 if (events & IPCT_RELATED &&
511                     ctnetlink_dump_master(skb, ct) < 0)
512                         goto nla_put_failure;
513
514                 if (events & IPCT_NATSEQADJ &&
515                     ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
516                         goto nla_put_failure;
517         }
518
519         nlh->nlmsg_len = skb->tail - b;
520         nfnetlink_send(skb, 0, group, 0);
521         return NOTIFY_DONE;
522
523 nlmsg_failure:
524 nla_put_failure:
525         kfree_skb(skb);
526         return NOTIFY_DONE;
527 }
528 #endif /* CONFIG_NF_CONNTRACK_EVENTS */
529
530 static int ctnetlink_done(struct netlink_callback *cb)
531 {
532         if (cb->args[1])
533                 nf_ct_put((struct nf_conn *)cb->args[1]);
534         return 0;
535 }
536
537 #define L3PROTO(ct) (ct)->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num
538
539 static int
540 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
541 {
542         struct nf_conn *ct, *last;
543         struct nf_conntrack_tuple_hash *h;
544         struct hlist_node *n;
545         struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
546         u_int8_t l3proto = nfmsg->nfgen_family;
547
548         rcu_read_lock();
549         last = (struct nf_conn *)cb->args[1];
550         for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
551 restart:
552                 hlist_for_each_entry_rcu(h, n, &nf_conntrack_hash[cb->args[0]],
553                                          hnode) {
554                         if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
555                                 continue;
556                         ct = nf_ct_tuplehash_to_ctrack(h);
557                         /* Dump entries of a given L3 protocol number.
558                          * If it is not specified, ie. l3proto == 0,
559                          * then dump everything. */
560                         if (l3proto && L3PROTO(ct) != l3proto)
561                                 continue;
562                         if (cb->args[1]) {
563                                 if (ct != last)
564                                         continue;
565                                 cb->args[1] = 0;
566                         }
567                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
568                                                 cb->nlh->nlmsg_seq,
569                                                 IPCTNL_MSG_CT_NEW,
570                                                 1, ct) < 0) {
571                                 if (!atomic_inc_not_zero(&ct->ct_general.use))
572                                         continue;
573                                 cb->args[1] = (unsigned long)ct;
574                                 goto out;
575                         }
576 #ifdef CONFIG_NF_CT_ACCT
577                         if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
578                                                 IPCTNL_MSG_CT_GET_CTRZERO)
579                                 memset(&ct->counters, 0, sizeof(ct->counters));
580 #endif
581                 }
582                 if (cb->args[1]) {
583                         cb->args[1] = 0;
584                         goto restart;
585                 }
586         }
587 out:
588         rcu_read_unlock();
589         if (last)
590                 nf_ct_put(last);
591
592         return skb->len;
593 }
594
595 static inline int
596 ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
597 {
598         struct nlattr *tb[CTA_IP_MAX+1];
599         struct nf_conntrack_l3proto *l3proto;
600         int ret = 0;
601
602         nla_parse_nested(tb, CTA_IP_MAX, attr, NULL);
603
604         l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
605
606         if (likely(l3proto->nlattr_to_tuple)) {
607                 ret = nla_validate_nested(attr, CTA_IP_MAX,
608                                           l3proto->nla_policy);
609                 if (ret == 0)
610                         ret = l3proto->nlattr_to_tuple(tb, tuple);
611         }
612
613         nf_ct_l3proto_put(l3proto);
614
615         return ret;
616 }
617
618 static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = {
619         [CTA_PROTO_NUM] = { .type = NLA_U8 },
620 };
621
622 static inline int
623 ctnetlink_parse_tuple_proto(struct nlattr *attr,
624                             struct nf_conntrack_tuple *tuple)
625 {
626         struct nlattr *tb[CTA_PROTO_MAX+1];
627         struct nf_conntrack_l4proto *l4proto;
628         int ret = 0;
629
630         ret = nla_parse_nested(tb, CTA_PROTO_MAX, attr, proto_nla_policy);
631         if (ret < 0)
632                 return ret;
633
634         if (!tb[CTA_PROTO_NUM])
635                 return -EINVAL;
636         tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
637
638         l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
639
640         if (likely(l4proto->nlattr_to_tuple)) {
641                 ret = nla_validate_nested(attr, CTA_PROTO_MAX,
642                                           l4proto->nla_policy);
643                 if (ret == 0)
644                         ret = l4proto->nlattr_to_tuple(tb, tuple);
645         }
646
647         nf_ct_l4proto_put(l4proto);
648
649         return ret;
650 }
651
652 static int
653 ctnetlink_parse_tuple(struct nlattr *cda[], struct nf_conntrack_tuple *tuple,
654                       enum ctattr_tuple type, u_int8_t l3num)
655 {
656         struct nlattr *tb[CTA_TUPLE_MAX+1];
657         int err;
658
659         memset(tuple, 0, sizeof(*tuple));
660
661         nla_parse_nested(tb, CTA_TUPLE_MAX, cda[type], NULL);
662
663         if (!tb[CTA_TUPLE_IP])
664                 return -EINVAL;
665
666         tuple->src.l3num = l3num;
667
668         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple);
669         if (err < 0)
670                 return err;
671
672         if (!tb[CTA_TUPLE_PROTO])
673                 return -EINVAL;
674
675         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple);
676         if (err < 0)
677                 return err;
678
679         /* orig and expect tuples get DIR_ORIGINAL */
680         if (type == CTA_TUPLE_REPLY)
681                 tuple->dst.dir = IP_CT_DIR_REPLY;
682         else
683                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
684
685         return 0;
686 }
687
688 #ifdef CONFIG_NF_NAT_NEEDED
689 static const struct nla_policy protonat_nla_policy[CTA_PROTONAT_MAX+1] = {
690         [CTA_PROTONAT_PORT_MIN] = { .type = NLA_U16 },
691         [CTA_PROTONAT_PORT_MAX] = { .type = NLA_U16 },
692 };
693
694 static int nfnetlink_parse_nat_proto(struct nlattr *attr,
695                                      const struct nf_conn *ct,
696                                      struct nf_nat_range *range)
697 {
698         struct nlattr *tb[CTA_PROTONAT_MAX+1];
699         const struct nf_nat_protocol *npt;
700         int err;
701
702         err = nla_parse_nested(tb, CTA_PROTONAT_MAX, attr, protonat_nla_policy);
703         if (err < 0)
704                 return err;
705
706         npt = nf_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
707
708         if (!npt->nlattr_to_range) {
709                 nf_nat_proto_put(npt);
710                 return 0;
711         }
712
713         /* nlattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
714         if (npt->nlattr_to_range(tb, range) > 0)
715                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
716
717         nf_nat_proto_put(npt);
718
719         return 0;
720 }
721
722 static const struct nla_policy nat_nla_policy[CTA_NAT_MAX+1] = {
723         [CTA_NAT_MINIP]         = { .type = NLA_U32 },
724         [CTA_NAT_MAXIP]         = { .type = NLA_U32 },
725 };
726
727 static inline int
728 nfnetlink_parse_nat(struct nlattr *nat,
729                     const struct nf_conn *ct, struct nf_nat_range *range)
730 {
731         struct nlattr *tb[CTA_NAT_MAX+1];
732         int err;
733
734         memset(range, 0, sizeof(*range));
735
736         err = nla_parse_nested(tb, CTA_NAT_MAX, nat, nat_nla_policy);
737         if (err < 0)
738                 return err;
739
740         if (tb[CTA_NAT_MINIP])
741                 range->min_ip = nla_get_be32(tb[CTA_NAT_MINIP]);
742
743         if (!tb[CTA_NAT_MAXIP])
744                 range->max_ip = range->min_ip;
745         else
746                 range->max_ip = nla_get_be32(tb[CTA_NAT_MAXIP]);
747
748         if (range->min_ip)
749                 range->flags |= IP_NAT_RANGE_MAP_IPS;
750
751         if (!tb[CTA_NAT_PROTO])
752                 return 0;
753
754         err = nfnetlink_parse_nat_proto(tb[CTA_NAT_PROTO], ct, range);
755         if (err < 0)
756                 return err;
757
758         return 0;
759 }
760 #endif
761
762 static inline int
763 ctnetlink_parse_help(struct nlattr *attr, char **helper_name)
764 {
765         struct nlattr *tb[CTA_HELP_MAX+1];
766
767         nla_parse_nested(tb, CTA_HELP_MAX, attr, NULL);
768
769         if (!tb[CTA_HELP_NAME])
770                 return -EINVAL;
771
772         *helper_name = nla_data(tb[CTA_HELP_NAME]);
773
774         return 0;
775 }
776
777 static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
778         [CTA_STATUS]            = { .type = NLA_U32 },
779         [CTA_TIMEOUT]           = { .type = NLA_U32 },
780         [CTA_MARK]              = { .type = NLA_U32 },
781         [CTA_USE]               = { .type = NLA_U32 },
782         [CTA_ID]                = { .type = NLA_U32 },
783 };
784
785 static int
786 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
787                         struct nlmsghdr *nlh, struct nlattr *cda[])
788 {
789         struct nf_conntrack_tuple_hash *h;
790         struct nf_conntrack_tuple tuple;
791         struct nf_conn *ct;
792         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
793         u_int8_t u3 = nfmsg->nfgen_family;
794         int err = 0;
795
796         if (cda[CTA_TUPLE_ORIG])
797                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
798         else if (cda[CTA_TUPLE_REPLY])
799                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
800         else {
801                 /* Flush the whole table */
802                 nf_conntrack_flush();
803                 return 0;
804         }
805
806         if (err < 0)
807                 return err;
808
809         h = nf_conntrack_find_get(&tuple);
810         if (!h)
811                 return -ENOENT;
812
813         ct = nf_ct_tuplehash_to_ctrack(h);
814
815         if (cda[CTA_ID]) {
816                 u_int32_t id = ntohl(nla_get_be32(cda[CTA_ID]));
817                 if (id != (u32)(unsigned long)ct) {
818                         nf_ct_put(ct);
819                         return -ENOENT;
820                 }
821         }
822         if (del_timer(&ct->timeout))
823                 ct->timeout.function((unsigned long)ct);
824
825         nf_ct_put(ct);
826
827         return 0;
828 }
829
830 static int
831 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
832                         struct nlmsghdr *nlh, struct nlattr *cda[])
833 {
834         struct nf_conntrack_tuple_hash *h;
835         struct nf_conntrack_tuple tuple;
836         struct nf_conn *ct;
837         struct sk_buff *skb2 = NULL;
838         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
839         u_int8_t u3 = nfmsg->nfgen_family;
840         int err = 0;
841
842         if (nlh->nlmsg_flags & NLM_F_DUMP) {
843 #ifndef CONFIG_NF_CT_ACCT
844                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) == IPCTNL_MSG_CT_GET_CTRZERO)
845                         return -ENOTSUPP;
846 #endif
847                 return netlink_dump_start(ctnl, skb, nlh, ctnetlink_dump_table,
848                                           ctnetlink_done);
849         }
850
851         if (cda[CTA_TUPLE_ORIG])
852                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
853         else if (cda[CTA_TUPLE_REPLY])
854                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
855         else
856                 return -EINVAL;
857
858         if (err < 0)
859                 return err;
860
861         h = nf_conntrack_find_get(&tuple);
862         if (!h)
863                 return -ENOENT;
864
865         ct = nf_ct_tuplehash_to_ctrack(h);
866
867         err = -ENOMEM;
868         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
869         if (!skb2) {
870                 nf_ct_put(ct);
871                 return -ENOMEM;
872         }
873
874         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
875                                   IPCTNL_MSG_CT_NEW, 1, ct);
876         nf_ct_put(ct);
877         if (err <= 0)
878                 goto free;
879
880         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
881         if (err < 0)
882                 goto out;
883
884         return 0;
885
886 free:
887         kfree_skb(skb2);
888 out:
889         return err;
890 }
891
892 static int
893 ctnetlink_change_status(struct nf_conn *ct, struct nlattr *cda[])
894 {
895         unsigned long d;
896         unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
897         d = ct->status ^ status;
898
899         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
900                 /* unchangeable */
901                 return -EINVAL;
902
903         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
904                 /* SEEN_REPLY bit can only be set */
905                 return -EINVAL;
906
907
908         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
909                 /* ASSURED bit can only be set */
910                 return -EINVAL;
911
912         if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
913 #ifndef CONFIG_NF_NAT_NEEDED
914                 return -EINVAL;
915 #else
916                 struct nf_nat_range range;
917
918                 if (cda[CTA_NAT_DST]) {
919                         if (nfnetlink_parse_nat(cda[CTA_NAT_DST], ct,
920                                                 &range) < 0)
921                                 return -EINVAL;
922                         if (nf_nat_initialized(ct, IP_NAT_MANIP_DST))
923                                 return -EEXIST;
924                         nf_nat_setup_info(ct, &range, IP_NAT_MANIP_DST);
925                 }
926                 if (cda[CTA_NAT_SRC]) {
927                         if (nfnetlink_parse_nat(cda[CTA_NAT_SRC], ct,
928                                                 &range) < 0)
929                                 return -EINVAL;
930                         if (nf_nat_initialized(ct, IP_NAT_MANIP_SRC))
931                                 return -EEXIST;
932                         nf_nat_setup_info(ct, &range, IP_NAT_MANIP_SRC);
933                 }
934 #endif
935         }
936
937         /* Be careful here, modifying NAT bits can screw up things,
938          * so don't let users modify them directly if they don't pass
939          * nf_nat_range. */
940         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
941         return 0;
942 }
943
944
945 static inline int
946 ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
947 {
948         struct nf_conntrack_helper *helper;
949         struct nf_conn_help *help = nfct_help(ct);
950         char *helpname;
951         int err;
952
953         /* don't change helper of sibling connections */
954         if (ct->master)
955                 return -EINVAL;
956
957         err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
958         if (err < 0)
959                 return err;
960
961         if (!strcmp(helpname, "")) {
962                 if (help && help->helper) {
963                         /* we had a helper before ... */
964                         nf_ct_remove_expectations(ct);
965                         rcu_assign_pointer(help->helper, NULL);
966                 }
967
968                 return 0;
969         }
970
971         helper = __nf_conntrack_helper_find_byname(helpname);
972         if (helper == NULL)
973                 return -EINVAL;
974
975         if (help) {
976                 if (help->helper == helper)
977                         return 0;
978                 if (help->helper)
979                         return -EBUSY;
980                 /* need to zero data of old helper */
981                 memset(&help->help, 0, sizeof(help->help));
982         } else {
983                 help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
984                 if (help == NULL)
985                         return -ENOMEM;
986         }
987
988         rcu_assign_pointer(help->helper, helper);
989
990         return 0;
991 }
992
993 static inline int
994 ctnetlink_change_timeout(struct nf_conn *ct, struct nlattr *cda[])
995 {
996         u_int32_t timeout = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
997
998         if (!del_timer(&ct->timeout))
999                 return -ETIME;
1000
1001         ct->timeout.expires = jiffies + timeout * HZ;
1002         add_timer(&ct->timeout);
1003
1004         return 0;
1005 }
1006
1007 static inline int
1008 ctnetlink_change_protoinfo(struct nf_conn *ct, struct nlattr *cda[])
1009 {
1010         struct nlattr *tb[CTA_PROTOINFO_MAX+1], *attr = cda[CTA_PROTOINFO];
1011         struct nf_conntrack_l4proto *l4proto;
1012         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
1013         u_int16_t l3num = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
1014         int err = 0;
1015
1016         nla_parse_nested(tb, CTA_PROTOINFO_MAX, attr, NULL);
1017
1018         l4proto = nf_ct_l4proto_find_get(l3num, npt);
1019
1020         if (l4proto->from_nlattr)
1021                 err = l4proto->from_nlattr(tb, ct);
1022         nf_ct_l4proto_put(l4proto);
1023
1024         return err;
1025 }
1026
1027 #ifdef CONFIG_NF_NAT_NEEDED
1028 static inline int
1029 change_nat_seq_adj(struct nf_nat_seq *natseq, struct nlattr *attr)
1030 {
1031         struct nlattr *cda[CTA_NAT_SEQ_MAX+1];
1032
1033         nla_parse_nested(cda, CTA_NAT_SEQ_MAX, attr, NULL);
1034
1035         if (!cda[CTA_NAT_SEQ_CORRECTION_POS])
1036                 return -EINVAL;
1037
1038         natseq->correction_pos =
1039                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_CORRECTION_POS]));
1040
1041         if (!cda[CTA_NAT_SEQ_OFFSET_BEFORE])
1042                 return -EINVAL;
1043
1044         natseq->offset_before =
1045                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_BEFORE]));
1046
1047         if (!cda[CTA_NAT_SEQ_OFFSET_AFTER])
1048                 return -EINVAL;
1049
1050         natseq->offset_after =
1051                 ntohl(nla_get_be32(cda[CTA_NAT_SEQ_OFFSET_AFTER]));
1052
1053         return 0;
1054 }
1055
1056 static int
1057 ctnetlink_change_nat_seq_adj(struct nf_conn *ct, struct nlattr *cda[])
1058 {
1059         int ret = 0;
1060         struct nf_conn_nat *nat = nfct_nat(ct);
1061
1062         if (!nat)
1063                 return 0;
1064
1065         if (cda[CTA_NAT_SEQ_ADJ_ORIG]) {
1066                 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_ORIGINAL],
1067                                          cda[CTA_NAT_SEQ_ADJ_ORIG]);
1068                 if (ret < 0)
1069                         return ret;
1070
1071                 ct->status |= IPS_SEQ_ADJUST;
1072         }
1073
1074         if (cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1075                 ret = change_nat_seq_adj(&nat->seq[IP_CT_DIR_REPLY],
1076                                          cda[CTA_NAT_SEQ_ADJ_REPLY]);
1077                 if (ret < 0)
1078                         return ret;
1079
1080                 ct->status |= IPS_SEQ_ADJUST;
1081         }
1082
1083         return 0;
1084 }
1085 #endif
1086
1087 static int
1088 ctnetlink_change_conntrack(struct nf_conn *ct, struct nlattr *cda[])
1089 {
1090         int err;
1091
1092         if (cda[CTA_HELP]) {
1093                 err = ctnetlink_change_helper(ct, cda);
1094                 if (err < 0)
1095                         return err;
1096         }
1097
1098         if (cda[CTA_TIMEOUT]) {
1099                 err = ctnetlink_change_timeout(ct, cda);
1100                 if (err < 0)
1101                         return err;
1102         }
1103
1104         if (cda[CTA_STATUS]) {
1105                 err = ctnetlink_change_status(ct, cda);
1106                 if (err < 0)
1107                         return err;
1108         }
1109
1110         if (cda[CTA_PROTOINFO]) {
1111                 err = ctnetlink_change_protoinfo(ct, cda);
1112                 if (err < 0)
1113                         return err;
1114         }
1115
1116 #if defined(CONFIG_NF_CONNTRACK_MARK)
1117         if (cda[CTA_MARK])
1118                 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1119 #endif
1120
1121 #ifdef CONFIG_NF_NAT_NEEDED
1122         if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
1123                 err = ctnetlink_change_nat_seq_adj(ct, cda);
1124                 if (err < 0)
1125                         return err;
1126         }
1127 #endif
1128
1129         return 0;
1130 }
1131
1132 static int
1133 ctnetlink_create_conntrack(struct nlattr *cda[],
1134                            struct nf_conntrack_tuple *otuple,
1135                            struct nf_conntrack_tuple *rtuple,
1136                            struct nf_conn *master_ct)
1137 {
1138         struct nf_conn *ct;
1139         int err = -EINVAL;
1140         struct nf_conn_help *help;
1141         struct nf_conntrack_helper *helper;
1142
1143         ct = nf_conntrack_alloc(otuple, rtuple);
1144         if (ct == NULL || IS_ERR(ct))
1145                 return -ENOMEM;
1146
1147         if (!cda[CTA_TIMEOUT])
1148                 goto err;
1149         ct->timeout.expires = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
1150
1151         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1152         ct->status |= IPS_CONFIRMED;
1153
1154         if (cda[CTA_STATUS]) {
1155                 err = ctnetlink_change_status(ct, cda);
1156                 if (err < 0)
1157                         goto err;
1158         }
1159
1160         if (cda[CTA_PROTOINFO]) {
1161                 err = ctnetlink_change_protoinfo(ct, cda);
1162                 if (err < 0)
1163                         goto err;
1164         }
1165
1166 #if defined(CONFIG_NF_CONNTRACK_MARK)
1167         if (cda[CTA_MARK])
1168                 ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
1169 #endif
1170
1171         rcu_read_lock();
1172         helper = __nf_ct_helper_find(rtuple);
1173         if (helper) {
1174                 help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
1175                 if (help == NULL) {
1176                         rcu_read_unlock();
1177                         err = -ENOMEM;
1178                         goto err;
1179                 }
1180                 /* not in hash table yet so not strictly necessary */
1181                 rcu_assign_pointer(help->helper, helper);
1182         }
1183
1184         /* setup master conntrack: this is a confirmed expectation */
1185         if (master_ct) {
1186                 __set_bit(IPS_EXPECTED_BIT, &ct->status);
1187                 ct->master = master_ct;
1188         }
1189
1190         add_timer(&ct->timeout);
1191         nf_conntrack_hash_insert(ct);
1192         rcu_read_unlock();
1193
1194         return 0;
1195
1196 err:
1197         nf_conntrack_free(ct);
1198         return err;
1199 }
1200
1201 static int
1202 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1203                         struct nlmsghdr *nlh, struct nlattr *cda[])
1204 {
1205         struct nf_conntrack_tuple otuple, rtuple;
1206         struct nf_conntrack_tuple_hash *h = NULL;
1207         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1208         u_int8_t u3 = nfmsg->nfgen_family;
1209         int err = 0;
1210
1211         if (cda[CTA_TUPLE_ORIG]) {
1212                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG, u3);
1213                 if (err < 0)
1214                         return err;
1215         }
1216
1217         if (cda[CTA_TUPLE_REPLY]) {
1218                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY, u3);
1219                 if (err < 0)
1220                         return err;
1221         }
1222
1223         write_lock_bh(&nf_conntrack_lock);
1224         if (cda[CTA_TUPLE_ORIG])
1225                 h = __nf_conntrack_find(&otuple, NULL);
1226         else if (cda[CTA_TUPLE_REPLY])
1227                 h = __nf_conntrack_find(&rtuple, NULL);
1228
1229         if (h == NULL) {
1230                 struct nf_conntrack_tuple master;
1231                 struct nf_conntrack_tuple_hash *master_h = NULL;
1232                 struct nf_conn *master_ct = NULL;
1233
1234                 if (cda[CTA_TUPLE_MASTER]) {
1235                         err = ctnetlink_parse_tuple(cda,
1236                                                     &master,
1237                                                     CTA_TUPLE_MASTER,
1238                                                     u3);
1239                         if (err < 0)
1240                                 goto out_unlock;
1241
1242                         master_h = __nf_conntrack_find(&master, NULL);
1243                         if (master_h == NULL) {
1244                                 err = -ENOENT;
1245                                 goto out_unlock;
1246                         }
1247                         master_ct = nf_ct_tuplehash_to_ctrack(master_h);
1248                         atomic_inc(&master_ct->ct_general.use);
1249                 }
1250
1251                 write_unlock_bh(&nf_conntrack_lock);
1252                 err = -ENOENT;
1253                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1254                         err = ctnetlink_create_conntrack(cda,
1255                                                          &otuple,
1256                                                          &rtuple,
1257                                                          master_ct);
1258                 if (err < 0 && master_ct)
1259                         nf_ct_put(master_ct);
1260
1261                 return err;
1262         }
1263         /* implicit 'else' */
1264
1265         /* We manipulate the conntrack inside the global conntrack table lock,
1266          * so there's no need to increase the refcount */
1267         err = -EEXIST;
1268         if (!(nlh->nlmsg_flags & NLM_F_EXCL)) {
1269                 /* we only allow nat config for new conntracks */
1270                 if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
1271                         err = -EINVAL;
1272                         goto out_unlock;
1273                 }
1274                 /* can't link an existing conntrack to a master */
1275                 if (cda[CTA_TUPLE_MASTER]) {
1276                         err = -EINVAL;
1277                         goto out_unlock;
1278                 }
1279                 err = ctnetlink_change_conntrack(nf_ct_tuplehash_to_ctrack(h),
1280                                                  cda);
1281         }
1282
1283 out_unlock:
1284         write_unlock_bh(&nf_conntrack_lock);
1285         return err;
1286 }
1287
1288 /***********************************************************************
1289  * EXPECT
1290  ***********************************************************************/
1291
1292 static inline int
1293 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1294                          const struct nf_conntrack_tuple *tuple,
1295                          enum ctattr_expect type)
1296 {
1297         struct nlattr *nest_parms;
1298
1299         nest_parms = nla_nest_start(skb, type | NLA_F_NESTED);
1300         if (!nest_parms)
1301                 goto nla_put_failure;
1302         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1303                 goto nla_put_failure;
1304         nla_nest_end(skb, nest_parms);
1305
1306         return 0;
1307
1308 nla_put_failure:
1309         return -1;
1310 }
1311
1312 static inline int
1313 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1314                         const struct nf_conntrack_tuple *tuple,
1315                         const struct nf_conntrack_tuple_mask *mask)
1316 {
1317         int ret;
1318         struct nf_conntrack_l3proto *l3proto;
1319         struct nf_conntrack_l4proto *l4proto;
1320         struct nf_conntrack_tuple m;
1321         struct nlattr *nest_parms;
1322
1323         memset(&m, 0xFF, sizeof(m));
1324         m.src.u.all = mask->src.u.all;
1325         memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3));
1326
1327         nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK | NLA_F_NESTED);
1328         if (!nest_parms)
1329                 goto nla_put_failure;
1330
1331         l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
1332         ret = ctnetlink_dump_tuples_ip(skb, &m, l3proto);
1333         nf_ct_l3proto_put(l3proto);
1334
1335         if (unlikely(ret < 0))
1336                 goto nla_put_failure;
1337
1338         l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
1339         ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto);
1340         nf_ct_l4proto_put(l4proto);
1341         if (unlikely(ret < 0))
1342                 goto nla_put_failure;
1343
1344         nla_nest_end(skb, nest_parms);
1345
1346         return 0;
1347
1348 nla_put_failure:
1349         return -1;
1350 }
1351
1352 static int
1353 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1354                           const struct nf_conntrack_expect *exp)
1355 {
1356         struct nf_conn *master = exp->master;
1357         long timeout = (exp->timeout.expires - jiffies) / HZ;
1358
1359         if (timeout < 0)
1360                 timeout = 0;
1361
1362         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1363                 goto nla_put_failure;
1364         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1365                 goto nla_put_failure;
1366         if (ctnetlink_exp_dump_tuple(skb,
1367                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1368                                  CTA_EXPECT_MASTER) < 0)
1369                 goto nla_put_failure;
1370
1371         NLA_PUT_BE32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout));
1372         NLA_PUT_BE32(skb, CTA_EXPECT_ID, htonl((unsigned long)exp));
1373
1374         return 0;
1375
1376 nla_put_failure:
1377         return -1;
1378 }
1379
1380 static int
1381 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1382                     int event,
1383                     int nowait,
1384                     const struct nf_conntrack_expect *exp)
1385 {
1386         struct nlmsghdr *nlh;
1387         struct nfgenmsg *nfmsg;
1388         unsigned char *b = skb_tail_pointer(skb);
1389
1390         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1391         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1392         nfmsg  = NLMSG_DATA(nlh);
1393
1394         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1395         nfmsg->nfgen_family = exp->tuple.src.l3num;
1396         nfmsg->version      = NFNETLINK_V0;
1397         nfmsg->res_id       = 0;
1398
1399         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1400                 goto nla_put_failure;
1401
1402         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
1403         return skb->len;
1404
1405 nlmsg_failure:
1406 nla_put_failure:
1407         nlmsg_trim(skb, b);
1408         return -1;
1409 }
1410
1411 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1412 static int ctnetlink_expect_event(struct notifier_block *this,
1413                                   unsigned long events, void *ptr)
1414 {
1415         struct nlmsghdr *nlh;
1416         struct nfgenmsg *nfmsg;
1417         struct nf_conntrack_expect *exp = (struct nf_conntrack_expect *)ptr;
1418         struct sk_buff *skb;
1419         unsigned int type;
1420         sk_buff_data_t b;
1421         int flags = 0;
1422
1423         if (events & IPEXP_NEW) {
1424                 type = IPCTNL_MSG_EXP_NEW;
1425                 flags = NLM_F_CREATE|NLM_F_EXCL;
1426         } else
1427                 return NOTIFY_DONE;
1428
1429         if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1430                 return NOTIFY_DONE;
1431
1432         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1433         if (!skb)
1434                 return NOTIFY_DONE;
1435
1436         b = skb->tail;
1437
1438         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1439         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1440         nfmsg = NLMSG_DATA(nlh);
1441
1442         nlh->nlmsg_flags    = flags;
1443         nfmsg->nfgen_family = exp->tuple.src.l3num;
1444         nfmsg->version      = NFNETLINK_V0;
1445         nfmsg->res_id       = 0;
1446
1447         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1448                 goto nla_put_failure;
1449
1450         nlh->nlmsg_len = skb->tail - b;
1451         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1452         return NOTIFY_DONE;
1453
1454 nlmsg_failure:
1455 nla_put_failure:
1456         kfree_skb(skb);
1457         return NOTIFY_DONE;
1458 }
1459 #endif
1460 static int ctnetlink_exp_done(struct netlink_callback *cb)
1461 {
1462         if (cb->args[1])
1463                 nf_ct_expect_put((struct nf_conntrack_expect *)cb->args[1]);
1464         return 0;
1465 }
1466
1467 static int
1468 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1469 {
1470         struct nf_conntrack_expect *exp, *last;
1471         struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
1472         struct hlist_node *n;
1473         u_int8_t l3proto = nfmsg->nfgen_family;
1474
1475         rcu_read_lock();
1476         last = (struct nf_conntrack_expect *)cb->args[1];
1477         for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) {
1478 restart:
1479                 hlist_for_each_entry(exp, n, &nf_ct_expect_hash[cb->args[0]],
1480                                      hnode) {
1481                         if (l3proto && exp->tuple.src.l3num != l3proto)
1482                                 continue;
1483                         if (cb->args[1]) {
1484                                 if (exp != last)
1485                                         continue;
1486                                 cb->args[1] = 0;
1487                         }
1488                         if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1489                                                     cb->nlh->nlmsg_seq,
1490                                                     IPCTNL_MSG_EXP_NEW,
1491                                                     1, exp) < 0) {
1492                                 if (!atomic_inc_not_zero(&exp->use))
1493                                         continue;
1494                                 cb->args[1] = (unsigned long)exp;
1495                                 goto out;
1496                         }
1497                 }
1498                 if (cb->args[1]) {
1499                         cb->args[1] = 0;
1500                         goto restart;
1501                 }
1502         }
1503 out:
1504         rcu_read_unlock();
1505         if (last)
1506                 nf_ct_expect_put(last);
1507
1508         return skb->len;
1509 }
1510
1511 static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = {
1512         [CTA_EXPECT_TIMEOUT]    = { .type = NLA_U32 },
1513         [CTA_EXPECT_ID]         = { .type = NLA_U32 },
1514 };
1515
1516 static int
1517 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1518                      struct nlmsghdr *nlh, struct nlattr *cda[])
1519 {
1520         struct nf_conntrack_tuple tuple;
1521         struct nf_conntrack_expect *exp;
1522         struct sk_buff *skb2;
1523         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1524         u_int8_t u3 = nfmsg->nfgen_family;
1525         int err = 0;
1526
1527         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1528                 return netlink_dump_start(ctnl, skb, nlh,
1529                                           ctnetlink_exp_dump_table,
1530                                           ctnetlink_exp_done);
1531         }
1532
1533         if (cda[CTA_EXPECT_MASTER])
1534                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER, u3);
1535         else
1536                 return -EINVAL;
1537
1538         if (err < 0)
1539                 return err;
1540
1541         exp = nf_ct_expect_find_get(&tuple);
1542         if (!exp)
1543                 return -ENOENT;
1544
1545         if (cda[CTA_EXPECT_ID]) {
1546                 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1547                 if (ntohl(id) != (u32)(unsigned long)exp) {
1548                         nf_ct_expect_put(exp);
1549                         return -ENOENT;
1550                 }
1551         }
1552
1553         err = -ENOMEM;
1554         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1555         if (!skb2)
1556                 goto out;
1557
1558         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1559                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1560                                       1, exp);
1561         if (err <= 0)
1562                 goto free;
1563
1564         nf_ct_expect_put(exp);
1565
1566         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1567
1568 free:
1569         kfree_skb(skb2);
1570 out:
1571         nf_ct_expect_put(exp);
1572         return err;
1573 }
1574
1575 static int
1576 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1577                      struct nlmsghdr *nlh, struct nlattr *cda[])
1578 {
1579         struct nf_conntrack_expect *exp;
1580         struct nf_conntrack_tuple tuple;
1581         struct nf_conntrack_helper *h;
1582         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1583         struct hlist_node *n, *next;
1584         u_int8_t u3 = nfmsg->nfgen_family;
1585         unsigned int i;
1586         int err;
1587
1588         if (cda[CTA_EXPECT_TUPLE]) {
1589                 /* delete a single expect by tuple */
1590                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1591                 if (err < 0)
1592                         return err;
1593
1594                 /* bump usage count to 2 */
1595                 exp = nf_ct_expect_find_get(&tuple);
1596                 if (!exp)
1597                         return -ENOENT;
1598
1599                 if (cda[CTA_EXPECT_ID]) {
1600                         __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]);
1601                         if (ntohl(id) != (u32)(unsigned long)exp) {
1602                                 nf_ct_expect_put(exp);
1603                                 return -ENOENT;
1604                         }
1605                 }
1606
1607                 /* after list removal, usage count == 1 */
1608                 nf_ct_unexpect_related(exp);
1609                 /* have to put what we 'get' above.
1610                  * after this line usage count == 0 */
1611                 nf_ct_expect_put(exp);
1612         } else if (cda[CTA_EXPECT_HELP_NAME]) {
1613                 char *name = nla_data(cda[CTA_EXPECT_HELP_NAME]);
1614                 struct nf_conn_help *m_help;
1615
1616                 /* delete all expectations for this helper */
1617                 write_lock_bh(&nf_conntrack_lock);
1618                 h = __nf_conntrack_helper_find_byname(name);
1619                 if (!h) {
1620                         write_unlock_bh(&nf_conntrack_lock);
1621                         return -EINVAL;
1622                 }
1623                 for (i = 0; i < nf_ct_expect_hsize; i++) {
1624                         hlist_for_each_entry_safe(exp, n, next,
1625                                                   &nf_ct_expect_hash[i],
1626                                                   hnode) {
1627                                 m_help = nfct_help(exp->master);
1628                                 if (m_help->helper == h
1629                                     && del_timer(&exp->timeout)) {
1630                                         nf_ct_unlink_expect(exp);
1631                                         nf_ct_expect_put(exp);
1632                                 }
1633                         }
1634                 }
1635                 write_unlock_bh(&nf_conntrack_lock);
1636         } else {
1637                 /* This basically means we have to flush everything*/
1638                 write_lock_bh(&nf_conntrack_lock);
1639                 for (i = 0; i < nf_ct_expect_hsize; i++) {
1640                         hlist_for_each_entry_safe(exp, n, next,
1641                                                   &nf_ct_expect_hash[i],
1642                                                   hnode) {
1643                                 if (del_timer(&exp->timeout)) {
1644                                         nf_ct_unlink_expect(exp);
1645                                         nf_ct_expect_put(exp);
1646                                 }
1647                         }
1648                 }
1649                 write_unlock_bh(&nf_conntrack_lock);
1650         }
1651
1652         return 0;
1653 }
1654 static int
1655 ctnetlink_change_expect(struct nf_conntrack_expect *x, struct nlattr *cda[])
1656 {
1657         return -EOPNOTSUPP;
1658 }
1659
1660 static int
1661 ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3)
1662 {
1663         struct nf_conntrack_tuple tuple, mask, master_tuple;
1664         struct nf_conntrack_tuple_hash *h = NULL;
1665         struct nf_conntrack_expect *exp;
1666         struct nf_conn *ct;
1667         struct nf_conn_help *help;
1668         int err = 0;
1669
1670         /* caller guarantees that those three CTA_EXPECT_* exist */
1671         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1672         if (err < 0)
1673                 return err;
1674         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK, u3);
1675         if (err < 0)
1676                 return err;
1677         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER, u3);
1678         if (err < 0)
1679                 return err;
1680
1681         /* Look for master conntrack of this expectation */
1682         h = nf_conntrack_find_get(&master_tuple);
1683         if (!h)
1684                 return -ENOENT;
1685         ct = nf_ct_tuplehash_to_ctrack(h);
1686         help = nfct_help(ct);
1687
1688         if (!help || !help->helper) {
1689                 /* such conntrack hasn't got any helper, abort */
1690                 err = -EINVAL;
1691                 goto out;
1692         }
1693
1694         exp = nf_ct_expect_alloc(ct);
1695         if (!exp) {
1696                 err = -ENOMEM;
1697                 goto out;
1698         }
1699
1700         exp->expectfn = NULL;
1701         exp->flags = 0;
1702         exp->master = ct;
1703         exp->helper = NULL;
1704         memcpy(&exp->tuple, &tuple, sizeof(struct nf_conntrack_tuple));
1705         memcpy(&exp->mask.src.u3, &mask.src.u3, sizeof(exp->mask.src.u3));
1706         exp->mask.src.u.all = mask.src.u.all;
1707
1708         err = nf_ct_expect_related(exp);
1709         nf_ct_expect_put(exp);
1710
1711 out:
1712         nf_ct_put(nf_ct_tuplehash_to_ctrack(h));
1713         return err;
1714 }
1715
1716 static int
1717 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1718                      struct nlmsghdr *nlh, struct nlattr *cda[])
1719 {
1720         struct nf_conntrack_tuple tuple;
1721         struct nf_conntrack_expect *exp;
1722         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
1723         u_int8_t u3 = nfmsg->nfgen_family;
1724         int err = 0;
1725
1726         if (!cda[CTA_EXPECT_TUPLE]
1727             || !cda[CTA_EXPECT_MASK]
1728             || !cda[CTA_EXPECT_MASTER])
1729                 return -EINVAL;
1730
1731         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
1732         if (err < 0)
1733                 return err;
1734
1735         write_lock_bh(&nf_conntrack_lock);
1736         exp = __nf_ct_expect_find(&tuple);
1737
1738         if (!exp) {
1739                 write_unlock_bh(&nf_conntrack_lock);
1740                 err = -ENOENT;
1741                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1742                         err = ctnetlink_create_expect(cda, u3);
1743                 return err;
1744         }
1745
1746         err = -EEXIST;
1747         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1748                 err = ctnetlink_change_expect(exp, cda);
1749         write_unlock_bh(&nf_conntrack_lock);
1750
1751         return err;
1752 }
1753
1754 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1755 static struct notifier_block ctnl_notifier = {
1756         .notifier_call  = ctnetlink_conntrack_event,
1757 };
1758
1759 static struct notifier_block ctnl_notifier_exp = {
1760         .notifier_call  = ctnetlink_expect_event,
1761 };
1762 #endif
1763
1764 static const struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1765         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1766                                             .attr_count = CTA_MAX,
1767                                             .policy = ct_nla_policy },
1768         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1769                                             .attr_count = CTA_MAX,
1770                                             .policy = ct_nla_policy },
1771         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1772                                             .attr_count = CTA_MAX,
1773                                             .policy = ct_nla_policy },
1774         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1775                                             .attr_count = CTA_MAX,
1776                                             .policy = ct_nla_policy },
1777 };
1778
1779 static const struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1780         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1781                                             .attr_count = CTA_EXPECT_MAX,
1782                                             .policy = exp_nla_policy },
1783         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1784                                             .attr_count = CTA_EXPECT_MAX,
1785                                             .policy = exp_nla_policy },
1786         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1787                                             .attr_count = CTA_EXPECT_MAX,
1788                                             .policy = exp_nla_policy },
1789 };
1790
1791 static const struct nfnetlink_subsystem ctnl_subsys = {
1792         .name                           = "conntrack",
1793         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1794         .cb_count                       = IPCTNL_MSG_MAX,
1795         .cb                             = ctnl_cb,
1796 };
1797
1798 static const struct nfnetlink_subsystem ctnl_exp_subsys = {
1799         .name                           = "conntrack_expect",
1800         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1801         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1802         .cb                             = ctnl_exp_cb,
1803 };
1804
1805 MODULE_ALIAS("ip_conntrack_netlink");
1806 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1807 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1808
1809 static int __init ctnetlink_init(void)
1810 {
1811         int ret;
1812
1813         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1814         ret = nfnetlink_subsys_register(&ctnl_subsys);
1815         if (ret < 0) {
1816                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1817                 goto err_out;
1818         }
1819
1820         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1821         if (ret < 0) {
1822                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1823                 goto err_unreg_subsys;
1824         }
1825
1826 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1827         ret = nf_conntrack_register_notifier(&ctnl_notifier);
1828         if (ret < 0) {
1829                 printk("ctnetlink_init: cannot register notifier.\n");
1830                 goto err_unreg_exp_subsys;
1831         }
1832
1833         ret = nf_ct_expect_register_notifier(&ctnl_notifier_exp);
1834         if (ret < 0) {
1835                 printk("ctnetlink_init: cannot expect register notifier.\n");
1836                 goto err_unreg_notifier;
1837         }
1838 #endif
1839
1840         return 0;
1841
1842 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1843 err_unreg_notifier:
1844         nf_conntrack_unregister_notifier(&ctnl_notifier);
1845 err_unreg_exp_subsys:
1846         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1847 #endif
1848 err_unreg_subsys:
1849         nfnetlink_subsys_unregister(&ctnl_subsys);
1850 err_out:
1851         return ret;
1852 }
1853
1854 static void __exit ctnetlink_exit(void)
1855 {
1856         printk("ctnetlink: unregistering from nfnetlink.\n");
1857
1858 #ifdef CONFIG_NF_CONNTRACK_EVENTS
1859         nf_ct_expect_unregister_notifier(&ctnl_notifier_exp);
1860         nf_conntrack_unregister_notifier(&ctnl_notifier);
1861 #endif
1862
1863         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1864         nfnetlink_subsys_unregister(&ctnl_subsys);
1865         return;
1866 }
1867
1868 module_init(ctnetlink_init);
1869 module_exit(ctnetlink_exit);